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 "xstrdup()" 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 unsigned long unpack_object_header_gently(const unsigned char *buf
, unsigned long len
, enum object_type
*type
, unsigned long *sizep
)
719 unsigned long used
= 0;
722 *type
= (c
>> 4) & 7;
728 if (sizeof(long) * 8 <= shift
)
731 size
+= (c
& 0x7f) << shift
;
738 static int unpack_sha1_header(z_stream
*stream
, unsigned char *map
, unsigned long mapsize
, void *buffer
, unsigned long bufsiz
)
740 unsigned long size
, used
;
741 static const char valid_loose_object_type
[8] = {
743 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
744 0, /* "delta" and others are invalid in a loose object */
746 enum object_type type
;
748 /* Get the data stream */
749 memset(stream
, 0, sizeof(*stream
));
750 stream
->next_in
= map
;
751 stream
->avail_in
= mapsize
;
752 stream
->next_out
= buffer
;
753 stream
->avail_out
= bufsiz
;
755 if (legacy_loose_object(map
)) {
757 return inflate(stream
, 0);
760 used
= unpack_object_header_gently(map
, mapsize
, &type
, &size
);
761 if (!used
|| !valid_loose_object_type
[type
])
766 /* Set up the stream for the rest.. */
767 stream
->next_in
= map
;
768 stream
->avail_in
= mapsize
;
771 /* And generate the fake traditional header */
772 stream
->total_out
= 1 + snprintf(buffer
, bufsiz
, "%s %lu",
773 type_names
[type
], size
);
777 static void *unpack_sha1_rest(z_stream
*stream
, void *buffer
, unsigned long size
)
779 int bytes
= strlen(buffer
) + 1;
780 unsigned char *buf
= xmalloc(1+size
);
783 n
= stream
->total_out
- bytes
;
786 memcpy(buf
, (char *) buffer
+ bytes
, n
);
789 stream
->next_out
= buf
+ bytes
;
790 stream
->avail_out
= size
- bytes
;
791 while (inflate(stream
, Z_FINISH
) == Z_OK
)
800 * We used to just use "sscanf()", but that's actually way
801 * too permissive for what we want to check. So do an anal
802 * object header parse by hand.
804 static int parse_sha1_header(char *hdr
, char *type
, unsigned long *sizep
)
810 * The type can be at most ten bytes (including the
811 * terminating '\0' that we add), and is followed by
826 * The length must follow immediately, and be in canonical
827 * decimal format (ie "010" is not valid).
834 unsigned long c
= *hdr
- '0';
838 size
= size
* 10 + c
;
844 * The length must be followed by a zero byte
846 return *hdr
? -1 : 0;
849 void * unpack_sha1_file(void *map
, unsigned long mapsize
, char *type
, unsigned long *size
)
855 ret
= unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
));
856 if (ret
< Z_OK
|| parse_sha1_header(hdr
, type
, size
) < 0)
859 return unpack_sha1_rest(&stream
, hdr
, *size
);
862 /* forward declaration for a mutually recursive function */
863 static int packed_object_info(struct pack_entry
*entry
,
864 char *type
, unsigned long *sizep
);
866 static int packed_delta_info(unsigned char *base_sha1
,
867 unsigned long delta_size
,
870 unsigned long *sizep
,
871 struct packed_git
*p
)
873 struct pack_entry base_ent
;
876 die("truncated pack file");
878 /* The base entry _must_ be in the same pack */
879 if (!find_pack_entry_one(base_sha1
, &base_ent
, p
))
880 die("failed to find delta-pack base object %s",
881 sha1_to_hex(base_sha1
));
883 /* We choose to only get the type of the base object and
884 * ignore potentially corrupt pack file that expects the delta
