dir: convert struct sha1_stat to use object_id
[git/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"
090ea126 25#include "streaming.h"
543c5caa 26#include "dir.h"
a73cdd21 27#include "mru.h"
12d95ef6 28#include "list.h"
c4c6effa 29#include "mergesort.h"
cf3c6352 30#include "quote.h"
4f39cd82 31#include "packfile.h"
e05db0fd 32
50c5cd58 33const unsigned char null_sha1[GIT_MAX_RAWSZ];
3e56e724 34const struct object_id null_oid;
8576fde6
JK
35const struct object_id empty_tree_oid = {
36 EMPTY_TREE_SHA1_BIN_LITERAL
37};
38const struct object_id empty_blob_oid = {
39 EMPTY_BLOB_SHA1_BIN_LITERAL
40};
88cd621d 41
f50e766b 42static void git_hash_sha1_init(void *ctx)
43{
44 git_SHA1_Init((git_SHA_CTX *)ctx);
45}
46
47static void git_hash_sha1_update(void *ctx, const void *data, size_t len)
48{
49 git_SHA1_Update((git_SHA_CTX *)ctx, data, len);
50}
51
52static void git_hash_sha1_final(unsigned char *hash, void *ctx)
53{
54 git_SHA1_Final(hash, (git_SHA_CTX *)ctx);
55}
56
57static void git_hash_unknown_init(void *ctx)
58{
59 die("trying to init unknown hash");
60}
61
62static void git_hash_unknown_update(void *ctx, const void *data, size_t len)
63{
64 die("trying to update unknown hash");
65}
66
67static void git_hash_unknown_final(unsigned char *hash, void *ctx)
68{
69 die("trying to finalize unknown hash");
70}
71
72const struct git_hash_algo hash_algos[GIT_HASH_NALGOS] = {
73 {
74 NULL,
75 0x00000000,
76 0,
77 0,
78 0,
79 git_hash_unknown_init,
80 git_hash_unknown_update,
81 git_hash_unknown_final,
82 NULL,
83 NULL,
84 },
85 {
86 "sha-1",
87 /* "sha1", big-endian */
88 0x73686131,
89 sizeof(git_SHA_CTX),
90 GIT_SHA1_RAWSZ,
91 GIT_SHA1_HEXSZ,
92 git_hash_sha1_init,
93 git_hash_sha1_update,
94 git_hash_sha1_final,
95 &empty_tree_oid,
96 &empty_blob_oid,
97 },
98};
99
c597ba80
NTND
100/*
101 * This is meant to hold a *small* number of objects that you would
102 * want read_sha1_file() to be able to return, but yet you do not want
103 * to write them into the object store (e.g. a browse-only
104 * application).
105 */
106static struct cached_object {
107 unsigned char sha1[20];
108 enum object_type type;
109 void *buf;
110 unsigned long size;
111} *cached_objects;
112static int cached_object_nr, cached_object_alloc;
113
114static struct cached_object empty_tree = {
dab0d410 115 EMPTY_TREE_SHA1_BIN_LITERAL,
c597ba80
NTND
116 OBJ_TREE,
117 "",
118 0
119};
120
121static struct cached_object *find_cached_object(const unsigned char *sha1)
122{
123 int i;
124 struct cached_object *co = cached_objects;
125
126 for (i = 0; i < cached_object_nr; i++, co++) {
127 if (!hashcmp(co->sha1, sha1))
128 return co;
129 }
130 if (!hashcmp(sha1, empty_tree.sha1))
131 return &empty_tree;
132 return NULL;
133}
134
9472935d
TB
135
136static enum safe_crlf get_safe_crlf(unsigned flags)
137{
138 if (flags & HASH_RENORMALIZE)
139 return SAFE_CRLF_RENORMALIZE;
140 else if (flags & HASH_WRITE_OBJECT)
141 return safe_crlf;
142 else
143 return SAFE_CRLF_FALSE;
144}
145
146
90a6464b
JH
147int mkdir_in_gitdir(const char *path)
148{
149 if (mkdir(path, 0777)) {
150 int saved_errno = errno;
151 struct stat st;
152 struct strbuf sb = STRBUF_INIT;
153
154 if (errno != EEXIST)
155 return -1;
156 /*
157 * Are we looking at a path in a symlinked worktree
158 * whose original repository does not yet have it?
159 * e.g. .git/rr-cache pointing at its original
160 * repository in which the user hasn't performed any
161 * conflict resolution yet?
162 */
163 if (lstat(path, &st) || !S_ISLNK(st.st_mode) ||
164 strbuf_readlink(&sb, path, st.st_size) ||
165 !is_absolute_path(sb.buf) ||
166 mkdir(sb.buf, 0777)) {
167 strbuf_release(&sb);
168 errno = saved_errno;
169 return -1;
170 }
171 strbuf_release(&sb);
172 }
173 return adjust_shared_perm(path);
174}
175
0be0521b 176enum scld_error safe_create_leading_directories(char *path)
b2cb9425 177{
26c8ae2a 178 char *next_component = path + offset_1st_component(path);
0be0521b 179 enum scld_error ret = SCLD_OK;
67d42212 180
0be0521b 181 while (ret == SCLD_OK && next_component) {
f0502332 182 struct stat st;
0f527403 183 char *slash = next_component, slash_character;
f0502332 184
0f527403
MH
185 while (*slash && !is_dir_sep(*slash))
186 slash++;
187
188 if (!*slash)
b2cb9425 189 break;
bf10cf70 190
26c8ae2a 191 next_component = slash + 1;
0f527403 192 while (is_dir_sep(*next_component))
bf10cf70 193 next_component++;
26c8ae2a 194 if (!*next_component)
5f0bdf50 195 break;
831651fd 196
0f527403 197 slash_character = *slash;
831651fd 198 *slash = '\0';
67d42212
JR
199 if (!stat(path, &st)) {
200 /* path exists */
204a047f
MH
201 if (!S_ISDIR(st.st_mode)) {
202 errno = ENOTDIR;
0be0521b 203 ret = SCLD_EXISTS;
204a047f 204 }
53a39721 205 } else if (mkdir(path, 0777)) {
928734d9 206 if (errno == EEXIST &&
9e6f885d 207 !stat(path, &st) && S_ISDIR(st.st_mode))
928734d9 208 ; /* somebody created it since we checked */
18d37e86
MH
209 else if (errno == ENOENT)
210 /*
211 * Either mkdir() failed because
212 * somebody just pruned the containing
213 * directory, or stat() failed because
214 * the file that was in our way was
215 * just removed. Either way, inform
216 * the caller that it might be worth
217 * trying again:
218 */
219 ret = SCLD_VANISHED;
9e6f885d 220 else
0be0521b 221 ret = SCLD_FAILED;
53a39721 222 } else if (adjust_shared_perm(path)) {
0be0521b 223 ret = SCLD_PERMS;
457f06d6 224 }
0f527403 225 *slash = slash_character;
b2cb9425 226 }
9e6f885d 227 return ret;
b2cb9425 228}
723c31fe 229
0be0521b 230enum scld_error safe_create_leading_directories_const(const char *path)
8e21d63b 231{
02944307 232 int save_errno;
8e21d63b
JK
233 /* path points to cache entries, so xstrdup before messing with it */
234 char *buf = xstrdup(path);
0be0521b 235 enum scld_error result = safe_create_leading_directories(buf);
02944307
MH
236
237 save_errno = errno;
8e21d63b 238 free(buf);
02944307 239 errno = save_errno;
8e21d63b
JK
240 return result;
241}
242
177978f5
MH
243int raceproof_create_file(const char *path, create_file_fn fn, void *cb)
244{
245 /*
246 * The number of times we will try to remove empty directories
247 * in the way of path. This is only 1 because if another
248 * process is racily creating directories that conflict with
249 * us, we don't want to fight against them.
250 */
251 int remove_directories_remaining = 1;
252
253 /*
254 * The number of times that we will try to create the
255 * directories containing path. We are willing to attempt this
256 * more than once, because another process could be trying to
257 * clean up empty directories at the same time as we are
258 * trying to create them.
259 */
260 int create_directories_remaining = 3;
261
262 /* A scratch copy of path, filled lazily if we need it: */
263 struct strbuf path_copy = STRBUF_INIT;
264
265 int ret, save_errno;
266
267 /* Sanity check: */
268 assert(*path);
269
270retry_fn:
271 ret = fn(path, cb);
272 save_errno = errno;
273 if (!ret)
274 goto out;
275
276 if (errno == EISDIR && remove_directories_remaining-- > 0) {
277 /*
278 * A directory is in the way. Maybe it is empty; try
279 * to remove it:
280 */
281 if (!path_copy.len)
282 strbuf_addstr(&path_copy, path);
283
284 if (!remove_dir_recursively(&path_copy, REMOVE_DIR_EMPTY_ONLY))
285 goto retry_fn;
286 } else if (errno == ENOENT && create_directories_remaining-- > 0) {
287 /*
288 * Maybe the containing directory didn't exist, or
289 * maybe it was just deleted by a process that is
290 * racing with us to clean up empty directories. Try
291 * to create it:
292 */
293 enum scld_error scld_result;
294
295 if (!path_copy.len)
296 strbuf_addstr(&path_copy, path);
297
298 do {
299 scld_result = safe_create_leading_directories(path_copy.buf);
300 if (scld_result == SCLD_OK)
301 goto retry_fn;
302 } while (scld_result == SCLD_VANISHED && create_directories_remaining-- > 0);
303 }
304
305out:
306 strbuf_release(&path_copy);
307 errno = save_errno;
308 return ret;
309}
310
f7b7774f 311static void fill_sha1_path(struct strbuf *buf, const unsigned char *sha1)
ace1534d
JH
312{
313 int i;
314 for (i = 0; i < 20; i++) {
315 static char hex[] = "0123456789abcdef";
316 unsigned int val = sha1[i];
f7b7774f
JK
317 strbuf_addch(buf, hex[val >> 4]);
318 strbuf_addch(buf, hex[val & 0xf]);
afbba2f0 319 if (!i)
f7b7774f 320 strbuf_addch(buf, '/');
ace1534d
JH
321 }
322}
323
30d6c6ea 324const char *sha1_file_name(const unsigned char *sha1)
0fcfd160 325{
f7b7774f 326 static struct strbuf buf = STRBUF_INIT;
0fcfd160 327
f7b7774f
JK
328 strbuf_reset(&buf);
329 strbuf_addf(&buf, "%s/", get_object_directory());
560fb6a1 330
f7b7774f
JK
331 fill_sha1_path(&buf, sha1);
332 return buf.buf;
0fcfd160
LT
333}
334
38dbe5f0
JK
335struct strbuf *alt_scratch_buf(struct alternate_object_database *alt)
336{
337 strbuf_setlen(&alt->scratch, alt->base_len);
338 return &alt->scratch;
339}
340
29ec6af2
JK
341static const char *alt_sha1_path(struct alternate_object_database *alt,
342 const unsigned char *sha1)
343{
38dbe5f0 344 struct strbuf *buf = alt_scratch_buf(alt);
f7b7774f 345 fill_sha1_path(buf, sha1);
38dbe5f0 346 return buf->buf;
0fcfd160
LT
347}
348
d5a63b99
JH
349struct alternate_object_database *alt_odb_list;
350static struct alternate_object_database **alt_odb_tail;
ace1534d 351
4ea82473
JK
352/*
353 * Return non-zero iff the path is usable as an alternate object database.
