]>
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 | */ | |
5e3f94df | 9 | #include "git-compat-util.h" |
0b027f6c | 10 | #include "abspath.h" |
b2141fc1 | 11 | #include "config.h" |
73359a9b | 12 | #include "convert.h" |
32a8f510 | 13 | #include "environment.h" |
f394e093 | 14 | #include "gettext.h" |
41771fa4 | 15 | #include "hex.h" |
6eac50d8 | 16 | #include "string-list.h" |
697cc8ef | 17 | #include "lockfile.h" |
1f688557 | 18 | #include "delta.h" |
a733cb60 | 19 | #include "pack.h" |
8e440259 PE |
20 | #include "blob.h" |
21 | #include "commit.h" | |
4dd1fbc7 | 22 | #include "run-command.h" |
8e440259 PE |
23 | #include "tag.h" |
24 | #include "tree.h" | |
c879daa2 | 25 | #include "tree-walk.h" |
f35a6d3b | 26 | #include "refs.h" |
70f5d5d3 | 27 | #include "pack-revindex.h" |
bc626927 | 28 | #include "hash-lookup.h" |
568508e7 | 29 | #include "bulk-checkin.h" |
031dc927 | 30 | #include "repository.h" |
47f351e9 | 31 | #include "replace-object.h" |
090ea126 | 32 | #include "streaming.h" |
543c5caa | 33 | #include "dir.h" |
12d95ef6 | 34 | #include "list.h" |
c4c6effa | 35 | #include "mergesort.h" |
cf3c6352 | 36 | #include "quote.h" |
4f39cd82 | 37 | #include "packfile.h" |
87bed179 | 38 | #include "object-file.h" |
90c62155 | 39 | #include "object-store.h" |
e3d2f20e | 40 | #include "oidtree.h" |
c339932b | 41 | #include "path.h" |
b14ed5ad | 42 | #include "promisor-remote.h" |
e38da487 | 43 | #include "setup.h" |
a35e03de | 44 | #include "submodule.h" |
69bbbe48 | 45 | #include "fsck.h" |
e05db0fd | 46 | |
1af64f73 | 47 | /* The maximum size for an object header. */ |
48 | #define MAX_HEADER_LEN 32 | |
49 | ||
e1ccd7e2 | 50 | |
51 | #define EMPTY_TREE_SHA1_BIN_LITERAL \ | |
52 | "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60" \ | |
53 | "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04" | |
13eeedb5 | 54 | #define EMPTY_TREE_SHA256_BIN_LITERAL \ |
55 | "\x6e\xf1\x9b\x41\x22\x5c\x53\x69\xf1\xc1" \ | |
56 | "\x04\xd4\x5d\x8d\x85\xef\xa9\xb0\x57\xb5" \ | |
57 | "\x3b\x14\xb4\xb9\xb9\x39\xdd\x74\xde\xcc" \ | |
58 | "\x53\x21" | |
e1ccd7e2 | 59 | |
60 | #define EMPTY_BLOB_SHA1_BIN_LITERAL \ | |
61 | "\xe6\x9d\xe2\x9b\xb2\xd1\xd6\x43\x4b\x8b" \ | |
62 | "\x29\xae\x77\x5a\xd8\xc2\xe4\x8c\x53\x91" | |
13eeedb5 | 63 | #define EMPTY_BLOB_SHA256_BIN_LITERAL \ |
64 | "\x47\x3a\x0f\x4c\x3b\xe8\xa9\x36\x81\xa2" \ | |
65 | "\x67\xe3\xb1\xe9\xa7\xdc\xda\x11\x85\x43" \ | |
66 | "\x6f\xe1\x41\xf7\x74\x91\x20\xa3\x03\x72" \ | |
67 | "\x18\x13" | |
e1ccd7e2 | 68 | |
e1ccd7e2 | 69 | static const struct object_id empty_tree_oid = { |
5a6dce70 | 70 | .hash = EMPTY_TREE_SHA1_BIN_LITERAL, |
71 | .algo = GIT_HASH_SHA1, | |
8576fde6 | 72 | }; |
e1ccd7e2 | 73 | static const struct object_id empty_blob_oid = { |
5a6dce70 | 74 | .hash = EMPTY_BLOB_SHA1_BIN_LITERAL, |
75 | .algo = GIT_HASH_SHA1, | |
8576fde6 | 76 | }; |
14228447 | 77 | static const struct object_id null_oid_sha1 = { |
78 | .hash = {0}, | |
79 | .algo = GIT_HASH_SHA1, | |
80 | }; | |
13eeedb5 | 81 | static const struct object_id empty_tree_oid_sha256 = { |
5a6dce70 | 82 | .hash = EMPTY_TREE_SHA256_BIN_LITERAL, |
83 | .algo = GIT_HASH_SHA256, | |
13eeedb5 | 84 | }; |
85 | static const struct object_id empty_blob_oid_sha256 = { | |
5a6dce70 | 86 | .hash = EMPTY_BLOB_SHA256_BIN_LITERAL, |
87 | .algo = GIT_HASH_SHA256, | |
13eeedb5 | 88 | }; |
14228447 | 89 | static const struct object_id null_oid_sha256 = { |
90 | .hash = {0}, | |
91 | .algo = GIT_HASH_SHA256, | |
92 | }; | |
88cd621d | 93 | |
ac73cedf | 94 | static void git_hash_sha1_init(git_hash_ctx *ctx) |
f50e766b | 95 | { |
ac73cedf | 96 | git_SHA1_Init(&ctx->sha1); |
f50e766b | 97 | } |
98 | ||
768e30ea | 99 | static void git_hash_sha1_clone(git_hash_ctx *dst, const git_hash_ctx *src) |
100 | { | |
101 | git_SHA1_Clone(&dst->sha1, &src->sha1); | |
102 | } | |
103 | ||
ac73cedf | 104 | static void git_hash_sha1_update(git_hash_ctx *ctx, const void *data, size_t len) |
f50e766b | 105 | { |
ac73cedf | 106 | git_SHA1_Update(&ctx->sha1, data, len); |
f50e766b | 107 | } |
108 | ||
ac73cedf | 109 | static void git_hash_sha1_final(unsigned char *hash, git_hash_ctx *ctx) |
f50e766b | 110 | { |
ac73cedf | 111 | git_SHA1_Final(hash, &ctx->sha1); |
f50e766b | 112 | } |
113 | ||
ab795f0d | 114 | static void git_hash_sha1_final_oid(struct object_id *oid, git_hash_ctx *ctx) |
115 | { | |
116 | git_SHA1_Final(oid->hash, &ctx->sha1); | |
117 | memset(oid->hash + GIT_SHA1_RAWSZ, 0, GIT_MAX_RAWSZ - GIT_SHA1_RAWSZ); | |
5a6dce70 | 118 | oid->algo = GIT_HASH_SHA1; |
ab795f0d | 119 | } |
120 | ||
13eeedb5 | 121 | |
122 | static void git_hash_sha256_init(git_hash_ctx *ctx) | |
123 | { | |
124 | git_SHA256_Init(&ctx->sha256); | |
125 | } | |
126 | ||
768e30ea | 127 | static void git_hash_sha256_clone(git_hash_ctx *dst, const git_hash_ctx *src) |
128 | { | |
129 | git_SHA256_Clone(&dst->sha256, &src->sha256); | |
130 | } | |
131 | ||
13eeedb5 | 132 | static void git_hash_sha256_update(git_hash_ctx *ctx, const void *data, size_t len) |
133 | { | |
134 | git_SHA256_Update(&ctx->sha256, data, len); | |
135 | } | |
136 | ||
137 | static void git_hash_sha256_final(unsigned char *hash, git_hash_ctx *ctx) | |
138 | { | |
139 | git_SHA256_Final(hash, &ctx->sha256); | |
140 | } | |
141 | ||
ab795f0d | 142 | static void git_hash_sha256_final_oid(struct object_id *oid, git_hash_ctx *ctx) |
143 | { | |
144 | git_SHA256_Final(oid->hash, &ctx->sha256); | |
145 | /* | |
146 | * This currently does nothing, so the compiler should optimize it out, | |
147 | * but keep it in case we extend the hash size again. | |
148 | */ | |
149 | memset(oid->hash + GIT_SHA256_RAWSZ, 0, GIT_MAX_RAWSZ - GIT_SHA256_RAWSZ); | |
5a6dce70 | 150 | oid->algo = GIT_HASH_SHA256; |
ab795f0d | 151 | } |
152 | ||
9eb6cdad | 153 | static void git_hash_unknown_init(git_hash_ctx *ctx UNUSED) |
f50e766b | 154 | { |
1a07e59c | 155 | BUG("trying to init unknown hash"); |
f50e766b | 156 | } |
157 | ||
9eb6cdad JK |
158 | static void git_hash_unknown_clone(git_hash_ctx *dst UNUSED, |
159 | const git_hash_ctx *src UNUSED) | |
768e30ea | 160 | { |
161 | BUG("trying to clone unknown hash"); | |
162 | } | |
163 | ||
9eb6cdad JK |
164 | static void git_hash_unknown_update(git_hash_ctx *ctx UNUSED, |
165 | const void *data UNUSED, | |
166 | size_t len UNUSED) | |
f50e766b | 167 | { |
1a07e59c | 168 | BUG("trying to update unknown hash"); |
f50e766b | 169 | } |
170 | ||
9eb6cdad JK |
171 | static void git_hash_unknown_final(unsigned char *hash UNUSED, |
172 | git_hash_ctx *ctx UNUSED) | |
f50e766b | 173 | { |
1a07e59c | 174 | BUG("trying to finalize unknown hash"); |
f50e766b | 175 | } |
176 | ||
9eb6cdad JK |
177 | static void git_hash_unknown_final_oid(struct object_id *oid UNUSED, |
178 | git_hash_ctx *ctx UNUSED) | |
ab795f0d | 179 | { |
180 | BUG("trying to finalize unknown hash"); | |
181 | } | |
182 | ||
f50e766b | 183 | const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = { |
184 | { | |
0cb9872e ÆAB |
185 | .name = NULL, |
186 | .format_id = 0x00000000, | |
187 | .rawsz = 0, | |
188 | .hexsz = 0, | |
189 | .blksz = 0, | |
190 | .init_fn = git_hash_unknown_init, | |
191 | .clone_fn = git_hash_unknown_clone, | |
192 | .update_fn = git_hash_unknown_update, | |
193 | .final_fn = git_hash_unknown_final, | |
194 | .final_oid_fn = git_hash_unknown_final_oid, | |
195 | .empty_tree = NULL, | |
196 | .empty_blob = NULL, | |
197 | .null_oid = NULL, | |
f50e766b | 198 | }, |
199 | { | |
0cb9872e ÆAB |
200 | .name = "sha1", |
201 | .format_id = GIT_SHA1_FORMAT_ID, | |
202 | .rawsz = GIT_SHA1_RAWSZ, | |
203 | .hexsz = GIT_SHA1_HEXSZ, | |
204 | .blksz = GIT_SHA1_BLKSZ, | |
205 | .init_fn = git_hash_sha1_init, | |
206 | .clone_fn = git_hash_sha1_clone, | |
207 | .update_fn = git_hash_sha1_update, | |
208 | .final_fn = git_hash_sha1_final, | |
209 | .final_oid_fn = git_hash_sha1_final_oid, | |
210 | .empty_tree = &empty_tree_oid, | |
211 | .empty_blob = &empty_blob_oid, | |
212 | .null_oid = &null_oid_sha1, | |
f50e766b | 213 | }, |
13eeedb5 | 214 | { |
0cb9872e ÆAB |
215 | .name = "sha256", |
216 | .format_id = GIT_SHA256_FORMAT_ID, | |
217 | .rawsz = GIT_SHA256_RAWSZ, | |
218 | .hexsz = GIT_SHA256_HEXSZ, | |
219 | .blksz = GIT_SHA256_BLKSZ, | |
220 | .init_fn = git_hash_sha256_init, | |
221 | .clone_fn = git_hash_sha256_clone, | |
222 | .update_fn = git_hash_sha256_update, | |
223 | .final_fn = git_hash_sha256_final, | |
224 | .final_oid_fn = git_hash_sha256_final_oid, | |
225 | .empty_tree = &empty_tree_oid_sha256, | |
226 | .empty_blob = &empty_blob_oid_sha256, | |
227 | .null_oid = &null_oid_sha256, | |
13eeedb5 | 228 | } |
f50e766b | 229 | }; |
230 | ||
14228447 | 231 | const struct object_id *null_oid(void) |
232 | { | |
233 | return the_hash_algo->null_oid; | |
234 | } | |
235 | ||
d8a92ced | 236 | const char *empty_tree_oid_hex(void) |
237 | { | |
238 | static char buf[GIT_MAX_HEXSZ + 1]; | |
239 | return oid_to_hex_r(buf, the_hash_algo->empty_tree); | |
240 | } | |
241 | ||
242 | const char *empty_blob_oid_hex(void) | |
243 | { | |
244 | static char buf[GIT_MAX_HEXSZ + 1]; | |
245 | return oid_to_hex_r(buf, the_hash_algo->empty_blob); | |
246 | } | |
247 | ||
2f90b9d9 | 248 | int hash_algo_by_name(const char *name) |
249 | { | |
250 | int i; | |
251 | if (!name) | |
252 | return GIT_HASH_UNKNOWN; | |
253 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
254 | if (!strcmp(name, hash_algos[i].name)) | |
255 | return i; | |
256 | return GIT_HASH_UNKNOWN; | |
257 | } | |
258 | ||
259 | int hash_algo_by_id(uint32_t format_id) | |
260 | { | |
261 | int i; | |
262 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
263 | if (format_id == hash_algos[i].format_id) | |
264 | return i; | |
265 | return GIT_HASH_UNKNOWN; | |
266 | } | |
267 | ||
95399788 | 268 | int hash_algo_by_length(int len) |
269 | { | |
270 | int i; | |
271 | for (i = 1; i < GIT_HASH_NALGOS; i++) | |
272 | if (len == hash_algos[i].rawsz) | |
273 | return i; | |
274 | return GIT_HASH_UNKNOWN; | |
275 | } | |
2f90b9d9 | 276 | |
c597ba80 NTND |
277 | /* |
278 | * This is meant to hold a *small* number of objects that you would | |
c7c33f50 | 279 | * want repo_read_object_file() to be able to return, but yet you do not want |
c597ba80 NTND |
280 | * to write them into the object store (e.g. a browse-only |
281 | * application). | |
282 | */ | |
283 | static struct cached_object { | |
62ba93ea | 284 | struct object_id oid; |
c597ba80 NTND |
285 | enum object_type type; |
286 | void *buf; | |
287 | unsigned long size; | |
288 | } *cached_objects; | |
289 | static int cached_object_nr, cached_object_alloc; | |
290 | ||
291 | static struct cached_object empty_tree = { | |
50103649 ÆAB |
292 | .oid = { |
293 | .hash = EMPTY_TREE_SHA1_BIN_LITERAL, | |
294 | }, | |
295 | .type = OBJ_TREE, | |
296 | .buf = "", | |
c597ba80 NTND |
297 | }; |
298 | ||
62ba93ea | 299 | static struct cached_object *find_cached_object(const struct object_id *oid) |
c597ba80 NTND |
300 | { |
301 | int i; | |
302 | struct cached_object *co = cached_objects; | |
303 | ||
304 | for (i = 0; i < cached_object_nr; i++, co++) { | |
4a7e27e9 | 305 | if (oideq(&co->oid, oid)) |
c597ba80 NTND |
306 | return co; |
307 | } | |
4a7e27e9 | 308 | if (oideq(oid, the_hash_algo->empty_tree)) |
c597ba80 NTND |
309 | return &empty_tree; |
310 | return NULL; | |
311 | } | |
312 | ||
9472935d | 313 | |
8462ff43 | 314 | static int get_conv_flags(unsigned flags) |
9472935d TB |
315 | { |
316 | if (flags & HASH_RENORMALIZE) | |
8462ff43 | 317 | return CONV_EOL_RENORMALIZE; |
9472935d | 318 | else if (flags & HASH_WRITE_OBJECT) |
107642fe | 319 | return global_conv_flags_eol | CONV_WRITE_OBJECT; |
9472935d | 320 | else |
8462ff43 | 321 | return 0; |
9472935d TB |
322 | } |
323 | ||
324 | ||
90a6464b JH |
325 | int mkdir_in_gitdir(const char *path) |
326 | { | |
327 | if (mkdir(path, 0777)) { | |
328 | int saved_errno = errno; | |
329 | struct stat st; | |
330 | struct strbuf sb = STRBUF_INIT; | |
331 | ||
332 | if (errno != EEXIST) | |
333 | return -1; | |
334 | /* | |
335 | * Are we looking at a path in a symlinked worktree | |
336 | * whose original repository does not yet have it? | |
337 | * e.g. .git/rr-cache pointing at its original | |
338 | * repository in which the user hasn't performed any | |
339 | * conflict resolution yet? | |
340 | */ | |
341 | if (lstat(path, &st) || !S_ISLNK(st.st_mode) || | |
342 | strbuf_readlink(&sb, path, st.st_size) || | |
343 | !is_absolute_path(sb.buf) || | |
344 | mkdir(sb.buf, 0777)) { | |
345 | strbuf_release(&sb); | |
346 | errno = saved_errno; | |
347 | return -1; | |
348 | } | |
349 | strbuf_release(&sb); | |
350 | } | |
351 | return adjust_shared_perm(path); | |
352 | } | |
353 | ||
eb3c027e | 354 | static enum scld_error safe_create_leading_directories_1(char *path, int share) |
b2cb9425 | 355 | { |
26c8ae2a | 356 | char *next_component = path + offset_1st_component(path); |
0be0521b | 357 | enum scld_error ret = SCLD_OK; |
67d42212 | 358 | |
0be0521b | 359 | while (ret == SCLD_OK && next_component) { |
f0502332 | 360 | struct stat st; |
0f527403 | 361 | char *slash = next_component, slash_character; |
f0502332 | 362 | |
0f527403 MH |
363 | while (*slash && !is_dir_sep(*slash)) |
364 | slash++; | |
365 | ||
366 | if (!*slash) | |
b2cb9425 | 367 | break; |
bf10cf70 | 368 | |
26c8ae2a | 369 | next_component = slash + 1; |
0f527403 | 370 | while (is_dir_sep(*next_component)) |
bf10cf70 | 371 | next_component++; |
26c8ae2a | 372 | if (!*next_component) |
5f0bdf50 | 373 | break; |
831651fd | 374 | |
0f527403 | 375 | slash_character = *slash; |
831651fd | 376 | *slash = '\0'; |
67d42212 JR |
377 | if (!stat(path, &st)) { |
378 | /* path exists */ | |
