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 static unsigned hexval(char c
)
31 if (c
>= '0' && c
<= '9')
33 if (c
>= 'a' && c
<= 'f')
35 if (c
>= 'A' && c
<= 'F')
40 int get_sha1_hex(const char *hex
, unsigned char *sha1
)
43 for (i
= 0; i
< 20; i
++) {
44 unsigned int val
= (hexval(hex
[0]) << 4) | hexval(hex
[1]);
53 int safe_create_leading_directories(char *path
)
62 pos
= strchr(pos
, '/');
66 if (!stat(path
, &st
)) {
68 if (!S_ISDIR(st
.st_mode
)) {
73 else if (mkdir(path
, 0777)) {
77 else if (adjust_shared_perm(path
)) {
86 char * sha1_to_hex(const unsigned char *sha1
)
89 static char hexbuffer
[4][50];
90 static const char hex
[] = "0123456789abcdef";
91 char *buffer
= hexbuffer
[3 & ++bufno
], *buf
= buffer
;
94 for (i
= 0; i
< 20; i
++) {
95 unsigned int val
= *sha1
++;
96 *buf
++ = hex
[val
>> 4];
97 *buf
++ = hex
[val
& 0xf];
104 static void fill_sha1_path(char *pathbuf
, const unsigned char *sha1
)
107 for (i
= 0; i
< 20; i
++) {
108 static char hex
[] = "0123456789abcdef";
109 unsigned int val
= sha1
[i
];
110 char *pos
= pathbuf
+ i
*2 + (i
> 0);
111 *pos
++ = hex
[val
>> 4];
112 *pos
= hex
[val
& 0xf];
117 * NOTE! This returns a statically allocated buffer, so you have to be
118 * careful about using it. Do a "strdup()" if you need to save the
121 * Also note that this returns the location for creating. Reading
122 * SHA1 file can happen from any alternate directory listed in the
123 * DB_ENVIRONMENT environment variable if it is not found in
124 * the primary object database.
126 char *sha1_file_name(const unsigned char *sha1
)
128 static char *name
, *base
;
131 const char *sha1_file_directory
= get_object_directory();
132 int len
= strlen(sha1_file_directory
);
133 base
= xmalloc(len
+ 60);
134 memcpy(base
, sha1_file_directory
, len
);
135 memset(base
+len
, 0, 60);
138 name
= base
+ len
+ 1;
140 fill_sha1_path(name
, sha1
);
144 char *sha1_pack_name(const unsigned char *sha1
)
146 static const char hex
[] = "0123456789abcdef";
147 static char *name
, *base
, *buf
;
151 const char *sha1_file_directory
= get_object_directory();
152 int len
= strlen(sha1_file_directory
);
153 base
= xmalloc(len
+ 60);
154 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory
);
155 name
= base
+ len
+ 11;
160 for (i
= 0; i
< 20; i
++) {
161 unsigned int val
= *sha1
++;
162 *buf
++ = hex
[val
>> 4];
163 *buf
++ = hex
[val
& 0xf];
169 char *sha1_pack_index_name(const unsigned char *sha1
)
171 static const char hex
[] = "0123456789abcdef";
172 static char *name
, *base
, *buf
;
176 const char *sha1_file_directory
= get_object_directory();
177 int len
= strlen(sha1_file_directory
);
178 base
= xmalloc(len
+ 60);
179 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory
);
180 name
= base
+ len
+ 11;
185 for (i
= 0; i
< 20; i
++) {
186 unsigned int val
= *sha1
++;
187 *buf
++ = hex
[val
>> 4];
188 *buf
++ = hex
[val
& 0xf];
194 struct alternate_object_database
*alt_odb_list
;
195 static struct alternate_object_database
**alt_odb_tail
;
197 static void read_info_alternates(const char * alternates
, int depth
);
200 * Prepare alternate object database registry.
202 * The variable alt_odb_list points at the list of struct
203 * alternate_object_database. The elements on this list come from
204 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
205 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
206 * whose contents is similar to that environment variable but can be
207 * LF separated. Its base points at a statically allocated buffer that
208 * contains "/the/directory/corresponding/to/.git/objects/...", while
209 * its name points just after the slash at the end of ".git/objects/"
210 * in the example above, and has enough space to hold 40-byte hex
211 * SHA1, an extra slash for the first level indirection, and the
214 static int link_alt_odb_entry(const char * entry
, int len
, const char * relative_base
, int depth
)
217 const char *objdir
= get_object_directory();
218 struct alternate_object_database
*ent
;
219 struct alternate_object_database
*alt
;
220 /* 43 = 40-byte + 2 '/' + terminating NUL */
222 int entlen
= pfxlen
+ 43;
225 if (*entry
!= '/' && relative_base
) {
226 /* Relative alt-odb */
228 base_len
= strlen(relative_base
) + 1;
232 ent
= xmalloc(sizeof(*ent
) + entlen
);
234 if (*entry
!= '/' && relative_base
) {
235 memcpy(ent
->base
, relative_base
, base_len
- 1);
236 ent
->base
[base_len
- 1] = '/';
237 memcpy(ent
->base
+ base_len
, entry
, len
);
240 memcpy(ent
->base
, entry
, pfxlen
);
242 ent
->name
= ent
->base
+ pfxlen
+ 1;
243 ent
->base
[pfxlen
+ 3] = '/';
244 ent
->base
[pfxlen
] = ent
->base
[entlen
-1] = 0;
246 /* Detect cases where alternate disappeared */
247 if (stat(ent
->base
, &st
) || !S_ISDIR(st
.st_mode
)) {
248 error("object directory %s does not exist; "
249 "check .git/objects/info/alternates.",
255 /* Prevent the common mistake of listing the same
256 * thing twice, or object directory itself.
