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] = { 0, };
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 adjust_shared_perm(const char *path
)
58 if (!shared_repository
)
60 if (lstat(path
, &st
) < 0)
71 if (chmod(path
, mode
) < 0)
76 int safe_create_leading_directories(char *path
)
85 pos
= strchr(pos
, '/');
89 if (!stat(path
, &st
)) {
91 if (!S_ISDIR(st
.st_mode
)) {
96 else if (mkdir(path
, 0777)) {
100 else if (adjust_shared_perm(path
)) {
109 char * sha1_to_hex(const unsigned char *sha1
)
112 static char hexbuffer
[4][50];
113 static const char hex
[] = "0123456789abcdef";
114 char *buffer
= hexbuffer
[3 & ++bufno
], *buf
= buffer
;
117 for (i
= 0; i
< 20; i
++) {
118 unsigned int val
= *sha1
++;
119 *buf
++ = hex
[val
>> 4];
120 *buf
++ = hex
[val
& 0xf];
127 static void fill_sha1_path(char *pathbuf
, const unsigned char *sha1
)
130 for (i
= 0; i
< 20; i
++) {
131 static char hex
[] = "0123456789abcdef";
132 unsigned int val
= sha1
[i
];
133 char *pos
= pathbuf
+ i
*2 + (i
> 0);
134 *pos
++ = hex
[val
>> 4];
135 *pos
= hex
[val
& 0xf];
140 * NOTE! This returns a statically allocated buffer, so you have to be
141 * careful about using it. Do a "strdup()" if you need to save the
144 * Also note that this returns the location for creating. Reading
145 * SHA1 file can happen from any alternate directory listed in the
146 * DB_ENVIRONMENT environment variable if it is not found in
147 * the primary object database.
149 char *sha1_file_name(const unsigned char *sha1
)
151 static char *name
, *base
;
154 const char *sha1_file_directory
= get_object_directory();
155 int len
= strlen(sha1_file_directory
);
156 base
= xmalloc(len
+ 60);
157 memcpy(base
, sha1_file_directory
, len
);
158 memset(base
+len
, 0, 60);
161 name
= base
+ len
+ 1;
163 fill_sha1_path(name
, sha1
);
167 char *sha1_pack_name(const unsigned char *sha1
)
169 static const char hex
[] = "0123456789abcdef";
170 static char *name
, *base
, *buf
;
174 const char *sha1_file_directory
= get_object_directory();
175 int len
= strlen(sha1_file_directory
);
176 base
= xmalloc(len
+ 60);
177 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory
);
178 name
= base
+ len
+ 11;
183 for (i
= 0; i
< 20; i
++) {
184 unsigned int val
= *sha1
++;
185 *buf
++ = hex
[val
>> 4];
186 *buf
++ = hex
[val
& 0xf];
192 char *sha1_pack_index_name(const unsigned char *sha1
)
194 static const char hex
[] = "0123456789abcdef";
195 static char *name
, *base
, *buf
;
199 const char *sha1_file_directory
= get_object_directory();
200 int len
= strlen(sha1_file_directory
);
201 base
= xmalloc(len
+ 60);
202 sprintf(base
, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory
);
203 name
= base
+ len
+ 11;
208 for (i
= 0; i
< 20; i
++) {
209 unsigned int val
= *sha1
++;
210 *buf
++ = hex
[val
>> 4];
211 *buf
++ = hex
[val
& 0xf];
217 struct alternate_object_database
*alt_odb_list
;
218 static struct alternate_object_database
**alt_odb_tail
;
220 static void read_info_alternates(const char * alternates
, int depth
);
223 * Prepare alternate object database registry.
225 * The variable alt_odb_list points at the list of struct
226 * alternate_object_database. The elements on this list come from
227 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
228 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
229 * whose contents is similar to that environment variable but can be
230 * LF separated. Its base points at a statically allocated buffer that
231 * contains "/the/directory/corresponding/to/.git/objects/...", while
232 * its name points just after the slash at the end of ".git/objects/"
233 * in the example above, and has enough space to hold 40-byte hex
234 * SHA1, an extra slash for the first level indirection, and the
237 static int link_alt_odb_entry(const char * entry
, int len
, const char * relative_base
, int depth
)
240 const char *objdir
= get_object_directory();
241 struct alternate_object_database
*ent
;
242 struct alternate_object_database
*alt
;
243 /* 43 = 40-byte + 2 '/' + terminating NUL */
245 int entlen
= pfxlen
+ 43;
248 if (*entry
!= '/' && relative_base
) {
249 /* Relative alt-odb */
251 base_len
= strlen(relative_base
) + 1;
255 ent
= xmalloc(sizeof(*ent
) + entlen
);
257 if (*entry
!= '/' && relative_base
) {
258 memcpy(ent
->base
, relative_base
, base_len
- 1);
259 ent
->base
[base_len
- 1] = '/';
260 memcpy(ent
->base
+ base_len
, entry
, len
);
263 memcpy(ent
->base
, entry
, pfxlen
);
265 ent
->name
= ent
->base
+ pfxlen
+ 1;
266 ent
->base
[pfxlen
+ 3] = '/';
267 ent
->base
[pfxlen
] = ent
->base
[entlen
-1] = 0;
269 /* Detect cases where alternate disappeared */
270 if (stat(ent
->base
, &st
) || !S_ISDIR(st
.st_mode
)) {
271 error("object directory %s does not exist; "
272 "check .git/objects/info/alternates.",
278 /* Prevent the common mistake of listing the same
279 * thing twice, or object directory itself.
