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
25 const unsigned char null_sha1
[20];
27 static unsigned int sha1_file_open_flag
= O_NOATIME
;
29 signed char hexval_table
[256] = {
30 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
31 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
32 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
33 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
34 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
35 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
36 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
37 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
38 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
39 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
40 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
41 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
42 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
43 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
44 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
45 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
46 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
47 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
48 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
49 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
50 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
51 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
52 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
53 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
54 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
55 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
56 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
57 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
58 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
59 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
60 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
61 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
64 int get_sha1_hex(const char *hex
, unsigned char *sha1
)
67 for (i
= 0; i
< 20; i
++) {
68 unsigned int val
= (hexval(hex
[0]) << 4) | hexval(hex
[1]);
77 int safe_create_leading_directories(char *path
)
86 pos
= strchr(pos
, '/');
90 if (!stat(path
, &st
)) {
92 if (!S_ISDIR(st
.st_mode
)) {
97 else if (mkdir(path
, 0777)) {
101 else if (adjust_shared_perm(path
)) {
110 char * sha1_to_hex(const unsigned char *sha1
)
113 static char hexbuffer
[4][50];
114 static const char hex
[] = "0123456789abcdef";
115 char *buffer
= hexbuffer
[3 & ++bufno
], *buf
= buffer
;
118 for (i
= 0; i
< 20; i
++) {
119 unsigned int val
= *sha1
++;
120 *buf
++ = hex
[val
>> 4];
121 *buf
++ = hex
[val
& 0xf];
128 static void fill_sha1_path(char *pathbuf
, const unsigned char *sha1
)
131 for (i
= 0; i
< 20; i
++) {
132 static char hex
[] = "0123456789abcdef";
133 unsigned int val
= sha1
[i
];
134 char *pos
= pathbuf
+ i
*2 + (i
> 0);
135 *pos
++ = hex
[val
>> 4];
136 *pos
= hex
[val
& 0xf];
141 * NOTE! This returns a statically allocated buffer, so you have to be
142 * careful about using it. Do a "xstrdup()" if you need to save the
145 * Also note that this returns the location for creating. Reading
146 * SHA1 file can happen from any alternate directory listed in the
147 * DB_ENVIRONMENT environment variable if it is not found in
148 * the primary object database.
150 char *sha1_file_name(const unsigned char *sha1
)
152 static char *name
, *base
;
155 const char *sha1_file_directory
= get_object_directory();
156 int len
= strlen(sha1_file_directory
);
157 base
= xmalloc(len
+ 60);
158 memcpy(base
, sha1_file_directory
, len
);
159 memset(base
+len
, 0, 60);
162 name
= base
+ len
+ 1;
164 fill_sha1_path(name
, sha1
);
168 char *sha1_pack_name(const unsigned char *sha1
)
170 static const char hex
[] = "0123456789abcdef";
171 static char *name
, *base
, *buf
;
175 const char *sha1_file_directory
= get_object_directory();
176 int len
= strlen(sha1_file_directory
);
177 base
= xmalloc(len
+ 60);
178 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory
);
179 name
= base
+ len
+ 11;
184 for (i
= 0; i
< 20; i
++) {
185 unsigned int val
= *sha1
++;
186 *buf
++ = hex
[val
>> 4];
187 *buf
++ = hex
[val
& 0xf];
193 char *sha1_pack_index_name(const unsigned char *sha1
)
195 static const char hex
[] = "0123456789abcdef";
196 static char *name
, *base
, *buf
;
200 const char *sha1_file_directory
= get_object_directory();
201 int len
= strlen(sha1_file_directory
);
202 base
= xmalloc(len
+ 60);
203 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory
);
204 name
= base
+ len
+ 11;
209 for (i
= 0; i
< 20; i
++) {
210 unsigned int val
= *sha1
++;
211 *buf
++ = hex
[val
>> 4];
212 *buf
++ = hex
[val
& 0xf];
218 struct alternate_object_database
*alt_odb_list
;
219 static struct alternate_object_database
**alt_odb_tail
;
221 static void read_info_alternates(const char * alternates
, int depth
);
224 * Prepare alternate object database registry.
226 * The variable alt_odb_list points at the list of struct
227 * alternate_object_database. The elements on this list come from
228 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
229 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
230 * whose contents is similar to that environment variable but can be
231 * LF separated. Its base points at a statically allocated buffer that
232 * contains "/the/directory/corresponding/to/.git/objects/...", while
233 * its name points just after the slash at the end of ".git/objects/"
234 * in the example above, and has enough space to hold 40-byte hex
235 * SHA1, an extra slash for the first level indirection, and the
238 static int link_alt_odb_entry(const char * entry
, int len
, const char * relative_base
, int depth
)
241 const char *objdir
= get_object_directory();
242 struct alternate_object_database
*ent
;
243 struct alternate_object_database
*alt
;
244 /* 43 = 40-byte + 2 '/' + terminating NUL */
246 int entlen
= pfxlen
+ 43;
249 if (*entry
!= '/' && relative_base
) {
250 /* Relative alt-odb */
252 base_len
= strlen(relative_base
) + 1;
256 ent
= xmalloc(sizeof(*ent
) + entlen
);
258 if (*entry
!= '/' && relative_base
) {
259 memcpy(ent
->base
, relative_base
, base_len
- 1);
260 ent
->base
[base_len
- 1] = '/';
261 memcpy(ent
->base
+ base_len
, entry
, len
);
264 memcpy(ent
->base
, entry
, pfxlen
);
266 ent
->name
= ent
->base
+ pfxlen
+ 1;
267 ent
->base
[pfxlen
+ 3] = '/';
268 ent
->base
[pfxlen
] = ent
->base
[entlen
-1] = 0;
270 /* Detect cases where alternate disappeared */
271 if (stat(ent
->base
, &st
) || !S_ISDIR(st
.st_mode
)) {
272 error("object directory %s does not exist; "
273 "check .git/objects/info/alternates.",
279 /* Prevent the common mistake of listing the same
280 * thing twice, or object directory itself.
