09456d23f8447711f783178ff4ea5af8f92828e2
[git/git.git] / sha1_file.c
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 */
9 #include "cache.h"
10 #include "delta.h"
11 #include "pack.h"
12 #include "blob.h"
13 #include "commit.h"
14 #include "tag.h"
15 #include "tree.h"
16
17 #ifndef O_NOATIME
18 #if defined(__linux__) && (defined(__i386__) || defined(__PPC__))
19 #define O_NOATIME 01000000
20 #else
21 #define O_NOATIME 0
22 #endif
23 #endif
24
25 const unsigned char null_sha1[20];
26
27 static unsigned int sha1_file_open_flag = O_NOATIME;
28
29 signed char hexval_table[256] = {
30 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
31 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
32 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
33 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
34 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
35 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
36 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
37 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
38 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
39 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
40 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
41 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
42 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
43 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
44 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
45 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
46 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
47 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
48 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
49 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
50 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
51 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
52 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
53 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
54 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
55 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
56 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
57 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
58 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
59 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
60 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
61 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
62 };
63
64 int get_sha1_hex(const char *hex, unsigned char *sha1)
65 {
66 int i;
67 for (i = 0; i < 20; i++) {
68 unsigned int val = (hexval(hex[0]) << 4) | hexval(hex[1]);
69 if (val & ~0xff)
70 return -1;
71 *sha1++ = val;
72 hex += 2;
73 }
74 return 0;
75 }
76
77 int safe_create_leading_directories(char *path)
78 {
79 char *pos = path;
80 struct stat st;
81
82 if (*pos == '/')
83 pos++;
84
85 while (pos) {
86 pos = strchr(pos, '/');
87 if (!pos)
88 break;
89 *pos = 0;
90 if (!stat(path, &st)) {
91 /* path exists */
92 if (!S_ISDIR(st.st_mode)) {
93 *pos = '/';
94 return -3;
95 }
96 }
97 else if (mkdir(path, 0777)) {
98 *pos = '/';
99 return -1;
100 }
101 else if (adjust_shared_perm(path)) {
102 *pos = '/';
103 return -2;
104 }
105 *pos++ = '/';
106 }
107 return 0;
108 }
109
110 char * sha1_to_hex(const unsigned char *sha1)
111 {
112 static int bufno;
113 static char hexbuffer[4][50];
114 static const char hex[] = "0123456789abcdef";
115 char *buffer = hexbuffer[3 & ++bufno], *buf = buffer;
116 int i;
117
118 for (i = 0; i < 20; i++) {
119 unsigned int val = *sha1++;
120 *buf++ = hex[val >> 4];
121 *buf++ = hex[val & 0xf];
122 }
123 *buf = '\0';
124
125 return buffer;
126 }
127
128 static void fill_sha1_path(char *pathbuf, const unsigned char *sha1)
129 {
130 int i;
131 for (i = 0; i < 20; i++) {
132 static char hex[] = "0123456789abcdef";
133 unsigned int val = sha1[i];
134 char *pos = pathbuf + i*2 + (i > 0);
135 *pos++ = hex[val >> 4];
136 *pos = hex[val & 0xf];
137 }
138 }
139
140 /*
141 * NOTE! This returns a statically allocated buffer, so you have to be
142 * careful about using it. Do a "xstrdup()" if you need to save the
143 * filename.
144 *
145 * Also note that this returns the location for creating. Reading
146 * SHA1 file can happen from any alternate directory listed in the
147 * DB_ENVIRONMENT environment variable if it is not found in
148 * the primary object database.
149 */
150 char *sha1_file_name(const unsigned char *sha1)
151 {
152 static char *name, *base;
153
154 if (!base) {
155 const char *sha1_file_directory = get_object_directory();
156 int len = strlen(sha1_file_directory);
157 base = xmalloc(len + 60);
158 memcpy(base, sha1_file_directory, len);
159 memset(base+len, 0, 60);
160 base[len] = '/';
161 base[len+3] = '/';
162 name = base + len + 1;
163 }
164 fill_sha1_path(name, sha1);
165 return base;
166 }
167
168 char *sha1_pack_name(const unsigned char *sha1)
169 {
170 static const char hex[] = "0123456789abcdef";
171 static char *name, *base, *buf;
172 int i;
173
174 if (!base) {
175 const char *sha1_file_directory = get_object_directory();
176 int len = strlen(sha1_file_directory);
177 base = xmalloc(len + 60);
178 sprintf(base, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory);
179 name = base + len + 11;
180 }
181
182 buf = name;
183
184 for (i = 0; i < 20; i++) {
185 unsigned int val = *sha1++;
186 *buf++ = hex[val >> 4];
187 *buf++ = hex[val & 0xf];
188 }
189
190 return base;
191 }
192
193 char *sha1_pack_index_name(const unsigned char *sha1)
194 {
195 static const char hex[] = "0123456789abcdef";
196 static char *name, *base, *buf;
197 int i;
198
199 if (!base) {
200 const char *sha1_file_directory = get_object_directory();
201 int len = strlen(sha1_file_directory);
202 base = xmalloc(len + 60);
203 sprintf(base, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory);
204 name = base + len + 11;
205 }
206
207 buf = name;
208
209 for (i = 0; i < 20; i++) {
210 unsigned int val = *sha1++;
211 *buf++ = hex[val >> 4];
212 *buf++ = hex[val & 0xf];
213 }
214
215 return base;
216 }
217
218 struct alternate_object_database *alt_odb_list;
219 static struct alternate_object_database **alt_odb_tail;
220
221 static void read_info_alternates(const char * alternates, int depth);
222
223 /*
224 * Prepare alternate object database registry.
225 *
226 * The variable alt_odb_list points at the list of struct
227 * alternate_object_database. The elements on this list come from
228 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
229 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
230 * whose contents is similar to that environment variable but can be
231 * LF separated. Its base points at a statically allocated buffer that
232 * contains "/the/directory/corresponding/to/.git/objects/...", while
233 * its name points just after the slash at the end of ".git/objects/"
234 * in the example above, and has enough space to hold 40-byte hex
235 * SHA1, an extra slash for the first level indirection, and the
236 * terminating NUL.
237 */
238 static int link_alt_odb_entry(const char * entry, int len, const char * relative_base, int depth)
239 {
240 struct stat st;
241 const char *objdir = get_object_directory();
242 struct alternate_object_database *ent;
243 struct alternate_object_database *alt;
244 /* 43 = 40-byte + 2 '/' + terminating NUL */
245 int pfxlen = len;
246 int entlen = pfxlen + 43;
247 int base_len = -1;
248
249 if (*entry != '/' && relative_base) {
250 /* Relative alt-odb */
251 if (base_len < 0)
252 base_len = strlen(relative_base) + 1;
253 entlen += base_len;
254 pfxlen += base_len;
255 }
256 ent = xmalloc(sizeof(*ent) + entlen);
257
258 if (*entry != '/' && relative_base) {
259 memcpy(ent->base, relative_base, base_len - 1);
260 ent->base[base_len - 1] = '/';
261 memcpy(ent->base + base_len, entry, len);
262 }
263 else
264 memcpy(ent->base, entry, pfxlen);
265
266 ent->name = ent->base + pfxlen + 1;
267 ent->base[pfxlen + 3] = '/';
268 ent->base[pfxlen] = ent->base[entlen-1] = 0;
269
270 /* Detect cases where alternate disappeared */
271 if (stat(ent->base, &st) || !S_ISDIR(st.st_mode)) {
272 error("object directory %s does not exist; "
273 "check .git/objects/info/alternates.",
274 ent->base);
275 free(ent);
276 return -1;
277 }
278
279 /* Prevent the common mistake of listing the same
280 * thing twice, or object directory itself.
