]> git.ipfire.org Git - thirdparty/git.git/blame - object-file.c
Git 2.33-rc1
[thirdparty/git.git] / object-file.c
CommitLineData
0fcfd160
LT
1/*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 *
e5afd444 6 * This handles basic git object files - packing, unpacking,
0fcfd160
LT
7 * creation etc.
8 */
0fcfd160 9#include "cache.h"
b2141fc1 10#include "config.h"
6eac50d8 11#include "string-list.h"
697cc8ef 12#include "lockfile.h"
1f688557 13#include "delta.h"
a733cb60 14#include "pack.h"
8e440259
PE
15#include "blob.h"
16#include "commit.h"
4dd1fbc7 17#include "run-command.h"
8e440259
PE
18#include "tag.h"
19#include "tree.h"
c879daa2 20#include "tree-walk.h"
f35a6d3b 21#include "refs.h"
70f5d5d3 22#include "pack-revindex.h"
bc626927 23#include "hash-lookup.h"
568508e7 24#include "bulk-checkin.h"
031dc927 25#include "repository.h"
47f351e9 26#include "replace-object.h"
090ea126 27#include "streaming.h"
543c5caa 28#include "dir.h"
12d95ef6 29#include "list.h"
c4c6effa 30#include "mergesort.h"
cf3c6352 31#include "quote.h"
4f39cd82 32#include "packfile.h"
90c62155 33#include "object-store.h"
b14ed5ad 34#include "promisor-remote.h"
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
e1ccd7e2 58static const struct object_id empty_tree_oid = {
5a6dce70 59 .hash = EMPTY_TREE_SHA1_BIN_LITERAL,
60 .algo = GIT_HASH_SHA1,
8576fde6 61};
e1ccd7e2 62static const struct object_id empty_blob_oid = {
5a6dce70 63 .hash = EMPTY_BLOB_SHA1_BIN_LITERAL,
64 .algo = GIT_HASH_SHA1,
8576fde6 65};
14228447 66static const struct object_id null_oid_sha1 = {
67 .hash = {0},
68 .algo = GIT_HASH_SHA1,
69};
13eeedb5 70static const struct object_id empty_tree_oid_sha256 = {
5a6dce70 71 .hash = EMPTY_TREE_SHA256_BIN_LITERAL,
72 .algo = GIT_HASH_SHA256,
13eeedb5 73};
74static const struct object_id empty_blob_oid_sha256 = {
5a6dce70 75 .hash = EMPTY_BLOB_SHA256_BIN_LITERAL,
76 .algo = GIT_HASH_SHA256,
13eeedb5 77};
14228447 78static const struct object_id null_oid_sha256 = {
79 .hash = {0},
80 .algo = GIT_HASH_SHA256,
81};
88cd621d 82
ac73cedf 83static void git_hash_sha1_init(git_hash_ctx *ctx)
f50e766b 84{
ac73cedf 85 git_SHA1_Init(&ctx->sha1);
f50e766b 86}
87
768e30ea 88static void git_hash_sha1_clone(git_hash_ctx *dst, const git_hash_ctx *src)
89{
90 git_SHA1_Clone(&dst->sha1, &src->sha1);
91}
92
ac73cedf 93static void git_hash_sha1_update(git_hash_ctx *ctx, const void *data, size_t len)
f50e766b 94{
ac73cedf 95 git_SHA1_Update(&ctx->sha1, data, len);
f50e766b 96}
97
ac73cedf 98static void git_hash_sha1_final(unsigned char *hash, git_hash_ctx *ctx)
f50e766b 99{
ac73cedf 100 git_SHA1_Final(hash, &ctx->sha1);
f50e766b 101}
102
ab795f0d 103static void git_hash_sha1_final_oid(struct object_id *oid, git_hash_ctx *ctx)
104{
105 git_SHA1_Final(oid->hash, &ctx->sha1);
106 memset(oid->hash + GIT_SHA1_RAWSZ, 0, GIT_MAX_RAWSZ - GIT_SHA1_RAWSZ);
5a6dce70 107 oid->algo = GIT_HASH_SHA1;
ab795f0d 108}
109
13eeedb5 110
111static void git_hash_sha256_init(git_hash_ctx *ctx)
112{
113 git_SHA256_Init(&ctx->sha256);
114}
115
768e30ea 116static void git_hash_sha256_clone(git_hash_ctx *dst, const git_hash_ctx *src)
117{
118 git_SHA256_Clone(&dst->sha256, &src->sha256);
119}
120
13eeedb5 121static void git_hash_sha256_update(git_hash_ctx *ctx, const void *data, size_t len)
122{
123 git_SHA256_Update(&ctx->sha256, data, len);
124}
125
126static void git_hash_sha256_final(unsigned char *hash, git_hash_ctx *ctx)
127{
128 git_SHA256_Final(hash, &ctx->sha256);
129}
130
ab795f0d 131static void git_hash_sha256_final_oid(struct object_id *oid, git_hash_ctx *ctx)
132{
133 git_SHA256_Final(oid->hash, &ctx->sha256);
134 /*
135 * This currently does nothing, so the compiler should optimize it out,
136 * but keep it in case we extend the hash size again.
137 */
138 memset(oid->hash + GIT_SHA256_RAWSZ, 0, GIT_MAX_RAWSZ - GIT_SHA256_RAWSZ);
5a6dce70 139 oid->algo = GIT_HASH_SHA256;
ab795f0d 140}
141
ac73cedf 142static void git_hash_unknown_init(git_hash_ctx *ctx)
f50e766b 143{
1a07e59c 144 BUG("trying to init unknown hash");
f50e766b 145}
146
768e30ea 147static void git_hash_unknown_clone(git_hash_ctx *dst, const git_hash_ctx *src)
148{
149 BUG("trying to clone unknown hash");
150}
151
ac73cedf 152static void git_hash_unknown_update(git_hash_ctx *ctx, const void *data, size_t len)
f50e766b 153{
1a07e59c 154 BUG("trying to update unknown hash");
f50e766b 155}
156
ac73cedf 157static void git_hash_unknown_final(unsigned char *hash, git_hash_ctx *ctx)
f50e766b 158{
1a07e59c 159 BUG("trying to finalize unknown hash");
f50e766b 160}
161
ab795f0d 162static void git_hash_unknown_final_oid(struct object_id *oid, git_hash_ctx *ctx)
163{
164 BUG("trying to finalize unknown hash");
165}
166
167
f50e766b 168const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = {
169 {
170 NULL,
171 0x00000000,
172 0,
173 0,
a2ce0a75 174 0,
f50e766b 175 git_hash_unknown_init,
768e30ea 176 git_hash_unknown_clone,
f50e766b 177 git_hash_unknown_update,
178 git_hash_unknown_final,
ab795f0d 179 git_hash_unknown_final_oid,
f50e766b 180 NULL,
181 NULL,
14228447 182 NULL,
f50e766b 183 },
184 {
1ccf07cb 185 "sha1",
f50e766b 186 /* "sha1", big-endian */
187 0x73686131,
f50e766b 188 GIT_SHA1_RAWSZ,
189 GIT_SHA1_HEXSZ,
a2ce0a75 190 GIT_SHA1_BLKSZ,
f50e766b 191 git_hash_sha1_init,
768e30ea 192 git_hash_sha1_clone,
f50e766b 193 git_hash_sha1_update,
194 git_hash_sha1_final,
ab795f0d 195 git_hash_sha1_final_oid,
f50e766b 196 &empty_tree_oid,
197 &empty_blob_oid,
14228447 198 &null_oid_sha1,
f50e766b 199 },
13eeedb5 200 {
201 "sha256",
202 /* "s256", big-endian */
203 0x73323536,
204 GIT_SHA256_RAWSZ,
205 GIT_SHA256_HEXSZ,
206 GIT_SHA256_BLKSZ,
207 git_hash_sha256_init,
768e30ea 208 git_hash_sha256_clone,
13eeedb5 209 git_hash_sha256_update,
210 git_hash_sha256_final,
ab795f0d 211 git_hash_sha256_final_oid,
13eeedb5 212 &empty_tree_oid_sha256,
213 &empty_blob_oid_sha256,
14228447 214 &null_oid_sha256,
13eeedb5 215 }
f50e766b 216};
217
14228447 218const struct object_id *null_oid(void)
219{
220 return the_hash_algo->null_oid;
221}
222
d8a92ced 223const char *empty_tree_oid_hex(void)
224{
225 static char buf[GIT_MAX_HEXSZ + 1];
226 return oid_to_hex_r(buf, the_hash_algo->empty_tree);
227}
228
229const char *empty_blob_oid_hex(void)
230{
231 static char buf[GIT_MAX_HEXSZ + 1];
232 return oid_to_hex_r(buf, the_hash_algo->empty_blob);
233}
234
2f90b9d9 235int hash_algo_by_name(const char *name)
236{
237 int i;
238 if (!name)
239 return GIT_HASH_UNKNOWN;
240 for (i = 1; i < GIT_HASH_NALGOS; i++)
241 if (!strcmp(name, hash_algos[i].name))
242 return i;
243 return GIT_HASH_UNKNOWN;
244}
245
246int hash_algo_by_id(uint32_t format_id)
247{
248 int i;
249 for (i = 1; i < GIT_HASH_NALGOS; i++)
250 if (format_id == hash_algos[i].format_id)
251 return i;
252 return GIT_HASH_UNKNOWN;
253}
254
95399788 255int hash_algo_by_length(int len)
256{
257 int i;
258 for (i = 1; i < GIT_HASH_NALGOS; i++)
259 if (len == hash_algos[i].rawsz)
260 return i;
261 return GIT_HASH_UNKNOWN;
262}
2f90b9d9 263
c597ba80
NTND
264/*
265 * This is meant to hold a *small* number of objects that you would
cb1c8d1d 266 * want read_object_file() to be able to return, but yet you do not want
c597ba80
NTND
267 * to write them into the object store (e.g. a browse-only
268 * application).
269 */
270static struct cached_object {
62ba93ea 271 struct object_id oid;
c597ba80
NTND
272 enum object_type type;
273 void *buf;
274 unsigned long size;
275} *cached_objects;
276static int cached_object_nr, cached_object_alloc;
277
278static struct cached_object empty_tree = {
62ba93ea 279 { EMPTY_TREE_SHA1_BIN_LITERAL },
c597ba80
NTND
280 OBJ_TREE,
281 "",
282 0
283};
284
62ba93ea 285static struct cached_object *find_cached_object(const struct object_id *oid)
c597ba80
NTND
286{
287 int i;
288 struct cached_object *co = cached_objects;
289
290 for (i = 0; i < cached_object_nr; i++, co++) {
4a7e27e9 291 if (oideq(&co->oid, oid))
c597ba80
NTND
292 return co;
293 }
4a7e27e9 294 if (oideq(oid, the_hash_algo->empty_tree))
c597ba80
NTND
295 return &empty_tree;
296 return NULL;
297}
298
9472935d 299
8462ff43 300static int get_conv_flags(unsigned flags)
9472935d
TB
301{
302 if (flags & HASH_RENORMALIZE)
8462ff43 303 return CONV_EOL_RENORMALIZE;
9472935d 304 else if (flags & HASH_WRITE_OBJECT)
107642fe 305 return global_conv_flags_eol | CONV_WRITE_OBJECT;
9472935d 306 else
8462ff43 307 return 0;
9472935d
TB
308}
309
310
90a6464b
JH
311int mkdir_in_gitdir(const char *path)
312{
313 if (mkdir(path, 0777)) {
314 int saved_errno = errno;
315 struct stat st;
316 struct strbuf sb = STRBUF_INIT;
317
318 if (errno != EEXIST)
319 return -1;
320 /*
321 * Are we looking at a path in a symlinked worktree
322 * whose original repository does not yet have it?
323 * e.g. .git/rr-cache pointing at its original
324 * repository in which the user hasn't performed any
325 * conflict resolution yet?
