]> git.ipfire.org Git - thirdparty/git.git/blame - sha1-file.c
Merge branch 'en/merge-ort-api-null-impl'
[thirdparty/git.git] / sha1-file.c
CommitLineData
0fcfd160
LT
1/*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 *
6 * This handles basic git sha1 object files - packing, unpacking,
7 * creation etc.
8 */
0fcfd160 9#include "cache.h"
b2141fc1 10#include "config.h"
6eac50d8 11#include "string-list.h"
697cc8ef 12#include "lockfile.h"
1f688557 13#include "delta.h"
a733cb60 14#include "pack.h"
8e440259
PE
15#include "blob.h"
16#include "commit.h"
4dd1fbc7 17#include "run-command.h"
8e440259
PE
18#include "tag.h"
19#include "tree.h"
c879daa2 20#include "tree-walk.h"
f35a6d3b 21#include "refs.h"
70f5d5d3 22#include "pack-revindex.h"
628522ec 23#include "sha1-lookup.h"
568508e7 24#include "bulk-checkin.h"
031dc927 25#include "repository.h"
47f351e9 26#include "replace-object.h"
090ea126 27#include "streaming.h"
543c5caa 28#include "dir.h"
12d95ef6 29#include "list.h"
c4c6effa 30#include "mergesort.h"
cf3c6352 31#include "quote.h"
4f39cd82 32#include "packfile.h"
90c62155 33#include "object-store.h"
b14ed5ad 34#include "promisor-remote.h"
e05db0fd 35
1af64f73 36/* The maximum size for an object header. */
37#define MAX_HEADER_LEN 32
38
e1ccd7e2 39
40#define EMPTY_TREE_SHA1_BIN_LITERAL \
41 "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60" \
42 "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04"
13eeedb5 43#define EMPTY_TREE_SHA256_BIN_LITERAL \
44 "\x6e\xf1\x9b\x41\x22\x5c\x53\x69\xf1\xc1" \
45 "\x04\xd4\x5d\x8d\x85\xef\xa9\xb0\x57\xb5" \
46 "\x3b\x14\xb4\xb9\xb9\x39\xdd\x74\xde\xcc" \
47 "\x53\x21"
e1ccd7e2 48
49#define EMPTY_BLOB_SHA1_BIN_LITERAL \
50 "\xe6\x9d\xe2\x9b\xb2\xd1\xd6\x43\x4b\x8b" \
51 "\x29\xae\x77\x5a\xd8\xc2\xe4\x8c\x53\x91"
13eeedb5 52#define EMPTY_BLOB_SHA256_BIN_LITERAL \
53 "\x47\x3a\x0f\x4c\x3b\xe8\xa9\x36\x81\xa2" \
54 "\x67\xe3\xb1\xe9\xa7\xdc\xda\x11\x85\x43" \
55 "\x6f\xe1\x41\xf7\x74\x91\x20\xa3\x03\x72" \
56 "\x18\x13"
e1ccd7e2 57
3e56e724 58const struct object_id null_oid;
e1ccd7e2 59static const struct object_id empty_tree_oid = {
8576fde6
JK
60 EMPTY_TREE_SHA1_BIN_LITERAL
61};
e1ccd7e2 62static const struct object_id empty_blob_oid = {
8576fde6
JK
63 EMPTY_BLOB_SHA1_BIN_LITERAL
64};
13eeedb5 65static const struct object_id empty_tree_oid_sha256 = {
66 EMPTY_TREE_SHA256_BIN_LITERAL
67};
68static const struct object_id empty_blob_oid_sha256 = {
69 EMPTY_BLOB_SHA256_BIN_LITERAL
70};
88cd621d 71
ac73cedf 72static void git_hash_sha1_init(git_hash_ctx *ctx)
f50e766b 73{
ac73cedf 74 git_SHA1_Init(&ctx->sha1);
f50e766b 75}
76
768e30ea 77static void git_hash_sha1_clone(git_hash_ctx *dst, const git_hash_ctx *src)
78{
79 git_SHA1_Clone(&dst->sha1, &src->sha1);
80}
81
ac73cedf 82static void git_hash_sha1_update(git_hash_ctx *ctx, const void *data, size_t len)
f50e766b 83{
ac73cedf 84 git_SHA1_Update(&ctx->sha1, data, len);
f50e766b 85}
86
ac73cedf 87static void git_hash_sha1_final(unsigned char *hash, git_hash_ctx *ctx)
f50e766b 88{
ac73cedf 89 git_SHA1_Final(hash, &ctx->sha1);
f50e766b 90}
91
13eeedb5 92
93static void git_hash_sha256_init(git_hash_ctx *ctx)
94{
95 git_SHA256_Init(&ctx->sha256);
96}
97
768e30ea 98static void git_hash_sha256_clone(git_hash_ctx *dst, const git_hash_ctx *src)
99{
100 git_SHA256_Clone(&dst->sha256, &src->sha256);
101}
102
13eeedb5 103static void git_hash_sha256_update(git_hash_ctx *ctx, const void *data, size_t len)
104{
105 git_SHA256_Update(&ctx->sha256, data, len);
106}
107
108static void git_hash_sha256_final(unsigned char *hash, git_hash_ctx *ctx)
109{
110 git_SHA256_Final(hash, &ctx->sha256);
111}
112
ac73cedf 113static void git_hash_unknown_init(git_hash_ctx *ctx)
f50e766b 114{
1a07e59c 115 BUG("trying to init unknown hash");
f50e766b 116}
117
768e30ea 118static void git_hash_unknown_clone(git_hash_ctx *dst, const git_hash_ctx *src)
119{
120 BUG("trying to clone unknown hash");
121}
122
ac73cedf 123static void git_hash_unknown_update(git_hash_ctx *ctx, const void *data, size_t len)
f50e766b 124{
1a07e59c 125 BUG("trying to update unknown hash");
f50e766b 126}
127
ac73cedf 128static void git_hash_unknown_final(unsigned char *hash, git_hash_ctx *ctx)
f50e766b 129{
1a07e59c 130 BUG("trying to finalize unknown hash");
f50e766b 131}
132
133const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = {
134 {
135 NULL,
136 0x00000000,
137 0,
138 0,
a2ce0a75 139 0,
f50e766b 140 git_hash_unknown_init,
768e30ea 141 git_hash_unknown_clone,
f50e766b 142 git_hash_unknown_update,
143 git_hash_unknown_final,
144 NULL,
145 NULL,
146 },
147 {
1ccf07cb 148 "sha1",
f50e766b 149 /* "sha1", big-endian */
150 0x73686131,
f50e766b 151 GIT_SHA1_RAWSZ,
152 GIT_SHA1_HEXSZ,
a2ce0a75 153 GIT_SHA1_BLKSZ,
f50e766b 154 git_hash_sha1_init,
768e30ea 155 git_hash_sha1_clone,
f50e766b 156 git_hash_sha1_update,
157 git_hash_sha1_final,
158 &empty_tree_oid,
159 &empty_blob_oid,
160 },
13eeedb5 161 {
162 "sha256",
163 /* "s256", big-endian */
164 0x73323536,
165 GIT_SHA256_RAWSZ,
166 GIT_SHA256_HEXSZ,
167 GIT_SHA256_BLKSZ,
168 git_hash_sha256_init,
768e30ea 169 git_hash_sha256_clone,
13eeedb5 170 git_hash_sha256_update,
171 git_hash_sha256_final,
172 &empty_tree_oid_sha256,
173 &empty_blob_oid_sha256,
174 }
f50e766b 175};
176
d8a92ced 177const char *empty_tree_oid_hex(void)
178{
179 static char buf[GIT_MAX_HEXSZ + 1];
180 return oid_to_hex_r(buf, the_hash_algo->empty_tree);
181}
182
183const char *empty_blob_oid_hex(void)
184{
185 static char buf[GIT_MAX_HEXSZ + 1];
186 return oid_to_hex_r(buf, the_hash_algo->empty_blob);
187}
188
2f90b9d9 189int hash_algo_by_name(const char *name)
190{
191 int i;
192 if (!name)
193 return GIT_HASH_UNKNOWN;
194 for (i = 1; i < GIT_HASH_NALGOS; i++)
195 if (!strcmp(name, hash_algos[i].name))
196 return i;
197 return GIT_HASH_UNKNOWN;
198}
199
200int hash_algo_by_id(uint32_t format_id)
201{
202 int i;
203 for (i = 1; i < GIT_HASH_NALGOS; i++)
204 if (format_id == hash_algos[i].format_id)
205 return i;
206 return GIT_HASH_UNKNOWN;
207}
208
95399788 209int hash_algo_by_length(int len)
210{
211 int i;
212 for (i = 1; i < GIT_HASH_NALGOS; i++)
213 if (len == hash_algos[i].rawsz)
214 return i;
215 return GIT_HASH_UNKNOWN;
216}
2f90b9d9 217
c597ba80
NTND
218/*
219 * This is meant to hold a *small* number of objects that you would
cb1c8d1d 220 * want read_object_file() to be able to return, but yet you do not want
c597ba80
NTND
221 * to write them into the object store (e.g. a browse-only
222 * application).
223 */
224static struct cached_object {
62ba93ea 225 struct object_id oid;
c597ba80
NTND
226 enum object_type type;
227 void *buf;
228 unsigned long size;
229} *cached_objects;
230static int cached_object_nr, cached_object_alloc;
231
232static struct cached_object empty_tree = {
62ba93ea 233 { EMPTY_TREE_SHA1_BIN_LITERAL },
c597ba80
NTND
234 OBJ_TREE,
235 "",
236 0
237};
238
62ba93ea 239static struct cached_object *find_cached_object(const struct object_id *oid)
c597ba80
NTND
240{
241 int i;
242 struct cached_object *co = cached_objects;
243
244 for (i = 0; i < cached_object_nr; i++, co++) {
4a7e27e9 245 if (oideq(&co->oid, oid))
c597ba80
NTND
246 return co;
247 }
4a7e27e9 248 if (oideq(oid, the_hash_algo->empty_tree))
c597ba80
NTND
249 return &empty_tree;
250 return NULL;
251}
252
9472935d 253
8462ff43 254static int get_conv_flags(unsigned flags)
9472935d
TB
255{
256 if (flags & HASH_RENORMALIZE)
8462ff43 257 return CONV_EOL_RENORMALIZE;
9472935d 258 else if (flags & HASH_WRITE_OBJECT)
107642fe 259 return global_conv_flags_eol | CONV_WRITE_OBJECT;
9472935d 260 else
8462ff43 261 return 0;
9472935d
TB
262}
263
264
90a6464b
JH
265int mkdir_in_gitdir(const char *path)
266{
267 if (mkdir(path, 0777)) {
268 int saved_errno = errno;
269 struct stat st;
270 struct strbuf sb = STRBUF_INIT;
271
272 if (errno != EEXIST)
273 return -1;
274 /*
275 * Are we looking at a path in a symlinked worktree
276 * whose original repository does not yet have it?
277 * e.g. .git/rr-cache pointing at its original
278 * repository in which the user hasn't performed any
279 * conflict resolution yet?
280 */
281 if (lstat(path, &st) || !S_ISLNK(st.st_mode) ||
282 strbuf_readlink(&sb, path, st.st_size) ||
283 !is_absolute_path(sb.buf) ||
284 mkdir(sb.buf, 0777)) {
285 strbuf_release(&sb);
286 errno = saved_errno;
287 return -1;
288 }
289 strbuf_release(&sb);
290 }
291 return adjust_shared_perm(path);
292}
293
0be0521b 294enum scld_error safe_create_leading_directories(char *path)
b2cb9425 295{
26c8ae2a 296 char *next_component = path + offset_1st_component(path);
0be0521b 297 enum scld_error ret = SCLD_OK;
67d42212 298
0be0521b 299 while (ret == SCLD_OK && next_component) {
f0502332 300 struct stat st;
0f527403 301 char *slash = next_component, slash_character;
f0502332 302
0f527403
MH
303 while (*slash && !is_dir_sep(*slash))
304 slash++;
305
306 if (!*slash)
b2cb9425 307 break;
bf10cf70 308
26c8ae2a 309 next_component = slash + 1;
0f527403 310 while (is_dir_sep(*next_component))
bf10cf70 311 next_component++;
26c8ae2a 312 if (!*next_component)
5f0bdf50 313 break;
831651fd 314
0f527403 315 slash_character = *slash;
831651fd 316 *slash = '\0';
67d42212
JR
317 if (!stat(path, &st)) {
318 /* path exists */
204a047f
MH
319 if (!S_ISDIR(st.st_mode)) {
320 errno = ENOTDIR;
0be0521b 321 ret = SCLD_EXISTS;
204a047f 322 }
53a39721 323 } else if (mkdir(path, 0777)) {
928734d9 324 if (errno == EEXIST &&
9e6f885d 325 !stat(path, &st) && S_ISDIR(st.st_mode))
928734d9 326 ; /* somebody created it since we checked */
18d37e86
MH
327 else if (errno == ENOENT)
328 /*
329 * Either mkdir() failed because
330 * somebody just pruned the containing
331 * directory, or stat() failed because
332 * the file that was in our way was
333 * just removed. Either way, inform
334 * the caller that it might be worth
335 * trying again:
336 */
337 ret = SCLD_VANISHED;
9e6f885d 338 else
0be0521b 339 ret = SCLD_FAILED;
53a39721 340 } else if (adjust_shared_perm(path)) {
0be0521b 341 ret = SCLD_PERMS;
457f06d6 342 }
0f527403 343 *slash = slash_character;
b2cb9425 344 }
9e6f885d 345 return ret;
b2cb9425 346}
723c31fe 347
0be0521b 348enum scld_error safe_create_leading_directories_const(const char *path)
8e21d63b 349{
02944307 350 int save_errno;
8e21d63b
JK
351 /* path points to cache entries, so xstrdup before messing with it */
352 char *buf = xstrdup(path);
0be0521b 353 enum scld_error result = safe_create_leading_directories(buf);
02944307
MH
354
355 save_errno = errno;
8e21d63b 356 free(buf);
02944307 357 errno = save_errno;
8e21d63b
JK
358 return result;
359}
360
177978f5
MH
361int raceproof_create_file(const char *path, create_file_fn fn, void *cb)
362{
363 /*
364 * The number of times we will try to remove empty directories
365 * in the way of path. This is only 1 because if another
366 * process is racily creating directories that conflict with
367 * us, we don't want to fight against them.