258 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
259 if (!memcmp(ent
->base
, alt
->base
, pfxlen
)) {
264 if (!memcmp(ent
->base
, objdir
, pfxlen
)) {
269 /* add the alternate entry */
271 alt_odb_tail
= &(ent
->next
);
274 /* recursively add alternates */
275 read_info_alternates(ent
->base
, depth
+ 1);
277 ent
->base
[pfxlen
] = '/';
282 static void link_alt_odb_entries(const char *alt
, const char *ep
, int sep
,
283 const char *relative_base
, int depth
)
285 const char *cp
, *last
;
288 error("%s: ignoring alternate object stores, nesting too deep.",
296 if (cp
< ep
&& *cp
== '#') {
297 while (cp
< ep
&& *cp
!= sep
)
302 while (cp
< ep
&& *cp
!= sep
)
305 if ((*last
!= '/') && depth
) {
306 error("%s: ignoring relative alternate object store %s",
307 relative_base
, last
);
309 link_alt_odb_entry(last
, cp
- last
,
310 relative_base
, depth
);
313 while (cp
< ep
&& *cp
== sep
)
319 static void read_info_alternates(const char * relative_base
, int depth
)
326 sprintf(path
, "%s/info/alternates", relative_base
);
327 fd
= open(path
, O_RDONLY
);
330 if (fstat(fd
, &st
) || (st
.st_size
== 0)) {
334 map
= mmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
336 if (map
== MAP_FAILED
)
339 link_alt_odb_entries(map
, map
+ st
.st_size
, '\n', relative_base
, depth
);
341 munmap(map
, st
.st_size
);
344 void prepare_alt_odb(void)
348 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
353 alt_odb_tail
= &alt_odb_list
;
354 link_alt_odb_entries(alt
, alt
+ strlen(alt
), ':', NULL
, 0);
356 read_info_alternates(get_object_directory(), 0);
359 static char *find_sha1_file(const unsigned char *sha1
, struct stat
*st
)
361 char *name
= sha1_file_name(sha1
);
362 struct alternate_object_database
*alt
;
367 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
369 fill_sha1_path(name
, sha1
);
370 if (!stat(alt
->base
, st
))
376 #define PACK_MAX_SZ (1<<26)
377 static int pack_used_ctr
;
378 static unsigned long pack_mapped
;
379 struct packed_git
*packed_git
;
381 static int check_packed_git_idx(const char *path
, unsigned long *idx_size_
,
386 unsigned long idx_size
;
388 int fd
= open(path
, O_RDONLY
);
392 if (fstat(fd
, &st
)) {
396 idx_size
= st
.st_size
;
397 idx_map
= mmap(NULL
, idx_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
399 if (idx_map
== MAP_FAILED
)
404 *idx_size_
= idx_size
;
406 /* check index map */
407 if (idx_size
< 4*256 + 20 + 20)
408 return error("index file too small");
410 for (i
= 0; i
< 256; i
++) {
411 unsigned int n
= ntohl(index
[i
]);
413 return error("non-monotonic index");
419 * - 256 index entries 4 bytes each
420 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
421 * - 20-byte SHA1 of the packfile
422 * - 20-byte SHA1 file checksum
424 if (idx_size
!= 4*256 + nr
* 24 + 20 + 20)
425 return error("wrong index file size");
430 static int unuse_one_packed_git(void)
432 struct packed_git
*p
, *lru
= NULL
;
434 for (p
= packed_git
; p
; p
= p
->next
) {
435 if (p
->pack_use_cnt
|| !p
->pack_base
)
437 if (!lru
|| p
->pack_last_used
< lru
->pack_last_used
)
442 munmap(lru
->pack_base
, lru
->pack_size
);
443 lru
->pack_base
= NULL
;
447 void unuse_packed_git(struct packed_git
*p
)
452 int use_packed_git(struct packed_git
*p
)
456 /* We created the struct before we had the pack */
457 stat(p
->pack_name
, &st
);
458 if (!S_ISREG(st
.st_mode
))
459 die("packfile %s not a regular file", p
->pack_name
);
460 p
->pack_size
= st
.st_size
;
466 struct pack_header
*hdr
;
468 pack_mapped
+= p
->pack_size
;
469 while (PACK_MAX_SZ
< pack_mapped
&& unuse_one_packed_git())
471 fd
= open(p
->pack_name
, O_RDONLY
);
473 die("packfile %s cannot be opened", p
->pack_name
);
474 if (fstat(fd
, &st
)) {
476 die("packfile %s cannot be opened", p
->pack_name
);
478 if (st
.st_size
!= p
->pack_size
)
479 die("packfile %s size mismatch.", p
->pack_name
);
480 map
= mmap(NULL
, p
->pack_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
482 if (map
== MAP_FAILED
)
483 die("packfile %s cannot be mapped.", p
->pack_name
);
486 /* Check if we understand this pack file. If we don't we're
487 * likely too old to handle it.
490 if (hdr
->hdr_signature
!= htonl(PACK_SIGNATURE
))
491 die("packfile %s isn't actually a pack.", p
->pack_name
);
492 if (!pack_version_ok(hdr
->hdr_version
))
493 die("packfile %s is version %i and not supported"
494 " (try upgrading GIT to a newer version)",
495 p
->pack_name
, ntohl(hdr
->hdr_version
));
497 /* Check if the pack file matches with the index file.
