Debian Codegraph @

ChangeLog

raw
1 2026-01-14 Daniel Kiper <[email protected]>
2
3 Release 2.14
4
5 2026-01-14 Daniel Kiper <[email protected]>
6
7 build: Add grub-core/tests/crypto_cipher_mode_vectors.h file to EXTRA_DIST
8 This file was not added to EXTRA_DIST during test creation.
9
10 Fixes: 51ebc6f67 (tests: Add functional tests for ecb/cbc helpers)
11
12 Reviewed-by: Alec Brown <[email protected]>
13 Reviewed-by: Srish Srinivasan <[email protected]>
14 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
15
16 2026-01-13 Radoslav Kolev <[email protected]>
17
18 configure: Print a more helpful error if autoconf-archive is not installed
19 ... because an undefined macro receives another macro as parameter and
20 autoconf is not smart enough to produce a useful error message.
21
22 Reviewed-by: Daniel Kiper <[email protected]>
23
24 2026-01-13 Avnish Chouhan <[email protected]>
25
26 kern/ieee1275/openfw: Add a check for invalid partition number
27 The grub_strtoul() may fail in several scenarios like invalid input,
28 overflow, etc. Lack of proper check may lead to unexpected failures
29 in the code further.
30
31 Reviewed-by: Daniel Kiper <[email protected]>
32
33 2026-01-08 Sudhakar Kuppusamy <[email protected]>
34
35 grub-mkimage: Do not generate empty SBAT metadata
36 When creating core.elf with SBAT the grub-mkimage does not check if
37 an SBAT metadata file contains at least an SBAT header or not. It leads to
38 adding an empty SBAT ELF note for PowerPC and the .sbat section for EFI.
39 Fix this by checking the SBAT metadata file size against the SBAT header
40 size before adding SBAT contents to the ELF note or .sbat section.
41
42 Reviewed-by: Daniel Kiper <[email protected]>
43
44 2026-01-08 Yao Zi <[email protected]>
45
46 configure: Defer check for -mcmodel=large until PIC/PIE checks are done
47 On RISC-V, large code model is only compatible with position-depedent
48 code. However, the configure script checks availability of -mcmodel=large
49 before determining whether PIC/PIE is enabled, and disable them.
50
51 This is problematic with toolchains that enable PIE by default, where
52 check for -mcmodel=large will always fail with,
53
54 cc1: sorry, unimplemented: code model 'large' with '-fPIC'
55
56 and -mcmodel=medany will be silently used instead, causing relocation
57 failures at runtime with some memory layouts since -mcmodel=medany
58 requires all data and code to stay within a contiguous 4 GiB range.
59
60 Let's defer the check for -mcmodel=large until PIC/PIE is ensured disabled.
61
62 Fixes: f1957dc8a334 (RISC-V: Add to build system)
63
64 Reported-by: Han Gao <[email protected]>
65 Reviewed-by: Daniel Kiper <[email protected]>
66
67 2026-01-08 Sudhakar Kuppusamy <[email protected]>
68
69 util/grub-mkimagexx: Stop generating unaligned appended signatures
70 When creating the core image with an unaligned appended signature size,
71 e.g. 479, for PowerPC, the grub-mkimage aligns the appended signature
72 size to a multiple of 4 bytes, but it does not add a padding needed to
73 align to multiple of 4 bytes appended signature size in the appended
74 signature ELF note. Therefore, after signing and installing this core
75 image, the firmware tries to read the magic string "~Module signature
76 appended~" from the appended signature ELF note but gets the partial
77 magic string like "Module signature appended~". It leads to the appended
78 signature magic string match failure.
79
80 Example:
81 grub-mkimage -O powerpc-ieee1275 -o core.elf -p /grub -x \
82 kernel.der --appended-signature-size 479 ...
83
84 sign-file SHA256 ./grub.key ./grub.pem ./core.elf ./core.elf.signed
85
86 Without padding: hexdump -C ./core.elf.signed
87 ...
88 00383550 00 00 00 13 00 00 01 e0 41 53 69 67 41 70 70 65 |........ASigAppe|
89 00383560 6e 64 65 64 2d 53 69 67 6e 61 74 75 72 65 00 00 |nded-Signature..|
90 ...
91 003836f0 dd 47 cd ed 02 8e 15 af 5b 09 2e 44 6f da 67 88 |.G......[..Do.g.|
92 00383700 4d 94 17 31 26 9d 47 95 d8 7c ad 36 00 d2 9c 53 |M..1&.G..|.6...S|
93 00383710 20 e0 af 60 78 cd 22 e6 ed 45 1e b1 e7 7e cf b5 | ..`x."..E...~..|
94 00383720 fc 58 ec df 1b ab 7a 00 00 02 00 00 00 00 00 00 |.X....z.........|
95 00383730 00 01 b7 7e 4d 6f 64 75 6c 65 20 73 69 67 6e 61 |...~Module signa|
96 00383740 74 75 72 65 20 61 70 70 65 6e 64 65 64 7e 0a |ture appended~.|
97
98 Fix this by adding a padding required to align appended signature size in the
99 appended signature ELF note to multiple of 4 bytes.
100
101 Example:
102 grub-mkimage -O powerpc-ieee1275 -o core.elf -p /grub -x \
103 kernel.der --appended-signature-size 479 ...
104
105 sign-file SHA256 ./grub.key ./grub.pem ./core.elf ./core.elf.signed
106
107 With padding: hexdump -C ./core.elf.signed
108 ...
109 00137460 62 00 00 00 00 00 00 13 00 00 01 ec 41 53 69 67 |b...........ASig|
110 00137470 41 70 70 65 6e 64 65 64 2d 53 69 67 6e 61 74 75 |Appended-Signatu|
111 ...
112 00137610 b7 07 cd b6 c8 ca 9a 5b 7c 13 8c 75 1d 1c 54 81 |.......[|..u..T.|
113 00137620 7f c4 9a 8b bd d7 73 8d 2f 7d d2 e6 d1 3c 52 a9 |......s./}...<R.|
114 00137630 4e 0b e5 24 ba 0a 82 aa 8e c5 86 fa e1 19 50 ec |N..$..........P.|
115 00137640 9f a7 9a ed e5 ed 13 35 00 00 02 00 00 00 00 00 |.......5........|
116 00137650 00 00 01 c2 7e 4d 6f 64 75 6c 65 20 73 69 67 6e |....~Module sign|
117 00137660 61 74 75 72 65 20 61 70 70 65 6e 64 65 64 7e 0a |ature appended~.|
118
119 Reviewed-by: Daniel Kiper <[email protected]>
120
121 2026-01-08 Srish Srinivasan <[email protected]>
122
123 tests: Add functional tests for ecb/cbc helpers
124 Test the following helper functions using AES with 128, 192, and
125 256 bit keys:
126 - grub_crypto_ecb_encrypt(),
127 - grub_crypto_ecb_decrypt(),
128 - grub_crypto_cbc_encrypt(),
129 - grub_crypto_cbc_decrypt().
130
131 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
132 Reviewed-by: Daniel Kiper <[email protected]>
133 Reviewed-by: Glenn Washburn <[email protected]>
134
135 2026-01-08 Srish Srinivasan <[email protected]>
136
137 osdep/aros/hostdisk: Fix use-after-free bug during MsgPort deletion
138 ... in function grub_util_fd_open() when creation of an I/O request or
139 opening a device fails. The "ret", the file descriptor, will be freed
140 before its associated MsgPort is deleted resulting in a use-after-free
141 condition.
142
143 Fix this issue by freeing "ret" after its associated MsgPort has been
144 deleted.
145
146 Reviewed-by: Avnish Chouhan <[email protected]>
147 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
148 Reviewed-by: Daniel Kiper <[email protected]>
149
150 2026-01-08 Ingo Breßler <[email protected]>
151
152 kern/efi/sb: Enable loading GRUB_FILE_TYPE_CRYPTODISK_ENCRYPTION_KEY and GRUB_FILE_TYPE_CRYPTODISK_DETACHED_HEADER
153 ... file types when UEFI Secure Boot is enabled. Otherwise it is not
154 possible to load cryptodisk encryption key or detached header.
155
156 Fixes: https://savannah.gnu.org/bugs/?65889
157
158 Reviewed-by: Alec Brown <[email protected]>
159
160 2025-12-21 Radoslav Kolev <[email protected]>
161
162 blsuki: Error out if unexpected arguments are supplied
163 This can be especially helpful, as the Fedora version of the blscfg
164 actually made use of positional arguments, but current implementation
165 switched to parameters. For example what used to be "blscfg (hd0,gpt2)/..."
166 now should be "blscfg --path (hd0,gpt2)/...)". In case of old configs/scripts
167 still supplying positional arguments we will now error out instead of just
168 ignoring them and falling back to defaults silently.
169
170 Reviewed-by: Daniel Kiper <[email protected]>
171
172 2025-12-21 Radoslav Kolev <[email protected]>
173
174 blsuki: Fix default location in comment to /loader/entries
175 Reviewed-by: Daniel Kiper <[email protected]>
176
177 2025-12-21 Radoslav Kolev <[email protected]>
178
179 blsuki: Use specified device in case of fallback
180 Currently if the fallback option is enabled and no files are found in
181 the specified directory it searches the default (loader/conf) directory
182 but always in the device set by the root environment variable. It makes
183 more sense and also the comment in the code implies, that the default
184 directory on the current device should be searched.
185
186 Reviewed-by: Daniel Kiper <[email protected]>
187
188 2025-12-21 Radoslav Kolev <[email protected]>
189
190 blsuki: Fix position of DIR parameter in blscfg command summary
191 The DIR parameter in the example should be specified after the -p|--path option
192 instead of after -f|fallback.
193
194 Reviewed-by: Daniel Kiper <[email protected]>
195
196 2025-12-21 Radoslav Kolev <[email protected]>
197
198 blsuki: Fix typo in entry parameter description
199 Change "specificUKII entries" to "specific UKI entries".
200
201 Reviewed-by: Daniel Kiper <[email protected]>
202
203 2025-12-21 Khalid Ali <[email protected]>
204
205 efi: Fix several memory leaks of UEFI handles
206 Fix possible and absolute memory leaks of "handles"
207 returned by grub_efi_locate_handle() using grub_malloc().
208
209 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
210 Reviewed-by: Daniel Kiper <[email protected]>
211
212 2025-12-21 Andreas K. Hüttel <[email protected]>
213
214 util/grub-install: Allow recursive copying of theme dirs
215 grub-install allows to pass a parameter to install a theme in the boot partition.
216 This works fine for the default starfield theme. However, in general themes can
217 contain subdirectories, as, e.g. "icons", and these are not copied by grub-install.
218 As a result, the icons are missing on the screen.
219
220 Fix this by simple recursive copying.
221
222 Reviewed-by: Daniel Kiper <[email protected]>
223
224 2025-12-21 Heinrich Schuchardt <[email protected]>
225
226 commands/efi/lsefisystab: Recognize EFI_MEMORY_ATTRIBUTES_TABLE_GUID and EFI_TCG2_FINAL_EVENTS_TABLE_GUID
227 Let the lsefisystab command recognize the following table GUIDs:
228 - EFI_MEMORY_ATTRIBUTES_TABLE_GUID,
229 - EFI_TCG2_FINAL_EVENTS_TABLE_GUID.
230
231 Reviewed-by: Vladimir Serbinenko <[email protected]>
232 Reviewed-by: Daniel Kiper <[email protected]>
233
234 2025-12-21 Glenn Washburn <[email protected]>
235
236 tests/util/grub-fs-tester: Use CSMACINTOSH encoding instead of macroman
237 From Debian 12 to 13, recode had a major overhaul and now does not support
238 the macroman encoding. Its unclear if this is a bug or intentional.
239 Regardless, use the CSMACINTOSH encoding instead as MacRoman and it are
240 aliases and CSMACINTOSH is supported on both Debian 12 and 13.
241
242 Reviewed-by: Daniel Kiper <[email protected]>
243
244 2025-12-21 Luca Boccassi <[email protected]>
245
246 commands/bli: Set UINT32_MAX in LoaderTpm2ActivePcrBanks if TPM2 present but no banks protocol
247 The implementation in sd-boot was changed to return UINT32_MAX when
248 the EFI environment detects a working TPM2, but with an older firmware
249 that doesn't implement the protocol to get the list of active banks.
250 This allows distinguishing with the case where there is no working TPM2,
251 in which case userspace just gives up, and instead lets userspace try to
252 figure it out later.
253
254 Fixes: f326c5c47 (commands/bli: Set LoaderTpm2ActivePcrBanks runtime variable)
255
256 Reviewed-by: Daniel Kiper <[email protected]>
257
258 2025-12-21 Sridhar Markonda <[email protected]>
259
260 script/execute: Add a NULL check after grub_calloc() call
261 ... in gettext_append() to handle allocation errors. This prevents NULL
262 pointer dereference and stops crashes during string translation.
263
264 Reviewed-by: Daniel Kiper <[email protected]>
265
266 2025-12-21 Avnish Chouhan <[email protected]>
267
268 disk/ieee1275/ofdisk: Fix memory leaks
269 In case of an overflow "p" and "p->grub_devpath" will not be freed.
270 Fix both issues.
271
272 Reviewed-by: Alec Brown <[email protected]>
273 Reviewed-by: Daniel Kiper <[email protected]>
274 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
275
276 2025-12-21 Avnish Chouhan <[email protected]>
277
278 efiemu/loadcore: Add grub_calloc() failure check
279 Add a failure check after grub_calloc() call. If grub_calloc()
280 fails, e.g., due to memory allocation failure, it returns NULL.
281 Then using grub_efiemu_elfsyms, which will be NULL, later will
282 result in a NULL pointer dereference.
283
284 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
285 Reviewed-by: Daniel Kiper <[email protected]>
286
287 2025-12-21 George Hu <[email protected]>
288
289 lib/x86_64/setjmp: Use 32-bit zero idiom for shorter encoding
290 Switch from "xorq %rax, %rax" to "xorl %eax, %eax". In 64-bit mode
291 zeroing EAX implicitly clears RAX and the 32-bit form encodes are one
292 byte smaller while keeping identical semantics.
293
294 Reviewed-by: Daniel Kiper <[email protected]>
295
296 2025-12-21 Glenn Washburn <[email protected]>
297
298 tests: Fix nonnative tests labeled as native
299 The tests asn1_test and tpm2_key_protector_test should be labelled as
300 nonnative tests because they run tests on the target. A clue that
301 indicates a nonnative test is the usage of the grub-shell script.
302
303 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
304 Reviewed-by: Daniel Kiper <[email protected]>
305
306 2025-12-21 Glenn Washburn <[email protected]>
307
308 INSTALL: Add note that the GNU Autoconf Archive may be needed
309 As of 1a5417f39a0c (configure: Check linker for --image-base support),
310 the GNU Autoconf Archive is now required to bootstrap GRUB.
311
312 Reviewed-by: Daniel Kiper <[email protected]>
313
314 2025-11-21 Glenn Washburn <[email protected]>
315
316 INSTALL: Fix a grammatical error
317 Also, add more documentation mentioning that the tests require
318 a "specially crafted environment" to run. Just running as root
319 is not enough.
320
321 Reviewed-by: Daniel Kiper <[email protected]>
322
323 2025-11-21 Glenn Washburn <[email protected]>
324
325 bootstrap: Condense and simplify LINGUAS generation
326 Remove unnecessary subshells. Loop over autogenerated po files only once.
327 Use existing LINGUAS created by bootstrap instead of finding po files
328 again.
329
330 Add wget as a soft requirement now that we are using bootstrap's code
331 for updating translation files. This should only be needed if updated
332 translations are desired, which is the default. There should be older
333 translation files already, and wget is not necessary if those will
334 suffice.
335
336 Reviewed-by: Daniel Kiper <[email protected]>
337
338 2025-11-21 Glenn Washburn <[email protected]>
339
340 bootstrap: Migrate linguas.sh into bootstrap.conf
341 Bootstrap has infrastructure for downloading/updating project po files
342 and generating the LINGUAS file. It uses wget instead of rsync, but
343 provides the same functionality, namely that only po files that have
344 a modification date before the corresponding one on the server will get
345 redownloaded. Bootstrap creates a pristine copy of the po files in
346 po/.reference, so update .gitignore to ignore that directory.
347
348 Bootstrap also creates the po/LINGUAS file, but it does not know to add
349 in GRUB's autogenerated po files. So move that code from linguas.sh into
350 the bootstrap epilogue.
351
352 Reviewed-by: Daniel Kiper <[email protected]>
353
354 2025-11-21 Glenn Washburn <[email protected]>
355
356 bootstrap: Run linguas.sh in bootstrap epilogue
357 Heretofore, linguas.sh had to be run by the user and a common mistake
358 made when building GRUB was to not run the command. By adding it to
359 the bootstrap epilogue it will by default get run at the end of the
360 bootstrap script. The user no longer needs to remember to run it.
361 If the --skip-po option is passed to bootstrap, do not run linguas.sh.
362 This allows for bootstrap to be run without updating the translations,
363 which might be desired in the future if we track po files so that
364 translations can be used as they were at time of release.
365
366 Update INSTALL file to reflect that it is no longer necessary to run
367 linguas.sh. Also, fix a list numbering error.
368
369 Fixes: 9f73ebd49be (* INSTALL: Document linguas.sh.)
370
371 Reviewed-by: Daniel Kiper <[email protected]>
372
373 2025-11-21 Avnish Chouhan <[email protected]>
374
375 normal/cmdline: Add grub_calloc() failure check and fix hist_lines state loss
376 If grub_calloc() fails hist_lines becomes NULL. It means we loose the
377 reference to the previously allocated hist_lines and leak memory. With
378 this change on failure hist_lines still points to the old memory. So,
379 no leak, no state corruption.
380
381 Reviewed-by: Daniel Kiper <[email protected]>
382
383 2025-11-20 Michael Chang <[email protected]>
384
385 blsuki: Fix grub_errno leakage in blsuki_is_default_entry()
386 The grub_strtol() call in blsuki_is_default_entry() can set grub_errno
387 to either GRUB_ERR_BAD_NUMBER or GRUB_ERR_OUT_OF_RANGE if the input
388 string is invalid or out of range.
389
390 This grub_errno value is currently left uncleared, which can lead to
391 unexpected behavior in subsequent functions that rely on checking
392 current state of grub_errno.
393
394 Clear grub_errno unconditionally when grub_strtol() reports error so
395 that we can plug the leak.
396
397 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
398 Reviewed-by: Avnish Chouhan <[email protected]>
399 Reviewed-by: Alec Brown <[email protected]>
400 Reviewed-by: Daniel Kiper <[email protected]>
401
402 2025-11-20 Glenn Washburn <[email protected]>
403
404 Revert "tests: Remove -w param from mkfs.hfsplus command"
405 The original commit removes testing of GRUB's support for HFS+
406 wrapping and replaces it with testing that is an exact duplicate of
407 another test, namely HFS+ without wrapping. To start, the change is
408 misleading in that it suggests that the testing of HFS+ wrapping is
409 still taking place, when it is not. If it was desired to remove support
410 for testing the HFS+ wrapping, then the test should have been removed
411 entirely. Second, having a series of tests that are exactly the same is
412 just a waste of testing resources. And third, the justification for the
413 change is nonsensical. Just because a required program may not have
414 a required feature on a particular distro is not a reason that a test
415 should be removed. Reducing test coverage because some distros do not
416 have the tools GRUB needs to run certain tests goes against the testing
417 priority to have test coverage be as broad as possible. The fact is
418 that Debian, the officially supported distro for running the tests, does
419 have a mkfs.hfsplus that supports the -w parameter.
420
421 This reverts commit 2bc0929a2 (tests: Remove -w param from mkfs.hfsplus command).
422
423 Reviewed-by: Daniel Kiper <[email protected]>
424
425 2025-11-20 Glenn Washburn <[email protected]>
426
427 Revert "tests: Skip tests if required tools are not available"
428 As explained in commit a21618c8a (tests: Test aborts due to missing
429 requirements should be marked as error instead of skipped) and in the
430 Automake manual[1], skipped tests are tests that should not be run, e.g.
431 running the ohci test on the powerpc-ieee1275 as there are no native ohci
432 drivers for that platform. Test that fail for reasons other than there is
433 a bug in GRUB code that is causing the test to fail are hard errors.
434 Commonly this is because the test is run in an improperly configured
435 environment, like required programs are missing. If a hard error condition
436 is identified with a SKIP return code, the person running the tests can not
437 know without investigating every skip if a SKIP in the tests was because
438 the test does not apply to the target being tested or because the user had
439 a misconfigured environment that was causing the test not to run. By
440 ensuring that a test is skipped only when it should not run, the person
441 running the test can be sure that there is no need to investigate why the
442 test was skipped.
443
444 This reverts commit bf13fed5f (tests: Skip tests if required tools are not available).
445
446 [1] https://www.gnu.org/software/automake/manual/automake.html#Generalities-about-Testing
447
448 Reviewed-by: Daniel Kiper <[email protected]>
449
450 2025-11-20 Sudhakar Kuppusamy <[email protected]>
451
452 osdep/linux/ofpath: Add missing strdup() failure checks
453 Segmentation faults or undefined behaviour may result from a NULL pointer
454 dereference in strip_trailing_digits() and grub_util_devname_to_ofpath()
455 if strdup() fails. Therefore, I added a NULL check to fix this.
456
457 Reviewed-by: Srish Srinivasan <[email protected]>
458 Reviewed-by: Avnish Chouhan <[email protected]>
459 Reviewed-by: Daniel Kiper <[email protected]>
460
461 2025-11-20 Vladimir Serbinenko <[email protected]>
462
463 lib/relocator: Fix dereference after NULL check
464 In the function free_subchunk(), after checking that subchu->post isn't NULL,
465 grub_memset() is called on subchu->pre->freebytes but it should be called on
466 subchu->post->freebytes. If subchu->pre is NULL but subchu->post isn't NULL,
467 then this could lead to a NULL pointer dereference.
468
469 Fixes: CID 473882
470
471 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
472 Reviewed-by: Daniel Kiper <[email protected]>
473
474 2025-11-20 Nicholas Vinson <[email protected]>
475
476 configure: Check linker for --image-base support
477 In several scenarios, configure tests assume it's safe to use
478 "-Wl,-Ttext,<address>", but starting with ld.lld-21, blindly using that
479 flag may result in configure-test failures due to ld.lld failing to
480 link. The failure is because ld.lld-21 no longer allows the specified
481 address is less than the base address.
482
483 However, ld.lld-21+ and ld.bfd-2.44+ both provide support for the
484 --image-base flag making it preferable over the older -Ttext flag.
485
486 Fixes: https://savannah.gnu.org/bugs/?67662
487
488 Reviewed-by: Daniel Kiper <[email protected]>
489
490 2025-11-20 Glenn Washburn <[email protected]>
491
492 INSTALL: Make note that Linux kernel 6.12.x or earlier is needed for reiserfs testing
493 Also, remove wording suggesting that tests may be skipped if prerequisites
494 are not installed. Tests should never be skipped because of an environment
495 misconfiguration, instead they should return a hard error (code 99).
496
497 Reviewed-by: Daniel Kiper <[email protected]>
498
499 2025-11-20 Glenn Washburn <[email protected]>
500
501 docs: Reorganize test section and add section on writing tests
502 Rename the main section to Tests and put the existing test section into
503 a subsection. A new subsection called "Writing tests" is added to give
504 a brief overview and make clear the difference in returning a SKIP code
505 versus a HARD ERROR code.
506
507 Reviewed-by: Daniel Kiper <[email protected]>
508
509 2025-11-20 Glenn Washburn <[email protected]>
510
511 docs: Add note and explanation that the privileged user is required for properly running the tests
512 Reviewed-by: Daniel Kiper <[email protected]>
513
514 2025-11-20 Glenn Washburn <[email protected]>
515
516 docs: Fix spelling, grammatical and usage issues with new Porting section
517 There are some other fixes outside of this section as well.
518
519 Reviewed-by: Daniel Kiper <[email protected]>
520
521 2025-11-20 Glenn Washburn <[email protected]>
522
523 util/grub-mkrescue: Fix spelling mistakes
524 Reviewed-by: Daniel Kiper <[email protected]>
525
526 2025-11-18 Jamie <[email protected]>
527
528 commands/usbtest: Ensure string length is sufficient in usb string processing
529 If descstrp->length is less than 2 this will result in underflow in
530 "descstrp->length / 2 - 1" math. Let's fix the check to make sure the
531 value is sufficient.
532
533 Reviewed-by: Daniel Kiper <[email protected]>
534
535 2025-11-18 Jamie <[email protected]>
536
537 commands/usbtest: Use correct string length field
538 An incorrect length field is used for buffer allocation. This leads to
539 grub_utf16_to_utf8() receiving an incorrect/different length and possibly
540 causing OOB write. This makes sure to use the correct length.
541
542 Fixes: CVE-2025-61661
543
544 Reported-by: Jamie <[email protected]>
545 Reviewed-by: Daniel Kiper <[email protected]>
546
547 2025-11-18 Alec Brown <[email protected]>
548
549 tests/lib/functional_test: Unregister commands on module unload
550 When the functional_test module is loaded, both the functional_test and
551 all_functional_test commands are registered but only the all_functional_test
552 command is being unregistered since it was the last to set the cmd variable
553 that gets unregistered when the module is unloaded. To unregister both
554 commands, we need to create an additional grub_extcmd_t variable.
555
556 Reviewed-by: Daniel Kiper <[email protected]>
557
558 2025-11-18 Alec Brown <[email protected]>
559
560 normal/main: Unregister commands on module unload
561 When the normal module is loaded, the normal and normal_exit commands
562 are registered but aren't unregistered when the module is unloaded. We
563 need to add calls to grub_unregister_command() when unloading the module
564 for these commands.
565
566 Fixes: CVE-2025-61663
567 Fixes: CVE-2025-61664
568
569 Reported-by: Alec Brown <[email protected]>
570 Reviewed-by: Daniel Kiper <[email protected]>
571
572 2025-11-18 Alec Brown <[email protected]>
573
574 gettext/gettext: Unregister gettext command on module unload
575 When the gettext module is loaded, the gettext command is registered but
576 isn't unregistered when the module is unloaded. We need to add a call to
577 grub_unregister_command() when unloading the module.
578
579 Fixes: CVE-2025-61662
580
581 Reported-by: Alec Brown <[email protected]>
582 Reviewed-by: Daniel Kiper <[email protected]>
583
584 2025-11-18 Thomas Frauendorfer | Miray Software <[email protected]>
585
586 net/net: Unregister net_set_vlan command on unload
587 The commit 954c48b9c (net/net: Add net_set_vlan command) added command
588 net_set_vlan to the net module. Unfortunately the commit only added the
589 grub_register_command() call on module load but missed the
590 grub_unregister_command() on unload. Let's fix this.
591
592 Fixes: CVE-2025-54770
593 Fixes: 954c48b9c (net/net: Add net_set_vlan command)
594
595 Reported-by: Thomas Frauendorfer | Miray Software <[email protected]>
596 Reviewed-by: Daniel Kiper <[email protected]>
597
598 2025-11-18 Thomas Frauendorfer | Miray Software <[email protected]>
599
600 kern/file: Call grub_dl_unref() after fs->fs_close()
601 With commit 16f196874 (kern/file: Implement filesystem reference
602 counting) files hold a reference to their file systems.
603
604 When closing a file in grub_file_close() we should not expect
605 file->fs to stay valid after calling grub_dl_unref() on file->fs->mod.
606 So, grub_dl_unref() should be called after file->fs->fs_close().
607
608 Fixes: CVE-2025-54771
609 Fixes: 16f196874 (kern/file: Implement filesystem reference counting)
610
611 Reported-by: Thomas Frauendorfer | Miray Software <[email protected]>
612 Reviewed-by: Daniel Kiper <[email protected]>
613
614 2025-11-18 Thomas Frauendorfer | Miray Software <[email protected]>
615
616 commands/test: Fix error in recursion depth calculation
617 The commit c68b7d236 (commands/test: Stack overflow due to unlimited
618 recursion depth) added recursion depth tests to the test command. But in
619 the error case it decrements the pointer to the depth value instead of
620 the value itself. Fix it.
621
622 Fixes: c68b7d236 (commands/test: Stack overflow due to unlimited recursion depth)
623
624 Reviewed-by: Daniel Kiper <[email protected]>
625
626 2025-11-18 Alec Brown <[email protected]>
627
628 fs/zfs/zfs: Avoid pointer downcasting in dnode_get()
629 Coverity marks multiple issues in grub-core/fs/zfs/zfs.c as either "Untrusted
630 value as argument", "Untrusted pointer read", or "Untrusted loop bound". Each
631 of these issues share a common cause where Coverity finds that data->dnode_buf
632 gets tainted by dnbuf since it is downcasting from (void *) to (dnode_phys_t *)
633 and could imply that the data the pointer points to is tainted. However, the
634 function zio_read(), which reads this data from disk, sanitizes this data by
635 verifying its checksum. To resolve the issues for Coverity, setting dnbuf to
636 (dnode_phys_t *) at the start of the function dnode_get() seems to do the trick.
637
638 Fixes: CID 314020
639 Fixes: CID 896330
640 Fixes: CID 896331
641 Fixes: CID 896334
642 Fixes: CID 896336
643 Fixes: CID 896340
644 Fixes: CID 897337
645
646 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
647 Reviewed-by: Daniel Kiper <[email protected]>
648
649 2025-11-18 Alec Brown <[email protected]>
650
651 mmap/mmap: Fix resource leak
652 In the function grub_mmap_iterate(), memory is allocated to
653 "ctx.scanline_events" and "present" but isn't freed when error handling
654 grub_malloc(). Prior to returning grub_errno, these variables should be
655 freed to prevent a resource leak.
656
657 Fixes: CID 96655
658
659 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
660 Reviewed-by: Daniel Kiper <[email protected]>
661
662 2025-11-18 Glenn Washburn <[email protected]>
663
664 linguas: Ensure that linguas.sh runs from the directory that it resides
665 The script assumes that it is run from the root of the source tree,
666 which is where it is located. So this should be enforced to prevent
667 accidental misuses.
668
669 realpath is used instead of readlink as that is recommended in Debian's
670 readlink manpage since at least Debian 11. Also, use the shell variable's
671 parameter expansion for removing a suffix pattern to get the directory
672 in which the script resides. This is preferable to using the dirname binary
673 as it avoids creating a new process.
674
675 Reviewed-by: Daniel Kiper <[email protected]>
676
677 2025-11-18 Glenn Washburn <[email protected]>
678
679 gitignore: Remove po/*.po and po/LINGUAS
680 po files will now be tracked. This has the following benefits as listed
681 by Colin Watson:
682
683 * Build processes would no longer be vulnerable to an external server
684 potentially going down for an extended period of time; they'd be
685 stuck with outdated translations until the server was fixed or came
686 up with a workaround, but that's better than nothing.
687
688 * It would be easier to manage branches of stable releases, rather than
689 assuming that translations downloaded for master will match the POT
690 files for a stable release.
691
692 * Tests would be able to pass from a clean git checkout without relying
693 on an external server, improving QA reliability.
694
695 * It would be easier to make and test branches while offline.
696
697 * The translations shipped with a release tarball could be tagged in
698 git so that it's easy to investigate bugs in them.
699
700 * Downstream distributors would be able to use git branches without
701 having to fill in additional files.
702
703 Reviewed-by: Daniel Kiper <[email protected]>
704
705 2025-11-18 Glenn Washburn <[email protected]>
706
707 tests/grub_cmd_cryptomount: Use builddir shell variable and quote
708 Using a shell variable instead of an autoconf macro creates less changes
709 to the file, which can be desirable when modifying by hand later and
710 then diffing with the unexpanded file. It also makes it simpler to
711 change the builddir after expansion, which may need to happen when
712 moving the build dir to a different path and not being able to rerun
713 the autoconf expansion.
714
715 Also, add quoting around the builddir variable as there may be spaces
716 in the path.
717
718 Reviewed-by: Daniel Kiper <[email protected]>
719 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
720
721 2025-11-18 Glenn Washburn <[email protected]>
722
723 tests/grub_cmd_cryptomount: Expect test success, now that GRUB supports Argon2
724 This test was included to verify that, while GRUB did not have Argon2
725 support, trying to open a LUKS2 volume with an Argon2 keyslot would fail
726 Now that Argon2 support is included, the test is failing because it
727 expected a failure, but is now getting success. Change the test to expect
728 success.
729
730 Reviewed-by: Daniel Kiper <[email protected]>
731 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
732
733 2025-11-18 Glenn Washburn <[email protected]>
734
735 tests/erofs_test: Improve accuracy of FSTIME check
736 On resource constrained test runs, the last modification time on the
737 image is an unreliable date to check against the filesystem creation
738 time. Use dump.erofs to get the filesystem creation time from the
739 superblock. This should get the timestamp as shown by GRUB's "ls -l".
740
741 Reviewed-by: Daniel Kiper <[email protected]>
742 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
743
744 2025-11-18 Glenn Washburn <[email protected]>
745
746 tests/erofs_test: Disable filetime check for erofs_compact
747 Compact EROFS inodes do not allow for modification times that are
748 different from FS creation times. The file modification time check is
749 done between the EROFS image and the file system where test temporary
750 files are written to, not the files as seen from the mounted EROFS image.
751 So its likely that the file modification time will be different, more
752 so when run on slower systems.
753
754 Reviewed-by: Daniel Kiper <[email protected]>
755 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
756
757 2025-11-18 Glenn Washburn <[email protected]>
758
759 tests/erofs_test: Remove root check
760 The erofs tests do not ever mount the generated erofs image. So root is
761 not needed, as with the squashfs and iso9660 filesystems.
762
763 Reviewed-by: Daniel Kiper <[email protected]>
764 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
765
766 2025-11-18 Glenn Washburn <[email protected]>
767
768 tests/erofs_test: Fix mkfs.erofs version test to not use process substitution bashism
769 The shell used to run the tests is generally /bin/sh, which does not
770 support process substitution.
771
772 Fixes: b990df0bef9e (tests/util/grub-fs-tester: Fix EROFS label tests in grub-fs-tester)
773
774 Reviewed-by: Daniel Kiper <[email protected]>
775 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
776
777 2025-11-18 Andrew Hamilton <[email protected]>
778
779 tests: Support changed mkfs.ext2 behavior for -r flag
780 Correct nuisance ext234_test failure on newer Linux distros.
781
782 Recently, the mkfs.ext2 utility removed support for the -r flag to
783 specify old (version 0) formats of ext2. A new flag was added to allow
784 the same behavior. Support both ways of specifying version 0 ext2 file
785 systems when testing ext2 in GRUB.
786
787 Reviewed-by: Daniel Kiper <[email protected]>
788 Reviewed-by: Glenn Washburn <[email protected]>
789
790 2025-11-18 Andrew Hamilton <[email protected]>
791
792 tests: Avoid test failure in erofs for label length
793 Recently, mkfs.erofs began to enforce that the file system
794 label is 15 characters or less (excluding NUL terminator).
795 This causes the current erofs test in GRUB to fail. Reduce
796 the test label used to fit in this limit allowing the test
797 to work as expected.
798
799 Reviewed-by: Daniel Kiper <[email protected]>
800 Reviewed-by: Glenn Washburn <[email protected]>
801
802 2025-11-18 Andrew Hamilton <[email protected]>
803
804 tests: Split ZFS ZSTD test into new file
805 Split ZFS ZSTD test into its own test script. Add a check to the new
806 test script to see if the zfs utility installed on the host supports
807 "zstd" compression before running the test and fail the test if not. It
808 seems at least some zfs-fuse binaries do not support zstd compression
809 and the current test will fail in that case. Splitting into a new file
810 will avoid masking other test failures due to missing zstd support.
811
812 Reviewed-by: Daniel Kiper <[email protected]>
813 Reviewed-by: Glenn Washburn <[email protected]>
814
815 2025-11-18 Andrew Hamilton <[email protected]>
816
817 util/grub.d/00_header.in: Disable loading all_video for EFI
818 Loading all_video for EFI can cause video issues in some cases
819 since GRUB Bochs/Cirrus drivers may conflict with native EFI drivers.
820 Change default behavior for EFI to only load EFI specific video
821 modules. Also include a new environment variable to restore the old
822 behavior if needed.
823
824 Fixes: https://savannah.gnu.org/bugs/?66200
825
826 Reviewed-by: Daniel Kiper <[email protected]>
827
828 2025-11-18 Glenn Washburn <[email protected]>
829
830 util/grub-mkrescue: Fix copy/paste issue referencing mdadm
831 The check_xorriso() function appears to have been copy/pasted from
832 somewhere that was originally checking the mdadm command. So the file
833 handle to the output of the xorriso command is named "mdadm". Instead
834 rename it to the more generic "fout". Also change a comment referencing
835 mdadm to reference xorriso.
836
837 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
838 Reviewed-by: Daniel Kiper <[email protected]>
839
840 2025-11-18 Avnish Chouhan <[email protected]>
841
842 osdep/linux/getroot: Add missing strdup() failure checks
843 If strdup() fails, it returns NULL and passing NULL further down to
844 the code can lead to segmentation fault or an undefined behavior.
845
846 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
847 Reviewed-by: Daniel Kiper <[email protected]>
848
849 2025-11-18 Avnish Chouhan <[email protected]>
850
851 kern/mips/arc/init: Add missing grub_strdup() failure check
852 If grub_strdup() fails, it returns NULL and passing NULL further down to
853 the code can lead to segmentation fault or an undefined behavior.
854
855 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
856 Reviewed-by: Daniel Kiper <[email protected]>
857
858 2025-11-18 Avnish Chouhan <[email protected]>
859
860 script/execute: Add missing grub_strdup() failure check
861 If grub_strdup() fails, it returns NULL and passing NULL further down to
862 the code can lead to segmentation fault or an undefined behavior.
863
864 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
865 Reviewed-by: Daniel Kiper <[email protected]>
866
867 2025-11-18 Avnish Chouhan <[email protected]>
868
869 kern/ieee1275/openfw: Add missing grub_strdup() failure checks
870 If grub_strdup() fails, it returns NULL and passing NULL further down to
871 the code can lead to segmentation fault or an undefined behavior.
872
873 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
874 Reviewed-by: Daniel Kiper <[email protected]>
875
876 2025-11-18 Andrew Hamilton <[email protected]>
877
878 docs: Add security hardening suggestions
879 Add some suggestions to the security section on maximizing the
880 security hardening of GRUB.
881
882 This change reveals sectioning issues introduced by commit 0b59d379f
883 (docs/grub: Document signing GRUB under UEFI) and commit 0f2dda8cf
884 (docs/grub: Document signing GRUB with an appended signature). Fix them
885 on the occasion.
886
887 Reviewed-by: Daniel Kiper <[email protected]>
888
889 2025-11-06 Leo Sandoval <[email protected]>
890
891 tests: Remove -w param from mkfs.hfsplus command
892 Apparently the man page is outdated because the option "-w" is shown
893 but not on "mkfs.hfsplus --usage". According to Gemini:
894
895 The -w option is used to add an HFS wrapper around an HFS Plus file
896 system, which is sometimes required for compatibility with older
897 Mac OS 9 systems. However, this is not a standard or commonly used
898 option and may not be available in all versions of the hfsprogs package,
899 especially on Linux.
900
901 Reviewed-by: Daniel Kiper <[email protected]>
902
903 2025-11-06 Leo Sandoval <[email protected]>
904
905 tests: Increase verbosity in *_test.in checks
906 In this case it does not hurt to increase bash execution verbosity so
907 we can get more insight in case of issues.
908
909 Reviewed-by: Daniel Kiper <[email protected]>
910
911 2025-11-06 Leo Sandoval <[email protected]>
912
913 tests: Skip tests if required tools are not available
914 There is no reason to fail a test if the required testing tool is not
915 present on the system, so skip the test instead of failing it.
916
917 Reviewed-by: Andrew Hamilton <[email protected]>
918 Reviewed-by: Neal Gompa <[email protected]>
919 Reviewed-by: Daniel Kiper <[email protected]>
920
921 2025-11-06 Srish Srinivasan <[email protected]>
922
923 tests: Extend pbkdf2_test to cover HMAC-SHA{256,512}
924 HMAC-SHA1 is the only HMAC variant tested in the existing vectors.
925 Add vectors to test HMAC-SHA{256,512} as well.
926
927 Reviewed-by: Stefan Berger <[email protected]>
928 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
929 Reviewed-by: Daniel Kiper <[email protected]>
930
931 2025-11-06 Avnish Chouhan <[email protected]>
932
933 lib/legacy_parse: Add missing grub_malloc() failure check
934 This patch adds a NULL check after grub_malloc() call. Missing a failure
935 check after calling grub_malloc() can lead to undefined behavior. If the
936 allocation fails and returns NULL subsequent dereferencing or writing to
937 the pointer will likely result in a runtime error such as a segmentation
938 fault.
939
940 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
941 Reviewed-by: Daniel Kiper <[email protected]>
942
943 2025-11-06 Avnish Chouhan <[email protected]>
944
945 mmap/mmap: Add missing grub_malloc() failure check
946 This patch adds a NULL check after grub_malloc() call. Missing a failure
947 check after calling grub_malloc() can lead to undefined behavior. If the
948 allocation fails and returns NULL subsequent dereferencing or writing to
949 the pointer will likely result in a runtime error such as a segmentation
950 fault.
951
952 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
953 Reviewed-by: Daniel Kiper <[email protected]>
954
955 2025-11-06 Avnish Chouhan <[email protected]>
956
957 partmap/msdos: Add missing grub_malloc() failure check
958 This patch adds a NULL check after grub_malloc() call. Missing a failure
959 check after calling grub_malloc() can lead to undefined behavior. If the
960 allocation fails and returns NULL subsequent dereferencing or writing to
961 the pointer will likely result in a runtime error such as a segmentation
962 fault.
963
964 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
965 Reviewed-by: Daniel Kiper <[email protected]>
966
967 2025-11-06 Avnish Chouhan <[email protected]>
968
969 normal/completion: Add missing grub_malloc() failure check
970 This patch adds a NULL check after grub_malloc() call. Missing a failure
971 check after calling grub_malloc() can lead to undefined behavior. If the
972 allocation fails and returns NULL subsequent dereferencing or writing to
973 the pointer will likely result in a runtime error such as a segmentation
974 fault.
975
976 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
977 Reviewed-by: Daniel Kiper <[email protected]>
978
979 2025-11-06 Avnish Chouhan <[email protected]>
980
981 term/ieee1275/serial: Fix memory leak
982 The grub_zalloc() allocates memory for port. If the allocation for
983 port->name fails the function returns NULL without freeing the
984 previously allocated port memory. This results in a memory leak.
985 To avoid this we must free port before return.
986
987 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
988 Reviewed-by: Daniel Kiper <[email protected]>
989
990 2025-11-06 Lars Wendler <[email protected]>
991
992 configure: Avoid bashisms
993 or else configure check doesn't succeed with non-bash shell (e.g. dash):
994
995 checking whether to enable AMD64 as(1) feature detection... /var/tmp/portage/sys-boot/grub-2.14_rc1/work/grub-2.14~rc1/configure: 39176: test: xx86_64: unexpected operator
996 no
997
998 and later build fails with
999
1000 /var/tmp/portage/sys-boot/grub-2.14_rc1/work/grub-2.14~rc1/grub-core/lib/libgcrypt-grub/src/hwf-x86.c: In function ‘detect_x86_gnuc’:
1001 /var/tmp/portage/sys-boot/grub-2.14_rc1/work/grub-2.14~rc1/grub-core/lib/libgcrypt-grub/src/hwf-x86.c:252:17: error: ‘HWF_INTEL_CPU’ undeclared (first use in this function)
1002 252 | result |= HWF_INTEL_CPU;
1003 | ^~~~~~~~~~~~~
1004
1005 and other corresponding HWF_INTEL_* definitions because HAVE_CPU_ARCH_X86 was
1006 erroneously not defined by configure script.
1007
1008 Reviewed-by: Gary Lin <[email protected]>
1009 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1010 Reviewed-by: Avnish Chouhan <[email protected]>
1011 Reviewed-by: Daniel Kiper <[email protected]>
1012
1013 2025-10-28 Daniel Kiper <[email protected]>
1014
1015 Release 2.14~rc1
1016
1017 2025-10-28 Daniel Kiper <[email protected]>
1018
1019 windows: Fix symbol table generation during module conversion from PE to ELF
1020 According to the System V Application Binary Interface specification [1]
1021 the sections holding a symbol table, SHT_SYMTAB and SHT_DYNSYM, have to
1022 have sh_info set to "One greater than the symbol table index of the last
1023 local symbol (binding STB_LOCAL)". Current code converting PE images to
1024 ELF files does not do that and readelf complains in following way:
1025
1026 ...
1027
1028 Section Headers:
1029 [Nr] Name Type Addr Off Size ES Flg Lk Inf Al
1030 [ 0] NULL 00000000 000000 000000 00 0 0 0
1031 [ 1] .text PROGBITS 00000000 000034 0014d4 00 AX 0 0 4
1032 [ 2] .data PROGBITS 00000000 001508 000040 00 WA 0 0 32
1033 [ 3] .rdata PROGBITS 00000000 001548 0006b8 00 A 0 0 4
1034 [ 4] .module_license PROGBITS 00000000 001c00 000010 00 0 0 4
1035 [ 5] .bss NOBITS 00000000 000000 000008 00 WA 0 0 4
1036 [ 6] .moddeps PROGBITS 00000000 001c10 000010 00 0 0 4
1037 [ 7] .modname PROGBITS 00000000 001c20 000008 00 0 0 4
1038 [ 8] .rel.text REL 00000000 001c28 0008c8 08 11 1 4
1039 [ 9] .rel.data REL 00000000 0024f0 000040 08 11 2 4
1040 [10] .rel.rdata REL 00000000 002530 000070 08 11 3 4
1041 [11] .symtab SYMTAB 00000000 0025a0 0001d0 10 12 0 4
1042 [12] .strtab STRTAB 00000000 002770 000237 00 0 0 1
1043
1044 ...
1045
1046 Symbol table '.symtab' contains 29 entries:
1047 Num: Value Size Type Bind Vis Ndx Name
1048 0: 00000000 0 NOTYPE LOCAL DEFAULT UND
1049 readelf: Warning: local symbol 0 found at index >= .symtab's sh_info value of 0
1050 1: 0000144a 0 FUNC LOCAL DEFAULT 1 grub_mod_init
1051 readelf: Warning: local symbol 1 found at index >= .symtab's sh_info value of 0
1052 2: 000014aa 0 FUNC LOCAL DEFAULT 1 grub_mod_fini
1053 readelf: Warning: local symbol 2 found at index >= .symtab's sh_info value of 0
1054 3: 00000000 0 SECTION LOCAL DEFAULT 1 .text
1055 readelf: Warning: local symbol 3 found at index >= .symtab's sh_info value of 0
1056 4: 00000000 0 SECTION LOCAL DEFAULT 2 .data
1057 readelf: Warning: local symbol 4 found at index >= .symtab's sh_info value of 0
1058 5: 00000000 0 SECTION LOCAL DEFAULT 5 .bss
1059 readelf: Warning: local symbol 5 found at index >= .symtab's sh_info value of 0
1060 6: 00000000 0 SECTION LOCAL DEFAULT 3 .rdata
1061 readelf: Warning: local symbol 6 found at index >= .symtab's sh_info value of 0
1062 7: 00000000 0 NOTYPE GLOBAL DEFAULT UND grub_dma_get_phys
1063 8: 00000000 0 NOTYPE GLOBAL DEFAULT UND grub_cs5536_write_msr
1064 9: 00000000 0 NOTYPE GLOBAL DEFAULT UND grub_dma_free
1065
1066 ...
1067
1068 Let's fix it...
1069
1070 [1] https://www.sco.com/developers/gabi/2012-12-31/contents.html
1071
1072 Reviewed-by: Ross Philipson <[email protected]>
1073 Reviewed-by: Alec Brown <[email protected]>
1074
1075 2025-10-28 Daniel Kiper <[email protected]>
1076
1077 windows: Fix relocation sections generation during module conversion from PE to ELF
1078 The commit 98ad84328 (kern/dl: Check for the SHF_INFO_LINK flag in
1079 grub_dl_relocate_symbols()) revealed a bug in the code converting PE
1080 module images to ELF files. The missing SHF_INFO_LINK flag for SHT_REL
1081 and SHT_RELA sections lead to hangs during GRUB load. This only happens
1082 for the GRUB images generated on Windows platforms. The *NIX platforms
1083 are not affected due to lack of PE to ELF conversion step.
1084
1085 This patch fixes the issue...
1086
1087 Reviewed-by: Ross Philipson <[email protected]>
1088 Reviewed-by: Alec Brown <[email protected]>
1089
1090 2025-10-28 Andrew Hamilton <[email protected]>
1091
1092 loader/efi/linux: Fix compile error with Clang
1093 Clang will produce a warning, which is treated as an error, that
1094 "vendor_defined_data" is uninitialized. This is a "zero length" array
1095 member of this struct. Add conditional compile pragma to allow this to
1096 compile with Clang.
1097
1098 Reviewed-by: Daniel Kiper <[email protected]>
1099
1100 2025-10-28 Andrew Hamilton <[email protected]>
1101
1102 build: Add tpm2key.asn file for reference to dist archive
1103 Add the tpm2key.asn file to the dist archive for reference by end users.
1104
1105 Reviewed-by: Daniel Kiper <[email protected]>
1106
1107 2025-10-28 Andrew Hamilton <[email protected]>
1108
1109 build: Include new zstd test support files in dist archive
1110 Include the two new zstd test support files in the dist archive
1111 so end users can successfully run this test.
1112
1113 Reviewed-by: Daniel Kiper <[email protected]>
1114
1115 2025-10-28 Andrew Hamilton <[email protected]>
1116
1117 build: Include MAINTAINERS and SECURITY files in dist archive
1118 Include the MAINTAINERS and SECURITY files in the dist archive
1119 for reference in distributed archives by end users.
1120
1121 Reviewed-by: Daniel Kiper <[email protected]>
1122
1123 2025-10-24 Daniel Kiper <[email protected]>
1124
1125 build: Add appended signatures header file to EXTRA_DIST
1126 This file was not added to EXTRA_DIST during the appended signatures merge.
1127
1128 Fixes: 3e4ff6ffb (appended signatures: Parse ASN1 node)
1129
1130 Reviewed-by: Alec Brown <[email protected]>
1131
1132 2025-10-24 Daniel Kiper <[email protected]>
1133
1134 lib/xzembed/xz_dec_stream: Replace grub_memcpy() call with memcpy()
1135 Make the code consistent.
1136
1137 Reviewed-by: Alec Brown <[email protected]>
1138
1139 2025-10-24 Daniel Kiper <[email protected]>
1140
1141 bootstrap: Fix patching warnings
1142 Currently bootstrap complains in the following way when
1143 patching gnulib files:
1144
1145 patching file regcomp.c
1146 Hunk #2 succeeded at 1029 with fuzz 2.
1147 Hunk #5 succeeded at 1716 with fuzz 2.
1148 patching file regexec.c
1149 patching file base64.c
1150 patching file regexec.c
1151 Hunk #1 succeeded at 807 (offset -21 lines).
1152
1153 Let's fix it by adding missing "\f" and amending line
1154 numbers in the patches.
1155
1156 Reviewed-by: Alec Brown <[email protected]>
1157
1158 2025-10-24 Yann Diorcet <[email protected]>
1159
1160 tss2: Always init out buffer before calling tpm2_submit_command_real()
1161 When tpm2_submit_command_real() is called for a retry, the content of
1162 out buffer can already be set with previous tpm2_submit_command_real()
1163 call's reply. Add a call to grub_tpm2_buffer_init() before tpm2_submit_command_real().
1164
1165 This solves the issues occurring during TPM_CC_Load command on the
1166 integrated TPM 2.0 in Intel Elkhart Lake chip.
1167
1168 Reviewed-by: Daniel Kiper <[email protected]>
1169
1170 2025-10-24 Andrew Hamilton <[email protected]>
1171
1172 fs/ntfs: Correct next_attribute validation
1173 Improved ad-hoc fuzzing coverage revealed a possible access violation
1174 around line 342 of grub-core/fs/ntfs.c when accessing the attr_cur
1175 pointer due to possibility of moving pointer "next" beyond of the end of
1176 the valid buffer inside next_attribute. Prevent this for cases where
1177 full attribute validation is not performed (such as on attribute lists)
1178 by performing a sanity check on the newly calculated next pointer.
1179
1180 Fixes: 06914b614 (fs/ntfs: Correct attribute vs attribute list validation)
1181
1182 Reviewed-by: Daniel Kiper <[email protected]>
1183
1184 2025-10-24 Michael Chang <[email protected]>
1185
1186 kern/ieee1275/init: Use net config for boot location instead of firmware bootpath
1187 On network boots grub_ieee1275_net_config() is used to determine the
1188 boot device but the path continues to be taken from the Open Firmware
1189 /chosen/bootpath property. This assumes the device node follows the
1190 generic IEEE 1275 syntax which is not always the case. Different drivers
1191 may extend or redefine the format and GRUB may then misinterpret the
1192 argument as a filename and set $prefix incorrectly.
1193
1194 The generic Open Firmware device path format is:
1195
1196 device-name[:device-argument]
1197 device-argument := [partition][,[filename]]
1198
1199 For example, a bootpath such as:
1200
1201 /vdevice/l-lan@30000002:speed=auto,duplex=auto,1.2.243.345,,9.8.76.543,1.2.34.5,5,5,255.255.255.0,512
1202
1203 does not follow this form. The section after the colon (the device-argument)
1204 contains driver-specific options and network parameters, not a valid filename.
1205 The GRUB interprets this string as a filename which results in $prefix being
1206 set to "/", effectively losing the intended boot directory.
1207
1208 The firmware is not at fault here since interpretation of device nodes
1209 is driver-specific. Instead, GRUB should use the filename provided in
1210 the cached DHCP packet which is consistent and reliable. This is also
1211 the same mechanism already used on UEFI and legacy BIOS platforms.
1212
1213 This patch updates grub_machine_get_bootlocation() to prefer the result
1214 from grub_ieee1275_net_config() when complete and only fall back to the
1215 firmware bootpath otherwise.
1216
1217 Reviewed-by: Avnish Chouhan <[email protected]>
1218 Reviewed-by: Daniel Kiper <[email protected]>
1219
1220 2025-10-24 Lidong Chen <[email protected]>
1221
1222 net/tftp: Fix NULL pointer dereference in grub_net_udp_close()
1223 A NULL pointer dereference can occur in grub_net_udp_close(data->sock)
1224 when handling a malformed TFTP OACK packet.
1225
1226 This issue was discovered via fuzzing. When a malformed OACK packet
1227 contains an invalid file size, "tsize", value tftp_receive() detects
1228 the error and saves it via grub_error_save(&data->save_err). Later,
1229 tftp_open() restores this error and calls grub_net_udp_close(data->sock)
1230 assuming the socket is still valid.
1231
1232 However, the socket may have already been closed and set to NULL after
1233 processing the final data block in tftp_receive() leading to a NULL
1234 pointer dereference when attempting to close it again.
1235
1236 Fix it by checking if the socket is non-NULL before closing.
1237
1238 Reviewed-by: Daniel Kiper <[email protected]>
1239 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1240
1241 2025-10-24 Lidong Chen <[email protected]>
1242
1243 net/dns: Prevent UAF and double free
1244 In recv_hook(), *data->addresses is freed without being set to NULL.
1245 Since *data->addresses can be cached in dns_cache[h].addresses, this
1246 can lead to UAF or double free if dns_cache[h].addresses is accessed
1247 or cleared later.
1248
1249 The fix sets *data->addresses to NULL after freeing to avoid dangling
1250 pointer.
1251
1252 Reviewed-by: Daniel Kiper <[email protected]>
1253
1254 2025-10-24 Lidong Chen <[email protected]>
1255
1256 net/bootp: Prevent a UAF in network interface unregister
1257 A UAF occurs in grub_net_network_level_interface_unregister()
1258 when inter->name is accessed after being freed in grub_cmd_bootp().
1259 Fix it by deferring grub_free(ifaces[j].name) until after
1260 grub_net_network_level_interface_unregister() completes.
1261
1262 Reviewed-by: Daniel Kiper <[email protected]>
1263
1264 2025-10-24 Andrew Hamilton <[email protected]>
1265
1266 docs: Document lsmemregions and memtools commands
1267 Add documentation of the new lsmemregions command as well as
1268 documenting the existing memtools module commands.
1269
1270 Reviewed-by: Daniel Kiper <[email protected]>
1271
1272 2025-10-24 Leo Sandoval <[email protected]>
1273
1274 commands/memtools: Add lsmemregions command
1275 Prints memory regions general information including size, number of
1276 blocks, total free and total allocated memory per region. The reason
1277 behind is to have a tool that shows general information about regions
1278 and how fragmented the memory is at some particular time.
1279
1280 Below is an example showing how this tool before and after memory stress.
1281
1282 grub> lsmemregions
1283
1284 Region 0x78f6e000 (size 33554368 blocks 1048574 free 27325472 alloc 6232768)
1285
1286 > stress_big_allocations
1287 ...
1288
1289 grub> lsmemregions
1290
1291 Region 0x7af8e000 (size 4032 blocks 126 free 2720 alloc 1312)
1292 Region 0x80c000 (size 81856 blocks 2558 free 81856 alloc 0)
1293 Region 0x7d165000 (size 167872 blocks 5246 free 167872 alloc 0)
1294 Region 0x7d0bf000 (size 655296 blocks 20478 free 655296 alloc 0)
1295 Region 0x7ee00000 (size 1331136 blocks 41598 free 1331136 alloc 0)
1296 Region 0x100000 (size 7385024 blocks 230782 free 7385024 alloc 0)
1297 Region 0x7af95000 (size 25382848 blocks 793214 free 25382848 alloc 0)
1298 Region 0x1780000 (size 2038357952 blocks 63698686 free 2077517536 alloc 5445568)
1299
1300 Reviewed-by: Andrew Hamilton <[email protected]>
1301 Reviewed-by: Avnish Chouhan <[email protected]>
1302 Reviewed-by: Daniel Kiper <[email protected]>
1303
1304 2025-10-24 Logan Gunthorpe <[email protected]>
1305
1306 tests/file_filter: Add zstd tests
1307 Test zstd decompression in the same way that other decompressors are tested.
1308
1309 Reviewed-by: Daniel Kiper <[email protected]>
1310
1311 2025-10-24 Logan Gunthorpe <[email protected]>
1312
1313 tests/file_filter: Add zstd test file
1314 Add a file.zstd similar to the other compression methods and generate
1315 a gpg signature with "gpg --detach-sign".
1316
1317 Reviewed-by: Daniel Kiper <[email protected]>
1318
1319 2025-10-24 Logan Gunthorpe <[email protected]>
1320
1321 tests/file_filter: Regenerate gpg keys
1322 The "keys" file is not a valid GPG secret key so it is not possible to
1323 generate new signatures.
1324
1325 Create a new key and use "gpg --export-secret-key" to export the key
1326 and "gpg --export" to export the public key. Then resign all the
1327 signatures with "gpg --detach-sign".
1328
1329 Reviewed-by: Daniel Kiper <[email protected]>
1330
1331 2025-10-24 Logan Gunthorpe <[email protected]>
1332
1333 io/zstdio: Implement zstdio decompression
1334 Add zstd based io decompression.
1335
1336 Based largely on the existing xzio, implement the same features using
1337 the zstd library already included in the project.
1338
1339 Reviewed-by: Avnish Chouhan <[email protected]>
1340 Reviewed-by: Daniel Kiper <[email protected]>
1341
1342 2025-10-23 Michael Chang <[email protected]>
1343
1344 fs/btrfs: Update doc link for bootloader support
1345 The old wiki link is obsolete and no longer updated. Change it to the
1346 current documentation.
1347
1348 Reviewed-by: Neal Gompa <[email protected]>
1349 Reviewed-by: Daniel Kiper <[email protected]>
1350
1351 2025-10-23 Michael Chang <[email protected]>
1352
1353 docs: Add Btrfs env block and special env vars
1354 Update grub.texi to describe the external environment block in the
1355 reserved area of Btrfs header used for grub-reboot and savedefault, and
1356 add a section documenting the saved_entry, next_entry, and env_block
1357 variables.
1358
1359 Reviewed-by: Neal Gompa <[email protected]>
1360 Reviewed-by: Daniel Kiper <[email protected]>
1361
1362 2025-10-23 Michael Chang <[email protected]>
1363
1364 util/grub.d/00_header.in: Wire grub.cfg to use env_block when present
1365 This patch extends the generated grub.cfg so that it can use the
1366 external environment block when the variable env_block is defined.
1367 During boot, if env_block is set, grub.cfg builds a device path for it,
1368 exports the variable, and then loads its contents in addition to the
1369 normal grubenv file.
1370
1371 When GRUB writes variables such as next_entry or saved_entry, the save
1372 commands are changed to write into env_block if it is set, and to fall
1373 back to the grubenv file otherwise. In this way the external environment
1374 block is used automatically, and existing commands like savedefault or
1375 save_env do not need to change.
1376
1377 Reviewed-by: Neal Gompa <[email protected]>
1378 Reviewed-by: Daniel Kiper <[email protected]>
1379
1380 2025-10-23 Michael Chang <[email protected]>
1381
1382 fs/btrfs: Add environment block to reserved header area
1383 This patch reserves space for the GRUB environment block inside the
1384 Btrfs header. The block is placed at an offset of GRUB_ENV_BTRFS_OFFSET,
1385 256 KiB from the start of the device, and occupies one sector. To
1386 protect the space, overflow guard sectors are placed before and after
1387 the reserved block.
1388
1389 The Btrfs header already defines regions for bootloader use. By adding
1390 this entry, GRUB gains a fixed and safe location to store the environment
1391 block without conflicting with other structures in the header.
1392
1393 Add Btrfs and its reserved area information to the fs_envblk_spec table.
1394 With the groundworks done in previous patches, the function is now
1395 complete and working in grub-editenv.
1396
1397 Reviewed-by: Neal Gompa <[email protected]>
1398 Reviewed-by: Daniel Kiper <[email protected]>
1399
1400 2025-10-23 Michael Chang <[email protected]>
1401
1402 util/grub-editenv: Add probe call for external envblk
1403 This patch adds the probe_fs_envblk() function to identify the root
1404 filesystem and invoke fs_envblk_init() with the probed filesystem type
1405 and device. This checks if the feature is available and initializes the
1406 handle, fs_envblk, to access the external environment block. It avoids
1407 configurations with diskfilter or cryptodisk where filesystem blocks may
1408 be remapped or encrypted.
1409
1410 The probe is only invoked when grub-editenv is working on the default
1411 environment file path. This restriction ensures that probing and
1412 possible raw device access are not triggered for arbitrary user supplied
1413 paths, but only for the standard grubenv file. In that case the code
1414 checks if the filename equals DEFAULT_ENVBLK_PATH and then calls
1415 probe_fs_envblk with fs_envblk_spec. The result is stored in the global
1416 fs_envblk handle. At this stage the external environment block is only
1417 detected and recorded, and the behavior of grub-editenv is unchanged.
1418
1419 Reviewed-by: Neal Gompa <[email protected]>
1420 Reviewed-by: Avnish Chouhan <[email protected]>
1421 Reviewed-by: Daniel Kiper <[email protected]>
1422
1423 2025-10-23 Michael Chang <[email protected]>
1424
1425 util/grub-editenv: Wire list_variables() to optional fs_envblk
1426 This patch updates list_variables() so that it also prints entries from
1427 the external environment block when one is present. The function first
1428 lists all variables from the file based envblk, then iterates over the
1429 external envblk and prints those as well.
1430
1431 The output format remains the same as before. The change makes it
1432 possible to inspect variables regardless of whether they are stored in
1433 the file envblk or in the reserved block.
1434
1435 Reviewed-by: Neal Gompa <[email protected]>
1436 Reviewed-by: Avnish Chouhan <[email protected]>
1437 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1438 Reviewed-by: Daniel Kiper <[email protected]>
1439
1440 2025-10-23 Michael Chang <[email protected]>
1441
1442 util/grub-editenv: Wire unset_variables() to optional fs_envblk
1443 This patch updates unset_variables() so that removals are also applied
1444 to the external environment block when it is present. The code opens the
1445 external block, deletes the same named keys there, and then writes the
1446 external block back using fs_envblk_write(). The file based envblk is
1447 still updated and written as before.
1448
1449 Reviewed-by: Neal Gompa <[email protected]>
1450 Reviewed-by: Avnish Chouhan <[email protected]>
1451 Reviewed-by: Daniel Kiper <[email protected]>
1452
1453 2025-10-23 Michael Chang <[email protected]>
1454
1455 util/grub-editenv: Wire set_variables() to optional fs_envblk
1456 This patch changes set_variables() so that it can use an external
1457 environment block when one is present. The variable next_entry is
1458 written into the external block, env_block is treated as read only, and
1459 all other variables are written into the normal file based envblk.
1460
1461 A cleanup step is added to handle cases where GRUB at runtime writes
1462 variables into the external block because file based updates are not
1463 safe on a copy on write filesystem such as Btrfs. For example, the
1464 savedefault command can update saved_entry, and on Btrfs GRUB will place
1465 that update in the external block instead of the file envblk. If an
1466 older copy remains in the external block, it would override the newer
1467 value from the file envblk when GRUB first loads the file and then
1468 applies the external block on top of it. To avoid this, whenever
1469 a variable is updated in the file envblk, any same named key in
1470 the external block is deleted.
1471
1472 Reviewed-by: Neal Gompa <[email protected]>
1473 Reviewed-by: Avnish Chouhan <[email protected]>
1474 Reviewed-by: Daniel Kiper <[email protected]>
1475
1476 2025-10-23 Michael Chang <[email protected]>
1477
1478 util/grub-editenv: Add fs_envblk write helper
1479 This patch adds the function fs_envblk_write to update the reserved
1480 environment block on disk. The helper takes an in memory envblk buffer
1481 and writes it back to the device at the location defined by the
1482 fs_envblk specification. It performs size checks and uses file sync to
1483 ensure that the updated data is flushed.
1484
1485 The helper is also added into the fs_envblk ops table, together with the
1486 open helper from the previous patch. With this change the basic input
1487 and output path for an external environment block is complete. The
1488 choice of which variables should be written externally will be handled
1489 by later patches.
1490
1491 Reviewed-by: Neal Gompa <[email protected]>
1492 Reviewed-by: Avnish Chouhan <[email protected]>
1493 Reviewed-by: Daniel Kiper <[email protected]>
1494
1495 2025-10-23 Michael Chang <[email protected]>
1496
1497 util/grub-editenv: Add fs_envblk open helper
1498 This patch adds the logic to locate and open an environment block that
1499 is stored in a reserved area on the device. It introduces the function
1500 fs_envblk_open() together with helper routines to read the block pointed
1501 to by the env_block variable, and to create the block on disk when it
1502 does not exist yet. When a block is created, the code records its
1503 location inside the file based envblk by setting env_block in block list
1504 syntax of offset plus size in sectors.
1505
1506 The env_block variable acts as a link from the file envblk to the raw
1507 disk region so that later runs of grub-editenv can follow it and access
1508 the external block. The helper is exposed through a small ops table
1509 attached to fs_envblk so that later patches can call
1510 fs_envblk->ops->open() without touching core code again. At this stage
1511 variables are still stored in the file envblk and no redirection has
1512 been applied.
1513
1514 In relation to this, the fs_envblk_spec table defines the file-system
1515 specific layout of the reserved raw blocks used for environment storage.
1516 It is prepared to facilitate integration in grub-editenv, with Btrfs to
1517 be added in the future once its reserved area is defined.
1518
1519 An fs_envblk_init() helper is added to prepare it for using the ops with
1520 its associated data context if the feature is available. It is not used
1521 yet, but will be used later when a filesystem and its device are probed
1522 to initialize the fs_envblk handle and enable access to the feature.
1523
1524 Reviewed-by: Neal Gompa <[email protected]>
1525 Reviewed-by: Avnish Chouhan <[email protected]>
1526 Reviewed-by: Daniel Kiper <[email protected]>
1527
1528 2025-10-23 Michael Chang <[email protected]>
1529
1530 tests: Add "z" length modifier printf tests
1531 Add unit tests for %zd, %zu and %zx to verify size_t and ssize_t
1532 formatting matches system snprintf().
1533
1534 Reviewed-by: Neal Gompa <[email protected]>
1535 Reviewed-by: Daniel Kiper <[email protected]>
1536
1537 2025-10-23 Michael Chang <[email protected]>
1538
1539 kern/misc: Add the "z" length modifier support
1540 Add support for the "z" length modifier in the printf code. This allows
1541 printing of size_t and ssize_t values using %zu, %zd and related
1542 formats. The parser maps "z" to the correct integer width based on
1543 sizeof(size_t).
1544
1545 Reviewed-by: Neal Gompa <[email protected]>
1546 Reviewed-by: Daniel Kiper <[email protected]>
1547
1548 2025-10-23 Gary Lin <[email protected]>
1549
1550 disk/cryptodisk: Add --hw-accel to enable hardware acceleration
1551 The --hw-accel option has been added to cryptomount to speed up
1552 decryption by temporarily enabling hardware-specific instruction
1553 sets (e.g., AVX, SSE) in libgcrypt.
1554
1555 A new feature, "feature_gcry_hw_accel", is also introduced to mark the
1556 availability of the new option.
1557
1558 Reviewed-by: Daniel Kiper <[email protected]>
1559
1560 2025-10-23 Gary Lin <[email protected]>
1561
1562 libgcrypt: Add hardware acceleration for gcry_sha512
1563 Enable hardware acceleration for the gcry_sha512 module when building
1564 for the x86_64 EFI target.
1565
1566 Reviewed-by: Daniel Kiper <[email protected]>
1567
1568 2025-10-23 Gary Lin <[email protected]>
1569
1570 libgcrypt: Add hardware acceleration for gcry_sha256
1571 Enable hardware acceleration for the gcry_sha256 module when building
1572 for the x86_64 EFI target.
1573
1574 Reviewed-by: Daniel Kiper <[email protected]>
1575
1576 2025-10-23 Gary Lin <[email protected]>
1577
1578 libgcrypt: Declare the sha256 shaext function
1579 There is no prototype of _gcry_sha256_transform_intel_shaext() defined
1580 in the header or libgcrypt-grub/cipher/sha256.c, and gcc may complain
1581 the missing-prototypes error when compiling sha256-intel-shaext.c.
1582
1583 Declare the prototype in sha256-intel-shaext.c to avoid the error.
1584
1585 Reviewed-by: Daniel Kiper <[email protected]>
1586
1587 2025-10-23 Gary Lin <[email protected]>
1588
1589 libgcrypt: Implement _gcry_get_hw_features()
1590 Implement _gcry_get_hw_features() and enable hardware feature detection
1591 for x86_64.
1592
1593 Reviewed-by: Daniel Kiper <[email protected]>
1594
1595 2025-10-23 Gary Lin <[email protected]>
1596
1597 libgcrypt: Copy sha512 x86_64 assembly files
1598 Copy the selected x86_64 assembly files to support hardware
1599 acceleration for sha512.
1600
1601 Reviewed-by: Daniel Kiper <[email protected]>
1602
1603 2025-10-23 Gary Lin <[email protected]>
1604
1605 libgcrypt: Copy sha256 x86_64 assembly files
1606 Copy the selected x86_64 assembly files to support hardware
1607 acceleration for sha256.
1608
1609 Reviewed-by: Daniel Kiper <[email protected]>
1610
1611 2025-10-23 Gary Lin <[email protected]>
1612
1613 lib/hwfeatures-gcry: Enable SSE and AVX for x86_64 EFI
1614 Implement the necessary functions to dynamically enable SSE and AVX
1615 on x86_64 EFI systems when the hardware is capable.
1616
1617 Reviewed-by: Daniel Kiper <[email protected]>
1618
1619 2025-10-23 Gary Lin <[email protected]>
1620
1621 lib/hwfeatures-gcry: Introduce functions to manage hardware features
1622 This commit introduces the generic functions to manage the hardware
1623 features in libgcrypt. These functions are stubs for future
1624 platform-specific implementations:
1625 - grub_gcry_hwf_enabled() returns __gcry_use_hwf which indicates if
1626 the hardware features are enabled specifically by grub_enable_gcry_hwf(),
1627 - grub_enable_gcry_hwf() invokes the architecture specific enablement
1628 functions and sets __gcry_use_hwf to true,
1629 - grub_reset_gcry_hwf() invokes the architecture specific reset
1630 functions and sets __gcry_use_hwf to false.
1631
1632 Reviewed-by: Daniel Kiper <[email protected]>
1633
1634 2025-10-23 Gary Lin <[email protected]>
1635
1636 configure: Tweak autoconf/automake files to detect x86_64 features
1637 To enable hardware acceleration, this commit ports the feature detection
1638 logic from libgcrypt. This allows us to check if the compiler supports
1639 specific assembly instructions, including SSSE3, Intel SHA extensions,
1640 SSE4.1, AVX, AVX2, AVX512, and BMI2.
1641
1642 To simplify the initial implementation, support for x86_64 feature
1643 detection is currently limited to the x86_64 EFI target.
1644
1645 Reviewed-by: Daniel Kiper <[email protected]>
1646
1647 2025-10-23 Gary Lin <[email protected]>
1648
1649 lib/pbkdf2: Optimize PBKDF2 by reusing HMAC handle
1650 The previous PBKDF2 implementation used grub_crypto_hmac_buffer() which
1651 allocates and frees an HMAC handle on every call. This approach caused
1652 significant performance overhead slowing down the boot process considerably.
1653
1654 This commit refactors the PBKDF2 code to use the new HMAC functions
1655 allowing the HMAC handle and its buffers to be allocated once and reused
1656 across multiple operations. This change significantly reduces disk
1657 unlocking time.
1658
1659 In a QEMU/OVMF test environment this patch reduced the time to unlock
1660 a LUKS2 (*) partition from approximately 15 seconds to 4 seconds.
1661
1662 (*) PBKDF2 SHA256 with 3454944 iterations.
1663
1664 Reviewed-by: Daniel Kiper <[email protected]>
1665
1666 2025-10-23 Gary Lin <[email protected]>
1667
1668 lib/crypto: Introduce new HMAC functions to reuse buffers
1669 To enable more efficient buffer reuse for HMAC operations three new
1670 functions have been introduced. This change prevents the need to
1671 reallocate memory for each HMAC operation:
1672 - grub_crypto_hmac_reset(): reinitializes the hash contexts in the HMAC handle,
1673 - grub_crypto_hmac_final(): provides the final HMAC result without freeing the
1674 handle allowing it to be reused immediately,
1675 - grub_crypto_hmac_free(): deallocates the HMAC handle and its associated memory.
1676
1677 To further facilitate buffer reuse ctx2 is now included within the HMAC handle
1678 struct and the initialization of ctx2 is moved to grub_crypto_hmac_init().
1679
1680 The intermediate hash states, ctx and ctx2, for the inner and outer padded
1681 keys are now cached. The grub_crypto_hmac_reset() restores these cached
1682 states for new operations which avoids redundant hashing of the keys.
1683
1684 Reviewed-by: Daniel Kiper <[email protected]>
1685
1686 2025-10-23 Gary Lin <[email protected]>
1687
1688 docs: Document argon2 and argon2_test modules
1689 Tested-By: Waldemar Brodkorb <[email protected]>
1690 Reviewed-by: Vladimir Serbinenko <[email protected]>
1691 Reviewed-by: Daniel Kiper <[email protected]>
1692
1693 2025-10-23 Gary Lin <[email protected]>
1694
1695 kern/misc: Implement faster grub_memcpy() for aligned buffers
1696 When both "dest" and "src" are aligned, copying the data in grub_addr_t
1697 sized chunks is more efficient than a byte-by-byte copy.
1698
1699 Also tweak __aeabi_memcpy(), __aeabi_memcpy4(), and __aeabi_memcpy8(),
1700 since grub_memcpy() is not inline anymore.
1701
1702 Optimization for unaligned buffers was omitted to maintain code
1703 simplicity and readability. The current chunk-copy optimization
1704 for aligned buffers already provides a noticeable performance
1705 improvement (*) for Argon2 keyslot decryption.
1706
1707 (*) On my system, for a LUKS2 keyslot configured with a 1 GB Argon2
1708 memory requirement, this patch reduces the decryption time from
1709 22 seconds to 12 seconds.
1710
1711 Reviewed-by: Daniel Kiper <[email protected]>
1712
1713 2025-10-21 Gary Lin <[email protected]>
1714
1715 tests/util/grub-fs-tester: Use Argon2id for LUKS2 test
1716 Given that the LUKS1 test already covers PBKDF2, the default KDF for the
1717 LUKS2 test has been switched to Argon2id to ensure both algorithms are
1718 validated.
1719
1720 Tested-By: Waldemar Brodkorb <[email protected]>
1721 Reviewed-by: Vladimir Serbinenko <[email protected]>
1722 Reviewed-by: Daniel Kiper <[email protected]>
1723
1724 2025-10-21 Gary Lin <[email protected]>
1725
1726 tests: Integrate Argon2 tests into functional_test
1727 Refactor the Argon2 tests to enable the module build and integrate the
1728 tests into function_test.
1729
1730 Tested-By: Waldemar Brodkorb <[email protected]>
1731 Reviewed-by: Vladimir Serbinenko <[email protected]>
1732 Reviewed-by: Daniel Kiper <[email protected]>
1733
1734 2025-10-21 Gary Lin <[email protected]>
1735
1736 tests: Import Argon2 tests from libgcrypt
1737 Copy the Argon2 test function, check_argon2(), from t-kdf.c in libgcrypt
1738 to grub-core/tests/argon2_test.c.
1739
1740 Tested-By: Waldemar Brodkorb <[email protected]>
1741 Reviewed-by: Daniel Kiper <[email protected]>
1742
1743 2025-10-21 Gary Lin <[email protected]>
1744
1745 disk/luks2: Add Argon2 support
1746 Leverage the new grub_crypto_argon2() function to add support for the
1747 Argon2i and Argon2id KDFs in LUKS2.
1748
1749 Tested-By: Waldemar Brodkorb <[email protected]>
1750 Reviewed-by: Daniel Kiper <[email protected]>
1751
1752 2025-10-21 Gary Lin <[email protected]>
1753
1754 argon2: Introduce grub_crypto_argon2()
1755 This commit introduces grub_crypto_argon2() which leverages the
1756 _gcry_kdf_*() functions from libgcrypt to provide Argon2 support.
1757
1758 Due to the dependency of the _gcry_kdf_*() functions, the order of
1759 "ldadd" entries have to be tweaked in Makefile.util.def so that the
1760 linker can discover these functions.
1761
1762 Reviewed-by: Vladimir Serbinenko <[email protected]>
1763 Reviewed-by: Daniel Kiper <[email protected]>
1764
1765 2025-10-21 Gary Lin <[email protected]>
1766
1767 libgcrypt/kdf: Fix 64-bit modulus on 32-bit platforms
1768 Use grub_divmod64() for the 64-bit modulus to prevent creation of
1769 special division calls such as __umoddi3() and __aeabi_uldivmod() on
1770 32-bit platforms.
1771
1772 Reviewed-by: Daniel Kiper <[email protected]>
1773
1774 2025-10-21 Gary Lin <[email protected]>
1775
1776 libgcrypt/kdf: Remove unsupported KDFs
1777 Clean up _gcry_kdf_*() to remove unsupported KDFs.
1778
1779 Reviewed-by: Daniel Kiper <[email protected]>
1780
1781 2025-10-21 Gary Lin <[email protected]>
1782
1783 libgcrypt/kdf: Get rid of gpg_err_code_from_errno()
1784 gpg_err_code_from_errno() requires libgcrypt_wrap/mem.c which is not in
1785 Makefile.utilgcry.def. This commit replaces gpg_err_code_from_errno()
1786 with GPG_ERR_* to avoid the build errors.
1787
1788 Reviewed-by: Daniel Kiper <[email protected]>
1789
1790 2025-10-21 Gary Lin <[email protected]>
1791
1792 libgcrypt/kdf: Implement hash_buffers() for BLAKE2b-512
1793 The hash_buffers() functions are disabled in GRUB by default but the
1794 Argon2 implementation requires hash_buffers() for BLAKE2b-512.
1795
1796 This commit implements argon2_blake2b_512_hash_buffers() as the
1797 replacement of _gcry_digest_spec_blake2b_512.hash_buffers().
1798
1799 Reviewed-by: Daniel Kiper <[email protected]>
1800
1801 2025-10-21 Gary Lin <[email protected]>
1802
1803 crypto: Update crypto.h for libgcrypt KDF functions
1804 This commit introduces the necessary changes to crypto.h in preparation
1805 for implementing Argon2 support via the generic KDF functions, _gcry_kdf_*():
1806 - add new GPG error types required by kdf.c,
1807 - declare _gcry_digest_spec_blake2b_512 to enable BLAKE2b-512 digest calculations,
1808 - define the gcrypt KDF algorithm IDs for Argon2,
1809 - add the prototypes of _gcry_kdf_*() functions.
1810
1811 Reviewed-by: Vladimir Serbinenko <[email protected]>
1812 Reviewed-by: Daniel Kiper <[email protected]>
1813
1814 2025-10-21 Gary Lin <[email protected]>
1815
1816 util/import_gcry: Import kdf.c for Argon2
1817 The import_gcry.py script now imports kdf.c from libgcrypt. To isolate
1818 the Argon2 implementation, all unrelated functions have been removed.
1819
1820 Reviewed-by: Daniel Kiper <[email protected]>
1821
1822 2025-10-11 Sudhakar Kuppusamy <[email protected]>
1823
1824 commands/menuentry: Fix for out of bound access
1825 A menu entry with an empty title leads to an out-of-bounds access at
1826 "ch = src[len - 1]", i.e., "src" is empty and "len" is zero. So, fixing
1827 this by checking the menu entry title length and throwing an error if
1828 the length is zero.
1829
1830 Reviewed-by: Daniel Kiper <[email protected]>
1831
1832 2025-10-11 Gary Lin <[email protected]>
1833
1834 tests/tpm2_key_protector_test: Add a test for PCR Capping
1835 A test is introduced to cap PCR 1 and track the PCR 1 value before and
1836 after key unsealing.
1837
1838 Reviewed-by: Stefan Berger <[email protected]>
1839 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1840 Reviewed-by: Daniel Kiper <[email protected]>
1841
1842 2025-10-11 Gary Lin <[email protected]>
1843
1844 tpm2_key_protector: Support PCR capping
1845 To prevent a sealed key from being unsealed again, a common and
1846 straightforward method is to "cap" the key by extending the associated
1847 PCRs. When the PCRs associated with the sealed key are extended, TPM will
1848 be unable to unseal the key, as the PCR values required for unsealing no
1849 longer match, effectively rendering the key unusable until the next
1850 system boot or a state where the PCRs are reset to their expected values.
1851
1852 To cap a specific set of PCRs, simply append the argument '-c pcr_list'
1853 to the tpm2_key_protector command. Upon successfully unsealing the key,
1854 the TPM2 key protector will then invoke tpm2_protector_cap_pcrs(). This
1855 function extends the selected PCRs with an EV_SEPARATOR event,
1856 effectively "capping" them. Consequently, the associated key cannot be
1857 unsealed in any subsequent attempts until these PCRs are reset to their
1858 original, pre-capped state, typically occurring upon the next system
1859 boot.
1860
1861 Reviewed-by: Stefan Berger <[email protected]>
1862 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1863 Reviewed-by: Daniel Kiper <[email protected]>
1864
1865 2025-10-11 Gary Lin <[email protected]>
1866
1867 tss2: Implement grub_tcg2_cap_pcr() for emu
1868 Since there is no system firmware for grub-emu, the TPM2_PCR_Event
1869 command becomes the only choice to implement grub_tcg2_cap_pcr().
1870
1871 Reviewed-by: Stefan Berger <[email protected]>
1872 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1873 Reviewed-by: Daniel Kiper <[email protected]>
1874
1875 2025-10-11 Gary Lin <[email protected]>
1876
1877 tss2: Implement grub_tcg2_cap_pcr() for ieee1275
1878 This commit implements grub_tcg2_cap_pcr() for ieee1275 with the
1879 firmware function, 2hash-ext-log, to extend the target PCR with an
1880 EV_SEPARATOR event and record the event into the TPM event log.
1881
1882 To avoid duplicate code, ibmvtpm_2hash_ext_log() is moved to tcg2.c
1883 and exported as a global function.
1884
1885 Reviewed-by: Daniel Kiper <[email protected]>
1886
1887 2025-10-11 Gary Lin <[email protected]>
1888
1889 tss2: Implement grub_tcg2_cap_pcr() for EFI
1890 This commit implements grub_tcg2_cap_pcr() for EFI by using the UEFI
1891 TCG2 protocol, HashLogExtendEvent, to extend the specified PCR with an
1892 EV_SEPARATOR event and ensure the event will be recorded properly in the
1893 TPM event log.
1894
1895 Reviewed-by: Stefan Berger <[email protected]>
1896 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1897 Reviewed-by: Daniel Kiper <[email protected]>
1898
1899 2025-10-11 Gary Lin <[email protected]>
1900
1901 tss2: Introduce grub_tcg2_cap_pcr()
1902 This commit introduces the definition of grub_tcg2_cap_pcr(), a new
1903 function designed to enhance the security of sealed keys. Its primary
1904 purpose is to "cap" a specific PCR by extending it with an EV_SEPARATOR
1905 event. This action cryptographically alters the PCR value, making it
1906 impossible to unseal any key that was previously sealed to the original
1907 PCR state. Consequently, the sealed key remains protected against
1908 unauthorized unsealing attempts until the associated PCRs are reset to
1909 their initial configuration, typically occurring during a subsequent
1910 system boot.
1911
1912 Reviewed-by: Stefan Berger <[email protected]>
1913 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1914 Reviewed-by: Daniel Kiper <[email protected]>
1915
1916 2025-10-11 Gary Lin <[email protected]>
1917
1918 tss2: Add TPM2_PCR_Event command
1919 The TPM2_PCR_Event command is introduced to tss2 to allow the user to
1920 extend a specific PCR. The related data structure and unmarshal function
1921 are also introduced.
1922
1923 However, simply invoking TPM2_PCR_Event does not automatically record
1924 the event into the TPM event log. The TPM event log is primarily
1925 maintained by the system firmware (e.g., BIOS/UEFI). Therefore, for most
1926 standard use cases, the recommended method for extending PCRs and
1927 ensuring proper event logging is to utilize the system firmware
1928 functions.
1929
1930 There are specific scenarios where direct use of TPM2_PCR_Event becomes
1931 necessary. For instance, in environments lacking system firmware support
1932 for PCR extension, such as the grub-emu, TPM2_PCR_Event serves as the
1933 only available method to extend PCRs.
1934
1935 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1936 Reviewed-by: Daniel Kiper <[email protected]>
1937
1938 2025-10-11 Thomas Zimmermann <[email protected]>
1939
1940 loader/i386/linux: Transfer EDID information to kernel
1941 The Linux kernel's struct bootparams provides a field at offset 0x140
1942 for storing an EDID header. Copy the video adapter's data to the field.
1943
1944 The edid_info field was added in 2003 (see "[FBDEV] EDID support from
1945 OpenFirmware on PPC platoforms and from the BIOS on intel platforms."),
1946 but only got useable in 2004 (see "[PATCH] Fix EDID_INFO in zero-page").
1947 The boot protocol was at version 2.03 at that time.
1948
1949 The field was never used much, but with the recent addition of the efidrm
1950 and vesadrm drivers to the kernel, it becomes much more useful. As with
1951 the initial screen setup, these drivers can make use of the provided
1952 EDID information for basic display output.
1953
1954 Reviewed-by: Neal Gompa <[email protected]>
1955 Reviewed-by: Michael Chang <[email protected]>
1956 Reviewed-by: Daniel Kiper <[email protected]>
1957
1958 2025-10-11 Dave Vasilevsky <[email protected]>
1959
1960 fs/hfsplus: Allow reading files created by Mac OS 9
1961 The "permissions" field of hfsplus files is only used by Mac OS X. This
1962 causes GRUB to skip reading files created by Mac OS 9, since their
1963 file mode is read as unknown. Instead, assume files with zero mode
1964 are regular files.
1965
1966 From Technote 1150:
1967
1968 The traditional Mac OS implementation of HFS Plus does not use the
1969 permissions field. Files created by traditional Mac OS have the
1970 entire field set to 0.
1971
1972 Reviewed-by: Daniel Kiper <[email protected]>
1973
1974 2025-10-11 Sridhar Markonda <[email protected]>
1975
1976 docs: Fix build warnings in libgcrypt and blsuki doc
1977 Following warnings are thrown during libgrcypt and bluski doc build:
1978
1979 grub.texi:4744: warning: node next pointer for `gcry_arcfour_module' is `gcry_blake2_module' but next is `gcry_aria_module' in menu
1980 grub.texi:4744: warning: node prev pointer for `gcry_arcfour_module' is `gcry_aria_module' but prev is `functional_test_module' in menu
1981 grub.texi:4751: warning: node prev pointer for `gcry_blake2_module' is `gcry_arcfour_module' but prev is `gcry_aria_module' in menu
1982 grub.texi:8532: warning: node next pointer for `trust' is `unset' but next is `uki' in menu
1983 grub.texi:8549: warning: node next pointer for `unset' is `uki' but next is `verify_detached' in menu
1984 grub.texi:8549: warning: node prev pointer for `unset' is `trust' but prev is `uki' in menu
1985 grub.texi:8557: warning: node next pointer for `uki' is `verify_detached' but next is `unset' in menu
1986 grub.texi:8557: warning: node prev pointer for `uki' is `unset' but prev is `trust' in menu
1987 grub.texi:8600: warning: node prev pointer for `verify_detached' is `uki' but prev is `unset' in menu
1988
1989 Fix order of gcry_aria_module and unset nodes.
1990
1991 Reviewed-by: Andrew Hamilton <[email protected]>
1992 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
1993 Reviewed-by: Avnish Chouhan <[email protected]>
1994 Reviewed-by: Daniel Kiper <[email protected]>
1995
1996 2025-10-11 Srish Srinivasan <[email protected]>
1997
1998 kern/command,commands/extcmd: Perform explicit NULL check in both the unregister helpers
1999 During command registration, grub_register_command_prio() returns
2000 a 0 when there is a failure in memory allocation. In such a situation,
2001 calls to grub_unregister_{command(), extcmd()} during command
2002 unregistration will result in dereferencing a NULL pointer.
2003
2004 Perform explicit NULL check in both unregister helpers to prevent
2005 undefined behaviour due to a NULL pointer dereference.
2006
2007 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
2008 Reviewed-by: Stefan Berger <[email protected]>
2009 Reviewed-by: Daniel Kiper <[email protected]>
2010
2011 2025-10-11 Luca Boccassi <[email protected]>
2012
2013 commands/efi/tpm: Call get_active_pcr_banks() only with TCG2 1.1 or newer
2014 The call was added in the 1.1 revision of the spec, 1.0 does
2015 not have it, and there are some machines out there with a TPM2
2016 and a UEFI firmware that only supports version 1.0, so the
2017 call fails in those cases. Check the reported version before
2018 calling get_active_pcr_banks().
2019
2020 See Table 4 in section 6.2 of the TCG EFI Protocol Specification:
2021
2022 https://trustedcomputinggroup.org/wp-content/uploads/EFI-Protocol-Specification-rev13-160330final.pdf
2023
2024 Fixes: f326c5c47 (commands/bli: Set LoaderTpm2ActivePcrBanks runtime variable)
2025
2026 Reviewed-by: Andrew Hamilton <[email protected]>
2027 Reviewed-by: Daniel Kiper <[email protected]>
2028
2029 2025-10-11 Leo Sandoval <[email protected]>
2030
2031 kern: Include function name on debug and error print functions
2032 With the following change, we see standard (grub_dprintf) and
2033 error (grub_error) logs with the function name embedded (see below)
2034 into the log which is particular useful when debugging:
2035
2036 commands/efi/tpm.c:grub_tpm_measure:281:tpm: log_event, pcr = 8, size = 0xb,
2037
2038 Including one more field on the print log impacts the binary sizes
2039 and in turn their respective distro packages. For Fedora rpm packages
2040 the increase is 20k approximately.
2041
2042 Reviewed-by: Daniel Kiper <[email protected]>
2043
2044 2025-10-11 Peter Jones <[email protected]>
2045
2046 kern: Make grub_error() more verbose
2047 Reviewed-by: Daniel Kiper <[email protected]>
2048
2049 2025-10-11 Michael Chang <[email protected]>
2050
2051 net/tcp: Fix TCP port number reused on reboot
2052 GRUB's TCP stack assigns source ports for outgoing connections starting
2053 at 21550 and increments sequentially by 1 (e.g., 21550, 21551, ...).
2054 While this generally works, it can lead to failures if the system
2055 reboots rapidly and reuses the same source port too soon.
2056
2057 This issue was observed on powerpc-ieee1275 platforms using CAS (Client
2058 Architecture Support) reboot. In such cases, loading the initrd over
2059 HTTP may fail with connection timeouts. Packet captures show the failed
2060 connections are flagged as "TCP Port Number Reused" by Wireshark.
2061
2062 The root cause is that GRUB reuses the same port shortly after reboot,
2063 while the server may still be tracking the previous connection in
2064 TIME_WAIT. This can result in the server rejecting the connection
2065 attempt or responding with a stale ACK or RST, leading to handshake
2066 failure.
2067
2068 This patch fixes the issue by introducing a time based source port
2069 selection strategy. Instead of always starting from port 21550, GRUB now
2070 computes an initial base port based on the current RTC time, divided
2071 into 5 minute windows. The purpose of this time based strategy is to
2072 ensure that GRUB avoids reusing the same source port within a 5 minute
2073 window, thereby preventing collisions with stale server side connection
2074 tracking that could interfere with a new TCP handshake.
2075
2076 A step size of 8 ensures that the same port will not be reused across
2077 reboots unless GRUB opens more than 8 TCP connections per second on
2078 average, something that is highly unlikely. In typical usage, a GRUB
2079 boot cycle lasts about 15 seconds and may open fewer than 100
2080 connections total, well below the reuse threshold. This makes the
2081 approach robust against short reboot intervals while keeping the logic
2082 simple and deterministic.
2083
2084 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
2085 Reviewed-by: Daniel Kiper <[email protected]>
2086
2087 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2088
2089 docs/grub: Document appended signature
2090 This explains how appended signatures can be used to form part of
2091 a secure boot chain, and documents the commands and variables
2092 introduced.
2093
2094 Reviewed-by: Stefan Berger <[email protected]>
2095 Reviewed-by: Avnish Chouhan <[email protected]>
2096 Reviewed-by: Daniel Kiper <[email protected]>
2097
2098 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2099
2100 docs/grub: Document signing GRUB with an appended signature
2101 Signing GRUB for firmware that verifies an appended signature is a
2102 bit fiddly. I don't want people to have to figure it out from scratch
2103 so document it here.
2104
2105 Reviewed-by: Stefan Berger <[email protected]>
2106 Reviewed-by: Avnish Chouhan <[email protected]>
2107 Reviewed-by: Daniel Kiper <[email protected]>
2108
2109 2025-10-11 Daniel Axtens <[email protected]>
2110
2111 docs/grub: Document signing GRUB under UEFI
2112 Before adding information about how GRUB is signed with an appended
2113 signature scheme, it's worth adding some information about how it
2114 can currently be signed for UEFI.
2115
2116 Reviewed-by: Stefan Berger <[email protected]>
2117 Reviewed-by: Avnish Chouhan <[email protected]>
2118 Reviewed-by: Daniel Kiper <[email protected]>
2119
2120 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2121
2122 appended signatures: Verification tests
2123 These tests are run through all_functional_test and test a range
2124 of commands and behaviours.
2125
2126 Reviewed-by: Avnish Chouhan <[email protected]>
2127 Reviewed-by: Daniel Kiper <[email protected]>
2128
2129 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2130
2131 appended signatures: GRUB commands to manage the hashes
2132 Introducing the following GRUB commands to manage certificate/binary
2133 hashes.
2134
2135 1. append_list_dbx:
2136 Show the list of distrusted certificates and binary/certificate
2137 hashes from the dbx list.
2138 2. append_add_db_hash:
2139 Add the trusted binary hash to the db list.
2140 3. append_add_dbx_hash:
2141 Add the distrusted certificate/binary hash to the dbx list.
2142
2143 Note that if signature verification (check_appended_signatures) is set to yes,
2144 the append_add_db_hash and append_add_dbx_hash commands only accept the file
2145 ‘hash_file’ that is signed with an appended signature.
2146
2147 Tested-by: Sridhar Markonda <[email protected]>
2148 Reviewed-by: Avnish Chouhan <[email protected]>
2149 Reviewed-by: Daniel Kiper <[email protected]>
2150
2151 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2152
2153 appended signatures: GRUB commands to manage the certificates
2154 Introducing the following GRUB commands to manage the certificates.
2155
2156 1. append_list_db:
2157 Show the list of trusted certificates from the db list
2158 2. append_add_db_cert:
2159 Add the trusted certificate to the db list
2160 3. append_add_dbx_cert:
2161 Add the distrusted certificate to the dbx list
2162 4. append_verify:
2163 Verify the signed file using db list
2164
2165 Note that if signature verification (check_appended_signatures) is set to yes,
2166 the append_add_db_cert and append_add_dbx_cert commands only accept the file
2167 ‘X509_certificate’ that is signed with an appended signature.
2168
2169 Tested-by: Sridhar Markonda <[email protected]>
2170 Reviewed-by: Avnish Chouhan <[email protected]>
2171 Reviewed-by: Daniel Kiper <[email protected]>
2172
2173 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2174
2175 appended signatures: Using db and dbx lists for signature verification
2176 Signature verification: verify the kernel against lists of hashes that are
2177 either in dbx or db list. If it is not in the dbx list then the trusted keys
2178 from the db list are used to verify the signature.
2179
2180 Reviewed-by: Avnish Chouhan <[email protected]>
2181 Reviewed-by: Daniel Kiper <[email protected]>
2182
2183 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2184
2185 appended signatures: Create db and dbx lists
2186 If secure boot is enabled with static key management mode, the trusted
2187 certificates will be extracted from the GRUB ELF Note and added to db list.
2188
2189 If secure boot is enabled with dynamic key management mode, the trusted
2190 certificates and certificate/binary hash will be extracted from the PKS
2191 and added to db list. The distrusted certificates, certificate/binary hash
2192 are read from the PKS and added to dbx list. Both dbx and db lists usage is
2193 added by a subsequent patch.
2194
2195 Note:
2196 - If db does not exist in the PKS storage, then read the static keys as a db
2197 default keys from the GRUB ELF Note and add them into the db list.
2198 - If the certificate or the certificate hash exists in the dbx list, then do not
2199 add that certificate/certificate hash to the db list.
2200
2201 Reviewed-by: Avnish Chouhan <[email protected]>
2202 Reviewed-by: Daniel Kiper <[email protected]>
2203
2204 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2205
2206 appended signatures: Introducing key management environment variable
2207 Introducing the appended signature key management environment variable. It is
2208 automatically set to either "static" or "dynamic" based on the Platform KeyStore.
2209
2210 "static": Enforce static key management signature verification. This is the
2211 default. When the GRUB is locked down, user cannot change the value
2212 by setting the appendedsig_key_mgmt variable back to "dynamic".
2213
2214 "dynamic": Enforce dynamic key management signature verification. When the GRUB
2215 is locked down, user cannot change the value by setting the
2216 appendedsig_key_mgmt variable back to "static".
2217
2218 Reviewed-by: Avnish Chouhan <[email protected]>
2219 Reviewed-by: Daniel Kiper <[email protected]>
2220
2221 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2222
2223 powerpc/ieee1275: Read the db and dbx secure boot variables
2224 Enhancing the infrastructure to enable the Platform Keystore (PKS) feature,
2225 which provides access to the SB_VERSION, db, and dbx secure boot variables
2226 from PKS.
2227
2228 If PKS is enabled, it will read secure boot variables such as db and dbx
2229 from PKS and extract EFI Signature List (ESL) from it. The ESLs would be
2230 saved in the Platform Keystore buffer, and the appendedsig module would
2231 read it later to extract the certificate's details from ESL.
2232
2233 In the following scenarios, static key management mode will be activated:
2234 1. When Secure Boot is enabled with static key management mode
2235 2. When SB_VERSION is unavailable but Secure Boot is enabled
2236 3. When PKS support is unavailable but Secure Boot is enabled
2237
2238 Note:
2239
2240 SB_VERSION: Key Management Mode
2241 1 - Enable dynamic key management mode. Read the db and dbx variables from PKS,
2242 and use them for signature verification.
2243 0 - Enable static key management mode. Read keys from the GRUB ELF Note and
2244 use it for signature verification.
2245
2246 Reviewed-by: Avnish Chouhan <[email protected]>
2247 Reviewed-by: Daniel Kiper <[email protected]>
2248
2249 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2250
2251 appended signatures: Support verifying appended signatures
2252 Building on the parsers and the ability to embed X.509 certificates, as well
2253 as the existing gcrypt functionality, add a module for verifying appended
2254 signatures.
2255
2256 This includes a signature verifier that requires that the Linux kernel and
2257 GRUB modules have appended signatures for verification.
2258
2259 Signature verification must be enabled by setting check_appended_signatures.
2260 If secure boot is enabled with enforce mode when the appendedsig module is
2261 loaded, signature verification will be enabled, and trusted keys will be
2262 extracted from the GRUB ELF Note and stored in the db and locked automatically.
2263
2264 Reviewed-by: Avnish Chouhan <[email protected]>
2265 Reviewed-by: Daniel Kiper <[email protected]>
2266
2267 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2268
2269 powerpc/ieee1275: Enter lockdown based on /ibm, secure-boot
2270 Read secure boot mode from 'ibm,secure-boot' property and if the secure boot
2271 mode is set to 2 (enforce), enter lockdown. Else it is considered as disabled.
2272 There are three secure boot modes. They are
2273
2274 0 - disabled
2275 No signature verification is performed. This is the default.
2276 1 - audit
2277 Signature verification is performed and if signature verification fails,
2278 display the errors and allow the boot to continue.
2279 2 - enforce
2280 Lockdown the GRUB. Signature verification is performed and if signature
2281 verification fails, display the errors and stop the boot.
2282
2283 Now, only support disabled and enforce.
2284
2285 Reviewed-by: Stefan Berger <[email protected]>
2286 Reviewed-by: Avnish Chouhan <[email protected]>
2287 Reviewed-by: Daniel Kiper <[email protected]>
2288
2289 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2290
2291 appended signatures: Parse X.509 certificates
2292 This code allows us to parse:
2293
2294 - X.509 certificates: at least enough to verify the signatures on the PKCS#7
2295 messages. We expect that the certificates embedded in GRUB will be leaf
2296 certificates, not CA certificates. The parser enforces this.
2297
2298 - X.509 certificates support the Extended Key Usage extension and handle it by
2299 verifying that the certificate has a Code Signing usage.
2300
2301 Signed-off-by: Javier Martinez Canillas <[email protected]> # EKU support
2302 Reported-by: Michal Suchanek <[email protected]> # key usage issue
2303 Reviewed-by: Avnish Chouhan <[email protected]>
2304 Reviewed-by: Daniel Kiper <[email protected]>
2305
2306 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2307
2308 appended signatures: Parse PKCS#7 signed data
2309 This code allows us to parse:
2310
2311 - PKCS#7 signed data messages. Only a single signer info is supported, which
2312 is all that the Linux sign-file utility supports creating out-of-the-box.
2313 Only RSA, SHA-256 and SHA-512 are supported. Any certificate embedded in
2314 the PKCS#7 message will be ignored.
2315
2316 Reviewed-by: Avnish Chouhan <[email protected]>
2317 Reviewed-by: Daniel Kiper <[email protected]>
2318
2319 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2320
2321 appended signatures: Parse ASN1 node
2322 This code allows us to parse ASN1 node and allocating memory to store it.
2323 It will work for anything where the size libtasn1 returns is right:
2324 - Integers
2325 - Octet strings
2326 - DER encoding of other structures
2327
2328 It will _not_ work for things where libtasn1 size requires adjustment:
2329 - Strings that require an extra NULL byte at the end
2330 - Bit strings because libtasn1 returns the length in bits, not bytes.
2331
2332 If the function returns a non-NULL value, the caller must free it.
2333
2334 Reviewed-by: Stefan Berger <[email protected]>
2335 Reviewed-by: Avnish Chouhan <[email protected]>
2336 Reviewed-by: Daniel Kiper <[email protected]>
2337
2338 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2339
2340 appended signatures: Import GNUTLS's ASN.1 description files
2341 In order to parse PKCS#7 messages and X.509 certificates with libtasn1, we need
2342 some information about how they are encoded. We get these from GNUTLS, which has
2343 the benefit that they support the features we need and are well tested.
2344
2345 The GNUTLS files are from:
2346
2347 - https://github.com/gnutls/gnutls/blob/master/lib/gnutls.asn
2348 - https://github.com/gnutls/gnutls/blob/master/lib/pkix.asn
2349
2350 The GNUTLS license is LGPLv2.1+, which is GPLv3 compatible, allowing us to import
2351 it without issue.
2352
2353 Reviewed-by: Stefan Berger <[email protected]>
2354 Reviewed-by: Avnish Chouhan <[email protected]>
2355 Reviewed-by: Daniel Kiper <[email protected]>
2356
2357 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2358
2359 grub-install: Support embedding x509 certificates
2360 To support verification of appended signatures, we need a way to embed the
2361 necessary public keys. Existing appended signature schemes in the Linux kernel
2362 use X.509 certificates, so allow certificates to be embedded in the GRUB core
2363 image in the same way as PGP keys.
2364
2365 Reviewed-by: Stefan Berger <[email protected]>
2366 Reviewed-by: Avnish Chouhan <[email protected]>
2367 Reviewed-by: Daniel Kiper <[email protected]>
2368
2369 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2370
2371 pgp: Rename OBJ_TYPE_PUBKEY to OBJ_TYPE_GPG_PUBKEY
2372 Prior to the addition of the X.509 public key support for appended signature,
2373 current PGP signature relied on the GPG public key. Changing the enum name
2374 from "OBJ_TYPE_PUBKEY" to "OBJ_TYPE_GPG_PUBKEY" to differentiate between x509
2375 certificate based appended signature and GPG certificate based PGP signature.
2376
2377 Reviewed-by: Stefan Berger <[email protected]>
2378 Reviewed-by: Avnish Chouhan <[email protected]>
2379 Reviewed-by: Daniel Kiper <[email protected]>
2380
2381 2025-10-11 Daniel Axtens <[email protected]>
2382
2383 crypto: Move storage for grub_crypto_pk_* to crypto.c
2384 The way gcry_rsa and friends (the asymmetric ciphers) are loaded for the
2385 pgp module is a bit quirky.
2386
2387 include/grub/crypto.h contains:
2388 extern struct gcry_pk_spec *grub_crypto_pk_rsa;
2389
2390 commands/pgp.c contains the actual storage:
2391 struct gcry_pk_spec *grub_crypto_pk_rsa;
2392
2393 And the module itself saves to the storage in pgp.c:
2394 GRUB_MOD_INIT(gcry_rsa)
2395 {
2396 grub_crypto_pk_rsa = &_gcry_pubkey_spec_rsa;
2397 }
2398
2399 This is annoying: gcry_rsa now has a dependency on pgp!
2400
2401 We want to be able to bring in gcry_rsa without bringing in PGP, so move the
2402 storage to crypto.c.
2403
2404 Previously, gcry_rsa depended on pgp and mpi. Now it depends on crypto and mpi.
2405 As pgp depends on crypto, this doesn't add any new module dependencies using
2406 the PGP verfier.
2407
2408 [FWIW, the story is different for the symmetric ciphers. cryptodisk and friends
2409 (zfs encryption etc) use grub_crypto_lookup_cipher_by_name() to get a cipher
2410 handle. That depends on grub_ciphers being populated by people calling
2411 grub_cipher_register. import_gcry.py ensures that the symmetric ciphers call it.]
2412
2413 Reviewed-by: Stefan Berger <[email protected]>
2414 Reviewed-by: Vladimir Serbinenko <[email protected]>
2415 Reviewed-by: Avnish Chouhan <[email protected]>
2416 Reviewed-by: Daniel Kiper <[email protected]>
2417
2418 2025-10-11 Sudhakar Kuppusamy <[email protected]>
2419
2420 powerpc/ieee1275: Add support for signing GRUB with an appended signature
2421 Add infrastructure to allow firmware to verify the integrity of GRUB
2422 by use of a Linux-kernel-module-style appended signature. We initially
2423 target powerpc-ieee1275, but the code should be extensible to other
2424 platforms.
2425
2426 Usually these signatures are appended to a file without modifying the
2427 ELF file itself. (This is what the 'sign-file' tool does, for example.)
2428 The verifier loads the signed file from the file system and looks at the
2429 end of the file for the appended signature. However, on powerpc-ieee1275
2430 platforms, the bootloader is often stored directly in the PReP partition
2431 as raw bytes without a file-system. This makes determining the location
2432 of an appended signature more difficult.
2433
2434 To address this, we add a new ELF Note.
2435
2436 The name field of shall be the string "Appended-Signature", zero-padded
2437 to 4 byte alignment. The type field shall be 0x41536967 (the ASCII values
2438 for the string "ASig"). It must be the final section in the ELF binary.
2439
2440 The description shall contain the appended signature structure as defined
2441 by the Linux kernel. The description will also be padded to be a multiple
2442 of 4 bytes. The padding shall be added before the appended signature
2443 structure (not at the end) so that the final bytes of a signed ELF file
2444 are the appended signature magic.
2445
2446 A subsequent patch documents how to create a GRUB core.img validly signed
2447 under this scheme.
2448
2449 Reviewed-by: Stefan Berger <[email protected]>
2450 Reviewed-by: Avnish Chouhan <[email protected]>
2451 Reviewed-by: Daniel Kiper <[email protected]>
2452
2453 2025-09-10 Anaëlle Cazuc <[email protected]>
2454
2455 lib/b64dec: Use grub_size_t instead of size_t for _gpgrt_b64dec_proc() function definition
2456 On some targets, size_t and grub_size_t may not be the same type
2457 (unsigned long / unsigned int). This breaks the compilation because the
2458 definition of _gpgrt_b64dec_proc() differs from gpgrt_b64dec_proc()
2459 declaration. Fix it by using grub_size_t in the _gpgrt_b64dec_proc()
2460 definition.
2461
2462 Reviewed-by: Daniel Kiper <[email protected]>
2463
2464 2025-09-10 Anaëlle Cazuc <[email protected]>
2465
2466 util/grub-mkimagexx: Fix riscv32 relocation offset
2467 When using grub-mkrescue for a riscv32 target, an invalid implicit cast
2468 on the offset calculation produces an error during the relocation process:
2469
2470 grub-mkrescue: error: target XXX not reachable from pc=fc.
2471
2472 This patch adds an explicit grub_int64_t cast to compute the offset
2473 as a 64-bit subtraction.
2474
2475 Reviewed-by: Daniel Kiper <[email protected]>
2476
2477 2025-09-10 Andrew Hamilton <[email protected]>
2478
2479 libgcrypt: Allow GRUB to build with Clang
2480 Attempts to build GRUB with Clang were failing due to errors such as:
2481
2482 error: redefinition of typedef 'gcry_md_hd_t' is a C11 feature
2483
2484 Correct this by adding a compiler pragma to disable the Clang
2485 "typedef-redefinition" warnings. This required an update to
2486 include/grub/crypto.h and the util/import_gcry.py script to add the
2487 pragma to libgcrypt-grub's types.h due to u16 and similar types.
2488
2489 Reviewed-by: Daniel Kiper <[email protected]>
2490
2491 2025-09-10 Andrew Hamilton <[email protected]>
2492
2493 tests: Add test ISO files to dist package
2494 Add test ISO files to dist package to allow ISO test to pass.
2495
2496 Reviewed-by: Daniel Kiper <[email protected]>
2497
2498 2025-09-10 Vladimir Serbinenko <[email protected]>
2499
2500 tests: Test dates outside of 32-bit Unix range
2501 Add tests outside the date range possible with 32-bit time calculation.
2502
2503 Reviewed-by: Daniel Kiper <[email protected]>
2504
2505 2025-09-10 Vladimir Serbinenko <[email protected]>
2506
2507 lib/datetime: Support dates outside of 1901..2038 range
2508 Fixes: https://savannah.gnu.org/bugs/?63894
2509 Fixes: https://savannah.gnu.org/bugs/?66301
2510
2511 Reviewed-by: Daniel Kiper <[email protected]>
2512
2513 2025-09-04 Andrew Hamilton <[email protected]>
2514
2515 bootstrap: Ensure shallow gnulib clone works on newer git
2516 Update the bootstrap script to be compatible with newer versions of git
2517 that changed the "git clone -h" output from containing:
2518
2519 --depth
2520
2521 to:
2522
2523 --[no-]depth
2524
2525 This bootstrap script is pulled the latest gnulib version from gnulib
2526 git, commit 9a1a6385 (Silence 'time-stamp' warnings with bleeding-edge
2527 Emacs.). This change avoids a full clone on gnulib, saving something
2528 like 50 MB.
2529
2530 Fixes: https://savannah.gnu.org/bugs/?66357
2531
2532 Reviewed-by: Daniel Kiper <[email protected]>
2533
2534 2025-09-04 Andrew Hamilton <[email protected]>
2535
2536 docs: Correct some URLs
2537 Correct some outdated links to various websites and change
2538 http to https in a few places.
2539
2540 Reviewed-by: Daniel Kiper <[email protected]>
2541
2542 2025-09-04 Andrew Hamilton <[email protected]>
2543
2544 docs: Update Future section to reflect current release
2545 Update the Future section of the GRUB manual to reflect
2546 current work on the 2.x series.
2547
2548 Reviewed-by: Daniel Kiper <[email protected]>
2549
2550 2025-09-04 Andrew Hamilton <[email protected]>
2551
2552 docs: Document new libgrypt modules
2553 Add documentation for new libgcrypt modules imported into GRUB.
2554
2555 Reviewed-by: Daniel Kiper <[email protected]>
2556
2557 2025-09-04 Andrew Hamilton <[email protected]>
2558
2559 docs: Clarify section heading and fix wording
2560 Update chapter name from "Outline" to "Platform-specific operations" to
2561 improve readability. Also slightly improve some wording in this section.
2562
2563 Reviewed-by: Daniel Kiper <[email protected]>
2564
2565 2025-09-04 Andrew Hamilton <[email protected]>
2566
2567 BUGS: Update to point to bug tracking system
2568 Update the BUGS file to just point to the GRUB bug tracking system.
2569
2570 Reviewed-by: Daniel Kiper <[email protected]>
2571
2572 2025-09-04 Andrew Hamilton <[email protected]>
2573
2574 INSTALL: Document libtasn1 needed for grub-protect
2575 Update INSTALL documentation to note that the optional grub-protect
2576 utility requires libtasn1 to build.
2577
2578 Reviewed-by: Daniel Kiper <[email protected]>
2579
2580 2025-09-04 Andrew Hamilton <[email protected]>
2581
2582 po: Update translations to build with gettext 0.26
2583 Gettext 0.26 validates format strings. In some cases before
2584 the GRUB build process was converting newlines sequences (\n)
2585 to (\<translated character>) which is invalid. Update the
2586 impacted language sed script files to ensure newlines use
2587 the correct escape sequence.
2588
2589 This avoids build errors such as:
2590
2591 [email protected]:8192: 'msgstr' is not a valid Shell printf format string, unlike 'msgid'. Reason: This escape sequence is invalid.
2592
2593 Fixes: https://savannah.gnu.org/bugs/?67353
2594
2595 Reviewed-by: Vladimir Serbinenko <[email protected]>
2596 Reviewed-by: Daniel Kiper <[email protected]>
2597
2598 2025-09-04 Kancy Joe <[email protected]>
2599
2600 term/efi/console: Treat key.scan_code 0x0102 (suspend) as Enter
2601 Some Qualcomm-based UEFI platforms only provide volume up, volume down,
2602 and power keys. The volume keys are already mapped to SCAN_UP and SCAN_DOWN,
2603 while the power key is mapped to SCAN_SUSPEND (key.scan_code 0x0102).
2604
2605 On such devices, the power key is commonly used as the Enter (confirm)
2606 button, since no dedicated Enter key exists. This patch treats key.scan_code
2607 0x0102 as Enter to improve usability on these platforms.
2608
2609 Reviewed-by: Daniel Kiper <[email protected]>
2610
2611 2025-09-04 Mate Kukri <[email protected]>
2612
2613 util/bash-completion.d/Makefile.am: s/mkrescure/mkrescue/g
2614 This is a typo that was stopping this bash-completion from being installed.
2615
2616 Reviewed-by: Daniel Kiper <[email protected]>
2617
2618 2025-09-04 Alec Brown <[email protected]>
2619
2620 blsuki: Add uki command to load Unified Kernel Image entries
2621 A Unified Kernel Image (UKI) is a single UEFI PE file that combines
2622 a UEFI boot stub, a Linux kernel image, an initrd, and further resources.
2623 The uki command will locate where the UKI file is and create a GRUB menu
2624 entry to load it.
2625
2626 The Unified Kernel Image Specification: https://uapi-group.org/specifications/specs/unified_kernel_image/
2627
2628 Reviewed-by: Daniel Kiper <[email protected]>
2629
2630 2025-09-04 Robbie Harwood <[email protected]>
2631
2632 blsuki: Check for mounted /boot in emu
2633 Irritatingly, BLS defines paths relative to the mountpoint of the
2634 filesystem which contains its snippets, not / or any other fixed
2635 location. So grub-emu needs to know whether /boot is a separate
2636 filesystem from / and conditionally prepend a path.
2637
2638 Reviewed-by: Daniel Kiper <[email protected]>
2639
2640 2025-09-04 Alec Brown <[email protected]>
2641
2642 util/misc.c: Change offset type for grub_util_write_image_at()
2643 Adding filevercmp support to grub-core/commands/blsuki.c from gnulib will cause
2644 issues with the type of the offset parameter for grub_util_write_image_at() for
2645 emu builds. To fix this issue, we can change the type from off_t to grub_off_t.
2646
2647 Reviewed-by: Daniel Kiper <[email protected]>
2648
2649 2025-09-04 Peter Jones <[email protected]>
2650
2651 blsuki: Add blscfg command to parse Boot Loader Specification snippets
2652 The BootLoaderSpec (BLS) defines a scheme where different bootloaders can
2653 share a format for boot items and a configuration directory that accepts
2654 these common configurations as drop-in files.
2655
2656 The BLS Specification: https://uapi-group.org/specifications/specs/boot_loader_specification/
2657
2658 Reviewed-by: Daniel Kiper <[email protected]>
2659
2660 2025-09-04 Alec Brown <[email protected]>
2661
2662 kern/misc: Implement grub_strtok()
2663 Add the functions grub_strtok() and grub_strtok_r() to help parse strings into
2664 tokens separated by characters in the "delim" parameter. These functions are
2665 present in gnulib but calling them directly from the gnulib code is quite
2666 challenging since the call "#include <string.h>" would include the header file
2667 grub-core/lib/posix_wrap/string.h instead of grub-core/lib/gnulib/string.h,
2668 where strtok() and strtok_r() are declared. Since this overlap is quite
2669 problematic, the simpler solution was to implement the code in the GRUB based
2670 on gnulib's implementation. For more information on these functions, visit the
2671 Linux Programmer's Manual, man strtok.
2672
2673 Reviewed-by: Daniel Kiper <[email protected]>
2674
2675 2025-09-04 Aaron Rainbolt <[email protected]>
2676
2677 kern/xen: Add Xen command line parsing
2678 Xen traditionally allows customizing guest behavior by passing arguments
2679 to the VM kernel via the kernel command line. This is no longer possible
2680 when using GRUB with Xen, as the kernel command line is decided by the
2681 GRUB configuration file within the guest, not data passed to the guest
2682 by Xen.
2683
2684 To work around this limitation, enable GRUB to parse a command line
2685 passed to it by Xen, and expose data from the command line to the GRUB
2686 configuration as environment variables. These variables can be used in
2687 the GRUB configuration for any desired purpose, such as extending the
2688 kernel command line passed to the guest. The command line format is
2689 inspired by the Linux kernel's command line format.
2690
2691 To reduce the risk of misuse, abuse, or accidents in production, the
2692 command line will only be parsed if it consists entirely of 7-bit ASCII
2693 characters, only alphabetical characters and underscores are permitted
2694 in variable names, and all variable names must start with the string
2695 "xen_grub_env_". This also allows room for expanding the command line
2696 arguments accepted by GRUB in the future, should other arguments end up
2697 becoming desirable in the future.
2698
2699 Reviewed-by: Daniel Kiper <[email protected]>
2700
2701 2025-09-03 Aaron Rainbolt <[email protected]>
2702
2703 include/xen/xen.h: Add warning comment for cmd_line
2704 The cmd_line field of the start_info struct is not guaranteed to be
2705 NUL-terminated, even though it is intended to contain a NUL-terminated
2706 string. Add a warning about this in a comment so future consumers of
2707 this field know to check it for a NUL terminator before using it.
2708
2709 Reviewed-by: Daniel Kiper <[email protected]>
2710
2711 2025-08-15 Doug Goldstein <[email protected]>
2712
2713 zfs: Fix LINUX_ROOT_DEVICE when grub-probe fails
2714 When grub-probe fails, the current code is to just stuff an empty result
2715 in which causes the user to not knowingly have a system that no longer
2716 boots. grub-probe can fail because the ZFS pool that contains the root
2717 filesystem might have features that GRUB does not yet support which is
2718 a common configuration for people with a rpool and a bpool. This behavior
2719 uses the zdb utility to dump the same value as the filesystem label
2720 would print.
2721
2722 Reviewed-by: Daniel Kiper <[email protected]>
2723
2724 2025-08-15 Vladimir Serbinenko <[email protected]>
2725
2726 relocator: Switch to own page table while moving chunks
2727 We need to avoid clobbering existing table between starting of chunk movers
2728 and the moment we install target page table. Generate temporary table for
2729 this rather than hoping that we don't clobber existing one.
2730
2731 Fixes 64-bit GhostBSD on 64-bit EFI.
2732
2733 Reviewed-by: Daniel Kiper <[email protected]>
2734
2735 2025-08-14 Mate Kukri <[email protected]>
2736
2737 configure: Generate tar-ustar tarball instead of tar-v7
2738 Some of our paths are too long for tar-v7 at this point but tar-ustar
2739 is supported by essentially everything. So, let's use that.
2740
2741 Reviewed-by: Daniel Kiper <[email protected]>
2742
2743 2025-08-14 Mate Kukri <[email protected]>
2744
2745 build: Add new libgcrypt and libtasn1 related files to EXTRA_DISTS
2746 These files were not added to EXTRA_DISTS during the libgcrypt
2747 and libtasn1 imports but are required for autogen.sh to work.
2748
2749 Reviewed-by: Daniel Kiper <[email protected]>
2750
2751 2025-08-14 Mate Kukri <[email protected]>
2752
2753 build: Add util/import_gcrypt_inth.sed to EXTRA_DISTS
2754 This new file was not added to the distribution tarball during the last
2755 libgcrypt import.
2756
2757 Fixes: 0739d24cd164 (libgcrypt: Adjust import script, definitions and API users for libgcrypt 1.11)
2758
2759 Reviewed-by: Daniel Kiper <[email protected]>
2760
2761 2025-08-14 Aaron Rainbolt <[email protected]>
2762
2763 include/xen/xen.h: Rename MAX_GUEST_CMDLINE to GRUB_XEN_MAX_GUEST_CMDLINE
2764 The include/xen/xen.h header was using an overly generic name to refer
2765 to the maximum length of the command line passed from Xen to a guest.
2766 Rename it to avoid confusion or conflicts in the future.
2767
2768 Reviewed-by: Daniel Kiper <[email protected]>
2769
2770 2025-08-14 Frediano Ziglio <[email protected]>
2771
2772 loader/arm64/xen_boot: Set correctly bootargs property for modules
2773 The cmdline_size already account for NUL terminator, you can see
2774 this in xen_boot_binary_load(). The same property is set correctly
2775 for Xen command line.
2776
2777 Reviewed-by: Daniel Kiper <[email protected]>
2778
2779 2025-08-14 Frediano Ziglio <[email protected]>
2780
2781 loader/efi/linux: Return correct size from LoadFile2
2782 From UEFI specifications 2.10, section 13.2.2, EFI_LOAD_FILE2_PROTOCOL.LoadFile
2783 (see https://uefi.org/specs/UEFI/2.10/13_Protocols_Media_Access.html), for BufferSize:
2784
2785 On input the size of Buffer in bytes. On output with a return code
2786 of EFI_SUCCESS, the amount of data transferred to Buffer. On output
2787 with a return code of EFI_BUFFER_TOO_SMALL, the size of Buffer
2788 required to retrieve the requested file.
2789
2790 So, set *buffer_size correctly.
2791
2792 Reviewed-by: Daniel Kiper <[email protected]>
2793
2794 2025-08-14 Luca Boccassi <[email protected]>
2795
2796 commands/bli: Set LoaderTpm2ActivePcrBanks runtime variable
2797 It turns out checking from userspace is not 100% reliable to figure out
2798 whether the firmware had TPM2 support enabled or not. For example with
2799 EDK2 arm64, the default upstream build config bundles TPM2 support with
2800 SecureBoot support, so if the latter is disabled, TPM2 is also unavailable.
2801 But still, the ACPI TPM2 table is created just as if it was enabled. So,
2802 /sys/firmware/acpi/tables/TPM2 exists and looks correct but there are no
2803 measurements, neither the firmware nor the loader/stub can do them, and
2804 /sys/kernel/security/tpm0/binary_bios_measurements does not exist.
2805 So, userspace cannot really tell what was going on in UEFI mode.
2806
2807 The loader can use the apposite UEFI protocol to check, which is a more
2808 definitive answer. Export the bitmask with the list of active banks as-is.
2809 If it's not 0, then in userspace we can be sure a working TPM2 was available
2810 in UEFI mode.
2811
2812 systemd-boot and systemd-stub v258 (current main) set this variable and
2813 userspace portion consumes it to be able to tell what was available in
2814 the firmware context.
2815
2816 Reviewed-by: Daniel Kiper <[email protected]>
2817
2818 2025-07-11 Vladimir Serbinenko <[email protected]>
2819
2820 docs: Write how to import new libgcrypt
2821 Reviewed-by: Daniel Kiper <[email protected]>
2822
2823 2025-07-11 Vladimir Serbinenko <[email protected]>
2824
2825 libgcrypt: Fix a memory leak
2826 Fixes: CID 468917
2827
2828 Reviewed-by: Daniel Kiper <[email protected]>
2829
2830 2025-07-11 Vladimir Serbinenko <[email protected]>
2831
2832 libgcrypt: Don't use 64-bit division on platforms where it's slow
2833 Reviewed-by: Daniel Kiper <[email protected]>
2834
2835 util/import_gcry: Fix pylint warnings
2836 Reviewed-by: Daniel Kiper <[email protected]>
2837
2838 util/import_gcry: Make compatible with Python 3.4
2839 Reviewed-by: Daniel Kiper <[email protected]>
2840
2841 libgcrypt: Import blake family of hashes
2842 Reviewed-by: Daniel Kiper <[email protected]>
2843
2844 2025-07-11 Vladimir Serbinenko <[email protected]>
2845
2846 libgcrypt: Ignore sign-compare warnings
2847 libgcrypt itself is compiled with -Wno-sign-compare. Do the same for consistency.
2848
2849 Reviewed-by: Daniel Kiper <[email protected]>
2850
2851 2025-07-11 Vladimir Serbinenko <[email protected]>
2852
2853 libgcrypt: Remove now unneeded compilation flag
2854 HAVE_STRTOUL is now defined in stdlib.h. Include it in g10lib.h rather
2855 than defining on command line.
2856
2857 Reviewed-by: Daniel Kiper <[email protected]>
2858
2859 2025-07-11 Vladimir Serbinenko <[email protected]>
2860
2861 libgcrypt: Fix Coverity warnings
2862 Reviewed-by: Daniel Kiper <[email protected]>
2863
2864 2025-07-11 Vladimir Serbinenko <[email protected]>
2865
2866 keccak: Disable acceleration with SSE asm
2867 Libgcrypt code assumes that on x64 all SSE registers are fair game.
2868 While it's true that CPUs in question support it, we disable it in
2869 our compilation options. Disable the offending optimization.
2870
2871 Reviewed-by: Daniel Kiper <[email protected]>
2872
2873 2025-07-11 Vladimir Serbinenko <[email protected]>
2874
2875 tests: Add DSA and RSA SEXP tests
2876 This allows us to test purely the integration of the implementation of
2877 DSA and RSA from libgcrypt without concerning with additional code.
2878
2879 Reviewed-by: Daniel Kiper <[email protected]>
2880
2881 2025-07-11 Vladimir Serbinenko <[email protected]>
2882
2883 libgcrypt: Adjust import script, definitions and API users for libgcrypt 1.11
2884 This patches modifies the GRUB-libgcrypt API to match new libgcrypt 1.11.
2885
2886 Reviewed-by: Daniel Kiper <[email protected]>
2887
2888 2025-07-11 Vladimir Serbinenko <[email protected]>
2889
2890 b64dec: Add harness for compilation in GRUB environment
2891 Reviewed-by: Daniel Kiper <[email protected]>
2892
2893 2025-07-11 Vladimir Serbinenko <[email protected]>
2894
2895 b64dec: Import b64dec from gpg-error
2896 Imported from libgpg-error 1.51.
2897
2898 Reviewed-by: Daniel Kiper <[email protected]>
2899
2900 2025-07-11 Vladimir Serbinenko <[email protected]>
2901
2902 libgcrypt: Import libgcrypt 1.11
2903 We currently use an old version of libgcrypt which results in us having
2904 fewer ciphers and missing on many other improvements.
2905
2906 Reviewed-by: Daniel Kiper <[email protected]>
2907
2908 2025-07-11 Mate Kukri <[email protected]>
2909
2910 loader/efi/linux: Use shim loader image handle where available
2911 Not reusing these handles will result in image measurements showing up
2912 twice in the event log.
2913
2914 On the occasion add missing grub_free() call.
2915
2916 Reviewed-by: Daniel Kiper <[email protected]>
2917
2918 2025-07-11 Mate Kukri <[email protected]>
2919
2920 loader/efi/chainloader: Use shim loader image handle where available
2921 Not reusing these handles will result in image measurements showing up
2922 twice in the event log.
2923
2924 Reviewed-by: Daniel Kiper <[email protected]>
2925
2926 2025-07-11 Mate Kukri <[email protected]>
2927
2928 efi/sb: Add API for retrieving shim loader image handles
2929 Not reusing these handles will result in image measurements showing up
2930 twice in the event log.
2931
2932 Reviewed-by: Daniel Kiper <[email protected]>
2933
2934 2025-07-11 Mate Kukri <[email protected]>
2935
2936 efi/sb: Add support for the shim loader protocol
2937 Use loader protocol for image verification where available, otherwise
2938 fall back to the old shim lock protocol.
2939
2940 Reviewed-by: Daniel Kiper <[email protected]>
2941
2942 2025-07-11 Julian Andres Klode <[email protected]>
2943
2944 efi: Provide wrappers for load_image, start_image and unload_image
2945 These can be used to register a different implementation later,
2946 for example, when shim provides a protocol with those functions.
2947
2948 Reviewed-by: Daniel Kiper <[email protected]>
2949
2950 2025-07-11 Frediano Ziglio <[email protected]>
2951
2952 loader/arm64/xen_boot: Consider alignment calling grub_arch_efi_linux_boot_image()
2953 The Xen image is loaded with an alignment, not always at "start".
2954
2955 Reviewed-by: Daniel Kiper <[email protected]>
2956 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
2957
2958 2025-07-11 Frediano Ziglio <[email protected]>
2959
2960 loader/arm64/xen_boot: Use bool instead of int
2961 More readable, could consume less space.
2962
2963 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
2964 Reviewed-by: Daniel Kiper <[email protected]>
2965
2966 2025-07-11 Frediano Ziglio <[email protected]>
2967
2968 loader/arm64/xen_boot: Remove correctly all modules loaded by xen_module command
2969 We need to use FOR_LIST_ELEMENTS_SAFE() instead of FOR_LIST_ELEMENTS()
2970 as single_binary_unload(), called during the loop, is changing the list
2971 using grub_list_remove(). Given the environment probably the old code
2972 simply removed only the first module on the list not freeing all the others.
2973
2974 Reviewed-by: Daniel Kiper <[email protected]>
2975 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
2976
2977 2025-07-11 Gary Lin <[email protected]>
2978
2979 dl: Fix grub_dl_is_persistent() for emu
2980 When attempting to build grub-emu the compilation failed with the
2981 following error message:
2982
2983 include/grub/dl.h: In function ‘grub_dl_is_persistent’:
2984 include/grub/dl.h:262:1: error: no return statement in function returning non-void [-Werror=return-type]
2985
2986 To avoid the error make the function always return 0.
2987
2988 Fixes: ba8eadde6be1 (dl: Provide a fake grub_dl_set_persistent() and grub_dl_is_persistent() for the emu target)
2989
2990 Cc: Daniel Axtens <[email protected]>
2991 Cc: Sudhakar Kuppusamy <[email protected]>
2992 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
2993 Reviewed-by: Daniel Kiper <[email protected]>
2994
2995 2025-06-26 Lidong Chen <[email protected]>
2996
2997 loader/i386/pc/linux: Fix resource leak
2998 In grub_cmd_initrd(), memory is allocated for variable initrd_ctx
2999 before calling grub_relocator_alloc_chunk_align_safe(). When the
3000 function call fails, initrd_ctx should be freed before exiting
3001 grub_cmd_initrd().
3002
3003 Fixes: CID 473852
3004
3005 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3006 Reviewed-by: Daniel Kiper <[email protected]>
3007
3008 2025-06-26 Adriano Cordova <[email protected]>
3009
3010 loader/efi/linux: Unload previous Linux kernel/initrd before updating kernel size
3011 Unload previous Linux kernel/initrd before updating the global variable
3012 kernel_size. Otherwise the previous Linux kernel gets deallocated with
3013 the kernel_size of the Linux kernel that is being currently loaded.
3014
3015 Reviewed-by: Daniel Kiper <[email protected]>
3016
3017 2025-06-26 Frediano Ziglio <[email protected]>
3018
3019 loader/efi/linux: Correctly terminate load_options member
3020 If a simple string for arguments are passed it should be NUL terminated.
3021 This is true for other code but not for "linux" command.
3022
3023 Reviewed-by: Daniel Kiper <[email protected]>
3024
3025 2025-06-26 Frediano Ziglio <[email protected]>
3026
3027 loader/efi/linux: Use sizeof() instead of constant
3028 This is more consistent with the above code using sizeof(grub_efi_char16_t).
3029
3030 Reviewed-by: Daniel Kiper <[email protected]>
3031
3032 2025-06-26 Frediano Ziglio <[email protected]>
3033
3034 loader/efi/linux: Use proper type for len variable
3035 Although the length should not exceed 2^31 grub_size_t is more
3036 suitable for that variable. len is used to compute the size
3037 of buffers which in C is a size_t, not a int. It is used
3038 for GRUB_EFI_BYTES_TO_PAGES which expects unsigned values.
3039 It is assigned to load_options_size which is unsigned, not signed.
3040
3041 Reviewed-by: Daniel Kiper <[email protected]>
3042
3043 2025-06-26 Frediano Ziglio <[email protected]>
3044
3045 loader/efi/linux: Do not pass excessive size for source string
3046 The size passed to grub_utf8_to_utf16() for the source string is
3047 used as a limit for the string if NUL character is not encountered.
3048 However, len, which is "strlen(src) * 2 + 2" is surely greater than
3049 strlen(src). Pass the exact correct length.
3050
3051 Reviewed-by: Daniel Kiper <[email protected]>
3052
3053 2025-06-26 Frediano Ziglio <[email protected]>
3054
3055 loader/efi/linux: Remove useless assignment
3056 If the following allocation fails this would leave load_options NULL
3057 while load_options_size not valid. If the allocation succeed
3058 load_options_size is overwritten.
3059
3060 Reviewed-by: Daniel Kiper <[email protected]>
3061
3062 2025-06-26 Frediano Ziglio <[email protected]>
3063
3064 include/grub/charset.h: Update documentation
3065 (grub_size_t) -1 is never returned, the function always return
3066 a not negative values. This is important for overflows considerations.
3067
3068 Reviewed-by: Daniel Kiper <[email protected]>
3069
3070 2025-06-26 Lidong Chen <[email protected]>
3071
3072 Revert "lzma: Make sure we don't dereference past array"
3073 Commit 40e261b89b71 (lib/LzmaEnc: Validate "len" before subtracting)
3074 ensures that the variable len is at least 2. As a result, GetLenToPosState(len)
3075 never returns a value greater than or equal to kNumLenToPosStates,
3076 making the changes introduced in the commit 16c0dbf4bc6a (lzma: Make
3077 sure we don't dereference past array) unreachable and no longer necessary.
3078
3079 This reverts commit 16c0dbf4bc6a (lzma: Make sure we don't dereference past array).
3080
3081 Fixes: CID 481982
3082
3083 Reviewed-by: Daniel Kiper <[email protected]>
3084
3085 2025-06-26 Andrew Hamilton <[email protected]>
3086
3087 tests/util/grub-shell: Correct netboot and file_filter test failure
3088 Correct a test failure in netboot_test and file_filter_test caused by an
3089 issue cleaning up the tmp directory created for netboot. Netboot creates
3090 a subdirectory in the tmp folder that causes the rmdir to fail - so
3091 cleanup the subdirectory first.
3092
3093 Fixes: 1d59f39b5f1b (tests/util/grub-shell: Remove the work directory on successful run and debug is not on)
3094
3095 Tested-by: Leo Sandoval <[email protected]>
3096 Reviewed-by: Daniel Kiper <[email protected]>
3097
3098 2025-06-26 Lidong Chen <[email protected]>
3099
3100 normal/charset: Fix underflow and overflow in loop init
3101 In bidi_line_wrap(), "kk - 1" in the for loop init, "i = kk - 1",
3102 underflows when "kk" (unsigned int) is 0. Assigning the result of
3103 "kk - 1" to signed int "i" may cause overflow. To address both
3104 issues, cast "kk" to a signed type before subtraction to ensure
3105 safe arithmetic and assignment.
3106
3107 Fixed: CID 473874
3108
3109 Reviewed-by: Daniel Kiper <[email protected]>
3110 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3111
3112 2025-06-26 Daniel Axtens <[email protected]>
3113
3114 dl: Provide a fake grub_dl_set_persistent() and grub_dl_is_persistent() for the emu target
3115 Trying to start grub-emu with a module that calls grub_dl_set_persistent()
3116 and grub_dl_is_persistent() will crash because grub-emu fakes modules and
3117 passes NULL to the module init function.
3118
3119 Provide an empty function for the emu case.
3120
3121 Fixes: ee7808e2197c (dl: Add support for persistent modules)
3122
3123 Reviewed-by: Stefan Berger <[email protected]>
3124 Reviewed-by: Avnish Chouhan <[email protected]>
3125 Reviewed-by: Daniel Kiper <[email protected]>
3126
3127 2025-06-26 Andrew Hamilton <[email protected]>
3128
3129 util/grub-protect: Correct uninit "err" variable
3130 In function protect_tpm2_export_tpm2key(), the "err" variable
3131 is uninitialized in the normal (error free) path, so ensure this
3132 defaults to GRUB_ERR_NONE.
3133
3134 This causes the GRUB build to fail with clang (observed with clang-14).
3135
3136 Fixes: 5934bf51c (util/grub-protect: Support NV index mode)
3137
3138 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3139 Reviewed-by: Daniel Kiper <[email protected]>
3140
3141 2025-06-26 Lidong Chen <[email protected]>
3142
3143 gnulib: Bring back the fix for resolving unused variable issue
3144 This patch resolved a minor issue spotted by Coverity:
3145 a983d36bd917 (gnulib/regexec: Resolve unused variable)
3146
3147 But, it was removed by the Gnulib update:
3148 2b7902459803 (Update gnulib version and drop most gnulib patches)
3149
3150 It caused Coverity to continue to flag the issue. Daniel Kiper
3151 suggested to bring back the patch a983d36bd917 (gnulib/regexec: Resolve
3152 unused variable).
3153
3154 Fixes: CID 292459
3155
3156 Reviewed-by: Daniel Kiper <[email protected]>
3157
3158 2025-06-26 Andrew Hamilton <[email protected]>
3159
3160 gnulib: Add patch to allow GRUB w/GCC-15 compile
3161 Pull in Gnulib fix to allow lib/base64.c to compile using GCC 15 or newer.
3162
3163 Pulled from Gnulib commit 25df6dc425 (Silence some
3164 -Wunterminated-string-initialization warnings.)
3165
3166 GCC 15 adds a new compiler warning "-Wunterminated-string-initialization"
3167 that will trigger what is considered a false-positive in lib/base64.c as
3168 this array is not treated as a string but an array of characters so the
3169 lack of NUL string terminator is expected.
3170
3171 GCC team has added ability to flag such instances of arrays that the
3172 compiler may think are strings as "nonstring" arrays to avoid this
3173 warning: __attribute__((nonstring)).
3174
3175 Fixes: https://savannah.gnu.org/bugs/?66470
3176
3177 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3178 Reviewed-by: Daniel Kiper <[email protected]>
3179
3180 2025-06-17 Alec Brown <[email protected]>
3181
3182 gnulib/regexec: Fix resource leak
3183 In the function merge_state_with_log(), memory is allocated for the variable
3184 next_nodes when creating a union of the variables table_nodes and log_nodes.
3185 However, if next_state->entrance_nodes is NULL, then table_nodes becomes NULL
3186 and we still allocate memory to copy the content of log_nodes. This can cause
3187 a resource leak since we only free the memory for next_nodes if table_nodes
3188 isn't NULL. To prevent this, we need to check that next_state->entrance_nodes
3189 isn't NULL before allocating memory for the union.
3190
3191 This issue has been fixed in the latest version of gnulib and I've backported
3192 this change to maintain consistency.
3193
3194 This issue was found by a Coverity scan of GRUB2 under the CID 473887.
3195
3196 Fixes: CID 473887
3197
3198 Reviewed-by: Daniel Kiper <[email protected]>
3199
3200 2025-06-17 Alec Brown <[email protected]>
3201
3202 gnulib/regcomp: Fix resource leak
3203 In the functions create_initial_state() and calc_eclosure_iter(), memory
3204 is allocated for the elems member of a re_node_set structure but that
3205 memory isn't freed on error. Before returning an error, a call to
3206 re_node_set_free() should be made to prevent the resource leak.
3207
3208 This issue has been fixed in the latest version of gnulib and I've
3209 backported this change to maintain consistency.
3210
3211 This issue was found by a Coverity scan of GRUB2 under the following
3212 CIDs: 473869, 473888.
3213
3214 Fixes: CID 473869
3215 Fixes: CID 473888
3216
3217 Reviewed-by: Daniel Kiper <[email protected]>
3218
3219 2025-06-17 Gary Lin <[email protected]>
3220
3221 tests/tpm2_key_protector_test: Add tests for SHA-384 PCR bank
3222 Add a few more tests to seal and unseal the key with the SHA-384 PCR
3223 bank instead of the default SHA-256 PCR bank.
3224
3225 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3226 Reviewed-by: Stefan Berger <[email protected]>
3227 Reviewed-by: Daniel Kiper <[email protected]>
3228
3229 2025-06-17 Gary Lin <[email protected]>
3230
3231 tpm2_key_protector: Dump the PCR bank for key unsealing
3232 TPM 2.0 Key File format stores the PCR selection in the parameters
3233 for TPM2_PolicyPCR and it already contains the selected PCR bank.
3234 Currently, tpm2_key_protector dumped the PCR bank specified by the
3235 --bank option, and it may not be the PCR bank for key unsealing.
3236
3237 To dump the real PCR bank for key unsealing, this commit records the PCR
3238 bank used by TPM2_PolicyPCR and dumps PCR values from that bank when
3239 necessary.
3240
3241 Reviewed-by: Stefan Berger <[email protected]>
3242 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3243 Reviewed-by: Daniel Kiper <[email protected]>
3244
3245 2025-06-17 Gary Lin <[email protected]>
3246
3247 util/grub-protect: Fix the hash algorithm of PCR digest
3248 For tpm2_key_protector and grub-protect, SHA-256 is chosen as the hash
3249 algorithm for the TPM session. However, grub-protect mistakenly used the
3250 hash algorithm of the PCR bank to calculate PCR digest. If the user
3251 chose a PCR bank other than SHA-256, grub-protect created a non-SHA-256
3252 PCR digest to seal the key. But, tpm2_key_protector expects a SHA-256
3253 PCR digest to the TPM unsealing session, so it would fail due to digest
3254 mismatch.
3255
3256 This commit fixes the hash algorithm of PCR digest in grub-protect to
3257 avoid the potential unsealing failure.
3258
3259 Fixes: https://github.com/lcp/grub2/issues/4
3260
3261 Reviewed-by: Stefan Berger <[email protected]>
3262 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3263 Reviewed-by: Daniel Kiper <[email protected]>
3264
3265 2025-06-17 Andrew Hamilton <[email protected]>
3266
3267 build: Add new header files to dist to allow building from tar
3268 Several new header files have been added to GRUB which need
3269 to be manually added to the dist archive. This allows building
3270 from the tar archive created by "make dist".
3271
3272 Reviewed-by: Daniel Kiper <[email protected]>
3273
3274 2025-06-17 Andrew Hamilton <[email protected]>
3275
3276 build: Remove extra_deps.lst from EXTRA_DIST
3277 This file is auto-generated based on the selected platform and should
3278 not be included in the source tarball.
3279
3280 Fixes: 6744840b (build: Track explicit module dependencies in Makefile.core.def)
3281
3282 Reviewed-by: Daniel Kiper <[email protected]>
3283
3284 2025-06-17 Lidong Chen <[email protected]>
3285
3286 lib/LzmaEnc: Validate "len" before subtracting
3287 In LzmaEnc_CodeOneBlock(), both GetOptimumFast() and GetOptimum()
3288 returns a value of greater or equal to 1, which is assigned to
3289 "len". But since LZMA_MATCH_LEN_MIN == 2, "len" should be validated
3290 before performing "len - LZMA_MATCH_LEN_MIN" to avoid underflow
3291 when "len" equals to 1.
3292
3293 Fixes: CID 51508
3294
3295 Reviewed-by: Daniel Kiper <[email protected]>
3296 Reviewed-by: Sudhakar Kuppusamy <[email protected]>
3297
3298 2025-06-12 Lidong Chen <[email protected]>
3299
3300 osdep/unix/hostdisk: Fix signed integer overflow
3301 The potential overflow issue arises at "size += ret;" because "size"
3302 is of type ssize_t (signed) while "len" is size_t (unsigned). Repeatedly
3303 adding read sizes, "ret", to "size" can potentially exceed the maximum
3304 value of ssize_t, causing it to overflow into a negative or incorrect value.
3305 The fix is to ensure "len" is within the range of SSIZE_MAX.
3306
3307 Fixes: CID 473850
3308 Fixes: CID 473863
3309
3310 Reviewed-by: Daniel Kiper <[email protected]>
3311
3312 2025-05-29 Egor Ignatov <[email protected]>
3313
3314 disk/luks2: Add attempting to decrypt message to align with luks and geli modules
3315 Reviewed-by: Daniel Kiper <[email protected]>
3316
3317 2025-05-29 Renaud Métrich <[email protected]>
3318
3319 osdep/linux/getroot: Detect DDF container similar to IMSM
3320 Similarly to Intel IMSM, there are BIOS and UEFI implementations that
3321 support DDF containers natively.
3322
3323 DDF and IMSM are very similar in handling, especially these should not
3324 be considered as RAID abstraction. This fixes the requirement of having
3325 a device map when probing DDF containers.
3326
3327 Fixes: https://issues.redhat.com/browse/RHEL-44336
3328
3329 Reviewed-by: Daniel Kiper <[email protected]>
3330
3331 2025-05-29 Andrew Hamilton <[email protected]>
3332
3333 fs/fshelp: Avoid possible NULL pointer deference
3334 Avoid attempting to defererence a NULL pointer to call read_symlink() when
3335 the given filesystem does not provide a read_symlink() function. This could
3336 be triggered if the calling filesystem had a file marked as a symlink.
3337 This appears possible for HFS and was observed during fuzzing of NTFS.
3338
3339 Reviewed-by: Vladimir Serbinenko <[email protected]>
3340 Reviewed-by: Daniel Kiper <[email protected]>
3341
3342 2025-05-29 Andrew Hamilton <[email protected]>
3343
3344 fs/ntfs: Correct possible infinite loops/hangs
3345 Correct several infinite loops/hangs found during fuzzing. The issues
3346 fixed here could occur if certain specific malformed NTFS file systems
3347 were presented to GRUB. Currently, GRUB does not allow NTFS file system
3348 access when lockdown mode is enforced, so these should be of minimal
3349 impact.
3350
3351 The changes made in this commit generally correct issues such as attempting
3352 to iterate through a buffer using a length read from the NTFS file system
3353 without confirming the length is larger than 0.
3354
3355 Reviewed-by: Daniel Kiper <[email protected]>
3356
3357 2025-05-29 Andrew Hamilton <[email protected]>
3358
3359 fs/ntfs: Correct possible access violations
3360 Correct several memory access violations found during fuzzing.
3361 The issues fixed here could occur if certain specific malformed NTFS
3362 file systems were presented to GRUB. Currently, GRUB does not allow NTFS
3363 file system access when lockdown mode is enforced, so these should be of
3364 minimal impact.
3365
3366 The changes made in this commit generally correct issues where pointers
3367 into data buffers were being calculated using lengths read from the
3368 NTFS file system without sufficient bounds/sanity checking; or
3369 attempting to access elements of a structure to free them, when the
3370 structure pointer is NULL.
3371
3372 Reviewed-by: Daniel Kiper <[email protected]>
3373
3374 2025-05-29 Andrew Hamilton <[email protected]>
3375
3376 fs/ntfs: Correct attribute vs attribute list validation
3377 Correct ntfs_test test failures around attempting to validate attribute
3378 list entries as attributes. The NTFS code uses common logic in some
3379 places to parse both attributes and attribute_lists which complicates
3380 validation. Attribute lists contain different headers including a
3381 different size of the length field (2 bytes) at offset 4 instead of the
3382 4 byte length field used in attributes at offset 4. There are other
3383 differences as well, but attempting to validate attribute list types
3384 using attribute header validation was causing failure of the NTFS test
3385 suite. This change restores some of the validation logic which may be
3386 shared between attributes and attribute lists to be closer to the
3387 original logic prior to fixes for previous CVEs. A following commit will
3388 address some of the implications of removing this validation logic by
3389 correcting some fuzzer failures (some which are exposed by removing the
3390 validation in some of the cases).
3391
3392 Fixes: 067b6d225 (fs/ntfs: Implement attribute verification)
3393
3394 Reviewed-by: Daniel Kiper <[email protected]>
3395
3396 2025-05-29 Andrew Hamilton <[email protected]>
3397
3398 fs/ntfs: Correct regression with run list calculation
3399 Correct ntfs_test test failures around attempting to validate attribute
3400 run list values. The calculation was incorrect for the "curr" variable.
3401 With previous calculation, some file systems would fail validation
3402 despite being well-formed and valid. This was caused by incrementing
3403 "curr" by min_size which included both the (already accounted for)
3404 min_size as well as the size of the run list. Correct by making a new
3405 variable "run_size" to denote the current run list size to increment
3406 both "curr" and "min_size" separately.
3407
3408 Fixes: 067b6d225 (fs/ntfs: Implement attribute verification)
3409
3410 Reviewed-by: Daniel Kiper <[email protected]>
3411
3412 2025-05-29 Shreenidhi Shedi <[email protected]>
3413
3414 lib/envblk: Ignore empty new lines while parsing env files
3415 Environment files may contain empty lines, which should be ignored
3416 during parsing. Currently, these lines are not skipped and resulting in
3417 incorrect behavior. This patch adds a check to skip empty lines along
3418 with those starting with "#".
3419
3420 Reviewed-by: Alexey Makhalov <[email protected]>
3421 Reviewed-by: Daniel Kiper <[email protected]>
3422
3423 2025-05-29 Glenn Washburn <[email protected]>
3424
3425 fs/zfs: Fix another memory leak in ZFS code
3426 Commit b66c6f918 (fs/zfs: Fix a number of memory leaks in ZFS code)
3427 fixes many of the same leaks detected in bug #63846 except one, which
3428 is fixed here.
3429
3430 Fixes: https://savannah.gnu.org/bugs/?63846
3431 Fixes: b66c6f918 (fs/zfs: Fix a number of memory leaks in ZFS code)
3432
3433 Reviewed-by: Daniel Kiper <[email protected]>
3434
3435 2025-05-29 Glenn Washburn <[email protected]>
3436
3437 tests: Disable gfxterm_menu and cmdline_cat tests
3438 Those tests fail depending on the version of unifont. As we don't distribute
3439 our own unifont it fails for most users. Disable them so that they don't mask
3440 real failures. They can be reinstated once we solve unifont problem.
3441
3442 Reviewed-by: Daniel Kiper <[email protected]>
3443
3444 2025-05-06 Michael Chang <[email protected]>
3445
3446 cryptocheck: Add --quiet option
3447 The option can be used to suppress output if we only want to test the
3448 return value of the command.
3449
3450 Also, mention this option in the documentation.
3451
3452 Reviewed-by: Daniel Kiper <[email protected]>
3453
3454 2025-05-06 Maxim Suhanov <[email protected]>
3455
3456 disk/cryptodisk: Wipe the passphrase from memory
3457 Switching to another EFI boot application while there are secrets in
3458 RAM is dangerous, because not all firmware is wiping memory on free.
3459
3460 To reduce the attack surface, wipe the passphrase acquired when
3461 unlocking an encrypted volume.
3462
3463 Reviewed-by: Daniel Kiper <[email protected]>
3464
3465 2025-05-06 Maxim Suhanov <[email protected]>
3466
3467 disk/cryptodisk: Add the "erase secrets" function
3468 This commit adds the grub_cryptodisk_erasesecrets() function to wipe
3469 master keys from all cryptodisks. This function is EFI-only.
3470
3471 Since there is no easy way to "force unmount" a given encrypted disk,
3472 this function renders all mounted cryptodisks unusable. An attempt to
3473 read them will return garbage.
3474
3475 This is why this function must be used in "no way back" conditions.
3476
3477 Currently, it is used when unloading the cryptodisk module and when
3478 performing the "exit" command (it is often used to switch to the next
3479 EFI application). This function is not called when performing the
3480 "chainloader" command, because the callee may return to GRUB. For this
3481 reason, users are encouraged to use "exit" instead of "chainloader" to
3482 execute third-party boot applications.
3483
3484 This function does not guarantee that all secrets are wiped from RAM.
3485 Console output, chunks from disk read requests and other may remain.
3486
3487 This function does not clear the IV prefix and rekey key for geli disks.
3488
3489 Also, this commit adds the relevant documentation improvements.
3490
3491 Reviewed-by: Daniel Kiper <[email protected]>
3492
3493 2025-05-06 Maxim Suhanov <[email protected]>
3494
3495 docs: Document available crypto disks checks
3496 Document the --cryptodisk-only argument. Also, document the
3497 "cryptocheck" command invoked when that argument is processed.
3498
3499 Reviewed-by: Daniel Kiper <[email protected]>
3500
3501 2025-05-06 Maxim Suhanov <[email protected]>
3502
3503 commands/search: Add the diskfilter support
3504 When the --cryptodisk-only argument is given, also check the target
3505 device using the "cryptocheck" command, if available.
3506
3507 This extends the checks to common layouts like LVM-on-LUKS, so the
3508 --cryptodisk-only argument transparently handles such setups.
3509
3510 Reviewed-by: Daniel Kiper <[email protected]>
3511
3512 2025-05-06 Maxim Suhanov <[email protected]>
3513
3514 disk/diskfilter: Introduce the "cryptocheck" command
3515 This command examines a given diskfilter device, e.g., an LVM disk,
3516 and checks if underlying disks, physical volumes, are cryptodisks,
3517 e.g., LUKS disks, this layout is called "LVM-on-LUKS".
3518
3519 The return value is 0 when all underlying disks (of a given device)
3520 are cryptodisks (1 if at least one disk is unencrypted or in an
3521 unknown state).
3522
3523 Users are encouraged to include the relevant check before loading
3524 anything from an LVM disk that is supposed to be encrypted.
3525
3526 This further supports the CLI authentication, blocking bypass
3527 attempts when booting from an encrypted LVM disk.
3528
3529 Reviewed-by: Daniel Kiper <[email protected]>
3530
3531 2025-05-06 Maxim Suhanov <[email protected]>
3532
3533 commands/search: Introduce the --cryptodisk-only argument
3534 This allows users to restrict the "search" command's scope to
3535 encrypted disks only.
3536
3537 Typically, this command is used to "rebase" $root and $prefix
3538 before loading additional configuration files via "source" or
3539 "configfile". Unfortunately, this leads to security problems,
3540 like CVE-2023-4001, when an unexpected, attacker-controlled
3541 device is chosen by the "search" command.
3542
3543 The --cryptodisk-only argument allows users to ensure that the
3544 file system picked is encrypted.
3545
3546 This feature supports the CLI authentication, blocking bypass
3547 attempts.
3548
3549 Reviewed-by: Daniel Kiper <[email protected]>
3550
3551 2025-05-06 Maxim Suhanov <[email protected]>
3552
3553 kern/rescue_reader: Block the rescue mode until the CLI authentication
3554 This further mitigates potential misuse of the CLI after the
3555 root device has been successfully unlocked via TPM.
3556
3557 Fixes: CVE-2025-4382
3558
3559 Reviewed-by: Daniel Kiper <[email protected]>
3560
3561 2025-04-23 Eric Sandeen <[email protected]>
3562
3563 fs/xfs: Fix large extent counters incompat feature support
3564 When large extent counter / NREXT64 support was added to GRUB, it missed
3565 a couple of direct reads of nextents which need to be changed to the new
3566 NREXT64-aware helper as well. Without this, we'll have mis-reads of some
3567 directories with this feature enabled.
3568
3569 The large extent counter fix likely raced on merge with commit 07318ee7e
3570 (fs/xfs: Fix XFS directory extent parsing) which added the new direct
3571 nextents reads just prior, causing this issue.
3572
3573 Fixes: aa7c1322671e (fs/xfs: Add large extent counters incompat feature support)
3574
3575 Reviewed-by: Anthony Iliopoulos <[email protected]>
3576 Reviewed-by: Jon DeVree <[email protected]>
3577 Reviewed-by: Vladimir Serbinenko <[email protected]>
3578 Reviewed-by: Daniel Kiper <[email protected]>
3579
3580 2025-04-10 Egor Ignatov <[email protected]>
3581
3582 util/grub-install: Include raid5rec module for RAID 4 as well
3583 RAID 4 requires the same recovery module as RAID 5. Extend the condition to
3584 cover both RAID levels.
3585
3586 Reviewed-by: Daniel Kiper <[email protected]>
3587
3588 2025-04-10 Vladimir Serbinenko <[email protected]>
3589
3590 loader/ia64/efi/linux: Reset grub_errno on failure to allocate
3591 The code goes on to allocate memory in another region on failure, hence
3592 it should discard the error.
3593
3594 Reviewed-by: Daniel Kiper <[email protected]>
3595
3596 2025-04-10 Vladimir Serbinenko <[email protected]>
3597
3598 lib/datetime: Specify license in emu module
3599 Other platforms specify license in platform-specific files but corresponding
3600 code for emu is in kernel, so datetime ends up without license section.
3601
3602 Reviewed-by: Daniel Kiper <[email protected]>
3603
3604 2025-04-10 Vladimir Serbinenko <[email protected]>
3605
3606 configure: Add -mno-relax on riscv*
3607 Without this option compiler sometimes emits R_RISCV_ALIGN relocs.
3608 Unlike other relocs this one requires the linker to do NOP deletions
3609 and we can't ignore them. Just instruct compiler not to emit them.
3610
3611 Reviewed-by: Daniel Kiper <[email protected]>
3612
3613 2025-04-10 Gary Lin <[email protected]>
3614
3615 docs: Document the long options of tpm2_key_protect_init
3616 Add the long options of tpm2_key_protect_init along with the short options.
3617
3618 Reviewed-by: Daniel Kiper <[email protected]>
3619
3620 2025-04-10 Gary Lin <[email protected]>
3621
3622 INSTALL: Document the packages needed for TPM2 key protector tests
3623 The TPM2 key protector tests require two external packages: swtpm-tools
3624 and tpm2-tools. Add those two packages to the INSTALL file to inform
3625 the user to install those packages before starting the TPM2 key protector
3626 tests.
3627
3628 Reviewed-by: Daniel Kiper <[email protected]>
3629 Reviewed-by: Stefan Berger <[email protected]>
3630
3631 2025-04-10 Gary Lin <[email protected]>
3632
3633 docs: Update NV index mode of TPM2 key protector
3634 This commit updates the NV index mode section and the grub-protect
3635 section to reflect the recent changes in TPM2 key protector and
3636 grub-protect.
3637
3638 Reviewed-by: Daniel Kiper <[email protected]>
3639
3640 2025-04-10 Gary Lin <[email protected]>
3641
3642 tests/tpm2_key_protector_test: Add more NV index mode tests
3643 Two more NV index test cases are added to test key sealing and
3644 unsealing with the NV index handle 0x1000000.
3645
3646 Reviewed-by: Stefan Berger <[email protected]>
3647 Reviewed-by: Daniel Kiper <[email protected]>
3648
3649 2025-04-10 Gary Lin <[email protected]>
3650
3651 tests/tpm2_key_protector_test: Reset "ret" on fail
3652 Reset "ret" to 0 when a test case fails so that the other test cases
3653 could continue.
3654
3655 Also set the exit status to 1 when encountering a failure to reflect the
3656 test result.
3657
3658 Reviewed-by: Stefan Berger <[email protected]>
3659 Reviewed-by: Daniel Kiper <[email protected]>
3660
3661 2025-04-10 Gary Lin <[email protected]>
3662
3663 tests/tpm2_key_protector_test: Simplify the NV index mode test
3664 Since grub-protect already supports NV index mode, tpm2_seal_nv() is
3665 replaced with one grub-protect command to simplify the test script.
3666
3667 "tpm2_evictcontrol" is also replaced with "grub-protect --tpm2-evict".
3668
3669 Reviewed-by: Stefan Berger <[email protected]>
3670 Reviewed-by: Daniel Kiper <[email protected]>
3671
3672 2025-04-10 Gary Lin <[email protected]>
3673
3674 util/grub-protect: Support NV index mode
3675 This commit implements the missing NV index mode support in grub-protect.
3676 NV index mode stores the sealed key in the TPM non-volatile memory (NVRAM)
3677 instead of a file. There are two supported types of TPM handles.
3678
3679 1. Persistent handle (0x81000000~0x81FFFFFF)
3680 Only the raw format is supported due to the limitation of persistent
3681 handles. This grub-protect command seals the key into the
3682 persistent handle 0x81000000.
3683
3684 # grub-protect \
3685 --protector=tpm2 \
3686 --action=add \
3687 --tpm2-bank=sha256 \
3688 --tpm2-pcrs=7,11 \
3689 --tpm2-keyfile=luks-key \
3690 --tpm2-nvindex=0x81000000
3691
3692 2. NV index handle (0x1000000~0x1FFFFFF)
3693 Both TPM 2.0 Key File format and the raw format are supported by NV
3694 index handles. Here is the grub-protect command to seal the key in
3695 TPM 2.0 Key File format into the NV index handle 0x1000000.
3696
3697 # grub-protect \
3698 --protector=tpm2 \
3699 --action=add \
3700 --tpm2key \
3701 --tpm2-bank=sha256 \
3702 --tpm2-pcrs=7,11 \
3703 --tpm2-keyfile=luks-key \
3704 --tpm2-nvindex=0x1000000
3705
3706 Besides the "add" action, the corresponding "remove" action is also
3707 introduced. To remove the data from a persistent or NV index handle,
3708 just use "--tpm2-nvindex=HANDLE" combining with "--tpm2-evict". This
3709 sample command removes the data from the NV index handle 0x1000000.
3710
3711 # grub-protect \
3712 --protector=tpm2 \
3713 --action=remove \
3714 --tpm2-evict \
3715 --tpm2-nvindex=0x1000000
3716
3717 Also set and check the boolean variables with true/false instead of 1/0.
3718
3719 Reviewed-by: Stefan Berger <[email protected]>
3720 Reviewed-by: Daniel Kiper <[email protected]>
3721
3722 2025-04-10 Gary Lin <[email protected]>
3723
3724 tpm2_key_protector: Support NV index handles
3725 Previously, NV index mode only supported persistent handles which are
3726 only for TPM objects.
3727
3728 On the other hand, the "NV index" handle allows the user-defined data,
3729 so it can be an alternative to the key file and support TPM 2.0 Key
3730 File format immediately.
3731
3732 The following tpm2-tools commands store the given key file, sealed.tpm,
3733 in either TPM 2.0 Key File format or the raw format into the NV index
3734 handle 0x1000000.
3735
3736 # tpm2_nvdefine -C o \
3737 -a "ownerread|ownerwrite" \
3738 -s $(stat -c %s sealed.tpm) \
3739 0x1000000
3740 # tpm2_nvwrite -C o -i sealed.tpm 0x1000000
3741
3742 To unseal the key in GRUB, add the "tpm2_key_protector_init" command to
3743 grub.cfg:
3744
3745 tpm2_key_protector_init --mode=nv --nvindex=0x1000000
3746 cryptomount -u <UUID> --protector tpm2
3747
3748 To remove the NV index handle:
3749
3750 # tpm2_nvundefine -C o 0x1000000
3751
3752 Reviewed-by: Daniel Kiper <[email protected]>
3753
3754 2025-04-10 Gary Lin <[email protected]>
3755
3756 tpm2_key_protector: Unseal key from a buffer
3757 Extract the logic to handle the file buffer from the SRK recover
3758 function to prepare to load the sealed key from the NV index handle,
3759 so the NV index mode can share the same code path in the later patch.
3760 The SRK recover function now only reads the file and sends the file
3761 buffer to the new function.
3762
3763 Besides this, to avoid introducing more options for the NV index mode,
3764 the file format is detected automatically before unmarshaling the data,
3765 so there is no need to use the command option to specify the file format
3766 anymore. In other words, "-T" and "-k" are the same now.
3767
3768 Also update grub.text to address the change.
3769
3770 Reviewed-by: Stefan Berger <[email protected]>
3771 Reviewed-by: Daniel Kiper <[email protected]>
3772
3773 2025-04-10 Gary Lin <[email protected]>
3774
3775 tss2: Add TPM 2.0 NV index commands
3776 The following TPM 2.0 commands are introduced to tss2 to access the
3777 TPM non-volatile memory associated with the NV index handles:
3778 - TPM2_NV_DefineSpace,
3779 - TPM2_NV_UndefineSpace,
3780 - TPM2_NV_ReadPublic,
3781 - TPM2_NV_Read,
3782 - TPM2_NV_Write.
3783
3784 The related marshal/unmarshal functions are also introduced.
3785
3786 Reviewed-by: Stefan Berger <[email protected]>
3787 Reviewed-by: Daniel Kiper <[email protected]>
3788
3789 2025-04-10 Gary Lin <[email protected]>
3790
3791 tss2: Fix the missing authCommand
3792 grub_tpm2_readpublic() and grub_tpm2_testparms() didn't check
3793 authCommand when marshaling the input data buffer. Currently, there is
3794 no caller using non-NULL authCommand. However, to avoid the potential
3795 issue, the conditional check is added to insert authCommand into the
3796 input buffer if necessary.
3797
3798 Also fix a few pointer checks.
3799
3800 Reviewed-by: Stefan Berger <[email protected]>
3801 Reviewed-by: Daniel Kiper <[email protected]>
3802
3803 2025-04-10 Gary Lin <[email protected]>
3804
3805 tpm2_key_protector: Add tpm2_dump_pcr command
3806 The user may need to inspect the TPM 2.0 PCR values with the GRUB shell,
3807 so the new tpm2_dump_pcr command is added to print all PCRs of the
3808 specified bank.
3809
3810 Also update the document for the new command.
3811
3812 Tested-by: Stefan Berger <[email protected]>
3813 Reviewed-by: Daniel Kiper <[email protected]>
3814
3815 2025-04-10 Gary Lin <[email protected]>
3816
3817 tpm2_key_protector: Dump PCRs on policy fail
3818 PCR mismatch is one common cause of TPM key unsealing fail. Since the
3819 system may be compromised, it is not safe to boot into OS to get the PCR
3820 values and TPM eventlog for the further investigation.
3821
3822 To provide some hints, GRUB now dumps PCRs on policy fail, so the user
3823 can check the current PCR values. PCR 0~15 are chosen to cover the
3824 firmware, bootloader, and OS.
3825
3826 The sample output:
3827
3828 PCR Mismatch! Check firmware and bootloader before typing passphrase!
3829 TPM PCR [sha256]:
3830 00: 17401f37710984c1d8a03a81fff3ab567ae9291bac61e21715b890ee28879738
3831 01: 7a114329ba388445a96e8db2a072785937c1b7a8803ed7cc682b87f3ff3dd7a8
3832 02: 11c2776849e8e24b7d80c926cbc4257871bffa744dadfefd3ed049ce25143e05
3833 03: 6c33b362073e28e30b47302bbdd3e6f9cee4debca3a304e646f8c68245724350
3834 04: 62d38838483ecfd2484ee3a2e5450d8ca3b35fc72cda6a8c620f9f43521c37d1
3835 05: d8a85cb37221ab7d1f2cc5f554dbe0463acb6784b5b8dc3164ccaa66d8fff0e1
3836 06: 9262e37cbe71ed4daf815b4a4881fb7251c9d371092dde827557d5368121e10e
3837 07: 219d542233be492d62b079ffe46cf13396a8c27e520e88b08eaf2e6d3b7e70f5
3838 08: de1f61c973b673e505adebe0d7e8fb65fde6c24dd4ab4fbaff9e28b18df6ecd3
3839 09: c1de7274fa3e879a16d7e6e7629e3463d95f68adcfd17c477183846dccc41c89
3840 10: 0000000000000000000000000000000000000000000000000000000000000000
3841 11: 0000000000000000000000000000000000000000000000000000000000000000
3842 12: 0000000000000000000000000000000000000000000000000000000000000000
3843 13: 0000000000000000000000000000000000000000000000000000000000000000
3844 14: 9ab9ebe4879a7f4dd00c04f37e79cfd69d0dd7a8bcc6b01135525b67676a3e40
3845 15: 0000000000000000000000000000000000000000000000000000000000000000
3846 16: 0000000000000000000000000000000000000000000000000000000000000000
3847 17: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
3848 18: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
3849 19: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
3850 20: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
3851 21: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
3852 22: ffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff
3853 23: 0000000000000000000000000000000000000000000000000000000000000000
3854 error: failed to unseal sealed key (TPM2_Unseal: 0x99d).