282 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
283 if (!memcmp(ent
->base
, alt
->base
, pfxlen
)) {
288 if (!memcmp(ent
->base
, objdir
, pfxlen
)) {
293 /* add the alternate entry */
295 alt_odb_tail
= &(ent
->next
);
298 /* recursively add alternates */
299 read_info_alternates(ent
->base
, depth
+ 1);
301 ent
->base
[pfxlen
] = '/';
306 static void link_alt_odb_entries(const char *alt
, const char *ep
, int sep
,
307 const char *relative_base
, int depth
)
309 const char *cp
, *last
;
312 error("%s: ignoring alternate object stores, nesting too deep.",
320 if (cp
< ep
&& *cp
== '#') {
321 while (cp
< ep
&& *cp
!= sep
)
326 while (cp
< ep
&& *cp
!= sep
)
329 if ((*last
!= '/') && depth
) {
330 error("%s: ignoring relative alternate object store %s",
331 relative_base
, last
);
333 link_alt_odb_entry(last
, cp
- last
,
334 relative_base
, depth
);
337 while (cp
< ep
&& *cp
== sep
)
343 static void read_info_alternates(const char * relative_base
, int depth
)
350 sprintf(path
, "%s/info/alternates", relative_base
);
351 fd
= open(path
, O_RDONLY
);
354 if (fstat(fd
, &st
) || (st
.st_size
== 0)) {
358 map
= mmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
360 if (map
== MAP_FAILED
)
363 link_alt_odb_entries(map
, map
+ st
.st_size
, '\n', relative_base
, depth
);
365 munmap(map
, st
.st_size
);
368 void prepare_alt_odb(void)
372 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
377 alt_odb_tail
= &alt_odb_list
;
378 link_alt_odb_entries(alt
, alt
+ strlen(alt
), ':', NULL
, 0);
380 read_info_alternates(get_object_directory(), 0);
383 static char *find_sha1_file(const unsigned char *sha1
, struct stat
*st
)
385 char *name
= sha1_file_name(sha1
);
386 struct alternate_object_database
*alt
;
391 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
393 fill_sha1_path(name
, sha1
);
394 if (!stat(alt
->base
, st
))
400 #define PACK_MAX_SZ (1<<26)
401 static int pack_used_ctr
;
402 static unsigned long pack_mapped
;
403 struct packed_git
*packed_git
;
405 static int check_packed_git_idx(const char *path
, unsigned long *idx_size_
,
410 unsigned long idx_size
;
412 int fd
= open(path
, O_RDONLY
);
416 if (fstat(fd
, &st
)) {
420 idx_size
= st
.st_size
;
421 idx_map
= mmap(NULL
, idx_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
423 if (idx_map
== MAP_FAILED
)
428 *idx_size_
= idx_size
;
430 /* check index map */
431 if (idx_size
< 4*256 + 20 + 20)
432 return error("index file too small");
434 for (i
= 0; i
< 256; i
++) {
435 unsigned int n
= ntohl(index
[i
]);
437 return error("non-monotonic index");
443 * - 256 index entries 4 bytes each
444 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
445 * - 20-byte SHA1 of the packfile
446 * - 20-byte SHA1 file checksum
448 if (idx_size
!= 4*256 + nr
* 24 + 20 + 20)
449 return error("wrong index file size");
454 static int unuse_one_packed_git(void)
456 struct packed_git
*p
, *lru
= NULL
;
458 for (p
= packed_git
; p
; p
= p
->next
) {
459 if (p
->pack_use_cnt
|| !p
->pack_base
)
461 if (!lru
|| p
->pack_last_used
< lru
->pack_last_used
)
466 munmap(lru
->pack_base
, lru
->pack_size
);
467 lru
->pack_base
= NULL
;
471 void unuse_packed_git(struct packed_git
*p
)
476 int use_packed_git(struct packed_git
*p
)
480 /* We created the struct before we had the pack */
481 stat(p
->pack_name
, &st
);
482 if (!S_ISREG(st
.st_mode
))
483 die("packfile %s not a regular file", p
->pack_name
);
484 p
->pack_size
= st
.st_size
;
490 struct pack_header
*hdr
;
492 pack_mapped
+= p
->pack_size
;
493 while (PACK_MAX_SZ
< pack_mapped
&& unuse_one_packed_git())
495 fd
= open(p
->pack_name
, O_RDONLY
);
497 die("packfile %s cannot be opened", p
->pack_name
);
498 if (fstat(fd
, &st
)) {
500 die("packfile %s cannot be opened", p
->pack_name
);
502 if (st
.st_size
!= p
->pack_size
)
503 die("packfile %s size mismatch.", p
->pack_name
);
504 map
= mmap(NULL
, p
->pack_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
506 if (map
== MAP_FAILED
)
507 die("packfile %s cannot be mapped.", p
->pack_name
);
510 /* Check if we understand this pack file. If we don't we're
511 * likely too old to handle it.