500 if (hashcmp((unsigned char *)(p
->index_base
) +
502 (unsigned char *)p
->pack_base
+
503 p
->pack_size
- 20)) {
504 die("packfile %s does not match index.", p
->pack_name
);
507 p
->pack_last_used
= pack_used_ctr
++;
512 struct packed_git
*add_packed_git(char *path
, int path_len
, int local
)
515 struct packed_git
*p
;
516 unsigned long idx_size
;
518 unsigned char sha1
[20];
520 if (check_packed_git_idx(path
, &idx_size
, &idx_map
))
523 /* do we have a corresponding .pack file? */
524 strcpy(path
+ path_len
- 4, ".pack");
525 if (stat(path
, &st
) || !S_ISREG(st
.st_mode
)) {
526 munmap(idx_map
, idx_size
);
529 /* ok, it looks sane as far as we can check without
530 * actually mapping the pack file.
532 p
= xmalloc(sizeof(*p
) + path_len
+ 2);
533 strcpy(p
->pack_name
, path
);
534 p
->index_size
= idx_size
;
535 p
->pack_size
= st
.st_size
;
536 p
->index_base
= idx_map
;
539 p
->pack_last_used
= 0;
541 p
->pack_local
= local
;
542 if ((path_len
> 44) && !get_sha1_hex(path
+ path_len
- 44, sha1
))
543 hashcpy(p
->sha1
, sha1
);
547 struct packed_git
*parse_pack_index(unsigned char *sha1
)
549 char *path
= sha1_pack_index_name(sha1
);
550 return parse_pack_index_file(sha1
, path
);
553 struct packed_git
*parse_pack_index_file(const unsigned char *sha1
, char *idx_path
)
555 struct packed_git
*p
;
556 unsigned long idx_size
;
560 if (check_packed_git_idx(idx_path
, &idx_size
, &idx_map
))
563 path
= sha1_pack_name(sha1
);
565 p
= xmalloc(sizeof(*p
) + strlen(path
) + 2);
566 strcpy(p
->pack_name
, path
);
567 p
->index_size
= idx_size
;
569 p
->index_base
= idx_map
;
572 p
->pack_last_used
= 0;
574 hashcpy(p
->sha1
, sha1
);
578 void install_packed_git(struct packed_git
*pack
)
580 pack
->next
= packed_git
;
584 static void prepare_packed_git_one(char *objdir
, int local
)
591 sprintf(path
, "%s/pack", objdir
);
596 error("unable to open object pack directory: %s: %s",
597 path
, strerror(errno
));
601 while ((de
= readdir(dir
)) != NULL
) {
602 int namelen
= strlen(de
->d_name
);
603 struct packed_git
*p
;
605 if (!has_extension(de
->d_name
, ".idx"))
608 /* we have .idx. Is it a file we can map? */
609 strcpy(path
+ len
, de
->d_name
);
610 for (p
= packed_git
; p
; p
= p
->next
) {
611 if (!memcmp(path
, p
->pack_name
, len
+ namelen
- 4))
616 p
= add_packed_git(path
, len
+ namelen
, local
);
619 p
->next
= packed_git
;
625 static int prepare_packed_git_run_once
= 0;
626 void prepare_packed_git(void)
628 struct alternate_object_database
*alt
;
630 if (prepare_packed_git_run_once
)
632 prepare_packed_git_one(get_object_directory(), 1);
634 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
636 prepare_packed_git_one(alt
->base
, 0);
639 prepare_packed_git_run_once
= 1;
642 static void reprepare_packed_git(void)
644 prepare_packed_git_run_once
= 0;
645 prepare_packed_git();
648 int check_sha1_signature(const unsigned char *sha1
, void *map
, unsigned long size
, const char *type
)
651 unsigned char real_sha1
[20];
655 SHA1_Update(&c
, header
, 1+sprintf(header
, "%s %lu", type
, size
));
656 SHA1_Update(&c
, map
, size
);
657 SHA1_Final(real_sha1
, &c
);
658 return hashcmp(sha1
, real_sha1
) ? -1 : 0;
661 void *map_sha1_file(const unsigned char *sha1
, unsigned long *size
)
666 char *filename
= find_sha1_file(sha1
, &st
);
672 fd
= open(filename
, O_RDONLY
| sha1_file_open_flag
);
674 /* See if it works without O_NOATIME */
675 switch (sha1_file_open_flag
) {
677 fd
= open(filename
, O_RDONLY
);
685 /* If it failed once, it will probably fail again.