3855 error: no key protector provided a usable key for luks (af16e48f-746b-4a12-aae1-c14dcee429e0).
3856
3857 If the user happens to have the PCR values for key sealing, the PCR dump
3858 can be used to identify the changed PCRs and narrow down the scope for
3859 closer inspection.
3860
3861 Please note that the PCR dump is trustworthy only if the GRUB binary is
3862 authentic, so the user has to check the GRUB binary thoroughly before
3863 using the PCR dump.
3864
3865 Reviewed-by: Stefan Berger <[email protected]>
3866 Reviewed-by: Daniel Kiper <[email protected]>
3867
3868 2025-04-04 Patrick Colp <[email protected]>
3869
3870 loader/i386/linux: Update linux_kernel_params to match upstream
3871 Update linux_kernel_params to match the v6.13.7 upstream version of boot_params.
3872 Refactor most things out into structs, as the Linux kernel does.
3873
3874 edid_info should be a struct with "unsigned char dummy[128]" and efi_info should
3875 be a struct as well, starting at 0x1c0. However, for backwards compatibility,
3876 GRUB can have efi_systab at 0x1b8 and padding at 0x1bc (or padding at both spots).
3877 This cuts into the end of edid_info. Make edid_info inline and only make it go
3878 up to 0x1b8.
3879
3880 Reviewed-by: Daniel Kiper <[email protected]>
3881
3882 2025-04-04 Lidong Chen <[email protected]>
3883
3884 loader/xnu: Fix memory leak
3885 In grub_xnu_load_kext_from_dir(), when the call to grub_device_open()
3886 failed, it simply cleaned up previously allocated memory and returned
3887 GRUB_ERR_NONE. However, it neglected to free ctx->newdirname which is
3888 allocated before the call to grub_device_open().
3889
3890 Fixes: CID 473859
3891
3892 Reviewed-by: Vladimir Serbinenko <[email protected]>
3893 Reviewed-by: Daniel Kiper <[email protected]>
3894
3895 2025-04-04 Lidong Chen <[email protected]>
3896
3897 fs/btrfs: Fix memory leaks
3898 Fix memory leaks in grub_btrfs_extent_read() and grub_btrfs_dir().
3899
3900 Fixes: CID 473842
3901 Fixes: CID 473871
3902
3903 Reviewed-by: Vladimir Serbinenko <[email protected]>
3904 Reviewed-by: Daniel Kiper <[email protected]>
3905
3906 2025-04-04 Lidong Chen <[email protected]>
3907
3908 loader/i386/linux: Fix resource leak
3909 In grub_cmd_initrd(), initrd_ctx is allocated before calling
3910 grub_relocator_alloc_chunk_align(). When that function fails,
3911 initrd_ctx should be freed before exiting grub_cmd_initrd().
3912
3913 Fixes: CID 473852
3914
3915 Reviewed-by: Daniel Kiper <[email protected]>
3916
3917 2025-04-04 Lidong Chen <[email protected]>
3918
3919 lib/reloacator: Fix memory leaks
3920 Fix memory leaks in grub_relocator_alloc_chunk_align().
3921
3922 Fixes: CID 473844
3923
3924 Reviewed-by: Vladimir Serbinenko <[email protected]>
3925 Reviewed-by: Daniel Kiper <[email protected]>
3926
3927 2025-04-04 Lidong Chen <[email protected]>
3928
3929 disk/ldm: Fix memory leaks
3930 Fix memory leaks in make_vg() with new helper functions, free_pv()
3931 and free_lv(). Additionally, correct a check after allocating
3932 comp->segments->nodes that mistakenly checked lv->segments->nodes
3933 instead, likely due to a copy-paste error.
3934
3935 Fixes: CID 473878
3936 Fixes: CID 473884
3937 Fixes: CID 473889
3938 Fixes: CID 473890
3939
3940 Reviewed-by: Vladimir Serbinenko <[email protected]>
3941 Reviewed-by: Daniel Kiper <[email protected]>
3942
3943 2025-03-26 Andrew Hamilton <[email protected]>
3944
3945 fs/ntfs: Fix NULL pointer dereference and possible infinite loop
3946 A regression was introduced recently as a part of the series of
3947 filesystem related patches to address some CVEs found in GRUB.
3948
3949 This issue may cause either an infinite loop at startup when
3950 accessing certain valid NTFS filesystems, or may cause a crash
3951 due to a NULL pointer dereference on systems where NULL address
3952 is invalid (such as may happen when calling grub-mount from
3953 the operating system level).
3954
3955 Correct this issue by checking that at->attr_cur is within bounds
3956 inside find_attr().
3957
3958 Fixes: https://savannah.gnu.org/bugs/?66855
3959 Fixes: aff263187 (fs/ntfs: Fix out-of-bounds read)
3960
3961 Reviewed-by: Daniel Kiper <[email protected]
3962 Reviewed-by: Vladimir Serbinenko <[email protected]>
3963
3964 2025-03-26 Nicolas Frayer <[email protected]>
3965
3966 net/drivers/ieee1275/ofnet: Add missing grub_malloc()
3967 The grub_malloc() has been inadvertently removed from the code after it
3968 has been modified to use safe math functions.
3969
3970 Fixes: 4beeff8a (net: Use safe math macros to prevent overflows)
3971
3972 Tested-by: Marta Lewandowska <[email protected]>
3973 Reviewed-by: Daniel Kiper <[email protected]>
3974
3975 2025-03-26 Avnish Chouhan <[email protected]>
3976
3977 kern/ieee1275/init: Increase MIN_RMA size for CAS negotiation on PowerPC machines
3978 Change RMA size from 512 MB to 768 MB which will result in more memory
3979 at boot time for PowerPC. When vTPM, Secure Boot or FADump are enabled
3980 on PowerPC the 512 MB RMA memory is not sufficient for boot. With this
3981 512 MB RMA, GRUB runs out of memory and fails to boot the machine.
3982 Sometimes even usage of CDROM requires more memory for installation and
3983 along with the options mentioned above exhausts the boot memory which
3984 results in boot failures. Increasing the RMA size will resolves multiple
3985 out of memory issues observed on PowerPC machines.
3986
3987 Failure details (GRUB debug console dump):
3988
3989 kern/ieee1275/init.c:550: mm requested region of size 8513000, flags 1
3990 kern/ieee1275/init.c:563: Cannot satisfy allocation and retain minimum runtime space
3991 kern/ieee1275/init.c:550: mm requested region of size 8513000, flags 0
3992 kern/ieee1275/init.c:563: Cannot satisfy allocation and retain minimum runtime space
3993 kern/file.c:215: Closing `/ppc/ppc64/initrd.img' ...
3994 kern/disk.c:297: Closing `ieee1275//vdevice/v-scsi@30000067/disk@8300000000000000'...
3995 kern/disk.c:311: Closing `ieee1275//vdevice/v-scsi@30000067/disk@8300000000000000' succeeded.
3996 kern/file.c:225: Closing `/ppc/ppc64/initrd.img' failed with 3.
3997 kern/file.c:148: Opening `/ppc/ppc64/initrd.img' succeeded.
3998 error: ../../grub-core/kern/mm.c:552:out of memory.
3999
4000 Reviewed-by: Daniel Kiper <[email protected]>
4001
4002 2025-03-26 Stuart Hayes <[email protected]>
4003
4004 fs/zfs: Fix a number of memory leaks in ZFS code
4005 Without this fix the GRUB failed to boot linux with "out of memory" after
4006 trying to run a "search --fs-uuid..." on a system that has 7 ZFS pools
4007 across about 80 drives.
4008
4009 Reviewed-by: Vladimir Serbinenko <[email protected]>
4010 Reviewed-by: Daniel Kiper <[email protected]>
4011
4012 2025-03-26 Glenn Washburn <[email protected]>
4013
4014 tests/util/grub-shell: Remove the work directory on successful run and debug is not on
4015 This removes a lot of empty grub-shell working directories in the TMPDIR directory.
4016
4017 Tested-by: Thomas Schmitt <[email protected]>
4018 Reviewed-by: Daniel Kiper <[email protected]>
4019
4020 2025-03-26 Thomas Schmitt <[email protected]>
4021
4022 tests/grub_cmd_cryptomount: Remove temporary directories if successful and debug is not on
4023 grub_cmd_cryptomount creates a directory per subtest. If a subtest is
4024 successful and debugging is not on, the directory should be empty.
4025 So, it can be deleted.
4026
4027 Tested-by: Thomas Schmitt <[email protected]>
4028 Reviewed-by: Daniel Kiper <[email protected]>
4029
4030 2025-03-26 Glenn Washburn <[email protected]>
4031
4032 tests/grub_cmd_cryptomount: Default TMPDIR to /tmp
4033 This fixes behavior where grub_cmd_cryptomount temporary files, which are
4034 some times not cleaned up, are left in the / directory. Set TMPDIR if your
4035 system does not have /tmp or it can not be used for some reason.
4036
4037 Reported-by: Thomas Schmitt <[email protected]>
4038 Tested-by: Thomas Schmitt <[email protected]>
4039 Reviewed-by: Daniel Kiper <[email protected]>
4040
4041 2025-03-26 Glenn Washburn <[email protected]>
4042
4043 tests/grub_cmd_cryptomount: Cleanup the cryptsetup script unless debug is enabled
4044 This fixes an issue where the grub_cmd_cryptomount test leaves a file
4045 with an ambiguous name in the / directory when TMPDIR is not set.
4046
4047 Reported-by: Thomas Schmitt <[email protected]>
4048 Tested-by: Thomas Schmitt <[email protected]>
4049 Reviewed-by: Daniel Kiper <[email protected]>
4050
4051 2025-03-26 Glenn Washburn <[email protected]>
4052
4053 tests: Cleanup generated files on expected failure in grub_cmd_cryptomount
4054 grub-shell-luks-tester only cleans up generated files when the test it
4055 runs returns success. Sometimes tests are run that should fail. Add
4056 a --xfail argument to grub-shell-luks-tester and pass it from
4057 grub_cmd_cryptomount when invoking a test that is expected to fail.
4058
4059 Reported-by: Thomas Schmitt <[email protected]>
4060 Tested-by: Thomas Schmitt <[email protected]>
4061 Reviewed-by: Daniel Kiper <[email protected]>
4062
4063 2025-03-26 Glenn Washburn <[email protected]>
4064
4065 tests/util/grub-shell-luks-tester: Add missing line to create RET variable in cleanup
4066 Set the RET variable to the exit status of the script, as was assumed in
4067 the cleanup() function.
4068
4069 Reported-by: Thomas Schmitt <[email protected]>
4070 Tested-by: Thomas Schmitt <[email protected]>
4071 Reviewed-by: Daniel Kiper <[email protected]>
4072
4073 2025-03-26 Glenn Washburn <[email protected]>
4074
4075 tests/util/grub-shell-luks-tester: Find cryptodisk by UUID
4076 GRUB has the capability to search all the disks for a cryptodisk of a
4077 given UUID. Use this instead of hardcoding which disk is the cryptodisk,
4078 which can change when devices are added or removed, or potentially when
4079 QEMU is upgraded. This can not be done for the detached header tests
4080 because the header contains the UUID.
4081
4082 Also, capitalize comment lines for consistency.
4083
4084 Reviewed-by: Daniel Kiper <[email protected]>
4085
4086 2025-03-26 Glenn Washburn <[email protected]>
4087
4088 tests/util/grub-shell: Default qemuopts to envvar $GRUB_QEMU_OPTS
4089 Fix a regression where qemuopts was mistakenly defaulted to the empty
4090 string. This prevents the sending of arbitrary QEMU options to tests,
4091 which may be desirable for overriding the machine type. There was a
4092 concern that allowing the tester to accept arbitrary options would add
4093 headaches for another developer trying to diagnose why a test failed on
4094 the testers machine because he could not be sure if any additional
4095 options were passed to make the test fail. However, all the options are
4096 recorded in the run.sh generated script, so this concern is unwarranted.
4097
4098 Fixes: 6d729ced70 (tests/util/grub-shell: Add $GRUB_QEMU_OPTS to run.sh to easily see unofficial QEMU arguments)
4099
4100 Reviewed-by: Daniel Kiper <[email protected]>
4101
4102 2025-03-05 Patrick Plenefisch <[email protected]>
4103
4104 disk/lvm: Add informational messages in error cases of ignored features
4105 Reviewed-by: Daniel Kiper <[email protected]>
4106
4107 2025-03-05 Patrick Plenefisch <[email protected]>
4108
4109 disk/lvm: Add support for cachevol LV
4110 Mark cachevol LV's as ignored features, which is true only if they are
4111 configured as "writethrough". This patch does not let GRUB boot from
4112 "writeback" cache-enabled LV's.
4113
4114 Reviewed-by: Daniel Kiper <[email protected]>
4115
4116 2025-03-05 Patrick Plenefisch <[email protected]>
4117
4118 disk/lvm: Add support for integrity LV
4119 The LV matching must be done after processing the ignored feature
4120 indirections, as integrity volumes & caches may have several levels
4121 of indirection that the segments must be shifted through.
4122
4123 Reviewed-by: Daniel Kiper <[email protected]>
4124
4125 2025-03-05 Patrick Plenefisch <[email protected]>
4126
4127 lvm: Match all LVM segments before validation
4128 The PV matching must be completely finished before validating a volume,
4129 otherwise referenced RAID stripes may not have PV data applied yet.
4130
4131 This change is required for integrity & cachevol support.
4132
4133 Reviewed-by: Daniel Kiper <[email protected]>
4134
4135 2025-03-05 Patrick Plenefisch <[email protected]>
4136
4137 disk/lvm: Remove unused cache_pool
4138 The cache_pool is never read or used, remove it.
4139
4140 Reviewed-by: Daniel Kiper <[email protected]>
4141
4142 2025-03-05 Patrick Plenefisch <[email protected]>
4143
4144 disk/lvm: Make cache_lv more generic as ignored_feature_lv
4145 This patch isn't necessary by itself, but when combined with subsequent
4146 patches it enhances readability as ignored_features_lv is then used for
4147 multiple types of extra LV's, not just cache LV's.
4148
4149 Reviewed-by: Daniel Kiper <[email protected]>
4150
4151 2025-03-05 Glenn Washburn <[email protected]>
4152
4153 commands/ls: Add directory header for dir args
4154 Like the GNU ls, first print a line with the directory path before printing
4155 files in the directory, which will not have a directory component, but only
4156 if there is more than one argument.
4157
4158 Reviewed-by: Daniel Kiper <[email protected]>
4159
4160 2025-03-05 Glenn Washburn <[email protected]>
4161
4162 commands/ls: Print full paths for file args
4163 For arguments that are paths to files, print the full path of the file.
4164
4165 Reviewed-by: Daniel Kiper <[email protected]>
4166
4167 2025-03-05 Glenn Washburn <[email protected]>
4168
4169 commands/ls: Output path for single file arguments given with path
4170 Reviewed-by: Daniel Kiper <[email protected]>
4171
4172 2025-03-05 Glenn Washburn <[email protected]>
4173
4174 commands/ls: Show modification time for file paths
4175 The modification time for paths to files was not being printed because
4176 the grub_dirhook_info, which contains the mtime, was initialized to NULL.
4177 Instead of calling print_file() directly, use fs->fs_dir() to call
4178 print_file() with a properly filled in grub_dirhook_info. This has the
4179 added benefit of reducing code complexity.
4180
4181 Reviewed-by: Daniel Kiper <[email protected]>
4182
4183 2025-03-05 Glenn Washburn <[email protected]>
4184
4185 commands/ls: Merge print_files_long() and print_files() into print_file()
4186 Simplify the code by removing logic around which file printer to call.
4187
4188 Reviewed-by: Daniel Kiper <[email protected]>
4189
4190 2025-03-05 Glenn Washburn <[email protected]>
4191
4192 commands/ls: Return proper GRUB_ERR_* for functions returning type grub_err_t
4193 Also, remove unused code.
4194
4195 Reviewed-by: Daniel Kiper <[email protected]>
4196
4197 2025-03-05 Glenn Washburn <[email protected]>
4198
4199 commands/acpi: Use options enum to index command options
4200 Reviewed-by: Daniel Kiper <[email protected]>
4201
4202 2025-03-05 Andrew Hamilton <[email protected]>
4203
4204 docs: Capture additional commands restricted by lockdown
4205 Update documentation to capture that all memrw commands, the minicmd
4206 dump command, and raw memory dumping via hexdump are restricted when
4207 lockdown is enabled. This aligns to recent GRUB code updates.
4208
4209 Reviewed-by: Daniel Kiper <[email protected]>
4210
4211 2025-03-05 Andrew Hamilton <[email protected]>
4212
4213 docs: Document restricted filesystems in lockdown
4214 Document which filesystems are not allowed when lockdown
4215 is enabled to align to recent GRUB changes.
4216
4217 Reviewed-by: Daniel Kiper <[email protected]>
4218
4219 2025-03-05 Vladimir Serbinenko <[email protected]>
4220
4221 loader/i386/bsd: Fix type passed for the kernel
4222 FreeBSD loader always passes "elf kernel". We currently pass "elf64 kernel"
4223 when loading 64-bit kernel. The -CURRENT, HEAD, kernel accepts only
4224 "elf kernel". Older kernel accepts either.
4225
4226 Tested with FreeBSD and DragonFlyBSD.
4227
4228 Reference: https://cgit.freebsd.org/src/commit/?id=b72ae900d4348118829fe04abdc11b620930c30f
4229
4230 Reviewed-by: Daniel Kiper <[email protected]>
4231
4232 2025-03-05 Vladimir Serbinenko <[email protected]>
4233
4234 kern/partition: Unbreak support for nested partitions
4235 When using syntax "hd0,gtp3,dfly1" then ptr points to trailing part, ",dfly1".
4236 So, it's improper to consider it as an invalid partition.
4237
4238 Reviewed-by: Daniel Kiper <[email protected]>
4239
4240 2025-03-05 Andrew Hamilton <[email protected]>
4241
4242 lib/tss2/tss2_structs.h: Fix clang build - remove duplicate typedef
4243 grub-core/lib/tss2/tss2_structs.h contains a duplicate typedef as follows:
4244 typedef TPMS_SCHEME_HASH_t TPMS_SCHEME_KDF2_t;
4245
4246 This causes a build failure when compiling with clang. Remove the
4247 duplicate typedef which allows successfully building GRUB with clang.
4248
4249 Reviewed-by: Ross Philipson <[email protected]>
4250 Reviewed-by: Gary Lin <[email protected]>
4251 Reviewed-by: Daniel Kiper <[email protected]>
4252
4253 2025-03-05 Yuri Zaporozhets <[email protected]>
4254
4255 include/grub/mm.h: Remove duplicate inclusion of grub/err.h
4256 The header is included twice. Fix that.
4257
4258 Reviewed-by: Daniel Kiper <[email protected]>
4259
4260 2025-03-05 James Le Cuirot <[email protected]>
4261
4262 script/execute: Don't let trailing blank lines determine the return code
4263 The grub_script_execute_sourcecode() parses and executes code one line
4264 at a time, updating the return code each time because only the last line
4265 determines the final status. However, trailing new lines were also
4266 executed, masking any failure on the previous line. Fix this by only
4267 trying to execute the command when there is actually one present.
4268
4269 This has presumably never been noticed because this code is not used by
4270 regular functions, only in special cases like eval and menu entries. The
4271 latter generally don't return at all, having booted an OS. When failing
4272 to boot, upstream GRUB triggers the fallback mechanism regardless of the
4273 return code.
4274
4275 We noticed the problem while using Red Hat's patches, which change this
4276 behaviour to take account of the return code. In that case, a failure
4277 takes you back to the menu rather than triggering a fallback.
4278
4279 Reviewed-by: Daniel Kiper <[email protected]>
4280
4281 2025-03-05 Glenn Washburn <[email protected]>
4282
4283 gitignore: Ignore generated files from libtasn
4284 The commit 504058e8 (libtasn1: Compile into asn1 module) generates files
4285 into the grub-core/lib/libtasn1-grub directory and commit 99cda678
4286 (asn1_test: Test module for libtasn1) generates files into the
4287 grub-core/tests/asn1/tests directory. Ignore these directories as they
4288 are not under revision control.
4289
4290 Reviewed-by: Daniel Kiper <[email protected]>
4291
4292 2025-03-05 Pascal Hambourg <[email protected]>
4293
4294 util/grub.d/30_os-prober.in: Conditionally show or hide chain and efi menu entries
4295 On systems which support multiple boot platforms such as BIOS and
4296 EFI, it makes no sense to show menu entries which are not supported
4297 by the current boot platform. Menu entries generated from os-prober
4298 "chain" boot type use boot sector chainloading which is supported
4299 on PC BIOS platform only.
4300
4301 Show "chain" menu entries only if boot platform is PC BIOS.
4302 Show "efi" menu entries only if boot platform is EFI.
4303
4304 This is aimed to allow os-prober to report both EFI and PC BIOS
4305 boot loaders regardless of the current boot mode on x86 systems
4306 which support both EFI and legacy BIOS boot, in order to generate
4307 a config file which can be used with either BIOS or EFI boot.
4308
4309 Reviewed-by: Daniel Kiper <[email protected]>
4310
4311 2025-02-26 Pascal Hambourg <[email protected]>
4312
4313 util/grub.d/30_os-prober.in: Fix GRUB_OS_PROBER_SKIP_LIST for non-EFI
4314 GRUB documentation states:
4315
4316 GRUB_OS_PROBER_SKIP_LIST
4317 List of space-separated FS UUIDs of filesystems to be ignored from
4318 os-prober output. For efi chainloaders it’s <UUID>@<EFI FILE>
4319
4320 But the actual behaviour does not match this description.
4321
4322 GRUB_OS_PROBER_SKIP_LIST="<UUID>"
4323
4324 does nothing. In order to skip non-EFI bootloaders, you must set
4325
4326 GRUB_OS_PROBER_SKIP_LIST="<UUID>@<DEVICE>"
4327
4328 which is both absurd, <UUID> and <DEVICE> are redundant, and wrong,
4329 <DEVICE> such as /dev/sd* may not be persistent across boots.
4330
4331 Also, any non-word character is accepted as a separator, including "-"
4332 and "@" which may be present in UUIDs. This can cause false positives
4333 because of partial UUID match.
4334
4335 This patch fixes these flaws while retaining some backward compatibility
4336 with previous behaviour which may be expected by existing setups:
4337 - also accept <UUID>@/dev/* (with warning) for non-EFI bootloaders,
4338 - also accept comma and semicolon as separator.
4339
4340 Fixes: 55e706c9 (Add GRUB_OS_PROBER_SKIP_LIST to selectively skipping systems)
4341
4342 Reviewed-by: Daniel Kiper <[email protected]>
4343
4344 2025-02-26 Glenn Washburn <[email protected]>
4345
4346 docs: Do not reference non-existent --dumb option
4347 This appears to be a relic from GRUB legacy that used a --dumb option for
4348 its terminal command. The proper way to do this in GRUB2 is to set the
4349 terminal to "dumb" via the terminfo command.
4350
4351 Fixes: https://savannah.gnu.org/bugs/?66302
4352
4353 Reported-by: Jernej Jakob <[email protected]>
4354 Reviewed-by: Daniel Kiper <[email protected]>
4355
4356 2025-02-26 Glenn Washburn <[email protected]>
4357
4358 docs: Replace @lbracechar{} and @rbracechar{} with @{ and @}
4359 Support for @lbracechar{} and @rbracechar{} was added in GNU Texinfo 5.0
4360 but many older systems may have versions lower than this. Use @{ and @}
4361 to support a wider range of GNU Texinfo versions.
4362
4363 Reviewed-by: Daniel Kiper <[email protected]>
4364
4365 2025-02-26 Egor Ignatov <[email protected]>
4366
4367 fs/xfs: Fix grub_xfs_iterate_dir() return value in case of failure
4368 Commit ef7850c757 (fs/xfs: Fix issues found while fuzzing the XFS
4369 filesystem) introduced multiple boundary checks in grub_xfs_iterate_dir()
4370 but handled the error incorrectly returning error code instead of 0.
4371 Fix it. Also change the error message so that it doesn't match the
4372 message in grub_xfs_read_inode().
4373
4374 Fixes: ef7850c757 (fs/xfs: Fix issues found while fuzzing the XFS filesystem)
4375
4376 Reviewed-by: Daniel Kiper <[email protected]>
4377
4378 2025-02-26 Darrick J. Wong <[email protected]>
4379
4380 fs/xfs: Add new superblock features added in Linux 6.12/6.13
4381 The Linux port of XFS added a few new features in 2024. The existing
4382 GRUB driver doesn't attempt to read or write any of the new metadata,
4383 so, all three can be added to the incompat allowlist.
4384
4385 On the occasion align XFS_SB_FEAT_INCOMPAT_NREXT64 value.
4386
4387 Reviewed-by: Daniel Kiper <[email protected]>
4388
4389 2025-02-26 Michael Chang <[email protected]>
4390
4391 fs/ext2: Rework out-of-bounds read for inline and external extents
4392 Previously, the number of extent entries was not properly capped based
4393 on the actual available space. This could lead to insufficient reads for
4394 external extents since the computation was based solely on the inline
4395 extent layout.
4396
4397 In this patch, when processing the extent header we determine whether
4398 the header is stored inline, i.e. at inode->blocks.dir_blocks, or in an
4399 external extent block. We then clamp the number of entries accordingly
4400 (using max_inline_ext for inline extents and max_external_ext for
4401 external extent blocks).
4402
4403 This change ensures that only the valid number of extent entries is
4404 processed preventing out-of-bound reads and potential filesystem
4405 corruption.
4406
4407 Fixes: 7e2f750f0a (fs/ext2: Fix out-of-bounds read for inline extents)
4408
4409 Tested-by: Christian Hesse <[email protected]>
4410 Reviewed-by: Daniel Kiper <[email protected]>
4411
4412 2025-02-26 Leo Sandoval <[email protected]>
4413
4414 disk/ahci: Remove conditional operator for endtime
4415 The conditional makes no sense when the two possible expressions have
4416 the same value, so, remove it (perhaps the compiler does it for us but
4417 better to remove it). This change makes spinup argument unused. So, drop
4418 it as well.
4419
4420 Reviewed-by: Daniel Kiper <[email protected]>
4421
4422 2025-02-26 Benjamin Herrenschmidt <[email protected]>
4423
4424 term/ns8250-spcr: Return if redirection is disabled
4425 The Microsoft spec for SPCR says "The base address of the Serial Port
4426 register set described using the ACPI Generic Address Structure, or
4427 0 if console redirection is disabled". So, return early if redirection
4428 is disabled (base address = 0). If this check is not done we may get
4429 invalid ports on machines with redirection disabled and boot may hang
4430 when reading the grub.cfg file.
4431
4432 Reviewed-by: Leo Sandoval <[email protected]>
4433 Reviewed-by: Daniel Kiper <[email protected]>
4434
4435 2025-02-26 Lukas Fink <[email protected]>
4436
4437 commands/file: Fix NULL dereference in the knetbsd tests
4438 The pointer returned by grub_elf_file() is not checked to verify it is
4439 not NULL before use. A NULL pointer may be returned when the given file
4440 does not have a valid ELF header.
4441
4442 Fixes: https://savannah.gnu.org/bugs/?61960
4443
4444 Reviewed-by: Ross Philipson <[email protected]>
4445 Reviewed-by: Daniel Kiper <[email protected]>
4446
4447 2025-02-26 Heinrich Schuchardt <[email protected]>
4448
4449 gdb_helper: Typo hueristic
4450 %s/hueristic/heuristic/
4451
4452 Reviewed-by: Ross Philipson <[email protected]>
4453 Reviewed-by: Daniel Kiper <[email protected]>
4454
4455 2025-02-26 Ruihan Li <[email protected]>
4456
4457 kern/efi/mm: Reset grub_mm_add_region_fn after ExitBootServices() call
4458 The EFI Boot Services can be used after ExitBootServices() call because
4459 the GRUB code still may allocate memory.
4460
4461 An example call stack is:
4462
4463 grub_multiboot_boot
4464 grub_multiboot2_make_mbi
4465 grub_efi_finish_boot_services
4466 b->exit_boot_services
4467 normal_boot
4468 grub_relocator32_boot
4469 grub_relocator_alloc_chunk_align_safe
4470 grub_relocator_alloc_chunk_align
4471 grub_malloc
4472 grub_memalign
4473 grub_mm_add_region_fn
4474 [= grub_efi_mm_add_regions]
4475 grub_efi_allocate_any_pages
4476 grub_efi_allocate_pages_real
4477 b->allocate_pages
4478
4479 This can lead to confusing errors. After ExitBootServices() call
4480 b->allocate_pages may point to the NULL address resulting in something like:
4481
4482 !!!! X64 Exception Type - 01(#DB - Debug) CPU Apic ID - 00000000 !!!!
4483 RIP - 000000000000201F, CS - 0000000000000038, RFLAGS - 0000000000200002
4484 RAX - 000000007F9EE010, RCX - 0000000000000001, RDX - 0000000000000002
4485 RBX - 0000000000000006, RSP - 00000000001CFBEC, RBP - 0000000000000000
4486 RSI - 0000000000000000, RDI - 00000000FFFFFFFF
4487 R8 - 0000000000000006, R9 - 000000007FEDFFB8, R10 - 0000000000000000
4488 R11 - 0000000000000475, R12 - 0000000000000001, R13 - 0000000000000002
4489 R14 - 00000000FFFFFFFF, R15 - 000000007E432C08
4490 DS - 0000000000000030, ES - 0000000000000030, FS - 0000000000000030
4491 GS - 0000000000000030, SS - 0000000000000030
4492 CR0 - 0000000080010033, CR2 - 0000000000000000, CR3 - 000000007FC01000
4493 CR4 - 0000000000000668, CR8 - 0000000000000000
4494 DR0 - 0000000000000000, DR1 - 0000000000000000, DR2 - 0000000000000000
4495 DR3 - 0000000000000000, DR6 - 00000000FFFF0FF0, DR7 - 0000000000000400
4496 GDTR - 000000007F9DE000 0000000000000047, LDTR - 0000000000000000
4497 IDTR - 000000007F470018 0000000000000FFF, TR - 0000000000000000
4498 FXSAVE_STATE - 00000000001CF840
4499
4500 Ideally we would like to avoid all memory allocations after exiting EFI
4501 Boot Services altogether but that requires significant code changes. This
4502 patch adds a simple workaround that resets grub_mm_add_region_fn to NULL
4503 after ExitBootServices() call, so:
4504
4505 - Memory allocations have a better chance of succeeding because grub_memalign()
4506 will try to reclaim the disk cache if it sees a NULL in grub_mm_add_region_fn.
4507
4508 - At worst it will fail to allocate memory but it will explicitly tell users
4509 that it's out of memory, which is still much better than the current
4510 situation where it fails in a fairly random way and triggers a CPU fault.
4511
4512 Reviewed-by: Daniel Kiper <[email protected]>
4513
4514 2025-02-26 Duan Yayong <[email protected]>
4515
4516 i386/tsc: The GRUB menu gets stuck due to unserialized rdtsc
4517 This patch is used to fix GRUB menu gets stuck in server AC
4518 poweron/poweroff stress test of x86_64, which is reproduced with
4519 1/200 ratio. The root cause analysis as below:
4520
4521 Q: What's the code logic?
4522
4523 A: The grub_tsc_init() function will init tsc by setting grub_tsc_rate,
4524 which call stack is:
4525
4526 grub_tsc_init() -> grub_tsc_calibrate_from_pmtimer() -> grub_divmod64()
4527
4528 Among, grub_divmod64() function needs tsc_diff as the second parameter.
4529 In grub_pmtimer_wait_count_tsc(), we will call grub_get_tsc() function
4530 to get time stamp counter value to assign to start_tsc variable, and
4531 get into while (1) loop space to get end_tsc variable value with same
4532 function, after 3580 ticks, return "end_tsc - start_tsc". Actually,
4533 rdtsc instruction will be called in grub_get_tsc, but rdtsc instruction
4534 is not reliable (for the reason see the next question), which will cause
4535 tsc_diff to be a very big number larger than (1UL << 32) or a negative
4536 number, so that grub_tsc_rate will be zero. When run_menu() function is
4537 startup, and calls grub_tsc_get_time_ms() function to get current time
4538 to check if timeout time reach, at this time, grub_tsc_get_time_ms()
4539 function will return zero due to zero grub_tsc_rate variable, then GRUB
4540 menu gets stuck...
4541
4542 Q: What's the difference between rdtsc and rdtscp instructions in x86_64
4543 architecture? Here is more explanations from Intel® 64 and IA-32
4544 Architectures Software Developer’s Manual Volume 2B (December 2024):
4545 https://cdrdv2.intel.com/v1/dl/getContent/671241
4546
4547 A: In page 4-558 -> RDTSC—Read Time-Stamp Counter:
4548 The RDTSC instruction is not a serializing instruction. It does not
4549 necessarily wait until all previous instructions have been executed
4550 before reading the counter. Similarly, subsequent instructions may
4551 begin execution before the read operation is performed. The following
4552 items may guide software seeking to order executions of RDTSC:
4553 - If software requires RDTSC to be executed only after all previous
4554 instructions have executed and all previous loads are globally
4555 visible, it can execute LFENCE immediately before RDTSC.
4556 - If software requires RDTSC to be executed only after all previous
4557 instructions have executed and all previous loads and stores are
4558 globally visible, it can execute the sequence MFENCE;LFENCE
4559 immediately before RDTSC.
4560 - If software requires RDTSC to be executed prior to execution of any
4561 subsequent instruction (including any memory accesses), it can execute
4562 the sequence LFENCE immediately after RDTSC.
4563
4564 A: In page 4-560 -> RDTSCP—Read Time-Stamp Counter and Processor ID:
4565 The RDTSCP instruction is not a serializing instruction, but it does wait
4566 until all previous instructions have executed and all previous loads are
4567 globally visible. But it does not wait for previous stores to be globally
4568 visible, and subsequent instructions may begin execution before the read
4569 operation is performed. The following items may guide software seeking to
4570 order executions of RDTSCP:
4571 - If software requires RDTSCP to be executed only after all previous
4572 stores are globally visible, it can execute MFENCE immediately before
4573 RDTSCP.
4574 - If software requires RDTSCP to be executed prior to execution of any
4575 subsequent instruction (including any memory accesses), it can execute
4576 LFENCE immediately after RDTSCP.
4577
4578 Q: Why there is a cpuid serializing instruction before rdtsc instruction,
4579 but "grub_get_tsc" still cannot work as expect?
4580
4581 A: From Intel® 64 and IA-32 Architectures Software Developer's Manual
4582 Volume 2A: Instruction Set Reference, A-L (December 2024):
4583 https://cdrdv2.intel.com/v1/dl/getContent/671199
4584
4585 In page 3-222 -> CPUID—CPU Identification:
4586 CPUID can be executed at any privilege level to serialize instruction execution.
4587 Serializing instruction execution guarantees that any modifications to flags,
4588 registers, and memory for previous instructions are completed before
4589 the next instruction is fetched and executed.
4590
4591 So we only kept the instruction rdtsc and its previous instruction in order
4592 currently. But it is still out-of-order possibility between rdtsc instruction
4593 and its subsequent instruction.
4594
4595 Q: Why do we do this fix?
4596
4597 A: In the one hand, add cpuid instruction after rdtsc instruction to make sure
4598 rdtsc instruction to be executed prior to execution of any subsequent instruction,
4599 about serializing execution that all previous instructions have been executed
4600 before rdtsc, there is a cpuid usage in original code. In the other hand, using
4601 cpuid instruction rather than lfence can make sure a forward compatibility for
4602 previous HW.
4603
4604 Base this fix, we did 1500 cycles power on/off stress test, and did not reproduce
4605 this issue again.
4606
4607 Fixes: https://savannah.gnu.org/bugs/?66257
4608
4609 Reviewed-by: Daniel Kiper <[email protected]>
4610
4611 2025-02-26 Duan Yayong <[email protected]>
4612
4613 kern/i386/tsc_pmtimer: The GRUB menu gets stuck due to failed calibration
4614 The grub_divmod64() may return 0 but grub_tsc_calibrate_from_pmtimer()
4615 still returns 1 saying calibration succeeded. Of course it is not true.
4616 So, return 0 when grub_divmod64() returns 0. This way other calibration
4617 functions can be called subsequently.
4618
4619 Reviewed-by: Daniel Kiper <[email protected]>
4620
4621 2025-02-26 Sergii Dmytruk <[email protected]>
4622
4623 loader/i386/linux: Fix cleanup if kernel doesn't support 64-bit addressing
4624 Simply returning from grub_cmd_linux() doesn't free "file" resource nor
4625 calls grub_dl_ref(my_mod). Jump to "fail" label for proper cleanup like
4626 other error checks do.
4627
4628 Reviewed-by: Daniel Kiper <[email protected]>
4629
4630 2025-02-13 Alec Brown <[email protected]>
4631
4632 loader/i386/bsd: Use safe math to avoid underflow
4633 The operation kern_end - kern_start may underflow when we input it into
4634 grub_relocator_alloc_chunk_addr() call. To avoid this we can use safe
4635 math for this subtraction.
4636
4637 Fixes: CID 73845
4638
4639 Reviewed-by: Daniel Kiper <[email protected]>
4640
4641 2025-02-13 Alec Brown <[email protected]>
4642
4643 loader/i386/linux: Cast left shift to grub_uint32_t
4644 The Coverity complains that we might overflow into a negative value when
4645 setting linux_params.kernel_alignment to (1 << align). We can remedy
4646 this by casting it to grub_uint32_t.
4647
4648 Fixes: CID 473876
4649
4650 Reviewed-by: Daniel Kiper <[email protected]>
4651
4652 2025-02-13 Lidong Chen <[email protected]>
4653
4654 kern/misc: Add sanity check after grub_strtoul() call
4655 When the format string, fmt0, includes a positional argument
4656 grub_strtoul() or grub_strtoull() is called to extract the argument
4657 position. However, the returned argument position isn't fully validated.
4658 If the format is something like "%0$x" then these functions return
4659 0 which leads to an underflow in the calculation of the args index, curn.
4660 The fix is to add a check to ensure the extracted argument position is
4661 greater than 0 before computing curn. Additionally, replace one
4662 grub_strtoull() with grub_strtoul() and change curn type to make code
4663 more correct.
4664
4665 Fixes: CID 473841
4666
4667 Reviewed-by: Daniel Kiper <[email protected]>
4668
4669 2025-02-13 Lidong Chen <[email protected]>
4670
4671 kern/partition: Add sanity check after grub_strtoul() call
4672 The current code incorrectly assumes that both the input and the values
4673 returned by grub_strtoul() are always valid which can lead to potential
4674 errors. This fix ensures proper validation to prevent any unintended issues.
4675
4676 Fixes: CID 473843
4677
4678 Reviewed-by: Daniel Kiper <[email protected]>
4679
4680 2025-02-13 Alec Brown <[email protected]>
4681
4682 normal/menu: Use safe math to avoid an integer overflow
4683 The Coverity indicates that the variable current_entry might overflow.
4684 To prevent this use safe math when adding GRUB_MENU_PAGE_SIZE to current_entry.
4685
4686 On the occasion fix limiting condition which was broken.
4687
4688 Fixes: CID 473853
4689
4690 Reviewed-by: Daniel Kiper <[email protected]>
4691
4692 2025-02-13 Alec Brown <[email protected]>
4693
4694 bus/usb/ehci: Define GRUB_EHCI_TOGGLE as grub_uint32_t
4695 The Coverity indicates that GRUB_EHCI_TOGGLE is an int that contains
4696 a negative value and we are using it for the variable token which is
4697 grub_uint32_t. To remedy this we can cast the definition to grub_uint32_t.
4698
4699 Fixes: CID 473851
4700
4701 Reviewed-by: Daniel Kiper <[email protected]>
4702
4703 2025-02-13 Lidong Chen <[email protected]>
4704
4705 misc: Ensure consistent overflow error messages
4706 Update the overflow error messages to make them consistent
4707 across the GRUB code.
4708
4709 Reviewed-by: Daniel Kiper <[email protected]>
4710
4711 2025-02-13 Lidong Chen <[email protected]>
4712
4713 osdep/unix/getroot: Fix potential underflow
4714 The entry_len is initialized in grub_find_root_devices_from_mountinfo()
4715 to 0 before the while loop iterates through /proc/self/mountinfo. If the
4716 file is empty or contains only invalid entries entry_len remains
4717 0 causing entry_len - 1 in the subsequent for loop initialization
4718 to underflow. To prevent this add a check to ensure entry_len > 0 before
4719 entering the for loop.
4720
4721 Fixes: CID 473877
4722
4723 Reviewed-by: Daniel Kiper <[email protected]>
4724 Reviewed-by: Ross Philipson <[email protected]>
4725
4726 2025-02-13 Lidong Chen <[email protected]>
4727
4728 script/execute: Fix potential underflow and NULL dereference
4729 The result is initialized to 0 in grub_script_arglist_to_argv().
4730 If the for loop condition is not met both result.args and result.argc
4731 remain 0 causing result.argc - 1 to underflow and/or result.args NULL
4732 dereference. Fix the issues by adding relevant checks.
4733
4734 Fixes: CID 473880
4735
4736 Reviewed-by: Daniel Kiper <[email protected]>
4737
4738 2025-02-13 Alec Brown <[email protected]>
4739
4740 fs/sfs: Check if allocated memory is NULL
4741 When using grub_zalloc(), if we are out of memory, this function can fail.
4742 After allocating memory, we should check if grub_zalloc() returns NULL.
4743 If so, we should handle this error.
4744
4745 Fixes: CID 473856
4746
4747 Reviewed-by: Ross Philipson <[email protected]>
4748 Reviewed-by: Daniel Kiper <[email protected]>
4749
4750 2025-02-13 Alec Brown <[email protected]>
4751
4752 net: Check if returned pointer for allocated memory is NULL
4753 When using grub_malloc(), the function can fail if we are out of memory.
4754 After allocating memory we should check if this function returned NULL
4755 and handle this error if it did.
4756
4757 Reviewed-by: Daniel Kiper <[email protected]>
4758
4759 2025-02-13 Lidong Chen <[email protected]>
4760
4761 net: Prevent overflows when allocating memory for arrays
4762 Use grub_calloc() when allocating memory for arrays to ensure proper
4763 overflow checks are in place.
4764
4765 Reviewed-by: Daniel Kiper <[email protected]>
4766
4767 2025-02-13 Lidong Chen <[email protected]>
4768
4769 net: Use safe math macros to prevent overflows
4770 Replace direct arithmetic operations with macros from include/grub/safemath.h
4771 to prevent potential overflow issues when calculating the memory sizes.
4772
4773 Reviewed-by: Daniel Kiper <[email protected]>
4774
4775 2025-02-13 Lidong Chen <[email protected]>
4776
4777 fs/zfs: Add missing NULL check after grub_strdup() call
4778 Reviewed-by: Daniel Kiper <[email protected]>
4779
4780 2025-02-13 Lidong Chen <[email protected]>
4781
4782 fs/zfs: Check if returned pointer for allocated memory is NULL
4783 When using grub_malloc() or grub_zalloc(), these functions can fail if
4784 we are out of memory. After allocating memory we should check if these
4785 functions returned NULL and handle this error if they did.
4786
4787 Reviewed-by: Daniel Kiper <[email protected]>
4788
4789 2025-02-13 Lidong Chen <[email protected]>
4790
4791 fs/zfs: Prevent overflows when allocating memory for arrays
4792 Use grub_calloc() when allocating memory for arrays to ensure proper
4793 overflow checks are in place.
4794
4795 Reviewed-by: Daniel Kiper <[email protected]>
4796
4797 2025-02-13 Lidong Chen <[email protected]>
4798
4799 fs/zfs: Use safe math macros to prevent overflows
4800 Replace direct arithmetic operations with macros from include/grub/safemath.h
4801 to prevent potential overflow issues when calculating the memory sizes.
4802
4803 Reviewed-by: Daniel Kiper <[email protected]>
4804
4805 2025-02-13 Lidong Chen <[email protected]>
4806
4807 fs: Prevent overflows when assigning returned values from read_number()
4808 The direct assignment of the unsigned long long value returned by
4809 read_number() can potentially lead to an overflow on a 32-bit systems.
4810 The fix replaces the direct assignments with calls to grub_cast()
4811 which detects the overflows and safely assigns the values if no
4812 overflow is detected.
4813
4814 Reviewed-by: Daniel Kiper <[email protected]>
4815
4816 2025-02-13 Lidong Chen <[email protected]>
4817
4818 fs: Prevent overflows when allocating memory for arrays
4819 Use grub_calloc() when allocating memory for arrays to ensure proper
4820 overflow checks are in place.
4821
4822 The HFS+ and squash4 security vulnerabilities were reported by
4823 Jonathan Bar Or <[email protected]>.
4824
4825 Fixes: CVE-2025-0678
4826 Fixes: CVE-2025-1125
4827
4828 Reviewed-by: Daniel Kiper <[email protected]>
4829
4830 2025-02-13 Lidong Chen <[email protected]>
4831
4832 fs: Use safe math macros to prevent overflows
4833 Replace direct arithmetic operations with macros from include/grub/safemath.h
4834 to prevent potential overflow issues when calculating the memory sizes.
4835
4836 Reviewed-by: Daniel Kiper <[email protected]>
4837
4838 2025-02-13 Alec Brown <[email protected]>
4839
4840 disk/ieee1275/ofdisk: Call grub_ieee1275_close() when grub_malloc() fails
4841 In the dev_iterate() function a handle is opened but isn't closed when
4842 grub_malloc() returns NULL. We should fix this by closing it on error.
4843
4844 Reviewed-by: Daniel Kiper <[email protected]>
4845
4846 2025-02-13 Alec Brown <[email protected]>
4847
4848 disk: Check if returned pointer for allocated memory is NULL
4849 When using grub_malloc(), grub_zalloc() or grub_calloc(), these functions can
4850 fail if we are out of memory. After allocating memory we should check if these
4851 functions returned NULL and handle this error if they did.
4852
4853 On the occasion make a NULL check in ATA code more obvious.
4854
4855 Reviewed-by: Daniel Kiper <[email protected]>
4856
4857 2025-02-13 Alec Brown <[email protected]>
4858
4859 disk: Prevent overflows when allocating memory for arrays
4860 Use grub_calloc() when allocating memory for arrays to ensure proper
4861 overflow checks are in place.
4862
4863 Reviewed-by: Daniel Kiper <[email protected]>
4864
4865 2025-02-13 Alec Brown <[email protected]>
4866
4867 disk: Use safe math macros to prevent overflows
4868 Replace direct arithmetic operations with macros from include/grub/safemath.h
4869 to prevent potential overflow issues when calculating the memory sizes.
4870
4871 Reviewed-by: Daniel Kiper <[email protected]>
4872
4873 2025-02-13 Daniel Axtens <[email protected]>
4874
4875 fs: Disable many filesystems under lockdown
4876 The idea is to permit the following: btrfs, cpio, exfat, ext, f2fs, fat,
4877 hfsplus, iso9660, squash4, tar, xfs and zfs.
4878
4879 The JFS, ReiserFS, romfs, UDF and UFS security vulnerabilities were
4880 reported by Jonathan Bar Or <[email protected]>.
4881
4882 Fixes: CVE-2025-0677
4883 Fixes: CVE-2025-0684
4884 Fixes: CVE-2025-0685
4885 Fixes: CVE-2025-0686
4886 Fixes: CVE-2025-0689
4887
4888 Suggested-by: Daniel Axtens <[email protected]>
4889 Reviewed-by: Daniel Kiper <[email protected]>
4890
4891 2025-02-13 Daniel Axtens <[email protected]>
4892
4893 fs/bfs: Disable under lockdown
4894 The BFS is not fuzz-clean. Don't allow it to be loaded under lockdown.
4895 This will also disable the AFS.
4896
4897 Fixes: CVE-2024-45778
4898 Fixes: CVE-2024-45779
4899
4900 Reported-by: Nils Langius <[email protected]>
4901 Reviewed-by: Daniel Kiper <[email protected]>
4902
4903 2025-02-13 B Horn <[email protected]>
4904
4905 commands/hexdump: Disable memory reading in lockdown mode
4906 Reported-by: B Horn <[email protected]>
4907 Reviewed-by: Daniel Kiper <[email protected]>
4908
4909 2025-02-13 B Horn <[email protected]>
4910
4911 commands/memrw: Disable memory reading in lockdown mode
4912 With the rest of module being blocked in lockdown mode it does not make
4913 a lot of sense to leave memory reading enabled. This also goes in par
4914 with disabling the dump command.
4915
4916 Reported-by: B Horn <[email protected]>
4917 Reviewed-by: Daniel Kiper <[email protected]>
4918
4919 2025-02-13 B Horn <[email protected]>
4920
4921 commands/minicmd: Block the dump command in lockdown mode
4922 The dump enables a user to read memory which should not be possible
4923 in lockdown mode.
4924
4925 Fixes: CVE-2025-1118
4926
4927 Reported-by: B Horn <[email protected]>
4928 Reported-by: Jonathan Bar Or <[email protected]>
4929 Reviewed-by: Daniel Kiper <[email protected]>
4930
4931 2025-02-13 Lidong Chen <[email protected]>
4932
4933 commands/test: Stack overflow due to unlimited recursion depth
4934 The test_parse() evaluates test expression recursively. Due to lack of
4935 recursion depth check a specially crafted expression may cause a stack
4936 overflow. The recursion is only triggered by the parentheses usage and
4937 it can be unlimited. However, sensible expressions are unlikely to
4938 contain more than a few parentheses. So, this patch limits the recursion
4939 depth to 100, which should be sufficient.
4940
4941 Reported-by: Nils Langius <[email protected]>
4942 Reviewed-by: Daniel Kiper <[email protected]>
4943
4944 2025-02-13 Jonathan Bar Or <[email protected]>
4945
4946 commands/read: Fix an integer overflow when supplying more than 2^31 characters
4947 The grub_getline() function currently has a signed integer variable "i"
4948 that can be overflown when user supplies more than 2^31 characters.
4949 It results in a memory corruption of the allocated line buffer as well
4950 as supplying large negative values to grub_realloc().
4951
4952 Fixes: CVE-2025-0690
4953
4954 Reported-by: Jonathan Bar Or <[email protected]>
4955 Reviewed-by: Daniel Kiper <[email protected]>
4956
4957 2025-02-13 Lidong Chen <[email protected]>
4958
4959 gettext: Integer overflow leads to heap OOB write
4960 The size calculation of the translation buffer in
4961 grub_gettext_getstr_from_position() may overflow
4962 to 0 leading to heap OOB write. This patch fixes
4963 the issue by using grub_add() and checking for
4964 an overflow.
4965
4966 Fixes: CVE-2024-45777
4967
4968 Reported-by: Nils Langius <[email protected]>
4969 Reviewed-by: Daniel Kiper <[email protected]>
4970 Reviewed-by: Alec Brown <[email protected]>
4971
4972 2025-02-13 Lidong Chen <[email protected]>
4973
4974 gettext: Integer overflow leads to heap OOB write or read
4975 Calculation of ctx->grub_gettext_msg_list size in grub_mofile_open() may
4976 overflow leading to subsequent OOB write or read. This patch fixes the
4977 issue by replacing grub_zalloc() and explicit multiplication with
4978 grub_calloc() which does the same thing in safe manner.
4979
4980 Fixes: CVE-2024-45776
4981
4982 Reported-by: Nils Langius <[email protected]>
4983 Reviewed-by: Daniel Kiper <[email protected]>
4984 Reviewed-by: Alec Brown <[email protected]>
4985
4986 2025-02-13 B Horn <[email protected]>
4987
4988 gettext: Remove variables hooks on module unload
4989 The gettext module does not entirely cleanup after itself in
4990 its GRUB_MOD_FINI() leaving a few variables hooks in place.
4991 It is not possible to unload gettext module because normal
4992 module depends on it. Though fix the issues for completeness.
4993
4994 Fixes: CVE-2025-0622
4995
4996 Reported-by: B Horn <[email protected]>
4997 Reviewed-by: Daniel Kiper <[email protected]>
4998
4999 2025-02-13 B Horn <[email protected]>
5000
5001 normal: Remove variables hooks on module unload
5002 The normal module does not entirely cleanup after itself in
5003 its GRUB_MOD_FINI() leaving a few variables hooks in place.
5004 It is not possible to unload normal module now but fix the
5005 issues for completeness.
5006
5007 On the occasion replace 0s with NULLs for "pager" variable
5008 hooks unregister.
5009
5010 Fixes: CVE-2025-0622
5011
5012 Reported-by: B Horn <[email protected]>
5013 Reviewed-by: Daniel Kiper <[email protected]>
5014
5015 2025-02-13 B Horn <[email protected]>
5016
5017 commands/pgp: Unregister the "check_signatures" hooks on module unload
5018 If the hooks are not removed they can be called after the module has
5019 been unloaded leading to an use-after-free.
5020
5021 Fixes: CVE-2025-0622
5022
5023 Reported-by: B Horn <[email protected]>
5024 Reviewed-by: Daniel Kiper <[email protected]>
5025
5026 2025-02-13 B Horn <[email protected]>
5027
5028 commands/ls: Fix NULL dereference
5029 The grub_strrchr() may return NULL when the dirname do not contain "/".
5030 This can happen on broken filesystems.
5031
5032 Reported-by: B Horn <[email protected]>
5033 Reviewed-by: Daniel Kiper <[email protected]>
5034
5035 2025-02-13 Lidong Chen <[email protected]>
5036
5037 commands/extcmd: Missing check for failed allocation
5038 The grub_extcmd_dispatcher() calls grub_arg_list_alloc() to allocate
5039 a grub_arg_list struct but it does not verify the allocation was successful.
5040 In case of failed allocation the NULL state pointer can be accessed in
5041 parse_option() through grub_arg_parse() which may lead to a security issue.
5042
5043 Fixes: CVE-2024-45775
5044
5045 Reported-by: Nils Langius <[email protected]>
5046 Reviewed-by: Daniel Kiper <[email protected]>
5047 Reviewed-by: Alec Brown <[email protected]>
5048
5049 2025-02-13 B Horn <[email protected]>
5050
5051 kern/dl: Check for the SHF_INFO_LINK flag in grub_dl_relocate_symbols()
5052 The grub_dl_relocate_symbols() iterates through the sections in
5053 an ELF looking for relocation sections. According to the spec [1]
5054 the SHF_INFO_LINK flag should be set if the sh_info field is meant
5055 to be a section index.
5056
5057 [1] https://refspecs.linuxbase.org/elf/gabi4+/ch4.sheader.html
5058
5059 Reported-by: B Horn <[email protected]>
5060 Reviewed-by: Daniel Kiper <[email protected]>
5061
5062 2025-02-13 B Horn <[email protected]>
5063
5064 kern/dl: Use correct segment in grub_dl_set_mem_attrs()
5065 The previous code would never actually call grub_update_mem_attrs()
5066 as sh_info will always be zero for the sections that exist in memory.
5067
5068 Reported-by: B Horn <[email protected]>
5069 Reviewed-by: Daniel Kiper <[email protected]>
5070
5071 2025-02-13 B Horn <[email protected]>
5072
5073 kern/dl: Fix for an integer overflow in grub_dl_ref()
5074 It was possible to overflow the value of mod->ref_count, a signed
5075 integer, by repeatedly invoking insmod on an already loaded module.
5076 This led to a use-after-free. As once ref_count was overflowed it became
5077 possible to unload the module while there was still references to it.
5078
5079 This resolves the issue by using grub_add() to check if the ref_count
5080 will overflow and then stops further increments. Further changes were
5081 also made to grub_dl_unref() to check for the underflow condition and
5082 the reference count was changed to an unsigned 64-bit integer.
5083
5084 Reported-by: B Horn <[email protected]>
5085 Reviewed-by: Daniel Kiper <[email protected]>
5086
5087 2025-02-13 Daniel Axtens <[email protected]>
5088
5089 video/readers/jpeg: Do not permit duplicate SOF0 markers in JPEG
5090 Otherwise a subsequent header could change the height and width
5091 allowing future OOB writes.
5092
5093 Fixes: CVE-2024-45774
5094
5095 Reported-by: Nils Langius <[email protected]>
5096 Reviewed-by: Daniel Kiper <[email protected]>
5097
5098 2025-02-13 B Horn <[email protected]>
5099
5100 net/tftp: Fix stack buffer overflow in tftp_open()
5101 An overly long filename can be passed to tftp_open() which would cause
5102 grub_normalize_filename() to write out of bounds.
5103
5104 Fixed by adding an extra argument to grub_normalize_filename() for the
5105 space available, making it act closer to a strlcpy(). As several fixed
5106 strings are strcpy()'d after into the same buffer, their total length is
5107 checked to see if they exceed the remaining space in the buffer. If so,
5108 return an error.
5109
5110 On the occasion simplify code a bit by removing unneeded rrqlen zeroing.
5111
5112 Reported-by: B Horn <[email protected]>
5113 Reviewed-by: Daniel Kiper <[email protected]>
5114
5115 2025-02-13 B Horn <[email protected]>
5116
5117 net: Fix OOB write in grub_net_search_config_file()
5118 The function included a call to grub_strcpy() which copied data from an
5119 environment variable to a buffer allocated in grub_cmd_normal(). The
5120 grub_cmd_normal() didn't consider the length of the environment variable.
5121 So, the copy operation could exceed the allocation and lead to an OOB
5122 write. Fix the issue by replacing grub_strcpy() with grub_strlcpy() and
5123 pass the underlying buffers size to the grub_net_search_config_file().
5124
5125 Fixes: CVE-2025-0624
5126
5127 Reported-by: B Horn <[email protected]>
5128 Reviewed-by: Daniel Kiper <[email protected]>
5129
5130 2025-01-23 B Horn <[email protected]>
5131
5132 net: Remove variables hooks when interface is unregisted
5133 The grub_net_network_level_interface_unregister(), previously
5134 implemented in a header, did not remove the variables hooks that
5135 were registered in grub_net_network_level_interface_register().
5136 Fix this by implementing the same logic used to register the
5137 variables and move the function into the grub-core/net/net.c.
5138
5139 Reviewed-by: Daniel Kiper <[email protected]>
5140
5141 2025-01-23 B Horn <[email protected]>
5142
5143 net: Unregister net_default_ip and net_default_mac variables hooks on unload
5144 The net module is a dependency of normal. So, it shouldn't be possible
5145 to unload the net. Though unregister variables hooks as a precaution.
5146 It also gets in line with unregistering the other net module hooks.
5147
5148 Reviewed-by: Daniel Kiper <[email protected]>
5149
5150 2025-01-23 B Horn <[email protected]>
5151
5152 script/execute: Limit the recursion depth
5153 If unbounded recursion is allowed it becomes possible to collide the
5154 stack with the heap. As UEFI firmware often lacks guard pages this
5155 becomes an exploitable issue as it is possible in some cases to do
5156 a controlled overwrite of a section of this heap region with
5157 arbitrary data.
5158
5159 Reported-by: B Horn <[email protected]>
5160 Reviewed-by: Daniel Kiper <[email protected]>
5161
5162 2025-01-23 B Horn <[email protected]>
5163
5164 kern/partition: Limit recursion in part_iterate()
5165 The part_iterate() is used by grub_partition_iterate() as a callback in
5166 the partition iterate functions. However, part_iterate() may also call
5167 the partition iterate functions which may lead to recursion. Fix potential
5168 issue by limiting the recursion depth.
5169
5170 Reviewed-by: Daniel Kiper <[email protected]>
5171
5172 2025-01-23 B Horn <[email protected]>
5173
5174 kern/disk: Limit recursion depth
5175 The grub_disk_read() may trigger other disk reads, e.g. via loopbacks.
5176 This may lead to very deep recursion which can corrupt the heap. So, fix
5177 the issue by limiting reads depth.
5178
5179 Reported-by: B Horn <[email protected]>
5180 Reviewed-by: Daniel Kiper <[email protected]>
5181
5182 2025-01-23 B Horn <[email protected]>
5183
5184 disk/loopback: Reference tracking for the loopback
5185 It was possible to delete a loopback while there were still references
5186 to it. This led to an exploitable use-after-free.
5187
5188 Fixed by implementing a reference counting in the grub_loopback struct.
5189
5190 Reported-by: B Horn <[email protected]>
5191 Reviewed-by: Daniel Kiper <[email protected]>
5192
5193 2025-01-23 Michael Chang <[email protected]>
5194
5195 disk/cryptodisk: Require authentication after TPM unlock for CLI access
5196 The GRUB may use TPM to verify the integrity of boot components and the
5197 result can determine whether a previously sealed key can be released. If
5198 everything checks out, showing nothing has been tampered with, the key
5199 is released and GRUB unlocks the encrypted root partition for the next
5200 stage of booting.
5201
5202 However, the liberal Command Line Interface (CLI) can be misused by
5203 anyone in this case to access files in the encrypted partition one way
5204 or another. Despite efforts to keep the CLI secure by preventing utility
5205 command output from leaking file content, many techniques in the wild
5206 could still be used to exploit the CLI, enabling attacks or learning
5207 methods to attack. It's nearly impossible to account for all scenarios
5208 where a hack could be applied.
5209
5210 Therefore, to mitigate potential misuse of the CLI after the root device
5211 has been successfully unlocked via TPM, the user should be required to
5212 authenticate using the LUKS password. This added layer of security
5213 ensures that only authorized users can access the CLI reducing the risk
5214 of exploitation or unauthorized access to the encrypted partition.
5215
5216 Fixes: CVE-2024-49504
5217
5218 Reviewed-by: Daniel Kiper <[email protected]>
5219
5220 2025-01-23 B Horn <[email protected]>
5221
5222 kern/file: Implement filesystem reference counting
5223 The grub_file_open() and grub_file_close() should be the only places
5224 that allow a reference to a filesystem to stay open. So, add grub_dl_t
5225 to grub_fs_t and set this in the GRUB_MOD_INIT() for each filesystem to
5226 avoid issues when filesystems forget to do it themselves or do not track
5227 their own references, e.g. squash4.
5228
5229 The fs_label(), fs_uuid(), fs_mtime() and fs_read() should all ref and
5230 unref in the same function but it is essentially redundant in GRUB
5231 single threaded model.
5232
5233 Reviewed-by: Daniel Kiper <[email protected]>
5234
5235 2025-01-23 B Horn <[email protected]>
5236
5237 kern/file: Ensure file->data is set
5238 This is to avoid a generic issue were some filesystems would not set
5239 data and also not set a grub_errno. This meant it was possible for many
5240 filesystems to grub_dl_unref() themselves multiple times resulting in
5241 it being possible to unload the filesystems while there were still
5242 references to them, e.g., via a loopback.
5243
5244 Reported-by: B Horn <[email protected]>
5245 Reviewed-by: Daniel Kiper <[email protected]>
5246
5247 2025-01-23 B Horn <[email protected]>
5248
5249 fs/xfs: Ensuring failing to mount sets a grub_errno
5250 It was previously possible for grub_xfs_mount() to return NULL without
5251 setting grub_errno if the XFS version was invalid. This resulted in it
5252 being possible for grub_dl_unref() to be called twice allowing the XFS
5253 module to be unloaded while there were still references to it.
5254
5255 Fixing this problem in general by ensuring a grub_errno is set if the
5256 fail label is reached.
5257
5258 Reported-by: B Horn <[email protected]>
5259 Reviewed-by: Daniel Kiper <[email protected]>
5260
5261 2025-01-23 Michael Chang <[email protected]>
5262
5263 fs/xfs: Fix out-of-bounds read
5264 The number of records in the root key array read from disk was not being
5265 validated against the size of the root node. This could lead to an
5266 out-of-bounds read.
5267
5268 This patch adds a check to ensure that the number of records in the root
5269 key array does not exceed the expected size of a root node read from
5270 disk. If this check detects an out-of-bounds condition the operation is
5271 aborted to prevent random errors due to metadata corruption.
5272
5273 Reported-by: Daniel Axtens <[email protected]>
5274 Reviewed-by: Daniel Kiper <[email protected]>
5275
5276 2025-01-23 B Horn <[email protected]>
5277
5278 fs/ntfs: Implement attribute verification
5279 It was possible to read OOB when an attribute had a size that exceeded
5280 the allocated buffer. This resolves that by making sure all attributes
5281 that get read are fully in the allocated space by implementing
5282 a function to validate them.
5283
5284 Defining the offsets in include/grub/ntfs.h but they are only used in
5285 the validation function and not across the rest of the NTFS code.
5286
5287 Reviewed-by: Daniel Kiper <[email protected]>
5288
5289 2025-01-23 B Horn <[email protected]>
5290
5291 fs/ntfs: Use a helper function to access attributes
5292 Right now to access the next attribute the code reads the length of the
5293 current attribute and adds that to the current pointer. This is error
5294 prone as bounds checking needs to be performed all over the place. So,
5295 implement a helper and ensure its used across find_attr() and read_attr().
5296
5297 This commit does *not* implement full bounds checking. It is just the
5298 preparation work for this to be added into the helper.
5299
5300 Reviewed-by: Daniel Kiper <[email protected]>
5301
5302 2025-01-23 B Horn <[email protected]>
5303
5304 fs/ntfs: Track the end of the MFT attribute buffer
5305 The end of the attribute buffer should be stored alongside the rest of
5306 the attribute struct as right now it is not possible to implement bounds
5307 checking when accessing attributes sequentially.
5308
5309 This is done via:
5310 - updating init_attr() to set at->end and check is is not initially out of bounds,
5311 - implementing checks as init_attr() had its type change in its callers,
5312 - updating the value of at->end when needed.
5313
5314 Reviewed-by: Daniel Kiper <[email protected]>
5315
5316 2025-01-23 Michael Chang <[email protected]>
5317
5318 fs/ntfs: Fix out-of-bounds read
5319 When parsing NTFS file records the presence of the 0xFF marker indicates
5320 the end of the attribute list. This value signifies that there are no
5321 more attributes to process.
5322
5323 However, when the end marker is missing due to corrupted metadata the
5324 loop continues to read beyond the attribute list resulting in out-of-bounds
5325 reads and potentially entering an infinite loop.
5326
5327 This patch adds a check to provide a stop condition for the loop ensuring
5328 it stops at the end of the attribute list or at the end of the Master File
5329 Table. This guards against out-of-bounds reads and prevents infinite loops.
5330
5331 Reported-by: Daniel Axtens <[email protected]>
5332 Reviewed-by: Daniel Kiper <[email protected]>
5333
5334 2025-01-23 Michael Chang <[email protected]>
5335
5336 fs/ext2: Fix out-of-bounds read for inline extents
5337 When inline extents are used, i.e. the extent tree depth equals zero,
5338 a maximum of four entries can fit into the inode's data block. If the
5339 extent header states a number of entries greater than four the current
5340 ext2 implementation causes an out-of-bounds read. Fix this issue by
5341 capping the number of extents to four when reading inline extents.
5342
5343 Reported-by: Daniel Axtens <[email protected]>
5344 Reviewed-by: Daniel Kiper <[email protected]>
5345
5346 2025-01-23 Lidong Chen <[email protected]>
5347
5348 fs/jfs: Inconsistent signed/unsigned types usage in return values
5349 The getblk() returns a value of type grub_int64_t which is assigned to
5350 iagblk and inoblk, both of type grub_uint64_t, in grub_jfs_read_inode()
5351 via grub_jfs_blkno(). This patch fixes the type mismatch in the
5352 functions. Additionally, the getblk() will return 0 instead of -1 on
5353 failure cases. This change is safe because grub_errno is always set in
5354 getblk() to indicate errors and it is later checked in the callers.
5355
5356 Reviewed-by: Alec Brown <[email protected]>
5357 Reviewed-by: Ross Philipson <[email protected]>
5358 Reviewed-by: Daniel Kiper <[email protected]>
5359
5360 2025-01-23 Lidong Chen <[email protected]>
5361
5362 fs/jfs: Use full 40 bits offset and address for a data extent
5363 An extent's logical offset and address are represented as a 40-bit value
5364 split into two parts: the most significant 8 bits and the least
5365 significant 32 bits. Currently the JFS code uses only the least
5366 significant 32 bits value for offsets and addresses assuming the data
5367 size will never exceed the 32-bit range. This approach ignores the most
5368 significant 8 bits potentially leading to incorrect offsets and
5369 addresses for larger values. The patch fixes it by incorporating the
5370 most significant 8 bits into the calculation to get the full 40-bits
5371 value for offsets and addresses.
5372
5373 https://jfs.sourceforge.net/project/pub/jfslayout.pdf
5374
5375 "off1,off2 is a 40-bit field, containing the logical offset of the first
5376 block in the extent.
5377 ...
5378 addr1,addr2 is a 40-bit field, containing the address of the extent."
5379
5380 Reviewed-by: Alec Brown <[email protected]>
5381 Reviewed-by: Ross Philipson <[email protected]>
5382 Reviewed-by: Daniel Kiper <[email protected]>
5383
5384 2025-01-23 Lidong Chen <[email protected]>
5385
5386 fs/jfs: Fix OOB read caused by invalid dir slot index
5387 While fuzz testing JFS with ASAN enabled an OOB read was detected in
5388 grub_jfs_opendir(). The issue occurred due to an invalid directory slot
5389 index in the first entry of the sorted directory slot array in the inode
5390 directory header. The fix ensures the slot index is validated before
5391 accessing it. Given that an internal or a leaf node in a directory B+
5392 tree is a 4 KiB in size and each directory slot is always 32 bytes, the
5393 max number of slots in a node is 128. The validation ensures that the
5394 slot index doesn't exceed this limit.
5395
5396 [1] https://jfs.sourceforge.net/project/pub/jfslayout.pdf
5397
5398 JFS will allocate 4K of disk space for an internal node of the B+ tree.
5399 An internal node looks the same as a leaf node.
5400 - page 10
5401
5402 Fixed number of Directory Slots depending on the size of the node. These are
5403 the slots to be used for storing the directory slot array and the directory
5404 entries or router entries. A directory slot is always 32 bytes.
5405 ...
5406 A Directory Slot Array which is a sorted array of indices to the directory
5407 slots that are currently in use.
5408 ...
5409 An internal or a leaf node in the directory B+ tree is a 4K page.
5410 - page 25
5411
5412 Reviewed-by: Daniel Kiper <[email protected]>
5413 Reviewed-by: Alec Brown <[email protected]>
5414
5415 2025-01-23 Lidong Chen <[email protected]>
5416
5417 fs/jfs: Fix OOB read in jfs_getent()
5418 The JFS fuzzing revealed an OOB read in grub_jfs_getent(). The crash
5419 was caused by an invalid leaf nodes count, diro->dirpage->header.count,
5420 which was larger than the maximum number of leaf nodes allowed in an
5421 inode. This fix is to ensure that the leaf nodes count is validated in
5422 grub_jfs_opendir() before calling grub_jfs_getent().
5423
5424 On the occasion replace existing raw numbers with newly defined constant.
5425
5426 Reviewed-by: Daniel Kiper <[email protected]>
5427 Reviewed-by: Alec Brown <[email protected]>
5428
5429 2025-01-23 Michael Chang <[email protected]>
5430
5431 fs/iso9660: Fix invalid free
5432 The ctx->filename can point to either a string literal or a dynamically
5433 allocated string. The ctx->filename_alloc field is used to indicate the
5434 type of allocation.
5435
5436 An issue has been identified where ctx->filename is reassigned to
5437 a string literal in susp_iterate_dir() but ctx->filename_alloc is not
5438 correctly handled. This oversight causes a memory leak and an invalid
5439 free operation later.
5440
5441 The fix involves checking ctx->filename_alloc, freeing the allocated
5442 string if necessary and clearing ctx->filename_alloc for string literals.
5443
5444 Reported-by: Daniel Axtens <[email protected]>
5445 Reviewed-by: Daniel Kiper <[email protected]>
5446
5447 2025-01-23 B Horn <[email protected]>
5448
5449 fs/iso9660: Set a grub_errno if mount fails
5450 It was possible for a grub_errno to not be set if mount of an ISO 9660
5451 filesystem failed when set_rockridge() returned 0.
5452
5453 This isn't known to be exploitable as the other filesystems due to
5454 filesystem helper checking the requested file type. Though fixing
5455 as a precaution.
5456
5457 Reported-by: B Horn <[email protected]>
5458 Reviewed-by: Daniel Kiper <[email protected]>
5459
5460 2025-01-23 B Horn <[email protected]>
5461
5462 fs/hfsplus: Set a grub_errno if mount fails
5463 It was possible for mount to fail but not set grub_errno. This led to
5464 a possible double decrement of the module reference count if the NULL
5465 page was mapped.
5466
5467 Fixing in general as a similar bug was fixed in commit 61b13c187
5468 (fs/hfsplus: Set grub_errno to prevent NULL pointer access) and there
5469 are likely more variants around.
5470
5471 Fixes: CVE-2024-45783
5472
5473 Reported-by: B Horn <[email protected]>
5474 Reviewed-by: Daniel Kiper <[email protected]>
5475
5476 2025-01-23 B Horn <[email protected]>
5477
5478 fs/f2fs: Set a grub_errno if mount fails
5479 It was previously possible for grub_errno to not be set when
5480 grub_f2fs_mount() failed if nat_bitmap_ptr() returned NULL.
5481
5482 This issue is solved by ensuring a grub_errno is set in the fail case.
5483
5484 Reported-by: B Horn <[email protected]>
5485 Reviewed-by: Daniel Kiper <[email protected]>
5486
5487 2025-01-23 Lidong Chen <[email protected]>
5488
5489 fs/tar: Integer overflow leads to heap OOB write
5490 Both namesize and linksize are derived from hd.size, a 12-digit octal
5491 number parsed by read_number(). Later direct arithmetic calculation like
5492 "namesize + 1" and "linksize + 1" may exceed the maximum value of
5493 grub_size_t leading to heap OOB write. This patch fixes the issue by
5494 using grub_add() and checking for an overflow.
5495
5496 Fixes: CVE-2024-45780
5497
5498 Reported-by: Nils Langius <[email protected]>
5499 Reviewed-by: Daniel Kiper <[email protected]>
5500 Reviewed-by: Alec Brown <[email protected]>
5501
5502 2025-01-16 B Horn <[email protected]>
5503
5504 fs/tar: Initialize name in grub_cpio_find_file()
5505 It was possible to iterate through grub_cpio_find_file() without
5506 allocating name and not setting mode to GRUB_ARCHELP_ATTR_END, which
5507 would cause the uninitialized value for name to be used as an argument
5508 for canonicalize() in grub_archelp_dir().
5509
5510 Reported-by: B Horn <[email protected]>
5511 Reviewed-by: Daniel Kiper <[email protected]>
5512
5513 2025-01-16 B Horn <[email protected]>
5514
5515 fs/hfs: Fix stack OOB write with grub_strcpy()
5516 Replaced with grub_strlcpy().
5517
5518 Fixes: CVE-2024-45782
5519 Fixes: CVE-2024-56737
5520 Fixes: https://savannah.gnu.org/bugs/?66599
5521
5522 Reported-by: B Horn <[email protected]>
5523 Reviewed-by: Daniel Kiper <[email protected]>
5524
5525 2024-12-02 B Horn <[email protected]>
5526
5527 fs/ufs: Fix a heap OOB write
5528 grub_strcpy() was used to copy a symlink name from the filesystem
5529 image to a heap allocated buffer. This led to a OOB write to adjacent
5530 heap allocations. Fix by using grub_strlcpy().
5531
5532 Fixes: CVE-2024-45781
5533
5534 Reported-by: B Horn <[email protected]>
5535 Reviewed-by: Daniel Kiper <[email protected]>
5536
5537 2024-12-02 B Horn <[email protected]>
5538
5539 misc: Implement grub_strlcpy()
5540 grub_strlcpy() acts the same way as strlcpy() does on most *NIX,
5541 returning the length of src and ensuring dest is always NUL
5542 terminated except when size is 0.
5543
5544 Reviewed-by: Daniel Kiper <[email protected]>
5545
5546 2024-11-28 Stefan Berger <[email protected]>
5547
5548 tpm2_key_protector: Enable build for powerpc_ieee1275
5549 Reviewed-by: Daniel Kiper <[email protected]>
5550
5551 2024-11-28 Stefan Berger <[email protected]>
5552
5553 ieee1275/tcg2: Add TCG2 driver for ieee1275 PowerPC firmware
5554 Follow recent extensions of EFI support providing a TCG2 driver with
5555 a public API for getting the maximum TPM command size and passing a TPM
5556 command through to the TPM 2. Implement this functionality using ieee1275
5557 PowerPC firmware API calls. Move tcg2.c into the TCG2 driver.
5558
5559 Reviewed-by: Daniel Kiper <[email protected]>
5560
5561 2024-11-28 Stefan Berger <[email protected]>
5562
5563 ieee1275/tcg2: Refactor grub_ieee1275_tpm_init()
5564 Move tpm_get_tpm_version() into grub_ieee1275_tpm_init() and invalidate
5565 grub_ieee1275_tpm_ihandle in case no TPM 2 could be detected. Try the
5566 initialization only once so that grub_tpm_present() will always return
5567 the same result. Use the grub_ieee1275_tpm_ihandle as indicator for an
5568 available TPM instead of grub_ieee1275_tpm_version, which can now be
5569 removed.
5570
5571 Reviewed-by: Daniel Kiper <[email protected]>
5572
5573 2024-11-28 Stefan Berger <[email protected]>
5574
5575 ieee1275/ibmvpm: Move TPM initialization functions to own file
5576 Move common initialization functions from the ibmvtpm driver module into
5577 tcg2.c that will be moved into the new TCG2 driver in a subsequent patch.
5578 Make the functions available to the ibmvtpm driver as public functions
5579 and variables.
5580
5581 Reviewed-by: Daniel Kiper <[email protected]>
5582
5583 2024-11-28 Stefan Berger <[email protected]>
5584
5585 ieee1275: Consolidate repeated definitions of IEEE1275_IHANDLE_INVALID
5586 Consolidate repeated definitions of IEEE1275_IHANDLE_INVALID that are cast
5587 to the type grub_ieee1275_ihandle_t. On the occasion add "GRUB_" prefix to
5588 the constant name.
5589
5590 Reviewed-by: Daniel Kiper <[email protected]>
5591
5592 2024-11-28 Stefan Berger <[email protected]>
5593
5594 term/ieee1275/serial: Cast 0 to proper type
5595 Cast 0 to proper type grub_ieee1275_ihandle_t. This type is
5596 used for struct grub_serial_port's handle that assigns or
5597 compares with IEEE1275_IHANDLE_INVALID.
5598
5599 Reviewed-by: Daniel Kiper <[email protected]>
5600
5601 2024-11-28 Stefan Berger <[email protected]>
5602
5603 tss2: Adjust bit fields for big endian targets
5604 The TPM bit fields need to be in reverse order for big endian targets,
5605 such as ieee1275 PowerPC platforms that run GRUB in big endian mode.
5606
5607 Reviewed-by: Gary Lin <[email protected]>
5608 Reviewed-by: Daniel Kiper <[email protected]>
5609
5610 2024-11-28 Gary Lin <[email protected]>
5611
5612 docs: Document TPM2 key protector
5613 Update the user manual to address TPM2 key protector including the two
5614 related commands, tpm2_key_protector_init and tpm2_key_protector_clear,
5615 and the user-space utility: grub-protect.
5616
5617 Reviewed-by: Daniel Kiper <[email protected]>
5618 Reviewed-by: Stefan Berger <[email protected]>
5619 Tested-by: Stefan Berger <[email protected]>
5620
5621 2024-11-28 Gary Lin <[email protected]>
5622
5623 tests: Add tpm2_key_protector_test
5624 For the tpm2_key_protector module, the TCG2 command submission function
5625 is the only difference between a QEMU instance and grub-emu. To test
5626 TPM2 key unsealing with a QEMU instance, it requires an extra OS image
5627 to invoke grub-protect to seal the LUKS key, rather than a simple
5628 grub-shell rescue CD image. On the other hand, grub-emu can share the
5629 emulated TPM2 device with the host, so that we can seal the LUKS key on
5630 host and test key unsealing with grub-emu.
5631
5632 This test script firstly creates a simple LUKS image to be loaded as a
5633 loopback device in grub-emu. Then an emulated TPM2 device is created by
5634 "swtpm chardev" and PCR 0 and 1 are extended.
5635
5636 There are several test cases in the script to test various settings. Each
5637 test case uses grub-protect or tpm2-tools to seal the LUKS password
5638 with PCR 0 and PCR 1. Then grub-emu is launched to load the LUKS image,
5639 try to mount the image with tpm2_key_protector_init and cryptomount, and
5640 verify the result.
5641
5642 Based on the idea from Michael Chang.
5643
5644 Cc: Michael Chang <[email protected]>
5645 Cc: Stefan Berger <[email protected]>
5646 Cc: Glenn Washburn <[email protected]>
5647 Reviewed-by: Daniel Kiper <[email protected]>
5648 Reviewed-by: Stefan Berger <[email protected]>
5649 Tested-by: Stefan Berger <[email protected]>
5650
5651 2024-11-28 Gary Lin <[email protected]>
5652
5653 tpm2_key_protector: Add grub-emu support
5654 As a preparation to test tpm2_key_protector with grub-emu, the new
5655 option, --tpm-device, is introduced to specify the TPM device for
5656 grub-emu so that grub-emu can access an emulated TPM device from
5657 the host.
5658
5659 Since grub-emu can directly access the device on host, it's easy to
5660 implement the essential TCG2 command submission function with the
5661 read/write functions and enable tpm2_key_protector module for grub-emu,
5662 so that we can further test TPM2 key unsealing with grub-emu.
5663
5664 Reviewed-by: Stefan Berger <[email protected]>
5665 Reviewed-by: Daniel Kiper <[email protected]>
5666 Tested-by: Stefan Berger <[email protected]>
5667
5668 2024-11-28 Gary Lin <[email protected]>
5669
5670 diskfilter: Look up cryptodisk devices first
5671 When using disk auto-unlocking with TPM 2.0, the typical grub.cfg may
5672 look like this:
5673
5674 tpm2_key_protector_init --tpm2key=(hd0,gpt1)/boot/grub/sealed.tpm
5675 cryptomount -u <PART-UUID> -P tpm2
5676 search --fs-uuid --set=root <FS-UUID>
5677
5678 Since the disk search order is based on the order of module loading, the
5679 attacker could insert a malicious disk with the same FS-UUID root to
5680 trick GRUB to boot into the malicious root and further dump memory to
5681 steal the unsealed key.
5682
5683 Do defend against such an attack, we can specify the hint provided by
5684 "grub-probe" to search the encrypted partition first:
5685
5686 search --fs-uuid --set=root --hint='cryptouuid/<PART-UUID>' <FS-UUID>
5687
5688 However, for LVM on an encrypted partition, the search hint provided by
5689 "grub-probe" is:
5690
5691 --hint='lvmid/<VG-UUID>/<LV-UUID>'
5692
5693 It doesn't guarantee to look up the logical volume from the encrypted
5694 partition, so the attacker may have the chance to fool GRUB to boot
5695 into the malicious disk.
5696
5697 To minimize the attack surface, this commit tweaks the disk device search
5698 in diskfilter to look up cryptodisk devices first and then others, so
5699 that the auto-unlocked disk will be found first, not the attacker's disk.
5700
5701 Cc: Fabian Vogt <[email protected]>
5702 Reviewed-by: Stefan Berger <[email protected]>
5703 Reviewed-by: Daniel Kiper <[email protected]>
5704 Tested-by: Stefan Berger <[email protected]>
5705
5706 2024-11-28 Gary Lin <[email protected]>
5707
5708 cryptodisk: Wipe out the cached keys from protectors
5709 An attacker may insert a malicious disk with the same crypto UUID and
5710 trick GRUB to mount the fake root. Even though the key from the key
5711 protector fails to unlock the fake root, it's not wiped out cleanly so
5712 the attacker could dump the memory to retrieve the secret key. To defend
5713 such attack, wipe out the cached key when we don't need it.
5714
5715 Cc: Fabian Vogt <[email protected]>
5716 Reviewed-by: Stefan Berger <[email protected]>
5717 Reviewed-by: Daniel Kiper <[email protected]>
5718 Tested-by: Stefan Berger <[email protected]>
5719
5720 2024-11-28 Patrick Colp <[email protected]>
5721
5722 cryptodisk: Fallback to passphrase
5723 If a protector is specified, but it fails to unlock the disk, fall back
5724 to asking for the passphrase.
5725
5726 Before requesting the passphrase, the error from the key protector(s)
5727 has to be cleared, or the later code, e.g., LUKS code, may stop as
5728 grub_errno is set. This commit prints error from the key protector(s)
5729 and sets grub_errno to GRUB_ERR_NONE to have a fresh start.
5730
5731 Reviewed-by: Stefan Berger <[email protected]>
5732 Reviewed-by: Daniel Kiper <[email protected]>
5733 Tested-by: Stefan Berger <[email protected]>
5734
5735 2024-11-28 Patrick Colp <[email protected]>
5736
5737 tpm2_key_protector: Implement NV index
5738 Currently with the TPM2 protector, only SRK mode is supported and
5739 NV index support is just a stub. Implement the NV index option.
5740
5741 Note: This only extends support on the unseal path. grub-protect
5742 has not been updated. tpm2-tools can be used to insert a key into
5743 the NV index.
5744
5745 An example of inserting a key using tpm2-tools:
5746
5747 # Get random key.
5748 tpm2_getrandom 32 > key.dat
5749
5750 # Create primary object.
5751 tpm2_createprimary -C o -g sha256 -G ecc -c primary.ctx
5752
5753 # Create policy object. `pcrs.dat` contains the PCR values to seal against.
5754 tpm2_startauthsession -S session.dat
5755 tpm2_policypcr -S session.dat -l sha256:7,11 -f pcrs.dat -L policy.dat
5756 tpm2_flushcontext session.dat
5757
5758 # Seal key into TPM.
5759 cat key.dat | tpm2_create -C primary.ctx -u key.pub -r key.priv -L policy.dat -i-
5760 tpm2_load -C primary.ctx -u key.pub -r key.priv -n sealing.name -c sealing.ctx
5761 tpm2_evictcontrol -C o -c sealing.ctx 0x81000000
5762
5763 Then to unseal the key in GRUB, add this to grub.cfg:
5764
5765 tpm2_key_protector_init --mode=nv --nvindex=0x81000000 --pcrs=7,11
5766 cryptomount -u <UUID> --protector tpm2
5767
5768 Reviewed-by: Stefan Berger <[email protected]>
5769 Reviewed-by: Daniel Kiper <[email protected]>
5770 Tested-by: Stefan Berger <[email protected]>
5771
5772 2024-11-28 Gary Lin <[email protected]>
5773
5774 tpm2_key_protector: Support authorized policy
5775 This commit handles the TPM2_PolicyAuthorize command from the key file
5776 in TPM 2.0 Key File format.
5777
5778 TPM2_PolicyAuthorize is the essential command to support authorized
5779 policy which allows the users to sign TPM policies with their own keys.
5780 Per TPM 2.0 Key File [1], CommandPolicy for TPM2_PolicyAuthorize
5781 comprises "TPM2B_PUBLIC pubkey", "TPM2B_DIGEST policy_ref", and
5782 "TPMT_SIGNATURE signature". To verify the signature, the current policy
5783 digest is hashed with the hash algorithm written in "signature", and then
5784 "signature" is verified with the hashed policy digest and "pubkey". Once
5785 TPM accepts "signature", TPM2_PolicyAuthorize is invoked to authorize the
5786 signed policy.
5787
5788 To create the key file with authorized policy, here are the pcr-oracle [2]
5789 commands:
5790
5791 # Generate the RSA key and create the authorized policy file
5792 $ pcr-oracle \
5793 --rsa-generate-key \
5794 --private-key policy-key.pem \
5795 --auth authorized.policy \
5796 create-authorized-policy 0,2,4,7,9
5797
5798 # Seal the secret with the authorized policy
5799 $ pcr-oracle \
5800 --key-format tpm2.0 \
5801 --auth authorized.policy \
5802 --input disk-secret.txt \
5803 --output sealed.key \
5804 seal-secret
5805
5806 # Sign the predicted PCR policy
5807 $ pcr-oracle \
5808 --key-format tpm2.0 \
5809 --private-key policy-key.pem \
5810 --from eventlog \
5811 --stop-event "grub-file=grub.cfg" \
5812 --after \
5813 --input sealed.key \
5814 --output /boot/efi/efi/grub/sealed.tpm \
5815 sign 0,2,4,7,9
5816
5817 Then specify the key file and the key protector to grub.cfg in the EFI
5818 system partition:
5819
5820 tpm2_key_protector_init -a RSA --tpm2key=(hd0,gpt1)/efi/grub/sealed.tpm
5821 cryptomount -u <PART_UUID> -P tpm2
5822
5823 For any change in the boot components, just run the "sign" command again
5824 to update the signature in sealed.tpm, and TPM can unseal the key file
5825 with the updated PCR policy.
5826
5827 [1] https://www.hansenpartnership.com/draft-bottomley-tpm2-keys.html
5828 [2] https://github.com/okirch/pcr-oracle
5829
5830 Reviewed-by: Stefan Berger <[email protected]>
5831 Reviewed-by: Daniel Kiper <[email protected]>
5832 Tested-by: Stefan Berger <[email protected]>
5833
5834 2024-11-28 Hernan Gatta <[email protected]>
5835
5836 util/grub-protect: Add new tool
5837 To utilize the key protectors framework, there must be a way to protect
5838 full-disk encryption keys in the first place. The grub-protect tool
5839 includes support for the TPM2 key protector but other protectors that
5840 require setup ahead of time can be supported in the future.
5841
5842 For the TPM2 key protector, the intended flow is for a user to have
5843 a LUKS 1 or LUKS 2-protected fully-encrypted disk. The user then creates
5844 a new LUKS key file, say by reading /dev/urandom into a file, and creates
5845 a new LUKS key slot for this key. Then, the user invokes the grub-protect
5846 tool to seal this key file to a set of PCRs using the system's TPM 2.0.
5847 The resulting sealed key file is stored in an unencrypted partition such
5848 as the EFI System Partition (ESP) so that GRUB may read it. The user also
5849 has to ensure the cryptomount command is included in GRUB's boot script
5850 and that it carries the requisite key protector (-P) parameter.
5851
5852 Sample usage:
5853
5854 $ dd if=/dev/urandom of=luks-key bs=1 count=32
5855 $ sudo cryptsetup luksAddKey /dev/sdb1 luks-key --pbkdf=pbkdf2 --hash=sha512
5856
5857 To seal the key with TPM 2.0 Key File (recommended):
5858
5859 $ sudo grub-protect --action=add \
5860 --protector=tpm2 \
5861 --tpm2-pcrs=0,2,4,7,9 \
5862 --tpm2key \
5863 --tpm2-keyfile=luks-key \
5864 --tpm2-outfile=/boot/efi/efi/grub/sealed.tpm
5865
5866 Or, to seal the key with the raw sealed key:
5867
5868 $ sudo grub-protect --action=add \
5869 --protector=tpm2 \
5870 --tpm2-pcrs=0,2,4,7,9 \
5871 --tpm2-keyfile=luks-key \
5872 --tpm2-outfile=/boot/efi/efi/grub/sealed.key
5873
5874 Then, in the boot script, for TPM 2.0 Key File:
5875
5876 tpm2_key_protector_init --tpm2key=(hd0,gpt1)/efi/grub/sealed.tpm
5877 cryptomount -u <SDB1_UUID> -P tpm2
5878
5879 Or, for the raw sealed key:
5880
5881 tpm2_key_protector_init --keyfile=(hd0,gpt1)/efi/grub/sealed.key --pcrs=0,2,4,7,9
5882 cryptomount -u <SDB1_UUID> -P tpm2
5883
5884 The benefit of using TPM 2.0 Key File is that the PCR set is already
5885 written in the key file, so there is no need to specify PCRs when
5886 invoking tpm2_key_protector_init.
5887
5888 Reviewed-by: Stefan Berger <[email protected]>
5889 Reviewed-by: Daniel Kiper <[email protected]>
5890 Tested-by: Stefan Berger <[email protected]>
5891
5892 2024-11-28 Hernan Gatta <[email protected]>
5893
5894 cryptodisk: Support key protectors
5895 Add a new parameter to cryptomount to support the key protectors framework: -P.
5896 The parameter is used to automatically retrieve a key from specified key
5897 protectors. The parameter may be repeated to specify any number of key
5898 protectors. These are tried in order until one provides a usable key for any
5899 given disk.
5900
5901 Reviewed-by: Glenn Washburn <[email protected]>
5902 Reviewed-by: Stefan Berger <[email protected]>
5903 Reviewed-by: Daniel Kiper <[email protected]>
5904 Tested-by: Stefan Berger <[email protected]>
5905
5906 2024-11-28 Hernan Gatta <[email protected]>
5907
5908 key_protector: Add TPM2 Key Protector
5909 The TPM2 key protector is a module that enables the automatic retrieval
5910 of a fully-encrypted disk's unlocking key from a TPM 2.0.
5911
5912 The theory of operation is such that the module accepts various
5913 arguments, most of which are optional and therefore possess reasonable
5914 defaults. One of these arguments is the keyfile/tpm2key parameter, which
5915 is mandatory. There are two supported key formats:
5916
5917 1. Raw Sealed Key (--keyfile)
5918 When sealing a key with TPM2_Create, the public portion of the sealed
5919 key is stored in TPM2B_PUBLIC, and the private portion is in
5920 TPM2B_PRIVATE. The raw sealed key glues the fully marshalled
5921 TPM2B_PUBLIC and TPM2B_PRIVATE into one file.
5922
5923 2. TPM 2.0 Key (--tpm2key)
5924 The following is the ASN.1 definition of TPM 2.0 Key File:
5925
5926 TPMPolicy ::= SEQUENCE {
5927 CommandCode [0] EXPLICIT INTEGER
5928 CommandPolicy [1] EXPLICIT OCTET STRING
5929 }
5930
5931 TPMAuthPolicy ::= SEQUENCE {
5932 Name [0] EXPLICIT UTF8STRING OPTIONAL
5933 Policy [1] EXPLICIT SEQUENCE OF TPMPolicy
5934 }
5935
5936 TPMKey ::= SEQUENCE {
5937 type OBJECT IDENTIFIER
5938 emptyAuth [0] EXPLICIT BOOLEAN OPTIONAL
5939 policy [1] EXPLICIT SEQUENCE OF TPMPolicy OPTIONAL
5940 secret [2] EXPLICIT OCTET STRING OPTIONAL
5941 authPolicy [3] EXPLICIT SEQUENCE OF TPMAuthPolicy OPTIONAL
5942 description [4] EXPLICIT UTF8String OPTIONAL,
5943 rsaParent [5] EXPLICIT BOOLEAN OPTIONAL,
5944 parent INTEGER
5945 pubkey OCTET STRING
5946 privkey OCTET STRING
5947 }
5948
5949 The TPM2 key protector only expects a "sealed" key in DER encoding,
5950 so "type" is always 2.23.133.10.1.5, "emptyAuth" is "TRUE", and
5951 "secret" is empty. "policy" and "authPolicy" are the possible policy
5952 command sequences to construct the policy digest to unseal the key.
5953 Similar to the raw sealed key, the public portion (TPM2B_PUBLIC) of
5954 the sealed key is stored in "pubkey", and the private portion
5955 (TPM2B_PRIVATE) is in "privkey".
5956
5957 For more details: https://www.hansenpartnership.com/draft-bottomley-tpm2-keys.html
5958
5959 This sealed key file is created via the grub-protect tool. The tool
5960 utilizes the TPM's sealing functionality to seal (i.e., encrypt) an
5961 unlocking key using a Storage Root Key (SRK) to the values of various
5962 Platform Configuration Registers (PCRs). These PCRs reflect the state
5963 of the system as it boots. If the values are as expected, the system
5964 may be considered trustworthy, at which point the TPM allows for a
5965 caller to utilize the private component of the SRK to unseal (i.e.,
5966 decrypt) the sealed key file. The caller, in this case, is this key
5967 protector.
5968
5969 The TPM2 key protector registers two commands:
5970
5971 - tpm2_key_protector_init: Initializes the state of the TPM2 key
5972 protector for later usage, clearing any
5973 previous state, too, if any.
5974
5975 - tpm2_key_protector_clear: Clears any state set by tpm2_key_protector_init.
5976
5977 The way this is expected to be used requires the user to, either
5978 interactively or, normally, via a boot script, initialize/configure
5979 the key protector and then specify that it be used by the "cryptomount"
5980 command (modifications to this command are in a different patch).
5981
5982 For instance, to unseal the raw sealed key file:
5983
5984 tpm2_key_protector_init --keyfile=(hd0,gpt1)/efi/grub/sealed-1.key
5985 cryptomount -u <PART1_UUID> -P tpm2
5986
5987 tpm2_key_protector_init --keyfile=(hd0,gpt1)/efi/grub/sealed-2.key --pcrs=7,11
5988 cryptomount -u <PART2_UUID> -P tpm2
5989
5990 Or, to unseal the TPM 2.0 Key file:
5991
5992 tpm2_key_protector_init --tpm2key=(hd0,gpt1)/efi/grub/sealed-1.tpm
5993 cryptomount -u <PART1_UUID> -P tpm2
5994
5995 tpm2_key_protector_init --tpm2key=(hd0,gpt1)/efi/grub/sealed-2.tpm --pcrs=7,11
5996 cryptomount -u <PART2_UUID> -P tpm2
5997
5998 If a user does not initialize the key protector and attempts to use it
5999 anyway, the protector returns an error.
6000
6001 Before unsealing the key, the TPM2 key protector follows the "TPMPolicy"
6002 sequences to enforce the TPM policy commands to construct a valid policy
6003 digest to unseal the key.
6004
6005 For the TPM 2.0 Key files, "authPolicy" may contain multiple "TPMPolicy"
6006 sequences, the TPM2 key protector iterates "authPolicy" to find a valid
6007 sequence to unseal key. If "authPolicy" is empty or all sequences in
6008 "authPolicy" fail, the protector tries the one from "policy". In case
6009 "policy" is also empty, the protector creates a "TPMPolicy" sequence
6010 based on the given PCR selection.
6011
6012 For the raw sealed key, the TPM2 key protector treats the key file as a
6013 TPM 2.0 Key file without "authPolicy" and "policy", so the "TPMPolicy"
6014 sequence is always based on the PCR selection from the command
6015 parameters.
6016
6017 This commit only supports one policy command: TPM2_PolicyPCR. The
6018 command set will be extended to support advanced features, such as
6019 authorized policy, in the later commits.
6020
6021 Cc: James Bottomley <[email protected]>
6022 Reviewed-by: Daniel Kiper <[email protected]>
6023 Reviewed-by: Stefan Berger <[email protected]>
6024 Tested-by: Stefan Berger <[email protected]>
6025
6026 2024-11-28 Gary Lin <[email protected]>
6027
6028 tss2: Add TPM2 Software Stack (TSS2) support
6029 A Trusted Platform Module (TPM) Software Stack (TSS) provides logic to
6030 compose and submit TPM commands and parse responses.
6031
6032 A limited number of TPM commands may be accessed via the EFI TCG2
6033 protocol. This protocol exposes functionality that is primarily geared
6034 toward TPM usage within the context of Secure Boot. For all other TPM
6035 commands, however, such as sealing and unsealing, this protocol does not
6036 provide any help, with the exception of passthrough command submission.
6037
6038 The SubmitCommand method allows a caller to send raw commands to the
6039 system's TPM and to receive the corresponding response. These
6040 command/response pairs are formatted using the TPM wire protocol. To
6041 construct commands in this way, and to parse the TPM's response, it is
6042 necessary to, first, possess knowledge of the various TPM structures, and,
6043 second, of the TPM wire protocol itself.
6044
6045 As such, this patch includes implementations of various grub_tpm2_* functions
6046 (inventoried below), and logic to write and read command and response
6047 buffers, respectively, using the TPM wire protocol.
6048
6049 Functions:
6050 - grub_tpm2_create(),
6051 - grub_tpm2_createprimary(),
6052 - grub_tpm2_evictcontrol(),
6053 - grub_tpm2_flushcontext(),
6054 - grub_tpm2_load(),
6055 - grub_tpm2_pcr_read(),
6056 - grub_tpm2_policygetdigest(),
6057 - grub_tpm2_policypcr(),
6058 - grub_tpm2_readpublic(),
6059 - grub_tpm2_startauthsession(),
6060 - grub_tpm2_unseal(),
6061 - grub_tpm2_loadexternal(),
6062 - grub_tpm2_hash(),
6063 - grub_tpm2_verifysignature(),
6064 - grub_tpm2_policyauthorize(),
6065 - grub_tpm2_testparms().
6066
6067 Reviewed-by: Daniel Kiper <[email protected]>
6068 Reviewed-by: Stefan Berger <[email protected]>
6069 Tested-by: Stefan Berger <[email protected]>
6070
6071 2024-11-28 Gary Lin <[email protected]>
6072
6073 tss2: Add TPM2 types and Marshal/Unmarshal functions
6074 This commit adds the necessary TPM2 types and structs as the preparation
6075 for the TPM2 Software Stack (TSS2) support. The Marshal/Unmarshal
6076 functions are also added to handle the data structure to be submitted to
6077 TPM2 commands and to be received from the response.
6078
6079 Reviewed-by: Daniel Kiper <[email protected]>
6080 Reviewed-by: Stefan Berger <[email protected]>
6081 Tested-by: Stefan Berger <[email protected]>
6082
6083 2024-11-28 Gary Lin <[email protected]>
6084
6085 tss2: Add TPM2 buffer handling functions
6086 As the preparation to support TPM2 Software Stack (TSS2), this commit
6087 implements the TPM2 buffer handling functions to pack data for the TPM2
6088 commands and unpack the data from the response.
6089
6090 Cc: Stefan Berger <[email protected]>
6091 Reviewed-by: Daniel Kiper <[email protected]>
6092 Tested-by: Stefan Berger <[email protected]>
6093
6094 2024-11-28 Hernan Gatta <[email protected]>
6095
6096 key_protector: Add key protectors framework
6097 A key protector encapsulates functionality to retrieve an unlocking key
6098 for a fully-encrypted disk from a specific source. A key protector
6099 module registers itself with the key protectors framework when it is
6100 loaded and unregisters when unloaded. Additionally, a key protector may
6101 accept parameters that describe how it should operate.
6102
6103 The key protectors framework, besides offering registration and
6104 unregistration functions, also offers a one-stop routine for finding and
6105 invoking a key protector by name. If a key protector with the specified
6106 name exists and if an unlocking key is successfully retrieved by it, the
6107 function returns to the caller the retrieved key and its length.
6108
6109 Cc: Vladimir Serbinenko <[email protected]>
6110 Reviewed-by: Stefan Berger <[email protected]>
6111 Reviewed-by: Daniel Kiper <[email protected]>
6112 Tested-by: Stefan Berger <[email protected]>
6113
6114 2024-11-28 Gary Lin <[email protected]>
6115
6116 libtasn1: Add the documentation
6117 Document libtasn1 in docs/grub-dev.texi and add the upgrade steps.
6118 Also add the patches to make libtasn1 compatible with GRUB code.
6119
6120 Reviewed-by: Vladimir Serbinenko <[email protected]>
6121 Reviewed-by: Daniel Kiper <[email protected]>
6122 Tested-by: Stefan Berger <[email protected]>
6123
6124 2024-11-28 Daniel Axtens <[email protected]>
6125
6126 asn1_test: Test module for libtasn1
6127 Import tests from libtasn1 that use functionality we import.
6128 This test module is integrated into functional_test so that the
6129 user can run the test in GRUB shell.
6130
6131 This doesn't test the full decoder but that will be exercised in
6132 test suites for coming patch sets.
6133
6134 Add testcase target in accordance with commit 5e10be48e5 (tests: Add
6135 check-native and check-nonnative make targets).
6136
6137 Cc: Vladimir Serbinenko <[email protected]>
6138 Reviewed-by: Daniel Kiper <[email protected]>
6139 Tested-by: Stefan Berger <[email protected]>
6140
6141 2024-11-28 Daniel Axtens <[email protected]>
6142
6143 libtasn1: Compile into asn1 module
6144 Create a wrapper file that specifies the module license.
6145 Set up the makefile so it is built.
6146
6147 Reviewed-by: Daniel Kiper <[email protected]>
6148 Tested-by: Stefan Berger <[email protected]>
6149
6150 2024-11-28 Gary Lin <[email protected]>
6151
6152 asn1_test: Enable the testcase only when GRUB_LONG_MAX is larger than GRUB_INT_MAX
6153 There is a testcase to test the values larger than "int" but smaller
6154 than "long". However, for some architectures, "long" and "int" are the
6155 same and the compiler may issue a warning like this:
6156
6157 grub-core/tests/asn1/tests/Test_overflow.c:48:50: error: left shift of negative value [-Werror=shift-negative-value]
6158 unsigned long num = ((long) GRUB_UINT_MAX) << 2;
6159 ^~
6160
6161 To avoid unnecessary error the testcase is enabled only when
6162 GRUB_LONG_MAX is larger than GRUB_INT_MAX.
6163
6164 Reviewed-by: Daniel Kiper <[email protected]>
6165 Tested-by: Stefan Berger <[email protected]>
6166
6167 2024-11-28 Gary Lin <[email protected]>
6168
6169 asn1_test: Use the grub-specific functions and types
6170 This commit converts functions and types to the grub-specific ones:
6171 - LONG_MAX -> GRUB_LONG_MAX,
6172 - INT_MAX -> GRUB_INT_MAX,
6173 - UINT_MAX -> GRUB_UINT_MAX,
6174 - size_t -> grub_size_t,
6175 - memcmp() -> grub_memcmp(),
6176 - memcpy() -> grub_memcpy(),
6177 - free() -> grub_free(),
6178 - strcmp() -> grub_strcmp().
6179
6180 Reviewed-by: Daniel Kiper <[email protected]>
6181 Tested-by: Stefan Berger <[email protected]>
6182
6183 2024-11-28 Gary Lin <[email protected]>
6184
6185 asn1_test: Print the error messages with grub_printf()
6186 This commit replaces printf() and fprintf() with grub_printf() to print
6187 the error messages for the testcases. Besides, asn1_strerror() is used
6188 to convert the result code to strings instead of asn1_perror().
6189
6190 Reviewed-by: Daniel Kiper <[email protected]>
6191 Tested-by: Stefan Berger <[email protected]>
6192
6193 2024-11-28 Gary Lin <[email protected]>
6194
6195 asn1_test: Remove "verbose" and the unnecessary printf()
6196 This commit removes the "verbose" variables and the unnecessary printf()
6197 to simplify the output.
6198
6199 Reviewed-by: Daniel Kiper <[email protected]>
6200 Tested-by: Stefan Berger <[email protected]>
6201
6202 2024-11-28 Gary Lin <[email protected]>
6203
6204 asn1_test: Return either 0 or 1 to reflect the results
6205 Some testcases use exit() to end the test. Since all the asn1 testcases
6206 are invoked as functions, this commit replaces exit() with return to
6207 reflect the test results, so that the main test function can check the
6208 results.
6209
6210 Reviewed-by: Daniel Kiper <[email protected]>
6211 Tested-by: Stefan Berger <[email protected]>
6212
6213 2024-11-28 Gary Lin <[email protected]>
6214
6215 asn1_test: Rename the main functions to the test names
6216 This commit changes the main functions in the testcases to the test
6217 names so that the real "main" test function can invokes them.
6218
6219 Reviewed-by: Daniel Kiper <[email protected]>
6220 Tested-by: Stefan Berger <[email protected]>
6221
6222 2024-11-28 Gary Lin <[email protected]>
6223
6224 asn1_test: Include asn1_test.h only
6225 This commit removes all the headers and only uses asn1_test.h.
6226 To avoid including int.h from grub-core/lib/libtasn1-grub/lib,
6227 CONST_DOWN is defined in reproducers.c.
6228
6229 Reviewed-by: Daniel Kiper <[email protected]>
6230 Tested-by: Stefan Berger <[email protected]>
6231
6232 2024-11-28 Gary Lin <[email protected]>
6233
6234 libtasn1: Fix the potential buffer overrun
6235 In _asn1_tag_der(), the first while loop for the long form may end up
6236 with a "k" value with "ASN1_MAX_TAG_SIZE" and cause the buffer overrun
6237 in the second while loop. This commit tweaks the conditional check to
6238 avoid producing a too large "k".
6239
6240 This is a quick fix and may differ from the official upstream fix.
6241
6242 libtasn1 issue: https://gitlab.com/gnutls/libtasn1/-/issues/49
6243
6244 Reviewed-by: Daniel Kiper <[email protected]>
6245 Tested-by: Stefan Berger <[email protected]>
6246
6247 2024-11-28 Gary Lin <[email protected]>
6248
6249 libtasn1: Use grub_divmod64() for division
6250 Replace a 64-bit division with a call to grub_divmod64(), preventing
6251 creation of __udivdi3() calls on 32-bit platforms.
6252
6253 Reviewed-by: Daniel Kiper <[email protected]>
6254 Tested-by: Stefan Berger <[email protected]>
6255
6256 2024-11-28 Gary Lin <[email protected]>
6257
6258 libtasn1: Adjust the header paths in libtasn1.h
6259 Since libtasn1.h is the header to be included by users, including the
6260 standard POSIX headers in libtasn1.h would force the user to add the
6261 CFLAGS/CPPFLAGS for the POSIX headers.
6262
6263 This commit adjusts the header paths to use the grub headers instead of
6264 the standard POSIX headers, so that users only need to include
6265 libtasn1.h to use libtasn1 functions.
6266
6267 Reviewed-by: Daniel Kiper <[email protected]>
6268 Tested-by: Stefan Berger <[email protected]>
6269
6270 2024-11-28 Gary Lin <[email protected]>
6271
6272 libtasn1: Replace strcat() with _asn1_str_cat()
6273 strcat() is not available in GRUB. This commit replaces strcat() and
6274 _asn1_strcat() with the bounds-checking _asn1_str_cat().
6275
6276 Reviewed-by: Daniel Kiper <[email protected]>
6277 Tested-by: Stefan Berger <[email protected]>
6278
6279 2024-11-28 Gary Lin <[email protected]>
6280
6281 libtasn1: Replace strcat() with strcpy() in _asn1_str_cat()
6282 strcat() is not available in GRUB. This commit replaces strcat() with
6283 strcpy() in _asn1_str_cat() as the preparation to replace other strcat()
6284 with the bounds-checking _asn1_str_cat().
6285
6286 Reviewed-by: Daniel Kiper <[email protected]>
6287 Tested-by: Stefan Berger <[email protected]>
6288
6289 2024-11-28 Gary Lin <[email protected]>
6290
6291 libtasn1: Disable code not needed in GRUB
6292 We don't expect to be able to write ASN.1, only read it,
6293 so we can disable some code.
6294
6295 Do that with #if 0/#endif, rather than deletion. This means
6296 that the difference between upstream and GRUB is smaller,
6297 which should make updating libtasn1 easier in the future.
6298
6299 With these exclusions we also avoid the need for minmax.h,
6300 which is convenient because it means we don't have to
6301 import it from gnulib.
6302
6303 Reviewed-by: Daniel Kiper <[email protected]>
6304 Tested-by: Stefan Berger <[email protected]>
6305
6306 2024-11-28 Daniel Axtens <[email protected]>
6307
6308 libtasn1: Import libtasn1-4.19.0
6309 Import a very trimmed-down set of libtasn1 files:
6310
6311 curl -L -O https://ftp.gnu.org/gnu/libtasn1/libtasn1-4.19.0.tar.gz
6312 tar xvzf libtasn1-4.19.0.tar.gz
6313 rm -rf grub-core/lib/libtasn1
6314 mkdir -p grub-core/lib/libtasn1/lib
6315 mkdir -p grub-core/lib/libtasn1/tests
6316 cp libtasn1-4.19.0/{README.md,COPYING} grub-core/lib/libtasn1
6317 cp libtasn1-4.19.0/lib/{coding.c,decoding.c,element.c,element.h,errors.c,gstr.c,gstr.h,int.h,parser_aux.c,parser_aux.h,structure.c,structure.h} grub-core/libtasn1/lib
6318 cp libtasn1-4.19.0/lib/includes/libtasn1.h grub-core/lib/libtasn1
6319 cp libtasn1-4.19.0/tests/{CVE-2018-1000654-1_asn1_tab.h,CVE-2018-1000654-2_asn1_tab.h,CVE-2018-1000654.c,object-id-decoding.c,object-id-encoding.c,octet-string.c,reproducers.c,Test_overflow.c,Test_simple.c,Test_strings.c} grub-core/lib/libtasn1/tests
6320 rm -rf libtasn1-4.19.0*
6321
6322 Reviewed-by: Vladimir Serbinenko <[email protected]>
6323 Reviewed-by: Daniel Kiper <[email protected]>
6324 Tested-by: Stefan Berger <[email protected]>
6325
6326 2024-11-28 Daniel Axtens <[email protected]>
6327
6328 posix_wrap: Tweaks in preparation for libtasn1
6329 Cc: Vladimir Serbinenko <[email protected]>
6330 Reviewed-by: Daniel Kiper <[email protected]>
6331 Tested-by: Stefan Berger <[email protected]>
6332
6333 2024-11-28 Rasmus Villemoes <[email protected]>
6334
6335 kern/fs: Honour file->read_hook() in grub_fs_blocklist_read()
6336 Unlike files accessed via a normal file system, the file->read_hook() is
6337 not honoured when using blocklist notation.
6338
6339 This means that when trying to use a dedicated, 1 KiB, raw partition
6340 for the environment block and hence does something like
6341
6342 save_env --file=(hd0,gpt9)0+2 X Y Z
6343
6344 this fails with "sparse file not allowed", which is rather unexpected,
6345 as I've explicitly said exactly which blocks should be used. Adding
6346 a little debugging reveals that grub_file_size(file) is 1024 as expected,
6347 but total_length is 0, simply because the callback was never invoked, so
6348 blocklists is an empty list.
6349
6350 Fix that by honouring the ->read_hook() set by the caller, also when
6351 a "file" is specified with blocklist notation.
6352
6353 Reviewed-by: Vladimir Serbinenko <[email protected]>
6354 Reviewed-by: Daniel Kiper <[email protected]>
6355
6356 2024-11-28 Glenn Washburn <[email protected]>
6357
6358 docs: Fix incorrect and potentially confusing language and minor formatting
6359 Reviewed-by: Vladimir Serbinenko <[email protected]>
6360 Reviewed-by: Daniel Kiper <[email protected]>
6361
6362 2024-10-31 Andrew Hamilton <[email protected]>
6363
6364 docs: Correct GRUB config file name for network boot
6365 Correct the documentation for the grub.cfg searching via network that
6366 will be done based on ethernet type, -01, which was missing, and a given
6367 MAC address.
6368
6369 Fixes: https://savannah.gnu.org/bugs/?65152
6370
6371 Reviewed-by: Daniel Kiper <[email protected]>
6372
6373 2024-10-31 Andrew Hamilton <[email protected]>
6374
6375 docs: Correct chainloader UEFI secure boot info
6376 Correct documentation for UEFI secure boot to remove statement that
6377 chainloader does not work with secure boot. This was fixed by the commit
6378 6d05264 (kern/efi/sb: Add chainloaded image as shim's verifiable object).
6379
6380 Fixes: https://savannah.gnu.org/bugs/?62004
6381
6382 Reviewed-by: Daniel Kiper <[email protected]>
6383
6384 2024-10-31 Andrew Hamilton <[email protected]>
6385
6386 docs: Correct PXE environment variables descriptions
6387 Correct documentation for pxe_default_server, pxe_default_gatway and
6388 pxe_blksize. Only pxe_default_server is actually used (alias for
6389 net_default_server). So, capture this and remove the other two.
6390
6391 Fixes: https://savannah.gnu.org/bugs/?54480
6392
6393 Reviewed-by: Daniel Kiper <[email protected]>
6394
6395 2024-10-31 Valentin Gehrke <[email protected]>
6396
6397 loader/multiboot: Do not add modules before successful download
6398 Multiboot modules that could not be read successfully, e.g. via network,
6399 should not be added to the list of modules to forward to the operating
6400 system that is to be booted subsequently.
6401
6402 This patch is necessary because even if a grub.cfg checks whether or not
6403 a module was successfully downloaded, it is futile to retry a failed
6404 download as the corrupted module will be forwarded either way.
6405
6406 Reviewed-by: Daniel Kiper <[email protected]>
6407
6408 2024-10-31 Sudhakar Kuppusamy <[email protected]>
6409
6410 grub-mkimage: Add SBAT metadata into ELF note for PowerPC targets
6411 The SBAT metadata is read from CSV file and transformed into an ELF note
6412 with the -s option.
6413
6414 Reviewed-by: Daniel Kiper <[email protected]>
6415
6416 2024-10-31 Sudhakar Kuppusamy <[email protected]>
6417
6418 grub-mkimage: Create new ELF note for SBAT
6419 In order to store the SBAT data we create a new ELF note. The string
6420 ".sbat", zero-padded to 4 byte alignment, shall be entered in the name
6421 field. The string "SBAT"'s ASCII values, 0x53424154, should be entered
6422 in the type field.
6423
6424 Reviewed-by: Daniel Kiper <[email protected]>
6425
6426 2024-10-31 Leo Sandoval <[email protected]>
6427
6428 commands/legacycfg: Avoid closing file twice
6429 An internal (at Red Hat) static soure code scan detected an
6430 use-after-free scenario:
6431
6432 Error: USE_AFTER_FREE (CWE-416):
6433 grub-2.06/grub-core/commands/legacycfg.c:194: freed_arg: "grub_file_close" frees "file".
6434 grub-2.06/grub-core/commands/legacycfg.c:201: deref_arg: Calling "grub_file_close" dereferences freed pointer "file".
6435 # 199| if (!args)
6436 # 200| {
6437 # 201|-> grub_file_close (file);
6438 # 202| grub_free (suffix);
6439 # 203| grub_free (entrysrc);
6440
6441 So, remove the extra file close call.
6442
6443 Reviewed-by: Daniel Kiper <[email protected]>
6444
6445 2024-10-31 Daniel Kiper <[email protected]>
6446
6447 nx: Rename GRUB_DL_ALIGN to DL_ALIGN
6448 Rename has been skipped by mistake in the original commit.
6449
6450 Fixes: 94649c026 (nx: Set page permissions for loaded modules)
6451
6452 Tested-by: Sudeep Holla <[email protected]>
6453 Reviewed-by: Ross Philipson <[email protected]>
6454
6455 2024-10-31 Benjamin Herrenschmidt <[email protected]>
6456
6457 kern/acpi: Fix out of bounds access in grub_acpi_xsdt_find_table()
6458 The calculation of the size of the table was incorrect (copy/pasta from
6459 grub_acpi_rsdt_find_table() I assume...). The entries are 64-bit long.
6460
6461 This causes us to access beyond the end of the table which is causing
6462 crashes during boot on some systems. Typically this is causing a crash
6463 on VMWare when using UEFI and enabling serial autodetection, as
6464
6465 grub_acpi_find_table (GRUB_ACPI_SPCR_SIGNATURE);
6466
6467 will goes past the end of the table (the SPCR table doesn't exits).
6468
6469 Reviewed-by: Vladimir Serbinenko <[email protected]>
6470 Reviewed-by: Daniel Kiper <[email protected]>
6471 Reviewed-by: Ross Philipson <[email protected]>
6472 Tested-by: Renata Ravanelli <[email protected]>
6473
6474 2024-10-11 Mate Kukri <[email protected]>
6475
6476 nx: Set the NX compatible flag for the GRUB EFI images
6477 For NX the GRUB binary has to announce that it is compatible with the
6478 NX feature. This implies that when loading the executable GRUB image
6479 several attributes are true:
6480 - the binary doesn't need an executable stack,
6481 - the binary doesn't need sections to be both executable and writable,
6482 - the binary knows how to use the EFI Memory Attributes Protocol on code
6483 it is loading.
6484
6485 This patch:
6486 - adds a definition for the PE DLL Characteristics flag GRUB_PE32_NX_COMPAT,
6487 - changes grub-mkimage to set that flag.
6488
6489 Reviewed-by: Daniel Kiper <[email protected]>
6490
6491 2024-10-11 Mate Kukri <[email protected]>
6492
6493 nx: Set page permissions for loaded modules
6494 For NX we need to set write and executable permissions on the sections
6495 of GRUB modules when we load them. All allocatable sections are marked
6496 readable. In addition:
6497 - SHF_WRITE sections are marked as writable,
6498 - and SHF_EXECINSTR sections are marked as executable.
6499
6500 Where relevant for the platform the tramp and GOT areas are marked non-writable.
6501
6502 Reviewed-by: Daniel Kiper <[email protected]>
6503
6504 2024-10-11 Mate Kukri <[email protected]>
6505
6506 nx: Add memory attribute get/set API
6507 For NX we need to set the page access permission attributes for write
6508 and execute permissions. This patch adds two new primitives, grub_set_mem_attrs()
6509 and grub_clear_mem_attrs(), and associated constants definitions used
6510 for that purpose. For most platforms it adds a dummy implementation.
6511 On EFI platforms it implements the primitives using the EFI Memory
6512 Attribute Protocol, defined in UEFI 2.10 specification.
6513
6514 Reviewed-by: Daniel Kiper <[email protected]>
6515
6516 2024-10-11 Mate Kukri <[email protected]>
6517
6518 modules: Load module sections at page-aligned addresses
6519 Currently we load module sections at whatever alignment gcc+ld happened
6520 to dump into the ELF section header which is often less then the page
6521 size. Since NX protections are page based this alignment must be rounded
6522 up to page size on platforms supporting NX protections. This patch
6523 switches EFI platforms to load module sections at 4 KiB page-aligned
6524 addresses. It then changes the allocation size computation and the
6525 loader code in grub_dl_load_segments() to align the locations and sizes
6526 up to these boundaries and fills any added padding with zeros. All of
6527 this happens before relocations are applied, so the relocations factor
6528 that in with no change.
6529
6530 Reviewed-by: Daniel Kiper <[email protected]>
6531
6532 2024-10-10 Peter Jones <[email protected]>
6533
6534 modules: Don't allocate space for non-allocable sections
6535 Currently when loading GRUB modules we allocate space for all sections
6536 including those without SHF_ALLOC set. We then copy the sections that
6537 /do/ have SHF_ALLOC set into the allocated memory leaving some of our
6538 allocation untouched forever. Additionally, on platforms with GOT fixups
6539 and trampolines we currently compute alignment round-ups for the
6540 sections and sections with sh_size = 0. This patch removes the extra
6541 space from the allocation computation and makes the allocation
6542 computation loop skip empty sections as the loading loop does.
6543
6544 Reviewed-By: Vladimir Serbinenko <[email protected]>
6545 Reviewed-by: Daniel Kiper <[email protected]>
6546
6547 2024-10-10 Peter Jones <[email protected]>
6548
6549 modules: Strip .llvm_addrsig sections and similar
6550 Currently GRUB modules built with Clang or GCC have several sections
6551 which we don't actually need or support. We already have a list of
6552 sections to skip in genmod.sh and this patch adds the following
6553 sections to that list (as well as a few newlines):
6554 - .note.gnu.property
6555 - .llvm*
6556
6557 Note that the glob there won't work without a new enough linker but the
6558 failure is just reversion to the status quo. So, that's not a big problem.
6559
6560 Reviewed-By: Vladimir Serbinenko <[email protected]>
6561 Reviewed-by: Daniel Kiper <[email protected]>
6562
6563 2024-10-10 Peter Jones <[email protected]>
6564
6565 modules: Make .module_license read-only
6566 Currently .module_license is set writable, that is, the section has the
6567 SHF_WRITE flag set, in the module's ELF headers. This probably never
6568 actually matters but it can't possibly be correct. The patch sets that
6569 data as "const" which causes that flag not to be set.
6570
6571 Reviewed-By: Vladimir Serbinenko <[email protected]>
6572 Reviewed-by: Daniel Kiper <[email protected]>
6573
6574 2024-10-10 Daniel Kiper <[email protected]>
6575
6576 i386/memory: Rename PAGE_SIZE to GRUB_PAGE_SIZE and make it global
6577 This is an x86-specific thing and should be available globally.
6578
6579 Reviewed-by: Ross Philipson <[email protected]>
6580 Reviewed-by: Daniel Kiper <[email protected]>
6581
6582 2024-10-10 Daniel Kiper <[email protected]>
6583
6584 i386/memory: Rename PAGE_SHIFT to GRUB_PAGE_SHIFT
6585 This fixes naming inconsistency that goes against coding style as well
6586 as helps to avoid potential conflicts and confusion as this constant is
6587 used in multiple places.
6588
6589 Reviewed-by: Ross Philipson <[email protected]>
6590 Reviewed-by: Daniel Kiper <[email protected]>
6591
6592 2024-10-10 Daniel Kiper <[email protected]>
6593
6594 i386/msr: Extract and improve MSR support detection code
6595 Currently rdmsr and wrmsr commands have own MSR support detection code.
6596 This code is the same. So, it is duplicated. Additionally, this code
6597 cannot be reused by others. Hence, extract this code to a function and
6598 make it public. By the way, improve a code a bit.
6599
6600 Additionally, use GRUB_ERR_BAD_DEVICE instead of GRUB_ERR_BUG to signal
6601 an error because errors encountered by this new routine are not bugs.
6602
6603 Reviewed-by: Ross Philipson <[email protected]>
6604 Reviewed-by: Daniel Kiper <[email protected]>
6605
6606 2024-10-10 Daniel Kiper <[email protected]>
6607
6608 i386/msr: Rename grub_msr_read() and grub_msr_write()
6609 Use more obvious names which match corresponding instructions:
6610 * grub_msr_read() => grub_rdmsr(),
6611 * grub_msr_write() => grub_wrmsr().
6612
6613 Reviewed-by: Ross Philipson <[email protected]>
6614 Reviewed-by: Daniel Kiper <[email protected]>
6615
6616 2024-10-10 Daniel Kiper <[email protected]>
6617
6618 i386/msr: Merge rdmsr.h and wrmsr.h into msr.h
6619 It does not make sense to have separate headers for individual static
6620 functions. So, make one common place to store them.
6621
6622 Reviewed-by: Ross Philipson <[email protected]>
6623 Reviewed-by: Daniel Kiper <[email protected]>
6624
6625 2024-10-10 Michael Chang <[email protected]>
6626
6627 commands/tpm: Skip loopback image measurement
6628 The loopback image is configured to function as a disk by being mapped
6629 as a block device. Instead of measuring the entire block device we
6630 should focus on tracking the individual files accessed from it. For
6631 example, we do not directly measure block devices like hd0 disk but the
6632 files opened from it.
6633
6634 This method is important to avoid running out of memory since loopback
6635 images can be very large. Trying to read and measure the whole image at
6636 once could cause out of memory errors and disrupt the boot process.
6637
6638 Reviewed-by: Daniel Kiper <[email protected]>
6639
6640 2024-10-10 Michael Chang <[email protected]>
6641
6642 net/drivers/efi/efinet: Skip virtual VLAN devices during card enumeration
6643 Similarly to the issue described in commit c52ae4057 (efinet: skip
6644 virtual IPv4 and IPv6 devices during card enumeration) the UEFI PXE
6645 driver creates additional VLAN child devices when a VLAN ID is
6646 configured on a network interface associated with a physical NIC. These
6647 virtual VLAN devices must be skipped during card enumeration to ensure
6648 that the subsequent SNP exclusive open operation targets the correct
6649 physical card instances. Otherwise packet transfer would fail.
6650
6651 A device path example with VLAN nodes:
6652
6653 /MAC(123456789ABC,0x1)/Vlan(20)/IPv4(0.0.0.0,0x0,DHCP,0.0.0.0,0.0.0.0,0.0.0.0)
6654
6655 Reviewed-by: Daniel Kiper <[email protected]>
6656
6657 2024-10-10 Michael Chang <[email protected]>
6658
6659 efi/console: Properly clear leftover artifacts from the screen
6660 A regression in GRUB 2.12 causes the GRUB screen to become cluttered
6661 with artifacts from the previous screen whether it's the UEFI post UI,
6662 UEFI shell or any graphical UI running before GRUB. This issue occurs
6663 in situations like booting GRUB from the UEFI shell and going straight
6664 to the rescue or command shell causing visual discomfort.
6665
6666 The regression was introduced by commit 2d7c3abd8 (efi/console: Do not
6667 set text-mode until it is actually needed). To address the screen
6668 flickering issue this commit suppresses the text-mode setting until the
6669 first output is requested. Before text-mode is set any attempt to clear
6670 the screen has no effect. This inactive period renders the clear screen
6671 ineffective in early boot stages, potentially leaving leftover artifacts
6672 that will clutter the GRUB console display, as there is no guarantee
6673 there will always be a clear screen after the first output.
6674
6675 The issue is fixed by ensuring grub_console_cls() to work through lazy
6676 mode-setting, while also avoiding screen clearing for the hidden menu
6677 which the flicker-free patch aims to improve.
6678
6679 Fixes: 2d7c3abd8 (efi/console: Do not set text-mode until we actually need it)
6680
6681 Reviewed-by: Daniel Kiper <[email protected]>
6682
6683 2024-10-10 Heinrich Schuchardt <[email protected]>
6684
6685 kern/riscv/efi/init: Use time register in grub_efi_get_time_ms()
6686 The cycle register is not guaranteed to count at constant frequency.
6687 If it is counting at all depends on the state the performance monitoring
6688 unit. Use the time register to measure time.
6689
6690 Reviewed-by: Daniel Kiper <[email protected]>
6691
6692 2024-09-05 Frediano Ziglio <[email protected]>
6693
6694 loader/efi/linux: Reset freed pointer
6695 Avoid dangling pointer. Code should not be reached but better safe than sorry.
6696
6697 Reviewed-by: Daniel Kiper <[email protected]>
6698
6699 2024-09-05 Frediano Ziglio <[email protected]>
6700
6701 loader/efi/linux: Reuse len variable
6702 Reviewed-by: Daniel Kiper <[email protected]>
6703
6704 2024-09-05 Frediano Ziglio <[email protected]>
6705
6706 lib/x86_64/relocator_asm: Use .quad instead of .long
6707 They are single 64-bit values. Used in other assembly files too.
6708
6709 Reviewed-by: Vladimir Serbinenko <[email protected]>
6710 Reviewed-by: Daniel Kiper <[email protected]>
6711
6712 2024-09-05 Frediano Ziglio <[email protected]>
6713
6714 lib/x86_64/relocator_asm: Fix comment in code
6715 The instruction uses a 64-bit immediate.
6716
6717 Reviewed-by: Daniel Kiper <[email protected]>
6718
6719 2024-09-05 Frediano Ziglio <[email protected]>
6720
6721 loader/efi/linux: Update comment
6722 The function called is grub_utf8_to_utf16().
6723
6724 Reviewed-by: Vladimir Serbinenko <[email protected]>
6725 Reviewed-by: Daniel Kiper <[email protected]>
6726
6727 2024-09-05 Vladimir Serbinenko <[email protected]>
6728
6729 util/grub-mkimagexx: Explicitly move modules to __bss_start for MIPS targets
6730 Assembly code looks for modules at __bss_start. Make this position explicit
6731 rather than matching BSS alignment and module alignment.
6732
6733 Reviewed-by: Daniel Kiper <[email protected]>
6734
6735 2024-09-05 Vladimir Serbinenko <[email protected]>
6736
6737 include/grub/offsets.h: Set mod_align to 4 on MIPS
6738 Module structure has natural alignment of 4. Respect it explicitly
6739 rather than relying on the fact that _end is usually aligned.
6740
6741 Reviewed-by: Daniel Kiper <[email protected]>
6742
6743 2024-09-05 Vladimir Serbinenko <[email protected]>
6744
6745 gentpl: Put boot/mips/startup_raw.S into beginning of the image
6746 Otherwise it breaks the decompressors for MIPS targets.
6747
6748 Reviewed-by: Daniel Kiper <[email protected]>
6749
6750 2024-09-05 Vladimir Serbinenko <[email protected]>
6751
6752 configure: Add -mno-gpopt option for mips and mipsel targets
6753 Without it compiler generates GPREL16 references which do not work
6754 with our memory layout.
6755
6756 Reviewed-by: Daniel Kiper <[email protected]>
6757
6758 2024-09-05 Vladimir Serbinenko <[email protected]>
6759
6760 lib/xzembed/xz_dec_bcj: Silence warning when no BCJ is available
6761 BCJ is not available for all platforms hence arguments may end up unused.
6762
6763 Reviewed-by: Daniel Kiper <[email protected]>
6764
6765 2024-09-05 Vladimir Serbinenko <[email protected]>
6766
6767 fs/erofs: Replace 64-bit modulo with bitwise operations
6768 Otherwise depending on compiler we end up with umoddi3 reference and
6769 failed module dependency resolution.
6770
6771 Reviewed-by: Daniel Kiper <[email protected]>
6772
6773 2024-09-05 Vladimir Serbinenko <[email protected]>
6774
6775 configure: Look for .otf fonts
6776 Reviewed-by: Daniel Kiper <[email protected]>
6777
6778 2024-09-05 Mate Kukri <[email protected]>
6779
6780 loader/efi/chainloader: Do not print device path of chainloaded file
6781 Users have no reason to see this and it can break graphical boot.
6782
6783 Reviewed-by: Daniel Kiper <[email protected]>
6784
6785 2024-09-05 Andrew Hamilton <[email protected]>
6786
6787 docs: Document all GRUB modules
6788 Add documentation for all GRUB modules contained in the source code tree.
6789 When possible, cross-references to additional detail on commands was added
6790 from their corresponding module documentation. In addition, documentation
6791 for the file command was added.
6792
6793 Reviewed-by: Daniel Kiper <[email protected]>
6794
6795 2024-09-05 Michael Chang <[email protected]>
6796
6797 commands/bli: Fix crash in get_part_uuid()
6798 The get_part_uuid() function made an assumption that the target GRUB
6799 device is a partition device and accessed device->disk->partition
6800 without checking for NULL. There are four situations where this
6801 assumption is problematic:
6802
6803 1. The device is a net device instead of a disk.
6804 2. The device is an abstraction device, like LVM, RAID, or CRYPTO, which
6805 is mostly logical "disk" ((lvmid/<UUID>) and so on).
6806 3. Firmware RAID may present the ESP to GRUB as an EFI disk (hd0) device
6807 if it is contained within a Linux software RAID.
6808 4. When booting from a CD-ROM, the ESP is a VFAT image indexed by the El
6809 Torito boot catalog. The boot device is set to (cd0), corresponding
6810 to the CD-ROM image mounted as an ISO 9660 filesystem.
6811
6812 As a result, get_part_uuid() could lead to a NULL pointer dereference
6813 and trigger a synchronous exception during boot if the ESP falls into
6814 one of these categories. This patch fixes the problem by adding the
6815 necessary checks to handle cases where the ESP is not a partition device.
6816
6817 Additionally, to avoid disrupting the boot process, this patch relaxes
6818 the severity of the errors in this context to non-critical. Errors will
6819 be logged, but they will not prevent the boot process from continuing.
6820
6821 Fixes: e0fa7dc84 (bli: Add a module for the Boot Loader Interface)
6822
6823 Reviewed-By: Oliver Steffen <[email protected]>
6824 Reviewed-by: Daniel Kiper <[email protected]>
6825
6826 2024-06-20 Thomas Schmitt <[email protected]>
6827
6828 util/grub-mkrescue: Check existence of option arguments
6829 As reported by Victoriia Egorova in bug 65880, grub-mkrescue does not
6830 verify that the expected argument of an option like -d or -k does really
6831 exist in argv. So, check the loop counter before incrementing it inside
6832 the loop which copies argv to argp_argv. Issue an error message similar
6833 to what older versions of grub-mkrescue did with a missing argument,
6834 e.g. 2.02.
6835
6836 Fixes: https://savannah.gnu.org/bugs/index.php?65880
6837
6838 Reviewed-by: Daniel Kiper <[email protected]>
6839
6840 2024-06-20 Tobias Heider <[email protected]>
6841
6842 loader/efi/fdt: Add fdtdump command to access device tree
6843 The fdtdump command allows dumping arbitrary device tree properties
6844 and saving them to a variable similar to the smbios command.
6845
6846 This is useful in scripts where further actions such as selecting
6847 a kernel or loading another device tree depend on the compatible
6848 or model values of the device tree provided by the firmware.
6849
6850 For now only the root level properties of the dtb are exposed.
6851
6852 Reviewed-by: Daniel Kiper <[email protected]>
6853
6854 2024-06-20 Vladimir Serbinenko <[email protected]>
6855
6856 osdep/devmapper/getroot: Unmark 2 strings for translation
6857 First they're use macros so they can't be translated as-is.
6858 Second there is no point in translating them as they're too technical.
6859
6860 Reviewed-by: Daniel Kiper <[email protected]>
6861
6862 2024-06-20 Vladimir Serbinenko <[email protected]>
6863
6864 loader/emu/linux: Fix determination of program name
6865 Current code works only if package matches binary name transformation rules.
6866 It's often true but is not guaranteed.
6867
6868 Fixes: https://savannah.gnu.org/bugs/?64410
6869
6870 Reviewed-by: Daniel Kiper <[email protected]>
6871
6872 2024-06-20 Vladimir Serbinenko <[email protected]>
6873
6874 disk/cryptodisk: Fix translatable message
6875 Fixes: https://savannah.gnu.org/bugs/?64408
6876
6877 Reviewed-by: Daniel Kiper <[email protected]>
6878
6879 2024-06-20 Vladimir Serbinenko <[email protected]>
6880
6881 tests: Add test for ZFS zstd
6882 Reviewed-by: Daniel Kiper <[email protected]>
6883
6884 fs/zfs/zfs: Add support for zstd compression
6885 Reviewed-by: Daniel Kiper <[email protected]>
6886
6887 2024-06-20 Mate Kukri <[email protected]>
6888
6889 kern/efi/mm: Detect calls to grub_efi_drop_alloc() with wrong page counts
6890 Silently keeping entries in the list if the address matches, but the
6891 page count doesn't is a bad idea, and can lead to double frees.
6892
6893 grub_efi_free_pages() have already freed parts of this block by this
6894 point, and thus keeping the whole block in the list and freeing it again
6895 at exit can lead to double frees.
6896
6897 Reviewed-by: Daniel Kiper <[email protected]>
6898
6899 2024-06-20 Mate Kukri <[email protected]>
6900
6901 kern/efi/mm: Change grub_efi_allocate_pages_real() to call semantically correct free function
6902 If the firmware happens to return 0 as an address of allocated pages,
6903 grub_efi_allocate_pages_real() tries to allocate a new set of pages,
6904 and then free the ones at address 0.
6905
6906 However at that point grub_efi_store_alloc() wasn't yet called, so
6907 freeing the pages at 0 using grub_efi_free_pages() which calls
6908 grub_efi_drop_alloc() isn't necessary, so let's call b->free_pages()
6909 instead.
6910
6911 The call to grub_efi_drop_alloc() doesn't seem particularly harmful,
6912 because it seems to do nothing if the allocation it is asked to drop
6913 isn't on the list, but the call to it is obviously unnecessary here.
6914
6915 Reviewed-by: Daniel Kiper <[email protected]>
6916
6917 2024-06-20 Mate Kukri <[email protected]>
6918
6919 kern/efi/mm: Change grub_efi_mm_add_regions() to keep track of map allocation size
6920 If the map was too big for the initial allocation, it was freed and replaced
6921 with a bigger one, but the free call still used the hard-coded size.
6922
6923 Seems like this wasn't hit for a long time, because most firmware maps
6924 fit into 12K.
6925
6926 This bug was triggered on Project Mu firmware with a big memory map, and
6927 results in the heap getting trashed and the firmware ASSERTING on
6928 corrupted heap guard values when GRUB exits.
6929
6930 Reviewed-by: Daniel Kiper <[email protected]>
6931
6932 2024-06-20 Yifan Zhao <[email protected]>
6933
6934 tests/util/grub-fs-tester: Fix EROFS label tests in grub-fs-tester
6935 mkfs.erofs with version < 1.6 does not support the -L option.
6936 Let's detect the version of mkfs.erofs and skip the label tests
6937 if it is not supported.
6938
6939 Suggested-by: Glenn Washburn <[email protected]>
6940 Reviewed-by: Daniel Kiper <[email protected]>
6941
6942 2024-06-20 Glenn Washburn <[email protected]>
6943
6944 tests: Switch to requiring exfatprogs from exfat-utils
6945 The current Debian stable, now 12, has dropped the exfat-utils package
6946 that the exfat filesystem test requires to run. There is an exfatprogs
6947 package that replaces exfat-utils, though it is not a drop-in replacement
6948 because mkfs.exfat has differing command line option names. Note, that
6949 we're not yet switching to using the exfat kernel module because this
6950 allows the testings on kernels that do not have the module.
6951
6952 Update mkfs.exfat usage to adhere to the different exfatprogs usage. Also,
6953 the exfatprogs mkfs.exfat, following the exfat specification more closely,
6954 only allows a maximum of 22 bytes of UTF-16 characters in the volume label
6955 compared to 30 bytes from exfat-utils. So the exfat label test is updated
6956 accordingly.
6957
6958 Update documentation to note that exfatprogs is now needed and also
6959 exfat-fuse, which is needed do the fuse mount.
6960
6961 Reviewed-by: Daniel Kiper <[email protected]>
6962
6963 2024-06-20 Glenn Washburn <[email protected]>
6964
6965 tests/util/grub-shell-luks-tester: Fix detached header test getting wrong header path
6966 When $detached_header was set 1, $luksdiskfile was set to the LUKS header
6967 file path with "${detached_header:-$luksfile}" appended, which evaluates
6968 to "1". Fix this by using two statements to set $luksdiskfile. The first
6969 sets it to the header file if $detached_header is set, otherwise leave it
6970 unset. The second statement sets it to itself if it is already set,
6971 otherwise it is set to $luksfile.
6972
6973 Fixes: a7b540e6e (tests: Add cryptomount functional test)
6974
6975 Reviewed-by: Daniel Kiper <[email protected]>
6976
6977 2024-06-20 Glenn Washburn <[email protected]>
6978
6979 tests/util/grub-shell: Add flexibility in QEMU firmware handling
6980 First look for firmware files in the source directory and then, if not
6981 found, look for them in locations where Debian installs them. Prefer to
6982 use the unified firmware file and, if not found, use the pflash firmware
6983 files split in to code and variables. By looking for files in the source
6984 directory first, system firmware files can be overridden and it can be
6985 ensured that the tests can be run regardless of the distro or where the
6986 system firmware files are stored. If no firmware files are found, print
6987 an error message and exit with error.
6988
6989 If a firmware VARS file is found, use it with snapshot mode enabled, which
6990 makes the VARS writable to the virtual machine, but does not write back
6991 the changes to the file. This allows using the readonly system VARS file
6992 without copying it or using it in readonly mode, which causes the ARM
6993 machine to fail. This also gives tests effectively their own ephemeral VARS
6994 file that can be written to without causing side-effects for other tests.
6995
6996 Reviewed-by: Daniel Kiper <[email protected]>
6997
6998 2024-06-20 Glenn Washburn <[email protected]>
6999
7000 tests/util/grub-shell: Use pflash instead of -bios to load UEFI firmware
7001 According to the OVMF whitepaper [1]:
7002
7003 IMPORTANT: Never pass OVMF.fd to qemu with the -bios option. That option
7004 maps the firmware image as ROM into the guest's address space, and forces
7005 OVMF to emulate non-volatile variables with a fallback driver that is
7006 bound to have insufficient and confusing semantics.
7007
7008 Use the pflash interface instead. Currently the unified firmware file is
7009 used, which contains both firmware code and variable sections. By enabling
7010 snapshot on the pflash device, the firmware can be loaded in such a way
7011 that variables can be written to without writing to the backing file.
7012
7013 Since pflash does no searching for firmware paths that are not absolute,
7014 unlike the -bios option, also make firmware paths absolute. Additionally,
7015 update the previous firmware paths or file names that did not correspond to
7016 ones installed by Debian.
7017
7018 Use the q35 machine, instead of the default i440fx, for i386-efi because
7019 the default machine type does not emulate a flash device, which is now
7020 needed to load the firmware.
7021
7022 [1] http://www.linux-kvm.org/downloads/lersek/ovmf-whitepaper-c770f8c.txt
7023
7024 Reviewed-by: Daniel Kiper <[email protected]>
7025
7026 2024-06-20 Glenn Washburn <[email protected]>
7027
7028 tests/util/grub-shell: Print gdbinfo if on EFI platform
7029 Allow using GDB to debug a failing QEMU test. This output does not cause
7030 issues for tests because it happens before the trim line, and so will be
7031 ignored.
7032
7033 Reviewed-by: Daniel Kiper <[email protected]>
7034
7035 2024-06-20 Glenn Washburn <[email protected]>
7036
7037 configure: Add Debian/Ubuntu DejaVu font path
7038 Reviewed-by: Daniel Kiper <[email protected]>
7039
7040 2024-06-20 Udo Steinberg <[email protected]>
7041
7042 term/ns8250-spcr: Add one more 16550 debug type
7043 Type 0x01 was introduced with the ACPI DBGP table and type 0x12 was introduced
7044 with the ACPI DBG2 table. Type 0x12 is used by the ACPI SPCR table on recent
7045 AWS bare-metal instances (c6i/c7i). Also give each debug type a proper name.
7046
7047 Reviewed-by: Glenn Washburn <[email protected]>
7048 Reviewed-by: Daniel Kiper <[email protected]>
7049
7050 2024-06-20 Vladimir Serbinenko <[email protected]>
7051
7052 loader/i386/multiboot_mbi: Fix handling of errors in broken aout-kludge
7053 Current code in some codepaths neither discards nor reports errors.
7054 Properly surface the error.
7055
7056 While on it split 2 cases of unrelated variables both named err.
7057
7058 Reviewed-by: Daniel Kiper <[email protected]>
7059
7060 2024-06-20 Michael Chang <[email protected]>
7061
7062 net/drivers/ieee1275/ofnet: Remove 200 ms timeout in get_card_packet() to reduce input latency
7063 When GRUB image is netbooted on ppc64le, the keyboard input exhibits
7064 significant latency, reports even say that characters are processed
7065 about once per second. This issue makes interactively trying to debug
7066 a ppc64le config very difficult.
7067
7068 It seems that the latency is largely caused by a 200 ms timeout in the
7069 idle event loop, during which the network card interface is consistently
7070 polled for incoming packets. Often, no packets arrive during this
7071 period, so the timeout nearly always expires, which blocks the response
7072 to key inputs.
7073
7074 Furthermore, this 200 ms timeout might not need to be enforced at this
7075 basic layer, considering that GRUB performs synchronous reads and its
7076 timeout management is actually handled by higher layers, not directly in
7077 the card instance. Additionally, the idle polling, which reacts to
7078 unsolicited packets like ICMP and SLAAC, would be fine at a less frequent
7079 polling interval, rather than needing a timeout for receiving a response.
7080
7081 For these reasons, we believe the timeout in get_card_packet() should be
7082 effectively removed. According to test results, the delay has disappeared,
7083 and it is now much easier to use interactively.
7084
7085 Signed-Off-by: Michael Chang <[email protected]>
7086 Tested-by: Tony Jones <[email protected]>
7087 Reviewed-by: Daniel Kiper <[email protected]>
7088
7089 2024-06-06 Hector Cao <[email protected]>
7090
7091 commands/efi/tpm: Re-enable measurements on confidential computing platforms
7092 The measurements for confidential computing has been introduced in the
7093 commit 4c76565b6 (efi/tpm: Add EFI_CC_MEASUREMENT_PROTOCOL support).
7094 Recently the patch 30708dfe3 (tpm: Disable the tpm verifier if the TPM
7095 device is not present) has been introduced to optimize the memory usage
7096 when a TPM device is not available on platforms. This fix prevents the
7097 tpm module to be loaded on confidential computing platforms, e.g. Intel
7098 machines with TDX enabled, where the TPM device is not available.
7099
7100 In this patch, we propose to load the tpm module for this use case by
7101 generalizing the tpm feature detection in order to cover CC platforms.
7102 Basically, we do it by detecting the availability of the
7103 EFI_CC_MEASUREMENT_PROTOCOL EFI protocol.
7104
7105 Fixes: https://savannah.gnu.org/bugs/?65821
7106 Fixes: 30708dfe3 (tpm: Disable the tpm verifier if the TPM device is not present)
7107
7108 Reviewed-by: Daniel Kiper <[email protected]>
7109 Reviewed-by: Kuppuswamy Sathyanarayanan <[email protected]>
7110
7111 2024-06-06 Tianjia Zhang <[email protected]>
7112
7113 util/grub-mkpasswd-pbkdf2: Simplify the main function implementation
7114 Allocate memory if needed, while saving the corresponding release
7115 operation, reducing the amount of code and code complexity.
7116
7117 Reviewed-by: Daniel Kiper <[email protected]>
7118
7119 2024-06-06 Avnish Chouhan <[email protected]>
7120
7121 kern/ieee1275/init: Add IEEE 1275 Radix support for KVM on Power
7122 This patch adds support for Radix, Xive and Radix_gtse in Options
7123 vector5 which is required for KVM LPARs. KVM LPARs ONLY support
7124 Radix and not the Hash. Not enabling Radix on any PowerVM KVM LPARs
7125 will result in boot failure.
7126
7127 Reviewed-by: Daniel Kiper <[email protected]>
7128
7129 2024-06-06 Vladimir Serbinenko <[email protected]>
7130
7131 fs/zfs/zfs: Mark vdev_zaps_v2 and head_errlog as supported
7132 We don't need any actual adjustments as we don't use the affected structures.
7133
7134 Reviewed-by: Daniel Kiper <[email protected]>
7135
7136 2024-06-06 Vladimir Serbinenko <[email protected]>
7137
7138 types: Add missing casts in compile-time byteswaps
7139 Without them, e.g., 0x80LL on 64-bit target is 32-bit byte-swapped to
7140 0xffffffff80000000 instead of correct 0x80000000.
7141
7142 Reviewed-by: Daniel Kiper <[email protected]>
7143
7144 2024-06-06 Vladimir Serbinenko <[email protected]>
7145
7146 font: Add Fedora-specific font paths
7147 Reviewed-by: Daniel Kiper <[email protected]>
7148
7149 fs/bfs: Fix improper grub_free() on non-existing files
7150 Reviewed-by: Daniel Kiper <[email protected]>
7151
7152 2024-06-06 Daniel Axtens <[email protected]>
7153
7154 io/gzio: Properly init a table
7155 ARRAY_SIZE() is the count of elements, but the element size is 4 bytes, so
7156 this was only initing the first 1/4th of the table. Detected with valgrind.
7157
7158 This should only matter in error paths, and I've not been able to identify
7159 any actual misbehaviour that results from reading in-bounds but uninited data.
7160
7161 Reviewed-by: Daniel Kiper <[email protected]>
7162
7163 2024-06-06 Daniel Axtens <[email protected]>
7164
7165 io/gzio: Abort early when get_byte() reads nothing
7166 This isn't intended to be a functional change, but it makes a lot of failures a lot
7167 faster, which is extremely helpful for fuzzing.
7168
7169 Without this change, we keep trying and trying to read more bytes into our buffer,
7170 never being able to (read always returns 0) and so we just return old buffer contents
7171 over and over until the decompression process fails some other way.
7172
7173 Reviewed-by: Daniel Kiper <[email protected]>
7174
7175 2024-06-06 Alec Brown <[email protected]>
7176
7177 cli_lock: Add build option to block command line interface
7178 Add functionality to disable command line interface access and editing of GRUB
7179 menu entries if GRUB image is built with --disable-cli.
7180
7181 Reviewed-by: Vladimir Serbinenko <[email protected]>
7182 Reviewed-by: Daniel Kiper <[email protected]>
7183
7184 2024-05-23 Yifan Zhao <[email protected]>
7185
7186 fs/erofs: Add tests for EROFS in grub-fs-tester
7187 This patch introduces three EROFS tests which cover compact, extended
7188 and chunk-based inodes respectively.
7189
7190 Reviewed-by: Glenn Washburn <[email protected]>
7191 Reviewed-by: Daniel Kiper <[email protected]>
7192
7193 2024-05-23 Yifan Zhao <[email protected]>
7194
7195 fs/erofs: Add support for the EROFS
7196 The EROFS [1] is a lightweight read-only filesystem designed for performance
7197 which has already been shipped in most Linux distributions as well as widely
7198 used in several scenarios, such as Android system partitions, container
7199 images and rootfs for embedded devices.
7200
7201 This patch brings in the EROFS uncompressed support. Now, it's possible to
7202 boot directly through GRUB with an EROFS rootfs.
7203
7204 Support for the EROFS compressed files will be added later.
7205
7206 [1] https://erofs.docs.kernel.org
7207
7208 Tested-by: Daniel Axtens <[email protected]>
7209 Reviewed-by: Daniel Kiper <[email protected]>
7210
7211 2024-05-23 Gao Xiang <[email protected]>
7212
7213 safemath: Add ALIGN_UP_OVF() which checks for an overflow
7214 The following EROFS patch will use this helper to handle
7215 ALIGN_UP() overflow.
7216
7217 Reviewed-by: Daniel Kiper <[email protected]>
7218
7219 2024-05-23 Jonathan Davies <[email protected]>
7220
7221 docs: Fix spelling mistakes
7222 Reviewed-by: Daniel Kiper <[email protected]>
7223
7224 2024-05-23 Pascal Hambourg <[email protected]>
7225
7226 util/grub.d/00_header.in: Quote background image pathname in output
7227 This is required if the pathname contains spaces or GRUB shell
7228 metacharacters else the generated config file check will fail.
7229
7230 Reviewed-by: Daniel Kiper <[email protected]>
7231
7232 2024-05-23 Rogier <[email protected]>
7233
7234 disk/lvm: GRUB fails to detect LVM volumes due to an incorrect computation of mda_end
7235 When handling a regular LVM volume, GRUB can fail with the message:
7236
7237 error: disk `lvmid/******-****-****-****-****-****-****/******-****-****-****-****-****-******' not found.
7238
7239 If the condition which triggers this exists, grub-probe will report the
7240 error mentioned above. Similarly, the GRUB boot code will fail to detect
7241 LVM volumes, resulting in a failure to boot off of LVM disks/partitions.
7242 The condition can be created on any LVM VG by an LVM configuration change,
7243 so any system with /boot on LVM can become unbootable at "any" time (after
7244 any LVM configuration change).
7245
7246 The problem is caused by an incorrect computation of mda_end in disk/lvm.c,
7247 when the metadata area wraps around. Apparently, this can start happening at
7248 around 220 metadata changes to the VG.
7249
7250 Fixes: 879c4a834 (lvm: Fix two more potential data-dependent alloc overflows)
7251 Fixes: https://savannah.gnu.org/bugs/?61620
7252
7253 Reviewed-by: Daniel Kiper <[email protected]>
7254 Tested-By: Michael Chang <[email protected]>
7255
7256 2024-05-09 Forest <[email protected]>
7257
7258 disk/cryptodisk: Allow user to retry failed passphrase
7259 Give the user a chance to re-enter their cryptodisk passphrase after a typo,
7260 rather than immediately failing (and likely dumping them into a GRUB shell).
7261
7262 By default, we allow 3 tries before giving up. A value in the
7263 cryptodisk_passphrase_tries environment variable will override this default.
7264
7265 The user can give up early by entering an empty passphrase, just as they
7266 could before this patch.
7267
7268 Reviewed-by: Daniel Kiper <[email protected]>
7269
7270 2024-05-09 Lidong Chen <[email protected]>
7271
7272 disk/mdraid1x_linux: Prevent infinite recursion
7273 The test corpus for version-1 RAID generated an infinite recursion
7274 in grub_partition_iterate() while attempting to read the superblock.
7275 The reason for the issue was that the data region overlapped with
7276 the superblock.
7277
7278 The infinite call loop looks like this:
7279 grub_partition_iterate() -> partmap->iterate() ->
7280 -> grub_disk_read() -> grub_disk_read_small() ->
7281 -> grub_disk_read_small_real() -> grub_diskfilter_read() ->
7282 -> read_lv() -> read_segment() -> grub_diskfilter_read_node() ->
7283 -> grub_disk_read() -> grub_disk_read_small() -> ...
7284
7285 The fix adds checks for both the superblock region and the data
7286 region when parsing the superblock metadata in grub_mdraid_detect().
7287
7288 Reviewed-by: Daniel Kiper <[email protected]>
7289
7290 2024-05-09 Ard Biesheuvel <[email protected]>
7291
7292 efi: Fix stack protector issues
7293 The "ground truth" stack protector cookie value is kept in a global
7294 variable, and loaded in every function prologue and epilogue to store
7295 it into resp. compare it with the stack slot holding the cookie.
7296
7297 If the comparison fails, the program aborts, and this might occur
7298 spuriously when the global variable changes values between the entry and
7299 exit of a function. This implies that assigning the global variable at
7300 boot should not involve any instrumented function calls, unless special
7301 care is taken to ensure that the live call stack is synchronized, which
7302 is non-trivial.
7303
7304 So avoid any function calls, including grub_memcpy(), which is
7305 unnecessary given that the stack cookie is always a suitably aligned
7306 variable of the native word size.
7307
7308 While at it, leave the last byte 0x0 to avoid inadvertent unbounded
7309 strings on the stack.
7310
7311 Note that the use of __attribute__((optimize)) is described as
7312 unsuitable for production use in the GCC documentation, so let's drop
7313 this as well now that it is no longer needed.
7314
7315 Reviewed-by: Daniel Kiper <[email protected]>
7316
7317 2024-05-09 Oliver Steffen <[email protected]>
7318
7319 build: Track explicit module dependencies in Makefile.core.def
7320 Add a new keyword, "depends", to the module definition syntax
7321 used in Makefile.core.def. This allows specifying explicit module
7322 dependencies together with the module definition.
7323
7324 Do not track the "extra_deps.lst" file in the repository anymore,
7325 it is now auto-generated.
7326
7327 Make use of this new keyword in the bli module definition.
7328
7329 Reviewed-by: Daniel Kiper <[email protected]>
7330
7331 2024-04-11 Daniel Kiper <[email protected]>
7332
7333 windows: Add _stack_chk_guard/_stack_chk_fail symbols for Windows 64-bit target
7334 Otherwise the GRUB cannot start due to missing symbols when stack
7335 protector is enabled on EFI platforms.
7336
7337 Reviewed-by: Vladimir Serbinenko <[email protected]>
7338
7339 2024-04-11 Gary Lin <[email protected]>
7340
7341 util/bash-completion: Fix for bash-completion 2.12
7342 _split_longopt() was the bash-completion private API and removed since
7343 bash-completion 2.12. This commit initializes the bash-completion
7344 general variables with _init_completion() to avoid the potential
7345 "command not found" error.
7346
7347 Although bash-completion 2.12 introduces _comp_initialize() to deprecate
7348 _init_completion(), _init_completion() is still chosen for the better
7349 backward compatibility.
7350
7351 Reviewed-by: Daniel Kiper <[email protected]>
7352
7353 2024-04-11 Vladimir 'phcoder' Serbinenko <[email protected]>
7354
7355 util/grub-fstest: Add a new command zfs-bootfs
7356 It is useful to check zfs-bootfs command.
7357
7358 Reviewed-by: Daniel Kiper <[email protected]>
7359
7360 2024-04-11 Vladimir 'phcoder' Serbinenko <[email protected]>
7361
7362 efi: Enable CMOS on x86 EFI platforms
7363 The CMOS actually exists on most EFI platforms and in some cases is used to
7364 store useful data that makes it justifiable for GRUB to read/write it.
7365
7366 As for date and time keep using EFI API and not CMOS one.
7367
7368 Reviewed-by: Daniel Kiper <[email protected]>
7369
7370 2024-04-11 Vladimir 'phcoder' Serbinenko <[email protected]>
7371
7372 acpi: Mark MADT entries as packed
7373 No alignment is guaranteed and in fact on my IA-64 SAPIC is aligned
7374 to 4 bytes instead of 8 and causes a trap. It affects only rarely used
7375 lsacpi command and so went unnoticed.
7376
7377 Reviewed-by: Daniel Kiper <[email protected]>
7378
7379 2024-04-11 Michael Chang <[email protected]>
7380
7381 gfxmenu/view: Resolve false grub_errno disrupting boot process
7382 When enabling gfxmenu and choosing to boot the Xen hypervisor from its
7383 menu, an error occurred:
7384
7385 error: ../../grub-core/video/bitmap_scale.c:42:null src bitmap in grub_video_create_scaled.
7386
7387 The error is returned by grub_video_bitmap_create_scaled() when the
7388 source pixmap is not there. The init_background() uses it to scale up
7389 the background image so it can fully fit into the screen resolution.
7390
7391 However not all backgrounds are set by a image, i.e. the "desktop-image"
7392 property of the theme file. Instead a color code may be used, for
7393 example OpenSUSE's green background uses "desktop-color" property:
7394
7395 desktop-color: "#0D202F"
7396
7397 So it is absolutely fine to call init_background() without a raw pixmap
7398 if color code is used. A missing check has to be added to ensure the
7399 grub_errno will not be erroneously set and gets in the way of ensuing
7400 boot process.
7401
7402 The reason it happens sporadically is due to grub_errno is reset to
7403 GRUB_ERR_NONE in other places if a function's error return can be
7404 ignored. In particular this hunk in grub_gfxmenu_create_box() does the
7405 majority of the reset of grub_errno returned by init_background(), but
7406 the path may not be always chosen.
7407
7408 grub_video_bitmap_load (&box->raw_pixmaps[i], path);
7409 grub_free (path);
7410
7411 /* Ignore missing pixmaps. */
7412 grub_errno = GRUB_ERR_NONE;
7413
7414 In any case, we cannot account on such random behavior and should only
7415 return grub_errno if it is justified.
7416
7417 On the occasion move the grub_video_bitmap struct definition to the
7418 beginning of the function.
7419
7420 Reviewed-by: Daniel Kiper <[email protected]>
7421
7422 2024-04-11 Jon DeVree <[email protected]>
7423
7424 fs/xfs: Handle non-continuous data blocks in directory extents
7425 The directory extent list does not have to be a continuous list of data
7426 blocks. When GRUB tries to read a non-existant member of the list,
7427 grub_xfs_read_file() will return a block of zero'ed memory. Checking for
7428 a zero'ed magic number is sufficient to skip this non-existant data block.
7429
7430 Prior to commit 07318ee7e (fs/xfs: Fix XFS directory extent parsing)
7431 this was handled as a subtle side effect of reading the (non-existant)
7432 tail data structure. Since the block was zero'ed the computation of the
7433 number of directory entries in the block would return 0 as well.
7434
7435 Fixes: 07318ee7e (fs/xfs: Fix XFS directory extent parsing)
7436 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=2254370
7437
7438 Reviewed-By: Vladimir Serbinenko <[email protected]>
7439 Reviewed-by: Daniel Kiper <[email protected]>
7440
7441 2024-02-15 Julian Andres Klode <[email protected]>
7442
7443 Revert "templates: Reinstate unused version comparison functions with warning"
7444 We reinstated these functions before the 2.12 release with a warning
7445 such that users upgrading to 2.12 who had custom scripts using them
7446 would not get broken in the upgrade and agreed to remove them after
7447 the 2.12 release. This removes them accordingly.
7448
7449 This reverts commit e7a831963 (templates: Reinstate unused version
7450 comparison functions with warning).
7451
7452 Cc: Mathieu Desnoyers <[email protected]>
7453 Cc: Daniel Kiper <[email protected]>
7454 Reviewed-by: Daniel Kiper <[email protected]>
7455
7456 2024-02-15 Gary Lin <[email protected]>
7457
7458 util/bash-completion: Load scripts on demand
7459 There are two system directories for bash-completion scripts. One is
7460 /usr/share/bash-completion/completions/ and the other is
7461 /etc/bash_completion.d/. The "etc" scripts are loaded in advance and
7462 for backward compatibility while the "usr" scripts are loaded on demand.
7463 To load scripts on demand it requires a corresponding script for every
7464 command. So, the main bash-completion script is split into several
7465 subscripts for different "grub-*" commands. To share the code the real
7466 completion functions are still implemented in "grub" and each
7467 subscript sources "grub" and invokes the corresponding function.
7468
7469 Reviewed-by: Daniel Kiper <[email protected]>
7470
7471 2024-01-25 Samuel Thibault <[email protected]>
7472
7473 util/grub.d/10_hurd.in: Find proper ld.so on 64-bit systems
7474 The 64-bit ABI defines ld.so to be /lib/ld-x86-64.so.1.
7475
7476 Reviewed-by: Daniel Kiper <[email protected]>
7477
7478 2024-01-25 Samuel Thibault <[email protected]>
7479
7480 osdep/hurd/getroot: Fix 64-bit build
7481 The file_get_fs_options() takes a mach_msg_type_number_t, 32-bit,
7482 not a size_t, 64-bit on 64-bit platforms.
7483
7484 Reviewed-by: Vladimir Serbinenko <[email protected]>
7485 Reviewed-by: Daniel Kiper <[email protected]>
7486
7487 2024-01-25 Alec Brown <[email protected]>
7488
7489 loader/i386/multiboot_mbi: Clean up redundant code
7490 In grub-core/loader/i386/multiboot_mbi.c, Coverity spotted redundant code where
7491 the variable err was being set to GRUB_ERR_NONE and then being overwritten
7492 later without being used. Since this is unnecessary, we can remove the code
7493 that sets err to GRUB_ERR_NONE.
7494
7495 Fixes: CID 428877
7496
7497 Reviewed-by: Daniel Kiper <[email protected]>
7498
7499 2024-01-25 Alec Brown <[email protected]>
7500
7501 osdep/unix/getroot: Clean up redundant code
7502 In grub-core/osdep/unix/getroot.c, Coverity spotted redundant code where the
7503 double pointer os_dev was being set to 0 and then being overwritten later
7504 without being used. Since this is unnecessary, we can remove the code that
7505 sets os_dev to 0.
7506
7507 Fixes: CID 428875
7508
7509 Reviewed-by: Daniel Kiper <[email protected]>
7510
7511 2024-01-25 Alec Brown <[email protected]>
7512
7513 fs/jfs: Clean up redundant code
7514 In grub-core/fs/jfs.c, Coverity spotted redundant code where the pointer diro
7515 was being set to 0 and then being overwritten later without being used. Since
7516 this is unnecessary, we can remove the code that sets diro to 0.
7517
7518 Fixes: CID 428876
7519
7520 Reviewed-by: Daniel Kiper <[email protected]>
7521
7522 2024-01-25 Gary Lin <[email protected]>
7523
7524 tests: Switch password quality check off for luks2 test
7525 When adding/changing the password for the luks2 partition, cryptsetup
7526 may reject the command due to the weak password. Since this is only for
7527 testing, add "--force-password" to switch password quality check off to
7528 avoid the unexpected failure.
7529
7530 Reviewed-by: Glenn Washburn <[email protected]>
7531 Reviewed-by: Daniel Kiper <[email protected]>
7532
7533 2023-12-22 Oskari Pirhonen <[email protected]>
7534
7535 build: Include grub-core/extra_deps.lst in dist
7536 Fixes build failure due to the extra_deps.lst file not existing in the
7537 tarball. Found while trying to package GRUB 2.12 for Gentoo.
7538
7539 make[3]: *** No rule to make target '/var/tmp/portage/sys-boot/grub-2.12/work/grub-2.12/grub-core/extra_deps.lst', needed by 'syminfo.lst'. Stop.
7540
7541 Fixes: 89fbe0cac (grub-core/Makefile.am: Make path to extra_deps.lst relative to $(top_srcdir)/grub-core)
7542 Fixes: 154dcb1ae (build: Allow explicit module dependencies)
7543
7544 Reviewed-by: Daniel Kiper <[email protected]>
7545
7546 2023-12-20 Daniel Kiper <[email protected]>
7547
7548 Bump version to 2.13
7549
7550 Release 2.12
7551
7552 2023-12-20 Glenn Washburn <[email protected]>
7553
7554 efi: Add support for reproducible builds
7555 Having randomly generated bytes in the binary output breaks reproducible
7556 builds. Since build timestamps are usually the source of irreproducibility
7557 there is a standard which defines an environment variable SOURCE_DATE_EPOCH
7558 to be used when set for build timestamps. According to the standard [1], the
7559 value of SOURCE_DATE_EPOCH is a base-10 integer of the number of seconds
7560 since the UNIX epoch. Currently, this is a 10 digit number that fits into
7561 32-bits, but will not shortly after the year 2100. So to be future-proof
7562 only use the least significant 32-bits. On 64-bit architectures, where the
7563 canary is also 64-bits, there is an extra 32-bits that can be filled to
7564 provide more entropy. The first byte is NUL to filter out string buffer
7565 overflow attacks and the remaining 24-bits are set to static random bytes.
7566
7567 [1] https://reproducible-builds.org/specs/source-date-epoch
7568
7569 Reviewed-by: Daniel Kiper <[email protected]>
7570
7571 2023-12-20 Glenn Washburn <[email protected]>
7572
7573 efi: Generate stack protector canary at build time if urandom is available
7574 Generating the canary at build time allows the canary to be different for
7575 every build which could limit the effectiveness of certain exploits.
7576 Fallback to the statically generated random bytes if /dev/urandom is not
7577 readable, e.g. Windows.
7578
7579 On 32-bit architectures, which use a 32-bit canary, reduce the canary to
7580 4 bytes with one byte being NUL to filter out string buffer overflow attacks.
7581
7582 Reviewed-by: Daniel Kiper <[email protected]>
7583
7584 2023-12-20 Glenn Washburn <[email protected]>
7585
7586 efi: Initialize canary to non-zero value
7587 The canary, __stack_chk_guard, is in the BSS and so will get initialized to
7588 zero if it is not explicitly initialized. If the UEFI firmware does not
7589 support the RNG protocol, then the canary will not be randomized and will
7590 be zero. This seems like a possibly easier value to write by an attacker.
7591 Initialize canary to static random bytes, so that it is still random when
7592 there is no RNG protocol. Set at least one byte to NUL to protect against
7593 string buffer overflow attacks [1]. Code that writes NUL terminated strings
7594 will terminate when a NUL is encountered in the input byte stream. So the
7595 attacker will not be able to forge the canary by including it in the input
7596 stream without terminating the string operation and thus limiting the
7597 stack corruption.
7598
7599 [1] https://www.sans.org/blog/stack-canaries-gingerly-sidestepping-the-cage/
7600
7601 Reviewed-by: Daniel Kiper <[email protected]>
7602
7603 2023-12-14 Alec Brown <[email protected]>
7604
7605 gfxmenu/gui_image: Fix double free of bitmap
7606 In grub-core/gfxmenu/gui_image.c, Coverity detected a double free in the
7607 function load_image(). The function checks if self->bitmap and self->raw_bitmap
7608 aren't NULL and then frees them. In the case self->bitmap and self->raw_bitmap
7609 are the same, only self->raw_bitmap is freed which would also free the memory
7610 used by self->bitmap. However, in this case self->bitmap isn't being set to NULL
7611 which could lead to a double free later in the code. After self->raw_bitmap is
7612 freed, it gets set to the variable bitmap. If this variable is NULL, the code
7613 could have a path that would free self->bitmap a second time in the function
7614 rescale_image().