368 */
369 int remove_directories_remaining = 1;
370
371 /*
372 * The number of times that we will try to create the
373 * directories containing path. We are willing to attempt this
374 * more than once, because another process could be trying to
375 * clean up empty directories at the same time as we are
376 * trying to create them.
377 */
378 int create_directories_remaining = 3;
379
380 /* A scratch copy of path, filled lazily if we need it: */
381 struct strbuf path_copy = STRBUF_INIT;
382
383 int ret, save_errno;
384
385 /* Sanity check: */
386 assert(*path);
387
388retry_fn:
389 ret = fn(path, cb);
390 save_errno = errno;
391 if (!ret)
392 goto out;
393
394 if (errno == EISDIR && remove_directories_remaining-- > 0) {
395 /*
396 * A directory is in the way. Maybe it is empty; try
397 * to remove it:
398 */
399 if (!path_copy.len)
400 strbuf_addstr(&path_copy, path);
401
402 if (!remove_dir_recursively(&path_copy, REMOVE_DIR_EMPTY_ONLY))
403 goto retry_fn;
404 } else if (errno == ENOENT && create_directories_remaining-- > 0) {
405 /*
406 * Maybe the containing directory didn't exist, or
407 * maybe it was just deleted by a process that is
408 * racing with us to clean up empty directories. Try
409 * to create it:
410 */
411 enum scld_error scld_result;
412
413 if (!path_copy.len)
414 strbuf_addstr(&path_copy, path);
415
416 do {
417 scld_result = safe_create_leading_directories(path_copy.buf);
418 if (scld_result == SCLD_OK)
419 goto retry_fn;
420 } while (scld_result == SCLD_VANISHED && create_directories_remaining-- > 0);
421 }
422
423out:
424 strbuf_release(&path_copy);
425 errno = save_errno;
426 return ret;
427}
428
514c5fdd 429static void fill_loose_path(struct strbuf *buf, const struct object_id *oid)
ace1534d
JH
430{
431 int i;
94b5e093 432 for (i = 0; i < the_hash_algo->rawsz; i++) {
ace1534d 433 static char hex[] = "0123456789abcdef";
514c5fdd 434 unsigned int val = oid->hash[i];
f7b7774f
JK
435 strbuf_addch(buf, hex[val >> 4]);
436 strbuf_addch(buf, hex[val & 0xf]);
afbba2f0 437 if (!i)
f7b7774f 438 strbuf_addch(buf, '/');
ace1534d
JH
439 }
440}
441
f0eaf638
JK
442static const char *odb_loose_path(struct object_directory *odb,
443 struct strbuf *buf,
514c5fdd 444 const struct object_id *oid)
0fcfd160 445{
b69fb867 446 strbuf_reset(buf);
f0eaf638 447 strbuf_addstr(buf, odb->path);
34498471 448 strbuf_addch(buf, '/');
514c5fdd 449 fill_loose_path(buf, oid);
f3f043a1 450 return buf->buf;
0fcfd160
LT
451}
452
f3f043a1 453const char *loose_object_path(struct repository *r, struct strbuf *buf,
514c5fdd 454 const struct object_id *oid)
29ec6af2 455{
514c5fdd 456 return odb_loose_path(r->objects->odb, buf, oid);
0fcfd160
LT
457}
458
4ea82473
JK
459/*
460 * Return non-zero iff the path is usable as an alternate object database.
461 */
13313fc3
SB
462static int alt_odb_usable(struct raw_object_store *o,
463 struct strbuf *path,
464 const char *normalized_objdir)
4ea82473 465{
263db403 466 struct object_directory *odb;
4ea82473
JK
467
468 /* Detect cases where alternate disappeared */
469 if (!is_directory(path->buf)) {
259328b7
NTND
470 error(_("object directory %s does not exist; "
471 "check .git/objects/info/alternates"),
4ea82473
JK
472 path->buf);
473 return 0;
474 }
475
476 /*
477 * Prevent the common mistake of listing the same
478 * thing twice, or object directory itself.
479 */
f0eaf638 480 for (odb = o->odb; odb; odb = odb->next) {
263db403 481 if (!fspathcmp(path->buf, odb->path))
4ea82473
JK
482 return 0;
483 }
484 if (!fspathcmp(path->buf, normalized_objdir))
485 return 0;
486
487 return 1;
488}
489
ddd5d056
JH
490/*
491 * Prepare alternate object database registry.
d5a63b99
JH
492 *
493 * The variable alt_odb_list points at the list of struct
263db403 494 * object_directory. The elements on this list come from
d5a63b99
JH
495 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
496 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
497 * whose contents is similar to that environment variable but can be
498 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
499 * contains "/the/directory/corresponding/to/.git/objects/...", while
500 * its name points just after the slash at the end of ".git/objects/"
501 * in the example above, and has enough space to hold 40-byte hex
502 * SHA1, an extra slash for the first level indirection, and the
503 * terminating NUL.
ddd5d056 504 */
77f012e8
SB
505static void read_info_alternates(struct repository *r,
506 const char *relative_base,
507 int depth);
508static int link_alt_odb_entry(struct repository *r, const char *entry,
cfc62fc9 509 const char *relative_base, int depth, const char *normalized_objdir)
ace1534d 510{
263db403 511 struct object_directory *ent;
5bdf0a84 512 struct strbuf pathbuf = STRBUF_INIT;
d5a63b99 513
85dadc38 514 if (!is_absolute_path(entry) && relative_base) {
4ac9006f 515 strbuf_realpath(&pathbuf, relative_base, 1);
5bdf0a84 516 strbuf_addch(&pathbuf, '/');
c2f493a4 517 }
6eac50d8 518 strbuf_addstr(&pathbuf, entry);
c2f493a4 519
37a95862 520 if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) {
259328b7 521 error(_("unable to normalize alternate object path: %s"),
670c359d
JK
522 pathbuf.buf);
523 strbuf_release(&pathbuf);
524 return -1;
525 }
5bdf0a84
HW
526
527 /*
528 * The trailing slash after the directory name is given by
529 * this function at the end. Remove duplicates.
530 */
4ea82473
JK
531 while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/')
532 strbuf_setlen(&pathbuf, pathbuf.len - 1);
5bdf0a84 533
77f012e8 534 if (!alt_odb_usable(r->objects, &pathbuf, normalized_objdir)) {
4ea82473 535 strbuf_release(&pathbuf);
c2f493a4
MW
536 return -1;
537 }
538
f0eaf638
JK
539 ent = xcalloc(1, sizeof(*ent));
540 ent->path = xstrdup(pathbuf.buf);
c2f493a4
MW
541
542 /* add the alternate entry */
f0eaf638
JK
543 *r->objects->odb_tail = ent;
544 r->objects->odb_tail = &(ent->next);
c2f493a4
MW
545 ent->next = NULL;
546
547 /* recursively add alternates */
77f012e8 548 read_info_alternates(r, pathbuf.buf, depth + 1);
c2f493a4 549
4ea82473 550 strbuf_release(&pathbuf);
c2f493a4
MW
551 return 0;
552}
553
cf3c6352
JK
554static const char *parse_alt_odb_entry(const char *string,
555 int sep,
556 struct strbuf *out)
557{
558 const char *end;
559
560 strbuf_reset(out);
561
562 if (*string == '#') {
563 /* comment; consume up to next separator */
564 end = strchrnul(string, sep);
565 } else if (*string == '"' && !unquote_c_style(out, string, &end)) {
566 /*
567 * quoted path; unquote_c_style has copied the
568 * data for us and set "end". Broken quoting (e.g.,
569 * an entry that doesn't end with a quote) falls
570 * back to the unquoted case below.
571 */
572 } else {
573 /* normal, unquoted path */
574 end = strchrnul(string, sep);
575 strbuf_add(out, string, end - string);
576 }
577
578 if (*end)
579 end++;
580 return end;
581}
582
77f012e8
SB
583static void link_alt_odb_entries(struct repository *r, const char *alt,
584 int sep, const char *relative_base, int depth)
c2f493a4 585{
539e7506 586 struct strbuf objdirbuf = STRBUF_INIT;
cf3c6352 587 struct strbuf entry = STRBUF_INIT;
c2f493a4 588
f28e3668
JK
589 if (!alt || !*alt)
590 return;
591
c2f493a4 592 if (depth > 5) {
259328b7 593 error(_("%s: ignoring alternate object stores, nesting too deep"),
c2f493a4
MW
594 relative_base);
595 return;
596 }
597
f0eaf638 598 strbuf_add_absolute_path(&objdirbuf, r->objects->odb->path);
670c359d 599 if (strbuf_normalize_path(&objdirbuf) < 0)
259328b7 600 die(_("unable to normalize object directory: %s"),
670c359d 601 objdirbuf.buf);
539e7506 602
cf3c6352
JK
603 while (*alt) {
604 alt = parse_alt_odb_entry(alt, sep, &entry);
605 if (!entry.len)
9577e7e3 606 continue;
77f012e8 607 link_alt_odb_entry(r, entry.buf,
cfc62fc9 608 relative_base, depth, objdirbuf.buf);
9577e7e3 609 }
cf3c6352 610 strbuf_release(&entry);
539e7506 611 strbuf_release(&objdirbuf);
d5a63b99
JH
612}
613
77f012e8
SB
614static void read_info_alternates(struct repository *r,
615 const char *relative_base,
616 int depth)
d5a63b99 617{
5015f01c 618 char *path;
dc732bd5 619 struct strbuf buf = STRBUF_INIT;
d5a63b99 620
5015f01c 621 path = xstrfmt("%s/info/alternates", relative_base);
dc732bd5 622 if (strbuf_read_file(&buf, path, 1024) < 0) {
f0f7bebe 623 warn_on_fopen_errors(path);
dc732bd5 624 free(path);
9a217f2a 625 return;
ace1534d 626 }
d5a63b99 627
77f012e8 628 link_alt_odb_entries(r, buf.buf, '\n', relative_base, depth);
dc732bd5
JK
629 strbuf_release(&buf);
630 free(path);
ace1534d
JH
631}
632
bef70b22
DB
633void add_to_alternates_file(const char *reference)
634{
f132a127 635 struct lock_file lock = LOCK_INIT;
77b9b1d1
JK
636 char *alts = git_pathdup("objects/info/alternates");
637 FILE *in, *out;
f132a127 638 int found = 0;
77b9b1d1 639
f132a127
640 hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR);
641 out = fdopen_lock_file(&lock, "w");
77b9b1d1 642 if (!out)
259328b7 643 die_errno(_("unable to fdopen alternates lockfile"));
77b9b1d1
JK
644
645 in = fopen(alts, "r");
646 if (in) {
647 struct strbuf line = STRBUF_INIT;
77b9b1d1 648
3f163962 649 while (strbuf_getline(&line, in) != EOF) {
77b9b1d1
JK
650 if (!strcmp(reference, line.buf)) {
651 found = 1;
652 break;
653 }
654 fprintf_or_die(out, "%s\n", line.buf);
655 }
656
657 strbuf_release(&line);
658 fclose(in);
77b9b1d1
JK
659 }
660 else if (errno != ENOENT)
259328b7 661 die_errno(_("unable to read alternates file"));
77b9b1d1 662
f132a127
663 if (found) {
664 rollback_lock_file(&lock);
665 } else {
77b9b1d1 666 fprintf_or_die(out, "%s\n", reference);
f132a127 667 if (commit_lock_file(&lock))
259328b7 668 die_errno(_("unable to move new alternates file into place"));
f0eaf638 669 if (the_repository->objects->loaded_alternates)
93d8d1e2
SB
670 link_alt_odb_entries(the_repository, reference,
671 '\n', NULL, 0);
77b9b1d1
JK
672 }
673 free(alts);
bef70b22
DB
674}
675
a5b34d21
JK
676void add_to_alternates_memory(const char *reference)
677{
678 /*
679 * Make sure alternates are initialized, or else our entry may be
680 * overwritten when they are.
