distinguish error versus short read from read_in_full()
[git.git] / packfile.c
blob263efb7151a4c4b1ef2dbc96d53b0636ff8ce287
1 #include "cache.h"
2 #include "mru.h"
3 #include "pack.h"
4 #include "dir.h"
5 #include "mergesort.h"
6 #include "packfile.h"
7 #include "delta.h"
8 #include "list.h"
9 #include "streaming.h"
10 #include "sha1-lookup.h"
12 char *odb_pack_name(struct strbuf *buf,
13 const unsigned char *sha1,
14 const char *ext)
16 strbuf_reset(buf);
17 strbuf_addf(buf, "%s/pack/pack-%s.%s", get_object_directory(),
18 sha1_to_hex(sha1), ext);
19 return buf->buf;
22 char *sha1_pack_name(const unsigned char *sha1)
24 static struct strbuf buf = STRBUF_INIT;
25 return odb_pack_name(&buf, sha1, "pack");
28 char *sha1_pack_index_name(const unsigned char *sha1)
30 static struct strbuf buf = STRBUF_INIT;
31 return odb_pack_name(&buf, sha1, "idx");
34 static unsigned int pack_used_ctr;
35 static unsigned int pack_mmap_calls;
36 static unsigned int peak_pack_open_windows;
37 static unsigned int pack_open_windows;
38 static unsigned int pack_open_fds;
39 static unsigned int pack_max_fds;
40 static size_t peak_pack_mapped;
41 static size_t pack_mapped;
42 struct packed_git *packed_git;
44 static struct mru packed_git_mru_storage;
45 struct mru *packed_git_mru = &packed_git_mru_storage;
47 #define SZ_FMT PRIuMAX
48 static inline uintmax_t sz_fmt(size_t s) { return s; }
50 void pack_report(void)
52 fprintf(stderr,
53 "pack_report: getpagesize() = %10" SZ_FMT "\n"
54 "pack_report: core.packedGitWindowSize = %10" SZ_FMT "\n"
55 "pack_report: core.packedGitLimit = %10" SZ_FMT "\n",
56 sz_fmt(getpagesize()),
57 sz_fmt(packed_git_window_size),
58 sz_fmt(packed_git_limit));
59 fprintf(stderr,
60 "pack_report: pack_used_ctr = %10u\n"
61 "pack_report: pack_mmap_calls = %10u\n"
62 "pack_report: pack_open_windows = %10u / %10u\n"
63 "pack_report: pack_mapped = "
64 "%10" SZ_FMT " / %10" SZ_FMT "\n",
65 pack_used_ctr,
66 pack_mmap_calls,
67 pack_open_windows, peak_pack_open_windows,
68 sz_fmt(pack_mapped), sz_fmt(peak_pack_mapped));
72 * Open and mmap the index file at path, perform a couple of
73 * consistency checks, then record its information to p. Return 0 on
74 * success.
76 static int check_packed_git_idx(const char *path, struct packed_git *p)
78 void *idx_map;
79 struct pack_idx_header *hdr;
80 size_t idx_size;
81 uint32_t version, nr, i, *index;
82 int fd = git_open(path);
83 struct stat st;
85 if (fd < 0)
86 return -1;
87 if (fstat(fd, &st)) {
88 close(fd);
89 return -1;
91 idx_size = xsize_t(st.st_size);
92 if (idx_size < 4 * 256 + 20 + 20) {
93 close(fd);
94 return error("index file %s is too small", path);
96 idx_map = xmmap(NULL, idx_size, PROT_READ, MAP_PRIVATE, fd, 0);
97 close(fd);
99 hdr = idx_map;
100 if (hdr->idx_signature == htonl(PACK_IDX_SIGNATURE)) {
101 version = ntohl(hdr->idx_version);
102 if (version < 2 || version > 2) {
103 munmap(idx_map, idx_size);
104 return error("index file %s is version %"PRIu32
105 " and is not supported by this binary"
106 " (try upgrading GIT to a newer version)",
107 path, version);
109 } else
110 version = 1;
112 nr = 0;
113 index = idx_map;
114 if (version > 1)
115 index += 2; /* skip index header */
116 for (i = 0; i < 256; i++) {
117 uint32_t n = ntohl(index[i]);
118 if (n < nr) {
119 munmap(idx_map, idx_size);
120 return error("non-monotonic index %s", path);
122 nr = n;
125 if (version == 1) {
127 * Total size:
128 * - 256 index entries 4 bytes each
129 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
130 * - 20-byte SHA1 of the packfile
131 * - 20-byte SHA1 file checksum
133 if (idx_size != 4*256 + nr * 24 + 20 + 20) {
134 munmap(idx_map, idx_size);
135 return error("wrong index v1 file size in %s", path);
137 } else if (version == 2) {
139 * Minimum size:
140 * - 8 bytes of header
141 * - 256 index entries 4 bytes each
142 * - 20-byte sha1 entry * nr
143 * - 4-byte crc entry * nr
144 * - 4-byte offset entry * nr
145 * - 20-byte SHA1 of the packfile
146 * - 20-byte SHA1 file checksum
147 * And after the 4-byte offset table might be a
148 * variable sized table containing 8-byte entries
149 * for offsets larger than 2^31.
151 unsigned long min_size = 8 + 4*256 + nr*(20 + 4 + 4) + 20 + 20;
152 unsigned long max_size = min_size;
153 if (nr)
154 max_size += (nr - 1)*8;
155 if (idx_size < min_size || idx_size > max_size) {
156 munmap(idx_map, idx_size);
157 return error("wrong index v2 file size in %s", path);
159 if (idx_size != min_size &&
161 * make sure we can deal with large pack offsets.
162 * 31-bit signed offset won't be enough, neither
163 * 32-bit unsigned one will be.
