/* SPDX-License-Identifier: LGPL-2.1-or-later */ #include #include #include #include #include #include "sd-device.h" #include "sd-id128.h" #include "sd-json.h" #include "sd-varlink.h" #include "architecture.h" #include "ask-password-api.h" #include "blkid-util.h" #include "blockdev-util.h" #include "btrfs-util.h" #include "chase.h" #include "conf-files.h" #include "constants.h" #include "copy.h" #include "crypto-util.h" #include "cryptsetup-util.h" #include "device-private.h" #include "devnum-util.h" #include "dissect-image.h" #include "dm-util.h" #include "env-file.h" #include "env-util.h" #include "errno-util.h" #include "extension-util.h" #include "extract-word.h" #include "fd-util.h" #include "fileio.h" #include "format-util.h" #include "fsck-util.h" #include "fstab-util.h" #include "gpt.h" #include "hash-funcs.h" #include "hexdecoct.h" #include "hostname-setup.h" #include "image-policy.h" #include "import-util.h" #include "io-util.h" #include "iovec-util.h" #include "json-util.h" #include "libmount-util.h" #include "loop-util.h" #include "mkdir.h" #include "mount-util.h" #include "mountpoint-util.h" #include "namespace-util.h" #include "nulstr-util.h" #include "os-util.h" #include "path-util.h" #include "pcrextend-util.h" #include "pidref.h" #include "proc-cmdline.h" #include "process-util.h" #include "resize-fs.h" #include "rm-rf.h" #include "runtime-scope.h" #include "siphash24.h" #include "stat-util.h" #include "stdio-util.h" #include "string-util.h" #include "strv.h" #include "time-util.h" #include "udev-util.h" #include "user-util.h" #include "varlink-util.h" #include "xattr-util.h" /* how many times to wait for the device nodes to appear */ #define N_DEVICE_NODE_LIST_ATTEMPTS 10 static int allowed_fstypes(char ***ret_strv) { _cleanup_strv_free_ char **l = NULL; const char *e; assert(ret_strv); e = secure_getenv("SYSTEMD_DISSECT_FILE_SYSTEMS"); if (e) l = strv_split(e, ":"); else l = strv_new("btrfs", "erofs", "ext4", "f2fs", "squashfs", "vfat", "xfs"); if (!l) return -ENOMEM; *ret_strv = TAKE_PTR(l); return 0; } int dissect_fstype_ok(const char *fstype) { _cleanup_strv_free_ char **l = NULL; int r; /* When we automatically mount file systems, be a bit conservative by default what we are willing to * mount, just as an extra safety net to not mount with badly maintained legacy file system * drivers. */ r = allowed_fstypes(&l); if (r < 0) return r; if (strv_contains(l, fstype)) return true; log_debug("File system type '%s' is not allowed to be mounted as result of automatic dissection.", fstype); return false; } #if HAVE_BLKID static const char *getenv_fstype(PartitionDesignator d) { if (d < 0 || partition_designator_is_verity(d) || d == PARTITION_SWAP) return NULL; char *v = strjoina("SYSTEMD_DISSECT_FSTYPE_", partition_designator_to_string(d)); return secure_getenv(ascii_strupper(v)); } #endif int probe_sector_size(int fd, uint32_t *ret) { assert(fd >= 0); assert(ret); ssize_t ssz = gpt_probe(fd, /* ret_header= */ NULL, /* ret_entries= */ NULL, /* ret_n_entries= */ NULL, /* ret_entry_size= */ NULL); if (ssz == -ENOTUNIQ) return log_debug_errno(ssz, "Detected valid partition table at offsets matching multiple sector sizes, refusing."); if (ssz < 0) return ssz; if (ssz == 0) { log_debug("Couldn't find any partition table to derive sector size of."); *ret = 512; /* pick the traditional default */ return 0; /* indicate we didn't find it */ } log_debug("Determined sector size %" PRIu32 " based on discovered partition table.", (uint32_t) ssz); *ret = ssz; return 1; /* indicate we *did* find it */ } int probe_sector_size_prefer_ioctl(int fd, uint32_t *ret) { struct stat st; assert(fd >= 0); assert(ret); /* Just like probe_sector_size(), but if we are looking at a block device, will use the already * configured sector size rather than probing by contents */ if (fstat(fd, &st) < 0) return -errno; if (S_ISBLK(st.st_mode)) { int r; r = blockdev_get_sector_size(fd, ret); if (r < 0) return r; return 1; /* indicate we *did* find it, like probe_sector_size() does */ } return probe_sector_size(fd, ret); } #if HAVE_BLKID static int probe_blkid_filter(blkid_probe p) { _cleanup_strv_free_ char **fstypes = NULL; int r; assert(p); r = allowed_fstypes(&fstypes); if (r < 0) return r; /* allowed_fstypes() returns the list of filesystem types that we are willing to mount. For the * blkid probe filter we additionally need to be able to detect crypto_LUKS (so that we can set up * LUKS decryption for encrypted partitions) and swap (so that we can identify swap partitions). */ r = strv_extend_many(&fstypes, "crypto_LUKS", "swap"); if (r < 0) return r; errno = 0; r = sym_blkid_probe_filter_superblocks_type(p, BLKID_FLTR_ONLYIN, fstypes); if (r != 0) return errno_or_else(EINVAL); /* Note: don't call blkid_probe_filter_superblocks_usage() here. Both filter functions share the * same bitmap internally, and each call resets it before applying its own filter — so a subsequent * usage filter would wipe the type filter we just set. The ONLYIN type filter above already * excludes everything not in the allowed list, including RAID superblocks, so a separate usage * filter is redundant anyway. */ return 0; } #endif int probe_filesystem_full( int fd, const char *path, uint64_t offset, uint64_t size, bool restrict_fstypes, char **ret_fstype) { /* Try to find device content type and return it in *ret_fstype. If nothing is found, * 0/NULL will be returned. -EUCLEAN will be returned for ambiguous results, and a * different error otherwise. */ #if HAVE_BLKID _cleanup_(blkid_free_probep) blkid_probe b = NULL; _cleanup_free_ char *path_by_fd = NULL; _cleanup_close_ int fd_close = -EBADF; const char *fstype; int r; assert(fd >= 0 || path); assert(ret_fstype); r = dlopen_libblkid(LOG_DEBUG); if (r < 0) return r; if (fd < 0) { fd_close = open(path, O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY); if (fd_close < 0) return -errno; fd = fd_close; } if (!path) { r = fd_get_path(fd, &path_by_fd); if (r < 0) return r; path = path_by_fd; } if (size == 0) /* empty size? nothing found! */ goto not_found; b = sym_blkid_new_probe(); if (!b) return -ENOMEM; if (restrict_fstypes) { r = probe_blkid_filter(b); if (r < 0) return r; } /* The Linux kernel maintains separate block device caches for main ("whole") and partition block * devices, which means making a change to one might not be reflected immediately when reading via * the other. That's massively confusing when mixing accesses to such devices. Let's address this in * a limited way: when probing a file system that is not at the beginning of the block device we * apparently probe a partition via the main block device, and in that case let's first flush the * main block device cache, so that we get the data that the per-partition block device last * sync'ed on. * * This only works under the assumption that any tools that write to the partition block devices * issue an syncfs()/fsync() on the device after making changes. Typically file system formatting * tools that write a superblock onto a partition block device do that, however. */ if (offset != 0) if (ioctl(fd, BLKFLSBUF, 0) < 0) log_debug_errno(errno, "Failed to flush block device cache, ignoring: %m"); errno = 0; r = sym_blkid_probe_set_device( b, fd, offset, size == UINT64_MAX ? 0 : size); /* when blkid sees size=0 it understands "everything". We prefer using UINT64_MAX for that */ if (r != 0) return errno_or_else(ENOMEM); sym_blkid_probe_enable_superblocks(b, 1); sym_blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE); errno = 0; r = sym_blkid_do_safeprobe(b); if (r == _BLKID_SAFEPROBE_NOT_FOUND) goto not_found; if (r == _BLKID_SAFEPROBE_AMBIGUOUS) return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN), "Results ambiguous for partition %s", path); if (r == _BLKID_SAFEPROBE_ERROR) return log_debug_errno(errno_or_else(EIO), "Failed to probe partition %s: %m", path); assert(r == _BLKID_SAFEPROBE_FOUND); (void) sym_blkid_probe_lookup_value(b, "TYPE", &fstype, NULL); if (fstype) { log_debug("Probed fstype '%s' on partition %s.", fstype, path); return strdup_to_full(ret_fstype, fstype); } not_found: log_debug("No type detected on partition %s", path); *ret_fstype = NULL; return 0; #else return -EOPNOTSUPP; #endif } int probe_partition_table(int fd, char **ret_pttype) { /* Probes the whole device referenced by fd for a partition table and returns its blkid type (e.g. * "gpt" or "dos") in *ret_pttype, or NULL if none is found. Returns a negative error on failure * (including -EUCLEAN for ambiguous results). */ #if HAVE_BLKID _cleanup_(blkid_free_probep) blkid_probe b = NULL; const char *pttype = NULL; int r; assert(fd >= 0); assert(ret_pttype); r = dlopen_libblkid(LOG_DEBUG); if (r < 0) return r; b = sym_blkid_new_probe(); if (!b) return -ENOMEM; errno = 0; r = sym_blkid_probe_set_device(b, fd, /* offset= */ 0, /* size= */ 0 /* i.e. everything */); if (r != 0) return errno_or_else(ENOMEM); sym_blkid_probe_enable_partitions(b, 1); errno = 0; r = sym_blkid_do_safeprobe(b); if (r == _BLKID_SAFEPROBE_NOT_FOUND) { log_debug("No partition table detected."); *ret_pttype = NULL; return 0; } if (r == _BLKID_SAFEPROBE_AMBIGUOUS) return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN), "Partition table results ambiguous."); if (r == _BLKID_SAFEPROBE_ERROR) return log_debug_errno(errno_or_else(EIO), "Failed to probe for partition table: %m"); assert(r == _BLKID_SAFEPROBE_FOUND); (void) sym_blkid_probe_lookup_value(b, "PTTYPE", &pttype, /* len= */ NULL); if (!pttype) { log_debug("No partition table detected."); *ret_pttype = NULL; return 0; } log_debug("Probed partition table type '%s'.", pttype); return strdup_to_full(ret_pttype, pttype); #else return log_debug_errno(SYNTHETIC_ERRNO(EOPNOTSUPP), "Compiled without blkid support, cannot probe for partition table."); #endif } #if HAVE_BLKID static int image_policy_may_use( const ImagePolicy *policy, PartitionDesignator designator) { PartitionPolicyFlags f; /* For each partition we find in the partition table do a first check if it may exist at all given * the policy, or if it shall be ignored. */ f = image_policy_get_exhaustively(policy, designator); if (f < 0) return f; if ((f & _PARTITION_POLICY_USE_MASK) == PARTITION_POLICY_ABSENT) /* only flag set in policy is "absent"? then this partition may not exist at all */ return log_debug_errno( SYNTHETIC_ERRNO(ERFKILL), "Partition of designator '%s' exists, but not allowed by policy, refusing.", partition_designator_to_string(designator)); if ((f & _PARTITION_POLICY_USE_MASK & ~PARTITION_POLICY_ABSENT) == PARTITION_POLICY_UNUSED) { /* only "unused" or "unused" + "absent" are set? then don't use it */ log_debug("Partition of designator '%s' exists, and policy dictates to ignore it, doing so.", partition_designator_to_string(designator)); return false; /* ignore! */ } return true; /* use! */ } static int image_policy_check_protection( const ImagePolicy *policy, PartitionDesignator designator, PartitionPolicyFlags found_flags) { PartitionPolicyFlags policy_flags; /* Checks if the flags in the policy for the designated partition overlap the flags of what we found */ if (found_flags < 0) return found_flags; policy_flags = image_policy_get_exhaustively(policy, designator); if (policy_flags < 0) return policy_flags; if ((found_flags & policy_flags) == 0) { _cleanup_free_ char *found_flags_string = NULL, *policy_flags_string = NULL; (void) partition_policy_flags_to_string(found_flags, /* simplify= */ true, &found_flags_string); (void) partition_policy_flags_to_string(policy_flags, /* simplify= */ true, &policy_flags_string); return log_debug_errno(SYNTHETIC_ERRNO(ERFKILL), "Partition %s discovered with policy '%s' but '%s' was required, refusing.", partition_designator_to_string(designator), strnull(found_flags_string), strnull(policy_flags_string)); } return 0; } /* internal LUKS2 header defines */ #define LUKS2_FIXED_HDR_SIZE UINT64_C(0x1000) #define LUKS2_MAGIC "LUKS\xba\xbe" /* Matches the beginning of 'struct luks2_hdr_disk' from cryptsetup */ struct luks_header_incomplete { char luks_magic[sizeof(LUKS2_MAGIC) - 1]; be16_t version; be64_t hdr_len; }; /* 'integrity' information from LUKS JSON header. Currently, only 'type' is extracted/checked. */ struct luks_integrity_data { char *type; }; static int integrity_information(const char *name, sd_json_variant *v, sd_json_dispatch_flags_t flags, void *userdata) { static const sd_json_dispatch_field table[] = { { "type", SD_JSON_VARIANT_STRING, sd_json_dispatch_const_string, offsetof(struct luks_integrity_data, type), SD_JSON_MANDATORY }, {} }; return sd_json_dispatch(v, table, flags, userdata); } /* cryptsetup needs a loop device to work with a partition which has offset/size but * dissect may be running unprivileged. Implement a minimal custom LUKS header parser * checking integrity protection information. */ static int partition_is_luks2_integrity(int part_fd, uint64_t offset, uint64_t size) { _cleanup_(sd_json_variant_unrefp) sd_json_variant *v = NULL; _cleanup_free_ char *json = NULL; sd_json_variant *w; const char *key; struct luks_header_incomplete header; ssize_t sz, json_len; int r; assert(part_fd >= 0); if (size < LUKS2_FIXED_HDR_SIZE) { log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Partition is too small to contain a LUKS header."); return 0; } sz = pread(part_fd, &header, sizeof(header), offset); if (sz < 0) return log_error_errno(errno, "Failed to read LUKS header."); if (sz != sizeof(header)) return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to read LUKS header."); if (memcmp(header.luks_magic, LUKS2_MAGIC, sizeof(header.luks_magic)) != 0) { log_debug("Partition does not have a LUKS magic header, assuming no integrity."); return 0; } if (be16toh(header.version) != 2) { log_debug("Partition is LUKS v%" PRIu16 ", not LUKS2, assuming no integrity.", be16toh(header.version)); return 0; } if (be64toh(header.hdr_len) > size) return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "LUKS header length exceeds partition size."); if (be64toh(header.hdr_len) <= LUKS2_FIXED_HDR_SIZE || offset > UINT64_MAX - be64toh(header.hdr_len)) return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "Invalid LUKS header length: %" PRIu64 ".", be64toh(header.hdr_len)); if (be64toh(header.hdr_len) - LUKS2_FIXED_HDR_SIZE > 16U * 1024U * 1024U) return log_error_errno(SYNTHETIC_ERRNO(EINVAL), "LUKS header JSON area too large: %" PRIu64 ".", be64toh(header.hdr_len)); json_len = be64toh(header.hdr_len) - LUKS2_FIXED_HDR_SIZE; json = malloc(json_len + 1); if (!json) return -ENOMEM; sz = pread(part_fd, json, json_len, offset + LUKS2_FIXED_HDR_SIZE); if (sz < 0) return log_error_errno(errno, "Failed to read LUKS JSON header."); if (sz != json_len) return log_error_errno(SYNTHETIC_ERRNO(EIO), "Failed to read LUKS JSON header."); json[sz] = '\0'; r = sd_json_parse(json, /* flags= */ 0, &v, /* reterr_line= */ NULL, /* reterr_column= */ NULL); if (r < 0) return log_error_errno(r, "Failed to parse LUKS JSON header."); v = sd_json_variant_by_key(v, "segments"); if (!v || !sd_json_variant_is_object(v)) { log_debug("LUKS JSON header lacks 'segments' information, assuming no integrity."); return 0; } /* Verify that all segments have integrity protection */ JSON_VARIANT_OBJECT_FOREACH(key, w, v) { struct luks_integrity_data data = {}; static const sd_json_dispatch_field dispatch_segment[] = { { "integrity", SD_JSON_VARIANT_OBJECT, integrity_information, 0, SD_JSON_MANDATORY }, {} }; r = sd_json_dispatch(w, dispatch_segment, SD_JSON_ALLOW_EXTENSIONS, &data); if (r < 0) { log_debug("Failed to get integrity information from LUKS JSON for segment %s, assuming no integrity.", key); return 0; } /* We don't require a particular integrity algorithm, everything but 'none' (which shouldn't * be there in the first place but is theoretically possible) works. */ if (streq(data.type, "none")) return 0; } return 1; } static int image_policy_check_partition_flags( const ImagePolicy *policy, PartitionDesignator designator, uint64_t gpt_flags) { PartitionPolicyFlags policy_flags; bool b; /* Checks if the partition flags in the policy match reality */ policy_flags = image_policy_get_exhaustively(policy, designator); if (policy_flags < 0) return policy_flags; b = FLAGS_SET(gpt_flags, SD_GPT_FLAG_READ_ONLY); if ((policy_flags & _PARTITION_POLICY_READ_ONLY_MASK) == (b ? PARTITION_POLICY_READ_ONLY_OFF : PARTITION_POLICY_READ_ONLY_ON)) return log_debug_errno(SYNTHETIC_ERRNO(ERFKILL), "Partition %s has 'read-only' flag incorrectly set (must be %s, is %s), refusing.", partition_designator_to_string(designator), one_zero(!b), one_zero(b)); b = FLAGS_SET(gpt_flags, SD_GPT_FLAG_GROWFS); if ((policy_flags & _PARTITION_POLICY_GROWFS_MASK) == (b ? PARTITION_POLICY_GROWFS_OFF : PARTITION_POLICY_GROWFS_ON)) return log_debug_errno(SYNTHETIC_ERRNO(ERFKILL), "Partition %s has 'growfs' flag incorrectly set (must be %s, is %s), refusing.", partition_designator_to_string(designator), one_zero(!b), one_zero(b)); return 0; } static int dissected_image_probe_filesystems( DissectedImage *m, int fd, const ImagePolicy *policy) { int r; assert(m); /* Fill in file system types if we don't know them yet. */ for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { DissectedPartition *p = m->partitions + i; PartitionPolicyFlags found_flags; if (!p->found) continue; if (!p->fstype) { /* If we have an fd referring to the partition block device, use that. Otherwise go * via the whole block device or backing regular file, and read via offset. */ if (p->mount_node_fd >= 0) r = probe_filesystem_full(p->mount_node_fd, p->node, 0, UINT64_MAX, /* bool restrict_fstypes= */ true, &p->fstype); else r = probe_filesystem_full(fd, p->node, p->offset, p->size, /* bool restrict_fstypes= */ true, &p->fstype); if (r < 0) return r; } if (streq_ptr(p->fstype, "crypto_LUKS")) { m->encrypted = true; if (p->mount_node_fd >= 0) r = partition_is_luks2_integrity(p->mount_node_fd, /* offset= */ 0, /* size= */ UINT64_MAX); else r = partition_is_luks2_integrity(fd, p->offset, p->size); if (r < 0) return r; /* found this one, it's definitely encrypted + with or without integrity checking */ found_flags = PARTITION_POLICY_UNUSED|(r > 0 ? PARTITION_POLICY_ENCRYPTEDWITHINTEGRITY : PARTITION_POLICY_ENCRYPTED); } else /* found it, but it's definitely not encrypted, hence mask the encrypted flag, but * set all other ways that indicate "present". */ found_flags = PARTITION_POLICY_UNUSED|PARTITION_POLICY_UNPROTECTED|PARTITION_POLICY_VERITY|PARTITION_POLICY_SIGNED; if (p->fstype && fstype_is_ro(p->fstype)) p->rw = false; if (!p->rw) p->growfs = false; /* We might have learnt more about the file system now (i.e. whether it is encrypted or not), * hence we need to validate this against policy again, to see if the policy still matches * with this new information. Note that image_policy_check_protection() will check for * overlap between what's allowed in the policy and what we pass as 'found_policy' here. In * the unencrypted case we thus might pass an overly unspecific mask here (i.e. unprotected * OR verity OR signed), but that's fine since the earlier policy check already checked more * specific which of those three cases where OK. Keep in mind that this function here only * looks at specific partitions (and thus can only deduce encryption or not) but not the * overall partition table (and thus cannot deduce verity or not). The earlier dissection * checks already did the relevant checks that look at the whole partition table, and * enforced policy there as needed. */ r = image_policy_check_protection(policy, i, found_flags); if (r < 0) return r; } return 0; } static void check_partition_flags( const char *node, unsigned long long pflags, unsigned long long supported) { assert(node); /* Mask away all flags supported by this partition's type and the three flags the UEFI spec defines generically */ pflags &= ~(supported | SD_GPT_FLAG_REQUIRED_PARTITION | SD_GPT_FLAG_NO_BLOCK_IO_PROTOCOL | SD_GPT_FLAG_LEGACY_BIOS_BOOTABLE); if (pflags == 0) return; /* If there are other bits set, then log about it, to make things discoverable */ for (unsigned i = 0; i < sizeof(pflags) * 8; i++) { unsigned long long bit = 1ULL << i; if (!FLAGS_SET(pflags, bit)) continue; log_debug("Unexpected