Merge git://git.kernel.org/pub/scm/gitk/gitk
[git/git.git] / sha1_file.c
CommitLineData
0fcfd160
LT
1/*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 *
6 * This handles basic git sha1 object files - packing, unpacking,
7 * creation etc.
8 */
0fcfd160 9#include "cache.h"
1f688557 10#include "delta.h"
a733cb60 11#include "pack.h"
8e440259
PE
12#include "blob.h"
13#include "commit.h"
14#include "tag.h"
15#include "tree.h"
f35a6d3b 16#include "refs.h"
70f5d5d3 17#include "pack-revindex.h"
628522ec 18#include "sha1-lookup.h"
0fcfd160 19
144bde78
LT
20#ifndef O_NOATIME
21#if defined(__linux__) && (defined(__i386__) || defined(__PPC__))
22#define O_NOATIME 01000000
23#else
24#define O_NOATIME 0
25#endif
26#endif
27
e05db0fd
PR
28#ifdef NO_C99_FORMAT
29#define SZ_FMT "lu"
9e42d6a1 30static unsigned long sz_fmt(size_t s) { return (unsigned long)s; }
e05db0fd
PR
31#else
32#define SZ_FMT "zu"
9e42d6a1 33static size_t sz_fmt(size_t s) { return s; }
e05db0fd
PR
34#endif
35
96f1e58f 36const unsigned char null_sha1[20];
88cd621d 37
144bde78
LT
38static unsigned int sha1_file_open_flag = O_NOATIME;
39
192a6be2 40const signed char hexval_table[256] = {
e49521b5
JH
41 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
42 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
43 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
44 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
45 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
46 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
47 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
48 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
49 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
50 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
51 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
52 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
53 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
54 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
55 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
56 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
57 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
58 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
59 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
60 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
61 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
62 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
63 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
64 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
65 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
66 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
67 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
68 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
69 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
70 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
71 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
72 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
73};
0fcfd160
LT
74
75int get_sha1_hex(const char *hex, unsigned char *sha1)
76{
77 int i;
78 for (i = 0; i < 20; i++) {
79 unsigned int val = (hexval(hex[0]) << 4) | hexval(hex[1]);
80 if (val & ~0xff)
81 return -1;
82 *sha1++ = val;
83 hex += 2;
84 }
85 return 0;
86}
87
b2cb9425
JH
88int safe_create_leading_directories(char *path)
89{
90 char *pos = path;
67d42212
JR
91 struct stat st;
92
85dadc38 93 if (is_absolute_path(path))
67ffdf4c 94 pos++;
b2cb9425
JH
95
96 while (pos) {
97 pos = strchr(pos, '/');
98 if (!pos)
99 break;
100 *pos = 0;
67d42212
JR
101 if (!stat(path, &st)) {
102 /* path exists */
103 if (!S_ISDIR(st.st_mode)) {
b2cb9425 104 *pos = '/';
67d42212 105 return -3;
b2cb9425 106 }
457f06d6 107 }
67d42212
JR
108 else if (mkdir(path, 0777)) {
109 *pos = '/';
110 return -1;
111 }
457f06d6
JS
112 else if (adjust_shared_perm(path)) {
113 *pos = '/';
114 return -2;
115 }
b2cb9425
JH
116 *pos++ = '/';
117 }
118 return 0;
119}
723c31fe 120
0fcfd160
LT
121char * sha1_to_hex(const unsigned char *sha1)
122{
dcb3450f
LT
123 static int bufno;
124 static char hexbuffer[4][50];
0fcfd160 125 static const char hex[] = "0123456789abcdef";
dcb3450f 126 char *buffer = hexbuffer[3 & ++bufno], *buf = buffer;
0fcfd160
LT
127 int i;
128
129 for (i = 0; i < 20; i++) {
130 unsigned int val = *sha1++;
131 *buf++ = hex[val >> 4];
132 *buf++ = hex[val & 0xf];
133 }
1e80e044
JS
134 *buf = '\0';
135
0fcfd160
LT
136 return buffer;
137}
138
ace1534d
JH
139static void fill_sha1_path(char *pathbuf, const unsigned char *sha1)
140{
141 int i;
142 for (i = 0; i < 20; i++) {
143 static char hex[] = "0123456789abcdef";
144 unsigned int val = sha1[i];
145 char *pos = pathbuf + i*2 + (i > 0);
146 *pos++ = hex[val >> 4];
147 *pos = hex[val & 0xf];
148 }
149}
150
0fcfd160
LT
151/*
152 * NOTE! This returns a statically allocated buffer, so you have to be
790296fd 153 * careful about using it. Do an "xstrdup()" if you need to save the
0fcfd160 154 * filename.
ace1534d
JH
155 *
156 * Also note that this returns the location for creating. Reading
157 * SHA1 file can happen from any alternate directory listed in the
d19938ab 158 * DB_ENVIRONMENT environment variable if it is not found in
ace1534d 159 * the primary object database.
0fcfd160
LT
160 */
161char *sha1_file_name(const unsigned char *sha1)
162{
0fcfd160
LT
163 static char *name, *base;
164
165 if (!base) {
d19938ab 166 const char *sha1_file_directory = get_object_directory();
0fcfd160 167 int len = strlen(sha1_file_directory);
812666c8 168 base = xmalloc(len + 60);
0fcfd160
LT
169 memcpy(base, sha1_file_directory, len);
170 memset(base+len, 0, 60);
171 base[len] = '/';
172 base[len+3] = '/';
173 name = base + len + 1;
174 }
ace1534d 175 fill_sha1_path(name, sha1);
0fcfd160
LT
176 return base;
177}
178
bf592c50
DB
179char *sha1_pack_name(const unsigned char *sha1)
180{
181 static const char hex[] = "0123456789abcdef";
182 static char *name, *base, *buf;
183 int i;
184
185 if (!base) {
186 const char *sha1_file_directory = get_object_directory();
187 int len = strlen(sha1_file_directory);
188 base = xmalloc(len + 60);
189 sprintf(base, "%s/pack/pack-1234567890123456789012345678901234567890.pack", sha1_file_directory);
190 name = base + len + 11;
191 }
192
193 buf = name;
194
195 for (i = 0; i < 20; i++) {
196 unsigned int val = *sha1++;
197 *buf++ = hex[val >> 4];
198 *buf++ = hex[val & 0xf];
199 }
a6080a0a 200
bf592c50
DB
201 return base;
202}
203
204char *sha1_pack_index_name(const unsigned char *sha1)
205{
206 static const char hex[] = "0123456789abcdef";
207 static char *name, *base, *buf;
208 int i;
209
210 if (!base) {
211 const char *sha1_file_directory = get_object_directory();
212 int len = strlen(sha1_file_directory);
213 base = xmalloc(len + 60);
214 sprintf(base, "%s/pack/pack-1234567890123456789012345678901234567890.idx", sha1_file_directory);
215 name = base + len + 11;
216 }
217
218 buf = name;
219
220 for (i = 0; i < 20; i++) {
221 unsigned int val = *sha1++;
222 *buf++ = hex[val >> 4];
223 *buf++ = hex[val & 0xf];
224 }
a6080a0a 225
bf592c50
DB
226 return base;
227}
228
d5a63b99
JH
229struct alternate_object_database *alt_odb_list;
230static struct alternate_object_database **alt_odb_tail;
ace1534d 231
c2f493a4
MW
232static void read_info_alternates(const char * alternates, int depth);
233
ddd5d056
JH
234/*
235 * Prepare alternate object database registry.
d5a63b99
JH
236 *
237 * The variable alt_odb_list points at the list of struct
238 * alternate_object_database. The elements on this list come from
239 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
240 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
241 * whose contents is similar to that environment variable but can be
242 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
243 * contains "/the/directory/corresponding/to/.git/objects/...", while
244 * its name points just after the slash at the end of ".git/objects/"
245 * in the example above, and has enough space to hold 40-byte hex
246 * SHA1, an extra slash for the first level indirection, and the
247 * terminating NUL.
ddd5d056 248 */
c2f493a4 249static int link_alt_odb_entry(const char * entry, int len, const char * relative_base, int depth)
ace1534d 250{
c2f493a4 251 struct stat st;
1494e038 252 const char *objdir = get_object_directory();
c2f493a4
MW
253 struct alternate_object_database *ent;
254 struct alternate_object_database *alt;
255 /* 43 = 40-byte + 2 '/' + terminating NUL */
256 int pfxlen = len;
257 int entlen = pfxlen + 43;
ccfd3e99 258 int base_len = -1;
d5a63b99 259
85dadc38 260 if (!is_absolute_path(entry) && relative_base) {
c2f493a4
MW
261 /* Relative alt-odb */
262 if (base_len < 0)
263 base_len = strlen(relative_base) + 1;
264 entlen += base_len;
265 pfxlen += base_len;
266 }
267 ent = xmalloc(sizeof(*ent) + entlen);
268
85dadc38 269 if (!is_absolute_path(entry) && relative_base) {
c2f493a4
MW
270 memcpy(ent->base, relative_base, base_len - 1);
271 ent->base[base_len - 1] = '/';
272 memcpy(ent->base + base_len, entry, len);
273 }
274 else
275 memcpy(ent->base, entry, pfxlen);
276
277 ent->name = ent->base + pfxlen + 1;
278 ent->base[pfxlen + 3] = '/';
279 ent->base[pfxlen] = ent->base[entlen-1] = 0;
280
281 /* Detect cases where alternate disappeared */
282 if (stat(ent->base, &st) || !S_ISDIR(st.st_mode)) {
283 error("object directory %s does not exist; "
284 "check .git/objects/info/alternates.",
285 ent->base);
286 free(ent);
287 return -1;
288 }
289
290 /* Prevent the common mistake of listing the same
291 * thing twice, or object directory itself.
