4 #include "repository.h"
10 #include "streaming.h"
11 #include "sha1-lookup.h"
15 #include "tree-walk.h"
17 #include "object-store.h"
19 char *odb_pack_name(struct strbuf
*buf
,
20 const unsigned char *sha1
,
24 strbuf_addf(buf
, "%s/pack/pack-%s.%s", get_object_directory(),
25 sha1_to_hex(sha1
), ext
);
29 char *sha1_pack_name(const unsigned char *sha1
)
31 static struct strbuf buf
= STRBUF_INIT
;
32 return odb_pack_name(&buf
, sha1
, "pack");
35 char *sha1_pack_index_name(const unsigned char *sha1
)
37 static struct strbuf buf
= STRBUF_INIT
;
38 return odb_pack_name(&buf
, sha1
, "idx");
41 static unsigned int pack_used_ctr
;
42 static unsigned int pack_mmap_calls
;
43 static unsigned int peak_pack_open_windows
;
44 static unsigned int pack_open_windows
;
45 static unsigned int pack_open_fds
;
46 static unsigned int pack_max_fds
;
47 static size_t peak_pack_mapped
;
48 static size_t pack_mapped
;
50 #define SZ_FMT PRIuMAX
51 static inline uintmax_t sz_fmt(size_t s
) { return s
; }
53 void pack_report(void)
56 "pack_report: getpagesize() = %10" SZ_FMT
"\n"
57 "pack_report: core.packedGitWindowSize = %10" SZ_FMT
"\n"
58 "pack_report: core.packedGitLimit = %10" SZ_FMT
"\n",
59 sz_fmt(getpagesize()),
60 sz_fmt(packed_git_window_size
),
61 sz_fmt(packed_git_limit
));
63 "pack_report: pack_used_ctr = %10u\n"
64 "pack_report: pack_mmap_calls = %10u\n"
65 "pack_report: pack_open_windows = %10u / %10u\n"
66 "pack_report: pack_mapped = "
67 "%10" SZ_FMT
" / %10" SZ_FMT
"\n",
70 pack_open_windows
, peak_pack_open_windows
,
71 sz_fmt(pack_mapped
), sz_fmt(peak_pack_mapped
));
75 * Open and mmap the index file at path, perform a couple of
76 * consistency checks, then record its information to p. Return 0 on
79 static int check_packed_git_idx(const char *path
, struct packed_git
*p
)
82 struct pack_idx_header
*hdr
;
84 uint32_t version
, nr
, i
, *index
;
85 int fd
= git_open(path
);
87 const unsigned int hashsz
= the_hash_algo
->rawsz
;
95 idx_size
= xsize_t(st
.st_size
);
96 if (idx_size
< 4 * 256 + hashsz
+ hashsz
) {
98 return error("index file %s is too small", path
);
100 idx_map
= xmmap(NULL
, idx_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
104 if (hdr
->idx_signature
== htonl(PACK_IDX_SIGNATURE
)) {
105 version
= ntohl(hdr
->idx_version
);
106 if (version
< 2 || version
> 2) {
107 munmap(idx_map
, idx_size
);
108 return error("index file %s is version %"PRIu32
109 " and is not supported by this binary"
110 " (try upgrading GIT to a newer version)",
119 index
+= 2; /* skip index header */
120 for (i
= 0; i
< 256; i
++) {
121 uint32_t n
= ntohl(index
[i
]);
123 munmap(idx_map
, idx_size
);
124 return error("non-monotonic index %s", path
);
132 * - 256 index entries 4 bytes each
133 * - 24-byte entries * nr (object ID + 4-byte offset)
134 * - hash of the packfile
137 if (idx_size
!= 4*256 + nr
* (hashsz
+ 4) + hashsz
+ hashsz
) {
138 munmap(idx_map
, idx_size
);
139 return error("wrong index v1 file size in %s", path
);
141 } else if (version
== 2) {
144 * - 8 bytes of header
145 * - 256 index entries 4 bytes each
146 * - object ID entry * nr
147 * - 4-byte crc entry * nr
148 * - 4-byte offset entry * nr
149 * - hash of the packfile
151 * And after the 4-byte offset table might be a
152 * variable sized table containing 8-byte entries
153 * for offsets larger than 2^31.
155 unsigned long min_size
= 8 + 4*256 + nr
*(hashsz
+ 4 + 4) + hashsz
+ hashsz
;
156 unsigned long max_size
= min_size
;
158 max_size
+= (nr
- 1)*8;
159 if (idx_size
< min_size
|| idx_size
> max_size
) {
160 munmap(idx_map
, idx_size
);
161 return error("wrong index v2 file size in %s", path
);
163 if (idx_size
!= min_size
&&
165 * make sure we can deal with large pack offsets.
166 * 31-bit signed offset won't be enough, neither
167 * 32-bit unsigned one will be.
169 (sizeof(off_t
) <= 4)) {
170 munmap(idx_map
, idx_size
);
171 return error("pack too large for current definition of off_t in %s", path
);
175 p
->index_version
= version
;
176 p
->index_data
= idx_map
;
177 p
->index_size
= idx_size
;
182 int open_pack_index(struct packed_git
*p
)
191 if (!strip_suffix(p
->pack_name
, ".pack", &len
))
192 BUG("pack_name does not end in .pack");
193 idx_name
= xstrfmt("%.*s.idx", (int)len
, p
->pack_name
);
194 ret
= check_packed_git_idx(idx_name
, p
);
199 uint32_t get_pack_fanout(struct packed_git
*p
, uint32_t value
)
201 const uint32_t *level1_ofs
= p
->index_data
;
204 if (open_pack_index(p
))
206 level1_ofs
= p
->index_data
;
209 if (p
->index_version
> 1) {
213 return ntohl(level1_ofs
[value
]);
216 static struct packed_git
*alloc_packed_git(int extra
)
218 struct packed_git
*p
= xmalloc(st_add(sizeof(*p
), extra
));
219 memset(p
, 0, sizeof(*p
));
224 struct packed_git
*parse_pack_index(unsigned char *sha1
, const char *idx_path
)
226 const char *path
= sha1_pack_name(sha1
);
227 size_t alloc
= st_add(strlen(path
), 1);
228 struct packed_git
*p
= alloc_packed_git(alloc
);
230 memcpy(p
->pack_name
, path
, alloc
); /* includes NUL */
231 hashcpy(p
->sha1
, sha1
);
232 if (check_packed_git_idx(idx_path
, p
)) {
240 static void scan_windows(struct packed_git
*p
,
241 struct packed_git
**lru_p
,
242 struct pack_window
**lru_w
,
243 struct pack_window
**lru_l
)
245 struct pack_window
*w
, *w_l
;
247 for (w_l
= NULL
, w
= p
->windows
; w
; w
= w
->next
) {
249 if (!*lru_w
|| w
->last_used
< (*lru_w
)->last_used
) {
259 static int unuse_one_window(struct packed_git
*current
)
261 struct packed_git
*p
, *lru_p
= NULL
;
262 struct pack_window
*lru_w
= NULL
, *lru_l
= NULL
;
265 scan_windows(current
, &lru_p
, &lru_w
, &lru_l
);
266 for (p
= the_repository
->objects
->packed_git
; p
; p
= p
->next
)
267 scan_windows(p
, &lru_p
, &lru_w
, &lru_l
);
269 munmap(lru_w
->base
, lru_w
->len
);
270 pack_mapped
-= lru_w
->len
;
272 lru_l
->next
= lru_w
->next
;
274 lru_p
->windows
= lru_w
->next
;
282 void release_pack_memory(size_t need
)
284 size_t cur
= pack_mapped
;
285 while (need
>= (cur
- pack_mapped
) && unuse_one_window(NULL
))
289 void close_pack_windows(struct packed_git
*p
)
292 struct pack_window
*w
= p
->windows
;
295 die("pack '%s' still has open windows to it",
297 munmap(w
->base
, w
->len
);
298 pack_mapped
-= w
->len
;
300 p
->windows
= w
->next
;
305 static int close_pack_fd(struct packed_git
*p
)
317 void close_pack_index(struct packed_git
*p
)
320 munmap((void *)p
->index_data
, p
->index_size
);
321 p
->index_data
= NULL
;
325 void close_pack(struct packed_git
*p
)
327 close_pack_windows(p
);
332 void close_all_packs(struct raw_object_store
*o
)
334 struct packed_git
*p
;
336 for (p
= o
->packed_git
; p
; p
= p
->next
)
338 BUG("want to close pack marked 'do-not-close'");
344 * The LRU pack is the one with the oldest MRU window, preferring packs
345 * with no used windows, or the oldest mtime if it has no windows allocated.