354 */
355static int alt_odb_usable(struct strbuf *path, const char *normalized_objdir)
356{
357 struct alternate_object_database *alt;
358
359 /* Detect cases where alternate disappeared */
360 if (!is_directory(path->buf)) {
361 error("object directory %s does not exist; "
362 "check .git/objects/info/alternates.",
363 path->buf);
364 return 0;
365 }
366
367 /*
368 * Prevent the common mistake of listing the same
369 * thing twice, or object directory itself.
370 */
371 for (alt = alt_odb_list; alt; alt = alt->next) {
ea0fc3b4 372 if (!fspathcmp(path->buf, alt->path))
4ea82473
JK
373 return 0;
374 }
375 if (!fspathcmp(path->buf, normalized_objdir))
376 return 0;
377
378 return 1;
379}
380
ddd5d056
JH
381/*
382 * Prepare alternate object database registry.
d5a63b99
JH
383 *
384 * The variable alt_odb_list points at the list of struct
385 * alternate_object_database. The elements on this list come from
386 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
387 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
388 * whose contents is similar to that environment variable but can be
389 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
390 * contains "/the/directory/corresponding/to/.git/objects/...", while
391 * its name points just after the slash at the end of ".git/objects/"
392 * in the example above, and has enough space to hold 40-byte hex
393 * SHA1, an extra slash for the first level indirection, and the
394 * terminating NUL.
ddd5d056 395 */
2456990d 396static void read_info_alternates(const char * relative_base, int depth);
539e7506
EK
397static int link_alt_odb_entry(const char *entry, const char *relative_base,
398 int depth, const char *normalized_objdir)
ace1534d 399{
c2f493a4 400 struct alternate_object_database *ent;
5bdf0a84 401 struct strbuf pathbuf = STRBUF_INIT;
d5a63b99 402
85dadc38 403 if (!is_absolute_path(entry) && relative_base) {
4ac9006f 404 strbuf_realpath(&pathbuf, relative_base, 1);
5bdf0a84 405 strbuf_addch(&pathbuf, '/');
c2f493a4 406 }
6eac50d8 407 strbuf_addstr(&pathbuf, entry);
c2f493a4 408
37a95862 409 if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) {
670c359d
JK
410 error("unable to normalize alternate object path: %s",
411 pathbuf.buf);
412 strbuf_release(&pathbuf);
413 return -1;
414 }
5bdf0a84
HW
415
416 /*
417 * The trailing slash after the directory name is given by
418 * this function at the end. Remove duplicates.
419 */
4ea82473
JK
420 while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/')
421 strbuf_setlen(&pathbuf, pathbuf.len - 1);
5bdf0a84 422
4ea82473
JK
423 if (!alt_odb_usable(&pathbuf, normalized_objdir)) {
424 strbuf_release(&pathbuf);
c2f493a4
MW
425 return -1;
426 }
427
7f0fa2c0 428 ent = alloc_alt_odb(pathbuf.buf);
c2f493a4
MW
429
430 /* add the alternate entry */
431 *alt_odb_tail = ent;
432 alt_odb_tail = &(ent->next);
433 ent->next = NULL;
434
435 /* recursively add alternates */
4ea82473 436 read_info_alternates(pathbuf.buf, depth + 1);
c2f493a4 437
4ea82473 438 strbuf_release(&pathbuf);
c2f493a4
MW
439 return 0;
440}
441
cf3c6352
JK
442static const char *parse_alt_odb_entry(const char *string,
443 int sep,
444 struct strbuf *out)
445{
446 const char *end;
447
448 strbuf_reset(out);
449
450 if (*string == '#') {
451 /* comment; consume up to next separator */
452 end = strchrnul(string, sep);
453 } else if (*string == '"' && !unquote_c_style(out, string, &end)) {
454 /*
455 * quoted path; unquote_c_style has copied the
456 * data for us and set "end". Broken quoting (e.g.,
457 * an entry that doesn't end with a quote) falls
458 * back to the unquoted case below.
459 */
460 } else {
461 /* normal, unquoted path */
462 end = strchrnul(string, sep);
463 strbuf_add(out, string, end - string);
464 }
465
466 if (*end)
467 end++;
468 return end;
469}
470
dc732bd5 471static void link_alt_odb_entries(const char *alt, int sep,
c2f493a4
MW
472 const char *relative_base, int depth)
473{
539e7506 474 struct strbuf objdirbuf = STRBUF_INIT;
cf3c6352 475 struct strbuf entry = STRBUF_INIT;
c2f493a4 476
f28e3668
JK
477 if (!alt || !*alt)
478 return;
479
c2f493a4
MW
480 if (depth > 5) {
481 error("%s: ignoring alternate object stores, nesting too deep.",
482 relative_base);
483 return;
484 }
485
f655651e 486 strbuf_add_absolute_path(&objdirbuf, get_object_directory());
670c359d
JK
487 if (strbuf_normalize_path(&objdirbuf) < 0)
488 die("unable to normalize object directory: %s",
489 objdirbuf.buf);
539e7506 490
cf3c6352
JK
491 while (*alt) {
492 alt = parse_alt_odb_entry(alt, sep, &entry);
493 if (!entry.len)
9577e7e3 494 continue;
cf3c6352 495 link_alt_odb_entry(entry.buf, relative_base, depth, objdirbuf.buf);
9577e7e3 496 }
cf3c6352 497 strbuf_release(&entry);
539e7506 498 strbuf_release(&objdirbuf);
d5a63b99
JH
499}
500
2456990d 501static void read_info_alternates(const char * relative_base, int depth)
d5a63b99 502{
5015f01c 503 char *path;
dc732bd5 504 struct strbuf buf = STRBUF_INIT;
d5a63b99 505
5015f01c 506 path = xstrfmt("%s/info/alternates", relative_base);
dc732bd5 507 if (strbuf_read_file(&buf, path, 1024) < 0) {
f0f7bebe 508 warn_on_fopen_errors(path);
dc732bd5 509 free(path);
9a217f2a 510 return;
ace1534d 511 }
d5a63b99 512
dc732bd5
JK
513 link_alt_odb_entries(buf.buf, '\n', relative_base, depth);
514 strbuf_release(&buf);
515 free(path);
ace1534d
JH
516}
517
7f0fa2c0
JK
518struct alternate_object_database *alloc_alt_odb(const char *dir)
519{
520 struct alternate_object_database *ent;
7f0fa2c0 521
597f9134 522 FLEX_ALLOC_STR(ent, path, dir);
38dbe5f0
JK
523 strbuf_init(&ent->scratch, 0);
524 strbuf_addf(&ent->scratch, "%s/", dir);
525 ent->base_len = ent->scratch.len;
7f0fa2c0
JK
526
527 return ent;
528}
529
bef70b22
DB
530void add_to_alternates_file(const char *reference)
531{
f132a127 532 struct lock_file lock = LOCK_INIT;
77b9b1d1
JK
533 char *alts = git_pathdup("objects/info/alternates");
534 FILE *in, *out;
f132a127 535 int found = 0;
77b9b1d1 536
f132a127
537 hold_lock_file_for_update(&lock, alts, LOCK_DIE_ON_ERROR);
538 out = fdopen_lock_file(&lock, "w");
77b9b1d1
JK
539 if (!out)
540 die_errno("unable to fdopen alternates lockfile");
541
542 in = fopen(alts, "r");
543 if (in) {
544 struct strbuf line = STRBUF_INIT;
77b9b1d1 545
3f163962 546 while (strbuf_getline(&line, in) != EOF) {
77b9b1d1
JK
547 if (!strcmp(reference, line.buf)) {
548 found = 1;
549 break;
550 }
551 fprintf_or_die(out, "%s\n", line.buf);
552 }
553
554 strbuf_release(&line);
555 fclose(in);
77b9b1d1
JK
556 }
557 else if (errno != ENOENT)
558 die_errno("unable to read alternates file");
559
f132a127
560 if (found) {
561 rollback_lock_file(&lock);
562 } else {
77b9b1d1 563 fprintf_or_die(out, "%s\n", reference);
f132a127 564 if (commit_lock_file(&lock))
77b9b1d1
JK
565 die_errno("unable to move new alternates file into place");
566 if (alt_odb_tail)
dc732bd5 567 link_alt_odb_entries(reference, '\n', NULL, 0);
77b9b1d1
JK
568 }
569 free(alts);
bef70b22
DB
570}
571
a5b34d21
JK
572void add_to_alternates_memory(const char *reference)
573{
574 /*
575 * Make sure alternates are initialized, or else our entry may be
576 * overwritten when they are.
577 */
578 prepare_alt_odb();
579
dc732bd5 580 link_alt_odb_entries(reference, '\n', NULL, 0);
a5b34d21
JK
581}
582
9eeea7d2
SB
583/*
584 * Compute the exact path an alternate is at and returns it. In case of
585 * error NULL is returned and the human readable error is added to `err`
586 * `path` may be relative and should point to $GITDIR.
587 * `err` must not be null.
