]>
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 | ||
f50e766b | 168 | const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = { |
169 | { | |
0cb9872e ÆAB |
170 | .name = NULL, |
171 | .format_id = 0x00000000, | |
172 | .rawsz = 0, | |
173 | .hexsz = 0, | |
174 | .blksz = 0, | |
175 | .init_fn = git_hash_unknown_init, | |
176 | .clone_fn = git_hash_unknown_clone, | |
177 | .update_fn = git_hash_unknown_update, | |
178 | .final_fn = git_hash_unknown_final, | |
179 | .final_oid_fn = git_hash_unknown_final_oid, | |
180 | .empty_tree = NULL, | |
181 | .empty_blob = NULL, | |
182 | .null_oid = NULL, | |
f50e766b | 183 | }, |
184 | { | |
0cb9872e ÆAB |
185 | .name = "sha1", |
186 | .format_id = GIT_SHA1_FORMAT_ID, | |
187 | .rawsz = GIT_SHA1_RAWSZ, | |
188 | .hexsz = GIT_SHA1_HEXSZ, | |
189 | .blksz = GIT_SHA1_BLKSZ, | |
190 | .init_fn = git_hash_sha1_init, | |
191 | .clone_fn = git_hash_sha1_clone, | |
192 | .update_fn = git_hash_sha1_update, | |
193 | .final_fn = git_hash_sha1_final, | |
194 | .final_oid_fn = git_hash_sha1_final_oid, | |
195 | .empty_tree = &empty_tree_oid, | |
196 | .empty_blob = &empty_blob_oid, | |
197 | .null_oid = &null_oid_sha1, | |
f50e766b | 198 | }, |
13eeedb5 | 199 | { |
0cb9872e ÆAB |
200 | .name = "sha256", |
201 | .format_id = GIT_SHA256_FORMAT_ID, | |
202 | .rawsz = GIT_SHA256_RAWSZ, | |
203 | .hexsz = GIT_SHA256_HEXSZ, | |
204 | .blksz = GIT_SHA256_BLKSZ, | |
205 | .init_fn = git_hash_sha256_init, | |
206 | .clone_fn = git_hash_sha256_clone, | |
207 | .update_fn = git_hash_sha256_update, | |
208 | .final_fn = git_hash_sha256_final, | |
209 | .final_oid_fn = git_hash_sha256_final_oid, | |
210 | .empty_tree = &empty_tree_oid_sha256, | |
211 | .empty_blob = &empty_blob_oid_sha256, | |
212 | .null_oid = &null_oid_sha256, | |
13eeedb5 | 213 | } |
f50e766b | 214 | }; |
215 | ||
14228447 | 216 | const struct object_id *null_oid(void) |
217 | { | |
218 | return the_hash_algo->null_oid; | |
219 | } | |
220 | ||
d8a92ced | 221 | const char *empty_tree_oid_hex(void) |
222 | { | |
223 | static char buf[GIT_MAX_HEXSZ + 1]; | |
224 | return oid_to_hex_r(buf, the_hash_algo->empty_tree); | |
225 | } | |
226 | ||
227 | const char *empty_blob_oid_hex(void) | |
228 | { | |
229 | static char buf[GIT_MAX_HEXSZ + 1]; | |
230 | return oid_to_hex_r(buf, the_hash_algo->empty_blob); | |
231 | } | |
232 | ||
2f90b9d9 | 233 | int hash_algo_by_name(const char *name) |
234 | { | |
235 | int i; | |
236 | if (!name) | |
237 | return GIT_HASH_UNKNOWN; | |
238 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
239 | if (!strcmp(name, hash_algos[i].name)) | |
240 | return i; | |
241 | return GIT_HASH_UNKNOWN; | |
242 | } | |
243 | ||
244 | int hash_algo_by_id(uint32_t format_id) | |
245 | { | |
246 | int i; | |
247 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
248 | if (format_id == hash_algos[i].format_id) | |
249 | return i; | |
250 | return GIT_HASH_UNKNOWN; | |
251 | } | |
252 | ||
95399788 | 253 | int hash_algo_by_length(int len) |
254 | { | |
255 | int i; | |
256 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
257 | if (len == hash_algos[i].rawsz) | |
258 | return i; | |
259 | return GIT_HASH_UNKNOWN; | |
260 | } | |
2f90b9d9 | 261 | |
c597ba80 NTND |
262 | /* |
263 | * This is meant to hold a *small* number of objects that you would | |
cb1c8d1d | 264 | * want read_object_file() to be able to return, but yet you do not want |
c597ba80 NTND |
265 | * to write them into the object store (e.g. a browse-only |
266 | * application). | |
267 | */ | |
268 | static struct cached_object { | |
62ba93ea | 269 | struct object_id oid; |
c597ba80 NTND |
270 | enum object_type type; |
271 | void *buf; | |
272 | unsigned long size; | |
273 | } *cached_objects; | |
274 | static int cached_object_nr, cached_object_alloc; | |
275 | ||
276 | static struct cached_object empty_tree = { | |
50103649 ÆAB |
277 | .oid = { |
278 | .hash = EMPTY_TREE_SHA1_BIN_LITERAL, | |
279 | }, | |
280 | .type = OBJ_TREE, | |
281 | .buf = "", | |
c597ba80 NTND |
282 | }; |
283 | ||
62ba93ea | 284 | static struct cached_object *find_cached_object(const struct object_id *oid) |
c597ba80 NTND |
285 | { |
286 | int i; | |
287 | struct cached_object *co = cached_objects; | |
288 | ||
289 | for (i = 0; i < cached_object_nr; i++, co++) { | |
4a7e27e9 | 290 | if (oideq(&co->oid, oid)) |
c597ba80 NTND |
291 | return co; |
292 | } | |
4a7e27e9 | 293 | if (oideq(oid, the_hash_algo->empty_tree)) |
c597ba80 NTND |
294 | return &empty_tree; |
295 | return NULL; | |
296 | } | |
297 | ||
9472935d | 298 | |
8462ff43 | 299 | static int get_conv_flags(unsigned flags) |
9472935d TB |
300 | { |
301 | if (flags & HASH_RENORMALIZE) | |
8462ff43 | 302 | return CONV_EOL_RENORMALIZE; |
9472935d | 303 | else if (flags & HASH_WRITE_OBJECT) |
107642fe | 304 | return global_conv_flags_eol | CONV_WRITE_OBJECT; |
9472935d | 305 | else |
8462ff43 | 306 | return 0; |
9472935d TB |
307 | } |
308 | ||
309 | ||
90a6464b JH |
310 | int mkdir_in_gitdir(const char *path) |
311 | { | |
312 | if (mkdir(path, 0777)) { | |
313 | int saved_errno = errno; | |
314 | struct stat st; | |
315 | struct strbuf sb = STRBUF_INIT; | |
316 | ||
317 | if (errno != EEXIST) | |
318 | return -1; | |
319 | /* | |
320 | * Are we looking at a path in a symlinked worktree | |
321 | * whose original repository does not yet have it? | |
322 | * e.g. .git/rr-cache pointing at its original | |
323 | * repository in which the user hasn't performed any | |
324 | * conflict resolution yet? | |
325 | */ | |
326 | if (lstat(path, &st) || !S_ISLNK(st.st_mode) || | |
327 | strbuf_readlink(&sb, path, st.st_size) || | |
328 | !is_absolute_path(sb.buf) || | |
329 | mkdir(sb.buf, 0777)) { | |
330 | strbuf_release(&sb); | |
331 | errno = saved_errno; | |
332 | return -1; | |
333 | } | |
334 | strbuf_release(&sb); | |
335 | } | |
336 | return adjust_shared_perm(path); | |
337 | } | |
338 | ||
eb3c027e | 339 | static enum scld_error safe_create_leading_directories_1(char *path, int share) |
b2cb9425 | 340 | { |
26c8ae2a | 341 | char *next_component = path + offset_1st_component(path); |
0be0521b | 342 | enum scld_error ret = SCLD_OK; |
67d42212 | 343 | |
0be0521b | 344 | while (ret == SCLD_OK && next_component) { |
f0502332 | 345 | struct stat st; |
0f527403 | 346 | char *slash = next_component, slash_character; |
f0502332 | 347 | |
0f527403 MH |
348 | while (*slash && !is_dir_sep(*slash)) |
349 | slash++; | |
350 | ||
351 | if (!*slash) | |
b2cb9425 | 352 | break; |
bf10cf70 | 353 | |
26c8ae2a | 354 | next_component = slash + 1; |
0f527403 | 355 | while (is_dir_sep(*next_component)) |
bf10cf70 | 356 | next_component++; |
26c8ae2a | 357 | if (!*next_component) |
5f0bdf50 | 358 | break; |
831651fd | 359 | |
0f527403 | 360 | slash_character = *slash; |
831651fd | 361 | *slash = '\0'; |
67d42212 JR |
362 | if (!stat(path, &st)) { |
363 | /* path exists */ | |
204a047f MH |
364 | if (!S_ISDIR(st.st_mode)) { |
365 | errno = ENOTDIR; | |
0be0521b | 366 | ret = SCLD_EXISTS; |
204a047f | 367 | } |
53a39721 | 368 | } else if (mkdir(path, 0777)) { |
928734d9 | 369 | if (errno == EEXIST && |
9e6f885d | 370 | !stat(path, &st) && S_ISDIR(st.st_mode)) |
928734d9 | 371 | ; /* somebody created it since we checked */ |
18d37e86 MH |
372 | else if (errno == ENOENT) |
373 | /* | |
374 | * Either mkdir() failed because | |
375 | * somebody just pruned the containing | |
376 | * directory, or stat() failed because | |
377 | * the file that was in our way was | |
378 | * just removed. Either way, inform | |
379 | * the caller that it might be worth | |
380 | * trying again: | |
381 | */ | |
382 | ret = SCLD_VANISHED; | |
9e6f885d | 383 | else |
0be0521b | 384 | ret = SCLD_FAILED; |
eb3c027e | 385 | } else if (share && adjust_shared_perm(path)) { |
0be0521b | 386 | ret = SCLD_PERMS; |
457f06d6 | 387 | } |
0f527403 | 388 | *slash = slash_character; |
b2cb9425 | 389 | } |
9e6f885d | 390 | return ret; |
b2cb9425 | 391 | } |
723c31fe | 392 | |
eb3c027e MT |
393 | enum scld_error safe_create_leading_directories(char *path) |
394 | { | |
395 | return safe_create_leading_directories_1(path, 1); | |
396 | } | |
397 | ||
398 | enum scld_error safe_create_leading_directories_no_share(char *path) | |
399 | { | |
400 | return safe_create_leading_directories_1(path, 0); | |
401 | } | |
402 | ||
0be0521b | 403 | enum scld_error safe_create_leading_directories_const(const char *path) |
8e21d63b | 404 | { |
02944307 | 405 | int save_errno; |
8e21d63b JK |
406 | /* path points to cache entries, so xstrdup before messing with it */ |
407 | char *buf = xstrdup(path); | |
0be0521b | 408 | enum scld_error result = safe_create_leading_directories(buf); |
02944307 MH |
409 | |
410 | save_errno = errno; | |
8e21d63b | 411 | free(buf); |
02944307 | 412 | errno = save_errno; |
8e21d63b JK |
413 | return result; |
414 | } | |
415 | ||
514c5fdd | 416 | static void fill_loose_path(struct strbuf *buf, const struct object_id *oid) |
ace1534d JH |
417 | { |
418 | int i; | |
94b5e093 | 419 | for (i = 0; i < the_hash_algo->rawsz; i++) { |
ace1534d | 420 | static char hex[] = "0123456789abcdef"; |
514c5fdd | 421 | unsigned int val = oid->hash[i]; |
f7b7774f JK |
422 | strbuf_addch(buf, hex[val >> 4]); |
423 | strbuf_addch(buf, hex[val & 0xf]); | |
afbba2f0 | 424 | if (!i) |
f7b7774f | 425 | strbuf_addch(buf, '/'); |
ace1534d JH |
426 | } |
427 | } | |
428 | ||
f0eaf638 JK |
429 | static const char *odb_loose_path(struct object_directory *odb, |
430 | struct strbuf *buf, | |
514c5fdd | 431 | const struct object_id *oid) |
0fcfd160 | 432 | { |
b69fb867 | 433 | strbuf_reset(buf); |
f0eaf638 | 434 | strbuf_addstr(buf, odb->path); |
34498471 | 435 | strbuf_addch(buf, '/'); |
514c5fdd | 436 | fill_loose_path(buf, oid); |
f3f043a1 | 437 | return buf->buf; |
0fcfd160 LT |
438 | } |
439 | ||
f3f043a1 | 440 | const char *loose_object_path(struct repository *r, struct strbuf *buf, |
514c5fdd | 441 | const struct object_id *oid) |
29ec6af2 | 442 | { |
514c5fdd | 443 | return odb_loose_path(r->objects->odb, buf, oid); |
0fcfd160 LT |
444 | } |
445 | ||
4ea82473 JK |
446 | /* |
447 | * Return non-zero iff the path is usable as an alternate object database. | |
448 | */ | |
13313fc3 SB |
449 | static int alt_odb_usable(struct raw_object_store *o, |
450 | struct strbuf *path, | |
cf2dc1c2 | 451 | const char *normalized_objdir, khiter_t *pos) |
4ea82473 | 452 | { |
cf2dc1c2 | 453 | int r; |
4ea82473 JK |
454 | |
455 | /* Detect cases where alternate disappeared */ | |
456 | if (!is_directory(path->buf)) { | |
259328b7 NTND |
457 | error(_("object directory %s does not exist; " |
458 | "check .git/objects/info/alternates"), | |
4ea82473 JK |
459 | path->buf); |
460 | return 0; | |
461 | } | |
462 | ||
463 | /* | |
464 | * Prevent the common mistake of listing the same | |
465 | * thing twice, or object directory itself. | |
466 | */ | |
cf2dc1c2 EW |
467 | if (!o->odb_by_path) { |
468 | khiter_t p; | |
469 | ||
470 | o->odb_by_path = kh_init_odb_path_map(); | |
471 | assert(!o->odb->next); | |
472 | p = kh_put_odb_path_map(o->odb_by_path, o->odb->path, &r); | |
473 | assert(r == 1); /* never used */ | |
474 | kh_value(o->odb_by_path, p) = o->odb; | |
4ea82473 | 475 | } |
cf2dc1c2 | 476 | if (fspatheq(path->buf, normalized_objdir)) |
4ea82473 | 477 | return 0; |
cf2dc1c2 EW |
478 | *pos = kh_put_odb_path_map(o->odb_by_path, path->buf, &r); |
479 | /* r: 0 = exists, 1 = never used, 2 = deleted */ | |
480 | return r == 0 ? 0 : 1; | |
4ea82473 JK |
481 | } |
482 | ||
ddd5d056 JH |
483 | /* |
484 | * Prepare alternate object database registry. | |
d5a63b99 JH |
485 | * |
486 | * The variable alt_odb_list points at the list of struct | |
263db403 | 487 | * object_directory. The elements on this list come from |
d5a63b99 JH |
488 | * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT |
489 | * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates, | |
1494e038 JH |
490 | * whose contents is similar to that environment variable but can be |
491 | * LF separated. Its base points at a statically allocated buffer that | |
d5a63b99 JH |
492 | * contains "/the/directory/corresponding/to/.git/objects/...", while |
493 | * its name points just after the slash at the end of ".git/objects/" | |
e5afd444 MÅ |
494 | * in the example above, and has enough space to hold all hex characters |
495 | * of the object ID, an extra slash for the first level indirection, and | |
496 | * the terminating NUL. | |
ddd5d056 | 497 | */ |
77f012e8 SB |
498 | static void read_info_alternates(struct repository *r, |
499 | const char *relative_base, | |
500 | int depth); | |
407532f8 | 501 | static int link_alt_odb_entry(struct repository *r, const struct strbuf *entry, |
cfc62fc9 | 502 | const char *relative_base, int depth, const char *normalized_objdir) |
ace1534d | 503 | { |
263db403 | 504 | struct object_directory *ent; |
5bdf0a84 | 505 | struct strbuf pathbuf = STRBUF_INIT; |
cf2dc1c2 | 506 | khiter_t pos; |
d5a63b99 | 507 | |