partition flag %llu set on %s!", bit, node); } } #endif int dissected_image_name_from_path(const char *path, char **ret) { int r; assert(path); assert(ret); _cleanup_free_ char *filename = NULL; r = path_extract_filename(path, &filename); if (r < 0) return r; _cleanup_free_ char *name = NULL; r = raw_strip_suffixes(filename, &name); if (r < 0) return r; if (!image_name_is_valid(name)) { log_debug("Image name %s is not valid, ignoring.", strna(name)); *ret = NULL; return 0; } *ret = TAKE_PTR(name); return 1; } #if HAVE_BLKID static int dissected_image_new(const char *path, DissectedImage **ret) { _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; _cleanup_free_ char *name = NULL; int r; assert(ret); if (path) { r = dissected_image_name_from_path(path, &name); if (r < 0) return r; } m = new(DissectedImage, 1); if (!m) return -ENOMEM; *m = (DissectedImage) { .has_init_system = -1, .image_name = TAKE_PTR(name), }; for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) m->partitions[i] = DISSECTED_PARTITION_NULL; *ret = TAKE_PTR(m); return 0; } #endif static void dissected_partition_done(DissectedPartition *p) { assert(p); free(p->fstype); free(p->node); free(p->label); free(p->decrypted_fstype); free(p->decrypted_node); free(p->mount_options); safe_close(p->mount_node_fd); safe_close(p->fsmount_fd); *p = DISSECTED_PARTITION_NULL; } static int acquire_sig_for_roothash( int fd, uint64_t partition_offset, uint64_t partition_size, struct iovec *ret_root_hash, struct iovec *ret_root_hash_sig) { int r; assert(fd >= 0); if (partition_offset == UINT64_MAX || partition_size == UINT64_MAX) return -EINVAL; if (partition_size > 4*1024*1024) /* Signature data cannot possible be larger than 4M, refuse that */ return log_debug_errno(SYNTHETIC_ERRNO(EFBIG), "Verity signature partition is larger than 4M, refusing."); _cleanup_free_ char *buf = new(char, partition_size+1); if (!buf) return -ENOMEM; ssize_t n = pread(fd, buf, partition_size, partition_offset); if (n < 0) return -errno; if ((uint64_t) n != partition_size) return -EIO; const char *e = memchr(buf, 0, partition_size); if (e) { /* If we found a NUL byte then the rest of the data must be NUL too */ if (!memeqzero(e, partition_size - (e - buf))) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Signature data contains embedded NUL byte."); } else buf[partition_size] = 0; _cleanup_(sd_json_variant_unrefp) sd_json_variant *v = NULL; r = sd_json_parse(buf, 0, &v, /* reterr_line= */ NULL, /* reterr_column= */ NULL); if (r < 0) return log_debug_errno(r, "Failed to parse signature JSON data: %m"); sd_json_variant *rh = sd_json_variant_by_key(v, "rootHash"); if (!rh) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Signature JSON object lacks 'rootHash' field."); _cleanup_(iovec_done) struct iovec root_hash = {}; r = sd_json_variant_unhex(rh, &root_hash.iov_base, &root_hash.iov_len); if (r < 0) return log_debug_errno(r, "Failed to parse root hash field: %m"); sd_json_variant *sig = sd_json_variant_by_key(v, "signature"); if (!sig) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Signature JSON object lacks 'signature' field."); _cleanup_(iovec_done) struct iovec root_hash_sig = {}; r = sd_json_variant_unbase64(sig, &root_hash_sig.iov_base, &root_hash_sig.iov_len); if (r < 0) return log_debug_errno(r, "Failed to parse signature field: %m"); if (ret_root_hash) *ret_root_hash = TAKE_STRUCT(root_hash); if (ret_root_hash_sig) *ret_root_hash_sig = TAKE_STRUCT(root_hash_sig); return 0; } #if HAVE_BLKID static int diskseq_should_be_used( const char *whole_devname, uint64_t diskseq, DissectImageFlags flags) { int r; assert(whole_devname); /* No diskseq. We cannot use by-diskseq symlink. */ if (diskseq == 0) return false; /* Do not use by-diskseq link unless DISSECT_IMAGE_DISKSEQ_DEVNODE flag is explicitly set. */ if (!FLAGS_SET(flags, DISSECT_IMAGE_DISKSEQ_DEVNODE)) return false; _cleanup_(sd_device_unrefp) sd_device *dev = NULL; r = sd_device_new_from_devname(&dev, whole_devname); if (r < 0) return r; /* When ID_IGNORE_DISKSEQ udev property is set, the by-diskseq symlink will not be created. */ r = device_get_property_bool(dev, "ID_IGNORE_DISKSEQ"); if (r >= 0) return !r; /* If explicitly specified, use it. */ if (r != -ENOENT) return r; return true; } static int make_partition_devname( const char *whole_devname, uint64_t diskseq, int nr, DissectImageFlags flags, char **ret) { _cleanup_free_ char *s = NULL; int r; assert(whole_devname); assert(nr != 0); /* zero is not a valid partition nr */ assert(ret); r = diskseq_should_be_used(whole_devname, diskseq, flags); if (r < 0) log_debug_errno(r, "Failed to determine if diskseq should be used for %s, assuming no, ignoring: %m", whole_devname); if (r <= 0) { /* Given a whole block device node name (e.g. /dev/sda or /dev/loop7) generate a partition * device name (e.g. /dev/sda7 or /dev/loop7p5). The rule the kernel uses is simple: if whole * block device node name ends in a digit, then suffix a 'p', followed by the partition * number. Otherwise, just suffix the partition number without any 'p'. */ if (nr < 0) { /* whole disk? */ s = strdup(whole_devname); if (!s) return -ENOMEM; } else { r = partition_node_of(whole_devname, nr, &s); if (r < 0) return r; } } else { if (nr < 0) /* whole disk? */ r = asprintf(&s, "/dev/disk/by-diskseq/%" PRIu64, diskseq); else r = asprintf(&s, "/dev/disk/by-diskseq/%" PRIu64 "-part%i", diskseq, nr); if (r < 0) return -ENOMEM; } *ret = TAKE_PTR(s); return 0; } static int open_partition( const char *node, bool is_partition, const LoopDevice *loop) { _cleanup_(sd_device_unrefp) sd_device *dev = NULL; _cleanup_close_ int fd = -EBADF; dev_t devnum; int r; assert(node); assert(loop || !is_partition); fd = open(node, O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY); if (fd < 0) return -errno; if (!loop) return TAKE_FD(fd); /* Check if the block device is a child of (or equivalent to) the originally provided one. */ r = block_device_new_from_fd(fd, is_partition ? BLOCK_DEVICE_LOOKUP_WHOLE_DISK : 0, &dev); if (r < 0) return r; r = sd_device_get_devnum(dev, &devnum); if (r < 0) return r; if (loop->devno != devnum) return -ENXIO; /* Also check diskseq. */ if (loop->diskseq != 0) { uint64_t diskseq; r = fd_get_diskseq(fd, &diskseq); if (r < 0) return r; if (loop->diskseq != diskseq) return -ENXIO; } log_debug("Opened %s (fd=%i, whole_block_devnum=" DEVNUM_FORMAT_STR ", diskseq=%" PRIu64 ").", node, fd, DEVNUM_FORMAT_VAL(loop->devno), loop->diskseq); return TAKE_FD(fd); } static int compare_arch(Architecture a, Architecture b) { if (a == b) return 0; if (a == native_architecture()) return 1; if (b == native_architecture()) return -1; #ifdef ARCHITECTURE_SECONDARY if (a == ARCHITECTURE_SECONDARY) return 1; if (b == ARCHITECTURE_SECONDARY) return -1; #endif return 0; } static bool image_filter_test(const ImageFilter *filter, PartitionDesignator d, const char *label) { assert(d < _PARTITION_DESIGNATOR_MAX); if (d < 0) /* For unspecified designators we have no filter expression */ return true; if (!filter || !filter->pattern[d]) return true; return fnmatch(filter->pattern[d], strempty(label), FNM_NOESCAPE) == 0; } static int dissect_image_from_unpartitioned( const char *devname, uint64_t diskseq, const sd_id128_t *uuid, bool encrypted, const VeritySettings *verity, const MountOptions *mount_options, const ImagePolicy *policy, const ImageFilter *filter, int *mount_node_fd, /* taken over on success */ char **fstype, /* taken over on success */ DissectedImage *m, DissectImageFlags flags, PartitionPolicyFlags found_flags) { _cleanup_free_ char *n = NULL, *o = NULL; const char *options = NULL; int r; assert(devname); assert(m); assert(fstype); POINTER_MAY_BE_NULL(mount_node_fd); if (!image_filter_test(filter, PARTITION_ROOT, /* label= */ NULL)) /* do a filter check with an empty partition label */ return -ECOMM; r = image_policy_may_use(policy, PARTITION_ROOT); if (r < 0) return r; if (r == 0) /* policy says ignore this, so we ignore it */ return -ENOPKG; r = image_policy_check_protection(policy, PARTITION_ROOT, found_flags); if (r < 0) return r; r = image_policy_check_partition_flags(policy, PARTITION_ROOT, 0); /* we have no gpt partition flags, hence check against all bits off */ if (r < 0) return r; r = make_partition_devname(devname, diskseq, /* nr= */ -1, flags, &n); if (r < 0) return r; m->single_file_system = true; m->encrypted = encrypted; m->has_verity = verity && verity->data_path; m->verity_ready = verity_settings_data_covers(verity, PARTITION_ROOT); m->has_verity_sig = false; /* signature not embedded, must be specified */ m->verity_sig_ready = m->verity_ready && iovec_is_set(&verity->root_hash); if (uuid) m->image_uuid = *uuid; options = mount_options_from_designator(mount_options, PARTITION_ROOT); if (options) { o = strdup(options); if (!o) return -ENOMEM; } m->partitions[PARTITION_ROOT] = (DissectedPartition) { .found = true, .rw = !m->verity_ready && !fstype_is_ro(*fstype), .partno = -1, .architecture = _ARCHITECTURE_INVALID, .fstype = TAKE_PTR(*fstype), .node = TAKE_PTR(n), .mount_options = TAKE_PTR(o), .mount_node_fd = mount_node_fd ? TAKE_FD(*mount_node_fd) : -EBADF, .size = UINT64_MAX, .fsmount_fd = -EBADF, }; return 0; } static int dissect_image( DissectedImage *m, int fd, const char *devname, const VeritySettings *verity, const MountOptions *mount_options, const ImagePolicy *policy, const ImageFilter *filter, DissectImageFlags flags) { sd_id128_t root_uuid = SD_ID128_NULL, root_verity_uuid = SD_ID128_NULL; sd_id128_t usr_uuid = SD_ID128_NULL, usr_verity_uuid = SD_ID128_NULL; bool is_gpt, is_mbr, multiple_generic = false, generic_rw = false, /* initialize to appease gcc */ generic_growfs = false; _cleanup_(blkid_free_probep) blkid_probe b = NULL; _cleanup_free_ char *generic_node = NULL; sd_id128_t generic_uuid = SD_ID128_NULL; const char *pttype = NULL; blkid_partlist pl; int r, generic_nr = -1, n_partitions; assert(m); assert(fd >= 0); assert(devname); assert(!verity || verity->designator < 0 || IN_SET(verity->designator, PARTITION_ROOT, PARTITION_USR)); assert(!verity || iovec_is_valid(&verity->root_hash)); assert(!verity || iovec_is_valid(&verity->root_hash_sig)); assert(!verity || iovec_is_set(&verity->root_hash) || !iovec_is_set(&verity->root_hash_sig)); assert(!((flags & DISSECT_IMAGE_GPT_ONLY) && (flags & DISSECT_IMAGE_NO_PARTITION_TABLE))); assert(m->sector_size > 0); /* Probes a disk image, and returns information about what it found in *ret. * * Returns -ENOPKG if no suitable partition table or file system could be found. * Returns -EADDRNOTAVAIL if a root hash was specified but no matching root/verity partitions found. * Returns -ENXIO if we couldn't find any partition suitable as root or /usr partition * Returns -ENOTUNIQ if we only found multiple generic partitions and thus don't know what to do with that * Returns -ERFKILL if image doesn't match image policy * Returns -EBADR if verity data was provided externally for an image that has a GPT partition table (i.e. is not just a naked fs) * Returns -EPROTONOSUPPORT if DISSECT_IMAGE_ADD_PARTITION_DEVICES is set but the block device does not have partition logic enabled * Returns -ENOMSG if we didn't find a single usable partition (and DISSECT_IMAGE_REFUSE_EMPTY is set) * Returns -EUCLEAN if some file system had an ambiguous file system superblock signature */ uint64_t diskseq = m->loop ? m->loop->diskseq : 0; if (verity && iovec_is_set(&verity->root_hash)) { sd_id128_t fsuuid, vuuid; /* If a root hash is supplied, then we use the root partition that has a UUID that match the * first 128-bit of the root hash. And we use the verity partition that has a UUID that match * the final 128-bit. */ if (verity->root_hash.iov_len < sizeof(sd_id128_t)) return -EINVAL; memcpy(&fsuuid, verity->root_hash.iov_base, sizeof(sd_id128_t)); memcpy(&vuuid, (const uint8_t*) verity->root_hash.iov_base + verity->root_hash.iov_len - sizeof(sd_id128_t), sizeof(sd_id128_t)); if (sd_id128_is_null(fsuuid)) return -EINVAL; if (sd_id128_is_null(vuuid)) return -EINVAL; /* If the verity data declares it's for the /usr partition, then search for that, in all * other cases assume it's for the root partition. */ if (verity->designator == PARTITION_USR) { usr_uuid = fsuuid; usr_verity_uuid = vuuid; } else { root_uuid = fsuuid; root_verity_uuid = vuuid; } } r = dlopen_libblkid(LOG_DEBUG); if (r < 0) return r; b = sym_blkid_new_probe(); if (!b) return -ENOMEM; r = probe_blkid_filter(b); if (r < 0) return r; errno = 0; r = sym_blkid_probe_set_device(b, fd, 0, 0); if (r != 0) return errno_or_else(ENOMEM); errno = 0; r = sym_blkid_probe_set_sectorsize(b, m->sector_size); if (r != 0) return errno_or_else(EIO); if ((flags & DISSECT_IMAGE_GPT_ONLY) == 0) { /* Look for file system superblocks, unless we only shall look for GPT partition tables */ sym_blkid_probe_enable_superblocks(b, 1); sym_blkid_probe_set_superblocks_flags(b, BLKID_SUBLKS_TYPE|BLKID_SUBLKS_USAGE|BLKID_SUBLKS_UUID); } sym_blkid_probe_enable_partitions(b, 1); sym_blkid_probe_set_partitions_flags(b, BLKID_PARTS_ENTRY_DETAILS); errno = 0; r = sym_blkid_do_safeprobe(b); if (r == _BLKID_SAFEPROBE_ERROR) return errno_or_else(EIO); if (IN_SET(r, _BLKID_SAFEPROBE_AMBIGUOUS, _BLKID_SAFEPROBE_NOT_FOUND)) return log_debug_errno(SYNTHETIC_ERRNO(ENOPKG), "Failed to identify any partition table."); assert(r == _BLKID_SAFEPROBE_FOUND); if ((!(flags & DISSECT_IMAGE_GPT_ONLY) && (flags & DISSECT_IMAGE_GENERIC_ROOT)) || (flags & DISSECT_IMAGE_NO_PARTITION_TABLE)) { _cleanup_free_ char *root_fstype_string = NULL; const char *usage = NULL; bool encrypted; r = partition_policy_determine_fstype(policy, PARTITION_ROOT, &encrypted, &root_fstype_string); if (r < 0) return r; /* If flags permit this, also allow using non-partitioned single-filesystem images */ (void) sym_blkid_probe_lookup_value(b, "USAGE", &usage, NULL); if (STRPTR_IN_SET(usage, "filesystem", "crypto")) { _cleanup_free_ char *t = NULL; const char *fstype = NULL; _cleanup_close_ int mount_node_fd = -EBADF; sd_id128_t uuid = SD_ID128_NULL; PartitionPolicyFlags found_flags; /* OK, we have found a file system, that's our root partition then. */ if (!root_fstype_string) { (void) sym_blkid_probe_lookup_value(b, "TYPE", &fstype, NULL); /* blkid will return FAT's serial number as UUID, hence it is quite possible that * parsing this will fail. We'll ignore the ID, since it's just too short to be * useful as true identifier. */ (void) blkid_probe_lookup_value_id128(b, "UUID", &uuid); } else /* The policy fstype flags translate to the literal fstype name of each filesystem. */ fstype = root_fstype_string; encrypted = encrypted || streq_ptr(fstype, "crypto_LUKS"); if (verity_settings_data_covers(verity, PARTITION_ROOT)) found_flags = iovec_is_set(&verity->root_hash_sig) ? PARTITION_POLICY_SIGNED : PARTITION_POLICY_VERITY; else if (encrypted) { r = partition_is_luks2_integrity(fd, /* offset= */ 0, /* size= */ UINT64_MAX); if (r < 0) return r; found_flags = r > 0 ? PARTITION_POLICY_ENCRYPTEDWITHINTEGRITY : PARTITION_POLICY_ENCRYPTED; } else found_flags = PARTITION_POLICY_UNPROTECTED; if (FLAGS_SET(flags, DISSECT_IMAGE_PIN_PARTITION_DEVICES)) { mount_node_fd = open_partition(devname, /* is_partition= */ false, m->loop); if (mount_node_fd < 0) return mount_node_fd; } if (fstype) { t = strdup(fstype); if (!t) return -ENOMEM; } return dissect_image_from_unpartitioned( devname, diskseq, &uuid, encrypted, verity, mount_options, policy, filter, &mount_node_fd, &t, m, flags, found_flags); } } (void) sym_blkid_probe_lookup_value(b, "PTTYPE", &pttype, NULL); if (!pttype) return -ENOPKG; is_gpt = streq_ptr(pttype, "gpt"); is_mbr = streq_ptr(pttype, "dos"); if (!is_gpt && ((flags & DISSECT_IMAGE_GPT_ONLY) || !is_mbr)) return -ENOPKG; /* We support external verity data partitions only if the image has no partition table */ if (verity && verity->data_path) return -EBADR; if (FLAGS_SET(flags, DISSECT_IMAGE_ADD_PARTITION_DEVICES)) { /* Safety check: refuse block devices that carry a partition table but for which the kernel doesn't * do partition scanning. */ r = blockdev_partscan_enabled_fd(fd); if (r < 0) return r; if (r == 0) return -EPROTONOSUPPORT; } (void) blkid_probe_lookup_value_id128(b, "PTUUID", &m->image_uuid); errno = 0; pl = sym_blkid_probe_get_partitions(b); if (!pl) return errno_or_else(ENOMEM); errno = 0; n_partitions = sym_blkid_partlist_numof_partitions(pl); if (n_partitions < 0) return errno_or_else(EIO); for (int i = 0; i < n_partitions; i++) { _cleanup_free_ char *node = NULL; unsigned long long pflags; blkid_loff_t start, size; blkid_partition pp; int nr; errno = 0; pp = sym_blkid_partlist_get_partition(pl, i); if (!pp) return errno_or_else(EIO); pflags = sym_blkid_partition_get_flags(pp); errno = 0; nr = sym_blkid_partition_get_partno(pp); if (nr < 0) return errno_or_else(EIO); errno = 0; start = sym_blkid_partition_get_start(pp); if (start < 0) return errno_or_else(EIO); assert((uint64_t) start < UINT64_MAX/512); errno = 0; size = sym_blkid_partition_get_size(pp); if (size < 0) return errno_or_else(EIO); assert((uint64_t) size < UINT64_MAX/512); /* While probing we need the non-diskseq device node name to access the thing, hence mask off * DISSECT_IMAGE_DISKSEQ_DEVNODE. */ r = make_partition_devname(devname, diskseq, nr, flags & ~DISSECT_IMAGE_DISKSEQ_DEVNODE, &node); if (r < 0) return r; /* So here's the thing: after the main ("whole") block device popped up it might take a while * before the kernel fully probed the partition table. Waiting for that to finish is icky in * userspace. So here's what we do instead. We issue the BLKPG_ADD_PARTITION ioctl to add the * partition ourselves, racing against the kernel. Good thing is: if this call fails with * EBUSY then the kernel was quicker than us, and that's totally OK, the outcome is good for * us: the device node will exist. If OTOH our call was successful we won the race. Which is * also good as the outcome is the same: the partition block device exists, and we can use * it. * * Kernel returns EBUSY if there's already a partition by that number or an overlapping * partition already existent. */ if (FLAGS_SET(flags, DISSECT_IMAGE_ADD_PARTITION_DEVICES)) { r = block_device_add_partition(fd, nr, (uint64_t) start * 512, (uint64_t) size * 512); if (r < 0) { if (r != -EBUSY) return log_debug_errno(r, "BLKPG_ADD_PARTITION failed: %m"); log_debug_errno(r, "Kernel was quicker than us in adding partition %i.", nr); } else log_debug("We were quicker than kernel in adding partition %i.", nr); } if (is_gpt) { const char *label; sd_id128_t type_id, id; GptPartitionType type; bool rw = true, growfs = false; r = blkid_partition_get_uuid_id128(pp, &id); if (r < 0) { log_debug_errno(r, "Failed to read partition UUID, ignoring: %m"); continue; } r = blkid_partition_get_type_id128(pp, &type_id); if (r < 0) { log_debug_errno(r, "Failed to read partition type UUID, ignoring: %m"); continue; } type = gpt_partition_type_from_uuid(type_id); label = sym_blkid_partition_get_name(pp); /* libblkid returns NULL here if empty */ /* systemd-sysupdate expects empty partitions to be marked with an "_empty" label, hence ignore them here. */ if (streq_ptr(label, "_empty")) continue; if (!image_filter_test(filter, type.designator, label)) continue; log_debug("Dissecting %s partition with label %s and UUID %s.", strna(partition_designator_to_string(type.designator)), strna(label), SD_ID128_TO_UUID_STRING(id)); /* Note that we don't check the SD_GPT_FLAG_NO_AUTO flag for the ESP, as it is * not defined there. We instead check the SD_GPT_FLAG_NO_BLOCK_IO_PROTOCOL, as * recommended by the UEFI spec (See "12.3.3 Number and Location of System * Partitions"). */ if (FLAGS_SET(pflags, SD_GPT_FLAG_NO_AUTO) && type.designator != PARTITION_ESP) { log_debug("Partition has 'no auto' flag set, ignoring."); continue; } if (!verity && partition_designator_is_verity(type.designator)) { log_debug("Partition is a verity hash or verity signature partition but no verity was requested, ignoring."); continue; } PartitionDesignator vd = partition_verity_to_data(type.designator); if (verity && verity->designator >= 0 && vd >= 0 && vd != verity->designator) { log_debug("Partition is a %s partition but verity was only requested for the %s partition, ignoring.", partition_designator_to_string(type.designator), partition_designator_to_string(verity->designator)); continue; } const char *fstype = getenv_fstype(type.designator); if (IN_SET(type.designator, PARTITION_HOME, PARTITION_SRV, PARTITION_XBOOTLDR, PARTITION_TMP)) { check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO | SD_GPT_FLAG_READ_ONLY | SD_GPT_FLAG_GROWFS); rw = !