165 (sizeof(off_t) <= 4)) {
166 munmap(idx_map, idx_size);
167 return error("pack too large for current definition of off_t in %s", path);
171 p->index_version = version;
172 p->index_data = idx_map;
173 p->index_size = idx_size;
174 p->num_objects = nr;
175 return 0;
178 int open_pack_index(struct packed_git *p)
180 char *idx_name;
181 size_t len;
182 int ret;
184 if (p->index_data)
185 return 0;
187 if (!strip_suffix(p->pack_name, ".pack", &len))
188 die("BUG: pack_name does not end in .pack");
189 idx_name = xstrfmt("%.*s.idx", (int)len, p->pack_name);
190 ret = check_packed_git_idx(idx_name, p);
191 free(idx_name);
192 return ret;
195 static struct packed_git *alloc_packed_git(int extra)
197 struct packed_git *p = xmalloc(st_add(sizeof(*p), extra));
198 memset(p, 0, sizeof(*p));
199 p->pack_fd = -1;
200 return p;
203 struct packed_git *parse_pack_index(unsigned char *sha1, const char *idx_path)
205 const char *path = sha1_pack_name(sha1);
206 size_t alloc = st_add(strlen(path), 1);
207 struct packed_git *p = alloc_packed_git(alloc);
209 memcpy(p->pack_name, path, alloc); /* includes NUL */
210 hashcpy(p->sha1, sha1);
211 if (check_packed_git_idx(idx_path, p)) {
212 free(p);
213 return NULL;
216 return p;
219 static void scan_windows(struct packed_git *p,
220 struct packed_git **lru_p,
221 struct pack_window **lru_w,
222 struct pack_window **lru_l)
224 struct pack_window *w, *w_l;
226 for (w_l = NULL, w = p->windows; w; w = w->next) {
227 if (!w->inuse_cnt) {
228 if (!*lru_w || w->last_used < (*lru_w)->last_used) {
229 *lru_p = p;
230 *lru_w = w;
231 *lru_l = w_l;
234 w_l = w;
238 static int unuse_one_window(struct packed_git *current)
240 struct packed_git *p, *lru_p = NULL;
241 struct pack_window *lru_w = NULL, *lru_l = NULL;
243 if (current)
244 scan_windows(current, &lru_p, &lru_w, &lru_l);
245 for (p = packed_git; p; p = p->next)
246 scan_windows(p, &lru_p, &lru_w, &lru_l);
247 if (lru_p) {
248 munmap(lru_w->base, lru_w->len);
249 pack_mapped -= lru_w->len;
250 if (lru_l)
251 lru_l->next = lru_w->next;
252 else
253 lru_p->windows = lru_w->next;
254 free(lru_w);
255 pack_open_windows--;
256 return 1;
258 return 0;
261 void release_pack_memory(size_t need)
263 size_t cur = pack_mapped;
264 while (need >= (cur - pack_mapped) && unuse_one_window(NULL))
265 ; /* nothing */
268 void close_pack_windows(struct packed_git *p)
270 while (p->windows) {
271 struct pack_window *w = p->windows;
273 if (w->inuse_cnt)
274 die("pack '%s' still has open windows to it",
275 p->pack_name);
276 munmap(w->base, w->len);
277 pack_mapped -= w->len;
278 pack_open_windows--;
279 p->windows = w->next;
280 free(w);
284 static int close_pack_fd(struct packed_git *p)
286 if (p->pack_fd < 0)
287 return 0;
289 close(p->pack_fd);
290 pack_open_fds--;
291 p->pack_fd = -1;
293 return 1;
296 void close_pack_index(struct packed_git *p)
298 if (p->index_data) {
299 munmap((void *)p->index_data, p->index_size);
300 p->index_data = NULL;
304 static void close_pack(struct packed_git *p)
306 close_pack_windows(p);
307 close_pack_fd(p);
308 close_pack_index(p);
311 void close_all_packs(void)
313 struct packed_git *p;
315 for (p = packed_git; p; p = p->next)
316 if (p->do_not_close)
317 die("BUG: want to close pack marked 'do-not-close'");
318 else
319 close_pack(p);
323 * The LRU pack is the one with the oldest MRU window, preferring packs
324 * with no used windows, or the oldest mtime if it has no windows allocated.
326 static void find_lru_pack(struct packed_git *p, struct packed_git **lru_p, struct pack_window **mru_w, int *accept_windows_inuse)
328 struct pack_window *w, *this_mru_w;
329 int has_windows_inuse = 0;
332 * Reject this pack if it has windows and the previously selected
333 * one does not. If this pack does not have windows, reject
334 * it if the pack file is newer than the previously selected one.
336 if (*lru_p && !*mru_w && (p->windows || p->mtime > (*lru_p)->mtime))
337 return;
339 for (w = this_mru_w = p->windows; w; w = w->next) {
341 * Reject this pack if any of its windows are in use,
342 * but the previously selected pack did not have any
343 * inuse windows. Otherwise, record that this pack
344 * has windows in use.
346 if (w->inuse_cnt) {
347 if (*accept_windows_inuse)
348 has_windows_inuse = 1;
349 else
350 return;
353 if (w->last_used > this_mru_w->last_used)
354 this_mru_w = w;
357 * Reject this pack if it has windows that have been
358 * used more recently than the previously selected pack.
359 * If the previously selected pack had windows inuse and
360 * we have not encountered a window in this pack that is
361 * inuse, skip this check since we prefer a pack with no
362 * inuse windows to one that has inuse windows.
364 if (*mru_w && *accept_windows_inuse == has_windows_inuse &&
365 this_mru_w->last_used > (*mru_w)->last_used)
366 return;
370 * Select this pack.
372 *mru_w = this_mru_w;
373 *lru_p = p;
374 *accept_windows_inuse = has_windows_inuse;
377 static int close_one_pack(void)
379 struct packed_git *p, *lru_p = NULL;
380 struct pack_window *mru_w = NULL;
381 int accept_windows_inuse = 1;
383 for (p = packed_git; p; p = p->next) {
384 if (p->pack_fd == -1)
385 continue;
386 find_lru_pack(p, &lru_p, &mru_w, &accept_windows_inuse);
389 if (lru_p)
390 return close_pack_fd(lru_p);
392 return 0;
395 static unsigned int get_max_fd_limit(void)
397 #ifdef RLIMIT_NOFILE
399 struct rlimit lim;
401 if (!getrlimit(RLIMIT_NOFILE, &lim))
402 return lim.rlim_cur;
404 #endif
406 #ifdef _SC_OPEN_MAX
408 long open_max = sysconf(_SC_OPEN_MAX);
409 if (0 < open_max)
410 return open_max;
412 * Otherwise, we got -1 for one of the two
413 * reasons:
415 * (1) sysconf() did not understand _SC_OPEN_MAX
416 * and signaled an error with -1; or
417 * (2) sysconf() said there is no limit.
419 * We _could_ clear errno before calling sysconf() to
420 * tell these two cases apart and return a huge number
421 * in the latter case to let the caller cap it to a
422 * value that is not so selfish, but letting the
423 * fallback OPEN_MAX codepath take care of these cases
424 * is a lot simpler.
427 #endif
429 #ifdef OPEN_MAX
430 return OPEN_MAX;
431 #else
432 return 1; /* see the caller ;-) */
433 #endif
437 * Do not call this directly as this leaks p->pack_fd on error return;
438 * call open_packed_git() instead.
440 static int open_packed_git_1(struct packed_git *p)
442 struct stat st;
443 struct pack_header hdr;
444 unsigned char sha1[20];
445 unsigned char *idx_sha1;
446 long fd_flag;
447 ssize_t read_result;
449 if (!p->index_data && open_pack_index(p))
450 return error("packfile %s index unavailable", p->pack_name);
452 if (!pack_max_fds) {
453 unsigned int max_fds = get_max_fd_limit();
455 /* Save 3 for stdin/stdout/stderr, 22 for work */
456 if (25 < max_fds)
457 pack_max_fds = max_fds - 25;
458 else
459 pack_max_fds = 1;
462 while (pack_max_fds <= pack_open_fds && close_one_pack())
463 ; /* nothing */
465 p->pack_fd = git_open(p->pack_name);
466 if (p->pack_fd < 0 || fstat(p->pack_fd, &st))
467 return -1;
468 pack_open_fds++;
470 /* If we created the struct before we had the pack we lack size. */
471 if (!p->pack_size) {
472 if (!S_ISREG(st.st_mode))
473 return error("packfile %s not a regular file", p->pack_name);
474 p->pack_size = st.st_size;
475 } else if (p->pack_size != st.st_size)
476 return error("packfile %s size changed", p->pack_name);
478 /* We leave these file descriptors open with sliding mmap;
479 * there is no point keeping them open across exec(), though.