407532f8 | 508 | if (!is_absolute_path(entry->buf) && relative_base) { |
4ac9006f | 509 | strbuf_realpath(&pathbuf, relative_base, 1); |
5bdf0a84 | 510 | strbuf_addch(&pathbuf, '/'); |
c2f493a4 | 511 | } |
407532f8 | 512 | strbuf_addbuf(&pathbuf, entry); |
c2f493a4 | 513 | |
37a95862 | 514 | if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) { |
259328b7 | 515 | error(_("unable to normalize alternate object path: %s"), |
670c359d JK |
516 | pathbuf.buf); |
517 | strbuf_release(&pathbuf); | |
518 | return -1; | |
519 | } | |
5bdf0a84 HW |
520 | |
521 | /* | |
522 | * The trailing slash after the directory name is given by | |
523 | * this function at the end. Remove duplicates. | |
524 | */ | |
4ea82473 JK |
525 | while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/') |
526 | strbuf_setlen(&pathbuf, pathbuf.len - 1); | |
5bdf0a84 | 527 | |
cf2dc1c2 | 528 | if (!alt_odb_usable(r->objects, &pathbuf, normalized_objdir, &pos)) { |
4ea82473 | 529 | strbuf_release(&pathbuf); |
c2f493a4 MW |
530 | return -1; |
531 | } | |
532 | ||
ca56dadb | 533 | CALLOC_ARRAY(ent, 1); |
cf2dc1c2 EW |
534 | /* pathbuf.buf is already in r->objects->odb_by_path */ |
535 | ent->path = strbuf_detach(&pathbuf, NULL); | |
c2f493a4 MW |
536 | |
537 | /* add the alternate entry */ | |
f0eaf638 JK |
538 | *r->objects->odb_tail = ent; |
539 | r->objects->odb_tail = &(ent->next); | |
c2f493a4 | 540 | ent->next = NULL; |
cf2dc1c2 EW |
541 | assert(r->objects->odb_by_path); |
542 | kh_value(r->objects->odb_by_path, pos) = ent; | |
c2f493a4 MW |
543 | |
544 | /* recursively add alternates */ | |
cf2dc1c2 | 545 | read_info_alternates(r, ent->path, depth + 1); |
c2f493a4 MW |
546 | |
547 | return 0; | |
548 | } | |
549 | ||
cf3c6352 JK |
550 | static const char *parse_alt_odb_entry(const char *string, |
551 | int sep, | |
552 | struct strbuf *out) | |
553 | { | |
554 | const char *end; | |
555 | ||
556 | strbuf_reset(out); | |
557 | ||
558 | if (*string == '#') { | |
559 | /* comment; consume up to next separator */ | |
560 | end = strchrnul(string, sep); | |
561 | } else if (*string == '"' && !unquote_c_style(out, string, &end)) { | |
562 | /* | |
563 | * quoted path; unquote_c_style has copied the | |
564 | * data for us and set "end". Broken quoting (e.g., | |
565 | * an entry that doesn't end with a quote) falls | |
566 | * back to the unquoted case below. | |
567 | */ | |
568 | } else { | |
569 | /* normal, unquoted path */ | |
570 | end = strchrnul(string, sep); | |
571 | strbuf_add(out, string, end - string); | |
572 | } | |
573 | ||
574 | if (*end) | |
575 | end++; | |
576 | return end; | |
577 | } | |
578 | ||
77f012e8 SB |
579 | static void link_alt_odb_entries(struct repository *r, const char *alt, |
580 | int sep, const char *relative_base, int depth) | |
c2f493a4 | 581 | { |
539e7506 | 582 | struct strbuf objdirbuf = STRBUF_INIT; |
cf3c6352 | 583 | struct strbuf entry = STRBUF_INIT; |
c2f493a4 | 584 | |
f28e3668 JK |
585 | if (!alt || !*alt) |
586 | return; | |
587 | ||
c2f493a4 | 588 | if (depth > 5) { |
259328b7 | 589 | error(_("%s: ignoring alternate object stores, nesting too deep"), |
c2f493a4 MW |
590 | relative_base); |
591 | return; | |
592 | } | |
593 | ||
f0eaf638 | 594 | strbuf_add_absolute_path(&objdirbuf, r->objects->odb->path); |
670c359d | 595 | if (strbuf_normalize_path(&objdirbuf) < 0) |
259328b7 | 596 | die(_("unable to normalize object directory: %s"), |
670c359d | 597 | objdirbuf.buf); |
539e7506 | 598 | |
cf3c6352 JK |
599 | while (*alt) { |
600 | alt = parse_alt_odb_entry(alt, sep, &entry); | |
601 | if (!entry.len) | |
9577e7e3 | 602 | continue; |
407532f8 | 603 | link_alt_odb_entry(r, &entry, |
cfc62fc9 | 604 | relative_base, depth, objdirbuf.buf); |
9577e7e3 | 605 | } |
cf3c6352 | 606 | strbuf_release(&entry); |
539e7506 | 607 | strbuf_release(&objdirbuf); |
d5a63b99 JH |
608 | } |
609 | ||
77f012e8 SB |
610 | static void read_info_alternates(struct repository *r, |
611 | const char *relative_base, | |
612 | int depth) | |
d5a63b99 | 613 | { |
5015f01c | 614 | char *path; |
dc732bd5 | 615 | struct strbuf buf = STRBUF_INIT; |
d5a63b99 | 616 | |
5015f01c | 617 | path = xstrfmt("%s/info/alternates", relative_base); |
dc732bd5 | 618 | if (strbuf_read_file(&buf, path, 1024) < 0) { |
f0f7bebe | 619 | warn_on_fopen_errors(path); |
dc732bd5 | 620 | free(path); |
9a217f2a | 621 | return; |
ace1534d | 622 | } |
d5a63b99 | 623 | |
77f012e8 | 624 | link_alt_odb_entries(r, buf.buf, '\n', relative_base, depth); |
dc732bd5 JK |
625 | strbuf_release(&buf); |
626 | free(path); | |
ace1534d JH |
627 | } |
628 | ||
bef70b22 DB |
629 | void add_to_alternates_file(const char *reference) |
630 | { | |
f132a127 | 631 | struct lock_file lock = LOCK_INIT; |
77b9b1d1 JK |
632 | char *alts = git_pathdup("objects/info/alternates"); |
633 | FILE *in, *out; | |
f132a127 | 634 | int found = 0; |
77b9b1d1 | 635 | |
f132a127 MÅ |
636 | hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR); |
637 | out = fdopen_lock_file(&lock, "w"); | |
77b9b1d1 | 638 | if (!out) |
259328b7 | 639 | die_errno(_("unable to fdopen alternates lockfile")); |
77b9b1d1 JK |
640 | |
641 | in = fopen(alts, "r"); | |
642 | if (in) { | |
643 | struct strbuf line = STRBUF_INIT; | |
77b9b1d1 | 644 | |
3f163962 | 645 | while (strbuf_getline(&line, in) != EOF) { |
77b9b1d1 JK |
646 | if (!strcmp(reference, line.buf)) { |
647 | found = 1; | |
648 | break; | |
649 | } | |
650 | fprintf_or_die(out, "%s\n", line.buf); | |
651 | } | |
652 | ||
653 | strbuf_release(&line); | |
654 | fclose(in); | |
77b9b1d1 JK |
655 | } |
656 | else if (errno != ENOENT) | |
259328b7 | 657 | die_errno(_("unable to read alternates file")); |
77b9b1d1 | 658 | |
f132a127 MÅ |
659 | if (found) { |
660 | rollback_lock_file(&lock); | |
661 | } else { | |
77b9b1d1 | 662 | fprintf_or_die(out, "%s\n", reference); |
f132a127 | 663 | if (commit_lock_file(&lock)) |
259328b7 | 664 | die_errno(_("unable to move new alternates file into place")); |
f0eaf638 | 665 | if (the_repository->objects->loaded_alternates) |
93d8d1e2 SB |
666 | link_alt_odb_entries(the_repository, reference, |
667 | '\n', NULL, 0); | |
77b9b1d1 JK |
668 | } |
669 | free(alts); | |
bef70b22 DB |
670 | } |
671 | ||
a5b34d21 JK |
672 | void add_to_alternates_memory(const char *reference) |
673 | { | |
674 | /* | |
675 | * Make sure alternates are initialized, or else our entry may be | |
676 | * overwritten when they are. | |
677 | */ | |
0b209034 | 678 | prepare_alt_odb(the_repository); |
a5b34d21 | 679 | |
93d8d1e2 SB |
680 | link_alt_odb_entries(the_repository, reference, |
681 | '\n', NULL, 0); | |
a5b34d21 JK |
682 | } |
683 | ||
b3cecf49 NS |
684 | struct object_directory *set_temporary_primary_odb(const char *dir, int will_destroy) |
685 | { | |
686 | struct object_directory *new_odb; | |
687 | ||
688 | /* | |
689 | * Make sure alternates are initialized, or else our entry may be | |
690 | * overwritten when they are. | |
691 | */ | |
692 | prepare_alt_odb(the_repository); | |
693 | ||
694 | /* | |
695 | * Make a new primary odb and link the old primary ODB in as an | |
696 | * alternate | |
697 | */ | |
698 | new_odb = xcalloc(1, sizeof(*new_odb)); | |
699 | new_odb->path = xstrdup(dir); | |
ecd81dfc NS |
700 | |
701 | /* | |
702 | * Disable ref updates while a temporary odb is active, since | |
703 | * the objects in the database may roll back. | |
704 | */ | |
705 | new_odb->disable_ref_updates = 1; | |
b3cecf49 NS |
706 | new_odb->will_destroy = will_destroy; |
707 | new_odb->next = the_repository->objects->odb; | |
708 | the_repository->objects->odb = new_odb; | |
709 | return new_odb->next; | |
710 | } | |
711 | ||
712 | void restore_primary_odb(struct object_directory *restore_odb, const char *old_path) | |
713 | { | |
714 | struct object_directory *cur_odb = the_repository->objects->odb; | |
715 | ||
716 | if (strcmp(old_path, cur_odb->path)) | |
717 | BUG("expected %s as primary object store; found %s", | |
718 | old_path, cur_odb->path); | |
719 | ||
720 | if (cur_odb->next != restore_odb) | |
721 | BUG("we expect the old primary object store to be the first alternate"); | |
722 | ||
723 | the_repository->objects->odb = restore_odb; | |
724 | free_object_directory(cur_odb); | |
725 | } | |
726 | ||
9eeea7d2 SB |
727 | /* |
728 | * Compute the exact path an alternate is at and returns it. In case of | |
729 | * error NULL is returned and the human readable error is added to `err` | |
efde7b72 | 730 | * `path` may be relative and should point to $GIT_DIR. |
9eeea7d2 SB |
731 | * `err` must not be null. |
732 | */ | |
733 | char *compute_alternate_path(const char *path, struct strbuf *err) | |
734 | { | |
735 | char *ref_git = NULL; | |
4530a85b | 736 | const char *repo; |
9eeea7d2 SB |
737 | int seen_error = 0; |
738 | ||
4530a85b AM |
739 | ref_git = real_pathdup(path, 0); |
740 | if (!ref_git) { | |
9eeea7d2 SB |
741 | seen_error = 1; |
742 | strbuf_addf(err, _("path '%s' does not exist"), path); | |
743 | goto out; | |
4530a85b | 744 | } |
9eeea7d2 SB |
745 | |
746 | repo = read_gitfile(ref_git); | |
747 | if (!repo) | |
748 | repo = read_gitfile(mkpath("%s/.git", ref_git)); | |
749 | if (repo) { | |
750 | free(ref_git); | |
751 | ref_git = xstrdup(repo); | |
752 | } | |
753 | ||
754 | if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) { | |
755 | char *ref_git_git = mkpathdup("%s/.git", ref_git); | |
756 | free(ref_git); | |
757 | ref_git = ref_git_git; | |
758 | } else if (!is_directory(mkpath("%s/objects", ref_git))) { | |
759 | struct strbuf sb = STRBUF_INIT; | |
760 | seen_error = 1; | |
761 | if (get_common_dir(&sb, ref_git)) { | |
762 | strbuf_addf(err, | |
763 | _("reference repository '%s' as a linked " | |
764 | "checkout is not supported yet."), | |
765 | path); | |
766 | goto out; | |
767 | } | |
768 | ||
769 | strbuf_addf(err, _("reference repository '%s' is not a " | |
770 | "local repository."), path); | |
771 | goto out; | |
772 | } | |
773 | ||
774 | if (!access(mkpath("%s/shallow", ref_git), F_OK)) { | |
775 | strbuf_addf(err, _("reference repository '%s' is shallow"), | |
776 | path); | |
777 | seen_error = 1; | |
778 | goto out; | |
779 | } | |
780 | ||
781 | if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) { | |
782 | strbuf_addf(err, | |
783 | _("reference repository '%s' is grafted"), | |
784 | path); | |
785 | seen_error = 1; | |
786 | goto out; | |
787 | } | |
788 | ||
789 | out: | |
790 | if (seen_error) { | |
6a83d902 | 791 | FREE_AND_NULL(ref_git); |
9eeea7d2 SB |
792 | } |
793 | ||
794 | return ref_git; | |
795 | } | |
796 | ||
f57a7396 TB |
797 | struct object_directory *find_odb(struct repository *r, const char *obj_dir) |
798 | { | |
799 | struct object_directory *odb; | |
800 | char *obj_dir_real = real_pathdup(obj_dir, 1); | |
801 | struct strbuf odb_path_real = STRBUF_INIT; | |
802 | ||
803 | prepare_alt_odb(r); | |
804 | for (odb = r->objects->odb; odb; odb = odb->next) { | |
805 | strbuf_realpath(&odb_path_real, odb->path, 1); | |
806 | if (!strcmp(obj_dir_real, odb_path_real.buf)) | |
807 | break; | |
808 | } | |
809 | ||
810 | free(obj_dir_real); | |
811 | strbuf_release(&odb_path_real); | |
812 | ||
813 | if (!odb) | |
814 | die(_("could not find object directory matching %s"), obj_dir); | |
815 | return odb; | |
816 | } | |
817 | ||
709dfa69 JK |
818 | static void fill_alternate_refs_command(struct child_process *cmd, |
819 | const char *repo_path) | |
820 | { | |
821 | const char *value; | |
822 | ||
823 | if (!git_config_get_value("core.alternateRefsCommand", &value)) { | |
824 | cmd->use_shell = 1; | |
825 | ||
c972bf4c JK |
826 | strvec_push(&cmd->args, value); |
827 | strvec_push(&cmd->args, repo_path); | |
709dfa69 JK |
828 | } else { |
829 | cmd->git_cmd = 1; | |
830 | ||
c972bf4c JK |
831 | strvec_pushf(&cmd->args, "--git-dir=%s", repo_path); |
832 | strvec_push(&cmd->args, "for-each-ref"); | |
833 | strvec_push(&cmd->args, "--format=%(objectname)"); | |
709dfa69 JK |
834 | |
835 | if (!git_config_get_value("core.alternateRefsPrefixes", &value)) { | |
c972bf4c JK |
836 | strvec_push(&cmd->args, "--"); |
837 | strvec_split(&cmd->args, value); | |
709dfa69 JK |
838 | } |
839 | } | |
840 | ||
29fda24d | 841 | strvec_pushv(&cmd->env, (const char **)local_repo_env); |
709dfa69 JK |
842 | cmd->out = -1; |
843 | } | |
844 | ||
845 | static void read_alternate_refs(const char *path, | |
846 | alternate_ref_fn *cb, | |
847 | void *data) | |
848 | { | |
849 | struct child_process cmd = CHILD_PROCESS_INIT; | |
850 | struct strbuf line = STRBUF_INIT; | |
851 | FILE *fh; | |
852 | ||
853 | fill_alternate_refs_command(&cmd, path); | |
854 | ||
855 | if (start_command(&cmd)) | |
856 | return; | |
857 | ||
858 | fh = xfdopen(cmd.out, "r"); | |
859 | while (strbuf_getline_lf(&line, fh) != EOF) { | |
860 | struct object_id oid; | |
861 | const char *p; | |
862 | ||
863 | if (parse_oid_hex(line.buf, &oid, &p) || *p) { | |
864 | warning(_("invalid line while parsing alternate refs: %s"), | |