326 */
327 if (lstat(path, &st) || !S_ISLNK(st.st_mode) ||
328 strbuf_readlink(&sb, path, st.st_size) ||
329 !is_absolute_path(sb.buf) ||
330 mkdir(sb.buf, 0777)) {
331 strbuf_release(&sb);
332 errno = saved_errno;
333 return -1;
334 }
335 strbuf_release(&sb);
336 }
337 return adjust_shared_perm(path);
338}
339
eb3c027e 340static enum scld_error safe_create_leading_directories_1(char *path, int share)
b2cb9425 341{
26c8ae2a 342 char *next_component = path + offset_1st_component(path);
0be0521b 343 enum scld_error ret = SCLD_OK;
67d42212 344
0be0521b 345 while (ret == SCLD_OK && next_component) {
f0502332 346 struct stat st;
0f527403 347 char *slash = next_component, slash_character;
f0502332 348
0f527403
MH
349 while (*slash && !is_dir_sep(*slash))
350 slash++;
351
352 if (!*slash)
b2cb9425 353 break;
bf10cf70 354
26c8ae2a 355 next_component = slash + 1;
0f527403 356 while (is_dir_sep(*next_component))
bf10cf70 357 next_component++;
26c8ae2a 358 if (!*next_component)
5f0bdf50 359 break;
831651fd 360
0f527403 361 slash_character = *slash;
831651fd 362 *slash = '\0';
67d42212
JR
363 if (!stat(path, &st)) {
364 /* path exists */
204a047f
MH
365 if (!S_ISDIR(st.st_mode)) {
366 errno = ENOTDIR;
0be0521b 367 ret = SCLD_EXISTS;
204a047f 368 }
53a39721 369 } else if (mkdir(path, 0777)) {
928734d9 370 if (errno == EEXIST &&
9e6f885d 371 !stat(path, &st) && S_ISDIR(st.st_mode))
928734d9 372 ; /* somebody created it since we checked */
18d37e86
MH
373 else if (errno == ENOENT)
374 /*
375 * Either mkdir() failed because
376 * somebody just pruned the containing
377 * directory, or stat() failed because
378 * the file that was in our way was
379 * just removed. Either way, inform
380 * the caller that it might be worth
381 * trying again:
382 */
383 ret = SCLD_VANISHED;
9e6f885d 384 else
0be0521b 385 ret = SCLD_FAILED;
eb3c027e 386 } else if (share && adjust_shared_perm(path)) {
0be0521b 387 ret = SCLD_PERMS;
457f06d6 388 }
0f527403 389 *slash = slash_character;
b2cb9425 390 }
9e6f885d 391 return ret;
b2cb9425 392}
723c31fe 393
eb3c027e
MT
394enum scld_error safe_create_leading_directories(char *path)
395{
396 return safe_create_leading_directories_1(path, 1);
397}
398
399enum scld_error safe_create_leading_directories_no_share(char *path)
400{
401 return safe_create_leading_directories_1(path, 0);
402}
403
0be0521b 404enum scld_error safe_create_leading_directories_const(const char *path)
8e21d63b 405{
02944307 406 int save_errno;
8e21d63b
JK
407 /* path points to cache entries, so xstrdup before messing with it */
408 char *buf = xstrdup(path);
0be0521b 409 enum scld_error result = safe_create_leading_directories(buf);
02944307
MH
410
411 save_errno = errno;
8e21d63b 412 free(buf);
02944307 413 errno = save_errno;
8e21d63b
JK
414 return result;
415}
416
177978f5
MH
417int raceproof_create_file(const char *path, create_file_fn fn, void *cb)
418{
419 /*
420 * The number of times we will try to remove empty directories
421 * in the way of path. This is only 1 because if another
422 * process is racily creating directories that conflict with
423 * us, we don't want to fight against them.
424 */
425 int remove_directories_remaining = 1;
426
427 /*
428 * The number of times that we will try to create the
429 * directories containing path. We are willing to attempt this
430 * more than once, because another process could be trying to
431 * clean up empty directories at the same time as we are
432 * trying to create them.
433 */
434 int create_directories_remaining = 3;
435
436 /* A scratch copy of path, filled lazily if we need it: */
437 struct strbuf path_copy = STRBUF_INIT;
438
439 int ret, save_errno;
440
441 /* Sanity check: */
442 assert(*path);
443
444retry_fn:
445 ret = fn(path, cb);
446 save_errno = errno;
447 if (!ret)
448 goto out;
449
450 if (errno == EISDIR && remove_directories_remaining-- > 0) {
451 /*
452 * A directory is in the way. Maybe it is empty; try
453 * to remove it:
454 */
455 if (!path_copy.len)
456 strbuf_addstr(&path_copy, path);
457
458 if (!remove_dir_recursively(&path_copy, REMOVE_DIR_EMPTY_ONLY))
459 goto retry_fn;
460 } else if (errno == ENOENT && create_directories_remaining-- > 0) {
461 /*
462 * Maybe the containing directory didn't exist, or
463 * maybe it was just deleted by a process that is
464 * racing with us to clean up empty directories. Try
465 * to create it:
466 */
467 enum scld_error scld_result;
468
469 if (!path_copy.len)
470 strbuf_addstr(&path_copy, path);
471
472 do {
473 scld_result = safe_create_leading_directories(path_copy.buf);
474 if (scld_result == SCLD_OK)
475 goto retry_fn;
476 } while (scld_result == SCLD_VANISHED && create_directories_remaining-- > 0);
477 }
478
479out:
480 strbuf_release(&path_copy);
481 errno = save_errno;
482 return ret;
483}
484
514c5fdd 485static void fill_loose_path(struct strbuf *buf, const struct object_id *oid)
ace1534d
JH
486{
487 int i;
94b5e093 488 for (i = 0; i < the_hash_algo->rawsz; i++) {
ace1534d 489 static char hex[] = "0123456789abcdef";
514c5fdd 490 unsigned int val = oid->hash[i];
f7b7774f
JK
491 strbuf_addch(buf, hex[val >> 4]);
492 strbuf_addch(buf, hex[val & 0xf]);
afbba2f0 493 if (!i)
f7b7774f 494 strbuf_addch(buf, '/');
ace1534d
JH
495 }
496}
497
f0eaf638
JK
498static const char *odb_loose_path(struct object_directory *odb,
499 struct strbuf *buf,
514c5fdd 500 const struct object_id *oid)
0fcfd160 501{
b69fb867 502 strbuf_reset(buf);
f0eaf638 503 strbuf_addstr(buf, odb->path);
34498471 504 strbuf_addch(buf, '/');
514c5fdd 505 fill_loose_path(buf, oid);
f3f043a1 506 return buf->buf;
0fcfd160
LT
507}
508
f3f043a1 509const char *loose_object_path(struct repository *r, struct strbuf *buf,
514c5fdd 510 const struct object_id *oid)
29ec6af2 511{
514c5fdd 512 return odb_loose_path(r->objects->odb, buf, oid);
0fcfd160
LT
513}
514
4ea82473
JK
515/*
516 * Return non-zero iff the path is usable as an alternate object database.
517 */
13313fc3
SB
518static int alt_odb_usable(struct raw_object_store *o,
519 struct strbuf *path,
cf2dc1c2 520 const char *normalized_objdir, khiter_t *pos)
4ea82473 521{
cf2dc1c2 522 int r;
4ea82473
JK
523
524 /* Detect cases where alternate disappeared */
525 if (!is_directory(path->buf)) {
259328b7
NTND
526 error(_("object directory %s does not exist; "
527 "check .git/objects/info/alternates"),
4ea82473
JK
528 path->buf);
529 return 0;
530 }
531
532 /*
533 * Prevent the common mistake of listing the same
534 * thing twice, or object directory itself.
535 */
cf2dc1c2
EW
536 if (!o->odb_by_path) {
537 khiter_t p;
538
539 o->odb_by_path = kh_init_odb_path_map();
540 assert(!o->odb->next);
541 p = kh_put_odb_path_map(o->odb_by_path, o->odb->path, &r);
542 assert(r == 1); /* never used */
543 kh_value(o->odb_by_path, p) = o->odb;
4ea82473 544 }
cf2dc1c2 545 if (fspatheq(path->buf, normalized_objdir))
4ea82473 546 return 0;
cf2dc1c2
EW
547 *pos = kh_put_odb_path_map(o->odb_by_path, path->buf, &r);
548 /* r: 0 = exists, 1 = never used, 2 = deleted */
549 return r == 0 ? 0 : 1;
4ea82473
JK
550}
551
ddd5d056
JH
552/*
553 * Prepare alternate object database registry.
d5a63b99
JH
554 *
555 * The variable alt_odb_list points at the list of struct
263db403 556 * object_directory. The elements on this list come from
d5a63b99
JH
557 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
558 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
559 * whose contents is similar to that environment variable but can be
560 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
561 * contains "/the/directory/corresponding/to/.git/objects/...", while
562 * its name points just after the slash at the end of ".git/objects/"
e5afd444
563 * in the example above, and has enough space to hold all hex characters
564 * of the object ID, an extra slash for the first level indirection, and
565 * the terminating NUL.
ddd5d056 566 */
77f012e8
SB
567static void read_info_alternates(struct repository *r,
568 const char *relative_base,
569 int depth);
407532f8 570static int link_alt_odb_entry(struct repository *r, const struct strbuf *entry,
cfc62fc9 571 const char *relative_base, int depth, const char *normalized_objdir)
ace1534d 572{
263db403 573 struct object_directory *ent;
5bdf0a84 574 struct strbuf pathbuf = STRBUF_INIT;
cf2dc1c2 575 khiter_t pos;
d5a63b99 576
407532f8 577 if (!is_absolute_path(entry->buf) && relative_base) {
4ac9006f 578 strbuf_realpath(&pathbuf, relative_base, 1);
5bdf0a84 579 strbuf_addch(&pathbuf, '/');
c2f493a4 580 }
407532f8 581 strbuf_addbuf(&pathbuf, entry);
c2f493a4 582
37a95862 583 if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) {
259328b7 584 error(_("unable to normalize alternate object path: %s"),
670c359d
JK
585 pathbuf.buf);
586 strbuf_release(&pathbuf);
587 return -1;
588 }
5bdf0a84
HW
589
590 /*
591 * The trailing slash after the directory name is given by
592 * this function at the end. Remove duplicates.
593 */
4ea82473
JK
594 while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/')
595 strbuf_setlen(&pathbuf, pathbuf.len - 1);
5bdf0a84 596
cf2dc1c2 597 if (!alt_odb_usable(r->objects, &pathbuf, normalized_objdir, &pos)) {
4ea82473 598 strbuf_release(&pathbuf);
c2f493a4
MW
599 return -1;
600 }
601
ca56dadb 602 CALLOC_ARRAY(ent, 1);
cf2dc1c2
EW
603 /* pathbuf.buf is already in r->objects->odb_by_path */
604 ent->path = strbuf_detach(&pathbuf, NULL);
c2f493a4
MW
605
606 /* add the alternate entry */
f0eaf638
JK
607 *r->objects->odb_tail = ent;
608 r->objects->odb_tail = &(ent->next);
c2f493a4 609 ent->next = NULL;
cf2dc1c2
EW
610 assert(r->objects->odb_by_path);
611 kh_value(r->objects->odb_by_path, pos) = ent;
c2f493a4
MW
612
613 /* recursively add alternates */
cf2dc1c2 614 read_info_alternates(r, ent->path, depth + 1);
c2f493a4
MW
615
616 return 0;
617}
618
cf3c6352
JK
619static const char *parse_alt_odb_entry(const char *string,
620 int sep,
621 struct strbuf *out)
622{
623 const char *end;
624
625 strbuf_reset(out);
626
627 if (*string == '#') {
628 /* comment; consume up to next separator */
629 end = strchrnul(string, sep);
630 } else if (*string == '"' && !unquote_c_style(out, string, &end)) {
631 /*
632 * quoted path; unquote_c_style has copied the
633 * data for us and set "end". Broken quoting (e.g.,
634 * an entry that doesn't end with a quote) falls
635 * back to the unquoted case below.
636 */
637 } else {
638 /* normal, unquoted path */
639 end = strchrnul(string, sep);
640 strbuf_add(out, string, end - string);
641 }
642
643 if (*end)
644 end++;
645 return end;
646}
647
77f012e8
SB
648static void link_alt_odb_entries(struct repository *r, const char *alt,
649 int sep, const char *relative_base, int depth)
c2f493a4 650{
539e7506 651 struct strbuf objdirbuf = STRBUF_INIT;
cf3c6352 652 struct strbuf entry = STRBUF_INIT;
c2f493a4 653
f28e3668
JK
654 if (!alt || !*alt)
655 return;
656
c2f493a4 657 if (depth > 5) {
259328b7 658 error(_("%s: ignoring alternate object stores, nesting too deep"),
c2f493a4
MW
659 relative_base);
660 return;
661 }
662
f0eaf638 663 strbuf_add_absolute_path(&objdirbuf, r->objects->odb->path);
670c359d 664 if (strbuf_normalize_path(&objdirbuf) < 0)
259328b7 665 die(_("unable to normalize object directory: %s"),
670c359d 666 objdirbuf.buf);
539e7506 667
cf3c6352
JK
668 while (*alt) {
669 alt = parse_alt_odb_entry(alt, sep, &entry);
670 if (!entry.len)
9577e7e3 671 continue;
407532f8 672 link_alt_odb_entry(r, &entry,
cfc62fc9 673 relative_base, depth, objdirbuf.buf);
9577e7e3 674 }
cf3c6352 675 strbuf_release(&entry);
539e7506 676 strbuf_release(&objdirbuf);
d5a63b99
JH
677}
678
77f012e8
SB
679static void read_info_alternates(struct repository *r,
680 const char *relative_base,
681 int depth)
d5a63b99 682{
5015f01c 683 char *path;
dc732bd5 684 struct strbuf buf = STRBUF_INIT;
d5a63b99 685
5015f01c 686 path = xstrfmt("%s/info/alternates", relative_base);
dc732bd5 687 if (strbuf_read_file(&buf, path, 1024) < 0) {
f0f7bebe 688 warn_on_fopen_errors(path);
dc732bd5 689 free(path);
9a217f2a 690 return;
ace1534d 691 }
d5a63b99 692
77f012e8 693 link_alt_odb_entries(r, buf.buf, '\n', relative_base, depth);
dc732bd5
JK
694 strbuf_release(&buf);
695 free(path);
ace1534d
JH
696}
697
bef70b22
DB
698void add_to_alternates_file(const char *reference)
699{
f132a127 700 struct lock_file lock = LOCK_INIT;
77b9b1d1
JK
701 char *alts = git_pathdup("objects/info/alternates");
702 FILE *in, *out;
f132a127 703 int found = 0;
77b9b1d1 704
f132a127
705 hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR);
706 out = fdopen_lock_file(&lock, "w");
77b9b1d1 707 if (!out)
259328b7 708 die_errno(_("unable to fdopen alternates lockfile"));
77b9b1d1
JK
709
710 in = fopen(alts, "r");
711 if (in) {
712 struct strbuf line = STRBUF_INIT;
77b9b1d1 713
3f163962 714 while (strbuf_getline(&line, in) != EOF) {
77b9b1d1
JK
715 if (!strcmp(reference, line.buf)) {
716 found = 1;
717 break;
718 }
719 fprintf_or_die(out, "%s\n", line.buf);
720 }
721
722 strbuf_release(&line);
723 fclose(in);
77b9b1d1
JK
724 }
725 else if (errno != ENOENT)
259328b7 726 die_errno(_("unable to read alternates file"));
77b9b1d1 727
f132a127
728 if (found) {
729 rollback_lock_file(&lock);
730 } else {
77b9b1d1 731 fprintf_or_die(out, "%s\n", reference);
f132a127 732 if (commit_lock_file(&lock))
259328b7 733 die_errno(_("unable to move new alternates file into place"));
f0eaf638 734 if (the_repository->objects->loaded_alternates)
93d8d1e2
SB
735 link_alt_odb_entries(the_repository, reference,
736 '\n', NULL, 0);
77b9b1d1
JK
737 }
738 free(alts);
bef70b22
DB
739}
740
a5b34d21
JK
741void add_to_alternates_memory(const char *reference)
742{
743 /*
744 * Make sure alternates are initialized, or else our entry may be
745 * overwritten when they are.