204a047f MH |
379 | if (!S_ISDIR(st.st_mode)) { |
380 | errno = ENOTDIR; | |
0be0521b | 381 | ret = SCLD_EXISTS; |
204a047f | 382 | } |
53a39721 | 383 | } else if (mkdir(path, 0777)) { |
928734d9 | 384 | if (errno == EEXIST && |
9e6f885d | 385 | !stat(path, &st) && S_ISDIR(st.st_mode)) |
928734d9 | 386 | ; /* somebody created it since we checked */ |
18d37e86 MH |
387 | else if (errno == ENOENT) |
388 | /* | |
389 | * Either mkdir() failed because | |
390 | * somebody just pruned the containing | |
391 | * directory, or stat() failed because | |
392 | * the file that was in our way was | |
393 | * just removed. Either way, inform | |
394 | * the caller that it might be worth | |
395 | * trying again: | |
396 | */ | |
397 | ret = SCLD_VANISHED; | |
9e6f885d | 398 | else |
0be0521b | 399 | ret = SCLD_FAILED; |
eb3c027e | 400 | } else if (share && adjust_shared_perm(path)) { |
0be0521b | 401 | ret = SCLD_PERMS; |
457f06d6 | 402 | } |
0f527403 | 403 | *slash = slash_character; |
b2cb9425 | 404 | } |
9e6f885d | 405 | return ret; |
b2cb9425 | 406 | } |
723c31fe | 407 | |
eb3c027e MT |
408 | enum scld_error safe_create_leading_directories(char *path) |
409 | { | |
410 | return safe_create_leading_directories_1(path, 1); | |
411 | } | |
412 | ||
413 | enum scld_error safe_create_leading_directories_no_share(char *path) | |
414 | { | |
415 | return safe_create_leading_directories_1(path, 0); | |
416 | } | |
417 | ||
0be0521b | 418 | enum scld_error safe_create_leading_directories_const(const char *path) |
8e21d63b | 419 | { |
02944307 | 420 | int save_errno; |
8e21d63b JK |
421 | /* path points to cache entries, so xstrdup before messing with it */ |
422 | char *buf = xstrdup(path); | |
0be0521b | 423 | enum scld_error result = safe_create_leading_directories(buf); |
02944307 MH |
424 | |
425 | save_errno = errno; | |
8e21d63b | 426 | free(buf); |
02944307 | 427 | errno = save_errno; |
8e21d63b JK |
428 | return result; |
429 | } | |
430 | ||
514c5fdd | 431 | static void fill_loose_path(struct strbuf *buf, const struct object_id *oid) |
ace1534d JH |
432 | { |
433 | int i; | |
94b5e093 | 434 | for (i = 0; i < the_hash_algo->rawsz; i++) { |
ace1534d | 435 | static char hex[] = "0123456789abcdef"; |
514c5fdd | 436 | unsigned int val = oid->hash[i]; |
f7b7774f JK |
437 | strbuf_addch(buf, hex[val >> 4]); |
438 | strbuf_addch(buf, hex[val & 0xf]); | |
afbba2f0 | 439 | if (!i) |
f7b7774f | 440 | strbuf_addch(buf, '/'); |
ace1534d JH |
441 | } |
442 | } | |
443 | ||
f0eaf638 JK |
444 | static const char *odb_loose_path(struct object_directory *odb, |
445 | struct strbuf *buf, | |
514c5fdd | 446 | const struct object_id *oid) |
0fcfd160 | 447 | { |
b69fb867 | 448 | strbuf_reset(buf); |
f0eaf638 | 449 | strbuf_addstr(buf, odb->path); |
34498471 | 450 | strbuf_addch(buf, '/'); |
514c5fdd | 451 | fill_loose_path(buf, oid); |
f3f043a1 | 452 | return buf->buf; |
0fcfd160 LT |
453 | } |
454 | ||
f3f043a1 | 455 | const char *loose_object_path(struct repository *r, struct strbuf *buf, |
514c5fdd | 456 | const struct object_id *oid) |
29ec6af2 | 457 | { |
514c5fdd | 458 | return odb_loose_path(r->objects->odb, buf, oid); |
0fcfd160 LT |
459 | } |
460 | ||
4ea82473 JK |
461 | /* |
462 | * Return non-zero iff the path is usable as an alternate object database. | |
463 | */ | |
13313fc3 SB |
464 | static int alt_odb_usable(struct raw_object_store *o, |
465 | struct strbuf *path, | |
cf2dc1c2 | 466 | const char *normalized_objdir, khiter_t *pos) |
4ea82473 | 467 | { |
cf2dc1c2 | 468 | int r; |
4ea82473 JK |
469 | |
470 | /* Detect cases where alternate disappeared */ | |
471 | if (!is_directory(path->buf)) { | |
259328b7 NTND |
472 | error(_("object directory %s does not exist; " |
473 | "check .git/objects/info/alternates"), | |
4ea82473 JK |
474 | path->buf); |
475 | return 0; | |
476 | } | |
477 | ||
478 | /* | |
479 | * Prevent the common mistake of listing the same | |
480 | * thing twice, or object directory itself. | |
481 | */ | |
cf2dc1c2 EW |
482 | if (!o->odb_by_path) { |
483 | khiter_t p; | |
484 | ||
485 | o->odb_by_path = kh_init_odb_path_map(); | |
486 | assert(!o->odb->next); | |
487 | p = kh_put_odb_path_map(o->odb_by_path, o->odb->path, &r); | |
488 | assert(r == 1); /* never used */ | |
489 | kh_value(o->odb_by_path, p) = o->odb; | |
4ea82473 | 490 | } |
cf2dc1c2 | 491 | if (fspatheq(path->buf, normalized_objdir)) |
4ea82473 | 492 | return 0; |
cf2dc1c2 EW |
493 | *pos = kh_put_odb_path_map(o->odb_by_path, path->buf, &r); |
494 | /* r: 0 = exists, 1 = never used, 2 = deleted */ | |
495 | return r == 0 ? 0 : 1; | |
4ea82473 JK |
496 | } |
497 | ||
ddd5d056 JH |
498 | /* |
499 | * Prepare alternate object database registry. | |
d5a63b99 JH |
500 | * |
501 | * The variable alt_odb_list points at the list of struct | |
263db403 | 502 | * object_directory. The elements on this list come from |
d5a63b99 JH |
503 | * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT |
504 | * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates, | |
1494e038 JH |
505 | * whose contents is similar to that environment variable but can be |
506 | * LF separated. Its base points at a statically allocated buffer that | |
d5a63b99 JH |
507 | * contains "/the/directory/corresponding/to/.git/objects/...", while |
508 | * its name points just after the slash at the end of ".git/objects/" | |
e5afd444 MÅ |
509 | * in the example above, and has enough space to hold all hex characters |
510 | * of the object ID, an extra slash for the first level indirection, and | |
511 | * the terminating NUL. | |
ddd5d056 | 512 | */ |
77f012e8 SB |
513 | static void read_info_alternates(struct repository *r, |
514 | const char *relative_base, | |
515 | int depth); | |
407532f8 | 516 | static int link_alt_odb_entry(struct repository *r, const struct strbuf *entry, |
cfc62fc9 | 517 | const char *relative_base, int depth, const char *normalized_objdir) |
ace1534d | 518 | { |
263db403 | 519 | struct object_directory *ent; |
5bdf0a84 | 520 | struct strbuf pathbuf = STRBUF_INIT; |
199337d6 | 521 | struct strbuf tmp = STRBUF_INIT; |
cf2dc1c2 | 522 | khiter_t pos; |
199337d6 | 523 | int ret = -1; |
d5a63b99 | 524 | |
407532f8 | 525 | if (!is_absolute_path(entry->buf) && relative_base) { |
4ac9006f | 526 | strbuf_realpath(&pathbuf, relative_base, 1); |
5bdf0a84 | 527 | strbuf_addch(&pathbuf, '/'); |
c2f493a4 | 528 | } |
407532f8 | 529 | strbuf_addbuf(&pathbuf, entry); |
c2f493a4 | 530 | |
199337d6 | 531 | if (!strbuf_realpath(&tmp, pathbuf.buf, 0)) { |
259328b7 | 532 | error(_("unable to normalize alternate object path: %s"), |
670c359d | 533 | pathbuf.buf); |
199337d6 | 534 | goto error; |
670c359d | 535 | } |
199337d6 | 536 | strbuf_swap(&pathbuf, &tmp); |
5bdf0a84 HW |
537 | |
538 | /* | |
539 | * The trailing slash after the directory name is given by | |
540 | * this function at the end. Remove duplicates. | |
541 | */ | |
4ea82473 JK |
542 | while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/') |
543 | strbuf_setlen(&pathbuf, pathbuf.len - 1); | |
5bdf0a84 | 544 | |
199337d6 GC |
545 | if (!alt_odb_usable(r->objects, &pathbuf, normalized_objdir, &pos)) |
546 | goto error; | |
c2f493a4 | 547 | |
ca56dadb | 548 | CALLOC_ARRAY(ent, 1); |
cf2dc1c2 EW |
549 | /* pathbuf.buf is already in r->objects->odb_by_path */ |
550 | ent->path = strbuf_detach(&pathbuf, NULL); | |
c2f493a4 MW |
551 | |
552 | /* add the alternate entry */ | |
f0eaf638 JK |
553 | *r->objects->odb_tail = ent; |
554 | r->objects->odb_tail = &(ent->next); | |
c2f493a4 | 555 | ent->next = NULL; |
cf2dc1c2 EW |
556 | assert(r->objects->odb_by_path); |
557 | kh_value(r->objects->odb_by_path, pos) = ent; | |
c2f493a4 MW |
558 | |
559 | /* recursively add alternates */ | |
cf2dc1c2 | 560 | read_info_alternates(r, ent->path, depth + 1); |
199337d6 GC |
561 | ret = 0; |
562 | error: | |
563 | strbuf_release(&tmp); | |
564 | strbuf_release(&pathbuf); | |
565 | return ret; | |
c2f493a4 MW |
566 | } |
567 | ||
cf3c6352 JK |
568 | static const char *parse_alt_odb_entry(const char *string, |
569 | int sep, | |
570 | struct strbuf *out) | |
571 | { | |
572 | const char *end; | |
573 | ||
574 | strbuf_reset(out); | |
575 | ||
576 | if (*string == '#') { | |
577 | /* comment; consume up to next separator */ | |
578 | end = strchrnul(string, sep); | |
579 | } else if (*string == '"' && !unquote_c_style(out, string, &end)) { | |
580 | /* | |
581 | * quoted path; unquote_c_style has copied the | |
582 | * data for us and set "end". Broken quoting (e.g., | |
583 | * an entry that doesn't end with a quote) falls | |
584 | * back to the unquoted case below. | |
585 | */ | |
586 | } else { | |
587 | /* normal, unquoted path */ | |
588 | end = strchrnul(string, sep); | |
589 | strbuf_add(out, string, end - string); | |
590 | } | |
591 | ||
592 | if (*end) | |
593 | end++; | |
594 | return end; | |
595 | } | |
596 | ||
77f012e8 SB |
597 | static void link_alt_odb_entries(struct repository *r, const char *alt, |
598 | int sep, const char *relative_base, int depth) | |
c2f493a4 | 599 | { |
539e7506 | 600 | struct strbuf objdirbuf = STRBUF_INIT; |
cf3c6352 | 601 | struct strbuf entry = STRBUF_INIT; |
c2f493a4 | 602 | |
f28e3668 JK |
603 | if (!alt || !*alt) |
604 | return; | |
605 | ||
c2f493a4 | 606 | if (depth > 5) { |
259328b7 | 607 | error(_("%s: ignoring alternate object stores, nesting too deep"), |
c2f493a4 MW |
608 | relative_base); |
609 | return; | |
610 | } | |
611 | ||
199337d6 | 612 | strbuf_realpath(&objdirbuf, r->objects->odb->path, 1); |
539e7506 | 613 | |
cf3c6352 JK |
614 | while (*alt) { |
615 | alt = parse_alt_odb_entry(alt, sep, &entry); | |
616 | if (!entry.len) | |
9577e7e3 | 617 | continue; |
407532f8 | 618 | link_alt_odb_entry(r, &entry, |
cfc62fc9 | 619 | relative_base, depth, objdirbuf.buf); |
9577e7e3 | 620 | } |
cf3c6352 | 621 | strbuf_release(&entry); |
539e7506 | 622 | strbuf_release(&objdirbuf); |
d5a63b99 JH |
623 | } |
624 | ||
77f012e8 SB |
625 | static void read_info_alternates(struct repository *r, |
626 | const char *relative_base, | |
627 | int depth) | |
d5a63b99 | 628 | { |
5015f01c | 629 | char *path; |
dc732bd5 | 630 | struct strbuf buf = STRBUF_INIT; |
d5a63b99 | 631 | |
5015f01c | 632 | path = xstrfmt("%s/info/alternates", relative_base); |
dc732bd5 | 633 | if (strbuf_read_file(&buf, path, 1024) < 0) { |
f0f7bebe | 634 | warn_on_fopen_errors(path); |
dc732bd5 | 635 | free(path); |
9a217f2a | 636 | return; |
ace1534d | 637 | } |
d5a63b99 | 638 | |
77f012e8 | 639 | link_alt_odb_entries(r, buf.buf, '\n', relative_base, depth); |
dc732bd5 JK |
640 | strbuf_release(&buf); |
641 | free(path); | |
ace1534d JH |
642 | } |
643 | ||
bef70b22 DB |
644 | void add_to_alternates_file(const char *reference) |
645 | { | |
f132a127 | 646 | struct lock_file lock = LOCK_INIT; |
77b9b1d1 JK |
647 | char *alts = git_pathdup("objects/info/alternates"); |
648 | FILE *in, *out; | |
f132a127 | 649 | int found = 0; |
77b9b1d1 | 650 | |
f132a127 MÅ |
651 | hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR); |
652 | out = fdopen_lock_file(&lock, "w"); | |
77b9b1d1 | 653 | if (!out) |
259328b7 | 654 | die_errno(_("unable to fdopen alternates lockfile")); |
77b9b1d1 JK |
655 | |
656 | in = fopen(alts, "r"); | |
657 | if (in) { | |
658 | struct strbuf line = STRBUF_INIT; | |
77b9b1d1 | 659 | |
3f163962 | 660 | while (strbuf_getline(&line, in) != EOF) { |
77b9b1d1 JK |
661 | if (!strcmp(reference, line.buf)) { |
662 | found = 1; | |
663 | break; | |
664 | } | |
665 | fprintf_or_die(out, "%s\n", line.buf); | |
666 | } | |
667 | ||
668 | strbuf_release(&line); | |
669 | fclose(in); | |
77b9b1d1 JK |
670 | } |
671 | else if (errno != ENOENT) | |
259328b7 | 672 | die_errno(_("unable to read alternates file")); |
77b9b1d1 | 673 | |
f132a127 MÅ |
674 | if (found) { |
675 | rollback_lock_file(&lock); | |
676 | } else { | |
77b9b1d1 | 677 | fprintf_or_die(out, "%s\n", reference); |
f132a127 | 678 | if (commit_lock_file(&lock)) |
259328b7 | 679 | die_errno(_("unable to move new alternates file into place")); |
f0eaf638 | 680 | if (the_repository->objects->loaded_alternates) |
93d8d1e2 SB |
681 | link_alt_odb_entries(the_repository, reference, |
682 | '\n', NULL, 0); | |
77b9b1d1 JK |
683 | } |
684 | free(alts); | |
bef70b22 DB |
685 | } |
686 | ||
a5b34d21 JK |
687 | void add_to_alternates_memory(const char *reference) |
688 | { | |
689 | /* | |
690 | * Make sure alternates are initialized, or else our entry may be | |
691 | * overwritten when they are. | |
692 | */ | |
0b209034 | 693 | prepare_alt_odb(the_repository); |
a5b34d21 | 694 | |
93d8d1e2 SB |
695 | link_alt_odb_entries(the_repository, reference, |
696 | '\n', NULL, 0); | |
a5b34d21 JK |
697 | } |
698 | ||
b3cecf49 NS |
699 | struct object_directory *set_temporary_primary_odb(const char *dir, int will_destroy) |
700 | { | |
701 | struct object_directory *new_odb; | |
702 | ||
703 | /* | |
704 | * Make sure alternates are initialized, or else our entry may be | |
705 | * overwritten when they are. | |
706 | */ | |
707 | prepare_alt_odb(the_repository); | |
708 | ||
709 | /* | |
710 | * Make a new primary odb and link the old primary ODB in as an | |
711 | * alternate | |
712 | */ | |
713 | new_odb = xcalloc(1, sizeof(*new_odb)); | |
714 | new_odb->path = xstrdup(dir); | |
ecd81dfc NS |
715 | |
716 | /* | |
717 | * Disable ref updates while a temporary odb is active, since | |
718 | * the objects in the database may roll back. | |
719 | */ | |
720 | new_odb->disable_ref_updates = 1; | |
b3cecf49 NS |
721 | new_odb->will_destroy = will_destroy; |
722 | new_odb->next = the_repository->objects->odb; | |
723 | the_repository->objects->odb = new_odb; | |