481 fd_flag = fcntl(p->pack_fd, F_GETFD, 0);
482 if (fd_flag < 0)
483 return error("cannot determine file descriptor flags");
484 fd_flag |= FD_CLOEXEC;
485 if (fcntl(p->pack_fd, F_SETFD, fd_flag) == -1)
486 return error("cannot set FD_CLOEXEC");
488 /* Verify we recognize this pack file format. */
489 read_result = read_in_full(p->pack_fd, &hdr, sizeof(hdr));
490 if (read_result < 0)
491 return error_errno("error reading from %s", p->pack_name);
492 if (read_result != sizeof(hdr))
493 return error("file %s is far too short to be a packfile", p->pack_name);
494 if (hdr.hdr_signature != htonl(PACK_SIGNATURE))
495 return error("file %s is not a GIT packfile", p->pack_name);
496 if (!pack_version_ok(hdr.hdr_version))
497 return error("packfile %s is version %"PRIu32" and not"
498 " supported (try upgrading GIT to a newer version)",
499 p->pack_name, ntohl(hdr.hdr_version));
501 /* Verify the pack matches its index. */
502 if (p->num_objects != ntohl(hdr.hdr_entries))
503 return error("packfile %s claims to have %"PRIu32" objects"
504 " while index indicates %"PRIu32" objects",
505 p->pack_name, ntohl(hdr.hdr_entries),
506 p->num_objects);
507 if (lseek(p->pack_fd, p->pack_size - sizeof(sha1), SEEK_SET) == -1)
508 return error("end of packfile %s is unavailable", p->pack_name);
509 read_result = read_in_full(p->pack_fd, sha1, sizeof(sha1));
510 if (read_result < 0)
511 return error_errno("error reading from %s", p->pack_name);
512 if (read_result != sizeof(sha1))
513 return error("packfile %s signature is unavailable", p->pack_name);
514 idx_sha1 = ((unsigned char *)p->index_data) + p->index_size - 40;
515 if (hashcmp(sha1, idx_sha1))
516 return error("packfile %s does not match index", p->pack_name);
517 return 0;
520 static int open_packed_git(struct packed_git *p)
522 if (!open_packed_git_1(p))
523 return 0;
524 close_pack_fd(p);
525 return -1;
528 static int in_window(struct pack_window *win, off_t offset)
530 /* We must promise at least 20 bytes (one hash) after the
531 * offset is available from this window, otherwise the offset
532 * is not actually in this window and a different window (which
533 * has that one hash excess) must be used. This is to support
534 * the object header and delta base parsing routines below.
536 off_t win_off = win->offset;
537 return win_off <= offset
538 && (offset + 20) <= (win_off + win->len);
541 unsigned char *use_pack(struct packed_git *p,
542 struct pack_window **w_cursor,
543 off_t offset,
544 unsigned long *left)
546 struct pack_window *win = *w_cursor;
548 /* Since packfiles end in a hash of their content and it's
549 * pointless to ask for an offset into the middle of that
550 * hash, and the in_window function above wouldn't match
551 * don't allow an offset too close to the end of the file.
553 if (!p->pack_size && p->pack_fd == -1 && open_packed_git(p))
554 die("packfile %s cannot be accessed", p->pack_name);
555 if (offset > (p->pack_size - 20))
556 die("offset beyond end of packfile (truncated pack?)");
557 if (offset < 0)
558 die(_("offset before end of packfile (broken .idx?)"));
560 if (!win || !in_window(win, offset)) {
561 if (win)
562 win->inuse_cnt--;
563 for (win = p->windows; win; win = win->next) {
564 if (in_window(win, offset))
565 break;
567 if (!win) {
568 size_t window_align = packed_git_window_size / 2;
569 off_t len;
571 if (p->pack_fd == -1 && open_packed_git(p))
572 die("packfile %s cannot be accessed", p->pack_name);
574 win = xcalloc(1, sizeof(*win));
575 win->offset = (offset / window_align) * window_align;
576 len = p->pack_size - win->offset;
577 if (len > packed_git_window_size)
578 len = packed_git_window_size;
579 win->len = (size_t)len;
580 pack_mapped += win->len;
581 while (packed_git_limit < pack_mapped
582 && unuse_one_window(p))
583 ; /* nothing */
584 win->base = xmmap(NULL, win->len,
585 PROT_READ, MAP_PRIVATE,
586 p->pack_fd, win->offset);
587 if (win->base == MAP_FAILED)
588 die_errno("packfile %s cannot be mapped",
589 p->pack_name);
590 if (!win->offset && win->len == p->pack_size
591 && !p->do_not_close)
592 close_pack_fd(p);
593 pack_mmap_calls++;
594 pack_open_windows++;
595 if (pack_mapped > peak_pack_mapped)
596 peak_pack_mapped = pack_mapped;
597 if (pack_open_windows > peak_pack_open_windows)
598 peak_pack_open_windows = pack_open_windows;
599 win->next = p->windows;
600 p->windows = win;
603 if (win != *w_cursor) {
604 win->last_used = pack_used_ctr++;
605 win->inuse_cnt++;
606 *w_cursor = win;
608 offset -= win->offset;
609 if (left)
610 *left = win->len - xsize_t(offset);
611 return win->base + offset;
614 void unuse_pack(struct pack_window **w_cursor)
616 struct pack_window *w = *w_cursor;
617 if (w) {
618 w->inuse_cnt--;
619 *w_cursor = NULL;
623 static void try_to_free_pack_memory(size_t size)
625 release_pack_memory(size);
628 struct packed_git *add_packed_git(const char *path, size_t path_len, int local)
630 static int have_set_try_to_free_routine;
631 struct stat st;
632 size_t alloc;
633 struct packed_git *p;
635 if (!have_set_try_to_free_routine) {
636 have_set_try_to_free_routine = 1;
637 set_try_to_free_routine(try_to_free_pack_memory);
641 * Make sure a corresponding .pack file exists and that
642 * the index looks sane.
644 if (!strip_suffix_mem(path, &path_len, ".idx"))
645 return NULL;
648 * ".pack" is long enough to hold any suffix we're adding (and
649 * the use xsnprintf double-checks that)
651 alloc = st_add3(path_len, strlen(".pack"), 1);
652 p = alloc_packed_git(alloc);
653 memcpy(p->pack_name, path, path_len);
655 xsnprintf(p->pack_name + path_len, alloc - path_len, ".keep");
656 if (!access(p->pack_name, F_OK))
657 p->pack_keep = 1;
659 xsnprintf(p->pack_name + path_len, alloc - path_len, ".pack");
660 if (stat(p->pack_name, &st) || !S_ISREG(st.st_mode)) {
661 free(p);
662 return NULL;
665 /* ok, it looks sane as far as we can check without
666 * actually mapping the pack file.