686 * Stop using O_NOATIME
688 sha1_file_open_flag
= 0;
690 map
= mmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
692 if (map
== MAP_FAILED
)
698 int legacy_loose_object(unsigned char *map
)
703 * Is it a zlib-compressed buffer? If so, the first byte
704 * must be 0x78 (15-bit window size, deflated), and the
705 * first 16-bit word is evenly divisible by 31
707 word
= (map
[0] << 8) + map
[1];
708 if (map
[0] == 0x78 && !(word
% 31))
714 static int unpack_sha1_header(z_stream
*stream
, unsigned char *map
, unsigned long mapsize
, void *buffer
, unsigned long bufsiz
)
719 static const char *typename
[8] = {
721 "commit", "tree", "blob", "tag",
726 /* Get the data stream */
727 memset(stream
, 0, sizeof(*stream
));
728 stream
->next_in
= map
;
729 stream
->avail_in
= mapsize
;
730 stream
->next_out
= buffer
;
731 stream
->avail_out
= bufsiz
;
733 if (legacy_loose_object(map
)) {
735 return inflate(stream
, 0);
740 type
= typename
[(c
>> 4) & 7];
747 if (bits
>= 8*sizeof(long))
750 size
+= (c
& 0x7f) << bits
;
755 /* Set up the stream for the rest.. */
756 stream
->next_in
= map
;
757 stream
->avail_in
= mapsize
;
760 /* And generate the fake traditional header */
761 stream
->total_out
= 1 + snprintf(buffer
, bufsiz
, "%s %lu", type
, size
);
765 static void *unpack_sha1_rest(z_stream
*stream
, void *buffer
, unsigned long size
)
767 int bytes
= strlen(buffer
) + 1;
768 unsigned char *buf
= xmalloc(1+size
);
771 n
= stream
->total_out
- bytes
;
774 memcpy(buf
, (char *) buffer
+ bytes
, n
);
777 stream
->next_out
= buf
+ bytes
;
778 stream
->avail_out
= size
- bytes
;
779 while (inflate(stream
, Z_FINISH
) == Z_OK
)
788 * We used to just use "sscanf()", but that's actually way
789 * too permissive for what we want to check. So do an anal
790 * object header parse by hand.
792 static int parse_sha1_header(char *hdr
, char *type
, unsigned long *sizep
)
798 * The type can be at most ten bytes (including the
799 * terminating '\0' that we add), and is followed by
814 * The length must follow immediately, and be in canonical
815 * decimal format (ie "010" is not valid).
822 unsigned long c
= *hdr
- '0';
826 size
= size
* 10 + c
;
832 * The length must be followed by a zero byte
834 return *hdr
? -1 : 0;
837 void * unpack_sha1_file(void *map
, unsigned long mapsize
, char *type
, unsigned long *size
)
843 ret
= unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
));
844 if (ret
< Z_OK
|| parse_sha1_header(hdr
, type
, size
) < 0)
847 return unpack_sha1_rest(&stream
, hdr
, *size
);
850 /* forward declaration for a mutually recursive function */
851 static int packed_object_info(struct pack_entry
*entry
,
852 char *type
, unsigned long *sizep
);
854 static int packed_delta_info(unsigned char *base_sha1
,
855 unsigned long delta_size
,
858 unsigned long *sizep
,
859 struct packed_git
*p
)
861 struct pack_entry base_ent
;
864 die("truncated pack file");
866 /* The base entry _must_ be in the same pack */
867 if (!find_pack_entry_one(base_sha1
, &base_ent
, p
))
868 die("failed to find delta-pack base object %s",
869 sha1_to_hex(base_sha1
));
871 /* We choose to only get the type of the base object and
872 * ignore potentially corrupt pack file that expects the delta
873 * based on a base with a wrong size. This saves tons of
877 if (packed_object_info(&base_ent
, type
, NULL
))
878 die("cannot get info for delta-pack base");
881 const unsigned char *data
;
882 unsigned char delta_head
[64];
883 unsigned long result_size
;
887 memset(&stream
, 0, sizeof(stream
));
889 data
= stream
.next_in
= base_sha1
+ 20;
890 stream
.avail_in
= left
- 20;
891 stream
.next_out
= delta_head
;
892 stream
.avail_out
= sizeof(delta_head
);
894 inflateInit(&stream
);
895 st
= inflate(&stream
, Z_FINISH
);
897 if ((st
!= Z_STREAM_END
) &&
898 stream
.total_out
!= sizeof(delta_head
))
899 die("delta data unpack-initial failed");
901 /* Examine the initial part of the delta to figure out
906 /* ignore base size */
907 get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
909 /* Read the result size */
910 result_size
= get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
911 *sizep
= result_size
;
916 static unsigned long unpack_object_header(struct packed_git
*p
, unsigned long offset
,
917 enum object_type
*type
, unsigned long *sizep
)
920 unsigned char *pack
, c
;
923 if (offset
>= p
->pack_size
)
924 die("object offset outside of pack file");
926 pack
= (unsigned char *) p
->pack_base
+ offset
;
929 *type
= (c
>> 4) & 7;
933 if (offset
>= p
->pack_size
)
934 die("object offset outside of pack file");
937 size
+= (c
& 0x7f) << shift
;
944 int check_reuse_pack_delta(struct packed_git
*p
, unsigned long offset
,
945 unsigned char *base