681 */
0b209034 682 prepare_alt_odb(the_repository);
a5b34d21 683
93d8d1e2
SB
684 link_alt_odb_entries(the_repository, reference,
685 '\n', NULL, 0);
a5b34d21
JK
686}
687
9eeea7d2
SB
688/*
689 * Compute the exact path an alternate is at and returns it. In case of
690 * error NULL is returned and the human readable error is added to `err`
efde7b72 691 * `path` may be relative and should point to $GIT_DIR.
9eeea7d2
SB
692 * `err` must not be null.
693 */
694char *compute_alternate_path(const char *path, struct strbuf *err)
695{
696 char *ref_git = NULL;
4530a85b 697 const char *repo;
9eeea7d2
SB
698 int seen_error = 0;
699
4530a85b
AM
700 ref_git = real_pathdup(path, 0);
701 if (!ref_git) {
9eeea7d2
SB
702 seen_error = 1;
703 strbuf_addf(err, _("path '%s' does not exist"), path);
704 goto out;
4530a85b 705 }
9eeea7d2
SB
706
707 repo = read_gitfile(ref_git);
708 if (!repo)
709 repo = read_gitfile(mkpath("%s/.git", ref_git));
710 if (repo) {
711 free(ref_git);
712 ref_git = xstrdup(repo);
713 }
714
715 if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) {
716 char *ref_git_git = mkpathdup("%s/.git", ref_git);
717 free(ref_git);
718 ref_git = ref_git_git;
719 } else if (!is_directory(mkpath("%s/objects", ref_git))) {
720 struct strbuf sb = STRBUF_INIT;
721 seen_error = 1;
722 if (get_common_dir(&sb, ref_git)) {
723 strbuf_addf(err,
724 _("reference repository '%s' as a linked "
725 "checkout is not supported yet."),
726 path);
727 goto out;
728 }
729
730 strbuf_addf(err, _("reference repository '%s' is not a "
731 "local repository."), path);
732 goto out;
733 }
734
735 if (!access(mkpath("%s/shallow", ref_git), F_OK)) {
736 strbuf_addf(err, _("reference repository '%s' is shallow"),
737 path);
738 seen_error = 1;
739 goto out;
740 }
741
742 if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) {
743 strbuf_addf(err,
744 _("reference repository '%s' is grafted"),
745 path);
746 seen_error = 1;
747 goto out;
748 }
749
750out:
751 if (seen_error) {
6a83d902 752 FREE_AND_NULL(ref_git);
9eeea7d2
SB
753 }
754
755 return ref_git;
756}
757
709dfa69
JK
758static void fill_alternate_refs_command(struct child_process *cmd,
759 const char *repo_path)
760{
761 const char *value;
762
763 if (!git_config_get_value("core.alternateRefsCommand", &value)) {
764 cmd->use_shell = 1;
765
c972bf4c
JK
766 strvec_push(&cmd->args, value);
767 strvec_push(&cmd->args, repo_path);
709dfa69
JK
768 } else {
769 cmd->git_cmd = 1;
770
c972bf4c
JK
771 strvec_pushf(&cmd->args, "--git-dir=%s", repo_path);
772 strvec_push(&cmd->args, "for-each-ref");
773 strvec_push(&cmd->args, "--format=%(objectname)");
709dfa69
JK
774
775 if (!git_config_get_value("core.alternateRefsPrefixes", &value)) {
c972bf4c
JK
776 strvec_push(&cmd->args, "--");
777 strvec_split(&cmd->args, value);
709dfa69
JK
778 }
779 }
780
781 cmd->env = local_repo_env;
782 cmd->out = -1;
783}
784
785static void read_alternate_refs(const char *path,
786 alternate_ref_fn *cb,
787 void *data)
788{
789 struct child_process cmd = CHILD_PROCESS_INIT;
790 struct strbuf line = STRBUF_INIT;
791 FILE *fh;
792
793 fill_alternate_refs_command(&cmd, path);
794
795 if (start_command(&cmd))
796 return;
797
798 fh = xfdopen(cmd.out, "r");
799 while (strbuf_getline_lf(&line, fh) != EOF) {
800 struct object_id oid;
801 const char *p;
802
803 if (parse_oid_hex(line.buf, &oid, &p) || *p) {
804 warning(_("invalid line while parsing alternate refs: %s"),
805 line.buf);
806 break;
807 }
808
809 cb(&oid, data);
810 }
811
812 fclose(fh);
813 finish_command(&cmd);
86ad3ea5 814 strbuf_release(&line);
709dfa69
JK
815}
816
817struct alternate_refs_data {
818 alternate_ref_fn *fn;
819 void *data;
820};
821
822static int refs_from_alternate_cb(struct object_directory *e,
823 void *data)
824{
825 struct strbuf path = STRBUF_INIT;
826 size_t base_len;
827 struct alternate_refs_data *cb = data;
828
829 if (!strbuf_realpath(&path, e->path, 0))
830 goto out;
831 if (!strbuf_strip_suffix(&path, "/objects"))
832 goto out;
833 base_len = path.len;
834
835 /* Is this a git repository with refs? */
836 strbuf_addstr(&path, "/refs");
837 if (!is_directory(path.buf))
838 goto out;
839 strbuf_setlen(&path, base_len);
840
841 read_alternate_refs(path.buf, cb->fn, cb->data);
842
843out:
844 strbuf_release(&path);
845 return 0;
846}
847
848void for_each_alternate_ref(alternate_ref_fn fn, void *data)
849{
850 struct alternate_refs_data cb;
851 cb.fn = fn;
852 cb.data = data;
853 foreach_alt_odb(refs_from_alternate_cb, &cb);
854}
855
fe1b2268 856int foreach_alt_odb(alt_odb_fn fn, void *cb)
d79796bc 857{
263db403 858 struct object_directory *ent;
fe1b2268 859 int r = 0;
d79796bc 860
0b209034 861 prepare_alt_odb(the_repository);
f0eaf638 862 for (ent = the_repository->objects->odb->next; ent; ent = ent->next) {
fe1b2268
JK
863 r = fn(ent, cb);
864 if (r)
865 break;
866 }
867 return r;
d79796bc
JH
868}
869
13068bf0 870void prepare_alt_odb(struct repository *r)
c2f493a4 871{
f0eaf638 872 if (r->objects->loaded_alternates)
7dc24aa5
SP
873 return;
874
13068bf0 875 link_alt_odb_entries(r, r->objects->alternate_db, PATH_SEP, NULL, 0);
c2f493a4 876
f0eaf638
JK
877 read_info_alternates(r, r->objects->odb->path, 0);
878 r->objects->loaded_alternates = 1;
c2f493a4
MW
879}
880
3096b2ec 881/* Returns 1 if we have successfully freshened the file, 0 otherwise. */
33d4221c 882static int freshen_file(const char *fn)
ace1534d 883{
312cd761 884 return !utime(fn, NULL);
0f4dc14a 885}
ace1534d 886
3096b2ec
JK
887/*
888 * All of the check_and_freshen functions return 1 if the file exists and was
889 * freshened (if freshening was requested), 0 otherwise. If they return
890 * 0, you should not assume that it is safe to skip a write of the object (it
891 * either does not exist on disk, or has a stale mtime and may be subject to
892 * pruning).
893 */
6a5e6f5e 894int check_and_freshen_file(const char *fn, int freshen)
33d4221c
JK
895{
896 if (access(fn, F_OK))
897 return 0;
3096b2ec 898 if (freshen && !freshen_file(fn))
33d4221c
JK
899 return 0;
900 return 1;
901}
902
f0eaf638
JK
903static int check_and_freshen_odb(struct object_directory *odb,
904 const struct object_id *oid,
905 int freshen)
33d4221c 906{
f0eaf638 907 static struct strbuf path = STRBUF_INIT;
514c5fdd 908 odb_loose_path(odb, &path, oid);
f0eaf638
JK
909 return check_and_freshen_file(path.buf, freshen);
910}
ea657730 911
f0eaf638
JK
912static int check_and_freshen_local(const struct object_id *oid, int freshen)
913{
914 return check_and_freshen_odb(the_repository->objects->odb, oid, freshen);
33d4221c
JK
915}
916
6862ebbf 917static int check_and_freshen_nonlocal(const struct object_id *oid, int freshen)
0f4dc14a 918{
263db403 919 struct object_directory *odb;
f3f043a1 920
0b209034 921 prepare_alt_odb(the_repository);
f0eaf638
JK
922 for (odb = the_repository->objects->odb->next; odb; odb = odb->next) {
923 if (check_and_freshen_odb(odb, oid, freshen))
c529d75a 924 return 1;
ace1534d 925 }
c529d75a 926 return 0;
ace1534d
JH
927}
928
6862ebbf 929static int check_and_freshen(const struct object_id *oid, int freshen)
33d4221c 930{
6862ebbf 931 return check_and_freshen_local(oid, freshen) ||
932 check_and_freshen_nonlocal(oid, freshen);
33d4221c
JK
933}
934
6862ebbf 935int has_loose_object_nonlocal(const struct object_id *oid)
33d4221c 936{
6862ebbf 937 return check_and_freshen_nonlocal(oid, 0);
33d4221c
JK
938}
939
6862ebbf 940static int has_loose_object(const struct object_id *oid)
0f4dc14a 941{
6862ebbf 942 return check_and_freshen(oid, 0);
0f4dc14a
BC
943}
944
02710228
SP
945static void mmap_limit_check(size_t length)
946{
947 static size_t limit = 0;
948 if (!limit) {
949 limit = git_env_ulong("GIT_MMAP_LIMIT", 0);
950 if (!limit)
951 limit = SIZE_MAX;
952 }
953 if (length > limit)
259328b7 954 die(_("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX),
02710228
SP
955 (uintmax_t)length, (uintmax_t)limit);
956}
957
1570856b
JK
958void *xmmap_gently(void *start, size_t length,
959 int prot, int flags, int fd, off_t offset)
58ecbd5e 960{
02710228
SP
961 void *ret;
962
963 mmap_limit_check(length);
964 ret = mmap(start, length, prot, flags, fd, offset);
9827d4c1
JK
965 if (ret == MAP_FAILED && !length)
966 ret = NULL;
58ecbd5e
JN
967 return ret;
968}
969
1570856b
JK
970void *xmmap(void *start, size_t length,
971 int prot, int flags, int fd, off_t offset)
972{
973 void *ret = xmmap_gently(start, length, prot, flags, fd, offset);
974 if (ret == MAP_FAILED)
259328b7 975 die_errno(_("mmap failed"));
1570856b
JK
976 return ret;
977}
978
88d0db55 979/*
090ea126 980 * With an in-core object data in "map", rehash it to make sure the
cb1c8d1d
JK
981 * object name actually matches "oid" to detect object corruption.
982 * With "map" == NULL, try reading the object named with "oid" using
090ea126 983 * the streaming interface and rehash it to do the same.
88d0db55 984 */
b98d1885
MT
985int check_object_signature(struct repository *r, const struct object_id *oid,
986 void *map, unsigned long size, const char *type)
88d0db55 987{
f070facc 988 struct object_id real_oid;
090ea126
NTND
989 enum object_type obj_type;
990 struct git_istream *st;
18e2588e 991 git_hash_ctx c;
1af64f73 992 char hdr[MAX_HEADER_LEN];
090ea126 993 int hdrlen;
88d0db55 994
090ea126 995 if (map) {
b98d1885 996 hash_object_file(r->hash_algo, map, size, type, &real_oid);
9001dc2a 997 return !oideq(oid, &real_oid) ? -1 : 0;
88d0db55
BC
998 }
999
b98d1885 1000 st = open_istream(r, oid, &obj_type, &size, NULL);
090ea126
NTND
1001 if (!st)
1002 return -1;
88d0db55 1003
090ea126 1004 /* Generate the header */
ca473cef 1005 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(obj_type), (uintmax_t)size) + 1;
88d0db55 1006
090ea126 1007 /* Sha1.. */
b98d1885
MT
1008 r->hash_algo->init_fn(&c);
1009 r->hash_algo->update_fn(&c, hdr, hdrlen);
090ea126
NTND
1010 for (;;) {
1011 char buf[1024 * 16];
1012 ssize_t readlen = read_istream(st, buf, sizeof(buf));
1f688557 1013
f54fac53
JK
1014 if (readlen < 0) {
1015 close_istream(st);
1016 return -1;
1017 }
090ea126
NTND
1018 if (!readlen)
1019 break;
b98d1885 1020 r->hash_algo->update_fn(&c, buf, readlen);
fa5fc15d 1021 }
b98d1885 1022 r->hash_algo->final_fn(real_oid.hash, &c);
090ea126 1023 close_istream(st);
9001dc2a 1024 return !oideq(oid, &real_oid) ? -1 : 0;
fa5fc15d
SP
1025}
1026
1b8ac5ea 1027int git_open_cloexec(const char *name, int flags)
a0788266 1028{
1e3001a8
JH
1029 int fd;
1030 static int o_cloexec = O_CLOEXEC;
a0788266 1031
1e3001a8
JH
1032 fd = open(name, flags | o_cloexec);
1033 if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) {
cd66ada0 1034 /* Try again w/o O_CLOEXEC: the kernel might not support it */
1e3001a8
JH
1035 o_cloexec &= ~O_CLOEXEC;
1036 fd = open(name, flags | o_cloexec);
491a8dec 1037 }
491a8dec 1038
9fb9495d 1039#if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
491a8dec 1040 {
1e3001a8 1041 static int fd_cloexec = FD_CLOEXEC;
a0788266 1042
1e3001a8
JH
1043 if (!o_cloexec && 0 <= fd && fd_cloexec) {
1044 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
9fb9495d
EW
1045 int flags = fcntl(fd, F_GETFD);
1046 if (fcntl(fd, F_SETFD, flags | fd_cloexec))
1e3001a8 1047 fd_cloexec = 0;
cd66ada0 1048 }
44d1c19e 1049 }
a0788266 1050#endif
1b8ac5ea 1051 return fd;
a0788266
JS
1052}
1053
3cf8b462 1054/*
514c5fdd 1055 * Find "oid" as a loose object in the local repository or in an alternate.