588 */
589char *compute_alternate_path(const char *path, struct strbuf *err)
590{
591 char *ref_git = NULL;
592 const char *repo, *ref_git_s;
593 int seen_error = 0;
594
595 ref_git_s = real_path_if_valid(path);
596 if (!ref_git_s) {
597 seen_error = 1;
598 strbuf_addf(err, _("path '%s' does not exist"), path);
599 goto out;
600 } else
601 /*
602 * Beware: read_gitfile(), real_path() and mkpath()
603 * return static buffer
604 */
605 ref_git = xstrdup(ref_git_s);
606
607 repo = read_gitfile(ref_git);
608 if (!repo)
609 repo = read_gitfile(mkpath("%s/.git", ref_git));
610 if (repo) {
611 free(ref_git);
612 ref_git = xstrdup(repo);
613 }
614
615 if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) {
616 char *ref_git_git = mkpathdup("%s/.git", ref_git);
617 free(ref_git);
618 ref_git = ref_git_git;
619 } else if (!is_directory(mkpath("%s/objects", ref_git))) {
620 struct strbuf sb = STRBUF_INIT;
621 seen_error = 1;
622 if (get_common_dir(&sb, ref_git)) {
623 strbuf_addf(err,
624 _("reference repository '%s' as a linked "
625 "checkout is not supported yet."),
626 path);
627 goto out;
628 }
629
630 strbuf_addf(err, _("reference repository '%s' is not a "
631 "local repository."), path);
632 goto out;
633 }
634
635 if (!access(mkpath("%s/shallow", ref_git), F_OK)) {
636 strbuf_addf(err, _("reference repository '%s' is shallow"),
637 path);
638 seen_error = 1;
639 goto out;
640 }
641
642 if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) {
643 strbuf_addf(err,
644 _("reference repository '%s' is grafted"),
645 path);
646 seen_error = 1;
647 goto out;
648 }
649
650out:
651 if (seen_error) {
6a83d902 652 FREE_AND_NULL(ref_git);
9eeea7d2
SB
653 }
654
655 return ref_git;
656}
657
fe1b2268 658int foreach_alt_odb(alt_odb_fn fn, void *cb)
d79796bc
JH
659{
660 struct alternate_object_database *ent;
fe1b2268 661 int r = 0;
d79796bc
JH
662
663 prepare_alt_odb();
fe1b2268
JK
664 for (ent = alt_odb_list; ent; ent = ent->next) {
665 r = fn(ent, cb);
666 if (r)
667 break;
668 }
669 return r;
d79796bc
JH
670}
671
c2f493a4
MW
672void prepare_alt_odb(void)
673{
554fe20d 674 const char *alt;
c2f493a4 675
7dc24aa5
SP
676 if (alt_odb_tail)
677 return;
678
c2f493a4 679 alt = getenv(ALTERNATE_DB_ENVIRONMENT);
c2f493a4 680
c2f493a4 681 alt_odb_tail = &alt_odb_list;
dc732bd5 682 link_alt_odb_entries(alt, PATH_SEP, NULL, 0);
c2f493a4
MW
683
684 read_info_alternates(get_object_directory(), 0);
685}
686
3096b2ec 687/* Returns 1 if we have successfully freshened the file, 0 otherwise. */
33d4221c 688static int freshen_file(const char *fn)
ace1534d 689{
33d4221c
JK
690 struct utimbuf t;
691 t.actime = t.modtime = time(NULL);
692 return !utime(fn, &t);
0f4dc14a 693}
ace1534d 694
3096b2ec
JK
695/*
696 * All of the check_and_freshen functions return 1 if the file exists and was
697 * freshened (if freshening was requested), 0 otherwise. If they return
698 * 0, you should not assume that it is safe to skip a write of the object (it
699 * either does not exist on disk, or has a stale mtime and may be subject to
700 * pruning).
701 */
6a5e6f5e 702int check_and_freshen_file(const char *fn, int freshen)
33d4221c
JK
703{
704 if (access(fn, F_OK))
705 return 0;
3096b2ec 706 if (freshen && !freshen_file(fn))
33d4221c
JK
707 return 0;
708 return 1;
709}
710
711static int check_and_freshen_local(const unsigned char *sha1, int freshen)
712{
713 return check_and_freshen_file(sha1_file_name(sha1), freshen);
714}
715
716static int check_and_freshen_nonlocal(const unsigned char *sha1, int freshen)
0f4dc14a
BC
717{
718 struct alternate_object_database *alt;
9a217f2a 719 prepare_alt_odb();
d5a63b99 720 for (alt = alt_odb_list; alt; alt = alt->next) {
29ec6af2
JK
721 const char *path = alt_sha1_path(alt, sha1);
722 if (check_and_freshen_file(path, freshen))
c529d75a 723 return 1;
ace1534d 724 }
c529d75a 725 return 0;
ace1534d
JH
726}
727
33d4221c
JK
728static int check_and_freshen(const unsigned char *sha1, int freshen)
729{
730 return check_and_freshen_local(sha1, freshen) ||
731 check_and_freshen_nonlocal(sha1, freshen);
732}
733
734int has_loose_object_nonlocal(const unsigned char *sha1)
735{
736 return check_and_freshen_nonlocal(sha1, 0);
737}
738
0f4dc14a
BC
739static int has_loose_object(const unsigned char *sha1)
740{
33d4221c 741 return check_and_freshen(sha1, 0);
0f4dc14a
BC
742}
743
02710228
SP
744static void mmap_limit_check(size_t length)
745{
746 static size_t limit = 0;
747 if (!limit) {
748 limit = git_env_ulong("GIT_MMAP_LIMIT", 0);
749 if (!limit)
750 limit = SIZE_MAX;
751 }
752 if (length > limit)
753 die("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX,
754 (uintmax_t)length, (uintmax_t)limit);
755}
756
1570856b
JK
757void *xmmap_gently(void *start, size_t length,
758 int prot, int flags, int fd, off_t offset)
58ecbd5e 759{
02710228
SP
760 void *ret;
761
762 mmap_limit_check(length);
763 ret = mmap(start, length, prot, flags, fd, offset);
58ecbd5e
JN
764 if (ret == MAP_FAILED) {
765 if (!length)
766 return NULL;
7c3ecb32 767 release_pack_memory(length);
58ecbd5e 768 ret = mmap(start, length, prot, flags, fd, offset);
58ecbd5e
JN
769 }
770 return ret;
771}
772
1570856b
JK
773void *xmmap(void *start, size_t length,
774 int prot, int flags, int fd, off_t offset)
775{
776 void *ret = xmmap_gently(start, length, prot, flags, fd, offset);
777 if (ret == MAP_FAILED)
9ca0aaf6 778 die_errno("mmap failed");
1570856b
JK
779 return ret;
780}
781
88d0db55 782/*
090ea126
NTND
783 * With an in-core object data in "map", rehash it to make sure the
784 * object name actually matches "sha1" to detect object corruption.
785 * With "map" == NULL, try reading the object named with "sha1" using
786 * the streaming interface and rehash it to do the same.
88d0db55 787 */
090ea126
NTND
788int check_sha1_signature(const unsigned char *sha1, void *map,
789 unsigned long size, const char *type)
88d0db55 790{
0fcfd160 791 unsigned char real_sha1[20];
090ea126
NTND
792 enum object_type obj_type;
793 struct git_istream *st;
794 git_SHA_CTX c;
795 char hdr[32];
796 int hdrlen;
88d0db55 797
090ea126
NTND
798 if (map) {
799 hash_sha1_file(map, size, type, real_sha1);
800 return hashcmp(sha1, real_sha1) ? -1 : 0;
88d0db55
BC
801 }
802
090ea126
NTND
803 st = open_istream(sha1, &obj_type, &size, NULL);
804 if (!st)
805 return -1;
88d0db55 806
090ea126 807 /* Generate the header */
ef1286d3 808 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %lu", typename(obj_type), size) + 1;
88d0db55 809
090ea126
NTND
810 /* Sha1.. */
811 git_SHA1_Init(&c);
812 git_SHA1_Update(&c, hdr, hdrlen);
813 for (;;) {
814 char buf[1024 * 16];
815 ssize_t readlen = read_istream(st, buf, sizeof(buf));
1f688557 816
f54fac53
JK
817 if (readlen < 0) {
818 close_istream(st);
819 return -1;
820 }
090ea126
NTND
821 if (!readlen)
822 break;
823 git_SHA1_Update(&c, buf, readlen);
fa5fc15d 824 }
090ea126
NTND
825 git_SHA1_Final(real_sha1, &c);
826 close_istream(st);
a89fccd2 827 return hashcmp(sha1, real_sha1) ? -1 : 0;
fa5fc15d
SP
828}
829
1b8ac5ea 830int git_open_cloexec(const char *name, int flags)
a0788266 831{
1e3001a8
JH
832 int fd;
833 static int o_cloexec = O_CLOEXEC;
a0788266 834
1e3001a8
JH
835 fd = open(name, flags | o_cloexec);
836 if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) {
cd66ada0 837 /* Try again w/o O_CLOEXEC: the kernel might not support it */
1e3001a8
JH
838 o_cloexec &= ~O_CLOEXEC;
839 fd = open(name, flags | o_cloexec);
491a8dec 840 }
491a8dec 841
9fb9495d 842#if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
491a8dec 843 {
1e3001a8 844 static int fd_cloexec = FD_CLOEXEC;
a0788266 845
1e3001a8
JH
846 if (!o_cloexec && 0 <= fd && fd_cloexec) {
847 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
9fb9495d
EW
848 int flags = fcntl(fd, F_GETFD);
849 if (fcntl(fd, F_SETFD, flags | fd_cloexec))
1e3001a8 850 fd_cloexec = 0;
cd66ada0 851 }
44d1c19e 852 }
a0788266 853#endif
1b8ac5ea 854 return fd;
a0788266
JS
855}
856
3cf8b462 857/*
771e7d57
JK
858 * Find "sha1" as a loose object in the local repository or in an alternate.
859 * Returns 0 on success, negative on failure.
860 *
861 * The "path" out-parameter will give the path of the object we found (if any).
862 * Note that it may point to static storage and is only valid until another
863 * call to sha1_file_name(), etc.
3cf8b462 864 */
771e7d57
JK
865static int stat_sha1_file(const unsigned char *sha1, struct stat *st,
866 const char **path)
1f688557 867{
052fe5ea 868 struct alternate_object_database *alt;
d079837e 869
771e7d57
JK
870 *path = sha1_file_name(sha1);
871 if (!lstat(*path, st))
052fe5ea 872 return 0;
c7934306 873
052fe5ea
JK
874 prepare_alt_odb();
875 errno = ENOENT;
876 for (alt = alt_odb_list; alt; alt = alt->next) {
771e7d57
JK
877 *path = alt_sha1_path(alt, sha1);
878 if (!lstat(*path, st))
052fe5ea 879 return 0;
c7934306
SP
880 }
881
3cf8b462
SP
882 return -1;
883}
884
771e7d57
JK
885/*
886 * Like stat_sha1_file(), but actually open the object and return the
887 * descriptor. See the caveats on the "path" parameter above.
888 */
889static int open_sha1_file(const unsigned char *sha1, const char **path)
60bb8b14 890{
44d1c19e 891 int fd;
44d1c19e 892 struct alternate_object_database *alt;
d6c8a05b 893 int most_interesting_errno;
60bb8b14 894
771e7d57
JK
895 *path = sha1_file_name(sha1);
896 fd = git_open(*path);
44d1c19e
LT
897 if (fd >= 0)
898 return fd;
d6c8a05b 899 most_interesting_errno = errno;
03e79c88 900
44d1c19e 901 prepare_alt_odb();
44d1c19e 902 for (alt = alt_odb_list; alt; alt = alt->next) {
771e7d57
JK
903 *path = alt_sha1_path(alt, sha1);
904 fd = git_open(*path);
44d1c19e
LT
905 if (fd >= 0)
906 return fd;
d6c8a05b
JK
907 if (most_interesting_errno == ENOENT)
908 most_interesting_errno = errno;
03e79c88 909 }
d6c8a05b 910 errno = most_interesting_errno;
44d1c19e 911 return -1;
1f688557
JH
912}
913
f6371f92
JK
914/*
915 * Map the loose object at "path" if it is not NULL, or the path found by
916 * searching for a loose object named "sha1".