281 */
282 for (alt = alt_odb_list; alt; alt = alt->next) {
283 if (!memcmp(ent->base, alt->base, pfxlen)) {
284 free(ent);
285 return -1;
286 }
287 }
288 if (!memcmp(ent->base, objdir, pfxlen)) {
289 free(ent);
290 return -1;
291 }
292
293 /* add the alternate entry */
294 *alt_odb_tail = ent;
295 alt_odb_tail = &(ent->next);
296 ent->next = NULL;
297
298 /* recursively add alternates */
299 read_info_alternates(ent->base, depth + 1);
300
301 ent->base[pfxlen] = '/';
302
303 return 0;
304 }
305
306 static void link_alt_odb_entries(const char *alt, const char *ep, int sep,
307 const char *relative_base, int depth)
308 {
309 const char *cp, *last;
310
311 if (depth > 5) {
312 error("%s: ignoring alternate object stores, nesting too deep.",
313 relative_base);
314 return;
315 }
316
317 last = alt;
318 while (last < ep) {
319 cp = last;
320 if (cp < ep && *cp == '#') {
321 while (cp < ep && *cp != sep)
322 cp++;
323 last = cp + 1;
324 continue;
325 }
326 while (cp < ep && *cp != sep)
327 cp++;
328 if (last != cp) {
329 if ((*last != '/') && depth) {
330 error("%s: ignoring relative alternate object store %s",
331 relative_base, last);
332 } else {
333 link_alt_odb_entry(last, cp - last,
334 relative_base, depth);
335 }
336 }
337 while (cp < ep && *cp == sep)
338 cp++;
339 last = cp;
340 }
341 }
342
343 static void read_info_alternates(const char * relative_base, int depth)
344 {
345 char *map;
346 struct stat st;
347 char path[PATH_MAX];
348 int fd;
349
350 sprintf(path, "%s/info/alternates", relative_base);
351 fd = open(path, O_RDONLY);
352 if (fd < 0)
353 return;
354 if (fstat(fd, &st) || (st.st_size == 0)) {
355 close(fd);
356 return;
357 }
358 map = mmap(NULL, st.st_size, PROT_READ, MAP_PRIVATE, fd, 0);
359 close(fd);
360 if (map == MAP_FAILED)
361 return;
362
363 link_alt_odb_entries(map, map + st.st_size, '\n', relative_base, depth);
364
365 munmap(map, st.st_size);
366 }
367
368 void prepare_alt_odb(void)
369 {
370 const char *alt;
371
372 alt = getenv(ALTERNATE_DB_ENVIRONMENT);
373 if (!alt) alt = "";
374
375 if (alt_odb_tail)
376 return;
377 alt_odb_tail = &alt_odb_list;
378 link_alt_odb_entries(alt, alt + strlen(alt), ':', NULL, 0);
379
380 read_info_alternates(get_object_directory(), 0);
381 }
382
383 static char *find_sha1_file(const unsigned char *sha1, struct stat *st)
384 {
385 char *name = sha1_file_name(sha1);
386 struct alternate_object_database *alt;
387
388 if (!stat(name, st))
389 return name;
390 prepare_alt_odb();
391 for (alt = alt_odb_list; alt; alt = alt->next) {
392 name = alt->name;
393 fill_sha1_path(name, sha1);
394 if (!stat(alt->base, st))
395 return alt->base;
396 }
397 return NULL;
398 }
399
400 #define PACK_MAX_SZ (1<<26)
401 static int pack_used_ctr;
402 static unsigned long pack_mapped;
403 struct packed_git *packed_git;
404
405 static int check_packed_git_idx(const char *path, unsigned long *idx_size_,
406 void **idx_map_)
407 {
408 void *idx_map;
409 unsigned int *index;
410 unsigned long idx_size;
411 int nr, i;
412 int fd = open(path, O_RDONLY);
413 struct stat st;
414 if (fd < 0)
415 return -1;
416 if (fstat(fd, &st)) {
417 close(fd);
418 return -1;
419 }
420 idx_size = st.st_size;
421 idx_map = mmap(NULL, idx_size, PROT_READ, MAP_PRIVATE, fd, 0);
422 close(fd);
423 if (idx_map == MAP_FAILED)
424 return -1;
425
426 index = idx_map;
427 *idx_map_ = idx_map;
428 *idx_size_ = idx_size;
429
430 /* check index map */
431 if (idx_size < 4*256 + 20 + 20)
432 return error("index file too small");
433 nr = 0;
434 for (i = 0; i < 256; i++) {
435 unsigned int n = ntohl(index[i]);
436 if (n < nr)
437 return error("non-monotonic index");
438 nr = n;
439 }
440
441 /*
442 * Total size:
443 * - 256 index entries 4 bytes each
444 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
445 * - 20-byte SHA1 of the packfile
446 * - 20-byte SHA1 file checksum
447 */
448 if (idx_size != 4*256 + nr * 24 + 20 + 20)
449 return error("wrong index file size");
450
451 return 0;
452 }
453
454 static int unuse_one_packed_git(void)
455 {
456 struct packed_git *p, *lru = NULL;
457
458 for (p = packed_git; p; p = p->next) {
459 if (p->pack_use_cnt || !p->pack_base)
460 continue;
461 if (!lru || p->pack_last_used < lru->pack_last_used)
462 lru = p;
463 }
464 if (!lru)
465 return 0;
466 munmap(lru->pack_base, lru->pack_size);
467 lru->pack_base = NULL;
468 return 1;
469 }
470
471 void unuse_packed_git(struct packed_git *p)
472 {
473 p->pack_use_cnt--;
474 }
475
476 int use_packed_git(struct packed_git *p)
477 {
478 if (!p->pack_size) {
479 struct stat st;
480 /* We created the struct before we had the pack */
481 stat(p->pack_name, &st);
482 if (!S_ISREG(st.st_mode))
483 die("packfile %s not a regular file", p->pack_name);
484 p->pack_size = st.st_size;
485 }
486 if (!p->pack_base) {
487 int fd;
488 struct stat st;
489 void *map;
490 struct pack_header *hdr;
491
492 pack_mapped += p->pack_size;
493 while (PACK_MAX_SZ < pack_mapped && unuse_one_packed_git())
494 ; /* nothing */
495 fd = open(p->pack_name, O_RDONLY);
496 if (fd < 0)
497 die("packfile %s cannot be opened", p->pack_name);
498 if (fstat(fd, &st)) {
499 close(fd);
500 die("packfile %s cannot be opened", p->pack_name);
501 }
502 if (st.st_size != p->pack_size)
503 die("packfile %s size mismatch.", p->pack_name);
504 map = mmap(NULL, p->pack_size, PROT_READ, MAP_PRIVATE, fd, 0);
505 close(fd);
506 if (map == MAP_FAILED)
507 die("packfile %s cannot be mapped.", p->pack_name);
508 p->pack_base = map;
509
510 /* Check if we understand this pack file. If we don't we're
511 * likely too old to handle it.
