2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
6 * This handles basic git sha1 object files - packing, unpacking,
18 #if defined(__linux__) && (defined(__i386__) || defined(__PPC__))
19 #define O_NOATIME 01000000
31 const unsigned char null_sha1[20];
33 static unsigned int sha1_file_open_flag = O_NOATIME;
35 signed char hexval_table[256] = {
36 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
37 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
38 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
39 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
40 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
41 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
42 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
43 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
44 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
45 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
46 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
47 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
48 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
49 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
50 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
51 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
52 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
53 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
54 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
55 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
56 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
57 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
58 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
59 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
60 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
61 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
62 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
63 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
64 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
65 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
66 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
67 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
70 int get_sha1_hex(const char *hex, unsigned char *sha1)
73 for (i = 0; i < 20; i++) {
74 unsigned int val = (hexval(hex[0]) << 4) | hexval(hex[1]);
83 int safe_create_leading_directories(char *path)
92 pos = strchr(pos, '/');
96 if (!stat(path, &st)) {
98 if (!S_ISDIR(st.st_mode)) {
103 else if (mkdir(path, 0777)) {
107 else if (adjust_shared_perm(path)) {
116 char * sha1_to_hex(const unsigned char *sha1)
119 static char hexbuffer[4][50];
120 static const char hex[] = "0123456789abcdef";
121 char *buffer = hexbuffer[3 & ++bufno], *buf = buffer;
124 for (i = 0; i < 20; i++) {
125 unsigned int val = *sha1++;
126 *buf++ = hex[val >> 4];
127 *buf++ = hex[val & 0xf];
134 static void fill_sha1_path(char *pathbuf, const unsigned char *sha1)
137 for (i = 0; i < 20; i++) {
138 static char hex[] = "0123456789abcdef";
139 unsigned int val = sha1[i];
140 char *pos = pathbuf + i*2 + (i > 0);
141 *pos++ = hex[val >> 4];
142 *pos = hex[val & 0xf];
147 * NOTE! This returns a statically allocated buffer, so you have to be
148 * careful about using it. Do a "xstrdup()" if you need to save the
151 * Also note that this returns the location for creating. Reading
152 * SHA1 file can happen from any alternate directory listed in the
153 * DB_ENVIRONMENT environment variable if it is not found in
154 * the primary object database.
156 char *sha1_file_name(const unsigned char *sha1)
158 static char *name, *base;
161 const char *sha1_file_directory = get_object_directory();
162 int len = strlen(sha1_file_directory);
163 base = xmalloc(len + 60);
164 memcpy(base, sha1_file_directory, len);
165 memset(base+len, 0, 60);
168 name = base + len + 1;
170 fill_sha1_path(name, sha1);
174 char *sha1_pack_name(const unsigned char *sha1)
176 static const char hex[] = "0123456789abcdef";
177 static char *name, *base, *buf;
181 const char *sha1_file_directory = get_object_directory();
182 int len = strlen(sha1_file_directory);
183 base = xmalloc(len + 60);
184 sprintf(base, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory);
185 name = base + len + 11;
190 for (i = 0; i < 20; i++) {
191 unsigned int val = *sha1++;
192 *buf++ = hex[val >> 4];
193 *buf++ = hex[val & 0xf];
199 char *sha1_pack_index_name(const unsigned char *sha1)
201 static const char hex[] = "0123456789abcdef";
202 static char *name, *base, *buf;
206 const char *sha1_file_directory = get_object_directory();
207 int len = strlen(sha1_file_directory);
208 base = xmalloc(len + 60);
209 sprintf(base, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory);
210 name = base + len + 11;
215 for (i = 0; i < 20; i++) {
216 unsigned int val = *sha1++;
217 *buf++ = hex[val >> 4];
218 *buf++ = hex[val & 0xf];
224 struct alternate_object_database *alt_odb_list;
225 static struct alternate_object_database **alt_odb_tail;
227 static void read_info_alternates(const char * alternates, int depth);
230 * Prepare alternate object database registry.
232 * The variable alt_odb_list points at the list of struct
233 * alternate_object_database. The elements on this list come from
234 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
235 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
236 * whose contents is similar to that environment variable but can be
237 * LF separated. Its base points at a statically allocated buffer that
238 * contains "/the/directory/corresponding/to/.git/objects/...", while
239 * its name points just after the slash at the end of ".git/objects/"
240 * in the example above, and has enough space to hold 40-byte hex
241 * SHA1, an extra slash for the first level indirection, and the
244 static int link_alt_odb_entry(const char * entry, int len, const char * relative_base, int depth)
247 const char *objdir = get_object_directory();
248 struct alternate_object_database *ent;
249 struct alternate_object_database *alt;
250 /* 43 = 40-byte + 2 '/' + terminating NUL */
252 int entlen = pfxlen + 43;
255 if (*entry != '/' && relative_base) {
256 /* Relative alt-odb */
258 base_len = strlen(relative_base) + 1;
262 ent = xmalloc(sizeof(*ent) + entlen);
264 if (*entry != '/' && relative_base) {
265 memcpy(ent->base, relative_base, base_len - 1);
266 ent->base[base_len - 1] = '/';
267 memcpy(ent->base + base_len, entry, len);
270 memcpy(ent->base, entry, pfxlen);
272 ent->name = ent->base + pfxlen + 1;
273 ent->base[pfxlen + 3] = '/';
274 ent->base[pfxlen] = ent->base[entlen-1] = 0;
276 /* Detect cases where alternate disappeared */
277 if (stat(ent->base, &st) || !S_ISDIR(st.st_mode)) {
278 error("object directory %s does not exist; "
279 "check .git/objects/info/alternates.",
285 /* Prevent the common mistake of listing the same
286 * thing twice, or object directory itself.
