Debian Codegraph @

ChangeLog

raw
1 2026-01-14 Fabio M. Di Nitto <[email protected]>
2
3 Merge pull request #466 from kronosnet/stable1-proposed
4 stable1-proposed
5
6 2026-01-13 Fabio M. Di Nitto <[email protected]>
7
8 update copyright to 2026
9
10 2025-12-01 Jan Friesse <[email protected]>
11
12 tests: Remove discards const qualifier warnings
13 gcc 15.2.1 started producing warning about "assignment discards
14 ‘const’ qualifier from pointer target type" so patch adds const
15 qualifier to fix this warning.
16
17 2025-11-24 Fabio M. Di Nitto <[email protected]>
18
19 build: add support for rust coreutils
20 ... next we drop support for Ubuntu if they don´t upstream their
21 breaking changes ...
22
23 2025-10-13 Chrissie Caulfield <[email protected]>
24
25 Annotate coverity reports so that we're not wading in them
26 every time a new version comes out
27
28 2025-10-13 Ferenc Wágner <[email protected]>
29
30 Distribute STYLE_GUIDE.md and m4/pkg_check_var.m4
31 For sake of completeness, mostly, since almost everybody uses Git and
32 pkg-config>=0.28 or pkgconf (where PKG_CHECK_VAR is already included)
33 for development nowadays.
34
35 2025-09-25 Fabio M. Di Nitto <[email protected]>
36
37 Merge pull request #446 from kronosnet/stable1-proposed
38 stable1-proposed
39
40 2025-09-24 Fabio M. Di Nitto <[email protected]>
41
42 libknet: deprecate SCTP support
43
44 2025-09-23 Fabio M. Di Nitto <[email protected]>
45
46 man: align doxyxml with libqb for solaris support
47
48 2025-09-23 Christine Caulfield <[email protected]>
49
50 Update README to explain libqb dependancy
51
52 Several test suite fixes
53 Increase the sleep in pmtud tests
54 Increase timeout in fun_onwire_upgrade
55 Try and make sure IPv6 test addresses are always valid
56
57 Make knet work on OpenIndiana (Solaris)
58 author: various
59
60 2025-09-06 Fabio M. Di Nitto <[email protected]>
61
62 libnozzle - Fix FreeBSD nozzle_down (take 2)
63 also, FreeBSD 13+ now allow tap0 to be delete.. if only the kernel
64 interface was stable....
65
66 2025-09-04 Chrissie Caulfield <[email protected]>
67
68 libnozzle - Fix FreeBSD nozzle_down (#458)
69 For reasons I am unlikely ever to understand, it's necessary to call
70 SIOCGIFFLAGS on a tap interface after SIOCIFDESTROY for the device to
71 actually be removed.
72
73 ...so do that
74
75 2025-08-30 David Hanisch <[email protected]>
76
77 fix wrong tests
78 Api calls fail due to NULL knet handle, but intention is to fail
79 due to wrong arguments.
80
81 2025-08-29 Fabio M. Di Nitto <[email protected]>
82
83 libknet: bump soname to reflect new symbol
84
85 2025-08-29 David Hanisch <[email protected]>
86
87 Allow to configure a dscp value for KNET_LINK_FLAG_TRAFFICHIPRIO
88 Currently IPTOS_LOWDELAY is used for IP_TOS. A setup may require to
89 set a certain dscp value to prioritize the IP traffic.
90
91 2025-08-28 Christine Caulfield <[email protected]>
92
93 Suppress some sudden coverity errors
94
95 2025-08-25 Dmitry Belyavskiy <[email protected]>
96
97 Some error checking for OpenSSL functions
98
99 2025-08-12 Fabio M. Di Nitto <[email protected]>
100
101 Fix wrong pointer in _handle_recv_from_links_thread()
102 Resolves: https://github.com/kronosnet/kronosnet/issues/449
103
104 2025-06-23 Christine Caulfield <[email protected]>
105
106 nozzle: Fix potential thread race in nozzle_close()
107
108 2025-06-16 google-labs-jules[bot] <161369871+google-labs-jules[bot]@users.noreply.github.com>
109
110 Create and refine STYLE_GUIDE.md for project contributions
111 This commit introduces STYLE_GUIDE.md, a comprehensive document
112 outlining coding standards and best practices for the Kronosnet project.
113 This guide aims to ensure consistency and clarity in contributions.
114
115 The following key areas are covered:
116
117 1. **C Language Style**:
118 * Minimum Dialect: C99.
119 * Indentation: Tabs (not spaces).
120 * Curly Braces:
121 * Control flow (if, for, while): Opening brace on the same line.
122 * Function definitions: Opening brace on the next line.
123 * Always use braces, even for single-line blocks.
124 * Variable Declaration: At the beginning of their scope.
125 * Line Length: Preferred maximum of 120 characters, flexible for
126 readability (e.g., nested structs).
127
128 2. **Naming Conventions**:
129 * General: `snake_case` for internal variables and function names.
130 * Public API Elements:
131 * `knet_` or `nozzle_` prefix for functions, structs, typedefs
132 (e.g., `knet_get_host_list`, `nozzle_device_config`).
133 * `KNET_` or `NOZZLE_` prefix for enums and defines (macros),
134 followed by uppercase `SNAKE_CASE` (e.g., `KNET_MAX_CLIENTS`,
135 `NOZZLE_BUFFER_SIZE_DEFAULT`).
136 * Internal Elements:
137 * Names should be descriptive, indicating purpose and subsystem.
138 * Functions shared across multiple threads: Prefix with `_`
139 (e.g., `_shared_data_access`).
140
141 3. **Comments**:
142 * `//` for single-line, `/* ... */` for multi-line.
143 * Emphasizes clear and concise comments for non-obvious code.
144
145 4. **Best Practices**:
146 * API Changes Require Tests: Any internal or external API
147 modification must be accompanied by corresponding tests in the
148 test suite.
149
150 5. **Copyright**:
151 * A copyright notice specific to `STYLE_GUIDE.md` (Year 2025,
152 Author: Jules <AI Agent>) has been included and formatted as
153 plain Markdown text.
154
155 This initial version of the style guide addresses issue #22 and incorporates
156 iterative feedback on formatting, content, and specificity. A trivial
157 change (extra newline at EOF) was included in this commit to facilitate
158 this comprehensive message.
159
160 2025-04-24 Fabio M. Di Nitto <[email protected]>
161
162 Merge pull request #438 from kronosnet/stable1-proposed
163 stable1-proposed
164
165 2025-04-22 Fabio M. Di Nitto <[email protected]>
166
167 Update copyright
168
169 udp: lower log levels triggered for each incoming packets
170
171 2025-03-20 Christine Caulfield <[email protected]>
172
173 compress: use lzo1x_decompress_safe
174 lzo1x_decompress() can SEGV if bad data is passed to it.
175 lzo1x_decompress_safe() is slower but ... safer.
176
177 Issue #435
178
179 2024-12-02 Fabio M. Di Nitto <[email protected]>
180
181 Merge pull request #429 from kronosnet/stable1-proposed
182 stable1 proposed
183
184 Convert time_t to unsigned long long before formatting (#431)
185 Conventionally it's signed long under the GNU C library, which is 32
186 bits on 32-bit architectures, but Glibc provides the __TIME_BITS==64
187 option, which makes it signed long long instead (64 bits) and makes no
188 difference on 64-bit architectures [1]. Since the ping_interval and
189 pong_timeout members of struct knet_link are unsigned, we can make
190 this conversion safe by rejecting negative values as well (besides
191 zero). In the end this enables using the uniform %llu conversion
192 specifier across all supported architectures. This got exposed by
193 Debian recently transitioning to 64-bit time_t [2].
194
195 [1] https://sourceware.org/glibc/wiki/Y2038ProofnessDesign
196 [2] https://wiki.debian.org/ReleaseGoals/64bit-time
197
198 Co-authored-by: Ferenc Wágner <[email protected]>
199
200 2024-12-02 Christine Caulfield <[email protected]>
201
202 Check packets come from the correct interface https://github.com/corosync/corosync/issues/750
203
204 2024-06-06 Fabio M. Di Nitto <[email protected]>
205
206 Merge pull request #423 from kronosnet/stable1-proposed
207 Stable1 proposed
208
209 2024-05-18 Fabio M. Di Nitto <[email protected]>
210
211 [man] update to latest doxyxml from libqb
212
213 Fix FORTIFY source detection
214 Problem spotted on FreeBSD 15 where gcc is currently broken.
215
216 The detection can see that the flags are not working, but
217 still propagating the last attempted value into the build,
218 causing a failure.
219
220 2024-01-10 Fabio M. Di Nitto <[email protected]>
221
222 Update Copyright year
223
224 2023-11-27 cglosner <[email protected]>
225
226 libknet/tests: fix potential overflow with sprintf
227
228 2023-11-24 Fabio M. Di Nitto <[email protected]>
229
230 [man] update to latest doxyxml from libqb
231 Resolves: https://github.com/kronosnet/kronosnet/issues/422
232
233 2023-10-16 Fabio M. Di Nitto <[email protected]>
234
235 links: compare apples with apples
236 spotted by build on armhf
237
238 2023-09-27 Fabio M. Di Nitto <[email protected]>
239
240 Merge pull request #418 from kronosnet/stable1-proposed
241 stable1 proposed
242
243 2023-09-26 Fabio M. Di Nitto <[email protected]>
244
245 links: fix conversion
246
247 Merge pull request #413 from kronosnet/stable1-proposed
248 stable1 proposed
249
250 2023-09-25 Fabio M. Di Nitto <[email protected]>
251
252 rx: allow links to be active faster
253 scenario: node A and node B
254
255 node A completes ping/pong link up process faster than B
256 node A starts to send data to node B
257
258 old behavior: node B would unconditionally discard data packets
259
260 new behavior:
261 node B will check internal link status to see if ping/pong link up
262 process is in progress.
263
264 if the link is sending pings and receiving pongs, it´s safe to assume
265 that the link will go up (since we are also receiving data from the
266 node A). node B will speed up link up and accept data immediately.
267
268 This should address a overall init race condition when nodes startup
269 are slighly off by enough milliseconds to cause 2 memeberships to form
270 and see fencing fireworks.
271
272 links: fix ping interval and pong timeout value checking
273 warn if ping interval is lower than thread timer resolution as
274 it has no effects
275
276 error if ping timeout is lower than thread timer resolution as
277 it causes links to be highly unstable
278
279 2023-09-18 Fabio M. Di Nitto <[email protected]>
280
281 nozzle: don´t leak tun/tap interfaces in test suite
282 the need_tun() test check to verify that the kernel module for tun/tap
283 is loaded, was opening a generic /dev/tun (or /dev/tap) to verify
284 the presence of the kernel module.
285
286 On linux, closing the fd to the tap/tun device is enough to destroy
287 the device.
288
289 On BSD the device is set to DOWN on closing the fd, but it´s not
290 destroyed.
291
292 Stop leaking by gathering ifr ioctl info from the generic interface
293 created by BSD and pass it to the ioctlfd to destroy.
294
295 2023-08-14 Jan Friesse <[email protected]>
296
297 [crypto] Try crypt and decrypt on crypto_init
298 Crypt (or decrypt) operation might fail in some cases even when
299 initialization succeeded. This problem is known to happen when
300 FIPS mode is enabled and combination of openssl/aes192 and md5 is used.
301 This was not caught during crypto_init so knet user (corosync) was
302 not aware of non-working crypto so it couldn't handle it - and couldn't
303 send/receive any packets.
304
305 Solution is to try crypt, decrypt and compare of result buffer right
306 after new crypto_instance is created. If any of these operation fails,
307 whole crypto_init fails, crypto_instance is freed and knet user is
308 notified by receiving non-zero return code.
309
310 2023-07-10 Fabio M. Di Nitto <[email protected]>
311
312 Merge pull request #404 from kronosnet/stable1-proposed
313 stable1-proposed
314
315 build: disable fortify check for test binaries
316
317 build: autodetect fortify source level automatically
318
319 build: annocheck disable cf-protection
320 currently reporting some MAYBE false positives in fedora
321
322 2023-06-07 Jan Friesse <[email protected]>
323
324 [spec] Migrate to SPDX license
325 Both Fedora and openSUSE now recommends to use SPDX shortname format for
326 License.
327
328 2023-05-24 Christine Caulfield <[email protected]>
329
330 nozzle: tests: Fix 'set_down' test on FreeBSD-devel
331 The new FreeBSD adds a LOWER_UP flag to interfaces that
332 indicates whether the driver is signalling L1 up. So we
333 need to exclude that from our grep when checking
334 the higher-level interface status.
335
336 2023-04-25 Jan Friesse <[email protected]>
337
338 [tx] Remove expensive memsets
339 Memset of dst_host_ids is relatively expensive because structure is
340 large (128KiB) and it is not really needed because it is always fully
341 set later.
342
343 2023-04-03 Christine Caulfield <[email protected]>
344
345 log: Add a TRACE log level
346 This also removes the #ifdef DEBUG' conditionals
347 so that those messages can be enabled without a rebuild.
348
349 backport of #403
350
351 2023-01-04 Fabio M. Di Nitto <[email protected]>
352
353 Merge pull request #381 from kronosnet/stable1-proposed
354 Stable1 proposed
355
356 2023-01-02 Fabio M. Di Nitto <[email protected]>
357
358 Update copyright
359
360 [man] update to latest doxyxml from libqb
361
362 2022-12-14 Christine Caulfield <[email protected]>
363
364 udp: Move EPERM to be alongside EMSGSIZE in tx error handler
365 EPERM can be returned (tested Linux 5.19 & FreeBSD 13) when a packet is
366 transmitted that is too large for the i/f MTU or an iptables
367 transmit rule - so it makes sense to treat it like EMSGSIZE
368 and trigger a pMTU run
369
370 2022-11-02 Christine Caulfield <[email protected]>
371
372 pmtud: Reset/restart pmtud when a node joins
373 If a new node joins with a 'black hole' that sets the effective
374 MTU to be less than the hardware MTU then it will fail to join.
375
376 So this patch restarts the pMTU processes each time a new node
377 joins so it has a chance to declare it's real MTU in time.
378
379 2022-10-14 Fabio M. Di Nitto <[email protected]>
380
381 Fix update copyright script to better deal with upper/lower cases
382
383 2022-08-22 Fabio M. Di Nitto <[email protected]>
384
385 Update link to google drive
386
387 2022-08-22 Ferenc Wágner <[email protected]>
388
389 m4/ax_pthread.m4: update to latest upstream version (serial 31)
390 To avoid the warning: $as_echo is obsolete message.
391
392 2022-07-27 Jan Friesse <[email protected]>
393
394 configure: Conditionalize AC_PROG_CC_C99
395 Autoconf 2.70 made AC_PROG_CC_C99 deprecated because
396 AC_PROG_CC is performing same set of tests and sets same
397 variables.
398
399 To remove warning, conditionalize AC_PROG_CC_C99.
400
401 2022-07-22 Chrissie Caulfield <[email protected]>
402
403 admin: Fix a couple of small things (#384)
404 Spotted by the internal CI
405
406 2022-06-09 Mingli Yu <[email protected]>
407
408 links.c: Fix build with gcc-12
409 Fix the build failure with gcc-12 when -Og pass to CFLAGS.
410 | /build/tmp-glibc/work/corei7-64-wrs-linux/kronosnet/1.22-r0/recipe-sysroot/usr/include/bits/string_fortified.h:59:10: error: 'link' may be used uninitialized [-Werror=maybe-uninitialized]
411 | 59 | return __builtin___memset_chk (__dest, __ch, __len,
412 | | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
413 | 60 | __glibc_objsize0 (__dest));
414 | | ~~~~~~~~~~~~~~~~~~~~~~~~~~
415 | ../../git/libknet/links.c: In function 'knet_link_set_config':
416 | ../../git/libknet/links.c:108:27: note: 'link' was declared here
417 | 108 | struct knet_link *link;
418 | | ^~~~
419 | cc1: all warnings being treated as errors
420
421 2022-06-02 Ferenc Wágner <[email protected]>
422
423 Fix comment typo: Minumum -> Minimum
424
425 Fix comment typo: compresion -> compression
426
427 2022-05-30 Fabio M. Di Nitto <[email protected]>
428
429 Merge pull request #370 from kronosnet/stable1-proposed
430 stable1 proposed
431
432 2022-04-29 Chrissie Caulfield <[email protected]>
433
434 [doc] To Do list moved to projects.clusterlabs.org (#379)
435
436 2022-04-01 Jan Friesse <[email protected]>
437
438 [logging] Fix Initializing typo
439
440 2022-03-23 Fabio M. Di Nitto <[email protected]>
441
442 [tx] fix epoll event size to cope with extra event on send to link
443 spotted by -Werror=stringop-overflow= new gcc feature
444
445 2022-03-21 Fabio M. Di Nitto <[email protected]>
446
447 [tests] deal with new annocheck symlink checks
448 command line has changed between f35 and rawhide, so go for
449 a neutral solution
450
451 2022-03-10 Fabio M. Di Nitto <[email protected]>
452
453 [tests] update memcheck exceptions
454
455 2022-03-10 Christine Caulfield <[email protected]>
456
457 tests: move all tests to new infrastructure
458 and update copyright years
459
460 2021-12-14 Christine Caulfield <[email protected]>
461
462 [UDP] Check for EHOSTUNREACH
463 EHOSTUNREACH seems to have been missed from the list of errors
464 checked at transmit time, so add it. Without this,
465 knet can spin under some conditions. (see issue #373)
466
467 2021-11-30 Fabio M. Di Nitto <[email protected]>
468
469 [build] fix flag detections for gcc
470 Resolves: https://github.com/kronosnet/kronosnet/issues/371
471
472 2021-11-17 Fabio M. Di Nitto <[email protected]>
473
474 [doc] update links to mailing lists and drop obsolete test suite
475
476 2021-11-15 Fabio M. Di Nitto <[email protected]>
477
478 Merge pull request #360 from kronosnet/stable1-proposed
479 Stable1 proposed
480
481 2021-11-12 Fabio M. Di Nitto <[email protected]>
482
483 [logging] slow down knet a bit when log socket is busy
484
485 [crypto] improve logging when failing to decrypt packets
486
487 2021-11-08 Fabio M. Di Nitto <[email protected]>
488
489 [udp] use ICMP error messages to trigger faster link down detection
490 this solves a possible race condition when:
491
492 - node1 is running
493 - node2 very fast
494 - node1 does NOT have enough time to detect that node2 has gone
495 and reset the local seq numbers / buffers
496 - node1 will start rejecting valid packets from node2
497
498 There is still a potential minor race condition where app
499 can restart so fast that kernel / network don't have time
500 to generate an ICMP error. This will be addressed using
501 instance id in onwire v2 protocol, as suggested by Jan F.
502
503 [host] fix dst_seq_num initialization race condition
504 There is a potential race condition where the sender
505 is overloaded, sending data packets before pings
506 can kick in and set the correct dst_seq_num.
507
508 if this node is starting up (dst_seq_num = 0),
509 it can start rejecing valid packets and get stuck.
510
511 Set the dst_seq_num to the first seen packet and
512 use that as reference instead.
513
514 2021-10-26 Fabio M. Di Nitto <[email protected]>
515
516 [crypto] fix nss crypto buffer boundaries usage
517 this issue was detected by temporary increasing knet header size
518 for debugging purposes, and was generating incorrect crypto data
519 by PK11_CipherOp because the new size was overriding the old data.
520
521 this issue does not affect onwire_v1 or stable1 as it is.
522
523 [build] fix debug build vs annocheck parameters
524 FORTIFY_SOURCE requires -O1 or higher
525
526 2021-10-19 Khem Raj <[email protected]>
527
528 libknet/tests: Correct include path for poll.h
529 Fixes
530 /usr/include/sys/poll.h:1:2: error: redirec
531 ting incorrect #include <sys/poll.h> to <poll.h> [-Werror,-W#warnings]
532 | #warning redirecting incorrect #include <sys/poll.h> to <poll.h>
533
534 2021-10-18 Fabio M. Di Nitto <[email protected]>
535
536 [tests] temporary disable annobin lto check for centos8 failure
537 adding -flto as suggested, triggers other annobin internal failures on
538 centos8.
539
540 2021-09-17 Christine Caulfield <[email protected]>
541
542 [build] Add annobin build/check support
543 the annobin check is enabled only if all compiler flags, linker flags
544 and annocheck binary are available.
545
546 the build will use as many of the hardening options required to pass
547 the annocheck regardless.
548
549 the check is performed / enabled only with gcc. clang currently suffers
550 from some limitations to automatically detect the annobin plugin, that
551 would increase drastacally the complexity of the build system
552 unnecessarely.
553
554 implementation based on:
555 - https://bugzilla.redhat.com/show_bug.cgi?id=1961686
556 - https://developers.redhat.com/blog/2019/02/04/annocheck-examining-the-contents-of-binary-files#
557 - https://access.redhat.com/documentation/en-us/red_hat_enterprise_linux/8/html/developing_c_and_cpp_applications_in_rhel_8/annobin_toolsets
558
559 Original idea by Christine Caulfield <[email protected]>
560
561 CentOS Stream CI insists on this but it's generally a 'good thing'
562
563 2021-09-16 Fabio M. Di Nitto <[email protected]>
564
565 [nozzle] fix tests on machines that don´t have bash
566
567 2021-09-16 Ferenc Wágner <[email protected]>
568
569 nozzle: preserve test result across cleanup
570
571 nozzle: skip tests if TUN support is not available
572
573 2021-08-30 Fabio M. Di Nitto <[email protected]>
574
575 Merge pull request #344 from kronosnet/stable1-proposed
576 stable1-proposed
577
578 2021-08-26 Christine Caulfield <[email protected]>
579
580 test: Use a pipe to synchronise wait_for_xxx events
581 The condition variable thing clearly not working for us.
582
583 2021-08-26 Fabio M. Di Nitto <[email protected]>
584
585 [tests] make execution of functional testing optionals
586
587 2021-08-26 Christine Caulfield <[email protected]>
588
589 send: Fix send_sync()
590 send_sync needs a filter to work - enforce this
591 Clear out some arrays that could have random values
592
593 2021-08-26 Fabio M. Di Nitto <[email protected]>
594
595 [transports] remove unused internal ACL API
596 with the new ACL code, this internal API is no longer used
597
598 [acl] major internal rework
599 ACL internal management was somewhat broken and didn´t really
600 do what was advertised for dynamic links.
601
602 - fix ACL to be truely per host/link as public API advertise
603 - move ACL checking when host/link are known
604 - add extensive tests
605 - fix various minor bugs that went undetected for a bit
606 - enable access list testings on more tests
607
608 [acl] make knet_link_add_acl a wrapper to knet_link_insert_acl
609 remove lots of duplicated code
610
611 2021-08-09 Fabio M. Di Nitto <[email protected]>
612
613 [host] use correct seq_num to reclaim buffers
614 we cannot use incoming packet seq_num to reclaim buffers as that causes
615 havoc.
616
617 use latest recorded (highest) packet number instead that follows the
618 reclaiming window
619
620 [rx] increase defrag buffers to better deal with network jitter
621
622 2021-07-02 Fabio M. Di Nitto <[email protected]>
623
624 [man] update doxygen2man to latest from libqb
625
626 2021-06-30 Christine Caulfield <[email protected]>
627
628 [doc] Add comments to structure members
629 Removes doxygen warnings at build time
630
631 Needs the latest doxygen2man to actually print anything in the
632 man pages.
633
634 Signed-Off-By: Christine Caulfield <[email protected]>
635
636 2021-06-30 Fabio M. Di Nitto <[email protected]>
637
638 [link] don´t allow mix of dynamic and static links
639 with same source address/port.
640
641 The configuration doesn´t make sense in itself and confuses ACL.
642
643 [link] use previously cached value as link info might not be accurate
644
645 [link] don´t leak resources if link configuration fails
646
647 2021-06-25 Fabio M. Di Nitto <[email protected]>
648
649 [tests] suppress more libnss3 errors
650
651 [spec] cleanup some krosnonetd leftovers
652
653 2021-06-25 Christine Caulfield <[email protected]>
654
655 Add comment about log_fd in knet_handle_new()
656
657 Sanitise sockaddr_storage structs coming in the API
658 Before we used to copy the whole sockaddr_storage from
659 the API into our structs. This meant that any junk left
660 at the end by the caller got into them and could
661 cause trouble with some comparisons.
662
663 This patch zeros out the extra bits of the sockaddr_storage
664 that are not valid.
665
666 2021-06-23 Fabio M. Di Nitto <[email protected]>
667
668 Update copyright around
669 manual backport from master/rust-bindings branches
670
671 2021-06-13 Chris Walker <[email protected]>
672
673 Require make utility
674
675 2021-06-11 Fabio M. Di Nitto <[email protected]>
676
677 [tx] minor optimizations to avoid recalculating socksize over and over
678
679 2021-06-11 Christine Caulfield <[email protected]>
680
681 send 'correct' sockaddr lengths to socket calls
682 FreeBSD-devel insists that the length passed into socket calls
683 matches the expected length of the sockaddr it describes.
684 So, eg, when passing a sockaddr_in, the length must be sizeof(sockaddr_in)
685 rather than sizeof(sockaddr_storage) which we were using
686
687 Signed-Off-By: Chrissie Caulfield <[email protected]>
688
689 2021-06-03 Jan Friesse <[email protected]>
690
691 [docs] IRC channel is now on Libera Chat
692
693 2021-05-07 Fabio M. Di Nitto <[email protected]>
694
695 [coverity] fix 2 buffer initialization (potential) issues
696 discovered by latest coverity scan update
697
698 2021-04-28 Fabio M. Di Nitto <[email protected]>
699
700 Merge pull request #331 from kronosnet/stable1-proposed
701 stable1-proposed
702
703 2021-04-27 Christine Caulfield <[email protected]>
704
705 [nozzle] fix nozzle_up() on FreeBSD
706 TBH I'm not sure how this ever worked, as we never seemed to call
707 SIOCIFCREATE, but this also fixes the race when the tests are run -j4
708 (or higher)
709
710 2021-04-26 Christine Caulfield <[email protected]>
711
712 [lib] Don't use -rpath
713 rpath seems, understandably, to annoy some analysis tools. We
714 use it to locate the crypto & compression plugins but libknet also needs
715 to use the plugins from the build-tree when running tests.
716
717 A simple way to do this is have the plugin directory as part of the knet
718 handle then we can override it in the test suite, which already does
719 'illegal' accesses to some handle fields.
720
721 OK, it's not the MOST elegant way of doing this perhaps, but it is
722 simple and effective.
723
724 2021-04-19 Christine Caulfield <[email protected]>
725
726 [nozzle] Remove inet_ntoa() call
727 inet_ntoa() is deprecated in favour of inet_ntop() (see man page for
728 inet_ntop()).
729
730 2021-04-14 Fabio M. Di Nitto <[email protected]>
731
732 [handle] validate handle in public API
733 - add _is_valid_handle() function to verify if a handle
734 is known, via qb_list*
735
736 - change all (and only) public API to use _is_valid_handle().
737
738 - move library config bits to lib_config.c and keep
739 header bit in internal.h.
740 this is more of a commodity need to simplify
741 accessability of above function to the test suite
742 at link time.
743
744 2021-04-09 Fabio M. Di Nitto <[email protected]>
745
746 [global] Update copyright
747
748 [docs] Update README for BSD section
749
750 [openssl] port knet top openssl3.0-alpha13
751 openssl 3.0 is not GA yet and there are still API changes around.
752 it´s currently only available in Debian experimental and while
753 not critical for knet, we continue to keep knet updated and be ready
754 for 3.0 GA.
755
756 2021-03-20 Christine Caulfield <[email protected]>
757
758 Add some more uses for event-based tests
759 This gets time for the whole test suite (on my machine) down
760 from 10m25 to 5.56 (make check) and
761 from >120m* to 55m29 (make check-memcheck)
762
763 * I accidentally hit ^C on these tests after 2 hours so it's
764 probably much longer
765
766 tests: use callbacks to wait for nodes
767 Rather than just wait for a arbitrary amount of time for
768 nodes to come online while testing, use the knet callback
769 mechanism. This should not only save some time in the running
770 of the tests, but also make them less susceptible to slowness
771 on the CI machines.
772
773 2021-03-01 Christine Caulfield <[email protected]>
774
775 tests: Fix some random CI failures.
776
777 2021-02-24 Fabio M. Di Nitto <[email protected]>
778
779 Drop travis support
780
781 2020-12-24 Fabio M. Di Nitto <[email protected]>
782
783 [tests] hide another internal libnss non recurring memory leak
784
785 2020-10-19 Fabio M. Di Nitto <[email protected]>
786
787 Merge pull request #319 from kronosnet/stable1-proposed
788 stable1-proposed
789
790 2020-10-12 Fabio M. Di Nitto <[email protected]>
791
792 [tests] add more exceptions to deal with Ubuntu latest rebuilds (clang counterpart)
793
794 2020-10-09 Fabio M. Di Nitto <[email protected]>
795
796 [tests] add more exceptions to deal with Ubuntu latest rebuilds
797
798 2020-09-21 Fabio M. Di Nitto <[email protected]>
799
800 [cleanup] remove unused infrastructure to send HOSTINFO messages
801 those messages are unused in 1.x and if necessary they will be
802 reintroduced in 2.x
803
804 2020-09-18 Fabio M. Di Nitto <[email protected]>
805
806 [sctp] check and warn about internal fd tracker
807 fixes new coverity scan error
808
809 2020-09-09 Fabio M. Di Nitto <[email protected]>
810
811 Revert "[heartbeat] don´t send ping if transport is not connected"
812 This reverts commit 2b73aff21f6341cfd00022694c47ff9b47e17a0e.
813
814 The same code already exists at the top of the function
815
816 [heartbeat] don´t send ping if transport is not connected
817 this could cause socket buffers to fill up with stale junk and
818 generate unecessary errors.
819
820 2020-08-31 Fabio M. Di Nitto <[email protected]>
821
822 [tests] hide internal nss memory leak (non-recurring)
823
824 2020-08-26 Fabio M. Di Nitto <[email protected]>
825
826 [rx] fix data stats on RX to match TX model
827
828 2020-08-20 Fabio M. Di Nitto <[email protected]>
829
830 [rx] fix per link stats for data packets
831 and cleanup all netutils common functions
832
833 2020-08-17 kronosnet CI bot <[email protected]>
834
835 Merge pull request #314 from kronosnet/stable1-proposed
836 stable1-proposed
837
838 2020-08-13 Fabio M. Di Nitto <[email protected]>
839
840 [threads] make sure to initialize events
841 spotted on fedora rawhide + valgrind
842
843 [cleanup] drop no longer used poc-code
844
845 [cleanup] kill all unsupported kronosnetd code
846 corosync can provide similar capabilities right now
847
848 2020-08-12 Fabio M. Di Nitto <[email protected]>
849
850 [cleanup] Move public APIs calls to more suitable location
851
852 2020-08-03 Fabio M. Di Nitto <[email protected]>
853
854 [crypto] add support for openssl 3.0
855
856 2020-07-14 Fabio M. Di Nitto <[email protected]>
857
858 Merge pull request #312 from kronosnet/stable1-proposed
859 Stable1 proposed
860
861 f6b11c05112c520ccda3b28e7676acbc6b8a7938
862
863 [man] fix type comparison
864
865 Merge pull request #305 from kronosnet/stable1-proposed
866 Stable1 proposed
867
868 2020-07-13 Fabio M. Di Nitto <[email protected]>
869
870 [tests] more ignore list of nss internal stuff
871
872 [tests] ignore openssl errors triggered when building with clang
873
874 [tests] fix variable increment logic (detected by clang)
875
876 2020-07-09 Fabio M. Di Nitto <[email protected]>
877
878 Update copyright
879
880 [crypto] add support for live reconfiguration
881
882 [build] update valgrind memcheck exception file
883
884 [tests] sets defaults for coverity scan check to match CI
885
886 [tests] add basic framework to create, join and stop many nodes
887 all nodes will be using loopback interface (lo) at different
888 port offset.
889
890 [tests] use link_get_enable instead of get_status
891
892 [tests] remove unnecessary shutdown locking
893 the test should be in charge not to shutdown too much at once
894
895 2020-07-09 Jan Friesse <[email protected]>
896
897 [build] Use git-version-gen during specfile build
898 Instead of copying parts of git-version-gen for spec target use
899 git-version-gen directly and parse final version into components
900 (rpmver, alphatag, numcomm) and use them.
901
902 Main reason is to simplify code a bit (sed scripts are a bit repetitive
903 tho), reuse the code and also allow building of RPM from dist tarball
904 generated from non-tagged commit or dirty git (not very useful).
905
906 The code relies on fact, that hyphen is never used in tagged release
907 name.
908
909 [build] Fix fallback handling in git-version-gen
910 When fallback is defined and gitarchive file doesn't exists, use fallback
911 rather than UNKNOWN.
912
913 2020-07-03 Fabio M. Di Nitto <[email protected]>
914
915 [man] sync doxygen2man from libqb master
916 this change is not required in master branch
917
918 2020-06-10 Fabio M. Di Nitto <[email protected]>
919
920 [openssl] allow use of deprecated HMAC interface with openssl 3.0
921 Port will be done after rekey feature is complete (too much code conflict atm)
922
923 2020-05-13 Fabio M. Di Nitto <[email protected]>
924
925 [tests] strip extra symbol information generated by new nm
926 spotted on debian experimental
927
928 2020-04-29 Fabio M. Di Nitto <[email protected]>
929
930 Merge pull request #304 from kronosnet/am_conditional
931 Always define the Automake conditional
932
933 2020-04-29 Ferenc Wágner <[email protected]>
934
935 Always define the Automake conditional
936 Otherwise ./configure --disable-man fails with
937
938 configure: error: conditional "BUILD_DOXYXML" was never defined.
939 Usually this means the macro was only invoked conditionally.
940
941 2020-04-23 Fabio M. Di Nitto <[email protected]>
942
943 Merge pull request #294 from kronosnet/stable1-proposed
944 stable1 proposed
945
946 2020-04-22 Fabio M. Di Nitto <[email protected]>
947
948 [transport] notify transports that a link is down
949
950 [sctp] fix ifup/ifdown handling for connected sockets
951
952 [netutils] move sockaddr_len definition where it belongs
953
954 [links] fix spacing
955
956 [sctp] make sure outgoing connections are bound to a specific IP address
957 https://github.com/kronosnet/kronosnet/issues/300
958
959 if 2 or more nodes in a knet network have a duplicate IP address,
960 for example a virtbr interface with default config, sctp
961 will fail to build the associations and abort all connections.
962
963 2020-03-25 Fabio M. Di Nitto <[email protected]>
964
965 [man] use libqb version of doxygen2man when available
966 and add pkg_check_var compat macro
967
968 2020-03-18 Fabio M. Di Nitto <[email protected]>
969
970 [build] make sure to use correct libqb include files across
971
972 2020-03-17 Fabio M. Di Nitto <[email protected]>
973
974 [lists] add missing build CFLAGS for test suite
975
976 [lists] drop internal implementation and use libqb one
977
978 2020-03-16 Fabio M. Di Nitto <[email protected]>
979
980 [lists] fix build with recent gcc changes
981
982 2020-03-09 Fabio M. Di Nitto <[email protected]>
983
984 [tests] fix return code from wait_for_packet
985
986 2020-03-04 Fabio M. Di Nitto <[email protected]>
987
988 Merge pull request #288 from kronosnet/stable1-proposed
989 Stable1 proposed
990
991 2020-02-28 Fabio M. Di Nitto <[email protected]>
992
993 [stats] allow knet_link_get_status to operate in readlock context
994 - add per link stats mutex
995 - use per link stats mutex across the board
996
997 note: some threads need to lock for a slightly longer period of time than
998 strictly necessary to avoid reverse-order locking with other mutexes.
999
1000 [stats] allow knet_handle_get_stats to operate in a readlock context
1001 - add global stat mutex lock to protect stats updates
1002 - use global stat mutex lock across all the threads
1003 - fix up some minor bugs:
1004 - update RX crypto stats only when crypto is enabled
1005 - update compress and crypto stats in a consistent fashion
1006
1007 [rx] kill unused variable
1008
1009 [tests] rework test suite link port allocation
1010 Logic is to try to configure a link with port X and if it fails, try the next
1011 port. This avoids port collisions between services and knet test suite.
1012
1013 Please note that the implementation in test-common.c is NOT super clean.
1014 There is still some redundant code in there that is left on purpose.
1015
1016 There is another branch, not yet merged, that implements functional testing
1017 framework that does heavy use of those functions.
1018
1019 We will clean test-common.c as we port the functional testing branch and make
1020 it ready for merging.
1021
1022 For now, this is good enough to have a more stable test suite.
1023
1024 [transports] use SO_REUSEADDR only for sctp
1025
1026 2020-02-26 Jan Friesse <[email protected]>
1027
1028 [man] Enhance prio description of POLICY_PASSIVE
1029 Some users found description of POLICY_PASSIVE priority confusing
1030 (probably because "priority" word is too overloaded) so add
1031 some redundancy to make description unambiguous.
1032
1033 2020-02-21 Christine Caulfield <[email protected]>
1034
1035 man: Change strcat to strncat
1036 Oddly, covscan doesn't compain about the use of strcat, but
1037 I'm going to pre-empt it, just in case it decides to.
1038
1039 man: Fix covscan reports in doxyxml.c
1040 This fixes most of the remaining covscan errors in doxyxml.c.
1041 The ones that remain are caused by malloced structures being
1042 stored in qb_hashtable_maps.
1043
1044 These still cause unfreed memory, because the contents of the maps
1045 are never explictly freed, but as they are used until the very end of
1046 the program (when the OS will free everything) I'm dubious as to
1047 whether it's worth doing it in the code - or whether covscan will
1048 work out what's going on anyway.