746 */
0b209034 747 prepare_alt_odb(the_repository);
a5b34d21 748
93d8d1e2
SB
749 link_alt_odb_entries(the_repository, reference,
750 '\n', NULL, 0);
a5b34d21
JK
751}
752
9eeea7d2
SB
753/*
754 * Compute the exact path an alternate is at and returns it. In case of
755 * error NULL is returned and the human readable error is added to `err`
efde7b72 756 * `path` may be relative and should point to $GIT_DIR.
9eeea7d2
SB
757 * `err` must not be null.
758 */
759char *compute_alternate_path(const char *path, struct strbuf *err)
760{
761 char *ref_git = NULL;
4530a85b 762 const char *repo;
9eeea7d2
SB
763 int seen_error = 0;
764
4530a85b
AM
765 ref_git = real_pathdup(path, 0);
766 if (!ref_git) {
9eeea7d2
SB
767 seen_error = 1;
768 strbuf_addf(err, _("path '%s' does not exist"), path);
769 goto out;
4530a85b 770 }
9eeea7d2
SB
771
772 repo = read_gitfile(ref_git);
773 if (!repo)
774 repo = read_gitfile(mkpath("%s/.git", ref_git));
775 if (repo) {
776 free(ref_git);
777 ref_git = xstrdup(repo);
778 }
779
780 if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) {
781 char *ref_git_git = mkpathdup("%s/.git", ref_git);
782 free(ref_git);
783 ref_git = ref_git_git;
784 } else if (!is_directory(mkpath("%s/objects", ref_git))) {
785 struct strbuf sb = STRBUF_INIT;
786 seen_error = 1;
787 if (get_common_dir(&sb, ref_git)) {
788 strbuf_addf(err,
789 _("reference repository '%s' as a linked "
790 "checkout is not supported yet."),
791 path);
792 goto out;
793 }
794
795 strbuf_addf(err, _("reference repository '%s' is not a "
796 "local repository."), path);
797 goto out;
798 }
799
800 if (!access(mkpath("%s/shallow", ref_git), F_OK)) {
801 strbuf_addf(err, _("reference repository '%s' is shallow"),
802 path);
803 seen_error = 1;
804 goto out;
805 }
806
807 if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) {
808 strbuf_addf(err,
809 _("reference repository '%s' is grafted"),
810 path);
811 seen_error = 1;
812 goto out;
813 }
814
815out:
816 if (seen_error) {
6a83d902 817 FREE_AND_NULL(ref_git);
9eeea7d2
SB
818 }
819
820 return ref_git;
821}
822
709dfa69
JK
823static void fill_alternate_refs_command(struct child_process *cmd,
824 const char *repo_path)
825{
826 const char *value;
827
828 if (!git_config_get_value("core.alternateRefsCommand", &value)) {
829 cmd->use_shell = 1;
830
c972bf4c
JK
831 strvec_push(&cmd->args, value);
832 strvec_push(&cmd->args, repo_path);
709dfa69
JK
833 } else {
834 cmd->git_cmd = 1;
835
c972bf4c
JK
836 strvec_pushf(&cmd->args, "--git-dir=%s", repo_path);
837 strvec_push(&cmd->args, "for-each-ref");
838 strvec_push(&cmd->args, "--format=%(objectname)");
709dfa69
JK
839
840 if (!git_config_get_value("core.alternateRefsPrefixes", &value)) {
c972bf4c
JK
841 strvec_push(&cmd->args, "--");
842 strvec_split(&cmd->args, value);
709dfa69
JK
843 }
844 }
845
846 cmd->env = local_repo_env;
847 cmd->out = -1;
848}
849
850static void read_alternate_refs(const char *path,
851 alternate_ref_fn *cb,
852 void *data)
853{
854 struct child_process cmd = CHILD_PROCESS_INIT;
855 struct strbuf line = STRBUF_INIT;
856 FILE *fh;
857
858 fill_alternate_refs_command(&cmd, path);
859
860 if (start_command(&cmd))
861 return;
862
863 fh = xfdopen(cmd.out, "r");
864 while (strbuf_getline_lf(&line, fh) != EOF) {
865 struct object_id oid;
866 const char *p;
867
868 if (parse_oid_hex(line.buf, &oid, &p) || *p) {
869 warning(_("invalid line while parsing alternate refs: %s"),
870 line.buf);
871 break;
872 }
873
874 cb(&oid, data);
875 }
876
877 fclose(fh);
878 finish_command(&cmd);
86ad3ea5 879 strbuf_release(&line);
709dfa69
JK
880}
881
882struct alternate_refs_data {
883 alternate_ref_fn *fn;
884 void *data;
885};
886
887static int refs_from_alternate_cb(struct object_directory *e,
888 void *data)
889{
890 struct strbuf path = STRBUF_INIT;
891 size_t base_len;
892 struct alternate_refs_data *cb = data;
893
894 if (!strbuf_realpath(&path, e->path, 0))
895 goto out;
896 if (!strbuf_strip_suffix(&path, "/objects"))
897 goto out;
898 base_len = path.len;
899
900 /* Is this a git repository with refs? */
901 strbuf_addstr(&path, "/refs");
902 if (!is_directory(path.buf))
903 goto out;
904 strbuf_setlen(&path, base_len);
905
906 read_alternate_refs(path.buf, cb->fn, cb->data);
907
908out:
909 strbuf_release(&path);
910 return 0;
911}
912
913void for_each_alternate_ref(alternate_ref_fn fn, void *data)
914{
915 struct alternate_refs_data cb;
916 cb.fn = fn;
917 cb.data = data;
918 foreach_alt_odb(refs_from_alternate_cb, &cb);
919}
920
fe1b2268 921int foreach_alt_odb(alt_odb_fn fn, void *cb)
d79796bc 922{
263db403 923 struct object_directory *ent;
fe1b2268 924 int r = 0;
d79796bc 925
0b209034 926 prepare_alt_odb(the_repository);
f0eaf638 927 for (ent = the_repository->objects->odb->next; ent; ent = ent->next) {
fe1b2268
JK
928 r = fn(ent, cb);
929 if (r)
930 break;
931 }
932 return r;
d79796bc
JH
933}
934
13068bf0 935void prepare_alt_odb(struct repository *r)
c2f493a4 936{
f0eaf638 937 if (r->objects->loaded_alternates)
7dc24aa5
SP
938 return;
939
13068bf0 940 link_alt_odb_entries(r, r->objects->alternate_db, PATH_SEP, NULL, 0);
c2f493a4 941
f0eaf638
JK
942 read_info_alternates(r, r->objects->odb->path, 0);
943 r->objects->loaded_alternates = 1;
c2f493a4
MW
944}
945
3096b2ec 946/* Returns 1 if we have successfully freshened the file, 0 otherwise. */
33d4221c 947static int freshen_file(const char *fn)
ace1534d 948{
312cd761 949 return !utime(fn, NULL);
0f4dc14a 950}
ace1534d 951
3096b2ec
JK
952/*
953 * All of the check_and_freshen functions return 1 if the file exists and was
954 * freshened (if freshening was requested), 0 otherwise. If they return
955 * 0, you should not assume that it is safe to skip a write of the object (it
956 * either does not exist on disk, or has a stale mtime and may be subject to
957 * pruning).
958 */
6a5e6f5e 959int check_and_freshen_file(const char *fn, int freshen)
33d4221c
JK
960{
961 if (access(fn, F_OK))
962 return 0;
3096b2ec 963 if (freshen && !freshen_file(fn))
33d4221c
JK
964 return 0;
965 return 1;
966}
967
f0eaf638
JK
968static int check_and_freshen_odb(struct object_directory *odb,
969 const struct object_id *oid,
970 int freshen)
33d4221c 971{
f0eaf638 972 static struct strbuf path = STRBUF_INIT;
514c5fdd 973 odb_loose_path(odb, &path, oid);
f0eaf638
JK
974 return check_and_freshen_file(path.buf, freshen);
975}
ea657730 976
f0eaf638
JK
977static int check_and_freshen_local(const struct object_id *oid, int freshen)
978{
979 return check_and_freshen_odb(the_repository->objects->odb, oid, freshen);
33d4221c
JK
980}
981
6862ebbf 982static int check_and_freshen_nonlocal(const struct object_id *oid, int freshen)
0f4dc14a 983{
263db403 984 struct object_directory *odb;
f3f043a1 985
0b209034 986 prepare_alt_odb(the_repository);
f0eaf638
JK
987 for (odb = the_repository->objects->odb->next; odb; odb = odb->next) {
988 if (check_and_freshen_odb(odb, oid, freshen))
c529d75a 989 return 1;
ace1534d 990 }
c529d75a 991 return 0;
ace1534d
JH
992}
993
6862ebbf 994static int check_and_freshen(const struct object_id *oid, int freshen)
33d4221c 995{
6862ebbf 996 return check_and_freshen_local(oid, freshen) ||
997 check_and_freshen_nonlocal(oid, freshen);
33d4221c
JK
998}
999
6862ebbf 1000int has_loose_object_nonlocal(const struct object_id *oid)
33d4221c 1001{
6862ebbf 1002 return check_and_freshen_nonlocal(oid, 0);
33d4221c
JK
1003}
1004
6862ebbf 1005static int has_loose_object(const struct object_id *oid)
0f4dc14a 1006{
6862ebbf 1007 return check_and_freshen(oid, 0);
0f4dc14a
BC
1008}
1009
02710228
SP
1010static void mmap_limit_check(size_t length)
1011{
1012 static size_t limit = 0;
1013 if (!limit) {
1014 limit = git_env_ulong("GIT_MMAP_LIMIT", 0);
1015 if (!limit)
1016 limit = SIZE_MAX;
1017 }
1018 if (length > limit)
259328b7 1019 die(_("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX),
02710228
SP
1020 (uintmax_t)length, (uintmax_t)limit);
1021}
1022
1570856b
JK
1023void *xmmap_gently(void *start, size_t length,
1024 int prot, int flags, int fd, off_t offset)
58ecbd5e 1025{
02710228
SP
1026 void *ret;
1027
1028 mmap_limit_check(length);
1029 ret = mmap(start, length, prot, flags, fd, offset);
9827d4c1
JK
1030 if (ret == MAP_FAILED && !length)
1031 ret = NULL;
58ecbd5e
JN
1032 return ret;
1033}
1034
dc059294
EW
1035const char *mmap_os_err(void)
1036{
1037 static const char blank[] = "";
1038#if defined(__linux__)
1039 if (errno == ENOMEM) {
1040 /* this continues an existing error message: */
1041 static const char enomem[] =
1042", check sys.vm.max_map_count and/or RLIMIT_DATA";
1043 return enomem;
1044 }
1045#endif /* OS-specific bits */
1046 return blank;
1047}
1048
1570856b
JK
1049void *xmmap(void *start, size_t length,
1050 int prot, int flags, int fd, off_t offset)
1051{
1052 void *ret = xmmap_gently(start, length, prot, flags, fd, offset);
1053 if (ret == MAP_FAILED)
dc059294 1054 die_errno(_("mmap failed%s"), mmap_os_err());
1570856b
JK
1055 return ret;
1056}
1057
88d0db55 1058/*
090ea126 1059 * With an in-core object data in "map", rehash it to make sure the
cb1c8d1d
JK
1060 * object name actually matches "oid" to detect object corruption.
1061 * With "map" == NULL, try reading the object named with "oid" using
090ea126 1062 * the streaming interface and rehash it to do the same.
