| 1 | #define DISABLE_SIGN_COMPARE_WARNINGS |
| 2 | |
| 3 | #include "git-compat-util.h" |
| 4 | #include "environment.h" |
| 5 | #include "hex.h" |
| 6 | #include "repository.h" |
| 7 | #include "pack.h" |
| 8 | #include "progress.h" |
| 9 | #include "packfile.h" |
| 10 | #include "object-file.h" |
| 11 | #include "odb.h" |
| 12 | #include "odb/streaming.h" |
| 13 | |
| 14 | struct idx_entry { |
| 15 | off_t offset; |
| 16 | unsigned int nr; |
| 17 | }; |
| 18 | |
| 19 | static int compare_entries(const void *e1, const void *e2) |
| 20 | { |
| 21 | const struct idx_entry *entry1 = e1; |
| 22 | const struct idx_entry *entry2 = e2; |
| 23 | if (entry1->offset < entry2->offset) |
| 24 | return -1; |
| 25 | if (entry1->offset > entry2->offset) |
| 26 | return 1; |
| 27 | return 0; |
| 28 | } |
| 29 | |
| 30 | int check_pack_crc(struct packed_git *p, struct pack_window **w_curs, |
| 31 | off_t offset, off_t len, unsigned int nr) |
| 32 | { |
| 33 | const uint32_t *index_crc; |
| 34 | uint32_t data_crc = crc32(0, NULL, 0); |
| 35 | |
| 36 | do { |
| 37 | unsigned long avail; |
| 38 | void *data = use_pack(p, w_curs, offset, &avail); |
| 39 | if (avail > len) |
| 40 | avail = len; |
| 41 | data_crc = crc32(data_crc, data, avail); |
| 42 | offset += avail; |
| 43 | len -= avail; |
| 44 | } while (len); |
| 45 | |
| 46 | index_crc = p->index_data; |
| 47 | index_crc += 2 + 256 + (size_t)p->num_objects * (p->repo->hash_algo->rawsz/4) + nr; |
| 48 | |
| 49 | return data_crc != ntohl(*index_crc); |
| 50 | } |
| 51 | |
| 52 | static int verify_packfile(struct repository *r, |
| 53 | struct packed_git *p, |
| 54 | struct pack_window **w_curs, |
| 55 | verify_fn fn, |
| 56 | void *fn_data, |
| 57 | struct progress *progress, uint32_t base_count) |
| 58 | |
| 59 | { |
| 60 | off_t index_size = p->index_size; |
| 61 | const unsigned char *index_base = p->index_data; |
| 62 | struct git_hash_ctx ctx; |
| 63 | unsigned char hash[GIT_MAX_RAWSZ], *pack_sig; |
| 64 | off_t offset = 0, pack_sig_ofs = 0; |
| 65 | uint32_t nr_objects, i; |
| 66 | int err = 0; |
| 67 | struct idx_entry *entries; |
| 68 | |
| 69 | if (!is_pack_valid(p)) |
| 70 | return error("packfile %s cannot be accessed", p->pack_name); |
| 71 | |
| 72 | r->hash_algo->init_fn(&ctx); |
| 73 | do { |
| 74 | unsigned long remaining; |
| 75 | unsigned char *in = use_pack(p, w_curs, offset, &remaining); |
| 76 | offset += remaining; |
| 77 | if (!pack_sig_ofs) |
| 78 | pack_sig_ofs = p->pack_size - r->hash_algo->rawsz; |
| 79 | if (offset > pack_sig_ofs) |
| 80 | remaining -= (unsigned int)(offset - pack_sig_ofs); |
| 81 | git_hash_update(&ctx, in, remaining); |
| 82 | } while (offset < pack_sig_ofs); |
| 83 | git_hash_final(hash, &ctx); |
| 84 | pack_sig = use_pack(p, w_curs, pack_sig_ofs, NULL); |
| 85 | if (!hasheq(hash, pack_sig, r->hash_algo)) |
| 86 | err = error("%s pack checksum mismatch", |
| 87 | p->pack_name); |
| 88 | if (!hasheq(index_base + index_size - r->hash_algo->hexsz, pack_sig, |
| 89 | r->hash_algo)) |
| 90 | err = error("%s pack checksum does not match its index", |
| 91 | p->pack_name); |
| 92 | unuse_pack(w_curs); |
| 93 | |
| 94 | /* Make sure everything reachable from idx is valid. Since we |
| 95 | * have verified that nr_objects matches between idx and pack, |
| 96 | * we do not do scan-streaming check on the pack file. |
| 97 | */ |
| 98 | nr_objects = p->num_objects; |
| 99 | ALLOC_ARRAY(entries, nr_objects + 1); |
| 100 | entries[nr_objects].offset = pack_sig_ofs; |
| 101 | /* first sort entries by pack offset, since unpacking them is more efficient that way */ |
| 102 | for (i = 0; i < nr_objects; i++) { |
| 103 | entries[i].offset = nth_packed_object_offset(p, i); |
| 104 | entries[i].nr = i; |
