2 * GIT - The information manager from hell
4 * Copyright (C) Linus Torvalds, 2005
6 * This handles basic git sha1 object files - packing, unpacking,
18 #if defined(__linux__) && (defined(__i386__) || defined(__PPC__))
19 #define O_NOATIME 01000000
31 const unsigned char null_sha1
[20];
33 static unsigned int sha1_file_open_flag
= O_NOATIME
;
35 signed char hexval_table
[256] = {
36 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
37 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
38 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
39 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
40 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
41 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
42 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
43 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
44 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
45 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
46 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
47 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
48 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
49 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
50 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
51 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
52 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
53 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
54 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
55 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
56 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
57 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
58 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
59 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
60 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
61 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
62 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
63 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
64 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
65 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
66 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
67 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
70 int get_sha1_hex(const char *hex
, unsigned char *sha1
)
73 for (i
= 0; i
< 20; i
++) {
74 unsigned int val
= (hexval(hex
[0]) << 4) | hexval(hex
[1]);
83 /* returns the number of chars to skip to first component */
84 static inline int is_path_absolute(const char *path
)
87 if (isalpha(path
[0]) && path
[1] == ':')
88 return 2 + (path
[2] == '/');
89 /* TODO: C:dir/file 'relative' paths are not catered for */
94 int safe_create_leading_directories(char *path
)
96 char *pos
= path
+ is_path_absolute(path
);
100 pos
= strchr(pos
, '/');
104 if (!stat(path
, &st
)) {
106 if (!S_ISDIR(st
.st_mode
)) {
111 else if (mkdir(path
, 0777)) {
115 else if (adjust_shared_perm(path
)) {
124 char * sha1_to_hex(const unsigned char *sha1
)
127 static char hexbuffer
[4][50];
128 static const char hex
[] = "0123456789abcdef";
129 char *buffer
= hexbuffer
[3 & ++bufno
], *buf
= buffer
;
132 for (i
= 0; i
< 20; i
++) {
133 unsigned int val
= *sha1
++;
134 *buf
++ = hex
[val
>> 4];
135 *buf
++ = hex
[val
& 0xf];
142 static void fill_sha1_path(char *pathbuf
, const unsigned char *sha1
)
145 for (i
= 0; i
< 20; i
++) {
146 static char hex
[] = "0123456789abcdef";
147 unsigned int val
= sha1
[i
];
148 char *pos
= pathbuf
+ i
*2 + (i
> 0);
149 *pos
++ = hex
[val
>> 4];
150 *pos
= hex
[val
& 0xf];
155 * NOTE! This returns a statically allocated buffer, so you have to be
156 * careful about using it. Do a "xstrdup()" if you need to save the
159 * Also note that this returns the location for creating. Reading
160 * SHA1 file can happen from any alternate directory listed in the
161 * DB_ENVIRONMENT environment variable if it is not found in
162 * the primary object database.
164 char *sha1_file_name(const unsigned char *sha1
)
166 static char *name
, *base
;
169 const char *sha1_file_directory
= get_object_directory();
170 int len
= strlen(sha1_file_directory
);
171 base
= xmalloc(len
+ 60);
172 memcpy(base
, sha1_file_directory
, len
);
173 memset(base
+len
, 0, 60);
176 name
= base
+ len
+ 1;
178 fill_sha1_path(name
, sha1
);
182 char *sha1_pack_name(const unsigned char *sha1
)
184 static const char hex
[] = "0123456789abcdef";
185 static char *name
, *base
, *buf
;
189 const char *sha1_file_directory
= get_object_directory();
190 int len
= strlen(sha1_file_directory
);
191 base
= xmalloc(len
+ 60);
192 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory
);
193 name
= base
+ len
+ 11;
198 for (i
= 0; i
< 20; i
++) {
199 unsigned int val
= *sha1
++;
200 *buf
++ = hex
[val
>> 4];
201 *buf
++ = hex
[val
& 0xf];
207 char *sha1_pack_index_name(const unsigned char *sha1
)
209 static const char hex
[] = "0123456789abcdef";
210 static char *name
, *base
, *buf
;
214 const char *sha1_file_directory
= get_object_directory();
215 int len
= strlen(sha1_file_directory
);
216 base
= xmalloc(len
+ 60);
217 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory
);
218 name
= base
+ len
+ 11;
223 for (i
= 0; i
< 20; i
++) {
224 unsigned int val
= *sha1
++;
225 *buf
++ = hex
[val
>> 4];
226 *buf
++ = hex
[val
& 0xf];
232 struct alternate_object_database
*alt_odb_list
;
233 static struct alternate_object_database
**alt_odb_tail
;
235 static void read_info_alternates(const char * alternates
, int depth
);
238 * Prepare alternate object database registry.
240 * The variable alt_odb_list points at the list of struct
241 * alternate_object_database. The elements on this list come from
242 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
243 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
244 * whose contents is similar to that environment variable but can be
245 * LF separated. Its base points at a statically allocated buffer that
246 * contains "/the/directory/corresponding/to/.git/objects/...", while
247 * its name points just after the slash at the end of ".git/objects/"
248 * in the example above, and has enough space to hold 40-byte hex
249 * SHA1, an extra slash for the first level indirection, and the
252 static int link_alt_odb_entry(const char * entry
, int len
, const char * relative_base
, int depth
)
255 const char *objdir
= get_object_directory();
256 struct alternate_object_database
*ent
;
257 struct alternate_object_database
*alt
;
258 /* 43 = 40-byte + 2 '/' + terminating NUL */
260 int entlen
= pfxlen
+ 43;
263 if (!is_path_absolute(entry
) && relative_base
) {
264 /* Relative alt-odb */
266 base_len
= strlen(relative_base
) + 1;
270 ent
= xmalloc(sizeof(*ent
) + entlen
);
272 if (!is_path_absolute(entry
) && relative_base
) {
273 memcpy(ent
->base
, relative_base
, base_len
- 1);
274 ent
->base
[base_len
- 1] = '/';
275 memcpy(ent
->base
+ base_len
, entry
, len
);
278 memcpy(ent
->base
, entry
, pfxlen
);
280 ent
->name
= ent
->base
+ pfxlen
+ 1;
281 ent
->base
[pfxlen
+ 3] = '/';
282 ent
->base
[pfxlen
] = ent
->base
[entlen
-1] = 0;
284 /* Detect cases where alternate disappeared */
285 if (stat(ent
->base
, &st
) || !S_ISDIR(st
.st_mode
)) {
286 error("object directory %s does not exist; "
287 "check .git/objects/info/alternates.",
293 /* Prevent the common mistake of listing the same
294 * thing twice, or object directory itself.