292 */
293 for (alt = alt_odb_list; alt; alt = alt->next) {
294 if (!memcmp(ent->base, alt->base, pfxlen)) {
295 free(ent);
296 return -1;
297 }
298 }
299 if (!memcmp(ent->base, objdir, pfxlen)) {
300 free(ent);
301 return -1;
302 }
303
304 /* add the alternate entry */
305 *alt_odb_tail = ent;
306 alt_odb_tail = &(ent->next);
307 ent->next = NULL;
308
309 /* recursively add alternates */
310 read_info_alternates(ent->base, depth + 1);
311
312 ent->base[pfxlen] = '/';
313
314 return 0;
315}
316
317static void link_alt_odb_entries(const char *alt, const char *ep, int sep,
318 const char *relative_base, int depth)
319{
320 const char *cp, *last;
321
322 if (depth > 5) {
323 error("%s: ignoring alternate object stores, nesting too deep.",
324 relative_base);
325 return;
326 }
327
d5a63b99 328 last = alt;
9577e7e3
JH
329 while (last < ep) {
330 cp = last;
331 if (cp < ep && *cp == '#') {
332 while (cp < ep && *cp != sep)
333 cp++;
334 last = cp + 1;
335 continue;
336 }
c2f493a4
MW
337 while (cp < ep && *cp != sep)
338 cp++;
d5a63b99 339 if (last != cp) {
85dadc38 340 if (!is_absolute_path(last) && depth) {
c2f493a4
MW
341 error("%s: ignoring relative alternate object store %s",
342 relative_base, last);
343 } else {
344 link_alt_odb_entry(last, cp - last,
345 relative_base, depth);
1494e038 346 }
d5a63b99 347 }
9577e7e3 348 while (cp < ep && *cp == sep)
d5a63b99
JH
349 cp++;
350 last = cp;
9577e7e3 351 }
d5a63b99
JH
352}
353
c2f493a4 354static void read_info_alternates(const char * relative_base, int depth)
d5a63b99 355{
9577e7e3 356 char *map;
dc49cd76 357 size_t mapsz;
d5a63b99 358 struct stat st;
9cb18f56
JM
359 const char alt_file_name[] = "info/alternates";
360 /* Given that relative_base is no longer than PATH_MAX,
361 ensure that "path" has enough space to append "/", the
362 file name, "info/alternates", and a trailing NUL. */
363 char path[PATH_MAX + 1 + sizeof alt_file_name];
c2f493a4 364 int fd;
d5a63b99 365
9cb18f56 366 sprintf(path, "%s/%s", relative_base, alt_file_name);
d5a63b99
JH
367 fd = open(path, O_RDONLY);
368 if (fd < 0)
369 return;
370 if (fstat(fd, &st) || (st.st_size == 0)) {
371 close(fd);
9a217f2a 372 return;
ace1534d 373 }
dc49cd76
SP
374 mapsz = xsize_t(st.st_size);
375 map = xmmap(NULL, mapsz, PROT_READ, MAP_PRIVATE, fd, 0);
d5a63b99 376 close(fd);
d5a63b99 377
dc49cd76 378 link_alt_odb_entries(map, map + mapsz, '\n', relative_base, depth);
c2f493a4 379
dc49cd76 380 munmap(map, mapsz);
ace1534d
JH
381}
382
c2f493a4
MW
383void prepare_alt_odb(void)
384{
554fe20d 385 const char *alt;
c2f493a4 386
7dc24aa5
SP
387 if (alt_odb_tail)
388 return;
389
c2f493a4
MW
390 alt = getenv(ALTERNATE_DB_ENVIRONMENT);
391 if (!alt) alt = "";
392
c2f493a4
MW
393 alt_odb_tail = &alt_odb_list;
394 link_alt_odb_entries(alt, alt + strlen(alt), ':', NULL, 0);
395
396 read_info_alternates(get_object_directory(), 0);
397}
398
ace1534d
JH
399static char *find_sha1_file(const unsigned char *sha1, struct stat *st)
400{
ace1534d 401 char *name = sha1_file_name(sha1);
d5a63b99 402 struct alternate_object_database *alt;
ace1534d
JH
403
404 if (!stat(name, st))
405 return name;
9a217f2a 406 prepare_alt_odb();
d5a63b99
JH
407 for (alt = alt_odb_list; alt; alt = alt->next) {
408 name = alt->name;
ace1534d 409 fill_sha1_path(name, sha1);
d5a63b99
JH
410 if (!stat(alt->base, st))
411 return alt->base;
ace1534d
JH
412 }
413 return NULL;
414}
415
60bb8b14 416static unsigned int pack_used_ctr;
a53128b6
SP
417static unsigned int pack_mmap_calls;
418static unsigned int peak_pack_open_windows;
419static unsigned int pack_open_windows;
420static size_t peak_pack_mapped;
60bb8b14 421static size_t pack_mapped;
9a217f2a 422struct packed_git *packed_git;
1f688557 423
b79d18c9 424void pack_report(void)
a53128b6
SP
425{
426 fprintf(stderr,
e05db0fd
PR
427 "pack_report: getpagesize() = %10" SZ_FMT "\n"
428 "pack_report: core.packedGitWindowSize = %10" SZ_FMT "\n"
429 "pack_report: core.packedGitLimit = %10" SZ_FMT "\n",
9e42d6a1
SP
430 sz_fmt(getpagesize()),
431 sz_fmt(packed_git_window_size),
432 sz_fmt(packed_git_limit));
a53128b6
SP
433 fprintf(stderr,
434 "pack_report: pack_used_ctr = %10u\n"
435 "pack_report: pack_mmap_calls = %10u\n"
436 "pack_report: pack_open_windows = %10u / %10u\n"
e05db0fd
PR
437 "pack_report: pack_mapped = "
438 "%10" SZ_FMT " / %10" SZ_FMT "\n",
a53128b6
SP
439 pack_used_ctr,
440 pack_mmap_calls,
441 pack_open_windows, peak_pack_open_windows,
9e42d6a1 442 sz_fmt(pack_mapped), sz_fmt(peak_pack_mapped));
a53128b6
SP
443}
444
42873078 445static int check_packed_git_idx(const char *path, struct packed_git *p)
1f688557
JH
446{
447 void *idx_map;
42873078 448 struct pack_idx_header *hdr;
dc49cd76 449 size_t idx_size;
74e34e1f 450 uint32_t version, nr, i, *index;
e93ec6f9 451 int fd = open(path, O_RDONLY);
1f688557 452 struct stat st;
42873078 453
1f688557
JH
454 if (fd < 0)
455 return -1;
456 if (fstat(fd, &st)) {
457 close(fd);
458 return -1;
459 }
dc49cd76 460 idx_size = xsize_t(st.st_size);
2d88451b
SP
461 if (idx_size < 4 * 256 + 20 + 20) {
462 close(fd);
463 return error("index file %s is too small", path);
464 }
c4712e45 465 idx_map = xmmap(NULL, idx_size, PROT_READ, MAP_PRIVATE, fd, 0);
1f688557 466 close(fd);
1f688557 467
42873078
NP
468 hdr = idx_map;
469 if (hdr->idx_signature == htonl(PACK_IDX_SIGNATURE)) {
74e34e1f
NP
470 version = ntohl(hdr->idx_version);
471 if (version < 2 || version > 2) {
472 munmap(idx_map, idx_size);
473 return error("index file %s is version %d"
474 " and is not supported by this binary"
475 " (try upgrading GIT to a newer version)",
476 path, version);
477 }
478 } else
479 version = 1;
df1b059d 480
1f688557 481 nr = 0;
42873078 482 index = idx_map;
74e34e1f
NP
483 if (version > 1)
484 index += 2; /* skip index header */
1f688557 485 for (i = 0; i < 256; i++) {
326bf396 486 uint32_t n = ntohl(index[i]);
2d88451b
SP
487 if (n < nr) {
488 munmap(idx_map, idx_size);
df1b059d 489 return error("non-monotonic index %s", path);
2d88451b 490 }
1f688557
JH
491 nr = n;
492 }
493
74e34e1f
NP
494 if (version == 1) {
495 /*
496 * Total size:
497 * - 256 index entries 4 bytes each
498 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
499 * - 20-byte SHA1 of the packfile
500 * - 20-byte SHA1 file checksum
501 */
502 if (idx_size != 4*256 + nr * 24 + 20 + 20) {
503 munmap(idx_map, idx_size);
eef427a0 504 return error("wrong index v1 file size in %s", path);
74e34e1f
NP
505 }
506 } else if (version == 2) {
507 /*
508 * Minimum size:
509 * - 8 bytes of header
510 * - 256 index entries 4 bytes each
511 * - 20-byte sha1 entry * nr
512 * - 4-byte crc entry * nr
513 * - 4-byte offset entry * nr
514 * - 20-byte SHA1 of the packfile
515 * - 20-byte SHA1 file checksum
516 * And after the 4-byte offset table might be a
517 * variable sized table containing 8-byte entries
518 * for offsets larger than 2^31.
519 */
520 unsigned long min_size = 8 + 4*256 + nr*(20 + 4 + 4) + 20 + 20;
1164f1e4
LT
521 unsigned long max_size = min_size;
522 if (nr)
523 max_size += (nr - 1)*8;
524 if (idx_size < min_size || idx_size > max_size) {
74e34e1f 525 munmap(idx_map, idx_size);
eef427a0 526 return error("wrong index v2 file size in %s", path);
74e34e1f 527 }
7109c889
JH
528 if (idx_size != min_size &&
529 /*
530 * make sure we can deal with large pack offsets.
531 * 31-bit signed offset won't be enough, neither
532 * 32-bit unsigned one will be.