88d0db55 1063 */
b98d1885
MT
1064int check_object_signature(struct repository *r, const struct object_id *oid,
1065 void *map, unsigned long size, const char *type)
88d0db55 1066{
f070facc 1067 struct object_id real_oid;
090ea126
NTND
1068 enum object_type obj_type;
1069 struct git_istream *st;
18e2588e 1070 git_hash_ctx c;
1af64f73 1071 char hdr[MAX_HEADER_LEN];
090ea126 1072 int hdrlen;
88d0db55 1073
090ea126 1074 if (map) {
b98d1885 1075 hash_object_file(r->hash_algo, map, size, type, &real_oid);
9001dc2a 1076 return !oideq(oid, &real_oid) ? -1 : 0;
88d0db55
BC
1077 }
1078
b98d1885 1079 st = open_istream(r, oid, &obj_type, &size, NULL);
090ea126
NTND
1080 if (!st)
1081 return -1;
88d0db55 1082
090ea126 1083 /* Generate the header */
ca473cef 1084 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(obj_type), (uintmax_t)size) + 1;
88d0db55 1085
090ea126 1086 /* Sha1.. */
b98d1885
MT
1087 r->hash_algo->init_fn(&c);
1088 r->hash_algo->update_fn(&c, hdr, hdrlen);
090ea126
NTND
1089 for (;;) {
1090 char buf[1024 * 16];
1091 ssize_t readlen = read_istream(st, buf, sizeof(buf));
1f688557 1092
f54fac53
JK
1093 if (readlen < 0) {
1094 close_istream(st);
1095 return -1;
1096 }
090ea126
NTND
1097 if (!readlen)
1098 break;
b98d1885 1099 r->hash_algo->update_fn(&c, buf, readlen);
fa5fc15d 1100 }
5951bf46 1101 r->hash_algo->final_oid_fn(&real_oid, &c);
090ea126 1102 close_istream(st);
9001dc2a 1103 return !oideq(oid, &real_oid) ? -1 : 0;
fa5fc15d
SP
1104}
1105
1b8ac5ea 1106int git_open_cloexec(const char *name, int flags)
a0788266 1107{
1e3001a8
JH
1108 int fd;
1109 static int o_cloexec = O_CLOEXEC;
a0788266 1110
1e3001a8
JH
1111 fd = open(name, flags | o_cloexec);
1112 if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) {
cd66ada0 1113 /* Try again w/o O_CLOEXEC: the kernel might not support it */
1e3001a8
JH
1114 o_cloexec &= ~O_CLOEXEC;
1115 fd = open(name, flags | o_cloexec);
491a8dec 1116 }
491a8dec 1117
9fb9495d 1118#if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
491a8dec 1119 {
1e3001a8 1120 static int fd_cloexec = FD_CLOEXEC;
a0788266 1121
1e3001a8
JH
1122 if (!o_cloexec && 0 <= fd && fd_cloexec) {
1123 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
9fb9495d
EW
1124 int flags = fcntl(fd, F_GETFD);
1125 if (fcntl(fd, F_SETFD, flags | fd_cloexec))
1e3001a8 1126 fd_cloexec = 0;
cd66ada0 1127 }
44d1c19e 1128 }
a0788266 1129#endif
1b8ac5ea 1130 return fd;
a0788266
JS
1131}
1132
3cf8b462 1133/*
514c5fdd 1134 * Find "oid" as a loose object in the local repository or in an alternate.
771e7d57
JK
1135 * Returns 0 on success, negative on failure.
1136 *
1137 * The "path" out-parameter will give the path of the object we found (if any).
1138 * Note that it may point to static storage and is only valid until another
514c5fdd 1139 * call to stat_loose_object().
3cf8b462 1140 */
514c5fdd
JK
1141static int stat_loose_object(struct repository *r, const struct object_id *oid,
1142 struct stat *st, const char **path)
1f688557 1143{
263db403 1144 struct object_directory *odb;
ea657730
CC
1145 static struct strbuf buf = STRBUF_INIT;
1146
d2607fa0 1147 prepare_alt_odb(r);
f0eaf638 1148 for (odb = r->objects->odb; odb; odb = odb->next) {
514c5fdd 1149 *path = odb_loose_path(odb, &buf, oid);
771e7d57 1150 if (!lstat(*path, st))
052fe5ea 1151 return 0;
c7934306
SP
1152 }
1153
3cf8b462
SP
1154 return -1;
1155}
1156
771e7d57 1157/*
514c5fdd 1158 * Like stat_loose_object(), but actually open the object and return the
771e7d57
JK
1159 * descriptor. See the caveats on the "path" parameter above.
1160 */
514c5fdd
JK
1161static int open_loose_object(struct repository *r,
1162 const struct object_id *oid, const char **path)
60bb8b14 1163{
44d1c19e 1164 int fd;
263db403 1165 struct object_directory *odb;
f0eaf638 1166 int most_interesting_errno = ENOENT;
ea657730
CC
1167 static struct strbuf buf = STRBUF_INIT;
1168
ec7283e5 1169 prepare_alt_odb(r);
f0eaf638 1170 for (odb = r->objects->odb; odb; odb = odb->next) {
514c5fdd 1171 *path = odb_loose_path(odb, &buf, oid);
771e7d57 1172 fd = git_open(*path);
44d1c19e
LT
1173 if (fd >= 0)
1174 return fd;
f0eaf638 1175
d6c8a05b
JK
1176 if (most_interesting_errno == ENOENT)
1177 most_interesting_errno = errno;
03e79c88 1178 }
d6c8a05b 1179 errno = most_interesting_errno;
44d1c19e 1180 return -1;
1f688557
JH
1181}
1182
61c7711c 1183static int quick_has_loose(struct repository *r,
d7a24573 1184 const struct object_id *oid)
61c7711c 1185{
61c7711c
JK
1186 struct object_directory *odb;
1187
61c7711c
JK
1188 prepare_alt_odb(r);
1189 for (odb = r->objects->odb; odb; odb = odb->next) {
92d8ed8a 1190 if (oidtree_contains(odb_loose_cache(odb, oid), oid))
61c7711c
JK
1191 return 1;
1192 }
1193 return 0;
1194}
1195
f6371f92
JK
1196/*
1197 * Map the loose object at "path" if it is not NULL, or the path found by
514c5fdd 1198 * searching for a loose object named "oid".
f6371f92 1199 */
514c5fdd
JK
1200static void *map_loose_object_1(struct repository *r, const char *path,
1201 const struct object_id *oid, unsigned long *size)
27d69a46 1202{
0fcfd160 1203 void *map;
144bde78 1204 int fd;
84dd81c1 1205
f6371f92
JK
1206 if (path)
1207 fd = git_open(path);
8261e1f1 1208 else
514c5fdd 1209 fd = open_loose_object(r, oid, &path);
44d1c19e
LT
1210 map = NULL;
1211 if (fd >= 0) {
1212 struct stat st;
1f688557 1213
44d1c19e
LT
1214 if (!fstat(fd, &st)) {
1215 *size = xsize_t(st.st_size);
33e42de0
MM
1216 if (!*size) {
1217 /* mmap() is forbidden on empty files */
259328b7 1218 error(_("object file %s is empty"), path);
6881925e 1219 close(fd);
33e42de0
MM
1220 return NULL;
1221 }
44d1c19e 1222 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
144bde78 1223 }
44d1c19e 1224 close(fd);
74e34e1f 1225 }
0fcfd160 1226 return map;
74e34e1f
NP
1227}
1228
514c5fdd
JK
1229void *map_loose_object(struct repository *r,
1230 const struct object_id *oid,
1231 unsigned long *size)
068f85e3 1232{
514c5fdd 1233 return map_loose_object_1(r, NULL, oid, size);
068f85e3 1234}
1235
00a7760e
JK
1236static int unpack_loose_short_header(git_zstream *stream,
1237 unsigned char *map, unsigned long mapsize,
1238 void *buffer, unsigned long bufsiz)
47fe3f6e 1239{
31877c9a
MT
1240 int ret;
1241
c4483576
LT
1242 /* Get the data stream */
1243 memset(stream, 0, sizeof(*stream));
1244 stream->next_in = map;
1245 stream->avail_in = mapsize;
1246 stream->next_out = buffer;
93821bd9 1247 stream->avail_out = bufsiz;
47fe3f6e 1248
39c68542 1249 git_inflate_init(stream);
31877c9a
MT
1250 obj_read_unlock();
1251 ret = git_inflate(stream, 0);
1252 obj_read_lock();
1253
1254 return ret;
9a217f2a
JH
1255}
1256
00a7760e
JK
1257int unpack_loose_header(git_zstream *stream,
1258 unsigned char *map, unsigned long mapsize,
1259 void *buffer, unsigned long bufsiz)
1f688557 1260{
00a7760e
JK
1261 int status = unpack_loose_short_header(stream, map, mapsize,
1262 buffer, bufsiz);
42873078 1263
d21f8426
JH
1264 if (status < Z_OK)
1265 return status;
628522ec 1266
d21f8426
JH
1267 /* Make sure we have the terminating NUL */
1268 if (!memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1269 return -1;
1f688557
JH
1270 return 0;
1271}
1272
00a7760e
JK
1273static int unpack_loose_header_to_strbuf(git_zstream *stream, unsigned char *map,
1274 unsigned long mapsize, void *buffer,
1275 unsigned long bufsiz, struct strbuf *header)
d131b7af 1276{
46f03448
KN
1277 int status;
1278
00a7760e 1279 status = unpack_loose_short_header(stream, map, mapsize, buffer, bufsiz);
d21f8426
JH
1280 if (status < Z_OK)
1281 return -1;
d131b7af 1282
46f03448
KN
1283 /*
1284 * Check if entire header is unpacked in the first iteration.
d131b7af 1285 */
46f03448
KN
1286 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1287 return 0;
d131b7af 1288
46f03448
KN
1289 /*
1290 * buffer[0..bufsiz] was not large enough. Copy the partial
1291 * result out to header, and then append the result of further
1292 * reading the stream.
1293 */
1294 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1295 stream->next_out = buffer;
1296 stream->avail_out = bufsiz;
d131b7af 1297
46f03448 1298 do {
31877c9a 1299 obj_read_unlock();
46f03448 1300 status = git_inflate(stream, 0);
31877c9a 1301 obj_read_lock();
46f03448
KN
1302 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1303 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1304 return 0;
1305 stream->next_out = buffer;
1306 stream->avail_out = bufsiz;
1307 } while (status != Z_STREAM_END);
1308 return -1;
d131b7af
SP
1309}
1310
00a7760e
JK
1311static void *unpack_loose_rest(git_zstream *stream,
1312 void *buffer, unsigned long size,
1313 const struct object_id *oid)
95099731 1314{
5180cacc 1315 int bytes = strlen(buffer) + 1;
3aee68aa 1316 unsigned char *buf = xmallocz(size);
93821bd9 1317 unsigned long n;
7efbff75 1318 int status = Z_OK;
95099731 1319
93821bd9
LT
1320 n = stream->total_out - bytes;
1321 if (n > size)
1322 n = size;
1323 memcpy(buf, (char *) buffer + bytes, n);
1324 bytes = n;
456cdf6e
LT
1325 if (bytes <= size) {
1326 /*
1327 * The above condition must be (bytes <= size), not
1328 * (bytes < size). In other words, even though we
ccf5ace0 1329 * expect no more output and set avail_out to zero,
456cdf6e
LT
1330 * the input zlib stream may have bytes that express
1331 * "this concludes the stream", and we *do* want to
1332 * eat that input.
1333 *
1334 * Otherwise we would not be able to test that we
1335 * consumed all the input to reach the expected size;
1336 * we also want to check that zlib tells us that all
1337 * went well with status == Z_STREAM_END at the end.
1338 */
5180cacc
LT
1339 stream->next_out = buf + bytes;
1340 stream->avail_out = size - bytes;
31877c9a
MT
1341 while (status == Z_OK) {
1342 obj_read_unlock();
39c68542 1343 status = git_inflate(stream, Z_FINISH);
31877c9a
MT
1344 obj_read_lock();
1345 }
5180cacc 1346 }
456cdf6e 1347 if (status == Z_STREAM_END && !stream->avail_in) {
39c68542 1348 git_inflate_end(stream);
7efbff75 1349 return buf;
95099731
NTND
1350 }
1351
7efbff75 1352 if (status < 0)
00a7760e 1353 error(_("corrupt loose object '%s'"), oid_to_hex(oid));
7efbff75 1354 else if (stream->avail_in)
259328b7 1355 error(_("garbage at end of loose object '%s'"),
00a7760e 1356 oid_to_hex(oid));
7efbff75
JH
1357 free(buf);
1358 return NULL;
95099731
NTND
1359}
1360
d40d535b 1361/*
5180cacc
LT
1362 * We used to just use "sscanf()", but that's actually way
1363 * too permissive for what we want to check. So do an anal
1364 * object header parse by hand.
d40d535b 1365 */
00a7760e
JK
1366static int parse_loose_header_extended(const char *hdr, struct object_info *oi,
1367 unsigned int flags)
1f688557 1368{
46f03448 1369 const char *type_buf = hdr;
5180cacc 1370 unsigned long size;
46f03448 1371 int type, type_len = 0;
43057304 1372
5180cacc 1373 /*
46f03448 1374 * The type can be of any size but is followed by
21666f1a 1375 * a space.
5180cacc 1376 */
5180cacc
LT
1377 for (;;) {
1378 char c = *hdr++;
d21f8426
JH
1379 if (!c)
1380 return -1;
5180cacc
LT
1381 if (c == ' ')
1382 break;
46f03448 1383 type_len++;
5180cacc 1384 }
1f688557 1385
46f03448 1386 type = type_from_string_gently(type_buf, type_len, 1);
6ca32f47
BW
1387 if (oi->type_name)
1388 strbuf_add(oi->type_name, type_buf, type_len);
46f03448
KN
1389 /*
1390 * Set type to 0 if its an unknown object and
2e3a16b2 1391 * we're obtaining the type using '--allow-unknown-type'
46f03448
KN
1392 * option.
1393 */
19fc5e84 1394 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0))
46f03448
KN
1395 type = 0;
1396 else if (type < 0)
259328b7 1397 die(_("invalid object type"));
46f03448
KN
1398 if (oi->typep)
1399 *oi->typep = type;
5180cacc
LT
1400
1401 /*
1402 * The length must follow immediately, and be in canonical
1403 * decimal format (ie "010" is not valid).