865 | line.buf); | |
866 | break; | |
867 | } | |
868 | ||
869 | cb(&oid, data); | |
870 | } | |
871 | ||
872 | fclose(fh); | |
873 | finish_command(&cmd); | |
86ad3ea5 | 874 | strbuf_release(&line); |
709dfa69 JK |
875 | } |
876 | ||
877 | struct alternate_refs_data { | |
878 | alternate_ref_fn *fn; | |
879 | void *data; | |
880 | }; | |
881 | ||
882 | static int refs_from_alternate_cb(struct object_directory *e, | |
883 | void *data) | |
884 | { | |
885 | struct strbuf path = STRBUF_INIT; | |
886 | size_t base_len; | |
887 | struct alternate_refs_data *cb = data; | |
888 | ||
889 | if (!strbuf_realpath(&path, e->path, 0)) | |
890 | goto out; | |
891 | if (!strbuf_strip_suffix(&path, "/objects")) | |
892 | goto out; | |
893 | base_len = path.len; | |
894 | ||
895 | /* Is this a git repository with refs? */ | |
896 | strbuf_addstr(&path, "/refs"); | |
897 | if (!is_directory(path.buf)) | |
898 | goto out; | |
899 | strbuf_setlen(&path, base_len); | |
900 | ||
901 | read_alternate_refs(path.buf, cb->fn, cb->data); | |
902 | ||
903 | out: | |
904 | strbuf_release(&path); | |
905 | return 0; | |
906 | } | |
907 | ||
908 | void for_each_alternate_ref(alternate_ref_fn fn, void *data) | |
909 | { | |
910 | struct alternate_refs_data cb; | |
911 | cb.fn = fn; | |
912 | cb.data = data; | |
913 | foreach_alt_odb(refs_from_alternate_cb, &cb); | |
914 | } | |
915 | ||
fe1b2268 | 916 | int foreach_alt_odb(alt_odb_fn fn, void *cb) |
d79796bc | 917 | { |
263db403 | 918 | struct object_directory *ent; |
fe1b2268 | 919 | int r = 0; |
d79796bc | 920 | |
0b209034 | 921 | prepare_alt_odb(the_repository); |
f0eaf638 | 922 | for (ent = the_repository->objects->odb->next; ent; ent = ent->next) { |
fe1b2268 JK |
923 | r = fn(ent, cb); |
924 | if (r) | |
925 | break; | |
926 | } | |
927 | return r; | |
d79796bc JH |
928 | } |
929 | ||
13068bf0 | 930 | void prepare_alt_odb(struct repository *r) |
c2f493a4 | 931 | { |
f0eaf638 | 932 | if (r->objects->loaded_alternates) |
7dc24aa5 SP |
933 | return; |
934 | ||
13068bf0 | 935 | link_alt_odb_entries(r, r->objects->alternate_db, PATH_SEP, NULL, 0); |
c2f493a4 | 936 | |
f0eaf638 JK |
937 | read_info_alternates(r, r->objects->odb->path, 0); |
938 | r->objects->loaded_alternates = 1; | |
c2f493a4 MW |
939 | } |
940 | ||
3096b2ec | 941 | /* Returns 1 if we have successfully freshened the file, 0 otherwise. */ |
33d4221c | 942 | static int freshen_file(const char *fn) |
ace1534d | 943 | { |
312cd761 | 944 | return !utime(fn, NULL); |
0f4dc14a | 945 | } |
ace1534d | 946 | |
3096b2ec JK |
947 | /* |
948 | * All of the check_and_freshen functions return 1 if the file exists and was | |
949 | * freshened (if freshening was requested), 0 otherwise. If they return | |
950 | * 0, you should not assume that it is safe to skip a write of the object (it | |
951 | * either does not exist on disk, or has a stale mtime and may be subject to | |
952 | * pruning). | |
953 | */ | |
6a5e6f5e | 954 | int check_and_freshen_file(const char *fn, int freshen) |
33d4221c JK |
955 | { |
956 | if (access(fn, F_OK)) | |
957 | return 0; | |
3096b2ec | 958 | if (freshen && !freshen_file(fn)) |
33d4221c JK |
959 | return 0; |
960 | return 1; | |
961 | } | |
962 | ||
f0eaf638 JK |
963 | static int check_and_freshen_odb(struct object_directory *odb, |
964 | const struct object_id *oid, | |
965 | int freshen) | |
33d4221c | 966 | { |
f0eaf638 | 967 | static struct strbuf path = STRBUF_INIT; |
514c5fdd | 968 | odb_loose_path(odb, &path, oid); |
f0eaf638 JK |
969 | return check_and_freshen_file(path.buf, freshen); |
970 | } | |
ea657730 | 971 | |
f0eaf638 JK |
972 | static int check_and_freshen_local(const struct object_id *oid, int freshen) |
973 | { | |
974 | return check_and_freshen_odb(the_repository->objects->odb, oid, freshen); | |
33d4221c JK |
975 | } |
976 | ||
6862ebbf | 977 | static int check_and_freshen_nonlocal(const struct object_id *oid, int freshen) |
0f4dc14a | 978 | { |
263db403 | 979 | struct object_directory *odb; |
f3f043a1 | 980 | |
0b209034 | 981 | prepare_alt_odb(the_repository); |
f0eaf638 JK |
982 | for (odb = the_repository->objects->odb->next; odb; odb = odb->next) { |
983 | if (check_and_freshen_odb(odb, oid, freshen)) | |
c529d75a | 984 | return 1; |
ace1534d | 985 | } |
c529d75a | 986 | return 0; |
ace1534d JH |
987 | } |
988 | ||
6862ebbf | 989 | static int check_and_freshen(const struct object_id *oid, int freshen) |
33d4221c | 990 | { |
6862ebbf | 991 | return check_and_freshen_local(oid, freshen) || |
992 | check_and_freshen_nonlocal(oid, freshen); | |
33d4221c JK |
993 | } |
994 | ||
6862ebbf | 995 | int has_loose_object_nonlocal(const struct object_id *oid) |
33d4221c | 996 | { |
6862ebbf | 997 | return check_and_freshen_nonlocal(oid, 0); |
33d4221c JK |
998 | } |
999 | ||
b7573536 | 1000 | int has_loose_object(const struct object_id *oid) |
0f4dc14a | 1001 | { |
6862ebbf | 1002 | return check_and_freshen(oid, 0); |
0f4dc14a BC |
1003 | } |
1004 | ||
02710228 SP |
1005 | static void mmap_limit_check(size_t length) |
1006 | { | |
1007 | static size_t limit = 0; | |
1008 | if (!limit) { | |
1009 | limit = git_env_ulong("GIT_MMAP_LIMIT", 0); | |
1010 | if (!limit) | |
1011 | limit = SIZE_MAX; | |
1012 | } | |
1013 | if (length > limit) | |
259328b7 | 1014 | die(_("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX), |
02710228 SP |
1015 | (uintmax_t)length, (uintmax_t)limit); |
1016 | } | |
1017 | ||
1570856b JK |
1018 | void *xmmap_gently(void *start, size_t length, |
1019 | int prot, int flags, int fd, off_t offset) | |
58ecbd5e | 1020 | { |
02710228 SP |
1021 | void *ret; |
1022 | ||
1023 | mmap_limit_check(length); | |
1024 | ret = mmap(start, length, prot, flags, fd, offset); | |
9827d4c1 JK |
1025 | if (ret == MAP_FAILED && !length) |
1026 | ret = NULL; | |
58ecbd5e JN |
1027 | return ret; |
1028 | } | |
1029 | ||
dc059294 EW |
1030 | const char *mmap_os_err(void) |
1031 | { | |
1032 | static const char blank[] = ""; | |
1033 | #if defined(__linux__) | |
1034 | if (errno == ENOMEM) { | |
1035 | /* this continues an existing error message: */ | |
1036 | static const char enomem[] = | |
1037 | ", check sys.vm.max_map_count and/or RLIMIT_DATA"; | |
1038 | return enomem; | |
1039 | } | |
1040 | #endif /* OS-specific bits */ | |
1041 | return blank; | |
1042 | } | |
1043 | ||
1570856b JK |
1044 | void *xmmap(void *start, size_t length, |
1045 | int prot, int flags, int fd, off_t offset) | |
1046 | { | |
1047 | void *ret = xmmap_gently(start, length, prot, flags, fd, offset); | |
1048 | if (ret == MAP_FAILED) | |
dc059294 | 1049 | die_errno(_("mmap failed%s"), mmap_os_err()); |
1570856b JK |
1050 | return ret; |
1051 | } | |
1052 | ||
b04cdea4 ÆAB |
1053 | static int format_object_header_literally(char *str, size_t size, |
1054 | const char *type, size_t objsize) | |
1055 | { | |
1056 | return xsnprintf(str, size, "%s %"PRIuMAX, type, (uintmax_t)objsize) + 1; | |
1057 | } | |
1058 | ||
1059 | int format_object_header(char *str, size_t size, enum object_type type, | |
1060 | size_t objsize) | |
1061 | { | |
1062 | const char *name = type_name(type); | |
1063 | ||
1064 | if (!name) | |
1065 | BUG("could not get a type name for 'enum object_type' value %d", type); | |
1066 | ||
1067 | return format_object_header_literally(str, size, name, objsize); | |
1068 | } | |
1069 | ||
b98d1885 | 1070 | int check_object_signature(struct repository *r, const struct object_id *oid, |
44439c1c ÆAB |
1071 | void *buf, unsigned long size, |
1072 | enum object_type type) | |
88d0db55 | 1073 | { |
0f156dbb ÆAB |
1074 | struct object_id real_oid; |
1075 | ||
1076 | hash_object_file(r->hash_algo, buf, size, type, &real_oid); | |
1077 | ||
1078 | return !oideq(oid, &real_oid) ? -1 : 0; | |
1079 | } | |
1080 | ||
1081 | int stream_object_signature(struct repository *r, const struct object_id *oid) | |
1082 | { | |
1083 | struct object_id real_oid; | |
1084 | unsigned long size; | |
090ea126 NTND |
1085 | enum object_type obj_type; |
1086 | struct git_istream *st; | |
18e2588e | 1087 | git_hash_ctx c; |
1af64f73 | 1088 | char hdr[MAX_HEADER_LEN]; |
090ea126 | 1089 | int hdrlen; |
88d0db55 | 1090 | |
b98d1885 | 1091 | st = open_istream(r, oid, &obj_type, &size, NULL); |
090ea126 NTND |
1092 | if (!st) |
1093 | return -1; | |
88d0db55 | 1094 | |
090ea126 | 1095 | /* Generate the header */ |
b04cdea4 | 1096 | hdrlen = format_object_header(hdr, sizeof(hdr), obj_type, size); |
88d0db55 | 1097 | |
090ea126 | 1098 | /* Sha1.. */ |
b98d1885 MT |
1099 | r->hash_algo->init_fn(&c); |
1100 | r->hash_algo->update_fn(&c, hdr, hdrlen); | |
090ea126 NTND |
1101 | for (;;) { |
1102 | char buf[1024 * 16]; | |
1103 | ssize_t readlen = read_istream(st, buf, sizeof(buf)); | |
1f688557 | 1104 | |
f54fac53 JK |
1105 | if (readlen < 0) { |
1106 | close_istream(st); | |
1107 | return -1; | |
1108 | } | |
090ea126 NTND |
1109 | if (!readlen) |
1110 | break; | |
b98d1885 | 1111 | r->hash_algo->update_fn(&c, buf, readlen); |
fa5fc15d | 1112 | } |
0f156dbb | 1113 | r->hash_algo->final_oid_fn(&real_oid, &c); |
090ea126 | 1114 | close_istream(st); |
0f156dbb | 1115 | return !oideq(oid, &real_oid) ? -1 : 0; |
fa5fc15d SP |
1116 | } |
1117 | ||
1b8ac5ea | 1118 | int git_open_cloexec(const char *name, int flags) |
a0788266 | 1119 | { |
1e3001a8 JH |
1120 | int fd; |
1121 | static int o_cloexec = O_CLOEXEC; | |
a0788266 | 1122 | |
1e3001a8 JH |
1123 | fd = open(name, flags | o_cloexec); |
1124 | if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) { | |
cd66ada0 | 1125 | /* Try again w/o O_CLOEXEC: the kernel might not support it */ |
1e3001a8 JH |
1126 | o_cloexec &= ~O_CLOEXEC; |
1127 | fd = open(name, flags | o_cloexec); | |
491a8dec | 1128 | } |
491a8dec | 1129 | |
9fb9495d | 1130 | #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC) |
491a8dec | 1131 | { |
1e3001a8 | 1132 | static int fd_cloexec = FD_CLOEXEC; |
a0788266 | 1133 | |
1e3001a8 JH |
1134 | if (!o_cloexec && 0 <= fd && fd_cloexec) { |
1135 | /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */ | |
9fb9495d EW |
1136 | int flags = fcntl(fd, F_GETFD); |
1137 | if (fcntl(fd, F_SETFD, flags | fd_cloexec)) | |
1e3001a8 | 1138 | fd_cloexec = 0; |
cd66ada0 | 1139 | } |
44d1c19e | 1140 | } |
a0788266 | 1141 | #endif |
1b8ac5ea | 1142 | return fd; |
a0788266 JS |
1143 | } |
1144 | ||
3cf8b462 | 1145 | /* |
514c5fdd | 1146 | * Find "oid" as a loose object in the local repository or in an alternate. |
771e7d57 JK |
1147 | * Returns 0 on success, negative on failure. |
1148 | * | |
1149 | * The "path" out-parameter will give the path of the object we found (if any). | |
1150 | * Note that it may point to static storage and is only valid until another | |
514c5fdd | 1151 | * call to stat_loose_object(). |
3cf8b462 | 1152 | */ |
514c5fdd JK |
1153 | static int stat_loose_object(struct repository *r, const struct object_id *oid, |
1154 | struct stat *st, const char **path) | |
1f688557 | 1155 | { |
263db403 | 1156 | struct object_directory *odb; |
ea657730 CC |
1157 | static struct strbuf buf = STRBUF_INIT; |
1158 | ||
d2607fa0 | 1159 | prepare_alt_odb(r); |
f0eaf638 | 1160 | for (odb = r->objects->odb; odb; odb = odb->next) { |
514c5fdd | 1161 | *path = odb_loose_path(odb, &buf, oid); |
771e7d57 | 1162 | if (!lstat(*path, st)) |
052fe5ea | 1163 | return 0; |
c7934306 SP |
1164 | } |
1165 | ||
3cf8b462 SP |
1166 | return -1; |
1167 | } | |
1168 | ||
771e7d57 | 1169 | /* |
514c5fdd | 1170 | * Like stat_loose_object(), but actually open the object and return the |
771e7d57 JK |
1171 | * descriptor. See the caveats on the "path" parameter above. |
1172 | */ | |
514c5fdd JK |
1173 | static int open_loose_object(struct repository *r, |
1174 | const struct object_id *oid, const char **path) | |
60bb8b14 | 1175 | { |
44d1c19e | 1176 | int fd; |
263db403 | 1177 | struct object_directory *odb; |
f0eaf638 | 1178 | int most_interesting_errno = ENOENT; |
ea657730 CC |
1179 | static struct strbuf buf = STRBUF_INIT; |
1180 | ||
ec7283e5 | 1181 | prepare_alt_odb(r); |
f0eaf638 | 1182 | for (odb = r->objects->odb; odb; odb = odb->next) { |
514c5fdd | 1183 | *path = odb_loose_path(odb, &buf, oid); |
771e7d57 | 1184 | fd = git_open(*path); |
44d1c19e LT |
1185 | if (fd >= 0) |
1186 | return fd; | |
f0eaf638 | 1187 | |
d6c8a05b JK |
1188 | if (most_interesting_errno == ENOENT) |
1189 | most_interesting_errno = errno; | |
03e79c88 | 1190 | } |
d6c8a05b | 1191 | errno = most_interesting_errno; |
44d1c19e | 1192 | return -1; |
1f688557 JH |
1193 | } |
1194 | ||
61c7711c | 1195 | static int quick_has_loose(struct repository *r, |
d7a24573 | 1196 | const struct object_id *oid) |
61c7711c | 1197 | { |
61c7711c JK |
1198 | struct object_directory *odb; |
1199 | ||
61c7711c JK |
1200 | prepare_alt_odb(r); |
1201 | for (odb = r->objects->odb; odb; odb = odb->next) { | |
92d8ed8a | 1202 | if (oidtree_contains(odb_loose_cache(odb, oid), oid)) |
61c7711c JK |
1203 | return 1; |
1204 | } | |
1205 | return 0; | |
1206 | } | |