281 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
282 if (!memcmp(ent
->base
, alt
->base
, pfxlen
)) {
287 if (!memcmp(ent
->base
, objdir
, pfxlen
)) {
292 /* add the alternate entry */
294 alt_odb_tail
= &(ent
->next
);
297 /* recursively add alternates */
298 read_info_alternates(ent
->base
, depth
+ 1);
300 ent
->base
[pfxlen
] = '/';
305 static void link_alt_odb_entries(const char *alt
, const char *ep
, int sep
,
306 const char *relative_base
, int depth
)
308 const char *cp
, *last
;
311 error("%s: ignoring alternate object stores, nesting too deep.",
319 if (cp
< ep
&& *cp
== '#') {
320 while (cp
< ep
&& *cp
!= sep
)
325 while (cp
< ep
&& *cp
!= sep
)
328 if ((*last
!= '/') && depth
) {
329 error("%s: ignoring relative alternate object store %s",
330 relative_base
, last
);
332 link_alt_odb_entry(last
, cp
- last
,
333 relative_base
, depth
);
336 while (cp
< ep
&& *cp
== sep
)
342 static void read_info_alternates(const char * relative_base
, int depth
)
349 sprintf(path
, "%s/info/alternates", relative_base
);
350 fd
= open(path
, O_RDONLY
);
353 if (fstat(fd
, &st
) || (st
.st_size
== 0)) {
357 map
= mmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
359 if (map
== MAP_FAILED
)
362 link_alt_odb_entries(map
, map
+ st
.st_size
, '\n', relative_base
, depth
);
364 munmap(map
, st
.st_size
);
367 void prepare_alt_odb(void)
371 alt
= getenv(ALTERNATE_DB_ENVIRONMENT
);
376 alt_odb_tail
= &alt_odb_list
;
377 link_alt_odb_entries(alt
, alt
+ strlen(alt
), ':', NULL
, 0);
379 read_info_alternates(get_object_directory(), 0);
382 static char *find_sha1_file(const unsigned char *sha1
, struct stat
*st
)
384 char *name
= sha1_file_name(sha1
);
385 struct alternate_object_database
*alt
;
390 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
392 fill_sha1_path(name
, sha1
);
393 if (!stat(alt
->base
, st
))
399 #define PACK_MAX_SZ (1<<26)
400 static int pack_used_ctr
;
401 static unsigned long pack_mapped
;
402 struct packed_git
*packed_git
;
404 static int check_packed_git_idx(const char *path
, unsigned long *idx_size_
,
409 unsigned long idx_size
;
411 int fd
= open(path
, O_RDONLY
);
415 if (fstat(fd
, &st
)) {
419 idx_size
= st
.st_size
;
420 idx_map
= mmap(NULL
, idx_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
422 if (idx_map
== MAP_FAILED
)
427 *idx_size_
= idx_size
;
429 /* check index map */
430 if (idx_size
< 4*256 + 20 + 20)
431 return error("index file too small");
433 for (i
= 0; i
< 256; i
++) {
434 unsigned int n
= ntohl(index
[i
]);
436 return error("non-monotonic index");
442 * - 256 index entries 4 bytes each
443 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
444 * - 20-byte SHA1 of the packfile
445 * - 20-byte SHA1 file checksum
447 if (idx_size
!= 4*256 + nr
* 24 + 20 + 20)
448 return error("wrong index file size");
453 static int unuse_one_packed_git(void)
455 struct packed_git
*p
, *lru
= NULL
;
457 for (p
= packed_git
; p
; p
= p
->next
) {
458 if (p
->pack_use_cnt
|| !p
->pack_base
)
460 if (!lru
|| p
->pack_last_used
< lru
->pack_last_used
)
465 munmap(lru
->pack_base
, lru
->pack_size
);
466 lru
->pack_base
= NULL
;
470 void unuse_packed_git(struct packed_git
*p
)
475 int use_packed_git(struct packed_git
*p
)
479 // We created the struct before we had the pack
480 stat(p
->pack_name
, &st
);
481 if (!S_ISREG(st
.st_mode
))
482 die("packfile %s not a regular file", p
->pack_name
);
483 p
->pack_size
= st
.st_size
;
490 pack_mapped
+= p
->pack_size
;
491 while (PACK_MAX_SZ
< pack_mapped
&& unuse_one_packed_git())
493 fd
= open(p
->pack_name
, O_RDONLY
);
495 die("packfile %s cannot be opened", p
->pack_name
);
496 if (fstat(fd
, &st
)) {
498 die("packfile %s cannot be opened", p
->pack_name
);
500 if (st
.st_size
!= p
->pack_size
)
501 die("packfile %s size mismatch.", p
->pack_name
);
502 map
= mmap(NULL
, p
->pack_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
504 if (map
== MAP_FAILED
)
505 die("packfile %s cannot be mapped.", p
->pack_name
);
508 /* Check if the pack file matches with the index file.
