pack: move nth_packed_object_{sha1,oid}
[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"
0fcfd160 32
96f1e58f 33const unsigned char null_sha1[20];
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
c597ba80
NTND
42/*
43 * This is meant to hold a *small* number of objects that you would
44 * want read_sha1_file() to be able to return, but yet you do not want
45 * to write them into the object store (e.g. a browse-only
46 * application).
47 */
48static struct cached_object {
49 unsigned char sha1[20];
50 enum object_type type;
51 void *buf;
52 unsigned long size;
53} *cached_objects;
54static int cached_object_nr, cached_object_alloc;
55
56static struct cached_object empty_tree = {
dab0d410 57 EMPTY_TREE_SHA1_BIN_LITERAL,
c597ba80
NTND
58 OBJ_TREE,
59 "",
60 0
61};
62
63static struct cached_object *find_cached_object(const unsigned char *sha1)
64{
65 int i;
66 struct cached_object *co = cached_objects;
67
68 for (i = 0; i < cached_object_nr; i++, co++) {
69 if (!hashcmp(co->sha1, sha1))
70 return co;
71 }
72 if (!hashcmp(sha1, empty_tree.sha1))
73 return &empty_tree;
74 return NULL;
75}
76
90a6464b
JH
77int mkdir_in_gitdir(const char *path)
78{
79 if (mkdir(path, 0777)) {
80 int saved_errno = errno;
81 struct stat st;
82 struct strbuf sb = STRBUF_INIT;
83
84 if (errno != EEXIST)
85 return -1;
86 /*
87 * Are we looking at a path in a symlinked worktree
88 * whose original repository does not yet have it?
89 * e.g. .git/rr-cache pointing at its original
90 * repository in which the user hasn't performed any
91 * conflict resolution yet?
92 */
93 if (lstat(path, &st) || !S_ISLNK(st.st_mode) ||
94 strbuf_readlink(&sb, path, st.st_size) ||
95 !is_absolute_path(sb.buf) ||
96 mkdir(sb.buf, 0777)) {
97 strbuf_release(&sb);
98 errno = saved_errno;
99 return -1;
100 }
101 strbuf_release(&sb);
102 }
103 return adjust_shared_perm(path);
104}
105
0be0521b 106enum scld_error safe_create_leading_directories(char *path)
b2cb9425 107{
26c8ae2a 108 char *next_component = path + offset_1st_component(path);
0be0521b 109 enum scld_error ret = SCLD_OK;
67d42212 110
0be0521b 111 while (ret == SCLD_OK && next_component) {
f0502332 112 struct stat st;
0f527403 113 char *slash = next_component, slash_character;
f0502332 114
0f527403
MH
115 while (*slash && !is_dir_sep(*slash))
116 slash++;
117
118 if (!*slash)
b2cb9425 119 break;
bf10cf70 120
26c8ae2a 121 next_component = slash + 1;
0f527403 122 while (is_dir_sep(*next_component))
bf10cf70 123 next_component++;
26c8ae2a 124 if (!*next_component)
5f0bdf50 125 break;
831651fd 126
0f527403 127 slash_character = *slash;
831651fd 128 *slash = '\0';
67d42212
JR
129 if (!stat(path, &st)) {
130 /* path exists */
204a047f
MH
131 if (!S_ISDIR(st.st_mode)) {
132 errno = ENOTDIR;
0be0521b 133 ret = SCLD_EXISTS;
204a047f 134 }
53a39721 135 } else if (mkdir(path, 0777)) {
928734d9 136 if (errno == EEXIST &&
9e6f885d 137 !stat(path, &st) && S_ISDIR(st.st_mode))
928734d9 138 ; /* somebody created it since we checked */
18d37e86
MH
139 else if (errno == ENOENT)
140 /*
141 * Either mkdir() failed because
142 * somebody just pruned the containing
143 * directory, or stat() failed because
144 * the file that was in our way was
145 * just removed. Either way, inform
146 * the caller that it might be worth
147 * trying again:
148 */
149 ret = SCLD_VANISHED;
9e6f885d 150 else
0be0521b 151 ret = SCLD_FAILED;
53a39721 152 } else if (adjust_shared_perm(path)) {
0be0521b 153 ret = SCLD_PERMS;
457f06d6 154 }
0f527403 155 *slash = slash_character;
b2cb9425 156 }
9e6f885d 157 return ret;
b2cb9425 158}
723c31fe 159
0be0521b 160enum scld_error safe_create_leading_directories_const(const char *path)
8e21d63b 161{
02944307 162 int save_errno;
8e21d63b
JK
163 /* path points to cache entries, so xstrdup before messing with it */
164 char *buf = xstrdup(path);
0be0521b 165 enum scld_error result = safe_create_leading_directories(buf);
02944307
MH
166
167 save_errno = errno;
8e21d63b 168 free(buf);
02944307 169 errno = save_errno;
8e21d63b
JK
170 return result;
171}
172
177978f5
MH
173int raceproof_create_file(const char *path, create_file_fn fn, void *cb)
174{
175 /*
176 * The number of times we will try to remove empty directories
177 * in the way of path. This is only 1 because if another
178 * process is racily creating directories that conflict with
179 * us, we don't want to fight against them.
180 */
181 int remove_directories_remaining = 1;
182
183 /*
184 * The number of times that we will try to create the
185 * directories containing path. We are willing to attempt this
186 * more than once, because another process could be trying to
187 * clean up empty directories at the same time as we are
188 * trying to create them.
189 */
190 int create_directories_remaining = 3;
191
192 /* A scratch copy of path, filled lazily if we need it: */
193 struct strbuf path_copy = STRBUF_INIT;
194
195 int ret, save_errno;
196
197 /* Sanity check: */
198 assert(*path);
199
200retry_fn:
201 ret = fn(path, cb);
202 save_errno = errno;
203 if (!ret)
204 goto out;
205
206 if (errno == EISDIR && remove_directories_remaining-- > 0) {
207 /*
208 * A directory is in the way. Maybe it is empty; try
209 * to remove it:
210 */
211 if (!path_copy.len)
212 strbuf_addstr(&path_copy, path);
213
214 if (!remove_dir_recursively(&path_copy, REMOVE_DIR_EMPTY_ONLY))
215 goto retry_fn;
216 } else if (errno == ENOENT && create_directories_remaining-- > 0) {
217 /*
218 * Maybe the containing directory didn't exist, or
219 * maybe it was just deleted by a process that is
220 * racing with us to clean up empty directories. Try
221 * to create it:
222 */
223 enum scld_error scld_result;
224
225 if (!path_copy.len)
226 strbuf_addstr(&path_copy, path);
227
228 do {
229 scld_result = safe_create_leading_directories(path_copy.buf);
230 if (scld_result == SCLD_OK)
231 goto retry_fn;
232 } while (scld_result == SCLD_VANISHED && create_directories_remaining-- > 0);
233 }
234
235out:
236 strbuf_release(&path_copy);
237 errno = save_errno;
238 return ret;
239}
240
f7b7774f 241static void fill_sha1_path(struct strbuf *buf, const unsigned char *sha1)
ace1534d
JH
242{
243 int i;
244 for (i = 0; i < 20; i++) {
245 static char hex[] = "0123456789abcdef";
246 unsigned int val = sha1[i];
f7b7774f
JK
247 strbuf_addch(buf, hex[val >> 4]);
248 strbuf_addch(buf, hex[val & 0xf]);
afbba2f0 249 if (!i)
f7b7774f 250 strbuf_addch(buf, '/');
ace1534d
JH
251 }
252}
253
30d6c6ea 254const char *sha1_file_name(const unsigned char *sha1)
0fcfd160 255{
f7b7774f 256 static struct strbuf buf = STRBUF_INIT;
0fcfd160 257
f7b7774f
JK
258 strbuf_reset(&buf);
259 strbuf_addf(&buf, "%s/", get_object_directory());
560fb6a1 260
f7b7774f
JK
261 fill_sha1_path(&buf, sha1);
262 return buf.buf;
0fcfd160
LT
263}
264
38dbe5f0
JK
265struct strbuf *alt_scratch_buf(struct alternate_object_database *alt)
266{
267 strbuf_setlen(&alt->scratch, alt->base_len);
268 return &alt->scratch;
269}
270
29ec6af2
JK
271static const char *alt_sha1_path(struct alternate_object_database *alt,
272 const unsigned char *sha1)
273{
38dbe5f0 274 struct strbuf *buf = alt_scratch_buf(alt);
f7b7774f 275 fill_sha1_path(buf, sha1);
38dbe5f0 276 return buf->buf;
0fcfd160
LT
277}
278
d5a63b99
JH
279struct alternate_object_database *alt_odb_list;
280static struct alternate_object_database **alt_odb_tail;
ace1534d 281
4ea82473
JK
282/*
283 * Return non-zero iff the path is usable as an alternate object database.
284 */
285static int alt_odb_usable(struct strbuf *path, const char *normalized_objdir)
286{
287 struct alternate_object_database *alt;
288
289 /* Detect cases where alternate disappeared */
290 if (!is_directory(path->buf)) {
291 error("object directory %s does not exist; "
292 "check .git/objects/info/alternates.",
293 path->buf);
294 return 0;
295 }
296
297 /*
298 * Prevent the common mistake of listing the same
299 * thing twice, or object directory itself.
300 */
301 for (alt = alt_odb_list; alt; alt = alt->next) {
ea0fc3b4 302 if (!fspathcmp(path->buf, alt->path))
4ea82473
JK
303 return 0;
304 }
305 if (!fspathcmp(path->buf, normalized_objdir))
306 return 0;
307
308 return 1;
309}
310
ddd5d056
JH
311/*
312 * Prepare alternate object database registry.
d5a63b99
JH
313 *
314 * The variable alt_odb_list points at the list of struct
315 * alternate_object_database. The elements on this list come from
316 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
317 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
318 * whose contents is similar to that environment variable but can be
319 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
320 * contains "/the/directory/corresponding/to/.git/objects/...", while
321 * its name points just after the slash at the end of ".git/objects/"
322 * in the example above, and has enough space to hold 40-byte hex
323 * SHA1, an extra slash for the first level indirection, and the
324 * terminating NUL.
ddd5d056 325 */
2456990d 326static void read_info_alternates(const char * relative_base, int depth);
539e7506
EK
327static int link_alt_odb_entry(const char *entry, const char *relative_base,
328 int depth, const char *normalized_objdir)
ace1534d 329{
c2f493a4 330 struct alternate_object_database *ent;
5bdf0a84 331 struct strbuf pathbuf = STRBUF_INIT;
d5a63b99 332
85dadc38 333 if (!is_absolute_path(entry) && relative_base) {
4ac9006f 334 strbuf_realpath(&pathbuf, relative_base, 1);
5bdf0a84 335 strbuf_addch(&pathbuf, '/');
c2f493a4 336 }
6eac50d8 337 strbuf_addstr(&pathbuf, entry);
c2f493a4 338
37a95862 339 if (strbuf_normalize_path(&pathbuf) < 0 && relative_base) {
670c359d
JK
340 error("unable to normalize alternate object path: %s",
341 pathbuf.buf);
342 strbuf_release(&pathbuf);
343 return -1;
344 }
5bdf0a84
HW
345
346 /*
347 * The trailing slash after the directory name is given by
348 * this function at the end. Remove duplicates.
