]>
Commit | Line | Data |
---|---|---|
0fcfd160 LT |
1 | /* |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | * | |
e5afd444 | 6 | * This handles basic git object files - packing, unpacking, |
0fcfd160 LT |
7 | * creation etc. |
8 | */ | |
0fcfd160 | 9 | #include "cache.h" |
b2141fc1 | 10 | #include "config.h" |
6eac50d8 | 11 | #include "string-list.h" |
697cc8ef | 12 | #include "lockfile.h" |
1f688557 | 13 | #include "delta.h" |
a733cb60 | 14 | #include "pack.h" |
8e440259 PE |
15 | #include "blob.h" |
16 | #include "commit.h" | |
4dd1fbc7 | 17 | #include "run-command.h" |
8e440259 PE |
18 | #include "tag.h" |
19 | #include "tree.h" | |
c879daa2 | 20 | #include "tree-walk.h" |
f35a6d3b | 21 | #include "refs.h" |
70f5d5d3 | 22 | #include "pack-revindex.h" |
bc626927 | 23 | #include "hash-lookup.h" |
568508e7 | 24 | #include "bulk-checkin.h" |
031dc927 | 25 | #include "repository.h" |
47f351e9 | 26 | #include "replace-object.h" |
090ea126 | 27 | #include "streaming.h" |
543c5caa | 28 | #include "dir.h" |
12d95ef6 | 29 | #include "list.h" |
c4c6effa | 30 | #include "mergesort.h" |
cf3c6352 | 31 | #include "quote.h" |
4f39cd82 | 32 | #include "packfile.h" |
90c62155 | 33 | #include "object-store.h" |
b14ed5ad | 34 | #include "promisor-remote.h" |
a35e03de | 35 | #include "submodule.h" |
e05db0fd | 36 | |
1af64f73 | 37 | /* The maximum size for an object header. */ |
38 | #define MAX_HEADER_LEN 32 | |
39 | ||
e1ccd7e2 | 40 | |
41 | #define EMPTY_TREE_SHA1_BIN_LITERAL \ | |
42 | "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60" \ | |
43 | "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04" | |
13eeedb5 | 44 | #define EMPTY_TREE_SHA256_BIN_LITERAL \ |
45 | "\x6e\xf1\x9b\x41\x22\x5c\x53\x69\xf1\xc1" \ | |
46 | "\x04\xd4\x5d\x8d\x85\xef\xa9\xb0\x57\xb5" \ | |
47 | "\x3b\x14\xb4\xb9\xb9\x39\xdd\x74\xde\xcc" \ | |
48 | "\x53\x21" | |
e1ccd7e2 | 49 | |
50 | #define EMPTY_BLOB_SHA1_BIN_LITERAL \ | |
51 | "\xe6\x9d\xe2\x9b\xb2\xd1\xd6\x43\x4b\x8b" \ | |
52 | "\x29\xae\x77\x5a\xd8\xc2\xe4\x8c\x53\x91" | |
13eeedb5 | 53 | #define EMPTY_BLOB_SHA256_BIN_LITERAL \ |
54 | "\x47\x3a\x0f\x4c\x3b\xe8\xa9\x36\x81\xa2" \ | |
55 | "\x67\xe3\xb1\xe9\xa7\xdc\xda\x11\x85\x43" \ | |
56 | "\x6f\xe1\x41\xf7\x74\x91\x20\xa3\x03\x72" \ | |
57 | "\x18\x13" | |
e1ccd7e2 | 58 | |
e1ccd7e2 | 59 | static const struct object_id empty_tree_oid = { |
5a6dce70 | 60 | .hash = EMPTY_TREE_SHA1_BIN_LITERAL, |
61 | .algo = GIT_HASH_SHA1, | |
8576fde6 | 62 | }; |
e1ccd7e2 | 63 | static const struct object_id empty_blob_oid = { |
5a6dce70 | 64 | .hash = EMPTY_BLOB_SHA1_BIN_LITERAL, |
65 | .algo = GIT_HASH_SHA1, | |
8576fde6 | 66 | }; |
14228447 | 67 | static const struct object_id null_oid_sha1 = { |
68 | .hash = {0}, | |
69 | .algo = GIT_HASH_SHA1, | |
70 | }; | |
13eeedb5 | 71 | static const struct object_id empty_tree_oid_sha256 = { |
5a6dce70 | 72 | .hash = EMPTY_TREE_SHA256_BIN_LITERAL, |
73 | .algo = GIT_HASH_SHA256, | |
13eeedb5 | 74 | }; |
75 | static const struct object_id empty_blob_oid_sha256 = { | |
5a6dce70 | 76 | .hash = EMPTY_BLOB_SHA256_BIN_LITERAL, |
77 | .algo = GIT_HASH_SHA256, | |
13eeedb5 | 78 | }; |
14228447 | 79 | static const struct object_id null_oid_sha256 = { |
80 | .hash = {0}, | |
81 | .algo = GIT_HASH_SHA256, | |
82 | }; | |
88cd621d | 83 | |
ac73cedf | 84 | static void git_hash_sha1_init(git_hash_ctx *ctx) |
f50e766b | 85 | { |
ac73cedf | 86 | git_SHA1_Init(&ctx->sha1); |
f50e766b | 87 | } |
88 | ||
768e30ea | 89 | static void git_hash_sha1_clone(git_hash_ctx *dst, const git_hash_ctx *src) |
90 | { | |
91 | git_SHA1_Clone(&dst->sha1, &src->sha1); | |
92 | } | |
93 | ||
ac73cedf | 94 | static void git_hash_sha1_update(git_hash_ctx *ctx, const void *data, size_t len) |
f50e766b | 95 | { |
ac73cedf | 96 | git_SHA1_Update(&ctx->sha1, data, len); |
f50e766b | 97 | } |
98 | ||
ac73cedf | 99 | static void git_hash_sha1_final(unsigned char *hash, git_hash_ctx *ctx) |
f50e766b | 100 | { |
ac73cedf | 101 | git_SHA1_Final(hash, &ctx->sha1); |
f50e766b | 102 | } |
103 | ||
ab795f0d | 104 | static void git_hash_sha1_final_oid(struct object_id *oid, git_hash_ctx *ctx) |
105 | { | |
106 | git_SHA1_Final(oid->hash, &ctx->sha1); | |
107 | memset(oid->hash + GIT_SHA1_RAWSZ, 0, GIT_MAX_RAWSZ - GIT_SHA1_RAWSZ); | |
5a6dce70 | 108 | oid->algo = GIT_HASH_SHA1; |
ab795f0d | 109 | } |
110 | ||
13eeedb5 | 111 | |
112 | static void git_hash_sha256_init(git_hash_ctx *ctx) | |
113 | { | |
114 | git_SHA256_Init(&ctx->sha256); | |
115 | } | |
116 | ||
768e30ea | 117 | static void git_hash_sha256_clone(git_hash_ctx *dst, const git_hash_ctx *src) |
118 | { | |
119 | git_SHA256_Clone(&dst->sha256, &src->sha256); | |
120 | } | |
121 | ||
13eeedb5 | 122 | static void git_hash_sha256_update(git_hash_ctx *ctx, const void *data, size_t len) |
123 | { | |
124 | git_SHA256_Update(&ctx->sha256, data, len); | |
125 | } | |
126 | ||
127 | static void git_hash_sha256_final(unsigned char *hash, git_hash_ctx *ctx) | |
128 | { | |
129 | git_SHA256_Final(hash, &ctx->sha256); | |
130 | } | |
131 | ||
ab795f0d | 132 | static void git_hash_sha256_final_oid(struct object_id *oid, git_hash_ctx *ctx) |
133 | { | |
134 | git_SHA256_Final(oid->hash, &ctx->sha256); | |
135 | /* | |
136 | * This currently does nothing, so the compiler should optimize it out, | |
137 | * but keep it in case we extend the hash size again. | |
138 | */ | |
139 | memset(oid->hash + GIT_SHA256_RAWSZ, 0, GIT_MAX_RAWSZ - GIT_SHA256_RAWSZ); | |
5a6dce70 | 140 | oid->algo = GIT_HASH_SHA256; |
ab795f0d | 141 | } |
142 | ||
ac73cedf | 143 | static void git_hash_unknown_init(git_hash_ctx *ctx) |
f50e766b | 144 | { |
1a07e59c | 145 | BUG("trying to init unknown hash"); |
f50e766b | 146 | } |
147 | ||
768e30ea | 148 | static void git_hash_unknown_clone(git_hash_ctx *dst, const git_hash_ctx *src) |
149 | { | |
150 | BUG("trying to clone unknown hash"); | |
151 | } | |
152 | ||
ac73cedf | 153 | static void git_hash_unknown_update(git_hash_ctx *ctx, const void *data, size_t len) |
f50e766b | 154 | { |
1a07e59c | 155 | BUG("trying to update unknown hash"); |
f50e766b | 156 | } |
157 | ||
ac73cedf | 158 | static void git_hash_unknown_final(unsigned char *hash, git_hash_ctx *ctx) |
f50e766b | 159 | { |
1a07e59c | 160 | BUG("trying to finalize unknown hash"); |
f50e766b | 161 | } |
162 | ||
ab795f0d | 163 | static void git_hash_unknown_final_oid(struct object_id *oid, git_hash_ctx *ctx) |
164 | { | |
165 | BUG("trying to finalize unknown hash"); | |
166 | } | |
167 | ||
168 | ||
f50e766b | 169 | const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = { |
170 | { | |
171 | NULL, | |
172 | 0x00000000, | |
173 | 0, | |
174 | 0, | |
a2ce0a75 | 175 | 0, |
f50e766b | 176 | git_hash_unknown_init, |
768e30ea | 177 | git_hash_unknown_clone, |
f50e766b | 178 | git_hash_unknown_update, |
179 | git_hash_unknown_final, | |
ab795f0d | 180 | git_hash_unknown_final_oid, |
f50e766b | 181 | NULL, |
182 | NULL, | |
14228447 | 183 | NULL, |
f50e766b | 184 | }, |
185 | { | |
1ccf07cb | 186 | "sha1", |
f50e766b | 187 | /* "sha1", big-endian */ |
188 | 0x73686131, | |
f50e766b | 189 | GIT_SHA1_RAWSZ, |
190 | GIT_SHA1_HEXSZ, | |
a2ce0a75 | 191 | GIT_SHA1_BLKSZ, |
f50e766b | 192 | git_hash_sha1_init, |
768e30ea | 193 | git_hash_sha1_clone, |
f50e766b | 194 | git_hash_sha1_update, |
195 | git_hash_sha1_final, | |
ab795f0d | 196 | git_hash_sha1_final_oid, |
f50e766b | 197 | &empty_tree_oid, |
198 | &empty_blob_oid, | |
14228447 | 199 | &null_oid_sha1, |
f50e766b | 200 | }, |
13eeedb5 | 201 | { |
202 | "sha256", | |
203 | /* "s256", big-endian */ | |
204 | 0x73323536, | |
205 | GIT_SHA256_RAWSZ, | |
206 | GIT_SHA256_HEXSZ, | |
207 | GIT_SHA256_BLKSZ, | |
208 | git_hash_sha256_init, | |
768e30ea | 209 | git_hash_sha256_clone, |
13eeedb5 | 210 | git_hash_sha256_update, |
211 | git_hash_sha256_final, | |
ab795f0d | 212 | git_hash_sha256_final_oid, |
13eeedb5 | 213 | &empty_tree_oid_sha256, |
214 | &empty_blob_oid_sha256, | |
14228447 | 215 | &null_oid_sha256, |
13eeedb5 | 216 | } |
f50e766b | 217 | }; |
218 | ||
14228447 | 219 | const struct object_id *null_oid(void) |
220 | { | |
221 | return the_hash_algo->null_oid; | |
222 | } | |
223 | ||
d8a92ced | 224 | const char *empty_tree_oid_hex(void) |
225 | { | |
226 | static char buf[GIT_MAX_HEXSZ + 1]; | |
227 | return oid_to_hex_r(buf, the_hash_algo->empty_tree); | |
228 | } | |
229 | ||
230 | const char *empty_blob_oid_hex(void) | |
231 | { | |
232 | static char buf[GIT_MAX_HEXSZ + 1]; | |
233 | return oid_to_hex_r(buf, the_hash_algo->empty_blob); | |
234 | } | |
235 | ||
2f90b9d9 | 236 | int hash_algo_by_name(const char *name) |
237 | { | |
238 | int i; | |
239 | if (!name) | |
240 | return GIT_HASH_UNKNOWN; | |
241 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
242 | if (!strcmp(name, hash_algos[i].name)) | |
243 | return i; | |
244 | return GIT_HASH_UNKNOWN; | |
245 | } | |
246 | ||
247 | int hash_algo_by_id(uint32_t format_id) | |
248 | { | |
249 | int i; | |
250 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
251 | if (format_id == hash_algos[i].format_id) | |
252 | return i; | |
253 | return GIT_HASH_UNKNOWN; | |
254 | } | |
255 | ||
95399788 | 256 | int hash_algo_by_length(int len) |
257 | { | |
258 | int i; | |
259 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
260 | if (len == hash_algos[i].rawsz) | |
261 | return i; | |
262 | return GIT_HASH_UNKNOWN; | |
263 | } | |
2f90b9d9 | 264 | |
c597ba80 NTND |
265 | /* |
266 | * This is meant to hold a *small* number of objects that you would | |
cb1c8d1d | 267 | * want read_object_file() to be able to return, but yet you do not want |
c597ba80 NTND |
268 | * to write them into the object store (e.g. a browse-only |
269 | * application). | |
270 | */ | |
271 | static struct cached_object { | |
62ba93ea | 272 | struct object_id oid; |
c597ba80 NTND |
273 | enum object_type type; |
274 | void *buf; | |
275 | unsigned long size; | |
276 | } *cached_objects; | |
277 | static int cached_object_nr, cached_object_alloc; | |
278 | ||
279 | static struct cached_object empty_tree = { | |
62ba93ea | 280 | { EMPTY_TREE_SHA1_BIN_LITERAL }, |
c597ba80 NTND |
281 | OBJ_TREE, |
282 | "", | |
283 | 0 | |
284 | }; | |
285 | ||
62ba93ea | 286 | static struct cached_object *find_cached_object(const struct object_id *oid) |
c597ba80 NTND |
287 | { |
288 | int i; | |
289 | struct cached_object *co = cached_objects; | |
290 | ||
291 | for (i = 0; i < cached_object_nr; i++, co++) { | |
4a7e27e9 | 292 | if (oideq(&co->oid, oid)) |
c597ba80 NTND |
293 | return co; |
294 | } | |
4a7e27e9 | 295 | if (oideq(oid, the_hash_algo->empty_tree)) |
c597ba80 NTND |
296 | return &empty_tree; |
297 | return NULL; | |
298 | } | |
299 | ||
9472935d | 300 | |
8462ff43 | 301 | static int get_conv_flags(unsigned flags) |
9472935d TB |
302 | { |
303 | if (flags & HASH_RENORMALIZE) | |
8462ff43 | 304 | return CONV_EOL_RENORMALIZE; |
9472935d | 305 | else if (flags & HASH_WRITE_OBJECT) |
107642fe | 306 | return global_conv_flags_eol | CONV_WRITE_OBJECT; |
9472935d | 307 | else |
8462ff43 | 308 | return 0; |
9472935d TB |
309 | } |
310 | ||
311 | ||
90a6464b JH |
312 | int mkdir_in_gitdir(const char *path) |
313 | { | |
314 | if (mkdir(path, 0777)) { | |
315 | int saved_errno = errno; | |
316 | struct stat st; | |
317 | struct strbuf sb = STRBUF_INIT; | |
318 | ||
319 | if (errno != EEXIST) | |
320 | return -1; | |
321 | /* | |
322 | * Are we looking at a path in a symlinked worktree | |
323 | * whose original repository does not yet have it? | |
324 | * e.g. .git/rr-cache pointing at its original | |
325 | * repository in which the user hasn't performed any | |
326 | * conflict resolution yet? | |
327 | */ | |
328 | if (lstat(path, &st) || !S_ISLNK(st.st_mode) || | |
329 | strbuf_readlink(&sb, path, st.st_size) || | |
330 | !is_absolute_path(sb.buf) || | |
331 | mkdir(sb.buf, 0777)) { | |
332 | strbuf_release(&sb); | |
333 | errno = saved_errno; | |
334 | return -1; | |
335 | } | |
336 | strbuf_release(&sb); | |
337 | } | |
338 | return adjust_shared_perm(path); | |
339 | } | |
340 | ||
eb3c027e | 341 | static enum scld_error safe_create_leading_directories_1(char *path, int share) |
b2cb9425 | 342 | { |
26c8ae2a | 343 | char *next_component = path + offset_1st_component(path); |
0be0521b | 344 | enum scld_error ret = SCLD_OK; |
67d42212 | 345 | |
0be0521b | 346 | while (ret == SCLD_OK && next_component) { |
f0502332 | 347 | struct stat st; |
0f527403 | 348 | char *slash = next_component, slash_character; |
f0502332 | 349 | |
0f527403 MH |
350 | while (*slash && !is_dir_sep(*slash)) |
351 | slash++; | |
352 | ||
353 | if (!*slash) | |
b2cb9425 | 354 | break; |
bf10cf70 | 355 | |
26c8ae2a | 356 | next_component = slash + 1; |
0f527403 | 357 | while (is_dir_sep(*next_component)) |
bf10cf70 | 358 | next_component++; |
26c8ae2a | 359 | if (!*next_component) |
5f0bdf50 | 360 | break; |
831651fd | 361 | |