724 | return new_odb->next; | |
725 | } | |
726 | ||
727 | void restore_primary_odb(struct object_directory *restore_odb, const char *old_path) | |
728 | { | |
729 | struct object_directory *cur_odb = the_repository->objects->odb; | |
730 | ||
731 | if (strcmp(old_path, cur_odb->path)) | |
732 | BUG("expected %s as primary object store; found %s", | |
733 | old_path, cur_odb->path); | |
734 | ||
735 | if (cur_odb->next != restore_odb) | |
736 | BUG("we expect the old primary object store to be the first alternate"); | |
737 | ||
738 | the_repository->objects->odb = restore_odb; | |
739 | free_object_directory(cur_odb); | |
740 | } | |
741 | ||
9eeea7d2 SB |
742 | /* |
743 | * Compute the exact path an alternate is at and returns it. In case of | |
744 | * error NULL is returned and the human readable error is added to `err` | |
efde7b72 | 745 | * `path` may be relative and should point to $GIT_DIR. |
9eeea7d2 SB |
746 | * `err` must not be null. |
747 | */ | |
748 | char *compute_alternate_path(const char *path, struct strbuf *err) | |
749 | { | |
750 | char *ref_git = NULL; | |
4530a85b | 751 | const char *repo; |
9eeea7d2 SB |
752 | int seen_error = 0; |
753 | ||
4530a85b AM |
754 | ref_git = real_pathdup(path, 0); |
755 | if (!ref_git) { | |
9eeea7d2 SB |
756 | seen_error = 1; |
757 | strbuf_addf(err, _("path '%s' does not exist"), path); | |
758 | goto out; | |
4530a85b | 759 | } |
9eeea7d2 SB |
760 | |
761 | repo = read_gitfile(ref_git); | |
762 | if (!repo) | |
763 | repo = read_gitfile(mkpath("%s/.git", ref_git)); | |
764 | if (repo) { | |
765 | free(ref_git); | |
766 | ref_git = xstrdup(repo); | |
767 | } | |
768 | ||
769 | if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) { | |
770 | char *ref_git_git = mkpathdup("%s/.git", ref_git); | |
771 | free(ref_git); | |
772 | ref_git = ref_git_git; | |
773 | } else if (!is_directory(mkpath("%s/objects", ref_git))) { | |
774 | struct strbuf sb = STRBUF_INIT; | |
775 | seen_error = 1; | |
776 | if (get_common_dir(&sb, ref_git)) { | |
777 | strbuf_addf(err, | |
778 | _("reference repository '%s' as a linked " | |
779 | "checkout is not supported yet."), | |
780 | path); | |
781 | goto out; | |
782 | } | |
783 | ||
784 | strbuf_addf(err, _("reference repository '%s' is not a " | |
785 | "local repository."), path); | |
786 | goto out; | |
787 | } | |
788 | ||
789 | if (!access(mkpath("%s/shallow", ref_git), F_OK)) { | |
790 | strbuf_addf(err, _("reference repository '%s' is shallow"), | |
791 | path); | |
792 | seen_error = 1; | |
793 | goto out; | |
794 | } | |
795 | ||
796 | if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) { | |
797 | strbuf_addf(err, | |
798 | _("reference repository '%s' is grafted"), | |
799 | path); | |
800 | seen_error = 1; | |
801 | goto out; | |
802 | } | |
803 | ||
804 | out: | |
805 | if (seen_error) { | |
6a83d902 | 806 | FREE_AND_NULL(ref_git); |
9eeea7d2 SB |
807 | } |
808 | ||
809 | return ref_git; | |
810 | } | |
811 | ||
f57a7396 TB |
812 | struct object_directory *find_odb(struct repository *r, const char *obj_dir) |
813 | { | |
814 | struct object_directory *odb; | |
815 | char *obj_dir_real = real_pathdup(obj_dir, 1); | |
816 | struct strbuf odb_path_real = STRBUF_INIT; | |
817 | ||
818 | prepare_alt_odb(r); | |
819 | for (odb = r->objects->odb; odb; odb = odb->next) { | |
820 | strbuf_realpath(&odb_path_real, odb->path, 1); | |
821 | if (!strcmp(obj_dir_real, odb_path_real.buf)) | |
822 | break; | |
823 | } | |
824 | ||
825 | free(obj_dir_real); | |
826 | strbuf_release(&odb_path_real); | |
827 | ||
828 | if (!odb) | |
829 | die(_("could not find object directory matching %s"), obj_dir); | |
830 | return odb; | |
831 | } | |
832 | ||
709dfa69 JK |
833 | static void fill_alternate_refs_command(struct child_process *cmd, |
834 | const char *repo_path) | |
835 | { | |
836 | const char *value; | |
837 | ||
838 | if (!git_config_get_value("core.alternateRefsCommand", &value)) { | |
839 | cmd->use_shell = 1; | |
840 | ||
c972bf4c JK |
841 | strvec_push(&cmd->args, value); |
842 | strvec_push(&cmd->args, repo_path); | |
709dfa69 JK |
843 | } else { |
844 | cmd->git_cmd = 1; | |
845 | ||
c972bf4c JK |
846 | strvec_pushf(&cmd->args, "--git-dir=%s", repo_path); |
847 | strvec_push(&cmd->args, "for-each-ref"); | |
848 | strvec_push(&cmd->args, "--format=%(objectname)"); | |
709dfa69 JK |
849 | |
850 | if (!git_config_get_value("core.alternateRefsPrefixes", &value)) { | |
c972bf4c JK |
851 | strvec_push(&cmd->args, "--"); |
852 | strvec_split(&cmd->args, value); | |
709dfa69 JK |
853 | } |
854 | } | |
855 | ||
29fda24d | 856 | strvec_pushv(&cmd->env, (const char **)local_repo_env); |
709dfa69 JK |
857 | cmd->out = -1; |
858 | } | |
859 | ||
860 | static void read_alternate_refs(const char *path, | |
861 | alternate_ref_fn *cb, | |
862 | void *data) | |
863 | { | |
864 | struct child_process cmd = CHILD_PROCESS_INIT; | |
865 | struct strbuf line = STRBUF_INIT; | |
866 | FILE *fh; | |
867 | ||
868 | fill_alternate_refs_command(&cmd, path); | |
869 | ||
870 | if (start_command(&cmd)) | |
871 | return; | |
872 | ||
873 | fh = xfdopen(cmd.out, "r"); | |
874 | while (strbuf_getline_lf(&line, fh) != EOF) { | |
875 | struct object_id oid; | |
876 | const char *p; | |
877 | ||
878 | if (parse_oid_hex(line.buf, &oid, &p) || *p) { | |
879 | warning(_("invalid line while parsing alternate refs: %s"), | |
880 | line.buf); | |
881 | break; | |
882 | } | |
883 | ||
884 | cb(&oid, data); | |
885 | } | |
886 | ||
887 | fclose(fh); | |
888 | finish_command(&cmd); | |
86ad3ea5 | 889 | strbuf_release(&line); |
709dfa69 JK |
890 | } |
891 | ||
892 | struct alternate_refs_data { | |
893 | alternate_ref_fn *fn; | |
894 | void *data; | |
895 | }; | |
896 | ||
897 | static int refs_from_alternate_cb(struct object_directory *e, | |
898 | void *data) | |
899 | { | |
900 | struct strbuf path = STRBUF_INIT; | |
901 | size_t base_len; | |
902 | struct alternate_refs_data *cb = data; | |
903 | ||
904 | if (!strbuf_realpath(&path, e->path, 0)) | |
905 | goto out; | |
906 | if (!strbuf_strip_suffix(&path, "/objects")) | |
907 | goto out; | |
908 | base_len = path.len; | |
909 | ||
910 | /* Is this a git repository with refs? */ | |
911 | strbuf_addstr(&path, "/refs"); | |
912 | if (!is_directory(path.buf)) | |
913 | goto out; | |
914 | strbuf_setlen(&path, base_len); | |
915 | ||
916 | read_alternate_refs(path.buf, cb->fn, cb->data); | |
917 | ||
918 | out: | |
919 | strbuf_release(&path); | |
920 | return 0; | |
921 | } | |
922 | ||
923 | void for_each_alternate_ref(alternate_ref_fn fn, void *data) | |
924 | { | |
925 | struct alternate_refs_data cb; | |
926 | cb.fn = fn; | |
927 | cb.data = data; | |
928 | foreach_alt_odb(refs_from_alternate_cb, &cb); | |
929 | } | |
930 | ||
fe1b2268 | 931 | int foreach_alt_odb(alt_odb_fn fn, void *cb) |
d79796bc | 932 | { |
263db403 | 933 | struct object_directory *ent; |
fe1b2268 | 934 | int r = 0; |
d79796bc | 935 | |
0b209034 | 936 | prepare_alt_odb(the_repository); |
f0eaf638 | 937 | for (ent = the_repository->objects->odb->next; ent; ent = ent->next) { |
fe1b2268 JK |
938 | r = fn(ent, cb); |
939 | if (r) | |
940 | break; | |
941 | } | |
942 | return r; | |
d79796bc JH |
943 | } |
944 | ||
13068bf0 | 945 | void prepare_alt_odb(struct repository *r) |
c2f493a4 | 946 | { |
f0eaf638 | 947 | if (r->objects->loaded_alternates) |
7dc24aa5 SP |
948 | return; |
949 | ||
13068bf0 | 950 | link_alt_odb_entries(r, r->objects->alternate_db, PATH_SEP, NULL, 0); |
c2f493a4 | 951 | |
f0eaf638 JK |
952 | read_info_alternates(r, r->objects->odb->path, 0); |
953 | r->objects->loaded_alternates = 1; | |
c2f493a4 MW |
954 | } |
955 | ||
d85cd187 PS |
956 | int has_alt_odb(struct repository *r) |
957 | { | |
958 | prepare_alt_odb(r); | |
959 | return !!r->objects->odb->next; | |
960 | } | |
961 | ||
3096b2ec | 962 | /* Returns 1 if we have successfully freshened the file, 0 otherwise. */ |
33d4221c | 963 | static int freshen_file(const char *fn) |
ace1534d | 964 | { |
312cd761 | 965 | return !utime(fn, NULL); |
0f4dc14a | 966 | } |
ace1534d | 967 | |
3096b2ec JK |
968 | /* |
969 | * All of the check_and_freshen functions return 1 if the file exists and was | |
970 | * freshened (if freshening was requested), 0 otherwise. If they return | |
971 | * 0, you should not assume that it is safe to skip a write of the object (it | |
972 | * either does not exist on disk, or has a stale mtime and may be subject to | |
973 | * pruning). | |
974 | */ | |
6a5e6f5e | 975 | int check_and_freshen_file(const char *fn, int freshen) |
33d4221c JK |
976 | { |
977 | if (access(fn, F_OK)) | |
978 | return 0; | |
3096b2ec | 979 | if (freshen && !freshen_file(fn)) |
33d4221c JK |
980 | return 0; |
981 | return 1; | |
982 | } | |
983 | ||
f0eaf638 JK |
984 | static int check_and_freshen_odb(struct object_directory *odb, |
985 | const struct object_id *oid, | |
986 | int freshen) | |
33d4221c | 987 | { |
f0eaf638 | 988 | static struct strbuf path = STRBUF_INIT; |
514c5fdd | 989 | odb_loose_path(odb, &path, oid); |
f0eaf638 JK |
990 | return check_and_freshen_file(path.buf, freshen); |
991 | } | |
ea657730 | 992 | |
f0eaf638 JK |
993 | static int check_and_freshen_local(const struct object_id *oid, int freshen) |
994 | { | |
995 | return check_and_freshen_odb(the_repository->objects->odb, oid, freshen); | |
33d4221c JK |
996 | } |
997 | ||
6862ebbf | 998 | static int check_and_freshen_nonlocal(const struct object_id *oid, int freshen) |
0f4dc14a | 999 | { |
263db403 | 1000 | struct object_directory *odb; |
f3f043a1 | 1001 | |
0b209034 | 1002 | prepare_alt_odb(the_repository); |
f0eaf638 JK |
1003 | for (odb = the_repository->objects->odb->next; odb; odb = odb->next) { |
1004 | if (check_and_freshen_odb(odb, oid, freshen)) | |
c529d75a | 1005 | return 1; |
ace1534d | 1006 | } |
c529d75a | 1007 | return 0; |
ace1534d JH |
1008 | } |
1009 | ||
6862ebbf | 1010 | static int check_and_freshen(const struct object_id *oid, int freshen) |
33d4221c | 1011 | { |
6862ebbf | 1012 | return check_and_freshen_local(oid, freshen) || |
1013 | check_and_freshen_nonlocal(oid, freshen); | |
33d4221c JK |
1014 | } |
1015 | ||
6862ebbf | 1016 | int has_loose_object_nonlocal(const struct object_id *oid) |
33d4221c | 1017 | { |
6862ebbf | 1018 | return check_and_freshen_nonlocal(oid, 0); |
33d4221c JK |
1019 | } |
1020 | ||
b7573536 | 1021 | int has_loose_object(const struct object_id *oid) |
0f4dc14a | 1022 | { |
6862ebbf | 1023 | return check_and_freshen(oid, 0); |
0f4dc14a BC |
1024 | } |
1025 | ||
02710228 SP |
1026 | static void mmap_limit_check(size_t length) |
1027 | { | |
1028 | static size_t limit = 0; | |
1029 | if (!limit) { | |
1030 | limit = git_env_ulong("GIT_MMAP_LIMIT", 0); | |
1031 | if (!limit) | |
1032 | limit = SIZE_MAX; | |
1033 | } | |
1034 | if (length > limit) | |
259328b7 | 1035 | die(_("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX), |
02710228 SP |
1036 | (uintmax_t)length, (uintmax_t)limit); |
1037 | } | |
1038 | ||
1570856b JK |
1039 | void *xmmap_gently(void *start, size_t length, |
1040 | int prot, int flags, int fd, off_t offset) | |
58ecbd5e | 1041 | { |
02710228 SP |
1042 | void *ret; |
1043 | ||
1044 | mmap_limit_check(length); | |
1045 | ret = mmap(start, length, prot, flags, fd, offset); | |
9827d4c1 JK |
1046 | if (ret == MAP_FAILED && !length) |
1047 | ret = NULL; | |
58ecbd5e JN |
1048 | return ret; |
1049 | } | |
1050 | ||
dc059294 EW |
1051 | const char *mmap_os_err(void) |
1052 | { | |
1053 | static const char blank[] = ""; | |
1054 | #if defined(__linux__) | |
1055 | if (errno == ENOMEM) { | |
1056 | /* this continues an existing error message: */ | |
1057 | static const char enomem[] = | |
1058 | ", check sys.vm.max_map_count and/or RLIMIT_DATA"; | |
1059 | return enomem; | |
1060 | } | |
1061 | #endif /* OS-specific bits */ | |
1062 | return blank; | |
1063 | } | |
1064 | ||
1570856b JK |
1065 | void *xmmap(void *start, size_t length, |
1066 | int prot, int flags, int fd, off_t offset) | |
1067 | { | |
1068 | void *ret = xmmap_gently(start, length, prot, flags, fd, offset); | |
1069 | if (ret == MAP_FAILED) | |
dc059294 | 1070 | die_errno(_("mmap failed%s"), mmap_os_err()); |
1570856b JK |
1071 | return ret; |
1072 | } | |
1073 | ||
b04cdea4 ÆAB |
1074 | static int format_object_header_literally(char *str, size_t size, |
1075 | const char *type, size_t objsize) | |
1076 | { | |
1077 | return xsnprintf(str, size, "%s %"PRIuMAX, type, (uintmax_t)objsize) + 1; | |
1078 | } | |
1079 | ||
1080 | int format_object_header(char *str, size_t size, enum object_type type, | |
1081 | size_t objsize) | |
1082 | { | |
1083 | const char *name = type_name(type); | |
1084 | ||
1085 | if (!name) | |
1086 | BUG("could not get a type name for 'enum object_type' value %d", type); | |
1087 | ||
1088 | return format_object_header_literally(str, size, name, objsize); | |
1089 | } | |
1090 | ||
b98d1885 | 1091 | int check_object_signature(struct repository *r, const struct object_id *oid, |
44439c1c ÆAB |
1092 | void *buf, unsigned long size, |
1093 | enum object_type type) | |
88d0db55 | 1094 | { |
0f156dbb ÆAB |
1095 | struct object_id real_oid; |
1096 | ||
1097 | hash_object_file(r->hash_algo, buf, size, type, &real_oid); | |
1098 | ||
1099 | return !oideq(oid, &real_oid) ? -1 : 0; | |
1100 | } | |
1101 | ||
1102 | int stream_object_signature(struct repository *r, const struct object_id *oid) | |
1103 | { | |
1104 | struct object_id real_oid; | |
1105 | unsigned long size; | |
090ea126 NTND |
1106 | enum object_type obj_type; |
1107 | struct git_istream *st; | |
18e2588e | 1108 | git_hash_ctx c; |
1af64f73 | 1109 | char hdr[MAX_HEADER_LEN]; |
090ea126 | 1110 | int hdrlen; |