349 */
4ea82473
JK
350 while (pathbuf.len && pathbuf.buf[pathbuf.len - 1] == '/')
351 strbuf_setlen(&pathbuf, pathbuf.len - 1);
5bdf0a84 352
4ea82473
JK
353 if (!alt_odb_usable(&pathbuf, normalized_objdir)) {
354 strbuf_release(&pathbuf);
c2f493a4
MW
355 return -1;
356 }
357
7f0fa2c0 358 ent = alloc_alt_odb(pathbuf.buf);
c2f493a4
MW
359
360 /* add the alternate entry */
361 *alt_odb_tail = ent;
362 alt_odb_tail = &(ent->next);
363 ent->next = NULL;
364
365 /* recursively add alternates */
4ea82473 366 read_info_alternates(pathbuf.buf, depth + 1);
c2f493a4 367
4ea82473 368 strbuf_release(&pathbuf);
c2f493a4
MW
369 return 0;
370}
371
cf3c6352
JK
372static const char *parse_alt_odb_entry(const char *string,
373 int sep,
374 struct strbuf *out)
375{
376 const char *end;
377
378 strbuf_reset(out);
379
380 if (*string == '#') {
381 /* comment; consume up to next separator */
382 end = strchrnul(string, sep);
383 } else if (*string == '"' && !unquote_c_style(out, string, &end)) {
384 /*
385 * quoted path; unquote_c_style has copied the
386 * data for us and set "end". Broken quoting (e.g.,
387 * an entry that doesn't end with a quote) falls
388 * back to the unquoted case below.
389 */
390 } else {
391 /* normal, unquoted path */
392 end = strchrnul(string, sep);
393 strbuf_add(out, string, end - string);
394 }
395
396 if (*end)
397 end++;
398 return end;
399}
400
c5950164 401static void link_alt_odb_entries(const char *alt, int len, int sep,
c2f493a4
MW
402 const char *relative_base, int depth)
403{
539e7506 404 struct strbuf objdirbuf = STRBUF_INIT;
cf3c6352 405 struct strbuf entry = STRBUF_INIT;
c2f493a4
MW
406
407 if (depth > 5) {
408 error("%s: ignoring alternate object stores, nesting too deep.",
409 relative_base);
410 return;
411 }
412
f655651e 413 strbuf_add_absolute_path(&objdirbuf, get_object_directory());
670c359d
JK
414 if (strbuf_normalize_path(&objdirbuf) < 0)
415 die("unable to normalize object directory: %s",
416 objdirbuf.buf);
539e7506 417
cf3c6352
JK
418 while (*alt) {
419 alt = parse_alt_odb_entry(alt, sep, &entry);
420 if (!entry.len)
9577e7e3 421 continue;
cf3c6352 422 link_alt_odb_entry(entry.buf, relative_base, depth, objdirbuf.buf);
9577e7e3 423 }
cf3c6352 424 strbuf_release(&entry);
539e7506 425 strbuf_release(&objdirbuf);
d5a63b99
JH
426}
427
2456990d 428static void read_info_alternates(const char * relative_base, int depth)
d5a63b99 429{
9577e7e3 430 char *map;
dc49cd76 431 size_t mapsz;
d5a63b99 432 struct stat st;
5015f01c 433 char *path;
c2f493a4 434 int fd;
d5a63b99 435
5015f01c 436 path = xstrfmt("%s/info/alternates", relative_base);
a5436b57 437 fd = git_open(path);
5015f01c 438 free(path);
d5a63b99
JH
439 if (fd < 0)
440 return;
441 if (fstat(fd, &st) || (st.st_size == 0)) {
442 close(fd);
9a217f2a 443 return;
ace1534d 444 }
dc49cd76
SP
445 mapsz = xsize_t(st.st_size);
446 map = xmmap(NULL, mapsz, PROT_READ, MAP_PRIVATE, fd, 0);
d5a63b99 447 close(fd);
d5a63b99 448
c5950164 449 link_alt_odb_entries(map, mapsz, '\n', relative_base, depth);
c2f493a4 450
dc49cd76 451 munmap(map, mapsz);
ace1534d
JH
452}
453
7f0fa2c0
JK
454struct alternate_object_database *alloc_alt_odb(const char *dir)
455{
456 struct alternate_object_database *ent;
7f0fa2c0 457
597f9134 458 FLEX_ALLOC_STR(ent, path, dir);
38dbe5f0
JK
459 strbuf_init(&ent->scratch, 0);
460 strbuf_addf(&ent->scratch, "%s/", dir);
461 ent->base_len = ent->scratch.len;
7f0fa2c0
JK
462
463 return ent;
464}
465
bef70b22
DB
466void add_to_alternates_file(const char *reference)
467{
468 struct lock_file *lock = xcalloc(1, sizeof(struct lock_file));
77b9b1d1
JK
469 char *alts = git_pathdup("objects/info/alternates");
470 FILE *in, *out;
471
472 hold_lock_file_for_update(lock, alts, LOCK_DIE_ON_ERROR);
473 out = fdopen_lock_file(lock, "w");
474 if (!out)
475 die_errno("unable to fdopen alternates lockfile");
476
477 in = fopen(alts, "r");
478 if (in) {
479 struct strbuf line = STRBUF_INIT;
480 int found = 0;
481
3f163962 482 while (strbuf_getline(&line, in) != EOF) {
77b9b1d1
JK
483 if (!strcmp(reference, line.buf)) {
484 found = 1;
485 break;
486 }
487 fprintf_or_die(out, "%s\n", line.buf);
488 }
489
490 strbuf_release(&line);
491 fclose(in);
492
493 if (found) {
494 rollback_lock_file(lock);
495 lock = NULL;
496 }
497 }
498 else if (errno != ENOENT)
499 die_errno("unable to read alternates file");
500
501 if (lock) {
502 fprintf_or_die(out, "%s\n", reference);
503 if (commit_lock_file(lock))
504 die_errno("unable to move new alternates file into place");
505 if (alt_odb_tail)
506 link_alt_odb_entries(reference, strlen(reference), '\n', NULL, 0);
507 }
508 free(alts);
bef70b22
DB
509}
510
a5b34d21
JK
511void add_to_alternates_memory(const char *reference)
512{
513 /*
514 * Make sure alternates are initialized, or else our entry may be
515 * overwritten when they are.
516 */
517 prepare_alt_odb();
518
519 link_alt_odb_entries(reference, strlen(reference), '\n', NULL, 0);
520}
521
9eeea7d2
SB
522/*
523 * Compute the exact path an alternate is at and returns it. In case of
524 * error NULL is returned and the human readable error is added to `err`
525 * `path` may be relative and should point to $GITDIR.
526 * `err` must not be null.
527 */
528char *compute_alternate_path(const char *path, struct strbuf *err)
529{
530 char *ref_git = NULL;
531 const char *repo, *ref_git_s;
532 int seen_error = 0;
533
534 ref_git_s = real_path_if_valid(path);
535 if (!ref_git_s) {
536 seen_error = 1;
537 strbuf_addf(err, _("path '%s' does not exist"), path);
538 goto out;
539 } else
540 /*
541 * Beware: read_gitfile(), real_path() and mkpath()
542 * return static buffer
543 */
544 ref_git = xstrdup(ref_git_s);
545
546 repo = read_gitfile(ref_git);
547 if (!repo)
548 repo = read_gitfile(mkpath("%s/.git", ref_git));
549 if (repo) {
550 free(ref_git);
551 ref_git = xstrdup(repo);
552 }
553
554 if (!repo && is_directory(mkpath("%s/.git/objects", ref_git))) {
555 char *ref_git_git = mkpathdup("%s/.git", ref_git);
556 free(ref_git);
557 ref_git = ref_git_git;
558 } else if (!is_directory(mkpath("%s/objects", ref_git))) {
559 struct strbuf sb = STRBUF_INIT;
560 seen_error = 1;
561 if (get_common_dir(&sb, ref_git)) {
562 strbuf_addf(err,
563 _("reference repository '%s' as a linked "
564 "checkout is not supported yet."),
565 path);
566 goto out;
567 }
568
569 strbuf_addf(err, _("reference repository '%s' is not a "
570 "local repository."), path);
571 goto out;
572 }
573
574 if (!access(mkpath("%s/shallow", ref_git), F_OK)) {
575 strbuf_addf(err, _("reference repository '%s' is shallow"),
576 path);
577 seen_error = 1;
578 goto out;
579 }
580
581 if (!access(mkpath("%s/info/grafts", ref_git), F_OK)) {
582 strbuf_addf(err,
583 _("reference repository '%s' is grafted"),
584 path);
585 seen_error = 1;
586 goto out;
587 }
588
589out:
590 if (seen_error) {
6a83d902 591 FREE_AND_NULL(ref_git);
9eeea7d2
SB
592 }
593
594 return ref_git;
595}
596
fe1b2268 597int foreach_alt_odb(alt_odb_fn fn, void *cb)
d79796bc
JH
598{
599 struct alternate_object_database *ent;
fe1b2268 600 int r = 0;
d79796bc
JH
601
602 prepare_alt_odb();
fe1b2268
JK
603 for (ent = alt_odb_list; ent; ent = ent->next) {
604 r = fn(ent, cb);
605 if (r)
606 break;
607 }
608 return r;
d79796bc
JH
609}
610
c2f493a4
MW
611void prepare_alt_odb(void)
612{
554fe20d 613 const char *alt;
c2f493a4 614
7dc24aa5
SP
615 if (alt_odb_tail)
616 return;
617
c2f493a4
MW
618 alt = getenv(ALTERNATE_DB_ENVIRONMENT);
619 if (!alt) alt = "";
620
c2f493a4 621 alt_odb_tail = &alt_odb_list;
c5950164 622 link_alt_odb_entries(alt, strlen(alt), PATH_SEP, NULL, 0);
c2f493a4
MW
623
624 read_info_alternates(get_object_directory(), 0);
625}
626
3096b2ec 627/* Returns 1 if we have successfully freshened the file, 0 otherwise. */
33d4221c 628static int freshen_file(const char *fn)
ace1534d 629{
33d4221c
JK
630 struct utimbuf t;
631 t.actime = t.modtime = time(NULL);
632 return !utime(fn, &t);
0f4dc14a 633}
ace1534d 634
3096b2ec
JK
635/*
636 * All of the check_and_freshen functions return 1 if the file exists and was
637 * freshened (if freshening was requested), 0 otherwise. If they return
638 * 0, you should not assume that it is safe to skip a write of the object (it
639 * either does not exist on disk, or has a stale mtime and may be subject to
640 * pruning).