296 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
297 if (!memcmp(ent
->base
, alt
->base
, pfxlen
)) {
302 if (!memcmp(ent
->base
, objdir
, pfxlen
)) {
307 /* add the alternate entry */
309 alt_odb_tail
= &(ent
->next
);
312 /* recursively add alternates */
313 read_info_alternates(ent
->base
, depth
+ 1);
315 ent
->base
[pfxlen
] = '/';
320 static void link_alt_odb_entries(const char *alt
, const char *ep
, int sep
,
321 const char *relative_base
, int depth
)
323 const char *cp
, *last
;
326 error("%s: ignoring alternate object stores, nesting too deep.",
334 if (cp
< ep
&& *cp
== '#') {
335 while (cp
< ep
&& *cp
!= sep
)
340 while (cp
< ep
&& *cp
!= sep
)
343 if (!is_path_absolute(last
) && depth
) {
344 error("%s: ignoring relative alternate object store %s",
345 relative_base
, last
);
347 link_alt_odb_entry(last
, cp
- last
,
348 relative_base
, depth
);
351 while (cp
< ep
&& *cp
== sep
)
357 static void read_info_alternates(const char * relative_base
, int depth
)
364 sprintf(path
, "%s/info/alternates", relative_base
);
365 fd
= open(path
, O_RDONLY
);
368 if (fstat(fd
, &st
) || (st
.st_size
== 0)) {
372 map
= xmmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
375 link_alt_odb_entries(map
, map
+ st
.st_size
, '\n', relative_base
, depth
);
377 munmap(map
, st
.st_size
);
380 void prepare_alt_odb(void)
384 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
389 alt_odb_tail
= &alt_odb_list
;
390 link_alt_odb_entries(alt
, alt
+ strlen(alt
), ':', NULL
, 0);
392 read_info_alternates(get_object_directory(), 0);
395 static char *find_sha1_file(const unsigned char *sha1
, struct stat
*st
)
397 char *name
= sha1_file_name(sha1
);
398 struct alternate_object_database
*alt
;
403 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
405 fill_sha1_path(name
, sha1
);
406 if (!stat(alt
->base
, st
))
412 static unsigned int pack_used_ctr
;
413 static unsigned int pack_mmap_calls
;
414 static unsigned int peak_pack_open_windows
;
415 static unsigned int pack_open_windows
;
416 static size_t peak_pack_mapped
;
417 static size_t pack_mapped
;
418 struct packed_git
*packed_git
;
423 "pack_report: getpagesize() = %10" SZ_FMT
"\n"
424 "pack_report: core.packedGitWindowSize = %10" SZ_FMT
"\n"
425 "pack_report: core.packedGitLimit = %10" SZ_FMT
"\n",
426 (size_t) getpagesize(),
427 packed_git_window_size
,
430 "pack_report: pack_used_ctr = %10u\n"
431 "pack_report: pack_mmap_calls = %10u\n"
432 "pack_report: pack_open_windows = %10u / %10u\n"
433 "pack_report: pack_mapped = "
434 "%10" SZ_FMT
" / %10" SZ_FMT
"\n",
437 pack_open_windows
, peak_pack_open_windows
,
438 pack_mapped
, peak_pack_mapped
);
441 static int check_packed_git_idx(const char *path
, unsigned long *idx_size_
,
446 unsigned long idx_size
;
448 int fd
= open(path
, O_RDONLY
);
452 if (fstat(fd
, &st
)) {
456 idx_size
= st
.st_size
;
457 idx_map
= xmmap(NULL
, idx_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
462 *idx_size_
= idx_size
;
464 /* check index map */
465 if (idx_size
< 4*256 + 20 + 20)
466 return error("index file %s is too small", path
);
468 /* a future index format would start with this, as older git
469 * binaries would fail the non-monotonic index check below.
470 * give a nicer warning to the user if we can.
472 if (index
[0] == htonl(PACK_IDX_SIGNATURE
))
473 return error("index file %s is a newer version"
474 " and is not supported by this binary"
475 " (try upgrading GIT to a newer version)",
479 for (i
= 0; i
< 256; i
++) {
480 unsigned int n
= ntohl(index
[i
]);
482 return error("non-monotonic index %s", path
);
488 * - 256 index entries 4 bytes each
489 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
490 * - 20-byte SHA1 of the packfile
491 * - 20-byte SHA1 file checksum
493 if (idx_size
!= 4*256 + nr
* 24 + 20 + 20)
494 return error("wrong index file size in %s", path
);
499 static void scan_windows(struct packed_git
*p
,
500 struct packed_git
**lru_p
,
501 struct pack_window
**lru_w
,
502 struct pack_window
**lru_l
)
504 struct pack_window
*w
, *w_l
;
506 for (w_l
= NULL
, w
= p
->windows
; w
; w
= w
->next
) {
508 if (!*lru_w
|| w
->last_used
< (*lru_w
)->last_used
) {
518 static int unuse_one_window(struct packed_git
*current
)
520 struct packed_git
*p
, *lru_p
= NULL
;
521 struct pack_window
*lru_w
= NULL
, *lru_l
= NULL
;
524 scan_windows(current
, &lru_p
, &lru_w
, &lru_l
);
525 for (p
= packed_git
; p
; p
= p
->next
)
526 scan_windows(p
, &lru_p
, &lru_w
, &lru_l
);
528 munmap(lru_w
->base
, lru_w
->len
);
529 pack_mapped
-= lru_w
->len
;
531 lru_l
->next
= lru_w
->next
;
533 lru_p
->windows
= lru_w
->next
;
534 if (!lru_p
->windows
&& lru_p
!= current
) {
535 close(lru_p
->pack_fd
);
546 void release_pack_memory(size_t need
)
548 size_t cur
= pack_mapped
;
549 while (need
>= (cur
- pack_mapped
) && unuse_one_window(NULL
))
553 void unuse_pack(struct pack_window
**w_cursor
)
555 struct pack_window
*w
= *w_cursor
;
563 * Do not call this directly as this leaks p->pack_fd on error return;
564 * call open_packed_git() instead.
566 static int open_packed_git_1(struct packed_git
*p
)
569 struct pack_header hdr
;
570 unsigned char sha1
[20];
571 unsigned char *idx_sha1
;
574 p
->pack_fd
= open(p
->pack_name
, O_RDONLY
);
575 if (p
->pack_fd
< 0 || fstat(p
->pack_fd
, &st
))
578 /* If we created the struct before we had the pack we lack size. */
580 if (!S_ISREG(st
.st_mode
))
581 return error("packfile %s not a regular file", p
->pack_name
);
582 p
->pack_size
= st
.st_size
;
583 } else if (p
->pack_size
!= st
.st_size
)
584 return error("packfile %s size changed", p
->pack_name
);
587 /* We leave these file descriptors open with sliding mmap;
588 * there is no point keeping them open across exec(), though.