533 */
534 (sizeof(off_t) <= 4)) {
535 munmap(idx_map, idx_size);
536 return error("pack too large for current definition of off_t in %s", path);
74e34e1f 537 }
2d88451b 538 }
1f688557 539
74e34e1f 540 p->index_version = version;
42873078
NP
541 p->index_data = idx_map;
542 p->index_size = idx_size;
57059091 543 p->num_objects = nr;
1f688557
JH
544 return 0;
545}
546
bc8e478a 547int open_pack_index(struct packed_git *p)
d079837e
SP
548{
549 char *idx_name;
550 int ret;
551
552 if (p->index_data)
553 return 0;
554
555 idx_name = xstrdup(p->pack_name);
556 strcpy(idx_name + strlen(idx_name) - strlen(".pack"), ".idx");
557 ret = check_packed_git_idx(idx_name, p);
558 free(idx_name);
559 return ret;
560}
561
11daf39b
SP
562static void scan_windows(struct packed_git *p,
563 struct packed_git **lru_p,
564 struct pack_window **lru_w,
565 struct pack_window **lru_l)
1f688557 566{
11daf39b
SP
567 struct pack_window *w, *w_l;
568
569 for (w_l = NULL, w = p->windows; w; w = w->next) {
570 if (!w->inuse_cnt) {
571 if (!*lru_w || w->last_used < (*lru_w)->last_used) {
572 *lru_p = p;
573 *lru_w = w;
574 *lru_l = w_l;
54044bf8 575 }
54044bf8 576 }
11daf39b 577 w_l = w;
f9253394 578 }
11daf39b
SP
579}
580
d1efefa4 581static int unuse_one_window(struct packed_git *current, int keep_fd)
11daf39b
SP
582{
583 struct packed_git *p, *lru_p = NULL;
584 struct pack_window *lru_w = NULL, *lru_l = NULL;
585
586 if (current)
587 scan_windows(current, &lru_p, &lru_w, &lru_l);
588 for (p = packed_git; p; p = p->next)
589 scan_windows(p, &lru_p, &lru_w, &lru_l);
54044bf8
SP
590 if (lru_p) {
591 munmap(lru_w->base, lru_w->len);
592 pack_mapped -= lru_w->len;
593 if (lru_l)
594 lru_l->next = lru_w->next;
595 else {
596 lru_p->windows = lru_w->next;
d1efefa4 597 if (!lru_p->windows && lru_p->pack_fd != keep_fd) {
54044bf8
SP
598 close(lru_p->pack_fd);
599 lru_p->pack_fd = -1;
600 }
601 }
602 free(lru_w);
a53128b6 603 pack_open_windows--;
54044bf8
SP
604 return 1;
605 }
606 return 0;
f9253394
JH
607}
608
d1efefa4 609void release_pack_memory(size_t need, int fd)
97bfeb34
SP
610{
611 size_t cur = pack_mapped;
d1efefa4 612 while (need >= (cur - pack_mapped) && unuse_one_window(NULL, fd))
97bfeb34
SP
613 ; /* nothing */
614}
615
c9ced051
SP
616void close_pack_windows(struct packed_git *p)
617{
618 while (p->windows) {
619 struct pack_window *w = p->windows;
620
621 if (w->inuse_cnt)
622 die("pack '%s' still has open windows to it",
623 p->pack_name);
624 munmap(w->base, w->len);
625 pack_mapped -= w->len;
626 pack_open_windows--;
627 p->windows = w->next;
628 free(w);
629 }
630}
631
03e79c88 632void unuse_pack(struct pack_window **w_cursor)
f9253394 633{
03e79c88
SP
634 struct pack_window *w = *w_cursor;
635 if (w) {
636 w->inuse_cnt--;
637 *w_cursor = NULL;
638 }
1f688557
JH
639}
640
3cf8b462
SP
641/*
642 * Do not call this directly as this leaks p->pack_fd on error return;
643 * call open_packed_git() instead.
644 */
645static int open_packed_git_1(struct packed_git *p)
1f688557 646{
9bc879c1
SP
647 struct stat st;
648 struct pack_header hdr;
649 unsigned char sha1[20];
650 unsigned char *idx_sha1;
2c039da8 651 long fd_flag;
9bc879c1 652
d079837e
SP
653 if (!p->index_data && open_pack_index(p))
654 return error("packfile %s index unavailable", p->pack_name);
655
9bc879c1
SP
656 p->pack_fd = open(p->pack_name, O_RDONLY);
657 if (p->pack_fd < 0 || fstat(p->pack_fd, &st))
072db278 658 return -1;
9bc879c1
SP
659
660 /* If we created the struct before we had the pack we lack size. */
bf592c50 661 if (!p->pack_size) {
bf592c50 662 if (!S_ISREG(st.st_mode))
072db278 663 return error("packfile %s not a regular file", p->pack_name);
bf592c50 664 p->pack_size = st.st_size;
9bc879c1 665 } else if (p->pack_size != st.st_size)
072db278 666 return error("packfile %s size changed", p->pack_name);
9bc879c1 667
2c039da8
JH
668 /* We leave these file descriptors open with sliding mmap;
669 * there is no point keeping them open across exec(), though.
670 */
671 fd_flag = fcntl(p->pack_fd, F_GETFD, 0);
672 if (fd_flag < 0)
072db278 673 return error("cannot determine file descriptor flags");
2c039da8
JH
674 fd_flag |= FD_CLOEXEC;
675 if (fcntl(p->pack_fd, F_SETFD, fd_flag) == -1)
072db278 676 return error("cannot set FD_CLOEXEC");
2c039da8 677
9bc879c1 678 /* Verify we recognize this pack file format. */
e6e2bd62 679 if (read_in_full(p->pack_fd, &hdr, sizeof(hdr)) != sizeof(hdr))
072db278 680 return error("file %s is far too short to be a packfile", p->pack_name);
9bc879c1 681 if (hdr.hdr_signature != htonl(PACK_SIGNATURE))
072db278 682 return error("file %s is not a GIT packfile", p->pack_name);
9bc879c1 683 if (!pack_version_ok(hdr.hdr_version))
072db278 684 return error("packfile %s is version %u and not supported"
9bc879c1
SP
685 " (try upgrading GIT to a newer version)",
686 p->pack_name, ntohl(hdr.hdr_version));
687
688 /* Verify the pack matches its index. */
57059091 689 if (p->num_objects != ntohl(hdr.hdr_entries))
072db278 690 return error("packfile %s claims to have %u objects"
57059091
NP
691 " while index indicates %u objects",
692 p->pack_name, ntohl(hdr.hdr_entries),
693 p->num_objects);
9bc879c1 694 if (lseek(p->pack_fd, p->pack_size - sizeof(sha1), SEEK_SET) == -1)
072db278 695 return error("end of packfile %s is unavailable", p->pack_name);
e6e2bd62 696 if (read_in_full(p->pack_fd, sha1, sizeof(sha1)) != sizeof(sha1))
072db278 697 return error("packfile %s signature is unavailable", p->pack_name);
42873078 698 idx_sha1 = ((unsigned char *)p->index_data) + p->index_size - 40;
9bc879c1 699 if (hashcmp(sha1, idx_sha1))
072db278
SP
700 return error("packfile %s does not match index", p->pack_name);
701 return 0;
9bc879c1
SP
702}
703
3cf8b462
SP
704static int open_packed_git(struct packed_git *p)
705{
706 if (!open_packed_git_1(p))
707 return 0;
708 if (p->pack_fd != -1) {
709 close(p->pack_fd);
710 p->pack_fd = -1;
711 }
712 return -1;
713}
714
c4001d92 715static int in_window(struct pack_window *win, off_t offset)
60bb8b14
SP
716{
717 /* We must promise at least 20 bytes (one hash) after the
718 * offset is available from this window, otherwise the offset
719 * is not actually in this window and a different window (which
720 * has that one hash excess) must be used. This is to support
721 * the object header and delta base parsing routines below.
722 */
723 off_t win_off = win->offset;
724 return win_off <= offset
725 && (offset + 20) <= (win_off + win->len);
726}
727
03e79c88
SP
728unsigned char* use_pack(struct packed_git *p,
729 struct pack_window **w_cursor,
c4001d92 730 off_t offset,
03e79c88 731 unsigned int *left)
9bc879c1 732{
60bb8b14 733 struct pack_window *win = *w_cursor;
03e79c88 734
072db278
SP
735 if (p->pack_fd == -1 && open_packed_git(p))
736 die("packfile %s cannot be accessed", p->pack_name);
60bb8b14
SP
737
738 /* Since packfiles end in a hash of their content and its
739 * pointless to ask for an offset into the middle of that
740 * hash, and the in_window function above wouldn't match
741 * don't allow an offset too close to the end of the file.
742 */
743 if (offset > (p->pack_size - 20))
744 die("offset beyond end of packfile (truncated pack?)");
745
746 if (!win || !in_window(win, offset)) {
747 if (win)
748 win->inuse_cnt--;
749 for (win = p->windows; win; win = win->next) {
750 if (in_window(win, offset))
751 break;
752 }
753 if (!win) {
78a28df9 754 size_t window_align = packed_git_window_size / 2;
dc49cd76 755 off_t len;
60bb8b14 756 win = xcalloc(1, sizeof(*win));
78a28df9 757 win->offset = (offset / window_align) * window_align;
dc49cd76
SP
758 len = p->pack_size - win->offset;
759 if (len > packed_git_window_size)
760 len = packed_git_window_size;
761 win->len = (size_t)len;
60bb8b14 762 pack_mapped += win->len;
11daf39b 763 while (packed_git_limit < pack_mapped
d1efefa4 764 && unuse_one_window(p, p->pack_fd))
60bb8b14 765 ; /* nothing */
c4712e45 766 win->base = xmmap(NULL, win->len,
60bb8b14
SP
767 PROT_READ, MAP_PRIVATE,
768 p->pack_fd, win->offset);
769 if (win->base == MAP_FAILED)
73b4e4be
SP
770 die("packfile %s cannot be mapped: %s",
771 p->pack_name,
772 strerror(errno));
a53128b6
SP
773 pack_mmap_calls++;
774 pack_open_windows++;
775 if (pack_mapped > peak_pack_mapped)
776 peak_pack_mapped = pack_mapped;
777 if (pack_open_windows > peak_pack_open_windows)
778 peak_pack_open_windows = pack_open_windows;
60bb8b14
SP
779 win->next = p->windows;
780 p->windows = win;
781 }
1f688557 782 }
03e79c88
SP
783 if (win != *w_cursor) {
784 win->last_used = pack_used_ctr++;
785 win->inuse_cnt++;
786 *w_cursor = win;
787 }
60bb8b14 788 offset -= win->offset;
03e79c88 789 if (left)
dc49cd76 790 *left = win->len - xsize_t(offset);
03e79c88 791 return win->base + offset;
1f688557
JH
792}
793
42873078 794struct packed_git *add_packed_git(const char *path, int path_len, int local)
1f688557
JH
795{
796 struct stat st;
42873078 797 struct packed_git *p = xmalloc(sizeof(*p) + path_len + 2);
1f688557 798
42873078
NP
799 /*
800 * Make sure a corresponding .pack file exists and that
801 * the index looks sane.