347 static void find_lru_pack(struct packed_git
*p
, struct packed_git
**lru_p
, struct pack_window
**mru_w
, int *accept_windows_inuse
)
349 struct pack_window
*w
, *this_mru_w
;
350 int has_windows_inuse
= 0;
353 * Reject this pack if it has windows and the previously selected
354 * one does not. If this pack does not have windows, reject
355 * it if the pack file is newer than the previously selected one.
357 if (*lru_p
&& !*mru_w
&& (p
->windows
|| p
->mtime
> (*lru_p
)->mtime
))
360 for (w
= this_mru_w
= p
->windows
; w
; w
= w
->next
) {
362 * Reject this pack if any of its windows are in use,
363 * but the previously selected pack did not have any
364 * inuse windows. Otherwise, record that this pack
365 * has windows in use.
368 if (*accept_windows_inuse
)
369 has_windows_inuse
= 1;
374 if (w
->last_used
> this_mru_w
->last_used
)
378 * Reject this pack if it has windows that have been
379 * used more recently than the previously selected pack.
380 * If the previously selected pack had windows inuse and
381 * we have not encountered a window in this pack that is
382 * inuse, skip this check since we prefer a pack with no
383 * inuse windows to one that has inuse windows.
385 if (*mru_w
&& *accept_windows_inuse
== has_windows_inuse
&&
386 this_mru_w
->last_used
> (*mru_w
)->last_used
)
395 *accept_windows_inuse
= has_windows_inuse
;
398 static int close_one_pack(void)
400 struct packed_git
*p
, *lru_p
= NULL
;
401 struct pack_window
*mru_w
= NULL
;
402 int accept_windows_inuse
= 1;
404 for (p
= the_repository
->objects
->packed_git
; p
; p
= p
->next
) {
405 if (p
->pack_fd
== -1)
407 find_lru_pack(p
, &lru_p
, &mru_w
, &accept_windows_inuse
);
411 return close_pack_fd(lru_p
);
416 static unsigned int get_max_fd_limit(void)
422 if (!getrlimit(RLIMIT_NOFILE
, &lim
))
429 long open_max
= sysconf(_SC_OPEN_MAX
);
433 * Otherwise, we got -1 for one of the two
436 * (1) sysconf() did not understand _SC_OPEN_MAX
437 * and signaled an error with -1; or
438 * (2) sysconf() said there is no limit.
440 * We _could_ clear errno before calling sysconf() to
441 * tell these two cases apart and return a huge number
442 * in the latter case to let the caller cap it to a
443 * value that is not so selfish, but letting the
444 * fallback OPEN_MAX codepath take care of these cases
453 return 1; /* see the caller ;-) */
458 * Do not call this directly as this leaks p->pack_fd on error return;
459 * call open_packed_git() instead.
461 static int open_packed_git_1(struct packed_git
*p
)
464 struct pack_header hdr
;
465 unsigned char hash
[GIT_MAX_RAWSZ
];
466 unsigned char *idx_hash
;
469 const unsigned hashsz
= the_hash_algo
->rawsz
;
471 if (!p
->index_data
&& open_pack_index(p
))
472 return error("packfile %s index unavailable", p
->pack_name
);
475 unsigned int max_fds
= get_max_fd_limit();
477 /* Save 3 for stdin/stdout/stderr, 22 for work */
479 pack_max_fds
= max_fds
- 25;
484 while (pack_max_fds
<= pack_open_fds
&& close_one_pack())
487 p
->pack_fd
= git_open(p
->pack_name
);
488 if (p
->pack_fd
< 0 || fstat(p
->pack_fd
, &st
))
492 /* If we created the struct before we had the pack we lack size. */
494 if (!S_ISREG(st
.st_mode
))
495 return error("packfile %s not a regular file", p
->pack_name
);
496 p
->pack_size
= st
.st_size
;
497 } else if (p
->pack_size
!= st
.st_size
)
498 return error("packfile %s size changed", p
->pack_name
);
500 /* We leave these file descriptors open with sliding mmap;
501 * there is no point keeping them open across exec(), though.
503 fd_flag
= fcntl(p
->pack_fd
, F_GETFD
, 0);
505 return error("cannot determine file descriptor flags");
506 fd_flag
|= FD_CLOEXEC
;
507 if (fcntl(p
->pack_fd
, F_SETFD
, fd_flag
) == -1)
508 return error("cannot set FD_CLOEXEC");
510 /* Verify we recognize this pack file format. */
511 read_result
= read_in_full(p
->pack_fd
, &hdr
, sizeof(hdr
));
513 return error_errno("error reading from %s", p
->pack_name
);
514 if (read_result
!= sizeof(hdr
))
515 return error("file %s is far too short to be a packfile", p
->pack_name
);
516 if (hdr
.hdr_signature
!= htonl(PACK_SIGNATURE
))
517 return error("file %s is not a GIT packfile", p
->pack_name
);
518 if (!pack_version_ok(hdr
.hdr_version
))
519 return error("packfile %s is version %"PRIu32
" and not"
520 " supported (try upgrading GIT to a newer version)",
521 p
->pack_name
, ntohl(hdr
.hdr_version
));
523 /* Verify the pack matches its index. */
524 if (p
->num_objects
!= ntohl(hdr
.hdr_entries
))
525 return error("packfile %s claims to have %"PRIu32
" objects"
526 " while index indicates %"PRIu32
" objects",
527 p
->pack_name
, ntohl(hdr
.hdr_entries
),
529 if (lseek(p
->pack_fd
, p
->pack_size
- hashsz
, SEEK_SET
) == -1)
530 return error("end of packfile %s is unavailable", p
->pack_name
);
531 read_result
= read_in_full(p
->pack_fd
, hash
, hashsz
);
533 return error_errno("error reading from %s", p
->pack_name
);
534 if (read_result
!= hashsz
)
535 return error("packfile %s signature is unavailable", p
->pack_name
);
536 idx_hash
= ((unsigned char *)p
->index_data
) + p
->index_size
- hashsz
* 2;
537 if (hashcmp(hash
, idx_hash
))
538 return error("packfile %s does not match index", p
->pack_name
);
542 static int open_packed_git(struct packed_git
*p
)
544 if (!open_packed_git_1(p
))
550 static int in_window(struct pack_window
*win
, off_t offset
)
552 /* We must promise at least one full hash after the
553 * offset is available from this window, otherwise the offset
554 * is not actually in this window and a different window (which
555 * has that one hash excess) must be used. This is to support
556 * the object header and delta base parsing routines below.