917 */
918static void *map_sha1_file_1(const char *path,
919 const unsigned char *sha1,
920 unsigned long *size)
27d69a46 921{
0fcfd160 922 void *map;
144bde78 923 int fd;
84dd81c1 924
f6371f92
JK
925 if (path)
926 fd = git_open(path);
8261e1f1 927 else
f6371f92 928 fd = open_sha1_file(sha1, &path);
44d1c19e
LT
929 map = NULL;
930 if (fd >= 0) {
931 struct stat st;
1f688557 932
44d1c19e
LT
933 if (!fstat(fd, &st)) {
934 *size = xsize_t(st.st_size);
33e42de0
MM
935 if (!*size) {
936 /* mmap() is forbidden on empty files */
771e7d57 937 error("object file %s is empty", path);
33e42de0
MM
938 return NULL;
939 }
44d1c19e 940 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
144bde78 941 }
44d1c19e 942 close(fd);
74e34e1f 943 }
0fcfd160 944 return map;
74e34e1f
NP
945}
946
f6371f92 947void *map_sha1_file(const unsigned char *sha1, unsigned long *size)
068f85e3 948{
f6371f92 949 return map_sha1_file_1(NULL, sha1, size);
068f85e3 950}
951
d21f8426
JH
952static int unpack_sha1_short_header(git_zstream *stream,
953 unsigned char *map, unsigned long mapsize,
954 void *buffer, unsigned long bufsiz)
47fe3f6e 955{
c4483576
LT
956 /* Get the data stream */
957 memset(stream, 0, sizeof(*stream));
958 stream->next_in = map;
959 stream->avail_in = mapsize;
960 stream->next_out = buffer;
93821bd9 961 stream->avail_out = bufsiz;
47fe3f6e 962
39c68542 963 git_inflate_init(stream);
cc5c54e7 964 return git_inflate(stream, 0);
9a217f2a
JH
965}
966
d21f8426
JH
967int unpack_sha1_header(git_zstream *stream,
968 unsigned char *map, unsigned long mapsize,
969 void *buffer, unsigned long bufsiz)
1f688557 970{
d21f8426
JH
971 int status = unpack_sha1_short_header(stream, map, mapsize,
972 buffer, bufsiz);
42873078 973
d21f8426
JH
974 if (status < Z_OK)
975 return status;
628522ec 976
d21f8426
JH
977 /* Make sure we have the terminating NUL */
978 if (!memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
979 return -1;
1f688557
JH
980 return 0;
981}
982
46f03448
KN
983static int unpack_sha1_header_to_strbuf(git_zstream *stream, unsigned char *map,
984 unsigned long mapsize, void *buffer,
985 unsigned long bufsiz, struct strbuf *header)
d131b7af 986{
46f03448
KN
987 int status;
988
d21f8426
JH
989 status = unpack_sha1_short_header(stream, map, mapsize, buffer, bufsiz);
990 if (status < Z_OK)
991 return -1;
d131b7af 992
46f03448
KN
993 /*
994 * Check if entire header is unpacked in the first iteration.
d131b7af 995 */
46f03448
KN
996 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
997 return 0;
d131b7af 998
46f03448
KN
999 /*
1000 * buffer[0..bufsiz] was not large enough. Copy the partial
1001 * result out to header, and then append the result of further
1002 * reading the stream.
1003 */
1004 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1005 stream->next_out = buffer;
1006 stream->avail_out = bufsiz;
d131b7af 1007
46f03448
KN
1008 do {
1009 status = git_inflate(stream, 0);
1010 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
1011 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
1012 return 0;
1013 stream->next_out = buffer;
1014 stream->avail_out = bufsiz;
1015 } while (status != Z_STREAM_END);
1016 return -1;
d131b7af
SP
1017}
1018
ef49a7a0 1019static void *unpack_sha1_rest(git_zstream *stream, void *buffer, unsigned long size, const unsigned char *sha1)
95099731 1020{
5180cacc 1021 int bytes = strlen(buffer) + 1;
3aee68aa 1022 unsigned char *buf = xmallocz(size);
93821bd9 1023 unsigned long n;
7efbff75 1024 int status = Z_OK;
95099731 1025
93821bd9
LT
1026 n = stream->total_out - bytes;
1027 if (n > size)
1028 n = size;
1029 memcpy(buf, (char *) buffer + bytes, n);
1030 bytes = n;
456cdf6e
LT
1031 if (bytes <= size) {
1032 /*
1033 * The above condition must be (bytes <= size), not
1034 * (bytes < size). In other words, even though we
ccf5ace0 1035 * expect no more output and set avail_out to zero,
456cdf6e
LT
1036 * the input zlib stream may have bytes that express
1037 * "this concludes the stream", and we *do* want to
1038 * eat that input.
1039 *
1040 * Otherwise we would not be able to test that we
1041 * consumed all the input to reach the expected size;
1042 * we also want to check that zlib tells us that all
1043 * went well with status == Z_STREAM_END at the end.
1044 */
5180cacc
LT
1045 stream->next_out = buf + bytes;
1046 stream->avail_out = size - bytes;
7efbff75 1047 while (status == Z_OK)
39c68542 1048 status = git_inflate(stream, Z_FINISH);
5180cacc 1049 }
456cdf6e 1050 if (status == Z_STREAM_END && !stream->avail_in) {
39c68542 1051 git_inflate_end(stream);
7efbff75 1052 return buf;
95099731
NTND
1053 }
1054
7efbff75
JH
1055 if (status < 0)
1056 error("corrupt loose object '%s'", sha1_to_hex(sha1));
1057 else if (stream->avail_in)
1058 error("garbage at end of loose object '%s'",
1059 sha1_to_hex(sha1));
1060 free(buf);
1061 return NULL;
95099731
NTND
1062}
1063
d40d535b 1064/*
5180cacc
LT
1065 * We used to just use "sscanf()", but that's actually way
1066 * too permissive for what we want to check. So do an anal
1067 * object header parse by hand.
d40d535b 1068 */
46f03448
KN
1069static int parse_sha1_header_extended(const char *hdr, struct object_info *oi,
1070 unsigned int flags)
1f688557 1071{
46f03448 1072 const char *type_buf = hdr;
5180cacc 1073 unsigned long size;
46f03448 1074 int type, type_len = 0;
43057304 1075
5180cacc 1076 /*
46f03448 1077 * The type can be of any size but is followed by
21666f1a 1078 * a space.
5180cacc 1079 */
5180cacc
LT
1080 for (;;) {
1081 char c = *hdr++;
d21f8426
JH
1082 if (!c)
1083 return -1;
5180cacc
LT
1084 if (c == ' ')
1085 break;
46f03448 1086 type_len++;
5180cacc 1087 }
1f688557 1088
46f03448
KN
1089 type = type_from_string_gently(type_buf, type_len, 1);
1090 if (oi->typename)
1091 strbuf_add(oi->typename, type_buf, type_len);
1092 /*
1093 * Set type to 0 if its an unknown object and
2e3a16b2 1094 * we're obtaining the type using '--allow-unknown-type'
46f03448
KN
1095 * option.
1096 */
19fc5e84 1097 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0))
46f03448
KN
1098 type = 0;
1099 else if (type < 0)
1100 die("invalid object type");
1101 if (oi->typep)
1102 *oi->typep = type;
5180cacc
LT
1103
1104 /*
1105 * The length must follow immediately, and be in canonical
1106 * decimal format (ie "010" is not valid).
1107 */
1108 size = *hdr++ - '0';
1109 if (size > 9)
1110 return -1;
1111 if (size) {
1112 for (;;) {
1113 unsigned long c = *hdr - '0';
1114 if (c > 9)
1115 break;
1116 hdr++;
1117 size = size * 10 + c;
1b1005d1 1118 }
c01f51cc 1119 }
46f03448
KN
1120
1121 if (oi->sizep)
1122 *oi->sizep = size;
5180cacc
LT
1123
1124 /*
1125 * The length must be followed by a zero byte
1126 */
46f03448 1127 return *hdr ? -1 : type;
1f688557
JH
1128}
1129
46f03448 1130int parse_sha1_header(const char *hdr, unsigned long *sizep)
bf592c50 1131{
27b5c1a0 1132 struct object_info oi = OBJECT_INFO_INIT;
21666f1a 1133
46f03448 1134 oi.sizep = sizep;
1f0c0d36 1135 return parse_sha1_header_extended(hdr, &oi, 0);
bf592c50
DB
1136}
1137
052fe5ea 1138static int sha1_loose_object_info(const unsigned char *sha1,
46f03448
KN
1139 struct object_info *oi,
1140 int flags)
65c2e0c3 1141{
46f03448
KN
1142 int status = 0;
1143 unsigned long mapsize;
65c2e0c3 1144 void *map;
ef49a7a0 1145 git_zstream stream;
d65a16f6 1146 char hdr[32];
46f03448 1147 struct strbuf hdrbuf = STRBUF_INIT;
c84a1f3e 1148 unsigned long size_scratch;
65c2e0c3 1149
5d642e75
JK
1150 if (oi->delta_base_sha1)
1151 hashclr(oi->delta_base_sha1);
1152
052fe5ea
JK
1153 /*
1154 * If we don't care about type or size, then we don't
4ef8d1dd
JH
1155 * need to look inside the object at all. Note that we
1156 * do not optimize out the stat call, even if the
1157 * caller doesn't care about the disk-size, since our
1158 * return value implicitly indicates whether the
1159 * object even exists.