514 if (hdr
->hdr_signature
!= htonl(PACK_SIGNATURE
))
515 die("packfile %s isn't actually a pack.", p
->pack_name
);
516 if (!pack_version_ok(hdr
->hdr_version
))
517 die("packfile %s is version %i and not supported"
518 " (try upgrading GIT to a newer version)",
519 p
->pack_name
, ntohl(hdr
->hdr_version
));
521 /* Check if the pack file matches with the index file.
524 if (hashcmp((unsigned char *)(p
->index_base
) +
526 (unsigned char *)p
->pack_base
+
527 p
->pack_size
- 20)) {
528 die("packfile %s does not match index.", p
->pack_name
);
531 p
->pack_last_used
= pack_used_ctr
++;
536 struct packed_git
*add_packed_git(char *path
, int path_len
, int local
)
539 struct packed_git
*p
;
540 unsigned long idx_size
;
542 unsigned char sha1
[20];
544 if (check_packed_git_idx(path
, &idx_size
, &idx_map
))
547 /* do we have a corresponding .pack file? */
548 strcpy(path
+ path_len
- 4, ".pack");
549 if (stat(path
, &st
) || !S_ISREG(st
.st_mode
)) {
550 munmap(idx_map
, idx_size
);
553 /* ok, it looks sane as far as we can check without
554 * actually mapping the pack file.
556 p
= xmalloc(sizeof(*p
) + path_len
+ 2);
557 strcpy(p
->pack_name
, path
);
558 p
->index_size
= idx_size
;
559 p
->pack_size
= st
.st_size
;
560 p
->index_base
= idx_map
;
563 p
->pack_last_used
= 0;
565 p
->pack_local
= local
;
566 if ((path_len
> 44) && !get_sha1_hex(path
+ path_len
- 44, sha1
))
567 hashcpy(p
->sha1
, sha1
);
571 struct packed_git
*parse_pack_index(unsigned char *sha1
)
573 char *path
= sha1_pack_index_name(sha1
);
574 return parse_pack_index_file(sha1
, path
);
577 struct packed_git
*parse_pack_index_file(const unsigned char *sha1
, char *idx_path
)
579 struct packed_git
*p
;
580 unsigned long idx_size
;
584 if (check_packed_git_idx(idx_path
, &idx_size
, &idx_map
))
587 path
= sha1_pack_name(sha1
);
589 p
= xmalloc(sizeof(*p
) + strlen(path
) + 2);
590 strcpy(p
->pack_name
, path
);
591 p
->index_size
= idx_size
;
593 p
->index_base
= idx_map
;
596 p
->pack_last_used
= 0;
598 hashcpy(p
->sha1
, sha1
);
602 void install_packed_git(struct packed_git
*pack
)
604 pack
->next
= packed_git
;
608 static void prepare_packed_git_one(char *objdir
, int local
)
615 sprintf(path
, "%s/pack", objdir
);
620 error("unable to open object pack directory: %s: %s",
621 path
, strerror(errno
));
625 while ((de
= readdir(dir
)) != NULL
) {
626 int namelen
= strlen(de
->d_name
);
627 struct packed_git
*p
;
629 if (!has_extension(de
->d_name
, ".idx"))
632 /* we have .idx. Is it a file we can map? */
633 strcpy(path
+ len
, de
->d_name
);
634 for (p
= packed_git
; p
; p
= p
->next
) {
635 if (!memcmp(path
, p
->pack_name
, len
+ namelen
- 4))
640 p
= add_packed_git(path
, len
+ namelen
, local
);
643 p
->next
= packed_git
;
649 static int prepare_packed_git_run_once
= 0;
650 void prepare_packed_git(void)
652 struct alternate_object_database
*alt
;
654 if (prepare_packed_git_run_once
)
656 prepare_packed_git_one(get_object_directory(), 1);
658 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
660 prepare_packed_git_one(alt
->base
, 0);
663 prepare_packed_git_run_once
= 1;
666 static void reprepare_packed_git(void)
668 prepare_packed_git_run_once
= 0;
669 prepare_packed_git();
672 int check_sha1_signature(const unsigned char *sha1
, void *map
, unsigned long size
, const char *type
)
675 unsigned char real_sha1
[20];
679 SHA1_Update(&c
, header
, 1+sprintf(header
, "%s %lu", type
, size
));
680 SHA1_Update(&c
, map
, size
);
681 SHA1_Final(real_sha1
, &c
);
682 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
685 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
690 char *filename
= find_sha1_file(sha1
, &st
);
696 fd
= open(filename
, O_RDONLY
| sha1_file_open_flag
);
698 /* See if it works without O_NOATIME */
699 switch (sha1_file_open_flag
) {
701 fd
= open(filename
, O_RDONLY
);
709 /* If it failed once, it will probably fail again.