1207 | ||
f6371f92 JK |
1208 | /* |
1209 | * Map the loose object at "path" if it is not NULL, or the path found by | |
514c5fdd | 1210 | * searching for a loose object named "oid". |
f6371f92 | 1211 | */ |
514c5fdd JK |
1212 | static void *map_loose_object_1(struct repository *r, const char *path, |
1213 | const struct object_id *oid, unsigned long *size) | |
27d69a46 | 1214 | { |
0fcfd160 | 1215 | void *map; |
144bde78 | 1216 | int fd; |
84dd81c1 | 1217 | |
f6371f92 JK |
1218 | if (path) |
1219 | fd = git_open(path); | |
8261e1f1 | 1220 | else |
514c5fdd | 1221 | fd = open_loose_object(r, oid, &path); |
44d1c19e LT |
1222 | map = NULL; |
1223 | if (fd >= 0) { | |
1224 | struct stat st; | |
1f688557 | 1225 | |
44d1c19e LT |
1226 | if (!fstat(fd, &st)) { |
1227 | *size = xsize_t(st.st_size); | |
33e42de0 MM |
1228 | if (!*size) { |
1229 | /* mmap() is forbidden on empty files */ | |
259328b7 | 1230 | error(_("object file %s is empty"), path); |
6881925e | 1231 | close(fd); |
33e42de0 MM |
1232 | return NULL; |
1233 | } | |
44d1c19e | 1234 | map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0); |
144bde78 | 1235 | } |
44d1c19e | 1236 | close(fd); |
74e34e1f | 1237 | } |
0fcfd160 | 1238 | return map; |
74e34e1f NP |
1239 | } |
1240 | ||
514c5fdd JK |
1241 | void *map_loose_object(struct repository *r, |
1242 | const struct object_id *oid, | |
1243 | unsigned long *size) | |
068f85e3 | 1244 | { |
514c5fdd | 1245 | return map_loose_object_1(r, NULL, oid, size); |
068f85e3 | 1246 | } |
1247 | ||
3b6a8db3 ÆAB |
1248 | enum unpack_loose_header_result unpack_loose_header(git_zstream *stream, |
1249 | unsigned char *map, | |
1250 | unsigned long mapsize, | |
1251 | void *buffer, | |
1252 | unsigned long bufsiz, | |
1253 | struct strbuf *header) | |
47fe3f6e | 1254 | { |
01cab976 | 1255 | int status; |
31877c9a | 1256 | |
c4483576 LT |
1257 | /* Get the data stream */ |
1258 | memset(stream, 0, sizeof(*stream)); | |
1259 | stream->next_in = map; | |
1260 | stream->avail_in = mapsize; | |
1261 | stream->next_out = buffer; | |
93821bd9 | 1262 | stream->avail_out = bufsiz; |
47fe3f6e | 1263 | |
39c68542 | 1264 | git_inflate_init(stream); |
31877c9a | 1265 | obj_read_unlock(); |
01cab976 | 1266 | status = git_inflate(stream, 0); |
31877c9a | 1267 | obj_read_lock(); |
d21f8426 | 1268 | if (status < Z_OK) |
3b6a8db3 | 1269 | return ULHR_BAD; |
d131b7af | 1270 | |
46f03448 KN |
1271 | /* |
1272 | * Check if entire header is unpacked in the first iteration. | |
d131b7af | 1273 | */ |
46f03448 | 1274 | if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) |
3b6a8db3 | 1275 | return ULHR_OK; |
d131b7af | 1276 | |
01cab976 ÆAB |
1277 | /* |
1278 | * We have a header longer than MAX_HEADER_LEN. The "header" | |
1279 | * here is only non-NULL when we run "cat-file | |
1280 | * --allow-unknown-type". | |
1281 | */ | |
1282 | if (!header) | |
5848fb11 | 1283 | return ULHR_TOO_LONG; |
d131b7af | 1284 | |
46f03448 KN |
1285 | /* |
1286 | * buffer[0..bufsiz] was not large enough. Copy the partial | |
1287 | * result out to header, and then append the result of further | |
1288 | * reading the stream. | |
1289 | */ | |
1290 | strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer); | |
1291 | stream->next_out = buffer; | |
1292 | stream->avail_out = bufsiz; | |
d131b7af | 1293 | |
46f03448 | 1294 | do { |
31877c9a | 1295 | obj_read_unlock(); |
46f03448 | 1296 | status = git_inflate(stream, 0); |
31877c9a | 1297 | obj_read_lock(); |
46f03448 KN |
1298 | strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer); |
1299 | if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) | |
1300 | return 0; | |
1301 | stream->next_out = buffer; | |
1302 | stream->avail_out = bufsiz; | |
1303 | } while (status != Z_STREAM_END); | |
5848fb11 | 1304 | return ULHR_TOO_LONG; |
d131b7af SP |
1305 | } |
1306 | ||
00a7760e JK |
1307 | static void *unpack_loose_rest(git_zstream *stream, |
1308 | void *buffer, unsigned long size, | |
1309 | const struct object_id *oid) | |
95099731 | 1310 | { |
5180cacc | 1311 | int bytes = strlen(buffer) + 1; |
3aee68aa | 1312 | unsigned char *buf = xmallocz(size); |
93821bd9 | 1313 | unsigned long n; |
7efbff75 | 1314 | int status = Z_OK; |
95099731 | 1315 | |
93821bd9 LT |
1316 | n = stream->total_out - bytes; |
1317 | if (n > size) | |
1318 | n = size; | |
1319 | memcpy(buf, (char *) buffer + bytes, n); | |
1320 | bytes = n; | |
456cdf6e LT |
1321 | if (bytes <= size) { |
1322 | /* | |
1323 | * The above condition must be (bytes <= size), not | |
1324 | * (bytes < size). In other words, even though we | |
ccf5ace0 | 1325 | * expect no more output and set avail_out to zero, |
456cdf6e LT |
1326 | * the input zlib stream may have bytes that express |
1327 | * "this concludes the stream", and we *do* want to | |
1328 | * eat that input. | |
1329 | * | |
1330 | * Otherwise we would not be able to test that we | |
1331 | * consumed all the input to reach the expected size; | |
1332 | * we also want to check that zlib tells us that all | |
1333 | * went well with status == Z_STREAM_END at the end. | |
1334 | */ | |
5180cacc LT |
1335 | stream->next_out = buf + bytes; |
1336 | stream->avail_out = size - bytes; | |
31877c9a MT |
1337 | while (status == Z_OK) { |
1338 | obj_read_unlock(); | |
39c68542 | 1339 | status = git_inflate(stream, Z_FINISH); |
31877c9a MT |
1340 | obj_read_lock(); |
1341 | } | |
5180cacc | 1342 | } |
456cdf6e | 1343 | if (status == Z_STREAM_END && !stream->avail_in) { |
39c68542 | 1344 | git_inflate_end(stream); |
7efbff75 | 1345 | return buf; |
95099731 NTND |
1346 | } |
1347 | ||
7efbff75 | 1348 | if (status < 0) |
00a7760e | 1349 | error(_("corrupt loose object '%s'"), oid_to_hex(oid)); |
7efbff75 | 1350 | else if (stream->avail_in) |
259328b7 | 1351 | error(_("garbage at end of loose object '%s'"), |
00a7760e | 1352 | oid_to_hex(oid)); |
7efbff75 JH |
1353 | free(buf); |
1354 | return NULL; | |
95099731 NTND |
1355 | } |
1356 | ||
d40d535b | 1357 | /* |
5180cacc LT |
1358 | * We used to just use "sscanf()", but that's actually way |
1359 | * too permissive for what we want to check. So do an anal | |
1360 | * object header parse by hand. | |
d40d535b | 1361 | */ |
dccb32bf | 1362 | int parse_loose_header(const char *hdr, struct object_info *oi) |
1f688557 | 1363 | { |
46f03448 | 1364 | const char *type_buf = hdr; |
d6a09e79 | 1365 | size_t size; |
46f03448 | 1366 | int type, type_len = 0; |
43057304 | 1367 | |
5180cacc | 1368 | /* |
46f03448 | 1369 | * The type can be of any size but is followed by |
21666f1a | 1370 | * a space. |
5180cacc | 1371 | */ |
5180cacc LT |
1372 | for (;;) { |
1373 | char c = *hdr++; | |
d21f8426 JH |
1374 | if (!c) |
1375 | return -1; | |
5180cacc LT |
1376 | if (c == ' ') |
1377 | break; | |
46f03448 | 1378 | type_len++; |
5180cacc | 1379 | } |
1f688557 | 1380 | |
46f03448 | 1381 | type = type_from_string_gently(type_buf, type_len, 1); |
6ca32f47 BW |
1382 | if (oi->type_name) |
1383 | strbuf_add(oi->type_name, type_buf, type_len); | |
46f03448 KN |
1384 | if (oi->typep) |
1385 | *oi->typep = type; | |
5180cacc LT |
1386 | |
1387 | /* | |
1388 | * The length must follow immediately, and be in canonical | |
1389 | * decimal format (ie "010" is not valid). | |
1390 | */ | |
1391 | size = *hdr++ - '0'; | |
1392 | if (size > 9) | |
1393 | return -1; | |
1394 | if (size) { | |
1395 | for (;;) { | |
1396 | unsigned long c = *hdr - '0'; | |
1397 | if (c > 9) | |
1398 | break; | |
1399 | hdr++; | |
d6a09e79 | 1400 | size = st_add(st_mult(size, 10), c); |
1b1005d1 | 1401 | } |
c01f51cc | 1402 | } |
46f03448 KN |
1403 | |
1404 | if (oi->sizep) | |
d6a09e79 | 1405 | *oi->sizep = cast_size_t_to_ulong(size); |
5180cacc LT |
1406 | |
1407 | /* | |
1408 | * The length must be followed by a zero byte | |
1409 | */ | |
dccb32bf ÆAB |
1410 | if (*hdr) |
1411 | return -1; | |
21666f1a | 1412 | |
dccb32bf ÆAB |
1413 | /* |
1414 | * The format is valid, but the type may still be bogus. The | |
1415 | * Caller needs to check its oi->typep. | |
1416 | */ | |
1417 | return 0; | |
bf592c50 DB |
1418 | } |
1419 | ||
514c5fdd JK |
1420 | static int loose_object_info(struct repository *r, |
1421 | const struct object_id *oid, | |
1422 | struct object_info *oi, int flags) | |
65c2e0c3 | 1423 | { |
46f03448 KN |
1424 | int status = 0; |
1425 | unsigned long mapsize; | |
65c2e0c3 | 1426 | void *map; |
ef49a7a0 | 1427 | git_zstream stream; |
1af64f73 | 1428 | char hdr[MAX_HEADER_LEN]; |
46f03448 | 1429 | struct strbuf hdrbuf = STRBUF_INIT; |
c84a1f3e | 1430 | unsigned long size_scratch; |
dccb32bf | 1431 | enum object_type type_scratch; |
01cab976 | 1432 | int allow_unknown = flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE; |
65c2e0c3 | 1433 | |
b99b6bcc JK |
1434 | if (oi->delta_base_oid) |
1435 | oidclr(oi->delta_base_oid); | |
5d642e75 | 1436 | |
052fe5ea JK |
1437 | /* |
1438 | * If we don't care about type or size, then we don't | |
4ef8d1dd JH |
1439 | * need to look inside the object at all. Note that we |
1440 | * do not optimize out the stat call, even if the | |
1441 | * caller doesn't care about the disk-size, since our | |
1442 | * return value implicitly indicates whether the | |
1443 | * object even exists. | |
052fe5ea | 1444 | */ |
6ca32f47 | 1445 | if (!oi->typep && !oi->type_name && !oi->sizep && !oi->contentp) { |
771e7d57 | 1446 | const char *path; |
4ef8d1dd | 1447 | struct stat st; |
61c7711c | 1448 | if (!oi->disk_sizep && (flags & OBJECT_INFO_QUICK)) |
d7a24573 | 1449 | return quick_has_loose(r, oid) ? 0 : -1; |
514c5fdd | 1450 | if (stat_loose_object(r, oid, &st, &path) < 0) |
4ef8d1dd JH |
1451 | return -1; |
1452 | if (oi->disk_sizep) | |
23c339c0 | 1453 | *oi->disk_sizep = st.st_size; |
052fe5ea JK |
1454 | return 0; |
1455 | } | |
1456 | ||
514c5fdd | 1457 | map = map_loose_object(r, oid, &mapsize); |
f0df4ed5 | 1458 | if (!map) |
dbea72a8 | 1459 | return -1; |
c84a1f3e JT |
1460 | |
1461 | if (!oi->sizep) | |
1462 | oi->sizep = &size_scratch; | |
dccb32bf ÆAB |
1463 | if (!oi->typep) |
1464 | oi->typep = &type_scratch; | |
c84a1f3e | 1465 | |
23c339c0 JK |
1466 | if (oi->disk_sizep) |
1467 | *oi->disk_sizep = mapsize; | |
01cab976 | 1468 | |
3b6a8db3 ÆAB |
1469 | switch (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr), |
1470 | allow_unknown ? &hdrbuf : NULL)) { | |
1471 | case ULHR_OK: | |
dccb32bf ÆAB |
1472 | if (parse_loose_header(hdrbuf.len ? hdrbuf.buf : hdr, oi) < 0) |
1473 | status = error(_("unable to parse %s header"), oid_to_hex(oid)); | |
1474 | else if (!allow_unknown && *oi->typep < 0) | |
1475 | die(_("invalid object type")); | |
1476 | ||
1477 | if (!oi->contentp) | |
1478 | break; | |
1479 | *oi->contentp = unpack_loose_rest(&stream, hdr, *oi->sizep, oid); | |
1480 | if (*oi->contentp) | |
1481 | goto cleanup; | |
1482 | ||
1483 | status = -1; | |
3b6a8db3 ÆAB |
1484 | break; |
1485 | case ULHR_BAD: | |
259328b7 | 1486 | status = error(_("unable to unpack %s header"), |
514c5fdd | 1487 | oid_to_hex(oid)); |
3b6a8db3 | 1488 | break; |
5848fb11 ÆAB |
1489 | case ULHR_TOO_LONG: |
1490 | status = error(_("header for %s too long, exceeds %d bytes"), | |
1491 | oid_to_hex(oid), MAX_HEADER_LEN); | |
1492 | break; | |
3b6a8db3 | 1493 | } |
c84a1f3e | 1494 | |
dccb32bf ÆAB |
1495 | git_inflate_end(&stream); |
1496 | cleanup: | |
65c2e0c3 | 1497 | munmap(map, mapsize); |
c84a1f3e JT |
1498 | if (oi->sizep == &size_scratch) |
1499 | oi->sizep = NULL; | |
46f03448 | 1500 | strbuf_release(&hdrbuf); |
dccb32bf ÆAB |
1501 | if (oi->typep == &type_scratch) |
1502 | oi->typep = NULL; | |
3ab0fb06 | 1503 | oi->whence = OI_LOOSE; |
dccb32bf | 1504 | return status; |
65c2e0c3 JH |
1505 | } |
1506 | ||
31877c9a MT |
1507 | int obj_read_use_lock = 0; |
1508 | pthread_mutex_t obj_read_mutex; | |
1509 | ||
1510 | void enable_obj_read_lock(void) | |
1511 | { | |
1512 | if (obj_read_use_lock) | |
1513 | return; | |
1514 | ||
1515 | obj_read_use_lock = 1; | |
1516 | init_recursive_mutex(&obj_read_mutex); | |
1517 | } | |
1518 | ||
1519 | void disable_obj_read_lock(void) | |
1520 | { | |
1521 | if (!obj_read_use_lock) | |
1522 | return; | |
1523 | ||
1524 | obj_read_use_lock = 0; | |
1525 | pthread_mutex_destroy(&obj_read_mutex); | |
1526 | } | |
1527 | ||
8b4c0103 JT |
1528 | int fetch_if_missing = 1; |
1529 | ||
31877c9a MT |
1530 | static int do_oid_object_info_extended(struct repository *r, |
1531 | const struct object_id *oid, | |
1532 | struct object_info *oi, unsigned flags) | |
f0df4ed5 | 1533 | { |
cd585e2a | 1534 | static struct object_info blank_oi = OBJECT_INFO_INIT; |
9c8a294a | 1535 | struct cached_object *co; |
f0df4ed5 | 1536 | struct pack_entry e; |
5b086407 | 1537 | int rtype; |
b383a13c | 1538 | const struct object_id *real = oid; |
8b4c0103 | 1539 | int already_retried = 0; |
f0df4ed5 | 1540 | |
31877c9a | 1541 | |
b383a13c | 1542 | if (flags & OBJECT_INFO_LOOKUP_REPLACE) |
9d98354f | 1543 | real = lookup_replace_object(r, oid); |
f0df4ed5 | 1544 | |