, unsigned long *sizep
,
946 enum object_type
*kindp
)
953 ptr
= unpack_object_header(p
, ptr
, kindp
, sizep
);
954 if (*kindp
!= OBJ_DELTA
)
956 hashcpy(base
, (unsigned char *) p
->pack_base
+ ptr
);
963 void packed_object_info_detail(struct pack_entry
*e
,
966 unsigned long *store_size
,
967 unsigned int *delta_chain_length
,
968 unsigned char *base_sha1
)
970 struct packed_git
*p
= e
->p
;
971 unsigned long offset
;
973 enum object_type kind
;
975 offset
= unpack_object_header(p
, e
->offset
, &kind
, size
);
976 pack
= (unsigned char *) p
->pack_base
+ offset
;
977 if (kind
!= OBJ_DELTA
)
978 *delta_chain_length
= 0;
980 unsigned int chain_length
= 0;
981 if (p
->pack_size
<= offset
+ 20)
982 die("pack file %s records an incomplete delta base",
984 hashcpy(base_sha1
, pack
);
986 struct pack_entry base_ent
;
989 find_pack_entry_one(pack
, &base_ent
, p
);
990 offset
= unpack_object_header(p
, base_ent
.offset
,
992 pack
= (unsigned char *) p
->pack_base
+ offset
;
994 } while (kind
== OBJ_DELTA
);
995 *delta_chain_length
= chain_length
;
999 strcpy(type
, commit_type
);
1002 strcpy(type
, tree_type
);
1005 strcpy(type
, blob_type
);
1008 strcpy(type
, tag_type
);
1011 die("corrupted pack file %s containing object of kind %d",
1012 p
->pack_name
, kind
);
1014 *store_size
= 0; /* notyet */
1017 static int packed_object_info(struct pack_entry
*entry
,
1018 char *type
, unsigned long *sizep
)
1020 struct packed_git
*p
= entry
->p
;
1021 unsigned long offset
, size
, left
;
1022 unsigned char *pack
;
1023 enum object_type kind
;
1026 if (use_packed_git(p
))
1027 die("cannot map packed file");
1029 offset
= unpack_object_header(p
, entry
->offset
, &kind
, &size
);
1030 pack
= (unsigned char *) p
->pack_base
+ offset
;
1031 left
= p
->pack_size
- offset
;
1035 retval
= packed_delta_info(pack
, size
, left
, type
, sizep
, p
);
1036 unuse_packed_git(p
);
1039 strcpy(type
, commit_type
);
1042 strcpy(type
, tree_type
);
1045 strcpy(type
, blob_type
);
1048 strcpy(type
, tag_type
);
1051 die("corrupted pack file %s containing object of kind %d",
1052 p
->pack_name
, kind
);
1056 unuse_packed_git(p
);
1060 static void *unpack_delta_entry(unsigned char *base_sha1
,
1061 unsigned long delta_size
,
1064 unsigned long *sizep
,
1065 struct packed_git
*p
)
1067 struct pack_entry base_ent
;
1068 void *data
, *delta_data
, *result
, *base
;
1069 unsigned long data_size
, result_size
, base_size
;
1074 die("truncated pack file");
1076 /* The base entry _must_ be in the same pack */
1077 if (!find_pack_entry_one(base_sha1
, &base_ent
, p
))
1078 die("failed to find delta-pack base object %s",
1079 sha1_to_hex(base_sha1
));
1080 base
= unpack_entry_gently(&base_ent
, type
, &base_size
);
1082 die("failed to read delta-pack base object %s",
1083 sha1_to_hex(base_sha1
));
1085 data
= base_sha1
+ 20;
1086 data_size
= left
- 20;
1087 delta_data
= xmalloc(delta_size
);
1089 memset(&stream
, 0, sizeof(stream
));
1091 stream
.next_in
= data
;
1092 stream
.avail_in
= data_size
;
1093 stream
.next_out
= delta_data
;
1094 stream
.avail_out
= delta_size
;
1096 inflateInit(&stream
);
1097 st
= inflate(&stream
, Z_FINISH
);
1098 inflateEnd(&stream
);
1099 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= delta_size
)
1100 die("delta data unpack failed");
1102 result
= patch_delta(base
, base_size
,
1103 delta_data
, delta_size
,
1106 die("failed to apply delta");
1109 *sizep
= result_size
;
1113 static void *unpack_non_delta_entry(unsigned char *data
,
1119 unsigned char *buffer
;
1121 buffer
= xmalloc(size
+ 1);
1123 memset(&stream
, 0, sizeof(stream
));
1124 stream
.next_in
= data
;
1125 stream
.avail_in
= left
;
1126 stream
.next_out
= buffer
;
1127 stream
.avail_out
= size
;
1129 inflateInit(&stream
);
1130 st
= inflate(&stream
, Z_FINISH
);
1131 inflateEnd(&stream
);
1132 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= size
) {
1140 static void *unpack_entry(struct pack_entry
*entry
,
1141 char *type
, unsigned long *sizep
)
1143 struct packed_git
*p
= entry
->p
;
1146 if (use_packed_git(p
))
1147 die("cannot map packed file");
1148 retval
= unpack_entry_gently(entry
, type
, sizep
);
1149 unuse_packed_git(p
);
1151 die("corrupted pack file %s", p
->pack_name
);
1155 /* The caller is responsible for use_packed_git()/unuse_packed_git() pair */
1156 void *unpack_entry_gently(struct pack_entry
*entry
,
1157 char *type
, unsigned long *sizep
)
1159 struct packed_git
*p
= entry
->p
;
1160 unsigned long offset
, size
, left
;
1161 unsigned char *pack
;
1162 enum object_type kind
;
1165 offset
= unpack_object_header(p
, entry
->offset
, &kind
, &size
);
1166 pack
= (unsigned char *) p
->pack_base
+ offset
;
1167 left