641 */
6a5e6f5e 642int check_and_freshen_file(const char *fn, int freshen)
33d4221c
JK
643{
644 if (access(fn, F_OK))
645 return 0;
3096b2ec 646 if (freshen && !freshen_file(fn))
33d4221c
JK
647 return 0;
648 return 1;
649}
650
651static int check_and_freshen_local(const unsigned char *sha1, int freshen)
652{
653 return check_and_freshen_file(sha1_file_name(sha1), freshen);
654}
655
656static int check_and_freshen_nonlocal(const unsigned char *sha1, int freshen)
0f4dc14a
BC
657{
658 struct alternate_object_database *alt;
9a217f2a 659 prepare_alt_odb();
d5a63b99 660 for (alt = alt_odb_list; alt; alt = alt->next) {
29ec6af2
JK
661 const char *path = alt_sha1_path(alt, sha1);
662 if (check_and_freshen_file(path, freshen))
c529d75a 663 return 1;
ace1534d 664 }
c529d75a 665 return 0;
ace1534d
JH
666}
667
33d4221c
JK
668static int check_and_freshen(const unsigned char *sha1, int freshen)
669{
670 return check_and_freshen_local(sha1, freshen) ||
671 check_and_freshen_nonlocal(sha1, freshen);
672}
673
674int has_loose_object_nonlocal(const unsigned char *sha1)
675{
676 return check_and_freshen_nonlocal(sha1, 0);
677}
678
0f4dc14a
BC
679static int has_loose_object(const unsigned char *sha1)
680{
33d4221c 681 return check_and_freshen(sha1, 0);
0f4dc14a
BC
682}
683
02710228
SP
684static void mmap_limit_check(size_t length)
685{
686 static size_t limit = 0;
687 if (!limit) {
688 limit = git_env_ulong("GIT_MMAP_LIMIT", 0);
689 if (!limit)
690 limit = SIZE_MAX;
691 }
692 if (length > limit)
693 die("attempting to mmap %"PRIuMAX" over limit %"PRIuMAX,
694 (uintmax_t)length, (uintmax_t)limit);
695}
696
1570856b
JK
697void *xmmap_gently(void *start, size_t length,
698 int prot, int flags, int fd, off_t offset)
58ecbd5e 699{
02710228
SP
700 void *ret;
701
702 mmap_limit_check(length);
703 ret = mmap(start, length, prot, flags, fd, offset);
58ecbd5e
JN
704 if (ret == MAP_FAILED) {
705 if (!length)
706 return NULL;
7c3ecb32 707 release_pack_memory(length);
58ecbd5e 708 ret = mmap(start, length, prot, flags, fd, offset);
58ecbd5e
JN
709 }
710 return ret;
711}
712
1570856b
JK
713void *xmmap(void *start, size_t length,
714 int prot, int flags, int fd, off_t offset)
715{
716 void *ret = xmmap_gently(start, length, prot, flags, fd, offset);
717 if (ret == MAP_FAILED)
9ca0aaf6 718 die_errno("mmap failed");
1570856b
JK
719 return ret;
720}
721
090ea126
NTND
722/*
723 * With an in-core object data in "map", rehash it to make sure the
724 * object name actually matches "sha1" to detect object corruption.
725 * With "map" == NULL, try reading the object named with "sha1" using
726 * the streaming interface and rehash it to do the same.
727 */
728int check_sha1_signature(const unsigned char *sha1, void *map,
729 unsigned long size, const char *type)
0fcfd160
LT
730{
731 unsigned char real_sha1[20];
090ea126
NTND
732 enum object_type obj_type;
733 struct git_istream *st;
734 git_SHA_CTX c;
735 char hdr[32];
736 int hdrlen;
737
738 if (map) {
739 hash_sha1_file(map, size, type, real_sha1);
740 return hashcmp(sha1, real_sha1) ? -1 : 0;
741 }
742
743 st = open_istream(sha1, &obj_type, &size, NULL);
744 if (!st)
745 return -1;
746
747 /* Generate the header */
ef1286d3 748 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %lu", typename(obj_type), size) + 1;
090ea126
NTND
749
750 /* Sha1.. */
751 git_SHA1_Init(&c);
752 git_SHA1_Update(&c, hdr, hdrlen);
753 for (;;) {
754 char buf[1024 * 16];
755 ssize_t readlen = read_istream(st, buf, sizeof(buf));
756
f54fac53
JK
757 if (readlen < 0) {
758 close_istream(st);
759 return -1;
760 }
090ea126
NTND
761 if (!readlen)
762 break;
763 git_SHA1_Update(&c, buf, readlen);
764 }
765 git_SHA1_Final(real_sha1, &c);
766 close_istream(st);
a89fccd2 767 return hashcmp(sha1, real_sha1) ? -1 : 0;
0fcfd160
LT
768}
769
1b8ac5ea 770int git_open_cloexec(const char *name, int flags)
44d1c19e 771{
1e3001a8
JH
772 int fd;
773 static int o_cloexec = O_CLOEXEC;
f2e872aa 774
1e3001a8
JH
775 fd = open(name, flags | o_cloexec);
776 if ((o_cloexec & O_CLOEXEC) && fd < 0 && errno == EINVAL) {
cd66ada0 777 /* Try again w/o O_CLOEXEC: the kernel might not support it */
1e3001a8
JH
778 o_cloexec &= ~O_CLOEXEC;
779 fd = open(name, flags | o_cloexec);
1b8ac5ea 780 }
f2e872aa 781
9fb9495d 782#if defined(F_GETFD) && defined(F_SETFD) && defined(FD_CLOEXEC)
1e3001a8
JH
783 {
784 static int fd_cloexec = FD_CLOEXEC;
785
786 if (!o_cloexec && 0 <= fd && fd_cloexec) {
787 /* Opened w/o O_CLOEXEC? try with fcntl(2) to add it */
9fb9495d
EW
788 int flags = fcntl(fd, F_GETFD);
789 if (fcntl(fd, F_SETFD, flags | fd_cloexec))
1e3001a8 790 fd_cloexec = 0;
cd66ada0 791 }
44d1c19e 792 }
1e3001a8 793#endif
1b8ac5ea 794 return fd;
44d1c19e
LT
795}
796
771e7d57
JK
797/*
798 * Find "sha1" as a loose object in the local repository or in an alternate.
799 * Returns 0 on success, negative on failure.
800 *
801 * The "path" out-parameter will give the path of the object we found (if any).
802 * Note that it may point to static storage and is only valid until another
803 * call to sha1_file_name(), etc.
804 */
805static int stat_sha1_file(const unsigned char *sha1, struct stat *st,
806 const char **path)
052fe5ea 807{
052fe5ea
JK
808 struct alternate_object_database *alt;
809
771e7d57
JK
810 *path = sha1_file_name(sha1);
811 if (!lstat(*path, st))
052fe5ea
JK
812 return 0;
813
814 prepare_alt_odb();
815 errno = ENOENT;
816 for (alt = alt_odb_list; alt; alt = alt->next) {
771e7d57
JK
817 *path = alt_sha1_path(alt, sha1);
818 if (!lstat(*path, st))
052fe5ea
JK
819 return 0;
820 }
821
822 return -1;
823}
824
771e7d57
JK
825/*
826 * Like stat_sha1_file(), but actually open the object and return the
827 * descriptor. See the caveats on the "path" parameter above.
828 */
829static int open_sha1_file(const unsigned char *sha1, const char **path)
44d1c19e
LT
830{
831 int fd;
44d1c19e 832 struct alternate_object_database *alt;
d6c8a05b 833 int most_interesting_errno;
44d1c19e 834
771e7d57
JK
835 *path = sha1_file_name(sha1);
836 fd = git_open(*path);
44d1c19e
LT
837 if (fd >= 0)
838 return fd;
d6c8a05b 839 most_interesting_errno = errno;
44d1c19e
LT
840
841 prepare_alt_odb();
44d1c19e 842 for (alt = alt_odb_list; alt; alt = alt->next) {
771e7d57
JK
843 *path = alt_sha1_path(alt, sha1);
844 fd = git_open(*path);
44d1c19e
LT
845 if (fd >= 0)
846 return fd;
d6c8a05b
JK
847 if (most_interesting_errno == ENOENT)
848 most_interesting_errno = errno;
44d1c19e 849 }
d6c8a05b 850 errno = most_interesting_errno;
44d1c19e
LT
851 return -1;
852}
853
f6371f92
JK
854/*
855 * Map the loose object at "path" if it is not NULL, or the path found by
856 * searching for a loose object named "sha1".
857 */
858static void *map_sha1_file_1(const char *path,
859 const unsigned char *sha1,
860 unsigned long *size)
0fcfd160 861{
0fcfd160 862 void *map;
144bde78 863 int fd;
ace1534d 864
f6371f92
JK
865 if (path)
866 fd = git_open(path);
867 else
868 fd = open_sha1_file(sha1, &path);
44d1c19e
LT
869 map = NULL;
870 if (fd >= 0) {
871 struct stat st;
0fcfd160 872
44d1c19e
LT
873 if (!fstat(fd, &st)) {
874 *size = xsize_t(st.st_size);
33e42de0
MM
875 if (!*size) {
876 /* mmap() is forbidden on empty files */
771e7d57 877 error("object file %s is empty", path);
33e42de0
MM
878 return NULL;
879 }
44d1c19e 880 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
144bde78 881 }
44d1c19e 882 close(fd);
0fcfd160 883 }
0fcfd160
LT
884 return map;
885}
886
f6371f92
JK
887void *map_sha1_file(const unsigned char *sha1, unsigned long *size)
888{
889 return map_sha1_file_1(NULL, sha1, size);
890}
891
d21f8426
JH
892static int unpack_sha1_short_header(git_zstream *stream,
893 unsigned char *map, unsigned long mapsize,
894 void *buffer, unsigned long bufsiz)
ad1ed5ee 895{
c4483576
LT
896 /* Get the data stream */
897 memset(stream, 0, sizeof(*stream));
898 stream->next_in = map;
899 stream->avail_in = mapsize;
900 stream->next_out = buffer;
93821bd9
LT
901 stream->avail_out = bufsiz;
902
39c68542 903 git_inflate_init(stream);
cc5c54e7 904 return git_inflate(stream, 0);
c4483576
LT
905}
906
d21f8426
JH
907int unpack_sha1_header(git_zstream *stream,
908 unsigned char *map, unsigned long mapsize,
909 void *buffer, unsigned long bufsiz)
910{
911 int status = unpack_sha1_short_header(stream, map, mapsize,
912 buffer, bufsiz);
913
914 if (status < Z_OK)
915 return status;
916
917 /* Make sure we have the terminating NUL */
918 if (!memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
919 return -1;
920 return 0;
921}
922
46f03448
KN
923static int unpack_sha1_header_to_strbuf(git_zstream *stream, unsigned char *map,
924 unsigned long mapsize, void *buffer,
925 unsigned long bufsiz, struct strbuf *header)
926{
927 int status;
928
d21f8426
JH
929 status = unpack_sha1_short_header(stream, map, mapsize, buffer, bufsiz);
930 if (status < Z_OK)
931 return -1;
46f03448
KN
932
933 /*
934 * Check if entire header is unpacked in the first iteration.
935 */
936 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
937 return 0;
938
939 /*
940 * buffer[0..bufsiz] was not large enough. Copy the partial
941 * result out to header, and then append the result of further
942 * reading the stream.
943 */
944 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
945 stream->next_out = buffer;
946 stream->avail_out = bufsiz;
947
948 do {
949 status = git_inflate(stream, 0);
950 strbuf_add(header, buffer, stream->next_out - (unsigned char *)buffer);
951 if (memchr(buffer, '\0', stream->next_out - (unsigned char *)buffer))
952 return 0;
953 stream->next_out = buffer;
954 stream->avail_out = bufsiz;
955 } while (status != Z_STREAM_END);
956 return -1;
957}
958
ef49a7a0 959static void *unpack_sha1_rest(git_zstream *stream, void *buffer, unsigned long size, const unsigned char *sha1)
5180cacc
LT
960{
961 int bytes = strlen(buffer) + 1;
3aee68aa 962 unsigned char *buf = xmallocz(size);
93821bd9 963 unsigned long n;
7efbff75 964 int status = Z_OK;
5180cacc 965
93821bd9
LT
966 n = stream->total_out - bytes;
967 if (n > size)
968 n = size;
969 memcpy(buf, (char *) buffer + bytes, n);
970 bytes = n;
456cdf6e
LT
971 if (bytes <= size) {
972 /*
973 * The above condition must be (bytes <= size), not
974 * (bytes < size). In other words, even though we
ccf5ace0 975 * expect no more output and set avail_out to zero,
456cdf6e
LT
976 * the input zlib stream may have bytes that express
977 * "this concludes the stream", and we *do* want to
978 * eat that input.