885 * based on a base with a wrong size. This saves tons of
889 if (packed_object_info(&base_ent
, type
, NULL
))
890 die("cannot get info for delta-pack base");
893 const unsigned char *data
;
894 unsigned char delta_head
[64];
895 unsigned long result_size
;
899 memset(&stream
, 0, sizeof(stream
));
901 data
= stream
.next_in
= base_sha1
+ 20;
902 stream
.avail_in
= left
- 20;
903 stream
.next_out
= delta_head
;
904 stream
.avail_out
= sizeof(delta_head
);
906 inflateInit(&stream
);
907 st
= inflate(&stream
, Z_FINISH
);
909 if ((st
!= Z_STREAM_END
) &&
910 stream
.total_out
!= sizeof(delta_head
))
911 die("delta data unpack-initial failed");
913 /* Examine the initial part of the delta to figure out
918 /* ignore base size */
919 get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
921 /* Read the result size */
922 result_size
= get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
923 *sizep
= result_size
;
928 static unsigned long unpack_object_header(struct packed_git
*p
, unsigned long offset
,
929 enum object_type
*type
, unsigned long *sizep
)
933 if (p
->pack_size
<= offset
)
934 die("object offset outside of pack file");
936 used
= unpack_object_header_gently((unsigned char *)p
->pack_base
+
938 p
->pack_size
- offset
, type
, sizep
);
940 die("object offset outside of pack file");
942 return offset
+ used
;
945 int check_reuse_pack_delta(struct packed_git
*p
, unsigned long offset
,
946 unsigned char *base
, unsigned long *sizep
,
947 enum object_type
*kindp
)
954 ptr
= unpack_object_header(p
, ptr
, kindp
, sizep
);
955 if (*kindp
!= OBJ_DELTA
)
957 hashcpy(base
, (unsigned char *) p
->pack_base
+ ptr
);
964 void packed_object_info_detail(struct pack_entry
*e
,
967 unsigned long *store_size
,
968 unsigned int *delta_chain_length
,
969 unsigned char *base_sha1
)
971 struct packed_git
*p
= e
->p
;
972 unsigned long offset
;
974 enum object_type kind
;
976 offset
= unpack_object_header(p
, e
->offset
, &kind
, size
);
977 pack
= (unsigned char *) p
->pack_base
+ offset
;
978 if (kind
!= OBJ_DELTA
)
979 *delta_chain_length
= 0;
981 unsigned int chain_length
= 0;
982 if (p
->pack_size
<= offset
+ 20)
983 die("pack file %s records an incomplete delta base",
985 hashcpy(base_sha1
, pack
);
987 struct pack_entry base_ent
;
990 find_pack_entry_one(pack
, &base_ent
, p
);
991 offset
= unpack_object_header(p
, base_ent
.offset
,
993 pack
= (unsigned char *) p
->pack_base
+ offset
;
995 } while (kind
== OBJ_DELTA
);
996 *delta_chain_length
= chain_length
;
1003 strcpy(type
, type_names
[kind
]);
1006 die("corrupted pack file %s containing object of kind %d",
1007 p
->pack_name
, kind
);
1009 *store_size
= 0; /* notyet */
1012 static int packed_object_info(struct pack_entry
*entry
,
1013 char *type
, unsigned long *sizep
)
1015 struct packed_git
*p
= entry
->p
;
1016 unsigned long offset
, size
, left
;
1017 unsigned char *pack
;
1018 enum object_type kind
;
1021 if (use_packed_git(p
))
1022 die("cannot map packed file");
1024 offset
= unpack_object_header(p
, entry
->offset
, &kind
, &size
);
1025 pack
= (unsigned char *) p
->pack_base
+ offset
;
1026 left
= p
->pack_size
- offset
;
1030 retval
= packed_delta_info(pack
, size
, left
, type
, sizep
, p
);
1031 unuse_packed_git(p
);
1037 strcpy(type
, type_names
[kind
]);
1040 die("corrupted pack file %s containing object of kind %d",