= p
->pack_size
- offset
;
1170 retval
= unpack_delta_entry(pack
, size
, left
, type
, sizep
, p
);
1173 strcpy(type
, commit_type
);
1176 strcpy(type
, tree_type
);
1179 strcpy(type
, blob_type
);
1182 strcpy(type
, tag_type
);
1188 retval
= unpack_non_delta_entry(pack
, size
, left
);
1192 int num_packed_objects(const struct packed_git
*p
)
1194 /* See check_packed_git_idx() */
1195 return (p
->index_size
- 20 - 20 - 4*256) / 24;
1198 int nth_packed_object_sha1(const struct packed_git
*p
, int n
,
1199 unsigned char* sha1
)
1201 void *index
= p
->index_base
+ 256;
1202 if (n
< 0 || num_packed_objects(p
) <= n
)
1204 hashcpy(sha1
, (unsigned char *) index
+ (24 * n
) + 4);
1208 int find_pack_entry_one(const unsigned char *sha1
,
1209 struct pack_entry
*e
, struct packed_git
*p
)
1211 unsigned int *level1_ofs
= p
->index_base
;
1212 int hi
= ntohl(level1_ofs
[*sha1
]);
1213 int lo
= ((*sha1
== 0x0) ? 0 : ntohl(level1_ofs
[*sha1
- 1]));
1214 void *index
= p
->index_base
+ 256;
1217 int mi
= (lo
+ hi
) / 2;
1218 int cmp
= hashcmp((unsigned char *)index
+ (24 * mi
) + 4, sha1
);
1220 e
->offset
= ntohl(*((unsigned int *) ((char *) index
+ (24 * mi
))));
1221 hashcpy(e
->sha1
, sha1
);
1233 static int find_pack_entry(const unsigned char *sha1
, struct pack_entry
*e
)
1235 struct packed_git
*p
;
1236 prepare_packed_git();
1238 for (p
= packed_git
; p
; p
= p
->next
) {
1239 if (find_pack_entry_one(sha1
, e
, p
))
1245 struct packed_git
*find_sha1_pack(const unsigned char *sha1
,
1246 struct packed_git
*packs
)
1248 struct packed_git
*p
;
1249 struct pack_entry e
;
1251 for (p
= packs
; p
; p
= p
->next
) {
1252 if (find_pack_entry_one(sha1
, &e
, p
))
1259 int sha1_object_info(const unsigned char *sha1
, char *type
, unsigned long *sizep
)
1262 unsigned long mapsize
, size
;
1267 map
= map_sha1_file(sha1
, &mapsize
);
1269 struct pack_entry e
;
1271 if (find_pack_entry(sha1
, &e
))
1272 return packed_object_info(&e
, type
, sizep
);
1273 reprepare_packed_git();
1274 if (find_pack_entry(sha1
, &e
))
1275 return packed_object_info(&e
, type
, sizep
);
1276 return error("unable to find %s", sha1_to_hex(sha1
));
1278 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1279 status
= error("unable to unpack %s header",
1281 if (parse_sha1_header(hdr
, type
, &size
) < 0)
1282 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1288 inflateEnd(&stream
);
1289 munmap(map
, mapsize
);
1293 static void *read_packed_sha1(const unsigned char *sha1
, char *type
, unsigned long *size
)
1295 struct pack_entry e
;
1297 if (!find_pack_entry(sha1
, &e
)) {
1298 error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1301 return unpack_entry(&e
, type
, size
);
1304 void * read_sha1_file(const unsigned char *sha1
, char *type
, unsigned long *size
)
1306 unsigned long mapsize
;
1308 struct pack_entry e
;
1310 if (find_pack_entry(sha1
, &e
))
1311 return read_packed_sha1(sha1
, type
, size
);
1312 map
= map_sha1_file(sha1
, &mapsize
);
1314 buf
= unpack_sha1_file(map
, mapsize
, type
, size
);
1315 munmap(map
, mapsize
);
1318 reprepare_packed_git();
1319 if (find_pack_entry(sha1
, &e
))
1320 return read_packed_sha1(sha1
, type
, size
);
1324 void *read_object_with_reference(const unsigned char *sha1
,
1325 const char *required_type
,
1326 unsigned long *size
,
1327 unsigned char *actual_sha1_return
)
1331 unsigned long isize
;
1332 unsigned char actual_sha1
[20];
1334 hashcpy(actual_sha1
, sha1
);
1336 int ref_length
= -1;
1337 const char *ref_type
= NULL
;
1339 buffer
= read_sha1_file(actual_sha1
, type
, &isize
);
1342 if (!strcmp(type
, required_type
)) {
1344 if (actual_sha1_return
)
1345 hashcpy(actual_sha1_return
, actual_sha1
);
1348 /* Handle references */
1349 else if (!strcmp(type
, commit_type
))
1351 else if (!strcmp(type
, tag_type
))
1352 ref_type
= "object ";
1357 ref_length
= strlen(ref_type
);
1359 if (memcmp(buffer
, ref_type
, ref_length
) ||
1360 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1365 /* Now we have the ID of the referred-to object in
1366 * actual_sha1. Check again. */
1370 char *write_sha1_file_prepare(void *buf
,
1373 unsigned char *sha1
,
1379 /* Generate the header */
1380 *hdrlen
= sprintf((char *)hdr
, "%s %lu", type
, len
)+1;
1384 SHA1_Update(&c
, hdr
, *hdrlen
);
1385 SHA1_Update(&c
, buf
, len
);
1386 SHA1_Final(sha1
, &c
);
1388 return sha1_file_name(sha1
);
1392 * Link the tempfile to the final place, possibly creating the
1393 * last directory level as you do so.
1395 * Returns the errno on failure, 0 on success.
1397 static int link_temp_to_file(const char *tmpfile
, char *filename
)
1402 if (!link(tmpfile
, filename
))
1406 * Try to mkdir the last path component if that failed.