512 */
513 hdr = map;
514 if (hdr->hdr_signature != htonl(PACK_SIGNATURE))
515 die("packfile %s isn't actually a pack.", p->pack_name);
516 if (!pack_version_ok(hdr->hdr_version))
517 die("packfile %s is version %i and not supported"
518 " (try upgrading GIT to a newer version)",
519 p->pack_name, ntohl(hdr->hdr_version));
520
521 /* Check if the pack file matches with the index file.
522 * this is cheap.
523 */
524 if (hashcmp((unsigned char *)(p->index_base) +
525 p->index_size - 40,
526 (unsigned char *)p->pack_base +
527 p->pack_size - 20)) {
528 die("packfile %s does not match index.", p->pack_name);
529 }
530 }
531 p->pack_last_used = pack_used_ctr++;
532 p->pack_use_cnt++;
533 return 0;
534 }
535
536 struct packed_git *add_packed_git(char *path, int path_len, int local)
537 {
538 struct stat st;
539 struct packed_git *p;
540 unsigned long idx_size;
541 void *idx_map;
542 unsigned char sha1[20];
543
544 if (check_packed_git_idx(path, &idx_size, &idx_map))
545 return NULL;
546
547 /* do we have a corresponding .pack file? */
548 strcpy(path + path_len - 4, ".pack");
549 if (stat(path, &st) || !S_ISREG(st.st_mode)) {
550 munmap(idx_map, idx_size);
551 return NULL;
552 }
553 /* ok, it looks sane as far as we can check without
554 * actually mapping the pack file.
555 */
556 p = xmalloc(sizeof(*p) + path_len + 2);
557 strcpy(p->pack_name, path);
558 p->index_size = idx_size;
559 p->pack_size = st.st_size;
560 p->index_base = idx_map;
561 p->next = NULL;
562 p->pack_base = NULL;
563 p->pack_last_used = 0;
564 p->pack_use_cnt = 0;
565 p->pack_local = local;
566 if ((path_len > 44) && !get_sha1_hex(path + path_len - 44, sha1))
567 hashcpy(p->sha1, sha1);
568 return p;
569 }
570
571 struct packed_git *parse_pack_index(unsigned char *sha1)
572 {
573 char *path = sha1_pack_index_name(sha1);
574 return parse_pack_index_file(sha1, path);
575 }
576
577 struct packed_git *parse_pack_index_file(const unsigned char *sha1, char *idx_path)
578 {
579 struct packed_git *p;
580 unsigned long idx_size;
581 void *idx_map;
582 char *path;
583
584 if (check_packed_git_idx(idx_path, &idx_size, &idx_map))
585 return NULL;
586
587 path = sha1_pack_name(sha1);
588
589 p = xmalloc(sizeof(*p) + strlen(path) + 2);
590 strcpy(p->pack_name, path);
591 p->index_size = idx_size;
592 p->pack_size = 0;
593 p->index_base = idx_map;
594 p->next = NULL;
595 p->pack_base = NULL;
596 p->pack_last_used = 0;
597 p->pack_use_cnt = 0;
598 hashcpy(p->sha1, sha1);
599 return p;
600 }
601
602 void install_packed_git(struct packed_git *pack)
603 {
604 pack->next = packed_git;
605 packed_git = pack;
606 }
607
608 static void prepare_packed_git_one(char *objdir, int local)
609 {
610 char path[PATH_MAX];
611 int len;
612 DIR *dir;
613 struct dirent *de;
614
615 sprintf(path, "%s/pack", objdir);
616 len = strlen(path);
617 dir = opendir(path);
618 if (!dir) {
619 if (errno != ENOENT)
620 error("unable to open object pack directory: %s: %s",
621 path, strerror(errno));
622 return;
623 }
624 path[len++] = '/';
625 while ((de = readdir(dir)) != NULL) {
626 int namelen = strlen(de->d_name);
627 struct packed_git *p;
628
629 if (!has_extension(de->d_name, ".idx"))
630 continue;
631
632 /* we have .idx. Is it a file we can map? */
633 strcpy(path + len, de->d_name);
634 for (p = packed_git; p; p = p->next) {
635 if (!memcmp(path, p->pack_name, len + namelen - 4))
636 break;
637 }
638 if (p)
639 continue;
640 p = add_packed_git(path, len + namelen, local);
641 if (!p)
642 continue;
643 p->next = packed_git;
644 packed_git = p;
645 }
646 closedir(dir);
647 }
648
649 static int prepare_packed_git_run_once = 0;
650 void prepare_packed_git(void)
651 {
652 struct alternate_object_database *alt;
653
654 if (prepare_packed_git_run_once)
655 return;
656 prepare_packed_git_one(get_object_directory(), 1);
657 prepare_alt_odb();
658 for (alt = alt_odb_list; alt; alt = alt->next) {
659 alt->name[-1] = 0;
660 prepare_packed_git_one(alt->base, 0);
661 alt->name[-1] = '/';
662 }
663 prepare_packed_git_run_once = 1;
664 }
665
666 void reprepare_packed_git(void)
667 {
668 prepare_packed_git_run_once = 0;
669 prepare_packed_git();
670 }
671
672 int check_sha1_signature(const unsigned char *sha1, void *map, unsigned long size, const char *type)
673 {
674 unsigned char real_sha1[20];
675 hash_sha1_file(map, size, type, real_sha1);
676 return hashcmp(sha1, real_sha1) ? -1 : 0;
677 }
678
679 void *map_sha1_file(const unsigned char *sha1, unsigned long *size)
680 {
681 struct stat st;
682 void *map;
683 int fd;
684 char *filename = find_sha1_file(sha1, &st);
685
686 if (!filename) {
687 return NULL;
688 }
689
690 fd = open(filename, O_RDONLY | sha1_file_open_flag);
691 if (fd < 0) {
692 /* See if it works without O_NOATIME */
693 switch (sha1_file_open_flag) {
694 default:
695 fd = open(filename, O_RDONLY);
696 if (fd >= 0)
697 break;
698 /* Fallthrough */
699 case 0:
700 return NULL;
701 }
702
703 /* If it failed once, it will probably fail again.