668 p->pack_size = st.st_size;
669 p->pack_local = local;
670 p->mtime = st.st_mtime;
671 if (path_len < 40 || get_sha1_hex(path + path_len - 40, p->sha1))
672 hashclr(p->sha1);
673 return p;
676 void install_packed_git(struct packed_git *pack)
678 if (pack->pack_fd != -1)
679 pack_open_fds++;
681 pack->next = packed_git;
682 packed_git = pack;
685 void (*report_garbage)(unsigned seen_bits, const char *path);
687 static void report_helper(const struct string_list *list,
688 int seen_bits, int first, int last)
690 if (seen_bits == (PACKDIR_FILE_PACK|PACKDIR_FILE_IDX))
691 return;
693 for (; first < last; first++)
694 report_garbage(seen_bits, list->items[first].string);
697 static void report_pack_garbage(struct string_list *list)
699 int i, baselen = -1, first = 0, seen_bits = 0;
701 if (!report_garbage)
702 return;
704 string_list_sort(list);
706 for (i = 0; i < list->nr; i++) {
707 const char *path = list->items[i].string;
708 if (baselen != -1 &&
709 strncmp(path, list->items[first].string, baselen)) {
710 report_helper(list, seen_bits, first, i);
711 baselen = -1;
712 seen_bits = 0;
714 if (baselen == -1) {
715 const char *dot = strrchr(path, '.');
716 if (!dot) {
717 report_garbage(PACKDIR_FILE_GARBAGE, path);
718 continue;
720 baselen = dot - path + 1;
721 first = i;
723 if (!strcmp(path + baselen, "pack"))
724 seen_bits |= 1;
725 else if (!strcmp(path + baselen, "idx"))
726 seen_bits |= 2;
728 report_helper(list, seen_bits, first, list->nr);
731 static void prepare_packed_git_one(char *objdir, int local)
733 struct strbuf path = STRBUF_INIT;
734 size_t dirnamelen;
735 DIR *dir;
736 struct dirent *de;
737 struct string_list garbage = STRING_LIST_INIT_DUP;
739 strbuf_addstr(&path, objdir);
740 strbuf_addstr(&path, "/pack");
741 dir = opendir(path.buf);
742 if (!dir) {
743 if (errno != ENOENT)
744 error_errno("unable to open object pack directory: %s",
745 path.buf);
746 strbuf_release(&path);
747 return;
749 strbuf_addch(&path, '/');
750 dirnamelen = path.len;
751 while ((de = readdir(dir)) != NULL) {
752 struct packed_git *p;
753 size_t base_len;
755 if (is_dot_or_dotdot(de->d_name))
756 continue;
758 strbuf_setlen(&path, dirnamelen);
759 strbuf_addstr(&path, de->d_name);
761 base_len = path.len;
762 if (strip_suffix_mem(path.buf, &base_len, ".idx")) {
763 /* Don't reopen a pack we already have. */
764 for (p = packed_git; p; p = p->next) {
765 size_t len;
766 if (strip_suffix(p->pack_name, ".pack", &len) &&
767 len == base_len &&
768 !memcmp(p->pack_name, path.buf, len))
769 break;
771 if (p == NULL &&
773 * See if it really is a valid .idx file with
774 * corresponding .pack file that we can map.
776 (p = add_packed_git(path.buf, path.len, local)) != NULL)
777 install_packed_git(p);
780 if (!report_garbage)
781 continue;
783 if (ends_with(de->d_name, ".idx") ||
784 ends_with(de->d_name, ".pack") ||
785 ends_with(de->d_name, ".bitmap") ||
786 ends_with(de->d_name, ".keep"))
787 string_list_append(&garbage, path.buf);
788 else
789 report_garbage(PACKDIR_FILE_GARBAGE, path.buf);
791 closedir(dir);
792 report_pack_garbage(&garbage);
793 string_list_clear(&garbage, 0);
794 strbuf_release(&path);
797 static int approximate_object_count_valid;
800 * Give a fast, rough count of the number of objects in the repository. This
801 * ignores loose objects completely. If you have a lot of them, then either
802 * you should repack because your performance will be awful, or they are
803 * all unreachable objects about to be pruned, in which case they're not really
804 * interesting as a measure of repo size in the first place.
806 unsigned long approximate_object_count(void)
808 static unsigned long count;
809 if (!approximate_object_count_valid) {
810 struct packed_git *p;
812 prepare_packed_git();
813 count = 0;
814 for (p = packed_git; p; p = p->next) {
815 if (open_pack_index(p))
816 continue;
817 count += p->num_objects;
820 return count;
823 static void *get_next_packed_git(const void *p)
825 return ((const struct packed_git *)p)->next;
828 static void set_next_packed_git(void *p, void *next)
830 ((struct packed_git *)p)->next = next;
833 static int sort_pack(const void *a_, const void *b_)
835 const struct packed_git *a = a_;
836 const struct packed_git *b = b_;
837 int st;
840 * Local packs tend to contain objects specific to our
841 * variant of the project than remote ones. In addition,
842 * remote ones could be on a network mounted filesystem.
843 * Favor local ones for these reasons.
845 st = a->pack_local - b->pack_local;
846 if (st)
847 return -st;
850 * Younger packs tend to contain more recent objects,
851 * and more recent objects tend to get accessed more
852 * often.
854 if (a->mtime < b->mtime)
855 return 1;
856 else if (a->mtime == b->mtime)
857 return 0;
858 return -1;
861 static void rearrange_packed_git(void)
863 packed_git = llist_mergesort(packed_git, get_next_packed_git,
864 set_next_packed_git, sort_pack);
867 static void prepare_packed_git_mru(void)
869 struct packed_git *p;
871 mru_clear(packed_git_mru);
872 for (p = packed_git; p; p = p->next)
873 mru_append(packed_git_mru, p);
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)
882 return;
883 prepare_packed_git_one(get_object_directory(), 1);
884 prepare_alt_odb();
885 for (alt = alt_odb_list; alt; alt = alt->next)
886 prepare_packed_git_one(alt->path, 0);
887 rearrange_packed_git();
888 prepare_packed_git_mru();
889 prepare_packed_git_run_once = 1;
892 void reprepare_packed_git(void)
894 approximate_object_count_valid = 0;
895 prepare_packed_git_run_once = 0;
896 prepare_packed_git();
899 unsigned long unpack_object_header_buffer(const unsigned char *buf,
900 unsigned long len, enum object_type *type, unsigned long *sizep)
902 unsigned shift;
903 unsigned long size, c;
904 unsigned long used = 0;
906 c = buf[used++];
907 *type = (c >> 4) & 7;
908 size = c & 15;
909 shift = 4;
910 while (c & 0x80) {
911 if (len <= used || bitsizeof(long) <= shift) {
912 error("bad object header");
913 size = used = 0;
914 break;
916 c = buf[used++];
917 size += (c & 0x7f) << shift;
918 shift += 7;
920 *sizep = size;
921 return used;
924 unsigned long get_size_from_delta(struct packed_git *p,
925 struct pack_window **w_curs,
926 off_t curpos)
928 const unsigned char *data;
929 unsigned char delta_head[20], *in;
930 git_zstream stream;
931 int st;
933 memset(&stream, 0, sizeof(stream));
934 stream.next_out = delta_head;
935 stream.avail_out = sizeof(delta_head);
937 git_inflate_init(&stream);
938 do {
939 in = use_pack(p, w_curs, curpos, &stream.avail_in);
940 stream.next_in = in;
941 st = git_inflate(&stream, Z_FINISH);
942 curpos += stream.next_in - in;
943 } while ((st == Z_OK || st == Z_BUF_ERROR) &&
944 stream.total_out < sizeof(delta_head));
945 git_inflate_end(&stream);
946 if ((st != Z_STREAM_END) && stream.total_out != sizeof(delta_head)) {
947 error("delta data unpack-initial failed");
948 return 0;
951 /* Examine the initial part of the delta to figure out
952 * the result size.
954 data = delta_head;
956 /* ignore base size */
957 get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
959 /* Read the result size */
960 return get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
963 int unpack_object_header(struct packed_git *p,
964 struct pack_window **w_curs,
965 off_t *curpos,
966 unsigned long *sizep)
968 unsigned char *base;
969 unsigned long left;
970 unsigned long used;
971 enum object_type type;
973 /* use_pack() assures us we have [base, base + 20) available
974 * as a range that we can look at. (Its actually the hash
975 * size that is assured.) With our object header encoding
976 * the maximum deflated object size is 2^137, which is just
977 * insane, so we know won't exceed what we have been given.