88d0db55 | 1111 | |
b98d1885 | 1112 | st = open_istream(r, oid, &obj_type, &size, NULL); |
090ea126 NTND |
1113 | if (!st) |
1114 | return -1; | |
88d0db55 | 1115 | |
090ea126 | 1116 | /* Generate the header */ |
b04cdea4 | 1117 | hdrlen = format_object_header(hdr, sizeof(hdr), obj_type, size); |
88d0db55 | 1118 | |
090ea126 | 1119 | /* Sha1.. */ |
b98d1885 MT |
1120 | r->hash_algo->init_fn(&c); |
1121 | r->hash_algo->update_fn(&c, hdr, hdrlen); | |
090ea126 NTND |
1122 | for (;;) { |
1123 | char buf[1024 * 16]; | |
1124 | ssize_t readlen = read_istream(st, buf, sizeof(buf)); | |
1f688557 | 1125 | |
f54fac53 JK |
1126 | if (readlen < 0) { |
1127 | close_istream(st); | |
1128 | return -1; | |
1129 | } | |
090ea126 NTND |
1130 | if (!readlen) |
1131 | break; | |
b98d1885 | 1132 | r->hash_algo->update_fn(&c, buf, readlen); |
fa5fc15d | 1133 | } |
0f156dbb | 1134 | r->hash_algo->final_oid_fn(&real_oid, &c); |
090ea126 | 1135 | close_istream(st); |
0f156dbb | 1136 | return !oideq(oid, &real_oid) ? -1 : 0; |
fa5fc15d SP |
1137 | } |
1138 | ||
1b8ac5ea | 1139 | int git_open_cloexec(const char *name, int flags) |
a0788266 | 1140 | { |
1e3001a8 JH |
1141 | int fd; |
1142 | static int o_cloexec = O_CLOEXEC; | |
a0788266 | 1143 | |
1e3001a8 JH |
1144 | fd = open(name, flags | o_cloexec); |
1145 | if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) { | |
cd66ada0 | 1146 | /* Try again w/o O_CLOEXEC: the kernel might not support it */ |
1e3001a8 JH |
1147 | o_cloexec &= ~O_CLOEXEC; |
1148 | fd = open(name, flags | o_cloexec); | |
491a8dec | 1149 | } |
491a8dec | 1150 | |
9fb9495d | 1151 | #if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC) |
491a8dec | 1152 | { |
1e3001a8 | 1153 | static int fd_cloexec = FD_CLOEXEC; |
a0788266 | 1154 | |
1e3001a8 JH |
1155 | if (!o_cloexec && 0 <= fd && fd_cloexec) { |
1156 | /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */ | |
9fb9495d EW |
1157 | int flags = fcntl(fd, F_GETFD); |
1158 | if (fcntl(fd, F_SETFD, flags | fd_cloexec)) | |
1e3001a8 | 1159 | fd_cloexec = 0; |
cd66ada0 | 1160 | } |
44d1c19e | 1161 | } |
a0788266 | 1162 | #endif |
1b8ac5ea | 1163 | return fd; |
a0788266 JS |
1164 | } |
1165 | ||
3cf8b462 | 1166 | /* |
514c5fdd | 1167 | * Find "oid" as a loose object in the local repository or in an alternate. |
771e7d57 JK |
1168 | * Returns 0 on success, negative on failure. |
1169 | * | |
1170 | * The "path" out-parameter will give the path of the object we found (if any). | |
1171 | * Note that it may point to static storage and is only valid until another | |
514c5fdd | 1172 | * call to stat_loose_object(). |
3cf8b462 | 1173 | */ |
514c5fdd JK |
1174 | static int stat_loose_object(struct repository *r, const struct object_id *oid, |
1175 | struct stat *st, const char **path) | |
1f688557 | 1176 | { |
263db403 | 1177 | struct object_directory *odb; |
ea657730 CC |
1178 | static struct strbuf buf = STRBUF_INIT; |
1179 | ||
d2607fa0 | 1180 | prepare_alt_odb(r); |
f0eaf638 | 1181 | for (odb = r->objects->odb; odb; odb = odb->next) { |
514c5fdd | 1182 | *path = odb_loose_path(odb, &buf, oid); |
771e7d57 | 1183 | if (!lstat(*path, st)) |
052fe5ea | 1184 | return 0; |
c7934306 SP |
1185 | } |
1186 | ||
3cf8b462 SP |
1187 | return -1; |
1188 | } | |
1189 | ||
771e7d57 | 1190 | /* |
514c5fdd | 1191 | * Like stat_loose_object(), but actually open the object and return the |
771e7d57 JK |
1192 | * descriptor. See the caveats on the "path" parameter above. |
1193 | */ | |
514c5fdd JK |
1194 | static int open_loose_object(struct repository *r, |
1195 | const struct object_id *oid, const char **path) | |
60bb8b14 | 1196 | { |
44d1c19e | 1197 | int fd; |
263db403 | 1198 | struct object_directory *odb; |
f0eaf638 | 1199 | int most_interesting_errno = ENOENT; |
ea657730 CC |
1200 | static struct strbuf buf = STRBUF_INIT; |
1201 | ||
ec7283e5 | 1202 | prepare_alt_odb(r); |
f0eaf638 | 1203 | for (odb = r->objects->odb; odb; odb = odb->next) { |
514c5fdd | 1204 | *path = odb_loose_path(odb, &buf, oid); |
771e7d57 | 1205 | fd = git_open(*path); |
44d1c19e LT |
1206 | if (fd >= 0) |
1207 | return fd; | |
f0eaf638 | 1208 | |
d6c8a05b JK |
1209 | if (most_interesting_errno == ENOENT) |
1210 | most_interesting_errno = errno; | |
03e79c88 | 1211 | } |
d6c8a05b | 1212 | errno = most_interesting_errno; |
44d1c19e | 1213 | return -1; |
1f688557 JH |
1214 | } |
1215 | ||
61c7711c | 1216 | static int quick_has_loose(struct repository *r, |
d7a24573 | 1217 | const struct object_id *oid) |
61c7711c | 1218 | { |
61c7711c JK |
1219 | struct object_directory *odb; |
1220 | ||
61c7711c JK |
1221 | prepare_alt_odb(r); |
1222 | for (odb = r->objects->odb; odb; odb = odb->next) { | |
92d8ed8a | 1223 | if (oidtree_contains(odb_loose_cache(odb, oid), oid)) |
61c7711c JK |
1224 | return 1; |
1225 | } | |
1226 | return 0; | |
1227 | } | |
1228 | ||
f6371f92 | 1229 | /* |
ae285ac4 JT |
1230 | * Map and close the given loose object fd. The path argument is used for |
1231 | * error reporting. | |
f6371f92 | 1232 | */ |
ae285ac4 | 1233 | static void *map_fd(int fd, const char *path, unsigned long *size) |
27d69a46 | 1234 | { |
ae285ac4 JT |
1235 | void *map = NULL; |
1236 | struct stat st; | |
1f688557 | 1237 | |
ae285ac4 JT |
1238 | if (!fstat(fd, &st)) { |
1239 | *size = xsize_t(st.st_size); | |
1240 | if (!*size) { | |
1241 | /* mmap() is forbidden on empty files */ | |
1242 | error(_("object file %s is empty"), path); | |
1243 | close(fd); | |
1244 | return NULL; | |
144bde78 | 1245 | } |
ae285ac4 | 1246 | map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0); |
74e34e1f | 1247 | } |
ae285ac4 | 1248 | close(fd); |
0fcfd160 | 1249 | return map; |
74e34e1f NP |
1250 | } |
1251 | ||
514c5fdd JK |
1252 | void *map_loose_object(struct repository *r, |
1253 | const struct object_id *oid, | |
1254 | unsigned long *size) | |
068f85e3 | 1255 | { |
ae285ac4 JT |
1256 | const char *p; |
1257 | int fd = open_loose_object(r, oid, &p); | |
1258 | ||
1259 | if (fd < 0) | |
1260 | return NULL; | |
1261 | return map_fd(fd, p, size); | |
068f85e3 | 1262 | } |
1263 | ||
3b6a8db3 ÆAB |
1264 | enum unpack_loose_header_result unpack_loose_header(git_zstream *stream, |
1265 | unsigned char *map, | |
1266 | unsigned long mapsize, | |
1267 | void *buffer, | |
1268 | unsigned long bufsiz, | |
1269 | struct strbuf *header) | |
47fe3f6e | 1270 | { |
01cab976 | 1271 | int status; |
31877c9a | 1272 | |
c4483576 LT |
1273 | /* Get the data stream */ |
1274 | memset(stream, 0, sizeof(*stream)); | |
1275 | stream->next_in = map; | |
1276 | stream->avail_in = mapsize; | |
1277 | stream->next_out = buffer; | |
93821bd9 | 1278 | stream->avail_out = bufsiz; |
47fe3f6e | 1279 | |
39c68542 | 1280 | git_inflate_init(stream); |
31877c9a | 1281 | obj_read_unlock(); |
01cab976 | 1282 | status = git_inflate(stream, 0); |
31877c9a | 1283 | obj_read_lock(); |
d21f8426 | 1284 | if (status < Z_OK) |
3b6a8db3 | 1285 | return ULHR_BAD; |
d131b7af | 1286 | |
46f03448 KN |
1287 | /* |
1288 | * Check if entire header is unpacked in the first iteration. | |
d131b7af | 1289 | */ |
46f03448 | 1290 | if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) |
3b6a8db3 | 1291 | return ULHR_OK; |
d131b7af | 1292 | |
01cab976 ÆAB |
1293 | /* |
1294 | * We have a header longer than MAX_HEADER_LEN. The "header" | |
1295 | * here is only non-NULL when we run "cat-file | |
1296 | * --allow-unknown-type". | |
1297 | */ | |
1298 | if (!header) | |
5848fb11 | 1299 | return ULHR_TOO_LONG; |
d131b7af | 1300 | |
46f03448 KN |
1301 | /* |
1302 | * buffer[0..bufsiz] was not large enough. Copy the partial | |
1303 | * result out to header, and then append the result of further | |
1304 | * reading the stream. | |
1305 | */ | |
1306 | strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer); | |
1307 | stream->next_out = buffer; | |
1308 | stream->avail_out = bufsiz; | |
d131b7af | 1309 | |
46f03448 | 1310 | do { |
31877c9a | 1311 | obj_read_unlock(); |
46f03448 | 1312 | status = git_inflate(stream, 0); |
31877c9a | 1313 | obj_read_lock(); |
46f03448 KN |
1314 | strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer); |
1315 | if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer)) | |
1316 | return 0; | |
1317 | stream->next_out = buffer; | |
1318 | stream->avail_out = bufsiz; | |
1319 | } while (status != Z_STREAM_END); | |
5848fb11 | 1320 | return ULHR_TOO_LONG; |
d131b7af SP |
1321 | } |
1322 | ||
00a7760e JK |
1323 | static void *unpack_loose_rest(git_zstream *stream, |
1324 | void *buffer, unsigned long size, | |
1325 | const struct object_id *oid) | |
95099731 | 1326 | { |
5180cacc | 1327 | int bytes = strlen(buffer) + 1; |
3aee68aa | 1328 | unsigned char *buf = xmallocz(size); |
93821bd9 | 1329 | unsigned long n; |
7efbff75 | 1330 | int status = Z_OK; |
95099731 | 1331 | |
93821bd9 LT |
1332 | n = stream->total_out - bytes; |
1333 | if (n > size) | |
1334 | n = size; | |
1335 | memcpy(buf, (char *) buffer + bytes, n); | |
1336 | bytes = n; | |
456cdf6e LT |
1337 | if (bytes <= size) { |
1338 | /* | |
1339 | * The above condition must be (bytes <= size), not | |
1340 | * (bytes < size). In other words, even though we | |
ccf5ace0 | 1341 | * expect no more output and set avail_out to zero, |
456cdf6e LT |
1342 | * the input zlib stream may have bytes that express |
1343 | * "this concludes the stream", and we *do* want to | |
1344 | * eat that input. | |
1345 | * | |
1346 | * Otherwise we would not be able to test that we | |
1347 | * consumed all the input to reach the expected size; | |
1348 | * we also want to check that zlib tells us that all | |
1349 | * went well with status == Z_STREAM_END at the end. | |
1350 | */ | |
5180cacc LT |
1351 | stream->next_out = buf + bytes; |
1352 | stream->avail_out = size - bytes; | |
31877c9a MT |
1353 | while (status == Z_OK) { |
1354 | obj_read_unlock(); | |
39c68542 | 1355 | status = git_inflate(stream, Z_FINISH); |
31877c9a MT |
1356 | obj_read_lock(); |
1357 | } | |
5180cacc | 1358 | } |
456cdf6e | 1359 | if (status == Z_STREAM_END && !stream->avail_in) { |
39c68542 | 1360 | git_inflate_end(stream); |
7efbff75 | 1361 | return buf; |
95099731 NTND |
1362 | } |
1363 | ||
7efbff75 | 1364 | if (status < 0) |
00a7760e | 1365 | error(_("corrupt loose object '%s'"), oid_to_hex(oid)); |
7efbff75 | 1366 | else if (stream->avail_in) |
259328b7 | 1367 | error(_("garbage at end of loose object '%s'"), |
00a7760e | 1368 | oid_to_hex(oid)); |
7efbff75 JH |
1369 | free(buf); |
1370 | return NULL; | |
95099731 NTND |
1371 | } |
1372 | ||
d40d535b | 1373 | /* |
5180cacc LT |
1374 | * We used to just use "sscanf()", but that's actually way |
1375 | * too permissive for what we want to check. So do an anal | |
1376 | * object header parse by hand. | |
d40d535b | 1377 | */ |
dccb32bf | 1378 | int parse_loose_header(const char *hdr, struct object_info *oi) |
1f688557 | 1379 | { |
46f03448 | 1380 | const char *type_buf = hdr; |
d6a09e79 | 1381 | size_t size; |
46f03448 | 1382 | int type, type_len = 0; |
43057304 | 1383 | |
5180cacc | 1384 | /* |
46f03448 | 1385 | * The type can be of any size but is followed by |
21666f1a | 1386 | * a space. |
5180cacc | 1387 | */ |
5180cacc LT |
1388 | for (;;) { |
1389 | char c = *hdr++; | |
d21f8426 JH |
1390 | if (!c) |
1391 | return -1; | |
5180cacc LT |
1392 | if (c == ' ') |
1393 | break; | |
46f03448 | 1394 | type_len++; |
5180cacc | 1395 | } |
1f688557 | 1396 | |
46f03448 | 1397 | type = type_from_string_gently(type_buf, type_len, 1); |
6ca32f47 BW |
1398 | if (oi->type_name) |
1399 | strbuf_add(oi->type_name, type_buf, type_len); | |
46f03448 KN |
1400 | if (oi->typep) |
1401 | *oi->typep = type; | |
5180cacc LT |
1402 | |
1403 | /* | |
1404 | * The length must follow immediately, and be in canonical | |
1405 | * decimal format (ie "010" is not valid). | |
1406 | */ | |
1407 | size = *hdr++ - '0'; | |
1408 | if (size > 9) | |
1409 | return -1; | |
1410 | if (size) { | |
1411 | for (;;) { | |
1412 | unsigned long c = *hdr - '0'; | |
1413 | if (c > 9) | |
1414 | break; | |
1415 | hdr++; | |
d6a09e79 | 1416 | size = st_add(st_mult(size, 10), c); |
1b1005d1 | 1417 | } |
c01f51cc | 1418 | } |
46f03448 KN |
1419 | |
1420 | if (oi->sizep) | |
d6a09e79 | 1421 | *oi->sizep = cast_size_t_to_ulong(size); |
5180cacc LT |
1422 | |
1423 | /* | |
1424 | * The length must be followed by a zero byte | |
1425 | */ | |
dccb32bf ÆAB |
1426 | if (*hdr) |
1427 | return -1; | |
21666f1a | 1428 | |
dccb32bf ÆAB |
1429 | /* |
1430 | * The format is valid, but the type may still be bogus. The | |
1431 | * Caller needs to check its oi->typep. | |
1432 | */ | |
1433 | return 0; | |
bf592c50 DB |
1434 | } |
1435 | ||
514c5fdd JK |
1436 | static int loose_object_info(struct repository *r, |
1437 | const struct object_id *oid, | |
1438 | struct object_info *oi, int flags) | |
65c2e0c3 | 1439 | { |
46f03448 | 1440 | int status = 0; |
9e59b38c | 1441 | int fd; |
46f03448 | 1442 | unsigned long mapsize; |
9e59b38c | 1443 | const char *path; |
65c2e0c3 | 1444 | void *map; |
ef49a7a0 | 1445 | git_zstream stream; |
1af64f73 | 1446 | char hdr[MAX_HEADER_LEN]; |
46f03448 | 1447 | struct strbuf hdrbuf = STRBUF_INIT; |
c84a1f3e | 1448 | unsigned long size_scratch; |
dccb32bf | 1449 | enum object_type type_scratch; |
01cab976 | 1450 | int allow_unknown = flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE; |
65c2e0c3 | 1451 | |
b99b6bcc JK |
1452 | if (oi->delta_base_oid) |
1453 | oidclr(oi->delta_base_oid); | |
5d642e75 | 1454 | |
052fe5ea JK |
1455 | /* |
1456 | * If we don't care about type or size, then we don't | |
4ef8d1dd JH |
1457 | * need to look inside the object at all. Note that we |
1458 | * do not optimize out the stat call, even if the | |