704 * Stop using O_NOATIME
705 */
706 sha1_file_open_flag = 0;
707 }
708 map = mmap(NULL, st.st_size, PROT_READ, MAP_PRIVATE, fd, 0);
709 close(fd);
710 if (map == MAP_FAILED)
711 return NULL;
712 *size = st.st_size;
713 return map;
714 }
715
716 int legacy_loose_object(unsigned char *map)
717 {
718 unsigned int word;
719
720 /*
721 * Is it a zlib-compressed buffer? If so, the first byte
722 * must be 0x78 (15-bit window size, deflated), and the
723 * first 16-bit word is evenly divisible by 31
724 */
725 word = (map[0] << 8) + map[1];
726 if (map[0] == 0x78 && !(word % 31))
727 return 1;
728 else
729 return 0;
730 }
731
732 unsigned long unpack_object_header_gently(const unsigned char *buf, unsigned long len, enum object_type *type, unsigned long *sizep)
733 {
734 unsigned shift;
735 unsigned char c;
736 unsigned long size;
737 unsigned long used = 0;
738
739 c = buf[used++];
740 *type = (c >> 4) & 7;
741 size = c & 15;
742 shift = 4;
743 while (c & 0x80) {
744 if (len <= used)
745 return 0;
746 if (sizeof(long) * 8 <= shift)
747 return 0;
748 c = buf[used++];
749 size += (c & 0x7f) << shift;
750 shift += 7;
751 }
752 *sizep = size;
753 return used;
754 }
755
756 static int unpack_sha1_header(z_stream *stream, unsigned char *map, unsigned long mapsize, void *buffer, unsigned long bufsiz)
757 {
758 unsigned long size, used;
759 static const char valid_loose_object_type[8] = {
760 0, /* OBJ_EXT */
761 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
762 0, /* "delta" and others are invalid in a loose object */
763 };
764 enum object_type type;
765
766 /* Get the data stream */
767 memset(stream, 0, sizeof(*stream));
768 stream->next_in = map;
769 stream->avail_in = mapsize;
770 stream->next_out = buffer;
771 stream->avail_out = bufsiz;
772
773 if (legacy_loose_object(map)) {
774 inflateInit(stream);
775 return inflate(stream, 0);
776 }
777
778 used = unpack_object_header_gently(map, mapsize, &type, &size);
779 if (!used || !valid_loose_object_type[type])
780 return -1;
781 map += used;
782 mapsize -= used;
783
784 /* Set up the stream for the rest.. */
785 stream->next_in = map;
786 stream->avail_in = mapsize;
787 inflateInit(stream);
788
789 /* And generate the fake traditional header */
790 stream->total_out = 1 + snprintf(buffer, bufsiz, "%s %lu",
791 type_names[type], size);
792 return 0;
793 }
794
795 static void *unpack_sha1_rest(z_stream *stream, void *buffer, unsigned long size)
796 {
797 int bytes = strlen(buffer) + 1;
798 unsigned char *buf = xmalloc(1+size);
799 unsigned long n;
800
801 n = stream->total_out - bytes;
802 if (n > size)
803 n = size;
804 memcpy(buf, (char *) buffer + bytes, n);
805 bytes = n;
806 if (bytes < size) {
807 stream->next_out = buf + bytes;
808 stream->avail_out = size - bytes;
809 while (inflate(stream, Z_FINISH) == Z_OK)
810 /* nothing */;
811 }
812 buf[size] = 0;
813 inflateEnd(stream);
814 return buf;
815 }
816
817 /*
818 * We used to just use "sscanf()", but that's actually way
819 * too permissive for what we want to check. So do an anal
820 * object header parse by hand.
821 */
822 static int parse_sha1_header(char *hdr, char *type, unsigned long *sizep)
823 {
824 int i;
825 unsigned long size;
826
827 /*
828 * The type can be at most ten bytes (including the
829 * terminating '\0' that we add), and is followed by
830 * a space.
831 */
832 i = 10;
833 for (;;) {
834 char c = *hdr++;
835 if (c == ' ')
836 break;
837 if (!--i)
838 return -1;
839 *type++ = c;
840 }
841 *type = 0;
842
843 /*
844 * The length must follow immediately, and be in canonical
845 * decimal format (ie "010" is not valid).
846 */
847 size = *hdr++ - '0';
848 if (size > 9)
849 return -1;
850 if (size) {
851 for (;;) {
852 unsigned long c = *hdr - '0';
853 if (c > 9)
854 break;
855 hdr++;
856 size = size * 10 + c;
857 }
858 }
859 *sizep = size;
860
861 /*
862 * The length must be followed by a zero byte
863 */
864 return *hdr ? -1 : 0;
865 }
866
867 void * unpack_sha1_file(void *map, unsigned long mapsize, char *type, unsigned long *size)
868 {
869 int ret;
870 z_stream stream;
871 char hdr[8192];
872
873 ret = unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr));
874 if (ret < Z_OK || parse_sha1_header(hdr, type, size) < 0)
875 return NULL;
876
877 return unpack_sha1_rest(&stream, hdr, *size);
878 }
879
880 static unsigned long get_delta_base(struct packed_git *p,
881 unsigned long offset,
882 enum object_type kind,
883 unsigned long delta_obj_offset,
884 unsigned long *base_obj_offset)
885 {
886 unsigned char *base_info = (unsigned char *) p->pack_base + offset;
887 unsigned long base_offset;
888
889 /* there must be at least 20 bytes left regardless of delta type */
890 if (p->pack_size <= offset + 20)
891 die("truncated pack file");
892
893 if (kind == OBJ_OFS_DELTA) {
894 unsigned used = 0;
895 unsigned char c = base_info[used++];
896 base_offset = c & 127;
897 while (c & 128) {
898 base_offset += 1;
899 if (!base_offset || base_offset & ~(~0UL >> 7))
900 die("offset value overflow for delta base object");
901 c = base_info[used++];
902 base_offset = (base_offset << 7) + (c & 127);
903 }
904 base_offset = delta_obj_offset - base_offset;
905 if (base_offset >= delta_obj_offset)
906 die("delta base offset out of bound");
907 offset += used;
908 } else if (kind == OBJ_REF_DELTA) {
909 /* The base entry _must_ be in the same pack */
910 base_offset = find_pack_entry_one(base_info, p);
911 if (!base_offset)
912 die("failed to find delta-pack base object %s",
913 sha1_to_hex(base_info));
914 offset += 20;
915 } else
916 die("I am totally screwed");
917 *base_obj_offset = base_offset;
918 return offset;
919 }
920
921 /* forward declaration for a mutually recursive function */
922 static int packed_object_info(struct packed_git *p, unsigned long offset,
923 char *type, unsigned long *sizep);
924
925 static int packed_delta_info(struct packed_git *p,
926 unsigned long offset,
927 enum object_type kind,
928 unsigned long obj_offset,
929 char *type,
930 unsigned long *sizep)
931 {
932 unsigned long base_offset;
933
934 offset = get_delta_base(p, offset, kind, obj_offset, &base_offset);
935
936 /* We choose to only get the type of the base object and
937 * ignore potentially corrupt pack file that expects the delta
938 * based on a base with a wrong size. This saves tons of
939 * inflate() calls.
940 */
941 if (packed_object_info(p, base_offset, type, NULL))
942 die("cannot get info for delta-pack base");
943
944 if (sizep) {
945 const unsigned char *data;
946 unsigned char delta_head[20];
947 unsigned long result_size;
948 z_stream stream;
949 int st;
950
951 memset(&stream, 0, sizeof(stream));
952
953 stream.next_in = (unsigned char *) p->pack_base + offset;
954 stream.avail_in = p->pack_size - offset;
955 stream.next_out = delta_head;
956 stream.avail_out = sizeof(delta_head);
957
958 inflateInit(&stream);
959 st = inflate(&stream, Z_FINISH);
960 inflateEnd(&stream);
961 if ((st != Z_STREAM_END) &&
962 stream.total_out != sizeof(delta_head))
963 die("delta data unpack-initial failed");
964
965 /* Examine the initial part of the delta to figure out
966 * the result size.