1041 p
->pack_name
, kind
);
1045 unuse_packed_git(p
);
1049 static void *unpack_compressed_entry(struct packed_git
*p
,
1050 unsigned long offset
,
1055 unsigned char *buffer
;
1057 buffer
= xmalloc(size
+ 1);
1059 memset(&stream
, 0, sizeof(stream
));
1060 stream
.next_in
= (unsigned char*)p
->pack_base
+ offset
;
1061 stream
.avail_in
= p
->pack_size
- offset
;
1062 stream
.next_out
= buffer
;
1063 stream
.avail_out
= size
;
1065 inflateInit(&stream
);
1066 st
= inflate(&stream
, Z_FINISH
);
1067 inflateEnd(&stream
);
1068 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= size
) {
1076 static void *unpack_delta_entry(struct packed_git
*p
,
1077 unsigned long offset
,
1078 unsigned long delta_size
,
1080 unsigned long *sizep
)
1082 struct pack_entry base_ent
;
1083 void *delta_data
, *result
, *base
;
1084 unsigned long result_size
, base_size
;
1085 unsigned char* base_sha1
;
1087 if ((offset
+ 20) >= p
->pack_size
)
1088 die("truncated pack file");
1090 /* The base entry _must_ be in the same pack */
1091 base_sha1
= (unsigned char*)p
->pack_base
+ offset
;
1092 if (!find_pack_entry_one(base_sha1
, &base_ent
, p
))
1093 die("failed to find delta-pack base object %s",
1094 sha1_to_hex(base_sha1
));
1095 base
= unpack_entry_gently(&base_ent
, type
, &base_size
);
1097 die("failed to read delta-pack base object %s",
1098 sha1_to_hex(base_sha1
));
1100 delta_data
= unpack_compressed_entry(p
, offset
+ 20, delta_size
);
1101 result
= patch_delta(base
, base_size
,
1102 delta_data
, delta_size
,
1105 die("failed to apply delta");
1108 *sizep
= result_size
;
1112 static void *unpack_entry(struct pack_entry
*entry
,
1113 char *type
, unsigned long *sizep
)
1115 struct packed_git
*p
= entry
->p
;
1118 if (use_packed_git(p
))
1119 die("cannot map packed file");
1120 retval
= unpack_entry_gently(entry
, type
, sizep
);
1121 unuse_packed_git(p
);
1123 die("corrupted pack file %s", p
->pack_name
);
1127 /* The caller is responsible for use_packed_git()/unuse_packed_git() pair */
1128 void *unpack_entry_gently(struct pack_entry
*entry
,
1129 char *type
, unsigned long *sizep
)
1131 struct packed_git
*p
= entry
->p
;
1132 unsigned long offset
, size
;
1133 enum object_type kind
;
1135 offset
= unpack_object_header(p
, entry
->offset
, &kind
, &size
);
1138 return unpack_delta_entry(p
, offset
, size
, type
, sizep
);
1143 strcpy(type
, type_names
[kind
]);
1145 return unpack_compressed_entry(p
, offset
, size
);
1151 int num_packed_objects(const struct packed_git
*p
)
1153 /* See check_packed_git_idx() */
1154 return (p
->index_size
- 20 - 20 - 4*256) / 24;
1157 int nth_packed_object_sha1(const struct packed_git
*p
, int n
,
1158 unsigned char* sha1
)
1160 void *index
= p
->index_base
+ 256;
1161 if (n
< 0 || num_packed_objects(p
) <= n
)
1163 hashcpy(sha1
, (unsigned char *) index
+ (24 * n
) + 4);
1167 int find_pack_entry_one(const unsigned char *sha1
,
1168 struct pack_entry
*e
, struct packed_git
*p
)
1170 unsigned int *level1_ofs
= p
->index_base
;
1171 int hi
= ntohl(level1_ofs
[*sha1
]);
1172 int lo
= ((*sha1
== 0x0) ? 0 : ntohl(level1_ofs
[*sha1
- 1]));
1173 void *index
= p
->index_base
+ 256;
1176 int mi
= (lo
+ hi
) / 2;
1177 int cmp
= hashcmp((unsigned char *)index
+ (24 * mi
) + 4, sha1
);
1179 e
->offset
= ntohl(*((unsigned int *) ((char *) index