979 base = use_pack(p, w_curs, *curpos, &left);
980 used = unpack_object_header_buffer(base, left, &type, sizep);
981 if (!used) {
982 type = OBJ_BAD;
983 } else
984 *curpos += used;
986 return type;
989 void mark_bad_packed_object(struct packed_git *p, const unsigned char *sha1)
991 unsigned i;
992 for (i = 0; i < p->num_bad_objects; i++)
993 if (!hashcmp(sha1, p->bad_object_sha1 + GIT_SHA1_RAWSZ * i))
994 return;
995 p->bad_object_sha1 = xrealloc(p->bad_object_sha1,
996 st_mult(GIT_MAX_RAWSZ,
997 st_add(p->num_bad_objects, 1)));
998 hashcpy(p->bad_object_sha1 + GIT_SHA1_RAWSZ * p->num_bad_objects, sha1);
999 p->num_bad_objects++;
1002 const struct packed_git *has_packed_and_bad(const unsigned char *sha1)
1004 struct packed_git *p;
1005 unsigned i;
1007 for (p = packed_git; p; p = p->next)
1008 for (i = 0; i < p->num_bad_objects; i++)
1009 if (!hashcmp(sha1, p->bad_object_sha1 + 20 * i))
1010 return p;
1011 return NULL;
1014 static off_t get_delta_base(struct packed_git *p,
1015 struct pack_window **w_curs,
1016 off_t *curpos,
1017 enum object_type type,
1018 off_t delta_obj_offset)
1020 unsigned char *base_info = use_pack(p, w_curs, *curpos, NULL);
1021 off_t base_offset;
1023 /* use_pack() assured us we have [base_info, base_info + 20)
1024 * as a range that we can look at without walking off the
1025 * end of the mapped window. Its actually the hash size
1026 * that is assured. An OFS_DELTA longer than the hash size
1027 * is stupid, as then a REF_DELTA would be smaller to store.
1029 if (type == OBJ_OFS_DELTA) {
1030 unsigned used = 0;
1031 unsigned char c = base_info[used++];
1032 base_offset = c & 127;
1033 while (c & 128) {
1034 base_offset += 1;
1035 if (!base_offset || MSB(base_offset, 7))
1036 return 0; /* overflow */
1037 c = base_info[used++];
1038 base_offset = (base_offset << 7) + (c & 127);
1040 base_offset = delta_obj_offset - base_offset;
1041 if (base_offset <= 0 || base_offset >= delta_obj_offset)
1042 return 0; /* out of bound */
1043 *curpos += used;
1044 } else if (type == OBJ_REF_DELTA) {
1045 /* The base entry _must_ be in the same pack */
1046 base_offset = find_pack_entry_one(base_info, p);
1047 *curpos += 20;
1048 } else
1049 die("I am totally screwed");
1050 return base_offset;
1054 * Like get_delta_base above, but we return the sha1 instead of the pack
1055 * offset. This means it is cheaper for REF deltas (we do not have to do
1056 * the final object lookup), but more expensive for OFS deltas (we
1057 * have to load the revidx to convert the offset back into a sha1).
1059 static const unsigned char *get_delta_base_sha1(struct packed_git *p,
1060 struct pack_window **w_curs,
1061 off_t curpos,
1062 enum object_type type,
1063 off_t delta_obj_offset)
1065 if (type == OBJ_REF_DELTA) {
1066 unsigned char *base = use_pack(p, w_curs, curpos, NULL);
1067 return base;
1068 } else if (type == OBJ_OFS_DELTA) {
1069 struct revindex_entry *revidx;
1070 off_t base_offset = get_delta_base(p, w_curs, &curpos,
1071 type, delta_obj_offset);
1073 if (!base_offset)
1074 return NULL;
1076 revidx = find_pack_revindex(p, base_offset);
1077 if (!revidx)
1078 return NULL;
1080 return nth_packed_object_sha1(p, revidx->nr);
1081 } else
1082 return NULL;
1085 static int retry_bad_packed_offset(struct packed_git *p, off_t obj_offset)
1087 int type;
1088 struct revindex_entry *revidx;
1089 const unsigned char *sha1;
1090 revidx = find_pack_revindex(p, obj_offset);
1091 if (!revidx)
1092 return OBJ_BAD;
1093 sha1 = nth_packed_object_sha1(p, revidx->nr);
1094 mark_bad_packed_object(p, sha1);
1095 type = sha1_object_info(sha1, NULL);
1096 if (type <= OBJ_NONE)
1097 return OBJ_BAD;
1098 return type;
1101 #define POI_STACK_PREALLOC 64
1103 static enum object_type packed_to_object_type(struct packed_git *p,
1104 off_t obj_offset,
1105 enum object_type type,
1106 struct pack_window **w_curs,
1107 off_t curpos)
1109 off_t small_poi_stack[POI_STACK_PREALLOC];
1110 off_t *poi_stack = small_poi_stack;
1111 int poi_stack_nr = 0, poi_stack_alloc = POI_STACK_PREALLOC;
1113 while (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
1114 off_t base_offset;
1115 unsigned long size;
1116 /* Push the object we're going to leave behind */
1117 if (poi_stack_nr >= poi_stack_alloc && poi_stack == small_poi_stack) {
1118 poi_stack_alloc = alloc_nr(poi_stack_nr);
1119 ALLOC_ARRAY(poi_stack, poi_stack_alloc);
1120 memcpy(poi_stack, small_poi_stack, sizeof(off_t)*poi_stack_nr);
1121 } else {
1122 ALLOC_GROW(poi_stack, poi_stack_nr+1, poi_stack_alloc);
1124 poi_stack[poi_stack_nr++] = obj_offset;
1125 /* If parsing the base offset fails, just unwind */
1126 base_offset = get_delta_base(p, w_curs, &curpos, type, obj_offset);
1127 if (!base_offset)
1128 goto unwind;
1129 curpos = obj_offset = base_offset;
1130 type = unpack_object_header(p, w_curs, &curpos, &size);
1131 if (type <= OBJ_NONE) {
1132 /* If getting the base itself fails, we first
1133 * retry the base, otherwise unwind */
1134 type = retry_bad_packed_offset(p, base_offset);
1135 if (type > OBJ_NONE)
1136 goto out;
1137 goto unwind;
1141 switch (type) {
1142 case OBJ_BAD:
1143 case OBJ_COMMIT:
1144 case OBJ_TREE:
1145 case OBJ_BLOB:
1146 case OBJ_TAG:
1147 break;
1148 default:
1149 error("unknown object type %i at offset %"PRIuMAX" in %s",
1150 type, (uintmax_t)obj_offset, p->pack_name);
1151 type = OBJ_BAD;
1154 out:
1155 if (poi_stack != small_poi_stack)
1156 free(poi_stack);
1157 return type;
1159 unwind:
1160 while (poi_stack_nr) {
1161 obj_offset = poi_stack[--poi_stack_nr];
1162 type = retry_bad_packed_offset(p, obj_offset);
1163 if (type > OBJ_NONE)
1164 goto out;
1166 type = OBJ_BAD;
1167 goto out;
1170 static struct hashmap delta_base_cache;
1171 static size_t delta_base_cached;
1173 static LIST_HEAD(delta_base_cache_lru);
1175 struct delta_base_cache_key {
1176 struct packed_git *p;
1177 off_t base_offset;
1180 struct delta_base_cache_entry {
1181 struct hashmap hash;
1182 struct delta_base_cache_key key;
1183 struct list_head lru;
1184 void *data;
1185 unsigned long size;
1186 enum object_type type;
1189 static unsigned int pack_entry_hash(struct packed_git *p, off_t base_offset)
1191 unsigned int hash;
1193 hash = (unsigned int)(intptr_t)p + (unsigned int)base_offset;
1194 hash += (hash >> 8) + (hash >> 16);
1195 return hash;
1198 static struct delta_base_cache_entry *
1199 get_delta_base_cache_entry(struct packed_git *p, off_t base_offset)
1201 struct hashmap_entry entry;
1202 struct delta_base_cache_key key;
1204 if (!delta_base_cache.cmpfn)
1205 return NULL;
1207 hashmap_entry_init(&entry, pack_entry_hash(p, base_offset));
1208 key.p = p;
1209 key.base_offset = base_offset;
1210 return hashmap_get(&delta_base_cache, &entry, &key);
1213 static int delta_base_cache_key_eq(const struct delta_base_cache_key *a,
1214 const struct delta_base_cache_key *b)
1216 return a->p == b->p && a->base_offset == b->base_offset;
1219 static int delta_base_cache_hash_cmp(const void *unused_cmp_data,
1220 const void *va, const void *vb,
1221 const void *vkey)
1223 const struct delta_base_cache_entry *a = va, *b = vb;
1224 const struct delta_base_cache_key *key = vkey;
1225 if (key)
1226 return !delta_base_cache_key_eq(&a->key, key);
1227 else
1228 return !delta_base_cache_key_eq(&a->key, &b->key);
1231 static int in_delta_base_cache(struct packed_git *p, off_t base_offset)
1233 return !!get_delta_base_cache_entry(p, base_offset);
1237 * Remove the entry from the cache, but do _not_ free the associated
1238 * entry data. The caller takes ownership of the "data" buffer, and
1239 * should copy out any fields it wants before detaching.