1408 * Re-try the "link()" regardless of whether the mkdir
1409 * succeeds, since a race might mean that somebody
1413 dir
= strrchr(filename
, '/');
1416 mkdir(filename
, 0777);
1417 if (adjust_shared_perm(filename
))
1420 if (!link(tmpfile
, filename
))
1428 * Move the just written object into its final resting place
1430 int move_temp_to_file(const char *tmpfile
, char *filename
)
1432 int ret
= link_temp_to_file(tmpfile
, filename
);
1435 * Coda hack - coda doesn't like cross-directory links,
1436 * so we fall back to a rename, which will mean that it
1437 * won't be able to check collisions, but that's not a
1440 * The same holds for FAT formatted media.
1442 * When this succeeds, we just return 0. We have nothing
1445 if (ret
&& ret
!= EEXIST
) {
1446 if (!rename(tmpfile
, filename
))
1452 if (ret
!= EEXIST
) {
1453 fprintf(stderr
, "unable to write sha1 filename %s: %s\n", filename
, strerror(ret
));
1456 /* FIXME!!! Collision check here ? */
1462 static int write_buffer(int fd
, const void *buf
, size_t len
)
1467 size
= write(fd
, buf
, len
);
1469 return error("file write: disk full");
1471 if (errno
== EINTR
|| errno
== EAGAIN
)
1473 return error("file write error (%s)", strerror(errno
));
1476 buf
= (char *) buf
+ size
;
1481 static int write_binary_header(unsigned char *hdr
, enum object_type type
, unsigned long len
)
1486 c
= (type
<< 4) | (len
& 15);
1499 static void setup_object_header(z_stream
*stream
, const char *type
, unsigned long len
)
1503 if (use_legacy_headers
) {
1504 while (deflate(stream
, 0) == Z_OK
)
1508 if (!strcmp(type
, blob_type
))
1509 obj_type
= OBJ_BLOB
;
1510 else if (!strcmp(type
, tree_type
))
1511 obj_type
= OBJ_TREE
;
1512 else if (!strcmp(type
, commit_type
))
1513 obj_type
= OBJ_COMMIT
;
1514 else if (!strcmp(type
, tag_type
))
1517 die("trying to generate bogus object of type '%s'", type
);
1518 hdr
= write_binary_header(stream
->next_out
, obj_type
, len
);
1519 stream
->total_out
= hdr
;
1520 stream
->next_out
+= hdr
;
1521 stream
->avail_out
-= hdr
;
1524 int write_sha1_file(void *buf
, unsigned long len
, const char *type
, unsigned char *returnsha1
)
1527 unsigned char *compressed
;
1529 unsigned char sha1
[20];
1531 static char tmpfile
[PATH_MAX
];
1532 unsigned char hdr
[50];
1535 /* Normally if we have it in the pack then we do not bother writing
1536 * it out into .git/objects/??/?{38} file.
1538 filename
= write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1540 hashcpy(returnsha1
, sha1
);
1541 if (has_sha1_file(sha1
))
1543 fd
= open(filename
, O_RDONLY
);
1546 * FIXME!!! We might do collision checking here, but we'd
1547 * need to uncompress the old file and check it. Later.
1553 if (errno
!= ENOENT
) {
1554 fprintf(stderr
, "sha1 file %s: %s\n", filename
, strerror(errno
));
1558 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1560 fd
= mkstemp(tmpfile
);
1562 fprintf(stderr
, "unable to create temporary sha1 filename %s: %s\n", tmpfile
, strerror(errno
));
1567 memset(&stream
, 0, sizeof(stream
));
1568 deflateInit(&stream
, zlib_compression_level
);
1569 size
= 8 + deflateBound(&stream
, len
+hdrlen
);
1570 compressed
= xmalloc(size
);
1573 stream
.next_out
= compressed
;
1574 stream
.avail_out
= size
;
1576 /* First header.. */
1577 stream
.next_in
= hdr
;
1578 stream
.avail_in
= hdrlen
;
1579 setup_object_header(&stream
, type
, len
);
1581 /* Then the data itself.. */
1582 stream
.next_in
= buf
;
1583 stream
.avail_in
= len
;
1584 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1586 deflateEnd(&stream
);
1587 size
= stream
.total_out
;
1589 if (write_buffer(fd
, compressed
, size
) < 0)
1590 die("unable to write sha1 file");
1595 return move_temp_to_file(tmpfile
, filename
);
1599 * We need to unpack and recompress the object for writing
1600 * it out to a different file.