(pflags & SD_GPT_FLAG_READ_ONLY); growfs = FLAGS_SET(pflags, SD_GPT_FLAG_GROWFS); /* XBOOTLDR cannot be integrity protected (since firmware needs to access * it), hence be restrictive with the fs choice when dissecting. */ if (type.designator == PARTITION_XBOOTLDR && !fstype) fstype = "vfat"; } else if (type.designator == PARTITION_ESP) { if (FLAGS_SET(pflags, SD_GPT_FLAG_NO_BLOCK_IO_PROTOCOL)) { log_debug("ESP Partition has 'no block io' flag set, ignoring."); continue; } /* Effectively the ESP has to be VFAT, let's enforce this */ if (!fstype) fstype = "vfat"; } else if (IN_SET(type.designator, PARTITION_ROOT, PARTITION_USR)) { sd_id128_t expected_uuid = type.designator == PARTITION_ROOT ? root_uuid : usr_uuid; check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO | SD_GPT_FLAG_READ_ONLY | SD_GPT_FLAG_GROWFS); if (!sd_id128_is_null(expected_uuid) && !sd_id128_equal(expected_uuid, id)) { log_debug("Partition UUID '%s' does not match expected UUID '%s' derived from %s verity hash, ignoring.", SD_ID128_TO_UUID_STRING(id), SD_ID128_TO_UUID_STRING(expected_uuid), partition_designator_to_string(type.designator)); continue; } rw = !(pflags & SD_GPT_FLAG_READ_ONLY); growfs = FLAGS_SET(pflags, SD_GPT_FLAG_GROWFS); } else if (IN_SET(type.designator, PARTITION_ROOT_VERITY, PARTITION_USR_VERITY)) { sd_id128_t expected_uuid = type.designator == PARTITION_ROOT_VERITY ? root_verity_uuid : usr_verity_uuid; check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO | SD_GPT_FLAG_READ_ONLY); m->has_verity = true; if (!sd_id128_is_null(expected_uuid) && !sd_id128_equal(expected_uuid, id)) { log_debug("Partition UUID '%s' does not match expected UUID '%s' derived from %s verity hash, ignoring.", SD_ID128_TO_UUID_STRING(id), SD_ID128_TO_UUID_STRING(expected_uuid), partition_designator_to_string(partition_verity_to_data(type.designator))); continue; } fstype = "DM_verity_hash"; rw = false; } else if (IN_SET(type.designator, PARTITION_ROOT_VERITY_SIG, PARTITION_USR_VERITY_SIG)) { if (verity && iovec_is_set(&verity->root_hash)) { _cleanup_(iovec_done) struct iovec root_hash = {}; r = acquire_sig_for_roothash( fd, (uint64_t) start * 512, (uint64_t) size * 512, &root_hash, /* ret_root_hash_sig= */ NULL); if (r < 0) return r; if (!iovec_equal(&verity->root_hash, &root_hash)) { if (DEBUG_LOGGING) { _cleanup_free_ char *found = NULL, *expected = NULL; found = hexmem(root_hash.iov_base, root_hash.iov_len); expected = hexmem(verity->root_hash.iov_base, verity->root_hash.iov_len); log_debug("Verity root hash in signature JSON data (%s) doesn't match configured hash (%s).", strna(found), strna(expected)); } continue; } } check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO | SD_GPT_FLAG_READ_ONLY); m->has_verity_sig = true; fstype = "verity_hash_signature"; rw = false; } else if (type.designator == PARTITION_SWAP) { check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO); /* Note: we don't set fstype = "swap" here, because we still need to probe if * it might be encrypted (i.e. fstype "crypt_LUKS") or unencrypted * (i.e. fstype "swap"), and the only way to figure that out is via fstype * probing. */ /* We don't have a designator for SD_GPT_LINUX_GENERIC so check the UUID instead. */ } else if (sd_id128_equal(type.uuid, SD_GPT_LINUX_GENERIC)) { if (!image_filter_test(filter, PARTITION_ROOT, label)) continue; check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO | SD_GPT_FLAG_READ_ONLY | SD_GPT_FLAG_GROWFS); if (generic_node) multiple_generic = true; else { generic_nr = nr; generic_rw = !(pflags & SD_GPT_FLAG_READ_ONLY); generic_growfs = FLAGS_SET(pflags, SD_GPT_FLAG_GROWFS); generic_uuid = id; generic_node = TAKE_PTR(node); } } else if (type.designator == PARTITION_VAR) { check_partition_flags(node, pflags, SD_GPT_FLAG_NO_AUTO | SD_GPT_FLAG_READ_ONLY | SD_GPT_FLAG_GROWFS); if (!FLAGS_SET(flags, DISSECT_IMAGE_RELAX_VAR_CHECK)) { sd_id128_t var_uuid; /* For /var we insist that the uuid of the partition matches the * HMAC-SHA256 of the /var GPT partition type uuid, keyed by machine * ID. Why? Unlike the other partitions /var is inherently * installation specific, hence we need to be careful not to mount it * in the wrong installation. By hashing the partition UUID from * /etc/machine-id we can securely bind the partition to the * installation. */ r = sd_id128_get_machine_app_specific(SD_GPT_VAR, &var_uuid); if (r < 0) return r; if (!sd_id128_equal(var_uuid, id)) { log_debug("Found a /var/ partition, but its UUID didn't match our expectations " "(found: " SD_ID128_UUID_FORMAT_STR ", expected: " SD_ID128_UUID_FORMAT_STR "), ignoring.", SD_ID128_FORMAT_VAL(id), SD_ID128_FORMAT_VAL(var_uuid)); continue; } } rw = !(pflags & SD_GPT_FLAG_READ_ONLY); growfs = FLAGS_SET(pflags, SD_GPT_FLAG_GROWFS); } if (type.designator != _PARTITION_DESIGNATOR_INVALID) { _cleanup_free_ char *t = NULL, *o = NULL, *l = NULL, *n = NULL; _cleanup_close_ int mount_node_fd = -EBADF; const char *options = NULL; r = image_policy_may_use(policy, type.designator); if (r < 0) return r; if (r == 0) { /* Policy says: ignore; Remember this fact, so that we later can distinguish between "found but ignored" and "not found at all" */ if (!m->partitions[type.designator].found) m->partitions[type.designator].ignored = true; continue; } if (m->partitions[type.designator].found) { int c; /* For most partition types the first one we see wins. Except for the * rootfs and /usr, where we do a version compare of the label, and * let the newest version win. This permits a simple A/B versioning * scheme in OS images. */ c = compare_arch(type.arch, m->partitions[type.designator].architecture); if (c < 0) /* the arch we already found is better than the one we found now */ continue; if (c == 0 && /* same arch? then go by version in label */ (!partition_designator_is_versioned(type.designator) || strverscmp_improved(label, m->partitions[type.designator].label) <= 0)) continue; dissected_partition_done(m->partitions + type.designator); } if (FLAGS_SET(flags, DISSECT_IMAGE_PIN_PARTITION_DEVICES) && type.designator != PARTITION_SWAP) { mount_node_fd = open_partition(node, /* is_partition= */ true, m->loop); if (mount_node_fd < 0) return mount_node_fd; } r = make_partition_devname(devname, diskseq, nr, flags, &n); if (r < 0) return r; /* Local override via env var or designator type wins */ if (fstype) { t = strdup(fstype); if (!t) return -ENOMEM; } else { r = partition_policy_determine_fstype(policy, type.designator, /* ret_encrypted= */ NULL, &t); if (r < 0) return r; } if (label) { l = strdup(label); if (!l) return -ENOMEM; } options = mount_options_from_designator(mount_options, type.designator); if (options) { o = strdup(options); if (!o) return -ENOMEM; } m->partitions[type.designator] = (DissectedPartition) { .found = true, .partno = nr, .rw = rw, .growfs = growfs, .architecture = type.arch, .node = TAKE_PTR(n), .fstype = TAKE_PTR(t), .label = TAKE_PTR(l), .uuid = id, .mount_options = TAKE_PTR(o), .mount_node_fd = TAKE_FD(mount_node_fd), .offset = (uint64_t) start * 512, .size = (uint64_t) size * 512, .gpt_flags = pflags, .fsmount_fd = -EBADF, }; } } else if (is_mbr) { switch (sym_blkid_partition_get_type(pp)) { case 0x83: /* Linux partition */ if (pflags != 0x80) /* Bootable flag */ continue; if (!image_filter_test(filter, PARTITION_ROOT, /* label= */ NULL)) continue; if (generic_node) multiple_generic = true; else { generic_nr = nr; generic_rw = true; generic_growfs = false; generic_node = TAKE_PTR(node); } break; case 0xEA: { /* Boot Loader Spec extended $BOOT partition */ _cleanup_close_ int mount_node_fd = -EBADF; _cleanup_free_ char *o = NULL, *n = NULL; sd_id128_t id = SD_ID128_NULL; const char *options = NULL; if (!image_filter_test(filter, PARTITION_XBOOTLDR, /* label= */ NULL)) continue; r = image_policy_may_use(policy, PARTITION_XBOOTLDR); if (r < 0) return r; if (r == 0) { /* policy says: ignore */ if (!m->partitions[PARTITION_XBOOTLDR].found) m->partitions[PARTITION_XBOOTLDR].ignored = true; continue; } /* First one wins */ if (m->partitions[PARTITION_XBOOTLDR].found) continue; if (FLAGS_SET(flags, DISSECT_IMAGE_PIN_PARTITION_DEVICES)) { mount_node_fd = open_partition(node, /* is_partition= */ true, m->loop); if (mount_node_fd < 0) return mount_node_fd; } (void) blkid_partition_get_uuid_id128(pp, &id); r = make_partition_devname(devname, diskseq, nr, flags, &n); if (r < 0) return r; options = mount_options_from_designator(mount_options, PARTITION_XBOOTLDR); if (options) { o = strdup(options); if (!o) return -ENOMEM; } m->partitions[PARTITION_XBOOTLDR] = (DissectedPartition) { .found = true, .partno = nr, .rw = true, .growfs = false, .architecture = _ARCHITECTURE_INVALID, .node = TAKE_PTR(n), .uuid = id, .mount_options = TAKE_PTR(o), .mount_node_fd = TAKE_FD(mount_node_fd), .offset = (uint64_t) start * 512, .size = (uint64_t) size * 512, .fsmount_fd = -EBADF, }; break; }} } } /* Verity found but no matching data partition? Something is off, refuse. */ FOREACH_ELEMENT(dd, ((const PartitionDesignator[]) { PARTITION_ROOT, PARTITION_USR })) { PartitionDesignator dv = partition_verity_hash_of(*dd); PartitionDesignator ds = partition_verity_sig_of(*dd); /* Hint to help static analyzers */ assert(dv >= 0); assert(ds >= 0); if (!m->partitions[*dd].found && (m->partitions[dv].found || m->partitions[ds].found)) return log_debug_errno( SYNTHETIC_ERRNO(EADDRNOTAVAIL), "Found %s verity hash partition without matching %s data partition.", partition_designator_to_string(*dd), partition_designator_to_string(*dd)); if (m->partitions[ds].found && !m->partitions[dv].found) return log_debug_errno( SYNTHETIC_ERRNO(EADDRNOTAVAIL), "Found %s verity signature partition without matching %s verity hash partition.", partition_designator_to_string(*dd), partition_designator_to_string(*dd)); } /* If root and /usr are combined then insist that the architecture matches */ if (m->partitions[PARTITION_ROOT].found && m->partitions[PARTITION_USR].found && (m->partitions[PARTITION_ROOT].architecture >= 0 && m->partitions[PARTITION_USR].architecture >= 0 && m->partitions[PARTITION_ROOT].architecture != m->partitions[PARTITION_USR].architecture)) return log_debug_errno(SYNTHETIC_ERRNO(EREMOTE), "Found root and usr partitions with different architectures (%s vs %s).", architecture_to_string(m->partitions[PARTITION_ROOT].architecture), architecture_to_string(m->partitions[PARTITION_USR].architecture)); if (!m->partitions[PARTITION_ROOT].found && !m->partitions[PARTITION_USR].found && (flags & DISSECT_IMAGE_GENERIC_ROOT) && (!verity || !iovec_is_set(&verity->root_hash) || verity->designator != PARTITION_USR)) { /* OK, we found nothing usable, then check if there's a single generic partition, and use * that. If the root hash was set however, then we won't fall back to a generic node, because * the root hash decides. */ /* If we didn't find a properly marked root partition, but we did find a single suitable * generic Linux partition, then use this as root partition, if the caller asked for it. */ if (multiple_generic) return -ENOTUNIQ; /* If we didn't find a generic node, then we can't fix this up either */ if (generic_node) { r = image_policy_may_use(policy, PARTITION_ROOT); if (r < 0) return r; if (r == 0) /* Policy says: ignore; remember that we did */ m->partitions[PARTITION_ROOT].ignored = true; else { _cleanup_close_ int mount_node_fd = -EBADF; _cleanup_free_ char *o = NULL, *n = NULL; const char *options; if (FLAGS_SET(flags, DISSECT_IMAGE_PIN_PARTITION_DEVICES)) { mount_node_fd = open_partition(generic_node, /* is_partition= */ true, m->loop); if (mount_node_fd < 0) return mount_node_fd; } r = make_partition_devname(devname, diskseq, generic_nr, flags, &n); if (r < 0) return r; options = mount_options_from_designator(mount_options, PARTITION_ROOT); if (options) { o = strdup(options); if (!o) return -ENOMEM; } assert(generic_nr >= 0); m->partitions[PARTITION_ROOT] = (DissectedPartition) { .found = true, .rw = generic_rw, .growfs = generic_growfs, .partno = generic_nr, .architecture = _ARCHITECTURE_INVALID, .node = TAKE_PTR(n), .uuid = generic_uuid, .mount_options = TAKE_PTR(o), .mount_node_fd = TAKE_FD(mount_node_fd), .offset = UINT64_MAX, .size = UINT64_MAX, .fsmount_fd = -EBADF, }; } } } /* Check if we have a root fs if we are told to do check. /usr alone is fine too, but only if appropriate flag for that is set too */ if (FLAGS_SET(flags, DISSECT_IMAGE_REQUIRE_ROOT) && !(m->partitions[PARTITION_ROOT].found || (m->partitions[PARTITION_USR].found && FLAGS_SET(flags, DISSECT_IMAGE_USR_NO_ROOT)))) return log_debug_errno(SYNTHETIC_ERRNO(ENXIO), "Root or usr partition requested but found neither."); if (m->partitions[PARTITION_ROOT_VERITY].found) { /* We only support one verity partition per image, i.e. can't do for both /usr and root fs */ if (m->partitions[PARTITION_USR_VERITY].found) return log_debug_errno(SYNTHETIC_ERRNO(ENOTUNIQ), "Found both root and usr verity enabled partitions which is not supported."); /* We don't support verity enabled root with a split out /usr. Neither with nor without * verity there. (Note that we do support verity-less root with verity-full /usr, though.) */ if (m->partitions[PARTITION_USR].found) return log_debug_errno(SYNTHETIC_ERRNO(EADDRNOTAVAIL), "Found verity enabled root partition with split usr partition which is not supported."); } if (verity) { /* If a verity designator is specified, then insist that the matching partition exists */ if (verity->designator >= 0 && !m->partitions[verity->designator].found) return log_debug_errno( SYNTHETIC_ERRNO(EADDRNOTAVAIL), "Explicit %s verity designator was specified but did not find %s partition.", partition_designator_to_string(verity->designator), partition_designator_to_string(verity->designator)); if (iovec_is_set(&verity->root_hash)) { /* If we have an explicit root hash and found the partitions for it, then we are ready to use * Verity, set things up for it */ PartitionDesignator d = verity->designator < 0 || verity->designator == PARTITION_ROOT ? PARTITION_ROOT : PARTITION_USR; PartitionDesignator dv = partition_verity_hash_of(d); assert(dv >= 0); if (!m->partitions[d].found) return log_debug_errno( SYNTHETIC_ERRNO(EADDRNOTAVAIL), "Verity enabled %s partition was requested but did not find a %s data partition.", partition_designator_to_string(d), partition_designator_to_string(d)); if (!m->partitions[dv].found) return log_debug_errno( SYNTHETIC_ERRNO(EADDRNOTAVAIL), "Verity enabled %s partition was requested but did not find a %s verity hash partition.", partition_designator_to_string(d), partition_designator_to_string(d)); /* If we found a verity setup, then the data partition is necessarily read-only. */ m->partitions[d].rw = false; m->verity_ready = true; if (iovec_is_set(&verity->root_hash_sig)) m->verity_sig_ready = true; } } bool any = false; /* After we discovered all partitions let's see if the verity requirements match the policy. (Note: * we don't check encryption requirements here, because we haven't probed the file system yet, hence * don't know if this is encrypted or not) */ for (PartitionDesignator di = 0; di < _PARTITION_DESIGNATOR_MAX; di++) { any = any || m->partitions[di].found; /* Determine the verity protection level for this partition. */ PartitionPolicyFlags found_flags; if (m->partitions[di].found) { found_flags = PARTITION_POLICY_ENCRYPTED|PARTITION_POLICY_ENCRYPTEDWITHINTEGRITY|PARTITION_POLICY_UNPROTECTED|PARTITION_POLICY_UNUSED; PartitionDesignator vi = partition_verity_hash_of(di); if (vi >= 0 && m->partitions[vi].found) { found_flags |= PARTITION_POLICY_VERITY; PartitionDesignator si = partition_verity_sig_of(di); if (si >= 0 && m->partitions[si].found) found_flags |= PARTITION_POLICY_SIGNED; } } else found_flags = m->partitions[di].ignored ? PARTITION_POLICY_UNUSED : PARTITION_POLICY_ABSENT; if (DEBUG_LOGGING) { _cleanup_free_ char *s = NULL; (void) partition_policy_flags_to_string(found_flags, /* simplify= */ false, &s); log_debug("Found for designator %s: %s.", partition_designator_to_string(di), strna(s)); } r = image_policy_check_protection(policy, di, found_flags); if (r < 0) return r; if (m->partitions[di].found) { r = image_policy_check_partition_flags(policy, di, m->partitions[di].gpt_flags); if (r < 0) return r; } } if (!any && !FLAGS_SET(flags, DISSECT_IMAGE_ALLOW_EMPTY)) return -ENOMSG; r = dissected_image_probe_filesystems(m, fd, policy); if (r < 0) return r; return 0; } #endif int dissected_image_new_from_existing_verity( const char *src, const VeritySettings *verity, const MountOptions *options, const ImagePolicy *image_policy, const ImageFilter *image_filter, RuntimeScope runtime_scope, DissectImageFlags dissect_image_flags, DissectedImage **ret) { /* Look for an already set up dm-verity device with a single filesystem, according to our naming * scheme and image policy, and if it is pinned by filesystem type set up the image directly. */ #if HAVE_BLKID _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; _cleanup_free_ char *node = NULL, *root_hash_encoded = NULL, *root_fstype_string = NULL; _cleanup_close_ int mount_node_fd = -EBADF; PartitionPolicyFlags found_flags; bool encrypted = false; int r; assert(!verity || verity->designator < 0 || IN_SET(verity->designator, PARTITION_ROOT, PARTITION_USR)); assert(!verity || iovec_is_valid(&verity->root_hash)); assert(!verity || iovec_is_valid(&verity->root_hash_sig)); assert(!verity || iovec_is_set(&verity->root_hash) || !iovec_is_set(&verity->root_hash_sig)); assert(ret); /* Shortcut: this deals only with verity images and requires a policy, and only for system services */ if (runtime_scope != RUNTIME_SCOPE_SYSTEM || !FLAGS_SET(dissect_image_flags, DISSECT_IMAGE_VERITY_SHARE) || !image_policy || !verity || !verity->data_path || (verity->designator >= 0 && verity->designator != PARTITION_ROOT) || !iovec_is_set(&verity->root_hash)) return -ENOPKG; /* The policy fstype flags translate to the literal fstype name of each filesystem. * Input must be a single filesystem image, if the policy specifies more than one, we need to dissect */ r = partition_policy_determine_fstype(image_policy, PARTITION_ROOT, &encrypted, &root_fstype_string); if (r < 0) return r; if (!root_fstype_string) return -ENOPKG; if (verity_settings_data_covers(verity, PARTITION_ROOT)) found_flags = iovec_is_set(&verity->root_hash_sig) ? PARTITION_POLICY_SIGNED : PARTITION_POLICY_VERITY; else found_flags = encrypted ? PARTITION_POLICY_ENCRYPTED : PARTITION_POLICY_UNPROTECTED; root_hash_encoded = hexmem(verity->root_hash.iov_base, verity->root_hash.iov_len); if (!root_hash_encoded) return -ENOMEM; node = strjoin("/dev/mapper/", root_hash_encoded, "-verity"); if (!node) return -ENOMEM; r = dissected_image_new(src, &dissected_image); if (r < 0) return r; mount_node_fd = open_partition(node, /* is_partition= */ false, /* loop= */ NULL); if (mount_node_fd < 0) return mount_node_fd; r = dissect_image_from_unpartitioned( node, /* diskseq= */ 0, /* uuid= */ NULL, encrypted, verity, options, image_policy, image_filter, &mount_node_fd, &root_fstype_string, dissected_image, dissect_image_flags, found_flags); if (r < 0) return r; *ret = TAKE_PTR(dissected_image); return 0; #else return -EOPNOTSUPP; #endif } int dissect_image_file( const char *path, const VeritySettings *verity, const MountOptions *mount_options, const ImagePolicy *image_policy, const ImageFilter *image_filter, DissectImageFlags flags, DissectedImage **ret) { #if HAVE_BLKID _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; _cleanup_close_ int fd = -EBADF; struct stat st; int r; assert(path); fd = open(path, O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_NOCTTY); if (fd < 0) return -errno; if (fstat(fd, &st) < 0) return -errno; r = stat_verify_regular(&st); if (r < 0) return r; r = dissected_image_new(path, &m); if (r < 0) return r; m->image_size = st.st_size; r = probe_sector_size(fd, &m->sector_size); if (r < 0) return r; r = dissect_image(m, fd, path, verity, mount_options, image_policy, image_filter, flags); if (r < 0) return r; if (ret) *ret = TAKE_PTR(m); return 0; #else return -EOPNOTSUPP; #endif } int dissect_log_error(int log_level, int r, const char *name, const VeritySettings *verity) { assert(log_level >= 0 && log_level <= LOG_DEBUG); assert(name); switch (r) { case 0 ... INT_MAX: /* success! */ return r; case -EOPNOTSUPP: return log_full_errno(log_level, r, "Dissecting images is not supported, compiled without blkid support."); case -ENOPKG: return log_full_errno(log_level, r, "%s: Couldn't identify a suitable partition table or file system.", name); case -ENOMEDIUM: return log_full_errno(log_level, r, "%s: The image does not pass os-release/extension-release validation.", name); case -EADDRNOTAVAIL: return log_full_errno(log_level, r, "%s: No root/usr partition for specified root/usr hash found.", name); case -EREMOTE: return log_full_errno(log_level, r, "%s: Found root and usr partitions with different architectures", name); case -ENOTUNIQ: return log_full_errno(log_level, r, "%s: Multiple suitable root partitions found in image.", name); case -ENXIO: return log_full_errno(log_level, r, "%s: No suitable root partition found in image.", name); case -EPROTONOSUPPORT: return log_full_errno(log_level, r, "Device '%s' is a loopback block device with partition scanning turned off, please turn it on.", name); case -ENOTBLK: return log_full_errno(log_level, r, "%s: Image is not a block device.", name); case -EBADR: return log_full_errno(log_level, r, "Combining partitioned images (such as '%s') with external Verity data (such as '%s') not supported. " "(Consider setting $SYSTEMD_DISSECT_VERITY_SIDECAR=0 to disable automatic discovery of external Verity data.)", name, strna(verity ? verity->data_path : NULL)); case -ERFKILL: return log_full_errno(log_level, r, "%s: Image does not match image policy.", name); case -ENOMSG: return log_full_errno(log_level, r, "%s: No suitable partitions found.", name); case -EUCLEAN: return log_full_errno(log_level, r, "%s: Partition with ambiguous file system superblock signature found.", name); default: return log_full_errno(log_level, r, "%s: Cannot dissect image: %m", name); } } int dissect_image_file_and_warn( const char *path, const VeritySettings *verity, const MountOptions *mount_options, const ImagePolicy *image_policy, const ImageFilter *image_filter, DissectImageFlags flags, DissectedImage **ret) { return dissect_log_error( LOG_ERR, dissect_image_file(path, verity, mount_options, image_policy, image_filter, flags, ret), path, verity); } void dissected_image_close(DissectedImage *m) { if (!m) return; /* Closes all fds we keep open associated with this, but nothing else */ FOREACH_ARRAY(p, m->partitions, _PARTITION_DESIGNATOR_MAX) { p->mount_node_fd = safe_close(p->mount_node_fd); p->fsmount_fd = safe_close(p->fsmount_fd); } m->loop = loop_device_unref(m->loop); } DissectedImage* dissected_image_unref(DissectedImage *m) { if (!m) return NULL; /* First, clear dissected partitions. */ for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) dissected_partition_done(m->partitions + i); /* Second, free decrypted images. This must be after dissected_partition_done(), as freeing * DecryptedImage may try to deactivate partitions. */ decrypted_image_unref(m->decrypted_image); /* Third, unref LoopDevice. This must be called after the above two, as freeing LoopDevice may try to * remove existing partitions on the loopback block device. */ loop_device_unref(m->loop); free(m->image_name); free(m->hostname); strv_free(m->machine_info); strv_free(m->os_release); strv_free(m->initrd_release); strv_free(m->confext_release); strv_free(m->sysext_release); return mfree(m); } static int is_loop_device(const char *path) { char s[SYS_BLOCK_PATH_MAX("/../loop/")]; struct stat st; int r; assert(path); if (stat(path, &st) < 0) return -errno; r = stat_verify_block(&st); if (r < 0) return r; xsprintf_sys_block_path(s, "/loop/", st.st_dev); if (access(s, F_OK) < 0) { if (errno != ENOENT) return -errno; /* The device itself isn't a loop device, but maybe it's a partition and its parent is? */ xsprintf_sys_block_path(s, "/../loop/", st.st_dev); if (access(s, F_OK) < 0) return errno == ENOENT ? false : -errno; } return true; } static int run_fsck(int node_fd, const char *fstype) { int r, exit_status; assert(node_fd >= 0); assert(fstype); r = fsck_exists_for_fstype(fstype); if (r < 0) { log_debug_errno(r, "Couldn't determine whether fsck for %s exists, proceeding anyway.", fstype); return 0; } if (r == 0) { log_debug("Not checking partition %s, as fsck for %s does not exist.", FORMAT_PROC_FD_PATH(node_fd), fstype); return 0; } _cleanup_(pidref_done) PidRef pidref = PIDREF_NULL; r = pidref_safe_fork_full( "(fsck)", NULL, &node_fd, 1, /* Leave the node fd open */ FORK_RESET_SIGNALS|FORK_CLOSE_ALL_FDS|FORK_RLIMIT_NOFILE_SAFE|FORK_DEATHSIG_SIGTERM|FORK_REARRANGE_STDIO|FORK_CLOEXEC_OFF, &pidref); if (r < 0) return log_debug_errno(r, "Failed to fork off fsck: %m"); if (r == 0) { /* Child */ execlp("fsck", "fsck", "-aT", FORMAT_PROC_FD_PATH(node_fd), NULL); log_open(); log_debug_errno(errno, "Failed to execl() fsck: %m"); _exit(FSCK_OPERATIONAL_ERROR); } exit_status = pidref_wait_for_terminate_and_check("fsck", &pidref, 0); if (exit_status < 0) return log_debug_errno(exit_status, "Failed to fork off fsck: %m"); if ((exit_status & ~FSCK_ERROR_CORRECTED) != FSCK_SUCCESS) { log_debug("fsck failed with exit status %i.", exit_status); if ((exit_status & (FSCK_SYSTEM_SHOULD_REBOOT|FSCK_ERRORS_LEFT_UNCORRECTED)) != 0) return log_debug_errno(SYNTHETIC_ERRNO(EUCLEAN), "File system is corrupted, refusing."); log_debug("Ignoring fsck error."); } return 0; } static int fs_grow(const char *node_path, int mount_fd, const char *mount_path) { _cleanup_close_ int _mount_fd = -EBADF, node_fd = -EBADF; uint64_t size, newsize; const char *id; int r; assert(node_path); assert(mount_fd >= 0 || mount_path); node_fd = open(node_path, O_RDONLY|O_CLOEXEC|O_NONBLOCK|O_NOCTTY); if (node_fd < 0) return log_debug_errno(errno, "Failed to open node device %s: %m", node_path); r = blockdev_get_device_size(node_fd, &size); if (r < 0) return log_debug_errno(r, "Failed to get block device size of %s: %m", node_path); if (mount_fd < 0) { assert(mount_path); _mount_fd = open(mount_path, O_RDONLY|O_DIRECTORY|O_CLOEXEC); if (_mount_fd < 0) return log_debug_errno(errno, "Failed to open mounted file system %s: %m", mount_path); mount_fd = _mount_fd; } else { mount_fd = fd_reopen_condition(mount_fd, O_RDONLY|O_DIRECTORY|O_CLOEXEC, O_RDONLY|O_DIRECTORY|O_CLOEXEC, &_mount_fd); if (mount_fd < 0) return log_debug_errno(errno, "Failed to reopen mount node: %m"); } id = mount_path ?: node_path; log_debug("Resizing \"%s\" to %"PRIu64" bytes...", id, size); r = resize_fs(mount_fd, size, &newsize); if (r < 0) return log_debug_errno(r, "Failed to resize \"%s\" to %"PRIu64" bytes: %m", id, size); if (newsize == size) log_debug("Successfully resized \"%s\" to %s bytes.", id, FORMAT_BYTES(newsize)); else { assert(newsize < size); log_debug("Successfully resized \"%s\" to %s bytes (%"PRIu64" bytes lost due to blocksize).", id, FORMAT_BYTES(newsize), size - newsize); } return 0; } int partition_pick_mount_options( PartitionDesignator d, const char *fstype, bool rw, bool discard, char **ret_options, unsigned long *ret_ms_flags) { _cleanup_free_ char *options = NULL; assert(ret_options); /* Selects a baseline of bind mount flags, that should always apply. * * Firstly, we set MS_NODEV universally on all mounts, since we don't want to allow device nodes outside of /dev/. * * On /var/tmp/ we'll also set MS_NOSUID, same as we set for /tmp/ on the host. * * On the ESP and XBOOTLDR partitions we'll also disable symlinks, and execution. These file systems * are generally untrusted (i.e. not encrypted or authenticated), and typically VFAT hence we should * be as restrictive as possible, and this shouldn't hurt, since the functionality is not available * there anyway. */ unsigned long flags = MS_NODEV; if (!rw) flags |= MS_RDONLY; switch (d) { case PARTITION_ESP: case PARTITION_XBOOTLDR: flags |= MS_NOSUID|MS_NOEXEC|MS_NOSYMFOLLOW; /* The ESP might contain a pre-boot random seed. Let's make this inaccessible to regular * userspace. ESP/XBOOTLDR is almost certainly VFAT, hence if we don't know assume it is. */ if (!fstype || fstype_can_fmask_dmask(fstype)) if (!strextend_with_separator(&options, ",", "fmask=0177,dmask=0077")) return -ENOMEM; break; case PARTITION_TMP: flags |= MS_NOSUID; break; default: ; } /* So, when you request MS_RDONLY from ext4, then this means nothing. It happily still writes to the * backing storage. What's worse, the BLKRO[GS]ET flag and (in case of loopback devices) * LO_FLAGS_READ_ONLY don't mean anything, they affect userspace accesses only, and write accesses * from the upper file system still get propagated through to the underlying file system, * unrestricted. To actually get ext4/xfs/btrfs to stop writing to the device we need to specify * "norecovery" as mount option, in addition to MS_RDONLY. Yes, this sucks, since it means we need to * carry a per file system table here. * * Note that this means that we might not be able to mount corrupted file systems as read-only * anymore (since in some cases the kernel implementations will refuse mounting when corrupted, * read-only and "norecovery" is specified). But I think for the case of automatically determined * mount options for loopback devices this is the right choice, since otherwise using the same * loopback file twice even in read-only mode, is going to fail badly sooner or later. The use case of * making reuse of the immutable images "just work" is more relevant to us than having read-only * access that actually modifies stuff work on such image files. Or to say this differently: if * people want their file systems to be fixed up they should just open them in writable mode, where * all these problems don't exist. */ if (!rw && fstype) { const char *option = fstype_norecovery_option(fstype); if (option && !strextend_with_separator(&options, ",", option)) return -ENOMEM; } if (discard && fstype && fstype_can_discard(fstype)) if (!strextend_with_separator(&options, ",", "discard")) return -ENOMEM; if (!ret_ms_flags) /* Fold flags into option string if ret_flags specified as NULL */ if (!strextend_with_separator(&options, ",", FLAGS_SET(flags, MS_RDONLY) ? "ro" : "rw", FLAGS_SET(flags, MS_NODEV) ? "nodev" : "dev", FLAGS_SET(flags, MS_NOSUID) ? "nosuid" : "suid", FLAGS_SET(flags, MS_NOEXEC) ? "noexec" : "exec", FLAGS_SET(flags, MS_NOSYMFOLLOW) ? "nosymfollow" : NULL)) /* NB: we suppress 'symfollow' here, since it's the default, and old /bin/mount might not know it */ return -ENOMEM; if (ret_ms_flags) *ret_ms_flags = flags; *ret_options = TAKE_PTR(options); return 0; } static bool need_user_mapping(uid_t uid_shift, uid_t uid_range) { if (!uid_is_valid(uid_shift)) return false; return uid_shift != 0 || uid_range != UINT32_MAX; } static int mount_partition( PartitionDesignator d, DissectedPartition *m, const char *where, const char *directory, uid_t uid_shift, uid_t uid_range, int userns_fd, DissectImageFlags flags) { _cleanup_free_ char *chased = NULL, *options = NULL; const char *p = NULL, *node, *fstype = NULL; bool rw, discard, grow; unsigned long ms_flags; int r; assert(m); if (!m->found) return 0; /* Check the various combinations when we can't do anything anymore */ if (m->fsmount_fd < 0 && m->mount_node_fd < 0) return 0; if (m->fsmount_fd >= 0 && !where) return 0; if (!where && m->mount_node_fd < 0) return 0; if (m->fsmount_fd < 0) { fstype = dissected_partition_fstype(m); if (!fstype) return -EAFNOSUPPORT; /* We are looking at an encrypted partition? This either means stacked encryption, or the * caller didn't call dissected_image_decrypt() beforehand. Let's return a recognizable error * for this case. */ if (streq(fstype, "crypto_LUKS")) return -EUNATCH; r = dissect_fstype_ok(fstype); if (r < 0) return r; if (!r) return -EIDRM; /* Recognizable error */ } node = m->mount_node_fd < 0 ? NULL : FORMAT_PROC_FD_PATH(m->mount_node_fd); rw = m->rw && !(flags & DISSECT_IMAGE_MOUNT_READ_ONLY); discard = ((flags & DISSECT_IMAGE_DISCARD) || ((flags & DISSECT_IMAGE_DISCARD_ON_LOOP) && (m->node && is_loop_device(m->node) > 0))); grow = rw && m->growfs && FLAGS_SET(flags, DISSECT_IMAGE_GROWFS); if (FLAGS_SET(flags, DISSECT_IMAGE_FSCK) && rw && m->mount_node_fd >= 0 && m->fsmount_fd < 0) { r = run_fsck(m->mount_node_fd, fstype); if (r < 0) return r; } if (where) { if (directory) { /* Automatically create missing mount points inside the image, if necessary. */ r = mkdir_p_root(where, directory, uid_shift, (gid_t) uid_shift, 0755); if (r < 0 && r != -EROFS) return r; r = chase(directory, where, CHASE_PREFIX_ROOT, &chased, NULL); if (r < 0) return r; p = chased; } else { /* Create top-level mount if missing – but only if this is asked for. This won't modify the * image (as the branch above does) but the host hierarchy, and the created directory might * survive our mount in the host hierarchy hence. */ if (FLAGS_SET(flags, DISSECT_IMAGE_MKDIR)) { r = mkdir_p(where, 0755); if (r < 0) return r; } p = where; } } if (m->fsmount_fd < 0) { r = partition_pick_mount_options(d, fstype, rw, discard, &options, &ms_flags); if (r < 0) return r; if (need_user_mapping(uid_shift, uid_range) && fstype_can_uid_gid(fstype)) { _cleanup_free_ char *uid_option = NULL; if (asprintf(&uid_option, "uid=" UID_FMT ",gid=" GID_FMT, uid_shift, (gid_t) uid_shift) < 0) return -ENOMEM; if (!strextend_with_separator(&options, ",", uid_option)) return -ENOMEM; userns_fd = -EBADF; /* Not needed */ } if (!isempty(m->mount_options)) if (!strextend_with_separator(&options, ",", m->mount_options)) return -ENOMEM; } if (p) { if (m->fsmount_fd >= 0) { /* Case #1: Attach existing fsmount fd to the file system */ r = mount_exchange_graceful( m->fsmount_fd, p, FLAGS_SET(flags, DISSECT_IMAGE_TRY_ATOMIC_MOUNT_EXCHANGE)); if (r < 0) return log_debug_errno(r, "Failed to mount image on '%s': %m", p); } else { assert(node); /* Case #2: Mount directly into place */ r = mount_nofollow_verbose(LOG_DEBUG, node, p, fstype, ms_flags, options); if (r < 0) return r; if (grow) (void) fs_grow(node, -EBADF, p); if (userns_fd >= 0) { r = remount_idmap_fd(STRV_MAKE(p), userns_fd, /* extra_mount_attr_set= */ 0); if (r < 0) return r; } } } else { assert(node); /* Case #3: Create fsmount fd */ m->fsmount_fd = make_fsmount(LOG_DEBUG, node, fstype, ms_flags, options, userns_fd); if (m->fsmount_fd < 0) return m->fsmount_fd; if (grow) (void) fs_grow(node, m->fsmount_fd, NULL); } return 1; } static int mount_root_tmpfs(const char *where, uid_t uid_shift, uid_t uid_range, DissectImageFlags flags) { _cleanup_free_ char *options = NULL; int r; assert(where); /* For images that contain /usr/ but no rootfs, let's mount rootfs as tmpfs */ if (FLAGS_SET(flags, DISSECT_IMAGE_MKDIR)) { r = mkdir_p(where, 0755); if (r < 0) return r; } if (need_user_mapping(uid_shift, uid_range)) { if (asprintf(&options, "uid=" UID_FMT ",gid=" GID_FMT, uid_shift, (gid_t) uid_shift) < 0) return -ENOMEM; } r = mount_nofollow_verbose(LOG_DEBUG, "rootfs", where, "tmpfs", MS_NODEV, options); if (r < 0) return r; return 1; } static int mount_point_is_available(const char *where, const char *path, bool missing_ok) { _cleanup_free_ char *p = NULL; int r; /* Check whether is suitable as a mountpoint, i.e. is an empty directory * or does not exist at all (when missing_ok). */ r = chase(path, where, CHASE_PREFIX_ROOT, &p, NULL); if (r == -ENOENT) return missing_ok; if (r < 0) return log_debug_errno(r, "Failed to chase \"%s\": %m", path); r = dir_is_empty(p, /* ignore_hidden_or_backup= */ false); if (r == -ENOTDIR) return false; if (r < 0) return log_debug_errno(r, "Failed to check directory \"%s\": %m", p); return r > 0; } int dissected_image_mount( DissectedImage *m, const char *where, uid_t uid_shift, uid_t uid_range, int userns_fd, DissectImageFlags flags) { _cleanup_close_ int my_userns_fd = -EBADF; int r; assert(m); if (FLAGS_SET(flags, DISSECT_IMAGE_FOREIGN_UID)) /* For image based mounts we currently require an identity mapping */ return -EOPNOTSUPP; /* If 'where' is NULL then we'll use the new mount API to create fsmount() fds for the mounts and * store them in DissectedPartition.fsmount_fd. * * If 'where' is not NULL then we'll either mount the partitions to the right places ourselves, * or use DissectedPartition.fsmount_fd and bind it to the right places. * * This allows splitting the setting up the superblocks and the binding to file systems paths into * two distinct and differently privileged components: one that gets the fsmount fds, and the other * that then applies them. * * Returns: * * -ENXIO → No root partition found * -EMEDIUMTYPE → DISSECT_IMAGE_VALIDATE_OS set but no os-release/extension-release file found * -EUNATCH → Encrypted partition found for which no dm-crypt was set up yet * -EUCLEAN → fsck for file system failed * -EBUSY → File system already mounted/used elsewhere (kernel) * -EAFNOSUPPORT → File system type not supported or not known * -EIDRM → File system is not among allowlisted "common" file systems */ if (!where && (flags & (DISSECT_IMAGE_VALIDATE_OS|DISSECT_IMAGE_VALIDATE_OS_EXT)) != 0) return -EOPNOTSUPP; /* for now, not supported */ if (!