1049
1050 2020-01-31 Fabio M. Di Nitto <[email protected]>
1051
1052 Merge pull request #267 from kronosnet/stable1-proposed
1053 stable1 proposed
1054
1055 [global] Update copyright across the board
1056
1057 [sctp] major surgery to use only SCTP events to determine socket status
1058 - drop concept of on_connected_epoll to determine if socket is ready or not
1059 - provide much better debugging output at all levels
1060 - incorporate fix from Xin Long <[email protected]> to gather socket status
1061 at the right time
1062 - deal with a recent kernel change on SCTP socket that broke knet (from rhel7):
1063 [net] sctp: allow delivering notifications after receiving SHUTDOWN
1064
1065 [rx] use defines to determine RX data types vs random numbers
1066 also extend a bit to make ready for SCTP extra return codes
1067
1068 2020-01-31 Christine Caulfield <[email protected]>
1069
1070 [tx] Don't Clear out msghdr for all transports.
1071 When sending a message to multiple links, if one of those links
1072 is not connection-oriented then msg_name & msg_namelen would be cleared,
1073 thus breaking the send to any subsequent non-connection-oriented links.
1074
1075 So now, if we need to clear out msg_name & msg_namelen, we take a copy of the
1076 msghdr and edit that instead,
1077
1078 [rx] Don't return 512 EOF messages from _recvmmsg
1079 If recvmsg() returns 0 for EOF then it's going to do so
1080 until the error is rectified or read with getsockopt(). But
1081 the _recvmmsg() wrapper keeps reading until the vector is full
1082 thus returning a block of 512 EOF messages all of which the caller
1083 has to plough through.
1084
1085 This patch causes _recvmmsg() to return as soon as it has got
1086 the first EOF so the the caller can deal with it in good time
1087 and not spin looking at the same thing over and over again.
1088
1089 I've also fixed a couple of typos in related comments
1090
1091 2020-01-31 Fabio M. Di Nitto <[email protected]>
1092
1093 [rx] send reply packets only when transport is connected
1094
1095 2020-01-30 Fabio M. Di Nitto <[email protected]>
1096
1097 [rx] unify latency values to a capped value to link precision
1098 keep the patch simple to avoid API/ABI breakage for now for easy backporting
1099
1100 [latency] fix incorrect math that could lead to bad latency calculation
1101 Also, document a bit better how latency is calculated
1102
1103 2020-01-30 Christine Caulfield <[email protected]>
1104
1105 [udp] Better fix for -ENETUNREACH
1106 This fix for the ENETUNREACH problem works better than the last one
1107 in that it also works with Linux kernels > 5.0.0 (which return
1108 -ENETUNREACH) if an interfaces is brought down, and also on FreeBSD
1109 which returns ENETDOWN.
1110
1111 2020-01-27 Fabio M. Di Nitto <[email protected]>
1112
1113 [udp] simplify code (same logic)
1114
1115 2020-01-24 Christine Caulfield <[email protected]>
1116
1117 [udp] don't make socket spin if a network I/F is down
1118 UDP treats ENETUNREACH as a temporary error and just retries,
1119 but this causes the TX thread to spin just doing sendto() therefore
1120 blocking all other traffic.
1121
1122 (To reproduce this try starting corosync with 2 links configured in
1123 corosync.conf but only one of them configured to the 'right' address
1124 - it will spin in a tight loop and need to be killed with -9)
1125
1126 SCTP does not seem to suffer from this.
1127
1128 2020-01-22 Fabio M. Di Nitto <[email protected]>
1129
1130 [nozzle] use interface name size consistently and drop strncpy in favour of memmove
1131
1132 [host] use KNET_MAX_HOST_LEN consistently
1133 detected by gcc10
1134
1135 2019-11-20 Fabio M. Di Nitto <[email protected]>
1136
1137 [tests] mark array as static
1138 fixes an odd segfault when running the test on ppc when built with clang
1139
1140 2019-11-04 wferi <[email protected]>
1141
1142 [handle] fix typo in error log message
1143
1144 2019-10-29 Fabio M. Di Nitto <[email protected]>
1145
1146 [handle] make sure to unlock config handle on failure
1147
1148 [RX] silence defrag buffer expiration debug error
1149 when using active-active links, it is simply too noisy and
1150 doesn't provide very useful information.
1151
1152 [TX] discard too big packets when reading from socketpairs
1153
1154 [RX] handle short write to the application properly
1155 this change affects only applications that are not using knet
1156 generated socketpairs to deliver/receive data to/from knet.
1157
1158 If an application uses a fd that is not SOCK_SEQPACKET (basically
1159 streaming), we have to handle short writes accordingly, and knet
1160 will continue delivering as long as there is progress.
1161
1162 The application is responsible to verify that the data packet
1163 is complete as the delivery is not guaranteed to be complete.
1164 The application can either embed the size of the packet in their
1165 data structure or use the socket error notification callback
1166 that will be invoked in case of errors or 0 data delivery.
1167
1168 [RX] Discard incoming packets if knet cannot reply back.
1169
1170 2019-10-27 Fabio M. Di Nitto <[email protected]>
1171
1172 [build] fix openssl version detection when not using pkg-config
1173
1174 2019-10-18 Ferenc Wágner <[email protected]>
1175
1176 [test] append newline to knet_send timeout message
1177
1178 2019-10-16 Fabio M. Di Nitto <[email protected]>
1179
1180 Merge pull request #259 from kronosnet/stable1-proposed
1181 Stable1 proposed
1182
1183 [test] add packet verification option to knet_bench
1184
1185 2019-10-15 Fabio M. Di Nitto <[email protected]>
1186
1187 [PMTUd] invalidate MTU for a link if the value is lower than minimum
1188 Under heavy network load and packet loss, calculated MTU can be
1189 too small. In that case we need to invalidate the link mtu,
1190 that would remove the link from the rotation (and traffic) and
1191 would give PMTUd time to get the right MTU in the next round.
1192
1193 [test] add ability to knet_bench to specify a fixed packet size for perf test
1194
1195 [rx] copy data into the defrag buffer only if we know the size of the frame
1196
1197 [host] fix defrag buffers reclaim logic
1198 The problem:
1199
1200 - let's assume a 2 nodes (A and B) cluster setup
1201 - node A sends fragmented packets to node B and there is
1202 packet loss on the network.
1203 - node B receives all those fragments and attempts to
1204 reassemble them.
1205 - node A sends packet seq_num X in Y fragments.
1206 - node B receives only part of the fragments and stores
1207 them in a defrag buf.
1208 - packet loss stops.
1209 - node A continues to send packets and a seq_num
1210 roll-over takes place.
1211 - node A sends a new packet seq_num X in Y fragments.
1212 - node B gets confused here because the parts of the old
1213 packet seq_num X are still stored and the buffer
1214 has not been reclaimed.
1215 - node B continues to rebuild packet seq_num X with
1216 old stale data and new data from after the roll-over.
1217 - node B completes reassembling the packet and delivers
1218 junk to the application.
1219
1220 The solution:
1221
1222 Add a much stronger buffer reclaim logic that will apply
1223 on each received packet and not only when defrag buffers
1224 are needed, as there might be a mix of fragmented and not
1225 fragmented packets in-flight.
1226
1227 The new logic creates a window of N packets that can be
1228 handled at the same time (based on the number of buffers)
1229 and clear everything else.
1230
1231 Fixes https://github.com/kronosnet/kronosnet/issues/261
1232
1233 [host] rename variables to make it easier to read the code
1234
1235 2019-09-26 Fabio M. Di Nitto <[email protected]>
1236
1237 [tests] add common function to sleep based on how the test suite is running
1238 Address issue while waiting for host to be up and PMTUd first run.
1239
1240 2019-09-25 Ferenc Wágner <[email protected]>
1241
1242 Fix typo: trasport -> transport
1243
1244 tests: skip the SCTP test if SCTP is not supported by the kernel
1245 For example, module loading is disabled on Debian build daemons.
1246 (In the vein of c5aa1c3343703455b480cef5c173f471e1bb020f.)
1247
1248 2019-09-20 Fabio M. Di Nitto <[email protected]>
1249
1250 Merge pull request #246 from kronosnet/stable1-proposed
1251 Stable1 proposed
1252
1253 2019-09-19 Fabio M. Di Nitto <[email protected]>
1254
1255 [links] fix memory corryption of link structure
1256 the index would overflow the buffer and overwrite data in the link
1257 structure. Depending on what was written the cluster could fall
1258 apart in many ways, from crashing, to hung.
1259
1260 Fixes: https://github.com/kronosnet/kronosnet/issues/255
1261
1262 thanks to the proxmox developers and community for reporting the issue
1263 and for all the help reproducing / debugging the problem.
1264
1265 2019-09-13 Fabio M. Di Nitto <[email protected]>
1266
1267 [tests] give PMTUd more time to redetect MTU
1268 Ideal fix would be to use PMTUd callback, but that requires a lot of
1269 extra test infrastructure. For now just workaround the problem.
1270
1271 [tests] fix ip generation boundaries
1272 https://ci.kronosnet.org/job/knet-build-all-voting/1450/knet-build-all-voting=rhel80z-s390x/console
1273
1274 and similar, when pid = 255, the secondary IP would hit 256 that is of course invalid.
1275
1276 2019-09-12 Fabio M. Di Nitto <[email protected]>
1277
1278 [nozzle] fix tapX range on newer FreeBSD
1279
1280 2019-09-10 Fabio M. Di Nitto <[email protected]>
1281
1282 [pmtud] switch to use async version of dstcache update due to locking context (read vs write)
1283
1284 2019-09-09 Fabio M. Di Nitto <[email protected]>
1285
1286 [links] stabilize latency calculation when nodes are not responsive
1287 The following scenario is more of a corner case than normal, but
1288 this change allows to better deal with this situation:
1289
1290 1) 2 nodes cluster (corosync) (node A and node B)
1291 2) kill -stop $(pidof corosync) on node A
1292 3) node B will continue to send ping packets to node A
1293 4) node A is accumulating those ping packets in the kernel network socket
1294 5) wait some seconds and unpause node A
1295 6) node A will start processing the ping packets in the queue
1296 and send pong replies to node B
1297 7) node B will see an extreme increase of latency due
1298 those "obsoleted" ping/pong packets
1299 8) node B, as latency increases, will take longer and longer
1300 to notice that node A is down due to the pong_timeout adjustment
1301 for latency (required for initial cluster spike).
1302
1303 the solution:
1304
1305 1) Use average latency to calculate pong_timeout_adj vs latency_max.
1306 Averate latency will go down again in time, while latency_max is never
1307 reset.
1308
1309 2) RX thread will filter out all pong packets that have higher latency
1310 than currently configure pong_timeout. This barrier should have
1311 been in place even before.
1312
1313 this solution reduces the latency spike on node B to a perfectly
1314 reasonable level and it will all eventually stabilize over time
1315 as latency samples increase and latency will reduce.
1316
1317 Please be aware that using a pong_timeout smaller than latency will
1318 simply mark the link down now.
1319
1320 2019-09-03 Jan Friesse <[email protected]>
1321
1322 [handle] Set thread stack size on create
1323 Musl libc has small stack size for threads. Knet needs ~300KiB (tested
1324 at the time when this patch was created). Glibc seems to use ~8MiB. As a
1325 compromise, 1MiB is used.
1326
1327 [common] Conditionalize RTLD_DI_ORIGIN
1328 RTLD_DI_ORIGIN is used to get absolute path of plugin. It is used only
1329 for logging useful info and not strictly needed, so use it only when it
1330 is defined (only musl is known to author of the patch)
1331
1332 [common] Include correct errno.h
1333 sys/errno.h is system-specific path and errno.h should be used instead.
1334
1335 2019-08-27 Jan Friesse <[email protected]>
1336
1337 [man] Fix priority description of POLICY_PASSIVE
1338 ... to match source code.
1339
1340 2019-08-21 Fabio M. Di Nitto <[email protected]>
1341
1342 [sctp] retry locking in case of failure
1343
1344 [tx] clean up channel management code for internal communications
1345 the code is still not in use but it's more clear and doesn't trigger
1346 memory overrun
1347
1348 [nozzle] fix a few coverity errors in the test suite
1349
1350 [tx] drop unnecessary usleep when sending to localhost
1351
1352 [common] make sure string is null terminated
1353
1354 [test] simplify flush log
1355 allocate on stack only once and make sure strings are null terminated
1356 drop useless read loop since log msg are always smaller than PAGE_SIZE
1357 and read are atomic at that level
1358
1359 [nozzle] avoid tons of possible buffer overruns
1360
1361 [PMTUd] do not double unlock global read lock
1362
1363 [coverity] add test targets to run coverity automatically
1364
1365 [compress] do not overrun allocated array for compress modules
1366
1367 [logging] make sure not to overrun buffers by pre-allocating them
1368
1369 [compress] don't leak memory in case of errors during zstd init
1370
1371 [nozzle] don't leak memory on error
1372
1373 [nozzle] fix negative return detected by coverity scan
1374
1375 [sctp] cleanup bugs detected in error paths by coverity scan
1376
1377 [acl] avoid forward null deferencing
1378
1379 [rx] better error report if we can't resolve hostname / port
1380
1381 [common] fix dlopen error handling
1382
1383 [tests] fix knet_bench coverity errors
1384
1385 [sctp] revalidate fd to make coverity scan happy
1386
1387 [handle] make sure that the pmtud buf contains at least knet header size
1388
1389 [rx] align data types
1390
1391 [sctp] free access list only if the socket is valid
1392
1393 [sctp] fix deference after null check
1394
1395 [coverity] add .travis.yml to integrate CI with coverity scan
1396
1397 [compress] fix #if def around BZIP2 testing
1398
1399 [compress] fix a few minor space vs tab and code formatting
1400
1401 2019-08-21 yuan ren <[email protected]>
1402
1403 [compress]Default compression level use
1404 1. add test casees for a module without default.Using default
1405 compression level.
1406 2. Discuss with Fabio, invalid compression level not the knet
1407 responsible for, so error logged. But if compress success but
1408 dstLen larger than srcLen, defualt compression level will be
1409 used, because the request level is not effective.
1410
1411 Default compress level use
1412 Discuss with Fabio, invalid compression level not the knet
1413 responsible for, so error logged. But if compress success but
1414 dstLen larger than srcLen, defualt compression level will be
1415 used, because the request level is not effective.
1416
1417 2019-08-21 Fabio M. Di Nitto <[email protected]>
1418
1419 Merge pull request #232 from kronosnet/stable1-proposed
1420 stable1-proposed
1421
1422 [build] bump soname to indicate new API call
1423
1424 [PMTUd] add ability to manually override MTU and disable PMTUd
1425
1426 [PMTUd] add dynamic pong timeout when using crypto
1427 problem originally reported by proxmox community, users
1428 observed that under pressure the MTU would flap back and forth
1429 between 2 values due to other node response timeout.
1430
1431 implement a dynamic timeout multiplier when using crypto that
1432 should solve the problem in a more flexible fashion.
1433
1434 When a timeout hits, those new logs will show:
1435
1436 [knet]: [info] host: host: 1 (passive) best link: 0 (pri: 0)
1437 [knet]: [debug] pmtud: Starting PMTUD for host: 1 link: 0
1438 [knet]: [debug] pmtud: Increasing PMTUd response timeout multiplier to (4) for host 1 link: 0
1439 [knet]: [info] pmtud: PMTUD link change for host: 1 link: 0 from 469 to 65429
1440 [knet]: [debug] pmtud: PMTUD completed for host: 1 link: 0 current link mtu: 65429
1441 [knet]: [info] pmtud: Global data MTU changed to: 65429
1442 [knet]: [debug] pmtud: Starting PMTUD for host: 1 link: 0
1443 [knet]: [debug] pmtud: Increasing PMTUd response timeout multiplier to (8) for host 1 link: 0
1444 [knet]: [debug] pmtud: Increasing PMTUd response timeout multiplier to (16) for host 1 link: 0
1445 [knet]: [debug] pmtud: Increasing PMTUd response timeout multiplier to (32) for host 1 link: 0
1446 [knet]: [debug] pmtud: Increasing PMTUd response timeout multiplier to (64) for host 1 link: 0
1447 [knet]: [debug] pmtud: PMTUD completed for host: 1 link: 0 current link mtu: 65429
1448 [knet]: [debug] pmtud: Starting PMTUD for host: 1 link: 0
1449 [knet]: [debug] pmtud: Increasing PMTUd response timeout multiplier to (128) for host 1 link: 0
1450 [knet]: [debug] pmtud: PMTUD completed for host: 1 link: 0 current link mtu: 65429
1451
1452 and when the latency reduces and it is safe to be more responsive again:
1453
1454 [knet]: [debug] pmtud: Starting PMTUD for host: 1 link: 0
1455 [knet]: [debug] pmtud: Decreasing PMTUd response timeout multiplier to (64) for host 1 link: 0
1456 [knet]: [debug] pmtud: PMTUD completed for host: 1 link: 0 current link mtu: 65429
1457
1458 ....
1459
1460 testing this patch on normal hosts is a bit challenging tho.
1461
1462 Patch was tested by hardcoding a super low timeout here:
1463
1464 diff --git a/libknet/threads_pmtud.c b/libknet/threads_pmtud.c
1465 index 4f0ba0f..5e2b89b 100644
1466 --- a/libknet/threads_pmtud.c
1467 +++ b/libknet/threads_pmtud.c
1468 @@ -261,7 +271,8 @@ retry:
1469 /*
1470 * crypto, under pressure, is a royal PITA
1471 */
1472 - pong_timeout_adj_tmp = dst_link->pong_timeout_adj * 2;
1473 + //pong_timeout_adj_tmp = dst_link->pong_timeout_adj * dst_link->pmtud_crypto_timeout_multiplier;
1474 + pong_timeout_adj_tmp = 30 * dst_link->pmtud_crypto_timeout_multiplier;
1475 } else {
1476 pong_timeout_adj_tmp = dst_link->pong_timeout_adj;
1477 }
1478
1479 and using a long running version of api_knet_send_crypto_test with a short PMTUd setfreq (10 sec).
1480
1481 [PMTUd] rework the whole math to calculate MTU
1482 internal changes:
1483 - drop the concept of sec_header_size that was completely wrong
1484 and unnecessary
1485 - bump crypto API to version 3 due to the above change
1486 - clarify the difference between link->proto_overhead and
1487 link->status->proto_overhead. We cannot rename the status
1488 one as it would also change ABI.
1489 - add onwire.c with documentation on the packet format
1490 and what various len(s) mean in context.
1491 - add 3 new functions to calculate MTUs back and forth
1492 and use them around, hopefully with enough clarification
1493 on why things are done in a given way.
1494 - heavily change thread_pmtud.c to use those new facilities.
1495 - fix major calculation issues when using crypto (non-crypto
1496 was not affected by the problem).
1497 - fix checks around to make sure they match the new math.
1498 - fix padding calculation.
1499 - add functional PMTUd crypto test
1500 this test can take several hours (12+) and should be executed
1501 on a controlled environment since it automatically changes
1502 loopback MTU to run tests.
1503 - fix way the lowest MTU is calculated during a PMTUd run
1504 to avoid spurious double notifications.
1505 - drop redundant checks.
1506
1507 user visible changes:
1508 - Global MTU is now calculated properly when using crypto
1509 and values will be in general bigger than before due
1510 to incorrect padding calculation in the previous implementation.
1511
1512 [PMTUd] fix MTU calculation when using crypto and add docs
1513
1514 [docs] add knet packet layout
1515
1516 [udp] log information about detected kernel MTU
1517
1518 [crypto] fix log information
1519
1520 [threads] allow knet_handle_setfwd to flush socket queues
1521
1522 [tests] ignore libnss errors from OpenSuse Tumbleweed
1523
1524 [tests] ignore libnss errors from OpenSuse 15
1525
1526 2019-06-12 Fabio M. Di Nitto <[email protected]>
1527
1528 Merge pull request #223 from kronosnet/stable1-proposed
1529 Stable1 proposed
1530
1531 [global] update copyrights
1532
1533 [global] clarify license entry per file to match README.licence
1534 libraries code: LGPL-2.0+
1535 binaries code and other files: GPL-2.0+
1536
1537 2019-06-11 Fabio M. Di Nitto <[email protected]>
1538
1539 [doc] fix a merge oversight from 541d7faf9068d10e12b4278c35825ce1353db081
1540
1541 [crypto] hide errors generated by openssl 1.1.1c
1542 see also:
1543 https://github.com/kronosnet/kronosnet/issues/226
1544 https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=930061#12
1545
1546 [crypto] openssl: drop calls to RAND_seed as they don´t really help RNG
1547 See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=930061#12 for reference
1548
1549 2019-06-06 yuan ren <[email protected]>
1550
1551 [crypto]openssl error strings release
1552 In versions prior to OpenSSL 1.1.0, ERR_free_strings() releases
1553 any resources created by ERR_load_crypto_strings.
1554
1555 2019-06-03 Fabio M. Di Nitto <[email protected]>
1556
1557 [transports] fix incorrect merge when cherry-picking 7033ddab505a0cf3655115fe5037579b7c882a8c
1558
1559 [crypto] fix openssl1.0 initialization code
1560
1561 [PMTUd] extend internal rerun API to allow full PMTUd reset
1562
1563 [crypto] rework knet_handle_crypto external API to be more solid
1564 the API was rather weak and could potentially leave traffic uncrypted
1565 in case of certain, corner case, failures.
1566
1567 this patch is a subset of a bigger rework of the crypto layer that
1568 will in future allow runtime reconfiguration without traffic disruption
1569 of the crypto config.
1570
1571 [crypto] make sure to trigger a PMTUd rerun on each good crypto config change
1572
1573 [PMTUd] create common/shared code to trigger PMTUd rerun
1574
1575 [crypto] make sure to clear all security info on crypto_fini
1576
1577 2019-06-03 Christine Caulfield <[email protected]>
1578
1579 misc: Fix more covscan warnings
1580 The only serious bug here is in transport_udp.c
1581 (see bottom of patch), the rest are mostly detail.
1582
1583 covscan still reports a lot of errors against doxyxml, most of
1584 which are because it doesn't understand the libqb hashtables.
1585
1586 2019-05-20 Christine Caulfield <[email protected]>
1587
1588 misc: some coverity fixes
1589 In rough order of seriousness:
1590
1591 1. Fix clock_gettime() in pmtud so that it's always called, as
1592 variable 'clock_now' is always read.
1593 2. Allow space for trailing NUL in libnozzle device names
1594 3. Fix api_nozzle_run_updown_test so it can run out of the build tree
1595 4. Disallow a 0 length prefix in libnozzle
1596 5. Fix potential use of NULL pointer on doxyxml
1597 6. Free 'name' in doxyxml as it's *not* in the map any more
1598 7. Fix dead code in libknet API functions left by code changes
1599
1600 2019-05-17 Fabio M. Di Nitto <[email protected]>
1601
1602 [spec] use ldconfig_scriptlets only when defined
1603
1604 [spec] drop support for init scripts
1605 no rpm distros left that support old fashion init scripts
1606
1607 [spec] fix a bunch of rpmlint errors
1608
1609 [spec] reconciliate fedora spec file into upstream spec file (part 1)
1610
1611 [spec] clean up useless conditionals and defines
1612 fix a couple of minor conditionals in the process
1613
1614 [spec] be more strict about plugins version and architecture depedencies
1615
1616 [spec] use distro conditionals to determine BuildRequires
1617
1618 [spec] fix upstream URLs to point to https and official release repo
1619 Fixes https://bugzilla.redhat.com/show_bug.cgi?id=1708616
1620
1621 also to be noted, the Source0: line is different from upstream and Fedora
1622 because upstream can handle tarballs during development
1623
1624 2019-05-09 Fabio M. Di Nitto <[email protected]>
1625
1626 [build] bump soname to indicate new API calls
1627
1628 [global] update copyright across the board
1629
1630 [man] fix libknet.h for errors detected by newly added test
1631
1632 [tests] improve wait for packet implementation to flush logs during wait
1633
1634 [tests] hide an arm internal memory leak (non-recurring)
1635
1636 [compress] add support for libzstd
1637
1638 2019-05-09 Chrissie Caulfield <[email protected]>
1639
1640 manpages: Document enums (#206)
1641 And also fix a bug in structure printing that caused it to print the wrong name for a struct.
1642
1643 2019-05-09 Fabio M. Di Nitto <[email protected]>
1644
1645 [tests] remove stray comment
1646
1647 [logging] fix log target of recently added API calls
1648 spotted during sctp testing
1649
1650 [access lists] add more extensive test for links_acl_ip
1651
1652 2019-05-09 Christine Caulfield <[email protected]>
1653
1654 acl: Fix English in commments
1655
1656 2019-05-09 Fabio M. Di Nitto <[email protected]>
1657
1658 [access lists] add public API tests
1659
1660 [access lists] improve checks on various data types
1661
1662 [access lists] test implicit access lists management for UDP, SCTP and LOOPBACK
1663
1664 [access lists] add external API calls to manage access lists
1665
1666 [access lists] add documentation for enable_access_list
1667
1668 [transports / access list] add internal API to gather which fd to use for access lists given a certain link struct
1669 this is required for the external API that has to be transport indepedent
1670
1671 [access lists] rename ip1/2 to ss1/2 to keep it more generic
1672
1673 [access lists] use arrays to access per-protocol functions
1674
1675 [access lists] use better name for fd_tracker structure
1676
1677 [access lists] confine access lists data structs within the protocol itself
1678
1679 [access lists] add errno around and start using them
1680
1681 [access lists] fix build on BSD and add some include files around
1682
1683 [access lists] make internal API consistent
1684
1685 [links] rename tranport_type to transport to avoid confusion (part 2)
1686 complements be9d053efafc822cabd696914d53b5dfe25fb4fd due to early
1687 cherry-pick of 7033ddab505a0cf3655115fe5037579b7c882a8c
1688
1689 [links] rename transport_type to transport to avoid confusion
1690
1691 [access lists] remove 2 unnecessary wrappers
1692
1693 [access lists] cleanup API a bit
1694
1695 [access lists] more use of generic wrappers and remove duplicate code
1696
1697 [access lists] move access lists structs and data types to links_acl.*
1698
1699 [access lists] move all acl wrappers to links_acl* and split links_acl_ip to their own files
1700
1701 [access lists] fix build on freebsd
1702 don't use malloc.h, obsoleted by stdlib.h
1703 define s6_addr32 that's only available in kernel space
1704
1705 [access lists] add access lists support to sctp
1706
1707 [access lists] enable generic access lists only for protocols that use them
1708 protocols such as SCTP that use their own access list tracking will
1709 need to setup access lists in transport_link_set/clear_config
1710
1711 [access lists] enable access lists for GENERIC_ACL protocols (udp for example)
1712
1713 [access lists] allow knet_bench to enable/disable access lists
1714
1715 [access lists] add tests for default access lists
1716
1717 [access lists] automatically add and remove point to point access lists
1718 those are not used just yet.
1719
1720 [handle] properly initialize fd tracker buffers
1721
1722 [access lists] make code more generic to accept more than IP protocol and start to bind it to each fd
1723 access lists are unique per file descriptor, each fd can have its own protocol and list.
1724
1725 remane around ipcheck* with check* to be more generic.
1726
1727 [transports] add information about the nature of the transport and supported access lists
1728
1729 [acl] add knet_handle_enable_access_lists api call
1730
1731 [acl] move poc-code into libknet dir and rename to links_acl.*
1732 code is not integrated yet and test suite can´t run standalone
1733
1734 2019-05-09 kronosnet CI bot <[email protected]>
1735
1736 Merge pull request #211 from kronosnet/stable1-proposed
1737 stable1-proposed
1738
1739 2019-05-02 Fabio M. Di Nitto <[email protected]>
1740
1741 [udp] improve error message decoding from ICMP errors
1742
1743 [udp] use defines vs hardcoded numbers
1744
1745 2019-05-01 Fabio M. Di Nitto <[email protected]>
1746
1747 [man] fix libknet.h for errors detected by newly added test
1748
1749 2019-04-30 Fabio M. Di Nitto <[email protected]>
1750
1751 [tests] add man page check to verify doxy header order and definitions
1752
1753 [man] fix knet_host_set_policy parameters order
1754
1755 2019-04-30 Christine Caulfield <[email protected]>
1756
1757 man: Tidy more man pages
1758 Followup to previous 'tidy'
1759
1760 2019-04-30 Chrissie Caulfield <[email protected]>
1761
1762 man: Tidy manpages (#215)
1763 * man: Tidy manpages for libnozzle
1764
1765 doxygen works in mysterious ways, adding a blank line before
1766 @brief makes the lines following that much tidier.
1767
1768 So now instead of
1769
1770 nozzle_close nozzle - pointer to the nozzle struct to destroy
1771
1772 we get:
1773
1774 nozzle_close
1775
1776 nozzle - pointer to the nozzle struct to destroy
1777
1778 * doxyxml: Cope with pointers-to-pointers passed as params
1779
1780 Double pointers showed as ' * *name' when they should be ' **name'.
1781
1782 Also tidy STRUCTURES display so that they are not indented too much,
1783
1784 * man: Similar @brief fixes for libknet.h
1785
1786 * doxyxml: Tidy descriptions of functions as parameters
1787
1788 If a complex function pointer was passed as a parameter then doxyxml
1789 tryied to line up all the other parameters with it - making a mess
1790 by having lots of blank space between the type and the name.
1791
1792 Now we enforce a maximum type length (a line-ish) so that shorter
1793 tyopes will line up OK and the really long ones will be left to their
1794 own devices.
1795
1796 2019-04-11 Fabio M. Di Nitto <[email protected]>
1797
1798 [crypto] remove libnss 3des support
1799
1800 2019-04-09 Fabian Grünbichler <[email protected]>
1801
1802 reduce minimum crypto key size to 1024bit
1803 Since the key is used for AES/3DES and HMAC operations only, this is
1804 safe. AES/3DES use keys in the 128- to 256-bit range, HMAC with
1805 MD5/SHA1/SHA2 should use keys with a minimum of 128- to 512-bit (in both
1806 cases, depending on the actual algorithm used).
1807
1808 This reduction also keeps knet compatible with existing Corosync 2.x
1809 keyfiles, which are 1024-bit.
1810
1811 2019-03-27 Fabio M. Di Nitto <[email protected]>
1812
1813 [build] add another exception to valgrind nss combo
1814
1815 Merge pull request #199 from kronosnet/stable1-proposed
1816 stable1-proposed
1817
1818 2019-03-26 Fabio M. Di Nitto <[email protected]>
1819
1820 [global] update copyright across the board
1821
1822 2019-03-26 Christine Caulfield <[email protected]>
1823
1824 link: Check address families on a link always match
1825 We can't create a link with a source address of a different
1826 address family than the destination as all sends will fail.
1827
1828 2019-03-26 Fabio M. Di Nitto <[email protected]>
1829
1830 [transports] fix usage after free of transport info memory
1831 spotted while implementing functional testing for rekey.
1832
1833 when configuring multiple hosts (3+) with 1 link sharing
1834 the same listener, but NOT enabling the link, if the user
1835 attempts to shutdown knet_h, the first call to knet_link_clear_config
1836 will not recognize that the listener is still in use by another
1837 link and will release all resources.
1838 At the time of invoking clear_config on the second link, transport_clear_config
1839 will access already freed memory, or alternatively, a call to
1840 link_enable would cause a crash because the listener does no
1841 longer exist.
1842
1843 [tests] test rpm should match currently installed version of libknet
1844
1845 2019-03-12 Chrissie Caulfield <[email protected]>
1846
1847 manpages: Document enums (#206)
1848 And also fix a bug in structure printing that caused it to print the wrong name for a struct.
1849
1850 2019-03-11 Fabio M. Di Nitto <[email protected]>
1851
1852 [tests] allow shipping of the test suite
1853
1854 [host] increase log level for currently used links
1855 bug reported on IRC by yuanren
1856
1857 [transport] don´t specify destination address for connection oriented protocols
1858 some implementations of sendto/sendmsg expects dst_addr / msg_name to be
1859 empty for connection oriented protocols such as SCTP (spotted on FreeBSD).
1860
1861 Make sure to clear or not pass those information on connection oriented protocol.
1862
1863 2019-03-10 Fabio M. Di Nitto <[email protected]>
1864
1865 [man] fix man page build dependency for BSD Make
1866
1867 [tests] add ability to create ipv6 sockets
1868
1869 [man] fix libnozzle man page generation target
1870
1871 [tests] fix knet_bench help text
1872
1873 2019-02-28 Christine Caulfield <[email protected]>
1874
1875 doxyxml: Don't crash if there's no doxygen stanza for a function
1876
1877 2019-02-28 Fabio M. Di Nitto <[email protected]>
1878
1879 [man] add test to make sure all generated man pages are being shipped / installed
1880
1881 2019-01-30 kronosnet CI bot <[email protected]>
1882
1883 Merge pull request #183 from kronosnet/stable1-proposed
1884 Stable1 proposed
1885
1886 2019-01-24 Jan Pokorný <[email protected]>
1887
1888 [common] fix not reporting a true dlinfo error cause
1889 This was automatically caught with GCC 9
1890 ("'%s' directive argument is null").
1891
1892 2019-01-07 Fabio M. Di Nitto <[email protected]>
1893
1894 [libknet] allow better handling of internal threads
1895 while playing around adding some dynamic internal threads, i did
1896 hit a limitation with allocating a dynamic THREAD_MAX value based
1897 on OS availability of certain features.
1898
1899 this patch changes from dynamic to static allocation of threads, that
1900 can always be increased if necessary without breaking anything.
1901
1902 simplify thread numbering and conditionals on OS features
1903
1904 improve debugging of thread status by adding a 2 phase thread registration
1905 process:
1906
1907 - register a thread before starting it with THREAD_REGISTERED before pthread_create
1908 - set a thread status as started with THREAD_STARTED
1909
1910 any failure in between can now be detected in debugging logs.
1911
1912 Note for developers: all threads *MUST* be registered/started before
1913 wait_all_threads_status check in handle.c.
1914
1915 2019-01-04 Fabio M. Di Nitto <[email protected]>
1916
1917 [build] enable all compilers warnings at once
1918
1919 2019-01-03 Fabio M. Di Nitto <[email protected]>
1920
1921 [libknet] expose API version in libknet.h to make it easier for applications to add conditional builds
1922
1923 2018-12-28 Ferenc Wágner <[email protected]>
1924
1925 Skip SCTP event initialization if the protocol is not supported
1926 We may not want to actually use SCTP after all, and if we do, we won't
1927 get to use the uninitialized subscribe struct anyway. Thus aborting
1928 handle initialization is premature here.
1929
1930 Fix nozzle man dates to SOURCE_EPOCH, just like we fix the libknet ones
1931
1932 2018-12-19 Fabio M. Di Nitto <[email protected]>
1933
1934 [rpm] ship libnozzle man pages
1935
1936 2018-12-19 Ferenc Wágner <[email protected]>
1937
1938 Fix typos
1939
1940 Install the nozzle man pages
1941
1942 2018-12-19 kronosnet CI bot <[email protected]>
1943
1944 Merge pull request #162 from kronosnet/stable1-proposed
1945 stable1 proposed
1946
1947 2018-12-18 Fabio M. Di Nitto <[email protected]>
1948
1949 [knet] fix build with older openssl versions
1950 spotted on freebsd-devel
1951
1952 [global] update notes to package maintainers
1953
1954 2018-12-18 Ferenc Wágner <[email protected]>
1955
1956 Tabify
1957
1958 Use the kernel struct size for SCTP event subscription
1959
1960 2018-12-18 Fabio M. Di Nitto <[email protected]>
1961
1962 [libnozzle] fix variable assignment order
1963
1964 [global] fix copyright across
1965
1966 [libnozzle] add nozzle_get_ips tests and kill nozzle_test
1967
1968 [libnozzle] add nozzle_del_ip tests
1969
1970 [libnozzle] fix del_ip removal of Ipv6 addresses when MTU < 1280
1971
1972 [libnozzle] add nozzle_add_ip tests
1973
1974 [libnozzle] add api_nozzle_run_updown tests
1975 and cleanup lots of craft around related to the old testing method
1976 for the same API
1977
1978 [libnozzle] add internal execute_bin_sh_command test
1979
1980 [libnozzle] cleanup tests Makefile.am
1981
1982 [libnozzle] add nozzle_get_fd test
1983
1984 [libnozzle] add nozzle_get_name_by_handle test
1985
1986 [libnozzle] add nozzle_get_handle_by_name tests
1987
1988 [libnozzle] add nozzle_get_mac / nozzle_set_mac / nozzle_reset_mac tests
1989
1990 [libnozzle] add nozzle_get_mtu / nozzle_set_mtu / nozzle_reset_mtu tests
1991
1992 [libnozzle] move multi nozzle device test to api_nozzle_open
1993
1994 [libnozzle] clean up code in set_up test
1995
1996 [libnozzle] improve set_down test
1997
1998 [libnozzle] add api_nozzle_set_(up|down) tests
1999
2000 [libnozzle] drop unnecessary indirection to call set_down
2001
2002 [libnozzle] drop redundant check
2003
2004 [libnozzle] add api_nozzle_close_test
2005 also, share more code in test-common, fix another BSD build issue
2006
2007 [libnozzle] fix test suite build on BSD
2008
2009 [libnozzle] add api_nozzle_open test
2010
2011 [libnozzle] prepare Makefile.am to explode with API tests
2012 also fix nozzle_test for using global compiler flags
2013
2014 [libnozzle] fix API check paths
2015
2016 [libnozzle] add per-API call test coverage check
2017
2018 [libnozzle] re-instate error checking in nozzle_get_ips
2019
2020 [libnozzle] add comment about BSD ip_add status
2021
2022 [libnozzle] cleanup nozzle_get_ips public API
2023
2024 [libnozzle] fix whitespaces
2025
2026 [global] Update copyright across
2027
2028 [libnozzle] Update copyright
2029
2030 [libnozzle] fix build on BSD
2031
2032 [libnozzle] remove error_string requirement from many API calls and mark them as final
2033
2034 [libtap] move nozzle_run_updown to completed section
2035
2036 [tests] properly check symbols on PPC64
2037
2038 [tests] remove debugging sleep()
2039
2040 [libnozzle] enable libnl3 < 3.3 build workaround
2041 fixes build on opensuse-42
2042
2043 [libnozzle] Use libnl3.0 for ip address management on Linux
2044 BSD will follow next using internal ioctls and drop usage of forking a shell
2045
2046 [libnozzle] whitespace cleanup
2047
2048 [nozzle] expand buffers to deal with strncpy and -Werror=stringop-truncatio
2049
2050 [nozzle] drop duplicated code
2051
2052 [nozzle] enable libnozzle build by default
2053
2054 [nozzle] move man pages to top level dir
2055
2056 [nozzle] fix man page generation
2057
2058 [nozzle] propagate errors all the way
2059
2060 [nozzle] fix build on BSD
2061
2062 [nozzle] nozzle_open and nozzle_close should be stable now
2063
2064 [nozzle] move set up and down to the stable API section
2065
2066 [nozzle] decouple running pre-up.d/up.d/down.d/post-down.d from interface status (part 4)
2067
2068 [nozzle] decouple running pre-up.d/up.d/down.d/post-down.d from interface status (part 3)
2069
2070 [nozzle] decouple running pre-up.d/up.d/down.d/post-down.d from interface status (part 2)
2071
2072 [nozzle] decouple running pre-up.d/up.d/down.d/post-down.d from interface status (part 1)
2073 provide the facility to do it via nozzle_run_updown but delegate the task to
2074 the application.