558 off_t win_off
= win
->offset
;
559 return win_off
<= offset
560 && (offset
+ the_hash_algo
->rawsz
) <= (win_off
+ win
->len
);
563 unsigned char *use_pack(struct packed_git
*p
,
564 struct pack_window
**w_cursor
,
568 struct pack_window
*win
= *w_cursor
;
570 /* Since packfiles end in a hash of their content and it's
571 * pointless to ask for an offset into the middle of that
572 * hash, and the in_window function above wouldn't match
573 * don't allow an offset too close to the end of the file.
575 if (!p
->pack_size
&& p
->pack_fd
== -1 && open_packed_git(p
))
576 die("packfile %s cannot be accessed", p
->pack_name
);
577 if (offset
> (p
->pack_size
- the_hash_algo
->rawsz
))
578 die("offset beyond end of packfile (truncated pack?)");
580 die(_("offset before end of packfile (broken .idx?)"));
582 if (!win
|| !in_window(win
, offset
)) {
585 for (win
= p
->windows
; win
; win
= win
->next
) {
586 if (in_window(win
, offset
))
590 size_t window_align
= packed_git_window_size
/ 2;
593 if (p
->pack_fd
== -1 && open_packed_git(p
))
594 die("packfile %s cannot be accessed", p
->pack_name
);
596 win
= xcalloc(1, sizeof(*win
));
597 win
->offset
= (offset
/ window_align
) * window_align
;
598 len
= p
->pack_size
- win
->offset
;
599 if (len
> packed_git_window_size
)
600 len
= packed_git_window_size
;
601 win
->len
= (size_t)len
;
602 pack_mapped
+= win
->len
;
603 while (packed_git_limit
< pack_mapped
604 && unuse_one_window(p
))
606 win
->base
= xmmap(NULL
, win
->len
,
607 PROT_READ
, MAP_PRIVATE
,
608 p
->pack_fd
, win
->offset
);
609 if (win
->base
== MAP_FAILED
)
610 die_errno("packfile %s cannot be mapped",
612 if (!win
->offset
&& win
->len
== p
->pack_size
617 if (pack_mapped
> peak_pack_mapped
)
618 peak_pack_mapped
= pack_mapped
;
619 if (pack_open_windows
> peak_pack_open_windows
)
620 peak_pack_open_windows
= pack_open_windows
;
621 win
->next
= p
->windows
;
625 if (win
!= *w_cursor
) {
626 win
->last_used
= pack_used_ctr
++;
630 offset
-= win
->offset
;
632 *left
= win
->len
- xsize_t(offset
);
633 return win
->base
+ offset
;
636 void unuse_pack(struct pack_window
**w_cursor
)
638 struct pack_window
*w
= *w_cursor
;
645 static void try_to_free_pack_memory(size_t size
)
647 release_pack_memory(size
);
650 struct packed_git
*add_packed_git(const char *path
, size_t path_len
, int local
)
652 static int have_set_try_to_free_routine
;
655 struct packed_git
*p
;
657 if (!have_set_try_to_free_routine
) {
658 have_set_try_to_free_routine
= 1;
659 set_try_to_free_routine(try_to_free_pack_memory
);
663 * Make sure a corresponding .pack file exists and that
664 * the index looks sane.
666 if (!strip_suffix_mem(path
, &path_len
, ".idx"))
670 * ".promisor" is long enough to hold any suffix we're adding (and
671 * the use xsnprintf double-checks that)
673 alloc
= st_add3(path_len
, strlen(".promisor"), 1);
674 p
= alloc_packed_git(alloc
);
675 memcpy(p
->pack_name
, path
, path_len
);
677 xsnprintf(p
->pack_name
+ path_len
, alloc
- path_len
, ".keep");
678 if (!access(p
->pack_name
, F_OK
))
681 xsnprintf(p
->pack_name
+ path_len
, alloc
- path_len
, ".promisor");
682 if (!access(p
->pack_name
, F_OK
))
683 p
->pack_promisor
= 1;
685 xsnprintf(p
->pack_name
+ path_len
, alloc
- path_len
, ".pack");
686 if (stat(p
->pack_name
, &st
) || !S_ISREG(st
.st_mode
)) {
691 /* ok, it looks sane as far as we can check without
692 * actually mapping the pack file.
694 p
->pack_size
= st
.st_size
;
695 p
->pack_local
= local
;
696 p
->mtime
= st
.st_mtime
;
697 if (path_len
< the_hash_algo
->hexsz
||
698 get_sha1_hex(path
+ path_len
- the_hash_algo
->hexsz
, p
->sha1
))
703 void install_packed_git(struct repository
*r
, struct packed_git
*pack
)
705 if (pack
->pack_fd
!= -1)
708 pack
->next
= r
->objects
->packed_git
;
709 r
->objects
->packed_git
= pack
;
712 void (*report_garbage
)(unsigned seen_bits
, const char *path
);
714 static void report_helper(const struct string_list
*list
,
715 int seen_bits
, int first
, int last
)
717 if (seen_bits
== (PACKDIR_FILE_PACK
|PACKDIR_FILE_IDX
))
720 for (; first
< last
; first
++)
721 report_garbage(seen_bits
, list
->items
[first
].string
);
724 static void report_pack_garbage(struct string_list
*list
)
726 int i
, baselen
= -1, first
= 0, seen_bits
= 0;
731 string_list_sort(list
);
733 for (i
= 0; i
< list
->nr
; i
++) {
734 const char *path
= list
->items
[i
].string
;
736 strncmp(path
, list
->items
[first
].string
, baselen
)) {
737 report_helper(list
, seen_bits
, first
, i
);
742 const char *dot
= strrchr(path
, '.');
744 report_garbage(PACKDIR_FILE_GARBAGE
, path
);
747 baselen
= dot
- path
+ 1;
750 if (!strcmp(path
+ baselen
, "pack"))
752 else if (!strcmp(path
+ baselen
, "idx"))
755 report_helper(list
, seen_bits
, first
, list
->nr
);
758 void for_each_file_in_pack_dir(const char *objdir
,
759 each_file_in_pack_dir_fn fn
,
762 struct strbuf path
= STRBUF_INIT
;
767 strbuf_addstr(&path
, objdir
);
768 strbuf_addstr(&path
, "/pack");
769 dir
= opendir(path
.buf
);
772 error_errno("unable to open object pack directory: %s",
774 strbuf_release(&path
);
777 strbuf_addch(&path
, '/');
778 dirnamelen
= path
.len
;
779 while ((de
= readdir(dir
)) != NULL
) {
780 if (is_dot_or_dotdot(de
->d_name
))
783 strbuf_setlen(&path
, dirnamelen
);
784 strbuf_addstr(&path
, de
->d_name
);
786 fn(path
.buf
, path
.len
, de
->d_name
, data
);
790 strbuf_release(&path
);
793 struct prepare_pack_data
{
794 struct repository
*r
;
795 struct string_list
*garbage
;
799 static void prepare_pack(const char *full_name
, size_t full_name_len
,
800 const char *file_name
, void *_data
)
802 struct prepare_pack_data
*data
= (struct prepare_pack_data
*)_data
;
803 struct packed_git
*p
;
804 size_t base_len
= full_name_len
;
806 if (strip_suffix_mem(full_name
, &base_len
, ".idx")) {
807 /* Don't reopen a pack we already have. */
808 for (p
= data
->r
->objects
->packed_git
; p
; p
= p
->next
) {
810 if (strip_suffix(p
->pack_name
, ".pack", &len
) &&
812 !memcmp(p
->pack_name
, full_name
, len
))
817 p
= add_packed_git(full_name
, full_name_len
, data
->local
);
819 install_packed_git(data
->r
, p
);
826 if (ends_with(file_name
, ".idx") ||
827 ends_with(file_name
, ".pack") ||
828 ends_with(file_name
, ".bitmap") ||
829 ends_with(file_name
, ".keep") ||
830 ends_with(file_name
, ".promisor"))
831 string_list_append(data
->garbage
, full_name
);
833 report_garbage(PACKDIR_FILE_GARBAGE
, full_name
);
836 static void prepare_packed_git_one(struct repository
*r
, char *objdir
, int local
)
838 struct prepare_pack_data data
;
839 struct string_list garbage
= STRING_LIST_INIT_DUP
;
842 data
.garbage
= &garbage
;
845 for_each_file_in_pack_dir(objdir
, prepare_pack
, &data
);
847 report_pack_garbage(data
.garbage
);
848 string_list_clear(data
.garbage
, 0);
851 static void prepare_packed_git(struct repository
*r
);
853 * Give a fast, rough count of the number of objects in the repository. This
854 * ignores loose objects completely. If you have a lot of them, then either
855 * you should repack because your performance will be awful, or they are
856 * all unreachable objects about to be pruned, in which case they're not really
857 * interesting as a measure of repo size in the first place.