288 for (alt = alt_odb_list; alt; alt = alt->next) {
289 if (!memcmp(ent->base, alt->base, pfxlen)) {
294 if (!memcmp(ent->base, objdir, pfxlen)) {
299 /* add the alternate entry */
301 alt_odb_tail = &(ent->next);
304 /* recursively add alternates */
305 read_info_alternates(ent->base, depth + 1);
307 ent->base[pfxlen] = '/';
312 static void link_alt_odb_entries(const char *alt, const char *ep, int sep,
313 const char *relative_base, int depth)
315 const char *cp, *last;
318 error("%s: ignoring alternate object stores, nesting too deep.",
326 if (cp < ep && *cp == '#') {
327 while (cp < ep && *cp != sep)
332 while (cp < ep && *cp != sep)
335 if ((*last != '/') && depth) {
336 error("%s: ignoring relative alternate object store %s",
337 relative_base, last);
339 link_alt_odb_entry(last, cp - last,
340 relative_base, depth);
343 while (cp < ep && *cp == sep)
349 static void read_info_alternates(const char * relative_base, int depth)
357 sprintf(path, "%s/info/alternates", relative_base);
358 fd = open(path, O_RDONLY);
361 if (fstat(fd, &st) || (st.st_size == 0)) {
365 mapsz = xsize_t(st.st_size);
366 map = xmmap(NULL, mapsz, PROT_READ, MAP_PRIVATE, fd, 0);
369 link_alt_odb_entries(map, map + mapsz, '\n', relative_base, depth);
374 void prepare_alt_odb(void)
378 alt = getenv(ALTERNATE_DB_ENVIRONMENT);
383 alt_odb_tail = &alt_odb_list;
384 link_alt_odb_entries(alt, alt + strlen(alt), ':', NULL, 0);
386 read_info_alternates(get_object_directory(), 0);
389 static char *find_sha1_file(const unsigned char *sha1, struct stat *st)
391 char *name = sha1_file_name(sha1);
392 struct alternate_object_database *alt;
397 for (alt = alt_odb_list; alt; alt = alt->next) {
399 fill_sha1_path(name, sha1);
400 if (!stat(alt->base, st))
406 static unsigned int pack_used_ctr;
407 static unsigned int pack_mmap_calls;
408 static unsigned int peak_pack_open_windows;
409 static unsigned int pack_open_windows;
410 static size_t peak_pack_mapped;
411 static size_t pack_mapped;
412 struct packed_git *packed_git;
417 "pack_report: getpagesize() = %10" SZ_FMT "\n"
418 "pack_report: core.packedGitWindowSize = %10" SZ_FMT "\n"
419 "pack_report: core.packedGitLimit = %10" SZ_FMT "\n",
420 (size_t) getpagesize(),
421 packed_git_window_size,
424 "pack_report: pack_used_ctr = %10u\n"
425 "pack_report: pack_mmap_calls = %10u\n"
426 "pack_report: pack_open_windows = %10u / %10u\n"
427 "pack_report: pack_mapped = "
428 "%10" SZ_FMT " / %10" SZ_FMT "\n",
431 pack_open_windows, peak_pack_open_windows,
432 pack_mapped, peak_pack_mapped);
435 static int check_packed_git_idx(const char *path, struct packed_git *p)
438 struct pack_idx_header *hdr;
440 uint32_t nr, i, *index;
441 int fd = open(path, O_RDONLY);
446 if (fstat(fd, &st)) {
450 idx_size = xsize_t(st.st_size);
451 if (idx_size < 4 * 256 + 20 + 20) {
453 return error("index file %s is too small", path);
455 idx_map = xmmap(NULL, idx_size, PROT_READ, MAP_PRIVATE, fd, 0);
458 /* a future index format would start with this, as older git
459 * binaries would fail the non-monotonic index check below.
460 * give a nicer warning to the user if we can.
463 if (hdr->idx_signature == htonl(PACK_IDX_SIGNATURE)) {
464 munmap(idx_map, idx_size);
465 return error("index file %s is a newer version"
466 " and is not supported by this binary"
467 " (try upgrading GIT to a newer version)",
473 for (i = 0; i < 256; i++) {
474 uint32_t n = ntohl(index[i]);
476 munmap(idx_map, idx_size);
477 return error("non-monotonic index %s", path);
484 * - 256 index entries 4 bytes each
485 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
486 * - 20-byte SHA1 of the packfile
487 * - 20-byte SHA1 file checksum
489 if (idx_size != 4*256 + nr * 24 + 20 + 20) {
490 munmap(idx_map, idx_size);
491 return error("wrong index file size in %s", path);
494 p->index_version = 1;
495 p->index_data = idx_map;
496 p->index_size = idx_size;
500 static void scan_windows(struct packed_git *p,
501 struct packed_git **lru_p,
502 struct pack_window **lru_w,
503 struct pack_window **lru_l)
505 struct pack_window *w, *w_l;
507 for (w_l = NULL, w = p->windows; w; w = w->next) {
509 if (!*lru_w || w->last_used < (*lru_w)->last_used) {
519 static int unuse_one_window(struct packed_git *current)
521 struct packed_git *p, *lru_p = NULL;
522 struct pack_window *lru_w = NULL, *lru_l = NULL;
525 scan_windows(current, &lru_p, &lru_w, &lru_l);
526 for (p = packed_git; p; p = p->next)
527 scan_windows(p, &lru_p, &lru_w, &lru_l);
529 munmap(lru_w->base, lru_w->len);
530 pack_mapped -= lru_w->len;
532 lru_l->next = lru_w->next;
534 lru_p->windows = lru_w->next;
535 if (!lru_p->windows && lru_p != current) {
536 close(lru_p->pack_fd);
547 void release_pack_memory(size_t need)
549 size_t cur = pack_mapped;
550 while (need >= (cur - pack_mapped) && unuse_one_window(NULL))
554 void unuse_pack(struct pack_window **w_cursor)
556 struct pack_window *w = *w_cursor;
564 * Do not call this directly as this leaks p->pack_fd on error return;
565 * call open_packed_git() instead.
567 static int open_packed_git_1(struct packed_git *p)
570 struct pack_header hdr;
571 unsigned char sha1[20];
572 unsigned char *idx_sha1;
575 p->pack_fd = open(p->pack_name, O_RDONLY);
576 if (p->pack_fd < 0 || fstat(p->pack_fd, &st))
579 /* If we created the struct before we had the pack we lack size. */
581 if (!S_ISREG(st.st_mode))
582 return error("packfile %s not a regular file", p->pack_name);
583 p->pack_size = st.st_size;
584 } else if (p->pack_size != st.st_size)
585 return error("packfile %s size changed", p->pack_name);
587 /* We leave these file descriptors open with sliding mmap;
588 * there is no point keeping them open across exec(), though.