1241 static void detach_delta_base_cache_entry(struct delta_base_cache_entry *ent)
1243 hashmap_remove(&delta_base_cache, ent, &ent->key);
1244 list_del(&ent->lru);
1245 delta_base_cached -= ent->size;
1246 free(ent);
1249 static void *cache_or_unpack_entry(struct packed_git *p, off_t base_offset,
1250 unsigned long *base_size, enum object_type *type)
1252 struct delta_base_cache_entry *ent;
1254 ent = get_delta_base_cache_entry(p, base_offset);
1255 if (!ent)
1256 return unpack_entry(p, base_offset, type, base_size);
1258 if (type)
1259 *type = ent->type;
1260 if (base_size)
1261 *base_size = ent->size;
1262 return xmemdupz(ent->data, ent->size);
1265 static inline void release_delta_base_cache(struct delta_base_cache_entry *ent)
1267 free(ent->data);
1268 detach_delta_base_cache_entry(ent);
1271 void clear_delta_base_cache(void)
1273 struct list_head *lru, *tmp;
1274 list_for_each_safe(lru, tmp, &delta_base_cache_lru) {
1275 struct delta_base_cache_entry *entry =
1276 list_entry(lru, struct delta_base_cache_entry, lru);
1277 release_delta_base_cache(entry);
1281 static void add_delta_base_cache(struct packed_git *p, off_t base_offset,
1282 void *base, unsigned long base_size, enum object_type type)
1284 struct delta_base_cache_entry *ent = xmalloc(sizeof(*ent));
1285 struct list_head *lru, *tmp;
1287 delta_base_cached += base_size;
1289 list_for_each_safe(lru, tmp, &delta_base_cache_lru) {
1290 struct delta_base_cache_entry *f =
1291 list_entry(lru, struct delta_base_cache_entry, lru);
1292 if (delta_base_cached <= delta_base_cache_limit)
1293 break;
1294 release_delta_base_cache(f);
1297 ent->key.p = p;
1298 ent->key.base_offset = base_offset;
1299 ent->type = type;
1300 ent->data = base;
1301 ent->size = base_size;
1302 list_add_tail(&ent->lru, &delta_base_cache_lru);
1304 if (!delta_base_cache.cmpfn)
1305 hashmap_init(&delta_base_cache, delta_base_cache_hash_cmp, NULL, 0);
1306 hashmap_entry_init(ent, pack_entry_hash(p, base_offset));
1307 hashmap_add(&delta_base_cache, ent);
1310 int packed_object_info(struct packed_git *p, off_t obj_offset,
1311 struct object_info *oi)
1313 struct pack_window *w_curs = NULL;
1314 unsigned long size;
1315 off_t curpos = obj_offset;
1316 enum object_type type;
1319 * We always get the representation type, but only convert it to
1320 * a "real" type later if the caller is interested.
1322 if (oi->contentp) {
1323 *oi->contentp = cache_or_unpack_entry(p, obj_offset, oi->sizep,
1324 &type);
1325 if (!*oi->contentp)
1326 type = OBJ_BAD;
1327 } else {
1328 type = unpack_object_header(p, &w_curs, &curpos, &size);
1331 if (!oi->contentp && oi->sizep) {
1332 if (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
1333 off_t tmp_pos = curpos;
1334 off_t base_offset = get_delta_base(p, &w_curs, &tmp_pos,
1335 type, obj_offset);
1336 if (!base_offset) {
1337 type = OBJ_BAD;
1338 goto out;
1340 *oi->sizep = get_size_from_delta(p, &w_curs, tmp_pos);
1341 if (*oi->sizep == 0) {
1342 type = OBJ_BAD;
1343 goto out;
1345 } else {
1346 *oi->sizep = size;
1350 if (oi->disk_sizep) {
1351 struct revindex_entry *revidx = find_pack_revindex(p, obj_offset);
1352 *oi->disk_sizep = revidx[1].offset - obj_offset;
1355 if (oi->typep || oi->typename) {
1356 enum object_type ptot;
1357 ptot = packed_to_object_type(p, obj_offset, type, &w_curs,
1358 curpos);
1359 if (oi->typep)
1360 *oi->typep = ptot;
1361 if (oi->typename) {
1362 const char *tn = typename(ptot);
1363 if (tn)
1364 strbuf_addstr(oi->typename, tn);
1366 if (ptot < 0) {
1367 type = OBJ_BAD;
1368 goto out;
1372 if (oi->delta_base_sha1) {
1373 if (type == OBJ_OFS_DELTA || type == OBJ_REF_DELTA) {
1374 const unsigned char *base;
1376 base = get_delta_base_sha1(p, &w_curs, curpos,
1377 type, obj_offset);
1378 if (!base) {
1379 type = OBJ_BAD;
1380 goto out;
1383 hashcpy(oi->delta_base_sha1, base);
1384 } else
1385 hashclr(oi->delta_base_sha1);
1388 oi->whence = in_delta_base_cache(p, obj_offset) ? OI_DBCACHED :
1389 OI_PACKED;
1391 out:
1392 unuse_pack(&w_curs);
1393 return type;
1396 static void *unpack_compressed_entry(struct packed_git *p,
1397 struct pack_window **w_curs,
1398 off_t curpos,
1399 unsigned long size)
1401 int st;
1402 git_zstream stream;
1403 unsigned char *buffer, *in;
1405 buffer = xmallocz_gently(size);
1406 if (!buffer)
1407 return NULL;
1408 memset(&stream, 0, sizeof(stream));
1409 stream.next_out = buffer;
1410 stream.avail_out = size + 1;
1412 git_inflate_init(&stream);
1413 do {
1414 in = use_pack(p, w_curs, curpos, &stream.avail_in);
1415 stream.next_in = in;
1416 st = git_inflate(&stream, Z_FINISH);
1417 if (!stream.avail_out)
1418 break; /* the payload is larger than it should be */
1419 curpos += stream.next_in - in;
1420 } while (st == Z_OK || st == Z_BUF_ERROR);
1421 git_inflate_end(&stream);
1422 if ((st != Z_STREAM_END) || stream.total_out != size) {
1423 free(buffer);
1424 return NULL;
1427 return buffer;
1430 static void write_pack_access_log(struct packed_git *p, off_t obj_offset)
1432 static struct trace_key pack_access = TRACE_KEY_INIT(PACK_ACCESS);
1433 trace_printf_key(&pack_access, "%s %"PRIuMAX"\n",
1434 p->pack_name, (uintmax_t)obj_offset);
1437 int do_check_packed_object_crc;
1439 #define UNPACK_ENTRY_STACK_PREALLOC 64
1440 struct unpack_entry_stack_ent {
1441 off_t obj_offset;
1442 off_t curpos;
1443 unsigned long size;