710 * Stop using O_NOATIME
712 sha1_file_open_flag
= 0;
714 map
= mmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
716 if (map
== MAP_FAILED
)
722 int legacy_loose_object(unsigned char *map
)
727 * Is it a zlib-compressed buffer? If so, the first byte
728 * must be 0x78 (15-bit window size, deflated), and the
729 * first 16-bit word is evenly divisible by 31
731 word
= (map
[0] << 8) + map
[1];
732 if (map
[0] == 0x78 && !(word
% 31))
738 unsigned long unpack_object_header_gently(const unsigned char *buf
, unsigned long len
, enum object_type
*type
, unsigned long *sizep
)
743 unsigned long used
= 0;
746 *type
= (c
>> 4) & 7;
752 if (sizeof(long) * 8 <= shift
)
755 size
+= (c
& 0x7f) << shift
;
762 static int unpack_sha1_header(z_stream
*stream
, unsigned char *map
, unsigned long mapsize
, void *buffer
, unsigned long bufsiz
)
764 unsigned long size
, used
;
765 static const char valid_loose_object_type
[8] = {
767 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
768 0, /* "delta" and others are invalid in a loose object */
770 enum object_type type
;
772 /* Get the data stream */
773 memset(stream
, 0, sizeof(*stream
));
774 stream
->next_in
= map
;
775 stream
->avail_in
= mapsize
;
776 stream
->next_out
= buffer
;
777 stream
->avail_out
= bufsiz
;
779 if (legacy_loose_object(map
)) {
781 return inflate(stream
, 0);
784 used
= unpack_object_header_gently(map
, mapsize
, &type
, &size
);
785 if (!used
|| !valid_loose_object_type
[type
])
790 /* Set up the stream for the rest.. */
791 stream
->next_in
= map
;
792 stream
->avail_in
= mapsize
;
795 /* And generate the fake traditional header */
796 stream
->total_out
= 1 + snprintf(buffer
, bufsiz
, "%s %lu",
797 type_names
[type
], size
);
801 static void *unpack_sha1_rest(z_stream
*stream
, void *buffer
, unsigned long size
)
803 int bytes
= strlen(buffer
) + 1;
804 unsigned char *buf
= xmalloc(1+size
);
807 n
= stream
->total_out
- bytes
;
810 memcpy(buf
, (char *) buffer
+ bytes
, n
);
813 stream
->next_out
= buf
+ bytes
;
814 stream
->avail_out
= size
- bytes
;
815 while (inflate(stream
, Z_FINISH
) == Z_OK
)
824 * We used to just use "sscanf()", but that's actually way
825 * too permissive for what we want to check. So do an anal
826 * object header parse by hand.
828 static int parse_sha1_header(char *hdr
, char *type
, unsigned long *sizep
)
834 * The type can be at most ten bytes (including the
835 * terminating '\0' that we add), and is followed by
850 * The length must follow immediately, and be in canonical
851 * decimal format (ie "010" is not valid).
858 unsigned long c
= *hdr
- '0';
862 size
= size
* 10 + c
;
868 * The length must be followed by a zero byte
870 return *hdr
? -1 : 0;
873 void * unpack_sha1_file(void *map
, unsigned long mapsize
, char *type
, unsigned long *size
)
879 ret
= unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
));
880 if (ret
< Z_OK
|| parse_sha1_header(hdr
, type
, size
) < 0)
883 return unpack_sha1_rest(&stream
, hdr
, *size
);
886 static unsigned long get_delta_base(struct packed_git
*p
,
887 unsigned long offset
,
888 enum object_type kind
,
889 unsigned long delta_obj_offset
,
890 unsigned long *base_obj_offset
)
892 unsigned char *base_info
= (unsigned char *) p
->pack_base
+ offset
;
893 unsigned long base_offset
;
895 /* there must be at least 20 bytes left regardless of delta type */
896 if (p
->pack_size
<= offset
+ 20)
897 die("truncated pack file");
899 if (kind
== OBJ_OFS_DELTA
) {
901 unsigned char c
= base_info
[used
++];
902 base_offset
= c
& 127;
905 if (!base_offset
|| base_offset
& ~(~0UL >> 7))
906 die("offset value overflow for delta base object");
907 c
= base_info
[used
++];
908 base_offset
= (base_offset
<< 7) + (c
& 127);
910 base_offset
= delta_obj_offset
- base_offset
;
911 if (base_offset
>= delta_obj_offset
)
912 die("delta base offset out of bound");
914 } else if (kind
== OBJ_REF_DELTA
) {
915 /* The base entry _must_ be in the same pack */
916 base_offset
= find_pack_entry_one(base_info
, p
);
918 die("failed to find delta-pack base object %s",
919 sha1_to_hex(base_info
));
922 die("I am totally screwed");
923 *base_obj_offset
= base_offset
;
927 /* forward declaration for a mutually recursive function */
928 static int packed_object_info(struct packed_git
*p
, unsigned long offset
,
929 char *type
, unsigned long *sizep
);
931 static int packed_delta_info(struct packed_git
*p
,
932 unsigned long offset
,
933 enum object_type kind
,
934 unsigned long obj_offset
,
936 unsigned long *sizep
)
938 unsigned long base_offset
;
940 offset
= get_delta_base(p
, offset
, kind
, obj_offset
, &base_offset
);
942 /* We choose to only get the type of the base object and
943 * ignore potentially corrupt pack file that expects the delta
944 * based on a base with a wrong size. This saves tons of
947 if (packed_object_info(p
, base_offset
, type
, NULL