802 */
803 path_len -= strlen(".idx");
804 if (path_len < 1)
1f688557 805 return NULL;
42873078
NP
806 memcpy(p->pack_name, path, path_len);
807 strcpy(p->pack_name + path_len, ".pack");
d079837e 808 if (stat(p->pack_name, &st) || !S_ISREG(st.st_mode)) {
42873078 809 free(p);
1f688557
JH
810 return NULL;
811 }
42873078 812
1f688557
JH
813 /* ok, it looks sane as far as we can check without
814 * actually mapping the pack file.
815 */
d079837e
SP
816 p->index_version = 0;
817 p->index_data = NULL;
818 p->index_size = 0;
819 p->num_objects = 0;
1f688557 820 p->pack_size = st.st_size;
1f688557 821 p->next = NULL;
c41ee586 822 p->windows = NULL;
9bc879c1 823 p->pack_fd = -1;
9d835df2 824 p->pack_local = local;
b867092f 825 p->mtime = st.st_mtime;
42873078
NP
826 if (path_len < 40 || get_sha1_hex(path + path_len - 40, p->sha1))
827 hashclr(p->sha1);
1f688557
JH
828 return p;
829}
830
bf592c50 831struct packed_git *parse_pack_index(unsigned char *sha1)
c508df5e
DB
832{
833 char *path = sha1_pack_index_name(sha1);
834 return parse_pack_index_file(sha1, path);
835}
836
42873078
NP
837struct packed_git *parse_pack_index_file(const unsigned char *sha1,
838 const char *idx_path)
bf592c50 839{
42873078
NP
840 const char *path = sha1_pack_name(sha1);
841 struct packed_git *p = xmalloc(sizeof(*p) + strlen(path) + 2);
bf592c50 842
42873078
NP
843 if (check_packed_git_idx(idx_path, p)) {
844 free(p);
bf592c50 845 return NULL;
42873078 846 }
bf592c50 847
bf592c50 848 strcpy(p->pack_name, path);
bf592c50 849 p->pack_size = 0;
bf592c50 850 p->next = NULL;
c41ee586 851 p->windows = NULL;
9bc879c1 852 p->pack_fd = -1;
e702496e 853 hashcpy(p->sha1, sha1);
bf592c50
DB
854 return p;
855}
856
857void install_packed_git(struct packed_git *pack)
858{
859 pack->next = packed_git;
860 packed_git = pack;
861}
862
9d835df2 863static void prepare_packed_git_one(char *objdir, int local)
1f688557 864{
9cb18f56
JM
865 /* Ensure that this buffer is large enough so that we can
866 append "/pack/" without clobbering the stack even if
867 strlen(objdir) were PATH_MAX. */
868 char path[PATH_MAX + 1 + 4 + 1 + 1];
1f688557
JH
869 int len;
870 DIR *dir;
871 struct dirent *de;
872
873 sprintf(path, "%s/pack", objdir);
874 len = strlen(path);
875 dir = opendir(path);
b5b16990 876 if (!dir) {
26125f6b 877 if (errno != ENOENT)
bd2afde8 878 error("unable to open object pack directory: %s: %s",
26125f6b 879 path, strerror(errno));
1f688557 880 return;
b5b16990 881 }
1f688557
JH
882 path[len++] = '/';
883 while ((de = readdir(dir)) != NULL) {
884 int namelen = strlen(de->d_name);
885 struct packed_git *p;
886
5bb1cda5 887 if (!has_extension(de->d_name, ".idx"))
1f688557
JH
888 continue;
889
9cb18f56
JM
890 if (len + namelen + 1 > sizeof(path))
891 continue;
892
54a15a8d 893 /* Don't reopen a pack we already have. */
1f688557 894 strcpy(path + len, de->d_name);
86f7780c
JK
895 for (p = packed_git; p; p = p->next) {
896 if (!memcmp(path, p->pack_name, len + namelen - 4))
897 break;
898 }
899 if (p)
900 continue;
54a15a8d
SP
901 /* See if it really is a valid .idx file with corresponding
902 * .pack file that we can map.
903 */
9d835df2 904 p = add_packed_git(path, len + namelen, local);
1f688557
JH
905 if (!p)
906 continue;
625e9421 907 install_packed_git(p);
1f688557 908 }
5b35bcd5 909 closedir(dir);
1f688557
JH
910}
911
b867092f
JH
912static int sort_pack(const void *a_, const void *b_)
913{
914 struct packed_git *a = *((struct packed_git **)a_);
915 struct packed_git *b = *((struct packed_git **)b_);
916 int st;
917
918 /*
919 * Local packs tend to contain objects specific to our
920 * variant of the project than remote ones. In addition,
921 * remote ones could be on a network mounted filesystem.
922 * Favor local ones for these reasons.
923 */
924 st = a->pack_local - b->pack_local;
925 if (st)
926 return -st;
927
928 /*
929 * Younger packs tend to contain more recent objects,
930 * and more recent objects tend to get accessed more
931 * often.
932 */
933 if (a->mtime < b->mtime)
934 return 1;
935 else if (a->mtime == b->mtime)
936 return 0;
937 return -1;
938}
939
940static void rearrange_packed_git(void)
941{
942 struct packed_git **ary, *p;
943 int i, n;
944
945 for (n = 0, p = packed_git; p; p = p->next)
946 n++;
947 if (n < 2)
948 return;
949
950 /* prepare an array of packed_git for easier sorting */
951 ary = xcalloc(n, sizeof(struct packed_git *));
952 for (n = 0, p = packed_git; p; p = p->next)
953 ary[n++] = p;
954
955 qsort(ary, n, sizeof(struct packed_git *), sort_pack);
956
957 /* link them back again */
958 for (i = 0; i < n - 1; i++)
959 ary[i]->next = ary[i + 1];
960 ary[n - 1]->next = NULL;
961 packed_git = ary[0];
962
963 free(ary);
964}
965
637cdd9d 966static int prepare_packed_git_run_once = 0;
9a217f2a 967void prepare_packed_git(void)
1f688557 968{
d5a63b99 969 struct alternate_object_database *alt;
1f688557 970
637cdd9d 971 if (prepare_packed_git_run_once)
1f688557 972 return;
9d835df2 973 prepare_packed_git_one(get_object_directory(), 1);
9a217f2a 974 prepare_alt_odb();
d5a63b99 975 for (alt = alt_odb_list; alt; alt = alt->next) {
1494e038 976 alt->name[-1] = 0;
9d835df2 977 prepare_packed_git_one(alt->base, 0);
1494e038 978 alt->name[-1] = '/';
1f688557 979 }
b867092f 980 rearrange_packed_git();
637cdd9d
JK
981 prepare_packed_git_run_once = 1;
982}
983
fc04c412 984void reprepare_packed_git(void)
637cdd9d
JK
985{
986 prepare_packed_git_run_once = 0;
987 prepare_packed_git();
1f688557
JH
988}
989
5d6ccf5c 990int check_sha1_signature(const unsigned char *sha1, void *map, unsigned long size, const char *type)
0fcfd160
LT
991{
992 unsigned char real_sha1[20];
7cfb5f36 993 hash_sha1_file(map, size, type, real_sha1);
a89fccd2 994 return hashcmp(sha1, real_sha1) ? -1 : 0;
0fcfd160
LT
995}
996
4175e9e3 997static void *map_sha1_file(const unsigned char *sha1, unsigned long *size)
0fcfd160 998{
0fcfd160
LT
999 struct stat st;
1000 void *map;
144bde78 1001 int fd;
ace1534d
JH
1002 char *filename = find_sha1_file(sha1, &st);
1003
1004 if (!filename) {
ace1534d
JH
1005 return NULL;
1006 }
0fcfd160 1007
144bde78 1008 fd = open(filename, O_RDONLY | sha1_file_open_flag);
0fcfd160 1009 if (fd < 0) {
144bde78
LT
1010 /* See if it works without O_NOATIME */
1011 switch (sha1_file_open_flag) {
1012 default:
1013 fd = open(filename, O_RDONLY);
1014 if (fd >= 0)
1015 break;
1016 /* Fallthrough */
1017 case 0:
144bde78
LT
1018 return NULL;
1019 }
1020
1f688557
JH
1021 /* If it failed once, it will probably fail again.
1022 * Stop using O_NOATIME
1023 */
144bde78 1024 sha1_file_open_flag = 0;
0fcfd160 1025 }
dc49cd76
SP
1026 *size = xsize_t(st.st_size);
1027 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
0fcfd160 1028 close(fd);
0fcfd160
LT
1029 return map;
1030}
1031
726f852b 1032static int legacy_loose_object(unsigned char *map)
bb6b8e4f
JH
1033{
1034 unsigned int word;
1035
1036 /*
1037 * Is it a zlib-compressed buffer? If so, the first byte
1038 * must be 0x78 (15-bit window size, deflated), and the
1039 * first 16-bit word is evenly divisible by 31
1040 */
1041 word = (map[0] << 8) + map[1];
1042 if (map[0] == 0x78 && !(word % 31))
1043 return 1;
1044 else
1045 return 0;
1046}
1047
72518e9c 1048unsigned long unpack_object_header_gently(const unsigned char *buf, unsigned long len, enum object_type *type, unsigned long *sizep)
c4483576 1049{
ad1ed5ee 1050 unsigned shift;
93821bd9 1051 unsigned char c;
93821bd9 1052 unsigned long size;
ad1ed5ee
JH
1053 unsigned long used = 0;
1054
1055 c = buf[used++];
1056 *type = (c >> 4) & 7;
1057 size = c & 15;
1058 shift = 4;
1059 while (c & 0x80) {
1060 if (len <= used)
1061 return 0;
1062 if (sizeof(long) * 8 <= shift)
1063 return 0;
1064 c = buf[used++];
1065 size += (c & 0x7f) << shift;
1066 shift += 7;
1067 }
1068 *sizep = size;
1069 return used;
1070}
1071
1072static int unpack_sha1_header(z_stream *stream, unsigned char *map, unsigned long mapsize, void *buffer, unsigned long bufsiz)
1073{
1074 unsigned long size, used;
1075 static const char valid_loose_object_type[8] = {
1076 0, /* OBJ_EXT */
1077 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
1078 0, /* "delta" and others are invalid in a loose object */
93821bd9 1079 };
ad1ed5ee 1080 enum object_type type;
93821bd9 1081
c4483576
LT
1082 /* Get the data stream */
1083 memset(stream, 0, sizeof(*stream));
1084 stream->next_in = map;
1085 stream->avail_in = mapsize;
1086 stream->next_out = buffer;
93821bd9
LT
1087 stream->avail_out = bufsiz;
1088
bb6b8e4f 1089 if (legacy_loose_object(map)) {
93821bd9
LT
1090 inflateInit(stream);
1091 return inflate(stream, 0);
1092 }
1093
726f852b
NP
1094
1095 /*
1096 * There used to be a second loose object header format which
1097 * was meant to mimic the in-pack format, allowing for direct
1098 * copy of the object data. This format turned up not to be
1099 * really worth it and we don't write it any longer. But we
1100 * can still read it.