511 if (memcmp((char*)(p
->index_base
) + p
->index_size
- 40,
512 p
->pack_base
+ p
->pack_size
- 20, 20)) {
514 die("packfile %s does not match index.", p
->pack_name
);
517 p
->pack_last_used
= pack_used_ctr
++;
522 struct packed_git
*add_packed_git(char *path
, int path_len
, int local
)
525 struct packed_git
*p
;
526 unsigned long idx_size
;
528 unsigned char sha1
[20];
530 if (check_packed_git_idx(path
, &idx_size
, &idx_map
))
533 /* do we have a corresponding .pack file? */
534 strcpy(path
+ path_len
- 4, ".pack");
535 if (stat(path
, &st
) || !S_ISREG(st
.st_mode
)) {
536 munmap(idx_map
, idx_size
);
539 /* ok, it looks sane as far as we can check without
540 * actually mapping the pack file.
542 p
= xmalloc(sizeof(*p
) + path_len
+ 2);
543 strcpy(p
->pack_name
, path
);
544 p
->index_size
= idx_size
;
545 p
->pack_size
= st
.st_size
;
546 p
->index_base
= idx_map
;
549 p
->pack_last_used
= 0;
551 p
->pack_local
= local
;
552 if ((path_len
> 44) && !get_sha1_hex(path
+ path_len
- 44, sha1
))
553 memcpy(p
->sha1
, sha1
, 20);
557 struct packed_git
*parse_pack_index(unsigned char *sha1
)
559 char *path
= sha1_pack_index_name(sha1
);
560 return parse_pack_index_file(sha1
, path
);
563 struct packed_git
*parse_pack_index_file(const unsigned char *sha1
, char *idx_path
)
565 struct packed_git
*p
;
566 unsigned long idx_size
;
570 if (check_packed_git_idx(idx_path
, &idx_size
, &idx_map
))
573 path
= sha1_pack_name(sha1
);
575 p
= xmalloc(sizeof(*p
) + strlen(path
) + 2);
576 strcpy(p
->pack_name
, path
);
577 p
->index_size
= idx_size
;
579 p
->index_base
= idx_map
;
582 p
->pack_last_used
= 0;
584 memcpy(p
->sha1
, sha1
, 20);
588 void install_packed_git(struct packed_git
*pack
)
590 pack
->next
= packed_git
;
594 static void prepare_packed_git_one(char *objdir
, int local
)
601 sprintf(path
, "%s/pack", objdir
);
606 error("unable to open object pack directory: %s: %s",
607 path
, strerror(errno
));
611 while ((de
= readdir(dir
)) != NULL
) {
612 int namelen
= strlen(de
->d_name
);
613 struct packed_git
*p
;
615 if (strcmp(de
->d_name
+ namelen
- 4, ".idx"))
618 /* we have .idx. Is it a file we can map? */
619 strcpy(path
+ len
, de
->d_name
);
620 for (p
= packed_git
; p
; p
= p
->next
) {
621 if (!memcmp(path
, p
->pack_name
, len
+ namelen
- 4))
626 p
= add_packed_git(path
, len
+ namelen
, local
);
629 p
->next
= packed_git
;
635 static int prepare_packed_git_run_once
= 0;
636 void prepare_packed_git(void)
638 struct alternate_object_database
*alt
;
640 if (prepare_packed_git_run_once
)
642 prepare_packed_git_one(get_object_directory(), 1);
644 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
) {
646 prepare_packed_git_one(alt
->base
, 0);
649 prepare_packed_git_run_once
= 1;
652 static void reprepare_packed_git(void)
654 prepare_packed_git_run_once
= 0;
655 prepare_packed_git();
658 int check_sha1_signature(const unsigned char *sha1
, void *map
, unsigned long size
, const char *type
)
661 unsigned char real_sha1
[20];
665 SHA1_Update(&c
, header
, 1+sprintf(header
, "%s %lu", type
, size
));
666 SHA1_Update(&c
, map
, size
);
667 SHA1_Final(real_sha1
, &c
);
668 return memcmp(sha1
, real_sha1
, 20) ? -1 : 0;
671 static void *map_sha1_file_internal(const unsigned char *sha1
,
677 char *filename
= find_sha1_file(sha1
, &st
);
683 fd
= open(filename
, O_RDONLY
| sha1_file_open_flag
);
685 /* See if it works without O_NOATIME */
686 switch (sha1_file_open_flag
) {
688 fd
= open(filename
, O_RDONLY
);
696 /* If it failed once, it will probably fail again.
697 * Stop using O_NOATIME
699 sha1_file_open_flag
= 0;
701 map
= mmap(NULL
, st
.st_size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
703 if (map
== MAP_FAILED
)
709 int unpack_sha1_header(z_stream
*stream
, void *map
, unsigned long mapsize
, void *buffer
, unsigned long size
)
711 /* Get the data stream */
712 memset(stream
, 0, sizeof(*stream
));
713 stream
->next_in
= map
;
714 stream
->avail_in
= mapsize
;
715 stream
->next_out
= buffer
;
716 stream
->avail_out
= size
;
719 return inflate(stream
, 0);
722 static void *unpack_sha1_rest(z_stream
*stream
, void *buffer
, unsigned long size
)
724 int bytes
= strlen(buffer
) + 1;
725 unsigned char *buf
= xmalloc(1+size
);
727 memcpy(buf
, buffer
+ bytes
, stream
->total_out
- bytes
);
728 bytes
= stream
->total_out
- bytes
;
730 stream
->next_out
= buf
+ bytes
;
731 stream
->avail_out
= size
- bytes
;
732 while (inflate(stream
, Z_FINISH
) == Z_OK
)
741 * We used to just use "sscanf()", but that's actually way
742 * too permissive for what we want to check. So do an anal
743 * object header parse by hand.