967 */
968 data = delta_head;
969
970 /* ignore base size */
971 get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
972
973 /* Read the result size */
974 result_size = get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
975 *sizep = result_size;
976 }
977 return 0;
978 }
979
980 static unsigned long unpack_object_header(struct packed_git *p, unsigned long offset,
981 enum object_type *type, unsigned long *sizep)
982 {
983 unsigned long used;
984
985 if (p->pack_size <= offset)
986 die("object offset outside of pack file");
987
988 used = unpack_object_header_gently((unsigned char *)p->pack_base +
989 offset,
990 p->pack_size - offset, type, sizep);
991 if (!used)
992 die("object offset outside of pack file");
993
994 return offset + used;
995 }
996
997 void packed_object_info_detail(struct packed_git *p,
998 unsigned long offset,
999 char *type,
1000 unsigned long *size,
1001 unsigned long *store_size,
1002 unsigned int *delta_chain_length,
1003 unsigned char *base_sha1)
1004 {
1005 unsigned long obj_offset, val;
1006 unsigned char *next_sha1;
1007 enum object_type kind;
1008
1009 *delta_chain_length = 0;
1010 obj_offset = offset;
1011 offset = unpack_object_header(p, offset, &kind, size);
1012
1013 for (;;) {
1014 switch (kind) {
1015 default:
1016 die("corrupted pack file %s containing object of kind %d",
1017 p->pack_name, kind);
1018 case OBJ_COMMIT:
1019 case OBJ_TREE:
1020 case OBJ_BLOB:
1021 case OBJ_TAG:
1022 strcpy(type, type_names[kind]);
1023 *store_size = 0; /* notyet */
1024 return;
1025 case OBJ_OFS_DELTA:
1026 get_delta_base(p, offset, kind, obj_offset, &offset);
1027 if (*delta_chain_length == 0) {
1028 /* TODO: find base_sha1 as pointed by offset */
1029 }
1030 break;
1031 case OBJ_REF_DELTA:
1032 if (p->pack_size <= offset + 20)
1033 die("pack file %s records an incomplete delta base",
1034 p->pack_name);
1035 next_sha1 = (unsigned char *) p->pack_base + offset;
1036 if (*delta_chain_length == 0)
1037 hashcpy(base_sha1, next_sha1);
1038 offset = find_pack_entry_one(next_sha1, p);
1039 break;
1040 }
1041 obj_offset = offset;
1042 offset = unpack_object_header(p, offset, &kind, &val);
1043 (*delta_chain_length)++;
1044 }
1045 }
1046
1047 static int packed_object_info(struct packed_git *p, unsigned long offset,
1048 char *type, unsigned long *sizep)
1049 {
1050 unsigned long size, obj_offset = offset;
1051 enum object_type kind;
1052
1053 offset = unpack_object_header(p, offset, &kind, &size);
1054
1055 switch (kind) {
1056 case OBJ_OFS_DELTA:
1057 case OBJ_REF_DELTA:
1058 return packed_delta_info(p, offset, kind, obj_offset, type, sizep);
1059 case OBJ_COMMIT:
1060 case OBJ_TREE:
1061 case OBJ_BLOB:
1062 case OBJ_TAG:
1063 strcpy(type, type_names[kind]);
1064 break;
1065 default:
1066 die("corrupted pack file %s containing object of kind %d",
1067 p->pack_name, kind);
1068 }
1069 if (sizep)
1070 *sizep = size;
1071 return 0;
1072 }
1073
1074 static void *unpack_compressed_entry(struct packed_git *p,
1075 unsigned long offset,
1076 unsigned long size)
1077 {
1078 int st;
1079 z_stream stream;
1080 unsigned char *buffer;
1081
1082 buffer = xmalloc(size + 1);
1083 buffer[size] = 0;
1084 memset(&stream, 0, sizeof(stream));
1085 stream.next_in = (unsigned char*)p->pack_base + offset;
1086 stream.avail_in = p->pack_size - offset;
1087 stream.next_out = buffer;
1088 stream.avail_out = size;
1089
1090 inflateInit(&stream);
1091 st = inflate(&stream, Z_FINISH);
1092 inflateEnd(&stream);
1093 if ((st != Z_STREAM_END) || stream.total_out != size) {
1094 free(buffer);
1095 return NULL;
1096 }
1097
1098 return buffer;
1099 }
1100
1101 static void *unpack_delta_entry(struct packed_git *p,
1102 unsigned long offset,
1103 unsigned long delta_size,
1104 enum object_type kind,
1105 unsigned long obj_offset,
1106 char *type,
1107 unsigned long *sizep)
1108 {
1109 void *delta_data, *result, *base;
1110 unsigned long result_size, base_size, base_offset;
1111
1112 offset = get_delta_base(p, offset, kind, obj_offset, &base_offset);
1113 base = unpack_entry_gently(p, base_offset, type, &base_size);
1114 if (!base)
1115 die("failed to read delta base object at %lu from %s",
1116 base_offset, p->pack_name);
1117
1118 delta_data = unpack_compressed_entry(p, offset, delta_size);
1119 result = patch_delta(base, base_size,
1120 delta_data, delta_size,
1121 &result_size);
1122 if (!result)
1123 die("failed to apply delta");
1124 free(delta_data);
1125 free(base);
1126 *sizep = result_size;
1127 return result;
1128 }
1129
1130 static void *unpack_entry(struct pack_entry *entry,
1131 char *type, unsigned long *sizep)
1132 {
1133 struct packed_git *p = entry->p;
1134 void *retval;
1135
1136 if (use_packed_git(p))
1137 die("cannot map packed file");
1138 retval = unpack_entry_gently(p, entry->offset, type, sizep);
1139 unuse_packed_git(p);
1140 if (!retval)
1141 die("corrupted pack file %s", p->pack_name);
1142 return retval;
1143 }
1144
1145 /* The caller is responsible for use_packed_git()/unuse_packed_git() pair */
1146 void *unpack_entry_gently(struct packed_git *p, unsigned long offset,
1147 char *type, unsigned long *sizep)
1148 {
1149 unsigned long size, obj_offset = offset;
1150 enum object_type kind;
1151
1152 offset = unpack_object_header(p, offset, &kind, &size);
1153 switch (kind) {
1154 case OBJ_OFS_DELTA:
1155 case OBJ_REF_DELTA:
1156 return unpack_delta_entry(p, offset, size, kind, obj_offset, type, sizep);
1157 case OBJ_COMMIT:
1158 case OBJ_TREE:
1159 case OBJ_BLOB:
1160 case OBJ_TAG:
1161 strcpy(type, type_names[kind]);
1162 *sizep = size;
1163 return unpack_compressed_entry(p, offset, size);
1164 default:
1165 return NULL;
1166 }
1167 }
1168
1169 int num_packed_objects(const struct packed_git *p)
1170 {
1171 /* See check_packed_git_idx() */
1172 return (p->index_size - 20 - 20 - 4*256) / 24;
1173 }
1174
1175 int nth_packed_object_sha1(const struct packed_git *p, int n,
1176 unsigned char* sha1)
1177 {
1178 void *index = p->index_base + 256;
1179 if (n < 0 || num_packed_objects(p) <= n)
1180 return -1;
1181 hashcpy(sha1, (unsigned char *) index + (24 * n) + 4);
1182 return 0;
1183 }
1184
1185 unsigned long find_pack_entry_one(const unsigned char *sha1,
1186 struct packed_git *p)
1187 {
1188 unsigned int *level1_ofs = p->index_base;
1189 int hi = ntohl(level1_ofs[*sha1]);
1190 int lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
1191 void *index = p->index_base + 256;
1192
1193 do {