2075
2076 This has the benefit of much better fine grained control over errors during those
2077 code paths.
2078
2079 [nozzle] move find_ip to internals
2080
2081 [nozzle] move ipv4 broadcast to internals
2082
2083 [nozzle] cleanup mac address functions
2084
2085 [nozzle] move code around to reflect public API and internal functions
2086 cleanup _get_mtu in the process
2087
2088 [nozzle] move code around to reflect completion and drop copy of code from internals.c
2089
2090 [nozzle] rename _check to is_valid_nozzle, make it more useful and fixup error code checking
2091
2092 [nozzle] be more explicit on function naming (part 2)
2093
2094 [nozzle] be more explicit on function naming
2095
2096 [nozzle] start cleaning internal function names
2097
2098 [nozzle] split code around to separate libnozzle and testsuite
2099
2100 [build] split tests, man pages and build dirs
2101
2102 [nozzle] rename lib_mutex to config_mutex
2103
2104 [build] fix build on BSD
2105
2106 [cleanup] document nozzle_iface struct, drop unnecessary ifr struct and cleanup names
2107
2108 [cleanup] rename ip_addr to ipaddr
2109
2110 [cleanup] rename sockfd to ioctlfd
2111
2112 [cleanup] rename struct _ip to struct nozzle_ip and cleanup related defines
2113
2114 [nozzle] rename _config to nozzle_lib_config
2115
2116 [cleanup] start splitting the code across multiple files
2117
2118 [nozzle] cleanup and document nozzle_get_fd
2119
2120 [nozzle] cleanup and document nozzle_get_name_by_handle
2121
2122 [nozzle] cleanup and document nozzle_get_handle_by_name
2123
2124 [nozzle] cleanup and document nozzle_(re)set_mac
2125
2126 [nozzle] cleanup and document nozzle_get_mac
2127
2128 2018-12-18 Christine Caulfield <[email protected]>
2129
2130 Tidy some English
2131
2132 2018-12-18 Fabio M. Di Nitto <[email protected]>
2133
2134 [nozzle] cleanup and document nozzle_set_mtu and nozzle_reset_mtu
2135 couldn't split those 2 because they are stricly related
2136
2137 [nozzle] cleanup and document nozzle_get_mtu
2138
2139 [nozzle] fix test on BSD
2140
2141 [nozzle] cleanup and document nozzle_get_ips (part 1)
2142
2143 [nozzle] cleanup and document nozzle_del_ip
2144
2145 [nozzle] cleanup and document nozzle_add_ip
2146
2147 [nozzle] re-order functions in order of importance
2148
2149 [nozzle] cleanup and document nozzle_set_down
2150
2151 [nozzle] cleanup and document nozzle_set_up
2152
2153 [nozzle] start to cleanup nozzle_close to be more informative and effective
2154
2155 [nozzle] cleanup nozzle_open coding style and return codes
2156 also update the header file for Doxygen man pages
2157
2158 [nozzle] rename last tap bits to nozzle
2159
2160 [nozzle] rename API from tap_ to nozzle_
2161
2162 [nozzle] fix pkg-config to point to nozzle
2163
2164 [nozzle] rename struct _iface to nozzle_iface (looks better on public headers)
2165
2166 [nozzle] rename tap_t to nozzle_t
2167
2168 [nozzle] include net/if.h to define IFNAMSIZ instead of redefine
2169
2170 [nozzle] add Doxygen header
2171
2172 [libnozzle] rename libtap to libnozzle (part 1)
2173 Our fearless Debian maintainer wferi made us aware that TAP is an overloaded
2174 term already and it would be very confusing to add yet another one.
2175
2176 nozzle is a good pick from http://www.thesaurus.com/browse/tap
2177
2178 I am sure lib*cock* would have cause havoc :-)
2179
2180 This commit only rename files and allow builds. The API needs to be renamed too,
2181 but that will come over time with major code cleanup as documented here:
2182 https://trello.com/c/pyEKd4UH/396-libnozzle-make-it-production-ga-ready
2183
2184 [libtap] BSD: switch to ioctl to destroy tap devices
2185
2186 [libtap] define IFNAMSIZ for applications that don't need/want to include net/if.h
2187
2188 [libtap] update README for BSD
2189
2190 [libtap] port to freebsd and fix a bunch of small bugs
2191 - tap_open: do mutex lock only after malloc to avoid deadlock
2192 - fix test suite to increase randomness when testing named tap devices
2193 - remove hardcoded paths to /bin/true /bin/false and /bin/grep
2194
2195 [libtap] tests: add better IPv6 private address space randomization
2196
2197 [libtap] tests: allow random localhost IPs for testing
2198 - avoids clashing with machine routing
2199 - allows parallel testing
2200
2201 2018-12-13 Christine Caulfield <[email protected]>
2202
2203 UDP: Cope with EPERM returned from sendmsg/sendto
2204 IPtables can cause -EPERM to be returned from socket sends, so
2205 we need to be able to handle it without going into a spin.
2206
2207 SCTP seems to be unaffected.
2208
2209 2018-12-03 Jan Pokorný <[email protected]>
2210
2211 [misc] fix a few typos/stylistics
2212
2213 2018-11-27 Chen Jingpiao <[email protected]>
2214
2215 [tests] update error message in api_knet_{addrtostr,strtoaddr}.c
2216
2217 [tests] fix check condition in api_knet_{addrtostr,strtoaddr}.c
2218
2219 clear the errno in each exported API when the call success
2220
2221 2018-11-20 Fabio M. Di Nitto <[email protected]>
2222
2223 [openssl] fix return type for pthread_self
2224 spotted on freebsd-devel
2225
2226 2018-11-16 Chen Jingpiao <[email protected]>
2227
2228 [tx] fix tx_uncompressed_packets stats collect in _parse_recv_from_sock
2229
2230 2018-11-13 Chen Jingpiao <[email protected]>
2231
2232 [rx] fix incorrect packet length in _parse_recv_from_links
2233
2234 [tx] remove goto statement to improve readability
2235
2236 Add static qualifier to {compress, crypto, transport}_modules_cmds
2237
2238 2018-11-12 Chen Jingpiao <[email protected]>
2239
2240 [transports] fix transport_modules_cmd array out of range
2241 Fixes: 18ab4888
2242
2243 2018-11-07 Fabio M. Di Nitto <[email protected]>
2244
2245 [build] fix supported compiler warning detection
2246 move from AC_PREPROC_IFELSE (strongly discouraged) to AC_COMPILE_IFELSE
2247
2248 our detection system was very weak and recent versions of clang did
2249 show that PREPROC_IFELFE (cpp) would enable warning options that
2250 the compiler does not support (clang).
2251
2252 use a full compilation test to detect what works and what doesn't.
2253
2254 Also expand the warning list to include new / renamed clang options
2255 of equivalents already enabled for older versions of clang and gcc.
2256
2257 2018-11-07 Chen Jingpiao <[email protected]>
2258
2259 tests: fix output message
2260 Fixes: 04537c57 ("[transports] cleanup API calls")
2261
2262 fix typo
2263
2264 man: update XML directory in doxyxml.c
2265 Fixes: 31945b18 ("[man] move libknet man pages and man pages build tools to top level dir")
2266
2267 2018-10-23 Christine Caulfield <[email protected]>
2268
2269 misc: keep covscan happy
2270 covscan complains about is ignoring returns from pthread_mutex_lock()
2271 but in those 2 circumstances we don't have much choice.
2272
2273 This patch just keeps coverity quiet so we don't miss any real bugs it
2274 might find.
2275
2276 2018-10-22 Fabio M. Di Nitto <[email protected]>
2277
2278 Merge pull request #155 from kronosnet/stable1-proposed
2279 Stable1 proposed
2280
2281 2018-10-21 Christine Caulfield <[email protected]>
2282
2283 links: Don't close loopback link
2284 knet_link_clear_config would incorrectly clear the local
2285 loopback link if the last 'real' link to an external node
2286 was closed. We didn't check the nodeid in the code, just
2287 the link number when clearing the loopback flag.
2288
2289 2018-10-08 Jan Friesse <[email protected]>
2290
2291 Import key with length not dividable by wrap key block size
2292 Wrapping of the key is standard crypto operation which needs data
2293 aligned to cipher block size, otherwise it fails.
2294
2295 Possible solution is to use a zero filled buffer with size aligned to
2296 required wrap key block size. Private key is copied to the beginning of
2297 the buffer and unwrap operation keeps using only required private key
2298 size.
2299
2300 2018-09-11 Fabio M. Di Nitto <[email protected]>
2301
2302 [tests] clear errno to avoid incorrect propagation from void fn call
2303
2304 [tests] fix select(2) handling (mostly affects slow machines)
2305
2306 2018-09-06 Fabio M. Di Nitto <[email protected]>
2307
2308 [threads] make it easier to debug thread startup / stop operations with pretty names
2309
2310 2018-09-05 Jan Friesse <[email protected]>
2311
2312 Use gitarchive-version for spec file
2313
2314 git-version-gen fail on UNKNOWN version
2315
2316 Do not match lightweight tags in git-version-gen
2317 Fix leftover from 175b517fd71d999150aa701ec847510490a18a6e which added
2318 matching of lightweight git tags.
2319
2320 Support for git archive stored tags
2321 Attempt to solve problem with git archive generated tarballs
2322 (used for example by github when release is downloaded) which are no
2323 longer git tree and (in contrast to officially released tarballs) also
2324 doesn't contain .tarball-version file so git-version-gen script simply
2325 cannot obtain valid version info.
2326
2327 Solution is based on using gitattributes which is instructs git to
2328 replace string in the .gitarchivever file by known ref names.
2329 git-version-gen is enhanced to support this file and tries to parse
2330 any string which looks like "tag: v[0-9][^,)]*". If such string
2331 is found it's used as a version. This file is used as a last attempt and
2332 other methods (.tarball-version, git abbrev) have precedence.
2333
2334 Based on idea stated by Jan Pokorný <[email protected]>.
2335
2336 2018-08-16 Fabio M. Di Nitto <[email protected]>
2337
2338 [tests] drop arm exceptions as they have been fixed in glibc
2339
2340 [tests] cleanup nss memcheck supports
2341 add one more exception for newer nss (spotted on fedora rawhide)
2342 drop FreeBSD excetions as we don't support running under valgrind on BSD
2343
2344 2018-08-07 Fabio M. Di Nitto <[email protected]>
2345
2346 Merge pull request #150 from kronosnet/stable1-proposed
2347 stable1-proposed
2348
2349 2018-08-03 Fabio M. Di Nitto <[email protected]>
2350
2351 [threads] use internal array to track status of threads (running/stopped) vs a random sleep()
2352
2353 [tests] don't redefine symbols from the main library
2354
2355 [tests] adds more overrides for fedora28/arm/glibc
2356
2357 [handle] reduce usage of handle_config_mutex to wrap only global lib init
2358
2359 [logging] report error in case logging buffer is bigger than PIPE_BUF
2360 See https://github.com/kronosnet/kronosnet/pull/148
2361
2362 [handle] drop init_lock_done as it was used only by locked version of log_msg
2363
2364 [logging] drop locking inside log_msg (part 1)
2365 looking at the code, there are only few calls to log_msg happening outside
2366 of locking context (error reports when failing to grab the lock).
2367 Everything else appears to be safe.
2368
2369 drop the attempt to lock inside log_msg for now, part 2 will address those
2370 few calls that can happen outside of locking context.
2371
2372 [handle] drop call to log_msg that would have no effect
2373 logging is not setup yet at this stage and we can't log just yet.
2374
2375 2018-07-17 Fabio M. Di Nitto <[email protected]>
2376
2377 [tests] ignore nss internal memory leak (3.38+) non recurring
2378
2379 [tests] reference correct library in memcheck comments
2380
2381 2018-06-08 Fabio M. Di Nitto <[email protected]>
2382
2383 [tests] add another lzma exception
2384 spotted on Debian Experimental, the latest toolchain appears to add
2385 full path to objects and valgrind would generate an error.
2386
2387 add a generic catch-all path exception.
2388
2389 2018-04-26 Ferenc Wágner <[email protected]>
2390
2391 coding style: add parentheses after the sizeof operator
2392
2393 log_msg: no point in putting NL after the final NUL
2394
2395 knet_log_msg: fix documentation: the closing newline is added by flush_logs
2396
2397 _logthread: while loop instead of goto
2398
2399 _logthread: select() returns the number of set bits, as an int
2400
2401 flush_logs: don't ever print past the message buffer
2402
2403 flush_logs: while loop instead of goto
2404
2405 flush_logs: a for loop is more straightforward here
2406
2407 flush_logs: don't repeat the type of msg
2408
2409 flush_logs: the reads will overwrite the entire msg structure
2410
2411 flush_logs: remove redundant condition
2412 The read loop is always entered, resetting len and returning from the
2413 function if it isn't greater than 0.
2414
2415 2018-04-23 Fabio M. Di Nitto <[email protected]>
2416
2417 Merge pull request #145 from kronosnet/stable1-proposed
2418 stable1-proposed
2419
2420 2018-04-20 Fabio M. Di Nitto <[email protected]>
2421
2422 [crypto] NSS_NoDB_Init: the parameter is reserved, must be NULL
2423 Thanks to Feri for spotting it in corosync
2424
2425 [crypto] improve checks around RX packet size
2426
2427 2018-04-19 Fabio M. Di Nitto <[email protected]>
2428
2429 Merge pull request #140 from kronosnet/stable1-proposed
2430 stable1-proposed
2431
2432 2018-03-27 Ferenc Wágner <[email protected]>
2433
2434 build: look for source_epoch in srcdir
2435
2436 build: check for SOURCE_DATE_EPOCH before turning to Git for an epoch
2437
2438 2018-03-27 Fabio M. Di Nitto <[email protected]>
2439
2440 [build] fail and provide proper error if source tree version is unknown
2441
2442 2018-03-27 Ferenc Wágner <[email protected]>
2443
2444 build: git -C is a relatively new option, avoid it
2445
2446 build: accomodate to the actual date utility
2447
2448 spec: use the source epoch here as well
2449
2450 man: use new doxyxml date options to build man pages
2451 Based on Fabio's original patch.
2452
2453 2018-03-27 Christine Caulfield <[email protected]>
2454
2455 man: Add date & year options to doxyxml
2456
2457 2018-03-27 Ferenc Wágner <[email protected]>
2458
2459 build: distribute the source_epoch file
2460
2461 build: determine SOURCE_EPOCH during configuration
2462
2463 2018-02-26 Fabio M. Di Nitto <[email protected]>
2464
2465 [compress] extend check to compare end buffer data
2466
2467 [compress] update valgrind exception file to match new lzma entry points
2468
2469 [compress] delegate compress_level validation to compress libraries
2470 - make val_level internal api call optional
2471 - keep lzo2 val_level around due to the specific nature of compress_level values
2472 - add internal compress_lib_test to do a round-robin (compress/decompress) check with provided values
2473 - drop unnecessary val_level checks around
2474 - update internal compress API include file
2475 - adjust api_knet_handle_compress test to use a known bad value for zlib validation
2476
2477 [tx/rx] improve error checking when running external code
2478 libknet cannot guarantee that the dst_host_filter is sane nor that it cannot
2479 be exploited to return garbage to knet.
2480
2481 there is an infinitesimal possibility that, if the plugin returns total crap,
2482 knet could crash by trying to access out-of-bound memory.
2483
2484 Prevent that by adding specific checks both in TX/RX.
2485
2486 2018-02-25 Fabio M. Di Nitto <[email protected]>
2487
2488 Merge pull request #116 from kronosnet/stable1-proposed
2489 stable1-proposed fixes
2490
2491 2018-02-23 Fabio M. Di Nitto <[email protected]>
2492
2493 [build] fix ChangeLog generation in release tarball
2494
2495 2018-02-20 Ferenc Wágner <[email protected]>
2496
2497 Indicate new API call in libtool version info
2498
2499 build: provide -version-info directly instead of -version-number
2500 The info components carry the real OS-independent library version
2501 information, from which the version number is computed differently
2502 on different systems. According to the libtool manual "new projects
2503 should use the -version-info flag instead".
2504
2505 New API function: knet_handle_new_ex() taking handle flags
2506 For now, the only supported flag is KNET_HANDLE_FLAG_PRIVILEGED,
2507 which is also the default in knet_handle_new().
2508
2509 The tests are adapted to switch this flag off, which makes them pass
2510 in default environments without privileges.
2511
2512 2018-02-20 Christine Caulfield <[email protected]>
2513
2514 UDP: Don't keep retrying send if we get EINVAL
2515 UDP can return EINVAL if the route is down or not configured, this can
2516 cause knet to 'spin' (well, ish, there is a usleep in there).
2517
2518 With this patch in place knet doesn't hog the CPU anywhere near so much
2519 (and it does recover if the link returns).
2520
2521 2018-02-19 Christine Caulfield <[email protected]>
2522
2523 doxyxml: Use overridden XML_DIR for reading structures
2524
2525 2018-02-19 Fabio M. Di Nitto <[email protected]>
2526
2527 [man] fix path to stamp file
2528
2529 [man] move kronosnetd man pages
2530
2531 [man] move libknet man pages and man pages build tools to top level dir
2532
2533 2018-02-18 Fabio M. Di Nitto <[email protected]>
2534
2535 [build] simplify Doxygen test based on Feri's input
2536
2537 [build] be consistent
2538
2539 [build] make man page build optional
2540
2541 [build] check for all *int types as we use pretty much all of them
2542
2543 [build] drop redundant check (it's duplicated just a few lines below)
2544
2545 [build] drop unnecessary function checks
2546
2547 [build] drop unnecessary include checks
2548
2549 [build] fail configure if we can't find Doxygen to build man pages
2550
2551 [build] drop obsolete AC_PROG_LIBTOOL included in LT_INIT
2552
2553 [build] enable error when checking --enable-new-dtags linker flags
2554
2555 2018-02-13 Ferenc Wágner <[email protected]>
2556
2557 build: no binaries in build-aux wander out of the build tree
2558
2559 build: better separation of flags during cross-compilation
2560
2561 build: kronosnetd still needs libqb on the host system
2562
2563 build: compile and link doxyxml for the build host
2564
2565 build: use build architecture libraries for doxyxml
2566
2567 2018-02-12 Ferenc Wágner <[email protected]>
2568
2569 Better wording suggested by Chrissie
2570
2571 Reword the documentation of knet_handle_compress_cfg
2572
2573 We support several compression libraries by now.
2574
2575 2018-02-03 Fabio M. Di Nitto <[email protected]>
2576
2577 [compress] lz4: fix builds for older lz4 releases
2578
2579 [compress] lz4: include all required headers
2580 upon reading lz4hc.h, lz4.h should always be included as depedency
2581
2582 2018-02-03 Ferenc Wágner <[email protected]>
2583
2584 doxyfile: description of return values may be missing
2585
2586 2018-02-03 Fabio M. Di Nitto <[email protected]>
2587
2588 [man] drop dist_man_MANS leftover
2589
2590 2018-02-01 Fabio M. Di Nitto <[email protected]>
2591
2592 [build] make gcc8 buffer boundaries check happy
2593
2594 2018-01-30 Ferenc Wágner <[email protected]>
2595
2596 doxyxml: use AM_CFLAGS for debugging info, optimization and warnings
2597 The rework of the FLAGS handling and the development of doxyxml happened
2598 on different temporary branches, and this slipped through on the merge.
2599
2600 tests: skip the SCTP test if SCTP is not supported by the kernel
2601 For example, module loading is disabled on Debian build daemons.
2602
2603 build: --export-dynamic is needed for dlopened modules only
2604 Not that it matters on ELF platforms the least. If Kronosnet is ever
2605 ported to Windows, however...
2606
2607 build: linking with libm is unnecessary on some systems (like OS X)
2608 Inlining of builtin functions can eliminate the libm dependency on other
2609 systems as well, but AC_SEARCH_LIBS can't detect this because it tests
2610 with a false prototype. Thus we use --as-needed linking already to trim
2611 such unnecessary dependencies.
2612
2613 build: use --as-needed to avoid unnecessary dependency on libm
2614 I removed --as-needed in 1b96703, but now I found a good reason to use
2615 it after all.
2616
2617 build: check for the --enable-new-dtags flag before using it
2618
2619 build: sanitize CFLAGS handling
2620 As for LDFLAGS previously: leave the CFLAGS "user variable" untouched to
2621 enable easy selective overriding of the flags used by the build system.
2622
2623 Using $lt_prog_compiler_pic shouldn't be necessary, let's drop it and
2624 see if anything breaks. WERROR_CFLAGS was undefined, so dropped as well.
2625
2626 build: we preserve CPPFLAGS, just remove the tempting hooks
2627
2628 build: sanitize LDFLAGS handling
2629 The Automake manual states that according to the GNU Coding Standards
2630 the so-called "Variables reserved for the user" mustn't be changed by
2631 the build system, and they must override the default settings from the
2632 build system. The "Flag Variables Ordering" section provides the
2633 recipes on which we build here.
2634
2635 Using $lt_prog_compiler_pic directly shouldn't be necessary, let's try
2636 leaving it out.
2637
2638 build: LIBS belong to LDADD for binaries
2639
2640 build: remove unused DEBUG Automake conditional
2641
2642 Propagate NSS initialization errors
2643
2644 EVP_CIPHER_block_size returns size_t, which is unsigned
2645
2646 Propagate OpenSSL initialization errors
2647
2648 Add some missing error propagation
2649
2650 2018-01-29 Fabio M. Di Nitto <[email protected]>
2651
2652 [man] do not ship precompiled doxyxml
2653
2654 [man] add build requires on doxygen
2655
2656 [man] fix man page generation external deps on doxyxml
2657
2658 [man] fix distcheck, build out of tree and tarball generation
2659
2660 [man] update .gitignore
2661
2662 [man] switch to autogenerated man pages
2663
2664 2018-01-29 Christine Caulfield <[email protected]>
2665
2666 doxyxml: Fix parameter ordering
2667 Also tidy pointer printing and remove extraneous quotes from
2668 the BITS IN CAPS (not sure why I did that, but it helps the diff
2669 from the old pages)
2670
2671 2018-01-29 Fabio M. Di Nitto <[email protected]>
2672
2673 [man] don't output unnecessary .PP sections
2674
2675 [man] fix comma separated list in See Also section
2676
2677 [man] reduce diff with the old man pages
2678 O...C...D...
2679
2680 [man] first attempt to plug doxyxml to replace doxy2man
2681
2682 [man] fix args parsing and help text
2683
2684 Update .gitignore
2685
2686 [man] init variable and silence gcc
2687
2688 [man] fix casting around
2689
2690 [man] define _DEFAULT_SOURCE for new features.h
2691
2692 [man] add libqb and libxml2 devel as general buildrequires for man pages
2693
2694 2018-01-29 Christine Caulfield <[email protected]>
2695
2696 doxyxml: Fix compiler warnings
2697
2698 2018-01-29 Fabio M. Di Nitto <[email protected]>
2699
2700 [man] add makefile.am
2701
2702 [man] move doxyman to build-aux
2703
2704 [man] drop redundant update-man-page code
2705
2706 2018-01-29 Christine Caulfield <[email protected]>
2707
2708 manpages: Generate man pages using our own command
2709 This removes the dependncy on doxy2man which is not
2710 packaged.
2711
2712 2018-01-19 Fabio M. Di Nitto <[email protected]>
2713
2714 Use LD_LIBRARY_PATH to find the build tree artifacts and disable RPATH across the project
2715 The libtool wrapper scripts of the test binaries set LD_LIBRARY_PATH to
2716 find the libknet library and its modules in the build tree, but RPATH
2717 overrides this setting. This is why RPATH is deprecated, so we switch
2718 to RUNPATH instead by using --enable-new-dtags, which is already the
2719 linker default on Debian for example.
2720
2721 Based on the original patch from Ferenc.
2722
2723 2018-01-18 Ferenc Wágner <[email protected]>
2724
2725 tests: check for exported symbols in the data section as well
2726 The PowerPC64 ELFv1 ABI puts the function descriptor symbols there,
2727 but why restrict the consistency check to the text section anyway?
2728
2729 2018-01-17 Bin Liu <[email protected]>
2730
2731 typo fix: change the url in spec file
2732
2733 2018-01-08 Fabio M. Di Nitto <[email protected]>
2734
2735 Merge pull request #112 from kronosnet/man-pages
2736 [docs] Update copyright and man pages pre-1.0
2737
2738 [docs] Update copyright and man pages pre-1.0
2739
2740 Merge pull request #111 from kronosnet/fixes
2741 Fixes
2742
2743 [PMTUd] drop (now) unnecessary and dangerous usleep
2744 prior to all threads being able to notify PMTUd of EMSGSIZE errors,
2745 we had this random usleep in there to have time to collect data.
2746 It was working at the time, but it's a bad idea.
2747
2748 On super large clusters (>66 nodes) with 4 links on each node, when
2749 applying heavy load (cpghum on all nodes at once), the average latency
2750 between nodes can increase so much that the PMTUd thread usleep
2751 could literally block corosync for seconds at a time.
2752
2753 Drop the usleep and live happily ever after
2754
2755 [PMTUd] reset value or suffer the pains of DEADLOCKING! errno is per thread, not per function
2756
2757 [PMTUd] only notify PMTUd when pmtud_running and don't force a run to avoid lock fighting
2758 When we request a writelock, we should notify PMTUd to abort and cond_signal only
2759 if PMTUd is running.
2760
2761 Also, don't request an immediate rerun as we cannot guarantee that the scheduling
2762 is favourable to our request and we could end up waiting forever.
2763
2764 [stats] show loopback links as connected
2765
2766 2018-01-05 Fabio M. Di Nitto <[email protected]>
2767
2768 Merge pull request #105 from kronosnet/scale
2769 Scale
2770
2771 2018-01-04 Fabio M. Di Nitto <[email protected]>
2772
2773 [PMTUd] more typo fixes
2774
2775 meeeehhhhhh fixed on the top comment, forgot the bottom one
2776
2777 [PMTUd] cleanup variable usage and code around it
2778
2779 2018-01-04 Christine Caulfield <[email protected]>
2780
2781 fix missing space
2782
2783 2018-01-03 Fabio M. Di Nitto <[email protected]>
2784
2785 [PMTUd] amend warning message
2786
2787 [PMTUd] generate warning when there is a timeout receiving a response
2788
2789 [PMTUd] if PMTUd is rescheduled, then enforce a new run
2790 the usleep between each PMTUd run will give time to the other threads to grab the wrlock
2791
2792 [PMTUd] make PMTUd force_run notification more quiet
2793
2794 [PTMUd] if any threads receives a EMSGSIZE outside of a PMTUD run, force PTMUd to run
2795 Scenario:
2796
2797 node X has MTU Y on the interface and application is sending packets with size >= Y.
2798
2799 The interface MTU is suddenly reduced to < Y
2800
2801 Before this change, the kernel would be dropping packets till the next PMTUd run.
2802
2803 After this change, the PMTUd will be informed that it has to rerun (overriding
2804 the pmtud_interval), reducing the packet drop to a minimum.
2805
2806 How to test:
2807
2808 force knet_bench to send 1500 size packets with ping_data (requires code change)
2809 and start it.
2810
2811 reduce MTU on the interface from 1500 to 1300 (for example)
2812
2813 Notice an immediate trigger of PMTUd run in debug mode
2814
2815 Note: going up, from 1300 to 1500 will wait for the next PMTUd re-run as there
2816 is no immediate way to detect this change unless we start listening to kernel
2817 netlink sockets with libnl3 (investigation in progress but not critical enough atm).
2818
2819 [PMTUd] rename pmtud_running to pmtud_waiting
2820
2821 [PTMUd] by using a safe PMTU to init a link, we can shave down corosync membership creation by 25%
2822 tested on a 9 node cluster with 2 links each node, going down from 4 seconds to 3 seconds
2823
2824 [PMTUd] Use kernel MTU information to determine next packet size during discovery
2825 Using this information we can, for good links (*), determine and verify the link
2826 MTU with 2 packets.
2827
2828 * good links means:
2829 node X has MTU Y configured on a given interface. Any network object between node X
2830 and destination is capable of handling MTU >= Y.
2831 In no case the kernel will allow us to send packets > Y.
2832
2833 [stats] report correct MTU for loopback links
2834
2835 Merge pull request #108 from kronosnet/minor-fixes
2836 Minor fixes
2837
2838 [tests] add missing break (detected by newer gcc)
2839
2840 [tests] knet_bench: allow to specify protocol (UDP/SCTP) per link
2841
2842 [tests] knet_bench: add machine parsable output
2843 this is done by adding prefixes to different messages to make it easier
2844 to grep/awk and changing format of [perf] to reflect cpghum syntax.
2845
2846 [tests] knet_bench allow to configure PMTUd interval from CLI
2847
2848 [tests] fix regression in knet_bench introduced accidentaly by 2052faab4376107fe99e8a3f955bbcbfa426f648
2849
2850 [openssl] drop unnecessary reference counter
2851
2852 2018-01-03 wferi <[email protected]>
2853
2854 Merge pull request #109 from kronosnet/lzo2
2855 build: work around broken pkg-config file in lzo2 version 2.10
2856
2857 2018-01-03 Ferenc Wágner <[email protected]>
2858
2859 build: work around broken pkg-config file in lzo2 version 2.10
2860 In principle this isn't specific to BSD, though 2.10 is only packaged
2861 for BSDs at the moment.
2862
2863 2018-01-02 Fabio M. Di Nitto <[email protected]>
2864
2865 Merge pull request #106 from kronosnet/openssl-fixes
2866 [openssl] fix openssl1.0 crash by implementing documented locking cal…
2867
2868 [openssl] fix openssl1.0 crash by implementing documented locking callbacks
2869 openssl < 1.1.0 do not support natively multithread applications and need
2870 some extra help from the application itself to provide locking callbacks.
2871
2872 https://www.openssl.org/docs/man1.0.2/crypto/threads.html
2873 https://www.openssl.org/blog/blog/2017/02/21/threads/
2874
2875 to test this fix is necessary to use knet_bench or corosync with openssl
2876 and lots of heavy load (perf benchmark) workload. Sooner or later
2877 the application will crash with some random tracebacks.
2878
2879 after this patch, the crash cannot be reproduced anymore.
2880
2881 tested using 9 nodes x2 active/active links all running corosync + cpghum
2882
2883 2017-12-19 Fabio M. Di Nitto <[email protected]>
2884
2885 Merge pull request #99 from kronosnet/skip
2886 Skip tests if socket buffers are inadequate
2887
2888 2017-12-19 Ferenc Wágner <[email protected]>
2889
2890 Remove spaces after function names
2891
2892 Anything that stops operation and returns an error should be log_err
2893
2894 Document knet-specific errno values
2895
2896 More fine-grained error handling
2897
2898 tests: skip if socket buffers are too small
2899
2900 tests: replace open coded error handling with knet_handle_start() calls
2901 Done by:
2902 $ sed -i -e '1h;2,$H;$!d;g' -e 's/knet_h = knet_handle_new(1, logfds\[1\], \(KNET_LOG_[A-Z]*\));\n\n\tif (!knet_h) {\n\t\tprintf("knet_handle_new failed: %s\\n", strerror(errno));\n\t\tflush_logs(logfds\[0\], stdout);\n\t\tclose_logpipes(logfds);\n\t\texit(FAIL);\n\t}/knet_h = knet_handle_start(logfds, \1);/g' libknet/tests/*.c
2903
2904 tests: introduce knet_handle_start helper
2905
2906 Report ENAMETOOLONG if socket buffers are too small
2907
2908 2017-12-19 Fabio M. Di Nitto <[email protected]>
2909
2910 Merge pull request #103 from kronosnet/mtu-lock
2911 [PMTUd] fix external API and PMTUd interaction
2912
2913 2017-12-18 Fabio M. Di Nitto <[email protected]>
2914
2915 [PMTUd] fix external API and PMTUd interaction
2916 The problem:
2917
2918 PMTUd can take a long time to release the global read lock, mostly due
2919 to the pthread_cond_timedwait required to ack/nack packets from the
2920 other hosts. This delay could block any wrlock operation for several seconds
2921 if not more.
2922
2923 The solution:
2924
2925 each call to the global pthread_rwlock_wrlock has been changed to a wrapper
2926 that will notify the PMTUd to interrupt its operations (and restart) first,
2927 then get a global write lock that is queued as soon as PMTUd is going out.
2928
2929 This solution also improves a lot shutdown speed.
2930
2931 How to test:
2932
2933 This is not super simple to test and verify. I used 2 VMs with known MTU of
2934 1500. Start knet_bench on both (normal ping_data -C is more than enough).
2935 Once they have established data exchange, change the MTU on one of the nodes
2936 to 1600 (or higher). This should guarantee that the PMTUd process will take
2937 a very long time to complete.
2938 First verify that the PMTUd process takes several seconds.
2939 Once the next PMTUd run starts, hit ctrl+c on the node that is executing
2940 the PMTUd and the process should exit much faster than before this patch.
2941
2942 Merge pull request #104 from kronosnet/spelling
2943 Spelling fixes
2944
2945 2017-12-18 Ferenc Wágner <[email protected]>
2946
2947 Fix typo and capitalization
2948
2949 Fix grammar (allow requires an object)
2950
2951 2017-12-15 Fabio M. Di Nitto <[email protected]>
2952
2953 Merge pull request #102 from kronosnet/clang
2954 Clang
2955
2956 [build] define conditionals only once
2957
2958 [build] workaround BSD / clang breakage
2959
2960 [build] fix clang builds
2961
2962 Merge pull request #101 from kronosnet/mtu2
2963 [PMTUd] fix multiple issues and stability problems
2964
2965 2017-12-14 Fabio M. Di Nitto <[email protected]>
2966
2967 [PMTUd] fix multiple issues and stability problems
2968 - resolve locking issue with thread_heartbeat that was causing
2969 spurious up/down link event.
2970 In the event of a PMTUd run taking too long, the heartbeat
2971 thread could hang for much longer than ping_timeout.
2972 Use backoff_mutex to sync between threads instead of the global lock.
2973
2974 - pause the DATA tx thread when sending any PMTUd related packets.
2975 Similar method as knet_send_sync, using the tx_mutex, allows a much
2976 more stable communication between nodes without any visible performance
2977 hit.
2978
2979 - calculate higher timeouts when using crypto to improve stability
2980
2981 - fix an odd race condition with the kernel where, during a single PMTUd run,
2982 the same packet size was marked both BAD and GOOD (via EMSGSIZE) by the
2983 kernel. That situation would cause our PMTUd to run away and calculate
2984 bad values.
2985
2986 - add a minor usleep between sending PMTUd packets to give time to the
2987 kernel to make its own mind about the link PMTU. This is based on average
2988 latency.
2989
2990 - since PMTUd can take several seconds before completion, use the "end time"
2991 to record the last run vs the start time.
2992
2993 - fix a major issue in sending PMTUd reply where an errno was not being
2994 passed down the link layer and would cause the RX thread to block forever.
2995
2996 2017-12-09 Fabio M. Di Nitto <[email protected]>
2997
2998 Merge pull request #98 from liu4480/add-debuginfo
2999 allow to choose whether to build debuginfo packages or not
3000
3001 2017-12-08 Bin Liu <[email protected]>
3002
3003 allow to choose whether to build debuginfo packages or not
3004 1. If not "--enable-rpm-debuginfo" or "--disable-rpm-debuginfo" is
3005 specified, follow the system default behavior
3006 2. If set, then build debuginfo package or not build based on the
3007 flag.