590 fd_flag = fcntl(p->pack_fd, F_GETFD, 0);
592 return error("cannot determine file descriptor flags");
593 fd_flag |= FD_CLOEXEC;
594 if (fcntl(p->pack_fd, F_SETFD, fd_flag) == -1)
595 return error("cannot set FD_CLOEXEC");
597 /* Verify we recognize this pack file format. */
598 if (read_in_full(p->pack_fd, &hdr, sizeof(hdr)) != sizeof(hdr))
599 return error("file %s is far too short to be a packfile", p->pack_name);
600 if (hdr.hdr_signature != htonl(PACK_SIGNATURE))
601 return error("file %s is not a GIT packfile", p->pack_name);
602 if (!pack_version_ok(hdr.hdr_version))
603 return error("packfile %s is version %u and not supported"
604 " (try upgrading GIT to a newer version)",
605 p->pack_name, ntohl(hdr.hdr_version));
607 /* Verify the pack matches its index. */
608 if (num_packed_objects(p) != ntohl(hdr.hdr_entries))
609 return error("packfile %s claims to have %u objects"
610 " while index size indicates %u objects",
611 p->pack_name, ntohl(hdr.hdr_entries),
612 num_packed_objects(p));
613 if (lseek(p->pack_fd, p->pack_size - sizeof(sha1), SEEK_SET) == -1)
614 return error("end of packfile %s is unavailable", p->pack_name);
615 if (read_in_full(p->pack_fd, sha1, sizeof(sha1)) != sizeof(sha1))
616 return error("packfile %s signature is unavailable", p->pack_name);
617 idx_sha1 = ((unsigned char *)p->index_data) + p->index_size - 40;
618 if (hashcmp(sha1, idx_sha1))
619 return error("packfile %s does not match index", p->pack_name);
623 static int open_packed_git(struct packed_git *p)
625 if (!open_packed_git_1(p))
627 if (p->pack_fd != -1) {
634 static int in_window(struct pack_window *win, off_t offset)
636 /* We must promise at least 20 bytes (one hash) after the
637 * offset is available from this window, otherwise the offset
638 * is not actually in this window and a different window (which
639 * has that one hash excess) must be used. This is to support
640 * the object header and delta base parsing routines below.
642 off_t win_off = win->offset;
643 return win_off <= offset
644 && (offset + 20) <= (win_off + win->len);
647 unsigned char* use_pack(struct packed_git *p,
648 struct pack_window **w_cursor,
652 struct pack_window *win = *w_cursor;
654 if (p->pack_fd == -1 && open_packed_git(p))
655 die("packfile %s cannot be accessed", p->pack_name);
657 /* Since packfiles end in a hash of their content and its
658 * pointless to ask for an offset into the middle of that
659 * hash, and the in_window function above wouldn't match
660 * don't allow an offset too close to the end of the file.
662 if (offset > (p->pack_size - 20))
663 die("offset beyond end of packfile (truncated pack?)");
665 if (!win || !in_window(win, offset)) {
668 for (win = p->windows; win; win = win->next) {
669 if (in_window(win, offset))
673 size_t window_align = packed_git_window_size / 2;
675 win = xcalloc(1, sizeof(*win));
676 win->offset = (offset / window_align) * window_align;
677 len = p->pack_size - win->offset;
678 if (len > packed_git_window_size)
679 len = packed_git_window_size;
680 win->len = (size_t)len;
681 pack_mapped += win->len;
682 while (packed_git_limit < pack_mapped
683 && unuse_one_window(p))
685 win->base = xmmap(NULL, win->len,
686 PROT_READ, MAP_PRIVATE,
687 p->pack_fd, win->offset);
688 if (win->base == MAP_FAILED)
689 die("packfile %s cannot be mapped: %s",
694 if (pack_mapped > peak_pack_mapped)
695 peak_pack_mapped = pack_mapped;
696 if (pack_open_windows > peak_pack_open_windows)
697 peak_pack_open_windows = pack_open_windows;
698 win->next = p->windows;
702 if (win != *w_cursor) {
703 win->last_used = pack_used_ctr++;
707 offset -= win->offset;
709 *left = win->len - xsize_t(offset);
710 return win->base + offset;
713 struct packed_git *add_packed_git(const char *path, int path_len, int local)
716 struct packed_git *p = xmalloc(sizeof(*p) + path_len + 2);
719 * Make sure a corresponding .pack file exists and that
720 * the index looks sane.
722 path_len -= strlen(".idx");
725 memcpy(p->pack_name, path, path_len);
726 strcpy(p->pack_name + path_len, ".pack");
727 if (stat(p->pack_name, &st) || !S_ISREG(st.st_mode) ||
728 check_packed_git_idx(path, p)) {
733 /* ok, it looks sane as far as we can check without
734 * actually mapping the pack file.
736 p->pack_size = st.st_size;
740 p->pack_local = local;
741 p->mtime = st.st_mtime;
742 if (path_len < 40 || get_sha1_hex(path + path_len - 40, p->sha1))
747 struct packed_git *parse_pack_index(unsigned char *sha1)
749 char *path = sha1_pack_index_name(sha1);
750 return parse_pack_index_file(sha1, path);
753 struct packed_git *parse_pack_index_file(const unsigned char *sha1,
754 const char *idx_path)
756 const char *path = sha1_pack_name(sha1);
757 struct packed_git *p = xmalloc(sizeof(*p) + strlen(path) + 2);
759 if (check_packed_git_idx(idx_path, p)) {
764 strcpy(p->pack_name, path);
769 hashcpy(p->sha1, sha1);
773 void install_packed_git(struct packed_git *pack)
775 pack->next = packed_git;
779 static void prepare_packed_git_one(char *objdir, int local)
786 sprintf(path, "%s/pack", objdir);
791 error("unable to open object pack directory: %s: %s",
792 path, strerror(errno));
796 while ((de = readdir(dir)) != NULL) {
797 int namelen = strlen(de->d_name);
798 struct packed_git *p;
800 if (!has_extension(de->d_name, ".idx"))
803 /* Don't reopen a pack we already have. */
804 strcpy(path + len, de->d_name);
805 for (p = packed_git; p; p = p->next) {
806 if (!memcmp(path, p->pack_name, len + namelen - 4))
811 /* See if it really is a valid .idx file with corresponding
812 * .pack file that we can map.
814 p = add_packed_git(path, len + namelen, local);
817 install_packed_git(p);
822 static int sort_pack(const void *a_, const void *b_)
824 struct packed_git *a = *((struct packed_git **)a_);
825 struct packed_git *b = *((struct packed_git **)b_);
829 * Local packs tend to contain objects specific to our
830 * variant of the project than remote ones. In addition,
831 * remote ones could be on a network mounted filesystem.
832 * Favor local ones for these reasons.