0f527403 | 362 | slash_character = *slash; |
831651fd | 363 | *slash = '\0'; |
67d42212 JR |
364 | if (!stat(path, &st)) { |
365 | /* path exists */ | |
204a047f MH |
366 | if (!S_ISDIR(st.st_mode)) { |
367 | errno = ENOTDIR; | |
0be0521b | 368 | ret = SCLD_EXISTS; |
204a047f | 369 | } |
53a39721 | 370 | } else if (mkdir(path, 0777)) { |
928734d9 | 371 | if (errno == EEXIST && |
9e6f885d | 372 | !stat(path, &st) && S_ISDIR(st.st_mode)) |
928734d9 | 373 | ; /* somebody created it since we checked */ |
18d37e86 MH |
374 | else if (errno == ENOENT) |
375 | /* | |
376 | * Either mkdir() failed because | |
377 | * somebody just pruned the containing | |
378 | * directory, or stat() failed because | |
379 | * the file that was in our way was | |
380 | * just removed. Either way, inform | |
381 | * the caller that it might be worth | |
382 | * trying again: | |
383 | */ | |
384 | ret = SCLD_VANISHED; | |
9e6f885d | 385 | else |
0be0521b | 386 | ret = SCLD_FAILED; |
eb3c027e | 387 | } else if (share && adjust_shared_perm(path)) { |
0be0521b | 388 | ret = SCLD_PERMS; |
457f06d6 | 389 | } |
0f527403 | 390 | *slash = slash_character; |
b2cb9425 | 391 | } |
9e6f885d | 392 | return ret; |
b2cb9425 | 393 | } |
723c31fe | 394 | |
eb3c027e MT |
395 | enum scld_error safe_create_leading_directories(char *path) |
396 | { | |
397 | return safe_create_leading_directories_1(path, 1); | |
398 | } | |
399 | ||
400 | enum scld_error safe_create_leading_directories_no_share(char *path) | |
401 | { | |
402 | return safe_create_leading_directories_1(path, 0); | |
403 | } | |
404 | ||
0be0521b | 405 | enum scld_error safe_create_leading_directories_const(const char *path) |
8e21d63b | 406 | { |
02944307 | 407 | int save_errno; |
8e21d63b JK |
408 | /* path points to cache entries, so xstrdup before messing with it */ |
409 | char *buf = xstrdup(path); | |
0be0521b | 410 | enum scld_error result = safe_create_leading_directories(buf); |
02944307 MH |
411 | |
412 | save_errno = errno; | |
8e21d63b | 413 | free(buf); |
02944307 | 414 | errno = save_errno; |
8e21d63b JK |
415 | return result; |
416 | } | |
417 | ||
177978f5 MH |
418 | int raceproof_create_file(const char *path, create_file_fn fn, void *cb) |
419 | { | |
420 | /* | |
421 | * The number of times we will try to remove empty directories | |
422 | * in the way of path. This is only 1 because if another | |
423 | * process is racily creating directories that conflict with | |
424 | * us, we don't want to fight against them. | |
425 | */ | |
426 | int remove_directories_remaining = 1; | |
427 | ||
428 | /* | |
429 | * The number of times that we will try to create the | |
430 | * directories containing path. We are willing to attempt this | |
431 | * more than once, because another process could be trying to | |
432 | * clean up empty directories at the same time as we are | |
433 | * trying to create them. | |
434 | */ | |
435 | int create_directories_remaining = 3; | |
436 | ||
437 | /* A scratch copy of path, filled lazily if we need it: */ | |
438 | struct strbuf path_copy = STRBUF_INIT; | |
439 | ||
440 | int ret, save_errno; | |
441 | ||
442 | /* Sanity check: */ | |
443 | assert(*path); | |
444 | ||
445 | retry_fn: | |
446 | ret = fn(path, cb); | |
447 | save_errno = errno; | |
448 | if (!ret) | |
449 | goto out; | |
450 | ||
451 | if (errno == EISDIR && remove_directories_remaining-- > 0) { | |
452 | /* | |
453 | * A directory is in the way. Maybe it is empty; try | |
454 | * to remove it: | |
455 | */ | |
456 | if (!path_copy.len) | |
457 | strbuf_addstr(&path_copy, path); | |
458 | ||
459 | if (!remove_dir_recursively(&path_copy, REMOVE_DIR_EMPTY_ONLY)) | |
460 | goto retry_fn; | |
461 | } else if (errno == ENOENT && create_directories_remaining-- > 0) { | |
462 | /* | |
463 | * Maybe the containing directory didn't exist, or | |
464 | * maybe it was just deleted by a process that is | |
465 | * racing with us to clean up empty directories. Try | |
466 | * to create it: | |
467 | */ | |
468 | enum scld_error scld_result; | |
469 | ||
470 | if (!path_copy.len) | |
471 | strbuf_addstr(&path_copy, path); | |
472 | ||
473 | do { | |
474 | scld_result = safe_create_leading_directories(path_copy.buf); | |
475 | if (scld_result == SCLD_OK) | |
476 | goto retry_fn; | |
477 | } while (scld_result == SCLD_VANISHED && create_directories_remaining-- > 0); | |
478 | } | |
479 | ||
480 | out: | |
481 | strbuf_release(&path_copy); | |
482 | errno = save_errno; | |
483 | return ret; | |
484 | } | |
485 | ||
514c5fdd | 486 | static void fill_loose_path(struct strbuf *buf, const struct object_id *oid) |
ace1534d JH |
487 | { |
488 | int i; | |
94b5e093 | 489 | for (i = 0; i < the_hash_algo->rawsz; i++) { |
ace1534d | 490 | static char hex[] = "0123456789abcdef"; |
514c5fdd | 491 | unsigned int val = oid->hash[i]; |
f7b7774f JK |
492 | strbuf_addch(buf, hex[val >> 4]); |
493 | strbuf_addch(buf, hex[val & 0xf]); | |
afbba2f0 | 494 | if (!i) |
f7b7774f | 495 | strbuf_addch(buf, '/'); |
ace1534d JH |
496 | } |
497 | } | |
498 | ||
f0eaf638 JK |
499 | static const char *odb_loose_path(struct object_directory *odb, |
500 | struct strbuf *buf, | |
514c5fdd | 501 | const struct object_id *oid) |
0fcfd160 | 502 | { |
b69fb867 | 503 | strbuf_reset(buf); |
f0eaf638 | 504 | strbuf_addstr(buf, odb->path); |
34498471 | 505 | strbuf_addch(buf, '/'); |
514c5fdd | 506 | fill_loose_path(buf, oid); |
f3f043a1 | 507 | return buf->buf; |
0fcfd160 LT |
508 | } |
509 | ||
f3f043a1 | 510 | const char *loose_object_path(struct repository *r, struct strbuf *buf, |
514c5fdd | 511 | const struct object_id *oid) |
29ec6af2 | 512 | { |
514c5fdd | 513 | return odb_loose_path(r->objects->odb, buf, oid); |
0fcfd160 LT |
514 | } |
515 | ||
4ea82473 JK |
516 | /* |
517 | * Return non-zero iff the path is usable as an alternate object database. | |
518 | */ | |
13313fc3 SB |
519 | static int alt_odb_usable(struct raw_object_store *o, |
520 | struct strbuf *path, | |
cf2dc1c2 | 521 | const char *normalized_objdir, khiter_t *pos) |
4ea82473 | 522 | { |
cf2dc1c2 | 523 | int r; |
4ea82473 JK |
524 | |
525 | /* Detect cases where alternate disappeared */ | |
526 | if (!is_directory(path->buf)) { | |
259328b7 NTND |
527 | error(_("object directory %s does not exist; " |
528 | "check .git/objects/info/alternates"), | |
4ea82473 JK |
529 | path->buf); |
530 | return 0; | |
531 | } | |
532 | ||
533 | /* | |
534 | * Prevent the common mistake of listing the same | |
535 | * thing twice, or object directory itself. | |
536 | */ | |
cf2dc1c2 EW |
537 | if (!o->odb_by_path) { |
538 | khiter_t p; | |
539 | ||
540 | o->odb_by_path = kh_init_odb_path_map(); | |
541 | assert(!o->odb->next); | |
542 | p = kh_put_odb_path_map(o->odb_by_path, o->odb->path, &r); | |
543 | assert(r == 1); /* never used */ | |
544 | kh_value(o->odb_by_path, p) = o->odb; | |
4ea82473 | 545 | } |
cf2dc1c2 | 546 | if (fspatheq(path->buf, normalized_objdir)) |
4ea82473 | 547 | return 0; |
cf2dc1c2 EW |
548 | *pos = kh_put_odb_path_map(o->odb_by_path, path->buf, &r); |
549 | /* r: 0 = exists, 1 = never used, 2 = deleted */ | |
550 | return r == 0 ? 0 : 1; | |
4ea82473 JK |
551 | } |
552 | ||
ddd5d056 JH |
553 | /* |
554 | * Prepare alternate object database registry. | |
d5a63b99 JH |
555 | * |
556 | * The variable alt_odb_list points at the list of struct | |
263db403 | 557 | * object_directory. The elements on this list come from |
d5a63b99 JH |
558 | * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT |
559 | * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates, | |
1494e038 JH |
560 | * whose contents is similar to that environment variable but can be |
561 | * LF separated. Its base points at a statically allocated buffer that | |
d5a63b99 JH |
562 | * contains "/the/directory/corresponding/to/.git/objects/...", while |
563 | * its name points just after the slash at the end of ".git/objects/" | |
e5afd444 MÅ |
564 | * in the example above, and has enough space to hold all hex characters |
565 | * of the object ID, an extra slash for the first level indirection, and | |
566 | * the terminating NUL. | |
ddd5d056 | 567 | */ |
77f012e8 SB |
568 | static void read_info_alternates(struct repository *r, |
569 | const char *relative_base, | |
570 | int depth); | |
407532f8 | 571 | static int link_alt_odb_entry(struct repository *r, const struct strbuf *entry, |
cfc62fc9 | 572 | const char *relative_base, int depth, const char *normalized_objdir) |
ace1534d | 573 | { |
263db403 | 574 | struct object_directory *ent; |
5bdf0a84 | 575 | struct strbuf pathbuf = STRBUF_INIT; |
cf2dc1c2 | 576 | khiter_t pos; |
d5a63b99 | 577 | |
407532f8 | 578 | if (!is_absolute_path(entry->buf) && relative_base) { |
4ac9006f | 579 | strbuf_realpath(&pathbuf, relative_base, 1); |
5bdf0a84 | 580 | strbuf_addch(&pathbuf, '/'); |
c2f493a4 | 581 | } |
407532f8 | 582 | strbuf_addbuf(&pathbuf, entry); |
c2f493a4 | 583 | |
37a95862 | 584 | if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) { |
259328b7 | 585 | error(_("unable to normalize alternate object path: %s"), |
670c359d JK |
586 | pathbuf.buf); |
587 | strbuf_release(&pathbuf); | |
588 | return -1; | |
589 | } | |
5bdf0a84 HW |
590 | |
591 | /* | |
592 | * The trailing slash after the directory name is given by | |
593 | * this function at the end. Remove duplicates. | |
594 | */ | |
4ea82473 JK |
595 | while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/') |
596 | strbuf_setlen(&pathbuf, pathbuf.len - 1); | |
5bdf0a84 | 597 | |
cf2dc1c2 | 598 | if (!alt_odb_usable(r->objects, &pathbuf, normalized_objdir, &pos)) { |
4ea82473 | 599 | strbuf_release(&pathbuf); |
c2f493a4 MW |
600 | return -1; |
601 | } | |
602 | ||
ca56dadb | 603 | CALLOC_ARRAY(ent, 1); |
cf2dc1c2 EW |
604 | /* pathbuf.buf is already in r->objects->odb_by_path */ |
605 | ent->path = strbuf_detach(&pathbuf, NULL); | |
c2f493a4 MW |
606 | |
607 | /* add the alternate entry */ | |
f0eaf638 JK |
608 | *r->objects->odb_tail = ent; |
609 | r->objects->odb_tail = &(ent->next); | |
c2f493a4 | 610 | ent->next = NULL; |
cf2dc1c2 EW |
611 | assert(r->objects->odb_by_path); |
612 | kh_value(r->objects->odb_by_path, pos) = ent; | |
c2f493a4 MW |
613 | |
614 | /* recursively add alternates */ | |
cf2dc1c2 | 615 | read_info_alternates(r, ent->path, depth + 1); |
c2f493a4 MW |
616 | |
617 | return 0; | |
618 | } | |
619 | ||
cf3c6352 JK |
620 | static const char *parse_alt_odb_entry(const char *string, |
621 | int sep, | |
622 | struct strbuf *out) | |
623 | { | |
624 | const char *end; | |
625 | ||
626 | strbuf_reset(out); | |
627 | ||
628 | if (*string == '#') { | |
629 | /* comment; consume up to next separator */ | |
630 | end = strchrnul(string, sep); | |
631 | } else if (*string == '"' && !unquote_c_style(out, string, &end)) { | |
632 | /* | |
633 | * quoted path; unquote_c_style has copied the | |
634 | * data for us and set "end". Broken quoting (e.g., | |
635 | * an entry that doesn't end with a quote) falls | |
636 | * back to the unquoted case below. | |
637 | */ | |
638 | } else { | |
639 | /* normal, unquoted path */ | |
640 | end = strchrnul(string, sep); | |
641 | strbuf_add(out, string, end - string); | |
642 | } | |
643 | ||
644 | if (*end) | |
645 | end++; | |
646 | return end; | |
647 | } | |
648 | ||
77f012e8 SB |
649 | static void link_alt_odb_entries(struct repository *r, const char *alt, |
650 | int sep, const char *relative_base, int depth) | |
c2f493a4 | 651 | { |
539e7506 | 652 | struct strbuf objdirbuf = STRBUF_INIT; |
cf3c6352 | 653 | struct strbuf entry = STRBUF_INIT; |
c2f493a4 | 654 | |
f28e3668 JK |
655 | if (!alt || !*alt) |
656 | return; | |
657 | ||
c2f493a4 | 658 | if (depth > 5) { |
259328b7 | 659 | error(_("%s: ignoring alternate object stores, nesting too deep"), |
c2f493a4 MW |
660 | relative_base); |
661 | return; | |
662 | } | |
663 | ||
f0eaf638 | 664 | strbuf_add_absolute_path(&objdirbuf, r->objects->odb->path); |
670c359d | 665 | if (strbuf_normalize_path(&objdirbuf) < 0) |
259328b7 | 666 | die(_("unable to normalize object directory: %s"), |
670c359d | 667 | objdirbuf.buf); |
539e7506 | 668 | |
cf3c6352 JK |
669 | while (*alt) { |
670 | alt = parse_alt_odb_entry(alt, sep, &entry); | |
671 | if (!entry.len) | |
9577e7e3 | 672 | continue; |
407532f8 | 673 | link_alt_odb_entry(r, &entry, |
cfc62fc9 | 674 | relative_base, depth, objdirbuf.buf); |
9577e7e3 | 675 | } |
cf3c6352 | 676 | strbuf_release(&entry); |
539e7506 | 677 | strbuf_release(&objdirbuf); |
d5a63b99 JH |
678 | } |
679 | ||
77f012e8 SB |
680 | static void read_info_alternates(struct repository *r, |
681 | const char *relative_base, | |
682 | int depth) | |
d5a63b99 | 683 | { |
5015f01c | 684 | char *path; |
dc732bd5 | 685 | struct strbuf buf = STRBUF_INIT; |
d5a63b99 | 686 | |
5015f01c | 687 | path = xstrfmt("%s/info/alternates", relative_base); |
dc732bd5 | 688 | if (strbuf_read_file(&buf, path, 1024) < 0) { |