1404 */
1405 size = *hdr++ - '0';
1406 if (size > 9)
1407 return -1;
1408 if (size) {
1409 for (;;) {
1410 unsigned long c = *hdr - '0';
1411 if (c > 9)
1412 break;
1413 hdr++;
1414 size = size * 10 + c;
1b1005d1 1415 }
c01f51cc 1416 }
46f03448
KN
1417
1418 if (oi->sizep)
1419 *oi->sizep = size;
5180cacc
LT
1420
1421 /*
1422 * The length must be followed by a zero byte
1423 */
46f03448 1424 return *hdr ? -1 : type;
1f688557
JH
1425}
1426
00a7760e 1427int parse_loose_header(const char *hdr, unsigned long *sizep)
bf592c50 1428{
27b5c1a0 1429 struct object_info oi = OBJECT_INFO_INIT;
21666f1a 1430
46f03448 1431 oi.sizep = sizep;
00a7760e 1432 return parse_loose_header_extended(hdr, &oi, 0);
bf592c50
DB
1433}
1434
514c5fdd
JK
1435static int loose_object_info(struct repository *r,
1436 const struct object_id *oid,
1437 struct object_info *oi, int flags)
65c2e0c3 1438{
46f03448
KN
1439 int status = 0;
1440 unsigned long mapsize;
65c2e0c3 1441 void *map;
ef49a7a0 1442 git_zstream stream;
1af64f73 1443 char hdr[MAX_HEADER_LEN];
46f03448 1444 struct strbuf hdrbuf = STRBUF_INIT;
c84a1f3e 1445 unsigned long size_scratch;
65c2e0c3 1446
b99b6bcc
JK
1447 if (oi->delta_base_oid)
1448 oidclr(oi->delta_base_oid);
5d642e75 1449
052fe5ea
JK
1450 /*
1451 * If we don't care about type or size, then we don't
4ef8d1dd
JH
1452 * need to look inside the object at all. Note that we
1453 * do not optimize out the stat call, even if the
1454 * caller doesn't care about the disk-size, since our
1455 * return value implicitly indicates whether the
1456 * object even exists.
052fe5ea 1457 */
6ca32f47 1458 if (!oi->typep && !oi->type_name && !oi->sizep && !oi->contentp) {
771e7d57 1459 const char *path;
4ef8d1dd 1460 struct stat st;
61c7711c 1461 if (!oi->disk_sizep && (flags & OBJECT_INFO_QUICK))
d7a24573 1462 return quick_has_loose(r, oid) ? 0 : -1;
514c5fdd 1463 if (stat_loose_object(r, oid, &st, &path) < 0)
4ef8d1dd
JH
1464 return -1;
1465 if (oi->disk_sizep)
23c339c0 1466 *oi->disk_sizep = st.st_size;
052fe5ea
JK
1467 return 0;
1468 }
1469
514c5fdd 1470 map = map_loose_object(r, oid, &mapsize);
f0df4ed5 1471 if (!map)
dbea72a8 1472 return -1;
c84a1f3e
JT
1473
1474 if (!oi->sizep)
1475 oi->sizep = &size_scratch;
1476
23c339c0
JK
1477 if (oi->disk_sizep)
1478 *oi->disk_sizep = mapsize;
19fc5e84 1479 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) {
00a7760e 1480 if (unpack_loose_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
259328b7 1481 status = error(_("unable to unpack %s header with --allow-unknown-type"),
514c5fdd 1482 oid_to_hex(oid));
00a7760e 1483 } else if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
259328b7 1484 status = error(_("unable to unpack %s header"),
514c5fdd 1485 oid_to_hex(oid));
46f03448
KN
1486 if (status < 0)
1487 ; /* Do nothing */
1488 else if (hdrbuf.len) {
00a7760e 1489 if ((status = parse_loose_header_extended(hdrbuf.buf, oi, flags)) < 0)
259328b7 1490 status = error(_("unable to parse %s header with --allow-unknown-type"),
514c5fdd 1491 oid_to_hex(oid));
00a7760e 1492 } else if ((status = parse_loose_header_extended(hdr, oi, flags)) < 0)
514c5fdd 1493 status = error(_("unable to parse %s header"), oid_to_hex(oid));
c84a1f3e 1494
b3ea7dd3 1495 if (status >= 0 && oi->contentp) {
00a7760e
JK
1496 *oi->contentp = unpack_loose_rest(&stream, hdr,
1497 *oi->sizep, oid);
b3ea7dd3
JK
1498 if (!*oi->contentp) {
1499 git_inflate_end(&stream);
1500 status = -1;
1501 }
1502 } else
c84a1f3e
JT
1503 git_inflate_end(&stream);
1504
65c2e0c3 1505 munmap(map, mapsize);
46f03448 1506 if (status && oi->typep)
23c339c0 1507 *oi->typep = status;
c84a1f3e
JT
1508 if (oi->sizep == &size_scratch)
1509 oi->sizep = NULL;
46f03448 1510 strbuf_release(&hdrbuf);
3ab0fb06 1511 oi->whence = OI_LOOSE;
93cff9a9 1512 return (status < 0) ? status : 0;
65c2e0c3
JH
1513}
1514
31877c9a
MT
1515int obj_read_use_lock = 0;
1516pthread_mutex_t obj_read_mutex;
1517
1518void enable_obj_read_lock(void)
1519{
1520 if (obj_read_use_lock)
1521 return;
1522
1523 obj_read_use_lock = 1;
1524 init_recursive_mutex(&obj_read_mutex);
1525}
1526
1527void disable_obj_read_lock(void)
1528{
1529 if (!obj_read_use_lock)
1530 return;
1531
1532 obj_read_use_lock = 0;
1533 pthread_mutex_destroy(&obj_read_mutex);
1534}
1535
8b4c0103
JT
1536int fetch_if_missing = 1;
1537
31877c9a
MT
1538static int do_oid_object_info_extended(struct repository *r,
1539 const struct object_id *oid,
1540 struct object_info *oi, unsigned flags)
f0df4ed5 1541{
cd585e2a 1542 static struct object_info blank_oi = OBJECT_INFO_INIT;
9c8a294a 1543 struct cached_object *co;
f0df4ed5 1544 struct pack_entry e;
5b086407 1545 int rtype;
b383a13c 1546 const struct object_id *real = oid;
8b4c0103 1547 int already_retried = 0;
f0df4ed5 1548
31877c9a 1549
b383a13c 1550 if (flags & OBJECT_INFO_LOOKUP_REPLACE)
9d98354f 1551 real = lookup_replace_object(r, oid);
f0df4ed5 1552
b383a13c 1553 if (is_null_oid(real))
87b5e236
JK
1554 return -1;
1555
cd585e2a
JT
1556 if (!oi)
1557 oi = &blank_oi;
1558
9c8a294a
JT
1559 co = find_cached_object(real);
1560 if (co) {
1561 if (oi->typep)
1562 *(oi->typep) = co->type;
1563 if (oi->sizep)
1564 *(oi->sizep) = co->size;
1565 if (oi->disk_sizep)
1566 *(oi->disk_sizep) = 0;
b99b6bcc
JK
1567 if (oi->delta_base_oid)
1568 oidclr(oi->delta_base_oid);
9c8a294a
JT
1569 if (oi->type_name)
1570 strbuf_addstr(oi->type_name, type_name(co->type));
1571 if (oi->contentp)
1572 *oi->contentp = xmemdupz(co->buf, co->size);
1573 oi->whence = OI_CACHED;
1574 return 0;
c4d9986f
NTND
1575 }
1576
8b4c0103 1577 while (1) {
42c8ce1c 1578 if (find_pack_entry(r, real, &e))
8b4c0103
JT
1579 break;
1580
024aa469
TI
1581 if (flags & OBJECT_INFO_IGNORE_LOOSE)
1582 return -1;
1583
ddd63e64 1584 /* Most likely it's a loose object. */
514c5fdd 1585 if (!loose_object_info(r, real, oi, flags))
5b086407 1586 return 0;
ddd63e64
SG
1587
1588 /* Not a loose object; someone else may have just packed it. */
2b7750c9 1589 if (!(flags & OBJECT_INFO_QUICK)) {
9d98354f 1590 reprepare_packed_git(r);
42c8ce1c 1591 if (find_pack_entry(r, real, &e))
2b7750c9
JT
1592 break;
1593 }
8b4c0103
JT
1594
1595 /* Check if it is a missing object */
ef830cc4
JT
1596 if (fetch_if_missing && repo_has_promisor_remote(r) &&
1597 !already_retried &&
31f5256c 1598 !(flags & OBJECT_INFO_SKIP_FETCH_OBJECT)) {
8b4c0103 1599 /*
b14ed5ad
CC
1600 * TODO Investigate checking promisor_remote_get_direct()
1601 * TODO return value and stopping on error here.
8b4c0103 1602 */
b14ed5ad 1603 promisor_remote_get_direct(r, real, 1);
8b4c0103
JT
1604 already_retried = 1;
1605 continue;
dfdd4afc 1606 }
8b4c0103
JT
1607
1608 return -1;
f0df4ed5 1609 }
3d77d877 1610
cd585e2a
JT
1611 if (oi == &blank_oi)
1612 /*
1613 * We know that the caller doesn't actually need the
1614 * information below, so return early.
1615 */
1616 return 0;
9d98354f 1617 rtype = packed_object_info(r, e.p, e.offset, oi);
412916ee 1618 if (rtype < 0) {
b383a13c 1619 mark_bad_packed_object(e.p, real->hash);
31877c9a 1620 return do_oid_object_info_extended(r, real, oi, 0);
3ab0fb06 1621 } else if (oi->whence == OI_PACKED) {
9a490590
JH
1622 oi->u.packed.offset = e.offset;
1623 oi->u.packed.pack = e.p;
1624 oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
1625 rtype == OBJ_OFS_DELTA);
3d77d877
NP
1626 }
1627
5b086407 1628 return 0;
f0df4ed5
JS
1629}
1630
31877c9a
MT
1631int oid_object_info_extended(struct repository *r, const struct object_id *oid,
1632 struct object_info *oi, unsigned flags)
1633{
1634 int ret;
1635 obj_read_lock();
1636 ret = do_oid_object_info_extended(r, oid, oi, flags);
1637 obj_read_unlock();
1638 return ret;
1639}
1640
1641
3fc0dca9 1642/* returns enum object_type or negative */
9d98354f
SB
1643int oid_object_info(struct repository *r,
1644 const struct object_id *oid,
1645 unsigned long *sizep)
9a490590 1646{
5b086407 1647 enum object_type type;
27b5c1a0 1648 struct object_info oi = OBJECT_INFO_INIT;
9a490590 1649
5b086407 1650 oi.typep = &type;
9a490590 1651 oi.sizep = sizep;
9d98354f
SB
1652 if (oid_object_info_extended(r, oid, &oi,
1653 OBJECT_INFO_LOOKUP_REPLACE) < 0)
5b086407
JK
1654 return -1;
1655 return type;
9a490590
JH
1656}
1657
1b9b5c69 1658static void *read_object(struct repository *r,
cba595ab 1659 const struct object_id *oid, enum object_type *type,
f1d8130b
JT
1660 unsigned long *size)
1661{
1662 struct object_info oi = OBJECT_INFO_INIT;
1663 void *content;
1664 oi.typep = type;
1665 oi.sizep = size;
1666 oi.contentp = &content;
1667
cba595ab 1668 if (oid_object_info_extended(r, oid, &oi, 0) < 0)
f1d8130b
JT
1669 return NULL;
1670 return content;
1671}
1672
829e5c3b
PO
1673int pretend_object_file(void *buf, unsigned long len, enum object_type type,
1674 struct object_id *oid)
d66b37bb
JH
1675{
1676 struct cached_object *co;
1677
2dcde20e 1678 hash_object_file(the_hash_algo, buf, len, type_name(type), oid);
a64d2aae
JT
1679 if (has_object_file_with_flags(oid, OBJECT_INFO_QUICK | OBJECT_INFO_SKIP_FETCH_OBJECT) ||
1680 find_cached_object(oid))
d66b37bb 1681 return 0;
c7353967 1682 ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc);
d66b37bb
JH
1683 co = &cached_objects[cached_object_nr++];
1684 co->size = len;
21666f1a 1685 co->type = type;
efa13f7b
JH
1686 co->buf = xmalloc(len);
1687 memcpy(co->buf, buf, len);
62ba93ea 1688 oidcpy(&co->oid, oid);
d66b37bb
JH
1689 return 0;
1690}
1691
b6c4cecc
JH
1692/*
1693 * This function dies on corrupt objects; the callers who want to
1694 * deal with them should arrange to call read_object() and give error
1695 * messages themselves.