7615
7616 Fixes: CID 292472
7617
7618 Reviewed-by: Daniel Kiper <[email protected]>
7619
7620 2023-12-13 Qiumiao Zhang <[email protected]>
7621
7622 commands/acpi: Fix calculation of ACPI tables addresses when processing RSDT and XSDT
7623 According to the ACPI specification the XSDT Entry field contains an array
7624 of 64-bit physical addresses which points to other DESCRIPTION_HEADERs. However,
7625 the entry_ptr iterator is defined as a 32-bit pointer. It means each 64-bit
7626 entry in the XSDT table is treated as two separate 32-bit entries then. Fix the
7627 issue by using correct addresses sizes when processing RSDT and XSDT tables.
7628
7629 Reviewed-by: Daniel Kiper <[email protected]>
7630
7631 2023-12-13 Vladimir Serbinenko <[email protected]>
7632
7633 libnvpair: Support prefixed nvlist symbol names as found on NetBSD
7634 NetBSD uses slightly different function names for the same functions.
7635
7636 Reviewed-by: Daniel Kiper <[email protected]>
7637
7638 2023-12-13 Vladimir Serbinenko <[email protected]>
7639
7640 bootstrap: Don't check gettext version
7641 NetBSD gettext is older than the check but we don't actually need 0.18.3,
7642 older one works fine. This is needed to make bootstrap work on NetBSD.
7643
7644 Reviewed-by: Daniel Kiper <[email protected]>
7645
7646 2023-12-13 Vladimir Serbinenko <[email protected]>
7647
7648 kern/mm: Use %x and cast for displaying sizeof()
7649 There is some variance in how compiler treats sizeof() especially
7650 on 32-bit platforms where it can be naturally either int or long.
7651 Explicit cast solves the issue.
7652
7653 Reviewed-by: Daniel Kiper <[email protected]>
7654
7655 2023-12-13 Vladimir Serbinenko <[email protected]>
7656
7657 configure: Add RPATH for freetype on NetBSD
7658 Without this build-time mkfont fails dynamic linking. This is not ideal
7659 but improves the situation until a better solution is available.
7660
7661 Reviewed-by: Daniel Kiper <[email protected]>
7662
7663 2023-12-13 Vladimir Serbinenko <[email protected]>
7664
7665 configure: Add *BSD font paths
7666 *BSD puts fonts in other places. Add them to the list.
7667
7668 Reviewed-by: Daniel Kiper <[email protected]>
7669
7670 2023-12-13 Vladimir Serbinenko <[email protected]>
7671
7672 autogen: Accept python3.10 as a python alternative
7673 NetBSD doesn't provide python or python3.
7674
7675 Reviewed-by: Daniel Kiper <[email protected]>
7676
7677 2023-12-12 Vladimir Serbinenko <[email protected]>
7678
7679 build: Rename HAVE_LIBZFS to USE_LIBZFS
7680 The HAVE_LIBZFS is defined by libzfs test and hence conflicts with
7681 manual definition. On NetBSD it ends up detecting zfs but not detecting
7682 nvpair and creates confusion. Split them.
7683
7684 Reviewed-by: Daniel Kiper <[email protected]>
7685
7686 2023-12-12 Vladimir Serbinenko <[email protected]>
7687
7688 gnulib: Tolerate always_inline attribute being ignored
7689 It's not critical, -Werror on it is inappropriate. We don't want to
7690 modify gnulib too much. This warning is pretty much irrelevant.
7691
7692 Reviewed-by: Daniel Kiper <[email protected]>
7693
7694 2023-12-12 Vladimir Serbinenko <[email protected]>
7695
7696 util/editenv: Don't use %m formatter
7697 It's not available on NetBSD outside of syslog. Using strerror() is more
7698 reliable as we retrieve errno immediately rather than down the stack.
7699
7700 Reviewed-by: Daniel Kiper <[email protected]>
7701
7702 2023-12-12 Vladimir Serbinenko <[email protected]>
7703
7704 osdep/bsd/hostdisk: Fix NetBSD compilation
7705 Wrong function and variable name cause a stupid compilation error on
7706 NetBSD and OpenBSD. Only NetBSD and OpenBSD use this file. No other
7707 platform is affected.
7708
7709 Additionally, define RAW_FLOPPY_MAJOR constant if it is missing.
7710
7711 Reviewed-by: Daniel Kiper <[email protected]>
7712
7713 2023-12-12 Vladimir Serbinenko <[email protected]>
7714
7715 osdep/generic/blocklist: Fix compilation
7716 After recent change in blocklist types we have a type mismatch. Fixing it
7717 requires a wrapper or large changes. I feel like wrapper makes more sense.
7718
7719 Without this patch we end up with a compilation problem and without wrapping
7720 callback data is not passed properly anymore.
7721
7722 Reviewed-by: Daniel Kiper <[email protected]>
7723
7724 2023-12-12 Vladimir Serbinenko <[email protected]>
7725
7726 disk/diskfilter: Remove unused variable
7727 Variable e is set but never used. We can just remove it now.
7728
7729 Reviewed-by: Daniel Kiper <[email protected]>
7730
7731 2023-12-12 Vladimir Serbinenko <[email protected]>
7732
7733 build: Tolerate unused-but-set in generated lexer/bison files
7734 We don't really control the small aspects of generated files and NetBSD
7735 version has an unused variable that is then detected by gcc as warning
7736 that is then promoted to error.
7737
7738 Reviewed-by: Daniel Kiper <[email protected]>
7739
7740 2023-12-12 Vladimir Serbinenko <[email protected]>
7741
7742 loader/i386/bsdXX: Fix loading after unaligned module
7743 Current code implicitly assumes that aligning chunk_size + *kern_end is
7744 the same as aligning on curload which is not the case because
7745 chunk_size starts at zero even if *kern_end is unaligned and ALIGN_PAGE
7746 moved curload to an aligned position but not *kern_end + chunk_size.
7747
7748 This fixes booting of FreeBSD with zfs module.
7749
7750 Reviewed-by: Daniel Kiper <[email protected]>
7751
7752 2023-12-12 Mate Kukri <[email protected]>
7753
7754 grub-core/Makefile.am: Make path to extra_deps.lst relative to $(top_srcdir)/grub-core
7755 The commit 154dcb1ae (build: Allow explicit module dependencies) broke
7756 out of tree builds by introducing the extra_deps.lst file into the
7757 source tree but referencing it just by name in grub-core/Makefile.am.
7758 Fix it by adding $(top_srcdir)/grub-core to the path.
7759
7760 Fixes: 154dcb1ae (build: Allow explicit module dependencies)
7761
7762 Reviewed-by: Daniel Kiper <[email protected]>
7763
7764 2023-12-12 Mate Kukri <[email protected]>
7765
7766 util/grub-install: Move platdir path canonicalization after files were copied to grubdir
7767 The commit 3f9eace2d (util/grub-install: Delay copying files to
7768 {grubdir,platdir} after install_device was validated) delaying
7769 copying of files caused a regression when installing without an
7770 existing directory structure.
7771
7772 This patch ensures that the platform directory actually exists by the
7773 time the code tries to canonicalize its filename.
7774
7775 Fixes: 3f9eace2d (util/grub-install: Delay copying files to {grubdir,platdir} after install_device was validated)
7776
7777 Reviewed-by: Daniel Kiper <[email protected]>
7778
7779 2023-12-12 Michael Chang <[email protected]>
7780
7781 util/grub-mkstandalone: Ensure deterministic tar file creation by sorting contents
7782 The add_tar_files() function currently iterates through a directory's
7783 content using readdir(), which doesn't guarantee a specific order. This
7784 lack of deterministic behavior impacts reproducibility in the build process.
7785
7786 This commit resolves the issue by introducing sorting functionality.
7787 The list retrieved by readdir() is now sorted alphabetically before
7788 incorporation into the tar archive, ensuring consistent and predictable
7789 file ordering within the archive.
7790
7791 On the occasion fix tfp memory leak.
7792
7793 Reviewed-by: Daniel Kiper <[email protected]>
7794
7795 2023-12-12 Michael Chang <[email protected]>
7796
7797 util/grub-mkstandalone: Ensure stable timestamps for generated images
7798 This change mirrors a previous fix [1] but is specific to images
7799 generated by grub-mkstandalone.
7800
7801 The former fix, commit 85a7be241 (util/mkimage: Use stable timestamp
7802 when generating binaries.), focused on utilizing a stable timestamp
7803 during binary generation in the util/mkimage context. This commit
7804 extends that approach to the images produced by grub-mkstandalone,
7805 ensuring consistency and stability in timestamps across all generated
7806 binaries.
7807
7808 [1] 85a7be241 util/mkimage: Use stable timestamp when generating binaries.
7809
7810 Reviewed-by: Daniel Kiper <[email protected]>
7811
7812 2023-12-05 Mate Kukri <[email protected]>
7813
7814 net/http: Fix gcc-13 errors relating to type signedness
7815 Replace definition of HTTP_PORT with a pre-processor macro that converts
7816 the constant to the correct grub_uint16_t type.
7817
7818 Change "port" local variable definition in http_establish() to have the
7819 same type.
7820
7821 Reviewed-by: Daniel Kiper <[email protected]
7822
7823 2023-12-05 Julian Andres Klode <[email protected]>
7824
7825 templates: Reinstate unused version comparison functions with warning
7826 Revert the commit a79c567f6 (templates: Remove unused version comparison
7827 functions) and add a warning to the functions that they are deprecated.
7828
7829 Removing the functions directly caused a lot of upgrade issues
7830 with custom user scripts that called the functions. In Debian and
7831 Ubuntu, grub-mkconfig is invoked as a post-installation script
7832 and would fail, causing upgrades to fail halfway through and
7833 putting the package manager into an inconsistent state.
7834
7835 FWIW, we get one bug per 2 weeks basically, for an interim Ubuntu
7836 release which generally does not receive much usage, that is a high
7837 number.
7838
7839 The proposal is to pick this for 2.12 and directly after the release
7840 remove it again. Then users will have time to fix their scripts without
7841 systems breaking immediately.
7842
7843 This reverts commit a79c567f6 (templates: Remove unused version
7844 comparison functions).
7845
7846 Cc: Mathieu Desnoyers <[email protected]>
7847 Cc: Daniel Kiper <[email protected]>
7848 Reviewed-by: Daniel Kiper <[email protected]>
7849
7850 2023-12-05 Mate Kukri <[email protected]>
7851
7852 util/grub-install: Delay copying files to {grubdir,platdir} after install_device was validated
7853 Previously grub-install copied modules to grubdir before doing any
7854 validation on the install_device.
7855
7856 When grub-install was called with an invalid install_device, modules
7857 were already copied to /boot before it found out and was forced to rely
7858 on atexit() rollback.
7859
7860 This patch delays copying the modules after at least some install_device
7861 validation was done, and thus reduces reliance on successful rollback.
7862
7863 Reviewed-by: Daniel Kiper <[email protected]>
7864
7865 2023-12-05 Julian Andres Klode <[email protected]>
7866
7867 efi: Set shim_lock_enabled even if validation is disabled
7868 If validation has been disabled via MokSbState, secure boot on the
7869 firmware is still enabled, and the kernel fails to boot.
7870
7871 This is a bit hacky, because shim_lock is not *fully* enabled, but
7872 it triggers the right code paths.
7873
7874 Ultimately, all this will be resolved by shim gaining it's own image
7875 loading and starting protocol, so this is more a temporary workaround.
7876
7877 Fixes: 6425c12cd (efi: Fallback to legacy mode if shim is loaded on x86 archs)
7878
7879 Cc: Peter Jones <[email protected]>
7880 Cc: Michael Chang <[email protected]>
7881 Reviewed-by: Daniel Kiper <[email protected]>
7882
7883 2023-12-05 Oliver Steffen <[email protected]>
7884
7885 docs: Improve bli module documentation
7886 Improve the documentation of the bli module and explain in more detail what
7887 it does. Make clear that GPT formatted drives are expected and other
7888 partition formats are ignored. Also reorder and reword this section a bit.
7889
7890 Reviewed-by: Daniel Kiper <[email protected]>
7891
7892 2023-12-05 Oliver Steffen <[email protected]>
7893
7894 bli: Add explicit dependency on the part_gpt module
7895 The bli module has a "hidden" dependency on the part_gpt module, which
7896 is not picked up automatically by the build system. One purpose of the
7897 bli module is to communicate the GPT UUID of the partition GRUB was
7898 launched from to Linux user-space (systemd-gpt-auto-generator).
7899 Without the part_gpt module, bli is not able to obtain the UUID. Since
7900 bli does its work in the module initialization function, the order in
7901 which the modules are loaded is also important: part_gpt needs to be
7902 loaded before the bli module.
7903
7904 To solve this, track this dependency explicitly.
7905
7906 Note that the Boot Loader Interface specification, which bli aims to
7907 implement, requires GPT formatted drives. The bli module ignores all
7908 other partition formats.
7909
7910 Reviewed-by: Daniel Kiper <[email protected]>
7911
7912 2023-12-05 Oliver Steffen <[email protected]>
7913
7914 build: Allow explicit module dependencies
7915 The build system deduces inter-module dependencies from the symbols
7916 required and exported by the modules. This works well, except for some
7917 rare cases where the dependency is indirect or hidden. A module might
7918 not make use of any function of some other module, but still expect its
7919 functionality to be available to GRUB.
7920
7921 To solve this, introduce a new file, currently empty, called extra_deps.lst
7922 to track these cases manually. This file gets processed in the same way
7923 as the automatically generated syminfo.lst, making it possible to inject
7924 data into the dependency resolver.
7925
7926 Since *.lst files are set to be ignored by git, add an exception for
7927 extra_deps.lst.
7928
7929 Additionally, introduce a new keyword for the syminfo.lst syntax:
7930 "depends" allows specifying a module dependency directly:
7931
7932 depends <module> <depdendency>...
7933
7934 Reviewed-by: Daniel Kiper <[email protected]>
7935
7936 2023-12-05 Stefan Berger <[email protected]>
7937
7938 kern/ieee1275/init/ppc64: Display upper_mem_limit when debugging
7939 Display upper_mem_limit and its rounded-down value in MiB.
7940
7941 Reviewed-by: Daniel Kiper <[email protected]>
7942
7943 2023-12-05 Stefan Berger <[email protected]>
7944
7945 kern/ieee1275/init/ppc64: Fix a comment
7946 Reviewed-by: Daniel Kiper <[email protected]>
7947
7948 2023-12-05 Stefan Berger <[email protected]>
7949
7950 kern/ieee1275/ieee1275: Display successful memory claims when debugging
7951 Display successful memory claims with exact address and rounded-down
7952 MiB location and rounded-up size in MiB.
7953
7954 Reviewed-by: Daniel Kiper <[email protected]>
7955 Cc: Eric Snowberg <[email protected]>
7956 Cc: Hari Bathini <[email protected]>
7957 Cc: Pavithra Prakash <[email protected]>
7958 Cc: Michael Ellerman <[email protected]>
7959 Cc: Carolyn Scherrer <[email protected]>
7960 Cc: Mahesh Salgaonkar <[email protected]>
7961 Cc: Sourabh Jain <[email protected]>
7962
7963 2023-12-05 Stefan Berger <[email protected]>
7964
7965 loader/powerpc/ieee1275: Use new allocation function for kernel and initrd
7966 On PowerVM and KVM on Power use the new memory allocation function that
7967 honors restrictions on which memory GRUB can actually use. In the request
7968 structure indicate the request for a single memory block along with
7969 address alignment restrictions. Request direct usage of the memory block
7970 by setting init_region to false (prevent it from being added to GRUB's
7971 heap). Initialize the found addr to -1, so that -1 will be returned
7972 to the loader in case no memory could be allocated.
7973
7974 Report an out-of-memory error in case the initrd could not be loaded.
7975
7976 Reviewed-by: Daniel Kiper <[email protected]>
7977 Cc: Hari Bathini <[email protected]>
7978 Cc: Pavithra Prakash <[email protected]>
7979 Cc: Michael Ellerman <[email protected]>
7980 Cc: Carolyn Scherrer <[email protected]>
7981 Cc: Mahesh Salgaonkar <[email protected]>
7982 Cc: Sourabh Jain <[email protected]>
7983
7984 2023-12-05 Stefan Berger <[email protected]>
7985
7986 kern/ieee1275/cmain/ppc64: Introduce flags to identify KVM and PowerVM
7987 Introduce flags to identify PowerVM and KVM on Power and set them where
7988 each type of host has been detected.
7989
7990 Reviewed-by: Daniel Kiper <[email protected]>
7991 Cc: Hari Bathini <[email protected]>
7992 Cc: Pavithra Prakash <[email protected]>
7993 Cc: Michael Ellerman <[email protected]>
7994 Cc: Carolyn Scherrer <[email protected]>
7995 Cc: Mahesh Salgaonkar <[email protected]>
7996 Cc: Sourabh Jain <[email protected]>
7997
7998 2023-12-05 Stefan Berger <[email protected]>
7999
8000 kern/ieee1275/init/ppc64: Rename regions_claim() to grub_regions_claim()
8001 Rename regions_claim() to grub_regions_claim() to make it available for
8002 memory allocation. The ieee1275 loader will use this function on PowerVM
8003 and KVM on Power and thus avoid usage of memory that it is not allowed
8004 to use.
8005
8006 Reviewed-by: Daniel Kiper <[email protected]>
8007 Cc: Hari Bathini <[email protected]>
8008 Cc: Pavithra Prakash <[email protected]>
8009 Cc: Michael Ellerman <[email protected]>
8010 Cc: Carolyn Scherrer <[email protected]>
8011 Cc: Mahesh Salgaonkar <[email protected]>
8012 Cc: Sourabh Jain <[email protected]>
8013
8014 2023-12-05 Stefan Berger <[email protected]>
8015
8016 kern/ieee1275/init/ppc64: Add support for alignment requirements
8017 Add support for memory alignment requirements and adjust a candidate
8018 address to it before checking whether the block is large enough. This
8019 must be done in this order since the alignment adjustment can make
8020 a block smaller than what was requested.
8021
8022 None of the current callers has memory alignment requirements but the
8023 ieee1275 loader for kernel and initrd will use it to convey them.
8024
8025 Reviewed-by: Daniel Kiper <[email protected]>
8026 Cc: Hari Bathini <[email protected]>
8027 Cc: Pavithra Prakash <[email protected]>
8028 Cc: Michael Ellerman <[email protected]>
8029 Cc: Carolyn Scherrer <[email protected]>
8030 Cc: Mahesh Salgaonkar <[email protected]>
8031 Cc: Sourabh Jain <[email protected]>
8032
8033 2023-12-05 Stefan Berger <[email protected]>
8034
8035 kern/ieee1275/init/ppc64: Return allocated address using context
8036 Return the allocated address of the memory block in the request structure
8037 if a memory allocation was actually done. Leave the address untouched
8038 otherwise. This enables a caller who wants to use the allocated memory
8039 directly, rather than adding the memory to the heap, to see where memory
8040 was allocated. None of the current callers need this but the converted
8041 ieee1275 loader will make use of it.
8042
8043 Reviewed-by: Daniel Kiper <[email protected]>
8044 Cc: Hari Bathini <[email protected]>
8045 Cc: Pavithra Prakash <[email protected]>
8046 Cc: Michael Ellerman <[email protected]>
8047 Cc: Carolyn Scherrer <[email protected]>
8048 Cc: Mahesh Salgaonkar <[email protected]>
8049 Cc: Sourabh Jain <[email protected]>
8050
8051 2023-12-05 Stefan Berger <[email protected]>
8052
8053 kern/ieee1275/init/ppc64: Decide by request whether to initialize region
8054 Let the regions_claim() request structure's init_region determine whether
8055 to call grub_mm_init_region() on it. This allows for adding memory to
8056 GRUB's memory heap if init_region is set to true, or direct usage of the
8057 memory otherwise. Set all current callers' init_region to true since they
8058 want to add memory regions to GRUB's heap.
8059
8060 Reviewed-by: Daniel Kiper <[email protected]>
8061 Cc: Hari Bathini <[email protected]>
8062 Cc: Pavithra Prakash <[email protected]>
8063 Cc: Michael Ellerman <[email protected]>
8064 Cc: Carolyn Scherrer <[email protected]>
8065 Cc: Mahesh Salgaonkar <[email protected]>
8066 Cc: Sourabh Jain <[email protected]>
8067
8068 2023-12-05 Stefan Berger <[email protected]>
8069
8070 kern/ieee1275/init/ppc64: Introduce a request for regions_claim()
8071 The regions_claim() function limits the allocation of memory regions
8072 by excluding certain memory areas from being used by GRUB. This for
8073 example includes a gap between 640MB and 768MB as well as an upper
8074 limit beyond which no memory may be used when an fadump is present.
8075 However, the ieee1275 loader for kernel and initrd currently does not
8076 use regions_claim() for memory allocation on PowerVM and KVM on Power
8077 and therefore may allocate memory in those areas that it should not use.
8078
8079 To make the regions_claim() function more flexible and ultimately usable
8080 for the ieee1275 loader, introduce a request structure to pass various
8081 parameters to the regions_claim() function that describe the properties
8082 of requested memory chunks. In a first step, move the total and flags
8083 variables into this structure.
8084
8085 Reviewed-by: Daniel Kiper <[email protected]>
8086 Cc: Hari Bathini <[email protected]>
8087 Cc: Pavithra Prakash <[email protected]>
8088 Cc: Michael Ellerman <[email protected]>
8089 Cc: Carolyn Scherrer <[email protected]>
8090 Cc: Mahesh Salgaonkar <[email protected]>
8091 Cc: Sourabh Jain <[email protected]>
8092
8093 2023-11-22 Anthony Iliopoulos <[email protected]>
8094
8095 fs/xfs: Add large extent counters incompat feature support
8096 XFS introduced 64-bit extent counters for inodes via a series of
8097 upstream commits and the feature was marked as stable in v6.5 via
8098 commit 61d7e8274cd8 (xfs: drop EXPERIMENTAL tag for large extent
8099 counts).
8100
8101 Further, xfsprogs release v6.5.0 switched this feature on by default
8102 in mkfs.xfs via commit e5b18d7d1d96 (mkfs: enable large extent counts
8103 by default).
8104
8105 Filesystems formatted with large extent count support, nrext64=1, are
8106 thus currently not recognizable by GRUB, since this is an incompat
8107 feature. Add the required support so that those filesystems and inodes
8108 with large extent counters can be read by GRUB.
8109
8110 Reviewed-by: Andrey Albershteyn <[email protected]>
8111 Reviewed-by: Daniel Kiper <[email protected]>
8112 Tested-by: Marta Lewandowska <[email protected]>
8113 Tested-by: Sebastian Andrzej Siewior <[email protected]>
8114
8115 2023-11-08 Vladimir Serbinenko <[email protected]>
8116
8117 gpt: Add compile time asserts for guid and gpt_partentry sizes
8118 With new alignment specification it's easy to screw up. Fortunately if it
8119 happens the size will be bigger than intended. Compile time assert will catch
8120 this.
8121
8122 Reviewed-by: Daniel Kiper <[email protected]>
8123
8124 2023-11-08 Vladimir Serbinenko <[email protected]>
8125
8126 types: Split aligned and packed guids
8127 On ia64 alignment requirements are strict. When we pass a pointer to
8128 UUID it needs to be at least 4-byte aligned or EFI will crash.
8129 On the other hand in device path there is no padding for UUID, so we
8130 need 2 types in one formor another. Make 4-byte aligned and unaligned types
8131
8132 The code is structured in a way to accept unaligned inputs
8133 in most cases and supply 4-byte aligned outputs.
8134
8135 Efiemu case is a bit ugly because there inputs and outputs are
8136 reversed and so we need careful casts to account for this
8137 inversion.
8138
8139 Reviewed-by: Daniel Kiper <[email protected]>
8140
8141 2023-11-06 Vladimir Serbinenko <[email protected]>
8142
8143 gpt_partition: Mark grub_gpt_partentry as having natural alignment
8144 gpt_partition contains grub_guid. We need to decide whether the whole
8145 structure is unaligned and then we need to use packed_guid. But we never
8146 have unaligned part entries as we read them in an aligned buffer from disk.
8147 Hence just make it all aligned.
8148
8149 2023-11-06 Vladimir Serbinenko <[email protected]>
8150
8151 efi: Deduplicate configuration table search function
8152 We do table search in many places doing exactly the same algorithm.
8153 The only minor variance in users is which table is used if several entries
8154 are present. As specification mandates uniqueness and even if it ever isn't,
8155 first entry is good enough, unify this code and always use the first entry.
8156
8157 Reviewed-by: Daniel Kiper <[email protected]>
8158
8159 2023-11-06 Vladimir Serbinenko <[email protected]>
8160
8161 lsefi: Add missing static qualifier
8162 known_protocols isn't used anywhere else and even misses grub_ prefix, so
8163 let's make it local (static).
8164
8165 Reviewed-by: Daniel Kiper <[email protected]>
8166
8167 2023-11-06 Vladimir Serbinenko <[email protected]>
8168
8169 types: Fix typo
8170 Just a small grammar mistake.
8171
8172 Reviewed-by: Daniel Kiper <[email protected]>
8173
8174 2023-10-30 Qiumiao Zhang <[email protected]>
8175
8176 util/grub-mount: Check file path sanity
8177 The function argp_parser() in util/grub-mount.c lacks a check on the
8178 sanity of the file path when parsing parameters. This results in
8179 a segmentation fault if a partition is mounted to a non-existent path.
8180
8181 Reviewed-by: Daniel Kiper <[email protected]>
8182
8183 2023-10-30 Richard Marko <[email protected]>
8184
8185 configure: Make the DJVU_FONT_SOURCE configurable with --with-dejavufont=FILE
8186 Font might be located in different location, the default font might
8187 not be available on all systems or other font might be preferred.
8188
8189 Reviewed-by: Daniel Kiper <[email protected]>
8190
8191 2023-10-30 Mads Kiilerich <[email protected]>
8192
8193 configure: Make the Unifont FONT_SOURCE configurable with --with-unifont=FILE
8194 Font might be located in different location, the default font might
8195 not be available on all systems or other font might be preferred.
8196
8197 Reviewed-by: Daniel Kiper <[email protected]>
8198
8199 2023-10-30 Jon DeVree <[email protected]>
8200
8201 fs/xfs: Fix XFS directory extent parsing
8202 The XFS directory entry parsing code has never been completely correct
8203 for extent based directories. The parser correctly handles the case
8204 where the directory is contained in a single extent, but then mistakenly
8205 assumes the data blocks for the multiple extent case are each identical
8206 to the single extent case. The difference in the format of the data
8207 blocks between the two cases is tiny enough that its gone unnoticed for
8208 a very long time.
8209
8210 A recent change introduced some additional bounds checking into the XFS
8211 parser. Like GRUB's existing parser, it is correct for the single extent
8212 case but incorrect for the multiple extent case. When parsing a directory
8213 with multiple extents, this new bounds checking is sometimes (but not
8214 always) tripped and triggers an "invalid XFS directory entry" error. This
8215 probably would have continued to go unnoticed but the /boot/grub/<arch>
8216 directory is large enough that it often has multiple extents.
8217
8218 The difference between the two cases is that when there are multiple
8219 extents, the data blocks do not contain a trailer nor do they contain
8220 any leaf information. That information is stored in a separate set of
8221 extents dedicated to just the leaf information. These extents come after
8222 the directory entry extents and are not included in the inode size. So
8223 the existing parser already ignores the leaf extents.
8224
8225 The only reason to read the trailer/leaf information at all is so that
8226 the parser can avoid misinterpreting that data as directory entries. So
8227 this updates the parser as follows:
8228
8229 For the single extent case the parser doesn't change much:
8230 1. Read the size of the leaf information from the trailer
8231 2. Set the end pointer for the parser to the start of the leaf
8232 information. (The previous bounds checking set the end pointer to the
8233 start of the trailer, so this is actually a small improvement.)
8234 3. Set the entries variable to the expected number of directory entries.
8235
8236 For the multiple extent case:
8237 1. Set the end pointer to the end of the block.
8238 2. Do not set up the entries variable. Figuring out how many entries are
8239 in each individual block is complex and does not seem worth it when
8240 it appears to be safe to just iterate over the entire block.
8241
8242 The bounds check itself was also dependent upon the faulty XFS parser
8243 because it accidentally used "filename + length - 1". Presumably this
8244 was able to pass the fuzzer because in the old parser there was always
8245 8 bytes of slack space between the tail pointer and the actual end of
8246 the block. Since this is no longer the case the bounds check needs to be
8247 updated to "filename + length + 1" in order to prevent a regression in
8248 the handling of corrupt fliesystems.
8249
8250 Notes:
8251 * When there is only one extent there will only ever be one block. If
8252 more than one block is required then XFS will always switch to holding
8253 leaf information in a separate extent.
8254 * B-tree based directories seems to be parsed properly by the same code
8255 that handles multiple extents. This is unlikely to ever occur within
8256 /boot though because its only used when there are an extremely large
8257 number of directory entries.
8258
8259 Fixes: ef7850c75 (fs/xfs: Fix issues found while fuzzing the XFS filesystem)
8260 Fixes: b2499b29c (Adds support for the XFS filesystem.)
8261 Fixes: https://savannah.gnu.org/bugs/?64376
8262
8263 Reviewed-by: Daniel Kiper <[email protected]>
8264 Tested-by: Sebastian Andrzej Siewior <[email protected]>
8265 Tested-by: Marta Lewandowska <[email protected]>
8266
8267 2023-10-30 Lidong Chen <[email protected]>
8268
8269 fs/xfs: Incorrect short form directory data boundary check
8270 After parsing of the current entry, the entry pointer is advanced
8271 to the next entry at the end of the "for" loop. In case where the
8272 last entry is at the end of the data boundary, the advanced entry
8273 pointer can point off the data boundary. The subsequent boundary
8274 check for the advanced entry pointer can cause a failure.
8275
8276 The fix is to include the boundary check into the "for" loop
8277 condition.
8278
8279 Reviewed-by: Daniel Kiper <[email protected]>
8280 Tested-by: Sebastian Andrzej Siewior <[email protected]>
8281 Tested-by: Marta Lewandowska <[email protected]>
8282
8283 2023-10-12 Vladimir 'phcoder' Serbinenko <[email protected]>
8284
8285 Revert "zfsinfo: Correct a check for error allocating memory"
8286 Original commit is wrong because grub_file_get_device_name() may return NULL
8287 if we use implicit $root. Additionally, the grub_errno is guaranteed to be
8288 GRUB_ERR_NONE at the beginning of a command. So, everything should work as
8289 expected and Coverity report, CID 73668, WRT to this code should be treated
8290 as false positive.
8291
8292 This reverts commit 7aab03418 (zfsinfo: Correct a check for error allocating memory).
8293
8294 Fixes: 7aab03418 (zfsinfo: Correct a check for error allocating memory)
8295
8296 Reviewed-by: Daniel Kiper <[email protected]>
8297
8298 2023-10-12 ValdikSS <[email protected]>
8299
8300 disk/i386/pc/biosdisk: Read up to 63 sectors in LBA mode
8301 Current code imposes limitations on the amount of sectors read in
8302 a single call according to CHS layout of the disk even in LBA
8303 read mode. There's no need to obey CHS layout restrictions for
8304 LBA reads on LBA disks. It only slows down booting process.
8305
8306 See: https://lore.kernel.org/grub-devel/[email protected]/
8307
8308 Reviewed-by: Daniel Kiper <[email protected]>
8309
8310 2023-10-12 ValdikSS <[email protected]>
8311
8312 kern/i386/pc/init: Flush cache only on VIA C3 and earlier
8313 The code flushes the cache on VIA processors unconditionally which
8314 is excessive. Check for cpuid family and execute wbinvd only on C3
8315 and earlier.
8316
8317 Fixes: https://savannah.gnu.org/bugs/?45149
8318 Fixes: 25492a0f0 (Add wbinvd around bios call.)
8319
8320 Reviewed-by: Daniel Kiper <[email protected]>
8321
8322 2023-10-12 Fabian Vogt <[email protected]>
8323
8324 fs/btrfs: Zero file data not backed by extents
8325 Implicit holes in file data need to be zeroed explicitly, instead of
8326 just leaving the data in the buffer uninitialized.
8327
8328 This led to kernels randomly failing to boot in "fun" ways when loaded
8329 from btrfs with the no_holes feature enabled, because large blocks of
8330 zeros in the kernel file contained random data instead.
8331
8332 Reviewed-by: Daniel Kiper <[email protected]>
8333 Reviewed-by: Qu Wenruo <[email protected]>
8334
8335 2023-10-12 Stefan Berger <[email protected]>
8336
8337 kern/ieee1275/init: Restrict high memory in presence of fadump on ppc64
8338 When a kernel dump is present then restrict the high memory regions to
8339 avoid allocating memory where the kernel dump resides. Use the
8340 ibm,kernel-dump node under /rtas to determine whether a kernel dump
8341 exists and up to which limit GRUB can use available memory. Set the
8342 upper_mem_limit to the size of the kernel dump section of type
8343 REAL_MODE_REGION and therefore only allow GRUB's memory usage for high
8344 addresses from RMO_ADDR_MAX to upper_mem_limit. This means that GRUB can
8345 use high memory in the range of RMO_ADDR_MAX (768MB) to upper_mem_limit
8346 and the kernel-dump memory regions above upper_mem_limit remain
8347 untouched. This change has no effect on memory allocations below
8348 linux_rmo_save (typically at 640MB).
8349
8350 Also, fall back to allocating below rmo_linux_save in case the chunk of
8351 memory there would be larger than the chunk of memory above RMO_ADDR_MAX.
8352 This can for example occur if a free memory area is found starting at 300MB
8353 extending up to 1GB but a kernel dump is located at 768MB and therefore
8354 does not allow the allocation of the high memory area but requiring to use
8355 the chunk starting at 300MB to avoid an unnecessary out-of-memory condition.
8356
8357 Reviewed-by: Hari Bathini <[email protected]>
8358 Cc: Pavithra Prakash <[email protected]>
8359 Cc: Michael Ellerman <[email protected]>
8360 Cc: Carolyn Scherrer <[email protected]>
8361 Cc: Mahesh Salgaonkar <[email protected]>
8362 Cc: Sourabh Jain <[email protected]>
8363 Reviewed-by: Daniel Kiper <[email protected]>
8364
8365 2023-10-12 Glenn Washburn <[email protected]>
8366
8367 tests/util/grub-shell: Enable RNG device to better test stack smashing
8368 In certain firmwares, e.g. OVMF, the RNG protocol is not enabled unless
8369 there is an RNG device. When not enabled, GRUB fails to initialize the
8370 stack guard with random bytes. For testing, this is not a big issue, but
8371 there have been bugs found in the initialization. So turn this on for EFI
8372 platforms to catch any regressions.
8373
8374 Reviewed-by: Daniel Kiper <[email protected]>
8375
8376 2023-10-12 Glenn Washburn <[email protected]>
8377
8378 kern/efi/init: Disable stack smashing protection on grub_efi_init()
8379 GCC is electing to instrument grub_efi_init() to give it stack smashing
8380 protection when configuring with --enable-stack-protector on the x86_64-efi
8381 target. In the function prologue, the canary at the top of the stack frame
8382 is set to the value of the stack guard. And in the epilogue, the canary is
8383 checked to verify if it is equal to the guard and if not to call the stack
8384 check fail function. The issue is that grub_efi_init() sets up the guard
8385 by initializing it with random bytes, if the firmware supports the RNG
8386 protocol. So in its prologue the canary will be set with the value of the
8387 uninitialized guard, likely NUL bytes. Then the guard is initialized, and
8388 finally the epilogue checks the canary against the guard, which will almost
8389 certainly be different. This causes the code path for a smashed stack to be
8390 taken, causing the machine to print out a message that stack smashing was
8391 detected, wait 5 seconds, and then reboot. Disable grub_efi_init()
8392 instrumentation so there is no stack smashing false positive generated.
8393
8394 Reviewed-by: Daniel Kiper <[email protected]>
8395
8396 2023-10-12 Glenn Washburn <[email protected]>
8397
8398 disk/cryptodisk: Add support for LUKS2 in (proc)/luks_script
8399 The sector size in bytes is added to each line and it is allowed to be
8400 6 decimal digits long, which covers the most common cases of 512 and 4096
8401 byte sectors with space for two additional digits as future-proofing. The
8402 size allocation is updated to reflect this additional field. Also make
8403 clearer the size allocation calculation.
8404
8405 Reviewed-by: Daniel Kiper <[email protected]>
8406
8407 2023-10-12 Glenn Washburn <[email protected]>
8408
8409 disk/cryptodisk: Optimize luks_script_get()
8410 Use the return value of grub_snprintf() to move the string pointer forward,
8411 instead of incrementing the string pointer iteratively until a NULL byte is
8412 reached. Move the space out of the format string argument, a small
8413 optimization, but also makes the spacing clearer. Also, use the new
8414 PRIxGRUB_OFFSET instead of PRIuGRUB_UINT64_T to accurately reflect the
8415 format string for this type.
8416
8417 Reviewed-by: Daniel Kiper <[email protected]>
8418
8419 2023-10-12 Glenn Washburn <[email protected]>
8420
8421 term/serial: Ensure proper NULL termination after grub_strncpy()
8422 A large enough argument to the --port option could cause a string buffer
8423 to be not NULL terminated because grub_strncpy() does not guarantee NULL
8424 termination if copied string is longer than max characters to copy.
8425
8426 Fixes: 712309eaae04 (term/serial: Use grub_strncpy() instead of grub_snprintf() when only copying string)
8427
8428 Reviewed-by: Daniel Kiper <[email protected]>
8429
8430 2023-10-12 Heinrich Schuchardt <[email protected]>
8431
8432 commands/efi/lsefisystab: Print the UEFI specification revision in human readable form
8433 E.g. 2.10 instead of 00020064 and 2.3.1 instead of 0002001f.
8434
8435 See UEFI 2.10 specification, chapter 4.2.1 EFI_TABLE_HEADER.
8436
8437 Reviewed-by: Daniel Kiper <[email protected]>
8438
8439 2023-10-03 Maxim Suhanov <[email protected]>
8440
8441 fs/ntfs: Make code more readable
8442 Move some calls used to access NTFS attribute header fields into
8443 functions with human-readable names.
8444
8445 Suggested-by: Daniel Kiper <[email protected]>
8446 Reviewed-by: Daniel Kiper <[email protected]>
8447
8448 2023-10-03 Maxim Suhanov <[email protected]>
8449
8450 fs/ntfs: Fix an OOB read when parsing a volume label
8451 This fix introduces checks to ensure that an NTFS volume label is always
8452 read from the corresponding file record segment.
8453
8454 The current NTFS code allows the volume label string to be read from an
8455 arbitrary, attacker-chosen memory location. However, the bytes read are
8456 always treated as UTF-16LE. So, the final string displayed is mostly
8457 unreadable and it can't be easily converted back to raw bytes.
8458
8459 The lack of this check is a minor issue, likely not causing a significant
8460 data leak.
8461
8462 Reported-by: Maxim Suhanov <[email protected]>
8463 Reviewed-by: Daniel Kiper <[email protected]>
8464
8465 2023-10-03 Maxim Suhanov <[email protected]>
8466
8467 fs/ntfs: Fix an OOB read when parsing bitmaps for index attributes
8468 This fix introduces checks to ensure that bitmaps for directory indices
8469 are never read beyond their actual sizes.
8470
8471 The lack of this check is a minor issue, likely not exploitable in any way.
8472
8473 Reported-by: Maxim Suhanov <[email protected]>
8474 Reviewed-by: Daniel Kiper <[email protected]>
8475
8476 2023-10-03 Maxim Suhanov <[email protected]>
8477
8478 fs/ntfs: Fix an OOB read when parsing directory entries from resident and non-resident index attributes
8479 This fix introduces checks to ensure that index entries are never read
8480 beyond the corresponding directory index.
8481
8482 The lack of this check is a minor issue, likely not exploitable in any way.
8483
8484 Reported-by: Maxim Suhanov <[email protected]>
8485 Reviewed-by: Daniel Kiper <[email protected]>
8486
8487 2023-10-03 Maxim Suhanov <[email protected]>
8488
8489 fs/ntfs: Fix an OOB read when reading data from the resident $DATA attribute
8490 When reading a file containing resident data, i.e., the file data is stored in
8491 the $DATA attribute within the NTFS file record, not in external clusters,
8492 there are no checks that this resident data actually fits the corresponding
8493 file record segment.
8494
8495 When parsing a specially-crafted file system image, the current NTFS code will
8496 read the file data from an arbitrary, attacker-chosen memory offset and of
8497 arbitrary, attacker-chosen length.
8498
8499 This allows an attacker to display arbitrary chunks of memory, which could
8500 contain sensitive information like password hashes or even plain-text,
8501 obfuscated passwords from BS EFI variables.
8502
8503 This fix implements a check to ensure that resident data is read from the
8504 corresponding file record segment only.
8505
8506 Fixes: CVE-2023-4693
8507
8508 Reported-by: Maxim Suhanov <[email protected]>
8509 Reviewed-by: Daniel Kiper <[email protected]>
8510
8511 2023-10-03 Maxim Suhanov <[email protected]>
8512
8513 fs/ntfs: Fix an OOB write when parsing the $ATTRIBUTE_LIST attribute for the $MFT file
8514 When parsing an extremely fragmented $MFT file, i.e., the file described
8515 using the $ATTRIBUTE_LIST attribute, current NTFS code will reuse a buffer
8516 containing bytes read from the underlying drive to store sector numbers,
8517 which are consumed later to read data from these sectors into another buffer.
8518
8519 These sectors numbers, two 32-bit integers, are always stored at predefined
8520 offsets, 0x10 and 0x14, relative to first byte of the selected entry within
8521 the $ATTRIBUTE_LIST attribute. Usually, this won't cause any problem.
8522
8523 However, when parsing a specially-crafted file system image, this may cause
8524 the NTFS code to write these integers beyond the buffer boundary, likely
8525 causing the GRUB memory allocator to misbehave or fail. These integers contain
8526 values which are controlled by on-disk structures of the NTFS file system.
8527
8528 Such modification and resulting misbehavior may touch a memory range not
8529 assigned to the GRUB and owned by firmware or another EFI application/driver.
8530
8531 This fix introduces checks to ensure that these sector numbers are never
8532 written beyond the boundary.
8533
8534 Fixes: CVE-2023-4692
8535
8536 Reported-by: Maxim Suhanov <[email protected]>
8537 Reviewed-by: Daniel Kiper <[email protected]>
8538
8539 2023-10-03 Michael Chang <[email protected]>
8540
8541 kern/acpi: Skip NULL entries in RSDT and XSDT
8542 During attempts to configure a serial console, a Page Fault Exception
8543 and system reset were encountered, specifically on release 2.12~rc1.
8544 This issue was not present in prior versions and seemed to affect only
8545 a specific machine, potentially pointing to hardware or firmware flaw.
8546
8547 After investigation, it was discovered that the invalid page access
8548 occurred during the discovery of serial MMIO ports as specified by
8549 ACPI's SPCR table [1]. The recent change uncovered an issue in GRUB's
8550 ACPI driver.
8551
8552 In certain cases, the XSDT/RSDT root table might contain a NULL entry as
8553 a terminator, depending on how the tables are assembled. GRUB cannot
8554 blindly trust the address in the root table to be valid and should
8555 perform a sanity check for NULL entries. This patch introduces this
8556 simple check.
8557
8558 This fix is also inspired by a related Linux kernel fix [2].
8559
8560 [1] 7b192ec4c term/ns8250: Use ACPI SPCR table when available to configure serial
8561 [2] 0f929fbf0 ACPICA: Tables: Add new mechanism to skip NULL entries in RSDT and XSDT.
8562
8563 Reviewed-by: Daniel Kiper <[email protected]>
8564
8565 2023-10-03 Glenn Washburn <[email protected]>
8566
8567 util/grub-install-common: Print usable grub-mkimage command
8568 When grub-install is run with the verbose option, it will print a log
8569 message indicating the grub-mkimage command and arguments used.
8570 GRUB no longer calls the grub-mkimage binary internally, however the
8571 command logged is a command that if run should effectively be what
8572 grub-install used. However, as this has changed some of the newer
8573 options have been incorrectly added so that the printed command fails
8574 when run separately. This change makes the displayed command run as
8575 intended.
8576
8577 Reviewed-by: Daniel Kiper <[email protected]>
8578
8579 2023-10-03 Glenn Washburn <[email protected]>
8580
8581 util/grub-install-common: Minor improvements to printing of grub-mkimage command
8582 This is a preparatory patch to make the following patch less cluttered. The
8583 only visible change made here is to not print extra spaces when either or
8584 both --note or --disable-shim-lock are not given and to not print an extra
8585 space at the end of the command. The latter is done by constructing the
8586 trailing argument string with spaces in front of each argument rather than
8587 trailing. The allocation of the argument string is made precise, which has
8588 the benefit of saving a few bytes, but more importantly self-documenting
8589 what the needed allocated bytes are. Also, unneeded braces are removed from
8590 an if block.
8591
8592 Reviewed-by: Daniel Kiper <[email protected]>
8593
8594 2023-10-03 Vladimir 'phcoder' Serbinenko <[email protected]>
8595
8596 lib/i386/relocator64: Fix 64-bit FreeBSD boot on BIOS
8597 The commit 80948f532d (lib/i386/relocator64: Build fixes for i386) has
8598 broken 64-bit FreeBSD boot on BIOS. This patch fixes the issue.
8599
8600 Fixes: 80948f532d (lib/i386/relocator64: Build fixes for i386)
8601
8602 Reviewed-by: Daniel Kiper <[email protected]>
8603
8604 2023-09-22 Anthony PERARD <[email protected]>
8605
8606 templates/linux_xen: Fix XSM entries generation
8607 It turns out that setting $xen_version in linux_entry_xsm() override
8608 $xen_version in the loop over $reverse_sorted_xen_list. This means
8609 that only one entry per Xen version is going to enable XSM, but all
8610 further entries are going to have "(XSM enabled)" in their titles
8611 without enabling XSM.
8612
8613 When a "xenpolicy-$xen_version" file was found for the current
8614 $xen_version, it would overwrite $xen_version to add "(XSM enabled)" to
8615 the menu entry title. Once updated, the next call to linux_entry_xsm()
8616 would also have this modified $xen_version and would look for the file
8617 "xenpolicy-*(XSM enabled)" and fail.
8618
8619 Reviewed-by: Daniel Kiper <[email protected]>
8620
8621 2023-09-22 Xiaotian Wu <[email protected]>
8622
8623 loongarch: Eliminate cmodel compilation warnings
8624 In the configure phase, the "-mcmodel=large" CFLAGS passed the test, but
8625 because it has not been implemented in gcc, the following warning will
8626 appear when compiling:
8627
8628 gcc: warning: 'large' is not supported, now cmodel is set to 'normal'
8629
8630 Reviewed-by: Daniel Kiper <[email protected]>
8631
8632 2023-09-22 Glenn Washburn <[email protected]>
8633
8634 configure: Enable -fno-omit-frame-pointer for backtrace module
8635 The backtrace module is written assuming that the frame pointer is in %ebp.
8636 By default, -Os optimization level is used, which enables the gcc option
8637 -fomit-frame-pointer. This breaks the backtrace functionality. Enabling
8638 this may cause an unnoticeable performance cost and virtually no size increase.
8639
8640 The backtrace command on x86_64 and probably i386 is broken due to the
8641 above rationale. I've not verified, but presumably the backtrace that used
8642 to be printed for an unhandled CPU exception is also broken. Do any distros
8643 handle this?
8644
8645 Considering that, to my knowledge, no one has complained about this in the
8646 over 13 years that -Os has been used, has this code actually been useful?
8647 Is it worth disabling -fomit-frame-pointer? Though, I don't see much downside
8648 right now in disabling it. Alternatively, we could disable/remove the
8649 backtrace code. I think it would be nice to keep it and have it working.
8650
8651 Nowadays, presumably QEMU makes the GDB stub rarely used as I imagine most
8652 are developing in a virtual machines. Also, the GDB stub does not work in UEFI.
8653 So, if anyone is using it on real hardware, they are doing so on pretty old
8654 machines. The lack of a GDB stub does not seem to be a pain point because
8655 no one has got it working on UEFI.
8656
8657 This patch gets the backtrace command working on x86_64-efi in QEMU for me.
8658 However, it hangs when run on my laptop. Not sure what's going on there.
8659
8660 Reviewed-by: Daniel Kiper <[email protected]>
8661
8662 2023-09-22 Ard Biesheuvel <[email protected]>
8663
8664 loader/efi/linux: Implement x86 mixed mode using legacy boot
8665 Recent mixed-mode Linux kernels, i.e., v4.0 or newer, can access EFI
8666 runtime services at OS runtime even when the OS was not entered via the
8667 EFI stub. This is because, instead of reverting back to the firmware's
8668 segment selectors, GDTs and IDTs, the 64-bit kernel simply calls 32-bit
8669 runtime services using compatibility mode, i.e., the same mode used for
8670 32-bit user space, without taking down all interrupt handling, exception
8671 handling, etc.
8672
8673 This means that GRUB's legacy x86 boot mode is sufficient to make use of
8674 this: 32-bit i686 builds of GRUB can already boot 64-bit kernels in EFI
8675 enlightened mode, but without going via the EFI stub, and provide all
8676 the metadata that the OS needs to map the EFI runtime regions and call
8677 EFI runtime services successfully.
8678
8679 It does mean that GRUB should not attempt to invoke the firmware's
8680 LoadImage()/StartImage() methods on kernel builds that it knows cannot
8681 be started natively. So, add a check for this in the native EFI boot
8682 path and fall back to legacy x86 mode in such cases.
8683
8684 Note that in the general case, booting non-native images of the same
8685 native word size, e.g., x64 EFI apps on arm64 firmware, might be
8686 supported by means of emulation. So, let's only disallow images that use
8687 a non-native word size. This will also permit booting i686 kernels on
8688 x86_64 builds, although without access to runtime services, as this is
8689 not supported by Linux.
8690
8691 This change on top of 2.12-rc1 is sufficient to boot ordinary Linux
8692 mixed mode builds and get full access to the EFI runtime services.
8693
8694 Cc: Daniel Kiper <[email protected]>
8695 Cc: Steve McIntyre <[email protected]>
8696 Cc: Julian Andres Klode <[email protected]>
8697 Acked-by: Dimitri John Ledkov <[email protected]>
8698 Reviewed-by: Daniel Kiper <[email protected]>
8699
8700 2023-09-22 Ard Biesheuvel <[email protected]>
8701
8702 loader/i386/linux: Prefer entry in long mode when booting via EFI
8703 The x86_64 Linux kernel can be booted in 32-bit mode, in which case the
8704 startup code creates a set of preliminary page tables that map the first
8705 4 GiB of physical memory 1:1 and enables paging. This is a prerequisite
8706 for 64-bit execution and can therefore only be implemented in 32-bit code.
8707
8708 The x86_64 Linux kernel can also be booted in 64-bit mode directly: this
8709 implies that paging is already enabled and it is the responsibility of
8710 the bootloader to ensure that the active page tables cover the entire
8711 loaded image, including its BSS space, the size of which is described in
8712 the image's setup header.
8713
8714 Given that the EFI spec mandates execution in long mode for x86_64 and
8715 stipulates that all system memory is mapped 1:1, the Linux/x86
8716 requirements for 64-bit entry can be met trivially when booting on
8717 x86_64 via EFI. So, enter via the 64-bit entry point in this case.
8718
8719 This involves inspecting the xloadflags field in the setup header to
8720 check whether the 64-bit entry point is supported. This field was
8721 introduced in Linux version v3.8 (early 2013).
8722
8723 This change ensures that all EFI firmware tables and other assets passed
8724 by the firmware or bootloader in memory remain mapped and accessible
8725 throughout the early startup code.
8726
8727 Avoiding the drop out of long mode will also be needed to support
8728 upcoming CPU designs that no longer implement 32-bit mode at all
8729 (as recently announced by Intel [0]).
8730
8731 [0] https://www.intel.com/content/www/us/en/developer/articles/technical/envisioning-future-simplified-architecture.html
8732
8733 Cc: Daniel Kiper <[email protected]>
8734 Cc: Julian Andres Klode <[email protected]>
8735 Reviewed-by: Daniel Kiper <[email protected]>
8736
8737 2023-09-18 Vladimir Serbinenko <[email protected]>
8738
8739 ZFS: Check bonustype in addition to dnode type
8740 Some dnodes are shared with properties zap. This is used
8741 e.g. for quotas. Then dnode type is 0xc4 and GRUB stumbles on
8742 this. Check bonus type and if it's ok then ignore dnode type mismatch
8743
8744 Reviewed-by: Daniel Kiper <[email protected]>
8745
8746 2023-09-18 Vladimir Serbinenko <[email protected]>
8747
8748 ZFS: Don't iterate over null objsets
8749 Reading them is harmless but useless as they are empty by definition
8750
8751 Reviewed-by: Daniel Kiper <[email protected]>
8752
8753 2023-09-18 Vladimir Serbinenko <[email protected]>
8754
8755 ZFS: Fix invalid memcmp
8756 We ended up comparing over unset values as we had dnode_phys on one side
8757 and dnode on another
8758
8759 Reviewed-by: Daniel Kiper <[email protected]>
8760
8761 2023-09-18 Vladimir Serbinenko <[email protected]>
8762
8763 ZFS: support inode type embed into its ID
8764 This is a speedup used in some ZFS version. This trips GRUB and makes it
8765 unable to access directories. Just skip it for now and revisit
8766 if we ever need this speedup.
8767
8768 Reviewed-by: Daniel Kiper <[email protected]>
8769
8770 2023-08-31 Heinrich Schuchardt <[email protected]>
8771
8772 video/efi_gop: Require shadow if PixelBltOnly
8773 If the EFI graphics pixel format is PixelBltOnly, we cannot write directly
8774 to the frame buffer. We need the shadow frame buffer which we copy via
8775 the BitBlt operation to the hardware.
8776
8777 If the pixel format is PixelBltOnly and allocation of the shadow frame
8778 buffer fails, we must raise an error to signal that the EFI GOP protocol
8779 is not usable.
8780
8781 Reviewed-by: Daniel Kiper <[email protected]>
8782
8783 2023-08-31 Glenn Washburn <[email protected]>
8784
8785 docs: Add menu to prevent older makeinfo versions from failing
8786 It has been reported that makeinfo version 4.13a complains and returns
8787 error when menus for chapter structuring commands are not present. It
8788 is also known that newer makeinfos, such as version 6.7, will create
8789 default menus when needed. Since the menu will be created regardless,
8790 explicitly create it to support older makeinfo versions. This also
8791 enables building to be successful when an older makeinfo is installed
8792 because in that case info files are attempted to be generated with the
8793 "all" target.
8794
8795 Reported-by: Olaf Hering <[email protected]>
8796 Reviewed-by: Daniel Kiper <[email protected]>
8797 Tested-by: Olaf Hering <[email protected]>
8798
8799 2023-08-31 Glenn Washburn <[email protected]>
8800
8801 docs: Use @ref instead of @xref
8802 The @xref command is meant to be used at the beginning of a sentence
8803 because its expansion creates a "See " prefix on all output formats, and
8804 on older makeinfo versions is strict about enforcing a "." or "," after
8805 the command. The @ref command has no such restriction and is just the
8806 link, which allows more control over output. This also fixes an issue
8807 where there was a repeated "see" in the output.
8808
8809 Reported-by: Olaf Hering <[email protected]>
8810 Reviewed-by: Daniel Kiper <[email protected]>
8811 Tested-by: Olaf Hering <[email protected]>
8812
8813 2023-08-31 Glenn Washburn <[email protected]>
8814
8815 tests/util/grub-shell-luks-tester: Allow setting timeout
8816 Allow using the envvar GRUB_SHELL_LUKS_TIMEOUT to change the default
8817 timeout. If not specified, use value of GRUB_SHELL_DEFAULT_TIMEOUT. And
8818 if that is not specified, fallback to original 600s timeout.
8819
8820 Reviewed-by: Daniel Kiper <[email protected]>
8821
8822 2023-08-31 Glenn Washburn <[email protected]>
8823
8824 disk/cryptodisk: Fix missing change when updating to use grub_uuidcasecmp()
8825 This was causing the cryptomount command to return failure even though
8826 the crypto device was successfully added. Of course, this meant that any
8827 script using the return code would behave unexpectedly.
8828
8829 Fixes: 3cf2e848bc03 (disk/cryptodisk: Allows UUIDs to be compared in a dash-insensitive manner)
8830
8831 Suggested-by: Olaf Hering <[email protected]>
8832 Reviewed-by: Patrich Steinhardt <[email protected]>
8833 Reviewed-by: Daniel Kiper <[email protected]>
8834
8835 2023-08-31 Glenn Washburn <[email protected]>
8836
8837 kern/misc: Make grub_vsnprintf() C99/POSIX conformant
8838 To comply with C99 and POSIX standards, snprintf() should return the
8839 number of bytes that would be written to the string (excluding the
8840 terminating NUL byte) if the buffer size was big enough. Before this
8841 change, the return value was the minimum of the standard return and the
8842 length of the buffer. Rarely is the return value of grub_snprintf() or
8843 grub_vsnprintf() used with current code, and the few places where it is
8844 used do not need to be changed.
8845
8846 Reviewed-by: Daniel Kiper <[email protected]>
8847
8848 2023-08-31 Glenn Washburn <[email protected]>
8849
8850 tests: Add serial_test
8851 This test is meant to test output via various serial devices. Currently,
8852 only the PCI serial device is tested.
8853
8854 Reviewed-by: Daniel Kiper <[email protected]>
8855
8856 2023-08-31 Glenn Washburn <[email protected]>
8857
8858 tests/util/grub-shell: Allow explicitly using other serial ports for output
8859 While here, move "-qemu=*" case to be next to the "--qemu-opts=*" case.
8860 This causes no change in logic, but is more logically located.
8861
8862 Reviewed-by: Daniel Kiper <[email protected]>
8863
8864 2023-08-31 Glenn Washburn <[email protected]>
8865
8866 tests/util/grub-shell-luks-tester: Do not remove generated files when test fails to allow debugging
8867 Reviewed-by: Daniel Kiper <[email protected]>
8868
8869 tests/util/grub-shell: Convert spaces to TABs
8870 Reviewed-by: Daniel Kiper <[email protected]>
8871
8872 2023-08-14 Glenn Washburn <[email protected]>
8873
8874 commands/ls: Print "????????????" if unable to get file size
8875 In long list mode, if the file can not be opened, the file is not printed.
8876 Instead, print the file but print the size as "????????????".
8877
8878 Reviewed-by: Daniel Kiper <[email protected]>
8879
8880 2023-08-14 Glenn Washburn <[email protected]>
8881
8882 commands/ls: Send correct dirname to print functions
8883 For each non-directory path argument to the ls command, the full path was
8884 being sent to the print functions, instead of the dirname. The long output
8885 print function expected dirname to be the directory containing the file
8886 and so could not open the file to get the file size because the generated
8887 path was incorrect. This caused the output to be a blank line.
8888
8889 Reviewed-by: Daniel Kiper <[email protected]>
8890
8891 2023-08-14 Glenn Washburn <[email protected]>
8892
8893 fs/archelp: If path given to grub_archelp_dir() is not a directory return error
8894 Specifically, return GRUB_ERR_BAD_FILE_TYPE because this is what is
8895 expected by the ls command when it is given a path to a non-directory.
8896 This fixes a bug where calling ls with a list of non-directory paths
8897 outputs a blank line for each such argument.
8898
8899 Reviewed-by: Daniel Kiper <[email protected]>
8900
8901 2023-08-14 Glenn Washburn <[email protected]>
8902
8903 commands/videoinfo: Prevent crash when run while video driver already active
8904 The videoinfo command will initialize all non-active video adapters. Video
8905 drivers tend to zero out the global framebuffer object on initialization.
8906 This is not a problem when there is no active video adapter. However, when
8907 there is, then outputting to the video adapter will cause a crash because
8908 methods in the framebuffer object are reinitialized. For example, this
8909 command sequence will cause a crash.
8910
8911 terminal_output --append gfxterm; videoinfo
8912
8913 When running in a QEMU headless with GRUB built for the x86_64-efi target,
8914 the first command initializes the Bochs video adapter, which, among
8915 other things, sets the set_page() member function. Then when videoinfo is
8916 run, all non-Bochs video adapters will be initialized, each one wiping
8917 the framebuffer and thus setting set_page to NULL. Soon after the videoinfo
8918 command finishes there will be a call to grub_refresh(), which will
8919 ultimately call the framebuffer's set_page which will be NULL and cause
8920 a crash when called.
8921
8922 Reviewed-by: Daniel Kiper <[email protected]>
8923
8924 2023-08-14 Glenn Washburn <[email protected]>
8925
8926 docs: Improve initrd documentation
8927 A list of improvements:
8928 * Remove reference to "initial ramdisk" and replace with "initrd". This
8929 then covers the case of ramdisk and ramfs, which is the usual method
8930 with kernels 2.6 and newer.
8931 * Add sentence with URL to initrd documentation Linux kernel.
8932 * Add a section documenting how to have the initrd command generate
8933 a new-style initrd via a specially crafted argument and include an example.
8934 * Update initrd16 to refer to the initrd section and make note that
8935 initrd16 is only on the pc platform.
8936
8937 Reviewed-by: Oskari Pirhonen <[email protected]>
8938 Reviewed-by: Paul Menzel <[email protected]>
8939 Reviewed-by: Daniel Kiper <[email protected]>
8940
8941 2023-08-14 Glenn Washburn <[email protected]>
8942
8943 term/ns8250-spcr: Continue processing SPCR table even if revision is < 2
8944 According to commit 0231d00082 (ACPI: SPCR: Make SPCR available to x86)
8945 to the Linux kernel, "On x86, many systems have a valid SPCR table but the
8946 table version is not 2 so the table version check must be a warning."
8947
8948 Reviewed-by: Benjamin Herrenschmidt <[email protected]>
8949 Reviewed-by: Daniel Kiper <[email protected]>
8950
8951 2023-08-14 Glenn Washburn <[email protected]>
8952
8953 docs: A note to cat that hexdump should be used for binary data
8954 The cat command should not be used to print binary data because it can
8955 show bytes not in the binary data and not show bytes that are in the data,
8956 which can lead to confusion. This happens because cat does some processing
8957 of the data stream, namely trying to decode substrings as UTF-8.
8958
8959 Reviewed-by: Oskari Pirhonen <[email protected]>
8960 Reviewed-by: Daniel Kiper <[email protected]>
8961
8962 2023-08-14 Glenn Washburn <[email protected]>
8963
8964 docs: Document hexdump command
8965 Reviewed-by: Oskari Pirhonen <[email protected]>
8966 Reviewed-by: Daniel Kiper <[email protected]>
8967
8968 docs: Group usage of user-space utilities into single chapter
8969 Reviewed-by: Oskari Pirhonen <[email protected]>
8970 Reviewed-by: Daniel Kiper <[email protected]>
8971
8972 2023-08-14 Qiumiao Zhang <[email protected]>
8973
8974 util/grub-mount: Fix memory leak in fuse_getattr()
8975 Reviewed-by: Daniel Kiper <[email protected]>
8976
8977 2023-08-14 Michał Grzelak <[email protected]>
8978
8979 configure: Fix SDL2 typo by referencing value
8980 During configuration of SDL2, variable enable_grub_emu_sdl2 is checked
8981 whether to throw an error message. However, error could not happen
8982 because two unequal strings were compared. Fix this by referencing
8983 value of enable_grub_emu_sdl2, not name.
8984
8985 Fixes: 17d6ac1a7 (emu: Add SDL2 support)
8986
8987 Reviewed-by: Julian Andres Klode <[email protected]>
8988 Reviewed-by: Daniel Kiper <[email protected]>
8989 Reviewed-by: Paul Menzel <[email protected]>
8990
8991 2023-08-14 Glenn Washburn <[email protected]>
8992
8993 docs: Add missing assumption
8994 Also reword a prior sentence to be more clear.
8995
8996 Fixes: 5a3d2b4742df (docs: Add debugging chapter to development documentation)
8997
8998 Reviewed-by: Oskari Pirhonen <[email protected]>
8999 Reviewed-by: Daniel Kiper <[email protected]>
9000
9001 2023-08-14 Oskari Pirhonen <[email protected]>
9002
9003 util/grub.d/25_bli.in: Fix shebang on unmerged-usr
9004 On an unmerged-usr system, grub-mkconfig errors out with the following
9005 error due to /usr/bin/sh not existing:
9006
9007 /usr/sbin/grub-mkconfig: /etc/grub.d/25_bli: /usr/bin/sh: bad interpreter: No such file or directory
9008
9009 Use a /bin/sh shebang to fix the error as well as match the other
9010 existing files.
9011
9012 Fixes: 158a6583e (util/grub.d/25_bli.in: Activate bli module on EFI)
9013
9014 Reviewed-by: Glenn Washburn <[email protected]>
9015 Reviewed-by: Daniel Kiper <[email protected]>
9016 Reviewed-by: Oliver Steffen <[email protected]>
9017
9018 2023-08-14 Glenn Washburn <[email protected]>
9019
9020 tests/util/grub-shell-luks-tester: Allow GRUB_SHELL_LUKS_DEFAULT_DEBUG and GRUB_TEST_DEFAULT_DEBUG to specify the debug level to grub-shell
9021 Reviewed-by: Daniel Kiper <[email protected]>
9022
9023 2023-08-14 Glenn Washburn <[email protected]>
9024
9025 tests/util/grub-shell: Allow setting the value of debug regardless of its previous state
9026 This allows an invocation of grub-shell to set the value of debug regardless
9027 of the global default environment variable GRUB_SHELL_DEFAULT_DEBUG.
9028
9029 Reviewed-by: Daniel Kiper <[email protected]>
9030
9031 2023-08-14 Glenn Washburn <[email protected]>
9032
9033 tests/util/grub-shell: Allow setting default timeout via GRUB_SHELL_DEFAULT_TIMEOUT envvar
9034 Reviewed-by: Daniel Kiper <[email protected]>
9035
9036 2023-08-14 Glenn Washburn <[email protected]>
9037
9038 tests/util/grub-shell: Add --verbose to grub-mkrescue when $debug is greater than 2
9039 Since this is fairly verbose output, do not enable first level of debug
9040 is turned on.
9041
9042 Reviewed-by: Daniel Kiper <[email protected]>
9043
9044 2023-07-10 Daniel Kiper <[email protected]>
9045
9046 Release 2.12~rc1
9047
9048 2023-07-03 Daniel Kiper <[email protected]>
9049
9050 efi: Fallback to legacy mode if shim is loaded on x86 archs
9051 The LoadImage() provided by the shim does not consult MOK when loading
9052 an image. So, simply signature verification fails when it should not.
9053 This means we cannot use Linux EFI stub to start the kernel when the
9054 shim is loaded. We have to fallback to legacy mode on x86 architectures.
9055 This is not possible on other architectures due to lack of legacy mode.
9056
9057 This is workaround which should disappear when the shim provides
9058 LoadImage() which looks up MOK during signature verification.
9059
9060 On the occasion align constants in include/grub/efi/sb.h.
9061
9062 Reviewed-by: Ard Biesheuvel <[email protected]>
9063
9064 2023-07-03 Daniel Kiper <[email protected]>
9065
9066 efi: Drop __grub_efi_api attribute from shim_lock->verify() function
9067 ... because (surprisingly) it does not use specific EFI calling convention...
9068
9069 Fixes: 6a080b9cd (efi: Add calling convention annotation to all prototypes)
9070
9071 Reviewed-by: Ard Biesheuvel <[email protected]>
9072
9073 2023-07-03 Samuel Thibault <[email protected]>
9074
9075 templates: Start pci-arbiter before acpi on Hurd
9076 acpi actually needs to access PCI, while pci-arbiter will not be making
9077 use of ACPI, so we need to start acpi first.
9078
9079 Reviewed-by: Daniel Kiper <[email protected]>
9080
9081 2023-07-03 Michał Grzelak <[email protected]>
9082
9083 configure.ac: Fix typo by adding missing $
9084 During configuration of SDL, variable enable_grub_emu_sdl is checked
9085 whether to throw an error message. However, error could not happen
9086 because two unequal strings were compared. Fix this by referencing
9087 value of enable_grub_emu_sdl, not name.
9088
9089 Fixes: 17d6ac1a7 (emu: Add SDL2 support)
9090
9091 Reviewed-by: Daniel Kiper <[email protected]>
9092
9093 2023-07-03 Glenn Washburn <[email protected]>
9094
9095 docs: Minor corrections
9096 When referring to initrd16 the link for initrd16 should be used, not a link
9097 for initrd. Also, correct the spelling of additionally and add a comma after
9098 it to correct its grammatical usage.
9099
9100 Reviewed-by: Daniel Kiper <[email protected]>
9101
9102 2023-07-03 Glenn Washburn <[email protected]>
9103
9104 kern/misc: Add space after comma in function argument list
9105 Reviewed-by: Daniel Kiper <[email protected]>
9106
9107 commands/regexp: Fix typo
9108 Reviewed-by: Daniel Kiper <[email protected]>
9109
9110 2023-07-03 Glenn Washburn <[email protected]>
9111
9112 term/serial: Use grub_strncpy() instead of grub_snprintf() when only copying string
9113 Using grub_strncpy() instead of grub_snprintf() is less overhead and
9114 indicates clearly that the dest should be the same string as the source.
9115
9116 Also fix indentation.
9117
9118 Reviewed-by: Daniel Kiper <[email protected]>
9119
9120 2023-07-03 Glenn Washburn <[email protected]>
9121
9122 loader/linux: Print debug message for each generated newc path generated
9123 Reviewed-by: Daniel Kiper <[email protected]>
9124
9125 2023-07-03 Glenn Washburn <[email protected]>
9126
9127 include/grub/types.h: Add PRI*GRUB_OFFSET and PRI*GRUB_DISK_ADDR
9128 These are currently always the same as PRI*GRUB_UINT64_T, but they may
9129 not be in the future.
9130
9131 Reviewed-by: Daniel Kiper <[email protected]>
9132
9133 2023-07-03 Glenn Washburn <[email protected]>
9134
9135 kern/misc: Support octal printf format code
9136 Also add parenthesis to nested ternary operator to improve clarity.
9137
9138 Reviewed-by: Daniel Kiper <[email protected]>
9139
9140 2023-07-03 Glenn Washburn <[email protected]>
9141
9142 gitignore: Ignore python bytecode files
9143 Python bytecode files, which end in .pyc, may be generated by the build
9144 system as needed and should not go into the git repository.
9145
9146 Reviewed-by: Daniel Kiper <[email protected]>
9147
9148 2023-07-03 Glenn Washburn <[email protected]>
9149
9150 loader/linux: Only emit newc directory once
9151 When creating at runtime a newc initrd via arguments to initrd with "newc:"
9152 prefixes, only emit a directory path record once. The original code
9153 intended to do that by bailing out of emitting the record when the record
9154 to be created matches an existing record. However, this does not happen
9155 because grub_memcmp() is improperly checked.
9156
9157 Generating duplicate newc directory records does not cause any problems
9158 because the Linux unpacker will skip it once it sees the directory already
9159 exists. This fix saves a little processing and makes the generated newc
9160 cpio archive a little smaller.
9161
9162 Fixes: 92750e4c60 (Add ability to generate newc additions on runtime.)
9163
9164 Reviewed-by: Daniel Kiper <[email protected]>
9165
9166 2023-07-03 Glenn Washburn <[email protected]>
9167
9168 loader/efi/linux: Fix formatting and remove unneeded parenthesis
9169 Reviewed-by: Daniel Kiper <[email protected]>
9170
9171 2023-07-03 Glenn Washburn <[email protected]>
9172
9173 loader/efi/linux: Print EFI status as hex number instead of uint
9174 EFI status codes are of different classes depending on the first byte and
9175 all error status codes defined in appendix D of the main spec start from
9176 1 and have the high bit set. When printing as a uint, the decimal is a very
9177 large number that needs have the high bit cleared get the spec error code.
9178 This can be easily visually done by a human if the number is printed as hex.
9179
9180 Reviewed-by: Daniel Kiper <[email protected]>
9181
9182 2023-07-03 Oskari Pirhonen <[email protected]>
9183
9184 docs: Minor edits to debugging chapter
9185 Small set of wording and grammatical edits which did not make it in time
9186 for the original review of the chapter.
9187
9188 Reviewed-by: Glenn Washburn <[email protected]>
9189 Reviewed-by: Daniel Kiper <[email protected]>
9190
9191 2023-06-23 Daniel Kiper <[email protected]>
9192
9193 lib/relocator: Fix OOB write when initializing lo->freebytes[]
9194 Fixes: CID 96636
9195
9196 Reviewed-by: Vladimir Serbinenko <[email protected]>
9197
9198 2023-06-23 Daniel Kiper <[email protected]>
9199
9200 lib/relocator: Enforce GRUB_RELOCATOR_FIRMWARE_REQUESTS_QUANT divisibility by 8
9201 Most of leftover code blindly assumes GRUB_RELOCATOR_FIRMWARE_REQUESTS_QUANT
9202 divisibility by 8. So, enforce this at compile time.
9203
9204 Reviewed-by: Vladimir Serbinenko <[email protected]>
9205
9206 2023-06-23 Julian Andres Klode <[email protected]>
9207
9208 emu: Add SDL2 support
9209 So all we did with the surface in SDL1 was split into window,
9210 surface, renderer and texture. Instead of drawing into the
9211 surface and then flipping, you build your pixels, then update
9212 a texture and then copy the texture to the renderer.
9213
9214 Here we use an empty RGB surface to hold our pixels, which enables
9215 us to keep most of the code the same. The SDL1 code has been adjusted
9216 to refer to "surface" instead of "window" when trying to access the
9217 properties of the surface.
9218
9219 This approaches the configuration by adding a new --enable-grub-emu-sdl2
9220 argument. If set to yes, or auto detected, it disables SDL1 support
9221 automatically.
9222
9223 This duplicates the sdl module block in Makefile.core.def which may
9224 be something to be aware of, but we also don't want to build separate
9225 module.
9226
9227 Fixes: https://bugs.debian.org/1038035
9228
9229 Reviewed-by: Daniel Kiper <[email protected]>
9230
9231 2023-06-23 Julian Andres Klode <[email protected]>
9232
9233 emu: SDL style fixes
9234 These should be quite obvious and will make the SDL2 patch easier
9235 to read then doing it inline there.
9236
9237 Reviewed-by: Daniel Kiper <[email protected]>
9238
9239 2023-06-23 Michał Grzelak <[email protected]>
9240
9241 tpm: Enable boot despite unknown firmware failure
9242 Currently booting the system is prevented when call to EFI firmware
9243 hash_log_extend_event() returns unknown error. Solve this by following
9244 convention used in commit a4356538d (commands/tpm: Don't propagate
9245 measurement failures to the verifiers layer).
9246
9247 Let the system to be bootable by default when unknown TPM error is
9248 encountered. Check environment variable tpm_fail_fatal to fallback to
9249 previous behaviour.
9250
9251 Reviewed-by: Daniel Kiper <[email protected]>
9252
9253 2023-06-23 Daniel Kiper <[email protected]>
9254
9255 bootstrap: Fix patching warnings
9256 Currently bootstrap complains in the following way when
9257 patching gnulib files:
9258
9259 patching file argp-help.c
9260 Hunk #1 succeeded at 52 (offset 1 line).
9261 Hunk #2 succeeded at 1548 (offset 115 lines).
9262 patching file mbswidth.c
9263 patching file mbswidth.h
9264 Hunk #1 succeeded at 40 (offset -5 lines).
9265
9266 Let's fix it by amending line numbers in the patch.
9267
9268 Reviewed-by: Alec Brown <[email protected]>
9269
9270 2023-06-23 Daniel Kiper <[email protected]>
9271
9272 efi: Add missing __grub_efi_api attributes
9273 The commit bb4aa6e06 (efi: Drop all uses of efi_call_XX() wrappers) did
9274 not add some __grub_efi_api attributes to the EFI calls. Lack of them
9275 led to hangs on x86_64-efi target. So, let's add missing __grub_efi_api
9276 attributes.
9277
9278 Fixes: bb4aa6e06 (efi: Drop all uses of efi_call_XX() wrappers)
9279
9280 Reported-by: Christian Hesse <[email protected]>
9281 Reported-by: Robin Candau <[email protected]>
9282 Tested-by: Robin Candau <[email protected]>
9283 Tested-by: Christian Hesse <[email protected]>
9284 Reviewed-by: Peter Jones <[email protected]>
9285
9286 2023-06-23 Julian Andres Klode <[email protected]>
9287
9288 disk: Generalize MD_MAX_DISKS to GRUB_MDRAID_MAX_DISKS
9289 Move the constant from grub-core/osdep/linux/getroot.c to
9290 include/grub/disk.h and then reuse it in place of the
9291 hardcoded 1024 limit in diskfilter.
9292
9293 Fixes: 2a5e3c1f2 (disk/diskfilter: Don't make a RAID array with more than 1024 disks)
9294
9295 Cc: Daniel Axtens <[email protected]>
9296 Cc: Kees Cook <[email protected]>
9297 Reviewed-by: Kees Cook <[email protected]>
9298 Reviewed-by: Daniel Kiper <[email protected]>
9299
9300 2023-06-23 Xiaotian Wu <[email protected]>
9301
9302 loongarch: Disable relaxation relocations
9303 A working GRUB cannot be built with upcoming binutils and GCC, because linker
9304 relaxation was added [1] causing new unsupported relocations to appear in modules.
9305
9306 So we pass -mno-relax to GCC if it is supported, to disable relaxation and make
9307 GRUB forward-compatible with new toolchains.
9308
9309 While similar code already exists for sparc64 in configure.ac, sparc64 sets
9310 LDFLAGS while LoongArch requires CFLAGS to be set. If we only set LDFLAGS on
9311 LoongArch, GCC will still generate relaxation relocations in the .o files, so
9312 the sparc64 code cannot be reused.
9313
9314 [1] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=56576f4a722b7398d35802ecf7d4185c27d6d69b
9315
9316 Reviewed-by: Daniel Kiper <[email protected]>
9317
9318 2023-06-13 Xiaotian Wu <[email protected]>
9319
9320 loongarch: Add ELF relocation types documentation and comments
9321 See https://github.com/loongson/la-abi-specs/blob/release/laelf.adoc#relocations
9322
9323 Reviewed-by: Daniel Kiper <[email protected]>
9324
9325 2023-06-13 Xiaotian Wu <[email protected]>
9326
9327 loongarch: Rename function names
9328 According to the relocation documentation, the following function names are
9329 renamed to show their exact meaning:
9330 - from grub_loongarch64_xxx64_hi12() to grub_loongarch64_abs64_hi12(),
9331 - from grub_loongarch64_xxx64_hi12() to grub_loongarch64_abs64_lo20().
9332
9333 Reviewed-by: Daniel Kiper <[email protected]>
9334
9335 2023-06-13 Xiaotian Wu <[email protected]>
9336
9337 util/grub-mkimagexx: Optimize code using pc variable
9338 We already have the pc variable, no need to calculate it again.
9339
9340 Reviewed-by: Daniel Kiper <[email protected]>
9341
9342 2023-06-13 Xiaotian Wu <[email protected]>
9343
9344 kern/{arm64,loongarch64}/dl_helper: Use the correct format specifier for formatted output
9345 Use PRIxGRUB_INT64_T format specifier for grub_int64_t type
9346 and drop redundant casts.
9347
9348 Reviewed-by: Daniel Kiper <[email protected]>
9349
9350 2023-06-13 Qiumiao Zhang <[email protected]>
9351
9352 kern/acpi: Use xsdt_addr if present
9353 According to the ACPI specification, in ACPI 2.0 or later, an
9354 ACPI-compatible OS must use the XSDT if present. So, we should
9355 use xsdt_addr instead of rsdt_addr if xsdt_addr is valid.
9356
9357 Reviewed-by: Daniel Kiper <[email protected]>
9358
9359 2023-06-13 Qiumiao Zhang <[email protected]>
9360
9361 commands/acpi: Use xsdt_addr if present
9362 According to the ACPI specification, in ACPI 2.0 or later, an
9363 ACPI-compatible OS must use the XSDT if present. So, we should
9364 use xsdt_addr instead of rsdt_addr if xsdt_addr is valid.
9365
9366 Reviewed-by: Daniel Kiper <[email protected]>
9367
9368 2023-06-13 Lidong Chen <[email protected]>
9369
9370 fs/udf: Fix out of bounds access
9371 Implemented a boundary check before advancing the allocation
9372 descriptors pointer.
9373
9374 Reviewed-by: Darren Kenny <[email protected]>
9375 Reviewed-by: Daniel Kiper <[email protected]>
9376
9377 2023-06-13 Glenn Washburn <[email protected]>
9378
9379 docs: Add debugging chapter to development documentation
9380 Debugging GRUB can be tricky and require arcane knowledge. This will
9381 help those unfamiliar with the process to get started debugging GRUB
9382 with less effort.
9383
9384 Reviewed-by: Daniel Kiper <[email protected]>
9385
9386 2023-06-13 Darren Kenny <[email protected]>
9387
9388 fs/xfs: Fix issues found while fuzzing the XFS filesystem
9389 While performing fuzz testing with XFS filesystem images with ASAN
9390 enabled, several issues were found where the memory accesses are made
9391 beyond the data that is allocated into the struct grub_xfs_data
9392 structure's data field.
9393
9394 The existing structure didn't store the size of the memory allocated into
9395 the buffer in the data field and had no way to check it. To resolve these
9396 issues, the data size is stored to enable checks into the data buffer.
9397
9398 With these checks in place, the fuzzing corpus no longer cause any crashes.
9399
9400 Reviewed-by: Daniel Kiper <[email protected]>
9401
9402 2023-06-13 Alexander Kanavin <[email protected]>
9403
9404 util/import_unicode.py: Ensure output is deterministic
9405 Ensure the generated unidata.c file is deterministic by sorting the
9406 keys of the dict.
9407
9408 Reviewed-by: Daniel Kiper <[email protected]>
9409
9410 2023-06-13 Alexander Kanavin <[email protected]>
9411
9412 grub-core/genmoddep.awk: Ensure output is deterministic
9413 The output in moddep.lst generated from syminfo.lst using genmoddep.awk
9414 is not deterministic since the order of the dependencies on each line
9415 can vary depending on how awk sorts the values in the array.
9416
9417 Be deterministic in the output by sorting the dependencies on each line.
9418
9419 Reviewed-by: Daniel Kiper <[email protected]>
9420
9421 2023-06-13 Alexander Kanavin <[email protected]>
9422
9423 gentpl.py: Ensure output is deterministic
9424 The output of the SOURCES lines in grub-core/Makefile.core.am, generated
9425 from grub-core/Makefile.core.def with gentpl.py is not deterministic due to
9426 missing sorting of the list used to generate it. Add such a sort.
9427
9428 Reviewed-by: Daniel Kiper <[email protected]>
9429
9430 2023-06-01 Glenn Washburn <[email protected]>
9431
9432 gdb: Add gdbinfo command for printing the load address of the EFI application
9433 EFI firmware determines where to load the GRUB EFI at runtime, and so the
9434 addresses of debug symbols are not known ahead of time. There is a command
9435 defined in the gdb_grub script which will load the debug symbols at the
9436 appropriate addresses, if given the application load address for GRUB.
9437 So add a command named "gdbinfo" to allow the user to print this GDB command
9438 string with the application load address on-demand. For the outputted GDB
9439 command to have any effect when entered into a GDB session, GDB should have
9440 been started with the script as an argument to the -x option or sourced into
9441 an active GDB session before running the outputted command.
9442
9443 Documentation for the gdbinfo command is also added.
9444
9445 Co-developed-by: Peter Jones <[email protected]>
9446 Reviewed-by: Daniel Kiper <[email protected]>
9447
9448 2023-06-01 Glenn Washburn <[email protected]>
9449
9450 loader/efi/chainloader: Do not require a $root visible to EFI firmware when chainloading
9451 The EFI chainloader checks that a device path can be created for the $root
9452 device before allowing chainloading to a given file. This is probably to
9453 ensure that the given file can be accessed and loaded by the firmware.
9454 However, since GRUB is loading the image itself, the firmware need not
9455 be able to access the file location of the image. So remove this check.
9456
9457 Also, this fixes an issue where chainloading an image file on a location
9458 that is accessible by the firmware, e.g. (hd0,1)/efi/boot.efi, would
9459 fail when root is a location inaccessible by the firmware, e.g. memdisk.
9460
9461 Use GRUB_EFI_BYTES_TO_PAGES() instead of doing the calculation explicitly.
9462
9463 Add comment noting the section where the load options for the chainloaded
9464 EFI application is constructed.
9465
9466 Reviewed-by: Ard Biesheuvel <[email protected]>
9467 Reviewed-by: Daniel Kiper <[email protected]>
9468
9469 2023-06-01 Glenn Washburn <[email protected]>
9470
9471 docs: Document extra arguments to chainloader on EFI
9472 Extra arguments given to chainloader on EFI platforms will be sent to
9473 the chainloaded application. Also, minor edit in the chainloading section
9474 to note that chainloading can be a jump via the firmware and not
9475 necessarily in real mode (which does not exist on some architectures).
9476
9477 Reviewed-by: Ard Biesheuvel <[email protected]>
9478 Reviewed-by: Daniel Kiper <[email protected]>
9479
9480 2023-06-01 Oliver Steffen <[email protected]>
9481
9482 util/grub.d/25_bli.in: Activate bli module on EFI
9483 Add a new configuration drop-in file that loads the bli module and runs
9484 the command if booting on the EFI platform.
9485
9486 Reviewed-by: Daniel Kiper <[email protected]>
9487
9488 2023-06-01 Oliver Steffen <[email protected]>
9489
9490 bli: Add a module for the Boot Loader Interface
9491 Add a new module named bli. It implements a small but quite useful part
9492 of the Boot Loader Interface [0]. This interface uses EFI variables for
9493 communication between the boot loader and the operating system.
9494
9495 When loaded, this module sets two EFI variables under the vendor GUID
9496 4a67b082-0a4c-41cf-b6c7-440b29bb8c4f:
9497
9498 - LoaderInfo: contains GRUB + <version number>.
9499 This allows the running operating system to identify the boot loader
9500 used during boot.
9501
9502 - LoaderDevicePartUUID: contains the partition UUID of the EFI System
9503 Partition (ESP). This is used by systemd-gpt-auto-generator [1] to
9504 find the root partitions (and others too), via partition type IDs [2].
9505
9506 This module is available on EFI platforms only. The bli module relies on
9507 the part_gpt module which has to be loaded beforehand to make the GPT
9508 partitions discoverable.
9509
9510 Update the documentation, add a new chapter "Modules" and describe the
9511 bli module there.
9512
9513 [0] https://systemd.io/BOOT_LOADER_INTERFACE/
9514 [1] https://www.freedesktop.org/software/systemd/man/systemd-gpt-auto-generator.html
9515 [2] https://uapi-group.org/specifications/specs/discoverable_partitions_specification/
9516
9517 Reviewed-by: Daniel Kiper <[email protected]>
9518
9519 2023-06-01 Oliver Steffen <[email protected]>
9520
9521 kern: Check for NULL when closing devices and disks
9522 Add checks for NULL pointers to grub_device_close() and
9523 grub_disk_close() to make these functions more robust.
9524
9525 Reviewed-by: Daniel Kiper <[email protected]>
9526
9527 2023-06-01 Oliver Steffen <[email protected]>
9528
9529 docs: Reword section headings
9530 Reword some section headings, remove "The List of" from titles. While
9531 grammatically correct, this phrase can be omitted to increase
9532 readability, especially in the table of contents.
9533
9534 Reviewed-by: Daniel Kiper <[email protected]>
9535
9536 2023-06-01 Oliver Steffen <[email protected]>
9537
9538 efi: Add grub_efi_set_variable_to_string()
9539 Add a function that sets an EFI variable to a string value.
9540 The string is converted from UTF-8 to UTF-16.
9541
9542 Reviewed-by: Daniel Kiper <[email protected]>
9543
9544 2023-06-01 Oliver Steffen <[email protected]>
9545
9546 kern/misc, kern/efi: Extract UTF-8 to UTF-16 code
9547 Create a new function for UTF-8 to UTF-16 conversion called
9548 grub_utf8_to_utf16_alloc() in the grub-code/kern/misc.c and replace
9549 charset conversion code used in some places in the EFI code. It is
9550 modeled after the grub_utf8_to_ucs4_alloc() like functions in
9551 include/grub/charset.h. It can't live in include/grub/charset.h,
9552 because it needs to be reachable from the kern/efi code.
9553
9554 Add a check for integer overflow and remove redundant NUL-termination.
9555
9556 Reviewed-by: Daniel Kiper <[email protected]>
9557
9558 2023-06-01 Oliver Steffen <[email protected]>
9559
9560 include/grub/types.h: Add GRUB_SSIZE_MAX
9561 In the same way as GRUB_SIZE_MAX, add GRUB_SSIZE_MAX.
9562
9563 Reviewed-by: Daniel Kiper <[email protected]>
9564
9565 2023-06-01 Oliver Steffen <[email protected]>
9566
9567 guid: Make use of GUID printf format specifier
9568 Use the new printf format specifier %pG.
9569
9570 Fixes the text representation of GUIDs in the output of the lsefisystab
9571 command (missing 4th dash).
9572
9573 Reviewed-by: Daniel Kiper <[email protected]>
9574
9575 2023-06-01 Oliver Steffen <[email protected]>
9576
9577 kern/misc: Add a format specifier GUIDs
9578 Extend the printf format specifier for pointers (%p) to accept a suffix
9579 specifier G to print GUIDs: %pG can be used to print grub_guid structs.
9580 This does not interfere with the -Wformat checking of gcc. Note that
9581 the data type is not checked though (%p accepts void *).
9582
9583 Reviewed-by: Daniel Kiper <[email protected]>
9584
9585 2023-06-01 Oliver Steffen <[email protected]>
9586
9587 guid: Unify GUID types
9588 There are 3 implementations of a GUID in GRUB. Replace them with
9589 a common one, placed in types.h.
9590
9591 It uses the "packed" flavor of the GUID structs, the alignment attribute
9592 is dropped, since it is not required.
9593
9594 Reviewed-by: Daniel Kiper <[email protected]>
9595
9596 2023-06-01 Oliver Steffen <[email protected]>
9597
9598 efi: Add grub_efi_set_variable_with_attributes()
9599 Add a function to the EFI module that allows setting EFI variables
9600 with specific attributes.
9601
9602 This is useful for marking variables as volatile, for example.
9603
9604 Reviewed-by: Daniel Kiper <[email protected]>
9605
9606 2023-05-25 Alec Brown <[email protected]>
9607
9608 kern/efi/mm: Fix use-after-free in finish boot services
9609 In grub-core/kern/efi/mm.c, grub_efi_finish_boot_services() has an instance
9610 where the memory for the variable finish_mmap_buf is freed, but on the next
9611 iteration of a while loop, grub_efi_get_memory_map() uses finish_mmap_buf. To
9612 prevent this, we can set finish_mmap_buf to NULL after the free.
9613
9614 Reviewed-by: Daniel Kiper <[email protected]>
9615
9616 2023-05-25 Ard Biesheuvel <[email protected]>
9617
9618 efi: Handle NULL return value when getting loaded image protocol
9619 The EFI spec mandates that the handle produced by the LoadImage boot
9620 service has a LoadedImage protocol instance installed on it, but for
9621 robustness, we should still deal with a NULL return value from the
9622 helper routine that obtains this protocol pointer.
9623
9624 If this happens, don't try to start the image but unload it and return
9625 an error.
9626
9627 Reviewed-by: Daniel Kiper <[email protected]>
9628
9629 2023-05-25 Ard Biesheuvel <[email protected]>
9630
9631 efi: Use generic EFI loader for x86_64 and i386
9632 Switch the x86 based EFI platform builds to the generic EFI loader,
9633 which exposes the initrd via the LoadFile2 protocol instead of the
9634 x86-specific setup header. This will launch the Linux kernel via its EFI
9635 stub, which performs its own initialization in the EFI boot services
9636 context before calling ExitBootServices() and performing the bare metal
9637 Linux boot.
9638
9639 Given that only Linux kernel versions v5.8 and later support this initrd
9640 loading method, the existing x86 loader is retained as a fallback, which
9641 will also be used for Linux kernels built without the EFI stub. In this
9642 case, GRUB calls ExitBootServices() before entering the Linux kernel,
9643 and all EFI related information is provided to the kernel via struct
9644 boot_params in the setup header, as before.
9645
9646 Note that this means that booting EFI stub kernels older than v5.8 is
9647 not supported even when not using an initrd at all. Also, the EFI
9648 handover protocol, which has no basis in the UEFI specification, is not
9649 implemented.
9650
9651 Reviewed-by: Daniel Kiper <[email protected]>
9652
9653 2023-05-25 Ard Biesheuvel <[email protected]>
9654
9655 efi: Remove x86_64 call wrappers
9656 The call wrappers are no longer needed now that GCC can generate
9657 function calls using MS calling convention, so let's get rid of them.
9658
9659 Reviewed-by: Daniel Kiper <[email protected]>
9660
9661 2023-05-25 Ard Biesheuvel <[email protected]>
9662
9663 efi: Drop all uses of efi_call_XX() wrappers
9664 Now that GCC can generate function calls using the correct calling
9665 convention for us, we can stop using the efi_call_XX() wrappers, and
9666 just dereference the function pointers directly.
9667
9668 This avoids the untyped variadic wrapper routines, which means better
9669 type checking for the method calls.
9670
9671 Reviewed-by: Daniel Kiper <[email protected]>
9672
9673 2023-05-25 Ard Biesheuvel <[email protected]>
9674
9675 efi: Add calling convention annotation to all prototypes
9676 UEFI mandates MS calling convention on x86_64, which was not supported
9677 on GCC when UEFI support was first introduced into GRUB. However, now we
9678 can use the ms_abi function type attribute to annotate functions and
9679 function pointers as adhering to the MS calling convention, and the
9680 compiler will generate the correct instruction sequence for us.
9681
9682 So let's add the appropriate annotation to all the function prototypes.
9683 This will allow us to drop the special call wrappers in a subsequent patch.
9684
9685 Reviewed-by: Daniel Kiper <[email protected]>
9686
9687 2023-05-25 Ard Biesheuvel <[email protected]>
9688
9689 efi: Make EFI PXE protocol methods non-callable
9690 The grub_efi_pxe_t struct definition has placeholders for the various
9691 protocol method pointers, given that they are never called in the code,
9692 and the prototypes have been omitted, and therefore do not comply with
9693 the UEFI spec.
9694
9695 So let's convert them into void* pointers, so they cannot be called
9696 inadvertently.
9697
9698 Reviewed-by: Daniel Kiper <[email protected]>
9699
9700 2023-05-25 Alec Brown <[email protected]>
9701
9702 loader/multiboot_elfxx: Check program header offset doesn't exceed constraints
9703 In grub-core/loader/multiboot_elfxx.c, we need to make sure that the program
9704 header offset is less than the file size along with the MULTIBOOT_SEARCH
9705 constant. We can do so by setting the variable phlimit to the minimum value of
9706 the two limits and check it each time we change program header index to insure
9707 that the program header offset isn't outside of the limits.
9708
9709 Fixes: CID 314029
9710 Fixes: CID 314038
9711
9712 Reviewed-by: Daniel Kiper <[email protected]>
9713
9714 2023-05-25 Alec Brown <[email protected]>
9715
9716 loader/multiboot_elfxx: Check section header region before allocating memory
9717 In grub-core/loader/multiboot_elfxx.c, space is being allocated for the section
9718 header region, but isn't verifying if the region is within the file's size.
9719 Before calling grub_calloc(), we can add a conditional to check if the section
9720 header region is smaller than the file size.
9721
9722 Fixes: CID 314029
9723 Fixes: CID 314038
9724
9725 Reviewed-by: Daniel Kiper <[email protected]>
9726
9727 2023-05-25 Alec Brown <[email protected]>
9728
9729 loader/multiboot_elfxx: Check program memory isn't larger than allocated memory size
9730 In grub-core/loader/multiboot_elfxx.c, the code is filling an area of memory
9731 with grub_memset() but doesn't check if there is space in the allocated memory
9732 before doing so. To make sure we aren't zeroing memory past the allocated memory
9733 region, we need to check that the offset into the allocated memory region plus
9734 the memory size of the program is smaller than the allocated memory size.
9735
9736 Fixes: CID 314029
9737 Fixes: CID 314038
9738
9739 Reviewed-by: Daniel Kiper <[email protected]>
9740
9741 2023-05-25 WANG Xuerui <[email protected]>
9742
9743 kern/loongarch64/dl_helper: Avoid undefined behavior when popping from an empty reloc stack
9744 The return value of grub_loongarch64_stack_pop() is unsigned, so -1 should
9745 not be used in the first place. Replacing with 0 is enough to avoid the
9746 UB in this edge case.
9747
9748 Technically though, proper error handling is needed throughout the
9749 management of the reloc stack, so no unexpected behavior will happen
9750 even in case of malformed object code input (right now, pushes become
9751 no-ops when the stack is full, and garbage results if the stack does not
9752 contain enough operands for an op). The refactor would touch some more
9753 places so would be best done in a separate series.
9754
9755 Fixes: CID 407777
9756 Fixes: CID 407778
9757
9758 Reviewed-by: Daniel Kiper <[email protected]>
9759
9760 2023-05-25 Peter Zijlstra (Intel) <[email protected]>
9761
9762 pci: Rename GRUB_PCI_CLASS_*
9763 Glenn suggested to rename the existing PCI_CLASS defines to have
9764 explicit class and subclass names.
9765
9766 Suggested-by: Glenn Washburn <[email protected]>
9767 Reviewed-by: Daniel Kiper <[email protected]>
9768
9769 2023-05-25 Peter Zijlstra (Intel) <[email protected]>
9770
9771 term/serial: Add support for PCI serial devices
9772 Loosely based on early_pci_serial_init() from Linux, allow GRUB to make
9773 use of PCI serial devices.
9774
9775 Specifically, my Alderlake NUC exposes the Intel AMT SoL UART as a PCI
9776 enumerated device but doesn't include it in the EFI tables.
9777
9778 Tested and confirmed working on a "Lenovo P360 Tiny" with Intel AMT
9779 enabled. This specific machine has (from lspci -vv):
9780
9781 00:16.3 Serial controller: Intel Corporation Device 7aeb (rev 11) (prog-if 02 [16550])
9782 DeviceName: Onboard - Other
9783 Subsystem: Lenovo Device 330e
9784 Control: I/O+ Mem+ BusMaster- SpecCycle- MemWINV- VGASnoop- ParErr- Stepping- SERR- FastB2B- DisINTx-
9785 Status: Cap+ 66MHz+ UDF- FastB2B+ ParErr- DEVSEL=fast >TAbort- <TAbort- <MAbort- >SERR- <PERR- INTx-
9786 Interrupt: pin D routed to IRQ 19
9787 Region 0: I/O ports at 40a0 [size=8]
9788 Region 1: Memory at b4224000 (32-bit, non-prefetchable) [size=4K]
9789 Capabilities: [40] MSI: Enable- Count=1/1 Maskable- 64bit+
9790 Address: 0000000000000000 Data: 0000
9791 Capabilities: [50] Power Management version 3
9792 Flags: PMEClk- DSI+ D1- D2- AuxCurrent=0mA PME(D0-,D1-,D2-,D3hot-,D3cold-)
9793 Status: D0 NoSoftRst+ PME-Enable- DSel=0 DScale=0 PME-
9794 Kernel driver in use: serial
9795
9796 From which the following config (/etc/default/grub) gets a working
9797 serial setup:
9798
9799 GRUB_CMDLINE_LINUX="console=tty0 earlyprintk=pciserial,00:16.3,115200 console=ttyS0,115200"
9800 GRUB_SERIAL_COMMAND="serial --port=0x40a0 --speed=115200"
9801 GRUB_TERMINAL="serial console"
9802
9803 Documentation is added to note that serial devices found on the PCI bus will
9804 be exposed as "pci,XX:XX.X" and how to find serial terminal logical names.
9805 Also, some minor documentation improvements were added.
9806
9807 This can be tested in QEMU by adding a pci-serial device, e.g. using the option
9808 "-device pci-serial".
9809
9810 Tested-by: Glenn Washburn <[email protected]>
9811 Reviewed-by: Daniel Kiper <[email protected]>
9812
9813 2023-05-17 Glenn Washburn <[email protected]>
9814
9815 tests/util/grub-fs-tester: Avoid failing some file system tests due to file system filling up
9816 On some systems /usr/share/dict/american-english can be larger than the
9817 available space on the filesystem being tested (e.g. vfat12a). This
9818 causes a failure of the filesystem test and is not a real test failure.
9819 Instead, use dd to copy at most 1 MiB of data to the filesystem, which is
9820 enough for our purposes and will not fill any of the tested filesystems.
9821
9822 Reviewed-by: Daniel Kiper <[email protected]>
9823
9824 2023-05-17 Glenn Washburn <[email protected]>
9825
9826 docs: Command-line and menu entry commands are now separated
9827 The menu entry commands now have their own section. Change the wording in
9828 the section that they were in to reflect this.
9829
9830 Reviewed-by: Daniel Kiper <[email protected]>
9831
9832 2023-05-17 Roger Pau Monné <[email protected]>
9833
9834 lib/relocator: Always enforce the requested alignment in malloc_in_range()
9835 On failure to allocate from grub_relocator_firmware_alloc_region() in
9836 malloc_in_range() the function would stop enforcing the alignment, and
9837 the following was returned:
9838
9839 lib/relocator.c:431: trying to allocate in 0x200000-0xffbf9fff aligned 0x200000 size 0x406000
9840 lib/relocator.c:1197: allocated: 0x74de2000+0x406000
9841 lib/relocator.c:1407: allocated 0x74de2000/0x74de2000
9842
9843 Fix this by making sure that target always contains a suitably aligned
9844 address. After the change the return from the function is:
9845
9846 lib/relocator.c:431: trying to allocate in 0x200000-0xffb87fff aligned 0x200000 size 0x478000
9847 lib/relocator.c:1204: allocated: 0x74c00000+0x478000
9848 lib/relocator.c:1414: allocated 0x74c00000/0x74c00000
9849
9850 Fixes: 3a5768645c05 (First version of allocation from firmware)
9851
9852 Reviewed-by: Daniel Kiper <[email protected]>
9853
9854 2023-05-17 Benjamin Herrenschmidt <[email protected]>
9855
9856 term/ns8250: Fix incorrect usage of access_size
9857 The access_size is part of a union, so doesn't technically exist for
9858 a PIO port (i.e., not MMIO), but we set it anyways.
9859
9860 This doesn't cause a bug today because the other leg of the union
9861 doesn't have anything overlapping with it now, but it's bad, I will
9862 punish myself for writing it that way :-) In the meantime, fix this
9863 and actually name the struct inside the union for clarity of intent
9864 and to avoid such issue in the future.
9865
9866 Reviewed-by: Daniel Kiper <[email protected]>
9867
9868 2023-05-17 Ákos Nagy <[email protected]>
9869
9870 util/grub-install-common: Fix the key of the --core-compress option
9871 Commit f23bc6510 (Transform -C option to grub-mkstandalone to
9872 --core-compress available in all grub-install flavours.) declared
9873 a new long option for specifying the compression method to use for
9874 the core image.
9875
9876 However, the option key has not been replaced in the parser function,
9877 it still expects the old one formerly used by grub-mkstandalone.
9878 Because of this the option is not recognized by any of the utils for
9879 which it is listed as supported.
9880
9881 Reviewed-by: Daniel Kiper <[email protected]>
9882
9883 2023-05-17 Lidong Chen <[email protected]>
9884
9885 fs/hfsplus: Set grub_errno to prevent NULL pointer access
9886 When an invalid node size is detected in grub_hfsplus_mount(), data
9887 pointer is freed. Thus, file->data is not set. The code should also
9888 set the grub_errno when that happens to indicate an error and to avoid
9889 accessing the uninitialized file->data in grub_file_close().
9890
9891 Reviewed-by: Daniel Kiper <[email protected]>
9892
9893 2023-05-17 Lidong Chen <[email protected]>
9894
9895 fs/hfsplus: Prevent out of bound access in catalog file
9896 A corrupted hfsplus can have a catalog key that is out of range. This
9897 can lead to out of bound access when advancing the pointer to access
9898 catalog file info. The valid range of a catalog key is specified in
9899 HFS Plus Technical Note TN1150 [1].
9900
9901 [1] https://developer.apple.com/library/archive/technotes/tn/tn1150.html
9902
9903 Reviewed-by: Daniel Kiper <[email protected]>
9904
9905 2023-05-17 Lidong Chen <[email protected]>
9906
9907 fs/hfsplus: Validate btree node size
9908 The invalid btree node size can cause crashes when parsing the btree.
9909 The fix is to ensure the btree node size is within the valid range
9910 defined in the HFS Plus technical note, TN1150 [1].
9911
9912 [1] https://developer.apple.com/library/archive/technotes/tn/tn1150.html
9913
9914 Reviewed-by: Daniel Kiper <[email protected]>
9915
9916 2023-05-17 Glenn Washburn <[email protected]>
9917
9918 INSTALL: Use exfat-utils package instead of exfatprogs
9919 The exfat-utils package is an older package complementing exfat-fuse, and
9920 was the only exfat tools for a long time. The exfat filesystem testing code
9921 was written with these tools in mind. A newer project exfatprogs appears to
9922 be of better quality and functionality and was written to complement the
9923 somewhat new exfat kernel module. Ideally we should be using the newer
9924 exfatprogs. However, the command line interface for mkfs.exfat is different
9925 between the two. So we can't use the exfatprogs tools until the test scripts
9926 have been updated to account for this. Recommend installing exfat-utils
9927 instead of exfatprogs for now.
9928
9929 Reviewed-by: Daniel Kiper <[email protected]>
9930
9931 2023-05-17 Glenn Washburn <[email protected]>
9932
9933 INSTALL: Document that building grub-mkfont requires xfonts-unifont
9934 Reviewed-by: Daniel Kiper <[email protected]>
9935
9936 2023-05-17 Renaud Métrich <[email protected]>
9937
9938 net/dns: Fix lookup error when no IPv6 is returned
9939 When trying to resolve DNS names into IP addresses, the DNS code fails
9940 from time to time with the following error:
9941 -------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
9942 error: ../../grub-core/net/dns.c:688:no DNS record found.
9943 -------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
9944
9945 This happens when both IPv4 and IPv6 queries are performed against the
9946 DNS server (e.g. 8.8.8.8) but there is no IP returned for IPv6 query, as
9947 shown below:
9948 -------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
9949 grub> net_del_dns 192.168.122.1
9950 grub> net_add_dns 8.8.8.8
9951 grub> net_nslookup ipv4.test-ipv6.com
9952 error: ../../grub-core/net/dns.c:688:no DNS record found.
9953 grub> net_nslookup ipv4.test-ipv6.com
9954 216.218.228.115
9955 -------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
9956
9957 The root cause is the code exiting prematurely when the data->addresses
9958 buffer has been allocated in recv_hook(), even if there was no address
9959 returned last time recv_hook() executed.
9960
9961 Reviewed-by: Daniel Kiper <[email protected]>
9962
9963 2023-05-17 Renaud Métrich <[email protected]>
9964
9965 net/dns: Add debugging messages in recv_hook() function
9966 Reviewed-by: Daniel Kiper <[email protected]>
9967
9968 net/dns: Simplify error handling of recv_hook() function
9969 Reviewed-by: Daniel Kiper <[email protected]>
9970
9971 2023-05-17 Renaud Métrich <[email protected]>
9972
9973 net/dns: Fix removal of DNS server
9974 When deleting the DNS server, we get the following error message:
9975 -------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
9976 grub> net_del_dns 192.168.122.1
9977 error: ../../grub-core/net/dns.c:646:no DNS reply received.
9978 -------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
9979
9980 This happens because the implementation is broken, it does a "add"
9981 internally instead of a "delete".
9982
9983 Reviewed-by: Daniel Kiper <[email protected]>
9984
9985 2023-05-17 Xiaotian Wu <[email protected]>
9986
9987 tests: Add LoongArch to various test cases
9988 I ran the test suite on a 3A5000 desktop, a LoongArch architecture machine,
9989 using Archlinux for LoongArch distro, see https://github.com/loongarchlinux.
9990
9991 Some software versions are:
9992 * linux 6.3.0-rc4
9993 * gcc 13.0.1 20230312
9994 * binutils 2.40
9995 * qemu 7.2.0
9996
9997 The test results of running "make check" with qemu 7.2 are as follows:
9998
9999 =================================
10000 GRUB 2.11: ./test-suite.log
10001 =================================
10002
10003 # TOTAL: 85
10004 # PASS: 73
10005 # SKIP: 8
10006 # XFAIL: 0
10007 # FAIL: 2
10008 # XPASS: 0
10009 # ERROR: 2
10010
10011 .. contents:: :depth: 2
10012
10013 ERROR: f2fs_test
10014 ================
10015
10016 mount: /tmp/grub-fs-tester.20230418175640563815408.f2fs.UDs/f2fs_rw: unknown filesystem type 'f2fs'.
10017 dmesg(1) may have more information after failed mount system call.
10018 MOUNT FAILED.
10019 ERROR f2fs_test (exit status: 99)
10020
10021 FAIL: hfs_test
10022 ==============
10023
10024 recode: Request `utf8..macroman' is erroneous
10025 mkfs.hfs: name required with -v option
10026 FAIL hfs_test (exit status: 1)
10027
10028 ERROR: zfs_test
10029 ===============
10030
10031 zpool not installed; cannot test zfs.
10032 ERROR zfs_test (exit status: 99)
10033
10034 SKIP: pata_test
10035 ===============
10036
10037 SKIP pata_test (exit status: 77)
10038
10039 SKIP: ahci_test
10040 ===============
10041
10042 SKIP ahci_test (exit status: 77)
10043
10044 SKIP: uhci_test
10045 ===============
10046
10047 SKIP uhci_test (exit status: 77)
10048
10049 SKIP: ohci_test
10050 ===============
10051
10052 SKIP ohci_test (exit status: 77)
10053
10054 SKIP: ehci_test
10055 ===============
10056
10057 SKIP ehci_test (exit status: 77)
10058
10059 SKIP: fddboot_test
10060 ==================
10061
10062 SKIP fddboot_test (exit status: 77)
10063
10064 SKIP: netboot_test
10065 ==================
10066
10067 SKIP netboot_test (exit status: 77)
10068
10069 SKIP: pseries_test
10070 ==================
10071
10072 SKIP pseries_test (exit status: 77)
10073
10074 FAIL: grub_func_test
10075 ====================
10076
10077 WARNING: Image format was not specified for '/tmp/grub-shell.HeTAD8Ty3U/grub.iso' and probing guessed raw.
10078 Automatically detecting the format is dangerous for raw images, write operations on block 0 will be restricted.
10079 Specify the 'raw' format explicitly to remove the restrictions.
10080 Functional test failure: shift_test:
10081 ...
10082 gfxterm_menu_640x480xi16:3 failed: 0xce34981e vs 0xd9f04953
10083 tests/video_checksum.c:checksum:615: assert failed: 0 Checksum
10084 gfxterm_menu_640x480xi16:2 failed: 0xa8fb749d vs 0xbf3fa5d0
10085 tests/video_checksum.c:checksum:615: assert failed: 0 Checksum
10086 gfxterm_menu_640x480xi16:1 failed: 0xce34981e vs 0xd9f04953
10087 gfxterm_menu: FAIL
10088 ...
10089 videotest_checksum:
10090 videotest_checksum: PASS
10091 exfctest:
10092 exfctest: PASS
10093 TEST FAILURE
10094 FAIL grub_func_test (exit status: 1)
10095
10096 We got 2 errors:
10097
10098 * f2fs_test
10099 The kernel uses 16k pages, causing failures when loading the f2fs kernel module,
10100 see https://github.com/torvalds/linux/blob/master/fs/f2fs/super.c#L4670
10101 This error can be ignored.
10102
10103 * zfs_test
10104 zfs does not support the LoongArch architecture and is not compatible with the
10105 6.3 kernel.
10106 This error can be ignored.
10107
10108 We got 2 failures:
10109
10110 * hfs_test
10111 I use recode 3.7.14-1 on Archlinux, running `recode -l` gives no output `MacRoman`,
10112 so we get this error.
10113 On Linux systems that support LoongArch, there is currently no need to use HFS,
10114 so this failure can be ignored.
10115
10116 * grub_func_test
10117 I don't know the reason for this failure. I guess it may be related to qemu's edk2.
10118 In the previous review, I was told that the failure here is the expected behavior.
10119 So, we can ignore this failure.
10120
10121 Reviewed-by: Daniel Kiper <[email protected]>
10122
10123 2023-05-17 Xiaotian Wu <[email protected]>
10124
10125 tests: Fix timezone inconsistency in squashfs_test
10126 The image timestamp was not returned in UTC, but the following logic
10127 expected and used UTC.
10128
10129 This patch fixes the test failure like described below:
10130
10131 unsquashfs -s /tmp/grub-fs-tester.20230407111703613257436.squash4_gzip.9R4/squash4_gzip_512_4096_1_0.img
10132 grep '^Creation'
10133 awk '{print $6 " " $7 " " $8 " " $9 " " $10; }'
10134 FSTIME='Fri Apr 7 11:17:05 2023'
10135 date -d 'Fri Apr 7 11:17:05 2023' -u '+%Y-%m-%d %H:%M:%S'
10136 FSTIME='2023-04-07 11:17:05'
10137 date -d '2023-04-07 11:17:05 UTC -1 second' -u '+%Y-%m-%d %H:%M:%S'
10138 FSTIMEM1='2023-04-07 11:17:04'
10139 date -d '2023-04-07 11:17:05 UTC -2 second' -u '+%Y-%m-%d %H:%M:%S'
10140 FSTIMEM2='2023-04-07 11:17:03'
10141 date -d '2023-04-07 11:17:05 UTC -3 second' -u '+%Y-%m-%d %H:%M:%S'
10142 FSTIMEM3='2023-04-07 11:17:02'
10143 grep -F 'Last modification time 2023-04-07 11:17:05'
10144 echo 'Device loop0: Filesystem type squash4 - Last modification time 2023-04-07 03:17:05 Friday - Sector size 512B - Total size 10680KiB'
10145 echo 'Device loop0: Filesystem type squash4 - Last modification time 2023-04-07 03:17:05 Friday - Sector size 512B - Total size 10680KiB'
10146 grep -F 'Last modification time 2023-04-07 11:17:04'
10147 echo 'Device loop0: Filesystem type squash4 - Last modification time 2023-04-07 03:17:05 Friday - Sector size 512B - Total size 10680KiB'
10148 grep -F 'Last modification time 2023-04-07 11:17:03'
10149 echo 'Device loop0: Filesystem type squash4 - Last modification time 2023-04-07 03:17:05 Friday - Sector size 512B - Total size 10680KiB'
10150 grep -F 'Last modification time 2023-04-07 11:17:02'
10151 echo FSTIME FAIL
10152
10153 Reviewed-by: Glenn Washburn <[email protected]>
10154 Reviewed-by: Daniel Kiper <[email protected]>
10155
10156 2023-05-17 Xiaotian Wu <[email protected]>
10157
10158 loongarch: Add to build system
10159 This patch adds LoongArch to the GRUB build system and various tools,
10160 so GRUB can be built on LoongArch as a UEFI application.
10161
10162 Reviewed-by: Daniel Kiper <[email protected]>
10163
10164 2023-05-17 Xiaotian Wu <[email protected]>
10165
10166 loongarch: Add auxiliary files
10167 Add support for manipulating architectural cache and timers, and EFI
10168 memory maps.
10169
10170 Reviewed-by: Daniel Kiper <[email protected]>
10171
10172 2023-05-17 Xiaotian Wu <[email protected]>
10173
10174 loongarch: Add support for ELF psABI v2.00 relocations
10175 A new set of relocation types was added in the LoongArch ELF psABI v2.00
10176 spec [1], [2] to replace the stack-based scheme in v1.00. Toolchain
10177 support is available from binutils 2.40 and gcc 13 onwards.
10178
10179 This patch adds support for the new relocation types, that are simpler
10180 to handle (in particular, stack operations are gone). Support for the
10181 v1.00 relocs are kept for now, for compatibility with older toolchains.
10182
10183 [1] https://github.com/loongson/LoongArch-Documentation/pull/57
10184 [2] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html#_appendix_revision_history
10185
10186 Reviewed-by: Daniel Kiper <[email protected]>
10187
10188 2023-05-17 Xiaotian Wu <[email protected]>
10189
10190 loongarch: Add support for ELF psABI v1.00 relocations
10191 This patch adds support of the stack-based LoongArch relocations
10192 throughout GRUB, including tools, dynamic linkage, and support for
10193 conversion of ELF relocations into PE ones. A stack machine is required
10194 to handle these per the spec [1] (see the R_LARCH_SOP types), of which
10195 a simple implementation is included.
10196
10197 These relocations are produced by binutils 2.38 and 2.39, while the newer
10198 v2.00 relocs require more recent toolchain (binutils 2.40+ & gcc 13+, or
10199 LLVM 16+). GCC 13 has not been officially released as of early 2023, so
10200 support for v1.00 relocs are expected to stay relevant for a while.
10201
10202 [1] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html#_relocations
10203
10204 Reviewed-by: Daniel Kiper <[email protected]>
10205
10206 2023-05-17 Xiaotian Wu <[email protected]>
10207
10208 loongarch: Add early startup code
10209 On entry, we need to save the system table pointer as well as our image
10210 handle. Add an early startup file that saves them and then brings us
10211 into our main function.
10212
10213 Reviewed-by: Daniel Kiper <[email protected]>
10214
10215 2023-05-17 Xiaotian Wu <[email protected]>
10216
10217 loongarch: Add setjmp implementation
10218 This patch adds a setjmp implementation for LoongArch.
10219
10220 Reviewed-by: Daniel Kiper <[email protected]>
10221
10222 2023-05-17 Xiaotian Wu <[email protected]>
10223
10224 elf: Add LoongArch definitions
10225 Add ELF e_machine ID [1] and relocations types [2] for LoongArch to
10226 the current in-repo definitions.
10227
10228 [1] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html#_e_machine_identifies_the_machine
10229 [2] https://loongson.github.io/LoongArch-Documentation/LoongArch-ELF-ABI-EN.html#_relocations
10230
10231 Reviewed-by: Daniel Kiper <[email protected]>
10232
10233 2023-05-17 Xiaotian Wu <[email protected]>
10234
10235 pe: Add LoongArch definitions
10236 Add PE machine types [1] and relocation types [2] for LoongArch to
10237 the current in-repo definitions.
10238
10239 [1] https://learn.microsoft.com/en-us/windows/win32/debug/pe-format#machine-types
10240 [2] https://learn.microsoft.com/en-us/windows/win32/debug/pe-format#base-relocation-types
10241
10242 Reviewed-by: Daniel Kiper <[email protected]>
10243
10244 2023-05-16 Chris Coulson <[email protected]>
10245
10246 font: Try opening fonts from the bundled memdisk
10247 GRUB since 93a786a00 (kern/efi/sb: Enforce verification of font files)
10248 has enforced verification of font files in secure boot mode. In order to
10249 continue to be able to load some default fonts, vendors may bundle them
10250 with their signed EFI image by adding them to the built-in memdisk.
10251
10252 This change makes the font loader try loading fonts from the memdisk
10253 before the prefix path when attempting to load a font file by specifying
10254 its filename, which avoids having to make changes to GRUB configurations
10255 in order to accommodate memdisk bundled fonts. It expects the directory
10256 structure to be the same as fonts stored in the prefix path,
10257 i.e. /fonts/<name>.pf2.
10258
10259 Reviewed-by: Steve McIntyre <[email protected]>
10260 Tested-by: Steve McIntyre <[email protected]>
10261 Reviewed-by: Robbie Harwood <[email protected]>
10262 Reviewed-by: Daniel Kiper <[email protected]>
10263
10264 2023-05-16 Robbie Harwood <[email protected]>
10265 Aaron Miller <[email protected]>
10266 Peter Jones <[email protected]>
10267
10268 net: Read bracketed IPv6 addrs and port numbers
10269 Allow specifying port numbers for http and tftp paths and allow IPv6
10270 addresses to be recognized with brackets around them, which is required
10271 to specify a port number.
10272
10273 Reviewed-by: Daniel Kiper <[email protected]>
10274
10275 2023-05-16 Robbie Harwood <[email protected]>
10276
10277 Revert "net/http: Allow use of non-standard TCP/IP ports"
10278 The notation introduced in ac8a37dda (net/http: Allow use of non-standard
10279 TCP/IP ports) contradicts that used in downstream distributions including
10280 Fedora, RHEL, Debian, Ubuntu, and others. Revert it and apply the downstream
10281 notation which was originally proposed to the GRUB in 2016.
10282
10283 This reverts commit ac8a37dda (net/http: Allow use of non-standard TCP/IP ports).
10284
10285 Reviewed-by: Daniel Kiper <[email protected]>
10286
10287 2023-05-16 Riku Viitanen <[email protected]>
10288
10289 term/at_keyboard: Add timeout to fix hang on HP EliteBooks
10290 This fixes the GRUB on Coreboot on HP EliteBooks by implementing
10291 a 200 ms timeout. The GRUB used to hang.
10292
10293 Fixes: https://ticket.coreboot.org/issues/141
10294
10295 Reviewed-by: Daniel Kiper <[email protected]>
10296
10297 2023-04-13 Glenn Washburn <[email protected]>
10298
10299 tests/util/grub-fs-tester: Add missing redirect to /dev/null
10300 In filesystem timestamp test, a check is done to verify that the timestamp
10301 for a file as reported in Linux by the filesystem is within a few seconds
10302 of the timestamp as reported by GRUB. This is done by grepping the output
10303 of GRUB's ls command for the timestamp as reported by the filesystem in
10304 Linux and for each of 3 seconds past that timestamp. All of these checks
10305 except one redirect the output of grep to /dev/null. Fix this exception
10306 to behave as the other checks.
10307
10308 Reviewed-by: Daniel Kiper <[email protected]>
10309
10310 2023-04-13 Mukesh Kumar Chaurasiya <[email protected]>
10311
10312 disk: Replace transform_sector() function with grub_disk_to_native_sector()
10313 The transform_sector() function is not very clear in what it's doing
10314 and confusing. The GRUB already has a function which is doing the same
10315 thing in a very self explanatory way, i.e., grub_disk_to_native_sector().
10316 So, it's much better to use self explanatory one than transform_sector().
10317
10318 Reviewed-by: Daniel Kiper <[email protected]>
10319
10320 2023-04-13 Thomas Schmitt <[email protected]>
10321
10322 tests: Add test for iso9660 delayed CE hop
10323 The ISO filesystem image iso9660_early_ce.iso exposes the unusual
10324 situation that the Rock Ridge name entry of its only file is located
10325 after a CE entry which points to the next continuation area.
10326
10327 The correct behavior is to read the Rock Ridge name and to only then
10328 load the next continuation area. If GRUB performs this correctly, then
10329 the name "RockRidgeName:x" will be read and reported by grub-fstest.
10330 If GRUB wrongly performs the CE hop immediately when encountering the CE
10331 entry, then the dull ISO 9660 name "rockridg" will not be overridden and
10332 be put out by grub-fstest.
10333
10334 Tested-by: Lidong Chen <[email protected]>
10335 Reviewed-by: Daniel Kiper <[email protected]>
10336
10337 2023-04-13 Thomas Schmitt <[email protected]>
10338
10339 fs/iso9660: Delay CE hop until end of current SUSP area
10340 The SUSP specs demand that the reading of the next SUSP area which is
10341 depicted by a CE entry shall be delayed until reading of the current
10342 SUSP area is completed. Up to now GRUB immediately ends reading of the
10343 current area and loads the new one. So, buffer the parameters of a found
10344 CE entry and perform checks and reading of new data only after the
10345 reader loop has ended.
10346
10347 Tested-by: Lidong Chen <[email protected]>
10348 Reviewed-by: Daniel Kiper <[email protected]>
10349
10350 2023-03-29 Avnish Chouhan <[email protected]>
10351
10352 kern/ieee1275/init: Extended support in Vec5
10353 This patch enables multiple options in Vec5 which are required and
10354 solves the boot issues seen on some machines which are looking for
10355 these specific options.
10356
10357 1. LPAR: Client program supports logical partitioning and
10358 associated hcall()s.
10359 2. SPLPAR: Client program supports the Shared
10360 Processor LPAR Option.
10361 3. DYN_RCON_MEM: Client program supports the
10362 “ibm,dynamic-reconfiguration-memory” property and it may be
10363 presented in the device tree.
10364 4. LARGE_PAGES: Client supports pages larger than 4 KB.
10365 5. DONATE_DCPU_CLS: Client supports donating dedicated processor cycles.
10366 6. PCI_EXP: Client supports PCI Express implementations
10367 utilizing Message Signaled Interrupts (MSIs).
10368
10369 7. CMOC: Enables the Cooperative Memory Over-commitment Option.
10370 8. EXT_CMO: Enables the Extended Cooperative Memory Over-commit Option.
10371
10372 9. ASSOC_REF: Enables “ibm,associativity” and
10373 “ibm,associativity-reference-points” properties.
10374 10. AFFINITY: Enables Platform Resource Reassignment Notification.
10375 11. NUMA: Supports NUMA Distance Lookup Table Option.
10376
10377 12. HOTPLUG_INTRPT: Supports Hotplug Interrupts.
10378 13. HPT_RESIZE: Enable Hash Page Table Resize Option.
10379
10380 14. MAX_CPU: Defines maximum number of CPUs supported.
10381
10382 15. PFO_HWRNG: Supports Random Number Generator.
10383 16. PFO_HW_COMP: Supports Compression Engine.
10384 17. PFO_ENCRYPT: Supports Encryption Engine.
10385
10386 18. SUB_PROCESSORS: Supports Sub-Processors.
10387
10388 19. DY_MEM_V2: Client program supports the “ibm,dynamic-memory-v2” property in the
10389 “ibm,dynamic-reconfiguration-memory” node and it may be presented in the device tree.
10390 20. DRC_INFO: Client program supports the “ibm,drc-info” property definition and it may be
10391 presented in the device tree.
10392
10393 Reviewed-by: Daniel Kiper <[email protected]>
10394
10395 2023-03-29 Avnish Chouhan <[email protected]>
10396
10397 kern/ieee1275/init: Convert plain numbers to constants in Vec5
10398 This patch converts the plain numbers used in Vec5 properties to constants.
10399
10400 1. LPAR: Client program supports logical partitioning and
10401 associated hcall()s.
10402 2. SPLPAR: Client program supports the Shared
10403 Processor LPAR Option.
10404 3. CMO: Enables the Cooperative Memory Over-commitment Option.
10405 4. MAX_CPU: Defines maximum number of CPUs supported.
10406
10407 Reviewed-by: Daniel Kiper <[email protected]>
10408
10409 2023-03-29 Robbie Harwood <[email protected]>
10410
10411 loader/emu/linux: Work around systemctl kexec returning
10412 Per systemctl(1), it "is asynchronous; it will return after the reboot
10413 operation is enqueued, without waiting for it to complete". This differs
10414 from kexec(8), which calls reboot(2) and therefore does not return.
10415
10416 When not using fallback, this confusingly results in:
10417
10418 error trying to perform 'systemctl kexec': 0
10419 Aborted. Press any key to exit.
10420
10421 on screen for a bit, followed by successful kexec.
10422
10423 To reduce the likelihood of hitting this case, add a delay on successful
10424 return. Ultimately, the systemd interface is racy: we can't avoid it
10425 entirely unless we never fallback on success.
10426
10427 Reviewed-by: Daniel Kiper <[email protected]>
10428
10429 2023-03-29 Michael Chang <[email protected]>
10430
10431 tpm: Disable the tpm verifier if the TPM device is not present
10432 When the tpm module is loaded, the verifier reads entire file into
10433 memory, measures it and uses verified content as a backing buffer for
10434 file accesses. However, this process may result in high memory
10435 utilization for file operations, sometimes causing a system to run out
10436 of memory which may finally lead to boot failure. To address this issue,
10437 among others, the commit 887f98f0d (mm: Allow dynamically requesting
10438 additional memory regions) have optimized memory management by
10439 dynamically allocating heap space to maximize memory usage and reduce
10440 threat of memory exhaustion. But in some cases problems may still arise,
10441 e.g., when large ISO images are mounted using loopback or when dealing
10442 with embedded systems with limited memory resources.
10443
10444 Unfortunately current implementation of the tpm module doesn't allow
10445 elimination of the back buffer once it is loaded. Even if the TPM device
10446 is not present or it has been explicitly disabled. This may unnecessary
10447 allocate a lot memory. To solve this issue, a patch has been developed
10448 to detect the TPM status at module load and skip verifier registration
10449 if the device is missing or deactivated. This prevents allocation of
10450 memory for the back buffer, avoiding wasting memory when no real measure
10451 boot functionality is performed. Disabling the TPM device in the system
10452 can reduce memory usage in the GRUB. It is useful in scenarios where
10453 high memory utilization is a concern and measurements of loaded
10454 artifacts are not necessary.
10455
10456 Reviewed-by: Daniel Kiper <[email protected]>
10457
10458 2023-03-29 Glenn Washburn <[email protected]>
10459
10460 INSTALL: Document programs and packages needed for using gdb_grub script
10461 Now that the gdb_grub script uses the Python API in GDB, a GDB with Python
10462 support must be used. Note that this means a GDB with version greater than
10463 7.0 must be used. This should not be an issue since that was released over
10464 a decade ago. Also, the minimum version of Python must be 3.5, which was
10465 released around 8 years ago.
10466
10467 Reviewed-by: Daniel Kiper <[email protected]>
10468
10469 2023-03-29 Atish Patra <[email protected]>
10470
10471 RISC-V: Use common linux loader
10472 RISC-V doesn't have to do anything very different from other architectures
10473 to loader EFI stub linux kernel. As a result, just use the common linux
10474 loader instead of defining a RISC-V specific linux loader.
10475
10476 Reviewed-by: Daniel Kiper <[email protected]>
10477
10478 2023-03-29 Atish Patra <[email protected]>
10479
10480 efi: Remove arch specific image headers for RISC-V, ARM64 and ARM
10481 The arch specific image header details are not very useful as most of
10482 the GRUB just looks at the PE/COFF spec parameters (PE32 magic and
10483 header offset).
10484
10485 Remove the arch specific images headers and define a generic arch
10486 headers that provide enough PE/COFF fields for the GRUB to parse
10487 kernel images correctly.
10488
10489 Reviewed-by: Daniel Kiper <[email protected]>
10490
10491 2023-03-29 Atish Patra <[email protected]>
10492
10493 loader/efi: Move ARM64 linux loader to common code
10494 ARM64 linux loader code is written in such a way that it can be reused
10495 across different architectures without much change. Move it to common
10496 code so that RISC-V doesn't have to define a separate loader.
10497
10498 Reviewed-by: Daniel Kiper <[email protected]>
10499
10500 2023-03-14 Alec Brown <[email protected]>
10501
10502 util/grub-module-verifierXX: Add module_size parameter to functions for sanity checking
10503 In grub-module-verifierXX.c, the function grub_module_verifyXX() performs an
10504 initial check that the ELF section headers are within the module's size, but
10505 doesn't check if the sections being accessed have contents that are within the
10506 module's size. In particular, we need to check that sh_offset and sh_size are
10507 less than the module's size. However, for some section header types we don't
10508 need to make these checks. For the type SHT_NULL, the section header is marked
10509 as inactive and the rest of the members within the section header have undefined
10510 values, so we don't need to check for sh_offset or sh_size. In the case of the
10511 type SHT_NOBITS, sh_offset has a conceptual offset which may be beyond the
10512 module size. Also, this type's sh_size may have a non-zero size, but a section
10513 of this type will take up no space in the module. This can all be checked in the
10514 function get_shdr(), but in order to do so, the parameter module_size must be
10515 added to functions so that the value of the module size can be used in
10516 get_shdr() from grub_module_verifyXX().
10517
10518 Also, had to rework some for loops to ensure the index passed to get_shdr() is
10519 within bounds.
10520
10521 Reviewed-by: Daniel Kiper <[email protected]>
10522
10523 2023-03-14 Glenn Washburn <[email protected]>
10524
10525 gdb: Add extra early initialization symbols for i386-pc
10526 Add symbols for boot.image, disk.image, and lzma_decompress.image if the
10527 target is i386-pc. This is only done for i386-pc because that is the only
10528 target that uses the images. By loading the symbols for these images,
10529 these images can be more easily debugged by allowing the setting of break-
10530 points in that code and to see easily get the value of data symbols.
10531
10532 Reviewed-by: Daniel Kiper <[email protected]>
10533
10534 2023-03-14 Glenn Washburn <[email protected]>
10535
10536 gdb: Modify gdb prompt when running gdb_grub script
10537 This will let users know that the GDB session is using the GRUB gdb scripts.
10538
10539 Reviewed-by: Daniel Kiper <[email protected]>
10540
10541 2023-03-14 Glenn Washburn <[email protected]>
10542
10543 gdb: Allow running user-defined commands at GRUB start
10544 A new command, run_on_start, for things to do before GRUB starts executing.
10545 Currently, this is setting up the loading of module symbols as they are
10546 loaded and allowing user-defined script to be run if a command named
10547 "onstart" exists.
10548
10549 On some platforms, notably x86, software breakpoints set in GDB before
10550 the GRUB image is loaded will be cleared when the image is loaded. This
10551 is because the breakpoints work by overwriting the memory of the break-
10552 point location with a special instruction which when hit will cause the
10553 debugger to stop execution. Just before execution is resumed by the
10554 debugger, the original instruction bytes are put back. When a breakpoint
10555 is set before the GRUB image is loaded, the special debugger instruction
10556 will be written to memory and when the GRUB image is loaded by the
10557 firmware, which has no knowledge of the debugger, the debugger instruction
10558 is overwritten. To the GDB user, GDB will show the breakpoint as set, but
10559 it will never be hit. Furthermore, GDB now becomes confused, such that
10560 even deleting and re-setting the breakpoint after the GRUB image is loaded
10561 will not allow for a working breakpoint.
10562
10563 To work around this, in run_on_start, first a watchpoint is set on _start,
10564 which will be triggered when the firmware starts loading the GRUB image.
10565 When the _start watchpoint is hit, the current breakpoints are saved to a
10566 file and then deleted by GDB before they can be overwritten by the firmware
10567 and confuse GDB. Then a temporary software breakpoint is set on _start,
10568 which will get triggered when the firmware hands off to GRUB to execute. In
10569 that breakpoint load the previously saved and deleted breakpoints now that
10570 there is no worry of them getting overwritten by the firmware. This is
10571 needed for runtime_load_module to work when it is run before the GRUB image
10572 is loaded.
10573
10574 Note that watchpoints are generally types of hardware breakpoints on x86, so
10575 its deleted as soon as it gets triggered so that a minimal set of hardware
10576 breakpoints are used, allowing more for the user.
10577
10578 Reviewed-by: Daniel Kiper <[email protected]>
10579
10580 2023-03-14 Glenn Washburn <[email protected]>
10581
10582 gdb: Add functions to make loading from dynamically positioned targets easier
10583 Many targets, such as EFI, load GRUB at addresses that are determined at
10584 runtime. So the load addresses in kernel.exec will almost certainly be
10585 wrong. Given the address of the start of the text segment, these
10586 functions will tell GDB to load the symbols at the proper locations. It
10587 is left up to the user to determine how to get the text address of the
10588 loaded GRUB image.
10589
10590 Reviewed-by: Daniel Kiper <[email protected]>
10591
10592 2023-03-14 Glenn Washburn <[email protected]>
10593
10594 gdb: Replace module symbol loading implementation with Python one
10595 Remove gmodule.pl and rewrite as a python in gdb_helper.py. This removes
10596 Perl dependency for the GRUB GDB script, but adds Python as a dependency.
10597 This is more desirable because Python is tightly integrated with GDB and
10598 can do things not even available to GDB native scripting language. GDB must
10599 be built with Python, however this is not a major limitation because every
10600 major distro non-end-of-life versions build GDB with Python support. And GDB
10601 has had support for Python since around 7.1-ish, which is about a decade.
10602
10603 This re-implementation has an added feature. If there is a user defined
10604 command named "onload_<module name>", then that command will be executed
10605 after the symbols for the specified module are loaded. When debugging a
10606 module it can be desirable to set break points on code in the module.
10607 This is difficult in GRUB because, at GDB start, the module is not loaded
10608 and on EFI platforms its not known ahead of time where the module will
10609 be loaded. So allow users to create an "onload_<modname>" command which
10610 will be run when the module with name "modname" is loaded.
10611
10612 Another addition is a new convenience function is defined
10613 $is_user_command(), which returns true if its string argument is
10614 the name of a user-defined command.
10615
10616 A secondary benefit of these changes is that the script does not write
10617 temporary files and has better error handling capabilities.
10618
10619 Reviewed-by: Daniel Kiper <[email protected]>
10620
10621 2023-03-14 Glenn Washburn <[email protected]>
10622
10623 gdb: Only connect to remote target once when first sourced
10624 The gdb_grub script was originally meant to be run once when GDB first
10625 starts up via the -x argument. So it runs commands unconditionally
10626 assuming that the script has not been run before. Its nice to be able
10627 to source the script again when developing the script to modify/add
10628 commands. So only run the commands not defined in user-defined commands,
10629 if a variable $runonce has already been set and when those commands have
10630 been run to set $runonce.
10631
10632 Reviewed-by: Daniel Kiper <[email protected]>
10633
10634 2023-03-14 Glenn Washburn <[email protected]>
10635
10636 gdb: Conditionally run GDB script logic for dynamically or statically positioned GRUB
10637 There are broadly two classes of targets to consider when loading symbols
10638 for GRUB, targets that determine where to load GRUB at runtime
10639 (dynamically positioned) and those that do not (statically positioned).
10640 For statically positioned targets, symbol loading is determined at link
10641 time, so nothing more needs to be known to load the symbols. For
10642 dynamically positioned targets, such as EFI targets, at runtime symbols
10643 should be offset by an amount that depends on where the runtime chose to
10644 load GRUB.
10645
10646 It is important to not load symbols statically for dynamic targets
10647 because then when subsequently loading the symbols correctly one must
10648 take care to remove the existing static symbols, otherwise there will be
10649 two sets of symbols and GDB seems to prefer the ones loaded first (i.e.
10650 the static ones).
10651
10652 Use autoconf variables to generate a gdb_grub for a particular target,
10653 which conditionally run startup code depending on if the target uses
10654 static or dynamic loading.
10655
10656 Reviewed-by: Daniel Kiper <[email protected]>
10657
10658 2023-03-14 Glenn Washburn <[email protected]>
10659
10660 gdb: Move runtime module loading into runtime_load_module
10661 By moving this code into a function, it can be run re-utilized while gdb is
10662 running, not just when loading the script. This will also be useful in
10663 some following changes which will make a separate script path for targets
10664 which statically vs dynamically position GRUB code.
10665
10666 Reviewed-by: Daniel Kiper <[email protected]>
10667
10668 2023-03-07 Michael Chang <[email protected]>
10669
10670 osdep/devmapper/getroot: Fix build error on 32-bit host
10671 The gcc build has failed for 32-bit host (e.g. i386-emu and arm-emu)
10672 due to mismatch between format specifier and data type.
10673
10674 ../grub-core/osdep/devmapper/getroot.c: In function
10675 'grub_util_pull_devmapper':
10676
10677 ../grub-core/osdep/devmapper/getroot.c:265:75: error: format '%lu'
10678 expects argument of type 'long unsigned int', but argument 2 has type
10679 'int' [-Werror=format=]
10680
10681 ../grub-core/osdep/devmapper/getroot.c:276:80: error: format '%lu'
10682 expects argument of type 'long unsigned int', but argument 2 has type
10683 'int' [-Werror=format=]
10684
10685 This patch fixes the problem by casting the type of calculated offset to
10686 grub_size_t and use platform PRIuGRUB_SIZE as format specifier.
10687
10688 Reviewed-by: Daniel Kiper <[email protected]>
10689
10690 2023-03-07 Stefan Berger <[email protected]>
10691
10692 commands/ieee1275/ibmvtpm: Add support for trusted boot using a vTPM 2.0
10693 Add support for trusted boot using a vTPM 2.0 on the IBM IEEE1275
10694 PowerPC platform. With this patch grub now measures text and binary data
10695 into the TPM's PCRs 8 and 9 in the same way as the x86_64 platform
10696 does.
10697
10698 This patch requires Daniel Axtens's patches for claiming more memory.
10699
10700 Note: The tpm_init() function cannot be called from GRUB_MOD_INIT() since
10701 it does not find the device nodes upon module initialization and
10702 therefore the call to tpm_init() must be deferred to grub_tpm_measure().
10703
10704 For vTPM support to work on PowerVM, system driver levels 1010.30
10705 or 1020.00 are required.
10706
10707 Note: Previous versions of firmware levels with the 2hash-ext-log
10708 API call have a bug that, once this API call is invoked, has the
10709 effect of disabling the vTPM driver under Linux causing an error
10710 message to be displayed in the Linux kernel log. Those users will
10711 have to update their machines to the firmware levels mentioned
10712 above.
10713
10714 Cc: Eric Snowberg <[email protected]>
10715 Reviewed-by: Daniel Kiper <[email protected]>
10716 Tested-by: Nageswara R Sastry <[email protected]>
10717 Reviewed-by: Robbie Harwood <[email protected]>
10718
10719 2023-03-07 Daniel Axtens <[email protected]>
10720
10721 commands/memtools: Add memtool module with memory allocation stress-test
10722 When working on memory, it's nice to be able to test your work.
10723
10724 Add a memtest module. When compiled with --enable-mm-debug, it exposes
10725 3 commands:
10726
10727 * lsmem - print all allocations and free space in all regions
10728 * lsfreemem - print free space in all regions
10729
10730 * stress_big_allocs - stress test large allocations:
10731 - how much memory can we allocate in one chunk?
10732 - how many 1MB chunks can we allocate?
10733 - check that gap-filling works with a 1MB aligned 900kB alloc + a
10734 100kB alloc.
10735
10736 Reviewed-by: Daniel Kiper <[email protected]>
10737 Tested-by: Nageswara R Sastry <[email protected]>
10738 Reviewed-by: Robbie Harwood <[email protected]>
10739
10740 2023-03-07 Diego Domingos <[email protected]>
10741
10742 ieee1275: Implement vec5 for cas negotiation
10743 As a legacy support, if the vector 5 is not implemented, Power Hypervisor will
10744 consider the max CPUs as 64 instead 256 currently supported during
10745 client-architecture-support negotiation.
10746
10747 This patch implements the vector 5 and set the MAX CPUs to 256 while setting the
10748 others values to 0 (default).
10749
10750 Acked-by: Daniel Axtens <[email protected]>
10751 Tested-by: Nageswara R Sastry <[email protected]>
10752 Reviewed-by: Robbie Harwood <[email protected]>
10753 Reviewed-by: Daniel Kiper <[email protected]>
10754
10755 2023-03-07 Daniel Axtens <[email protected]>
10756
10757 ieee1275: Support runtime memory claiming
10758 On powerpc-ieee1275, we are running out of memory trying to verify
10759 anything. This is because:
10760
10761 - we have to load an entire file into memory to verify it. This is
10762 difficult to change with appended signatures.
10763 - We only have 32MB of heap.
10764 - Distro kernels are now often around 30MB.
10765
10766 So we want to be able to claim more memory from OpenFirmware for our heap
10767 at runtime.
10768
10769 There are some complications:
10770
10771 - The grub mm code isn't the only thing that will make claims on
10772 memory from OpenFirmware:
10773
10774 * PFW/SLOF will have claimed some for their own use.
10775
10776 * The ieee1275 loader will try to find other bits of memory that we
10777 haven't claimed to place the kernel and initrd when we go to boot.
10778
10779 * Once we load Linux, it will also try to claim memory. It claims
10780 memory without any reference to /memory/available, it just starts
10781 at min(top of RMO, 768MB) and works down. So we need to avoid this
10782 area. See arch/powerpc/kernel/prom_init.c as of v5.11.
10783
10784 - The smallest amount of memory a ppc64 KVM guest can have is 256MB.
10785 It doesn't work with distro kernels but can work with custom kernels.
10786 We should maintain support for that. (ppc32 can boot with even less,
10787 and we shouldn't break that either.)
10788
10789 - Even if a VM has more memory, the memory OpenFirmware makes available
10790 as Real Memory Area can be restricted. Even with our CAS work, an LPAR
10791 on a PowerVM box is likely to have only 512MB available to OpenFirmware
10792 even if it has many gigabytes of memory allocated.
10793
10794 What should we do?
10795
10796 We don't know in advance how big the kernel and initrd are going to be,
10797 which makes figuring out how much memory we can take a bit tricky.
10798
10799 To figure out how much memory we should leave unused, I looked at:
10800
10801 - an Ubuntu 20.04.1 ppc64le pseries KVM guest:
10802 vmlinux: ~30MB
10803 initrd: ~50MB
10804
10805 - a RHEL8.2 ppc64le pseries KVM guest:
10806 vmlinux: ~30MB
10807 initrd: ~30MB
10808
10809 So to give us a little wriggle room, I think we want to leave at least
10810 128MB for the loader to put vmlinux and initrd in memory and leave Linux
10811 with space to satisfy its early allocations.
10812
10813 Allow other space to be allocated at runtime.
10814
10815 Tested-by: Stefan Berger <[email protected]>
10816 Tested-by: Nageswara R Sastry <[email protected]>
10817 Reviewed-by: Robbie Harwood <[email protected]>
10818 Reviewed-by: Daniel Kiper <[email protected]>
10819
10820 2023-03-07 Daniel Axtens <[email protected]>
10821
10822 ieee1275: Drop len -= 1 quirk in heap_init
10823 This was apparently "required by some firmware": commit dc9468500919
10824 (2007-02-12 Hollis Blanchard <[email protected]>).
10825
10826 It's not clear what firmware that was, and what platform from 14 years ago
10827 which exhibited the bug then is still both in use and buggy now.
10828
10829 It doesn't cause issues on qemu (mac99 or pseries) or under PFW for Power8.
10830
10831 I don't have access to old Mac hardware, but if anyone feels especially
10832 strongly we can put it under some feature flag. I really want to disable
10833 it under pseries because it will mess with region merging.
10834
10835 Reviewed-by: Daniel Kiper <[email protected]>
10836 Tested-by: Nageswara R Sastry <[email protected]>
10837 Reviewed-by: Robbie Harwood <[email protected]>
10838
10839 2023-03-07 Daniel Axtens <[email protected]>
10840
10841 ieee1275: Request memory with ibm, client-architecture-support
10842 On PowerVM, the first time we boot a Linux partition, we may only get
10843 256MB of real memory area, even if the partition has more memory.
10844
10845 This isn't enough to reliably verify a kernel. Fortunately, the Power
10846 Architecture Platform Reference (PAPR) defines a method we can call to ask
10847 for more memory: the broad and powerful ibm,client-architecture-support
10848 (CAS) method.
10849
10850 CAS can do an enormous amount of things on a PAPR platform: as well as
10851 asking for memory, you can set the supported processor level, the interrupt
10852 controller, hash vs radix mmu, and so on.
10853
10854 If:
10855
10856 - we are running under what we think is PowerVM (compatible property of /
10857 begins with "IBM"), and
10858
10859 - the full amount of RMA is less than 512MB (as determined by the reg
10860 property of /memory)
10861
10862 then call CAS as follows: (refer to the Linux on Power Architecture
10863 Reference, LoPAR, which is public, at B.5.2.3):
10864
10865 - Use the "any" PVR value and supply 2 option vectors.
10866
10867 - Set option vector 1 (PowerPC Server Processor Architecture Level)
10868 to "ignore".
10869
10870 - Set option vector 2 with default or Linux-like options, including a
10871 min-rma-size of 512MB.
10872
10873 - Set option vector 3 to request Floating Point, VMX and Decimal Floating
10874 point, but don't abort the boot if we can't get them.
10875
10876 - Set option vector 4 to request a minimum VP percentage to 1%, which is
10877 what Linux requests, and is below the default of 10%. Without this,
10878 some systems with very large or very small configurations fail to boot.
10879
10880 This will cause a CAS reboot and the partition will restart with 512MB
10881 of RMA. Importantly, grub will notice the 512MB and not call CAS again.
10882
10883 Notes about the choices of parameters:
10884
10885 - A partition can be configured with only 256MB of memory, which would
10886 mean this request couldn't be satisfied, but PFW refuses to load with
10887 only 256MB of memory, so it's a bit moot. SLOF will run fine with 256MB,
10888 but we will never call CAS under qemu/SLOF because /compatible won't
10889 begin with "IBM".)
10890
10891 - unspecified CAS vectors take on default values. Some of these values
10892 might restrict the ability of certain hardware configurations to boot.
10893 This is why we need to specify the VP percentage in vector 4, which is
10894 in turn why we need to specify vector 3.
10895
10896 Finally, we should have enough memory to verify a kernel, and we will
10897 reach Linux. One of the first things Linux does while still running under
10898 OpenFirmware is to call CAS with a much fuller set of options (including
10899 asking for 512MB of memory). Linux includes a much more restrictive set of
10900 PVR values and processor support levels, and this CAS invocation will likely
10901 induce another reboot. On this reboot grub will again notice the higher RMA,
10902 and not call CAS. We will get to Linux again, Linux will call CAS again, but
10903 because the values are now set for Linux this will not induce another CAS
10904 reboot and we will finally boot all the way to userspace.
10905
10906 On all subsequent boots, everything will be configured with 512MB of RMA,
10907 so there will be no further CAS reboots from grub. (phyp is super sticky
10908 with the RMA size - it persists even on cold boots. So if you've ever booted
10909 Linux in a partition, you'll probably never have grub call CAS. It'll only
10910 ever fire the first time a partition loads grub, or if you deliberately lower
10911 the amount of memory your partition has below 512MB.)
10912
10913 Reviewed-by: Daniel Kiper <[email protected]>
10914 Tested-by: Nageswara R Sastry <[email protected]>
10915 Reviewed-by: Robbie Harwood <[email protected]>
10916
10917 2023-02-28 Khem Raj <[email protected]>
10918
10919 RISC-V: Handle R_RISCV_CALL_PLT reloc
10920 GNU assembler starting 2.40 release always generates R_RISCV_CALL_PLT
10921 reloc for call in assembler [1], similarly LLVM does not make
10922 distinction between R_RISCV_CALL_PLT and R_RISCV_CALL [2].
10923
10924 Fixes "grub-mkimage: error: relocation 0x13 is not implemented yet.".
10925
10926 [1] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=70f35d72ef04cd23771875c1661c9975044a749c
10927 [2] https://reviews.llvm.org/D132530
10928
10929 Reviewed-by: Daniel Kiper <[email protected]>
10930
10931 2023-02-28 John Paul Adrian Glaubitz <[email protected]>
10932
10933 osdep/hurd/getroot: Remove unused variables in grub_util_find_hurd_root_device()
10934 Found during a test build on Debian/hurd-i386 with --disable-werror enabled:
10935
10936 In file included from grub-core/osdep/getroot.c:12:
10937 grub-core/osdep/hurd/getroot.c: In function ‘grub_util_find_hurd_root_device’:
10938 grub-core/osdep/hurd/getroot.c:126:13: error: unused variable ‘next’ [-Werror=unused-variable]
10939 126 | char *next;
10940 | ^~~~
10941 grub-core/osdep/hurd/getroot.c:125:14: error: unused variable ‘size’ [-Werror=unused-variable]
10942 125 | size_t size;
10943 | ^~~~
10944
10945 Fixes: e981b0a24 (osdep/hurd/getroot: Use "part:" qualifier)
10946
10947 Reviewed-by: Samuel Thibault <[email protected]>
10948 Reviewed-by: Daniel Kiper <[email protected]>
10949
10950 2023-02-28 Glenn Washburn <[email protected]>
10951
10952 gdb: If no modules have been loaded, do not try to load module symbols
10953 This prevents load_all_modules from failing when called before any
10954 modules have been loaded. Failures in GDB user-defined functions cause
10955 any function which called them to also fail.
10956
10957 Reviewed-by: Daniel Kiper <[email protected]>
10958
10959 2023-02-28 Glenn Washburn <[email protected]>
10960
10961 gdb: Prevent wrapping when writing to .segments.tmp
10962 GDB logging is redirected to write .segments.tmp, which means that GDB
10963 will wrap lines longer than what it thinks is the screen width
10964 (typically 80 characters). When wrapping does occur it causes gmodule.pl
10965 to misbehave. So disable line wrapping by using GDB's "with" command so
10966 that its guaranteed to return the width to the previous value upon
10967 command completion.
10968
10969 Also disable command tracing when dumping the module sections because that
10970 output will go to .segments.tmp and thus cause gmodule.pl to misbehave.
10971
10972 Reviewed-by: Daniel Kiper <[email protected]>
10973
10974 2023-02-28 Glenn Washburn <[email protected]>
10975
10976 gdb: Fix redirection issue in dump_module_sections
10977 An error in any GDB command causes it to immediately abort with an error,
10978 this includes any command that calls that command. This leads to an issue
10979 in dump_module_sections where an error causes the command to exit without
10980 turning off file redirection. The user then ends up with a GDB command
10981 line where commands output nothing to the console.
10982
10983 Instead do the work of dump_module_sections in the command
10984 dump_module_sections_helper and run the command using GDB's pipe command
10985 which does the redirection and undoes the redirection when it finishes
10986 regardless of any errors in the command.
10987
10988 Also, remove .segments.tmp file prior to loading modules in case one was
10989 left from a previous run.
10990
10991 Reviewed-by: Daniel Kiper <[email protected]>
10992
10993 2023-02-28 Glenn Washburn <[email protected]>
10994
10995 efi: Allow expression as func argument to efi_call_* macros on all platforms
10996 On EFI platforms where EFI calls do not require a wrapper (notably i386-efi
10997 and arm64-efi), the func argument needs to be wrapped in parenthesis to
10998 allow valid syntax when func is an expression which evaluates to a function
10999 pointer. On EFI platforms that do need a wrapper, this was never an issue
11000 because func is passed to the C function wrapper as an argument and thus
11001 does not need parenthesis to be evaluated.
11002
11003 Reviewed-by: Daniel Kiper <[email protected]>
11004
11005 2023-02-28 Jeremy Szu <[email protected]>
11006
11007 loader/i386/linux: Correct wrong initrd address for debug
11008 The "addr" is used to request the memory with specific ranges but the real
11009 loadable address come from the relocator. Thus, print the final retrieved
11010 addresses, virtual and physical, for initrd.
11011
11012 On the occasion migrate to PRIxGRUB_ADDR and PRIxGRUB_SIZE format specifiers.
11013
11014 Reviewed-by: Daniel Kiper <[email protected]>
11015
11016 2023-02-28 Glenn Washburn <[email protected]>
11017
11018 INSTALL: Document that the functional test requires the package xfonts-unifont
11019 Reviewed-by: Thomas Schmitt <[email protected]>
11020 Reviewed-by: Daniel Kiper <[email protected]>
11021
11022 2023-02-28 Glenn Washburn <[email protected]>
11023
11024 tests: Return hard error for functional test when unicode.pf2 does not exist
11025 The functional test requires unicode.pf2 to run successfully, so
11026 explicitly have the test return ERROR when its not found.
11027
11028 Tested-by: Thomas Schmitt <[email protected]>
11029 Reviewed-by: Daniel Kiper <[email protected]>
11030
11031 2023-02-28 Glenn Washburn <[email protected]>
11032
11033 tests: grub_cmd_cryptomount should hard error when pre-requisites are not met
11034 Tests should be SKIP'd only when they do not apply to a particular target.
11035 Hard errors are for when the test should run but can not be setup properly.
11036
11037 Reviewed-by: Daniel Kiper <[email protected]>
11038
11039 2023-02-28 Glenn Washburn <[email protected]>
11040
11041 tests: Add pathological iso9660 filesystem tests
11042 These are not added to grub-fs-tester because they are not generated and
11043 none of the filesystem tests are run on these ISOs. The test is to run the
11044 command "ls /" on the ISO, and a failure is determined if the command
11045 times out, has non-zero return value or has any output.
11046
11047 Tested-by: Thomas Schmitt <[email protected]>
11048 Reviewed-by: Daniel Kiper <[email protected]>
11049
11050 2023-02-14 Mukesh Kumar Chaurasiya <[email protected]>
11051
11052 osdep/linux/hostdisk: Modify sector by sysfs as disk sector
11053 The disk sector size provided by sysfs file system considers the sector
11054 size of 512 irrespective of disk sector size, thus causing the read by
11055 the GRUB to an incorrect offset from what was originally intended.
11056
11057 Considering the 512 sector size of sysfs data the actual sector needs to
11058 be modified corresponding to disk sector size.
11059
11060 Reviewed-by: Daniel Kiper <[email protected]>
11061
11062 2023-02-14 Glenn Washburn <[email protected]>
11063
11064 tests/util/grub-fs-tester: Use shell variable instead of autoconf
11065 By using a shell variable that is set once by the expansion of an autoconf
11066 variable, the resulting script is more readable.
11067
11068 Reviewed-by: Daniel Kiper <[email protected]>
11069
11070 2023-02-14 Glenn Washburn <[email protected]>
11071
11072 tests/util/grub-fs-tester: Remove unused variable
11073 Reviewed-by: Daniel Kiper <[email protected]>
11074
11075 2023-02-14 Alec Brown <[email protected]>
11076
11077 net/bootp: Fix unchecked return value
11078 In the function send_dhcp_packet(), added an error check for the return
11079 value of grub_netbuff_push().
11080
11081 Fixes: CID 404614
11082
11083 Reviewed-by: Daniel Kiper <[email protected]>
11084
11085 2023-02-02 Zhang Boyang <[email protected]>
11086
11087 mm: Avoid complex heap growth math in hot path
11088 We do a lot of math about heap growth in hot path of grub_memalign().
11089 However, the result is only used if out of memory is encountered, which
11090 is seldom.
11091
11092 This patch moves these calculations away from hot path. These
11093 calculations are now only done if out of memory is encountered. This
11094 change can also help compiler to optimize integer overflow checks away.
11095
11096 Reviewed-by: Daniel Kiper <[email protected]>
11097
11098 2023-02-02 Zhang Boyang <[email protected]>
11099
11100 mm: Preallocate some space when adding new regions
11101 When grub_memalign() encounters out-of-memory, it will try
11102 grub_mm_add_region_fn() to request more memory from system firmware.
11103 However, it doesn't preallocate memory space for future allocation
11104 requests. In extreme cases, it requires one call to
11105 grub_mm_add_region_fn() for each memory allocation request. This can
11106 be very slow.
11107
11108 This patch introduces GRUB_MM_HEAP_GROW_EXTRA, the minimal heap growth
11109 granularity. The new region size is now set to the bigger one of its
11110 original value and GRUB_MM_HEAP_GROW_EXTRA. Thus, it will result in some
11111 memory space preallocated if current allocations request is small.
11112
11113 The value of GRUB_MM_HEAP_GROW_EXTRA is set to 1MB. If this value is
11114 smaller, the cost of small memory allocations will be higher. If this
11115 value is larger, more memory will be wasted and it might cause
11116 out-of-memory on machines with small amount of RAM.
11117
11118 Reviewed-by: Daniel Kiper <[email protected]>
11119
11120 2023-02-02 Zhang Boyang <[email protected]>
11121
11122 mm: Adjust new region size to take management overhead into account
11123 When grub_memalign() encounters out-of-memory, it will try
11124 grub_mm_add_region_fn() to request more memory from system firmware.
11125 However, the size passed to it doesn't take region management overhead
11126 into account. Adding a memory area of "size" bytes may result in a heap
11127 region of less than "size" bytes really available. Thus, the new region
11128 may not be adequate for current allocation request, confusing
11129 out-of-memory handling code.
11130
11131 This patch introduces GRUB_MM_MGMT_OVERHEAD to address the region
11132 management overhead (e.g. metadata, padding). The value of this new
11133 constant must be large enough to make sure grub_memalign(align, size)
11134 always succeeds after a successful call to
11135 grub_mm_init_region(addr, size + align + GRUB_MM_MGMT_OVERHEAD),
11136 for any given addr and size (assuming no integer overflow).
11137
11138 The size passed to grub_mm_add_region_fn() is now correctly adjusted,
11139 thus if grub_mm_add_region_fn() succeeded, current allocation request
11140 can always succeed.
11141
11142 Reviewed-by: Daniel Kiper <[email protected]>
11143
11144 2023-02-02 Glenn Washburn <[email protected]>
11145
11146 tests/util/grub-shell: Add $GRUB_QEMU_OPTS to run.sh to easily see unofficial QEMU arguments
11147 When re-running a failed test, even the non-standard grub-shell QEMU
11148 arguments should be preserved in the run.sh to more precisely replay
11149 the failed test run.
11150
11151 Reviewed-by: Daniel Kiper <[email protected]>
11152
11153 2023-02-02 Glenn Washburn <[email protected]>
11154
11155 tests/util/grub-shell: Create run.sh in working directory for easily running test again
11156 Now it becomes trivial to re-run a test from the output in its working
11157 directory. This also makes it easy to send a reproducible failing test to
11158 the mailing list. This has allowed a refactor so that the duplicated code
11159 to call QEMU has be condensed (e.g. the use of timeout and file descriptor
11160 redirection). The run.sh script will pass any arguments given to QEMU.
11161 This allows QEMU to be easily started in a state ready for GDB to be
11162 attached.
11163
11164 Reviewed-by: Daniel Kiper <[email protected]>
11165
11166 2023-02-02 Glenn Washburn <[email protected]>
11167
11168 tests: Allow turning on shell tracing from environment variables
11169 This allows turning on shell tracing for grub-shell and grub-fs-tester
11170 when its not practical or not possible to use command line arguments
11171 (e.g. from "make check"). Turn on tracing when the envvar is an integer
11172 greater than 1, since these can generate a lot of output. Since this
11173 change uses the environment variables to set the default value for debug
11174 in grub-shell, this allows enabling grub-shell's debug mode which will
11175 preserve various generated output files that are helpful for debugging
11176 tests.
11177
11178 Reviewed-by: Daniel Kiper <[email protected]>
11179
11180 2023-02-02 Glenn Washburn <[email protected]>
11181
11182 misc: Move *printf function declarations to same location
11183 Reviewed-by: Daniel Kiper <[email protected]>
11184
11185 2023-02-02 Thomas Schmitt <[email protected]>
11186
11187 fs/iso9660: Prevent skipping CE or ST at start of continuation area
11188 If processing of a SUSP CE entry leads to a continuation area which
11189 begins by entry CE or ST, then these entries were skipped without
11190 interpretation. In case of CE this would lead to premature end of
11191 processing the SUSP entries of the file. In case of ST this could
11192 cause following non-SUSP bytes to be interpreted as SUSP entries.
11193
11194 Tested-by: Lidong Chen <[email protected]>
11195 Reviewed-by: Thomas Schmitt <[email protected]>
11196 Reviewed-by: Daniel Kiper <[email protected]>
11197
11198 2023-02-02 Lidong Chen <[email protected]>
11199
11200 fs/iso9660: Incorrect check for entry boundary
11201 An SL entry consists of the entry info and the component area.
11202 The entry info should take up 5 bytes instead of sizeof(*entry).
11203 The area after the first 5 bytes is the component area. It is
11204 incorrect to use the sizeof(*entry) to check the entry boundary.
11205
11206 Reviewed-by: Thomas Schmitt <[email protected]>
11207 Reviewed-by: Daniel Kiper <[email protected]>
11208
11209 2023-02-02 Lidong Chen <[email protected]>
11210
11211 fs/iso9660: Avoid reading past the entry boundary
11212 Added a check for the SP entry data boundary before reading it.
11213
11214 Reviewed-by: Thomas Schmitt <[email protected]>
11215 Reviewed-by: Daniel Kiper <[email protected]>
11216
11217 2023-02-02 Lidong Chen <[email protected]>
11218
11219 fs/iso9660: Prevent read past the end of system use area
11220 In the code, the for loop advanced the entry pointer to the next entry before
11221 checking if the next entry is within the system use area boundary. Another
11222 issue in the code was that there is no check for the size of system use area.
11223 For a corrupted system, the size of system use area can be less than the size
11224 of minimum SUSP entry size (4 bytes). These can cause buffer overrun. The fixes
11225 added the checks to ensure the read is valid and within the boundary.
11226
11227 Reviewed-by: Thomas Schmitt <[email protected]>
11228 Reviewed-by: Daniel Kiper <[email protected]>
11229
11230 2023-02-02 Lidong Chen <[email protected]>
11231
11232 fs/iso9660: Add check to prevent infinite loop
11233 There is no check for the end of block when reading
11234 directory extents. It resulted in read_node() always
11235 read from the same offset in the while loop, thus
11236 caused infinite loop. The fix added a check for the
11237 end of the block and ensure the read is within directory
11238 boundary.
11239
11240 Reviewed-by: Thomas Schmitt <[email protected]>
11241 Reviewed-by: Daniel Kiper <[email protected]>
11242
11243 2023-02-01 Pierre-Louis Bonicoli <[email protected]>
11244
11245 grub-fs-tester: Add LUKS1 and LUKS2 support
11246 The logical sector size used by LUKS1 is 512 bytes and LUKS2 uses 512 to
11247 4069 bytes. The default password used is "pass", but can be overridden
11248 by setting the PASS environment variable. The device mapper name is set
11249 to the name of the temp directory so that its easy to correlate device
11250 mapper name with a particular test run. Also since this name is unique
11251 per test run, multiple simultaneous test runs are allowed.
11252
11253 Note that cryptsetup is passing the --disable-locks parameter to allow
11254 cryptsetup run successfully when /run/lock/cryptsetup is not accessible.
11255 Since the device mapper name is unique per test run, there is no need to
11256 worry about locking the device to serialize access.
11257
11258 Tested-by: Glenn Washburn <[email protected]>
11259 Reviewed-by: Patrick Steinhardt <[email protected]>
11260 Reviewed-by: Daniel Kiper <[email protected]>
11261
11262 2023-02-01 Josselin Poiret <[email protected]>
11263
11264 osdep/devmapper/getroot: Set up cheated LUKS2 cryptodisk mount from DM parameters
11265 This lets a LUKS2 cryptodisk have its cipher and hash filled out,
11266 otherwise they wouldn't be initialized if cheat mounted.
11267
11268 Tested-by: Glenn Washburn <[email protected]>
11269 Reviewed-by: Patrick Steinhardt <[email protected]>
11270 Reviewed-by: Daniel Kiper <[email protected]>
11271
11272 2023-02-01 Josselin Poiret <[email protected]>
11273
11274 osdep/devmapper/getroot: Have devmapper recognize LUKS2
11275 Changes UUID comparisons so that LUKS1 and LUKS2 are both recognized
11276 as being LUKS cryptodisks.
11277
11278 Tested-by: Glenn Washburn <[email protected]>
11279 Reviewed-by: Patrick Steinhardt <[email protected]>
11280 Reviewed-by: Daniel Kiper <[email protected]>
11281
11282 2023-02-01 Fabian Vogt <[email protected]>
11283
11284 disk/cryptodisk: When cheatmounting, use the sector info of the cheat device
11285 When using grub-probe with cryptodisk, the mapped block device from the host
11286 is used directly instead of decrypting the source device in GRUB code.
11287 In that case, the sector size and count of the host device needs to be used.
11288 This is especially important when using LUKS2, which does not assign
11289 total_sectors and log_sector_size when scanning, but only later when the
11290 segments in the JSON area are evaluated. With an unset log_sector_size,
11291 grub_device_open() complains.
11292
11293 This fixes grub-probe failing with
11294 "error: sector sizes of 1 bytes aren't supported yet.".
11295
11296 Reviewed-by: Patrick Steinhardt <[email protected]>
11297 Tested-by: Glenn Washburn <[email protected]>
11298 Reviewed-by: Glenn Washburn <[email protected]>
11299 Reviewed-by: Patrick Steinhardt <[email protected]>
11300 Reviewed-by: Daniel Kiper <[email protected]>
11301
11302 2023-01-19 Daniel Axtens <[email protected]>
11303
11304 fs/f2fs: Fix off-by-one error in nat journal entries check
11305 Oops. You're allowed to have up to n = NAT_JOURNAL_ENTRIES entries
11306 _inclusive_, because the loop below uses i < n, not i <= n. D'oh.
11307
11308 Fixes: 4bd9877f6216 (fs/f2fs: Do not read past the end of nat journal entries)
11309
11310 Reported-by: программист нект <[email protected]>
11311 Tested-by: программист нект <[email protected]>
11312 Reviewed-by: Daniel Kiper <[email protected]>
11313
11314 2023-01-19 Nicholas Vinson <[email protected]>
11315
11316 gentpl.py: Remove .interp section from .img files
11317 When building .img files, a .interp section from the .image files will
11318 sometimes be copied into the .img file. This additional section pushes
11319 the .img file beyond the 512-byte limit and causes grub-install to fail
11320 to run for i386-pc platforms.
11321
11322 Reviewed-by: Daniel Kiper <[email protected]>
11323
11324 2023-01-19 Glenn Washburn <[email protected]>
11325
11326 tests: Add cryptomount functional test
11327 The grub_cmd_cryptomount make check test performs some functional testing
11328 of cryptomount and by extension the underlying cryptodisk infrastructure.
11329
11330 A utility test script named grub-shell-luks-tester is created to handle the
11331 complexities of the testing, making it simpler to add new test cases in
11332 grub_cmd_cryptomount.
11333
11334 Reviewed-by: Daniel Kiper <[email protected]>
11335
11336 2023-01-19 Glenn Washburn <[email protected]>
11337
11338 tests/util/grub-shell: Add halt_cmd variable to testcase namespace
11339 This allows test case scripts to use the appropriate halt command for
11340 the built architecture to end execution early. Otherwise, test case
11341 scripts have no way to know the appropriate mechanism for halting the
11342 test case early.
11343
11344 Reviewed-by: Daniel Kiper <[email protected]>
11345
11346 2023-01-19 Glenn Washburn <[email protected]>
11347
11348 tests/util/grub-shell: Trim line should always be matched from the beginning of the line
11349 When turning on shell tracing the trim line will be output before we
11350 actually want to start the trim. However, in this case the trim line never
11351 starts from the beginning of the line. So start trimming from the correct
11352 line by matching from the beginning of the line.
11353
11354 Reviewed-by: Daniel Kiper <[email protected]>
11355
11356 2023-01-19 Glenn Washburn <[email protected]>
11357
11358 tests/util/grub-shell: Allow specifying non-default trim line contents
11359 This will be useful for tests that have unwanted output from setup. This is
11360 not documented because its only intended to be internal at the moment. Also,
11361 --no-trim is allowed to explicitly turn off trim.