1101 */
ad1ed5ee
JH
1102 used = unpack_object_header_gently(map, mapsize, &type, &size);
1103 if (!used || !valid_loose_object_type[type])
93821bd9 1104 return -1;
ad1ed5ee
JH
1105 map += used;
1106 mapsize -= used;
93821bd9
LT
1107
1108 /* Set up the stream for the rest.. */
1109 stream->next_in = map;
1110 stream->avail_in = mapsize;
c4483576 1111 inflateInit(stream);
93821bd9
LT
1112
1113 /* And generate the fake traditional header */
ad1ed5ee 1114 stream->total_out = 1 + snprintf(buffer, bufsiz, "%s %lu",
df843662 1115 typename(type), size);
93821bd9 1116 return 0;
c4483576
LT
1117}
1118
7efbff75 1119static void *unpack_sha1_rest(z_stream *stream, void *buffer, unsigned long size, const unsigned char *sha1)
5180cacc
LT
1120{
1121 int bytes = strlen(buffer) + 1;
d565b341 1122 unsigned char *buf = xmalloc(1+size);
93821bd9 1123 unsigned long n;
7efbff75 1124 int status = Z_OK;
5180cacc 1125
93821bd9
LT
1126 n = stream->total_out - bytes;
1127 if (n > size)
1128 n = size;
1129 memcpy(buf, (char *) buffer + bytes, n);
1130 bytes = n;
456cdf6e
LT
1131 if (bytes <= size) {
1132 /*
1133 * The above condition must be (bytes <= size), not
1134 * (bytes < size). In other words, even though we
1135 * expect no more output and set avail_out to zer0,
1136 * the input zlib stream may have bytes that express
1137 * "this concludes the stream", and we *do* want to
1138 * eat that input.
1139 *
1140 * Otherwise we would not be able to test that we
1141 * consumed all the input to reach the expected size;
1142 * we also want to check that zlib tells us that all
1143 * went well with status == Z_STREAM_END at the end.
1144 */
5180cacc
LT
1145 stream->next_out = buf + bytes;
1146 stream->avail_out = size - bytes;
7efbff75
JH
1147 while (status == Z_OK)
1148 status = inflate(stream, Z_FINISH);
5180cacc
LT
1149 }
1150 buf[size] = 0;
456cdf6e 1151 if (status == Z_STREAM_END && !stream->avail_in) {
7efbff75
JH
1152 inflateEnd(stream);
1153 return buf;
1154 }
1155
1156 if (status < 0)
1157 error("corrupt loose object '%s'", sha1_to_hex(sha1));
1158 else if (stream->avail_in)
1159 error("garbage at end of loose object '%s'",
1160 sha1_to_hex(sha1));
1161 free(buf);
1162 return NULL;
5180cacc
LT
1163}
1164
1165/*
1166 * We used to just use "sscanf()", but that's actually way
1167 * too permissive for what we want to check. So do an anal
1168 * object header parse by hand.
1169 */
21666f1a 1170static int parse_sha1_header(const char *hdr, unsigned long *sizep)
5180cacc 1171{
21666f1a 1172 char type[10];
5180cacc
LT
1173 int i;
1174 unsigned long size;
1175
1176 /*
a6080a0a 1177 * The type can be at most ten bytes (including the
5180cacc 1178 * terminating '\0' that we add), and is followed by
21666f1a 1179 * a space.
5180cacc 1180 */
21666f1a 1181 i = 0;
5180cacc
LT
1182 for (;;) {
1183 char c = *hdr++;
1184 if (c == ' ')
1185 break;
21666f1a
NP
1186 type[i++] = c;
1187 if (i >= sizeof(type))
5180cacc 1188 return -1;
5180cacc 1189 }
21666f1a 1190 type[i] = 0;
5180cacc
LT
1191
1192 /*
1193 * The length must follow immediately, and be in canonical
1194 * decimal format (ie "010" is not valid).
1195 */
1196 size = *hdr++ - '0';
1197 if (size > 9)
1198 return -1;
1199 if (size) {
1200 for (;;) {
1201 unsigned long c = *hdr - '0';
1202 if (c > 9)
1203 break;
1204 hdr++;
1205 size = size * 10 + c;
1206 }
1207 }
1208 *sizep = size;
1209
1210 /*
1211 * The length must be followed by a zero byte
1212 */
21666f1a 1213 return *hdr ? -1 : type_from_string(type);
5180cacc
LT
1214}
1215
7efbff75 1216static void *unpack_sha1_file(void *map, unsigned long mapsize, enum object_type *type, unsigned long *size, const unsigned char *sha1)
0fcfd160 1217{
5180cacc 1218 int ret;
0fcfd160 1219 z_stream stream;
5180cacc 1220 char hdr[8192];
0fcfd160 1221
5180cacc 1222 ret = unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr));
21666f1a 1223 if (ret < Z_OK || (*type = parse_sha1_header(hdr, size)) < 0)
0fcfd160
LT
1224 return NULL;
1225
7efbff75 1226 return unpack_sha1_rest(&stream, hdr, *size, sha1);
0fcfd160
LT
1227}
1228
54dab52a
NP
1229unsigned long get_size_from_delta(struct packed_git *p,
1230 struct pack_window **w_curs,
1231 off_t curpos)
1232{
1233 const unsigned char *data;
1234 unsigned char delta_head[20], *in;
1235 z_stream stream;
1236 int st;
1237
1238 memset(&stream, 0, sizeof(stream));
1239 stream.next_out = delta_head;
1240 stream.avail_out = sizeof(delta_head);
1241
1242 inflateInit(&stream);
1243 do {
1244 in = use_pack(p, w_curs, curpos, &stream.avail_in);
1245 stream.next_in = in;
1246 st = inflate(&stream, Z_FINISH);
1247 curpos += stream.next_in - in;
1248 } while ((st == Z_OK || st == Z_BUF_ERROR) &&
1249 stream.total_out < sizeof(delta_head));
1250 inflateEnd(&stream);
1251 if ((st != Z_STREAM_END) && stream.total_out != sizeof(delta_head))
1252 die("delta data unpack-initial failed");
1253
1254 /* Examine the initial part of the delta to figure out
1255 * the result size.
1256 */
1257 data = delta_head;
1258
1259 /* ignore base size */
1260 get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
1261
1262 /* Read the result size */
1263 return get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
1264}
1265
c4001d92 1266static off_t get_delta_base(struct packed_git *p,
03e79c88 1267 struct pack_window **w_curs,
c4001d92 1268 off_t *curpos,
21666f1a 1269 enum object_type type,
c4001d92 1270 off_t delta_obj_offset)
eb32d236 1271{
2b87c45b 1272 unsigned char *base_info = use_pack(p, w_curs, *curpos, NULL);
c4001d92 1273 off_t base_offset;
eb32d236 1274
8d8a4ea5
SP
1275 /* use_pack() assured us we have [base_info, base_info + 20)
1276 * as a range that we can look at without walking off the
1277 * end of the mapped window. Its actually the hash size
1278 * that is assured. An OFS_DELTA longer than the hash size
1279 * is stupid, as then a REF_DELTA would be smaller to store.
1280 */
21666f1a 1281 if (type == OBJ_OFS_DELTA) {
eb32d236
NP
1282 unsigned used = 0;
1283 unsigned char c = base_info[used++];
1284 base_offset = c & 127;
1285 while (c & 128) {
1286 base_offset += 1;
8723f216 1287 if (!base_offset || MSB(base_offset, 7))
eb32d236
NP
1288 die("offset value overflow for delta base object");
1289 c = base_info[used++];
1290 base_offset = (base_offset << 7) + (c & 127);
1291 }
1292 base_offset = delta_obj_offset - base_offset;
1293 if (base_offset >= delta_obj_offset)
1294 die("delta base offset out of bound");
2b87c45b 1295 *curpos += used;
21666f1a 1296 } else if (type == OBJ_REF_DELTA) {
eb32d236
NP
1297 /* The base entry _must_ be in the same pack */
1298 base_offset = find_pack_entry_one(base_info, p);
1299 if (!base_offset)
1300 die("failed to find delta-pack base object %s",
1301 sha1_to_hex(base_info));
2b87c45b 1302 *curpos += 20;
eb32d236
NP
1303 } else
1304 die("I am totally screwed");
2b87c45b 1305 return base_offset;
eb32d236
NP
1306}
1307
f3bf9224 1308/* forward declaration for a mutually recursive function */
c4001d92 1309static int packed_object_info(struct packed_git *p, off_t offset,
21666f1a 1310 unsigned long *sizep);
f3bf9224 1311
43057304 1312static int packed_delta_info(struct packed_git *p,
03e79c88 1313 struct pack_window **w_curs,
c4001d92 1314 off_t curpos,
21666f1a 1315 enum object_type type,
c4001d92 1316 off_t obj_offset,
43057304 1317 unsigned long *sizep)
5db47c2b 1318{
c4001d92 1319 off_t base_offset;
f3bf9224 1320
21666f1a
NP
1321 base_offset = get_delta_base(p, w_curs, &curpos, type, obj_offset);
1322 type = packed_object_info(p, base_offset, NULL);
f3bf9224 1323
c62266f3
JH
1324 /* We choose to only get the type of the base object and
1325 * ignore potentially corrupt pack file that expects the delta
1326 * based on a base with a wrong size. This saves tons of
1327 * inflate() calls.