1459 | * caller doesn't care about the disk-size, since our | |
1460 | * return value implicitly indicates whether the | |
1461 | * object even exists. | |
052fe5ea | 1462 | */ |
6ca32f47 | 1463 | if (!oi->typep && !oi->type_name && !oi->sizep && !oi->contentp) { |
4ef8d1dd | 1464 | struct stat st; |
61c7711c | 1465 | if (!oi->disk_sizep && (flags & OBJECT_INFO_QUICK)) |
d7a24573 | 1466 | return quick_has_loose(r, oid) ? 0 : -1; |
514c5fdd | 1467 | if (stat_loose_object(r, oid, &st, &path) < 0) |
4ef8d1dd JH |
1468 | return -1; |
1469 | if (oi->disk_sizep) | |
23c339c0 | 1470 | *oi->disk_sizep = st.st_size; |
052fe5ea JK |
1471 | return 0; |
1472 | } | |
1473 | ||
9e59b38c JT |
1474 | fd = open_loose_object(r, oid, &path); |
1475 | if (fd < 0) { | |
1476 | if (errno != ENOENT) | |
1477 | error_errno(_("unable to open loose object %s"), oid_to_hex(oid)); | |
1478 | return -1; | |
1479 | } | |
1480 | map = map_fd(fd, path, &mapsize); | |
f0df4ed5 | 1481 | if (!map) |
dbea72a8 | 1482 | return -1; |
c84a1f3e JT |
1483 | |
1484 | if (!oi->sizep) | |
1485 | oi->sizep = &size_scratch; | |
dccb32bf ÆAB |
1486 | if (!oi->typep) |
1487 | oi->typep = &type_scratch; | |
c84a1f3e | 1488 | |
23c339c0 JK |
1489 | if (oi->disk_sizep) |
1490 | *oi->disk_sizep = mapsize; | |
01cab976 | 1491 | |
3b6a8db3 ÆAB |
1492 | switch (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr), |
1493 | allow_unknown ? &hdrbuf : NULL)) { | |
1494 | case ULHR_OK: | |
dccb32bf ÆAB |
1495 | if (parse_loose_header(hdrbuf.len ? hdrbuf.buf : hdr, oi) < 0) |
1496 | status = error(_("unable to parse %s header"), oid_to_hex(oid)); | |
1497 | else if (!allow_unknown && *oi->typep < 0) | |
1498 | die(_("invalid object type")); | |
1499 | ||
1500 | if (!oi->contentp) | |
1501 | break; | |
1502 | *oi->contentp = unpack_loose_rest(&stream, hdr, *oi->sizep, oid); | |
1503 | if (*oi->contentp) | |
1504 | goto cleanup; | |
1505 | ||
1506 | status = -1; | |
3b6a8db3 ÆAB |
1507 | break; |
1508 | case ULHR_BAD: | |
259328b7 | 1509 | status = error(_("unable to unpack %s header"), |
514c5fdd | 1510 | oid_to_hex(oid)); |
3b6a8db3 | 1511 | break; |
5848fb11 ÆAB |
1512 | case ULHR_TOO_LONG: |
1513 | status = error(_("header for %s too long, exceeds %d bytes"), | |
1514 | oid_to_hex(oid), MAX_HEADER_LEN); | |
1515 | break; | |
3b6a8db3 | 1516 | } |
c84a1f3e | 1517 | |
9e59b38c JT |
1518 | if (status && (flags & OBJECT_INFO_DIE_IF_CORRUPT)) |
1519 | die(_("loose object %s (stored in %s) is corrupt"), | |
1520 | oid_to_hex(oid), path); | |
1521 | ||
dccb32bf ÆAB |
1522 | git_inflate_end(&stream); |
1523 | cleanup: | |
65c2e0c3 | 1524 | munmap(map, mapsize); |
c84a1f3e JT |
1525 | if (oi->sizep == &size_scratch) |
1526 | oi->sizep = NULL; | |
46f03448 | 1527 | strbuf_release(&hdrbuf); |
dccb32bf ÆAB |
1528 | if (oi->typep == &type_scratch) |
1529 | oi->typep = NULL; | |
3ab0fb06 | 1530 | oi->whence = OI_LOOSE; |
dccb32bf | 1531 | return status; |
65c2e0c3 JH |
1532 | } |
1533 | ||
31877c9a MT |
1534 | int obj_read_use_lock = 0; |
1535 | pthread_mutex_t obj_read_mutex; | |
1536 | ||
1537 | void enable_obj_read_lock(void) | |
1538 | { | |
1539 | if (obj_read_use_lock) | |
1540 | return; | |
1541 | ||
1542 | obj_read_use_lock = 1; | |
1543 | init_recursive_mutex(&obj_read_mutex); | |
1544 | } | |
1545 | ||
1546 | void disable_obj_read_lock(void) | |
1547 | { | |
1548 | if (!obj_read_use_lock) | |
1549 | return; | |
1550 | ||
1551 | obj_read_use_lock = 0; | |
1552 | pthread_mutex_destroy(&obj_read_mutex); | |
1553 | } | |
1554 | ||
8b4c0103 JT |
1555 | int fetch_if_missing = 1; |
1556 | ||
31877c9a MT |
1557 | static int do_oid_object_info_extended(struct repository *r, |
1558 | const struct object_id *oid, | |
1559 | struct object_info *oi, unsigned flags) | |
f0df4ed5 | 1560 | { |
cd585e2a | 1561 | static struct object_info blank_oi = OBJECT_INFO_INIT; |
9c8a294a | 1562 | struct cached_object *co; |
f0df4ed5 | 1563 | struct pack_entry e; |
5b086407 | 1564 | int rtype; |
b383a13c | 1565 | const struct object_id *real = oid; |
8b4c0103 | 1566 | int already_retried = 0; |
f0df4ed5 | 1567 | |
31877c9a | 1568 | |
b383a13c | 1569 | if (flags & OBJECT_INFO_LOOKUP_REPLACE) |
9d98354f | 1570 | real = lookup_replace_object(r, oid); |
f0df4ed5 | 1571 | |
b383a13c | 1572 | if (is_null_oid(real)) |
87b5e236 JK |
1573 | return -1; |
1574 | ||
cd585e2a JT |
1575 | if (!oi) |
1576 | oi = &blank_oi; | |
1577 | ||
9c8a294a JT |
1578 | co = find_cached_object(real); |
1579 | if (co) { | |
1580 | if (oi->typep) | |
1581 | *(oi->typep) = co->type; | |
1582 | if (oi->sizep) | |
1583 | *(oi->sizep) = co->size; | |
1584 | if (oi->disk_sizep) | |
1585 | *(oi->disk_sizep) = 0; | |
b99b6bcc JK |
1586 | if (oi->delta_base_oid) |
1587 | oidclr(oi->delta_base_oid); | |
9c8a294a JT |
1588 | if (oi->type_name) |
1589 | strbuf_addstr(oi->type_name, type_name(co->type)); | |
1590 | if (oi->contentp) | |
1591 | *oi->contentp = xmemdupz(co->buf, co->size); | |
1592 | oi->whence = OI_CACHED; | |
1593 | return 0; | |
c4d9986f NTND |
1594 | } |
1595 | ||
8b4c0103 | 1596 | while (1) { |
42c8ce1c | 1597 | if (find_pack_entry(r, real, &e)) |
8b4c0103 JT |
1598 | break; |
1599 | ||
ddd63e64 | 1600 | /* Most likely it's a loose object. */ |
514c5fdd | 1601 | if (!loose_object_info(r, real, oi, flags)) |
5b086407 | 1602 | return 0; |
ddd63e64 SG |
1603 | |
1604 | /* Not a loose object; someone else may have just packed it. */ | |
2b7750c9 | 1605 | if (!(flags & OBJECT_INFO_QUICK)) { |
9d98354f | 1606 | reprepare_packed_git(r); |
42c8ce1c | 1607 | if (find_pack_entry(r, real, &e)) |
2b7750c9 JT |
1608 | break; |
1609 | } | |
8b4c0103 | 1610 | |
eef71904 JT |
1611 | /* |
1612 | * If r is the_repository, this might be an attempt at | |
1613 | * accessing a submodule object as if it were in the_repository | |
1614 | * (having called add_submodule_odb() on that submodule's ODB). | |
1615 | * If any such ODBs exist, register them and try again. | |
1616 | */ | |
1617 | if (r == the_repository && | |
1618 | register_all_submodule_odb_as_alternates()) | |
a35e03de JT |
1619 | /* We added some alternates; retry */ |
1620 | continue; | |
1621 | ||
8b4c0103 | 1622 | /* Check if it is a missing object */ |
ef830cc4 JT |
1623 | if (fetch_if_missing && repo_has_promisor_remote(r) && |
1624 | !already_retried && | |
31f5256c | 1625 | !(flags & OBJECT_INFO_SKIP_FETCH_OBJECT)) { |
b14ed5ad | 1626 | promisor_remote_get_direct(r, real, 1); |
8b4c0103 JT |
1627 | already_retried = 1; |
1628 | continue; | |
dfdd4afc | 1629 | } |
8b4c0103 | 1630 | |
9e59b38c JT |
1631 | if (flags & OBJECT_INFO_DIE_IF_CORRUPT) { |
1632 | const struct packed_git *p; | |
1633 | if ((flags & OBJECT_INFO_LOOKUP_REPLACE) && !oideq(real, oid)) | |
1634 | die(_("replacement %s not found for %s"), | |
1635 | oid_to_hex(real), oid_to_hex(oid)); | |
1636 | if ((p = has_packed_and_bad(r, real))) | |
1637 | die(_("packed object %s (stored in %s) is corrupt"), | |
1638 | oid_to_hex(real), p->pack_name); | |
1639 | } | |
8b4c0103 | 1640 | return -1; |
f0df4ed5 | 1641 | } |
3d77d877 | 1642 | |
cd585e2a JT |
1643 | if (oi == &blank_oi) |
1644 | /* | |
1645 | * We know that the caller doesn't actually need the | |
1646 | * information below, so return early. | |
1647 | */ | |
1648 | return 0; | |
9d98354f | 1649 | rtype = packed_object_info(r, e.p, e.offset, oi); |
412916ee | 1650 | if (rtype < 0) { |
751530de | 1651 | mark_bad_packed_object(e.p, real); |
31877c9a | 1652 | return do_oid_object_info_extended(r, real, oi, 0); |
3ab0fb06 | 1653 | } else if (oi->whence == OI_PACKED) { |
9a490590 JH |
1654 | oi->u.packed.offset = e.offset; |
1655 | oi->u.packed.pack = e.p; | |
1656 | oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA || | |
1657 | rtype == OBJ_OFS_DELTA); | |
3d77d877 NP |
1658 | } |
1659 | ||
5b086407 | 1660 | return 0; |
f0df4ed5 JS |
1661 | } |
1662 | ||
31877c9a MT |
1663 | int oid_object_info_extended(struct repository *r, const struct object_id *oid, |
1664 | struct object_info *oi, unsigned flags) | |
1665 | { | |
1666 | int ret; | |
1667 | obj_read_lock(); | |
1668 | ret = do_oid_object_info_extended(r, oid, oi, flags); | |
1669 | obj_read_unlock(); | |
1670 | return ret; | |
1671 | } | |
1672 | ||
1673 | ||
3fc0dca9 | 1674 | /* returns enum object_type or negative */ |
9d98354f SB |
1675 | int oid_object_info(struct repository *r, |
1676 | const struct object_id *oid, | |
1677 | unsigned long *sizep) | |
9a490590 | 1678 | { |
5b086407 | 1679 | enum object_type type; |
27b5c1a0 | 1680 | struct object_info oi = OBJECT_INFO_INIT; |
9a490590 | 1681 | |
5b086407 | 1682 | oi.typep = &type; |
9a490590 | 1683 | oi.sizep = sizep; |
9d98354f SB |
1684 | if (oid_object_info_extended(r, oid, &oi, |
1685 | OBJECT_INFO_LOOKUP_REPLACE) < 0) | |
5b086407 JK |
1686 | return -1; |
1687 | return type; | |
9a490590 JH |
1688 | } |
1689 | ||
829e5c3b PO |
1690 | int pretend_object_file(void *buf, unsigned long len, enum object_type type, |
1691 | struct object_id *oid) | |
d66b37bb JH |
1692 | { |
1693 | struct cached_object *co; | |
1694 | ||
44439c1c | 1695 | hash_object_file(the_hash_algo, buf, len, type, oid); |
bc726bd0 | 1696 | if (repo_has_object_file_with_flags(the_repository, oid, OBJECT_INFO_QUICK | OBJECT_INFO_SKIP_FETCH_OBJECT) || |
a64d2aae | 1697 | find_cached_object(oid)) |
d66b37bb | 1698 | return 0; |
c7353967 | 1699 | ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc); |
d66b37bb JH |
1700 | co = &cached_objects[cached_object_nr++]; |
1701 | co->size = len; | |
21666f1a | 1702 | co->type = type; |
efa13f7b JH |
1703 | co->buf = xmalloc(len); |
1704 | memcpy(co->buf, buf, len); | |
62ba93ea | 1705 | oidcpy(&co->oid, oid); |
d66b37bb JH |
1706 | return 0; |
1707 | } | |
1708 | ||
b6c4cecc JH |
1709 | /* |
1710 | * This function dies on corrupt objects; the callers who want to | |
b25562e6 JK |
1711 | * deal with them should arrange to call oid_object_info_extended() and give |
1712 | * error messages themselves. | |
b6c4cecc | 1713 | */ |
0ba05cf2 JK |
1714 | void *repo_read_object_file(struct repository *r, |
1715 | const struct object_id *oid, | |
1716 | enum object_type *type, | |
1717 | unsigned long *size) | |
ac939109 | 1718 | { |
b25562e6 | 1719 | struct object_info oi = OBJECT_INFO_INIT; |
7be13f5f | 1720 | unsigned flags = OBJECT_INFO_DIE_IF_CORRUPT | OBJECT_INFO_LOOKUP_REPLACE; |
3ba7a065 | 1721 | void *data; |
b6c4cecc | 1722 | |
b25562e6 JK |
1723 | oi.typep = type; |
1724 | oi.sizep = size; | |
1725 | oi.contentp = &data; | |
b25562e6 | 1726 | if (oid_object_info_extended(r, oid, &oi, flags)) |
15b63689 | 1727 | return NULL; |
68095570 | 1728 | |
b25562e6 | 1729 | return data; |
ac939109 NP |
1730 | } |
1731 | ||
d3b4705a NTND |
1732 | void *read_object_with_reference(struct repository *r, |
1733 | const struct object_id *oid, | |
6aea6bae | 1734 | enum object_type required_type, |
40469ee9 | 1735 | unsigned long *size, |
02f0547e | 1736 | struct object_id *actual_oid_return) |
f4913f91 | 1737 | { |
6aea6bae | 1738 | enum object_type type; |
f4913f91 JH |
1739 | void *buffer; |
1740 | unsigned long isize; | |
02f0547e | 1741 | struct object_id actual_oid; |
f4913f91 | 1742 | |
02f0547e | 1743 | oidcpy(&actual_oid, oid); |
40469ee9 JH |
1744 | while (1) { |
1745 | int ref_length = -1; | |
1746 | const char *ref_type = NULL; | |
f4913f91 | 1747 | |
d3b4705a | 1748 | buffer = repo_read_object_file(r, &actual_oid, &type, &isize); |
40469ee9 JH |
1749 | if (!buffer) |
1750 | return NULL; | |
21666f1a | 1751 | if (type == required_type) { |
40469ee9 | 1752 | *size = isize; |
02f0547e | 1753 | if (actual_oid_return) |
1754 | oidcpy(actual_oid_return, &actual_oid); | |
40469ee9 JH |
1755 | return buffer; |
1756 | } | |
1757 | /* Handle references */ | |
21666f1a | 1758 | else if (type == OBJ_COMMIT) |
40469ee9 | 1759 | ref_type = "tree "; |
21666f1a | 1760 | else if (type == OBJ_TAG) |
40469ee9 JH |
1761 | ref_type = "object "; |
1762 | else { | |
1763 | free(buffer); | |
1764 | return NULL; | |
1765 | } | |
1766 | ref_length = strlen(ref_type); | |
f4913f91 | 1767 | |
94b5e093 | 1768 | if (ref_length + the_hash_algo->hexsz > isize || |
50974ec9 | 1769 | memcmp(buffer, ref_type, ref_length) || |
02f0547e | 1770 | get_oid_hex((char *) buffer + ref_length, &actual_oid)) { |
40469ee9 JH |
1771 | free(buffer); |
1772 | return NULL; | |
1773 | } | |
1cf58e72 | 1774 | free(buffer); |
40469ee9 | 1775 | /* Now we have the ID of the referred-to object in |
02f0547e | 1776 | * actual_oid. Check again. */ |
f4913f91 | 1777 | } |
f4913f91 JH |
1778 | } |
1779 | ||
2bbb28a3 ÆAB |
1780 | static void hash_object_body(const struct git_hash_algo *algo, git_hash_ctx *c, |
1781 | const void *buf, unsigned long len, | |
1782 | struct object_id *oid, | |
1783 | char *hdr, int *hdrlen) | |
1784 | { | |
1785 | algo->init_fn(c); | |
1786 | algo->update_fn(c, hdr, *hdrlen); | |
1787 | algo->update_fn(c, buf, len); | |
1788 | algo->final_oid_fn(oid, c); | |
1789 | } | |
1790 | ||
7ad5c44d MT |
1791 | static void write_object_file_prepare(const struct git_hash_algo *algo, |
1792 | const void *buf, unsigned long len, | |
2bbb28a3 | 1793 | enum object_type type, struct object_id *oid, |
a09c985e | 1794 | char *hdr, int *hdrlen) |
d410c0f5 | 1795 | { |
18e2588e | 1796 | git_hash_ctx c; |
d410c0f5 JH |
1797 | |
1798 | /* Generate the header */ | |
2bbb28a3 | 1799 | *hdrlen = format_object_header(hdr, *hdrlen, type, len); |
d410c0f5 JH |
1800 | |
1801 | /* Sha1.. */ | |
2bbb28a3 ÆAB |
1802 | hash_object_body(algo, &c, buf, len, oid, hdr, hdrlen); |
1803 | } | |
1804 | ||
1805 | static void write_object_file_prepare_literally(const struct git_hash_algo *algo, | |