1696 */
a3b72c89
SB
1697void *read_object_file_extended(struct repository *r,
1698 const struct object_id *oid,
b4f5aca4 1699 enum object_type *type,
1700 unsigned long *size,
1701 int lookup_replace)
ac939109 1702{
3ba7a065 1703 void *data;
b6c4cecc 1704 const struct packed_git *p;
771e7d57
JK
1705 const char *path;
1706 struct stat st;
1f2e7cea 1707 const struct object_id *repl = lookup_replace ?
a3b72c89 1708 lookup_replace_object(r, oid) : oid;
b6c4cecc 1709
3ba7a065 1710 errno = 0;
cba595ab 1711 data = read_object(r, repl, type, size);
4bbf5a26 1712 if (data)
b6c4cecc 1713 return data;
68095570 1714
31877c9a 1715 obj_read_lock();
25f3af3f 1716 if (errno && errno != ENOENT)
259328b7 1717 die_errno(_("failed to read object %s"), oid_to_hex(oid));
3ba7a065 1718
68095570 1719 /* die if we replaced an object with one that does not exist */
b383a13c 1720 if (repl != oid)
259328b7 1721 die(_("replacement %s not found for %s"),
b383a13c 1722 oid_to_hex(repl), oid_to_hex(oid));
68095570 1723
cba595ab 1724 if (!stat_loose_object(r, repl, &st, &path))
259328b7 1725 die(_("loose object %s (stored in %s) is corrupt"),
b383a13c 1726 oid_to_hex(repl), path);
68095570 1727
a3b72c89 1728 if ((p = has_packed_and_bad(r, repl->hash)) != NULL)
259328b7 1729 die(_("packed object %s (stored in %s) is corrupt"),
b383a13c 1730 oid_to_hex(repl), p->pack_name);
31877c9a 1731 obj_read_unlock();
f5552aee 1732
b6c4cecc 1733 return NULL;
ac939109
NP
1734}
1735
d3b4705a
NTND
1736void *read_object_with_reference(struct repository *r,
1737 const struct object_id *oid,
21666f1a 1738 const char *required_type_name,
40469ee9 1739 unsigned long *size,
02f0547e 1740 struct object_id *actual_oid_return)
f4913f91 1741{
21666f1a 1742 enum object_type type, required_type;
f4913f91
JH
1743 void *buffer;
1744 unsigned long isize;
02f0547e 1745 struct object_id actual_oid;
f4913f91 1746
21666f1a 1747 required_type = type_from_string(required_type_name);
02f0547e 1748 oidcpy(&actual_oid, oid);
40469ee9
JH
1749 while (1) {
1750 int ref_length = -1;
1751 const char *ref_type = NULL;
f4913f91 1752
d3b4705a 1753 buffer = repo_read_object_file(r, &actual_oid, &type, &isize);
40469ee9
JH
1754 if (!buffer)
1755 return NULL;
21666f1a 1756 if (type == required_type) {
40469ee9 1757 *size = isize;
02f0547e 1758 if (actual_oid_return)
1759 oidcpy(actual_oid_return, &actual_oid);
40469ee9
JH
1760 return buffer;
1761 }
1762 /* Handle references */
21666f1a 1763 else if (type == OBJ_COMMIT)
40469ee9 1764 ref_type = "tree ";
21666f1a 1765 else if (type == OBJ_TAG)
40469ee9
JH
1766 ref_type = "object ";
1767 else {
1768 free(buffer);
1769 return NULL;
1770 }
1771 ref_length = strlen(ref_type);
f4913f91 1772
94b5e093 1773 if (ref_length + the_hash_algo->hexsz > isize ||
50974ec9 1774 memcmp(buffer, ref_type, ref_length) ||
02f0547e 1775 get_oid_hex((char *) buffer + ref_length, &actual_oid)) {
40469ee9
JH
1776 free(buffer);
1777 return NULL;
1778 }
1cf58e72 1779 free(buffer);
40469ee9 1780 /* Now we have the ID of the referred-to object in
02f0547e 1781 * actual_oid. Check again. */
f4913f91 1782 }
f4913f91
JH
1783}
1784
7ad5c44d
MT
1785static void write_object_file_prepare(const struct git_hash_algo *algo,
1786 const void *buf, unsigned long len,
a09c985e
PO
1787 const char *type, struct object_id *oid,
1788 char *hdr, int *hdrlen)
d410c0f5 1789{
18e2588e 1790 git_hash_ctx c;
d410c0f5
JH
1791
1792 /* Generate the header */
ca473cef 1793 *hdrlen = xsnprintf(hdr, *hdrlen, "%s %"PRIuMAX , type, (uintmax_t)len)+1;
d410c0f5
JH
1794
1795 /* Sha1.. */
7ad5c44d
MT
1796 algo->init_fn(&c);
1797 algo->update_fn(&c, hdr, *hdrlen);
1798 algo->update_fn(&c, buf, len);
5951bf46 1799 algo->final_oid_fn(oid, &c);
d410c0f5
JH
1800}
1801
230f1322 1802/*
5a688fe4 1803 * Move the just written object into its final resting place.
230f1322 1804 */
cb5add58 1805int finalize_object_file(const char *tmpfile, const char *filename)
230f1322 1806{
e32c0a9c 1807 int ret = 0;
5a688fe4 1808
348df166 1809 if (object_creation_mode == OBJECT_CREATION_USES_RENAMES)
be66a6c4
JS
1810 goto try_rename;
1811 else if (link(tmpfile, filename))
e32c0a9c 1812 ret = errno;
7ebb6fca
LT
1813
1814 /*
1815 * Coda hack - coda doesn't like cross-directory links,
1816 * so we fall back to a rename, which will mean that it
1817 * won't be able to check collisions, but that's not a
1818 * big deal.
1819 *
1820 * The same holds for FAT formatted media.
1821 *
3be1f18e 1822 * When this succeeds, we just return. We have nothing
7ebb6fca
LT
1823 * left to unlink.
1824 */
1825 if (ret && ret != EEXIST) {
be66a6c4 1826 try_rename:
7ebb6fca 1827 if (!rename(tmpfile, filename))
3be1f18e 1828 goto out;
9e48b389 1829 ret = errno;
230f1322 1830 }
691f1a28 1831 unlink_or_warn(tmpfile);
230f1322
LT
1832 if (ret) {
1833 if (ret != EEXIST) {
2c319886 1834 return error_errno(_("unable to write file %s"), filename);
230f1322
LT
1835 }
1836 /* FIXME!!! Collision check here ? */
1837 }
1838
3be1f18e 1839out:
5256b006 1840 if (adjust_shared_perm(filename))
259328b7 1841 return error(_("unable to set permission to '%s'"), filename);
230f1322
LT
1842 return 0;
1843}
1844
4d548150
LT
1845static int write_buffer(int fd, const void *buf, size_t len)
1846{
d34cf19b 1847 if (write_in_full(fd, buf, len) < 0)
259328b7 1848 return error_errno(_("file write error"));
4d548150
LT
1849 return 0;
1850}
1851
2dcde20e
MT
1852int hash_object_file(const struct git_hash_algo *algo, const void *buf,
1853 unsigned long len, const char *type,
f070facc 1854 struct object_id *oid)
abdc3fc8 1855{
1af64f73 1856 char hdr[MAX_HEADER_LEN];
ef1286d3 1857 int hdrlen = sizeof(hdr);
2dcde20e 1858 write_object_file_prepare(algo, buf, len, type, oid, hdr, &hdrlen);
abdc3fc8
RS
1859 return 0;
1860}
1861
e9039dd3 1862/* Finalize a file on disk, and close it. */
514c5fdd 1863static void close_loose_object(int fd)
e9039dd3 1864{
aafe9fba 1865 if (fsync_object_files)
76011357 1866 fsync_or_die(fd, "loose object file");
e9039dd3 1867 if (close(fd) != 0)
76011357 1868 die_errno(_("error when closing loose object file"));
e9039dd3
LT
1869}
1870
5723fe7e
LT
1871/* Size of directory component, including the ending '/' */
1872static inline int directory_size(const char *filename)
1873{
1874 const char *s = strrchr(filename, '/');
1875 if (!s)
1876 return 0;
1877 return s - filename + 1;
1878}
1879
1880/*
1881 * This creates a temporary file in the same directory as the final
1882 * 'filename'
1883 *
1884 * We want to avoid cross-directory filename renames, because those
1885 * can have problems on various filesystems (FAT, NFS, Coda).
1886 */
d4b3d11a 1887static int create_tmpfile(struct strbuf *tmp, const char *filename)
5723fe7e
LT
1888{
1889 int fd, dirlen = directory_size(filename);
1890
d4b3d11a
JK
1891 strbuf_reset(tmp);
1892 strbuf_add(tmp, filename, dirlen);
1893 strbuf_addstr(tmp, "tmp_obj_XXXXXX");
1894 fd = git_mkstemp_mode(tmp->buf, 0444);
cbacbf4e 1895 if (fd < 0 && dirlen && errno == ENOENT) {
d4b3d11a
JK
1896 /*
1897 * Make sure the directory exists; note that the contents
1898 * of the buffer are undefined after mkstemp returns an
1899 * error, so we have to rewrite the whole buffer from
1900 * scratch.
1901 */
1902 strbuf_reset(tmp);
1903 strbuf_add(tmp, filename, dirlen - 1);
1904 if (mkdir(tmp->buf, 0777) && errno != EEXIST)
b2476a60 1905 return -1;
d4b3d11a 1906 if (adjust_shared_perm(tmp->buf))
5723fe7e
LT
1907 return -1;
1908
1909 /* Try again */
d4b3d11a
JK
1910 strbuf_addstr(tmp, "/tmp_obj_XXXXXX");
1911 fd = git_mkstemp_mode(tmp->buf, 0444);
5723fe7e
LT
1912 }
1913 return fd;
1914}
1915
3fc7281f
PO
1916static int write_loose_object(const struct object_id *oid, char *hdr,
1917 int hdrlen, const void *buf, unsigned long len,
1918 time_t mtime)
0fcfd160 1919{
915308b1 1920 int fd, ret;
9892beba 1921 unsigned char compressed[4096];
ef49a7a0 1922 git_zstream stream;
18e2588e 1923 git_hash_ctx c;
3fc7281f 1924 struct object_id parano_oid;
d4b3d11a 1925 static struct strbuf tmp_file = STRBUF_INIT;
ea657730
CC
1926 static struct strbuf filename = STRBUF_INIT;
1927
514c5fdd 1928 loose_object_path(the_repository, &filename, oid);
a44c9a5e 1929
ea657730 1930 fd = create_tmpfile(&tmp_file, filename.buf);
aac17941 1931 if (fd < 0) {
35243577 1932 if (errno == EACCES)
259328b7 1933 return error(_("insufficient permission for adding an object to repository database %s"), get_object_directory());
916d081b 1934 else
259328b7 1935 return error_errno(_("unable to create temporary file"));
aac17941
LT
1936 }
1937
0fcfd160 1938 /* Set it up */
55bb5c91 1939 git_deflate_init(&stream, zlib_compression_level);
0fcfd160 1940 stream.next_out = compressed;
9892beba 1941 stream.avail_out = sizeof(compressed);
18e2588e 1942 the_hash_algo->init_fn(&c);
a44c9a5e
LT
1943
1944 /* First header.. */
d65a16f6 1945 stream.next_in = (unsigned char *)hdr;
a44c9a5e 1946 stream.avail_in = hdrlen;
55bb5c91
JH
1947 while (git_deflate(&stream, 0) == Z_OK)
1948 ; /* nothing */
18e2588e 1949 the_hash_algo->update_fn(&c, hdr, hdrlen);
a44c9a5e
LT
1950
1951 /* Then the data itself.. */
c00e657d 1952 stream.next_in = (void *)buf;
a44c9a5e 1953 stream.avail_in = len;
9892beba 1954 do {
748af44c 1955 unsigned char *in0 = stream.next_in;
55bb5c91 1956 ret = git_deflate(&stream, Z_FINISH);
18e2588e 1957 the_hash_algo->update_fn(&c, in0, stream.next_in - in0);
9892beba 1958 if (write_buffer(fd, compressed, stream.next_out - compressed) < 0)
76011357 1959 die(_("unable to write loose object file"));
9892beba
NP
1960 stream.next_out = compressed;
1961 stream.avail_out = sizeof(compressed);
1962 } while (ret == Z_OK);
1963
ac54c277 1964 if (ret != Z_STREAM_END)
259328b7 1965 die(_("unable to deflate new object %s (%d)"), oid_to_hex(oid),
3fc7281f 1966 ret);
55bb5c91 1967 ret = git_deflate_end_gently(&stream);
ac54c277 1968 if (ret != Z_OK)
259328b7 1969 die(_("deflateEnd on object %s failed (%d)"), oid_to_hex(oid),
3fc7281f 1970 ret);
5951bf46 1971 the_hash_algo->final_oid_fn(&parano_oid, &c);
9001dc2a 1972 if (!oideq(oid, &parano_oid))
259328b7 1973 die(_("confused by unstable object source data for %s"),
3fc7281f 1974 oid_to_hex(oid));
ac54c277 1975
514c5fdd 1976 close_loose_object(fd);
0fcfd160 1977
bbac7311
NP
1978 if (mtime) {
1979 struct utimbuf utb;
1980 utb.actime = mtime;
1981 utb.modtime = mtime;
d4b3d11a 1982 if (utime(tmp_file.buf, &utb) < 0)
259328b7 1983 warning_errno(_("failed utime() on %s"), tmp_file.buf);
bbac7311
NP
1984 }
1985
ea657730 1986 return finalize_object_file(tmp_file.buf, filename.buf);
0fcfd160 1987}
8237b185 1988
6862ebbf 1989static int freshen_loose_object(const struct object_id *oid)
33d4221c 1990{
6862ebbf 1991 return check_and_freshen(oid, 1);
33d4221c
JK
1992}
1993
6862ebbf 1994static int freshen_packed_object(const struct object_id *oid)
33d4221c
JK
1995{
1996 struct pack_entry e;
544443cb 1997 if (!find_pack_entry(the_repository, oid, &e))
ee1c6c34
JK
1998 return 0;
1999 if (e.p->freshened)
2000 return 1;
2001 if (!freshen_file(e.p->pack_name))
2002 return 0;
2003 e.p->freshened = 1;
2004 return 1;
33d4221c
JK
2005}
2006
a09c985e
PO
2007int write_object_file(const void *buf, unsigned long len, const char *type,
2008 struct object_id *oid)
bbac7311 2009{
1af64f73 2010 char hdr[MAX_HEADER_LEN];
ef1286d3 2011 int hdrlen = sizeof(hdr);
bbac7311
NP
2012
2013 /* Normally if we have it in the pack then we do not bother writing
2014 * it out into .git/objects/??/?{38} file.