1328 */
54dab52a
NP
1329 if (sizep)
1330 *sizep = get_size_from_delta(p, w_curs, curpos);
21666f1a
NP
1331
1332 return type;
5db47c2b
JH
1333}
1334
2b87c45b
NP
1335static int unpack_object_header(struct packed_git *p,
1336 struct pack_window **w_curs,
c4001d92 1337 off_t *curpos,
2b87c45b 1338 unsigned long *sizep)
a733cb60 1339{
03e79c88
SP
1340 unsigned char *base;
1341 unsigned int left;
ad1ed5ee 1342 unsigned long used;
2b87c45b 1343 enum object_type type;
a733cb60 1344
8d8a4ea5
SP
1345 /* use_pack() assures us we have [base, base + 20) available
1346 * as a range that we can look at at. (Its actually the hash
3dff5379 1347 * size that is assured.) With our object header encoding
8d8a4ea5
SP
1348 * the maximum deflated object size is 2^137, which is just
1349 * insane, so we know won't exceed what we have been given.
1350 */
2b87c45b
NP
1351 base = use_pack(p, w_curs, *curpos, &left);
1352 used = unpack_object_header_gently(base, left, &type, sizep);
ad1ed5ee
JH
1353 if (!used)
1354 die("object offset outside of pack file");
2b87c45b 1355 *curpos += used;
ad1ed5ee 1356
2b87c45b 1357 return type;
a733cb60
LT
1358}
1359
21666f1a 1360const char *packed_object_info_detail(struct packed_git *p,
c4001d92 1361 off_t obj_offset,
21666f1a
NP
1362 unsigned long *size,
1363 unsigned long *store_size,
1364 unsigned int *delta_chain_length,
1365 unsigned char *base_sha1)
ad8c80a5 1366{
03e79c88 1367 struct pack_window *w_curs = NULL;
c4001d92
SP
1368 off_t curpos;
1369 unsigned long dummy;
43057304 1370 unsigned char *next_sha1;
21666f1a 1371 enum object_type type;
70f5d5d3 1372 struct revindex_entry *revidx;
ad8c80a5 1373
43057304 1374 *delta_chain_length = 0;
2b87c45b 1375 curpos = obj_offset;
21666f1a 1376 type = unpack_object_header(p, &w_curs, &curpos, size);
43057304 1377
70f5d5d3
NP
1378 revidx = find_pack_revindex(p, obj_offset);
1379 *store_size = revidx[1].offset - obj_offset;
1380
43057304 1381 for (;;) {
21666f1a 1382 switch (type) {
43057304 1383 default:
08a19d87 1384 die("pack %s contains unknown object type %d",
21666f1a 1385 p->pack_name, type);
43057304
NP
1386 case OBJ_COMMIT:
1387 case OBJ_TREE:
1388 case OBJ_BLOB:
1389 case OBJ_TAG:
03e79c88 1390 unuse_pack(&w_curs);
21666f1a 1391 return typename(type);
eb32d236 1392 case OBJ_OFS_DELTA:
21666f1a 1393 obj_offset = get_delta_base(p, &w_curs, &curpos, type, obj_offset);
eb32d236 1394 if (*delta_chain_length == 0) {
70f5d5d3
NP
1395 revidx = find_pack_revindex(p, obj_offset);
1396 hashcpy(base_sha1, nth_packed_object_sha1(p, revidx->nr));
eb32d236
NP
1397 }
1398 break;
1399 case OBJ_REF_DELTA:
2b87c45b 1400 next_sha1 = use_pack(p, &w_curs, curpos, NULL);
43057304
NP
1401 if (*delta_chain_length == 0)
1402 hashcpy(base_sha1, next_sha1);
2b87c45b 1403 obj_offset = find_pack_entry_one(next_sha1, p);
43057304
NP
1404 break;
1405 }
43057304 1406 (*delta_chain_length)++;
2b87c45b 1407 curpos = obj_offset;
21666f1a 1408 type = unpack_object_header(p, &w_curs, &curpos, &dummy);
ad8c80a5 1409 }
ad8c80a5
JH
1410}
1411
c4001d92 1412static int packed_object_info(struct packed_git *p, off_t obj_offset,
21666f1a 1413 unsigned long *sizep)
1f688557 1414{
03e79c88 1415 struct pack_window *w_curs = NULL;
c4001d92
SP
1416 unsigned long size;
1417 off_t curpos = obj_offset;
21666f1a 1418 enum object_type type;
5db47c2b 1419
21666f1a 1420 type = unpack_object_header(p, &w_curs, &curpos, &size);
5db47c2b 1421
21666f1a 1422 switch (type) {
eb32d236
NP
1423 case OBJ_OFS_DELTA:
1424 case OBJ_REF_DELTA:
21666f1a
NP
1425 type = packed_delta_info(p, &w_curs, curpos,
1426 type, obj_offset, sizep);
2b87c45b 1427 break;
a733cb60 1428 case OBJ_COMMIT:
a733cb60 1429 case OBJ_TREE:
a733cb60 1430 case OBJ_BLOB:
a733cb60 1431 case OBJ_TAG:
2b87c45b
NP
1432 if (sizep)
1433 *sizep = size;
a69d0943 1434 break;
1f688557 1435 default:
08a19d87 1436 die("pack %s contains unknown object type %d",
21666f1a 1437 p->pack_name, type);
1f688557 1438 }
2b87c45b 1439 unuse_pack(&w_curs);
21666f1a 1440 return type;
1f688557
JH
1441}
1442
eb950c19 1443static void *unpack_compressed_entry(struct packed_git *p,
03e79c88 1444 struct pack_window **w_curs,
c4001d92 1445 off_t curpos,
eb950c19 1446 unsigned long size)
de530aaa
SP
1447{
1448 int st;
1449 z_stream stream;
079afb18 1450 unsigned char *buffer, *in;
de530aaa
SP
1451
1452 buffer = xmalloc(size + 1);
1453 buffer[size] = 0;
1454 memset(&stream, 0, sizeof(stream));
de530aaa
SP
1455 stream.next_out = buffer;
1456 stream.avail_out = size;
1457
1458 inflateInit(&stream);
079afb18 1459 do {
2b87c45b 1460 in = use_pack(p, w_curs, curpos, &stream.avail_in);
079afb18
SP
1461 stream.next_in = in;
1462 st = inflate(&stream, Z_FINISH);
2b87c45b 1463 curpos += stream.next_in - in;
079afb18 1464 } while (st == Z_OK || st == Z_BUF_ERROR);
de530aaa
SP
1465 inflateEnd(&stream);
1466 if ((st != Z_STREAM_END) || stream.total_out != size) {
1467 free(buffer);
1468 return NULL;
1469 }
1470
1471 return buffer;
1472}
1473
e5e01619
LT
1474#define MAX_DELTA_CACHE (256)
1475
18bdec11 1476static size_t delta_base_cached;
5e08ecbf
NP
1477
1478static struct delta_base_cache_lru_list {
1479 struct delta_base_cache_lru_list *prev;
1480 struct delta_base_cache_lru_list *next;
1481} delta_base_cache_lru = { &delta_base_cache_lru, &delta_base_cache_lru };
1482
e5e01619 1483static struct delta_base_cache_entry {
5e08ecbf
NP
1484 struct delta_base_cache_lru_list lru;
1485 void *data;
e5e01619
LT
1486 struct packed_git *p;
1487 off_t base_offset;
1488 unsigned long size;
e5e01619
LT
1489 enum object_type type;
1490} delta_base_cache[MAX_DELTA_CACHE];
1491
1492static unsigned long pack_entry_hash(struct packed_git *p, off_t base_offset)
1493{
1494 unsigned long hash;
1495
1496 hash = (unsigned long)p + (unsigned long)base_offset;
1497 hash += (hash >> 8) + (hash >> 16);
3358004a 1498 return hash % MAX_DELTA_CACHE;
e5e01619
LT
1499}
1500
62f255ad 1501static void *cache_or_unpack_entry(struct packed_git *p, off_t base_offset,
a0cba108 1502 unsigned long *base_size, enum object_type *type, int keep_cache)
62f255ad 1503{
e5e01619
LT
1504 void *ret;
1505 unsigned long hash = pack_entry_hash(p, base_offset);
1506 struct delta_base_cache_entry *ent = delta_base_cache + hash;
1507
1508 ret = ent->data;
1509 if (ret && ent->p == p && ent->base_offset == base_offset)
1510 goto found_cache_entry;
62f255ad 1511 return unpack_entry(p, base_offset, type, base_size);
e5e01619
LT
1512
1513found_cache_entry:
18bdec11 1514 if (!keep_cache) {
a0cba108 1515 ent->data = NULL;
5e08ecbf
NP
1516 ent->lru.next->prev = ent->lru.prev;
1517 ent->lru.prev->next = ent->lru.next;
18bdec11 1518 delta_base_cached -= ent->size;
182af834
PH
1519 } else {
1520 ret = xmemdupz(ent->data, ent->size);
a0cba108 1521 }
e5e01619
LT
1522 *type = ent->type;
1523 *base_size = ent->size;
1524 return ret;
62f255ad
LT
1525}
1526
18bdec11
SP
1527static inline void release_delta_base_cache(struct delta_base_cache_entry *ent)
1528{
1529 if (ent->data) {
1530 free(ent->data);
1531 ent->data = NULL;
5e08ecbf
NP
1532 ent->lru.next->prev = ent->lru.prev;
1533 ent->lru.prev->next = ent->lru.next;
18bdec11
SP
1534 delta_base_cached -= ent->size;
1535 }
1536}
1537
62f255ad
LT
1538static void add_delta_base_cache(struct packed_git *p, off_t base_offset,
1539 void *base, unsigned long base_size, enum object_type type)
1540{
5e08ecbf 1541 unsigned long hash = pack_entry_hash(p, base_offset);
e5e01619 1542 struct delta_base_cache_entry *ent = delta_base_cache + hash;
5e08ecbf 1543 struct delta_base_cache_lru_list *lru;
e5e01619 1544
18bdec11
SP
1545 release_delta_base_cache(ent);
1546 delta_base_cached += base_size;