771e7d57
JK
1056 * Returns 0 on success, negative on failure.
1057 *
1058 * The "path" out-parameter will give the path of the object we found (if any).
1059 * Note that it may point to static storage and is only valid until another
514c5fdd 1060 * call to stat_loose_object().
3cf8b462 1061 */
514c5fdd
JK
1062static int stat_loose_object(struct repository *r, const struct object_id *oid,
1063 struct stat *st, const char **path)
1f688557 1064{
263db403 1065 struct object_directory *odb;
ea657730
CC
1066 static struct strbuf buf = STRBUF_INIT;
1067
d2607fa0 1068 prepare_alt_odb(r);
f0eaf638 1069 for (odb = r->objects->odb; odb; odb = odb->next) {
514c5fdd 1070 *path = odb_loose_path(odb, &buf, oid);
771e7d57 1071 if (!lstat(*path, st))
052fe5ea 1072 return 0;
c7934306
SP
1073 }
1074
3cf8b462
SP
1075 return -1;
1076}
1077
771e7d57 1078/*
514c5fdd 1079 * Like stat_loose_object(), but actually open the object and return the
771e7d57
JK
1080 * descriptor. See the caveats on the "path" parameter above.
1081 */
514c5fdd
JK
1082static int open_loose_object(struct repository *r,
1083 const struct object_id *oid, const char **path)
60bb8b14 1084{
44d1c19e 1085 int fd;
263db403 1086 struct object_directory *odb;
f0eaf638 1087 int most_interesting_errno = ENOENT;
ea657730
CC
1088 static struct strbuf buf = STRBUF_INIT;
1089
ec7283e5 1090 prepare_alt_odb(r);
f0eaf638 1091 for (odb = r->objects->odb; odb; odb = odb->next) {
514c5fdd 1092 *path = odb_loose_path(odb, &buf, oid);
771e7d57 1093 fd = git_open(*path);
44d1c19e
LT
1094 if (fd >= 0)
1095 return fd;
f0eaf638 1096
d6c8a05b
JK
1097 if (most_interesting_errno == ENOENT)
1098 most_interesting_errno = errno;
03e79c88 1099 }
d6c8a05b 1100 errno = most_interesting_errno;
44d1c19e 1101 return -1;
1f688557
JH
1102}
1103
61c7711c 1104static int quick_has_loose(struct repository *r,
d7a24573 1105 const struct object_id *oid)
61c7711c 1106{
61c7711c
JK
1107 struct object_directory *odb;
1108
61c7711c
JK
1109 prepare_alt_odb(r);
1110 for (odb = r->objects->odb; odb; odb = odb->next) {
d7a24573 1111 if (oid_array_lookup(odb_loose_cache(odb, oid), oid) >= 0)
61c7711c
JK
1112 return 1;
1113 }
1114 return 0;
1115}
1116
f6371f92
JK
1117/*
1118 * Map the loose object at "path" if it is not NULL, or the path found by
514c5fdd 1119 * searching for a loose object named "oid".
f6371f92 1120 */
514c5fdd
JK
1121static void *map_loose_object_1(struct repository *r, const char *path,
1122 const struct object_id *oid, unsigned long *size)
27d69a46 1123{
0fcfd160 1124 void *map;
144bde78 1125 int fd;
84dd81c1 1126
f6371f92
JK
1127 if (path)
1128 fd = git_open(path);
8261e1f1 1129 else
514c5fdd 1130 fd = open_loose_object(r, oid, &path);
44d1c19e
LT
1131 map = NULL;
1132 if (fd >= 0) {
1133 struct stat st;
1f688557 1134
44d1c19e
LT
1135 if (!fstat(fd, &st)) {
1136 *size = xsize_t(st.st_size);
33e42de0
MM
1137 if (!*size) {
1138 /* mmap() is forbidden on empty files */
259328b7 1139 error(_("object file %s is empty"), path);
6881925e 1140 close(fd);
33e42de0
MM
1141 return NULL;
1142 }
44d1c19e 1143 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
144bde78 1144 }
44d1c19e 1145 close(fd);
74e34e1f 1146 }
0fcfd160 1147 return map;
74e34e1f
NP
1148}
1149
514c5fdd
JK
1150void *map_loose_object(struct repository *r,
1151 const struct object_id *oid,
1152 unsigned long *size)
068f85e3 1153{
514c5fdd 1154 return map_loose_object_1(r, NULL, oid, size);
068f85e3 1155}
1156
00a7760e
JK
1157static int unpack_loose_short_header(git_zstream *stream,
1158 unsigned char *map, unsigned long mapsize,
1159 void *buffer, unsigned long bufsiz)
47fe3f6e 1160{
31877c9a
MT
1161 int ret;
1162
c4483576
LT
1163 /* Get the data stream */
1164 memset(stream, 0, sizeof(*stream));
1165 stream->next_in = map;
1166 stream->avail_in = mapsize;
1167 stream->next_out = buffer;
93821bd9 1168 stream->avail_out = bufsiz;
47fe3f6e 1169
39c68542 1170 git_inflate_init(stream);
31877c9a
MT
1171 obj_read_unlock();
1172 ret = git_inflate(stream, 0);
1173 obj_read_lock();
1174
1175 return ret;
9a217f2a
JH
1176}
1177
00a7760e
JK
1178int unpack_loose_header(git_zstream *stream,
1179 unsigned char *map, unsigned long mapsize,
1180 void *buffer, unsigned long bufsiz)
1f688557 1181{
00a7760e
JK
1182 int status = unpack_loose_short_header(stream, map, mapsize,
1183 buffer, bufsiz);
42873078 1184
d21f8426
JH
1185 if (status < Z_OK)
1186 return status;
628522ec 1187
d21f8426
JH
1188 /* Make sure we have the terminating NUL */
1189 if (!memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1190 return -1;
1f688557
JH
1191 return 0;
1192}
1193
00a7760e
JK
1194static int unpack_loose_header_to_strbuf(git_zstream *stream, unsigned char *map,
1195 unsigned long mapsize, void *buffer,
1196 unsigned long bufsiz, struct strbuf *header)
d131b7af 1197{
46f03448
KN
1198 int status;
1199
00a7760e 1200 status = unpack_loose_short_header(stream, map, mapsize, buffer, bufsiz);
d21f8426
JH
1201 if (status < Z_OK)
1202 return -1;
d131b7af 1203
46f03448
KN
1204 /*
1205 * Check if entire header is unpacked in the first iteration.
d131b7af 1206 */
46f03448
KN
1207 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1208 return 0;
d131b7af 1209
46f03448
KN
1210 /*
1211 * buffer[0..bufsiz] was not large enough. Copy the partial
1212 * result out to header, and then append the result of further
1213 * reading the stream.
1214 */
1215 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1216 stream->next_out = buffer;
1217 stream->avail_out = bufsiz;
d131b7af 1218
46f03448 1219 do {
31877c9a 1220 obj_read_unlock();
46f03448 1221 status = git_inflate(stream, 0);
31877c9a 1222 obj_read_lock();
46f03448
KN
1223 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1224 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1225 return 0;
1226 stream->next_out = buffer;
1227 stream->avail_out = bufsiz;
1228 } while (status != Z_STREAM_END);
1229 return -1;
d131b7af
SP
1230}
1231
00a7760e
JK
1232static void *unpack_loose_rest(git_zstream *stream,
1233 void *buffer, unsigned long size,
1234 const struct object_id *oid)
95099731 1235{
5180cacc 1236 int bytes = strlen(buffer) + 1;
3aee68aa 1237 unsigned char *buf = xmallocz(size);
93821bd9 1238 unsigned long n;
7efbff75 1239 int status = Z_OK;
95099731 1240
93821bd9
LT
1241 n = stream->total_out - bytes;
1242 if (n > size)
1243 n = size;
1244 memcpy(buf, (char *) buffer + bytes, n);
1245 bytes = n;
456cdf6e
LT
1246 if (bytes <= size) {
1247 /*
1248 * The above condition must be (bytes <= size), not
1249 * (bytes < size). In other words, even though we
ccf5ace0 1250 * expect no more output and set avail_out to zero,
456cdf6e
LT
1251 * the input zlib stream may have bytes that express
1252 * "this concludes the stream", and we *do* want to
1253 * eat that input.
1254 *
1255 * Otherwise we would not be able to test that we
1256 * consumed all the input to reach the expected size;
1257 * we also want to check that zlib tells us that all
1258 * went well with status == Z_STREAM_END at the end.
1259 */
5180cacc
LT
1260 stream->next_out = buf + bytes;
1261 stream->avail_out = size - bytes;
31877c9a
MT
1262 while (status == Z_OK) {
1263 obj_read_unlock();
39c68542 1264 status = git_inflate(stream, Z_FINISH);
31877c9a
MT
1265 obj_read_lock();
1266 }
5180cacc 1267 }
456cdf6e 1268 if (status == Z_STREAM_END && !stream->avail_in) {
39c68542 1269 git_inflate_end(stream);
7efbff75 1270 return buf;
95099731
NTND
1271 }
1272
7efbff75 1273 if (status < 0)
00a7760e 1274 error(_("corrupt loose object '%s'"), oid_to_hex(oid));
7efbff75 1275 else if (stream->avail_in)
259328b7 1276 error(_("garbage at end of loose object '%s'"),
00a7760e 1277 oid_to_hex(oid));
7efbff75
JH
1278 free(buf);
1279 return NULL;
95099731
NTND
1280}
1281
d40d535b 1282/*
5180cacc
LT
1283 * We used to just use "sscanf()", but that's actually way
1284 * too permissive for what we want to check. So do an anal
1285 * object header parse by hand.
d40d535b 1286 */
00a7760e
JK
1287static int parse_loose_header_extended(const char *hdr, struct object_info *oi,
1288 unsigned int flags)
1f688557 1289{
46f03448 1290 const char *type_buf = hdr;
5180cacc 1291 unsigned long size;
46f03448 1292 int type, type_len = 0;
43057304 1293
5180cacc 1294 /*
46f03448 1295 * The type can be of any size but is followed by
21666f1a 1296 * a space.
5180cacc 1297 */
5180cacc
LT
1298 for (;;) {
1299 char c = *hdr++;
d21f8426
JH
1300 if (!c)
1301 return -1;
5180cacc
LT
1302 if (c == ' ')
1303 break;
46f03448 1304 type_len++;
5180cacc 1305 }
1f688557 1306
46f03448 1307 type = type_from_string_gently(type_buf, type_len, 1);
6ca32f47
BW
1308 if (oi->type_name)
1309 strbuf_add(oi->type_name, type_buf, type_len);
46f03448
KN
1310 /*
1311 * Set type to 0 if its an unknown object and
2e3a16b2 1312 * we're obtaining the type using '--allow-unknown-type'
46f03448
KN
1313 * option.
1314 */
19fc5e84 1315 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0))
46f03448
KN
1316 type = 0;
1317 else if (type < 0)
259328b7 1318 die(_("invalid object type"));
46f03448
KN
1319 if (oi->typep)
1320 *oi->typep = type;
5180cacc
LT
1321
1322 /*
1323 * The length must follow immediately, and be in canonical
1324 * decimal format (ie "010" is not valid).
1325 */
1326 size = *hdr++ - '0';
1327 if (size > 9)
1328 return -1;
1329 if (size) {
1330 for (;;) {
1331 unsigned long c = *hdr - '0';
1332 if (c > 9)
1333 break;
1334 hdr++;
1335 size = size * 10 + c;
1b1005d1 1336 }
c01f51cc 1337 }
46f03448
KN
1338
1339 if (oi->sizep)
1340 *oi->sizep = size;
5180cacc
LT
1341
1342 /*
1343 * The length must be followed by a zero byte
1344 */
46f03448 1345 return *hdr ? -1 : type;
1f688557
JH
1346}
1347
00a7760e 1348int parse_loose_header(const char *hdr, unsigned long *sizep)
bf592c50 1349{
27b5c1a0 1350 struct object_info oi = OBJECT_INFO_INIT;
21666f1a 1351
46f03448 1352 oi.sizep = sizep;
00a7760e 1353 return parse_loose_header_extended(hdr, &oi, 0);
bf592c50
DB
1354}
1355
514c5fdd
JK
1356static int loose_object_info(struct repository *r,
1357 const struct object_id *oid,
1358 struct object_info *oi, int flags)
65c2e0c3 1359{
46f03448
KN
1360 int status = 0;
1361 unsigned long mapsize;
65c2e0c3 1362 void *map;
ef49a7a0 1363 git_zstream stream;
1af64f73 1364 char hdr[MAX_HEADER_LEN];
46f03448 1365 struct strbuf hdrbuf = STRBUF_INIT;
c84a1f3e 1366 unsigned long size_scratch;
65c2e0c3 1367
b99b6bcc
JK
1368 if (oi->delta_base_oid)
1369 oidclr(oi->delta_base_oid);
5d642e75 1370
052fe5ea
JK
1371 /*
1372 * If we don't care about type or size, then we don't
4ef8d1dd
JH
1373 * need to look inside the object at all. Note that we
1374 * do not optimize out the stat call, even if the
1375 * caller doesn't care about the disk-size, since our
1376 * return value implicitly indicates whether the
1377 * object even exists.