1806 | const void *buf, unsigned long len, | |
1807 | const char *type, struct object_id *oid, | |
1808 | char *hdr, int *hdrlen) | |
1809 | { | |
1810 | git_hash_ctx c; | |
1811 | ||
1812 | *hdrlen = format_object_header_literally(hdr, *hdrlen, type, len); | |
1813 | hash_object_body(algo, &c, buf, len, oid, hdr, hdrlen); | |
d410c0f5 JH |
1814 | } |
1815 | ||
230f1322 | 1816 | /* |
5a688fe4 | 1817 | * Move the just written object into its final resting place. |
230f1322 | 1818 | */ |
cb5add58 | 1819 | int finalize_object_file(const char *tmpfile, const char *filename) |
230f1322 | 1820 | { |
e32c0a9c | 1821 | int ret = 0; |
5a688fe4 | 1822 | |
348df166 | 1823 | if (object_creation_mode == OBJECT_CREATION_USES_RENAMES) |
be66a6c4 JS |
1824 | goto try_rename; |
1825 | else if (link(tmpfile, filename)) | |
e32c0a9c | 1826 | ret = errno; |
7ebb6fca LT |
1827 | |
1828 | /* | |
1829 | * Coda hack - coda doesn't like cross-directory links, | |
1830 | * so we fall back to a rename, which will mean that it | |
1831 | * won't be able to check collisions, but that's not a | |
1832 | * big deal. | |
1833 | * | |
1834 | * The same holds for FAT formatted media. | |
1835 | * | |
3be1f18e | 1836 | * When this succeeds, we just return. We have nothing |
7ebb6fca LT |
1837 | * left to unlink. |
1838 | */ | |
1839 | if (ret && ret != EEXIST) { | |
be66a6c4 | 1840 | try_rename: |
7ebb6fca | 1841 | if (!rename(tmpfile, filename)) |
3be1f18e | 1842 | goto out; |
9e48b389 | 1843 | ret = errno; |
230f1322 | 1844 | } |
691f1a28 | 1845 | unlink_or_warn(tmpfile); |
230f1322 LT |
1846 | if (ret) { |
1847 | if (ret != EEXIST) { | |
2c319886 | 1848 | return error_errno(_("unable to write file %s"), filename); |
230f1322 LT |
1849 | } |
1850 | /* FIXME!!! Collision check here ? */ | |
1851 | } | |
1852 | ||
3be1f18e | 1853 | out: |
5256b006 | 1854 | if (adjust_shared_perm(filename)) |
259328b7 | 1855 | return error(_("unable to set permission to '%s'"), filename); |
230f1322 LT |
1856 | return 0; |
1857 | } | |
1858 | ||
44439c1c ÆAB |
1859 | static void hash_object_file_literally(const struct git_hash_algo *algo, |
1860 | const void *buf, unsigned long len, | |
1861 | const char *type, struct object_id *oid) | |
abdc3fc8 | 1862 | { |
1af64f73 | 1863 | char hdr[MAX_HEADER_LEN]; |
ef1286d3 | 1864 | int hdrlen = sizeof(hdr); |
44439c1c | 1865 | |
2bbb28a3 | 1866 | write_object_file_prepare_literally(algo, buf, len, type, oid, hdr, &hdrlen); |
abdc3fc8 RS |
1867 | } |
1868 | ||
44439c1c ÆAB |
1869 | void hash_object_file(const struct git_hash_algo *algo, const void *buf, |
1870 | unsigned long len, enum object_type type, | |
1871 | struct object_id *oid) | |
1872 | { | |
1873 | hash_object_file_literally(algo, buf, len, type_name(type), oid); | |
abdc3fc8 RS |
1874 | } |
1875 | ||
e9039dd3 | 1876 | /* Finalize a file on disk, and close it. */ |
c4e707f8 | 1877 | static void close_loose_object(int fd, const char *filename) |
e9039dd3 | 1878 | { |
020406ea NS |
1879 | if (the_repository->objects->odb->will_destroy) |
1880 | goto out; | |
b3cecf49 | 1881 | |
c0f4752e | 1882 | if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT)) |
83937e95 | 1883 | fsync_loose_object_bulk_checkin(fd, filename); |
c0f4752e | 1884 | else if (fsync_object_files > 0) |
c4e707f8 | 1885 | fsync_or_die(fd, filename); |
020406ea NS |
1886 | else |
1887 | fsync_component_or_die(FSYNC_COMPONENT_LOOSE_OBJECT, fd, | |
c4e707f8 | 1888 | filename); |
020406ea NS |
1889 | |
1890 | out: | |
e9039dd3 | 1891 | if (close(fd) != 0) |
76011357 | 1892 | die_errno(_("error when closing loose object file")); |
e9039dd3 LT |
1893 | } |
1894 | ||
5723fe7e LT |
1895 | /* Size of directory component, including the ending '/' */ |
1896 | static inline int directory_size(const char *filename) | |
1897 | { | |
1898 | const char *s = strrchr(filename, '/'); | |
1899 | if (!s) | |
1900 | return 0; | |
1901 | return s - filename + 1; | |
1902 | } | |
1903 | ||
1904 | /* | |
1905 | * This creates a temporary file in the same directory as the final | |
1906 | * 'filename' | |
1907 | * | |
1908 | * We want to avoid cross-directory filename renames, because those | |
1909 | * can have problems on various filesystems (FAT, NFS, Coda). | |
1910 | */ | |
d4b3d11a | 1911 | static int create_tmpfile(struct strbuf *tmp, const char *filename) |
5723fe7e LT |
1912 | { |
1913 | int fd, dirlen = directory_size(filename); | |
1914 | ||
d4b3d11a JK |
1915 | strbuf_reset(tmp); |
1916 | strbuf_add(tmp, filename, dirlen); | |
1917 | strbuf_addstr(tmp, "tmp_obj_XXXXXX"); | |
1918 | fd = git_mkstemp_mode(tmp->buf, 0444); | |
cbacbf4e | 1919 | if (fd < 0 && dirlen && errno == ENOENT) { |
d4b3d11a JK |
1920 | /* |
1921 | * Make sure the directory exists; note that the contents | |
1922 | * of the buffer are undefined after mkstemp returns an | |
1923 | * error, so we have to rewrite the whole buffer from | |
1924 | * scratch. | |
1925 | */ | |
1926 | strbuf_reset(tmp); | |
1927 | strbuf_add(tmp, filename, dirlen - 1); | |
1928 | if (mkdir(tmp->buf, 0777) && errno != EEXIST) | |
b2476a60 | 1929 | return -1; |
d4b3d11a | 1930 | if (adjust_shared_perm(tmp->buf)) |
5723fe7e LT |
1931 | return -1; |
1932 | ||
1933 | /* Try again */ | |
d4b3d11a JK |
1934 | strbuf_addstr(tmp, "/tmp_obj_XXXXXX"); |
1935 | fd = git_mkstemp_mode(tmp->buf, 0444); | |
5723fe7e LT |
1936 | } |
1937 | return fd; | |
1938 | } | |
1939 | ||
97a9db6f HX |
1940 | /** |
1941 | * Common steps for loose object writers to start writing loose | |
1942 | * objects: | |
1943 | * | |
1944 | * - Create tmpfile for the loose object. | |
1945 | * - Setup zlib stream for compression. | |
1946 | * - Start to feed header to zlib stream. | |
1947 | * | |
1948 | * Returns a "fd", which should later be provided to | |
1949 | * end_loose_object_common(). | |
1950 | */ | |
1951 | static int start_loose_object_common(struct strbuf *tmp_file, | |
1952 | const char *filename, unsigned flags, | |
1953 | git_zstream *stream, | |
1954 | unsigned char *buf, size_t buflen, | |
1955 | git_hash_ctx *c, | |
1956 | char *hdr, int hdrlen) | |
1957 | { | |
1958 | int fd; | |
1959 | ||
1960 | fd = create_tmpfile(tmp_file, filename); | |
1961 | if (fd < 0) { | |
1962 | if (flags & HASH_SILENT) | |
1963 | return -1; | |
1964 | else if (errno == EACCES) | |
1965 | return error(_("insufficient permission for adding " | |
1966 | "an object to repository database %s"), | |
1967 | get_object_directory()); | |
1968 | else | |
1969 | return error_errno( | |
1970 | _("unable to create temporary file")); | |
1971 | } | |
1972 | ||
1973 | /* Setup zlib stream for compression */ | |
1974 | git_deflate_init(stream, zlib_compression_level); | |
1975 | stream->next_out = buf; | |
1976 | stream->avail_out = buflen; | |
1977 | the_hash_algo->init_fn(c); | |
1978 | ||
1979 | /* Start to feed header to zlib stream */ | |
1980 | stream->next_in = (unsigned char *)hdr; | |
1981 | stream->avail_in = hdrlen; | |
1982 | while (git_deflate(stream, 0) == Z_OK) | |
1983 | ; /* nothing */ | |
1984 | the_hash_algo->update_fn(c, hdr, hdrlen); | |
1985 | ||
1986 | return fd; | |
1987 | } | |
1988 | ||
21e7d881 ÆAB |
1989 | /** |
1990 | * Common steps for the inner git_deflate() loop for writing loose | |
1991 | * objects. Returns what git_deflate() returns. | |
1992 | */ | |
1993 | static int write_loose_object_common(git_hash_ctx *c, | |
1994 | git_zstream *stream, const int flush, | |
1995 | unsigned char *in0, const int fd, | |
1996 | unsigned char *compressed, | |
1997 | const size_t compressed_len) | |
1998 | { | |
1999 | int ret; | |
2000 | ||
2001 | ret = git_deflate(stream, flush ? Z_FINISH : 0); | |
2002 | the_hash_algo->update_fn(c, in0, stream->next_in - in0); | |
d422d061 RS |
2003 | if (write_in_full(fd, compressed, stream->next_out - compressed) < 0) |
2004 | die_errno(_("unable to write loose object file")); | |
21e7d881 ÆAB |
2005 | stream->next_out = compressed; |
2006 | stream->avail_out = compressed_len; | |
2007 | ||
2008 | return ret; | |
2009 | } | |
2010 | ||
97a9db6f HX |
2011 | /** |
2012 | * Common steps for loose object writers to end writing loose objects: | |
2013 | * | |
2014 | * - End the compression of zlib stream. | |
2015 | * - Get the calculated oid to "oid". | |
2016 | */ | |
2017 | static int end_loose_object_common(git_hash_ctx *c, git_zstream *stream, | |
2018 | struct object_id *oid) | |
2019 | { | |
2020 | int ret; | |
2021 | ||
2022 | ret = git_deflate_end_gently(stream); | |
2023 | if (ret != Z_OK) | |
2024 | return ret; | |
2025 | the_hash_algo->final_oid_fn(oid, c); | |
2026 | ||
2027 | return Z_OK; | |
2028 | } | |
2029 | ||
3fc7281f PO |
2030 | static int write_loose_object(const struct object_id *oid, char *hdr, |
2031 | int hdrlen, const void *buf, unsigned long len, | |
4ef91a2d | 2032 | time_t mtime, unsigned flags) |
0fcfd160 | 2033 | { |
915308b1 | 2034 | int fd, ret; |
9892beba | 2035 | unsigned char compressed[4096]; |
ef49a7a0 | 2036 | git_zstream stream; |
18e2588e | 2037 | git_hash_ctx c; |
3fc7281f | 2038 | struct object_id parano_oid; |
d4b3d11a | 2039 | static struct strbuf tmp_file = STRBUF_INIT; |
ea657730 CC |
2040 | static struct strbuf filename = STRBUF_INIT; |
2041 | ||
c0f4752e NS |
2042 | if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT)) |
2043 | prepare_loose_object_bulk_checkin(); | |
2044 | ||
514c5fdd | 2045 | loose_object_path(the_repository, &filename, oid); |
a44c9a5e | 2046 | |
97a9db6f HX |
2047 | fd = start_loose_object_common(&tmp_file, filename.buf, flags, |
2048 | &stream, compressed, sizeof(compressed), | |
2049 | &c, hdr, hdrlen); | |
2050 | if (fd < 0) | |
2051 | return -1; | |
a44c9a5e LT |
2052 | |
2053 | /* Then the data itself.. */ | |
c00e657d | 2054 | stream.next_in = (void *)buf; |
a44c9a5e | 2055 | stream.avail_in = len; |
9892beba | 2056 | do { |
748af44c | 2057 | unsigned char *in0 = stream.next_in; |
21e7d881 ÆAB |
2058 | |
2059 | ret = write_loose_object_common(&c, &stream, 1, in0, fd, | |
2060 | compressed, sizeof(compressed)); | |
9892beba NP |
2061 | } while (ret == Z_OK); |
2062 | ||
ac54c277 | 2063 | if (ret != Z_STREAM_END) |
259328b7 | 2064 | die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid), |
3fc7281f | 2065 | ret); |
97a9db6f | 2066 | ret = end_loose_object_common(&c, &stream, ¶no_oid); |
ac54c277 | 2067 | if (ret != Z_OK) |
259328b7 | 2068 | die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid), |
3fc7281f | 2069 | ret); |
9001dc2a | 2070 | if (!oideq(oid, ¶no_oid)) |
259328b7 | 2071 | die(_("confused by unstable object source data for %s"), |
3fc7281f | 2072 | oid_to_hex(oid)); |
ac54c277 | 2073 | |
c4e707f8 | 2074 | close_loose_object(fd, tmp_file.buf); |
0fcfd160 | 2075 | |
bbac7311 NP |
2076 | if (mtime) { |
2077 | struct utimbuf utb; | |
2078 | utb.actime = mtime; | |
2079 | utb.modtime = mtime; | |
4ef91a2d ÆAB |
2080 | if (utime(tmp_file.buf, &utb) < 0 && |
2081 | !(flags & HASH_SILENT)) | |
259328b7 | 2082 | warning_errno(_("failed utime() on %s"), tmp_file.buf); |
bbac7311 NP |
2083 | } |
2084 | ||
ea657730 | 2085 | return finalize_object_file(tmp_file.buf, filename.buf); |
0fcfd160 | 2086 | } |
8237b185 | 2087 | |
6862ebbf | 2088 | static int freshen_loose_object(const struct object_id *oid) |
33d4221c | 2089 | { |
6862ebbf | 2090 | return check_and_freshen(oid, 1); |
33d4221c JK |
2091 | } |
2092 | ||
6862ebbf | 2093 | static int freshen_packed_object(const struct object_id *oid) |
33d4221c JK |
2094 | { |
2095 | struct pack_entry e; | |
544443cb | 2096 | if (!find_pack_entry(the_repository, oid, &e)) |
ee1c6c34 | 2097 | return 0; |
a6131642 TB |
2098 | if (e.p->is_cruft) |
2099 | return 0; | |
ee1c6c34 JK |
2100 | if (e.p->freshened) |
2101 | return 1; | |
2102 | if (!freshen_file(e.p->pack_name)) | |
2103 | return 0; | |
2104 | e.p->freshened = 1; | |
2105 | return 1; | |
33d4221c JK |
2106 | } |
2107 | ||
2b6070ac HX |
2108 | int stream_loose_object(struct input_stream *in_stream, size_t len, |
2109 | struct object_id *oid) | |
2110 | { | |
2111 | int fd, ret, err = 0, flush = 0; | |
2112 | unsigned char compressed[4096]; | |
2113 | git_zstream stream; | |
2114 | git_hash_ctx c; | |
2115 | struct strbuf tmp_file = STRBUF_INIT; | |
2116 | struct strbuf filename = STRBUF_INIT; | |
2117 | int dirlen; | |
2118 | char hdr[MAX_HEADER_LEN]; | |
2119 | int hdrlen; | |
2120 | ||
2121 | if (batch_fsync_enabled(FSYNC_COMPONENT_LOOSE_OBJECT)) | |
2122 | prepare_loose_object_bulk_checkin(); | |
2123 | ||
2124 | /* Since oid is not determined, save tmp file to odb path. */ | |
2125 | strbuf_addf(&filename, "%s/", get_object_directory()); | |
2126 | hdrlen = format_object_header(hdr, sizeof(hdr), OBJ_BLOB, len); | |
2127 | ||
2128 | /* | |
2129 | * Common steps for write_loose_object and stream_loose_object to | |
2130 | * start writing loose objects: | |
2131 | * | |
2132 | * - Create tmpfile for the loose object. | |
2133 | * - Setup zlib stream for compression. | |
2134 | * - Start to feed header to zlib stream. | |
2135 | */ | |
2136 | fd = start_loose_object_common(&tmp_file, filename.buf, 0, | |
2137 | &stream, compressed, sizeof(compressed), | |
2138 | &c, hdr, hdrlen); | |
2139 | if (fd < 0) { | |
2140 | err = -1; | |
2141 | goto cleanup; | |
2142 | } | |
2143 | ||
2144 | /* Then the data itself.. */ | |
2145 | do { | |
2146 | unsigned char *in0 = stream.next_in; | |
2147 | ||
2148 | if (!stream.avail_in && !in_stream->is_finished) { | |
2149 | const void *in = in_stream->read(in_stream, &stream.avail_in); | |
2150 | stream.next_in = (void *)in; | |
2151 | in0 = (unsigned char *)in; | |
2152 | /* All data has been read. */ | |
2153 | if (in_stream->is_finished) | |
2154 | flush = 1; | |
2155 | } | |