11362
11363 Reviewed-by: Daniel Kiper <[email protected]>
11364
11365 2023-01-19 Glenn Washburn <[email protected]>
11366
11367 tests/util/grub-shell: Only cleanup working directory file if QEMU does not fail or timeout
11368 This keeps the generated files to aid in diagnosing the source of the failure.
11369
11370 Reviewed-by: Daniel Kiper <[email protected]>
11371
11372 2023-01-19 Glenn Washburn <[email protected]>
11373
11374 tests/util/grub-shell: Set exit status to QEMU exit status
11375 This allows us to test if unexpected output in test scripts is because of
11376 a bug in GRUB, because there was an error in QEMU, or QEMU was killed due
11377 to a timeout.
11378
11379 Reviewed-by: Daniel Kiper <[email protected]>
11380
11381 2023-01-19 Glenn Washburn <[email protected]>
11382
11383 io/gzio: Remove confusing, out-dated comment
11384 The "transparent" parameter to grub_gzio_open() was removed in 2010, fc2ef1172c
11385 (* grub-core/io/gzio.c (grub_gzio_open): Removed "transparent" parameter.)
11386
11387 Reviewed-by: Daniel Kiper <[email protected]>
11388
11389 2023-01-19 Glenn Washburn <[email protected]>
11390
11391 efi: Fix spacing
11392 Reviewed-by: Daniel Kiper <[email protected]>
11393
11394 misc: Fix spacing
11395 Reviewed-by: Daniel Kiper <[email protected]>
11396
11397 misc: Spelling fixes
11398 Reviewed-by: Daniel Kiper <[email protected]>
11399
11400 gdb: Unregister gdbstub_break command when unloading module
11401 Reviewed-by: Daniel Kiper <[email protected]>
11402
11403 2023-01-19 Glenn Washburn <[email protected]>
11404
11405 tests: Fix help test to reflect updated help output
11406 Commit f5759a878 (normal/help: Add paging instructions to normal and help
11407 prompts) changed the output of the help command, which broke the help
11408 test. This change allows the test to pass.
11409
11410 On the occasion do s/outpu/output/.
11411
11412 Reviewed-by: Daniel Kiper <[email protected]>
11413
11414 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11415
11416 term/serial: Improve detection of duplicate serial ports
11417 We currently rely on some pretty fragile comparison by name to
11418 identify whether a serial port being configured is identical
11419
11420 Reviewed-by: Daniel Kiper <[email protected]>
11421
11422 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11423
11424 term/serial: Avoid double lookup of serial ports
11425 The various functions to add a port used to return port->name, and
11426 the callers would immediately iterate all registered ports to "find"
11427 the one just created by comparing that return value with ... port->name.
11428
11429 This is a waste of cycles and code. Instead, have those functions
11430 return "port" directly.
11431
11432 Reviewed-by: Daniel Kiper <[email protected]>
11433
11434 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11435
11436 term/serial: Replace usage of memcmp() with strncmp()
11437 We are comparing strings after all.
11438
11439 Reviewed-by: Daniel Kiper <[email protected]>
11440
11441 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11442
11443 term/serial: Add ability to specify MMIO ports via "serial" command
11444 This adds the ability to explicitly add an MMIO based serial port
11445 via the "serial" command. The syntax is:
11446
11447 serial --port=mmio,<hex_address>{.b,.w,.l,.q}
11448
11449 Reviewed-by: Daniel Kiper <[email protected]>
11450
11451 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11452
11453 term/ns8250: Support more MMIO access sizes
11454 It is common for PCI based UARTs to use larger than one byte access
11455 sizes. This adds support for this and uses the information present
11456 in SPCR accordingly.
11457
11458 Reviewed-by: Daniel Kiper <[email protected]>
11459
11460 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11461
11462 term/ns8250: Use ACPI SPCR table when available to configure serial
11463 "serial auto" is now equivalent to just "serial" and will use the
11464 SPCR to discover the port if present, otherwise defaults to "com0"
11465 as before.
11466
11467 This allows to support MMIO ports specified by ACPI which is needed
11468 on AWS EC2 "metal" instances, and will enable GRUB to pickup the
11469 port configuration specified by ACPI in other cases.
11470
11471 Reviewed-by: Daniel Kiper <[email protected]>
11472
11473 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11474
11475 term/ns8250: Add configuration parameter when adding ports
11476 This will allow ports to be added with a pre-set configuration.
11477
11478 Reviewed-by: Daniel Kiper <[email protected]>
11479
11480 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11481
11482 term/ns8250: Move base clock definition to a header
11483 And while at it, unify it as clock frequency in Hz, to match the value in
11484 grub_serial_config struct and do the division by 16 in one common place.
11485
11486 This will simplify adding SPCR support.
11487
11488 Reviewed-by: Daniel Kiper <[email protected]>
11489
11490 2023-01-19 Benjamin Herrenschmidt <[email protected]>
11491
11492 term/ns8250: Add base support for MMIO UARTs
11493 This adds the ability for the driver to access UARTs via MMIO instead
11494 of PIO selectively at runtime, and exposes a new function to add an
11495 MMIO port.
11496
11497 In an ideal world, MMIO accessors would be generic and have architecture
11498 specific memory barriers. However, existing drivers don't have them and
11499 most of those "bare metal" drivers tend to be for x86 which doesn't need
11500 them. If necessary, those can be added later.
11501
11502 Reviewed-by: Daniel Kiper <[email protected]>
11503
11504 2023-01-18 Benjamin Herrenschmidt <[email protected]>
11505
11506 acpi: Add SPCR and generic address definitions
11507 This adds the definition of the two ACPI tables according to the spec.
11508
11509 Reviewed-by: Daniel Kiper <[email protected]>
11510
11511 2023-01-18 Benjamin Herrenschmidt <[email protected]>
11512
11513 kern/acpi: Export a generic grub_acpi_find_table()
11514 And convert grub_acpi_find_fadt() to use it.
11515
11516 Reviewed-by: Daniel Kiper <[email protected]>
11517
11518 2023-01-10 Maxim Fomin <[email protected]>
11519
11520 kern/fs: Fix possible integer overflow in i386-pc mode with large partitions
11521 The i386-pc mode supports MBR partition scheme where maximum partition
11522 size is 2 TiB. In case of large partitions left shift expression with
11523 unsigned long int "length" object may cause integer overflow making
11524 calculated partition size less than true value. This issue is fixed by
11525 increasing the size of "length" integer type.
11526
11527 Reviewed-by: Daniel Kiper <[email protected]>
11528
11529 2023-01-10 Glenn Washburn <[email protected]>
11530
11531 commands/cmp: Only return success when both files have the same contents
11532 This allows the cmp command to be used in GRUB scripts to conditionally
11533 run commands based on whether two files are the same.
11534
11535 The command is now quiet by default and the -v switch can be given to enable
11536 verbose mode, the previous behavior.
11537
11538 Update documentation accordingly.
11539
11540 Suggested-by: Li Gen <[email protected]>
11541 Reviewed-by: Daniel Kiper <[email protected]>
11542
11543 2023-01-10 Glenn Washburn <[email protected]>
11544
11545 docs: Remove text about cryptodisk UUIDs no being able to use dashes
11546 This was fixed here: 3cf2e848bc (disk/cryptodisk: Allows UUIDs to be compared
11547 in a dash-insensitive manner).
11548
11549 Reviewed-by: Daniel Kiper <[email protected]>
11550
11551 2023-01-10 Glenn Washburn <[email protected]>
11552
11553 tests/util/grub-shell: Add GRUB output logfile with grub-shell --debug
11554 This allows seeing full QEMU output of grub-shell, which can be invaluable
11555 when debugging failing tests.
11556
11557 Reviewed-by: Daniel Kiper <[email protected]>
11558
11559 2023-01-10 Marek Marczykowski-Górecki <[email protected]>
11560
11561 templates/linux_xen: Fix detecting XSM policy
11562 The xenpolicy variable was left set from previous function call. This
11563 resulted in all-but-first menu entries including XSM policy, even if it
11564 did not exist.
11565
11566 Fix this by initializing the xenpolicy variable.
11567
11568 Reviewed-by: Daniel Kiper <[email protected]>
11569
11570 2023-01-10 Zhang Boyang <[email protected]>
11571
11572 font: Reject fonts with negative max_char_width or max_char_height
11573 If max_char_width or max_char_height are negative wrong values can be propagated
11574 by grub_font_get_max_char_width() or grub_font_get_max_char_height(). Prevent
11575 this from happening.
11576
11577 Reviewed-by: Daniel Kiper <[email protected]>
11578
11579 2023-01-10 Zhang Boyang <[email protected]>
11580
11581 font: Assign null_font to unknown_glyph
11582 Like glyphs in ascii_font_glyph[], assign null_font to
11583 unknown_glyph->font in order to prevent grub_font_get_*() from
11584 dereferencing NULL pointer.
11585
11586 Reviewed-by: Daniel Kiper <[email protected]>
11587
11588 2023-01-10 Zhang Boyang <[email protected]>
11589
11590 font: Check return value of grub_malloc() in ascii_glyph_lookup()
11591 There is a problem in ascii_glyph_lookup(). It doesn't check the return
11592 value of grub_malloc(). If memory can't be allocated, then NULL pointer
11593 will be written to.
11594
11595 This patch fixes the problem by fallbacking to unknown_glyph when
11596 grub_malloc() returns NULL.
11597
11598 Reviewed-by: Daniel Kiper <[email protected]>
11599
11600 2023-01-10 Maxim Fomin <[email protected]>
11601
11602 disk/plainmount: Support plain encryption mode
11603 This patch adds support for plain encryption mode, plain dm-crypt, via
11604 new module/command named "plainmount".
11605
11606 Reviewed-by: Daniel Kiper <[email protected]>
11607 Reviewed-by: Glenn Washburn <[email protected]>
11608
11609 2023-01-10 Pete Batard <[email protected]>
11610
11611 util/grub-mkrescue: Search by file UUID rather than partition UUID for EFI boot
11612 The final piece needed to add UEFI file system transposition support is to
11613 ensure the boot media can be located regardless of how the boot partition
11614 was instantiated. Especially, we do not want to be reliant on brittle
11615 partition UUIDs, as these only work if a boot media is duplicated at the
11616 block level and not at the file system level.
11617
11618 To accomplish this for EFI boot, we now create a UUID file in a .disk/
11619 directory, that can then be searched for.
11620
11621 Note: The switch from make_image_fwdisk_abs() to make_image_abs() is
11622 needed in order to use the search functionality.
11623
11624 Reviewed-by: Daniel Kiper <[email protected]>
11625
11626 2023-01-10 Pete Batard <[email protected]>
11627
11628 util/grub-mkrescue: Preserve a copy of the EFI bootloaders on the ISO 9660 file system
11629 To enable file system transposition support for UEFI, we also must ensure that
11630 there exists a copy of the EFI bootloaders, that are currently embedded in the
11631 efi.img for xorriso, at their expected UEFI location on the ISO 9660 file system.
11632
11633 This is accomplished by removing the use of a temporary directory to create the
11634 efi/ content, to instead place it at the root of the ISO 9660 content.
11635
11636 Reviewed-by: Daniel Kiper <[email protected]>
11637
11638 2023-01-10 Pete Batard <[email protected]>
11639
11640 util/grub-mkrescue: Add support for FAT and NTFS on EFI boot
11641 In order to add file system transposition support for UEFI, i.e. the ability
11642 to copy the content of an grub-mkrescue ISO 9660 image onto user-formatted
11643 media, and have that boot on UEFI systems, the first thing we need to do is
11644 add support for the file systems that are natively handled by UEFI. This
11645 mandatorily includes FAT, but we also include NTFS as the latter is also
11646 commonly supported on modern x64 platforms.
11647
11648 Reviewed-by: Daniel Kiper <[email protected]>
11649
11650 2022-12-07 t.feng <[email protected]>
11651
11652 util/bash-completion: Disable SC2120 shellcheck warning
11653 SC2120 (warning): function references arguments, but none are ever passed.
11654
11655 In grub-completion.bash.in line 63:
11656 __grub_get_options_from_help () {
11657 ^-- SC2120 (warning)
11658 local prog
11659
11660 if [ $# -ge 1 ]; then
11661 prog="$1"
11662
11663 The arg of __grub_get_options_from_help() is optional. So, the current
11664 code meets the exception and does not need to be modified. Ignoring the
11665 warning then.
11666
11667 More: https://github.com/koalaman/shellcheck/wiki/SC2120
11668
11669 Reviewed-by: Daniel Kiper <[email protected]>
11670
11671 2022-12-07 t.feng <[email protected]>
11672
11673 util/bash-completion: Fix SC2155 shellcheck warning
11674 SC2155 (warning): Declare and assign separately to avoid masking return values.
11675
11676 The exit status of the command is overridden by the exit status of the
11677 creation of the local variable.
11678
11679 In grub-completion.bash.in line 115:
11680 local config_file=$(__grub_dir)/grub.cfg
11681 ^---------^ SC2155 (warning)
11682
11683 In grub-completion.bash.in line 126:
11684 local grub_dir=$(__grub_dir)
11685 ^------^ SC2155 (warning)
11686
11687 More: https://github.com/koalaman/shellcheck/wiki/SC2155
11688
11689 Reviewed-by: Daniel Kiper <[email protected]>
11690
11691 2022-12-07 t.feng <[email protected]>
11692
11693 util/bash-completion: Fix SC2207 shellcheck warning
11694 SC2207 (warning): Prefer mapfile or read -a to split
11695 command output (or quote to avoid splitting).
11696
11697 In grub-completion.bash.in line 56:
11698 COMPREPLY=($(compgen -P "${2-}" -W "${1-}" -S "${4-}" -- "$cur"))
11699 ^-- SC2207 (warning)
11700
11701 In grub-completion.bash.in line 119:
11702 COMPREPLY=( $(compgen \
11703 ^-- SC2207 (warning)
11704
11705 In grub-completion.bash.in line 128:
11706 COMPREPLY=( $( compgen -f -X '!*/*.mod' -- "${grub_dir}/$cur" | {
11707 ^-- SC2207 (warning)
11708
11709 COMPREPLY=($(command)) are doing unquoted command expansion in an array.
11710 This will invoke the shell's sloppy word splitting and glob expansion.
11711
11712 If we want to split the output into lines or words, use read -r and
11713 loops will be better. This prevents the shell from doing unwanted
11714 splitting and glob expansion, and therefore avoiding problems with
11715 output containing spaces or special characters.
11716
11717 More: https://github.com/koalaman/shellcheck/wiki/SC2207
11718
11719 Reviewed-by: Daniel Kiper <[email protected]>
11720
11721 2022-12-07 t.feng <[email protected]>
11722
11723 util/bash-completion: Fix SC2070 shellcheck error
11724 SC2070 (error): -n doesn't work with unquoted arguments.
11725 Quote or use [[ ]].
11726 In grub-completion.bash.in line 130:
11727 [ -n $tmp ] && {
11728 ^--^ SC2070 (error)
11729
11730 More: https://github.com/koalaman/shellcheck/wiki/SC2070
11731
11732 Reviewed-by: Daniel Kiper <[email protected]>
11733
11734 2022-12-07 Steve McIntyre <[email protected]>
11735
11736 kern/file: Fix error handling in grub_file_open()
11737 grub_file_open() calls grub_file_get_device_name(), but doesn't check
11738 the return. Instead, it checks if grub_errno is set.
11739
11740 However, nothing initialises grub_errno here when grub_file_open()
11741 starts. This means that trying to open one file that doesn't exist and
11742 then trying to open another file that does will (incorrectly) also
11743 fail to open that second file.
11744
11745 Let's fix that.
11746
11747 Reviewed-by: Daniel Kiper <[email protected]>
11748
11749 2022-12-07 Jeremy Szu <[email protected]>
11750
11751 loader/i386/linux: Fix initrd maximum address overflow
11752 The current i386 initrd is limited under 1 GiB memory and it works with
11753 most compressed initrds (also initrd_addr_max case reported by kernel).
11754
11755 addr = (addr_max - aligned_size) & ~0xFFF;
11756
11757 Above line is used to calculate the reasonable address to store the initrd.
11758
11759 However, if initrd size is greater than 1 GiB or initrd_addr_max, then it
11760 will get overflow, especially on x86_64 arch.
11761
11762 Therefore, add a check point to prevent it overflows as well as having
11763 a debug log for complex story of initrd addresses.
11764
11765 Reviewed-by: Daniel Kiper <[email protected]>
11766
11767 2022-12-07 Dimitri John Ledkov <[email protected]>
11768
11769 templates: Enable fwsetup on EFI platforms only
11770 Only perform call to fwsetup if one is on EFI platform. On all other
11771 platforms fwsetup command does not exists, and thus returns 0 and
11772 a useless uefi-firmware menu entry gets generated.
11773
11774 Reviewed-by: Daniel Kiper <[email protected]>
11775
11776 2022-12-07 t.feng <[email protected]>
11777
11778 fs/xfs: Fix memory leaks in XFS module
11779 Reviewed-by: Daniel Kiper <[email protected]>
11780
11781 2022-12-07 t.feng <[email protected]>
11782
11783 fs/squash4: Fix memory leaks in grub_squash_iterate_dir()
11784 Fixes: 20dd511c8 (Handle "." and ".." on squashfs)
11785
11786 Reviewed-by: Daniel Kiper <[email protected]>
11787
11788 2022-12-07 t.feng <[email protected]>
11789
11790 fs/iso9660: Fix memory leaks in grub_iso9660_susp_iterate()
11791 Fixes: 99373ce47 (* grub-core/fs/iso9660.c: Remove nested functions)
11792
11793 Reviewed-by: Thomas Schmitt <[email protected]>
11794 Reviewed-by: Daniel Kiper <[email protected]>
11795
11796 2022-12-07 t.feng <[email protected]>
11797
11798 fs/hfsplus: Fix memory leak in grub_hfsplus_btree_search()
11799 Fixes: 58ea11d5b (fs/hfsplus: Don't fetch a key beyond the end of the node)
11800
11801 Reviewed-by: Daniel Kiper <[email protected]>
11802
11803 2022-12-07 t.feng <[email protected]>
11804
11805 fs/bfs: Fix memory leak in read_bfs_file()
11806 The l1_entries and l2_entries were not freed at the end of file read.
11807
11808 Fixes: 5825b3794 (BFS implementation based on the specification)
11809
11810 Reviewed-by: Daniel Kiper <[email protected]>
11811
11812 2022-12-07 t.feng <[email protected]>
11813
11814 fs/ntfs: Fix memory leaks in grub_ntfs_read_symlink()
11815 Fixes: 5773fb641 (Support NTFS reparse points)
11816
11817 Reviewed-by: Daniel Kiper <[email protected]>
11818
11819 2022-12-07 t.feng <[email protected]>
11820
11821 fs/minix: Fix memory leaks in grub_minix_lookup_symlink()
11822 Fixes: a07e6ad01 (* grub-core/fs/minix.c: Remove variable length arrays)
11823
11824 Reviewed-by: Daniel Kiper <[email protected]>
11825
11826 2022-12-07 t.feng <[email protected]>
11827
11828 fs/btrfs: Fix memory leak in find_path()
11829 Fixes: 82591fa6e (Make / in btrfs refer to real root)
11830
11831 Reviewed-by: Daniel Kiper <[email protected]>
11832
11833 2022-12-07 t.feng <[email protected]>
11834
11835 fs/affs: Fix memory leaks in grub_affs_create_node()
11836 The hashtable is not freed if GRUB_AFFS_FILETYPE_HARDLINK and
11837 grub_disk_read() failed. If grub_affs_create_node() returns non-zero
11838 the hashtable should be freed too.
11839
11840 By the way, the hashtable argument is unused in grub_affs_create_node().
11841 So, we can remove the argument and free it in grub_affs_iterate_dir().
11842 It allocates the memory and it should be responsible for releasing it.
11843
11844 This is why commit ebf32bc4e9 (fs/affs: Fix resource leaks) missed
11845 this memory leak.
11846
11847 Fixes: ebf32bc4e9 (fs/affs: Fix resource leaks)
11848
11849 Reviewed-by: Daniel Kiper <[email protected]>
11850
11851 2022-12-07 Ryan Cohen <[email protected]>
11852
11853 normal/cmdline: Fix two related integer underflows
11854 An unchecked decrement operation in cl_print() would cause a few
11855 integers to underflow. Where an output terminal's state is stored in
11856 cl_term, the values cl_term->ystart and cl_term->pos.y both underflow.
11857
11858 This can be replicated with the following steps:
11859
11860 1. Get to the GRUB command line
11861 2. Hold down the "d" key (or any key that enters a visible character)
11862 until it fills the entire row
11863 3. Press "HOME" and then press "CTRL-k". This will clear every
11864 character entered in step 2
11865 4. Continuously press "CTRL-y" until the terminal scrolls the original
11866 prompt ("grub> ") passed the terminal's top row. Now, no prompt
11867 should be visible. This step causes cl_term->ystart to underflow
11868 5. Press "HOME" and then "d" (or any visible character). This can have
11869 different visual effects for different systems, but it will always
11870 cause cl_term->pos.y to underflow
11871
11872 On BIOS systems, these underflows cause the output terminal to
11873 completely stop displaying anything. Characters can still be
11874 entered and commands can be run, but nothing will display on the
11875 terminal. From here, you can only get the display working by running
11876 a command to switch the current output terminal to a different type:
11877
11878 terminal_output <OTHER_TERMINAL>
11879
11880 On UEFI systems, these replication steps do not break the output
11881 terminal. Until you press "ENTER", the cursor stops responding to input,
11882 but you can press "ENTER" after step 5 and the command line will
11883 work properly again. This patch is mostly important for BIOS systems
11884 where the output terminal is rendered unusable after the underflows
11885 occur.
11886
11887 This patch adds two checks, one for each variable. It ensures that
11888 cl_term->ystart does not decrement passed 0. It also ensures that
11889 cl_term->pos.y does not get set passed the terminal's bottom row.
11890
11891 When the previously listed replication steps are followed with this
11892 patch, the terminal's cursor will be set to the top row and the command
11893 line is still usable, even on BIOS systems.
11894
11895 Reviewed-by: Daniel Kiper <[email protected]>
11896
11897 2022-12-07 Ryan Cohen <[email protected]>
11898
11899 term/i386/pc/vga_text: Prevent out-of-bounds writes to VGA text buffer
11900 Coordinates passed to screen_write_char() did not have any checks to
11901 ensure they are not out-of-bounds. This adds an if statement to prevent
11902 out-of-bounds writes to the VGA text buffer.
11903
11904 Reviewed-by: Daniel Kiper <[email protected]>
11905
11906 2022-12-07 Gary Lin <[email protected]>
11907
11908 loader/linux: Ensure the newc pathname is NULL-terminated
11909 Per "man 5 cpio", the namesize in the cpio header includes the trailing
11910 NUL byte of the pathname and the pathname is followed by NUL bytes, but
11911 the current implementation ignores the trailing NUL byte when making
11912 the newc header. Although make_header() tries to pad the pathname string,
11913 the padding won't happen when strlen(name) + sizeof(struct newc_head)
11914 is a multiple of 4, and the non-NULL-terminated pathname may lead to
11915 unexpected results.
11916
11917 Assume that a file is created with 'echo -n aaaa > /boot/test12' and
11918 loaded by grub2:
11919
11920 linux /boot/vmlinuz
11921 initrd newc:test12:/boot/test12 /boot/initrd
11922
11923 The initrd command eventually invoked grub_initrd_load() and sent
11924 't''e''s''t''1''2' to make_header() to generate the header:
11925
11926 00000070 30 37 30 37 30 31 33 30 31 43 41 30 44 45 30 30 |070701301CA0DE00|
11927 00000080 30 30 38 31 41 34 30 30 30 30 30 33 45 38 30 30 |0081A4000003E800|
11928 00000090 30 30 30 30 36 34 30 30 30 30 30 30 30 31 36 33 |0000640000000163|
11929 000000a0 37 36 45 34 35 32 30 30 30 30 30 30 30 34 30 30 |76E4520000000400|
11930 000000b0 30 30 30 30 30 38 30 30 30 30 30 30 31 33 30 30 |0000080000001300|
11931 000000c0 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 |0000000000000000|
11932 000000d0 30 30 30 30 30 36 30 30 30 30 30 30 30 30 74 65 |00000600000000te|
11933 ^namesize
11934 000000e0 73 74 31 32 61 61 61 61 30 37 30 37 30 31 30 30 |st12aaaa07070100|
11935 ^^ end of the pathname
11936
11937 Since strlen("test12") + sizeof(struct newc_head) is 116 = 29 * 4,
11938 make_header() didn't pad the pathname, and the file content followed
11939 "test12" immediately. This violates the cpio format and may trigger such
11940 error during linux boot:
11941
11942 Initramfs unpacking failed: ZSTD-compressed data is trunc
11943
11944 To avoid the potential problems, this commit counts the trailing NUL byte
11945 in when calling make_header() and adjusts the initrd size accordingly.
11946
11947 Now the header becomes
11948
11949 00000070 30 37 30 37 30 31 33 30 31 43 41 30 44 45 30 30 |070701301CA0DE00|
11950 00000080 30 30 38 31 41 34 30 30 30 30 30 33 45 38 30 30 |0081A4000003E800|
11951 00000090 30 30 30 30 36 34 30 30 30 30 30 30 30 31 36 33 |0000640000000163|
11952 000000a0 37 36 45 34 35 32 30 30 30 30 30 30 30 34 30 30 |76E4520000000400|
11953 000000b0 30 30 30 30 30 38 30 30 30 30 30 30 31 33 30 30 |0000080000001300|
11954 000000c0 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 30 |0000000000000000|
11955 000000d0 30 30 30 30 30 37 30 30 30 30 30 30 30 30 74 65 |00000700000000te|
11956 ^namesize
11957 000000e0 73 74 31 32 00 00 00 00 61 61 61 61 30 37 30 37 |st12....aaaa0707|
11958 ^^ end of the pathname
11959
11960 Besides the trailing NUL byte, make_header() pads 3 more NUL bytes, and
11961 the user can safely read the pathname without a further check.
11962
11963 To conform to the cpio format, the headers for "TRAILER!!!" are also
11964 adjusted to include the trailing NUL byte, not ignore it.
11965
11966 Reviewed-by: Daniel Kiper <[email protected]>
11967
11968 2022-12-07 Jagannathan Raman <[email protected]>
11969
11970 fs/udf: Validate length of AED in grub_udf_read_block()
11971 Validate the length of Allocation Extent Descriptor in grub_udf_read_block(),
11972 based on the details in UDF spec. v2.01 section 2.3.11.
11973
11974 Fixes: CID 314037
11975
11976 Reviewed-by: Daniel Kiper <[email protected]>
11977
11978 2022-12-07 Ismael Luceno <[email protected]>
11979
11980 util/grub-install: Ensure a functional /dev/nvram
11981 This enables an early failure; for i386-ieee1275 and powerpc-ieee1275 on
11982 Linux, without /dev/nvram the system may be left in an unbootable state.
11983
11984 Reviewed-by: Daniel Kiper <[email protected]>
11985
11986 2022-12-07 Ismael Luceno <[email protected]>
11987
11988 templates: Set defaults using var substitution
11989 Reviewed-by: Daniel Kiper <[email protected]>
11990
11991 2022-12-07 Glenn Washburn <[email protected]>
11992
11993 tests: Put all generated files into working dir and use better file names
11994 When running tests there are many invocations of grub-shell, and because
11995 the output files are all random names in the same tmp directory, it
11996 becomes more work to figure out which files went with which grub-shell
11997 invocations. So all generated files from one invocation of grub-shell
11998 are put into a randomly named directory, so as not to collide with other
11999 grub-shell invocations. And now that the generated files can be put in
12000 a location where they will not get stepped on, and they can be named
12001 sensible names.
12002
12003 Reviewed-by: Daniel Kiper <[email protected]>
12004
12005 2022-11-14 Zhang Boyang <[email protected]>
12006
12007 normal/charset: Fix an integer overflow in grub_unicode_aglomerate_comb()
12008 The out->ncomb is a bit-field of 8 bits. So, the max possible value is 255.
12009 However, code in grub_unicode_aglomerate_comb() doesn't check for an
12010 overflow when incrementing out->ncomb. If out->ncomb is already 255,
12011 after incrementing it will get 0 instead of 256, and cause illegal
12012 memory access in subsequent processing.
12013
12014 This patch introduces GRUB_UNICODE_NCOMB_MAX to represent the max
12015 acceptable value of ncomb. The code now checks for this limit and
12016 ignores additional combining characters when limit is reached.
12017
12018 Reported-by: Daniel Axtens <[email protected]>
12019 Reviewed-by: Daniel Kiper <[email protected]>
12020
12021 2022-11-14 Zhang Boyang <[email protected]>
12022
12023 font: Assign null_font to glyphs in ascii_font_glyph[]
12024 The calculations in blit_comb() need information from glyph's font, e.g.
12025 grub_font_get_xheight(main_glyph->font). However, main_glyph->font is
12026 NULL if main_glyph comes from ascii_font_glyph[]. Therefore
12027 grub_font_get_*() crashes because of NULL pointer.
12028
12029 There is already a solution, the null_font. So, assign it to those glyphs
12030 in ascii_font_glyph[].
12031
12032 Reported-by: Daniel Axtens <[email protected]>
12033 Reviewed-by: Daniel Kiper <[email protected]>
12034
12035 2022-11-14 Zhang Boyang <[email protected]>
12036
12037 font: Harden grub_font_blit_glyph() and grub_font_blit_glyph_mirror()
12038 As a mitigation and hardening measure add sanity checks to
12039 grub_font_blit_glyph() and grub_font_blit_glyph_mirror(). This patch
12040 makes these two functions do nothing if target blitting area isn't fully
12041 contained in target bitmap. Therefore, if complex calculations in caller
12042 overflows and malicious coordinates are given, we are still safe because
12043 any coordinates which result in out-of-bound-write are rejected. However,
12044 this patch only checks for invalid coordinates, and doesn't provide any
12045 protection against invalid source glyph or destination glyph, e.g.
12046 mismatch between glyph size and buffer size.
12047
12048 This hardening measure is designed to mitigate possible overflows in
12049 blit_comb(). If overflow occurs, it may return invalid bounding box
12050 during dry run and call grub_font_blit_glyph() with malicious
12051 coordinates during actual blitting. However, we are still safe because
12052 the scratch glyph itself is valid, although its size makes no sense, and
12053 any invalid coordinates are rejected.
12054
12055 It would be better to call grub_fatal() if illegal parameter is detected.
12056 However, doing this may end up in a dangerous recursion because grub_fatal()
12057 would print messages to the screen and we are in the progress of drawing
12058 characters on the screen.
12059
12060 Reported-by: Daniel Axtens <[email protected]>
12061 Reviewed-by: Daniel Kiper <[email protected]>
12062
12063 2022-11-14 Zhang Boyang <[email protected]>
12064
12065 font: Fix an integer underflow in blit_comb()
12066 The expression (ctx.bounds.height - combining_glyphs[i]->height) / 2 may
12067 evaluate to a very big invalid value even if both ctx.bounds.height and
12068 combining_glyphs[i]->height are small integers. For example, if
12069 ctx.bounds.height is 10 and combining_glyphs[i]->height is 12, this
12070 expression evaluates to 2147483647 (expected -1). This is because
12071 coordinates are allowed to be negative but ctx.bounds.height is an
12072 unsigned int. So, the subtraction operates on unsigned ints and
12073 underflows to a very big value. The division makes things even worse.
12074 The quotient is still an invalid value even if converted back to int.
12075
12076 This patch fixes the problem by casting ctx.bounds.height to int. As
12077 a result the subtraction will operate on int and grub_uint16_t which
12078 will be promoted to an int. So, the underflow will no longer happen. Other
12079 uses of ctx.bounds.height (and ctx.bounds.width) are also casted to int,
12080 to ensure coordinates are always calculated on signed integers.
12081
12082 Fixes: CVE-2022-3775
12083
12084 Reported-by: Daniel Axtens <[email protected]>
12085 Reviewed-by: Daniel Kiper <[email protected]>
12086
12087 2022-11-14 Zhang Boyang <[email protected]>
12088
12089 fbutil: Fix integer overflow
12090 Expressions like u64 = u32 * u32 are unsafe because their products are
12091 truncated to u32 even if left hand side is u64. This patch fixes all
12092 problems like that one in fbutil.
12093
12094 To get right result not only left hand side have to be u64 but it's also
12095 necessary to cast at least one of the operands of all leaf operators of
12096 right hand side to u64, e.g. u64 = u32 * u32 + u32 * u32 should be
12097 u64 = (u64)u32 * u32 + (u64)u32 * u32.
12098
12099 For 1-bit bitmaps grub_uint64_t have to be used. It's safe because any
12100 combination of values in (grub_uint64_t)u32 * u32 + u32 expression will
12101 not overflow grub_uint64_t.
12102
12103 Other expressions like ptr + u32 * u32 + u32 * u32 are also vulnerable.
12104 They should be ptr + (grub_addr_t)u32 * u32 + (grub_addr_t)u32 * u32.
12105
12106 This patch also adds a comment to grub_video_fb_get_video_ptr() which
12107 says it's arguments must be valid and no sanity check is performed
12108 (like its siblings in grub-core/video/fb/fbutil.c).
12109
12110 Reviewed-by: Daniel Kiper <[email protected]>
12111
12112 2022-11-14 Zhang Boyang <[email protected]>
12113
12114 kern/efi/sb: Enforce verification of font files
12115 As a mitigation and hardening measure enforce verification of font
12116 files. Then only trusted font files can be load. This will reduce the
12117 attack surface at cost of losing the ability of end-users to customize
12118 fonts if e.g. UEFI Secure Boot is enabled. Vendors can always customize
12119 fonts because they have ability to pack fonts into their GRUB bundles.
12120
12121 This goal is achieved by:
12122
12123 * Removing GRUB_FILE_TYPE_FONT from shim lock verifier's
12124 skip-verification list.
12125
12126 * Adding GRUB_FILE_TYPE_FONT to lockdown verifier's defer-auth list,
12127 so font files must be verified by a verifier before they can be loaded.
12128
12129 Suggested-by: Daniel Kiper <[email protected]>
12130 Reviewed-by: Daniel Kiper <[email protected]>
12131
12132 2022-11-14 Zhang Boyang <[email protected]>
12133
12134 font: Fix integer underflow in binary search of char index
12135 If search target is less than all entries in font->index then "hi"
12136 variable is set to -1, which translates to SIZE_MAX and leads to errors.
12137
12138 This patch fixes the problem by replacing the entire binary search code
12139 with the libstdc++'s std::lower_bound() implementation.
12140
12141 Reviewed-by: Daniel Kiper <[email protected]>
12142
12143 2022-11-14 Zhang Boyang <[email protected]>
12144
12145 font: Fix integer overflow in BMP index
12146 The BMP index (font->bmp_idx) is designed as a reverse lookup table of
12147 char entries (font->char_index), in order to speed up lookups for BMP
12148 chars (i.e. code < 0x10000). The values in BMP index are the subscripts
12149 of the corresponding char entries, stored in grub_uint16_t, while 0xffff
12150 means not found.
12151
12152 This patch fixes the problem of large subscript truncated to grub_uint16_t,
12153 leading BMP index to return wrong char entry or report false miss. The
12154 code now checks for bounds and uses BMP index as a hint, and fallbacks
12155 to binary-search if necessary.
12156
12157 On the occasion add a comment about BMP index is initialized to 0xffff.
12158
12159 Reviewed-by: Daniel Kiper <[email protected]>
12160
12161 2022-11-14 Zhang Boyang <[email protected]>
12162
12163 font: Fix integer overflow in ensure_comb_space()
12164 In fact it can't overflow at all because glyph_id->ncomb is only 8-bit
12165 wide. But let's keep safe if somebody changes the width of glyph_id->ncomb
12166 in the future. This patch also fixes the inconsistency between
12167 render_max_comb_glyphs and render_combining_glyphs when grub_malloc()
12168 returns NULL.
12169
12170 Reviewed-by: Daniel Kiper <[email protected]>
12171
12172 2022-11-14 Zhang Boyang <[email protected]>
12173
12174 font: Remove grub_font_dup_glyph()
12175 Remove grub_font_dup_glyph() since nobody is using it since 2013, and
12176 I'm too lazy to fix the integer overflow problem in it.
12177
12178 Reviewed-by: Daniel Kiper <[email protected]>
12179
12180 2022-11-14 Zhang Boyang <[email protected]>
12181
12182 font: Fix several integer overflows in grub_font_construct_glyph()
12183 This patch fixes several integer overflows in grub_font_construct_glyph().
12184 Glyphs of invalid size, zero or leading to an overflow, are rejected.
12185 The inconsistency between "glyph" and "max_glyph_size" when grub_malloc()
12186 returns NULL is fixed too.
12187
12188 Fixes: CVE-2022-2601
12189
12190 Reported-by: Zhang Boyang <[email protected]>
12191 Reviewed-by: Daniel Kiper <[email protected]>
12192
12193 2022-11-14 Zhang Boyang <[email protected]>
12194
12195 font: Fix size overflow in grub_font_get_glyph_internal()
12196 The length of memory allocation and file read may overflow. This patch
12197 fixes the problem by using safemath macros.
12198
12199 There is a lot of code repetition like "(x * y + 7) / 8". It is unsafe
12200 if overflow happens. This patch introduces grub_video_bitmap_calc_1bpp_bufsz().
12201 It is safe replacement for such code. It has safemath-like prototype.
12202
12203 This patch also introduces grub_cast(value, pointer), it casts value to
12204 typeof(*pointer) then store the value to *pointer. It returns true when
12205 overflow occurs or false if there is no overflow. The semantics of arguments
12206 and return value are designed to be consistent with other safemath macros.
12207
12208 Reviewed-by: Daniel Kiper <[email protected]>
12209
12210 2022-11-14 Zhang Boyang <[email protected]>
12211
12212 font: Reject glyphs exceeds font->max_glyph_width or font->max_glyph_height
12213 Check glyph's width and height against limits specified in font's
12214 metadata. Reject the glyph (and font) if such limits are exceeded.
12215
12216 Reviewed-by: Daniel Kiper <[email protected]>
12217
12218 2022-11-14 t.feng <[email protected]>
12219
12220 loader/multiboot_elfxx: Fix memory leak
12221 The commit eb33e61b3 (multiboot: fix memory leak) did not fix all
12222 issues. Fix all of them right now.
12223
12224 Fixes: eb33e61b3 (multiboot: fix memory leak)
12225
12226 Reviewed-by: Daniel Kiper <[email protected]>
12227
12228 2022-11-14 Damian Szuberski <[email protected]>
12229
12230 docs: Correct GRUB_DISABLE_LINUX_PARTUUID documentation
12231 Reviewed-by: Daniel Kiper <[email protected]>
12232
12233 2022-11-14 Arsen Arsenović <[email protected]>
12234
12235 osdep/unix/getroot: Pass -P to zpool status
12236 zpool status by default prints basenames of VDEVs, which means that GRUB
12237 would have to go around guessing to see whether a VDEV exists. Instead,
12238 it'd be more robust to simply tell zpool to give us full paths to VDEVs
12239 via -P.
12240
12241 Reviewed-by: Daniel Kiper <[email protected]>
12242
12243 2022-11-14 Robbie Harwood <[email protected]>
12244
12245 normal/help: Add paging instructions to normal and help prompts
12246 This is not an ideal solution, as interactive users must always run
12247 a command in order to get the behavior they want, but it avoids
12248 problematic interactions between prompting and sourcing files.
12249
12250 Reviewed-by: Daniel Kiper <[email protected]>
12251
12252 2022-11-14 Robbie Harwood <[email protected]>
12253
12254 commands/tpm: Don't propagate measurement failures to the verifiers layer
12255 Currently if an EFI firmware fails to do a TPM measurement for a file,
12256 the error will be propagated to the verifiers framework which will
12257 prevent it to be opened. This mean that buggy firmwares will lead to
12258 the system not booting because files won't be allowed to be loaded. But
12259 a failure to do a TPM measurement isn't expected to be a fatal error
12260 that causes the system to be unbootable.
12261
12262 To avoid this, don't return errors from .write and .verify_string
12263 callbacks and just print a debug message in the case of a TPM
12264 measurement failure. Add an environment variable, tpm_fail_fatal, to
12265 restore the previous behavior.
12266
12267 Also-authored-by: Javier Martinez Canillas <[email protected]>
12268 Reviewed-by: Daniel Kiper <[email protected]>
12269
12270 2022-11-14 Robbie Harwood <[email protected]>
12271
12272 kern/env: Add function for retrieving variables as booleans
12273 Reviewed-by: Daniel Kiper <[email protected]>
12274
12275 2022-11-14 Robbie Harwood <[email protected]>
12276
12277 types: Make bool generally available
12278 Add an include on stdbool.h, making the bool type generally available
12279 within the GRUB without needing to add a file-specific include every
12280 time it would be used.
12281
12282 Reviewed-by: Daniel Kiper <[email protected]>
12283
12284 2022-11-14 Raymund Will <[email protected]>
12285
12286 loader: Add support for grub-emu to kexec Linux menu entries
12287 The GRUB emulator is used as a debugging utility but it could also be
12288 used as a user-space bootloader if there is support to boot an operating
12289 system.
12290
12291 The Linux kernel is already able to (re)boot another kernel via the
12292 kexec boot mechanism. So the grub-emu tool could rely on this feature
12293 and have linux and initrd commands that are used to pass a kernel,
12294 initramfs image and command line parameters to kexec for booting
12295 a selected menu entry.
12296
12297 By default the systemctl kexec option is used so systemd can shutdown
12298 all of the running services before doing a reboot using kexec. But if
12299 this is not present, it can fall back to executing the kexec user-space
12300 tool directly. The ability to force a kexec-reboot when systemctl kexec
12301 fails must only be used in controlled environments to avoid possible
12302 filesystem corruption and data loss.
12303
12304 Reviewed-by: Daniel Kiper <[email protected]>
12305
12306 2022-11-14 Denton Liu <[email protected]>
12307
12308 templates: Introduce GRUB_TOP_LEVEL_* vars
12309 A user may wish to use an image that is not sorted as the "latest"
12310 version as the top-level entry. For example, in Arch Linux, if a user
12311 has the LTS and regular kernels installed, "/boot/vmlinuz-linux-lts"
12312 gets sorted as the "latest" compared to "/boot/vmlinuz-linux", meaning
12313 the LTS kernel becomes the top-level entry. However, a user may wish to
12314 use the regular kernel as the top-level default with the LTS only
12315 existing as a backup.
12316
12317 This need can be seen in Arch Linux's AUR with two user-submitted
12318 packages[0][1] providing an update hook which patches /etc/grub.d/10_linux
12319 to move the desired kernel to the top-level. This patch serves to solve
12320 this in a more generic way.
12321
12322 Introduce the GRUB_TOP_LEVEL, GRUB_TOP_LEVEL_XEN and GRUB_TOP_LEVEL_OS_PROBER
12323 variables to allow users to specify the top-level entry.
12324
12325 Create grub_move_to_front() as a helper function which moves entries to
12326 the front of a list. This function does the heavy lifting of moving
12327 the menu entry to the front in each script.
12328
12329 In 10_netbsd, since there isn't an explicit list variable, extract the
12330 items that are being iterated through into a list so that we can
12331 optionally apply grub_move_to_front() to the list before the loop.
12332
12333 [0]: https://aur.archlinux.org/packages/grub-linux-default-hook
12334 [1]: https://aur.archlinux.org/packages/grub-linux-rt-default-hook
12335
12336 Reviewed-by: Oskari Pirhonen <[email protected]>
12337 Reviewed-by: Daniel Kiper <[email protected]>
12338
12339 2022-10-27 Alec Brown <[email protected]>
12340
12341 video/readers: Add artificial limit to image dimensions
12342 In grub-core/video/readers/jpeg.c, the height and width of a JPEG image don't
12343 have an upper limit for how big the JPEG image can be. In Coverity, this is
12344 getting flagged as an untrusted loop bound. This issue can also seen in PNG and
12345 TGA format images as well but Coverity isn't flagging it. To prevent this, the
12346 constant IMAGE_HW_MAX_PX is being added to include/grub/bitmap.h, which has
12347 a value of 16384, to act as an artificial limit and restrict the height and
12348 width of images. This value was picked as it is double the current max
12349 resolution size, which is 8K.
12350
12351 Fixes: CID 292450
12352
12353 Reviewed-by: Darren Kenny <[email protected]>
12354 Reviewed-by: Daniel Kiper <[email protected]>
12355
12356 2022-10-27 Daniel Axtens <[email protected]>
12357
12358 disk/diskfilter: Don't make a RAID array with more than 1024 disks
12359 This is "belt and braces" with commit 12e20a6a695f (disk/diskfilter:
12360 Check calloc() result for NULL): we end up trying to use too much memory
12361 in situations like corrupted Linux software RAID setups purporting to
12362 use a huge number of disks. Simply refuse to permit such configurations.
12363
12364 1024 is a bit arbitrary, yes, and I feel a bit like I'm tempting fate
12365 here, but I think 1024 disks in an array (that GRUB has to read to boot!)
12366 should be enough for anyone.
12367
12368 Reviewed-by: Daniel Kiper <[email protected]>
12369
12370 2022-10-27 Ard Biesheuvel <[email protected]>
12371
12372 arm64/efi/linux: Ignore FDT unless we need to modify it
12373 Now that we implemented support for the LoadFile2 protocol for initrd
12374 loading, there is no longer a need to pass the initrd parameters via
12375 the device tree. This means that when the LoadFile2 protocol is being
12376 used, there is no reason to update the device tree in the first place,
12377 and so we can ignore it entirely.
12378
12379 The only remaining reason to deal with the devicetree is if we are
12380 using the "devicetree" command to load one from disk, so tweak the
12381 logic in grub_fdt_install() to take that into account.
12382
12383 Reviewed-by: Leif Lindholm <[email protected]>
12384 Reviewed-by: Daniel Kiper <[email protected]>
12385
12386 2022-10-27 Ard Biesheuvel <[email protected]>
12387
12388 arm64/efi/linux: Implement LoadFile2 initrd loading protocol for Linux
12389 Recent Linux kernels will invoke the LoadFile2 protocol installed on
12390 a well-known vendor media path to load the initrd if it is exposed by
12391 the firmware. Using this method is preferred for two reasons:
12392 - the Linux kernel is in charge of allocating the memory, and so it can
12393 implement any placement policy it wants (given that these tend to
12394 change between kernel versions),
12395 - it is no longer necessary to modify the device tree provided by the
12396 firmware.
12397
12398 So let's install this protocol when handling the "initrd" command if
12399 such a recent kernel was detected (based on the PE/COFF image version),
12400 and defer loading the initrd contents until the point where the kernel
12401 invokes the LoadFile2 protocol.
12402
12403 Reviewed-by: Heinrich Schuchardt <[email protected]>
12404 Tested-by: Ilias Apalodimas <[email protected]>
12405 Reviewed-by: Ilias Apalodimas <[email protected]>
12406 Reviewed-by: Daniel Kiper <[email protected]>
12407
12408 2022-10-27 Ard Biesheuvel <[email protected]>
12409
12410 efi/efinet: Don't close connections at fini_hw() time
12411 When GRUB runs on top of EFI firmware, it only has access to block and
12412 network device abstractions exposed by the firmware, and it is up to the
12413 firmware to quiesce the underlying hardware when exiting boot services
12414 and handing over to the OS.
12415
12416 This is especially important for network devices, to prevent incoming
12417 packets from being DMA'd straight into memory after the OS has taken
12418 over but before it has managed to reconfigure the network hardware.
12419
12420 GRUB handles this by means of the grub_net_fini_hw() preboot hook, which
12421 is executed before calling into the booted image. This means that all
12422 network devices disappear or become inoperable before the EFI stub
12423 executes on EFI targeted builds. This is problematic as it prevents the
12424 EFI stub from calling back into GRUB provided protocols such as
12425 LoadFile2 for the initrd, which we will provide in a subsequent patch.
12426
12427 So add a flag that indicates to the network core that EFI network
12428 devices should not be closed when grub_net_fini_hw() is called.
12429
12430 Reviewed-by: Heinrich Schuchardt <[email protected]>
12431 Reviewed-by: Daniel Kiper <[email protected]>
12432
12433 2022-10-27 Ard Biesheuvel <[email protected]>
12434
12435 loader/arm64/linux: Account for COFF headers appearing at unexpected offsets
12436 The way we load the Linux and PE/COFF image headers depends on a fixed
12437 placement of the COFF header at offset 0x40 into the file. This is
12438 a reasonable default, given that this is where Linux emits it today.
12439 However, in order to comply with the PE/COFF spec, which allows this
12440 header to appear anywhere in the file, let's ensure that we read the
12441 header from where it actually appears in the file if it is not located
12442 at offset 0x40.
12443
12444 Reviewed-by: Daniel Kiper <[email protected]>
12445
12446 2022-10-27 Ard Biesheuvel <[email protected]>
12447
12448 arm/linux: Unify ARM/arm64 vs Xen PE/COFF header handling
12449 Xen has its own version of the image header, to account for the
12450 additional PE/COFF header fields. Since we are adding references to
12451 those in the shared EFI loader code, update the common definitions
12452 and drop the Xen specific one which no longer has a purpose.
12453
12454 Since in both cases, the call to grub_arch_efi_linux_check_image() is
12455 preceded by a load of the image header, let's move the load into that
12456 function, and rename it to grub_arch_efi_linux_load_image_header().
12457
12458 Reviewed-by: Daniel Kiper <[email protected]>
12459
12460 2022-10-27 Ard Biesheuvel <[email protected]>
12461
12462 efi: Move MS-DOS stub out of generic PE header definition
12463 The PE/COFF spec permits the COFF signature and file header to appear
12464 anywhere in the file, and the actual offset is recorded in 4 byte
12465 little endian field at offset 0x3c of the image.
12466
12467 When GRUB is emitted as a PE/COFF binary, we reuse the 128 byte MS-DOS
12468 stub (even for non-x86 architectures), putting the COFF signature and
12469 file header at offset 0x80. However, other PE/COFF images may use
12470 different values, and non-x86 Linux kernels use an offset of 0x40
12471 instead.
12472
12473 So let's get rid of the grub_pe32_header struct from pe32.h, given that
12474 it does not represent anything defined by the PE/COFF spec. Instead,
12475 introduce a minimal struct grub_msdos_image_header type based on the
12476 PE/COFF spec's description of the image header, and use the offset
12477 recorded at file position 0x3c to discover the actual location of the PE
12478 signature and the COFF image header.
12479
12480 The remaining fields are moved into a struct grub_pe_image_header,
12481 which we will use later to access COFF header fields of arbitrary
12482 images (and which may therefore appear at different offsets)
12483
12484 Reviewed-by: Daniel Kiper <[email protected]>
12485
12486 2022-10-27 Jagannathan Raman <[email protected]>
12487
12488 kern/buffer: Handle NULL input pointer in grub_buffer_free()
12489 The grub_buffer_free() should handle NULL input pointer, similar to
12490 grub_free(). If the pointer is not referencing any memory location,
12491 grub_buffer_free() need not perform any function.
12492
12493 Fixes: CID 396931
12494
12495 Reviewed-by: Ross Philipson <[email protected]>
12496 Reviewed-by: Daniel Kiper <[email protected]>
12497
12498 2022-10-27 Jagannathan Raman <[email protected]>
12499
12500 fs/zfs/zfs: Update dangling dn_new pointer in dnode_get_path()
12501 The dnode_get_path() traverses dnode structures to locate the dnode leaf
12502 of a given path. When the leaf is a symlink to another path, it restarts
12503 the traversal either from root or from a different path. In such cases,
12504 dn_new must be re-initialized
12505
12506 Passes "make check".
12507
12508 Fixes: CID 86750
12509
12510 Reviewed-by: Ross Philipson <[email protected]>
12511 Reviewed-by: Daniel Kiper <[email protected]>
12512
12513 2022-10-27 Darren Kenny <[email protected]>
12514
12515 build: Update to reflect minimum clang version 8.0
12516 After doing some validation with clang from versions 3.8 and up, the
12517 builds prior to version 8.0.0 fail due to the use of safemath functions
12518 at link time.
12519
12520 Reviewed-by: Daniel Kiper <[email protected]>
12521
12522 2022-10-27 Darren Kenny <[email protected]>
12523
12524 configure: Fix building with clang
12525 Building the current code with clang and the latest gnulib fails due to
12526 the use of a variable-length-array (vla) warning, which turns in to an
12527 error due to the presence of the -Werror during the build.
12528
12529 The gnulib team stated that their code should not be built with -Werror.
12530
12531 At present, the only way to do this is for the complete code-base, by
12532 using the --disable-werror option to configure.
12533
12534 Rather than doing this, and failing to gain any benefit that it provides,
12535 instead, if building with clang, this patch makes it possible to specifically
12536 not error on vlas, while retaining the -Werror functionality otherwise.
12537
12538 Reviewed-by: Daniel Kiper <[email protected]>
12539
12540 2022-10-27 Darren Kenny <[email protected]>
12541
12542 gnulib: Provide abort() implementation for gnulib
12543 The recent gnulib updates require an implementation of abort(), but the
12544 current macro provided by changeset:
12545
12546 cd37d3d3916c gnulib: Drop no-abort.patch
12547
12548 to config.h.in does not work with the clang compiler since it doesn't
12549 provide a __builtin_trap() implementation, so this element of the
12550 changeset needs to be reverted, and replaced.
12551
12552 After some discussion with Vladimir 'phcoder' Serbinenko and Daniel Kiper
12553 it was suggested to bring back in the change from the changeset:
12554
12555 db7337a3d353 * grub-core/gnulib/regcomp.c (regerror): ...
12556
12557 Which implements abort() as an inline call to grub_abort(), but since
12558 that was made static by changeset:
12559
12560 a8f15bceeafe * grub-core/kern/misc.c (grub_abort): Make static
12561
12562 it is also necessary to revert the specific part that makes it a static
12563 function too.
12564
12565 Another implementation of abort() was found in grub-core/kern/compiler-rt.c
12566 which needs to also be removed to be consistent.
12567
12568 Reviewed-by: Daniel Kiper <[email protected]>
12569
12570 2022-10-27 Alec Brown <[email protected]>
12571
12572 disk/cryptodisk: Fix unintentional integer overflow
12573 In the function grub_cryptodisk_endecrypt(), a for loop is incrementing the
12574 variable i by (1U << log_sector_size). The variable i is of type grub_size_t
12575 which is a 64-bit unsigned integer on x86_64 architecture. On the other hand, 1U
12576 is a 32-bit unsigned integer. By performing a left shift on a 32-bit value and
12577 assigning it to a 64-bit variable, the 64-bit variable may have incorrect values
12578 in the high 32-bits if the shift has an overflow. To avoid this, we replace 1U
12579 with (grub_size_t)1.
12580
12581 Fixes: CID 307788
12582
12583 Reviewed-by: Darren Kenny <[email protected]>
12584 Reviewed-by: Patrick Steinhardt <[email protected]>
12585 Reviewed-by: Daniel Kiper <[email protected]>
12586
12587 2022-10-27 Zhang Boyang <[email protected]>
12588
12589 mm: Try invalidate disk caches last when out of memory
12590 Every heap grow will cause all disk caches invalidated which decreases
12591 performance severely. This patch moves disk cache invalidation code to
12592 the last of memory squeezing measures. So, disk caches are released only
12593 when there are no other ways to get free memory.
12594
12595 Reviewed-by: Daniel Kiper <[email protected]>
12596 Reviewed-by: Patrick Steinhardt <[email protected]>
12597
12598 2022-10-27 Qiumiao Zhang <[email protected]>
12599
12600 util/grub-mkfont: Use valid conversion specifiers in printf() and fprintf()
12601 For printf()/fprintf() functions, unsigned integers should use %u as the
12602 valid conversion specifier instead of %d.
12603
12604 Reviewed-by: Daniel Kiper <[email protected]>
12605
12606 2022-10-27 Chris Coulson <[email protected]>
12607
12608 efi: Compile kernel.img with -fshort-wchar on all EFI targets
12609 The stack check logs a console message on failure, and the EFI API expects
12610 a NULL terminated UCS-2 string. In order to define a UCS-2 string literal,
12611 kernel.img on amd64 and i386 EFI targets is built with -fshort-wchar.
12612
12613 Also compile kernel.img on other EFI targets with -fshort-wchar.
12614
12615 Fixes: 37ddd94 (kern/efi/init: Log a console error during a stack check failure)
12616
12617 Reported-by: Glenn Washburn <[email protected]>
12618 Reviewed-by: Daniel Kiper <[email protected]>
12619
12620 2022-10-11 Benjamin Herrenschmidt <[email protected]>
12621
12622 normal/menu: Add Ctrl-L to refresh the menu
12623 This is useful on cloud instances with remote serial ports as it can be
12624 difficult to connect "fast enough" to get the initial menu display
12625
12626 Reviewed-by: Daniel Kiper <[email protected]>
12627
12628 2022-10-11 Michael Chang <[email protected]>
12629
12630 util/grub-install: Set point of no return for powerpc-ieee1275 install
12631 The point of no return is used to define a point where no change should
12632 be reverted in a wake of fatal error that consequently aborts the
12633 process. The powerpc-ieee1275 install apparently missed this point of no
12634 return definition that newly installed modules could be inadvertently
12635 reverted after successful image embedding so that boot failure is
12636 incurred due to inconsistent state.
12637
12638 Reviewed-by: Daniel Kiper <[email protected]>
12639
12640 2022-10-11 Daniel Axtens <[email protected]>
12641
12642 disk/diskfilter: Check calloc() result for NULL
12643 With wildly corrupt inputs, we can end up trying to calloc a very
12644 large amount of memory, which will fail and give us a NULL pointer.
12645 We need to check that to avoid a crash. (And, even if we blocked
12646 such inputs, it is good practice to check the results of allocations
12647 anyway.)
12648
12649 Reviewed-by: Daniel Kiper <[email protected]>
12650
12651 2022-10-11 Glenn Washburn <[email protected]>
12652
12653 disk/cryptodisk: Allows UUIDs to be compared in a dash-insensitive manner
12654 A user can now specify UUID strings with dashes, instead of having to remove
12655 dashes. This is backwards-compatibility preserving and also fixes a source
12656 of user confusion over the inconsistency with how UUIDs are specified
12657 between file system UUIDs and cryptomount UUIDs. Since cryptsetup, the
12658 reference implementation for LUKS, displays and generates UUIDs with dashes
12659 there has been additional confusion when using the UUID strings from
12660 cryptsetup as exact input into GRUB does not find the expected cryptodisk.
12661
12662 A new function grub_uuidcasecmp() is added that is general enough to be used
12663 other places where UUIDs are being compared.
12664
12665 Reviewed-by: Patrick Steinhardt <[email protected]>
12666 Reviewed-by: Daniel Kiper <[email protected]>
12667
12668 2022-10-11 Glenn Washburn <[email protected]>
12669
12670 kern/corecmd: Quote variable values when displayed by the set command
12671 Variable values may contain spaces at the end or newlines. However, when
12672 displayed without quotes this is not obvious and can lead to confusion as
12673 to the actual contents of variables. Also for some variables grub_env_get()
12674 returns a NULL pointer instead of a pointer to an empty string and
12675 previously would be printed as "var=(null)". Now such variables will be
12676 displayed as "var=''".
12677
12678 Reviewed-by: Daniel Kiper <[email protected]>
12679
12680 2022-10-11 Samuel Thibault <[email protected]>
12681
12682 templates: Add support for acpi on Hurd
12683 This adds acpi as bootstrap module whenever it is available. This opens the
12684 path for proper IRQ routing for fully-userland disk drivers.
12685
12686 Reviewed-by: Daniel Kiper <[email protected]>
12687
12688 2022-10-11 Peter Jones <[email protected]>
12689
12690 util/grub-module-verifierXX: Enable running standalone checkers
12691 Allow treating util/grub-module-verifierXX.c as a file you can build
12692 directly so syntax checkers like vim's "syntastic" plugin, which uses
12693 "gcc -x c -fsyntax-only" to build it, will work.
12694
12695 One still has to do whatever setup is required to make it pick the
12696 right include dirs, which -I options we use, etc., but this makes
12697 it so you can do the checking on the file you're editing, rather
12698 than on a different file.
12699
12700 Reviewed-by: Daniel Kiper <[email protected]>
12701
12702 2022-10-04 Tuan Phan <[email protected]>
12703
12704 kern/compiler-rt: Fix __clzsi2() logic
12705 Fix the incorrect return value of __clzsi2() function.
12706
12707 Fixes: e795b90 (RISC-V: Add libgcc helpers for clz)
12708
12709 Reviewed-by: Daniel Kiper <[email protected]>
12710
12711 2022-10-04 Daniel Axtens <[email protected]>
12712
12713 efi: Increase default memory allocation to 32 MiB
12714 We have multiple reports of things being slower with a 1 MiB initial static
12715 allocation, and a report (more difficult to nail down) of a boot failure
12716 as a result of the smaller initial allocation.
12717
12718 Make the initial memory allocation 32 MiB.
12719
12720 Reviewed-by: Daniel Kiper <[email protected]>
12721
12722 2022-10-04 Christian Hesse <[email protected]>
12723
12724 templates: Filter C.UTF-8 locale for translation
12725 In addition to C locale there is also C.UTF-8 locale now. Filter that as
12726 well, by using ${grub_lang}, which contains a stripped value.
12727 This fixes the following message and resulting boot failure:
12728
12729 error: file `/boot/grub/locale/C.gmo' not found.
12730
12731 Reviewed-by: Daniel Kiper <[email protected]>
12732
12733 2022-10-04 Steve McIntyre <[email protected]>
12734
12735 tests: Explicitly unset SOURCE_DATE_EPOCH before running fs tests
12736 In some filesystem utils like mksquashfs, they will silently change
12737 behaviour and cause timestamps to unexpectedly change. Build
12738 environments like Debian's set SOURCE_DATE_EPOCH in the environment,
12739 so remove it. Reproducible builds are good and useful for shipped
12740 artifacts, but this causes build-time tests to fail.
12741
12742 Reviewed-by: Daniel Kiper <[email protected]>
12743
12744 2022-10-04 Heinrich Schuchardt <[email protected]>
12745
12746 commands/efi/lsefisystab: Short text for EFI_CONFORMANCE_PROFILES_TABLE
12747 The EFI_CONFORMANCE_PROFILES_TABLE_GUID is used for a table of GUIDs for conformance
12748 profiles (cf. UEFI specification 2.10, 4.6.5 EFI_CONFORMANCE_PROFILE_TABLE).
12749
12750 The lsefisystab command is used to display installed EFI configuration tables.
12751 Currently it only shows the GUID but not a short text for the table.
12752
12753 Provide a short text for the EFI_CONFORMANCE_PROFILES_TABLE_GUID.
12754
12755 Reviewed-by: Daniel Kiper <[email protected]>
12756
12757 2022-10-04 Theodore Ts'o <[email protected]>
12758
12759 fs/ext2: Ignore the large_dir incompat feature
12760 Recently, ext4 added the large_dir feature, which adds support for
12761 a 3 level htree directory support.
12762
12763 The GRUB supports existing file systems with htree directories by
12764 ignoring their existence, and since the index nodes for the hash tree
12765 look like deleted directory entries (by design), the GRUB can simply do
12766 a brute force O(n) linear search of directories. The same is true for
12767 3 level deep htrees indicated by large_dir feature flag.
12768
12769 Hence, it is safe for the GRUB to ignore the large_dir incompat feature.
12770
12771 Fixes: https://savannah.gnu.org/bugs/?61606
12772
12773 Reviewed-by: Daniel Kiper <[email protected]>
12774
12775 2022-10-04 Glenn Washburn <[email protected]>
12776
12777 disk/loopback: Support transparent decompression of backing file
12778 A new option is added to the loopback command, -D or --decompress, which
12779 when specified transparently decompresses the backing file. This allows
12780 compressed images to be used as if they were uncompressed.
12781
12782 Add documentation to support this change.
12783
12784 Suggested-by: Li Gen <[email protected]>
12785 Reviewed-by: Daniel Kiper <[email protected]>
12786
12787 2022-10-04 Glenn Washburn <[email protected]>
12788
12789 configure: Add -DGRUB_HAS_PCI when compiling C/C++ files on targets that support PCI
12790 The list of targets that support PCI is in gentpl.py. However, there is no
12791 support for generating makefile script from a .def file that will apply
12792 globally to the makefile, but on a per target basis. So instead, use
12793 gentpl.py in configure to get the list of targets and check if the current
12794 build target is one of them. If it is, set the automake conditional
12795 COND_HAVE_PCI. Then in conf/Makefile.common add -DGRUB_HAS_PCI for the
12796 platform if COND_HAVE_PCI is true.
12797
12798 Reviewed-by: Daniel Kiper <[email protected]>
12799
12800 2022-10-04 Li Gen <[email protected]>
12801
12802 commands/read: Fix overflow in grub_getline()
12803 Store returned value from grub_getkey() in int instead of char to
12804 prevent throwing away the extended bits. This was a problem because,
12805 for instance, the left arrow key press would return
12806 (GRUB_TERM_EXTENDED | 0x4b), which would have the GRUB_TERM_EXTENDED
12807 thrown away leaving 0x4b or 'K'. These extended keys should either
12808 work as intended or do nothing. This change has them do nothing,
12809 instead of inserting a key not pressed by the user.
12810
12811 Reviewed-by: Daniel Kiper <[email protected]>
12812
12813 2022-10-04 Li Gen <[email protected]>
12814
12815 efi: Correct function prototype for register_key_notify() method of grub_efi_simple_text_input_ex_interface
12816 The register_key_notify() method should have an output parameter which is
12817 a pointer to the unique handle assigned to the registered notification.
12818
12819 Reviewed-by: Daniel Kiper <[email protected]>
12820
12821 2022-10-04 Masahiro Matsuya <[email protected]>
12822
12823 net/drivers/ieee1275/ofnet: Fix incorrect netmask
12824 The netmask configured in firmware is not respected on ppc64 (big endian).
12825 When 255.255.252.0 is set as netmask in firmware, the following is the
12826 value of bootpath string in grub_ieee1275_parse_bootpath():
12827
12828 /vdevice/l-lan@30000002:speed=auto,duplex=auto,192.168.88.10,,192.168.89.113,192.168.88.1,5,5,255.255.252.0,512
12829
12830 The netmask in this bootpath is not a problem, since it's a value specified
12831 in firmware. But the value of subnet_mask.ipv4 was set with 0xfffffc00, and
12832 __builtin_ctz(~grub_le_to_cpu32(subnet_mask.ipv4)) returned 16 (not 22).
12833 As a result, 16 was used for netmask wrongly:
12834
12835 1111 1111 1111 1111 1111 1100 0000 0000 # subnet_mask.ipv4(=0xfffffc00)
12836 0000 0000 1111 1100 1111 1111 1111 1111 # grub_le_to_cpu32(subnet_mask.ipv4)
12837 1111 1111 0000 0011 0000 0000 0000 0000 # ~grub_le_to_cpu32(subnet_mask.ipv4)
12838
12839 and the count of zero with __builtin_ctz() can be 16. This patch changes
12840 it as below:
12841
12842 1111 1111 1111 1111 1111 1100 0000 0000 # subnet_mask.ipv4(=0xfffffc00)
12843 0000 0000 1111 1100 1111 1111 1111 1111 # grub_le_to_cpu32(subnet_mask.ipv4)
12844 1111 1111 1111 1111 1111 1100 0000 0000 # grub_be_to_cpu32(subnet_mask.ipv4)
12845 0000 0000 0000 0000 0000 0011 1111 1111 # ~grub_be_to_cpu32(subnet_mask.ipv4)
12846
12847 The count of zero with __builtin_clz() can be 22 (clz counts the number
12848 of one bits preceding the most significant zero bit).
12849
12850 Reviewed-by: Daniel Kiper <[email protected]>
12851
12852 2022-10-04 Ross Philipson <[email protected]>
12853
12854 loader/i386/bsd: Initialize BSD relocator state variables
12855 Numerous register fields in the relocator state are simply not
12856 used depending on the relocator. This causes Coverity to flag
12857 these fields but there is no real bug here. Simply initializing
12858 the variable to {0} solves the issue. Fixed in the else case too
12859 for consistency.
12860
12861 Fixes: CID 396932
12862
12863 Reviewed-by: Daniel Kiper <[email protected]>
12864
12865 2022-08-20 Andrea G. Monaco <[email protected]>
12866
12867 docs: Add a link to environment variables
12868 This is trivial, but it might save some time to beginners.
12869
12870 Reviewed-by: Glenn Washburn <[email protected]>
12871 Reviewed-by: Daniel Kiper <[email protected]>
12872
12873 2022-08-20 Robbie Harwood <[email protected]>
12874
12875 docs: Fix mismatched brackets in halt command
12876 Reviewed-by: Daniel Kiper <[email protected]>
12877
12878 docs: Document fwsetup command
12879 Reviewed-by: Daniel Kiper <[email protected]>
12880
12881 2022-08-20 Robbie Harwood <[email protected]>
12882
12883 efi: Don't display a uefi-firmware entry if it's not supported
12884 Add a new --is-supported option to commands/efi/efifwsetup and
12885 conditionalize display on it.
12886
12887 Reviewed-by: Daniel Kiper <[email protected]>
12888
12889 2022-08-20 Javier Martinez Canillas <[email protected]>
12890
12891 commands/efi/efifwsetup: Print an error if boot to firmware setup is not supported
12892 The "fwsetup" command is only registered if the firmware supports booting
12893 to the firmware setup UI. But it could be possible that the GRUB config
12894 already contains a "fwsetup" entry, because it was generated in a machine
12895 that has support for this feature.
12896
12897 To prevent users getting an error like:
12898
12899 error: ../../grub-core/script/function.c:109:can't find command `fwsetup'.
12900
12901 if it is not supported by the firmware, let's just always register the
12902 command but print a more accurate message if the firmware doesn't
12903 support this option.
12904
12905 Reviewed-by: Daniel Kiper <[email protected]>
12906
12907 2022-08-20 Javier Martinez Canillas <[email protected]>
12908
12909 templates: Check for EFI at runtime instead of config generation time
12910 The 30_uefi-firmware template checks if an OsIndicationsSupported UEFI var
12911 exists and EFI_OS_INDICATIONS_BOOT_TO_FW_UI bit is set, to decide whether
12912 a "fwsetup" menu entry would be added or not to the GRUB menu.
12913
12914 But this has the problem that it will only work if the configuration file
12915 was created on an UEFI machine that supports booting to a firmware UI.
12916
12917 This for example doesn't support creating GRUB config files when executing
12918 on systems that support both UEFI and legacy BIOS booting. Since creating
12919 the config file from legacy BIOS wouldn't allow to access the firmware UI.
12920
12921 To prevent this, make the template to unconditionally create the grub.cfg
12922 snippet but check at runtime if was booted through UEFI to decide if this
12923 entry should be added. That way it won't be added when booting with BIOS.
12924
12925 There's no need to check if EFI_OS_INDICATIONS_BOOT_TO_FW_UI bit is set,
12926 since that's already done by the "fwsetup" command when is executed.
12927
12928 Reviewed-by: Daniel Kiper <[email protected]>
12929
12930 2022-08-20 Robbie Harwood <[email protected]>
12931
12932 efi: Make all grub_efi_guid_t variables static
12933 This is believed to result in smaller code.
12934
12935 Reviewed-by: Daniel Kiper <[email protected]>
12936
12937 2022-08-20 Robbie Harwood <[email protected]>
12938
12939 commands/efi/efifwsetup: Add missing grub_free()s
12940 Each call of grub_efi_get_variable() needs a grub_free().
12941
12942 Reviewed-by: Daniel Kiper <[email protected]>
12943
12944 2022-08-19 Jagannathan Raman <[email protected]>
12945
12946 fs/zfs/zfs: Pass pointer to dnode_end_t instead of value to fill_fs_info()
12947 Coverity reports that dnode_end_t argument of fill_fs_info() is too
12948 large to pass-by-value. Therefore, replace the argument with a pointer.
12949
12950 Fixes: CID 73631
12951
12952 Reviewed-by: Darren Kenny <[email protected]>
12953 Reviewed-by: Daniel Kiper <[email protected]>
12954
12955 2022-08-19 Patrick Steinhardt <[email protected]>
12956
12957 disk/luks2: Fix decoding of digests and salts with escaped chars
12958 It was reported in the #grub IRC channel on Libera that decryption of
12959 LUKS2 partitions fails with errors about invalid digests and/or salts.
12960 In all of these cases, what failed was decoding the Base64
12961 representation of these, where the encoded data contained invalid
12962 characters.
12963
12964 As it turns out, the root cause is that json-c, which is used by
12965 cryptsetup to read and write the JSON header, will escape some
12966 characters by prepending a backslash when writing JSON strings by
12967 default. Most importantly, json-c also escapes the forward slash, which
12968 is part of the Base64 alphabet. Because GRUB doesn't know to unescape
12969 such characters, decoding this string will rightfully fail.
12970
12971 Interestingly, this issue has until now only been reported by users of
12972 Ubuntu 18.04. And a bit of digging in fact reveals that cryptsetup has
12973 changed the logic in a054206d (Suppress useless slash escaping in json
12974 lib, 2018-04-20), which has been released with cryptsetup v2.0.3. Ubuntu
12975 18.04 is still shipping with cryptsetup v2.0.2 though, which explains
12976 why this is not a more frequent issue.
12977
12978 Fix the issue by using our new grub_json_unescape() helper function
12979 that handles unescaping for us.
12980
12981 Reported-by: Afdal
12982 Reviewed-by: Daniel Kiper <[email protected]>
12983
12984 2022-08-19 Patrick Steinhardt <[email protected]>
12985
12986 lib/json/json: Add function to unescape JSON-encoded strings
12987 JSON strings require certain characters to be encoded, either by using
12988 a single reverse solidus character "\" for a set of popular characters,
12989 or by using a Unicode representation of "\uXXXXX". The jsmn library
12990 doesn't handle unescaping for us, so we must implement this functionality
12991 for ourselves.
12992
12993 Add a new function grub_json_unescape() that takes a potentially
12994 escaped JSON string as input and returns a new unescaped string.
12995
12996 Reviewed-by: Daniel Kiper <[email protected]>
12997
12998 2022-08-19 Nikita Ermakov <[email protected]>
12999
13000 loader: Drop argv[] argument in grub_initrd_load()
13001 In the case of an error grub_initrd_load() uses argv[] to print the
13002 filename that caused the error. It is also possible to obtain the
13003 filename from the file handles and there is no need to duplicate that
13004 information in argv[], so let's drop it.
13005
13006 Reviewed-by: Daniel Kiper <[email protected]>
13007
13008 2022-08-19 Alec Brown <[email protected]>
13009
13010 loader: Update error conditionals to use enums
13011 In grub-core/loader/i386/bsdXX.c and grub-core/loader/multiboot_elfxx.c, error
13012 conditionals are simplified to statements such as "if (err)". Even though the
13013 assumption that non-zero values give errors is correct, it would be clearer and
13014 more consistent to compare these conditionals to GRUB_ERR_NONE.
13015
13016 Reviewed-by: Darren Kenny <[email protected]>
13017 Reviewed-by: Daniel Kiper <[email protected]>
13018
13019 2022-08-19 Alec Brown <[email protected]>
13020
13021 util/grub-module-verifierXX: Changed get_shnum() return type
13022 In util/grub-module-verifierXX.c, the function get_shnum() returns the variable
13023 shnum, which is of the type Elf_Word. In the function, shnum can be obtained by
13024 the e_shnum member of an Elf_Ehdr or the sh_size member of an Elf_Shdr. The
13025 sh_size member can either be grub_uint32_t or grub_uint64_t, depending on the
13026 architecture, but Elf_Word is only grub_uint32_t. To account for when sh_size is
13027 grub_uint64_t, we can set shnum to have type Elf_Shnum and have get_shnum()
13028 return an Elf_Shnum.
13029
13030 Reviewed-by: Darren Kenny <[email protected]>
13031 Reviewed-by: Daniel Kiper <[email protected]>
13032
13033 2022-08-19 Alec Brown <[email protected]>
13034
13035 elf: Validate number of elf program header table entries
13036 In bsdXX.c and multiboot_elfxx.c, e_phnum is used to obtain the number of
13037 program header table entries, but it wasn't being checked if the value was
13038 there.
13039
13040 According to the elf(5) manual page,
13041 "If the number of entries in the program header table is larger than or equal to
13042 PN_XNUM (0xffff), this member holds PN_XNUM (0xffff) and the real number of
13043 entries in the program header table is held in the sh_info member of the
13044 initial entry in section header table. Otherwise, the sh_info member of the
13045 initial entry contains the value zero."
13046
13047 Since this check wasn't being made, grub_elfXX_get_phnum() is being added to
13048 elfXX.c to make this check and use e_phnum if it doesn't have PN_XNUM as a
13049 value, else use sh_info. We also need to make sure e_phnum isn't greater than
13050 PN_XNUM and sh_info isn't less than PN_XNUM.
13051
13052 Note that even though elf.c and elfXX.c are located in grub-core/kern, they are
13053 compiled as modules and don't need the EXPORT_FUNC() macro to define the functions
13054 in elf.h.
13055
13056 Also, changed casts of phnum to match variables being set as well as dropped
13057 casts when unnecessary.
13058
13059 Reviewed-by: Darren Kenny <[email protected]>
13060 Reviewed-by: Daniel Kiper <[email protected]>
13061
13062 2022-08-19 Alec Brown <[email protected]>
13063
13064 elf: Validate elf section header table index for section name string table
13065 In multiboot_elfxx.c, e_shstrndx is used to obtain the section header table
13066 index of the section name string table, but it wasn't being checked if the value
13067 was there.
13068
13069 According to the elf(5) manual page,
13070 "If the index of section name string table section is larger than or equal to
13071 SHN_LORESERVE (0xff00), this member holds SHN_XINDEX (0xffff) and the real
13072 index of the section name string table section is held in the sh_link member of
13073 the initial entry in section header table. Otherwise, the sh_link member of the
13074 initial entry in section header table contains the value zero."
13075
13076 Since this check wasn't being made, grub_elfXX_get_shstrndx() is being added to
13077 elfXX.c to make this check and use e_shstrndx if it doesn't have SHN_XINDEX as a
13078 value, else use sh_link. We also need to make sure e_shstrndx isn't greater than
13079 or equal to SHN_LORESERVE and sh_link isn't less than SHN_LORESERVE.
13080
13081 Note that even though elf.c and elfXX.c are located in grub-core/kern, they are
13082 compiled as modules and don't need the EXPORT_FUNC() macro to define the functions
13083 in elf.h.
13084
13085 Reviewed-by: Darren Kenny <[email protected]>
13086 Reviewed-by: Daniel Kiper <[email protected]>
13087
13088 2022-08-19 Alec Brown <[email protected]>
13089
13090 elf: Validate number of elf section header table entries
13091 In bsdXX.c and multiboot_elfxx.c, e_shnum is used to obtain the number of
13092 section header table entries, but it wasn't being checked if the value was
13093 there.
13094
13095 According to the elf(5) manual page,
13096 "If the number of entries in the section header table is larger than or equal to
13097 SHN_LORESERVE (0xff00), e_shnum holds the value zero and the real number of
13098 entries in the section header table is held in the sh_size member of the initial
13099 entry in section header table. Otherwise, the sh_size member of the initial
13100 entry in the section header table holds the value zero."
13101
13102 Since this check wasn't being made, grub_elfXX_get_shnum() is being added to
13103 elfXX.c to make this check and use whichever member doesn't have a value of
13104 zero. If both are zero, then we must return an error. We also need to make sure
13105 that e_shnum doesn't have a value greater than or equal to SHN_LORESERVE and
13106 sh_size isn't less than SHN_LORESERVE.
13107
13108 In order to get this function to work, the type ElfXX_Shnum is being added where
13109 Elf32_Shnum defines Elf32_Word and Elf64_Shnum defines Elf64_Xword. This new
13110 type is needed because if shnum obtains a value from sh_size, sh_size could be
13111 of type El32_Word for Elf32_Shdr structures or Elf64_Xword for Elf64_Shdr
13112 structures.
13113
13114 Note that even though elf.c and elfXX.c are located in grub-core/kern, they are
13115 compiled as modules and don't need the EXPORT_FUNC() macro to define the functions
13116 in elf.h.
13117
13118 For a few smaller changes, changed casts of shnum to match variables being set
13119 as well as dropped casts when unnecessary and fixed spacing errors in bsdXX.c.
13120 Also, shnum is an unsigned integer and is compared to int i in multiboot_elfxx.c,
13121 it should be unsigned to match shnum.
13122
13123 Reviewed-by: Darren Kenny <[email protected]>
13124 Reviewed-by: Daniel Kiper <[email protected]>
13125
13126 2022-08-19 Mauricio Faria de Oliveira <[email protected]>
13127
13128 templates/linux_xen: Properly order the multiple initrd files
13129 The linux_xen template orders the "early" initrd file(s) _first_
13130 (i.e., before the "real" initrd files) and that seems reasonable,
13131 as microcode updates usually come first.
13132
13133 However, this usually breaks Linux boot with initrd under Xen
13134 because Xen assumes the real initrd is the first multiboot[2]
13135 module after the kernel, passing its address over to Linux in
13136 Xen's start_info struct.
13137
13138 So, if a microcode-only initrd (i.e., without init/userspace)
13139 is found by grub-mkconfig, it ends up considered as a normal
13140 initrd by the Linux kernel, which cannot do anything with it
13141 (as it has no other files) and panic()s unable to mount root
13142 if it depends on a initrd to do that (e.g., root=UUID=...).
13143
13144 ...
13145
13146 Well, since Xen doesn't actually use the provided microcode
13147 by default / unless the 'ucode=<module number|scan>' option
13148 is enabled, this isn't used in the general case (and breaks).
13149
13150 Additionally, if an user enables the 'ucode=' option, that
13151 either specifies which module is to be used for microcode,
13152 or scans all modules (regardless of being first) for that.
13153
13154 Thus, for Xen:
13155 - it is *not required* to have microcode first,
13156 - but it is *required* to have real initrd first
13157
13158 So, fix it by ordering the real initrd before early initrd(s).
13159
13160 After:
13161
13162 # touch /boot/xen /boot/microcode.cpio
13163 # grub-mkconfig 2>/dev/null | grep -P '^\t(multiboot|module)'
13164 multiboot /boot/xen ...
13165 module /boot/vmlinuz-5.4.0-122-generic ...
13166 module --nounzip /boot/initrd.img-5.4.0-122-generic
13167 module --nounzip /boot/microcode.cpio
13168
13169 ...
13170
13171 Corner case specific to Xen implementation details:
13172
13173 It is actually _possible_ to have a microcode initrd first,
13174 but that requires a non-default option (so can't rely on it),
13175 and it turns out to be inconsistent with its counterpart
13176 (really shouldn't rely on it, as it may get confusing; below).
13177
13178 'ucode=1' does manually specify the first module is microcode
13179 _AND_ clears its bit in the module bitmap. The next module is
13180 now the 'new first', and gets passed to Linux as initrd. Good.
13181
13182 'ucode=scan' checks all modules for microcode, but does _NOT_
13183 clear a bit if it finds one (reasonable, as it can find that
13184 prepended in a "real" initrd anyway, which needs to be used).
13185 The first module still gets passed to Linux as initrd. Bad.
13186
13187 Fixes: e86f6aafb8de (grub-mkconfig/20_linux_xen: Support multiple early initrd images)
13188
13189 Acked-by: Juergen Gross <[email protected]>
13190 Reviewed-by: Daniel Kiper <[email protected]>
13191
13192 2022-08-19 Mauricio Faria de Oliveira <[email protected]>
13193
13194 templates/linux_xen: Properly load multiple initrd files
13195 The linux_xen template can put multiple initrd files in the
13196 same multiboot[2] module[2] command, which is against specs.
13197
13198 This causes ONLY the _first_ initrd file to be loaded; other
13199 files just have filenames in a "cmdline" string of the first
13200 initrd file and are NOT loaded.
13201
13202 Fix this by inserting a module[2] command per initrd file.
13203
13204 Before:
13205
13206 # touch /boot/xen /boot/microcode.cpio
13207 # grub-mkconfig 2>/dev/null | grep -P '^\t(multiboot|module)'
13208 multiboot /boot/xen ...
13209 module /boot/vmlinuz-5.4.0-122-generic ...
13210 module --nounzip /boot/microcode.cpio /boot/initrd.img-5.4.0-122-generic
13211
13212 After:
13213
13214 # touch /boot/xen /boot/microcode.cpio
13215 # grub-mkconfig 2>/dev/null | grep -P '^\t(multiboot|module)'
13216 multiboot /boot/xen ...
13217 module /boot/vmlinuz-5.4.0-122-generic ...
13218 module --nounzip /boot/microcode.cpio
13219 module --nounzip /boot/initrd.img-5.4.0-122-generic
13220
13221 Cause:
13222
13223 The code was copied from the linux template, which is *apparently*
13224 equivalent.. but its initrd command grub_cmd_initrd() *supports*
13225 multiple files (see grub_initrd_init()), while module/module2 in
13226 grub_cmd_module() *does not* (see grub_multiboot[2]_add_module()).
13227
13228 See commit e86f6aafb8de (grub-mkconfig/20_linux_xen: Support multiple early initrd images):
13229 'This is basically a copy of a698240d "grub-mkconfig/10_linux:
13230 Support multiple early initrd images" ...'
13231
13232 Specs:
13233
13234 Both multiboot and multiboot2 specifications mention support for
13235 'multiple boot modules' (struct/tag used for kernel/initrd files):
13236
13237 "Boot loaders don’t have to support multiple boot modules,
13238 but they are strongly encouraged to" [1,2]
13239
13240 However, there is a 1:1 relationship between boot modules and files,
13241 more or less clearly; note the usage of singular/plural "module(s)".
13242 (Multiboot2, clearly: "One tag appears per module".)
13243
13244 Multiboot [1]:
13245
13246 "the ‘mods’ fields indicate ... what boot modules
13247 were loaded ..., and where they can be found.
13248 ‘mods_count’ contains the number of modules loaded"
13249
13250 "The first two fields contain the start and end addresses
13251 of the boot module itself."
13252
13253 Multiboot2 [2]:
13254
13255 "This tag indicates ... what boot module was loaded ...,
13256 and where it can be found."
13257
13258 "The ‘mod_start’ and ‘mod_end’ contain the start and end
13259 physical addresses of the boot module itself."
13260
13261 "One tag appears per module.
13262 This tag type may appear multiple times."
13263
13264 And both clearly mention the 'string' field of a boot module,
13265 which is to be used by the operating system, not boot loader:
13266
13267 "The ‘string’ field provides an arbitrary string to be
13268 associated with that particular boot module ...
13269 its exact use is specific to the operating system."
13270
13271 Links:
13272
13273 [1] https://www.gnu.org/software/grub/manual/multiboot/multiboot.html
13274 3.3 Boot information format
13275
13276 [2] https://www.gnu.org/software/grub/manual/multiboot2/multiboot.html
13277 3.6.6 Modules
13278
13279 Fixes: e86f6aafb8de (grub-mkconfig/20_linux_xen: Support multiple early initrd images)
13280
13281 Acked-by: Juergen Gross <[email protected]>
13282 Reviewed-by: Daniel Kiper <[email protected]>
13283
13284 2022-08-19 Glenn Washburn <[email protected]>
13285
13286 misc: Add cast in grub_strncasecmp() to drop sign when calling grub_tolower()
13287 Note this cast was fixed in grub_strcasecmp() in commit ce41ab7aab
13288 (* grub-core/kern/misc.c (grub_strcmp): Use unsigned comparison as per
13289 common usage and preffered in several parts of code.), but this commit
13290 omitted fixing it in grub_strncasecmp().
13291
13292 Reviewed-by: Daniel Kiper <[email protected]>
13293
13294 2022-08-19 Glenn Washburn <[email protected]>
13295
13296 tests/util/grub-shell: Only show grub-mkrescue output if it returns an error
13297 The previous behavior ignored an error and the output from grub-mkrescue.
13298 This made it difficult to discover that grub-mkrescue was the reason that
13299 tests which rely on grub-shell were failing. Even after discovering
13300 grub-mkrescue was the culprit, there was no output to indicate why it was
13301 failing. It turns out that grub-mkrescue is a thin wrapper around xorriso.
13302 So if you do not have xorriso installed it will fail with an error message
13303 about not being able to find xorriso.
13304
13305 This change will allow grub-mkrescue output to be written to stderr, only
13306 if grub-mkrescue fails. If grub-mkrescue succeeds, there will be no output
13307 from grub-mkrescue so as not to interfere with the functioning of tests.
13308 This change should have no effect on the running of tests or other uses of
13309 grub-shell as it only modifies the error path.
13310
13311 Also, if grub-mkrescue fails, the script exits early. Since grub-shell
13312 needs the ISO image created by grub-mkresue to boot the QEMU instance,
13313 a failure here should be considered fatal.
13314
13315 Reviewed-by: Daniel Kiper <[email protected]>
13316
13317 2022-08-19 Ard Biesheuvel <[email protected]>
13318
13319 loader/arm64/linux: Remove magic number header field check
13320 The "ARM\x64" magic number in the file header identifies an image as one
13321 that implements the bare metal boot protocol, allowing the loader to
13322 simply move the file to a suitably aligned address in memory, with
13323 sufficient headroom for the trailing .bss segment (the required memory
13324 size is described in the header as well).
13325
13326 Note of this matters for GRUB, as it only supports EFI boot. EFI does
13327 not care about this magic number, and nor should GRUB: this prevents us
13328 from booting other PE linux images, such as the generic EFI zboot
13329 decompressor, which is a pure PE/COFF image, and does not implement the
13330 bare metal boot protocol.
13331
13332 So drop the magic number check.
13333
13334 Reviewed-by: Daniel Kiper <[email protected]>
13335
13336 2022-08-19 Darren Kenny <[email protected]>
13337
13338 util/grub-install-common: Confirm directory creation in grub_install_mkdir_p()
13339 Because grub_util_mkdir() is implemented to not return a value on any
13340 platform, grub_instal_mkdir_p() can test for success by confirming that
13341 the directory requested exists after attempting to create it, otherwise
13342 it should fail with an error and exit.
13343
13344 While fixing this, a flaw in the logic was shown, where the first match
13345 of the path separator, which almost always was the first character in
13346 the path (e.g. /boot/grub2) would result in creating a directory with an
13347 empty name (i.e. ""). To avoid that, it should skip the handling of the
13348 path separator where p is pointing to the first character.
13349
13350 Reviewed-by: Daniel Kiper <[email protected]>
13351
13352 2022-08-19 Darren Kenny <[email protected]>
13353
13354 util: Ignore return value for grub_util_mkdir() on all platforms
13355 Coverity signaled 2 issues where the return value of grub_util_mkdir()
13356 was not being tested.
13357
13358 The Windows variant of this code defines the function as having no
13359 return value (void), but the UNIX variants all are mapped using a macro
13360 to the libc mkdir() function, which returns an int value.
13361
13362 To be consistent, the mapping should cast to void to for these too.
13363
13364 Fixes: CID 73583
13365 Fixes: CID 73617
13366
13367 Reviewed-by: Daniel Kiper <[email protected]>
13368
13369 2022-08-19 Glenn Washburn <[email protected]>
13370
13371 disk/cryptodisk: Support encrypted volumes using detached headers on a partition
13372 Update the read hook to take into account encrypted volumes on a partition.
13373 GRUB disk read hooks supply an absolute sector number at which the read is
13374 started from. If the encrypted volume is in a partition, the sector number
13375 given to the read hook will be offset by the number of the sector at the
13376 start of the partition. The read hook then needs to subtract the partition
13377 start from the supplied sector to get the correct start sector for the read
13378 into the detached header file.
13379
13380 Reported-by: brutser <[email protected]>
13381 Tested-by: brutser <[email protected]>
13382 Reviewed-by: Daniel Kiper <[email protected]>
13383
13384 2022-08-10 Glenn Washburn <[email protected]>
13385
13386 tests/util/grub-shell: Use shell variable instead of autoconf
13387 By using shell variable that are set once by the expansion of an autoconf
13388 variable, the resulting shell script is more easily moved and modified
13389 from the build/install directory it was generated for. The resulting
13390 script is more readable as well.
13391
13392 Reviewed-by: Daniel Kiper <[email protected]>
13393
13394 2022-08-10 Stefan Agner <[email protected]>
13395
13396 Makefile: Make grub_fstest.pp depend on config-util.h
13397 If you build with "make -j25", sometimes you see:
13398
13399 /build/output_generic_x86_64/host/bin/x86_64-buildroot-linux-gnu-gcc -E -DHAVE_CONFIG_H -I. -I.. -Wall -W -DGRUB_UTIL=1 -D_FILE_OFFSET_BITS=64 -I./include -DGRUB_FILE=\"util/grub-fstest.c\" -I. -I.. -I. -I.. -I../include -I./include -I../grub-core/lib/libgcrypt-grub/src/ -I./grub-core/lib/gnulib -I../grub-core/lib/gnulib -D_LARGEFILE_SOURCE -D_LARGEFILE64_SOURCE -D_FILE_OFFSET_BITS=64 -Os -fno-stack-protector -D_FILE_OFFSET_BITS=64 \
13400 -D'GRUB_MOD_INIT(x)=@MARKER@x@' ../util/grub-fstest.c ../grub-core/kern/emu/hostfs.c ../grub-core/disk/host.c ../grub-core/osdep/init.c > grub_fstest.pp || (rm -f grub_fstest.pp; exit 1)
13401 config.status: creating config-util.h
13402 ../grub-core/kern/emu/hostfs.c:20:10: fatal error: config-util.h: No such file or directory
13403 20 | #include <config-util.h>
13404 | ^~~~~~~~~~~~~~~
13405 compilation terminated.
13406
13407 Reviewed-by: Daniel Kiper <[email protected]>
13408
13409 2022-08-10 Qiumiao Zhang <[email protected]>
13410
13411 util/grub-mkfont: Fix resource leaks
13412 Reviewed-by: Daniel Kiper <[email protected]>
13413
13414 2022-08-10 Peter Jones <[email protected]>
13415
13416 kern/i386/tsc_pmtimer: Make pmtimer tsc calibration not take 51 seconds to fail
13417 On my laptop running at 2.4GHz, if I run a VM where tsc calibration
13418 using pmtimer will fail presuming a broken pmtimer, it takes ~51 seconds
13419 to do so (as measured with the stopwatch on my phone), with a tsc delta
13420 of 0x1cd1c85300, or around 125 billion cycles.
13421
13422 If instead of trying to wait for 5-200ms to show up on the pmtimer, we
13423 try to wait for 5-200us, it decides it's broken in ~0x2626aa0 TSCs, aka
13424 ~2.4 million cycles, or more or less instantly.
13425
13426 Additionally, this reading the pmtimer was returning 0xffffffff anyway,
13427 and that's obviously an invalid return. I've added a check for that and
13428 0 so we don't bother waiting for the test if what we're seeing is dead
13429 pins with no response at all.
13430
13431 If "debug" includes "pmtimer", you will see one of the following three
13432 outcomes. If pmtimer gives all 0 or all 1 bits, you will see:
13433
13434 pmtimer: 0xffffff bad_reads: 1
13435 pmtimer: 0xffffff bad_reads: 2
13436 pmtimer: 0xffffff bad_reads: 3
13437 pmtimer: 0xffffff bad_reads: 4
13438 pmtimer: 0xffffff bad_reads: 5
13439 pmtimer: 0xffffff bad_reads: 6
13440 pmtimer: 0xffffff bad_reads: 7
13441 pmtimer: 0xffffff bad_reads: 8
13442 pmtimer: 0xffffff bad_reads: 9
13443 pmtimer: 0xffffff bad_reads: 10
13444 timer is broken; giving up.
13445
13446 This outcome was tested using qemu+kvm with UEFI (OVMF) firmware and
13447 these options: -machine pc-q35-2.10 -cpu Broadwell-noTSX
13448
13449 If pmtimer gives any other bit patterns but is not actually marching
13450 forward fast enough to use for clock calibration, you will see:
13451
13452 pmtimer delta is 0x0 (1904 iterations)
13453 tsc delta is implausible: 0x2626aa0
13454
13455 This outcome was tested using GRUB patched to not ignore bad reads using
13456 qemu+kvm with UEFI (OVMF) firmware, and these options:
13457 -machine pc-q35-2.10 -cpu Broadwell-noTSX
13458
13459 If pmtimer actually works, you'll see something like:
13460
13461 pmtimer delta is 0xdff
13462 tsc delta is 0x278756
13463
13464 This outcome was tested using qemu+kvm with UEFI (OVMF) firmware, and
13465 these options: -machine pc-i440fx-2.4 -cpu Broadwell-noTSX
13466
13467 I've also tested this outcome on a real Intel Xeon E3-1275v3 on an Intel
13468 Server Board S1200V3RPS using the SDV.RP.B8 "Release" build here:
13469 https://www.intel.com/content/www/us/en/download/674448/firmware-update-for-the-intel-server-board-s1200rp-uefi-development-kit-release-vb8.html
13470
13471 Reviewed-by: Daniel Kiper <[email protected]>
13472
13473 2022-08-10 Glenn Washburn <[email protected]>
13474
13475 disk/luks2: Continue trying all keyslots even if there are some failures
13476 luks2_get_keyslot() can fail for a variety of reasons that do not necessarily
13477 mean the next keyslot should not be tried (e.g. a new kdf type). So always
13478 try the next slot. This will make GRUB more resilient to non-spec json data
13479 that 3rd party systems may add. We do not care if some of the keyslots are
13480 unusable, only if there is at least one that is.
13481
13482 Reviewed-by: Daniel Kiper <[email protected]>
13483
13484 2022-08-10 Glenn Washburn <[email protected]>
13485
13486 efi: Add efitextmode command for getting/setting the text mode resolution
13487 This command is meant to behave similarly to the "mode" command of the EFI
13488 Shell application. In addition to allowing mode selection by giving the
13489 number of columns and rows as arguments, the command allows specifying the
13490 mode number to select the mode. Also supported are the arguments "min" and
13491 "max", which set the mode to the minimum and maximum mode respectively as
13492 calculated by the columns * rows of that mode.
13493
13494 Reviewed-by: Daniel Kiper <[email protected]>
13495
13496 2022-07-27 Robbie Harwood <[email protected]>
13497
13498 fs/fat: Don't error when mtime is 0
13499 In the wild, we occasionally see valid ESPs where some file modification
13500 times are 0. For instance:
13501
13502 ├── [Dec 31 1979] EFI
13503 │ ├── [Dec 31 1979] BOOT
13504 │ │ ├── [Dec 31 1979] BOOTX64.EFI
13505 │ │ └── [Dec 31 1979] fbx64.efi
13506 │ └── [Jun 27 02:41] fedora
13507 │ ├── [Dec 31 1979] BOOTX64.CSV
13508 │ ├── [Dec 31 1979] fonts
13509 │ ├── [Mar 14 03:35] fw
13510 │ │ ├── [Mar 14 03:35] fwupd-359c1169-abd6-4a0d-8bce-e4d4713335c1.cap
13511 │ │ ├── [Mar 14 03:34] fwupd-9d255c4b-2d88-4861-860d-7ee52ade9463.cap
13512 │ │ └── [Mar 14 03:34] fwupd-b36438d8-9128-49d2-b280-487be02d948b.cap
13513 │ ├── [Dec 31 1979] fwupdx64.efi
13514 │ ├── [May 10 10:47] grub.cfg
13515 │ ├── [Jun 3 12:38] grub.cfg.new.new
13516 │ ├── [May 10 10:41] grub.cfg.old
13517 │ ├── [Jun 27 02:41] grubenv
13518 │ ├── [Dec 31 1979] grubx64.efi
13519 │ ├── [Dec 31 1979] mmx64.efi
13520 │ ├── [Dec 31 1979] shim.efi
13521 │ ├── [Dec 31 1979] shimx64.efi
13522 │ └── [Dec 31 1979] shimx64-fedora.efi
13523 └── [Dec 31 1979] FSCK0000.REC
13524
13525 5 directories, 17 files
13526
13527 This causes grub-probe failure, which in turn causes grub-mkconfig
13528 failure. They are valid filesystems that appear intact, and the Linux
13529 FAT stack is able to mount and manipulate them without complaint.
13530
13531 The check for mtime of 0 has been present since
13532 20def1a3c3952982395cd7c3ea7e78638527962b (fat: support file
13533 modification times).
13534
13535 Reviewed-by: Daniel Kiper <[email protected]>
13536
13537 2022-07-27 Robbie Harwood <[email protected]>
13538
13539 kern/fs: The grub_fs_probe() should dprint errors from filesystems
13540 When filesystem detection fails, all that's currently debug-logged is
13541 a series of messages like:
13542
13543 grub-core/kern/fs.c:56:fs: Detecting ntfs...
13544 grub-core/kern/fs.c:76:fs: ntfs detection failed.
13545
13546 repeated for each filesystem. Any messages provided to grub_error() by
13547 the filesystem are lost, and one has to break out gdb to figure out what
13548 went wrong.
13549
13550 With this change, one instead sees:
13551
13552 grub-core/kern/fs.c:56:fs: Detecting fat...
13553 grub-core/osdep/hostdisk.c:357:hostdisk: reusing open device
13554 `/path/to/device'
13555 grub-core/kern/fs.c:77:fs: error: invalid modification timestamp for /.
13556 grub-core/kern/fs.c:79:fs: fat detection failed.
13557
13558 in the debug prints.
13559
13560 Reviewed-by: Daniel Kiper <[email protected]>
13561
13562 2022-07-27 Robbie Harwood <[email protected]>
13563
13564 util/grub-probe: Document the behavior of multiple -v
13565 Reviewed-by: Daniel Kiper <[email protected]>
13566
13567 2022-07-27 Ross Philipson <[email protected]>
13568
13569 lib/relocator: Initialize local relocator subchunk struct to all zeros
13570 The way the code is written the tofree variable would never be passed to
13571 the free_subchunk() function uninitialized. Coverity cannot determine
13572 this and flags the situation as "Using uninitialized value...". The fix
13573 is just to initialize the local struct.
13574
13575 Fixes: CID 314016
13576
13577 Reviewed-by: Darren Kenny <[email protected]>
13578 Tested-by: Alec Brown <[email protected]>
13579 Reviewed-by: Daniel Kiper <[email protected]>
13580
13581 2022-07-27 Lu Ken <[email protected]>
13582
13583 efi/tpm: Add EFI_CC_MEASUREMENT_PROTOCOL support
13584 The EFI_CC_MEASUREMENT_PROTOCOL abstracts the measurement for virtual firmware
13585 in confidential computing environment. It is similar to the EFI_TCG2_PROTOCOL.
13586 It was proposed by Intel and ARM and approved by UEFI organization.
13587
13588 It is defined in Intel GHCI specification: https://cdrdv2.intel.com/v1/dl/getContent/726790 .
13589 The EDKII header file is available at https://github.com/tianocore/edk2/blob/master/MdePkg/Include/Protocol/CcMeasurement.h .
13590
13591 Reviewed-by: Daniel Kiper <[email protected]>
13592
13593 2022-07-27 Lu Ken <[email protected]>
13594
13595 commands/efi/tpm: Use grub_strcpy() instead of grub_memcpy()
13596 The event description is a string, so using grub_strcpy() is cleaner than
13597 using grub_memcpy().
13598
13599 Reviewed-by: Daniel Kiper <[email protected]>
13600
13601 2022-07-27 Lu Ken <[email protected]>
13602
13603 commands/efi/tpm: Refine the status of log event
13604 1. Use macro GRUB_ERR_NONE instead of hard code 0.
13605 2. Keep lowercase of the first char for the status string of log event.
13606
13607 Reviewed-by: Daniel Kiper <[email protected]>
13608
13609 2022-07-12 Nicholas Vinson <[email protected]>
13610
13611 configure: Warn if stack protector is not allowed
13612 Introduce ERROR_PLATFORM_NOT_SUPPORT_SSP environment variable to treat
13613 the "--enable-stack-protector is only supported on EFI platforms" message
13614 as a warning instead of an error. If ERROR_PLATFORM_NOT_SUPPORT_SSP is
13615 set to "no" (case-insensitive), then the message will be printed as
13616 a warning. Otherwise, it prints as an error. The default behavior is to
13617 print the message as an error.
13618
13619 For any wrapper build script that has some variation of:
13620
13621 for p in SELECTED_GRUB_PLATFORMS; do \
13622 configure --enable-stack-protector \
13623 --with-platform${P} ... || die; \
13624 done
13625 make
13626
13627 The GRUB will fail to build if SELECTED_GRUB_PLATFORMS contains a platform
13628 that does not support SSP.
13629
13630 Such wrapper scripts need to work-around this issue by modifying the
13631 above for-loop, so it conditionally passes --enable-stack-protector to
13632 configure for the proper GRUB platform(s).
13633
13634 However, if the above example is modified to have to conditionally pass
13635 in --enable-stack-protector, its behavior is effectively the same as the
13636 proposed change. Additionally, The list of SSP supported platforms is
13637 now in 2 places. One in the configure script and one in the build wrapper
13638 script. If the second list is not properly maintained it could mistakenly
13639 disable SSP for a platform that later gained support for it.
13640
13641 Reviewed-by: Daniel Kiper <[email protected]>
13642
13643 2022-07-12 Darren Kenny <[email protected]>
13644
13645 util/grub-mkfont: Fix tainted loop boundary issues with substitutions
13646 With gsub substitutions the offsets should be validated against the
13647 number of glyphs in a font face and the memory allocated for the gsub
13648 substitution data.
13649
13650 Both the number of glyphs and the last address in the allocated data are
13651 passed in to process_cursive(), where the number of glyphs validates the end
13652 of the range.
13653
13654 Enabling memory allocation validation uses two macros, one to simply check the
13655 address against the allocated space, and the other to check that the number of
13656 items of a given size doesn't extend outside of the allocated space.
13657
13658 Fixes: CID 73770
13659 Fixes: CID 314040
13660
13661 Reviewed-by: Daniel Kiper <[email protected]>
13662
13663 2022-07-12 Glenn Washburn <[email protected]>
13664
13665 efi: Add missing header from include/grub/efi/console_control.h
13666 Reviewed-by: Daniel Kiper <[email protected]>
13667
13668 2022-07-04 Glenn Washburn <[email protected]>
13669
13670 disk: Replace code that calculates the log of sector size with grub_log2ull()
13671 Reviewed-by: Daniel Kiper <[email protected]>
13672
13673 2022-07-04 Mathieu Desnoyers <[email protected]>
13674
13675 templates: Remove unused version comparison functions
13676 There are no users left of version_find_latest(), version_test_gt(), and
13677 version_test_numeric(). Remove those unused helper functions. Using
13678 those helper functions is what caused the quadratic sorting performance
13679 issues in the first place, so removing them is a net win.
13680
13681 Reviewed-by: Robbie Harwood <[email protected]>
13682 Reviewed-by: Daniel Kiper <[email protected]>
13683
13684 2022-07-04 Mathieu Desnoyers <[email protected]>
13685
13686 templates/kfreebsd: Fix quadratic algorithm for sorting menu items
13687 The current implementation of the 10_kfreebsd script implements its menu
13688 items sorting in bash with a quadratic algorithm, calling "sed", "sort",
13689 "head", and "grep" to compare versions between individual lines, which
13690 is annoyingly slow for kernel developers who can easily end up with
13691 50-100 kernels in their boot partition.
13692
13693 This fix is ported from the 10_linux script, which has a similar
13694 quadratic code pattern.
13695
13696 Cc: [email protected]
13697 Reviewed-by: Daniel Kiper <[email protected]>
13698
13699 2022-07-04 Mathieu Desnoyers <[email protected]>
13700
13701 templates/hurd: Fix quadratic algorithm for sorting menu items
13702 The current implementation of the 10_hurd script implements its menu
13703 items sorting in bash with a quadratic algorithm, calling "sed", "sort",
13704 "head", and "grep" to compare versions between individual lines, which
13705 is annoyingly slow for kernel developers who can easily end up with
13706 50-100 kernels in their boot partition.
13707
13708 This fix is ported from the 10_linux script, which has a similar
13709 quadratic code pattern.
13710
13711 Cc: Samuel Thibault <[email protected]>
13712 Tested-by: Samuel Thibault <[email protected]>
13713 Reviewed-by: Daniel Kiper <[email protected]>
13714
13715 2022-07-04 Mathieu Desnoyers <[email protected]>
13716
13717 templates/linux_xen: Fix quadratic algorithm for sorting menu items
13718 The current implementation of the 20_linux_xen script implements its
13719 menu items sorting in bash with a quadratic algorithm, calling "sed",
13720 "sort", "head", and "grep" to compare versions between individual lines,
13721 which is annoyingly slow for kernel developers who can easily end up
13722 with 50-100 kernels in their boot partition.
13723
13724 This fix is ported from the 10_linux script, which has a similar
13725 quadratic code pattern.
13726
13727 Cc: [email protected]
13728 Tested-by: Jason Andryuk <[email protected]>
13729 Reviewed-by: Daniel Kiper <[email protected]>
13730
13731 2022-07-04 Mathieu Desnoyers <[email protected]>
13732
13733 templates/linux: Fix quadratic algorithm for sorting menu items
13734 The current implementation of the 10_linux script implements its menu
13735 items sorting in bash with a quadratic algorithm, calling "sed", "sort",
13736 "head", and "grep" to compare versions between individual lines, which
13737 is annoyingly slow for kernel developers who can easily end up with
13738 50-100 kernels in /boot.
13739
13740 As an example, on a Intel(R) Core(TM) i7-8650U CPU @ 1.90GHz, running:
13741
13742 /usr/sbin/grub-mkconfig > /dev/null
13743
13744 With 44 kernels in /boot, this command takes 10-15 seconds to complete.
13745 After this fix, the same command runs in 5 seconds.
13746
13747 With 116 kernels in /boot, this command takes 40 seconds to complete.
13748 After this fix, the same command runs in 8 seconds.
13749
13750 For reference, the quadratic algorithm here is:
13751
13752 while [ "x$list" != "x" ] ; do <--- outer loop
13753 linux=`version_find_latest $list`
13754 version_find_latest()
13755 for i in "$@" ; do <--- inner loop
13756 version_test_gt()
13757 fork+exec sed
13758 version_test_numeric()
13759 version_sort
13760 fork+exec sort
13761 fork+exec head -n 1
13762 fork+exec grep
13763 list=`echo $list | tr ' ' '\n' | fgrep -vx "$linux" | tr '\n' ' '`
13764 tr
13765 fgrep
13766 tr
13767
13768 So all commands executed under version_test_gt() are executed
13769 O(n^2) times where n is the number of kernel images in /boot.
13770
13771 Here is the improved algorithm proposed:
13772 - Prepare a list with all the relevant information for ordering by a single
13773 sort(1) execution. This is done by renaming ".old" suffixes by " 1" and
13774 by suffixing all other files with " 2", thus making sure the ".old" entries
13775 will follow the non-old entries in reverse-sorted-order.
13776 - Call version_reverse_sort on the list (sort -r -V): A single execution of
13777 sort(1). For instance, GNU coreutils' sort will reverse-sort the list in
13778 O(n*log(n)) with a merge sort.
13779 - Replace the " 1" suffixes by ".old", and remove the " 2" suffixes.
13780 - Iterate on the reverse-sorted list to output each menu entry item.
13781
13782 Therefore, the algorithm proposed has O(n*log(n)) complexity with GNU
13783 coreutils' sort compared to the prior O(n^2) complexity. Moreover, the
13784 constant time required for each list entry is much less because sorting
13785 is done within a single execution of sort(1) rather than requiring
13786 O(n^2) executions of sed(1), sort(1), head(1), and grep(1) in
13787 sub-shells.
13788
13789 Reviewed-by: Robbie Harwood <[email protected]>
13790 Reviewed-by: Daniel Kiper <[email protected]>
13791
13792 2022-07-04 Glenn Washburn <[email protected]>
13793
13794 docs: Add documentation on detached header option to cryptomount
13795 Reviewed-by: Patrick Steinhardt <[email protected]>
13796 Reviewed-by: Daniel Kiper <[email protected]>
13797
13798 2022-07-04 Glenn Washburn <[email protected]>
13799
13800 cryptodisk: Add support for using detached header files
13801 Using the disk read hook mechanism, setup a read hook on the source disk
13802 which will read from the given header file during the scan and recovery
13803 cryptodisk backend functions. Disk read hooks are executed after the data
13804 has been read from the disk. This is okay, because the read hook is given
13805 the read buffer before its sent back to the caller. In this case, the hook
13806 can then overwrite the data read from the disk device with data from the
13807 header file sent in as the read hook data. This is transparent to the
13808 read caller. Since the callers of this function have just opened the
13809 source disk, there are no current read hooks, so there's no need to
13810 save/restore them nor consider if they should be called or not.
13811
13812 This hook assumes that the header is at the start of the volume, which
13813 is not the case for some formats (e.g. GELI). So GELI will return an
13814 error if a detached header is specified. It also can only be used
13815 with formats where the detached header file can be written to the
13816 first blocks of the volume and the volume could still be unlocked.
13817 So the header file can not be formatted differently from the on-disk
13818 header. If these assumpts are not met, detached header file processing
13819 must be specially handled in the cryptodisk backend module.
13820
13821 The hook will be called potentially many times by a backend. This is fine
13822 because of the assumptions mentioned and the read hook reads from absolute
13823 offsets and is stateless.
13824
13825 Also add a --header (short -H) option to cryptomount which takes a file
13826 argument.
13827
13828 Reviewed-by: Patrick Steinhardt <[email protected]>
13829 Reviewed-by: Daniel Kiper <[email protected]>
13830
13831 2022-07-04 Glenn Washburn <[email protected]>
13832
13833 disk: Allow read hook callback to take read buffer to potentially modify it
13834 It will be desirable in the future to allow having the read hook modify the
13835 data passed back from a read function call on a disk or file. This adds that
13836 infrastructure and has no impact on code flow for existing uses of the read
13837 hook. Also changed is that now when the read hook callback is called it can
13838 also indicate what error code should be sent back to the read caller.
13839
13840 Reviewed-by: Patrick Steinhardt <[email protected]>
13841 Reviewed-by: Daniel Kiper <[email protected]>
13842
13843 2022-07-04 Glenn Washburn <[email protected]>
13844
13845 docs: Document undocumented variables
13846 Document the variables net_<interface>_clientid, net_<interface>_clientuuid,
13847 lockdown, and shim_lock in the list of special environment variables.
13848
13849 Reviewed-by: Daniel Kiper <[email protected]>
13850
13851 2022-07-04 Patrick Steinhardt <[email protected]>
13852
13853 kern/efi/mm: Implement runtime addition of pages
13854 Adjust the interface of grub_efi_mm_add_regions() to take a set of
13855 GRUB_MM_ADD_REGION_* flags, which most notably is currently only the
13856 GRUB_MM_ADD_REGION_CONSECUTIVE flag. This allows us to set the function
13857 up as callback for the memory subsystem and have it call out to us in
13858 case there's not enough pages available in the current heap.
13859
13860 Reviewed-by: Daniel Kiper <[email protected]>
13861 Tested-by: Patrick Steinhardt <[email protected]>
13862
13863 2022-07-04 Patrick Steinhardt <[email protected]>
13864
13865 kern/efi/mm: Pass up errors from add_memory_regions()
13866 The function add_memory_regions() is currently only called on system
13867 initialization to allocate a fixed amount of pages. As such, it didn't
13868 need to return any errors: in case it failed, we cannot proceed anyway.
13869 This will change with the upcoming support for requesting more memory
13870 from the firmware at runtime, where it doesn't make sense anymore to
13871 fail hard.
13872
13873 Refactor the function to return an error to prepare for this. Note that
13874 this does not change the behaviour when initializing the memory system
13875 because grub_efi_mm_init() knows to call grub_fatal() in case
13876 grub_efi_mm_add_regions() returns an error.
13877
13878 Reviewed-by: Daniel Kiper <[email protected]>
13879 Tested-by: Patrick Steinhardt <[email protected]>
13880
13881 2022-07-04 Patrick Steinhardt <[email protected]>
13882
13883 kern/efi/mm: Extract function to add memory regions
13884 In preparation of support for runtime-allocating additional memory
13885 region, this patch extracts the function to retrieve the EFI memory
13886 map and add a subset of it to GRUB's own memory regions.
13887
13888 Reviewed-by: Daniel Kiper <[email protected]>
13889 Tested-by: Patrick Steinhardt <[email protected]>
13890
13891 2022-07-04 Patrick Steinhardt <[email protected]>
13892
13893 kern/efi/mm: Always request a fixed number of pages on init
13894 When initializing the EFI memory subsystem, we will by default request
13895 a quarter of the available memory, bounded by a minimum/maximum value.
13896 Given that we're about to extend the EFI memory system to dynamically
13897 request additional pages from the firmware as required, this scaling of
13898 requested memory based on available memory will not make a lot of sense
13899 anymore.
13900
13901 Remove this logic as a preparatory patch such that we'll instead defer
13902 to the runtime memory allocator. Note that ideally, we'd want to change
13903 this after dynamic requesting of pages has been implemented for the EFI
13904 platform. But because we'll need to split up initialization of the
13905 memory subsystem and the request of pages from the firmware, we'd have
13906 to duplicate quite some logic at first only to remove it afterwards
13907 again. This seems quite pointless, so we instead have patches slightly
13908 out of order.
13909
13910 Reviewed-by: Daniel Kiper <[email protected]>
13911 Tested-by: Patrick Steinhardt <[email protected]>
13912
13913 2022-07-04 Patrick Steinhardt <[email protected]>
13914
13915 mm: Allow dynamically requesting additional memory regions
13916 Currently, all platforms will set up their heap on initialization of the
13917 platform code. While this works mostly fine, it poses some limitations
13918 on memory management on us. Most notably, allocating big chunks of
13919 memory in the gigabyte range would require us to pre-request this many
13920 bytes from the firmware and add it to the heap from the beginning on
13921 some platforms like EFI. As this isn't needed for most configurations,
13922 it is inefficient and may even negatively impact some usecases when,
13923 e.g., chainloading. Nonetheless, allocating big chunks of memory is
13924 required sometimes, where one example is the upcoming support for the
13925 Argon2 key derival function in LUKS2.
13926
13927 In order to avoid pre-allocating big chunks of memory, this commit
13928 implements a runtime mechanism to add more pages to the system. When
13929 a given allocation cannot be currently satisfied, we'll call a given
13930 callback set up by the platform's own memory management subsystem,
13931 asking it to add a memory area with at least "n" bytes. If this
13932 succeeds, we retry searching for a valid memory region, which should
13933 now succeed.
13934
13935 If this fails, we try asking for "n" bytes, possibly spread across
13936 multiple regions, in hopes that region merging means that we end up
13937 with enough memory for things to work out.
13938
13939 Tested-by: Stefan Berger <[email protected]>
13940 Reviewed-by: Daniel Kiper <[email protected]>
13941 Tested-by: Patrick Steinhardt <[email protected]>
13942
13943 2022-07-04 Patrick Steinhardt <[email protected]>
13944
13945 mm: Drop unused unloading of modules on OOM
13946 In grub_memalign(), there's a commented section which would allow for
13947 unloading of unneeded modules in case where there is not enough free
13948 memory available to satisfy a request. Given that this code is never
13949 compiled in, let's remove it together with grub_dl_unload_unneeded().
13950
13951 Reviewed-by: Daniel Kiper <[email protected]>
13952 Tested-by: Patrick Steinhardt <[email protected]>
13953
13954 2022-07-04 Daniel Axtens <[email protected]>
13955
13956 mm: Debug support for region operations
13957 This is handy for debugging. Enable with "set debug=regions".
13958
13959 Reviewed-by: Daniel Kiper <[email protected]>
13960 Tested-by: Patrick Steinhardt <[email protected]>
13961
13962 2022-07-04 Daniel Axtens <[email protected]>
13963
13964 mm: When adding a region, merge with region after as well as before
13965 On x86_64-efi (at least) regions seem to be added from top down. The mm
13966 code will merge a new region with an existing region that comes
13967 immediately before the new region. This allows larger allocations to be
13968 satisfied that would otherwise be the case.
13969
13970 On powerpc-ieee1275, however, regions are added from bottom up. So if
13971 we add 3x 32MB regions, we can still only satisfy a 32MB allocation,
13972 rather than the 96MB allocation we might otherwise be able to satisfy.
13973
13974 * Define 'post_size' as being bytes lost to the end of an allocation
13975 due to being given weird sizes from firmware that are not multiples
13976 of GRUB_MM_ALIGN.
13977
13978 * Allow merging of regions immediately _after_ existing regions, not
13979 just before. As with the other approach, we create an allocated
13980 block to represent the new space and the pass it to grub_free() to
13981 get the metadata right.
13982
13983 Tested-by: Stefan Berger <[email protected]>
13984 Reviewed-by: Daniel Kiper <[email protected]>
13985 Tested-by: Patrick Steinhardt <[email protected]>
13986
13987 2022-06-29 Daniel Axtens <[email protected]>
13988
13989 mm: Assert that we preserve header vs region alignment
13990 grub_mm_region_init() does:
13991
13992 h = (grub_mm_header_t) (r + 1);
13993
13994 where h is a grub_mm_header_t and r is a grub_mm_region_t.
13995
13996 Cells are supposed to be GRUB_MM_ALIGN aligned, but while grub_mm_dump
13997 ensures this vs the region header, grub_mm_region_init() does not.
13998
13999 It's better to be explicit than implicit here: rather than changing
14000 grub_mm_region_init() to ALIGN_UP(), require that the struct is
14001 explicitly a multiple of the header size.
14002
14003 Reviewed-by: Daniel Kiper <[email protected]>
14004 Tested-by: Patrick Steinhardt <[email protected]>
14005
14006 2022-06-28 Daniel Axtens <[email protected]>
14007
14008 tests: Only pass SeaBIOS fw_opt for x86 non-EFI platforms
14009 This breaks the tests on pseries - just restrict it to x86 platforms
14010 that don't specify an EFI.
14011
14012 Reviewed-by: Daniel Kiper <[email protected]>
14013
14014 2022-06-07 Darren Kenny <[email protected]>
14015
14016 fs/btrfs: Fix more fuzz issues related to chunks
14017 The corpus was generating issues in grub_btrfs_read_logical() when
14018 attempting to iterate over stripe entries in the superblock's
14019 bootmapping.
14020
14021 In most cases the reason for the failure was that the number of stripes
14022 in chunk->nstripes exceeded the possible space statically allocated in
14023 superblock bootmapping space. Each stripe entry in the bootmapping block
14024 consists of a grub_btrfs_key followed by a grub_btrfs_chunk_stripe.
14025
14026 Another issue that came up was that while calculating the chunk size,
14027 in an earlier piece of code in that function, depending on the data
14028 provided in the btrfs file system, it would end up calculating a size
14029 that was too small to contain even 1 grub_btrfs_chunk_item, which is
14030 obviously invalid too.
14031
14032 Reviewed-by: Daniel Kiper <[email protected]>
14033
14034 2022-06-07 Darren Kenny <[email protected]>
14035
14036 fs/btrfs: Fix more ASAN and SEGV issues found with fuzzing
14037 The fuzzer is generating btrfs file systems that have chunks with
14038 invalid combinations of stripes and substripes for the given RAID
14039 configurations.
14040
14041 After examining the Linux kernel fs/btrfs/tree-checker.c code, it
14042 appears that sub-stripes should only be applied to RAID10, and in that
14043 case there should only ever be 2 of them.
14044
14045 Similarly, RAID single should only have 1 stripe, and RAID1/1C3/1C4
14046 should have 2. 3 or 4 stripes respectively, which is what redundancy
14047 corresponds.
14048
14049 Some of the chunks ended up with a size of 0, which grub_malloc() still
14050 returned memory for and in turn generated ASAN errors later when
14051 accessed.
14052
14053 While it would be possible to specifically limit the number of stripes,
14054 a more correct test was on the combination of the chunk item, and the
14055 number of stripes by the size of the chunk stripe structure in
14056 comparison to the size of the chunk itself.
14057
14058 Reviewed-by: Daniel Kiper <[email protected]>
14059
14060 2022-06-07 Darren Kenny <[email protected]>
14061
14062 fs/btrfs: Fix several fuzz issues with invalid dir item sizing
14063 According to the btrfs code in Linux, the structure of a directory item
14064 leaf should be of the form:
14065
14066 |struct btrfs_dir_item|name|data|
14067
14068 in GRUB the name len and data len are in the grub_btrfs_dir_item
14069 structure's n and m fields respectively.
14070
14071 The combined size of the structure, name and data should be less than
14072 the allocated memory, a difference to the Linux kernel's struct
14073 btrfs_dir_item is that the grub_btrfs_dir_item has an extra field for
14074 where the name is stored, so we adjust for that too.
14075
14076 Reviewed-by: Daniel Kiper <[email protected]>
14077
14078 2022-06-07 Sudhakar Kuppusamy <[email protected]>
14079
14080 fs/f2fs: Do not copy file names that are too long
14081 A corrupt f2fs file system might specify a name length which is greater
14082 than the maximum name length supported by the GRUB f2fs driver.
14083
14084 We will allocate enough memory to store the overly long name, but there
14085 are only F2FS_NAME_LEN bytes in the source, so we would read past the end
14086 of the source.
14087
14088 While checking directory entries, do not copy a file name with an invalid
14089 length.
14090
14091 Reviewed-by: Daniel Kiper <[email protected]>
14092
14093 2022-06-07 Sudhakar Kuppusamy <[email protected]>
14094
14095 fs/f2fs: Do not read past the end of nat bitmap
14096 A corrupt f2fs filesystem could have a block offset or a bitmap
14097 offset that would cause us to read beyond the bounds of the nat
14098 bitmap.
14099
14100 Introduce the nat_bitmap_size member in grub_f2fs_data which holds
14101 the size of nat bitmap.
14102
14103 Set the size when loading the nat bitmap in nat_bitmap_ptr(), and
14104 catch when an invalid offset would create a pointer past the end of
14105 the allocated space.
14106
14107 Check against the bitmap size in grub_f2fs_test_bit() test bit to avoid
14108 reading past the end of the nat bitmap.
14109
14110 Reviewed-by: Daniel Kiper <[email protected]>
14111
14112 2022-06-07 Sudhakar Kuppusamy <[email protected]>
14113
14114 fs/f2fs: Do not read past the end of nat journal entries
14115 A corrupt f2fs file system could specify a nat journal entry count
14116 that is beyond the maximum NAT_JOURNAL_ENTRIES.
14117
14118 Check if the specified nat journal entry count before accessing the
14119 array, and throw an error if it is too large.
14120
14121 Reviewed-by: Daniel Kiper <[email protected]>
14122
14123 2022-06-07 Daniel Axtens <[email protected]>
14124
14125 net/http: Error out on headers with LF without CR
14126 In a similar vein to the previous patch, parse_line() would write
14127 a NUL byte past the end of the buffer if there was an HTTP header
14128 with a LF rather than a CRLF.
14129
14130 RFC-2616 says:
14131
14132 Many HTTP/1.1 header field values consist of words separated by LWS
14133 or special characters. These special characters MUST be in a quoted
14134 string to be used within a parameter value (as defined in section 3.6).
14135
14136 We don't support quoted sections or continuation lines, etc.
14137
14138 If we see an LF that's not part of a CRLF, bail out.
14139
14140 Fixes: CVE-2022-28734
14141
14142 Reviewed-by: Daniel Kiper <[email protected]>
14143
14144 2022-06-07 Daniel Axtens <[email protected]>
14145
14146 net/http: Fix OOB write for split http headers
14147 GRUB has special code for handling an http header that is split
14148 across two packets.
14149
14150 The code tracks the end of line by looking for a "\n" byte. The
14151 code for split headers has always advanced the pointer just past the
14152 end of the line, whereas the code that handles unsplit headers does
14153 not advance the pointer. This extra advance causes the length to be
14154 one greater, which breaks an assumption in parse_line(), leading to
14155 it writing a NUL byte one byte past the end of the buffer where we
14156 reconstruct the line from the two packets.
14157
14158 It's conceivable that an attacker controlled set of packets could
14159 cause this to zero out the first byte of the "next" pointer of the
14160 grub_mm_region structure following the current_line buffer.
14161
14162 Do not advance the pointer in the split header case.
14163
14164 Fixes: CVE-2022-28734
14165
14166 Reviewed-by: Daniel Kiper <[email protected]>
14167
14168 2022-06-07 Daniel Axtens <[email protected]>
14169
14170 net/http: Do not tear down socket if it's already been torn down
14171 It's possible for data->sock to get torn down in tcp error handling.
14172 If we unconditionally tear it down again we will end up doing writes
14173 to an offset of the NULL pointer when we go to tear it down again.
14174
14175 Detect if it has been torn down and don't do it again.
14176
14177 Reviewed-by: Daniel Kiper <[email protected]>
14178
14179 2022-06-07 Daniel Axtens <[email protected]>
14180
14181 net/tftp: Avoid a trivial UAF
14182 Under tftp errors, we print a tftp error message from the tftp header.
14183 However, the tftph pointer is a pointer inside nb, the netbuff. Previously,
14184 we were freeing the nb and then dereferencing it. Don't do that, use it
14185 and then free it later.
14186
14187 This isn't really _bad_ per se, especially as we're single-threaded, but
14188 it trips up fuzzers.
14189
14190 Reviewed-by: Daniel Kiper <[email protected]>
14191
14192 2022-06-07 Daniel Axtens <[email protected]>
14193
14194 net/tftp: Prevent a UAF and double-free from a failed seek
14195 A malicious tftp server can cause UAFs and a double free.
14196
14197 An attempt to read from a network file is handled by grub_net_fs_read(). If
14198 the read is at an offset other than the current offset, grub_net_seek_real()
14199 is invoked.
14200
14201 In grub_net_seek_real(), if a backwards seek cannot be satisfied from the
14202 currently received packets, and the underlying transport does not provide
14203 a seek method, then grub_net_seek_real() will close and reopen the network
14204 protocol layer.
14205
14206 For tftp, the ->close() call goes to tftp_close() and frees the tftp_data_t
14207 file->data. The file->data pointer is not nulled out after the free.
14208
14209 If the ->open() call fails, the file->data will not be reallocated and will
14210 continue point to a freed memory block. This could happen from a server
14211 refusing to send the requisite ack to the new tftp request, for example.
14212
14213 The seek and the read will then fail, but the grub_file continues to exist:
14214 the failed seek does not necessarily cause the entire file to be thrown
14215 away (e.g. where the file is checked to see if it is gzipped/lzio/xz/etc.,
14216 a read failure is interpreted as a decompressor passing on the file, not as
14217 an invalidation of the entire grub_file_t structure).
14218
14219 This means subsequent attempts to read or seek the file will use the old
14220 file->data after free. Eventually, the file will be close()d again and
14221 file->data will be freed again.
14222
14223 Mark a net_fs file that doesn't reopen as broken. Do not permit read() or
14224 close() on a broken file (seek is not exposed directly to the file API -
14225 it is only called as part of read, so this blocks seeks as well).
14226
14227 As an additional defence, null out the ->data pointer if tftp_open() fails.
14228 That would have lead to a simple null pointer dereference rather than
14229 a mess of UAFs.
14230
14231 This may affect other protocols, I haven't checked.
14232
14233 Reviewed-by: Daniel Kiper <[email protected]>
14234
14235 2022-06-07 Daniel Axtens <[email protected]>
14236
14237 net/dns: Don't read past the end of the string we're checking against
14238 I don't really understand what's going on here but fuzzing found
14239 a bug where we read past the end of check_with. That's a C string,
14240 so use grub_strlen() to make sure we don't overread it.
14241
14242 Reviewed-by: Daniel Kiper <[email protected]>
14243
14244 2022-06-07 Daniel Axtens <[email protected]>
14245
14246 net/dns: Fix double-free addresses on corrupt DNS response
14247 grub_net_dns_lookup() takes as inputs a pointer to an array of addresses
14248 ("addresses") for the given name, and pointer to a number of addresses
14249 ("naddresses"). grub_net_dns_lookup() is responsible for allocating
14250 "addresses", and the caller is responsible for freeing it if
14251 "naddresses" > 0.
14252
14253 The DNS recv_hook will sometimes set and free the addresses array,
14254 for example if the packet is too short:
14255
14256 if (ptr + 10 >= nb->tail)
14257 {
14258 if (!*data->naddresses)
14259 grub_free (*data->addresses);
14260 grub_netbuff_free (nb);
14261 return GRUB_ERR_NONE;
14262 }
14263
14264 Later on the nslookup command code unconditionally frees the "addresses"
14265 array. Normally this is fine: the array is either populated with valid
14266 data or is NULL. But in these sorts of error cases it is neither NULL
14267 nor valid and we get a double-free.
14268
14269 Only free "addresses" if "naddresses" > 0.
14270
14271 It looks like the other use of grub_net_dns_lookup() is not affected.
14272
14273 Reviewed-by: Daniel Kiper <[email protected]>
14274
14275 2022-06-07 Daniel Axtens <[email protected]>
14276
14277 net/netbuff: Block overly large netbuff allocs
14278 A netbuff shouldn't be too huge. It's bounded by MTU and TCP segment
14279 reassembly. If we are asked to create one that is unreasonably big, refuse.
14280
14281 This is a hardening measure: if we hit this code, there's a bug somewhere
14282 else that we should catch and fix.
14283
14284 This commit:
14285 - stops the bug propagating any further.
14286 - provides a spot to instrument in e.g. fuzzing to try to catch these bugs.
14287
14288 I have put instrumentation (e.g. __builtin_trap() to force a crash) here and
14289 have not been able to find any more crashes.
14290
14291 Reviewed-by: Daniel Kiper <[email protected]>
14292
14293 2022-06-07 Daniel Axtens <[email protected]>
14294
14295 net/ip: Do IP fragment maths safely
14296 We can receive packets with invalid IP fragmentation information. This
14297 can lead to rsm->total_len underflowing and becoming very large.
14298
14299 Then, in grub_netbuff_alloc(), we add to this very large number, which can
14300 cause it to overflow and wrap back around to a small positive number.
14301 The allocation then succeeds, but the resulting buffer is too small and
14302 subsequent operations can write past the end of the buffer.
14303
14304 Catch the underflow here.
14305
14306 Fixes: CVE-2022-28733
14307
14308 Reviewed-by: Daniel Kiper <[email protected]>
14309
14310 2022-06-07 Daniel Axtens <[email protected]>
14311
14312 normal/charset: Fix array out-of-bounds formatting unicode for display
14313 In some cases attempting to display arbitrary binary strings leads
14314 to ASAN splats reading the widthspec array out of bounds.
14315
14316 Check the index. If it would be out of bounds, return a width of 1.
14317 I don't know if that's strictly correct, but we're not really expecting
14318 great display of arbitrary binary data, and it's certainly not worse than
14319 an OOB read.
14320
14321 Reviewed-by: Daniel Kiper <[email protected]>
14322
14323 2022-06-07 Daniel Axtens <[email protected]>
14324
14325 video/readers/jpeg: Block int underflow -> wild pointer write
14326 Certain 1 px wide images caused a wild pointer write in
14327 grub_jpeg_ycrcb_to_rgb(). This was caused because in grub_jpeg_decode_data(),
14328 we have the following loop:
14329
14330 for (; data->r1 < nr1 && (!data->dri || rst);
14331 data->r1++, data->bitmap_ptr += (vb * data->image_width - hb * nc1) * 3)
14332
14333 We did not check if vb * width >= hb * nc1.
14334
14335 On a 64-bit platform, if that turns out to be negative, it will underflow,
14336 be interpreted as unsigned 64-bit, then be added to the 64-bit pointer, so
14337 we see data->bitmap_ptr jump, e.g.:
14338
14339 0x6180_0000_0480 to
14340 0x6181_0000_0498
14341 ^
14342 ~--- carry has occurred and this pointer is now far away from
14343 any object.
14344
14345 On a 32-bit platform, it will decrement the pointer, creating a pointer
14346 that won't crash but will overwrite random data.
14347
14348 Catch the underflow and error out.
14349
14350 Fixes: CVE-2021-3697
14351
14352 Reviewed-by: Daniel Kiper <[email protected]>
14353
14354 2022-06-07 Daniel Axtens <[email protected]>
14355
14356 video/readers/jpeg: Refuse to handle multiple start of streams
14357 An invalid file could contain multiple start of stream blocks, which
14358 would cause us to reallocate and leak our bitmap. Refuse to handle
14359 multiple start of streams.
14360
14361 Additionally, fix a grub_error() call formatting.
14362
14363 Reviewed-by: Daniel Kiper <[email protected]>
14364
14365 2022-06-07 Daniel Axtens <[email protected]>
14366
14367 video/readers/jpeg: Do not reallocate a given huff table
14368 Fix a memory leak where an invalid file could cause us to reallocate
14369 memory for a huffman table we had already allocated memory for.
14370
14371 Reviewed-by: Daniel Kiper <[email protected]>
14372
14373 2022-06-07 Daniel Axtens <[email protected]>
14374
14375 video/readers/jpeg: Abort sooner if a read operation fails
14376 Fuzzing revealed some inputs that were taking a long time, potentially
14377 forever, because they did not bail quickly upon encountering an I/O error.
14378
14379 Try to catch I/O errors sooner and bail out.
14380
14381 Reviewed-by: Daniel Kiper <[email protected]>
14382
14383 2022-06-07 Daniel Axtens <[email protected]>
14384
14385 video/readers/png: Sanity check some huffman codes
14386 ASAN picked up two OOB global reads: we weren't checking if some code
14387 values fit within the cplens or cpdext arrays. Check and throw an error
14388 if not.
14389
14390 Reviewed-by: Daniel Kiper <[email protected]>
14391
14392 2022-06-07 Daniel Axtens <[email protected]>
14393
14394 video/readers/png: Avoid heap OOB R/W inserting huff table items
14395 In fuzzing we observed crashes where a code would attempt to be inserted
14396 into a huffman table before the start, leading to a set of heap OOB reads
14397 and writes as table entries with negative indices were shifted around and
14398 the new code written in.
14399
14400 Catch the case where we would underflow the array and bail.
14401
14402 Fixes: CVE-2021-3696
14403
14404 Reviewed-by: Daniel Kiper <[email protected]>
14405
14406 2022-06-07 Daniel Axtens <[email protected]>
14407
14408 video/readers/png: Drop greyscale support to fix heap out-of-bounds write
14409 A 16-bit greyscale PNG without alpha is processed in the following loop:
14410
14411 for (i = 0; i < (data->image_width * data->image_height);
14412 i++, d1 += 4, d2 += 2)
14413 {
14414 d1[R3] = d2[1];
14415 d1[G3] = d2[1];
14416 d1[B3] = d2[1];
14417 }
14418
14419 The increment of d1 is wrong. d1 is incremented by 4 bytes per iteration,
14420 but there are only 3 bytes allocated for storage. This means that image
14421 data will overwrite somewhat-attacker-controlled parts of memory - 3 bytes
14422 out of every 4 following the end of the image.
14423
14424 This has existed since greyscale support was added in 2013 in commit
14425 3ccf16dff98f (grub-core/video/readers/png.c: Support grayscale).
14426
14427 Saving starfield.png as a 16-bit greyscale image without alpha in the gimp
14428 and attempting to load it causes grub-emu to crash - I don't think this code
14429 has ever worked.
14430
14431 Delete all PNG greyscale support.
14432
14433 Fixes: CVE-2021-3695
14434
14435 Reviewed-by: Daniel Kiper <[email protected]>
14436
14437 2022-06-07 Daniel Axtens <[email protected]>
14438
14439 video/readers/png: Refuse to handle multiple image headers
14440 This causes the bitmap to be leaked. Do not permit multiple image headers.
14441
14442 Reviewed-by: Daniel Kiper <[email protected]>
14443
14444 2022-06-07 Daniel Axtens <[email protected]>
14445
14446 video/readers/png: Abort sooner if a read operation fails
14447 Fuzzing revealed some inputs that were taking a long time, potentially
14448 forever, because they did not bail quickly upon encountering an I/O error.
14449
14450 Try to catch I/O errors sooner and bail out.
14451
14452 Reviewed-by: Daniel Kiper <[email protected]>
14453
14454 2022-06-07 Daniel Axtens <[email protected]>
14455
14456 kern/file: Do not leak device_name on error in grub_file_open()
14457 If we have an error in grub_file_open() before we free device_name, we
14458 will leak it.
14459
14460 Free device_name in the error path and null out the pointer in the good
14461 path once we free it there.
14462
14463 Reviewed-by: Daniel Kiper <[email protected]>
14464
14465 2022-06-07 Julian Andres Klode <[email protected]>
14466
14467 kern/efi/sb: Reject non-kernel files in the shim_lock verifier
14468 We must not allow other verifiers to pass things like the GRUB modules.
14469 Instead of maintaining a blocklist, maintain an allowlist of things
14470 that we do not care about.
14471
14472 This allowlist really should be made reusable, and shared by the
14473 lockdown verifier, but this is the minimal patch addressing
14474 security concerns where the TPM verifier was able to mark modules
14475 as verified (or the OpenPGP verifier for that matter), when it
14476 should not do so on shim-powered secure boot systems.
14477
14478 Fixes: CVE-2022-28735
14479
14480 Reviewed-by: Daniel Kiper <[email protected]>
14481
14482 2022-06-07 Chris Coulson <[email protected]>
14483
14484 loader/efi/chainloader: Use grub_loader_set_ex()
14485 This ports the EFI chainloader to use grub_loader_set_ex() in order to fix
14486 a use-after-free bug that occurs when grub_cmd_chainloader() is executed
14487 more than once before a boot attempt is performed.
14488
14489 Fixes: CVE-2022-28736
14490
14491 Reviewed-by: Daniel Kiper <[email protected]>
14492
14493 2022-06-07 Chris Coulson <[email protected]>
14494
14495 commands/boot: Add API to pass context to loader
14496 Loaders rely on global variables for saving context which is consumed
14497 in the boot hook and freed in the unload hook. In the case where a loader
14498 command is executed twice, calling grub_loader_set() a second time executes
14499 the unload hook, but in some cases this runs when the loader's global
14500 context has already been updated, resulting in the updated context being
14501 freed and potential use-after-free bugs when the boot hook is subsequently
14502 called.
14503
14504 This adds a new API, grub_loader_set_ex(), which allows a loader to specify
14505 context that is passed to its boot and unload hooks. This is an alternative
14506 to requiring that loaders call grub_loader_unset() before mutating their
14507 global context.
14508
14509 Reviewed-by: Daniel Kiper <[email protected]>
14510
14511 2022-06-07 Chris Coulson <[email protected]>
14512
14513 loader/efi/chainloader: Simplify the loader state
14514 The chainloader command retains the source buffer and device path passed
14515 to LoadImage(), requiring the unload hook passed to grub_loader_set() to
14516 free them. It isn't required to retain this state though - they aren't
14517 required by StartImage() or anything else in the boot hook, so clean them
14518 up before grub_cmd_chainloader() finishes.
14519
14520 Reviewed-by: Daniel Kiper <[email protected]>
14521
14522 2022-06-07 Jagannathan Raman <[email protected]>
14523
14524 fs/zfs/zfs: zfs_mount() - avoid pointer downcasting
14525 Coverity reports that while loopis in the following functions uses
14526 tainted data as boundary:
14527 zfs_mount() -> check_mos_features() -> dnode_get() -> zfs_log2()
14528 zfs_mount() -> grub_memmove()
14529
14530 The defect type is "Untrusted loop bound" caused as a result of
14531 "tainted_data_downcast". Coverity does not like the pointer downcast
14532 here and we need to address it.
14533
14534 We believe Coverity flags pointer downcast for the following two
14535 reasons:
14536 1. External data: The pointer downcast could indicate that the source is
14537 external data, which we need to further sanitize - such as verifying its
14538 limits. In this case, the data is read from an external source, which is
14539 a disk. But, zio_read(), which reads the data from the disk, sanitizes it
14540 using a checksum. checksum is the best facility that ZFS offers to verify
14541 external data, and we don't believe a better way exists. Therefore, no
14542 further action is possible for this.
14543
14544 2. Corruption due to alignment: downcasting a pointer from a strict type
14545 to less strict type could result in data corruption. For example, the
14546 following cast would corrupt because uint32_t is 4-byte aligned, and
14547 won't be able to point to 0x1003 which is not 4-byte aligned.
14548 uint8_t *ptr = 0x1003;
14549 uint32_t *word = ptr; (incorrect, alignment issues)
14550
14551 This patch converts the "osp" pointer in zfs_mount() from a "void" type
14552 to "objset_phys_t" type to address this issue.
14553
14554 We are not sure if there are any other reasons why Coverity flags the
14555 downcast. However, the fix for alignment issue masks/suppresses any
14556 other issues from showing up.
14557
14558 Fixes: CID 314023
14559
14560 Reviewed-by: Darren Kenny <[email protected]>
14561 Reviewed-by: Daniel Kiper <[email protected]>
14562
14563 2022-06-07 Jagannathan Raman <[email protected]>
14564
14565 fs/zfs/zfs: make_mdn() - avoid pointer downcasting
14566 Coverity reports that the while loop in the following function uses
14567 tainted data as boundary:
14568 fill_fs_info() -> dnode_get() -> zfs_log2()
14569
14570 The tainted originated from:
14571 fill_fs_info() -> make_mdn()
14572
14573 The defect type is "Untrusted loop bound" caused as a result of
14574 "tainted_data_downcast". Coverity does not like the pointer downcast
14575 here and we need to address it.
14576
14577 We believe Coverity flags pointer downcast for the following two
14578 reasons:
14579 1. External data: The pointer downcast could indicate that the source is
14580 external data, which we need to further sanitize - such as verifying its
14581 limits. In this case, the data is read from an external source, which is
14582 a disk. But, zio_read(), which reads the data from the disk, sanitizes it
14583 using a checksum. checksum is the best facility that ZFS offers to verify
14584 external data, and we don't believe a better way exists. Therefore, no
14585 further action is possible for this.
14586
14587 2. Corruption due to alignment: downcasting a pointer from a strict type
14588 to less strict type could result in data corruption. For example, the
14589 following cast would corrupt because uint32_t is 4-byte aligned, and
14590 won't be able to point to 0x1003 which is not 4-byte aligned.
14591 uint8_t *ptr = 0x1003;
14592 uint32_t *word = ptr; (incorrect, alignment issues)
14593
14594 This patch converts the "osp" pointer in make_mdn() from a "void" type
14595 to "objset_phys_t" type to address the issue.
14596
14597 We are not sure if there are any other reasons why Coverity flags the
14598 downcast. However, the fix for alignment issue masks/suppresses any
14599 other issues from showing up.
14600
14601 Fixes: CID 314020
14602
14603 Reviewed-by: Darren Kenny <[email protected]>
14604 Reviewed-by: Daniel Kiper <[email protected]>
14605
14606 2022-06-07 Alec Brown <[email protected]>
14607
14608 util/grub-module-verifierXX: Add e_shoff check in get_shdr()
14609 In util/grub-module-verifierXX.c, the function get_shdr() is used to obtain the
14610 section header at a given index but isn't checking that there is an offset for
14611 the section header table. To validate that there is, we can check that e_shoff
14612 isn't 0.
14613
14614 Reviewed-by: Daniel Kiper <[email protected]>
14615 Reviewed-by: Darren Kenny <[email protected]>
14616
14617 2022-06-07 Alec Brown <[email protected]>
14618
14619 grub-core/loader/i386/bsdXX: Avoid downcasting (char *) to (Elf_Shdr *)
14620 In bsdXX.c, a couple of untrusted loop bound and untrusted allocation size bugs
14621 were flagged by Coverity in the functions grub_openbsd_find_ramdisk() and
14622 grub_freebsd_load_elfmodule(). These bugs were flagged by coverity because the
14623 variable shdr was downcasting from a char pointer to an Elf_Shdr pointer
14624 whenever it was used to set the base value in for loops. To avoid this, we need
14625 to set shdr as an Elf_Shdr pointer where it is initialized.
14626
14627 In the function read_headers(), the function is reading elf section header data
14628 from a file and passing it to the variable shdr as data for a char pointer. If
14629 we switch the type of shdr to an Elf_Shdr pointer in read_headers() as well as
14630 other functions, then we won't need to downcast to an Elf_Shdr pointer. By doing
14631 this, the issue becomes masked from Coverity's view. In the following patches,
14632 we check limits to ensure the data isn't tainted.
14633
14634 Also, switched use of (char *) to (grub_uint8_t *) to give a better indication
14635 of pointer arithmetic and not suggest use of a C string.
14636
14637 Fixes: CID 314018
14638 Fixes: CID 314030
14639 Fixes: CID 314031
14640 Fixes: CID 314039
14641
14642 Reviewed-by: Daniel Kiper <[email protected]>
14643 Reviewed-by: Darren Kenny <[email protected]>
14644
14645 2022-06-07 Stefan Agner <[email protected]>
14646
14647 disk/efi/efidisk: Pass buffers with higher alignment
14648 Some devices report IoAlign values but seem to require buffers with
14649 higher alignment.
14650
14651 The UEFI specification is saying: "IoAlign values of 0 and 1 mean that
14652 the buffer can be placed anywhere in memory. Otherwise, IoAlign must
14653 be a power of 2, and the requirement is that the start address of
14654 a buffer must be evenly divisible by IoAlign with no remainder."
14655
14656 Some devices report IoAlign of 2, however seem to require 4 bytes
14657 aligned buffers. It seems that this got misinterpreted by some vendors
14658 assuming IoAlign is 2^IoAlign. There is also such a hint in an example
14659 in earlier versions of the Driver Writer's Guide:
14660
14661 ScsiPassThruMode.IoAlign = 2; // Data must be alligned on 4-byte boundary
14662
14663 Some devices report no alignment requirements at all but seem to read
14664 corrupted data or report read errors when passing unaligned buffers.
14665
14666 Work around by using an alignment of at least BlockSize (typically 512
14667 bytes) in any case. If IoAlign (interpreted as per UEFI specification)
14668 requests a higher alignment than BlockSize, follow IoAlign still.
14669
14670 Note: The problem has only noticed with compressed squashfs. It seems
14671 that ext4 (and presumably other file system drivers) pass buffers with
14672 a higher alignment already.
14673
14674 Acked-by: Heinrich Schuchardt <[email protected]>
14675 Reviewed-by: Daniel Kiper <[email protected]>
14676
14677 2022-06-07 Samuel Thibault <[email protected]>
14678
14679 osdep/hurd/getroot: Use "part:" qualifier
14680 When using userland drivers such as rumpdisk, we'd rather make ext2fs use
14681 parted-based libstore partitioning support. That can be used for kernelland
14682 drivers as well, so we can just make GRUB always use the "part:" qualifier
14683 to switch ext2fs to it.
14684
14685 grub_util_find_hurd_root_device() then has to understand this syntax and
14686 translate it into the /dev/ entry name.
14687
14688 Reviewed-by: Daniel Kiper <[email protected]>
14689
14690 2022-06-07 Glenn Washburn <[email protected]>
14691
14692 docs: Add documentation on keyfile option to cryptomount
14693 Reviewed-by: Daniel Kiper <[email protected]>
14694
14695 disk/cryptodisk: Use enum constants as indexes into cryptomount option array
14696 Reviewed-by: Daniel Kiper <[email protected]>
14697
14698 2022-06-07 John Lane <[email protected]>
14699
14700 disk/cryptodisk: Add options to cryptomount to support keyfiles
14701 Add the options --key-file, --keyfile-offset, and --keyfile-size to
14702 cryptomount and code to put read the requested key file data and pass
14703 via the cargs struct. Note, key file data is for all intents and purposes
14704 equivalent to a password given to cryptomount. So there is no need to
14705 enable support for key files in the various crypto backends (e.g. LUKS1)
14706 because the key data is passed just as if it were a password.
14707
14708 Reviewed-by: Daniel Kiper <[email protected]>
14709
14710 2022-06-07 Denis 'GNUtoo' Carikli <[email protected]>
14711
14712 disk/geli: Unify grub_cryptodisk_dev function names
14713 Reviewed-by: Patrick Steinhardt <[email protected]>
14714 Reviewed-by: Daniel Kiper <[email protected]>
14715
14716 disk/luks: Unify grub_cryptodisk_dev function names
14717 Reviewed-by: Patrick Steinhardt <[email protected]>
14718 Reviewed-by: Daniel Kiper <[email protected]>
14719
14720 2022-06-07 Glenn Washburn <[email protected]>
14721
14722 util/probe: Remove unused header includes
14723 Reviewed-by: Daniel Kiper <[email protected]>
14724
14725 commands/macbless: Remove whitespace between N_ macro and open parenthesis
14726 Reviewed-by: Daniel Kiper <[email protected]>
14727
14728 2022-06-07 Glenn Washburn <[email protected]>
14729
14730 tests: Add /sbin and /usr/sbin to path in partmap test
14731 The partmap test requires no elevated privileges. However, it uses parted
14732 which can be used as a normal user, but is usually located in /sbin or
14733 /usr/bin (eg. on Debian systems). Whereas the normal user does not usually
14734 have /sbin or /usr/sbin added to their path, thus parted will not be found
14735 causing the test to abort. Add /sbin and /usr/sbin to the path for the
14736 partmap test so that the test can run successfully as an unprivileged user.
14737
14738 Reviewed-by: Daniel Kiper <[email protected]>
14739
14740 2022-06-07 Glenn Washburn <[email protected]>
14741
14742 tests: Show host determined fs UUID when hfs UUID test fails
14743 On failure, the hfs test should show both the host and GRUB determined fs
14744 UUID. Prior to this change, both outputs where generated by GRUB, which is
14745 less helpful in determining the cause of failure.
14746
14747 Reviewed-by: Daniel Kiper <[email protected]>
14748
14749 2022-05-24 Glenn Washburn <[email protected]>
14750
14751 docs: Add section for general undocumented commands
14752 The section is an itemized list of commands that are not listed else where
14753 in the command sections.
14754
14755 Reviewed-by: Daniel Kiper <[email protected]>
14756
14757 2022-05-24 Glenn Washburn <[email protected]>
14758
14759 docs: Add under documented loader commands to beginning of loader section
14760 Reviewed-by: Daniel Kiper <[email protected]>
14761
14762 2022-05-24 Glenn Washburn <[email protected]>
14763
14764 docs: Create command section for loader commands
14765 Move loader commands documented in the general commands list into the
14766 loader command section.
14767
14768 Reviewed-by: Daniel Kiper <[email protected]>
14769
14770 2022-05-24 Glenn Washburn <[email protected]>
14771
14772 docs: Markup loader commands with @command tag
14773 Also, add period to terminate sentence.
14774
14775 Reviewed-by: Daniel Kiper <[email protected]>
14776
14777 2022-05-24 Glenn Washburn <[email protected]>
14778
14779 docs: Make note of i386-pc specific usage of halt command
14780 The --no-apm option is only available on the i396-pc target.
14781
14782 Reviewed-by: Daniel Kiper <[email protected]>
14783
14784 2022-05-24 Glenn Washburn <[email protected]>
14785
14786 docs: Make note that sendkey is only available on i386-pc
14787 Reviewed-by: Daniel Kiper <[email protected]>
14788
14789 docs: Fix spelling typo and remove unnecessary spaces
14790 Reviewed-by: Daniel Kiper <[email protected]>
14791
14792 2022-05-24 Glenn Washburn <[email protected]>
14793
14794 net/net: Fix incorrect condition for calling grub_net_tcp_retransmit()
14795 The commit 848724273e4 (net/net: Avoid unnecessary calls to
14796 grub_net_tcp_retransmit()) needs to have its condition inverted to avoid
14797 unnecessary calls to grub_net_tcp_retransmit(). As it is, it creates many
14798 unnecessary calls and does not call grub_net_tcp_retransmit() when needed.
14799 The call to grub_net_tcp_retransmit() should only be made when
14800 grub_net_cards does _not_ equal NULL, meaning that there are potentially
14801 network cards that need TCP retransmission.
14802
14803 Fixes: 848724273e4 (net/net: Avoid unnecessary calls to grub_net_tcp_retransmit())
14804
14805 Reviewed-by: Daniel Kiper <[email protected]>
14806
14807 2022-05-24 Oskari Pirhonen <[email protected]>
14808
14809 templates: Improve initramfs detection
14810 Add detection for initramfs of the form *.img.old. For example, Gentoo's
14811 sys-kernel/genkernel installs it as initramfs-*.img and moves any existing
14812 one to initramfs-*.img.old.
14813
14814 Apply the same scheme to initrd-*.img and initrd-*.gz files for consistency.
14815
14816 Reviewed-by: Daniel Kiper <[email protected]>
14817
14818 2022-05-24 Samuel Thibault <[email protected]>
14819
14820 osdep/hurd: Support device entries with @/dev/disk: qualifier
14821 Those are used with non-bootstrap disk drivers, for which libstore has to
14822 open /dev/disk before calling device_open on it instead of on the device
14823 master port. Normally in that case all /dev/ entries also have the @/dev/disk:
14824 qualifier, so we can just drop it.
14825
14826 Reviewed-by: Daniel Kiper <[email protected]>
14827
14828 2022-05-24 Darren Kenny <[email protected]>
14829
14830 grub-mkimage: Creating aarch64 images from x86 host is broken
14831 A recent fix that made appears to have broken the ability to create an
14832 aarch64 boot image on a x86-based host.
14833
14834 This was due to an overzealous testing of the architecture when building
14835 grub-mkimage and removing the code that build an ARM image when not built
14836 on ARM.
14837
14838 On the occasion remove redundant break.
14839
14840 Fixes: 8541f319 (grub-mkimage: Only check aarch64 relocations when built for aarch64)
14841
14842 Tested-by: Selva Ganesan <[email protected]>
14843 Reviewed-by: Daniel Kiper <[email protected]>
14844
14845 2022-05-24 Icenowy Zheng <[email protected]>
14846
14847 grub-install: Allow to install to non-EFI ESP when --force
14848 Although the EFI specification enforces support for FAT ESP, it's free
14849 for EFI implementations to implement support for ESPs with other formats
14850 (e.g. ext4, ntfs, etc), and at least U-Boot EFI will support ext4 ESP if
14851 U-Boot is built with ext4 support. In some situations a GRUB installation
14852 on such a non-FAT ESP could be useful (e.g. a NTFS-based USB disk that
14853 can dual boot a Windows installation media and a Linux LiveCD).
14854
14855 As this is advanced and implementation-dependent behavior, let grub-install
14856 allow this kind of installation, but only when --force is specified.
14857
14858 Reviewed-by: Daniel Kiper <[email protected]>
14859
14860 2022-04-26 Qiumiao Zhang <[email protected]>
14861
14862 net: Fix NULL pointer dereference when parsing ICMP6_ROUTER_ADVERTISE messages
14863 During UEFI PXE boot in IPv6 network, if the DHCP server adopts stateful
14864 automatic configuration, then the client receives a ICMP6_ROUTER_ADVERTISE
14865 multicast message from the server. This may be received without the interface
14866 having a configured network address, so orig_inf will be NULL, which can lead
14867 to a NULL dereference when creating the default route. Actually, in this case,
14868 the client obtains the default route through DHCPv6 instead of RA messages.
14869 So if orig_inf == NULL and route_inf == NULL, we should not set the
14870 default route.
14871
14872 Fixes: https://savannah.gnu.org/bugs/?62072
14873
14874 Reviewed-by: Daniel Kiper <[email protected]>
14875
14876 2022-04-26 Glenn Washburn <[email protected]>
14877
14878 tests: Ensure that loopback devices and zfs devices are cleaned up
14879 ZFS file systems are not unmounted using umount, but instead by exporting
14880 them. So export the ZFS file system that has the same label as the one that
14881 was created during the test, if such one exists. This is required to delete
14882 the loopback device that uses the ZFS image file. Otherwise the added code
14883 to delete all loopback devices setup during the test run will never be able
14884 to finish because the loopback device can not be deleted while in use.
14885
14886 Reviewed-by: Daniel Kiper <[email protected]>
14887
14888 2022-04-26 Glenn Washburn <[email protected]>
14889
14890 tests: Ensure that mountpoints are unmounted before exiting
14891 When all tests complete successfully, filesystems mounted by grub-fs-tester
14892 will be unmounted before exiting. However, on certain test failures the
14893 tester will exit with a failure code and not unmount previously mounted
14894 filesystems. Now keep track of mounts and umounts and run an exit handler
14895 on exit or process interruption that will umount all mounts that haven't
14896 already been unmounted.
14897
14898 Reviewed-by: Daniel Kiper <[email protected]>
14899
14900 2022-04-20 Glenn Washburn <[email protected]>
14901
14902 docs: Use correct list format
14903 Using "*" to prefix list items leads to undesirable display output for
14904 at least the generation of the html documentation. Use the @itemize and
14905 @item directives to get itemized list output.
14906
14907 Also fix some wording and punctuation issues.
14908
14909 Reviewed-by: Daniel Kiper <[email protected]>
14910
14911 2022-04-20 Glenn Washburn <[email protected]>
14912
14913 docs: Clarify meaning of "list" and "cond" for "if" and "while" commands respectively
14914 It is not clear from the documentation what a "list" is in the context
14915 of the "if" command. Note that its a list of simple commands separated
14916 by a ";" and that only the exit status of the last command matters.
14917 The same is true for the "cond" parameter to the "while" command.
14918
14919 Reviewed-by: Daniel Kiper <[email protected]>
14920
14921 2022-04-20 Glenn Washburn <[email protected]>
14922
14923 docs: Add note that drivemap is only available on i386-pc
14924 Reviewed-by: Daniel Kiper <[email protected]>
14925
14926 2022-04-20 Glenn Washburn <[email protected]>
14927
14928 tests: Give grub-fs-tester temp directory a better name
14929 Instead of "tmp" the name is prefixed by the name of the scripts (e.g.
14930 grub-fs-tester). A timestamp is added in the name to allow for easily
14931 seeing a chronological sorting of runs and the name of the filesystem
14932 being tested. The random component is set to the minimal possible,
14933 3 characters, because the timestamp should provide enough uniqueness.
14934
14935 Reviewed-by: Daniel Kiper <[email protected]>
14936
14937 2022-04-20 Glenn Washburn <[email protected]>
14938
14939 tests: Disable blkid cache usage
14940 Using the blkid cache can cause issues when running many file system tests
14941 in parallel. We do not need it, as its only there to improve performance,
14942 and using the cache does not provide significant performance improvements.
14943
14944 Reviewed-by: Daniel Kiper <[email protected]>
14945
14946 2022-04-20 Glenn Washburn <[email protected]>
14947
14948 configure: Fix default -O2 being added when CFLAGS not set
14949 Autoconf will set a default CFLAGS of "-g -O2" if CFLAGS is not set.
14950 CFLAGS was defaulted to "" early in configure to prevent this. A recent
14951 commit ad9ccf660 (configure: Fix various new autotools warnings) added
14952 AC_USE_SYSTEM_EXTENSIONS, which pulls in the autoconf CFLAGS check,
14953 before we default CFLAGS and thus setting the autoconf default for
14954 CFLAGS. Move the default setting of CFLAGS to before AC_USE_SYSTEM_EXTENSIONS
14955 so that autoconf will see CFLAGS as set and not give it a default.
14956
14957 CFLAGS is also moved above AC_CONFIG_AUX_DIR, because CFLAGS should be
14958 defaulted to "" as soon as possible to catch any autoconf macros that try
14959 to use some other default. Regardless, this currently has no effect as that
14960 macro does not consider the CFLAGS variable.
14961
14962 Reviewed-by: Robbie Harwood <[email protected]>
14963 Reviewed-by: Daniel Kiper <[email protected]>
14964
14965 2022-04-20 Darren Kenny <[email protected]>
14966
14967 video/readers/jpeg: Fix possible invalid loop boundary condition
14968 The value of next_marker is adjusted based on the word sized value
14969 read from data->file.
14970
14971 The updated next_marker value should reference a location in the file
14972 just beyond the huffman table, and as such should not have a value
14973 larger than the size of the file.
14974
14975 Fixes: CID 73657
14976
14977 Reviewed-by: Daniel Kiper <[email protected]>
14978
14979 2022-04-20 Michael Chang <[email protected]>
14980
14981 lib/reed_solomon: Fix array subscript 0 is outside array bounds
14982 The grub_absolute_pointer() is a compound expression that can only work
14983 within a function. We are out of luck here when the pointer variables
14984 require global definition due to ATTRIBUTE_TEXT that have to use fully
14985 initialized global definition because of the way linkers work.
14986
14987 static gf_single_t * const gf_powx ATTRIBUTE_TEXT = (void *) 0x100000;
14988
14989 For the reason given above, use GCC diagnostic pragmas to suppress the
14990 array-bounds warning.
14991
14992 Reviewed-by: Daniel Kiper <[email protected]>
14993
14994 2022-04-20 Michael Chang <[email protected]>
14995
14996 build: Fix -Werror=array-bounds array subscript 0 is outside array bounds
14997 The GRUB is failing to build with GCC-12 in many places like this:
14998
14999 In function 'init_cbfsdisk',
15000 inlined from 'grub_mod_init' at ../../grub-core/fs/cbfs.c:391:3:
15001 ../../grub-core/fs/cbfs.c:345:7: error: array subscript 0 is outside array bounds of 'grub_uint32_t[0]' {aka 'unsigned int[]'} [-Werror=array-bounds]
15002 345 | ptr = *(grub_uint32_t *) 0xfffffffc;
15003 | ~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
15004
15005 This is caused by GCC regression in 11/12 [1]. In a nut shell, the
15006 warning is about detected invalid accesses at non-zero offsets to NULL
15007 pointers. Since hardwired constant address is treated as NULL plus an
15008 offset in the same underlying code, the warning is therefore triggered.
15009
15010 Instead of inserting #pragma all over the places where literal pointers
15011 are accessed to avoid diagnosing array-bounds, we can try to borrow the
15012 idea from Linux kernel that the absolute_pointer() macro [2][3] is used
15013 to disconnect a pointer using literal address from it's original object,
15014 hence GCC won't be able to make assumptions on the boundary while doing
15015 pointer arithmetic. With that we can greatly reduce the code we have to
15016 cover up by making initial literal pointer assignment to use the new
15017 wrapper but not having to track everywhere literal pointers are
15018 accessed. This also makes code looks cleaner.
15019
15020 Please note the grub_absolute_pointer() macro requires to be invoked in
15021 a function as long as it is compound expression. Some global variables
15022 with literal pointers has been changed to local ones in order to use
15023 grub_absolute_pointer() to initialize it. The shuffling is basically done
15024 in a selective and careful way that the variable's scope doesn't matter
15025 being local or global, for example, the global variable must not get
15026 modified at run time throughout. For the record, here's the list of
15027 global variables got shuffled in this patch:
15028
15029 grub-core/commands/i386/pc/drivemap.c:int13slot
15030 grub-core/term/i386/pc/console.c:bios_data_area
15031 grub-core/term/ns8250.c:serial_hw_io_addr
15032
15033 [1] https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99578
15034 [2] https://elixir.bootlin.com/linux/v5.16.14/source/include/linux/compiler.h#L180
15035 [3] https://elixir.bootlin.com/linux/v5.16.14/source/include/linux/compiler-gcc.h#L31
15036
15037 Reviewed-by: Daniel Kiper <[email protected]>
15038
15039 2022-04-20 Michael Chang <[email protected]>
15040
15041 util/mkimage: Fix dangling pointer may be used error
15042 The warning is real as long as dangling pointer to tmp_ may be used if
15043 o32 and o64 are both NULL. However that is not going to happen and can
15044 be ignored safely because the PE_OHDR is being used in a context that
15045 either o32 or o64 must have been properly initialized. Sadly compiler
15046 seems not to always optimize that unused tmp_ away so explicit
15047 suppression remain needed here.
15048
15049 ../util/mkimage.c: In function 'grub_install_generate_image':
15050 ../util/mkimage.c:1422:41: error: dangling pointer to 'tmp_' may be used [-Werror=dangling-pointer=]
15051 1422 | PE_OHDR (o32, o64, header_size) = grub_host_to_target32 (header_size);
15052 ../util/mkimage.c:857:28: note: 'tmp_' declared here
15053 857 | __typeof__((o64)->field) tmp_; \
15054 | ^~~~
15055
15056 Reviewed-by: Daniel Kiper <[email protected]>
15057
15058 2022-04-20 Chad Kimes <[email protected]>
15059
15060 net/drivers/efi/efinet: Configure VLAN from UEFI device used for PXE
15061 This patch handles automatic configuration of VLAN when booting from PXE
15062 on UEFI hardware.
15063
15064 Reviewed-by: Daniel Kiper <[email protected]>
15065
15066 2022-04-20 Chad Kimes <[email protected]>
15067
15068 kern/efi/efi: Print VLAN info in EFI device path
15069 Reviewed-by: Daniel Kiper <[email protected]>
15070
15071 2022-04-20 Chad Kimes <[email protected]>
15072
15073 net/net: Add net_set_vlan command
15074 Previously there was no way to set the 802.1Q VLAN identifier, despite
15075 support for vlantag in the net module. The only location vlantag was
15076 being populated was from PXE boot and only for Open Firmware hardware.
15077 This commit allows users to manually configure VLAN information for any
15078 interface.
15079
15080 Example usage:
15081 grub> net_ls_addr
15082 efinet1 00:11:22:33:44:55 192.0.2.100
15083 grub> net_set_vlan efinet1 100
15084 grub> net_ls_addr
15085 efinet1 00:11:22:33:44:55 192.0.2.100 vlan100
15086 grub> net_set_vlan efinet1 0
15087 efinet1 00:11:22:33:44:55 192.0.2.100
15088
15089 Reviewed-by: Daniel Kiper <[email protected]>
15090
15091 2022-04-20 Chad Kimes <[email protected]>
15092
15093 net/net: Add vlan information to net_ls_addr output
15094 Example output:
15095 grub> net_ls_addr
15096 efinet1 00:11:22:33:44:55 192.0.2.100 vlan100
15097
15098 Reviewed-by: Daniel Kiper <[email protected]>
15099
15100 2022-04-04 Chris Coulson <[email protected]>
15101
15102 kern/efi/init: Log a console error during a stack check failure
15103 The initial implementation of the stack protector just busy looped
15104 in __stack_chk_fail in order to reduce the amount of code being
15105 executed after the stack has been compromised because of a lack of
15106 firmware memory protections. With future firmware implementations
15107 incorporating memory protections such as W^X, call in to boot services
15108 when an error occurs in order to log a message to the console before
15109 automatically rebooting the machine.
15110
15111 Reviewed-by: Daniel Kiper <[email protected]>
15112
15113 2022-04-04 Alec Brown <[email protected]>
15114
15115 loader/i386/xnu: Fix uninitialized scalar variable
15116 In the function grub_xnu_boot(), struct grub_relocator32_state state is called
15117 but isn't being initialized. This results in the members grub_uint32_t ebx,
15118 grub_uint32_t ecx, grub_uint32_t edx, grub_uint32_t edi, and grub_uint32_t esi
15119 being filled with junk data from the stack since none of them are being set to
15120 any values. We can prevent this by setting state to {0}.