3008
3009 2017-12-05 Fabio M. Di Nitto <[email protected]>
3010
3011 Merge pull request #96 from kronosnet/modules2
3012 Modules2
3013
3014 2017-12-04 Ferenc Wágner <[email protected]>
3015
3016 build: remove useless assignment
3017
3018 build: use AS_HELP_STRING for consistent formatting
3019
3020 build: recover original option ordering
3021
3022 build: move more repetitive module work into our Autoconf macro
3023
3024 Check ABI version on module load
3025
3026 2017-12-04 Fabio M. Di Nitto <[email protected]>
3027
3028 [tests] drop obsoleted tests that are broken after moving to the new modules format
3029
3030 [tests] make module tests dynamic by using internal knowledge of what is built or not
3031 drop all related #ifdef as well
3032
3033 2017-11-30 Fabio M. Di Nitto <[email protected]>
3034
3035 [modules] update valgrind BSD exceptions
3036
3037 [modules] spec file: fix directory shipping
3038
3039 [modules] update spec file to ship modules in separate rpms + meta packages
3040
3041 2017-11-30 Ferenc Wágner <[email protected]>
3042
3043 Unify compress and crypto module loaders
3044
3045 Unify crypto module initialization functions
3046 Taken from Fabio M. Di Nitto.
3047
3048 libknet.h uses ssize_t, include its definition directly
3049 This fixes the BSD build.
3050
3051 Pass log_msg to the plugins
3052
3053 Switch over all plugins to the module system
3054
3055 2017-11-29 Ferenc Wágner <[email protected]>
3056
3057 Make the bzip2 compress plugin a proper module
3058 Our current practice of dlopening foreign shared libraries is problematic
3059 for several reasons:
3060 * not portable: modules and shared libraries can be different object types
3061 * dependency information is invisible (our canaries mostly solve this)
3062 * hardwiring SONAMES breaks on transitions (KNET_PKG_SONAME solves this)
3063 * symbol versioning information is lost (theoretically solvable)
3064
3065 The preferred way out is generating dynamically loaded private modules
3066 from the main source, which then rely on the dynamic linker to load the
3067 external symbols as usual.
3068
3069 Separate compress_model_t definition into a new header file
3070 We'll use this type together with the zlib headers, which clash with
3071 our compress.h because both define a function named compress.
3072
3073 2017-11-28 Fabio M. Di Nitto <[email protected]>
3074
3075 Merge pull request #94 from kronosnet/sock
3076 [transport] make the new derooted sock management less chatty
3077
3078 [transports] add logging around setsockopts that have specific per distro/include requirements
3079 this is helpful to determine if a given option has been built-in and/or activated
3080 properly
3081
3082 [transport] add missing include files and really fix 86c81a32cfccde1dea
3083
3084 [transport] consistency is not optional :-)
3085
3086 [transport] fix errno handling
3087
3088 [transport] make the new derooted sock management less chatty
3089
3090 2017-11-27 Fabio M. Di Nitto <[email protected]>
3091
3092 Merge pull request #91 from kronosnet/no-root
3093 Enable testing without root privileges
3094
3095 2017-11-27 Ferenc Wágner <[email protected]>
3096
3097 Detailed error reporting about socket buffer sizes
3098
3099 2017-11-26 Ferenc Wágner <[email protected]>
3100
3101 Remove the need_root() function
3102 The point is not having root privileges, but having high enough socket
3103 buffer caps. Very few of the test can run without it.
3104
3105 2017-11-25 Ferenc Wágner <[email protected]>
3106
3107 Don't require root privileges unless necessary
3108 On Linux, if /proc/sys/net/core/[rw]mem_max are set to at least
3109 8388608 (KNET_RING_RCVBUFF), setting the socket buffer sizes
3110 doesn't require root privileges.
3111
3112 FreeBSD uses the kern.ipc.maxsockbuf sysctl MIB variable for
3113 capping user buffer requests.
3114
3115 Linux doubles the requested amount for administrative overhead,
3116 but FreeBSD does not, so we can't be too strict when checking
3117 the results.
3118
3119 2017-11-25 Fabio M. Di Nitto <[email protected]>
3120
3121 Merge pull request #93 from kronosnet/dynamic-pong
3122 Dynamic pong
3123
3124 [tests] workaround buggy valgrind-freebsd implementation.
3125 valgrind is not officially supported on FreeBSD. The current valgrind-freebsd
3126 port is maintained separately and it's lagging several releases behind
3127 vs upstream valgrind.
3128
3129 When running knet+openssl+valgrind on x86_64, with high pthread_cond_timedwait
3130 configuration, valgrind appears to be stuck tracing openssl internal operations
3131 to the point that the internal knet RX thread is not scheduled for minutes.
3132 At that point all processes will go kaboom (heartbeat, PMTUd, etc) almost constantly
3133 failing this test.
3134
3135 Given that all other architectures and OSes can run this test happily, I am adding
3136 this exception to the test suite, to be re-evaluated in future if newer versions
3137 of valgrind will be available on FreeBSD.
3138
3139 2017-11-24 Fabio M. Di Nitto <[email protected]>
3140
3141 [tests] don't duplicate test
3142
3143 [tests] use a new tunable functions to wait for packets
3144
3145 [tests] when running under valgrind, allow a lot more time for host to be reachable
3146
3147 [links] fix description
3148
3149 [links] add pong_timeout backoff mechanism to deal with high latency links
3150
3151 Merge pull request #92 from kronosnet/timer-fixes
3152 Timer fixes
3153
3154 [links] extend knet_link_set_ping_timers ABI to be ready for future expansion
3155
3156 [tests] implement common and more flexible wait_for_host implementation
3157
3158 [docs] update header to be more specific about ping timers and relative man pages
3159
3160 [PTMUd] don't harcode 2 seconds timeout to receive a PMTUd response from the remote node
3161
3162 2017-11-23 Fabio M. Di Nitto <[email protected]>
3163
3164 Merge pull request #87 from chrissie-c/balance_crypt_stats
3165 stats: Make the crypt_packet stats balance
3166
3167 2017-11-23 Christine Caulfield <[email protected]>
3168
3169 stats: Only collect crypt over stats for received data packets
3170 This is more for consistency really. The TX code only accumulates the
3171 crypto overhead for data packets because it's not worth adding all the
3172 accounting code around the ping/pong & pmtu transmitters.
3173
3174 stats: Make the crypt_packet stats balance
3175 The tx_crypt stats only included data packets whereas the rx_crypt
3176 stats included everything, because the TX was done in different places
3177 and the RX in just the one.
3178
3179 I've added a 'stats_extra' structure so that the other threads can
3180 update their own stats without extra locking, the get_stats call adds
3181 them in as necessary.
3182
3183 2017-11-21 Fabio M. Di Nitto <[email protected]>
3184
3185 Merge pull request #90 from kronosnet/soname
3186 build: determine the plugin SONAMEs automatically
3187
3188 2017-11-20 Ferenc Wágner <[email protected]>
3189
3190 build: determine the plugin SONAMEs automatically
3191 Most importantly, this avoids dlopening libcrypto.so, which is a symlink
3192 in the OpenSSL development packages on Linux. Rather, we use the SONAME
3193 of the first library added by pkg-config, which seems to work well across
3194 the board. The strange case is NSS, which ends up using libssl3.so on
3195 CentOS, Fedora and RedHat, but libnss3.so on Debian, FreeBSD and Ubuntu.
3196 The tests pass regardless, so this might be tolerable.
3197
3198 2017-11-17 Fabio M. Di Nitto <[email protected]>
3199
3200 Merge pull request #89 from chrissie-c/knet_bench_stats
3201 [knet_bench] Add option to display stats
3202
3203 2017-11-17 Christine Caulfield <[email protected]>
3204
3205 [knet_bench] Add option to display stats
3206
3207 2017-11-15 Fabio M. Di Nitto <[email protected]>
3208
3209 Merge pull request #88 from kronosnet/nss-workaround
3210 [build] workaround memory alignment bug in nss in combination with va…
3211
3212 [build] workaround memory alignment bug in nss in combination with valgrind on Linux/i386
3213 Over the past few weeks, we noticed CI failing on i386 when running make check-memcheck
3214 (test suite executed with valgring). Those failures were not consistent and sometime random.
3215
3216 After some heavy debugging it turns out that those failures are a combination of libnss3
3217 internal memory allocator bug and valgrind internal memory allocator.
3218
3219 Current libnss3 memory allocator expects a 16 bytes memory aligned
3220 buffer and in the event the memory is not aligned, it would round it.
3221 The issue is that, in performing this rounding, nss does not track properly the
3222 original address of the memory, and the equivalent of subsequent free of this buffer
3223 would fail, leaking memory and potentially accessing memory not allocated or corrpting
3224 memory.
3225
3226 The nss bug is already fixed upstream by commit: changeset: 13557:52e38f913220
3227
3228 Valgrind on Linux/i386 memory allocator can return memory that is not aligned to the 16 bytes,
3229 contrary to what malloc/glibc does. Valgrind was returning (somehow randomly) a non aligned
3230 buffer to nss triggering the nss bug and the leak.
3231
3232 By forcing valgrind to use aligned memory, nss bug does not trigger and there is no memory leak
3233 during the execution of the test suite.
3234
3235 The memory leak never triggered when running knet in normal conditions.
3236
3237 2017-11-14 Fabio M. Di Nitto <[email protected]>
3238
3239 Merge pull request #86 from kronosnet/pthread
3240 build: adopt ax_pthread.m4 to provide all necessary flags
3241
3242 2017-11-14 Ferenc Wágner <[email protected]>
3243
3244 build: adopt ax_pthread.m4 to provide all necessary flags
3245 Previously we only linked with -lpthread, but for example the -pthread
3246 GCC option provides a CPP define as well.
3247
3248 2017-11-14 Fabio M. Di Nitto <[email protected]>
3249
3250 Merge pull request #85 from wferi/libcrypto
3251 We only use the crypto part of OpenSSL
3252
3253 2017-11-14 Ferenc Wágner <[email protected]>
3254
3255 Revert "build: link as needed"
3256 This reverts commit e3c019fde32f78bf2e3a9bcf6c2fff93245708a0.
3257
3258 Once LIBS is kept empty and OpenSSL isn't linked to in full, overlinking
3259 shouldn't be a problem anymore.
3260
3261 build: only kronosnetd needs PAM, so keep the linker options separate
3262
3263 build: remove unused function checks
3264 These have no corresponding HAVE_X checks in the code.
3265
3266 build: the check_lib_no_libs function was unused
3267 The optional arguments of AC_CHECK_LIB work good enough.
3268
3269 build: don't add the dl library to LIBS, keep it in dl_LIBS only
3270 This leaves us with an empty LIBS in the most common cases.
3271
3272 2017-11-13 Ferenc Wágner <[email protected]>
3273
3274 build: clock_gettime does not require the rt library for glibc 2.17+
3275 So stop overlinking with -lrt on modern systems.
3276
3277 build: we only use the crypto part of OpenSSL
3278
3279 2017-11-13 Fabio M. Di Nitto <[email protected]>
3280
3281 Merge pull request #84 from kronosnet/api-fixes
3282 Api fixes
3283
3284 [libknet] fix comments and update man pages
3285
3286 [libknet] add padding to get *_info structs
3287
3288 [tests] fix tests differently
3289
3290 [tests] fix build on i386
3291
3292 [tests] add api_knet_send_crypto
3293
3294 [transports] knet_get_transport_list should behaves as knet_get_compress/crypto_list
3295
3296 [tests] fix error message in api_knet_get_compress_list.c
3297
3298 [docs] update copyright and manpagea
3299
3300 [compress] move knet_handle_get_compress_list to knet_get_compress_list
3301 now matches the same API as get_transport_list
3302
3303 [crypto] move knet_handle_get_crypto_list to knet_get_crypto_list
3304 now matches the same API as get_transport_list
3305
3306 [logging] don't hardcode the max, use max value for the typedef
3307
3308 [transports] make KNET_MAX_TRANSPORTS = UINT8_MAX now that the APIs allow holes in the IDs
3309 this way we don't need to break ABI if we add new transports later on.
3310
3311 [transports] cleanup API calls
3312 knet_get_transport_id_by_name, knet_get_transport_list and knet_get_transport_name_by_id don't need
3313 a handle to be functional as they only access build-time info that don't change at runtime.
3314
3315 [transports] move to use the same internal API as crypto and compress modules
3316
3317 Merge pull request #83 from wferi/canary
3318 Automatic canaries
3319
3320 2017-11-13 Ferenc Wágner <[email protected]>
3321
3322 Extend the preprocessor schema to the NSS crypto plugin
3323
3324 Extend the preprocessor schema to the OpenSSL crypto plugin
3325
3326 2017-11-12 Ferenc Wágner <[email protected]>
3327
3328 Validate dynamic symbol prototypes
3329 Practically, all imported symbols are functions returning an int, but
3330 the LZMA API defines it as an enum.
3331
3332 Extend the preprocessor schema to all compression plugins
3333
3334 Use the preprocessor for the zlib symbol remapping as well
3335 This ensures that the canary uses all dynamically loaded symbols,
3336 enabling precise versioning of the optional dependencies.
3337
3338 Use the preprocessor to reference all remapped zlib symbols in the canary
3339
3340 2017-11-11 Ferenc Wágner <[email protected]>
3341
3342 Switch libknet/compress_zlib.c to using the remap_symbol helper
3343
3344 Define a common helper function for symbol remapping
3345
3346 git ignore libknet/{compress,crypto}_canary
3347
3348 2017-11-10 Fabio M. Di Nitto <[email protected]>
3349
3350 Merge pull request #80 from kronosnet/clang-fixes
3351 Clang fixes
3352
3353 [clang] fix build on freebsd
3354
3355 [tests] add exception for lzma on Ubuntu 17.10 i386 (clang) - take 2
3356
3357 [tests] add exception for lzma on Ubuntu 17.10 i386 (clang)
3358
3359 Merge pull request #81 from wferi/canary
3360 Generate dependency information for plugin libraries
3361
3362 2017-11-10 Ferenc Wágner <[email protected]>
3363
3364 Generate dependency information for plugin libraries
3365
3366 2017-11-09 Fabio M. Di Nitto <[email protected]>
3367
3368 Merge pull request #79 from wferi/git-version-gen
3369 build: update git-version-gen
3370
3371 2017-11-09 Ferenc Wágner <[email protected]>
3372
3373 build: update git-version-gen
3374 This newer version copied from libqb does not speak -dirty if the
3375 tarball is extracted into a GIT repository, which happens during
3376 packaging.
3377
3378 https://lists.gnu.org/archive/html/coreutils/2011-02/msg00069.html
3379
3380 2017-11-07 Fabio M. Di Nitto <[email protected]>
3381
3382 Merge pull request #78 from kronosnet/docs
3383 Docs
3384
3385 [docs] update man pages with proper copyright info
3386
3387 2017-11-06 Fabio M. Di Nitto <[email protected]>
3388
3389 Add copyright to the README file
3390
3391 [docs] update README file
3392
3393 [docs] update man pages with copyright info
3394
3395 [docs] drop TODO list
3396
3397 [docs] remove empty publican docbooks
3398
3399 2017-11-06 Christine Caulfield <[email protected]>
3400
3401 doc: Add copyright section to man pages
3402
3403 2017-11-06 Fabio M. Di Nitto <[email protected]>
3404
3405 [docs] add copyright to more files
3406
3407 Cleanup export sym naming
3408 14:42 < chrissie> Here's the quote in the manual if you need it https://ftp.gnu.org/old-gnu/Manuals/ld-2.9.1/html_mono/ld.html#SEC25
3409 14:42 < fabbione> lol, i trust you :-)
3410 14:51 < chrissie> most I wondered if you wnated a reference for the commit
3411
3412 [docs] more copyright fixing
3413
3414 Merge pull request #76 from kronosnet/docs
3415 [docs] Update copyright info and add tools to automate it
3416
3417 [docs] fix git startdate to follow file renames and set enddate to this year
3418
3419 [docs] Update copyright info and add tools to automate it
3420
3421 Merge pull request #75 from kronosnet/docs
3422 [docs] update COPYRIGHT
3423
3424 [docs] update COPYRIGHT
3425
3426 2017-11-02 Fabio M. Di Nitto <[email protected]>
3427
3428 Merge pull request #71 from chrissie-c/parallelise_tests
3429 tests: Parallelize tests
3430
3431 2017-11-02 Christine Caulfield <[email protected]>
3432
3433 tests: Fix [s]printf of uint32
3434
3435 tests: Parallelize tests
3436 Randomize (using PID) the port numbers used for the tests so they
3437 don't clash.
3438
3439 2017-11-02 Fabio M. Di Nitto <[email protected]>
3440
3441 Merge pull request #74 from chrissie-c/mark-link-configured-later
3442 link: Only mark a link as 'configured' when it actually is
3443
3444 2017-11-02 Christine Caulfield <[email protected]>
3445
3446 link: Only mark a link as 'configured' when it actually is
3447 If the transport-specific config fails then the link can be left
3448 unconfigured, but marked as 'configured'.
3449
3450 I've only moved the setting of 'configured' flag itself rather than
3451 the rest of the values in case the transport-specific code wants
3452 to change them for any reason.
3453
3454 2017-11-01 Fabio M. Di Nitto <[email protected]>
3455
3456 Merge pull request #73 from kronosnet/build-fixes
3457 Build fixes
3458
3459 [build] fix make distcheck on bsd machines
3460 make sure to clean temporary files created by sed on the way of building
3461
3462 [build] fix git dirty detection
3463
3464 [build] fix changelog generation
3465
3466 [build] fix make distcheck target
3467
3468 [build] don't ship prebuilt spec file with release tarball
3469
3470 [build] disable kronosnetd build with distcheck
3471
3472 Merge pull request #72 from kronosnet/build-fixes
3473 [rpmbuild] fix BuildRequires to allow rpmbuild for both Fedora and Op…
3474
3475 [rpmbuild] fix BuildRequires to allow rpmbuild for both Fedora and OpenSUSE based distros
3476
3477 2017-10-31 Fabio M. Di Nitto <[email protected]>
3478
3479 Merge pull request #70 from wferi/dist
3480 Add the notes and the docbook to the distribution
3481
3482 2017-10-31 Ferenc Wágner <[email protected]>
3483
3484 Add the notes and the docbook to the distribution
3485
3486 2017-10-31 Fabio M. Di Nitto <[email protected]>
3487
3488 Merge pull request #69 from kronosnet/build-fixes
3489 [bsd] fix builds and testsuite on BSD x86_64/i386
3490
3491 [bsd] fix builds and testsuite on BSD x86_64/i386
3492 I don't particularly enjoy the use of host detection in configure.ac,
3493 but there is no easier way to set the library names for dlopen without
3494 some complex and fragile scripts.
3495
3496 BSD valgrind + nss complains about some non-initialized memory access.
3497 being internal to nss we can ignore it.
3498
3499 on BSD i386 size of time_t is int32_t vs all other platforms where is int64_t.
3500 BSD doesn't appear to be offerign any PRIx conversion and since it's contained
3501 within the testsuite only (for printf purposes), cast and get away with it.
3502
3503 add check to avoid building libtap on BSD for now.
3504
3505 2017-10-30 Fabio M. Di Nitto <[email protected]>
3506
3507 Merge pull request #68 from chrissie-c/dont-send-channel-32
3508 [tx] Don't send to channel 32 if there are no channels available
3509
3510 2017-10-30 Christine Caulfield <[email protected]>
3511
3512 [tx] Don't send to channel 32 if there are no channels available
3513 For some reason this only triggers on FreeBSD (that we've seen).
3514
3515 2017-10-30 Fabio M. Di Nitto <[email protected]>
3516
3517 Merge pull request #67 from kronosnet/minor-cleanup
3518 [libknet] rename transport_info to knet_transport_info
3519
3520 [libknet] rename transport_info to knet_transport_info
3521 match the namespace of all our struct. This is an API change, but doesn't affect ABI.
3522
3523 Merge pull request #66 from kronosnet/doxyfication
3524 Doxyfication
3525
3526 [man] update man pages
3527
3528 white space cleanup
3529
3530 2017-10-30 Christine Caulfield <[email protected]>
3531
3532 doc: Fix minor oddities in the doxygen comments of libknet.h
3533 host_status_change_notify_fn() still generates a duff man page but
3534 that seems to be a bug in doxy2man
3535
3536 2017-10-30 Fabio M. Di Nitto <[email protected]>
3537
3538 add autogenerated man pages and build target to update them on demand
3539
3540 2017-10-28 Fabio M. Di Nitto <[email protected]>
3541
3542 [docs] generate libknet .3 man pages at build time
3543
3544 2017-10-27 Christine Caulfield <[email protected]>
3545
3546 doc: First attempt at Doxyfying libknet.h
3547
3548 2017-10-27 Fabio M. Di Nitto <[email protected]>
3549
3550 Merge pull request #65 from kronosnet/release-fixes
3551 Release fixes
3552
3553 [release] fix scp call for publishing files
3554
3555 [release] update .gitignore
3556
3557 [release] sign all tarballs and file list
3558
3559 Merge pull request #64 from kronosnet/libtap-fixes
3560 [libtap] reduce size of max updownpath to adjust for max action, / an…
3561
3562 2017-10-26 Fabio M. Di Nitto <[email protected]>
3563
3564 [libtap] reduce size of max updownpath to adjust for max action, / and max tapdev name size
3565
3566 Merge pull request #63 from kronosnet/nodebian
3567 [build] drop debian support from upstream
3568
3569 [build] drop debian support from upstream
3570 On Debian HA team request, we are dropping the debian/ dir
3571
3572 Merge pull request #61 from kronosnet/bsd
3573 [build] update Makefile.am to work with BSD make
3574
3575 [build] update Makefile.am to work with BSD make
3576
3577 Merge pull request #62 from chrissie-c/new-apis2
3578 libknet: Add APIs to query crypto & compression models
3579
3580 2017-10-26 Christine Caulfield <[email protected]>
3581
3582 libknet: Add APIs to query crypto & compression models
3583 Also add API call to clear the handle and link stats
3584
3585 2017-10-26 Fabio M. Di Nitto <[email protected]>
3586
3587 Merge pull request #60 from wferi/autotools
3588 Autotools
3589
3590 2017-10-26 Ferenc Wágner <[email protected]>
3591
3592 configure: initialize pkg-config with its standard macro
3593
3594 configure: remove some superfluous checks
3595 We've got no C++ code in here.
3596 No need for ranlib when using libtool.
3597
3598 2017-10-25 Fabio M. Di Nitto <[email protected]>
3599
3600 Merge pull request #59 from kronosnet/ci-fixes
3601 Ci fixes
3602
3603 [build] enable valgrind automatic supressions generation during make check-*
3604 this is useful to create exception when false errors are found on CI run
3605
3606 [tests] add exception for lzma on Ubuntu 17.10 i386
3607
3608 Merge pull request #58 from liu4480/run-autogen
3609 add "runautogen" option for compiling
3610
3611 2017-10-25 Bin Liu <[email protected]>
3612
3613 add "runautogen" option for compiling
3614
3615 2017-10-19 Fabio M. Di Nitto <[email protected]>
3616
3617 Merge pull request #56 from kronosnet/fabbione-cleanup
3618 [testing] drop travis support as we are moving all builders to https:…
3619
3620 [testing] drop travis support as we are moving all builders to https://ci.kronosnet.org/
3621
3622 Merge pull request #57 from kronosnet/fabbione-udp-fixes
3623 [udp] fix incorrect return code
3624
3625 [udp] fix incorrect return code
3626 UDP TX is tricky and racy and in theory TX never fails, as UDP is allowed
3627 to drop packets internally anyway.
3628
3629 In some race condition situations, we can endup in a situation where:
3630 - thread X attempts to send a packet
3631 - socket receives an ICMP back (for whatever reasons) generated
3632 by some other random thread/packet combo
3633 - thread X detects an error in TX and decode the error (correct)
3634
3635 pre patch:
3636 - thread X would treat that as error and drop the packet
3637 post patch:
3638 - thread X should simply resend the packet
3639
3640 simply also the ordering of parsing socket error code to make it easier to read
3641
3642 2017-10-18 Fabio M. Di Nitto <[email protected]>
3643
3644 Merge pull request #53 from digimer/master
3645 Added 'gcc' to .spec BuildRequires
3646
3647 2017-10-18 Digimer <[email protected]>
3648
3649 * Added a requirement for 'gcc' as that is not installed on fedora minimal by default, breaking rpmbuild's build attempt.
3650
3651 2017-10-16 Fabio M. Di Nitto <[email protected]>
3652
3653 [global] use memmove instead of memcpy
3654
3655 2017-10-15 Fabio M. Di Nitto <[email protected]>
3656
3657 [build] more casting fixes
3658
3659 [build] more signed/unsigned fixes
3660
3661 [build] fix some signed/unsigned declarations
3662
3663 [build] allow make check-memcheck and check-hellgrind only if valgrind is avalable
3664
3665 [build] bump minimum requirements for automake
3666 this is required for distclean + subdir-object to work properly
3667
3668 2017-09-26 Fabio M. Di Nitto <[email protected]>
3669
3670 [tests] fix build and cast aligment on Debian/sparc64 unstable (gcc7)
3671
3672 2017-09-25 Fabio M. Di Nitto <[email protected]>
3673
3674 [build] cleanup whitespace
3675
3676 [build] disable poc code by default
3677
3678 2017-09-21 Fabio M. Di Nitto <[email protected]>
3679
3680 [dlopen] drop support for unload of libraries
3681
3682 [openssl] check for errors when initializing openssl1.1
3683
3684 [build] fix build on freebsd11 (and others)
3685
3686 2017-09-19 Fabio M. Di Nitto <[email protected]>
3687
3688 [tests] fix build with no nss
3689
3690 [tests] move dump_all to callback section
3691
3692 [tests] add exception for lzma on Debian Unstable
3693
3694 [test] drop useless library count check
3695
3696 [crypto] add support for openssl1.0.2 and openssl1.1.0 to accomodate Debian unstable
3697 Please note that openssl1.1.0 needs more testing
3698
3699 2017-09-19 Christine Caulfield <[email protected]>
3700
3701 [host] Never mark loopback-connected hosts as unreachable
3702 There's no point in checking the links of a host with a LOOPBACK
3703 link - it's always up and reachable.
3704
3705 2017-09-19 Fabio M. Di Nitto <[email protected]>
3706
3707 [openssl] update libknet.h to document openssl support
3708
3709 [openssl] extend lib load/unload to cover for openssl
3710
3711 [openssl] filter "none" from *getbyname openssl functions
3712
3713 [openssl] drop unrequired OPENSSL_config
3714
3715 [openssl] revert debugging code
3716
3717 [openssl] go back to the original minimal keysize
3718
3719 [openssl] move to dlopen model
3720
3721 [nss] protect shutdown from segfaults in case we cannot resolve all syms
3722
3723 [openssl] add api_knet_handle_crypto test
3724
3725 [openssl] drop incorrect keylen management code
3726
3727 [crypto] add openssl support (first cut)
3728 still needs dlopen and memcheck love + sync with nss key size
3729
3730 [cryto] nss: add namespace to internal functions
3731
3732 2017-09-18 Fabio M. Di Nitto <[email protected]>
3733
3734 [build] fix casting with gcc-7 (debian unstable)
3735
3736 [debian] fix tests/ linking with debian unstable
3737
3738 [debian] fix build deps (pass 3 or 4)
3739
3740 [debian] be more specific about builddeps
3741
3742 [debian] fix build dependency in control file
3743
3744 2017-09-14 Fabio M. Di Nitto <[email protected]>
3745
3746 [test] fix conditional build of int_lib_load_unload
3747
3748 [global] bump lib soname to 1.0.0
3749
3750 [dlopen] compress: drop unnecessary tracker and use init array instead
3751
3752 [compress] move tracking update into a write lock context
3753
3754 [dlopen] compress: fix load/unload of libraries and add extensive test
3755 each handle should only unload libraries that has loaded or referenced
3756 on exit, attempt to close all libraries from that handle
3757 on compress_fini close only the one that's open
3758
3759 [compress] make sure to unload resources before re-init the new config
3760
3761 [global] rename link.(c|h) to links.(c|h) as they conflict with global /usr/include/link.h
3762
3763 [compress] split check_lib_init into 2 functions
3764
3765 [crypto] simplify lib init code and fix error handling
3766 also update the valgrind exception file
3767
3768 [dlopen] move all dlopen lib names at the top of the respective files
3769 makes it easier to figure out library supported version
3770
3771 [dlopen] add suppress entries for internal NSPR/NSS memory leaks (part 2)
3772
3773 [docs] fix a typo
3774
3775 [dlopen] fix crypto/compress internal name space for similar functions
3776
3777 [dlopen] add suppress entries for internal NSPR/NSS memory leaks
3778
3779 [dlopen] use RTLD_NOW instead of RTLD_LAZY to avoid perf hit on fast path
3780
3781 [crypto] partially revert and partially fix b22ef0ca65d687dbc
3782 nss/nspr still leaks approx 2kb on unload, but it's much cleaner
3783 now than the many Kb before those fixes.
3784
3785 https://developer.mozilla.org/en-US/docs/Mozilla/Projects/NSS/Memory_allocation
3786
3787 [handle] remove knet handle tracker
3788 reload of shlib should be done on a per handle base to avoid complex
3789 global functions.
3790
3791 For example, if you application is using 2 knet_h and you need to reload
3792 libbz2, then switch both knet_h to use another compression method (or none)
3793 and switch back to libbz2.
3794
3795 Similar can be done with nss/openssl once openssl support is added.
3796
3797 A global function to reload all library would a very complex error reporting
3798 mechanism and could leave some handles improperly configured.
3799
3800 The application would have a better understanding of the needs at any given
3801 time.
3802
3803 [crypto] fix nss unload/reload memory leak
3804
3805 [dlopen] if the library is a symlink, display the symlink info to the final binary
3806 it cannot prevent symlink attacks but definetely helps to identify them
3807
3808 [common] don't hardcode max path size
3809
3810 [dlclose] add logging messages around unloading libraries
3811
3812 [common] use a standard fn to dlopen libraries and print load path info
3813 might be useful to detect LD_PRELOAD attack
3814
3815 [crypto] nss: be more clear about shut down/unload situation
3816
3817 [tests] fix crypto test to deal with new nss dlopen space
3818
3819 [nss] clear error before dlopen
3820
3821 [nss] warn the users that library cannot be unloaded
3822
3823 [crypto] fix libnss horror show to avoid leaking memory
3824 libnss simply can't be unloaded at runtime as it leaks memory.
3825 switch back to the hold atexit code and don't allow unloading
3826 of the library.
3827
3828 [crypto] handle properly errors reported from .init functions
3829
3830 [crypto] fix libnss unload crash and remove that atexit horror show
3831
3832 [crypto] fix libref count the same way as compress
3833
3834 [compress] centralize handling of libref in compress abstraction layer
3835 the complexity come from a mix and match of those modules that
3836 have/don't have .is_init/.init functions, and makes handling
3837 more delicate.
3838
3839 This commit ensures that:
3840 - library is loaded only once (makes load_lib mandatory for every modules)
3841 - use knet_h->compress_int_data to track per handle init.
3842 modules providing .is_init,.init don't need set anything there
3843 unless they need to.
3844 - modules don't need to track libref anymore but they must
3845 provide a unload_lib/load_lib.
3846 - library are closed only once libref is 0.
3847
3848 [crypto] move libnss to a dlopen model (pass 1)
3849 unload and exit will coredump for now
3850
3851 [dlopen] implement dlopen rate limit to avoid RX DoS
3852
3853 [global] update notes to package maintainers
3854
3855 [compress] fix internal ref count for library unload
3856
3857 [compress] switch to new load_lib / unload_lib API
3858
3859 [compress] fix error message to be more explicit
3860
3861 [dlopen] start to cleanup the compress module API
3862
3863 [build] consistent error message
3864
3865 [build] move hardcoding of linking libraries in one place and enforce better check
3866
3867 [compress] move bzip2 to dlopen model
3868
3869 [compress] cleanup dlerror before dlopen for lzma
3870
3871 [compress] cosmetic cleanup... (part 2)
3872
3873 [compress] cosmetic cleanup
3874
3875 [compress] move lzma to dlopen model
3876
3877 [compress] move lz4 to dlopen model
3878
3879 [compress] make sure lib handles are clean after dlclose
3880
3881 [compress] move zlib to dlopen model
3882
3883 [compress] fix library init to work for multiple knet_handle
3884
3885 [compression] simplify lzo2 dlopen code handling
3886 code is still broken when using multiple knet_handle
3887
3888 [compression] move lzo2 to dlopen model and fix internal api
3889 this commit breaks all other compression modules for now
3890
3891 [test] add the test file
3892
3893 [compress] only initialize compress library on demand
3894
3895 [test] don't run api_knet_handle_new_limit by default
3896 it takes more than 20GB of ram and special ulimit settings.
3897
3898 [handle] implement open knet_handle tracker
3899 limit the number of open knet_handle to UINT8_MAX for now.
3900
3901 UINT8_MAX requires approx 20GB of RAM only for init and more
3902 than 1024 fds (tune your ulimit as necessary) without taking
3903 into account fds required for links
3904
3905 2017-08-24 Christine Caulfield <[email protected]>
3906
3907 libknet: some FreeBSD compatibility fixes.
3908 Also add support for KNET_LINK_FLAG_TRAFFICHIPRIO on FreeBSD
3909
3910 2017-08-21 Chrissie Caulfield <[email protected]>
3911
3912 add knet_handle_get_stats (#49)
3913 * stats: Add compression and crypto statistics
3914
3915 New stats structures shows how effective the compression is, and
3916 also the compression/decompression overhead time.
3917
3918 Crypto stats show the byte overhead of the crypto library as well as the time taken
3919 to sign/encrypt and decrypt the packets.
3920
3921 Signed-off-by: Christine Caulfield <[email protected]>
3922
3923 2017-08-21 Fabio M. Di Nitto <[email protected]>
3924
3925 [build] drop -lpthread from pkg-config
3926
3927 2017-08-18 Fabio M. Di Nitto <[email protected]>
3928
3929 [tests] fix knet_bench usage
3930
3931 [build] fix several hanging bits after compress/crypto optional builds
3932
3933 [build] allow optional build of nss crypto support
3934
3935 [build] rename nsscrypto to crypto_nss and match compress module names
3936
3937 [build] fix SCTP support build
3938
3939 2017-08-17 Fabio M. Di Nitto <[email protected]>
3940
3941 [compress] fix build to allow --disable-compress-all --enable-compress-$foo
3942 disable all but build $foo...
3943
3944 [compression] change test from ERROR to SKIP
3945
3946 [compress] report error from knet_handle_compress api test if there is no zlib builtin
3947
3948 [compress] update api_knet_send_compress test
3949
3950 [compress] split init and config functions
3951
3952 2017-08-17 Christine Caulfield <[email protected]>
3953
3954 [compress] fix some Itaglish & a trailing space
3955
3956 2017-08-17 Fabio M. Di Nitto <[email protected]>
3957
3958 [compress] fix test suite build
3959
3960 [compress] allow conditional builds of compress modules
3961
3962 [compress] prepare internal structs and code to allow selection of compression models
3963
3964 [compress] add pkg-config support for lzo2
3965 recent versions of lzo2 have pkg-config support. Use it when available or
3966 fall back to old detection method.
3967
3968 [compress] add bzip2 support
3969
3970 [compression] update libknet.h for lzma
3971
3972 [compress] add lzma support
3973 update also the memcheck exception file for valgrind. those updates
3974 are strictly for lzma internal calls and they work on both rhel7
3975 and fedora 25.
3976
3977 [compression] remove debugging code
3978
3979 [compress] add support for lzo2 compression
3980
3981 [testing] add knet_send with all crypto modules test
3982
3983 [compress] add support for threshold (don't compress packets smaller than X)
3984
3985 [compress zlib] fix log levels around
3986
3987 [compress lz4] fix log levels around
3988
3989 [compress] remove stray debugging entry
3990
3991 [compress] add support for lz4 and lz4hc compress mechanism
3992
3993 [compress] fix header define
3994
3995 [packaging] add explicit build dependencies
3996
3997 [compress] add userdata compression support
3998 - implement generic plugin for compression
3999 - add zlib compression support
4000
4001 2017-08-07 Christine Caulfield <[email protected]>
4002
4003 loopback: Optimise loopback sending
4004 If a packet is only sent via loopback then don't bother
4005 fragmenting or encypting it.
4006
4007 2017-08-04 Fabio M. Di Nitto <[email protected]>
4008
4009 Merge pull request #48 from chrissie-c/loopback-single
4010 loopback: Only allow 1 link to localhost if loopback is used.
4011
4012 2017-08-04 Christine Caulfield <[email protected]>
4013
4014 loopback: Fix some errno assignments that should have been to savederrno
4015
4016 2017-08-03 Christine Caulfield <[email protected]>
4017
4018 loopback: Only allow 1 link to localhost if loopback is used.
4019 If loopback is configured as a link transport to localhost then
4020 no other transport makes sense and could actually cause
4021 duplicate packets to be received. So now it's enforced and tested.
4022
4023 2017-08-03 Fabio M. Di Nitto <[email protected]>
4024
4025 [build] minor "make my OCD more quiet" cleanups
4026
4027 [tx] use loopback logging
4028
4029 [crypto] use unique name to avoid collision with other plugins
4030
4031 [logging] add loopback nice name
4032
4033 2017-08-02 Fabio M. Di Nitto <[email protected]>
4034
4035 [transport] minor nitpick
4036
4037 2017-08-01 Christine Caulfield <[email protected]>
4038
4039 libknet: Add LOOPBACK transport
4040 KNET_TRANSPORT_LOOPBACK is a transport designed for
4041 high-speed communication internal to the local node. Data
4042 packets are copied directly from the sendfd to the
4043 receiving fd. The calling program must make sure that
4044 data is removed from the receiving fd in a timely manner.
4045
4046 Reviewed-by: Fabio M. Di Nitto <[email protected]>
4047
4048 2017-07-31 Fabio M. Di Nitto <[email protected]>
4049
4050 Merge pull request #47 from jnpkrn/build+init
4051 Build+init cleanups
4052
4053 2017-07-31 Jan Pokorný <[email protected]>
4054
4055 [init] service file: drop no longer available ControlGroup= option
4056 Support for that was dropped in systemd 205[1]. To be able to get RT
4057 priority regardless of the runtime control groups (Linux-specific),
4058 there's a precedent solution for corosync[2], should it be needed.