834 st = a->pack_local - b->pack_local;
839 * Younger packs tend to contain more recent objects,
840 * and more recent objects tend to get accessed more
843 if (a->mtime < b->mtime)
845 else if (a->mtime == b->mtime)
850 static void rearrange_packed_git(void)
852 struct packed_git **ary, *p;
855 for (n = 0, p = packed_git; p; p = p->next)
860 /* prepare an array of packed_git for easier sorting */
861 ary = xcalloc(n, sizeof(struct packed_git *));
862 for (n = 0, p = packed_git; p; p = p->next)
865 qsort(ary, n, sizeof(struct packed_git *), sort_pack);
867 /* link them back again */
868 for (i = 0; i < n - 1; i++)
869 ary[i]->next = ary[i + 1];
870 ary[n - 1]->next = NULL;
876 static int prepare_packed_git_run_once = 0;
877 void prepare_packed_git(void)
879 struct alternate_object_database *alt;
881 if (prepare_packed_git_run_once)
883 prepare_packed_git_one(get_object_directory(), 1);
885 for (alt = alt_odb_list; alt; alt = alt->next) {
887 prepare_packed_git_one(alt->base, 0);
890 rearrange_packed_git();
891 prepare_packed_git_run_once = 1;
894 void reprepare_packed_git(void)
896 prepare_packed_git_run_once = 0;
897 prepare_packed_git();
900 int check_sha1_signature(const unsigned char *sha1, void *map, unsigned long size, const char *type)
902 unsigned char real_sha1[20];
903 hash_sha1_file(map, size, type, real_sha1);
904 return hashcmp(sha1, real_sha1) ? -1 : 0;
907 void *map_sha1_file(const unsigned char *sha1, unsigned long *size)
912 char *filename = find_sha1_file(sha1, &st);
918 fd = open(filename, O_RDONLY | sha1_file_open_flag);
920 /* See if it works without O_NOATIME */
921 switch (sha1_file_open_flag) {
923 fd = open(filename, O_RDONLY);
931 /* If it failed once, it will probably fail again.
932 * Stop using O_NOATIME
934 sha1_file_open_flag = 0;
936 *size = xsize_t(st.st_size);
937 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
942 int legacy_loose_object(unsigned char *map)
947 * Is it a zlib-compressed buffer? If so, the first byte
948 * must be 0x78 (15-bit window size, deflated), and the
949 * first 16-bit word is evenly divisible by 31
951 word = (map[0] << 8) + map[1];
952 if (map[0] == 0x78 && !(word % 31))
958 unsigned long unpack_object_header_gently(const unsigned char *buf, unsigned long len, enum object_type *type, unsigned long *sizep)
963 unsigned long used = 0;
966 *type = (c >> 4) & 7;
972 if (sizeof(long) * 8 <= shift)
975 size += (c & 0x7f) << shift;
982 static int unpack_sha1_header(z_stream *stream, unsigned char *map, unsigned long mapsize, void *buffer, unsigned long bufsiz)
984 unsigned long size, used;
985 static const char valid_loose_object_type[8] = {
987 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
988 0, /* "delta" and others are invalid in a loose object */
990 enum object_type type;
992 /* Get the data stream */
993 memset(stream, 0, sizeof(*stream));
994 stream->next_in = map;
995 stream->avail_in = mapsize;
996 stream->next_out = buffer;
997 stream->avail_out = bufsiz;
999 if (legacy_loose_object(map)) {
1000 inflateInit(stream);
1001 return inflate(stream, 0);
1004 used = unpack_object_header_gently(map, mapsize, &type, &size);
1005 if (!used || !valid_loose_object_type[type])
1010 /* Set up the stream for the rest.. */
1011 stream->next_in = map;
1012 stream->avail_in = mapsize;
1013 inflateInit(stream);
1015 /* And generate the fake traditional header */
1016 stream->total_out = 1 + snprintf(buffer, bufsiz, "%s %lu",
1017 typename(type), size);
1021 static void *unpack_sha1_rest(z_stream *stream, void *buffer, unsigned long size, const unsigned char *sha1)
1023 int bytes = strlen(buffer) + 1;
1024 unsigned char *buf = xmalloc(1+size);
1028 n = stream->total_out - bytes;
1031 memcpy(buf, (char *) buffer + bytes, n);
1034 stream->next_out = buf + bytes;
1035 stream->avail_out = size - bytes;
1036 while (status == Z_OK)
1037 status = inflate(stream, Z_FINISH);
1040 if ((status == Z_OK || status == Z_STREAM_END) && !stream->avail_in) {
1046 error("corrupt loose object '%s'", sha1_to_hex(sha1));
1047 else if (stream->avail_in)
1048 error("garbage at end of loose object '%s'",
1055 * We used to just use "sscanf()", but that's actually way
1056 * too permissive for what we want to check. So do an anal
1057 * object header parse by hand.
1059 static int parse_sha1_header(const char *hdr, unsigned long *sizep)
1066 * The type can be at most ten bytes (including the
1067 * terminating '\0' that we add), and is followed by
1076 if (i >= sizeof(type))
1082 * The length must follow immediately, and be in canonical
1083 * decimal format (ie "010" is not valid).
1085 size = *hdr++ - '0';
1090 unsigned long c = *hdr - '0';
1094 size = size * 10 + c;
1100 * The length must be followed by a zero byte
1102 return *hdr ? -1 : type_from_string(type);
1105 static void *unpack_sha1_file(void *map, unsigned long mapsize, enum object_type *type, unsigned long *size, const unsigned char *sha1)
1111 ret = unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr));
1112 if (ret < Z_OK || (*type = parse_sha1_header(hdr, size)) < 0)
1115 return unpack_sha1_rest(&stream, hdr, *size, sha1);
1118 static off_t get_delta_base(struct packed_git *p,
1119 struct pack_window **w_curs,
1121 enum object_type type,
1122 off_t delta_obj_offset)
1124 unsigned char *base_info = use_pack(p, w_curs, *curpos, NULL);
1127 /* use_pack() assured us we have [base_info, base_info + 20)
1128 * as a range that we can look at without walking off the
1129 * end of the mapped window. Its actually the hash size
1130 * that is assured. An OFS_DELTA longer than the hash size
1131 * is stupid, as then a REF_DELTA would be smaller to store.