2156 | ret = write_loose_object_common(&c, &stream, flush, in0, fd, | |
2157 | compressed, sizeof(compressed)); | |
2158 | /* | |
2159 | * Unlike write_loose_object(), we do not have the entire | |
2160 | * buffer. If we get Z_BUF_ERROR due to too few input bytes, | |
2161 | * then we'll replenish them in the next input_stream->read() | |
2162 | * call when we loop. | |
2163 | */ | |
2164 | } while (ret == Z_OK || ret == Z_BUF_ERROR); | |
2165 | ||
2166 | if (stream.total_in != len + hdrlen) | |
2167 | die(_("write stream object %ld != %"PRIuMAX), stream.total_in, | |
2168 | (uintmax_t)len + hdrlen); | |
2169 | ||
2170 | /* | |
2171 | * Common steps for write_loose_object and stream_loose_object to | |
2172 | * end writing loose oject: | |
2173 | * | |
2174 | * - End the compression of zlib stream. | |
2175 | * - Get the calculated oid. | |
2176 | */ | |
2177 | if (ret != Z_STREAM_END) | |
2178 | die(_("unable to stream deflate new object (%d)"), ret); | |
2179 | ret = end_loose_object_common(&c, &stream, oid); | |
2180 | if (ret != Z_OK) | |
2181 | die(_("deflateEnd on stream object failed (%d)"), ret); | |
2182 | close_loose_object(fd, tmp_file.buf); | |
2183 | ||
2184 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) { | |
2185 | unlink_or_warn(tmp_file.buf); | |
2186 | goto cleanup; | |
2187 | } | |
2188 | ||
2189 | loose_object_path(the_repository, &filename, oid); | |
2190 | ||
2191 | /* We finally know the object path, and create the missing dir. */ | |
2192 | dirlen = directory_size(filename.buf); | |
2193 | if (dirlen) { | |
2194 | struct strbuf dir = STRBUF_INIT; | |
2195 | strbuf_add(&dir, filename.buf, dirlen); | |
2196 | ||
2197 | if (mkdir_in_gitdir(dir.buf) && errno != EEXIST) { | |
2198 | err = error_errno(_("unable to create directory %s"), dir.buf); | |
2199 | strbuf_release(&dir); | |
2200 | goto cleanup; | |
2201 | } | |
2202 | strbuf_release(&dir); | |
2203 | } | |
2204 | ||
2205 | err = finalize_object_file(tmp_file.buf, filename.buf); | |
2206 | cleanup: | |
2207 | strbuf_release(&tmp_file); | |
2208 | strbuf_release(&filename); | |
2209 | return err; | |
2210 | } | |
2211 | ||
4ef91a2d | 2212 | int write_object_file_flags(const void *buf, unsigned long len, |
c80d226a | 2213 | enum object_type type, struct object_id *oid, |
4ef91a2d | 2214 | unsigned flags) |
bbac7311 | 2215 | { |
1af64f73 | 2216 | char hdr[MAX_HEADER_LEN]; |
ef1286d3 | 2217 | int hdrlen = sizeof(hdr); |
bbac7311 NP |
2218 | |
2219 | /* Normally if we have it in the pack then we do not bother writing | |
2220 | * it out into .git/objects/??/?{38} file. | |
2221 | */ | |
7ad5c44d MT |
2222 | write_object_file_prepare(the_hash_algo, buf, len, type, oid, hdr, |
2223 | &hdrlen); | |
6862ebbf | 2224 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) |
bbac7311 | 2225 | return 0; |
4ef91a2d | 2226 | return write_loose_object(oid, hdr, hdrlen, buf, len, 0, flags); |
bbac7311 NP |
2227 | } |
2228 | ||
0ff7b4f9 ÆAB |
2229 | int write_object_file_literally(const void *buf, unsigned long len, |
2230 | const char *type, struct object_id *oid, | |
2231 | unsigned flags) | |
0c3db67c ES |
2232 | { |
2233 | char *header; | |
2234 | int hdrlen, status = 0; | |
2235 | ||
2236 | /* type string, SP, %lu of the length plus NUL must fit this */ | |
1af64f73 | 2237 | hdrlen = strlen(type) + MAX_HEADER_LEN; |
ef1286d3 | 2238 | header = xmalloc(hdrlen); |
2bbb28a3 ÆAB |
2239 | write_object_file_prepare_literally(the_hash_algo, buf, len, type, |
2240 | oid, header, &hdrlen); | |
0c3db67c ES |
2241 | |
2242 | if (!(flags & HASH_WRITE_OBJECT)) | |
2243 | goto cleanup; | |
6862ebbf | 2244 | if (freshen_packed_object(oid) || freshen_loose_object(oid)) |
0c3db67c | 2245 | goto cleanup; |
4ef91a2d | 2246 | status = write_loose_object(oid, header, hdrlen, buf, len, 0, 0); |
0c3db67c ES |
2247 | |
2248 | cleanup: | |
2249 | free(header); | |
2250 | return status; | |
2251 | } | |
2252 | ||
4bdb70a4 | 2253 | int force_object_loose(const struct object_id *oid, time_t mtime) |
bbac7311 | 2254 | { |
bbac7311 NP |
2255 | void *buf; |
2256 | unsigned long len; | |
b25562e6 | 2257 | struct object_info oi = OBJECT_INFO_INIT; |
bbac7311 | 2258 | enum object_type type; |
1af64f73 | 2259 | char hdr[MAX_HEADER_LEN]; |
bbac7311 | 2260 | int hdrlen; |
1fb23e65 | 2261 | int ret; |
bbac7311 | 2262 | |
6862ebbf | 2263 | if (has_loose_object(oid)) |
bbac7311 | 2264 | return 0; |
b25562e6 JK |
2265 | oi.typep = &type; |
2266 | oi.sizep = &len; | |
2267 | oi.contentp = &buf; | |
2268 | if (oid_object_info_extended(the_repository, oid, &oi, 0)) | |
2c319886 | 2269 | return error(_("cannot read object for %s"), oid_to_hex(oid)); |
b04cdea4 | 2270 | hdrlen = format_object_header(hdr, sizeof(hdr), type, len); |
4ef91a2d | 2271 | ret = write_loose_object(oid, hdr, hdrlen, buf, len, mtime, 0); |
1fb23e65 BS |
2272 | free(buf); |
2273 | ||
2274 | return ret; | |
bbac7311 NP |
2275 | } |
2276 | ||
1d8d9cb6 JT |
2277 | int has_object(struct repository *r, const struct object_id *oid, |
2278 | unsigned flags) | |
2279 | { | |
2280 | int quick = !(flags & HAS_OBJECT_RECHECK_PACKED); | |
2281 | unsigned object_info_flags = OBJECT_INFO_SKIP_FETCH_OBJECT | | |
2282 | (quick ? OBJECT_INFO_QUICK : 0); | |
2283 | ||
2284 | if (!startup_info->have_repository) | |
2285 | return 0; | |
2286 | return oid_object_info_extended(r, oid, NULL, object_info_flags) >= 0; | |
2287 | } | |
2288 | ||
cba595ab JH |
2289 | int repo_has_object_file_with_flags(struct repository *r, |
2290 | const struct object_id *oid, int flags) | |
8237b185 | 2291 | { |
3e8b7d3c JN |
2292 | if (!startup_info->have_repository) |
2293 | return 0; | |
9c8a294a | 2294 | return oid_object_info_extended(r, oid, NULL, flags) >= 0; |
b419aa25 | 2295 | } |
2296 | ||
9b45f499 SB |
2297 | int repo_has_object_file(struct repository *r, |
2298 | const struct object_id *oid) | |
b419aa25 | 2299 | { |
cba595ab | 2300 | return repo_has_object_file_with_flags(r, oid, 0); |
5827a035 JK |
2301 | } |
2302 | ||
69bbbe48 JK |
2303 | /* |
2304 | * We can't use the normal fsck_error_function() for index_mem(), | |
2305 | * because we don't yet have a valid oid for it to report. Instead, | |
2306 | * report the minimal fsck error here, and rely on the caller to | |
2307 | * give more context. | |
2308 | */ | |
0b4e9013 JK |
2309 | static int hash_format_check_report(struct fsck_options *opts UNUSED, |
2310 | const struct object_id *oid UNUSED, | |
2311 | enum object_type object_type UNUSED, | |
2312 | enum fsck_msg_type msg_type UNUSED, | |
2313 | enum fsck_msg_id msg_id UNUSED, | |
69bbbe48 JK |
2314 | const char *message) |
2315 | { | |
2316 | error(_("object fails fsck: %s"), message); | |
2317 | return 1; | |
c879daa2 NTND |
2318 | } |
2319 | ||
58bf2a4c NTND |
2320 | static int index_mem(struct index_state *istate, |
2321 | struct object_id *oid, void *buf, size_t size, | |
c4ce46fc JH |
2322 | enum object_type type, |
2323 | const char *path, unsigned flags) | |
e7332f96 | 2324 | { |
63e05f90 | 2325 | int ret = 0; |
bbea0dde | 2326 | int re_allocated = 0; |
c4ce46fc | 2327 | int write_object = flags & HASH_WRITE_OBJECT; |
74400e71 | 2328 | |
7672db20 | 2329 | if (!type) |
edaec3fb | 2330 | type = OBJ_BLOB; |
6c510bee LT |
2331 | |
2332 | /* | |
2333 | * Convert blobs to git internal format | |
2334 | */ | |
43df4f86 | 2335 | if ((type == OBJ_BLOB) && path) { |
f285a2d7 | 2336 | struct strbuf nbuf = STRBUF_INIT; |
58bf2a4c | 2337 | if (convert_to_git(istate, path, buf, size, &nbuf, |
8462ff43 | 2338 | get_conv_flags(flags))) { |
b315c5c0 | 2339 | buf = strbuf_detach(&nbuf, &size); |
6c510bee LT |
2340 | re_allocated = 1; |
2341 | } | |
2342 | } | |
c4ce46fc | 2343 | if (flags & HASH_FORMAT_CHECK) { |
69bbbe48 JK |
2344 | struct fsck_options opts = FSCK_OPTIONS_DEFAULT; |
2345 | ||
2346 | opts.strict = 1; | |
2347 | opts.error_func = hash_format_check_report; | |
2348 | if (fsck_buffer(null_oid(), type, buf, size, &opts)) | |
2349 | die(_("refusing to create malformed object")); | |
2350 | fsck_finish(&opts); | |
c879daa2 | 2351 | } |
6c510bee | 2352 | |
7672db20 | 2353 | if (write_object) |
c80d226a | 2354 | ret = write_object_file(buf, size, type, oid); |
abdc3fc8 | 2355 | else |
44439c1c | 2356 | hash_object_file(the_hash_algo, buf, size, type, oid); |
43df4f86 | 2357 | if (re_allocated) |
6c510bee | 2358 | free(buf); |
43df4f86 DP |
2359 | return ret; |
2360 | } | |
2361 | ||
58bf2a4c NTND |
2362 | static int index_stream_convert_blob(struct index_state *istate, |
2363 | struct object_id *oid, | |
2364 | int fd, | |
2365 | const char *path, | |
2366 | unsigned flags) | |
9035d75a | 2367 | { |
63e05f90 | 2368 | int ret = 0; |
9035d75a SP |
2369 | const int write_object = flags & HASH_WRITE_OBJECT; |
2370 | struct strbuf sbuf = STRBUF_INIT; | |
2371 | ||
2372 | assert(path); | |
58bf2a4c | 2373 | assert(would_convert_to_git_filter_fd(istate, path)); |
9035d75a | 2374 | |
58bf2a4c | 2375 | convert_to_git_filter_fd(istate, path, fd, &sbuf, |
8462ff43 | 2376 | get_conv_flags(flags)); |
9035d75a SP |
2377 | |
2378 | if (write_object) | |
c80d226a | 2379 | ret = write_object_file(sbuf.buf, sbuf.len, OBJ_BLOB, |
a09c985e | 2380 | oid); |
9035d75a | 2381 | else |
44439c1c ÆAB |
2382 | hash_object_file(the_hash_algo, sbuf.buf, sbuf.len, OBJ_BLOB, |
2383 | oid); | |
9035d75a SP |
2384 | strbuf_release(&sbuf); |
2385 | return ret; | |
2386 | } | |
2387 | ||
58bf2a4c NTND |
2388 | static int index_pipe(struct index_state *istate, struct object_id *oid, |
2389 | int fd, enum object_type type, | |
7b41e1e1 JH |
2390 | const char *path, unsigned flags) |
2391 | { | |
2392 | struct strbuf sbuf = STRBUF_INIT; | |
2393 | int ret; | |
2394 | ||
2395 | if (strbuf_read(&sbuf, fd, 4096) >= 0) | |
58bf2a4c | 2396 | ret = index_mem(istate, oid, sbuf.buf, sbuf.len, type, path, flags); |
7b41e1e1 JH |
2397 | else |
2398 | ret = -1; | |
2399 | strbuf_release(&sbuf); | |
2400 | return ret; | |
2401 | } | |
2402 | ||
ea68b0ce DP |
2403 | #define SMALL_FILE_SIZE (32*1024) |
2404 | ||
58bf2a4c NTND |
2405 | static int index_core(struct index_state *istate, |
2406 | struct object_id *oid, int fd, size_t size, | |
7b41e1e1 JH |
2407 | enum object_type type, const char *path, |
2408 | unsigned flags) | |
43df4f86 DP |
2409 | { |
2410 | int ret; | |
43df4f86 | 2411 | |
7b41e1e1 | 2412 | if (!size) { |
58bf2a4c | 2413 | ret = index_mem(istate, oid, "", size, type, path, flags); |
ea68b0ce DP |
2414 | } else if (size <= SMALL_FILE_SIZE) { |
2415 | char *buf = xmalloc(size); | |
90dca671 JK |
2416 | ssize_t read_result = read_in_full(fd, buf, size); |
2417 | if (read_result < 0) | |
259328b7 | 2418 | ret = error_errno(_("read error while indexing %s"), |
90dca671 JK |
2419 | path ? path : "<unknown>"); |
2420 | else if (read_result != size) | |
259328b7 | 2421 | ret = error(_("short read while indexing %s"), |
90dca671 | 2422 | path ? path : "<unknown>"); |
ea68b0ce | 2423 | else |
58bf2a4c | 2424 | ret = index_mem(istate, oid, buf, size, type, path, flags); |
ea68b0ce | 2425 | free(buf); |
08bda208 | 2426 | } else { |
43df4f86 | 2427 | void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0); |
58bf2a4c | 2428 | ret = index_mem(istate, oid, buf, size, type, path, flags); |
aac17941 | 2429 | munmap(buf, size); |
08bda208 | 2430 | } |
7b41e1e1 JH |
2431 | return ret; |
2432 | } | |
2433 | ||
4dd1fbc7 | 2434 | /* |
568508e7 JH |
2435 | * This creates one packfile per large blob unless bulk-checkin |
2436 | * machinery is "plugged". | |
4dd1fbc7 JH |
2437 | * |
2438 | * This also bypasses the usual "convert-to-git" dance, and that is on | |
2439 | * purpose. We could write a streaming version of the converting | |
2440 | * functions and insert that before feeding the data to fast-import | |
4f22b101 JK |
2441 | * (or equivalent in-core API described above). However, that is |
2442 | * somewhat complicated, as we do not know the size of the filter | |
2443 | * result, which we need to know beforehand when writing a git object. | |
2444 | * Since the primary motivation for trying to stream from the working | |
2445 | * tree file and to avoid mmaping it in core is to deal with large | |
2446 | * binary blobs, they generally do not want to get any conversion, and | |
2447 | * callers should avoid this code path when filters are requested. | |
4dd1fbc7 | 2448 | */ |
7d5e1dc3 | 2449 | static int index_stream(struct object_id *oid, int fd, size_t size, |
4dd1fbc7 JH |
2450 | enum object_type type, const char *path, |
2451 | unsigned flags) | |
2452 | { | |
68ee6dfc | 2453 | return index_bulk_checkin(oid, fd, size, type, path, flags); |
4dd1fbc7 JH |
2454 | } |
2455 | ||
58bf2a4c NTND |
2456 | int index_fd(struct index_state *istate, struct object_id *oid, |
2457 | int fd, struct stat *st, | |
7b41e1e1 JH |
2458 | enum object_type type, const char *path, unsigned flags) |
2459 | { | |
2460 | int ret; | |
7b41e1e1 | 2461 | |
9079ab7c SP |
2462 | /* |
2463 | * Call xsize_t() only when needed to avoid potentially unnecessary | |
2464 | * die() for large files. | |
2465 | */ | |
58bf2a4c NTND |
2466 | if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(istate, path)) |
2467 | ret = index_stream_convert_blob(istate, oid, fd, path, flags); | |
9035d75a | 2468 | else if (!S_ISREG(st->st_mode)) |
58bf2a4c | 2469 | ret = index_pipe(istate, oid, fd, type, path, flags); |
9079ab7c | 2470 | else if (st->st_size <= big_file_threshold || type != OBJ_BLOB || |
58bf2a4c NTND |
2471 | (path && would_convert_to_git(istate, path))) |
2472 | ret = index_core(istate, oid, fd, xsize_t(st->st_size), | |
2473 | type, path, flags); | |
4dd1fbc7 | 2474 | else |
7d5e1dc3 | 2475 | ret = index_stream(oid, fd, xsize_t(st->st_size), type, path, |