2015 */
7ad5c44d
MT
2016 write_object_file_prepare(the_hash_algo, buf, len, type, oid, hdr,
2017 &hdrlen);
6862ebbf 2018 if (freshen_packed_object(oid) || freshen_loose_object(oid))
bbac7311 2019 return 0;
3fc7281f 2020 return write_loose_object(oid, hdr, hdrlen, buf, len, 0);
bbac7311
NP
2021}
2022
1752cbbc
PO
2023int hash_object_file_literally(const void *buf, unsigned long len,
2024 const char *type, struct object_id *oid,
2025 unsigned flags)
0c3db67c
ES
2026{
2027 char *header;
2028 int hdrlen, status = 0;
2029
2030 /* type string, SP, %lu of the length plus NUL must fit this */
1af64f73 2031 hdrlen = strlen(type) + MAX_HEADER_LEN;
ef1286d3 2032 header = xmalloc(hdrlen);
7ad5c44d
MT
2033 write_object_file_prepare(the_hash_algo, buf, len, type, oid, header,
2034 &hdrlen);
0c3db67c
ES
2035
2036 if (!(flags & HASH_WRITE_OBJECT))
2037 goto cleanup;
6862ebbf 2038 if (freshen_packed_object(oid) || freshen_loose_object(oid))
0c3db67c 2039 goto cleanup;
3fc7281f 2040 status = write_loose_object(oid, header, hdrlen, buf, len, 0);
0c3db67c
ES
2041
2042cleanup:
2043 free(header);
2044 return status;
2045}
2046
4bdb70a4 2047int force_object_loose(const struct object_id *oid, time_t mtime)
bbac7311 2048{
bbac7311
NP
2049 void *buf;
2050 unsigned long len;
2051 enum object_type type;
1af64f73 2052 char hdr[MAX_HEADER_LEN];
bbac7311 2053 int hdrlen;
1fb23e65 2054 int ret;
bbac7311 2055
6862ebbf 2056 if (has_loose_object(oid))
bbac7311 2057 return 0;
cba595ab 2058 buf = read_object(the_repository, oid, &type, &len);
bbac7311 2059 if (!buf)
2c319886 2060 return error(_("cannot read object for %s"), oid_to_hex(oid));
ca473cef 2061 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %"PRIuMAX , type_name(type), (uintmax_t)len) + 1;
3fc7281f 2062 ret = write_loose_object(oid, hdr, hdrlen, buf, len, mtime);
1fb23e65
BS
2063 free(buf);
2064
2065 return ret;
bbac7311
NP
2066}
2067
1d8d9cb6
JT
2068int has_object(struct repository *r, const struct object_id *oid,
2069 unsigned flags)
2070{
2071 int quick = !(flags & HAS_OBJECT_RECHECK_PACKED);
2072 unsigned object_info_flags = OBJECT_INFO_SKIP_FETCH_OBJECT |
2073 (quick ? OBJECT_INFO_QUICK : 0);
2074
2075 if (!startup_info->have_repository)
2076 return 0;
2077 return oid_object_info_extended(r, oid, NULL, object_info_flags) >= 0;
2078}
2079
cba595ab
JH
2080int repo_has_object_file_with_flags(struct repository *r,
2081 const struct object_id *oid, int flags)
8237b185 2082{
3e8b7d3c
JN
2083 if (!startup_info->have_repository)
2084 return 0;
9c8a294a 2085 return oid_object_info_extended(r, oid, NULL, flags) >= 0;
b419aa25 2086}
2087
9b45f499
SB
2088int repo_has_object_file(struct repository *r,
2089 const struct object_id *oid)
b419aa25 2090{
cba595ab 2091 return repo_has_object_file_with_flags(r, oid, 0);
5827a035
JK
2092}
2093
c879daa2
NTND
2094static void check_tree(const void *buf, size_t size)
2095{
2096 struct tree_desc desc;
2097 struct name_entry entry;
2098
2099 init_tree_desc(&desc, buf, size);
2100 while (tree_entry(&desc, &entry))
2101 /* do nothing
2102 * tree_entry() will die() on malformed entries */
2103 ;
2104}
2105
2106static void check_commit(const void *buf, size_t size)
2107{
2108 struct commit c;
2109 memset(&c, 0, sizeof(c));
08f4f445 2110 if (parse_commit_buffer(the_repository, &c, buf, size, 0))
259328b7 2111 die(_("corrupt commit"));
c879daa2
NTND
2112}
2113
2114static void check_tag(const void *buf, size_t size)
2115{
2116 struct tag t;
2117 memset(&t, 0, sizeof(t));
0e740fed 2118 if (parse_tag_buffer(the_repository, &t, buf, size))
259328b7 2119 die(_("corrupt tag"));
c879daa2
NTND
2120}
2121
58bf2a4c
NTND
2122static int index_mem(struct index_state *istate,
2123 struct object_id *oid, void *buf, size_t size,
c4ce46fc
JH
2124 enum object_type type,
2125 const char *path, unsigned flags)
e7332f96 2126{
6c510bee 2127 int ret, re_allocated = 0;
c4ce46fc 2128 int write_object = flags & HASH_WRITE_OBJECT;
74400e71 2129
7672db20 2130 if (!type)
edaec3fb 2131 type = OBJ_BLOB;
6c510bee
LT
2132
2133 /*
2134 * Convert blobs to git internal format
2135 */
43df4f86 2136 if ((type == OBJ_BLOB) && path) {
f285a2d7 2137 struct strbuf nbuf = STRBUF_INIT;
58bf2a4c 2138 if (convert_to_git(istate, path, buf, size, &nbuf,
8462ff43 2139 get_conv_flags(flags))) {
b315c5c0 2140 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
2141 re_allocated = 1;
2142 }
2143 }
c4ce46fc 2144 if (flags & HASH_FORMAT_CHECK) {
c879daa2
NTND
2145 if (type == OBJ_TREE)
2146 check_tree(buf, size);
2147 if (type == OBJ_COMMIT)
2148 check_commit(buf, size);
2149 if (type == OBJ_TAG)
2150 check_tag(buf, size);
2151 }
6c510bee 2152
7672db20 2153 if (write_object)
169c9c01 2154 ret = write_object_file(buf, size, type_name(type), oid);
abdc3fc8 2155 else
2dcde20e
MT
2156 ret = hash_object_file(the_hash_algo, buf, size,
2157 type_name(type), oid);
43df4f86 2158 if (re_allocated)
6c510bee 2159 free(buf);
43df4f86
DP
2160 return ret;
2161}
2162
58bf2a4c
NTND
2163static int index_stream_convert_blob(struct index_state *istate,
2164 struct object_id *oid,
2165 int fd,
2166 const char *path,
2167 unsigned flags)
9035d75a
SP
2168{
2169 int ret;
2170 const int write_object = flags & HASH_WRITE_OBJECT;
2171 struct strbuf sbuf = STRBUF_INIT;
2172
2173 assert(path);
58bf2a4c 2174 assert(would_convert_to_git_filter_fd(istate, path));
9035d75a 2175
58bf2a4c 2176 convert_to_git_filter_fd(istate, path, fd, &sbuf,
8462ff43 2177 get_conv_flags(flags));
9035d75a
SP
2178
2179 if (write_object)
169c9c01 2180 ret = write_object_file(sbuf.buf, sbuf.len, type_name(OBJ_BLOB),
a09c985e 2181 oid);
9035d75a 2182 else
2dcde20e
MT
2183 ret = hash_object_file(the_hash_algo, sbuf.buf, sbuf.len,
2184 type_name(OBJ_BLOB), oid);
9035d75a
SP
2185 strbuf_release(&sbuf);
2186 return ret;
2187}
2188
58bf2a4c
NTND
2189static int index_pipe(struct index_state *istate, struct object_id *oid,
2190 int fd, enum object_type type,
7b41e1e1
JH
2191 const char *path, unsigned flags)
2192{
2193 struct strbuf sbuf = STRBUF_INIT;
2194 int ret;
2195
2196 if (strbuf_read(&sbuf, fd, 4096) >= 0)
58bf2a4c 2197 ret = index_mem(istate, oid, sbuf.buf, sbuf.len, type, path, flags);
7b41e1e1
JH
2198 else
2199 ret = -1;
2200 strbuf_release(&sbuf);
2201 return ret;
2202}
2203
ea68b0ce
DP
2204#define SMALL_FILE_SIZE (32*1024)
2205
58bf2a4c
NTND
2206static int index_core(struct index_state *istate,
2207 struct object_id *oid, int fd, size_t size,
7b41e1e1
JH
2208 enum object_type type, const char *path,
2209 unsigned flags)
43df4f86
DP
2210{
2211 int ret;
43df4f86 2212
7b41e1e1 2213 if (!size) {
58bf2a4c 2214 ret = index_mem(istate, oid, "", size, type, path, flags);
ea68b0ce
DP
2215 } else if (size <= SMALL_FILE_SIZE) {
2216 char *buf = xmalloc(size);
90dca671
JK
2217 ssize_t read_result = read_in_full(fd, buf, size);
2218 if (read_result < 0)
259328b7 2219 ret = error_errno(_("read error while indexing %s"),
90dca671
JK
2220 path ? path : "<unknown>");
2221 else if (read_result != size)
259328b7 2222 ret = error(_("short read while indexing %s"),
90dca671 2223 path ? path : "<unknown>");
ea68b0ce 2224 else
58bf2a4c 2225 ret = index_mem(istate, oid, buf, size, type, path, flags);
ea68b0ce 2226 free(buf);
08bda208 2227 } else {
43df4f86 2228 void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
58bf2a4c 2229 ret = index_mem(istate, oid, buf, size, type, path, flags);
aac17941 2230 munmap(buf, size);
08bda208 2231 }
7b41e1e1
JH
2232 return ret;
2233}
2234
4dd1fbc7 2235/*
568508e7
JH
2236 * This creates one packfile per large blob unless bulk-checkin
2237 * machinery is "plugged".
4dd1fbc7
JH
2238 *
2239 * This also bypasses the usual "convert-to-git" dance, and that is on
2240 * purpose. We could write a streaming version of the converting
2241 * functions and insert that before feeding the data to fast-import
4f22b101
JK
2242 * (or equivalent in-core API described above). However, that is
2243 * somewhat complicated, as we do not know the size of the filter
2244 * result, which we need to know beforehand when writing a git object.
2245 * Since the primary motivation for trying to stream from the working
2246 * tree file and to avoid mmaping it in core is to deal with large
2247 * binary blobs, they generally do not want to get any conversion, and
2248 * callers should avoid this code path when filters are requested.
4dd1fbc7 2249 */
7d5e1dc3 2250static int index_stream(struct object_id *oid, int fd, size_t size,
4dd1fbc7
JH
2251 enum object_type type, const char *path,
2252 unsigned flags)
2253{
68ee6dfc 2254 return index_bulk_checkin(oid, fd, size, type, path, flags);
4dd1fbc7
JH
2255}
2256
58bf2a4c
NTND
2257int index_fd(struct index_state *istate, struct object_id *oid,
2258 int fd, struct stat *st,
7b41e1e1
JH
2259 enum object_type type, const char *path, unsigned flags)
2260{
2261 int ret;
7b41e1e1 2262
9079ab7c
SP
2263 /*
2264 * Call xsize_t() only when needed to avoid potentially unnecessary
2265 * die() for large files.