f0f7bebe | 689 | warn_on_fopen_errors(path); |
dc732bd5 | 690 | free(path); |
9a217f2a | 691 | return; |
ace1534d | 692 | } |
d5a63b99 | 693 | |
77f012e8 | 694 | link_alt_odb_entries(r, buf.buf, '\n', relative_base, depth); |
dc732bd5 JK |
695 | strbuf_release(&buf); |
696 | free(path); | |
ace1534d JH |
697 | } |
698 | ||
bef70b22 DB |
699 | void add_to_alternates_file(const char *reference) |
700 | { | |
f132a127 | 701 | struct lock_file lock = LOCK_INIT; |
77b9b1d1 JK |
702 | char *alts = git_pathdup("objects/info/alternates"); |
703 | FILE *in, *out; | |
f132a127 | 704 | int found = 0; |
77b9b1d1 | 705 | |
f132a127 MÅ |
706 | hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR); |
707 | out = fdopen_lock_file(&lock, "w"); | |
77b9b1d1 | 708 | if (!out) |
259328b7 | 709 | die_errno(_("unable to fdopen alternates lockfile")); |
77b9b1d1 JK |
710 | |
711 | in = fopen(alts, "r"); | |
712 | if (in) { | |
713 | struct strbuf line = STRBUF_INIT; | |
77b9b1d1 | 714 | |
3f163962 | 715 | while (strbuf_getline(&line, in) != EOF) { |
77b9b1d1 JK |
716 | if (!strcmp(reference, line.buf)) { |
717 | found = 1; | |
718 | break; | |
719 | } | |
720 | fprintf_or_die(out, "%s\n", line.buf); | |
721 | } | |
722 | ||
723 | strbuf_release(&line); | |
724 | fclose(in); | |
77b9b1d1 JK |
725 | } |
726 | else if (errno != ENOENT) | |
259328b7 | 727 | die_errno(_("unable to read alternates file")); |
77b9b1d1 | 728 | |
f132a127 MÅ |
729 | if (found) { |
730 | rollback_lock_file(&lock); | |
731 | } else { | |
77b9b1d1 | 732 | fprintf_or_die(out, "%s\n", reference); |
f132a127 | 733 | if (commit_lock_file(&lock)) |
259328b7 | 734 | die_errno(_("unable to move new alternates file into place")); |
f0eaf638 | 735 | if (the_repository->objects->loaded_alternates) |
93d8d1e2 SB |
736 | link_alt_odb_entries(the_repository, reference, |
737 | '\n', NULL, 0); | |
77b9b1d1 JK |
738 | } |
739 | free(alts); | |
bef70b22 DB |
740 | } |
741 | ||
a5b34d21 JK |
742 | void add_to_alternates_memory(const char *reference) |
743 | { | |
744 | /* | |
745 | * Make sure alternates are initialized, or else our entry may be | |
746 | * overwritten when they are. | |
747 | */ | |
0b209034 | 748 | prepare_alt_odb(the_repository); |
a5b34d21 | 749 | |
93d8d1e2 SB |
750 | link_alt_odb_entries(the_repository, reference, |
751 | '\n', NULL, 0); | |
a5b34d21 JK |
752 | } |
753 | ||
9eeea7d2 SB |
754 | /* |
755 | * Compute the exact path an alternate is at and returns it. In case of | |
756 | * error NULL is returned and the human readable error is added to `err` | |
efde7b72 | 757 | * `path` may be relative and should point to $GIT_DIR. |
9eeea7d2 SB |
758 | * `err` must not be null. |
759 | */ | |
760 | char *compute_alternate_path(const char *path, struct strbuf *err) | |
761 | { | |
762 | char *ref_git = NULL; | |
4530a85b | 763 | const char *repo; |
9eeea7d2 SB |
764 | int seen_error = 0; |
765 | ||
4530a85b AM |
766 | ref_git = real_pathdup(path, 0); |
767 | if (!ref_git) { | |
9eeea7d2 SB |
768 | seen_error = 1; |
769 | strbuf_addf(err, _("path '%s' does not exist"), path); | |
770 | goto out; | |
4530a85b | 771 | } |
9eeea7d2 SB |
772 | |
773 | repo = read_gitfile(ref_git); | |
774 | if (!repo) | |
775 | repo = read_gitfile(mkpath("%s/.git", ref_git)); | |
776 | if (repo) { | |
777 | free(ref_git); | |
778 | ref_git = xstrdup(repo); | |
779 | } | |
780 | ||
781 | if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) { | |
782 | char *ref_git_git = mkpathdup("%s/.git", ref_git); | |
783 | free(ref_git); | |
784 | ref_git = ref_git_git; | |
785 | } else if (!is_directory(mkpath("%s/objects", ref_git))) { | |
786 | struct strbuf sb = STRBUF_INIT; | |
787 | seen_error = 1; | |
788 | if (get_common_dir(&sb, ref_git)) { | |
789 | strbuf_addf(err, | |
790 | _("reference repository '%s' as a linked " | |
791 | "checkout is not supported yet."), | |
792 | path); | |
793 | goto out; | |
794 | } | |
795 | ||
796 | strbuf_addf(err, _("reference repository '%s' is not a " | |
797 | "local repository."), path); | |
798 | goto out; | |
799 | } | |
800 | ||
801 | if (!access(mkpath("%s/shallow", ref_git), F_OK)) { | |
802 | strbuf_addf(err, _("reference repository '%s' is shallow"), | |
803 | path); | |
804 | seen_error = 1; | |
805 | goto out; | |
806 | } | |
807 | ||
808 | if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) { | |
809 | strbuf_addf(err, | |
810 | _("reference repository '%s' is grafted"), | |
811 | path); | |
812 | seen_error = 1; | |
813 | goto out; | |
814 | } | |
815 | ||
816 | out: | |
817 | if (seen_error) { | |
6a83d902 | 818 | FREE_AND_NULL(ref_git); |
9eeea7d2 SB |
819 | } |
820 | ||
821 | return ref_git; | |
822 | } | |
823 | ||
f57a7396 TB |
824 | struct object_directory *find_odb(struct repository *r, const char *obj_dir) |
825 | { | |
826 | struct object_directory *odb; | |
827 | char *obj_dir_real = real_pathdup(obj_dir, 1); | |
828 | struct strbuf odb_path_real = STRBUF_INIT; | |
829 | ||
830 | prepare_alt_odb(r); | |
831 | for (odb = r->objects->odb; odb; odb = odb->next) { | |
832 | strbuf_realpath(&odb_path_real, odb->path, 1); | |
833 | if (!strcmp(obj_dir_real, odb_path_real.buf)) | |
834 | break; | |
835 | } | |
836 | ||
837 | free(obj_dir_real); | |
838 | strbuf_release(&odb_path_real); | |
839 | ||
840 | if (!odb) | |
841 | die(_("could not find object directory matching %s"), obj_dir); | |
842 | return odb; | |
843 | } | |
844 | ||
709dfa69 JK |
845 | static void fill_alternate_refs_command(struct child_process *cmd, |
846 | const char *repo_path) | |
847 | { | |
848 | const char *value; | |
849 | ||
850 | if (!git_config_get_value("core.alternateRefsCommand", &value)) { | |
851 | cmd->use_shell = 1; | |
852 | ||
c972bf4c JK |
853 | strvec_push(&cmd->args, value); |
854 | strvec_push(&cmd->args, repo_path); | |
709dfa69 JK |
855 | } else { |
856 | cmd->git_cmd = 1; | |
857 | ||
c972bf4c JK |
858 | strvec_pushf(&cmd->args, "--git-dir=%s", repo_path); |
859 | strvec_push(&cmd->args, "for-each-ref"); | |
860 | strvec_push(&cmd->args, "--format=%(objectname)"); | |
709dfa69 JK |
861 | |
862 | if (!git_config_get_value("core.alternateRefsPrefixes", &value)) { | |
c972bf4c JK |
863 | strvec_push(&cmd->args, "--"); |
864 | strvec_split(&cmd->args, value); | |
709dfa69 JK |
865 | } |
866 | } | |
867 | ||
868 | cmd->env = local_repo_env; | |
869 | cmd->out = -1; | |
870 | } | |
871 | ||
872 | static void read_alternate_refs(const char *path, | |
873 | alternate_ref_fn *cb, | |
874 | void *data) | |
875 | { | |
876 | struct child_process cmd = CHILD_PROCESS_INIT; | |
877 | struct strbuf line = STRBUF_INIT; | |
878 | FILE *fh; | |
879 | ||
880 | fill_alternate_refs_command(&cmd, path); | |
881 | ||
882 | if (start_command(&cmd)) | |
883 | return; | |
884 | ||
885 | fh = xfdopen(cmd.out, "r"); | |
886 | while (strbuf_getline_lf(&line, fh) != EOF) { | |
887 | struct object_id oid; | |
888 | const char *p; | |
889 | ||
890 | if (parse_oid_hex(line.buf, &oid, &p) || *p) { | |
891 | warning(_("invalid line while parsing alternate refs: %s"), | |
892 | line.buf); | |
893 | break; | |
894 | } | |
895 | ||
896 | cb(&oid, data); | |
897 | } | |
898 | ||
899 | fclose(fh); | |
900 | finish_command(&cmd); | |
86ad3ea5 | 901 | strbuf_release(&line); |
709dfa69 JK |
902 | } |
903 | ||
904 | struct alternate_refs_data { | |
905 | alternate_ref_fn *fn; | |
906 | void *data; | |
907 | }; | |
908 | ||
909 | static int refs_from_alternate_cb(struct object_directory *e, | |
910 | void *data) | |
911 | { | |
912 | struct strbuf path = STRBUF_INIT; | |
913 | size_t base_len; | |
914 | struct alternate_refs_data *cb = data; | |
915 | ||
916 | if (!strbuf_realpath(&path, e->path, 0)) | |
917 | goto out; | |
918 | if (!strbuf_strip_suffix(&path, "/objects")) | |
919 | goto out; | |
920 | base_len = path.len; | |
921 | ||
922 | /* Is this a git repository with refs? */ | |
923 | strbuf_addstr(&path, "/refs"); | |
924 | if (!is_directory(path.buf)) | |
925 | goto out; | |
926 | strbuf_setlen(&path, base_len); | |
927 | ||
928 | read_alternate_refs(path.buf, cb->fn, cb->data); | |
929 | ||
930 | out: | |
931 | strbuf_release(&path); | |
932 | return 0; | |
933 | } | |
934 | ||
935 | void for_each_alternate_ref(alternate_ref_fn fn, void *data) | |
936 | { | |
937 | struct alternate_refs_data cb; | |
938 | cb.fn = fn; | |
939 | cb.data = data; | |
940 | foreach_alt_odb(refs_from_alternate_cb, &cb); | |
941 | } | |
942 | ||
fe1b2268 | 943 | int foreach_alt_odb(alt_odb_fn fn, void *cb) |
d79796bc | 944 | { |
263db403 | 945 | struct object_directory *ent; |
fe1b2268 | 946 | int r = 0; |
d79796bc | 947 | |
0b209034 | 948 | prepare_alt_odb(the_repository); |
f0eaf638 | 949 | for (ent = the_repository->objects->odb->next; ent; ent = ent->next) { |
fe1b2268 JK |
950 | r = fn(ent, cb); |
951 | if (r) | |
952 | break; | |
953 | } | |
954 | return r; | |
d79796bc JH |
955 | } |
956 | ||
13068bf0 | 957 | void prepare_alt_odb(struct repository *r) |
c2f493a4 | 958 | { |
f0eaf638 | 959 | if (r->objects->loaded_alternates) |
7dc24aa5 SP |
960 | return; |
961 | ||
13068bf0 | 962 | link_alt_odb_entries(r, r->objects->alternate_db, PATH_SEP, NULL, 0); |
c2f493a4 | 963 | |
f0eaf638 JK |
964 | read_info_alternates(r, r->objects->odb->path, 0); |
965 | r->objects->loaded_alternates = 1; | |
c2f493a4 MW |
966 | } |
967 | ||
3096b2ec | 968 | /* Returns 1 if we have successfully freshened the file, 0 otherwise. */ |
33d4221c | 969 | static int freshen_file(const char *fn) |
ace1534d | 970 | { |
312cd761 | 971 | return !utime(fn, NULL); |
0f4dc14a | 972 | } |
ace1534d | 973 | |
3096b2ec JK |
974 | /* |
975 | * All of the check_and_freshen functions return 1 if the file exists and was | |
976 | * freshened (if freshening was requested), 0 otherwise. If they return | |
977 | * 0, you should not assume that it is safe to skip a write of the object (it | |
978 | * either does not exist on disk, or has a stale mtime and may be subject to | |
979 | * pruning). | |
980 | */ | |
6a5e6f5e | 981 | int check_and_freshen_file(const char *fn, int freshen) |
33d4221c JK |
982 | { |
983 | if (access(fn, F_OK)) | |
984 | return 0; | |
3096b2ec | 985 | if (freshen && !freshen_file(fn)) |
33d4221c JK |
986 | return 0; |
987 | return 1; | |
988 | } | |
989 | ||
f0eaf638 JK |
990 | static int check_and_freshen_odb(struct object_directory *odb, |
991 | const struct object_id *oid, | |
992 | int freshen) | |
33d4221c | 993 | { |
f0eaf638 | 994 | static struct strbuf path = STRBUF_INIT; |
514c5fdd | 995 | odb_loose_path(odb, &path, oid); |
f0eaf638 JK |
996 | return check_and_freshen_file(path.buf, freshen); |
997 | } | |
ea657730 | 998 | |
f0eaf638 JK |
999 | static int check_and_freshen_local(const struct object_id *oid, int freshen) |
1000 | { | |
1001 | return check_and_freshen_odb(the_repository->objects->odb, oid, freshen); | |
33d4221c JK |
1002 | } |
1003 | ||
6862ebbf | 1004 | static int check_and_freshen_nonlocal(const struct object_id *oid, int freshen) |
0f4dc14a | 1005 | { |
263db403 | 1006 | struct object_directory *odb; |
f3f043a1 | 1007 | |
0b209034 | 1008 | prepare_alt_odb(the_repository); |
f0eaf638 JK |
1009 | for (odb = the_repository->objects->odb->next; odb; odb = odb->next) { |
1010 | if (check_and_freshen_odb(odb, oid, freshen)) | |
c529d75a | 1011 | return 1; |
ace1534d | 1012 | } |
c529d75a | 1013 | return 0; |
ace1534d JH |
1014 | } |
1015 | ||
6862ebbf | 1016 | static int check_and_freshen(const struct object_id *oid, int freshen) |
33d4221c | 1017 | { |
6862ebbf | 1018 | return check_and_freshen_local(oid, freshen) || |
1019 | check_and_freshen_nonlocal(oid, freshen); | |
33d4221c JK |
1020 | } |
1021 | ||
6862ebbf | 1022 | int has_loose_object_nonlocal(const struct object_id *oid) |
33d4221c | 1023 | { |
6862ebbf | 1024 | return check_and_freshen_nonlocal(oid, 0); |
33d4221c JK |
1025 | } |
1026 | ||
6862ebbf | 1027 | static int has_loose_object(const struct object_id *oid) |
0f4dc14a | 1028 | { |
6862ebbf | 1029 | return check_and_freshen(oid, 0); |
0f4dc14a BC |
1030 | } |
1031 | ||
02710228 SP |
1032 | static void mmap_limit_check(size_t length) |
1033 | { | |
1034 | static size_t limit = 0; | |
1035 | if (!limit) { | |
1036 | limit = git_env_ulong("GIT_MMAP_LIMIT", 0); | |
1037 | if (!limit) | |
1038 | limit = SIZE_MAX; | |
1039 | } | |
1040 | if (length > limit) | |
259328b7 | 1041 | die(_("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX), |
02710228 SP |
1042 | (uintmax_t)length, (uintmax_t)limit); |
1043 | } | |
1044 | ||
1570856b JK |
1045 | void *xmmap_gently(void *start, size_t length, |
1046 | int prot, int flags, int fd, off_t offset) | |
58ecbd5e | 1047 | { |
02710228 SP |
1048 | void *ret; |
1049 | ||
1050 | mmap_limit_check(length); | |
1051 | ret = mmap(start, length, prot, flags, fd, offset); | |
9827d4c1 JK |
1052 | if (ret == MAP_FAILED && !length) |
1053 | ret = NULL; | |
58ecbd5e JN |
1054 | return ret; |
1055 | } | |
1056 | ||