1194 int mi = (lo + hi) / 2;
1195 int cmp = hashcmp((unsigned char *)index + (24 * mi) + 4, sha1);
1196 if (!cmp)
1197 return ntohl(*((unsigned int *) ((char *) index + (24 * mi))));
1198 if (cmp > 0)
1199 hi = mi;
1200 else
1201 lo = mi+1;
1202 } while (lo < hi);
1203 return 0;
1204 }
1205
1206 static int matches_pack_name(struct packed_git *p, const char *ig)
1207 {
1208 const char *last_c, *c;
1209
1210 if (!strcmp(p->pack_name, ig))
1211 return 0;
1212
1213 for (c = p->pack_name, last_c = c; *c;)
1214 if (*c == '/')
1215 last_c = ++c;
1216 else
1217 ++c;
1218 if (!strcmp(last_c, ig))
1219 return 0;
1220
1221 return 1;
1222 }
1223
1224 static int find_pack_entry(const unsigned char *sha1, struct pack_entry *e, const char **ignore_packed)
1225 {
1226 struct packed_git *p;
1227 unsigned long offset;
1228
1229 prepare_packed_git();
1230
1231 for (p = packed_git; p; p = p->next) {
1232 if (ignore_packed) {
1233 const char **ig;
1234 for (ig = ignore_packed; *ig; ig++)
1235 if (!matches_pack_name(p, *ig))
1236 break;
1237 if (*ig)
1238 continue;
1239 }
1240 offset = find_pack_entry_one(sha1, p);
1241 if (offset) {
1242 e->offset = offset;
1243 e->p = p;
1244 hashcpy(e->sha1, sha1);
1245 return 1;
1246 }
1247 }
1248 return 0;
1249 }
1250
1251 struct packed_git *find_sha1_pack(const unsigned char *sha1,
1252 struct packed_git *packs)
1253 {
1254 struct packed_git *p;
1255
1256 for (p = packs; p; p = p->next) {
1257 if (find_pack_entry_one(sha1, p))
1258 return p;
1259 }
1260 return NULL;
1261
1262 }
1263
1264 int sha1_object_info(const unsigned char *sha1, char *type, unsigned long *sizep)
1265 {
1266 int status;
1267 unsigned long mapsize, size;
1268 void *map;
1269 z_stream stream;
1270 char hdr[128];
1271
1272 map = map_sha1_file(sha1, &mapsize);
1273 if (!map) {
1274 struct pack_entry e;
1275
1276 if (!find_pack_entry(sha1, &e, NULL)) {
1277 reprepare_packed_git();
1278 if (!find_pack_entry(sha1, &e, NULL))
1279 return error("unable to find %s", sha1_to_hex(sha1));
1280 }
1281 if (use_packed_git(e.p))
1282 die("cannot map packed file");
1283 status = packed_object_info(e.p, e.offset, type, sizep);
1284 unuse_packed_git(e.p);
1285 return status;
1286 }
1287 if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
1288 status = error("unable to unpack %s header",
1289 sha1_to_hex(sha1));
1290 if (parse_sha1_header(hdr, type, &size) < 0)
1291 status = error("unable to parse %s header", sha1_to_hex(sha1));
1292 else {
1293 status = 0;
1294 if (sizep)
1295 *sizep = size;
1296 }
1297 inflateEnd(&stream);
1298 munmap(map, mapsize);
1299 return status;
1300 }
1301
1302 static void *read_packed_sha1(const unsigned char *sha1, char *type, unsigned long *size)
1303 {
1304 struct pack_entry e;
1305
1306 if (!find_pack_entry(sha1, &e, NULL)) {
1307 error("cannot read sha1_file for %s", sha1_to_hex(sha1));
1308 return NULL;
1309 }
1310 return unpack_entry(&e, type, size);
1311 }
1312
1313 void * read_sha1_file(const unsigned char *sha1, char *type, unsigned long *size)
1314 {
1315 unsigned long mapsize;
1316 void *map, *buf;
1317 struct pack_entry e;
1318
1319 if (find_pack_entry(sha1, &e, NULL))
1320 return read_packed_sha1(sha1, type, size);
1321 map = map_sha1_file(sha1, &mapsize);
1322 if (map) {
1323 buf = unpack_sha1_file(map, mapsize, type, size);
1324 munmap(map, mapsize);
1325 return buf;
1326 }
1327 reprepare_packed_git();
1328 if (find_pack_entry(sha1, &e, NULL))
1329 return read_packed_sha1(sha1, type, size);
1330 return NULL;
1331 }
1332
1333 void *read_object_with_reference(const unsigned char *sha1,
1334 const char *required_type,
1335 unsigned long *size,
1336 unsigned char *actual_sha1_return)
1337 {
1338 char type[20];
1339 void *buffer;
1340 unsigned long isize;
1341 unsigned char actual_sha1[20];
1342
1343 hashcpy(actual_sha1, sha1);
1344 while (1) {
1345 int ref_length = -1;
1346 const char *ref_type = NULL;
1347
1348 buffer = read_sha1_file(actual_sha1, type, &isize);
1349 if (!buffer)
1350 return NULL;
1351 if (!strcmp(type, required_type)) {
1352 *size = isize;
1353 if (actual_sha1_return)
1354 hashcpy(actual_sha1_return, actual_sha1);
1355 return buffer;
1356 }
1357 /* Handle references */
1358 else if (!strcmp(type, commit_type))
1359 ref_type = "tree ";
1360 else if (!strcmp(type, tag_type))
1361 ref_type = "object ";
1362 else {
1363 free(buffer);
1364 return NULL;
1365 }
1366 ref_length = strlen(ref_type);
1367
1368 if (memcmp(buffer, ref_type, ref_length) ||
1369 get_sha1_hex((char *) buffer + ref_length, actual_sha1)) {
1370 free(buffer);
1371 return NULL;
1372 }
1373 free(buffer);
1374 /* Now we have the ID of the referred-to object in
1375 * actual_sha1. Check again. */
1376 }
1377 }
1378
1379 static void write_sha1_file_prepare(void *buf, unsigned long len,
1380 const char *type, unsigned char *sha1,
1381 unsigned char *hdr, int *hdrlen)
1382 {
1383 SHA_CTX c;
1384
1385 /* Generate the header */
1386 *hdrlen = sprintf((char *)hdr, "%s %lu", type, len)+1;
1387
1388 /* Sha1.. */
1389 SHA1_Init(&c);
1390 SHA1_Update(&c, hdr, *hdrlen);
1391 SHA1_Update(&c, buf, len);
1392 SHA1_Final(sha1, &c);
1393 }
1394
1395 /*
1396 * Link the tempfile to the final place, possibly creating the
1397 * last directory level as you do so.
1398 *
1399 * Returns the errno on failure, 0 on success.
1400 */
1401 static int link_temp_to_file(const char *tmpfile, const char *filename)
1402 {
1403 int ret;
1404 char *dir;
1405
1406 if (!link(tmpfile, filename))
1407 return 0;
1408
1409 /*
1410 * Try to mkdir the last path component if that failed.
1411 *
1412 * Re-try the "link()" regardless of whether the mkdir
1413 * succeeds, since a race might mean that somebody
1414 * else succeeded.
1415 */
1416 ret = errno;
1417 dir = strrchr(filename, '/');
1418 if (dir) {
1419 *dir = 0;
1420 if (!mkdir(filename, 0777) && adjust_shared_perm(filename)) {
1421 *dir = '/';
1422 return -2;
1423 }
1424 *dir = '/';
1425 if (!link(tmpfile, filename))
1426 return 0;
1427 ret = errno;
1428 }
1429 return ret;
1430 }
1431
1432 /*
1433 * Move the just written object into its final resting place
1434 */
1435 int move_temp_to_file(const char *tmpfile, const char *filename)
1436 {
1437 int ret = link_temp_to_file(tmpfile, filename);
1438
1439 /*
1440 * Coda hack - coda doesn't like cross-directory links,
1441 * so we fall back to a rename, which will mean that it
1442 * won't be able to check collisions, but that's not a
1443 * big deal.