052fe5ea 1378 */
6ca32f47 1379 if (!oi->typep && !oi->type_name && !oi->sizep && !oi->contentp) {
771e7d57 1380 const char *path;
4ef8d1dd 1381 struct stat st;
61c7711c 1382 if (!oi->disk_sizep && (flags & OBJECT_INFO_QUICK))
d7a24573 1383 return quick_has_loose(r, oid) ? 0 : -1;
514c5fdd 1384 if (stat_loose_object(r, oid, &st, &path) < 0)
4ef8d1dd
JH
1385 return -1;
1386 if (oi->disk_sizep)
23c339c0 1387 *oi->disk_sizep = st.st_size;
052fe5ea
JK
1388 return 0;
1389 }
1390
514c5fdd 1391 map = map_loose_object(r, oid, &mapsize);
f0df4ed5 1392 if (!map)
dbea72a8 1393 return -1;
c84a1f3e
JT
1394
1395 if (!oi->sizep)
1396 oi->sizep = &size_scratch;
1397
23c339c0
JK
1398 if (oi->disk_sizep)
1399 *oi->disk_sizep = mapsize;
19fc5e84 1400 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) {
00a7760e 1401 if (unpack_loose_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
259328b7 1402 status = error(_("unable to unpack %s header with --allow-unknown-type"),
514c5fdd 1403 oid_to_hex(oid));
00a7760e 1404 } else if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
259328b7 1405 status = error(_("unable to unpack %s header"),
514c5fdd 1406 oid_to_hex(oid));
46f03448
KN
1407 if (status < 0)
1408 ; /* Do nothing */
1409 else if (hdrbuf.len) {
00a7760e 1410 if ((status = parse_loose_header_extended(hdrbuf.buf, oi, flags)) < 0)
259328b7 1411 status = error(_("unable to parse %s header with --allow-unknown-type"),
514c5fdd 1412 oid_to_hex(oid));
00a7760e 1413 } else if ((status = parse_loose_header_extended(hdr, oi, flags)) < 0)
514c5fdd 1414 status = error(_("unable to parse %s header"), oid_to_hex(oid));
c84a1f3e 1415
b3ea7dd3 1416 if (status >= 0 && oi->contentp) {
00a7760e
JK
1417 *oi->contentp = unpack_loose_rest(&stream, hdr,
1418 *oi->sizep, oid);
b3ea7dd3
JK
1419 if (!*oi->contentp) {
1420 git_inflate_end(&stream);
1421 status = -1;
1422 }
1423 } else
c84a1f3e
JT
1424 git_inflate_end(&stream);
1425
65c2e0c3 1426 munmap(map, mapsize);
46f03448 1427 if (status && oi->typep)
23c339c0 1428 *oi->typep = status;
c84a1f3e
JT
1429 if (oi->sizep == &size_scratch)
1430 oi->sizep = NULL;
46f03448 1431 strbuf_release(&hdrbuf);
3ab0fb06 1432 oi->whence = OI_LOOSE;
93cff9a9 1433 return (status < 0) ? status : 0;
65c2e0c3
JH
1434}
1435
31877c9a
MT
1436int obj_read_use_lock = 0;
1437pthread_mutex_t obj_read_mutex;
1438
1439void enable_obj_read_lock(void)
1440{
1441 if (obj_read_use_lock)
1442 return;
1443
1444 obj_read_use_lock = 1;
1445 init_recursive_mutex(&obj_read_mutex);
1446}
1447
1448void disable_obj_read_lock(void)
1449{
1450 if (!obj_read_use_lock)
1451 return;
1452
1453 obj_read_use_lock = 0;
1454 pthread_mutex_destroy(&obj_read_mutex);
1455}
1456
8b4c0103
JT
1457int fetch_if_missing = 1;
1458
31877c9a
MT
1459static int do_oid_object_info_extended(struct repository *r,
1460 const struct object_id *oid,
1461 struct object_info *oi, unsigned flags)
f0df4ed5 1462{
cd585e2a 1463 static struct object_info blank_oi = OBJECT_INFO_INIT;
9c8a294a 1464 struct cached_object *co;
f0df4ed5 1465 struct pack_entry e;
5b086407 1466 int rtype;
b383a13c 1467 const struct object_id *real = oid;
8b4c0103 1468 int already_retried = 0;
f0df4ed5 1469
31877c9a 1470
b383a13c 1471 if (flags & OBJECT_INFO_LOOKUP_REPLACE)
9d98354f 1472 real = lookup_replace_object(r, oid);
f0df4ed5 1473
b383a13c 1474 if (is_null_oid(real))
87b5e236
JK
1475 return -1;
1476
cd585e2a
JT
1477 if (!oi)
1478 oi = &blank_oi;
1479
9c8a294a
JT
1480 co = find_cached_object(real);
1481 if (co) {
1482 if (oi->typep)
1483 *(oi->typep) = co->type;
1484 if (oi->sizep)
1485 *(oi->sizep) = co->size;
1486 if (oi->disk_sizep)
1487 *(oi->disk_sizep) = 0;
b99b6bcc
JK
1488 if (oi->delta_base_oid)
1489 oidclr(oi->delta_base_oid);
9c8a294a
JT
1490 if (oi->type_name)
1491 strbuf_addstr(oi->type_name, type_name(co->type));
1492 if (oi->contentp)
1493 *oi->contentp = xmemdupz(co->buf, co->size);
1494 oi->whence = OI_CACHED;
1495 return 0;
c4d9986f
NTND
1496 }
1497
8b4c0103 1498 while (1) {
42c8ce1c 1499 if (find_pack_entry(r, real, &e))
8b4c0103
JT
1500 break;
1501
024aa469
TI
1502 if (flags & OBJECT_INFO_IGNORE_LOOSE)
1503 return -1;
1504
ddd63e64 1505 /* Most likely it's a loose object. */
514c5fdd 1506 if (!loose_object_info(r, real, oi, flags))
5b086407 1507 return 0;
ddd63e64
SG
1508
1509 /* Not a loose object; someone else may have just packed it. */
2b7750c9 1510 if (!(flags & OBJECT_INFO_QUICK)) {
9d98354f 1511 reprepare_packed_git(r);
42c8ce1c 1512 if (find_pack_entry(r, real, &e))
2b7750c9
JT
1513 break;
1514 }
8b4c0103
JT
1515
1516 /* Check if it is a missing object */
b14ed5ad 1517 if (fetch_if_missing && has_promisor_remote() &&
0f4a4fb1 1518 !already_retried && r == the_repository &&
31f5256c 1519 !(flags & OBJECT_INFO_SKIP_FETCH_OBJECT)) {
8b4c0103 1520 /*
b14ed5ad
CC
1521 * TODO Investigate checking promisor_remote_get_direct()
1522 * TODO return value and stopping on error here.
1523 * TODO Pass a repository struct through
1524 * promisor_remote_get_direct(), such that arbitrary
1525 * repositories work.
8b4c0103 1526 */
b14ed5ad 1527 promisor_remote_get_direct(r, real, 1);
8b4c0103
JT
1528 already_retried = 1;
1529 continue;
dfdd4afc 1530 }
8b4c0103
JT
1531
1532 return -1;
f0df4ed5 1533 }
3d77d877 1534
cd585e2a
JT
1535 if (oi == &blank_oi)
1536 /*
1537 * We know that the caller doesn't actually need the
1538 * information below, so return early.
1539 */
1540 return 0;
9d98354f 1541 rtype = packed_object_info(r, e.p, e.offset, oi);
412916ee 1542 if (rtype < 0) {
b383a13c 1543 mark_bad_packed_object(e.p, real->hash);
31877c9a 1544 return do_oid_object_info_extended(r, real, oi, 0);
3ab0fb06 1545 } else if (oi->whence == OI_PACKED) {
9a490590
JH
1546 oi->u.packed.offset = e.offset;
1547 oi->u.packed.pack = e.p;
1548 oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
1549 rtype == OBJ_OFS_DELTA);
3d77d877
NP
1550 }
1551
5b086407 1552 return 0;
f0df4ed5
JS
1553}
1554
31877c9a
MT
1555int oid_object_info_extended(struct repository *r, const struct object_id *oid,
1556 struct object_info *oi, unsigned flags)
1557{
1558 int ret;
1559 obj_read_lock();
1560 ret = do_oid_object_info_extended(r, oid, oi, flags);
1561 obj_read_unlock();
1562 return ret;
1563}
1564
1565
3fc0dca9 1566/* returns enum object_type or negative */
9d98354f
SB
1567int oid_object_info(struct repository *r,
1568 const struct object_id *oid,
1569 unsigned long *sizep)
9a490590 1570{
5b086407 1571 enum object_type type;
27b5c1a0 1572 struct object_info oi = OBJECT_INFO_INIT;
9a490590 1573
5b086407 1574 oi.typep = &type;
9a490590 1575 oi.sizep = sizep;
9d98354f
SB
1576 if (oid_object_info_extended(r, oid, &oi,
1577 OBJECT_INFO_LOOKUP_REPLACE) < 0)
5b086407
JK
1578 return -1;
1579 return type;
9a490590
JH
1580}
1581
1b9b5c69 1582static void *read_object(struct repository *r,
cba595ab 1583 const struct object_id *oid, enum object_type *type,
f1d8130b
JT
1584 unsigned long *size)
1585{
1586 struct object_info oi = OBJECT_INFO_INIT;
1587 void *content;
1588 oi.typep = type;
1589 oi.sizep = size;
1590 oi.contentp = &content;
1591
cba595ab 1592 if (oid_object_info_extended(r, oid, &oi, 0) < 0)
f1d8130b
JT
1593 return NULL;
1594 return content;
1595}
1596
829e5c3b
PO
1597int pretend_object_file(void *buf, unsigned long len, enum object_type type,
1598 struct object_id *oid)
d66b37bb
JH
1599{
1600 struct cached_object *co;
1601
2dcde20e 1602 hash_object_file(the_hash_algo, buf, len, type_name(type), oid);
a64d2aae
JT
1603 if (has_object_file_with_flags(oid, OBJECT_INFO_QUICK | OBJECT_INFO_SKIP_FETCH_OBJECT) ||
1604 find_cached_object(oid))
d66b37bb 1605 return 0;
c7353967 1606 ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc);
d66b37bb
JH
1607 co = &cached_objects[cached_object_nr++];
1608 co->size = len;
21666f1a 1609 co->type = type;
efa13f7b
JH
1610 co->buf = xmalloc(len);
1611 memcpy(co->buf, buf, len);
62ba93ea 1612 oidcpy(&co->oid, oid);
d66b37bb
JH
1613 return 0;
1614}
1615
b6c4cecc
JH
1616/*
1617 * This function dies on corrupt objects; the callers who want to
1618 * deal with them should arrange to call read_object() and give error
1619 * messages themselves.