b383a13c | 1545 | if (is_null_oid(real)) |
87b5e236 JK |
1546 | return -1; |
1547 | ||
cd585e2a JT |
1548 | if (!oi) |
1549 | oi = &blank_oi; | |
1550 | ||
9c8a294a JT |
1551 | co = find_cached_object(real); |
1552 | if (co) { | |
1553 | if (oi->typep) | |
1554 | *(oi->typep) = co->type; | |
1555 | if (oi->sizep) | |
1556 | *(oi->sizep) = co->size; | |
1557 | if (oi->disk_sizep) | |
1558 | *(oi->disk_sizep) = 0; | |
b99b6bcc JK |
1559 | if (oi->delta_base_oid) |
1560 | oidclr(oi->delta_base_oid); | |
9c8a294a JT |
1561 | if (oi->type_name) |
1562 | strbuf_addstr(oi->type_name, type_name(co->type)); | |
1563 | if (oi->contentp) | |
1564 | *oi->contentp = xmemdupz(co->buf, co->size); | |
1565 | oi->whence = OI_CACHED; | |
1566 | return 0; | |
c4d9986f NTND |
1567 | } |
1568 | ||
8b4c0103 | 1569 | while (1) { |
42c8ce1c | 1570 | if (find_pack_entry(r, real, &e)) |
8b4c0103 JT |
1571 | break; |
1572 | ||
024aa469 TI |
1573 | if (flags & OBJECT_INFO_IGNORE_LOOSE) |
1574 | return -1; | |
1575 | ||
ddd63e64 | 1576 | /* Most likely it's a loose object. */ |
514c5fdd | 1577 | if (!loose_object_info(r, real, oi, flags)) |
5b086407 | 1578 | return 0; |
ddd63e64 SG |
1579 | |
1580 | /* Not a loose object; someone else may have just packed it. */ | |
2b7750c9 | 1581 | if (!(flags & OBJECT_INFO_QUICK)) { |
9d98354f | 1582 | reprepare_packed_git(r); |
42c8ce1c | 1583 | if (find_pack_entry(r, real, &e)) |
2b7750c9 JT |
1584 | break; |
1585 | } | |
8b4c0103 | 1586 | |
eef71904 JT |
1587 | /* |
1588 | * If r is the_repository, this might be an attempt at | |
1589 | * accessing a submodule object as if it were in the_repository | |
1590 | * (having called add_submodule_odb() on that submodule's ODB). | |
1591 | * If any such ODBs exist, register them and try again. | |
1592 | */ | |
1593 | if (r == the_repository && | |
1594 | register_all_submodule_odb_as_alternates()) | |
a35e03de JT |
1595 | /* We added some alternates; retry */ |
1596 | continue; | |
1597 | ||
8b4c0103 | 1598 | /* Check if it is a missing object */ |
ef830cc4 JT |
1599 | if (fetch_if_missing && repo_has_promisor_remote(r) && |
1600 | !already_retried && | |
31f5256c | 1601 | !(flags & OBJECT_INFO_SKIP_FETCH_OBJECT)) { |
8b4c0103 | 1602 | /* |
b14ed5ad CC |
1603 | * TODO Investigate checking promisor_remote_get_direct() |
1604 | * TODO return value and stopping on error here. | |
8b4c0103 | 1605 | */ |
b14ed5ad | 1606 | promisor_remote_get_direct(r, real, 1); |
8b4c0103 JT |
1607 | already_retried = 1; |
1608 | continue; | |
dfdd4afc | 1609 | } |
8b4c0103 JT |
1610 | |
1611 | return -1; | |
f0df4ed5 | 1612 | } |
3d77d877 | 1613 | |
cd585e2a JT |
1614 | if (oi == &blank_oi) |
1615 | /* | |
1616 | * We know that the caller doesn't actually need the | |
1617 | * information below, so return early. | |
1618 | */ | |
1619 | return 0; | |
9d98354f | 1620 | rtype = packed_object_info(r, e.p, e.offset, oi); |
412916ee | 1621 | if (rtype < 0) { |
751530de | 1622 | mark_bad_packed_object(e.p, real); |
31877c9a | 1623 | return do_oid_object_info_extended(r, real, oi, 0); |
3ab0fb06 | 1624 | } else if (oi->whence == OI_PACKED) { |
9a490590 JH |
1625 | oi->u.packed.offset = e.offset; |
1626 | oi->u.packed.pack = e.p; | |
1627 | oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA || | |
1628 | rtype == OBJ_OFS_DELTA); | |
3d77d877 NP |
1629 | } |
1630 | ||
5b086407 | 1631 | return 0; |
f0df4ed5 JS |
1632 | } |
1633 | ||
31877c9a MT |
1634 | int oid_object_info_extended(struct repository *r, const struct object_id *oid, |
1635 | struct object_info *oi, unsigned flags) | |
1636 | { | |
1637 | int ret; | |
1638 | obj_read_lock(); | |
1639 | ret = do_oid_object_info_extended(r, oid, oi, flags); | |
1640 | obj_read_unlock(); | |
1641 | return ret; | |
1642 | } | |
1643 | ||
1644 | ||
3fc0dca9 | 1645 | /* returns enum object_type or negative */ |
9d98354f SB |
1646 | int oid_object_info(struct repository *r, |
1647 | const struct object_id *oid, | |
1648 | unsigned long *sizep) | |
9a490590 | 1649 | { |
5b086407 | 1650 | enum object_type type; |
27b5c1a0 | 1651 | struct object_info oi = OBJECT_INFO_INIT; |
9a490590 | 1652 | |
5b086407 | 1653 | oi.typep = &type; |
9a490590 | 1654 | oi.sizep = sizep; |
9d98354f SB |
1655 | if (oid_object_info_extended(r, oid, &oi, |
1656 | OBJECT_INFO_LOOKUP_REPLACE) < 0) | |
5b086407 JK |
1657 | return -1; |
1658 | return type; | |
9a490590 JH |
1659 | } |
1660 | ||
1b9b5c69 | 1661 | static void *read_object(struct repository *r, |
cba595ab | 1662 | const struct object_id *oid, enum object_type *type, |
f1d8130b JT |
1663 | unsigned long *size) |
1664 | { | |
1665 | struct object_info oi = OBJECT_INFO_INIT; | |
1666 | void *content; | |
1667 | oi.typep = type; | |
1668 | oi.sizep = size; | |
1669 | oi.contentp = &content; | |
1670 | ||
cba595ab | 1671 | if (oid_object_info_extended(r, oid, &oi, 0) < 0) |
f1d8130b JT |
1672 | return NULL; |
1673 | return content; | |
1674 | } | |
1675 | ||
829e5c3b PO |
1676 | int pretend_object_file(void *buf, unsigned long len, enum object_type type, |
1677 | struct object_id *oid) | |
d66b37bb JH |
1678 | { |
1679 | struct cached_object *co; | |
1680 | ||
44439c1c | 1681 | hash_object_file(the_hash_algo, buf, len, type, oid); |
a64d2aae JT |
1682 | if (has_object_file_with_flags(oid, OBJECT_INFO_QUICK | OBJECT_INFO_SKIP_FETCH_OBJECT) || |
1683 | find_cached_object(oid)) | |
d66b37bb | 1684 | return 0; |
c7353967 | 1685 | ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc); |
d66b37bb JH |
1686 | co = &cached_objects[cached_object_nr++]; |
1687 | co->size = len; | |
21666f1a | 1688 | co->type = type; |
efa13f7b JH |
1689 | co->buf = xmalloc(len); |
1690 | memcpy(co->buf, buf, len); | |
62ba93ea | 1691 | oidcpy(&co->oid, oid); |
d66b37bb JH |
1692 | return 0; |
1693 | } | |
1694 | ||
b6c4cecc JH |
1695 | /* |
1696 | * This function dies on corrupt objects; the callers who want to | |
1697 | * deal with them should arrange to call read_object() and give error | |
1698 | * messages themselves. | |
1699 | */ | |
a3b72c89 SB |
1700 | void *read_object_file_extended(struct repository *r, |
1701 | const struct object_id *oid, | |
b4f5aca4 | 1702 | enum object_type *type, |
1703 | unsigned long *size, | |
1704 | int lookup_replace) | |
ac939109 | 1705 | { |
3ba7a065 | 1706 | void *data; |
b6c4cecc | 1707 | const struct packed_git *p; |
771e7d57 JK |
1708 | const char *path; |
1709 | struct stat st; | |
1f2e7cea | 1710 | const struct object_id *repl = lookup_replace ? |
a3b72c89 | 1711 | lookup_replace_object(r, oid) : oid; |
b6c4cecc | 1712 | |
3ba7a065 | 1713 | errno = 0; |
cba595ab | 1714 | data = read_object(r, repl, type, size); |
4bbf5a26 | 1715 | if (data) |
b6c4cecc | 1716 | return data; |
68095570 | 1717 | |
31877c9a | 1718 | obj_read_lock(); |
25f3af3f | 1719 | if (errno && errno != ENOENT) |
259328b7 | 1720 | die_errno(_("failed to read object %s"), oid_to_hex(oid)); |
3ba7a065 | 1721 | |
68095570 | 1722 | /* die if we replaced an object with one that does not exist */ |
b383a13c | 1723 | if (repl != oid) |
259328b7 | 1724 | die(_("replacement %s not found for %s"), |
b383a13c | 1725 | oid_to_hex(repl), oid_to_hex(oid)); |
68095570 | 1726 | |
cba595ab | 1727 | if (!stat_loose_object(r, repl, &st, &path)) |
259328b7 | 1728 | die(_("loose object %s (stored in %s) is corrupt"), |
b383a13c | 1729 | oid_to_hex(repl), path); |
68095570 | 1730 | |
afe8a907 | 1731 | if ((p = has_packed_and_bad(r, repl))) |
259328b7 | 1732 | die(_("packed object %s (stored in %s) is corrupt"), |
b383a13c | 1733 | oid_to_hex(repl), p->pack_name); |
31877c9a | 1734 | obj_read_unlock(); |
f5552aee | 1735 | |
b6c4cecc | 1736 | return NULL; |
ac939109 NP |
1737 | } |
1738 | ||
d3b4705a NTND |
1739 | void *read_object_with_reference(struct repository *r, |
1740 | const struct object_id *oid, | |
6aea6bae | 1741 | enum object_type required_type, |
40469ee9 | 1742 | unsigned long *size, |
02f0547e | 1743 | struct object_id *actual_oid_return) |
f4913f91 | 1744 | { |
6aea6bae | 1745 | enum object_type type; |
f4913f91 JH |
1746 | void *buffer; |
1747 | unsigned long isize; | |
02f0547e | 1748 | struct object_id actual_oid; |
f4913f91 | 1749 | |
02f0547e | 1750 | oidcpy(&actual_oid, oid); |
40469ee9 JH |
1751 | while (1) { |
1752 | int ref_length = -1; | |
1753 | const char *ref_type = NULL; | |
f4913f91 | 1754 | |
d3b4705a | 1755 | buffer = repo_read_object_file(r, &actual_oid, &type, &isize); |
40469ee9 JH |
1756 | if (!buffer) |
1757 | return NULL; | |
21666f1a | 1758 | if (type == required_type) { |
40469ee9 | 1759 | *size = isize; |
02f0547e | 1760 | if (actual_oid_return) |
1761 | oidcpy(actual_oid_return, &actual_oid); | |
40469ee9 JH |
1762 | return buffer; |
1763 | } | |
1764 | /* Handle references */ | |
21666f1a | 1765 | else if (type == OBJ_COMMIT) |
40469ee9 | 1766 | ref_type = "tree "; |
21666f1a | 1767 | else if (type == OBJ_TAG) |
40469ee9 JH |
1768 | ref_type = "object "; |
1769 | else { | |
1770 | free(buffer); | |
1771 | return NULL; | |
1772 | } | |
1773 | ref_length = strlen(ref_type); | |
f4913f91 | 1774 | |
94b5e093 | 1775 | if (ref_length + the_hash_algo->hexsz > isize || |
50974ec9 | 1776 | memcmp(buffer, ref_type, ref_length) || |
02f0547e | 1777 | get_oid_hex((char *) buffer + ref_length, &actual_oid)) { |
40469ee9 JH |
1778 | free(buffer); |
1779 | return NULL; | |
1780 | } | |
1cf58e72 | 1781 | free(buffer); |
40469ee9 | 1782 | /* Now we have the ID of the referred-to object in |
02f0547e | 1783 | * actual_oid. Check again. */ |
f4913f91 | 1784 | } |
f4913f91 JH |
1785 | } |
1786 | ||
2bbb28a3 ÆAB |
1787 | static void hash_object_body(const struct git_hash_algo *algo, git_hash_ctx *c, |
1788 | const void *buf, unsigned long len, | |
1789 | struct object_id *oid, | |
1790 | char *hdr, int *hdrlen) | |
1791 | { | |
1792 | algo->init_fn(c); | |
1793 | algo->update_fn(c, hdr, *hdrlen); | |
1794 | algo->update_fn(c, buf, len); | |
1795 | algo->final_oid_fn(oid, c); | |
1796 | } | |
1797 | ||
7ad5c44d MT |
1798 | static void write_object_file_prepare(const struct git_hash_algo *algo, |
1799 | const void *buf, unsigned long len, | |
2bbb28a3 | 1800 | enum object_type type, struct object_id *oid, |
a09c985e | 1801 | char *hdr, int *hdrlen) |
d410c0f5 | 1802 | { |
18e2588e | 1803 | git_hash_ctx c; |
d410c0f5 JH |
1804 | |
1805 | /* Generate the header */ | |
2bbb28a3 | 1806 | *hdrlen = format_object_header(hdr, *hdrlen, type, len); |
d410c0f5 JH |
1807 | |
1808 | /* Sha1.. */ | |
2bbb28a3 ÆAB |
1809 | hash_object_body(algo, &c, buf, len, oid, hdr, hdrlen); |
1810 | } | |
1811 | ||
1812 | static void write_object_file_prepare_literally(const struct git_hash_algo *algo, | |
1813 | const void *buf, unsigned long len, | |
1814 | const char *type, struct object_id *oid, | |
1815 | char *hdr, int *hdrlen) | |
1816 | { | |
1817 | git_hash_ctx c; | |
1818 | ||
1819 | *hdrlen = format_object_header_literally(hdr, *hdrlen, type, len); | |
1820 | hash_object_body(algo, &c, buf, len, oid, hdr, hdrlen); | |
d410c0f5 JH |
1821 | } |
1822 | ||
230f1322 | 1823 | /* |
5a688fe4 | 1824 | * Move the just written object into its final resting place. |
230f1322 | 1825 | */ |
cb5add58 | 1826 | int finalize_object_file(const char *tmpfile, const char *filename) |
230f1322 | 1827 | { |
e32c0a9c | 1828 | int ret = 0; |
5a688fe4 | 1829 | |
348df166 | 1830 | if (object_creation_mode == OBJECT_CREATION_USES_RENAMES) |
be66a6c4 JS |
1831 | goto try_rename; |
1832 | else if (link(tmpfile, filename)) | |
e32c0a9c | 1833 | ret = errno; |
7ebb6fca LT |
1834 | |
1835 | /* | |
1836 | * Coda hack - coda doesn't like cross-directory links, | |
1837 | * so we fall back to a rename, which will mean that it | |
1838 | * won't be able to check collisions, but that's not a | |
1839 | * big deal. | |
1840 | * | |
1841 | * The same holds for FAT formatted media. | |
1842 | * | |
3be1f18e | 1843 | * When this succeeds, we just return. We have nothing |
7ebb6fca LT |
1844 | * left to unlink. |
1845 | */ | |
1846 | if (ret && ret != EEXIST) { | |
be66a6c4 | 1847 | try_rename: |
7ebb6fca | 1848 | if (!rename(tmpfile, filename)) |
3be1f18e | 1849 | goto out; |
9e48b389 | 1850 | ret = errno; |
230f1322 | 1851 | } |
691f1a28 | 1852 | unlink_or_warn(tmpfile); |
230f1322 LT |
1853 | if (ret) { |
1854 | if (ret != EEXIST) { | |
2c319886 | 1855 | return error_errno(_("unable to write file %s"), filename); |
230f1322 LT |
1856 | } |
1857 | /* FIXME!!! Collision check here ? */ | |
1858 | } | |
1859 | ||
3be1f18e | 1860 | out: |
5256b006 | 1861 | if (adjust_shared_perm(filename)) |
259328b7 | 1862 | return error(_("unable to set permission to '%s'"), filename); |
230f1322 LT |
1863 | return 0; |
1864 | } | |
1865 | ||
4d548150 LT |
1866 | static int write_buffer(int fd, const void *buf, size_t len) |
1867 | { | |
d34cf19b | 1868 | if (write_in_full(fd, buf, len) < 0) |
259328b7 | 1869 | return error_errno(_("file write error")); |
4d548150 LT |
1870 | return 0; |
1871 | } | |
1872 | ||