590 fd_flag
= fcntl(p
->pack_fd
, F_GETFD
, 0);
592 return error("cannot determine file descriptor flags");
593 fd_flag
|= FD_CLOEXEC
;
594 if (fcntl(p
->pack_fd
, F_SETFD
, fd_flag
) == -1)
595 return error("cannot set FD_CLOEXEC");
598 /* Verify we recognize this pack file format. */
599 if (read_in_full(p
->pack_fd
, &hdr
, sizeof(hdr
)) != sizeof(hdr
))
600 return error("file %s is far too short to be a packfile", p
->pack_name
);
601 if (hdr
.hdr_signature
!= htonl(PACK_SIGNATURE
))
602 return error("file %s is not a GIT packfile", p
->pack_name
);
603 if (!pack_version_ok(hdr
.hdr_version
))
604 return error("packfile %s is version %u and not supported"
605 " (try upgrading GIT to a newer version)",
606 p
->pack_name
, ntohl(hdr
.hdr_version
));
608 /* Verify the pack matches its index. */
609 if (num_packed_objects(p
) != ntohl(hdr
.hdr_entries
))
610 return error("packfile %s claims to have %u objects"
611 " while index size indicates %u objects",
612 p
->pack_name
, ntohl(hdr
.hdr_entries
),
613 num_packed_objects(p
));
614 if (lseek(p
->pack_fd
, p
->pack_size
- sizeof(sha1
), SEEK_SET
) == -1)
615 return error("end of packfile %s is unavailable", p
->pack_name
);
616 if (read_in_full(p
->pack_fd
, sha1
, sizeof(sha1
)) != sizeof(sha1
))
617 return error("packfile %s signature is unavailable", p
->pack_name
);
618 idx_sha1
= ((unsigned char *)p
->index_base
) + p
->index_size
- 40;
619 if (hashcmp(sha1
, idx_sha1
))
620 return error("packfile %s does not match index", p
->pack_name
);
624 static int open_packed_git(struct packed_git
*p
)
626 if (!open_packed_git_1(p
))
628 if (p
->pack_fd
!= -1) {
635 static int in_window(struct pack_window
*win
, unsigned long offset
)
637 /* We must promise at least 20 bytes (one hash) after the
638 * offset is available from this window, otherwise the offset
639 * is not actually in this window and a different window (which
640 * has that one hash excess) must be used. This is to support
641 * the object header and delta base parsing routines below.
643 off_t win_off
= win
->offset
;
644 return win_off
<= offset
645 && (offset
+ 20) <= (win_off
+ win
->len
);
648 unsigned char* use_pack(struct packed_git
*p
,
649 struct pack_window
**w_cursor
,
650 unsigned long offset
,
653 struct pack_window
*win
= *w_cursor
;
655 if (p
->pack_fd
== -1 && open_packed_git(p
))
656 die("packfile %s cannot be accessed", p
->pack_name
);
658 /* Since packfiles end in a hash of their content and its
659 * pointless to ask for an offset into the middle of that
660 * hash, and the in_window function above wouldn't match
661 * don't allow an offset too close to the end of the file.
663 if (offset
> (p
->pack_size
- 20))
664 die("offset beyond end of packfile (truncated pack?)");
666 if (!win
|| !in_window(win
, offset
)) {
669 for (win
= p
->windows
; win
; win
= win
->next
) {
670 if (in_window(win
, offset
))
674 size_t window_align
= packed_git_window_size
/ 2;
675 win
= xcalloc(1, sizeof(*win
));
676 win
->offset
= (offset
/ window_align
) * window_align
;
677 win
->len
= p
->pack_size
- win
->offset
;
678 if (win
->len
> packed_git_window_size
)
679 win
->len
= packed_git_window_size
;
680 pack_mapped
+= win
->len
;
681 while (packed_git_limit
< pack_mapped
682 && unuse_one_window(p
))
684 win
->base
= xmmap(NULL
, win
->len
,
685 PROT_READ
, MAP_PRIVATE
,
686 p
->pack_fd
, win
->offset
);
687 if (win
->base
== MAP_FAILED
)
688 die("packfile %s cannot be mapped: %s",
693 if (pack_mapped
> peak_pack_mapped
)
694 peak_pack_mapped
= pack_mapped
;
695 if (pack_open_windows
> peak_pack_open_windows
)
696 peak_pack_open_windows
= pack_open_windows
;
697 win
->next
= p
->windows
;
701 if (win
!= *w_cursor
) {
702 win
->last_used
= pack_used_ctr
++;
706 offset
-= win
->offset
;
708 *left
= win
->len
- offset
;
709 return win
->base
+ offset
;
712 struct packed_git
*add_packed_git(char *path
, int path_len
, int local
)
715 struct packed_git
*p
;
716 unsigned long idx_size
;
718 unsigned char sha1
[20];
720 if (check_packed_git_idx(path
, &idx_size
, &idx_map
))
723 /* do we have a corresponding .pack file? */
724 strcpy(path
+ path_len
- 4, ".pack");
725 if (stat(path
, &st
) || !S_ISREG(st
.st_mode
)) {
726 munmap(idx_map
, idx_size
);
729 /* ok, it looks sane as far as we can check without
730 * actually mapping the pack file.
732 p
= xmalloc(sizeof(*p
) + path_len
+ 2);
733 strcpy(p
->pack_name
, path
);
734 p
->index_size
= idx_size
;
735 p
->pack_size
= st
.st_size
;
736 p
->index_base
= idx_map
;
740 p
->pack_local
= local
;
741 if ((path_len
> 44) && !get_sha1_hex(path
+ path_len
- 44, sha1
))
742 hashcpy(p
->sha1
, sha1
);
746 struct packed_git
*parse_pack_index(unsigned char *sha1
)
748 char *path
= sha1_pack_index_name(sha1
);
749 return parse_pack_index_file(sha1
, path
);
752 struct packed_git
*parse_pack_index_file(const unsigned char *sha1
, char *idx_path
)
754 struct packed_git
*p
;
755 unsigned long idx_size
;
759 if (check_packed_git_idx(idx_path
, &idx_size
, &idx_map
))
762 path
= sha1_pack_name(sha1
);
764 p
= xmalloc(sizeof(*p
) + strlen(path
) + 2);
765 strcpy(p
->pack_name
, path
);
766 p
->index_size
= idx_size
;
768 p
->index_base
= idx_map
;
772 hashcpy(p
->sha1
, sha1
);
776 void install_packed_git(struct packed_git
*pack
)
778 pack
->next
= packed_git
;
782 static void prepare_packed_git_one(char *objdir
, int local
)
789 sprintf(path
, "%s/pack", objdir
);
794 error("unable to open object pack directory: %s: %s",
795 path
, strerror(errno
));
799 while ((de
= readdir(dir
)) != NULL
) {
800 int namelen
= strlen(de
->d_name
);
801 struct packed_git
*p
;
803 if (!has_extension(de
->d_name
, ".idx"))
806 /* Don't reopen a pack we already have. */
807 strcpy(path
+ len
, de
->d_name
);
808 for (p
= packed_git
; p
; p
= p
->next
) {
809 if (!memcmp(path
, p
->pack_name
, len
+ namelen
- 4))
814 /* See if it really is a valid .idx file with corresponding
815 * .pack file that we can map.