1446 static void *read_object(const unsigned char *sha1, enum object_type *type,
1447 unsigned long *size)
1449 struct object_info oi = OBJECT_INFO_INIT;
1450 void *content;
1451 oi.typep = type;
1452 oi.sizep = size;
1453 oi.contentp = &content;
1455 if (sha1_object_info_extended(sha1, &oi, 0) < 0)
1456 return NULL;
1457 return content;
1460 void *unpack_entry(struct packed_git *p, off_t obj_offset,
1461 enum object_type *final_type, unsigned long *final_size)
1463 struct pack_window *w_curs = NULL;
1464 off_t curpos = obj_offset;
1465 void *data = NULL;
1466 unsigned long size;
1467 enum object_type type;
1468 struct unpack_entry_stack_ent small_delta_stack[UNPACK_ENTRY_STACK_PREALLOC];
1469 struct unpack_entry_stack_ent *delta_stack = small_delta_stack;
1470 int delta_stack_nr = 0, delta_stack_alloc = UNPACK_ENTRY_STACK_PREALLOC;
1471 int base_from_cache = 0;
1473 write_pack_access_log(p, obj_offset);
1475 /* PHASE 1: drill down to the innermost base object */
1476 for (;;) {
1477 off_t base_offset;
1478 int i;
1479 struct delta_base_cache_entry *ent;
1481 ent = get_delta_base_cache_entry(p, curpos);
1482 if (ent) {
1483 type = ent->type;
1484 data = ent->data;
1485 size = ent->size;
1486 detach_delta_base_cache_entry(ent);
1487 base_from_cache = 1;
1488 break;
1491 if (do_check_packed_object_crc && p->index_version > 1) {
1492 struct revindex_entry *revidx = find_pack_revindex(p, obj_offset);
1493 off_t len = revidx[1].offset - obj_offset;
1494 if (check_pack_crc(p, &w_curs, obj_offset, len, revidx->nr)) {
1495 const unsigned char *sha1 =
1496 nth_packed_object_sha1(p, revidx->nr);
1497 error("bad packed object CRC for %s",
1498 sha1_to_hex(sha1));
1499 mark_bad_packed_object(p, sha1);
1500 data = NULL;
1501 goto out;
1505 type = unpack_object_header(p, &w_curs, &curpos, &size);
1506 if (type != OBJ_OFS_DELTA && type != OBJ_REF_DELTA)
1507 break;
1509 base_offset = get_delta_base(p, &w_curs, &curpos, type, obj_offset);
1510 if (!base_offset) {
1511 error("failed to validate delta base reference "
1512 "at offset %"PRIuMAX" from %s",
1513 (uintmax_t)curpos, p->pack_name);
1514 /* bail to phase 2, in hopes of recovery */
1515 data = NULL;
1516 break;
1519 /* push object, proceed to base */
1520 if (delta_stack_nr >= delta_stack_alloc
1521 && delta_stack == small_delta_stack) {
1522 delta_stack_alloc = alloc_nr(delta_stack_nr);
1523 ALLOC_ARRAY(delta_stack, delta_stack_alloc);
1524 memcpy(delta_stack, small_delta_stack,
1525 sizeof(*delta_stack)*delta_stack_nr);
1526 } else {
1527 ALLOC_GROW(delta_stack, delta_stack_nr+1, delta_stack_alloc);
1529 i = delta_stack_nr++;
1530 delta_stack[i].obj_offset = obj_offset;
1531 delta_stack[i].curpos = curpos;
1532 delta_stack[i].size = size;
1534 curpos = obj_offset = base_offset;
1537 /* PHASE 2: handle the base */
1538 switch (type) {
1539 case OBJ_OFS_DELTA:
1540 case OBJ_REF_DELTA:
1541 if (data)
1542 die("BUG: unpack_entry: left loop at a valid delta");
1543 break;
1544 case OBJ_COMMIT:
1545 case OBJ_TREE:
1546 case OBJ_BLOB:
1547 case OBJ_TAG:
1548 if (!base_from_cache)
1549 data = unpack_compressed_entry(p, &w_curs, curpos, size);
1550 break;
1551 default:
1552 data = NULL;
1553 error("unknown object type %i at offset %"PRIuMAX" in %s",
1554 type, (uintmax_t)obj_offset, p->pack_name);
1557 /* PHASE 3: apply deltas in order */
1559 /* invariants:
1560 * 'data' holds the base data, or NULL if there was corruption
1562 while (delta_stack_nr) {
1563 void *delta_data;
1564 void *base = data;
1565 void *external_base = NULL;
1566 unsigned long delta_size, base_size = size;
1567 int i;
1569 data = NULL;
1571 if (base)
1572 add_delta_base_cache(p, obj_offset, base, base_size, type);
1574 if (!base) {
1576 * We're probably in deep shit, but let's try to fetch
1577 * the required base anyway from another pack or loose.
1578 * This is costly but should happen only in the presence
1579 * of a corrupted pack, and is better than failing outright.
1581 struct revindex_entry *revidx;
1582 const unsigned char *base_sha1;
1583 revidx = find_pack_revindex(p, obj_offset);
1584 if (revidx) {
1585 base_sha1 = nth_packed_object_sha1(p, revidx->nr);
1586 error("failed to read delta base object %s"
1587 " at offset %"PRIuMAX" from %s",
1588 sha1_to_hex(base_sha1), (uintmax_t)obj_offset,
1589 p->pack_name);
1590 mark_bad_packed_object(p, base_sha1);
1591 base = read_object(base_sha1, &type, &base_size);
1592 external_base = base;
1596 i = --delta_stack_nr;
1597 obj_offset = delta_stack[i].obj_offset;
1598 curpos = delta_stack[i].curpos;
1599 delta_size = delta_stack[i].size;
1601 if (!base)
1602 continue;
1604 delta_data = unpack_compressed_entry(p, &w_curs, curpos, delta_size);
1606 if (!delta_data) {
1607 error("failed to unpack compressed delta "
1608 "at offset %"PRIuMAX" from %s",
1609 (uintmax_t)curpos, p->pack_name);
1610 data = NULL;
1611 free(external_base);
1612 continue;
1615 data = patch_delta(base, base_size,
1616 delta_data, delta_size,
1617 &size);
1620 * We could not apply the delta; warn the user, but keep going.
1621 * Our failure will be noticed either in the next iteration of
1622 * the loop, or if this is the final delta, in the caller when
1623 * we return NULL. Those code paths will take care of making
1624 * a more explicit warning and retrying with another copy of
1625 * the object.