dc059294 EW |
1057 | const char *mmap_os_err(void) |
1058 | { | |
1059 | static const char blank[] = ""; | |
1060 | #if defined(__linux__) | |
1061 | if (errno == ENOMEM) { | |
1062 | /* this continues an existing error message: */ | |
1063 | static const char enomem[] = | |
1064 | ", check sys.vm.max_map_count and/or RLIMIT_DATA"; | |
1065 | return enomem; | |
1066 | } | |
1067 | #endif /* OS-specific bits */ | |
1068 | return blank; | |
1069 | } | |
1070 | ||
1570856b JK |
1071 | void *xmmap(void *start, size_t length, |
1072 | int prot, int flags, int fd, off_t offset) | |
1073 | { | |
1074 | void *ret = xmmap_gently(start, length, prot, flags, fd, offset); | |
1075 | if (ret == MAP_FAILED) | |
dc059294 | 1076 | die_errno(_("mmap failed%s"), mmap_os_err()); |
1570856b JK |
1077 | return ret; |
1078 | } | |
1079 | ||
88d0db55 | 1080 | /* |
090ea126 | 1081 | * With an in-core object data in "map", rehash it to make sure the |
cb1c8d1d JK |
1082 | * object name actually matches "oid" to detect object corruption. |
1083 | * With "map" == NULL, try reading the object named with "oid" using | |
090ea126 | 1084 | * the streaming interface and rehash it to do the same. |
88d0db55 | 1085 | */ |
b98d1885 MT |
1086 | int check_object_signature(struct repository *r, const struct object_id *oid, |
1087 | void *map, unsigned long size, const char *type) | |
88d0db55 | 1088 | { |
f070facc | 1089 | struct object_id real_oid; |
090ea126 NTND |
1090 | enum object_type obj_type; |
1091 | struct git_istream *st; | |
18e2588e | 1092 | git_hash_ctx c; |
1af64f73 | 1093 | char hdr[MAX_HEADER_LEN]; |
090ea126 | 1094 | int hdrlen; |
88d0db55 | 1095 | |
090ea126 | 1096 | if (map) { |
b98d1885 | 1097 | hash_object_file(r->hash_algo, map, size, type, &real_oid); |
9001dc2a | 1098 | return !oideq(oid, &real_oid) ? -1 : 0; |
88d0db55 BC |
1099 | } |
1100 | ||
b98d1885 | 1101 | st = open_istream(r, oid, &obj_type, &size, NULL); |
090ea126 NTND |
1102 | if (!st) |
1103 | return -1; | |
88d0db55 | 1104 | |
090ea126 | 1105 | /* Generate the header */ |
ca473cef | 1106 | hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(obj_type), (uintmax_t)size) + 1; |
88d0db55 | 1107 | |
090ea126 | 1108 | /* Sha1.. */ |
b98d1885 MT |
1109 | r->hash_algo->init_fn(&c); |
1110 | r->hash_algo->update_fn(&c, hdr, hdrlen); | |
090ea126 NTND |
1111 | for (;;) { |
1112 | char buf[1024 * 16]; | |
1113 | ssize_t readlen = read_istream(st, buf, sizeof(buf)); | |
1f688557 | 1114 | |
f54fac53 JK |
1115 | if (readlen < 0) { |
1116 | close_istream(st); | |
1117 | return -1; | |
1118 | } | |
090ea126 NTND |
1119 | if (!readlen) |
1120 | break; | |
b98d1885 | 1121 | r->hash_algo->update_fn(&c, buf, readlen); |
fa5fc15d | 1122 | } |
5951bf46 | 1123 | r->hash_algo->final_oid_fn(&real_oid, &c); |
090ea126 | 1124 | close_istream(st); |
9001dc2a | 1125 | return !oideq(oid, &real_oid) ? -1 : 0; |
fa5fc15d SP |
1126 | } |
1127 | ||
1b8ac5ea | 1128 | int git_open_cloexec(const char *name, int flags) |
a0788266 | 1129 | { |
1e3001a8 JH |
1130 | int fd; |
1131 | static int o_cloexec = O_CLOEXEC; | |
a0788266 | 1132 | |
1e3001a8 JH |
1133 | fd = open(name, flags | o_cloexec); |
1134 | if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) { | |
cd66ada0 | 1135 | /* Try again w/o O_CLOEXEC: the kernel might not support it */ |
1e3001a8 JH |
1136 | o_cloexec &= ~O_CLOEXEC; |
1137 | fd = open(name, flags | o_cloexec); | |
491a8dec | 1138 | } |
491a8dec | 1139 | |
9fb9495d | 1140 | #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC) |
491a8dec | 1141 | { |
1e3001a8 | 1142 | static int fd_cloexec = FD_CLOEXEC; |
a0788266 | 1143 | |
1e3001a8 JH |
1144 | if (!o_cloexec && 0 <= fd && fd_cloexec) { |
1145 | /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */ | |
9fb9495d EW |
1146 | int flags = fcntl(fd, F_GETFD); |
1147 | if (fcntl(fd, F_SETFD, flags | fd_cloexec)) | |
1e3001a8 | 1148 | fd_cloexec = 0; |
cd66ada0 | 1149 | } |
44d1c19e | 1150 | } |
a0788266 | 1151 | #endif |
1b8ac5ea | 1152 | return fd; |
a0788266 JS |
1153 | } |
1154 | ||
3cf8b462 | 1155 | /* |
514c5fdd | 1156 | * Find "oid" as a loose object in the local repository or in an alternate. |
771e7d57 JK |
1157 | * Returns 0 on success, negative on failure. |
1158 | * | |
1159 | * The "path" out-parameter will give the path of the object we found (if any). | |
1160 | * Note that it may point to static storage and is only valid until another | |
514c5fdd | 1161 | * call to stat_loose_object(). |
3cf8b462 | 1162 | */ |
514c5fdd JK |
1163 | static int stat_loose_object(struct repository *r, const struct object_id *oid, |
1164 | struct stat *st, const char **path) | |
1f688557 | 1165 | { |
263db403 | 1166 | struct object_directory *odb; |
ea657730 CC |
1167 | static struct strbuf buf = STRBUF_INIT; |
1168 | ||
d2607fa0 | 1169 | prepare_alt_odb(r); |
f0eaf638 | 1170 | for (odb = r->objects->odb; odb; odb = odb->next) { |
514c5fdd | 1171 | *path = odb_loose_path(odb, &buf, oid); |
771e7d57 | 1172 | if (!lstat(*path, st)) |
052fe5ea | 1173 | return 0; |
c7934306 SP |
1174 | } |
1175 | ||
3cf8b462 SP |
1176 | return -1; |
1177 | } | |
1178 | ||
771e7d57 | 1179 | /* |
514c5fdd | 1180 | * Like stat_loose_object(), but actually open the object and return the |
771e7d57 JK |
1181 | * descriptor. See the caveats on the "path" parameter above. |
1182 | */ | |
514c5fdd JK |
1183 | static int open_loose_object(struct repository *r, |
1184 | const struct object_id *oid, const char **path) | |
60bb8b14 | 1185 | { |
44d1c19e | 1186 | int fd; |
263db403 | 1187 | struct object_directory *odb; |
f0eaf638 | 1188 | int most_interesting_errno = ENOENT; |
ea657730 CC |
1189 | static struct strbuf buf = STRBUF_INIT; |
1190 | ||
ec7283e5 | 1191 | prepare_alt_odb(r); |
f0eaf638 | 1192 | for (odb = r->objects->odb; odb; odb = odb->next) { |
514c5fdd | 1193 | *path = odb_loose_path(odb, &buf, oid); |
771e7d57 | 1194 | fd = git_open(*path); |
44d1c19e LT |
1195 | if (fd >= 0) |
1196 | return fd; | |
f0eaf638 | 1197 | |
d6c8a05b JK |
1198 | if (most_interesting_errno == ENOENT) |
1199 | most_interesting_errno = errno; | |
03e79c88 | 1200 | } |
d6c8a05b | 1201 | errno = most_interesting_errno; |
44d1c19e | 1202 | return -1; |
1f688557 JH |
1203 | } |
1204 | ||
61c7711c | 1205 | static int quick_has_loose(struct repository *r, |
d7a24573 | 1206 | const struct object_id *oid) |
61c7711c | 1207 | { |
61c7711c JK |
1208 | struct object_directory *odb; |
1209 | ||
61c7711c JK |
1210 | prepare_alt_odb(r); |
1211 | for (odb = r->objects->odb; odb; odb = odb->next) { | |
92d8ed8a | 1212 | if (oidtree_contains(odb_loose_cache(odb, oid), oid)) |
61c7711c JK |
1213 | return 1; |
1214 | } | |
1215 | return 0; | |
1216 | } | |
1217 | ||
f6371f92 JK |
1218 | /* |
1219 | * Map the loose object at "path" if it is not NULL, or the path found by | |
514c5fdd | 1220 | * searching for a loose object named "oid". |
f6371f92 | 1221 | */ |
514c5fdd JK |
1222 | static void *map_loose_object_1(struct repository *r, const char *path, |
1223 | const struct object_id *oid, unsigned long *size) | |
27d69a46 | 1224 | { |
0fcfd160 | 1225 | void *map; |
144bde78 | 1226 | int fd; |
84dd81c1 | 1227 | |
f6371f92 JK |
1228 | if (path) |
1229 | fd = git_open(path); | |
8261e1f1 | 1230 | else |
514c5fdd | 1231 | fd = open_loose_object(r, oid, &path); |
44d1c19e LT |
1232 | map = NULL; |
1233 | if (fd >= 0) { | |
1234 | struct stat st; | |
1f688557 | 1235 | |
44d1c19e LT |
1236 | if (!fstat(fd, &st)) { |
1237 | *size = xsize_t(st.st_size); | |
33e42de0 MM |
1238 | if (!*size) { |
1239 | /* mmap() is forbidden on empty files */ | |
259328b7 | 1240 | error(_("object file %s is empty"), path); |
6881925e | 1241 | close(fd); |
33e42de0 MM |
1242 | return NULL; |
1243 | } | |
44d1c19e | 1244 | map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0); |
144bde78 | 1245 | } |
44d1c19e | 1246 | close(fd); |
74e34e1f | 1247 | } |
0fcfd160 | 1248 | return map; |
74e34e1f NP |
1249 | } |
1250 | ||
514c5fdd JK |
1251 | void *map_loose_object(struct repository *r, |
1252 | const struct object_id *oid, | |
1253 | unsigned long *size) | |
068f85e3 | 1254 | { |
514c5fdd | 1255 | return map_loose_object_1(r, NULL, oid, size); |
068f85e3 | 1256 | } |
1257 | ||
00a7760e JK |
1258 | static int unpack_loose_short_header(git_zstream *stream, |
1259 | unsigned char *map, unsigned long mapsize, | |
1260 | void *buffer, unsigned long bufsiz) | |
47fe3f6e | 1261 | { |
31877c9a MT |
1262 | int ret; |
1263 | ||
c4483576 LT |
1264 | /* Get the data stream */ |
1265 | memset(stream, 0, sizeof(*stream)); | |
1266 | stream->next_in = map; | |
1267 | stream->avail_in = mapsize; | |
1268 | stream->next_out = buffer; | |
93821bd9 | 1269 | stream->avail_out = bufsiz; |
47fe3f6e | 1270 | |
39c68542 | 1271 | git_inflate_init(stream); |
31877c9a MT |
1272 | obj_read_unlock(); |
1273 | ret = git_inflate(stream, 0); | |
1274 | obj_read_lock(); | |
1275 | ||
1276 | return ret; | |
9a217f2a JH |
1277 | } |
1278 | ||
00a7760e JK |
1279 | int unpack_loose_header(git_zstream *stream, |
1280 | unsigned char *map, unsigned long mapsize, | |
1281 | void *buffer, unsigned long bufsiz) | |
1f688557 | 1282 | { |
00a7760e JK |
1283 | int status = unpack_loose_short_header(stream, map, mapsize, |
1284 | buffer, bufsiz); | |
42873078 | 1285 | |
d21f8426 JH |
1286 | if (status < Z_OK) |
1287 | return status; | |
628522ec | 1288 | |
d21f8426 JH |
1289 | /* Make sure we have the terminating NUL */ |
1290 | if (!memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) | |
1291 | return -1; | |
1f688557 JH |
1292 | return 0; |
1293 | } | |
1294 | ||
00a7760e JK |
1295 | static int unpack_loose_header_to_strbuf(git_zstream *stream, unsigned char *map, |
1296 | unsigned long mapsize, void *buffer, | |
1297 | unsigned long bufsiz, struct strbuf *header) | |
d131b7af | 1298 | { |
46f03448 KN |
1299 | int status; |
1300 | ||
00a7760e | 1301 | status = unpack_loose_short_header(stream, map, mapsize, buffer, bufsiz); |
d21f8426 JH |
1302 | if (status < Z_OK) |
1303 | return -1; | |
d131b7af | 1304 | |
46f03448 KN |
1305 | /* |
1306 | * Check if entire header is unpacked in the first iteration. | |
d131b7af | 1307 | */ |
46f03448 KN |
1308 | if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) |
1309 | return 0; | |
d131b7af | 1310 | |
46f03448 KN |
1311 | /* |
1312 | * buffer[0..bufsiz] was not large enough. Copy the partial | |
1313 | * result out to header, and then append the result of further | |
1314 | * reading the stream. | |
1315 | */ | |
1316 | strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer); | |
1317 | stream->next_out = buffer; | |
1318 | stream->avail_out = bufsiz; | |
d131b7af | 1319 | |
46f03448 | 1320 | do { |
31877c9a | 1321 | obj_read_unlock(); |
46f03448 | 1322 | status = git_inflate(stream, 0); |
31877c9a | 1323 | obj_read_lock(); |
46f03448 KN |
1324 | strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer); |
1325 | if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) | |
1326 | return 0; | |
1327 | stream->next_out = buffer; | |
1328 | stream->avail_out = bufsiz; | |
1329 | } while (status != Z_STREAM_END); | |
1330 | return -1; | |
d131b7af SP |
1331 | } |
1332 | ||
00a7760e JK |
1333 | static void *unpack_loose_rest(git_zstream *stream, |
1334 | void *buffer, unsigned long size, | |
1335 | const struct object_id *oid) | |
95099731 | 1336 | { |
5180cacc | 1337 | int bytes = strlen(buffer) + 1; |
3aee68aa | 1338 | unsigned char *buf = xmallocz(size); |
93821bd9 | 1339 | unsigned long n; |
7efbff75 | 1340 | int status = Z_OK; |
95099731 | 1341 | |
93821bd9 LT |
1342 | n = stream->total_out - bytes; |
1343 | if (n > size) | |
1344 | n = size; | |
1345 | memcpy(buf, (char *) buffer + bytes, n); | |
1346 | bytes = n; | |
456cdf6e LT |
1347 | if (bytes <= size) { |
1348 | /* | |
1349 | * The above condition must be (bytes <= size), not | |
1350 | * (bytes < size). In other words, even though we | |
ccf5ace0 | 1351 | * expect no more output and set avail_out to zero, |
456cdf6e LT |
1352 | * the input zlib stream may have bytes that express |
1353 | * "this concludes the stream", and we *do* want to | |
1354 | * eat that input. | |
1355 | * | |
1356 | * Otherwise we would not be able to test that we | |
1357 | * consumed all the input to reach the expected size; | |
1358 | * we also want to check that zlib tells us that all | |
1359 | * went well with status == Z_STREAM_END at the end. | |
1360 | */ | |
5180cacc LT |
1361 | stream->next_out = buf + bytes; |
1362 | stream->avail_out = size - bytes; | |
31877c9a MT |
1363 | while (status == Z_OK) { |
1364 | obj_read_unlock(); | |
39c68542 | 1365 | status = git_inflate(stream, Z_FINISH); |
31877c9a MT |
1366 | obj_read_lock(); |
1367 | } | |
5180cacc | 1368 | } |
456cdf6e | 1369 | if (status == Z_STREAM_END && !stream->avail_in) { |
39c68542 | 1370 | git_inflate_end(stream); |
7efbff75 | 1371 | return buf; |