1133 if (type == OBJ_OFS_DELTA) {
1135 unsigned char c = base_info[used++];
1136 base_offset = c & 127;
1139 if (!base_offset || base_offset & ~(~0UL >> 7))
1140 die("offset value overflow for delta base object");
1141 c = base_info[used++];
1142 base_offset = (base_offset << 7) + (c & 127);
1144 base_offset = delta_obj_offset - base_offset;
1145 if (base_offset >= delta_obj_offset)
1146 die("delta base offset out of bound");
1148 } else if (type == OBJ_REF_DELTA) {
1149 /* The base entry _must_ be in the same pack */
1150 base_offset = find_pack_entry_one(base_info, p);
1152 die("failed to find delta-pack base object %s",
1153 sha1_to_hex(base_info));
1156 die("I am totally screwed");
1160 /* forward declaration for a mutually recursive function */
1161 static int packed_object_info(struct packed_git *p, off_t offset,
1162 unsigned long *sizep);
1164 static int packed_delta_info(struct packed_git *p,
1165 struct pack_window **w_curs,
1167 enum object_type type,
1169 unsigned long *sizep)
1173 base_offset = get_delta_base(p, w_curs, &curpos, type, obj_offset);
1174 type = packed_object_info(p, base_offset, NULL);
1176 /* We choose to only get the type of the base object and
1177 * ignore potentially corrupt pack file that expects the delta
1178 * based on a base with a wrong size. This saves tons of
1182 const unsigned char *data;
1183 unsigned char delta_head[20], *in;
1187 memset(&stream, 0, sizeof(stream));
1188 stream.next_out = delta_head;
1189 stream.avail_out = sizeof(delta_head);
1191 inflateInit(&stream);
1193 in = use_pack(p, w_curs, curpos, &stream.avail_in);
1194 stream.next_in = in;
1195 st = inflate(&stream, Z_FINISH);
1196 curpos += stream.next_in - in;
1197 } while ((st == Z_OK || st == Z_BUF_ERROR)
1198 && stream.total_out < sizeof(delta_head));
1199 inflateEnd(&stream);
1200 if ((st != Z_STREAM_END) &&
1201 stream.total_out != sizeof(delta_head))
1202 die("delta data unpack-initial failed");
1204 /* Examine the initial part of the delta to figure out
1209 /* ignore base size */
1210 get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
1212 /* Read the result size */
1213 *sizep = get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
1219 static int unpack_object_header(struct packed_git *p,
1220 struct pack_window **w_curs,
1222 unsigned long *sizep)
1224 unsigned char *base;
1227 enum object_type type;
1229 /* use_pack() assures us we have [base, base + 20) available
1230 * as a range that we can look at at. (Its actually the hash
1231 * size that is assured.) With our object header encoding
1232 * the maximum deflated object size is 2^137, which is just
1233 * insane, so we know won't exceed what we have been given.
1235 base = use_pack(p, w_curs, *curpos, &left);
1236 used = unpack_object_header_gently(base, left, &type, sizep);
1238 die("object offset outside of pack file");
1244 const char *packed_object_info_detail(struct packed_git *p,
1246 unsigned long *size,
1247 unsigned long *store_size,
1248 unsigned int *delta_chain_length,
1249 unsigned char *base_sha1)
1251 struct pack_window *w_curs = NULL;
1253 unsigned long dummy;
1254 unsigned char *next_sha1;
1255 enum object_type type;
1257 *delta_chain_length = 0;
1258 curpos = obj_offset;
1259 type = unpack_object_header(p, &w_curs, &curpos, size);
1264 die("pack %s contains unknown object type %d",
1265 p->pack_name, type);
1270 *store_size = 0; /* notyet */
1271 unuse_pack(&w_curs);
1272 return typename(type);
1274 obj_offset = get_delta_base(p, &w_curs, &curpos, type, obj_offset);
1275 if (*delta_chain_length == 0) {
1276 /* TODO: find base_sha1 as pointed by curpos */
1281 next_sha1 = use_pack(p, &w_curs, curpos, NULL);
1282 if (*delta_chain_length == 0)
1283 hashcpy(base_sha1, next_sha1);
1284 obj_offset = find_pack_entry_one(next_sha1, p);
1287 (*delta_chain_length)++;
1288 curpos = obj_offset;
1289 type = unpack_object_header(p, &w_curs, &curpos, &dummy);
1293 static int packed_object_info(struct packed_git *p, off_t obj_offset,
1294 unsigned long *sizep)
1296 struct pack_window *w_curs = NULL;
1298 off_t curpos = obj_offset;
1299 enum object_type type;
1301 type = unpack_object_header(p, &w_curs, &curpos, &size);
1306 type = packed_delta_info(p, &w_curs, curpos,
1307 type, obj_offset, sizep);
1317 die("pack %s contains unknown object type %d",
1318 p->pack_name, type);
1320 unuse_pack(&w_curs);
1324 static void *unpack_compressed_entry(struct packed_git *p,
1325 struct pack_window **w_curs,
1331 unsigned char *buffer, *in;
1333 buffer = xmalloc(size + 1);
1335 memset(&stream, 0, sizeof(stream));
1336 stream.next_out = buffer;
1337 stream.avail_out = size;
1339 inflateInit(&stream);
1341 in = use_pack(p, w_curs, curpos, &stream.avail_in);
1342 stream.next_in = in;
1343 st = inflate(&stream, Z_FINISH);
1344 curpos += stream.next_in - in;
1345 } while (st == Z_OK || st == Z_BUF_ERROR);
1346 inflateEnd(&stream);
1347 if ((st != Z_STREAM_END) || stream.total_out != size) {
1355 #define MAX_DELTA_CACHE (256)
1357 static size_t delta_base_cached;
1358 static struct delta_base_cache_entry {
1359 struct packed_git *p;
1363 enum object_type type;
1364 } delta_base_cache[MAX_DELTA_CACHE];
1366 static unsigned long pack_entry_hash(struct packed_git *p, off_t base_offset)
1370 hash = (unsigned long)p + (unsigned long)base_offset;
1371 hash += (hash >> 8) + (hash >> 16);
1375 static void *cache_or_unpack_entry(struct packed_git *p, off_t base_offset,
1376 unsigned long *base_size, enum object_type *type, int keep_cache)
1379 unsigned long hash = pack_entry_hash(p, base_offset);
1380 struct delta_base_cache_entry *ent = delta_base_cache + hash;
1383 if (ret && ent->p == p && ent->base_offset == base_offset)
1384 goto found_cache_entry;
1385 return unpack_entry(p, base_offset, type, base_size);
1390 delta_base_cached -= ent->size;
1393 ret = xmalloc(ent->size + 1);
1394 memcpy(ret, ent->data, ent->size);
1395 ((char *)ret)[ent->size] = 0;
1398 *base_size = ent->size;
1402 static inline void release_delta_base_cache(struct delta_base_cache_entry *ent)