052fe5ea 1160 */
c84a1f3e 1161 if (!oi->typep && !oi->typename && !oi->sizep && !oi->contentp) {
771e7d57 1162 const char *path;
4ef8d1dd 1163 struct stat st;
771e7d57 1164 if (stat_sha1_file(sha1, &st, &path) < 0)
4ef8d1dd
JH
1165 return -1;
1166 if (oi->disk_sizep)
23c339c0 1167 *oi->disk_sizep = st.st_size;
052fe5ea
JK
1168 return 0;
1169 }
1170
bb6b8e4f 1171 map = map_sha1_file(sha1, &mapsize);
f0df4ed5 1172 if (!map)
dbea72a8 1173 return -1;
c84a1f3e
JT
1174
1175 if (!oi->sizep)
1176 oi->sizep = &size_scratch;
1177
23c339c0
JK
1178 if (oi->disk_sizep)
1179 *oi->disk_sizep = mapsize;
19fc5e84 1180 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) {
46f03448
KN
1181 if (unpack_sha1_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
1182 status = error("unable to unpack %s header with --allow-unknown-type",
1183 sha1_to_hex(sha1));
1184 } else if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
36e4d74a
JH
1185 status = error("unable to unpack %s header",
1186 sha1_to_hex(sha1));
46f03448
KN
1187 if (status < 0)
1188 ; /* Do nothing */
1189 else if (hdrbuf.len) {
1190 if ((status = parse_sha1_header_extended(hdrbuf.buf, oi, flags)) < 0)
1191 status = error("unable to parse %s header with --allow-unknown-type",
1192 sha1_to_hex(sha1));
1193 } else if ((status = parse_sha1_header_extended(hdr, oi, flags)) < 0)
36e4d74a 1194 status = error("unable to parse %s header", sha1_to_hex(sha1));
c84a1f3e 1195
b3ea7dd3 1196 if (status >= 0 && oi->contentp) {
c84a1f3e
JT
1197 *oi->contentp = unpack_sha1_rest(&stream, hdr,
1198 *oi->sizep, sha1);
b3ea7dd3
JK
1199 if (!*oi->contentp) {
1200 git_inflate_end(&stream);
1201 status = -1;
1202 }
1203 } else
c84a1f3e
JT
1204 git_inflate_end(&stream);
1205
65c2e0c3 1206 munmap(map, mapsize);
46f03448 1207 if (status && oi->typep)
23c339c0 1208 *oi->typep = status;
c84a1f3e
JT
1209 if (oi->sizep == &size_scratch)
1210 oi->sizep = NULL;
46f03448 1211 strbuf_release(&hdrbuf);
3ab0fb06 1212 oi->whence = OI_LOOSE;
93cff9a9 1213 return (status < 0) ? status : 0;
65c2e0c3
JH
1214}
1215
de7b5d62 1216int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi, unsigned flags)
f0df4ed5 1217{
cd585e2a 1218 static struct object_info blank_oi = OBJECT_INFO_INIT;
f0df4ed5 1219 struct pack_entry e;
5b086407 1220 int rtype;
1f0c0d36
JT
1221 const unsigned char *real = (flags & OBJECT_INFO_LOOKUP_REPLACE) ?
1222 lookup_replace_object(sha1) :
1223 sha1;
f0df4ed5 1224
87b5e236
JK
1225 if (is_null_sha1(real))
1226 return -1;
1227
cd585e2a
JT
1228 if (!oi)
1229 oi = &blank_oi;
1230
dfdd4afc
JT
1231 if (!(flags & OBJECT_INFO_SKIP_CACHED)) {
1232 struct cached_object *co = find_cached_object(real);
1233 if (co) {
1234 if (oi->typep)
1235 *(oi->typep) = co->type;
1236 if (oi->sizep)
1237 *(oi->sizep) = co->size;
1238 if (oi->disk_sizep)
1239 *(oi->disk_sizep) = 0;
1240 if (oi->delta_base_sha1)
1241 hashclr(oi->delta_base_sha1);
1242 if (oi->typename)
1243 strbuf_addstr(oi->typename, typename(co->type));
1244 if (oi->contentp)
1245 *oi->contentp = xmemdupz(co->buf, co->size);
1246 oi->whence = OI_CACHED;
1247 return 0;
1248 }
c4d9986f
NTND
1249 }
1250
1f7117ef 1251 if (!find_pack_entry(real, &e)) {
ddd63e64 1252 /* Most likely it's a loose object. */
3ab0fb06 1253 if (!sha1_loose_object_info(real, oi, flags))
5b086407 1254 return 0;
ddd63e64
SG
1255
1256 /* Not a loose object; someone else may have just packed it. */
dfdd4afc 1257 if (flags & OBJECT_INFO_QUICK) {
052fe5ea 1258 return -1;
dfdd4afc
JT
1259 } else {
1260 reprepare_packed_git();
1261 if (!find_pack_entry(real, &e))
1262 return -1;
1263 }
f0df4ed5 1264 }
3d77d877 1265
cd585e2a
JT
1266 if (oi == &blank_oi)
1267 /*
1268 * We know that the caller doesn't actually need the
1269 * information below, so return early.
1270 */
1271 return 0;
46f03448 1272
23c339c0 1273 rtype = packed_object_info(e.p, e.offset, oi);
412916ee 1274 if (rtype < 0) {
1f7117ef
CC
1275 mark_bad_packed_object(e.p, real);
1276 return sha1_object_info_extended(real, oi, 0);
3ab0fb06 1277 } else if (oi->whence == OI_PACKED) {
9a490590
JH
1278 oi->u.packed.offset = e.offset;
1279 oi->u.packed.pack = e.p;
1280 oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
1281 rtype == OBJ_OFS_DELTA);
3d77d877
NP
1282 }
1283
5b086407 1284 return 0;
f0df4ed5
JS
1285}
1286
3fc0dca9 1287/* returns enum object_type or negative */
9a490590
JH
1288int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
1289{
5b086407 1290 enum object_type type;
27b5c1a0 1291 struct object_info oi = OBJECT_INFO_INIT;
9a490590 1292
5b086407 1293 oi.typep = &type;
9a490590 1294 oi.sizep = sizep;
1f0c0d36
JT
1295 if (sha1_object_info_extended(sha1, &oi,
1296 OBJECT_INFO_LOOKUP_REPLACE) < 0)
5b086407
JK
1297 return -1;
1298 return type;
9a490590
JH
1299}
1300
f1d8130b
JT
1301static void *read_object(const unsigned char *sha1, enum object_type *type,
1302 unsigned long *size)
1303{
1304 struct object_info oi = OBJECT_INFO_INIT;
1305 void *content;
1306 oi.typep = type;
1307 oi.sizep = size;
1308 oi.contentp = &content;
1309
1310 if (sha1_object_info_extended(sha1, &oi, 0) < 0)
1311 return NULL;
1312 return content;
1313}
1314
829e5c3b
PO
1315int pretend_object_file(void *buf, unsigned long len, enum object_type type,
1316 struct object_id *oid)
d66b37bb
JH
1317{
1318 struct cached_object *co;
1319
829e5c3b
PO
1320 hash_sha1_file(buf, len, typename(type), oid->hash);
1321 if (has_sha1_file(oid->hash) || find_cached_object(oid->hash))
d66b37bb 1322 return 0;
c7353967 1323 ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc);
d66b37bb
JH
1324 co = &cached_objects[cached_object_nr++];
1325 co->size = len;
21666f1a 1326 co->type = type;
efa13f7b
JH
1327 co->buf = xmalloc(len);
1328 memcpy(co->buf, buf, len);
829e5c3b 1329 hashcpy(co->sha1, oid->hash);
d66b37bb
JH
1330 return 0;
1331}
1332
b6c4cecc
JH
1333/*
1334 * This function dies on corrupt objects; the callers who want to
1335 * deal with them should arrange to call read_object() and give error
1336 * messages themselves.
1337 */
5bf29b95
JH
1338void *read_sha1_file_extended(const unsigned char *sha1,
1339 enum object_type *type,
1340 unsigned long *size,
1f0c0d36 1341 int lookup_replace)
ac939109 1342{
3ba7a065 1343 void *data;
b6c4cecc 1344 const struct packed_git *p;
771e7d57
JK
1345 const char *path;
1346 struct stat st;
1f0c0d36
JT
1347 const unsigned char *repl = lookup_replace ? lookup_replace_object(sha1)
1348 : sha1;
b6c4cecc 1349
3ba7a065
JH
1350 errno = 0;
1351 data = read_object(repl, type, size);
4bbf5a26 1352 if (data)
b6c4cecc 1353 return data;
68095570 1354
25f3af3f 1355 if (errno && errno != ENOENT)
3ba7a065
JH
1356 die_errno("failed to read object %s", sha1_to_hex(sha1));
1357
68095570 1358 /* die if we replaced an object with one that does not exist */
b6c4cecc 1359 if (repl != sha1)
68095570
CC
1360 die("replacement %s not found for %s",
1361 sha1_to_hex(repl), sha1_to_hex(sha1));
1362
771e7d57 1363 if (!stat_sha1_file(repl, &st, &path))
b6c4cecc
JH
1364 die("loose object %s (stored in %s) is corrupt",
1365 sha1_to_hex(repl), path);
68095570 1366
b6c4cecc
JH
1367 if ((p = has_packed_and_bad(repl)) != NULL)
1368 die("packed object %s (stored in %s) is corrupt",
1369 sha1_to_hex(repl), p->pack_name);
f5552aee 1370
b6c4cecc 1371 return NULL;
ac939109
NP
1372}
1373
40469ee9 1374void *read_object_with_reference(const unsigned char *sha1,
21666f1a 1375 const char *required_type_name,
40469ee9
JH
1376 unsigned long *size,
1377 unsigned char *actual_sha1_return)
f4913f91 1378{
21666f1a 1379 enum object_type type, required_type;
f4913f91
JH
1380 void *buffer;
1381 unsigned long isize;
40469ee9 1382 unsigned char actual_sha1[20];
f4913f91 1383
21666f1a 1384 required_type = type_from_string(required_type_name);
e702496e 1385 hashcpy(actual_sha1, sha1);
40469ee9
JH
1386 while (1) {
1387 int ref_length = -1;
1388 const char *ref_type = NULL;
f4913f91 1389
21666f1a 1390 buffer = read_sha1_file(actual_sha1, &type, &isize);
40469ee9
JH
1391 if (!buffer)
1392 return NULL;
21666f1a 1393 if (type == required_type) {
40469ee9
JH
1394 *size = isize;
1395 if (actual_sha1_return)
e702496e 1396 hashcpy(actual_sha1_return, actual_sha1);
40469ee9
JH
1397 return buffer;
1398 }
1399 /* Handle references */
21666f1a 1400 else if (type == OBJ_COMMIT)
40469ee9 1401 ref_type = "tree ";
21666f1a 1402 else if (type == OBJ_TAG)
40469ee9
JH
1403 ref_type = "object ";
1404 else {
1405 free(buffer);
1406 return NULL;
1407 }
1408 ref_length = strlen(ref_type);
f4913f91 1409
50974ec9
MK
1410 if (ref_length + 40 > isize ||
1411 memcmp(buffer, ref_type, ref_length) ||
1d7f171c 1412 get_sha1_hex((char *) buffer + ref_length, actual_sha1)) {
40469ee9
JH
1413 free(buffer);
1414 return NULL;
1415 }
1cf58e72 1416 free(buffer);
40469ee9
JH
1417 /* Now we have the ID of the referred-to object in
1418 * actual_sha1. Check again. */
f4913f91 1419 }
f4913f91
JH
1420}
1421
ce9fbf16 1422static void write_sha1_file_prepare(const void *buf, unsigned long len,
972a9155 1423 const char *type, unsigned char *sha1,
d65a16f6 1424 char *hdr, int *hdrlen)
d410c0f5 1425{
9126f009 1426 git_SHA_CTX c;
d410c0f5
JH
1427
1428 /* Generate the header */
ef1286d3 1429 *hdrlen = xsnprintf(hdr, *hdrlen, "%s %lu", type, len)+1;
d410c0f5
JH
1430
1431 /* Sha1.. */
9126f009
NP
1432 git_SHA1_Init(&c);
1433 git_SHA1_Update(&c, hdr, *hdrlen);
1434 git_SHA1_Update(&c, buf, len);
1435 git_SHA1_Final(sha1, &c);
d410c0f5
JH
1436}
1437
230f1322 1438/*
5a688fe4 1439 * Move the just written object into its final resting place.