859 unsigned long approximate_object_count(void)
861 if (!the_repository
->objects
->approximate_object_count_valid
) {
863 struct packed_git
*p
;
865 prepare_packed_git(the_repository
);
867 for (p
= the_repository
->objects
->packed_git
; p
; p
= p
->next
) {
868 if (open_pack_index(p
))
870 count
+= p
->num_objects
;
872 the_repository
->objects
->approximate_object_count
= count
;
874 return the_repository
->objects
->approximate_object_count
;
877 static void *get_next_packed_git(const void *p
)
879 return ((const struct packed_git
*)p
)->next
;
882 static void set_next_packed_git(void *p
, void *next
)
884 ((struct packed_git
*)p
)->next
= next
;
887 static int sort_pack(const void *a_
, const void *b_
)
889 const struct packed_git
*a
= a_
;
890 const struct packed_git
*b
= b_
;
894 * Local packs tend to contain objects specific to our
895 * variant of the project than remote ones. In addition,
896 * remote ones could be on a network mounted filesystem.
897 * Favor local ones for these reasons.
899 st
= a
->pack_local
- b
->pack_local
;
904 * Younger packs tend to contain more recent objects,
905 * and more recent objects tend to get accessed more
908 if (a
->mtime
< b
->mtime
)
910 else if (a
->mtime
== b
->mtime
)
915 static void rearrange_packed_git(struct repository
*r
)
917 r
->objects
->packed_git
= llist_mergesort(
918 r
->objects
->packed_git
, get_next_packed_git
,
919 set_next_packed_git
, sort_pack
);
922 static void prepare_packed_git_mru(struct repository
*r
)
924 struct packed_git
*p
;
926 INIT_LIST_HEAD(&r
->objects
->packed_git_mru
);
928 for (p
= r
->objects
->packed_git
; p
; p
= p
->next
)
929 list_add_tail(&p
->mru
, &r
->objects
->packed_git_mru
);
932 static void prepare_packed_git(struct repository
*r
)
934 struct alternate_object_database
*alt
;
936 if (r
->objects
->packed_git_initialized
)
938 prepare_packed_git_one(r
, r
->objects
->objectdir
, 1);
940 for (alt
= r
->objects
->alt_odb_list
; alt
; alt
= alt
->next
)
941 prepare_packed_git_one(r
, alt
->path
, 0);
942 rearrange_packed_git(r
);
943 prepare_packed_git_mru(r
);
944 r
->objects
->packed_git_initialized
= 1;
947 void reprepare_packed_git(struct repository
*r
)
949 r
->objects
->approximate_object_count_valid
= 0;
950 r
->objects
->packed_git_initialized
= 0;
951 prepare_packed_git(r
);
954 struct packed_git
*get_packed_git(struct repository
*r
)
956 prepare_packed_git(r
);
957 return r
->objects
->packed_git
;
960 struct list_head
*get_packed_git_mru(struct repository
*r
)
962 prepare_packed_git(r
);
963 return &r
->objects
->packed_git_mru
;
966 unsigned long unpack_object_header_buffer(const unsigned char *buf
,
967 unsigned long len
, enum object_type
*type
, unsigned long *sizep
)
970 unsigned long size
, c
;
971 unsigned long used
= 0;
974 *type
= (c
>> 4) & 7;
978 if (len
<= used
|| bitsizeof(long) <= shift
) {
979 error("bad object header");
984 size
+= (c
& 0x7f) << shift
;
991 unsigned long get_size_from_delta(struct packed_git
*p
,
992 struct pack_window
**w_curs
,
995 const unsigned char *data
;
996 unsigned char delta_head
[20], *in
;
1000 memset(&stream
, 0, sizeof(stream
));
1001 stream
.next_out
= delta_head
;
1002 stream
.avail_out
= sizeof(delta_head
);
1004 git_inflate_init(&stream
);
1006 in
= use_pack(p
, w_curs
, curpos
, &stream
.avail_in
);
1007 stream
.next_in
= in
;
1008 st
= git_inflate(&stream
, Z_FINISH
);
1009 curpos
+= stream
.next_in
- in
;
1010 } while ((st
== Z_OK
|| st
== Z_BUF_ERROR
) &&
1011 stream
.total_out
< sizeof(delta_head
));
1012 git_inflate_end(&stream
);
1013 if ((st
!= Z_STREAM_END
) && stream
.total_out
!= sizeof(delta_head
)) {
1014 error("delta data unpack-initial failed");
1018 /* Examine the initial part of the delta to figure out
1023 /* ignore base size */
1024 get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
1026 /* Read the result size */
1027 return get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
1030 int unpack_object_header(struct packed_git
*p
,
1031 struct pack_window
**w_curs
,
1033 unsigned long *sizep
)
1035 unsigned char *base
;
1038 enum object_type type
;
1040 /* use_pack() assures us we have [base, base + 20) available
1041 * as a range that we can look at. (Its actually the hash
1042 * size that is assured.) With our object header encoding
1043 * the maximum deflated object size is 2^137, which is just
1044 * insane, so we know won't exceed what we have been given.