5e08ecbf
NP
1547
1548 for (lru = delta_base_cache_lru.next;
1549 delta_base_cached > delta_base_cache_limit
1550 && lru != &delta_base_cache_lru;
1551 lru = lru->next) {
1552 struct delta_base_cache_entry *f = (void *)lru;
18bdec11
SP
1553 if (f->type == OBJ_BLOB)
1554 release_delta_base_cache(f);
1555 }
5e08ecbf
NP
1556 for (lru = delta_base_cache_lru.next;
1557 delta_base_cached > delta_base_cache_limit
1558 && lru != &delta_base_cache_lru;
1559 lru = lru->next) {
1560 struct delta_base_cache_entry *f = (void *)lru;
1561 release_delta_base_cache(f);
1562 }
18bdec11 1563
e5e01619
LT
1564 ent->p = p;
1565 ent->base_offset = base_offset;
1566 ent->type = type;
1567 ent->data = base;
1568 ent->size = base_size;
5e08ecbf
NP
1569 ent->lru.next = &delta_base_cache_lru;
1570 ent->lru.prev = delta_base_cache_lru.prev;
1571 delta_base_cache_lru.prev->next = &ent->lru;
1572 delta_base_cache_lru.prev = &ent->lru;
62f255ad
LT
1573}
1574
eb950c19 1575static void *unpack_delta_entry(struct packed_git *p,
03e79c88 1576 struct pack_window **w_curs,
c4001d92 1577 off_t curpos,
1f688557 1578 unsigned long delta_size,
c4001d92 1579 off_t obj_offset,
21666f1a 1580 enum object_type *type,
eb950c19 1581 unsigned long *sizep)
1f688557 1582{
7c3e8be3 1583 void *delta_data, *result, *base;
c4001d92
SP
1584 unsigned long base_size;
1585 off_t base_offset;
43057304 1586
21666f1a 1587 base_offset = get_delta_base(p, w_curs, &curpos, *type, obj_offset);
a0cba108 1588 base = cache_or_unpack_entry(p, base_offset, &base_size, type, 0);
67686d95 1589 if (!base)
c4001d92
SP
1590 die("failed to read delta base object"
1591 " at %"PRIuMAX" from %s",
1592 (uintmax_t)base_offset, p->pack_name);
67686d95 1593
2b87c45b 1594 delta_data = unpack_compressed_entry(p, w_curs, curpos, delta_size);
9064d87b
SP
1595 if (!delta_data)
1596 die("failed to unpack compressed delta"
1597 " at %"PRIuMAX" from %s",
1598 (uintmax_t)curpos, p->pack_name);
1f688557
JH
1599 result = patch_delta(base, base_size,
1600 delta_data, delta_size,
21666f1a 1601 sizep);
1f688557
JH
1602 if (!result)
1603 die("failed to apply delta");
1604 free(delta_data);
62f255ad 1605 add_delta_base_cache(p, base_offset, base, base_size, *type);
1f688557
JH
1606 return result;
1607}
1608
c4001d92 1609void *unpack_entry(struct packed_git *p, off_t obj_offset,
21666f1a 1610 enum object_type *type, unsigned long *sizep)
1f688557 1611{
03e79c88 1612 struct pack_window *w_curs = NULL;
c4001d92 1613 off_t curpos = obj_offset;
21666f1a 1614 void *data;
1f688557 1615
21666f1a
NP
1616 *type = unpack_object_header(p, &w_curs, &curpos, sizep);
1617 switch (*type) {
eb32d236
NP
1618 case OBJ_OFS_DELTA:
1619 case OBJ_REF_DELTA:
21666f1a
NP
1620 data = unpack_delta_entry(p, &w_curs, curpos, *sizep,
1621 obj_offset, type, sizep);
4d703a1a 1622 break;
a733cb60 1623 case OBJ_COMMIT:
a733cb60 1624 case OBJ_TREE:
a733cb60 1625 case OBJ_BLOB:
a733cb60 1626 case OBJ_TAG:
21666f1a 1627 data = unpack_compressed_entry(p, &w_curs, curpos, *sizep);
4d703a1a 1628 break;
1f688557 1629 default:
21666f1a 1630 die("unknown object type %i in %s", *type, p->pack_name);
1f688557 1631 }
03e79c88 1632 unuse_pack(&w_curs);
21666f1a 1633 return data;
1f688557
JH
1634}
1635
d079837e 1636const unsigned char *nth_packed_object_sha1(struct packed_git *p,
d72308e0 1637 uint32_t n)
9a217f2a 1638{
42873078 1639 const unsigned char *index = p->index_data;
d079837e
SP
1640 if (!index) {
1641 if (open_pack_index(p))
1642 return NULL;
1643 index = p->index_data;
1644 }
57059091 1645 if (n >= p->num_objects)
d72308e0 1646 return NULL;
74e34e1f
NP
1647 index += 4 * 256;
1648 if (p->index_version == 1) {
1649 return index + 24 * n + 4;
1650 } else {
1651 index += 8;
1652 return index + 20 * n;
1653 }
1654}
1655
1656static off_t nth_packed_object_offset(const struct packed_git *p, uint32_t n)
1657{
1658 const unsigned char *index = p->index_data;
1659 index += 4 * 256;
1660 if (p->index_version == 1) {
1661 return ntohl(*((uint32_t *)(index + 24 * n)));
1662 } else {
1663 uint32_t off;
1664 index += 8 + p->num_objects * (20 + 4);
1665 off = ntohl(*((uint32_t *)(index + 4 * n)));
1666 if (!(off & 0x80000000))
1667 return off;
1668 index += p->num_objects * 4 + (off & 0x7fffffff) * 8;
1669 return (((uint64_t)ntohl(*((uint32_t *)(index + 0)))) << 32) |
1670 ntohl(*((uint32_t *)(index + 4)));
1671 }
9a217f2a
JH
1672}
1673
c4001d92 1674off_t find_pack_entry_one(const unsigned char *sha1,
43057304 1675 struct packed_git *p)
1f688557 1676{
42873078 1677 const uint32_t *level1_ofs = p->index_data;
42873078 1678 const unsigned char *index = p->index_data;
628522ec
JH
1679 unsigned hi, lo, stride;
1680 static int use_lookup = -1;
1681 static int debug_lookup = -1;
1682
1683 if (debug_lookup < 0)
1684 debug_lookup = !!getenv("GIT_DEBUG_LOOKUP");
42873078 1685
d079837e
SP
1686 if (!index) {
1687 if (open_pack_index(p))
1688 return 0;
1689 level1_ofs = p->index_data;
1690 index = p->index_data;
1691 }
74e34e1f
NP
1692 if (p->index_version > 1) {
1693 level1_ofs += 2;
1694 index += 8;
1695 }
42873078 1696 index += 4 * 256;
74e34e1f
NP
1697 hi = ntohl(level1_ofs[*sha1]);
1698 lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
628522ec
JH
1699 if (p->index_version > 1) {
1700 stride = 20;
1701 } else {
1702 stride = 24;
1703 index += 4;
1704 }
1705
1706 if (debug_lookup)
1707 printf("%02x%02x%02x... lo %u hi %u nr %u\n",
1708 sha1[0], sha1[1], sha1[2], lo, hi, p->num_objects);
1709
1710 if (use_lookup < 0)
1711 use_lookup = !!getenv("GIT_USE_LOOKUP");
1712 if (use_lookup) {
1713 int pos = sha1_entry_pos(index, stride, 0,
1714 lo, hi, p->num_objects, sha1);
1715 if (pos < 0)
1716 return 0;
1717 return nth_packed_object_offset(p, pos);
1718 }
1f688557
JH
1719
1720 do {
74e34e1f 1721 unsigned mi = (lo + hi) / 2;
628522ec
JH
1722 int cmp = hashcmp(index + mi * stride, sha1);
1723
1724 if (debug_lookup)
1725 printf("lo %u hi %u rg %u mi %u\n",
1726 lo, hi, hi - lo, mi);
43057304 1727 if (!cmp)
74e34e1f 1728 return nth_packed_object_offset(p, mi);
1f688557
JH
1729 if (cmp > 0)
1730 hi = mi;
1731 else
1732 lo = mi+1;
1733 } while (lo < hi);
1734 return 0;
1735}
1736
000dfd3f 1737int matches_pack_name(struct packed_git *p, const char *name)
d4ff6d92
SP
1738{
1739 const char *last_c, *c;
1740
000dfd3f
JH
1741 if (!strcmp(p->pack_name, name))
1742 return 1;
d4ff6d92
SP
1743
1744 for (c = p->pack_name, last_c = c; *c;)
1745 if (*c == '/')
1746 last_c = ++c;
1747 else
1748 ++c;
000dfd3f
JH
1749 if (!strcmp(last_c, name))
1750 return 1;
d4ff6d92 1751
000dfd3f 1752 return 0;
d4ff6d92
SP
1753}
1754
106d710b 1755static int find_pack_entry(const unsigned char *sha1, struct pack_entry *e, const char **ignore_packed)
1f688557 1756{
f7c22cc6 1757 static struct packed_git *last_found = (void *)1;
1f688557 1758 struct packed_git *p;
c4001d92 1759 off_t offset;
43057304 1760
1f688557 1761 prepare_packed_git();
f7c22cc6
NP
1762 if (!packed_git)
1763 return 0;
1764 p = (last_found == (void *)1) ? packed_git : last_found;
1f688557 1765
f7c22cc6 1766 do {
106d710b
JH
1767 if (ignore_packed) {
1768 const char **ig;
1769 for (ig = ignore_packed; *ig; ig++)
000dfd3f 1770 if (matches_pack_name(p, *ig))
106d710b
JH
1771 break;
1772 if (*ig)
f7c22cc6 1773 goto next;
106d710b 1774 }
f7c22cc6 1775
43057304
NP
1776 offset = find_pack_entry_one(sha1, p);
1777 if (offset) {
c715f783
SP
1778 /*
1779 * We are about to tell the caller where they can
1780 * locate the requested object. We better make
1781 * sure the packfile is still here and can be
1782 * accessed before supplying that answer, as
1783 * it may have been deleted since the index
1784 * was loaded!