979 *
980 * Otherwise we would not be able to test that we
981 * consumed all the input to reach the expected size;
982 * we also want to check that zlib tells us that all
983 * went well with status == Z_STREAM_END at the end.
984 */
5180cacc
LT
985 stream->next_out = buf + bytes;
986 stream->avail_out = size - bytes;
7efbff75 987 while (status == Z_OK)
39c68542 988 status = git_inflate(stream, Z_FINISH);
5180cacc 989 }
456cdf6e 990 if (status == Z_STREAM_END && !stream->avail_in) {
39c68542 991 git_inflate_end(stream);
7efbff75
JH
992 return buf;
993 }
994
995 if (status < 0)
996 error("corrupt loose object '%s'", sha1_to_hex(sha1));
997 else if (stream->avail_in)
998 error("garbage at end of loose object '%s'",
999 sha1_to_hex(sha1));
1000 free(buf);
1001 return NULL;
5180cacc
LT
1002}
1003
1004/*
1005 * We used to just use "sscanf()", but that's actually way
1006 * too permissive for what we want to check. So do an anal
1007 * object header parse by hand.
1008 */
46f03448
KN
1009static int parse_sha1_header_extended(const char *hdr, struct object_info *oi,
1010 unsigned int flags)
5180cacc 1011{
46f03448 1012 const char *type_buf = hdr;
5180cacc 1013 unsigned long size;
46f03448 1014 int type, type_len = 0;
5180cacc
LT
1015
1016 /*
46f03448 1017 * The type can be of any size but is followed by
21666f1a 1018 * a space.
5180cacc 1019 */
5180cacc
LT
1020 for (;;) {
1021 char c = *hdr++;
d21f8426
JH
1022 if (!c)
1023 return -1;
5180cacc
LT
1024 if (c == ' ')
1025 break;
46f03448 1026 type_len++;
5180cacc 1027 }
46f03448
KN
1028
1029 type = type_from_string_gently(type_buf, type_len, 1);
1030 if (oi->typename)
1031 strbuf_add(oi->typename, type_buf, type_len);
1032 /*
1033 * Set type to 0 if its an unknown object and
2e3a16b2 1034 * we're obtaining the type using '--allow-unknown-type'
46f03448
KN
1035 * option.
1036 */
19fc5e84 1037 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE) && (type < 0))
46f03448
KN
1038 type = 0;
1039 else if (type < 0)
1040 die("invalid object type");
1041 if (oi->typep)
1042 *oi->typep = type;
5180cacc
LT
1043
1044 /*
1045 * The length must follow immediately, and be in canonical
1046 * decimal format (ie "010" is not valid).
1047 */
1048 size = *hdr++ - '0';
1049 if (size > 9)
1050 return -1;
1051 if (size) {
1052 for (;;) {
1053 unsigned long c = *hdr - '0';
1054 if (c > 9)
1055 break;
1056 hdr++;
1057 size = size * 10 + c;
1058 }
1059 }
46f03448
KN
1060
1061 if (oi->sizep)
1062 *oi->sizep = size;
5180cacc
LT
1063
1064 /*
1065 * The length must be followed by a zero byte
1066 */
46f03448
KN
1067 return *hdr ? -1 : type;
1068}
1069
1070int parse_sha1_header(const char *hdr, unsigned long *sizep)
1071{
27b5c1a0 1072 struct object_info oi = OBJECT_INFO_INIT;
46f03448
KN
1073
1074 oi.sizep = sizep;
1f0c0d36 1075 return parse_sha1_header_extended(hdr, &oi, 0);
0fcfd160
LT
1076}
1077
47fe3f6e
JK
1078void check_pack_index_ptr(const struct packed_git *p, const void *vptr)
1079{
1080 const unsigned char *ptr = vptr;
1081 const unsigned char *start = p->index_data;
1082 const unsigned char *end = start + p->index_size;
1083 if (ptr < start)
7465feba 1084 die(_("offset before start of pack index for %s (corrupt index?)"),
47fe3f6e
JK
1085 p->pack_name);
1086 /* No need to check for underflow; .idx files must be at least 8 bytes */
1087 if (ptr >= end - 8)
7465feba 1088 die(_("offset beyond end of pack index for %s (truncated index?)"),
47fe3f6e
JK
1089 p->pack_name);
1090}
1091
99093238 1092off_t nth_packed_object_offset(const struct packed_git *p, uint32_t n)
74e34e1f
NP
1093{
1094 const unsigned char *index = p->index_data;
1095 index += 4 * 256;
1096 if (p->index_version == 1) {
1097 return ntohl(*((uint32_t *)(index + 24 * n)));
1098 } else {
1099 uint32_t off;
1100 index += 8 + p->num_objects * (20 + 4);
1101 off = ntohl(*((uint32_t *)(index + 4 * n)));
1102 if (!(off & 0x80000000))
1103 return off;
1104 index += p->num_objects * 4 + (off & 0x7fffffff) * 8;
47fe3f6e 1105 check_pack_index_ptr(p, index);
74e34e1f
NP
1106 return (((uint64_t)ntohl(*((uint32_t *)(index + 0)))) << 32) |
1107 ntohl(*((uint32_t *)(index + 4)));
1108 }
9a217f2a
JH
1109}
1110
c4001d92 1111off_t find_pack_entry_one(const unsigned char *sha1,
43057304 1112 struct packed_git *p)
1f688557 1113{
42873078 1114 const uint32_t *level1_ofs = p->index_data;
42873078 1115 const unsigned char *index = p->index_data;
628522ec 1116 unsigned hi, lo, stride;
628522ec
JH
1117 static int debug_lookup = -1;
1118
1119 if (debug_lookup < 0)
1120 debug_lookup = !!getenv("GIT_DEBUG_LOOKUP");
42873078 1121
d079837e
SP
1122 if (!index) {
1123 if (open_pack_index(p))
1124 return 0;
1125 level1_ofs = p->index_data;
1126 index = p->index_data;
1127 }
74e34e1f
NP
1128 if (p->index_version > 1) {
1129 level1_ofs += 2;
1130 index += 8;
1131 }
42873078 1132 index += 4 * 256;
74e34e1f
NP
1133 hi = ntohl(level1_ofs[*sha1]);
1134 lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
628522ec
JH
1135 if (p->index_version > 1) {
1136 stride = 20;
1137 } else {
1138 stride = 24;
1139 index += 4;
1140 }
1141
1142 if (debug_lookup)
6e1c2344 1143 printf("%02x%02x%02x... lo %u hi %u nr %"PRIu32"\n",
628522ec
JH
1144 sha1[0], sha1[1], sha1[2], lo, hi, p->num_objects);
1145
6355a768 1146 while (lo < hi) {
74e34e1f 1147 unsigned mi = (lo + hi) / 2;
628522ec
JH
1148 int cmp = hashcmp(index + mi * stride, sha1);
1149
1150 if (debug_lookup)
1151 printf("lo %u hi %u rg %u mi %u\n",
1152 lo, hi, hi - lo, mi);
43057304 1153 if (!cmp)
74e34e1f 1154 return nth_packed_object_offset(p, mi);
1f688557
JH
1155 if (cmp > 0)
1156 hi = mi;
1157 else
1158 lo = mi+1;
6355a768 1159 }
1f688557
JH
1160 return 0;
1161}
1162
4c080182 1163int is_pack_valid(struct packed_git *p)
d131b7af
SP
1164{
1165 /* An already open pack is known to be valid. */
1166 if (p->pack_fd != -1)
1167 return 1;
1168
1169 /* If the pack has one window completely covering the
1170 * file size, the pack is known to be valid even if
1171 * the descriptor is not currently open.
1172 */
1173 if (p->windows) {
1174 struct pack_window *w = p->windows;
1175
1176 if (!w->offset && w->len == p->pack_size)
1177 return 1;
1178 }
1179
1180 /* Force the pack to open to prove its valid. */
1181 return !open_packed_git(p);
1182}
1183
95099731
NTND
1184static int fill_pack_entry(const unsigned char *sha1,
1185 struct pack_entry *e,
1186 struct packed_git *p)
1187{
1188 off_t offset;
1189
1190 if (p->num_bad_objects) {
1191 unsigned i;
1192 for (i = 0; i < p->num_bad_objects; i++)
1193 if (!hashcmp(sha1, p->bad_object_sha1 + 20 * i))
1194 return 0;
1195 }
1196
1197 offset = find_pack_entry_one(sha1, p);
1198 if (!offset)
1199 return 0;
1200
1201 /*
1202 * We are about to tell the caller where they can locate the
1203 * requested object. We better make sure the packfile is
1204 * still here and can be accessed before supplying that
1205 * answer, as it may have been deleted since the index was
1206 * loaded!
1207 */
319b678a 1208 if (!is_pack_valid(p))
95099731 1209 return 0;
95099731
NTND
1210 e->offset = offset;
1211 e->p = p;
1212 hashcpy(e->sha1, sha1);
1213 return 1;
1214}
1215
d40d535b
MH
1216/*
1217 * Iff a pack file contains the object named by sha1, return true and
1218 * store its location to e.
1219 */
4d6acb70 1220static int find_pack_entry(const unsigned char *sha1, struct pack_entry *e)
1f688557 1221{
a73cdd21 1222 struct mru_entry *p;
43057304 1223
1f688557 1224 prepare_packed_git();
f7c22cc6
NP
1225 if (!packed_git)
1226 return 0;
1f688557 1227
a73cdd21
JK
1228 for (p = packed_git_mru->head; p; p = p->next) {
1229 if (fill_pack_entry(sha1, e, p->item)) {
1230 mru_mark(packed_git_mru, p);
1b1005d1
MH
1231 return 1;
1232 }
c01f51cc 1233 }
1f688557
JH
1234 return 0;
1235}
1236
a6080a0a 1237struct packed_git *find_sha1_pack(const unsigned char *sha1,
bf592c50
DB
1238 struct packed_git *packs)
1239{
1240 struct packed_git *p;
bf592c50
DB
1241
1242 for (p = packs; p; p = p->next) {
43057304 1243 if (find_pack_entry_one(sha1, p))
bf592c50
DB
1244 return p;
1245 }
1246 return NULL;
21666f1a 1247
bf592c50
DB
1248}
1249
052fe5ea 1250static int sha1_loose_object_info(const unsigned char *sha1,
46f03448
KN
1251 struct object_info *oi,
1252 int flags)
65c2e0c3 1253{
46f03448
KN
1254 int status = 0;
1255 unsigned long mapsize;
65c2e0c3 1256 void *map;
ef49a7a0 1257 git_zstream stream;
d65a16f6 1258 char hdr[32];
46f03448 1259 struct strbuf hdrbuf = STRBUF_INIT;
c84a1f3e 1260 unsigned long size_scratch;
65c2e0c3 1261
5d642e75
JK
1262 if (oi->delta_base_sha1)
1263 hashclr(oi->delta_base_sha1);
1264
052fe5ea
JK
1265 /*
1266 * If we don't care about type or size, then we don't
4ef8d1dd
JH
1267 * need to look inside the object at all. Note that we
1268 * do not optimize out the stat call, even if the
1269 * caller doesn't care about the disk-size, since our
1270 * return value implicitly indicates whether the
1271 * object even exists.