| 105 | } |
| 106 | QSORT(entries, nr_objects, compare_entries); |
| 107 | |
| 108 | for (i = 0; i < nr_objects; i++) { |
| 109 | struct odb_read_stream *stream = NULL; |
| 110 | void *data; |
| 111 | struct object_id oid; |
| 112 | enum object_type type; |
| 113 | size_t size; |
| 114 | off_t curpos; |
| 115 | int data_valid; |
| 116 | |
| 117 | if (nth_packed_object_id(&oid, p, entries[i].nr) < 0) |
| 118 | BUG("unable to get oid of object %lu from %s", |
| 119 | (unsigned long)entries[i].nr, p->pack_name); |
| 120 | |
| 121 | if (p->index_version > 1) { |
| 122 | off_t offset = entries[i].offset; |
| 123 | off_t len = entries[i+1].offset - offset; |
| 124 | unsigned int nr = entries[i].nr; |
| 125 | if (check_pack_crc(p, w_curs, offset, len, nr)) |
| 126 | err = error("index CRC mismatch for object %s " |
| 127 | "from %s at offset %"PRIuMAX"", |
| 128 | oid_to_hex(&oid), |
| 129 | p->pack_name, (uintmax_t)offset); |
| 130 | } |
| 131 | |
| 132 | curpos = entries[i].offset; |
| 133 | type = unpack_object_header(p, w_curs, &curpos, &size); |
| 134 | unuse_pack(w_curs); |
| 135 | |
| 136 | if (type == OBJ_BLOB && |
| 137 | repo_settings_get_big_file_threshold(r) <= size) { |
| 138 | /* |
| 139 | * Let stream_object_signature() check it with |
| 140 | * the streaming interface; no point slurping |
| 141 | * the data in-core only to discard. |
| 142 | */ |
| 143 | data = NULL; |
| 144 | data_valid = 0; |
| 145 | } else { |
| 146 | data = unpack_entry(r, p, entries[i].offset, &type, |
| 147 | &size); |
| 148 | data_valid = 1; |
| 149 | } |
| 150 | |
| 151 | if (data_valid && !data) |
| 152 | err = error("cannot unpack %s from %s at offset %"PRIuMAX"", |
| 153 | oid_to_hex(&oid), p->pack_name, |
| 154 | (uintmax_t)entries[i].offset); |
| 155 | else if (data && check_object_signature(r, &oid, data, size, |
| 156 | type) < 0) |
| 157 | err = error("packed %s from %s is corrupt", |
| 158 | oid_to_hex(&oid), p->pack_name); |
| 159 | else if (!data && |
| 160 | (packfile_read_object_stream(&stream, &oid, p, entries[i].offset) < 0 || |
| 161 | stream_object_signature(r, stream, &oid) < 0)) |
| 162 | err = error("packed %s from %s is corrupt", |
| 163 | oid_to_hex(&oid), p->pack_name); |
| 164 | else if (fn) { |
| 165 | int eaten = 0; |
| 166 | err |= fn(&oid, type, size, data, &eaten, fn_data); |
| 167 | if (eaten) |
| 168 | data = NULL; |
| 169 | } |
| 170 | if (((base_count + i) & 1023) == 0) |
| 171 | display_progress(progress, base_count + i); |
| 172 | |
| 173 | if (stream) |
| 174 | odb_read_stream_close(stream); |
| 175 | free(data); |
| 176 | } |
| 177 | |
| 178 | display_progress(progress, base_count + i); |
| 179 | free(entries); |
| 180 | return err; |
| 181 | } |
| 182 | |
| 183 | int verify_pack_index(struct packed_git *p) |
| 184 | { |
| 185 | int err = 0; |
| 186 | |
| 187 | if (open_pack_index(p)) |
| 188 | return error("packfile %s index not opened", p->pack_name); |
| 189 | |
| 190 | /* Verify SHA1 sum of the index file */ |
| 191 | if (!hashfile_checksum_valid(p->repo->hash_algo, p->index_data, p->index_size)) |
| 192 | err = error("Packfile index for %s hash mismatch", |
| 193 | p->pack_name); |
| 194 | return err; |
| 195 | } |
| 196 | |
| 197 | int verify_pack(struct repository *r, struct packed_git *p, verify_fn fn, void *fn_data, |
| 198 | struct progress *progress, uint32_t base_count) |
| 199 | { |
| 200 | int err = 0; |
| 201 | struct pack_window *w_curs = NULL; |
| 202 | |
| 203 | err |= verify_pack_index(p); |
| 204 | if (!p->index_data) |
| 205 | return -1; |
| 206 | |
| 207 | err |= verify_packfile(r, p, &w_curs, fn, fn_data, progress, base_count); |
| 208 | unuse_pack(&w_curs); |
| 209 | |
| 210 | return err; |
| 211 | } |