230f1322 1440 */
cb5add58 1441int finalize_object_file(const char *tmpfile, const char *filename)
230f1322 1442{
e32c0a9c 1443 int ret = 0;
5a688fe4 1444
348df166 1445 if (object_creation_mode == OBJECT_CREATION_USES_RENAMES)
be66a6c4
JS
1446 goto try_rename;
1447 else if (link(tmpfile, filename))
e32c0a9c 1448 ret = errno;
7ebb6fca
LT
1449
1450 /*
1451 * Coda hack - coda doesn't like cross-directory links,
1452 * so we fall back to a rename, which will mean that it
1453 * won't be able to check collisions, but that's not a
1454 * big deal.
1455 *
1456 * The same holds for FAT formatted media.
1457 *
3be1f18e 1458 * When this succeeds, we just return. We have nothing
7ebb6fca
LT
1459 * left to unlink.
1460 */
1461 if (ret && ret != EEXIST) {
be66a6c4 1462 try_rename:
7ebb6fca 1463 if (!rename(tmpfile, filename))
3be1f18e 1464 goto out;
9e48b389 1465 ret = errno;
230f1322 1466 }
691f1a28 1467 unlink_or_warn(tmpfile);
230f1322
LT
1468 if (ret) {
1469 if (ret != EEXIST) {
7616c6ca 1470 return error_errno("unable to write sha1 filename %s", filename);
230f1322
LT
1471 }
1472 /* FIXME!!! Collision check here ? */
1473 }
1474
3be1f18e 1475out:
5256b006 1476 if (adjust_shared_perm(filename))
5a688fe4 1477 return error("unable to set permission to '%s'", filename);
230f1322
LT
1478 return 0;
1479}
1480
4d548150
LT
1481static int write_buffer(int fd, const void *buf, size_t len)
1482{
d34cf19b 1483 if (write_in_full(fd, buf, len) < 0)
7616c6ca 1484 return error_errno("file write error");
4d548150
LT
1485 return 0;
1486}
1487
ce9fbf16 1488int hash_sha1_file(const void *buf, unsigned long len, const char *type,
abdc3fc8
RS
1489 unsigned char *sha1)
1490{
d65a16f6 1491 char hdr[32];
ef1286d3 1492 int hdrlen = sizeof(hdr);
abdc3fc8
RS
1493 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
1494 return 0;
1495}
1496
e9039dd3
LT
1497/* Finalize a file on disk, and close it. */
1498static void close_sha1_file(int fd)
1499{
aafe9fba
LT
1500 if (fsync_object_files)
1501 fsync_or_die(fd, "sha1 file");
e9039dd3 1502 if (close(fd) != 0)
d824cbba 1503 die_errno("error when closing sha1 file");
e9039dd3
LT
1504}
1505
5723fe7e
LT
1506/* Size of directory component, including the ending '/' */
1507static inline int directory_size(const char *filename)
1508{
1509 const char *s = strrchr(filename, '/');
1510 if (!s)
1511 return 0;
1512 return s - filename + 1;
1513}
1514
1515/*
1516 * This creates a temporary file in the same directory as the final
1517 * 'filename'
1518 *
1519 * We want to avoid cross-directory filename renames, because those
1520 * can have problems on various filesystems (FAT, NFS, Coda).
1521 */
d4b3d11a 1522static int create_tmpfile(struct strbuf *tmp, const char *filename)
5723fe7e
LT
1523{
1524 int fd, dirlen = directory_size(filename);
1525
d4b3d11a
JK
1526 strbuf_reset(tmp);
1527 strbuf_add(tmp, filename, dirlen);
1528 strbuf_addstr(tmp, "tmp_obj_XXXXXX");
1529 fd = git_mkstemp_mode(tmp->buf, 0444);
cbacbf4e 1530 if (fd < 0 && dirlen && errno == ENOENT) {
d4b3d11a
JK
1531 /*
1532 * Make sure the directory exists; note that the contents
1533 * of the buffer are undefined after mkstemp returns an
1534 * error, so we have to rewrite the whole buffer from
1535 * scratch.
1536 */
1537 strbuf_reset(tmp);
1538 strbuf_add(tmp, filename, dirlen - 1);
1539 if (mkdir(tmp->buf, 0777) && errno != EEXIST)
b2476a60 1540 return -1;
d4b3d11a 1541 if (adjust_shared_perm(tmp->buf))
5723fe7e
LT
1542 return -1;
1543
1544 /* Try again */
d4b3d11a
JK
1545 strbuf_addstr(tmp, "/tmp_obj_XXXXXX");
1546 fd = git_mkstemp_mode(tmp->buf, 0444);
5723fe7e
LT
1547 }
1548 return fd;
1549}
1550
bbac7311 1551static int write_loose_object(const unsigned char *sha1, char *hdr, int hdrlen,
c00e657d 1552 const void *buf, unsigned long len, time_t mtime)
0fcfd160 1553{
915308b1 1554 int fd, ret;
9892beba 1555 unsigned char compressed[4096];
ef49a7a0 1556 git_zstream stream;
748af44c
NP
1557 git_SHA_CTX c;
1558 unsigned char parano_sha1[20];
d4b3d11a 1559 static struct strbuf tmp_file = STRBUF_INIT;
30d6c6ea 1560 const char *filename = sha1_file_name(sha1);
a44c9a5e 1561
d4b3d11a 1562 fd = create_tmpfile(&tmp_file, filename);
aac17941 1563 if (fd < 0) {
35243577 1564 if (errno == EACCES)
82247e9b 1565 return error("insufficient permission for adding an object to repository database %s", get_object_directory());
916d081b 1566 else
7616c6ca 1567 return error_errno("unable to create temporary file");
aac17941
LT
1568 }
1569
0fcfd160 1570 /* Set it up */
55bb5c91 1571 git_deflate_init(&stream, zlib_compression_level);
0fcfd160 1572 stream.next_out = compressed;
9892beba 1573 stream.avail_out = sizeof(compressed);
748af44c 1574 git_SHA1_Init(&c);
a44c9a5e
LT
1575
1576 /* First header.. */
d65a16f6 1577 stream.next_in = (unsigned char *)hdr;
a44c9a5e 1578 stream.avail_in = hdrlen;
55bb5c91
JH
1579 while (git_deflate(&stream, 0) == Z_OK)
1580 ; /* nothing */
748af44c 1581 git_SHA1_Update(&c, hdr, hdrlen);
a44c9a5e
LT
1582
1583 /* Then the data itself.. */
c00e657d 1584 stream.next_in = (void *)buf;
a44c9a5e 1585 stream.avail_in = len;
9892beba 1586 do {
748af44c 1587 unsigned char *in0 = stream.next_in;
55bb5c91 1588 ret = git_deflate(&stream, Z_FINISH);
748af44c 1589 git_SHA1_Update(&c, in0, stream.next_in - in0);
9892beba
NP
1590 if (write_buffer(fd, compressed, stream.next_out - compressed) < 0)
1591 die("unable to write sha1 file");
1592 stream.next_out = compressed;
1593 stream.avail_out = sizeof(compressed);
1594 } while (ret == Z_OK);
1595
ac54c277
LT
1596 if (ret != Z_STREAM_END)
1597 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1), ret);
55bb5c91 1598 ret = git_deflate_end_gently(&stream);
ac54c277
LT
1599 if (ret != Z_OK)
1600 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1), ret);
748af44c
NP
1601 git_SHA1_Final(parano_sha1, &c);
1602 if (hashcmp(sha1, parano_sha1) != 0)
1603 die("confused by unstable object source data for %s", sha1_to_hex(sha1));
ac54c277 1604
e9039dd3 1605 close_sha1_file(fd);
0fcfd160 1606
bbac7311
NP
1607 if (mtime) {
1608 struct utimbuf utb;
1609 utb.actime = mtime;
1610 utb.modtime = mtime;
d4b3d11a 1611 if (utime(tmp_file.buf, &utb) < 0)
7616c6ca 1612 warning_errno("failed utime() on %s", tmp_file.buf);
bbac7311
NP
1613 }
1614
d4b3d11a 1615 return finalize_object_file(tmp_file.buf, filename);
0fcfd160 1616}
8237b185 1617
33d4221c
JK
1618static int freshen_loose_object(const unsigned char *sha1)
1619{
1620 return check_and_freshen(sha1, 1);
1621}
1622
1623static int freshen_packed_object(const unsigned char *sha1)
1624{
1625 struct pack_entry e;
ee1c6c34
JK
1626 if (!find_pack_entry(sha1, &e))
1627 return 0;
1628 if (e.p->freshened)
1629 return 1;
1630 if (!freshen_file(e.p->pack_name))
1631 return 0;
1632 e.p->freshened = 1;
1633 return 1;
33d4221c
JK
1634}
1635
1427a7ff 1636int write_sha1_file(const void *buf, unsigned long len, const char *type, unsigned char *sha1)
bbac7311 1637{
bbac7311 1638 char hdr[32];
ef1286d3 1639 int hdrlen = sizeof(hdr);
bbac7311
NP
1640
1641 /* Normally if we have it in the pack then we do not bother writing
1642 * it out into .git/objects/??/?{38} file.