745 int parse_sha1_header(char *hdr
, char *type
, unsigned long *sizep
)
751 * The type can be at most ten bytes (including the
752 * terminating '\0' that we add), and is followed by
767 * The length must follow immediately, and be in canonical
768 * decimal format (ie "010" is not valid).
775 unsigned long c
= *hdr
- '0';
779 size
= size
* 10 + c
;
785 * The length must be followed by a zero byte
787 return *hdr
? -1 : 0;
790 void * unpack_sha1_file(void *map
, unsigned long mapsize
, char *type
, unsigned long *size
)
796 ret
= unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
));
797 if (ret
< Z_OK
|| parse_sha1_header(hdr
, type
, size
) < 0)
800 return unpack_sha1_rest(&stream
, hdr
, *size
);
803 /* forward declaration for a mutually recursive function */
804 static int packed_object_info(struct pack_entry
*entry
,
805 char *type
, unsigned long *sizep
);
807 static int packed_delta_info(unsigned char *base_sha1
,
808 unsigned long delta_size
,
811 unsigned long *sizep
,
812 struct packed_git
*p
)
814 struct pack_entry base_ent
;
817 die("truncated pack file");
819 /* The base entry _must_ be in the same pack */
820 if (!find_pack_entry_one(base_sha1
, &base_ent
, p
))
821 die("failed to find delta-pack base object %s",
822 sha1_to_hex(base_sha1
));
824 /* We choose to only get the type of the base object and
825 * ignore potentially corrupt pack file that expects the delta
826 * based on a base with a wrong size. This saves tons of
830 if (packed_object_info(&base_ent
, type
, NULL
))
831 die("cannot get info for delta-pack base");
834 const unsigned char *data
;
835 unsigned char delta_head
[64];
836 unsigned long result_size
;
840 memset(&stream
, 0, sizeof(stream
));
842 data
= stream
.next_in
= base_sha1
+ 20;
843 stream
.avail_in
= left
- 20;
844 stream
.next_out
= delta_head
;
845 stream
.avail_out
= sizeof(delta_head
);
847 inflateInit(&stream
);
848 st
= inflate(&stream
, Z_FINISH
);
850 if ((st
!= Z_STREAM_END
) &&
851 stream
.total_out
!= sizeof(delta_head
))
852 die("delta data unpack-initial failed");
854 /* Examine the initial part of the delta to figure out
859 /* ignore base size */
860 get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
862 /* Read the result size */
863 result_size
= get_delta_hdr_size(&data
, delta_head
+sizeof(delta_head
));
864 *sizep
= result_size
;
869 static unsigned long unpack_object_header(struct packed_git
*p
, unsigned long offset
,
870 enum object_type
*type
, unsigned long *sizep
)
873 unsigned char *pack
, c
;
876 if (offset
>= p
->pack_size
)
877 die("object offset outside of pack file");
879 pack
= p
->pack_base
+ offset
;
882 *type
= (c
>> 4) & 7;
886 if (offset
>= p
->pack_size
)
887 die("object offset outside of pack file");
890 size
+= (c
& 0x7f) << shift
;
897 int check_reuse_pack_delta(struct packed_git
*p
, unsigned long offset
,
898 unsigned char *base
, unsigned long *sizep
,
899 enum object_type
*kindp
)
906 ptr
= unpack_object_header(p
, ptr
, kindp
, sizep
);
907 if (*kindp
!= OBJ_DELTA
)
909 memcpy(base
, p
->pack_base
+ ptr
, 20);
916 void packed_object_info_detail(struct pack_entry
*e
,
919 unsigned long *store_size
,
920 unsigned int *delta_chain_length
,
921 unsigned char *base_sha1
)
923 struct packed_git
*p
= e
->p
;
924 unsigned long offset
;
926 enum object_type kind
;
928 offset
= unpack_object_header(p
, e
->offset
, &kind
, size
);
929 pack
= p
->pack_base
+ offset
;
930 if (kind
!= OBJ_DELTA
)
931 *delta_chain_length
= 0;
933 unsigned int chain_length
= 0;
934 if (p
->pack_size
<= offset
+ 20)
935 die("pack file %s records an incomplete delta base",
937 memcpy(base_sha1
, pack
, 20);
939 struct pack_entry base_ent
;
942 find_pack_entry_one(pack
, &base_ent
, p
);
943 offset
= unpack_object_header(p
, base_ent
.offset
,
945 pack
= p
->pack_base
+ offset
;
947 } while (kind
== OBJ_DELTA
);
948 *delta_chain_length
= chain_length
;
952 strcpy(type
, commit_type
);
955 strcpy(type
, tree_type
);
958 strcpy(type
, blob_type
);
961 strcpy(type
, tag_type
);