052fe5ea 1272 */
c84a1f3e 1273 if (!oi->typep && !oi->typename && !oi->sizep && !oi->contentp) {
771e7d57 1274 const char *path;
4ef8d1dd 1275 struct stat st;
771e7d57 1276 if (stat_sha1_file(sha1, &st, &path) < 0)
4ef8d1dd
JH
1277 return -1;
1278 if (oi->disk_sizep)
23c339c0 1279 *oi->disk_sizep = st.st_size;
052fe5ea
JK
1280 return 0;
1281 }
1282
bb6b8e4f 1283 map = map_sha1_file(sha1, &mapsize);
f0df4ed5 1284 if (!map)
dbea72a8 1285 return -1;
c84a1f3e
JT
1286
1287 if (!oi->sizep)
1288 oi->sizep = &size_scratch;
1289
23c339c0
JK
1290 if (oi->disk_sizep)
1291 *oi->disk_sizep = mapsize;
19fc5e84 1292 if ((flags & OBJECT_INFO_ALLOW_UNKNOWN_TYPE)) {
46f03448
KN
1293 if (unpack_sha1_header_to_strbuf(&stream, map, mapsize, hdr, sizeof(hdr), &hdrbuf) < 0)
1294 status = error("unable to unpack %s header with --allow-unknown-type",
1295 sha1_to_hex(sha1));
1296 } else if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
36e4d74a
JH
1297 status = error("unable to unpack %s header",
1298 sha1_to_hex(sha1));
46f03448
KN
1299 if (status < 0)
1300 ; /* Do nothing */
1301 else if (hdrbuf.len) {
1302 if ((status = parse_sha1_header_extended(hdrbuf.buf, oi, flags)) < 0)
1303 status = error("unable to parse %s header with --allow-unknown-type",
1304 sha1_to_hex(sha1));
1305 } else if ((status = parse_sha1_header_extended(hdr, oi, flags)) < 0)
36e4d74a 1306 status = error("unable to parse %s header", sha1_to_hex(sha1));
c84a1f3e
JT
1307
1308 if (status >= 0 && oi->contentp)
1309 *oi->contentp = unpack_sha1_rest(&stream, hdr,
1310 *oi->sizep, sha1);
1311 else
1312 git_inflate_end(&stream);
1313
65c2e0c3 1314 munmap(map, mapsize);
46f03448 1315 if (status && oi->typep)
23c339c0 1316 *oi->typep = status;
c84a1f3e
JT
1317 if (oi->sizep == &size_scratch)
1318 oi->sizep = NULL;
46f03448 1319 strbuf_release(&hdrbuf);
3ab0fb06 1320 oi->whence = OI_LOOSE;
93cff9a9 1321 return (status < 0) ? status : 0;
65c2e0c3
JH
1322}
1323
de7b5d62 1324int sha1_object_info_extended(const unsigned char *sha1, struct object_info *oi, unsigned flags)
f0df4ed5 1325{
cd585e2a 1326 static struct object_info blank_oi = OBJECT_INFO_INIT;
f0df4ed5 1327 struct pack_entry e;
5b086407 1328 int rtype;
1f0c0d36
JT
1329 const unsigned char *real = (flags & OBJECT_INFO_LOOKUP_REPLACE) ?
1330 lookup_replace_object(sha1) :
1331 sha1;
f0df4ed5 1332
cd585e2a
JT
1333 if (!oi)
1334 oi = &blank_oi;
1335
dfdd4afc
JT
1336 if (!(flags & OBJECT_INFO_SKIP_CACHED)) {
1337 struct cached_object *co = find_cached_object(real);
1338 if (co) {
1339 if (oi->typep)
1340 *(oi->typep) = co->type;
1341 if (oi->sizep)
1342 *(oi->sizep) = co->size;
1343 if (oi->disk_sizep)
1344 *(oi->disk_sizep) = 0;
1345 if (oi->delta_base_sha1)
1346 hashclr(oi->delta_base_sha1);
1347 if (oi->typename)
1348 strbuf_addstr(oi->typename, typename(co->type));
1349 if (oi->contentp)
1350 *oi->contentp = xmemdupz(co->buf, co->size);
1351 oi->whence = OI_CACHED;
1352 return 0;
1353 }
c4d9986f
NTND
1354 }
1355
1f7117ef 1356 if (!find_pack_entry(real, &e)) {
ddd63e64 1357 /* Most likely it's a loose object. */
3ab0fb06 1358 if (!sha1_loose_object_info(real, oi, flags))
5b086407 1359 return 0;
ddd63e64
SG
1360
1361 /* Not a loose object; someone else may have just packed it. */
dfdd4afc 1362 if (flags & OBJECT_INFO_QUICK) {
052fe5ea 1363 return -1;
dfdd4afc
JT
1364 } else {
1365 reprepare_packed_git();
1366 if (!find_pack_entry(real, &e))
1367 return -1;
1368 }
f0df4ed5 1369 }
3d77d877 1370
cd585e2a
JT
1371 if (oi == &blank_oi)
1372 /*
1373 * We know that the caller doesn't actually need the
1374 * information below, so return early.
1375 */
1376 return 0;
46f03448 1377
23c339c0 1378 rtype = packed_object_info(e.p, e.offset, oi);
412916ee 1379 if (rtype < 0) {
1f7117ef
CC
1380 mark_bad_packed_object(e.p, real);
1381 return sha1_object_info_extended(real, oi, 0);
3ab0fb06 1382 } else if (oi->whence == OI_PACKED) {
9a490590
JH
1383 oi->u.packed.offset = e.offset;
1384 oi->u.packed.pack = e.p;
1385 oi->u.packed.is_delta = (rtype == OBJ_REF_DELTA ||
1386 rtype == OBJ_OFS_DELTA);
3d77d877
NP
1387 }
1388
5b086407 1389 return 0;
f0df4ed5
JS
1390}
1391
3fc0dca9 1392/* returns enum object_type or negative */
9a490590
JH
1393int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
1394{
5b086407 1395 enum object_type type;
27b5c1a0 1396 struct object_info oi = OBJECT_INFO_INIT;
9a490590 1397
5b086407 1398 oi.typep = &type;
9a490590 1399 oi.sizep = sizep;
1f0c0d36
JT
1400 if (sha1_object_info_extended(sha1, &oi,
1401 OBJECT_INFO_LOOKUP_REPLACE) < 0)
5b086407
JK
1402 return -1;
1403 return type;
9a490590
JH
1404}
1405
f1d8130b
JT
1406static void *read_object(const unsigned char *sha1, enum object_type *type,
1407 unsigned long *size)
1408{
1409 struct object_info oi = OBJECT_INFO_INIT;
1410 void *content;
1411 oi.typep = type;
1412 oi.sizep = size;
1413 oi.contentp = &content;
1414
1415 if (sha1_object_info_extended(sha1, &oi, 0) < 0)
1416 return NULL;
1417 return content;
1418}
1419
21666f1a
NP
1420int pretend_sha1_file(void *buf, unsigned long len, enum object_type type,
1421 unsigned char *sha1)
d66b37bb
JH
1422{
1423 struct cached_object *co;
1424
21666f1a 1425 hash_sha1_file(buf, len, typename(type), sha1);
d66b37bb
JH
1426 if (has_sha1_file(sha1) || find_cached_object(sha1))
1427 return 0;
c7353967 1428 ALLOC_GROW(cached_objects, cached_object_nr + 1, cached_object_alloc);
d66b37bb
JH
1429 co = &cached_objects[cached_object_nr++];
1430 co->size = len;
21666f1a 1431 co->type = type;
efa13f7b
JH
1432 co->buf = xmalloc(len);
1433 memcpy(co->buf, buf, len);
d66b37bb
JH
1434 hashcpy(co->sha1, sha1);
1435 return 0;
1436}
1437
b6c4cecc
JH
1438/*
1439 * This function dies on corrupt objects; the callers who want to
1440 * deal with them should arrange to call read_object() and give error
1441 * messages themselves.
1442 */
5bf29b95
JH
1443void *read_sha1_file_extended(const unsigned char *sha1,
1444 enum object_type *type,
1445 unsigned long *size,
1f0c0d36 1446 int lookup_replace)
ac939109 1447{
3ba7a065 1448 void *data;
b6c4cecc 1449 const struct packed_git *p;
771e7d57
JK
1450 const char *path;
1451 struct stat st;
1f0c0d36
JT
1452 const unsigned char *repl = lookup_replace ? lookup_replace_object(sha1)
1453 : sha1;
b6c4cecc 1454
3ba7a065
JH
1455 errno = 0;
1456 data = read_object(repl, type, size);
4bbf5a26 1457 if (data)
b6c4cecc 1458 return data;
68095570 1459
25f3af3f 1460 if (errno && errno != ENOENT)
3ba7a065
JH
1461 die_errno("failed to read object %s", sha1_to_hex(sha1));
1462
68095570 1463 /* die if we replaced an object with one that does not exist */
b6c4cecc 1464 if (repl != sha1)
68095570
CC
1465 die("replacement %s not found for %s",
1466 sha1_to_hex(repl), sha1_to_hex(sha1));
1467
771e7d57 1468 if (!stat_sha1_file(repl, &st, &path))
b6c4cecc
JH
1469 die("loose object %s (stored in %s) is corrupt",
1470 sha1_to_hex(repl), path);
68095570 1471
b6c4cecc
JH
1472 if ((p = has_packed_and_bad(repl)) != NULL)
1473 die("packed object %s (stored in %s) is corrupt",
1474 sha1_to_hex(repl), p->pack_name);
f5552aee 1475
b6c4cecc 1476 return NULL;
ac939109
NP
1477}
1478
40469ee9 1479void *read_object_with_reference(const unsigned char *sha1,
21666f1a 1480 const char *required_type_name,
40469ee9
JH
1481 unsigned long *size,
1482 unsigned char *actual_sha1_return)
f4913f91 1483{
21666f1a 1484 enum object_type type, required_type;
f4913f91
JH
1485 void *buffer;
1486 unsigned long isize;
40469ee9 1487 unsigned char actual_sha1[20];
f4913f91 1488
21666f1a 1489 required_type = type_from_string(required_type_name);
e702496e 1490 hashcpy(actual_sha1, sha1);
40469ee9
JH
1491 while (1) {
1492 int ref_length = -1;
1493 const char *ref_type = NULL;
f4913f91 1494
21666f1a 1495 buffer = read_sha1_file(actual_sha1, &type, &isize);
40469ee9
JH
1496 if (!buffer)
1497 return NULL;
21666f1a 1498 if (type == required_type) {
40469ee9
JH
1499 *size = isize;
1500 if (actual_sha1_return)
e702496e 1501 hashcpy(actual_sha1_return, actual_sha1);
40469ee9
JH
1502 return buffer;
1503 }
1504 /* Handle references */
21666f1a 1505 else if (type == OBJ_COMMIT)
40469ee9 1506 ref_type = "tree ";
21666f1a 1507 else if (type == OBJ_TAG)
40469ee9
JH
1508 ref_type = "object ";
1509 else {
1510 free(buffer);
1511 return NULL;
1512 }
1513 ref_length = strlen(ref_type);
f4913f91 1514
50974ec9
MK
1515 if (ref_length + 40 > isize ||
1516 memcmp(buffer, ref_type, ref_length) ||
1d7f171c 1517 get_sha1_hex((char *) buffer + ref_length, actual_sha1)) {
40469ee9
JH
1518 free(buffer);
1519 return NULL;
1520 }
1cf58e72 1521 free(buffer);
40469ee9
JH
1522 /* Now we have the ID of the referred-to object in
1523 * actual_sha1. Check again. */
f4913f91 1524 }
f4913f91
JH
1525}
1526
ce9fbf16 1527static void write_sha1_file_prepare(const void *buf, unsigned long len,
972a9155 1528 const char *type, unsigned char *sha1,
d65a16f6 1529 char *hdr, int *hdrlen)
d410c0f5 1530{
9126f009 1531 git_SHA_CTX c;
d410c0f5
JH
1532
1533 /* Generate the header */
ef1286d3 1534 *hdrlen = xsnprintf(hdr, *hdrlen, "%s %lu", type, len)+1;
d410c0f5
JH
1535
1536 /* Sha1.. */
9126f009
NP
1537 git_SHA1_Init(&c);
1538 git_SHA1_Update(&c, hdr, *hdrlen);
1539 git_SHA1_Update(&c, buf, len);
1540 git_SHA1_Final(sha1, &c);
d410c0f5
JH
1541}
1542
230f1322 1543/*
5a688fe4 1544 * Move the just written object into its final resting place.