1643 */
1644 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
b5f52f37 1645 if (freshen_packed_object(sha1) || freshen_loose_object(sha1))
bbac7311
NP
1646 return 0;
1647 return write_loose_object(sha1, hdr, hdrlen, buf, len, 0);
1648}
1649
0c3db67c 1650int hash_sha1_file_literally(const void *buf, unsigned long len, const char *type,
da77611d 1651 struct object_id *oid, unsigned flags)
0c3db67c
ES
1652{
1653 char *header;
1654 int hdrlen, status = 0;
1655
1656 /* type string, SP, %lu of the length plus NUL must fit this */
ef1286d3
JK
1657 hdrlen = strlen(type) + 32;
1658 header = xmalloc(hdrlen);
da77611d 1659 write_sha1_file_prepare(buf, len, type, oid->hash, header, &hdrlen);
0c3db67c
ES
1660
1661 if (!(flags & HASH_WRITE_OBJECT))
1662 goto cleanup;
da77611d 1663 if (freshen_packed_object(oid->hash) || freshen_loose_object(oid->hash))
0c3db67c 1664 goto cleanup;
da77611d 1665 status = write_loose_object(oid->hash, header, hdrlen, buf, len, 0);
0c3db67c
ES
1666
1667cleanup:
1668 free(header);
1669 return status;
1670}
1671
bbac7311
NP
1672int force_object_loose(const unsigned char *sha1, time_t mtime)
1673{
bbac7311
NP
1674 void *buf;
1675 unsigned long len;
1676 enum object_type type;
1677 char hdr[32];
1678 int hdrlen;
1fb23e65 1679 int ret;
bbac7311 1680
c529d75a 1681 if (has_loose_object(sha1))
bbac7311 1682 return 0;
789bf26b 1683 buf = read_object(sha1, &type, &len);
bbac7311
NP
1684 if (!buf)
1685 return error("cannot read sha1_file for %s", sha1_to_hex(sha1));
ef1286d3 1686 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %lu", typename(type), len) + 1;
1fb23e65
BS
1687 ret = write_loose_object(sha1, hdr, hdrlen, buf, len, mtime);
1688 free(buf);
1689
1690 return ret;
bbac7311
NP
1691}
1692
0eeb077b 1693int has_sha1_file_with_flags(const unsigned char *sha1, int flags)
8237b185 1694{
3e8b7d3c
JN
1695 if (!startup_info->have_repository)
1696 return 0;
e83e71c5
JT
1697 return sha1_object_info_extended(sha1, NULL,
1698 flags | OBJECT_INFO_SKIP_CACHED) >= 0;
b419aa25 1699}
1700
1701int has_object_file(const struct object_id *oid)
1702{
1703 return has_sha1_file(oid->hash);
8237b185 1704}
74400e71 1705
5827a035
JK
1706int has_object_file_with_flags(const struct object_id *oid, int flags)
1707{
1708 return has_sha1_file_with_flags(oid->hash, flags);
1709}
1710
c879daa2
NTND
1711static void check_tree(const void *buf, size_t size)
1712{
1713 struct tree_desc desc;
1714 struct name_entry entry;
1715
1716 init_tree_desc(&desc, buf, size);
1717 while (tree_entry(&desc, &entry))
1718 /* do nothing
1719 * tree_entry() will die() on malformed entries */
1720 ;
1721}
1722
1723static void check_commit(const void *buf, size_t size)
1724{
1725 struct commit c;
1726 memset(&c, 0, sizeof(c));
1727 if (parse_commit_buffer(&c, buf, size))
1728 die("corrupt commit");
1729}
1730
1731static void check_tag(const void *buf, size_t size)
1732{
1733 struct tag t;
1734 memset(&t, 0, sizeof(t));
1735 if (parse_tag_buffer(&t, buf, size))
1736 die("corrupt tag");
1737}
1738
bcd29864 1739static int index_mem(struct object_id *oid, void *buf, size_t size,
c4ce46fc
JH
1740 enum object_type type,
1741 const char *path, unsigned flags)
e7332f96 1742{
6c510bee 1743 int ret, re_allocated = 0;
c4ce46fc 1744 int write_object = flags & HASH_WRITE_OBJECT;
74400e71 1745
7672db20 1746 if (!type)
edaec3fb 1747 type = OBJ_BLOB;
6c510bee
LT
1748
1749 /*
1750 * Convert blobs to git internal format
1751 */
43df4f86 1752 if ((type == OBJ_BLOB) && path) {
f285a2d7 1753 struct strbuf nbuf = STRBUF_INIT;
82b474e0 1754 if (convert_to_git(&the_index, path, buf, size, &nbuf,
9472935d 1755 get_safe_crlf(flags))) {
b315c5c0 1756 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
1757 re_allocated = 1;
1758 }
1759 }
c4ce46fc 1760 if (flags & HASH_FORMAT_CHECK) {
c879daa2
NTND
1761 if (type == OBJ_TREE)
1762 check_tree(buf, size);
1763 if (type == OBJ_COMMIT)
1764 check_commit(buf, size);
1765 if (type == OBJ_TAG)
1766 check_tag(buf, size);
1767 }
6c510bee 1768
7672db20 1769 if (write_object)
bcd29864 1770 ret = write_sha1_file(buf, size, typename(type), oid->hash);
abdc3fc8 1771 else
bcd29864 1772 ret = hash_sha1_file(buf, size, typename(type), oid->hash);
43df4f86 1773 if (re_allocated)
6c510bee 1774 free(buf);
43df4f86
DP
1775 return ret;
1776}
1777
bcd29864 1778static int index_stream_convert_blob(struct object_id *oid, int fd,
9035d75a
SP
1779 const char *path, unsigned flags)
1780{
1781 int ret;
1782 const int write_object = flags & HASH_WRITE_OBJECT;
1783 struct strbuf sbuf = STRBUF_INIT;
1784
1785 assert(path);
1786 assert(would_convert_to_git_filter_fd(path));
1787
d6c41c20 1788 convert_to_git_filter_fd(&the_index, path, fd, &sbuf,
9472935d 1789 get_safe_crlf(flags));
9035d75a
SP
1790
1791 if (write_object)
1792 ret = write_sha1_file(sbuf.buf, sbuf.len, typename(OBJ_BLOB),
bcd29864 1793 oid->hash);
9035d75a
SP
1794 else
1795 ret = hash_sha1_file(sbuf.buf, sbuf.len, typename(OBJ_BLOB),
bcd29864 1796 oid->hash);
9035d75a
SP
1797 strbuf_release(&sbuf);
1798 return ret;
1799}
1800
bcd29864 1801static int index_pipe(struct object_id *oid, int fd, enum object_type type,
7b41e1e1
JH
1802 const char *path, unsigned flags)
1803{
1804 struct strbuf sbuf = STRBUF_INIT;
1805 int ret;
1806
1807 if (strbuf_read(&sbuf, fd, 4096) >= 0)
bcd29864 1808 ret = index_mem(oid, sbuf.buf, sbuf.len, type, path, flags);
7b41e1e1
JH
1809 else
1810 ret = -1;
1811 strbuf_release(&sbuf);
1812 return ret;
1813}
1814
ea68b0ce
DP
1815#define SMALL_FILE_SIZE (32*1024)
1816
bcd29864 1817static int index_core(struct object_id *oid, int fd, size_t size,
7b41e1e1
JH
1818 enum object_type type, const char *path,
1819 unsigned flags)
43df4f86
DP
1820{
1821 int ret;
43df4f86 1822
7b41e1e1 1823 if (!size) {
bcd29864 1824 ret = index_mem(oid, "", size, type, path, flags);
ea68b0ce
DP
1825 } else if (size <= SMALL_FILE_SIZE) {
1826 char *buf = xmalloc(size);
90dca671
JK
1827 ssize_t read_result = read_in_full(fd, buf, size);
1828 if (read_result < 0)
1829 ret = error_errno("read error while indexing %s",
1830 path ? path : "<unknown>");
1831 else if (read_result != size)
1832 ret = error("short read while indexing %s",
1833 path ? path : "<unknown>");
ea68b0ce 1834 else
bcd29864 1835 ret = index_mem(oid, buf, size, type, path, flags);
ea68b0ce 1836 free(buf);
08bda208 1837 } else {
43df4f86 1838 void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
bcd29864 1839 ret = index_mem(oid, buf, size, type, path, flags);
aac17941 1840 munmap(buf, size);
08bda208 1841 }
7b41e1e1
JH
1842 return ret;
1843}
1844
4dd1fbc7 1845/*
568508e7
JH
1846 * This creates one packfile per large blob unless bulk-checkin
1847 * machinery is "plugged".
4dd1fbc7
JH
1848 *
1849 * This also bypasses the usual "convert-to-git" dance, and that is on
1850 * purpose. We could write a streaming version of the converting
1851 * functions and insert that before feeding the data to fast-import
4f22b101
JK
1852 * (or equivalent in-core API described above). However, that is
1853 * somewhat complicated, as we do not know the size of the filter
1854 * result, which we need to know beforehand when writing a git object.
1855 * Since the primary motivation for trying to stream from the working
1856 * tree file and to avoid mmaping it in core is to deal with large
1857 * binary blobs, they generally do not want to get any conversion, and
1858 * callers should avoid this code path when filters are requested.
4dd1fbc7 1859 */
7d5e1dc3 1860static int index_stream(struct object_id *oid, int fd, size_t size,
4dd1fbc7
JH
1861 enum object_type type, const char *path,
1862 unsigned flags)
1863{
7d5e1dc3 1864 return index_bulk_checkin(oid->hash, fd, size, type, path, flags);
4dd1fbc7
JH
1865}
1866
e3506559 1867int index_fd(struct object_id *oid, int fd, struct stat *st,
7b41e1e1
JH
1868 enum object_type type, const char *path, unsigned flags)
1869{
1870 int ret;
7b41e1e1 1871
9079ab7c
SP
1872 /*
1873 * Call xsize_t() only when needed to avoid potentially unnecessary
1874 * die() for large files.