817 p
= add_packed_git(path
, len
+ namelen
, local
);
820 install_packed_git(p
);
825 static int prepare_packed_git_run_once
= 0;
826 void prepare_packed_git(void)
828 struct alternate_object_database
*alt
;
830 if (prepare_packed_git_run_once
)
832 prepare_packed_git_one(get_object_directory(), 1);
834 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
836 prepare_packed_git_one(alt
->base
, 0);
839 prepare_packed_git_run_once
= 1;
842 void reprepare_packed_git(void)
844 prepare_packed_git_run_once
= 0;
845 prepare_packed_git();
848 int check_sha1_signature(const unsigned char *sha1
, void *map
, unsigned long size
, const char *type
)
850 unsigned char real_sha1
[20];
851 hash_sha1_file(map
, size
, type
, real_sha1
);
852 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
855 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
860 char *filename
= find_sha1_file(sha1
, &st
);
866 fd
= open(filename
, O_RDONLY
| sha1_file_open_flag
);
868 /* See if it works without O_NOATIME */
869 switch (sha1_file_open_flag
) {
871 fd
= open(filename
, O_RDONLY
);
879 /* If it failed once, it will probably fail again.
880 * Stop using O_NOATIME
882 sha1_file_open_flag
= 0;
884 map
= xmmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
890 int legacy_loose_object(unsigned char *map
)
895 * Is it a zlib-compressed buffer? If so, the first byte
896 * must be 0x78 (15-bit window size, deflated), and the
897 * first 16-bit word is evenly divisible by 31
899 word
= (map
[0] << 8) + map
[1];
900 if (map
[0] == 0x78 && !(word
% 31))
906 unsigned long unpack_object_header_gently(const unsigned char *buf
, unsigned long len
, enum object_type
*type
, unsigned long *sizep
)
911 unsigned long used
= 0;
914 *type
= (c
>> 4) & 7;
920 if (sizeof(long) * 8 <= shift
)
923 size
+= (c
& 0x7f) << shift
;
930 static int unpack_sha1_header(z_stream
*stream
, unsigned char *map
, unsigned long mapsize
, void *buffer
, unsigned long bufsiz
)
932 unsigned long size
, used
;
933 static const char valid_loose_object_type
[8] = {
935 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
936 0, /* "delta" and others are invalid in a loose object */
938 enum object_type type
;
940 /* Get the data stream */
941 memset(stream
, 0, sizeof(*stream
));
942 stream
->next_in
= map
;
943 stream
->avail_in
= mapsize
;
944 stream
->next_out
= buffer
;
945 stream
->avail_out
= bufsiz
;
947 if (legacy_loose_object(map
)) {
949 return inflate(stream
, 0);
952 used
= unpack_object_header_gently(map
, mapsize
, &type
, &size
);
953 if (!used
|| !valid_loose_object_type
[type
])
958 /* Set up the stream for the rest.. */
959 stream
->next_in
= map
;
960 stream
->avail_in
= mapsize
;
963 /* And generate the fake traditional header */
964 stream
->total_out
= 1 + snprintf(buffer
, bufsiz
, "%s %lu",
965 type_names
[type
], size
);
969 static void *unpack_sha1_rest(z_stream
*stream
, void *buffer
, unsigned long size
)
971 int bytes
= strlen(buffer
) + 1;
972 unsigned char *buf
= xmalloc(1+size
);
975 n
= stream
->total_out
- bytes
;
978 memcpy(buf
, (char *) buffer
+ bytes
, n
);
981 stream
->next_out
= buf
+ bytes
;
982 stream
->avail_out
= size
- bytes
;
983 while (inflate(stream
, Z_FINISH
) == Z_OK
)
992 * We used to just use "sscanf()", but that's actually way
993 * too permissive for what we want to check. So do an anal
994 * object header parse by hand.
996 static int parse_sha1_header(char *hdr
, char *type
, unsigned long *sizep
)
1002 * The type can be at most ten bytes (including the
1003 * terminating '\0' that we add), and is followed by
1018 * The length must follow immediately, and be in canonical
1019 * decimal format (ie "010" is not valid).
1021 size
= *hdr
++ - '0';
1026 unsigned long c
= *hdr
- '0';
1030 size
= size
* 10 + c
;
1036 * The length must be followed by a zero byte
1038 return *hdr
? -1 : 0;
1041 void * unpack_sha1_file(void *map
, unsigned long mapsize
, char *type
, unsigned long *size
)
1047 ret
= unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
));
1048 if (ret
< Z_OK
|| parse_sha1_header(hdr
, type
, size
) < 0)
1051 return unpack_sha1_rest(&stream
, hdr
, *size
);
1054 static unsigned long get_delta_base(struct packed_git
*p
,
1055 struct pack_window
**w_curs
,
1056 unsigned long offset
,
1057 enum object_type kind
,
1058 unsigned long delta_obj_offset
,
1059 unsigned long *base_obj_offset
)
1061 unsigned char *base_info
= use_pack(p
, w_curs
, offset
, NULL
);
1062 unsigned long base_offset
;
1064 /* use_pack() assured us we have [base_info, base_info + 20)
1065 * as a range that we can look at without walking off the
1066 * end of the mapped window. Its actually the hash size
1067 * that is assured. An OFS_DELTA longer than the hash size
1068 * is stupid, as then a REF_DELTA would be smaller to store.