))
948 die("cannot get info for delta-pack base");
951 const unsigned char *data
;
952 unsigned char delta_head
[64];
953 unsigned long result_size
;
957 memset(&stream
, 0, sizeof(stream
));
959 stream
.next_in
= (unsigned char *) p
->pack_base
+ offset
;
960 stream
.avail_in
= p
->pack_size
- offset
;
961 stream
.next_out
= delta_head
;
962 stream
.avail_out
= sizeof(delta_head
);
964 inflateInit(&stream
);
965 st
= inflate(&stream
, Z_FINISH
);
967 if ((st
!= Z_STREAM_END
) &&
968 stream
.total_out
!= sizeof(delta_head
))
969 die("delta data unpack-initial failed");
971 /* Examine the initial part of the delta to figure out
976 /* ignore base size */
977 get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
979 /* Read the result size */
980 result_size
= get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
981 *sizep
= result_size
;
986 static unsigned long unpack_object_header(struct packed_git
*p
, unsigned long offset
,
987 enum object_type
*type
, unsigned long *sizep
)
991 if (p
->pack_size
<= offset
)
992 die("object offset outside of pack file");
994 used
= unpack_object_header_gently((unsigned char *)p
->pack_base
+
996 p
->pack_size
- offset
, type
, sizep
);
998 die("object offset outside of pack file");
1000 return offset
+ used
;
1003 void packed_object_info_detail(struct packed_git
*p
,
1004 unsigned long offset
,
1006 unsigned long *size
,
1007 unsigned long *store_size
,
1008 unsigned int *delta_chain_length
,
1009 unsigned char *base_sha1
)
1011 unsigned long obj_offset
, val
;
1012 unsigned char *next_sha1
;
1013 enum object_type kind
;
1015 *delta_chain_length
= 0;
1016 obj_offset
= offset
;
1017 offset
= unpack_object_header(p
, offset
, &kind
, size
);
1022 die("corrupted pack file %s containing object of kind %d",
1023 p
->pack_name
, kind
);
1028 strcpy(type
, type_names
[kind
]);
1029 *store_size
= 0; /* notyet */
1032 get_delta_base(p
, offset
, kind
, obj_offset
, &offset
);
1033 if (*delta_chain_length
== 0) {
1034 /* TODO: find base_sha1 as pointed by offset */
1038 if (p
->pack_size
<= offset
+ 20)
1039 die("pack file %s records an incomplete delta base",
1041 next_sha1
= (unsigned char *) p
->pack_base
+ offset
;
1042 if (*delta_chain_length
== 0)
1043 hashcpy(base_sha1
, next_sha1
);
1044 offset
= find_pack_entry_one(next_sha1
, p
);
1047 obj_offset
= offset
;
1048 offset
= unpack_object_header(p
, offset
, &kind
, &val
);
1049 (*delta_chain_length
)++;
1053 static int packed_object_info(struct packed_git
*p
, unsigned long offset
,
1054 char *type
, unsigned long *sizep
)
1056 unsigned long size
, obj_offset
= offset
;
1057 enum object_type kind
;
1059 offset
= unpack_object_header(p
, offset
, &kind
, &size
);
1064 return packed_delta_info(p
, offset
, kind
, obj_offset
, type
, sizep
);
1069 strcpy(type
, type_names
[kind
]);
1072 die("corrupted pack file %s containing object of kind %d",
1073 p
->pack_name
, kind
);
1080 static void *unpack_compressed_entry(struct packed_git
*p
,
1081 unsigned long offset
,
1086 unsigned char *buffer
;
1088 buffer
= xmalloc(size
+ 1);
1090 memset(&stream
, 0, sizeof(stream
));
1091 stream
.next_in
= (unsigned char*)p
->pack_base
+ offset
;
1092 stream
.avail_in
= p
->pack_size
- offset
;
1093 stream
.next_out
= buffer
;
1094 stream
.avail_out
= size
;
1096 inflateInit(&stream
);
1097 st
= inflate(&stream
, Z_FINISH
);
1098 inflateEnd(&stream
);
1099 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= size
) {
1107 static void *unpack_delta_entry(struct packed_git
*p
,
1108 unsigned long offset
,
1109 unsigned long delta_size
,
1110 enum object_type kind
,
1111 unsigned long obj_offset
,
1113 unsigned long *sizep
)
1115 void *delta_data
, *result
, *base
;
1116 unsigned long result_size
, base_size
, base_offset
;
1118 offset
= get_delta_base(p
, offset
, kind
, obj_offset
, &base_offset
);
1119 base
= unpack_entry_gently(p
, base_offset
, type
, &base_size
);
1121 die("failed to read delta base object at %lu from %s",
1122 base_offset
, p
->pack_name
);
1124 delta_data
= unpack_compressed_entry(p
, offset
, delta_size
);
1125 result
= patch_delta(base
, base_size
,
1126 delta_data
, delta_size
,
1129 die("failed to apply delta");
1132 *sizep
= result_size
;
1136 static void *unpack_entry(struct pack_entry
*entry
,
1137 char *type
, unsigned long *sizep
)
1139 struct packed_git
*p
= entry
->p
;
1142 if (use_packed_git(p
))
1143 die("cannot map packed file");
1144 retval
= unpack_entry_gently(p
, entry
->offset
, type
, sizep
);
1145 unuse_packed_git(p
);
1147 die("corrupted pack file %s", p
->pack_name
);
1151 /* The caller is responsible for use_packed_git()/unuse_packed_git() pair */
1152 void *unpack_entry_gently(struct packed_git
*p
, unsigned long offset
,
1153 char *type
, unsigned long *sizep
)
1155 unsigned long size
, obj_offset
= offset
;
1156 enum object_type kind
;
1158 offset