44439c1c ÆAB |
1873 | static void hash_object_file_literally(const struct git_hash_algo *algo, |
1874 | const void *buf, unsigned long len, | |
1875 | const char *type, struct object_id *oid) | |
abdc3fc8 | 1876 | { |
1af64f73 | 1877 | char hdr[MAX_HEADER_LEN]; |
ef1286d3 | 1878 | int hdrlen = sizeof(hdr); |
44439c1c | 1879 | |
2bbb28a3 | 1880 | write_object_file_prepare_literally(algo, buf, len, type, oid, hdr, &hdrlen); |
abdc3fc8 RS |
1881 | } |
1882 | ||
44439c1c ÆAB |
1883 | void hash_object_file(const struct git_hash_algo *algo, const void *buf, |
1884 | unsigned long len, enum object_type type, | |
1885 | struct object_id *oid) | |
1886 | { | |
1887 | hash_object_file_literally(algo, buf, len, type_name(type), oid); | |
abdc3fc8 RS |
1888 | } |
1889 | ||
e9039dd3 | 1890 | /* Finalize a file on disk, and close it. */ |
c4e707f8 | 1891 | static void close_loose_object(int fd, const char *filename) |
e9039dd3 | 1892 | { |
020406ea NS |
1893 | if (the_repository->objects->odb->will_destroy) |
1894 | goto out; | |
b3cecf49 | 1895 | |
c0f4752e | 1896 | if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT)) |
83937e95 | 1897 | fsync_loose_object_bulk_checkin(fd, filename); |
c0f4752e | 1898 | else if (fsync_object_files > 0) |
c4e707f8 | 1899 | fsync_or_die(fd, filename); |
020406ea NS |
1900 | else |
1901 | fsync_component_or_die(FSYNC_COMPONENT_LOOSE_OBJECT, fd, | |
c4e707f8 | 1902 | filename); |
020406ea NS |
1903 | |
1904 | out: | |
e9039dd3 | 1905 | if (close(fd) != 0) |
76011357 | 1906 | die_errno(_("error when closing loose object file")); |
e9039dd3 LT |
1907 | } |
1908 | ||
5723fe7e LT |
1909 | /* Size of directory component, including the ending '/' */ |
1910 | static inline int directory_size(const char *filename) | |
1911 | { | |
1912 | const char *s = strrchr(filename, '/'); | |
1913 | if (!s) | |
1914 | return 0; | |
1915 | return s - filename + 1; | |
1916 | } | |
1917 | ||
1918 | /* | |
1919 | * This creates a temporary file in the same directory as the final | |
1920 | * 'filename' | |
1921 | * | |
1922 | * We want to avoid cross-directory filename renames, because those | |
1923 | * can have problems on various filesystems (FAT, NFS, Coda). | |
1924 | */ | |
d4b3d11a | 1925 | static int create_tmpfile(struct strbuf *tmp, const char *filename) |
5723fe7e LT |
1926 | { |
1927 | int fd, dirlen = directory_size(filename); | |
1928 | ||
d4b3d11a JK |
1929 | strbuf_reset(tmp); |
1930 | strbuf_add(tmp, filename, dirlen); | |
1931 | strbuf_addstr(tmp, "tmp_obj_XXXXXX"); | |
1932 | fd = git_mkstemp_mode(tmp->buf, 0444); | |
cbacbf4e | 1933 | if (fd < 0 && dirlen && errno == ENOENT) { |
d4b3d11a JK |
1934 | /* |
1935 | * Make sure the directory exists; note that the contents | |
1936 | * of the buffer are undefined after mkstemp returns an | |
1937 | * error, so we have to rewrite the whole buffer from | |
1938 | * scratch. | |
1939 | */ | |
1940 | strbuf_reset(tmp); | |
1941 | strbuf_add(tmp, filename, dirlen - 1); | |
1942 | if (mkdir(tmp->buf, 0777) && errno != EEXIST) | |
b2476a60 | 1943 | return -1; |
d4b3d11a | 1944 | if (adjust_shared_perm(tmp->buf)) |
5723fe7e LT |
1945 | return -1; |
1946 | ||
1947 | /* Try again */ | |
d4b3d11a JK |
1948 | strbuf_addstr(tmp, "/tmp_obj_XXXXXX"); |
1949 | fd = git_mkstemp_mode(tmp->buf, 0444); | |
5723fe7e LT |
1950 | } |
1951 | return fd; | |
1952 | } | |
1953 | ||
97a9db6f HX |
1954 | /** |
1955 | * Common steps for loose object writers to start writing loose | |
1956 | * objects: | |
1957 | * | |
1958 | * - Create tmpfile for the loose object. | |
1959 | * - Setup zlib stream for compression. | |
1960 | * - Start to feed header to zlib stream. | |
1961 | * | |
1962 | * Returns a "fd", which should later be provided to | |
1963 | * end_loose_object_common(). | |
1964 | */ | |
1965 | static int start_loose_object_common(struct strbuf *tmp_file, | |
1966 | const char *filename, unsigned flags, | |
1967 | git_zstream *stream, | |
1968 | unsigned char *buf, size_t buflen, | |
1969 | git_hash_ctx *c, | |
1970 | char *hdr, int hdrlen) | |
1971 | { | |
1972 | int fd; | |
1973 | ||
1974 | fd = create_tmpfile(tmp_file, filename); | |
1975 | if (fd < 0) { | |
1976 | if (flags & HASH_SILENT) | |
1977 | return -1; | |
1978 | else if (errno == EACCES) | |
1979 | return error(_("insufficient permission for adding " | |
1980 | "an object to repository database %s"), | |
1981 | get_object_directory()); | |
1982 | else | |
1983 | return error_errno( | |
1984 | _("unable to create temporary file")); | |
1985 | } | |
1986 | ||
1987 | /* Setup zlib stream for compression */ | |
1988 | git_deflate_init(stream, zlib_compression_level); | |
1989 | stream->next_out = buf; | |
1990 | stream->avail_out = buflen; | |
1991 | the_hash_algo->init_fn(c); | |
1992 | ||
1993 | /* Start to feed header to zlib stream */ | |
1994 | stream->next_in = (unsigned char *)hdr; | |
1995 | stream->avail_in = hdrlen; | |
1996 | while (git_deflate(stream, 0) == Z_OK) | |
1997 | ; /* nothing */ | |
1998 | the_hash_algo->update_fn(c, hdr, hdrlen); | |
1999 | ||
2000 | return fd; | |
2001 | } | |
2002 | ||
21e7d881 ÆAB |
2003 | /** |
2004 | * Common steps for the inner git_deflate() loop for writing loose | |
2005 | * objects. Returns what git_deflate() returns. | |
2006 | */ | |
2007 | static int write_loose_object_common(git_hash_ctx *c, | |
2008 | git_zstream *stream, const int flush, | |
2009 | unsigned char *in0, const int fd, | |
2010 | unsigned char *compressed, | |
2011 | const size_t compressed_len) | |
2012 | { | |
2013 | int ret; | |
2014 | ||
2015 | ret = git_deflate(stream, flush ? Z_FINISH : 0); | |
2016 | the_hash_algo->update_fn(c, in0, stream->next_in - in0); | |
2017 | if (write_buffer(fd, compressed, stream->next_out - compressed) < 0) | |
2018 | die(_("unable to write loose object file")); | |
2019 | stream->next_out = compressed; | |
2020 | stream->avail_out = compressed_len; | |
2021 | ||
2022 | return ret; | |
2023 | } | |
2024 | ||
97a9db6f HX |
2025 | /** |
2026 | * Common steps for loose object writers to end writing loose objects: | |
2027 | * | |
2028 | * - End the compression of zlib stream. | |
2029 | * - Get the calculated oid to "oid". | |
2030 | */ | |
2031 | static int end_loose_object_common(git_hash_ctx *c, git_zstream *stream, | |
2032 | struct object_id *oid) | |
2033 | { | |
2034 | int ret; | |
2035 | ||
2036 | ret = git_deflate_end_gently(stream); | |
2037 | if (ret != Z_OK) | |
2038 | return ret; | |
2039 | the_hash_algo->final_oid_fn(oid, c); | |
2040 | ||
2041 | return Z_OK; | |
2042 | } | |
2043 | ||
3fc7281f PO |
2044 | static int write_loose_object(const struct object_id *oid, char *hdr, |
2045 | int hdrlen, const void *buf, unsigned long len, | |
4ef91a2d | 2046 | time_t mtime, unsigned flags) |
0fcfd160 | 2047 | { |
915308b1 | 2048 | int fd, ret; |
9892beba | 2049 | unsigned char compressed[4096]; |
ef49a7a0 | 2050 | git_zstream stream; |
18e2588e | 2051 | git_hash_ctx c; |
3fc7281f | 2052 | struct object_id parano_oid; |
d4b3d11a | 2053 | static struct strbuf tmp_file = STRBUF_INIT; |
ea657730 CC |
2054 | static struct strbuf filename = STRBUF_INIT; |
2055 | ||
c0f4752e NS |
2056 | if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT)) |
2057 | prepare_loose_object_bulk_checkin(); | |
2058 | ||
514c5fdd | 2059 | loose_object_path(the_repository, &filename, oid); |
a44c9a5e | 2060 | |
97a9db6f HX |
2061 | fd = start_loose_object_common(&tmp_file, filename.buf, flags, |
2062 | &stream, compressed, sizeof(compressed), | |
2063 | &c, hdr, hdrlen); | |
2064 | if (fd < 0) | |
2065 | return -1; | |
a44c9a5e LT |
2066 | |
2067 | /* Then the data itself.. */ | |
c00e657d | 2068 | stream.next_in = (void *)buf; |
a44c9a5e | 2069 | stream.avail_in = len; |
9892beba | 2070 | do { |
748af44c | 2071 | unsigned char *in0 = stream.next_in; |
21e7d881 ÆAB |
2072 | |
2073 | ret = write_loose_object_common(&c, &stream, 1, in0, fd, | |
2074 | compressed, sizeof(compressed)); | |
9892beba NP |
2075 | } while (ret == Z_OK); |
2076 | ||
ac54c277 | 2077 | if (ret != Z_STREAM_END) |
259328b7 | 2078 | die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid), |
3fc7281f | 2079 | ret); |
97a9db6f | 2080 | ret = end_loose_object_common(&c, &stream, ¶no_oid); |
ac54c277 | 2081 | if (ret != Z_OK) |
259328b7 | 2082 | die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid), |
3fc7281f | 2083 | ret); |
9001dc2a | 2084 | if (!oideq(oid, ¶no_oid)) |
259328b7 | 2085 | die(_("confused by unstable object source data for %s"), |
3fc7281f | 2086 | oid_to_hex(oid)); |
ac54c277 | 2087 | |
c4e707f8 | 2088 | close_loose_object(fd, tmp_file.buf); |
0fcfd160 | 2089 | |
bbac7311 NP |
2090 | if (mtime) { |
2091 | struct utimbuf utb; | |
2092 | utb.actime = mtime; | |
2093 | utb.modtime = mtime; | |
4ef91a2d ÆAB |
2094 | if (utime(tmp_file.buf, &utb) < 0 && |
2095 | !(flags & HASH_SILENT)) | |
259328b7 | 2096 | warning_errno(_("failed utime() on %s"), tmp_file.buf); |
bbac7311 NP |
2097 | } |
2098 | ||
ea657730 | 2099 | return finalize_object_file(tmp_file.buf, filename.buf); |
0fcfd160 | 2100 | } |
8237b185 | 2101 | |
6862ebbf | 2102 | static int freshen_loose_object(const struct object_id *oid) |
33d4221c | 2103 | { |
6862ebbf | 2104 | return check_and_freshen(oid, 1); |
33d4221c JK |
2105 | } |
2106 | ||
6862ebbf | 2107 | static int freshen_packed_object(const struct object_id *oid) |
33d4221c JK |
2108 | { |
2109 | struct pack_entry e; | |
544443cb | 2110 | if (!find_pack_entry(the_repository, oid, &e)) |
ee1c6c34 | 2111 | return 0; |
a6131642 TB |
2112 | if (e.p->is_cruft) |
2113 | return 0; | |
ee1c6c34 JK |
2114 | if (e.p->freshened) |
2115 | return 1; | |
2116 | if (!freshen_file(e.p->pack_name)) | |
2117 | return 0; | |
2118 | e.p->freshened = 1; | |
2119 | return 1; | |
33d4221c JK |
2120 | } |
2121 | ||
2b6070ac HX |
2122 | int stream_loose_object(struct input_stream *in_stream, size_t len, |
2123 | struct object_id *oid) | |
2124 | { | |
2125 | int fd, ret, err = 0, flush = 0; | |
2126 | unsigned char compressed[4096]; | |
2127 | git_zstream stream; | |
2128 | git_hash_ctx c; | |
2129 | struct strbuf tmp_file = STRBUF_INIT; | |
2130 | struct strbuf filename = STRBUF_INIT; | |
2131 | int dirlen; | |
2132 | char hdr[MAX_HEADER_LEN]; | |
2133 | int hdrlen; | |
2134 | ||
2135 | if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT)) | |
2136 | prepare_loose_object_bulk_checkin(); | |
2137 | ||
2138 | /* Since oid is not determined, save tmp file to odb path. */ | |
2139 | strbuf_addf(&filename, "%s/", get_object_directory()); | |
2140 | hdrlen = format_object_header(hdr, sizeof(hdr), OBJ_BLOB, len); | |
2141 | ||
2142 | /* | |
2143 | * Common steps for write_loose_object and stream_loose_object to | |
2144 | * start writing loose objects: | |
2145 | * | |
2146 | * - Create tmpfile for the loose object. | |
2147 | * - Setup zlib stream for compression. | |
2148 | * - Start to feed header to zlib stream. | |
2149 | */ | |
2150 | fd = start_loose_object_common(&tmp_file, filename.buf, 0, | |
2151 | &stream, compressed, sizeof(compressed), | |
2152 | &c, hdr, hdrlen); | |
2153 | if (fd < 0) { | |
2154 | err = -1; | |
2155 | goto cleanup; | |
2156 | } | |
2157 | ||
2158 | /* Then the data itself.. */ | |
2159 | do { | |
2160 | unsigned char *in0 = stream.next_in; | |
2161 | ||
2162 | if (!stream.avail_in && !in_stream->is_finished) { | |
2163 | const void *in = in_stream->read(in_stream, &stream.avail_in); | |
2164 | stream.next_in = (void *)in; | |
2165 | in0 = (unsigned char *)in; | |
2166 | /* All data has been read. */ | |
2167 | if (in_stream->is_finished) | |
2168 | flush = 1; | |
2169 | } | |
2170 | ret = write_loose_object_common(&c, &stream, flush, in0, fd, | |
2171 | compressed, sizeof(compressed)); | |
2172 | /* | |
2173 | * Unlike write_loose_object(), we do not have the entire | |
2174 | * buffer. If we get Z_BUF_ERROR due to too few input bytes, | |
2175 | * then we'll replenish them in the next input_stream->read() | |
2176 | * call when we loop. | |
2177 | */ | |
2178 | } while (ret == Z_OK || ret == Z_BUF_ERROR); | |
2179 | ||
2180 | if (stream.total_in != len + hdrlen) | |
2181 | die(_("write stream object %ld != %"PRIuMAX), stream.total_in, | |
2182 | (uintmax_t)len + hdrlen); | |
2183 | ||
2184 | /* | |
2185 | * Common steps for write_loose_object and stream_loose_object to | |
2186 | * end writing loose oject: | |
2187 | * | |
2188 | * - End the compression of zlib stream. | |
2189 | * - Get the calculated oid. | |
2190 | */ | |
2191 | if (ret != Z_STREAM_END) | |
2192 | die(_("unable to stream deflate new object (%d)"), ret); | |
2193 | ret = end_loose_object_common(&c, &stream, oid); | |
2194 | if (ret != Z_OK) | |
2195 | die(_("deflateEnd on stream object failed (%d)"), ret); | |
2196 | close_loose_object(fd, tmp_file.buf); | |
2197 | ||
2198 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) { | |
2199 | unlink_or_warn(tmp_file.buf); | |
2200 | goto cleanup; | |
2201 | } | |
2202 | ||