964 die("corrupted pack file %s containing object of kind %d",
967 *store_size
= 0; /* notyet */
970 static int packed_object_info(struct pack_entry
*entry
,
971 char *type
, unsigned long *sizep
)
973 struct packed_git
*p
= entry
->p
;
974 unsigned long offset
, size
, left
;
976 enum object_type kind
;
979 if (use_packed_git(p
))
980 die("cannot map packed file");
982 offset
= unpack_object_header(p
, entry
->offset
, &kind
, &size
);
983 pack
= p
->pack_base
+ offset
;
984 left
= p
->pack_size
- offset
;
988 retval
= packed_delta_info(pack
, size
, left
, type
, sizep
, p
);
992 strcpy(type
, commit_type
);
995 strcpy(type
, tree_type
);
998 strcpy(type
, blob_type
);
1001 strcpy(type
, tag_type
);
1004 die("corrupted pack file %s containing object of kind %d",
1005 p
->pack_name
, kind
);
1009 unuse_packed_git(p
);
1013 /* forward declaration for a mutually recursive function */
1014 static void *unpack_entry(struct pack_entry
*, char *, unsigned long *);
1016 static void *unpack_delta_entry(unsigned char *base_sha1
,
1017 unsigned long delta_size
,
1020 unsigned long *sizep
,
1021 struct packed_git
*p
)
1023 struct pack_entry base_ent
;
1024 void *data
, *delta_data
, *result
, *base
;
1025 unsigned long data_size
, result_size
, base_size
;
1030 die("truncated pack file");
1032 /* The base entry _must_ be in the same pack */
1033 if (!find_pack_entry_one(base_sha1
, &base_ent
, p
))
1034 die("failed to find delta-pack base object %s",
1035 sha1_to_hex(base_sha1
));
1036 base
= unpack_entry_gently(&base_ent
, type
, &base_size
);
1038 die("failed to read delta-pack base object %s",
1039 sha1_to_hex(base_sha1
));
1041 data
= base_sha1
+ 20;
1042 data_size
= left
- 20;
1043 delta_data
= xmalloc(delta_size
);
1045 memset(&stream
, 0, sizeof(stream
));
1047 stream
.next_in
= data
;
1048 stream
.avail_in
= data_size
;
1049 stream
.next_out
= delta_data
;
1050 stream
.avail_out
= delta_size
;
1052 inflateInit(&stream
);
1053 st
= inflate(&stream
, Z_FINISH
);
1054 inflateEnd(&stream
);
1055 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= delta_size
)
1056 die("delta data unpack failed");
1058 result
= patch_delta(base
, base_size
,
1059 delta_data
, delta_size
,
1062 die("failed to apply delta");
1065 *sizep
= result_size
;
1069 static void *unpack_non_delta_entry(unsigned char *data
,
1075 unsigned char *buffer
;
1077 buffer
= xmalloc(size
+ 1);
1079 memset(&stream
, 0, sizeof(stream
));
1080 stream
.next_in
= data
;
1081 stream
.avail_in
= left
;
1082 stream
.next_out
= buffer
;
1083 stream
.avail_out
= size
;
1085 inflateInit(&stream
);
1086 st
= inflate(&stream
, Z_FINISH
);
1087 inflateEnd(&stream
);
1088 if ((st
!= Z_STREAM_END
) || stream
.total_out
!= size
) {
1096 static void *unpack_entry(struct pack_entry
*entry
,
1097 char *type
, unsigned long *sizep
)
1099 struct packed_git
*p
= entry
->p
;
1102 if (use_packed_git(p
))
1103 die("cannot map packed file");
1104 retval
= unpack_entry_gently(entry
, type
, sizep
);
1105 unuse_packed_git(p
);
1107 die("corrupted pack file %s", p
->pack_name
);
1111 /* The caller is responsible for use_packed_git()/unuse_packed_git() pair */
1112 void *unpack_entry_gently(struct pack_entry
*entry
,
1113 char *type
, unsigned long *sizep
)
1115 struct packed_git
*p
= entry
->p
;
1116 unsigned long offset
, size
, left
;
1117 unsigned char *pack
;
1118 enum object_type kind
;
1121 offset
= unpack_object_header(p
, entry
->offset
, &kind
, &size
);
1122 pack
= p
->pack_base
+ offset
;
1123 left
= p
->pack_size
- offset
;
1126 retval
= unpack_delta_entry(pack
, size
, left
, type
, sizep
, p
);
1129 strcpy(type
, commit_type
);
1132 strcpy(type
, tree_type
);
1135 strcpy(type
, blob_type
);
1138 strcpy(type
, tag_type
);
1144 retval
= unpack_non_delta_entry(pack
, size
, left
);
1148 int num_packed_objects(const struct packed_git
*p
)
1150 /* See check_packed_git_idx() */
1151 return (p
->index_size
- 20 - 20 - 4*256) / 24;
1154 int nth_packed_object_sha1(const struct packed_git
*p
, int n
,
1155 unsigned char* sha1