4059
4060 [1] https://lists.freedesktop.org/archives/systemd-devel/2013-July/011679.html
4061 [2] https://github.com/corosync/corosync/pull/231
4062
4063 [build] configure: avoid confusion wrt. what's enabled by default
4064
4065 2017-07-30 Fabio M. Di Nitto <[email protected]>
4066
4067 [build] use newly create kronosnet release team gpgkey
4068
4069 [build] enable release script / publishing bits
4070
4071 [build] make maintainer-clean clean also after a release
4072
4073 [build] add Cluster Release Team signing key
4074
4075 2017-07-28 Fabio M. Di Nitto <[email protected]>
4076
4077 Merge pull request #46 from jnpkrn/build-release.mk
4078 [build] release.mk: adapt recent changes from libqb (fence-agents)
4079
4080 2017-07-28 Jan Pokorný <[email protected]>
4081
4082 [build] release.mk: reflect current release publishing practice
4083 Core author(s) had previous affiliation with project hosting on
4084 fedorahosted.org, but since the sunset of that hosting[1], it's no
4085 longer an option.
4086
4087 Unfortunately, publish-by-scp convenience of fedorahosted.org won't fly
4088 with GitHub that insists on the (primarily REST) API to get things
4089 automated, so update "publish" recipe to that effect, providing some
4090 hints on possible adaption of the process.
4091
4092 Also, turn "git push --tags" to a bit safer "git push --follow-tags"
4093 for good measure.
4094
4095 [1] https://lists.fedoraproject.org/archives/list/[email protected]/message/BWSMCGZPPNG3JOCFQ6Z74MIBU7FG3KGB/
4096
4097 [build] release.mk: simplify the default goal, declare .PHONY targets
4098 The default goal line now serves mainly the documentation purpose,
4099 otherwise it could be squashed down to "sign" target.
4100
4101 [build] release.mk: move soft guard for no GPG key up the supply chain
4102 It also means the missing key specification is now a hard error when
4103 ASCII armored files are to be created (directly or by other rules).
4104
4105 Leveraging the previous commit, it is now unnecessary to guard against
4106 release mode not being requested, so this guard is dropped completely
4107 in the transition, allowing for the full-fledged (incl. GPG signing)
4108 release delivery testing without mangling with in-tree tags.
4109
4110 To allow "sign" target to (possibly eventually) become the implicit goal
4111 even if "gpgsignkey" variable is explicitly unspecified, make it depend
4112 on $(deliverables).
4113
4114 [build] release.mk: fix no-release conflict (implied-required version)
4115 Previously, generating custom versioned tarballs without requiring
4116 a "release" (which comprises also tagging hence in turn making the
4117 implied version match the required one as desired) would not allow
4118 to test the release routines thoroughly incl. GPG signing.
4119 Unfortunately, allowing for a temporary tag is prone to accidentally
4120 slipping it to the distributed tree, so that should rather be avoided
4121 Now it can.
4122
4123 Another step to blur the boundaries between whether "release" is
4124 specified or not is to make checksums file generated unconditionally.
4125
4126 Also both pre-existing and the newly created (per the above scenario)
4127 "overhead" files arising from make/release.mk are git-ignored properly.
4128
4129 [build] release.mk: simplify/generalize GPG signing rule
4130 ...especially in case we want to easily switch to signing also
4131 particular archives, which may be a reasonable expectation from some
4132 downstream/distro maintainers (as demonstrated with the libqb's
4133 example[1] where this changeset was picked from a followup PR[2],
4134 respectively its adaptation for fence-agents[3]).
4135 Since this commit, such a change is a matter of only swapping
4136 a prerequisities specification for "sign" target per the commented
4137 out suggestion.
4138
4139 For that to work reliably even after such a change, for the complete
4140 dependency backtrack, though, a pre-existing issue with parallelized
4141 work on the prerequisites needs to be fixed first. The (currently
4142 hypothetical) problem is "tarballs" (phony) target missing in the
4143 linearized supply chain for the signed tarballs, as opposed to supply
4144 chain for the checksums file that can be worked on in an (unordered!)
4145 interleave with the work on these signed tarballs. Hence fix this
4146 (currently hypothetical) race condition by stretching "tarballs" target
4147 as progress barrier (using it as a prerequisite) uniformly for all
4148 the deliverables that are going to be signed (as opposed to the
4149 currently only "sha256" target). And when at it, make "deliverables"
4150 variable list them conveniently (along with reformatting definitions
4151 of other variables for uniformity).
4152
4153 [1] https://github.com/ClusterLabs/libqb/issues/145#issuecomment-306640694
4154 [2] https://github.com/ClusterLabs/libqb/pull/260
4155 [3] https://github.com/ClusterLabs/fence-agents/pull/123
4156
4157 2017-07-24 Christine Caulfield <[email protected]>
4158
4159 libknet: Small fixes for errors found by GCC 7.1
4160
4161 2017-07-18 Christine Caulfield <[email protected]>
4162
4163 libknet.h: change some status values from bitfield to uint8_t
4164 Bitfields are a pain and you can't take the address of them
4165 which makes it hard for cosoync stats to deal with the values
4166
4167 Reviewed-by: Fabio M. Di Nitto <[email protected]>
4168
4169 2017-07-12 Fabio M. Di Nitto <[email protected]>
4170
4171 [nss] expand error reporting with proper error strings
4172 thanks Honza for adding some of them in the previous commits
4173
4174 [crypto] set default case for key_type and avoid build warnings
4175 when key_type is unknown, gcc complains that cipher can be used
4176 uninitialized.
4177
4178 add default case to catch wrong callers and avoid build warning
4179
4180 [crypto] cleanup inconsistent whitespacing
4181
4182 Merge pull request #44 from jfriesse/fips-enable
4183 Fips enable
4184
4185 2017-07-11 Jan Friesse <[email protected]>
4186
4187 [knet.h] Reduce KNET_MIN_KEY_LEN
4188 By reducing KNET_MIN_KEY_LEN we allow FIPS enabled system using NSS to
4189 import (unwrap) the key.
4190
4191 [nsscrypto] Use different method to import key
4192 PK11_ImportSymKey doesn't work when FIPS is enabled because NSS is
4193 targeting to FIPS Level 2 where loading of unencrypted symmetric
4194 key is prohibited.
4195
4196 FIPS Level 2 is hard to achieve without breaking compatibility so patch
4197 implements "workaround" to make NSS behave like FIPS Level 1
4198 (where is allowed to load unencrypted symmetric key).
4199
4200 Workaround is about using temporal key to encrypt corosync authkey in
4201 memory and then to unwrap it into valid NSS key.
4202
4203 Also when FIPS is enabled, it's not currently (nss 3.30.2) possible to
4204 unwrap key larger than 256 bytes.
4205
4206 [nsscrypto] Refactor symmetric key importing
4207
4208 2017-06-24 Fabio M. Di Nitto <[email protected]>
4209
4210 [stats] add doc details in the stat header
4211
4212 [link] switch link_get_status to write lock
4213 previous to the introduction of detailed link stats, the values
4214 contained in the status structure would hardly change over time
4215 making a rdlock more than enough to get the right data.
4216
4217 Now that the values are updated constantly, we need to remove
4218 the window of thread X writing values and thread Y to read them
4219 while being written.
4220
4221 2017-06-20 Chrissie Caulfield <[email protected]>
4222
4223 libknet: Add link stats (#43)
4224 * libknet: Add link stats
4225
4226 This patch also includes an ABI change to knet_link_get_status()
4227 (as well as returning a load of stats) that should allow us to
4228 add more stats at a later stage without ballooning memory use
4229 with pads or crashing earlier-linked programs with structure
4230 overwrites.
4231
4232 kronosnetd compiles but has not been tested...
4233
4234 2017-06-12 Fabio M. Di Nitto <[email protected]>
4235
4236 [rx/tx] Increase RX buffers to 512 (new-tx branch)
4237 latency is still much lower than current udpu and throughput
4238 for small packets is much better
4239
4240 [kronosnetd] fix build with new link flags
4241
4242 2017-06-09 Christine Caulfield <[email protected]>
4243
4244 libknet: Add link flag for 'hi' (interactive) priority
4245 Rather than make it the default, this is now a link option.
4246 Corosync needs it but other applications probably will not.
4247
4248 Reviewed-by: Fabio M. Di Nitto <[email protected]>
4249
4250 2017-06-08 Christine Caulfield <[email protected]>
4251
4252 libknet: Add 'flags' parameter to knet_link_[gs]et_config()
4253 I have at least 1 flag planned but I'm sure there's plenty of scope
4254 for more. Also trying to get the API changes out of the way ready
4255 for the cluster summit in September.
4256
4257 2017-06-04 Fabio M. Di Nitto <[email protected]>
4258
4259 Reduce TX memory further down by some MB
4260
4261 [tx/rx] reduce memory footprint and latency
4262 reduce memory usage by approx 16MB and drop min/avg latency
4263 for some workloads
4264
4265 2017-06-03 Fabio M. Di Nitto <[email protected]>
4266
4267 Resolves signed-unsigned comparison warnings
4268 Patch based on:
4269
4270 commit 730a8b6a56820741b361e0907eecb6297a9f0348
4271 Author: Michael Jones <[email protected]>
4272
4273 Resolves: https://github.com/fabbione/kronosnet/pull/40
4274
4275 2017-06-03 Michael Jones <[email protected]>
4276
4277 Fixes format=signedness related warnings
4278
4279 Adds additional warning flags to the compiler, no new warnings unhidden
4280
4281 2017-05-26 Christine Caulfield <[email protected]>
4282
4283 [sctp] Don't hold global_rwlock while sleeping.
4284 The SCTP connect and accept threads can do small sleeps
4285 when there are errors. The global_rwlock should not be held
4286 while this is happening.
4287
4288 While the lock is a read lock and doesn't block sending and
4289 receiving in other threads it can cause a deadlock on shutdown.
4290
4291 [libknet] Unlock the global lock if logging pipe is full.
4292 If a write to the logging pipe fails (eg it's busy
4293 waiting on something else, I've seen this in corosync
4294 at shutdown) then we must unlock the global_rwlock
4295 or we'll never get out of this alive.
4296
4297 Reviewed-by: Fabio M. Di Nitto <[email protected]>
4298
4299 2017-05-11 Christine Caulfield <[email protected]>
4300
4301 [libknet] Add TC_PRIO_INTERACTIVE to sockets
4302
4303 2017-04-19 Fabio M. Di Nitto <[email protected]>
4304
4305 Merge pull request #36 from liu4480/remove-if-defined-i386
4306 use PRIu64 to replace #if defined(__i386__)
4307
4308 2017-04-19 Bin Liu <[email protected]>
4309
4310 use PRIu64 to replace #if defined(__i386__)
4311
4312 2017-04-13 Christine Caulfield <[email protected]>
4313
4314 [libknet] Add some includes to allow FreeBSD compilation
4315 Reviewed-by: Fabio M. Di Nitto <[email protected]>
4316
4317 2017-04-13 Fabio M. Di Nitto <[email protected]>
4318
4319 Merge pull request #35 from liu4480/type-fixes
4320 fix compile errors on i386/i686 for knet_bench.c
4321 Resolves: #34
4322
4323 2017-04-13 bliu <[email protected]>
4324
4325 fix compile errors on i386/i686 for knet_bench.c
4326
4327 2017-04-13 Fabio M. Di Nitto <[email protected]>
4328
4329 Merge pull request #33 from liu4480/type-fixes
4330 fix compile errors on i686
4331
4332 2017-04-13 bliu <[email protected]>
4333
4334 fix compile errors on i686
4335
4336 2017-03-17 Fabio M. Di Nitto <[email protected]>
4337
4338 [docs] update slids a tiny bit
4339
4340 2017-03-03 Fabio M. Di Nitto <[email protected]>
4341
4342 [tx] reduce memory usage and memmove on big pckts via better iovec mapping
4343
4344 2017-03-01 Fabio M. Di Nitto <[email protected]>
4345
4346 [build] fix clang errors
4347 this should fix https://github.com/fabbione/kronosnet/issues/32
4348
4349 2017-02-28 Fabio M. Di Nitto <[email protected]>
4350
4351 Merge pull request #30 from jfriesse/shadow-time
4352 [knet_bench]: Fix shadow declaration of time
4353
4354 2017-02-27 Jan Friesse <[email protected]>
4355
4356 [knet_bench]: Fix shadow declaration of time
4357
4358 2017-02-25 Fabio M. Di Nitto <[email protected]>
4359
4360 [transports] add option to configure reconnect interval
4361
4362 2017-02-24 Fabio M. Di Nitto <[email protected]>
4363
4364 [hb] use savederrno correctly
4365
4366 2017-02-23 Fabio M. Di Nitto <[email protected]>
4367
4368 [tests] knet_bench add support for perf-by-time
4369
4370 [tests] add first cut of perf-by-size implementation to knet_bench
4371 Also port to _sendmmsg/_recvmmsg and fix a couple of minor shutdown issues
4372
4373 2017-02-22 Fabio M. Di Nitto <[email protected]>
4374
4375 [tests] don't spam for no logs
4376
4377 [transport] fix sendmmsg and recvmmsg wrappers to return correct data
4378 for sendmmsg also don't overwrite the msg_len
4379
4380 2017-02-21 Fabio M. Di Nitto <[email protected]>
4381
4382 [libknet] drop bad doc copy/paste
4383
4384 Merge pull request #29 from liu4480/master
4385 add doc for knet_handle_get_channel in libknet/libknet.h
4386
4387 2017-02-21 Bin Liu <[email protected]>
4388
4389 add doc for knet_handle_get_channel in libknet/libknet.h
4390
4391 2017-02-21 Fabio M. Di Nitto <[email protected]>
4392
4393 [build] add better header protection
4394 fixes: https://github.com/fabbione/kronosnet/issues/23
4395
4396 [build] Enforce C99 compile option
4397 fixes: https://github.com/fabbione/kronosnet/pull/25
4398
4399 [sctp] fix more occurrences of link -> kn_link
4400 complete 940833bf99b18e14b08c0a5a24402026131a0884
4401
4402 fixes: https://github.com/fabbione/kronosnet/issues/28
4403
4404 [build] don't relink the whole test suite on each change
4405 fixes: https://github.com/fabbione/kronosnet/issues/27
4406
4407 2017-02-17 Fabio M. Di Nitto <[email protected]>
4408
4409 Merge pull request #21 from liu4480/master
4410 Fix: remove duplicate judgement of (have_name < 0) in kronosnetd/vty_cli_cmds.c
4411
4412 2017-02-17 Bin Liu <[email protected]>
4413
4414 Fix: remove duplicate judgement of (have_name < 0) in kronosnetd/vty_cli_cmds.c
4415 In line 1087 of kronosnetd/vty_cli_cmds.c, there is a judgement of
4416 (have_name < 0), when have_name < 0, it will return 0, which make
4417 the judge of have_name < 0 in line 1094 will never be executed.
4418 Remove the unnecessary judgement.
4419
4420 2017-02-15 Fabio M. Di Nitto <[email protected]>
4421
4422 [tests] port knet_bench to use internal version recvmmsg
4423
4424 [global] drop usage of recvmmsg inside libknet
4425 as requested by different kernel developers, we should stop
4426 using both sendmmsg and recvmmsg.
4427
4428 as temporary solution use Jan's compat wrappers. the whole TX/RX
4429 code will need review to do a full proper switch since all
4430 error codes will change and propagate differently to transport hooks
4431 and knet_send_sync users
4432
4433 [global] drop usage of sendmmsg inside libknet
4434 as requested by different kernel developers, we should stop
4435 using both sendmmsg and recvmmsg.
4436
4437 as temporary solution use Jan's compat wrappers. the whole TX/RX
4438 code will need review to do a full proper switch since all
4439 error codes will change and propagate differently to transport hooks
4440 and knet_send_sync users
4441
4442 [global] create our own copy of mmsghdr in preparation to drop sendmmsg and recvmmsg
4443 knet_bench will drop mmsghdr when dropping sendmmsg and recvmmsg
4444
4445 [poc] drop PoC sctp-defrag-bug code now that's merged into master
4446
4447 2017-02-14 Fabio M. Di Nitto <[email protected]>
4448
4449 [udp] remember to memset freshly allocated memory
4450
4451 [sctp] drop duplicated definition
4452
4453 [global] revert b525ee6a6b74ec, go back to 64K nodes, and use typedef for nodeid
4454
4455 [tx] increase timeres on TX pressure and reduce log noise
4456 when TX sockets are overloaded, we spend more time spitting out
4457 logs than recovering from the overload. ifdef the logging on
4458 critical path out (still available with debug build).
4459
4460 also drastically reduce the waiting time by 64x.
4461
4462 this changes increases UDP perf on 3 nodes by 200%
4463
4464 [tx] pass pointers around and save some memory
4465
4466 [tx] minor optimization in handling packets < MTU
4467 - reduce of one loop
4468 - reduce one memcpy
4469 - reduce setting a bunch of data
4470
4471 [tx] map any outgoing packets only once and reduce memsets around
4472
4473 [tx] map outgoing packet once per host vs once per host per link
4474 - reduces memsets by 8x
4475 - reduces complexity in resending packets on socket overload
4476 by changing start mmsghdr vs remapping everything all over again
4477
4478 2017-02-07 Fabio M. Di Nitto <[email protected]>
4479
4480 [transport] define value only once
4481
4482 2017-02-06 Fabio M. Di Nitto <[email protected]>
4483
4484 [tx] fix error log
4485
4486 2017-02-05 Fabio M. Di Nitto <[email protected]>
4487
4488 Move TODO list to trello
4489
4490 2017-02-03 Fabio M. Di Nitto <[email protected]>
4491
4492 [global] Reduce the number of support nodes in a knet from 64K to 256
4493 realistically speaking nobody is ever going to deploy more than 32/64 nodes.
4494
4495 NOTE: this commit changes both onwire and API!
4496
4497 2017-02-02 Fabio M. Di Nitto <[email protected]>
4498
4499 [send/recv] split RX and TX threads in separate files
4500
4501 [send/recv] split code into separate files (step 1)
4502
4503 [send/recv] Unify and simplify usage of seq_num in packets
4504 IMPORTANT: this commit changes onwire protocol in an incompatible way!
4505
4506 - remove the concet of bcast and mcast seq num and use one tx_seq_num
4507 - stop using LINK_UP_DOWN messages to broadcast node seq num and
4508 transfer this data inside heartbeat messages
4509 - LINK_UP_DOWN messages are currently unused but let's keep the
4510 infrastructure around for future
4511 - minor cleanup in host_set_policy to confirm change of switching
4512 policy in the logs
4513 - _link_updown should use async call to host dstcache update
4514 due to locking context
4515 - switch knet_link_set_enable to use write locking context
4516 since the only reason it was read lock, was due to the need
4517 to send LINK_UP_DOWN messages
4518 - knet_link_set_priority can now use dstcache in sync mode
4519 - add seq_num and heartbeat type (timed/untimed) data to heartbeat
4520 messages. timed messages are generated regularly by hb_thread.
4521 untimed messages are generated by the TX thread to sync seq_num
4522 on heavy load across all connected node. (see comments in the code)
4523 - access to the node seq_num is now mutex locked
4524 - abstract ability to send pings from multiple threads
4525 - special case seq_num == 0 to detect a node crash and coming back
4526 to life before hb_thread can detect the disconnection
4527 - forcefully send ping in the TX thread every SEQ_MAX / 8 packets
4528 to allow nodes to sync seq_num
4529 - optimize TX thread code to prepare the outgoing buffers once
4530 vs multiple times. There is still work that can be done here
4531 to optimize sending to multiple host, but this change
4532 is intrusive enough already as it is
4533 - add logic to clear circular buffers when receiving pings
4534
4535 [sctp] Allow internal IPC to operate under RX pressure
4536
4537 2017-01-28 Fabio M. Di Nitto <[email protected]>
4538
4539 [rpm] disable build of poc code
4540
4541 2017-01-27 Christine Caulfield <[email protected]>
4542
4543 [compat] Fix FreeBSD defines
4544 We need to undefine HAVE_RECVMMSG & HAVE_SENDMMSG for it to compile
4545 on FreeBSD (not sure how that ever worked before).
4546
4547 There is a PR for FreeBSD to fix the recvmmsg bug, but it's not in
4548 a released update yet. So with luck we can remove this soon-ish
4549
4550 https://github.com/freebsd/freebsd/commit/3fa64907439c00965fb0cbb222a6bf073c452630
4551
4552 2017-01-27 Fabio M. Di Nitto <[email protected]>
4553
4554 [host] remove completely broken host to host communication locking system
4555 the original idea was to have a host-to-host (semi-)reliable communication protocol
4556 but that just isn't possible without flow control and retransmit
4557
4558 IIRC the only side affect of this missing lock is a corner case where:
4559 1) node A totally crashes
4560 2) node A come backs to life, sends it's status info (seq_num information)
4561 3) node B does NOT receive the status info
4562 4) node A starts sending traffic and a few packets might get lost
4563
4564 this will be solved when rewriting the TX thread to optimize the seq_num handling
4565
4566 [send/recv] make sure to unlock dsthost thread under certain errors
4567
4568 [transport] make sctp tx error handling more robust and avoid spinning on dead sockets
4569
4570 Update TODO
4571
4572 [build] add build option to enable/disable poc code
4573
4574 2017-01-26 Fabio M. Di Nitto <[email protected]>
4575
4576 [transport] fix support for dynamic links connections
4577 - add internal transport API for handling incoming dynamic connections (both UDP and SCTP)
4578 - fix copy/compare address code in RX thread
4579 - make sure to reset sockaddr_storage len in iov
4580
4581 [netutils] cleanup cmpaddr and add cpyaddrport
4582 cpyaddrport specifically copies _only_ address and port and skip
4583 over other fields that are unnecessary for what we need
4584
4585 2017-01-26 Christine Caulfield <[email protected]>
4586
4587 compat: make it work on FreeBSD
4588 Although FreeBSD 11 has sendmmsg & recvmmsg wrappers, they
4589 don't quite work the same as Linux so I've enabled the (fixed)
4590 compat versions for that platform.
4591
4592 make check now works on FreeBSD 11
4593
4594 2017-01-26 Fabio M. Di Nitto <[email protected]>
4595
4596 [kronosnetd] add support for multiple transports
4597
4598 2017-01-25 Fabio M. Di Nitto <[email protected]>
4599
4600 Update TODO
4601
4602 [send/recv] simplify _parse_recv_from_links internal API
4603
4604 [transports] add support for transport_tx_sock_error
4605 add IP_RECVERR for UDP to better handle errors on TX sockets
4606
4607 [PMTUd] Fix global pmtud calculation
4608
4609 2017-01-24 Fabio M. Di Nitto <[email protected]>
4610
4611 Update TODO list
4612
4613 Merge pull request #20 from jonesmz/travis-ci-for-pull-request
4614 Adds Travis CI continuous integration configuration file
4615
4616 [transports] SCTP: handle short reads on sockets
4617 - fix trasport is_data API
4618 - handle per accepted socket reassembly buffer by changing fd_tracker
4619 data for incoming connections
4620 - allow in-kernel SCTP fragmentation again
4621 - use MSG_EOR on a per socket base to reassemble partial packet delivery
4622 - fix some whitespaces around
4623
4624 2017-01-24 Michael Jones <[email protected]>
4625
4626 Adds Travis CI continuous integration configuration file
4627
4628 2017-01-24 Christine Caulfield <[email protected]>
4629
4630 transport: Rename 'link' to 'kn_link' so it doesn't conflict with libc
4631
4632 2017-01-23 Christine Caulfield <[email protected]>
4633
4634 poc: add option to check CRCs in sctp_defrag_bug
4635
4636 2017-01-20 Fabio M. Di Nitto <[email protected]>
4637
4638 [PoC] implement EOR handling code for short reads
4639
4640 2017-01-19 Fabio M. Di Nitto <[email protected]>
4641
4642 [poc] add sctp potential fragmentation bug trigger
4643
4644 [build] drop last bits of common/
4645
4646 [transport] remove duplicate code
4647
4648 2017-01-18 Fabio M. Di Nitto <[email protected]>
4649
4650 Update TODO
4651
4652 [transport] add knet_handle_get_transport_id_by_name api
4653
4654 [transport] add knet_handle_get_transport_name_by_id api call
4655 also fix a counter in knet_handle_get_transport_name_by_id
4656
4657 [transport] add knet_handle_get_transport_list api call
4658
4659 [handle] fix error messages
4660
4661 2017-01-17 Christine Caulfield <[email protected]>
4662
4663 [compat] Fix some includes for FreeBSD
4664
4665 2017-01-17 Fabio M. Di Nitto <[email protected]>
4666
4667 [compat] minor cleanup on the autotool stuff
4668
4669 2017-01-17 Christine Caulfield <[email protected]>
4670
4671 general: Add compatibility code & #ifdefs
4672 This is enough to get knet compiling on FreeBSD 11 and bits of it
4673 working. It's nowhere near ready on BSD though, more work is needed,
4674 but given the fast pace of development it's best to get this in now
4675 rather than track it in a separate branch.
4676
4677 2017-01-17 Fabio M. Di Nitto <[email protected]>
4678
4679 Update TODO
4680
4681 Update TODO
4682
4683 2017-01-16 Fabio M. Di Nitto <[email protected]>
4684
4685 [api] add commodity functions to convert to/from strings/sockaddress
4686 functions are nothing more than wrappers for getnameinfo and getaddrinfo
4687 with some sanity checks, but exposing them around saves lots of
4688 maintanance of duplicate code across different stuff.
4689
4690 2017-01-16 Christine Caulfield <[email protected]>
4691
4692 host: Remove host_tail as it's not used.
4693 it also had a bug which could cause knet to crash after a host was
4694 removed.
4695
4696 This patch puts new nodes at the front of the list rather than at the
4697 end, but I don't think this is a problem.
4698
4699 2017-01-16 Fabio M. Di Nitto <[email protected]>
4700
4701 [test] fix comment
4702
4703 [sctp] enable workaround for in-kernel fragmentation issue
4704
4705 [test] kill obsoleted ping_test
4706
4707 Update TODO
4708
4709 [sctp] port sctp to the new API and fix many issues
4710 - use the new transport API
4711 - fix locking context around to avoid race conditions and deadlocking
4712 - fix shutdown code (segfaults and core dumps)
4713 - properly differentiate between connecting sockets and accepted sockets
4714 and better use of fd_tracker
4715 - abstract as much as possible socket management code from threads
4716 - add lots of comments and debugging messages around
4717 - simplify socket error management reported from RX thread
4718 - rework loop timers for connect_thread to avoid thread overload
4719 - reduce usage of _transport_addrtostr to one call and reuse
4720 link information around
4721 - improve error handling across the board
4722 - stop using data structs inside epolls and switch back to using fds
4723 this was necessary to avoid processing stale data from epolls
4724 and look up data from fd_tracket instead
4725 - add listener stop function
4726 - make functions safer to be called multiple times
4727 - probably more.. but can't remember
4728
4729 NOTE: this is not the most elegant code, but it seems to be doing its job
4730 fine.
4731
4732 [heartbeat] reset PMTUd on link down to trigger a new PMTUd when link come back online
4733
4734 [PMTUd] fail early if transport is not connected
4735
4736 [heartbeat] and make it a tiny little bit faster when transport is not connected
4737
4738 [heartbeat] make better use of transport_connected option
4739
4740 2017-01-15 Fabio M. Di Nitto <[email protected]>
4741
4742 [sockets] move socketpair code into transport common code
4743 also use _configure_common_socket for cloexec/nonblock and rx/tx buffers
4744 that should be the same for all socketpairs, outgoing and incoming sockets
4745
4746 [dstthread] make sure not to get caught in signal when reading from socket
4747
4748 [host] make sure not to get caught in signal when sending host updates
4749
4750 2017-01-12 Fabio M. Di Nitto <[email protected]>
4751
4752 [transports] drop fd_tracker locking in favour of global locking
4753
4754 [udp] drop unnecessary list locking
4755
4756 2017-01-11 Fabio M. Di Nitto <[email protected]>
4757
4758 [SCTP] temporary disable SCTP support while we fix the implementation
4759 not the best solution, but it's a quick one to allow master to build again
4760
4761 [send/recv] report correct errno
4762
4763 Revert "[heartbeat] temporary disable usage of transport_connected"
4764 This reverts commit 75ca10991fa3e0d9bc8cd9706395ff0e2bc62c38.
4765
4766 whops.. pushed in the wrong branch by mistake
4767
4768 [heartbeat] temporary disable usage of transport_connected
4769 this will make it easier to stabilize SCTP and connection oriented
4770 protocols
4771
4772 2017-01-09 Fabio M. Di Nitto <[email protected]>
4773
4774 Update TODO list
4775
4776 [transports] major rework of APIs and ABIs!
4777 This commit breaks APIs and ABIs and this changelog might be missing
4778 a bit or 5.
4779
4780 External visible changes:
4781
4782 - Change link initialization process API:
4783
4784 The old method:
4785 - link_set_config -> link_enable
4786 where link_set_config would simply store config data (sockaddr and
4787 such) in the link struct and link_enable would create sockets and
4788 do the magic. This method didn't work well for complex transports
4789 such as SCTP and introduced a series of race conditions and deadlocks.
4790
4791 The new method:
4792 - link_set_config would now store config in link struct AND create
4793 all related sockets and such. The link will not be used for data
4794 traffic till enabled.
4795 - link_set_enable will enable/disable the link for traffic (including
4796 heartbeat) and requires a link to be configured (as before)
4797 - link_clear_config (new API) can/has to be called after disabling
4798 a link to close all connections and sockets (free resources).
4799
4800 Internal changes:
4801
4802 - Drop the concept of listeners.* and delegate those to
4803 underlying transports.
4804
4805 - Add the concept of fd_tracker. Each transport is required and it is
4806 responsible to update the fd_tracking array for the fd that the
4807 transport itself is creating/using/clearing.
4808 The fd_tracker is required to perform fast lookups on socket
4809 errors and RX thread to determine what code is responsible to parse
4810 given conditions such as errors from the sockets or OOB data/notifications.
4811
4812 - Introduce the concept of link->transport_connected.
4813 In case the transport requires socket connection to the other end (SCTP for
4814 example), set to 0 while disconnected and set 1 once connected to the other
4815 side. This flag will avoid unnecessary errors generated from the TX threads.
4816 NOTE: still needs better plumbing around. For now it's only partially used
4817 in the heartbeat thread. UDP sets to 1 by default.
4818
4819 - Rework the transport API to be easier to use.
4820 - If a transport is not available, get_XXXX_transport() should return NULL.
4821 - A transport that provides &XXXX_transport_ops MUST have all operations
4822 implemented as described in internals.h. This is required to skip tons
4823 of if/else checks on fast code paths.
4824 - Improve documentation of the transport API in internals.h.
4825 - transport common: drop some unnecessary functions for now, they might
4826 can back later in better format once SCTP is working again.
4827 - provide a locked/unlocked version of _set_fd_tracker but this is an
4828 artifact of trying to fix SCTP deadlocking. _set_fd_tracker should
4829 always be locked.
4830 - cleanup transport_udp.c to match the new API and perform better
4831 error handling and better cleaning in case of errors.
4832 - switch mtu_overhead to constant from call into a function.
4833 - drop usage of _transport_addrtostr/_transport_addrtostr_free for all but
4834 accepting incoming SCTP connections. All other datas are already available
4835 and it's unnecessary to perform extra lookups. This will eventually
4836 move to a proper knet_api and avoid completely the need to build with --debug.
4837
4838 - RX thread:
4839 - use the new transport hooks to handle socket errors and
4840 determine if a packet is data or internal OOB info.
4841 - remove the last transport specific bits and move them into
4842 transport implementation itself.
4843
4844 - Test suite:
4845 - Update the test suite to deal with the new API changes
4846 - Fix a couple of shutdown problems in knet_bench
4847 - Fix a shutdown issue in test-common.c
4848 - Document the whole recvmmsg API in details, based on kernel
4849 implementation. recvmmsg man page is incorrect.
4850
4851 - Fix kronosnetd to use the new APIs
4852
4853 2016-12-21 Fabio M. Di Nitto <[email protected]>
4854
4855 [send/recv] remove stray unlock operation
4856
4857 2016-12-20 Fabio M. Di Nitto <[email protected]>
4858
4859 [logging] rename KNET_SUB_SCTP_LINK_T to KNET_SUB_TRANSP_SCTP
4860
4861 [logging] rename KNET_SUB_UDP_LINK_T to KNET_SUB_TRANSP_UDP
4862
4863 [logging] rename KNET_SUB_LINK_T to KNET_SUB_RX
4864
4865 [logging] rename KNET_SUB_SEND_T to KNET_SUB_TX
4866
4867 [logging] rename KNET_SUB_PMTUD_T to KNET_SUB_PMTUD
4868
4869 [logging] rename KNET_SUB_HB_T to KNET_SUB_HEARTBEAT
4870
4871 [logging] rename KNET_SUB_TRANSPORT_T to KNET_SUB_TRANSPORT
4872
4873 [logging] rename KNET_SUB_SWITCH_T to KNET_SUB_DSTCACHE
4874
4875 [logging] assign blocks to different logging subsystems
4876 - drop KNET_SUB_PMTUD that was unused
4877 - add KNET_SUB_TRANSPORT_T that was missing
4878 - switch from "common" to "unknown" for odd logging
4879 - fix up code around to allow holes in the structs
4880 - fix up test suite
4881
4882 2016-12-19 Fabio M. Di Nitto <[email protected]>
4883
4884 [PMTUd] adjust all values around new transports
4885
4886 [build] cleanup conditionals to build SCTP and other protocols
4887 Unsupported protocols just need to return NULL on get_PROTO_transport
4888 and contain all protocol conditionals within the transport_PROTO.c.
4889
4890 There is still one conditional in RX/TX thread that needs cleaning
4891 but it's a rather complex operation and leave it in for now.
4892
4893 [transports] initialize transports with knet_handle and protects from missing proto ops
4894
4895 [listener] drop unnecessary debugging code
4896
4897 2016-12-18 Fabio M. Di Nitto <[email protected]>
4898
4899 [sctp] _handle_incoming_sctp don't leak accepted_socks on errors
4900 add a check to verify max connections or we will leak FDs somewhere at some point
4901
4902 [logging] threads should have _t naming
4903
4904 2016-12-17 Fabio M. Di Nitto <[email protected]>
4905
4906 [sctp] don't pass notification packets to the decoder
4907
4908 2016-12-16 Fabio M. Di Nitto <[email protected]>
4909
4910 [build] make SCTP build optional (default: on)
4911
4912 2016-12-16 Christine Caulfield <[email protected]>
4913
4914 sctp: Fix a couple of crashes & tidy on_epoll usage
4915 Also fix locking in sctp shutdown
4916
4917 2016-12-16 Fabio M. Di Nitto <[email protected]>
4918
4919 [build] don't leak pam libraries all over the place
4920
4921 [links] fix segfault when using dynamic ip addresses
4922
4923 Update TODO list
4924
4925 [build] cleanup configure.ac and Makefile.am
4926 - check for libqb only if building kronosnetd
4927 - check for SCTP headers and fails if we don't find them
4928 - export DEBUG conditional in Makefiles
4929 - conditionally include netutils.c for DEBUG
4930
4931 [rpm] fix BuildRequires for SCTP
4932
4933 [kronosnetd] Fix build after SCTP merge
4934 Default to UDP for now, add support for multiple protocols later on
4935
4936 [libknet] fix typo
4937
4938 2016-12-15 Fabio M. Di Nitto <[email protected]>
4939
4940 Update TODO list
4941
4942 Merge pull request #18 from chrissie-c/topic-cc-transports
4943 Factor out protocol code and add new ones
4944
4945 2016-12-13 Christine Caulfield <[email protected]>
4946
4947 internals: Finish renaming QB_LISt macros to KNET_LIST
4948
4949 libknet: Fix things raised in review
4950
4951 2016-12-02 Christine Caulfield <[email protected]>
4952
4953 libknet:add TCP transport
4954
4955 libknet: add SCTP transport
4956
4957 libknet: Add support for multiple transports
4958 Changes the API to add transport to
4959 knet_link_set_config & knet_link_get_config
4960
4961 libknet: Factor out UDP into its own file
4962 With a transport_common file for protocol-independant things.
4963
4964 2016-11-25 Fabio M. Di Nitto <[email protected]>
4965
4966 [poc] remove iov-enc code now merged in master
4967
4968 [tests] Add knet_bench as more flexible replacement for ping_test
4969
4970 [send/recv] Fix bug in handling HOSTINFO packets seq num
4971 the bcast/ucast seq num were being mixed up, causing packet
4972 loss on link up/down (also in active mode).
4973
4974 2016-11-21 Fabio M. Di Nitto <[email protected]>
4975
4976 [tests] add facility to stop a whole knet handle in one shot
4977
4978 [tests] allow scheduler to be configured and allow start_logging to be called multiple times safely
4979 also test crypto_bench with multiple schedulers
4980
4981 2016-11-19 Fabio M. Di Nitto <[email protected]>
4982
4983 [tests] use SCHED_RR for crypto_bench
4984
4985 2016-11-16 Fabio M. Di Nitto <[email protected]>
4986
4987 [tests] add options to configure scheduler
4988
4989 [tests] move crypto bench to use start_logging
4990
4991 [tests] add knet logs: prefix when no logs are being recorded
4992
4993 [tests] add warning on when to use or not to use start_logging
4994
4995 [tests] Fix Makefile.am to build check_PROGRAMS
4996
4997 [tests] add logging thread support to simplify writing tests
4998
4999 2016-11-15 Fabio M. Di Nitto <[email protected]>
5000
5001 Merge pull request #16 from miz-take/prototype
5002 saved errno when epoll_ctl failed
5003
5004 2016-11-15 Takeshi MIZUTA <[email protected]>
5005
5006 saved errno when epoll_ctl failed
5007
5008 2016-11-15 Fabio M. Di Nitto <[email protected]>
5009
5010 Update TODO list
5011
5012 [tests] rename fun_crypto_benchmark to crypto_bench
5013 and create a benchmark section in tests/Makefile.am
5014
5015 [tests] fix up input handling on crypto tests
5016
5017 [crypto] drop the unnecessary hash_block code
5018 this was a leftover from corosync port of the code and it was
5019 a failsafe for buffer allocation apparently. buffer allocation
5020 was completely reworked in corosync to dynamically account for
5021 headers and so on.