2203 | loose_object_path(the_repository, &filename, oid); | |
2204 | ||
2205 | /* We finally know the object path, and create the missing dir. */ | |
2206 | dirlen = directory_size(filename.buf); | |
2207 | if (dirlen) { | |
2208 | struct strbuf dir = STRBUF_INIT; | |
2209 | strbuf_add(&dir, filename.buf, dirlen); | |
2210 | ||
2211 | if (mkdir_in_gitdir(dir.buf) && errno != EEXIST) { | |
2212 | err = error_errno(_("unable to create directory %s"), dir.buf); | |
2213 | strbuf_release(&dir); | |
2214 | goto cleanup; | |
2215 | } | |
2216 | strbuf_release(&dir); | |
2217 | } | |
2218 | ||
2219 | err = finalize_object_file(tmp_file.buf, filename.buf); | |
2220 | cleanup: | |
2221 | strbuf_release(&tmp_file); | |
2222 | strbuf_release(&filename); | |
2223 | return err; | |
2224 | } | |
2225 | ||
4ef91a2d | 2226 | int write_object_file_flags(const void *buf, unsigned long len, |
c80d226a | 2227 | enum object_type type, struct object_id *oid, |
4ef91a2d | 2228 | unsigned flags) |
bbac7311 | 2229 | { |
1af64f73 | 2230 | char hdr[MAX_HEADER_LEN]; |
ef1286d3 | 2231 | int hdrlen = sizeof(hdr); |
bbac7311 NP |
2232 | |
2233 | /* Normally if we have it in the pack then we do not bother writing | |
2234 | * it out into .git/objects/??/?{38} file. | |
2235 | */ | |
7ad5c44d MT |
2236 | write_object_file_prepare(the_hash_algo, buf, len, type, oid, hdr, |
2237 | &hdrlen); | |
6862ebbf | 2238 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) |
bbac7311 | 2239 | return 0; |
4ef91a2d | 2240 | return write_loose_object(oid, hdr, hdrlen, buf, len, 0, flags); |
bbac7311 NP |
2241 | } |
2242 | ||
0ff7b4f9 ÆAB |
2243 | int write_object_file_literally(const void *buf, unsigned long len, |
2244 | const char *type, struct object_id *oid, | |
2245 | unsigned flags) | |
0c3db67c ES |
2246 | { |
2247 | char *header; | |
2248 | int hdrlen, status = 0; | |
2249 | ||
2250 | /* type string, SP, %lu of the length plus NUL must fit this */ | |
1af64f73 | 2251 | hdrlen = strlen(type) + MAX_HEADER_LEN; |
ef1286d3 | 2252 | header = xmalloc(hdrlen); |
2bbb28a3 ÆAB |
2253 | write_object_file_prepare_literally(the_hash_algo, buf, len, type, |
2254 | oid, header, &hdrlen); | |
0c3db67c ES |
2255 | |
2256 | if (!(flags & HASH_WRITE_OBJECT)) | |
2257 | goto cleanup; | |
6862ebbf | 2258 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) |
0c3db67c | 2259 | goto cleanup; |
4ef91a2d | 2260 | status = write_loose_object(oid, header, hdrlen, buf, len, 0, 0); |
0c3db67c ES |
2261 | |
2262 | cleanup: | |
2263 | free(header); | |
2264 | return status; | |
2265 | } | |
2266 | ||
4bdb70a4 | 2267 | int force_object_loose(const struct object_id *oid, time_t mtime) |
bbac7311 | 2268 | { |
bbac7311 NP |
2269 | void *buf; |
2270 | unsigned long len; | |
2271 | enum object_type type; | |
1af64f73 | 2272 | char hdr[MAX_HEADER_LEN]; |
bbac7311 | 2273 | int hdrlen; |
1fb23e65 | 2274 | int ret; |
bbac7311 | 2275 | |
6862ebbf | 2276 | if (has_loose_object(oid)) |
bbac7311 | 2277 | return 0; |
cba595ab | 2278 | buf = read_object(the_repository, oid, &type, &len); |
bbac7311 | 2279 | if (!buf) |
2c319886 | 2280 | return error(_("cannot read object for %s"), oid_to_hex(oid)); |
b04cdea4 | 2281 | hdrlen = format_object_header(hdr, sizeof(hdr), type, len); |
4ef91a2d | 2282 | ret = write_loose_object(oid, hdr, hdrlen, buf, len, mtime, 0); |
1fb23e65 BS |
2283 | free(buf); |
2284 | ||
2285 | return ret; | |
bbac7311 NP |
2286 | } |
2287 | ||
1d8d9cb6 JT |
2288 | int has_object(struct repository *r, const struct object_id *oid, |
2289 | unsigned flags) | |
2290 | { | |
2291 | int quick = !(flags & HAS_OBJECT_RECHECK_PACKED); | |
2292 | unsigned object_info_flags = OBJECT_INFO_SKIP_FETCH_OBJECT | | |
2293 | (quick ? OBJECT_INFO_QUICK : 0); | |
2294 | ||
2295 | if (!startup_info->have_repository) | |
2296 | return 0; | |
2297 | return oid_object_info_extended(r, oid, NULL, object_info_flags) >= 0; | |
2298 | } | |
2299 | ||
cba595ab JH |
2300 | int repo_has_object_file_with_flags(struct repository *r, |
2301 | const struct object_id *oid, int flags) | |
8237b185 | 2302 | { |
3e8b7d3c JN |
2303 | if (!startup_info->have_repository) |
2304 | return 0; | |
9c8a294a | 2305 | return oid_object_info_extended(r, oid, NULL, flags) >= 0; |
b419aa25 | 2306 | } |
2307 | ||
9b45f499 SB |
2308 | int repo_has_object_file(struct repository *r, |
2309 | const struct object_id *oid) | |
b419aa25 | 2310 | { |
cba595ab | 2311 | return repo_has_object_file_with_flags(r, oid, 0); |
5827a035 JK |
2312 | } |
2313 | ||
c879daa2 NTND |
2314 | static void check_tree(const void *buf, size_t size) |
2315 | { | |
2316 | struct tree_desc desc; | |
2317 | struct name_entry entry; | |
2318 | ||
2319 | init_tree_desc(&desc, buf, size); | |
2320 | while (tree_entry(&desc, &entry)) | |
2321 | /* do nothing | |
2322 | * tree_entry() will die() on malformed entries */ | |
2323 | ; | |
2324 | } | |
2325 | ||
2326 | static void check_commit(const void *buf, size_t size) | |
2327 | { | |
2328 | struct commit c; | |
2329 | memset(&c, 0, sizeof(c)); | |
08f4f445 | 2330 | if (parse_commit_buffer(the_repository, &c, buf, size, 0)) |
259328b7 | 2331 | die(_("corrupt commit")); |
c879daa2 NTND |
2332 | } |
2333 | ||
2334 | static void check_tag(const void *buf, size_t size) | |
2335 | { | |
2336 | struct tag t; | |
2337 | memset(&t, 0, sizeof(t)); | |
0e740fed | 2338 | if (parse_tag_buffer(the_repository, &t, buf, size)) |
259328b7 | 2339 | die(_("corrupt tag")); |
c879daa2 NTND |
2340 | } |
2341 | ||
58bf2a4c NTND |
2342 | static int index_mem(struct index_state *istate, |
2343 | struct object_id *oid, void *buf, size_t size, | |
c4ce46fc JH |
2344 | enum object_type type, |
2345 | const char *path, unsigned flags) | |
e7332f96 | 2346 | { |
63e05f90 | 2347 | int ret = 0; |
bbea0dde | 2348 | int re_allocated = 0; |
c4ce46fc | 2349 | int write_object = flags & HASH_WRITE_OBJECT; |
74400e71 | 2350 | |
7672db20 | 2351 | if (!type) |
edaec3fb | 2352 | type = OBJ_BLOB; |
6c510bee LT |
2353 | |
2354 | /* | |
2355 | * Convert blobs to git internal format | |
2356 | */ | |
43df4f86 | 2357 | if ((type == OBJ_BLOB) && path) { |
f285a2d7 | 2358 | struct strbuf nbuf = STRBUF_INIT; |
58bf2a4c | 2359 | if (convert_to_git(istate, path, buf, size, &nbuf, |
8462ff43 | 2360 | get_conv_flags(flags))) { |
b315c5c0 | 2361 | buf = strbuf_detach(&nbuf, &size); |
6c510bee LT |
2362 | re_allocated = 1; |
2363 | } | |
2364 | } | |
c4ce46fc | 2365 | if (flags & HASH_FORMAT_CHECK) { |
c879daa2 NTND |
2366 | if (type == OBJ_TREE) |
2367 | check_tree(buf, size); | |
2368 | if (type == OBJ_COMMIT) | |
2369 | check_commit(buf, size); | |
2370 | if (type == OBJ_TAG) | |
2371 | check_tag(buf, size); | |
2372 | } | |
6c510bee | 2373 | |
7672db20 | 2374 | if (write_object) |
c80d226a | 2375 | ret = write_object_file(buf, size, type, oid); |
abdc3fc8 | 2376 | else |
44439c1c | 2377 | hash_object_file(the_hash_algo, buf, size, type, oid); |
43df4f86 | 2378 | if (re_allocated) |
6c510bee | 2379 | free(buf); |
43df4f86 DP |
2380 | return ret; |
2381 | } | |
2382 | ||
58bf2a4c NTND |
2383 | static int index_stream_convert_blob(struct index_state *istate, |
2384 | struct object_id *oid, | |
2385 | int fd, | |
2386 | const char *path, | |
2387 | unsigned flags) | |
9035d75a | 2388 | { |
63e05f90 | 2389 | int ret = 0; |
9035d75a SP |
2390 | const int write_object = flags & HASH_WRITE_OBJECT; |
2391 | struct strbuf sbuf = STRBUF_INIT; | |
2392 | ||
2393 | assert(path); | |
58bf2a4c | 2394 | assert(would_convert_to_git_filter_fd(istate, path)); |
9035d75a | 2395 | |
58bf2a4c | 2396 | convert_to_git_filter_fd(istate, path, fd, &sbuf, |
8462ff43 | 2397 | get_conv_flags(flags)); |
9035d75a SP |
2398 | |
2399 | if (write_object) | |
c80d226a | 2400 | ret = write_object_file(sbuf.buf, sbuf.len, OBJ_BLOB, |
a09c985e | 2401 | oid); |
9035d75a | 2402 | else |
44439c1c ÆAB |
2403 | hash_object_file(the_hash_algo, sbuf.buf, sbuf.len, OBJ_BLOB, |
2404 | oid); | |
9035d75a SP |
2405 | strbuf_release(&sbuf); |
2406 | return ret; | |
2407 | } | |
2408 | ||
58bf2a4c NTND |
2409 | static int index_pipe(struct index_state *istate, struct object_id *oid, |
2410 | int fd, enum object_type type, | |
7b41e1e1 JH |
2411 | const char *path, unsigned flags) |
2412 | { | |
2413 | struct strbuf sbuf = STRBUF_INIT; | |
2414 | int ret; | |
2415 | ||
2416 | if (strbuf_read(&sbuf, fd, 4096) >= 0) | |
58bf2a4c | 2417 | ret = index_mem(istate, oid, sbuf.buf, sbuf.len, type, path, flags); |
7b41e1e1 JH |
2418 | else |
2419 | ret = -1; | |
2420 | strbuf_release(&sbuf); | |
2421 | return ret; | |
2422 | } | |
2423 | ||
ea68b0ce DP |
2424 | #define SMALL_FILE_SIZE (32*1024) |
2425 | ||
58bf2a4c NTND |
2426 | static int index_core(struct index_state *istate, |
2427 | struct object_id *oid, int fd, size_t size, | |
7b41e1e1 JH |
2428 | enum object_type type, const char *path, |
2429 | unsigned flags) | |
43df4f86 DP |
2430 | { |
2431 | int ret; | |
43df4f86 | 2432 | |
7b41e1e1 | 2433 | if (!size) { |
58bf2a4c | 2434 | ret = index_mem(istate, oid, "", size, type, path, flags); |
ea68b0ce DP |
2435 | } else if (size <= SMALL_FILE_SIZE) { |
2436 | char *buf = xmalloc(size); | |
90dca671 JK |
2437 | ssize_t read_result = read_in_full(fd, buf, size); |
2438 | if (read_result < 0) | |
259328b7 | 2439 | ret = error_errno(_("read error while indexing %s"), |
90dca671 JK |
2440 | path ? path : "<unknown>"); |
2441 | else if (read_result != size) | |
259328b7 | 2442 | ret = error(_("short read while indexing %s"), |
90dca671 | 2443 | path ? path : "<unknown>"); |
ea68b0ce | 2444 | else |
58bf2a4c | 2445 | ret = index_mem(istate, oid, buf, size, type, path, flags); |
ea68b0ce | 2446 | free(buf); |
08bda208 | 2447 | } else { |
43df4f86 | 2448 | void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0); |
58bf2a4c | 2449 | ret = index_mem(istate, oid, buf, size, type, path, flags); |
aac17941 | 2450 | munmap(buf, size); |
08bda208 | 2451 | } |
7b41e1e1 JH |
2452 | return ret; |
2453 | } | |
2454 | ||
4dd1fbc7 | 2455 | /* |
568508e7 JH |
2456 | * This creates one packfile per large blob unless bulk-checkin |
2457 | * machinery is "plugged". | |
4dd1fbc7 JH |
2458 | * |
2459 | * This also bypasses the usual "convert-to-git" dance, and that is on | |
2460 | * purpose. We could write a streaming version of the converting | |
2461 | * functions and insert that before feeding the data to fast-import | |
4f22b101 JK |
2462 | * (or equivalent in-core API described above). However, that is |
2463 | * somewhat complicated, as we do not know the size of the filter | |
2464 | * result, which we need to know beforehand when writing a git object. | |
2465 | * Since the primary motivation for trying to stream from the working | |
2466 | * tree file and to avoid mmaping it in core is to deal with large | |
2467 | * binary blobs, they generally do not want to get any conversion, and | |
2468 | * callers should avoid this code path when filters are requested. | |
4dd1fbc7 | 2469 | */ |
7d5e1dc3 | 2470 | static int index_stream(struct object_id *oid, int fd, size_t size, |
4dd1fbc7 JH |
2471 | enum object_type type, const char *path, |
2472 | unsigned flags) | |
2473 | { | |
68ee6dfc | 2474 | return index_bulk_checkin(oid, fd, size, type, path, flags); |
4dd1fbc7 JH |
2475 | } |
2476 | ||
58bf2a4c NTND |
2477 | int index_fd(struct index_state *istate, struct object_id *oid, |
2478 | int fd, struct stat *st, | |
7b41e1e1 JH |
2479 | enum object_type type, const char *path, unsigned flags) |
2480 | { | |
2481 | int ret; | |
7b41e1e1 | 2482 | |
9079ab7c SP |
2483 | /* |
2484 | * Call xsize_t() only when needed to avoid potentially unnecessary | |
2485 | * die() for large files. | |
2486 | */ | |
58bf2a4c NTND |
2487 | if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(istate, path)) |
2488 | ret = index_stream_convert_blob(istate, oid, fd, path, flags); | |
9035d75a | 2489 | else if (!S_ISREG(st->st_mode)) |
58bf2a4c | 2490 | ret = index_pipe(istate, oid, fd, type, path, flags); |
9079ab7c | 2491 | else if (st->st_size <= big_file_threshold || type != OBJ_BLOB || |
58bf2a4c NTND |
2492 | (path && would_convert_to_git(istate, path))) |
2493 | ret = index_core(istate, oid, fd, xsize_t(st->st_size), | |
2494 | type, path, flags); | |
4dd1fbc7 | 2495 | else |
7d5e1dc3 | 2496 | ret = index_stream(oid, fd, xsize_t(st->st_size), type, path, |
9079ab7c | 2497 | flags); |
43df4f86 | 2498 | close(fd); |
aac17941 | 2499 | return ret; |
74400e71 | 2500 | } |
ec1fcc16 | 2501 | |
58bf2a4c NTND |
2502 | int index_path(struct index_state *istate, struct object_id *oid, |
2503 | const char *path, struct stat *st, unsigned flags) | |
ec1fcc16 JH |
2504 | { |
2505 | int fd; | |
b760d3aa | 2506 | struct strbuf sb = STRBUF_INIT; |
ea8e0297 | 2507 | int rc = 0; |
ec1fcc16 JH |
2508 | |
2509 | switch (st->st_mode & S_IFMT) { | |
2510 | case S_IFREG: | |
2511 | fd = open(path, O_RDONLY); | |
2512 | if (fd < 0) | |
7616c6ca | 2513 | return error_errno("open(\"%s\")", path); |