1407 delta_base_cached -= ent->size;
1411 static void add_delta_base_cache(struct packed_git *p, off_t base_offset,
1412 void *base, unsigned long base_size, enum object_type type)
1414 unsigned long i, hash = pack_entry_hash(p, base_offset);
1415 struct delta_base_cache_entry *ent = delta_base_cache + hash;
1417 release_delta_base_cache(ent);
1418 delta_base_cached += base_size;
1419 for (i = 0; delta_base_cached > delta_base_cache_limit
1420 && i < ARRAY_SIZE(delta_base_cache); i++) {
1421 struct delta_base_cache_entry *f = delta_base_cache + i;
1422 if (f->type == OBJ_BLOB)
1423 release_delta_base_cache(f);
1425 for (i = 0; delta_base_cached > delta_base_cache_limit
1426 && i < ARRAY_SIZE(delta_base_cache); i++)
1427 release_delta_base_cache(delta_base_cache + i);
1430 ent->base_offset = base_offset;
1433 ent->size = base_size;
1436 static void *unpack_delta_entry(struct packed_git *p,
1437 struct pack_window **w_curs,
1439 unsigned long delta_size,
1441 enum object_type *type,
1442 unsigned long *sizep)
1444 void *delta_data, *result, *base;
1445 unsigned long base_size;
1448 base_offset = get_delta_base(p, w_curs, &curpos, *type, obj_offset);
1449 base = cache_or_unpack_entry(p, base_offset, &base_size, type, 0);
1451 die("failed to read delta base object"
1452 " at %"PRIuMAX" from %s",
1453 (uintmax_t)base_offset, p->pack_name);
1455 delta_data = unpack_compressed_entry(p, w_curs, curpos, delta_size);
1456 result = patch_delta(base, base_size,
1457 delta_data, delta_size,
1460 die("failed to apply delta");
1462 add_delta_base_cache(p, base_offset, base, base_size, *type);
1466 void *unpack_entry(struct packed_git *p, off_t obj_offset,
1467 enum object_type *type, unsigned long *sizep)
1469 struct pack_window *w_curs = NULL;
1470 off_t curpos = obj_offset;
1473 *type = unpack_object_header(p, &w_curs, &curpos, sizep);
1477 data = unpack_delta_entry(p, &w_curs, curpos, *sizep,
1478 obj_offset, type, sizep);
1484 data = unpack_compressed_entry(p, &w_curs, curpos, *sizep);
1487 die("unknown object type %i in %s", *type, p->pack_name);
1489 unuse_pack(&w_curs);
1493 uint32_t num_packed_objects(const struct packed_git *p)
1495 /* See check_packed_git_idx() */
1496 return (uint32_t)((p->index_size - 20 - 20 - 4*256) / 24);
1499 int nth_packed_object_sha1(const struct packed_git *p, uint32_t n,
1500 unsigned char* sha1)
1502 const unsigned char *index = p->index_data;
1504 if (num_packed_objects(p) <= n)
1506 hashcpy(sha1, index + 24 * n + 4);
1510 off_t find_pack_entry_one(const unsigned char *sha1,
1511 struct packed_git *p)
1513 const uint32_t *level1_ofs = p->index_data;
1514 int hi = ntohl(level1_ofs[*sha1]);
1515 int lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
1516 const unsigned char *index = p->index_data;
1521 int mi = (lo + hi) / 2;
1522 int cmp = hashcmp(index + 24 * mi + 4, sha1);
1524 return ntohl(*((uint32_t *)((char *)index + (24 * mi))));
1533 static int matches_pack_name(struct packed_git *p, const char *ig)
1535 const char *last_c, *c;
1537 if (!strcmp(p->pack_name, ig))
1540 for (c = p->pack_name, last_c = c; *c;)
1545 if (!strcmp(last_c, ig))
1551 static int find_pack_entry(const unsigned char *sha1, struct pack_entry *e, const char **ignore_packed)
1553 struct packed_git *p;
1556 prepare_packed_git();
1558 for (p = packed_git; p; p = p->next) {
1559 if (ignore_packed) {
1561 for (ig = ignore_packed; *ig; ig++)
1562 if (!matches_pack_name(p, *ig))
1567 offset = find_pack_entry_one(sha1, p);
1570 * We are about to tell the caller where they can
1571 * locate the requested object. We better make
1572 * sure the packfile is still here and can be
1573 * accessed before supplying that answer, as
1574 * it may have been deleted since the index
1577 if (p->pack_fd == -1 && open_packed_git(p)) {
1578 error("packfile %s cannot be accessed", p->pack_name);
1583 hashcpy(e->sha1, sha1);
1590 struct packed_git *find_sha1_pack(const unsigned char *sha1,
1591 struct packed_git *packs)
1593 struct packed_git *p;
1595 for (p = packs; p; p = p->next) {
1596 if (find_pack_entry_one(sha1, p))
1603 static int sha1_loose_object_info(const unsigned char *sha1, unsigned long *sizep)
1606 unsigned long mapsize, size;
1611 map = map_sha1_file(sha1, &mapsize);
1613 return error("unable to find %s", sha1_to_hex(sha1));
1614 if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
1615 status = error("unable to unpack %s header",
1617 else if ((status = parse_sha1_header(hdr, &size)) < 0)
1618 status = error("unable to parse %s header", sha1_to_hex(sha1));
1621 inflateEnd(&stream);
1622 munmap(map, mapsize);
1626 int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
1628 struct pack_entry e;
1630 if (!find_pack_entry(sha1, &e, NULL)) {
1631 reprepare_packed_git();
1632 if (!find_pack_entry(sha1, &e, NULL))
1633 return sha1_loose_object_info(sha1, sizep);
1635 return packed_object_info(e.p, e.offset, sizep);
1638 static void *read_packed_sha1(const unsigned char *sha1,
1639 enum object_type *type, unsigned long *size)
1641 struct pack_entry e;
1643 if (!find_pack_entry(sha1, &e, NULL))
1646 return cache_or_unpack_entry(e.p, e.offset, size, type, 1);
1650 * This is meant to hold a *small* number of objects that you would
1651 * want read_sha1_file() to be able to return, but yet you do not want
1652 * to write them into the object store (e.g. a browse-only
1655 static struct cached_object {
1656 unsigned char sha1[20];
1657 enum object_type type;
1661 static int cached_object_nr, cached_object_alloc;
1663 static struct cached_object *find_cached_object(const unsigned char *sha1)
1666 struct cached_object *co = cached_objects;
1668 for (i = 0; i < cached_object_nr; i++, co++) {
1669 if (!hashcmp(co->sha1, sha1))
1675 int pretend_sha1_file(void *buf, unsigned long len, enum object_type type,
1676 unsigned char *sha1)
1678 struct cached_object *co;
1680 hash_sha1_file(buf, len, typename(type), sha1);
1681 if (has_sha1_file(sha1) || find_cached_object(sha1))
1683 if (cached_object_alloc <= cached_object_nr) {