+ (24 * mi
))));
1180 hashcpy(e
->sha1
, sha1
);
1192 static int find_pack_entry(const unsigned char *sha1
, struct pack_entry
*e
)
1194 struct packed_git
*p
;
1195 prepare_packed_git();
1197 for (p
= packed_git
; p
; p
= p
->next
) {
1198 if (find_pack_entry_one(sha1
, e
, p
))
1204 struct packed_git
*find_sha1_pack(const unsigned char *sha1
,
1205 struct packed_git
*packs
)
1207 struct packed_git
*p
;
1208 struct pack_entry e
;
1210 for (p
= packs
; p
; p
= p
->next
) {
1211 if (find_pack_entry_one(sha1
, &e
, p
))
1218 int sha1_object_info(const unsigned char *sha1
, char *type
, unsigned long *sizep
)
1221 unsigned long mapsize
, size
;
1226 map
= map_sha1_file(sha1
, &mapsize
);
1228 struct pack_entry e
;
1230 if (find_pack_entry(sha1
, &e
))
1231 return packed_object_info(&e
, type
, sizep
);
1232 reprepare_packed_git();
1233 if (find_pack_entry(sha1
, &e
))
1234 return packed_object_info(&e
, type
, sizep
);
1235 return error("unable to find %s", sha1_to_hex(sha1
));
1237 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1238 status
= error("unable to unpack %s header",
1240 if (parse_sha1_header(hdr
, type
, &size
) < 0)
1241 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1247 inflateEnd(&stream
);
1248 munmap(map
, mapsize
);
1252 static void *read_packed_sha1(const unsigned char *sha1
, char *type
, unsigned long *size
)
1254 struct pack_entry e
;
1256 if (!find_pack_entry(sha1
, &e
)) {
1257 error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1260 return unpack_entry(&e
, type
, size
);
1263 void * read_sha1_file(const unsigned char *sha1
, char *type
, unsigned long *size
)
1265 unsigned long mapsize
;
1267 struct pack_entry e
;
1269 if (find_pack_entry(sha1
, &e
))
1270 return read_packed_sha1(sha1
, type
, size
);
1271 map
= map_sha1_file(sha1
, &mapsize
);
1273 buf
= unpack_sha1_file(map
, mapsize
, type
, size
);
1274 munmap(map
, mapsize
);
1277 reprepare_packed_git();
1278 if (find_pack_entry(sha1
, &e
))
1279 return read_packed_sha1(sha1
, type
, size
);
1283 void *read_object_with_reference(const unsigned char *sha1
,
1284 const char *required_type
,
1285 unsigned long *size
,
1286 unsigned char *actual_sha1_return
)
1290 unsigned long isize
;
1291 unsigned char actual_sha1
[20];
1293 hashcpy(actual_sha1
, sha1
);
1295 int ref_length
= -1;
1296 const char *ref_type
= NULL
;
1298 buffer
= read_sha1_file(actual_sha1
, type
, &isize
);
1301 if (!strcmp(type
, required_type
)) {
1303 if (actual_sha1_return
)
1304 hashcpy(actual_sha1_return
, actual_sha1
);
1307 /* Handle references */
1308 else if (!strcmp(type
, commit_type
))
1310 else if (!strcmp(type
, tag_type
))
1311 ref_type
= "object ";
1316 ref_length
= strlen(ref_type
);
1318 if (memcmp(buffer
, ref_type
, ref_length
) ||
1319 get_sha1_hex((char *) buffer
+ ref_length
, actual_sha1
)) {
1324 /* Now we have the ID of the referred-to object in
1325 * actual_sha1. Check again. */
1329 char *write_sha1_file_prepare(void *buf
,
1332 unsigned char *sha1
,
1338 /* Generate the header */
1339 *hdrlen
= sprintf((char *)hdr
, "%s %lu", type
, len
)+1;
1343 SHA1_Update(&c
, hdr
, *hdrlen
);
1344 SHA1_Update(&c
, buf
, len
);
1345 SHA1_Final(sha1
, &c
);
1347 return sha1_file_name(sha1
);
1351 * Link the tempfile to the final place, possibly creating the
1352 * last directory level as you do so.