5022
5023 this also removes one memcpy and shave some more time in hashing
5024 operations.
5025
5026 [tests] dynamically allocate memory for int_crypto_test
5027 and make it easier to extend text to some input boundaries.
5028
5029 also make check-memcheck doesn't appear to spot bad memory access
5030 on allocated-on-stack buffers
5031
5032 [test] move benchmark outside of regular testing
5033
5034 [crypto] cleanup some headers and drop duplicate code
5035
5036 [crypto] introduce (nss)crypto_encrypt_and_signv
5037 and move crypto_encrypt_and_sign to use the v version.
5038
5039 [poc] add io-hashing PoC
5040
5041 [tests] fix hello world! test
5042
5043 2016-11-07 Fabio M. Di Nitto <[email protected]>
5044
5045 [tests] Add basic functional crypto benchmark test
5046
5047 2016-11-06 Fabio M. Di Nitto <[email protected]>
5048
5049 [tests] add internal crypto basic test
5050
5051 Fix build with different gcc
5052
5053 Integrate it all with build
5054
5055 Simplify build of poc-code
5056
5057 Simplify even further. EncContext tracks the location of the buffer and doesn't care about tmp_len
5058
5059 fix PoC to handle buffer data alignment
5060
5061 2016-11-05 Fabio M. Di Nitto <[email protected]>
5062
5063 Add libnss iovec PoC
5064 this is a small modified version of the code example available here:
5065
5066 https://developer.mozilla.org/en-US/docs/Mozilla/Projects/NSS/nss_sample_code/NSS_Sample_Code_sample2
5067
5068 to show that it is possible to encrypt from multiple src buffers
5069 into a final buffer.
5070
5071 this approach will allow us to remove a lot of memcpy around the
5072 recv/send thread and replace them with iovec mapping all across.
5073
5074 Possible side effects:
5075 - less memory consumption (mapping vs allocating)
5076 - more complex code / crypto (internal) API changes
5077
5078 Add poc-code directory and move access-list code from Chrissie there
5079 Over time we have created a lot of small PoC code to make knet better.
5080 Most of it is already integrated but other chunks are still floating around.
5081
5082 I prefer to keep them here so that:
5083
5084 1) they don't get lost
5085 2) it will be easier to integrate them once we get to that point
5086 3) we can share PoC and work on them together
5087
5088 Add notes for package maintainers
5089
5090 2016-10-22 Fabio M. Di Nitto <[email protected]>
5091
5092 [build] use rpm bcond for a more elegant solution to override ./configure defaults
5093
5094 [build] fix spec/srpm/rpm build
5095 build only packages that we are configuring for (--enable-libtap,
5096 --enable-kronosnetd) and have spec file depends on the latest
5097 configure run to regenerate the info
5098
5099 2016-10-21 Fabio M. Di Nitto <[email protected]>
5100
5101 [build] fix make distcheck to include kronostned/libtap coverage
5102
5103 [build] fix conditional build to pass make distcheck
5104
5105 [tests] fix return code checking
5106 spotted by gcc 5.3.1
5107
5108 Merge pull request #15 from chrissie-c/master
5109 config: Make libtap & kronosnetd optional
5110
5111 2016-10-21 Christine Caulfield <[email protected]>
5112
5113 Make sure we build libtap if kronosnetd is also being built
5114 I didn't think it worth checking for and using local libtap,
5115 if anybody else wants that I will take patches.
5116
5117 config: Make libtap & kronosnetd optional
5118
5119 2016-10-19 Fabio M. Di Nitto <[email protected]>
5120
5121 Merge pull request #14 from jfriesse/include-compat-h
5122 [libknet] Really use compat when needed
5123
5124 2016-10-19 Jan Friesse <[email protected]>
5125
5126 [libknet] Really use compat when needed
5127 Previous patch added compatatibility layer for systems without
5128 (recv|send)mmsg but sadly forgot include compat.h when needed.
5129
5130 2016-10-18 Fabio M. Di Nitto <[email protected]>
5131
5132 Merge pull request #13 from jfriesse/genericmmsg
5133 [libknet] Support systems without (send|recv)mmsg
5134
5135 2016-10-18 Jan Friesse <[email protected]>
5136
5137 [libknet] Support systems without (send|recv)mmsg
5138 Implemented are generic (slightly reduced functionality) functions and
5139 for systems with syscall available but libc prototypes unavailable
5140 also syscall wrapper.
5141
5142 2016-10-15 Fabio M. Di Nitto <[email protected]>
5143
5144 [send/recv] err is never <0 in that code path
5145
5146 [listener] fix listener sock configuration to match other sockets
5147 add send buffer and avoid EAGAIN issues on high load by delegating
5148 buffering to the kernel
5149
5150 2016-10-13 Fabio M. Di Nitto <[email protected]>
5151
5152 Merge pull request #12 from jfriesse/lib64
5153 [build] Make sure lib64 is used when available
5154
5155 Merge pull request #11 from jfriesse/donotoverridesyncindex
5156 [libknet] Do not override symbols
5157
5158 Merge pull request #10 from jfriesse/removeprlimit
5159 [libknet] Use getrlimit instead of prlimit
5160
5161 2016-10-13 Jan Friesse <[email protected]>
5162
5163 [build] Make sure lib64 is used when available
5164 Default location for libraries is PREFIX/usr/lib. This doesn't work well
5165 on systems with mixed 64 and 32-bit libs.
5166
5167 [libknet] Do not override symbols
5168 sync and index are both stdlib symbols so it's better not to use them as
5169 function arguments.
5170
5171 [libknet] Use getrlimit instead of prlimit
5172 prlimit is Linux specific. Because passed pid is 0, we can replace
5173 prlimit call by portable getrlimit without loose of functionality.
5174
5175 2016-09-15 Fabio M. Di Nitto <[email protected]>
5176
5177 [tests] Enable pmtud setfreq test in helgrind context
5178 update helgrind suppression file
5179
5180 [PMTUd] rewrite thread to follow the same footsteps of heartbeat thread
5181 drop all those notification mechanisms that are a locking nightmare
5182 with very little to none gain and use a loop similar to heartbeat.
5183
5184 Also break down the code in 3 sections:
5185 - main thread loop
5186 - pmtud data gathering per link
5187 - pmtud data gathering for all hosts/links
5188
5189 [threads] move thread timer resolution to common section
5190
5191 2016-09-13 Fabio M. Di Nitto <[email protected]>
5192
5193 [tests] move all API checks makefile bits to a separate file
5194 prepare for int(ernal)_ and fun(ctional)_ tests
5195
5196 [tests] drop normal info and leave only progress/warnings/errors
5197
5198 [tests] Add knet_handle_enable_pmtud_notify tests
5199
5200 [tests] remove unnecessary host_remove
5201
5202 [tests] Add knet_handle_pmtud_get tests
5203
5204 [PMTUd] Fix knet_handle_pmtud_get API sanity checks
5205
5206 [tests] Add knet_handle_pmtud_getfreq tests
5207
5208 [PMTUd] Fix getfreq API check
5209
5210 Revert "[tests] enable memcheck and helgrind checks by default"
5211 This reverts commit 7d3ca6729c224c6a592a71ba8813f3d5b811912c.
5212
5213 [tests] enable memcheck and helgrind checks by default
5214
5215 2016-09-11 Fabio M. Di Nitto <[email protected]>
5216
5217 [tests] add api_knet_handle_pmtud_setfreq tests
5218 currently not enabled for helgrind checks. PMTUd thread needs a good rewrite first
5219 but the API is fine.
5220
5221 [tests] expand knet_send for basic send checks
5222
5223 [tests] add api_knet_send_sync tests and update helgrind suppression file
5224
5225 [send/recv] Fix checks for knet_send_sync and move sync test earlier
5226
5227 [tests] print warnings for tests with exception enabled
5228
5229 [tests] export check type to all tests
5230 This allows to skip/warn on some tests while we fix the PMTUd thread
5231
5232 2016-09-06 Fabio M. Di Nitto <[email protected]>
5233
5234 [tests] remove stray line
5235
5236 2016-09-04 Fabio M. Di Nitto <[email protected]>
5237
5238 [tests] add knet_send tests
5239
5240 [handle] Allow threads to recognized shutdown before stopping them.
5241 On busy threads and shutdown, there is a chance that the thread
5242 didn't have time to process the shutdown requests and even tho
5243 we are closing down, some locks can be destroyed while still locked.
5244
5245 In theory we could have each thread report that it has ack'ed the shutdown
5246 and then close, but a sleep seems to be more than good enough for now.
5247
5248 [send/recv] reworking locking on send to link
5249 protect reading of datafd when processing outgoing data
5250 and avoid unnecessary double locking in _handle_send_to_links
5251 (calling functions both uses locking already).
5252
5253 [handle] expand knet_send input sanity checks
5254
5255 [tests] fix ping_test to deal with knet_recv API fixes
5256
5257 [tests] fix memcpm test and typos
5258
5259 2016-09-03 Fabio M. Di Nitto <[email protected]>
5260
5261 [tests] Add knet_recv tests
5262
5263 [handle] fix checks around knet_recv
5264
5265 2016-08-28 Fabio M. Di Nitto <[email protected]>
5266
5267 [tests] add knet_handle_get_datafd tests
5268
5269 [tests] add knet_handle_get_channel tests
5270
5271 [tests] add knet_handle_remove_datafd tests
5272
5273 [tests] add knet_handle_add_datafd tests
5274
5275 [tests] drop suprious cleanup entry
5276
5277 [tests] add knet_handle_enable_sock_notify tests
5278
5279 [tests] add knet_handle_enable_filter tests
5280
5281 2016-08-26 Fabio M. Di Nitto <[email protected]>
5282
5283 [tests] add knet_host_enable_status_change_notify tests
5284
5285 [tests] add knet_host_get_status tests
5286
5287 2016-08-21 Fabio M. Di Nitto <[email protected]>
5288
5289 [tests] add knet_link_get_status tests
5290
5291 [tests] re-order test executions based on class
5292
5293 [tests] align coding style and error checking
5294
5295 [tests] add knet_handle_setfwd tests
5296
5297 2016-08-17 Fabio M. Di Nitto <[email protected]>
5298
5299 [crypto] fix nss init/shutdown memory leaks
5300 make nss_init more robust across invokations
5301
5302 [build] separate memcheck and helgrind suppressions files
5303
5304 2016-08-13 Fabio M. Di Nitto <[email protected]>
5305
5306 [tests] add knet_handle_crypto tests
5307
5308 [crypto] fix knet_handle_crypto config checks
5309 remove redundant and duplicated code from nsscrypto
5310
5311 2016-08-12 Fabio M. Di Nitto <[email protected]>
5312
5313 [tests] fix out-of-tree path to valgrind suppression file
5314
5315 [build] make clang happy
5316
5317 [crypto] drop redudant checks
5318 spotted by clang
5319
5320 [build] make clang happy
5321
5322 [handle] Fix check for malloc result
5323 spotted by clang
5324
5325 [build] fix configure.ac to deal with clang return codes being weird
5326 when testing for an option alone clang will return a warning and RC=0
5327 unless -Werror is also specified, and that warning switch to RC=1.
5328
5329 [tests] fix missing check for api_knet_link_get_priority
5330
5331 [tests] add knet_link_get_link_list tests
5332
5333 [tests] add knet_link_get_enable tests and fix a potential segfault
5334
5335 [tests] fixup knet_link_set_enable tests
5336 add disable code checking and silence a helgrind warning.
5337
5338 helgrind complains about the heartbeat code writing back to the link
5339 status vs knet_host_add.
5340
5341 knet_host_add is using an wr lock, heartbeat code a read lock.
5342
5343 the heartbeat code is the only one changing two ping timers values.
5344 nothing else in the code is using/changing them.
5345
5346 knet_host_add does a memset at alloc time.
5347
5348 tho, it's impossible to hit the HB code _before_ or _during_ knet_host_add,
5349 therefor it's safe to silence the warning.
5350
5351 2016-08-11 Fabio M. Di Nitto <[email protected]>
5352
5353 [tests] add knet_link_set_enable tests
5354
5355 2016-08-10 Fabio M. Di Nitto <[email protected]>
5356
5357 [tests] add knet_link_get_priority tests
5358
5359 2016-08-09 Fabio M. Di Nitto <[email protected]>
5360
5361 [tests] add knet_link_set_priority tests
5362
5363 2016-08-05 Fabio M. Di Nitto <[email protected]>
5364
5365 [tests] add progress message
5366
5367 2016-07-02 Fabio M. Di Nitto <[email protected]>
5368
5369 [build] chose between the lesser of 2 evils
5370
5371 [tests] rename all tests to add postfix _test
5372
5373 2016-07-01 Fabio M. Di Nitto <[email protected]>
5374
5375 [link] rename knet_link_get_timeout to knet_link_get_ping_timers and add related tests
5376 Also in the process, make sure to init timers when configuring a link
5377 and export link defaults in libknet.h
5378
5379 [link] rename knet_link_set_timeout to knet_link_set_ping_timers and related tests
5380
5381 [tests] add knet_link_get_pong_count tests
5382
5383 [tests] fix knet_link_set_pong_count check
5384
5385 [tests] Add knet_link_set_pong_count tests
5386
5387 [tests] remove unused variable
5388
5389 [link] fix knet_link_get_config API and add related tests
5390
5391 2016-06-30 Fabio M. Di Nitto <[email protected]>
5392
5393 [libtap] make testing of symbols more consistent with libknet
5394
5395 [build] be consistent in building libraries and use same export standards
5396
5397 [tests] make it consistent and easier to run check-memcheck and check-helgrind
5398
5399 2016-06-29 Fabio M. Di Nitto <[email protected]>
5400
5401 [tests] add api_knet_link_set_config tests
5402
5403 2016-06-27 Fabio M. Di Nitto <[email protected]>
5404
5405 [tests] don't sort output and use header order to display API calls
5406 makes it easier to follow the logic
5407
5408 2016-06-26 Fabio M. Di Nitto <[email protected]>
5409
5410 [tests] extend api test coverage to match header and binary exported symbols
5411
5412 [libknet] exports only public API symbols
5413
5414 [tests] fix api_knet_log_set_loglevel incorrect exit sequence
5415
5416 [tests] Add knet_log_get_loglevel tests
5417
5418 [tests] add knet_log_set_loglevel tests
5419
5420 2016-06-25 Fabio M. Di Nitto <[email protected]>
5421
5422 [build] add comment
5423
5424 [logging] fix knet_log_get_loglevel_id crash, return code and add related tests
5425
5426 [logging] fix knet_log_get_loglevel_name crash, return correct values and add related tests
5427
5428 [logging] Fix knet_log_get_subsystem_name crash, make call consistent and add related tests
5429
5430 [logging] fix knet_log_get_subsystem_id crash and add related tests
5431
5432 2016-06-24 Fabio M. Di Nitto <[email protected]>
5433
5434 [host] fix possible race condition in accessing host data
5435 bonus: remove a lock used on all threads all the time
5436 cons: traffic can pause more often towards a host if another host links
5437 are not stable
5438
5439 [host] fix knet_host_get_policy API and add related tests
5440
5441 [host] Fix knet_host_set_policy API and add tests
5442
5443 [tests] add knet_host_get_host_list test
5444
5445 [host] fix knet_host_get_id_by_host_name API and add related tests
5446
5447 [host] fix knet_host_get_name_by_host_id API and add related tests
5448
5449 2016-06-21 Fabio M. Di Nitto <[email protected]>
5450
5451 [tests] add simple tool to report api testing coverage
5452
5453 2016-06-20 Fabio M. Di Nitto <[email protected]>
5454
5455 [host] return an error on names that are too long vs truncating
5456
5457 [tests] add api_knet_host_set_name test
5458
5459 [host] fix knet_host_set_name api checks
5460 verify that we actually get a name from the caller
5461 and make sure name is unique or search would break
5462
5463 2016-06-19 Fabio M. Di Nitto <[email protected]>
5464
5465 [tests] add knet valgrind suppression file
5466 remove some error reporting for now, since they are known and
5467 harmless race conditions.
5468
5469 [tests] make ping test run under valgrind and init all memory correctly
5470
5471 [send/recv] make sure to initialize memory properly
5472
5473 [send/recv] fix invalid memory access when checking channel values
5474
5475 2016-06-18 Fabio M. Di Nitto <[email protected]>
5476
5477 [tests] add api_knet_host_remove test
5478
5479 2016-06-17 Fabio M. Di Nitto <[email protected]>
5480
5481 [global] move netutils to common dir (take 2)
5482
5483 [host] fix build failure
5484
5485 [global] move netutils to common dir
5486 those will be handy for libknet testing
5487
5488 [tests] add api_knet_host_add test
5489
5490 [host] perform check before locking
5491
5492 [tests] add api_knet_handle_free tests
5493
5494 [logging] rework locking around logging
5495 make sure to check/fill data after locking and complete write
5496 before release the lock.
5497
5498 also use trylock vs lock since most log are called while already
5499 locked and it's cleaner.
5500
5501 [handle] drop redundant check in knet_handle_free
5502 If there are no hosts configured, there are also no listeners configured
5503
5504 [tests] make it clear where log is coming from
5505 and avoid confusion from logs generated from our test suite
5506
5507 [tests] fix some exit paths to flush logs and close fds
5508
5509 2016-06-16 Fabio M. Di Nitto <[email protected]>
5510
5511 [handle] drop redundant check now the the whole process is locked at global level
5512
5513 [tests] add api_knet_handle_new test
5514
5515 [PMTUd] Allow thread to have time to react to shutdown before cancelling the thread
5516 and solves an issue deleting locks (spotted by helgrind) on exit
5517
5518 [handle] cleanup socketpair creation for good
5519 Better version of 1f890b897474352a376b27cfe79365237791d66f
5520
5521 Don't accept fd values higher than max fd allowed by the OS
5522
5523 2016-06-14 Fabio M. Di Nitto <[email protected]>
5524
5525 [threads] fix locking across the board for shutdown
5526
5527 2016-06-13 Fabio M. Di Nitto <[email protected]>
5528
5529 [libtap] make sure memory is init properly (according to valgrind)
5530
5531 [tests] make helgrind errors a failure
5532
5533 [handle] be consistent in applying sockopts around
5534
5535 [handle] fix fd leak and sock option settings
5536 apparently int math has never been my strong side
5537
5538 [tests] fix running tests with valgrin under libtool
5539
5540 Update gitignore
5541
5542 [tests] fix special check include
5543
5544 Revert "[tests] move all tests to top level directory"
5545 This reverts commit d9adee9b923776ea5813b5a024c16a58b8d989c0.
5546
5547 [tests] move all tests to top level directory
5548
5549 [tests] fix libtap test
5550
5551 [test] Add common check for root user
5552 knet tests can't really run as normal user
5553
5554 [tests] add some common facility to run shell codes within tests
5555 code stolen from libtap, but useful to drive iptables and alike during
5556 some tests
5557
5558 Fix core dump on exit path when running as user and related test case.
5559 Drop the max host to 1024 for testing.
5560
5561 [tests] add common return codes as understood by automake
5562
5563 2016-06-12 Fabio M. Di Nitto <[email protected]>
5564
5565 Move libknet tests in their subdir in preparation for proper test suite
5566
5567 [PMTUd] export protocol overhead information
5568
5569 2016-06-11 Fabio M. Di Nitto <[email protected]>
5570
5571 Update TODO list
5572
5573 [PMTUd] Drop concept of link_mtu and fix data mtu exported values (pass 1)
5574 Cleanup a poor attempt to export all MTU data to the users.
5575
5576 The original idea was to provide information about real link onwire mtu
5577 and max user data mtu (that could be sent on wire without fragmentation)
5578 that adjusted the value for IP protocol, knet header and crypto (if configured).
5579
5580 Effectively there is no use for the real link onwire mtu and it only involves
5581 extra complexity to export lots of extra data to the end user that would have
5582 to redo the calculations internally to gather data mtu.
5583
5584 Also fix the status.mtu value to match knet view of every link.
5585
5586 2016-06-03 Fabio M. Di Nitto <[email protected]>
5587
5588 Update TODO list
5589
5590 Update TODO list
5591
5592 [PMTUd] fix link instability due to too short PMTUd timeouts
5593 Partially roll back changes from 004bd26a
5594
5595 2016-06-01 Fabio M. Di Nitto <[email protected]>
5596
5597 Update TODO
5598
5599 [socket] internal socketpair should use same recvbuf as any other socket
5600
5601 2016-05-16 Fabio M. Di Nitto <[email protected]>
5602
5603 Merge pull request #4 from mbaldessari/arm-fix
5604 Initialize savederrno variable
5605
5606 2016-05-15 Michele Baldessari <[email protected]>
5607
5608 Initialize savederrno variable
5609 This fixes an error seen on F23 arm/gcc:
5610 In file included from threads_send_recv.c:26:0:
5611 threads_send_recv.c: In function '_handle_send_to_links_thread':
5612 logging.h:19:2: error: 'savederrno' may be used uninitialized in this
5613 function [-Werror=maybe-uninitialized]
5614 log_msg(knet_h, subsys, KNET_LOG_ERR, fmt, ##args)
5615 ^
5616 threads_send_recv.c:475:6: note: 'savederrno' was declared here
5617 int savederrno, docallback = 0;
5618 ^
5619 cc1: all warnings being treated as errors
5620 Makefile:843: recipe for target 'libknet_la-threads_send_recv.lo' failed
5621 make[2]: *** [libknet_la-threads_send_recv.lo] Error 1
5622 make[2]: Leaving directory
5623 '/home/jenkins/workspace/kronosnet/arm-fedora-23-gcc/libknet'
5624 Makefile:506: recipe for target 'all-recursive' failed
5625 make[1]: *** [all-recursive] Error 1
5626 make[1]: Leaving directory
5627 '/home/jenkins/workspace/kronosnet/arm-fedora-23-gcc'
5628 Makefile:417: recipe for target 'all' failed
5629 make: *** [all] Error 2
5630
5631 2016-03-06 Fabio M. Di Nitto <[email protected]>
5632
5633 [PMTUd] fix several issues around PMTUd and improve discovery performances
5634 - Fix a race condition around first time MTU discovery where MTU value
5635 would be out of the wazooo
5636 - Fix a case where onwire MTU value would be smaller than crypto header
5637 and generate errors
5638
5639 - Make MTU value a decision factor if a link is usable or not.
5640 If MTU is not valid (either the discovery process failed) or
5641 the value is not sane, the link will be removed from the available
5642 data link pool till the problem is resolved. Once the problem is
5643 solved the link will recover automatically.
5644
5645 - Have link up/down status trigger a PMTUd process to speed up
5646 detection.
5647
5648 - Add ability to detect smaller MTU than minimum defined by protocol.
5649
5650 - Improve error reporting across the board.
5651
5652 - Make PMTUd timeout dynamic based on current link latency. This change
5653 makes detection of small MTU very fast.
5654
5655 - Sanity check values returned by PMTUD to be in range between
5656 minimum MTU defined by given protocol and maximum that knet can handle.
5657
5658 - Trigger host cache reculaculation on MTU validity change to reduce
5659 risk of packet loss on bad links.
5660
5661 2016-02-16 Fabio M. Di Nitto <[email protected]>
5662
5663 [send/recv] fix handling of duplicated HOSTINFO pckt with active-active links
5664
5665 [internal] rename list_rwlock to global_rwlock
5666
5667 2016-02-15 Fabio M. Di Nitto <[email protected]>
5668
5669 [host] fix memory leak in knet_host_add in case of failure
5670 make it easier to read knet_host_remove (no functional changes)
5671
5672 [internal] document usage of mutex and locks
5673
5674 Update TODO list
5675
5676 Update TODO list
5677
5678 2016-02-11 Fabio M. Di Nitto <[email protected]>
5679
5680 Update TODO
5681
5682 2016-02-10 Michele Baldessari <[email protected]>
5683
5684 [docs] Fix docbook typo
5685
5686 2016-02-10 Fabio M. Di Nitto <[email protected]>
5687
5688 [crypto] fix crash when receiving non crypto packets on a crypto socket
5689
5690 [onwire] let's try to be smart and preallocate space for future features
5691
5692 2016-02-09 Michele Baldessari <[email protected]>
5693
5694 Merge pull request #3 from mbaldessari/wip_overflow_fedora
5695 Fix small string overflow
5696
5697 Fix small string overflow
5698 In file included from /usr/include/stdio.h:936:0,
5699 from ping_test.c:12: In function ‘snprintf’,
5700 inlined from ‘main’ at ping_test.c:504:3: /usr/include/bits/stdio2.h:64:10: error: call to
5701 __builtin___snprintf_chk will always overflow destination buffer [-Werror]
5702 return __builtin___snprintf_chk (__s, __n, __USE_FORTIFY_LEVEL - 1,
5703
5704 Merge pull request #2 from mbaldessari/smatch_fixes
5705 Assorted fixes found via smatch
5706
5707 Fix another indentation issue
5708
5709 Fix equality test
5710
5711 Fix broken indentations
5712
5713 Move len to ssize_t
5714 len is used as a return code for both select()(int) and read()(ssize_t)
5715 so make it an ssize_t type instead of size_t which is unsigned.
5716
5717 Fix error path test equality
5718
5719 2016-02-09 Fabio M. Di Nitto <[email protected]>
5720
5721 Update TODO list
5722
5723 2016-02-07 Fabio M. Di Nitto <[email protected]>
5724
5725 Update TODO list
5726
5727 2016-02-06 Fabio M. Di Nitto <[email protected]>
5728
5729 [send/recv] add knet_send_sync function
5730 read carefully the doc in libknet.h or suffer the ethernal doom in hell
5731
5732 2016-01-10 Fabio M. Di Nitto <[email protected]>
5733
5734 [sockets] first cut at making socket management more consistent
5735 and fix sendmmsg handling
5736
5737 2016-01-05 Fabio M. Di Nitto <[email protected]>
5738
5739 Update TODO list
5740
5741 2015-12-24 Fabio M. Di Nitto <[email protected]>
5742
5743 Update TODO
5744
5745 2015-12-20 Fabio M. Di Nitto <[email protected]>
5746
5747 [send/recv] rework handling of 0 byte read/write on local sockets
5748 considering there is no obvious use case trying to send 0 byte packets
5749 and that 0 byte packtes are filtered by iovec calls across the all
5750 code, there is no point trying to handle 0 byte differently from
5751 any other socket error.
5752
5753 this commit makes sure that every time there is an error (-1) or a 0
5754 byte read from the locally provided sockets, a call back is issued.
5755
5756 on read errors (-1) the socket will be removed from the epoll
5757 to avoid spinning and it is safe to call knet_handle_remove_datafd
5758 afterwards to remove it completely.
5759
5760 it is now mandatory to enable a sock_notify callback before adding
5761 datafd.
5762
5763 read libknet.h carefully on what the callback is supposed to do
5764 based on the type of socket your application is adding.
5765
5766 2015-12-18 Fabio M. Di Nitto <[email protected]>
5767
5768 Update TODO list
5769
5770 remove stray comment
5771
5772 2015-12-18 Christine Caulfield <[email protected]>
5773
5774 [lib] Fix typos & English in libknet.h file
5775
5776 2015-12-18 Fabio M. Di Nitto <[email protected]>
5777
5778 [send/recv] deal properly with 0 bytes packets on local sockets
5779
5780 [send/recv] be consistent between readv and recvmmsg handling 0 bytes packets
5781
5782 [PMTUd] improve debug message to add iface mtu
5783
5784 [global] add support for multiple local sockets/fds
5785 WARNING: this commit changes API and onwire protocol
5786
5787 It is now possible to configure multiple local sockets to send/recv
5788 data over the same knet handle.
5789
5790 to every send/recv socket a channel is assigned that behaves more
5791 or less like a VLAN.
5792 hostA channel 0 will be delivered to datafd on hostB channel 0
5793 hostA channel 1 to hostB channel 1
5794 etc.
5795
5796 It is possible to configure up to 32 channel, but there is space
5797 to increase this number if necessary.
5798
5799 New API calls:
5800 knet_handle_add_datafd
5801 knet_handle_remove_datafd
5802 knet_handle_enable_sock_notify
5803 knet_handle_get_channel
5804 knet_handle_get_datafd
5805
5806 Notification has been added in cases where a local socket is the result
5807 of accepting a tcp connection that gets disconnected and corrective
5808 action needs to be taken by the application.
5809
5810 Changed API calls:
5811 knet_handle_new
5812 knet_recv
5813 knet_send
5814 knet_handle_enable_filter
5815
5816 onwire changes:
5817 add one byte to transport channel information
5818
5819 channel information are also passed down to the dst_host_filter
5820 in the event the application has needs to modify that on TX/RX
5821 events.
5822
5823 Internal changes:
5824 cleanup how sockpairs are handled in general to make it simpler
5825 to map application side and internal side.
5826
5827 NOTE: callback function has not been properly tested.
5828
5829 2015-12-16 Fabio M. Di Nitto <[email protected]>
5830
5831 Update TODO list
5832
5833 [internal] switch from pipe to socketpairs for consistency
5834
5835 [send/recv] add support for recvmmsg on local socket
5836
5837 Update TODO list
5838
5839 2015-12-14 Fabio M. Di Nitto <[email protected]>
5840
5841 Update TODO
5842
5843 2015-12-13 Fabio M. Di Nitto <[email protected]>
5844
5845 [send/recv] save some more times dealing with unreachable hosts for mcast pckts
5846
5847 [send/recv] rework commit 1de7acb8 a bit
5848 on the exit thread, kh_node is knet_h->host_id, no point
5849 to convert it N times per packet/frag.
5850
5851 [send/recv] drop too verbose log message and fix comment
5852
5853 [send/recv] remove TODO comment
5854 splitting frags across multiple links is not worth the complexity of
5855 the code or the complexity in re-assemblying the packet if links
5856 are at different speed/status
5857
5858 [send/recv] sort code around
5859
5860 Update TODO list
5861
5862 [send/recv] fix bcast data entry for hostinfo packets
5863 this should address the issue Chrissie reported where
5864 first data packet would vanish in void
5865
5866 2015-12-12 Fabio M. Di Nitto <[email protected]>
5867
5868 [send/recv] drop redundant check
5869
5870 [send/recv] fix logging levels
5871
5872 Update configure.ac for sendmmsg
5873
5874 [host] cleanup circular buffer function names
5875
5876 Sort TODO list
5877
5878 [host] add host status change callback notifications
5879 API is also ready for further expansion (external and remote support),
5880 but they are currently not implemented.
5881
5882 allow both notifications and polling of status via knet_host_get_status
5883
5884 use new notification and api polling in ping_test.c as unit test
5885
5886 use the new host status information in the TX thread to skip some
5887 time consuming pckt processing if the destination host(s) is not
5888 reachable
5889
5890 2015-12-11 Fabio M. Di Nitto <[email protected]>
5891
5892 [send/recv] fix data len passed to dst host filter on the RX end
5893
5894 2015-12-10 Fabio M. Di Nitto <[email protected]>
5895
5896 [filters] be consistent and provide private data for all notifications/callbacks
5897
5898 2015-12-09 Fabio M. Di Nitto <[email protected]>
5899
5900 [send/recv] improve dst_host_filter API to include more data to determine destination
5901
5902 [send/recv] add support for dynamic buffer allocation for packet re-assembly
5903 this commit changes the on-wire protocol.
5904
5905 2015-12-08 Fabio M. Di Nitto <[email protected]>
5906
5907 [send/recv] fix host_id value in host_filter
5908 thanks Chrissie for reporting the problem
5909
5910 2015-11-30 Fabio M. Di Nitto <[email protected]>
5911
5912 [send/recv] Fix handling of HOSTINFO packets
5913 HOSTINFO packets are nothing more than special cased DATA packets.
5914
5915 The contents of HOSTINFO is stored inside DATA of DATA packets and
5916 up to the point where we need to use the HOSTINFO data, we need
5917 to treat the packets as DATA for defragmentation and deduplication.
5918
5919 This also fixes a problem when using round-robin and active switching
5920 policy where the HOSTINFO exchange would loop forever in a storm
5921 because duplicate pckts were being handled at once.
5922
5923 2015-11-29 Fabio M. Di Nitto <[email protected]>
5924
5925 Update TODO list
5926
5927 2015-11-28 Fabio M. Di Nitto <[email protected]>
5928
5929 Replace memcpy with memmove across the board
5930 memmove is just a tiny littlebit faster than memcpy.
5931
5932 Update TODO list
5933
5934 Update TODO
5935
5936 [PMTUd] Fix crash and calculation when using crypto
5937 most of the problem was around incorrect typing of variables
5938 and incorrect calculation when using signing without encryption.
5939
5940 2015-11-27 Fabio M. Di Nitto <[email protected]>
5941
5942 [libtap] automatic restore IPv6 address when recovering from MTU flapping
5943 NOTE: if a user manually changes MTU < 1280, your IPv6 addresses
5944 are gone for good.
5945
5946 Update TODO list
5947
5948 2015-11-26 Fabio M. Di Nitto <[email protected]>
5949
5950 Update TODO list
5951
5952 2015-11-25 Fabio M. Di Nitto <[email protected]>
5953
5954 [crypto] align crypto_authenticate_and_decrypt API to crypto_encrypt_and_sign
5955 2 issues solved by this change:
5956
5957 1) API were different and crypto_authenticate_and_decrypt would trash the
5958 incoming packet, that we might want to keep for later usage
5959 (re-switch for example)
5960
5961 2) By using an extre pre-allocated buffer while decrypting incoming packets
5962 we save a whole memcpy and that reduces latency on crypto communications:
5963
5964 pre-patch:
5965
5966 [root@kronosnet-node1-br0 ~]# ping 192.168.12.2 -f -c 1000 -s 65000
5967 PING 192.168.12.2 (192.168.12.2) 65000(65028) bytes of data.
5968
5969 --- 192.168.12.2 ping statistics ---
5970 1000 packets transmitted, 1000 received, 0% packet loss, time 5302ms
5971 rtt min/avg/max/mdev = 5.102/5.283/22.679/0.947 ms, pipe 3, ipg/ewma 5.307/5.223 ms
5972
5973 post-patch:
5974
5975 [root@kronosnet-node1-br0 ~]# ping 192.168.12.2 -f -c 1000 -s 65000
5976 PING 192.168.12.2 (192.168.12.2) 65000(65028) bytes of data.
5977
5978 --- 192.168.12.2 ping statistics ---
5979 1000 packets transmitted, 1000 received, 0% packet loss, time 5196ms
5980 rtt min/avg/max/mdev = 4.997/5.154/6.661/0.129 ms, ipg/ewma 5.201/5.173 ms
5981
5982 (using aes256 and sha1)
5983
5984 2015-11-22 Fabio M. Di Nitto <[email protected]>
5985
5986 [send/recv] init recv buffer array only once
5987
5988 [internal] define/use PCKT_FRAG_MAX instead of UINT8_MAX
5989 makes the code more readable
5990
5991 Fix error message
5992
5993 [send/recv] use recvmmsg on network sockets
5994 On my test systems this change reduces the latency of fragmented packets
5995 from 0.27ms to 0.235ms in average. There are still some optimizations
5996 on how buffers can be initialized that could shave some time on this code
5997 path.
5998
5999 Update TODO
6000
6001 2015-11-18 Fabio M. Di Nitto <[email protected]>
6002
6003 Update copyright dates
6004
6005 [build] update configure.ac
6006
6007 2015-11-16 Fabio M. Di Nitto <[email protected]>
6008
6009 [send/recv] fix data filling in the frag code
6010
6011 [send/recv] use sendmmsg to reduce latency
6012 On my test systems this change reduces the latency from 0.7ms to 0.25ms
6013 when sending 65000 bytes packet and using internal re-assembly code.
6014
6015 Internally use N buffers to split the incoming data and reduce by N frags
6016 the amount of memcpy between kernel and userland.
6017
6018 [internal] reduce memory footprint further down
6019
6020 2015-10-24 Fabio M. Di Nitto <[email protected]>
6021
6022 [global] reduce memory footprint by a long shot
6023
6024 [send/recv] add simplistic packet re-assembly code
6025 this will probably explode if you are using multiple links
6026 in round-robin or active-active mode.
6027
6028 2015-10-21 Fabio M. Di Nitto <[email protected]>
6029
6030 Revert "[global] reduce memory footprint by a long shot"
6031 This reverts commit 4b4290a2fafbeb4d08bf489caa35ea4c205b18b9.
6032
6033 2015-09-20 Fabio M. Di Nitto <[email protected]>
6034
6035 [global] reduce memory footprint by a long shot
6036
6037 [threads] cleanup header
6038
6039 [cleanup] remove unused headers around
6040
6041 [threads] split send recv threads in separate file
6042
6043 [threads] split dsthandler thread in separate file
6044
6045 [threads] split pmtud thread in separate file
6046
6047 [threads] split heartbeat thread in separate file
6048
6049 [threads] implement outgoing pckt fragmentation based on PMTUd
6050
6051 2015-09-16 Fabio M. Di Nitto <[email protected]>
6052
6053 Update TODO list
6054
6055 [global] restrict max packet size to 65536 bytes
6056 and apply workaround in data buf allocation while we cleanup
6057 the whole buffer allocation handling situation
6058
6059 2015-09-15 Fabio M. Di Nitto <[email protected]>
6060
6061 [test] disable crypto rekey test for now
6062
6063 [knetd] fix libqb logging filter
6064 Thanks to Chrissie for the awesome work as usual
6065
6066 [test] fix ping_test data print and add some more info
6067
6068 [handle] add missing include
6069
6070 [handle] fix knet_send sock handling to use correct data flow
6071
6072 [handle] fix init order to have more debug in case of sockpair error
6073
6074 [commodity] add knet_send/knet_recv api calls
6075 based on readv/writev man page, it is highly discouraged to mix
6076 different kind of write/read operations on a given fd (for example
6077 a read with a writev).
6078
6079 since not many finds using iovec that pleasent, those 2 functions
6080 do wrap the more common buff/buff_len calls into iovec implementations.