58bf2a4c | 2514 | if (index_fd(istate, oid, fd, st, OBJ_BLOB, path, flags) < 0) |
259328b7 | 2515 | return error(_("%s: failed to insert into database"), |
ec1fcc16 JH |
2516 | path); |
2517 | break; | |
2518 | case S_IFLNK: | |
7616c6ca NTND |
2519 | if (strbuf_readlink(&sb, path, st->st_size)) |
2520 | return error_errno("readlink(\"%s\")", path); | |
c4ce46fc | 2521 | if (!(flags & HASH_WRITE_OBJECT)) |
2dcde20e | 2522 | hash_object_file(the_hash_algo, sb.buf, sb.len, |
44439c1c | 2523 | OBJ_BLOB, oid); |
c80d226a | 2524 | else if (write_object_file(sb.buf, sb.len, OBJ_BLOB, oid)) |
259328b7 | 2525 | rc = error(_("%s: failed to insert into database"), path); |
b760d3aa | 2526 | strbuf_release(&sb); |
ec1fcc16 | 2527 | break; |
f35a6d3b | 2528 | case S_IFDIR: |
a98e6101 | 2529 | return resolve_gitlink_ref(path, "HEAD", oid); |
ec1fcc16 | 2530 | default: |
259328b7 | 2531 | return error(_("%s: unsupported file type"), path); |
ec1fcc16 | 2532 | } |
ea8e0297 | 2533 | return rc; |
ec1fcc16 | 2534 | } |
a69e5429 JH |
2535 | |
2536 | int read_pack_header(int fd, struct pack_header *header) | |
2537 | { | |
f48ecd38 | 2538 | if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header)) |
c697ad14 HO |
2539 | /* "eof before pack header was fully read" */ |
2540 | return PH_ERROR_EOF; | |
2541 | ||
a69e5429 JH |
2542 | if (header->hdr_signature != htonl(PACK_SIGNATURE)) |
2543 | /* "protocol error (pack signature mismatch detected)" */ | |
2544 | return PH_ERROR_PACK_SIGNATURE; | |
2545 | if (!pack_version_ok(header->hdr_version)) | |
2546 | /* "protocol error (pack version unsupported)" */ | |
2547 | return PH_ERROR_PROTOCOL; | |
2548 | return 0; | |
2549 | } | |
40d52ff7 | 2550 | |
e816caa0 | 2551 | void assert_oid_type(const struct object_id *oid, enum object_type expect) |
40d52ff7 | 2552 | { |
0df8e965 | 2553 | enum object_type type = oid_object_info(the_repository, oid, NULL); |
40d52ff7 | 2554 | if (type < 0) |
259328b7 | 2555 | die(_("%s is not a valid object"), oid_to_hex(oid)); |
40d52ff7 | 2556 | if (type != expect) |
259328b7 | 2557 | die(_("%s is not a valid '%s' object"), oid_to_hex(oid), |
debca9d2 | 2558 | type_name(expect)); |
40d52ff7 | 2559 | } |
27e1e22d | 2560 | |
70c49050 | 2561 | int for_each_file_in_obj_subdir(unsigned int subdir_nr, |
cc817ca3 RS |
2562 | struct strbuf *path, |
2563 | each_loose_object_fn obj_cb, | |
2564 | each_loose_cruft_fn cruft_cb, | |
2565 | each_loose_subdir_fn subdir_cb, | |
2566 | void *data) | |
27e1e22d | 2567 | { |
0375f472 RS |
2568 | size_t origlen, baselen; |
2569 | DIR *dir; | |
27e1e22d JK |
2570 | struct dirent *de; |
2571 | int r = 0; | |
62a24c89 | 2572 | struct object_id oid; |
27e1e22d | 2573 | |
70c49050 RS |
2574 | if (subdir_nr > 0xff) |
2575 | BUG("invalid loose object subdirectory: %x", subdir_nr); | |
2576 | ||
0375f472 RS |
2577 | origlen = path->len; |
2578 | strbuf_complete(path, '/'); | |
2579 | strbuf_addf(path, "%02x", subdir_nr); | |
0375f472 RS |
2580 | |
2581 | dir = opendir(path->buf); | |
27e1e22d | 2582 | if (!dir) { |
0375f472 | 2583 | if (errno != ENOENT) |
259328b7 | 2584 | r = error_errno(_("unable to open %s"), path->buf); |
0375f472 RS |
2585 | strbuf_setlen(path, origlen); |
2586 | return r; | |
27e1e22d JK |
2587 | } |
2588 | ||
62a24c89 | 2589 | oid.hash[0] = subdir_nr; |
163ee5e6 DS |
2590 | strbuf_addch(path, '/'); |
2591 | baselen = path->len; | |
62a24c89 | 2592 | |
b548f0f1 | 2593 | while ((de = readdir_skip_dot_and_dotdot(dir))) { |
163ee5e6 | 2594 | size_t namelen; |
27e1e22d | 2595 | |
163ee5e6 | 2596 | namelen = strlen(de->d_name); |
27e1e22d | 2597 | strbuf_setlen(path, baselen); |
163ee5e6 | 2598 | strbuf_add(path, de->d_name, namelen); |
94b5e093 | 2599 | if (namelen == the_hash_algo->hexsz - 2 && |
62a24c89 | 2600 | !hex_to_bytes(oid.hash + 1, de->d_name, |
94b5e093 | 2601 | the_hash_algo->rawsz - 1)) { |
5a6dce70 | 2602 | oid_set_algo(&oid, the_hash_algo); |
62a24c89 RS |
2603 | if (obj_cb) { |
2604 | r = obj_cb(&oid, path->buf, data); | |
2605 | if (r) | |
2606 | break; | |
27e1e22d | 2607 | } |
62a24c89 | 2608 | continue; |
27e1e22d JK |
2609 | } |
2610 | ||
2611 | if (cruft_cb) { | |
2612 | r = cruft_cb(de->d_name, path->buf, data); | |
2613 | if (r) | |
2614 | break; | |
2615 | } | |
2616 | } | |
094c7e63 | 2617 | closedir(dir); |
27e1e22d | 2618 | |
163ee5e6 | 2619 | strbuf_setlen(path, baselen - 1); |
27e1e22d JK |
2620 | if (!r && subdir_cb) |
2621 | r = subdir_cb(subdir_nr, path->buf, data); | |
2622 | ||
0375f472 RS |
2623 | strbuf_setlen(path, origlen); |
2624 | ||
27e1e22d JK |
2625 | return r; |
2626 | } | |
2627 | ||
e6f875e0 | 2628 | int for_each_loose_file_in_objdir_buf(struct strbuf *path, |
27e1e22d JK |
2629 | each_loose_object_fn obj_cb, |
2630 | each_loose_cruft_fn cruft_cb, | |
2631 | each_loose_subdir_fn subdir_cb, | |
2632 | void *data) | |
2633 | { | |
27e1e22d JK |
2634 | int r = 0; |
2635 | int i; | |
2636 | ||
27e1e22d | 2637 | for (i = 0; i < 256; i++) { |
e6f875e0 | 2638 | r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb, |
27e1e22d | 2639 | subdir_cb, data); |
27e1e22d JK |
2640 | if (r) |
2641 | break; | |
2642 | } | |
2643 | ||
e6f875e0 JK |
2644 | return r; |
2645 | } | |
2646 | ||
2647 | int for_each_loose_file_in_objdir(const char *path, | |
2648 | each_loose_object_fn obj_cb, | |
2649 | each_loose_cruft_fn cruft_cb, | |
2650 | each_loose_subdir_fn subdir_cb, | |
2651 | void *data) | |
2652 | { | |
2653 | struct strbuf buf = STRBUF_INIT; | |
2654 | int r; | |
2655 | ||
2656 | strbuf_addstr(&buf, path); | |
2657 | r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb, | |
2658 | subdir_cb, data); | |
27e1e22d | 2659 | strbuf_release(&buf); |
e6f875e0 | 2660 | |
27e1e22d JK |
2661 | return r; |
2662 | } | |
660c889e | 2663 | |
a7ff6f5a JK |
2664 | int for_each_loose_object(each_loose_object_fn cb, void *data, |
2665 | enum for_each_object_flags flags) | |
660c889e | 2666 | { |
f0eaf638 | 2667 | struct object_directory *odb; |
b0a42642 | 2668 | |
f0eaf638 JK |
2669 | prepare_alt_odb(the_repository); |
2670 | for (odb = the_repository->objects->odb; odb; odb = odb->next) { | |
2671 | int r = for_each_loose_file_in_objdir(odb->path, cb, NULL, | |
2672 | NULL, data); | |
2673 | if (r) | |
2674 | return r; | |
660c889e | 2675 | |
f0eaf638 JK |
2676 | if (flags & FOR_EACH_OBJECT_LOCAL_ONLY) |
2677 | break; | |
2678 | } | |
1385bb7b | 2679 | |
f0eaf638 | 2680 | return 0; |
660c889e JK |
2681 | } |
2682 | ||
3a2e0824 JK |
2683 | static int append_loose_object(const struct object_id *oid, const char *path, |
2684 | void *data) | |
660c889e | 2685 | { |
92d8ed8a | 2686 | oidtree_insert(data, oid); |
3a2e0824 JK |
2687 | return 0; |
2688 | } | |
660c889e | 2689 | |
92d8ed8a | 2690 | struct oidtree *odb_loose_cache(struct object_directory *odb, |
0000d654 RS |
2691 | const struct object_id *oid) |
2692 | { | |
2693 | int subdir_nr = oid->hash[0]; | |
3a2e0824 | 2694 | struct strbuf buf = STRBUF_INIT; |
33f379ee EW |
2695 | size_t word_bits = bitsizeof(odb->loose_objects_subdir_seen[0]); |
2696 | size_t word_index = subdir_nr / word_bits; | |
26de1fc0 | 2697 | size_t mask = (size_t)1u << (subdir_nr % word_bits); |
33f379ee | 2698 | uint32_t *bitmap; |
660c889e | 2699 | |
3a2e0824 | 2700 | if (subdir_nr < 0 || |
33f379ee | 2701 | subdir_nr >= bitsizeof(odb->loose_objects_subdir_seen)) |
3a2e0824 | 2702 | BUG("subdir_nr out of range"); |
1385bb7b | 2703 | |
33f379ee EW |
2704 | bitmap = &odb->loose_objects_subdir_seen[word_index]; |
2705 | if (*bitmap & mask) | |
92d8ed8a EW |
2706 | return odb->loose_objects_cache; |
2707 | if (!odb->loose_objects_cache) { | |
2708 | ALLOC_ARRAY(odb->loose_objects_cache, 1); | |
2709 | oidtree_init(odb->loose_objects_cache); | |
2710 | } | |
3a2e0824 JK |
2711 | strbuf_addstr(&buf, odb->path); |
2712 | for_each_file_in_obj_subdir(subdir_nr, &buf, | |
2713 | append_loose_object, | |
2714 | NULL, NULL, | |
92d8ed8a | 2715 | odb->loose_objects_cache); |
33f379ee | 2716 | *bitmap |= mask; |
7317aa71 | 2717 | strbuf_release(&buf); |
92d8ed8a | 2718 | return odb->loose_objects_cache; |
660c889e JK |
2719 | } |
2720 | ||
d4e19e51 RS |
2721 | void odb_clear_loose_cache(struct object_directory *odb) |
2722 | { | |
92d8ed8a EW |
2723 | oidtree_clear(odb->loose_objects_cache); |
2724 | FREE_AND_NULL(odb->loose_objects_cache); | |
d4e19e51 RS |
2725 | memset(&odb->loose_objects_subdir_seen, 0, |
2726 | sizeof(odb->loose_objects_subdir_seen)); | |
2727 | } | |
2728 | ||
00a7760e JK |
2729 | static int check_stream_oid(git_zstream *stream, |
2730 | const char *hdr, | |
2731 | unsigned long size, | |
2732 | const char *path, | |
2733 | const struct object_id *expected_oid) | |
f6371f92 | 2734 | { |
18e2588e | 2735 | git_hash_ctx c; |
00a7760e | 2736 | struct object_id real_oid; |
f6371f92 JK |
2737 | unsigned char buf[4096]; |
2738 | unsigned long total_read; | |
2739 | int status = Z_OK; | |
2740 | ||
18e2588e | 2741 | the_hash_algo->init_fn(&c); |
2742 | the_hash_algo->update_fn(&c, hdr, stream->total_out); | |
f6371f92 JK |
2743 | |
2744 | /* | |
2745 | * We already read some bytes into hdr, but the ones up to the NUL | |
2746 | * do not count against the object's content size. | |
2747 | */ | |
2748 | total_read = stream->total_out - strlen(hdr) - 1; | |
2749 | ||
2750 | /* | |
2751 | * This size comparison must be "<=" to read the final zlib packets; | |
00a7760e | 2752 | * see the comment in unpack_loose_rest for details. |
f6371f92 JK |
2753 | */ |
2754 | while (total_read <= size && | |
18ad13e5 JH |
2755 | (status == Z_OK || |
2756 | (status == Z_BUF_ERROR && !stream->avail_out))) { | |
f6371f92 JK |
2757 | stream->next_out = buf; |
2758 | stream->avail_out = sizeof(buf); | |
2759 | if (size - total_read < stream->avail_out) | |
2760 | stream->avail_out = size - total_read; | |
2761 | status = git_inflate(stream, Z_FINISH); | |
18e2588e | 2762 | the_hash_algo->update_fn(&c, buf, stream->next_out - buf); |
f6371f92 JK |
2763 | total_read += stream->next_out - buf; |
2764 | } | |
2765 | git_inflate_end(stream); | |
2766 | ||
2767 | if (status != Z_STREAM_END) { | |
00a7760e | 2768 | error(_("corrupt loose object '%s'"), oid_to_hex(expected_oid)); |
f6371f92 JK |
2769 | return -1; |
2770 | } | |
cce044df | 2771 | if (stream->avail_in) { |
259328b7 | 2772 | error(_("garbage at end of loose object '%s'"), |
00a7760e | 2773 | oid_to_hex(expected_oid)); |
cce044df JK |
2774 | return -1; |
2775 | } | |
f6371f92 | 2776 | |
5951bf46 | 2777 | the_hash_algo->final_oid_fn(&real_oid, &c); |
00a7760e | 2778 | if (!oideq(expected_oid, &real_oid)) { |
01f8d594 | 2779 | error(_("hash mismatch for %s (expected %s)"), path, |
00a7760e | 2780 | oid_to_hex(expected_oid)); |
f6371f92 JK |
2781 | return -1; |
2782 | } | |
2783 | ||
2784 | return 0; | |
2785 | } | |
2786 | ||
2787 | int read_loose_object(const char *path, | |
d61d87bd | 2788 | const struct object_id *expected_oid, |
96e41f58 | 2789 | struct object_id *real_oid, |
31deb28f ÆAB |
2790 | void **contents, |
2791 | struct object_info *oi) | |
f6371f92 JK |
2792 | { |
2793 | int ret = -1; | |
f6371f92 JK |
2794 | void *map = NULL; |
2795 | unsigned long mapsize; | |
2796 | git_zstream stream; | |
1af64f73 | 2797 | char hdr[MAX_HEADER_LEN]; |
31deb28f | 2798 | unsigned long *size = oi->sizep; |
f6371f92 | 2799 | |
514c5fdd | 2800 | map = map_loose_object_1(the_repository, path, NULL, &mapsize); |
f6371f92 | 2801 | if (!map) { |
259328b7 | 2802 | error_errno(_("unable to mmap %s"), path); |
f6371f92 JK |
2803 | goto out; |
2804 | } | |
2805 | ||
8a50571a DS |
2806 | if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr), |
2807 | NULL) != ULHR_OK) { | |
259328b7 | 2808 | error(_("unable to unpack header of %s"), path); |
f6371f92 JK |
2809 | goto out; |
2810 | } | |
2811 | ||
31deb28f | 2812 | if (parse_loose_header(hdr, oi) < 0) { |
259328b7 | 2813 | error(_("unable to parse header of %s"), path); |
f6371f92 JK |
2814 | git_inflate_end(&stream); |
2815 | goto out; | |
2816 | } | |
2817 | ||
31deb28f | 2818 | if (*oi->typep == OBJ_BLOB && *size > big_file_threshold) { |
00a7760e | 2819 | if (check_stream_oid(&stream, hdr, *size, path, expected_oid) < 0) |
f6371f92 JK |
2820 | goto out; |
2821 | } else { | |
00a7760e | 2822 | *contents = unpack_loose_rest(&stream, hdr, *size, expected_oid); |
f6371f92 | 2823 | if (!*contents) { |
259328b7 | 2824 | error(_("unable to unpack contents of %s"), path); |
f6371f92 JK |
2825 | git_inflate_end(&stream); |
2826 | goto out; | |
2827 | } | |
44439c1c | 2828 | hash_object_file_literally(the_repository->hash_algo, |
16235e3b | 2829 | *contents, *size, |
44439c1c | 2830 | oi->type_name->buf, real_oid); |
0f156dbb | 2831 | if (!oideq(expected_oid, real_oid)) |
f6371f92 | 2832 | goto out; |
f6371f92 JK |
2833 | } |
2834 | ||
2835 | ret = 0; /* everything checks out */ | |
2836 | ||
2837 | out: | |
2838 | if (map) | |
2839 | munmap(map, mapsize); | |
f6371f92 JK |
2840 | return ret; |
2841 | } |