1602 static void *repack_object(const unsigned char *sha1
, unsigned long *objsize
)
1606 unsigned char *unpacked
;
1613 /* need to unpack and recompress it by itself */
1614 unpacked
= read_packed_sha1(sha1
, type
, &len
);
1616 hdrlen
= sprintf(hdr
, "%s %lu", type
, len
) + 1;
1619 memset(&stream
, 0, sizeof(stream
));
1620 deflateInit(&stream
, zlib_compression_level
);
1621 size
= deflateBound(&stream
, len
+ hdrlen
);
1622 buf
= xmalloc(size
);
1625 stream
.next_out
= buf
;
1626 stream
.avail_out
= size
;
1628 /* First header.. */
1629 stream
.next_in
= (void *)hdr
;
1630 stream
.avail_in
= hdrlen
;
1631 while (deflate(&stream
, 0) == Z_OK
)
1634 /* Then the data itself.. */
1635 stream
.next_in
= unpacked
;
1636 stream
.avail_in
= len
;
1637 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1639 deflateEnd(&stream
);
1642 *objsize
= stream
.total_out
;
1646 int write_sha1_to_fd(int fd
, const unsigned char *sha1
)
1649 unsigned long objsize
;
1650 void *buf
= map_sha1_file(sha1
, &objsize
);
1653 retval
= write_buffer(fd
, buf
, objsize
);
1654 munmap(buf
, objsize
);
1658 buf
= repack_object(sha1
, &objsize
);
1659 retval
= write_buffer(fd
, buf
, objsize
);
1664 int write_sha1_from_fd(const unsigned char *sha1
, int fd
, char *buffer
,
1665 size_t bufsize
, size_t *bufposn
)
1667 char tmpfile
[PATH_MAX
];
1670 unsigned char real_sha1
[20];
1671 unsigned char discard
[4096];
1675 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1677 local
= mkstemp(tmpfile
);
1679 return error("Couldn't open %s for %s",
1680 tmpfile
, sha1_to_hex(sha1
));
1682 memset(&stream
, 0, sizeof(stream
));
1684 inflateInit(&stream
);
1691 stream
.avail_in
= *bufposn
;
1692 stream
.next_in
= (unsigned char *) buffer
;
1694 stream
.next_out
= discard
;
1695 stream
.avail_out
= sizeof(discard
);
1696 ret
= inflate(&stream
, Z_SYNC_FLUSH
);
1697 SHA1_Update(&c
, discard
, sizeof(discard
) -
1699 } while (stream
.avail_in
&& ret
== Z_OK
);
1700 if (write_buffer(local
, buffer
, *bufposn
- stream
.avail_in
) < 0)
1701 die("unable to write sha1 file");
1702 memmove(buffer
, buffer
+ *bufposn
- stream
.avail_in
,
1704 *bufposn
= stream
.avail_in
;
1708 size
= read(fd
, buffer
+ *bufposn
, bufsize
- *bufposn
);
1713 return error("Connection closed?");
1714 perror("Reading from connection");
1719 inflateEnd(&stream
);
1722 SHA1_Final(real_sha1
, &c
);
1723 if (ret
!= Z_STREAM_END
) {
1725 return error("File %s corrupted", sha1_to_hex(sha1
));
1727 if (hashcmp(sha1
, real_sha1
)) {
1729 return error("File %s has bad hash", sha1_to_hex(sha1
));
1732 return move_temp_to_file(tmpfile
, sha1_file_name(sha1
));
1735 int has_pack_index(const unsigned char *sha1
)
1738 if (stat(sha1_pack_index_name(sha1
), &st
))
1743 int has_pack_file(const unsigned char *sha1
)
1746 if (stat(sha1_pack_name(sha1
), &st
))
1751 int has_sha1_pack(const unsigned char *sha1
)
1753 struct pack_entry e
;
1754 return find_pack_entry(sha1
, &e
);
1757 int has_sha1_file(const unsigned char *sha1
)
1760 struct pack_entry e
;
1762 if (find_pack_entry(sha1
, &e
))
1764 return find_sha1_file(sha1
, &st
) ? 1 : 0;
1768 * reads from fd as long as possible into a supplied buffer of size bytes.
1769 * If necessary the buffer's size is increased using realloc()
1771 * returns 0 if anything went fine and -1 otherwise
1773 * NOTE: both buf and size may change, but even when -1 is returned
1774 * you still have to free() it yourself.
1776 int read_pipe(int fd
, char** return_buf
, unsigned long* return_size
)
1778 char* buf
= *return_buf
;
1779 unsigned long size
= *return_size
;
1781 unsigned long off
= 0;
1784 iret
= xread(fd
, buf
+ off
, size
- off
);
1789 buf
= realloc(buf
, size
);
1802 int index_pipe(unsigned char *sha1
, int fd
, const char *type
, int write_object
)
1804 unsigned long size
= 4096;
1805 char *buf
= malloc(size
);
1807 unsigned char hdr
[50];
1810 if (read_pipe(fd
, &buf
, &size
)) {
1818 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1820 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1827 int index_fd(unsigned char *sha1
, int fd
, struct stat
*st
, int write_object
, const char *type
)
1829 unsigned long size
= st
->st_size
;
1832 unsigned char hdr
[50];
1837 buf
= mmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1839 if (buf
== MAP_FAILED
)
1845 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1847 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1855 int index_path(unsigned char *sha1
, const char *path
, struct stat
*st
, int write_object
)
1860 switch (st
->st_mode
& S_IFMT
) {
1862 fd
= open(path
, O_RDONLY
);
1864 return error("open(\"%s\"): %s", path
,
1866 if (index_fd(sha1
, fd
, st
, write_object
, NULL
) < 0)
1867 return error("%s: failed to insert into database",
1871 target
= xmalloc(st
->st_size
+1);
1872 if (readlink(path
, target
, st
->st_size
+1) != st
->st_size
) {
1873 char *errstr
= strerror(errno
);
1875 return error("readlink(\"%s\"): %s", path
,
1878 if (!write_object
) {
1879 unsigned char hdr
[50];
1881 write_sha1_file_prepare(target
, st
->st_size
, blob_type
,
1882 sha1
, hdr
, &hdrlen
);
1883 } else if (write_sha1_file(target
, st
->st_size
, blob_type
, sha1
))
1884 return error("%s: failed to insert into database",
1889 return error("%s: unsupported file type", path
);