1070 if (kind
== OBJ_OFS_DELTA
) {
1072 unsigned char c
= base_info
[used
++];
1073 base_offset
= c
& 127;
1076 if (!base_offset
|| base_offset
& ~(~0UL >> 7))
1077 die("offset value overflow for delta base object");
1078 c
= base_info
[used
++];
1079 base_offset
= (base_offset
<< 7) + (c
& 127);
1081 base_offset
= delta_obj_offset
- base_offset
;
1082 if (base_offset
>= delta_obj_offset
)
1083 die("delta base offset out of bound");
1085 } else if (kind
== OBJ_REF_DELTA
) {
1086 /* The base entry _must_ be in the same pack */
1087 base_offset
= find_pack_entry_one(base_info
, p
);
1089 die("failed to find delta-pack base object %s",
1090 sha1_to_hex(base_info
));
1093 die("I am totally screwed");
1094 *base_obj_offset
= base_offset
;
1098 /* forward declaration for a mutually recursive function */
1099 static int packed_object_info(struct packed_git
*p
, unsigned long offset
,
1100 char *type
, unsigned long *sizep
);
1102 static int packed_delta_info(struct packed_git
*p
,
1103 struct pack_window
**w_curs
,
1104 unsigned long offset
,
1105 enum object_type kind
,
1106 unsigned long obj_offset
,
1108 unsigned long *sizep
)
1110 unsigned long base_offset
;
1112 offset
= get_delta_base(p
, w_curs
, offset
, kind
,
1113 obj_offset
, &base_offset
);
1115 /* We choose to only get the type of the base object and
1116 * ignore potentially corrupt pack file that expects the delta
1117 * based on a base with a wrong size. This saves tons of
1120 if (packed_object_info(p
, base_offset
, type
, NULL
))
1121 die("cannot get info for delta-pack base");
1124 const unsigned char *data
;
1125 unsigned char delta_head
[20], *in
;
1126 unsigned long result_size
;
1130 memset(&stream
, 0, sizeof(stream
));
1131 stream
.next_out
= delta_head
;
1132 stream
.avail_out
= sizeof(delta_head
);
1134 inflateInit(&stream
);
1136 in
= use_pack(p
, w_curs
, offset
, &stream
.avail_in
);
1137 stream
.next_in
= in
;
1138 st
= inflate(&stream
, Z_FINISH
);
1139 offset
+= stream
.next_in
- in
;
1140 } while ((st
== Z_OK
|| st
== Z_BUF_ERROR
)
1141 && stream
.total_out
< sizeof(delta_head
));
1142 inflateEnd(&stream
);
1143 if ((st
!= Z_STREAM_END
) &&
1144 stream
.total_out
!= sizeof(delta_head
))
1145 die("delta data unpack-initial failed");
1147 /* Examine the initial part of the delta to figure out
1152 /* ignore base size */
1153 get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
1155 /* Read the result size */
1156 result_size
= get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
1157 *sizep
= result_size
;
1162 static unsigned long unpack_object_header(struct packed_git
*p
,
1163 struct pack_window
**w_curs
,
1164 unsigned long offset
,
1165 enum object_type
*type
,
1166 unsigned long *sizep
)
1168 unsigned char *base
;
1172 /* use_pack() assures us we have [base, base + 20) available
1173 * as a range that we can look at at. (Its actually the hash
1174 * size that is assured.) With our object header encoding
1175 * the maximum deflated object size is 2^137, which is just
1176 * insane, so we know won't exceed what we have been given.
1178 base
= use_pack(p
, w_curs
, offset
, &left
);
1179 used
= unpack_object_header_gently(base
, left
, type
, sizep
);
1181 die("object offset outside of pack file");
1183 return offset
+ used
;
1186 void packed_object_info_detail(struct packed_git
*p
,
1187 unsigned long offset
,
1189 unsigned long *size
,
1190 unsigned long *store_size
,
1191 unsigned int *delta_chain_length
,
1192 unsigned char *base_sha1
)
1194 struct pack_window
*w_curs
= NULL
;
1195 unsigned long obj_offset
, val
;
1196 unsigned char *next_sha1
;
1197 enum object_type kind
;
1199 *delta_chain_length
= 0;
1200 obj_offset
= offset
;
1201 offset
= unpack_object_header(p
, &w_curs
, offset
, &kind
, size
);
1206 die("pack %s contains unknown object type %d",
1207 p
->pack_name
, kind
);
1212 strcpy(type
, type_names
[kind
]);
1213 *store_size
= 0; /* notyet */
1214 unuse_pack(&w_curs
);
1217 get_delta_base(p
, &w_curs
, offset
, kind
,
1218 obj_offset
, &offset
);
1219 if (*delta_chain_length
== 0) {
1220 /* TODO: find base_sha1 as pointed by offset */
1224 next_sha1
= use_pack(p
, &w_curs
, offset
, NULL
);
1225 if (*delta_chain_length
== 0)
1226 hashcpy(base_sha1
, next_sha1
);
1227 offset
= find_pack_entry_one(next_sha1
, p
);
1230 obj_offset
= offset
;
1231 offset
= unpack_object_header(p
, &w_curs
, offset
, &kind
, &val
);
1232 (*delta_chain_length
)++;
1236 static int packed_object_info(struct packed_git
*p
, unsigned long offset
,
1237 char *type
, unsigned long *sizep
)
1239 struct pack_window
*w_curs
= NULL
;
1240 unsigned long size
, obj_offset
= offset
;
1241 enum object_type kind
;
1244 offset
= unpack_object_header(p
, &w_curs
, offset
, &kind
, &size
);
1249 r
= packed_delta_info(p
, &w_curs
, offset
, kind
,
1250 obj_offset
, type
, sizep
);
1251 unuse_pack(&w_curs
);
1257 strcpy(type
, type_names
[kind
]);
1258 unuse_pack(&w_curs
);
1261 die("pack %s contains unknown object type %d",
1262 p
->pack_name
, kind
);
1269 static void *unpack_compressed_entry(struct packed_git
*p
,
1270 struct pack_window
**w_curs
,
1271 unsigned long offset
,
1276 unsigned char *buffer
, *in
;
1278 buffer
= xmalloc(size
+ 1);
1280 memset(&stream
, 0, sizeof(stream
));
1281 stream
.next_out
= buffer
;
1282 stream
.avail_out
= size
;
1284 inflateInit(&stream
);
1286 in
= use_pack(p
, w_curs
, offset
, &stream
.avail_in
);
1287 stream
.next_in
= in
;
1288 st
= inflate(&stream
, Z_FINISH
);
1289 offset
+= stream
.next_in
- in
;
1290 } while (st
== Z_OK
|| st
== Z_BUF_ERROR
);
1291 inflateEnd(&stream
);
1292 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= size
) {
1300 static void *unpack_delta_entry(struct packed_git
*p
,
1301 struct pack_window
**w_curs
,
1302 unsigned long offset
,
1303 unsigned long delta_size
,
1304 enum object_type kind
,
1305 unsigned long obj_offset
,
1307 unsigned long *sizep
)
1309 void *delta_data
, *result
, *base
;
1310 unsigned long result_size
, base_size
, base_offset
;
1312 offset
= get_delta_base(p
, w_curs
, offset
, kind
,
1313 obj_offset
, &base_offset
);
1314 base
= unpack_entry(p
, base_offset
, type
, &base_size
);
1316 die("failed to read delta base object at %lu from %s",
1317 base_offset
, p
->pack_name
);
1319 delta_data
= unpack_compressed_entry(p
, w_curs
, offset
, delta_size
);
1320 result
= patch_delta(base
, base_size
,
1321 delta_data
, delta_size
,
1324 die("failed to apply delta");
1327 *sizep
= result_size
;
1331 void *unpack_entry(struct packed_git
*p
, unsigned long offset
,
1332 char *type
, unsigned long *sizep
)
1334 struct pack_window
*w_curs
= NULL
;
1335 unsigned long size
, obj_offset
= offset
;
1336 enum object_type kind
;
1339 offset