230f1322 1545 */
cb5add58 1546int finalize_object_file(const char *tmpfile, const char *filename)
230f1322 1547{
e32c0a9c 1548 int ret = 0;
5a688fe4 1549
348df166 1550 if (object_creation_mode == OBJECT_CREATION_USES_RENAMES)
be66a6c4
JS
1551 goto try_rename;
1552 else if (link(tmpfile, filename))
e32c0a9c 1553 ret = errno;
7ebb6fca
LT
1554
1555 /*
1556 * Coda hack - coda doesn't like cross-directory links,
1557 * so we fall back to a rename, which will mean that it
1558 * won't be able to check collisions, but that's not a
1559 * big deal.
1560 *
1561 * The same holds for FAT formatted media.
1562 *
3be1f18e 1563 * When this succeeds, we just return. We have nothing
7ebb6fca
LT
1564 * left to unlink.
1565 */
1566 if (ret && ret != EEXIST) {
be66a6c4 1567 try_rename:
7ebb6fca 1568 if (!rename(tmpfile, filename))
3be1f18e 1569 goto out;
9e48b389 1570 ret = errno;
230f1322 1571 }
691f1a28 1572 unlink_or_warn(tmpfile);
230f1322
LT
1573 if (ret) {
1574 if (ret != EEXIST) {
7616c6ca 1575 return error_errno("unable to write sha1 filename %s", filename);
230f1322
LT
1576 }
1577 /* FIXME!!! Collision check here ? */
1578 }
1579
3be1f18e 1580out:
5256b006 1581 if (adjust_shared_perm(filename))
5a688fe4 1582 return error("unable to set permission to '%s'", filename);
230f1322
LT
1583 return 0;
1584}
1585
4d548150
LT
1586static int write_buffer(int fd, const void *buf, size_t len)
1587{
d34cf19b 1588 if (write_in_full(fd, buf, len) < 0)
7616c6ca 1589 return error_errno("file write error");
4d548150
LT
1590 return 0;
1591}
1592
ce9fbf16 1593int hash_sha1_file(const void *buf, unsigned long len, const char *type,
abdc3fc8
RS
1594 unsigned char *sha1)
1595{
d65a16f6 1596 char hdr[32];
ef1286d3 1597 int hdrlen = sizeof(hdr);
abdc3fc8
RS
1598 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
1599 return 0;
1600}
1601
e9039dd3
LT
1602/* Finalize a file on disk, and close it. */
1603static void close_sha1_file(int fd)
1604{
aafe9fba
LT
1605 if (fsync_object_files)
1606 fsync_or_die(fd, "sha1 file");
e9039dd3 1607 if (close(fd) != 0)
d824cbba 1608 die_errno("error when closing sha1 file");
e9039dd3
LT
1609}
1610
5723fe7e
LT
1611/* Size of directory component, including the ending '/' */
1612static inline int directory_size(const char *filename)
1613{
1614 const char *s = strrchr(filename, '/');
1615 if (!s)
1616 return 0;
1617 return s - filename + 1;
1618}
1619
1620/*
1621 * This creates a temporary file in the same directory as the final
1622 * 'filename'
1623 *
1624 * We want to avoid cross-directory filename renames, because those
1625 * can have problems on various filesystems (FAT, NFS, Coda).
1626 */
d4b3d11a 1627static int create_tmpfile(struct strbuf *tmp, const char *filename)
5723fe7e
LT
1628{
1629 int fd, dirlen = directory_size(filename);
1630
d4b3d11a
JK
1631 strbuf_reset(tmp);
1632 strbuf_add(tmp, filename, dirlen);
1633 strbuf_addstr(tmp, "tmp_obj_XXXXXX");
1634 fd = git_mkstemp_mode(tmp->buf, 0444);
cbacbf4e 1635 if (fd < 0 && dirlen && errno == ENOENT) {
d4b3d11a
JK
1636 /*
1637 * Make sure the directory exists; note that the contents
1638 * of the buffer are undefined after mkstemp returns an
1639 * error, so we have to rewrite the whole buffer from
1640 * scratch.
1641 */
1642 strbuf_reset(tmp);
1643 strbuf_add(tmp, filename, dirlen - 1);
1644 if (mkdir(tmp->buf, 0777) && errno != EEXIST)
b2476a60 1645 return -1;
d4b3d11a 1646 if (adjust_shared_perm(tmp->buf))
5723fe7e
LT
1647 return -1;
1648
1649 /* Try again */
d4b3d11a
JK
1650 strbuf_addstr(tmp, "/tmp_obj_XXXXXX");
1651 fd = git_mkstemp_mode(tmp->buf, 0444);
5723fe7e
LT
1652 }
1653 return fd;
1654}
1655
bbac7311 1656static int write_loose_object(const unsigned char *sha1, char *hdr, int hdrlen,
c00e657d 1657 const void *buf, unsigned long len, time_t mtime)
0fcfd160 1658{
915308b1 1659 int fd, ret;
9892beba 1660 unsigned char compressed[4096];
ef49a7a0 1661 git_zstream stream;
748af44c
NP
1662 git_SHA_CTX c;
1663 unsigned char parano_sha1[20];
d4b3d11a 1664 static struct strbuf tmp_file = STRBUF_INIT;
30d6c6ea 1665 const char *filename = sha1_file_name(sha1);
a44c9a5e 1666
d4b3d11a 1667 fd = create_tmpfile(&tmp_file, filename);
aac17941 1668 if (fd < 0) {
35243577 1669 if (errno == EACCES)
82247e9b 1670 return error("insufficient permission for adding an object to repository database %s", get_object_directory());
916d081b 1671 else
7616c6ca 1672 return error_errno("unable to create temporary file");
aac17941
LT
1673 }
1674
0fcfd160 1675 /* Set it up */
55bb5c91 1676 git_deflate_init(&stream, zlib_compression_level);
0fcfd160 1677 stream.next_out = compressed;
9892beba 1678 stream.avail_out = sizeof(compressed);
748af44c 1679 git_SHA1_Init(&c);
a44c9a5e
LT
1680
1681 /* First header.. */
d65a16f6 1682 stream.next_in = (unsigned char *)hdr;
a44c9a5e 1683 stream.avail_in = hdrlen;
55bb5c91
JH
1684 while (git_deflate(&stream, 0) == Z_OK)
1685 ; /* nothing */
748af44c 1686 git_SHA1_Update(&c, hdr, hdrlen);
a44c9a5e
LT
1687
1688 /* Then the data itself.. */
c00e657d 1689 stream.next_in = (void *)buf;
a44c9a5e 1690 stream.avail_in = len;
9892beba 1691 do {
748af44c 1692 unsigned char *in0 = stream.next_in;
55bb5c91 1693 ret = git_deflate(&stream, Z_FINISH);
748af44c 1694 git_SHA1_Update(&c, in0, stream.next_in - in0);
9892beba
NP
1695 if (write_buffer(fd, compressed, stream.next_out - compressed) < 0)
1696 die("unable to write sha1 file");
1697 stream.next_out = compressed;
1698 stream.avail_out = sizeof(compressed);
1699 } while (ret == Z_OK);
1700
ac54c277
LT
1701 if (ret != Z_STREAM_END)
1702 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1), ret);
55bb5c91 1703 ret = git_deflate_end_gently(&stream);
ac54c277
LT
1704 if (ret != Z_OK)
1705 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1), ret);
748af44c
NP
1706 git_SHA1_Final(parano_sha1, &c);
1707 if (hashcmp(sha1, parano_sha1) != 0)
1708 die("confused by unstable object source data for %s", sha1_to_hex(sha1));
ac54c277 1709
e9039dd3 1710 close_sha1_file(fd);
0fcfd160 1711
bbac7311
NP
1712 if (mtime) {
1713 struct utimbuf utb;
1714 utb.actime = mtime;
1715 utb.modtime = mtime;
d4b3d11a 1716 if (utime(tmp_file.buf, &utb) < 0)
7616c6ca 1717 warning_errno("failed utime() on %s", tmp_file.buf);
bbac7311
NP
1718 }
1719
d4b3d11a 1720 return finalize_object_file(tmp_file.buf, filename);
0fcfd160 1721}
8237b185 1722
33d4221c
JK
1723static int freshen_loose_object(const unsigned char *sha1)
1724{
1725 return check_and_freshen(sha1, 1);
1726}
1727
1728static int freshen_packed_object(const unsigned char *sha1)
1729{
1730 struct pack_entry e;
ee1c6c34
JK
1731 if (!find_pack_entry(sha1, &e))
1732 return 0;
1733 if (e.p->freshened)
1734 return 1;
1735 if (!freshen_file(e.p->pack_name))
1736 return 0;
1737 e.p->freshened = 1;
1738 return 1;
33d4221c
JK
1739}
1740
1427a7ff 1741int write_sha1_file(const void *buf, unsigned long len, const char *type, unsigned char *sha1)
bbac7311 1742{
bbac7311 1743 char hdr[32];
ef1286d3 1744 int hdrlen = sizeof(hdr);
bbac7311
NP
1745
1746 /* Normally if we have it in the pack then we do not bother writing
1747 * it out into .git/objects/??/?{38} file.