6081
6082 [handle] API change warning: add support for knet_handle_new to create proper sockpairs
6083 this is a small facility for applications that don't want/need to create
6084 socketpairs to read/write data to/from knet and avoids lots of duplicated
6085 socket management on the client side.
6086
6087 [threads] add warning on fragmented packets
6088 also prepare ping_test for it
6089
6090 2015-09-08 Fabio M. Di Nitto <[email protected]>
6091
6092 [ping] fix sockpair creation to deal with sequence of packets and not streaming
6093
6094 [threads] use iovec calls to read/write from datafd
6095
6096 [cosmetic] rename net_fd to datafd
6097
6098 2015-09-06 Fabio M. Di Nitto <[email protected]>
6099
6100 [handle] add missing errno on failure
6101
6102 [libtap] change error check code to avoid a segfault and make it more readable
6103
6104 2015-04-21 Fabio M. Di Nitto <[email protected]>
6105
6106 [listener] fix 2 cosmetic checks and keep around a useful debugging fn
6107
6108 [threads] make ping error message more verbose
6109
6110 Update TODO list
6111
6112 2015-04-21 Christine Caulfield <[email protected]>
6113
6114 libtap: don't close pipe fd twice
6115 libtap was closing the parent's pipe() fd twice. Sometimes, between the
6116 first and second close, that fd gets reopened by the logging thread as
6117 PF_LOCAL, and used for logging.
6118
6119 Then libtap wakes up again after the fork and closes the fd thinking
6120 it's its pipe fd. knet then opens it again as AF_INET for pinging.
6121
6122 libqb logging then errors on a sendto() and closes it AGAIN, and reopens
6123 it as a PF_LOCAL logging fd. Ping then has a brain haemorrage next time
6124 it comes round to calling sendto().
6125
6126 2015-03-20 Fabio M. Di Nitto <[email protected]>
6127
6128 Update TODO list
6129
6130 2015-03-19 Fabio M. Di Nitto <[email protected]>
6131
6132 [host] move dst cache updates to the right logging subsystem
6133 and return error codes
6134
6135 Update TODO list
6136
6137 [host] create dst_cache sync and async functions and fix link lookup race condition
6138 Prior to this change, there was a potential race condition in which, when shutting
6139 down a link, the destination cache would not be updated in time and pckts would
6140 be lost in an invalid socket.
6141
6142 This change introduced a sync version of the dst cache update and protects
6143 outgoing traffic socket lookup with a mutex.
6144
6145 The only downside of this change is that, in future, when we will introduce
6146 complex switching algorithms, we might have to spend a bit more time
6147 recalculating the network topology on link shutdown and halt traffic switching.
6148
6149 Pro is that we have a smaller gap in lost traffic when reconfiguring the local
6150 node.
6151
6152 2015-03-18 Fabio M. Di Nitto <[email protected]>
6153
6154 [threads] simplify code to send data to links
6155 in preparation to remove race condition in _dst_cache update,
6156 simplify the access path to common code
6157
6158 [onwire] fix regression in macro expansions
6159
6160 2015-03-12 Fabio M. Di Nitto <[email protected]>
6161
6162 [onwire] introduce khpi_link_status_ defines to make code simpler to parse
6163
6164 2015-03-11 Fabio M. Di Nitto <[email protected]>
6165
6166 [onwire] add define for knet_hostinfo structures size
6167
6168 2015-03-10 Fabio M. Di Nitto <[email protected]>
6169
6170 [onwire] rename kf_plink/kf_psize/kf_pdata to their respective more descriptive versions
6171
6172 [onwire] rename kf_link and kf_time to their respective more descriptive versions
6173
6174 [onwire] rename kf_seq_num and kf_data to their respective more descriptive versions
6175
6176 [onwire] add comment
6177
6178 [onwire] more cleanup around _SIZE define usage
6179
6180 2015-03-09 Fabio M. Di Nitto <[email protected]>
6181
6182 [onwire] rename various KNET_FRAME_ definitions to KNET_HEADER_
6183
6184 [onwire] cleanup and use defines
6185 make it easier to understand values and usage of other status info
6186
6187 [onwire] rename link_up_down struct to knet_hostinfo_payload_link_status
6188
6189 [onwire] rename knet_hinfo_data struct to knet_hostinfo
6190 and add comments around the various fields
6191
6192 2015-03-08 Fabio M. Di Nitto <[email protected]>
6193
6194 [onwire] add comment in ifout section
6195
6196 [onwire] start cleaning host_info data structures
6197
6198 [debug] add tool to print out network pckt header size
6199
6200 [onwire] clenaup knet_header_payload_data structure naming
6201
6202 [onwire] clenaup knet_header_payload_ping structure naming
6203
6204 2015-03-07 Fabio M. Di Nitto <[email protected]>
6205
6206 [cleanup] move some values from int to unsigned int
6207
6208 [onwire] clenaup knet_header_payload_pmtud structure naming
6209 and add more comments
6210
6211 [onwire] add more comments
6212
6213 [onwire] add extra comments
6214
6215 [onwire] cleanup knet_header_payload union
6216 use longer name to make it easier to identify nesting of structures
6217
6218 [onwire] rename knet_frame_data to knet_header_payload
6219
6220 [onwire] rename knet_frame to knet_header
6221 also start renaming variables around to make it easier to understand
6222 the level of nesting and add comments for each field
6223
6224 [internal] start renaming things to make code look decent
6225
6226 [internals] Simplify code maintanance by shuffling around some variables
6227 this allows creation of better #define to maintain some parts of the
6228 code related to buffers allocations.
6229
6230 As a bonus, gcc does a better job at packing structs outside of a union
6231 and we save 4 bytes onwire.
6232
6233 2015-03-05 Fabio M. Di Nitto <[email protected]>
6234
6235 [knetd] Fix internal MTU calculation to include tap ethernet frame
6236
6237 2015-03-04 Fabio M. Di Nitto <[email protected]>
6238
6239 [PMTUd] fix mutex deadlock on socket error
6240
6241 [listeners] Use IP_FREEBIND options on sockets
6242 this allows us to bind to IPs that are not yet on the system and
6243 survive the infamous ifdown/ifup <device> test
6244
6245 2015-03-03 Fabio M. Di Nitto <[email protected]>
6246
6247 [pingd] fix ping flood on sock errors
6248 if we get an error on sendto, we would ping flood the link because
6249 the ping_last timestamp was not being updated.
6250
6251 [libknet] rename tap to send to link
6252 plain and simple cleanup
6253
6254 2015-03-02 Fabio M. Di Nitto <[email protected]>
6255
6256 Update TODO list
6257
6258 [knetd] add support for pmtud frequency
6259
6260 [PMUTd] add call to get pmtud interval value
6261
6262 [knetd] Use pmtud values to set tap device MTU
6263
6264 2015-02-28 Fabio M. Di Nitto <[email protected]>
6265
6266 [PMTUd] fix notification callback to add private data
6267 this is necessary to identify the callback when driving multiple
6268 knet instances within the same daemon.
6269
6270 Update TODO list
6271
6272 2014-12-01 Fabio M. Di Nitto <[email protected]>
6273
6274 [PMTUd] add support for pmtud_get for who doesn't want notifications
6275
6276 [PMTUd] add support for PMTUd changes notifications
6277
6278 2014-11-30 Fabio M. Di Nitto <[email protected]>
6279
6280 [PMTUd] distinguish between link_mtu and data_mtu
6281 link_mtu is the pure MTU over the link and doesn't account for knet headers.
6282 data_mtu is what can be really transmitted onwire by the application.
6283
6284 2014-10-23 Fabio M. Di Nitto <[email protected]>
6285
6286 [PMTUd] add support for timer config and fix critical shutdown deadlock in condwait
6287
6288 [PMTUD] change error message
6289
6290 [PMTUd] fix scary message at startup
6291
6292 2014-10-21 Fabio M. Di Nitto <[email protected]>
6293
6294 [PMTUD] fix threads to work with crypto
6295 temporary disable re-keying test as it seems to be break PMTUd
6296
6297 2014-10-20 Fabio M. Di Nitto <[email protected]>
6298
6299 [PMTUD] change variable names to make it easier to read
6300 still doesn't work with crypto enable, but part of the code is getting
6301 ready for it.
6302
6303 2014-08-29 Fabio M. Di Nitto <[email protected]>
6304
6305 [nss] fix calculation for crypto header to include SALT_SIZE
6306 and propagate block size up in the layers to allow PMTUd thread
6307 to round buffers and do crypted PMTUd packets.
6308
6309 2014-08-28 Fabio M. Di Nitto <[email protected]>
6310
6311 [link] add rudimentary PMTUD support.
6312 1) It won't work with crypto enable yet
6313 2) values are not really used yet for anything
6314 3) no API calls implemented yet
6315
6316 2014-08-11 Fabio M. Di Nitto <[email protected]>
6317
6318 [listener] check setsockopts return codes
6319
6320 2014-07-23 Fabio M. Di Nitto <[email protected]>
6321
6322 [nss] make sure that any crypto config requires a proper hashing value
6323 this is mandatory to avoid some collision checks when random junk
6324 is received over the network socket
6325
6326 2014-07-21 Fabio M. Di Nitto <[email protected]>
6327
6328 [libtap] check for error buffers only when necessary
6329
6330 [build] fix pkg-config files to link correctly
6331
6332 [build] libtap requires sys/types.h for size_t definition
6333
6334 2014-07-20 Fabio M. Di Nitto <[email protected]>
6335
6336 [build] cleanup after test-driver
6337 update .gitignore and make maintainer-clean target
6338
6339 2013-11-16 Michele Baldessari <[email protected]>
6340
6341 This fixes build on F20 with gcc 4.8.2
6342
6343 2013-08-31 Fabio M. Di Nitto <[email protected]>
6344
6345 Update slides
6346
6347 2013-06-15 Fabio M. Di Nitto <[email protected]>
6348
6349 libknet: add extra error checking
6350
6351 2013-06-10 Fabio M. Di Nitto <[email protected]>
6352
6353 kronosnetd: add support for pong_count
6354
6355 link: add pong count handling to avoid excessive flapping of traffic
6356
6357 2013-06-09 Fabio M. Di Nitto <[email protected]>
6358
6359 cleanup TODO list
6360
6361 crypto: track size of crypto overhead data
6362
6363 libknet: comment out un-used onwire data for now
6364
6365 2013-03-11 Fabio M. Di Nitto <[email protected]>
6366
6367 Update TODO list
6368
6369 2013-03-10 Fabio M. Di Nitto <[email protected]>
6370
6371 build: fix install target
6372
6373 listener: fix more unistd.h breakage
6374
6375 2013-02-01 Fabio M. Di Nitto <[email protected]>
6376
6377 Every set should have a get...
6378
6379 cleanup a few bits around logging and interal API
6380
6381 2013-01-30 Fabio M. Di Nitto <[email protected]>
6382
6383 libknet: keep host list cached in knet handle
6384
6385 Update TODO list
6386
6387 2013-01-29 Fabio M. Di Nitto <[email protected]>
6388
6389 libknet: add link api call to get list of configured links
6390
6391 Finish to cleanup current logging API
6392
6393 logging: start documenting API and other bits
6394
6395 logging: fix formatting
6396
6397 fix README pam section
6398
6399 Fix ping_test to deal with EINVAL return codes
6400
6401 Update TODO list
6402
6403 libknet: fix knet_link_get_status and document it
6404
6405 kronosnetd: deal with proper error codes from knet_link_get calls
6406
6407 libknet: fix set/get_priority and document them
6408
6409 libknet: fix set/get_timeout and document them
6410
6411 Add knet_link_get_enable call
6412
6413 ping_test: some more fixes
6414 how the hell did this thing ever worked? needs a rewrite
6415
6416 link: fix return code when exit path is clean
6417
6418 2013-01-28 Fabio M. Di Nitto <[email protected]>
6419
6420 fix \n logging
6421
6422 listener: rename link to lnk to avoid conflict with unistd.h
6423
6424 ping_test: fix argc corruption and make output configurable
6425
6426 threads: make one error more specific
6427
6428 libknet: rework knet_link_enable into knet_link_set_enable
6429 make it thread safe and fix all return codes (ret vs errno)
6430
6431 2013-01-27 Fabio M. Di Nitto <[email protected]>
6432
6433 Update TODO list
6434
6435 2013-01-20 Fabio M. Di Nitto <[email protected]>
6436
6437 link: protect set/get config from doing damages
6438
6439 link: rename configured to enabled
6440 we need configured internally for other stuff
6441
6442 2013-01-11 Fabio M. Di Nitto <[email protected]>
6443
6444 libknet: cleanup knet_link_set/get_config
6445 this also allows to kill the madness around dynamic ip management
6446 and saves enother bytes in the ping header
6447
6448 Fix make dist/clean
6449
6450 2013-01-10 Fabio M. Di Nitto <[email protected]>
6451
6452 Fix typo
6453
6454 debian: fix a few bits around and resync with rpm packaging
6455
6456 Update TODO list
6457
6458 Update TODO list
6459
6460 2013-01-08 Fabio M. Di Nitto <[email protected]>
6461
6462 systemd: allow kronosnetd to set real time scheduler
6463 see also https://bugzilla.redhat.com/show_bug.cgi?id=893015
6464
6465 Update TODO list
6466
6467 logging: move string to the beginning of msg
6468 this allow (to some extents) to print log messages directly to
6469 stdout and/or stderr without extra parsing.
6470
6471 NOTE:
6472 - messages are sent from differnet threads and there for they are less readable
6473 when they arrive all together
6474 - there is no printing of message level and subsystem
6475
6476 for simple coding/debugging purposes this is useful, but otherwise
6477 look at how ping_test.c or kronosnetd create a logpipe and read
6478 sequencial messages from there.
6479
6480 Update TODO list
6481
6482 rewrite TODO list to be more useful
6483
6484 2013-01-07 Fabio M. Di Nitto <[email protected]>
6485
6486 libknet: fix HMAC key init
6487
6488 2013-01-06 Fabio M. Di Nitto <[email protected]>
6489
6490 start cleaning TODO list and make it more useful
6491
6492 Sort TODO list
6493
6494 2013-01-02 Fabio M. Di Nitto <[email protected]>
6495
6496 Update TODO list
6497
6498 2013-01-01 Fabio M. Di Nitto <[email protected]>
6499
6500 Update TODO list
6501
6502 2012-12-31 Fabio M. Di Nitto <[email protected]>
6503
6504 pam: use password-auth on systems like fedora that ship it
6505 or passwd as a fallback (debian and others)
6506
6507 rpm: fix integration with systemd (missing sysconfig file) and pam/groupadd
6508
6509 rpm: fix systemd integration with spec file
6510
6511 2012-12-25 Fabio M. Di Nitto <[email protected]>
6512
6513 kronosnetd: fix integration with fedora pam
6514
6515 2012-12-24 Fabio M. Di Nitto <[email protected]>
6516
6517 Update TODO list
6518
6519 Downgrade error from set scheduler to warning
6520
6521 Fix build dependencies for parallel build
6522
6523 Add missing header to final tarball
6524
6525 2012-12-21 Fabio M. Di Nitto <[email protected]>
6526
6527 Start documenting knet_link_* api
6528
6529 2012-12-20 Fabio M. Di Nitto <[email protected]>
6530
6531 Update TODO list
6532
6533 fix crash when multiple links are sharing the same listener
6534
6535 Fix shutdown race conditions between locks/threads/signals
6536 that in turns make knet_handle_free thread safe
6537
6538 Fix log subsystem on _dst_cache_update
6539
6540 make knet_host_(get|set)_policy behave as documented
6541 also fix _dst_cache_update to return errno and use host_id
6542
6543 Update TODO list
6544
6545 make knet_host_get_host_list work as documented in the API
6546
6547 make knet_host_get_id_by_host_name work as documented in the API
6548
6549 make knet_host_get_name_by_host_id work as documented in the API
6550
6551 Fix pthread error to errno relation
6552
6553 make knet_host_set_name comply with documented API
6554
6555 make knet_host_remove comply with documented API
6556 also fix a bunch of other bits
6557
6558 make knet_host_add comply with documented API
6559 also continue renaming node_id to host_id as we go
6560
6561 handle: fix errno return value
6562
6563 2012-12-19 Fabio M. Di Nitto <[email protected]>
6564
6565 Make knet_handle_* call compliant to documented API
6566 allow knet_handle_crypto to be invoked at runtime
6567
6568 libknet: make knet_handle_new / knet_handle_free do what the API says they should
6569
6570 move timespec_diff to threads.h (only user)
6571
6572 Make define name clear
6573
6574 Cleanup internal headers
6575
6576 Move _dst_cache_update to host function set
6577
6578 and maybe add the threads.* files to git might help.. just a tiny little bit
6579
6580 Move all threads code into threads.{c,h}
6581
6582 Move all link related code to link.{c,h}
6583
6584 Move listener function definitions in listener.h
6585
6586 move logging code/definitions to logging.{c,h}
6587
6588 move host.c function definitions into host.h
6589
6590 move some definitions out of internals.h into common.h
6591
6592 Start renaming node_id to host_id
6593
6594 Document knet_handle_crypto
6595
6596 document a couple of more knet_handle_* calls
6597
6598 Update TODO list
6599
6600 2012-12-18 Fabio M. Di Nitto <[email protected]>
6601
6602 libknet: document knet_handle_enable_filter and dst_host_filter_fn
6603
6604 One more cosmetic
6605
6606 libknet: cosmetic fixes to knet_handle_new
6607
6608 libknet: knet_handle_new API cleanup and documentation
6609
6610 libknet: document limits
6611
6612 move comment where it belongs
6613
6614 libknet: make dst_filter a runtime config option
6615
6616 libknet doesn't need a default port
6617 it's a UI issue to deal with
6618
6619 Reduce logging buffer size to allow more messages
6620
6621 Add comments to knet_link and re-arrange
6622
6623 move knet_link into internals.h
6624 hack fest still in place, we need to sort out ABI and locking
6625
6626 2012-12-17 Fabio M. Di Nitto <[email protected]>
6627
6628 libknet api cleanup: move knet_host out of public API
6629 the knet_host_* calls are pretty much set for now.
6630
6631 the new knet_link_* calls are temporary to interface with the old API.
6632
6633 knet_link will die next.
6634
6635 Move --version to library only bits
6636
6637 Update TODO list
6638
6639 2012-12-14 Fabio M. Di Nitto <[email protected]>
6640
6641 reorganize libknet.h in sections
6642 just makes it easier to read as we move stuff around
6643
6644 Drop tests/ subfolder
6645
6646 Move all tests in the respective folders
6647 this makes it easier to isolate what needs changing where and when
6648
6649 Move hellgrind and valgrind checks to a common make snippet
6650
6651 Cleanup tests for kronosnetd
6652
6653 remove obsoleted test tool
6654
6655 Drop one more unrequired config bit
6656
6657 move all logging definitions in the same section
6658
6659 drop redundant dst_filter config check
6660
6661 fix segfault in ping_test src_port handling
6662
6663 move listener only define
6664
6665 cleanup crypto headers
6666
6667 move ping size to onwire and make name more clear
6668
6669 drop unrequired header
6670
6671 libknet: create onwire.h to contain/define all packet(s) data/info
6672
6673 Make rename clean
6674
6675 rename libknet-private.h to internals.h
6676
6677 2012-12-13 Fabio M. Di Nitto <[email protected]>
6678
6679 Move etherfilter where it belongs
6680
6681 2012-12-07 Fabio M. Di Nitto <[email protected]>
6682
6683 libknet: fix host info propagation on kronosnetd restart
6684 this avoids lots of incorrect chit/chat and make sure
6685 to send data when required
6686
6687 Update TODO list
6688
6689 2012-12-06 Fabio M. Di Nitto <[email protected]>
6690
6691 kronosnetd: fix init script return codes
6692
6693 2012-12-04 Fabio M. Di Nitto <[email protected]>
6694
6695 One more attempt to fix make dist out-of-tree
6696
6697 better fix to fix git fix for build out of tree fix
6698
6699 Fix out-of-tree build
6700
6701 Update TODO list
6702
6703 fix O_NONBLOCK fd setting functions
6704 *facepalm*
6705
6706 logging: cleanup logging to avoid clogging the logfd pipe
6707 move it to its own thread and clean up logfd handling
6708
6709 Fix debian generation
6710
6711 debian: fix date spacing/formatting
6712
6713 libtap: small step back to .so.0.0.0
6714 no point to have 1.0.0 when we are not even sure the API is final
6715
6716 build: fix maintainer-clean when building on != debian based distro
6717
6718 Update .gitignore
6719
6720 libknet: fix data forwarding check
6721
6722 Update TODO list
6723
6724 2012-12-03 Fabio M. Di Nitto <[email protected]>
6725
6726 Update TODO list
6727
6728 build: fix make dist and make maintainer-clean
6729
6730 add support to build deb packages
6731 use: make debs
6732
6733 2012-12-02 Fabio M. Di Nitto <[email protected]>
6734
6735 generate package version from git/configure
6736
6737 2012-11-29 Fabio M. Di Nitto <[email protected]>
6738
6739 Update TODO list
6740
6741 Fix copyright/authors all over
6742
6743 kronostned: add logrotate file
6744
6745 update TODO list
6746
6747 Fix super annoying kronosnet/kronosnetd discrepancies around
6748 and fix a few rpmlint warnings/errors
6749
6750 Update .gitignore
6751
6752 kronosnetd: add basic spec file
6753
6754 add native debian packaging
6755 this is based on debhelper version 7 to allow build
6756 also on Ubuntu Lucid
6757
6758 update from autoscan
6759
6760 kronosnetd: add man pages
6761
6762 init script: shutup lintian warning
6763
6764 2012-11-26 Fabio M. Di Nitto <[email protected]>
6765
6766 libknet: fix cbuffers cleaning after normal shutdown/link down/crash
6767
6768 2012-11-25 Fabio M. Di Nitto <[email protected]>
6769
6770 Update TODO list
6771
6772 libknet: clean up circular buffer counters when remote goes dead
6773
6774 libknet: record status from incoming HOST_INFO link up-down messages
6775
6776 2012-11-23 Fabio M. Di Nitto <[email protected]>
6777
6778 libknet: notify remote host of link status changes
6779 receiver still can't use those data for anything
6780
6781 libknet: fix locking within logging... for real
6782 what the hell was i thinking with 1ab701ce16
6783
6784 libknet: fix locking within logging
6785 we cannot grab a lock during logging because something else
6786 might be already using a lock.
6787
6788 libknet: add support for HOST_INFO unicast dispatch
6789
6790 libknet: fix data write to fd
6791
6792 Update TODO
6793
6794 2012-11-17 Fabio M. Di Nitto <[email protected]>
6795
6796 libknet: remove hardcoded magic numbers
6797
6798 libknet: basic prototyping of HOST_INFO messages
6799
6800 libknet: make all data packed
6801 gcc doesn't propagate packed attribute down to a union
6802
6803 libknet: fix unlocking in case of errors in tap_to_links
6804
6805 2012-11-16 Fabio M. Di Nitto <[email protected]>
6806
6807 Update TODO list
6808
6809 libknet: add sync host to host messagging
6810
6811 kronosnetd: make baseport endian clean
6812
6813 kronosnet: merge concept of RINGID and baseport
6814 this avoids extra confusion with ringid
6815
6816 2012-11-14 Fabio M. Di Nitto <[email protected]>
6817
6818 Update TODO list
6819
6820 libtap: store ip in the same order as they entered
6821
6822 kronosnetd: save interfaces in the same order as input
6823
6824 kronosnetd: allow vty timeout to be loaded from config
6825
6826 Update TODO list
6827
6828 2012-11-13 Fabio M. Di Nitto <[email protected]>
6829
6830 Update TODO list
6831
6832 libtap: attempt to fix ioctl usage
6833
6834 kronosnetd: fix parsing of ip_list return from libtap
6835
6836 2012-11-12 Fabio M. Di Nitto <[email protected]>
6837
6838 Update TODO list (bug list)
6839
6840 kronosnetd: increase default vty timeout and save it to config
6841 loading still has no effect
6842
6843 kronosnetd: add vty timeout command
6844 the value is not persistent or saved in config
6845
6846 kronosnetd: add vty config node
6847
6848 Update TODO list
6849 cannot be done without major config rewrite right now
6850
6851 kronosnetd: fix spacing in login prompt
6852
6853 Update TODO list
6854
6855 2012-11-10 Fabio M. Di Nitto <[email protected]>
6856
6857 Update TODO list
6858
6859 2012-11-09 Fabio M. Di Nitto <[email protected]>
6860
6861 libknet: fix link_enable logic
6862
6863 2012-11-08 Fabio M. Di Nitto <[email protected]>
6864
6865 kronosnetd: perform clean shutdown on daemon exit
6866
6867 libknet: fix listener remove return code
6868
6869 libknet: misc cleanup
6870 remove fds from epolls on exit
6871
6872 don't mingle with listeners if there is no link status change
6873
6874 libknet: fix unlocking in log message
6875
6876 libknet: make basic infrastructure ready for host to host comm
6877
6878 libknet: rename pipefd to dstpipefd
6879
6880 Update TODO list
6881
6882 2012-11-06 Fabio M. Di Nitto <[email protected]>
6883
6884 crypto: move crypto buffer allocation back to handle
6885 nsscrypto doesn't need to know how many buffers to allocate/free
6886
6887 libknet: drop magic number in header
6888 save 4 bytes per packet and avoid plain text attack
6889
6890 Update TODO
6891
6892 libknet: make error message more useful
6893
6894 kronosnetd: make config safer to copy paste
6895
6896 kronosnetd: add ringid to interface creation to spicy up mac generation
6897
6898 fix multihoming support
6899 major change across the board:
6900
6901 - drop concept of listeners completely, they are now handled internally
6902 - link needs src/dst
6903 - fix different segfaults
6904 - cleanup API
6905 - port kronosnetd UI to new format
6906
6907 2012-11-02 Fabio M. Di Nitto <[email protected]>
6908
6909 libknet: first big hammer on logging
6910 still needs fine tuning as we go to understand exactly what to log
6911 in debugging mode
6912
6913 kronosnet: use a different macro for debug from libknet
6914
6915 Update TODO list
6916
6917 kronosnetd: add logging from libknet to logfile/stdout
6918 no logging to vty yet
6919
6920 2012-11-01 Fabio M. Di Nitto <[email protected]>
6921
6922 libknet: add basic logging infrastructure to the library
6923
6924 2012-10-28 Fabio M. Di Nitto <[email protected]>
6925
6926 Update TODO list
6927
6928 Rename config fd to make space for logging from knet
6929
6930 2012-10-27 Fabio M. Di Nitto <[email protected]>
6931
6932 Update TODO list for libknet logging infrastructure
6933
6934 Update TODO list
6935 Using multiple dynamic links requires:
6936
6937 link dynamicXXXXXXXX
6938
6939 where XXXXXXXX can be any identifier
6940
6941 Use bindv6only by default and make code simpler and safer
6942
6943 2012-10-26 Fabio M. Di Nitto <[email protected]>
6944
6945 Update TODO list
6946
6947 libknet: fix bindv6only socket handling for dynamic connections
6948
6949 2012-10-25 Fabio M. Di Nitto <[email protected]>
6950
6951 libknet/kronosnetd: add support for bindv6only
6952
6953 kronosnetd: fix error message and exit path
6954
6955 kronosnetd: add support for dual stack CLI (bindv6only yes/no)
6956
6957 kronosnetd: increase CLI default timeout
6958
6959 libknet: fix timespec handling
6960 guess what.. it ain't portable...
6961
6962 Update TODO
6963
6964 Update TODO list
6965
6966 2012-10-24 Fabio M. Di Nitto <[email protected]>
6967
6968 kronosnetd: don't buffer overflow
6969
6970 update .gitignore
6971
6972 build: link as needed
6973
6974 general: add init script and service file
6975
6976 libknet: ship .pc file and sort linking
6977
6978 libknet: link against -lrt for clock_gettime symbol
6979
6980 build: fix build for old version of nss/nspr
6981
6982 build: add missing file from tarball
6983
6984 Add preliminary support for handling dynamic ip connections
6985 max on dynamic link per host for now, we need to fix link_id handling
6986 first
6987
6988 2012-10-23 Fabio M. Di Nitto <[email protected]>
6989
6990 kronosnetd: provide basic status functionality
6991
6992 ping_test: fix unit measure to be bearable
6993
6994 ping_test: fix usage example
6995
6996 kronosnet: add config support for link timers
6997
6998 kronosnetd: add link priority configuration
6999
7000 libknet: add link priority api call
7001
7002 2012-10-15 Fabio M. Di Nitto <[email protected]>
7003
7004 libtap: workaround bug in SETMTU ioctl
7005 apparently SETMTU does parse flags as well, make it clean
7006
7007 kronosnetd: allow configuration of switching-policy engine
7008 passive/active/round-robin
7009
7010 update TODO list
7011
7012 2012-10-14 Fabio M. Di Nitto <[email protected]>
7013
7014 libknet: worst commit ever....
7015 fix a series of segfaults in crypto init code
7016 add api call to set switching policy (that was the original idea
7017 of this change, till things started exploding on me)
7018
7019 libknet: readd dst_cache_update function for internal use
7020 this is required to trigger a recalculation when we change switching
7021 policy at runtime from active/passive to active/active or round-robin
7022
7023 libknet: don't use black magic..
7024
7025 libknet: simplify link updown notification code
7026
7027 libknet: remove that horrid dst_cache update and mask it around link_enabled
7028 makes the API a bit more readable.
7029
7030 2012-10-13 Fabio M. Di Nitto <[email protected]>
7031
7032 libknet: fix 3des key size
7033
7034 libknet: add support for 3des nss crypto
7035
7036 libknet: add support for nss aes128 and aes192
7037
7038 Update TODO list
7039
7040 2012-10-12 Fabio M. Di Nitto <[email protected]>
7041
7042 libknet: move crypto to fully modular system
7043
7044 libknet: start implementing generic crypto vs nsscrypto framework
7045 it will save us some time later if we decide to plug something else
7046
7047 crypto: rework the API and make it a separate init call
7048 this also saves us a few memcpy around when not using crypto
7049
7050 2012-10-07 Fabio M. Di Nitto <[email protected]>
7051
7052 libknet: make host add a O(1) operation
7053
7054 2012-10-06 Fabio M. Di Nitto <[email protected]>
7055
7056 libknet: keep host ordered by insert
7057
7058 kronosnetd: drop trailing space when saving config
7059
7060 libknet: fix memmove offset and make it simpler
7061
7062 libknet: first cut for RR support per packet
7063
7064 2012-10-05 Fabio M. Di Nitto <[email protected]>
7065
7066 libknet: add handling of link priority for active passive
7067 also update the TODO list
7068
7069 (this commit has not been tested)
7070
7071 update TODO list based on rekey brainstorm
7072
7073 2012-10-03 Fabio M. Di Nitto <[email protected]>
7074
7075 kronosnetd: use fixed mac addresses to reduce arp work
7076
7077 libknet: export function to allow dst cache update on link removal
7078
7079 update TODO list
7080
7081 link policy handler: add first cut of code to have a switching policy engine
7082 for now only ACTIVE and PASSIVE are supported. RR needs dst_cache_update/lookup.
7083
7084 2012-10-02 Fabio M. Di Nitto <[email protected]>
7085
7086 libknet: rename var to be more explicit
7087
7088 libknet: fix endianess and memory alignment issues
7089 also check for some return codes
7090
7091 build: drop warning check for enum-compare
7092 in some older versions of gcc, the flag is supported only for C++,
7093 our flag test succeed, but we fail to build.
7094
7095 2012-10-01 Fabio M. Di Nitto <[email protected]>
7096
7097 libknet: add unicast switching support
7098
7099 2012-09-30 Fabio M. Di Nitto <[email protected]>
7100
7101 update TODO with some ideas for improvements
7102
7103 2012-09-29 Fabio M. Di Nitto <[email protected]>
7104
7105 build: LIBQB -> libqb
7106
7107 libknet: rename fields in knet_link to be more meaningful
7108
7109 libknet: rename i,j,k to some proper vars and make code a bit more readable
7110
7111 libknet: make ready for unicast handling and fix unlocking if crypto fails
7112
7113 update TODO list
7114
7115 2012-09-28 Fabio M. Di Nitto <[email protected]>
7116
7117 libknet: packet inspector first cut
7118 add packet destination filter API and provide an example
7119 for ethernet packet inspector.
7120
7121 also start to plug the inspector into the transmitting end of
7122 libknet
7123
7124 libknet: implement basic ring buffer for bcast packets
7125 fix KNET_FRAME_SIZE handling around
7126
7127 don't try to send messages to links that are not ready
7128
7129 gather host info for all incoming pkcts (data/ping) since there
7130 is no point to go crazy figuring it out
7131
7132 add basic ring buffer functions based on playground implementation
7133 (still missing reset of the buffer)
7134
7135 kronosnetd: fix data initialization
7136 we can't memset knet_link because khost_add does some initialization
7137 there that we overwrite otherwise.
7138
7139 drop incorrect clean up path. knet_links are not malloc'ed but statically
7140 allocated.
7141
7142 2012-09-27 Fabio M. Di Nitto <[email protected]>
7143
7144 kronosnetd: remove spurios debugging messages
7145
7146 libknet: add seq_num to sender host
7147 bcast for packets that needs to be sent to all nodes
7148 ucast for direct packets
7149
7150 the difference is necessary to avoid unncessary re-encryption
7151 of packets. This will make more sense once packet inspector
7152 is in place, but basically seq_num are per host and not all
7153 packets will go everywhere, creating undesirable holes in the
7154 packet_ring_buffer, that in turns, in a non-encrypted network
7155 could lead to some annoyance.
7156
7157 libknet: change packet definitions to include seq_num for data packets
7158 this is required for pckt deduplication
7159
7160 kronosnetd: node_id have to match knet size for packet inspector to work
7161
7162 2012-09-26 Fabio M. Di Nitto <[email protected]>
7163
7164 libknet: add notes on packet inspector
7165 and where it should hook for notifications
7166
7167 build: fix out of tree build
7168
7169 clean up some hardcoded values and export them via header correctly
7170
7171 kronosnetd: add support for encryption keys
7172
7173 kronosnetd: add knet-keygen
7174
7175 2012-09-25 Fabio M. Di Nitto <[email protected]>
7176
7177 fix a few int vs ssize_t bits around
7178
7179 vty: add partial crypto/hash support
7180
7181 libknet: fix nss types and add crypto/hash support to ping_test
7182
7183 2012-09-24 Fabio M. Di Nitto <[email protected]>
7184
7185 libknet: plug nss for good into threads
7186
7187 libknet: crypto simplify even more the code path and data storage
7188
7189 libknet: add macro for debugging of nsscrypto and simplify more code around
7190
7191 libknet: more crypto cleanup and allocate crypto buffers when necessary
7192
7193 libknet: add a few comments
7194
7195 libknet: more nss crypto work
7196 make crypto_init/fini work.
7197
7198 make libknet shared (necessary for correct linking with nss).
7199
7200 2012-09-21 Fabio M. Di Nitto <[email protected]>
7201
7202 libknet: start plugging in crypto support
7203
7204 2012-09-20 Fabio M. Di Nitto <[email protected]>
7205
7206 libknet: make tap to links and links to tap full duplex
7207 this cuts down by 40% latency on ping test
7208
7209 ping_test: clarify unit measure...
7210
7211 2012-09-17 Fabio M. Di Nitto <[email protected]>
7212
7213 libknet: add totemcrypto.c|h from corosync (pristine)
7214 permission to copy and change licence to GPLv2+ has been granted
7215 in writing by current authors of the code.
7216
7217 let's be more realistic on what we can do here
7218 libvty is not going to happen. Too complex and it's not the goal
7219 for this project. For now, let's move it back into the daemon.
7220
7221 2012-09-16 Fabio M. Di Nitto <[email protected]>
7222
7223 logging: add logging support for syslog/stderr/logfile
7224 hook into libqb for logging facilities.
7225
7226 build: cleanup escaping in exporting vars
7227
7228 2012-09-15 Fabio M. Di Nitto <[email protected]>
7229
7230 drop generic utils.(c|h) and start implementing proper logging
7231
7232 logging: add concept of logfile and fix fprintf output
7233
7234 build: depends on nss and libqb
7235
7236 build: move daemon in its own dir
7237
7238 libtap: remove external header (useless really)
7239
7240 2012-09-14 Fabio M. Di Nitto <[email protected]>
7241
7242 libtap: mark library 1.0.0
7243 drop fPIC (managed correctly by libtool)
7244
7245 and make it 1.0.0 (number vs version)
7246
7247 build: fix some minor bits around
7248
7249 2011-03-23 Fabio M. Di Nitto <[email protected]>
7250
7251 main: print errors on stderr before daemonizing
7252
7253 testsuite: decouple from utils.c
7254
7255 libtap: decouple test suite from utils.h
7256
7257 cfg: drop util.h
7258
7259 vty: add missing include and fix build
7260
7261 decouple cloexec calls from utils
7262
7263 test: rename test to match what it really does
7264
7265 move netutils in libvty/ since vty is the only user
7266
7267 Revert "libknet: add missing include"
7268 This reverts commit c5edcc14952da3e9914f1ff8df98413ab4728c2c.
7269
7270 logging: fix utils_syslog handling
7271 set default to syslog to 0 again and enable it only in kronosnetd,
7272 otherwise all test suite would log to syslog too.