1444 *
1445 * The same holds for FAT formatted media.
1446 *
1447 * When this succeeds, we just return 0. We have nothing
1448 * left to unlink.
1449 */
1450 if (ret && ret != EEXIST) {
1451 if (!rename(tmpfile, filename))
1452 return 0;
1453 ret = errno;
1454 }
1455 unlink(tmpfile);
1456 if (ret) {
1457 if (ret != EEXIST) {
1458 return error("unable to write sha1 filename %s: %s\n", filename, strerror(ret));
1459 }
1460 /* FIXME!!! Collision check here ? */
1461 }
1462
1463 return 0;
1464 }
1465
1466 static int write_buffer(int fd, const void *buf, size_t len)
1467 {
1468 while (len) {
1469 ssize_t size;
1470
1471 size = write(fd, buf, len);
1472 if (!size)
1473 return error("file write: disk full");
1474 if (size < 0) {
1475 if (errno == EINTR || errno == EAGAIN)
1476 continue;
1477 return error("file write error (%s)", strerror(errno));
1478 }
1479 len -= size;
1480 buf = (char *) buf + size;
1481 }
1482 return 0;
1483 }
1484
1485 static int write_binary_header(unsigned char *hdr, enum object_type type, unsigned long len)
1486 {
1487 int hdr_len;
1488 unsigned char c;
1489
1490 c = (type << 4) | (len & 15);
1491 len >>= 4;
1492 hdr_len = 1;
1493 while (len) {
1494 *hdr++ = c | 0x80;
1495 hdr_len++;
1496 c = (len & 0x7f);
1497 len >>= 7;
1498 }
1499 *hdr = c;
1500 return hdr_len;
1501 }
1502
1503 static void setup_object_header(z_stream *stream, const char *type, unsigned long len)
1504 {
1505 int obj_type, hdr;
1506
1507 if (use_legacy_headers) {
1508 while (deflate(stream, 0) == Z_OK)
1509 /* nothing */;
1510 return;
1511 }
1512 if (!strcmp(type, blob_type))
1513 obj_type = OBJ_BLOB;
1514 else if (!strcmp(type, tree_type))
1515 obj_type = OBJ_TREE;
1516 else if (!strcmp(type, commit_type))
1517 obj_type = OBJ_COMMIT;
1518 else if (!strcmp(type, tag_type))
1519 obj_type = OBJ_TAG;
1520 else
1521 die("trying to generate bogus object of type '%s'", type);
1522 hdr = write_binary_header(stream->next_out, obj_type, len);
1523 stream->total_out = hdr;
1524 stream->next_out += hdr;
1525 stream->avail_out -= hdr;
1526 }
1527
1528 int hash_sha1_file(void *buf, unsigned long len, const char *type,
1529 unsigned char *sha1)
1530 {
1531 unsigned char hdr[50];
1532 int hdrlen;
1533 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
1534 return 0;
1535 }
1536
1537 int write_sha1_file(void *buf, unsigned long len, const char *type, unsigned char *returnsha1)
1538 {
1539 int size;
1540 unsigned char *compressed;
1541 z_stream stream;
1542 unsigned char sha1[20];
1543 char *filename;
1544 static char tmpfile[PATH_MAX];
1545 unsigned char hdr[50];
1546 int fd, hdrlen;
1547
1548 /* Normally if we have it in the pack then we do not bother writing
1549 * it out into .git/objects/??/?{38} file.
1550 */
1551 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
1552 filename = sha1_file_name(sha1);
1553 if (returnsha1)
1554 hashcpy(returnsha1, sha1);
1555 if (has_sha1_file(sha1))
1556 return 0;
1557 fd = open(filename, O_RDONLY);
1558 if (fd >= 0) {
1559 /*
1560 * FIXME!!! We might do collision checking here, but we'd
1561 * need to uncompress the old file and check it. Later.
1562 */
1563 close(fd);
1564 return 0;
1565 }
1566
1567 if (errno != ENOENT) {
1568 return error("sha1 file %s: %s\n", filename, strerror(errno));
1569 }
1570
1571 snprintf(tmpfile, sizeof(tmpfile), "%s/obj_XXXXXX", get_object_directory());
1572
1573 fd = mkstemp(tmpfile);
1574 if (fd < 0) {
1575 if (errno == EPERM)
1576 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
1577 else
1578 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
1579 }
1580
1581 /* Set it up */
1582 memset(&stream, 0, sizeof(stream));
1583 deflateInit(&stream, zlib_compression_level);
1584 size = 8 + deflateBound(&stream, len+hdrlen);
1585 compressed = xmalloc(size);
1586
1587 /* Compress it */
1588 stream.next_out = compressed;
1589 stream.avail_out = size;
1590
1591 /* First header.. */
1592 stream.next_in = hdr;
1593 stream.avail_in = hdrlen;
1594 setup_object_header(&stream, type, len);
1595
1596 /* Then the data itself.. */
1597 stream.next_in = buf;
1598 stream.avail_in = len;
1599 while (deflate(&stream, Z_FINISH) == Z_OK)
1600 /* nothing */;
1601 deflateEnd(&stream);
1602 size = stream.total_out;
1603
1604 if (write_buffer(fd, compressed, size) < 0)
1605 die("unable to write sha1 file");
1606 fchmod(fd, 0444);
1607 close(fd);
1608 free(compressed);
1609
1610 return move_temp_to_file(tmpfile, filename);
1611 }
1612
1613 /*
1614 * We need to unpack and recompress the object for writing
1615 * it out to a different file.