1046 base
= use_pack(p
, w_curs
, *curpos
, &left
);
1047 used
= unpack_object_header_buffer(base
, left
, &type
, sizep
);
1056 void mark_bad_packed_object(struct packed_git
*p
, const unsigned char *sha1
)
1059 for (i
= 0; i
< p
->num_bad_objects
; i
++)
1060 if (!hashcmp(sha1
, p
->bad_object_sha1
+ GIT_SHA1_RAWSZ
* i
))
1062 p
->bad_object_sha1
= xrealloc(p
->bad_object_sha1
,
1063 st_mult(GIT_MAX_RAWSZ
,
1064 st_add(p
->num_bad_objects
, 1)));
1065 hashcpy(p
->bad_object_sha1
+ GIT_SHA1_RAWSZ
* p
->num_bad_objects
, sha1
);
1066 p
->num_bad_objects
++;
1069 const struct packed_git
*has_packed_and_bad(const unsigned char *sha1
)
1071 struct packed_git
*p
;
1074 for (p
= the_repository
->objects
->packed_git
; p
; p
= p
->next
)
1075 for (i
= 0; i
< p
->num_bad_objects
; i
++)
1077 p
->bad_object_sha1
+ the_hash_algo
->rawsz
* i
))
1082 static off_t
get_delta_base(struct packed_git
*p
,
1083 struct pack_window
**w_curs
,
1085 enum object_type type
,
1086 off_t delta_obj_offset
)
1088 unsigned char *base_info
= use_pack(p
, w_curs
, *curpos
, NULL
);
1091 /* use_pack() assured us we have [base_info, base_info + 20)
1092 * as a range that we can look at without walking off the
1093 * end of the mapped window. Its actually the hash size
1094 * that is assured. An OFS_DELTA longer than the hash size
1095 * is stupid, as then a REF_DELTA would be smaller to store.
1097 if (type
== OBJ_OFS_DELTA
) {
1099 unsigned char c
= base_info
[used
++];
1100 base_offset
= c
& 127;
1103 if (!base_offset
|| MSB(base_offset
, 7))
1104 return 0; /* overflow */
1105 c
= base_info
[used
++];
1106 base_offset
= (base_offset
<< 7) + (c
& 127);
1108 base_offset
= delta_obj_offset
- base_offset
;
1109 if (base_offset
<= 0 || base_offset
>= delta_obj_offset
)
1110 return 0; /* out of bound */
1112 } else if (type
== OBJ_REF_DELTA
) {
1113 /* The base entry _must_ be in the same pack */
1114 base_offset
= find_pack_entry_one(base_info
, p
);
1115 *curpos
+= the_hash_algo
->rawsz
;
1117 die("I am totally screwed");
1122 * Like get_delta_base above, but we return the sha1 instead of the pack
1123 * offset. This means it is cheaper for REF deltas (we do not have to do
1124 * the final object lookup), but more expensive for OFS deltas (we
1125 * have to load the revidx to convert the offset back into a sha1).
1127 static const unsigned char *get_delta_base_sha1(struct packed_git
*p
,
1128 struct pack_window
**w_curs
,
1130 enum object_type type
,
1131 off_t delta_obj_offset
)
1133 if (type
== OBJ_REF_DELTA
) {
1134 unsigned char *base
= use_pack(p
, w_curs
, curpos
, NULL
);
1136 } else if (type
== OBJ_OFS_DELTA
) {
1137 struct revindex_entry
*revidx
;
1138 off_t base_offset
= get_delta_base(p
, w_curs
, &curpos
,
1139 type
, delta_obj_offset
);
1144 revidx
= find_pack_revindex(p
, base_offset
);
1148 return nth_packed_object_sha1(p
, revidx
->nr
);
1153 static int retry_bad_packed_offset(struct repository
*r
,
1154 struct packed_git
*p
,
1158 struct revindex_entry
*revidx
;
1159 struct object_id oid
;
1160 revidx
= find_pack_revindex(p
, obj_offset
);
1163 nth_packed_object_oid(&oid
, p
, revidx
->nr
);
1164 mark_bad_packed_object(p
, oid
.hash
);
1165 type
= oid_object_info(r
, &oid
, NULL
);
1166 if (type
<= OBJ_NONE
)
1171 #define POI_STACK_PREALLOC 64
1173 static enum object_type
packed_to_object_type(struct repository
*r
,
1174 struct packed_git
*p
,
1176 enum object_type type
,
1177 struct pack_window
**w_curs
,
1180 off_t small_poi_stack
[POI_STACK_PREALLOC
];
1181 off_t
*poi_stack
= small_poi_stack
;
1182 int poi_stack_nr
= 0, poi_stack_alloc
= POI_STACK_PREALLOC
;
1184 while (type
== OBJ_OFS_DELTA
|| type
== OBJ_REF_DELTA
) {
1187 /* Push the object we're going to leave behind */
1188 if (poi_stack_nr
>= poi_stack_alloc
&& poi_stack
== small_poi_stack
) {
1189 poi_stack_alloc
= alloc_nr(poi_stack_nr
);
1190 ALLOC_ARRAY(poi_stack
, poi_stack_alloc
);
1191 memcpy(poi_stack
, small_poi_stack
, sizeof(off_t
)*poi_stack_nr
);
1193 ALLOC_GROW(poi_stack
, poi_stack_nr
+1, poi_stack_alloc
);
1195 poi_stack
[poi_stack_nr
++] = obj_offset
;
1196 /* If parsing the base offset fails, just unwind */
1197 base_offset
= get_delta_base(p
, w_curs
, &curpos
, type
, obj_offset
);
1200 curpos
= obj_offset
= base_offset
;
1201 type
= unpack_object_header(p
, w_curs
, &curpos
, &size
);
1202 if (type
<= OBJ_NONE
) {
1203 /* If getting the base itself fails, we first
1204 * retry the base, otherwise unwind */
1205 type
= retry_bad_packed_offset(r
, p
, base_offset
);
1206 if (type
> OBJ_NONE
)
1220 error("unknown object type %i at offset %"PRIuMAX
" in %s",
1221 type
, (uintmax_t)obj_offset
, p
->pack_name
);
1226 if (poi_stack
!= small_poi_stack
)
1231 while (poi_stack_nr
) {
1232 obj_offset
= poi_stack
[--poi_stack_nr
];
1233 type
= retry_bad_packed_offset(r
, p
, obj_offset
);
1234 if (type
> OBJ_NONE
)
1241 static struct hashmap delta_base_cache
;
1242 static size_t delta_base_cached
;
1244 static LIST_HEAD(delta_base_cache_lru
);
1246 struct delta_base_cache_key
{
1247 struct packed_git
*p
;
1251 struct delta_base_cache_entry
{
1252 struct hashmap hash
;
1253 struct delta_base_cache_key key
;
1254 struct list_head lru
;
1257 enum object_type type
;
1260 static unsigned int pack_entry_hash(struct packed_git
*p
, off_t base_offset
)
1264 hash
= (unsigned int)(intptr_t)p
+ (unsigned int)base_offset
;
1265 hash
+= (hash
>> 8) + (hash
>> 16);
1269 static struct delta_base_cache_entry
*
1270 get_delta_base_cache_entry(struct packed_git
*p
, off_t base_offset
)
1272 struct hashmap_entry entry
;
1273 struct delta_base_cache_key key
;
1275 if (!delta_base_cache
.cmpfn
)
1278 hashmap_entry_init(&entry
, pack_entry_hash(p
, base_offset
));
1280 key
.base_offset
= base_offset
;
1281 return hashmap_get(&delta_base_cache
, &entry
, &key
);
1284 static int delta_base_cache_key_eq(const struct delta_base_cache_key
*a
,
1285 const struct delta_base_cache_key
*b
)
1287 return a
->p
== b
->p
&& a
->base_offset
== b
->base_offset
;
1290 static int delta_base_cache_hash_cmp(const void *unused_cmp_data
,
1291 const void *va
, const void *vb
,
1294 const struct delta_base_cache_entry
*a
= va
, *b
= vb
;
1295 const struct delta_base_cache_key
*key
= vkey
;
1297 return !delta_base_cache_key_eq(&a
->key
, key
);
1299 return !delta_base_cache_key_eq(&a
->key
, &b
->key
);
1302 static int in_delta_base_cache(struct packed_git
*p
, off_t base_offset
)
1304 return !!get_delta_base_cache_entry(p
, base_offset
);
1308 * Remove the entry from the cache, but do _not_ free the associated
1309 * entry data. The caller takes ownership of the "data" buffer, and
1310 * should copy out any fields it wants before detaching.