1748 */
1749 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
b5f52f37 1750 if (freshen_packed_object(sha1) || freshen_loose_object(sha1))
bbac7311
NP
1751 return 0;
1752 return write_loose_object(sha1, hdr, hdrlen, buf, len, 0);
1753}
1754
0c3db67c
ES
1755int hash_sha1_file_literally(const void *buf, unsigned long len, const char *type,
1756 unsigned char *sha1, unsigned flags)
1757{
1758 char *header;
1759 int hdrlen, status = 0;
1760
1761 /* type string, SP, %lu of the length plus NUL must fit this */
ef1286d3
JK
1762 hdrlen = strlen(type) + 32;
1763 header = xmalloc(hdrlen);
0c3db67c
ES
1764 write_sha1_file_prepare(buf, len, type, sha1, header, &hdrlen);
1765
1766 if (!(flags & HASH_WRITE_OBJECT))
1767 goto cleanup;
051086b9 1768 if (freshen_packed_object(sha1) || freshen_loose_object(sha1))
0c3db67c
ES
1769 goto cleanup;
1770 status = write_loose_object(sha1, header, hdrlen, buf, len, 0);
1771
1772cleanup:
1773 free(header);
1774 return status;
1775}
1776
bbac7311
NP
1777int force_object_loose(const unsigned char *sha1, time_t mtime)
1778{
bbac7311
NP
1779 void *buf;
1780 unsigned long len;
1781 enum object_type type;
1782 char hdr[32];
1783 int hdrlen;
1fb23e65 1784 int ret;
bbac7311 1785
c529d75a 1786 if (has_loose_object(sha1))
bbac7311 1787 return 0;
789bf26b 1788 buf = read_object(sha1, &type, &len);
bbac7311
NP
1789 if (!buf)
1790 return error("cannot read sha1_file for %s", sha1_to_hex(sha1));
ef1286d3 1791 hdrlen = xsnprintf(hdr, sizeof(hdr), "%s %lu", typename(type), len) + 1;
1fb23e65
BS
1792 ret = write_loose_object(sha1, hdr, hdrlen, buf, len, mtime);
1793 free(buf);
1794
1795 return ret;
bbac7311
NP
1796}
1797
bf592c50
DB
1798int has_pack_index(const unsigned char *sha1)
1799{
1800 struct stat st;
1801 if (stat(sha1_pack_index_name(sha1), &st))
1802 return 0;
1803 return 1;
bf592c50
DB
1804}
1805
cd673c1f
JH
1806int has_sha1_pack(const unsigned char *sha1)
1807{
1808 struct pack_entry e;
1809 return find_pack_entry(sha1, &e);
1810}
1811
0eeb077b 1812int has_sha1_file_with_flags(const unsigned char *sha1, int flags)
8237b185 1813{
3e8b7d3c
JN
1814 if (!startup_info->have_repository)
1815 return 0;
e83e71c5
JT
1816 return sha1_object_info_extended(sha1, NULL,
1817 flags | OBJECT_INFO_SKIP_CACHED) >= 0;
b419aa25 1818}
1819
1820int has_object_file(const struct object_id *oid)
1821{
1822 return has_sha1_file(oid->hash);
8237b185 1823}
74400e71 1824
5827a035
JK
1825int has_object_file_with_flags(const struct object_id *oid, int flags)
1826{
1827 return has_sha1_file_with_flags(oid->hash, flags);
1828}
1829
c879daa2
NTND
1830static void check_tree(const void *buf, size_t size)
1831{
1832 struct tree_desc desc;
1833 struct name_entry entry;
1834
1835 init_tree_desc(&desc, buf, size);
1836 while (tree_entry(&desc, &entry))
1837 /* do nothing
1838 * tree_entry() will die() on malformed entries */
1839 ;
1840}
1841
1842static void check_commit(const void *buf, size_t size)
1843{
1844 struct commit c;
1845 memset(&c, 0, sizeof(c));
1846 if (parse_commit_buffer(&c, buf, size))
1847 die("corrupt commit");
1848}
1849
1850static void check_tag(const void *buf, size_t size)
1851{
1852 struct tag t;
1853 memset(&t, 0, sizeof(t));
1854 if (parse_tag_buffer(&t, buf, size))
1855 die("corrupt tag");
1856}
1857
43df4f86 1858static int index_mem(unsigned char *sha1, void *buf, size_t size,
c4ce46fc
JH
1859 enum object_type type,
1860 const char *path, unsigned flags)
e7332f96 1861{
6c510bee 1862 int ret, re_allocated = 0;
c4ce46fc 1863 int write_object = flags & HASH_WRITE_OBJECT;
74400e71 1864
7672db20 1865 if (!type)
edaec3fb 1866 type = OBJ_BLOB;
6c510bee
LT
1867
1868 /*
1869 * Convert blobs to git internal format
1870 */
43df4f86 1871 if ((type == OBJ_BLOB) && path) {
f285a2d7 1872 struct strbuf nbuf = STRBUF_INIT;
82b474e0 1873 if (convert_to_git(&the_index, path, buf, size, &nbuf,
5e12e78e 1874 write_object ? safe_crlf : SAFE_CRLF_FALSE)) {
b315c5c0 1875 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
1876 re_allocated = 1;
1877 }
1878 }
c4ce46fc 1879 if (flags & HASH_FORMAT_CHECK) {
c879daa2
NTND
1880 if (type == OBJ_TREE)
1881 check_tree(buf, size);
1882 if (type == OBJ_COMMIT)
1883 check_commit(buf, size);
1884 if (type == OBJ_TAG)
1885 check_tag(buf, size);
1886 }
6c510bee 1887
7672db20 1888 if (write_object)
edaec3fb 1889 ret = write_sha1_file(buf, size, typename(type), sha1);
abdc3fc8 1890 else
edaec3fb 1891 ret = hash_sha1_file(buf, size, typename(type), sha1);
43df4f86 1892 if (re_allocated)
6c510bee 1893 free(buf);
43df4f86
DP
1894 return ret;
1895}
1896
9035d75a
SP
1897static int index_stream_convert_blob(unsigned char *sha1, int fd,
1898 const char *path, unsigned flags)
1899{
1900 int ret;
1901 const int write_object = flags & HASH_WRITE_OBJECT;
1902 struct strbuf sbuf = STRBUF_INIT;
1903
1904 assert(path);
1905 assert(would_convert_to_git_filter_fd(path));
1906
d6c41c20 1907 convert_to_git_filter_fd(&the_index, path, fd, &sbuf,
9035d75a
SP
1908 write_object ? safe_crlf : SAFE_CRLF_FALSE);
1909
1910 if (write_object)
1911 ret = write_sha1_file(sbuf.buf, sbuf.len, typename(OBJ_BLOB),
1912 sha1);
1913 else
1914 ret = hash_sha1_file(sbuf.buf, sbuf.len, typename(OBJ_BLOB),
1915 sha1);
1916 strbuf_release(&sbuf);
1917 return ret;
1918}
1919
7b41e1e1
JH
1920static int index_pipe(unsigned char *sha1, int fd, enum object_type type,
1921 const char *path, unsigned flags)
1922{
1923 struct strbuf sbuf = STRBUF_INIT;
1924 int ret;
1925
1926 if (strbuf_read(&sbuf, fd, 4096) >= 0)
1927 ret = index_mem(sha1, sbuf.buf, sbuf.len, type, path, flags);
1928 else
1929 ret = -1;
1930 strbuf_release(&sbuf);
1931 return ret;
1932}
1933
ea68b0ce
DP
1934#define SMALL_FILE_SIZE (32*1024)
1935
7b41e1e1
JH
1936static int index_core(unsigned char *sha1, int fd, size_t size,
1937 enum object_type type, const char *path,
1938 unsigned flags)
43df4f86
DP
1939{
1940 int ret;
43df4f86 1941
7b41e1e1 1942 if (!size) {
f6a1e1e2 1943 ret = index_mem(sha1, "", size, type, path, flags);
ea68b0ce
DP
1944 } else if (size <= SMALL_FILE_SIZE) {
1945 char *buf = xmalloc(size);
1946 if (size == read_in_full(fd, buf, size))
c4ce46fc 1947 ret = index_mem(sha1, buf, size, type, path, flags);
ea68b0ce 1948 else
7616c6ca 1949 ret = error_errno("short read");
ea68b0ce 1950 free(buf);
08bda208 1951 } else {
43df4f86 1952 void *buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
c4ce46fc 1953 ret = index_mem(sha1, buf, size, type, path, flags);
aac17941 1954 munmap(buf, size);
08bda208 1955 }
7b41e1e1
JH
1956 return ret;
1957}
1958
4dd1fbc7 1959/*
568508e7
JH
1960 * This creates one packfile per large blob unless bulk-checkin
1961 * machinery is "plugged".
4dd1fbc7
JH
1962 *
1963 * This also bypasses the usual "convert-to-git" dance, and that is on
1964 * purpose. We could write a streaming version of the converting
1965 * functions and insert that before feeding the data to fast-import
4f22b101
JK
1966 * (or equivalent in-core API described above). However, that is
1967 * somewhat complicated, as we do not know the size of the filter
1968 * result, which we need to know beforehand when writing a git object.
1969 * Since the primary motivation for trying to stream from the working
1970 * tree file and to avoid mmaping it in core is to deal with large
1971 * binary blobs, they generally do not want to get any conversion, and
1972 * callers should avoid this code path when filters are requested.
4dd1fbc7
JH
1973 */
1974static int index_stream(unsigned char *sha1, int fd, size_t size,
1975 enum object_type type, const char *path,
1976 unsigned flags)
1977{
568508e7 1978 return index_bulk_checkin(sha1, fd, size, type, path, flags);
4dd1fbc7
JH
1979}
1980
7b41e1e1
JH
1981int index_fd(unsigned char *sha1, int fd, struct stat *st,
1982 enum object_type type, const char *path, unsigned flags)
1983{
1984 int ret;
7b41e1e1 1985
9079ab7c
SP
1986 /*
1987 * Call xsize_t() only when needed to avoid potentially unnecessary
1988 * die() for large files.