1684 cached_object_alloc = alloc_nr(cached_object_alloc);
1685 cached_objects = xrealloc(cached_objects,
1686 sizeof(*cached_objects) *
1687 cached_object_alloc);
1689 co = &cached_objects[cached_object_nr++];
1692 co->buf = xmalloc(len);
1693 memcpy(co->buf, buf, len);
1694 hashcpy(co->sha1, sha1);
1698 void *read_sha1_file(const unsigned char *sha1, enum object_type *type,
1699 unsigned long *size)
1701 unsigned long mapsize;
1703 struct cached_object *co;
1705 co = find_cached_object(sha1);
1707 buf = xmalloc(co->size + 1);
1708 memcpy(buf, co->buf, co->size);
1709 ((char*)buf)[co->size] = 0;
1715 buf = read_packed_sha1(sha1, type, size);
1718 map = map_sha1_file(sha1, &mapsize);
1720 buf = unpack_sha1_file(map, mapsize, type, size, sha1);
1721 munmap(map, mapsize);
1724 reprepare_packed_git();
1725 return read_packed_sha1(sha1, type, size);
1728 void *read_object_with_reference(const unsigned char *sha1,
1729 const char *required_type_name,
1730 unsigned long *size,
1731 unsigned char *actual_sha1_return)
1733 enum object_type type, required_type;
1735 unsigned long isize;
1736 unsigned char actual_sha1[20];
1738 required_type = type_from_string(required_type_name);
1739 hashcpy(actual_sha1, sha1);
1741 int ref_length = -1;
1742 const char *ref_type = NULL;
1744 buffer = read_sha1_file(actual_sha1, &type, &isize);
1747 if (type == required_type) {
1749 if (actual_sha1_return)
1750 hashcpy(actual_sha1_return, actual_sha1);
1753 /* Handle references */
1754 else if (type == OBJ_COMMIT)
1756 else if (type == OBJ_TAG)
1757 ref_type = "object ";
1762 ref_length = strlen(ref_type);
1764 if (memcmp(buffer, ref_type, ref_length) ||
1765 get_sha1_hex((char *) buffer + ref_length, actual_sha1)) {
1770 /* Now we have the ID of the referred-to object in
1771 * actual_sha1. Check again. */
1775 static void write_sha1_file_prepare(void *buf, unsigned long len,
1776 const char *type, unsigned char *sha1,
1777 char *hdr, int *hdrlen)
1781 /* Generate the header */
1782 *hdrlen = sprintf(hdr, "%s %lu", type, len)+1;
1786 SHA1_Update(&c, hdr, *hdrlen);
1787 SHA1_Update(&c, buf, len);
1788 SHA1_Final(sha1, &c);
1792 * Link the tempfile to the final place, possibly creating the
1793 * last directory level as you do so.
1795 * Returns the errno on failure, 0 on success.
1797 static int link_temp_to_file(const char *tmpfile, const char *filename)
1802 if (!link(tmpfile, filename))
1806 * Try to mkdir the last path component if that failed.
1808 * Re-try the "link()" regardless of whether the mkdir
1809 * succeeds, since a race might mean that somebody
1813 dir = strrchr(filename, '/');
1816 if (!mkdir(filename, 0777) && adjust_shared_perm(filename)) {
1821 if (!link(tmpfile, filename))
1829 * Move the just written object into its final resting place
1831 int move_temp_to_file(const char *tmpfile, const char *filename)
1833 int ret = link_temp_to_file(tmpfile, filename);
1836 * Coda hack - coda doesn't like cross-directory links,
1837 * so we fall back to a rename, which will mean that it
1838 * won't be able to check collisions, but that's not a
1841 * The same holds for FAT formatted media.
1843 * When this succeeds, we just return 0. We have nothing
1846 if (ret && ret != EEXIST) {
1847 if (!rename(tmpfile, filename))
1853 if (ret != EEXIST) {
1854 return error("unable to write sha1 filename %s: %s\n", filename, strerror(ret));
1856 /* FIXME!!! Collision check here ? */
1862 static int write_buffer(int fd, const void *buf, size_t len)
1864 if (write_in_full(fd, buf, len) < 0)
1865 return error("file write error (%s)", strerror(errno));
1869 static int write_binary_header(unsigned char *hdr, enum object_type type, unsigned long len)
1874 c = (type << 4) | (len & 15);
1887 static void setup_object_header(z_stream *stream, const char *type, unsigned long len)
1889 int obj_type, hdrlen;
1891 if (use_legacy_headers) {
1892 while (deflate(stream, 0) == Z_OK)
1896 obj_type = type_from_string(type);
1897 hdrlen = write_binary_header(stream->next_out, obj_type, len);
1898 stream->total_out = hdrlen;
1899 stream->next_out += hdrlen;
1900 stream->avail_out -= hdrlen;
1903 int hash_sha1_file(void *buf, unsigned long len, const char *type,
1904 unsigned char *sha1)
1908 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
1912 int write_sha1_file(void *buf, unsigned long len, const char *type, unsigned char *returnsha1)
1915 unsigned char *compressed;
1917 unsigned char sha1[20];
1919 static char tmpfile[PATH_MAX];
1923 /* Normally if we have it in the pack then we do not bother writing
1924 * it out into .git/objects/??/?{38} file.
1926 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
1927 filename = sha1_file_name(sha1);
1929 hashcpy(returnsha1, sha1);
1930 if (has_sha1_file(sha1))
1932 fd = open(filename, O_RDONLY);
1935 * FIXME!!! We might do collision checking here, but we'd
1936 * need to uncompress the old file and check it. Later.
1942 if (errno != ENOENT) {
1943 return error("sha1 file %s: %s\n", filename, strerror(errno));
1946 snprintf(tmpfile, sizeof(tmpfile), "%s/obj_XXXXXX", get_object_directory());
1948 fd = mkstemp(tmpfile);
1951 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
1953 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
1957 memset(&stream, 0, sizeof(stream));
1958 deflateInit(&stream, zlib_compression_level);
1959 size = 8 + deflateBound(&stream, len+hdrlen);
1960 compressed = xmalloc(size);
1963 stream.next_out = compressed;
1964 stream.avail_out = size;
1966 /* First header.. */
1967 stream.next_in = (unsigned char *)hdr;
1968 stream.avail_in = hdrlen;
1969 setup_object_header(&stream, type, len);
1971 /* Then the data itself.. */
1972 stream.next_in = buf;
1973 stream.avail_in = len;
1974 while (deflate(&stream, Z_FINISH) == Z_OK)
1976 deflateEnd(&stream);
1977 size = stream.total_out;
1979 if (write_buffer(fd, compressed, size) < 0)
1980 die("unable to write sha1 file");
1985 return move_temp_to_file(tmpfile, filename);
1989 * We need to unpack and recompress the object for writing
1990 * it out to a different file.