= unpack_object_header(p
, offset
, &kind
, &size
);
1162 return unpack_delta_entry(p
, offset
, size
, kind
, obj_offset
, type
, sizep
);
1167 strcpy(type
, type_names
[kind
]);
1169 return unpack_compressed_entry(p
, offset
, size
);
1175 int num_packed_objects(const struct packed_git
*p
)
1177 /* See check_packed_git_idx() */
1178 return (p
->index_size
- 20 - 20 - 4*256) / 24;
1181 int nth_packed_object_sha1(const struct packed_git
*p
, int n
,
1182 unsigned char* sha1
)
1184 void *index
= p
->index_base
+ 256;
1185 if (n
< 0 || num_packed_objects(p
) <= n
)
1187 hashcpy(sha1
, (unsigned char *) index
+ (24 * n
) + 4);
1191 unsigned long find_pack_entry_one(const unsigned char *sha1
,
1192 struct packed_git
*p
)
1194 unsigned int *level1_ofs
= p
->index_base
;
1195 int hi
= ntohl(level1_ofs
[*sha1
]);
1196 int lo
= ((*sha1
== 0x0) ? 0 : ntohl(level1_ofs
[*sha1
- 1]));
1197 void *index
= p
->index_base
+ 256;
1200 int mi
= (lo
+ hi
) / 2;
1201 int cmp
= hashcmp((unsigned char *)index
+ (24 * mi
) + 4, sha1
);
1203 return ntohl(*((unsigned int *) ((char *) index
+ (24 * mi
))));
1212 static int find_pack_entry(const unsigned char *sha1
, struct pack_entry
*e
, const char **ignore_packed
)
1214 struct packed_git
*p
;
1215 unsigned long offset
;
1217 prepare_packed_git();
1219 for (p
= packed_git
; p
; p
= p
->next
) {
1220 if (ignore_packed
) {
1222 for (ig
= ignore_packed
; *ig
; ig
++)
1223 if (!strcmp(p
->pack_name
, *ig
))
1228 offset
= find_pack_entry_one(sha1
, p
);
1232 hashcpy(e
->sha1
, sha1
);
1239 struct packed_git
*find_sha1_pack(const unsigned char *sha1
,
1240 struct packed_git
*packs
)
1242 struct packed_git
*p
;
1244 for (p
= packs
; p
; p
= p
->next
) {
1245 if (find_pack_entry_one(sha1
, p
))
1252 int sha1_object_info(const unsigned char *sha1
, char *type
, unsigned long *sizep
)
1255 unsigned long mapsize
, size
;
1260 map
= map_sha1_file(sha1
, &mapsize
);
1262 struct pack_entry e
;
1264 if (!find_pack_entry(sha1
, &e
, NULL
)) {
1265 reprepare_packed_git();
1266 if (!find_pack_entry(sha1
, &e
, NULL
))
1267 return error("unable to find %s", sha1_to_hex(sha1
));
1269 if (use_packed_git(e
.p
))
1270 die("cannot map packed file");
1271 status
= packed_object_info(e
.p
, e
.offset
, type
, sizep
);
1272 unuse_packed_git(e
.p
);
1275 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1276 status
= error("unable to unpack %s header",
1278 if (parse_sha1_header(hdr
, type
, &size
) < 0)
1279 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1285 inflateEnd(&stream
);
1286 munmap(map
, mapsize
);
1290 static void *read_packed_sha1(const unsigned char *sha1
, char *type
, unsigned long *size
)
1292 struct pack_entry e
;
1294 if (!find_pack_entry(sha1
, &e
, NULL
)) {
1295 error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1298 return unpack_entry(&e
, type
, size
);
1301 void * read_sha1_file(const unsigned char *sha1
, char *type
, unsigned long *size
)
1303 unsigned long mapsize
;
1305 struct pack_entry e
;
1307 if (find_pack_entry(sha1
, &e
, NULL
))
1308 return read_packed_sha1(sha1
, type
, size
);
1309 map
= map_sha1_file(sha1
, &mapsize
);
1311 buf
= unpack_sha1_file(map
, mapsize
, type
, size
);
1312 munmap(map
, mapsize
);
1315 reprepare_packed_git();
1316 if (find_pack_entry(sha1
, &e
, NULL
))
1317 return read_packed_sha1(sha1
, type
, size
);
1321 void *read_object_with_reference(const unsigned char *sha1
,
1322 const char *required_type
,
1323 unsigned long *size
,
1324 unsigned char *actual_sha1_return
)
1328 unsigned long isize
;
1329 unsigned char actual_sha1
[20];
1331 hashcpy(actual_sha1
, sha1
);
1333 int ref_length
= -1;
1334 const char *ref_type
= NULL
;
1336 buffer
= read_sha1_file(actual_sha1
, type
, &isize
);
1339 if (!strcmp(type
, required_type
)) {
1341 if (actual_sha1_return
)
1342 hashcpy(actual_sha1_return
, actual_sha1
);
1345 /* Handle references */
1346 else if (!strcmp(type
, commit_type
))
1348 else if (!strcmp(type
, tag_type
))
1349 ref_type
= "object ";
1354 ref_length
= strlen(ref_type
);
1356 if (memcmp(buffer
, ref_type
, ref_length
) ||
1357 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1362 /* Now we have the ID of the referred-to object in
1363 * actual_sha1. Check again. */
1367 char *write_sha1_file_prepare(void *buf
,
1370 unsigned char *sha1
,
1376 /* Generate the header */
1377 *hdrlen
= sprintf((char *)hdr
, "%s %lu", type
, len
)+1;
1381 SHA1_Update(&c
, hdr
, *hdrlen
);
1382 SHA1_Update(&c
, buf
, len
);
1383 SHA1_Final(sha1
, &c
);
1385 return sha1_file_name(sha1
);
1389 * Link the tempfile to the final place, possibly creating the
1390 * last directory level as you do so.
1392 * Returns the errno on failure, 0 on success.
1394 static int link_temp_to_file(const char *tmpfile
, const char *filename
)
1399 if (!link(tmpfile
, filename
))
1403 * Try to mkdir the last path component if that failed.