1620 */
a3b72c89
SB
1621void *read_object_file_extended(struct repository *r,
1622 const struct object_id *oid,
b4f5aca4 1623 enum object_type *type,
1624 unsigned long *size,
1625 int lookup_replace)
ac939109 1626{
3ba7a065 1627 void *data;
b6c4cecc 1628 const struct packed_git *p;
771e7d57
JK
1629 const char *path;
1630 struct stat st;
1f2e7cea 1631 const struct object_id *repl = lookup_replace ?
a3b72c89 1632 lookup_replace_object(r, oid) : oid;
b6c4cecc 1633
3ba7a065 1634 errno = 0;
cba595ab 1635 data = read_object(r, repl, type, size);
4bbf5a26 1636 if (data)
b6c4cecc 1637 return data;
68095570 1638
31877c9a 1639 obj_read_lock();
25f3af3f 1640 if (errno && errno != ENOENT)
259328b7 1641 die_errno(_("failed to read object %s"), oid_to_hex(oid));
3ba7a065 1642
68095570 1643 /* die if we replaced an object with one that does not exist */
b383a13c 1644 if (repl != oid)
259328b7 1645 die(_("replacement %s not found for %s"),
b383a13c 1646 oid_to_hex(repl), oid_to_hex(oid));
68095570 1647
cba595ab 1648 if (!stat_loose_object(r, repl, &st, &path))
259328b7 1649 die(_("loose object %s (stored in %s) is corrupt"),
b383a13c 1650 oid_to_hex(repl), path);
68095570 1651
a3b72c89 1652 if ((p = has_packed_and_bad(r, repl->hash)) != NULL)
259328b7 1653 die(_("packed object %s (stored in %s) is corrupt"),
b383a13c 1654 oid_to_hex(repl), p->pack_name);
31877c9a 1655 obj_read_unlock();
f5552aee 1656
b6c4cecc 1657 return NULL;
ac939109
NP
1658}
1659
d3b4705a
NTND
1660void *read_object_with_reference(struct repository *r,
1661 const struct object_id *oid,
21666f1a 1662 const char *required_type_name,
40469ee9 1663 unsigned long *size,
02f0547e 1664 struct object_id *actual_oid_return)
f4913f91 1665{
21666f1a 1666 enum object_type type, required_type;
f4913f91
JH
1667 void *buffer;
1668 unsigned long isize;
02f0547e 1669 struct object_id actual_oid;
f4913f91 1670
21666f1a 1671 required_type = type_from_string(required_type_name);
02f0547e 1672 oidcpy(&actual_oid, oid);
40469ee9
JH
1673 while (1) {
1674 int ref_length = -1;
1675 const char *ref_type = NULL;
f4913f91 1676
d3b4705a 1677 buffer = repo_read_object_file(r, &actual_oid, &type, &isize);
40469ee9
JH
1678 if (!buffer)
1679 return NULL;
21666f1a 1680 if (type == required_type) {
40469ee9 1681 *size = isize;
02f0547e 1682 if (actual_oid_return)
1683 oidcpy(actual_oid_return, &actual_oid);
40469ee9
JH
1684 return buffer;
1685 }
1686 /* Handle references */
21666f1a 1687 else if (type == OBJ_COMMIT)
40469ee9 1688 ref_type = "tree ";
21666f1a 1689 else if (type == OBJ_TAG)
40469ee9
JH
1690 ref_type = "object ";
1691 else {
1692 free(buffer);
1693 return NULL;
1694 }
1695 ref_length = strlen(ref_type);
f4913f91 1696
94b5e093 1697 if (ref_length + the_hash_algo->hexsz > isize ||
50974ec9 1698 memcmp(buffer, ref_type, ref_length) ||
02f0547e 1699 get_oid_hex((char *) buffer + ref_length, &actual_oid)) {
40469ee9
JH
1700 free(buffer);
1701 return NULL;
1702 }
1cf58e72 1703 free(buffer);
40469ee9 1704 /* Now we have the ID of the referred-to object in
02f0547e 1705 * actual_oid. Check again. */
f4913f91 1706 }
f4913f91
JH
1707}
1708
7ad5c44d
MT
1709static void write_object_file_prepare(const struct git_hash_algo *algo,
1710 const void *buf, unsigned long len,
a09c985e
PO
1711 const char *type, struct object_id *oid,
1712 char *hdr, int *hdrlen)
d410c0f5 1713{
18e2588e 1714 git_hash_ctx c;
d410c0f5
JH
1715
1716 /* Generate the header */
ca473cef 1717 *hdrlen = xsnprintf(hdr, *hdrlen, "%s %"PRIuMAX , type, (uintmax_t)len)+1;
d410c0f5
JH
1718
1719 /* Sha1.. */
7ad5c44d
MT
1720 algo->init_fn(&c);
1721 algo->update_fn(&c, hdr, *hdrlen);
1722 algo->update_fn(&c, buf, len);
1723 algo->final_fn(oid->hash, &c);
d410c0f5
JH
1724}
1725
230f1322 1726/*
5a688fe4 1727 * Move the just written object into its final resting place.
230f1322 1728 */
cb5add58 1729int finalize_object_file(const char *tmpfile, const char *filename)
230f1322 1730{
e32c0a9c 1731 int ret = 0;
5a688fe4 1732
348df166 1733 if (object_creation_mode == OBJECT_CREATION_USES_RENAMES)
be66a6c4
JS
1734 goto try_rename;
1735 else if (link(tmpfile, filename))
e32c0a9c 1736 ret = errno;
7ebb6fca
LT
1737
1738 /*
1739 * Coda hack - coda doesn't like cross-directory links,
1740 * so we fall back to a rename, which will mean that it
1741 * won't be able to check collisions, but that's not a
1742 * big deal.
1743 *
1744 * The same holds for FAT formatted media.
1745 *
3be1f18e 1746 * When this succeeds, we just return. We have nothing
7ebb6fca
LT
1747 * left to unlink.
1748 */
1749 if (ret && ret != EEXIST) {
be66a6c4 1750 try_rename:
7ebb6fca 1751 if (!rename(tmpfile, filename))
3be1f18e 1752 goto out;
9e48b389 1753 ret = errno;
230f1322 1754 }
691f1a28 1755 unlink_or_warn(tmpfile);
230f1322
LT
1756 if (ret) {
1757 if (ret != EEXIST) {
2c319886 1758 return error_errno(_("unable to write file %s"), filename);
230f1322
LT
1759 }
1760 /* FIXME!!! Collision check here ? */
1761 }
1762
3be1f18e 1763out:
5256b006 1764 if (adjust_shared_perm(filename))
259328b7 1765 return error(_("unable to set permission to '%s'"), filename);
230f1322
LT
1766 return 0;
1767}
1768
4d548150
LT
1769static int write_buffer(int fd, const void *buf, size_t len)
1770{
d34cf19b 1771 if (write_in_full(fd, buf, len) < 0)
259328b7 1772 return error_errno(_("file write error"));
4d548150
LT
1773 return 0;
1774}
1775
2dcde20e
MT
1776int hash_object_file(const struct git_hash_algo *algo, const void *buf,
1777 unsigned long len, const char *type,
f070facc 1778 struct object_id *oid)
abdc3fc8 1779{
1af64f73 1780 char hdr[MAX_HEADER_LEN];
ef1286d3 1781 int hdrlen = sizeof(hdr);
2dcde20e 1782 write_object_file_prepare(algo, buf, len, type, oid, hdr, &hdrlen);
abdc3fc8
RS
1783 return 0;
1784}
1785
e9039dd3 1786/* Finalize a file on disk, and close it. */
514c5fdd 1787static void close_loose_object(int fd)
e9039dd3 1788{
aafe9fba 1789 if (fsync_object_files)
76011357 1790 fsync_or_die(fd, "loose object file");
e9039dd3 1791 if (close(fd) != 0)
76011357 1792 die_errno(_("error when closing loose object file"));
e9039dd3
LT
1793}
1794
5723fe7e
LT
1795/* Size of directory component, including the ending '/' */
1796static inline int directory_size(const char *filename)
1797{
1798 const char *s = strrchr(filename, '/');
1799 if (!s)
1800 return 0;
1801 return s - filename + 1;
1802}
1803
1804/*
1805 * This creates a temporary file in the same directory as the final
1806 * 'filename'
1807 *
1808 * We want to avoid cross-directory filename renames, because those
1809 * can have problems on various filesystems (FAT, NFS, Coda).
1810 */
d4b3d11a 1811static int create_tmpfile(struct strbuf *tmp, const char *filename)
5723fe7e
LT
1812{
1813 int fd, dirlen = directory_size(filename);
1814
d4b3d11a
JK
1815 strbuf_reset(tmp);
1816 strbuf_add(tmp, filename, dirlen);
1817 strbuf_addstr(tmp, "tmp_obj_XXXXXX");
1818 fd = git_mkstemp_mode(tmp->buf, 0444);
cbacbf4e 1819 if (fd < 0 && dirlen && errno == ENOENT) {
d4b3d11a
JK
1820 /*
1821 * Make sure the directory exists; note that the contents
1822 * of the buffer are undefined after mkstemp returns an
1823 * error, so we have to rewrite the whole buffer from
1824 * scratch.
1825 */
1826 strbuf_reset(tmp);
1827 strbuf_add(tmp, filename, dirlen - 1);
1828 if (mkdir(tmp->buf, 0777) && errno != EEXIST)
b2476a60 1829 return -1;
d4b3d11a 1830 if (adjust_shared_perm(tmp->buf))
5723fe7e
LT
1831 return -1;
1832
1833 /* Try again */
d4b3d11a
JK
1834 strbuf_addstr(tmp, "/tmp_obj_XXXXXX");
1835 fd = git_mkstemp_mode(tmp->buf, 0444);
5723fe7e
LT
1836 }
1837 return fd;
1838}
1839
3fc7281f
PO
1840static int write_loose_object(const struct object_id *oid, char *hdr,
1841 int hdrlen, const void *buf, unsigned long len,
1842 time_t mtime)
0fcfd160 1843{
915308b1 1844 int fd, ret;
9892beba 1845 unsigned char compressed[4096];
ef49a7a0 1846 git_zstream stream;
18e2588e 1847 git_hash_ctx c;
3fc7281f 1848 struct object_id parano_oid;
d4b3d11a 1849 static struct strbuf tmp_file = STRBUF_INIT;
ea657730
CC
1850 static struct strbuf filename = STRBUF_INIT;
1851
514c5fdd 1852 loose_object_path(the_repository, &filename, oid);
a44c9a5e 1853
ea657730 1854 fd = create_tmpfile(&tmp_file, filename.buf);
aac17941 1855 if (fd < 0) {
35243577 1856 if (errno == EACCES)
259328b7 1857 return error(_("insufficient permission for adding an object to repository database %s"), get_object_directory());
916d081b 1858 else
259328b7 1859 return error_errno(_("unable to create temporary file"));
aac17941
LT
1860 }
1861
0fcfd160 1862 /* Set it up */
55bb5c91 1863 git_deflate_init(&stream, zlib_compression_level);
0fcfd160 1864 stream.next_out = compressed;
9892beba 1865 stream.avail_out = sizeof(compressed);
18e2588e 1866 the_hash_algo->init_fn(&c);
a44c9a5e
LT
1867
1868 /* First header.. */
d65a16f6 1869 stream.next_in = (unsigned char *)hdr;
a44c9a5e 1870 stream.avail_in = hdrlen;
55bb5c91
JH
1871 while (git_deflate(&stream, 0) == Z_OK)
1872 ; /* nothing */
18e2588e 1873 the_hash_algo->update_fn(&c, hdr, hdrlen);
a44c9a5e
LT
1874
1875 /* Then the data itself.. */
c00e657d 1876 stream.next_in = (void *)buf;
a44c9a5e 1877 stream.avail_in = len;
9892beba 1878 do {
748af44c 1879 unsigned char *in0 = stream.next_in;
55bb5c91 1880 ret = git_deflate(&stream, Z_FINISH);
18e2588e 1881 the_hash_algo->update_fn(&c, in0, stream.next_in - in0);
9892beba 1882 if (write_buffer(fd, compressed, stream.next_out - compressed) < 0)
76011357 1883 die(_("unable to write loose object file"));
9892beba
NP
1884 stream.next_out = compressed;
1885 stream.avail_out = sizeof(compressed);
1886 } while (ret == Z_OK);
1887
ac54c277 1888 if (ret != Z_STREAM_END)
259328b7 1889 die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid),
3fc7281f 1890 ret);
55bb5c91 1891 ret = git_deflate_end_gently(&stream);
ac54c277 1892 if (ret != Z_OK)
259328b7 1893 die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid),
3fc7281f 1894 ret);
18e2588e 1895 the_hash_algo->final_fn(parano_oid.hash, &c);
9001dc2a 1896 if (!oideq(oid, &parano_oid))
259328b7 1897 die(_("confused by unstable object source data for %s"),
3fc7281f 1898 oid_to_hex(oid));
ac54c277 1899
514c5fdd 1900 close_loose_object(fd);
0fcfd160 1901
bbac7311
NP
1902 if (mtime) {
1903 struct utimbuf utb;
1904 utb.actime = mtime;
1905 utb.modtime = mtime;
d4b3d11a 1906 if (utime(tmp_file.buf, &utb) < 0)
259328b7 1907 warning_errno(_("failed utime() on %s"), tmp_file.buf);
bbac7311
NP
1908 }
1909
ea657730 1910 return finalize_object_file(tmp_file.buf, filename.buf);
0fcfd160 1911}
8237b185 1912
6862ebbf 1913static int freshen_loose_object(const struct object_id *oid)
33d4221c 1914{
6862ebbf 1915 return check_and_freshen(oid, 1);
33d4221c
JK
1916}
1917
6862ebbf 1918static int freshen_packed_object(const struct object_id *oid)
33d4221c
JK
1919{
1920 struct pack_entry e;
544443cb 1921 if (!find_pack_entry(the_repository, oid, &e))
ee1c6c34
JK
1922 return 0;
1923 if (e.p->freshened)
1924 return 1;
1925 if (!freshen_file(e.p->pack_name))
1926 return 0;
1927 e.p->freshened = 1;
1928 return 1;
33d4221c
JK
1929}
1930
a09c985e
PO
1931int write_object_file(const void *buf, unsigned long len, const char *type,
1932 struct object_id *oid)
bbac7311 1933{
1af64f73 1934 char hdr[MAX_HEADER_LEN];
ef1286d3 1935 int hdrlen = sizeof(hdr);
bbac7311
NP
1936
1937 /* Normally if we have it in the pack then we do not bother writing
1938 * it out into .git/objects/??/?{38} file.