15121
15122 Fixes: CID 375035
15123
15124 Reviewed-by: Darren Kenny <[email protected]>
15125 Reviewed-by: Daniel Kiper <[email protected]>
15126
15127 2022-04-04 Alec Brown <[email protected]>
15128
15129 loader/i386/xnu: Fix uninitialized scalar variable
15130 In the function grub_xnu_boot_resume(), struct grub_relocator32_state state is
15131 called but isn't being initialized. This results in the members grub_uint32_t
15132 ebx, grub_uint32_t ecx, grub_uint32_t edx, grub_uint32_t esi, and grub_uint32_t
15133 edi being filled with junk data from the stack since none of them are being set
15134 to any values. We can prevent this by setting state to {0}.
15135
15136 Fixes: CID 375031
15137
15138 Reviewed-by: Darren Kenny <[email protected]>
15139 Reviewed-by: Daniel Kiper <[email protected]>
15140
15141 2022-04-04 Alec Brown <[email protected]>
15142
15143 loader/i386/pc/linux: Fix uninitialized scalar variable
15144 In the function grub_linux16_boot(), struct grub_relocator16_state state is
15145 called but isn't being initialized. This results in the members grub_uint32_t
15146 ebx, grub_uint32_t edx, grub_uint32_t esi, and grub_uint32_t ebp being filled
15147 with junk data from the stack since none of them are being set to any values.
15148 We can prevent this by setting state to {0}.
15149
15150 Fixes: CID 375028
15151
15152 Reviewed-by: Darren Kenny <[email protected]>
15153 Reviewed-by: Daniel Kiper <[email protected]>
15154
15155 2022-04-04 Alec Brown <[email protected]>
15156
15157 loader/i386/bsd: Fix uninitialized scalar variable
15158 In the function grub_netbsd_setup_video(), struct grub_netbsd_btinfo_framebuf
15159 params is called but isn't being initialized. The member grub_uint8_t
15160 reserved[16] isn't set to any values and is instead filled with junk data from
15161 the stack. We can prevent this by setting params to {0}.
15162
15163 Fixes: CID 375026
15164
15165 Reviewed-by: Darren Kenny <[email protected]>
15166 Reviewed-by: Daniel Kiper <[email protected]>
15167
15168 2022-04-04 Alec Brown <[email protected]>
15169
15170 net/net: Fix uninitialized scalar variable
15171 In the function grub_net_ipv6_get_link_local(), grub_net_network_level_address_t
15172 addr is called but isn't being initialized. This results in the member
15173 grub_dns_option_t option being filled with junk data from the stack. We can
15174 prevent this by setting the option member in addr to 0.
15175
15176 Fixes: CID 375033
15177
15178 Reviewed-by: Darren Kenny <[email protected]>
15179 Reviewed-by: Daniel Kiper <[email protected]>
15180
15181 2022-04-04 Alec Brown <[email protected]>
15182
15183 net/bootp: Fix uninitialized scalar variable
15184 In the function grub_net_configure_by_dhcp_ack(),
15185 grub_net_network_level_address_t addr is called but isn't being initialized.
15186 This results in the member grub_dns_option_t option being filled with junk data
15187 from the stack. To prevent this, we can set the option member in addr to 0.
15188
15189 Fixes: CID 375036
15190
15191 Reviewed-by: Darren Kenny <[email protected]>
15192 Reviewed-by: Daniel Kiper <[email protected]>
15193
15194 2022-04-04 Alec Brown <[email protected]>
15195
15196 net/arp: Fix uninitialized scalar variable
15197 In the function grub_net_arp_receive(), grub_net_network_level_address_t
15198 sender_addr and target_addr are being called but aren't being initialized.
15199 In both of these structs, each member is being set to a value except for
15200 grub_dns_option_t option. This results in this member being filled with junk
15201 data from the stack. To prevent this, we can set the option member in both
15202 structs to 0.
15203
15204 Fixes: CID 375030
15205
15206 Reviewed-by: Darren Kenny <[email protected]>
15207 Reviewed-by: Daniel Kiper <[email protected]>
15208
15209 2022-04-04 Glenn Washburn <[email protected]>
15210
15211 net/tcp: Only call grub_get_time_ms() when there are sockets to potentially retransmit for
15212 If the machine has network cards found, but there are no tcp open sockets
15213 (because the user doesn't use the network to boot), then grub_net_tcp_retransmit()
15214 should be a noop. Thus GRUB doesn't need to call grub_get_time_ms(), which
15215 does a call into firmware on powerpc-ieee1275, and probably other targets.
15216 So only call grub_get_time_ms() if there are tcp sockets.
15217
15218 Aside from improving performance, its also useful to stay out of the firmware
15219 as much as possible when debugging via QEMU because its a pain to get back
15220 in to GRUB execution. grub_net_tcp_retransmit() can get called very frequently
15221 via grub_net_poll_cards_idle() when GRUB is waiting for a keypress
15222 (grub_getkey_noblock() calls grub_net_poll_cards_idle()). This can be annoying
15223 when debugging an issue in GRUB on PowerPC in QEMU with GDB when GRUB is waiting
15224 for a keypress because interrupting via GDB nearly always lands in the OpenBIOS
15225 firmware's milliseconds call.
15226
15227 Reviewed-by: Daniel Kiper <[email protected]>
15228
15229 2022-04-04 Glenn Washburn <[email protected]>
15230
15231 net/net: Avoid unnecessary calls to grub_net_tcp_retransmit()
15232 In grub_net_poll_cards_idle_real(), only call grub_net_tcp_retransmit() if there
15233 are network cards found. If there are no network card found, there can be no
15234 tcp sockets to transmit on. So no need to go through that logic.
15235
15236 Reviewed-by: Daniel Kiper <[email protected]>
15237
15238 2022-04-04 Glenn Washburn <[email protected]>
15239
15240 net/net: Unset grub_net_poll_cards_idle when net module has been unloaded
15241 This looks like it was a copy/paste error. If the net module is unloaded,
15242 grub_net_poll_cards_idle should be NULL so that GRUB does not try to call
15243 a function which now doesn't exist.
15244
15245 Reviewed-by: Daniel Kiper <[email protected]>
15246
15247 2022-04-04 Glenn Washburn <[email protected]>
15248
15249 INSTALL: Add information on using --build when cross-compiling
15250 The autoconf 2.65 manual [1] strongly recommends specifying the --build
15251 option when the --host is used. Add this to the example and add a note
15252 that this is recommended.
15253
15254 [1] https://www.gnu.org/software/autoconf/manual/autoconf-2.65/html_node/Hosts-and-Cross_002dCompilation.html
15255
15256 Reviewed-by: Daniel Kiper <[email protected]>
15257
15258 2022-04-04 Glenn Washburn <[email protected]>
15259
15260 configure: Whitespace changes to improve readability
15261 Reviewed-by: Daniel Kiper <[email protected]>
15262
15263 2022-04-04 Glenn Washburn <[email protected]>
15264
15265 configure: Remove unused CFLAGS definitions
15266 These CFLAGS definitions are reset below them before they have a change to
15267 affect anything. The exception is the *-emu case, which is put in the next
15268 if block, which is the only place its used before getting reset.
15269
15270 Reviewed-by: Daniel Kiper <[email protected]>
15271
15272 2022-04-04 Glenn Washburn <[email protected]>
15273
15274 configure: Remove dead code
15275 It appears as though the intent of this code is to define abort() and main()
15276 symbols for some configure tests. However, it never gets used because the if
15277 is only entered when not building for *-emu, but the next if block only runs
15278 when building for *-emu. And the if block after that unconditionally resets
15279 CFLAGS. So this code can have no effect.
15280
15281 Additionally, s/aclocal.m4/acinclude.m4/ and move grub_ASM_USCORE to put
15282 with other marcos defined in acinclude.m4.
15283
15284 Reviewed-by: Daniel Kiper <[email protected]>
15285
15286 2022-04-04 Glenn Washburn <[email protected]>
15287
15288 configure: Sort AM_CONDITIONALs alphabetically
15289 Reviewed-by: Daniel Kiper <[email protected]>
15290
15291 2022-04-04 Glenn Washburn <[email protected]>
15292
15293 configure: Allow HOST_CC to override CC
15294 According to the INSTALL, "The HOST_* variables override not prefixed
15295 variables". This change makes it so, instead of previous behavior, which
15296 was to ignore the HOST_CC environment variable.
15297
15298 Reviewed-by: Daniel Kiper <[email protected]>
15299
15300 2022-04-04 Glenn Washburn <[email protected]>
15301
15302 gdb: Add malloc and free symbols to kernel.exec to improve gdb functionality
15303 Add linker flags when linking kernel.exec to have malloc and free point to
15304 grub_malloc() and grub_free() respectively. Some gdb functionality depends on
15305 gdb locating the symbols "malloc" and "free", such as dynamically creating
15306 strings for arguments to injected function calls. A trivial example would
15307 the gdb command 'p strlen("astring")'. Make sure not to do this on emu
15308 platforms, or an infinite loop occurs because emu has a special
15309 grub_malloc() that calls malloc().
15310
15311 Reviewed-by: Daniel Kiper <[email protected]>
15312
15313 2022-04-04 Renaud Métrich <[email protected]>
15314
15315 commands/search: Add new --efidisk-only option for EFI systems
15316 When using "search" on EFI systems, we sometimes want to exclude devices
15317 that are not EFI disks, e.g. md, lvm. This is typically used when
15318 wanting to chainload when having a software raid (md) for EFI partition:
15319 with no option, "search --file /EFI/redhat/shimx64.efi" sets root envvar
15320 to "md/boot_efi" which cannot be used for chainloading since there is no
15321 effective EFI device behind.
15322
15323 Reviewed-by: Daniel Kiper <[email protected]>
15324
15325 2022-04-04 Renaud Métrich <[email protected]>
15326
15327 commands/search: Refactor --no-floppy option to have something generic
15328 Reviewed-by: Daniel Kiper <[email protected]>
15329
15330 2022-04-04 Hans de Goede <[email protected]>
15331
15332 kern/main: Suppress the "Welcome to GRUB!" message in EFI builds
15333 GRUB EFI builds are now often used in combination with flicker-free
15334 boot, but this breaks with upstream GRUB because the "Welcome to GRUB!"
15335 message will kick the EFI fb into text mode and show the msg, breaking
15336 the flicker-free experience.
15337
15338 EFI systems are so fast, that when the menu or the countdown are
15339 enabled the message will be immediately overwritten, so in these cases
15340 not printing the message does not matter.
15341
15342 And in case when the timeout_style is set to TIMEOUT_STYLE_HIDDEN,
15343 the user has asked GRUB to be quiet (for example to allow flickfree
15344 boot) and thus the message should not be printed.
15345
15346 Reviewed-by: Robbie Harwood <[email protected]>
15347 Reviewed-by: Daniel Kiper <[email protected]>
15348
15349 2022-04-04 Hans de Goede <[email protected]>
15350
15351 normal/menu: Don't show "Booting `%s'" msg when auto-booting with TIMEOUT_STYLE_HIDDEN
15352 When the user has asked the menu code to be hidden/quiet and the current
15353 entry is being autobooted because the timeout has expired don't show
15354 the "Booting `%s'" msg.
15355
15356 This is necessary to let flicker-free boots really be flicker free,
15357 otherwise the "Booting `%s'" msg will kick the EFI fb into text mode
15358 and show the msg, breaking the flicker-free experience.
15359
15360 Reviewed-by: Robbie Harwood <[email protected]>
15361 Reviewed-by: Daniel Kiper <[email protected]>
15362
15363 2022-03-21 Hans de Goede <[email protected]>
15364
15365 term/efi/console: Do not set cursor until the first text output
15366 To allow flickerfree boot the EFI console code does not call
15367 grub_efi_set_text_mode(1) until some text is actually output. Depending
15368 on if the output text is because of an error loading, e.g. the .cfg
15369 file, or because of showing the menu the cursor needs to be on or off
15370 when the first text is shown. So far the cursor was hardcoded to being
15371 on, but this is causing drawing artifacts + slow drawing of the menu as
15372 reported here: https://bugzilla.redhat.com/show_bug.cgi?id=1946969
15373 Handle the cursorstate in the same way as the colorstate to fix this,
15374 when no text has been output yet, just cache the cursorstate and then
15375 use the last set value when the first text is output.
15376
15377 Fixes: 2d7c3abd871f (efi/console: Do not set text-mode until we actually need it)
15378 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1946969
15379
15380 Reviewed-by: Daniel Kiper <[email protected]>
15381
15382 2022-03-21 Hans de Goede <[email protected]>
15383
15384 term/efi/console: Do not set colorstate until the first text output
15385 GRUB_MOD_INIT(normal) does an unconditional:
15386
15387 grub_env_set ("color_normal", "light-gray/black");
15388
15389 which triggers a grub_term_setcolorstate() call. The original version
15390 of the "efi/console: Do not set text-mode until we actually need it" patch,
15391 https://lists.gnu.org/archive/html/grub-devel/2018-03/msg00125.html,
15392 protected against this by caching the requested state in
15393 grub_console_setcolorstate() and then only applying it when the first
15394 text output actually happens. During refactoring to move the
15395 grub_console_setcolorstate() up higher in the grub-core/term/efi/console.c
15396 file the code to cache the color-state + bail early was accidentally dropped.
15397 Restore the cache the color-state + bail early behavior from the original.
15398
15399 Fixes: 2d7c3abd871f (efi/console: Do not set text-mode until we actually need it)
15400
15401 Cc: Javier Martinez Canillas <[email protected]>
15402 Reviewed-by: Daniel Kiper <[email protected]>
15403
15404 2022-03-21 Darren Kenny <[email protected]>
15405
15406 kern/rescue_parser: Ensure that parser allocated memory is not leaked
15407 While it would appear unlikely that the memory allocated in *argv in
15408 grub_parser_split_cmdline() would be leaked, we should try ensure that
15409 it doesn't leak by calling grub_free() before we return from
15410 grub_rescue_parse_line().
15411
15412 To avoid a possible double-free, grub_parser_split_cmdline() is being
15413 changed to assign *argv = NULL when we've called grub_free() in the fail
15414 section.
15415
15416 Fixes: CID 96680
15417
15418 Reviewed-by: Daniel Kiper <[email protected]>
15419
15420 2022-03-21 Darren Kenny <[email protected]>
15421
15422 grub-mkimage: Only check aarch64 relocations when built for aarch64
15423 Coverity flagged the switch checks for R_AARCH64_* as being logically
15424 dead code, since it could never happen on x86 due to the masking of the
15425 values earlier in the code.
15426
15427 A check for building on __arm__ (which gcc and clang define) and for
15428 MKIMAGE_ELF64 (which GRUB defines) has been added to avoid this dead
15429 code being built in.
15430
15431 Fixes: CID 158599
15432
15433 Reviewed-by: Daniel Kiper <[email protected]>
15434
15435 2022-03-21 Daniel Kiper <[email protected]>
15436
15437 lib/posix_wrap/errno.h: Add __set_errno() macro
15438 $ ./configure --target=x86_64-w64-mingw32 --with-platform=efi --host=x86_64-w64-mingw32
15439 $ make
15440
15441 [...]
15442
15443 cat syminfo.lst | sort | gawk -f ./genmoddep.awk > moddep.lst || (rm -f moddep.lst; exit 1)
15444 __imp__errno in regexp is not defined
15445
15446 This happens because grub-core/lib/gnulib/malloc/dynarray_resize.c and
15447 grub-core/lib/gnulib/malloc/dynarray_emplace_enlarge.c (both are used by
15448 regexp module) from the latest Gnulib call __set_errno() which originally
15449 sets errno variable (Windows builds add __imp__ prefix). Of course it is
15450 not defined and grub_errno should be used instead.
15451
15452 Reviewed-by: Daniel Kiper <[email protected]>
15453
15454 2022-03-21 Robbie Harwood <[email protected]>
15455
15456 configure: Fix various new autotools warnings
15457 Reviewed-by: Daniel Kiper <[email protected]>
15458
15459 2022-03-21 Robbie Harwood <[email protected]>
15460
15461 gnulib: Handle warnings introduced by updated gnulib
15462 - Fix type of size variable in luks2_verify_key()
15463 - Avoid redefinition of SIZE_MAX and ATTRIBUTE_ERROR
15464 - Work around gnulib's int types on older compilers
15465
15466 Reviewed-by: Daniel Kiper <[email protected]>
15467
15468 2022-03-21 Robbie Harwood <[email protected]>
15469
15470 gnulib: Update gnulib version and drop most gnulib patches
15471 In addition to the changes carried in our gnulib patches, several
15472 Coverity and code hygiene fixes that were previously downstream are also
15473 included in this 3-year gnulib increment.
15474
15475 Unfortunately, fix-width.patch is retained.
15476
15477 Bump minimum autoconf version from 2.63 to 2.64 and automake from 1.11
15478 to 1.14, as required by gnulib.
15479
15480 Sync bootstrap script itself with gnulib.
15481
15482 Update regexp module for new dynarray dependency.
15483
15484 Reviewed-by: Daniel Kiper <[email protected]>
15485
15486 2022-03-21 Robbie Harwood <[email protected]>
15487
15488 gnulib: Drop no-abort.patch
15489 Originally added in commit db7337a3d (grub-core/lib/posix_wrap/stdlib.h
15490 (abort): Removed), this patched out all relevant invocations of abort()
15491 in gnulib. While it was not documented why at the time, testing suggests
15492 that there's no abort() implementation available for gnulib to use.
15493
15494 gnulib's position is that the use of abort() is correct here, since it
15495 happens when input violates a "shall" from POSIX. Additionally, the
15496 code in question is probably not reachable. Since abort() is more
15497 friendly to user-space, they prefer to make no change, so we can just
15498 carry a define instead (suggested by Paul Eggert).
15499
15500 Reviewed-by: Daniel Kiper <[email protected]>
15501
15502 2022-03-21 Robbie Harwood <[email protected]>
15503
15504 gnulib: Drop fix-base64.patch
15505 Originally added in commit 9fbdec2f (bootstrap: Add gnulib's base64
15506 module) and subsequently modified in commit 552c9fd08 (gnulib: Fix build
15507 of base64 when compiling with memory debugging), fix-base64.patch
15508 handled two problems we have using gnulib, which are exercised by the
15509 base64 module but not directly caused by it.
15510
15511 First, GRUB defines its own bool type, while gnulib expects the
15512 equivalent of stdbool.h to be present. Rather than patching gnulib,
15513 instead use gnulib's stdbool module to provide a bool type if needed
15514 (suggested by Simon Josefsson).
15515
15516 Second, our config.h doesn't always inherit config-util.h, which is
15517 where gnulib-related options like _GL_ATTRIBUTE_CONST end up.
15518 fix-base64.h worked around this by defining the attribute away, but this
15519 workaround is better placed in config.h itself, not a gnulib patch.
15520
15521 Reviewed-by: Daniel Kiper <[email protected]>
15522
15523 2022-03-21 Robbie Harwood <[email protected]>
15524
15525 config: Where present, ensure config-util.h precedes config.h
15526 gnulib defines go in config-util.h, and we need to know whether to
15527 provide duplicates in config.h or not.
15528
15529 Reviewed-by: Daniel Kiper <[email protected]>
15530
15531 2022-03-21 Robbie Harwood <[email protected]>
15532
15533 config.h.in: Use visual indentation
15534 Reviewed-by: Daniel Kiper <[email protected]>
15535
15536 2022-03-14 Robbie Harwood <[email protected]>
15537
15538 INSTALL: Drop mention of libusb
15539 The commit 9d25b0da9 (Remove emu libusb support.) dropped use of libusb,
15540 but did not remove mention of it from INSTALL file.
15541
15542 Reviewed-by: Daniel Kiper <[email protected]>
15543
15544 2022-03-14 Daniel Kiper <[email protected]>
15545
15546 INSTALL: Add more cross-compiling Debian packages
15547 The mingw-w64-tools is especially important because with out it some
15548 Windows builds may fail due to lack of proper pkg-config.
15549
15550 Reviewed-by: Robbie Harwood <[email protected]>
15551
15552 2022-03-14 Daniel Kiper <[email protected]>
15553
15554 configure: Drop ${grub_coredir} unneeded references
15555 These are probably stray references left after earlier removals.
15556
15557 Reviewed-by: Robbie Harwood <[email protected]>
15558
15559 2022-03-14 Daniel Kiper <[email protected]>
15560
15561 conf/i386-cygwin-img-ld: Do not discard .data and .edata sections
15562 $ ./configure --target=i686-w64-mingw32 --with-platform=efi --host=i686-w64-mingw32
15563
15564 [...]
15565
15566 checking if __bss_start is defined by the compiler... no
15567 checking if edata is defined by the compiler... no
15568 checking if _edata is defined by the compiler... no
15569 configure: error: none of __bss_start, edata or _edata is defined
15570
15571 This happens on machines with quite recent ld due to an error:
15572
15573 `edata' referenced in section `.text' of /tmp/cc72w9E4.o: defined in discarded section `.data' of conftest.exe
15574 collect2: error: ld returned 1 exit status
15575
15576 So, we have to tell linker to not discard .data and .edata sections.
15577 The trick comes from ld documentation:
15578
15579 3.6.7 Output Section Discarding
15580
15581 The linker will not normally create output sections with no contents.
15582 This is for convenience when referring to input sections that may or may
15583 not be present in any of the input files. For example:
15584
15585 .foo : { *(.foo) }
15586
15587 will only create a ‘.foo’ section in the output file if there is a
15588 ‘.foo’ section in at least one input file, and if the input sections are
15589 not all empty. Other link script directives that allocate space in an
15590 output section will also create the output section. So too will
15591 assignments to dot even if the assignment does not create space, except
15592 for ‘. = 0’, ‘. = . + 0’, ‘. = sym’, ‘. = . + sym’ and ‘. = ALIGN (. !=
15593 0, expr, 1)’ when ‘sym’ is an absolute symbol of value 0 defined in the
15594 script. This allows you to force output of an empty section with ‘. = .’.
15595
15596 This change does not impact generated binaries because the
15597 conf/i386-cygwin-img-ld.sc linker script is used only when
15598 you run configure.
15599
15600 Reviewed-by: Robbie Harwood <[email protected]>
15601
15602 2022-03-14 Daniel Kiper <[email protected]>
15603
15604 commands/i386/pc/sendkey: Fix "writing 1 byte into a region of size 0" build error
15605 Latest GCC may complain in that way:
15606
15607 commands/i386/pc/sendkey.c: In function ‘grub_sendkey_postboot’:
15608 commands/i386/pc/sendkey.c:223:21: error: writing 1 byte into a region of size 0 [-Werror=stringop-overflow=]
15609 223 | *((char *) 0x41a) = 0x1e;
15610 | ~~~~~~~~~~~~~~~~~~^~~~~~
15611
15612 The volatile keyword addition helps and additionally assures us the
15613 compiler will not optimize out fixed assignments.
15614
15615 Reviewed-by: Robbie Harwood <[email protected]>
15616
15617 2022-03-14 Daniel Kiper <[email protected]>
15618
15619 loader/i386/bsd: Initialize ptr variable in grub_bsd_add_meta()
15620 Latest GCC may complain in that way:
15621
15622 In file included from ../include/grub/disk.h:31,
15623 from ../include/grub/file.h:26,
15624 from ../include/grub/loader.h:23,
15625 from loader/i386/bsd.c:19:
15626 loader/i386/bsd.c: In function ‘grub_cmd_openbsd’:
15627 ../include/grub/misc.h:71:10: error: ‘ptr’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
15628 71 | return grub_memmove (dest, src, n);
15629 | ^~~~~~~~~~~~~~~~~~~~~~~~~~~
15630 loader/i386/bsd.c:266:9: note: ‘ptr’ was declared here
15631 266 | void *ptr;
15632 | ^~~
15633
15634 So, let's fix it by assigning NULL to ptr in grub_bsd_add_meta().
15635
15636 Reviewed-by: Robbie Harwood <[email protected]>
15637
15638 2022-03-14 Daniel Kiper <[email protected]>
15639
15640 osdep/windows/platform: Disable gcc9 -Waddress-of-packed-member
15641 $ ./configure --target=x86_64-w64-mingw32 --with-platform=efi --host=x86_64-w64-mingw32
15642 $ make
15643
15644 [...]
15645
15646 In file included from grub-core/osdep/platform.c:4:
15647 grub-core/osdep/windows/platform.c: In function ‘grub_install_register_efi’:
15648 grub-core/osdep/windows/platform.c:382:41: error: taking address of packed member of ‘struct grub_efi_file_path_device_path’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
15649 382 | path16_len = grub_utf8_to_utf16 (filep->path_name,
15650 | ~~~~~^~~~~~~~~~~
15651
15652 Disable the -Wadress-of-packaed-member diagnostic for grub_utf8_to_utf16()
15653 call which contains filep->path_name reference. It seems safe because the
15654 structure is defined according to the UEFI spec and we hope authors did not
15655 make any mistake... :-)
15656
15657 This fix is similar to the fix in the commit 8e8723a6b
15658 (f2fs: Disable gcc9 -Waddress-of-packed-member).
15659
15660 Reviewed-by: Robbie Harwood <[email protected]>
15661
15662 2022-03-14 Glenn Washburn <[email protected]>
15663
15664 po: Un-transliterate the %zu format code
15665 Commit 45bffae13 (util/resolve: Bail with error if moddep.lst file line is
15666 too long) uses the %zu format specifier which has not been used in
15667 any translated strings yet. So the sed scripts used for transliterating
15668 certain languages need to be updated otherwise creation of the message
15669 indexes will fail on an unknown format code. This is essentially the same
15670 issue fixed for the %m format code in commit 2e246b6f (po: Fix replacement
15671 of %m in sed programs).
15672
15673 Also reorder transliteration lines to be more lexicographically ordered.
15674
15675 Reviewed-by: Daniel Kiper <[email protected]>
15676
15677 2022-03-14 Daniel Axtens <[email protected]>
15678
15679 net: Check against nb->tail in grub_netbuff_pull()
15680 GRUB netbuff structure members track 2 different things: the extent of memory
15681 allocated for the packet, and the extent of memory currently being worked on.
15682
15683 This works out in the structure as follows:
15684
15685 nb->head: beginning of the allocation
15686 nb->data: beginning of the working data
15687 nb->tail: end of the working data
15688 nb->end: end of the allocation
15689
15690 The head and end pointers are set in grub_netbuff_alloc() and do not change.
15691 The data and tail pointers are initialised to point at start of the
15692 allocation (that is, head == data == tail initially), and are then
15693 manipulated by grub_netbuff_*() functions. Key functions are as follows:
15694
15695 - grub_netbuff_put(): "put" more data into the packet - advance nb->tail
15696 - grub_netbuff_unput(): trim the tail of the packet - retract nb->tail
15697 - grub_netbuff_pull(): "consume" some packet data - advance nb->data
15698 - grub_netbuff_reserve(): reserve space for future headers - advance nb->data and nb->tail
15699 - grub_netbuff_push(): "un-consume" data to allow headers to be written - retract nb->data
15700
15701 Each of those functions does some form of error checking. For example,
15702 grub_netbuff_put() does not allow nb->tail to exceed nb->end, and
15703 grub_netbuff_push() does not allow nb->data to be before nb->head.
15704
15705 However, grub_netbuff_pull()'s error checking is a bit weird. It advances nb->data
15706 and checks that it does not exceed nb->end. That allows you to get into the
15707 situation where nb->data > nb->tail, which should not be.
15708
15709 Make grub_netbuff_pull() check against both nb->tail and nb->end. In theory just
15710 checking against ->tail should be sufficient but the extra check should be
15711 cheap and seems like good defensive practice.
15712
15713 Reviewed-by: Daniel Kiper <[email protected]>
15714
15715 2022-03-14 Fabian Vogt <[email protected]>
15716
15717 grub-mount: Add support for libfuse3
15718 The libfuse 3.0.0 got released in 2016, with some API changes compared to 2.x.
15719 This commit introduces support for 3.x while keeping it compatible with 2.6
15720 as a fallback still.
15721
15722 To detect fuse3, switch configure over to use pkg-config, which is simpler yet
15723 more reliable than looking for library and header manually. Also set
15724 FUSE_USE_VERSION that way, as it depends on the used libfuse version.
15725
15726 Now that the CFLAGS are read from pkg-config, use just <fuse.h>, which works
15727 with 2.x as well as 3.x and is recommended by libfuse upstream.
15728
15729 One behavior change of libfuse3 is that FUSE_ATOMIC_O_TRUNC is set by default,
15730 which means that open with O_TRUNC is passed as-is instead of calling the
15731 truncate operation. With libfuse2, truncate failed with -ENOSYS and that was
15732 returned to the application. To make O_TRUNC fail with libfuse3, return -EROFS
15733 explicitly if writing was requested.
15734
15735 Reviewed-by: Daniel Kiper <[email protected]>
15736
15737 2022-03-14 Elyes Haouas <[email protected]>
15738
15739 include: Remove trailing whitespaces
15740 Reviewed-by: Daniel Kiper <[email protected]>
15741
15742 util: Remove trailing whitespaces
15743 Reviewed-by: Daniel Kiper <[email protected]>
15744
15745 video: Remove trailing whitespaces
15746 Reviewed-by: Daniel Kiper <[email protected]>
15747
15748 tests: Remove trailing whitespaces
15749 Reviewed-by: Daniel Kiper <[email protected]>
15750
15751 term: Remove trailing whitespaces
15752 Reviewed-by: Daniel Kiper <[email protected]>
15753
15754 script: Remove trailing whitespaces
15755 Reviewed-by: Daniel Kiper <[email protected]>
15756
15757 partmap: Remove trailing whitespaces
15758 Reviewed-by: Daniel Kiper <[email protected]>
15759
15760 osdep: Remove trailing whitespaces
15761 Reviewed-by: Daniel Kiper <[email protected]>
15762
15763 normal: Remove trailing whitespaces
15764 Reviewed-by: Daniel Kiper <[email protected]>
15765
15766 net: Remove trailing whitespaces
15767 Reviewed-by: Daniel Kiper <[email protected]>
15768
15769 loader: Remove trailing whitespaces
15770 Reviewed-by: Daniel Kiper <[email protected]>
15771
15772 lib: Remove trailing whitespaces
15773 Reviewed-by: Daniel Kiper <[email protected]>
15774
15775 kern: Remove trailing whitespaces
15776 Reviewed-by: Daniel Kiper <[email protected]>
15777
15778 io: Remove trailing whitespaces
15779 Reviewed-by: Daniel Kiper <[email protected]>
15780
15781 gfxmenu: Remove trailing whitespaces
15782 Reviewed-by: Daniel Kiper <[email protected]>
15783
15784 gfxmenu: Remove trailing whitespaces
15785 Reviewed-by: Daniel Kiper <[email protected]>
15786
15787 fs: Remove trailing whitespaces
15788 Reviewed-by: Daniel Kiper <[email protected]>
15789
15790 font: Remove trailing whitespaces
15791 Reviewed-by: Daniel Kiper <[email protected]>
15792
15793 disk: Remove trailing whitespaces
15794 Reviewed-by: Daniel Kiper <[email protected]>
15795
15796 commands: Remove trailing whitespaces
15797 Reviewed-by: Daniel Kiper <[email protected]>
15798
15799 bus: Remove trailing whitespaces
15800 Reviewed-by: Daniel Kiper <[email protected]>
15801
15802 2022-03-07 Chad Kimes <[email protected]>
15803
15804 net/ethernet: Fix VLAN networking on little-endian systems
15805 VLAN configuration seems to have never worked on little-endian systems.
15806 This is likely because VLANTAG_IDENTIFIER is not byte-swapped before
15807 copying into the net buffer, nor is inf->vlantag. We can resolve this by
15808 using grub_cpu_to_be16{_compile_time}() and its inverse when copying
15809 VLAN info to/from the net buffer.
15810
15811 Reviewed-by: Daniel Kiper <[email protected]>
15812
15813 2022-03-07 Heinrich Schuchardt <[email protected]>
15814
15815 commands/efi/lsefisystab: Short text EFI_IMAGE_SECURITY_DATABASE_GUID
15816 The EFI_IMAGE_SECURITY_DATABASE_GUID is used for the image execution
15817 information table (cf. UEFI specification 2.9, 32.5.3.1 Using The Image
15818 Execution Information Table).
15819
15820 The lsefisystab command is used to display installed EFI configuration
15821 tables. Currently it only shows the GUID but not a short text for the
15822 table.
15823
15824 Provide a short text for the EFI_IMAGE_SECURITY_DATABASE_GUID.
15825
15826 Reviewed-by: Daniel Kiper <[email protected]>
15827
15828 2022-03-07 Glenn Washburn <[email protected]>
15829
15830 tests: Fix whitespace formatting
15831 Reviewed-by: Daniel Kiper <[email protected]>
15832
15833 2022-03-07 Peter Jones <[email protected]>
15834
15835 ChangeLog: Retire ChangeLog-2015
15836 ChangeLog-2015 has been untouched for over 7 years now, and any
15837 information in it is purely for historical purposes. At the same time,
15838 grepping for code winds up matching this file quite a bit, almost never
15839 accomplishing anything other than cluttering up your grep results. We
15840 don't need this in the main repo, and "git show" will find it if you're
15841 looking at the old history of commits on some file.
15842
15843 This patch deletes it and the Makefile.am rule to distribute it.
15844
15845 Reviewed-by: Daniel Axtens <[email protected]>
15846 Reviewed-by: Robbie Harwood <[email protected]>
15847 Reviewed-by: Javier Martinez Canillas <[email protected]>
15848 Reviewed-by: Daniel Kiper <[email protected]>
15849
15850 2022-03-07 Peter Levine <[email protected]>
15851
15852 templates: Properly handle multiple initrd paths in 30_os-prober
15853 os-prober now effectively handles multiple paths passed to initrd, but
15854 grub-mkconfig still truncates off any subsequent space-delimited paths.
15855
15856 Support proper parsing of space-delimited initrd paths passed from
15857 os-prober for distributions, like Manjaro, that require it.
15858
15859 Fixes: https://savannah.gnu.org/bugs/?47681
15860
15861 Reviewed-by: Daniel Kiper <[email protected]>
15862
15863 2022-03-07 Samuel Thibault <[email protected]>
15864
15865 templates: Add support for pci-arbiter and rumpdisk on Hurd
15866 This adds pci-arbiter and rumpdisk as bootstrap modules whenever they are
15867 available. This opens the path for fully-userland disk support.
15868
15869 Reviewed-by: Daniel Kiper <[email protected]>
15870
15871 2022-03-07 Glenn Washburn <[email protected]>
15872
15873 mm: Temporarily disable grub_mm_debug while calling grub_vprintf() in grub_printf()
15874 To prevent infinite recursion when grub_mm_debug is on, disable it when
15875 calling grub_vprintf(). One such call loop is:
15876 grub_vprintf() -> parse_printf_args() -> parse_printf_arg_fmt() ->
15877 grub_debug_calloc() -> grub_printf() -> grub_vprintf().
15878
15879 Reviewed-by: Daniel Kiper <[email protected]>
15880
15881 2022-03-07 Glenn Washburn <[email protected]>
15882
15883 mm: Export grub_mm_dump() and grub_mm_dump_free()
15884 These functions may be useful within modules as well. Export them so that
15885 modules can use them.
15886
15887 Reviewed-by: Daniel Kiper <[email protected]>
15888
15889 2022-03-07 Glenn Washburn <[email protected]>
15890
15891 configure: Properly handle MM_DEBUG
15892 Define MM_DEBUG in config.h when --enable-mm-debug is passed to configure.
15893 It was being defined in config-util.h which only gets used when building
15894 GRUB utilities for the host side. The enabling of debugging for memory
15895 management in include/grub/mm.h explicitly does not happen when compiling
15896 for the GRUB utilities. So this debugging code effectively could never be
15897 enabled. Note, that MM_DEBUG is defined in an #if directive because the
15898 enabling of debugging checks if MM_DEBUG is defined, not what its value is.
15899 So even if MM_DEBUG were defined to nothing, the debugging code would
15900 still be enabled.
15901
15902 Reviewed-by: Daniel Kiper <[email protected]>
15903
15904 2022-03-07 Fangrui Song <[email protected]>
15905
15906 configure: Replace -Wl,-r,-d with -Wl,-r and add -fno-common
15907 In GNU ld and ld.lld, -d is used with -r to allocate space to COMMON symbols.
15908 This behavior is presumably to work around legacy projects which inspect
15909 relocatable output by themselves and do not handle COMMON symbols. The GRUB
15910 does not do this.
15911
15912 See https://github.com/llvm/llvm-project/issues/53660
15913 -d is quite useless and ld.lld 15.0.0 will make -d no-op.
15914
15915 COMMON symbols have special symbol resolution semantics which can cause surprise
15916 (see https://maskray.me/blog/2022-02-06-all-about-common-symbols). GCC<10 and
15917 Clang<11 defaulted to -fcommon. Just use -fno-common to avoid COMMON symbols.
15918
15919 Reviewed-by: Daniel Kiper <[email protected]>
15920
15921 2022-03-07 Glenn Washburn <[email protected]>
15922
15923 tests: Add check-native and check-nonnative make targets
15924 This allows for testing only tests that run directly on the build machine or
15925 only tests that run in a virtualized environment. When testing multiple
15926 targets on the same build machine the native tests only need to be run once
15927 for all targets. Whereas, the nonnative tests must be run for each target
15928 because the test is potentially compiled differently for each target.
15929
15930 Reviewed-by: Daniel Kiper <[email protected]>
15931
15932 2022-03-07 Renaud Métrich <[email protected]>
15933
15934 commands/search: Fix bug stopping iteration when --no-floppy is used
15935 When using --no-floppy and a floppy was encountered, iterate_device()
15936 was returning 1, causing the iteration to stop instead of continuing.
15937
15938 Reviewed-by: Daniel Kiper <[email protected]>
15939
15940 2022-03-07 Glenn Washburn <[email protected]>
15941
15942 Revert "iee1275/datetime: Fix off-by-1 error."
15943 This is causing the test grub_cmd_date() to fail because the returned
15944 date is one day more than it should be.
15945
15946 This reverts commit 607d66116 (iee1275/datetime: Fix off-by-1 error.).
15947
15948 Tested-by: Daniel Axtens <[email protected]>
15949 Reviewed-by: Daniel Kiper <[email protected]>
15950
15951 2022-02-08 Glenn Washburn <[email protected]>
15952
15953 tests: Remove $((BASE#NUM)) bashism in grub-fs-tester
15954 This bashism allows converting NUM in base BASE to decimal. Its not needed
15955 because the only place its used is to convert from hexidecimal and this can
15956 also be done with the more portable $((0xHEXNUM)) syntax.
15957
15958 Reviewed-by: Daniel Kiper <[email protected]>
15959
15960 2022-02-08 Glenn Washburn <[email protected]>
15961
15962 tests: Skip pata_test on i386-efi
15963 In comparison to other i386 targets, on i386-efi the Q35 QEMU machine type
15964 is used to do the testing to be able to make use of the EFI firmware in
15965 QEMU. On the Q35 machine type there is no way to use ATA to communicate with
15966 an IDE, only AHCI.
15967
15968 Reviewed-by: Daniel Kiper <[email protected]>
15969
15970 2022-02-08 Glenn Washburn <[email protected]>
15971
15972 tests: Do not remove image file on error in pata_test
15973 The image file can be useful in debugging an issue when the test fails.
15974
15975 Reviewed-by: Daniel Kiper <[email protected]>
15976
15977 2022-02-08 Alec Brown <[email protected]>
15978
15979 util/grub-module-verifierXX: Validate elf section header table index for section name string table
15980 In grub-module-verifierXX.c, the function find_section() uses the value from
15981 grub_target_to_host16 (e->e_shstrndx) to obtain the section header table index
15982 of the section name string table, but it wasn't being checked if the value was
15983 there.
15984
15985 According to the elf(5) manual page,
15986 "If the index of section name string table section is larger than or equal
15987 to SHN_LORESERVE (0xff00), this member holds SHN_XINDEX (0xffff) and the real
15988 index of the section name string table section is held in the sh_link member of
15989 the initial entry in section header table. Otherwise, the sh_link member of the
15990 initial entry in section header table contains the value zero."
15991
15992 Since this check wasn't being made, the function get_shstrndx() is being added
15993 to make this check and use e_shstrndx if it doesn't have SHN_XINDEX as a value,
15994 else use sh_link. We also need to make sure e_shstrndx isn't greater than or
15995 equal to SHN_LORESERVE and sh_link isn't less than SHN_LORESERVE.
15996
15997 Note that it may look as though the argument *arch isn't being used, it's
15998 actually required in order to use the macros grub_target_to_host*(x) which are
15999 unwinded to grub_target_to_host*_real(arch, (x)) based on defines earlier in
16000 the file.
16001
16002 Reviewed-by: Darren Kenny <[email protected]>
16003 Reviewed-by: Daniel Kiper <[email protected]>
16004
16005 2022-02-08 Alec Brown <[email protected]>
16006
16007 util/grub-module-verifierXX: Validate number of elf section header table entries
16008 In grub-module-verifierXX.c, grub_target_to_host16 (e->e_shnum) is used to
16009 obtain the number of section header table entries, but it wasn't being
16010 checked if the value was there.
16011
16012 According to the elf(5) manual page,
16013 "If the number of entries in the section header table is larger than or equal
16014 to SHN_LORESERVE (0xff00), e_shnum holds the value zero and the real number of
16015 entries in the section header table is held in the sh_size member of the intial
16016 entry in section header table. Otherwise, the sh_size member of the initial
16017 entry in the section header table holds the value zero."
16018
16019 Since this check wasn't being made, the function get_shnum() is being added to
16020 make this check and use whichever member doesn't have a value of zero. If both
16021 are zero, then we must return an error. We also need to make sure that e_shnum
16022 doesn't have a value greater than or equal to SHN_LORESERVE and sh_size isn't
16023 less than SHN_LORESERVE.
16024
16025 Note that it may look as though the argument *arch isn't being used, it's
16026 actually required in order to use the macros grub_target_to_host*(x) which are
16027 unwinded to grub_target_to_host*_real(arch, (x)) based on defines earlier in
16028 the file.
16029
16030 Fixes: CID 314021
16031 Fixes: CID 314027
16032 Fixes: CID 314033
16033
16034 Reviewed-by: Darren Kenny <[email protected]>
16035 Reviewed-by: Daniel Kiper <[email protected]>
16036
16037 2022-02-08 Alec Brown <[email protected]>
16038
16039 util/grub-module-verifierXX: Add function to calculate section headers
16040 Added the function get_shdr() which returns the section header at a given index
16041 parameter passed into this function. This helps traverse the section header
16042 table and reduces repeated calls to lengthy equations used to obtain section
16043 headers.
16044
16045 Note that it may look as though the argument *arch isn't being used, it's
16046 actually required in order to use the macros grub_target_to_host*(x) which are
16047 unwinded to grub_target_to_host*_real(arch, (x)) based on defines earlier in the
16048 file.
16049
16050 Reviewed-by: Darren Kenny <[email protected]>
16051 Reviewed-by: Daniel Kiper <[email protected]>
16052
16053 2022-02-08 Alec Brown <[email protected]>
16054
16055 fs/affs: Fix resource leaks
16056 In commit 178ac5107389 (affs: Fix memory leaks), fixes were made to
16057 grub_affs_iterate_dir() to prevent memory leaks from occurring after it
16058 returns without freeing node. However, there were still some instances
16059 where node was causing a memory leak when the function returns after
16060 calling grub_affs_create_node(). In this function, new memory is
16061 allocated to node but doesn't get freed until the hook() function is
16062 called near the end. Before hook() is called, node should be freed in
16063 grub_affs_create_node() before returning out of it.
16064
16065 Fixes: 178ac5107389 (affs: Fix memory leaks)
16066 Fixes: CID 73759
16067
16068 Reviewed-by: Darren Kenny <[email protected]>
16069 Reviewed-by: Daniel Kiper <[email protected]>
16070
16071 2022-02-08 Heinrich Schuchardt <[email protected]>
16072
16073 RISC-V: Adjust -march flags for binutils 2.38
16074 As of version 2.38 binutils defaults to ISA specification version
16075 2019-12-13. This version of the specification has has separated the
16076 the csr read/write (csrr*/csrw*) instructions and the fence.i from
16077 the I extension and put them into separate Zicsr and Zifencei
16078 extensions.
16079
16080 This implies that we have to adjust the -march flag passed to the
16081 compiler accordingly.
16082
16083 Reviewed-by: Daniel Kiper <[email protected]>
16084
16085 2022-02-08 Heinrich Schuchardt <[email protected]>
16086
16087 efi: Correct struct grub_efi_boot_services
16088 The UEFI specification defines that the EFI_BOOT_SERVICES.Exit(() service may return
16089 EFI_SUCCESS or EFI_INVALID_PARAMETER. So it cannot be __attribute__((noreturn)).
16090
16091 Reviewed-by: Daniel Kiper <[email protected]>
16092
16093 2022-02-08 Glenn Washburn <[email protected]>
16094
16095 conf/Makefile.common: Order alphabetically variables
16096 Reviewed-by: Daniel Kiper <[email protected]>
16097
16098 2022-02-08 Stephen Balousek <[email protected]>
16099
16100 net/http: Allow use of non-standard TCP/IP ports
16101 Allow the use of HTTP servers listening on ports other 80. This is done
16102 with an extension to the http notation:
16103
16104 (http[,server[,port]])
16105
16106 - or -
16107
16108 (http[,server[:port]])
16109
16110 Reviewed-by: Daniel Kiper <[email protected]>
16111
16112 2022-02-08 Glenn Washburn <[email protected]>
16113
16114 Makefile: Only look for @MARKER@ at the start of a line when generating libgrub_a_init.lst
16115 Under certain conditions libgrub.pp gets generated with a such that it
16116 contains a bunch of CPP defines, at least one of which contains "@MARKER@".
16117 This line should not be used when generating libgrub_a_init.lst, otherwise
16118 we get compiler errors like:
16119
16120 libgrub_a_init.c:22:18: error: stray ‘#’ in program
16121 22 | extern void grub_#define_init (void);
16122 | ^
16123 libgrub_a_init.c:22:19: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘define_init’
16124 22 | extern void grub_#define_init (void);
16125 | ^~~~~~~~~~~
16126 libgrub_a_init.c:23:18: error: stray ‘#’ in program
16127 23 | extern void grub_#define_fini (void);
16128 | ^
16129 libgrub_a_init.c:23:19: error: expected ‘=’, ‘,’, ‘;’, ‘asm’ or ‘__attribute__’ before ‘define_fini’
16130 23 | extern void grub_#define_fini (void);
16131 | ^~~~~~~~~~~
16132 ...
16133
16134 When generating libgrub_a_init.lst only lines starting with "@MARKER@"
16135 are desired.
16136
16137 Reviewed-by: Daniel Kiper <[email protected]>
16138
16139 2022-02-08 Glenn Washburn <[email protected]>
16140
16141 gentpl.py: Fix issue where sometimes marker files have CPP defines
16142 When generating video.lst, modules whose marker file contains the string
16143 VIDEO_LIST_MARKER are selected. But when the marker file contains the CPP
16144 defines, one of the defines is VIDEO_LIST_MARKER and is present in all
16145 marker files, so they are all selected. By removing the defines, the correct
16146 modules are selected.
16147
16148 Reviewed-by: Daniel Kiper <[email protected]>
16149
16150 2022-02-08 Glenn Washburn <[email protected]>
16151
16152 util/resolve: Bail with error if moddep.lst file line is too long
16153 The code reads each line into a buffer of size 1024 and does not check if
16154 the line is longer. So a line longer than 1024 will be read as a valid line
16155 followed by an invalid line. Then an error confusing to the user is sent
16156 with the test "invalid line format". But the line format is perfectly fine,
16157 the problem is in GRUB's parser. Check if we've hit a line longer than the
16158 size of the buffer, and if so send a more correct and reasonable error.
16159
16160 Reviewed-by: Daniel Kiper <[email protected]>
16161
16162 2022-02-08 Glenn Washburn <[email protected]>
16163
16164 util/resolve: Do not read past the end of the array in read_dep_list()
16165 If the last non-NULL byte of "buf" is not a white-space character (such as
16166 when a read line is longer than the size of "buf"), then "p" will eventually
16167 point to the byte after the last byte in "buf". After which "p" will be
16168 dereferenced in the while conditional leading to an out of bounds read. Make
16169 sure that "p" is inside "buf" before dereferencing it.
16170
16171 Reviewed-by: Daniel Kiper <[email protected]>
16172
16173 2022-02-07 Glenn Washburn <[email protected]>
16174
16175 kern/misc: Allow selective disabling of debug facility names
16176 Sometimes you only know which debug logging facility names you want to
16177 turn off, not necessarily all the ones you want enabled. This patch allows
16178 the debug string to contain facility names in the $debug variable which are
16179 prefixed with a "-" to disable debug log messages for that conditional. Say
16180 you want all debug logging on except for btrfs and scripting, then do:
16181 "set debug=all,-btrfs,-scripting"
16182
16183 Note, that only the last occurrence of the facility name with or without a
16184 leading "-" is considered. So simply appending ",-facilityname" to the
16185 $debug variable will disable that conditional. To illustrate, the command
16186 "set debug=all,-btrfs,-scripting,btrfs" will enable btrfs.
16187
16188 Also, add documentation explaining this new behavior.
16189
16190 Reviewed-by: Daniel Kiper <[email protected]>
16191
16192 2022-02-07 Glenn Washburn <[email protected]>
16193
16194 cryptodisk: Fix Coverity use after free bug
16195 The Coverity output is:
16196
16197 *** CID 366905: Memory - illegal accesses (USE_AFTER_FREE)
16198 /grub-core/disk/cryptodisk.c: 1064 in grub_cryptodisk_scan_device_real()
16199 1058 cleanup:
16200 1059 if (askpass)
16201 1060 {
16202 1061 cargs->key_len = 0;
16203 1062 grub_free (cargs->key_data);
16204 1063 }
16205 >>> CID 366905: Memory - illegal accesses (USE_AFTER_FREE)
16206 >>> Using freed pointer "dev".
16207 1064 return dev;
16208 1065 }
16209 1066
16210 1067 #ifdef GRUB_UTIL
16211 1068 #include <grub/util/misc.h>
16212 1069 grub_err_t
16213
16214 Here the "dev" variable can point to a freed cryptodisk device if the
16215 function grub_cryptodisk_insert() fails. This can happen only on a OOM
16216 condition, but when this happens grub_cryptodisk_insert() calls grub_free on
16217 the passed device. Since grub_cryptodisk_scan_device_real() assumes that
16218 grub_cryptodisk_insert() is always successful, it will return the device,
16219 though the device was freed.
16220
16221 Change grub_cryptodisk_insert() to not free the passed device on failure.
16222 Then on grub_cryptodisk_insert() failure, free the device pointer. This is
16223 done by going to the label "error", which will call cryptodisk_close() to
16224 free the device and set the device pointer to NULL, so that a pointer to
16225 freed memory is not returned.
16226
16227 Fixes: CID 366905
16228
16229 Reviewed-by: Daniel Kiper <[email protected]>
16230
16231 2021-12-23 Daniel Axtens <[email protected]>
16232
16233 mm: Document grub_mm_init_region()
16234 The grub_mm_init_region() does some things that seem magical, especially
16235 around region merging. Make it a bit clearer.
16236
16237 Reviewed-by: Daniel Kiper <[email protected]>
16238
16239 2021-12-23 Daniel Axtens <[email protected]>
16240
16241 mm: Document grub_free()
16242 The grub_free() possesses a surprising number of quirks, and also
16243 uses single-letter variable names confusingly to iterate through
16244 the free list.
16245
16246 Document what's going on.
16247
16248 Use prev and cur to iterate over the free list.
16249
16250 Reviewed-by: Daniel Kiper <[email protected]>
16251
16252 2021-12-23 Daniel Axtens <[email protected]>
16253
16254 mm: grub_real_malloc(): Make small allocs comment match code
16255 Small allocations move the region's *first pointer. The comment
16256 says that this happens for allocations under 64K. The code says
16257 it's for allocations under 32K. Commit 45bf8b3a7549 changed the
16258 code intentionally: make the comment match.
16259
16260 Fixes: 45bf8b3a7549 (* grub-core/kern/mm.c (grub_real_malloc): Decrease cut-off of moving the)
16261
16262 Reviewed-by: Daniel Kiper <[email protected]>
16263
16264 2021-12-23 Daniel Axtens <[email protected]>
16265
16266 mm: Clarify grub_real_malloc()
16267 When iterating through the singly linked list of free blocks,
16268 grub_real_malloc() uses p and q for the current and previous blocks
16269 respectively. This isn't super clear, so swap to using prev and cur.
16270
16271 This makes another quirk more obvious. The comment at the top of
16272 grub_real_malloc() might lead you to believe that the function will
16273 allocate from *first if there is space in that block.
16274
16275 It actually doesn't do that, and it can't do that with the current
16276 data structures. If we used up all of *first, we would need to change
16277 the ->next of the previous block to point to *first->next, but we
16278 can't do that because it's a singly linked list and we don't have
16279 access to *first's previous block.
16280
16281 What grub_real_malloc() actually does is set *first to the initial
16282 previous block, and *first->next is the block we try to allocate
16283 from. That allows us to keep all the data structures consistent.
16284
16285 Document that.
16286
16287 Reviewed-by: Daniel Kiper <[email protected]>
16288
16289 2021-12-23 Daniel Axtens <[email protected]>
16290
16291 mm: Document GRUB internal memory management structures
16292 I spent more than a trivial quantity of time figuring out pre_size and
16293 whether a memory region's size contains the header cell or not.
16294
16295 Document the meanings of all the properties. Hopefully now no-one else
16296 has to figure it out!
16297
16298 Reviewed-by: Daniel Kiper <[email protected]>
16299
16300 2021-12-23 Michael Chang <[email protected]>
16301
16302 fs/btrfs: Use full btrfs bootloader area
16303 Up to now GRUB can only embed to the first 64 KiB before primary
16304 superblock of btrfs, effectively limiting the GRUB core size. That
16305 could consequently pose restrictions to feature enablement like
16306 advanced zstd compression.
16307
16308 This patch attempts to utilize full unused area reserved by btrfs for
16309 the bootloader outlined in the document [1]:
16310
16311 The first 1MiB on each device is unused with the exception of primary
16312 superblock that is on the offset 64KiB and spans 4KiB.
16313
16314 Apart from that, adjacent sectors to superblock and first block group
16315 are not used for embedding in case of overflow and logged access to
16316 adjacent sectors could be useful for tracing it up.
16317
16318 This patch has been tested to provide out of the box support for btrfs
16319 zstd compression with which GRUB has been installed to the partition.
16320
16321 [1] https://btrfs.wiki.kernel.org/index.php/Manpage/btrfs(5)#BOOTLOADER_SUPPORT
16322
16323 Reviewed-by: Daniel Kiper <[email protected]>
16324
16325 2021-12-23 Glenn Washburn <[email protected]>
16326
16327 tests: Refactor building xorriso command for iso9660 tests
16328 The iso9660 tests test creating isos with different combinations of
16329 Joliet, Rock Ridge, and ISO 9660 conformance level. Refactor xorriso
16330 argument generation for more readability and extensibility.
16331
16332 Reviewed-by: Thomas Schmitt <[email protected]>
16333 Reviewed-by: Daniel Kiper <[email protected]>
16334
16335 2021-12-23 Glenn Washburn <[email protected]>
16336
16337 cryptodisk: Improve handling of partition name in cryptomount password prompt
16338 Call grub_partition_get_name() unconditionally to initialize the part
16339 variable. Then part will only be NULL when grub_partition_get_name() errors.
16340 Note that when source->partition is NULL, then grub_partition_get_name()
16341 returns an allocated empty string. So no comma or partition will be printed,
16342 as desired.
16343
16344 Reviewed-by: Daniel Kiper <[email protected]>
16345
16346 2021-12-23 Glenn Washburn <[email protected]>
16347
16348 cryptodisk: Move global variables into grub_cryptomount_args struct
16349 Note that cargs.search_uuid does not need to be initialized in various parts
16350 of the cryptomount argument parsing, just once when cargs is declared with
16351 a struct initializer. The previous code used a global variable which would
16352 retain the value across cryptomount invocations.
16353
16354 Reviewed-by: Daniel Kiper <[email protected]>
16355
16356 2021-12-23 Glenn Washburn <[email protected]>
16357
16358 cryptodisk: Refactor password input out of crypto dev modules into cryptodisk
16359 The crypto device modules should only be setting up the crypto devices and
16360 not getting user input. This has the added benefit of simplifying the code
16361 such that three essentially duplicate pieces of code are merged into one.
16362
16363 Add documentation of passphrase option for cryptomount as it is now usable.
16364
16365 Reviewed-by: Daniel Kiper <[email protected]>
16366
16367 2021-12-23 Glenn Washburn <[email protected]>
16368
16369 cryptodisk: Add infrastructure to pass data from cryptomount to cryptodisk modules
16370 Previously, the cryptomount arguments were passed by global variable and
16371 function call argument, neither of which are ideal. This change passes data
16372 via a grub_cryptomount_args struct, which can be added to over time as
16373 opposed to continually adding arguments to the cryptodisk scan and
16374 recover_key.
16375
16376 As an example, passing a password as a cryptomount argument is implemented.
16377 However, the backends are not implemented, so testing this will return a not
16378 implemented error.
16379
16380 Also, add comments to cryptomount argument parsing to make it more obvious
16381 which argument states are being handled.
16382
16383 Reviewed-by: Daniel Kiper <[email protected]>
16384
16385 2021-12-23 Glenn Washburn <[email protected]>
16386
16387 cryptodisk: Improve cryptomount -u error message
16388 When a cryptmount is specified with a UUID, but no cryptodisk backends find
16389 a disk with that UUID, return a more detailed message giving telling the
16390 user that they might not have a needed cryptobackend module loaded.
16391
16392 Reviewed-by: Daniel Kiper <[email protected]>
16393
16394 2021-12-23 Glenn Washburn <[email protected]>
16395
16396 cryptodisk: Improve error messaging in cryptomount invocations
16397 Update such that "cryptomount -u UUID" will not print two error messages
16398 when an invalid passphrase is given and the most relevant error message
16399 will be displayed.
16400
16401 Reviewed-by: Daniel Kiper <[email protected]>
16402
16403 2021-12-23 Glenn Washburn <[email protected]>
16404
16405 cryptodisk: Return failure in cryptomount when no cryptodisk modules are loaded
16406 This displays an error notifying the user that they'll want to load
16407 a backend module to make cryptomount useful.
16408
16409 Reviewed-by: Daniel Kiper <[email protected]>
16410
16411 2021-12-23 Glenn Washburn <[email protected]>
16412
16413 cryptodisk: Refactor to discard have_it global
16414 The global "have_it" was never used by the crypto-backends, but was used to
16415 determine if a crypto-backend successfully mounted a cryptodisk with a given
16416 UUID. This is not needed however, because grub_device_iterate() will return
16417 1 if and only if grub_cryptodisk_scan_device() returns 1. And
16418 grub_cryptodisk_scan_device() will now only return 1 if a search_uuid has
16419 been specified and a cryptodisk was successfully setup by a crypto-backend or
16420 a cryptodisk of the requested UUID is already open.
16421
16422 To implement this grub_cryptodisk_scan_device_real() is modified to return
16423 a cryptodisk or NULL on failure and having the appropriate grub_errno set to
16424 indicated failure. Note that grub_cryptodisk_scan_device_real() will fail now
16425 with a new errno GRUB_ERR_BAD_MODULE when none of the cryptodisk backend
16426 modules succeed in identifying the source disk.
16427
16428 With this change grub_device_iterate() will return 1 when a crypto device is
16429 successfully decrypted or when the source device has already been successfully
16430 opened. Prior to this change, trying to mount an already successfully opened
16431 device would trigger an error with the message "no such cryptodisk found",
16432 which is at best misleading. The mount should silently succeed in this case,
16433 which is what happens with this patch.
16434
16435 Reviewed-by: Daniel Kiper <[email protected]>
16436
16437 2021-12-23 Glenn Washburn <[email protected]>
16438
16439 luks2: Add debug message to align with luks and geli modules
16440 Reviewed-by: Daniel Kiper <[email protected]>
16441
16442 configure: Fix misspelled variable BUILD_LDFAGS -> BUILD_LDFLAGS
16443 Reviewed-by: Daniel Kiper <[email protected]>
16444
16445 2021-12-23 Michael Chang <[email protected]>
16446
16447 grub-mkconfig: Restore umask for the grub.cfg
16448 The commit ab2e53c8a (grub-mkconfig: Honor a symlink when generating
16449 configuration by grub-mkconfig) has inadvertently discarded umask for
16450 creating grub.cfg in the process of running grub-mkconfig. The resulting
16451 wrong permission (0644) would allow unprivileged users to read GRUB
16452 configuration file content. This presents a low confidentiality risk
16453 as grub.cfg may contain non-secured plain-text passwords.
16454
16455 This patch restores the missing umask and sets the creation file mode
16456 to 0600 preventing unprivileged access.
16457
16458 Fixes: CVE-2021-3981
16459
16460 Reviewed-by: Daniel Kiper <[email protected]>
16461
16462 2021-12-23 Heinrich Schuchardt <[email protected]>
16463
16464 efi: Create the grub_efi_close_protocol() library function
16465 Create a library function for CloseProtocol() and use it for the SNP driver.
16466
16467 Reviewed-by: Daniel Kiper <[email protected]>
16468
16469 2021-12-23 Heinrich Schuchardt <[email protected]>
16470
16471 efinet: Correct closing of SNP protocol
16472 In the context of the implementation of the EFI_LOAD_FILE2_PROTOCOL for
16473 the initial ramdisk it was observed that opening the SNP protocol failed.
16474 https://lists.gnu.org/archive/html/grub-devel/2021-10/msg00020.html
16475 This is due to an incorrect call to CloseProtocol().
16476
16477 The first parameter of CloseProtocol() is the handle, not the interface.
16478
16479 We call OpenProtocol() with ControllerHandle == NULL. Hence we must also
16480 call CloseProtcol() with ControllerHandel == NULL.
16481
16482 Each call of OpenProtocol() for the same network card handle is expected to
16483 return the same interface pointer. If we want to close the protocol which
16484 we opened non-exclusively when searching for a card, we have to do this
16485 before opening the protocol exclusively.
16486
16487 As there is no guarantee that we successfully open the protocol add checks
16488 in the transmit and receive functions.
16489
16490 Reported-by: Andreas Schwab <[email protected]>
16491 Reviewed-by: Daniel Kiper <[email protected]>
16492
16493 2021-12-23 Colin Watson <[email protected]>
16494
16495 minilzo: Update to minilzo-2.10
16496 minilzo fails to build on a number of Debian release architectures
16497 (armel, mips64el, mipsel, ppc64el) with errors such as:
16498
16499 ../../grub-core/lib/minilzo/minilzo.c: In function 'lzo_memops_get_le16':
16500 ../../grub-core/lib/minilzo/minilzo.c:3479:11: error: dereferencing type-punned pointer will break strict-aliasing rules [-Werror=strict-aliasing]
16501 3479 | * (lzo_memops_TU2p) (lzo_memops_TU0p) (dd) = * (const lzo_memops_TU2p) (const lzo_memops_TU0p) (ss); \
16502 | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
16503 ../../grub-core/lib/minilzo/minilzo.c:3530:5: note: in expansion of macro 'LZO_MEMOPS_COPY2'
16504 3530 | LZO_MEMOPS_COPY2(&v, ss);
16505 | ^~~~~~~~~~~~~~~~
16506
16507 The latest upstream version is 2.10, so updating to it seems like a good
16508 idea on general principles, and it fixes builds on all the above
16509 architectures.
16510
16511 The update procedure documented in the GRUB Developers Manual worked; I
16512 just updated the version numbers to make it clear that it's been
16513 executed recently.
16514
16515 Reviewed-by: Daniel Kiper <[email protected]>
16516
16517 2021-12-23 Glenn Washburn <[email protected]>
16518
16519 docs: Add documentation on packages for building documentation
16520 Reviewed-by: Daniel Kiper <[email protected]>
16521
16522 2021-12-23 Glenn Washburn <[email protected]>
16523
16524 docs: Fix broken links in development docs
16525 Use the Git Book as a reference for documentation on Git as no other link
16526 was provided. Other links were broken because they used @url instead of
16527 @uref and needed a comma separator between link and link text.
16528
16529 Reviewed-by: Daniel Kiper <[email protected]>
16530
16531 2021-12-23 Glenn Washburn <[email protected]>
16532
16533 docs: Update development docs to include information on running test suite
16534 Add a section with minimal description on setting up and running the test
16535 suite with a link to the INSTALL documentation which is a little more
16536 detailed in terms of package requirements.
16537
16538 Reviewed-by: Daniel Kiper <[email protected]>
16539
16540 2021-12-23 Glenn Washburn <[email protected]>
16541
16542 docs: Add sentence on where Debian packages can be searched for online
16543 Reviewed-by: Daniel Kiper <[email protected]>
16544
16545 2021-12-23 Qu Wenruo <[email protected]>
16546
16547 fs/btrfs: Make extent item iteration to handle gaps
16548 The GRUB btrfs implementation can't handle two very basic btrfs
16549 file layouts:
16550
16551 1. Mixed inline/regualr extents
16552 # mkfs.btrfs -f test.img
16553 # mount test.img /mnt/btrfs
16554 # xfs_io -f -c "pwrite 0 1k" -c "sync" -c "falloc 0 4k" \
16555 -c "pwrite 4k 4k" /mnt/btrfs/file
16556 # umount /mnt/btrfs
16557 # ./grub-fstest ./grub-fstest --debug=btrfs ~/test.img hex "/file"
16558
16559 Such mixed inline/regular extents case is not recommended layout,
16560 but all existing tools and kernel can handle it without problem.
16561
16562 2. NO_HOLES feature
16563 # mkfs.btrfs -f test.img -O no_holes
16564 # mount test.img /mnt/btrfs
16565 # xfs_io -f -c "pwrite 0 4k" -c "pwrite 8k 4k" /mnt/btrfs/file
16566 # umount /mnt/btrfs
16567 # ./grub-fstest ./grub-fstest --debug=btrfs ~/test.img hex "/file"
16568
16569 NO_HOLES feature is going to be the default mkfs feature in the incoming
16570 v5.15 release, and kernel has support for it since v4.0.
16571
16572 The way GRUB btrfs code iterates through file extents relies on no gap
16573 between extents.
16574
16575 If any gap is hit, then GRUB btrfs will error out, without any proper
16576 reason to help debug the bug.
16577
16578 This is a bad assumption, since a long long time ago btrfs has a new
16579 feature called NO_HOLES to allow btrfs to skip the padding hole extent
16580 to reduce metadata usage.
16581
16582 The NO_HOLES feature is already stable since kernel v4.0 and is going to
16583 be the default mkfs feature in the incoming v5.15 btrfs-progs release.
16584
16585 When there is a extent gap, instead of error out, just try next item.
16586
16587 This is still not ideal, as kernel/progs/U-boot all do the iteration
16588 item by item, not relying on the file offset continuity.
16589
16590 But it will be way more time consuming to correct the whole behavior than
16591 starting from scratch to build a proper designed btrfs module for GRUB.
16592
16593 Reviewed-by: Daniel Kiper <[email protected]>
16594
16595 2021-11-22 Alec Brown <[email protected]>
16596
16597 disk/ldm: Fix resource leak
16598 Commit 23e39f50ca7a (disk/ldm: Make sure comp data is freed before exiting from
16599 make_vg()) fixed several spots in make_vg() where comp data was leaking memory
16600 when an error was being handled but missed one. To avoid leaking memory, comp
16601 should be freed when an error is being handled after comp has been successfully
16602 allocated memory in the for loop.
16603
16604 Fixes: 23e39f50ca7a (disk/ldm: Make sure comp data is freed before exiting from make_vg())
16605 Fixes: CID 73804
16606
16607 Reviewed-by: Daniel Kiper <[email protected]>
16608
16609 2021-11-22 Alec Brown <[email protected]>
16610
16611 commands/probe: Fix resource leaks
16612 Commit 1fc860bb76bb (commands/probe: Fix a resource leak when probing disks),
16613 missed other cases where grub_device_close() should be called before a return
16614 statement is called. Also found that grub_disk_close() wasn't being called when
16615 an error is being returned. To avoid conflict with grub_errno, grub_error_push()
16616 should be called before either grub_device_close() or grub_disk_close() is
16617 called and grub_error_pop() should be called before grub_errno is returned.
16618
16619 Fixes: 1fc860bb76bb (commands/probe: Fix a resource leak when probing disks)
16620 Fixes: CID 292443
16621
16622 Reviewed-by: Daniel Kiper <[email protected]>
16623
16624 2021-11-22 Michael Chang <[email protected]>
16625
16626 templates: Filter out POSIX locale for translation
16627 The POSIX locale is default or native operating system's locale
16628 identical to the C locale, so no translation to human speaking languages
16629 are provided. For this reason we should filter out LANG=POSIX as well as
16630 LANG=C upon generating grub.cfg to avoid looking up for it's gettext's
16631 message catalogs that will consequently result in an unpleasant message:
16632
16633 error: file `/boot/grub/locale/POSIX.gmo' not found
16634
16635 Reviewed-by: Daniel Kiper <[email protected]>
16636
16637 2021-11-02 Darren Kenny <[email protected]>
16638
16639 io/gzio: Fix possible use of uninitialized variable in huft_build()
16640 In huft_build() it is possible to reach the for loop where "r" is being
16641 assigned to "q[j]" without "r.v" ever being initialized.
16642
16643 Fixes: CID 314024
16644
16645 Reviewed-by: Daniel Kiper <[email protected]>
16646
16647 2021-11-02 Darren Kenny <[email protected]>
16648
16649 fs/zfs/zfs: Fix possible insecure use of chunk size in zap_leaf_array_get()
16650 In zap_leaf_array_get() the chunk size passed in is considered tainted
16651 by Coverity, and is being used before it is tested for validity. To fix
16652 this the assignment of "la" is moved until after the test of the value
16653 of "chunk".
16654
16655 Fixes: CID 314014
16656
16657 Reviewed-by: Daniel Kiper <[email protected]>
16658
16659 2021-11-02 Darren Kenny <[email protected]>
16660
16661 util/grub-mkfont: Fix memory leak in write_font_pf2()
16662 In the function write_font_pf2() memory is allocated for font_name to
16663 construct a new name, but it is not released before returning from the
16664 function, leaking the allocated memory.
16665
16666 Fixes: CID 314015
16667
16668 Reviewed-by: Daniel Kiper <[email protected]>
16669
16670 2021-11-02 Darren Kenny <[email protected]>
16671
16672 util/grub-fstest: Fix resource leaks in cmd_cmp()
16673 In the function cmd_cmp() within the while loop, srcnew and destnew are
16674 being allocated but are never freed either before leaving scope or in
16675 the recursive calls being made to cmd_cmp().
16676
16677 Fixes: CID 314032
16678 Fixes: CID 314045
16679
16680 Reviewed-by: Daniel Kiper <[email protected]>
16681
16682 2021-11-02 Darren Kenny <[email protected]>
16683
16684 util/grub-mkrescue: Fix memory leak in write_part()
16685 In the function write_part(), the value of inname is not used beyond
16686 the grub_util_fopen() call, so it should be freed to avoid leakage.
16687
16688 Fixes: CID 314028
16689
16690 Reviewed-by: Daniel Kiper <[email protected]>
16691
16692 2021-11-02 Darren Kenny <[email protected]>
16693
16694 util/grub-install-common: Fix memory leak in copy_all()
16695 The copy_all() function skips a section of code using continue, but
16696 fails to free the memory in srcf first, leaking it.
16697
16698 Fixes: CID 314026
16699
16700 Reviewed-by: Daniel Kiper <[email protected]>
16701
16702 2021-11-02 Robbie Harwood <[email protected]>
16703
16704 kern/dl: Print module name on license check failure
16705 Prior to this change, the GRUB would only indicate that the check had
16706 been failed, but not by what module. This made it difficult to track
16707 down either the problem module, or debug the false positive further.
16708
16709 Before performing the license check, resolve the module name so that
16710 it can be printed if the license check fails.
16711
16712 Reviewed-by: Daniel Kiper <[email protected]>
16713
16714 2021-10-25 Glenn Washburn <[email protected]>
16715
16716 kern/misc: Add debug log condition to log output
16717 Adding the conditional to debug log messages allows the GRUB user to
16718 construct the $debug variable without needing to consult the source to
16719 find the conditional (especially useful for situations where the source
16720 is not readily available).
16721
16722 Reviewed-by: Daniel Kiper <[email protected]>
16723
16724 2021-10-25 Glenn Washburn <[email protected]>
16725
16726 tests: In partmap_test, use ${parted} variable when checking for binary
16727 Reviewed-by: Daniel Kiper <[email protected]>
16728
16729 2021-10-25 Glenn Washburn <[email protected]>
16730
16731 tests: Test aborts due to missing requirements should be marked as error instead of skipped
16732 Many tests abort due to not being root or missing tools, for instance mkfs
16733 commands for file system tests. The tests are exited with code 77, which
16734 means they were skipped. A skipped test is a test that should not be run,
16735 e.g. a test specific to ARM64 should not be run on an x86 build. These aborts
16736 are actually a hard error, code 99. That means that the test could not be
16737 completed, but not because what was supposed to be tested failed, e.g. in
16738 these cases where a missing tool prevents the running of a test.
16739
16740 Reviewed-by: Daniel Kiper <[email protected]>
16741
16742 2021-10-25 Glenn Washburn <[email protected]>
16743
16744 tests: Boot PowerPC using PMU instead of CUDA for power management
16745 A recent refactoring of CUDA command code has exposed a bug in OpenBIOS [1]
16746 which was causing system powerdown and system reset to fail, thus causing
16747 the QEMU instance to hang. This in turn caused the grub-shell command to
16748 timeout causing it to return an error code when the test actually completed
16749 successfully.
16750
16751 Since it could be a while before the patch fixing this issue in OpenBIOS
16752 filters down to the average distro, switch to PMU to allow powerdowns and
16753 reboots to work as expected.
16754
16755 [1] https://gitlab.com/qemu-project/qemu/-/issues/624
16756
16757 Reviewed-by: Daniel Kiper <[email protected]>
16758
16759 2021-10-14 Kees Cook <[email protected]>
16760
16761 osdep/linux: Fix md array device enumeration
16762 GET_ARRAY_INFO's info.nr_disks does not map to GET_DISK_INFO's
16763 disk.number, which is an internal kernel index. If an array has had drives
16764 added, removed, etc., there may be gaps in GET_DISK_INFO's results. But
16765 since the consumer of devicelist cannot tolerate gaps (it expects to walk
16766 a NULL-terminated list of device name strings), the devicelist index (j)
16767 must be tracked separately from the disk.number index (i).
16768
16769 As part of this, since GRUB wants to only examine active (i.e. present
16770 and non-failed) disks, the count of remaining disks (remaining) must be
16771 tracked separately from the devicelist index (j).
16772
16773 Additionally, drop a line with empty spaces only.
16774
16775 Fixes: 49de079bbe1c (... (grub_util_raid_getmembers): Handle "removed" disks)
16776 Fixes: 2b00217369ac (... Added support for RAID and LVM)
16777 Fixes: https://bugs.launchpad.net/ubuntu/+source/grub2/+bug/1912043
16778 Fixes: https://savannah.gnu.org/bugs/index.php?59887
16779
16780 Reviewed-by: Petr Vorel <[email protected]>
16781 Reviewed-by: Daniel Kiper <[email protected]>
16782
16783 2021-10-14 Glenn Washburn <[email protected]>
16784
16785 docs: Add fuller accounting of "make check" prerequisites
16786 Many of the prerequisites for exercising the full "make check" test suite
16787 have not been documented. This adds them along with a note that some tests
16788 require elevated privileges to run.
16789
16790 Add an incomplete list of cross compiling toolchain packages for Debian
16791 and trusted sources for other distros.
16792
16793 Add statement at the start of the document to clarify that package names
16794 are from Debian 11.
16795
16796 Reviewed-by: Daniel Kiper <[email protected]>
16797
16798 2021-10-14 Glenn Washburn <[email protected]>
16799
16800 tests: Do not delete filesystem images on error
16801 The filesystem images created for the filesystem test can be useful when
16802 debugging why a filesystem test failed. So, keep them around and let the
16803 user clean them up.
16804
16805 Reviewed-by: Daniel Kiper <[email protected]>
16806
16807 2021-10-14 Glenn Washburn <[email protected]>
16808
16809 tests: Output list of devices when partmap fails
16810 Reviewed-by: Daniel Kiper <[email protected]>
16811
16812 2021-10-14 Glenn Washburn <[email protected]>
16813
16814 tests: Skip HFS test only when mac_roman module is not loaded and not loadable
16815 Allow the HFS tests to not be skipped if the mac_roman modules is loaded in
16816 the kernel, but not accessible to modprobe.
16817
16818 Reviewed-by: Daniel Kiper <[email protected]>
16819
16820 2021-10-14 Glenn Washburn <[email protected]>
16821
16822 tests: Change FAT volume label to be with in the valid character range
16823 The ";", semi-colon, character is not a valid character for a FAT filesystem
16824 label. This test used to succeed because prior to v4.2 of dosfstools
16825 mkfs.vfat did not enforce the character restrictions for volume labels. So,
16826 change the volume label string to be valid but contain symbol characters to
16827 test odd volume labels.
16828
16829 Reviewed-by: Daniel Kiper <[email protected]>
16830
16831 2021-10-14 Glenn Washburn <[email protected]>
16832
16833 tests: Only test MINIX3 volumes of 1 KiB block size
16834 Apparently there used to be a -B option for mkfs.minix to create a volume
16835 with a specified block size. This version is hard to come by and does not
16836 appear to be available in Debian distributions. So, remove support for
16837 testing a variety of blocks sizes for MINIX3. This allows the MINIX tests
16838 to run because they were being skipped due to not finding a mkfs.minix with
16839 the -B option.
16840
16841 Reviewed-by: Daniel Kiper <[email protected]>
16842
16843 2021-10-14 Glenn Washburn <[email protected]>
16844
16845 tests: mkfs.btrfs now supports only 4 KiB sector sizes and above
16846 Reviewed-by: Daniel Kiper <[email protected]>
16847
16848 tests: Disable ReiserFS tests for old format because newer kernels do not support them
16849 Reviewed-by: Daniel Kiper <[email protected]>
16850
16851 tests: mkreiserfs only supports 4096 block size
16852 Reviewed-by: Daniel Kiper <[email protected]>
16853
16854 tests: Rename variable filtime -> filetime as its meant to be
16855 Reviewed-by: Daniel Kiper <[email protected]>
16856
16857 2021-10-14 Glenn Washburn <[email protected]>
16858
16859 tests: Use @BUILD_SHEBANG@ autoconf var instead of literal shell
16860 This bring this test in line with the rest of the test scripts.
16861
16862 Reviewed-by: Daniel Kiper <[email protected]>
16863
16864 2021-10-14 Glenn Washburn <[email protected]>
16865
16866 tests: Exit with skipped exit code when test not performed
16867 These tests were not performed and therefore did not pass, nor fail. This
16868 fixes misleading test exit code where, for instance, the pseries_test will
16869 pass on i386-pc, which is not a pseries architecture.
16870
16871 Reviewed-by: Daniel Kiper <[email protected]>
16872
16873 2021-10-14 Glenn Washburn <[email protected]>
16874
16875 tests: A failure of mktemp should cause the test script to exit with code 99
16876 A test exiting with code 99 means that there was an error in the test itself
16877 and not a failure in the thing being tested (also known as a hard error).
16878
16879 Reviewed-by: Daniel Kiper <[email protected]>
16880
16881 2021-10-14 Glenn Washburn <[email protected]>
16882
16883 tests: Make setup errors in grub-fs-tester hard errors
16884 When a test program fails because it failed to setup the test properly, this
16885 does not indicate a failure in what is attempting to be tested because the
16886 test is never run. So exit with a hard error exit status to note this
16887 difference. This will allow easier detection of tests that are not actually
16888 being run and those that are really failing.
16889
16890 Reviewed-by: Daniel Kiper <[email protected]>
16891
16892 2021-10-14 Glenn Washburn <[email protected]>
16893
16894 tests: Do not occlude grub-shell return code
16895 The script grub-shell does the bulk of the testing. If it returns an error
16896 code, that means that the test failed and the test should immediately exit
16897 with that error code. When grub-shell is used as a non-terminating command
16898 in a pipeline, e.g. when data needs to be extracted from its output, its
16899 error code will be occluded by the last command in the pipeline. Refactor
16900 tests so that the shell will error with the exit code of grub-shell by
16901 breaking up pipelines such that grub-shell is always the last command in
16902 the pipeline that it is used in.
16903
16904 Reviewed-by: Daniel Kiper <[email protected]>
16905
16906 2021-10-14 Glenn Washburn <[email protected]>
16907
16908 tests: Do not occlude subshell error codes when used as input to the test command
16909 When using the output of a subshell as input, its error code is ignored in
16910 the context of "set -e". Many test scripts use grub-shell in a subshell with
16911 output used as an argument to the test command to test for expected output.
16912 Refactor these tests so that the subshell output goes to a shell variable,
16913 so that if the subshell errors the script will immediately exit with an
16914 error code.
16915
16916 Reviewed-by: Daniel Kiper <[email protected]>
16917
16918 2021-10-14 Glenn Washburn <[email protected]>
16919
16920 tests: Add set -e to missing tests
16921 This helps to ensure that error codes do not get ignored.
16922
16923 Reviewed-by: Daniel Kiper <[email protected]>
16924
16925 2021-10-14 Glenn Washburn <[email protected]>
16926
16927 tests: When checking squashfs fstime, use superblock last modified time
16928 Currently, the filesystem timestamp check in grub-fs-tester uses the
16929 squashfs image file's last modified timestamp and checks to see if that
16930 time stamp is within 3 seconds of the superblock timestamp as determined by
16931 grub. The image file's timestamp could be more than 3 seconds off if
16932 mksquashfs takes more than 3 seconds to generate the image, as is the case
16933 on a virtual machine. Instead use squashfs tools to get the filesystem
16934 timestamp directly.
16935
16936 Reviewed-by: Daniel Kiper <[email protected]>
16937
16938 2021-10-14 Glenn Washburn <[email protected]>
16939
16940 tests: Fix partmap_test for arm*-efi, disk numbering has changed
16941 Perhaps using a newer UEFI firmware is the reason for the created test disk
16942 showing up as hd2 instead of hd3.
16943
16944 Reviewed-by: Daniel Kiper <[email protected]>
16945
16946 2021-10-04 Nikolai Kostrigin <[email protected]>
16947
16948 docs/grub-dev: Fix typos
16949 Reviewed-by: Daniel Kiper <[email protected]>
16950
16951 2021-10-04 Michael Chang <[email protected]>
16952
16953 build: Fix build error with binutils 2.36
16954 The following procedure to build xen/pvgrub is broken.
16955
16956 git clone https://git.savannah.gnu.org/git/grub.git
16957 cd grub
16958 ./bootstrap
16959 mkdir build-xen
16960 cd build-xen
16961 ../configure --with-platform=xen
16962 make
16963
16964 It fails with the message:
16965
16966 /usr/lib64/gcc/x86_64-suse-linux/10/../../../../x86_64-suse-linux/bin/ld:
16967 section .note.gnu.property VMA [0000000000400158,0000000000400187]
16968 overlaps section .bss VMA [000000000000f000,000000000041e1af]
16969
16970 The most significant factor is that new assembler (GNU as) generates the
16971 .note.gnu.property section as default. This note section overlaps with
16972 .bss because it doesn't reposition with -Wl,-Ttext,0 with which the base
16973 address of .text section is set, rather the address of .note.gnu.property
16974 is calculated for some reason from 0x400000 where the ELF executable
16975 defaults to start.
16976
16977 Using -Ttext-segment doesn't help either, though it is said to set the
16978 address of the first byte of the text segment according to "man ld".
16979 What it actually does is to override the default 0x400000, aka the image
16980 base address, to something else. The entire process can be observed in
16981 the default linker script used by gcc [1]. Therefore we can't expect it
16982 to achieve the same thing as -Ttext given that the first segment where
16983 .text resides is offset by SIZEOF_HEADERS plus some sections may be
16984 preceding it within the first segment. The end result is .text always
16985 has to start with non-zero address with -Wl,-Ttext-segment,0 if using
16986 default linker script.
16987
16988 It is also worth mentioning that binutils upstream apparently doesn't
16989 seem to consider this as a bug [2] and proposed to use -Wl,-Ttext-segment,0
16990 which is not fruitful as what has been tested by Gentoo [3].
16991
16992 As long as GRUB didn't use ISA information encoded in .note.gnu.property,
16993 we can safely drop it via -Wa,-mx86-used-note=no assembler option to
16994 fix the linker error above.
16995
16996 This is considered a better approach than using custom linker script to
16997 drop the .note.gnu.property section because object file manipulation can
16998 also be hampered one way or the other in that linker script may not be
16999 helpful. See also this commit removing the section in the process of objcopy.
17000
17001 6643507ce build: Fix GRUB i386-pc build with Ubuntu gcc
17002
17003 [1] In /usr/lib64/ldscripts/elf_x86_64.x or use 'gcc -Wl,--verbose ...'
17004 PROVIDE (__executable_start = SEGMENT_START("text-segment", 0x400000));
17005 . = SEGMENT_START("text-segment", 0x400000) + SIZEOF_HEADERS;
17006 [2] https://sourceware.org/bugzilla/show_bug.cgi?id=27377
17007 [3] https://bugs.gentoo.org/787221
17008
17009 Reviewed-by: Daniel Kiper <[email protected]>
17010
17011 2021-10-04 Michael Chang <[email protected]>
17012
17013 disk/diskfilter: Use nodes in logical volume's segment as member device
17014 Currently the grub_diskfilter_memberlist() function returns all physical
17015 volumes added to a volume group to which a logical volume (LV) belongs.
17016 However, this is suboptimal as it doesn't fit the intended behavior of
17017 returning underlying devices that make up the LV. To give a clear
17018 picture, the result should be identical to running commands below to
17019 display the logical volumes with underlying physical volumes in use.
17020
17021 localhost:~ # lvs -o lv_name,vg_name,devices /dev/system/root
17022 LV VG Devices
17023 root system /dev/vda2(512)
17024
17025 localhost:~ # lvdisplay --maps /dev/system/root
17026 --- Logical volume ---
17027 ...
17028 --- Segments ---
17029 Logical extents 0 to 4604:
17030 Type linear
17031 Physical volume /dev/vda2
17032 Physical extents 512 to 5116
17033
17034 As shown above, we can know system-root LV uses only /dev/vda2 to
17035 allocate it's extents, or we can say that /dev/vda2 is the member device
17036 comprising the system-root LV.
17037
17038 It is important to be precise on the member devices, because that helps
17039 to avoid pulling in excessive dependency. Let's use an example to
17040 demonstrate why it is needed.
17041
17042 localhost:~ # findmnt /
17043 TARGET SOURCE FSTYPE OPTIONS
17044 / /dev/mapper/system-root ext4 rw,relatime
17045
17046 localhost:~ # pvs
17047 PV VG Fmt Attr PSize PFree
17048 /dev/mapper/data system lvm2 a-- 1020.00m 0
17049 /dev/vda2 system lvm2 a-- 19.99g 0
17050
17051 localhost:~ # cryptsetup status /dev/mapper/data
17052 /dev/mapper/data is active and is in use.
17053 type: LUKS1
17054 cipher: aes-xts-plain64
17055 keysize: 512 bits
17056 key location: dm-crypt
17057 device: /dev/vdb
17058 sector size: 512
17059 offset: 4096 sectors
17060 size: 2093056 sectors
17061 mode: read/write
17062
17063 localhost:~ # vgs
17064 VG #PV #LV #SN Attr VSize VFree
17065 system 2 3 0 wz--n- 20.98g 0
17066
17067 localhost:~ # lvs -o lv_name,vg_name,devices
17068 LV VG Devices
17069 data system /dev/mapper/data(0)
17070 root system /dev/vda2(512)
17071 swap system /dev/vda2(0)
17072
17073 We can learn from above that /dev/mapper/data is an encrypted volume and
17074 also gets assigned to volume group "system" as one of it's physical
17075 volumes. And also it is not used by root device, /dev/mapper/system-root,
17076 for allocating extents, so it shouldn't be taking part in the process of
17077 setting up GRUB to access root device.
17078
17079 However, running grub-install reports error as volume group "system"
17080 contains encrypted volume.
17081
17082 error: attempt to install to encrypted disk without cryptodisk
17083 enabled. Set `GRUB_ENABLE_CRYPTODISK=y' in file `/etc/default/grub'.
17084
17085 Certainly we can enable GRUB_ENABLE_CRYPTODISK=y and move on, but that
17086 is not always acceptable since the server may need to be booted unattended.
17087 Additionally, typing passphrase for every system startup can be a big
17088 hassle of which most users would like to avoid.
17089
17090 This patch solves the problem by returning exact physical volume, /dev/vda2,
17091 rightly used by system-root from the example above, thus grub-install will
17092 not error out because the excessive encrypted device to boot the root device
17093 is not configured.
17094
17095 Tested-by: Olav Reinert <[email protected]>
17096 Reviewed-by: Daniel Kiper <[email protected]>
17097
17098 2021-10-04 Krzysztof Nowicki <[email protected]>
17099
17100 fs/ext2: Fix handling of missing sparse extent leafs
17101 When a file on ext4 is stored as sparse the data belonging to
17102 zero-filled blocks is not written to storage and the extent map is
17103 missing entries for these blocks. Such case can happen both for depth
17104 0 extents (leafs) as well as higher-level tables.
17105
17106 Consider a scenario of a file which has a zero-filled beginning (e.g.
17107 ISO image). In such case real data starts at block 8. If such a file is
17108 stored using 2-level extent structure the extent list in the inode will
17109 be depth 1 and will have an entry to a depth 0 (leaf) extent header for
17110 blocks 8-n.
17111
17112 Unfortunately existing GRUB2 ext2 driver is only able to handle missing
17113 entries in leaf extent tables, for which the grub_ext2_read_block()
17114 function returns 0. In case the whole leaf extent list is missing for
17115 a block the function fails with "invalid extent" error.
17116
17117 The fix for this problem relies on the grub_ext4_find_leaf() helper
17118 function to distinguish two error cases: missing extent and error
17119 walking through the extent tree. The existing error message is raised
17120 only for the latter case, while for the missing leaf extent zero is
17121 returned from grub_ext2_read_block() indicating a sparse block.
17122
17123 Reviewed-by: Daniel Kiper <[email protected]>
17124
17125 2021-10-04 Daniel Axtens <[email protected]>
17126
17127 powerpc: Drop Open Hack'Ware - remove GRUB_IEEE1275_FLAG_NO_ANSI
17128 Open Hack'Ware was the only user.
17129
17130 Reviewed-by: Daniel Kiper <[email protected]>
17131
17132 2021-10-04 Daniel Axtens <[email protected]>
17133
17134 powerpc: Drop Open Hack'Ware - remove GRUB_IEEE1275_FLAG_CANNOT_INTERPRET
17135 Open Hack'Ware was the only user.
17136
17137 Reviewed-by: Daniel Kiper <[email protected]>
17138
17139 2021-10-04 Daniel Axtens <[email protected]>
17140
17141 powerpc: Drop Open Hack'Ware - remove GRUB_IEEE1275_FLAG_CANNOT_SET_COLORS
17142 Open Hack'Ware was the only user.
17143
17144 Reviewed-by: Daniel Kiper <[email protected]>
17145
17146 2021-10-04 Daniel Axtens <[email protected]>
17147
17148 powerpc: Drop Open Hack'Ware - remove GRUB_IEEE1275_FLAG_FORCE_CLAIM
17149 Open Hack'Ware was the only user. It added a lot of complexity.
17150
17151 Reviewed-by: Daniel Kiper <[email protected]>
17152
17153 2021-10-04 Daniel Axtens <[email protected]>
17154
17155 powerpc: Drop Open Hack'Ware
17156 Open Hack'Ware was an alternative firmware of powerpc under QEMU.
17157
17158 The last commit to any Open Hack'Ware repo I can find is from 2014 [1].
17159
17160 Open Hack'Ware was used for the QEMU "prep" machine type, which was
17161 deprecated in QEMU in commit 54c86f5a4844 (hw/ppc: deprecate the
17162 machine type 'prep', replaced by '40p') in QEMU v3.1, and had reportedly
17163 been broken for years before without anyone noticing. Support was removed
17164 in February 2020 by commit b2ce76a0730e (hw/ppc/prep: Remove the
17165 deprecated "prep" machine and the OpenHackware BIOS).
17166
17167 Open Hack'Ware's limitations require some messy code in GRUB. This
17168 complexity is not worth carrying any more.
17169
17170 Remove detection of Open Hack'Ware. We will clean up the feature flags
17171 in following commits.
17172
17173 [1]: https://github.com/qemu/openhackware and
17174 https://repo.or.cz/w/openhackware.git are QEMU submodules. They have
17175 only small changes on top of OHW v0.4.1, which was imported into
17176 QEMU SCM in 2010. I can't find anything resembling an official repo
17177 any more.
17178
17179 Reviewed-by: Daniel Kiper <[email protected]>
17180
17181 2021-09-20 Glenn Washburn <[email protected]>
17182
17183 docs/grub: Improve search documentation, by adding short options and section on hints
17184 Reviewed-by: Daniel Kiper <[email protected]>
17185
17186 2021-09-20 Glenn Washburn <[email protected]>
17187
17188 fs/udf: Fix regression which is preventing symlink access
17189 This code was broken by commit 3f05d693 (malloc: Use overflow checking
17190 primitives where we do complex allocations), which added overflow
17191 checking in many areas. The problem here is that the changes update the
17192 local variable sz, which was already in use and which was not updated
17193 before the change. So the code using sz was getting a different value of
17194 than it would have previously for the same UDF image. This causes the
17195 logic getting the destination of the symlink to not realize that its
17196 gotten the full destination, but keeps trying to read past the end of
17197 the destination. The bytes after the end are generally NULL padding
17198 bytes, but that's not a valid component type (ECMA-167 14.16.1.1). So
17199 grub_udf_read_symlink() branches to error logic, returning NULL, instead
17200 of the symlink destination path.
17201
17202 The result of this bug is that the UDF filesystem tests were failing in
17203 the symlink test with the grub-fstest error message:
17204
17205 grub-fstest: error: cannot open `(loop0)/sym': invalid symlink.
17206
17207 This change stores the result of doubling sz in another local variable s,
17208 so as not to modify sz. Also remove unnecessary grub_add(), which increased
17209 the output by 1, presumably to account for a NULL byte. This isn't needed
17210 because an output buffer of size twice sz is already guaranteed to be more
17211 than enough to contain the path components converted to UTF-8. The value of
17212 sz contains at least 4 bytes for the path component header (ECMA-167 14.16.1),
17213 which means that 2 * 4 bytes are allocated but will not be used for UTF-8
17214 characters, so the NULL byte is accounted for.
17215
17216 Reviewed-by: Daniel Kiper <[email protected]>
17217
17218 2021-09-20 Chris Vogel <[email protected]>
17219
17220 templates: Add GRUB_CMDLINE_LINUX_RECOVERY
17221 When generating grub.cfg using grub-mkconfig and the scripts 10_linux and
17222 20_linux_xen there is no way to add kernel command line parameters _only_ to
17223 the recovery entries generated.
17224
17225 This is needed to e.g. start a debug shell in installations using systemd
17226 using the kernel command line parameter "systemd.debug-shell" or to recover
17227 in a system with encrypted root in situations where the decryption of the
17228 root filesystem per crypttab in the intiramfs image is broken and the recovery
17229 entry should contain information how to decrypt the rootfs (cryptopts=).
17230
17231 This patch does not change the default behaviour of the GRUB if
17232 GRUB_CMDLINE_LINUX_RECOVERY is not set.
17233
17234 If GRUB_CMDLINE_LINUX_RECOVERY is set and the generated recovery entry should
17235 include the kernel parameter "single" the parameter must be explicitly included
17236 in GRUB_CMDLINE_LINUX_RECOVERY.
17237
17238 As far as I know all credits for the idea and the initial implementation go to
17239 Kyle Ranking of Purism.
17240
17241 Reviewed-by: Daniel Kiper <[email protected]>
17242
17243 2021-09-20 Michael Chang <[email protected]>
17244
17245 emu: Fix executable stack marking
17246 The gcc by default assumes executable stack is required if the source
17247 object file doesn't have .note.GNU-stack section in place. If any of the
17248 source objects doesn't incorporate the GNU-stack note, the resulting
17249 program will have executable stack flag set in PT_GNU_STACK program
17250 header to instruct program loader or kernel to set up the executable
17251 stack when program loads to memory.
17252
17253 Usually the .note.GNU-stack section will be generated by gcc
17254 automatically if it finds that executable stack is not required. However
17255 it doesn't take care of generating .note.GNU-stack section for those
17256 object files built from assembler sources. This leads to unnecessary
17257 risk of security of exploiting the executable stack because those
17258 assembler sources don't actually require stack to be executable to work.
17259
17260 The grub-emu and grub-emu-lite are found to flag stack as executable
17261 revealed by execstack tool.
17262
17263 $ mkdir -p build-emu && cd build-emu
17264 $ ../configure --with-platform=emu && make
17265 $ execstack -q grub-core/grub-emu grub-core/grub-emu-lite
17266 X grub-core/grub-emu
17267 X grub-core/grub-emu-lite
17268
17269 This patch will add the missing GNU-stack note to the assembler source
17270 used by both utilities, therefore the result doesn't count on gcc
17271 default behavior and the executable stack is disabled.
17272
17273 $ execstack -q grub-core/grub-emu grub-core/grub-emu-lite
17274 - grub-core/grub-emu
17275 - grub-core/grub-emu-lite
17276
17277 Reviewed-by: Daniel Kiper <[email protected]>
17278
17279 2021-09-13 Thomas Schmitt <[email protected]>
17280
17281 tests: Keep grub-fs-tester ziso9660 from failing for wrong reasons
17282 The test for the ability to decompress zisofs encoded files is supposed
17283 to fail due to the lack of this ability in GRUB. But it fails early with
17284 xorriso : FAILURE : -volid: Text too long (1650 > 32)
17285 because "ziso9660" is not in the list of filesystems which accept at most
17286 32 bytes in their FSLABEL. If this is fixed, the test returns false
17287 success because the xorriso run does not produce any zisofs compressed
17288 files. The problem is in the sequence of native xorriso commands used.
17289 The command -set_filter_r applies only to the files which are already
17290 inserted into the emerging ISO filesystem. In the current sequence no
17291 files have been inserted yet by command -add when the last of two
17292 -set_filter_r commands is executed. After this is corrected, xorriso
17293 refuses to work because the global settings of command -zisofs can be
17294 made only before command -set_filter_r has attached zisofs filters to
17295 the data files in the emerging ISO. Further: A bug in xorriso causes
17296 a false warning about FSLABEL being too long for Joliet. Shortcomings
17297 of Joliet cause warnings about symbolic links. Such warnings might
17298 distract from the actual reason why the test is expected to fail.
17299
17300 So, add "ziso9660" to the 32-byte FSLABEL list.
17301
17302 Fix the xorriso run to produce compressed files which for now cause
17303 righteous failure of the test. Do this by removing a surplus group of
17304 -set_filter_r and -zisofs commands, by moving the other such group
17305 behind -add, and by swapping -set_filter_r and -zisofs.
17306
17307 Remove the -as mkisofs options which produce a Joliet filesystem tree.
17308
17309 Reviewed-by: Daniel Kiper <[email protected]>
17310
17311 2021-09-13 Glenn Washburn <[email protected]>
17312
17313 commands/read: Add silent mode to read command to suppress input echo
17314 This conforms to the behavior of the -s option of the Bash read command.
17315
17316 docs/grub: Document the -s option for the read command.
17317
17318 Reviewed-by: Daniel Kiper <[email protected]>
17319
17320 2021-09-13 Glenn Washburn <[email protected]>
17321
17322 kern/fs: Allow number of blocks in block list to be optional, defaulting length to device length
17323 This is primarily useful to do something like "loopback newdev (dev)8+" to
17324 create a device that skips the first 4 KiB, which may contain a container
17325 header, e.g. a non-standard RAID1 header, that GRUB does not recognize. This
17326 would allow that container data to be potentially accessed up to the end of
17327 container, which may be necessary for some layouts that store data at the
17328 end. There is currently not a good way to programmatically get the number
17329 of sectors on a disk to set the appropriate length of the blocklist.
17330
17331 Reviewed-by: Daniel Kiper <[email protected]>
17332
17333 2021-09-06 Petr Vorel <[email protected]>
17334
17335 autogen.sh: Detect python
17336 It helps to avoid an error on distros which has only python3 binary:
17337 ./autogen.sh: line 20: python: command not found
17338
17339 Use python3 as the default as python2 is EOL since Jan 2020. However,
17340 check also for python which is on most distros, if not all, python2
17341 because code still works on python2.
17342
17343 Although it should not be needed keep the possibility to define PYTHON
17344 variable.
17345
17346 For detection use "command -v" which is POSIX and supported on all
17347 common shells (bash, zsh, dash, busybox sh, mksh) instead requiring
17348 "which" as an extra dependency (usable on containers).
17349
17350 Update the INSTALL file too.
17351
17352 Reviewed-by: Daniel Kiper <[email protected]>
17353
17354 2021-09-06 Petr Vorel <[email protected]>
17355
17356 bootstrap: Require GNU patch
17357 The bootstrap.conf uses patch, let's require it.
17358
17359 Better than multiple messages:
17360 ./bootstrap.conf: line 84: patch: command not found
17361
17362 Mention it also in the INSTALL file.
17363
17364 Reviewed-by: Daniel Kiper <[email protected]>
17365
17366 2021-09-06 Thomas Schmitt <[email protected]>
17367
17368 tests: Let xorriso fixely assume UTF-8 as local character set
17369 The iso9660_test fails if the effective locale is not UTF-8. This happens
17370 because xorriso needs to convert file names and FSLABEL to UCS-2 when
17371 preparing a Joliet tree. The grub-fs-tester obviously intends to use UTF-8
17372 as character set, but xorriso assumes by default the result of nl_langinfo(3)
17373 with item CODESET. So, override the result of nl_langinfo(CODESET) by options
17374 of xorriso -as mkisofs.
17375
17376 Reviewed-by: Daniel Kiper <[email protected]>
17377
17378 2021-09-06 Fangrui Song via Grub-devel <[email protected]>
17379
17380 configure: Check for -falign-jumps=1 beside -falign-loops=1
17381 The Clang does not support -falign-jumps and only recently gained support
17382 for -falign-loops. The -falign-jumps=1 should be tested beside
17383 -fliang-loops=1 to avoid passing unrecognized options to the Clang:
17384
17385 clang-14: error: optimization flag '-falign-jumps=1' is not supported [-Werror,-Wignored-optimization-argument]
17386
17387 The -falign-functions=1 is supported by GCC 5.1.0/Clang 3.8.0. So, just
17388 add the option unconditionally.
17389
17390 Acked-by: Paul Menzel <[email protected]>
17391 Reviewed-by: Daniel Kiper <[email protected]>
17392
17393 2021-09-06 Fangrui Song via Grub-devel <[email protected]>
17394
17395 configure: Remove obsoleted -malign-{jumps, loops, functions}
17396 The GCC warns "cc1: warning: ‘-malign-loops’ is obsolete, use ‘-falign-loops’".
17397 The Clang silently ignores -malign-{jumps,loops,functions}.
17398
17399 The preferred -falign-* forms have been supported since GCC 3.2. So, just
17400 remove -malign-{jumps,loops,functions}.
17401
17402 Acked-by: Paul Menzel <[email protected]>
17403 Reviewed-by: Daniel Kiper <[email protected]>
17404
17405 2021-09-06 Erwan Velu <[email protected]>
17406
17407 fs/xfs: Fix unreadable filesystem with v4 superblock
17408 The commit 8b1e5d193 (fs/xfs: Add bigtime incompat feature support)
17409 introduced the bigtime support by adding some features in v3 inodes.
17410 This change extended grub_xfs_inode struct by 76 bytes but also changed
17411 the computation of XFS_V2_INODE_SIZE and XFS_V3_INODE_SIZE. Prior this
17412 commit, XFS_V2_INODE_SIZE was 100 bytes. After the commit it's 84 bytes
17413 XFS_V2_INODE_SIZE becomes 16 bytes too small.
17414
17415 As a result, the data structures aren't properly aligned and the GRUB
17416 generates "attempt to read or write outside of partition" errors when
17417 trying to read the XFS filesystem:
17418
17419 GNU GRUB version 2.11
17420 ....
17421 grub> set debug=efi,gpt,xfs
17422 grub> insmod part_gpt
17423 grub> ls (hd0,gpt1)/
17424 partmap/gpt.c:93: Read a valid GPT header
17425 partmap/gpt.c:115: GPT entry 0: start=4096, length=1953125
17426 fs/xfs.c:931: Reading sb
17427 fs/xfs.c:270: Validating superblock
17428 fs/xfs.c:295: XFS v4 superblock detected
17429 fs/xfs.c:962: Reading root ino 128
17430 fs/xfs.c:515: Reading inode (128) - 64, 0
17431 fs/xfs.c:515: Reading inode (739521961424144223) - 344365866970255880, 3840
17432 error: attempt to read or write outside of partition.
17433
17434 This commit change the XFS_V2_INODE_SIZE computation by subtracting 76
17435 bytes instead of 92 bytes from the actual size of grub_xfs_inode struct.
17436 This 76 bytes value comes from added members:
17437 20 grub_uint8_t unused5
17438 1 grub_uint64_t flags2
17439 48 grub_uint8_t unused6
17440
17441 This patch explicitly splits the v2 and v3 parts of the structure.
17442 The unused4 is still ending of the v2 structures and the v3 starts
17443 at unused5. Thanks to this we will avoid future corruptions of v2
17444 or v3 inodes.
17445
17446 The XFS_V2_INODE_SIZE is returning to its expected size and the
17447 filesystem is back to a readable state:
17448
17449 GNU GRUB version 2.11
17450 ....
17451 grub> set debug=efi,gpt,xfs
17452 grub> insmod part_gpt
17453 grub> ls (hd0,gpt1)/
17454 partmap/gpt.c:93: Read a valid GPT header
17455 partmap/gpt.c:115: GPT entry 0: start=4096, length=1953125
17456 fs/xfs.c:931: Reading sb
17457 fs/xfs.c:270: Validating superblock
17458 fs/xfs.c:295: XFS v4 superblock detected
17459 fs/xfs.c:962: Reading root ino 128
17460 fs/xfs.c:515: Reading inode (128) - 64, 0
17461 fs/xfs.c:515: Reading inode (128) - 64, 0
17462 fs/xfs.c:931: Reading sb
17463 fs/xfs.c:270: Validating superblock
17464 fs/xfs.c:295: XFS v4 superblock detected
17465 fs/xfs.c:962: Reading root ino 128
17466 fs/xfs.c:515: Reading inode (128) - 64, 0
17467 fs/xfs.c:515: Reading inode (128) - 64, 0
17468 fs/xfs.c:515: Reading inode (128) - 64, 0
17469 fs/xfs.c:515: Reading inode (131) - 64, 768
17470 efi/ fs/xfs.c:515: Reading inode (3145856) - 1464904, 0
17471 grub2/ fs/xfs.c:515: Reading inode (132) - 64, 1024
17472 grub/ fs/xfs.c:515: Reading inode (139) - 64, 2816
17473 grub>
17474
17475 Fixes: 8b1e5d193 (fs/xfs: Add bigtime incompat feature support)
17476
17477 Tested-by: Carlos Maiolino <[email protected]>
17478 Reviewed-by: Daniel Kiper <[email protected]>
17479
17480 2021-09-06 Heinrich Schuchardt <[email protected]>
17481
17482 libgcrypt: Avoid -Wempty-body in rijndael do_setkey()
17483 Avoid a warning
17484
17485 lib/libgcrypt-grub/cipher/rijndael.c:229:9:
17486 warning: suggest braces around empty body in an ‘if’ statement [-Wempty-body]
17487 229 | ;
17488 | ^
17489
17490 Reviewed-by: Daniel Kiper <[email protected]>
17491
17492 2021-09-06 Heinrich Schuchardt <[email protected]>
17493
17494 libgcrypt: Avoid -Wsign-compare in rijndael do_setkey()
17495 Avoid a warning
17496
17497 lib/libgcrypt-grub/cipher/rijndael.c:352:21: warning:
17498 comparison of integer expressions of different signedness:
17499 ‘int’ and ‘unsigned int’ [-Wsign-compare]
17500 352 | for (i = 0; i < keylen; i++)
17501 |
17502
17503 Reviewed-by: Daniel Kiper <[email protected]>
17504
17505 2021-09-06 Wouter van Kesteren <[email protected]>
17506
17507 commands/setpci: Honor write mask argument
17508 In the case that one passes a write mask with ":" the write_mask is
17509 obtained from grub_strtoul() and then promptly overwritten by 0xffffffff
17510 three lines later.
17511
17512 This appears to have been so since the initial version of setpci in 2009.
17513 I'm surprised no one else has hit this issue in the past 12 years...
17514
17515 Reviewed-by: Daniel Kiper <[email protected]>
17516
17517 2021-07-22 Jeff Mahoney <[email protected]>
17518
17519 osdep/linux/hostdisk: Use stat() instead of udevadm for partition lookup
17520 The sysfs_partition_path() calls udevadm to resolve the sysfs path for
17521 a block device. That can be accomplished by stating the device node
17522 and using the major/minor to follow the symlinks in /sys/dev/block/.
17523
17524 This cuts the execution time of grub-mkconfig to somewhere near 55% on
17525 system without LVM (which uses libdevmapper instead sysfs_partition_path()).
17526
17527 Remove udevadm call as it does not help us more than calling stat() directly.
17528
17529 Reviewed-by: Daniel Kiper <[email protected]>
17530
17531 2021-07-22 Petr Vorel <[email protected]>
17532
17533 osdep: Introduce include/grub/osdep/major.h and use it
17534 ... to factor out fix for glibc 2.25 introduced in 7a5b301e3 (build: Use
17535 AC_HEADER_MAJOR to find device macros).
17536
17537 Note: Once glibc 2.25 is old enough and this fix is not needed also
17538 AC_HEADER_MAJOR in configure.ac should be removed.
17539
17540 Reviewed-by: Daniel Kiper <[email protected]>
17541
17542 2021-07-22 Daniel Axtens <[email protected]>
17543
17544 ieee1275: Drop HEAP_MAX_ADDR and HEAP_MIN_SIZE constants
17545 The HEAP_MAX_ADDR is confusing. Currently it is set to 32MB, except on
17546 ieee1275 on x86, where it is 64MB.
17547
17548 There is a comment which purports to explain it:
17549
17550 /* If possible, we will avoid claiming heap above this address, because it
17551 seems to cause relocation problems with OSes that link at 4 MiB */
17552
17553 This doesn't make a lot of sense when the constants are well above 4MB
17554 already. It was not always this way. Prior to commit 7b5d0fe4440c
17555 (Increase heap limit) in 2010, HEAP_MAX_SIZE and HEAP_MAX_ADDR were
17556 indeed 4MB. However, when the constants were increased the comment was
17557 left unchanged.
17558
17559 It's been over a decade. It doesn't seem like we have problems with
17560 claims over 4MB on powerpc or x86 ieee1275. The SPARC does things
17561 completely differently and never used the constant.
17562
17563 Drop the constant and the check.
17564
17565 The only use of HEAP_MIN_SIZE was to potentially override the
17566 HEAP_MAX_ADDR check. It is now unused. Remove it too.
17567
17568 Tested-by: Stefan Berger <[email protected]>
17569 Reviewed-by: Daniel Kiper <[email protected]>
17570
17571 2021-07-22 Marius Bakke <[email protected]>
17572
17573 tests/ahci: Change "ide-drive" deprecated QEMU device name to "ide-hd"
17574 The "ide-drive" device was removed in QEMU 6.0. The "ide-hd" has been
17575 available for more than 10 years now in QEMU. Thus there shouldn't be
17576 any need for backwards compatible names.
17577
17578 Reviewed-by: Daniel Kiper <[email protected]>
17579
17580 2021-07-22 Javier Martinez Canillas <[email protected]>
17581
17582 fs/ext2: Ignore checksum seed incompat feature
17583 This incompat feature is used to denote that the filesystem stored its
17584 metadata checksum seed in the superblock. This is used to allow tune2fs
17585 changing the UUID on a mounted metdata_csum filesystem without having
17586 to rewrite all the disk metadata. However, the GRUB doesn't use the
17587 metadata checksum at all. So, it can just ignore this feature if it
17588 is enabled. This is consistent with the GRUB filesystem code in general
17589 which just does a best effort to access the filesystem's data.
17590
17591 The checksum seed incompat feature has to be removed from the ignore
17592 list if the support for metadata checksum verification is added to the
17593 GRUB ext2 driver later.
17594
17595 Suggested-by: Eric Sandeen <[email protected]>
17596 Suggested-by: Lukas Czerner <[email protected]>
17597 Reviewed-by: Lukas Czerner <[email protected]>
17598 Reviewed-by: Daniel Kiper <[email protected]>
17599
17600 2021-06-08 Glenn Washburn <[email protected]>
17601
17602 zfs: Use grub_uint64_t instead of 1ULL in BF64_*CODE() macros
17603 The underlying type of 1ULL does not change across architectures but
17604 grub_uint64_t does. This allows using the BF64_*CODE() macros as
17605 arguments to format string functions that use the PRI* format string
17606 macros that also vary with architecture.
17607
17608 Change the grub_error() call, where this was previously an issue and
17609 temporarily fixed by casting and using a format string literal code,
17610 to now use PRI* macros and remove casting.
17611
17612 Reviewed-by: Daniel Kiper <[email protected]>
17613
17614 2021-06-08 Daniel Kiper <[email protected]>
17615
17616 Bump version to 2.11
17617 Skip versions between 2.07 and 2.10 to avoid leading zeros in minor
17618 version number. This way version parsing in scripts should be easier.
17619
17620 Release 2.06
17621
17622 2021-06-08 Daniel Kiper <[email protected]>
17623
17624 SECURITY: Add SECURITY file
17625 The SECURITY file describes the GRUB project security policy.
17626
17627 It is based on https://github.com/wireapp/wire/blob/master/SECURITY.md
17628
17629 2021-06-08 Daniel Kiper <[email protected]>
17630
17631 MAINTAINERS: Add MAINTAINERS file
17632 The MAINTAINERS file provides basic information about the GRUB project
17633 and its maintainers.
17634
17635 2021-06-01 Dimitri John Ledkov <[email protected]>
17636
17637 grub-install: Add backup and restore
17638 Refactor clean_grub_dir() to create a backup of all the files, instead
17639 of just irrevocably removing them as the first action. If available,
17640 register atexit() handler to restore the backup if errors occur before
17641 point of no return, or remove the backup if everything was successful.
17642 If atexit() is not available, the backup remains on disk for manual
17643 recovery.
17644
17645 Some platforms defined a point of no return, i.e. after modules & core
17646 images were updated. Failures from any commands after that stage are
17647 ignored, and backup is cleaned up. For example, on EFI platforms update
17648 is not reverted when efibootmgr fails.
17649
17650 Extra care is taken to ensure atexit() handler is only invoked by the
17651 parent process and not any children forks. Some older GRUB codebases
17652 can invoke parent atexit() hooks from forks, which can mess up the
17653 backup.
17654
17655 This allows safer upgrades of MBR & modules, such that
17656 modules/images/fonts/translations are consistent with MBR in case of
17657 errors. For example accidental grub-install /dev/non-existent-disk
17658 currently clobbers and upgrades modules in /boot/grub, despite not
17659 actually updating any MBR.
17660
17661 This patch only handles backup and restore of files copied to /boot/grub.
17662 This patch does not perform backup (or restoration) of MBR itself or
17663 blocklists. Thus when installing i386-pc platform, corruption may still
17664 occur with MBR and blocklists which will not be attempted to be
17665 automatically recovered.
17666
17667 Also add modinfo.sh and *.efi to the cleanup/backup/restore code path,
17668 to ensure it is also cleaned, backed up and restored.
17669
17670 Reviewed-by: Daniel Kiper <[email protected]>
17671
17672 2021-06-01 Dimitri John Ledkov <[email protected]>
17673
17674 osdep/unix/exec: Avoid atexit() handlers when child execvp() fails
17675 The functions grub_util_exec_pipe() and grub_util_exec_pipe_stderr()
17676 currently call execvp(). If the call fails for any reason, the child
17677 currently calls exit(127). This in turn executes the parents
17678 atexit() handlers from the forked child, and then the same handlers
17679 are called again from parent. This is usually not desired, and can
17680 lead to deadlocks, and undesired behavior. So, change the exit() calls
17681 to _exit() calls to avoid calling atexit() handlers from child.
17682
17683 Fixes: e75cf4a58 (unix exec: avoid atexit handlers when child exits)
17684
17685 Reviewed-by: Daniel Kiper <[email protected]>
17686
17687 2021-06-01 Jan (janneke) Nieuwenhuizen <[email protected]>
17688
17689 lib/i386/relocator64: Build fixes for i386
17690 This fixes cross-compiling to x86 (e.g., the Hurd) from x86-linux of
17691
17692 grub-core/lib/i386/relocator64.S
17693
17694 This file has six sections that only build with a 64-bit assembler,
17695 yet only the first two sections had support for a 32-bit assembler.
17696 This patch completes this for the remaining sections.
17697
17698 To reproduce, update the GRUB source description in your local Guix
17699 archive and run
17700
17701 ./pre-inst-env guix build --system=i686-linux --target=i586-pc-gnu grub
17702
17703 or install an x86 cross-build environment on x86-linux (32-bit!) and
17704 configure to cross build and make, e.g., do something like
17705
17706 ./configure \
17707 CC_FOR_BUILD=gcc \
17708 --build=i686-unknown-linux-gnu \
17709 --host=i586-pc-gnu
17710 make
17711
17712 Additionally, remove a line with redundant spaces.
17713
17714 Reviewed-by: Daniel Kiper <[email protected]>
17715
17716 2021-06-01 Javier Martinez Canillas <[email protected]>
17717
17718 fs/xfs: Add needsrepair incompat feature support
17719 The XFS now has an incompat feature flag to indicate that a filesystem
17720 needs to be repaired. The Linux kernel refuses to mount the filesystem
17721 that has it set and only the xfs_repair tool is able to clear that flag.
17722
17723 The GRUB doesn't have the concept of mounting filesystems and just
17724 attempts to read the files. But it does some sanity checking before
17725 attempting to read from the filesystem. Among the things which are tested,
17726 is if the super block only has set of incompatible features flags that
17727 are supported by GRUB. If it contains any flags that are not listed as
17728 supported, reading the XFS filesystem fails.
17729
17730 Since the GRUB doesn't attempt to detect if the filesystem is inconsistent
17731 nor replays the journal, the filesystem access is a best effort. For this
17732 reason, ignore if the filesystem needs to be repaired and just print a debug
17733 message. That way, if reading or booting fails later, the user is able to
17734 figure out that the failures can be related to broken XFS filesystem.
17735
17736 Suggested-by: Eric Sandeen <[email protected]>
17737 Reviewed-by: Daniel Kiper <[email protected]>
17738
17739 2021-06-01 Carlos Maiolino <[email protected]>
17740
17741 fs/xfs: Add bigtime incompat feature support
17742 The XFS filesystem supports a bigtime feature to overcome y2038 problem.
17743 This patch makes the GRUB able to support the XFS filesystems with this
17744 feature enabled.
17745
17746 The XFS counter for the bigtime enabled timestamps starts at 0, which
17747 translates to GRUB_INT32_MIN (Dec 31 20:45:52 UTC 1901) in the legacy
17748 timestamps. The conversion to Unix timestamps is made before passing the
17749 value to other GRUB functions.
17750
17751 For this to work properly, GRUB requires an access to flags2 field in the
17752 XFS ondisk inode. So, the grub_xfs_inode structure has been updated to
17753 cover full ondisk inode.
17754
17755 Reviewed-by: Daniel Kiper <[email protected]>
17756
17757 2021-06-01 Carlos Maiolino <[email protected]>
17758
17759 fs: Use 64-bit type for filesystem timestamp
17760 Some filesystems nowadays use 64-bit types for timestamps. So, update
17761 grub_dirhook_info struct to use an grub_int64_t type to store mtime.
17762 This also updates the grub_unixtime2datetime() function to receive
17763 a 64-bit timestamp argument and do 64-bit-safe divisions.
17764
17765 All the remaining conversion from 32-bit to 64-bit should be safe, as
17766 32-bit to 64-bit attributions will be implicitly casted. The most
17767 critical part in the 32-bit to 64-bit conversion is in the function
17768 grub_unixtime2datetime() where it needs to deal with the 64-bit type.
17769 So, for that, the grub_divmod64() helper has been used.
17770
17771 These changes enables the GRUB to support dates beyond y2038.
17772
17773 Reviewed-by: Daniel Kiper <[email protected]>
17774
17775 2021-05-28 Javier Martinez Canillas <[email protected]>
17776
17777 types: Define PRI{x,d}GRUB_INT{32,64}_T format specifiers
17778 There are already PRI*_T constants defined for unsigned integers but not
17779 for signed integers. Add format specifiers for the latter.
17780
17781 Suggested-by: Daniel Kiper <[email protected]>
17782 Reviewed-by: Daniel Kiper <[email protected]>
17783
17784 2021-05-28 Tianjia Zhang <[email protected]>
17785
17786 kern/efi/sb: Remove duplicate efi_shim_lock_guid variable
17787 The efi_shim_lock_guid local variable and shim_lock_guid global variable
17788 have the same GUID value. Only the latter is retained.
17789
17790 Reviewed-by: Daniel Kiper <[email protected]>
17791
17792 2021-05-10 Javier Martinez Canillas <[email protected]>
17793
17794 util/mkimage: Fix wrong PE32+ section sizes for some arches
17795 The commit f60ba9e5945 (util/mkimage: Refactor section setup to use a helper)
17796 added a helper function to setup PE sections. But it also changed how the
17797 raw data offsets were calculated since all the section sizes are aligned.
17798 However, for some platforms, i.e ia64-efi and arm64-efi, the kernel image
17799 size is not aligned using the section alignment. This leads to the situation
17800 in which the mods section offset in its PE section header does not match its
17801 real placement in the PE file. So, finally the GRUB is not able to locate
17802 and load built-in modules.
17803
17804 The problem surfaces on ia64-efi and arm64-efi because both platforms
17805 require additional relocation data which is added behind .bss section.
17806 So, we have to add some padding behind this extra data to make the
17807 beginning of mods section properly aligned in the PE file. Fix it by
17808 aligning the kernel_size to the section alignment. That makes the sizes
17809 and offsets in the PE section headers to match relevant sections in the
17810 PE32+ binary file.
17811
17812 Reported-by: John Paul Adrian Glaubitz <[email protected]>
17813 Tested-by: John Paul Adrian Glaubitz <[email protected]>
17814 Reviewed-by: Daniel Kiper <[email protected]>
17815
17816 2021-05-10 Daniel Kiper <[email protected]>
17817
17818 term/terminfo: Fix the terminfo command help and documentation
17819 Additionally, fix the terminfo spelling mistake in
17820 the GRUB development documentation.
17821
17822 Reviewed-by: Javier Martinez Canillas <[email protected]>
17823
17824 2021-05-10 Daniel Kiper <[email protected]>
17825
17826 i18n: Align N_() formatting with the rest of GRUB code
17827 Reviewed-by: Javier Martinez Canillas <[email protected]>
17828
17829 2021-05-10 Daniel Kiper <[email protected]>
17830
17831 i18n: Format large integers before the translation message - take 2
17832 This is an additional fix which has been missing from the commit 837fe48de
17833 (i18n: Format large integers before the translation message).
17834
17835 Reviewed-by: Javier Martinez Canillas <[email protected]>
17836
17837 2021-04-13 Miguel Ángel Arruga Vivas <[email protected]>
17838
17839 i18n: Format large integers before the translation message
17840 The GNU gettext only supports the ISO C99 macros for integral
17841 types. If there is a need to use unsupported formatting macros,
17842 e.g. PRIuGRUB_UINT64_T, according to [1] the number to a string
17843 conversion should be separated from the code printing message
17844 requiring the internationalization. So, the function grub_snprintf()
17845 is used to print the numeric values to an intermediate buffer and
17846 the internationalized message contains a string format directive.
17847
17848 [1] https://www.gnu.org/software/gettext/manual/html_node/Preparing-Strings.html#No-string-concatenation
17849
17850 Reviewed-by: Daniel Kiper <[email protected]>
17851
17852 2021-04-12 Daniel Axtens <[email protected]>
17853
17854 video/fb/fbfill: Use unsigned integers for width/height
17855 Since commit 7ce3259f67ac (video/fb/fbfill: Fix potential integer
17856 overflow), clang builds of grub-emu have failed with messages like:
17857
17858 /usr/bin/ld: libgrubmods.a(libgrubmods_a-fbfill.o): in function `grub_video_fbfill_direct24':
17859 fbfill.c:(.text+0x28e): undefined reference to `__muloti4'
17860
17861 This appears to be due to a weird quirk in how clang compiles
17862
17863 grub_mul(dst->mode_info->bytes_per_pixel, width, &rowskip)
17864
17865 which is grub_mul(unsigned int, int, &grub_size_t).
17866
17867 It looks like clang somewhere promotes everything to 128-bit maths
17868 before ultimately reducing down to 64 bit for grub_size_t. I think
17869 this is because width is signed, and indeed converting width to an
17870 unsigned int makes the problem go away.
17871
17872 This conversion also makes more sense generally:
17873 - the caller of all the fbfill_directN functions is
17874 grub_video_fb_fill_dispatch() and it takes width and height as
17875 unsigned ints already,
17876 - it doesn't make sense to fill a negative width or height.
17877
17878 Convert the width and height arguments and associated loop counters
17879 to unsigned ints.
17880
17881 Fixes: 7ce3259f67ac (video/fb/fbfill: Fix potential integer overflow)
17882
17883 Reviewed-by: Daniel Kiper <[email protected]>
17884
17885 2021-04-12 Glenn Washburn <[email protected]>
17886
17887 docs: Conform badmem and cutmem description indentations with other commands
17888 Reviewed-by: Daniel Kiper <[email protected]>
17889
17890 docs: Add note to cryptomount that UUIDs should be specified without dashes
17891 Reviewed-by: Daniel Kiper <[email protected]>
17892
17893 2021-04-12 Aru Sahni <[email protected]>
17894
17895 templates: Fix user-facing typo with an incorrect use of "it's"
17896 Since the possessive form of "it" is being used, the apostrophe must be omitted.
17897
17898 Reviewed-by: Daniel Kiper <[email protected]>
17899
17900 2021-04-12 Colin Watson <[email protected]>
17901
17902 buffer: Sync up out-of-range error message
17903 The messages associated with other similar GRUB_ERR_OUT_OF_RANGE errors
17904 were lacking the trailing full stop. Syncing up the strings saves a small
17905 amount of precious core image space on i386-pc.
17906
17907 DOWN: obj/i386-pc/grub-core/kernel.img (31740 > 31708) - change: -32
17908 DOWN: i386-pc core image (biosdisk ext2 part_msdos) (27453 > 27452) - change: -1
17909 DOWN: i386-pc core image (biosdisk ext2 part_msdos diskfilter mdraid09) (32367 > 32359) - change: -8
17910
17911 Reviewed-by: Daniel Kiper <[email protected]>
17912
17913 2021-04-12 Glenn Washburn <[email protected]>
17914
17915 usb/usbhub: Use GRUB_USB_MAX_CONF macro instead of literal in hub for maximum configs
17916 Reviewed-by: Daniel Kiper <[email protected]>
17917
17918 2021-04-12 Daniel Drake <[email protected]>
17919
17920 fs/minix: Avoid mistakenly probing ext2 filesystems
17921 The ext2 (and ext3, ext4) filesystems write the number of free inodes to
17922 location 0x410.
17923
17924 On a MINIX filesystem, that same location is used for the MINIX superblock
17925 magic number.
17926
17927 If the number of free inodes on an ext2 filesystem is equal to any
17928 of the four MINIX superblock magic values plus any multiple of 65536,
17929 GRUB's MINIX filesystem code will probe it as a MINIX filesystem.
17930
17931 In the case of an OS using ext2 as the root filesystem, since there will
17932 ordinarily be some amount of file creation and deletion on every bootup,
17933 it effectively means that this situation has a 1:16384 chance of being hit
17934 on every reboot.
17935
17936 This will cause GRUB's filesystem probing code to mistakenly identify an
17937 ext2 filesystem as MINIX. This can be seen by e.g. "search --label"
17938 incorrectly indicating that no such ext2 partition with matching label
17939 exists, whereas in fact it does.
17940
17941 After spotting the rough cause of the issue I was facing here, I borrowed
17942 much of the diagnosis/explanation from meierfra who found and investigated
17943 the same issue in util-linux in 2010:
17944
17945 https://bugs.launchpad.net/ubuntu/+source/util-linux/+bug/518582
17946
17947 This was fixed in util-linux by having the MINIX code check for the
17948 ext2 magic. Do the same here.
17949
17950 Reviewed-by: Derek Foreman <[email protected]>
17951 Reviewed-by: Daniel Kiper <[email protected]>
17952
17953 2021-03-12 Daniel Kiper <[email protected]>
17954
17955 Release 2.06~rc1
17956
17957 2021-03-11 Ard Biesheuvel <[email protected]>
17958
17959 arm/linux: Fix ARM Linux header layout
17960 The hdr_offset member of the ARM Linux image header appears at
17961 offset 0x3c, matching the PE/COFF spec's placement of the COFF
17962 header offset in the MS-DOS header. We're currently off by four,
17963 so fix that.
17964
17965 Reviewed-by: Daniel Kiper <[email protected]>
17966
17967 2021-03-10 Glenn Washburn <[email protected]>
17968
17969 style: Format string macro should have a space between quotes
17970 Reviewed-by: Daniel Kiper <[email protected]>
17971
17972 2021-03-10 Glenn Washburn <[email protected]>
17973
17974 grub/err: Do compile-time format string checking on grub_error()
17975 This should help prevent format string errors and thus improve the quality
17976 of error reporting.
17977
17978 Reviewed-by: Daniel Kiper <[email protected]>
17979
17980 2021-03-10 Glenn Washburn <[email protected]>
17981
17982 fs/zfs/zfs: Use format code "%llu" for 64-bit uint bp->blk_prop in grub_error()
17983 This is a temporary, less-intrusive change to get the build to success with
17984 compiler format string checking turned on. There is a better fix which
17985 addresses this issue, but it needs more testing. Use this change so that
17986 format string checking on grub_error() can be turned on until the better
17987 change is fully tested.
17988
17989 Reviewed-by: Daniel Kiper <[email protected]>
17990
17991 2021-03-10 Glenn Washburn <[email protected]>
17992
17993 fs/hfsplus: Use format code PRIuGRUB_UINT64_T for 64-bit typed fileblock in grub_error()
17994 Reviewed-by: Daniel Kiper <[email protected]>
17995
17996 2021-03-10 Glenn Washburn <[email protected]>
17997
17998 dl/elf: Use format code PRIxGRUB_UINT64_T for 64-bit arg in grub_error()
17999 The macro ELF_R_TYPE does not change the underlying type. Here its argument
18000 is a 64-bit Elf64_Xword. Make sure the format code matches.
18001
18002 For the RISC-V architecture, rel->r_info could be either Elf32_Xword or
18003 Elf64_Xword depending on if 32 or 64-bit RISC-V is being built. So cast
18004 to 64-bit value regardless.
18005
18006 Reviewed-by: Daniel Kiper <[email protected]>
18007
18008 2021-03-10 Glenn Washburn <[email protected]>
18009
18010 disk/ata: Use format code PRIxGRUB_UINT64_T for 64-bit uint argument in grub_error()
18011 Reviewed-by: Daniel Kiper <[email protected]>
18012
18013 2021-03-10 Glenn Washburn <[email protected]>
18014
18015 loader/i386/pc/linux: Use PRI* macros to get correct format string code across architectures
18016 Also remove casting of format string args so that the architecture dependent
18017 type is preserved.
18018
18019 Reviewed-by: Daniel Kiper <[email protected]>
18020
18021 2021-03-10 Glenn Washburn <[email protected]>
18022
18023 kern/efi/mm: Format string error in grub_error()
18024 The second format string argument, GRUB_EFI_MAX_USABLE_ADDRESS, is a macro
18025 to a number literal. However, depending on what the target architecture, the
18026 type can be 32 or 64 bits. Cast to a 64-bit integer. Also, change the
18027 format string literals "%llx" to use PRIxGRUB_UINT64_T.
18028
18029 Reviewed-by: Daniel Kiper <[email protected]>
18030
18031 2021-03-10 Glenn Washburn <[email protected]>
18032
18033 commands/pgp: Format code for grub_error() is incorrect
18034 The format code is for a 32-bit int, but the argument, keyid, is declared as
18035 a 64 bit int. The comment above says keyid is 32-bit. I'm not sure if the
18036 comment or declaration is wrong, so force the display of a 64-bit int for now.
18037
18038 Reviewed-by: Daniel Kiper <[email protected]>
18039
18040 2021-03-10 Glenn Washburn <[email protected]>
18041
18042 grub_error: Use format code PRIuGRUB_SIZE for variables of type grub_size_t
18043 Reviewed-by: Daniel Kiper <[email protected]>
18044
18045 2021-03-10 Glenn Washburn <[email protected]>
18046
18047 disk/dmraid_nvidia: Format string error in grub_error()
18048 The grub_error() has a format string expecting two arguments, but only one
18049 provided. According to the comments in the struct grub_nv_super definition,
18050 the version field looks like a version number where major.minor is encoded
18051 as each a byte in the two-byte short.
18052
18053 Reviewed-by: Daniel Kiper <[email protected]>
18054
18055 2021-03-10 Glenn Washburn <[email protected]>
18056
18057 video/bochs: grub_error() format string add missing format code
18058 Reviewed-by: Daniel Kiper <[email protected]>
18059
18060 2021-03-10 Glenn Washburn <[email protected]>
18061
18062 parttool/msdospart: grub_error() missing format string argument
18063 Its obvious from the error message that the variable named "type" was
18064 accidentally omitted.
18065
18066 Reviewed-by: Daniel Kiper <[email protected]>
18067
18068 2021-03-10 Glenn Washburn <[email protected]>
18069
18070 misc: Format string for grub_error() should be a literal
18071 Reviewed-by: Daniel Kiper <[email protected]>
18072
18073 2021-03-10 Philip Müller <[email protected]>
18074
18075 templates: Properly disable the os-prober by default
18076 This patch does the following:
18077 - really disables os-prober by default in the util/grub-mkconfig.in
18078 by setting GRUB_DISABLE_OS_PROBER to true,
18079 - fixes the logic in the util/grub.d/30_os-prober.in,
18080 - updates the grub_warn() lines.
18081
18082 Reason for the code shuffling in the util/grub-mkconfig.in:
18083
18084 The default was GRUB_DISABLE_OS_PROBER=false if you don't set
18085 GRUB_DISABLE_OS_PROBER at all. To prevent os-prober from starting we
18086 have to set it by default to true and shuffle GRUB_DISABLE_OS_PROBER to
18087 code section, which is executed by the script. However we still give an
18088 option to the user to overwrite it with false, if he wants to execute
18089 os-prober after all.
18090
18091 Fixes: e3464147 (templates: Disable the os-prober by default)
18092
18093 Reported-by: Didier Spaier <[email protected]>
18094 Reported-by: Lennart Sorensen <[email protected]>
18095 Reported-by: John Paul Adrian Glaubitz <[email protected]>
18096 Reviewed-by: Daniel Kiper <[email protected]>
18097
18098 2021-03-10 Michael Chang <[email protected]>
18099
18100 kern/efi/sb: Add chainloaded image as shim's verifiable object
18101 While attempting to dual boot Microsoft Windows with UEFI chainloader,
18102 it failed with below error when UEFI Secure Boot was enabled:
18103
18104 error ../../grub-core/kern/verifiers.c:119:verification requested but
18105 nobody cares: /EFI/Microsoft/Boot/bootmgfw.efi.
18106
18107 It is a regression, as previously it worked without any problem.
18108
18109 It turns out chainloading PE image has been locked down by commit
18110 578c95298 (kern: Add lockdown support). However, we should consider it
18111 as verifiable object by shim to allow booting in UEFI Secure Boot mode.
18112 The chainloaded PE image could also have trusted signature created by
18113 vendor with their pubkey cert in db. For that matters it's usage should
18114 not be locked down under UEFI Secure Boot, and instead shim should be
18115 allowed to validate a PE binary signature before running it.
18116
18117 Fixes: 578c95298 (kern: Add lockdown support)
18118
18119 Reviewed-by: Daniel Kiper <[email protected]>
18120
18121 2021-03-10 Glenn Washburn <[email protected]>
18122
18123 disk/pata: Suppress error message "no device connected"
18124 This error message comes from the grub_print_error() in
18125 grub_pata_device_initialize(), which does not pass on the error, and is
18126 raised in check_device(). The function check_device() needs to return this
18127 as an error because check_device() is also used in grub_pata_open(), which
18128 does pass on this error to indicate that the device can not be used.
18129
18130 This is actually not an error when displayed by grub_pata_device_initialize()
18131 because it just indicates that there are no pata devices seen. This may be
18132 confusing to end users who do not have pata devices yet are loading the
18133 pata module (perhaps implicitly via nativedisk). This also causes unnecessary
18134 output which may need to be accounted for in functional testing.
18135
18136 Instead print to the debug log when check_device() raises this "error" and
18137 pop the error from the error stack. If there is another error on the stack
18138 then print the error stack as those should be real errors.
18139
18140 Acked-by: Paul Menzel <[email protected]>
18141 Reviewed-by: Daniel Kiper <[email protected]>
18142
18143 2021-03-10 Yi Zhao <[email protected]>
18144
18145 fs/ext2: Fix a file not found error when a symlink filesize is equal to 60
18146 We encountered a file not found error when the symlink filesize is
18147 equal to 60:
18148
18149 $ ls -l initrd
18150 lrwxrwxrwx 1 root root 60 Jan 6 16:37 initrd -> secure-core-image-initramfs-5.10.2-yoctodev-standard.cpio.gz
18151
18152 When booting, we got the following error in the GRUB:
18153
18154 error: file `/initrd' not found
18155
18156 The root cause is that the size of diro->inode.symlink is equal to 60
18157 and a symlink name has to be terminated with NUL there. So, if the
18158 symlink filesize is exactly 60 then it is also stored in a separate
18159 block rather than in the inode itself.
18160
18161 Reviewed-by: Daniel Kiper <[email protected]>
18162
18163 2021-03-02 Tianjia Zhang <[email protected]>
18164
18165 loader/i386/linux: Do not use grub_le_to_cpu32() for relocatable variable
18166 The relocatable variable is defined as grub_uint8_t. Relevant
18167 member in setup_header structure is also defined as one byte
18168 in Linux boot protocol. By semantic definition it is a bool type.
18169 It is not appropriate to treat it as a four bytes. This patch
18170 fixes the issue.
18171
18172 Reviewed-by: Daniel Kiper <[email protected]>
18173
18174 2021-03-02 Tianjia Zhang <[email protected]>
18175
18176 loader/i386/linux: Remove redundant code from in grub_cmd_linux()
18177 The preferred_address has been assigned to GRUB_LINUX_BZIMAGE_ADDR
18178 during initialization in grub_cmd_linux(). The assignment here
18179 is redundant and should be removed.
18180
18181 Reviewed-by: Daniel Kiper <[email protected]>
18182
18183 2021-03-02 Heinrich Schuchardt <[email protected]>
18184
18185 efi: The device-tree must be in EfiACPIReclaimMemory
18186 According to the Embedded Base Boot Requirements (EBBR) specification the
18187 device-tree passed to Linux as a configuration table must reside in
18188 EfiACPIReclaimMemory.
18189
18190 Reviewed-by: Daniel Kiper <[email protected]>
18191
18192 2021-03-02 Heinrich Schuchardt <[email protected]>
18193
18194 commands/efi/lsefisystab: Add short text for EFI_RT_PROPERTIES_TABLE_GUID
18195 UEFI specification 2.8 errata B introduced the EFI_RT_PROPERTIES_TABLE
18196 describing the services available at runtime.
18197
18198 The lsefisystab command is used to display installed EFI configuration
18199 tables. Currently it only shows the GUID but not a short text for the
18200 new table.
18201
18202 Provide a short text for the EFI_RT_PROPERTIES_TABLE_GUID.
18203
18204 Reviewed-by: Daniel Kiper <[email protected]>
18205
18206 2021-03-02 Petr Vorel <[email protected]>
18207
18208 docs/luks2: Mention key derivation function support
18209 To give users hint why Argon2, the default in cryptsetup for LUKS2, does
18210 not work.
18211
18212 Acked-by: Paul Menzel <[email protected]>
18213 Reviewed-by: Patrick Steinhardt <[email protected]>
18214 Reviewed-by: Daniel Kiper <[email protected]>
18215
18216 2021-03-02 Derek Foreman <[email protected]>
18217
18218 commands/file: Fix array/enum desync
18219 The commit f1957dc8a (RISC-V: Add to build system) added two entries to
18220 the options array, but only 1 entry to the enum. This resulted in
18221 everything after the insertion point being off by one.
18222
18223 This broke at least the "file --is-hibernated-hiberfil" command.
18224
18225 Bring the two back in sync by splitting the IS_RISCV_EFI enum entry into
18226 two, as is done for other architectures.
18227
18228 Reviewed-by: Daniel Kiper <[email protected]>
18229
18230 2021-03-02 Marco A Benatto <[email protected]>
18231
18232 kern/mm: Fix grub_debug_calloc() compilation error
18233 Fix compilation error due to missing parameter to
18234 grub_printf() when MM_DEBUG is defined.
18235
18236 Fixes: 64e26162e (calloc: Make sure we always have an overflow-checking calloc() available)
18237
18238 Reviewed-by: Daniel Kiper <[email protected]>
18239
18240 2021-03-02 Alex Burmashev <[email protected]>
18241
18242 templates: Disable the os-prober by default
18243 The os-prober is enabled by default what may lead to potentially
18244 dangerous use cases and borderline opening attack vectors. This
18245 patch disables the os-prober, adds warning messages and updates
18246 GRUB_DISABLE_OS_PROBER configuration option documentation. This
18247 way we make it clear that the os-prober usage is not recommended.
18248
18249 Simplistic nature of this change allows downstream vendors, who
18250 really want os-prober to be enabled out of the box in their
18251 relevant products, easily revert to it's old behavior.
18252
18253 Reported-by: NyankoSec (<[email protected]>, https://twitter.com/NyankoSec),
18254 working with SSD Secure Disclosure
18255 Reviewed-by: Daniel Kiper <[email protected]>
18256
18257 2021-03-02 Thomas Frauendorfer | Miray Software <[email protected]>
18258
18259 gfxmenu/gui: Check printf() format in the gui_progress_bar and gui_label
18260 The gui_progress_bar and gui_label components can display the timeout
18261 value. The format string can be set through a theme file. This patch
18262 adds a validation step to the format string.
18263
18264 If a user loads a theme file into the GRUB without this patch then
18265 a GUI label with the following settings
18266
18267 + label {
18268 ...
18269 id = "__timeout__"
18270 text = "%s"
18271 }
18272
18273 will interpret the current timeout value as string pointer and print the
18274 memory at that position on the screen. It is not desired behavior.
18275
18276 Reviewed-by: Daniel Kiper <[email protected]>
18277
18278 2021-03-02 Thomas Frauendorfer | Miray Software <[email protected]>
18279
18280 kern/misc: Add function to check printf() format against expected format
18281 The grub_printf_fmt_check() function parses the arguments of an untrusted
18282 printf() format and an expected printf() format and then compares the
18283 arguments counts and arguments types. The arguments count in the untrusted
18284 format string must be less or equal to the arguments count in the expected
18285 format string and both arguments types must match.
18286
18287 To do this the parse_printf_arg_fmt() helper function is extended in the
18288 following way:
18289
18290 1. Add a return value to report errors to the grub_printf_fmt_check().
18291
18292 2. Add the fmt_check argument to enable stricter format verification:
18293 - the function expects that arguments definitions are always
18294 terminated by a supported conversion specifier.
18295 - positional parameters, "$", are not allowed, as they cannot be
18296 validated correctly with the current implementation. For example
18297 "%s%1$d" would assign the first args entry twice while leaving the
18298 second one unchanged.
18299 - Return an error if preallocated space in args is too small and
18300 allocation fails for the needed size. The grub_printf_fmt_check()
18301 should verify all arguments. So, if validation is not possible for
18302 any reason it should return an error.
18303 This also adds a case entry to handle "%%", which is the escape
18304 sequence to print "%" character.
18305
18306 3. Add the max_args argument to check for the maximum allowed arguments
18307 count in a printf() string. This should be set to the arguments count
18308 of the expected format. Then the parse_printf_arg_fmt() function will
18309 return an error if the arguments count is exceeded.
18310
18311 The two additional arguments allow us to use parse_printf_arg_fmt() in
18312 printf() and grub_printf_fmt_check() calls.
18313
18314 When parse_printf_arg_fmt() is used by grub_printf_fmt_check() the
18315 function parse user provided untrusted format string too. So, in
18316 that case it is better to be too strict than too lenient.
18317
18318 Reviewed-by: Daniel Kiper <[email protected]>
18319
18320 2021-03-02 Thomas Frauendorfer | Miray Software <[email protected]>
18321
18322 kern/misc: Add STRING type for internal printf() format handling
18323 Set printf() argument type for "%s" to new type STRING. This is in
18324 preparation for a follow up patch to compare a printf() format string
18325 against an expected printf() format string.
18326
18327 For "%s" the corresponding printf() argument is dereferenced as pointer
18328 while all other argument types are defined as integer value. However,
18329 when validating a printf() format it is necessary to differentiate "%s"
18330 from "%p" and other integers. So, let's do that.
18331
18332 Reviewed-by: Daniel Kiper <[email protected]>
18333
18334 2021-03-02 Thomas Frauendorfer | Miray Software <[email protected]>
18335
18336 kern/misc: Split parse_printf_args() into format parsing and va_list handling
18337 This patch is preparing for a follow up patch which will use
18338 the format parsing part to compare the arguments in a printf()
18339 format from an external source against a printf() format with
18340 expected arguments.
18341
18342 Reviewed-by: Daniel Kiper <[email protected]>
18343
18344 2021-03-02 Dimitri John Ledkov <[email protected]>
18345
18346 shim_lock: Only skip loading shim_lock verifier with explicit consent
18347 Commit 32ddc42c (efi: Only register shim_lock verifier if shim_lock
18348 protocol is found and SB enabled) reintroduced CVE-2020-15705 which
18349 previously only existed in the out-of-tree linuxefi patches and was
18350 fixed as part of the BootHole patch series.
18351
18352 Under Secure Boot enforce loading shim_lock verifier. Allow skipping
18353 shim_lock verifier if SecureBoot/MokSBState EFI variables indicate
18354 skipping validations, or if GRUB image is built with --disable-shim-lock.
18355
18356 Fixes: 132ddc42c (efi: Only register shim_lock verifier if shim_lock
18357 protocol is found and SB enabled)
18358 Fixes: CVE-2020-15705
18359 Fixes: CVE-2021-3418
18360
18361 Reported-by: Dimitri John Ledkov <[email protected]>
18362 Reviewed-by: Daniel Kiper <[email protected]>
18363
18364 2021-03-02 Dimitri John Ledkov <[email protected]>
18365
18366 grub-install-common: Add --sbat option
18367 Reviewed-by: Daniel Kiper <[email protected]>
18368
18369 2021-03-02 Peter Jones <[email protected]>
18370
18371 util/mkimage: Add an option to import SBAT metadata into a .sbat section
18372 Add a --sbat option to the grub-mkimage tool which allows us to import
18373 an SBAT metadata formatted as a CSV file into a .sbat section of the
18374 EFI binary.
18375
18376 Reviewed-by: Daniel Kiper <[email protected]>
18377
18378 2021-03-02 Peter Jones <[email protected]>
18379
18380 util/mkimage: Refactor section setup to use a helper
18381 Add a init_pe_section() helper function to setup PE sections. This makes
18382 the code simpler and easier to read.
18383
18384 Reviewed-by: Daniel Kiper <[email protected]>
18385
18386 2021-03-02 Peter Jones <[email protected]>
18387
18388 util/mkimage: Improve data_size value calculation
18389 According to "Microsoft Portable Executable and Common Object File Format
18390 Specification", the Optional Header SizeOfInitializedData field contains:
18391
18392 Size of the initialized data section, or the sum of all such sections if
18393 there are multiple data sections.
18394
18395 Make this explicit by adding the GRUB kernel data size to the sum of all
18396 the modules sizes. The ALIGN_UP() is not required by the PE spec but do
18397 it to avoid alignment issues.
18398
18399 Reviewed-by: Daniel Kiper <[email protected]>
18400
18401 2021-03-02 Peter Jones <[email protected]>
18402
18403 util/mkimage: Reorder PE optional header fields set-up
18404 This makes the PE32 and PE32+ header fields set-up easier to follow by
18405 setting them closer to the initialization of their related sections.
18406
18407 Reviewed-by: Daniel Kiper <[email protected]>
18408
18409 2021-03-02 Peter Jones <[email protected]>
18410
18411 util/mkimage: Unify more of the PE32 and PE32+ header set-up
18412 There's quite a bit of code duplication in the code that sets the optional
18413 header for PE32 and PE32+. The two are very similar with the exception of
18414 a few fields that have type grub_uint64_t instead of grub_uint32_t.
18415
18416 Factor out the common code and add a PE_OHDR() macro that simplifies the
18417 set-up and make the code more readable.
18418
18419 Reviewed-by: Daniel Kiper <[email protected]>
18420
18421 2021-03-02 Peter Jones <[email protected]>
18422
18423 util/mkimage: Always use grub_host_to_target32() to initialize PE stack and heap stuff
18424 This change does not impact final result of initialization itself.
18425 However, it eases PE code unification in subsequent patches.
18426
18427 Reviewed-by: Daniel Kiper <[email protected]>
18428
18429 2021-03-02 Peter Jones <[email protected]>
18430
18431 util/mkimage: Use grub_host_to_target32() instead of grub_cpu_to_le32()
18432 The latter doesn't take into account the target image endianness. There is
18433 a grub_cpu_to_le32_compile_time() but no compile time variant for function
18434 grub_host_to_target32(). So, let's keep using the other one for this case.
18435
18436 Reviewed-by: Daniel Kiper <[email protected]>
18437
18438 2021-03-02 Javier Martinez Canillas <[email protected]>
18439
18440 util/mkimage: Remove unused code to add BSS section
18441 The code is compiled out so there is no reason to keep it.
18442
18443 Additionally, don't set bss_size field since we do not add a BSS section.
18444
18445 Reviewed-by: Daniel Kiper <[email protected]>
18446
18447 2021-03-02 Chris Coulson <[email protected]>
18448
18449 kern/efi: Add initial stack protector implementation
18450 It works only on UEFI platforms but can be quite easily extended to
18451 others architectures and platforms if needed.
18452
18453 Reviewed-by: Marco A Benatto <[email protected]>
18454 Reviewed-by: Javier Martinez Canillas <[email protected]>
18455
18456 2021-03-02 Chris Coulson <[email protected]>
18457
18458 kern/parser: Fix a stack buffer overflow
18459 grub_parser_split_cmdline() expands variable names present in the supplied
18460 command line in to their corresponding variable contents and uses a 1 kiB
18461 stack buffer for temporary storage without sufficient bounds checking. If
18462 the function is called with a command line that references a variable with
18463 a sufficiently large payload, it is possible to overflow the stack
18464 buffer via tab completion, corrupt the stack frame and potentially
18465 control execution.
18466
18467 Fixes: CVE-2020-27749
18468
18469 Reported-by: Chris Coulson <[email protected]>
18470 Reviewed-by: Daniel Kiper <[email protected]>
18471
18472 2021-03-02 Chris Coulson <[email protected]>
18473
18474 kern/buffer: Add variable sized heap buffer
18475 Add a new variable sized heap buffer type (grub_buffer_t) with simple
18476 operations for appending data, accessing the data and maintaining
18477 a read cursor.
18478
18479 Reviewed-by: Daniel Kiper <[email protected]>
18480
18481 2021-03-02 Chris Coulson <[email protected]>
18482
18483 kern/parser: Refactor grub_parser_split_cmdline() cleanup
18484 Introduce a common function epilogue used for cleaning up on all
18485 return paths, which will simplify additional error handling to be
18486 introduced in a subsequent commit.
18487
18488 Reviewed-by: Daniel Kiper <[email protected]>
18489
18490 2021-03-02 Chris Coulson <[email protected]>
18491
18492 kern/parser: Introduce terminate_arg() helper
18493 process_char() and grub_parser_split_cmdline() use similar code for
18494 terminating the most recent argument. Add a helper function for this.
18495
18496 Reviewed-by: Daniel Kiper <[email protected]>
18497
18498 2021-03-02 Chris Coulson <[email protected]>
18499
18500 kern/parser: Introduce process_char() helper
18501 grub_parser_split_cmdline() iterates over each command line character.
18502 In order to add error checking and to simplify the subsequent error
18503 handling, split the character processing in to a separate function.
18504
18505 Reviewed-by: Daniel Kiper <[email protected]>
18506
18507 2021-03-02 Chris Coulson <[email protected]>
18508
18509 kern/parser: Fix a memory leak
18510 The getline() function supplied to grub_parser_split_cmdline() returns
18511 a newly allocated buffer and can be called multiple times, but the
18512 returned buffer is never freed.
18513
18514 Reviewed-by: Daniel Kiper <[email protected]>
18515
18516 2021-03-02 Daniel Axtens <[email protected]>
18517
18518 fs/btrfs: Squash some uninitialized reads
18519 We need to check errors before calling into a function that uses the result.
18520
18521 Reviewed-by: Daniel Kiper <[email protected]>
18522
18523 2021-03-02 Daniel Axtens <[email protected]>
18524
18525 fs/btrfs: Validate the number of stripes/parities in RAID5/6
18526 This prevents a divide by zero if nstripes == nparities, and
18527 also prevents propagation of invalid values if nstripes ends up
18528 less than nparities.
18529
18530 Reviewed-by: Daniel Kiper <[email protected]>
18531
18532 2021-03-02 Daniel Axtens <[email protected]>
18533
18534 disk/lvm: Do not allow a LV to be it's own segment's node's LV
18535 This prevents infinite recursion in the diskfilter verification code.
18536
18537 Reviewed-by: Daniel Kiper <[email protected]>
18538
18539 2021-03-02 Daniel Axtens <[email protected]>
18540
18541 disk/lvm: Sanitize rlocn->offset to prevent wild read
18542 rlocn->offset is read directly from disk and added to the metadatabuf
18543 pointer to create a pointer to a block of metadata. It's a 64-bit
18544 quantity so as long as you don't overflow you can set subsequent
18545 pointers to point anywhere in memory.
18546
18547 Require that rlocn->offset fits within the metadata buffer size.
18548
18549 Reviewed-by: Daniel Kiper <[email protected]>
18550
18551 2021-03-02 Daniel Axtens <[email protected]>
18552
18553 disk/lvm: Do not overread metadata
18554 We could reach the end of valid metadata and not realize, leading to
18555 some buffer overreads. Check if we have reached the end and bail.
18556
18557 Reviewed-by: Daniel Kiper <[email protected]>
18558
18559 2021-03-02 Daniel Axtens <[email protected]>
18560
18561 disk/lvm: Do not crash if an expected string is not found
18562 Clean up a bunch of cases where we could have strstr() fail and lead to
18563 us dereferencing NULL.
18564
18565 We'll still leak memory in some cases (loops don't clean up allocations
18566 from earlier iterations if a later iteration fails) but at least we're
18567 not crashing.
18568
18569 Reviewed-by: Daniel Kiper <[email protected]>
18570
18571 2021-03-02 Daniel Axtens <[email protected]>
18572
18573 disk/lvm: Bail on missing PV list
18574 There's an if block for the presence of "physical_volumes {", but if
18575 that block is absent, then p remains NULL and a NULL-deref will result
18576 when looking for logical volumes.
18577
18578 It doesn't seem like LVM makes sense without physical volumes, so error
18579 out rather than crashing.
18580
18581 Reviewed-by: Daniel Kiper <[email protected]>
18582
18583 2021-03-02 Daniel Axtens <[email protected]>
18584
18585 disk/lvm: Don't blast past the end of the circular metadata buffer
18586 This catches at least some OOB reads, and it's possible I suppose that
18587 if 2 * mda_size is less than GRUB_LVM_MDA_HEADER_SIZE it might catch some
18588 OOB writes too (although that hasn't showed up as a crash in fuzzing yet).
18589
18590 It's a bit ugly and I'd appreciate better suggestions.
18591
18592 Reviewed-by: Daniel Kiper <[email protected]>
18593
18594 2021-03-02 Daniel Axtens <[email protected]>
18595
18596 disk/lvm: Don't go beyond the end of the data we read from disk
18597 We unconditionally trusted offset_xl from the LVM label header, even if
18598 it told us that the PV header/disk locations were way off past the end
18599 of the data we read from disk.
18600
18601 Require that the offset be sane, fixing an OOB read and crash.
18602
18603 Fixes: CID 314367, CID 314371
18604
18605 Reviewed-by: Daniel Kiper <[email protected]>
18606
18607 2021-03-02 Daniel Axtens <[email protected]>
18608
18609 io/gzio: Zero gzio->tl/td in init_dynamic_block() if huft_build() fails
18610 If huft_build() fails, gzio->tl or gzio->td could contain pointers that
18611 are no longer valid. Zero them out.
18612
18613 This prevents a double free when grub_gzio_close() comes through and
18614 attempts to free them again.
18615
18616 Reviewed-by: Daniel Kiper <[email protected]>
18617
18618 2021-03-02 Daniel Axtens <[email protected]>
18619
18620 io/gzio: Catch missing values in huft_build() and bail
18621 In huft_build(), "v" is a table of values in order of bit length.
18622 The code later (when setting up table entries in "r") assumes that all
18623 elements of this array corresponding to a code are initialized and less
18624 than N_MAX. However, it doesn't enforce this.
18625
18626 With sufficiently manipulated inputs (e.g. from fuzzing), there can be
18627 elements of "v" that are not filled. Therefore a lookup into "e" or "d"
18628 will use an uninitialized value. This can lead to an invalid/OOB read on
18629 those values, often leading to a crash.
18630
18631 Reviewed-by: Daniel Kiper <[email protected]>
18632
18633 2021-03-02 Daniel Axtens <[email protected]>
18634
18635 io/gzio: Add init_dynamic_block() clean up if unpacking codes fails
18636 init_dynamic_block() didn't clean up gzio->tl and td in some error
18637 paths. This left td pointing to part of tl. Then in grub_gzio_close(),
18638 when tl was freed the storage for td would also be freed. The code then
18639 attempts to free td explicitly, performing a UAF and then a double free.
18640
18641 Explicitly clean up tl and td in the error paths.
18642
18643 Reviewed-by: Daniel Kiper <[email protected]>
18644
18645 2021-03-02 Daniel Axtens <[email protected]>
18646
18647 io/gzio: Bail if gzio->tl/td is NULL
18648 This is an ugly fix that doesn't address why gzio->tl comes to be NULL.
18649 However, it seems to be sufficient to patch up a bunch of NULL derefs.
18650
18651 It would be good to revisit this in future and see if we can have
18652 a cleaner solution that addresses some of the causes of the unexpected
18653 NULL pointers.
18654
18655 Reviewed-by: Daniel Kiper <[email protected]>
18656
18657 2021-03-02 Daniel Axtens <[email protected]>
18658
18659 fs/nilfs2: Properly bail on errors in grub_nilfs2_btree_node_lookup()
18660 We just introduced an error return in grub_nilfs2_btree_node_lookup().
18661 Make sure the callers catch it.
18662
18663 At the same time, make sure that grub_nilfs2_btree_node_lookup() always
18664 inits the index pointer passed to it.
18665
18666 Reviewed-by: Daniel Kiper <[email protected]>
18667
18668 2021-03-02 Daniel Axtens <[email protected]>
18669
18670 fs/nilfs2: Don't search children if provided number is too large
18671 NILFS2 reads the number of children a node has from the node. Unfortunately,
18672 that's not trustworthy. Check if it's beyond what the filesystem permits and
18673 reject it if so.
18674
18675 This blocks some OOB reads. I'm not sure how controllable the read is and what
18676 could be done with invalidly read data later on.
18677
18678 Reviewed-by: Daniel Kiper <[email protected]>
18679
18680 2021-03-02 Daniel Axtens <[email protected]>
18681
18682 fs/nilfs2: Reject too-large keys
18683 NILFS2 has up to 7 keys, per the data structure. Do not permit array
18684 indices in excess of that.
18685
18686 This catches some OOB reads. I don't know how controllable the invalidly
18687 read data is or if that could be used later in the program.
18688
18689 Reviewed-by: Daniel Kiper <[email protected]>
18690
18691 2021-03-02 Daniel Axtens <[email protected]>
18692
18693 fs/jfs: Catch infinite recursion
18694 It's possible with a fuzzed filesystem for JFS to keep getblk()-ing
18695 the same data over and over again, leading to stack exhaustion.
18696
18697 Check if we'd be calling the function with exactly the same data as
18698 was passed in, and if so abort.
18699
18700 I'm not sure what the performance impact of this is and am open to
18701 better ideas.
18702
18703 Reviewed-by: Daniel Kiper <[email protected]>
18704
18705 2021-03-02 Daniel Axtens <[email protected]>
18706
18707 fs/jfs: Limit the extents that getblk() can consider
18708 getblk() implicitly trusts that treehead->count is an accurate count of
18709 the number of extents. However, that value is read from disk and is not
18710 trustworthy, leading to OOB reads and crashes. I am not sure to what
18711 extent the data read from OOB can influence subsequent program execution.
18712
18713 Require callers to pass in the maximum number of extents for which
18714 they have storage.
18715
18716 Reviewed-by: Daniel Kiper <[email protected]>
18717
18718 2021-03-02 Daniel Axtens <[email protected]>
18719
18720 fs/jfs: Do not move to leaf level if name length is negative
18721 Fuzzing JFS revealed crashes where a negative number would be passed
18722 to le_to_cpu16_copy(). There it would be cast to a large positive number
18723 and the copy would read and write off the end of the respective buffers.
18724
18725 Catch this at the top as well as the bottom of the loop.
18726
18727 Reviewed-by: Daniel Kiper <[email protected]>
18728
18729 2021-03-02 Daniel Axtens <[email protected]>
18730
18731 fs/sfs: Fix over-read of root object name
18732 There's a read of the name of the root object that assumes that the name
18733 is nul-terminated within the root block. This isn't guaranteed - it seems
18734 SFS would require you to read multiple blocks to get a full name in general,
18735 but maybe that doesn't apply to the root object.
18736
18737 Either way, figure out how much space is left in the root block and don't
18738 over-read it. This fixes some OOB reads.
18739
18740 Reviewed-by: Daniel Kiper <[email protected]>
18741
18742 2021-03-02 Daniel Axtens <[email protected]>
18743
18744 fs/hfs: Disable under lockdown
18745 HFS has issues such as infinite mutual recursion that are simply too
18746 complex to fix for such a legacy format. So simply do not permit
18747 it to be loaded under lockdown.
18748
18749 Reviewed-by: Daniel Kiper <[email protected]>
18750
18751 2021-03-02 Daniel Axtens <[email protected]>
18752
18753 fs/hfsplus: Don't use uninitialized data on corrupt filesystems
18754 Valgrind identified the following use of uninitialized data:
18755
18756 ==2782220== Conditional jump or move depends on uninitialised value(s)
18757 ==2782220== at 0x42B364: grub_hfsplus_btree_search (hfsplus.c:566)
18758 ==2782220== by 0x42B21D: grub_hfsplus_read_block (hfsplus.c:185)
18759 ==2782220== by 0x42A693: grub_fshelp_read_file (fshelp.c:386)
18760 ==2782220== by 0x42C598: grub_hfsplus_read_file (hfsplus.c:219)
18761 ==2782220== by 0x42C598: grub_hfsplus_mount (hfsplus.c:330)
18762 ==2782220== by 0x42B8C5: grub_hfsplus_dir (hfsplus.c:958)
18763 ==2782220== by 0x4C1AE6: grub_fs_probe (fs.c:73)
18764 ==2782220== by 0x407C94: grub_ls_list_files (ls.c:186)
18765 ==2782220== by 0x407C94: grub_cmd_ls (ls.c:284)
18766 ==2782220== by 0x4D7130: grub_extcmd_dispatcher (extcmd.c:55)
18767 ==2782220== by 0x4045A6: execute_command (grub-fstest.c:59)
18768 ==2782220== by 0x4045A6: fstest (grub-fstest.c:433)
18769 ==2782220== by 0x4045A6: main (grub-fstest.c:772)
18770 ==2782220== Uninitialised value was created by a heap allocation
18771 ==2782220== at 0x483C7F3: malloc (in /usr/lib/x86_64-linux-gnu/valgrind/vgpreload_memcheck-amd64-linux.so)
18772 ==2782220== by 0x4C0305: grub_malloc (mm.c:42)
18773 ==2782220== by 0x42C21D: grub_hfsplus_mount (hfsplus.c:239)
18774 ==2782220== by 0x42B8C5: grub_hfsplus_dir (hfsplus.c:958)
18775 ==2782220== by 0x4C1AE6: grub_fs_probe (fs.c:73)
18776 ==2782220== by 0x407C94: grub_ls_list_files (ls.c:186)
18777 ==2782220== by 0x407C94: grub_cmd_ls (ls.c:284)
18778 ==2782220== by 0x4D7130: grub_extcmd_dispatcher (extcmd.c:55)
18779 ==2782220== by 0x4045A6: execute_command (grub-fstest.c:59)
18780 ==2782220== by 0x4045A6: fstest (grub-fstest.c:433)
18781 ==2782220== by 0x4045A6: main (grub-fstest.c:772)
18782
18783 This happens when the process of reading the catalog file goes sufficiently
18784 wrong that there's an attempt to read the extent overflow file, which has
18785 not yet been loaded. Keep track of when the extent overflow file is
18786 fully loaded and refuse to use it before then.
18787
18788 The load valgrind doesn't like is btree->nodesize, and that's then used
18789 to allocate a data structure. It looks like there are subsequently a lot
18790 of reads based on that pointer so OOB reads are likely, and indeed crashes
18791 (albeit difficult-to-replicate ones) have been observed in fuzzing.
18792
18793 Reviewed-by: Daniel Kiper <[email protected]>
18794
18795 2021-03-02 Daniel Axtens <[email protected]>
18796
18797 fs/hfsplus: Don't fetch a key beyond the end of the node
18798 Otherwise you get a wild pointer, leading to a bunch of invalid reads.
18799 Check it falls inside the given node.
18800
18801 Reviewed-by: Daniel Kiper <[email protected]>
18802
18803 2021-03-02 Daniel Axtens <[email protected]>
18804
18805 fs/fshelp: Catch impermissibly large block sizes in read helper
18806 A fuzzed HFS+ filesystem had log2blocksize = 22. This gave
18807 log2blocksize + GRUB_DISK_SECTOR_BITS = 31. 1 << 31 = 0x80000000,
18808 which is -1 as an int. This caused some wacky behavior later on in
18809 the function, leading to out-of-bounds writes on the destination buffer.
18810
18811 Catch log2blocksize + GRUB_DISK_SECTOR_BITS >= 31. We could be stricter,
18812 but this is the minimum that will prevent integer size weirdness.
18813
18814 Reviewed-by: Daniel Kiper <[email protected]>
18815
18816 2021-03-02 Daniel Axtens <[email protected]>
18817
18818 term/gfxterm: Don't set up a font with glyphs that are too big
18819 Catch the case where we have a font so big that it causes the number of
18820 rows or columns to be 0. Currently we continue and allocate a
18821 virtual_screen.text_buffer of size 0. We then try to use that for glpyhs
18822 and things go badly.
18823
18824 On the emu platform, malloc() may give us a valid pointer, in which case
18825 we'll access heap memory which we shouldn't. Alternatively, it may give us
18826 NULL, in which case we'll crash. For other platforms, if I understand
18827 grub_memalign() correctly, we will receive a valid but small allocation
18828 that we will very likely later overrun.
18829
18830 Prevent the creation of a virtual screen that isn't at least 40 cols
18831 by 12 rows. This is arbitrary, but it seems that if your width or height
18832 is half a standard 80x24 terminal, you're probably going to struggle to
18833 read anything anyway.
18834
18835 Reviewed-by: Daniel Kiper <[email protected]>
18836
18837 2021-03-02 Daniel Axtens <[email protected]>
18838
18839 video/readers/jpeg: Don't decode data before start of stream
18840 When a start of stream marker is encountered, we call grub_jpeg_decode_sos()
18841 which allocates space for a bitmap.
18842
18843 When a restart marker is encountered, we call grub_jpeg_decode_data() which
18844 then fills in that bitmap.
18845
18846 If we get a restart marker before the start of stream marker, we will
18847 attempt to write to a bitmap_ptr that hasn't been allocated. Catch this
18848 and bail out. This fixes an attempt to write to NULL.
18849
18850 Reviewed-by: Daniel Kiper <[email protected]>
18851
18852 2021-03-02 Daniel Axtens <[email protected]>
18853
18854 video/readers/jpeg: Catch OOB reads/writes in grub_jpeg_decode_du()
18855 The key line is:
18856
18857 du[jpeg_zigzag_order[pos]] = val * (int) data->quan_table[qt][pos];
18858
18859 jpeg_zigzag_order is grub_uint8_t[64].
18860
18861 I don't understand JPEG decoders quite well enough to explain what's
18862 going on here. However, I observe sometimes pos=64, which leads to an
18863 OOB read of the jpeg_zigzag_order global then an OOB write to du.
18864 That leads to various unpleasant memory corruption conditions.
18865
18866 Catch where pos >= ARRAY_SIZE(jpeg_zigzag_order) and bail.
18867
18868 Reviewed-by: Daniel Kiper <[email protected]>
18869
18870 2021-03-02 Daniel Axtens <[email protected]>
18871
18872 video/readers/jpeg: Catch files with unsupported quantization or Huffman tables
18873 Our decoder only supports 2 quantization tables. If a file asks for
18874 a quantization table with index > 1, reject it.
18875
18876 Similarly, our decoder only supports 4 Huffman tables. If a file asks
18877 for a Huffman table with index > 3, reject it.
18878
18879 This fixes some out of bounds reads. It's not clear what degree of control
18880 over subsequent execution could be gained by someone who can carefully
18881 set up the contents of memory before loading an invalid JPEG file.