(m->partitions[PARTITION_ROOT].found || (m->partitions[PARTITION_USR].found && FLAGS_SET(flags, DISSECT_IMAGE_USR_NO_ROOT)))) return -ENXIO; /* Require a root fs or at least a /usr/ fs (the latter is subject to a flag of its own) */ if (userns_fd < 0 && need_user_mapping(uid_shift, uid_range) && FLAGS_SET(flags, DISSECT_IMAGE_MOUNT_IDMAPPED)) { my_userns_fd = make_userns(uid_shift, uid_range, UID_INVALID, UID_INVALID, REMOUNT_IDMAPPING_HOST_ROOT); if (my_userns_fd < 0) return my_userns_fd; userns_fd = my_userns_fd; } if ((flags & DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY) == 0) { /* First mount the root fs. If there's none we use a tmpfs. */ if (m->partitions[PARTITION_ROOT].found) { r = mount_partition(PARTITION_ROOT, m->partitions + PARTITION_ROOT, where, NULL, uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; } else if (where) { r = mount_root_tmpfs(where, uid_shift, uid_range, flags); if (r < 0) return r; } /* For us mounting root always means mounting /usr as well */ r = mount_partition(PARTITION_USR, m->partitions + PARTITION_USR, where, "/usr", uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; } if ((flags & DISSECT_IMAGE_MOUNT_NON_ROOT_ONLY) == 0 && (flags & (DISSECT_IMAGE_VALIDATE_OS|DISSECT_IMAGE_VALIDATE_OS_EXT)) != 0) { /* If either one of the validation flags are set, ensure that the image qualifies as * one or the other (or both). */ bool ok = false; assert(where); if (FLAGS_SET(flags, DISSECT_IMAGE_VALIDATE_OS)) { log_debug("Checking if '%s' is an OS tree", where); r = path_is_os_tree(where); if (r < 0) return log_debug_errno(r, "Failed to check is '%s' is an OS tree: %m", where); if (r > 0) { log_debug("Successfully identified '%s' as an OS tree", where); ok = true; } } if (!ok && FLAGS_SET(flags, DISSECT_IMAGE_VALIDATE_OS_EXT) && m->image_name) { log_debug("Checking if '%s' is an extension tree", where); r = extension_has_forbidden_content(where); if (r < 0) return log_debug_errno(r, "Failed to check if '%s' contains content forbidden for an extension image: %m", where); if (r == 0) { ImageClass class = IMAGE_SYSEXT; r = path_is_extension_tree(class, where, m->image_name, FLAGS_SET(flags, DISSECT_IMAGE_RELAX_EXTENSION_CHECK)); if (r == 0) { class = IMAGE_CONFEXT; r = path_is_extension_tree(class, where, m->image_name, FLAGS_SET(flags, DISSECT_IMAGE_RELAX_EXTENSION_CHECK)); } if (r < 0) return log_debug_errno(r, "Failed to check if '%s' is an extension tree: %m", where); if (r > 0) { log_debug("Successfully identified '%s' as a %s extension tree", where, image_class_to_string(class)); ok = true; } } } if (!ok) return -ENOMEDIUM; } if (flags & DISSECT_IMAGE_MOUNT_ROOT_ONLY) return 0; r = mount_partition(PARTITION_HOME, m->partitions + PARTITION_HOME, where, "/home", uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; r = mount_partition(PARTITION_SRV, m->partitions + PARTITION_SRV, where, "/srv", uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; r = mount_partition(PARTITION_VAR, m->partitions + PARTITION_VAR, where, "/var", uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; r = mount_partition(PARTITION_TMP, m->partitions + PARTITION_TMP, where, "/var/tmp", uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; int slash_boot_is_available = 0; if (where) { r = slash_boot_is_available = mount_point_is_available(where, "/boot", /* missing_ok= */ true); if (r < 0) return r; } if (!where || slash_boot_is_available) { r = mount_partition(PARTITION_XBOOTLDR, m->partitions + PARTITION_XBOOTLDR, where, "/boot", uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; slash_boot_is_available = !r; } if (m->partitions[PARTITION_ESP].found) { const char *esp_path = NULL; if (where) { /* Mount the ESP to /boot/ if it exists and is empty and we didn't already mount the * XBOOTLDR partition into it. Otherwise, use /efi instead, but only if it exists * and is empty. */ if (slash_boot_is_available) { r = mount_point_is_available(where, "/boot", /* missing_ok= */ false); if (r < 0) return r; if (r > 0) esp_path = "/boot"; } if (!esp_path) { r = mount_point_is_available(where, "/efi", /* missing_ok= */ true); if (r < 0) return r; if (r > 0) esp_path = "/efi"; } } /* OK, let's mount the ESP now (possibly creating the dir if missing) */ r = mount_partition(PARTITION_ESP, m->partitions + PARTITION_ESP, where, esp_path, uid_shift, uid_range, userns_fd, flags); if (r < 0) return r; } return 0; } int dissected_image_mount_and_warn( DissectedImage *m, const char *where, uid_t uid_shift, uid_t uid_range, int userns_fd, DissectImageFlags flags) { int r; assert(m); r = dissected_image_mount(m, where, uid_shift, uid_range, userns_fd, flags); if (r == -ENXIO) return log_error_errno(r, "Failed to mount image: No root file system found in image."); if (r == -EMEDIUMTYPE) return log_error_errno(r, "Failed to mount image: No suitable os-release/extension-release file in image found."); if (r == -EUNATCH) return log_error_errno(r, "Failed to mount image: Encrypted file system discovered, but decryption not requested."); if (r == -EUCLEAN) return log_error_errno(r, "Failed to mount image: File system check on image failed."); if (r == -EBUSY) return log_error_errno(r, "Failed to mount image: File system already mounted elsewhere."); if (r == -EAFNOSUPPORT) return log_error_errno(r, "Failed to mount image: File system type not supported or not known."); if (r == -EIDRM) return log_error_errno(r, "Failed to mount image: File system is too uncommon, refused."); if (r < 0) return log_error_errno(r, "Failed to mount image: %m"); return r; } #if HAVE_LIBCRYPTSETUP struct DecryptedPartition { struct crypt_device *device; char *name; bool relinquished; }; #endif typedef struct DecryptedPartition DecryptedPartition; struct DecryptedImage { unsigned n_ref; DecryptedPartition *decrypted; size_t n_decrypted; }; static DecryptedImage* decrypted_image_free(DecryptedImage *d) { #if HAVE_LIBCRYPTSETUP int r; if (!d) return NULL; for (size_t i = 0; i < d->n_decrypted; i++) { DecryptedPartition *p = d->decrypted + i; if (p->device && p->name && !p->relinquished) { _cleanup_free_ char *node = NULL; node = path_join("/dev/mapper", p->name); if (node) { r = btrfs_forget_device(node); if (r < 0 && r != -ENOENT) log_debug_errno(r, "Failed to forget btrfs device %s, ignoring: %m", node); } else log_oom_debug(); /* Let's deactivate lazily, as the dm volume may be already/still used by other processes. */ r = sym_crypt_deactivate_by_name(p->device, p->name, CRYPT_DEACTIVATE_DEFERRED); if (r < 0) log_debug_errno(r, "Failed to deactivate encrypted partition %s", p->name); } if (p->device) sym_crypt_free(p->device); free(p->name); } free(d->decrypted); free(d); #endif return NULL; } DEFINE_TRIVIAL_REF_UNREF_FUNC(DecryptedImage, decrypted_image, decrypted_image_free); #if HAVE_LIBCRYPTSETUP static int decrypted_image_new(DecryptedImage **ret) { _cleanup_(decrypted_image_unrefp) DecryptedImage *d = NULL; assert(ret); d = new(DecryptedImage, 1); if (!d) return -ENOMEM; *d = (DecryptedImage) { .n_ref = 1, }; *ret = TAKE_PTR(d); return 0; } static uint64_t dissected_image_diskseq(const DissectedImage *di) { assert(di); return di->loop ? di->loop->diskseq : 0; } static int make_dm_name_and_node( const char *base, uint64_t diskseq, const char *suffix, char **ret_name, char **ret_node) { assert(base); assert(suffix); assert(ret_name); assert(ret_node); _cleanup_free_ char *name = NULL; if (diskseq != 0) name = asprintf_safe("%s-%" PRIu64 "%s", base, diskseq, suffix); else name = strjoin(base, suffix); if (!name) return -ENOMEM; if (!filename_is_valid(name)) return -EINVAL; _cleanup_free_ char *node = path_join(sym_crypt_get_dir(), name); if (!node) return -ENOMEM; *ret_name = TAKE_PTR(name); *ret_node = TAKE_PTR(node); return 0; } static int make_dm_name_and_node_from_node( const char *original_node, uint64_t diskseq, const char *suffix, char **ret_name, char **ret_node) { int r; assert(original_node); _cleanup_free_ char *base = NULL; r = path_extract_filename(original_node, &base); if (r < 0) return r; return make_dm_name_and_node(base, diskseq, suffix, ret_name, ret_node); } static int decrypt_partition( DissectedImage *di, DissectedPartition *m, const char *passphrase, DissectImageFlags flags, PartitionPolicyFlags policy_flags, DecryptedImage *d) { _cleanup_free_ char *node = NULL, *name = NULL; _cleanup_(crypt_freep) struct crypt_device *cd = NULL; _cleanup_close_ int fd = -EBADF; int r; assert(di); assert(m); assert(d); if (!m->found || !m->node || !m->fstype) return 0; if (!streq(m->fstype, "crypto_LUKS")) return 0; if (!passphrase) return -ENOKEY; if (!FLAGS_SET(policy_flags, PARTITION_POLICY_ENCRYPTED)) return log_debug_errno(SYNTHETIC_ERRNO(ERFKILL), "Attempted to unlock partition via LUKS, but it's prohibited."); r = dlopen_cryptsetup(LOG_DEBUG); if (r < 0) return r; r = make_dm_name_and_node_from_node(m->node, dissected_image_diskseq(di), "-decrypted", &name, &node); if (r < 0) return r; if (!GREEDY_REALLOC0(d->decrypted, d->n_decrypted + 1)) return -ENOMEM; r = sym_crypt_init(&cd, m->node); if (r < 0) return log_debug_errno(r, "Failed to initialize dm-crypt: %m"); cryptsetup_enable_logging(cd); r = sym_crypt_load(cd, CRYPT_LUKS, NULL); if (r < 0) return log_debug_errno(r, "Failed to load LUKS metadata: %m"); r = sym_crypt_activate_by_passphrase(cd, name, CRYPT_ANY_SLOT, passphrase, strlen(passphrase), ((flags & DISSECT_IMAGE_DEVICE_READ_ONLY) ? CRYPT_ACTIVATE_READONLY : 0) | ((flags & DISSECT_IMAGE_DISCARD_ON_CRYPTO) ? CRYPT_ACTIVATE_ALLOW_DISCARDS : 0)); if (r < 0) { log_debug_errno(r, "Failed to activate LUKS device: %m"); return r == -EPERM ? -EKEYREJECTED : r; } fd = open(node, O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY); if (fd < 0) return log_debug_errno(errno, "Failed to open %s: %m", node); d->decrypted[d->n_decrypted++] = (DecryptedPartition) { .name = TAKE_PTR(name), .device = TAKE_PTR(cd), }; m->decrypted_node = TAKE_PTR(node); close_and_replace(m->mount_node_fd, fd); return 0; } static int verity_can_reuse( const VeritySettings *verity, const char *name, struct crypt_device **ret_cd) { /* If the same volume was already open, check that the root hashes match, and reuse it if they do */ _cleanup_(crypt_freep) struct crypt_device *cd = NULL; struct crypt_params_verity crypt_params = {}; int r; assert(verity); assert(name); assert(ret_cd); r = sym_crypt_init_by_name(&cd, name); if (r < 0) return log_debug_errno(r, "Error opening verity device, crypt_init_by_name failed: %m"); cryptsetup_enable_logging(cd); r = sym_crypt_get_verity_info(cd, &crypt_params); if (r < 0) return log_debug_errno(r, "Error opening verity device, crypt_get_verity_info failed: %m"); _cleanup_(iovec_done) struct iovec root_hash_existing = { .iov_base = malloc0(verity->root_hash.iov_len), .iov_len = verity->root_hash.iov_len, }; if (!root_hash_existing.iov_base) return -ENOMEM; r = sym_crypt_volume_key_get(cd, CRYPT_ANY_SLOT, root_hash_existing.iov_base, &root_hash_existing.iov_len, NULL, 0); if (r < 0) return log_debug_errno(r, "Error opening verity device, crypt_volume_key_get failed: %m"); if (!iovec_equal(&verity->root_hash, &root_hash_existing)) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Error opening verity device, it already exists but root hashes are different."); /* Ensure that, if signatures are supported, we only reuse the device if the previous mount used the * same settings, so that a previous unsigned mount will not be reused if the user asks to use * signing for the new one, and vice versa. */ if (iovec_is_set(&verity->root_hash_sig) != FLAGS_SET(crypt_params.flags, CRYPT_VERITY_ROOT_HASH_SIGNATURE)) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Error opening verity device, it already exists but signature settings are not the same."); *ret_cd = TAKE_PTR(cd); return 0; } static char* dm_deferred_remove_clean(char *name) { if (!name) return NULL; (void) sym_crypt_deactivate_by_name(NULL, name, CRYPT_DEACTIVATE_DEFERRED); return mfree(name); } DEFINE_TRIVIAL_CLEANUP_FUNC(char *, dm_deferred_remove_clean); static int validate_signature_userspace(const VeritySettings *verity, const char *root, DissectImageFlags flags) { int r; /* Returns > 0 if signature checks out, == 0 if not, < 0 on unexpected errors */ if (!FLAGS_SET(flags, DISSECT_IMAGE_ALLOW_USERSPACE_VERITY)) { log_debug("Userspace dm-verity signature authentication disabled via flag."); return 0; } r = secure_getenv_bool("SYSTEMD_ALLOW_USERSPACE_VERITY"); if (r < 0 && r != -ENXIO) { log_debug_errno(r, "Failed to parse $SYSTEMD_ALLOW_USERSPACE_VERITY environment variable, refusing userspace dm-verity signature authentication."); return 0; } if (!r) { log_debug("Userspace dm-verity signature authentication disabled via $SYSTEMD_ALLOW_USERSPACE_VERITY environment variable."); return 0; } bool b; r = proc_cmdline_get_bool("systemd.allow_userspace_verity", PROC_CMDLINE_TRUE_WHEN_MISSING, &b); if (r < 0) { log_debug_errno(r, "Failed to parse systemd.allow_userspace_verity= kernel command line option, refusing userspace dm-verity signature authentication."); return 0; } if (!b) { log_debug("Userspace dm-verity signature authentication disabled via systemd.allow_userspace_verity= kernel command line option."); return 0; } #if HAVE_OPENSSL _cleanup_(sk_X509_free_allp) STACK_OF(X509) *sk = NULL; _cleanup_strv_free_ char **certs = NULL; _cleanup_(PKCS7_freep) PKCS7 *p7 = NULL; _cleanup_free_ char *s = NULL; _cleanup_(BIO_freep) BIO *bio = NULL; /* 'bio' must be freed first, 's' second, hence keep this order * of declaration in place, please */ assert(verity); assert(iovec_is_set(&verity->root_hash)); assert(iovec_is_set(&verity->root_hash_sig)); r = dlopen_libcrypto(LOG_DEBUG); if (r < 0) return r; /* Because installing a signature certificate into the kernel chain is so messy, let's optionally do * userspace validation. */ r = conf_files_list_nulstr(&certs, ".crt", root, CONF_FILES_REGULAR|CONF_FILES_FILTER_MASKED, CONF_PATHS_NULSTR("verity.d")); if (r < 0) return log_debug_errno(r, "Failed to enumerate certificates: %m"); if (strv_isempty(certs)) { log_debug("No userspace dm-verity certificates found."); return 0; } const unsigned char *d = verity->root_hash_sig.iov_base; p7 = sym_d2i_PKCS7(NULL, &d, (long) verity->root_hash_sig.iov_len); if (!p7) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Failed to parse PKCS7 DER signature data."); s = hexmem(verity->root_hash.iov_base, verity->root_hash.iov_len); if (!s) return log_oom_debug(); bio = sym_BIO_new_mem_buf(s, strlen(s)); if (!bio) return log_oom_debug(); sk = sym_sk_X509_new_null(); if (!sk) return log_oom_debug(); STRV_FOREACH(i, certs) { _cleanup_(X509_freep) X509 *c = NULL; _cleanup_fclose_ FILE *f = NULL; f = fopen(*i, "re"); if (!f) { log_debug_errno(errno, "Failed to open '%s', ignoring: %m", *i); continue; } c = sym_PEM_read_X509(f, NULL, NULL, NULL); if (!c) { log_debug("Failed to load X509 certificate '%s', ignoring.", *i); continue; } if (sym_sk_X509_push(sk, c) == 0) return log_oom_debug(); TAKE_PTR(c); } r = sym_PKCS7_verify(p7, sk, NULL, bio, NULL, PKCS7_NOINTERN|PKCS7_NOVERIFY); if (r) log_debug("Userspace PKCS#7 validation succeeded."); else log_debug("Userspace PKCS#7 validation failed: %s", sym_ERR_error_string(sym_ERR_get_error(), NULL)); return r; #else log_debug("Not doing client-side validation of dm-verity root hash signatures, OpenSSL support disabled."); return 0; #endif } static int do_crypt_activate_verity( struct crypt_device *cd, const char *root, const char *name, const VeritySettings *verity, DissectImageFlags flags, PartitionPolicyFlags policy_flags) { int r; assert(cd); assert(name); assert(verity); bool check_signature; if (iovec_is_set(&verity->root_hash_sig) && FLAGS_SET(policy_flags, PARTITION_POLICY_SIGNED)) { r = secure_getenv_bool("SYSTEMD_DISSECT_VERITY_SIGNATURE"); if (r < 0 && r != -ENXIO) log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_SIGNATURE"); check_signature = r != 0; } else check_signature = false; bool measure_signature; if (check_signature) { int k; /* First, if we have support for signed keys in the kernel, then try that first. */ r = sym_crypt_activate_by_signed_key( cd, name, verity->root_hash.iov_base, verity->root_hash.iov_len, verity->root_hash_sig.iov_base, verity->root_hash_sig.iov_len, CRYPT_ACTIVATE_READONLY); if (r >= 0) { log_debug("Verity activation via kernel signature logic worked."); measure_signature = true; goto done; } log_debug_errno(r, "Validation of dm-verity signature failed via the kernel, trying userspace validation instead: %m"); /* Let's mangle ENOKEY → EDESTADDRREQ, so that we return a clear, recognizable error if * there's a signature we don't recognize, that is distinct from the LUKS/encryption * -ENOKEY, which means "password required, but I have none". */ if (r == -ENOKEY) r = -EDESTADDRREQ; /* So this didn't work via the kernel, then let's try userspace validation instead. If that * works we'll try to activate without telling the kernel the signature. */ /* Preferably propagate the original kernel error, so that the fallback logic can work, * as the device-mapper is finicky around concurrent activations of the same volume */ k = validate_signature_userspace(verity, root, flags); if (k < 0) return k; if (k == 0) { log_debug("Activation of signed Verity volume worked neither via the kernel nor in userspace, can't activate."); /* So if we had a signature and we're supposed to exclusively allow * signature-based activation, then return the error now */ if (!FLAGS_SET(policy_flags, PARTITION_POLICY_VERITY)) return r < 0 ? r : -EDESTADDRREQ; log_debug("Activation of signed Verity volume without validating signature is permitted by policy. Continuing."); } else log_debug("Verity activation via userspace signature logic worked, activating by root hash."); /* Otherwise let's see what signature-less activation results in. */ measure_signature = true; } else if (!FLAGS_SET(policy_flags, PARTITION_POLICY_VERITY)) return log_debug_errno(SYNTHETIC_ERRNO(ERFKILL), "No-signature activation of Verity volume not allowed by policy, refusing."); else measure_signature = false; r = sym_crypt_activate_by_volume_key( cd, name, verity->root_hash.iov_base, verity->root_hash.iov_len, CRYPT_ACTIVATE_READONLY); if (r < 0) return log_debug_errno(r, "Activation of Verity via root hash failed: %m"); log_debug("Activation of Verity via root hash succeeded."); done: (void) pcrextend_verity_now( name, &verity->root_hash, measure_signature ? &verity->root_hash_sig : NULL); return 0; } static usec_t verity_timeout(void) { usec_t t = 100 * USEC_PER_MSEC; const char *e; int r; /* On slower machines, like non-KVM vm, setting up device may take a long time. * Let's make the timeout configurable. */ e = getenv("SYSTEMD_DISSECT_VERITY_TIMEOUT_SEC"); if (!e) return t; r = parse_sec(e, &t); if (r < 0) log_debug_errno(r, "Failed to parse timeout specified in $SYSTEMD_DISSECT_VERITY_TIMEOUT_SEC, " "using the default timeout (%s).", FORMAT_TIMESPAN(t, USEC_PER_MSEC)); return t; } static int verity_partition( DissectedImage *di, PartitionDesignator designator, DissectedPartition *m, /* data partition */ DissectedPartition *v, /* verity partition */ const char *root, /* The root to get user verity certs from (for a sysext) */ const VeritySettings *verity, DissectImageFlags flags, PartitionPolicyFlags policy_flags, DecryptedImage *d) { _cleanup_(crypt_freep) struct crypt_device *cd = NULL; _cleanup_free_ char *node = NULL, *name = NULL; _cleanup_close_ int mount_node_fd = -EBADF; int r; assert(di); assert(m); assert(v || (verity && verity->data_path)); if (!verity || !iovec_is_set(&verity->root_hash)) return 0; if (!((verity->designator < 0 && designator == PARTITION_ROOT) || (verity->designator == designator))) return 0; if (!m->found || !m->node || !m->fstype) return 0; if (!verity->data_path) { if (!v->found || !v->node || !v->fstype) return 0; if (!streq(v->fstype, "DM_verity_hash")) return 0; } if (!(policy_flags & (PARTITION_POLICY_VERITY|PARTITION_POLICY_SIGNED))) { log_debug("Attempted to unlock partition via Verity, but it's prohibited, skipping."); return 0; } r = dlopen_cryptsetup(LOG_DEBUG); if (r < 0) return r; if (FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) { /* Use the roothash, which is unique per volume, as the device node name, so that it can be reused */ _cleanup_free_ char *root_hash_encoded = NULL; root_hash_encoded = hexmem(verity->root_hash.iov_base, verity->root_hash.iov_len); if (!root_hash_encoded) return -ENOMEM; r = make_dm_name_and_node(root_hash_encoded, /* diskseq= */ 0, "-verity", &name, &node); } else r = make_dm_name_and_node_from_node(m->node, dissected_image_diskseq(di), "-verity", &name, &node); if (r < 0) return r; r = sym_crypt_init(&cd, verity->data_path ?: v->node); if (r < 0) return r; cryptsetup_enable_logging(cd); r = sym_crypt_load(cd, CRYPT_VERITY, NULL); if (r < 0) return r; r = sym_crypt_set_data_device(cd, m->node); if (r < 0) return r; if (!GREEDY_REALLOC0(d->decrypted, d->n_decrypted + 1)) return -ENOMEM; /* If activating fails because the device already exists, check the metadata and reuse it if it matches. * In case of ENODEV/ENOENT, which can happen if another process is activating at the exact same time, * retry a few times before giving up. */ for (unsigned i = 0; i < N_DEVICE_NODE_LIST_ATTEMPTS; i++) { _cleanup_(dm_deferred_remove_cleanp) char *restore_deferred_remove = NULL; _cleanup_(crypt_freep) struct crypt_device *existing_cd = NULL; _cleanup_close_ int fd = -EBADF; /* First, check if the device already exists. */ fd = open(node, O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY); if (fd < 0 && !ERRNO_IS_DEVICE_ABSENT(errno)) return log_debug_errno(errno, "Failed to open verity device %s: %m", node); if (fd >= 0) goto check; /* The device already exists. Let's check it. */ /* The symlink to the device node does not exist yet. Assume not activated, and let's activate it. */ r = do_crypt_activate_verity(cd, root, name, verity, flags, policy_flags); if (r >= 0) goto try_open; /* The device is activated. Let's open it. */ /* libdevmapper can return EINVAL when the device is already in the activation stage. * There's no way to distinguish this situation from a genuine error due to invalid * parameters, so immediately fall back to activating the device with a unique name. * Improvements in libcrypsetup can ensure this never happens: * https://gitlab.com/cryptsetup/cryptsetup/-/merge_requests/96 */ if (r == -EINVAL && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) break; /* Volume is being opened but not ready, crypt_init_by_name would fail, try to open again if * sharing is enabled. */ if (r == -ENODEV && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) goto try_again; if (!IN_SET(r, -EEXIST, /* Volume has already been opened and ready to be used. */ -EBUSY /* Volume is being opened but not ready, crypt_init_by_name() can fetch details. */)) return log_debug_errno(r, "Failed to activate verity device %s: %m", node); check: /* To avoid races, disable automatic removal on umount