= unpack_object_header(p
, &w_curs
, offset
, &kind
, &size
);
1343 retval
= unpack_delta_entry(p
, &w_curs
, offset
, size
,
1344 kind
, obj_offset
, type
, sizep
);
1350 strcpy(type
, type_names
[kind
]);
1352 retval
= unpack_compressed_entry(p
, &w_curs
, offset
, size
);
1355 die("unknown object type %i in %s", kind
, p
->pack_name
);
1357 unuse_pack(&w_curs
);
1361 int num_packed_objects(const struct packed_git
*p
)
1363 /* See check_packed_git_idx() */
1364 return (p
->index_size
- 20 - 20 - 4*256) / 24;
1367 int nth_packed_object_sha1(const struct packed_git
*p
, int n
,
1368 unsigned char* sha1
)
1370 void *index
= p
->index_base
+ 256;
1371 if (n
< 0 || num_packed_objects(p
) <= n
)
1373 hashcpy(sha1
, (unsigned char *) index
+ (24 * n
) + 4);
1377 unsigned long find_pack_entry_one(const unsigned char *sha1
,
1378 struct packed_git
*p
)
1380 uint32_t *level1_ofs
= p
->index_base
;
1381 int hi
= ntohl(level1_ofs
[*sha1
]);
1382 int lo
= ((*sha1
== 0x0) ? 0 : ntohl(level1_ofs
[*sha1
- 1]));
1383 void *index
= p
->index_base
+ 256;
1386 int mi
= (lo
+ hi
) / 2;
1387 int cmp
= hashcmp((unsigned char *)index
+ (24 * mi
) + 4, sha1
);
1389 return ntohl(*((uint32_t *)((char *)index
+ (24 * mi
))));
1398 static int matches_pack_name(struct packed_git
*p
, const char *ig
)
1400 const char *last_c
, *c
;
1402 if (!strcmp(p
->pack_name
, ig
))
1405 for (c
= p
->pack_name
, last_c
= c
; *c
;)
1410 if (!strcmp(last_c
, ig
))
1416 static int find_pack_entry(const unsigned char *sha1
, struct pack_entry
*e
, const char **ignore_packed
)
1418 struct packed_git
*p
;
1419 unsigned long offset
;
1421 prepare_packed_git();
1423 for (p
= packed_git
; p
; p
= p
->next
) {
1424 if (ignore_packed
) {
1426 for (ig
= ignore_packed
; *ig
; ig
++)
1427 if (!matches_pack_name(p
, *ig
))
1432 offset
= find_pack_entry_one(sha1
, p
);
1435 * We are about to tell the caller where they can
1436 * locate the requested object. We better make
1437 * sure the packfile is still here and can be
1438 * accessed before supplying that answer, as
1439 * it may have been deleted since the index
1442 if (p
->pack_fd
== -1 && open_packed_git(p
)) {
1443 error("packfile %s cannot be accessed", p
->pack_name
);
1448 hashcpy(e
->sha1
, sha1
);
1455 struct packed_git
*find_sha1_pack(const unsigned char *sha1
,
1456 struct packed_git
*packs
)
1458 struct packed_git
*p
;
1460 for (p
= packs
; p
; p
= p
->next
) {
1461 if (find_pack_entry_one(sha1
, p
))
1468 static int sha1_loose_object_info(const unsigned char *sha1
, char *type
, unsigned long *sizep
)
1471 unsigned long mapsize
, size
;
1476 map
= map_sha1_file(sha1
, &mapsize
);
1478 return error("unable to find %s", sha1_to_hex(sha1
));
1479 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1480 status
= error("unable to unpack %s header",
1482 if (parse_sha1_header(hdr
, type
, &size
) < 0)
1483 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1489 inflateEnd(&stream
);
1490 munmap(map
, mapsize
);
1494 int sha1_object_info(const unsigned char *sha1
, char *type
, unsigned long *sizep
)
1496 struct pack_entry e
;
1498 if (!find_pack_entry(sha1
, &e
, NULL
)) {
1499 reprepare_packed_git();
1500 if (!find_pack_entry(sha1
, &e
, NULL
))
1501 return sha1_loose_object_info(sha1
, type
, sizep
);
1503 return packed_object_info(e
.p
, e
.offset
, type
, sizep
);
1506 static void *read_packed_sha1(const unsigned char *sha1
, char *type
, unsigned long *size
)
1508 struct pack_entry e
;
1510 if (!find_pack_entry(sha1
, &e
, NULL
))
1513 return unpack_entry(e
.p
, e
.offset
, type
, size
);
1517 * This is meant to hold a *small* number of objects that you would
1518 * want read_sha1_file() to be able to return, but yet you do not want
1519 * to write them into the object store (e.g. a browse-only
1522 static struct cached_object
{
1523 unsigned char sha1
[20];
1528 static int cached_object_nr
, cached_object_alloc
;
1530 static struct cached_object
*find_cached_object(const unsigned char *sha1
)
1533 struct cached_object
*co
= cached_objects
;
1535 for (i
= 0; i
< cached_object_nr
; i
++, co
++) {
1536 if (!hashcmp(co
->sha1
, sha1
))
1542 int pretend_sha1_file(void *buf
, unsigned long len
, const char *type
, unsigned char *sha1
)
1544 struct cached_object
*co
;
1546 hash_sha1_file(buf
, len
, type
, sha1
);
1547 if (has_sha1_file(sha1
) || find_cached_object(sha1
))
1549 if (cached_object_alloc
<= cached_object_nr
) {
1550 cached_object_alloc
= alloc_nr(cached_object_alloc
);
1551 cached_objects
= xrealloc(cached_objects
,
1552 sizeof(*cached_objects
) *
1553 cached_object_alloc
);
1555 co
= &cached_objects
[cached_object_nr
++];
1557 co
->type
= strdup(type
);
1558 co
->buf
= xmalloc(len
);
1559 memcpy(co
->buf
, buf
, len
);
1560 hashcpy(co
->sha1
, sha1
);
1564 void *read_sha1_file(const unsigned char *sha1
, char *type
, unsigned long *size
)
1566 unsigned long mapsize
;
1568 struct cached_object
*co
;
1570 co
= find_cached_object(sha1
);
1572 buf
= xmalloc(co
->size
+ 1);
1573 memcpy(buf
, co
->buf
, co
->size
);
1574 ((char*)buf
)[co
->size
] = 0;
1575 strcpy(type
, co
->type
);
1580 buf
= read_packed_sha1(sha1
, type
, size
);
1583 map
= map_sha1_file(sha1
, &mapsize
);
1585 buf
= unpack_sha1_file(map
, mapsize
, type
, size
);
1586 munmap(map
, mapsize
);
1589 reprepare_packed_git();
1590 return read_packed_sha1(sha1
, type
, size
);
1593 void *read_object_with_reference(const unsigned char *sha1
,
1594 const char *required_type
,
1595 unsigned long *size
,
1596 unsigned char *actual_sha1_return
)
1600 unsigned long isize
;
1601 unsigned char actual_sha1
[20];
1603 hashcpy(actual_sha1
, sha1
);
1605 int ref_length
= -1;
1606 const char *ref_type
= NULL
;
1608 buffer
= read_sha1_file(actual_sha1
, type
, &isize
);
1611 if (!strcmp(type
, required_type
)) {
1613 if (actual_sha1_return
)
1614 hashcpy(actual_sha1_return
, actual_sha1
);
1617 /* Handle references */
1618 else if (!strcmp(type
, commit_type
))
1620 else if (!strcmp(type
, tag_type
))
1621 ref_type
= "object ";
1626 ref_length
= strlen(ref_type
);
1628 if (memcmp(buffer
, ref_type
, ref_length
) ||
1629 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1634 /* Now we have the ID of the referred-to object in
1635 * actual_sha1. Check again. */
1639 static void write_sha1_file_prepare(void *buf
, unsigned long len
,
1640 const char *type
, unsigned char *sha1
,
1641 unsigned char *hdr
, int *hdrlen
)
1645 /* Generate the header */
1646 *hdrlen
= sprintf((char *)hdr
, "%s %lu", type
, len
)+1;
1650 SHA1_Update(&c
, hdr
, *hdrlen
);
1651 SHA1_Update(&c
, buf
, len
);
1652 SHA1_Final(sha1
, &c
);
1656 * Link the tempfile to the final place, possibly creating the
1657 * last directory level as you do so.