95099731 NTND |
1372 | } |
1373 | ||
7efbff75 | 1374 | if (status < 0) |
00a7760e | 1375 | error(_("corrupt loose object '%s'"), oid_to_hex(oid)); |
7efbff75 | 1376 | else if (stream->avail_in) |
259328b7 | 1377 | error(_("garbage at end of loose object '%s'"), |
00a7760e | 1378 | oid_to_hex(oid)); |
7efbff75 JH |
1379 | free(buf); |
1380 | return NULL; | |
95099731 NTND |
1381 | } |
1382 | ||
d40d535b | 1383 | /* |
5180cacc LT |
1384 | * We used to just use "sscanf()", but that's actually way |
1385 | * too permissive for what we want to check. So do an anal | |
1386 | * object header parse by hand. | |
d40d535b | 1387 | */ |
00a7760e JK |
1388 | static int parse_loose_header_extended(const char *hdr, struct object_info *oi, |
1389 | unsigned int flags) | |
1f688557 | 1390 | { |
46f03448 | 1391 | const char *type_buf = hdr; |
5180cacc | 1392 | unsigned long size; |
46f03448 | 1393 | int type, type_len = 0; |
43057304 | 1394 | |
5180cacc | 1395 | /* |
46f03448 | 1396 | * The type can be of any size but is followed by |
21666f1a | 1397 | * a space. |
5180cacc | 1398 | */ |
5180cacc LT |
1399 | for (;;) { |
1400 | char c = *hdr++; | |
d21f8426 JH |
1401 | if (!c) |
1402 | return -1; | |
5180cacc LT |
1403 | if (c == ' ') |
1404 | break; | |
46f03448 | 1405 | type_len++; |
5180cacc | 1406 | } |
1f688557 | 1407 | |
46f03448 | 1408 | type = type_from_string_gently(type_buf, type_len, 1); |
6ca32f47 BW |
1409 | if (oi->type_name) |
1410 | strbuf_add(oi->type_name, type_buf, type_len); | |
46f03448 KN |
1411 | /* |
1412 | * Set type to 0 if its an unknown object and | |
2e3a16b2 | 1413 | * we're obtaining the type using '--allow-unknown-type' |
46f03448 KN |
1414 | * option. |
1415 | */ | |
19fc5e84 | 1416 | if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0)) |
46f03448 KN |
1417 | type = 0; |
1418 | else if (type < 0) | |
259328b7 | 1419 | die(_("invalid object type")); |
46f03448 KN |
1420 | if (oi->typep) |
1421 | *oi->typep = type; | |
5180cacc LT |
1422 | |
1423 | /* | |
1424 | * The length must follow immediately, and be in canonical | |
1425 | * decimal format (ie "010" is not valid). | |
1426 | */ | |
1427 | size = *hdr++ - '0'; | |
1428 | if (size > 9) | |
1429 | return -1; | |
1430 | if (size) { | |
1431 | for (;;) { | |
1432 | unsigned long c = *hdr - '0'; | |
1433 | if (c > 9) | |
1434 | break; | |
1435 | hdr++; | |
1436 | size = size * 10 + c; | |
1b1005d1 | 1437 | } |
c01f51cc | 1438 | } |
46f03448 KN |
1439 | |
1440 | if (oi->sizep) | |
1441 | *oi->sizep = size; | |
5180cacc LT |
1442 | |
1443 | /* | |
1444 | * The length must be followed by a zero byte | |
1445 | */ | |
46f03448 | 1446 | return *hdr ? -1 : type; |
1f688557 JH |
1447 | } |
1448 | ||
00a7760e | 1449 | int parse_loose_header(const char *hdr, unsigned long *sizep) |
bf592c50 | 1450 | { |
27b5c1a0 | 1451 | struct object_info oi = OBJECT_INFO_INIT; |
21666f1a | 1452 | |
46f03448 | 1453 | oi.sizep = sizep; |
00a7760e | 1454 | return parse_loose_header_extended(hdr, &oi, 0); |
bf592c50 DB |
1455 | } |
1456 | ||
514c5fdd JK |
1457 | static int loose_object_info(struct repository *r, |
1458 | const struct object_id *oid, | |
1459 | struct object_info *oi, int flags) | |
65c2e0c3 | 1460 | { |
46f03448 KN |
1461 | int status = 0; |
1462 | unsigned long mapsize; | |
65c2e0c3 | 1463 | void *map; |
ef49a7a0 | 1464 | git_zstream stream; |
1af64f73 | 1465 | char hdr[MAX_HEADER_LEN]; |
46f03448 | 1466 | struct strbuf hdrbuf = STRBUF_INIT; |
c84a1f3e | 1467 | unsigned long size_scratch; |
65c2e0c3 | 1468 | |
b99b6bcc JK |
1469 | if (oi->delta_base_oid) |
1470 | oidclr(oi->delta_base_oid); | |
5d642e75 | 1471 | |
052fe5ea JK |
1472 | /* |
1473 | * If we don't care about type or size, then we don't | |
4ef8d1dd JH |
1474 | * need to look inside the object at all. Note that we |
1475 | * do not optimize out the stat call, even if the | |
1476 | * caller doesn't care about the disk-size, since our | |
1477 | * return value implicitly indicates whether the | |
1478 | * object even exists. | |
052fe5ea | 1479 | */ |
6ca32f47 | 1480 | if (!oi->typep && !oi->type_name && !oi->sizep && !oi->contentp) { |
771e7d57 | 1481 | const char *path; |
4ef8d1dd | 1482 | struct stat st; |
61c7711c | 1483 | if (!oi->disk_sizep && (flags & OBJECT_INFO_QUICK)) |
d7a24573 | 1484 | return quick_has_loose(r, oid) ? 0 : -1; |
514c5fdd | 1485 | if (stat_loose_object(r, oid, &st, &path) < 0) |
4ef8d1dd JH |
1486 | return -1; |
1487 | if (oi->disk_sizep) | |
23c339c0 | 1488 | *oi->disk_sizep = st.st_size; |
052fe5ea JK |
1489 | return 0; |
1490 | } | |
1491 | ||
514c5fdd | 1492 | map = map_loose_object(r, oid, &mapsize); |
f0df4ed5 | 1493 | if (!map) |
dbea72a8 | 1494 | return -1; |
c84a1f3e JT |
1495 | |
1496 | if (!oi->sizep) | |
1497 | oi->sizep = &size_scratch; | |
1498 | ||
23c339c0 JK |
1499 | if (oi->disk_sizep) |
1500 | *oi->disk_sizep = mapsize; | |
19fc5e84 | 1501 | if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) { |
00a7760e | 1502 | if (unpack_loose_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0) |
259328b7 | 1503 | status = error(_("unable to unpack %s header with --allow-unknown-type"), |
514c5fdd | 1504 | oid_to_hex(oid)); |
00a7760e | 1505 | } else if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) |
259328b7 | 1506 | status = error(_("unable to unpack %s header"), |
514c5fdd | 1507 | oid_to_hex(oid)); |
46f03448 KN |
1508 | if (status < 0) |
1509 | ; /* Do nothing */ | |
1510 | else if (hdrbuf.len) { | |
00a7760e | 1511 | if ((status = parse_loose_header_extended(hdrbuf.buf, oi, flags)) < 0) |
259328b7 | 1512 | status = error(_("unable to parse %s header with --allow-unknown-type"), |
514c5fdd | 1513 | oid_to_hex(oid)); |
00a7760e | 1514 | } else if ((status = parse_loose_header_extended(hdr, oi, flags)) < 0) |
514c5fdd | 1515 | status = error(_("unable to parse %s header"), oid_to_hex(oid)); |
c84a1f3e | 1516 | |
b3ea7dd3 | 1517 | if (status >= 0 && oi->contentp) { |
00a7760e JK |
1518 | *oi->contentp = unpack_loose_rest(&stream, hdr, |
1519 | *oi->sizep, oid); | |
b3ea7dd3 JK |
1520 | if (!*oi->contentp) { |
1521 | git_inflate_end(&stream); | |
1522 | status = -1; | |
1523 | } | |
1524 | } else | |
c84a1f3e JT |
1525 | git_inflate_end(&stream); |
1526 | ||
65c2e0c3 | 1527 | munmap(map, mapsize); |
46f03448 | 1528 | if (status && oi->typep) |
23c339c0 | 1529 | *oi->typep = status; |
c84a1f3e JT |
1530 | if (oi->sizep == &size_scratch) |
1531 | oi->sizep = NULL; | |
46f03448 | 1532 | strbuf_release(&hdrbuf); |
3ab0fb06 | 1533 | oi->whence = OI_LOOSE; |
93cff9a9 | 1534 | return (status < 0) ? status : 0; |
65c2e0c3 JH |
1535 | } |
1536 | ||
31877c9a MT |
1537 | int obj_read_use_lock = 0; |
1538 | pthread_mutex_t obj_read_mutex; | |
1539 | ||
1540 | void enable_obj_read_lock(void) | |
1541 | { | |
1542 | if (obj_read_use_lock) | |
1543 | return; | |
1544 | ||
1545 | obj_read_use_lock = 1; | |
1546 | init_recursive_mutex(&obj_read_mutex); | |
1547 | } | |
1548 | ||
1549 | void disable_obj_read_lock(void) | |
1550 | { | |
1551 | if (!obj_read_use_lock) | |
1552 | return; | |
1553 | ||
1554 | obj_read_use_lock = 0; | |
1555 | pthread_mutex_destroy(&obj_read_mutex); | |
1556 | } | |
1557 | ||
8b4c0103 JT |
1558 | int fetch_if_missing = 1; |
1559 | ||
31877c9a MT |
1560 | static int do_oid_object_info_extended(struct repository *r, |
1561 | const struct object_id *oid, | |
1562 | struct object_info *oi, unsigned flags) | |
f0df4ed5 | 1563 | { |
cd585e2a | 1564 | static struct object_info blank_oi = OBJECT_INFO_INIT; |
9c8a294a | 1565 | struct cached_object *co; |
f0df4ed5 | 1566 | struct pack_entry e; |
5b086407 | 1567 | int rtype; |
b383a13c | 1568 | const struct object_id *real = oid; |
8b4c0103 | 1569 | int already_retried = 0; |
f0df4ed5 | 1570 | |
31877c9a | 1571 | |
b383a13c | 1572 | if (flags & OBJECT_INFO_LOOKUP_REPLACE) |
9d98354f | 1573 | real = lookup_replace_object(r, oid); |
f0df4ed5 | 1574 | |
b383a13c | 1575 | if (is_null_oid(real)) |
87b5e236 JK |
1576 | return -1; |
1577 | ||
cd585e2a JT |
1578 | if (!oi) |
1579 | oi = &blank_oi; | |
1580 | ||
9c8a294a JT |
1581 | co = find_cached_object(real); |
1582 | if (co) { | |
1583 | if (oi->typep) | |
1584 | *(oi->typep) = co->type; | |
1585 | if (oi->sizep) | |
1586 | *(oi->sizep) = co->size; | |
1587 | if (oi->disk_sizep) | |
1588 | *(oi->disk_sizep) = 0; | |
b99b6bcc JK |
1589 | if (oi->delta_base_oid) |
1590 | oidclr(oi->delta_base_oid); | |
9c8a294a JT |
1591 | if (oi->type_name) |
1592 | strbuf_addstr(oi->type_name, type_name(co->type)); | |
1593 | if (oi->contentp) | |
1594 | *oi->contentp = xmemdupz(co->buf, co->size); | |
1595 | oi->whence = OI_CACHED; | |
1596 | return 0; | |
c4d9986f NTND |
1597 | } |
1598 | ||
8b4c0103 | 1599 | while (1) { |
42c8ce1c | 1600 | if (find_pack_entry(r, real, &e)) |
8b4c0103 JT |
1601 | break; |
1602 | ||
024aa469 TI |
1603 | if (flags & OBJECT_INFO_IGNORE_LOOSE) |
1604 | return -1; | |
1605 | ||
ddd63e64 | 1606 | /* Most likely it's a loose object. */ |
514c5fdd | 1607 | if (!loose_object_info(r, real, oi, flags)) |
5b086407 | 1608 | return 0; |
ddd63e64 SG |
1609 | |
1610 | /* Not a loose object; someone else may have just packed it. */ | |
2b7750c9 | 1611 | if (!(flags & OBJECT_INFO_QUICK)) { |
9d98354f | 1612 | reprepare_packed_git(r); |
42c8ce1c | 1613 | if (find_pack_entry(r, real, &e)) |
2b7750c9 JT |
1614 | break; |
1615 | } | |
8b4c0103 | 1616 | |
a35e03de JT |
1617 | if (register_all_submodule_odb_as_alternates()) |
1618 | /* We added some alternates; retry */ | |
1619 | continue; | |
1620 | ||
8b4c0103 | 1621 | /* Check if it is a missing object */ |
ef830cc4 JT |
1622 | if (fetch_if_missing && repo_has_promisor_remote(r) && |
1623 | !already_retried && | |
31f5256c | 1624 | !(flags & OBJECT_INFO_SKIP_FETCH_OBJECT)) { |
8b4c0103 | 1625 | /* |
b14ed5ad CC |
1626 | * TODO Investigate checking promisor_remote_get_direct() |
1627 | * TODO return value and stopping on error here. | |
8b4c0103 | 1628 | */ |
b14ed5ad | 1629 | promisor_remote_get_direct(r, real, 1); |
8b4c0103 JT |
1630 | already_retried = 1; |
1631 | continue; | |
dfdd4afc | 1632 | } |
8b4c0103 JT |
1633 | |
1634 | return -1; | |
f0df4ed5 | 1635 | } |
3d77d877 | 1636 | |
cd585e2a JT |
1637 | if (oi == &blank_oi) |
1638 | /* | |
1639 | * We know that the caller doesn't actually need the | |
1640 | * information below, so return early. | |
1641 | */ | |
1642 | return 0; | |
9d98354f | 1643 | rtype = packed_object_info(r, e.p, e.offset, oi); |
412916ee | 1644 | if (rtype < 0) { |
751530de | 1645 | mark_bad_packed_object(e.p, real); |
31877c9a | 1646 | return do_oid_object_info_extended(r, real, oi, 0); |
3ab0fb06 | 1647 | } else if (oi->whence == OI_PACKED) { |
9a490590 JH |
1648 | oi->u.packed.offset = e.offset; |
1649 | oi->u.packed.pack = e.p; | |
1650 | oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA || | |
1651 | rtype == OBJ_OFS_DELTA); | |
3d77d877 NP |
1652 | } |
1653 | ||
5b086407 | 1654 | return 0; |
f0df4ed5 JS |
1655 | } |
1656 | ||
31877c9a MT |
1657 | int oid_object_info_extended(struct repository *r, const struct object_id *oid, |
1658 | struct object_info *oi, unsigned flags) | |
1659 | { | |
1660 | int ret; | |
1661 | obj_read_lock(); | |
1662 | ret = do_oid_object_info_extended(r, oid, oi, flags); | |
1663 | obj_read_unlock(); | |
1664 | return ret; | |
1665 | } | |
1666 | ||
1667 | ||
3fc0dca9 | 1668 | /* returns enum object_type or negative */ |
9d98354f SB |
1669 | int oid_object_info(struct repository *r, |
1670 | const struct object_id *oid, | |
1671 | unsigned long *sizep) | |
9a490590 | 1672 | { |
5b086407 | 1673 | enum object_type type; |
27b5c1a0 | 1674 | struct object_info oi = OBJECT_INFO_INIT; |
9a490590 | 1675 | |
5b086407 | 1676 | oi.typep = &type; |
9a490590 | 1677 | oi.sizep = sizep; |
9d98354f SB |
1678 | if (oid_object_info_extended(r, oid, &oi, |
1679 | OBJECT_INFO_LOOKUP_REPLACE) < 0) | |
5b086407 JK |
1680 | return -1; |
1681 | return type; | |
9a490590 JH |
1682 | } |
1683 | ||
1b9b5c69 | 1684 | static void *read_object(struct repository *r, |
cba595ab | 1685 | const struct object_id *oid, enum object_type *type, |
f1d8130b JT |
1686 | unsigned long *size) |
1687 | { | |
1688 | struct object_info oi = OBJECT_INFO_INIT; | |
1689 | void *content; | |
1690 | oi.typep = type; | |
1691 | oi.sizep = size; | |
1692 | oi.contentp = &content; | |
1693 | ||
cba595ab | 1694 | if (oid_object_info_extended(r, oid, &oi, 0) < 0) |
f1d8130b JT |
1695 | return NULL; |
1696 | return content; | |
1697 | } | |
1698 | ||
829e5c3b PO |
1699 | int pretend_object_file(void *buf, unsigned long len, enum object_type type, |
1700 | struct object_id *oid) | |
d66b37bb JH |
1701 | { |
1702 | struct cached_object *co; | |
1703 | ||
2dcde20e | 1704 | hash_object_file(the_hash_algo, buf, len, type_name(type), oid); |
a64d2aae JT |
1705 | if (has_object_file_with_flags(oid, OBJECT_INFO_QUICK | OBJECT_INFO_SKIP_FETCH_OBJECT) || |