while setting up the new device. Restore it on failure. */ r = dm_deferred_remove_cancel(name); /* -EBUSY and -ENXIO: the device has already been removed or being removed. We cannot * use the device, try to open again. See target_message() in drivers/md/dm-ioctl.c * and dm_cancel_deferred_remove() in drivers/md/dm.c */ if (IN_SET(r, -EBUSY, -ENXIO)) goto try_again; if (r < 0) return log_debug_errno(r, "Failed to disable automated deferred removal for verity device %s: %m", node); restore_deferred_remove = strdup(name); if (!restore_deferred_remove) return log_oom_debug(); r = verity_can_reuse(verity, name, &existing_cd); /* Same as above, -EINVAL can randomly happen when it actually means -EEXIST */ if (r == -EINVAL && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) break; if (IN_SET(r, -ENOENT, /* Removed?? */ -EBUSY, /* Volume is being opened but not ready, crypt_init_by_name() can fetch details. */ -ENODEV /* Volume is being opened but not ready, crypt_init_by_name() would fail, try to open again. */ )) goto try_again; if (r < 0) return log_debug_errno(r, "Failed to check if existing verity device %s can be reused: %m", node); if (fd < 0) { /* devmapper might say that the device exists, but the devlink might not yet have been * created. Check and wait for the udev event in that case. */ r = device_wait_for_devlink(node, "block", verity_timeout(), NULL); /* Fallback to activation with a unique device if it's taking too long */ if (r == -ETIMEDOUT && FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) break; if (r < 0) return log_debug_errno(r, "Failed to wait device node symlink %s: %m", node); } try_open: if (fd < 0) { /* Now, the device is activated and devlink is created. Let's open it. */ fd = open(node, O_RDONLY|O_NONBLOCK|O_CLOEXEC|O_NOCTTY); if (fd < 0) { if (!ERRNO_IS_DEVICE_ABSENT(errno)) return log_debug_errno(errno, "Failed to open verity device %s: %m", node); /* The device has already been removed?? */ goto try_again; } } /* Everything looks good and we'll be able to mount the device, so deferred remove will be re-enabled at that point. */ restore_deferred_remove = mfree(restore_deferred_remove); mount_node_fd = TAKE_FD(fd); if (existing_cd) crypt_free_and_replace(cd, existing_cd); goto success; try_again: /* Device is being removed by another process. Let's wait for a while. */ (void) usleep_safe(2 * USEC_PER_MSEC); } /* All trials failed or a conflicting verity device exists. Let's try to activate with a unique name. */ if (FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)) { /* Before trying to activate with unique name, we need to free crypt_device object. * Otherwise, we get error from libcryptsetup like the following: * ------ * systemd[1234]: Cannot use device /dev/loop5 which is in use (already mapped or mounted). * ------ */ sym_crypt_free(cd); cd = NULL; return verity_partition( di, designator, m, v, root, verity, flags & ~DISSECT_IMAGE_VERITY_SHARE, policy_flags, d); } return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), "All attempts to activate verity device %s failed.", name); success: d->decrypted[d->n_decrypted++] = (DecryptedPartition) { .name = TAKE_PTR(name), .device = TAKE_PTR(cd), }; m->decrypted_node = TAKE_PTR(node); close_and_replace(m->mount_node_fd, mount_node_fd); return 0; } #endif int dissected_image_decrypt( DissectedImage *m, const char *root, /* The root to get user verity certs from (for a sysext) */ const char *passphrase, const VeritySettings *verity, const ImagePolicy *policy, DissectImageFlags flags) { #if HAVE_LIBCRYPTSETUP _cleanup_(decrypted_image_unrefp) DecryptedImage *d = NULL; #endif int r; assert(m); assert(!verity || iovec_is_valid(&verity->root_hash)); assert(!verity || iovec_is_valid(&verity->root_hash_sig)); /* Returns: * * = 0 → There was nothing to decrypt/setup * > 0 → Decrypted/setup successfully * -ENOKEY → dm-crypt: there's something to decrypt but no decryption key was supplied * -EKEYREJECTED → dm-crypt: Passed key was not correct * -EDESTADDRREQ → dm-verity: there's something to setup but no signature was supplied * -EBUSY → dm-verity: Generic Verity error (kernel is not very explanatory) * -ERFKILL → image policy not compatible with request * -EEXIST → DM device already exists under the specified name */ if (verity && iovec_is_set(&verity->root_hash) && verity->root_hash.iov_len < sizeof(sd_id128_t)) return -EINVAL; if (!m->encrypted && !m->verity_ready) return 0; r = secure_getenv_bool("SYSTEMD_VERITY_SHARING"); if (r >= 0) SET_FLAG(flags, DISSECT_IMAGE_VERITY_SHARE, r); #if HAVE_LIBCRYPTSETUP r = decrypted_image_new(&d); if (r < 0) return r; for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { DissectedPartition *p = m->partitions + i; PartitionDesignator k; if (!p->found) continue; PartitionPolicyFlags fl = image_policy_get_exhaustively(policy, i); r = decrypt_partition(m, p, passphrase, flags, fl, d); if (r < 0) return r; k = partition_verity_hash_of(i); if (k >= 0) { r = verity_partition(m, i, p, m->partitions + k, root, verity, flags, fl, d); if (r < 0) return r; } if (!p->decrypted_fstype && p->mount_node_fd >= 0 && p->decrypted_node) { r = probe_filesystem_full(p->mount_node_fd, p->decrypted_node, 0, UINT64_MAX, /* bool restrict_fstypes= */ true, &p->decrypted_fstype); if (r < 0 && r != -EUCLEAN) return r; } } m->decrypted_image = TAKE_PTR(d); return 1; #else return -EOPNOTSUPP; #endif } int dissected_image_decrypt_interactively( DissectedImage *m, const char *passphrase, const VeritySettings *verity, const ImagePolicy *image_policy, DissectImageFlags flags) { _cleanup_strv_free_erase_ char **z = NULL; int n = 3, r; if (passphrase) n--; for (;;) { r = dissected_image_decrypt(m, /* root= */ NULL, passphrase, verity, image_policy, flags); if (r >= 0) return r; if (r == -EKEYREJECTED) log_error_errno(r, "Incorrect passphrase, try again!"); else if (r == -EDESTADDRREQ) return log_error_errno(r, "Image lacks recognized signature."); else if (r == -ERFKILL) return log_error_errno(r, "Unlocking of Verity/LUKS volumes not permitted by policy."); else if (r != -ENOKEY) return log_error_errno(r, "Failed to decrypt/set up image: %m"); if (--n < 0) return log_error_errno(SYNTHETIC_ERRNO(EKEYREJECTED), "Too many retries."); z = strv_free_erase(z); static const AskPasswordRequest req = { .tty_fd = -EBADF, .message = "Please enter image passphrase:", .id = "dissect", .keyring = "dissect", .credential = "dissect.passphrase", .until = USEC_INFINITY, .hup_fd = -EBADF, }; r = ask_password_auto(&req, /* flags= */ 0, &z); if (r < 0) return log_error_errno(r, "Failed to query for passphrase: %m"); assert(!strv_isempty(z)); passphrase = z[0]; } } static int decrypted_image_relinquish(DecryptedImage *d) { assert(d); /* Turns on automatic removal after the last use ended for all DM devices of this image, and sets a * boolean so that we don't clean it up ourselves either anymore */ #if HAVE_LIBCRYPTSETUP int r; for (size_t i = 0; i < d->n_decrypted; i++) { DecryptedPartition *p = d->decrypted + i; if (p->relinquished) continue; r = sym_crypt_deactivate_by_name(NULL, p->name, CRYPT_DEACTIVATE_DEFERRED); if (r < 0) return log_debug_errno(r, "Failed to mark %s for auto-removal: %m", p->name); p->relinquished = true; } #endif return 0; } int dissected_image_relinquish(DissectedImage *m) { int r; assert(m); if (m->decrypted_image) { r = decrypted_image_relinquish(m->decrypted_image); if (r < 0) return r; } if (m->loop) loop_device_relinquish(m->loop); return 0; } void image_filter_done(ImageFilter *f) { assert(f); FOREACH_ELEMENT(p, f->pattern) *p = mfree(*p); } ImageFilter *image_filter_free(ImageFilter *f) { if (!f) return NULL; image_filter_done(f); return mfree(f); } int image_filter_parse(const char *s, ImageFilter **ret) { _cleanup_(image_filter_freep) ImageFilter *f = NULL; int r; if (isempty(s)) { if (ret) *ret = NULL; return 0; } for (;;) { _cleanup_free_ char *word = NULL; r = extract_first_word(&s, &word, ":", EXTRACT_UNQUOTE|EXTRACT_DONT_COALESCE_SEPARATORS); if (r < 0) return log_debug_errno(r, "Failed to extract word: %m"); if (r == 0) break; _cleanup_free_ char *designator = NULL, *pattern = NULL; const char *x = word; r = extract_many_words(&x, "=", EXTRACT_UNQUOTE|EXTRACT_DONT_COALESCE_SEPARATORS, &designator, &pattern); if (r < 0) return log_debug_errno(r, "Failed to extract designator: %m"); if (r != 2 || !isempty(x)) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Unable to split: %s", word); PartitionDesignator d = partition_designator_from_string(designator); if (d < 0) return log_debug_errno(d, "Failed to parse partition designator: %s", designator); if (!f) { f = new0(ImageFilter, 1); if (!f) return log_oom_debug(); } else if (f->pattern[d]) return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Duplicate pattern for '%s', refusing.", partition_designator_to_string(d)); f->pattern[d] = TAKE_PTR(pattern); } if (ret) *ret = TAKE_PTR(f); return 0; } static char *build_auxiliary_path(const char *image, const char *suffix) { const char *e; char *n; assert(image); assert(suffix); e = endswith(image, ".raw"); if (!e) return strjoin(image, suffix); n = new(char, e - image + strlen(suffix) + 1); if (!n) return NULL; strcpy(mempcpy(n, image, e - image), suffix); return n; } void verity_settings_done(VeritySettings *v) { assert(v); iovec_done(&v->root_hash); iovec_done(&v->root_hash_sig); v->data_path = mfree(v->data_path); } VeritySettings* verity_settings_free(VeritySettings *v) { if (!v) return NULL; verity_settings_done(v); return mfree(v); } void verity_settings_hash_func(const VeritySettings *s, struct siphash *state) { assert(s); siphash24_compress_typesafe(s->root_hash.iov_len, state); siphash24_compress(s->root_hash.iov_base, s->root_hash.iov_len, state); } int verity_settings_compare_func(const VeritySettings *x, const VeritySettings *y) { assert(x); assert(y); return iovec_memcmp(&x->root_hash, &y->root_hash); } DEFINE_HASH_OPS_WITH_VALUE_DESTRUCTOR(verity_settings_hash_ops, VeritySettings, verity_settings_hash_func, verity_settings_compare_func, VeritySettings, verity_settings_free); int verity_settings_load( VeritySettings *verity, const char *image, const char *root_hash_path, const char *root_hash_sig_path) { PartitionDesignator designator; int r; assert(verity); assert(image); assert(verity->designator < 0 || IN_SET(verity->designator, PARTITION_ROOT, PARTITION_USR)); /* If we are asked to load the root hash for a device node, exit early */ if (is_device_path(image)) return 0; r = secure_getenv_bool("SYSTEMD_DISSECT_VERITY_SIDECAR"); if (r < 0 && r != -ENXIO) log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_SIDECAR, ignoring: %m"); if (r == 0) return 0; designator = verity->designator; /* We only fill in what isn't already filled in */ _cleanup_(iovec_done) struct iovec root_hash = {}; if (!iovec_is_set(&verity->root_hash)) { _cleanup_free_ char *text = NULL; if (root_hash_path) { /* If explicitly specified it takes precedence */ r = read_one_line_file(root_hash_path, &text); if (r < 0) return r; if (designator < 0) designator = PARTITION_ROOT; } else { /* Otherwise look for xattr and separate file, and first for the data for root and if * that doesn't exist for /usr */ if (designator < 0 || designator == PARTITION_ROOT) { r = getxattr_malloc(image, "user.verity.roothash", &text, /* ret_size= */ NULL); if (r < 0) { _cleanup_free_ char *p = NULL; if (!ERRNO_IS_XATTR_ABSENT(r)) return r; p = build_auxiliary_path(image, ".roothash"); if (!p) return -ENOMEM; r = read_one_line_file(p, &text); if (r < 0 && r != -ENOENT) return r; } if (text) designator = PARTITION_ROOT; } if (!text && (designator < 0 || designator == PARTITION_USR)) { /* So in the "roothash" xattr/file name above the "root" of course primarily * refers to the root of the Verity Merkle tree. But coincidentally it also * is the hash for the *root* file system, i.e. the "root" neatly refers to * two distinct concepts called "root". Taking benefit of this happy * coincidence we call the file with the root hash for the /usr/ file system * `usrhash`, because `usrroothash` or `rootusrhash` would just be too * confusing. We thus drop the reference to the root of the Merkle tree, and * just indicate which file system it's about. */ r = getxattr_malloc(image, "user.verity.usrhash", &text, /* ret_size= */ NULL); if (r < 0) { _cleanup_free_ char *p = NULL; if (!ERRNO_IS_XATTR_ABSENT(r)) return r; p = build_auxiliary_path(image, ".usrhash"); if (!p) return -ENOMEM; r = read_one_line_file(p, &text); if (r < 0 && r != -ENOENT) return r; } if (text) designator = PARTITION_USR; } } if (text) { r = unhexmem(text, &root_hash.iov_base, &root_hash.iov_len); if (r < 0) return r; if (root_hash.iov_len < sizeof(sd_id128_t)) return -EINVAL; } } _cleanup_(iovec_done) struct iovec root_hash_sig = {}; if ((iovec_is_set(&root_hash) || iovec_is_set(&verity->root_hash)) && !iovec_is_set(&verity->root_hash_sig)) { if (root_hash_sig_path) { r = read_full_file(root_hash_sig_path, (char**) &root_hash_sig.iov_base, &root_hash_sig.iov_len); if (r < 0 && r != -ENOENT) return r; if (r >= 0 && root_hash_sig.iov_len == 0) /* refuse empty size signatures */ return -EINVAL; if (designator < 0) designator = PARTITION_ROOT; } else { if (designator < 0 || designator == PARTITION_ROOT) { _cleanup_free_ char *p = NULL; /* Follow naming convention recommended by the relevant RFC: * https://tools.ietf.org/html/rfc5751#section-3.2.1 */ p = build_auxiliary_path(image, ".roothash.p7s"); if (!p) return -ENOMEM; r = read_full_file(p, (char**) &root_hash_sig.iov_base, &root_hash_sig.iov_len); if (r < 0 && r != -ENOENT) return r; if (r >= 0) { designator = PARTITION_ROOT; if (root_hash_sig.iov_len == 0) /* refuse empty size signatures */ return -EINVAL; } } if (!iovec_is_set(&root_hash_sig) && (designator < 0 || designator == PARTITION_USR)) { _cleanup_free_ char *p = NULL; p = build_auxiliary_path(image, ".usrhash.p7s"); if (!p) return -ENOMEM; r = read_full_file(p, (char**) &root_hash_sig.iov_base, &root_hash_sig.iov_len); if (r < 0 && r != -ENOENT) return r; if (r >= 0) { designator = PARTITION_USR; if (root_hash_sig.iov_len == 0) /* refuse empty size signatures */ return -EINVAL; } } } } _cleanup_free_ char *verity_data_path = NULL; if (!verity->data_path) { _cleanup_free_ char *p = NULL; p = build_auxiliary_path(image, ".verity"); if (!p) return -ENOMEM; if (access(p, F_OK) < 0) { if (errno != ENOENT) return -errno; } else verity_data_path = TAKE_PTR(p); } if (iovec_is_set(&root_hash)) verity->root_hash = TAKE_STRUCT(root_hash); if (iovec_is_set(&root_hash_sig)) verity->root_hash_sig = TAKE_STRUCT(root_hash_sig); if (verity_data_path) verity->data_path = TAKE_PTR(verity_data_path); if (verity->designator < 0) verity->designator = designator; return 1; } int verity_settings_copy(VeritySettings *dest, const VeritySettings *source) { assert(dest); if (!source) { *dest = VERITY_SETTINGS_DEFAULT; return 0; } _cleanup_(iovec_done) struct iovec rh = {}; if (iovec_is_set(&source->root_hash)) { if (!iovec_memdup(&source->root_hash, &rh)) return log_oom_debug(); } _cleanup_(iovec_done) struct iovec sig = {}; if (iovec_is_set(&source->root_hash_sig)) { if (!iovec_memdup(&source->root_hash_sig, &sig)) return log_oom_debug(); } _cleanup_free_ char *p = NULL; if (source->data_path) { p = strdup(source->data_path); if (!p) return log_oom_debug(); } *dest = (VeritySettings) { .root_hash = TAKE_STRUCT(rh), .root_hash_sig = TAKE_STRUCT(sig), .data_path = TAKE_PTR(p), .designator = source->designator, }; return 1; } int dissected_image_load_verity_sig_partition( DissectedImage *m, int fd, VeritySettings *verity) { int r; assert(m); assert(fd >= 0); assert(verity); if (iovec_is_set(&verity->root_hash) && iovec_is_set(&verity->root_hash_sig)) /* Already loaded? */ return 0; r = secure_getenv_bool("SYSTEMD_DISSECT_VERITY_EMBEDDED"); if (r < 0 && r != -ENXIO) log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_EMBEDDED, ignoring: %m"); if (r == 0) return 0; PartitionDesignator dd = verity->designator; if (dd < 0) { if (m->partitions[PARTITION_ROOT_VERITY].found) dd = PARTITION_ROOT; else if (m->partitions[PARTITION_USR_VERITY].found) dd = PARTITION_USR; else return 0; } if (!m->partitions[dd].found) return 0; PartitionDesignator dv = partition_verity_hash_of(dd); assert(dv >= 0); if (!m->partitions[dv].found) return 0; PartitionDesignator ds = partition_verity_sig_of(dd); assert(ds >= 0); DissectedPartition *p = m->partitions + ds; if (!p->found) return 0; _cleanup_(iovec_done) struct iovec root_hash = {}, root_hash_sig = {}; r = acquire_sig_for_roothash( fd, p->offset, p->size, &root_hash, &root_hash_sig); if (r < 0) return r; /* Check if specified root hash matches if it is specified */ if (iovec_is_set(&verity->root_hash) && !iovec_equal(&verity->root_hash, &root_hash)) { _cleanup_free_ char *a = NULL, *b = NULL; a = hexmem(root_hash.iov_base, root_hash.iov_len); b = hexmem(verity->root_hash.iov_base, verity->root_hash.iov_len); return log_debug_errno(SYNTHETIC_ERRNO(EINVAL), "Root hash in signature JSON data (%s) doesn't match configured hash (%s).", strna(a), strna(b)); } iovec_done(&verity->root_hash); verity->root_hash = TAKE_STRUCT(root_hash); iovec_done(&verity->root_hash_sig); verity->root_hash_sig = TAKE_STRUCT(root_hash_sig); verity->designator = dd; m->verity_ready = true; m->verity_sig_ready = true; m->partitions[dd].rw = false; return 1; } int dissected_image_guess_verity_roothash( DissectedImage *m, VeritySettings *verity) { int r; assert(m); assert(verity); /* Guesses the Verity root hash from the partitions we found, taking into account that as per * https://uapi-group.org/specifications/specs/discoverable_partitions_specification/ the UUIDS of * the data and verity partitions are respectively the first and second halves of the dm-verity * roothash. * * Note of course that relying on this guesswork is mostly useful for later attestation, not so much * for a-priori security. */ if (iovec_is_set(&verity->root_hash)) /* Already loaded? */ return 0; r = secure_getenv_bool("SYSTEMD_DISSECT_VERITY_GUESS"); if (r < 0 && r != -ENXIO) log_debug_errno(r, "Failed to parse $SYSTEMD_DISSECT_VERITY_GUESS, ignoring: %m"); if (r == 0) return 0; PartitionDesignator dd = verity->designator; if (dd < 0) { if (m->partitions[PARTITION_ROOT_VERITY].found) dd = PARTITION_ROOT; else if (m->partitions[PARTITION_USR_VERITY].found) dd = PARTITION_USR; else return 0; } DissectedPartition *d = m->partitions + dd; if (!d->found) return 0; PartitionDesignator dv = partition_verity_hash_of(dd); assert(dv >= 0); DissectedPartition *p = m->partitions + dv; if (!p->found) return 0; _cleanup_free_ void *rh = malloc(sizeof(sd_id128_t) * 2); if (!rh) return log_oom_debug(); memcpy(mempcpy(rh, &d->uuid, sizeof(sd_id128_t)), &p->uuid, sizeof(sd_id128_t)); verity->root_hash = IOVEC_MAKE(TAKE_PTR(rh), sizeof(sd_id128_t) * 2); verity->designator = dd; m->verity_ready = true; m->partitions[dd].rw = false; return 0; } int dissected_image_acquire_metadata( DissectedImage *m, int userns_fd, DissectImageFlags extra_flags) { enum { META_HOSTNAME, META_MACHINE_ID, META_MACHINE_INFO, META_OS_RELEASE, META_INITRD_RELEASE, META_SYSEXT_RELEASE, META_CONFEXT_RELEASE, META_HAS_INIT_SYSTEM, _META_MAX, }; static const char *const paths[_META_MAX] = { [META_HOSTNAME] = "/etc/hostname\0", [META_MACHINE_ID] = "/etc/machine-id\0", [META_MACHINE_INFO] = "/etc/machine-info\0", [META_OS_RELEASE] = "/etc/os-release\0" "/usr/lib/os-release\0", [META_INITRD_RELEASE] = "/etc/initrd-release\0" "/usr/lib/initrd-release\0", [META_SYSEXT_RELEASE] = "sysext-release\0", /* String used only for logging. */ [META_CONFEXT_RELEASE] = "confext-release\0", /* ditto */ [META_HAS_INIT_SYSTEM] = "has-init-system\0", /* ditto */ }; _cleanup_strv_free_ char **machine_info = NULL, **os_release = NULL, **initrd_release = NULL, **sysext_release = NULL, **confext_release = NULL; _cleanup_free_ char *hostname = NULL, *t = NULL; _cleanup_close_pair_ int error_pipe[2] = EBADF_PAIR; _cleanup_(pidref_done_sigkill_wait) PidRef child = PIDREF_NULL; sd_id128_t machine_id = SD_ID128_NULL; unsigned n_meta_initialized = 0; int fds[2 * _META_MAX], r, v; int has_init_system = -1; ssize_t n; assert(m); r = dlopen_libmount(LOG_DEBUG); if (r < 0) return r; for (; n_meta_initialized < _META_MAX; n_meta_initialized++) { assert(paths[n_meta_initialized]); if (pipe2(fds + 2*n_meta_initialized, O_CLOEXEC) < 0) { r = -errno; goto finish; } } r = get_common_dissect_directory(&t); if (r < 0) goto finish; if (pipe2(error_pipe, O_CLOEXEC) < 0) { r = -errno; goto finish; } r = pidref_safe_fork("(sd-dissect)", FORK_RESET_SIGNALS|FORK_DEATHSIG_SIGTERM, &child); if (r < 0) goto finish; if (r == 0) { /* Child */ error_pipe[0] = safe_close(error_pipe[0]); if (userns_fd < 0) r = detach_mount_namespace_harder(0, 0); else r = detach_mount_namespace_userns(userns_fd); if (r < 0) { log_debug_errno(r, "Failed to detach mount namespace: %m"); report_errno_and_exit(error_pipe[1], r); } r = dissected_image_mount( m, t, /* uid_shift= */ UID_INVALID, /* uid_range= */ UID_INVALID, /* userns_fd= */ -EBADF, extra_flags | DISSECT_IMAGE_READ_ONLY | DISSECT_IMAGE_MOUNT_ROOT_ONLY | DISSECT_IMAGE_USR_NO_ROOT); if (r < 0) { log_debug_errno(r, "Failed to mount dissected image: %m"); report_errno_and_exit(error_pipe[1], r); } for (unsigned k = 0; k < _META_MAX; k++) { _cleanup_close_ int fd = -ENOENT; assert(paths[k]); fds[2*k] = safe_close(fds[2*k]); switch (k) { case META_SYSEXT_RELEASE: if (!m->image_name) goto next; /* As per the os-release spec, if the image is an extension it will have a * file named after the image name in extension-release.d/ - we use the image * name and try to resolve it with the extension-release helpers, as * sometimes the image names are mangled on deployment and do not match * anymore. Unlike other paths this is not fixed, and the image name can be * mangled on deployment, so by calling into the helper we allow a fallback * that matches on the first extension-release file found in the directory, * if one named after the image cannot be found first. */ r = open_extension_release( t, IMAGE_SYSEXT, m->image_name, /* relax_extension_release_check= */ false, /* ret_path= */ NULL, &fd); if (r < 0) fd = r; break; case META_CONFEXT_RELEASE: if (!m->image_name) goto next; /* As above */ r = open_extension_release( t, IMAGE_CONFEXT, m->image_name, /* relax_extension_release_check= */ false, /* ret_path= */ NULL, &fd); if (r < 0) fd = r; break; case META_HAS_INIT_SYSTEM: { bool found = false; FOREACH_STRING(init, "/usr/lib/systemd/systemd", /* systemd on /usr/ merged system */ "/lib/systemd/systemd", /* systemd on /usr/ non-merged systems */ "/sbin/init") { /* traditional path the Linux kernel invokes */ r = chase(init, t, CHASE_PREFIX_ROOT, NULL, NULL); if (r < 0) { if (r != -ENOENT) log_debug_errno(r, "Failed to resolve %s, ignoring: %m", init); } else { found = true; break; } } r = loop_write(fds[2*k+1], &found, sizeof(found)); if (r < 0) report_errno_and_exit(error_pipe[1], r); goto next; } default: NULSTR_FOREACH(p, paths[k]) { fd = chase_and_open(p, t, CHASE_PREFIX_ROOT, O_RDONLY|O_CLOEXEC|O_NOCTTY, NULL); if (fd >= 0) break; } } if (fd < 0) { log_debug_errno(fd, "Failed to read %s file of image, ignoring: %m", paths[k]); goto next; } r = copy_bytes(fd, fds[2*k+1], UINT64_MAX, 0); if (r < 0) report_errno_and_exit(error_pipe[1], r); next: fds[2*k+1] = safe_close(fds[2*k+1]); } _exit(EXIT_SUCCESS); } error_pipe[1] = safe_close(error_pipe[1]); for (unsigned k = 0; k < _META_MAX; k++) { _cleanup_fclose_ FILE *f = NULL; assert(paths[k]); fds[2*k+1] = safe_close(fds[2*k+1]); f = take_fdopen(&fds[2*k], "r"); if (!f) { r = -errno; goto finish; } switch (k) { case META_HOSTNAME: r = read_etc_hostname_stream(f, /* substitute_wildcards= */ false, &hostname); if (r < 0) log_debug_errno(r, "Failed to read /etc/hostname of image: %m"); break; case META_MACHINE_ID: { _cleanup_free_ char *line = NULL; r = read_line(f, LONG_LINE_MAX, &line); if (r < 0) log_debug_errno(r, "Failed to read /etc/machine-id of image: %m"); else if (r == 33) { r = sd_id128_from_string(line, &machine_id); if (r < 0) log_debug_errno(r, "Image contains invalid /etc/machine-id: %s", line); } else if (r == 0) log_debug("/etc/machine-id file of image is empty."); else if (streq(line, "uninitialized")) log_debug("/etc/machine-id file of image is uninitialized (likely aborted first boot)."); else log_debug("/etc/machine-id file of image has unexpected length %i.", r); break; } case META_MACHINE_INFO: r = load_env_file_pairs(f, "machine-info", &machine_info); if (r < 0) log_debug_errno(r, "Failed to read /etc/machine-info of image: %m"); break; case META_OS_RELEASE: r = load_env_file_pairs(f, "os-release", &os_release); if (r < 0) log_debug_errno(r, "Failed to read OS release file of image: %m"); break; case META_INITRD_RELEASE: r = load_env_file_pairs(f, "initrd-release", &initrd_release); if (r < 0) log_debug_errno(r, "Failed to read initrd release file of image: %m"); break; case META_SYSEXT_RELEASE: r = load_env_file_pairs(f, "sysext-release", &sysext_release); if (r < 0) log_debug_errno(r, "Failed to read sysext release file of image: %m"); break; case META_CONFEXT_RELEASE: r = load_env_file_pairs(f, "confext-release", &confext_release); if (r < 0) log_debug_errno(r, "Failed to read confext release file of image: %m"); break; case META_HAS_INIT_SYSTEM: { bool b = false; size_t nr; errno = 0; nr = fread(&b, 1, sizeof(b), f); if (nr != sizeof(b)) log_debug_errno(errno_or_else(EIO), "Failed to read has-init-system boolean: %m"); else has_init_system = b; break; }} } r = pidref_wait_for_terminate_and_check("(sd-dissect)", &child, 0); if (r < 0) goto finish; pidref_done(&child); n = read(error_pipe[0], &v, sizeof(v)); if (n < 0) { r = -errno; goto finish; } if (n == sizeof(v)) { r = v; /* propagate error sent to us from child */ goto finish; } if (n != 0) { r = -EIO; goto finish; } if (r != EXIT_SUCCESS) { r = -EPROTO; goto finish; } free_and_replace(m->hostname, hostname); m->machine_id = machine_id; strv_free_and_replace(m->machine_info, machine_info); strv_free_and_replace(m->os_release, os_release); strv_free_and_replace(m->initrd_release, initrd_release); strv_free_and_replace(m->sysext_release, sysext_release); strv_free_and_replace(m->confext_release, confext_release); m->has_init_system = has_init_system; finish: for (unsigned k = 0; k < n_meta_initialized; k++) safe_close_pair(fds + 2*k); return r; } Architecture dissected_image_architecture(DissectedImage *m) { assert(m); if (m->partitions[PARTITION_ROOT].found && m->partitions[PARTITION_ROOT].architecture >= 0) return m->partitions[PARTITION_ROOT].architecture; if (m->partitions[PARTITION_USR].found && m->partitions[PARTITION_USR].architecture >= 0) return m->partitions[PARTITION_USR].architecture; return _ARCHITECTURE_INVALID; } bool dissected_image_is_portable(DissectedImage *m) { return m && (strv_env_pairs_get(m->os_release, "PORTABLE_PREFIXES") || strv_env_pairs_get(m->os_release, "PORTABLE_SCOPE")); } bool dissected_image_is_initrd(DissectedImage *m) { return m && !strv_isempty(m->initrd_release); } int dissect_loop_device( LoopDevice *loop, const VeritySettings *verity, const MountOptions *mount_options, const ImagePolicy *image_policy, const ImageFilter *image_filter, DissectImageFlags flags, DissectedImage **ret) { #if HAVE_BLKID _cleanup_(dissected_image_unrefp) DissectedImage *m = NULL; int r; assert(loop); r = dissected_image_new(loop->backing_file ?: loop->node, &m); if (r < 0) return r; m->loop = loop_device_ref(loop); m->image_size = m->loop->device_size; m->sector_size = m->loop->sector_size; r = dissect_image( m, loop->fd, loop->node, verity, mount_options, image_policy, image_filter, flags); if (r < 0) return r; if (ret) *ret = TAKE_PTR(m); return 0; #else return -EOPNOTSUPP; #endif } int dissect_loop_device_and_warn( LoopDevice *loop, const VeritySettings *verity, const MountOptions *mount_options, const ImagePolicy *image_policy, const ImageFilter *image_filter, DissectImageFlags flags, DissectedImage **ret) { assert(loop); return dissect_log_error( LOG_ERR, dissect_loop_device(loop, verity, mount_options, image_policy, image_filter, flags, ret), loop->backing_file ?: loop->node, verity); } bool dissected_image_verity_candidate(const DissectedImage *image, PartitionDesignator partition_designator) { assert(image); /* Checks if this partition could theoretically do Verity. For non-partitioned images this only works * if there's an external verity file supplied, for which we can consult .has_verity. For partitioned * images we only check the partition type. * * This call is used to decide whether to suppress or show a verity column in tabular output of the * image. */ if (image->single_file_system) return partition_designator == PARTITION_ROOT && image->has_verity; return partition_verity_hash_of(partition_designator) >= 0; } bool dissected_image_verity_ready(const DissectedImage *image, PartitionDesignator partition_designator) { PartitionDesignator k; assert(image); /* Checks if this partition has verity data available that we can activate. For non-partitioned this * works for the root partition, for others only if the associated verity partition was found. */ if (!image->verity_ready) return false; if (image->single_file_system) return partition_designator == PARTITION_ROOT; k = partition_verity_hash_of(partition_designator); return k >= 0 && image->partitions[k].found; } bool dissected_image_verity_sig_ready(const DissectedImage *image, PartitionDesignator partition_designator) { PartitionDesignator k; assert(image); /* Checks if this partition has verity signature data available that we can use. */ if (!image->verity_sig_ready) return false; if (image->single_file_system) return partition_designator == PARTITION_ROOT; k = partition_verity_sig_of(partition_designator); return k >= 0 && image->partitions[k].found; } int mount_options_set_and_consume(MountOptions **options, PartitionDesignator d, char *s) { assert(options); assert(d >= 0); if (!*options) { *options = new0(MountOptions, 1); if (!*options) { free(s); return log_oom(); } } return free_and_replace((*options)->options[d], s); } int mount_options_dup(const MountOptions *source, MountOptions **ret) { _cleanup_(mount_options_free_allp) MountOptions *options = NULL; assert(source); assert(ret); options = new0(MountOptions, 1); if (!options) return log_oom_debug(); for (PartitionDesignator d = 0; d < _PARTITION_DESIGNATOR_MAX; d++) if (source->options[d]) { options->options[d] = strdup(source->options[d]); if (!options->options[d]) return log_oom_debug(); } *ret = TAKE_PTR(options); return 0; } int mount_options_to_string(const MountOptions *mount_options, char **ret) { _cleanup_free_ char *s = NULL; assert(mount_options); assert(ret); for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) if (!isempty(mount_options->options[i])) if (!strextend_with_separator(&s, ":", "%s:%s", partition_designator_to_string(i), mount_options->options[i])) return log_oom_debug(); *ret = TAKE_PTR(s); return 0; } MountOptions* mount_options_free_all(MountOptions *options) { if (!options) return NULL; free_many_charp(options->options, _PARTITION_DESIGNATOR_MAX); return mfree(options); } const char* mount_options_from_designator(const MountOptions *options, PartitionDesignator designator) { assert(designator >= 0 && designator < _PARTITION_DESIGNATOR_MAX); if (!options) return NULL; return options->options[designator]; } int mount_image_privately_interactively( const char *image, const ImagePolicy *image_policy, DissectImageFlags flags, char **ret_directory, int *ret_dir_fd, LoopDevice **ret_loop_device) { _cleanup_(verity_settings_done) VeritySettings verity = VERITY_SETTINGS_DEFAULT; _cleanup_(loop_device_unrefp) LoopDevice *d = NULL; _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; _cleanup_free_ char *dir = NULL; int r; /* Mounts an OS image at a temporary place, inside a newly created mount namespace of our own. This * is used by tools such as systemd-tmpfiles or systemd-firstboot to operate on some disk image * easily. */ assert(image); assert(ret_loop_device); /* We intend to mount this right-away, hence add the partitions if needed and pin them. */ flags |= DISSECT_IMAGE_ADD_PARTITION_DEVICES | DISSECT_IMAGE_PIN_PARTITION_DEVICES; r = verity_settings_load(&verity, image, NULL, NULL); if (r < 0) return log_error_errno(r, "Failed to load root hash data: %m"); r = loop_device_make_by_path( image, FLAGS_SET(flags, DISSECT_IMAGE_DEVICE_READ_ONLY) ? O_RDONLY : -1, /* sector_size= */ UINT32_MAX, FLAGS_SET(flags, DISSECT_IMAGE_NO_PARTITION_TABLE) ? 0 : LO_FLAGS_PARTSCAN, LOCK_SH, &d); if (r < 0) return log_error_errno(r, "Failed to set up loopback device for %s: %m", image); r = dissect_loop_device_and_warn( d, &verity, /* mount_options= */ NULL, image_policy, /* image_filter= */ NULL, flags, &dissected_image); if (r < 0) return r; r = dissected_image_load_verity_sig_partition(dissected_image, d->fd, &verity); if (r < 0) return r; r = dissected_image_guess_verity_roothash(dissected_image, &verity); if (r < 0) return r; r = dissected_image_decrypt_interactively(dissected_image, NULL, &verity, image_policy, flags); if (r < 0) return r; r = detach_mount_namespace(); if (r < 0) return log_error_errno(r, "Failed to detach mount namespace: %m"); r = mkdir_p("/run/systemd/mount-rootfs", 0555); if (r < 0) return log_error_errno(r, "Failed to create mount point: %m"); r = dissected_image_mount_and_warn( dissected_image, "/run/systemd/mount-rootfs", /* uid_shift= */ UID_INVALID, /* uid_range= */ UID_INVALID, /* userns_fd= */ -EBADF, flags); if (r < 0) return r; r = loop_device_flock(d, LOCK_UN); if (r < 0) return r; r = dissected_image_relinquish(dissected_image); if (r < 0) return log_error_errno(r, "Failed to relinquish DM and loopback block devices: %m"); if (ret_directory) { dir = strdup("/run/systemd/mount-rootfs"); if (!dir) return log_oom(); } if (ret_dir_fd) { _cleanup_close_ int dir_fd = -EBADF; dir_fd = open("/run/systemd/mount-rootfs", O_CLOEXEC|O_DIRECTORY); if (dir_fd < 0) return log_error_errno(errno, "Failed to open mount point directory: %m"); *ret_dir_fd = TAKE_FD(dir_fd); } if (ret_directory) *ret_directory = TAKE_PTR(dir); *ret_loop_device = TAKE_PTR(d); return 0; } static bool mount_options_relax_extension_release_checks(const MountOptions *options) { if (!options) return false; return string_contains_word(mount_options_from_designator(options, PARTITION_ROOT), ",", "x-systemd.relax-extension-release-check") || string_contains_word(mount_options_from_designator(options, PARTITION_USR), ",", "x-systemd.relax-extension-release-check"); } int verity_dissect_and_mount( int src_fd, const char *src, const char *dest, const MountOptions *options, const ImagePolicy *image_policy, const ImageFilter *image_filter, const ExtensionReleaseData *extension_release_data, ImageClass required_class, VeritySettings *verity, RuntimeScope runtime_scope, DissectedImage **ret_image) { _cleanup_(loop_device_unrefp) LoopDevice *loop_device = NULL; _cleanup_(dissected_image_unrefp) DissectedImage *dissected_image = NULL; _cleanup_(verity_settings_done) VeritySettings local_verity = VERITY_SETTINGS_DEFAULT; _cleanup_close_ int userns_fd = -EBADF; DissectImageFlags dissect_image_flags; bool relax_extension_release_check; int r; assert(src); /* Verifying release metadata requires mounted image for now, so ensure the check is skipped when * opening an image without mounting it immediately (i.e.: 'dest' is NULL). */ assert(!extension_release_data || dest); relax_extension_release_check = mount_options_relax_extension_release_checks(options); /* We might get an FD for the image, but we use the original path to look for the dm-verity files. * The caller might also give us a pre-loaded VeritySettings, in which case we just use it. It will * also be extended, as dissected_image_load_verity_sig_partition() is invoked. */ if (!verity) { r = verity_settings_load(&local_verity, src, NULL, NULL); if (r < 0) return log_debug_errno(r, "Failed to load root hash: %m"); verity = &local_verity; } dissect_image_flags = (verity->data_path ? DISSECT_IMAGE_NO_PARTITION_TABLE : 0) | (relax_extension_release_check ? DISSECT_IMAGE_RELAX_EXTENSION_CHECK : 0) | DISSECT_IMAGE_ADD_PARTITION_DEVICES | DISSECT_IMAGE_PIN_PARTITION_DEVICES | DISSECT_IMAGE_ALLOW_USERSPACE_VERITY | DISSECT_IMAGE_VERITY_SHARE; /* First check if we have a verity device already open and with a fstype pinned by policy. If it * cannot be found, then fallback to the slow path (full dissect). */ r = dissected_image_new_from_existing_verity( src_fd >= 0 ? FORMAT_PROC_FD_PATH(src_fd) : src, verity, options, image_policy, image_filter, runtime_scope, dissect_image_flags, &dissected_image); if (r < 0 && !ERRNO_IS_NEG_DEVICE_ABSENT(r) && r != -ENOPKG) return r; if (r >= 0) log_debug("Reusing pre-existing verity-protected image %s", src_fd >= 0 ? FORMAT_PROC_FD_PATH(src_fd) : src); else { if (runtime_scope == RUNTIME_SCOPE_SYSTEM) { /* Note that we don't use loop_device_make here, as the FD is most likely O_PATH which would not be * accepted by LOOP_CONFIGURE, so just let loop_device_make_by_path reopen it as a regular FD. */ r = loop_device_make_by_path( src_fd >= 0 ? FORMAT_PROC_FD_PATH(src_fd) : src, /* open_flags= */ -1, /* sector_size= */ UINT32_MAX, verity->data_path ? 0 : LO_FLAGS_PARTSCAN, LOCK_SH, &loop_device); if (r < 0) return log_debug_errno(r, "Failed to create loop device for image: %m"); r = dissect_loop_device( loop_device, verity, options, image_policy, image_filter, dissect_image_flags, &dissected_image); /* No partition table? Might be a single-filesystem image, try again */ if (!verity->data_path && r == -ENOPKG) r = dissect_loop_device( loop_device, verity, options, image_policy, image_filter, dissect_image_flags | DISSECT_IMAGE_NO_PARTITION_TABLE, &dissected_image); if (r < 0) return log_debug_errno(r, "Failed to dissect image: %m"); r = dissected_image_load_verity_sig_partition(dissected_image, loop_device->fd, verity); if (r < 0) return r; r = dissected_image_guess_verity_roothash(dissected_image, verity); if (r < 0) return r; r = dissected_image_decrypt( dissected_image, /* root= */ NULL, /* passphrase= */ NULL, verity, image_policy, dissect_image_flags); if (r < 0) return log_debug_errno(r, "Failed to decrypt dissected image: %m"); } else { userns_fd = namespace_open_by_type(NAMESPACE_USER); if (userns_fd < 0) return log_debug_errno(userns_fd, "Failed to open our own user namespace: %m"); r = mountfsd_mount_image( /* vl= */ NULL, src_fd >= 0 ? FORMAT_PROC_FD_PATH(src_fd) : src, userns_fd, options, image_policy, verity, dissect_image_flags, &dissected_image); if (r < 0) return r; } } if (dest) { r = mkdir_p_label(dest, 0755); if (r < 0) return log_debug_errno(r, "Failed to create destination directory %s: %m", dest); r = umount_recursive(dest, 0); if (r < 0) return log_debug_errno(r, "Failed to umount under destination directory %s: %m", dest); } r = dissected_image_mount( dissected_image, dest, /* uid_shift= */ UID_INVALID, /* uid_range= */ UID_INVALID, userns_fd, dissect_image_flags); if (r < 0) return log_debug_errno(r, "Failed to mount image: %m"); if (loop_device) { r = loop_device_flock(loop_device, LOCK_UN); if (r < 0) return log_debug_errno(r, "Failed to unlock loopback device: %m"); } /* If we got os-release values from the caller, then we need to match them with the image's * extension-release.d/ content. Return -EINVAL if there's any mismatch. * First, check the distro ID. If that matches, then check the new SYSEXT_LEVEL value if * available, or else fallback to VERSION_ID. If neither is present (eg: rolling release), * then a simple match on the ID will be performed. Also if an extension class was specified, * check that it matches or return ENOCSI (which looks like error-no-class if one squints enough). */ if ((extension_release_data && extension_release_data->os_release_id) || required_class >= 0) { _cleanup_strv_free_ char **extension_release = NULL; ImageClass class = IMAGE_SYSEXT; r = load_extension_release_pairs(dest, required_class >= 0 ? required_class : IMAGE_SYSEXT, dissected_image->image_name, relax_extension_release_check, &extension_release); if (r == -ENOENT) { if (required_class >= 0) return log_debug_errno(SYNTHETIC_ERRNO(ENOCSI), "Image %s extension-release metadata does not match the expected class", dissected_image->image_name); r = load_extension_release_pairs(dest, IMAGE_CONFEXT, dissected_image->image_name, relax_extension_release_check, &extension_release); if (r >= 0) class = IMAGE_CONFEXT; } if (r < 0) return log_debug_errno(r, "Failed to parse image %s extension-release metadata: %m", dissected_image->image_name); if (extension_release_data && !isempty(extension_release_data->os_release_id)) { r = extension_release_validate( dissected_image->image_name, extension_release_data->os_release_id, extension_release_data->os_release_id_like, extension_release_data->os_release_version_id, class == IMAGE_SYSEXT ? extension_release_data->os_release_sysext_level : extension_release_data->os_release_confext_level, extension_release_data->os_release_extension_scope, extension_release, class); if (r == 0) return log_debug_errno(SYNTHETIC_ERRNO(ESTALE), "Image %s extension-release metadata does not match the root's", dissected_image->image_name); if (r < 0) return log_debug_errno(r, "Failed to compare image %s extension-release metadata with the root's os-release: %m", dissected_image->image_name); } } r = dissected_image_relinquish(dissected_image); if (r < 0) return log_debug_errno(r, "Failed to relinquish dissected image: %m"); if (ret_image) *ret_image = TAKE_PTR(dissected_image); return 0; } void extension_release_data_done(ExtensionReleaseData *data) { assert(data); data->os_release_id = mfree(data->os_release_id); data->os_release_id_like = mfree(data->os_release_id_like); data->os_release_version_id = mfree(data->os_release_version_id); data->os_release_sysext_level = mfree(data->os_release_sysext_level); data->os_release_confext_level = mfree(data->os_release_confext_level); data->os_release_extension_scope = mfree(data->os_release_extension_scope); } int get_common_dissect_directory(char **ret) { _cleanup_free_ char *t = NULL; int r; /* A common location we mount dissected images to. The assumption is that everyone who uses this * function runs in their own private mount namespace (with mount propagation off on /run/systemd/, * and thus can mount something here without affecting anyone else). */ t = strdup("/run/systemd/dissect-root"); if (!t) return log_oom_debug(); r = mkdir_parents(t, 0755); if (r < 0) return log_debug_errno(r, "Failed to create parent dirs of mount point '%s': %m", t); r = RET_NERRNO(mkdir(t, 0000)); /* It's supposed to be overmounted, hence let's make this inaccessible */ if (r < 0 && r != -EEXIST) return log_debug_errno(r, "Failed to create mount point '%s': %m", t); if (ret) *ret = TAKE_PTR(t); return 0; } #if HAVE_BLKID static JSON_DISPATCH_ENUM_DEFINE(dispatch_architecture, Architecture, architecture_from_string); static JSON_DISPATCH_ENUM_DEFINE(dispatch_partition_designator, PartitionDesignator, partition_designator_from_string); typedef struct PartitionFields { PartitionDesignator designator; bool rw; bool growfs; unsigned partno; Architecture architecture; sd_id128_t uuid; char *fstype; char *label; uint64_t size; uint64_t offset; unsigned fsmount_fd_idx; } PartitionFields; static void partition_fields_done(PartitionFields *f) { assert(f); f->fstype = mfree(f->fstype); f->label = mfree(f->label); } typedef struct MountImageReplyParameters { sd_json_variant *partitions; bool single_file_system; char *image_policy; uint64_t image_size; uint32_t sector_size; sd_id128_t image_uuid; } MountImageReplyParameters; static void mount_image_reply_parameters_done(MountImageReplyParameters *p) { assert(p); p->image_policy = mfree(p->image_policy); p->partitions = sd_json_variant_unref(p->partitions); } #endif int mountfsd_connect(sd_varlink **ret) { int r; assert(ret); _cleanup_(sd_varlink_unrefp) sd_varlink *vl = NULL; r = sd_varlink_connect_address(&vl, "/run/systemd/io.systemd.MountFileSystem"); if (r < 0) return log_debug_errno(r, "Failed to connect to mountfsd: %m"); r = sd_varlink_set_allow_fd_passing_input(vl, true); if (r < 0) return log_debug_errno(r, "Failed to enable varlink fd passing for read: %m"); r = sd_varlink_set_allow_fd_passing_output(vl, true); if (r < 0) return log_debug_errno(r, "Failed to enable varlink fd passing for write: %m"); *ret = TAKE_PTR(vl); return 0; } int mountfsd_mount_image_fd( sd_varlink *vl, int image_fd, int userns_fd, const MountOptions *options, const ImagePolicy *image_policy, const VeritySettings *verity, DissectImageFlags flags, DissectedImage **ret) { #if HAVE_BLKID _cleanup_(mount_image_reply_parameters_done) MountImageReplyParameters p = {}; static const sd_json_dispatch_field dispatch_table[] = { { "partitions", SD_JSON_VARIANT_ARRAY, sd_json_dispatch_variant, offsetof(struct MountImageReplyParameters, partitions), SD_JSON_MANDATORY }, { "singleFileSystem", SD_JSON_VARIANT_BOOLEAN, sd_json_dispatch_stdbool, offsetof(struct MountImageReplyParameters, single_file_system), 0 }, { "imagePolicy", SD_JSON_VARIANT_STRING, sd_json_dispatch_string, offsetof(struct MountImageReplyParameters, image_policy), 0 }, { "imageSize", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint64, offsetof(struct MountImageReplyParameters, image_size), SD_JSON_MANDATORY }, { "sectorSize", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint32, offsetof(struct MountImageReplyParameters, sector_size), SD_JSON_MANDATORY }, { "imageUuid", SD_JSON_VARIANT_STRING, sd_json_dispatch_id128, offsetof(struct MountImageReplyParameters, image_uuid), 0 }, {} }; _cleanup_(dissected_image_unrefp) DissectedImage *di = NULL; _cleanup_close_ int verity_data_fd = -EBADF; _cleanup_free_ char *ps = NULL; const char *error_id; int r; assert(image_fd >= 0); assert(ret); _cleanup_(sd_varlink_unrefp) sd_varlink *_vl = NULL; if (!vl) { r = mountfsd_connect(&_vl); if (r < 0) return r; vl = _vl; } _cleanup_close_ int reopened_fd = -EBADF; image_fd = fd_reopen_condition(image_fd, O_CLOEXEC|O_NOCTTY|O_NONBLOCK|(FLAGS_SET(flags, DISSECT_IMAGE_MOUNT_READ_ONLY) ? O_RDONLY : O_RDWR), O_PATH, &reopened_fd); if (image_fd < 0) return log_debug_errno(image_fd, "Failed to reopen fd: %m"); r = sd_varlink_push_dup_fd(vl, image_fd); if (r < 0) return log_debug_errno(r, "Failed to push image fd into varlink connection: %m"); if (userns_fd >= 0) { r = sd_varlink_push_dup_fd(vl, userns_fd); if (r < 0) return log_debug_errno(r, "Failed to push image fd into varlink connection: %m"); } if (image_policy) { r = image_policy_to_string(image_policy, /* simplify= */ false, &ps); if (r < 0) return log_debug_errno(r, "Failed to format image policy to string: %m"); } if (verity && verity->data_path) { verity_data_fd = open(verity->data_path, O_RDONLY|O_CLOEXEC); if (verity_data_fd < 0) return log_debug_errno(errno, "Failed to open verity data file '%s': %m", verity->data_path); r = sd_varlink_push_dup_fd(vl, verity_data_fd); if (r < 0) return log_debug_errno(r, "Failed to push verity data fd into varlink connection: %m"); } _cleanup_(sd_json_variant_unrefp) sd_json_variant *mount_options = NULL; for (PartitionDesignator i = 0; i < _PARTITION_DESIGNATOR_MAX; i++) { _cleanup_free_ char *filtered = NULL; const char *o = mount_options_from_designator(options, i); if (!o) continue; /* We communicate relaxExtensionReleaseCheck separately via the varlink API, so filter it out * from the mount options we pass to mountfsd. */ if (IN_SET(i, PARTITION_ROOT, PARTITION_USR)) { r = fstab_filter_options( o, "x-systemd.relax-extension-release-check\0", /* ret_namefound= */ NULL, /* ret_value= */ NULL, /* ret_values= */ NULL, &filtered); if (r < 0) return log_debug_errno(r, "Failed to filter mount options: %m"); if (isempty(filtered)) continue; } r = sd_json_variant_merge_objectbo( &mount_options, SD_JSON_BUILD_PAIR_STRING(partition_designator_to_string(i), filtered ?: o)); if (r < 0) return log_debug_errno(r, "Failed to build mount options array: %m"); } sd_json_variant *reply = NULL; r = varlink_callbo_and_log( vl, "io.systemd.MountFileSystem.MountImage", &reply, &error_id, SD_JSON_BUILD_PAIR_UNSIGNED("imageFileDescriptor", 0), SD_JSON_BUILD_PAIR_CONDITION(userns_fd >= 0, "userNamespaceFileDescriptor", SD_JSON_BUILD_UNSIGNED(1)), SD_JSON_BUILD_PAIR_BOOLEAN("readOnly", FLAGS_SET(flags, DISSECT_IMAGE_MOUNT_READ_ONLY)), SD_JSON_BUILD_PAIR_BOOLEAN("growFileSystems", FLAGS_SET(flags, DISSECT_IMAGE_GROWFS)), SD_JSON_BUILD_PAIR_CONDITION(!!ps, "imagePolicy", SD_JSON_BUILD_STRING(ps)), JSON_BUILD_PAIR_VARIANT_NON_NULL("mountOptions", mount_options), SD_JSON_BUILD_PAIR_BOOLEAN("relaxExtensionReleaseChecks", mount_options_relax_extension_release_checks(options)), SD_JSON_BUILD_PAIR_BOOLEAN("veritySharing", FLAGS_SET(flags, DISSECT_IMAGE_VERITY_SHARE)), SD_JSON_BUILD_PAIR_CONDITION(verity_data_fd >= 0, "verityDataFileDescriptor", SD_JSON_BUILD_UNSIGNED(userns_fd >= 0 ? 2 : 1)), SD_JSON_BUILD_PAIR_CONDITION(verity && iovec_is_set(&verity->root_hash), "verityRootHash", JSON_BUILD_IOVEC_HEX(&verity->root_hash)), SD_JSON_BUILD_PAIR_CONDITION(verity && iovec_is_set(&verity->root_hash_sig), "verityRootHashSignature", JSON_BUILD_IOVEC_BASE64(&verity->root_hash_sig)), SD_JSON_BUILD_PAIR_BOOLEAN("allowInteractiveAuthentication", FLAGS_SET(flags, DISSECT_IMAGE_ALLOW_INTERACTIVE_AUTH))); if (r < 0) return r; r = sd_json_dispatch(reply, dispatch_table, SD_JSON_ALLOW_EXTENSIONS, &p); if (r < 0) return log_debug_errno(r, "Failed to parse MountImage() reply: %m"); log_debug("Effective image policy: %s", p.image_policy); sd_json_variant *i; JSON_VARIANT_ARRAY_FOREACH(i, p.partitions) { _cleanup_close_ int fsmount_fd = -EBADF; _cleanup_(partition_fields_done) PartitionFields pp = { .designator = _PARTITION_DESIGNATOR_INVALID, .architecture = _ARCHITECTURE_INVALID, .size = UINT64_MAX, .offset = UINT64_MAX, .fsmount_fd_idx = UINT_MAX, }; static const sd_json_dispatch_field partition_dispatch_table[] = { { "designator", SD_JSON_VARIANT_STRING, dispatch_partition_designator, offsetof(struct PartitionFields, designator), SD_JSON_MANDATORY }, { "writable", SD_JSON_VARIANT_BOOLEAN, sd_json_dispatch_stdbool, offsetof(struct PartitionFields, rw), SD_JSON_MANDATORY }, { "growFileSystem", SD_JSON_VARIANT_BOOLEAN, sd_json_dispatch_stdbool, offsetof(struct PartitionFields, growfs), SD_JSON_MANDATORY }, { "partitionNumber", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint, offsetof(struct PartitionFields, partno), 0 }, { "architecture", SD_JSON_VARIANT_STRING, dispatch_architecture, offsetof(struct PartitionFields, architecture), 0 }, { "partitionUuid", SD_JSON_VARIANT_STRING, sd_json_dispatch_id128, offsetof(struct PartitionFields, uuid), 0 }, { "fileSystemType", SD_JSON_VARIANT_STRING, sd_json_dispatch_string, offsetof(struct PartitionFields, fstype), SD_JSON_MANDATORY }, { "partitionLabel", SD_JSON_VARIANT_STRING, sd_json_dispatch_string, offsetof(struct PartitionFields, label), 0 }, { "size", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint64, offsetof(struct PartitionFields, size), SD_JSON_MANDATORY }, { "offset", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint64, offsetof(struct PartitionFields, offset), SD_JSON_MANDATORY }, { "mountFileDescriptor", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint, offsetof(struct PartitionFields, fsmount_fd_idx), SD_JSON_MANDATORY }, {} }; r = sd_json_dispatch(i, partition_dispatch_table, SD_JSON_ALLOW_EXTENSIONS, &pp); if (r < 0) return log_debug_errno(r, "Failed to parse partition data: %m"); if (pp.fsmount_fd_idx != UINT_MAX) { fsmount_fd = sd_varlink_take_fd(vl, pp.fsmount_fd_idx); if (fsmount_fd < 0) return fsmount_fd; } assert(pp.designator >= 0); if (!di) { r = dissected_image_new(/* path= */ NULL, &di); if (r < 0) return log_debug_errno(r, "Failed to allocated new dissected image structure: %m"); } if (di->partitions[pp.designator].found) return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), "Duplicate partition data for '%s'.", partition_designator_to_string(pp.designator)); di->partitions[pp.designator] = (DissectedPartition) { .found = true, .rw = pp.rw, .growfs = pp.growfs, .partno = pp.partno, .architecture = pp.architecture, .uuid = pp.uuid, .fstype = TAKE_PTR(pp.fstype), .label = TAKE_PTR(pp.label), .mount_node_fd = -EBADF, .size = pp.size, .offset = pp.offset, .fsmount_fd = TAKE_FD(fsmount_fd), }; } if (!di) return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), "Partition list is empty."); di->single_file_system = p.single_file_system; di->image_size = p.image_size; di->sector_size = p.sector_size; di->image_uuid = p.image_uuid; *ret = TAKE_PTR(di); return 0; #else return -EOPNOTSUPP; #endif } int mountfsd_mount_image( sd_varlink *vl, const char *path, int userns_fd, const MountOptions *options, const ImagePolicy *image_policy, const VeritySettings *verity, DissectImageFlags flags, DissectedImage **ret) { int r; assert(path); assert(ret); _cleanup_close_ int image_fd = open(path, O_RDONLY|O_CLOEXEC); if (image_fd < 0) return log_debug_errno(errno, "Failed to open '%s': %m", path); _cleanup_(dissected_image_unrefp) DissectedImage *di = NULL; r = mountfsd_mount_image_fd(vl, image_fd, userns_fd, options, image_policy, verity, flags, &di); if (r < 0) return r; if (!di->image_name) { r = dissected_image_name_from_path(path, &di->image_name); if (r < 0) return r; } *ret = TAKE_PTR(di); return 0; } int mountfsd_mount_directory_fd( sd_varlink *vl, int directory_fd, int userns_fd, DissectImageFlags flags, int *ret_mount_fd) { int r; assert(directory_fd >= 0); assert(ret_mount_fd); /* Pick one identity, not both, that makes no sense. */ assert(!FLAGS_SET(flags, DISSECT_IMAGE_FOREIGN_UID|DISSECT_IMAGE_IDENTITY_UID)); _cleanup_(sd_varlink_unrefp) sd_varlink *_vl = NULL; if (!vl) { r = mountfsd_connect(&_vl); if (r < 0) return r; vl = _vl; } r = sd_varlink_push_dup_fd(vl, directory_fd); if (r < 0) return log_debug_errno(r, "Failed to push directory fd into varlink connection: %m"); if (userns_fd >= 0) { r = sd_varlink_push_dup_fd(vl, userns_fd); if (r < 0) return log_debug_errno(r, "Failed to push user namespace fd into varlink connection: %m"); } sd_json_variant *reply = NULL; const char *error_id = NULL; r = varlink_callbo_and_log( vl, "io.systemd.MountFileSystem.MountDirectory", &reply, &error_id, SD_JSON_BUILD_PAIR_UNSIGNED("directoryFileDescriptor", 0), SD_JSON_BUILD_PAIR_CONDITION(userns_fd >= 0, "userNamespaceFileDescriptor", SD_JSON_BUILD_UNSIGNED(1)), SD_JSON_BUILD_PAIR_BOOLEAN("readOnly", FLAGS_SET(flags, DISSECT_IMAGE_MOUNT_READ_ONLY)), SD_JSON_BUILD_PAIR_STRING("mode", FLAGS_SET(flags, DISSECT_IMAGE_FOREIGN_UID) ? "foreign" : FLAGS_SET(flags, DISSECT_IMAGE_IDENTITY_UID) ? "identity" : "auto"), SD_JSON_BUILD_PAIR_BOOLEAN("allowInteractiveAuthentication", FLAGS_SET(flags, DISSECT_IMAGE_ALLOW_INTERACTIVE_AUTH))); if (r < 0) return r; static const sd_json_dispatch_field dispatch_table[] = { { "mountFileDescriptor", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint, 0, SD_JSON_MANDATORY }, {} }; unsigned fsmount_fd_idx = UINT_MAX; r = sd_json_dispatch(reply, dispatch_table, SD_JSON_ALLOW_EXTENSIONS, &fsmount_fd_idx); if (r < 0) return log_debug_errno(r, "Failed to parse MountImage() reply: %m"); _cleanup_close_ int fsmount_fd = sd_varlink_take_fd(vl, fsmount_fd_idx); if (fsmount_fd < 0) return log_debug_errno(fsmount_fd, "Failed to take mount fd from Varlink connection: %m"); *ret_mount_fd = TAKE_FD(fsmount_fd); return 0; } int mountfsd_mount_directory( sd_varlink *vl, const char *path, int userns_fd, DissectImageFlags flags, int *ret_mount_fd) { assert(path); assert(ret_mount_fd); _cleanup_close_ int directory_fd = open(path, O_DIRECTORY|O_RDONLY|O_CLOEXEC|O_PATH); if (directory_fd < 0) return log_debug_errno(errno, "Failed to open '%s': %m", path); return mountfsd_mount_directory_fd(vl, directory_fd, userns_fd, flags, ret_mount_fd); } int mountfsd_make_directory_fd( sd_varlink *vl, int parent_fd, const char *name, mode_t mode, DissectImageFlags flags, int *ret_directory_fd) { int r; assert(parent_fd >= 0); assert(name); _cleanup_(sd_varlink_unrefp) sd_varlink *_vl = NULL; if (!vl) { r = mountfsd_connect(&_vl); if (r < 0) return r; vl = _vl; } r = sd_varlink_push_dup_fd(vl, parent_fd); if (r < 0) return log_debug_errno(r, "Failed to push parent fd into varlink connection: %m"); sd_json_variant *reply = NULL; const char *error_id = NULL; r = varlink_callbo_and_log( vl, "io.systemd.MountFileSystem.MakeDirectory", &reply, &error_id, SD_JSON_BUILD_PAIR_UNSIGNED("parentFileDescriptor", 0), SD_JSON_BUILD_PAIR_STRING("name", name), SD_JSON_BUILD_PAIR_CONDITION(!IN_SET(mode, MODE_INVALID, 0700), "mode", SD_JSON_BUILD_UNSIGNED(mode)), /* suppress this field if default/unset */ SD_JSON_BUILD_PAIR_BOOLEAN("allowInteractiveAuthentication", FLAGS_SET(flags, DISSECT_IMAGE_ALLOW_INTERACTIVE_AUTH))); if (r < 0) return r; static const sd_json_dispatch_field dispatch_table[] = { { "directoryFileDescriptor", _SD_JSON_VARIANT_TYPE_INVALID, sd_json_dispatch_uint, 0, SD_JSON_MANDATORY }, {} }; unsigned directory_fd_idx = UINT_MAX; r = sd_json_dispatch(reply, dispatch_table, SD_JSON_ALLOW_EXTENSIONS, &directory_fd_idx); if (r < 0) return log_debug_errno(r, "Failed to parse MountImage() reply: %m"); _cleanup_close_ int directory_fd = sd_varlink_take_fd(vl, directory_fd_idx); if (directory_fd < 0) return log_debug_errno(directory_fd, "Failed to take directory fd from Varlink connection: %m"); if (ret_directory_fd) *ret_directory_fd = TAKE_FD(directory_fd); return 0; } int mountfsd_make_directory( sd_varlink *vl, const char *path, mode_t mode, DissectImageFlags flags, int *ret_directory_fd) { int r; _cleanup_free_ char *parent = NULL; r = path_extract_directory(path, &parent); if (r < 0) return log_debug_errno(r, "Failed to extract parent directory from '%s': %m", path); _cleanup_free_ char *dirname = NULL; r = path_extract_filename(path, &dirname); if (r < 0) return log_debug_errno(r, "Failed to extract directory name from '%s': %m", path); _cleanup_close_ int fd = open(parent, O_DIRECTORY|O_CLOEXEC); if (fd < 0) return log_debug_errno(errno, "Failed to open '%s': %m", parent); return mountfsd_make_directory_fd(vl, fd, dirname, mode, flags, ret_directory_fd); } int copy_tree_at_foreign(int source_fd, int target_fd, int userns_fd) { int r; assert(source_fd >= 0); assert(target_fd >= 0); assert(userns_fd >= 0); /* Copies dir referenced by source_fd into dir referenced by source_fd, moves to the specified userns * for that, which should be foreign UID range */ r = pidref_safe_fork_full( "copy-tree", /* stdio_fds= */ NULL, (int[]) { userns_fd, source_fd, target_fd }, 3, FORK_RESET_SIGNALS|FORK_CLOSE_ALL_FDS|FORK_DEATHSIG_SIGTERM|FORK_WAIT|FORK_REOPEN_LOG, /* ret= */ NULL); if (r < 0) return r; if (r == 0) { r = namespace_enter( /* pidns_fd= */ -EBADF, /* mntns_fd= */ -EBADF, /* netns_fd= */ -EBADF, userns_fd, /* root_fd= */ -EBADF); if (r < 0) { log_debug_errno(r, "Failed to join user namespace: %m"); _exit(EXIT_FAILURE); } r = copy_tree_at( source_fd, /* from= */ NULL, target_fd, /* to= */ NULL, /* override_uid= */ UID_INVALID, /* override_gid= */ GID_INVALID, COPY_REFLINK|COPY_HARDLINKS|COPY_MERGE_EMPTY|COPY_MERGE_APPLY_STAT|COPY_SAME_MOUNT|COPY_ALL_XATTRS, /* denylist= */ NULL, /* subvolumes= */ NULL); if (r < 0) { log_debug_errno(r, "Failed to copy tree: %m"); _exit(EXIT_FAILURE); } _exit(EXIT_SUCCESS); } return 0; } int remove_tree_foreign(const char *path, int userns_fd) { int r; assert(path); assert(userns_fd >= 0); _cleanup_close_ int tree_fd = -EBADF; r = mountfsd_mount_directory( /* vl= */ NULL, path, userns_fd, DISSECT_IMAGE_FOREIGN_UID, &tree_fd); if (r < 0) return r; r = pidref_safe_fork_full( "rm-tree", /* stdio_fds= */ NULL, (int[]) { userns_fd, tree_fd }, 2, FORK_RESET_SIGNALS|FORK_CLOSE_ALL_FDS|FORK_DEATHSIG_SIGTERM|FORK_WAIT|FORK_REOPEN_LOG, /* ret= */ NULL); if (r < 0) return r; if (r == 0) { /* child */ r = namespace_enter( /* pidns_fd= */ -EBADF, /* mntns_fd= */ -EBADF, /* netns_fd= */ -EBADF, userns_fd, /* root_fd= */ -EBADF); if (r < 0) { log_debug_errno(r, "Failed to join user namespace: %m"); _exit(EXIT_FAILURE); } _cleanup_close_ int dfd = fd_reopen(tree_fd, O_DIRECTORY|O_CLOEXEC); if (dfd < 0) { log_debug_errno(r, "Failed to reopen tree fd: %m"); _exit(EXIT_FAILURE); } r = rm_rf_children(dfd, REMOVE_PHYSICAL|REMOVE_SUBVOLUME|REMOVE_CHMOD, /* root_dev= */ NULL); if (r < 0) log_debug_errno(r, "Failed to empty '%s' directory in foreign UID mode, ignoring: %m", path); _exit(EXIT_SUCCESS); } return 0; }