1659 * Returns the errno on failure, 0 on success.
1661 static int link_temp_to_file(const char *tmpfile
, const char *filename
)
1666 if (!link(tmpfile
, filename
))
1670 * Try to mkdir the last path component if that failed.
1672 * Re-try the "link()" regardless of whether the mkdir
1673 * succeeds, since a race might mean that somebody
1677 dir
= strrchr(filename
, '/');
1680 if (!mkdir(filename
, 0777) && adjust_shared_perm(filename
)) {
1685 if (!link(tmpfile
, filename
))
1693 * Move the just written object into its final resting place
1695 int move_temp_to_file(const char *tmpfile
, const char *filename
)
1697 int ret
= link_temp_to_file(tmpfile
, filename
);
1700 * Coda hack - coda doesn't like cross-directory links,
1701 * so we fall back to a rename, which will mean that it
1702 * won't be able to check collisions, but that's not a
1705 * The same holds for FAT formatted media.
1707 * When this succeeds, we just return 0. We have nothing
1710 if (ret
&& ret
!= EEXIST
) {
1711 if (!rename(tmpfile
, filename
))
1717 if (ret
!= EEXIST
) {
1718 return error("unable to write sha1 filename %s: %s\n", filename
, strerror(ret
));
1720 /* FIXME!!! Collision check here ? */
1726 static int write_buffer(int fd
, const void *buf
, size_t len
)
1728 if (write_in_full(fd
, buf
, len
) < 0)
1729 return error("file write error (%s)", strerror(errno
));
1733 static int write_binary_header(unsigned char *hdr
, enum object_type type
, unsigned long len
)
1738 c
= (type
<< 4) | (len
& 15);
1751 static void setup_object_header(z_stream
*stream
, const char *type
, unsigned long len
)
1755 if (use_legacy_headers
) {
1756 while (deflate(stream
, 0) == Z_OK
)
1760 if (!strcmp(type
, blob_type
))
1761 obj_type
= OBJ_BLOB
;
1762 else if (!strcmp(type
, tree_type
))
1763 obj_type
= OBJ_TREE
;
1764 else if (!strcmp(type
, commit_type
))
1765 obj_type
= OBJ_COMMIT
;
1766 else if (!strcmp(type
, tag_type
))
1769 die("trying to generate bogus object of type '%s'", type
);
1770 hdr
= write_binary_header(stream
->next_out
, obj_type
, len
);
1771 stream
->total_out
= hdr
;
1772 stream
->next_out
+= hdr
;
1773 stream
->avail_out
-= hdr
;
1776 int hash_sha1_file(void *buf
, unsigned long len
, const char *type
,
1777 unsigned char *sha1
)
1779 unsigned char hdr
[50];
1781 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1785 int write_sha1_file(void *buf
, unsigned long len
, const char *type
, unsigned char *returnsha1
)
1788 unsigned char *compressed
;
1790 unsigned char sha1
[20];
1792 static char tmpfile
[PATH_MAX
];
1793 unsigned char hdr
[50];
1796 /* Normally if we have it in the pack then we do not bother writing
1797 * it out into .git/objects/??/?{38} file.
1799 write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1800 filename
= sha1_file_name(sha1
);
1802 hashcpy(returnsha1
, sha1
);
1803 if (has_sha1_file(sha1
))
1805 fd
= open(filename
, O_RDONLY
);
1808 * FIXME!!! We might do collision checking here, but we'd
1809 * need to uncompress the old file and check it. Later.
1815 if (errno
!= ENOENT
) {
1816 return error("sha1 file %s: %s\n", filename
, strerror(errno
));
1819 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1821 fd
= mkstemp(tmpfile
);
1824 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
1826 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile
, strerror(errno
));
1830 memset(&stream
, 0, sizeof(stream
));
1831 deflateInit(&stream
, zlib_compression_level
);
1832 size
= 8 + deflateBound(&stream
, len
+hdrlen
);
1833 compressed
= xmalloc(size
);
1836 stream
.next_out
= compressed
;
1837 stream
.avail_out
= size
;
1839 /* First header.. */
1840 stream
.next_in
= hdr
;
1841 stream
.avail_in
= hdrlen
;
1842 setup_object_header(&stream
, type
, len
);
1844 /* Then the data itself.. */
1845 stream
.next_in
= buf
;
1846 stream
.avail_in
= len
;
1847 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1849 deflateEnd(&stream
);
1850 size
= stream
.total_out
;
1852 if (write_buffer(fd
, compressed
, size
) < 0)
1853 die("unable to write sha1 file");
1858 return move_temp_to_file(tmpfile
, filename
);
1862 * We need to unpack and recompress the object for writing
1863 * it out to a different file.