)
1157 void *index
= p
->index_base
+ 256;
1158 if (n
< 0 || num_packed_objects(p
) <= n
)
1160 memcpy(sha1
, (index
+ 24 * n
+ 4), 20);
1164 int find_pack_entry_one(const unsigned char *sha1
,
1165 struct pack_entry
*e
, struct packed_git
*p
)
1167 unsigned int *level1_ofs
= p
->index_base
;
1168 int hi
= ntohl(level1_ofs
[*sha1
]);
1169 int lo
= ((*sha1
== 0x0) ? 0 : ntohl(level1_ofs
[*sha1
- 1]));
1170 void *index
= p
->index_base
+ 256;
1173 int mi
= (lo
+ hi
) / 2;
1174 int cmp
= memcmp(index
+ 24 * mi
+ 4, sha1
, 20);
1176 e
->offset
= ntohl(*((unsigned int *)(index
+ 24 * mi
)));
1177 memcpy(e
->sha1
, sha1
, 20);
1189 static int find_pack_entry(const unsigned char *sha1
, struct pack_entry
*e
)
1191 struct packed_git
*p
;
1192 prepare_packed_git();
1194 for (p
= packed_git
; p
; p
= p
->next
) {
1195 if (find_pack_entry_one(sha1
, e
, p
))
1201 struct packed_git
*find_sha1_pack(const unsigned char *sha1
,
1202 struct packed_git
*packs
)
1204 struct packed_git
*p
;
1205 struct pack_entry e
;
1207 for (p
= packs
; p
; p
= p
->next
) {
1208 if (find_pack_entry_one(sha1
, &e
, p
))
1215 int sha1_object_info(const unsigned char *sha1
, char *type
, unsigned long *sizep
)
1218 unsigned long mapsize
, size
;
1223 map
= map_sha1_file_internal(sha1
, &mapsize
);
1225 struct pack_entry e
;
1227 if (find_pack_entry(sha1
, &e
))
1228 return packed_object_info(&e
, type
, sizep
);
1229 reprepare_packed_git();
1230 if (find_pack_entry(sha1
, &e
))
1231 return packed_object_info(&e
, type
, sizep
);
1232 return error("unable to find %s", sha1_to_hex(sha1
));
1234 if (unpack_sha1_header(&stream
, map
, mapsize
, hdr
, sizeof(hdr
)) < 0)
1235 status
= error("unable to unpack %s header",
1237 if (parse_sha1_header(hdr
, type
, &size
) < 0)
1238 status
= error("unable to parse %s header", sha1_to_hex(sha1
));
1244 inflateEnd(&stream
);
1245 munmap(map
, mapsize
);
1249 static void *read_packed_sha1(const unsigned char *sha1
, char *type
, unsigned long *size
)
1251 struct pack_entry e
;
1253 if (!find_pack_entry(sha1
, &e
)) {
1254 error("cannot read sha1_file for %s", sha1_to_hex(sha1
));
1257 return unpack_entry(&e
, type
, size
);
1260 void * read_sha1_file(const unsigned char *sha1
, char *type
, unsigned long *size
)
1262 unsigned long mapsize
;
1264 struct pack_entry e
;
1266 if (find_pack_entry(sha1
, &e
))
1267 return read_packed_sha1(sha1
, type
, size
);
1268 map
= map_sha1_file_internal(sha1
, &mapsize
);
1270 buf
= unpack_sha1_file(map
, mapsize
, type
, size
);
1271 munmap(map
, mapsize
);
1274 reprepare_packed_git();
1275 if (find_pack_entry(sha1
, &e
))
1276 return read_packed_sha1(sha1
, type
, size
);
1280 void *read_object_with_reference(const unsigned char *sha1
,
1281 const char *required_type
,
1282 unsigned long *size
,
1283 unsigned char *actual_sha1_return
)
1287 unsigned long isize
;
1288 unsigned char actual_sha1
[20];
1290 memcpy(actual_sha1
, sha1
, 20);
1292 int ref_length
= -1;
1293 const char *ref_type
= NULL
;
1295 buffer
= read_sha1_file(actual_sha1
, type
, &isize
);
1298 if (!strcmp(type
, required_type
)) {
1300 if (actual_sha1_return
)
1301 memcpy(actual_sha1_return
, actual_sha1
, 20);
1304 /* Handle references */
1305 else if (!strcmp(type
, commit_type
))
1307 else if (!strcmp(type
, tag_type
))
1308 ref_type
= "object ";
1313 ref_length
= strlen(ref_type
);
1315 if (memcmp(buffer
, ref_type
, ref_length
) ||
1316 get_sha1_hex(buffer
+ ref_length
, actual_sha1
)) {
1321 /* Now we have the ID of the referred-to object in
1322 * actual_sha1. Check again. */
1326 char *write_sha1_file_prepare(void *buf
,
1329 unsigned char *sha1
,
1335 /* Generate the header */
1336 *hdrlen
= sprintf((char *)hdr
, "%s %lu", type
, len
)+1;
1340 SHA1_Update(&c
, hdr
, *hdrlen
);
1341 SHA1_Update(&c
, buf
, len
);
1342 SHA1_Final(sha1
, &c
);
1344 return sha1_file_name(sha1
);
1348 * Link the tempfile to the final place, possibly creating the
1349 * last directory level as you do so.
1351 * Returns the errno on failure, 0 on success.