1939 */
7ad5c44d
MT
1940 write_object_file_prepare(the_hash_algo, buf, len, type, oid, hdr,
1941 &hdrlen);
6862ebbf 1942 if (freshen_packed_object(oid) || freshen_loose_object(oid))
bbac7311 1943 return 0;
3fc7281f 1944 return write_loose_object(oid, hdr, hdrlen, buf, len, 0);
bbac7311
NP
1945}
1946
1752cbbc
PO
1947int hash_object_file_literally(const void *buf, unsigned long len,
1948 const char *type, struct object_id *oid,
1949 unsigned flags)
0c3db67c
ES
1950{
1951 char *header;
1952 int hdrlen, status = 0;
1953
1954 /* type string, SP, %lu of the length plus NUL must fit this */
1af64f73 1955 hdrlen = strlen(type) + MAX_HEADER_LEN;
ef1286d3 1956 header = xmalloc(hdrlen);
7ad5c44d
MT
1957 write_object_file_prepare(the_hash_algo, buf, len, type, oid, header,
1958 &hdrlen);
0c3db67c
ES
1959
1960 if (!(flags & HASH_WRITE_OBJECT))
1961 goto cleanup;
6862ebbf 1962 if (freshen_packed_object(oid) || freshen_loose_object(oid))
0c3db67c 1963 goto cleanup;
3fc7281f 1964 status = write_loose_object(oid, header, hdrlen, buf, len, 0);
0c3db67c
ES
1965
1966cleanup:
1967 free(header);
1968 return status;
1969}
1970
4bdb70a4 1971int force_object_loose(const struct object_id *oid, time_t mtime)
bbac7311 1972{
bbac7311
NP
1973 void *buf;
1974 unsigned long len;
1975 enum object_type type;
1af64f73 1976 char hdr[MAX_HEADER_LEN];
bbac7311 1977 int hdrlen;
1fb23e65 1978 int ret;
bbac7311 1979
6862ebbf 1980 if (has_loose_object(oid))
bbac7311 1981 return 0;
cba595ab 1982 buf = read_object(the_repository, oid, &type, &len);
bbac7311 1983 if (!buf)
2c319886 1984 return error(_("cannot read object for %s"), oid_to_hex(oid));
ca473cef 1985 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(type), (uintmax_t)len) + 1;
3fc7281f 1986 ret = write_loose_object(oid, hdr, hdrlen, buf, len, mtime);
1fb23e65
BS
1987 free(buf);
1988
1989 return ret;
bbac7311
NP
1990}
1991
1d8d9cb6
JT
1992int has_object(struct repository *r, const struct object_id *oid,
1993 unsigned flags)
1994{
1995 int quick = !(flags & HAS_OBJECT_RECHECK_PACKED);
1996 unsigned object_info_flags = OBJECT_INFO_SKIP_FETCH_OBJECT |
1997 (quick ? OBJECT_INFO_QUICK : 0);
1998
1999 if (!startup_info->have_repository)
2000 return 0;
2001 return oid_object_info_extended(r, oid, NULL, object_info_flags) >= 0;
2002}
2003
cba595ab
JH
2004int repo_has_object_file_with_flags(struct repository *r,
2005 const struct object_id *oid, int flags)
8237b185 2006{
3e8b7d3c
JN
2007 if (!startup_info->have_repository)
2008 return 0;
9c8a294a 2009 return oid_object_info_extended(r, oid, NULL, flags) >= 0;
b419aa25 2010}
2011
9b45f499
SB
2012int repo_has_object_file(struct repository *r,
2013 const struct object_id *oid)
b419aa25 2014{
cba595ab 2015 return repo_has_object_file_with_flags(r, oid, 0);
5827a035
JK
2016}
2017
c879daa2
NTND
2018static void check_tree(const void *buf, size_t size)
2019{
2020 struct tree_desc desc;
2021 struct name_entry entry;
2022
2023 init_tree_desc(&desc, buf, size);
2024 while (tree_entry(&desc, &entry))
2025 /* do nothing
2026 * tree_entry() will die() on malformed entries */
2027 ;
2028}
2029
2030static void check_commit(const void *buf, size_t size)
2031{
2032 struct commit c;
2033 memset(&c, 0, sizeof(c));
08f4f445 2034 if (parse_commit_buffer(the_repository, &c, buf, size, 0))
259328b7 2035 die(_("corrupt commit"));
c879daa2
NTND
2036}
2037
2038static void check_tag(const void *buf, size_t size)
2039{
2040 struct tag t;
2041 memset(&t, 0, sizeof(t));
0e740fed 2042 if (parse_tag_buffer(the_repository, &t, buf, size))
259328b7 2043 die(_("corrupt tag"));
c879daa2
NTND
2044}
2045
58bf2a4c
NTND
2046static int index_mem(struct index_state *istate,
2047 struct object_id *oid, void *buf, size_t size,
c4ce46fc
JH
2048 enum object_type type,
2049 const char *path, unsigned flags)
e7332f96 2050{
6c510bee 2051 int ret, re_allocated = 0;
c4ce46fc 2052 int write_object = flags & HASH_WRITE_OBJECT;
74400e71 2053
7672db20 2054 if (!type)
edaec3fb 2055 type = OBJ_BLOB;
6c510bee
LT
2056
2057 /*
2058 * Convert blobs to git internal format
2059 */
43df4f86 2060 if ((type == OBJ_BLOB) && path) {
f285a2d7 2061 struct strbuf nbuf = STRBUF_INIT;
58bf2a4c 2062 if (convert_to_git(istate, path, buf, size, &nbuf,
8462ff43 2063 get_conv_flags(flags))) {
b315c5c0 2064 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
2065 re_allocated = 1;
2066 }
2067 }
c4ce46fc 2068 if (flags & HASH_FORMAT_CHECK) {
c879daa2
NTND
2069 if (type == OBJ_TREE)
2070 check_tree(buf, size);
2071 if (type == OBJ_COMMIT)
2072 check_commit(buf, size);
2073 if (type == OBJ_TAG)
2074 check_tag(buf, size);
2075 }
6c510bee 2076
7672db20 2077 if (write_object)
169c9c01 2078 ret = write_object_file(buf, size, type_name(type), oid);
abdc3fc8 2079 else
2dcde20e
MT
2080 ret = hash_object_file(the_hash_algo, buf, size,
2081 type_name(type), oid);
43df4f86 2082 if (re_allocated)
6c510bee 2083 free(buf);
43df4f86
DP
2084 return ret;
2085}
2086
58bf2a4c
NTND
2087static int index_stream_convert_blob(struct index_state *istate,
2088 struct object_id *oid,
2089 int fd,
2090 const char *path,
2091 unsigned flags)
9035d75a
SP
2092{
2093 int ret;
2094 const int write_object = flags & HASH_WRITE_OBJECT;
2095 struct strbuf sbuf = STRBUF_INIT;
2096
2097 assert(path);
58bf2a4c 2098 assert(would_convert_to_git_filter_fd(istate, path));
9035d75a 2099
58bf2a4c 2100 convert_to_git_filter_fd(istate, path, fd, &sbuf,
8462ff43 2101 get_conv_flags(flags));
9035d75a
SP
2102
2103 if (write_object)
169c9c01 2104 ret = write_object_file(sbuf.buf, sbuf.len, type_name(OBJ_BLOB),
a09c985e 2105 oid);
9035d75a 2106 else
2dcde20e
MT
2107 ret = hash_object_file(the_hash_algo, sbuf.buf, sbuf.len,
2108 type_name(OBJ_BLOB), oid);
9035d75a
SP
2109 strbuf_release(&sbuf);
2110 return ret;
2111}
2112
58bf2a4c
NTND
2113static int index_pipe(struct index_state *istate, struct object_id *oid,
2114 int fd, enum object_type type,
7b41e1e1
JH
2115 const char *path, unsigned flags)
2116{
2117 struct strbuf sbuf = STRBUF_INIT;
2118 int ret;
2119
2120 if (strbuf_read(&sbuf, fd, 4096) >= 0)
58bf2a4c 2121 ret = index_mem(istate, oid, sbuf.buf, sbuf.len, type, path, flags);
7b41e1e1
JH
2122 else
2123 ret = -1;
2124 strbuf_release(&sbuf);
2125 return ret;
2126}
2127
ea68b0ce
DP
2128#define SMALL_FILE_SIZE (32*1024)
2129
58bf2a4c
NTND
2130static int index_core(struct index_state *istate,
2131 struct object_id *oid, int fd, size_t size,
7b41e1e1
JH
2132 enum object_type type, const char *path,
2133 unsigned flags)
43df4f86
DP
2134{
2135 int ret;
43df4f86 2136
7b41e1e1 2137 if (!size) {
58bf2a4c 2138 ret = index_mem(istate, oid, "", size, type, path, flags);
ea68b0ce
DP
2139 } else if (size <= SMALL_FILE_SIZE) {
2140 char *buf = xmalloc(size);
90dca671
JK
2141 ssize_t read_result = read_in_full(fd, buf, size);
2142 if (read_result < 0)
259328b7 2143 ret = error_errno(_("read error while indexing %s"),
90dca671
JK
2144 path ? path : "<unknown>");
2145 else if (read_result != size)
259328b7 2146 ret = error(_("short read while indexing %s"),
90dca671 2147 path ? path : "<unknown>");
ea68b0ce 2148 else
58bf2a4c 2149 ret = index_mem(istate, oid, buf, size, type, path, flags);
ea68b0ce 2150 free(buf);
08bda208 2151 } else {
43df4f86 2152 void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
58bf2a4c 2153 ret = index_mem(istate, oid, buf, size, type, path, flags);
aac17941 2154 munmap(buf, size);
08bda208 2155 }
7b41e1e1
JH
2156 return ret;
2157}
2158
4dd1fbc7 2159/*
568508e7
JH
2160 * This creates one packfile per large blob unless bulk-checkin
2161 * machinery is "plugged".
4dd1fbc7
JH
2162 *
2163 * This also bypasses the usual "convert-to-git" dance, and that is on
2164 * purpose. We could write a streaming version of the converting
2165 * functions and insert that before feeding the data to fast-import
4f22b101
JK
2166 * (or equivalent in-core API described above). However, that is
2167 * somewhat complicated, as we do not know the size of the filter
2168 * result, which we need to know beforehand when writing a git object.
2169 * Since the primary motivation for trying to stream from the working
2170 * tree file and to avoid mmaping it in core is to deal with large
2171 * binary blobs, they generally do not want to get any conversion, and
2172 * callers should avoid this code path when filters are requested.
4dd1fbc7 2173 */
7d5e1dc3 2174static int index_stream(struct object_id *oid, int fd, size_t size,
4dd1fbc7
JH
2175 enum object_type type, const char *path,
2176 unsigned flags)
2177{
68ee6dfc 2178 return index_bulk_checkin(oid, fd, size, type, path, flags);
4dd1fbc7
JH
2179}
2180
58bf2a4c
NTND
2181int index_fd(struct index_state *istate, struct object_id *oid,
2182 int fd, struct stat *st,
7b41e1e1
JH
2183 enum object_type type, const char *path, unsigned flags)
2184{
2185 int ret;
7b41e1e1 2186
9079ab7c
SP
2187 /*
2188 * Call xsize_t() only when needed to avoid potentially unnecessary
2189 * die() for large files.