1865 static void *repack_object(const unsigned char *sha1
, unsigned long *objsize
)
1869 unsigned char *unpacked
;
1876 /* need to unpack and recompress it by itself */
1877 unpacked
= read_packed_sha1(sha1
, type
, &len
);
1879 error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1881 hdrlen
= sprintf(hdr
, "%s %lu", type
, len
) + 1;
1884 memset(&stream
, 0, sizeof(stream
));
1885 deflateInit(&stream
, zlib_compression_level
);
1886 size
= deflateBound(&stream
, len
+ hdrlen
);
1887 buf
= xmalloc(size
);
1890 stream
.next_out
= buf
;
1891 stream
.avail_out
= size
;
1893 /* First header.. */
1894 stream
.next_in
= (void *)hdr
;
1895 stream
.avail_in
= hdrlen
;
1896 while (deflate(&stream
, 0) == Z_OK
)
1899 /* Then the data itself.. */
1900 stream
.next_in
= unpacked
;
1901 stream
.avail_in
= len
;
1902 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1904 deflateEnd(&stream
);
1907 *objsize
= stream
.total_out
;
1911 int write_sha1_to_fd(int fd
, const unsigned char *sha1
)
1914 unsigned long objsize
;
1915 void *buf
= map_sha1_file(sha1
, &objsize
);
1918 retval
= write_buffer(fd
, buf
, objsize
);
1919 munmap(buf
, objsize
);
1923 buf
= repack_object(sha1
, &objsize
);
1924 retval
= write_buffer(fd
, buf
, objsize
);
1929 int write_sha1_from_fd(const unsigned char *sha1
, int fd
, char *buffer
,
1930 size_t bufsize
, size_t *bufposn
)
1932 char tmpfile
[PATH_MAX
];
1935 unsigned char real_sha1
[20];
1936 unsigned char discard
[4096];
1940 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1942 local
= mkstemp(tmpfile
);
1945 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
1947 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile
, strerror(errno
));
1950 memset(&stream
, 0, sizeof(stream
));
1952 inflateInit(&stream
);
1959 stream
.avail_in
= *bufposn
;
1960 stream
.next_in
= (unsigned char *) buffer
;
1962 stream
.next_out
= discard
;
1963 stream
.avail_out
= sizeof(discard
);
1964 ret
= inflate(&stream
, Z_SYNC_FLUSH
);
1965 SHA1_Update(&c
, discard
, sizeof(discard
) -
1967 } while (stream
.avail_in
&& ret
== Z_OK
);
1968 if (write_buffer(local
, buffer
, *bufposn
- stream
.avail_in
) < 0)
1969 die("unable to write sha1 file");
1970 memmove(buffer
, buffer
+ *bufposn
- stream
.avail_in
,
1972 *bufposn
= stream
.avail_in
;
1976 size
= xread(fd
, buffer
+ *bufposn
, bufsize
- *bufposn
);
1981 return error("Connection closed?");
1982 perror("Reading from connection");
1987 inflateEnd(&stream
);
1990 SHA1_Final(real_sha1
, &c
);
1991 if (ret
!= Z_STREAM_END
) {
1993 return error("File %s corrupted", sha1_to_hex(sha1
));
1995 if (hashcmp(sha1
, real_sha1
)) {
1997 return error("File %s has bad hash", sha1_to_hex(sha1
));
2000 return move_temp_to_file(tmpfile
, sha1_file_name(sha1
));
2003 int has_pack_index(const unsigned char *sha1
)
2006 if (stat(sha1_pack_index_name(sha1
), &st
))
2011 int has_pack_file(const unsigned char *sha1
)
2014 if (stat(sha1_pack_name(sha1
), &st
))
2019 int has_sha1_pack(const unsigned char *sha1
, const char **ignore_packed
)
2021 struct pack_entry e
;
2022 return find_pack_entry(sha1
, &e
, ignore_packed
);
2025 int has_sha1_file(const unsigned char *sha1
)
2028 struct pack_entry e
;
2030 if (find_pack_entry(sha1
, &e
, NULL
))
2032 return find_sha1_file(sha1
, &st
) ? 1 : 0;
2036 * reads from fd as long as possible into a supplied buffer of size bytes.
2037 * If necessary the buffer's size is increased using realloc()
2039 * returns 0 if anything went fine and -1 otherwise
2041 * NOTE: both buf and size may change, but even when -1 is returned
2042 * you still have to free() it yourself.
2044 int read_pipe(int fd
, char** return_buf
, unsigned long* return_size
)
2046 char* buf
= *return_buf
;
2047 unsigned long size
= *return_size
;
2049 unsigned long off
= 0;
2052 iret
= xread(fd
, buf
+ off
, size
- off
);
2057 buf
= xrealloc(buf
, size
);
2070 int index_pipe(unsigned char *sha1
, int fd
, const char *type
, int write_object
)
2072 unsigned long size
= 4096;
2073 char *buf
= xmalloc(size
);
2076 if (read_pipe(fd
, &buf
, &size
)) {
2084 ret
= write_sha1_file(buf
, size
, type
, sha1
);
2086 ret
= hash_sha1_file(buf
, size
, type
, sha1
);
2091 int index_fd(unsigned char *sha1
, int fd
, struct stat
*st
, int write_object
, const char *type
)
2093 unsigned long size
= st
->st_size
;
2095 int ret
, re_allocated
= 0;
2099 buf
= xmmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
2106 * Convert blobs to git internal format
2108 if (!strcmp(type
, blob_type
)) {
2109 unsigned long nsize
= size
;
2111 if (convert_to_git(NULL
, &nbuf
, &nsize
)) {
2121 ret
= write_sha1_file(buf
, size
, type
, sha1
);
2123 ret
= hash_sha1_file(buf
, size
, type
, sha1
);
2133 int index_path(unsigned char *sha1
, const char *path
, struct stat
*st
, int write_object
)
2138 switch (st
->st_mode
& S_IFMT
) {
2140 fd
= open(path
, O_RDONLY
);
2142 return error("open(\"%s\"): %s", path
,
2144 if (index_fd(sha1
, fd
, st
, write_object
, NULL
) < 0)
2145 return error("%s: failed to insert into database",
2149 target
= xmalloc(st
->st_size
+1);
2150 if (readlink(path
, target
, st
->st_size
+1) != st
->st_size
) {
2151 char *errstr
= strerror(errno
);
2153 return error("readlink(\"%s\"): %s", path
,
2157 hash_sha1_file(target
, st
->st_size
, blob_type
, sha1
);
2158 else if (write_sha1_file(target
, st
->st_size
, blob_type
, sha1
))
2159 return error("%s: failed to insert into database",
2164 return error("%s: unsupported file type", path
);
2169 int read_pack_header(int fd
, struct pack_header
*header
)
2171 char *c
= (char*)header
;
2172 ssize_t remaining
= sizeof(struct pack_header
);
2174 ssize_t r
= xread(fd
, c
, remaining
);
2176 /* "eof before pack header was fully read" */
2177 return PH_ERROR_EOF
;
2180 } while (remaining
> 0);
2181 if (header
->hdr_signature
!= htonl(PACK_SIGNATURE
))
2182 /* "protocol error (pack signature mismatch detected)" */
2183 return PH_ERROR_PACK_SIGNATURE
;
2184 if (!pack_version_ok(header
->hdr_version
))
2185 /* "protocol error (pack version unsupported)" */
2186 return PH_ERROR_PROTOCOL
;