7273
7274 utils: remove timespec_diff that's relevant only to libknet
7275
7276 libknet: add missing include
7277
7278 utils: drop concept of STATIC vs static
7279
7280 vty: make more functions static
7281
7282 2011-03-22 Fabio M. Di Nitto <[email protected]>
7283
7284 logging: enable syslog by default
7285 also inject log_debug into syslog if configured
7286
7287 main: move defaults into its own function
7288
7289 Move vty into its own dir for future "librarification"
7290
7291 Move libtap doc to the next release
7292
7293 fix file permission
7294
7295 2011-03-09 Fabio M. Di Nitto <[email protected]>
7296
7297 Add some extra warnings at build time
7298
7299 2011-01-24 Fabio M. Di Nitto <[email protected]>
7300
7301 Update TODO: API renames are done
7302
7303 build: use libtap vs libtap_test
7304
7305 2011-01-19 Fabio M. Di Nitto <[email protected]>
7306
7307 libtap: embed test suite in the library
7308
7309 libtap: move tests in libtap/
7310
7311 update TODO for libtap
7312
7313 2011-01-18 Fabio M. Di Nitto <[email protected]>
7314
7315 libtap: fix linking and set exported symbols
7316
7317 2011-01-05 Fabio M. Di Nitto <[email protected]>
7318
7319 main: remove oom_adj
7320 kernel ABI will change, not really worth the troubles for now
7321
7322 2011-01-03 Fabio M. Di Nitto <[email protected]>
7323
7324 libtap: align API calls to correct names
7325
7326 libtap: improve error reporting for up/down operations
7327
7328 2010-12-23 Fabio M. Di Nitto <[email protected]>
7329
7330 libtap: propagate errors from CLI calls to upper layers
7331
7332 2010-12-22 Federico Simoncelli <[email protected]>
7333
7334 libknet: use naming convention for static functions
7335
7336 misc: more file renaming
7337
7338 libknet: split code into multiple files
7339
7340 libknet: update naming convention, rename header files
7341
7342 2010-12-22 Fabio M. Di Nitto <[email protected]>
7343
7344 libtap: first pass to remove calls to log_*
7345 use internal error buffer instead across the non public API.
7346
7347 pass2 will export it to users.
7348
7349 2010-12-21 Federico Simoncelli <[email protected]>
7350
7351 knet: move knet code to its own library
7352
7353 2010-12-20 Federico Simoncelli <[email protected]>
7354
7355 ring: fix node_id endianness, use ping/pong packet mask
7356
7357 ring: use hosts index to send pings and receive pongs
7358
7359 2010-12-20 Fabio M. Di Nitto <[email protected]>
7360
7361 vty_cli_cmds: bind pre-up/up/down/post-down
7362
7363 libtap: add updown tests and fix off by one check
7364
7365 libtap: add exec_updown wrapper for ifup/down operations
7366
7367 Update TODO
7368
7369 libtap: several improvments across the board
7370 start to add pre-up/up/down/post-down support
7371
7372 cleanup internal API function names
7373
7374 fix memory leak on close
7375
7376 add forward declearation for all static functions
7377
7378 add an extra test to be executed under valgrind
7379
7380 test: add track-fds to valgrind check
7381
7382 2010-12-18 Fabio M. Di Nitto <[email protected]>
7383
7384 Update TODO list
7385
7386 libtap: complete ip_list management and last integration with CLI
7387
7388 libtap: make libtap more thread safe
7389
7390 2010-12-17 Fabio M. Di Nitto <[email protected]>
7391
7392 build: clean some more
7393
7394 vty_cli_cmds: prepare for new knet_handle_new API
7395
7396 Merge branch 'master' of github.com:fabbione/kronosnet
7397
7398 Update TODO
7399
7400 libtap: move ip management into libtap
7401
7402 2010-12-17 Federico Simoncelli <[email protected]>
7403
7404 utils: fix timespec diff on x86 platforms
7405
7406 2010-12-17 Fabio M. Di Nitto <[email protected]>
7407
7408 libtap: rename internal struct from knet_tap to tap_
7409
7410 libtap: fold interface name into libtap and avoid dup use
7411
7412 libtap: add knet_tap_find, switch to ifname macro
7413 also revert a0cbd79cefdbfed1d41f8842d6786f5b55484aa4 to fix unnamed tap device creation
7414
7415 add more TODO entries
7416
7417 vty_cli_cmds/cfg: drop local knowledge of MTU and delegate to libtap
7418
7419 libtap: add reset to default functions for mtu/mac
7420
7421 TODO list update for libtap
7422
7423 libtap: do proper initialization and close of the library
7424 store default mtu and mac at init. This operation is required also
7425 for direct access to set_mtu and set_mac.
7426
7427 libtap: cleanup tap socket on last close
7428
7429 2010-12-16 Federico Simoncelli <[email protected]>
7430
7431 Merge branch 'master' of github.com:fabbione/kronosnet
7432
7433 todo: update todo list
7434
7435 2010-12-16 Fabio M. Di Nitto <[email protected]>
7436
7437 libtap: start introducing internal list/config tracking
7438
7439 libtap: remove unrequired strcpy
7440
7441 vty: remove debugging entry
7442
7443 libtap: switch to use opaque struct
7444
7445 build: I HATE MAKEFILE EXPANSION.... fix libtap build kthxbye
7446
7447 build: link test tools with shared libtap
7448 do not link kronosnetd yet.
7449
7450 build: need two versions for each lib
7451 test and final.
7452
7453 build: include libtap subdir in the build
7454
7455 libtap: rename knet_test to tap_test
7456
7457 libtap: add shared version of the library
7458 NOT TO BE USED YET
7459
7460 libtap: align API to name convention
7461
7462 First cut to isolate tap handling in its own library
7463
7464 vty_cli_cmds: port VTY to new API
7465 needs lot more work to get clean
7466
7467 vty_cli_cmds: port "interface" to new ring API
7468
7469 2010-12-10 Federico Simoncelli <[email protected]>
7470
7471 knet: new api design
7472
7473 2010-12-03 Federico Simoncelli <[email protected]>
7474
7475 ring: improve looping over links
7476
7477 2010-12-02 Federico Simoncelli <[email protected]>
7478
7479 ring: allocate links with host
7480
7481 2010-11-30 Federico Simoncelli <[email protected]>
7482
7483 ring: fix major bugs in knet frame length
7484
7485 ring: add host index for fast lookup
7486
7487 tests: fix long integer syntax
7488
7489 2010-11-29 Federico Simoncelli <[email protected]>
7490
7491 ring: improve knet_frame structure
7492
7493 utils: fix timespec difference
7494
7495 2010-11-27 Fabio M. Di Nitto <[email protected]>
7496
7497 knet: make internal functions static
7498
7499 TODO: remove redundant entry
7500
7501 TODO: add more items for shared libs
7502
7503 build: add .bz2 tarball to release
7504
7505 build: fix uninstall section
7506
7507 move kronosnet.conf to /etc/kronosnet/kronosnet.conf
7508
7509 add more todo items
7510
7511 2010-11-26 Fabio M. Di Nitto <[email protected]>
7512
7513 update TODO
7514
7515 Add TODO list and release planning
7516
7517 docs: add architecture doc
7518
7519 Add README file
7520
7521 2010-11-25 Fabio M. Di Nitto <[email protected]>
7522
7523 vty/cfg: remove two unrequired wrappers
7524
7525 vty: remove bridge struct
7526
7527 vty: destroy knet_h on error and "no interface"
7528
7529 2010-11-24 Federico Simoncelli <[email protected]>
7530
7531 git: add benchmarks to ignored files
7532
7533 benchmark: add lookup benchmark
7534
7535 utils: improve knet_ts_diff speed
7536
7537 Merge branch 'master' of github.com:fabbione/kronosnet
7538
7539 knet: add sigint handler
7540
7541 2010-11-24 Fabio M. Di Nitto <[email protected]>
7542
7543 cfg: remove unused struct
7544
7545 2010-11-23 Federico Simoncelli <[email protected]>
7546
7547 knet_handle: stop heartbt thread on knet_handle_free
7548
7549 knet_handle: check pthread exit status
7550
7551 knet_handle: add knet_handle_free
7552
7553 knet_handle: do not generate new sockets for knet_handle
7554
7555 2010-11-23 Fabio M. Di Nitto <[email protected]>
7556
7557 vty: add link menu and clean up links on object destroy path
7558
7559 vty: save bridge status in config file
7560
7561 ring: avoid lots of unnecessary casting, fix build on sparc and fix link loop
7562 also avoid a segfault when scanning for multiple links
7563
7564 2010-11-22 Fabio M. Di Nitto <[email protected]>
7565
7566 vty: add start/stop commands to interface
7567 configure
7568 interface kronosnet0 2
7569 baseport 50000
7570 mtu 9000
7571 ip 192.168.7.2 24
7572 peer node1 1
7573 link 192.168.5.1
7574 link 192.168.4.1
7575 exit
7576 start
7577 exit
7578 exit
7579 exit
7580
7581 now starts the bridge and traffic is flowing!
7582
7583 ring: fix address comparison to determine if link is up/down
7584 also initialize addrlen properly and fix make dist
7585
7586 build: update configure.ac after autoscan
7587
7588 build: fix make dist
7589
7590 vty: remove concept of auto-listeners/manual listeners
7591 switch to a much simpler baseport and port range concept.
7592
7593 for now, only baseport is implementend and needs extra checks implementation.
7594 portrange will come soon enough.
7595
7596 ring: fix ping calculation for good
7597 finally I have been able to reproduce the problem and it's related to
7598 the "unknown" source in MONOTONIC clock. It can rollover making the signed
7599 comparison useless. Force unsigned long long for the calculation that
7600 fixes the issue.
7601
7602 Switching from:
7603 Notice: timediff: -345483470 interval: 1000000
7604
7605 To:
7606 Notice: timediff: 18446744073387515858 interval: 1000000
7607
7608 ring: use long long instead of suseconds_t
7609 it appears (but not totally sure) that suseconds_t is not easily portable.
7610 On sparc it was showing, at random, some different behaviour.
7611
7612 2010-11-19 Fabio M. Di Nitto <[email protected]>
7613
7614 Merge branch 'master' of github.com:fabbione/kronosnet
7615
7616 vty: fix no link list handling
7617
7618 2010-11-19 Federico Simoncelli <[email protected]>
7619
7620 Merge branch 'master' of github.com:fabbione/kronosnet
7621
7622 vty: add 'O' to pre-escape sequences
7623
7624 2010-11-19 Fabio M. Di Nitto <[email protected]>
7625
7626 vty: initialize timeout values
7627
7628 vty: add basic link support
7629
7630 build: fix sparc build
7631
7632 2010-11-19 Federico Simoncelli <[email protected]>
7633
7634 tests: fix printf format for latency
7635
7636 tests: display errno for functions supposed to fail
7637
7638 knet: fix maximum socket receive buffer size (unit is bytes)
7639
7640 Merge branch 'master' of github.com:fabbione/kronosnet
7641
7642 knet: add listener test cases
7643
7644 2010-11-19 Fabio M. Di Nitto <[email protected]>
7645
7646 vty: get autolisteners to work somehow...
7647 this is getting ugly and needs cleanup
7648
7649 2010-11-19 Federico Simoncelli <[email protected]>
7650
7651 knet: add knet_link_timeout and knet_listener_remove
7652
7653 2010-11-18 Fabio M. Di Nitto <[email protected]>
7654
7655 vty: add concept of peers/hosts
7656 sort ring.h header to group functions based on target
7657
7658 vty/config: collect handles in relative config structs
7659
7660 vty: add basic options for auto-listeners support
7661
7662 2010-11-16 Federico Simoncelli <[email protected]>
7663
7664 knet: use -1 timeout in epoll to wait indefinitely
7665
7666 knet: add acquire/release functions for the knet_listener list
7667
7668 2010-11-15 Fabio M. Di Nitto <[email protected]>
7669
7670 vty: add write/read config support
7671
7672 vty: add config show command
7673 also fix assignment again...
7674
7675 vty_cli_cmds: Remove unrequired assignments
7676
7677 vty_cli_cmds: make help text more generic
7678
7679 2010-11-15 Federico Simoncelli <[email protected]>
7680
7681 test: add latency/timeout testing comment
7682
7683 heartbeat: avoid flaky link when latency is higher than timeout
7684
7685 knet: compute latency average
7686
7687 2010-11-12 Federico Simoncelli <[email protected]>
7688
7689 Merge branch 'master' of github.com:fabbione/kronosnet
7690
7691 knet: add link latency statistics (last value)
7692
7693 2010-11-12 Fabio M. Di Nitto <[email protected]>
7694
7695 vty: add support to configure ip addresses on knet ifaces
7696
7697 2010-11-12 Federico Simoncelli <[email protected]>
7698
7699 Merge branch 'master' of github.com:fabbione/kronosnet
7700
7701 knet: add timestamp into ping packet
7702
7703 2010-11-12 Fabio M. Di Nitto <[email protected]>
7704
7705 cfg: move ether config in its own struct
7706
7707 cfg: drop unrequired locking
7708 all config changes are already locked by the vty
7709
7710 vty_cli_cmds: fix create interface invokation
7711
7712 vty: add mtu config support in interface menu
7713
7714 vty_cli_cmds: add a few comments
7715
7716 vty_cli_cmds: more sorting
7717
7718 cfg: add concept of default mtu and configurable mtu
7719
7720 vty_cli_cmds: sort code around
7721 move all helper functions at the top and all menu/exec cmd at the bottom
7722
7723 2010-11-11 Fabio M. Di Nitto <[email protected]>
7724
7725 vty_cli_cmds: set mac address at iface creation time
7726 also fix an off-by-one check
7727
7728 2010-11-11 Federico Simoncelli <[email protected]>
7729
7730 Merge branch 'master' of github.com:fabbione/kronosnet
7731
7732 knet: split control thread and heartbeat thread
7733
7734 2010-11-11 Fabio M. Di Nitto <[email protected]>
7735
7736 vty/cfg: allow multiple parameters per command
7737
7738 2010-11-10 Fabio M. Di Nitto <[email protected]>
7739
7740 cfg: fix interface name detection
7741 this one needs to be precise
7742
7743 vty: start adding bindings to core
7744 this first snap allows to create/remove knet interfaces
7745
7746 2010-11-10 Federico Simoncelli <[email protected]>
7747
7748 knet: move debug into ping_test
7749
7750 knet: add ping_interval and pong_timeout configuration per link
7751
7752 knet: add trivial heartbeat feature
7753
7754 cosmetic: fix typo
7755
7756 2010-11-10 Fabio M. Di Nitto <[email protected]>
7757
7758 whops.. remove debugging/testing code
7759
7760 vty_cli: fix tab completion
7761
7762 vty_cli: add parameter concept and tab completion
7763
7764 2010-11-09 Fabio M. Di Nitto <[email protected]>
7765
7766 vty: add basic support for partial command execution and help
7767
7768 vty_auth: fix off-by-one in group checking
7769 a user with the same name of a group is automatically part of that group
7770
7771 2010-11-04 Fabio M. Di Nitto <[email protected]>
7772
7773 vty_cli: add ctrl+z handler
7774
7775 vty: clear config lock on exit
7776
7777 vty_cli_cmds: simply revert action infrastructure
7778
7779 knet: check if created tap device has same name of requested
7780
7781 vty_cli_cmds: add "no" infrastructure
7782
7783 vty_cli_cmds: add interface configuration node layer
7784 and fix a segfault in the help function
7785
7786 vty_cli_cmds: cosmetic output for who command
7787
7788 vty: add configuration handler
7789
7790 2010-11-03 Fabio M. Di Nitto <[email protected]>
7791
7792 vty_cli: allow to logout via ctrl+d and rename node names to be more useful
7793
7794 vty_cli_cmds: add logout/exit commands
7795
7796 2010-11-02 Fabio M. Di Nitto <[email protected]>
7797
7798 vty_cli_cmds: group code together and make command search non-destructive
7799
7800 vty_cli: end of history next should give a clean line
7801
7802 2010-11-01 Fabio M. Di Nitto <[email protected]>
7803
7804 vty_cli_cmds: add help command
7805
7806 vty_cli_cmds: add default help handler
7807
7808 vty: handle static allocation consistently
7809
7810 vty: laydown basic code to execute commands
7811
7812 2010-10-31 Fabio M. Di Nitto <[email protected]>
7813
7814 vty_utils: fix segfault when free'ing history
7815
7816 2010-10-29 Fabio M. Di Nitto <[email protected]>
7817
7818 vty: inform CLI users that the server is going away
7819
7820 vty: add note on exit path
7821
7822 vty: fix pam memory leakage in case of failure/timeout
7823
7824 vty: don't leak memory on pthread_cancel
7825
7826 2010-10-28 Fabio M. Di Nitto <[email protected]>
7827
7828 cosmetic: add extra return to login info
7829
7830 vty_cli: add basic concept of prompt
7831 fix line redraw when parsing history
7832
7833 vty_cli: fix telnet backchar on x86
7834
7835 vty_cli: add history handler
7836
7837 2010-10-28 Federico Simoncelli <[email protected]>
7838
7839 vty: cosmetics changes
7840
7841 2010-10-28 Fabio M. Di Nitto <[email protected]>
7842
7843 vty_cli: add support for insert mode
7844
7845 vty_cli: bind extra keys
7846
7847 vty_cli: add more home/end escape codes
7848
7849 2010-10-28 Federico Simoncelli <[email protected]>
7850
7851 vty: add extra escape sequences (home, ins, del, end, pg-up/down)
7852
7853 2010-10-28 Fabio M. Di Nitto <[email protected]>
7854
7855 vty_cli: add ctrl+t handler
7856
7857 vty: fix off by one vty allocation
7858
7859 vty_cli: add forward word handler
7860
7861 vty_cli: add kill forward word handler
7862
7863 vty_cli: add ctrl+w handler
7864
7865 vty_cli: fix refresh when adding in the middle of the buffer
7866
7867 vty_cli: add backward word handler
7868
7869 vty_cli: make add_to_buf echo to terminal
7870 temporary disable knet_vty_rewrite_line as it's not used right now
7871
7872 this change makes the cli a lot more snappy :)
7873
7874 fix an off-by-one buffer init
7875
7876 2010-10-27 Fabio M. Di Nitto <[email protected]>
7877
7878 vty_cli: add ctrl+u handler
7879
7880 vty_cli: add ctrl+h / del handlers
7881
7882 vty_cli: add ctrl+e handler
7883
7884 vty_cli: add ctrl+k handler
7885
7886 vty_cli: add ctrl+c handler
7887 temporary change enter handler to differentiate in testing
7888
7889 vty_cli: add ctrl+a handler
7890
7891 vty_cli: add backward/forward char handlers
7892
7893 vty_cli: remove debugging
7894
7895 vty_cli: fix parsing of escape chars
7896
7897 2010-10-27 Federico Simoncelli <[email protected]>
7898
7899 Merge branch 'master' of github.com:fabbione/kronosnet
7900
7901 netutils: check name resolution in strtoaddr
7902
7903 2010-10-27 Fabio M. Di Nitto <[email protected]>
7904
7905 vty_cli: add support for escape chars
7906
7907 vty_cli: lay down scheleton for parsing CLI input
7908
7909 vty_cli: drop idle update in favour of the generic one
7910
7911 vty: make timeout global for all CLI's
7912 move the checks into the main vty thread that's idle most of the time
7913 anyway.
7914
7915 vty cli: add basic infrastructure for CLI
7916
7917 vty_utils: some iacs send a response
7918 MUST make that dynamic... really
7919
7920 2010-10-27 Federico Simoncelli <[email protected]>
7921
7922 Merge branch 'master' of github.com:fabbione/kronosnet
7923
7924 netutils: add tests for strtoaddr and addrtostr
7925
7926 2010-10-27 Fabio M. Di Nitto <[email protected]>
7927
7928 vty: configure telnet session and add better error reporting
7929 all manual steps at init requires manual checking, the vty loop will do that
7930 in an a more elegant fashion :)
7931
7932 vty utils: return error if user requested to write more than VTY_MAX_BUFFER_SIZE
7933 Probably at somepoint we should allocate this buffer dynamically
7934
7935 vty_utils: add connection barrier if socket is closed underneath our feet
7936
7937 cosmetic: change welcome message
7938
7939 vty: remove debugging setting
7940
7941 build: check for pthreads at configure.ac
7942
7943 2010-10-26 Federico Simoncelli <[email protected]>
7944
7945 knet: add knet_listener structure (replaces knet_handle_bind)
7946
7947 2010-10-25 Fabio M. Di Nitto <[email protected]>
7948
7949 vty: cleanup error/info reporting code
7950 OMG.. did I really write that crap?
7951
7952 Merge branch 'master' of github.com:fabbione/kronosnet
7953
7954 build: scan for more headers
7955
7956 2010-10-25 Federico Simoncelli <[email protected]>
7957
7958 knet: do not resolve host names
7959
7960 2010-10-25 Fabio M. Di Nitto <[email protected]>
7961
7962 vty: free memory
7963
7964 2010-10-22 Fabio M. Di Nitto <[email protected]>
7965
7966 vty: more fixes on the auth sequence and better error reports
7967 Will I ever get this one right?
7968
7969 vty_auth: make pam_authenticate errors fatal
7970
7971 vty: fix PAM init sequence and error report
7972 deal with the lack of /etc/pam.d/kronosnet without crashing
7973
7974 vty: log all connections, including source ip and port
7975
7976 2010-10-22 Federico Simoncelli <[email protected]>
7977
7978 Merge branch 'master' of github.com:fabbione/kronosnet
7979
7980 knet: check malloc return code in addrtostr
7981
7982 2010-10-22 Fabio M. Di Nitto <[email protected]>
7983
7984 vty: init_listener use strtoaddr for the winner!
7985
7986 2010-10-22 Federico Simoncelli <[email protected]>
7987
7988 Merge branch 'master' of github.com:fabbione/kronosnet
7989
7990 knet: add addrtostr
7991
7992 2010-10-22 Fabio M. Di Nitto <[email protected]>
7993
7994 build: allow default admin group to be configurable and clean up
7995
7996 2010-10-22 Federico Simoncelli <[email protected]>
7997
7998 Merge branch 'master' of github.com:fabbione/kronosnet
7999
8000 knet: improve strtoaddr using getaddrinfo
8001
8002 2010-10-22 Fabio M. Di Nitto <[email protected]>
8003
8004 Drop log_debug entries for now
8005
8006 build: update configure.ac
8007
8008 vty: perform group check before allowing users to enable
8009
8010 vty: associate username with vty and drastically improve error detection/report
8011 also make knet_vty_auth_user reusable
8012
8013 2010-10-21 Fabio M. Di Nitto <[email protected]>
8014
8015 vty: tell the user on which vty he/she is logging in
8016
8017 vty: use struct knet_vty all around
8018
8019 vty: address several issues with vty allocation
8020 use knet_vty structure as thread arg to pass more data around
8021
8022 initialize knet_vtys memory
8023
8024 make sure to reuse knet_vtys slots
8025
8026 pass data to vty thread
8027
8028 clear memory on entry/exit
8029
8030 vty: fix set_max_connections sanity checks
8031
8032 2010-10-21 Federico Simoncelli <[email protected]>
8033
8034 Merge branch 'master' of github.com:fabbione/kronosnet
8035
8036 knet: add netutils stub
8037
8038 2010-10-20 Fabio M. Di Nitto <[email protected]>
8039
8040 vty: add CLI threads and PAM authentication
8041 NOTE: /etc/pam.d/kronosnet is now required to telnet to CLI
8042
8043 2010-10-20 Federico Simoncelli <[email protected]>
8044
8045 knet: add acquire/release functions to lock the host structure list
8046
8047 2010-10-18 Federico Simoncelli <[email protected]>
8048
8049 knet: use KNET_MAX_EVENTS in epoll_create
8050
8051 knet: fix typo
8052
8053 knet: use knet_fdset_cloexec, minor lock fix
8054
8055 2010-10-18 Fabio M. Di Nitto <[email protected]>
8056
8057 Add generic util function to set CLOEXEC on fds
8058
8059 2010-10-18 Federico Simoncelli <[email protected]>
8060
8061 knet: new knet api interface
8062
8063 2010-10-18 Fabio M. Di Nitto <[email protected]>
8064
8065 main/vty: return errors from main loop
8066
8067 vty: add main loop and bind it to main.c
8068
8069 update .gitignore
8070
8071 vty: add config options for max connections
8072 add also basic scheleton to handle incoming connections
8073
8074 build: add default CFLAGS/LDFLAGS for test progs
8075
8076 main: parse all command line options
8077 add better error handling
8078
8079 2010-10-18 Federico Simoncelli <[email protected]>
8080
8081 knet: add control thread
8082
8083 2010-10-18 Fabio M. Di Nitto <[email protected]>
8084
8085 main: add basic daemon infrastructure
8086
8087 vty: make vty listener thread safe
8088
8089 Add knet_vty_close_listener
8090
8091 2010-10-16 Fabio M. Di Nitto <[email protected]>
8092
8093 build: fix missing initialization
8094 spotted building with F-14 gcc
8095
8096 2010-10-15 Fabio M. Di Nitto <[email protected]>
8097
8098 vty: fix idiotic use of memory
8099
8100 vty: don't use options that are too new...
8101
8102 build: add more options to configure.ac
8103
8104 vty: add tcp listener
8105
8106 2010-10-15 Federico Simoncelli <[email protected]>
8107
8108 knet: cleanup, minor fix for printf format
8109
8110 knet: add loop count
8111
8112 knet: thread-safe list cleanup
8113
8114 Merge branch 'master' of github.com:fabbione/kronosnet
8115
8116 knet: add thread-safe host list
8117
8118 2010-10-15 Fabio M. Di Nitto <[email protected]>
8119
8120 ring: fix endian conversion of KNET_FRAME_MAGIC
8121
8122 2010-10-14 Federico Simoncelli <[email protected]>
8123
8124 knet: add packet inspection placeholder
8125
8126 knet: complete rework in network communication
8127
8128 2010-10-14 Fabio M. Di Nitto <[email protected]>
8129
8130 knet: make knet config functions thread safe
8131 also export knet_execute_shell
8132
8133 knet: add support to add/remove ip
8134
8135 2010-10-13 Fabio M. Di Nitto <[email protected]>
8136
8137 knet: cleanup a few things
8138 remote unrequired var
8139 use knet_close internally
8140 avoid unnecessary casts to void *
8141 determine ifindex at creation time
8142
8143 knet: rework API to allow multiple knet ethernets
8144
8145 2010-10-11 Fabio M. Di Nitto <[email protected]>
8146
8147 Add dependency on automake and color tests
8148
8149 build: add more checks to configure.ac
8150
8151 Build test tools automatically
8152
8153 2010-10-11 Federico Simoncelli <[email protected]>
8154
8155 Add multiple destination hosts for each ring
8156
8157 Add ping test, ring test cleanup
8158
8159 2010-10-08 Federico Simoncelli <[email protected]>
8160
8161 Knet ring test cleanup
8162
8163 Merge branch 'master' of github.com:fabbione/kronosnet
8164
8165 Add knet_frame structure for ring messages
8166
8167 2010-10-08 Fabio M. Di Nitto <[email protected]>
8168
8169 knet test: use ip rather than ifconfig that requires /proc mounted
8170
8171 knet: add knet_set_up/down
8172
8173 2010-10-08 Federico Simoncelli <[email protected]>
8174
8175 Use unconnected sockets for ring messages
8176
8177 Fix code inconsistencies and variable naming
8178
8179 Merge branch 'master' of github.com:fabbione/kronosnet
8180
8181 Use INADDR_LOOPBACK macro for localhost
8182
8183 2010-10-08 Fabio M. Di Nitto <[email protected]>
8184
8185 build: update configure.ac
8186
8187 2010-10-08 Federico Simoncelli <[email protected]>
8188
8189 Merge branch 'master' of github.com:fabbione/kronosnet
8190
8191 Don't use ipv6 by default, use sockaddr_storage in knet_ring structure
8192
8193 2010-10-08 Fabio M. Di Nitto <[email protected]>
8194
8195 general: clean up usage of log_error
8196
8197 knet: add knet_execute_shell and tests
8198 this will be used to handle preup/up/down/post-down calls to shell
8199
8200 in case the script fails, both stdout and stderr outputs are logged
8201
8202 2010-10-08 Federico Simoncelli <[email protected]>
8203
8204 Display more information in ring test, minor fixes
8205
8206 2010-10-08 Fabio M. Di Nitto <[email protected]>
8207
8208 knet: drop redundant checks in set_mtu
8209 all those checks are performed by the kernel too, that has a much
8210 better idea of what MTU should be.
8211
8212 knet: add get/set mac address
8213 and update test cases
8214
8215 2010-10-07 Federico Simoncelli <[email protected]>
8216
8217 Add ring connection and unit test
8218
8219 Add syslog option, print error description in log_error
8220
8221 2010-10-07 Fabio M. Di Nitto <[email protected]>
8222
8223 Add knet_get_mtu/set_mtu functions
8224 and add test cases
8225
8226 knet: create static knet socket to handle all ioctl work
8227 also make ifreq static and global instead of polling everytime.
8228
8229 Introduce STATIC macro to access global static variables from tests/
8230
8231 Adjust test case to verify for sock errors
8232
8233 Add knet_open and knet_close functions
8234
8235 2010-10-07 Federico Simoncelli <[email protected]>
8236
8237 Add logging macros
8238
8239 2010-10-07 Fabio M. Di Nitto <[email protected]>
8240
8241 Clean master to start refactoring
8242
8243 2010-10-07 Federico Simoncelli <[email protected]>
8244
8245 Identify target node id by mac address
8246
8247 2010-10-06 Federico Simoncelli <[email protected]>
8248
8249 Set kronosnet hw address using nodeid value
8250
8251 Packet tests cleanup
8252
8253 Packet inspection should return the node id
8254
8255 Fix ipv6 multicast broadcast check
8256
8257 Add packet inspection for unicast delivery
8258
8259 Improve knet_header and payload allocation
8260
8261 Merge branch 'master' of github.com:fabbione/kronosnet
8262
8263 Skip documentation build if publican is not available
8264
8265 2010-10-06 Fabio M. Di Nitto <[email protected]>
8266
8267 debug: add code to decode src/dst mac address
8268 so it doesn't get lost...
8269
8270 2010-10-05 Fabio M. Di Nitto <[email protected]>
8271
8272 fix example config file
8273
8274 reduce logging verbosity in main
8275
8276 docs: add more format and simpler push to website
8277
8278 docs: add basic Makefile.am
8279
8280 Add basic publican docs
8281
8282 2010-09-30 Federico Simoncelli <[email protected]>
8283
8284 Merge branch 'master' of github.com:fabbione/kronosnet
8285
8286 Initial handshake implementation
8287
8288 2010-09-30 Fabio M. Di Nitto <[email protected]>
8289
8290 Revert "Test commit"
8291 This reverts commit 7f856ba6006ab644b8bc8fb08c83ac0cd83d0885.
8292
8293 Test commit
8294
8295 2010-09-27 Federico Simoncelli <[email protected]>
8296
8297 Increase socket receive buffer.
8298
8299 Extensive optimization in circular buffer algorithm.
8300
8301 2010-09-24 Jan Friesse <[email protected]>
8302
8303 Fix buffer overflow in sending data
8304 Previous dispatch_buffer function added sizeof(struct knet_header) to
8305 sended data. This is not needed and causes reading of not initialized
8306 data.
8307
8308 2010-09-23 Federico Simoncelli <[email protected]>
8309
8310 Several changes across the whole tree to use logsys instead of logthread.
8311
8312 2010-09-23 Jan Friesse <[email protected]>
8313
8314 Process knet_ips and mtu
8315 Ability to set link mtu and add ip address(es). Current implementation
8316 uses /sbin/ip command.
8317
8318 2010-09-14 Jan Friesse <[email protected]>
8319
8320 Netdevname again and (hopefully) done right
8321 Patch fixes netname option to exists in node config (where it logically
8322 belongs).
8323
8324 It also introduces config preup function to handle interface pre up
8325 (only netdevname is currently copied)
8326
8327 Add str_explode function
8328 Function supports explode of string to 1-n pieces separated by space or
8329 tab. This is useful in convert_ip and also in future parsing of
8330 knet_ips.
8331
8332 2010-09-14 Federico Simoncelli <[email protected]>
8333
8334 Circular buffer code cleanup
8335
8336 2010-09-13 Federico Simoncelli <[email protected]>
8337
8338 Improve protection against packet duplication (2)
8339
8340 2010-09-10 Federico Simoncelli <[email protected]>
8341
8342 Improve protection against packet duplication
8343
8344 2010-09-10 Jan Friesse <[email protected]>
8345
8346 Rename project to Kronosnet
8347
8348 2010-09-10 Federico Simoncelli <[email protected]>
8349
8350 Merge branch 'master' of github.com:fabbione/clusternet
8351
8352 Fix ring buffer cleanup boundary
8353
8354 2010-09-10 Jan Friesse <[email protected]>
8355
8356 Support for netname change in config file
8357 Default tun name can be changed in configuration file.
8358
8359 Default name also contains %d string now, so if more than one daemon is
8360 running, each device has unique name.
8361
8362 2010-09-09 Federico Simoncelli <[email protected]>
8363
8364 Use trivial algorithm to clear ring buffer
8365
8366 2010-09-09 Fabio M. Di Nitto <[email protected]>
8367
8368 Start adding ring buffer to handle packet duplication detection
8369
8370 2010-09-08 Fabio M. Di Nitto <[email protected]>
8371
8372 Fix check for rollover and add a sequence check
8373 the sequence check traps the out-of-order udp packets. We need
8374 to fix that with a circular buffer.
8375
8376 Add version of cnet_header, src_nodeid and dst_nodeid
8377
8378 2010-09-06 Fabio M. Di Nitto <[email protected]>
8379
8380 Add 4-rings example
8381
8382 2010-09-03 Fabio M. Di Nitto <[email protected]>
8383
8384 logging: fix config parsing and make logging conf static
8385
8386 Add support for packet sequence rollover
8387
8388 2010-09-01 Fabio M. Di Nitto <[email protected]>
8389
8390 Go back one step.. disable ping/pong, add redundant-ring support via pckt duplication
8391 XXX: must add handling of seq_num rollover or clusternet will halt after uint32_t pkcts
8392
8393 Add redundant ring in example.conf
8394
8395 Rename variable to something unique
8396
8397 add basic (but still useless) heartbeat process
8398
8399 Add packet parsing when received on the netsocket
8400 Laydown heartbeat thread
8401
8402 2010-08-31 Fabio M. Di Nitto <[email protected]>
8403
8404 Propagte our nodeid in the pckts
8405
8406 Start checking incoming network packages for our MAGIC signature
8407
8408 Start adding network control structure
8409 also update example.conf to match my basic testing setup
8410
8411 Allow command line to override configure file for debug options
8412
8413 2010-08-30 Fabio M. Di Nitto <[email protected]>
8414
8415 Re-enable IPv6 and remove some more dead code
8416
8417 Set standard port
8418
8419 Switch to UDP and remove a lot of useless code
8420
8421 Bridge incoming connections with the ethernet
8422 tested on 4 nodes, almost works :)
8423
8424 Detect local IPs and don't iniziate useless connections
8425
8426 Associate incoming and outgoing connections
8427 XXX: temporary drop IPv6 to preserve life on earth for a few more days
8428
8429 remove more paranoia logs
8430
8431 Remove some paranoia logging
8432
8433 add error and change variable name
8434
8435 Start briding FDs around
8436 lots of dirty hacks here..
8437
8438 Move RX on eth code in a separate thread
8439
8440 don't use MTU to set the buffer, just allocate a static one
8441 max ipv6 mtu * 2
8442
8443 change config keyword from "ip" to "nodeips"
8444
8445 Update example config
8446 we want to configure the interface for the user to avoid lots
8447 of complex polling at runtime.
8448
8449 2010-08-27 Fabio M. Di Nitto <[email protected]>
8450
8451 prepare to deliver pckts to localnet from remote
8452
8453 Allocate buffers dynamically based on MTU size
8454
8455 Start to create a main loop that's somewhere close to be useful
8456
8457 Add netsocket
8458
8459 2010-08-26 Fabio M. Di Nitto <[email protected]>
8460
8461 add common functions for socket comm and implement status command
8462
8463 clusternetd: allow cnet_tool to stop the daemon
8464
8465 cnet_tool: laydown structures and allow to send quit command
8466
8467 TEST: add simple main loop for testing
8468
8469 fix socket handling, add connection to cmd line tool
8470
8471 add cmd line tool
8472
8473 add basic local commands
8474
8475 add local socket data structure
8476
8477 control thread: set close on exit and start accepting connections
8478
8479 control thread: listen to local socket
8480
8481 control thread: add check to see if the thread does init properly
8482
8483 start adding the control thread to the daemon
8484 it will take care to handle local socket connections from the management
8485 tools
8486
8487 Add licence/copyright
8488
8489 build: update configure.ac for new headers/functions/types
8490
8491 add generic utils bit where to add all random useful things
8492
8493 nodes: allow selection of inet family for each node
8494
8495 2010-08-25 Fabio M. Di Nitto <[email protected]>
8496
8497 nodes: fix lookups parameters
8498
8499 nodes: allocate conn structs based on data resolution
8500
8501 nodes: rename tempip to avoid terms overload
8502
8503 parse space separated list of ip addresses
8504
8505 starts to add config parsing for nodes and nodes internal data structs
8506
8507 2010-08-16 Fabio M. Di Nitto <[email protected]>
8508
8509 add example.conf
8510
8511 update ignore file
8512
8513 conf: start to add node entries config parsing code
8514
8515 main: improve drastically error reporting from bottom layers
8516 init logging earlier to make config parsers re-usable for reload
8517
8518 logging: avoid leaking memory on error
8519
8520 main: move messages to DEBUG level
8521
8522 config: add support for rerouting and statistics collector config options
8523
8524 config: don't double free some bits on error
8525
8526 cnet: rename functions
8527
8528 config: add missing include
8529
8530 cnet: rename files to c(luster)net.{c,h}
8531
8532 tun: make it a tap/ether device
8533
8534 config: cleanup confdb after usage
8535
8536 config: don't leak memory on exit
8537
8538 config: reuse corosync.conf format via coroparser and add full config of logging
8539
8540 2010-08-13 Fabio M. Di Nitto <[email protected]>
8541
8542 add basic config file reading functionality
8543
8544 start implementing config file reading
8545
8546 clusternetd: add basic stuff