18882
18883 Reviewed-by: Daniel Kiper <[email protected]>
18884
18885 2021-03-02 Daniel Axtens <[email protected]>
18886
18887 kern/misc: Always set *end in grub_strtoull()
18888 Currently, if there is an error in grub_strtoull(), *end is not set.
18889 This differs from the usual behavior of strtoull(), and also means that
18890 some callers may use an uninitialized value for *end.
18891
18892 Set *end unconditionally.
18893
18894 Reviewed-by: Daniel Kiper <[email protected]>
18895
18896 2021-03-02 Daniel Axtens <[email protected]>
18897
18898 commands/menuentry: Fix quoting in setparams_prefix()
18899 Commit 9acdcbf32542 (use single quotes in menuentry setparams command)
18900 says that expressing a quoted single quote will require 3 characters. It
18901 actually requires (and always did require!) 4 characters:
18902
18903 str: a'b => a'\''b
18904 len: 3 => 6 (2 for the letters + 4 for the quote)
18905
18906 This leads to not allocating enough memory and thus out of bounds writes
18907 that have been observed to cause heap corruption.
18908
18909 Allocate 4 bytes for each single quote.
18910
18911 Commit 22e7dbb2bb81 (Fix quoting in legacy parser.) does the same
18912 quoting, but it adds 3 as extra overhead on top of the single byte that
18913 the quote already needs. So it's correct.
18914
18915 Fixes: 9acdcbf32542 (use single quotes in menuentry setparams command)
18916 Fixes: CVE-2021-20233
18917
18918 Reported-by: Daniel Axtens <[email protected]>
18919 Reviewed-by: Daniel Kiper <[email protected]>
18920
18921 2021-03-02 Daniel Axtens <[email protected]>
18922
18923 script/execute: Don't crash on a "for" loop with no items
18924 The following crashes the parser:
18925
18926 for x in; do
18927 0
18928 done
18929
18930 This is because grub_script_arglist_to_argv() doesn't consider the
18931 possibility that arglist is NULL. Catch that explicitly.
18932
18933 This avoids a NULL pointer dereference.
18934
18935 Reviewed-by: Daniel Kiper <[email protected]>
18936
18937 2021-03-02 Daniel Axtens <[email protected]>
18938
18939 lib/arg: Block repeated short options that require an argument
18940 Fuzzing found the following crash:
18941
18942 search -hhhhhhhhhhhhhf
18943
18944 We didn't allocate enough option space for 13 hints because the
18945 allocation code counts the number of discrete arguments (i.e. argc).
18946 However, the shortopt parsing code will happily keep processing
18947 a combination of short options without checking if those short
18948 options require an argument. This means you can easily end writing
18949 past the allocated option space.
18950
18951 This fixes a OOB write which can cause heap corruption.
18952
18953 Fixes: CVE-2021-20225
18954
18955 Reported-by: Daniel Axtens <[email protected]>
18956 Reviewed-by: Daniel Kiper <[email protected]>
18957
18958 2021-03-02 Daniel Axtens <[email protected]>
18959
18960 script/execute: Avoid crash when using "$#" outside a function scope
18961 "$#" represents the number of arguments to a function. It is only
18962 defined in a function scope, where "scope" is non-NULL. Currently,
18963 if we attempt to evaluate "$#" outside a function scope, "scope" will
18964 be NULL and we will crash with a NULL pointer dereference.
18965
18966 Do not attempt to count arguments for "$#" if "scope" is NULL. This
18967 will result in "$#" being interpreted as an empty string if evaluated
18968 outside a function scope.
18969
18970 Reviewed-by: Daniel Kiper <[email protected]>
18971
18972 2021-03-02 Daniel Axtens <[email protected]>
18973
18974 commands/ls: Require device_name is not NULL before printing
18975 This can be triggered with:
18976 ls -l (0 0*)
18977 and causes a NULL deref in grub_normal_print_device_info().
18978
18979 I'm not sure if there's any implication with the IEEE 1275 platform.
18980
18981 Reviewed-by: Daniel Kiper <[email protected]>
18982
18983 2021-03-02 Daniel Axtens <[email protected]>
18984
18985 script/execute: Fix NULL dereference in grub_script_execute_cmdline()
18986 Reviewed-by: Daniel Kiper <[email protected]>
18987
18988 2021-03-02 Darren Kenny <[email protected]>
18989
18990 util/glue-efi: Fix incorrect use of a possibly negative value
18991 It is possible for the ftell() function to return a negative value,
18992 although it is fairly unlikely here, we should be checking for
18993 a negative value before we assign it to an unsigned value.
18994
18995 Fixes: CID 73744
18996
18997 Reviewed-by: Daniel Kiper <[email protected]>
18998
18999 2021-03-02 Darren Kenny <[email protected]>
19000
19001 util/grub-editenv: Fix incorrect casting of a signed value
19002 The return value of ftell() may be negative (-1) on error. While it is
19003 probably unlikely to occur, we should not blindly cast to an unsigned
19004 value without first testing that it is not negative.
19005
19006 Fixes: CID 73856
19007
19008 Reviewed-by: Daniel Kiper <[email protected]>
19009
19010 2021-03-02 Daniel Kiper <[email protected]>
19011
19012 util/grub-install: Fix NULL pointer dereferences
19013 Two grub_device_open() calls does not have associated NULL checks
19014 for returned values. Fix that and appease the Coverity.
19015
19016 Fixes: CID 314583
19017
19018 Reviewed-by: Javier Martinez Canillas <[email protected]>
19019
19020 2021-03-02 Paulo Flabiano Smorigo <[email protected]>
19021
19022 loader/xnu: Check if pointer is NULL before using it
19023 Fixes: CID 73654
19024
19025 Reviewed-by: Daniel Kiper <[email protected]>
19026
19027 2021-03-02 Marco A Benatto <[email protected]>
19028
19029 loader/xnu: Free driverkey data when an error is detected in grub_xnu_writetree_toheap()
19030 ... to avoid memory leaks.
19031
19032 Fixes: CID 96640
19033
19034 Reviewed-by: Daniel Kiper <[email protected]>
19035
19036 2021-03-02 Darren Kenny <[email protected]>
19037
19038 loader/xnu: Fix memory leak
19039 The code here is finished with the memory stored in name, but it only
19040 frees it if there curvalue is valid, while it could actually free it
19041 regardless.
19042
19043 The fix is a simple relocation of the grub_free() to before the test
19044 of curvalue.
19045
19046 Fixes: CID 96646
19047
19048 Reviewed-by: Daniel Kiper <[email protected]>
19049
19050 2021-03-02 Darren Kenny <[email protected]>
19051
19052 loader/bsd: Check for NULL arg up-front
19053 The code in the next block suggests that it is possible for .set to be
19054 true but .arg may still be NULL.
19055
19056 This code assumes that it is never NULL, yet later is testing if it is
19057 NULL - that is inconsistent.
19058
19059 So we should check first if .arg is not NULL, and remove this check that
19060 is being flagged by Coverity since it is no longer required.
19061
19062 Fixes: CID 292471
19063
19064 Reviewed-by: Daniel Kiper <[email protected]>
19065
19066 2021-03-02 Darren Kenny <[email protected]>
19067
19068 gfxmenu/gui_list: Remove code that coverity is flagging as dead
19069 The test of value for NULL before calling grub_strdup() is not required,
19070 since the if condition prior to this has already tested for value being
19071 NULL and cannot reach this code if it is.
19072
19073 Fixes: CID 73659
19074
19075 Reviewed-by: Daniel Kiper <[email protected]>
19076
19077 2021-03-02 Darren Kenny <[email protected]>
19078
19079 video/readers/jpeg: Test for an invalid next marker reference from a jpeg file
19080 While it may never happen, and potentially could be caught at the end of
19081 the function, it is worth checking up front for a bad reference to the
19082 next marker just in case of a maliciously crafted file being provided.
19083
19084 Fixes: CID 73694
19085
19086 Reviewed-by: Daniel Kiper <[email protected]>
19087
19088 2021-03-02 Darren Kenny <[email protected]>
19089
19090 video/fb/video_fb: Fix possible integer overflow
19091 It is minimal possibility that the values being used here will overflow.
19092 So, change the code to use the safemath function grub_mul() to ensure
19093 that doesn't happen.
19094
19095 Fixes: CID 73761
19096
19097 Reviewed-by: Daniel Kiper <[email protected]>
19098
19099 2021-03-02 Darren Kenny <[email protected]>
19100
19101 video/fb/video_fb: Fix multiple integer overflows
19102 The calculation of the unsigned 64-bit value is being generated by
19103 multiplying 2, signed or unsigned, 32-bit integers which may overflow
19104 before promotion to unsigned 64-bit. Fix all of them.
19105
19106 Fixes: CID 73703, CID 73767, CID 73833
19107
19108 Reviewed-by: Daniel Kiper <[email protected]>
19109
19110 2021-03-02 Darren Kenny <[email protected]>
19111
19112 video/fb/fbfill: Fix potential integer overflow
19113 The multiplication of 2 unsigned 32-bit integers may overflow before
19114 promotion to unsigned 64-bit. We should ensure that the multiplication
19115 is done with overflow detection. Additionally, use grub_sub() for
19116 subtraction.
19117
19118 Fixes: CID 73640, CID 73697, CID 73702, CID 73823
19119
19120 Reviewed-by: Daniel Kiper <[email protected]>
19121
19122 2021-03-02 Darren Kenny <[email protected]>
19123
19124 video/efi_gop: Remove unnecessary return value of grub_video_gop_fill_mode_info()
19125 The return value of grub_video_gop_fill_mode_info() is never able to be
19126 anything other than GRUB_ERR_NONE. So, rather than continue to return
19127 a value and checking it each time, it is more correct to redefine the
19128 function to not return anything and remove checks of its return value
19129 altogether.
19130
19131 Fixes: CID 96701
19132
19133 Reviewed-by: Daniel Kiper <[email protected]>
19134
19135 2021-03-02 Darren Kenny <[email protected]>
19136
19137 commands/probe: Fix a resource leak when probing disks
19138 Every other return statement in this code is calling grub_device_close()
19139 to clean up dev before returning. This one should do that too.
19140
19141 Fixes: CID 292443
19142
19143 Reviewed-by: Daniel Kiper <[email protected]>
19144
19145 2021-03-02 Chris Coulson <[email protected]>
19146
19147 commands/hashsum: Fix a memory leak
19148 check_list() uses grub_file_getline(), which allocates a buffer.
19149 If the hash list file contains invalid lines, the function leaks
19150 this buffer when it returns an error.
19151
19152 Fixes: CID 176635
19153
19154 Reviewed-by: Daniel Kiper <[email protected]>
19155
19156 2021-03-02 Darren Kenny <[email protected]>
19157
19158 normal/completion: Fix leaking of memory when processing a completion
19159 It is possible for the code to reach the end of the function without
19160 freeing the memory allocated to argv and argc still to be 0.
19161
19162 We should always call grub_free(argv). The grub_free() will handle
19163 a NULL argument correctly if it reaches that code without the memory
19164 being allocated.
19165
19166 Fixes: CID 96672
19167
19168 Reviewed-by: Daniel Kiper <[email protected]>
19169
19170 2021-03-02 Darren Kenny <[email protected]>
19171
19172 syslinux: Fix memory leak while parsing
19173 In syslinux_parse_real() the 2 points where return is being called
19174 didn't release the memory stored in buf which is no longer required.
19175
19176 Fixes: CID 176634
19177
19178 Reviewed-by: Daniel Kiper <[email protected]>
19179
19180 2021-03-02 Darren Kenny <[email protected]>
19181
19182 libgcrypt/mpi: Fix possible NULL dereference
19183 The code in gcry_mpi_scan() assumes that buffer is not NULL, but there
19184 is no explicit check for that, so we add one.
19185
19186 Fixes: CID 73757
19187
19188 Reviewed-by: Daniel Kiper <[email protected]>
19189
19190 2021-03-02 Darren Kenny <[email protected]>
19191
19192 libgcrypt/mpi: Fix possible unintended sign extension
19193 The array of unsigned char gets promoted to a signed 32-bit int before
19194 it is finally promoted to a size_t. There is the possibility that this
19195 may result in the signed-bit being set for the intermediate signed
19196 32-bit int. We should ensure that the promotion is to the correct type
19197 before we bitwise-OR the values.
19198
19199 Fixes: CID 96697
19200
19201 Reviewed-by: Daniel Kiper <[email protected]>
19202
19203 2021-03-02 Darren Kenny <[email protected]>
19204
19205 affs: Fix memory leaks
19206 The node structure reference is being allocated but not freed if it
19207 reaches the end of the function. If any of the hooks had returned
19208 a non-zero value, then node would have been copied in to the context
19209 reference, but otherwise node is not stored and should be freed.
19210
19211 Similarly, the call to grub_affs_create_node() replaces the allocated
19212 memory in node with a newly allocated structure, leaking the existing
19213 memory pointed by node.
19214
19215 Finally, when dir->parent is set, then we again replace node with newly
19216 allocated memory, which seems unnecessary when we copy in the values
19217 from dir->parent immediately after.
19218
19219 Fixes: CID 73759
19220
19221 Reviewed-by: Daniel Kiper <[email protected]>
19222
19223 2021-03-02 Darren Kenny <[email protected]>
19224
19225 zfsinfo: Correct a check for error allocating memory
19226 While arguably the check for grub_errno is correct, we should really be
19227 checking the return value from the function since it is always possible
19228 that grub_errno was set elsewhere, making this code behave incorrectly.
19229
19230 Fixes: CID 73668
19231
19232 Reviewed-by: Daniel Kiper <[email protected]>
19233
19234 2021-03-02 Darren Kenny <[email protected]>
19235
19236 zfs: Fix possible integer overflows
19237 In all cases the problem is that the value being acted upon by
19238 a left-shift is a 32-bit number which is then being used in the
19239 context of a 64-bit number.
19240
19241 To avoid overflow we ensure that the number being shifted is 64-bit
19242 before the shift is done.
19243
19244 Fixes: CID 73684, CID 73695, CID 73764
19245
19246 Reviewed-by: Daniel Kiper <[email protected]>
19247
19248 2021-03-02 Paulo Flabiano Smorigo <[email protected]>
19249
19250 zfs: Fix resource leaks while constructing path
19251 There are several exit points in dnode_get_path() that are causing possible
19252 memory leaks.
19253
19254 In the while(1) the correct exit mechanism should not be to do a direct return,
19255 but to instead break out of the loop, setting err first if it is not already set.
19256
19257 The reason behind this is that the dnode_path is a linked list, and while doing
19258 through this loop, it is being allocated and built up - the only way to
19259 correctly unravel it is to traverse it, which is what is being done at the end
19260 of the function outside of the loop.
19261
19262 Several of the existing exit points correctly did a break, but not all so this
19263 change makes that more consistent and should resolve the leaking of memory as
19264 found by Coverity.
19265
19266 Fixes: CID 73741
19267
19268 Reviewed-by: Daniel Kiper <[email protected]>
19269
19270 2021-03-02 Darren Kenny <[email protected]>
19271
19272 zfs: Fix possible negative shift operation
19273 While it is possible for the return value from zfs_log2() to be zero
19274 (0), it is quite unlikely, given that the previous assignment to blksz
19275 is shifted up by SPA_MINBLOCKSHIFT (9) before 9 is subtracted at the
19276 assignment to epbs.
19277
19278 But, while unlikely during a normal operation, it may be that a carefully
19279 crafted ZFS filesystem could result in a zero (0) value to the
19280 dn_datalbkszsec field, which means that the shift left does nothing
19281 and assigns zero (0) to blksz, resulting in a negative epbs value.
19282
19283 Fixes: CID 73608
19284
19285 Reviewed-by: Daniel Kiper <[email protected]>
19286
19287 2021-03-02 Darren Kenny <[email protected]>
19288
19289 hfsplus: Check that the volume name length is valid
19290 HFS+ documentation suggests that the maximum filename and volume name is
19291 255 Unicode characters in length.
19292
19293 So, when converting from big-endian to little-endian, we should ensure
19294 that the name of the volume has a length that is between 0 and 255,
19295 inclusive.
19296
19297 Fixes: CID 73641
19298
19299 Reviewed-by: Daniel Kiper <[email protected]>
19300
19301 2021-03-02 Darren Kenny <[email protected]>
19302
19303 disk/cryptodisk: Fix potential integer overflow
19304 The encrypt and decrypt functions expect a grub_size_t. So, we need to
19305 ensure that the constant bit shift is using grub_size_t rather than
19306 unsigned int when it is performing the shift.
19307
19308 Fixes: CID 307788
19309
19310 Reviewed-by: Daniel Kiper <[email protected]>
19311
19312 2021-03-02 Darren Kenny <[email protected]>
19313
19314 disk/ldm: Fix memory leak on uninserted lv references
19315 The problem here is that the memory allocated to the variable lv is not
19316 yet inserted into the list that is being processed at the label fail2.
19317
19318 As we can already see at line 342, which correctly frees lv before going
19319 to fail2, we should also be doing that at these earlier jumps to fail2.
19320
19321 Fixes: CID 73824
19322
19323 Reviewed-by: Daniel Kiper <[email protected]>
19324
19325 2021-03-02 Paulo Flabiano Smorigo <[email protected]>
19326
19327 disk/ldm: If failed then free vg variable too
19328 Fixes: CID 73809
19329
19330 Reviewed-by: Daniel Kiper <[email protected]>
19331
19332 2021-03-02 Marco A Benatto <[email protected]>
19333
19334 disk/ldm: Make sure comp data is freed before exiting from make_vg()
19335 Several error handling paths in make_vg() do not free comp data before
19336 jumping to fail2 label and returning from the function. This will leak
19337 memory. So, let's fix all issues of that kind.
19338
19339 Fixes: CID 73804
19340
19341 Reviewed-by: Daniel Kiper <[email protected]>
19342
19343 2021-03-02 Darren Kenny <[email protected]>
19344
19345 kern/partition: Check for NULL before dereferencing input string
19346 There is the possibility that the value of str comes from an external
19347 source and continuing to use it before ever checking its validity is
19348 wrong. So, needs fixing.
19349
19350 Additionally, drop unneeded part initialization.
19351
19352 Fixes: CID 292444
19353
19354 Reviewed-by: Daniel Kiper <[email protected]>
19355
19356 2021-03-02 Darren Kenny <[email protected]>
19357
19358 zstd: Initialize seq_t structure fully
19359 While many compilers will initialize this to zero, not all will, so it
19360 is better to be sure that fields not being explicitly set are at known
19361 values, and there is code that checks this fields value elsewhere in the
19362 code.
19363
19364 Fixes: CID 292440
19365
19366 Reviewed-by: Daniel Kiper <[email protected]>
19367
19368 2021-03-02 Darren Kenny <[email protected]>
19369
19370 io/lzopio: Resolve unnecessary self-assignment errors
19371 These 2 assignments are unnecessary since they are just assigning
19372 to themselves.
19373
19374 Fixes: CID 73643
19375
19376 Reviewed-by: Daniel Kiper <[email protected]>
19377
19378 2021-03-02 Darren Kenny <[email protected]>
19379
19380 gnulib/regcomp: Fix uninitialized re_token
19381 This issue has been fixed in the latest version of gnulib, so to
19382 maintain consistency, I've backported that change rather than doing
19383 something different.
19384
19385 Fixes: CID 73828
19386
19387 Reviewed-by: Daniel Kiper <[email protected]>
19388
19389 2021-03-02 Darren Kenny <[email protected]>
19390
19391 gnulib/regexec: Fix possible null-dereference
19392 It appears to be possible that the mctx->state_log field may be NULL,
19393 and the name of this function, clean_state_log_if_needed(), suggests
19394 that it should be checking that it is valid to be cleaned before
19395 assuming that it does.
19396
19397 Fixes: CID 86720
19398
19399 Reviewed-by: Daniel Kiper <[email protected]>
19400
19401 2021-03-02 Darren Kenny <[email protected]>
19402
19403 gnulib/argp-help: Fix dereference of a possibly NULL state
19404 All other instances of call to __argp_failure() where there is
19405 a dgettext() call is first checking whether state is NULL before
19406 attempting to dereference it to get the root_argp->argp_domain.
19407
19408 Fixes: CID 292436
19409
19410 Reviewed-by: Daniel Kiper <[email protected]>
19411
19412 2021-03-02 Darren Kenny <[email protected]>
19413
19414 gnulib/regcomp: Fix uninitialized token structure
19415 The code is assuming that the value of br_token.constraint was
19416 initialized to zero when it wasn't.
19417
19418 While some compilers will ensure that, not all do, so it is better to
19419 fix this explicitly than leave it to chance.
19420
19421 Fixes: CID 73749
19422
19423 Reviewed-by: Daniel Kiper <[email protected]>
19424
19425 2021-03-02 Darren Kenny <[email protected]>
19426
19427 gnulib/regexec: Resolve unused variable
19428 This is a really minor issue where a variable is being assigned to but
19429 not checked before it is overwritten again.
19430
19431 The reason for this issue is that we are not building with DEBUG set and
19432 this in turn means that the assert() that reads the value of the
19433 variable match_last is being processed out.
19434
19435 The solution, move the assignment to match_last in to an ifdef DEBUG too.
19436
19437 Fixes: CID 292459
19438
19439 Reviewed-by: Daniel Kiper <[email protected]>
19440
19441 2021-03-02 Darren Kenny <[email protected]>
19442
19443 kern/efi/mm: Fix possible NULL pointer dereference
19444 The model of grub_efi_get_memory_map() is that if memory_map is NULL,
19445 then the purpose is to discover how much memory should be allocated to
19446 it for the subsequent call.
19447
19448 The problem here is that with grub_efi_is_finished set to 1, there is no
19449 check at all that the function is being called with a non-NULL memory_map.
19450
19451 While this MAY be true, we shouldn't assume it.
19452
19453 The solution to this is to behave as expected, and if memory_map is NULL,
19454 then don't try to use it and allow memory_map_size to be filled in, and
19455 return 0 as is done later in the code if the buffer is too small (or NULL).
19456
19457 Additionally, drop unneeded ret = 1.
19458
19459 Fixes: CID 96632
19460
19461 Reviewed-by: Daniel Kiper <[email protected]>
19462
19463 2021-03-02 Darren Kenny <[email protected]>
19464
19465 kern/efi: Fix memory leak on failure
19466 Free the memory allocated to name before returning on failure.
19467
19468 Fixes: CID 296222
19469
19470 Reviewed-by: Daniel Kiper <[email protected]>
19471
19472 2021-03-02 Darren Kenny <[email protected]>
19473
19474 kern/parser: Fix resource leak if argc == 0
19475 After processing the command-line yet arriving at the point where we are
19476 setting argv, we are allocating memory, even if argc == 0, which makes
19477 no sense since we never put anything into the allocated argv.
19478
19479 The solution is to simply return that we've successfully processed the
19480 arguments but that argc == 0, and also ensure that argv is NULL when
19481 we're not allocating anything in it.
19482
19483 There are only 2 callers of this function, and both are handling a zero
19484 value in argc assuming nothing is allocated in argv.
19485
19486 Fixes: CID 96680
19487
19488 Reviewed-by: Daniel Kiper <[email protected]>
19489
19490 2021-03-02 Darren Kenny <[email protected]>
19491
19492 net/tftp: Fix dangling memory pointer
19493 The static code analysis tool, Parfait, reported that the valid of
19494 file->data was left referencing memory that was freed by the call to
19495 grub_free(data) where data was initialized from file->data.
19496
19497 To ensure that there is no unintentional access to this memory
19498 referenced by file->data we should set the pointer to NULL.
19499
19500 Reviewed-by: Daniel Kiper <[email protected]>
19501
19502 2021-03-02 Darren Kenny <[email protected]>
19503
19504 net/net: Fix possible dereference to of a NULL pointer
19505 It is always possible that grub_zalloc() could fail, so we should check for
19506 a NULL return. Otherwise we run the risk of dereferencing a NULL pointer.
19507
19508 Fixes: CID 296221
19509
19510 Reviewed-by: Daniel Kiper <[email protected]>
19511
19512 2021-03-02 Darren Kenny <[email protected]>
19513
19514 mmap: Fix memory leak when iterating over mapped memory
19515 When returning from grub_mmap_iterate() the memory allocated to present
19516 is not being released causing it to leak.
19517
19518 Fixes: CID 96655
19519
19520 Reviewed-by: Daniel Kiper <[email protected]>
19521
19522 2021-03-02 Javier Martinez Canillas <[email protected]>
19523
19524 usb: Avoid possible out-of-bound accesses caused by malicious devices
19525 The maximum number of configurations and interfaces are fixed but there is
19526 no out-of-bound checking to prevent a malicious USB device to report large
19527 values for these and cause accesses outside the arrays' memory.
19528
19529 Fixes: CVE-2020-25647
19530
19531 Reported-by: Joseph Tartaro <[email protected]>
19532 Reported-by: Ilja Van Sprundel <[email protected]>
19533 Reviewed-by: Daniel Kiper <[email protected]>
19534
19535 2021-03-02 Javier Martinez Canillas <[email protected]>
19536
19537 dl: Only allow unloading modules that are not dependencies
19538 When a module is attempted to be removed its reference counter is always
19539 decremented. This means that repeated rmmod invocations will cause the
19540 module to be unloaded even if another module depends on it.
19541
19542 This may lead to a use-after-free scenario allowing an attacker to execute
19543 arbitrary code and by-pass the UEFI Secure Boot protection.
19544
19545 While being there, add the extern keyword to some function declarations in
19546 that header file.
19547
19548 Fixes: CVE-2020-25632
19549
19550 Reported-by: Chris Coulson <[email protected]>
19551 Reviewed-by: Daniel Kiper <[email protected]>
19552
19553 2021-03-02 Javier Martinez Canillas <[email protected]>
19554
19555 docs: Document the cutmem command
19556 The command is not present in the docs/grub.texi user documentation.
19557
19558 Reported-by: Daniel Kiper <[email protected]>
19559 Reviewed-by: Javier Martinez Canillas <[email protected]>
19560
19561 2021-03-02 Javier Martinez Canillas <[email protected]>
19562
19563 loader/xnu: Don't allow loading extension and packages when locked down
19564 The shim_lock verifier validates the XNU kernels but no its extensions
19565 and packages. Prevent these to be loaded when the GRUB is locked down.
19566
19567 Reviewed-by: Daniel Kiper <[email protected]>
19568
19569 2021-03-02 Javier Martinez Canillas <[email protected]>
19570
19571 gdb: Restrict GDB access when locked down
19572 The gdbstub* commands allow to start and control a GDB stub running on
19573 local host that can be used to connect from a remote debugger. Restrict
19574 this functionality when the GRUB is locked down.
19575
19576 Reviewed-by: Daniel Kiper <[email protected]>
19577
19578 2021-03-02 Javier Martinez Canillas <[email protected]>
19579
19580 commands/hdparm: Restrict hdparm command when locked down
19581 The command can be used to get/set ATA disk parameters. Some of these can
19582 be dangerous since change the disk behavior. Restrict it when locked down.
19583
19584 Reviewed-by: Daniel Kiper <[email protected]>
19585
19586 2021-03-02 Javier Martinez Canillas <[email protected]>
19587
19588 commands/setpci: Restrict setpci command when locked down
19589 This command can set PCI devices register values, which makes it dangerous
19590 in a locked down configuration. Restrict it so can't be used on this setup.
19591
19592 Reviewed-by: Daniel Kiper <[email protected]>
19593
19594 2021-03-02 Javier Martinez Canillas <[email protected]>
19595
19596 commands: Restrict commands that can load BIOS or DT blobs when locked down
19597 There are some more commands that should be restricted when the GRUB is
19598 locked down. Following is the list of commands and reasons to restrict:
19599
19600 * fakebios: creates BIOS-like structures for backward compatibility with
19601 existing OSes. This should not be allowed when locked down.
19602
19603 * loadbios: reads a BIOS dump from storage and loads it. This action
19604 should not be allowed when locked down.
19605
19606 * devicetree: loads a Device Tree blob and passes it to the OS. It replaces
19607 any Device Tree provided by the firmware. This also should
19608 not be allowed when locked down.
19609
19610 Reviewed-by: Daniel Kiper <[email protected]>
19611
19612 2021-03-02 Javier Martinez Canillas <[email protected]>
19613
19614 mmap: Don't register cutmem and badram commands when lockdown is enforced
19615 The cutmem and badram commands can be used to remove EFI memory regions
19616 and potentially disable the UEFI Secure Boot. Prevent the commands to be
19617 registered if the GRUB is locked down.
19618
19619 Fixes: CVE-2020-27779
19620
19621 Reported-by: Teddy Reed <[email protected]>
19622 Reviewed-by: Daniel Kiper <[email protected]>
19623
19624 2021-03-02 Javier Martinez Canillas <[email protected]>
19625
19626 acpi: Don't register the acpi command when locked down
19627 The command is not allowed when lockdown is enforced. Otherwise an
19628 attacker can instruct the GRUB to load an SSDT table to overwrite
19629 the kernel lockdown configuration and later load and execute
19630 unsigned code.
19631
19632 Fixes: CVE-2020-14372
19633
19634 Reported-by: Máté Kukri <[email protected]>
19635 Reviewed-by: Daniel Kiper <[email protected]>
19636
19637 2021-03-02 Javier Martinez Canillas <[email protected]>
19638
19639 efi: Use grub_is_lockdown() instead of hardcoding a disabled modules list
19640 Now the GRUB can check if it has been locked down and this can be used to
19641 prevent executing commands that can be utilized to circumvent the UEFI
19642 Secure Boot mechanisms. So, instead of hardcoding a list of modules that
19643 have to be disabled, prevent the usage of commands that can be dangerous.
19644
19645 This not only allows the commands to be disabled on other platforms, but
19646 also properly separate the concerns. Since the shim_lock verifier logic
19647 should be only about preventing to run untrusted binaries and not about
19648 defining these kind of policies.
19649
19650 Reviewed-by: Daniel Kiper <[email protected]>
19651
19652 2021-03-02 Javier Martinez Canillas <[email protected]>
19653
19654 efi: Lockdown the GRUB when the UEFI Secure Boot is enabled
19655 If the UEFI Secure Boot is enabled then the GRUB must be locked down
19656 to prevent executing code that can potentially be used to subvert its
19657 verification mechanisms.
19658
19659 Reviewed-by: Daniel Kiper <[email protected]>
19660
19661 2021-03-02 Javier Martinez Canillas <[email protected]>
19662
19663 kern/lockdown: Set a variable if the GRUB is locked down
19664 It may be useful for scripts to determine whether the GRUB is locked
19665 down or not. Add the lockdown variable which is set to "y" when the GRUB
19666 is locked down.
19667
19668 Suggested-by: Dimitri John Ledkov <[email protected]>
19669 Reviewed-by: Daniel Kiper <[email protected]>
19670
19671 2021-03-02 Javier Martinez Canillas <[email protected]>
19672
19673 kern: Add lockdown support
19674 When the GRUB starts on a secure boot platform, some commands can be
19675 used to subvert the protections provided by the verification mechanism and
19676 could lead to booting untrusted system.
19677
19678 To prevent that situation, allow GRUB to be locked down. That way the code
19679 may check if GRUB has been locked down and further restrict the commands
19680 that are registered or what subset of their functionality could be used.
19681
19682 The lockdown support adds the following components:
19683
19684 * The grub_lockdown() function which can be used to lockdown GRUB if,
19685 e.g., UEFI Secure Boot is enabled.
19686
19687 * The grub_is_lockdown() function which can be used to check if the GRUB
19688 was locked down.
19689
19690 * A verifier that flags OS kernels, the GRUB modules, Device Trees and ACPI
19691 tables as GRUB_VERIFY_FLAGS_DEFER_AUTH to defer verification to other
19692 verifiers. These files are only successfully verified if another registered
19693 verifier returns success. Otherwise, the whole verification process fails.
19694
19695 For example, PE/COFF binaries verification can be done by the shim_lock
19696 verifier which validates the signatures using the shim_lock protocol.
19697 However, the verification is not deferred directly to the shim_lock verifier.
19698 The shim_lock verifier is hooked into the verification process instead.
19699
19700 * A set of grub_{command,extcmd}_lockdown functions that can be used by
19701 code registering command handlers, to only register unsafe commands if
19702 the GRUB has not been locked down.
19703
19704 Reviewed-by: Daniel Kiper <[email protected]>
19705
19706 2021-03-02 Marco A Benatto <[email protected]>
19707
19708 efi: Move the shim_lock verifier to the GRUB core
19709 Move the shim_lock verifier from its own module into the core image. The
19710 Secure Boot lockdown mechanism has the intent to prevent the load of any
19711 unsigned code or binary when Secure Boot is enabled.
19712
19713 The reason is that GRUB must be able to prevent executing untrusted code
19714 if UEFI Secure Boot is enabled, without depending on external modules.
19715
19716 Reviewed-by: Daniel Kiper <[email protected]>
19717
19718 2021-03-02 Marco A Benatto <[email protected]>
19719
19720 verifiers: Move verifiers API to kernel image
19721 Move verifiers API from a module to the kernel image, so it can be
19722 used there as well. There are no functional changes in this patch.
19723
19724 Reviewed-by: Daniel Kiper <[email protected]>
19725
19726 2020-12-18 Glenn Washburn <[email protected]>
19727
19728 docs: Add documentation of disk size limitations
19729 Document the artificially imposed 1 EiB disk size limit and size limitations
19730 with LUKS volumes.
19731
19732 Fix a few punctuation issues.
19733
19734 Reviewed-by: Daniel Kiper <[email protected]>
19735
19736 2020-12-18 Glenn Washburn <[email protected]>
19737
19738 luks2: Use grub_log2ull() to calculate log_sector_size and improve readability
19739 Reviewed-by: Daniel Kiper <[email protected]>
19740
19741 misc: Add grub_log2ull() macro for calculating log base 2 of 64-bit integers
19742 Reviewed-by: Daniel Kiper <[email protected]>
19743
19744 2020-12-18 Glenn Washburn <[email protected]>
19745
19746 mips: Enable __clzdi2()
19747 This patch is similar to commit 9dab2f51e (sparc: Enable __clzsi2() and
19748 __clzdi2()) but for MIPS target and __clzdi2() only, __clzsi2() was
19749 already enabled.
19750
19751 Suggested-by: Daniel Kiper <[email protected]>
19752 Reviewed-by: Daniel Kiper <[email protected]>
19753
19754 2020-12-18 Glenn Washburn <[email protected]>
19755
19756 luks2: Better error handling when setting up the cryptodisk
19757 Do some sanity checking on data coming from the LUKS2 header. If segment.size
19758 is "dynamic", verify that the offset is not past the end of disk. Otherwise,
19759 check for errors from grub_strtoull() when converting segment size from
19760 string. If a GRUB_ERR_BAD_NUMBER error was returned, then the string was
19761 not a valid parsable number, so skip the key. If GRUB_ERR_OUT_OF_RANGE was
19762 returned, then there was an overflow in converting to a 64-bit unsigned
19763 integer. So this could be a very large disk (perhaps large RAID array).
19764 In this case skip the key too. Additionally, enforce some other limits
19765 and fail if needed.
19766
19767 Reviewed-by: Daniel Kiper <[email protected]>
19768
19769 2020-12-18 Glenn Washburn <[email protected]>
19770
19771 luks2: Do not handle disks of size GRUB_DISK_SIZE_UNKNOWN for now
19772 Check to make sure that source disk has a known size. If not, print
19773 a message and return error. There are 4 cases where GRUB_DISK_SIZE_UNKNOWN
19774 is set (biosdisk, obdisk, ofdisk, and uboot), and in all those cases
19775 processing continues. So this is probably a bit conservative. However,
19776 3 of the cases seem pathological, and the other, biosdisk, happens when
19777 booting from a CD-ROM. Since I doubt booting from a LUKS2 volume on
19778 a CD-ROM is a big use case, we'll error until someone complains.
19779
19780 Reviewed-by: Daniel Kiper <[email protected]>
19781
19782 2020-12-18 Glenn Washburn <[email protected]>
19783
19784 luks2: Convert to crypt sectors from GRUB native sectors
19785 The function grub_disk_native_sectors(source) returns the number of sectors
19786 of source in GRUB native (512-byte) sectors, not source sized sectors. So
19787 the conversion needs to use GRUB_DISK_SECTOR_BITS, the GRUB native sector
19788 size.
19789
19790 Reviewed-by: Daniel Kiper <[email protected]>
19791
19792 2020-12-12 Glenn Washburn <[email protected]>
19793
19794 luks2: Error check segment.sector_size
19795 Reviewed-by: Daniel Kiper <[email protected]>
19796
19797 2020-12-12 Glenn Washburn <[email protected]>
19798
19799 cryptodisk: Properly handle non-512 byte sized sectors
19800 By default, dm-crypt internally uses an IV that corresponds to 512-byte
19801 sectors, even when a larger sector size is specified. What this means is
19802 that when using a larger sector size, the IV is incremented every sector.
19803 However, the amount the IV is incremented is the number of 512 byte blocks
19804 in a sector (i.e. 8 for 4K sectors). Confusingly the IV does not correspond
19805 to the number of, for example, 4K sectors. So each 512 byte cipher block in
19806 a sector will be encrypted with the same IV and the IV will be incremented
19807 afterwards by the number of 512 byte cipher blocks in the sector.
19808
19809 There are some encryption utilities which do it the intuitive way and have
19810 the IV equal to the sector number regardless of sector size (ie. the fifth
19811 sector would have an IV of 4 for each cipher block). And this is supported
19812 by dm-crypt with the iv_large_sectors option and also cryptsetup as of 2.3.3
19813 with the --iv-large-sectors, though not with LUKS headers (only with --type
19814 plain). However, support for this has not been included as grub does not
19815 support plain devices right now.
19816
19817 One gotcha here is that the encrypted split keys are encrypted with a hard-
19818 coded 512-byte sector size. So even if your data is encrypted with 4K sector
19819 sizes, the split key encrypted area must be decrypted with a block size of
19820 512 (ie the IV increments every 512 bytes). This made these changes less
19821 aesthetically pleasing than desired.
19822
19823 Reviewed-by: Daniel Kiper <[email protected]>
19824
19825 2020-12-12 Glenn Washburn <[email protected]>
19826
19827 luks2: grub_cryptodisk_t->total_sectors is the max number of device native sectors
19828 We need to convert the sectors from the size of the underlying device to the
19829 cryptodisk sector size; segment.size is in bytes which need to be converted
19830 to cryptodisk sectors as well.
19831
19832 Also, removed an empty statement.
19833
19834 Reviewed-by: Daniel Kiper <[email protected]>
19835
19836 2020-12-12 Glenn Washburn <[email protected]>
19837
19838 cryptodisk: Add macros GRUB_TYPE_U_MAX/MIN(type) to replace literals
19839 Add GRUB_TYPE_U_MAX/MIN(type) macros to get the max/min values for an
19840 unsigned number with size of type.
19841
19842 Reviewed-by: Daniel Kiper <[email protected]>
19843
19844 2020-12-12 Glenn Washburn <[email protected]>
19845
19846 cryptodisk: Add macro GRUB_TYPE_BITS() to replace some literals
19847 The new macro GRUB_TYPE_BITS(type) returns the number of bits
19848 allocated for type.
19849
19850 Reviewed-by: Daniel Kiper <[email protected]>
19851
19852 2020-12-12 Glenn Washburn <[email protected]>
19853
19854 luks2: Add string "index" to user strings using a json index
19855 This allows error messages to be more easily distinguishable between indexes
19856 and slot keys. The former include the string "index" in the error/debug
19857 string, and the later are surrounded in quotes.
19858
19859 Reviewed-by: Daniel Kiper <[email protected]>
19860
19861 2020-12-12 Glenn Washburn <[email protected]>
19862
19863 luks2: Rename json index variables to names that they are obviously json indexes
19864 Reviewed-by: Daniel Kiper <[email protected]>
19865
19866 2020-12-12 Glenn Washburn <[email protected]>
19867
19868 luks2: Use more intuitive object name instead of json index in user messages
19869 Use the object name in the json array rather than the 0 based index in the
19870 json array for keyslots, segments, and digests. This is less confusing for
19871 the end user. For example, say you have a LUKS2 device with a key in slot 1
19872 and slot 4. When using the password for slot 4 to unlock the device, the
19873 messages using the index of the keyslot will mention keyslot 1 (its a
19874 zero-based index). Furthermore, with this change the keyslot number will
19875 align with the number used to reference the keyslot when using the
19876 --key-slot argument to cryptsetup.
19877
19878 Reviewed-by: Daniel Kiper <[email protected]>
19879
19880 2020-12-12 Glenn Washburn <[email protected]>
19881
19882 luks2: Add idx member to struct grub_luks2_keyslot/segment/digest
19883 This allows code using these structs to know the named key associated with
19884 these json data structures. In the future we can use these to provide better
19885 error messages to the user.
19886
19887 Get rid of idx local variable in luks2_get_keyslot() which was overloaded to
19888 be used for both keyslot and segment slot keys.
19889
19890 Reviewed-by: Daniel Kiper <[email protected]>
19891
19892 2020-12-12 Glenn Washburn <[email protected]>
19893
19894 luks2: Make sure all fields of output argument in luks2_parse_digest() are written to
19895 We should assume that the output argument "out" is uninitialized and could
19896 have random data. So, make sure to initialize the segments and keyslots bit
19897 fields because potentially not all bits of those fields are written to.
19898 Otherwise, the digest could say it belongs to keyslots and segments that it
19899 does not.
19900
19901 Reviewed-by: Patrick Steinhardt <[email protected]>
19902 Reviewed-by: Daniel Kiper <[email protected]>
19903
19904 2020-12-12 Glenn Washburn <[email protected]>
19905
19906 luks2: Remove unused argument in grub_error() call
19907 Reviewed-by: Patrick Steinhardt <[email protected]>
19908 Reviewed-by: Daniel Kiper <[email protected]>
19909
19910 luks2: Convert 8 spaces to tabs
19911 Reviewed-by: Patrick Steinhardt <[email protected]>
19912 Reviewed-by: Daniel Kiper <[email protected]>
19913
19914 2020-12-12 Glenn Washburn <[email protected]>
19915
19916 misc: Add parentheses around ALIGN_UP() and ALIGN_DOWN() arguments
19917 This ensures that expected order of operations is preserved when arguments
19918 are expressions.
19919
19920 Reviewed-by: Patrick Steinhardt <[email protected]>
19921 Reviewed-by: Daniel Kiper <[email protected]>
19922
19923 2020-12-12 Glenn Washburn <[email protected]>
19924
19925 disk: Rename grub_disk_get_size() to grub_disk_native_sectors()
19926 The function grub_disk_get_size() is confusingly named because it actually
19927 returns a sector count where the sectors are sized in the GRUB native sector
19928 size. Rename to something more appropriate.
19929
19930 Suggested-by: Daniel Kiper <[email protected]>
19931 Reviewed-by: Patrick Steinhardt <[email protected]>
19932 Reviewed-by: Daniel Kiper <[email protected]>
19933
19934 2020-12-12 Glenn Washburn <[email protected]>
19935
19936 loopback: Do not automaticaly replace existing loopback dev, error instead
19937 If there is a loopback device with the same name as the one to be created,
19938 instead of closing the old one and replacing it with the new one, return an
19939 error instead. If the loopback device was created, its probably being used
19940 by something and just replacing it may cause GRUB to crash unexpectedly.
19941 This fixes obvious problems like "loopback d (d)/somefile". Its not too
19942 onerous to force the user to delete the loopback first with the "-d" switch.
19943
19944 Reviewed-by: Daniel Kiper <[email protected]>
19945
19946 2020-12-12 Glenn Washburn <[email protected]>
19947
19948 disk: Move hardcoded max disk size literal to a GRUB_DISK_MAX_SECTORS in disk.h
19949 There is a hardcoded maximum disk size that can be read or written from,
19950 currently set at 1 EiB in grub_disk_adjust_range(). Move the literal into a
19951 macro in disk.h, so our assumptions are more visible. This hard coded limit
19952 does not prevent using larger disks, just GRUB won't read/write past the
19953 limit. The comment accompanying this restriction didn't quite make sense to
19954 me, so its been modified too.
19955
19956 Reviewed-by: Daniel Kiper <[email protected]>
19957
19958 2020-12-12 Glenn Washburn <[email protected]>
19959
19960 fs: Fix block lists not being able to address to end of disk sometimes
19961 When checking if a block list goes past the end of the disk, make sure
19962 the total size of the disk is in GRUB native sector sizes, otherwise there
19963 will be blocks at the end of the disk inaccessible by block lists.
19964
19965 Reviewed-by: Daniel Kiper <[email protected]>
19966
19967 2020-12-12 Vladimir Serbinenko <[email protected]>
19968
19969 mbr: Document new limitations on MBR gap support
19970 Reviewed-by: Daniel Kiper <[email protected]>
19971
19972 2020-12-12 Vladimir Serbinenko <[email protected]>
19973
19974 mbr: Warn if MBR gap is small and user uses advanced modules
19975 We don't want to support small MBR gap in pair with anything but the
19976 simplest config of biosdisk + part_msdos + simple filesystem. In this
19977 path "simple filesystems" are all current filesystems except ZFS and
19978 Btrfs.
19979
19980 Reviewed-by: Daniel Kiper <[email protected]>
19981
19982 2020-12-12 Tianjia Zhang <[email protected]>
19983
19984 efi/tpm: Extract duplicate code into independent functions
19985 Part of the code logic for processing the return value of efi
19986 log_extend_event is repetitive and complicated. Extract the
19987 repetitive code into an independent function.
19988
19989 Reviewed-by: Daniel Kiper <[email protected]>
19990
19991 2020-12-12 Tianjia Zhang <[email protected]>
19992
19993 efi/tpm: Add debug information for device protocol and eventlog
19994 Add a number of debug logs to the tpm module. The condition tag
19995 for opening debugging is "tpm". On TPM machines, this will bring
19996 great convenience to diagnosis and debugging.
19997
19998 Reviewed-by: Daniel Kiper <[email protected]>
19999
20000 2020-12-12 Daniel Kiper <[email protected]>
20001
20002 loader/linux: Report the UEFI Secure Boot status to the Linux kernel
20003 Now that the GRUB has a grub_efi_get_secureboot() function to check the
20004 UEFI Secure Boot status, use it to report that to the Linux kernel.
20005
20006 Reviewed-by: Daniel Kiper <[email protected]>
20007
20008 2020-12-12 Javier Martinez Canillas <[email protected]>
20009
20010 efi: Only register shim_lock verifier if shim_lock protocol is found and SB enabled
20011 The shim_lock module registers a verifier to call shim's verify, but the
20012 handler is registered even when the shim_lock protocol was not installed.
20013
20014 This doesn't cause a NULL pointer dereference in shim_lock_write() because
20015 the shim_lock_init() function just returns GRUB_ERR_NONE if sl isn't set.
20016
20017 But in that case there's no point to even register the shim_lock verifier
20018 since won't do anything. Additionally, it is only useful when Secure Boot
20019 is enabled.
20020
20021 Finally, don't assume that the shim_lock protocol will always be present
20022 when the shim_lock_write() function is called, and check for it on every
20023 call to this function.
20024
20025 Reported-by: Michael Chang <[email protected]>
20026 Reported-by: Peter Jones <[email protected]>
20027 Reviewed-by: Daniel Kiper <[email protected]>
20028
20029 2020-12-11 Daniel Kiper <[email protected]>
20030
20031 efi: Add secure boot detection
20032 Introduce grub_efi_get_secureboot() function which returns whether
20033 UEFI Secure Boot is enabled or not on UEFI systems.
20034
20035 Reviewed-by: Daniel Kiper <[email protected]>
20036
20037 2020-12-11 Daniel Kiper <[email protected]>
20038
20039 efi: Add a function to read EFI variables with attributes
20040 It will be used to properly detect and report UEFI Secure Boot status to
20041 the x86 Linux kernel. The functionality will be added by subsequent patches.
20042
20043 Reviewed-by: Daniel Kiper <[email protected]>
20044
20045 2020-12-11 Daniel Kiper <[email protected]>
20046
20047 efi: Return grub_efi_status_t from grub_efi_get_variable()
20048 This is needed to properly detect and report UEFI Secure Boot status
20049 to the x86 Linux kernel. The functionality will be added by subsequent
20050 patches.
20051
20052 Reviewed-by: Daniel Kiper <[email protected]>
20053
20054 2020-12-11 Daniel Kiper <[email protected]>
20055
20056 efi: Make shim_lock GUID and protocol type public
20057 The GUID will be used to properly detect and report UEFI Secure Boot
20058 status to the x86 Linux kernel. The functionality will be added by
20059 subsequent patches. The shim_lock protocol type is made public for
20060 completeness.
20061
20062 Additionally, fix formatting of four preceding GUIDs.
20063
20064 Reviewed-by: Daniel Kiper <[email protected]>
20065
20066 2020-12-11 Javier Martinez Canillas <[email protected]>
20067
20068 arm/term: Fix linking error due multiple ps2_state definitions
20069 When building with --target=arm-linux-gnu --with-platform=coreboot
20070 a linking error occurs caused by multiple definitions of the
20071 ps2_state variable.
20072
20073 Mark them as static since they aren't used outside their compilation unit.
20074
20075 Reviewed-by: Daniel Kiper <[email protected]>
20076
20077 2020-12-11 Javier Martinez Canillas <[email protected]>
20078
20079 include/grub/i386/linux.h: Include missing <grub/types.h> header
20080 This header uses types defined in <grub/types.h> but does not include it,
20081 which leads to compile errors like the following:
20082
20083 In file included from ../include/grub/cpu/linux.h:19,
20084 from kern/efi/sb.c:21:
20085 ../include/grub/i386/linux.h:80:3: error: unknown type name ‘grub_uint64_t’
20086 80 | grub_uint64_t addr;
20087
20088 Reviewed-by: Daniel Kiper <[email protected]>
20089
20090 2020-12-11 Javier Martinez Canillas <[email protected]>
20091
20092 i386: Don't include <grub/cpu/linux.h> in coreboot and ieee1275 startup.S
20093 Nothing defined in the header file is used in the assembly code but it
20094 may lead to build errors if some headers are included through this and
20095 contains definitions that are not recognized by the assembler, e.g.:
20096
20097 ../include/grub/types.h: Assembler messages:
20098 ../include/grub/types.h:76: Error: no such instruction: `typedef signed char grub_int8_t'
20099 ../include/grub/types.h:77: Error: no such instruction: `typedef short grub_int16_t'
20100 ../include/grub/types.h:78: Error: no such instruction: `typedef int grub_int32_t'
20101
20102 Reviewed-by: Daniel Kiper <[email protected]>
20103
20104 2020-11-20 Glenn Washburn <[email protected]>
20105
20106 luks2: Rename index variable "j" to "i" in luks2_get_keyslot()
20107 Looping variable "j" was named such because the variable name "i" was taken.
20108 Since "i" has been renamed in the previous patch, we can rename "j" to "i".
20109
20110 Reviewed-by: Patrick Steinhardt <[email protected]>
20111 Reviewed-by: Daniel Kiper <[email protected]>
20112
20113 2020-11-20 Glenn Washburn <[email protected]>
20114
20115 luks2: Rename variable "i" to "keyslot_idx" in luks2_get_keyslot()
20116 Variables named "i" are usually looping variables. So, rename it to
20117 "keyslot_idx" to ease luks2_get_keyslot() reading.
20118
20119 Reviewed-by: Patrick Steinhardt <[email protected]>
20120 Reviewed-by: Daniel Kiper <[email protected]>
20121
20122 2020-11-20 Glenn Washburn <[email protected]>
20123
20124 luks2: Use correct index variable when looping in luks2_get_keyslot()
20125 The loop variable "j" should be used to index the digests and segments json
20126 array, instead of the variable "i", which is the keyslot index.
20127
20128 Reviewed-by: Patrick Steinhardt <[email protected]>
20129 Reviewed-by: Daniel Kiper <[email protected]>
20130
20131 2020-11-20 Glenn Washburn <[email protected]>
20132
20133 luks2: Rename source disk variable named "disk" to "source" as in luks.c
20134 This makes it more obvious to the reader that the disk referred to is the
20135 source disk, as opposed to say the disk holding the cryptodisk.
20136
20137 Reviewed-by: Patrick Steinhardt <[email protected]>
20138 Reviewed-by: Daniel Kiper <[email protected]>
20139
20140 2020-11-20 Glenn Washburn <[email protected]>
20141
20142 cryptodisk: Rename "offset" in grub_cryptodisk_t to "offset_sectors"
20143 This makes it clear that the offset represents sectors, not bytes, in
20144 order to improve readability.
20145
20146 Reviewed-by: Patrick Steinhardt <[email protected]>
20147 Reviewed-by: Daniel Kiper <[email protected]>
20148
20149 2020-11-20 Glenn Washburn <[email protected]>
20150
20151 cryptodisk: Rename "total_length" field in grub_cryptodisk_t to "total_sectors"
20152 This creates an alignment with grub_disk_t naming of the same field and is
20153 more intuitive as to how it should be used.
20154
20155 Reviewed-by: Patrick Steinhardt <[email protected]>
20156 Reviewed-by: Daniel Kiper <[email protected]>
20157
20158 2020-11-20 Glenn Washburn <[email protected]>
20159
20160 types: Define GRUB_CHAR_BIT based on compiler macro instead of using literal
20161 Reviewed-by: Daniel Kiper <[email protected]>
20162
20163 2020-11-20 Javier Martinez Canillas <[email protected]>
20164
20165 include/grub/arm64/linux.h: Include missing <grub/types.h> header
20166 This header uses types defined in <grub/types.h> but does not include it,
20167 which leads to compile errors like the following:
20168
20169 ../include/grub/cpu/linux.h:27:3: error: unknown type name ‘grub_uint32_t’
20170 27 | grub_uint32_t code0; /* Executable code */
20171 | ^~~~~~~~~~~~~
20172
20173 Reviewed-by: Daniel Kiper <[email protected]>
20174
20175 2020-11-20 Javier Martinez Canillas <[email protected]>
20176
20177 include/grub/arm/system.h: Include missing <grub/symbol.h> header
20178 The header uses the EXPORT_FUNC() macro defined in <grub/types.h> but
20179 doesn't include it, which leads to the following compile error on arm:
20180
20181 ../include/grub/cpu/system.h:12:13: error: ‘EXPORT_FUNC’ declared as function returning a function
20182 12 | extern void EXPORT_FUNC(grub_arm_disable_caches_mmu) (void);
20183 | ^~~~~~~~~~~
20184 ../include/grub/cpu/system.h:12:1: warning: parameter names (without types) in function declaration
20185 12 | extern void EXPORT_FUNC(grub_arm_disable_caches_mmu) (void);
20186 | ^~~~~~
20187 make[3]: *** [Makefile:36581: kern/efi/kernel_exec-sb.o] Error 1
20188
20189 Reviewed-by: Daniel Kiper <[email protected]>
20190
20191 2020-11-20 Daniel Axtens <[email protected]>
20192
20193 docs: grub-install --pubkey has been supported for some time
20194 grub-install --pubkey is supported, so we can now document it.
20195
20196 Reviewed-by: Daniel Kiper <[email protected]>
20197
20198 2020-11-20 Daniel Axtens <[email protected]>
20199
20200 docs: grub-install is no longer a shell script
20201 Since commit cd46aa6cefab in 2013, grub-install hasn't been a shell
20202 script. The para doesn't really add that much, especially since it's
20203 the user manual, so just drop it.
20204
20205 Reviewed-by: Daniel Kiper <[email protected]>
20206
20207 2020-10-30 Jacob Kroon <[email protected]>
20208
20209 Makefile: Remove unused GRUB_PKGLIBDIR definition
20210 Reviewed-by: Daniel Kiper <[email protected]>
20211
20212 2020-10-30 Daniel Axtens <[email protected]>
20213
20214 lzma: Fix compilation error under clang 10
20215 Compiling under clang 10 gives:
20216
20217 grub-core/lib/LzmaEnc.c:1362:9: error: misleading indentation; statement is not part of the previous 'if' [-Werror,-Wmisleading-indentation]
20218 {
20219 ^
20220 grub-core/lib/LzmaEnc.c:1358:7: note: previous statement is here
20221 if (repIndex == 0)
20222 ^
20223 1 error generated.
20224
20225 It's not really that unclear in context: there's a commented-out
20226 if-statement. But tweak the alignment anyway so that clang is happy.
20227
20228 Reviewed-by: Daniel Kiper <[email protected]>
20229
20230 2020-10-30 Cao jin <[email protected]>
20231
20232 kern/i386/realmode: Update comment
20233 Commit b81d609e4c did not update it.
20234
20235 Reviewed-by: Daniel Kiper <[email protected]>
20236
20237 2020-10-30 Glenn Washburn <[email protected]>
20238
20239 cryptodisk: Fix cipher IV mode "plain64" always being set as "plain"
20240 When setting cipher IV mode, detection is done by prefix matching the
20241 cipher IV mode part of the cipher mode string. Since "plain" matches
20242 "plain64", we must check for "plain64" first. Otherwise, "plain64" will
20243 be detected as "plain".
20244
20245 Reviewed-by: Patrick Steinhardt <[email protected]>
20246 Reviewed-by: Daniel Kiper <[email protected]>
20247
20248 2020-09-18 Glenn Washburn <[email protected]>
20249
20250 crypto: Remove GPG_ERROR_CFLAGS from gpg_err_code_t enum
20251 This was probably added by accident when originally creating the file.
20252
20253 Reviewed-by: Daniel Kiper <[email protected]>
20254
20255 2020-09-18 Glenn Washburn <[email protected]>
20256
20257 script: Do not allow a delimiter between function name and block start
20258 Currently the following is valid syntax but should be a syntax error:
20259
20260 grub> function f; { echo HERE; }
20261 grub> f
20262 HERE
20263
20264 This fix is not backward compatible, but current syntax is not documented
20265 either and has no functional value. So any scripts with this unintended
20266 syntax are technically syntactically incorrect and should not be relying
20267 on this behavior.
20268
20269 Reviewed-by: Daniel Kiper <[email protected]>
20270
20271 2020-09-18 Glenn Washburn <[email protected]>
20272
20273 docs: Support for loading and concatenating multiple initrds
20274 This has been available since January of 2012 but has not been documented.
20275
20276 Reviewed-by: Daniel Kiper <[email protected]>
20277
20278 2020-09-18 Glenn Washburn <[email protected]>
20279
20280 lexer: char const * should be const char *
20281 Reviewed-by: Daniel Kiper <[email protected]>
20282
20283 cryptodisk: Use cipher name instead of object in error message
20284 Reviewed-by: Daniel Kiper <[email protected]>
20285
20286 2020-09-18 Glenn Washburn <[email protected]>
20287
20288 tests: F2FS test should use MOUNTDEVICE like other tests
20289 LODEVICES is not an array variable and should not be accessed as such.
20290 This allows the f2fs test to pass as it was failing because a device
20291 name had a space prepended to the path.
20292
20293 Acked-by: Jaegeuk Kim <[email protected]>
20294 Tested-by: Paul Menzel <[email protected]>
20295 Reviewed-by: Daniel Kiper <[email protected]>
20296
20297 2020-09-18 Florian La Roche <[email protected]>
20298
20299 grub-mkconfig: If $hints is not set reduce the output into grub.cfg to just 1 line
20300 Reviewed-by: Daniel Kiper <[email protected]>
20301
20302 2020-09-18 Petr Vorel <[email protected]>
20303
20304 travis: Run bootstrap to fix build
20305 autogen.sh isn't enough:
20306
20307 $ ./autogen.sh
20308 Gnulib not yet bootstrapped; run ./bootstrap instead.
20309 The command "./autogen.sh" exited with 1.
20310
20311 Additionally, using bootstrap requires to install autopoint package.
20312
20313 Reviewed-by: Daniel Kiper <[email protected]>
20314
20315 2020-09-18 Patrick Steinhardt <[email protected]>
20316
20317 luks2: Strip dashes off of the UUID
20318 The UUID header for LUKS2 uses a format with dashes, same as for
20319 LUKS(1). But while we strip these dashes for the latter, we don't for
20320 the former. This isn't wrong per se, but it's definitely inconsistent
20321 for users as they need to use the dashed format for LUKS2 and the
20322 non-dashed format for LUKS when e.g. calling "cryptomount -u $UUID".
20323
20324 Fix this inconsistency by stripping dashes off of the LUKS2 UUID.
20325
20326 Reviewed-by: Daniel Kiper <[email protected]>
20327
20328 2020-09-18 Tianjia Zhang <[email protected]>
20329
20330 efi/tpm: Remove unused functions and structures
20331 Although the tpm_execute() series of functions are defined they are not
20332 used anywhere. Several structures in the include/grub/efi/tpm.h header
20333 file are not used too. There is even nonexistent grub_tpm_init()
20334 declaration in this header. Delete all that unneeded stuff.
20335
20336 If somebody needs the functionality implemented in the dropped code then
20337 he/she can re-add it later. Now it needlessly increases the GRUB
20338 code/image size.
20339
20340 Reviewed-by: Daniel Kiper <[email protected]>
20341
20342 2020-09-18 Tianjia Zhang <[email protected]>
20343
20344 shim_lock: Enable module for all EFI architectures
20345 Like the tpm the shim_lock module is only enabled for x86_64 target.
20346 However, there's nothing specific to x86_64 in the implementation and
20347 it can be enabled for all EFI architectures.
20348
20349 Reviewed-by: Daniel Kiper <[email protected]>
20350
20351 2020-09-18 Daniel Kiper <[email protected]>
20352
20353 efi/tpm: Fix typo in grub_efi_tpm2_protocol struct
20354 Rename get_active_pcr_blanks() to get_active_pcr_banks().
20355
20356 Reviewed-by: Javier Martinez Canillas <[email protected]>
20357
20358 2020-09-18 Daniel Kiper <[email protected]>
20359
20360 i386/efi/init: Drop bogus include
20361 Reviewed-by: Javier Martinez Canillas <[email protected]>
20362
20363 2020-09-18 Daniel Kiper <[email protected]>
20364
20365 docs: Fix devicetree command description
20366 Specifically fix the subsection and drop bogus reference to the GNU/Linux.
20367
20368 Reported-by: Patrick Higgins <[email protected]>
20369 Reviewed-by: Javier Martinez Canillas <[email protected]>
20370
20371 2020-09-18 Martin Whitaker <[email protected]>
20372
20373 grub-install: Fix inverted test for NLS enabled when copying locales
20374 Commit 3d8439da8 (grub-install: Locale depends on nls) attempted to avoid
20375 copying locale files to the target directory when NLS was disabled.
20376 However the test is inverted, and it does the opposite.
20377
20378 Reviewed-by: Javier Martinez Canillas <[email protected]>
20379
20380 2020-09-11 Javier Martinez Canillas <[email protected]>
20381
20382 tftp: Roll-over block counter to prevent data packets timeouts
20383 Commit 781b3e5efc3 (tftp: Do not use priority queue) caused a regression
20384 when fetching files over TFTP whose size is bigger than 65535 * block size.
20385
20386 grub> linux /images/pxeboot/vmlinuz
20387 grub> echo $?
20388 0
20389 grub> initrd /images/pxeboot/initrd.img
20390 error: timeout reading '/images/pxeboot/initrd.img'.
20391 grub> echo $?
20392 28
20393
20394 It is caused by the block number counter being a 16-bit field, which leads
20395 to a maximum file size of ((1 << 16) - 1) * block size. Because GRUB sets
20396 the block size to 1024 octets (by using the TFTP Blocksize Option from RFC
20397 2348 [0]), the maximum file size that can be transferred is 67107840 bytes.
20398
20399 The TFTP PROTOCOL (REVISION 2) RFC 1350 [1] does not mention what a client
20400 should do when a file size is bigger than the maximum, but most TFTP hosts
20401 support the block number counter to be rolled over. That is, acking a data
20402 packet with a block number of 0 is taken as if the 65356th block was acked.
20403
20404 It was working before because the block counter roll-over was happening due
20405 an overflow. But that got fixed by the mentioned commit, which led to the
20406 regression when attempting to fetch files larger than the maximum size.
20407
20408 To allow TFTP file transfers of unlimited size again, re-introduce a block
20409 counter roll-over so the data packets are acked preventing the timeouts.
20410
20411 [0]: https://tools.ietf.org/html/rfc2348
20412 [1]: https://tools.ietf.org/html/rfc1350
20413
20414 Fixes: 781b3e5efc3 (tftp: Do not use priority queue)
20415
20416 Suggested-by: Peter Jones <[email protected]>
20417 Reviewed-by: Daniel Kiper <[email protected]>
20418
20419 2020-09-11 Florian La Roche <[email protected]>
20420
20421 templates: Remove unnecessary trailing semicolon
20422 Reviewed-by: Daniel Kiper <[email protected]>
20423
20424 2020-09-11 Glenn Washburn <[email protected]>
20425
20426 cryptodisk: Fix incorrect calculation of start sector
20427 Here dev is a grub_cryptodisk_t and dev->offset is offset in sectors of size
20428 native to the cryptodisk device. The sector is correctly transformed into
20429 native grub sector size, but then added to dev->offset which is not
20430 transformed. It would be nice if the type system would help us with this.
20431
20432 Reviewed-by: Patrick Steinhardt <[email protected]>
20433 Reviewed-by: Daniel Kiper <[email protected]>
20434
20435 2020-09-11 Glenn Washburn <[email protected]>
20436
20437 cryptodisk: Unregister cryptomount command when removing module
20438 Reviewed-by: Patrick Steinhardt <[email protected]>
20439 Reviewed-by: Daniel Kiper <[email protected]>
20440
20441 2020-09-11 Patrick Steinhardt <[email protected]>
20442
20443 luks2: Improve error reporting when decrypting/verifying key
20444 While we already set up error messages in both luks2_verify_key() and
20445 luks2_decrypt_key(), we do not ever print them. This makes it really
20446 hard to discover why a given key actually failed to decrypt a disk.
20447
20448 Improve this by including the error message in the user-visible output.
20449
20450 Reviewed-by: Daniel Kiper <[email protected]>
20451
20452 2020-09-11 Patrick Steinhardt <[email protected]>
20453
20454 luks: Fix out-of-bounds copy of UUID
20455 When configuring a LUKS disk, we copy over the UUID from the LUKS header
20456 into the new grub_cryptodisk_t structure via grub_memcpy(). As size
20457 we mistakenly use the size of the grub_cryptodisk_t UUID field, which
20458 is guaranteed to be strictly bigger than the LUKS UUID field we're
20459 copying. As a result, the copy always goes out-of-bounds and copies some
20460 garbage from other surrounding fields. During runtime, this isn't
20461 noticed due to the fact that we always NUL-terminate the UUID and thus
20462 never hit the trailing garbage.
20463
20464 Fix the issue by using the size of the local stripped UUID field.
20465
20466 Reviewed-by: Daniel Kiper <[email protected]>
20467
20468 2020-09-11 Patrick Steinhardt <[email protected]>
20469
20470 json: Remove invalid typedef redefinition
20471 The C standard does not allow for typedef redefinitions, even if they
20472 map to the same underlying type. In order to avoid including the
20473 jsmn.h in json.h and thus exposing jsmn's internals, we have exactly
20474 such a forward-declaring typedef in json.h. If enforcing the GNU99 C
20475 standard, clang may generate a warning about this non-standard
20476 construct.
20477
20478 Fix the issue by using a simple "struct jsmntok" forward declaration
20479 instead of using a typedef.
20480
20481 Tested-by: Chuck Tuffli <[email protected]>
20482 Reviewed-by: Daniel Kiper <[email protected]>
20483
20484 2020-09-11 Cao jin <[email protected]>
20485
20486 i386/relocator_common: Drop empty #ifdef
20487 Reviewed-by: Daniel Kiper <[email protected]>
20488
20489 2020-09-11 Ave Milia <[email protected]>
20490
20491 video/bochs: Fix typo
20492 Reviewed-by: Daniel Kiper <[email protected]>
20493
20494 2020-07-29 Colin Watson <[email protected]>
20495
20496 linux: Fix integer overflows in initrd size handling
20497 These could be triggered by a crafted filesystem with very large files.
20498
20499 Fixes: CVE-2020-15707
20500
20501 Reviewed-by: Jan Setje-Eilers <[email protected]>
20502 Reviewed-by: Daniel Kiper <[email protected]>
20503
20504 2020-07-29 Peter Jones <[email protected]>
20505
20506 loader/linux: Avoid overflow on initrd size calculation
20507 Reviewed-by: Daniel Kiper <[email protected]>
20508
20509 2020-07-29 Alexey Makhalov <[email protected]>
20510
20511 efi: Fix use-after-free in halt/reboot path
20512 commit 92bfc33db984 ("efi: Free malloc regions on exit")
20513 introduced memory freeing in grub_efi_fini(), which is
20514 used not only by exit path but by halt/reboot one as well.
20515 As result of memory freeing, code and data regions used by
20516 modules, such as halt, reboot, acpi (used by halt) also got
20517 freed. After return to module code, CPU executes, filled
20518 by UEFI firmware (tested with edk2), 0xAFAFAFAF pattern as
20519 a code. Which leads to #UD exception later.
20520
20521 grub> halt
20522 !!!! X64 Exception Type - 06(#UD - Invalid Opcode) CPU Apic ID - 00000000 !!!!
20523 RIP - 0000000003F4EC28, CS - 0000000000000038, RFLAGS - 0000000000200246
20524 RAX - 0000000000000000, RCX - 00000000061DA188, RDX - 0A74C0854DC35D41
20525 RBX - 0000000003E10E08, RSP - 0000000007F0F860, RBP - 0000000000000000
20526 RSI - 00000000064DB768, RDI - 000000000832C5C3
20527 R8 - 0000000000000002, R9 - 0000000000000000, R10 - 00000000061E2E52
20528 R11 - 0000000000000020, R12 - 0000000003EE5C1F, R13 - 00000000061E0FF4
20529 R14 - 0000000003E10D80, R15 - 00000000061E2F60
20530 DS - 0000000000000030, ES - 0000000000000030, FS - 0000000000000030
20531 GS - 0000000000000030, SS - 0000000000000030
20532 CR0 - 0000000080010033, CR2 - 0000000000000000, CR3 - 0000000007C01000
20533 CR4 - 0000000000000668, CR8 - 0000000000000000
20534 DR0 - 0000000000000000, DR1 - 0000000000000000, DR2 - 0000000000000000
20535 DR3 - 0000000000000000, DR6 - 00000000FFFF0FF0, DR7 - 0000000000000400
20536 GDTR - 00000000079EEA98 0000000000000047, LDTR - 0000000000000000
20537 IDTR - 0000000007598018 0000000000000FFF, TR - 0000000000000000
20538 FXSAVE_STATE - 0000000007F0F4C0
20539
20540 Proposal here is to continue to free allocated memory for
20541 exit boot services path but keep it for halt/reboot path
20542 as it won't be much security concern here.
20543 Introduced GRUB_LOADER_FLAG_EFI_KEEP_ALLOCATED_MEMORY
20544 loader flag to be used by efi halt/reboot path.
20545
20546 Reviewed-by: Darren Kenny <[email protected]>
20547 Reviewed-by: Daniel Kiper <[email protected]>
20548
20549 2020-07-29 Daniel Kiper <[email protected]>
20550
20551 efi/chainloader: Propagate errors from copy_file_path()
20552 Without any error propagated to the caller, make_file_path()
20553 would then try to advance the invalid device path node with
20554 GRUB_EFI_NEXT_DEVICE_PATH(), which would fail, returning a NULL
20555 pointer that would subsequently be dereferenced. Hence, propagate
20556 errors from copy_file_path().
20557
20558 Reviewed-by: Daniel Kiper <[email protected]>
20559
20560 2020-07-29 Peter Jones <[email protected]>
20561
20562 efi: Fix some malformed device path arithmetic errors
20563 Several places we take the length of a device path and subtract 4 from
20564 it, without ever checking that it's >= 4. There are also cases where
20565 this kind of malformation will result in unpredictable iteration,
20566 including treating the length from one dp node as the type in the next
20567 node. These are all errors, no matter where the data comes from.
20568
20569 This patch adds a checking macro, GRUB_EFI_DEVICE_PATH_VALID(), which
20570 can be used in several places, and makes GRUB_EFI_NEXT_DEVICE_PATH()
20571 return NULL and GRUB_EFI_END_ENTIRE_DEVICE_PATH() evaluate as true when
20572 the length is too small. Additionally, it makes several places in the
20573 code check for and return errors in these cases.
20574
20575 Reviewed-by: Daniel Kiper <[email protected]>
20576
20577 2020-07-29 Peter Jones <[email protected]>
20578
20579 emu: Make grub_free(NULL) safe
20580 The grub_free() implementation in grub-core/kern/mm.c safely handles
20581 NULL pointers, and code at many places depends on this. We don't know
20582 that the same is true on all host OSes, so we need to handle the same
20583 behavior in grub-emu's implementation.
20584
20585 Reviewed-by: Darren Kenny <[email protected]>
20586 Reviewed-by: Daniel Kiper <[email protected]>
20587
20588 2020-07-29 Peter Jones <[email protected]>
20589
20590 lvm: Fix two more potential data-dependent alloc overflows
20591 It appears to be possible to make a (possibly invalid) lvm PV with
20592 a metadata size field that overflows our type when adding it to the
20593 address we've allocated. Even if it doesn't, it may be possible to do so
20594 with the math using the outcome of that as an operand. Check them both.
20595
20596 Reviewed-by: Daniel Kiper <[email protected]>
20597
20598 2020-07-29 Peter Jones <[email protected]>
20599
20600 hfsplus: Fix two more overflows
20601 Both node->size and node->namelen come from the supplied filesystem,
20602 which may be user-supplied. We can't trust them for the math unless we
20603 know they don't overflow. Making sure they go through grub_add() or
20604 grub_calloc() first will give us that.
20605
20606 Reviewed-by: Darren Kenny <[email protected]>
20607 Reviewed-by: Daniel Kiper <[email protected]>
20608
20609 2020-07-29 Alexey Makhalov <[email protected]>
20610
20611 relocator: Fix grub_relocator_alloc_chunk_align() top memory allocation
20612 Current implementation of grub_relocator_alloc_chunk_align()
20613 does not allow allocation of the top byte.
20614
20615 Assuming input args are:
20616 max_addr = 0xfffff000;
20617 size = 0x1000;
20618
20619 And this is valid. But following overflow protection will
20620 unnecessarily move max_addr one byte down (to 0xffffefff):
20621 if (max_addr > ~size)
20622 max_addr = ~size;
20623
20624 ~size + 1 will fix the situation. In addition, check size
20625 for non zero to do not zero max_addr.
20626
20627 Reviewed-by: Daniel Kiper <[email protected]>
20628
20629 2020-07-29 Chris Coulson <[email protected]>
20630
20631 script: Avoid a use-after-free when redefining a function during execution
20632 Defining a new function with the same name as a previously defined
20633 function causes the grub_script and associated resources for the
20634 previous function to be freed. If the previous function is currently
20635 executing when a function with the same name is defined, this results
20636 in use-after-frees when processing subsequent commands in the original
20637 function.
20638
20639 Instead, reject a new function definition if it has the same name as
20640 a previously defined function, and that function is currently being
20641 executed. Although a behavioural change, this should be backwards
20642 compatible with existing configurations because they can't be
20643 dependent on the current behaviour without being broken.
20644
20645 Fixes: CVE-2020-15706
20646
20647 Reviewed-by: Daniel Kiper <[email protected]>
20648
20649 2020-07-29 Chris Coulson <[email protected]>
20650
20651 script: Remove unused fields from grub_script_function struct
20652 Reviewed-by: Daniel Kiper <[email protected]>
20653
20654 2020-07-29 Alexey Makhalov <[email protected]>
20655
20656 relocator: Protect grub_relocator_alloc_chunk_align() max_addr against integer underflow
20657 This commit introduces integer underflow mitigation in max_addr calculation
20658 in grub_relocator_alloc_chunk_align() invocation.
20659
20660 It consists of 2 fixes:
20661 1. Introduced grub_relocator_alloc_chunk_align_safe() wrapper function to perform
20662 sanity check for min/max and size values, and to make safe invocation of
20663 grub_relocator_alloc_chunk_align() with validated max_addr value. Replace all
20664 invocations such as grub_relocator_alloc_chunk_align(..., min_addr, max_addr - size, size, ...)
20665 by grub_relocator_alloc_chunk_align_safe(..., min_addr, max_addr, size, ...).
20666 2. Introduced UP_TO_TOP32(s) macro for the cases where max_addr is 32-bit top
20667 address (0xffffffff - size + 1) or similar.
20668
20669 Reviewed-by: Daniel Kiper <[email protected]>
20670
20671 2020-07-29 Alexey Makhalov <[email protected]>
20672
20673 relocator: Protect grub_relocator_alloc_chunk_addr() input args against integer underflow/overflow
20674 Use arithmetic macros from safemath.h to accomplish it. In this commit,
20675 I didn't want to be too paranoid to check every possible math equation
20676 for overflow/underflow. Only obvious places (with non zero chance of
20677 overflow/underflow) were refactored.
20678
20679 Reviewed-by: Daniel Kiper <[email protected]>
20680
20681 2020-07-29 Alexey Makhalov <[email protected]>
20682
20683 tftp: Do not use priority queue
20684 There is not need to reassemble the order of blocks. Per RFC 1350,
20685 server must wait for the ACK, before sending next block. Data packets
20686 can be served immediately without putting them to priority queue.
20687
20688 Logic to handle incoming packet is this:
20689 - if packet block id equal to expected block id, then
20690 process the packet,
20691 - if packet block id is less than expected - this is retransmit
20692 of old packet, then ACK it and drop the packet,
20693 - if packet block id is more than expected - that shouldn't
20694 happen, just drop the packet.
20695
20696 It makes the tftp receive path code simpler, smaller and faster.
20697 As a benefit, this change fixes CID# 73624 and CID# 96690, caused
20698 by following while loop:
20699
20700 while (cmp_block (grub_be_to_cpu16 (tftph->u.data.block), data->block + 1) == 0)
20701
20702 where tftph pointer is not moving from one iteration to another, causing
20703 to serve same packet again. Luckily, double serving didn't happen due to
20704 data->block++ during the first iteration.
20705
20706 Fixes: CID 73624, CID 96690
20707
20708 Reviewed-by: Daniel Kiper <[email protected]>
20709
20710 2020-07-29 Konrad Rzeszutek Wilk <[email protected]>
20711
20712 multiboot2: Fix memory leak if grub_create_loader_cmdline() fails
20713 Fixes: CID 292468
20714
20715 Reviewed-by: Daniel Kiper <[email protected]>
20716
20717 2020-07-29 Konrad Rzeszutek Wilk <[email protected]>
20718
20719 udf: Fix memory leak
20720 Fixes: CID 73796
20721
20722 Reviewed-by: Daniel Kiper <[email protected]>
20723 Reviewed-by: Jan Setje-Eilers <[email protected]>
20724
20725 2020-07-29 Konrad Rzeszutek Wilk <[email protected]>
20726
20727 term: Fix overflow on user inputs
20728 This requires a very weird input from the serial interface but can cause
20729 an overflow in input_buf (keys) overwriting the next variable (npending)
20730 with the user choice:
20731
20732 (pahole output)
20733
20734 struct grub_terminfo_input_state {
20735 int input_buf[6]; /* 0 24 */
20736 int npending; /* 24 4 */ <- CORRUPT
20737 ...snip...
20738
20739 The magic string requires causing this is "ESC,O,],0,1,2,q" and we overflow
20740 npending with "q" (aka increase npending to 161). The simplest fix is to
20741 just to disallow overwrites input_buf, which exactly what this patch does.
20742
20743 Fixes: CID 292449
20744
20745 Reviewed-by: Daniel Kiper <[email protected]>
20746
20747 2020-07-29 Konrad Rzeszutek Wilk <[email protected]>
20748
20749 lzma: Make sure we don't dereference past array
20750 The two dimensional array p->posSlotEncoder[4][64] is being dereferenced
20751 using the GetLenToPosState() macro which checks if len is less than 5,
20752 and if so subtracts 2 from it. If len = 0, that is 0 - 2 = 4294967294.
20753 Obviously we don't want to dereference that far out so we check if the
20754 position found is greater or equal kNumLenToPosStates (4) and bail out.
20755
20756 N.B.: Upstream LZMA 18.05 and later has this function completely rewritten
20757 without any history.
20758
20759 Fixes: CID 51526
20760
20761 Reviewed-by: Daniel Kiper <[email protected]>
20762
20763 2020-07-29 Chris Coulson <[email protected]>
20764
20765 json: Avoid a double-free when parsing fails.
20766 When grub_json_parse() succeeds, it returns the root object which
20767 contains a pointer to the provided JSON string. Callers are
20768 responsible for ensuring that this string outlives the root
20769 object and for freeing its memory when it's no longer needed.
20770
20771 If grub_json_parse() fails to parse the provided JSON string,
20772 it frees the string before returning an error. This results
20773 in a double free in luks2_recover_key(), which also frees the
20774 same string after grub_json_parse() returns an error.
20775
20776 This changes grub_json_parse() to never free the JSON string
20777 passed to it, and updates the documentation for it to make it
20778 clear that callers are responsible for ensuring that the string
20779 outlives the root JSON object.
20780
20781 Fixes: CID 292465
20782
20783 Reviewed-by: Daniel Kiper <[email protected]>
20784
20785 2020-07-29 Alexey Makhalov <[email protected]>
20786
20787 xnu: Fix double free in grub_xnu_devprop_add_property()
20788 grub_xnu_devprop_add_property() should not free utf8 and utf16 as it get
20789 allocated and freed in the caller.
20790
20791 Minor improvement: do prop fields initialization after memory allocations.
20792
20793 Fixes: CID 292442, CID 292457, CID 292460, CID 292466
20794
20795 Reviewed-by: Daniel Kiper <[email protected]>
20796
20797 2020-07-29 Alexey Makhalov <[email protected]>
20798
20799 gfxmenu: Fix double free in load_image()
20800 self->bitmap should be zeroed after free. Otherwise, there is a chance
20801 to double free (USE_AFTER_FREE) it later in rescale_image().
20802
20803 Fixes: CID 292472
20804
20805 Reviewed-by: Daniel Kiper <[email protected]>
20806
20807 2020-07-29 Daniel Kiper <[email protected]>
20808
20809 font: Do not load more than one NAME section
20810 The GRUB font file can have one NAME section only. Though if somebody
20811 crafts a broken font file with many NAME sections and loads it then the
20812 GRUB leaks memory. So, prevent against that by loading first NAME
20813 section and failing in controlled way on following one.
20814
20815 Reported-by: Chris Coulson <[email protected]>
20816 Reviewed-by: Jan Setje-Eilers <[email protected]>
20817
20818 2020-07-29 Peter Jones <[email protected]>
20819
20820 iso9660: Don't leak memory on realloc() failures
20821 Reviewed-by: Daniel Kiper <[email protected]>
20822
20823 2020-07-29 Peter Jones <[email protected]>
20824
20825 malloc: Use overflow checking primitives where we do complex allocations
20826 This attempts to fix the places where we do the following where
20827 arithmetic_expr may include unvalidated data:
20828
20829 X = grub_malloc(arithmetic_expr);
20830
20831 It accomplishes this by doing the arithmetic ahead of time using grub_add(),
20832 grub_sub(), grub_mul() and testing for overflow before proceeding.
20833
20834 Among other issues, this fixes:
20835 - allocation of integer overflow in grub_video_bitmap_create()
20836 reported by Chris Coulson,
20837 - allocation of integer overflow in grub_png_decode_image_header()
20838 reported by Chris Coulson,
20839 - allocation of integer overflow in grub_squash_read_symlink()
20840 reported by Chris Coulson,
20841 - allocation of integer overflow in grub_ext2_read_symlink()
20842 reported by Chris Coulson,
20843 - allocation of integer overflow in read_section_as_string()
20844 reported by Chris Coulson.
20845
20846 Fixes: CVE-2020-14309, CVE-2020-14310, CVE-2020-14311
20847
20848 Reviewed-by: Daniel Kiper <[email protected]>
20849
20850 2020-07-29 Peter Jones <[email protected]>
20851
20852 calloc: Use calloc() at most places
20853 This modifies most of the places we do some form of:
20854
20855 X = malloc(Y * Z);
20856
20857 to use calloc(Y, Z) instead.
20858
20859 Among other issues, this fixes:
20860 - allocation of integer overflow in grub_png_decode_image_header()
20861 reported by Chris Coulson,
20862 - allocation of integer overflow in luks_recover_key()
20863 reported by Chris Coulson,
20864 - allocation of integer overflow in grub_lvm_detect()
20865 reported by Chris Coulson.
20866
20867 Fixes: CVE-2020-14308
20868
20869 Reviewed-by: Daniel Kiper <[email protected]>
20870
20871 2020-07-29 Peter Jones <[email protected]>
20872
20873 calloc: Make sure we always have an overflow-checking calloc() available
20874 This tries to make sure that everywhere in this source tree, we always have
20875 an appropriate version of calloc() (i.e. grub_calloc(), xcalloc(), etc.)
20876 available, and that they all safely check for overflow and return NULL when
20877 it would occur.
20878
20879 Reviewed-by: Daniel Kiper <[email protected]>
20880
20881 2020-07-29 Peter Jones <[email protected]>
20882
20883 safemath: Add some arithmetic primitives that check for overflow
20884 This adds a new header, include/grub/safemath.h, that includes easy to
20885 use wrappers for __builtin_{add,sub,mul}_overflow() declared like:
20886
20887 bool OP(a, b, res)
20888
20889 where OP is grub_add, grub_sub or grub_mul. OP() returns true in the
20890 case where the operation would overflow and res is not modified.
20891 Otherwise, false is returned and the operation is executed.
20892
20893 These arithmetic primitives require newer compiler versions. So, bump
20894 these requirements in the INSTALL file too.
20895
20896 Reviewed-by: Daniel Kiper <[email protected]>
20897
20898 2020-07-29 Peter Jones <[email protected]>
20899
20900 yylex: Make lexer fatal errors actually be fatal
20901 When presented with a command that can't be tokenized to anything
20902 smaller than YYLMAX characters, the parser calls YY_FATAL_ERROR(errmsg),
20903 expecting that will stop further processing, as such:
20904
20905 #define YY_DO_BEFORE_ACTION \
20906 yyg->yytext_ptr = yy_bp; \
20907 yyleng = (int) (yy_cp - yy_bp); \
20908 yyg->yy_hold_char = *yy_cp; \
20909 *yy_cp = '\0'; \
20910 if ( yyleng >= YYLMAX ) \
20911 YY_FATAL_ERROR( "token too large, exceeds YYLMAX" ); \
20912 yy_flex_strncpy( yytext, yyg->yytext_ptr, yyleng + 1 , yyscanner); \
20913 yyg->yy_c_buf_p = yy_cp;
20914
20915 The code flex generates expects that YY_FATAL_ERROR() will either return
20916 for it or do some form of longjmp(), or handle the error in some way at
20917 least, and so the strncpy() call isn't in an "else" clause, and thus if
20918 YY_FATAL_ERROR() is *not* actually fatal, it does the call with the
20919 questionable limit, and predictable results ensue.
20920
20921 Unfortunately, our implementation of YY_FATAL_ERROR() is:
20922
20923 #define YY_FATAL_ERROR(msg) \
20924 do { \
20925 grub_printf (_("fatal error: %s\n"), _(msg)); \
20926 } while (0)
20927
20928 The same pattern exists in yyless(), and similar problems exist in users
20929 of YY_INPUT(), several places in the main parsing loop,
20930 yy_get_next_buffer(), yy_load_buffer_state(), yyensure_buffer_stack,
20931 yy_scan_buffer(), etc.
20932
20933 All of these callers expect YY_FATAL_ERROR() to actually be fatal, and
20934 the things they do if it returns after calling it are wildly unsafe.
20935
20936 Fixes: CVE-2020-10713
20937
20938 Reviewed-by: Daniel Kiper <[email protected]>
20939
20940 2020-05-25 Marc Zyngier <[email protected]>
20941
20942 arm: Fix 32-bit ARM handling of the CTR register
20943 When booting on an ARMv8 core that implements either CTR.IDC or CTR.DIC
20944 (indicating that some of the cache maintenance operations can be
20945 removed when dealing with I/D-cache coherency, GRUB dies with a
20946 "Unsupported cache type 0x........" message.
20947
20948 This is pretty likely to happen when running in a virtual machine
20949 hosted on an arm64 machine (I've triggered it on a system built around
20950 a bunch of Cortex-A55 cores, which implements CTR.IDC).
20951
20952 It turns out that the way GRUB deals with the CTR register is a bit
20953 harsh for anything from ARMv7 onwards. The layout of the register is
20954 backward compatible, meaning that nothing that gets added is allowed to
20955 break earlier behaviour. In this case, ignoring IDC is completely fine,
20956 and only results in unnecessary cache maintenance.
20957
20958 We can thus avoid being paranoid, and align the 32bit behaviour with
20959 its 64bit equivalent.
20960
20961 This patch has the added benefit that it gets rid of a (gnu-specific)
20962 case range too.
20963
20964 Reviewed-by: Leif Lindholm <[email protected]>
20965 Reviewed-by: Daniel Kiper <[email protected]>
20966
20967 2020-05-25 Ian Jackson <[email protected]>
20968
20969 templates/20_linux_xen: Support Xen Security Modules (XSM/FLASK)
20970 XSM is enabled by adding "flask=enforcing" as a Xen command line
20971 argument, and providing the policy file as a grub module.
20972
20973 We make entries for both with and without XSM. If XSM is not compiled
20974 into Xen, then there are no policy files, so no change to the boot
20975 options.
20976
20977 Reviewed-by: Daniel Kiper <[email protected]>
20978
20979 2020-05-25 Ian Jackson <[email protected]>
20980
20981 templates/20_linux_xen: Ignore xenpolicy and config files too
20982 file_is_not_sym() currently only checks for xen-syms. Extend it to
20983 disregard xenpolicy (XSM policy files) and files ending .config (which
20984 are built by the Xen upstream build system in some configurations and
20985 can therefore end up in /boot).
20986
20987 Rename the function accordingly, to file_is_not_xen_garbage().
20988
20989 Reviewed-by: Daniel Kiper <[email protected]>
20990
20991 2020-05-25 Javier Martinez Canillas <[email protected]>
20992
20993 net: Break out nested function
20994 Nested functions are not supported in C, but are permitted as an extension
20995 in the GNU C dialect. Commit cb2f15c5448 ("normal/main: Search for specific
20996 config files for netboot") added a nested function which caused the build
20997 to break when compiling with clang.
20998
20999 Break that out into a static helper function to make the code portable again.
21000
21001 Reported-by: Daniel Axtens <[email protected]>
21002 Tested-by: Daniel Axtens <[email protected]>
21003 Reviewed-by: Daniel Kiper <[email protected]>
21004
21005 2020-05-25 Javier Martinez Canillas <[email protected]>
21006
21007 tpm: Enable module for all EFI platforms
21008 The module is only enabled for x86_64, but there's nothing specific to
21009 x86_64 in the implementation and can be enabled for all EFI platforms.
21010
21011 Reviewed-by: Daniel Kiper <[email protected]>
21012
21013 2020-05-25 Daniel Kiper <[email protected]>
21014
21015 INSTALL/configure: Update install doc and configure comment
21016 ..to reflect the GRUB build reality in them.
21017
21018 Additionally, fix text formatting a bit.
21019
21020 Reviewed-by: Leif Lindholm <[email protected]>
21021
21022 2020-05-25 Daniel Kiper <[email protected]>
21023
21024 configure: Set gnu99 C language standard by default
21025 Commit d5a32255d (misc: Make grub_strtol() "end" pointers have safer
21026 const qualifiers) introduced "restrict" keyword into some functions
21027 definitions. This keyword was introduced in C99 standard. However, some
21028 compilers by default may use C89 or something different. This behavior
21029 leads to the breakage during builds when c89 or gnu89 is in force. So,
21030 let's set gnu99 C language standard for all compilers by default. This
21031 way a bit random build issue will be fixed and the GRUB source will be
21032 build consistently regardless of type and version of the compiler.
21033
21034 It was decided to use gnu99 C language standard because it fixes the
21035 issue mentioned above and also provides some useful extensions which are
21036 used here and there in the GRUB source. Potentially we can use gnu11
21037 too. However, this may reduce pool of older compilers which can be used
21038 to build the GRUB. So, let's live with gnu99 until we discover that we
21039 strongly require a feature from newer C standard.
21040
21041 The user is still able to override C language standard using relevant
21042 *_CFLAGS variables.
21043
21044 Reviewed-by: Leif Lindholm <[email protected]>
21045
21046 2020-05-15 Tianjia Zhang <[email protected]>
21047
21048 tpm: Rename function grub_tpm_log_event() to grub_tpm_measure()
21049 grub_tpm_log_event() and grub_tpm_measure() are two functions that
21050 have the same effect. So, keep grub_tpm_log_event() and rename it
21051 to grub_tpm_measure(). This way we get also a more clear semantics.
21052
21053 Reviewed-by: Daniel Kiper <[email protected]>
21054
21055 2020-05-15 Daniel Kiper <[email protected]>
21056
21057 autogen: Replace -iname with -ipath in find command
21058 ..because -iname cannot be used to match paths.
21059
21060 Reviewed-by: Javier Martinez Canillas <[email protected]>
21061 Reviewed-by: Leif Lindholm <[email protected]>
21062 Reviewed-by: Daniel Axtens <[email protected]>
21063
21064 2020-05-15 Daniel Kiper <[email protected]>
21065
21066 INSTALL: Update configure example
21067 ..to make it more relevant.
21068
21069 Reviewed-by: Leif Lindholm <[email protected]>
21070
21071 2020-05-15 Daniel Kiper <[email protected]>
21072
21073 configure: Drop unneeded TARGET_CFLAGS expansion
21074 Reviewed-by: Javier Martinez Canillas <[email protected]>
21075 Reviewed-by: Leif Lindholm <[email protected]>
21076
21077 2020-05-15 Jacob Kroon <[email protected]>
21078
21079 docs/grub: Support for probing partition UUID on MSDOS disks
21080 Support was implemented in commit c7cb11b21 (probe: Support probing for
21081 msdos PARTUUID).
21082
21083 Reviewed-by: Daniel Kiper <[email protected]>
21084
21085 2020-05-15 Tianjia Zhang <[email protected]>
21086
21087 verifiers: Add verify string debug message
21088 Like grub_verifiers_open(), the grub_verify_string() should also
21089 display this debug message, which is very helpful for debugging.
21090
21091 Reviewed-by: Daniel Kiper <[email protected]>
21092
21093 2020-05-15 Javier Martinez Canillas <[email protected]>
21094
21095 envblk: Fix buffer overrun when attempting to shrink a variable value
21096 If an existing variable is set with a value whose length is smaller than
21097 the current value, a memory corruption can happen due copying padding '#'
21098 characters outside of the environment block buffer.
21099
21100 This is caused by a wrong calculation of the previous free space position
21101 after moving backward the characters that followed the old variable value.
21102
21103 That position is calculated to fill the remaining of the buffer with the
21104 padding '#' characters. But since isn't calculated correctly, it can lead
21105 to copies outside of the buffer.
21106
21107 The issue can be reproduced by creating a variable with a large value and
21108 then try to set a new value that is much smaller:
21109
21110 $ grub2-editenv --version
21111 grub2-editenv (GRUB) 2.04
21112
21113 $ grub2-editenv env create
21114
21115 $ grub2-editenv env set a="$(for i in {1..500}; do var="b$var"; done; echo $var)"
21116
21117 $ wc -c env
21118 1024 grubenv
21119
21120 $ grub2-editenv env set a="$(for i in {1..50}; do var="b$var"; done; echo $var)"
21121 malloc(): corrupted top size
21122 Aborted (core dumped)
21123
21124 $ wc -c env
21125 0 grubenv
21126
21127 Reported-by: Renaud Métrich <[email protected]>
21128 Reviewed-by: Daniel Kiper <[email protected]>
21129
21130 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21131
21132 docs: Remove docs for non-existing uppermem command
21133 Remove all documentation of and mentions of the uppermem
21134 command from the docs/grub.texi file.
21135
21136 The uppermem command is not implemented in the GRUB source
21137 at all and appears to never have been implemented despite
21138 former plans to add an uppermem command.
21139
21140 To reduce user confusion, this even removes the paragraph
21141 describing how GRUB's uppermem command was supposed to
21142 complement the Linux kernel's mem= parameter.
21143
21144 Reviewed-by: Daniel Kiper <[email protected]>
21145
21146 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21147
21148 docs: Remove docs for non-existing pxe_unload command
21149 Remove the documentation of the pxe_unload command from the
21150 docs/grub.texi file.
21151
21152 The pxe_unload command is not implemented in the grub source
21153 at this time at all. It appears to have been removed in commit
21154 671a78acb (cleanup pxe and efi network release).
21155
21156 Reviewed-by: Daniel Kiper <[email protected]>
21157
21158 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21159
21160 gitignore: Add a few forgotten file patterns
21161 Add a few patterns to .gitignore to cover files which are generated
21162 by building grub ("make", "make check", "make dist") but which have
21163 been forgotten to add to .gitignore in the past.
21164
21165 Reviewed-by: Daniel Kiper <[email protected]>
21166
21167 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21168
21169 gitignore: Add leading slashes where appropriate
21170 Going through the list of gitignore patterns without a leading slash,
21171 this adds a leading slash where it appears to have been forgotten.
21172
21173 Some gitignore patterns like ".deps/" or "Makefile" clearly should
21174 match everywhere, so those definitively need no leading slash.
21175
21176 For some patterns like "ascii.bitmaps", it is unclear where in the
21177 source tree they should match. Those patterns are kept as they are,
21178 matching the patterns in the whole tree of subdirectories.
21179
21180 Reviewed-by: Daniel Kiper <[email protected]>
21181
21182 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21183
21184 gitignore: Add trailing slashes for directories
21185 Add trailing slashes for all patterns matching directories.
21186
21187 Note that we do *not* add trailing slashes for *symlinks*
21188 to directories.
21189
21190 Reviewed-by: Daniel Kiper <[email protected]>
21191
21192 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21193
21194 gitignore: Sort both pattern groups alphabetically
21195 Alphabetically sort the two groups of gitignore patterns:
21196
21197 * The group of patterns without slashes, matching anywhere
21198 in the directory subtree.
21199
21200 * The group of patterns with slashes, matching relative to the
21201 .gitignore file's directory
21202
21203 Reviewed-by: Daniel Kiper <[email protected]>
21204
21205 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21206
21207 gitignore: Group patterns with and without slash
21208 Group the .gitignore patterns into two groups:
21209
21210 * Pattern not including a slash, i.e. matching files anywhere in
21211 the .gitignore file's directory and all of its subdirectories.
21212
21213 * Patterns including a slash, i.e. matching only relative to the
21214 .gitignore file's directory.
21215
21216 Reviewed-by: Daniel Kiper <[email protected]>
21217
21218 2020-05-15 Hans Ulrich Niedermann <[email protected]>
21219
21220 gitignore: Consistent leading slash is easier to read
21221 As all gitignore patterns containing a left or middle slash match
21222 only relative to the .gitignore file's directory, we write them
21223 all in the same manner with a leading slash.
21224
21225 This makes the file significantly easier to read.
21226
21227 Reviewed-by: Daniel Kiper <[email protected]>
21228
21229 2020-05-15 Daniel Kiper <[email protected]>
21230
21231 mips/cache: Add missing nop's in delay slots
21232 Lack of them causes random instructions to be executed before the
21233 jump really happens.
21234
21235 Reviewed-by: Daniel Kiper <[email protected]>
21236
21237 2020-04-21 Patrick Steinhardt <[email protected]>
21238
21239 luks2: Propagate error when reading area key fails
21240 When decrypting a given keyslot, all error cases except for one set up
21241 an error and return the error code. The only exception is when we try to
21242 read the area key: instead of setting up an error message, we directly
21243 print it via grub_dprintf().
21244
21245 Convert the outlier to use grub_error() to allow more uniform handling
21246 of errors.
21247
21248 Reviewed-by: Daniel Kiper <[email protected]>
21249
21250 2020-04-21 Patrick Steinhardt <[email protected]>
21251
21252 json: Get rid of casts for "jsmntok_t"
21253 With the upstream change having landed that adds a name to the
21254 previously anonymous "jsmntok" typedef, we can now add a forward
21255 declaration for that struct in our code. As a result, we no longer have
21256 to store the "tokens" member of "struct grub_json" as a void pointer but
21257 can instead use the forward declaration, allowing us to get rid of casts
21258 of that field.
21259
21260 Reviewed-by: Daniel Kiper <[email protected]>
21261
21262 2020-04-21 Patrick Steinhardt <[email protected]>
21263
21264 json: Update jsmn library to upstream commit 053d3cd
21265 Update our embedded version of the jsmn library to upstream commit
21266 053d3cd (Merge pull request #175 from pks-t/pks/struct-type,
21267 2020-04-02).
21268
21269 Reviewed-by: Daniel Kiper <[email protected]>
21270
21271 2020-04-21 Steve Langasek <[email protected]>
21272
21273 templates: Output a menu entry for firmware setup on UEFI FastBoot systems
21274 The fwsetup command allows to reboot into the EFI firmware setup menu, add
21275 a template to include a menu entry on EFI systems that makes use of that
21276 command to reboot into the EFI firmware settings.
21277
21278 This is useful for users since the hotkey to enter into the EFI setup menu
21279 may not be the same on all systems so users can use the menu entry without
21280 needing to figure out what key needs to be pressed.
21281
21282 Also, if fastboot is enabled in the BIOS then often it is not possible to
21283 enter the firmware setup menu. So the entry is again useful for this case.
21284
21285 Reviewed-by: Daniel Kiper <[email protected]>
21286
21287 2020-04-21 Hans de Goede <[email protected]>
21288
21289 kern/term: Accept ESC, F4 and holding SHIFT as user interrupt keys
21290 On some devices the ESC key is the hotkey to enter the BIOS/EFI setup
21291 screen, making it really hard to time pressing it right. Besides that
21292 ESC is also pretty hard to discover for a user who does not know it
21293 will unhide the menu.
21294
21295 This commit makes F4, which was chosen because is not used as a hotkey
21296 to enter the BIOS setup by any vendor, also interrupt sleeps / stop the
21297 menu countdown.
21298
21299 This solves the ESC gets into the BIOS setup and also somewhat solves
21300 the discoverability issue, but leaves the timing issue unresolved.
21301
21302 This commit fixes the timing issue by also adding support for keeping
21303 SHIFT pressed during boot to stop the menu countdown. This matches
21304 what Ubuntu is doing, which should also help with discoverability.
21305
21306 Reviewed-by: Daniel Kiper <[email protected]>
21307
21308 2020-04-21 Hans de Goede <[email protected]>
21309
21310 efi/console: Do not set text-mode until we actually need it
21311 If we're running with a hidden menu we may never need text mode, so do not
21312 change the video-mode to text until we actually need it.
21313
21314 This allows to boot a machine without unnecessary graphical transitions and
21315 provide a seamless boot experience to users.
21316
21317 Reviewed-by: Daniel Kiper <[email protected]>
21318
21319 2020-04-21 Hans de Goede <[email protected]>
21320
21321 efi/console: Implement getkeystatus() support
21322 Implement getkeystatus() support in the EFI console driver.
21323
21324 This is needed because the logic to determine if a key was pressed to make
21325 the menu countdown stop will be changed by a later patch to also take into
21326 account the SHIFT key being held down.
21327
21328 For this reason the EFI console driver has to support getkeystatus() to
21329 allow detecting that event.
21330
21331 Note that if a non-modifier key gets pressed and repeated calls to
21332 getkeystatus() are made then it will return the modifier status at the
21333 time of the non-modifier key, until that key-press gets consumed by a
21334 getkey() call.
21335
21336 This is a side-effect of how the EFI simple-text-input protocol works
21337 and cannot be avoided.
21338
21339 Reviewed-by: Daniel Kiper <[email protected]>
21340
21341 2020-04-21 Hans de Goede <[email protected]>
21342
21343 efi/console: Add grub_console_read_key_stroke() helper function
21344 This is a preparatory patch for adding getkeystatus() support to the
21345 EFI console driver.
21346
21347 We can get modifier status through the simple_text_input read_key_stroke()
21348 method, but if a non-modifier key is (also) pressed the read_key_stroke()
21349 call will consume that key from the firmware's queue.
21350
21351 The new grub_console_read_key_stroke() helper buffers upto 1 key-stroke.
21352 If it has a non-modifier key buffered, it will return that one, if its
21353 buffer is empty, it will fills its buffer by getting a new key-stroke.
21354
21355 If called with consume=1 it will empty its buffer after copying the
21356 key-data to the callers buffer, this is how getkey() will use it.
21357
21358 If called with consume=0 it will keep the last key-stroke buffered, this
21359 is how getkeystatus() will call it. This means that if a non-modifier
21360 key gets pressed, repeated getkeystatus() calls will return the modifiers
21361 of that key-press until it is consumed by a getkey() call.
21362
21363 Reviewed-by: Daniel Kiper <[email protected]>
21364
21365 2020-04-21 Hans de Goede <[email protected]>
21366
21367 kern/term: Make grub_getkeystatus() helper function available everywhere
21368 Move grub_getkeystatushelper() function from grub-core/commands/keystatus.c
21369 to grub-core/kern/term.c and export it so that it can be used outside of
21370 the keystatus command code too.
21371
21372 There's no logic change in this patch. The function definition is moved so
21373 it can be called from grub-core/kern/term.c in a subsequent patch. It will
21374 be used to determine if a SHIFT key has was held down and use that also to
21375 interrupt the countdown, without the need to press a key at the right time.
21376
21377 Reviewed-by: Daniel Kiper <[email protected]>
21378
21379 2020-04-21 Javier Martinez Canillas <[email protected]>
21380
21381 efi/console: Move grub_console_set{colorstate,cursor} higher in the file
21382 This is just a preparatory patch to move the functions higher in the file,
21383 since these will be called by the grub_prepare_for_text_output() function
21384 that will be introduced in a later patch.
21385
21386 The logic is unchanged by this patch. Functions definitions are just moved
21387 to avoid a forward declaration in a later patch, keeping the code clean.
21388
21389 Reviewed-by: Daniel Kiper <[email protected]>
21390
21391 2020-04-21 Paul Menzel <[email protected]>
21392
21393 docs/grub: Fix typo in *preferred*
21394 Reviewed-by: Daniel Kiper <[email protected]>
21395
21396 2020-04-21 Daniel Axtens <[email protected]>
21397
21398 powerpc/mkimage: Fix CHRP note descsz
21399 Currently, an image generated with 'grub-mkimage -n' causes an error when
21400 read with 'readelf -a':
21401
21402 Displaying notes found at file offset 0x000106f0 with length 0x0000002c:
21403 Owner Data size Description
21404 readelf: Warning: note with invalid namesz and/or descsz found at offset 0x0
21405 readelf: Warning: type: 0x1275, namesize: 0x00000008, descsize: 0x0000002c, alignment: 4
21406
21407 This is because the descsz of the CHRP note is set to
21408 sizeof (struct grub_ieee1275_note)
21409 which is the size of the entire note, including name and elf header. The
21410 desczs should contain only the contents, not the name and header sizes.
21411
21412 Set the descsz instead to 'sizeof (struct grub_ieee1275_note_desc)'
21413
21414 Resultant readelf output:
21415
21416 Displaying notes found at file offset 0x00010710 with length 0x0000002c:
21417 Owner Data size Description
21418 PowerPC 0x00000018 Unknown note type: (0x00001275)
21419 description data: ff ff ff ff 00 c0 00 00 ff ff ff ff ff ff ff ff ff ff ff ff 00 00 40 00
21420
21421 So far as I can tell this issue has existed for as long as the note
21422 generation code has existed, but I guess nothing really checks descsz.
21423
21424 Reviewed-by: Daniel Kiper <[email protected]>
21425
21426 2020-03-31 Flavio Suligoi <[email protected]>
21427
21428 efi: Add missed space in GRUB_EFI_GLOBAL_VARIABLE_GUID
21429 Reviewed-by: Daniel Kiper <[email protected]>
21430
21431 2020-03-31 Michael Chang <[email protected]>
21432
21433 zfs: Fix gcc10 error -Werror=zero-length-bounds
21434 We bumped into the build error while testing gcc-10 pre-release.
21435
21436 In file included from ../../include/grub/file.h:22,
21437 from ../../grub-core/fs/zfs/zfs.c:34:
21438 ../../grub-core/fs/zfs/zfs.c: In function 'zap_leaf_lookup':
21439 ../../grub-core/fs/zfs/zfs.c:2263:44: error: array subscript '<unknown>' is outside the bounds of an interior zero-length array 'grub_uint16_t[0]' {aka 'short unsigned int[0]'} [-Werror=zero-length-bounds]
21440 2263 | for (chunk = grub_zfs_to_cpu16 (l->l_hash[LEAF_HASH (blksft, h, l)], endian);
21441 ../../include/grub/types.h:241:48: note: in definition of macro 'grub_le_to_cpu16'
21442 241 | # define grub_le_to_cpu16(x) ((grub_uint16_t) (x))
21443 | ^
21444 ../../grub-core/fs/zfs/zfs.c:2263:16: note: in expansion of macro 'grub_zfs_to_cpu16'
21445 2263 | for (chunk = grub_zfs_to_cpu16 (l->l_hash[LEAF_HASH (blksft, h, l)], endian);
21446 | ^~~~~~~~~~~~~~~~~
21447 In file included from ../../grub-core/fs/zfs/zfs.c:48:
21448 ../../include/grub/zfs/zap_leaf.h:72:16: note: while referencing 'l_hash'
21449 72 | grub_uint16_t l_hash[0];
21450 | ^~~~~~
21451
21452 Here I'd like to quote from the gcc document [1] which seems best to
21453 explain what is going on here.
21454
21455 "Although the size of a zero-length array is zero, an array member of
21456 this kind may increase the size of the enclosing type as a result of
21457 tail padding. The offset of a zero-length array member from the
21458 beginning of the enclosing structure is the same as the offset of an
21459 array with one or more elements of the same type. The alignment of a
21460 zero-length array is the same as the alignment of its elements.
21461
21462 Declaring zero-length arrays in other contexts, including as interior
21463 members of structure objects or as non-member objects, is discouraged.
21464 Accessing elements of zero-length arrays declared in such contexts is
21465 undefined and may be diagnosed."
21466
21467 The l_hash[0] is apparnetly an interior member to the enclosed structure
21468 while l_entries[0] is the trailing member. And the offending code tries
21469 to access members in l_hash[0] array that triggers the diagnose.
21470
21471 Given that the l_entries[0] is used to get proper alignment to access
21472 leaf chunks, we can accomplish the same thing through the ALIGN_UP macro
21473 thus eliminating l_entries[0] from the structure. In this way we can
21474 pacify the warning as l_hash[0] now becomes the last member to the
21475 enclosed structure.
21476
21477 [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html
21478
21479 Reviewed-by: Daniel Kiper <[email protected]>
21480
21481 2020-03-31 Michael Chang <[email protected]>
21482
21483 mdraid1x_linux: Fix gcc10 error -Werror=array-bounds
21484 We bumped into the build error while testing gcc-10 pre-release.
21485
21486 ../../grub-core/disk/mdraid1x_linux.c: In function 'grub_mdraid_detect':
21487 ../../grub-core/disk/mdraid1x_linux.c:181:15: error: array subscript <unknown> is outside array bounds of 'grub_uint16_t[0]' {aka 'short unsigned int[0]'} [-Werror=array-bounds]
21488 181 | (char *) &sb.dev_roles[grub_le_to_cpu32 (sb.dev_number)]
21489 | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
21490 ../../grub-core/disk/mdraid1x_linux.c:98:17: note: while referencing 'dev_roles'
21491 98 | grub_uint16_t dev_roles[0]; /* Role in array, or 0xffff for a spare, or 0xfffe for faulty. */
21492 | ^~~~~~~~~
21493 ../../grub-core/disk/mdraid1x_linux.c:127:33: note: defined here 'sb'
21494 127 | struct grub_raid_super_1x sb;
21495 | ^~
21496 cc1: all warnings being treated as errors
21497
21498 Apparently gcc issues the warning when trying to access sb.dev_roles
21499 array's member, since it is a zero length array as the last element of
21500 struct grub_raid_super_1x that is allocated sparsely without extra
21501 chunks for the trailing bits, so the warning looks legitimate in this
21502 regard.
21503
21504 As the whole thing here is doing offset computation, it is undue to use
21505 syntax that would imply array member access then take address from it
21506 later. Instead we could accomplish the same thing through basic array
21507 pointer arithmetic to pacify the warning.
21508
21509 Reviewed-by: Daniel Kiper <[email protected]>
21510
21511 2020-03-31 Simon Hardy <[email protected]>
21512
21513 build: Fix GRUB i386-pc build with Ubuntu gcc
21514 With recent versions of gcc on Ubuntu a very large lzma_decompress.img file is
21515 output. (e.g. 134479600 bytes instead of 2864.) This causes grub-mkimage to
21516 fail with: "error: Decompressor is too big."
21517
21518 This seems to be caused by a section .note.gnu.property that is placed at an
21519 offset such that objcopy needs to pad the img file with zeros.
21520
21521 This issue is present on:
21522 Ubuntu 19.10 with gcc (Ubuntu 8.3.0-26ubuntu1~19.10) 8.3.0
21523 Ubuntu 19.10 with gcc (Ubuntu 9.2.1-9ubuntu2) 9.2.1 20191008
21524
21525 This issue is not present on:
21526 Ubuntu 19.10 with gcc (Ubuntu 7.5.0-3ubuntu1~19.10) 7.5.0
21527 RHEL 8.0 with gcc 8.3.1 20190507 (Red Hat 8.3.1-4)
21528
21529 The issue can be fixed by removing the section using objcopy as shown in
21530 this patch.
21531
21532 Reviewed-by: Daniel Kiper <[email protected]>
21533
21534 2020-03-31 Tianjia Zhang <[email protected]>
21535
21536 efi/tpm: Fix memory leak in grub_tpm1/2_log_event()
21537 The memory requested for the event is not released here,
21538 causing memory leaks. This patch fixes this problem.
21539
21540 Reviewed-by: Javier Martinez Canillas <[email protected]>
21541 Reviewed-by: Daniel Kiper <[email protected]>
21542
21543 2020-03-31 Michael Chang <[email protected]>
21544
21545 docs: Document notes on LVM cache booting
21546 Add notes on LVM cache booting to the GRUB manual to help user understanding
21547 the outstanding issue and status.
21548
21549 Reviewed-by: Daniel Kiper <[email protected]>
21550
21551 2020-03-31 Michael Chang <[email protected]>
21552
21553 lvm: Add LVM cache logical volume handling
21554 The LVM cache logical volume is the logical volume consisting of the original
21555 and the cache pool logical volume. The original is usually on a larger and
21556 slower storage device while the cache pool is on a smaller and faster one. The
21557 performance of the original volume can be improved by storing the frequently
21558 used data on the cache pool to utilize the greater performance of faster
21559 device.
21560
21561 The default cache mode "writethrough" ensures that any data written will be
21562 stored both in the cache and on the origin LV, therefore grub can be straight
21563 to read the original lv as no data loss is guarenteed.
21564
21565 The second cache mode is "writeback", which delays writing from the cache pool
21566 back to the origin LV to have increased performance. The drawback is potential
21567 data loss if losing the associated cache device.
21568
21569 During the boot time grub reads the LVM offline i.e. LVM volumes are not
21570 activated and mounted, hence it should be fine to read directly from original
21571 lv since all cached data should have been flushed back in the process of taking
21572 it offline.
21573
21574 It is also not much helpful to the situation by adding fsync calls to the
21575 install code. The fsync did not force to write back dirty cache to the original
21576 device and rather it would update associated cache metadata to complete the
21577 write transaction with the cache device. IOW the writes to cached blocks still
21578 go only to the cache device.
21579
21580 To write back dirty cache, as LVM cache did not support dirty cache flush per
21581 block range, there'no way to do it for file. On the other hand the "cleaner"
21582 policy is implemented and can be used to write back "all" dirty blocks in a
21583 cache, which effectively drain all dirty cache gradually to attain and last in
21584 the "clean" state, which can be useful for shrinking or decommissioning a
21585 cache. The result and effect is not what we are looking for here.
21586
21587 In conclusion, as it seems no way to enforce file writes to the original
21588 device, grub may suffer from power failure as it cannot assemble the cache
21589 device and read the dirty data from it. However since the case is only
21590 applicable to writeback mode which is sensitive to data lost in nature, I'd
21591 still like to propose my (relatively simple) patch and treat reading dirty
21592 cache as improvement.
21593
21594 Reviewed-by: Daniel Kiper <[email protected]>
21595
21596 2020-03-10 Patrick Steinhardt <[email protected]>
21597
21598 gnulib: Fix build of base64 when compiling with memory debugging
21599 When building GRUB with memory management debugging enabled, then the
21600 build fails because of `grub_debug_malloc()` and `grub_debug_free()`
21601 being undefined in the luks2 module. The cause is that we patch
21602 "base64.h" to unconditionaly include "config-util.h", which shouldn't be
21603 included for modules at all. As a result, `MM_DEBUG` is defined when
21604 building the module, causing it to use the debug memory allocation
21605 functions. As these are not built into modules, we end up with a linker
21606 error.
21607
21608 Fix the issue by removing the <config-util.h> include altogether. The
21609 sole reason it was included was for the `_GL_ATTRIBUTE_CONST` macro,
21610 which we can simply define as empty in case it's not set.
21611
21612 Reviewed-by: Daniel Kiper <[email protected]>
21613
21614 2020-03-10 Patrick Steinhardt <[email protected]>
21615
21616 build: Fix option to explicitly disable memory debugging
21617 The memory management system supports a debug mode that can be enabled
21618 at build time by passing "--enable-mm-debug" to the configure script.
21619 Passing the option will cause us define MM_DEBUG as expected, but in
21620 fact the reverse option "--disable-mm-debug" will do the exact same
21621 thing and also set up the define. This currently causes the build of
21622 "lib/gnulib/base64.c" to fail as it tries to use `grub_debug_malloc()`
21623 and `grub_debug_free()` even though both symbols aren't defined.
21624
21625 Seemingly, `AC_ARG_ENABLE()` will always execute the third argument if
21626 either the positive or negative option was passed. Let's thus fix the
21627 issue by moving the call to`AC_DEFINE()` into an explicit `if test
21628 $xenable_mm_debug` block, similar to how other defines work.
21629
21630 Reviewed-by: Daniel Kiper <[email protected]>
21631 Reviewed-by: Paul Menzel <[email protected]>
21632
21633 2020-03-10 David Michael <[email protected]>
21634
21635 fat: Support file modification times
21636 This allows comparing file ages on EFI system partitions.
21637
21638 Reviewed-by: Daniel Kiper <[email protected]>
21639
21640 2020-03-10 David Michael <[email protected]>
21641
21642 exfat: Save the matching directory entry struct when searching
21643 This provides the node's attributes outside the iterator function
21644 so the file modification time can be accessed and reported.
21645
21646 Reviewed-by: Daniel Kiper <[email protected]>
21647
21648 2020-03-10 Mike Gilbert <[email protected]>
21649
21650 datetime: Enable the datetime module for the emu platform
21651 Fixes a build failure:
21652
21653 grub-core/commands/date.c:49: undefined reference to `grub_get_weekday_name'
21654 grub-core/commands/ls.c:155: undefined reference to `grub_unixtime2datetime'
21655
21656 Bug: https://bugs.gentoo.org/711512
21657
21658 Reviewed-by: Javier Martinez Canillas <[email protected]>
21659 Tested-by: Javier Martinez Canillas <[email protected]>
21660 Reviewed-by: Daniel Kiper <[email protected]>
21661
21662 2020-03-10 John Paul Adrian Glaubitz <[email protected]>
21663
21664 build: Add soft-float handling for SuperH (sh4)
21665 While GRUB has no platform support for SuperH (sh4) yet, this change
21666 adds the target-specific handling of soft-floats such that the GRUB
21667 utilities can be built on this target.
21668
21669 Reviewed-by: Daniel Kiper <[email protected]>
21670
21671 2020-03-10 Peter Jones <[email protected]>
21672
21673 efi: Fix the type of grub_efi_status_t
21674 Currently, in some builds with some checkers, we see:
21675
21676 1. grub-core/disk/efi/efidisk.c:601: error[shiftTooManyBitsSigned]: Shifting signed 64-bit value by 63 bits is undefined behaviour
21677
21678 This is because grub_efi_status_t is defined as grub_efi_intn_t, which is
21679 signed, and shifting into the sign bit is not defined behavior. UEFI fixed
21680 this in the spec in 2.3:
21681
21682 2.3 | Change the defined type of EFI_STATUS from INTN to UINTN | May 7, 2009
21683
21684 And the current EDK2 code has:
21685 MdePkg/Include/Base.h-//
21686 MdePkg/Include/Base.h-// Status codes common to all execution phases
21687 MdePkg/Include/Base.h-//
21688 MdePkg/Include/Base.h:typedef UINTN RETURN_STATUS;
21689 MdePkg/Include/Base.h-
21690 MdePkg/Include/Base.h-/**
21691 MdePkg/Include/Base.h- Produces a RETURN_STATUS code with the highest bit set.
21692 MdePkg/Include/Base.h-
21693 MdePkg/Include/Base.h- @param StatusCode The status code value to convert into a warning code.
21694 MdePkg/Include/Base.h- StatusCode must be in the range 0x00000000..0x7FFFFFFF.
21695 MdePkg/Include/Base.h-
21696 MdePkg/Include/Base.h- @return The value specified by StatusCode with the highest bit set.
21697 MdePkg/Include/Base.h-
21698 MdePkg/Include/Base.h-**/
21699 MdePkg/Include/Base.h-#define ENCODE_ERROR(StatusCode) ((RETURN_STATUS)(MAX_BIT | (StatusCode)))
21700 MdePkg/Include/Base.h-
21701 MdePkg/Include/Base.h-/**
21702 MdePkg/Include/Base.h- Produces a RETURN_STATUS code with the highest bit clear.
21703 MdePkg/Include/Base.h-
21704 MdePkg/Include/Base.h- @param StatusCode The status code value to convert into a warning code.
21705 MdePkg/Include/Base.h- StatusCode must be in the range 0x00000000..0x7FFFFFFF.
21706 MdePkg/Include/Base.h-
21707 MdePkg/Include/Base.h- @return The value specified by StatusCode with the highest bit clear.
21708 MdePkg/Include/Base.h-
21709 MdePkg/Include/Base.h-**/
21710 MdePkg/Include/Base.h-#define ENCODE_WARNING(StatusCode) ((RETURN_STATUS)(StatusCode))
21711 MdePkg/Include/Base.h-
21712 MdePkg/Include/Base.h-/**
21713 MdePkg/Include/Base.h- Returns TRUE if a specified RETURN_STATUS code is an error code.
21714 MdePkg/Include/Base.h-
21715 MdePkg/Include/Base.h- This function returns TRUE if StatusCode has the high bit set. Otherwise, FALSE is returned.
21716 MdePkg/Include/Base.h-
21717 MdePkg/Include/Base.h- @param StatusCode The status code value to evaluate.
21718 MdePkg/Include/Base.h-
21719 MdePkg/Include/Base.h- @retval TRUE The high bit of StatusCode is set.
21720 MdePkg/Include/Base.h- @retval FALSE The high bit of StatusCode is clear.
21721 MdePkg/Include/Base.h-
21722 MdePkg/Include/Base.h-**/
21723 MdePkg/Include/Base.h-#define RETURN_ERROR(StatusCode) (((INTN)(RETURN_STATUS)(StatusCode)) < 0)
21724 ...
21725 Uefi/UefiBaseType.h:typedef RETURN_STATUS EFI_STATUS;
21726
21727 This patch makes grub's implementation match the Edk2 declaration with regards
21728 to the signedness of the type.
21729
21730 Reviewed-by: Daniel Kiper <[email protected]>
21731
21732 2020-03-10 Peter Jones <[email protected]>
21733
21734 efi/gop: Add debug output on GOP probing
21735 Add debug information to EFI GOP video driver probing function.
21736
21737 Reviewed-by: Daniel Kiper <[email protected]>
21738
21739 2020-03-10 Peter Jones <[email protected]>
21740
21741 efi/uga: Use video instead of fb as debug condition
21742 All other video drivers use "video" as the debug condition instead of "fb"
21743 so change this in the efi/uga driver to make it consistent with the others.
21744
21745 Reviewed-by: Daniel Kiper <[email protected]>
21746
21747 2020-03-10 Peter Jones <[email protected]>
21748
21749 efi: Print error messages to grub_efi_allocate_pages_real()
21750 No messages were printed in this function, add some to ease debugging.
21751
21752 Also, the function returns a void * pointer so return NULL instead of
21753 0 to make the code more readable.
21754
21755 Reviewed-by: Daniel Kiper <[email protected]>
21756
21757 2020-03-10 Andrei Borzenkov <[email protected]>
21758
21759 efi/uga: Use 64 bit for fb_base
21760 We get 64 bit from PCI BAR but then truncate by assigning to 32 bit.
21761 Make sure to check that pointer does not overflow on 32 bit platform.
21762
21763 Closes: 50931
21764
21765 Reviewed-by: Daniel Kiper <[email protected]>
21766
21767 2020-03-10 Alexander Graf <[email protected]>
21768
21769 efi/gop: Add support for BLT_ONLY adapters
21770 EFI GOP has support for multiple different bitness types of frame buffers
21771 and for a special "BLT only" type which is always defined to be RGBx.
21772
21773 Because grub2 doesn't ever directly access the frame buffer but instead
21774 only renders graphics via the BLT interface anyway, we can easily support
21775 these adapters.
21776
21777 The reason this has come up now is the emerging support for virtio-gpu
21778 in OVMF. That adapter does not have the notion of a memory mapped frame
21779 buffer and thus is BLT only.
21780
21781 Reviewed-by: Daniel Kiper <[email protected]>
21782
21783 2020-03-10 Peter Jones <[email protected]>
21784
21785 normal/completion: Fix possible NULL pointer dereference
21786 Coverity Scan reports that the grub_strrchr() function can return NULL if
21787 the character is not found. Check if that's the case for dirfile pointer.
21788
21789 Reviewed-by: Daniel Kiper <[email protected]>
21790
21791 2020-03-10 Peter Jones <[email protected]>
21792
21793 kern: Add grub_debug_enabled()
21794 Add a grub_debug_enabled() helper function instead of open coding it.
21795
21796 Reviewed-by: Daniel Kiper <[email protected]>
21797
21798 2020-03-10 Peter Jones <[email protected]>
21799
21800 Makefile: Make libgrub.pp depend on config-util.h
21801 If you build with "make -j48" a lot, sometimes you see:
21802
21803 gcc -E -DHAVE_CONFIG_H -I. -I.. -Wall -W -DGRUB_UTIL=1 -D_FILE_OFFSET_BITS=64 -I./include -DGRUB_FILE=\"grub_script.tab.h\" -I. -I.. -I. -I.. -I../include -I./include -I../grub-core/lib/libgcrypt-grub/src/ -I../grub-core/lib/minilzo -I../grub-core/lib/xzembed -DMINILZO_HAVE_CONFIG_H -Wall -W -DGRUB_UTIL=1 -D_FILE_OFFSET_BITS=64 -I./include -DGRUB_FILE=\"grub_script.tab.h\" -I. -I.. -I. -I.. -I../include -I./include -I../grub-core/lib/libgcrypt-grub/src/ -I./grub-core/gnulib -I../grub-core/gnulib -I/builddir/build/BUILD/grub-2.02/grub-aarch64-efi-2.02 -D_FILE_OFFSET_BITS=64 \
21804 -D'GRUB_MOD_INIT(x)=@MARKER@x@' grub_script.tab.h grub_script.yy.h ../grub-core/commands/blocklist.c ../grub-core/commands/macbless.c ../grub-core/commands/xnu_uuid.c ../grub-core/commands/testload.c ../grub-core/commands/ls.c ../grub-core/disk/dmraid_nvidia.c ../grub-core/disk/loopback.c ../grub-core/disk/lvm.c ../grub-core/disk/mdraid_linux.c ../grub-core/disk/mdraid_linux_be.c ../grub-core/disk/mdraid1x_linux.c ../grub-core/disk/raid5_recover.c ../grub-core/disk/raid6_recover.c ../grub-core/font/font.c ../grub-core/gfxmenu/font.c ../grub-core/normal/charset.c ../grub-core/video/fb/fbblit.c ../grub-core/video/fb/fbutil.c ../grub-core/video/fb/fbfill.c ../grub-core/video/fb/video_fb.c ../grub-core/video/video.c ../grub-core/video/capture.c ../grub-core/video/colors.c ../grub-core/unidata.c ../grub-core/io/bufio.c ../grub-core/fs/affs.c ../grub-core/fs/afs.c ../grub-core/fs/bfs.c ../grub-core/fs/btrfs.c ../grub-core/fs/cbfs.c ../grub-core/fs/cpio.c ../grub-core/fs/cpio_be.c ../grub-core/fs/odc.c ../grub-core/fs/newc.c ../grub-core/fs/ext2.c ../grub-core/fs/fat.c ../grub-core/fs/exfat.c ../grub-core/fs/fshelp.c ../grub-core/fs/hfs.c ../grub-core/fs/hfsplus.c ../grub-core/fs/hfspluscomp.c ../grub-core/fs/iso9660.c ../grub-core/fs/jfs.c ../grub-core/fs/minix.c ../grub-core/fs/minix2.c ../grub-core/fs/minix3.c ../grub-core/fs/minix_be.c ../grub-core/fs/minix2_be.c ../grub-core/fs/minix3_be.c ../grub-core/fs/nilfs2.c ../grub-core/fs/ntfs.c ../grub-core/fs/ntfscomp.c ../grub-core/fs/reiserfs.c ../grub-core/fs/romfs.c ../grub-core/fs/sfs.c ../grub-core/fs/squash4.c ../grub-core/fs/tar.c ../grub-core/fs/udf.c ../grub-core/fs/ufs2.c ../grub-core/fs/ufs.c ../grub-core/fs/ufs_be.c ../grub-core/fs/xfs.c ../grub-core/fs/zfs/zfscrypt.c ../grub-core/fs/zfs/zfs.c ../grub-core/fs/zfs/zfsinfo.c ../grub-core/fs/zfs/zfs_lzjb.c ../grub-core/fs/zfs/zfs_lz4.c ../grub-core/fs/zfs/zfs_sha256.c ../grub-core/fs/zfs/zfs_fletcher.c ../grub-core/lib/envblk.c ../grub-core/lib/hexdump.c ../grub-core/lib/LzFind.c ../grub-core/lib/LzmaEnc.c ../grub-core/lib/crc.c ../grub-core/lib/adler32.c ../grub-core/lib/crc64.c ../grub-core/normal/datetime.c ../grub-core/normal/misc.c ../grub-core/partmap/acorn.c ../grub-core/partmap/amiga.c ../grub-core/partmap/apple.c ../grub-core/partmap/sun.c ../grub-core/partmap/plan.c ../grub-core/partmap/dvh.c ../grub-core/partmap/sunpc.c ../grub-core/partmap/bsdlabel.c ../grub-core/partmap/dfly.c ../grub-core/script/function.c ../grub-core/script/lexer.c ../grub-core/script/main.c ../grub-core/script/script.c ../grub-core/script/argv.c ../grub-core/io/gzio.c ../grub-core/io/xzio.c ../grub-core/io/lzopio.c ../grub-core/kern/ia64/dl_helper.c ../grub-core/kern/arm/dl_helper.c ../grub-core/kern/arm64/dl_helper.c ../grub-core/lib/minilzo/minilzo.c ../grub-core/lib/xzembed/xz_dec_bcj.c ../grub-core/lib/xzembed/xz_dec_lzma2.c ../grub-core/lib/xzembed/xz_dec_stream.c ../util/misc.c ../grub-core/kern/command.c ../grub-core/kern/device.c ../grub-core/kern/disk.c ../grub-core/lib/disk.c ../util/getroot.c ../grub-core/osdep/unix/getroot.c ../grub-core/osdep/getroot.c ../grub-core/osdep/devmapper/getroot.c ../grub-core/osdep/relpath.c ../grub-core/kern/emu/hostdisk.c ../grub-core/osdep/devmapper/hostdisk.c ../grub-core/osdep/hostdisk.c ../grub-core/osdep/unix/hostdisk.c ../grub-core/osdep/exec.c ../grub-core/osdep/sleep.c ../grub-core/osdep/password.c ../grub-core/kern/emu/misc.c ../grub-core/kern/emu/mm.c ../grub-core/kern/env.c ../grub-core/kern/err.c ../grub-core/kern/file.c ../grub-core/kern/fs.c ../grub-core/kern/list.c ../grub-core/kern/misc.c ../grub-core/kern/partition.c ../grub-core/lib/crypto.c ../grub-core/disk/luks.c ../grub-core/disk/geli.c ../grub-core/disk/cryptodisk.c ../grub-core/disk/AFSplitter.c ../grub-core/lib/pbkdf2.c ../grub-core/commands/extcmd.c ../grub-core/lib/arg.c ../grub-core/disk/ldm.c ../grub-core/disk/diskfilter.c ../grub-core/partmap/gpt.c ../grub-core/partmap/msdos.c ../grub-core/fs/proc.c ../grub-core/fs/archelp.c > libgrub.pp || (rm -f libgrub.pp; exit 1)
21805 rm -f stamp-h1
21806 touch ../config-util.h.in
21807 cd . && /bin/sh ./config.status config-util.h
21808 config.status: creating config-util.h
21809 In file included from ../include/grub/mm.h:25:0,
21810 from ../include/grub/disk.h:29,
21811 from ../include/grub/file.h:26,
21812 from ../grub-core/fs/btrfs.c:21:
21813 ./config.h:38:10: fatal error: ./config-util.h: No such file or directory
21814 #include <config-util.h>
21815 ^~~~~~~~~~~~~~~
21816 compilation terminated.
21817 make: *** [Makefile:13098: libgrub.pp] Error 1
21818
21819 This is because libgrub.pp is built with -DGRUB_UTIL=1, which means
21820 it'll try to include config-util.h, but a parallel make is actually
21821 building that file. I think.
21822
21823 Reviewed-by: Daniel Kiper <[email protected]>
21824
21825 2020-03-10 Peter Jones <[email protected]>
21826
21827 efi: Print more debug info in our module loader
21828 The function that searches the mods section base address does not have
21829 any debug information. Add some debugging outputs that could be useful.
21830
21831 Reviewed-by: Daniel Kiper <[email protected]>
21832
21833 2020-03-10 Peter Jones <[email protected]>
21834
21835 linux/getroot: Handle rssd storage device names
21836 The Micron PCIe SSDs Linux driver (mtip32xx) exposes block devices
21837 as /dev/rssd[a-z]+[0-9]*. Add support for these rssd device names.
21838
21839 Reviewed-by: Daniel Kiper <[email protected]>
21840
21841 2020-03-10 Julian Andres Klode <[email protected]>
21842
21843 smbios: Add a --linux argument to apply linux modalias-like filtering
21844 Linux creates modalias strings by filtering out non-ASCII, space,
21845 and colon characters. Provide an option that does the same filtering
21846 so people can create a modalias string in GRUB, and then match their
21847 modalias patterns against it.
21848
21849 Reviewed-by: Daniel Kiper <[email protected]>
21850
21851 2020-03-10 Mike Gilbert <[email protected]>
21852
21853 po: Fix replacement of %m in sed programs
21854 When running make dist, I hit this error:
21855
21856 rm -f [email protected] && /usr/bin/gmsgfmt -c --statistics --verbose -o [email protected] [email protected]
21857 [email protected]:5312: 'msgstr' is not a valid C format string, unlike 'msgid'.
21858 Reason: The character that terminates the directive number 3 is not a valid conversion specifier.
21859 /usr/bin/gmsgfmt: found 1 fatal error
21860
21861 This was caused by "%m" being replaced with foreign Unicode characters.
21862 For example:
21863
21864 msgid "cannot rename the file %s to %s: %m"
21865 msgstr "ﺹﺎﻨﻧﻮﺗ ﺮﻌﻧﺎﻤﻋ ﺖﻬﻋ ﻒִﻴﻠﻋ %s ﺕﻭ %s: %ﻡ"
21866
21867 Mimic the workaround used for "%s" by reversing the replacement of "%m" at
21868 the end of the sed programs.
21869
21870 Reviewed-by: Daniel Kiper <[email protected]>
21871
21872 2020-03-10 Colin Watson <[email protected]>
21873
21874 gettext: Restore patches to po/Makefile.in.in
21875 These were inadvertently lost during the conversion to Gnulib (gnulib:
21876 Upgrade Gnulib and switch to bootstrap tool; commit 35b909062). The
21877 files in po/gettext-patches/ can be imported using "git am" on top of
21878 the gettext tag corresponding to AM_GNU_GETTEXT_VERSION in configure.ac
21879 (currently 0.18.3). They handle translation of messages in shell files,
21880 make msgfmt output in little-endian format, and arrange to use @SHELL@
21881 rather than /bin/sh.
21882
21883 There were some changes solely for the purpose of distributing extra
21884 files; for ease of maintenance, I've added these to
21885 conf/Makefile.extra-dist instead.
21886
21887 Fixes: https://savannah.gnu.org/bugs/?57298
21888
21889 Reviewed-by: Daniel Kiper <[email protected]>
21890
21891 2020-02-28 Peter Jones <[email protected]>
21892
21893 misc: Make grub_strtol() "end" pointers have safer const qualifiers
21894 Currently the string functions grub_strtol(), grub_strtoul(), and
21895 grub_strtoull() don't declare the "end" pointer in such a way as to
21896 require the pointer itself or the character array to be immutable to the
21897 implementation, nor does the C standard do so in its similar functions,
21898 though it does require us not to change any of it.
21899
21900 The typical declarations of these functions follow this pattern:
21901
21902 long
21903 strtol(const char * restrict nptr, char ** restrict endptr, int base);
21904
21905 Much of the reason for this is historic, and a discussion of that
21906 follows below, after the explanation of this change. (GRUB currently
21907 does not include the "restrict" qualifiers, and we name the arguments a
21908 bit differently.)
21909
21910 The implementation is semantically required to treat the character array
21911 as immutable, but such accidental modifications aren't stopped by the
21912 compiler, and the semantics for both the callers and the implementation
21913 of these functions are sometimes also helped by adding that requirement.
21914
21915 This patch changes these declarations to follow this pattern instead:
21916
21917 long
21918 strtol(const char * restrict nptr,
21919 const char ** const restrict endptr,
21920 int base);
21921
21922 This means that if any modification to these functions accidentally
21923 introduces either an errant modification to the underlying character
21924 array, or an accidental assignment to endptr rather than *endptr, the
21925 compiler should generate an error. (The two uses of "restrict" in this
21926 case basically mean strtol() isn't allowed to modify the character array
21927 by going through *endptr, and endptr isn't allowed to point inside the
21928 array.)
21929
21930 It also means the typical use case changes to:
21931
21932 char *s = ...;
21933 const char *end;
21934 long l;
21935
21936 l = strtol(s, &end, 10);
21937
21938 Or even:
21939
21940 const char *p = str;
21941 while (p && *p) {
21942 long l = strtol(p, &p, 10);
21943 ...
21944 }
21945
21946 This fixes 26 places where we discard our attempts at treating the data
21947 safely by doing:
21948
21949 const char *p = str;
21950 long l;
21951
21952 l = strtol(p, (char **)&ptr, 10);
21953
21954 It also adds 5 places where we do:
21955
21956 char *p = str;
21957 while (p && *p) {
21958 long l = strtol(p, (const char ** const)&p, 10);
21959 ...
21960 /* more calls that need p not to be pointer-to-const */
21961 }
21962
21963 While moderately distasteful, this is a better problem to have.
21964
21965 With one minor exception, I have tested that all of this compiles
21966 without relevant warnings or errors, and that /much/ of it behaves
21967 correctly, with gcc 9 using 'gcc -W -Wall -Wextra'. The one exception
21968 is the changes in grub-core/osdep/aros/hostdisk.c , which I have no idea
21969 how to build.
21970
21971 Because the C standard defined type-qualifiers in a way that can be
21972 confusing, in the past there's been a slow but fairly regular stream of
21973 churn within our patches, which add and remove the const qualifier in many
21974 of the users of these functions. This change should help avoid that in
21975 the future, and in order to help ensure this, I've added an explanation
21976 in misc.h so that when someone does get a compiler warning about a type
21977 error, they have the fix at hand.
21978
21979 The reason we don't have "const" in these calls in the standard is
21980 purely anachronistic: C78 (de facto) did not have type qualifiers in the
21981 syntax, and the "const" type qualifier was added for C89 (I think; it
21982 may have been later). strtol() appears to date from 4.3BSD in 1986,
21983 which means it could not be added to those functions in the standard
21984 without breaking compatibility, which is usually avoided.
21985
21986 The syntax chosen for type qualifiers is what has led to the churn
21987 regarding usage of const, and is especially confusing on string
21988 functions due to the lack of a string type. Quoting from C99, the
21989 syntax is:
21990
21991 declarator:
21992 pointer[opt] direct-declarator
21993 direct-declarator:
21994 identifier
21995 ( declarator )
21996 direct-declarator [ type-qualifier-list[opt] assignment-expression[opt] ]
21997 ...
21998 direct-declarator [ type-qualifier-list[opt] * ]
21999 ...
22000 pointer:
22001 * type-qualifier-list[opt]
22002 * type-qualifier-list[opt] pointer
22003 type-qualifier-list:
22004 type-qualifier
22005 type-qualifier-list type-qualifier
22006 ...
22007 type-qualifier:
22008 const
22009 restrict
22010 volatile
22011
22012 So the examples go like:
22013
22014 const char foo; // immutable object
22015 const char *foo; // mutable pointer to object
22016 char * const foo; // immutable pointer to mutable object
22017 const char * const foo; // immutable pointer to immutable object
22018 const char const * const foo; // XXX extra const keyword in the middle
22019 const char * const * const foo; // immutable pointer to immutable
22020 // pointer to immutable object
22021 const char ** const foo; // immutable pointer to mutable pointer
22022 // to immutable object
22023
22024 Making const left-associative for * and right-associative for everything
22025 else may not have been the best choice ever, but here we are, and the
22026 inevitable result is people using trying to use const (as they should!),
22027 putting it at the wrong place, fighting with the compiler for a bit, and
22028 then either removing it or typecasting something in a bad way. I won't
22029 go into describing restrict, but its syntax has exactly the same issue
22030 as with const.
22031
22032 Anyway, the last example above actually represents the *behavior* that's
22033 required of strtol()-like functions, so that's our choice for the "end"
22034 pointer.
22035
22036 Reviewed-by: Daniel Kiper <[email protected]>
22037
22038 2020-02-28 Mike Gilbert <[email protected]>
22039
22040 build: Disable PIE in TARGET_CCASFLAGS if needed
22041 PIE should be disabled in assembly sources as well, or else GRUB will
22042 fail to boot.
22043
22044 Bug: https://bugs.gentoo.org/667852
22045
22046 Reviewed-by: Daniel Kiper <[email protected]>
22047 Tested-by: John Paul Adrian Glaubitz <[email protected]>
22048
22049 2020-02-28 Mike Gilbert <[email protected]>
22050
22051 build: Move TARGET_* assignments earlier
22052 On a 32-bit SPARC userland, configure fails to compile assembly and the
22053 build fails:
22054
22055 checking for options to compile assembly... configure: error: could not compile assembly
22056
22057 config.log shows:
22058
22059 asm-tests/sparc64.S: Assembler messages:
22060 asm-tests/sparc64.S:5: Error: Architecture mismatch on "lduw [%o4+4],%o4".
22061 asm-tests/sparc64.S:5: (Requires v9|v9a|v9b|v9c|v9d|v9e|v9v|v9m|m8; requested architecture is sparclite.)
22062 asm-tests/sparc64.S:7: Error: Architecture mismatch on "stw %o5,[%o3]".
22063 asm-tests/sparc64.S:7: (Requires v9|v9a|v9b|v9c|v9d|v9e|v9v|v9m|m8; requested architecture is sparclite.)
22064 asm-tests/sparc64.S:8: Error: Architecture mismatch on "bne,pt %icc,1b ,pt %icc,1b".
22065 asm-tests/sparc64.S:8: (Requires v9|v9a|v9b|v9c|v9d|v9e|v9v|v9m|m8; requested architecture is sparclite.)
22066
22067 Simply moving these blocks earlier in configure.ac is sufficient to
22068 ensure that the tests are executed with the appropriate flags
22069 (specifically -m64 in this case).
22070
22071 Bug: https://bugs.gentoo.org/667850
22072
22073 Reviewed-by: Daniel Kiper <[email protected]>
22074 Tested-by: John Paul Adrian Glaubitz <[email protected]>
22075
22076 2020-02-28 Patrick Steinhardt <[email protected]>
22077
22078 luks2: Add missing newline to debug message
22079 The debug message printed when decryption with a keyslot fails is
22080 missing its trailing newline. Add it to avoid mangling it with
22081 subsequent output.
22082
22083 Reviewed-by: Daniel Kiper <[email protected]>
22084
22085 2020-02-18 Michael Chang <[email protected]>
22086
22087 verifiers: Fix calling uninitialized function pointer
22088 The necessary check for NULL before use of function ver->close is not
22089 taking place in the failure path. This patch simply adds the missing
22090 check and fixes the problem that GRUB hangs indefinitely after booting
22091 rogue image without valid signature if secure boot is turned on.
22092
22093 Now it displays like this for booting rogue UEFI image:
22094
22095 error: bad shim signature
22096 error: you need to load the kernel first
22097
22098 Press any key to continue...
22099
22100 and then you can go back to boot menu by pressing any key or after a few
22101 seconds expired.
22102
22103 Reviewed-by: Javier Martinez Canillas <[email protected]>
22104 Reviewed-by: Daniel Kiper <[email protected]>
22105
22106 2020-02-18 Peter Jones <[email protected]>
22107
22108 grub-editenv: Make grub-editenv chase symlinks including those across devices
22109 The grub-editenv create command will wrongly overwrite /boot/grub2/grubenv
22110 with a regular file if grubenv is a symbolic link. But instead, it should
22111 create a new file in the path the symlink points to.
22112
22113 This lets /boot/grub2/grubenv be a symlink to /boot/efi/EFI/fedora/grubenv
22114 even when they're different mount points, which allows grub2-editenv to be
22115 the same across platforms (i.e. UEFI vs BIOS).
22116
22117 For example, in Fedora the GRUB EFI builds have prefix set to /EFI/fedora
22118 (on the EFI System Partition), but for BIOS machine it'll be /boot/grub2
22119 (which may or may not be its own mountpoint).
22120
22121 With this patch, on EFI machines we can make /boot/grub2/grubenv a symlink
22122 to /boot/efi/EFI/fedora/grubenv, and the same copy of grub-set-default will
22123 work on both kinds of systems.
22124
22125 Windows doesn't implement a readlink primitive, so the current behaviour is
22126 maintained for this operating system.
22127
22128 Reviewed-by: Adam Jackson <[email protected]>
22129 Reviewed-by: Daniel Kiper <[email protected]>
22130
22131 2020-02-18 Peter Jones <[email protected]>
22132
22133 grub-editenv: Add grub_util_readlink()
22134 Currently grub-editenv and related tools are not able to follow symbolic
22135 links when finding their config file. For example the grub-editenv create
22136 command will wrongly overwrite a symlink in /boot/grub2/grubenv with a new
22137 regular file, instead of creating a file in the path the symlink points to.
22138
22139 A following patch will change that and add support in grub-editenv to
22140 follow symbolic links when finding the grub environment variables file.
22141
22142 Add a grub_util_readlink() helper function that is just a wrapper around
22143 the platform specific function to read the value of a symbolic link. This
22144 helper function will be used by the following patch for grub-editenv.
22145
22146 The helper function is not added for Windows, since this operating system
22147 doesn't have a primitive to read the contents of a symbolic link.
22148
22149 Reviewed-by: Adam Jackson <[email protected]>
22150 Reviewed-by: Daniel Kiper <[email protected]>
22151
22152 2020-02-18 Robert Marshall <[email protected]>
22153
22154 docs: Update info with grub.cfg netboot selection order
22155 Add documentation to the GRUB manual that specifies the order netboot
22156 clients use to select a GRUB configuration file.
22157
22158 Also explain that the feature is enabled by default but can be disabled
22159 by setting the "feature_net_search_cfg" environment variable to "n" in
22160 an embedded configuration file.
22161
22162 Reviewed-by: Daniel Kiper <[email protected]>
22163
22164 2020-02-18 Paulo Flabiano Smorigo <[email protected]>
22165
22166 normal/main: Search for specific config files for netboot
22167 This patch implements a search for a specific configuration when the config
22168 file is on a remoteserver. It uses the following order:
22169 1) DHCP client UUID option.
22170 2) MAC address (in lower case hexadecimal with dash separators);
22171 3) IP (in upper case hexadecimal) or IPv6;
22172 4) The original grub.cfg file.
22173
22174 This procedure is similar to what is used by pxelinux and yaboot:
22175 http://www.syslinux.org/wiki/index.php/PXELINUX#config
22176
22177 It is enabled by default but can be disabled by setting the environment
22178 variable "feature_net_search_cfg" to "n" in an embedded configuration.
22179
22180 Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=873406
22181
22182 Reviewed-by: Daniel Kiper <[email protected]>
22183
22184 2020-02-18 Paulo Flabiano Smorigo <[email protected]>
22185
22186 net/dhcp: Set net_<interface>_client{id, uuid} variables from DHCP options
22187 This patch sets a net_<interface>_clientid and net_<interface>_clientuuid
22188 GRUB environment variables, using the DHCP client ID and UUID options if
22189 these are found.
22190
22191 In the same way than net_<interface>_<option> variables are set for other
22192 options such domain name, boot file, next server, etc.
22193
22194 Reviewed-by: Daniel Kiper <[email protected]>
22195
22196 2020-02-18 Javier Martinez Canillas <[email protected]>
22197
22198 net/dhcp: Consistently use decimal numbers for DHCP/BOOTP options enum
22199 The DHCP Options and BOOTP Vendor Extensions enum values are a mixture of
22200 decimal and hexadecimal numbers. Change this to consistently use decimal
22201 numbers for all since that is how these values are defined by RFC 2132.
22202
22203 Suggested-by: Daniel Kiper <[email protected]>
22204 Reviewed-by: Daniel Kiper <[email protected]>
22205
22206 2020-02-18 Paulo Flabiano Smorigo <[email protected]>
22207
22208 kern: Add %X option to printf functions
22209 The printf(3) function has support for the %X format specifier, to output
22210 an unsigned hexadecimal integer in uppercase.
22211
22212 This can be achived in GRUB using the %x format specifier in grub_printf()
22213 and calling grub_toupper(), but it is more convenient if there is support
22214 for %X in grub_printf().
22215
22216 Reviewed-by: Daniel Kiper <[email protected]>
22217
22218 2020-02-18 Javier Martinez Canillas <[email protected]>
22219
22220 normal: Move common datetime functions out of the normal module
22221 The common datetime helper functions are currently included in the normal
22222 module, but this makes any other module that calls these functions to have
22223 a dependency with the normal module only for this reason.
22224
22225 Since the normal module does a lot of stuff, it calls functions from other
22226 modules. But since other modules may depend on it for calling the datetime
22227 helpers, this could lead to circular dependencies between modules.
22228
22229 As an example, when platform == xen the grub_get_datetime() function from
22230 the datetime module calls to the grub_unixtime2datetime() helper function
22231 from the normal module. Which leads to the following module dependency:
22232
22233 datetime -> normal
22234
22235 and send_dhcp_packet() from the net module calls the grub_get_datetime()
22236 function, which leads to the following module dependency:
22237
22238 net -> datetime -> normal
22239
22240 but that means that the normal module is not allowed to depend on net or
22241 any other module that depends on it due the transitive dependency caused
22242 by datetime. A recent patch attempted to add support to fetch the config
22243 file over the network, which leads to the following circular dependency:
22244
22245 normal -> net -> datetime -> normal
22246
22247 So having the datetime helpers in the normal module makes it quite fragile
22248 and easy to add circular dependencies like these, that break the build due
22249 the genmoddep.awk script catching the issues.
22250
22251 Fix this by taking the datetime helper functions out of the normal module
22252 and instead add them to the datetime module itself. Besides fixing these
22253 issues, it makes more sense to have these helper functions there anyways.
22254
22255 Reported-by: Daniel Kiper <[email protected]>
22256 Reviewed-by: Daniel Kiper <[email protected]>
22257
22258 2020-02-11 Peter Jones <[email protected]>
22259
22260 minilzo: Update to minilzo-2.08
22261 This patch updates the miniLZO library to a newer version, which among other
22262 things fixes "CVE-2014-4607 - lzo: lzo1x_decompress_safe() integer overflow"
22263 that is present in the current used in GRUB.
22264
22265 It also updates the "GRUB Developers Manual", to mention that the library is
22266 used and describes the process to update it to a newer release when needed.
22267
22268 Resolves: http://savannah.gnu.org/bugs/?42635
22269
22270 Reviewed-by: Daniel Kiper <[email protected]>
22271
22272 2020-01-28 Peter Jones <[email protected]>
22273
22274 squash4: Fix an uninitialized variable
22275 gcc says:
22276
22277 grub-core/fs/squash4.c: In function ‘direct_read’:
22278 grub-core/fs/squash4.c:868:10: error: ‘err’ may be used uninitialized in
22279 this function [-Werror=maybe-uninitialized]
22280 868 | if (err)
22281 | ^
22282 cc1: all warnings being treated as errors
22283
22284 This patch initializes it to GRUB_ERR_NONE.
22285
22286 Reviewed-by: Daniel Kiper <[email protected]>
22287
22288 2020-01-28 C. Masloch <[email protected]>
22289
22290 freedos: Fix FreeDOS command booting large files (near or above 64 KiB)
22291 While testing the 86-DOS lDebug [1] booting from GRUB2, newer versions of the
22292 debugger would fail to load when booted using GRUB's freedos command. The
22293 behaviour observed in a qemu i386 machine was that the ROM-BIOS's boot load
22294 would start anew, instead of loading the selected debugger as kernel.
22295
22296 It came to light that there was a size limit: Kernel files that were 58880
22297 bytes (E600h) long or shorter succeeded to boot, while files that were 64000
22298 bytes or longer failed in the manner described.
22299
22300 Eventually it turned out that the relocator16 stub succeeded whenever it was
22301 placed completely within the first 64 KiB of the Low Memory Area. The chunk
22302 for the relocator is allocated with a minimum address of 0x8010 and a maximum
22303 address just below 0xA0000 [2]. That means if the kernel is, for instance,
22304 E600h bytes long, then the kernel will be allocated memory starting at 00600h
22305 (the fixed FreeDOS kernel load address) up to E600h + 00600h = 0EC00h, which
22306 leaves 1400h (5120) bytes for the relocator to stay in the first 64 KiB.
22307 If the kernel is 64000 bytes (FA00h) long, then the relocator must go to
22308 FA00h + 00600h = 10000h at least which is outside the first 64 KiB.
22309
22310 The problem is that the relocator16 initialises the DS register with a
22311 "pseudo real mode" descriptor, which is defined with a segment limit of
22312 64 KiB and a segment base of zero. After that, the relocator addressed
22313 parts of itself (implicitly) using the DS register, with an offset from
22314 ESI, which holds the linear address of the relocator's base [3]. With the
22315 larger kernel files this would lead to accessing data beyond the 64 KiB
22316 segment limit, presumably leading to a fault and perhaps a subsequent
22317 triple-fault or such.
22318
22319 This patch fixes the relocator to set the segment base of the descriptors
22320 to the base address of the relocator; then, the subsequent accesses to
22321 the relocator's variables are done without the ESI register as an index.
22322 This does not interfere with the relocator's or its target's normal
22323 operation; the segment limits are still loaded with 64 KiB and all the
22324 segment bases are subsequently reset by the relocator anyway.
22325
22326 Current versions of the debugger to test are uploaded to [4]. The file
22327 ldebugnh.com (LZ4-compressed and built with -D_EXTHELP=0) at 58368 bytes
22328 loads successfully, whereas ldebug.com at 64000 bytes fails. Loading one
22329 of these files requires setting root to a FAT FS partition and using the
22330 freedos command to specify the file as kernel:
22331
22332 set root='(hd0,msdos1)'
22333 freedos /ldebug.com
22334 boot
22335
22336 Booting the file using the multiboot command (which uses a WIP entrypoint
22337 of the debugger) works, as it does not use GRUB's relocator16 but instead
22338 includes a loader in the kernel itself, which drops it back to 86 Mode.
22339
22340 [1]: https://hg.ulukai.org/ecm/ldebug
22341 [2]: http://git.savannah.gnu.org/cgit/grub.git/tree/grub-core/lib/i386/relocator.c?id=495781f5ed1b48bf27f16c53940d6700c181c74c#n127
22342 [3]: http://git.savannah.gnu.org/cgit/grub.git/tree/grub-core/lib/i386/relocator16.S?id=495781f5ed1b48bf27f16c53940d6700c181c74c#n97
22343 [4]: https://ulukai.org/ecm/lDebug-5479a7988d21-nohelp.zip
22344
22345 Reviewed-by: Daniel Kiper <[email protected]>
22346
22347 2020-01-10 Patrick Steinhardt <[email protected]>
22348
22349 disk: Implement support for LUKS2
22350 With cryptsetup 2.0, a new version of LUKS was introduced that breaks
22351 compatibility with the previous version due to various reasons. GRUB
22352 currently lacks any support for LUKS2, making it impossible to decrypt
22353 disks encrypted with that version. This commit implements support for
22354 this new format.
22355
22356 Note that LUKS1 and LUKS2 are quite different data formats. While they
22357 do share the same disk signature in the first few bytes, representation
22358 of encryption parameters is completely different between both versions.
22359 While the former version one relied on a single binary header, only,
22360 LUKS2 uses the binary header only in order to locate the actual metadata
22361 which is encoded in JSON. Furthermore, the new data format is a lot more
22362 complex to allow for more flexible setups, like e.g. having multiple
22363 encrypted segments and other features that weren't previously possible.
22364 Because of this, it was decided that it doesn't make sense to keep both
22365 LUKS1 and LUKS2 support in the same module and instead to implement it
22366 in two different modules luks and luks2.
22367
22368 The proposed support for LUKS2 is able to make use of the metadata to
22369 decrypt such disks. Note though that in the current version, only the
22370 PBKDF2 key derival function is supported. This can mostly attributed to
22371 the fact that the libgcrypt library currently has no support for either
22372 Argon2i or Argon2id, which are the remaining KDFs supported by LUKS2. It
22373 wouldn't have been much of a problem to bundle those algorithms with
22374 GRUB itself, but it was decided against that in order to keep down the
22375 number of patches required for initial LUKS2 support. Adding it in the
22376 future would be trivial, given that the code structure is already in
22377 place.
22378
22379 Reviewed-by: Daniel Kiper <[email protected]>
22380
22381 2020-01-10 Patrick Steinhardt <[email protected]>
22382
22383 luks: Move configuration of ciphers into cryptodisk
22384 The luks module contains quite a lot of logic to parse cipher and
22385 cipher-mode strings like aes-xts-plain64 into constants to apply them
22386 to the grub_cryptodisk_t structure. This code will be required by the
22387 upcoming luks2 module, as well, which is why this commit moves it into
22388 its own function grub_cryptodisk_setcipher in the cryptodisk module.
22389 While the strings are probably rather specific to the LUKS modules, it
22390 certainly does make sense that the cryptodisk module houses code to set
22391 up its own internal ciphers instead of hosting that code in the luks
22392 module.
22393
22394 Except for necessary adjustments around error handling, this commit does
22395 an exact move of the cipher configuration logic from luks.c to
22396 cryptodisk.c. Any behavior changes are unintentional.
22397
22398 Reviewed-by: Daniel Kiper <[email protected]>
22399
22400 2020-01-10 Patrick Steinhardt <[email protected]>
22401
22402 afsplitter: Move into its own module
22403 While the AFSplitter code is currently used only by the luks module,
22404 upcoming support for luks2 will add a second module that depends on it.
22405 To avoid any linker errors when adding the code to both modules because
22406 of duplicated symbols, this commit moves it into its own standalone
22407 module afsplitter as a preparatory step.
22408
22409 Reviewed-by: Daniel Kiper <[email protected]>
22410
22411 2020-01-10 Patrick Steinhardt <[email protected]>
22412
22413 bootstrap: Add gnulib's base64 module
22414 The upcoming support for LUKS2 disc encryption requires us to include a
22415 parser for base64-encoded data, as it is used to represent salts and
22416 digests. As gnulib already has code to decode such data, we can just
22417 add it to the boostrapping configuration in order to make it available
22418 in GRUB.
22419
22420 The gnulib module makes use of booleans via the <stdbool.h> header. As
22421 GRUB does not provide any POSIX wrapper header for this, but instead
22422 implements support for bool in <sys/types.h>, we need to patch
22423 base64.h to not use <stdbool.h> anymore. We unfortunately cannot include
22424 <sys/types.h> instead, as it would then use gnulib's internal header
22425 while compiling the gnulib object but our own <sys/types.h> when
22426 including it in a GRUB module. Because of this, the patch replaces the
22427 include with a direct typedef.
22428
22429 A second fix is required to make available _GL_ATTRIBUTE_CONST, which
22430 is provided by the configure script. As base64.h does not include
22431 <config.h>, it is thus not available and results in a compile error.
22432 This is fixed by adding an include of <config-util.h>.
22433
22434 Reviewed-by: Daniel Kiper <[email protected]>
22435
22436 2020-01-10 Patrick Steinhardt <[email protected]>
22437
22438 json: Implement wrapping interface
22439 While the newly added jsmn library provides the parsing interface, it
22440 does not provide any kind of interface to act on parsed tokens. Instead,
22441 the caller is expected to handle pointer arithmetics inside of the token
22442 array in order to extract required information. While simple, this
22443 requires users to know some of the inner workings of the library and is
22444 thus quite an unintuitive interface.
22445
22446 This commit adds a new interface on top of the jsmn parser that provides
22447 convenience functions to retrieve values from the parsed json type, grub_json_t.
22448
22449 Reviewed-by: Daniel Kiper <[email protected]>
22450
22451 2020-01-10 Patrick Steinhardt <[email protected]>
22452
22453 json: Import upstream jsmn-1.1.0
22454 The upcoming support for LUKS2 encryption will require a JSON parser to
22455 decode all parameters required for decryption of a drive. As there is
22456 currently no other tool that requires JSON, and as gnulib does not
22457 provide a parser, we need to introduce a new one into the code base. The
22458 backend for the JSON implementation is going to be the jsmn library [1].
22459 It has several benefits that make it a very good fit for inclusion in
22460 GRUB:
22461
22462 - It is licensed under MIT.
22463 - It is written in C89.
22464 - It has no dependencies, not even libc.
22465 - It is small with only about 500 lines of code.
22466 - It doesn't do any dynamic memory allocation.
22467 - It is testen on x86, amd64, ARM and AVR.
22468
22469 The library itself comes as a single header, only, that contains both
22470 declarations and definitions. The exposed interface is kind of
22471 simplistic, though, and does not provide any convenience features
22472 whatsoever. Thus there will be a separate interface provided by GRUB
22473 around this parser that is going to be implemented in the following
22474 commit. This change only imports jsmn.h from tag v1.1.0 and adds it
22475 unmodified to a new json module with the following command:
22476
22477 curl -L https://raw.githubusercontent.com/zserge/jsmn/v1.1.0/jsmn.h \
22478 -o grub-core/lib/json/jsmn.h
22479
22480 Upstream jsmn commit hash: fdcef3ebf886fa210d14956d3c068a653e76a24e
22481 Upstream jsmn commit name: Modernize (#149), 2019-04-20
22482
22483 [1]: https://github.com/zserge/jsmn
22484
22485 Reviewed-by: Daniel Kiper <[email protected]>
22486
22487 2019-12-20 Lukasz Hawrylko <[email protected]>
22488
22489 multiboot2: Set min address for mbi allocation to 0x1000
22490 In some cases GRUB2 allocates multiboot2 structure at 0 address, that is
22491 a confusing behavior. Consumers of that structure can have internal NULL-checks
22492 that will throw an error when get a pointer to data allocated at address 0.
22493 To prevent that, define min address for mbi allocation on x86 and x86_64
22494 platforms.
22495
22496 Reviewed-by: Daniel Kiper <[email protected]>
22497
22498 2019-12-20 Paul Menzel <[email protected]>
22499
22500 docs: Export "superusers" variable to apply to submenus
22501 Reviewed-by: Daniel Kiper <[email protected]>
22502
22503 2019-12-20 Daniel Kiper <[email protected]>
22504
22505 loader/i386/linux: Fix an underflow in the setup_header length calculation
22506 Recent work around x86 Linux kernel loader revealed an underflow in the
22507 setup_header length calculation and another related issue. Both lead to
22508 the memory overwrite and later machine crash.
22509
22510 Currently when the GRUB copies the setup_header into the linux_params
22511 (struct boot_params, traditionally known as "zero page") it assumes the
22512 setup_header size as sizeof(linux_i386_kernel_header/lh). This is
22513 incorrect. It should use the value calculated accordingly to the Linux
22514 kernel boot protocol. Otherwise in case of pretty old kernel, to be
22515 exact Linux kernel boot protocol, the GRUB may write more into
22516 linux_params than it was expected to. Fortunately this is not very big
22517 issue. Though it has to be fixed. However, there is also an underflow
22518 which is grave. It happens when
22519
22520 sizeof(linux_i386_kernel_header/lh) > "real size of the setup_header".
22521
22522 Then len value wraps around and grub_file_read() reads whole kernel into
22523 the linux_params overwriting memory past it. This leads to the GRUB
22524 memory allocator breakage and finally to its crash during boot.
22525
22526 The patch fixes both issues. Additionally, it moves the code not related to
22527 grub_memset(linux_params)/grub_memcpy(linux_params)/grub_file_read(linux_params)
22528 section outside of it to not confuse the reader.
22529
22530 Fixes: e683cfb0cf5 (loader/i386/linux: Calculate the setup_header length)
22531
22532 Reviewed-by: Javier Martinez Canillas <[email protected]>
22533 Reviewed-by: Ross Philipson <[email protected]>
22534 Reviewed-by: Krystian Hebel <[email protected]>
22535
22536 2019-12-06 David Sterba <[email protected]>
22537
22538 btrfs: Add support for new RAID1C34 profiles
22539 New 3- and 4-copy variants of RAID1 were merged into Linux kernel 5.5.
22540 Add the two new profiles to the list of recognized ones. As this builds
22541 on the same code as RAID1, only the redundancy level needs to be
22542 adjusted, the rest is done by the existing code.
22543
22544 Reviewed-by: Daniel Kiper <[email protected]>
22545
22546 2019-12-06 Lenny Szubowicz <[email protected]>
22547
22548 tftp: Normalize slashes in TFTP paths
22549 Some TFTP servers do not handle multiple consecutive slashes correctly.
22550 This patch avoids sending TFTP requests with non-normalized paths.
22551
22552 Reviewed-by: Daniel Kiper <[email protected]>
22553
22554 2019-11-18 Michael Chang <[email protected]>
22555
22556 grub-editenv: Warn a user against editing environment block
22557 The environment block is a preallocated 1024-byte file which serves as
22558 persistent storage for environment variables. It has its own format
22559 which is sensitive to corruption if an editor does not know how to
22560 process it. Besides that the editor may inadvertently change grubenv
22561 file size and/or make it sparse which can lead to unexpected results.
22562
22563 This patch adds a message to the grubenv file to warn a user against
22564 editing it by tools other than grub-editenv.
22565
22566 Reviewed-by: Javier Martinez Canillas <[email protected]>
22567 Reviewed-by: Daniel Kiper <[email protected]>
22568
22569 2019-11-18 Michael Chang <[email protected]>
22570
22571 hostdisk: Set linux file descriptor to O_CLOEXEC as default
22572 We are often bothered by this sort of lvm warning while running grub-install
22573 every now and then:
22574
22575 File descriptor 4 (/dev/vda1) leaked on vgs invocation. Parent PID 1991: /usr/sbin/grub2-install
22576
22577 The requirement related to the warning is dictated in the lvm man page:
22578
22579 "On invocation, lvm requires that only the standard file descriptors stdin,
22580 stdout and stderr are available. If others are found, they get closed and
22581 messages are issued warning about the leak. This warning can be suppressed by
22582 setting the environment variable LVM_SUPPRESS_FD_WARNINGS."
22583
22584 While it could be disabled through settings, most Linux distributions seem to
22585 enable it by default and the justification provided by the developer looks to
22586 be valid to me: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=466138#15
22587
22588 Rather than trying to close and reopen the file descriptor to the same file
22589 multiple times, which is rather cumbersome, for the sake of no vgs invocation
22590 could happen in between. This patch enables the close-on-exec flag (O_CLOEXEC)
22591 for new file descriptor returned by the open() system call, making it closed
22592 thus not inherited by the child process forked and executed by the exec()
22593 family of functions.
22594
22595 Fixes Debian bug #466138.
22596
22597 Reviewed-by: Daniel Kiper <[email protected]>
22598
22599 2019-10-28 Eli Schwartz <[email protected]>
22600
22601 grub-mkconfig: Use portable "command -v" to detect installed programs
22602 The "which" utility is not guaranteed to be installed either, and if it
22603 is, its behavior is not portable either.
22604
22605 Conversely, the "command -v" shell builtin is required to exist in all
22606 POSIX 2008 compliant shells, and is thus guaranteed to work everywhere.
22607
22608 Examples of open-source shells likely to be installed as /bin/sh on
22609 Linux, which implement the 11-year-old standard: ash, bash, busybox,
22610 dash, ksh, mksh and zsh.
22611
22612 A side benefit of using the POSIX portable option is that it requires
22613 neither an external disk executable, nor (because unlike "which", the
22614 exit code is reliable) a subshell fork. This therefore represents a mild
22615 speedup.
22616
22617 Reviewed-by: Daniel Kiper <[email protected]>
22618
22619 2019-10-28 Peter Jones <[email protected]>
22620
22621 templates: Add GRUB_DISABLE_UUID
22622 The grub-mkconfig and 10_linux scripts by default attempt to use a UUID to
22623 set the root kernel command line parameter and the $root GRUB environment
22624 variable.
22625
22626 The former can be disabled by setting the GRUB_DISABLE_LINUX_UUID variable
22627 to "true", but there is currently no way to disable the latter.
22628
22629 The generated grub config uses the search command with the --fs-uuid option
22630 to find the device that has to be set as $root, i.e:
22631
22632 search --no-floppy --fs-uuid --set=root ...
22633
22634 This is usually more reliable but in some cases it may not be appropriate,
22635 so this patch introduces a new GRUB_DISABLE_UUID variable that can be used
22636 to disable searching for the $root device by filesystem UUID.
22637
22638 When disabled, the $root device will be set to the value specified in the
22639 device.map as found by the grub-probe --target=compatibility_hint option.
22640
22641 When setting GRUB_DISABLE_UUID=true, the GRUB_DISABLE_LINUX_UUID and
22642 GRUB_DISABLE_LINUX_PARTUUID variables will also be set to "true" unless
22643 these have been explicitly set to "false".