1353 static int link_temp_to_file(const char *tmpfile
, char *filename
)
1357 if (!link(tmpfile
, filename
))
1361 * Try to mkdir the last path component if that failed
1364 * Re-try the "link()" regardless of whether the mkdir
1365 * succeeds, since a race might mean that somebody
1369 if (ret
== ENOENT
) {
1370 char *dir
= strrchr(filename
, '/');
1373 mkdir(filename
, 0777);
1374 if (adjust_shared_perm(filename
))
1377 if (!link(tmpfile
, filename
))
1386 * Move the just written object into its final resting place
1388 int move_temp_to_file(const char *tmpfile
, char *filename
)
1390 int ret
= link_temp_to_file(tmpfile
, filename
);
1393 * Coda hack - coda doesn't like cross-directory links,
1394 * so we fall back to a rename, which will mean that it
1395 * won't be able to check collisions, but that's not a
1398 * The same holds for FAT formatted media.
1400 * When this succeeds, we just return 0. We have nothing
1403 if (ret
&& ret
!= EEXIST
) {
1404 if (!rename(tmpfile
, filename
))
1410 if (ret
!= EEXIST
) {
1411 fprintf(stderr
, "unable to write sha1 filename %s: %s\n", filename
, strerror(ret
));
1414 /* FIXME!!! Collision check here ? */
1420 static int write_buffer(int fd
, const void *buf
, size_t len
)
1425 size
= write(fd
, buf
, len
);
1427 return error("file write: disk full");
1429 if (errno
== EINTR
|| errno
== EAGAIN
)
1431 return error("file write error (%s)", strerror(errno
));
1439 int write_sha1_file(void *buf
, unsigned long len
, const char *type
, unsigned char *returnsha1
)
1442 unsigned char *compressed
;
1444 unsigned char sha1
[20];
1446 static char tmpfile
[PATH_MAX
];
1447 unsigned char hdr
[50];
1450 /* Normally if we have it in the pack then we do not bother writing
1451 * it out into .git/objects/??/?{38} file.
1453 filename
= write_sha1_file_prepare(buf
, len
, type
, sha1
, hdr
, &hdrlen
);
1455 memcpy(returnsha1
, sha1
, 20);
1456 if (has_sha1_file(sha1
))
1458 fd
= open(filename
, O_RDONLY
);
1461 * FIXME!!! We might do collision checking here, but we'd
1462 * need to uncompress the old file and check it. Later.
1468 if (errno
!= ENOENT
) {
1469 fprintf(stderr
, "sha1 file %s: %s\n", filename
, strerror(errno
));
1473 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1475 fd
= mkstemp(tmpfile
);
1477 fprintf(stderr
, "unable to create temporary sha1 filename %s: %s\n", tmpfile
, strerror(errno
));
1482 memset(&stream
, 0, sizeof(stream
));
1483 deflateInit(&stream
, Z_BEST_COMPRESSION
);
1484 size
= deflateBound(&stream
, len
+hdrlen
);
1485 compressed
= xmalloc(size
);
1488 stream
.next_out
= compressed
;
1489 stream
.avail_out
= size
;
1491 /* First header.. */
1492 stream
.next_in
= hdr
;
1493 stream
.avail_in
= hdrlen
;
1494 while (deflate(&stream
, 0) == Z_OK
)
1497 /* Then the data itself.. */
1498 stream
.next_in
= buf
;
1499 stream
.avail_in
= len
;
1500 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1502 deflateEnd(&stream
);
1503 size
= stream
.total_out
;
1505 if (write_buffer(fd
, compressed
, size
) < 0)
1506 die("unable to write sha1 file");
1511 return move_temp_to_file(tmpfile
, filename
);
1515 * We need to unpack and recompress the object for writing
1516 * it out to a different file.
1518 static void *repack_object(const unsigned char *sha1
, unsigned long *objsize
)
1522 unsigned char *unpacked
;
1529 // need to unpack and recompress it by itself
1530 unpacked
= read_packed_sha1(sha1
, type
, &len
);
1532 hdrlen
= sprintf(hdr
, "%s %lu", type
, len
) + 1;
1535 memset(&stream
, 0, sizeof(stream
));
1536 deflateInit(&stream
, Z_BEST_COMPRESSION
);
1537 size
= deflateBound(&stream
, len
+ hdrlen
);
1538 buf
= xmalloc(size
);
1541 stream
.next_out
= buf
;
1542 stream
.avail_out
= size
;
1544 /* First header.. */
1545 stream
.next_in
= (void *)hdr
;
1546 stream
.avail_in
= hdrlen
;
1547 while (deflate(&stream
, 0) == Z_OK
)
1550 /* Then the data itself.. */
1551 stream
.next_in
= unpacked
;
1552 stream
.avail_in
= len
;
1553 while (deflate(&stream
, Z_FINISH
) == Z_OK
)
1555 deflateEnd(&stream
);
1558 *objsize
= stream
.total_out
;
1562 int write_sha1_to_fd(int fd
, const unsigned char *sha1
)
1565 unsigned long objsize
;
1566 void *buf
= map_sha1_file_internal(sha1
, &objsize
);
1569 retval
= write_buffer(fd
, buf
, objsize