1312 static void detach_delta_base_cache_entry(struct delta_base_cache_entry
*ent
)
1314 hashmap_remove(&delta_base_cache
, ent
, &ent
->key
);
1315 list_del(&ent
->lru
);
1316 delta_base_cached
-= ent
->size
;
1320 static void *cache_or_unpack_entry(struct repository
*r
, struct packed_git
*p
,
1321 off_t base_offset
, unsigned long *base_size
,
1322 enum object_type
*type
)
1324 struct delta_base_cache_entry
*ent
;
1326 ent
= get_delta_base_cache_entry(p
, base_offset
);
1328 return unpack_entry(r
, p
, base_offset
, type
, base_size
);
1333 *base_size
= ent
->size
;
1334 return xmemdupz(ent
->data
, ent
->size
);
1337 static inline void release_delta_base_cache(struct delta_base_cache_entry
*ent
)
1340 detach_delta_base_cache_entry(ent
);
1343 void clear_delta_base_cache(void)
1345 struct list_head
*lru
, *tmp
;
1346 list_for_each_safe(lru
, tmp
, &delta_base_cache_lru
) {
1347 struct delta_base_cache_entry
*entry
=
1348 list_entry(lru
, struct delta_base_cache_entry
, lru
);
1349 release_delta_base_cache(entry
);
1353 static void add_delta_base_cache(struct packed_git
*p
, off_t base_offset
,
1354 void *base
, unsigned long base_size
, enum object_type type
)
1356 struct delta_base_cache_entry
*ent
= xmalloc(sizeof(*ent
));
1357 struct list_head
*lru
, *tmp
;
1359 delta_base_cached
+= base_size
;
1361 list_for_each_safe(lru
, tmp
, &delta_base_cache_lru
) {
1362 struct delta_base_cache_entry
*f
=
1363 list_entry(lru
, struct delta_base_cache_entry
, lru
);
1364 if (delta_base_cached
<= delta_base_cache_limit
)
1366 release_delta_base_cache(f
);
1370 ent
->key
.base_offset
= base_offset
;
1373 ent
->size
= base_size
;
1374 list_add_tail(&ent
->lru
, &delta_base_cache_lru
);
1376 if (!delta_base_cache
.cmpfn
)
1377 hashmap_init(&delta_base_cache
, delta_base_cache_hash_cmp
, NULL
, 0);
1378 hashmap_entry_init(ent
, pack_entry_hash(p
, base_offset
));
1379 hashmap_add(&delta_base_cache
, ent
);
1382 int packed_object_info(struct repository
*r
, struct packed_git
*p
,
1383 off_t obj_offset
, struct object_info
*oi
)
1385 struct pack_window
*w_curs
= NULL
;
1387 off_t curpos
= obj_offset
;
1388 enum object_type type
;
1391 * We always get the representation type, but only convert it to
1392 * a "real" type later if the caller is interested.
1395 *oi
->contentp
= cache_or_unpack_entry(r
, p
, obj_offset
, oi
->sizep
,
1400 type
= unpack_object_header(p
, &w_curs
, &curpos
, &size
);
1403 if (!oi
->contentp
&& oi
->sizep
) {
1404 if (type
== OBJ_OFS_DELTA
|| type
== OBJ_REF_DELTA
) {
1405 off_t tmp_pos
= curpos
;
1406 off_t base_offset
= get_delta_base(p
, &w_curs
, &tmp_pos
,
1412 *oi
->sizep
= get_size_from_delta(p
, &w_curs
, tmp_pos
);
1413 if (*oi
->sizep
== 0) {
1422 if (oi
->disk_sizep
) {
1423 struct revindex_entry
*revidx
= find_pack_revindex(p
, obj_offset
);
1424 *oi
->disk_sizep
= revidx
[1].offset
- obj_offset
;
1427 if (oi
->typep
|| oi
->type_name
) {
1428 enum object_type ptot
;
1429 ptot
= packed_to_object_type(r
, p
, obj_offset
,
1430 type
, &w_curs
, curpos
);
1433 if (oi
->type_name
) {
1434 const char *tn
= type_name(ptot
);
1436 strbuf_addstr(oi
->type_name
, tn
);
1444 if (oi
->delta_base_sha1
) {
1445 if (type
== OBJ_OFS_DELTA
|| type
== OBJ_REF_DELTA
) {
1446 const unsigned char *base
;
1448 base
= get_delta_base_sha1(p
, &w_curs
, curpos
,
1455 hashcpy(oi
->delta_base_sha1
, base
);
1457 hashclr(oi
->delta_base_sha1
);
1460 oi
->whence
= in_delta_base_cache(p
, obj_offset
) ? OI_DBCACHED
:
1464 unuse_pack(&w_curs
);
1468 static void *unpack_compressed_entry(struct packed_git
*p
,
1469 struct pack_window
**w_curs
,
1475 unsigned char *buffer
, *in
;
1477 buffer
= xmallocz_gently(size
);
1480 memset(&stream
, 0, sizeof(stream
));
1481 stream
.next_out
= buffer
;
1482 stream
.avail_out
= size
+ 1;
1484 git_inflate_init(&stream
);
1486 in
= use_pack(p
, w_curs
, curpos
, &stream
.avail_in
);
1487 stream
.next_in
= in
;
1488 st
= git_inflate(&stream
, Z_FINISH
);
1489 if (!stream
.avail_out
)
1490 break; /* the payload is larger than it should be */
1491 curpos
+= stream
.next_in
- in
;
1492 } while (st
== Z_OK
|| st
== Z_BUF_ERROR
);
1493 git_inflate_end(&stream
);
1494 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= size
) {
1499 /* versions of zlib can clobber unconsumed portion of outbuf */
1500 buffer
[size
] = '\0';
1505 static void write_pack_access_log(struct packed_git
*p
, off_t obj_offset
)
1507 static struct trace_key pack_access
= TRACE_KEY_INIT(PACK_ACCESS
);
1508 trace_printf_key(&pack_access