22644
22645 That way, the GRUB_DISABLE_UUID variable can be used to force using the
22646 device names for both GRUB and Linux.
22647
22648 Reviewed-by: Daniel Kiper <[email protected]>
22649 Reviewed-by: Nicholas Vinson <[email protected]>
22650
22651 2019-10-21 Michael Bideau <[email protected]>
22652
22653 at_keyboard: Fix unreliable key presses
22654 This patch fixes an issue that prevented the at_keyboard module to work
22655 (for me). The cause was a bad/wrong return value in the
22656 grub_at_keyboard_getkey() function in grub-core/term/at_keyboard.c file
22657 at line 237. My symptoms were to have an unresponsive keyboard. Keys
22658 needed to be pressed 10x and more to effectively be printed sometimes
22659 generating multiple key presses (after 1 or 2 sec of no printing). It
22660 was very problematic when typing passphrase in early stage (with
22661 GRUB_ENABLE_CRYPTODISK). When switched to "console" terminal input
22662 keyboard worked perfectly. It also worked great with the GRUB 2.02
22663 packaged by Debian (2.02+dfsg1-20). It was not an output issue but an
22664 input one.
22665
22666 I've managed to analyze the issue and found that it came from the commit
22667 216950a4e (at_keyboard: Split protocol from controller code.). Three
22668 lines where moved from the fetch_key() function in
22669 grub-core/term/at_keyboard.c file to the beginning of
22670 grub_at_keyboard_getkey() function (same file). However, returning -1
22671 made sense when it happened in fetch_key() function but not anymore in
22672 grub_at_keyboard_getkey() function which should return GRUB_TERM_NO_KEY.
22673 I think it was just an incomplete cut-paste missing a small manual
22674 correction. Let's fix it.
22675
22676 Note: Commit message updated by Daniel Kiper.
22677
22678 Reviewed-by: Daniel Kiper <[email protected]>
22679
22680 2019-10-21 Prarit Bhargava <[email protected]>
22681
22682 templates: Fix bad test on GRUB_DISABLE_SUBMENU
22683 The GRUB_DISABLE_SUBMENU option is different than the others in the sense
22684 that it has to be set to "y" instead of "true" to be enabled.
22685
22686 That causes a lot of confusion to users, some may wrongly set it to "true"
22687 expecting that will work the same than with most options, and some may set
22688 it to "yes" since for other options the value to set is a word and not a
22689 single character.
22690
22691 This patch changes all the grub.d scripts using the GRUB_DISABLE_SUBMENU
22692 option, so they check if it was set to "true" instead of "y", making it
22693 consistent with all the other options.
22694
22695 But to keep backward compatibility for users that set the option to "y" in
22696 /etc/default/grub file, keep testing for this value. And also do it for
22697 "yes", since it is a common mistake made by users caused by this option
22698 being inconsistent with the others.
22699
22700 Reviewed-by: Daniel Kiper <[email protected]>
22701
22702 2019-10-21 Nicholas Vinson <[email protected]>
22703
22704 probe: Support probing for msdos PARTUUID
22705 Extend partition UUID probing support in GRUB core to display pseudo
22706 partition UUIDs for MBR (MSDOS) partitions.
22707
22708 Reviewed-by: Daniel Kiper <[email protected]>
22709
22710 2019-09-23 Colin Watson <[email protected]>
22711
22712 grub-mkconfig: Fix typo in --help output
22713 The short form of "--version" that grub-mkconfig accepts is "-V", not "-v".
22714
22715 Fixes Debian bug #935504.
22716
22717 Reviewed-by: Vladimir 'phcoder' Serbinenko <[email protected]>
22718 Reviewed-by: Daniel Kiper <[email protected]>
22719
22720 2019-09-23 Andreas Schwab <[email protected]>
22721
22722 grub-install: Define default platform for RISC-V
22723 Reviewed-by: Daniel Kiper <[email protected]>
22724 Reviewed-by: Alexander Graf <[email protected]>
22725
22726 2019-09-23 Andreas Schwab <[email protected]>
22727
22728 RISC-V: Add __clzdi2 symbol
22729 This is needed for the zstd module build for riscv64-emu.
22730
22731 Reviewed-by: Daniel Kiper <[email protected]>
22732
22733 2019-09-23 Peter Jones <[email protected]>
22734
22735 gitattributes: Mark po/exclude.pot as binary so git won't try to diff nonprintables
22736 Reviewed-by: Daniel Kiper <[email protected]>
22737
22738 2019-09-23 Marcel Kolaja <[email protected]>
22739
22740 grub-mkconfig: Honor a symlink when generating configuration by grub-mkconfig
22741 Honor a symlink when generating configuration by grub-mkconfig, so that
22742 the -o option follows it rather than overwriting it with a regular file.
22743
22744 Reviewed-by: Daniel Kiper <[email protected]>
22745
22746 2019-09-23 Gustavo Luiz Duarte <[email protected]>
22747
22748 net: Fix crash on http
22749 Don't free file->data on receiving FIN flag since it is used all over
22750 without checking. http_close() will be called later to free that memory.
22751
22752 Fixes bug: https://bugzilla.redhat.com/show_bug.cgi?id=860834
22753
22754 Reviewed-by: Daniel Kiper <[email protected]>
22755
22756 2019-09-23 Andre Przywara <[email protected]>
22757
22758 docs: Document newly introduced net_dhcp command
22759 Commit 5bc41db756c5 ("net/dhcp: Add explicit net_dhcp command")
22760 introduced the new command "net_dhcp", which (for now) is an alias for
22761 the existing "net_bootp". Unfortunately the TEXI documentation was not
22762 adjusted accordingly.
22763
22764 Rename the existing paragraph about net_bootp to read net_dhcp instead,
22765 and make the net_bootp stanza point to this new command.
22766
22767 On the way add the newly parsed TFTP_SERVER_NAME and BOOTFILE_NAME
22768 packets to the list of supported DHCP options.
22769
22770 Fixes bug: https://savannah.gnu.org/bugs/?56725
22771
22772 Reported-by: Heinrich Schuchardt <[email protected]>
22773 Reviewed-by: Daniel Kiper <[email protected]>
22774
22775 2019-07-18 James Clarke <[email protected]>
22776
22777 [PATCH] sparc64: Fix BIOS Boot Partition support
22778 Currently, gpt_offset is uninitialised when using a BIOS Boot Partition
22779 but is used unconditionally inside save_blocklists. Instead, ensure it
22780 is always initialised to 0 (note that there is already separate code to
22781 do the equivalent adjustment after we call save_blocklists on this code
22782 path).
22783
22784 This patch has been tested on a T5-2 LDOM.
22785
22786 Tested-by: John Paul Adrian Glaubitz <[email protected]>
22787 Reviewed-by: Vladimir Serbinenko <[email protected]>
22788 Reviewed-by: Eric Snowberg <[email protected]>
22789
22790 ---
22791 util/setup.c | 4 +++-
22792 1 file changed, 3 insertions(+), 1 deletion(-)
22793
22794 2019-07-11 Vladimir Serbinenko <[email protected]>
22795
22796 configure: Add -fno-ident when available
22797 MinGW for i386-pc without this option generates a .rdata$zzz symbol that is
22798 page-aligned and hence lzma_decompress no longer fits in its allocated space.
22799 Additionally, MinGW with -fno-ident also saves a bit of space in modules. In
22800 case of other compilers we already strip the relevant sections, so, this
22801 option has no effect.
22802
22803 More info can be found at https://github.com/msys2/MINGW-packages/issues/21
22804
22805 Reviewed-by: Daniel Kiper <[email protected]>
22806
22807 2019-07-11 Heinrich Schuchardt <[email protected]>
22808
22809 lsefisystab: Add support for device tree table
22810 The device tree may passed by the firmware as UEFI configuration
22811 table. Let lsefisystab display a short text and not only the GUID
22812 for the device tree.
22813
22814 Here is an example output:
22815
22816 grub> lsefisystab
22817 Address: 0xbff694d8
22818 Signature: 5453595320494249 revision: 00020046
22819 Vendor: Das U-Boot, Version=20190700
22820 2 tables:
22821 0xbe741000 eb9d2d31-2d88-11d3-9a160090273fc14d SMBIOS
22822 0x87f00000 b1b621d5-f19c-41a5-830bd9152c69aae0 DEVICE TREE
22823
22824 Reviewed-by: Leif Lindholm <[email protected]>
22825 Reviewed-by: Daniel Kiper <[email protected]>
22826
22827 2019-07-11 David Michael <[email protected]>
22828
22829 smbios: Add a module for retrieving SMBIOS information
22830 The following are two use cases from Rajat Jain <[email protected]>:
22831
22832 1) We have a board that boots Linux and this board itself can be plugged
22833 into one of different chassis types. We need to pass different
22834 parameters to the kernel based on the "CHASSIS_TYPE" information
22835 that is passed by the bios in the DMI/SMBIOS tables.
22836
22837 2) We may have a USB stick that can go into multiple boards, and the
22838 exact kernel to be loaded depends on the machine information
22839 (PRODUCT_NAME etc) passed via the DMI.
22840
22841 Reviewed-by: Daniel Kiper <[email protected]>
22842
22843 2019-07-11 David Michael <[email protected]>
22844
22845 lsefisystab: Define SMBIOS3 entry point structures for EFI
22846 This adds the GUID and includes it in lsefisystab output.
22847
22848 Reviewed-by: Leif Lindholm <[email protected]>
22849 Reviewed-by: Daniel Kiper <[email protected]>
22850
22851 2019-07-11 David Michael <[email protected]>
22852
22853 verifiers: Blocklist fallout cleanup
22854 Blocklist fallout cleanup after commit 5c6f9bc15 (generic/blocklist: Fix
22855 implicit declaration of function grub_file_filter_disable_compression()).
22856
22857 Reviewed-by: Daniel Kiper <[email protected]>
22858
22859 2019-07-11 Andreas Schwab <[email protected]>
22860
22861 RISC-V: Fix computation of pc-relative relocation offset
22862 The offset calculation was missing the relocation addend.
22863
22864 Tested-by: Chester Lin <[email protected]>
22865 Reviewed-by: Daniel Kiper <[email protected]>
22866
22867 2019-07-11 Leif Lindholm <[email protected]>
22868
22869 configure: Disable arm movw/movt relocations for GCC
22870 When building for arm, we already disable movw/movt relocations for clang,
22871 since they are incompatible with PE.
22872
22873 When building with bare metal GCC toolchains (like the one used in the
22874 travis ci scripts), we end up with these relocations again. So add an
22875 additional test for the '-mword-relocations' flag used by GCC.
22876
22877 Reported-by: Alexander Graf <[email protected]>
22878 Reviewed-by: Daniel Kiper <[email protected]>
22879
22880 2019-07-11 Jacob Kroon <[email protected]>
22881
22882 probe: Support probing for partition UUID with --part-uuid
22883 Linux supports root=PARTUUID=<partuuid> boot argument, so add
22884 support for probing it. Compared to the fs UUID, the partition
22885 UUID does not change when reformatting a partition.
22886
22887 For now, only disks using a GPT partition table are supported.
22888
22889 Reviewed-by: Paul Menzel <[email protected]>
22890 Reviewed-by: Daniel Kiper <[email protected]>
22891
22892 2019-07-05 Daniel Kiper <[email protected]>
22893
22894 Bump version to 2.05
22895
22896 2019-07-04 Daniel Kiper <[email protected]>
22897
22898 Release 2.04
22899
22900 2019-06-24 Thomas Schmitt <[email protected]>
22901
22902 docs: Document workaround for grub-mkrescue with older MacBooks
22903 Add a description of the workaround for firmware of older MacBooks
22904 which stalls with a grub-mkrescue ISO image for x86_64-efi target
22905 on an USB stick.
22906
22907 Reviewed-by: Daniel Kiper <[email protected]>
22908
22909 2019-06-24 Eric Snowberg <[email protected]>
22910
22911 docs: Bootstrap changes required for older distros
22912 Some older distros do not contain gettext 0.18. Document the workaround
22913 to use the bootstrap utility on these systems.
22914
22915 Reviewed-by: Daniel Kiper <[email protected]>
22916
22917 2019-06-07 Leif Lindholm <[email protected]>
22918
22919 ia64: build fix in cache.h
22920 Add IA64 to the architectures excluding a declaration for
22921 grub_arch_sync_dma_caches().
22922
22923 IA64 does not include any of the source files that require the function,
22924 but was overlooked for d8901e3ba115 ("cache: Fix compilation for ppc,
22925 sparc and arm64").
22926
22927 Add it to the list of excluding architectures in order to not get
22928 missing symbol errors when running grub-mkimage.
22929
22930 Reported-by: Alexander Graf <[email protected]>
22931 Tested-by: John Paul Adrian Glaubitz <[email protected]>
22932 Reviewed-by: Daniel Kiper <[email protected]>
22933
22934 2019-06-07 Vladimir 'phcoder' Serbinenko <[email protected]>
22935
22936 hostfs: #undef open and close.
22937 Unlike in case of disks in this case it's just a single place, so it's easier
22938 to just #undef
22939
22940 Reviewed-by: Daniel Kiper <[email protected]>
22941
22942 2019-06-03 John Paul Adrian Glaubitz <[email protected]>
22943
22944 f2fs: Disable gcc9 -Waddress-of-packed-member
22945 Disable the -Wadress-of-packaed-member diagnostic for the grub_f2fs_label
22946 function since the result is found to be false postive.
22947
22948 A pointer to the 'volume_name' member of 'struct grub_f2fs_superblock' is
22949 guaranteed to be aligned as the offset of 'volume_name' within the struct
22950 is dividable by the natural alignment on both 32- and 64-bit targets.
22951
22952 grub-core/fs/f2fs.c: In function ‘grub_f2fs_label’:
22953 grub-core/fs/f2fs.c:1253:60: error: taking address of packed member of ‘struct grub_f2fs_superblock’ may result in an unaligned pointer value [-Werror=address-of-packed-member]
22954 1253 | *label = (char *) grub_f2fs_utf16_to_utf8 (data->sblock.volume_name);
22955 | ~~~~~~~~~~~~^~~~~~~~~~~~
22956 cc1: all warnings being treated as errors
22957
22958 Reported-by: Neil MacLeod <[email protected]>
22959 Tested-by: Neil MacLeod <[email protected]>
22960 Reviewed-by: Daniel Kiper <[email protected]>
22961
22962 2019-05-20 Vincent Legoll <[email protected]>
22963
22964 grub-mkrescue: Fix error message about the wrong command having failed: mformat instead of mcopy
22965 Reviewed-by: Daniel Kiper <[email protected]>
22966
22967 2019-05-20 Mathieu Trudel-Lapierre <[email protected]>
22968
22969 video: skip 'text' gfxpayload if not supported, to fallback to default
22970 On UEFI, 'text' gfxpayload is not supported, but we still reach parse_modespec()
22971 with it, which will obviously fail. Fortunately, whatever gfxpayload is set,
22972 we still still have the 'auto' default to fall back to. Allow getting to this
22973 fallback by not trying to parse 'text' as a modespec.
22974
22975 This is because 'text' correctly doesn't parse as a modespec, and ought to have
22976 been ignored before we got to that point, just like it is immediately picked if
22977 we're running on a system where 'text' is a supported video mode.
22978
22979 Bug: https://savannah.gnu.org/bugs/index.php?56217
22980
22981 Reviewed-by: Daniel Kiper <[email protected]>
22982
22983 2019-05-20 Ovidiu Panait <[email protected]>
22984
22985 grub-mkconfig: Use -c instead of --printf for stat
22986 "--printf" only works with the stat variant provided by coreutils.
22987
22988 With busybox, stat will fail with the following error:
22989 stat: unrecognized option '--printf=%T'
22990
22991 Usage: stat [OPTIONS] FILE...
22992
22993 Reviewed-by: Daniel Kiper <[email protected]>
22994
22995 2019-05-20 Michael Chang <[email protected]>
22996
22997 f2fs: Fix gcc9 error -Werror=maybe-uninitialized
22998 The function grub_get_node_path() could return uninitialized offset with
22999 level == 0 if the block is greater than direct_index + 2 * direct_blks +
23000 2 * indirect_blks + dindirect_blks. The uninitialized offset is then used
23001 by function grub_f2fs_get_block() because level == 0 is valid and
23002 meaningful return to be processed.
23003
23004 The fix is to set level = -1 as return value by grub_get_node_path() to
23005 signify an error that the input block cannot be handled. Any caller
23006 should therefore check level is negative or not before processing the
23007 output.
23008
23009 Reported-by: Neil MacLeod <[email protected]>
23010 Tested-by: Neil MacLeod <[email protected]>
23011 Reviewed-by: Daniel Kiper <[email protected]>
23012
23013 2019-05-06 Alexander Graf <[email protected]>
23014
23015 arm: Align section alignment with manual relocation offset code
23016 The arm relocation code has a manual special case for EFI binaries to
23017 add the natural alignment to its own relocation awareness.
23018
23019 Since commit a51f953f4ee87 ("mkimage: Align efi sections on 4k
23020 boundary") we changed that alignment from 0x400 to 0x1000 bytes. Reflect
23021 the change in that branch that we forgot as well.
23022
23023 This fixes running 32bit arm grub efi binaries for me again.
23024
23025 Fixes: a51f953f4ee87 ("mkimage: Align efi sections on 4k boundary")
23026 Reported-by: Heinrich Schuchardt <[email protected]>
23027 Reported-by: Steve McIntyre <[email protected]>
23028 Reviewed-by: Daniel Kiper <[email protected]>
23029 Tested-by: Julien ROBIN <[email protected]>
23030 Reviewed-by: Leif Lindholm <[email protected]>
23031 Tested-by: Leif Lindholm <[email protected]>
23032 Reviewed-by: Daniel Kiper <[email protected]>
23033
23034 2019-05-06 Alexander Graf <[email protected]>
23035
23036 arm: Move trampolines into code section
23037 When creating T32->A32 transition jumps, the relocation code in grub
23038 will generate trampolines. These trampolines live in the .data section
23039 of our PE binary which means they are not marked as executable.
23040
23041 This misbehavior was unmasked by commit a51f953f4ee87 ("mkimage: Align
23042 efi sections on 4k boundary") which made the X/NX boundary more obvious
23043 because everything became page aligned.
23044
23045 To put things into proper order, let's move the arm trampolines into the
23046 .text section instead. That way everyone knows they are executable.
23047
23048 Fixes: a51f953f4ee87 ("mkimage: Align efi sections on 4k boundary")
23049 Reported-by: Julien ROBIN <[email protected]>
23050 Reported-by: Leif Lindholm <[email protected]>
23051 Tested-by: Julien ROBIN <[email protected]>
23052 Reviewed-by: Leif Lindholm <[email protected]>
23053 Tested-by: Leif Lindholm <[email protected]>
23054 Reviewed-by: Daniel Kiper <[email protected]>
23055
23056 2019-04-23 Michael Chang <[email protected]>
23057
23058 efi: Fix gcc9 error -Waddress-of-packed-member
23059 The address of fp->path_name could be unaligned since seeking into the
23060 device path buffer for a given node could end in byte boundary.
23061
23062 The fix is allocating aligned buffer by grub_malloc for holding the
23063 UTF16 string copied from fp->path_name, and after using that buffer as
23064 argument for grub_utf16_to_utf8 to convert it to UTF8 string.
23065
23066 [ 255s] ../../grub-core/kern/efi/efi.c: In function 'grub_efi_get_filename':
23067 [ 255s] ../../grub-core/kern/efi/efi.c:410:60: error: taking address of packed member of 'struct grub_efi_file_path_device_path' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23068 [ 255s] 410 | p = (char *) grub_utf16_to_utf8 ((unsigned char *) p, fp->path_name, len);
23069 [ 255s] | ~~^~~~~~~~~~~
23070 [ 255s] ../../grub-core/kern/efi/efi.c: In function 'grub_efi_print_device_path':
23071 [ 255s] ../../grub-core/kern/efi/efi.c:900:33: error: taking address of packed member of 'struct grub_efi_file_path_device_path' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23072 [ 255s] 900 | *grub_utf16_to_utf8 (buf, fp->path_name,
23073 [ 255s] | ~~^~~~~~~~~~~
23074
23075 Reviewed-by: Daniel Kiper <[email protected]>
23076
23077 2019-04-23 Michael Chang <[email protected]>
23078
23079 chainloader: Fix gcc9 error -Waddress-of-packed-member
23080 The address of fp->path_name could be unaligned since seeking into the
23081 device path buffer for a given node could end in byte boundary.
23082
23083 The fix is using aligned buffer allocated by grub_malloc for receiving
23084 the converted UTF16 string by grub_utf8_to_utf16 and also the processing
23085 after. The resulting string then gets copied to fp->path_name.
23086
23087 [ 243s] ../../grub-core/loader/efi/chainloader.c: In function 'copy_file_path':
23088 [ 243s] ../../grub-core/loader/efi/chainloader.c:136:32: error: taking address of packed member of 'struct grub_efi_file_path_device_path' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23089 [ 243s] 136 | size = grub_utf8_to_utf16 (fp->path_name, len * GRUB_MAX_UTF16_PER_UTF8,
23090 [ 243s] | ~~^~~~~~~~~~~
23091 [ 243s] ../../grub-core/loader/efi/chainloader.c:138:12: error: taking address of packed member of 'struct grub_efi_file_path_device_path' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23092 [ 243s] 138 | for (p = fp->path_name; p < fp->path_name + size; p++)
23093 [ 243s] | ^~
23094
23095 Reviewed-by: Daniel Kiper <[email protected]>
23096
23097 2019-04-23 Michael Chang <[email protected]>
23098
23099 usbtest: Disable gcc9 -Waddress-of-packed-member
23100 Disable the -Wadress-of-packaed-member diagnostic for the
23101 grub_usb_get_string function since the result is false postive. The
23102 descstrp->str is found to be aligned in the buffer allocated for 'struct
23103 grub_usb_desc_str'.
23104
23105 [ 229s] ../../grub-core/commands/usbtest.c: In function 'grub_usb_get_string':
23106 [ 229s] ../../grub-core/commands/usbtest.c:104:58: error: taking address of packed member of 'struct grub_usb_desc_str' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23107 [ 229s] 104 | *grub_utf16_to_utf8 ((grub_uint8_t *) *string, descstrp->str,
23108 [ 229s] | ~~~~~~~~^~~~~
23109
23110 Reviewed-by: Daniel Kiper <[email protected]>
23111
23112 2019-04-23 Michael Chang <[email protected]>
23113
23114 acpi: Fix gcc9 error -Waddress-of-packed-member
23115 Simply adds the missing packed attribute to 'struct grub_acpi_madt'.
23116
23117 [ 233s] ../../grub-core/commands/lsacpi.c: In function 'disp_acpi_xsdt_table':
23118 [ 233s] ../../grub-core/commands/lsacpi.c:201:27: error: converting a packed 'struct grub_acpi_table_header' pointer (alignment 1) to a 'struct grub_acpi_madt' pointer (alignment 4) may result in an unaligned pointer value [-Werror=address-of-packed-member]
23119 [ 233s] 201 | disp_madt_table ((struct grub_acpi_madt *) t);
23120 [ 233s] | ^~~~~~~~~~~~~~
23121 [ 233s] In file included from ../../grub-core/commands/lsacpi.c:23:
23122 [ 233s] ../../include/grub/acpi.h:50:8: note: defined here
23123 [ 233s] 50 | struct grub_acpi_table_header
23124 [ 233s] | ^~~~~~~~~~~~~~~~~~~~~~
23125 [ 233s] ../../include/grub/acpi.h:90:8: note: defined here
23126 [ 233s] 90 | struct grub_acpi_madt
23127 [ 233s] | ^~~~~~~~~~~~~~
23128 [ 233s] ../../grub-core/commands/lsacpi.c: In function 'disp_acpi_rsdt_table':
23129 [ 233s] ../../grub-core/commands/lsacpi.c:225:27: error: converting a packed 'struct grub_acpi_table_header' pointer (alignment 1) to a 'struct grub_acpi_madt' pointer (alignment 4) may result in an unaligned pointer value [-Werror=address-of-packed-member]
23130 [ 233s] 225 | disp_madt_table ((struct grub_acpi_madt *) t);
23131 [ 233s] | ^~~~~~~~~~~~~~
23132 [ 233s] In file included from ../../grub-core/commands/lsacpi.c:23:
23133 [ 233s] ../../include/grub/acpi.h:50:8: note: defined here
23134 [ 233s] 50 | struct grub_acpi_table_header
23135 [ 233s] | ^~~~~~~~~~~~~~~~~~~~~~
23136 [ 233s] ../../include/grub/acpi.h:90:8: note: defined here
23137 [ 233s] 90 | struct grub_acpi_madt
23138 [ 233s] | ^~~~~~~~~~~~~~
23139
23140 Reviewed-by: Daniel Kiper <[email protected]>
23141
23142 2019-04-23 Michael Chang <[email protected]>
23143
23144 hfsplus: Fix gcc9 error with -Waddress-of-packed-member
23145 The catkey->name could be unaligned since the address of 'void* record'
23146 is calculated as offset in bytes to a malloc buffer.
23147
23148 The fix is using aligned buffer allocated by grub_malloc for holding
23149 the UTF16 string copied from catkey->name. And use that buffer as
23150 argument for grub_utf16_to_utf8 to convert to UTF8 strings.
23151
23152 In addition, using a new copy of buffer rather than catkey->name itself
23153 for processing the endianess conversion, we can also get rid of the hunk
23154 restoring byte order of catkey->name to what it was previously.
23155
23156 [ 59s] ../grub-core/fs/hfsplus.c: In function 'list_nodes':
23157 [ 59s] ../grub-core/fs/hfsplus.c:738:57: error: taking address of packed member of 'struct grub_hfsplus_catkey' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23158 [ 59s] 738 | *grub_utf16_to_utf8 ((grub_uint8_t *) filename, catkey->name,
23159 [ 59s] | ~~~~~~^~~~~~
23160 [ 59s] ../grub-core/fs/hfsplus.c: In function 'grub_hfsplus_label':
23161 [ 59s] ../grub-core/fs/hfsplus.c:1019:57: error: taking address of packed member of 'struct grub_hfsplus_catkey' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23162 [ 59s] 1019 | *grub_utf16_to_utf8 ((grub_uint8_t *) (*label), catkey->name,
23163 [ 59s] | ~~~~~~^~~~~~
23164
23165 Reviewed-by: Daniel Kiper <[email protected]>
23166
23167 2019-04-23 Michael Chang <[email protected]>
23168
23169 hfs: Fix gcc9 error -Waddress-of-packed-member
23170 Simply adds the missing packed attribute to 'struct grub_hfs_extent'.
23171
23172 [ 83s] ../grub-core/fs/hfs.c: In function 'grub_hfs_iterate_records':
23173 [ 83s] ../grub-core/fs/hfs.c:699:9: error: taking address of packed member of 'struct grub_hfs_sblock' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23174 [ 83s] 699 | ? (&data->sblock.catalog_recs)
23175 [ 83s] | ~^~~~~~~~~~~~~~~~~~~~~~~~~~~
23176 [ 83s] ../grub-core/fs/hfs.c:700:9: error: taking address of packed member of 'struct grub_hfs_sblock' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23177 [ 83s] 700 | : (&data->sblock.extent_recs));
23178 [ 83s] | ~^~~~~~~~~~~~~~~~~~~~~~~~~~
23179
23180 Reviewed-by: Daniel Kiper <[email protected]>
23181
23182 2019-04-23 Michael Chang <[email protected]>
23183
23184 jfs: Disable gcc9 -Waddress-of-packed-member
23185 Disable the -Wadress-of-packaed-member diagnostic for the
23186 grub_jfs_getent function since the result is found to be false postive.
23187
23188 The leaf is read into memory as continous chunks in size of 32 bytes and
23189 the pointer to its base is aligned, which also guarentee its member
23190 leaf->namepart is aligned.
23191
23192 [ 60s] ../grub-core/fs/jfs.c: In function 'grub_jfs_getent':
23193 [ 60s] ../grub-core/fs/jfs.c:557:44: error: taking address of packed member of 'struct grub_jfs_leaf_dirent' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23194 [ 60s] 557 | le_to_cpu16_copy (filename + strpos, leaf->namepart, len < diro->data->namecomponentlen ? len
23195 [ 60s] | ~~~~^~~~~~~~~~
23196 [ 60s] ../grub-core/fs/jfs.c:570:48: error: taking address of packed member of 'struct grub_jfs_leaf_next_dirent' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23197 [ 60s] 570 | le_to_cpu16_copy (filename + strpos, next_leaf->namepart, len < 15 ? len : 15);
23198 [ 60s] | ~~~~~~~~~^~~~~~~~~~
23199 [ 60s] cc1: all warnings being treated as errors
23200
23201 Reviewed-by: Daniel Kiper <[email protected]>
23202
23203 2019-04-23 Michael Chang <[email protected]>
23204
23205 cpio: Disable gcc9 -Waddress-of-packed-member
23206 Disable the -Wadress-of-packaed-member diagnostic for the
23207 grub_cpio_find_file function since the result is found to be false
23208 postive. Any pointers to member of the 'struct head hd' is aligned even
23209 if the structure is packed without paddings.
23210
23211 [ 59s] In file included from ../grub-core/fs/cpio.c:51:
23212 [ 59s] ../grub-core/fs/cpio_common.c: In function 'grub_cpio_find_file':
23213 [ 59s] ../grub-core/fs/cpio_common.c:58:31: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23214 [ 59s] 58 | data->size = read_number (hd.filesize, ARRAY_SIZE (hd.filesize));
23215 [ 59s] | ~~^~~~~~~~~
23216 [ 59s] ../grub-core/fs/cpio_common.c:60:29: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23217 [ 59s] 60 | *mtime = read_number (hd.mtime, ARRAY_SIZE (hd.mtime));
23218 [ 59s] | ~~^~~~~~
23219 [ 59s] ../grub-core/fs/cpio_common.c:61:28: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23220 [ 59s] 61 | modeval = read_number (hd.mode, ARRAY_SIZE (hd.mode));
23221 [ 59s] | ~~^~~~~
23222 [ 59s] ../grub-core/fs/cpio_common.c:62:29: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23223 [ 59s] 62 | namesize = read_number (hd.namesize, ARRAY_SIZE (hd.namesize));
23224 [ 59s] | ~~^~~~~~~~~
23225 [ 59s] In file included from ../grub-core/fs/cpio_be.c:51:
23226 [ 59s] ../grub-core/fs/cpio_common.c: In function 'grub_cpio_find_file':
23227 [ 59s] ../grub-core/fs/cpio_common.c:58:31: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23228 [ 59s] 58 | data->size = read_number (hd.filesize, ARRAY_SIZE (hd.filesize));
23229 [ 59s] | ~~^~~~~~~~~
23230 [ 59s] ../grub-core/fs/cpio_common.c:60:29: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23231 [ 59s] 60 | *mtime = read_number (hd.mtime, ARRAY_SIZE (hd.mtime));
23232 [ 59s] | ~~^~~~~~
23233 [ 59s] ../grub-core/fs/cpio_common.c:61:28: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23234 [ 59s] 61 | modeval = read_number (hd.mode, ARRAY_SIZE (hd.mode));
23235 [ 59s] | ~~^~~~~
23236 [ 59s] ../grub-core/fs/cpio_common.c:62:29: error: taking address of packed member of 'struct head' may result in an unaligned pointer value [-Werror=address-of-packed-member]
23237 [ 59s] 62 | namesize = read_number (hd.namesize, ARRAY_SIZE (hd.namesize));
23238 [ 59s] | ~~^~~~~~~~~
23239
23240 Reviewed-by: Daniel Kiper <[email protected]>
23241
23242 2019-04-23 Heinrich Schuchardt <[email protected]>
23243
23244 efi: Avoid NULL dereference if FilePath is NULL
23245 The UEFI specification allows LoadImage() to be called with a memory
23246 location only and without a device path. In this case FilePath will not be
23247 set in the EFI_LOADED_IMAGE_PROTOCOL.
23248
23249 So in function grub_efi_get_filename() the device path argument may be
23250 NULL. As we cannot determine the device path in this case just return NULL
23251 from the function.
23252
23253 Reviewed-by: Leif Lindholm <[email protected]>
23254 Reviewed-by: Daniel Kiper <[email protected]>
23255
23256 2019-04-23 Daniel Kiper <[email protected]>
23257
23258 x86/msr: Fix build with older GCC versions
23259 Some older GCC versions produce following error when x86 MSR modules are build:
23260
23261 In file included from commands/i386/rdmsr.c:29:0:
23262 ../include/grub/i386/rdmsr.h:27:29: error: no previous prototype for ‘grub_msr_read’ [-Werror=missing-prototypes]
23263 extern inline grub_uint64_t grub_msr_read (grub_uint32_t msr_id)
23264 ^
23265 cc1: all warnings being treated as errors
23266
23267 This happens due to lack of support for a such usage of extern keyword
23268 in older GCCs. Additionally, this usage is not consistent with the rest
23269 of codebase. So, replace it with static keyword.
23270
23271 Additionally, fix incorrect coding style.
23272
23273 Reported-by: Eric Snowberg <[email protected]>
23274 Reported-by: adrian15 <[email protected]>
23275 Reviewed-by: Vladimir 'phcoder' Serbinenko <[email protected]>
23276 Reviewed-by: Eric Snowberg <[email protected]>
23277 Tested-by: adrian15 <[email protected]>
23278
23279 2019-04-09 Vladimir Serbinenko <[email protected]>
23280
23281 Release 2.04~rc1
23282
23283 2019-04-09 Vladimir Serbinenko <[email protected]>
23284
23285 Change fs functions to add fs_ prefix
23286 This avoid conflict with gnulib
23287
23288 Reviewed-by: Daniel Kiper <[email protected]>
23289
23290 2019-04-08 Vladimir Serbinenko <[email protected]>
23291
23292 A workaround for clang problem assembling startup_raw.S
23293 Reviewed-by: Daniel Kiper <[email protected]>
23294
23295 2019-04-04 Eric Snowberg <[email protected]>
23296
23297 ieee1275: NULL pointer dereference in grub_ieee1275_encode_devname()
23298 Function grub_strndup() may return NULL, this is called from
23299 function grub_ieee1275_get_devname() which is then called from
23300 function grub_ieee1275_encode_devname() to set device. The device
23301 variable could then be used with a NULL pointer.
23302
23303 Reviewed-by: Colin Watson <[email protected]>
23304 Reviewed-by: Daniel Kiper <[email protected]>
23305
23306 2019-04-02 Daniel Kiper <[email protected]>
23307
23308 docs/grub-dev: Change comments rules
23309 Current comments forms are annoying, so, some of them are disallowed
23310 starting from now. New rules are more flexible and mostly aligned
23311 with, e.g., Linux kernel comments rules.
23312
23313 Reviewed-by: Vladimir Serbinenko <[email protected]>
23314
23315 2019-04-02 Andrew Jeddeloh <[email protected]>
23316
23317 loader/i386/linux: Calculate the setup_header length
23318 Previously the setup_header length was just assumed to be the size of the
23319 linux_kernel_params struct. The linux x86 32-bit boot protocol says that the
23320 end of the linux_i386_kernel_header is at 0x202 + the byte value at 0x201 in
23321 the linux_i386_kernel_header. So, calculate the size of the header using the
23322 end of the linux_i386_kernel_header, rather than assume it is the size of the
23323 linux_kernel_params struct.
23324
23325 Additionally, add some required members to the linux_kernel_params
23326 struct and align the content of linux_i386_kernel_header struct with
23327 it. New members naming was taken directly from Linux kernel source.
23328
23329 linux_kernel_params and linux_i386_kernel_header structs require more
23330 cleanup. However, this is not urgent, so, let's do this after release.
23331 Just in case...
23332
23333 Reviewed-by: Vladimir Serbinenko <[email protected]>
23334 Reviewed-by: Ross Philipson <[email protected]>
23335
23336 2019-04-02 Eric Snowberg <[email protected]>
23337
23338 efidisk: NULL pointer dereference in grub_efidisk_get_device_name()
23339 Function grub_efi_find_last_device_path() may return NULL when called
23340 from grub_efidisk_get_device_name().
23341
23342 Reviewed-by: Daniel Kiper <[email protected]>
23343
23344 2019-04-02 Eric Snowberg <[email protected]>
23345
23346 efidisk: NULL pointer dereference in is_child()
23347 Function grub_efi_find_last_device() path may return NULL when called
23348 from is_child().
23349
23350 Reviewed-by: Daniel Kiper <[email protected]>
23351
23352 2019-04-02 Eric Snowberg <[email protected]>
23353
23354 efidisk: Write to NULL pointer ldp
23355 Function grub_efi_find_last_device_path() may return constant NULL when
23356 called from find_parent_device().
23357
23358 Reviewed-by: Daniel Kiper <[email protected]>
23359
23360 2019-04-02 Vladimir Serbinenko <[email protected]>
23361
23362 clang: Pair -Qn with -Qunused-arguments.
23363 When assembling module wirh clang -Qn ends up on command line but later ignored
23364 To avoid it breaking the compile, add -Qunused-arguments.
23365
23366 Reviewed-by: Daniel Kiper <[email protected]>
23367
23368 2019-03-28 John Paul Adrian Glaubitz <[email protected]>
23369
23370 ieee1275: Fix path reference in comment of sparc64 boot loader code
23371 Reviewed-by: Daniel Kiper <[email protected]>
23372
23373 2019-03-28 John Paul Adrian Glaubitz <[email protected]>
23374
23375 ieee1275: Include a.out header in assembly of sparc64 boot loader
23376 Recent versions of binutils dropped support for the a.out and COFF
23377 formats on sparc64 targets. Since the boot loader on sparc64 is
23378 supposed to be an a.out binary and the a.out header entries are
23379 rather simple to calculate in our case, we just write the header
23380 ourselves instead of relying on external tools to do that.
23381
23382 Reviewed-by: Daniel Kiper <[email protected]>
23383
23384 2019-03-26 Vladimir Serbinenko <[email protected]>
23385
23386 Propagate GNU_PRINTF from gnulib vfprintf
23387 gnulib now replaces vfprintf and hence its format becomes GNU_PRINTF format
23388
23389 This also fixes matching definitions to always use GNU format
23390
23391 Reviewed-by: Daniel Kiper <[email protected]>
23392
23393 2019-03-26 Vladimir Serbinenko <[email protected]>
23394
23395 efi/tpm.c: Add missing casts
23396 Without those casts we get a warning about implicit conversion of pointer
23397 to integer.
23398
23399 2019-03-26 Vladimir Serbinenko <[email protected]>
23400
23401 POTFILES: Don't include gnulib in grub.pot
23402 They're translated as a separate project, so we
23403 don't want to submit them again.
23404
23405 Reviewed-by: Daniel Kiper <[email protected]>
23406
23407 2019-03-26 Vladimir Serbinenko <[email protected]>
23408
23409 configure.ac: Use nostdlib when checking for nostdinc
23410 With clang nostdinc behaviour is influenced by nostdlib. Since we
23411 always add nostdlib, add it in test as well
23412
23413 Reviewed-by: Daniel Kiper <[email protected]>
23414
23415 2019-03-25 Vladimir Serbinenko <[email protected]>
23416
23417 efi/tpm.h: Fix hash_log_extend_event definition.
23418 I didn't check the spec but pointer to address doesn't make much sense
23419 and doesn't match the code.
23420
23421 Rename grub_disk members
23422 Otherwise it horribly clashes with gnulib when it's
23423 replacing open/write/read/close
23424
23425 grub-mkimagexx: Fix RISCV error message
23426 Outputting a raw pointer doesn't match the format and is
23427 also useless. Output offset instead.
23428
23429 kern/emu/misc.c: Don't include config-util.h when running as GRUB_BUILD
23430
23431 Support R_PPC_PLTREL24
23432 It's emitted by clang 7. It's the same as R_PPC_REL24.
23433
23434 2019-03-20 Daniel Kiper <[email protected]>
23435
23436 sparc: Enable __clzsi2() and __clzdi2()
23437 This patch is similiar to commit e795b9011 (RISC-V: Add libgcc helpers
23438 for clz) but for SPARC target.
23439
23440 Reviewed-by: Ross Philipson <[email protected]>
23441
23442 2019-03-20 Daniel Kiper <[email protected]>
23443
23444 mips: Enable __clzsi2()
23445 This patch is similiar to commit e795b9011 (RISC-V: Add libgcc helpers
23446 for clz) but for MIPS target.
23447
23448 Reviewed-by: Ross Philipson <[email protected]>
23449
23450 2019-03-20 Daniel Kiper <[email protected]>
23451
23452 verifiers: MIPS fallout cleanup
23453 MIPS fallout cleanup after commit 4d4a8c96e (verifiers: Add possibility
23454 to verify kernel and modules command lines).
23455
23456 Reviewed-by: Ross Philipson <[email protected]>
23457
23458 2019-03-20 Daniel Kiper <[email protected]>
23459
23460 verifiers: PowerPC fallout cleanup
23461 PowerPC fallout cleanup after commit 4d4a8c96e (verifiers: Add possibility
23462 to verify kernel and modules command lines) and ca0a4f689 (verifiers: File
23463 type for fine-grained signature-verification controlling).
23464
23465 Reviewed-by: Ross Philipson <[email protected]>
23466
23467 2019-03-20 Daniel Kiper <[email protected]>
23468
23469 verifiers: IA-64 fallout cleanup
23470 IA-64 fallout cleanup after commit 4d4a8c96e (verifiers: Add possibility
23471 to verify kernel and modules command lines).
23472
23473 Reviewed-by: Ross Philipson <[email protected]>
23474
23475 2019-03-20 Colin Watson <[email protected]>
23476
23477 posix_wrap: Flesh out posix_wrap/limits.h a little more
23478 In addition to what was already there, Gnulib's <intprops.h> needs SCHAR_MIN,
23479 SCHAR_MAX, SHRT_MIN, INT_MIN, LONG_MIN, and LONG_MAX. Fixes build on CentOS 7.
23480
23481 Reported-by: "Chen, Farrah" <[email protected]>
23482 Reviewed-by: Daniel Kiper <[email protected]>
23483
23484 2019-03-19 Marek Marczykowski-Górecki <[email protected]>
23485
23486 xen: Look for Xen notes in section headers too
23487 Mirror behaviour of ELF loader in libxc: first look for Xen notes in
23488 PT_NOTE segment, then in SHT_NOTE section and only then fallback to
23489 a section with __xen_guest name. This fixes loading PV kernels that
23490 Xen note have outside of PT_NOTE. While this may be result of a buggy
23491 linker script, loading such kernel directly works fine, so make it work
23492 with GRUB too. Specifically, this applies to binaries built from Unikraft.
23493
23494 Reviewed-by: Daniel Kiper <[email protected]>
23495
23496 2019-03-19 Colin Watson <[email protected]>
23497
23498 getroot: Save/restore CWD more reliably on Unix
23499 Various GRUB utilities fail if the current directory doesn't exist,
23500 because grub_find_device() chdirs to a different directory and then
23501 fails when trying to chdir back. Gnulib's save-cwd module uses fchdir()
23502 instead when it can, avoiding this category of problem.
23503
23504 Fixes Debian bug #918700.
23505
23506 Reviewed-by: Daniel Kiper <[email protected]>
23507
23508 2019-03-12 Andrei Borzenkov <[email protected]>
23509
23510 net/dhcp: Add explicit net_dhcp command
23511 Mostly for cosmetic reasons, we add a "net_dhcp" command, which is (at the
23512 moment) identical to the existing "net_bootp" command. Both actually trigger
23513 a DHCP handshake now, and both should be able to deal with pure BOOTP servers.
23514 We could think about dropping the DHCP options from the initial DISCOVER packet
23515 when the user issues the net_bootp command, but it's unclear whether this is
23516 really useful, as both protocols should be able to coexist.
23517
23518 Reviewed-by: Daniel Kiper <[email protected]>
23519
23520 2019-03-12 Andrei Borzenkov <[email protected]>
23521
23522 net/dhcp: Actually send out DHCPv4 DISCOVER and REQUEST messages
23523 Even though we were parsing some DHCP options sent by the server, so far
23524 we are only using the BOOTP 2-way handshake, even when talking to a DHCP
23525 server.
23526
23527 Change this by actually sending out DHCP DISCOVER packets instead of the
23528 generic (mostly empty) BOOTP BOOTREQUEST packets.
23529
23530 A pure BOOTP server would ignore the extra DHCP options in the DISCOVER
23531 packet and would just reply with a BOOTREPLY packet, which we also
23532 handle in the code.
23533
23534 Reviewed-by: Daniel Kiper <[email protected]>
23535
23536 2019-03-12 Andrei Borzenkov <[email protected]>
23537
23538 net/dhcp: Allow receiving DHCP OFFER and ACK packets
23539 In respone to a BOOTREQUEST packet a BOOTP server would answer with a BOOTREPLY
23540 packet, which ends the conversation for good. DHCP uses a 4-way handshake,
23541 where the initial server respone is an OFFER, which has to be answered with
23542 REQUEST by the client again, only to be completed by an ACKNOWLEDGE packet
23543 from the server.
23544
23545 Teach the grub_net_process_dhcp() function to deal with OFFER packets,
23546 and treat ACK packets the same es BOOTREPLY packets.
23547
23548 Reviewed-by: Daniel Kiper <[email protected]>
23549
23550 2019-03-12 Andrei Borzenkov <[email protected]>
23551
23552 net/dhcp: Use DHCP options for name and bootfile
23553 The BOOTP RFC describes the boot file name and the server name as being part
23554 of the integral BOOTP data structure, with some limits on the size of them.
23555 DHCP extends this by allowing them to be separate DHCP options, which is more
23556 flexible.
23557
23558 Teach the code dealing with those fields to check for those DHCP options first
23559 and use this information, if provided. We fall back to using the BOOTP
23560 information if those options are not used.
23561
23562 Reviewed-by: Daniel Kiper <[email protected]>
23563
23564 2019-03-12 Andrei Borzenkov <[email protected]>
23565
23566 net/dhcp: Introduce per-interface timeout
23567 Currently we have a global timeout for all network cards in the BOOTP/DHCP
23568 discovery process.
23569
23570 Make this timeout a per-interface one, so better accommodate the upcoming
23571 4-way DHCP handshake and to also cover the lease time limit a DHCP offer
23572 will come with.
23573
23574 Reviewed-by: Daniel Kiper <[email protected]>
23575
23576 2019-03-12 Andrei Borzenkov <[email protected]>
23577
23578 net/dhcp: Make grub_net_process_dhcp() take an interface
23579 Change the interface of the function dealing with incoming BOOTP packets
23580 to take an interface instead of a card, to allow more fine per-interface
23581 state (timeout, handshake state) later on.
23582
23583 Use the opportunity to clean up the code a bit.
23584
23585 Reviewed-by: Daniel Kiper <[email protected]>
23586
23587 2019-03-12 Andrei Borzenkov <[email protected]>
23588
23589 net/dhcp: Refactor DHCP packet transmission into separate function
23590 In contrast to BOOTP, DHCP uses a 4-way handshake, so requires to send
23591 packets more often.
23592
23593 Refactor the generation and sending of the BOOTREQUEST packet into
23594 a separate function, so that future code can more easily reuse this.
23595
23596 Reviewed-by: Daniel Kiper <[email protected]>
23597
23598 2019-03-12 Andrei Borzenkov <[email protected]>
23599
23600 net/dhcp: Allow overloading legacy bootfile and name field
23601 DHCP specifies a special dummy option OVERLOAD, to allow DHCP options to
23602 spill over into the (legacy) BOOTFILE and SNAME fields.
23603
23604 Parse and handle this option properly.
23605
23606 Reviewed-by: Daniel Kiper <[email protected]>
23607
23608 2019-03-12 Andrei Borzenkov <[email protected]>
23609
23610 net/dhcp: Replace parse_dhcp_vendor() with find_dhcp_option()
23611 For proper DHCP support we will need to parse DHCP options from a packet
23612 more often and at various places.
23613
23614 Refactor the option parsing into a new function, which will scan a packet to
23615 find *a particular* option field. Use that new function in places where we
23616 were dealing with DHCP options before.
23617
23618 Reviewed-by: Daniel Kiper <[email protected]>
23619
23620 2019-03-12 Andrei Borzenkov <[email protected]>
23621
23622 net/dhcp: Remove dead code
23623 The comment is right, the "giaddr" fields holds the IP address of the BOOTP
23624 relay, not a general purpose router address. Just remove the commented code,
23625 archeologists can find it in the git history.
23626
23627 Reviewed-by: Daniel Kiper <[email protected]>
23628
23629 2019-03-12 Jesús Diéguez Fernández <[email protected]>
23630
23631 msr: Add new MSR modules (rdmsr/wrmsr)
23632 In order to be able to read from and write to model-specific registers,
23633 two new modules are added. They are i386 specific, as the cpuid module.
23634
23635 rdmsr module registers the command rdmsr that allows reading from a MSR.
23636 wrmsr module registers the command wrmsr that allows writing to a MSR.
23637
23638 wrmsr module is disabled if UEFI secure boot is enabled.
23639
23640 Please note that on SMP systems, interacting with a MSR that has a scope
23641 per hardware thread, implies that the value only applies to the
23642 particular cpu/core/thread that ran the command.
23643
23644 Also, if you specify a reserved or unimplemented MSR address, it will
23645 cause a general protection exception (which is not currently being
23646 handled) and the system will reboot.
23647
23648 Reviewed-by: Daniel Kiper <[email protected]>
23649
23650 2019-03-12 Jesús Diéguez Fernández <[email protected]>
23651
23652 asm: Replace "__asm__ __volatile__" with "asm volatile"
23653 In order to maintain the coding style consistency, it was requested to
23654 replace the methods that use "__asm__ __volatile__" with "asm volatile".
23655
23656 Reviewed-by: Daniel Kiper <[email protected]>
23657
23658 2019-03-12 Eric Snowberg <[email protected]>
23659
23660 sparc64: Add bios boot partition support
23661 Add BIOS Boot Partition support for sparc64 platforms. This will work a
23662 little different than x86. With GPT, both the OBP "load" and "boot" commands
23663 are partition aware and neither command can see the partition table. Therefore
23664 the entire boot-loader is stored within the BIOS Boot Partition and nothing
23665 is stored within the bootstrap code area of MBR.
23666
23667 To use it, the end user will issue the boot command with the path pointing to
23668 the BIOS Boot Partition.
23669
23670 For example with the disk below:
23671
23672 Model: Unknown (unknown)
23673 Disk /dev/nvme1n1: 1600GB
23674 Sector size (logical/physical): 512B/512B
23675 Partition Table: gpt
23676
23677 Number Start End Size File system Name Flags
23678 1 1049kB 1075MB 1074MB ext3
23679 2 1075MB 1076MB 1049kB bios_grub
23680 3 1076MB 1600GB 1599GB lvm
23681
23682 To boot grub2 from OBP, you would use:
23683
23684 boot /pci@302/pci@1/pci@0/pci@13/nvme@0/disk@1:b
23685
23686 Reviewed-by: Daniel Kiper <[email protected]>
23687
23688 2019-03-12 Eric Snowberg <[email protected]>
23689
23690 ieee1275: obdisk driver
23691 Add a new disk driver called obdisk for IEEE1275 platforms. Currently
23692 the only platform using this disk driver is SPARC, however other IEEE1275
23693 platforms could start using it if they so choose. While the functionality
23694 within the current IEEE1275 ofdisk driver may be suitable for PPC and x86, it
23695 presented too many problems on SPARC hardware.
23696
23697 Within the old ofdisk, there is not a way to determine the true canonical
23698 name for the disk. Within Open Boot, the same disk can have multiple names
23699 but all reference the same disk. For example the same disk can be referenced
23700 by its SAS WWN, using this form:
23701
23702 /pci@302/pci@2/pci@0/pci@17/LSI,sas@0/disk@w5000cca02f037d6d,0
23703
23704 It can also be referenced by its PHY identifier using this form:
23705
23706 /pci@302/pci@2/pci@0/pci@17/LSI,sas@0/disk@p0
23707
23708 It can also be referenced by its Target identifier using this form:
23709
23710 /pci@302/pci@2/pci@0/pci@17/LSI,sas@0/disk@0
23711
23712 Also, when the LUN=0, it is legal to omit the ,0 from the device name. So with
23713 the disk above, before taking into account the device aliases, there are 6 ways
23714 to reference the same disk.
23715
23716 Then it is possible to have 0 .. n device aliases all representing the same disk.
23717 Within this new driver the true canonical name is determined using the the
23718 IEEE1275 encode-unit and decode-unit commands when address_cells == 4. This
23719 will determine the true single canonical name for the device so multiple ihandles
23720 are not opened for the same device. This is what frequently happens with the old
23721 ofdisk driver. With some devices when they are opened multiple times it causes
23722 the entire system to hang.
23723
23724 Another problem solved with this driver is devices that do not have a device
23725 alias can be booted and used within GRUB. Within the old ofdisk, this was not
23726 possible, unless it was the original boot device. All devices behind a SAS
23727 or SCSI parent can be found. Within the old ofdisk, finding these disks
23728 relied on there being an alias defined. The alias requirement is not
23729 necessary with this new driver. It can also find devices behind a parent
23730 after they have been hot-plugged. This is something that is not possible
23731 with the old ofdisk driver.
23732
23733 The old ofdisk driver also incorrectly assumes that the device pointing to by a
23734 device alias is in its true canonical form. This assumption is never made with
23735 this new driver.
23736
23737 Another issue solved with this driver is that it properly caches the ihandle
23738 for all open devices. The old ofdisk tries to do this by caching the last
23739 opened ihandle. However this does not work properly because the layer above
23740 does not use a consistent device name for the same disk when calling into the
23741 driver. This is because the upper layer uses the bootpath value returned within
23742 /chosen, other times it uses the device alias, and other times it uses the
23743 value within grub.cfg. It does not have a way to figure out that these devices
23744 are the same disk. This is not a problem with this new driver.
23745
23746 Due to the way GRUB repeatedly opens and closes the same disk. Caching the
23747 ihandle is important on SPARC. Without caching, some SAS devices can take
23748 15 - 20 minutes to get to the GRUB menu. This ihandle caching is not possible
23749 without correctly having the canonical disk name.
23750
23751 When available, this driver also tries to use the deblocker #blocks and
23752 a way of determining the disk size.
23753
23754 Finally and probably most importantly, this new driver is also capable of
23755 seeing all partitions on a GPT disk. With the old driver, the GPT
23756 partition table can not be read and only the first partition on the disk
23757 can be seen.
23758
23759 Reviewed-by: Daniel Kiper <[email protected]>
23760
23761 2019-03-12 Paul Menzel <[email protected]>
23762
23763 Makefile: Allow to set file systems modules for default_payload.elf
23764 By default all file system modules are added to the GRUB coreboot
23765 payload `default_payload.elf`. This makes the image quite big,
23766 especially as often not all modules are needed.
23767
23768 Introduce the variable `FS_PAYLOAD_MODULES`, which can be used to
23769 explicitly set file systems modules to be added.
23770
23771 $ make default_payload.elf
23772 test -f default_payload.elf && rm default_payload.elf || true
23773 pkgdatadir=. ./grub-mkstandalone --grub-mkimage=./grub-mkimage -O i386-coreboot -o default_payload.elf --modules='ahci pata ehci uhci ohci usb_keyboard usbms part_msdos ext2 fat at_keyboard part_gpt usbserial_usbdebug cbfs' --install-modules='ls linux search configfile normal cbtime cbls memrw iorw minicmd lsmmap lspci halt reboot hexdump pcidump regexp setpci lsacpi chain test serial multiboot cbmemc linux16 gzio echo help syslinuxcfg xnu affs afs bfs btrfs cbfs cpio cpio_be exfat ext2 f2fs fat hfs hfsplus iso9660 jfs minix minix2 minix2_be minix3 minix3_be minix_be newc nilfs2 ntfs odc procfs reiserfs romfs sfs squash4 tar udf ufs1 ufs1_be ufs2 xfs zfs password_pbkdf2 ' --fonts= --themes= --locales= -d grub-core/ /boot/grub/grub.cfg=./coreboot.cfg
23774 $ ls -l default_payload.elf
23775 -rw-rw---- 1 joey joey 1199568 Mar 6 13:58 default_payload.elf
23776
23777 $ make default_payload.elf FS_PAYLOAD_MODULES="" # ext2 already in `--modules`
23778 test -f default_payload.elf && rm default_payload.elf || true
23779 pkgdatadir=. ./grub-mkstandalone --grub-mkimage=./grub-mkimage -O i386-coreboot -o default_payload.elf --modules='ahci pata ehci uhci ohci usb_keyboard usbms part_msdos ext2 fat at_keyboard part_gpt usbserial_usbdebug cbfs' --install-modules='ls linux search configfile normal cbtime cbls memrw iorw minicmd lsmmap lspci halt reboot hexdump pcidump regexp setpci lsacpi chain test serial multiboot cbmemc linux16 gzio echo help syslinuxcfg xnu password_pbkdf2 ' --fonts= --themes= --locales= -d grub-core/ /boot/grub/grub.cfg=./coreboot.cfg
23780 $ ls -l default_payload.elf
23781 -rw-rw---- 1 joey joey 832976 Mar 7 12:13 default_payload.elf
23782
23783 So, the resulting payload size is around 370 kB smaller. (Adding it to
23784 the CBFS, it will be compressed, so the effective size difference will
23785 be smaller.)
23786
23787 Reviewed-by: Daniel Kiper <[email protected]>
23788
23789 2019-03-07 Vladimir Serbinenko <[email protected]>
23790
23791 windows/platform.c: Fix compilation errors
23792
23793 2019-03-05 Colin Watson <[email protected]>
23794
23795 gnulib: Upgrade Gnulib and switch to bootstrap tool
23796 Upgrade Gnulib files to 20190105.
23797
23798 It's much easier to maintain GRUB's use of portability support files
23799 from Gnulib when the process is automatic and driven by a single
23800 configuration file, rather than by maintainers occasionally running
23801 gnulib-tool and committing the result. Removing these
23802 automatically-copied files from revision control also removes the
23803 temptation to hack the output in ways that are difficult for future
23804 maintainers to follow. Gnulib includes a "bootstrap" program which is
23805 designed for this.
23806
23807 The canonical way to bootstrap GRUB from revision control is now
23808 "./bootstrap", but "./autogen.sh" is still useful if you just want to
23809 generate the GRUB-specific parts of the build system.
23810
23811 GRUB now requires Autoconf >= 2.63 and Automake >= 1.11, in line with
23812 Gnulib.
23813
23814 Gnulib source code is now placed in grub-core/lib/gnulib/ (which should
23815 not be edited directly), and GRUB's patches are in
23816 grub-core/lib/gnulib-patches/. I've added a few notes to the developer
23817 manual on how to maintain this.
23818
23819 Reviewed-by: Daniel Kiper <[email protected]>
23820
23821 2019-03-05 Colin Watson <[email protected]>
23822
23823 syslinux: Fix syslinux_test in out-of-tree builds
23824 syslinux_parse simplifies some filenames by removing things like ".."
23825 segments, but the tests assumed that @abs_top_srcdir@ would be
23826 untouched, which is not true in the case of out-of-tree builds where
23827 @abs_top_srcdir@ may contain ".." segments.
23828
23829 Performing the substitution requires some awkwardness in Makefile.am due
23830 to details of how config.status works.
23831
23832 Reviewed-by: Daniel Kiper <[email protected]>
23833
23834 2019-03-05 Colin Watson <[email protected]>
23835
23836 util: Detect more I/O errors
23837 Many of GRUB's utilities don't check anywhere near all the possible
23838 write errors. For example, if grub-install runs out of space when
23839 copying a file, it won't notice. There were missing checks for the
23840 return values of write, fflush, fsync, and close (or the equivalents on
23841 other OSes), all of which must be checked.
23842
23843 I tried to be consistent with the existing logging practices of the
23844 various hostdisk implementations, but they weren't entirely consistent
23845 to start with so I used my judgement. The result at least looks
23846 reasonable on GNU/Linux when I provoke a write error:
23847
23848 Installing for x86_64-efi platform.
23849 grub-install: error: cannot copy `/usr/lib/grub/x86_64-efi-signed/grubx64.efi.signed' to `/boot/efi/EFI/debian/grubx64.efi': No space left on device.
23850
23851 There are more missing checks in other utilities, but this should fix
23852 the most critical ones.
23853
23854 Fixes Debian bug #922741.
23855
23856 Reviewed-by: Steve McIntyre <[email protected]>
23857 Reviewed-by: Daniel Kiper <[email protected]>
23858
23859 2019-03-05 James Clarke <[email protected]>
23860
23861 osdep/freebsd: Fix partition calculation for EBR entries
23862 For EBR partitions, "start" is the relative starting sector of the EBR
23863 header itself, whereas "offset" is the relative starting byte of the
23864 partition's contents, excluding the EBR header and any padding. Thus we
23865 must use "offset", and divide by the sector size to convert to sectors.
23866
23867 Fixes Debian bug #923253.
23868
23869 Reviewed-by: Colin Watson <[email protected]>
23870 Reviewed-by: Daniel Kiper <[email protected]>
23871
23872 2019-02-26 Steve McIntyre <[email protected]>
23873
23874 grub-install: Check for arm-efi as a default target
23875 Much like on x86, we can work out if the system is running on top of EFI
23876 firmware. If so, return "arm-efi". If not, fall back to "arm-uboot" as
23877 previously.
23878
23879 Split out the code to (maybe) load the efivar module and check for
23880 /sys/firmware/efi into a common helper routine is_efi_system().
23881
23882 Reviewed-by: Leif Lindholm <[email protected]>
23883 Reviewed-by: Daniel Kiper <[email protected]>
23884
23885 2019-02-26 Daniel Kiper <[email protected]>
23886
23887 Revert "grub-install: Check for arm-efi as a default target"
23888 This reverts commit 082fd84d525f8d6602f892160b77c0a948308a78.
23889
23890 Incorrect version of the patch was pushed into the git repo.
23891
23892 Reported-by: Leif Lindholm <[email protected]>
23893
23894 2019-02-25 Alexander Graf <[email protected]>
23895
23896 travis: Add Travis CI config file
23897 There is a really convenient service for open source project from Travis
23898 CI: They allow for free CI testing using their infrastructure.
23899
23900 GRUB has had issues with broken builds for various targets for a long time
23901 already. The main reason is a lack of CI to just do smoke tests on whether
23902 all targets still at least compile.
23903
23904 This patch adds a Travis config file which builds (almost) all currently
23905 available targets.
23906
23907 On top of that, this Travis config also runs a small execution test on the
23908 x86_64-efi target.
23909
23910 All of this config file can easily be extended further on. It probably
23911 makes sense to do something similar to the u-boot test infrastructure
23912 that communicates with the payload properly. Going forward, we also will
23913 want to do more QEMU runtime checks for other targets.
23914
23915 Currently, with this config alone, I already see about half of the available
23916 targets as broken. So it's definitely desperately needed :).
23917
23918 Reviewed-by: Daniel Kiper <[email protected]>
23919
23920 2019-02-25 Steve McIntyre <[email protected]>
23921
23922 grub-install: Check for arm-efi as a default target
23923 Much like on x86, we can work out if the system is running on top
23924 of EFI firmware. If so, return "arm-efi". If not, fall back to
23925 "arm-uboot" as previously.
23926
23927 Heavily inspired by the existing code for x86.
23928
23929 Reviewed-by: Leif Lindholm <[email protected]>
23930 Reviewed-by: Daniel Kiper <[email protected]>
23931
23932 2019-02-25 Leif Lindholm <[email protected]>
23933
23934 arm64/efi: Fix grub_efi_get_ram_base()
23935 grub_efi_get_ram_base() looks for the lowest available RAM address by
23936 traversing the memory map, comparing lowest address found so far.
23937 Due to a brain glitch, that "so far" was initialized to GRUB_UINT_MAX -
23938 completely preventing boot on systems without RAM below 4GB.
23939
23940 Change the initial value to GRUB_EFI_MAX_USABLE_ADDRESS, as originally
23941 intended.
23942
23943 Reported-by: Steve McIntyre <[email protected]>
23944 Tested-by: Steve McIntyre <[email protected]>
23945 Reviewed-by: Daniel Kiper <[email protected]>
23946
23947 2019-02-25 Paul Menzel <[email protected]>
23948
23949 normal/menu: Do not treat error values as key presses
23950 Some terminals, like `grub-core/term/at_keyboard.c`, return `-1` in case
23951 they are not ready yet.
23952
23953 if (! KEYBOARD_ISREADY (grub_inb (KEYBOARD_REG_STATUS)))
23954 return -1;
23955
23956 Currently, that is treated as a key press, and the menu time-out is
23957 cancelled/cleared. This is unwanted, as the boot is stopped and the user
23958 manually has to select a menu entry. Therefore, adapt the condition to
23959 require the key value also to be greater than 0.
23960
23961 `GRUB_TERM_NO_KEY` is defined as 0, so the condition could be collapsed
23962 to greater or equal than (≥) 0, but the compiler will probably do that
23963 for us anyway, so keep the cases separate for clarity.
23964
23965 This is tested with coreboot, the GRUB default payload, and the
23966 configuration file `grub.cfg` below.
23967
23968 For GRUB:
23969
23970 $ ./autogen.sh
23971 $ ./configure --with-platform=coreboot
23972 $ make -j`nproc`
23973 $ make default_payload.elf
23974
23975 For coreboot:
23976
23977 $ more grub.cfg
23978 serial --unit 0 --speed 115200
23979 set timeout=5
23980
23981 menuentry 'halt' {
23982 halt
23983 }
23984 $ build/cbfstool build/coreboot.rom add-payload \
23985 -f /dev/shm/grub/default_payload.elf -n fallback/payload -c lzma
23986 $ build/cbfstool build/coreboot.rom add -f grub.cfg -n etc/grub.cfg -t raw
23987 $ qemu-system-x86_64 --version
23988 QEMU emulator version 3.1.0 (Debian 1:3.1+dfsg-2+b1)
23989 Copyright (c) 2003-2018 Fabrice Bellard and the QEMU Project developers
23990 $ qemu-system-x86_64 -M pc -bios build/coreboot.rom -serial stdio -nic none
23991
23992 Currently, the time-out is cancelled/cleared. With the commit, it is not.
23993 With a small GRUB payload, this the problem is also reproducible on the
23994 ASRock E350M1.
23995
23996 Link: http://lists.gnu.org/archive/html/grub-devel/2019-01/msg00037.html
23997
23998 Reviewed-by: Daniel Kiper <[email protected]>
23999
24000 2019-02-25 Alexander Graf <[email protected]>
24001
24002 fdt: Treat device tree file type like ACPI
24003 We now have signature check logic in grub which allows us to treat
24004 files differently depending on their file type.
24005
24006 Treat a loaded device tree like an overlayed ACPI table.
24007 Both describe hardware, so I suppose their threat level is the same.
24008
24009 Reviewed-by: Daniel Kiper <[email protected]>
24010 Reviewed-by: Alistair Francis <[email protected]>
24011
24012 2019-02-25 Alexander Graf <[email protected]>
24013
24014 RISC-V: Add to build system
24015 This patch adds support for RISC-V to the grub build system. With this
24016 patch, I can successfully build grub on RISC-V as a UEFI application.
24017
24018 Reviewed-by: Alistair Francis <[email protected]>
24019 Reviewed-by: Bin Meng <[email protected]>
24020 Tested-by: Bin Meng <[email protected]>
24021 Reviewed-by: Daniel Kiper <[email protected]>
24022
24023 2019-02-25 Alexander Graf <[email protected]>
24024
24025 RISC-V: Add libgcc helpers for clz
24026 Gcc may decide it wants to call helper functions to execute clz. Provide
24027 them in our own copy of libgcc.
24028
24029 Reviewed-by: Daniel Kiper <[email protected]>
24030
24031 2019-02-25 Alexander Graf <[email protected]>
24032
24033 RISC-V: Add auxiliary files
24034 To support a new architecture we need to provide a few helper functions
24035 for memory, cache, timer, etc support.
24036
24037 This patch adds the remainders of those. Some bits are still disabled,
24038 as I couldn't guarantee that we're always running on models / in modes
24039 where the respective hardware is available.
24040
24041 Reviewed-by: Alistair Francis <[email protected]>
24042 Reviewed-by: Daniel Kiper <[email protected]>
24043
24044 2019-02-25 Alexander Graf <[email protected]>
24045
24046 RISC-V: Add awareness for RISC-V reloations
24047 This patch adds awareness of RISC-V relocations throughout the grub tools
24048 as well as dynamic linkage and elf->PE relocation conversion support.
24049
24050 Reviewed-by: Alistair Francis <[email protected]>
24051 Reviewed-by: Daniel Kiper <[email protected]>
24052
24053 2019-02-25 Alexander Graf <[email protected]>
24054
24055 RISC-V: Add Linux load logic
24056 We currently only support to run grub on RISC-V as UEFI payload. Ideally,
24057 we also only want to support running Linux underneath as UEFI payload.
24058
24059 Prepare that with some Linux boot stub code. Once the arm64 target is
24060 generalized, we can hook into that one and gain boot functionality.
24061
24062 Reviewed-by: Alistair Francis <[email protected]>
24063 Reviewed-by: Daniel Kiper <[email protected]>
24064
24065 2019-02-25 Alexander Graf <[email protected]>
24066
24067 RISC-V: Add early startup code
24068 On entry, we need to save the system table pointer as well as our image
24069 handle. Add an early startup file that saves them and then brings us
24070 into our main function.
24071
24072 Reviewed-by: Alistair Francis <[email protected]>
24073 Reviewed-by: Bin Meng <[email protected]>
24074 Tested-by: Bin Meng <[email protected]>
24075 Reviewed-by: Daniel Kiper <[email protected]>
24076
24077 2019-02-25 Alexander Graf <[email protected]>
24078
24079 RISC-V: Add setjmp implementation
24080 This patch adds a 32/64 capable setjmp implementation for RISC-V.
24081
24082 Reviewed-by: Alistair Francis <[email protected]>
24083 Reviewed-by: Bin Meng <[email protected]>
24084 Tested-by: Bin Meng <[email protected]>
24085 Reviewed-by: Daniel Kiper <[email protected]>
24086
24087 2019-02-25 Alexander Graf <[email protected]>
24088
24089 elf.h: Add RISC-V definitions
24090 The RISC-V ABI document outlines ELF header structure and relocation
24091 information. Pull the respective magic numbers into our elf header
24092 so we can make use of them.
24093
24094 Reviewed-by: Alistair Francis <[email protected]>
24095 Reviewed-by: Bin Meng <[email protected]>
24096 Tested-by: Bin Meng <[email protected]>
24097 Reviewed-by: Daniel Kiper <[email protected]>
24098
24099 2019-02-25 Alexander Graf <[email protected]>
24100
24101 PE: Add RISC-V definitions
24102 The PE format defines magic numbers as well as relocation identifiers for
24103 RISC-V. Add them to our include file, so we can make use of them.
24104
24105 Reviewed-by: Leif Lindholm <[email protected]>
24106 Reviewed-by: Alistair Francis <[email protected]>
24107 Reviewed-by: Bin Meng <[email protected]>
24108 Tested-by: Bin Meng <[email protected]>
24109 Reviewed-by: Daniel Kiper <[email protected]>
24110
24111 2019-02-25 Alexander Graf <[email protected]>
24112
24113 efi: Rename armxx to arch
24114 Some architectures want to boot Linux as plain UEFI binary. Today that
24115 really only encompasses ARM and AArch64, but going forward more
24116 architectures may adopt that model.
24117
24118 So rename our internal API accordingly.
24119
24120 Acked-by: Leif Lindholm <[email protected]>
24121 Reviewed-by: Alistair Francis <[email protected]>
24122 Reviewed-by: Bin Meng <[email protected]>
24123 Tested-by: Bin Meng <[email protected]>
24124 Reviewed-by: Daniel Kiper <[email protected]>
24125
24126 2019-02-06 Alexander Graf <[email protected]>
24127
24128 mkimage: Clarify file alignment in efi case
24129 There are a few spots in the PE generation code for EFI binaries that uses
24130 the section alignment rather than file alignment, even though the alignment
24131 is really only file bound.
24132
24133 Replace those cases with the file alignment constant instead.
24134
24135 Reported-by: Daniel Kiper <[email protected]>
24136 Reviewed-by: Daniel Kiper <[email protected]>
24137 Tested-by: Julien ROBIN <[email protected]>
24138
24139 2019-02-06 Alexander Graf <[email protected]>
24140
24141 mkimage: Align efi sections on 4k boundary
24142 There is UEFI firmware popping up in the wild now that implements stricter
24143 permission checks using NX and write protect page table entry bits.
24144
24145 This means that firmware now may fail to load binaries if its individual
24146 sections are not page aligned, as otherwise it can not ensure permission
24147 boundaries.
24148
24149 So let's bump all efi section alignments up to 4k (EFI page size). That way
24150 we will stay compatible going forward.
24151
24152 Unfortunately our internals can't deal very well with a mismatch of alignment
24153 between the virtual and file offsets, so we have to also pad our target
24154 binary a bit.
24155
24156 Reviewed-by: Daniel Kiper <[email protected]>
24157 Tested-by: Julien ROBIN <[email protected]>
24158
24159 2019-02-06 Alexander Graf <[email protected]>
24160
24161 mkimage: Use EFI32_HEADER_SIZE define in arm-efi case
24162 The efi-arm case was defining its own header size calculation, even though it's
24163 100% identical to the common EFI32_HEADER_SIZE definition.
24164
24165 So let's clean it up to use the common define.
24166
24167 Reviewed-by: Daniel Kiper <[email protected]>
24168 Tested-by: Julien ROBIN <[email protected]>
24169
24170 2019-02-06 Guillaume GARDET <[email protected]>
24171
24172 arm: Move initrd upper to leave more space for kernel
24173 This patch allows to have bigger kernels. If the kernel grows, then it will
24174 overwrite the initrd when it is extracted.
24175
24176 Acked-by: Alexander Graf <[email protected]>
24177 Reviewed-by: Daniel Kiper <[email protected]>
24178
24179 2019-01-23 Leif Lindholm <[email protected]>
24180
24181 linux, efi, arm*, fdt: Break FDT extra allocation space out into a #define
24182 A certain amount of dynamic space is required for the handover from
24183 GRUB/Linux-EFI-stub. This entails things like initrd addresses,
24184 address-cells entries and associated strings.
24185
24186 But move this into a proper centralised #define rather than live-code
24187 it in the loader.
24188
24189 Reviewed-by: Daniel Kiper <[email protected]>
24190
24191 2019-01-22 Cristian Ciocaltea <[email protected]>
24192
24193 uboot: Add the missing disk write operation support
24194 uboot_disk_write() is currently lacking the write support
24195 to storage devices because, historically, those devices did not
24196 implement block_write() in U-Boot.
24197
24198 The solution has been tested using a patched U-Boot loading
24199 and booting GRUB in a QEMU vexpress-a9 environment.
24200 The disk write operations were triggered with GRUB's save_env
24201 command.
24202
24203 Reviewed-by: Daniel Kiper <[email protected]>
24204
24205 2019-01-21 Max Tottenham <[email protected]>
24206
24207 tpm: Fix bug in GRUB2 TPM module
24208 The value of tpm_handle changes between successive calls to grub_tpm_handle_find(),
24209 as instead of simply copying the stored pointer we end up taking the address of
24210 said pointer when using the cached value of grub_tpm_handle.
24211
24212 This causes grub_efi_open_protocol() to return a nullptr in grub_tpm2_execute()
24213 and grub_tpm2_log_event(). Said nullptr goes unchecked and
24214 efi_call_5(tpm->hash_log_extend_event,...) ends up jumping to 0x0, Qemu crashes
24215 once video ROM is reached at 0xb0000.
24216
24217 This patch seems to do the trick of fixing that bug, but we should also ensure
24218 that all calls to grub_efi_open_protocol() are checked so that we don't start
24219 executing low memory.
24220
24221 Reviewed-by: Matthew Garrett <[email protected]>
24222 Reviewed-by: Daniel Kiper <[email protected]>
24223
24224 2019-01-14 Colin Watson <[email protected]>
24225
24226 pgp: Fix emu build and tests after pgp module renaming
24227 Commit b07feb8746c3bb845e3f0d33d37c0bded704d14d (verifiers: Rename
24228 verify module to pgp module) renamed the "verify" module to "pgp", but
24229 the GRUB_MOD_INIT and GRUB_MOD_FINI macros were left as "verify", which
24230 broke the emu target build; and file_filter_test still referred to the
24231 now non-existent "verify" module. Fix both of these.
24232
24233 Reviewed-by: Daniel Kiper <[email protected]>
24234
24235 2019-01-14 Peter Große <[email protected]>
24236
24237 grub-mkconfig/20_linux_xen: Support multiple early initrd images
24238 Add support for multiple, shared, early initrd images. These early
24239 images will be loaded in the order declared, and all will be loaded
24240 before the initrd image.
24241
24242 While many classes of data can be provided by early images, the
24243 immediate use case would be for distributions to provide CPU
24244 microcode to mitigate the Meltdown and Spectre vulnerabilities.
24245
24246 Xen has also support to load microcode updates provided as additional
24247 modules by the bootloader.
24248
24249 There are two environment variables provided for declaring the early
24250 images.
24251
24252 * GRUB_EARLY_INITRD_LINUX_STOCK is for the distribution declare
24253 images that are provided by the distribution or installed packages.
24254 If undeclared, this will default to a set of common microcode image
24255 names.
24256
24257 * GRUB_EARLY_INITRD_LINUX_CUSTOM is for user created images. User
24258 images will be loaded after the stock images.
24259
24260 These separate configurations allow the distribution and user to
24261 declare different image sets without clobbering each other.
24262
24263 This also makes a minor update to ensure that UUID partition labels
24264 stay disabled when no initrd image is found, even if early images are
24265 present.
24266
24267 This is basically a copy of a698240d "grub-mkconfig/10_linux: Support
24268 multiple early initrd images" by Matthew S. Turnbull.
24269
24270 Reviewed-by: Daniel Kiper <[email protected]>
24271
24272 2019-01-14 Heinrich Schuchardt <[email protected]>
24273
24274 grub-core/loader/efi/fdt.c: Do not copy random memory
24275 We should not try to copy any memory area which is outside of the original
24276 fdt. If this extra memory is controlled by a hypervisor this might end
24277 with a crash.
24278
24279 Reviewed-by: Leif Lindholm <[email protected]>
24280 Reviewed-by: Daniel Kiper <[email protected]>
24281
24282 2018-12-12 Matthew Garrett <[email protected]>
24283
24284 verifiers: Add TPM documentation
24285 Describe the behaviour of GRUB when the TPM module is in use.
24286
24287 Reviewed-by: Daniel Kiper <[email protected]>
24288
24289 2018-12-12 Matthew Garrett <[email protected]>
24290
24291 verifiers: Core TPM support
24292 Add support for performing basic TPM measurements. Right now this only
24293 supports extending PCRs statically and only on UEFI. In future we might
24294 want to have some sort of mechanism for choosing which events get logged
24295 to which PCRs, but this seems like a good default policy and we can wait
24296 to see whether anyone has a use case before adding more complexity.
24297
24298 Reviewed-by: Daniel Kiper <[email protected]>
24299
24300 2018-12-12 Matthew Garrett <[email protected]>
24301
24302 verifiers: Verify commands executed by grub
24303 Pass all commands executed by GRUB to the verifiers layer. Most verifiers will
24304 ignore this, but some (such as the TPM verifier) want to be able to measure and
24305 log each command executed in order to ensure that the boot state is as expected.
24306
24307 Reviewed-by: Daniel Kiper <[email protected]>
24308
24309 2018-12-12 Juergen Gross <[email protected]>
24310
24311 xen_pvh: Add support to configure
24312 Support platform i386/xen_pvh in configure.
24313
24314 Reviewed-by: Daniel Kiper <[email protected]>
24315 Tested-by: Hans van Kranenburg <[email protected]>
24316
24317 2018-12-12 Juergen Gross <[email protected]>
24318
24319 xen_pvh: Support grub-install for xen_pvh
24320 Add xen_pvh support to grub-install.
24321
24322 Reviewed-by: Daniel Kiper <[email protected]>
24323 Tested-by: Hans van Kranenburg <[email protected]>
24324
24325 2018-12-12 Juergen Gross <[email protected]>
24326
24327 xen_pvh: Support building a standalone image
24328 Support mkimage for xen_pvh.
24329
24330 Reviewed-by: Daniel Kiper <[email protected]>
24331 Tested-by: Hans van Kranenburg <[email protected]>
24332
24333 2018-12-12 Juergen Gross <[email protected]>
24334
24335 xen: Use elfnote defines instead of plain numbers
24336 In order to avoid using plain integers for the ELF notes use the
24337 available Xen include instead.
24338
24339 Reviewed-by: Daniel Kiper <[email protected]>
24340 Tested-by: Hans van Kranenburg <[email protected]>
24341
24342 2018-12-12 Hans van Kranenburg <[email protected]>
24343
24344 grub-module-verifier: Ignore all_video for xen_pvh
24345 This solves the build failing with "Error: no symbol table and no
24346 .moddeps section"
24347
24348 Also see:
24349 - 6371e9c10433578bb236a8284ddb9ce9e201eb59
24350 - https://savannah.gnu.org/bugs/?49012
24351
24352 Reviewed-by: Daniel Kiper <[email protected]>
24353 Tested-by: Hans van Kranenburg <[email protected]>
24354
24355 2018-12-12 Juergen Gross <[email protected]>
24356
24357 xen_pvh: Add build runes for grub-core
24358 Add the modifications to the build system needed to build a xen_pvh
24359 grub.
24360
24361 Reviewed-by: Daniel Kiper <[email protected]>
24362 Tested-by: Hans van Kranenburg <[email protected]>
24363
24364 2018-12-12 Juergen Gross <[email protected]>
24365
24366 xen: Init memory regions for PVH
24367 Add all usable memory regions to grub memory management and add the
24368 needed mmap iterate code, which will be used by grub core (e.g.
24369 grub-core/lib/relocator.c or grub-core/mmap/mmap.c).
24370
24371 As we are running in 32-bit mode don't add memory above 4GB.
24372
24373 Reviewed-by: Daniel Kiper <[email protected]>
24374 Tested-by: Hans van Kranenburg <[email protected]>
24375
24376 2018-12-12 Juergen Gross <[email protected]>
24377
24378 xen: Setup Xen specific data for PVH
24379 Initialize the needed Xen specific data. This is:
24380
24381 - the Xen start of day page containing the console and Xenstore ring
24382 page PFN and event channel
24383 - the grant table
24384 - the shared info page
24385
24386 Write back the possibly modified memory map to the hypervisor in case
24387 the guest is reading it from there again.
24388
24389 Set the RSDP address for the guest from the start_info page passed
24390 as boot parameter.
24391
24392 Reviewed-by: Daniel Kiper <[email protected]>
24393 Reviewed-by: Roger Pau Monné <[email protected]>
24394 Tested-by: Hans van Kranenburg <[email protected]>
24395
24396 2018-12-12 Juergen Gross <[email protected]>
24397
24398 xen: Get memory map from hypervisor for PVH
24399 Retrieve the memory map from the hypervisor and normalize it to contain
24400 no overlapping entries and to be sorted by address.
24401
24402 Reviewed-by: Daniel Kiper <[email protected]>
24403 Tested-by: Hans van Kranenburg <[email protected]>
24404
24405 2018-12-12 Juergen Gross <[email protected]>
24406
24407 xen: Setup hypercall page for PVH
24408 Add the needed code to setup the hypercall page for calling into the
24409 Xen hypervisor.
24410
24411 Import the XEN_HVM_DEBUGCONS_IOPORT define from Xen unstable into
24412 include/xen/arch-x86/xen.h
24413
24414 Reviewed-by: Roger Pau Monné <[email protected]>
24415 Reviewed-by: Daniel Kiper <[email protected]>
24416 Tested-by: Hans van Kranenburg <[email protected]>
24417
24418 2018-12-12 Juergen Gross <[email protected]>
24419
24420 xen: Add PVH boot entry code
24421 Add the code for the Xen PVH mode boot entry.
24422
24423 Reviewed-by: Daniel Kiper <[email protected]>
24424 Tested-by: Hans van Kranenburg <[email protected]>
24425
24426 2018-12-12 Juergen Gross <[email protected]>
24427
24428 xen: Add basic hooks for PVH in current code
24429 Add the hooks to current code needed for Xen PVH. They will be filled
24430 with code later when the related functionality is being added.
24431
24432 loader/i386/linux.c needs to include machine/kernel.h now as it needs
24433 to get GRUB_KERNEL_USE_RSDP_ADDR from there. This in turn requires to
24434 add an empty kernel.h header for some i386 platforms (efi, coreboot,
24435 ieee1275, xen) and for x86_64 efi.
24436
24437 Reviewed-by: Daniel Kiper <[email protected]>
24438 Tested-by: Hans van Kranenburg <[email protected]>
24439
24440 2018-12-12 Juergen Gross <[email protected]>
24441
24442 xen: Add PVH specific defines to offset.h
24443 include/grub/offsets.h needs some defines for Xen PVH mode.
24444
24445 Add them. While at it line up the values in the surrounding lines to
24446 start at the same column.
24447
24448 Reviewed-by: Daniel Kiper <[email protected]>
24449 Tested-by: Hans van Kranenburg <[email protected]>
24450
24451 2018-12-12 Juergen Gross <[email protected]>
24452
24453 xen: Modify grub_xen_ptr2mfn() for Xen PVH
24454 grub_xen_ptr2mfn() returns the machine frame number for a given pointer
24455 value. For Xen-PVH guests this is just the PFN. Add the PVH specific
24456 variant.
24457
24458 Reviewed-by: Daniel Kiper <[email protected]>
24459 Tested-by: Hans van Kranenburg <[email protected]>
24460
24461 2018-12-12 Juergen Gross <[email protected]>
24462
24463 xen: Rearrange xen/init.c to prepare it for Xen PVH mode
24464 Rearrange grub-core/kern/xen/init.c to prepare adding PVH mode support
24465 to it. This includes putting some code under #ifdef GRUB_MACHINE_XEN
24466 as it will not be used when running as PVH.
24467
24468 Reviewed-by: Daniel Kiper <[email protected]>
24469 Tested-by: Hans van Kranenburg <[email protected]>
24470
24471 2018-12-12 Juergen Gross <[email protected]>
24472
24473 xen: Add some dummy headers for PVH mode
24474 With Xen PVH mode adding a new machine type the machine related headers
24475 need to be present for the build to succeed. Most of the headers just
24476 need to include the related common i386 headers. Add those to the tree.
24477
24478 Note that xen_pvh/int.h needs to include pc/int_types.h instead of
24479 pc/int.h in order to avoid the definition of grub_bios_interrupt().
24480
24481 xen_pvh/memory.h needs to include coreboot/memory.h (like some other
24482 <machine>/memory.h do as well) as this contains just the needed stubs.
24483
24484 Reviewed-by: Daniel Kiper <[email protected]>
24485 Tested-by: Hans van Kranenburg <[email protected]>
24486
24487 2018-12-12 Juergen Gross <[email protected]>
24488
24489 xen: Prepare common code for Xen PVH support
24490 Some common code needs to be special cased for Xen PVH mode. This hits
24491 mostly Xen PV mode specific areas.
24492
24493 Split include/grub/i386/pc/int_types.h off from
24494 include/grub/i386/pc/int.h to support including this file later from
24495 xen_pvh code without the grub_bios_interrupt definition.
24496
24497 Move definition of struct grub_e820_mmap_entry from
24498 grub-core/mmap/i386/pc/mmap.c to include/grub/i386/memory.h in order
24499 to make it usable from xen_pvh code.
24500
24501 Reviewed-by: Daniel Kiper <[email protected]>
24502 Tested-by: Hans van Kranenburg <[email protected]>
24503
24504 2018-12-12 Juergen Gross <[email protected]>
24505
24506 xen: Carve out grant tab initialization into dedicated function
24507 Initialize the grant tab in a dedicated function. This will enable
24508 using it for PVH guests, too.
24509
24510 Call the new function from grub_machine_init() as this will later
24511 be common between Xen PV and Xen PVH mode.
24512
24513 Reviewed-by: Daniel Kiper <[email protected]>
24514 Tested-by: Hans van Kranenburg <[email protected]>
24515
24516 2018-12-12 Juergen Gross <[email protected]>
24517
24518 loader/linux: Support passing RSDP address via boot params
24519 Xen PVH guests will have the RSDP at an arbitrary address. Support that
24520 by passing the RSDP address via the boot parameters to Linux.
24521
24522 Reviewed-by: Daniel Kiper <[email protected]>
24523 Tested-by: Hans van Kranenburg <[email protected]>
24524
24525 2018-12-12 Juergen Gross <[email protected]>
24526
24527 xen: Add some Xen headers
24528 In order to support grub2 in Xen PVH environment some additional Xen
24529 headers are needed as grub2 will be started in PVH mode requiring to
24530 use several HVM hypercalls and structures.
24531
24532 Add the needed headers from Xen 4.10 being the first Xen version with
24533 full (not only experimental) PVH guest support.
24534
24535 Reviewed-by: Daniel Kiper <[email protected]>
24536 Tested-by: Hans van Kranenburg <[email protected]>
24537
24538 2018-12-07 Daniel Kiper <[email protected]>
24539
24540 verifiers: ARM Xen fallout cleanup
24541 ARM Xen fallout cleanup after commit ca0a4f689 (verifiers: File type for
24542 fine-grained signature-verification controlling).
24543
24544 Reviewed-by: Ross Philipson <[email protected]>
24545
24546 2018-12-07 Daniel Kiper <[email protected]>
24547
24548 verifiers: Xen fallout cleanup
24549 Xen fallout cleanup after commit ca0a4f689 (verifiers: File type for
24550 fine-grained signature-verification controlling).
24551
24552 Reviewed-by: Ross Philipson <[email protected]>
24553
24554 2018-11-28 Eric Snowberg <[email protected]>
24555
24556 ofnet: Fix build regression in grub_ieee1275_parse_bootpath()
24557 The grub_ieee1275_parse_bootpath() function (commit a661a32, ofnet: Initialize
24558 structs in bootpath parser) introduces a build regression on SPARC:
24559
24560 cc1: warnings being treated as errors
24561 net/drivers/ieee1275/ofnet.c: In function 'grub_ieee1275_parse_bootpath':
24562 net/drivers/ieee1275/ofnet.c:156: error: missing initializer
24563 net/drivers/ieee1275/ofnet.c:156: error: (near initialization for 'client_addr.type')
24564 net/drivers/ieee1275/ofnet.c:156: error: missing initializer
24565 net/drivers/ieee1275/ofnet.c:156: error: (near initialization for 'gateway_addr.type')
24566 net/drivers/ieee1275/ofnet.c:156: error: missing initializer
24567 net/drivers/ieee1275/ofnet.c:156: error: (near initialization for 'subnet_mask.type')
24568 net/drivers/ieee1275/ofnet.c:157: error: missing initializer
24569 net/drivers/ieee1275/ofnet.c:157: error: (near initialization for 'hw_addr.type')
24570 make[3]: *** [net/drivers/ieee1275/ofnet_module-ofnet.o] Error 1
24571
24572 Initialize the entire structure.
24573
24574 More info can be found here:
24575 http://lists.gnu.org/archive/html/grub-devel/2018-03/msg00034.html
24576
24577 Reviewed-by: Daniel Kiper <[email protected]>
24578
24579 2018-11-26 Nick Terrell <[email protected]>
24580
24581 btrfs: Add zstd support to grub btrfs
24582 - Adds zstd support to the btrfs module.
24583 - Adds a test case for btrfs zstd support.
24584 - Changes top_srcdir to srcdir in the btrfs module's lzo include
24585 following comments from Daniel Kiper about the zstd include.
24586
24587 Tested on Ubuntu-18.04 with a btrfs /boot partition with and without zstd
24588 compression. A test case was also added to the test suite that fails before
24589 the patch, and passes after.
24590
24591 Reviewed-by: Daniel Kiper <[email protected]>
24592
24593 2018-11-26 Nick Terrell <[email protected]>
24594
24595 zstd: Import upstream zstd-1.3.6
24596 - Import zstd-1.3.6 from upstream
24597 - Add zstd's module.c file
24598 - Add the zstd module to Makefile.core.def
24599
24600 Import zstd-1.3.6 from upstream [1]. Only the files need for decompression
24601 are imported. I used the latest zstd release, which includes patches [2] to
24602 build cleanly in GRUB.
24603
24604 I included the script used to import zstd-1.3.6 below at the bottom of the
24605 commit message.
24606
24607 Upstream zstd commit hash: 4fa456d7f12f8b27bd3b2f5dfd4f46898cb31c24
24608 Upstream zstd commit name: Merge pull request #1354 from facebook/dev
24609
24610 Zstd requires some posix headers, which it gets from posix_wrap.
24611 This can be checked by inspecting the .Po files generated by automake,
24612 which contain the header dependencies. After building run the command
24613 `cat grub-core/lib/zstd/.deps-core/*.Po` to see the dependencies [3].
24614 The only OS dependencies are:
24615
24616 - stddef.h, which is already a dependency in posix_wrap, and used for size_t
24617 by lzo and xz.
24618 - stdarg.h, which comes from the grub/misc.h header, and we don't use in zstd.
24619
24620 All the types like uint64_t are typedefed to grub_uint64_t under the hood.
24621 The only exception is size_t, which comes from stddef.h. This is already the
24622 case for lzo and xz. I don't think there are any cross-compilation concerns,
24623 because cross-compilers provide their own system headers (and it would already
24624 be broken).
24625
24626 [1] https://github.com/facebook/zstd/releases/tag/v1.3.6
24627 [2] https://github.com/facebook/zstd/pull/1344
24628 [3] https://gist.github.com/terrelln/7a16b92f5a1b3aecf980f944b4a966c4
24629
24630 ```
24631
24632 curl -L -O https://github.com/facebook/zstd/releases/download/v1.3.6/zstd-1.3.6.tar.gz
24633 curl -L -O https://github.com/facebook/zstd/releases/download/v1.3.6/zstd-1.3.6.tar.gz.sha256
24634 sha256sum --check zstd-1.3.6.tar.gz.sha256
24635 tar xzf zstd-1.3.6.tar.gz
24636
24637 SRC_LIB="zstd-1.3.6/lib"
24638 DST_LIB="grub-core/lib/zstd"
24639 rm -rf $DST_LIB
24640 mkdir -p $DST_LIB
24641 cp $SRC_LIB/zstd.h $DST_LIB/
24642 cp $SRC_LIB/common/*.[hc] $DST_LIB/
24643 cp $SRC_LIB/decompress/*.[hc] $DST_LIB/
24644 rm $DST_LIB/{pool.[hc],threading.[hc]}
24645 rm -rf zstd-1.3.6*
24646 echo SUCCESS!
24647 ```
24648
24649 Reviewed-by: Daniel Kiper <[email protected]>
24650
24651 2018-11-21 Michael Chang <[email protected]>
24652
24653 verifiers: fix double close on pgp's sig file descriptor
24654 An error emerged as when I was testing the verifiers branch, so instead
24655 of putting it in pgp prefix, the verifiers is used to reflect what the
24656 patch is based on.
24657
24658 While running verify_detached, grub aborts with error.
24659
24660 verify_detached /@/.snapshots/1/snapshot/boot/grub/grub.cfg
24661 /@/.snapshots/1/snapshot/boot/grub/grub.cfg.sig
24662
24663 alloc magic is broken at 0x7beea660: 0
24664 Aborted. Press any key to exit.
24665
24666 The error is caused by sig file descriptor been closed twice, first time
24667 in grub_verify_signature() to which it is passed as parameter. Second in
24668 grub_cmd_verify_signature() or in whichever opens the sig file
24669 descriptor. The second close is not consider as bug to me either, as in
24670 common rule of what opens a file has to close it to avoid file
24671 descriptor leakage.
24672
24673 After all the design of grub_verify_signature() makes it difficult to keep
24674 a good trace on opened file descriptor from it's caller. Let's refine
24675 the application interface to accept file path rather than descriptor, in
24676 this way the caller doesn't have to care about closing the descriptor by
24677 delegating it to grub_verify_signature() with full tracing to opened
24678 file descriptor by itself.
24679
24680 Also making it clear that sig descriptor is not referenced in error
24681 returning path of grub_verify_signature_init(), so it can be closed
24682 directly by it's caller. This also makes delegating it to
24683 grub_pubkey_close() infeasible to help in relieving file descriptor
24684 leakage as it has to depend on uncertainty of ctxt fields in error
24685 returning path.
24686
24687 Reviewed-by: Daniel Kiper <[email protected]>
24688
24689 2018-11-21 Lee Jones <[email protected]>
24690
24691 generic/blocklist: Fix implicit declaration of function grub_file_filter_disable_compression()
24692 grub_file_filter_disable_compression() no longer exists.
24693
24694 Reviewed-by: Daniel Kiper <[email protected]>
24695
24696 2018-11-21 Lee Jones <[email protected]>
24697
24698 arm64/xen: Fix too few arguments to function grub_create_loader_cmdline()
24699 Without this fix, building xen_boot.c omits:
24700
24701 loader/arm64/xen_boot.c: In function ‘xen_boot_binary_load’:
24702 loader/arm64/xen_boot.c:370:7: error: too few arguments to function ‘grub_create_loader_cmdline’
24703 grub_create_loader_cmdline (argc - 1, argv + 1, binary->cmdline,
24704 ^~~~~~~~~~~~~~~~~~~~~~~~~~
24705 In file included from loader/arm64/xen_boot.c:36:0:
24706 ../include/grub/lib/cmdline.h:29:12: note: declared here
24707 grub_err_t grub_create_loader_cmdline (int argc, char *argv[], char *buf,
24708
24709 Reviewed-by: Julien Grall <[email protected]>
24710 Reviewed-by: Daniel Kiper <[email protected]>
24711
24712 2018-11-16 Leif Lindholm <[email protected]>
24713
24714 arm-uboot, ia64, sparc64: Fix up grub_file_open() calls
24715 The verifiers framework changed the grub_file_open() interface, breaking all
24716 non-x86 linux loaders. Add file types to the grub_file_open() calls to make
24717 them build again.
24718
24719 Reviewed-by: Daniel Kiper <[email protected]>
24720
24721 2018-11-16 Leif Lindholm <[email protected]>
24722
24723 arm64/efi: Fix breakage caused by verifiers
24724 - add variable "err" (used but not defined),
24725 - add GRUB_FILE_TYPE_LINUX_KERNEL to grub_file_open() call.
24726
24727 Reviewed-by: Daniel Kiper <[email protected]>
24728
24729 2018-11-16 Leif Lindholm <[email protected]>
24730
24731 grub-core/loader/efi/fdt.c: Fixup grub_file_open() call
24732 The verifiers framework changed the API of grub_file_open(), but did not
24733 fix up all users. Add the file type GRUB_FILE_TYPE_DEVICE_TREE_IMAGE
24734 to the "devicetree" command handler call.
24735
24736 Reviewed-by: Daniel Kiper <[email protected]>
24737
24738 2018-11-16 Leif Lindholm <[email protected]>
24739
24740 include/grub/file.h: Add device tree file type
24741 The API change of grub_file_open() for adding verifiers did not include
24742 a type for device tree blobs. Add GRUB_FILE_TYPE_DEVICE_TREE_IMAGE to
24743 the grub_file_type enum.
24744
24745 Reviewed-by: Daniel Kiper <[email protected]>
24746
24747 2018-11-16 Leif Lindholm <[email protected]>
24748
24749 include/grub/verify.h: Add include guard
24750 verify.h was added without include guards. This means compiling anything
24751 including both include/grub/verify.h and include/grub/lib/cmdline.h fails
24752 (at least grub-core/loader/arm64/linux.c.
24753
24754 Add the necessary include guard.
24755
24756 Reviewed-by: Daniel Kiper <[email protected]>
24757
24758 2018-11-16 Matthew Daley <[email protected]>
24759
24760 mkimage: Pad DTBs to target-specific pointer size
24761 Device tree (DTB) lengths are being padded to a multiple of 4 bytes
24762 rather than the target-specific pointer size. This causes objects
24763 following OBJ_TYPE_DTB objects to be incorrectly parsed during GRUB
24764 execution on arm64.
24765
24766 Fix by using ALIGN_ADDR(), not ALIGN_UP().
24767
24768 Signed-by-off: Matthew Daley <[email protected]>
24769 Reviewed-by: Daniel Kiper <[email protected]>
24770
24771 2018-11-09 Colin Watson <[email protected]>
24772
24773 Cope with / being on a ZFS root dataset
24774 If / is on the root dataset in a ZFS pool, then ${bootfs} will be set to
24775 "/" (whereas if it is on a non-root dataset, there will be no trailing
24776 slash). Passing "root=ZFS=${rpool}/" will fail to boot, but
24777 "root=ZFS=${rpool}" works fine, so strip the trailing slash.
24778
24779 Fixes: https://savannah.gnu.org/bugs/?52746
24780
24781 Tested-by: Fejes József <[email protected]>
24782 Reviewed-by: Daniel Kiper <[email protected]>
24783
24784 2018-11-09 Paul Menzel <[email protected]>
24785
24786 unix/platform: Initialize variable to fix grub-install on UEFI system
24787 On a UEFI system, were no boot entry *grub* is present, currently,
24788 `grub-install` fails with an error.
24789
24790 $ efibootmgr
24791 BootCurrent: 0000
24792 Timeout: 0 seconds
24793 BootOrder: 0001,0006,0003,0004,0005
24794 Boot0001 Diskette Drive
24795 Boot0003* USB Storage Device
24796 Boot0004* CD/DVD/CD-RW Drive
24797 Boot0005 Onboard NIC
24798 Boot0006* WDC WD2500AAKX-75U6AA0
24799 $ sudo grub-install /dev/sda
24800 Installing for x86_64-efi platform.
24801 grub-install: error: efibootmgr failed to register the boot entry: Unknown error 22020.
24802
24803 The error code is always different, and the error message (incorrectly)
24804 points to efibootmgr.
24805
24806 But, the error is in GRUB’s function
24807 `grub_install_remove_efi_entries_by_distributor()`, where the variable
24808 `rc` for the return value, is uninitialized and never set, when no boot
24809 entry for the distributor is found.
24810
24811 The content of that uninitialized variable is then returned as the error
24812 code of efibootmgr.
24813
24814 Set the variable to 0, so that success is returned, when no entry needs
24815 to be deleted.
24816
24817 Tested on Dell OptiPlex 7010 with firmware A28.
24818
24819 $ sudo ./grub-install /dev/sda
24820 Installing for x86_64-efi platform.
24821 Installation finished. No error reported.
24822
24823 [1]: https://github.com/rhboot/efibootmgr/issues/100
24824
24825 Reviewed-by: Daniel Kiper <[email protected]>
24826
24827 2018-11-09 Daniel Kiper <[email protected]>
24828
24829 efi: Add EFI shim lock verifier
24830 This module provides shim lock verification for various kernels
24831 if UEFI secure boot is enabled on a machine.
24832
24833 It is recommended to put this module into GRUB2 standalone image
24834 (avoid putting iorw and memrw modules into it; they are disallowed
24835 if UEFI secure boot is enabled). However, it is also possible to use
24836 it as a normal module. Though such configurations are more fragile
24837 and less secure due to various limitations.
24838
24839 If the module is loaded and UEFI secure boot is enabled then:
24840 - module itself cannot be unloaded (persistent module),
24841 - the iorw and memrw modules cannot be loaded,
24842 - if the iorw and memrw modules are loaded then
24843 machine boot is disabled,
24844 - GRUB2 defers modules and ACPI tables verification to
24845 other verifiers.
24846
24847 Reviewed-by: Ross Philipson <[email protected]>
24848
24849 2018-11-09 Daniel Kiper <[email protected]>
24850
24851 dl: Add support for persistent modules
24852 This type of modules cannot be unloaded. This is useful if a given
24853 functionality, e.g. UEFI secure boot shim signature verification, should
24854 not be disabled if it was enabled at some point in time. Somebody may
24855 say that we can use standalone GRUB2 here. That is true. However, the
24856 code is not so big nor complicated hence it make sense to support
24857 modularized configs too.
24858
24859 Reviewed-by: Ross Philipson <[email protected]>
24860
24861 2018-11-09 Vladimir Serbinenko <[email protected]>
24862
24863 verifiers: Add the documentation
24864 Reviewed-by: Ross Philipson <[email protected]>
24865
24866 2018-11-09 Daniel Kiper <[email protected]>
24867
24868 verifiers: Rename verify module to pgp module
24869 Just for clarity. No functional change.
24870
24871 Reviewed-by: Ross Philipson <[email protected]>
24872
24873 2018-11-09 Daniel Kiper <[email protected]>
24874
24875 verifiers: Add possibility to defer verification to other verifiers
24876 This way if a verifier requires verification of a given file it can defer task
24877 to another verifier (another authority) if it is not able to do it itself. E.g.
24878 shim_lock verifier, posted as a subsequent patch, is able to verify only PE
24879 files. This means that it is not able to verify any of GRUB2 modules which have
24880 to be trusted on UEFI systems with secure boot enabled. So, it can defer
24881 verification to other verifier, e.g. PGP one.
24882
24883 I silently assume that other verifiers are trusted and will do good job for us.
24884 Or at least they will not do any harm.
24885
24886 Reviewed-by: Ross Philipson <[email protected]>
24887
24888 2018-11-09 Vladimir Serbinenko <[email protected]>
24889
24890 verifiers: Add possibility to verify kernel and modules command lines
24891 Reviewed-by: Ross Philipson <[email protected]>
24892
24893 2018-11-09 Vladimir Serbinenko <[email protected]>
24894
24895 verifiers: Framework core
24896 Verifiers framework provides core file verification functionality which
24897 can be used by various security mechanisms, e.g., UEFI secure boot, TPM,
24898 PGP signature verification, etc.
24899
24900 The patch contains PGP code changes and probably they should be extracted
24901 to separate patch for the sake of clarity.
24902
24903 Reviewed-by: Ross Philipson <[email protected]>
24904
24905 2018-11-09 Vladimir Serbinenko <[email protected]>
24906
24907 verifiers: File type for fine-grained signature-verification controlling
24908 Let's provide file type info to the I/O layer. This way verifiers
24909 framework and its users will be able to differentiate files and verify
24910 only required ones.
24911
24912 This is preparatory patch.
24913
24914 Reviewed-by: Ross Philipson <[email protected]>
24915
24916 2018-11-09 Daniel Kiper <[email protected]>
24917
24918 bufio: Use grub_size_t instead of plain int for size
24919 Reviewed-by: Ross Philipson <[email protected]>
24920
24921 2018-10-31 Goffredo Baroncelli <[email protected]>
24922
24923 btrfs: Add RAID 6 recovery for a btrfs filesystem
24924 Add the RAID 6 recovery, in order to use a RAID 6 filesystem even if some
24925 disks (up to two) are missing. This code use the md RAID 6 code already
24926 present in grub.
24927
24928 Reviewed-by: Daniel Kiper <[email protected]>
24929
24930 2018-10-31 Goffredo Baroncelli <[email protected]>
24931
24932 btrfs: Make more generic the code for RAID 6 rebuilding
24933 The original code which handles the recovery of a RAID 6 disks array
24934 assumes that all reads are multiple of 1 << GRUB_DISK_SECTOR_BITS and it
24935 assumes that all the I/O is done via the struct grub_diskfilter_segment.
24936 This is not true for the btrfs code. In order to reuse the native
24937 grub_raid6_recover() code, it is modified to not call
24938 grub_diskfilter_read_node() directly, but to call an handler passed
24939 as an argument.
24940
24941 Reviewed-by: Daniel Kiper <[email protected]>
24942
24943 2018-10-31 Goffredo Baroncelli <[email protected]>
24944
24945 btrfs: Add support for recovery for a RAID 5 btrfs profiles
24946 Add support for recovery for a RAID 5 btrfs profile. In addition
24947 it is added some code as preparatory work for RAID 6 recovery code.
24948
24949 Reviewed-by: Daniel Kiper <[email protected]>
24950
24951 2018-10-31 Goffredo Baroncelli <[email protected]>
24952
24953 btrfs: Refactor the code that read from disk
24954 Move the code in charge to read the data from disk into a separate
24955 function. This helps to separate the error handling logic (which
24956 depends on the different raid profiles) from the read from disk
24957 logic. Refactoring this code increases the general readability too.
24958
24959 This is a preparatory patch, to help the adding of the RAID 5/6 recovery code.
24960
24961 Reviewed-by: Daniel Kiper <[email protected]>
24962
24963 2018-10-31 Goffredo Baroncelli <[email protected]>
24964
24965 btrfs: Move logging code in grub_btrfs_read_logical()
24966 A portion of the logging code is moved outside of internal for(;;). The part
24967 that is left inside is the one which depends on the internal for(;;) index.
24968
24969 This is a preparatory patch. The next one will refactor the code inside
24970 the for(;;) into an another function.
24971
24972 Reviewed-by: Daniel Kiper <[email protected]>
24973
24974 2018-10-31 Goffredo Baroncelli <[email protected]>
24975
24976 btrfs: Avoid a rescan for a device which was already not found
24977 Currently read from missing device triggers rescan. However, it is never
24978 recorded that the device is missing. So, each read of a missing device
24979 triggers rescan again and again. This behavior causes a lot of unneeded
24980 rescans leading to huge slowdowns.
24981
24982 This patch fixes above mentioned issue. Information about missing devices
24983 is stored in the data->devices_attached[] array as NULL value in dev
24984 member. Rescan is triggered only if no information is found for a given
24985 device. This means that only first time read triggers rescan.
24986
24987 The patch drops premature return. This way data->devices_attached[] is
24988 filled even when a given device is missing.
24989
24990 Reviewed-by: Daniel Kiper <[email protected]>
24991
24992 2018-10-31 Goffredo Baroncelli <[email protected]>
24993
24994 btrfs: Move the error logging from find_device() to its caller
24995 The caller knows better if this error is fatal or not, i.e. another disk is
24996 available or not.
24997
24998 This is a preparatory patch.
24999
25000 Reviewed-by: Daniel Kiper <[email protected]>
25001
25002 2018-10-31 Goffredo Baroncelli <[email protected]>
25003
25004 btrfs: Add helper to check the btrfs header
25005 This helper is used in a few places to help the debugging. As
25006 conservative approach the error is only logged.
25007 This does not impact the error handling.
25008
25009 Reviewed-by: Daniel Kiper <[email protected]>
25010
25011 2018-10-31 Goffredo Baroncelli <[email protected]>
25012
25013 btrfs: Add support for reading a filesystem with a RAID 5 or RAID 6 profile
25014 Reviewed-by: Daniel Kiper <[email protected]>
25015
25016 2018-09-27 Michael Chang <[email protected]>
25017
25018 msdos: Fix overflow in converting partition start and length into 512B blocks
25019 When booting from NVME SSD with 4k sector size, it fails with the message.
25020
25021 error: attempt to read or write outside of partition.
25022
25023 This patch fixes the problem by fixing overflow in converting partition start
25024 and length into 512B blocks.
25025
25026 Reviewed-by: Daniel Kiper <[email protected]>
25027
25028 2018-09-27 Mihai Moldovan <[email protected]>
25029
25030 osdep/linux: Convert partition start to disk sector length
25031 When reading data off a disk, sector values are based on the disk sector
25032 length.
25033
25034 Within grub_util_fd_open_device(), the start of the partition was taken
25035 directly from grub's partition information structure, which uses the
25036 internal sector length (currently 512b), but never transformed to the
25037 disk's sector length.
25038
25039 Subsequent calculations were all wrong for devices that have a diverging
25040 sector length and the functions eventually skipped to the wrong stream
25041 location, reading invalid data.
25042
25043 Reviewed-by: Daniel Kiper <[email protected]>
25044
25045 2018-09-27 Adam Williamson <[email protected]>
25046
25047 python: Use AM_PATH_PYTHON to determine interpreter for gentpl.py
25048 gentpl.py is python2/3-agnostic, but there's no way to cause it
25049 to be run with any interpreter other than 'python', it's just
25050 hard-coded into Makefile.common that way. Adjust that to use
25051 AM_PATH_PYTHON (provided by automake) to find an interpreter
25052 and run gentpl.py with that instead. This makes grub buildable
25053 when `python` does not exist (but rather `python3` or `python2`
25054 or `python2.7`, etc.) Minimum version is set to 2.6 as this is
25055 the first version with `__future__.print_function` available.
25056
25057 Note, AM_PATH_PYTHON respects the PYTHON environment variable
25058 and will treat its value as the *only* candidate for a valid
25059 interpreter if it is set - when PYTHON is set, AM_PATH_PYTHON
25060 will not try to find any alternative interpreter, it will only
25061 check whether the interpreter set as the value of PYTHON meets
25062 the requirements and use it if so or fail if not. This means
25063 that when using grub's `autogen.sh`, as it too uses the value
25064 of the PYTHON environment variable (and if it is not set, just
25065 sets it to 'python') you cannot rely on AM_PATH_PYTHON
25066 interpreter discovery. If your desired Python interpreter is
25067 not just 'python', you must set the PYTHON environment variable,
25068 e.g. 'PYTHON=/usr/local/bin/python3 ./autogen.sh'. The specified
25069 interpreter will then be used both by autogen.sh itself and by
25070 the autotools-driven build scripts.
25071
25072 Reviewed-by: Daniel Kiper <[email protected]>
25073
25074 2018-09-27 Colin Watson <[email protected]>
25075
25076 build: Use pkg-config to find FreeType
25077 pkg-config is apparently preferred over freetype-config these days (see
25078 the BUGS section of freetype-config(1)). pkg-config support was added
25079 to FreeType in version 2.1.5, which was released in 2003, so it should
25080 comfortably be available everywhere by now.
25081
25082 We no longer need to explicitly substitute FREETYPE_CFLAGS and
25083 FREETYPE_LIBS, since PKG_CHECK_MODULES does that automatically.
25084
25085 Fixes Debian bug #887721.
25086
25087 Reported-by: Hugh McMaster <[email protected]>
25088 Reviewed-by: Daniel Kiper <[email protected]>
25089
25090 2018-09-27 Colin Watson <[email protected]>
25091
25092 build: Capitalise *freetype_* variables
25093 Using FREETYPE_CFLAGS and FREETYPE_LIBS is more in line with the naming
25094 scheme used by pkg-config macros.
25095
25096 Reviewed-by: Daniel Kiper <[email protected]>
25097
25098 2018-09-13 Julian Andres Klode <[email protected]>
25099
25100 ofnet: Initialize structs in bootpath parser
25101 Code later on checks if variables inside the struct are
25102 0 to see if they have been set, like if there were addresses
25103 in the bootpath.
25104
25105 The variables were not initialized however, so the check
25106 might succeed with uninitialized data, and a new interface
25107 with random addresses and the same name is added. This causes
25108 $net_default_mac to point to the random one, so, for example,
25109 using that variable to load per-mac config files fails.
25110
25111 Bug-Ubuntu: https://bugs.launchpad.net/bugs/1785859
25112
25113 Reviewed-by: Daniel Kiper <[email protected]>
25114
25115 2018-09-13 dann frazier <[email protected]>
25116
25117 grub-reboot: Warn when "for the next boot only" promise cannot be kept
25118 The "for the next boot only" property of grub-reboot is dependent upon
25119 GRUB being able to clear the next_entry variable in the environment
25120 block. However, GRUB cannot write to devices using the diskfilter
25121 and lvm abstractions.
25122
25123 Ref: https://lists.gnu.org/archive/html/grub-devel/2009-12/msg00276.html
25124 Ref: https://bugs.launchpad.net/bugs/788298
25125
25126 Reviewed-by: Daniel Kiper <[email protected]>
25127
25128 2018-09-13 Cao jin <[email protected]>
25129
25130 relocator16: Comments update
25131 Reviewed-by: Daniel Kiper <[email protected]>
25132
25133 2018-09-13 Paul Menzel <[email protected]>
25134
25135 ahci: Increase time-out from 10 s to 32 s
25136 This is a cryptographically signed message in MIME format.
25137
25138 Date: Thu, 9 Aug 2018 07:27:35 +0200
25139
25140 Currently, the GRUB payload for coreboot does not detect the Western
25141 Digital hard disk WDC WD20EARS-60M AB51 connected to the ASRock E350M1,
25142 as that takes over ten seconds to spin up.
25143
25144 ```
25145 disk/ahci.c:533: port 0, err: 0
25146 disk/ahci.c:539: port 0, err: 0
25147 disk/ahci.c:543: port 0, err: 0
25148 disk/ahci.c:549: port 0, offset: 120, tfd:80, CMD: 6016
25149 disk/ahci.c:552: port 0, err: 0
25150 disk/ahci.c:563: port 0, offset: 120, tfd:80, CMD: 6016
25151 disk/ahci.c:566: port: 0, err: 0
25152 disk/ahci.c:593: port 0 is busy
25153 disk/ahci.c:621: cleaning up failed devs
25154 ```
25155
25156 GRUB detects the drive, when either unloading the module *ahci*, and
25157 then loading it again, or when doing a warm reset.
25158
25159 As the ten second time-out is too short, increase it to 32 seconds,
25160 used by SeaBIOS. which detects the drive successfully.
25161
25162 The AHCI driver in libpayload uses 30 seconds, and that time-out was
25163 added in commit 354066e1 (libpayload: ahci: Increase timeout for
25164 signature reading) with the description below.
25165
25166 > We can't read the drives signature before it's ready, i.e. spun up.
25167 > So set the timeout to the standard 30s. Also put a notice on the
25168 > console, so the user knows why the signature reading failed.
25169
25170 Reviewed-by: Daniel Kiper <[email protected]>
25171
25172 2018-09-13 Cao jin <[email protected]>
25173
25174 linux16: Code cleanup
25175 1. move relocator related code more close to each other
25176 2. use variable "len" since it has correct assignment, and keep coding
25177 style with upper code
25178
25179 Reviewed-by: Daniel Kiper <[email protected]>
25180
25181 2018-09-13 Colin Watson <[email protected]>
25182
25183 tests: Fix qemu options for UHCI test
25184 qemu 2.12 removed the -usbdevice option. Use a more modern spelling
25185 instead, in line with other USB-related tests.
25186
25187 Reviewed-by: Daniel Kiper <[email protected]>
25188
25189 2018-09-13 Colin Watson <[email protected]>
25190
25191 tests: Disable sercon in SeaBIOS
25192 SeaBIOS 1.11.0 added support for VGA emulation over a serial port, which
25193 interferes with grub-shell. Turn it off.
25194
25195 Reviewed-by: Daniel Kiper <[email protected]>
25196
25197 2018-09-12 Peter Jones <[email protected]>
25198
25199 grub-module-verifier: Report the filename or modname in errors
25200 Make it so that when grub-module-verifier complains of an issue, it tells you
25201 which module the issue was with.
25202
25203 Reviewed-by: Daniel Kiper <[email protected]>
25204
25205 2018-09-12 Peter Jones <[email protected]>
25206
25207 configure: Fix an 8 year old typo
25208 Reviewed-by: Daniel Kiper <[email protected]>
25209
25210 2018-09-12 Leif Lindholm <[email protected]>
25211
25212 loader/multiboot_mbi2: Use central copy of grub_efi_find_mmap_size()
25213 Delete local copy of function to determine required buffer size for the
25214 UEFI memory map, use helper in kern/efi/mm.c.
25215
25216 Reviewed-by: Daniel Kiper <[email protected]>
25217
25218 2018-09-12 Leif Lindholm <[email protected]>
25219
25220 loader/ia64/linux: Use central copy of grub_efi_find_mmap_size()
25221 Delete local copy of function to determine required buffer size for the
25222 UEFI memory map, use helper in kern/efi/mm.c.
25223
25224 Reviewed-by: Daniel Kiper <[email protected]>
25225
25226 2018-09-12 Leif Lindholm <[email protected]>
25227
25228 loader/i386/linux: Use central copy of grub_efi_find_mmap_size()
25229 Delete local copy of function to determine required buffer size for the
25230 UEFI memory map, use helper in kern/efi/mm.c.
25231
25232 Reviewed-by: Daniel Kiper <[email protected]>
25233
25234 2018-07-25 Leif Lindholm <[email protected]>
25235
25236 i386: Don't include lib/i386/reset.c in EFI builds
25237 Commit 0ba90a7f0178 ("efi: Move grub_reboot() into kernel") broke
25238 the build on i386-efi - genmoddep.awk bails out with message
25239 grub_reboot in reboot is duplicated in kernel
25240 This is because both lib/i386/reset.c and kern/efi/efi.c now provide
25241 this function.
25242
25243 Rather than explicitly list each i386 platform variant in
25244 Makefile.core.def, include the contents of lib/i386/reset.c only when
25245 GRUB_MACHINE_EFI is not set.
25246
25247 Reviewed-by: Daniel Kiper <[email protected]>
25248
25249 2018-07-25 Leif Lindholm <[email protected]>
25250
25251 efi: Restrict arm/arm64 linux loader initrd placement
25252 The 32-bit arm Linux kernel is built as a zImage, which self-decompresses
25253 down to near start of RAM. In order for an initrd/initramfs to be
25254 accessible, it needs to be placed within the first ~768MB of RAM.
25255 The initrd loader built into the kernel EFI stub restricts this down to
25256 512MB for simplicity - so enable the same restriction in grub.
25257
25258 For arm64, the requirement is within a 1GB aligned 32GB window also
25259 covering the (runtime) kernel image. Since the EFI stub loader itself
25260 will attempt to relocate to near start of RAM, force initrd to be loaded
25261 completely within the first 32GB of RAM.
25262
25263 Reviewed-by: Daniel Kiper <[email protected]>
25264
25265 2018-07-25 Leif Lindholm <[email protected]>
25266
25267 arm: Delete unused efi support from loader/arm
25268 The 32-bit arm efi port now shares the 64-bit linux loader, so delete
25269 the now unused bits from the 32-bit linux loader.
25270
25271 This in turn leaves the grub-core/kern/arm/efi/misc.c unused, so
25272 delete that too.
25273
25274 Reviewed-by: Daniel Kiper <[email protected]>
25275
25276 2018-07-25 Leif Lindholm <[email protected]>
25277
25278 arm/efi: Switch to arm64 linux loader
25279 The arm64 and arm linux kernel EFI-stub support presents pretty much
25280 identical interfaces, so the same linux loader source can be used for
25281 both architectures.
25282
25283 Switch 32-bit ARM UEFI platforms over to the existing EFI-stub aware
25284 loader initially developed for arm64.
25285
25286 This *WILL* stop non-efistub Linux kernels from booting on arm-efi.
25287
25288 Reviewed-by: Daniel Kiper <[email protected]>
25289
25290 2018-07-25 Leif Lindholm <[email protected]>
25291
25292 arm64/linux/loader: Rename functions and macros and move to common headers
25293 In preparation for using the linux loader for 32-bit and 64-bit platforms,
25294 rename grub_arm64*/GRUB_ARM64* to grub_armxx*/GRUB_ARMXX*.
25295
25296 Move prototypes for now-common functions to efi/efi.h.
25297
25298 Reviewed-by: Daniel Kiper <[email protected]>
25299
25300 2018-07-25 Leif Lindholm <[email protected]>
25301
25302 efi: Add grub_efi_get_ram_base() function for arm64
25303 Since ARM platforms do not have a common memory map, add a helper
25304 function that finds the lowest address region with the EFI_MEMORY_WB
25305 attribute set in the UEFI memory map.
25306
25307 Required for the arm64 efi linux loader to restrict the initrd
25308 location to where it will be accessible by the kernel at runtime.
25309
25310 Reviewed-by: Daniel Kiper <[email protected]>
25311
25312 2018-07-25 Leif Lindholm <[email protected]>
25313
25314 efi: Add central copy of grub_efi_find_mmap_size
25315 There are several implementations of this function in the tree.
25316 Add a central version in grub-core/efi/mm.c.
25317
25318 Reviewed-by: Daniel Kiper <[email protected]>
25319
25320 2018-07-25 Arindam Nath <[email protected]>
25321
25322 i386/linux: Add support for ext_lfb_base
25323 The EFI Graphics Output Protocol can return a 64-bit
25324 linear frame buffer address in some firmware/BIOS
25325 implementations. We currently only store the lower
25326 32-bits in the lfb_base. This will eventually be
25327 passed to Linux kernel and the efifb driver will
25328 incorrectly interpret the framebuffer address as
25329 32-bit address.
25330
25331 The Linux kernel has already added support to handle
25332 64-bit linear framebuffer address in the efifb driver
25333 since quite some time now.
25334
25335 This patch adds the support for 64-bit linear frame
25336 buffer address in GRUB to address the above mentioned
25337 scenario.
25338
25339 Reviewed-by: Daniel Kiper <[email protected]>
25340
25341 2018-07-11 Leif Lindholm <[email protected]>
25342
25343 commands/file: Use definitions from arm64/linux.h
25344 Clean up code for matching IS_ARM64 slightly by making use of struct
25345 linux_arm64_kernel_header and GRUB_LINUX_ARM64_MAGIC_SIGNATURE.
25346
25347 Reviewed-by: Daniel Kiper <[email protected]>
25348
25349 2018-07-11 Leif Lindholm <[email protected]>
25350
25351 commands/file: Use definitions from arm/linux.h
25352 Clean up code for matching IS_ARM slightly by making use of struct
25353 linux_arm_kernel_header and GRUB_LINUX_ARM_MAGIC_SIGNATURE.
25354
25355 Reviewed-by: Daniel Kiper <[email protected]>
25356
25357 2018-07-11 Hans de Goede <[email protected]>
25358
25359 efi/console: Fix the "enter" key not working on x86 tablets
25360 Most 8" or 7" x86 Windows 10 tablets come with volume up/down buttons and
25361 a power-button. In their UEFI these are almost always mapped to arrow
25362 up/down and enter.
25363
25364 Pressing the volume buttons (sometimes by accident) will stop the
25365 menu countdown, but the power-button / "enter" key was not being recognized
25366 as enter, so the user would be stuck at the grub menu.
25367
25368 The problem is that these tablets send scan_code 13 or 0x0d for the
25369 power-button, which officialy maps to the F3 key. They also set
25370 unicode_char to 0x0d.
25371
25372 This commit recognizes the special case of both scan_code and unicode_char
25373 being set to 0x0d and treats this as an enter key press.
25374
25375 This fixes things getting stuck at the grub-menu and allows the user
25376 to choice a grub-menu entry using the buttons on the tablet.
25377
25378 Reviewed-by: Daniel Kiper <[email protected]>
25379
25380 2018-07-11 Cao jin <[email protected]>
25381
25382 grub-setup: Debug message cleanup
25383 Variable "root" is initialized after root device probing and is null in
25384 current place, so, drop it.
25385
25386 Reviewed-by: Daniel Kiper <[email protected]>
25387
25388 2018-07-02 Denis 'GNUtoo' Carikli <[email protected]>
25389
25390 multiboot_elfxx.c: Fix compilation by fixing undeclared variable
25391 Without that fix we have:
25392 In file included from ../../include/grub/command.h:25:0,
25393 from ../../grub-core/loader/multiboot.c:30:
25394 ../../grub-core/loader/multiboot_elfxx.c: In function 'grub_multiboot_load_elf64':
25395 ../../grub-core/loader/multiboot_elfxx.c:130:28: error: 'relocatable' undeclared (first use in this function)
25396 "load_base_addr=0x%x\n", relocatable,
25397
25398 This happens due to mistake in the commit 14ec665
25399 (mbi: Use per segment a separate relocator chunk).
25400
25401 So, let's fix it.
25402
25403 2018-06-23 Leif Lindholm <[email protected]>
25404
25405 efi/fdt: Set address/size cells to 2 for empty tree
25406 When booting an arm* system on UEFI with an empty device tree (currently
25407 only when hardware description comes from ACPI), we don't currently set
25408 default to 1 cell (32 bits).
25409
25410 Set both of these properties, to 2 cells (64 bits), to resolve issues
25411 with kexec on some platforms.
25412
25413 This change corresponds with linux kernel commit ae8a442dfdc4
25414 ("efi/libstub/arm*: Set default address and size cells values for an empty dtb")
25415 and ensures booting through grub does not behave differently from booting
25416 the stub loader directly.
25417
25418 See also https://patchwork.kernel.org/patch/9561201/
25419
25420 Reviewed-by: Daniel Kiper <[email protected]>
25421
25422 2018-06-23 Leif Lindholm <[email protected]>
25423
25424 fdt: Move prop_entry_size to fdt.h
25425 To be able to resuse the prop_entry_size macro, move it to
25426 <grub/fdt.h> and rename it grub_fdt_prop_entry_size.
25427
25428 Reviewed-by: Daniel Kiper <[email protected]>
25429
25430 2018-06-23 Will Thompson <[email protected]>
25431
25432 grub-fs-tester: Fix losetup race
25433 If something else on the system is using loopback devices, then the
25434 device that's free at the call to `losetup -f` may not be free in the
25435 following call to try to use it. Instead, find and use the first free
25436 loopback device in a single call to losetup.
25437
25438 Reviewed-by: Daniel Kiper <[email protected]>
25439
25440 2018-06-23 Alexander Boettcher <[email protected]>
25441
25442 mbi: Use per segment a separate relocator chunk
25443 Instead of setting up a all comprising relocator chunk for all segments,
25444 use per segment a separate relocator chunk.
25445
25446 Currently, if the ELF is non-relocatable, a single relocator chunk will
25447 comprise memory (between the segments) which gets overridden by the relst()
25448 invocation of the movers code in grub_relocator16/32/64_boot().
25449
25450 The overridden memory may contain reserved ranges like VGA memory or ACPI
25451 tables, which may lead to crashes or at least to strange boot behaviour.
25452
25453 Reviewed-by: Daniel Kiper <[email protected]>
25454
25455 2018-06-05 Daniel Kiper <[email protected]>
25456
25457 templates: Add missing "]"
25458 Commit 51be337 (templates: Update grub script template files)
25459 lacked one "]", so, add it.
25460
25461 Reported-by: Philip <[email protected]>
25462
25463 2018-05-29 Daniel Kiper <[email protected]>
25464
25465 xfs: Accept filesystem with sparse inodes
25466 The sparse inode metadata format became a mkfs.xfs default in
25467 xfsprogs-4.16.0, and such filesystems are now rejected by grub as
25468 containing an incompatible feature.
25469
25470 In essence, this feature allows xfs to allocate inodes into fragmented
25471 freespace. (Without this feature, if xfs could not allocate contiguous
25472 space for 64 new inodes, inode creation would fail.)
25473
25474 In practice, the disk format change is restricted to the inode btree,
25475 which as far as I can tell is not used by grub. If all you're doing
25476 today is parsing a directory, reading an inode number, and converting
25477 that inode number to a disk location, then ignoring this feature
25478 should be fine, so I've added it to XFS_SB_FEAT_INCOMPAT_SUPPORTED
25479
25480 I did some brief testing of this patch by hacking up the regression
25481 tests to completely fragment freespace on the test xfs filesystem, and
25482 then write a large-ish number of inodes to consume any existing
25483 contiguous 64-inode chunk. This way any files the grub tests add and
25484 traverse would be in such a fragmented inode allocation. Tests passed,
25485 but I'm not sure how to cleanly integrate that into the test harness.
25486
25487 Reviewed-by: Daniel Kiper <[email protected]>
25488 Tested-by: Chris Murphy <[email protected]>
25489
25490 2018-05-29 Oleg Solovyov <[email protected]>
25491
25492 grub-probe: Don't skip /dev/mapper/dm-* devices
25493 This patch ensures that grub-probe will find the root device placed in
25494 /dev/mapper/dm-[0-9]+-.* e.g. device named /dev/mapper/dm-0-luks will be
25495 found and grub.cfg will be updated properly, enabling the system to boot.
25496
25497 Reviewed-by: Daniel Kiper <[email protected]>
25498
25499 2018-05-08 Michael Chang <[email protected]>
25500
25501 bufio: Round up block size to power of 2
25502 Rounding up the bufio->block_size to meet power of 2 to facilitate next_buf
25503 calculation in grub_bufio_read().
25504
25505 Reviewed-by: Daniel Kiper <[email protected]>
25506
25507 2018-04-23 Nicholas Vinson <[email protected]>
25508
25509 templates: Update grub script template files
25510 Update grub-mkconfig.in and 10_linux.in to support grub-probe's new
25511 partuuid target. Update grub.texi documentation. The following table
25512 shows how GRUB_DISABLE_LINUX_UUID, GRUB_DISABLE_LINUX_PARTUUID, and
25513 initramfs detection interact:
25514
25515 Initramfs GRUB_DISABLE_LINUX_PARTUUID GRUB_DISABLE_LINUX_UUID Linux Root
25516 detected Set Set ID Method
25517
25518 false false false part UUID
25519 false false true part UUID
25520 false true false dev name
25521 false true true dev name
25522 true false false fs UUID
25523 true false true part UUID
25524 true true false fs UUID
25525 true true true dev name
25526
25527 Note: GRUB_DISABLE_LINUX_PARTUUID and GRUB_DISABLE_LINUX_UUID equate to
25528 'false' when unset or set to any value other than 'true'.
25529 GRUB_DISABLE_LINUX_PARTUUID defaults to 'true'.
25530
25531 Reviewed-by: Daniel Kiper <[email protected]>
25532
25533 2018-04-23 Nicholas Vinson <[email protected]>
25534
25535 grub-probe: Add PARTUUID detection support
25536 Add PARTUUID detection support grub-probe for MBR and GPT partition schemes.
25537
25538 Reviewed-by: Daniel Kiper <[email protected]>
25539
25540 2018-04-23 Nicholas Vinson <[email protected]>
25541
25542 disk: Update grub_gpt_partentry
25543 Rename grub_gpt_part_type to grub_gpt_part_guid and update grub_gpt_partentry
25544 to use this type for both the partition type GUID string and the partition GUID
25545 string entries. This change ensures that the two GUID fields are handled more
25546 consistently and helps to simplify the changes needed to add Linux partition
25547 GUID support.
25548
25549 Reviewed-by: Daniel Kiper <[email protected]>
25550
25551 2018-04-23 Nicholas Vinson <[email protected]>
25552
25553 grub-probe: Centralize GUID prints
25554 Define print_gpt_guid(), so there is a central function for printing
25555 GUID strings. This change is a precursor for later patches which rely
25556 on this logic.
25557
25558 Reviewed-by: Daniel Kiper <[email protected]>
25559
25560 2018-04-23 Olaf Hering <[email protected]>
25561
25562 grub-install: Locale depends on nls
25563 With --disable-nls no locales exist.
25564
25565 Avoid runtime error by moving code that copies locales into its own
25566 function. Return early in case nls was disabled. That way the compiler
25567 will throw away unreachable code, no need to put preprocessor
25568 conditionals everywhere to avoid warnings about unused code.
25569
25570 Fix memleak by freeing srcf and dstf.
25571 Convert tabs to spaces in moved code.
25572
25573 Reviewed-by: Daniel Kiper <[email protected]>
25574
25575 2018-04-23 Cao jin <[email protected]>
25576
25577 diskboot: Trivial correction on stale comments
25578 diskboot.img now is loaded at 0x8000 and is jumped to with 0:0x8000.
25579
25580 Reviewed-by: Daniel Kiper <[email protected]>
25581
25582 2018-04-10 Jaegeuk Kim <[email protected]>
25583
25584 fs: Add F2FS support
25585 "F2FS (Flash-Friendly File System) is flash-friendly file system which was merged
25586 into Linux kernel v3.8 in 2013.
25587
25588 The motive for F2FS was to build a file system that from the start, takes into
25589 account the characteristics of NAND flash memory-based storage devices (such as
25590 solid-state disks, eMMC, and SD cards).
25591
25592 F2FS was designed on a basis of a log-structured file system approach, which
25593 remedies some known issues of the older log structured file systems, such as
25594 the snowball effect of wandering trees and high cleaning overhead. In addition,
25595 since a NAND-based storage device shows different characteristics according to
25596 its internal geometry or flash memory management scheme (such as the Flash
25597 Translation Layer or FTL), it supports various parameters not only for
25598 configuring on-disk layout, but also for selecting allocation and cleaning
25599 algorithm.", quote by https://en.wikipedia.org/wiki/F2FS.
25600
25601 The source codes for F2FS are available from:
25602
25603 http://git.kernel.org/cgit/linux/kernel/git/jaegeuk/f2fs.git
25604 http://git.kernel.org/cgit/linux/kernel/git/jaegeuk/f2fs-tools.git
25605
25606 This patch has been integrated in OpenMandriva Lx 3.
25607 https://www.openmandriva.org/
25608
25609 Reviewed-by: Daniel Kiper <[email protected]>
25610
25611 2018-04-04 Michael Chang <[email protected]>
25612
25613 Fix packed-not-aligned error on GCC 8
25614 When building with GCC 8, there are several errors regarding packed-not-aligned.
25615
25616 ./include/grub/gpt_partition.h:79:1: error: alignment 1 of ‘struct grub_gpt_partentry’ is less than 8 [-Werror=packed-not-aligned]
25617
25618 This patch fixes the build error by cleaning up the ambiguity of placing
25619 aligned structure in a packed one. In "struct grub_btrfs_time" and "struct
25620 grub_gpt_part_type", the aligned attribute seems to be superfluous, and also
25621 has to be packed, to ensure the structure is bit-to-bit mapped to the format
25622 laid on disk. I think we could blame to copy and paste error here for the
25623 mistake. In "struct efi_variable", we have to use grub_efi_packed_guid_t, as
25624 the name suggests. :)
25625
25626 Tested-by: Michael Chang <[email protected]>
25627 Tested-by: Paul Menzel <[email protected]>
25628 Reviewed-by: Daniel Kiper <[email protected]>
25629
25630 2018-04-04 [email protected] <[email protected]>
25631
25632 efi/uga: Fix PCIe LER when GRUB2 accesses non-enabled MMIO data from VGA
25633 A GPU inserted into a PCIe I/O slot disappears during system startup.
25634 The problem centers around GRUB and a specific VGA init function in
25635 efi_uga.c. This causes an LER (Link Error Recorvery) because the MMIO
25636 memory has not been enabled before attempting access.
25637
25638 The fix is to add the same coding used in other VGA drivers, specifically
25639 to add a check to insure that it is indeed a VGA controller. And then
25640 enable the MMIO address space with the specific bits.
25641
25642 Reviewed-by: Daniel Kiper <[email protected]>
25643
25644 2018-03-26 Eric Snowberg <[email protected]>
25645
25646 ieee1275: NULL pointer dereference in grub_machine_get_bootlocation()
25647 Read from NULL pointer canon in function grub_machine_get_bootlocation().
25648 Function grub_ieee1275_canonicalise_devname() may return NULL.
25649
25650 Reviewed-by: Daniel Kiper <[email protected]>
25651
25652 2018-03-14 Eric Snowberg <[email protected]>
25653
25654 ieee1275: split up grub_machine_get_bootlocation
25655 Split up some of the functionality in grub_machine_get_bootlocation into
25656 grub_ieee1275_get_boot_dev. This will allow for code reuse in a follow on
25657 patch.
25658
25659 Reviewed-by: Daniel Kiper <[email protected]>
25660
25661 2018-03-14 C. Masloch <[email protected]>
25662
25663 chainloader: patch in BPB's sectors_per_track and num_heads
25664 These fields must reflect the ROM-BIOS's geometry for CHS-based
25665 loaders to correctly load their next stage. Most loaders do not
25666 query the ROM-BIOS (Int13.08), relying on the BPB fields to hold
25667 the correct values already.
25668
25669 Tested with lDebug booted in qemu via grub2's
25670 FreeDOS direct loading support, refer to
25671 https://bitbucket.org/ecm/ldosboot + https://bitbucket.org/ecm/ldebug
25672 (For this test, lDebug's iniload.asm must be assembled with
25673 -D_QUERY_GEOMETRY=0 to leave the BPB values provided by grub.)
25674
25675 Reviewed-by: Daniel Kiper <[email protected]>
25676
25677 2018-03-14 Matthew S. Turnbull <[email protected]>
25678
25679 grub-mkconfig/10_linux: Support multiple early initrd images
25680 Add support for multiple, shared, early initrd images. These early
25681 images will be loaded in the order declared, and all will be loaded
25682 before the initrd image.
25683
25684 While many classes of data can be provided by early images, the
25685 immediate use case would be for distributions to provide CPU
25686 microcode to mitigate the Meltdown and Spectre vulnerabilities.
25687
25688 There are two environment variables provided for declaring the early
25689 images.
25690
25691 * GRUB_EARLY_INITRD_LINUX_STOCK is for the distribution declare
25692 images that are provided by the distribution or installed packages.
25693 If undeclared, this will default to a set of common microcode image
25694 names.
25695
25696 * GRUB_EARLY_INITRD_LINUX_CUSTOM is for user created images. User
25697 images will be loaded after the stock images.
25698
25699 These separate configurations allow the distribution and user to
25700 declare different image sets without clobbering each other.
25701
25702 This also makes a minor update to ensure that UUID partition labels
25703 stay disabled when no initrd image is found, even if early images are
25704 present.
25705
25706 This is a continuation of a previous patch published by Christian
25707 Hesse in 2016:
25708 http://lists.gnu.org/archive/html/grub-devel/2016-02/msg00025.html
25709
25710 Down stream Gentoo bug:
25711 https://bugs.gentoo.org/645088
25712
25713 Reviewed-by: Daniel Kiper <[email protected]>
25714
25715 2018-03-07 Eric Snowberg <[email protected]>
25716
25717 mkimage: fix build regression in grub_mkimage_load_image
25718 The grub_mkimage_load_image function (commit 7542af6, mkimage: refactor a bunch
25719 of section data into a struct.) introduces a build regression on SPARC:
25720
25721 cc1: warnings being treated as errors
25722 In file included from util/grub-mkimage32.c:23:
25723 util/grub-mkimagexx.c: In function 'grub_mkimage_load_image32':
25724 util/grub-mkimagexx.c:1968: error: missing initializer
25725 util/grub-mkimagexx.c:1968: error: (near initialization for 'smd.sections')
25726 make[2]: *** [util/grub_mkimage-grub-mkimage32.o] Error 1
25727
25728 Initialize the entire section_metadata structure.
25729
25730 Reviewed-by: Daniel Kiper <[email protected]>
25731
25732 2018-03-05 dann frazier <[email protected]>
25733
25734 Revert "Keep the native terminal active when enabling gfxterm"
25735 This can cause an issue where GRUB is trying to display both a text and
25736 graphical menu on the display at the same time, resulting in a flickering
25737 effect when e.g. scrolling quickly through a menu (LP: #1752767).
25738
25739 Revert for now while we look for a better solution for the original issue.
25740
25741 This reverts commit 52ef7b23f528ce844716661d586497a177e80d5b.
25742
25743 Reviewed-by: Daniel Kiper <[email protected]>
25744
25745 2018-03-05 Eric Snowberg <[email protected]>
25746
25747 sparc64: #blocks64 disk node method
25748 Return the 64bit number of blocks of storage associated with the device or
25749 instance. Where a "block" is a unit of storage consisting of the number of
25750 bytes returned by the package's "block-size" method. If the size cannot be
25751 determined, or if the number of blocks exceeds the range return -1.
25752
25753 Reviewed-by: Daniel Kiper <[email protected]>
25754
25755 2018-03-05 Eric Snowberg <[email protected]>
25756
25757 sparc64: #blocks disk node method
25758 Return the number of blocks of storage associated with the device or
25759 instance. Where a "block" is a unit of storage consisting of the number
25760 of bytes returned by the package's "block-size" method. If the size cannot
25761 be determined, the #blocks method returns the maximum unsigned integer
25762 (which, because of Open Firmware's assumption of two's complement arithmetic,
25763 is equivalent to the signed number -1). If the number of blocks exceeds
25764 the range of an unsigned number, return 0 to alert the caller to try
25765 the #blocks64 command.
25766
25767 Reviewed-by: Daniel Kiper <[email protected]>
25768
25769 2018-03-05 Eric Snowberg <[email protected]>
25770
25771 ieee1275: block-size deblocker support method
25772 IEEE Std 1275-1994 Standard for Boot (Initialization Configuration)
25773 Firmware: Core Requirements and Practices
25774
25775 3.8.3 deblocker support package
25776
25777 Any package that uses the "deblocker" support package must define
25778 the following method, which the deblocker uses as a low-level
25779 interface to the device
25780
25781 block-size ( -- block-len ) Return "granularity" for accesses to this
25782 device.
25783
25784 Reviewed-by: Daniel Kiper <[email protected]>
25785
25786 2018-03-05 Daniel Kiper <[email protected]>
25787
25788 ieee1275: no-data-command bus specific method
25789 IEEE 1275-1994 Standard for Boot (Initialization Configuration)
25790 Firmware: Core Requirements and Practices
25791
25792 E.3.2.2 Bus-specific methods for bus nodes
25793
25794 A package implementing the scsi-2 device type shall implement the
25795 following bus-specific method:
25796
25797 no-data-command ( cmd-addr -- error? )
25798 Executes a simple SCSI command, automatically retrying under
25799 certain conditions. cmd-addr is the address of a 6-byte command buffer
25800 containing an SCSI command that does not have a data transfer phase.
25801 Executes the command, retrying indefinitely with the same retry criteria
25802 as retry-command.
25803
25804 error? is nonzero if an error occurred, zero otherwise.
25805 NOTE no-data-command is a convenience function. It provides
25806 no capabilities that are not present in retry-command, but for
25807 those commands that meet its restrictions, it is easier to use.
25808
25809 Reviewed-by: Daniel Kiper <[email protected]>
25810
25811 2018-03-05 Eric Snowberg <[email protected]>
25812
25813 ieee1275: set-address bus specific method
25814 IEEE 1275-1994 Standard for Boot (Initialization Configuration)
25815 Firmware: Core Requirements and Practices
25816 E.3.2.2 Bus-specific methods for bus nodes
25817
25818 A package implementing the scsi-2 device type shall implement the
25819 following bus-specific method:
25820
25821 set-address ( unit# target# -- )
25822 Sets the SCSI target number (0x0..0xf) and unit number (0..7) to which
25823 subsequent commands apply.
25824
25825 This function is for devices with #address-cells == 2
25826
25827 Reviewed-by: Daniel Kiper <[email protected]>
25828
25829 2018-03-05 Eric Snowberg <[email protected]>
25830
25831 ieee1275: encode-unit command for 4 addr cell devs
25832 Convert physical address to text unit-string.
25833
25834 Convert phys.lo ... phys-high, the numerical representation, to unit-string,
25835 the text string representation of a physical address within the address
25836 space defined by this device node. The number of cells in the list
25837 phys.lo ... phys.hi is determined by the value of the #address-cells property
25838 of this node.
25839
25840 This function is for devices with #address-cells == 4
25841
25842 Reviewed-by: Daniel Kiper <[email protected]>
25843
25844 2018-03-05 Eric Snowberg <[email protected]>
25845
25846 ieee1275: decode-unit command for 4 addr cell devs
25847 decode-unit ( addr len -- phys.lo ... phys.hi )
25848
25849 Convert text unit-string to physical address.
25850
25851 Convert unit-string, the text string representation, to phys.lo ... phys.hi,
25852 the numerical representation of a physical address within the address space
25853 defined by this device node. The number of cells in the list
25854 phys.lo ... phys.hi is determined by the value of the #address-cells
25855 property of this node.
25856
25857 This function is for devices with #address-cells == 4
25858
25859 Reviewed-by: Daniel Kiper <[email protected]>
25860
25861 2018-03-05 Eric Snowberg <[email protected]>
25862
25863 sparc64: Limit nvme of_path_of_nvme to just SPARC
25864 Limit NVMe of_path_of_nvme to just SPARC hardware for now. It has been
25865 found that non-Open Firmware hardware platforms can some how access
25866 this function.
25867
25868 Reviewed-by: Daniel Kiper <[email protected]>
25869
25870 2018-03-05 John Paul Adrian Glaubitz <[email protected]>
25871
25872 ieee1275: Fix crash in of_path_of_nvme when of_path is empty
25873 The of_path_of_nvme function (commit 2391d57, ieee1275: add nvme
25874 support within ofpath) introduced a functional regression:
25875
25876 On systems which are not based on Open Firmware but have at
25877 least one NVME device, find_obppath will return NULL and thus
25878 trying to append the disk name to of_path will result in a
25879 crash.
25880
25881 The proper behavior of of_path_of_nvme is, however, to just
25882 return NULL in such cases, like other users of find_obppath,
25883 such as of_path_of_scsi.
25884
25885 Reviewed-by: Eric Snowberg <[email protected]>
25886 Reviewed-by: Daniel Kiper <[email protected]>
25887
25888 2018-03-05 Peter Jones <[email protected]>
25889
25890 .mod files: Strip annobin annotations and .eh_frame, and their relocations
25891 This way debuginfo built from the .module will still include this
25892 information, but the final result won't have the data we don't actually
25893 need in the modules, either on-disk, loaded at runtime, or in prebuilt
25894 images.
25895
25896 Reviewed-by: Daniel Kiper <[email protected]>
25897
25898 2018-03-05 Peter Jones <[email protected]>
25899
25900 mkimage: avoid copying relocations for sections that won't be copied.
25901 Some versions of gcc include a plugin called "annobin", and in some
25902 build systems this is enabled by default. This plugin creates special
25903 ELF note sections to track which ABI-breaking features are used by a
25904 binary, as well as a series of relocations to annotate where.
25905
25906 If grub is compiled with this feature, then when grub-mkimage translates
25907 the binary to another file format which does not strongly associate
25908 relocation data with sections (i.e. when platform is *-efi), these
25909 relocations appear to be against the .text section rather than the
25910 original note section. When the binary is loaded by the PE runtime
25911 loader, hilarity ensues.
25912
25913 This issue is not necessarily limited to the annobin, but could arise
25914 any time there are relocations in sections that are not represented in
25915 grub-mkimage's output.
25916
25917 This patch seeks to avoid this issue by only including relocations that
25918 refer to sections which will be included in the final binary.
25919
25920 As an aside, this should also obviate the need to avoid -funwind-tables,
25921 -fasynchronous-unwind-tables, and any sections similar to .eh_frame in
25922 the future. I've tested it on x86-64-efi with the following gcc command
25923 line options (as recorded by -grecord-gcc-flags), but I still need to
25924 test the result on some other platforms that have been problematic in
25925 the past (especially ARM Aarch64) before I feel comfortable making
25926 changes to the configure.ac bits:
25927
25928 GNU C11 7.2.1 20180116 (Red Hat 7.2.1-7) -mno-mmx -mno-sse -mno-sse2 -mno-sse3 -mno-3dnow -msoft-float -mno-stack-arg-probe -mcmodel=large -mno-red-zone -m64 -mtune=generic -march=x86-64 -g3 -Os -freg-struct-return -fno-stack-protector -ffreestanding -funwind-tables -fasynchronous-unwind-tables -fno-strict-aliasing -fstack-clash-protection -fno-ident -fplugin=annobin
25929
25930 Reviewed-by: Daniel Kiper <[email protected]>
25931
25932 2018-03-05 Peter Jones <[email protected]>
25933
25934 mkimage: refactor a bunch of section data into a struct.
25935 This basically moves a bunch of the section information we pass around a
25936 lot into a struct, and passes a pointer to a single one of those
25937 instead.
25938
25939 This shouldn't change the binary file output or the "grub-mkimage -v"
25940 output in any way.
25941
25942 Reviewed-by: Daniel Kiper <[email protected]>
25943
25944 2018-03-05 Peter Jones <[email protected]>
25945
25946 mkimage: make locate_sections() set up vaddresses as well.
25947 This puts both kinds of address initialization at the same place, and also lets
25948 us iterate through the section list one time fewer.
25949
25950 Reviewed-by: Daniel Kiper <[email protected]>
25951
25952 2018-03-05 Peter Jones <[email protected]>
25953
25954 mkimage: rename a couple of things to be less confusing later.
25955 This renames some things:
25956
25957 - the "strtab" and "strtab_section" in relocate_symbols are changed to "symtab"
25958 instead, so as to be less confusing when "strtab" is moved to a struct in a
25959 later patch.
25960
25961 - The places where we pass section_vaddresses to functions are changed to also
25962 be called section_vaddresses"inside those functions, so I get less confused
25963 when I put addresses and vaddresses in a struct in a later patch.
25964
25965 Reviewed-by: Daniel Kiper <[email protected]>
25966
25967 2018-03-05 Peter Jones <[email protected]>
25968
25969 mkimage: make it easier to run syntax checkers on grub-mkimagexx.c
25970 This makes it so you can treat grub-mkimagexx.c as a file you can build
25971 directly, so syntax checkers like vim's "syntastic" plugin, which uses
25972 "gcc -x c -fsyntax-only" to build it, will work.
25973
25974 One still has to do whatever setup is required to make it pick the right
25975 include dirs, which -W options we use, etc., but this makes it so you
25976 can do the checking on the file you're editing, rather than on a
25977 different file.
25978
25979 Reviewed-by: Daniel Kiper <[email protected]>
25980
25981 2018-03-05 Peter Jones <[email protected]>
25982
25983 aout.h: Fix missing include.
25984 grub_aout_load() has a grub_file_t parameter, and depending on what order
25985 includes land in, it's sometimes not defined. This patch explicitly adds
25986 file.h to aout.h so that it will always be defined.
25987
25988 Reviewed-by: Daniel Kiper <[email protected]>
25989
25990 2018-02-26 Joakim Bech <[email protected]>
25991
25992 ieee1275: fix build regression in of_path_of_nvme
25993 The of_path_of_nvme function (commit 2391d57, ieee1275: add nvme
25994 support within ofpath) introduced a build regression:
25995 grub-core/osdep/linux/ofpath.c:365:21: error: comparison between pointer
25996 and zero character constant [-Werror=pointer-compare]
25997 if ((digit_string != '\0') && (*part_end == 'p'))
25998
25999 Update digit_string to compare against the char instead of the pointer.
26000
26001 Reviewed-by: Daniel Kiper <[email protected]>
26002
26003 2018-02-23 Leif Lindholm <[email protected]>
26004
26005 arm: make linux.h safe to include for non-native builds
26006 <grub/machine/loader.h> (for machine arm/efi) and
26007 <grub/machine/kernel.h> (for machine arm/coreboot) will not always
26008 resolve (and will likely not be valid to) if pulled in when building
26009 non-native commands, such as host tools or the "file" command.
26010 So explicitly include them with their expanded pathnames.
26011
26012 Reviewed-by: Daniel Kiper <[email protected]>
26013
26014 2018-02-23 Leif Lindholm <[email protected]>
26015
26016 arm: switch linux loader to linux_arm_kernel_header struct
26017 Use kernel header struct and magic definition to align (and coexist) with
26018 i386/arm64 ports.
26019
26020 Reviewed-by: Daniel Kiper <[email protected]>
26021
26022 2018-02-23 Leif Lindholm <[email protected]>
26023
26024 arm64: align linux kernel magic macro naming with i386
26025 Change GRUB_ARM64_LINUX_MAGIC to GRUB_LINUX_ARM64_MAGIC_SIGNATURE.
26026
26027 Reviewed-by: Daniel Kiper <[email protected]>
26028
26029 2018-02-23 Leif Lindholm <[email protected]>
26030
26031 arm64: align linux kernel header struct naming with i386
26032 Rename struct grub_arm64_linux_kernel_header -> linux_arm64_kernel_header.
26033
26034 Reviewed-by: Daniel Kiper <[email protected]>
26035
26036 2018-02-23 Leif Lindholm <[email protected]>
26037
26038 i386: make struct linux_kernel_header architecture specific
26039 struct linux_kernel_header -> struct linux_i386_kernel_header
26040
26041 Reviewed-by: Daniel Kiper <[email protected]>
26042
26043 2018-02-23 Leif Lindholm <[email protected]>
26044
26045 make GRUB_LINUX_MAGIC_SIGNATURE architecture-specific
26046 Rename GRUB_LINUX_MAGIC_SIGNATURE GRUB_LINUX_I386_MAGIC_SIGNATURE,
26047 to be usable in code that supports more than one image type.
26048
26049 Reviewed-by: Daniel Kiper <[email protected]>
26050
26051 2018-02-23 Leif Lindholm <[email protected]>
26052
26053 Make arch-specific linux.h include guards architecture unique
26054 Replace uses of GRUB_LINUX_MACHINE_HEADER and GRUB_LINUX_CPU_HEADER
26055 with GRUB_<arch>_LINUX_HEADER include guards to prevent issues when
26056 including more than one of them.
26057
26058 Reviewed-by: Daniel Kiper <[email protected]>
26059
26060 2018-02-23 Leif Lindholm <[email protected]>
26061
26062 arm64/efi: move EFI_PAGE definitions to efi/memory.h
26063 The EFI page definitions and macros are generic and should not be confined
26064 to arm64 headers - so move to efi/memory.h.
26065 Also add EFI_PAGE_SIZE macro.
26066
26067 Update loader sources to reflect new header location.
26068
26069 Reviewed-by: Daniel Kiper <[email protected]>
26070
26071 2018-02-23 Colin Watson <[email protected]>
26072
26073 libgcrypt: Import replacement CRC operations
26074 The CRC implementation imported from libgcrypt 1.5.3 is arguably
26075 non-free, due to being encumbered by the restrictive Internet Society
26076 licence on RFCs (see e.g. https://wiki.debian.org/NonFreeIETFDocuments).
26077 Fortunately, libgcrypt has since replaced it with a version that is both
26078 reportedly better-optimised and doesn't suffer from this encumbrance.
26079
26080 The ideal solution would be to update to a new version of libgcrypt, and
26081 I spent some time trying to do that. However, util/import_gcry.py
26082 requires complex modifications to cope with the new version, and I
26083 stalled part-way through; furthermore, GRUB's libgcrypt tree already
26084 contains some backports of upstream changes. Rather than allowing the
26085 perfect to be the enemy of the good, I think it's best to backport this
26086 single change to at least sort out the licensing situation. Doing so
26087 won't make things any harder for a future wholesale upgrade.
26088
26089 This commit is mostly a straightforward backport of
26090 https://git.gnupg.org/cgi-bin/gitweb.cgi?p=libgcrypt.git;a=commitdiff;h=06e122baa3321483a47bbf82fd2a4540becfa0c9,
26091 but I also imported bufhelp.h from libgcrypt 1.7.0 (newer versions
26092 required further changes elsewhere).
26093
26094 I've tested that "hashsum -h crc32" still produces correct output for a
26095 variety of files on both i386-pc and x86_64-emu targets.
26096
26097 Reviewed-by: Daniel Kiper <[email protected]>
26098
26099 2018-02-23 Eric Snowberg <[email protected]>
26100
26101 ieee1275: add nvme support within ofpath
26102 Add NVMe support within ofpath.
26103
26104 The Open Firmware text representation for a NVMe device contains the
26105 Namespace ID. An invalid namespace ID is one whose value is zero or whose
26106 value is greater than the value reported by the Number of Namespaces (NN)
26107 field in the Identify Controller data structure. At the moment only a
26108 single Namespace is supported, therefore the value is currently hard coded
26109 to one.
26110
26111 Reviewed-by: Daniel Kiper <[email protected]>
26112
26113 2018-02-23 Daniel Kiper <[email protected]>
26114
26115 chainloader: Fix wrong break condition (must be AND not, OR)
26116 The definition of bpb's num_total_sectors_16 and num_total_sectors_32
26117 is that either the 16-bit field is non-zero and is used (in which case
26118 eg mkfs.fat sets the 32-bit field to zero), or it is zero and the
26119 32-bit field is used. Therefore, a BPB is invalid only if *both*
26120 fields are zero; having one field as zero and the other as non-zero is
26121 the case to be expected. (Indeed, according to Microsoft's specification
26122 one of the fields *must* be zero, and the other non-zero.)
26123
26124 This affects all users of grub_chainloader_patch_bpb which are in
26125 chainloader.c, freedos.c, and ntldr.c
26126
26127 Some descriptions of the semantics of these two fields:
26128
26129 https://www.win.tue.nl/~aeb/linux/fs/fat/fat-1.html
26130
26131 The old 2-byte fields "total number of sectors" and "number of
26132 sectors per FAT" are now zero; this information is now found in
26133 the new 4-byte fields.
26134
26135 (Here given in the FAT32 EBPB section but the total sectors 16/32 bit
26136 fields semantic is true of FAT12 and FAT16 too.)
26137
26138 https://wiki.osdev.org/FAT#BPB_.28BIOS_Parameter_Block.29
26139
26140 19 | 2 | The total sectors in the logical volume. If this value is 0,
26141 it means there are more than 65535 sectors in the volume, and the actual
26142 count is stored in "Large Sectors (bytes 32-35).
26143
26144 32 | 4 | Large amount of sector on media. This field is set if there
26145 are more than 65535 sectors in the volume.
26146
26147 (Doesn't specify what the "large" field is set to when unused, but as
26148 mentioned mkfs.fat sets it to zero then.)
26149
26150 https://technet.microsoft.com/en-us/library/cc976796.aspx
26151
26152 0x13 | WORD | 0x0000 |
26153 Small Sectors . The number of sectors on the volume represented in 16
26154 bits (< 65,536). For volumes larger than 65,536 sectors, this field
26155 has a value of zero and the Large Sectors field is used instead.
26156
26157 0x20 | DWORD | 0x01F03E00 |
26158 Large Sectors . If the value of the Small Sectors field is zero, this
26159 field contains the total number of sectors in the FAT16 volume. If the
26160 value of the Small Sectors field is not zero, the value of this field
26161 is zero.
26162
26163 https://staff.washington.edu/dittrich/misc/fatgen103.pdf page 10
26164
26165 BPB_TotSec16 | 19 | 2 |
26166 This field is the old 16-bit total count of sectors on the volume.
26167 This count includes the count of all sectors in all four regions of the
26168 volume. This field can be 0; if it is 0, then BPB_TotSec32 must be
26169 non-zero. For FAT32 volumes, this field must be 0. For FAT12 and
26170 FAT16 volumes, this field contains the sector count, and
26171 BPB_TotSec32 is 0 if the total sector count “fits” (is less than
26172 0x10000).
26173
26174 BPB_TotSec32 | 32 | 4 |
26175 This field is the new 32-bit total count of sectors on the volume.
26176 This count includes the count of all sectors in all four regions of the
26177 volume. This field can be 0; if it is 0, then BPB_TotSec16 must be
26178 non-zero. For FAT32 volumes, this field must be non-zero. For
26179 FAT12/FAT16 volumes, this field contains the sector count if
26180 BPB_TotSec16 is 0 (count is greater than or equal to 0x10000).
26181
26182 (This specifies that an unused BPB_TotSec32 field is set to zero.)
26183
26184 By the way fix offsets in include/grub/fat.h.
26185
26186 Tested with lDebug booted in qemu via grub2's
26187 FreeDOS direct loading support, refer to
26188 https://bitbucket.org/ecm/ldosboot + https://bitbucket.org/ecm/ldebug
26189
26190 Reviewed-by: Daniel Kiper <[email protected]>
26191
26192 2018-02-23 H.J. Lu <[email protected]>
26193
26194 x86-64: Treat R_X86_64_PLT32 as R_X86_64_PC32
26195 Starting from binutils commit bd7ab16b4537788ad53521c45469a1bdae84ad4a:
26196
26197 https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=bd7ab16b4537788ad53521c45469a1bdae84ad4a
26198
26199 x86-64 assembler generates R_X86_64_PLT32, instead of R_X86_64_PC32, for
26200 32-bit PC-relative branches. Grub2 should treat R_X86_64_PLT32 as
26201 R_X86_64_PC32.
26202
26203 Reviewed-by: Daniel Kiper <[email protected]>
26204
26205 2018-02-14 Steve McIntyre <[email protected]>
26206
26207 Make grub-install check for errors from efibootmgr
26208 Code is currently ignoring errors from efibootmgr, giving users
26209 clearly bogus output like:
26210
26211 Setting up grub-efi-amd64 (2.02~beta3-4) ...
26212 Installing for x86_64-efi platform.
26213 Could not delete variable: No space left on device
26214 Could not prepare Boot variable: No space left on device
26215 Installation finished. No error reported.
26216
26217 and then potentially unbootable systems. If efibootmgr fails, grub-install
26218 should know that and report it!
26219
26220 We've been using similar patch in Debian now for some time, with no ill effects.
26221
26222 Reviewed-by: Daniel Kiper <[email protected]>
26223
26224 2018-02-14 Eric Snowberg <[email protected]>
26225
26226 sparc64: fix OF path names for sun4v systems
26227 Fix the Open Firmware (OF) path property for sun4v SPARC systems.
26228 These platforms do not have a /sas/ within their path. Over time
26229 different OF addressing schemes have been supported. There
26230 is no generic addressing scheme that works across every HBA.
26231
26232 It looks that this functionality will not work if you try to cross-install
26233 SPARC GRUB2 binary using e.g. x86 grub-install. By default it should work.
26234 However, we will also have other issues here, like lack of access to OF
26235 firmware/paths, which make such configs unusable anyway. So, let's leave
26236 this patch as is for time being. If somebody cares then he/she should fix
26237 the issue(s) at some point.
26238
26239 Reviewed-by: Daniel Kiper <[email protected]>
26240
26241 2018-02-14 Eric Snowberg <[email protected]>
26242
26243 sparc64: Add blocklist GPT support for SPARC
26244 Add block-list GPT support for SPARC. The OBP "load" and "boot" methods
26245 are partition aware and neither command can see the partition table. Also
26246 neither command can address the entire physical disk. When the install
26247 happens, grub generates the block-list entries based on the beginning of the
26248 physical disk, not the beginning of the partition. This patch fixes the
26249 block-list entries so they match what OBP expects during boot for a GPT disk.
26250
26251 T5 and above now supports GPT as well as VTOC.
26252
26253 This patch has been tested on T5-2 and newer SPARC systems.
26254
26255 Reviewed-by: Daniel Kiper <[email protected]>
26256
26257 2018-01-29 Stefan Fritsch <[email protected]>
26258
26259 ahci: Improve error handling
26260 Check the error bits in the interrupt status register. According to the
26261 AHCI 1.2 spec, "Interrupt sources that are disabled (‘0’) are still
26262 reflected in the status registers.", so this should work even though
26263 grub uses polling
26264
26265 This fixes the following problem on a Fujitsu E744 laptop:
26266
26267 Sometimes there is a very long delay (up to several minutes) when
26268 booting from hard disk. It seems accessing the DVD drive (which has no
26269 disk inserted) sometimes fails with some errors, which leads to each
26270 access being stalled until the 20s timeout triggers. This seems to
26271 happen when grub is trying to read filesystem/partition data.
26272
26273 The problem is that the command_issue bit that is checked in the loop is
26274 only reset if the "HBA receives a FIS which clears the BSY, DRQ, and ERR
26275 bits for the command", but the ERR bit is never cleared. Therefore
26276 command_issue is never reset and grub waits for the timeout.
26277
26278 The relevant bit in our case is the Task File Error Status (TFES), which
26279 is equivalent to the ERR bit 0 in tfd. But this patch also checks
26280 the other error bits except for the "Interface non-fatal error status"
26281 bit.
26282
26283 Reviewed-by: Daniel Kiper <[email protected]>
26284
26285 2018-01-29 dann frazier <[email protected]>
26286
26287 Keep the native terminal active when enabling gfxterm
26288 grub-mkconfig will set GRUB_TERMINAL_OUTPUT to "gfxterm" unless the user
26289 has overridden it. On EFI systems, this will stop output from going to the
26290 default "console" terminal. When the EFI fw console is configured to output to
26291 both serial and video, this will cause GRUB to only display on video - while
26292 continuing to accept input from both video and serial.
26293
26294 Instead of switching from "console" to "gfxterm", let's output to both.
26295
26296 Reviewed-by: Daniel Kiper <[email protected]>
26297
26298 2017-12-06 Julien Grall <[email protected]>
26299
26300 arm64/xen: Add missing #address-cells and #size-cells properties
26301 The properties #address-cells and #size-cells are used to know the
26302 number of cells for ranges provided by "regs". If they don't exist, the
26303 value are resp. 2 and 1.
26304
26305 Currently, when multiboot nodes are created it is assumed that #address-cells
26306 and #size-cells are exactly 2. However, they are never set by GRUB and
26307 will result to later failure when the device-tree is generated by GRUB
26308 or contain different values.
26309
26310 To prevent this failure, create the both properties in the chosen nodes.
26311
26312 Reviewed-by: Daniel Kiper <[email protected]>
26313
26314 2017-12-06 Jordan Glover <[email protected]>
26315
26316 grub-mkconfig: Fix detecting .sig files as system images
26317 grub-mkconfig detects detached RSA signatures for kernel images used for
26318 signature checking as valid images and adds them to grub.cfg as separate
26319 menu entries. This patch adds .sig extension to common blacklist.
26320
26321 Reviewed-by: Daniel Kiper <[email protected]>
26322
26323 2017-12-06 Eric Snowberg <[email protected]>
26324
26325 ieee1275: Fix segfault in grub-ofpathname
26326 Reviewed-by: Daniel Kiper <[email protected]>
26327
26328 2017-11-28 Eric Snowberg <[email protected]>
26329
26330 grub-install: Fix memory leak
26331 Reviewed-by: Daniel Kiper <[email protected]>
26332
26333 2017-11-24 Eric Snowberg <[email protected]>
26334
26335 ls: prevent double open
26336 Prevent a double open. This can cause problems with some ieee1275
26337 devices, causing the system to hang. The double open can occur
26338 as follows:
26339
26340 grub_ls_list_files (char *dirname, int longlist, int all, int human)
26341 dev = grub_device_open (device_name);
26342 dev remains open while:
26343 grub_normal_print_device_info (device_name);
26344 dev = grub_device_open (name);
26345
26346 Reviewed-by: Daniel Kiper <[email protected]>
26347
26348 2017-10-06 David E. Box <[email protected]>
26349
26350 tsc: Change default tsc calibration method to pmtimer on EFI systems
26351 On efi systems, make pmtimer based tsc calibration the default over the
26352 pit. This prevents Grub from hanging on Intel SoC systems that power gate
26353 the pit.
26354
26355 Reviewed-by: Daniel Kiper <[email protected]>
26356
26357 2017-09-07 Alexander Graf <[email protected]>
26358
26359 efi: Free malloc regions on exit
26360 When we exit grub, we don't free all the memory that we allocated earlier
26361 for our heap region. This can cause problems with setups where you try
26362 to descend the boot order using "exit" entries, such as PXE -> HD boot
26363 scenarios.
26364
26365 Reviewed-by: Daniel Kiper <[email protected]>
26366
26367 2017-09-07 Alexander Graf <[email protected]>
26368
26369 efi: Move grub_reboot() into kernel
26370 The reboot function calls machine_fini() and then reboots the system.
26371 Currently it lives in lib/ which means it gets compiled into the
26372 reboot module which lives on the heap.
26373
26374 In a following patch, I want to free the heap on machine_fini()
26375 though, so we would free the memory that the code is running in. That
26376 obviously breaks with smarter UEFI implementations.
26377
26378 So this patch moves it into the core. That way we ensure that all
26379 code running after machine_fini() in the UEFI case is running from
26380 memory that got allocated (and gets deallocated) by the UEFI core.