);
1570 munmap(buf
, objsize
);
1574 buf
= repack_object(sha1
, &objsize
);
1575 retval
= write_buffer(fd
, buf
, objsize
);
1580 int write_sha1_from_fd(const unsigned char *sha1
, int fd
, char *buffer
,
1581 size_t bufsize
, size_t *bufposn
)
1583 char tmpfile
[PATH_MAX
];
1586 unsigned char real_sha1
[20];
1587 unsigned char discard
[4096];
1591 snprintf(tmpfile
, sizeof(tmpfile
), "%s/obj_XXXXXX", get_object_directory());
1593 local
= mkstemp(tmpfile
);
1595 return error("Couldn't open %s for %s",
1596 tmpfile
, sha1_to_hex(sha1
));
1598 memset(&stream
, 0, sizeof(stream
));
1600 inflateInit(&stream
);
1607 stream
.avail_in
= *bufposn
;
1608 stream
.next_in
= (unsigned char *) buffer
;
1610 stream
.next_out
= discard
;
1611 stream
.avail_out
= sizeof(discard
);
1612 ret
= inflate(&stream
, Z_SYNC_FLUSH
);
1613 SHA1_Update(&c
, discard
, sizeof(discard
) -
1615 } while (stream
.avail_in
&& ret
== Z_OK
);
1616 if (write_buffer(local
, buffer
, *bufposn
- stream
.avail_in
) < 0)
1617 die("unable to write sha1 file");
1618 memmove(buffer
, buffer
+ *bufposn
- stream
.avail_in
,
1620 *bufposn
= stream
.avail_in
;
1624 size
= read(fd
, buffer
+ *bufposn
, bufsize
- *bufposn
);
1629 return error("Connection closed?");
1630 perror("Reading from connection");
1635 inflateEnd(&stream
);
1638 SHA1_Final(real_sha1
, &c
);
1639 if (ret
!= Z_STREAM_END
) {
1641 return error("File %s corrupted", sha1_to_hex(sha1
));
1643 if (memcmp(sha1
, real_sha1
, 20)) {
1645 return error("File %s has bad hash", sha1_to_hex(sha1
));
1648 return move_temp_to_file(tmpfile
, sha1_file_name(sha1
));
1651 int has_pack_index(const unsigned char *sha1
)
1654 if (stat(sha1_pack_index_name(sha1
), &st
))
1659 int has_pack_file(const unsigned char *sha1
)
1662 if (stat(sha1_pack_name(sha1
), &st
))
1667 int has_sha1_pack(const unsigned char *sha1
)
1669 struct pack_entry e
;
1670 return find_pack_entry(sha1
, &e
);
1673 int has_sha1_file(const unsigned char *sha1
)
1676 struct pack_entry e
;
1678 if (find_pack_entry(sha1
, &e
))
1680 return find_sha1_file(sha1
, &st
) ? 1 : 0;
1684 * reads from fd as long as possible into a supplied buffer of size bytes.
1685 * If neccessary the buffer's size is increased using realloc()
1687 * returns 0 if anything went fine and -1 otherwise
1689 * NOTE: both buf and size may change, but even when -1 is returned
1690 * you still have to free() it yourself.
1692 int read_pipe(int fd
, char** return_buf
, unsigned long* return_size
)
1694 char* buf
= *return_buf
;
1695 unsigned long size
= *return_size
;
1697 unsigned long off
= 0;
1700 iret
= xread(fd
, buf
+ off
, size
- off
);
1705 buf
= realloc(buf
, size
);
1718 int index_pipe(unsigned char *sha1
, int fd
, const char *type
, int write_object
)
1720 unsigned long size
= 4096;
1721 char *buf
= malloc(size
);
1723 unsigned char hdr
[50];
1726 if (read_pipe(fd
, &buf
, &size
)) {
1734 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1736 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1743 int index_fd(unsigned char *sha1
, int fd
, struct stat
*st
, int write_object
, const char *type
)
1745 unsigned long size
= st
->st_size
;
1748 unsigned char hdr
[50];
1753 buf
= mmap(NULL
, size
, PROT_READ
, MAP_PRIVATE
, fd
, 0);
1755 if (buf
== MAP_FAILED
)
1761 ret
= write_sha1_file(buf
, size
, type
, sha1
);
1763 write_sha1_file_prepare(buf
, size
, type
, sha1
, hdr
, &hdrlen
);
1771 int index_path(unsigned char *sha1
, const char *path
, struct stat
*st
, int write_object
)
1776 switch (st
->st_mode
& S_IFMT
) {
1778 fd
= open(path
, O_RDONLY
);
1780 return error("open(\"%s\"): %s", path
,
1782 if (index_fd(sha1
, fd
, st
, write_object
, NULL
) < 0)
1783 return error("%s: failed to insert into database",
1787 target
= xmalloc(st
->st_size
+1);
1788 if (readlink(path
, target
, st
->st_size
+1) != st
->st_size
) {
1789 char *errstr
= strerror(errno
);
1791 return error("readlink(\"%s\"): %s", path
,
1794 if (!write_object
) {
1795 unsigned char hdr
[50];
1797 write_sha1_file_prepare(target
, st
->st_size
, blob_type
,
1798 sha1
, hdr
, &hdrlen
);
1799 } else if (write_sha1_file(target
, st
->st_size
, blob_type
, sha1
))
1800 return error("%s: failed to insert into database",
1805 return error("%s: unsupported file type", path
);