1627 if (!data)
1628 error("failed to apply delta");
1630 free(delta_data);
1631 free(external_base);
1634 if (final_type)
1635 *final_type = type;
1636 if (final_size)
1637 *final_size = size;
1639 out:
1640 unuse_pack(&w_curs);
1642 if (delta_stack != small_delta_stack)
1643 free(delta_stack);
1645 return data;
1648 const unsigned char *nth_packed_object_sha1(struct packed_git *p,
1649 uint32_t n)
1651 const unsigned char *index = p->index_data;
1652 if (!index) {
1653 if (open_pack_index(p))
1654 return NULL;
1655 index = p->index_data;
1657 if (n >= p->num_objects)
1658 return NULL;
1659 index += 4 * 256;
1660 if (p->index_version == 1) {
1661 return index + 24 * n + 4;
1662 } else {
1663 index += 8;
1664 return index + 20 * n;
1668 const struct object_id *nth_packed_object_oid(struct object_id *oid,
1669 struct packed_git *p,
1670 uint32_t n)
1672 const unsigned char *hash = nth_packed_object_sha1(p, n);
1673 if (!hash)
1674 return NULL;
1675 hashcpy(oid->hash, hash);
1676 return oid;
1679 void check_pack_index_ptr(const struct packed_git *p, const void *vptr)
1681 const unsigned char *ptr = vptr;
1682 const unsigned char *start = p->index_data;
1683 const unsigned char *end = start + p->index_size;
1684 if (ptr < start)
1685 die(_("offset before start of pack index for %s (corrupt index?)"),
1686 p->pack_name);
1687 /* No need to check for underflow; .idx files must be at least 8 bytes */
1688 if (ptr >= end - 8)
1689 die(_("offset beyond end of pack index for %s (truncated index?)"),
1690 p->pack_name);
1693 off_t nth_packed_object_offset(const struct packed_git *p, uint32_t n)
1695 const unsigned char *index = p->index_data;
1696 index += 4 * 256;
1697 if (p->index_version == 1) {
1698 return ntohl(*((uint32_t *)(index + 24 * n)));
1699 } else {
1700 uint32_t off;
1701 index += 8 + p->num_objects * (20 + 4);
1702 off = ntohl(*((uint32_t *)(index + 4 * n)));
1703 if (!(off & 0x80000000))
1704 return off;
1705 index += p->num_objects * 4 + (off & 0x7fffffff) * 8;
1706 check_pack_index_ptr(p, index);
1707 return (((uint64_t)ntohl(*((uint32_t *)(index + 0)))) << 32) |
1708 ntohl(*((uint32_t *)(index + 4)));
1712 off_t find_pack_entry_one(const unsigned char *sha1,
1713 struct packed_git *p)
1715 const uint32_t *level1_ofs = p->index_data;
1716 const unsigned char *index = p->index_data;
1717 unsigned hi, lo, stride;
1718 static int debug_lookup = -1;
1720 if (debug_lookup < 0)
1721 debug_lookup = !!getenv("GIT_DEBUG_LOOKUP");
1723 if (!index) {
1724 if (open_pack_index(p))
1725 return 0;
1726 level1_ofs = p->index_data;
1727 index = p->index_data;
1729 if (p->index_version > 1) {
1730 level1_ofs += 2;
1731 index += 8;
1733 index += 4 * 256;
1734 hi = ntohl(level1_ofs[*sha1]);
1735 lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
1736 if (p->index_version > 1) {
1737 stride = 20;
1738 } else {
1739 stride = 24;
1740 index += 4;
1743 if (debug_lookup)
1744 printf("%02x%02x%02x... lo %u hi %u nr %"PRIu32"\n",
1745 sha1[0], sha1[1], sha1[2], lo, hi, p->num_objects);
1747 while (lo < hi) {
1748 unsigned mi = (lo + hi) / 2;
1749 int cmp = hashcmp(index + mi * stride, sha1);
1751 if (debug_lookup)
1752 printf("lo %u hi %u rg %u mi %u\n",
1753 lo, hi, hi - lo, mi);
1754 if (!cmp)
1755 return nth_packed_object_offset(p, mi);
1756 if (cmp > 0)
1757 hi = mi;
1758 else
1759 lo = mi+1;
1761 return 0;
1764 int is_pack_valid(struct packed_git *p)
1766 /* An already open pack is known to be valid. */
1767 if (p->pack_fd != -1)
1768 return 1;
1770 /* If the pack has one window completely covering the
1771 * file size, the pack is known to be valid even if
1772 * the descriptor is not currently open.
1774 if (p->windows) {
1775 struct pack_window *w = p->windows;
1777 if (!w->offset && w->len == p->pack_size)
1778 return 1;
1781 /* Force the pack to open to prove its valid. */
1782 return !open_packed_git(p);
1785 struct packed_git *find_sha1_pack(const unsigned char *sha1,
1786 struct packed_git *packs)
1788 struct packed_git *p;
1790 for (p = packs; p; p = p->next) {
1791 if (find_pack_entry_one(sha1, p))
1792 return p;
1794 return NULL;
1798 static int fill_pack_entry(const unsigned char *sha1,
1799 struct pack_entry *e,
1800 struct packed_git *p)
1802 off_t offset;
1804 if (p->num_bad_objects) {
1805 unsigned i;
1806 for (i = 0; i < p->num_bad_objects; i++)
1807 if (!hashcmp(sha1, p->bad_object_sha1 + 20 * i))
1808 return 0;
1811 offset = find_pack_entry_one(sha1, p);
1812 if (!offset)
1813 return 0;
1816 * We are about to tell the caller where they can locate the
1817 * requested object. We better make sure the packfile is
1818 * still here and can be accessed before supplying that
1819 * answer, as it may have been deleted since the index was
1820 * loaded!
1822 if (!is_pack_valid(p))
1823 return 0;
1824 e->offset = offset;
1825 e->p = p;
1826 hashcpy(e->sha1, sha1);
1827 return 1;
1831 * Iff a pack file contains the object named by sha1, return true and
1832 * store its location to e.
1834 int find_pack_entry(const unsigned char *sha1, struct pack_entry *e)
1836 struct mru_entry *p;
1838 prepare_packed_git();
1839 if (!packed_git)
1840 return 0;
1842 for (p = packed_git_mru->head; p; p = p->next) {
1843 if (fill_pack_entry(sha1, e, p->item)) {
1844 mru_mark(packed_git_mru, p);
1845 return 1;
1848 return 0;
1851 int has_sha1_pack(const unsigned char *sha1)
1853 struct pack_entry e;
1854 return find_pack_entry(sha1, &e);
1857 int has_pack_index(const unsigned char *sha1)
1859 struct stat st;
1860 if (stat(sha1_pack_index_name(sha1), &st))
1861 return 0;
1862 return 1;
1865 static int for_each_object_in_pack(struct packed_git *p, each_packed_object_fn cb, void *data)
1867 uint32_t i;
1868 int r = 0;
1870 for (i = 0; i < p->num_objects; i++) {
1871 struct object_id oid;
1873 if (!nth_packed_object_oid(&oid, p, i))
1874 return error("unable to get sha1 of object %u in %s",
1875 i, p->pack_name);
1877 r = cb(&oid, p, i, data);
1878 if (r)
1879 break;
1881 return r;
1884 int for_each_packed_object(each_packed_object_fn cb, void *data, unsigned flags)
1886 struct packed_git *p;
1887 int r = 0;
1888 int pack_errors = 0;
1890 prepare_packed_git();
1891 for (p = packed_git; p; p = p->next) {
1892 if ((flags & FOR_EACH_OBJECT_LOCAL_ONLY) && !p->pack_local)
1893 continue;
1894 if (open_pack_index(p)) {
1895 pack_errors = 1;
1896 continue;
1898 r = for_each_object_in_pack(p, cb, data);
1899 if (r)
1900 break;
1902 return r ? r : pack_errors;