1354 * Returns the errno on failure, 0 on success.
1356 static int link_temp_to_file(const char *tmpfile
, const char *filename
)
1361 if (!link(tmpfile
, filename
))
1365 * Try to mkdir the last path component if that failed.
1367 * Re-try the "link()" regardless of whether the mkdir
1368 * succeeds, since a race might mean that somebody
1372 dir
= strrchr(filename
, '/');
1375 mkdir(filename
, 0777);
1376 if (adjust_shared_perm(filename
))
1379 if (!link(tmpfile
, filename
))
1387 * Move the just written object into its final resting place
1389 int move_temp_to_file(const char *tmpfile
, const char *filename
)
1391 int ret
= link_temp_to_file(tmpfile
, filename
);
1394 * Coda hack - coda doesn't like cross-directory links,
1395 * so we fall back to a rename, which will mean that it
1396 * won't be able to check collisions, but that's not a
1399 * The same holds for FAT formatted media.
1401 * When this succeeds, we just return 0. We have nothing
1404 if (ret
&& ret
!= EEXIST
) {
1405 if (!rename(tmpfile
, filename
))
1411 if (ret
!= EEXIST
) {
1412 fprintf(stderr
, "unable to write sha1 filename %s: %s\n", filename
, strerror(ret
));
1415 /* FIXME!!! Collision check here ? */
1421 static int write_buffer(int fd
, const void *buf
, size_t len
)
1426 size
= write(fd
, buf
, len
);
1428 return error("file write: disk full");
1430 if (errno
== EINTR
|| errno
== EAGAIN
)
1432 return error("file write error (%s)", strerror(errno
));
1435 buf
= (char *) buf
+ size
;
1440 static int write_binary_header(unsigned char *hdr
, enum object_type type
, unsigned long len
)
1445 c
= (type
<< 4) | (len
& 15);
1458 static void setup_object_header(z_stream
*stream
, const char *type
, unsigned long len
)
1462 if (use_legacy_headers
) {
1463 while (deflate(stream
, 0) == Z_OK
)
1467 if (!strcmp(type
, blob_type
))
1468 obj_type
= OBJ_BLOB
;
1469 else if (!strcmp(type
, tree_type
))
1470 obj_type
= OBJ_TREE
;
1471 else if (!strcmp(type
, commit_type
))
1472 obj_type
= OBJ_COMMIT
;
1473 else if (!strcmp(type
, tag_type
))
1476 die("trying to generate bogus object of type '%s'", type
);
1477 hdr
= write_binary_header(stream
->next_out
, obj_type
, len
);
1478 stream
->total_out
= hdr
;
1479 stream
->next_out
+= hdr
;
1480 stream
->avail_out
-= hdr
;
1483 int write_sha1_file(void *buf
, unsigned long len
, const char *type
, unsigned char *returnsha1
)
1486 unsigned char *compressed
;
1488 unsigned char sha1
[20];
1490 static char tmpfile
[PATH_MAX
];
1491 unsigned char hdr
[50];
1494 /* Normally if we have it in the pack then we do not bother writing
1495 * it out into .git/objects/??/?{38} file.
1497 filename
= write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1499 hashcpy(returnsha1
, sha1
);
1500 if (has_sha1_file(sha1
))
1502 fd
= open(filename
, O_RDONLY
);
1505 * FIXME!!! We might do collision checking here, but we'd
1506 * need to uncompress the old file and check it. Later.