2190 */
58bf2a4c
NTND
2191 if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(istate, path))
2192 ret = index_stream_convert_blob(istate, oid, fd, path, flags);
9035d75a 2193 else if (!S_ISREG(st->st_mode))
58bf2a4c 2194 ret = index_pipe(istate, oid, fd, type, path, flags);
9079ab7c 2195 else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
58bf2a4c
NTND
2196 (path && would_convert_to_git(istate, path)))
2197 ret = index_core(istate, oid, fd, xsize_t(st->st_size),
2198 type, path, flags);
4dd1fbc7 2199 else
7d5e1dc3 2200 ret = index_stream(oid, fd, xsize_t(st->st_size), type, path,
9079ab7c 2201 flags);
43df4f86 2202 close(fd);
aac17941 2203 return ret;
74400e71 2204}
ec1fcc16 2205
58bf2a4c
NTND
2206int index_path(struct index_state *istate, struct object_id *oid,
2207 const char *path, struct stat *st, unsigned flags)
ec1fcc16
JH
2208{
2209 int fd;
b760d3aa 2210 struct strbuf sb = STRBUF_INIT;
ea8e0297 2211 int rc = 0;
ec1fcc16
JH
2212
2213 switch (st->st_mode & S_IFMT) {
2214 case S_IFREG:
2215 fd = open(path, O_RDONLY);
2216 if (fd < 0)
7616c6ca 2217 return error_errno("open(\"%s\")", path);
58bf2a4c 2218 if (index_fd(istate, oid, fd, st, OBJ_BLOB, path, flags) < 0)
259328b7 2219 return error(_("%s: failed to insert into database"),
ec1fcc16
JH
2220 path);
2221 break;
2222 case S_IFLNK:
7616c6ca
NTND
2223 if (strbuf_readlink(&sb, path, st->st_size))
2224 return error_errno("readlink(\"%s\")", path);
c4ce46fc 2225 if (!(flags & HASH_WRITE_OBJECT))
2dcde20e
MT
2226 hash_object_file(the_hash_algo, sb.buf, sb.len,
2227 blob_type, oid);
a09c985e 2228 else if (write_object_file(sb.buf, sb.len, blob_type, oid))
259328b7 2229 rc = error(_("%s: failed to insert into database"), path);
b760d3aa 2230 strbuf_release(&sb);
ec1fcc16 2231 break;
f35a6d3b 2232 case S_IFDIR:
a98e6101 2233 return resolve_gitlink_ref(path, "HEAD", oid);
ec1fcc16 2234 default:
259328b7 2235 return error(_("%s: unsupported file type"), path);
ec1fcc16 2236 }
ea8e0297 2237 return rc;
ec1fcc16 2238}
a69e5429
JH
2239
2240int read_pack_header(int fd, struct pack_header *header)
2241{
f48ecd38 2242 if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header))
c697ad14
HO
2243 /* "eof before pack header was fully read" */
2244 return PH_ERROR_EOF;
2245
a69e5429
JH
2246 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2247 /* "protocol error (pack signature mismatch detected)" */
2248 return PH_ERROR_PACK_SIGNATURE;
2249 if (!pack_version_ok(header->hdr_version))
2250 /* "protocol error (pack version unsupported)" */
2251 return PH_ERROR_PROTOCOL;
2252 return 0;
2253}
40d52ff7 2254
e816caa0 2255void assert_oid_type(const struct object_id *oid, enum object_type expect)
40d52ff7 2256{
0df8e965 2257 enum object_type type = oid_object_info(the_repository, oid, NULL);
40d52ff7 2258 if (type < 0)
259328b7 2259 die(_("%s is not a valid object"), oid_to_hex(oid));
40d52ff7 2260 if (type != expect)
259328b7 2261 die(_("%s is not a valid '%s' object"), oid_to_hex(oid),
debca9d2 2262 type_name(expect));
40d52ff7 2263}
27e1e22d 2264
70c49050 2265int for_each_file_in_obj_subdir(unsigned int subdir_nr,
cc817ca3
RS
2266 struct strbuf *path,
2267 each_loose_object_fn obj_cb,
2268 each_loose_cruft_fn cruft_cb,
2269 each_loose_subdir_fn subdir_cb,
2270 void *data)
27e1e22d 2271{
0375f472
RS
2272 size_t origlen, baselen;
2273 DIR *dir;
27e1e22d
JK
2274 struct dirent *de;
2275 int r = 0;
62a24c89 2276 struct object_id oid;
27e1e22d 2277
70c49050
RS
2278 if (subdir_nr > 0xff)
2279 BUG("invalid loose object subdirectory: %x", subdir_nr);
2280
0375f472
RS
2281 origlen = path->len;
2282 strbuf_complete(path, '/');
2283 strbuf_addf(path, "%02x", subdir_nr);
0375f472
RS
2284
2285 dir = opendir(path->buf);
27e1e22d 2286 if (!dir) {
0375f472 2287 if (errno != ENOENT)
259328b7 2288 r = error_errno(_("unable to open %s"), path->buf);
0375f472
RS
2289 strbuf_setlen(path, origlen);
2290 return r;
27e1e22d
JK
2291 }
2292
62a24c89 2293 oid.hash[0] = subdir_nr;
163ee5e6
DS
2294 strbuf_addch(path, '/');
2295 baselen = path->len;
62a24c89 2296
27e1e22d 2297 while ((de = readdir(dir))) {
163ee5e6 2298 size_t namelen;
27e1e22d
JK
2299 if (is_dot_or_dotdot(de->d_name))
2300 continue;
2301
163ee5e6 2302 namelen = strlen(de->d_name);
27e1e22d 2303 strbuf_setlen(path, baselen);
163ee5e6 2304 strbuf_add(path, de->d_name, namelen);
94b5e093 2305 if (namelen == the_hash_algo->hexsz - 2 &&
62a24c89 2306 !hex_to_bytes(oid.hash + 1, de->d_name,
94b5e093 2307 the_hash_algo->rawsz - 1)) {
62a24c89
RS
2308 if (obj_cb) {
2309 r = obj_cb(&oid, path->buf, data);
2310 if (r)
2311 break;
27e1e22d 2312 }
62a24c89 2313 continue;
27e1e22d
JK
2314 }
2315
2316 if (cruft_cb) {
2317 r = cruft_cb(de->d_name, path->buf, data);
2318 if (r)
2319 break;
2320 }
2321 }
094c7e63 2322 closedir(dir);
27e1e22d 2323
163ee5e6 2324 strbuf_setlen(path, baselen - 1);
27e1e22d
JK
2325 if (!r && subdir_cb)
2326 r = subdir_cb(subdir_nr, path->buf, data);
2327
0375f472
RS
2328 strbuf_setlen(path, origlen);
2329
27e1e22d
JK
2330 return r;
2331}
2332
e6f875e0 2333int for_each_loose_file_in_objdir_buf(struct strbuf *path,
27e1e22d
JK
2334 each_loose_object_fn obj_cb,
2335 each_loose_cruft_fn cruft_cb,
2336 each_loose_subdir_fn subdir_cb,
2337 void *data)
2338{
27e1e22d
JK
2339 int r = 0;
2340 int i;
2341
27e1e22d 2342 for (i = 0; i < 256; i++) {
e6f875e0 2343 r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
27e1e22d 2344 subdir_cb, data);
27e1e22d
JK
2345 if (r)
2346 break;
2347 }
2348
e6f875e0
JK
2349 return r;
2350}
2351
2352int for_each_loose_file_in_objdir(const char *path,
2353 each_loose_object_fn obj_cb,
2354 each_loose_cruft_fn cruft_cb,
2355 each_loose_subdir_fn subdir_cb,
2356 void *data)
2357{
2358 struct strbuf buf = STRBUF_INIT;
2359 int r;
2360
2361 strbuf_addstr(&buf, path);
2362 r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb,
2363 subdir_cb, data);
27e1e22d 2364 strbuf_release(&buf);
e6f875e0 2365
27e1e22d
JK
2366 return r;
2367}
660c889e 2368
a7ff6f5a
JK
2369int for_each_loose_object(each_loose_object_fn cb, void *data,
2370 enum for_each_object_flags flags)
660c889e 2371{
f0eaf638 2372 struct object_directory *odb;
b0a42642 2373
f0eaf638
JK
2374 prepare_alt_odb(the_repository);
2375 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
2376 int r = for_each_loose_file_in_objdir(odb->path, cb, NULL,
2377 NULL, data);
2378 if (r)
2379 return r;
660c889e 2380
f0eaf638
JK
2381 if (flags & FOR_EACH_OBJECT_LOCAL_ONLY)
2382 break;
2383 }
1385bb7b 2384
f0eaf638 2385 return 0;
660c889e
JK
2386}
2387
3a2e0824
JK
2388static int append_loose_object(const struct object_id *oid, const char *path,
2389 void *data)
660c889e 2390{
3a2e0824
JK
2391 oid_array_append(data, oid);
2392 return 0;
2393}
660c889e 2394
0000d654
RS
2395struct oid_array *odb_loose_cache(struct object_directory *odb,
2396 const struct object_id *oid)
2397{
2398 int subdir_nr = oid->hash[0];
3a2e0824 2399 struct strbuf buf = STRBUF_INIT;
660c889e 2400
3a2e0824
JK
2401 if (subdir_nr < 0 ||
2402 subdir_nr >= ARRAY_SIZE(odb->loose_objects_subdir_seen))
2403 BUG("subdir_nr out of range");
1385bb7b 2404
3a2e0824 2405 if (odb->loose_objects_subdir_seen[subdir_nr])
8be88dbc 2406 return &odb->loose_objects_cache[subdir_nr];
3a2e0824
JK
2407
2408 strbuf_addstr(&buf, odb->path);
2409 for_each_file_in_obj_subdir(subdir_nr, &buf,
2410 append_loose_object,
2411 NULL, NULL,
4cea1ce0 2412 &odb->loose_objects_cache[subdir_nr]);
3a2e0824 2413 odb->loose_objects_subdir_seen[subdir_nr] = 1;
7317aa71 2414 strbuf_release(&buf);
8be88dbc 2415 return &odb->loose_objects_cache[subdir_nr];
660c889e
JK
2416}
2417
d4e19e51
RS
2418void odb_clear_loose_cache(struct object_directory *odb)
2419{
4cea1ce0
RS
2420 int i;
2421
2422 for (i = 0; i < ARRAY_SIZE(odb->loose_objects_cache); i++)
2423 oid_array_clear(&odb->loose_objects_cache[i]);
d4e19e51
RS
2424 memset(&odb->loose_objects_subdir_seen, 0,
2425 sizeof(odb->loose_objects_subdir_seen));
2426}
2427
00a7760e
JK
2428static int check_stream_oid(git_zstream *stream,
2429 const char *hdr,
2430 unsigned long size,
2431 const char *path,
2432 const struct object_id *expected_oid)
f6371f92 2433{
18e2588e 2434 git_hash_ctx c;
00a7760e 2435 struct object_id real_oid;
f6371f92
JK
2436 unsigned char buf[4096];
2437 unsigned long total_read;
2438 int status = Z_OK;
2439
18e2588e 2440 the_hash_algo->init_fn(&c);
2441 the_hash_algo->update_fn(&c, hdr, stream->total_out);
f6371f92
JK
2442
2443 /*
2444 * We already read some bytes into hdr, but the ones up to the NUL
2445 * do not count against the object's content size.
2446 */
2447 total_read = stream->total_out - strlen(hdr) - 1;
2448
2449 /*
2450 * This size comparison must be "<=" to read the final zlib packets;
00a7760e 2451 * see the comment in unpack_loose_rest for details.
f6371f92
JK
2452 */
2453 while (total_read <= size &&
18ad13e5
JH
2454 (status == Z_OK ||
2455 (status == Z_BUF_ERROR && !stream->avail_out))) {
f6371f92
JK
2456 stream->next_out = buf;
2457 stream->avail_out = sizeof(buf);
2458 if (size - total_read < stream->avail_out)
2459 stream->avail_out = size - total_read;
2460 status = git_inflate(stream, Z_FINISH);
18e2588e 2461 the_hash_algo->update_fn(&c, buf, stream->next_out - buf);
f6371f92
JK
2462 total_read += stream->next_out - buf;
2463 }
2464 git_inflate_end(stream);
2465
2466 if (status != Z_STREAM_END) {
00a7760e 2467 error(_("corrupt loose object '%s'"), oid_to_hex(expected_oid));
f6371f92
JK
2468 return -1;
2469 }
cce044df 2470 if (stream->avail_in) {
259328b7 2471 error(_("garbage at end of loose object '%s'"),
00a7760e 2472 oid_to_hex(expected_oid));
cce044df
JK
2473 return -1;
2474 }
f6371f92 2475
00a7760e
JK
2476 the_hash_algo->final_fn(real_oid.hash, &c);
2477 if (!oideq(expected_oid, &real_oid)) {
01f8d594 2478 error(_("hash mismatch for %s (expected %s)"), path,
00a7760e 2479 oid_to_hex(expected_oid));
f6371f92
JK
2480 return -1;
2481 }
2482
2483 return 0;
2484}
2485
2486int read_loose_object(const char *path,
d61d87bd 2487 const struct object_id *expected_oid,
f6371f92
JK
2488 enum object_type *type,
2489 unsigned long *size,
2490 void **contents)
2491{
2492 int ret = -1;
f6371f92
JK
2493 void *map = NULL;
2494 unsigned long mapsize;
2495 git_zstream stream;
1af64f73 2496 char hdr[MAX_HEADER_LEN];
f6371f92
JK
2497
2498 *contents = NULL;
2499
514c5fdd 2500 map = map_loose_object_1(the_repository, path, NULL, &mapsize);
f6371f92 2501 if (!map) {
259328b7 2502 error_errno(_("unable to mmap %s"), path);
f6371f92
JK
2503 goto out;
2504 }
2505
00a7760e 2506 if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) {
259328b7 2507 error(_("unable to unpack header of %s"), path);
f6371f92
JK
2508 goto out;
2509 }
2510
00a7760e 2511 *type = parse_loose_header(hdr, size);
f6371f92 2512 if (*type < 0) {
259328b7 2513 error(_("unable to parse header of %s"), path);
f6371f92
JK
2514 git_inflate_end(&stream);
2515 goto out;
2516 }
2517
7ac4f3a0 2518 if (*type == OBJ_BLOB && *size > big_file_threshold) {
00a7760e 2519 if (check_stream_oid(&stream, hdr, *size, path, expected_oid) < 0)
f6371f92
JK
2520 goto out;
2521 } else {
00a7760e 2522 *contents = unpack_loose_rest(&stream, hdr, *size, expected_oid);
f6371f92 2523 if (!*contents) {
259328b7 2524 error(_("unable to unpack contents of %s"), path);
f6371f92
JK
2525 git_inflate_end(&stream);
2526 goto out;
2527 }
b98d1885
MT
2528 if (check_object_signature(the_repository, expected_oid,
2529 *contents, *size,
2530 type_name(*type))) {
01f8d594 2531 error(_("hash mismatch for %s (expected %s)"), path,
d61d87bd 2532 oid_to_hex(expected_oid));
f6371f92
JK
2533 free(*contents);
2534 goto out;
2535 }
2536 }
2537
2538 ret = 0; /* everything checks out */
2539
2540out:
2541 if (map)
2542 munmap(map, mapsize);
f6371f92
JK
2543 return ret;
2544}