, "%s %"PRIuMAX
"\n",
1509 p
->pack_name
, (uintmax_t)obj_offset
);
1512 int do_check_packed_object_crc
;
1514 #define UNPACK_ENTRY_STACK_PREALLOC 64
1515 struct unpack_entry_stack_ent
{
1521 static void *read_object(struct repository
*r
,
1522 const struct object_id
*oid
,
1523 enum object_type
*type
,
1524 unsigned long *size
)
1526 struct object_info oi
= OBJECT_INFO_INIT
;
1530 oi
.contentp
= &content
;
1532 if (oid_object_info_extended(r
, oid
, &oi
, 0) < 0)
1537 void *unpack_entry(struct repository
*r
, struct packed_git
*p
, off_t obj_offset
,
1538 enum object_type
*final_type
, unsigned long *final_size
)
1540 struct pack_window
*w_curs
= NULL
;
1541 off_t curpos
= obj_offset
;
1544 enum object_type type
;
1545 struct unpack_entry_stack_ent small_delta_stack
[UNPACK_ENTRY_STACK_PREALLOC
];
1546 struct unpack_entry_stack_ent
*delta_stack
= small_delta_stack
;
1547 int delta_stack_nr
= 0, delta_stack_alloc
= UNPACK_ENTRY_STACK_PREALLOC
;
1548 int base_from_cache
= 0;
1550 write_pack_access_log(p
, obj_offset
);
1552 /* PHASE 1: drill down to the innermost base object */
1556 struct delta_base_cache_entry
*ent
;
1558 ent
= get_delta_base_cache_entry(p
, curpos
);
1563 detach_delta_base_cache_entry(ent
);
1564 base_from_cache
= 1;
1568 if (do_check_packed_object_crc
&& p
->index_version
> 1) {
1569 struct revindex_entry
*revidx
= find_pack_revindex(p
, obj_offset
);
1570 off_t len
= revidx
[1].offset
- obj_offset
;
1571 if (check_pack_crc(p
, &w_curs
, obj_offset
, len
, revidx
->nr
)) {
1572 struct object_id oid
;
1573 nth_packed_object_oid(&oid
, p
, revidx
->nr
);
1574 error("bad packed object CRC for %s",
1576 mark_bad_packed_object(p
, oid
.hash
);
1582 type
= unpack_object_header(p
, &w_curs
, &curpos
, &size
);
1583 if (type
!= OBJ_OFS_DELTA
&& type
!= OBJ_REF_DELTA
)
1586 base_offset
= get_delta_base(p
, &w_curs
, &curpos
, type
, obj_offset
);
1588 error("failed to validate delta base reference "
1589 "at offset %"PRIuMAX
" from %s",
1590 (uintmax_t)curpos
, p
->pack_name
);
1591 /* bail to phase 2, in hopes of recovery */
1596 /* push object, proceed to base */
1597 if (delta_stack_nr
>= delta_stack_alloc
1598 && delta_stack
== small_delta_stack
) {
1599 delta_stack_alloc
= alloc_nr(delta_stack_nr
);
1600 ALLOC_ARRAY(delta_stack
, delta_stack_alloc
);
1601 memcpy(delta_stack
, small_delta_stack
,
1602 sizeof(*delta_stack
)*delta_stack_nr
);
1604 ALLOC_GROW(delta_stack
, delta_stack_nr
+1, delta_stack_alloc
);
1606 i
= delta_stack_nr
++;
1607 delta_stack
[i
].obj_offset
= obj_offset
;
1608 delta_stack
[i
].curpos
= curpos
;
1609 delta_stack
[i
].size
= size
;
1611 curpos
= obj_offset
= base_offset
;
1614 /* PHASE 2: handle the base */
1619 BUG("unpack_entry: left loop at a valid delta");
1625 if (!base_from_cache
)
1626 data
= unpack_compressed_entry(p
, &w_curs
, curpos
, size
);
1630 error("unknown object type %i at offset %"PRIuMAX
" in %s",
1631 type
, (uintmax_t)obj_offset
, p
->pack_name
);
1634 /* PHASE 3: apply deltas in order */
1637 * 'data' holds the base data, or NULL if there was corruption
1639 while (delta_stack_nr
) {
1642 void *external_base
= NULL
;
1643 unsigned long delta_size
, base_size
= size
;
1649 add_delta_base_cache(p
, obj_offset
, base
, base_size
, type
);
1653 * We're probably in deep shit, but let's try to fetch
1654 * the required base anyway from another pack or loose.
1655 * This is costly but should happen only in the presence
1656 * of a corrupted pack, and is better than failing outright.
1658 struct revindex_entry
*revidx
;
1659 struct object_id base_oid
;
1660 revidx
= find_pack_revindex(p
, obj_offset
);
1662 nth_packed_object_oid(&base_oid
, p
, revidx
->nr
);
1663 error("failed to read delta base object %s"
1664 " at offset %"PRIuMAX
" from %s",
1665 oid_to_hex(&base_oid
), (uintmax_t)obj_offset
,
1667 mark_bad_packed_object(p
, base_oid
.hash
);
1668 base
= read_object(r
, &base_oid
, &type
, &base_size
);
1669 external_base
= base
;
1673 i
= --delta_stack_nr
;
1674 obj_offset
= delta_stack
[i
].obj_offset
;
1675 curpos
= delta_stack
[i
].curpos
;
1676 delta_size
= delta_stack
[i
].size
;
1681 delta_data
= unpack_compressed_entry(p
, &w_curs
, curpos
, delta_size
);
1684 error("failed to unpack compressed delta "
1685 "at offset %"PRIuMAX
" from %s",
1686 (uintmax_t)curpos
, p
->pack_name
);
1688 free(external_base
);
1692 data
= patch_delta(base
, base_size
,
1693 delta_data
, delta_size
,
1697 * We could not apply the delta; warn the user, but keep going.