1512 if (errno
!= ENOENT
) {
1513 fprintf(stderr
, "sha1 file %s: %s\n", filename
, strerror(errno
));
1517 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1519 fd
= mkstemp(tmpfile
);
1521 fprintf(stderr
, "unable to create temporary sha1 filename %s: %s\n", tmpfile
, strerror(errno
));
1526 memset(&stream
, 0, sizeof(stream
));
1527 deflateInit(&stream
, zlib_compression_level
);
1528 size
= 8 + deflateBound(&stream
, len
+hdrlen
);
1529 compressed
= xmalloc(size
);
1532 stream
.next_out
= compressed
;
1533 stream
.avail_out
= size
;
1535 /* First header.. */
1536 stream
.next_in
= hdr
;
1537 stream
.avail_in
= hdrlen
;
1538 setup_object_header(&stream
, type
, len
);
1540 /* Then the data itself.. */
1541 stream
.next_in
= buf
;
1542 stream
.avail_in
= len
;
1543 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1545 deflateEnd(&stream
);
1546 size
= stream
.total_out
;
1548 if (write_buffer(fd
, compressed
, size
) < 0)
1549 die("unable to write sha1 file");
1554 return move_temp_to_file(tmpfile
, filename
);
1558 * We need to unpack and recompress the object for writing
1559 * it out to a different file.
1561 static void *repack_object(const unsigned char *sha1
, unsigned long *objsize
)
1565 unsigned char *unpacked
;
1572 /* need to unpack and recompress it by itself */
1573 unpacked
= read_packed_sha1(sha1
, type
, &len
);
1575 hdrlen
= sprintf(hdr
, "%s %lu", type
, len
) + 1;
1578 memset(&stream
, 0, sizeof(stream
));
1579 deflateInit(&stream
, zlib_compression_level
);
1580 size
= deflateBound(&stream
, len
+ hdrlen
);
1581 buf
= xmalloc(size
);
1584 stream
.next_out
= buf
;
1585 stream
.avail_out
= size
;
1587 /* First header.. */
1588 stream
.next_in
= (void *)hdr
;
1589 stream
.avail_in
= hdrlen
;
1590 while (deflate(&stream
, 0) == Z_OK
)
1593 /* Then the data itself.. */
1594 stream
.next_in
= unpacked
;
1595 stream
.avail_in
= len
;
1596 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1598 deflateEnd(&stream
);
1601 *objsize
= stream
.total_out
;
1605 int write_sha1_to_fd(int fd
, const unsigned char *sha1
)
1608 unsigned long objsize
;
1609 void *buf
= map_sha1_file(sha1
, &objsize
);
1612 retval
= write_buffer(fd
, buf
, objsize
);
1613 munmap(buf
, objsize
);
1617 buf
= repack_object(sha1
, &objsize
);
1618 retval
= write_buffer(fd
, buf
, objsize
);
1623 int write_sha1_from_fd(const unsigned char *sha1
, int fd
, char *buffer
,
1624 size_t bufsize
, size_t *bufposn
)
1626 char tmpfile
[PATH_MAX
];
1629 unsigned char real_sha1
[20];
1630 unsigned char discard
[4096];
1634 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1636 local
= mkstemp(tmpfile
);
1638 return error("Couldn't open %s for %s",
1639 tmpfile
, sha1_to_hex(sha1
));
1641 memset(&stream
, 0, sizeof(stream
));
1643 inflateInit(&stream
);
1650 stream
.avail_in
= *bufposn
;
1651 stream
.next_in
= (unsigned char *) buffer
;
1653 stream
.next_out
= discard
;
1654 stream
.avail_out
= sizeof(discard
);
1655 ret
= inflate(&stream
, Z_SYNC_FLUSH
);
1656 SHA1_Update(&c
, discard
, sizeof(discard
) -
1658 } while (stream
.avail_in
&& ret
== Z_OK
);
1659 if (write_buffer(local
, buffer
, *bufposn
- stream
.avail_in
) < 0)
1660 die("unable to write sha1 file");