1992 static void *repack_object(const unsigned char *sha1, unsigned long *objsize)
1996 unsigned char *unpacked;
1998 enum object_type type;
2003 /* need to unpack and recompress it by itself */
2004 unpacked = read_packed_sha1(sha1, &type, &len);
2006 error("cannot read sha1_file for %s", sha1_to_hex(sha1));
2008 hdrlen = sprintf(hdr, "%s %lu", typename(type), len) + 1;
2011 memset(&stream, 0, sizeof(stream));
2012 deflateInit(&stream, zlib_compression_level);
2013 size = deflateBound(&stream, len + hdrlen);
2014 buf = xmalloc(size);
2017 stream.next_out = buf;
2018 stream.avail_out = size;
2020 /* First header.. */
2021 stream.next_in = (void *)hdr;
2022 stream.avail_in = hdrlen;
2023 while (deflate(&stream, 0) == Z_OK)
2026 /* Then the data itself.. */
2027 stream.next_in = unpacked;
2028 stream.avail_in = len;
2029 while (deflate(&stream, Z_FINISH) == Z_OK)
2031 deflateEnd(&stream);
2034 *objsize = stream.total_out;
2038 int write_sha1_to_fd(int fd, const unsigned char *sha1)
2041 unsigned long objsize;
2042 void *buf = map_sha1_file(sha1, &objsize);
2045 retval = write_buffer(fd, buf, objsize);
2046 munmap(buf, objsize);
2050 buf = repack_object(sha1, &objsize);
2051 retval = write_buffer(fd, buf, objsize);
2056 int write_sha1_from_fd(const unsigned char *sha1, int fd, char *buffer,
2057 size_t bufsize, size_t *bufposn)
2059 char tmpfile[PATH_MAX];
2062 unsigned char real_sha1[20];
2063 unsigned char discard[4096];
2067 snprintf(tmpfile, sizeof(tmpfile), "%s/obj_XXXXXX", get_object_directory());
2069 local = mkstemp(tmpfile);
2072 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
2074 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
2077 memset(&stream, 0, sizeof(stream));
2079 inflateInit(&stream);
2086 stream.avail_in = *bufposn;
2087 stream.next_in = (unsigned char *) buffer;
2089 stream.next_out = discard;
2090 stream.avail_out = sizeof(discard);
2091 ret = inflate(&stream, Z_SYNC_FLUSH);
2092 SHA1_Update(&c, discard, sizeof(discard) -
2094 } while (stream.avail_in && ret == Z_OK);
2095 if (write_buffer(local, buffer, *bufposn - stream.avail_in) < 0)
2096 die("unable to write sha1 file");
2097 memmove(buffer, buffer + *bufposn - stream.avail_in,
2099 *bufposn = stream.avail_in;
2103 size = xread(fd, buffer + *bufposn, bufsize - *bufposn);
2108 return error("Connection closed?");
2109 perror("Reading from connection");
2114 inflateEnd(&stream);
2117 SHA1_Final(real_sha1, &c);
2118 if (ret != Z_STREAM_END) {
2120 return error("File %s corrupted", sha1_to_hex(sha1));
2122 if (hashcmp(sha1, real_sha1)) {
2124 return error("File %s has bad hash", sha1_to_hex(sha1));
2127 return move_temp_to_file(tmpfile, sha1_file_name(sha1));
2130 int has_pack_index(const unsigned char *sha1)
2133 if (stat(sha1_pack_index_name(sha1), &st))
2138 int has_pack_file(const unsigned char *sha1)
2141 if (stat(sha1_pack_name(sha1), &st))
2146 int has_sha1_pack(const unsigned char *sha1, const char **ignore_packed)
2148 struct pack_entry e;
2149 return find_pack_entry(sha1, &e, ignore_packed);
2152 int has_sha1_file(const unsigned char *sha1)
2155 struct pack_entry e;
2157 if (find_pack_entry(sha1, &e, NULL))
2159 return find_sha1_file(sha1, &st) ? 1 : 0;
2163 * reads from fd as long as possible into a supplied buffer of size bytes.
2164 * If necessary the buffer's size is increased using realloc()
2166 * returns 0 if anything went fine and -1 otherwise
2168 * NOTE: both buf and size may change, but even when -1 is returned
2169 * you still have to free() it yourself.
2171 int read_pipe(int fd, char** return_buf, unsigned long* return_size)
2173 char* buf = *return_buf;
2174 unsigned long size = *return_size;
2176 unsigned long off = 0;
2179 iret = xread(fd, buf + off, size - off);
2184 buf = xrealloc(buf, size);
2197 int index_pipe(unsigned char *sha1, int fd, const char *type, int write_object)
2199 unsigned long size = 4096;
2200 char *buf = xmalloc(size);
2203 if (read_pipe(fd, &buf, &size)) {
2211 ret = write_sha1_file(buf, size, type, sha1);
2213 ret = hash_sha1_file(buf, size, type, sha1);
2218 int index_fd(unsigned char *sha1, int fd, struct stat *st, int write_object,
2219 enum object_type type, const char *path)
2221 size_t size = xsize_t(st->st_size);
2223 int ret, re_allocated = 0;
2226 buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
2233 * Convert blobs to git internal format
2235 if ((type == OBJ_BLOB) && S_ISREG(st->st_mode)) {
2236 unsigned long nsize = size;
2238 if (convert_to_git(path, &nbuf, &nsize)) {
2248 ret = write_sha1_file(buf, size, typename(type), sha1);
2250 ret = hash_sha1_file(buf, size, typename(type), sha1);
2260 int index_path(unsigned char *sha1, const char *path, struct stat *st, int write_object)
2266 switch (st->st_mode & S_IFMT) {
2268 fd = open(path, O_RDONLY);
2270 return error("open(\"%s\"): %s", path,
2272 if (index_fd(sha1, fd, st, write_object, OBJ_BLOB, path) < 0)
2273 return error("%s: failed to insert into database",
2277 len = xsize_t(st->st_size);
2278 target = xmalloc(len + 1);
2279 if (readlink(path, target, len + 1) != st->st_size) {
2280 char *errstr = strerror(errno);
2282 return error("readlink(\"%s\"): %s", path,
2286 hash_sha1_file(target, len, blob_type, sha1);
2287 else if (write_sha1_file(target, len, blob_type, sha1))
2288 return error("%s: failed to insert into database",
2293 return error("%s: unsupported file type", path);
2298 int read_pack_header(int fd, struct pack_header *header)
2300 char *c = (char*)header;
2301 ssize_t remaining = sizeof(struct pack_header);
2303 ssize_t r = xread(fd, c, remaining);
2305 /* "eof before pack header was fully read" */
2306 return PH_ERROR_EOF;
2309 } while (remaining > 0);
2310 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2311 /* "protocol error (pack signature mismatch detected)" */
2312 return PH_ERROR_PACK_SIGNATURE;
2313 if (!pack_version_ok(header->hdr_version))
2314 /* "protocol error (pack version unsupported)" */
2315 return PH_ERROR_PROTOCOL;