1875 */
9035d75a 1876 if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(path))
bcd29864 1877 ret = index_stream_convert_blob(oid, fd, path, flags);
9035d75a 1878 else if (!S_ISREG(st->st_mode))
bcd29864 1879 ret = index_pipe(oid, fd, type, path, flags);
9079ab7c 1880 else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
82b474e0 1881 (path && would_convert_to_git(&the_index, path)))
bcd29864 1882 ret = index_core(oid, fd, xsize_t(st->st_size), type, path,
9079ab7c 1883 flags);
4dd1fbc7 1884 else
7d5e1dc3 1885 ret = index_stream(oid, fd, xsize_t(st->st_size), type, path,
9079ab7c 1886 flags);
43df4f86 1887 close(fd);
aac17941 1888 return ret;
74400e71 1889}
ec1fcc16 1890
98e019b0 1891int index_path(struct object_id *oid, const char *path, struct stat *st, unsigned flags)
ec1fcc16
JH
1892{
1893 int fd;
b760d3aa 1894 struct strbuf sb = STRBUF_INIT;
ea8e0297 1895 int rc = 0;
ec1fcc16
JH
1896
1897 switch (st->st_mode & S_IFMT) {
1898 case S_IFREG:
1899 fd = open(path, O_RDONLY);
1900 if (fd < 0)
7616c6ca 1901 return error_errno("open(\"%s\")", path);
e3506559 1902 if (index_fd(oid, fd, st, OBJ_BLOB, path, flags) < 0)
ec1fcc16
JH
1903 return error("%s: failed to insert into database",
1904 path);
1905 break;
1906 case S_IFLNK:
7616c6ca
NTND
1907 if (strbuf_readlink(&sb, path, st->st_size))
1908 return error_errno("readlink(\"%s\")", path);
c4ce46fc 1909 if (!(flags & HASH_WRITE_OBJECT))
98e019b0
PO
1910 hash_sha1_file(sb.buf, sb.len, blob_type, oid->hash);
1911 else if (write_sha1_file(sb.buf, sb.len, blob_type, oid->hash))
ea8e0297 1912 rc = error("%s: failed to insert into database", path);
b760d3aa 1913 strbuf_release(&sb);
ec1fcc16 1914 break;
f35a6d3b 1915 case S_IFDIR:
a98e6101 1916 return resolve_gitlink_ref(path, "HEAD", oid);
ec1fcc16
JH
1917 default:
1918 return error("%s: unsupported file type", path);
1919 }
ea8e0297 1920 return rc;
ec1fcc16 1921}
a69e5429
JH
1922
1923int read_pack_header(int fd, struct pack_header *header)
1924{
f48ecd38 1925 if (read_in_full(fd, header, sizeof(*header)) != sizeof(*header))
c697ad14
HO
1926 /* "eof before pack header was fully read" */
1927 return PH_ERROR_EOF;
1928
a69e5429
JH
1929 if (header->hdr_signature != htonl(PACK_SIGNATURE))
1930 /* "protocol error (pack signature mismatch detected)" */
1931 return PH_ERROR_PACK_SIGNATURE;
1932 if (!pack_version_ok(header->hdr_version))
1933 /* "protocol error (pack version unsupported)" */
1934 return PH_ERROR_PROTOCOL;
1935 return 0;
1936}
40d52ff7
JK
1937
1938void assert_sha1_type(const unsigned char *sha1, enum object_type expect)
1939{
1940 enum object_type type = sha1_object_info(sha1, NULL);
1941 if (type < 0)
1942 die("%s is not a valid object", sha1_to_hex(sha1));
1943 if (type != expect)
1944 die("%s is not a valid '%s' object", sha1_to_hex(sha1),
1945 typename(expect));
1946}
27e1e22d 1947
70c49050 1948int for_each_file_in_obj_subdir(unsigned int subdir_nr,
cc817ca3
RS
1949 struct strbuf *path,
1950 each_loose_object_fn obj_cb,
1951 each_loose_cruft_fn cruft_cb,
1952 each_loose_subdir_fn subdir_cb,
1953 void *data)
27e1e22d 1954{
0375f472
RS
1955 size_t origlen, baselen;
1956 DIR *dir;
27e1e22d
JK
1957 struct dirent *de;
1958 int r = 0;
62a24c89 1959 struct object_id oid;
27e1e22d 1960
70c49050
RS
1961 if (subdir_nr > 0xff)
1962 BUG("invalid loose object subdirectory: %x", subdir_nr);
1963
0375f472
RS
1964 origlen = path->len;
1965 strbuf_complete(path, '/');
1966 strbuf_addf(path, "%02x", subdir_nr);
0375f472
RS
1967
1968 dir = opendir(path->buf);
27e1e22d 1969 if (!dir) {
0375f472
RS
1970 if (errno != ENOENT)
1971 r = error_errno("unable to open %s", path->buf);
1972 strbuf_setlen(path, origlen);
1973 return r;
27e1e22d
JK
1974 }
1975
62a24c89 1976 oid.hash[0] = subdir_nr;
163ee5e6
DS
1977 strbuf_addch(path, '/');
1978 baselen = path->len;
62a24c89 1979
27e1e22d 1980 while ((de = readdir(dir))) {
163ee5e6 1981 size_t namelen;
27e1e22d
JK
1982 if (is_dot_or_dotdot(de->d_name))
1983 continue;
1984
163ee5e6 1985 namelen = strlen(de->d_name);
27e1e22d 1986 strbuf_setlen(path, baselen);
163ee5e6
DS
1987 strbuf_add(path, de->d_name, namelen);
1988 if (namelen == GIT_SHA1_HEXSZ - 2 &&
62a24c89
RS
1989 !hex_to_bytes(oid.hash + 1, de->d_name,
1990 GIT_SHA1_RAWSZ - 1)) {
1991 if (obj_cb) {
1992 r = obj_cb(&oid, path->buf, data);
1993 if (r)
1994 break;
27e1e22d 1995 }
62a24c89 1996 continue;
27e1e22d
JK
1997 }
1998
1999 if (cruft_cb) {
2000 r = cruft_cb(de->d_name, path->buf, data);
2001 if (r)
2002 break;
2003 }
2004 }
094c7e63 2005 closedir(dir);
27e1e22d 2006
163ee5e6 2007 strbuf_setlen(path, baselen - 1);
27e1e22d
JK
2008 if (!r && subdir_cb)
2009 r = subdir_cb(subdir_nr, path->buf, data);
2010
0375f472
RS
2011 strbuf_setlen(path, origlen);
2012
27e1e22d
JK
2013 return r;
2014}
2015
e6f875e0 2016int for_each_loose_file_in_objdir_buf(struct strbuf *path,
27e1e22d
JK
2017 each_loose_object_fn obj_cb,
2018 each_loose_cruft_fn cruft_cb,
2019 each_loose_subdir_fn subdir_cb,
2020 void *data)
2021{
27e1e22d
JK
2022 int r = 0;
2023 int i;
2024
27e1e22d 2025 for (i = 0; i < 256; i++) {
e6f875e0 2026 r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
27e1e22d 2027 subdir_cb, data);
27e1e22d
JK
2028 if (r)
2029 break;
2030 }
2031
e6f875e0
JK
2032 return r;
2033}
2034
2035int for_each_loose_file_in_objdir(const char *path,
2036 each_loose_object_fn obj_cb,
2037 each_loose_cruft_fn cruft_cb,
2038 each_loose_subdir_fn subdir_cb,
2039 void *data)
2040{
2041 struct strbuf buf = STRBUF_INIT;
2042 int r;
2043
2044 strbuf_addstr(&buf, path);
2045 r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb,
2046 subdir_cb, data);
27e1e22d 2047 strbuf_release(&buf);
e6f875e0 2048
27e1e22d
JK
2049 return r;
2050}
660c889e
JK
2051
2052struct loose_alt_odb_data {
2053 each_loose_object_fn *cb;
2054 void *data;
2055};
2056
2057static int loose_from_alt_odb(struct alternate_object_database *alt,
2058 void *vdata)
2059{
2060 struct loose_alt_odb_data *data = vdata;
b0a42642
JM
2061 struct strbuf buf = STRBUF_INIT;
2062 int r;
2063
597f9134 2064 strbuf_addstr(&buf, alt->path);
b0a42642
JM
2065 r = for_each_loose_file_in_objdir_buf(&buf,
2066 data->cb, NULL, NULL,
2067 data->data);
2068 strbuf_release(&buf);
2069 return r;
660c889e
JK
2070}
2071
1385bb7b 2072int for_each_loose_object(each_loose_object_fn cb, void *data, unsigned flags)
660c889e
JK
2073{
2074 struct loose_alt_odb_data alt;
2075 int r;
2076
2077 r = for_each_loose_file_in_objdir(get_object_directory(),
2078 cb, NULL, NULL, data);
2079 if (r)
2080 return r;
2081
1385bb7b
JK
2082 if (flags & FOR_EACH_OBJECT_LOCAL_ONLY)
2083 return 0;
2084
660c889e
JK
2085 alt.cb = cb;
2086 alt.data = data;
2087 return foreach_alt_odb(loose_from_alt_odb, &alt);
2088}
2089
f6371f92
JK
2090static int check_stream_sha1(git_zstream *stream,
2091 const char *hdr,
2092 unsigned long size,
2093 const char *path,
2094 const unsigned char *expected_sha1)
2095{
2096 git_SHA_CTX c;
cd02599c 2097 unsigned char real_sha1[GIT_MAX_RAWSZ];
f6371f92
JK
2098 unsigned char buf[4096];
2099 unsigned long total_read;
2100 int status = Z_OK;
2101
2102 git_SHA1_Init(&c);
2103 git_SHA1_Update(&c, hdr, stream->total_out);
2104
2105 /*
2106 * We already read some bytes into hdr, but the ones up to the NUL
2107 * do not count against the object's content size.
2108 */
2109 total_read = stream->total_out - strlen(hdr) - 1;
2110
2111 /*
2112 * This size comparison must be "<=" to read the final zlib packets;
2113 * see the comment in unpack_sha1_rest for details.
2114 */
2115 while (total_read <= size &&
2116 (status == Z_OK || status == Z_BUF_ERROR)) {
2117 stream->next_out = buf;
2118 stream->avail_out = sizeof(buf);
2119 if (size - total_read < stream->avail_out)
2120 stream->avail_out = size - total_read;
2121 status = git_inflate(stream, Z_FINISH);
2122 git_SHA1_Update(&c, buf, stream->next_out - buf);
2123 total_read += stream->next_out - buf;
2124 }
2125 git_inflate_end(stream);
2126
2127 if (status != Z_STREAM_END) {
2128 error("corrupt loose object '%s'", sha1_to_hex(expected_sha1));
2129 return -1;
2130 }
cce044df
JK
2131 if (stream->avail_in) {
2132 error("garbage at end of loose object '%s'",
2133 sha1_to_hex(expected_sha1));
2134 return -1;
2135 }
f6371f92
JK
2136
2137 git_SHA1_Final(real_sha1, &c);
2138 if (hashcmp(expected_sha1, real_sha1)) {
2139 error("sha1 mismatch for %s (expected %s)", path,
2140 sha1_to_hex(expected_sha1));
2141 return -1;
2142 }
2143
2144 return 0;
2145}
2146
2147int read_loose_object(const char *path,
2148 const unsigned char *expected_sha1,
2149 enum object_type *type,
2150 unsigned long *size,
2151 void **contents)
2152{
2153 int ret = -1;
f6371f92
JK
2154 void *map = NULL;
2155 unsigned long mapsize;
2156 git_zstream stream;
2157 char hdr[32];
2158
2159 *contents = NULL;
2160
2161 map = map_sha1_file_1(path, NULL, &mapsize);
2162 if (!map) {
2163 error_errno("unable to mmap %s", path);
2164 goto out;
2165 }
2166
2167 if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) {
2168 error("unable to unpack header of %s", path);
2169 goto out;
2170 }
2171
2172 *type = parse_sha1_header(hdr, size);
2173 if (*type < 0) {
2174 error("unable to parse header of %s", path);
2175 git_inflate_end(&stream);
2176 goto out;
2177 }
2178
2179 if (*type == OBJ_BLOB) {
2180 if (check_stream_sha1(&stream, hdr, *size, path, expected_sha1) < 0)
2181 goto out;
2182 } else {
2183 *contents = unpack_sha1_rest(&stream, hdr, *size, expected_sha1);
2184 if (!*contents) {
2185 error("unable to unpack contents of %s", path);
2186 git_inflate_end(&stream);
2187 goto out;
2188 }
2189 if (check_sha1_signature(expected_sha1, *contents,
2190 *size, typename(*type))) {
2191 error("sha1 mismatch for %s (expected %s)", path,
2192 sha1_to_hex(expected_sha1));
2193 free(*contents);
2194 goto out;
2195 }
2196 }
2197
2198 ret = 0; /* everything checks out */
2199
2200out:
2201 if (map)
2202 munmap(map, mapsize);
f6371f92
JK
2203 return ret;
2204}