1706 | find_cached_object(oid)) | |
d66b37bb | 1707 | return 0; |
c7353967 | 1708 | ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc); |
d66b37bb JH |
1709 | co = &cached_objects[cached_object_nr++]; |
1710 | co->size = len; | |
21666f1a | 1711 | co->type = type; |
efa13f7b JH |
1712 | co->buf = xmalloc(len); |
1713 | memcpy(co->buf, buf, len); | |
62ba93ea | 1714 | oidcpy(&co->oid, oid); |
d66b37bb JH |
1715 | return 0; |
1716 | } | |
1717 | ||
b6c4cecc JH |
1718 | /* |
1719 | * This function dies on corrupt objects; the callers who want to | |
1720 | * deal with them should arrange to call read_object() and give error | |
1721 | * messages themselves. | |
1722 | */ | |
a3b72c89 SB |
1723 | void *read_object_file_extended(struct repository *r, |
1724 | const struct object_id *oid, | |
b4f5aca4 | 1725 | enum object_type *type, |
1726 | unsigned long *size, | |
1727 | int lookup_replace) | |
ac939109 | 1728 | { |
3ba7a065 | 1729 | void *data; |
b6c4cecc | 1730 | const struct packed_git *p; |
771e7d57 JK |
1731 | const char *path; |
1732 | struct stat st; | |
1f2e7cea | 1733 | const struct object_id *repl = lookup_replace ? |
a3b72c89 | 1734 | lookup_replace_object(r, oid) : oid; |
b6c4cecc | 1735 | |
3ba7a065 | 1736 | errno = 0; |
cba595ab | 1737 | data = read_object(r, repl, type, size); |
4bbf5a26 | 1738 | if (data) |
b6c4cecc | 1739 | return data; |
68095570 | 1740 | |
31877c9a | 1741 | obj_read_lock(); |
25f3af3f | 1742 | if (errno && errno != ENOENT) |
259328b7 | 1743 | die_errno(_("failed to read object %s"), oid_to_hex(oid)); |
3ba7a065 | 1744 | |
68095570 | 1745 | /* die if we replaced an object with one that does not exist */ |
b383a13c | 1746 | if (repl != oid) |
259328b7 | 1747 | die(_("replacement %s not found for %s"), |
b383a13c | 1748 | oid_to_hex(repl), oid_to_hex(oid)); |
68095570 | 1749 | |
cba595ab | 1750 | if (!stat_loose_object(r, repl, &st, &path)) |
259328b7 | 1751 | die(_("loose object %s (stored in %s) is corrupt"), |
b383a13c | 1752 | oid_to_hex(repl), path); |
68095570 | 1753 | |
7407d733 | 1754 | if ((p = has_packed_and_bad(r, repl)) != NULL) |
259328b7 | 1755 | die(_("packed object %s (stored in %s) is corrupt"), |
b383a13c | 1756 | oid_to_hex(repl), p->pack_name); |
31877c9a | 1757 | obj_read_unlock(); |
f5552aee | 1758 | |
b6c4cecc | 1759 | return NULL; |
ac939109 NP |
1760 | } |
1761 | ||
d3b4705a NTND |
1762 | void *read_object_with_reference(struct repository *r, |
1763 | const struct object_id *oid, | |
21666f1a | 1764 | const char *required_type_name, |
40469ee9 | 1765 | unsigned long *size, |
02f0547e | 1766 | struct object_id *actual_oid_return) |
f4913f91 | 1767 | { |
21666f1a | 1768 | enum object_type type, required_type; |
f4913f91 JH |
1769 | void *buffer; |
1770 | unsigned long isize; | |
02f0547e | 1771 | struct object_id actual_oid; |
f4913f91 | 1772 | |
21666f1a | 1773 | required_type = type_from_string(required_type_name); |
02f0547e | 1774 | oidcpy(&actual_oid, oid); |
40469ee9 JH |
1775 | while (1) { |
1776 | int ref_length = -1; | |
1777 | const char *ref_type = NULL; | |
f4913f91 | 1778 | |
d3b4705a | 1779 | buffer = repo_read_object_file(r, &actual_oid, &type, &isize); |
40469ee9 JH |
1780 | if (!buffer) |
1781 | return NULL; | |
21666f1a | 1782 | if (type == required_type) { |
40469ee9 | 1783 | *size = isize; |
02f0547e | 1784 | if (actual_oid_return) |
1785 | oidcpy(actual_oid_return, &actual_oid); | |
40469ee9 JH |
1786 | return buffer; |
1787 | } | |
1788 | /* Handle references */ | |
21666f1a | 1789 | else if (type == OBJ_COMMIT) |
40469ee9 | 1790 | ref_type = "tree "; |
21666f1a | 1791 | else if (type == OBJ_TAG) |
40469ee9 JH |
1792 | ref_type = "object "; |
1793 | else { | |
1794 | free(buffer); | |
1795 | return NULL; | |
1796 | } | |
1797 | ref_length = strlen(ref_type); | |
f4913f91 | 1798 | |
94b5e093 | 1799 | if (ref_length + the_hash_algo->hexsz > isize || |
50974ec9 | 1800 | memcmp(buffer, ref_type, ref_length) || |
02f0547e | 1801 | get_oid_hex((char *) buffer + ref_length, &actual_oid)) { |
40469ee9 JH |
1802 | free(buffer); |
1803 | return NULL; | |
1804 | } | |
1cf58e72 | 1805 | free(buffer); |
40469ee9 | 1806 | /* Now we have the ID of the referred-to object in |
02f0547e | 1807 | * actual_oid. Check again. */ |
f4913f91 | 1808 | } |
f4913f91 JH |
1809 | } |
1810 | ||
7ad5c44d MT |
1811 | static void write_object_file_prepare(const struct git_hash_algo *algo, |
1812 | const void *buf, unsigned long len, | |
a09c985e PO |
1813 | const char *type, struct object_id *oid, |
1814 | char *hdr, int *hdrlen) | |
d410c0f5 | 1815 | { |
18e2588e | 1816 | git_hash_ctx c; |
d410c0f5 JH |
1817 | |
1818 | /* Generate the header */ | |
ca473cef | 1819 | *hdrlen = xsnprintf(hdr, *hdrlen, "%s %"PRIuMAX , type, (uintmax_t)len)+1; |
d410c0f5 JH |
1820 | |
1821 | /* Sha1.. */ | |
7ad5c44d MT |
1822 | algo->init_fn(&c); |
1823 | algo->update_fn(&c, hdr, *hdrlen); | |
1824 | algo->update_fn(&c, buf, len); | |
5951bf46 | 1825 | algo->final_oid_fn(oid, &c); |
d410c0f5 JH |
1826 | } |
1827 | ||
230f1322 | 1828 | /* |
5a688fe4 | 1829 | * Move the just written object into its final resting place. |
230f1322 | 1830 | */ |
cb5add58 | 1831 | int finalize_object_file(const char *tmpfile, const char *filename) |
230f1322 | 1832 | { |
e32c0a9c | 1833 | int ret = 0; |
5a688fe4 | 1834 | |
348df166 | 1835 | if (object_creation_mode == OBJECT_CREATION_USES_RENAMES) |
be66a6c4 JS |
1836 | goto try_rename; |
1837 | else if (link(tmpfile, filename)) | |
e32c0a9c | 1838 | ret = errno; |
7ebb6fca LT |
1839 | |
1840 | /* | |
1841 | * Coda hack - coda doesn't like cross-directory links, | |
1842 | * so we fall back to a rename, which will mean that it | |
1843 | * won't be able to check collisions, but that's not a | |
1844 | * big deal. | |
1845 | * | |
1846 | * The same holds for FAT formatted media. | |
1847 | * | |
3be1f18e | 1848 | * When this succeeds, we just return. We have nothing |
7ebb6fca LT |
1849 | * left to unlink. |
1850 | */ | |
1851 | if (ret && ret != EEXIST) { | |
be66a6c4 | 1852 | try_rename: |
7ebb6fca | 1853 | if (!rename(tmpfile, filename)) |
3be1f18e | 1854 | goto out; |
9e48b389 | 1855 | ret = errno; |
230f1322 | 1856 | } |
691f1a28 | 1857 | unlink_or_warn(tmpfile); |
230f1322 LT |
1858 | if (ret) { |
1859 | if (ret != EEXIST) { | |
2c319886 | 1860 | return error_errno(_("unable to write file %s"), filename); |
230f1322 LT |
1861 | } |
1862 | /* FIXME!!! Collision check here ? */ | |
1863 | } | |
1864 | ||
3be1f18e | 1865 | out: |
5256b006 | 1866 | if (adjust_shared_perm(filename)) |
259328b7 | 1867 | return error(_("unable to set permission to '%s'"), filename); |
230f1322 LT |
1868 | return 0; |
1869 | } | |
1870 | ||
4d548150 LT |
1871 | static int write_buffer(int fd, const void *buf, size_t len) |
1872 | { | |
d34cf19b | 1873 | if (write_in_full(fd, buf, len) < 0) |
259328b7 | 1874 | return error_errno(_("file write error")); |
4d548150 LT |
1875 | return 0; |
1876 | } | |
1877 | ||
2dcde20e MT |
1878 | int hash_object_file(const struct git_hash_algo *algo, const void *buf, |
1879 | unsigned long len, const char *type, | |
f070facc | 1880 | struct object_id *oid) |
abdc3fc8 | 1881 | { |
1af64f73 | 1882 | char hdr[MAX_HEADER_LEN]; |
ef1286d3 | 1883 | int hdrlen = sizeof(hdr); |
2dcde20e | 1884 | write_object_file_prepare(algo, buf, len, type, oid, hdr, &hdrlen); |
abdc3fc8 RS |
1885 | return 0; |
1886 | } | |
1887 | ||
e9039dd3 | 1888 | /* Finalize a file on disk, and close it. */ |
514c5fdd | 1889 | static void close_loose_object(int fd) |
e9039dd3 | 1890 | { |
aafe9fba | 1891 | if (fsync_object_files) |
76011357 | 1892 | fsync_or_die(fd, "loose object file"); |
e9039dd3 | 1893 | if (close(fd) != 0) |
76011357 | 1894 | die_errno(_("error when closing loose object file")); |
e9039dd3 LT |
1895 | } |
1896 | ||
5723fe7e LT |
1897 | /* Size of directory component, including the ending '/' */ |
1898 | static inline int directory_size(const char *filename) | |
1899 | { | |
1900 | const char *s = strrchr(filename, '/'); | |
1901 | if (!s) | |
1902 | return 0; | |
1903 | return s - filename + 1; | |
1904 | } | |
1905 | ||
1906 | /* | |
1907 | * This creates a temporary file in the same directory as the final | |
1908 | * 'filename' | |
1909 | * | |
1910 | * We want to avoid cross-directory filename renames, because those | |
1911 | * can have problems on various filesystems (FAT, NFS, Coda). | |
1912 | */ | |
d4b3d11a | 1913 | static int create_tmpfile(struct strbuf *tmp, const char *filename) |
5723fe7e LT |
1914 | { |
1915 | int fd, dirlen = directory_size(filename); | |
1916 | ||
d4b3d11a JK |
1917 | strbuf_reset(tmp); |
1918 | strbuf_add(tmp, filename, dirlen); | |
1919 | strbuf_addstr(tmp, "tmp_obj_XXXXXX"); | |
1920 | fd = git_mkstemp_mode(tmp->buf, 0444); | |
cbacbf4e | 1921 | if (fd < 0 && dirlen && errno == ENOENT) { |
d4b3d11a JK |
1922 | /* |
1923 | * Make sure the directory exists; note that the contents | |
1924 | * of the buffer are undefined after mkstemp returns an | |
1925 | * error, so we have to rewrite the whole buffer from | |
1926 | * scratch. | |
1927 | */ | |
1928 | strbuf_reset(tmp); | |
1929 | strbuf_add(tmp, filename, dirlen - 1); | |
1930 | if (mkdir(tmp->buf, 0777) && errno != EEXIST) | |
b2476a60 | 1931 | return -1; |
d4b3d11a | 1932 | if (adjust_shared_perm(tmp->buf)) |
5723fe7e LT |
1933 | return -1; |
1934 | ||
1935 | /* Try again */ | |
d4b3d11a JK |
1936 | strbuf_addstr(tmp, "/tmp_obj_XXXXXX"); |
1937 | fd = git_mkstemp_mode(tmp->buf, 0444); | |
5723fe7e LT |
1938 | } |
1939 | return fd; | |
1940 | } | |
1941 | ||
3fc7281f PO |
1942 | static int write_loose_object(const struct object_id *oid, char *hdr, |
1943 | int hdrlen, const void *buf, unsigned long len, | |
1944 | time_t mtime) | |
0fcfd160 | 1945 | { |
915308b1 | 1946 | int fd, ret; |
9892beba | 1947 | unsigned char compressed[4096]; |
ef49a7a0 | 1948 | git_zstream stream; |
18e2588e | 1949 | git_hash_ctx c; |
3fc7281f | 1950 | struct object_id parano_oid; |
d4b3d11a | 1951 | static struct strbuf tmp_file = STRBUF_INIT; |
ea657730 CC |
1952 | static struct strbuf filename = STRBUF_INIT; |
1953 | ||
514c5fdd | 1954 | loose_object_path(the_repository, &filename, oid); |
a44c9a5e | 1955 | |
ea657730 | 1956 | fd = create_tmpfile(&tmp_file, filename.buf); |
aac17941 | 1957 | if (fd < 0) { |
35243577 | 1958 | if (errno == EACCES) |
259328b7 | 1959 | return error(_("insufficient permission for adding an object to repository database %s"), get_object_directory()); |
916d081b | 1960 | else |
259328b7 | 1961 | return error_errno(_("unable to create temporary file")); |
aac17941 LT |
1962 | } |
1963 | ||
0fcfd160 | 1964 | /* Set it up */ |
55bb5c91 | 1965 | git_deflate_init(&stream, zlib_compression_level); |
0fcfd160 | 1966 | stream.next_out = compressed; |
9892beba | 1967 | stream.avail_out = sizeof(compressed); |
18e2588e | 1968 | the_hash_algo->init_fn(&c); |
a44c9a5e LT |
1969 | |
1970 | /* First header.. */ | |
d65a16f6 | 1971 | stream.next_in = (unsigned char *)hdr; |
a44c9a5e | 1972 | stream.avail_in = hdrlen; |
55bb5c91 JH |
1973 | while (git_deflate(&stream, 0) == Z_OK) |
1974 | ; /* nothing */ | |
18e2588e | 1975 | the_hash_algo->update_fn(&c, hdr, hdrlen); |
a44c9a5e LT |
1976 | |
1977 | /* Then the data itself.. */ | |
c00e657d | 1978 | stream.next_in = (void *)buf; |
a44c9a5e | 1979 | stream.avail_in = len; |
9892beba | 1980 | do { |
748af44c | 1981 | unsigned char *in0 = stream.next_in; |
55bb5c91 | 1982 | ret = git_deflate(&stream, Z_FINISH); |
18e2588e | 1983 | the_hash_algo->update_fn(&c, in0, stream.next_in - in0); |
9892beba | 1984 | if (write_buffer(fd, compressed, stream.next_out - compressed) < 0) |
76011357 | 1985 | die(_("unable to write loose object file")); |
9892beba NP |
1986 | stream.next_out = compressed; |
1987 | stream.avail_out = sizeof(compressed); | |
1988 | } while (ret == Z_OK); | |
1989 | ||
ac54c277 | 1990 | if (ret != Z_STREAM_END) |
259328b7 | 1991 | die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid), |
3fc7281f | 1992 | ret); |
55bb5c91 | 1993 | ret = git_deflate_end_gently(&stream); |
ac54c277 | 1994 | if (ret != Z_OK) |
259328b7 | 1995 | die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid), |
3fc7281f | 1996 | ret); |
5951bf46 | 1997 | the_hash_algo->final_oid_fn(¶no_oid, &c); |
9001dc2a | 1998 | if (!oideq(oid, ¶no_oid)) |
259328b7 | 1999 | die(_("confused by unstable object source data for %s"), |
3fc7281f | 2000 | oid_to_hex(oid)); |
ac54c277 | 2001 | |
514c5fdd | 2002 | close_loose_object(fd); |
0fcfd160 | 2003 | |
bbac7311 NP |
2004 | if (mtime) { |
2005 | struct utimbuf utb; | |
2006 | utb.actime = mtime; | |
2007 | utb.modtime = mtime; | |
d4b3d11a | 2008 | if (utime(tmp_file.buf, &utb) < 0) |
259328b7 | 2009 | warning_errno(_("failed utime() on %s"), tmp_file.buf); |
bbac7311 NP |
2010 | } |
2011 | ||
ea657730 | 2012 | return finalize_object_file(tmp_file.buf, filename.buf); |