9079ab7c | 2476 | flags); |
43df4f86 | 2477 | close(fd); |
aac17941 | 2478 | return ret; |
74400e71 | 2479 | } |
ec1fcc16 | 2480 | |
58bf2a4c NTND |
2481 | int index_path(struct index_state *istate, struct object_id *oid, |
2482 | const char *path, struct stat *st, unsigned flags) | |
ec1fcc16 JH |
2483 | { |
2484 | int fd; | |
b760d3aa | 2485 | struct strbuf sb = STRBUF_INIT; |
ea8e0297 | 2486 | int rc = 0; |
ec1fcc16 JH |
2487 | |
2488 | switch (st->st_mode & S_IFMT) { | |
2489 | case S_IFREG: | |
2490 | fd = open(path, O_RDONLY); | |
2491 | if (fd < 0) | |
7616c6ca | 2492 | return error_errno("open(\"%s\")", path); |
58bf2a4c | 2493 | if (index_fd(istate, oid, fd, st, OBJ_BLOB, path, flags) < 0) |
259328b7 | 2494 | return error(_("%s: failed to insert into database"), |
ec1fcc16 JH |
2495 | path); |
2496 | break; | |
2497 | case S_IFLNK: | |
7616c6ca NTND |
2498 | if (strbuf_readlink(&sb, path, st->st_size)) |
2499 | return error_errno("readlink(\"%s\")", path); | |
c4ce46fc | 2500 | if (!(flags & HASH_WRITE_OBJECT)) |
2dcde20e | 2501 | hash_object_file(the_hash_algo, sb.buf, sb.len, |
44439c1c | 2502 | OBJ_BLOB, oid); |
c80d226a | 2503 | else if (write_object_file(sb.buf, sb.len, OBJ_BLOB, oid)) |
259328b7 | 2504 | rc = error(_("%s: failed to insert into database"), path); |
b760d3aa | 2505 | strbuf_release(&sb); |
ec1fcc16 | 2506 | break; |
f35a6d3b | 2507 | case S_IFDIR: |
a98e6101 | 2508 | return resolve_gitlink_ref(path, "HEAD", oid); |
ec1fcc16 | 2509 | default: |
259328b7 | 2510 | return error(_("%s: unsupported file type"), path); |
ec1fcc16 | 2511 | } |
ea8e0297 | 2512 | return rc; |
ec1fcc16 | 2513 | } |
a69e5429 JH |
2514 | |
2515 | int read_pack_header(int fd, struct pack_header *header) | |
2516 | { | |
f48ecd38 | 2517 | if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header)) |
c697ad14 HO |
2518 | /* "eof before pack header was fully read" */ |
2519 | return PH_ERROR_EOF; | |
2520 | ||
a69e5429 JH |
2521 | if (header->hdr_signature != htonl(PACK_SIGNATURE)) |
2522 | /* "protocol error (pack signature mismatch detected)" */ | |
2523 | return PH_ERROR_PACK_SIGNATURE; | |
2524 | if (!pack_version_ok(header->hdr_version)) | |
2525 | /* "protocol error (pack version unsupported)" */ | |
2526 | return PH_ERROR_PROTOCOL; | |
2527 | return 0; | |
2528 | } | |
40d52ff7 | 2529 | |
e816caa0 | 2530 | void assert_oid_type(const struct object_id *oid, enum object_type expect) |
40d52ff7 | 2531 | { |
0df8e965 | 2532 | enum object_type type = oid_object_info(the_repository, oid, NULL); |
40d52ff7 | 2533 | if (type < 0) |
259328b7 | 2534 | die(_("%s is not a valid object"), oid_to_hex(oid)); |
40d52ff7 | 2535 | if (type != expect) |
259328b7 | 2536 | die(_("%s is not a valid '%s' object"), oid_to_hex(oid), |
debca9d2 | 2537 | type_name(expect)); |
40d52ff7 | 2538 | } |
27e1e22d | 2539 | |
70c49050 | 2540 | int for_each_file_in_obj_subdir(unsigned int subdir_nr, |
cc817ca3 RS |
2541 | struct strbuf *path, |
2542 | each_loose_object_fn obj_cb, | |
2543 | each_loose_cruft_fn cruft_cb, | |
2544 | each_loose_subdir_fn subdir_cb, | |
2545 | void *data) | |
27e1e22d | 2546 | { |
0375f472 RS |
2547 | size_t origlen, baselen; |
2548 | DIR *dir; | |
27e1e22d JK |
2549 | struct dirent *de; |
2550 | int r = 0; | |
62a24c89 | 2551 | struct object_id oid; |
27e1e22d | 2552 | |
70c49050 RS |
2553 | if (subdir_nr > 0xff) |
2554 | BUG("invalid loose object subdirectory: %x", subdir_nr); | |
2555 | ||
0375f472 RS |
2556 | origlen = path->len; |
2557 | strbuf_complete(path, '/'); | |
2558 | strbuf_addf(path, "%02x", subdir_nr); | |
0375f472 RS |
2559 | |
2560 | dir = opendir(path->buf); | |
27e1e22d | 2561 | if (!dir) { |
0375f472 | 2562 | if (errno != ENOENT) |
259328b7 | 2563 | r = error_errno(_("unable to open %s"), path->buf); |
0375f472 RS |
2564 | strbuf_setlen(path, origlen); |
2565 | return r; | |
27e1e22d JK |
2566 | } |
2567 | ||
62a24c89 | 2568 | oid.hash[0] = subdir_nr; |
163ee5e6 DS |
2569 | strbuf_addch(path, '/'); |
2570 | baselen = path->len; | |
62a24c89 | 2571 | |
b548f0f1 | 2572 | while ((de = readdir_skip_dot_and_dotdot(dir))) { |
163ee5e6 | 2573 | size_t namelen; |
27e1e22d | 2574 | |
163ee5e6 | 2575 | namelen = strlen(de->d_name); |
27e1e22d | 2576 | strbuf_setlen(path, baselen); |
163ee5e6 | 2577 | strbuf_add(path, de->d_name, namelen); |
94b5e093 | 2578 | if (namelen == the_hash_algo->hexsz - 2 && |
62a24c89 | 2579 | !hex_to_bytes(oid.hash + 1, de->d_name, |
94b5e093 | 2580 | the_hash_algo->rawsz - 1)) { |
5a6dce70 | 2581 | oid_set_algo(&oid, the_hash_algo); |
62a24c89 RS |
2582 | if (obj_cb) { |
2583 | r = obj_cb(&oid, path->buf, data); | |
2584 | if (r) | |
2585 | break; | |
27e1e22d | 2586 | } |
62a24c89 | 2587 | continue; |
27e1e22d JK |
2588 | } |
2589 | ||
2590 | if (cruft_cb) { | |
2591 | r = cruft_cb(de->d_name, path->buf, data); | |
2592 | if (r) | |
2593 | break; | |
2594 | } | |
2595 | } | |
094c7e63 | 2596 | closedir(dir); |
27e1e22d | 2597 | |
163ee5e6 | 2598 | strbuf_setlen(path, baselen - 1); |
27e1e22d JK |
2599 | if (!r && subdir_cb) |
2600 | r = subdir_cb(subdir_nr, path->buf, data); | |
2601 | ||
0375f472 RS |
2602 | strbuf_setlen(path, origlen); |
2603 | ||
27e1e22d JK |
2604 | return r; |
2605 | } | |
2606 | ||
e6f875e0 | 2607 | int for_each_loose_file_in_objdir_buf(struct strbuf *path, |
27e1e22d JK |
2608 | each_loose_object_fn obj_cb, |
2609 | each_loose_cruft_fn cruft_cb, | |
2610 | each_loose_subdir_fn subdir_cb, | |
2611 | void *data) | |
2612 | { | |
27e1e22d JK |
2613 | int r = 0; |
2614 | int i; | |
2615 | ||
27e1e22d | 2616 | for (i = 0; i < 256; i++) { |
e6f875e0 | 2617 | r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb, |
27e1e22d | 2618 | subdir_cb, data); |
27e1e22d JK |
2619 | if (r) |
2620 | break; | |
2621 | } | |
2622 | ||
e6f875e0 JK |
2623 | return r; |
2624 | } | |
2625 | ||
2626 | int for_each_loose_file_in_objdir(const char *path, | |
2627 | each_loose_object_fn obj_cb, | |
2628 | each_loose_cruft_fn cruft_cb, | |
2629 | each_loose_subdir_fn subdir_cb, | |
2630 | void *data) | |
2631 | { | |
2632 | struct strbuf buf = STRBUF_INIT; | |
2633 | int r; | |
2634 | ||
2635 | strbuf_addstr(&buf, path); | |
2636 | r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb, | |
2637 | subdir_cb, data); | |
27e1e22d | 2638 | strbuf_release(&buf); |
e6f875e0 | 2639 | |
27e1e22d JK |
2640 | return r; |
2641 | } | |
660c889e | 2642 | |
a7ff6f5a JK |
2643 | int for_each_loose_object(each_loose_object_fn cb, void *data, |
2644 | enum for_each_object_flags flags) | |
660c889e | 2645 | { |
f0eaf638 | 2646 | struct object_directory *odb; |
b0a42642 | 2647 | |
f0eaf638 JK |
2648 | prepare_alt_odb(the_repository); |
2649 | for (odb = the_repository->objects->odb; odb; odb = odb->next) { | |
2650 | int r = for_each_loose_file_in_objdir(odb->path, cb, NULL, | |
2651 | NULL, data); | |
2652 | if (r) | |
2653 | return r; | |
660c889e | 2654 | |
f0eaf638 JK |
2655 | if (flags & FOR_EACH_OBJECT_LOCAL_ONLY) |
2656 | break; | |
2657 | } | |
1385bb7b | 2658 | |
f0eaf638 | 2659 | return 0; |
660c889e JK |
2660 | } |
2661 | ||
be252d33 JK |
2662 | static int append_loose_object(const struct object_id *oid, |
2663 | const char *path UNUSED, | |
3a2e0824 | 2664 | void *data) |
660c889e | 2665 | { |
92d8ed8a | 2666 | oidtree_insert(data, oid); |
3a2e0824 JK |
2667 | return 0; |
2668 | } | |
660c889e | 2669 | |
92d8ed8a | 2670 | struct oidtree *odb_loose_cache(struct object_directory *odb, |
0000d654 RS |
2671 | const struct object_id *oid) |
2672 | { | |
2673 | int subdir_nr = oid->hash[0]; | |
3a2e0824 | 2674 | struct strbuf buf = STRBUF_INIT; |
33f379ee EW |
2675 | size_t word_bits = bitsizeof(odb->loose_objects_subdir_seen[0]); |
2676 | size_t word_index = subdir_nr / word_bits; | |
26de1fc0 | 2677 | size_t mask = (size_t)1u << (subdir_nr % word_bits); |
33f379ee | 2678 | uint32_t *bitmap; |
660c889e | 2679 | |
3a2e0824 | 2680 | if (subdir_nr < 0 || |
33f379ee | 2681 | subdir_nr >= bitsizeof(odb->loose_objects_subdir_seen)) |
3a2e0824 | 2682 | BUG("subdir_nr out of range"); |
1385bb7b | 2683 | |
33f379ee EW |
2684 | bitmap = &odb->loose_objects_subdir_seen[word_index]; |
2685 | if (*bitmap & mask) | |
92d8ed8a EW |
2686 | return odb->loose_objects_cache; |
2687 | if (!odb->loose_objects_cache) { | |
2688 | ALLOC_ARRAY(odb->loose_objects_cache, 1); | |
2689 | oidtree_init(odb->loose_objects_cache); | |
2690 | } | |
3a2e0824 JK |
2691 | strbuf_addstr(&buf, odb->path); |
2692 | for_each_file_in_obj_subdir(subdir_nr, &buf, | |
2693 | append_loose_object, | |
2694 | NULL, NULL, | |
92d8ed8a | 2695 | odb->loose_objects_cache); |
33f379ee | 2696 | *bitmap |= mask; |
7317aa71 | 2697 | strbuf_release(&buf); |
92d8ed8a | 2698 | return odb->loose_objects_cache; |
660c889e JK |
2699 | } |
2700 | ||
d4e19e51 RS |
2701 | void odb_clear_loose_cache(struct object_directory *odb) |
2702 | { | |
92d8ed8a EW |
2703 | oidtree_clear(odb->loose_objects_cache); |
2704 | FREE_AND_NULL(odb->loose_objects_cache); | |
d4e19e51 RS |
2705 | memset(&odb->loose_objects_subdir_seen, 0, |
2706 | sizeof(odb->loose_objects_subdir_seen)); | |
2707 | } | |
2708 | ||
00a7760e JK |
2709 | static int check_stream_oid(git_zstream *stream, |
2710 | const char *hdr, | |
2711 | unsigned long size, | |
2712 | const char *path, | |
2713 | const struct object_id *expected_oid) | |
f6371f92 | 2714 | { |
18e2588e | 2715 | git_hash_ctx c; |
00a7760e | 2716 | struct object_id real_oid; |
f6371f92 JK |
2717 | unsigned char buf[4096]; |
2718 | unsigned long total_read; | |
2719 | int status = Z_OK; | |
2720 | ||
18e2588e | 2721 | the_hash_algo->init_fn(&c); |
2722 | the_hash_algo->update_fn(&c, hdr, stream->total_out); | |
f6371f92 JK |
2723 | |
2724 | /* | |
2725 | * We already read some bytes into hdr, but the ones up to the NUL | |
2726 | * do not count against the object's content size. | |
2727 | */ | |
2728 | total_read = stream->total_out - strlen(hdr) - 1; | |
2729 | ||
2730 | /* | |
2731 | * This size comparison must be "<=" to read the final zlib packets; | |
00a7760e | 2732 | * see the comment in unpack_loose_rest for details. |
f6371f92 JK |
2733 | */ |
2734 | while (total_read <= size && | |
18ad13e5 JH |
2735 | (status == Z_OK || |
2736 | (status == Z_BUF_ERROR && !stream->avail_out))) { | |
f6371f92 JK |
2737 | stream->next_out = buf; |
2738 | stream->avail_out = sizeof(buf); | |
2739 | if (size - total_read < stream->avail_out) | |
2740 | stream->avail_out = size - total_read; | |
2741 | status = git_inflate(stream, Z_FINISH); | |
18e2588e | 2742 | the_hash_algo->update_fn(&c, buf, stream->next_out - buf); |
f6371f92 JK |
2743 | total_read += stream->next_out - buf; |
2744 | } | |
2745 | git_inflate_end(stream); | |
2746 | ||
2747 | if (status != Z_STREAM_END) { | |
00a7760e | 2748 | error(_("corrupt loose object '%s'"), oid_to_hex(expected_oid)); |
f6371f92 JK |
2749 | return -1; |
2750 | } | |
cce044df | 2751 | if (stream->avail_in) { |
259328b7 | 2752 | error(_("garbage at end of loose object '%s'"), |
00a7760e | 2753 | oid_to_hex(expected_oid)); |
cce044df JK |
2754 | return -1; |
2755 | } | |
f6371f92 | 2756 | |
5951bf46 | 2757 | the_hash_algo->final_oid_fn(&real_oid, &c); |
00a7760e | 2758 | if (!oideq(expected_oid, &real_oid)) { |
01f8d594 | 2759 | error(_("hash mismatch for %s (expected %s)"), path, |
00a7760e | 2760 | oid_to_hex(expected_oid)); |
f6371f92 JK |
2761 | return -1; |
2762 | } | |
2763 | ||
2764 | return 0; | |
2765 | } | |
2766 | ||
2767 | int read_loose_object(const char *path, | |
d61d87bd | 2768 | const struct object_id *expected_oid, |
96e41f58 | 2769 | struct object_id *real_oid, |
31deb28f ÆAB |
2770 | void **contents, |
2771 | struct object_info *oi) | |
f6371f92 JK |
2772 | { |
2773 | int ret = -1; | |
ae285ac4 | 2774 | int fd; |
f6371f92 JK |
2775 | void *map = NULL; |
2776 | unsigned long mapsize; | |
2777 | git_zstream stream; | |
1af64f73 | 2778 | char hdr[MAX_HEADER_LEN]; |
31deb28f | 2779 | unsigned long *size = oi->sizep; |
f6371f92 | 2780 | |
ae285ac4 JT |
2781 | fd = git_open(path); |
2782 | if (fd >= 0) | |
2783 | map = map_fd(fd, path, &mapsize); | |
f6371f92 | 2784 | if (!map) { |
259328b7 | 2785 | error_errno(_("unable to mmap %s"), path); |
f6371f92 JK |
2786 | goto out; |
2787 | } | |
2788 | ||
8a50571a DS |
2789 | if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr), |
2790 | NULL) != ULHR_OK) { | |
259328b7 | 2791 | error(_("unable to unpack header of %s"), path); |
f6371f92 JK |
2792 | goto out; |
2793 | } | |
2794 | ||
31deb28f | 2795 | if (parse_loose_header(hdr, oi) < 0) { |
259328b7 | 2796 | error(_("unable to parse header of %s"), path); |
f6371f92 JK |
2797 | git_inflate_end(&stream); |
2798 | goto out; | |
2799 | } | |
2800 | ||
31deb28f | 2801 | if (*oi->typep == OBJ_BLOB && *size > big_file_threshold) { |
00a7760e | 2802 | if (check_stream_oid(&stream, hdr, *size, path, expected_oid) < 0) |
f6371f92 JK |
2803 | goto out; |
2804 | } else { | |
00a7760e | 2805 | *contents = unpack_loose_rest(&stream, hdr, *size, expected_oid); |
f6371f92 | 2806 | if (!*contents) { |
259328b7 | 2807 | error(_("unable to unpack contents of %s"), path); |
f6371f92 JK |
2808 | git_inflate_end(&stream); |
2809 | goto out; | |
2810 | } | |
44439c1c | 2811 | hash_object_file_literally(the_repository->hash_algo, |
16235e3b | 2812 | *contents, *size, |
44439c1c | 2813 | oi->type_name->buf, real_oid); |
0f156dbb | 2814 | if (!oideq(expected_oid, real_oid)) |
f6371f92 | 2815 | goto out; |
f6371f92 JK |
2816 | } |
2817 | ||
2818 | ret = 0; /* everything checks out */ | |
2819 | ||
2820 | out: | |
2821 | if (map) | |
2822 | munmap(map, mapsize); | |
f6371f92 JK |
2823 | return ret; |
2824 | } |