1405 * Re-try the "link()" regardless of whether the mkdir
1406 * succeeds, since a race might mean that somebody
1410 dir
= strrchr(filename
, '/');
1413 mkdir(filename
, 0777);
1414 if (adjust_shared_perm(filename
))
1417 if (!link(tmpfile
, filename
))
1425 * Move the just written object into its final resting place
1427 int move_temp_to_file(const char *tmpfile
, const char *filename
)
1429 int ret
= link_temp_to_file(tmpfile
, filename
);
1432 * Coda hack - coda doesn't like cross-directory links,
1433 * so we fall back to a rename, which will mean that it
1434 * won't be able to check collisions, but that's not a
1437 * The same holds for FAT formatted media.
1439 * When this succeeds, we just return 0. We have nothing
1442 if (ret
&& ret
!= EEXIST
) {
1443 if (!rename(tmpfile
, filename
))
1449 if (ret
!= EEXIST
) {
1450 fprintf(stderr
, "unable to write sha1 filename %s: %s\n", filename
, strerror(ret
));
1453 /* FIXME!!! Collision check here ? */
1459 static int write_buffer(int fd
, const void *buf
, size_t len
)
1464 size
= write(fd
, buf
, len
);
1466 return error("file write: disk full");
1468 if (errno
== EINTR
|| errno
== EAGAIN
)
1470 return error("file write error (%s)", strerror(errno
));
1473 buf
= (char *) buf
+ size
;
1478 static int write_binary_header(unsigned char *hdr
, enum object_type type
, unsigned long len
)
1483 c
= (type
<< 4) | (len
& 15);
1496 static void setup_object_header(z_stream
*stream
, const char *type
, unsigned long len
)
1500 if (use_legacy_headers
) {
1501 while (deflate(stream
, 0) == Z_OK
)
1505 if (!strcmp(type
, blob_type
))
1506 obj_type
= OBJ_BLOB
;
1507 else if (!strcmp(type
, tree_type
))
1508 obj_type
= OBJ_TREE
;
1509 else if (!strcmp(type
, commit_type
))
1510 obj_type
= OBJ_COMMIT
;
1511 else if (!strcmp(type
, tag_type
))
1514 die("trying to generate bogus object of type '%s'", type
);
1515 hdr
= write_binary_header(stream
->next_out
, obj_type
, len
);
1516 stream
->total_out
= hdr
;
1517 stream
->next_out
+= hdr
;
1518 stream
->avail_out
-= hdr
;
1521 int write_sha1_file(void *buf
, unsigned long len
, const char *type
, unsigned char *returnsha1
)
1524 unsigned char *compressed
;
1526 unsigned char sha1
[20];
1528 static char tmpfile
[PATH_MAX
];
1529 unsigned char hdr
[50];
1532 /* Normally if we have it in the pack then we do not bother writing
1533 * it out into .git/objects/??/?{38} file.
1535 filename
= write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1537 hashcpy(returnsha1
, sha1
);
1538 if (has_sha1_file(sha1
))
1540 fd
= open(filename
, O_RDONLY
);
1543 * FIXME!!! We might do collision checking here, but we'd
1544 * need to uncompress the old file and check it. Later.
1550 if (errno
!= ENOENT
) {
1551 fprintf(stderr
, "sha1 file %s: %s\n", filename
, strerror(errno
));
1555 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1557 fd
= mkstemp(tmpfile
);
1559 fprintf(stderr
, "unable to create temporary sha1 filename %s: %s\n", tmpfile
, strerror(errno
));
1564 memset(&stream
, 0, sizeof(stream
));
1565 deflateInit(&stream
, zlib_compression_level
);
1566 size
= 8 + deflateBound(&stream
, len
+hdrlen
);
1567 compressed
= xmalloc(size
);
1570 stream
.next_out
= compressed
;
1571 stream
.avail_out
= size
;
1573 /* First header.. */
1574 stream
.next_in
= hdr
;
1575 stream
.avail_in
= hdrlen
;
1576 setup_object_header(&stream
, type
, len
);
1578 /* Then the data itself.. */
1579 stream
.next_in
= buf
;
1580 stream
.avail_in
= len
;
1581 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1583 deflateEnd(&stream
);
1584 size
= stream
.total_out
;
1586 if (write_buffer(fd
, compressed
, size
) < 0)
1587 die("unable to write sha1 file");
1592 return move_temp_to_file(tmpfile
, filename
);
1596 * We need to unpack and recompress the object for writing
1597 * it out to a different file.