26381
26382 Reviewed-by: Daniel Kiper <[email protected]>
26383
26384 2017-09-07 Konrad Rzeszutek Wilk <[email protected]>
26385
26386 Use grub-file to figure out whether multiboot2 should be used for Xen.gz
26387 The multiboot2 is much more preferable than multiboot. Especiall
26388 if booting under EFI where multiboot does not have the functionality
26389 to pass ImageHandler.
26390
26391 Reviewed-by: Daniel Kiper <[email protected]>
26392
26393 2017-09-07 Konrad Rzeszutek Wilk <[email protected]>
26394
26395 Fix util/grub.d/20_linux_xen.in: Add xen_boot command support for aarch64
26396 Commit d33045ce7ffcb7c1e4a60c14d5ca64b36e3c5abe introduced
26397 the support for this, but it does not work under x86 (as it stops
26398 20_linux_xen from running).
26399
26400 The 20_linux_xen is run under a shell and any exits from within it:
26401
26402 (For example on x86):
26403 + /usr/bin/grub2-file --is-arm64-efi /boot/xen-4.9.0.gz
26404 [root@tst063 grub]# echo $?
26405 1
26406
26407 will result in 20_linux_xen exiting without continuing
26408 and also causing grub2-mkconfig to stop processing.
26409
26410 As in:
26411
26412 [root@tst063 grub]# ./grub-mkconfig | tail
26413 Generating grub configuration file ...
26414 Found linux image: /boot/vmlinuz-4.13.0-0.rc5.git1.1.fc27.x86_64
26415 Found initrd image: /boot/initramfs-4.13.0-0.rc5.git1.1.fc27.x86_64.img
26416 Found linux image: /boot/vmlinuz-0-rescue-ec082ee24aea41b9b16aca52a6d10cc2
26417 Found initrd image: /boot/initramfs-0-rescue-ec082ee24aea41b9b16aca52a6d10cc2.img
26418 echo 'Loading Linux 0-rescue-ec082ee24aea41b9b16aca52a6d10cc2 ...'
26419 linux /vmlinuz-0-rescue-ec082ee24aea41b9b16aca52a6d10cc2 root=/dev/mapper/fedora_tst063-root ro single
26420 echo 'Loading initial ramdisk ...'
26421 initrd /initramfs-0-rescue-ec082ee24aea41b9b16aca52a6d10cc2.img
26422 }
26423 }
26424
26425 ### END /usr/local/etc/grub.d/10_linux ###
26426
26427 ### BEGIN /usr/local/etc/grub.d/20_linux_xen ###
26428
26429 root@tst063 grub]#
26430
26431 And no more.
26432
26433 This patch wraps the invocation of grub-file to be a in subshell
26434 and to process the return value in a conditional. That fixes
26435 the issue.
26436
26437 RH-BZ 1486002: grub2-mkconfig does not work if xen.gz is installed.
26438
26439 CC: Fu Wei <[email protected]>
26440 Reviewed-by: Daniel Kiper <[email protected]>
26441
26442 2017-09-07 Vladimir Serbinenko <[email protected]>
26443
26444 Fix compilation for x86_64-efi.
26445
26446 2017-09-05 Vladimir Serbinenko <[email protected]>
26447
26448 Add a file missing in multiboot2 commit.
26449
26450 2017-08-30 Vladimir Serbinenko <[email protected]>
26451
26452 gzio: fix unaligned access
26453
26454 grub-fs-tester: Fix bashism
26455
26456 2017-08-30 Vladimir Serbinenko <[email protected]>
26457
26458 Regenerate checksum.h with newer unifont.
26459 Old link is broken. New unifont is
26460 http://ftp.de.debian.org/debian/pool/main/u/unifont/xfonts-unifont_9.0.06-2_all.deb
26461
26462 printf_unit_test: Disable Wformat-truncation on GCC >= 7
26463 We intentionally pass NULL as argument to format, hence disable the warning.
26464
26465 qemu, coreboot, multiboot: Change linking address to 0x9000.
26466 It's common for distros to use a defective ld which links at 0x9000. Instead
26467 of fighting it, just move link target to 0x9000.
26468
26469 2017-08-30 Stefan Fritsch <[email protected]>
26470
26471 Implement checksum verification for gunzip
26472 This implements the crc32 check for the gzip format. Support for zlib's
26473 adler checksum is not included, yet.
26474
26475 2017-08-30 Vladimir Serbinenko <[email protected]>
26476
26477 xfs: Don't attempt to iterate over empty directory.
26478 Reported by: Tuomas Tynkkynen
26479
26480 2017-08-30 Patrick Steinhardt <[email protected]>
26481
26482 unix exec: avoid atexit handlers when child exits
26483 The `grub_util_exec_redirect_all` helper function can be used to
26484 spawn an executable and redirect its output to some files. After calling
26485 `fork()`, the parent will wait for the child to terminate with
26486 `waitpid()` while the child prepares its file descriptors, environment
26487 and finally calls `execvp()`. If something in the children's setup
26488 fails, it will stop by calling `exit(127)`.
26489
26490 Calling `exit()` will cause any function registered via `atexit()` to be
26491 executed, which is usually the wrong thing to do in a child. And
26492 actually, one can easily observe faulty behaviour on musl-based systems
26493 without modprobe(8) installed: executing `grub-install --help` will call
26494 `grub_util_exec_redirect_all` with "modprobe", which obviously fails if
26495 modprobe(8) is not installed. Due to the child now exiting and invoking
26496 the `atexit()` handlers, it will clean up some data structures of the
26497 parent and cause it to be deadlocked in the `waitpid()` syscall.
26498
26499 The issue can easily be fixed by calling `_exit(127)` instead, which is
26500 especially designed to be called when the atexit-handlers should not be
26501 executed.
26502
26503 2017-08-30 Vladimir Serbinenko <[email protected]>
26504
26505 arc: Do not create spurious variable grub_arc_memory_type_t.
26506
26507 2017-08-14 Xuan Guo <nbdd0121>
26508
26509 Set have_exec to y on cygwin so we have grub_mkrescue.
26510
26511 2017-08-14 Vladimir Serbinenko <[email protected]>
26512
26513 enforcing fixup
26514
26515 multiboot fixup
26516
26517 linux fixup
26518
26519 yylex: Explicilty cast fprintf to void.
26520 It's needed to avoid warning on recent GCC.
26521
26522 genmoddep: Check that no modules provide the same symbol.
26523 The semantics of 2 modules providing the same symbol are undefined. So
26524 ensure that it doesn't happen.
26525
26526 Fix symbols appearing in several modules in linux*.
26527 If same symbol is provided by 2 modules its semantics are undefined.
26528 Avoid this by depending rather than double-including files.
26529
26530 2017-08-14 Vladimir Serbinenko <[email protected]>
26531
26532 multiboot: disentangle multiboot and multiboot2.
26533 Previously we had multiboot and multiboot2 declaring the same symbols.
26534 This can potentially lead to aliasing and strange behaviours when e.g.
26535 module instead of module2 is used with multiboot2.
26536
26537 Bug: #51137
26538
26539 2017-08-14 Vladimir Serbinenko <[email protected]>
26540
26541 hdparm: Depend on hexdump rather than having a second copy of hexdump.
26542
26543 grub.texi: Fix typo
26544 Reported by: Ori Avtalion <saltyhorse>
26545
26546 2017-08-07 Pete Batard <[email protected]>
26547
26548 io: add a GRUB_GZ prefix to gzio specific defines
26549 * This is done to avoid a conflict with a PACKED define in the EDK2
26550
26551 core: use GRUB_TERM_ definitions when handling term characters
26552 * Also use hex value for GRUB_TERM_ESC as '\e' is not in the C standard and is not understood by some compilers
26553
26554 2017-08-07 Leif Lindholm <[email protected]>
26555
26556 efi: change heap allocation type to GRUB_EFI_LOADER_CODE
26557 With upcoming changes to EDK2, allocations of type EFI_LOADER_DATA may
26558 not return regions with execute ability. Since modules are loaded onto
26559 the heap, change the heap allocation type to GRUB_EFI_LOADER_CODE in
26560 order to permit execution on systems with this feature enabled.
26561
26562 Closes: 50420
26563
26564 2017-08-07 Leif Lindholm <[email protected]>
26565
26566 arm64 linux loader: improve type portability
26567 In preparation for turning this into a common loader for 32-bit and 64-bit
26568 platforms, ensure the code will compile cleanly for either.
26569
26570 2017-08-07 Leif Lindholm <[email protected]>
26571
26572 efi: Add GRUB_PE32_MAGIC definition
26573 Add a generic GRUB_PE32_MAGIC definition for the PE 'MZ' tag and delete
26574 the existing one in arm64/linux.h.
26575
26576 Update arm64 Linux loader to use this new definition.
26577
26578 2017-08-07 Leif Lindholm <[email protected]>
26579
26580 efi: move fdt helper library
26581 There is nothing ARM64 (or even ARM) specific about the efi fdt helper
26582 library, which is used for locating or overriding a firmware-provided
26583 devicetree in a UEFI system - so move it to loader/efi for reuse.
26584
26585 Move the fdtload.h include file to grub/efi and update path to
26586 efi/fdtload.h in source code referring to it.
26587
26588 2017-08-07 Vladimir Serbinenko <[email protected]>
26589
26590 Remove grub_efi_allocate_pages.
26591 grub_efi_allocate_pages Essentially does 2 unrelated things:
26592 * Allocate at fixed address.
26593 * Allocate at any address.
26594
26595 To switch between 2 different functions it uses address == 0 as magic
26596 value which is wrong as 0 is a perfectly valid fixed adress to allocate at.
26597
26598 2017-08-07 Leif Lindholm <[email protected]>
26599
26600 efi: refactor grub_efi_allocate_pages
26601 Expose a new function, grub_efi_allocate_pages_real(), making it possible
26602 to specify allocation type and memory type as supported by the UEFI
26603 AllocatePages boot service.
26604
26605 Make grub_efi_allocate_pages() a consumer of the new function,
26606 maintaining its old functionality.
26607
26608 Also delete some left-around #if 1/#else blocks in the affected
26609 functions.
26610
26611 2017-08-07 Vladimir Serbinenko <[email protected]>
26612
26613 Fail if xorriso failed.
26614 If xorriso failed most likely we didn't generate a meaningful image.
26615
26616 mkrescue: Check xorriso presence before doing anything else.
26617 mkrescue can't do anything useful without xorriso, so abort early if it's
26618 not available.
26619
26620 2017-08-07 Pali Rohár <[email protected]>
26621
26622 * grub-core/fs/udf.c: Add support for UUID
26623 Use same algorithm as in libblkid from util-linux v2.30.
26624
26625 1. Take first 16 bytes from UTF-8 encoded string of VolumeSetIdentifier
26626 2. If all bytes are hexadecimal digits, convert to lowercase and use as UUID
26627 3. If first 8 bytes are not all hexadecimal digits, convert those 8 bytes
26628 to their hexadecimal representation, resulting in 16 bytes for UUID
26629 4. Otherwise, compose UUID from two parts:
26630 1. part: converted first 8 bytes (which are hexadecimal digits) to lowercase
26631 2. part: encoded following 4 bytes to their hexadecimal representation (16 bytes)
26632
26633 So UUID would always have 16 hexadecimal digits in lowercase variant.
26634
26635 According to UDF specification, first 16 Unicode characters of
26636 VolumeSetIdentifier should be unique value and first 8 should be
26637 hexadecimal characters.
26638
26639 In most cases all 16 characters are hexadecimal, but e.g. MS Windows
26640 format.exe set only first 8 as hexadecimal and remaining as fixed
26641 (non-unique) which violates specification.
26642
26643 2017-08-07 Pali Rohár <[email protected]>
26644
26645 udf: Fix reading label, lvd.ident is dstring
26646 UDF dstring has stored length in the last byte of buffer. Therefore last
26647 byte is not part of recorded characters. And empty string in dstring is
26648 encoded as empty buffer, including first byte (compression id).
26649
26650 2017-08-07 Pete Batard <[email protected]>
26651
26652 zfs: remove size_t typedef and use grub_size_t instead
26653 * Prevents some toolchains from issuing a warning on size_t redef.
26654
26655 2017-08-03 Rob Clark <[email protected]>
26656
26657 Fix a segfault in lsefi
26658 when protocols_per_handle returns error, we can't use the pointers we
26659 passed to it, and that includes trusting num_protocols.
26660
26661 Reviewed-by: Daniel Kiper <[email protected]>
26662
26663 2017-07-10 Vladimir Serbinenko <[email protected]>
26664
26665 fdt: silence clang warning.
26666
26667 2017-07-09 Vladimir Serbinenko <[email protected]>
26668
26669 arm-efi: Fix compilation
26670
26671 2017-07-09 AppChecker <appchecker>
26672
26673 crypto: Fix use after free.
26674 Reported by: AppChecker
26675 Transformed to patch by: Satish Govindarajan
26676
26677 2017-07-09 Vladimir Serbinenko <[email protected]>
26678
26679 ehci: Fix compilation on i386
26680
26681 2017-07-09 phcoder <[email protected]>
26682
26683 cache: Fix compilation for ppc, sparc and arm64
26684
26685 ehci: Fix compilation for amd64
26686
26687 2017-06-29 Eric Biggers <[email protected]>
26688
26689 Allow GRUB to mount ext2/3/4 filesystems that have the encryption feature.
26690 On such a filesystem, inodes may have EXT4_ENCRYPT_FLAG set.
26691 For a regular file, this means its contents are encrypted; for a
26692 directory, this means the filenames in its directory entries are
26693 encrypted; and for a symlink, this means its target is encrypted. Since
26694 GRUB cannot decrypt encrypted contents or filenames, just issue an error
26695 if it would need to do so. This is sufficient to allow unencrypted boot
26696 files to co-exist with encrypted files elsewhere on the filesystem.
26697
26698 (Note that encrypted regular files and symlinks will not normally be
26699 encountered outside an encrypted directory; however, it's possible via
26700 hard links, so they still need to be handled.)
26701
26702 Tested by booting from an ext4 /boot partition on which I had run
26703 'tune2fs -O encrypt'. I also verified that the expected error messages
26704 are printed when trying to access encrypted directories, files, and
26705 symlinks from the GRUB command line. Also ran 'sudo ./grub-fs-tester
26706 ext4_encrypt'; note that this requires e2fsprogs v1.43+ and Linux v4.1+.
26707
26708 2017-05-29 Eric Snowberg <[email protected]>
26709
26710 sparc64: Don't use devspec to determine the OBP path
26711 Don't use devspec to determine the OBP path on SPARC hardware. Within all
26712 versions of Linux on SPARC, the devspec returns one of three values:
26713 "none", "vnet-port", or "vdisk". Unlike on PPC, none of these values
26714 are useful in determining the OBP path.
26715
26716 Before this patch grub-ofpathname always returned the wrong value
26717 for a virtual disk. For example:
26718
26719 % grub-ofpathname /dev/vdiskc2
26720 vdisk/disk@2:b
26721
26722 After this patch it now returns the correct value:
26723
26724 % grub-ofpathname /dev/vdiskc2
26725 /virtual-devices@100/channel-devices@200/disk@2:b
26726
26727 Orabug: 24459765
26728
26729 Reviewed-by: Daniel Kiper <[email protected]>
26730
26731 2017-05-18 Fu Wei <[email protected]>
26732
26733 arm64: Update the introduction of Xen boot commands in docs/grub.texi
26734 delete: xen_linux, xen_initrd, xen_xsm
26735 add: xen_module
26736
26737 This update bases on
26738 commit 0edd750e50698854068358ea53528100a9192902
26739 Author: Vladimir Serbinenko <[email protected]>
26740 Date: Fri Jan 22 10:18:47 2016 +0100
26741
26742 xen_boot: Remove obsolete module type distinctions.
26743
26744 Also bases on the module loading mechanism of Xen code:
26745 488c2a8 docs/arm64: clarify the documention for loading XSM support
26746 67831c4 docs/arm64: update the documentation for loading XSM support
26747 ca32012 xen/arm64: check XSM Magic from the second unknown module.
26748
26749 Reviewed-by: Julien Grall <[email protected]>
26750 Reviewed-by: Daniel Kiper <[email protected]>
26751
26752 2017-05-18 Fu Wei <[email protected]>
26753
26754 util/grub.d/20_linux_xen.in: Add xen_boot command support for aarch64
26755 This patch adds the support of xen_boot command for aarch64:
26756 xen_hypervisor
26757 xen_module
26758 These two commands are only for aarch64, since it has its own protocol and
26759 commands to boot xen hypervisor and Dom0, but not multiboot.
26760
26761 For other architectures, they are still using multiboot and module
26762 commands.
26763
26764 Reviewed-by: Daniel Kiper <[email protected]>
26765
26766 2017-05-18 Fu Wei <[email protected]>
26767
26768 arm64: Add "--nounzip" option support in xen_module command
26769 This patch adds "--nounzip" option support in order to
26770 be compatible with the module command of multiboot on other architecture,
26771 by this way we can simplify grub-mkconfig support code.
26772
26773 This patch also allow us to use zip compressed module(like Linux kernel
26774 for Dom0).
26775
26776 Reviewed-by: Daniel Kiper <[email protected]>
26777
26778 2017-05-18 Julien Grall <[email protected]>
26779
26780 arm64/xen_boot: Fix Xen boot using GRUB2 on AARCH64
26781 Xen is currently crashing because of malformed compatible property for
26782 the boot module. This is because the property string is not
26783 null-terminated as requested by the ePAR spec.
26784
26785 Tested-by: Fu Wei <[email protected]>
26786 Reviewed-by: Daniel Kiper <[email protected]>
26787
26788 2017-05-18 Eric Snowberg <[email protected]>
26789
26790 sparc64: Close cdboot ihandle
26791 The ihandle is left open with a cd-core image. This will cause a delay
26792 booting grub from a virtual cdrom in a LDOM. It will also cause problems
26793 as Linux boots, since it expects the ihandle to be closed during init.
26794
26795 Orabug: 25911275
26796
26797 Reviewed-by: Daniel Kiper <[email protected]>
26798
26799 2017-05-09 Vladimir Serbinenko <[email protected]>
26800
26801 at_keyboard: Fix falco chromebook case.
26802 EC is slow, so we need few delays for it to toggle the bits correctly.
26803
26804 Command to enable clock and keyboard were not sent.
26805
26806 2017-05-09 Julius Werner <[email protected]>
26807
26808 coreboot: Changed cbmemc to support updated console format from coreboot.
26809
26810 2017-05-09 Vladimir Serbinenko <[email protected]>
26811
26812 Missing parts of previous commit
26813
26814 arm_coreboot: Add Chromebook keyboard driver.
26815
26816 rk3288_spi: Add SPI driver
26817
26818 fdtbus: Add ability to send/receive messages on parent busses.
26819
26820 Fix bug on FDT nodes with compatible property
26821
26822 2017-05-08 Vladimir Serbinenko <[email protected]>
26823
26824 arm_coreboot: Support EHCI.
26825
26826 ehci: Split core code from PCI part.
26827 On ARM often EHCI is present without PCI and just declared in device
26828 tree. So splitcore from PCI part.
26829
26830 arm_coreboot: Support DMA.
26831 This is needed to support USB and some other busses.
26832
26833 arm_coreboot: Support loading linux images.
26834
26835 arm_coreboot: Support grub-mkstandalone.
26836
26837 arm_coreboot: Support keyboard for vexpress.
26838
26839 at_keyboard: Split protocol from controller code.
26840 On vexpress controller is different but protocol is the same, so reuse the
26841 code.
26842
26843 arm-coreboot: Export FDT routines.
26844 We need to use them from modules as well.
26845
26846 arm-coreboot: Support for vexpress timer.
26847
26848 Add support for device-tree-based drivers.
26849
26850 arm-coreboot: Start new port.
26851
26852 Rename uboot/datetime to dummy/datetime.
26853 It's just a stub and is not UBoot-specific.
26854
26855 Rename uboot/halt.c to dummy/halt.c.
26856 It's not U-Boot specific and it's a stub.
26857
26858 coreboot: Split parts that are platform-independent.
26859 We currently assume that coreboot is always i386, it's no longer the case,
26860 so split i386-coreboot parts from generic coreboot code.
26861
26862 Refactor arm-uboot code to make it genereic.
26863 arm-coreboot startup code can be very similar to arm-uboot but current code has
26864 U-Boot specific references. So split U-Boot part from generic part.
26865
26866 mkimage: Pass layout to mkimage_generate_elfXX rather than some fields.
26867 This allows easier extension of this function without having too long of
26868 arguments list.
26869
26870 2017-05-03 Paulo Flabiano Smorigo <[email protected]>
26871
26872 Add Virtual LAN support.
26873 This patch adds support for virtual LAN (VLAN) tagging. VLAN tagging allows
26874 multiple VLANs in a bridged network to share the same physical network link
26875 but maintain isolation:
26876
26877 http://en.wikipedia.org/wiki/IEEE_802.1Q
26878
26879 * grub-core/net/ethernet.c: Add check, get, and set vlan tag id.
26880 * grub-core/net/drivers/ieee1275/ofnet.c: Get vlan tag id from bootargs.
26881 * grub-core/net/arp.c: Add check.
26882 * grub-core/net/ip.c: Likewise.
26883 * include/grub/net/arp.h: Add vlantag attribute.
26884 * include/grub/net/ip.h: Likewise.
26885
26886 2017-05-03 Vladimir Serbinenko <[email protected]>
26887
26888 strtoull: Fix behaviour on chars between '9' and 'a'.
26889 Reported by: Aaron Miller <[email protected]>
26890
26891 Add strtoull test.
26892
26893 Fix shebang for termux.
26894 Termux doesn't have a /bin/sh. So we needto use $SHELL.
26895 Keep /bin/sh as much as possible.
26896
26897 Add termux path to dict.
26898
26899 po: Use @SHELL@ rather than /bin/sh.
26900 /bin/sh might not exist.
26901
26902 Use $(SHELL) rather than /bin/sh.
26903 /bin/sh doesn't exist under termux.
26904
26905 Support lseek64.
26906 Android doesn't have 64-bit off_t, so use off64_t instead.
26907
26908 Don't retrieve fstime when it's not useful.
26909
26910 support busybox date.
26911 Busybox date doesn't understand weekdays in -d input,
26912 so strip them beforehand.
26913
26914 fs-tester: make sh-compatible
26915
26916 Remove bashisms from tests.
26917 Those tests don't actually need bash. Just use common shebang.
26918
26919 Bump version to 2.03
26920
26921 2017-04-25 Vladimir Serbinenko <[email protected]>
26922
26923 Increase version to 2.02.
26924
26925 2017-04-12 Vladimir Serbinenko <[email protected]>
26926
26927 Fix remaining cases of gcc 7 fallthrough warning.
26928 They are all intended, so just add the relevant comment.
26929
26930 2017-04-04 Andrei Borzenkov <[email protected]>
26931
26932 Add gnulib-fix-gcc7-fallthrough.diff
26933 As long as the code is not upstream, add it as explicit patch for the
26934 case of gnulib refresh.
26935
26936 2017-04-04 Andrei Borzenkov <[email protected]>
26937
26938 i386, x86_64, ppc: fix switch fallthrough cases with GCC7
26939 In util/getroot and efidisk slightly modify exitsing comment to mostly
26940 retain it but still make GCC7 compliant with respect to fall through
26941 annotation.
26942
26943 In grub-core/lib/xzembed/xz_dec_lzma2.c it adds same comments as
26944 upstream.
26945
26946 In grub-core/tests/setjmp_tets.c declare functions as "noreturn" to
26947 suppress GCC7 warning.
26948
26949 In grub-core/gnulib/regexec.c use new __attribute__, because existing
26950 annotation is not recognized by GCC7 parser (which requires that comment
26951 immediately precedes case statement).
26952
26953 Otherwise add FALLTHROUGH comment.
26954
26955 Closes: 50598
26956
26957 2017-04-04 Andrei Borzenkov <[email protected]>
26958
26959 btrfs: avoid "used uninitialized" error with GCC7
26960 sblock was local and so considered new variable on every loop
26961 iteration.
26962
26963 Closes: 50597
26964
26965 2017-04-02 Andrei Borzenkov <[email protected]>
26966
26967 acpi: add missing efi_call wrapper to acpi command
26968 Fixed loading of ACPI tables on EFI (side effect was apparent memory
26969 corruption ranging from unpredictable behavior to system reset).
26970
26971 Reported by Nando Eva <[email protected]>
26972
26973 2017-03-15 Vladimir Serbinenko <[email protected]>
26974
26975 Increment version to GRUB 2.02~rc2.
26976
26977 Use core2duo for bootcheck test on 64-bit EFI.
26978 Obviously pentium2 can't run efi64.
26979
26980 2017-03-14 Andrei Borzenkov <[email protected]>
26981
26982 efi: skip iPXE block device.
26983 iPXE adds Simple File System Protocol to loaded image handle, as side
26984 effect it also adds Block IO protocol (according to comments, to work
26985 around some bugs in EDK2). GRUB assumes that every device with Block IO
26986 is disk and skips network initialization entirely. But iPXE Block IO
26987 implementation is just a stub which always fails for every operation
26988 so cannot be used. Attempt to detect and skip such devices.
26989
26990 We are using media ID which iPXE sets to "iPXE" and block IO size in
26991 hope that no real device would announce 1B block ...
26992
26993 Closes: 50518
26994
26995 2017-03-05 phcoder <[email protected]>
26996
26997 xen: Fix wrong register in relocator.
26998 This fixes chainloading of some GRUB variants.
26999
27000 2017-02-27 Vladimir Serbinenko <[email protected]>
27001
27002 video_fb: Fix blue collor if using unoptimized blitter.
27003 when unmapping the color what matters is the mode of source, not target.
27004
27005 legacy_initrd: Strip any additional arguments to initrd.
27006
27007 2017-02-26 Andrei Borzenkov <[email protected]>
27008
27009 grub-fs-tester: improve squash4 tests
27010 1. Make sure files are not multiple of block size. This will ensure tail packing
27011 for squash4 and may also trigger more codes paths in other filesystems.
27012
27013 2. Call mksquashfs with -always-use-fragments to force tail packing.
27014
27015 2017-02-25 Andrei Borzenkov <[email protected]>
27016
27017 efi: strip off final NULL from File Path in grub_efi_get_filename
27018 UEFI 2.6 9.3.6.4 File Path Media Device Path says that Path Name is
27019 "A NULL-terminated Path string including directory and file names".
27020
27021 Strip final NULL from Path Name in each File Path node when constructing
27022 full path. To be on safe side, strip all of them.
27023
27024 Fixes failure chainloading grub from grub, when loaded grub truncates
27025 image path and does not find its grub.cfg.
27026
27027 https://bugzilla.opensuse.org/show_bug.cgi?id=1026344
27028
27029 This was triggered by commit ce95549cc54b5d6f494608a7c390dba3aab4fba7;
27030 before it we built Path Name without trailing NULL, and apparently all
27031 other bootloaders use single File Path node, thus not exposing this bug.
27032
27033 2017-02-24 Andrei Borzenkov <[email protected]>
27034
27035 squash4: fix handling of fragments and sparse files
27036 1. Do not assume block list and fragment are mutually exclusive. Squash
27037 can pack file tail as fragment (unless -no-fragments is specified); so
27038 check read offset and read either from block list or from fragments as
27039 appropriate.
27040
27041 2. Support sparse files with zero blocks.
27042
27043 3. Fix fragment read - frag.offset is absolute fragment position,
27044 not offset relative to ino.chunk.
27045
27046 Reported and tested by Carlo Caione <[email protected]>
27047
27048 2017-02-22 Vladimir Serbinenko <[email protected]>
27049
27050 Whitelist sparc64-ieee1275 as having no video modules.
27051 ieee1275_fb is not built on sparc64 due to virtual address issues.
27052
27053 2017-02-12 Andrei Borzenkov <[email protected]>
27054
27055 script: fix double free in lexer
27056 yylex_destroy() already frees scanner.
27057
27058 Found by: Coverity scan.
27059 CID: 176636
27060
27061 2017-02-07 Vladimir Serbinenko <[email protected]>
27062
27063 xen: Fix parsing of XZ kernel.
27064 In case of xz, the uncompressed size is appended to xz data which confuses
27065 our xz decompressor. Trim it.
27066
27067 2017-02-07 Vladimir Serbinenko <[email protected]>
27068
27069 xen: Fix handling of GRUB chainloading.
27070 In case of GRUB we put remapper after domain pages and not at 0x0.
27071 In this case we use max_addr to put remapper. Unfortunately we increment
27072 max_addr as well in this case resulting in virt mapping mapping page
27073 at old max_addr and trying to boot using new max_addr.
27074
27075 Closes 46014.
27076
27077 2017-02-04 Vladimir Serbinenko <[email protected]>
27078
27079 linguas: Don't skip ko.po.
27080 Translation project doesn't require copyright disclaimers. They're independant
27081 from us. They're responsible for their copyright story.
27082
27083 2017-02-03 Vladimir Serbinenko <[email protected]>
27084
27085 Fix truncated checksum.h.
27086
27087 Regenerate checksums.h
27088 Screenshots contain version, so we need new checksums.
27089
27090 Release 2.02-rc1.
27091
27092 Fix mingw compilation.
27093
27094 2017-02-03 Daniel Kahn Gillmor <[email protected]>
27095
27096 documentation: Clarify documentation for special environment variable "default".
27097 The current documentation for the special environment variable
27098 "default" is confusing and unclear. This patch attempts to clean it
27099 up.
27100
27101 In particular, the current documentation refers to the "number or
27102 title", but then in the example it gives, the menu entries and
27103 submenus all have numbers *in* their title; furthermore, there is no
27104 example given about how to choose the number, or any indication about
27105 whether counting is zero-indexed or 1-indexed.
27106
27107 Having a cleaner example and presenting all variants (numeric, title,
27108 and id) should make it clearer to the user.
27109
27110 2017-02-03 Vladimir Serbinenko <[email protected]>
27111
27112 Avoid causing kernel oops in nilfs2 test.
27113 1024-byte and 2048-byte blocks don't really work with some kernels, skip
27114 them as we don't want any oops'es.
27115
27116 btrfs: Shorten label by one character.
27117 mkfs.btrfs imposes a slightly lower limit than would be possible in btrfs.
27118
27119 2017-02-02 Vladimir Serbinenko <[email protected]>
27120
27121 grub-fs-tester: Fix mkudffs invocation.
27122 With current invocation order of arguments is wrong and path is hardcoded.
27123
27124 grub-fs-tester: Fix fat test.
27125 mkfs.vfat ignores -S when invoked on a disk, including loopback device,
27126 so do an mkfs on underlying image.
27127
27128 2017-02-02 Daniel Kiper <[email protected]>
27129
27130 i386/relocator: Align stack in grub_relocator64_efi relocator
27131 Unified Extensible Firmware Interface Specification, Version 2.6,
27132 section 2.3.4, x64 Platforms, boot services, says among others:
27133 The stack must be 16-byte aligned. So, do it. Otherwise OS may
27134 boot only by chance as it happens right now.
27135
27136 2017-02-02 Vladimir Serbinenko <[email protected]>
27137
27138 i386-ieee1275: Add missing bootcheck target.
27139
27140 bootcheck-linux-i386: Use -cpu pentium2.
27141 Most modern kernels are compiled for i686, so use -cpu pentium2
27142 to avoid spurious failures.
27143
27144 Use -fPIC with arm64 with clang.
27145 Currently it doesn't work either way but with -fPIC it should work once
27146 clang bug is fixed.
27147
27148 INSTALL: Fix mention of thumb-clang.
27149
27150 Fix thumb compilation with clang.
27151 According to EABI only STT_FUNC has convention of lowest bit indicating
27152 execution mode. R_THM_{JUMP,CALL}* relocations are assumed to be pointing
27153 to thumb mode unless they use STT_FUNC.
27154
27155 2017-02-01 Vladimir Serbinenko <[email protected]>
27156
27157 Add missing strtoull_test.c
27158 It was forgotten in my local directory.
27159
27160 arm64: Add support for GOT and PCREL32 relocations.
27161
27162 mkimage: Fix memory leak.
27163
27164 arm/arm64: Fix improper use of start address.
27165 It was used instead of loading address of current section or of entire buffer.
27166
27167 ia64: Fix iterator for relocation entries.
27168 Don't assume relocation entry size and use sh_entsize properly.
27169
27170 arm: Fix trampoline generation.
27171 We used the wrong pointer in this case. It worked only by accident.
27172
27173 Fix bootcheck-related files compilation.
27174 We need -static as otherwise linker will set interpreter field and ld.so
27175 is not available on our initrd's.
27176 Strip all sections we don't need on binary tests.
27177
27178 2017-01-31 Vladimir Serbinenko <[email protected]>
27179
27180 Regenerate checksum.h.
27181 Screenshots checked.
27182 Using unifont from http://ftp.us.debian.org/debian/pool/main/u/unifont/xfonts-unifont_7.0.06-1_all.deb.
27183
27184 grub-mkfont: Remove leftover debug statement.
27185
27186 charset: Trim away RLM and LRM.
27187 They are not visible but would otherwise end up as [LRM] or [RLM] squares
27188 with some fonts.
27189
27190 gfxterm: Fix clearing of cursor.
27191 If ascent is bigger than height - 2, then we draw over character box but then
27192 to clear cursor we only draw over character box. So trim ascent if necessarry.
27193
27194 ia64: Add support for R_IA64_GPREL64I.
27195 Recent GCC generates those relocations, so we need to support them.
27196
27197 2017-01-30 Vladimir Serbinenko <[email protected]>
27198
27199 grub-module-verifier: Add mips to all_video whitelist.
27200 On MIPS video is compiled-in. So all_video is empty. Whitelist it.
27201
27202 Fix -nopie/-nopie check.
27203 We don't use lgcc_s but missing lgcc_s or another library cause test to fail.
27204 So use -nostdlib.
27205 We need to use -Werror to avoid warning-generated case to be accepted.
27206 Clang uses -nopie rather than -no-pie. Check both and use whichever one works.
27207 Additionally android clang passes -pie to the linker even though it doesn't
27208 define __PIE__. So if compilation without no-pie logic fails add -nopie/-no-pie
27209 even if __PIE__ is not defined.
27210
27211 grub-module-verifier: Ignore all_video emptiness on xen.
27212 It's intentional that it's empty when no video modules
27213 are available.
27214
27215 2017-01-28 Vladimir Serbinenko <[email protected]>
27216
27217 Support arm clang 3.8 amd later.
27218 clang 3.8 and later doesn't support -mllvm -use-arm-movt=0
27219 whereas older clang doesn't know -mno-movt. So use
27220 -mno-movt whenever possible and fallback to mllvm variant.
27221
27222 2017-01-27 Carlo Caione <[email protected]>
27223
27224 exfat: Support files over 4GiB
27225 file size in grub_fat_data was 32-bit on exfat.
27226
27227 2017-01-27 Vladimir Serbinenko <[email protected]>
27228
27229 Ensure that grub_reboot doesn't return on emu.
27230 Use grub_fatal if longjmp fails.
27231
27232 grub_reboot is marked as noreturn so return would cause
27233 a crash.
27234
27235 2017-01-27 Vladimir Serbinenko <[email protected]>
27236
27237 grub-shell: skip font copying when no font is available.
27238
27239 Don't use -mlong-calls on arm.
27240 We don't really need it and it's flaky and creates
27241 bogus symbols with clang.
27242
27243 configure: Disable movw/movt with clang.
27244 Those relocations are not compatible with PE and also
27245 not compatible with custom uboot relocator.
27246 Disable them.
27247
27248 grub-fs-tester: Delete directory once we're done.
27249
27250 grub-fs-tester: Accomodate for slower systems.
27251 fstime can be more different with xz squashfs.
27252 Allow difference up to 3 seconds.
27253 This code is ugly now but rewriting it now is not on the
27254 table.
27255
27256 grub-fs-tester: Accomodate for testing in proot containers.
27257 proot creates hidden files with .proot prefix and name
27258 derived from real file name. So decrease file name length
27259 and path depth. For some reason depth 85 also results in
27260 undeleteable directory, so use 84 instead of 85.
27261
27262 2017-01-24 Andrei Borzenkov <[email protected]>
27263
27264 osdep/linux: handle autofs entries in /proc/self/mountinfo
27265 These entries have placeholder for device name and so are useless for our
27266 purpose. grub failed with something like
27267
27268 grub-install: error: failed to get canonical path of `systemd-1'.
27269
27270 When we see autofs entry, record it (to keep parent-child relationship) but
27271 continue to look for real mount. If it is found, we process it as usual. If
27272 only autofs entry exists, attempt to trigger mount by opening mount point
27273 and retry. Mount point itself is then kept open to avoid timeout.
27274
27275 Recent systemd is by default using automount for /boot/efi so this should
27276 become more popular problem on EFI systems.
27277
27278 Closes: 49942
27279
27280 2017-01-08 Andrei Borzenkov <[email protected]>
27281
27282 linux: fix "vga=XX deprecated" warning for text mode
27283 Arguments were in reverse order which resulted in
27284
27285 text is deprecated. Use set gfxpayload=vga=0 before linux command instead.
27286
27287 2016-12-22 Andrei Borzenkov <[email protected]>
27288
27289 configure: fix check for sys/sysmacros.h under glibc 2.25+
27290 glibc 2.25 still includes sys/sysmacros.h in sys/types.h but also emits
27291 deprecation warning. So test for sys/types.h succeeds in configure but later
27292 compilation fails because we use -Werror by default.
27293
27294 While this is fixed in current autoconf GIT, we really cannot force everyone
27295 to use bleeding edge (that is not even released right now). So run test under
27296 -Werror as well to force proper detection.
27297
27298 This should have no impact on autoconf 2.70+ as AC_HEADER_MAJOR in this version
27299 simply checks for header existence.
27300
27301 Reported and tested by Khem Raj <[email protected]>
27302
27303 2016-12-22 Michael Chang <[email protected]>
27304
27305 Fix fwpath in efi netboot
27306 The path returned by grub_efi_net_config has already been stripped for the
27307 directory part extracted from cached bootp packet. We should just return the
27308 result to avoild it be stripped again.
27309
27310 It fixed the problem that grub.efi as NBP image always looking for grub.cfg and
27311 platform directory in upper folder rather than current one it gets loaded while
27312 $prefix is empty. The behavior is inconsistent with other architecture and how
27313 we would expect empty $prefix going to be in general.
27314
27315 The only exception to the general rule of empty $prefix is that when loaded
27316 from platform directory itself, the platform part is stripped thus upper folder
27317 is used for looking up files. It meets the case for how grub-mknetdir lay out
27318 the files under tftp root directory, but also hide away this issue to be
27319 identified as it appears to be just works.
27320
27321 Also fix possible memory leak by moving grub_efi_get_filename() call after
27322 grub_efi_net_config().
27323
27324 2016-12-15 Andrei Borzenkov <[email protected]>
27325
27326 efi: properly terminate filepath with NULL in chainloader
27327 EFI File Path Media Device Path is defined as NULL terminated string;
27328 but chainloader built file paths without final NULL. This caused error
27329 with Secure Boot and Linux Foundation PreLoader on Acer with InsydeH20 BIOS.
27330 Apparently firmware failed verification with EFI_INVALID_PARAMETER which is
27331 considered fatal error by PreLoader.
27332
27333 Reported and tested by Giovanni Santini <[email protected]>
27334
27335 2016-12-14 Magnus Granberg <[email protected]>
27336
27337 configure: add check for -no-pie if the compiler default to -fPIE
27338 When Grub is compile with gcc 6.1 that have --enable-defult-pie set.
27339 It fail with.
27340 -ffreestanding -m32 -Wl,-melf_i386 -Wl,--build-id=none -nostdlib -Wl,-N -Wl,-r,-d -
27341 o trig.module trig_module-trigtables.o
27342 grep 'MARKER' gcry_whirlpool.marker.new > gcry_whirlpool.marker; rm -f
27343 gcry_whirlpool.marker.new
27344 /usr/lib/gcc/x86_64-pc-linux-gnu/6.1.0/../../../../x86_64-pc-linux-gnu/bin/ld: -r and -
27345 shared may not be used together
27346 collect2: error: ld returned 1 exit status
27347 Makefile:26993: recipe for target 'trig.module' failed
27348
27349 Check that compiler supports -no-pie and add it to linker flags.
27350
27351 2016-12-14 Stanislav Kholmanskikh <[email protected]>
27352
27353 ofnet: implement the receive buffer
27354 get_card_packet() from ofnet.c allocates a netbuff based on the device's MTU:
27355
27356 nb = grub_netbuff_alloc (dev->mtu + 64 + 2);
27357
27358 In the case when the MTU is large, and the received packet is
27359 relatively small, this leads to allocation of significantly more memory,
27360 than it's required. An example could be transmission of TFTP packets
27361 with 0x400 blksize via a network card with 0x10000 MTU.
27362
27363 This patch implements a per-card receive buffer in a way similar to efinet.c,
27364 and makes get_card_packet() allocate a netbuff of the received data size.
27365
27366 Reviewed-by: Daniel Kiper <[email protected]>
27367
27368 2016-12-14 Stanislav Kholmanskikh <[email protected]>
27369
27370 ofnet: move the allocation of the transmit buffer into a function
27371 In the current code search_net_devices() uses the "alloc-mem" command
27372 from the IEEE1275 User Interface for allocation of the transmit buffer
27373 for the case when GRUB_IEEE1275_FLAG_VIRT_TO_REAL_BROKEN is set.
27374
27375 I don't have hardware where this flag is set to verify if this
27376 workaround is still needed. However, further changes to ofnet will
27377 require to execute this workaround one more time. Therefore, to
27378 avoid possible duplication of code I'm moving this piece of
27379 code into a function.
27380
27381 Reviewed-by: Daniel Kiper <[email protected]>
27382
27383 2016-11-24 Alexander Graf <[email protected]>
27384
27385 efi: Move fdt helper into own file
27386 We only support FDT files with EFI on arm and arm64 systems, not
27387 on x86. So move the helper that finds a prepopulated FDT UUID
27388 into its own file and only build it for architectures where it
27389 also gets called.
27390
27391 Reviewed-by: Daniel Kiper <[email protected]>
27392
27393 2016-11-22 Andrei Borzenkov <[email protected]>
27394
27395 NEWS updates
27396
27397 2016-11-22 Andrei Borzenkov <[email protected]>
27398
27399 bootp: export next server IP as environment variable
27400 Network boot autoconfiguration sets default server to next server IP
27401 (siaddr) from BOOTP/DHCP reply, but manual configuration using net_bootp
27402 exports only server name. Unfortunately semantic of server name is not
27403 clearly defined. BOOTP RFC 951 defines it only for client request, and
27404 DHCP RFC 1541 only mentions it, without any implied usage. It looks like
27405 this field is mostly empty in server replies.
27406
27407 Export next server IP as net_<interface>_next_server variable. This allows
27408 grub configuration script to set $root/$prefix based on information obtained
27409 by net_bootp.
27410
27411 Reported and tested by: Nikunj A Dadhania <[email protected]>
27412 Cc: [email protected]
27413
27414 v2: change variable name to net_<interface>_next_server as discussed on the list
27415
27416 2016-11-22 Aaro Koskinen <[email protected]>
27417
27418 configure.ac: don't require build time grub-mkfont on powerpc-ieee1275
27419 Don't require build time grub-mkfont on powerpc-ieee1275.
27420
27421 2016-11-14 Dirk Mueller <[email protected]>
27422
27423 grub-mknetdir: Add support for ARM64 EFI
27424
27425 2016-11-12 Joonas Lahtinen <[email protected]>
27426
27427 .gitignore: Add grub-core/build-grub-module-verifier
27428
27429 2016-11-10 Alexander Graf <[email protected]>
27430
27431 arm efi: Use fdt from firmware when available
27432 If EFI is nice enough to pass us an FDT using configuration tables on 32bit
27433 ARM, we should really try and make use of it.
27434
27435 Reviewed-by: Daniel Kiper <[email protected]>
27436
27437 2016-11-10 Alexander Graf <[email protected]>
27438
27439 arm64: Move firmware fdt search into global function
27440 Searching for a device tree that EFI passes to us via configuration tables
27441 is nothing architecture specific. Move it into generic code.
27442
27443 Reviewed-by: Daniel Kiper <[email protected]>
27444
27445 2016-11-05 Corey Hickey <[email protected]>
27446
27447 fix detection of non-LUKS CRYPT
27448 grub_util_get_dm_abstraction() does a string comparison of insufficient
27449 length. When using a UUID such as "CRYPT-PLAIN-sda6_crypt", the function
27450 returns GRUB_DEV_ABSTRACTION_LUKS.
27451
27452 This results in the error:
27453 ./grub-probe: error: disk `cryptouuid/sda6_crypt' not found.
27454
27455 This appears to be a copy/paste error introduced in:
27456 a10e7a5a8918bea6e2632055129fa9b516fe965a
27457
27458 The bug was (apparently) latent until revealed by:
27459 3bca85b4184f74995a7cc2791e432173fde26d34
27460
27461 2016-10-27 Juergen Gross <[email protected]>
27462
27463 xen: add capability to load p2m list outside of kernel mapping
27464 Modern pvops linux kernels support a p2m list not covered by the
27465 kernel mapping. This capability is flagged by an elf-note specifying
27466 the virtual address the kernel is expecting the p2m list to be mapped
27467 to.
27468
27469 In case the elf-note is set by the kernel don't place the p2m list
27470 into the kernel mapping, but map it to the given address. This will
27471 allow to support domains with larger memory, as the kernel mapping is
27472 limited to 2GB and a domain with huge memory in the TB range will have
27473 a p2m list larger than this.
27474
27475 Reviewed-by: Daniel Kiper <[email protected]>
27476
27477 2016-10-27 Juergen Gross <[email protected]>
27478
27479 xen: modify page table construction
27480 Modify the page table construction to allow multiple virtual regions
27481 to be mapped. This is done as preparation for removing the p2m list
27482 from the initial kernel mapping in order to support huge pv domains.
27483
27484 This allows a cleaner approach for mapping the relocator page by
27485 using this capability.
27486
27487 The interface to the assembler level of the relocator has to be changed
27488 in order to be able to process multiple page table areas.
27489
27490 Reviewed-by: Daniel Kiper <[email protected]>
27491
27492 2016-10-27 Juergen Gross <[email protected]>
27493
27494 xen: add capability to load initrd outside of initial mapping
27495 Modern pvops linux kernels support an initrd not covered by the initial
27496 mapping. This capability is flagged by an elf-note.
27497
27498 In case the elf-note is set by the kernel don't place the initrd into
27499 the initial mapping. This will allow to load larger initrds and/or
27500 support domains with larger memory, as the initial mapping is limited
27501 to 2GB and it is containing the p2m list.
27502
27503 Reviewed-by: Daniel Kiper <[email protected]>
27504
27505 2016-10-27 Juergen Gross <[email protected]>
27506
27507 xen: factor out allocation of page tables into separate function
27508 Do the allocation of page tables in a separate function. This will
27509 allow to do the allocation at different times of the boot preparations
27510 depending on the features the kernel is supporting.
27511
27512 Reviewed-by: Daniel Kiper <[email protected]>
27513
27514 2016-10-27 Juergen Gross <[email protected]>
27515
27516 xen: factor out allocation of special pages into separate function
27517 Do the allocation of special pages (start info, console and xenbus
27518 ring buffers) in a separate function. This will allow to do the
27519 allocation at different times of the boot preparations depending on
27520 the features the kernel is supporting.
27521
27522 Reviewed-by: Daniel Kiper <[email protected]>
27523
27524 2016-10-27 Juergen Gross <[email protected]>
27525
27526 xen: factor out p2m list allocation into separate function
27527 Do the p2m list allocation of the to be loaded kernel in a separate
27528 function. This will allow doing the p2m list allocation at different
27529 times of the boot preparations depending on the features the kernel
27530 is supporting.
27531
27532 While at this remove superfluous setting of first_p2m_pfn and
27533 nr_p2m_frames as those are needed only in case of the p2m list not
27534 being mapped by the initial kernel mapping.
27535
27536 Reviewed-by: Daniel Kiper <[email protected]>
27537
27538 2016-10-27 Juergen Gross <[email protected]>
27539
27540 xen: synchronize xen header
27541 Get actual version of include/xen/xen.h from the Xen repository in
27542 order to be able to use constants defined there.
27543
27544 Reviewed-by: Daniel Kiper <[email protected]>
27545
27546 2016-10-27 Juergen Gross <[email protected]>
27547
27548 xen: add elfnote.h to avoid using numbers instead of constants
27549 Various features and parameters of a pv-kernel are specified via
27550 elf notes in the kernel image. Those notes are part of the interface
27551 between the Xen hypervisor and the kernel.
27552
27553 Instead of using num,bers in the code when interpreting the elf notes
27554 make use of the header supplied by Xen for that purpose.
27555
27556 Reviewed-by: Daniel Kiper <[email protected]>
27557
27558 2016-10-27 Juergen Gross <[email protected]>
27559
27560 xen: reduce number of global variables in xen loader
27561 The loader for xen paravirtualized environment is using lots of global
27562 variables. Reduce the number by making them either local or by putting
27563 them into a single state structure.
27564
27565 Reviewed-by: Daniel Kiper <[email protected]>
27566
27567 2016-10-27 Juergen Gross <[email protected]>
27568
27569 xen: avoid memleaks on error
27570 When loading a Xen pv-kernel avoid memory leaks in case of errors.
27571
27572 Reviewed-by: Daniel Kiper <[email protected]>
27573
27574 2016-10-27 Juergen Gross <[email protected]>
27575
27576 xen: make xen loader callable multiple times
27577 The loader for xen paravirtualized environment isn't callable multiple
27578 times as it won't free any memory in case of failure.
27579
27580 Call grub_relocator_unload() as other modules do it before allocating
27581 a new relocator or when unloading the module.
27582
27583 Reviewed-by: Daniel Kiper <[email protected]>
27584
27585 2016-10-27 Daniel Kiper <[email protected]>
27586
27587 multiboot2: Add support for relocatable images
27588 Currently multiboot2 protocol loads image exactly at address specified in
27589 ELF or multiboot2 header. This solution works quite well on legacy BIOS
27590 platforms. It is possible because memory regions are placed at predictable
27591 addresses (though I was not able to find any spec which says that it is
27592 strong requirement, so, it looks that it is just a goodwill of hardware
27593 designers). However, EFI platforms are more volatile. Even if required
27594 memory regions live at specific addresses then they are sometimes simply
27595 not free (e.g. used by boot/runtime services on Dell PowerEdge R820 and
27596 OVMF). This means that you are not able to just set up final image
27597 destination on build time. You have to provide method to relocate image
27598 contents to real load address which is usually different than load address
27599 specified in ELF and multiboot2 headers.
27600
27601 This patch provides all needed machinery to do self relocation in image code.
27602 First of all GRUB2 reads min_addr (min. load addr), max_addr (max. load addr),
27603 align (required image alignment), preference (it says which memory regions are
27604 preferred by image, e.g. none, low, high) from multiboot_header_tag_relocatable
27605 header tag contained in binary (at this stage load addresses from multiboot2
27606 and/or ELF headers are ignored). Later loader tries to fulfill request (not only
27607 that one) and if it succeeds then it informs image about real load address via
27608 multiboot_tag_load_base_addr tag. At this stage GRUB2 role is finished. Starting
27609 from now executable must cope with relocations itself using whole static and
27610 dynamic knowledge provided by boot loader.
27611
27612 This patch does not provide functionality which could do relocations using
27613 ELF relocation data. However, I was asked by Konrad Rzeszutek Wilk and Vladimir
27614 'phcoder' Serbinenko to investigate that thing. It looks that relevant machinery
27615 could be added to existing code (including this patch) without huge effort.
27616 Additionally, ELF relocation could live in parallel with self relocation provided
27617 by this patch. However, during research I realized that first of all we should
27618 establish the details how ELF relocatable image should look like and how it should
27619 be build. At least to build proper test/example files.
27620
27621 So, this patch just provides support for self relocatable images. If ELF file
27622 with relocs is loaded then GRUB2 complains loudly and ignores it. Support for
27623 such files will be added later.
27624
27625 This patch was tested with Xen image which uses that functionality. However, this Xen
27626 feature is still under development and new patchset will be released in about 2-3 weeks.
27627
27628 Reviewed-by: Vladimir Serbinenko <[email protected]>
27629
27630 2016-10-27 Daniel Kiper <[email protected]>
27631
27632 multiboot2: Do not pass memory maps to image if EFI boot services are enabled
27633 If image requested EFI boot services then skip multiboot2 memory maps.
27634 Main reason for not providing maps is because they will likely be
27635 invalid. We do a few allocations after filling them, e.g. for relocator
27636 needs. Usually we do not care as we would have finished boot services.
27637 If we keep boot services then it is easier/safer to not provide maps.
27638 However, if image needs memory maps and they are not provided by bootloader
27639 then it should get itself just before ExitBootServices() call.
27640
27641 Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
27642 Reviewed-by: Vladimir Serbinenko <[email protected]>
27643
27644 2016-10-27 Daniel Kiper <[email protected]>
27645
27646 multiboot2: Add tags used to pass ImageHandle to loaded image
27647 Add tags used to pass ImageHandle to loaded image if requested.
27648 It is used by at least ExitBootServices() function.
27649
27650 Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
27651 Reviewed-by: Vladimir Serbinenko <[email protected]>
27652
27653 2016-10-27 Daniel Kiper <[email protected]>
27654
27655 i386/relocator: Add grub_relocator64_efi relocator
27656 Add grub_relocator64_efi relocator. It will be used on EFI 64-bit platforms
27657 when multiboot2 compatible image requests MULTIBOOT_TAG_TYPE_EFI_BS. Relocator
27658 will set lower parts of %rax and %rbx accordingly to multiboot2 specification.
27659 On the other hand processor mode, just before jumping into loaded image, will
27660 be set accordingly to Unified Extensible Firmware Interface Specification,
27661 Version 2.4 Errata B, section 2.3.4, x64 Platforms, boot services. This way
27662 loaded image will be able to use EFI boot services without any issues.
27663
27664 Reviewed-by: Konrad Rzeszutek Wilk <[email protected]>
27665 Reviewed-by: Vladimir Serbinenko <[email protected]>
27666
27667 2016-10-18 Sakar Arora <[email protected]>
27668
27669 net/ip: Fix limit_time calculation in freeing old fragments
27670 limit_time underflows when current time is less than 90000ms.
27671 This causes packet fragments received during this time, i.e.,
27672 till 90000ms pass since timer init, to be rejected.
27673
27674 Hence, set it to 0 if its less than 90000.
27675
27676 2016-09-28 Andrei Borzenkov <[email protected]>
27677
27678 asm-tests/i386-pc: Check that movl is 5 bytes.
27679 LLVM 3.9 now emits short form of jump instructions, but it is still using
27680 32 bit addresses for some movl instructions. Fortunately it was caught early:
27681
27682 clang ... boot/i386/pc/boot.S
27683 clang -cc1as: fatal error: error in backend: invalid .org offset '440' (at offset '441')
27684
27685 Add additional check to catch it during configure run and force -no-integrated-as.
27686
27687 Closes: 49200
27688
27689 More details in
27690 https://lists.gnu.org/archive/html/grub-devel/2015-02/msg00099.html
27691 https://llvm.org/bugs/show_bug.cgi?id=22662
27692
27693 2016-08-13 Pete Batard <[email protected]>
27694
27695 Add missing va_end() to xasprintf() in grub-emu.
27696
27697 2016-07-27 Andrei Borzenkov <[email protected]>
27698
27699 at_keyboard: fix numpad "0" and "." mapping
27700 Reported for set 1 by fgndevelop <[email protected]>. Apparently
27701 set 2 was reversed too.
27702
27703 2016-07-26 Andrei Borzenkov <[email protected]>
27704
27705 dns: fix buffer overflow for data->addresses in recv_hook
27706 We may get more than one response before exiting out of loop in
27707 grub_net_dns_lookup, but buffer was allocated for the first response only,
27708 so storing answers from subsequent replies wrote past allocated size.
27709 We never really use more than the very first address during lookup so there
27710 is little point in collecting all of them. Just quit early if we already have
27711 some reply.
27712
27713 Code needs serious redesign to actually collect multiple answers
27714 and select the best fit according to requested type (IPv4 or IPv6).
27715
27716 Reported and tested by Michael Chang <[email protected]>
27717
27718 2016-07-26 Andrei Borzenkov <[email protected]>
27719
27720 xfs: accept filesystem with meta_uuid
27721 XFS V5 stores UUID in metadata and compares them with superblock UUID.
27722 To allow changing of user-visible UUID it stores original value in new
27723 superblock field (meta_uuid) and sets incompatible flag to indicate that
27724 new field must be used to verify metadata. Our driver currently does not
27725 check metadata UUID so simply accept such filesystem.
27726
27727 Reported-By: Marcos Mello <[email protected]>
27728 Reviewd by Jan Kara <[email protected]>
27729
27730 2016-05-03 Andrei Borzenkov <[email protected]>
27731
27732 net: translate pxe prefix to tftp when checking for self-load
27733 Commit ba218c1 missed legacy pxe and pxe: prefixes which are
27734 translated to tftp, so comparison failed.
27735
27736 2016-04-30 Andrei Borzenkov <[email protected]>
27737
27738 net: reset net->stall in grub_net_seek_real
27739 If we open new connection, we need to reset stall indication, otherwise
27740 nothing will ever be polled (low level code rely on this field being
27741 zero when establishing connection).
27742
27743 2016-04-30 Stefan Fritsch <[email protected]>
27744
27745 http: reset EOF indication in http_seek
27746 Otherwise next read will stop polling too early due to stale EOF
27747 indicator, returning incomplete data to caller.
27748
27749 2016-04-24 Mike Gilbert <[email protected]>
27750
27751 build: Use AC_HEADER_MAJOR to find device macros
27752 Depending on the OS/libc, device macros are defined in different
27753 headers. This change ensures we include the right one.
27754
27755 sys/types.h - BSD
27756 sys/mkdev.h - Sun
27757 sys/sysmacros.h - glibc (Linux)
27758
27759 glibc currently pulls sys/sysmacros.h into sys/types.h, but this may
27760 change in a future release.
27761
27762 https://sourceware.org/ml/libc-alpha/2015-11/msg00253.html
27763
27764 2016-04-09 Michael Chang <[email protected]>
27765
27766 http: fix superfluous null line in range request header
27767 At least the apache sever is very unhappy with that extra null line and will
27768 take more than ten seconds in responding to each range request, which slows
27769 down a lot the entire http file transfer process or even time out.
27770
27771 2016-03-22 Andrei Borzenkov <[email protected]>
27772
27773 configure: set -fno-pie together with -fno-PIE
27774 OpenBSD 5.9 apparently defaults to -fpie. We use -fno-PIE when appropriate
27775 already, but that is not enough - it does not turn off -fpie.
27776
27777 Actually check for -fPIE is not precise enough. __PIE__ is set for both
27778 -fpie and -fPIE but with different values. As far as I can tell, both
27779 options were introduced at the same time, so both should always be supported.
27780
27781 This fixes compilation on OpenBSD 5.9 which otherwise created insanely big
27782 lzma_decompress.img.
27783
27784 Reported, suggested and tested by: Jiri B <[email protected]>
27785
27786 2016-03-20 Andrei Borzenkov <[email protected]>
27787
27788 bootp: check that interface is not NULL in configure_by_dhcp_ack
27789 grub_net_add_addr may fail with OOM and we use returned interface
27790 later without any checks.
27791
27792 2016-03-19 Andrei Borzenkov <[email protected]>
27793
27794 bootp: fix memory leak in grub_cmd_dhcpopt
27795
27796 2016-03-15 Aaron Luft <[email protected]>
27797
27798 Remove the variable oldname which is attempting to free stack space.
27799 Historically this variable hold previous value of filename that
27800 had to be freed if allocated previously. Currently this branch
27801 is entered only if filename was not allocated previously so it
27802 became redundant. It did not cause real problems because grub_free
27803 was not called, but code is confusing and causes compilation error
27804 in some cases.
27805
27806 2016-03-13 Andrei Borzenkov <[email protected]>
27807
27808 Makefile.util.def: add $LIBINTL to grub-macbless flags
27809 Fixes compilation on OpenBSD 5.9.
27810
27811 Reported by Jiri B <[email protected]>
27812
27813 2016-03-11 Robert Marshall <[email protected]>
27814
27815 Failed config now returns exit code (#1252311)
27816 Grub would notify the user if the new config was invalid, however, it
27817 did not exit properly with exit code 1. Added the proper exit code.
27818
27819 Resolves: rhbz#1252311
27820
27821 2016-03-11 Michael Chang <[email protected]>
27822
27823 xen_file: Fix invalid payload size
27824
27825 2016-03-10 Vladimir Serbinenko <[email protected]>
27826
27827 multiboot2: Remove useless GRUB_PACKED
27828 Reported by: Daniel Kiper
27829
27830 2016-03-06 Andrei Borzenkov <[email protected]>
27831
27832 20_linux_xen: fix test for GRUB_DEVICE
27833 Same fix as in 082bc9f.
27834
27835 2016-03-06 Mike Gilbert <[email protected]>
27836
27837 10_linux: Fix grouping of tests for GRUB_DEVICE
27838 Commit 7290bb562 causes GRUB_DISABLE_LINUX_UUID to be ignored due to
27839 mixing of || and && operators. Add some parens to help with that.
27840
27841 2016-02-28 Andrei Borzenkov <[email protected]>
27842
27843 NEWS update
27844
27845 2016-02-28 Vladimir Serbinenko <[email protected]>
27846
27847 Release 2.02~beta3
27848
27849 grub_arch_sync_dma_caches: Accept volatile address
27850
27851 2016-02-27 Leif Lindholm <[email protected]>
27852
27853 efidisk: Respect block_io_protocol buffer alignment
27854 Returned from the OpenProtocol operation, the grub_efi_block_io_media
27855 structure contains the io_align field, specifying the minimum alignment
27856 required for buffers used in any data transfers with the device.
27857
27858 Make grub_efidisk_readwrite() allocate a temporary buffer, aligned to
27859 this boundary, if the buffer passed to it does not already meet the
27860 requirements.
27861
27862 Also sanity check the io_align field in grub_efidisk_open() for
27863 power-of-two-ness and bail if invalid.
27864
27865 2016-02-27 Vladimir Serbinenko <[email protected]>
27866
27867 usbtrans: Fix memory coherence and use-after-free.
27868
27869 ehci: Fix memory coherence
27870 This is a no-op on x86 but necessarry on ARM and may be necessarry on MIPS.
27871
27872 arm-uboot: Make self-relocatable to allow loading at any address
27873
27874 Allow _start == 0 with relocatable images
27875
27876 2016-02-27 Vladimir Serbinenko <[email protected]>
27877
27878 Provide __bss_start and _end symbols in grub-mkimage.
27879 For this ensure that all bss sections are merged.
27880
27881 We need this to correctly prelink non-PE relocatable images.
27882
27883 2016-02-27 Vladimir Serbinenko <[email protected]>
27884
27885 Encapsulate image layout into a separate structure.
27886 Currently we pass around a lot of pointer. Instead put all relevant data
27887 into one structure.
27888
27889 mkimagexx: Split PE and generic part for relocations.
27890 As a preparation for U-Boot relocations, split emitting PE-relocations
27891 from parsing source ELF-relocations.
27892
27893 mkimage.c: Split into separate files.
27894 util/grub-mkimagexx.c is included in a special way into mkimage.c.
27895 Interoperation between defines makes this very tricky. Instead
27896 just have a clean interface and compile util/grub-mkimage*.c separately
27897 from mkimage.c
27898
27899 bsd: Ensure that kernel is loaded before loading module.
27900 kernel_type may be set to the type of failed kernel. This patching-up is
27901 easier than to reflow kernel loading routines.
27902
27903 cat: Don't switch terminal mode when there is nothing to highlight.
27904 This just pollutes serial console.
27905
27906 Use console rather than serial_efi0 on arm64-efi in tests
27907
27908 2016-02-27 Andrei Borzenkov <[email protected]>
27909
27910 efidisk: fix misplaced parenthesis in b00e4c2
27911
27912 2016-02-26 Andrei Borzenkov <[email protected]>
27913
27914 efidisk: prevent errors from diskfilter scan of removable drives
27915 Map EFI_NO_MEDIA to GRUB_ERR_OUT_OF_RANGE that is ignored by diskfilter. This
27916 actually matches pretty close (we obviously attempt to read outside of media)
27917 and avoids adding more error codes.
27918
27919 This affects only internally initiated scans. If read/write from removable is
27920 explicitly requested, we still return an error and text explanation is more
27921 clear for user than generic error.
27922
27923 Reported and tested by Andreas Loew <[email protected]>
27924
27925 2016-02-26 Vladimir Serbinenko <[email protected]>
27926
27927 Regenerate checksums
27928
27929 Makefile: Don't delete default_payload.elf if it doesn't exist.
27930
27931 2016-02-25 Josef Bacik <[email protected]>
27932
27933 net: fix ipv6 routing
27934 ipv6 routing in grub2 is broken, we cannot talk to anything outside our local
27935 network or anything that doesn't route in our global namespace. This patch
27936 fixes this by doing a couple of things
27937
27938 1) Read the router information off of the router advertisement. If we have a
27939 router lifetime we need to take the source address and create a route from it.
27940
27941 2) Changes the routing stuff slightly to allow you to specify a gateway _and_ an
27942 interface. Since the router advertisements come in on the link local address we
27943 need to associate it with the global address on the card. So when we are
27944 processing the router advertisement, either use the SLAAC interface we create
27945 and add the route to that interface, or loop through the global addresses we
27946 currently have on our interface and associate it with one of those addresses.
27947 We need to have a special case here for the default route so that it gets used,
27948 we do this by setting the masksize to 0 to mean it encompasses all networks.
27949 The routing code will automatically select the best route so if there is a
27950 closer match we will use that.
27951
27952 With this patch I can now talk to ipv6 addresses outside of my local network.
27953 Thanks,
27954
27955 2016-02-24 Vladimir Serbinenko <[email protected]>
27956
27957 ieee1275: fix signed comparison
27958
27959 2016-02-23 Andrei Borzenkov <[email protected]>
27960
27961 search: actually skip floppy with --no-floppy
27962 grub_device_iterate() ignores device when iterator returns 1, not 0.
27963
27964 Reported by Carlos E. R. <[email protected]>
27965
27966 2016-02-23 Andrei Borzenkov <[email protected]>
27967
27968 multiboot2: zero reserved field in memory map
27969 Documentation says, bootloader should set reserved field to zero.
27970
27971 Reported by Wink Saville <[email protected]>
27972
27973 2016-02-22 Vladimir Serbinenko <[email protected]>
27974
27975 Improve EHCI logging
27976 Add dprintf's on common error paths and remove some entries which are too
27977 noisy.
27978
27979 usb_keyboard: Remove useless include
27980 This prevents non-PCI machines from having USB.
27981
27982 Refresh before abort
27983 This ensures that abort message is actually visible to the user.
27984
27985 2016-02-22 Eric Snowberg <[email protected]>
27986
27987 ieee1275: prevent buffer over-read
27988 Prevent buffer over-read in grub_machine_mmap_iterate. This was
27989 causing phys_base from being calculated properly. This then
27990 caused the wrong value to be placed in ramdisk_image within
27991 struct linux_hdrs. Which prevented the ramdisk from loading on
27992 boot.
27993
27994 Newer SPARC systems contain more than 8 available memory entries.
27995
27996 For example on a T5-8 with 2TB of memory, the memory layout could
27997 look like this:
27998
27999 T5-8 Memory
28000 reg 00000000 30000000 0000003f b0000000
28001 00000800 00000000 00000040 00000000
28002 00001000 00000000 00000040 00000000
28003 00001800 00000000 00000040 00000000
28004 00002000 00000000 00000040 00000000
28005 00002800 00000000 00000040 00000000
28006 00003000 00000000 00000040 00000000
28007 00003800 00000000 00000040 00000000
28008 available 00003800 00000000 0000003f ffcae000
28009 00003000 00000000 00000040 00000000
28010 00002800 00000000 00000040 00000000
28011 00002000 00000000 00000040 00000000
28012 00001800 00000000 00000040 00000000
28013 00001000 00000000 00000040 00000000
28014 00000800 00000000 00000040 00000000
28015 00000000 70000000 0000003f 70000000
28016 00000000 6eef8000 00000000 00002000
28017 00000000 30400000 00000000 3eaf6000
28018 name memory
28019
28020 2016-02-22 Thomas Huth <[email protected]>
28021
28022 menu_entry: Disable cursor during update_screen()
28023 When running grub in a VGA console of a KVM pseries guest on PowerPC,
28024 you can see the cursor sweeping over the whole line when entering a
28025 character in editor mode. This is visible because grub always refreshes
28026 the whole line when entering a character in editor mode, and drawing
28027 characters is quite a slow operation with the firmware used for the
28028 powerpc pseries guests (SLOF).
28029 To avoid this ugliness, the cursor should be disabled when refreshing
28030 the screen contents during update_screen().
28031
28032 2016-02-17 Vladimir Serbinenko <[email protected]>
28033
28034 default_payload.elf: Always rebuild and remove before build.
28035 It's difficult to know all dependencies. Since it's manual and cheap
28036 target anyway, simply always rebuild it.
28037
28038 default_payload.elf: Include password_pbkdf2.
28039 Withoout this module we may end up in a system where no password is
28040 accepted.
28041
28042 default_payload.elf: Add modules from $(EXTRA_PAYLOAD_MODULES).
28043 This allows coreboot building system to add extra modules depending
28044 on user config.
28045
28046 mm: Avoid integer overflow.
28047
28048 Remove -Wno-maybe-uninitialized as it may not be present.
28049
28050 Fix warnings when compiling with -O3
28051
28052 2016-02-14 Vladimir Serbinenko <[email protected]>
28053
28054 Add wbinvd around bios call.
28055 Via C3 has problems with cache coherency when transitioning between the modes,
28056 so flush it around bios calls.
28057
28058 2016-02-12 Eric Snowberg <[email protected]>
28059
28060 OBP available region contains grub. Start at grub_phys_end.
28061 This prevents a problem where grub was being overwritten since
28062 grub_phys_start does not start at a zero offset within the memory
28063 map.
28064
28065 2016-02-12 Andreas Freimuth <[email protected]>
28066
28067 Add Thinkpad T410s button cmos address.
28068
28069 2016-02-12 Vladimir Serbinenko <[email protected]>
28070
28071 TODO: Remove obsolete link
28072
28073 2016-02-12 Toomas Soome <[email protected]>
28074
28075 lz4: Fix pointer overflow
28076
28077 2016-02-12 Vladimir Serbinenko <[email protected]>
28078
28079 grub-shell: Update 32-bit OVMF binary name.
28080
28081 2016-02-12 Daniel Kiper <[email protected]>
28082
28083 relocator: Fix integer underflow.
28084
28085 2016-02-12 Vladimir Serbinenko <[email protected]>
28086
28087 Change -v to -V for version of shell utils.
28088
28089 xnu: Add new kernel path to autoconfig.
28090
28091 arm64: Use cpu timer for timekeeping.
28092
28093 powerpc: Trim header in tests.
28094
28095 default_payload: Include syslinuxcfg, all filesystems and xnu.
28096
28097 xnu: Supply random seed.
28098 Now we're able to load kernels up to El Capitan.
28099
28100 Add RNG module.
28101
28102 yylex: use grub_fatal for exit.
28103 lexer calls yylex_fatal on fatal internal errors. yylex_fatal itself is
28104 declared as noreturn and calls exit. Returning from noreturn function has
28105 unpredictable consequences.
28106
28107 printf: Fix and test %% behaviour in presence of subsequenbt args.
28108
28109 Split pmtimer wait and tsc measurement from pmtimer tsc calibration.
28110
28111 Make grub_cpu_is_tsc_supported generally available.
28112
28113 Make grub_acpi_find_fadt accessible generically
28114
28115 Make unaligned types public.
28116 This simplifies code which has to handle those types.
28117
28118 Fix emu compilation error on arm.
28119
28120 2016-02-11 Vladimir Serbinenko <[email protected]>
28121
28122 xnu: Include relocated EFI in heap size.
28123
28124 xnu: supply ramsize to the kernel.
28125 Without this info recent kernels crash as they allocate no heap.
28126
28127 2016-02-03 Andrei Borzenkov <[email protected]>
28128
28129 support modules without symbol table
28130 all_video module does not have any code or data and exists solely for
28131 .moddeps section to pull in dependencies. This makes all symbols unneeded.
28132
28133 While in current binutils (last released version as of this commit is 2.26)
28134 ``strip --strip-unneeded'' unintentionally adds section symbols for each
28135 existing section, this behavior was considered a bug and changed in commit
28136 14f2c699ddca1e2f706342dffc59a6c7e23e844c to completely strip symbol table
28137 in this case.
28138
28139 Older binutils (verified with 2.17) and some other toolchains (at least
28140 elftoolchain r3223M), both used in FreeBSD, remove symbol table in all_video
28141 as well.
28142
28143 Relax run-time check and do not return error for modules without symbol table.
28144 Add additional checks to module verifier to make sure such modules
28145
28146 a) have non-empty .moddeps section. Without either externally visible symbols
28147 or .moddeps modules are completely useless and should not be built.
28148
28149 b) do not have any relocations.
28150
28151 Closes: 46986
28152
28153 v2: add run-time check for empty symbol table if relocations are present as
28154 suggested by Vladimir.
28155
28156 2016-02-01 Andrei Borzenkov <[email protected]>
28157
28158 10_linux: avoid multi-device root= kernel argument
28159 If root filesystem is multidev btrfs, do not attempt to pass all devices as
28160 kernel root= argument. This results in splitting command line in GRUB due to
28161 embedded newline and even if we managed to quote it, kernel does not know how
28162 to interpret it anyway. Multidev btrfs requires user space device scanning,
28163 so passing single device would not work too.
28164
28165 This still respects user settings GRUB_DISABLE_LINUX_UUID. Not sure what we
28166 should do in this case.
28167
28168 Closes: 45709
28169
28170 2016-01-22 Vladimir Serbinenko <[email protected]>
28171
28172 Error out if mtools invocation fails.
28173
28174 arm64: Add support for relocations needed for linaro gcc
28175
28176 efiemu: Fix compilation failure
28177
28178 Document cpuid -p
28179
28180 2016-01-22 Robert Elliott <[email protected]>
28181
28182 efiemu: Handle persistent RAM and unknown possible future additions.
28183
28184 2016-01-22 Vladimir Serbinenko <[email protected]>
28185
28186 Document expr1 expr2 syntax for test command
28187
28188 2016-01-22 Michael Chang <[email protected]>
28189
28190 Restore terminal settings on grub-emu exit.
28191
28192 2016-01-22 Vladimir Serbinenko <[email protected]>
28193
28194 xen_boot: Remove obsolete module type distinctions.
28195
28196 arm: Ignore qemu clock bug
28197
28198 i386-ieee1275: Increase maximum heap size to accomodate highres graphi tests
28199
28200 2016-01-20 Colin Watson <[email protected]>
28201
28202 Remove pragmas related to -Wunreachable-code
28203 -Wunreachable-code has been a no-op since GCC 4.5; GRUB hasn't been
28204 compiled with it since 2012; and GCC 6 produces "error:
28205 '-Wunreachable-code' is not an option that controls warnings" for these.
28206
28207 Fixes Debian bug #812047.
28208
28209 2016-01-16 Colin Watson <[email protected]>
28210
28211 loader/bsd: Fix signed/unsigned comparison
28212
28213 ahci, ehci: Fix typos
28214
28215 2016-01-16 Andrei Borzenkov <[email protected]>
28216
28217 grub-probe: fix memory leak
28218 Found by: Coverity scan.
28219 CID: 73783
28220
28221 2016-01-16 Andrei Borzenkov <[email protected]>
28222
28223 tftp: fix memory leaks in open
28224 If protocol open fails, file is immediately freed, so data was leaked.
28225
28226 Found by: Coverity scan.
28227 CID: 96659
28228
28229 2016-01-16 Andrei Borzenkov <[email protected]>
28230
28231 tcp: fix memory leaks
28232 Found by: Coverity scan.
28233 CID: 96639, 96647
28234
28235 net: fix memory leaks
28236 Found by: Coverity scan.
28237 CID: 96638, 96648
28238
28239 legacycfg: fix memory leaks and add NULL check
28240 Memory leaks found by Coverity scan.
28241 CID: 96642, 96645
28242
28243 2016-01-15 Andrei Borzenkov <[email protected]>
28244
28245 loader: Unintended sign extension
28246 CID: 96707, 96699, 96693, 96691, 96711, 96709, 96708, 96703, 96702,
28247 96700, 96698, 96696, 96695, 96692, 96710, 96705
28248
28249 2016-01-12 Andrei Borzenkov <[email protected]>
28250
28251 script: fix memory leak
28252 Found by: Coverity scan.
28253 CID: 96637
28254
28255 normal: fix memory leak
28256 Found by: Coverity scan.
28257 CID: 96641, 96670, 96667
28258
28259 xnu: fix memory leak
28260 Found by: Coverity scan.
28261 CID: 96663
28262
28263 truecrypt: fix memory leak
28264 Found by: Coverity scan.
28265 CID: 156611
28266
28267 gfxmenu: fix memory leak
28268 Found by: Coverity scan.
28269 CID: 96657
28270
28271 efiemu: fix memory leak
28272 Found by: Coverity scan.
28273 CID: 156610
28274
28275 efidisk: fix memory leak
28276 Found by: Coverity scan.
28277 CID: 96644
28278
28279 verify: fix memory leak
28280 Found by: Coverity scan.
28281 CID: 96643
28282
28283 password_pbkdf2: fix memory leak
28284 Found by: Coverity scan.
28285 CID: 96656
28286
28287 parttool: fix memory leak
28288 Found by: Coverity scan.
28289 CID: 96652
28290
28291 2016-01-12 Andrei Borzenkov <[email protected]>
28292
28293 nativedisk: fix memory leak
28294 Based on Coverity scan.
28295 CID: 96660
28296
28297 Extended to also cover other error return places.
28298
28299 2016-01-12 Andrei Borzenkov <[email protected]>
28300
28301 acpi: fix memory leak
28302 Found by: Coverity scan.
28303 CID: 96673
28304
28305 2016-01-10 Andrei Borzenkov <[email protected]>
28306
28307 grub-install: include ehci in list of native modules
28308 This matches behavior of "nativedisk" command.
28309
28310 Reported and tested by Smith Henry <[email protected]>
28311
28312 2016-01-10 Andrei Borzenkov <[email protected]>
28313
28314 grub-mkimage: remove redundant NULL check
28315 Found by: Coverity scan.
28316 CID: 73737
28317
28318 2016-01-10 Andrei Borzenkov <[email protected]>
28319
28320 net: remove dead and redundant code
28321 server cannot be NULL at this point (we return error earlier if it is).
28322 Also structure is zalloc'ed, so no need to explicitly initialize
28323 members to 0.
28324
28325 Found by: Coverity scan.
28326 CID: 73837
28327
28328 2016-01-10 Andrei Borzenkov <[email protected]>
28329
28330 hostdisk: fix device detection
28331 Condition was apparently reversed so GRUB assumed all devices were
28332 files. This later made it skip BLKFLSBUF ioctl on Linux which caused
28333 various page cache coherency issues. Observed were
28334
28335 - failure to validate blocklist install (read content did not match
28336 just written)
28337
28338 - failure to detect Linux MD on disk after online hot addition
28339 (GRUB got stale superblock)
28340
28341 Closes: 46691
28342
28343 2016-01-09 Andrei Borzenkov <[email protected]>
28344
28345 setup: fix NULL pointer dereference
28346 Check return value of grub_guess_root_devices
28347
28348 Found by: Coverity scan.
28349 CID: 73638, 73751
28350
28351 2016-01-09 Andrei Borzenkov <[email protected]>
28352
28353 mkimage: fix unintended sign extension
28354 Found by: Coverity scan.
28355 CID: 73691, 73717
28356
28357 2016-01-09 Andrei Borzenkov <[email protected]>
28358
28359 util/getroot: delete dead code
28360 is_part cannot be non-zero at this point.
28361
28362 Found by: Coveruty scan.
28363 CID: 73838
28364
28365 2016-01-09 Andrei Borzenkov <[email protected]>
28366
28367 loader/multiboot: fix unintended sign extension
28368 Found by: Coveruty scan.
28369 CID: 73700, 73763
28370
28371 kern/elf: fix unintended sign extension
28372 Found by: Coverity scan.
28373 CID: 73729, 73735, 73758, 73760
28374
28375 2016-01-09 Andrei Borzenkov <[email protected]>
28376
28377 xfs: fix possible inode corruption in directory scan
28378 grub_xfs_iterate_dir did not restore first character after inline
28379 name when match was found. Dependning on XFS format this character
28380 could be inode number and we could return to the same node later in
28381 find_file if processing cycled symlinks.
28382
28383 CID: 86724
28384
28385 2016-01-09 Andrei Borzenkov <[email protected]>
28386
28387 rescue_parser: restructure code to avoid Coverity false positive
28388 If line contains single word, line and argv[0] are aliases, so
28389 no NULL dereference is possible, but Coverity does not know it.
28390 Change code to avoid ambiguity and also remove redundant call to
28391 grub_strchr.
28392
28393 CID: 86725
28394
28395 2016-01-09 Andrei Borzenkov <[email protected]>
28396
28397 grub-mklayout: check subscript bounds
28398 Found by: Coverity scan.
28399 CID: 73686
28400
28401 grub-probe: fix memory leak
28402 Found by: Coverity scan.
28403 CID: 73783
28404
28405 gfxmenu: fix memory leak
28406 Found by: Coverity scan.
28407 CID: 73766
28408
28409 2016-01-09 Andrei Borzenkov <[email protected]>
28410
28411 util/setup: fix grub_util_path_list leak
28412 Add helper grub_util_free_path_list and use it where appropriate.
28413
28414 Found by: Coverity scan.
28415 CID: 73727
28416
28417 2016-01-09 Andrei Borzenkov <[email protected]>
28418
28419 setup: fix memory leak
28420 Found by: Coverity scan.
28421 CID: 73680, 73715
28422
28423 efiemu: check return value of grub_efiemu_write_value
28424 Found by: Coverity scan.
28425 CID: 73590
28426
28427 efiemu: change code to avoid Coverity false positive
28428 CID: 73623
28429
28430 efiemu: fix unintended sign extension
28431 Found by: Coverity scan.
28432 CID: 73883, 73637
28433
28434 hfs: fix memory leak
28435 Found by: Coverity scan.
28436 CID: 156531
28437
28438 grub-module-verifier: fix unintended sign extension
28439 Found by: Coverity scan.
28440 CID: 156533, 156532
28441
28442 2016-01-08 Vladimir Serbinenko <[email protected]>
28443
28444 Tests: Support arm-efi
28445
28446 2016-01-07 Vladimir Serbinenko <[email protected]>
28447
28448 arm64/setjmp: Add missing move for arg1 == 0 case.
28449
28450 grub-shell: Support arm64-efi
28451
28452 2016-01-07 Mark Salter <[email protected]>
28453
28454 arm-efi: Reduce timer event frequency by 10
28455 Timer event to keep grub msec counter was running at 1000HZ. This was too
28456 fast for UEFI timer driver and resulted in a 10x slowdown in grub time
28457 versus wallclock. Reduce the timer event frequency and increase tick
28458 increment accordingly to keep better time.
28459
28460 2016-01-07 Vladimir Serbinenko <[email protected]>
28461
28462 x86_64-efi: Automatically add -bios OVMF.fd to qemu in tests.
28463
28464 Allow GRUB_QEMU_OPTS to override machine.
28465
28466 arm64: Disable tests that need native drivers.
28467
28468 Disable NetBSD bootcheck on EFI until it supports ACPI on EFI.
28469
28470 grub-shell: Use new cbfstool syntax.
28471
28472 grub-shell: On i386-ieee1275 don't try to switch to console.
28473 console goes to serial as well, so this doesn't stop garbage from going
28474 to serial. But it creates garbage itself.
28475
28476 hddboot_test: reenable on OVMF
28477 OVMF now supports booting from disks.
28478
28479 iee1275/datetime: Fix off-by-1 error.
28480
28481 2016-01-07 Vladimir Serbinenko <[email protected]>
28482
28483 Adjust bootcheck tests for multiboot/coreboot/qemu to match real support.
28484 coreboot has ACPI while 2 others don't. *BSD need ACPI and have trouble
28485 without it. Don't even attempt to boot *BSD on multiboot or qemu targets.
28486
28487 On coreboot boot all *BSD except 32-bit NetBSD which apparently does some
28488 early BIOS calls.
28489
28490 2016-01-05 Vladimir Serbinenko <[email protected]>
28491
28492 minixfs_test: Check if mkfs.minixfs supports -B option.
28493
28494 Add memdisk support to grub-emu.
28495 Use it to add custom files, so that tests which need them work.
28496
28497 Move file loading functions to grub-emu.
28498 So that we can use it in grub-emu as well as utils.
28499
28500 Disable progress indicator in grub-shell.
28501 This disables progress indicator for tests. This in turn fixes test
28502 flakiness as they ended up timing-dependent.
28503
28504 Update checksums
28505
28506 2016-01-02 Andrei Borzenkov <[email protected]>
28507
28508 acpihalt: add GRUB_ACPI_OPCODE_CREATE_DWORD_FIELD (0x8a)
28509 Fixes ACPI halt on ASUSTeK P8B75-V,
28510 Bios: American Megatrends v: 0414 date: 04/24/2012
28511
28512 Reported-By: Goh Lip <[email protected]>
28513
28514 2016-01-02 Andrei Borzenkov <[email protected]>
28515
28516 acpihalt: fix GRUB_DSDT_TEST compilation
28517
28518 2016-01-01 Andrei Borzenkov <[email protected]>
28519
28520 Add missing BUILD_EXEEXT
28521
28522 2015-12-31 Vladimir Serbinenko <[email protected]>
28523
28524 configure.ac: Reorder efiemu check to after link format check.
28525 efiemu is supposed to be disabled when compiling through exe format.
28526 Unfortunately format was determined only after efiemu check. Reorder to fix the
28527 problem
28528
28529 2015-12-31 Andrey Borzenkov <[email protected]>
28530
28531 remove temporary .bin files (kernel and modules)
28532
28533 add dejavu built fonts to cleanfiles
28534
28535 2015-12-31 Andrei Borzenkov <[email protected]>
28536
28537 Add grub-module-verifier files to EXTRA_DIST
28538
28539 2015-12-31 Vladimir Serbinenko <[email protected]>
28540
28541 configure: Add -fno-unwind-tables if supported.
28542 Unwind tables are useless for us bt consume space if present. Ensure that they
28543 are not.
28544
28545 module-verifier: allow limited-range relocations on sparc64.
28546 clang as incomplete mcmodel=large support. As we don't currently need full
28547 mcmodel=large support for sparc64, relax those checks.
28548
28549 Disable build-time module check on emu.
28550 On emu some checks can be laxer like check for relocation range. Additionally
28551 module loading in emu is rarely used. So skip this check rather than making
28552 it laxer for all platforms. In ideal we may want to have slightly different
28553 check for emu but for now this is good enough.
28554
28555 configure: Fix grub_cv_cc_fno_unwind_tables check.
28556 Check tries -fno-dwarf2-cfi-asm but adds -fno-asynchronous-unwind-tables
28557 to TARGET_CFLAGS. Fix this.
28558
28559 Add -mno-stack-arg-probe on mingw.
28560 This argument disables generation of calls to __chkstk_ms. Those calls are
28561 useless on GRUB as function is dummy. Yet they increase module size and
28562 use limited-range relocations which may not work under some memory layouts.
28563 We currently don't use such layouts on concerned platforms but lt's correct
28564 this.
28565
28566 Strip .ARM.exidx
28567 This section is generated by clang and is useful only for debugging.
28568 It contains exotic relocations, so strip them to avoid them interferring
28569 with module loading.
28570
28571 module-verifier: Check range-limited relative relocations.
28572 Check that they point to the same module, so will end up in the same
28573 chunk of memory.
28574
28575 xen/relocator: Use local symbol to ensure that code is relocation-free.
28576
28577 backtrace: Fix register call syntax
28578
28579 Verify modules on build-time rather than failing in runtime.
28580
28581 sparc64: Fix assembly to let compiler to fill in memory references.
28582 This fixes the use of not fully relocatable (they assume that variables are
28583 under 4G limit in virtual memory) references.
28584
28585 2015-12-30 Andrey Borzenkov <[email protected]>
28586
28587 30_os-prober: derive --class from os-prober generated label
28588 Currently only Windows gets distinguished icons, everything else is displayed
28589 using the same generic one. Add additional --class based on os-prober returned
28590 label, which usually is expected to match primary distribution name.
28591
28592 Also use it for Windows as well - chainloader prober may actually return
28593 different strings (Windows, MS-DOS, Windows9xME).
28594
28595 2015-12-30 Vladimir Serbinenko <[email protected]>
28596
28597 backtrace: Remove assembly assumption that grub_backtrace_pointer is under 4G
28598
28599 2015-12-30 Andrei Borzenkov <[email protected]>
28600
28601 menu: fix line count calculation for long lines
28602 It gave one extra screen line if length was exactly equal to screen
28603 width.
28604
28605 Reported by Michael Chang.
28606 Also-By: Michael Chang <[email protected]>
28607
28608 2015-12-29 Vladimir Serbinenko <[email protected]>
28609
28610 grub-mkrescue: Delete temporary file
28611 Reported by: Thomas Schmitt
28612
28613 grub-mount: Fix oath parsing.
28614 Brackets detection was copied from somewhere else and makes no sense in case
28615 of grub-mount and prevents user from accessing and files with ) in them.
28616
28617 exfat: Fix stream extension flag parsing.
28618
28619 2015-12-26 Andrei Borzenkov <[email protected]>
28620
28621 devmapper: check for valid device abstraction in get_grub_dev
28622 This was lost when code was refactored. Patch restores previous behavior.
28623
28624 It is still not clear whether this is the right one. Due to the way we
28625 detect DM abstraction, partitions on DM are skipped, we fall through to
28626 generic detection which ends up in assuming parent device is BIOS disk.
28627
28628 It is useful to install GRUB on VM disk from the host. But it also means
28629 that GRUB will mistakenly allow install on real system as well.
28630
28631 For now let's fix regression; future behavior needs to be discussed.
28632
28633 Closes: 45163
28634
28635 2015-12-19 Andrei Borzenkov <[email protected]>
28636
28637 windows: correct LBA in generated EFI HDD media paths
28638 GRUB keeps partition offset and size in units of 512B sectors. Media paths
28639 are defined in terms of LBA which are presumed to match HDD sector size.
28640
28641 This is probably cosmetic (EFI requires that partition is searched by GUID)
28642 and still incorrect if GPT was created using different logical block size.
28643 But current code is obviously wrong and new has better chances to be correct.
28644
28645 2015-12-17 Robert Elliott <[email protected]>
28646
28647 lsefimmap: support persistent memory and other UEFI 2.5 features
28648 This should accompany
28649 76ce1de740 Translate UEFI persistent memory type
28650
28651 1. Add a string for the EfiPersistentMemory type 14 that was
28652 added in UEFI 2.5.
28653
28654 2. Decode the memory attributes that were added in UEFI 2.5:
28655 * NV (non-volatile)
28656 * MORE_RELIABLE (higher reliable, e.g., mirrored memory in a system
28657 with partial memory mirroring)
28658 * RO (read-only)
28659
28660 3. Use proper IEC binary units (KiB, MiB, etc.) for power-of-two
28661 values rather than misusing SI power-of-ten units (KB, MB, etc.)
28662
28663 4. The lsmmap command only decodes memory ranges sizes up to GiB scale
28664 units. Persistent memory ranges will reach into the TiB scale.
28665 Since 64-bit size field supports TiB, PiB, and EiB, decode all of
28666 them for completeness.
28667
28668 5. In the lsefimmap command, rewrite the print statements to
28669 * avoid rounding
28670 * avoid a big nested if/else tree.
28671
28672 For example: In the sixth entry below, the value of 309MB implies
28673 316416KB but is really reporting 316436KB.
28674
28675 Widen the size column to 6 digits to accommodate typical cases.
28676 The worst case value would require 14 digits; if that happens,
28677 let the columns get out of sync.
28678
28679 Old format:
28680 Type Physical start - end #Pages Size Attributes
28681 conv-mem 0000000000000000-0000000000092fff 00000093 588KB UC WC WT WB
28682 reserved 0000000000093000-0000000000093fff 00000001 4KB UC WC WT WB
28683 conv-mem 0000000000094000-000000000009ffff 0000000c 48KB UC WC WT WB
28684 conv-mem 0000000000100000-000000000fffffff 0000ff00 255MB UC WC WT WB
28685 BS-code 0000000010000000-0000000010048fff 00000049 292KB UC WC WT WB
28686 conv-mem 0000000010049000-000000002354dfff 00013505 309MB UC WC WT WB
28687 ldr-data 000000002354e000-000000003ecfffff 0001b7b2 439MB UC WC WT WB
28688 BS-data 000000003ed00000-000000003ed7ffff 00000080 512KB UC WC WT WB
28689 conv-mem 000000003ed80000-000000006af5ffff 0002c1e0 705MB UC WC WT WB
28690 reserved 000000006af60000-000000006b55ffff 00000600 6MB UC WC WT WB
28691 BS-data 000000006b560000-000000006b560fff 00000001 4KB UC WC WT WB
28692 RT-data 000000006b561000-000000006b5e1fff 00000081 516KB RT UC WC WT WB
28693 BS-data 000000006b5e2000-000000006ecfafff 00003719 55MB UC WC WT WB
28694 BS-code 000000006ecfb000-000000006ecfbfff 00000001 4KB UC WC WT WB
28695 conv-mem 000000006ecfc000-00000000711fafff 000024ff 36MB UC WC WT WB
28696 BS-data 00000000711fb000-000000007128dfff 00000093 588KB UC WC WT WB
28697 Unk 0d 0000000880000000-0000000e7fffffff 00600000 24GB UC WC WT WB NV
28698 reserved 0000001680000000-0000001c7fffffff 00600000 24GB UC WC WT WB NV
28699
28700 New format:
28701 Type Physical start - end #Pages Size Attributes
28702 conv-mem 0000000000000000-0000000000092fff 00000093 588KiB UC WC WT WB
28703 reserved 0000000000093000-0000000000093fff 00000001 4KiB UC WC WT WB
28704 conv-mem 0000000000094000-000000000009ffff 0000000c 48KiB UC WC WT WB
28705 conv-mem 0000000000100000-000000000fffffff 0000ff00 255MiB UC WC WT WB
28706 BS-code 0000000010000000-0000000010048fff 00000049 292KiB UC WC WT WB
28707 conv-mem 0000000010049000-000000002354dfff 00013505 316436KiB UC WC WT WB
28708 ldr-data 000000002354e000-000000003ecfffff 0001b7b2 450248KiB UC WC WT WB
28709 BS-data 000000003ed00000-000000003ed7ffff 00000080 512KiB UC WC WT WB
28710 conv-mem 000000003ed80000-000000006af5ffff 0002c1e0 722816KiB UC WC WT WB
28711 reserved 000000006af60000-000000006b55ffff 00000600 6MiB UC WC WT WB
28712 BS-data 000000006b560000-000000006b560fff 00000001 4KiB UC WC WT WB
28713 RT-data 000000006b561000-000000006b5e1fff 00000081 516KiB RT UC WC WT WB
28714 BS-data 000000006b5e2000-000000006ecfafff 00003719 56420KiB UC WC WT WB
28715 BS-code 000000006ecfb000-000000006ecfbfff 00000001 4KiB UC WC WT WB
28716 conv-mem 000000006ecfc000-0000000071222fff 00002527 38044KiB UC WC WT WB
28717 BS-data 0000000071223000-00000000712ddfff 000000bb 748KiB UC WC WT WB
28718 persist 0000000880000000-0000000e7fffffff 00600000 24GiB UC WC WT WB NV
28719 reserved 0000001680000000-0000001c7fffffff 00600000 24GiB UC WC WT WB NV
28720
28721 2015-12-16 Andrei Borzenkov <[email protected]>
28722
28723 kernel: print and reset grub_errno after each embedded config line
28724 Otherwise it causes subsequent file open to fail, because grub_file_open
28725 misinterprets set grub_errno for grub_file_get_device_name failure.
28726
28727 Closes: 46540
28728
28729 2015-12-16 Andrei Borzenkov <[email protected]>
28730
28731 Erase backspaced character in grub_username_get
28732 It probably does not work across linefeed, but hopefully user names are not
28733 that long (and nobody is using terminal that small).
28734
28735 2015-12-16 Hector Marco-Gisbert <[email protected]>
28736
28737 Fix security issue when reading username and password
28738 This patch fixes two integer underflows at:
28739 * grub-core/lib/crypto.c
28740 * grub-core/normal/auth.c
28741
28742 CVE-2015-8370
28743
28744 Also-By: Andrey Borzenkov <[email protected]>
28745
28746 2015-12-15 Andrei Borzenkov <[email protected]>
28747
28748 NEWS: more additions
28749 Also-By: Robert Elliott <[email protected]>
28750
28751 2015-12-15 Robert Elliott <[email protected]>
28752
28753 Translate UEFI persistent memory type
28754 Define
28755 * GRUB_EFI_PERSISTENT_MEMORY (UEFI memory map type 14) per UEFI 2.5
28756 * GRUB_MEMORY_PERSISTENT (E820 type 7) per ACPI 3.0
28757 * GRUB_MEMORY_PERSISTENT_LEGACY (E820 unofficial type 12) per ACPI 3.0
28758
28759 and translate GRUB_EFI_PERSISTENT_MEMORY to GRUB_MEMORY_PERSISTENT in
28760 grub_efi_mmap_iterate().
28761
28762 Includes
28763 * adding the E820 names to lsmmap
28764 * handling the E820 types in make_efi_memtype()
28765
28766 Suggested-by: Vladimir 'φ-coder/phcoder' Serbinenko <[email protected]>
28767 Suggested-by: Andrei Borzenkov <[email protected]>
28768
28769 2015-12-14 Vladimir Serbinenko <[email protected]>
28770
28771 Document bootlocation discovery limitations and xen platform limitations
28772
28773 2015-12-07 Josef Bacik <[email protected]>
28774
28775 tcp: ack when we get an OOO/lost packet
28776 While adding tcp window scaling support I was finding that I'd get some packet
28777 loss or reordering when transferring from large distances and grub would just
28778 timeout. This is because we weren't ack'ing when we got our OOO packet, so the
28779 sender didn't know it needed to retransmit anything, so eventually it would fill
28780 the window and stop transmitting, and we'd time out. Fix this by ACK'ing when
28781 we don't find our next sequence numbered packet. With this fix I no longer time
28782 out. Thanks,
28783
28784 2015-12-01 Michael Chang <[email protected]>
28785
28786 i386: fix TSC calibration using PIT
28787 Condition was accidentally reversed, so PIT calibration always failed
28788 when PIT was present and always succeeded when PIT was missing, but in
28789 the latter case resulted in absurdly fast clock.
28790
28791 Reported and tested by Vitaly Kuznetsov <[email protected]>
28792
28793 2015-11-28 Andrei Borzenkov <[email protected]>
28794
28795 Do not include generated gnulib headers in tarball
28796 gnulib files are already handled by recursive make distdir invocation.
28797 Including all generated headers (after make completed) causes build
28798 failure if target system is different (different compile version etc).
28799
28800 2015-11-27 Andrei Borzenkov <[email protected]>
28801
28802 Replace numbers with grub_memory_type_t enums
28803
28804 2015-11-27 Andrei Borzenkov <[email protected]>
28805
28806 configure: fix macports flex version detection
28807 Macports add extra information after version itself:
28808
28809 $flex --version
28810 flex 2.5.35 Apple(flex-31)
28811
28812 We require at least felx 2.5.35 so do not need to care about prehistoric
28813 "flex version n.n.n"; just use second field always.
28814
28815 Reported by Peter Cheung <[email protected]>
28816
28817 2015-11-27 Vladimir Serbinenko <[email protected]>
28818
28819 tsc: Use alternative delay sources whenever appropriate.
28820 PIT isn't available on some of new hardware including Hyper-V. So
28821 use pmtimer for calibration. Moreover pmtimer calibration is faster, so
28822 use it on coreboor where booting time is important.
28823
28824 Based on patch by Michael Chang.
28825
28826 2015-11-26 Andrei Borzenkov <[email protected]>
28827
28828 efi: really mark memory of unknown type as reserved
28829 9be4c45dbe3c877d1f4856e99ee15133c6cd2261 added switch case between
28830 fall through cases, causing all memory regions of unknown type to be
28831 marked as available.
28832
28833 Move default case into its own block and add explicit FALLTHROUGH
28834 annotation.
28835
28836 Reported by Elliott, Robert (Persistent Memory) <[email protected]>
28837
28838 2015-11-24 Josef Bacik <[email protected]>
28839
28840 net: reset nb->data per dns record lookup loop
28841 We were resetting nb->data every time we tried a new server, but we need to do
28842 it every time we try for a different record, otherwise we don't end up falling
28843 back to the A record properly. Thanks,
28844
28845 2015-11-18 Andrei Borzenkov <[email protected]>
28846
28847 unix: do not close stdin in grub_passwd_get
28848 This makes it impossible to read from stdin without controlling tty:
28849
28850 10:/mnt # echo -e passwd\\npasswd | setsid ./grub-mkpasswd-pbkdf2
28851 Enter password:
28852 Reenter password: ./grub-mkpasswd-pbkdf2: error: failure to read password.
28853 10:/mnt
28854
28855 2015-11-17 Andrei Borzenkov <[email protected]>
28856
28857 lsefisystab: add missing comma after 7994077
28858
28859 2015-11-14 Pavel Bludov <[email protected]>
28860
28861 Add some UUIDs found in the hardware
28862
28863 2015-11-13 Konstantin Vlasov <[email protected]>
28864
28865 gfxterm: fix calculation of terminal-top and terminal-height
28866 They used screen width, not height.
28867
28868 2015-11-12 Paulo Flabiano Smorigo <[email protected]>
28869
28870 ofdisk: add sas disks to the device list
28871
28872 2015-11-12 Vladimir Serbinenko <[email protected]>
28873
28874 multiboot: Don't rely on particular ordering of options.
28875
28876 multiboot_mbi: Fix handling of --quirk-bad-kludge.
28877
28878 2015-11-12 Fu Wei <[email protected]>
28879
28880 xen_boot: Remove useless file_name_index variable.
28881
28882 Document ARM64 xen commands
28883
28884 2015-11-11 Vladimir Serbinenko <[email protected]>
28885
28886 asm-tests/i386-pc: Check that near jumps are 2 bytes.
28887 We already check that jump over 300 bytes gap is 3 bytes in code16-mode.
28888 Some clang versions generate 3-byte opcode for short jumps which makes
28889 boot.img blow over 512-byte limit. Enforce -no-integrated-as in such cases
28890
28891 2015-11-11 Paulo Flabiano Smorigo <[email protected]>
28892
28893 ofdisk: add a comment about vscsi method
28894
28895 2015-11-09 Vladimir Serbinenko <[email protected]>
28896
28897 fdt.mod: Move license tag to the right file.
28898
28899 2015-11-09 Fu Wei <[email protected]>
28900
28901 fdt.mod: Add missing license tag.
28902
28903 2015-11-09 Vladimir Serbinenko <[email protected]>
28904
28905 kern/elf: Ignore cast-align warnings
28906
28907 2015-11-09 Vladimir Serbinenko <[email protected]>
28908
28909 cbfs: Fix corner case and compilation with recdent gcc
28910 Accept the header to touch the jump address at 0xfffffff0.
28911
28912 Fix compilation for 64-bit EFI with recent GCC.
28913
28914 2015-11-08 Vladimir Serbinenko <[email protected]>
28915
28916 fstester: Enforce LC_ALL=C
28917
28918 Adapt build-system to use imported xen headers.
28919
28920 Import xen headers directly into GRUB
28921
28922 cbfs: Check for ptr range sanity.
28923 Triaged by Andrei and enhanced with suggestions by Aaron Durbin
28924 Also-By: Andrei Borzenkov <[email protected]>
28925
28926 Remove reliance C.UTF-8
28927
28928 genmoddep.awk: Add a test that we have no circular dependencies
28929
28930 Makefile.core.def: Break circular dependency on arm64.
28931
28932 autogen: Use cp instead of ln -s.
28933 libgcrypt-grub shouldn't be modified directly anyway. With this patch
28934 tarball without contrib can be unpacked on FAT and stay usable for
28935 out-of-tree compile on full POSIX FS (compile on FAT not tested).
28936
28937 2015-11-07 Andrei Borzenkov <[email protected]>
28938
28939 partmap_test: check that parted is available
28940 Skip test if parted is unavailable instead of returning false failure.
28941
28942 2015-11-07 [email protected] <[email protected]>
28943
28944 cryptodisk: teach grub_cryptodisk_insert() about partitions (bug #45889)
28945 It is not possible to configure encrypted containers on multiple partitions of
28946 the same disk; after the first one all subsequent fail with
28947
28948 disk/cryptodisk.c:978: already mounted as crypto0
28949
28950 Store partition offset in cryptomount descriptor to distinguish between them.
28951
28952 2015-11-07 Andrey Borzenkov <[email protected]>
28953
28954 doc: document config_directory and config_file variables
28955
28956 2015-11-07 Andrei Borzenkov <[email protected]>
28957
28958 unix/getroot: remove unused MAJOR definition
28959 We use major() everywhere, these definitions just add to confusion.
28960
28961 Add comments to code for commit d313218
28962
28963 2015-11-07 Andrei Borzenkov <[email protected]>
28964
28965 devmapper/getroot: use makedev instead of direct shift
28966 Fixes device detection with large number of devices.
28967
28968 Reported by Tim Wallberg <[email protected]>
28969
28970 2015-11-06 Andrei Borzenkov <[email protected]>
28971
28972 mkimage: zero fill alignment space
28973 This did not cause real problem but is good for reproducible builds. I hit
28974 it with recent bootinfoscript that displays embedded config; I was puzzled
28975 by random garbage at the end.
28976
28977 Prezero memory buffer used to assemble core.img. This makes individual
28978 memset redundant. Also ensure buffer is filled with zeroes in several other
28979 places.
28980
28981 Also remove redundant zeroing code where we fill in the whole memory block
28982 anyway.
28983
28984 2015-11-06 Vladimir Serbinenko <[email protected]>
28985
28986 configure.ac: Explicitly add -mno-sse3 on x86.
28987
28988 README: Remove dead link to the wiki
28989
28990 2015-10-29 Andrei Borzenkov <[email protected]>
28991
28992 NEWS: mention powerpc64le support
28993
28994 2015-10-29 Ignat Korchagin <ignat>
28995
28996 tcp: Fix uninited mac address when accepting connection.
28997
28998 2015-10-29 Fu Wei <[email protected]>
28999
29000 arm64: Add support for xen boot protocol.
29001
29002 2015-10-29 Vladimir Serbinenko <[email protected]>
29003
29004 arm64: Move FDT functions to separate module
29005
29006 2015-10-27 Andrei Borzenkov <[email protected]>
29007
29008 efi: fix warnings with recent GCC
29009 ../../grub-core/term/efi/console.c:128:32: error: suggest parentheses around '&&' within '||' [-Werror=parentheses]
29010 if (key.unicode_char >= 0x20 && key.unicode_char <= 0x7f
29011
29012 2015-10-26 Eric Snowberg <[email protected]>
29013
29014 ofdisk: Fix devpath freeing logic.
29015
29016 2015-10-26 Paulo Flabiano Smorigo <[email protected]>
29017
29018 Implement cross-endian ELF load for powerpc
29019
29020 2015-10-25 Peter Jones <[email protected]>
29021
29022 Use EFI_SIMPLE_TEXT_INPUT_EX to support key combinations.
29023
29024 2015-10-14 Andrei Borzenkov <[email protected]>
29025
29026 configure: find options to force endian on MIPS
29027
29028 2015-10-14 Andrei Borzenkov <[email protected]>
29029
29030 configure: force o32 ABI on MIPS
29031 GRUB code expects O32 or N32. N32 is less tested than O32, so we prefer to
29032 compile with O32. Some systems (e.g. GNU Guix) default to using newer
29033 n64 or n32 ABI. Try to find suitable options to force o32.
29034
29035 For GCC this is simply -mabi=32. While clang supports this option as well,
29036 o32 ABI is valid for MIPS target and n32/64 ABI are valid for MIPS64 target
29037 only, so use "-target mips/mipsel -mabi=32".
29038
29039 Reported-By: Mark H Weaver <[email protected]>
29040 Also-By: Mark H Weaver <[email protected]>
29041
29042 2015-10-12 Andrei Borzenkov <[email protected]>
29043
29044 net: avoid closing NULL socket in DNS lookup
29045 Refactor code so that we do not store NULL pointers in array
29046 of in-flight DNS servers.
29047
29048 Reported-By: Josef Bacik <[email protected]>
29049
29050 2015-10-11 Andrei Borzenkov <[email protected]>
29051
29052 install: --compress argument is not optional
29053 Fixes crash if argument is not specified. Also use `|' to separate choices
29054 in list of compression methods to align it with --core-compress.
29055
29056 2015-10-11 Vladimir Serbinenko <[email protected]>
29057
29058 mips: Make setjmp code N32-compliant.
29059
29060 mips: Make the assembly-code N32-compatible.
29061 There are no $t4 or $t5 in N32 but there are $a4 and $a5.
29062
29063 2015-10-10 Andrei Borzenkov <[email protected]>
29064
29065 progress: avoid NULL dereference for net files
29066 From original patch by dann frazier <[email protected]>:
29067
29068 grub_net_fs_open() saves off a copy of the file structure it gets passed and
29069 uses it to create a bufio structure. It then overwrites the passed in file
29070 structure with this new bufio structure. Since file->name doesn't get set
29071 until we return back to grub_file_open(), it means that only the bufio
29072 structure gets a valid file->name. The "real" file's name is left
29073 uninitialized. This leads to a crash when the progress module hook is called
29074 on it.
29075
29076 grub_net_fs_open() already saved copy of file name as ->net->name, so change
29077 progress module to use it.
29078
29079 Also, grub_file_open may leave file->name as NULL if grub_strdup fails. Check
29080 for it.
29081
29082 Also-By: dann frazier <[email protected]>
29083
29084 2015-10-10 Andrei Borzenkov <[email protected]>
29085
29086 file: ignore host disk in blocklist check
29087 It cannot work anyway because host disk cannot be read. This fixes hostfs access
29088 on native Windows build where filenames start with '\' or do not have initial
29089 separator at all (d:\foo).
29090
29091 Issue was observed when running grub-fstest on Windows. On UNIX image name is
29092 canonicalized to always start with `/' so this was not noticed.
29093
29094 This has side effect of allowing relative path names on host, but this already
29095 was the case with `ls' command, so it just extends it to all commands.
29096
29097 Reported-By: Arch Stack <[email protected]>
29098 Also-By: Arch Stack <[email protected]>
29099
29100 2015-10-09 Vladimir Serbinenko <[email protected]>
29101
29102 mips/dl: Handle addend in RELA entries.
29103
29104 gfxmenu/model: Delete empty file.
29105
29106 2015-10-09 Alexander Bluhm <[email protected]>
29107
29108 ufs: Fix parameters to grub_memset.
29109 len = 0 made simply no sense. Fix parameters to be in line with read.
29110
29111 2015-10-07 Stanislav Kholmanskikh <[email protected]>
29112
29113 ofnet: Do not set SUFFIX for sun4v network devices
29114 sun4v vnet devices do not implement the support of duplex and speed
29115 instance attributes. An attempt to open such a device with
29116 the attributes will fail:
29117
29118 ok select net:speed=auto,duplex=auto
29119 Unknown key 'speed'
29120 Unknown key 'duplex'
29121 Manual Configuration: Host IP, boot server and filename must be specified
29122 WARNING: /virtual-devices@100/channel-devices@200/network@0: Can't open OBP standard TFTP package
29123
29124 Can't open device
29125 ok
29126
29127 Therefore, let's not set SUFFIX for such devices.
29128
29129 2015-10-07 Eric Snowberg <[email protected]>
29130
29131 sparc64 - use correct drive name within grub_util_sparc_setup
29132 Incorrect drive name was being passed into grub_util_sparc_setup,
29133 causing the grub-install to fail.
29134
29135 2015-09-13 Andrei Borzenkov <[email protected]>
29136
29137 cryptodisk: strip parenthesis from backing device name
29138 Otherwise subsequent disk open fails.
29139
29140 Reported-By: Klemens Nanni <[email protected]>
29141
29142 2015-08-22 Felix Zielcke <[email protected]>
29143
29144 disk/ldm, partmap/msdos.c: fix spelling error
29145
29146 2015-08-13 Andrei Borzenkov <[email protected]>
29147
29148 net: do not try to load protocol module via itself
29149 Otherwise we get infinite recursion.
29150
29151 Closes: 45729
29152
29153 2015-08-09 Josef Bacik <[email protected]>
29154
29155 efinet: handle get_status() on buggy firmware properly
29156 The EFI spec indicates that get_status() should return the address of the buffer
29157 we passed into transmit to indicate the the buffer was transmitted. However we
29158 have boxes where the firmware returns some arbitrary address instead, which
29159 makes grub think that we've not sent anything. So since we have the SNP stuff
29160 opened in exclusive mode just assume any non-NULL txbuf means that our transmit
29161 occurred properly. This makes grub able to do its networking stuff properly on
29162 our broken firmware. Thanks,
29163
29164 cc: Peter Jones <[email protected]>
29165
29166 2015-08-09 Andrei Borzenkov <[email protected]>
29167
29168 linguas.sh: fix error when removing non-existing autogenerated files
29169
29170 2015-07-28 Vladimir Serbinenko <[email protected]>
29171
29172 ahci: Ensure that bus mastering is set.
29173 Fixes ahci_test failing on several platforms.
29174
29175 2015-07-27 Vladimir Serbinenko <[email protected]>
29176
29177 archelp: Never pass NULL as mtime.
29178 Moves complexity from fs code (NULL check) to common code (passing non-NULL).
29179
29180 HFS: Convert to fshelp.
29181 HFS doesn't handle "." and ".." properly. Convert it to fshelp to reuse the
29182 logic.
29183
29184 FAT: Convert to fshelp.
29185 exFAT doesn't handle "." and ".." correctly, convert it to fshelp to
29186 reuse the same logic.
29187
29188 BFS: Convert to fshelp.
29189 BFS doesn't handle ".." correctly, so convert it to fshelp to reuse the logic.
29190
29191 fshelp: Add handling of "." and ".." and grub_fshelp_find_file_lookup.
29192 Recent tests have discovered that many of our filesystems have flawed
29193 handling of "." and "..". Rather than attempting to fix it in filesystems
29194 themselves, make the common code fshelp aware of "." and ".." and handle
29195 them in this layer. Add grub_fshelp_find_file_lookup for easy conversion
29196 of BFS, HFS and exFAT which have the same problem and don't use fshelp.
29197
29198 Switch procfs to use archelp.
29199 This fixes handling of "." and "..".
29200
29201 grub-install: Use a+ in fopen rather than r+.
29202 r+ does not create a file if none exists.
29203
29204 Add transform_data as a variant of data with substitutions.
29205 This fixrs name mismatch for grub.chrp with
29206 transform_program_name='s,grub,grub2,g'
29207
29208 2015-07-24 Ignat Korchagin <[email protected]>
29209
29210 efi: fix GetVariable return status check in 81ca24a
29211 GetVariable should return EFI_BUFFER_TOO_SMALL if given buffer of size
29212 zero; commit incorrectly checked for EFI_SUCCESS.
29213
29214 2015-07-24 Vladimir Serbinenko <[email protected]>
29215
29216 zfs_test: Skip dotdot in volume root test.
29217 Given special semantics of ZFS it's far from clear what the expected
29218 result is. Just skip it for now
29219
29220 xfs_test: Test both crc and non-crc filesystems.
29221
29222 xfs: Fix handling of symlink with crc-enabled filesystem.
29223
29224 reiserfs: Fix handling of first entry in the directory.
29225 Fixes garbage being added to "." filename.
29226
29227 2015-07-23 Ignat Korchagin <[email protected]>
29228
29229 efi: fix memory leak in variable handling
29230
29231 2015-07-23 Vladimir Serbinenko <[email protected]>
29232
29233 exclude.pot: Add missing blacklisted strings.
29234
29235 archelp: Fix handling of dot and dotdot at the end of the name.
29236 Fixes cpio_test and tar_test.
29237
29238 arm-emu: Add __aeabi_memcpy* and __aeabi_memclr* symbols.
29239 Fixes compilation with clang.
29240
29241 2015-07-22 Vladimir Serbinenko <[email protected]>
29242
29243 fwstart: Fix loading of address of read_spd_fail.
29244
29245 fwstart: Add missing argument to p2align.
29246 Resulting binary is unchanged as it happens we were already aligned
29247 by chance.
29248
29249 2015-07-22 Vladimir Serbinenko <[email protected]>
29250
29251 fwstart: Replace blt with bltz.
29252 blt A, $zero, B and bltz A, B are equivalent but clang recognizes only
29253 later, so use it.
29254
29255 Resulting binary is unchanged.
29256
29257 2015-07-22 Vladimir Serbinenko <[email protected]>
29258
29259 Remove mips_attributes.
29260 mips_attributes was introduced to work around clang problems with
29261 -msoft-float. Those problems are now fixed and moreover .gnu_attributes
29262 itself is unportable and creates problem with clang.
29263
29264 Revert "mips: Fix soft-float handling."
29265
29266 This partially reverts commit 6a4ecd276ed39f66be0ad6ff0f8ff67598098605.
29267
29268 2015-07-22 Vladimir Serbinenko <[email protected]>
29269
29270 ARM: provide __aeabi_memclr* and __aeabi_memcpy* symbols
29271 Fixes compilation with recent clang.
29272
29273 diskfilter: Make name a const char to fix compilation error.
29274
29275 dmraid_nvidia: Set a name to usable value to avoid null dereference.
29276 Reported by: Andrei Borzenkov
29277
29278 configure.ac: Handle powerpc64le compiler
29279 Also-by: Paulo Flabiano Smorigo <[email protected]>
29280
29281 2015-07-20 Bernhard Übelacker <[email protected]>
29282
29283 loader/linux: Make trailer initrd entry aligned again.
29284 Regression from commit:
29285 loader/linux: do not pad initrd with zeroes at the end
29286 a8c473288d3f0a5e17a903a5121dea1a695dda3b
29287
29288 Wimboot fails since the change above because it expects the "trailer"
29289 initrd element on an aligned address.
29290 This issue shows only when newc_name is used and the last initrd
29291 entry has a not aligned size.
29292
29293 2015-07-16 Vladimir Serbinenko <[email protected]>
29294
29295 XFS: Fix wrong alignment treatment.
29296
29297 grub_ext2_read_block: Fix return type on error.
29298
29299 2015-07-05 Andrei Borzenkov <[email protected]>
29300
29301 use TARGET_LDFLAGS in grub_PROG_OBJCOPY_ABSOLUTE
29302 That's what Makefile will use and it is required if unusual flags
29303 must be passed to linker (e.g. to build ppc32 code on ppc64le with clang).
29304
29305 2015-06-26 Michael Chang <[email protected]>
29306
29307 Fix missing byte order conversion in get_btrfs_fs_prefix function
29308 Since btrfs on-disk format uses little-endian, the searched item types
29309 (ROOT_REF, INODE_REF) need converting the byte order in order to
29310 function properly on big-endian systems.
29311
29312 2015-06-26 Andrei Borzenkov <[email protected]>
29313
29314 chainloader: fix resoource leak
29315 Found by: Coverity scan.
29316 CID: 96651
29317
29318 loader/bsd: fix memory leak
29319 Found by: Coverity scan.
29320 CID: 96662, 96665
29321
29322 2015-06-20 Andrei Borzenkov <[email protected]>
29323
29324 loader/bsd: free memory leaks
29325 Found by: Coverity scan.
29326 CID: 96671, 96658, 96653
29327
29328 search_wrap: fix memory leak
29329 Found by: Coverity scan.
29330 CID: 96675
29331
29332 password_pbkdf2: fix memory leak
29333 Found by: Coverity scan.
29334 CID: 96676
29335
29336 normal: fix memory leak
29337 Found by: Coverity scan.
29338 CID: 96677
29339
29340 efi/serial: fix memory leak
29341 Found by: Coverity scan.
29342 CID: 96678
29343
29344 ohci: fix memory leak
29345 Found by: Coverity scan.
29346 CID: 96679
29347
29348 loader/bsd: free memory leaks
29349 Found by: Coverity scan.
29350 CID: 96682
29351
29352 multiboot: fix memory leak
29353 Found by: Coverity scan.
29354 CID: 96684
29355
29356 normal: fix memory leak
29357 Found by: Coverity scan.
29358 CID: 96685
29359
29360 loader/bsd: fix memory leak
29361 Found by: Coverity scan.
29362 CID: 96686
29363
29364 reed_solomon: fix memory leak
29365 Found by: Coverity scan.
29366 CID: 96688
29367
29368 usb: fix use after free
29369 Found by: Coverity scan.
29370 CID: 96704
29371
29372 xnu: fix use after free
29373 Found by: Coverity scan.
29374 CID: 96706
29375
29376 disk/scsi: fix use after free
29377 Found by: Coverity scan.
29378 CID: 96713
29379
29380 efi/chainloader: fix use after free
29381 Found by: Coverity scan.
29382 CID: 96714
29383
29384 search: fix use after free
29385 Found by: Coverity scan.
29386 CID: 96715
29387
29388 NEWS: emu libusb support removed
29389
29390 2015-06-19 Andrei Borzenkov <[email protected]>
29391
29392 grub-probe: fix memory leak in probe (ofpath)
29393 Found by: Coverity scan.
29394 CID: 73772
29395
29396 2015-06-19 Andrei Borzenkov <[email protected]>
29397
29398 grub-probe: restructure code to make static analysis easier
29399 Current code in probe() could not be verified to not contain memory leaks.
29400 Restructure code and ensure grub_device_close is always called at the end of
29401 loop.
29402
29403 Calms down Coverity scan.
29404 CID: 73739
29405
29406 2015-06-19 Andrei Borzenkov <[email protected]>
29407
29408 zfs: fix memory leak
29409 Found by: Coverity scan.
29410 CID: 73647
29411
29412 xfs: silence Coverity overflow warning
29413 inode size cannot really overflow integer, but Coverity does not know it.
29414 CID: 96602
29415
29416 zfs: memory leak
29417 Found by Coverity scan.
29418 CID: 96603
29419
29420 unix/getroot: memory leak
29421 Found by Coverity scan.
29422 CID: 96605
29423
29424 unix/relpath: memory leak
29425 Found by Coverity scan.
29426 CID: 96606
29427
29428 2015-06-19 Andrei Borzenkov <[email protected]>
29429
29430 syslinux_parse: assorted issues found by Coverity
29431 1. Remove unneeded NULL check
29432 CID: 96607
29433
29434 2. Do not allocate storage for initrd, copy it directly from input
29435 buffer. Avoids memory leak in failure path.
29436 CID: 96604
29437
29438 3. Unchecked error return from print()
29439 CID: 96601, 73595
29440
29441 2015-06-19 Andrei Borzenkov <[email protected]>
29442
29443 syslinux_parse: make print_escaped actually stop before `to'
29444 The only current user is mboot.c32 which unfortunately is not covered
29445 by regression tests.
29446
29447 2015-06-18 Andrei Borzenkov <[email protected]>
29448
29449 fat: fix handling of "." and ".." directory entries
29450 Emulate dot and dotdot in root directory. For other directories do not
29451 add separator between name and extension for these two special entries.
29452
29453 Closes: 45335
29454
29455 2015-06-18 Andrei Borzenkov <[email protected]>
29456
29457 tests: regression tests for "." and ".." directory entries
29458
29459 2015-06-16 Andrei Borzenkov <[email protected]>
29460
29461 efinet: enable hardware filters when opening interface
29462 Exclusive open on SNP will close all existing protocol instances which
29463 may disable all receive filters on interface. Reinstall them after we
29464 opened protocol exclusively.
29465
29466 Also follow UEFI specification recommendation and stop interfaces when
29467 closing them:
29468
29469 Unexpected system errors, reboots and hangs can occur if an OS is loaded
29470 and the network devices are not Shutdown() and Stopped().
29471
29472 Also by: Mark Salter <[email protected]>
29473 Closes: 45204
29474
29475 2015-06-16 Andrei Borzenkov <[email protected]>
29476
29477 NEWS: mention libgcc removal
29478
29479 2015-06-15 Paulo Flabiano Smorigo <[email protected]>
29480
29481 Add flag for powerpc ieee1275 to avoid unneeded optimizations
29482
29483 2015-06-12 Mark Salter <[email protected]>
29484
29485 Fix exit to EFI firmware
29486 The current code for EFI grub_exit() calls grub_efi_fini() before
29487 returning to firmware. In the case of ARM, this leaves a timer
29488 event running which could lead to a firmware crash. This patch
29489 changes this so that grub_machine_fini() is called with a NORETURN
29490 flag. This allows machine-specific shutdown to happen as well
29491 as the shutdown done by grub_efi_fini().
29492
29493 2015-06-12 Paul Menzel <[email protected]>
29494
29495 disk/ahci.c: Use defines `GRUB_AHCI_HBA_PORT_CMD_SPIN_UP` and `GRUB_AHCI_HBA_PORT_CMD_POWER_ON`
29496 Instead of hard coding `2` and `4` use the macros defined already at the
29497 top of the file. As a consequence, wrap the now too long line.
29498
29499 2015-06-12 Andrei Borzenkov <[email protected]>
29500
29501 NEWS: XFS v5 support
29502
29503 2015-06-12 Jan Kara <[email protected]>
29504
29505 xfs: V5 filesystem format support
29506 Add support for new XFS on disk format. We have to handle optional
29507 filetype fields in directory entries, additional CRC, LSN, UUID entries
29508 in some structures, etc.
29509
29510 xfs: Add helpers for inode size
29511 Add helpers to return size of XFS inode on disk and when loaded in
29512 memory.
29513
29514 2015-06-04 Toomas Soome <[email protected]>
29515
29516 multiboot_header_tag_module_align fix to confirm multiboot specification
29517
29518 2015-06-02 Leif Lindholm <[email protected]>
29519
29520 configure.ac: clean up arm64 soft-float handling
29521 Fix compilation with gcc 5.1 (avoid internal compiler error), by
29522 replacing explicit -march +nofp+nosimd options with -mgeneral-regs-only.
29523
29524 This also enables the removal of some further conditional build flag
29525 setting.
29526
29527 2015-06-01 dann frazier <[email protected]>
29528
29529 arm64/setjmp: Add missing license macro
29530 Including the setjmp module in an arm64-efi image will cause it to
29531 immediately exit with an "incompatible license" error.
29532
29533 The source file includes a GPLv3+ boilerplate, so fix this by declaring a
29534 GPLv3+ license using the GRUB_MOD_LICENSE macro.
29535
29536 2015-05-31 Paul Menzel <[email protected]>
29537
29538 disk/ahci.c: Add port number to port debug messages
29539 Currently, some messages cannot be mapped to the port they belong to as
29540 the port number is missing from the output. So add `port: n` to the
29541 debug messages.
29542
29543 2015-05-30 Andrei Borzenkov <[email protected]>
29544
29545 Clarify use of superusers variable and menu entry access
29546 superusers controls both CLI and editing. Also explicitly mention that
29547 empty superusers disables them.
29548
29549 "Access to menuentry" is a bit vague - change to "execute menuentry"
29550 to make it obvious, what access is granted.
29551
29552 2015-05-30 Paul Menzel <[email protected]>
29553
29554 Correct spelling of *scheduled*
29555 Run the command below
29556
29557 $ git grep -l schedulded | xargs sed -i 's/schedulded/scheduled/g'
29558
29559 and revert the change in `ChangeLog-2015`.
29560
29561 Including "miscellaneous" spelling fix noted by [email protected]
29562
29563 2015-05-30 Toomas Soome <[email protected]>
29564
29565 zfs extensible_dataset and large_blocks feature support
29566 large blocks basically use extensible dataset feature, or to be exact,
29567 setting recordsize above 128k will trigger large_block feature to be
29568 enabled and storing such blocks is using feature extensible dataset. so
29569 the extensible dataset is prerequisite.
29570
29571 Changes implement read support extensible dataset… instead of fixed DMU
29572 types they dont specify type, making it possible to use fat zap objects
29573 from bonus area.
29574
29575 2015-05-27 Vladimir Serbinenko <[email protected]>
29576
29577 multiboot1: never place modules in low memory.
29578 While in theory permitted by the spec, modules rarely fit in low memory
29579 anyway and not every kernel is able to handle modules in low memory anyway.
29580 At least VMWare is known not to be able to handle modules at arbitrary
29581 locations.
29582
29583 2015-05-24 Paul Menzel <[email protected]>
29584
29585 disk/ahci: Use defines `GRUB_ATA_STATUS_BUSY` and `GRUB_ATA_STATUS_DRQ`
29586 Instead of hard coding `0x88` use the macros defined in `disk/ata.h`.
29587
29588 2015-05-19 Paul Menzel <[email protected]>
29589
29590 cb_timestamps.c: Add new time stamp descriptions
29591 Add the descriptions of the “core”, that means no vendorcode or payload,
29592 coreboot time stamps added up to coreboot commit a7d92441 (timestamps:
29593 You can never have enough of them!) [1].
29594
29595 Running `coreboot_boottime` in the GRUB command line interface now shows
29596 descriptions for all time stamps again on the ASRock E350M1.
29597
29598 [1] http://review.coreboot.org/9608
29599
29600 2015-05-17 Andrei Borzenkov <[email protected]>
29601
29602 bootp: ignore gateway_ip (relay) field.
29603 From RFC1542:
29604
29605 The 'giaddr' field is rather poorly named. It exists to facilitate
29606 the transfer of BOOTREQUEST messages from a client, through BOOTP
29607 relay agents, to servers on different networks than the client.
29608 Similarly, it facilitates the delivery of BOOTREPLY messages from the
29609 servers, through BOOTP relay agents, back to the client. In no case
29610 does it represent a general IP router to be used by the client. A
29611 BOOTP client MUST set the 'giaddr' field to zero (0.0.0.0) in all
29612 BOOTREQUEST messages it generates.
29613
29614 A BOOTP client MUST NOT interpret the 'giaddr' field of a BOOTREPLY
29615 message to be the IP address of an IP router. A BOOTP client SHOULD
29616 completely ignore the contents of the 'giaddr' field in BOOTREPLY
29617 messages.
29618
29619 Leave code ifdef'd out for the time being in case we see regression.
29620
29621 Suggested by: Rink Springer <[email protected]>
29622 Closes: 43396
29623
29624 2015-05-17 Andrei Borzenkov <[email protected]>
29625
29626 hostdisk: fix crash with NULL device.map
29627 grub-macbless calls grub_util_biosdisk_init with NULL device.map.
29628
29629 2015-05-14 Andrei Borzenkov <[email protected]>
29630
29631 zfs: fix integer truncation in zap_lookup
29632 Size after shift could exceed 16 bits; use grub_unit32_t for result.
29633
29634 Reported and tested by: Kostya Berger <[email protected]>
29635 Closes: 44448
29636
29637 2015-05-13 Andrei Borzenkov <[email protected]>
29638
29639 remove extra newlines in grub_util_* strings
29640 grub_util_{info,warn,error} already add trailing newlines, so remove
29641 them from format strings. Also trailing full stops are already added.
29642
29643 2015-05-12 Jan Kara <[email protected]>
29644
29645 xfs: Convert inode numbers to cpu endianity immediately after reading
29646 Currently XFS driver converted inode numbers to native endianity only
29647 when using them to compute inode position. Although this works, it is
29648 somewhat confusing. So convert inode numbers when reading them from disk
29649 structures as every other field.
29650
29651 2015-05-11 Jan Kara <[email protected]>
29652
29653 xfs: Fix termination loop for directory iteration
29654 Directory iteration used wrong position (sizeof wrong structure) for
29655 termination of iteration inside a directory block. Luckily the position
29656 ended up being wrong by just 1 byte and directory entries are larger so
29657 things worked out fine in practice. But fix the problem anyway.
29658
29659 2015-05-08 Andrei Borzenkov <[email protected]>
29660
29661 acpi: do not skip BIOS scan if EBDA length is zero
29662 EBDA layout is not standardized so we cannot assume first two bytes
29663 are length. Neither is it required by ACPI standard. HP 8710W is known
29664 to contain zeroes here.
29665
29666 Closes: 45002
29667
29668 2015-05-07 Andrei Borzenkov <[email protected]>
29669
29670 Add asm-tests to tarball
29671
29672 2015-05-07 Vladimir Serbinenko <[email protected]>
29673
29674 util/grub-mkrescue: Fix compilation
29675
29676 2015-05-07 Andrei Borzenkov <[email protected]>
29677
29678 efinet: open Simple Network Protocol exclusively
29679 EDK2 network stack is based on Managed Network Protocol which is layered
29680 on top of Simple Management Protocol and does background polling. This
29681 polling races with grub for received (and probably trasmitted) packets
29682 which causes either serious slowdown or complete failure to load files.
29683
29684 Open SNP device exclusively. This destroys all child MNP instances and
29685 stops background polling.
29686
29687 Exclusive open cannot be done when enumerating cards, as it would destroy
29688 PXE information we need to autoconfigure interface; and it cannot be done
29689 during autoconfiguration as we need to do it for non-PXE boot as well. So
29690 move SNP open to card ->open method and add matching ->close to clean up.
29691
29692 Based on patch from Mark Salter <[email protected]>
29693
29694 Also-By: Mark Salter <[email protected]>
29695 Closes: 41731
29696
29697 2015-05-07 Andrei Borzenkov <[email protected]>
29698
29699 efinet: skip virtual IPv4 and IPv6 devices when enumerating cards
29700 EDK2 PXE driver creates two child devices - IPv4 and IPv6 - with
29701 bound SNP instance. This means we get three cards for every physical
29702 adapter when enumerating. Not only is this confusing, this may result
29703 in grub ignoring packets that come in via the "wrong" card.
29704
29705 Example of device hierarchy is
29706
29707 Ctrl[91] PciRoot(0x0)/Pci(0x3,0x0)
29708 Ctrl[95] PciRoot(0x0)/Pci(0x3,0x0)/MAC(525400123456,0x1)
29709 Ctrl[B4] PciRoot(0x0)/Pci(0x3,0x0)/MAC(525400123456,0x1)/IPv4(0.0.0.0)
29710 Ctrl[BC] PciRoot(0x0)/Pci(0x3,0x0)/MAC(525400123456,0x1)/IPv6(0000:0000:0000:0000:0000:0000:0000:0000)
29711
29712 Skip PXE created virtual devices when enumerating cards. Make sure to
29713 find real card when applying initial autoconfiguration during PXE boot,
29714 this information is associated with one of child devices.
29715
29716 2015-05-07 Andrei Borzenkov <[email protected]>
29717
29718 efidisk: move device path helpers in core for efinet
29719
29720 convert to, not from, CPU byte order in DNS receive function
29721
29722 2015-05-07 Andrei Borzenkov <[email protected]>
29723
29724 loader/linux: do not pad initrd with zeroes at the end
29725 Syslinux memdisk is using initrd image and needs to know uncompressed
29726 size in advance. For gzip uncompressed size is at the end of compressed
29727 stream. Grub padded each input file to 4 bytes at the end, which means
29728 syslinux got wrong size.
29729
29730 Linux initramfs loader apparently does not care about trailing alignment.
29731 So change code to align beginning of each file instead which atomatically
29732 gives us the correct size for single file.
29733
29734 Reported-By: David Shaw <[email protected]>
29735
29736 2015-05-07 Daniel Kiper <[email protected]>
29737
29738 i386/relocator: Remove unused extern grub_relocator64_rip_addr
29739
29740 2015-05-07 Vladimir Serbinenko <[email protected]>
29741
29742 grub-install-common: Increase buf size to 8192 as modinfo.sh is bigger.
29743
29744 2015-05-07 Vladimir Serbinenko <[email protected]>
29745
29746 grub-mkrescue: Recognize -output as an alias of --output.
29747 This helps us to be in line with xorriso -as mkisofs.
29748
29749 Suggested by: Thomas Schmitt
29750
29751 2015-05-07 Vladimir Serbinenko <[email protected]>
29752
29753 linux.c: Ensure that initrd is page-aligned.
29754
29755 Revert parts accidentally committed 2 commits ago.
29756
29757 2015-05-07 Fu Wei <[email protected]>
29758
29759 fdt.h: Add grub_fdt_set_reg64 macro
29760
29761 arm64: Export useful functions from linux.c
29762
29763 2015-05-04 Andrei Borzenkov <[email protected]>
29764
29765 Revert "efinet: memory leak on module removal"
29766 This reverts commits 47b2bee3ef0ea60fc3f5bfc37f3784e559385297
29767 and 8d3c4544ffdd0289a4b0bdeb0cdc6355f801a4b3. It is not safe
29768 to free allocated cards, dangling pointers main remain. Such
29769 cleanup requires more changes in net core.
29770
29771 efinet: cannot free const char * pointer
29772
29773 efinet: memory leak on module removal
29774
29775 2015-05-03 Andrei Borzenkov <[email protected]>
29776
29777 zfs: add missing NULL check and fix incorrect buffer overwrite
29778 grub_memset should zero out padding after data end. It is not clear
29779 why it is needed at all - ZFS block is at least 512 bytes and power
29780 of two, so it is always multiple of 16 bytes. This grub_memset
29781 apparently never did anything.
29782
29783 2015-05-03 Toomas Soome <[email protected]>
29784
29785 zfs: com.delphix:embedded_data feature support
29786
29787 zfs: com.delphix:hole_birth feature support
29788 In the past birth was always zero for holes. This feature started
29789 to make use of birth for holes as well, so change code to test for
29790 valid DVA address instead.
29791
29792 2015-04-29 Andrei Borzenkov <[email protected]>
29793
29794 grub-mkconfig: use $pkgdatadir in scripts
29795 Otherwise scripts will source wrong grub-mkconfig_lib.
29796
29797 2015-04-24 Vladimir Serbinenko <[email protected]>
29798
29799 Remove -V in grub-mkrescue.c
29800 It clashhes with -V which is alias to -volid.
29801
29802 2015-04-13 Toomas Soome <[email protected]>
29803
29804 getroot: include sys/mkdev.h for makedev
29805 Solaris (like) systems need to include sys/mkdev.h for makedev() function.
29806
29807 2015-04-13 Toomas Soome <[email protected]>
29808
29809 core/partmap: rename 'sun' to avoid clash with predefined symbol
29810 the symbol “sun” is defined macro in solaris derived systems, from
29811 gcc -dM -E:
29812
29813 and therefore can not be used as name.
29814
29815 2015-04-12 Paul Menzel <[email protected]>
29816
29817 docs/grub.texi: Fix spelling of cbfstool
29818
29819 2015-04-06 Andrei Borzenkov <[email protected]>
29820
29821 core: avoid NULL derefrence in grub_divmod64s
29822 It can be called with NULL for third argument. grub_divmod32* for
29823 now are called only from within wrappers, so skip check.
29824
29825 Reported-By: Michael Zimmermann <[email protected]>
29826
29827 2015-03-28 Andrei Borzenkov <[email protected]>
29828
29829 do not emit cryptomount without crypto UUID
29830
29831 2015-03-28 Sarah Newman <[email protected]>
29832
29833 grub-core/loader/i386/xen.c: Initialized initrd_ctx so we don't free a random pointer from the stack.
29834
29835 2015-03-27 Andrei Borzenkov <[email protected]>
29836
29837 net: trivial grub_cpu_to_XX_compile_time cleanup
29838
29839 2015-03-27 Lunar <[email protected]>
29840
29841 syslinux: Support {vesa,}menu.c32.
29842
29843 2015-03-27 Steve McIntyre <[email protected]>
29844
29845 Recognize EFI platform even in case of mismatch between Linux and EFI.
29846 Some x86 systems might be capable of running a 64-bit Linux kernel but
29847 only use a 32-bit EFI (e.g. Intel Bay Trail systems). It's useful for
29848 grub-install to be able to recognise such systems, to set the default
29849 x86 platform correctly.
29850
29851 To allow grub-install to know the size of the firmware rather than
29852 just the size of the kernel, there is now an extra EFI sysfs file to
29853 describe the underlying firmware. Read that if possible, otherwise
29854 fall back to the kernel type as before.
29855
29856 2015-03-27 Michael Zimmermann <[email protected]>
29857
29858 Add missing initializers to silence suprious warnings.
29859
29860 2015-03-27 Leif Lindholm <[email protected]>
29861
29862 dl_helper: Cleanup
29863 Use the new thumb_get_instruction_word/thumb_set_instruction_word
29864 helpers throughout.
29865
29866 Style cleanup (missing spaces).
29867
29868 Move Thumb MOVW/MOVT handlers into Thumb relocation section of file.
29869
29870 2015-03-27 Martin Wilck <[email protected]>
29871
29872 efinet: Check for immediate completition.
29873 This both speeds GRUB up and workarounds unexpected EFI behaviour.
29874
29875 2015-03-27 Vladimir Serbinenko <[email protected]>
29876
29877 Make Makefile.util.def independent of platform.
29878
29879 2015-03-27 Daniel Kahn Gillmor <[email protected]>
29880
29881 util/mkimage: Use stable timestamp when generating binaries.
29882
29883 2015-03-27 Vladimir Serbinenko <[email protected]>
29884
29885 modinfo.sh.in: Add missing config variables.
29886
29887 Makefile.core.def: Remove obsolete LDADD_KERNEL
29888
29889 arp, icmp: Fix handling in case of oversized or invalid packets.
29890 This restrict ARP handling to MAC and IP addresses but in practice we need
29891 only this case anyway and other cases are very rar if exist at all. It makes
29892 code much simpler and less error-prone.
29893
29894 2015-03-23 Colin Watson <[email protected]>
29895
29896 hostfs: Drop unnecessary feature test macros
29897 _BSD_SOURCE was added to allow the use of DT_DIR, but that was removed
29898 in e768b77068a0b030a07576852bd0f121c9a077eb. While adding
29899 _DEFAULT_SOURCE as well works around problems with current glibc,
29900 neither is in fact needed nowadays.
29901
29902 2015-03-20 Vladimir Serbinenko <[email protected]>
29903
29904 compiler-rt-emu: Add missing file.
29905
29906 emunet: Fix init error checking.
29907 Otherwise emunet doesn't expose any cards.
29908
29909 fddboot_test: Add -no-pad to xorriso.
29910
29911 grub-mkrescue: pass all unrecognized options unchanged to xorriso.
29912
29913 cacheinfo: Add missing license information.
29914
29915 2015-03-19 Andrei Borzenkov <[email protected]>
29916
29917 grub-fs-tester: add LVM RAID1 support
29918 LVM miscalculates bitmap size with small extent, so start with 16K as
29919 for other RAID types.
29920
29921 Until version 2.02.103 LVM counts metadata segments twice when checking
29922 available space, reduce segment count by one to account for this bug.
29923
29924 2015-03-19 Andrei Borzenkov <[email protected]>
29925
29926 core: add LVM RAID1 support
29927 Closes 44534.
29928
29929 2015-03-16 Andrei Borzenkov <[email protected]>
29930
29931 grub-fs-tester: explicitly set segment type for LVM mirror
29932 LVM mirror defaults to RAID1 today and can be different on different
29933 systems as set in lvm.conf.
29934
29935 2015-03-15 Andrei Borzenkov <[email protected]>
29936
29937 grub-fs-tester: better estimation of filesystem time for LVM/RAID
29938 Write activity with LVM/RAID can happen after filesystem is unmounted.
29939 In my testing modification time of loop files was 15 - 20 seconds
29940 after unmount. So use time as close to unmount as possible as
29941 reference instead.
29942
29943 2015-03-06 Vladimir Serbinenko <[email protected]>
29944
29945 hfsplus: Fix potential access to uninited memory on invalid FS
29946
29947 2015-03-06 Jon McCune <[email protected]>
29948
29949 autogen.sh: Allow overriding the python to be used by setting $PYTHON.
29950 Some installations have several python versions installed. Allow user
29951 to choose which one to use by setting $PYTHON.
29952
29953 2015-03-05 Andrei Borzenkov <[email protected]>
29954
29955 update gnulib/argp-help.c to fix garbage in grub-mknetdir --help output
29956 argp_help attempts to translate empty string, which results in printing
29957 meta information about translation, like in
29958
29959 bor@opensuse:~/build/grub> grub2-mknetdir --help
29960 Использование: grub2-mknetdir [ПАРАМЕТР…]
29961 Project-Id-Version: grub 2.02-pre2
29962 Report-Msgid-Bugs-To: [email protected]
29963 ...
29964
29965 Update gnulib/argp-help.c to the current version which fixes this
29966 (commit b9bfe78424b871f5b92e5ee9e7d21ef951a6801d).
29967
29968 2015-03-05 Andrey Borzenkov <[email protected]>
29969
29970 update m4/extern-inline.m4 to upstream version to fix compilation on FreeBSD
29971 In file included from util/grub-mkimage.c:54:0:
29972 ./grub-core/gnulib/argp.h:627:49: error: '__sbistype' is static but
29973 used in inline function '_option_is_short' which is not static
29974 [-Werror] cc1: all warnings being treated as errors gmake[2]: ***
29975 [util/grub_mkimage-grub-mkimage.o] Error 1
29976
29977 Update m4/extern-inline.m4 to current upstream gnulib version that
29978 contains fix for this (commit b9bfe78424b871f5b92e5ee9e7d21ef951a6801d).
29979
29980 Reported-By: Beeblebrox <[email protected]>
29981
29982 2015-03-04 Vladimir Serbinenko <[email protected]>
29983
29984 syslinux_parse: Fix the case of unknown localboot.
29985 Reported by: Jordan Uggla
29986
29987 configure.ac: Fix the name of pciaccess header.
29988
29989 Fix canonicalize_file_name clash.
29990 canonicalize_file_name clashed with gnulib function. Additionally
29991 it was declared in 2 places: emu/misc.h and util/misc.h. Added
29992 grub_ prefix and removed second declaration.
29993
29994 2015-03-03 Vladimir Serbinenko <[email protected]>
29995
29996 Remove emu libusb support.
29997 It's disabled by default and has been broken for a long time.
29998 As nobody is interested in fixing and maintaining it, remove it.
29999
30000 configure.ac: Remove unused COND_clang
30001
30002 Remove libgcc dependency.
30003 libgcc for boot environment isn't always present and compatible.
30004 libgcc is often absent if endianness or bit-size at boot is different
30005 from running OS.
30006 libgcc may use optimised opcodes that aren't available on boot time.
30007 So instead of relying on libgcc shipped with the compiler, supply
30008 the functions in GRUB directly.
30009 Tests are present to ensure that those replacement functions behave the
30010 way compiler expects them to.
30011
30012 types.h: Use __builtin_bswap* with clang.
30013 clang pretends to be GCC 4.2 but we use __builtin_bswap* only with GCC 4.3+.
30014 clang support __builtin_bswap*, so use it.
30015
30016 configure.ac: Set $CPPFLAGS when checking for no_app_regs.
30017 Fixes compilation for sparc64 with clang.
30018
30019 Don't continue to query block-size if disk doesn't have it.
30020 Stops poluting screen with a lot of "block-size: exception -21".
30021
30022 2015-02-28 Andrei Borzenkov <[email protected]>
30023
30024 grub-probe: free temporary variable
30025
30026 2015-02-28 Vladimir Serbinenko <[email protected]>
30027
30028 exclude.pot: Add new technical strings
30029
30030 grub-probe: Mark a "[default=]" for translation.
30031
30032 grub-shell: Add missing --locale-directory.
30033 Fixes the language tests is no make install was done.
30034
30035 ntfs_test: Skip is setfattr is unavailable.
30036
30037 2015-02-26 Vladimir Serbinenko <[email protected]>
30038
30039 emu/cache: Change declaration of __clear_cache to match builtin declaration.
30040 Fixes compile of arm64-emu.
30041
30042 arm/dl: Fix handling of nonstandard relocation sizes
30043
30044 gzio: Optimize by removing division.
30045
30046 raid6: Optimize by removing division.
30047
30048 dmraid_nvidia: Fix division by 0 and missing byte-swap.
30049
30050 crypto: restrict cipher block size to power of 2.
30051 All current ciphers have blocks which are power of 2 and it's
30052 unlikely to change. Other block length would be tricky to handle anyway.
30053 This restriction allows avoiding extra divisions.
30054
30055 jpeg: Optimise by replacing division with shifts.
30056
30057 png: Optimize by avoiding divisions.
30058
30059 Add missing lib/division.c
30060
30061 fbblit: Optimize by replacing division with additions and shifts.
30062
30063 bitmap_scale: Optimize by moving division out of the loop.
30064
30065 minilzo: Skip parts tha we don't need.
30066
30067 2015-02-23 Vladimir Serbinenko <[email protected]>
30068
30069 mips: Fix soft-float handling.
30070 Add -msoft-float alongside clang arguments to specify ABI.
30071 Specify ABI in asm files explicitly.
30072 This trigers asm warning due to gcc failing to propagate -msoft-float
30073 but it's tolerable.
30074
30075 Add missing grub_ prefix in memcpy invocation
30076
30077 Allow clang compilation for thumb with -mthumb-interwork.
30078 clang already uses -mthumb-interwork behaviour even thout it doesn't
30079 support the option.
30080
30081 arm64: Fix compilation failure.
30082 Don't supply +nosimd to asm files.
30083 Otherwise +nosimd coming from flags forbids some of instructions
30084 used in cache_flush.
30085
30086 Supply signed division to fix ARM compilation.
30087 Previously we supplied only unsigned divisions on platforms that need software
30088 division.
30089 Yet compiler may itself use a signed division. A typical example would be a
30090 difference between 2 pointers which involves division by object size.
30091
30092 2015-02-22 Vladimir Serbinenko <[email protected]>
30093
30094 acpi: Fix unused function warning.
30095
30096 configure.ac: Add ia64-specific way to disable floats.
30097
30098 i386/tsc: Fix unused function warning on xen.
30099
30100 2015-02-22 Vladimir Serbinenko <[email protected]>
30101
30102 Experimental support for clang for sparc64.
30103 Automatically discover command line options to make clang and
30104 gcc behave in same way.
30105
30106 Tested with qemu.
30107
30108 2015-02-22 Vladimir Serbinenko <[email protected]>
30109
30110 Discover which option provides soft-float on configure stage.
30111 Deals with clang needing other arguments to stop issuing floating
30112 instructions than gcc.
30113
30114 2015-02-21 Vladimir Serbinenko <[email protected]>
30115
30116 mips: Switch to more portable .org
30117 Binary is unchanged.
30118
30119 sparc64: Switch to more portable .org.
30120 Binaries are unchanged.
30121
30122 kernel-8086: Switch to more portable .org.
30123
30124 Relax requirements on asm for non-BIOS i386 platforms.
30125 These platforms don't have a hard limit on size of resulting code16
30126 code, so we don't care if assembly is bigger than necessarry.
30127
30128 qemu: Switch to more portable .org
30129 Binary is checked identical.
30130
30131 qemu: Fix GateA20 enabling.
30132 GateA20 code was inactive due to address error.
30133
30134 qemu: Fix compilation
30135
30136 Remove realmode.S from coreboot and qemu.
30137 It's not used there.
30138
30139 Remove obsolete ADDR32 and DATA32 checks.
30140
30141 i386: Remove needless ADDR32 prefixes when address is known and fixed.
30142 Shaves off 6 bytes in lzma_decompress.img.
30143
30144 i386-pc/boot: Explicitly mark kernel_address[_high] as local.
30145 Otherwise apple asm might try to make accesses relocatable.
30146
30147 Change dot assignmnet to more portable .org.
30148 Binary is unchanged (verified)
30149
30150 i386: Move from explicit ADDR32/DATA32 prefixes to instruction suffixes.
30151 Is more portable.
30152 Binary is unchanged (verified).
30153
30154 Test which flags make our asm compile.
30155 Previously we relied on assumption that clang always needs -no-integrated-as
30156 but it's not always true.
30157
30158 INSTALL: clarify that clang support is experimental
30159
30160 zfs/mzap_lookup: Fix argument types
30161
30162 wildcard: Mark unused argument as such.
30163
30164 ofdisk: Exclude floppies from scanning.
30165 It causes similar hang as CD on at least the qemu.
30166
30167 configure: Add -msoft-float to CCASFLAGS
30168 Otherwise mismatch between API flags triggers linker failure
30169
30170 mips/startup_raw: Use more portable .asciz
30171
30172 Provide __aeabi_mem{cpy,set}
30173 Fixes ARM compilation
30174
30175 div_test: Don't try to divide by zero
30176
30177 INSTALL: Fix names of host flags to match actual behaviour
30178
30179 Strip .MIPS.abiflags which causes compile failure
30180
30181 2015-02-20 Vladimir Serbinenko <[email protected]>
30182
30183 configure: Move adding of include options to the very end to avoid subshell.
30184
30185 configure: Add missing comma.
30186
30187 2015-02-16 Vladimir Serbinenko <[email protected]>
30188
30189 ext2: Ignore INCOMPAT_MMP.
30190 It's not really incompatible as long as driver never writes to FS.
30191
30192 ext2: Support META_BG.
30193 This fixes bug that system would become unbootable after ext*
30194 online resize if no resize_inode was created at ext* format time.
30195
30196 2015-02-16 Andrei Borzenkov <[email protected]>
30197
30198 tests: remove hardcoded paths from syslinux_test
30199 abs_top_srcdir appeared in Autoconf 2.52f. Minimal grub requirement
30200 is 2.60 so we should be good here.
30201
30202 build-sys: add syslinux test files to tarball
30203
30204 2015-02-16 Vladimir Serbinenko <[email protected]>
30205
30206 Add test for syslinux converter
30207
30208 2015-02-16 Vladimir Serbinenko <[email protected]>
30209
30210 Don't remove initrd= parameter.
30211 Based on simplified patch by Lunar.
30212
30213 Reported by: Lunar
30214
30215 2015-02-16 Vladimir Serbinenko <[email protected]>
30216
30217 syslinux_parse: Always output comments even if no entries are found.
30218
30219 2015-02-15 Andrei Borzenkov <[email protected]>
30220
30221 diskfilter_make_raid: more memory leaks in failure path
30222
30223 2015-02-14 Vladimir Serbinenko <[email protected]>
30224
30225 disk/lvm: Use zalloc to ensure that segments are initialised to sane value.
30226 Reported by: EmanueL Czirai.
30227
30228 2015-02-14 Daniel Kiper <[email protected]>
30229
30230 multiboot2: Fix information request tag size calculation
30231
30232 2015-02-14 Andrei Borzenkov <[email protected]>
30233
30234 diskfilter: fix double free of lv names for mdraid
30235 Avoid micro-optimization in grub_diskfilter_make_raid and make sure
30236 name and fullname are independent strings. This avoids need to special
30237 case it everywhere else.
30238
30239 Also fix memory leak in failure case in grub_diskfilter_make_raid.
30240
30241 Closes: 41582
30242
30243 2015-02-14 Andrei Borzenkov <[email protected]>
30244
30245 diskfilter: fix crash in validate_lv for mdraid arrays
30246 Commit 750f4bacd3262376ced3f837d8dc78f834ca233a put LV validation before
30247 actual vg assignment. Make grub_diskfilter_make_raid to assign ->vg as
30248 happens in other cases for consistency. Also clean up redundant code and add
30249 explicit NULL lv->vg check in validate_lv.
30250
30251 Also fix segment validation in validate_lv; it became obvious when crash
30252 was fixed.
30253
30254 Closes: 44199
30255
30256 2015-02-12 Jiri Slaby <[email protected]>
30257
30258 util: mkimage, fix gcc5 build failure
30259 gcc5 reports:
30260 ../util/mkimage.c: In function 'grub_install_get_image_target':
30261 ../util/mkimage.c:954:5: error: loop exit may only be reached after undefined behavior [-Werror=aggressive-loop-optimizations]
30262 && j < ARRAY_SIZE (image_targets[i].names); j++)
30263 ^
30264 ../util/mkimage.c:953:39: note: possible undefined statement is here
30265 for (j = 0; image_targets[i].names[j]
30266 ^
30267
30268 Well, let's move the index 'j' test before accesing the array to:
30269 1) make the loop obvious
30270 2) make gcc happy
30271
30272 2015-02-03 Leif Lindholm <[email protected]>
30273
30274 arm: implement additional relocations generated by gcc 4.9 at -O3
30275 GCC 4.9 also generates R_ARM_THM_MOVW_ABS_NC and R_ARM_THM_MOVT_ABS,
30276 as an alternative to ABS32.
30277
30278 2015-01-30 Andrei Borzenkov <[email protected]>
30279
30280 setup: fix blocklist size calculation
30281 Found by: Coverity scan.
30282
30283 grub-fstest: fix descriptor leak
30284 Found by: Coverity scan.
30285
30286 2015-01-30 Andrei Borzenkov <[email protected]>
30287
30288 net/pxe: fix error condition
30289 Test return value of grub_netbuff_reserve(), buf itself cannot be
30290 NULL here.
30291
30292 Found by: Coverity scan.
30293
30294 2015-01-30 Andrei Borzenkov <[email protected]>
30295
30296 grub-mkimage: fix potential NULL pointer dereference
30297 Move fatal check whether symtab_section is NULL before first reference.
30298
30299 Found by: Coverity scan.
30300
30301 2015-01-30 Andrei Borzenkov <[email protected]>
30302
30303 net/ip: check result of grub_netbuff_push
30304 Found by: Coverity scan.
30305
30306 tests: add test command file tests
30307 This requires access to files in both host and grub image, so
30308 implementing as separate test unit instead of script test was
30309 more easy.
30310
30311 test: consistently use TMPDIR and same name pattern for temp files
30312
30313 test: fix previous commit - we need to return from subexpression
30314 ( ... ) was processed recursively, we need to return from it. Revert
30315 this change.
30316
30317 test: do not stop after first file test or closing bracket
30318 Closes: 44115
30319
30320 2015-01-28 Leif Lindholm <[email protected]>
30321
30322 configure.ac: don't use -msoft-float for arm64
30323 aarch64 toolchains do not support the -msoft-float option added by
30324 commit 3661261f. Insted, for arm64 use -march=armv8-a+nofp+nosimd.
30325
30326 Reported-by: Ryan Harkin <[email protected]>
30327
30328 2015-01-28 Andrei Borzenkov <[email protected]>
30329
30330 script/execute.c: fix memory leak.
30331 Make sure to continue loop over array after failure to free
30332 allocated strings.
30333
30334 Found by: Coverity scan.
30335
30336 2015-01-28 Andrei Borzenkov <[email protected]>
30337
30338 syslinux_parse: fix memory leak.
30339 Found by: Coverity scan.
30340
30341 2015-01-27 Andrei Borzenkov <[email protected]>
30342
30343 Change quotes to match overall style in NEWS
30344
30345 loader/xnu: fix memory leak.
30346 Foound by: Coverity scan.
30347
30348 util/grub-probe: fix memory leaks.
30349 Found by: Coverity scan.
30350
30351 fs/hfsplus: fix memory leak.
30352 Found by: Coverity scan.
30353
30354 fs/zfs/zfscrypt.c: fix indentation.
30355
30356 fs/zfs/zfscrypt.c: fix memory leaks.
30357 Found by: Coverity scan.
30358
30359 commands/parttool: fix memory leak.
30360 Found by: Coverity scan.
30361
30362 fs/zfs/zfs.c: fix memory leak.
30363 Found by: Coverity scan.
30364
30365 linux/ofpath: fix descriptor leak
30366 Found by: Coverity scan
30367
30368 linux/hostdisk: use strncpy instead of strlcpy
30369 strlcpy is not available on Linux as part of standard libraries.
30370 It probably is not worth extra configure checks espicially as we
30371 need to handle missing function anyway.
30372
30373 2015-01-27 Vladimir Serbinenko <[email protected]>
30374
30375 Document intentional fallthroughs.
30376 Found by: Coverity scan.
30377
30378 linux/ofpath: Fix error handling.
30379 Found by: Coverity Scan.
30380
30381 linux/hostdisk: Limit strcpy size to buffer size.
30382 Found by: Coverity scan.
30383
30384 fs/zfscrypt: Add missing explicit cast.
30385 Found by: Coverity scan.
30386
30387 fs/zfs: Fix error handling.
30388 Found by: Coverity Scan.
30389
30390 2015-01-27 Vladimir Serbinenko <[email protected]>
30391
30392 fs/{cbfs,cpio}: Remove useless check if mode is NULL.
30393 Callers already ensure that it's not null.
30394
30395 Found by: Coverity Scan.
30396
30397 2015-01-27 Vladimir Serbinenko <[email protected]>
30398
30399 commands/acpi: Use ALIGN_UP rather than manual expression.
30400 Improves readability and hopefully automatic scanning.
30401
30402 Found by: Coverity Scan.
30403
30404 2015-01-26 Andrei Borzenkov <[email protected]>
30405
30406 util/setup: fix memory leak.
30407 Found by: Coverity scan.
30408
30409 util/mkimage: fix memory leaks.
30410 Found by: Coverity scan.
30411
30412 util/grub-mount: fix descriptor leak.
30413 Found by: Coverity scan.
30414
30415 util/grub-mkstandalone: fix memory leak.
30416 Found by: Coverity scan.
30417
30418 util/grub-install: rearrange code to avoid memory leak.
30419 Found by: Coverity scan.
30420
30421 linux/getroot: fix memory leak.
30422 Found by: Coverity scan.
30423
30424 util/install: fix memory leak.
30425 Found by: Coverity scan.
30426
30427 util/setup: fix memory leak.
30428 Found by: Coverity scan.
30429
30430 linux/ofpath: fix various memory leaks.
30431 Found by: Coverity scan.
30432
30433 linux/getroot: fix descriptor leak.
30434 Found by: Coverity scan.
30435
30436 2015-01-26 Vladimir Serbinenko <[email protected]>
30437
30438 util/misc.c: Check ftello return value.
30439 Found by: Coverity scan.
30440
30441 grub-macbless: Fix resource leak.
30442 Found by: Coverity scan.
30443
30444 grub-install: Fix memory leak.
30445 Found by: Coverity scan.
30446
30447 grub-install-common: Fix sizeof usage.
30448 Found by: Coverity scan.
30449
30450 util/getroot: Add missing grub_disk_close.
30451 Found by: Coverity scan.
30452
30453 vbe: Fix incorrect register usage.
30454 Found by: Coverity scan.
30455
30456 unix/password: Fix file descriptor leak.
30457 Found by: Coverity scan.
30458
30459 linux/getroot: Fix error handling.
30460 Found by: Coverity scan.
30461
30462 linux/blocklist: Fix memory leak.
30463 Found by: Coverity scan.
30464
30465 devmapper/getroot: Fix memory leak.
30466 Found by: Coverity scan.
30467
30468 normal/misc: Close device on all pathes.
30469 Found by: Coverity scan.
30470
30471 normal/main: Fix error handling.
30472 Found by: Coverity scan.
30473
30474 xnu: Add missing error check.
30475 Found by: Coveriy scan.
30476
30477 plan9: Add missing grub_device_close.
30478 Found by: Coverity scan.
30479
30480 multiboot: Simplify to avoid confusing assignment.
30481 Found by: Coverity scan.
30482
30483 bsd: Add missing null-pointer check.
30484 Found by: Coverity scan.
30485
30486 lib/syslinux_parse: Add missing error check.
30487 Found by: Coverity scan.
30488
30489 lib/syslinux_parse: Fix memory leak.
30490 Found by: Coveriy scan.
30491
30492 lib/syslinux_parse: Add missing alloc check.
30493 Found by: Coverity scan.
30494
30495 i386/pc/mmap: Fix memset size.
30496 Found by: Coverity scan.
30497
30498 gfxmenu/theme_loader: Add missing allos error check.
30499 Found by: Coverity scan.
30500
30501 gfxmenu/icon_manager: Fix null pointer dereference.
30502 Found by: Coverity scan.
30503
30504 fs/ufs: Add missing error check.
30505 Found by: Coverity scan.
30506
30507 configure.ac: Always add -D_FILE_OFFSET_BITS=64.
30508
30509 2015-01-25 Vladimir Serbinenko <[email protected]>
30510
30511 fs/sfs: Fix error check and add sanity check.
30512 Found by: Coverity scan.
30513
30514 fs/reiserfs: Fix sector count overflow.
30515 Found by: Coverity scan.
30516
30517 fs/ntfs: Add sizes sanity checks.
30518 Found by: Coverity scan.
30519
30520 fs/ntfs: Add missing free.
30521 Found by: Coverity scan.
30522
30523 2015-01-25 Vladimir Serbinenko <[email protected]>
30524
30525 fs/minix: Fix sector promotion to 64-bit.
30526 While on it make GRUB_MINIX_ZONE2SECT into function.
30527
30528 Found by: Coverity scan
30529
30530 2015-01-25 Vladimir Serbinenko <[email protected]>
30531
30532 grub_iso9660_read: Explicitly check read_node return value.
30533 Not really needed as grub_errno is already checked but is nicer.
30534
30535 Found by: Coverity scan.
30536
30537 2015-01-25 Andrei Borzenkov <[email protected]>
30538
30539 commands/fileXX: Fix remaining memory leak.
30540 Found by: Coverity Scan.
30541
30542 2015-01-25 Vladimir Serbinenko <[email protected]>
30543
30544 fs/hfs: Add pointer sanity checks.
30545 Found by: Coverity scan.
30546
30547 fs/hfs/hfs_open: Check that mount succeeded.
30548 Found by: Coverity scan.
30549
30550 fs/fat: Fix codepath to properly free on error.
30551 Found by: Coverity scan.
30552
30553 fs/cpio_common: Add a sanity check on namesize.
30554 Found by: Coverity scan.
30555
30556 fs/cbfs: Add missing free.
30557 Found by: Coverity scan.
30558
30559 2015-01-24 Vladimir Serbinenko <[email protected]>
30560
30561 font: Add missing free.
30562 Found by: Coverity Scan.
30563
30564 biosdisk: Add missing cast.
30565 Found by: Coverity scan.
30566
30567 disk/geli: Add missing free.
30568 Found by: Coverity scan.
30569
30570 disk/geli: Add missing seek success check.
30571 Found by: Coverity scan.
30572
30573 disk/diskfilter: Add missing lv presence check.
30574 Found by: Coverity scan.
30575
30576 disk/cryptodisk: Add missing error check.
30577 Found by: Coverity scan.
30578
30579 2015-01-24 Vladimir Serbinenko <[email protected]>
30580
30581 disk/ahci: Fix device_map_range argument.
30582 Argument is not used on x86, hence it's gone unnoticed.
30583
30584 Found by: Coverity scan.
30585
30586 2015-01-24 Vladimir Serbinenko <[email protected]>
30587
30588 disk/AFsplitter: check argument validity before doing any allocs.
30589 This avoids possible memory leaks.
30590
30591 Found by: Coverity scan.
30592
30593 2015-01-24 Vladimir Serbinenko <[email protected]>
30594
30595 commands/wildcard: Add missing free.
30596 Found by: Coverity scan.
30597
30598 commands/verify: Fix sha1 context zeroing-out.
30599 Current code doesn't zero-out context completely. It's a minor issue
30600 really as sha1 init already takes care of initing the context.
30601
30602 commands/tr: Simplify and fix missing parameter test.
30603 Found by: Coverity scan
30604
30605 commands/syslinux: Add missing free.
30606 Found by: Coverity scan.
30607
30608 commands/parttool: Add missing device close.
30609 Found by: Coverity scan.
30610
30611 commands/nativedisk: Add missing device_close.
30612 Found by: Coverity scan.
30613
30614 2015-01-24 Vladimir Serbinenko <[email protected]>
30615
30616 commands/macbless: Handle device opening errors correctly.
30617 Wrong variable was checked for errors.
30618
30619 Found by: Coverity scan.
30620
30621 2015-01-24 Vladimir Serbinenko <[email protected]>
30622
30623 commands/macbless: Fix potential overflow.
30624 Is a minor concern as no such FS would be created under normal circumstances
30625 and failure was benign.
30626
30627 Found by: Coverity scan.
30628
30629 2015-01-24 Vladimir Serbinenko <[email protected]>
30630
30631 commands/macbless: Remove incorrect grub_free.
30632 Found by: Coverity Scan
30633
30634 commands/legacycfg: Fix resource leaks.
30635
30636 zfs: Fix disk-matching logic.
30637 Reported by: Tim Chase <dweeezil>
30638
30639 commands/hdparm: Add missing grub_disk_close.
30640 Found by: Coverity scan.
30641
30642 gptsync: Add missing device_close.
30643 Found by: Coverity scan
30644
30645 commands/fileXX: Fix memory leak.
30646 Found by: Coverity Scan.
30647
30648 2015-01-24 Vladimir Serbinenko <[email protected]>
30649
30650 commands/file: Change the confusing loop stop condition.
30651 Old condition was used to zero-out header variable on exit of the loop.
30652 This is correct but confusing. Replace with in-loop logic.
30653
30654 Found by: Coverity Scan.
30655
30656 2015-01-24 Vladimir Serbinenko <[email protected]>
30657
30658 commands/acpi: Use ALIGN_UP rather than manual expression.
30659 Improves readability and hopefully automatic scanning.
30660
30661 Found by: Coverity Scan.
30662
30663 2015-01-24 Vladimir Serbinenko <[email protected]>
30664
30665 uhci: Fix null pointer dereference.
30666 Found by: Coverity scan.
30667
30668 Always add -msoft-float to avoid compiler generating float arithmetics.
30669
30670 2015-01-24 Vladimir Serbinenko <[email protected]>
30671
30672 Generate empty ChangeLog if no .git is available.
30673 When making dist from a git snapshot without repo available make dist would
30674 fail to find ChangeLog. Generate empty ChangeLog if no ChangeLog is already
30675 present and repo is not available.
30676
30677 Reported by: Andrei Borzenkov <[email protected]>
30678
30679 2015-01-24 Vladimir Serbinenko <[email protected]>
30680
30681 Makefile.am: Fix Changelog cutoff address.
30682 gitlog-to-changelog Doesn't generate entries for cutoff day, only
30683 for days after the cutoff date, adjust by one to compensate.
30684
30685 efidisk: Return the determined root disk even if partition is unknown.
30686
30687 util/grub-mkrescue.c: Always include part_msdos and part_gpt on EFI.
30688 When booted from stick, EFI would use GPT partition and our root
30689 device detection algortihm depends on GRUB's ability to see the same
30690 partitions. Hence include msdos and gpt partmap modules on EFI even when
30691 they're not needed to access root filesystem.
30692
30693 conf/Makefile.common: Remove unused {LD,C}FLAGS_CPU.
30694
30695 Autogenerate ChangeLog from git changelog.
30696 Old ChangeLog is moved to ChangeLog-2015. For all changes starting from
30697 this one ChangeLog will be generated from gitlog only on explicit make
30698 invocation and make dist.