2266 */
58bf2a4c
NTND
2267 if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(istate, path))
2268 ret = index_stream_convert_blob(istate, oid, fd, path, flags);
9035d75a 2269 else if (!S_ISREG(st->st_mode))
58bf2a4c 2270 ret = index_pipe(istate, oid, fd, type, path, flags);
9079ab7c 2271 else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
58bf2a4c
NTND
2272 (path && would_convert_to_git(istate, path)))
2273 ret = index_core(istate, oid, fd, xsize_t(st->st_size),
2274 type, path, flags);
4dd1fbc7 2275 else
7d5e1dc3 2276 ret = index_stream(oid, fd, xsize_t(st->st_size), type, path,
9079ab7c 2277 flags);
43df4f86 2278 close(fd);
aac17941 2279 return ret;
74400e71 2280}
ec1fcc16 2281
58bf2a4c
NTND
2282int index_path(struct index_state *istate, struct object_id *oid,
2283 const char *path, struct stat *st, unsigned flags)
ec1fcc16
JH
2284{
2285 int fd;
b760d3aa 2286 struct strbuf sb = STRBUF_INIT;
ea8e0297 2287 int rc = 0;
ec1fcc16
JH
2288
2289 switch (st->st_mode & S_IFMT) {
2290 case S_IFREG:
2291 fd = open(path, O_RDONLY);
2292 if (fd < 0)
7616c6ca 2293 return error_errno("open(\"%s\")", path);
58bf2a4c 2294 if (index_fd(istate, oid, fd, st, OBJ_BLOB, path, flags) < 0)
259328b7 2295 return error(_("%s: failed to insert into database"),
ec1fcc16
JH
2296 path);
2297 break;
2298 case S_IFLNK:
7616c6ca
NTND
2299 if (strbuf_readlink(&sb, path, st->st_size))
2300 return error_errno("readlink(\"%s\")", path);
c4ce46fc 2301 if (!(flags & HASH_WRITE_OBJECT))
2dcde20e
MT
2302 hash_object_file(the_hash_algo, sb.buf, sb.len,
2303 blob_type, oid);
a09c985e 2304 else if (write_object_file(sb.buf, sb.len, blob_type, oid))
259328b7 2305 rc = error(_("%s: failed to insert into database"), path);
b760d3aa 2306 strbuf_release(&sb);
ec1fcc16 2307 break;
f35a6d3b 2308 case S_IFDIR:
a98e6101 2309 return resolve_gitlink_ref(path, "HEAD", oid);
ec1fcc16 2310 default:
259328b7 2311 return error(_("%s: unsupported file type"), path);
ec1fcc16 2312 }
ea8e0297 2313 return rc;
ec1fcc16 2314}
a69e5429
JH
2315
2316int read_pack_header(int fd, struct pack_header *header)
2317{
f48ecd38 2318 if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header))
c697ad14
HO
2319 /* "eof before pack header was fully read" */
2320 return PH_ERROR_EOF;
2321
a69e5429
JH
2322 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2323 /* "protocol error (pack signature mismatch detected)" */
2324 return PH_ERROR_PACK_SIGNATURE;
2325 if (!pack_version_ok(header->hdr_version))
2326 /* "protocol error (pack version unsupported)" */
2327 return PH_ERROR_PROTOCOL;
2328 return 0;
2329}
40d52ff7 2330
e816caa0 2331void assert_oid_type(const struct object_id *oid, enum object_type expect)
40d52ff7 2332{
0df8e965 2333 enum object_type type = oid_object_info(the_repository, oid, NULL);
40d52ff7 2334 if (type < 0)
259328b7 2335 die(_("%s is not a valid object"), oid_to_hex(oid));
40d52ff7 2336 if (type != expect)
259328b7 2337 die(_("%s is not a valid '%s' object"), oid_to_hex(oid),
debca9d2 2338 type_name(expect));
40d52ff7 2339}
27e1e22d 2340
70c49050 2341int for_each_file_in_obj_subdir(unsigned int subdir_nr,
cc817ca3
RS
2342 struct strbuf *path,
2343 each_loose_object_fn obj_cb,
2344 each_loose_cruft_fn cruft_cb,
2345 each_loose_subdir_fn subdir_cb,
2346 void *data)
27e1e22d 2347{
0375f472
RS
2348 size_t origlen, baselen;
2349 DIR *dir;
27e1e22d
JK
2350 struct dirent *de;
2351 int r = 0;
62a24c89 2352 struct object_id oid;
27e1e22d 2353
70c49050
RS
2354 if (subdir_nr > 0xff)
2355 BUG("invalid loose object subdirectory: %x", subdir_nr);
2356
0375f472
RS
2357 origlen = path->len;
2358 strbuf_complete(path, '/');
2359 strbuf_addf(path, "%02x", subdir_nr);
0375f472
RS
2360
2361 dir = opendir(path->buf);
27e1e22d 2362 if (!dir) {
0375f472 2363 if (errno != ENOENT)
259328b7 2364 r = error_errno(_("unable to open %s"), path->buf);
0375f472
RS
2365 strbuf_setlen(path, origlen);
2366 return r;
27e1e22d
JK
2367 }
2368
62a24c89 2369 oid.hash[0] = subdir_nr;
163ee5e6
DS
2370 strbuf_addch(path, '/');
2371 baselen = path->len;
62a24c89 2372
b548f0f1 2373 while ((de = readdir_skip_dot_and_dotdot(dir))) {
163ee5e6 2374 size_t namelen;
27e1e22d 2375
163ee5e6 2376 namelen = strlen(de->d_name);
27e1e22d 2377 strbuf_setlen(path, baselen);
163ee5e6 2378 strbuf_add(path, de->d_name, namelen);
94b5e093 2379 if (namelen == the_hash_algo->hexsz - 2 &&
62a24c89 2380 !hex_to_bytes(oid.hash + 1, de->d_name,
94b5e093 2381 the_hash_algo->rawsz - 1)) {
5a6dce70 2382 oid_set_algo(&oid, the_hash_algo);
62a24c89
RS
2383 if (obj_cb) {
2384 r = obj_cb(&oid, path->buf, data);
2385 if (r)
2386 break;
27e1e22d 2387 }
62a24c89 2388 continue;
27e1e22d
JK
2389 }
2390
2391 if (cruft_cb) {
2392 r = cruft_cb(de->d_name, path->buf, data);
2393 if (r)
2394 break;
2395 }
2396 }
094c7e63 2397 closedir(dir);
27e1e22d 2398
163ee5e6 2399 strbuf_setlen(path, baselen - 1);
27e1e22d
JK
2400 if (!r && subdir_cb)
2401 r = subdir_cb(subdir_nr, path->buf, data);
2402
0375f472
RS
2403 strbuf_setlen(path, origlen);
2404
27e1e22d
JK
2405 return r;
2406}
2407
e6f875e0 2408int for_each_loose_file_in_objdir_buf(struct strbuf *path,
27e1e22d
JK
2409 each_loose_object_fn obj_cb,
2410 each_loose_cruft_fn cruft_cb,
2411 each_loose_subdir_fn subdir_cb,
2412 void *data)
2413{
27e1e22d
JK
2414 int r = 0;
2415 int i;
2416
27e1e22d 2417 for (i = 0; i < 256; i++) {
e6f875e0 2418 r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
27e1e22d 2419 subdir_cb, data);
27e1e22d
JK
2420 if (r)
2421 break;
2422 }
2423
e6f875e0
JK
2424 return r;
2425}
2426
2427int for_each_loose_file_in_objdir(const char *path,
2428 each_loose_object_fn obj_cb,
2429 each_loose_cruft_fn cruft_cb,
2430 each_loose_subdir_fn subdir_cb,
2431 void *data)
2432{
2433 struct strbuf buf = STRBUF_INIT;
2434 int r;
2435
2436 strbuf_addstr(&buf, path);
2437 r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb,
2438 subdir_cb, data);
27e1e22d 2439 strbuf_release(&buf);
e6f875e0 2440
27e1e22d
JK
2441 return r;
2442}
660c889e 2443
a7ff6f5a
JK
2444int for_each_loose_object(each_loose_object_fn cb, void *data,
2445 enum for_each_object_flags flags)
660c889e 2446{
f0eaf638 2447 struct object_directory *odb;
b0a42642 2448
f0eaf638
JK
2449 prepare_alt_odb(the_repository);
2450 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
2451 int r = for_each_loose_file_in_objdir(odb->path, cb, NULL,
2452 NULL, data);
2453 if (r)
2454 return r;
660c889e 2455
f0eaf638
JK
2456 if (flags & FOR_EACH_OBJECT_LOCAL_ONLY)
2457 break;
2458 }
1385bb7b 2459
f0eaf638 2460 return 0;
660c889e
JK
2461}
2462
3a2e0824
JK
2463static int append_loose_object(const struct object_id *oid, const char *path,
2464 void *data)
660c889e 2465{
92d8ed8a 2466 oidtree_insert(data, oid);
3a2e0824
JK
2467 return 0;
2468}
660c889e 2469
92d8ed8a 2470struct oidtree *odb_loose_cache(struct object_directory *odb,
0000d654
RS
2471 const struct object_id *oid)
2472{
2473 int subdir_nr = oid->hash[0];
3a2e0824 2474 struct strbuf buf = STRBUF_INIT;
33f379ee
EW
2475 size_t word_bits = bitsizeof(odb->loose_objects_subdir_seen[0]);
2476 size_t word_index = subdir_nr / word_bits;
2477 size_t mask = 1 << (subdir_nr % word_bits);
2478 uint32_t *bitmap;
660c889e 2479
3a2e0824 2480 if (subdir_nr < 0 ||
33f379ee 2481 subdir_nr >= bitsizeof(odb->loose_objects_subdir_seen))
3a2e0824 2482 BUG("subdir_nr out of range");
1385bb7b 2483
33f379ee
EW
2484 bitmap = &odb->loose_objects_subdir_seen[word_index];
2485 if (*bitmap & mask)
92d8ed8a
EW
2486 return odb->loose_objects_cache;
2487 if (!odb->loose_objects_cache) {
2488 ALLOC_ARRAY(odb->loose_objects_cache, 1);
2489 oidtree_init(odb->loose_objects_cache);
2490 }
3a2e0824
JK
2491 strbuf_addstr(&buf, odb->path);
2492 for_each_file_in_obj_subdir(subdir_nr, &buf,
2493 append_loose_object,
2494 NULL, NULL,
92d8ed8a 2495 odb->loose_objects_cache);
33f379ee 2496 *bitmap |= mask;
7317aa71 2497 strbuf_release(&buf);
92d8ed8a 2498 return odb->loose_objects_cache;
660c889e
JK
2499}
2500
d4e19e51
RS
2501void odb_clear_loose_cache(struct object_directory *odb)
2502{
92d8ed8a
EW
2503 oidtree_clear(odb->loose_objects_cache);
2504 FREE_AND_NULL(odb->loose_objects_cache);
d4e19e51
RS
2505 memset(&odb->loose_objects_subdir_seen, 0,
2506 sizeof(odb->loose_objects_subdir_seen));
2507}
2508
00a7760e
JK
2509static int check_stream_oid(git_zstream *stream,
2510 const char *hdr,
2511 unsigned long size,
2512 const char *path,
2513 const struct object_id *expected_oid)
f6371f92 2514{
18e2588e 2515 git_hash_ctx c;
00a7760e 2516 struct object_id real_oid;
f6371f92
JK
2517 unsigned char buf[4096];
2518 unsigned long total_read;
2519 int status = Z_OK;
2520
18e2588e 2521 the_hash_algo->init_fn(&c);
2522 the_hash_algo->update_fn(&c, hdr, stream->total_out);
f6371f92
JK
2523
2524 /*
2525 * We already read some bytes into hdr, but the ones up to the NUL
2526 * do not count against the object's content size.
2527 */
2528 total_read = stream->total_out - strlen(hdr) - 1;
2529
2530 /*
2531 * This size comparison must be "<=" to read the final zlib packets;
00a7760e 2532 * see the comment in unpack_loose_rest for details.
f6371f92
JK
2533 */
2534 while (total_read <= size &&
18ad13e5
JH
2535 (status == Z_OK ||
2536 (status == Z_BUF_ERROR && !stream->avail_out))) {
f6371f92
JK
2537 stream->next_out = buf;
2538 stream->avail_out = sizeof(buf);
2539 if (size - total_read < stream->avail_out)
2540 stream->avail_out = size - total_read;
2541 status = git_inflate(stream, Z_FINISH);
18e2588e 2542 the_hash_algo->update_fn(&c, buf, stream->next_out - buf);
f6371f92
JK
2543 total_read += stream->next_out - buf;
2544 }
2545 git_inflate_end(stream);
2546
2547 if (status != Z_STREAM_END) {
00a7760e 2548 error(_("corrupt loose object '%s'"), oid_to_hex(expected_oid));
f6371f92
JK
2549 return -1;
2550 }
cce044df 2551 if (stream->avail_in) {
259328b7 2552 error(_("garbage at end of loose object '%s'"),
00a7760e 2553 oid_to_hex(expected_oid));
cce044df
JK
2554 return -1;
2555 }
f6371f92 2556
5951bf46 2557 the_hash_algo->final_oid_fn(&real_oid, &c);
00a7760e 2558 if (!oideq(expected_oid, &real_oid)) {
01f8d594 2559 error(_("hash mismatch for %s (expected %s)"), path,
00a7760e 2560 oid_to_hex(expected_oid));
f6371f92
JK
2561 return -1;
2562 }
2563
2564 return 0;
2565}
2566
2567int read_loose_object(const char *path,
d61d87bd 2568 const struct object_id *expected_oid,
f6371f92
JK
2569 enum object_type *type,
2570 unsigned long *size,
2571 void **contents)
2572{
2573 int ret = -1;
f6371f92
JK
2574 void *map = NULL;
2575 unsigned long mapsize;
2576 git_zstream stream;
1af64f73 2577 char hdr[MAX_HEADER_LEN];
f6371f92
JK
2578
2579 *contents = NULL;
2580
514c5fdd 2581 map = map_loose_object_1(the_repository, path, NULL, &mapsize);
f6371f92 2582 if (!map) {
259328b7 2583 error_errno(_("unable to mmap %s"), path);
f6371f92
JK
2584 goto out;
2585 }
2586
00a7760e 2587 if (unpack_loose_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) {
259328b7 2588 error(_("unable to unpack header of %s"), path);
f6371f92
JK
2589 goto out;
2590 }
2591
00a7760e 2592 *type = parse_loose_header(hdr, size);
f6371f92 2593 if (*type < 0) {
259328b7 2594 error(_("unable to parse header of %s"), path);
f6371f92
JK
2595 git_inflate_end(&stream);
2596 goto out;
2597 }
2598
7ac4f3a0 2599 if (*type == OBJ_BLOB && *size > big_file_threshold) {
00a7760e 2600 if (check_stream_oid(&stream, hdr, *size, path, expected_oid) < 0)
f6371f92
JK
2601 goto out;
2602 } else {
00a7760e 2603 *contents = unpack_loose_rest(&stream, hdr, *size, expected_oid);
f6371f92 2604 if (!*contents) {
259328b7 2605 error(_("unable to unpack contents of %s"), path);
f6371f92
JK
2606 git_inflate_end(&stream);
2607 goto out;
2608 }
b98d1885
MT
2609 if (check_object_signature(the_repository, expected_oid,
2610 *contents, *size,
2611 type_name(*type))) {
01f8d594 2612 error(_("hash mismatch for %s (expected %s)"), path,
d61d87bd 2613 oid_to_hex(expected_oid));
f6371f92
JK
2614 free(*contents);
2615 goto out;
2616 }
2617 }
2618
2619 ret = 0; /* everything checks out */
2620
2621out:
2622 if (map)
2623 munmap(map, mapsize);
f6371f92
JK
2624 return ret;
2625}