0fcfd160 | 2013 | } |
8237b185 | 2014 | |
6862ebbf | 2015 | static int freshen_loose_object(const struct object_id *oid) |
33d4221c | 2016 | { |
6862ebbf | 2017 | return check_and_freshen(oid, 1); |
33d4221c JK |
2018 | } |
2019 | ||
6862ebbf | 2020 | static int freshen_packed_object(const struct object_id *oid) |
33d4221c JK |
2021 | { |
2022 | struct pack_entry e; | |
544443cb | 2023 | if (!find_pack_entry(the_repository, oid, &e)) |
ee1c6c34 JK |
2024 | return 0; |
2025 | if (e.p->freshened) | |
2026 | return 1; | |
2027 | if (!freshen_file(e.p->pack_name)) | |
2028 | return 0; | |
2029 | e.p->freshened = 1; | |
2030 | return 1; | |
33d4221c JK |
2031 | } |
2032 | ||
a09c985e PO |
2033 | int write_object_file(const void *buf, unsigned long len, const char *type, |
2034 | struct object_id *oid) | |
bbac7311 | 2035 | { |
1af64f73 | 2036 | char hdr[MAX_HEADER_LEN]; |
ef1286d3 | 2037 | int hdrlen = sizeof(hdr); |
bbac7311 NP |
2038 | |
2039 | /* Normally if we have it in the pack then we do not bother writing | |
2040 | * it out into .git/objects/??/?{38} file. | |
2041 | */ | |
7ad5c44d MT |
2042 | write_object_file_prepare(the_hash_algo, buf, len, type, oid, hdr, |
2043 | &hdrlen); | |
6862ebbf | 2044 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) |
bbac7311 | 2045 | return 0; |
3fc7281f | 2046 | return write_loose_object(oid, hdr, hdrlen, buf, len, 0); |
bbac7311 NP |
2047 | } |
2048 | ||
1752cbbc PO |
2049 | int hash_object_file_literally(const void *buf, unsigned long len, |
2050 | const char *type, struct object_id *oid, | |
2051 | unsigned flags) | |
0c3db67c ES |
2052 | { |
2053 | char *header; | |
2054 | int hdrlen, status = 0; | |
2055 | ||
2056 | /* type string, SP, %lu of the length plus NUL must fit this */ | |
1af64f73 | 2057 | hdrlen = strlen(type) + MAX_HEADER_LEN; |
ef1286d3 | 2058 | header = xmalloc(hdrlen); |
7ad5c44d MT |
2059 | write_object_file_prepare(the_hash_algo, buf, len, type, oid, header, |
2060 | &hdrlen); | |
0c3db67c ES |
2061 | |
2062 | if (!(flags & HASH_WRITE_OBJECT)) | |
2063 | goto cleanup; | |
6862ebbf | 2064 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) |
0c3db67c | 2065 | goto cleanup; |
3fc7281f | 2066 | status = write_loose_object(oid, header, hdrlen, buf, len, 0); |
0c3db67c ES |
2067 | |
2068 | cleanup: | |
2069 | free(header); | |
2070 | return status; | |
2071 | } | |
2072 | ||
4bdb70a4 | 2073 | int force_object_loose(const struct object_id *oid, time_t mtime) |
bbac7311 | 2074 | { |
bbac7311 NP |
2075 | void *buf; |
2076 | unsigned long len; | |
2077 | enum object_type type; | |
1af64f73 | 2078 | char hdr[MAX_HEADER_LEN]; |
bbac7311 | 2079 | int hdrlen; |
1fb23e65 | 2080 | int ret; |
bbac7311 | 2081 | |
6862ebbf | 2082 | if (has_loose_object(oid)) |
bbac7311 | 2083 | return 0; |
cba595ab | 2084 | buf = read_object(the_repository, oid, &type, &len); |
bbac7311 | 2085 | if (!buf) |
2c319886 | 2086 | return error(_("cannot read object for %s"), oid_to_hex(oid)); |
ca473cef | 2087 | hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(type), (uintmax_t)len) + 1; |
3fc7281f | 2088 | ret = write_loose_object(oid, hdr, hdrlen, buf, len, mtime); |
1fb23e65 BS |
2089 | free(buf); |
2090 | ||
2091 | return ret; | |
bbac7311 NP |
2092 | } |
2093 | ||
1d8d9cb6 JT |
2094 | int has_object(struct repository *r, const struct object_id *oid, |
2095 | unsigned flags) | |
2096 | { | |
2097 | int quick = !(flags & HAS_OBJECT_RECHECK_PACKED); | |
2098 | unsigned object_info_flags = OBJECT_INFO_SKIP_FETCH_OBJECT | | |
2099 | (quick ? OBJECT_INFO_QUICK : 0); | |
2100 | ||
2101 | if (!startup_info->have_repository) | |
2102 | return 0; | |
2103 | return oid_object_info_extended(r, oid, NULL, object_info_flags) >= 0; | |
2104 | } | |
2105 | ||
cba595ab JH |
2106 | int repo_has_object_file_with_flags(struct repository *r, |
2107 | const struct object_id *oid, int flags) | |
8237b185 | 2108 | { |
3e8b7d3c JN |
2109 | if (!startup_info->have_repository) |
2110 | return 0; | |
9c8a294a | 2111 | return oid_object_info_extended(r, oid, NULL, flags) >= 0; |
b419aa25 | 2112 | } |
2113 | ||
9b45f499 SB |
2114 | int repo_has_object_file(struct repository *r, |
2115 | const struct object_id *oid) | |
b419aa25 | 2116 | { |
cba595ab | 2117 | return repo_has_object_file_with_flags(r, oid, 0); |
5827a035 JK |
2118 | } |
2119 | ||
c879daa2 NTND |
2120 | static void check_tree(const void *buf, size_t size) |
2121 | { | |
2122 | struct tree_desc desc; | |
2123 | struct name_entry entry; | |
2124 | ||
2125 | init_tree_desc(&desc, buf, size); | |
2126 | while (tree_entry(&desc, &entry)) | |
2127 | /* do nothing | |
2128 | * tree_entry() will die() on malformed entries */ | |
2129 | ; | |
2130 | } | |
2131 | ||
2132 | static void check_commit(const void *buf, size_t size) | |
2133 | { | |
2134 | struct commit c; | |
2135 | memset(&c, 0, sizeof(c)); | |
08f4f445 | 2136 | if (parse_commit_buffer(the_repository, &c, buf, size, 0)) |
259328b7 | 2137 | die(_("corrupt commit")); |
c879daa2 NTND |
2138 | } |
2139 | ||
2140 | static void check_tag(const void *buf, size_t size) | |
2141 | { | |
2142 | struct tag t; | |
2143 | memset(&t, 0, sizeof(t)); | |
0e740fed | 2144 | if (parse_tag_buffer(the_repository, &t, buf, size)) |
259328b7 | 2145 | die(_("corrupt tag")); |
c879daa2 NTND |
2146 | } |
2147 | ||
58bf2a4c NTND |
2148 | static int index_mem(struct index_state *istate, |
2149 | struct object_id *oid, void *buf, size_t size, | |
c4ce46fc JH |
2150 | enum object_type type, |
2151 | const char *path, unsigned flags) | |
e7332f96 | 2152 | { |
6c510bee | 2153 | int ret, re_allocated = 0; |
c4ce46fc | 2154 | int write_object = flags & HASH_WRITE_OBJECT; |
74400e71 | 2155 | |
7672db20 | 2156 | if (!type) |
edaec3fb | 2157 | type = OBJ_BLOB; |
6c510bee LT |
2158 | |
2159 | /* | |
2160 | * Convert blobs to git internal format | |
2161 | */ | |
43df4f86 | 2162 | if ((type == OBJ_BLOB) && path) { |
f285a2d7 | 2163 | struct strbuf nbuf = STRBUF_INIT; |
58bf2a4c | 2164 | if (convert_to_git(istate, path, buf, size, &nbuf, |
8462ff43 | 2165 | get_conv_flags(flags))) { |
b315c5c0 | 2166 | buf = strbuf_detach(&nbuf, &size); |
6c510bee LT |
2167 | re_allocated = 1; |
2168 | } | |
2169 | } | |
c4ce46fc | 2170 | if (flags & HASH_FORMAT_CHECK) { |
c879daa2 NTND |
2171 | if (type == OBJ_TREE) |
2172 | check_tree(buf, size); | |
2173 | if (type == OBJ_COMMIT) | |
2174 | check_commit(buf, size); | |
2175 | if (type == OBJ_TAG) | |
2176 | check_tag(buf, size); | |
2177 | } | |
6c510bee | 2178 | |
7672db20 | 2179 | if (write_object) |
169c9c01 | 2180 | ret = write_object_file(buf, size, type_name(type), oid); |
abdc3fc8 | 2181 | else |
2dcde20e MT |
2182 | ret = hash_object_file(the_hash_algo, buf, size, |
2183 | type_name(type), oid); | |
43df4f86 | 2184 | if (re_allocated) |
6c510bee | 2185 | free(buf); |
43df4f86 DP |
2186 | return ret; |
2187 | } | |
2188 | ||
58bf2a4c NTND |
2189 | static int index_stream_convert_blob(struct index_state *istate, |
2190 | struct object_id *oid, | |
2191 | int fd, | |
2192 | const char *path, | |
2193 | unsigned flags) | |
9035d75a SP |
2194 | { |
2195 | int ret; | |
2196 | const int write_object = flags & HASH_WRITE_OBJECT; | |
2197 | struct strbuf sbuf = STRBUF_INIT; | |
2198 | ||
2199 | assert(path); | |
58bf2a4c | 2200 | assert(would_convert_to_git_filter_fd(istate, path)); |
9035d75a | 2201 | |
58bf2a4c | 2202 | convert_to_git_filter_fd(istate, path, fd, &sbuf, |
8462ff43 | 2203 | get_conv_flags(flags)); |
9035d75a SP |
2204 | |
2205 | if (write_object) | |
169c9c01 | 2206 | ret = write_object_file(sbuf.buf, sbuf.len, type_name(OBJ_BLOB), |
a09c985e | 2207 | oid); |
9035d75a | 2208 | else |
2dcde20e MT |
2209 | ret = hash_object_file(the_hash_algo, sbuf.buf, sbuf.len, |
2210 | type_name(OBJ_BLOB), oid); | |
9035d75a SP |
2211 | strbuf_release(&sbuf); |
2212 | return ret; | |
2213 | } | |
2214 | ||
58bf2a4c NTND |
2215 | static int index_pipe(struct index_state *istate, struct object_id *oid, |
2216 | int fd, enum object_type type, | |
7b41e1e1 JH |
2217 | const char *path, unsigned flags) |
2218 | { | |
2219 | struct strbuf sbuf = STRBUF_INIT; | |
2220 | int ret; | |
2221 | ||
2222 | if (strbuf_read(&sbuf, fd, 4096) >= 0) | |
58bf2a4c | 2223 | ret = index_mem(istate, oid, sbuf.buf, sbuf.len, type, path, flags); |
7b41e1e1 JH |
2224 | else |
2225 | ret = -1; | |
2226 | strbuf_release(&sbuf); | |
2227 | return ret; | |
2228 | } | |
2229 | ||
ea68b0ce DP |
2230 | #define SMALL_FILE_SIZE (32*1024) |
2231 | ||
58bf2a4c NTND |
2232 | static int index_core(struct index_state *istate, |
2233 | struct object_id *oid, int fd, size_t size, | |
7b41e1e1 JH |
2234 | enum object_type type, const char *path, |
2235 | unsigned flags) | |
43df4f86 DP |
2236 | { |
2237 | int ret; | |
43df4f86 | 2238 | |
7b41e1e1 | 2239 | if (!size) { |
58bf2a4c | 2240 | ret = index_mem(istate, oid, "", size, type, path, flags); |
ea68b0ce DP |
2241 | } else if (size <= SMALL_FILE_SIZE) { |
2242 | char *buf = xmalloc(size); | |
90dca671 JK |
2243 | ssize_t read_result = read_in_full(fd, buf, size); |
2244 | if (read_result < 0) | |
259328b7 | 2245 | ret = error_errno(_("read error while indexing %s"), |
90dca671 JK |
2246 | path ? path : "<unknown>"); |
2247 | else if (read_result != size) | |
259328b7 | 2248 | ret = error(_("short read while indexing %s"), |
90dca671 | 2249 | path ? path : "<unknown>"); |
ea68b0ce | 2250 | else |
58bf2a4c | 2251 | ret = index_mem(istate, oid, buf, size, type, path, flags); |
ea68b0ce | 2252 | free(buf); |
08bda208 | 2253 | } else { |
43df4f86 | 2254 | void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0); |
58bf2a4c | 2255 | ret = index_mem(istate, oid, buf, size, type, path, flags); |
aac17941 | 2256 | munmap(buf, size); |
08bda208 | 2257 | } |
7b41e1e1 JH |
2258 | return ret; |
2259 | } | |
2260 | ||
4dd1fbc7 | 2261 | /* |
568508e7 JH |
2262 | * This creates one packfile per large blob unless bulk-checkin |
2263 | * machinery is "plugged". | |
4dd1fbc7 JH |
2264 | * |
2265 | * This also bypasses the usual "convert-to-git" dance, and that is on | |
2266 | * purpose. We could write a streaming version of the converting | |
2267 | * functions and insert that before feeding the data to fast-import | |
4f22b101 JK |
2268 | * (or equivalent in-core API described above). However, that is |
2269 | * somewhat complicated, as we do not know the size of the filter | |
2270 | * result, which we need to know beforehand when writing a git object. | |
2271 | * Since the primary motivation for trying to stream from the working | |
2272 | * tree file and to avoid mmaping it in core is to deal with large | |
2273 | * binary blobs, they generally do not want to get any conversion, and | |
2274 | * callers should avoid this code path when filters are requested. | |
4dd1fbc7 | 2275 | */ |
7d5e1dc3 | 2276 | static int index_stream(struct object_id *oid, int fd, size_t size, |
4dd1fbc7 JH |
2277 | enum object_type type, const char *path, |
2278 | unsigned flags) | |
2279 | { | |
68ee6dfc | 2280 | return index_bulk_checkin(oid, fd, size, type, path, flags); |
4dd1fbc7 JH |
2281 | } |
2282 | ||
58bf2a4c NTND |
2283 | int index_fd(struct index_state *istate, struct object_id *oid, |
2284 | int fd, struct stat *st, | |
7b41e1e1 JH |
2285 | enum object_type type, const char *path, unsigned flags) |
2286 | { | |
2287 | int ret; | |
7b41e1e1 | 2288 | |
9079ab7c SP |
2289 | /* |
2290 | * Call xsize_t() only when needed to avoid potentially unnecessary | |
2291 | * die() for large files. | |
2292 | */ | |
58bf2a4c NTND |
2293 | if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(istate, path)) |
2294 | ret = index_stream_convert_blob(istate, oid, fd, path, flags); | |
9035d75a | 2295 | else if (!S_ISREG(st->st_mode)) |
58bf2a4c | 2296 | ret = index_pipe(istate, oid, fd, type, path, flags); |
9079ab7c | 2297 | else if (st->st_size <= big_file_threshold || type != OBJ_BLOB || |
58bf2a4c NTND |
2298 | (path && would_convert_to_git(istate, path))) |
2299 | ret = index_core(istate, oid, fd, xsize_t(st->st_size), | |
2300 | type, path, flags); | |
4dd1fbc7 | 2301 | else |
7d5e1dc3 | 2302 | ret = index_stream(oid, fd, xsize_t(st->st_size), type, path, |
9079ab7c | 2303 | flags); |
43df4f86 | 2304 | close(fd); |
aac17941 | 2305 | return ret; |
74400e71 | 2306 | } |
ec1fcc16 | 2307 | |
58bf2a4c NTND |
2308 | int index_path(struct index_state *istate, struct object_id *oid, |
2309 | const char *path, struct stat *st, unsigned flags) | |
ec1fcc16 JH |
2310 | { |
2311 | int fd; | |
b760d3aa | 2312 | struct strbuf sb = STRBUF_INIT; |
ea8e0297 | 2313 | int rc = 0; |
ec1fcc16 JH |
2314 | |
2315 | switch (st->st_mode & S_IFMT) { | |
2316 | case S_IFREG: | |
2317 | fd = open(path, O_RDONLY); | |
2318 | if (fd < 0) | |
7616c6ca | 2319 | return error_errno("open(\"%s\")", path); |
58bf2a4c | 2320 | if (index_fd(istate, oid, fd, st, OBJ_BLOB, path, flags) < 0) |