1616 */
1617 static void *repack_object(const unsigned char *sha1, unsigned long *objsize)
1618 {
1619 size_t size;
1620 z_stream stream;
1621 unsigned char *unpacked;
1622 unsigned long len;
1623 char type[20];
1624 char hdr[50];
1625 int hdrlen;
1626 void *buf;
1627
1628 /* need to unpack and recompress it by itself */
1629 unpacked = read_packed_sha1(sha1, type, &len);
1630
1631 hdrlen = sprintf(hdr, "%s %lu", type, len) + 1;
1632
1633 /* Set it up */
1634 memset(&stream, 0, sizeof(stream));
1635 deflateInit(&stream, zlib_compression_level);
1636 size = deflateBound(&stream, len + hdrlen);
1637 buf = xmalloc(size);
1638
1639 /* Compress it */
1640 stream.next_out = buf;
1641 stream.avail_out = size;
1642
1643 /* First header.. */
1644 stream.next_in = (void *)hdr;
1645 stream.avail_in = hdrlen;
1646 while (deflate(&stream, 0) == Z_OK)
1647 /* nothing */;
1648
1649 /* Then the data itself.. */
1650 stream.next_in = unpacked;
1651 stream.avail_in = len;
1652 while (deflate(&stream, Z_FINISH) == Z_OK)
1653 /* nothing */;
1654 deflateEnd(&stream);
1655 free(unpacked);
1656
1657 *objsize = stream.total_out;
1658 return buf;
1659 }
1660
1661 int write_sha1_to_fd(int fd, const unsigned char *sha1)
1662 {
1663 int retval;
1664 unsigned long objsize;
1665 void *buf = map_sha1_file(sha1, &objsize);
1666
1667 if (buf) {
1668 retval = write_buffer(fd, buf, objsize);
1669 munmap(buf, objsize);
1670 return retval;
1671 }
1672
1673 buf = repack_object(sha1, &objsize);
1674 retval = write_buffer(fd, buf, objsize);
1675 free(buf);
1676 return retval;
1677 }
1678
1679 int write_sha1_from_fd(const unsigned char *sha1, int fd, char *buffer,
1680 size_t bufsize, size_t *bufposn)
1681 {
1682 char tmpfile[PATH_MAX];
1683 int local;
1684 z_stream stream;
1685 unsigned char real_sha1[20];
1686 unsigned char discard[4096];
1687 int ret;
1688 SHA_CTX c;
1689
1690 snprintf(tmpfile, sizeof(tmpfile), "%s/obj_XXXXXX", get_object_directory());
1691
1692 local = mkstemp(tmpfile);
1693 if (local < 0) {
1694 if (errno == EPERM)
1695 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
1696 else
1697 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
1698 }
1699
1700 memset(&stream, 0, sizeof(stream));
1701
1702 inflateInit(&stream);
1703
1704 SHA1_Init(&c);
1705
1706 do {
1707 ssize_t size;
1708 if (*bufposn) {
1709 stream.avail_in = *bufposn;
1710 stream.next_in = (unsigned char *) buffer;
1711 do {
1712 stream.next_out = discard;
1713 stream.avail_out = sizeof(discard);
1714 ret = inflate(&stream, Z_SYNC_FLUSH);
1715 SHA1_Update(&c, discard, sizeof(discard) -
1716 stream.avail_out);
1717 } while (stream.avail_in && ret == Z_OK);
1718 if (write_buffer(local, buffer, *bufposn - stream.avail_in) < 0)
1719 die("unable to write sha1 file");
1720 memmove(buffer, buffer + *bufposn - stream.avail_in,
1721 stream.avail_in);
1722 *bufposn = stream.avail_in;
1723 if (ret != Z_OK)
1724 break;
1725 }
1726 size = read(fd, buffer + *bufposn, bufsize - *bufposn);
1727 if (size <= 0) {
1728 close(local);
1729 unlink(tmpfile);
1730 if (!size)
1731 return error("Connection closed?");
1732 perror("Reading from connection");
1733 return -1;
1734 }
1735 *bufposn += size;
1736 } while (1);
1737 inflateEnd(&stream);
1738
1739 close(local);
1740 SHA1_Final(real_sha1, &c);
1741 if (ret != Z_STREAM_END) {
1742 unlink(tmpfile);
1743 return error("File %s corrupted", sha1_to_hex(sha1));
1744 }
1745 if (hashcmp(sha1, real_sha1)) {
1746 unlink(tmpfile);
1747 return error("File %s has bad hash", sha1_to_hex(sha1));
1748 }
1749
1750 return move_temp_to_file(tmpfile, sha1_file_name(sha1));
1751 }
1752
1753 int has_pack_index(const unsigned char *sha1)
1754 {
1755 struct stat st;
1756 if (stat(sha1_pack_index_name(sha1), &st))
1757 return 0;
1758 return 1;
1759 }
1760
1761 int has_pack_file(const unsigned char *sha1)
1762 {
1763 struct stat st;
1764 if (stat(sha1_pack_name(sha1), &st))
1765 return 0;
1766 return 1;
1767 }
1768
1769 int has_sha1_pack(const unsigned char *sha1, const char **ignore_packed)
1770 {
1771 struct pack_entry e;
1772 return find_pack_entry(sha1, &e, ignore_packed);
1773 }
1774
1775 int has_sha1_file(const unsigned char *sha1)
1776 {
1777 struct stat st;
1778 struct pack_entry e;
1779
1780 if (find_pack_entry(sha1, &e, NULL))
1781 return 1;
1782 return find_sha1_file(sha1, &st) ? 1 : 0;
1783 }
1784
1785 /*
1786 * reads from fd as long as possible into a supplied buffer of size bytes.
1787 * If necessary the buffer's size is increased using realloc()
1788 *
1789 * returns 0 if anything went fine and -1 otherwise
1790 *
1791 * NOTE: both buf and size may change, but even when -1 is returned
1792 * you still have to free() it yourself.
1793 */
1794 int read_pipe(int fd, char** return_buf, unsigned long* return_size)
1795 {
1796 char* buf = *return_buf;
1797 unsigned long size = *return_size;
1798 int iret;
1799 unsigned long off = 0;
1800
1801 do {
1802 iret = xread(fd, buf + off, size - off);
1803 if (iret > 0) {
1804 off += iret;
1805 if (off == size) {
1806 size *= 2;
1807 buf = xrealloc(buf, size);
1808 }
1809 }
1810 } while (iret > 0);
1811
1812 *return_buf = buf;
1813 *return_size = off;
1814
1815 if (iret < 0)
1816 return -1;
1817 return 0;
1818 }
1819
1820 int index_pipe(unsigned char *sha1, int fd, const char *type, int write_object)
1821 {
1822 unsigned long size = 4096;
1823 char *buf = xmalloc(size);
1824 int ret;
1825
1826 if (read_pipe(fd, &buf, &size)) {
1827 free(buf);
1828 return -1;
1829 }
1830
1831 if (!type)
1832 type = blob_type;
1833 if (write_object)
1834 ret = write_sha1_file(buf, size, type, sha1);
1835 else
1836 ret = hash_sha1_file(buf, size, type, sha1);
1837 free(buf);
1838 return ret;
1839 }
1840
1841 int index_fd(unsigned char *sha1, int fd, struct stat *st, int write_object, const char *type)
1842 {
1843 unsigned long size = st->st_size;
1844 void *buf;
1845 int ret;
1846
1847 buf = "";
1848 if (size)
1849 buf = mmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
1850 close(fd);
1851 if (buf == MAP_FAILED)
1852 return -1;
1853
1854 if (!type)
1855 type = blob_type;
1856 if (write_object)
1857 ret = write_sha1_file(buf, size, type, sha1);
1858 else
1859 ret = hash_sha1_file(buf, size, type, sha1);
1860 if (size)
1861 munmap(buf, size);
1862 return ret;
1863 }
1864
1865 int index_path(unsigned char *sha1, const char *path, struct stat *st, int write_object)
1866 {
1867 int fd;
1868 char *target;
1869
1870 switch (st->st_mode & S_IFMT) {
1871 case S_IFREG:
1872 fd = open(path, O_RDONLY);
1873 if (fd < 0)
1874 return error("open(\"%s\"): %s", path,
1875 strerror(errno));
1876 if (index_fd(sha1, fd, st, write_object, NULL) < 0)
1877 return error("%s: failed to insert into database",
1878 path);
1879 break;
1880 case S_IFLNK:
1881 target = xmalloc(st->st_size+1);
1882 if (readlink(path, target, st->st_size+1) != st->st_size) {
1883 char *errstr = strerror(errno);
1884 free(target);
1885 return error("readlink(\"%s\"): %s", path,
1886 errstr);
1887 }
1888 if (!write_object)
1889 hash_sha1_file(target, st->st_size, blob_type, sha1);
1890 else if (write_sha1_file(target, st->st_size, blob_type, sha1))
1891 return error("%s: failed to insert into database",
1892 path);
1893 free(target);
1894 break;
1895 default:
1896 return error("%s: unsupported file type", path);
1897 }
1898 return 0;
1899 }