1599 static void *repack_object(const unsigned char *sha1
, unsigned long *objsize
)
1603 unsigned char *unpacked
;
1610 /* need to unpack and recompress it by itself */
1611 unpacked
= read_packed_sha1(sha1
, type
, &len
);
1613 hdrlen
= sprintf(hdr
, "%s %lu", type
, len
) + 1;
1616 memset(&stream
, 0, sizeof(stream
));
1617 deflateInit(&stream
, zlib_compression_level
);
1618 size
= deflateBound(&stream
, len
+ hdrlen
);
1619 buf
= xmalloc(size
);
1622 stream
.next_out
= buf
;
1623 stream
.avail_out
= size
;
1625 /* First header.. */
1626 stream
.next_in
= (void *)hdr
;
1627 stream
.avail_in
= hdrlen
;
1628 while (deflate(&stream
, 0) == Z_OK
)
1631 /* Then the data itself.. */
1632 stream
.next_in
= unpacked
;
1633 stream
.avail_in
= len
;
1634 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1636 deflateEnd(&stream
);
1639 *objsize
= stream
.total_out
;
1643 int write_sha1_to_fd(int fd
, const unsigned char *sha1
)
1646 unsigned long objsize
;
1647 void *buf
= map_sha1_file(sha1
, &objsize
);
1650 retval
= write_buffer(fd
, buf
, objsize
);
1651 munmap(buf
, objsize
);
1655 buf
= repack_object(sha1
, &objsize
);
1656 retval
= write_buffer(fd
, buf
, objsize
);
1661 int write_sha1_from_fd(const unsigned char *sha1
, int fd
, char *buffer
,
1662 size_t bufsize
, size_t *bufposn
)
1664 char tmpfile
[PATH_MAX
];
1667 unsigned char real_sha1
[20];
1668 unsigned char discard
[4096];
1672 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1674 local
= mkstemp(tmpfile
);
1676 return error("Couldn't open %s for %s",
1677 tmpfile
, sha1_to_hex(sha1
));
1679 memset(&stream
, 0, sizeof(stream
));
1681 inflateInit(&stream
);
1688 stream
.avail_in
= *bufposn
;
1689 stream
.next_in
= (unsigned char *) buffer
;
1691 stream
.next_out
= discard
;
1692 stream
.avail_out
= sizeof(discard
);
1693 ret
= inflate(&stream
, Z_SYNC_FLUSH
);
1694 SHA1_Update(&c
, discard
, sizeof(discard
) -
1696 } while (stream
.avail_in
&& ret
== Z_OK
);
1697 if (write_buffer(local
, buffer
, *bufposn
- stream
.avail_in
) < 0)
1698 die("unable to write sha1 file");
1699 memmove(buffer
, buffer
+ *bufposn
- stream
.avail_in
,
1701 *bufposn
= stream
.avail_in
;
1705 size
= read(fd
, buffer
+ *bufposn
, bufsize
- *bufposn
);
1710 return error("Connection closed?");
1711 perror("Reading from connection");
1716 inflateEnd(&stream
);
1719 SHA1_Final(real_sha1
, &c
);
1720 if (ret
!= Z_STREAM_END
) {
1722 return error("File %s corrupted", sha1_to_hex(sha1
));
1724 if (hashcmp(sha1
, real_sha1
)) {
1726 return error("File %s has bad hash", sha1_to_hex(sha1
));
1729 return move_temp_to_file(tmpfile
, sha1_file_name(sha1
));
1732 int has_pack_index(const unsigned char *sha1
)
1735 if (stat(sha1_pack_index_name(sha1
), &st
))
1740 int has_pack_file(const unsigned char *sha1
)
1743 if (stat(sha1_pack_name(sha1
), &st
))
1748 int has_sha1_pack(const unsigned char *sha1
, const char **ignore_packed
)
1750 struct pack_entry e
;
1751 return find_pack_entry(sha1
, &e
, ignore_packed
);
1754 int has_sha1_file(const unsigned char *sha1
)
1757 struct pack_entry e
;
1759 if (find_pack_entry(sha1
, &e
, NULL
))
1761 return find_sha1_file(sha1
, &st
) ? 1 : 0;
1765 * reads from fd as long as possible into a supplied buffer of size bytes.
1766 * If necessary the buffer's size is increased using realloc()
1768 * returns 0 if anything went fine and -1 otherwise
1770 * NOTE: both buf and size may change, but even when -1 is returned
1771 * you still have to free() it yourself.
1773 int read_pipe(int fd
, char** return_buf
, unsigned long* return_size
)
1775 char* buf
= *return_buf
;
1776 unsigned long size
= *return_size
;
1778 unsigned long off
= 0;
1781 iret
= xread(fd
, buf
+ off
, size
- off
);
1786 buf
= xrealloc(buf
, size
);
1799 int index_pipe(unsigned char *sha1
, int fd
, const char *type
, int write_object
)
1801 unsigned long size
= 4096;
1802 char *buf
= xmalloc(size
);
1804 unsigned char hdr
[50];
1807 if (read_pipe(fd
, &buf
, &size
)) {
1815 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1817 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1824 int index_fd(unsigned char *sha1
, int fd
, struct stat
*st
, int write_object
, const char *type
)
1826 unsigned long size
= st
->st_size
;
1829 unsigned char hdr
[50];
1834 buf
= mmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1836 if (buf
== MAP_FAILED
)
1842 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1844 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1852 int index_path(unsigned char *sha1
, const char *path
, struct stat
*st
, int write_object
)
1857 switch (st
->st_mode
& S_IFMT
) {
1859 fd
= open(path
, O_RDONLY
);
1861 return error("open(\"%s\"): %s", path
,
1863 if (index_fd(sha1
, fd
, st
, write_object
, NULL
) < 0)
1864 return error("%s: failed to insert into database",
1868 target
= xmalloc(st
->st_size
+1);
1869 if (readlink(path
, target
, st
->st_size
+1) != st
->st_size
) {
1870 char *errstr
= strerror(errno
);
1872 return error("readlink(\"%s\"): %s", path
,
1875 if (!write_object
) {
1876 unsigned char hdr
[50];
1878 write_sha1_file_prepare(target
, st
->st_size
, blob_type
,
1879 sha1
, hdr
, &hdrlen
);
1880 } else if (write_sha1_file(target
, st
->st_size
, blob_type
, sha1
))
1881 return error("%s: failed to insert into database",
1886 return error("%s: unsupported file type", path
);