259328b7 | 2321 | return error(_("%s: failed to insert into database"), |
ec1fcc16 JH |
2322 | path); |
2323 | break; | |
2324 | case S_IFLNK: | |
7616c6ca NTND |
2325 | if (strbuf_readlink(&sb, path, st->st_size)) |
2326 | return error_errno("readlink(\"%s\")", path); | |
c4ce46fc | 2327 | if (!(flags & HASH_WRITE_OBJECT)) |
2dcde20e MT |
2328 | hash_object_file(the_hash_algo, sb.buf, sb.len, |
2329 | blob_type, oid); | |
a09c985e | 2330 | else if (write_object_file(sb.buf, sb.len, blob_type, oid)) |
259328b7 | 2331 | rc = error(_("%s: failed to insert into database"), path); |
b760d3aa | 2332 | strbuf_release(&sb); |
ec1fcc16 | 2333 | break; |
f35a6d3b | 2334 | case S_IFDIR: |
a98e6101 | 2335 | return resolve_gitlink_ref(path, "HEAD", oid); |
ec1fcc16 | 2336 | default: |
259328b7 | 2337 | return error(_("%s: unsupported file type"), path); |
ec1fcc16 | 2338 | } |
ea8e0297 | 2339 | return rc; |
ec1fcc16 | 2340 | } |
a69e5429 JH |
2341 | |
2342 | int read_pack_header(int fd, struct pack_header *header) | |
2343 | { | |
f48ecd38 | 2344 | if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header)) |
c697ad14 HO |
2345 | /* "eof before pack header was fully read" */ |
2346 | return PH_ERROR_EOF; | |
2347 | ||
a69e5429 JH |
2348 | if (header->hdr_signature != htonl(PACK_SIGNATURE)) |
2349 | /* "protocol error (pack signature mismatch detected)" */ | |
2350 | return PH_ERROR_PACK_SIGNATURE; | |
2351 | if (!pack_version_ok(header->hdr_version)) | |
2352 | /* "protocol error (pack version unsupported)" */ | |
2353 | return PH_ERROR_PROTOCOL; | |
2354 | return 0; | |
2355 | } | |
40d52ff7 | 2356 | |
e816caa0 | 2357 | void assert_oid_type(const struct object_id *oid, enum object_type expect) |
40d52ff7 | 2358 | { |
0df8e965 | 2359 | enum object_type type = oid_object_info(the_repository, oid, NULL); |
40d52ff7 | 2360 | if (type < 0) |
259328b7 | 2361 | die(_("%s is not a valid object"), oid_to_hex(oid)); |
40d52ff7 | 2362 | if (type != expect) |
259328b7 | 2363 | die(_("%s is not a valid '%s' object"), oid_to_hex(oid), |
debca9d2 | 2364 | type_name(expect)); |
40d52ff7 | 2365 | } |
27e1e22d | 2366 | |
70c49050 | 2367 | int for_each_file_in_obj_subdir(unsigned int subdir_nr, |
cc817ca3 RS |
2368 | struct strbuf *path, |
2369 | each_loose_object_fn obj_cb, | |
2370 | each_loose_cruft_fn cruft_cb, | |
2371 | each_loose_subdir_fn subdir_cb, | |
2372 | void *data) | |
27e1e22d | 2373 | { |
0375f472 RS |
2374 | size_t origlen, baselen; |
2375 | DIR *dir; | |
27e1e22d JK |
2376 | struct dirent *de; |
2377 | int r = 0; | |
62a24c89 | 2378 | struct object_id oid; |
27e1e22d | 2379 | |
70c49050 RS |
2380 | if (subdir_nr > 0xff) |
2381 | BUG("invalid loose object subdirectory: %x", subdir_nr); | |
2382 | ||
0375f472 RS |
2383 | origlen = path->len; |
2384 | strbuf_complete(path, '/'); | |
2385 | strbuf_addf(path, "%02x", subdir_nr); | |
0375f472 RS |
2386 | |
2387 | dir = opendir(path->buf); | |
27e1e22d | 2388 | if (!dir) { |
0375f472 | 2389 | if (errno != ENOENT) |
259328b7 | 2390 | r = error_errno(_("unable to open %s"), path->buf); |
0375f472 RS |
2391 | strbuf_setlen(path, origlen); |
2392 | return r; | |
27e1e22d JK |
2393 | } |
2394 | ||
62a24c89 | 2395 | oid.hash[0] = subdir_nr; |
163ee5e6 DS |
2396 | strbuf_addch(path, '/'); |
2397 | baselen = path->len; | |
62a24c89 | 2398 | |
b548f0f1 | 2399 | while ((de = readdir_skip_dot_and_dotdot(dir))) { |
163ee5e6 | 2400 | size_t namelen; |
27e1e22d | 2401 | |
163ee5e6 | 2402 | namelen = strlen(de->d_name); |
27e1e22d | 2403 | strbuf_setlen(path, baselen); |
163ee5e6 | 2404 | strbuf_add(path, de->d_name, namelen); |
94b5e093 | 2405 | if (namelen == the_hash_algo->hexsz - 2 && |
62a24c89 | 2406 | !hex_to_bytes(oid.hash + 1, de->d_name, |
94b5e093 | 2407 | the_hash_algo->rawsz - 1)) { |
5a6dce70 | 2408 | oid_set_algo(&oid, the_hash_algo); |
62a24c89 RS |
2409 | if (obj_cb) { |
2410 | r = obj_cb(&oid, path->buf, data); | |
2411 | if (r) | |
2412 | break; | |
27e1e22d | 2413 | } |
62a24c89 | 2414 | continue; |
27e1e22d JK |
2415 | } |
2416 | ||
2417 | if (cruft_cb) { | |
2418 | r = cruft_cb(de->d_name, path->buf, data); | |
2419 | if (r) | |
2420 | break; | |
2421 | } | |
2422 | } | |
094c7e63 | 2423 | closedir(dir); |
27e1e22d | 2424 | |
163ee5e6 | 2425 | strbuf_setlen(path, baselen - 1); |
27e1e22d JK |
2426 | if (!r && subdir_cb) |
2427 | r = subdir_cb(subdir_nr, path->buf, data); | |
2428 | ||
0375f472 RS |
2429 | strbuf_setlen(path, origlen); |
2430 | ||
27e1e22d JK |
2431 | return r; |
2432 | } | |
2433 | ||
e6f875e0 | 2434 | int for_each_loose_file_in_objdir_buf(struct strbuf *path, |
27e1e22d JK |
2435 | each_loose_object_fn obj_cb, |
2436 | each_loose_cruft_fn cruft_cb, | |
2437 | each_loose_subdir_fn subdir_cb, | |
2438 | void *data) | |
2439 | { | |
27e1e22d JK |
2440 | int r = 0; |
2441 | int i; | |
2442 | ||
27e1e22d | 2443 | for (i = 0; i < 256; i++) { |
e6f875e0 | 2444 | r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb, |
27e1e22d | 2445 | subdir_cb, data); |
27e1e22d JK |
2446 | if (r) |
2447 | break; | |
2448 | } | |
2449 | ||
e6f875e0 JK |
2450 | return r; |
2451 | } | |
2452 | ||
2453 | int for_each_loose_file_in_objdir(const char *path, | |
2454 | each_loose_object_fn obj_cb, | |
2455 | each_loose_cruft_fn cruft_cb, | |
2456 | each_loose_subdir_fn subdir_cb, | |
2457 | void *data) | |
2458 | { | |
2459 | struct strbuf buf = STRBUF_INIT; | |
2460 | int r; | |
2461 | ||
2462 | strbuf_addstr(&buf, path); | |
2463 | r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb, | |
2464 | subdir_cb, data); | |
27e1e22d | 2465 | strbuf_release(&buf); |
e6f875e0 | 2466 | |
27e1e22d JK |
2467 | return r; |
2468 | } | |
660c889e | 2469 | |
a7ff6f5a JK |
2470 | int for_each_loose_object(each_loose_object_fn cb, void *data, |
2471 | enum for_each_object_flags flags) | |
660c889e | 2472 | { |
f0eaf638 | 2473 | struct object_directory *odb; |
b0a42642 | 2474 | |
f0eaf638 JK |
2475 | prepare_alt_odb(the_repository); |
2476 | for (odb = the_repository->objects->odb; odb; odb = odb->next) { | |
2477 | int r = for_each_loose_file_in_objdir(odb->path, cb, NULL, | |
2478 | NULL, data); | |
2479 | if (r) | |
2480 | return r; | |
660c889e | 2481 | |
f0eaf638 JK |
2482 | if (flags & FOR_EACH_OBJECT_LOCAL_ONLY) |
2483 | break; | |
2484 | } | |
1385bb7b | 2485 | |
f0eaf638 | 2486 | return 0; |
660c889e JK |
2487 | } |
2488 | ||
3a2e0824 JK |
2489 | static int append_loose_object(const struct object_id *oid, const char *path, |
2490 | void *data) | |
660c889e | 2491 | { |
92d8ed8a | 2492 | oidtree_insert(data, oid); |
3a2e0824 JK |
2493 | return 0; |
2494 | } | |
660c889e | 2495 | |
92d8ed8a | 2496 | struct oidtree *odb_loose_cache(struct object_directory *odb, |
0000d654 RS |
2497 | const struct object_id *oid) |
2498 | { | |
2499 | int subdir_nr = oid->hash[0]; | |
3a2e0824 | 2500 | struct strbuf buf = STRBUF_INIT; |
33f379ee EW |
2501 | size_t word_bits = bitsizeof(odb->loose_objects_subdir_seen[0]); |
2502 | size_t word_index = subdir_nr / word_bits; | |
581a3bb1 | 2503 | size_t mask = 1u << (subdir_nr % word_bits); |
33f379ee | 2504 | uint32_t *bitmap; |
660c889e | 2505 | |
3a2e0824 | 2506 | if (subdir_nr < 0 || |
33f379ee | 2507 | subdir_nr >= bitsizeof(odb->loose_objects_subdir_seen)) |
3a2e0824 | 2508 | BUG("subdir_nr out of range"); |
1385bb7b | 2509 | |
33f379ee EW |
2510 | bitmap = &odb->loose_objects_subdir_seen[word_index]; |
2511 | if (*bitmap & mask) | |
92d8ed8a EW |
2512 | return odb->loose_objects_cache; |
2513 | if (!odb->loose_objects_cache) { | |
2514 | ALLOC_ARRAY(odb->loose_objects_cache, 1); | |
2515 | oidtree_init(odb->loose_objects_cache); | |
2516 | } | |
3a2e0824 JK |
2517 | strbuf_addstr(&buf, odb->path); |
2518 | for_each_file_in_obj_subdir(subdir_nr, &buf, | |
2519 | append_loose_object, | |
2520 | NULL, NULL, | |
92d8ed8a | 2521 | odb->loose_objects_cache); |
33f379ee | 2522 | *bitmap |= mask; |
7317aa71 | 2523 | strbuf_release(&buf); |
92d8ed8a | 2524 | return odb->loose_objects_cache; |
660c889e JK |
2525 | } |
2526 | ||
d4e19e51 RS |
2527 | void odb_clear_loose_cache(struct object_directory *odb) |
2528 | { | |
92d8ed8a EW |
2529 | oidtree_clear(odb->loose_objects_cache); |
2530 | FREE_AND_NULL(odb->loose_objects_cache); | |
d4e19e51 RS |
2531 | memset(&odb->loose_objects_subdir_seen, 0, |
2532 | sizeof(odb->loose_objects_subdir_seen)); | |
2533 | } | |
2534 | ||
00a7760e JK |
2535 | static int check_stream_oid(git_zstream *stream, |
2536 | const char *hdr, | |
2537 | unsigned long size, | |
2538 | const char *path, | |
2539 | const struct object_id *expected_oid) | |
f6371f92 | 2540 | { |
18e2588e | 2541 | git_hash_ctx c; |
00a7760e | 2542 | struct object_id real_oid; |
f6371f92 JK |
2543 | unsigned char buf[4096]; |
2544 | unsigned long total_read; | |
2545 | int status = Z_OK; | |
2546 | ||
18e2588e | 2547 | the_hash_algo->init_fn(&c); |
2548 | the_hash_algo->update_fn(&c, hdr, stream->total_out); | |
f6371f92 JK |
2549 | |
2550 | /* | |
2551 | * We already read some bytes into hdr, but the ones up to the NUL | |
2552 | * do not count against the object's content size. | |
2553 | */ | |
2554 | total_read = stream->total_out - strlen(hdr) - 1; | |
2555 | ||
2556 | /* | |
2557 | * This size comparison must be "<=" to read the final zlib packets; | |
00a7760e | 2558 | * see the comment in unpack_loose_rest for details. |
f6371f92 JK |
2559 | */ |
2560 | while (total_read <= size && | |
18ad13e5 JH |
2561 | (status == Z_OK || |
2562 | (status == Z_BUF_ERROR && !stream->avail_out))) { | |
f6371f92 JK |
2563 | stream->next_out = buf; |
2564 | stream->avail_out = sizeof(buf); | |
2565 | if (size - total_read < stream->avail_out) | |
2566 | stream->avail_out = size - total_read; | |
2567 | status = git_inflate(stream, Z_FINISH); | |
18e2588e | 2568 | the_hash_algo->update_fn(&c, buf, stream->next_out - buf); |
f6371f92 JK |
2569 | total_read += stream->next_out - buf; |
2570 | } | |
2571 | git_inflate_end(stream); | |
2572 | ||
2573 | if (status != Z_STREAM_END) { | |
00a7760e | 2574 | error(_("corrupt loose object '%s'"), oid_to_hex(expected_oid)); |
f6371f92 JK |
2575 | return -1; |
2576 | } | |
cce044df | 2577 | if (stream->avail_in) { |
259328b7 | 2578 | error(_("garbage at end of loose object '%s'"), |
00a7760e | 2579 | oid_to_hex(expected_oid)); |
cce044df JK |
2580 | return -1; |
2581 | } | |
f6371f92 | 2582 | |
5951bf46 | 2583 | the_hash_algo->final_oid_fn(&real_oid, &c); |
00a7760e | 2584 | if (!oideq(expected_oid, &real_oid)) { |
01f8d594 | 2585 | error(_("hash mismatch for %s (expected %s)"), path, |
00a7760e | 2586 | oid_to_hex(expected_oid)); |
f6371f92 JK |
2587 | return -1; |
2588 | } | |
2589 | ||
2590 | return 0; | |
2591 | } | |
2592 | ||
2593 | int read_loose_object(const char *path, | |
d61d87bd | 2594 | const struct object_id *expected_oid, |
f6371f92 JK |
2595 | enum object_type *type, |
2596 | unsigned long *size, | |
2597 | void **contents) | |
2598 | { | |
2599 | int ret = -1; | |
f6371f92 JK |
2600 | void *map = NULL; |
2601 | unsigned long mapsize; | |
2602 | git_zstream stream; | |
1af64f73 | 2603 | char hdr[MAX_HEADER_LEN]; |
f6371f92 JK |
2604 | |
2605 | *contents = NULL; | |
2606 | ||
514c5fdd | 2607 | map = map_loose_object_1(the_repository, path, NULL, &mapsize); |
f6371f92 | 2608 | if (!map) { |
259328b7 | 2609 | error_errno(_("unable to mmap %s"), path); |
f6371f92 JK |
2610 | goto out; |
2611 | } | |
2612 | ||
00a7760e | 2613 | if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) { |
259328b7 | 2614 | error(_("unable to unpack header of %s"), path); |
f6371f92 JK |
2615 | goto out; |
2616 | } | |
2617 | ||
00a7760e | 2618 | *type = parse_loose_header(hdr, size); |
f6371f92 | 2619 | if (*type < 0) { |
259328b7 | 2620 | error(_("unable to parse header of %s"), path); |
f6371f92 JK |
2621 | git_inflate_end(&stream); |
2622 | goto out; | |
2623 | } | |
2624 | ||
7ac4f3a0 | 2625 | if (*type == OBJ_BLOB && *size > big_file_threshold) { |
00a7760e | 2626 | if (check_stream_oid(&stream, hdr, *size, path, expected_oid) < 0) |
f6371f92 JK |
2627 | goto out; |
2628 | } else { | |
00a7760e | 2629 | *contents = unpack_loose_rest(&stream, hdr, *size, expected_oid); |
f6371f92 | 2630 | if (!*contents) { |
259328b7 | 2631 | error(_("unable to unpack contents of %s"), path); |
f6371f92 JK |
2632 | git_inflate_end(&stream); |
2633 | goto out; | |
2634 | } | |
b98d1885 MT |
2635 | if (check_object_signature(the_repository, expected_oid, |
2636 | *contents, *size, | |
2637 | type_name(*type))) { | |
01f8d594 | 2638 | error(_("hash mismatch for %s (expected %s)"), path, |
d61d87bd | 2639 | oid_to_hex(expected_oid)); |
f6371f92 JK |
2640 | free(*contents); |
2641 | goto out; | |
2642 | } | |
2643 | } | |
2644 | ||
2645 | ret = 0; /* everything checks out */ | |
2646 | ||
2647 | out: | |
2648 | if (map) | |
2649 | munmap(map, mapsize); | |
f6371f92 JK |
2650 | return ret; |
2651 | } |