1661 memmove(buffer
, buffer
+ *bufposn
- stream
.avail_in
,
1663 *bufposn
= stream
.avail_in
;
1667 size
= read(fd
, buffer
+ *bufposn
, bufsize
- *bufposn
);
1672 return error("Connection closed?");
1673 perror("Reading from connection");
1678 inflateEnd(&stream
);
1681 SHA1_Final(real_sha1
, &c
);
1682 if (ret
!= Z_STREAM_END
) {
1684 return error("File %s corrupted", sha1_to_hex(sha1
));
1686 if (hashcmp(sha1
, real_sha1
)) {
1688 return error("File %s has bad hash", sha1_to_hex(sha1
));
1691 return move_temp_to_file(tmpfile
, sha1_file_name(sha1
));
1694 int has_pack_index(const unsigned char *sha1
)
1697 if (stat(sha1_pack_index_name(sha1
), &st
))
1702 int has_pack_file(const unsigned char *sha1
)
1705 if (stat(sha1_pack_name(sha1
), &st
))
1710 int has_sha1_pack(const unsigned char *sha1
)
1712 struct pack_entry e
;
1713 return find_pack_entry(sha1
, &e
);
1716 int has_sha1_file(const unsigned char *sha1
)
1719 struct pack_entry e
;
1721 if (find_pack_entry(sha1
, &e
))
1723 return find_sha1_file(sha1
, &st
) ? 1 : 0;
1727 * reads from fd as long as possible into a supplied buffer of size bytes.
1728 * If necessary the buffer's size is increased using realloc()
1730 * returns 0 if anything went fine and -1 otherwise
1732 * NOTE: both buf and size may change, but even when -1 is returned
1733 * you still have to free() it yourself.
1735 int read_pipe(int fd
, char** return_buf
, unsigned long* return_size
)
1737 char* buf
= *return_buf
;
1738 unsigned long size
= *return_size
;
1740 unsigned long off
= 0;
1743 iret
= xread(fd
, buf
+ off
, size
- off
);
1748 buf
= xrealloc(buf
, size
);
1761 int index_pipe(unsigned char *sha1
, int fd
, const char *type
, int write_object
)
1763 unsigned long size
= 4096;
1764 char *buf
= xmalloc(size
);
1766 unsigned char hdr
[50];
1769 if (read_pipe(fd
, &buf
, &size
)) {
1777 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1779 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1786 int index_fd(unsigned char *sha1
, int fd
, struct stat
*st
, int write_object
, const char *type
)
1788 unsigned long size
= st
->st_size
;
1791 unsigned char hdr
[50];
1796 buf
= mmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1798 if (buf
== MAP_FAILED
)
1804 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1806 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1814 int index_path(unsigned char *sha1
, const char *path
, struct stat
*st
, int write_object
)
1819 switch (st
->st_mode
& S_IFMT
) {
1821 fd
= open(path
, O_RDONLY
);
1823 return error("open(\"%s\"): %s", path
,
1825 if (index_fd(sha1
, fd
, st
, write_object
, NULL
) < 0)
1826 return error("%s: failed to insert into database",
1830 target
= xmalloc(st
->st_size
+1);
1831 if (readlink(path
, target
, st
->st_size
+1) != st
->st_size
) {
1832 char *errstr
= strerror(errno
);
1834 return error("readlink(\"%s\"): %s", path
,
1837 if (!write_object
) {
1838 unsigned char hdr
[50];
1840 write_sha1_file_prepare(target
, st
->st_size
, blob_type
,
1841 sha1
, hdr
, &hdrlen
);
1842 } else if (write_sha1_file(target
, st
->st_size
, blob_type
, sha1
))
1843 return error("%s: failed to insert into database",
1848 return error("%s: unsupported file type", path
);