1698 * Our failure will be noticed either in the next iteration of
1699 * the loop, or if this is the final delta, in the caller when
1700 * we return NULL. Those code paths will take care of making
1701 * a more explicit warning and retrying with another copy of
1705 error("failed to apply delta");
1708 free(external_base
);
1717 unuse_pack(&w_curs
);
1719 if (delta_stack
!= small_delta_stack
)
1725 int bsearch_pack(const struct object_id
*oid
, const struct packed_git
*p
, uint32_t *result
)
1727 const unsigned char *index_fanout
= p
->index_data
;
1728 const unsigned char *index_lookup
;
1729 const unsigned int hashsz
= the_hash_algo
->rawsz
;
1730 int index_lookup_width
;
1733 BUG("bsearch_pack called without a valid pack-index");
1735 index_lookup
= index_fanout
+ 4 * 256;
1736 if (p
->index_version
== 1) {
1737 index_lookup_width
= hashsz
+ 4;
1740 index_lookup_width
= hashsz
;
1745 return bsearch_hash(oid
->hash
, (const uint32_t*)index_fanout
,
1746 index_lookup
, index_lookup_width
, result
);
1749 const unsigned char *nth_packed_object_sha1(struct packed_git
*p
,
1752 const unsigned char *index
= p
->index_data
;
1753 const unsigned int hashsz
= the_hash_algo
->rawsz
;
1755 if (open_pack_index(p
))
1757 index
= p
->index_data
;
1759 if (n
>= p
->num_objects
)
1762 if (p
->index_version
== 1) {
1763 return index
+ (hashsz
+ 4) * n
+ 4;
1766 return index
+ hashsz
* n
;
1770 const struct object_id
*nth_packed_object_oid(struct object_id
*oid
,
1771 struct packed_git
*p
,
1774 const unsigned char *hash
= nth_packed_object_sha1(p
, n
);
1777 hashcpy(oid
->hash
, hash
);
1781 void check_pack_index_ptr(const struct packed_git
*p
, const void *vptr
)
1783 const unsigned char *ptr
= vptr
;
1784 const unsigned char *start
= p
->index_data
;
1785 const unsigned char *end
= start
+ p
->index_size
;
1787 die(_("offset before start of pack index for %s (corrupt index?)"),
1789 /* No need to check for underflow; .idx files must be at least 8 bytes */
1791 die(_("offset beyond end of pack index for %s (truncated index?)"),
1795 off_t
nth_packed_object_offset(const struct packed_git
*p
, uint32_t n
)
1797 const unsigned char *index
= p
->index_data
;
1798 const unsigned int hashsz
= the_hash_algo
->rawsz
;
1800 if (p
->index_version
== 1) {
1801 return ntohl(*((uint32_t *)(index
+ (hashsz
+ 4) * n
)));
1804 index
+= 8 + p
->num_objects
* (hashsz
+ 4);
1805 off
= ntohl(*((uint32_t *)(index
+ 4 * n
)));
1806 if (!(off
& 0x80000000))
1808 index
+= p
->num_objects
* 4 + (off
& 0x7fffffff) * 8;
1809 check_pack_index_ptr(p
, index
);
1810 return get_be64(index
);
1814 off_t
find_pack_entry_one(const unsigned char *sha1
,
1815 struct packed_git
*p
)
1817 const unsigned char *index
= p
->index_data
;
1818 struct object_id oid
;
1822 if (open_pack_index(p
))
1826 hashcpy(oid
.hash
, sha1
);
1827 if (bsearch_pack(&oid
, p
, &result
))
1828 return nth_packed_object_offset(p
, result
);
1832 int is_pack_valid(struct packed_git
*p
)
1834 /* An already open pack is known to be valid. */
1835 if (p
->pack_fd
!= -1)
1838 /* If the pack has one window completely covering the
1839 * file size, the pack is known to be valid even if
1840 * the descriptor is not currently open.
1843 struct pack_window
*w
= p
->windows
;
1845 if (!w
->offset
&& w
->len
== p
->pack_size
)
1849 /* Force the pack to open to prove its valid. */
1850 return !open_packed_git(p
);
1853 struct packed_git
*find_sha1_pack(const unsigned char *sha1
,
1854 struct packed_git
*packs
)
1856 struct packed_git
*p
;
1858 for (p
= packs
; p
; p
= p
->next
) {
1859 if (find_pack_entry_one(sha1
, p
))
1866 static int fill_pack_entry(const struct object_id
*oid
,
1867 struct pack_entry
*e
,
1868 struct packed_git
*p
)
1872 if (p
->num_bad_objects
) {
1874 for (i
= 0; i
< p
->num_bad_objects
; i
++)
1875 if (!hashcmp(oid
->hash
,
1876 p
->bad_object_sha1
+ the_hash_algo
->rawsz
* i
))
1880 offset
= find_pack_entry_one(oid
->hash
, p
);
1885 * We are about to tell the caller where they can locate the
1886 * requested object. We better make sure the packfile is
1887 * still here and can be accessed before supplying that
1888 * answer, as it may have been deleted since the index was
1891 if (!is_pack_valid(p
))
1898 int find_pack_entry(struct repository
*r
, const struct object_id
*oid
, struct pack_entry
*e
)
1900 struct list_head
*pos
;
1902 prepare_packed_git(r
);
1903 if (!r
->objects
->packed_git
)
1906 list_for_each(pos
, &r
->objects
->packed_git_mru
) {
1907 struct packed_git
*p
= list_entry(pos
, struct packed_git
, mru
);
1908 if (fill_pack_entry(oid
, e
, p
)) {
1909 list_move(&p
->mru
, &r
->objects
->packed_git_mru
);
1916 int has_object_pack(const struct object_id
*oid
)
1918 struct pack_entry e
;
1919 return find_pack_entry(the_repository
, oid
, &e
);
1922 int has_pack_index(const unsigned char *sha1
)
1925 if (stat(sha1_pack_index_name(sha1
), &st
))
1930 int for_each_object_in_pack(struct packed_git
*p
, each_packed_object_fn cb
, void *data
)
1935 for (i
= 0; i
< p
->num_objects
; i
++) {
1936 struct object_id oid
;
1938 if (!nth_packed_object_oid(&oid
, p
, i
))
1939 return error("unable to get sha1 of object %u in %s",
1942 r
= cb(&oid
, p
, i
, data
);
1949 int for_each_packed_object(each_packed_object_fn cb
, void *data
, unsigned flags
)
1951 struct packed_git
*p
;
1953 int pack_errors
= 0;
1955 prepare_packed_git(the_repository
);
1956 for (p
= the_repository
->objects
->packed_git
; p
; p
= p
->next
) {
1957 if ((flags
& FOR_EACH_OBJECT_LOCAL_ONLY
) && !p
->pack_local
)
1959 if ((flags
& FOR_EACH_OBJECT_PROMISOR_ONLY
) &&
1962 if (open_pack_index(p
)) {
1966 r
= for_each_object_in_pack(p
, cb
, data
);
1970 return r
? r
: pack_errors
;
1973 static int add_promisor_object(const struct object_id
*oid
,
1974 struct packed_git
*pack
,
1978 struct oidset
*set
= set_
;
1979 struct object
*obj
= parse_object(oid
);
1983 oidset_insert(set
, oid
);
1986 * If this is a tree, commit, or tag, the objects it refers
1987 * to are also promisor objects. (Blobs refer to no objects->)
1989 if (obj
->type
== OBJ_TREE
) {
1990 struct tree
*tree
= (struct tree
*)obj
;
1991 struct tree_desc desc
;
1992 struct name_entry entry
;
1993 if (init_tree_desc_gently(&desc
, tree
->buffer
, tree
->size
))
1995 * Error messages are given when packs are
1996 * verified, so do not print any here.
1999 while (tree_entry_gently(&desc
, &entry
))
2000 oidset_insert(set
, entry
.oid
);
2001 } else if (obj
->type
== OBJ_COMMIT
) {
2002 struct commit
*commit
= (struct commit
*) obj
;
2003 struct commit_list
*parents
= commit
->parents
;
2005 oidset_insert(set
, get_commit_tree_oid(commit
));
2006 for (; parents
; parents
= parents
->next
)
2007 oidset_insert(set
, &parents
->item
->object
.oid
);
2008 } else if (obj
->type
== OBJ_TAG
) {
2009 struct tag
*tag
= (struct tag
*) obj
;
2010 oidset_insert(set
, &tag
->tagged
->oid
);
2015 int is_promisor_object(const struct object_id
*oid
)
2017 static struct oidset promisor_objects
;
2018 static int promisor_objects_prepared
;
2020 if (!promisor_objects_prepared
) {
2021 if (repository_format_partial_clone
) {
2022 for_each_packed_object(add_promisor_object
,
2024 FOR_EACH_OBJECT_PROMISOR_ONLY
);
2026 promisor_objects_prepared
= 1;
2028 return oidset_contains(&promisor_objects
, oid
);