1989 */
9035d75a
SP
1990 if (type == OBJ_BLOB && path && would_convert_to_git_filter_fd(path))
1991 ret = index_stream_convert_blob(sha1, fd, path, flags);
1992 else if (!S_ISREG(st->st_mode))
7b41e1e1 1993 ret = index_pipe(sha1, fd, type, path, flags);
9079ab7c 1994 else if (st->st_size <= big_file_threshold || type != OBJ_BLOB ||
82b474e0 1995 (path && would_convert_to_git(&the_index, path)))
9079ab7c
SP
1996 ret = index_core(sha1, fd, xsize_t(st->st_size), type, path,
1997 flags);
4dd1fbc7 1998 else
9079ab7c
SP
1999 ret = index_stream(sha1, fd, xsize_t(st->st_size), type, path,
2000 flags);
43df4f86 2001 close(fd);
aac17941 2002 return ret;
74400e71 2003}
ec1fcc16 2004
c4ce46fc 2005int index_path(unsigned char *sha1, const char *path, struct stat *st, unsigned flags)
ec1fcc16
JH
2006{
2007 int fd;
b760d3aa 2008 struct strbuf sb = STRBUF_INIT;
ec1fcc16
JH
2009
2010 switch (st->st_mode & S_IFMT) {
2011 case S_IFREG:
2012 fd = open(path, O_RDONLY);
2013 if (fd < 0)
7616c6ca 2014 return error_errno("open(\"%s\")", path);
c4ce46fc 2015 if (index_fd(sha1, fd, st, OBJ_BLOB, path, flags) < 0)
ec1fcc16
JH
2016 return error("%s: failed to insert into database",
2017 path);
2018 break;
2019 case S_IFLNK:
7616c6ca
NTND
2020 if (strbuf_readlink(&sb, path, st->st_size))
2021 return error_errno("readlink(\"%s\")", path);
c4ce46fc 2022 if (!(flags & HASH_WRITE_OBJECT))
b760d3aa
LT
2023 hash_sha1_file(sb.buf, sb.len, blob_type, sha1);
2024 else if (write_sha1_file(sb.buf, sb.len, blob_type, sha1))
ec1fcc16
JH
2025 return error("%s: failed to insert into database",
2026 path);
b760d3aa 2027 strbuf_release(&sb);
ec1fcc16 2028 break;
f35a6d3b
LT
2029 case S_IFDIR:
2030 return resolve_gitlink_ref(path, "HEAD", sha1);
ec1fcc16
JH
2031 default:
2032 return error("%s: unsupported file type", path);
2033 }
2034 return 0;
2035}
a69e5429
JH
2036
2037int read_pack_header(int fd, struct pack_header *header)
2038{
c697ad14
HO
2039 if (read_in_full(fd, header, sizeof(*header)) < sizeof(*header))
2040 /* "eof before pack header was fully read" */
2041 return PH_ERROR_EOF;
2042
a69e5429
JH
2043 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2044 /* "protocol error (pack signature mismatch detected)" */
2045 return PH_ERROR_PACK_SIGNATURE;
2046 if (!pack_version_ok(header->hdr_version))
2047 /* "protocol error (pack version unsupported)" */
2048 return PH_ERROR_PROTOCOL;
2049 return 0;
2050}
40d52ff7
JK
2051
2052void assert_sha1_type(const unsigned char *sha1, enum object_type expect)
2053{
2054 enum object_type type = sha1_object_info(sha1, NULL);
2055 if (type < 0)
2056 die("%s is not a valid object", sha1_to_hex(sha1));
2057 if (type != expect)
2058 die("%s is not a valid '%s' object", sha1_to_hex(sha1),
2059 typename(expect));
2060}
27e1e22d 2061
70c49050 2062int for_each_file_in_obj_subdir(unsigned int subdir_nr,
cc817ca3
RS
2063 struct strbuf *path,
2064 each_loose_object_fn obj_cb,
2065 each_loose_cruft_fn cruft_cb,
2066 each_loose_subdir_fn subdir_cb,
2067 void *data)
27e1e22d 2068{
0375f472
RS
2069 size_t origlen, baselen;
2070 DIR *dir;
27e1e22d
JK
2071 struct dirent *de;
2072 int r = 0;
2073
70c49050
RS
2074 if (subdir_nr > 0xff)
2075 BUG("invalid loose object subdirectory: %x", subdir_nr);
2076
0375f472
RS
2077 origlen = path->len;
2078 strbuf_complete(path, '/');
2079 strbuf_addf(path, "%02x", subdir_nr);
2080 baselen = path->len;
2081
2082 dir = opendir(path->buf);
27e1e22d 2083 if (!dir) {
0375f472
RS
2084 if (errno != ENOENT)
2085 r = error_errno("unable to open %s", path->buf);
2086 strbuf_setlen(path, origlen);
2087 return r;
27e1e22d
JK
2088 }
2089
2090 while ((de = readdir(dir))) {
2091 if (is_dot_or_dotdot(de->d_name))
2092 continue;
2093
2094 strbuf_setlen(path, baselen);
2095 strbuf_addf(path, "/%s", de->d_name);
2096
76c1d9a0 2097 if (strlen(de->d_name) == GIT_SHA1_HEXSZ - 2) {
dc01505f 2098 char hex[GIT_MAX_HEXSZ+1];
76c1d9a0 2099 struct object_id oid;
27e1e22d 2100
1a168e5c
JK
2101 xsnprintf(hex, sizeof(hex), "%02x%s",
2102 subdir_nr, de->d_name);
76c1d9a0 2103 if (!get_oid_hex(hex, &oid)) {
27e1e22d 2104 if (obj_cb) {
76c1d9a0 2105 r = obj_cb(&oid, path->buf, data);
27e1e22d
JK
2106 if (r)
2107 break;
2108 }
2109 continue;
2110 }
2111 }
2112
2113 if (cruft_cb) {
2114 r = cruft_cb(de->d_name, path->buf, data);
2115 if (r)
2116 break;
2117 }
2118 }
094c7e63 2119 closedir(dir);
27e1e22d 2120
094c7e63 2121 strbuf_setlen(path, baselen);
27e1e22d
JK
2122 if (!r && subdir_cb)
2123 r = subdir_cb(subdir_nr, path->buf, data);
2124
0375f472
RS
2125 strbuf_setlen(path, origlen);
2126
27e1e22d
JK
2127 return r;
2128}
2129
e6f875e0 2130int for_each_loose_file_in_objdir_buf(struct strbuf *path,
27e1e22d
JK
2131 each_loose_object_fn obj_cb,
2132 each_loose_cruft_fn cruft_cb,
2133 each_loose_subdir_fn subdir_cb,
2134 void *data)
2135{
27e1e22d
JK
2136 int r = 0;
2137 int i;
2138
27e1e22d 2139 for (i = 0; i < 256; i++) {
e6f875e0 2140 r = for_each_file_in_obj_subdir(i, path, obj_cb, cruft_cb,
27e1e22d 2141 subdir_cb, data);
27e1e22d
JK
2142 if (r)
2143 break;
2144 }
2145
e6f875e0
JK
2146 return r;
2147}
2148
2149int for_each_loose_file_in_objdir(const char *path,
2150 each_loose_object_fn obj_cb,
2151 each_loose_cruft_fn cruft_cb,
2152 each_loose_subdir_fn subdir_cb,
2153 void *data)
2154{
2155 struct strbuf buf = STRBUF_INIT;
2156 int r;
2157
2158 strbuf_addstr(&buf, path);
2159 r = for_each_loose_file_in_objdir_buf(&buf, obj_cb, cruft_cb,
2160 subdir_cb, data);
27e1e22d 2161 strbuf_release(&buf);
e6f875e0 2162
27e1e22d
JK
2163 return r;
2164}
660c889e
JK
2165
2166struct loose_alt_odb_data {
2167 each_loose_object_fn *cb;
2168 void *data;
2169};
2170
2171static int loose_from_alt_odb(struct alternate_object_database *alt,
2172 void *vdata)
2173{
2174 struct loose_alt_odb_data *data = vdata;
b0a42642
JM
2175 struct strbuf buf = STRBUF_INIT;
2176 int r;
2177
597f9134 2178 strbuf_addstr(&buf, alt->path);
b0a42642
JM
2179 r = for_each_loose_file_in_objdir_buf(&buf,
2180 data->cb, NULL, NULL,
2181 data->data);
2182 strbuf_release(&buf);
2183 return r;
660c889e
JK
2184}
2185
1385bb7b 2186int for_each_loose_object(each_loose_object_fn cb, void *data, unsigned flags)
660c889e
JK
2187{
2188 struct loose_alt_odb_data alt;
2189 int r;
2190
2191 r = for_each_loose_file_in_objdir(get_object_directory(),
2192 cb, NULL, NULL, data);
2193 if (r)
2194 return r;
2195
1385bb7b
JK
2196 if (flags & FOR_EACH_OBJECT_LOCAL_ONLY)
2197 return 0;
2198
660c889e
JK
2199 alt.cb = cb;
2200 alt.data = data;
2201 return foreach_alt_odb(loose_from_alt_odb, &alt);
2202}
2203
2204static int for_each_object_in_pack(struct packed_git *p, each_packed_object_fn cb, void *data)
2205{
2206 uint32_t i;
2207 int r = 0;
2208
2209 for (i = 0; i < p->num_objects; i++) {
068f85e3 2210 struct object_id oid;
660c889e 2211
068f85e3 2212 if (!nth_packed_object_oid(&oid, p, i))
660c889e
JK
2213 return error("unable to get sha1 of object %u in %s",
2214 i, p->pack_name);
2215
76c1d9a0 2216 r = cb(&oid, p, i, data);
660c889e
JK
2217 if (r)
2218 break;
2219 }
2220 return r;
2221}
2222
1385bb7b 2223int for_each_packed_object(each_packed_object_fn cb, void *data, unsigned flags)
660c889e
JK
2224{
2225 struct packed_git *p;
2226 int r = 0;
f813e9ea 2227 int pack_errors = 0;
660c889e
JK
2228
2229 prepare_packed_git();
2230 for (p = packed_git; p; p = p->next) {
1385bb7b
JK
2231 if ((flags & FOR_EACH_OBJECT_LOCAL_ONLY) && !p->pack_local)
2232 continue;
f813e9ea
JK
2233 if (open_pack_index(p)) {
2234 pack_errors = 1;
2235 continue;
2236 }
660c889e
JK
2237 r = for_each_object_in_pack(p, cb, data);
2238 if (r)
2239 break;
2240 }
f813e9ea 2241 return r ? r : pack_errors;
660c889e 2242}
f6371f92
JK
2243
2244static int check_stream_sha1(git_zstream *stream,
2245 const char *hdr,
2246 unsigned long size,
2247 const char *path,
2248 const unsigned char *expected_sha1)
2249{
2250 git_SHA_CTX c;
cd02599c 2251 unsigned char real_sha1[GIT_MAX_RAWSZ];
f6371f92
JK
2252 unsigned char buf[4096];
2253 unsigned long total_read;
2254 int status = Z_OK;
2255
2256 git_SHA1_Init(&c);
2257 git_SHA1_Update(&c, hdr, stream->total_out);
2258
2259 /*
2260 * We already read some bytes into hdr, but the ones up to the NUL
2261 * do not count against the object's content size.
2262 */
2263 total_read = stream->total_out - strlen(hdr) - 1;
2264
2265 /*
2266 * This size comparison must be "<=" to read the final zlib packets;
2267 * see the comment in unpack_sha1_rest for details.
2268 */
2269 while (total_read <= size &&
2270 (status == Z_OK || status == Z_BUF_ERROR)) {
2271 stream->next_out = buf;
2272 stream->avail_out = sizeof(buf);
2273 if (size - total_read < stream->avail_out)
2274 stream->avail_out = size - total_read;
2275 status = git_inflate(stream, Z_FINISH);
2276 git_SHA1_Update(&c, buf, stream->next_out - buf);
2277 total_read += stream->next_out - buf;
2278 }
2279 git_inflate_end(stream);
2280
2281 if (status != Z_STREAM_END) {
2282 error("corrupt loose object '%s'", sha1_to_hex(expected_sha1));
2283 return -1;
2284 }
cce044df
JK
2285 if (stream->avail_in) {
2286 error("garbage at end of loose object '%s'",
2287 sha1_to_hex(expected_sha1));
2288 return -1;
2289 }
f6371f92
JK
2290
2291 git_SHA1_Final(real_sha1, &c);
2292 if (hashcmp(expected_sha1, real_sha1)) {
2293 error("sha1 mismatch for %s (expected %s)", path,
2294 sha1_to_hex(expected_sha1));
2295 return -1;
2296 }
2297
2298 return 0;
2299}
2300
2301int read_loose_object(const char *path,
2302 const unsigned char *expected_sha1,
2303 enum object_type *type,
2304 unsigned long *size,
2305 void **contents)
2306{
2307 int ret = -1;
f6371f92
JK
2308 void *map = NULL;
2309 unsigned long mapsize;
2310 git_zstream stream;
2311 char hdr[32];
2312
2313 *contents = NULL;
2314
2315 map = map_sha1_file_1(path, NULL, &mapsize);
2316 if (!map) {
2317 error_errno("unable to mmap %s", path);
2318 goto out;
2319 }
2320
2321 if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0) {
2322 error("unable to unpack header of %s", path);
2323 goto out;
2324 }
2325
2326 *type = parse_sha1_header(hdr, size);
2327 if (*type < 0) {
2328 error("unable to parse header of %s", path);
2329 git_inflate_end(&stream);
2330 goto out;
2331 }
2332
2333 if (*type == OBJ_BLOB) {
2334 if (check_stream_sha1(&stream, hdr, *size, path, expected_sha1) < 0)
2335 goto out;
2336 } else {
2337 *contents = unpack_sha1_rest(&stream, hdr, *size, expected_sha1);
2338 if (!*contents) {
2339 error("unable to unpack contents of %s", path);
2340 git_inflate_end(&stream);
2341 goto out;
2342 }
2343 if (check_sha1_signature(expected_sha1, *contents,
2344 *size, typename(*type))) {
2345 error("sha1 mismatch for %s (expected %s)", path,
2346 sha1_to_hex(expected_sha1));
2347 free(*contents);
2348 goto out;
2349 }
2350 }
2351
2352 ret = 0; /* everything checks out */
2353
2354out:
2355 if (map)
2356 munmap(map, mapsize);
f6371f92
JK
2357 return ret;
2358}