1785 */
1786 if (p->pack_fd == -1 && open_packed_git(p)) {
1787 error("packfile %s cannot be accessed", p->pack_name);
f7c22cc6 1788 goto next;
c715f783 1789 }
43057304
NP
1790 e->offset = offset;
1791 e->p = p;
1792 hashcpy(e->sha1, sha1);
f7c22cc6 1793 last_found = p;
1f688557 1794 return 1;
43057304 1795 }
f7c22cc6
NP
1796
1797 next:
1798 if (p == last_found)
1799 p = packed_git;
1800 else
1801 p = p->next;
1802 if (p == last_found)
1803 p = p->next;
1804 } while (p);
1f688557
JH
1805 return 0;
1806}
1807
a6080a0a 1808struct packed_git *find_sha1_pack(const unsigned char *sha1,
bf592c50
DB
1809 struct packed_git *packs)
1810{
1811 struct packed_git *p;
bf592c50
DB
1812
1813 for (p = packs; p; p = p->next) {
43057304 1814 if (find_pack_entry_one(sha1, p))
bf592c50
DB
1815 return p;
1816 }
1817 return NULL;
21666f1a 1818
bf592c50
DB
1819}
1820
21666f1a 1821static int sha1_loose_object_info(const unsigned char *sha1, unsigned long *sizep)
65c2e0c3 1822{
36e4d74a 1823 int status;
65c2e0c3
JH
1824 unsigned long mapsize, size;
1825 void *map;
1826 z_stream stream;
d65a16f6 1827 char hdr[32];
65c2e0c3 1828
bb6b8e4f 1829 map = map_sha1_file(sha1, &mapsize);
f0df4ed5
JS
1830 if (!map)
1831 return error("unable to find %s", sha1_to_hex(sha1));
36e4d74a
JH
1832 if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
1833 status = error("unable to unpack %s header",
1834 sha1_to_hex(sha1));
21666f1a 1835 else if ((status = parse_sha1_header(hdr, &size)) < 0)
36e4d74a 1836 status = error("unable to parse %s header", sha1_to_hex(sha1));
21666f1a
NP
1837 else if (sizep)
1838 *sizep = size;
65c2e0c3
JH
1839 inflateEnd(&stream);
1840 munmap(map, mapsize);
1841 return status;
1842}
1843
21666f1a 1844int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
f0df4ed5 1845{
f0df4ed5
JS
1846 struct pack_entry e;
1847
1848 if (!find_pack_entry(sha1, &e, NULL)) {
1849 reprepare_packed_git();
1850 if (!find_pack_entry(sha1, &e, NULL))
21666f1a 1851 return sha1_loose_object_info(sha1, sizep);
f0df4ed5 1852 }
21666f1a 1853 return packed_object_info(e.p, e.offset, sizep);
f0df4ed5
JS
1854}
1855
21666f1a
NP
1856static void *read_packed_sha1(const unsigned char *sha1,
1857 enum object_type *type, unsigned long *size)
1f688557
JH
1858{
1859 struct pack_entry e;
1860
8276c007 1861 if (!find_pack_entry(sha1, &e, NULL))
1f688557 1862 return NULL;
8276c007 1863 else
a0cba108 1864 return cache_or_unpack_entry(e.p, e.offset, size, type, 1);
1f688557
JH
1865}
1866
d66b37bb
JH
1867/*
1868 * This is meant to hold a *small* number of objects that you would
1869 * want read_sha1_file() to be able to return, but yet you do not want
1870 * to write them into the object store (e.g. a browse-only
1871 * application).
1872 */
1873static struct cached_object {
1874 unsigned char sha1[20];
21666f1a 1875 enum object_type type;
d66b37bb
JH
1876 void *buf;
1877 unsigned long size;
1878} *cached_objects;
1879static int cached_object_nr, cached_object_alloc;
1880
346245a1
JK
1881static struct cached_object empty_tree = {
1882 /* empty tree sha1: 4b825dc642cb6eb9a060e54bf8d69288fbee4904 */
1883 "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60"
1884 "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04",
1885 OBJ_TREE,
1886 "",
1887 0
1888};
1889
d66b37bb
JH
1890static struct cached_object *find_cached_object(const unsigned char *sha1)
1891{
1892 int i;
1893 struct cached_object *co = cached_objects;
1894
1895 for (i = 0; i < cached_object_nr; i++, co++) {
1896 if (!hashcmp(co->sha1, sha1))
1897 return co;
1898 }
346245a1
JK
1899 if (!hashcmp(sha1, empty_tree.sha1))
1900 return &empty_tree;
d66b37bb
JH
1901 return NULL;
1902}
1903
21666f1a
NP
1904int pretend_sha1_file(void *buf, unsigned long len, enum object_type type,
1905 unsigned char *sha1)
d66b37bb
JH
1906{
1907 struct cached_object *co;
1908
21666f1a 1909 hash_sha1_file(buf, len, typename(type), sha1);
d66b37bb
JH
1910 if (has_sha1_file(sha1) || find_cached_object(sha1))
1911 return 0;
1912 if (cached_object_alloc <= cached_object_nr) {
1913 cached_object_alloc = alloc_nr(cached_object_alloc);
1914 cached_objects = xrealloc(cached_objects,
1915 sizeof(*cached_objects) *
1916 cached_object_alloc);
1917 }
1918 co = &cached_objects[cached_object_nr++];
1919 co->size = len;
21666f1a 1920 co->type = type;
efa13f7b
JH
1921 co->buf = xmalloc(len);
1922 memcpy(co->buf, buf, len);
d66b37bb
JH
1923 hashcpy(co->sha1, sha1);
1924 return 0;
1925}
1926
21666f1a
NP
1927void *read_sha1_file(const unsigned char *sha1, enum object_type *type,
1928 unsigned long *size)
0fcfd160
LT
1929{
1930 unsigned long mapsize;
1931 void *map, *buf;
d66b37bb
JH
1932 struct cached_object *co;
1933
1934 co = find_cached_object(sha1);
1935 if (co) {
21666f1a 1936 *type = co->type;
d66b37bb 1937 *size = co->size;
182af834 1938 return xmemdupz(co->buf, co->size);
d66b37bb 1939 }
0fcfd160 1940
8276c007
PE
1941 buf = read_packed_sha1(sha1, type, size);
1942 if (buf)
1943 return buf;
bb6b8e4f 1944 map = map_sha1_file(sha1, &mapsize);
0fcfd160 1945 if (map) {
7efbff75 1946 buf = unpack_sha1_file(map, mapsize, type, size, sha1);
0fcfd160
LT
1947 munmap(map, mapsize);
1948 return buf;
1949 }
637cdd9d 1950 reprepare_packed_git();
8276c007 1951 return read_packed_sha1(sha1, type, size);
0fcfd160
LT
1952}
1953
40469ee9 1954void *read_object_with_reference(const unsigned char *sha1,
21666f1a 1955 const char *required_type_name,
40469ee9
JH
1956 unsigned long *size,
1957 unsigned char *actual_sha1_return)
f4913f91 1958{
21666f1a 1959 enum object_type type, required_type;
f4913f91
JH
1960 void *buffer;
1961 unsigned long isize;
40469ee9 1962 unsigned char actual_sha1[20];
f4913f91 1963
21666f1a 1964 required_type = type_from_string(required_type_name);
e702496e 1965 hashcpy(actual_sha1, sha1);
40469ee9
JH
1966 while (1) {
1967 int ref_length = -1;
1968 const char *ref_type = NULL;
f4913f91 1969
21666f1a 1970 buffer = read_sha1_file(actual_sha1, &type, &isize);
40469ee9
JH
1971 if (!buffer)
1972 return NULL;
21666f1a 1973 if (type == required_type) {
40469ee9
JH
1974 *size = isize;
1975 if (actual_sha1_return)
e702496e 1976 hashcpy(actual_sha1_return, actual_sha1);
40469ee9
JH
1977 return buffer;
1978 }
1979 /* Handle references */
21666f1a 1980 else if (type == OBJ_COMMIT)
40469ee9 1981 ref_type = "tree ";
21666f1a 1982 else if (type == OBJ_TAG)
40469ee9
JH
1983 ref_type = "object ";
1984 else {
1985 free(buffer);
1986 return NULL;
1987 }
1988 ref_length = strlen(ref_type);
f4913f91 1989
50974ec9
MK
1990 if (ref_length + 40 > isize ||
1991 memcmp(buffer, ref_type, ref_length) ||
1d7f171c 1992 get_sha1_hex((char *) buffer + ref_length, actual_sha1)) {
40469ee9
JH
1993 free(buffer);
1994 return NULL;
1995 }
1cf58e72 1996 free(buffer);
40469ee9
JH
1997 /* Now we have the ID of the referred-to object in
1998 * actual_sha1. Check again. */
f4913f91 1999 }
f4913f91
JH
2000}
2001
ce9fbf16 2002static void write_sha1_file_prepare(const void *buf, unsigned long len,
972a9155 2003 const char *type, unsigned char *sha1,
d65a16f6 2004 char *hdr, int *hdrlen)
d410c0f5
JH
2005{
2006 SHA_CTX c;
2007
2008 /* Generate the header */
d65a16f6 2009 *hdrlen = sprintf(hdr, "%s %lu", type, len)+1;
d410c0f5
JH
2010
2011 /* Sha1.. */
2012 SHA1_Init(&c);
2013 SHA1_Update(&c, hdr, *hdrlen);
2014 SHA1_Update(&c, buf, len);
2015 SHA1_Final(sha1, &c);
d410c0f5
JH
2016}
2017
230f1322
LT
2018/*
2019 * Link the tempfile to the final place, possibly creating the
2020 * last directory level as you do so.
2021 *
2022 * Returns the errno on failure, 0 on success.
2023 */
839837b9 2024static int link_temp_to_file(const char *tmpfile, const char *filename)
230f1322
LT
2025{
2026 int ret;
756aaf4a 2027 char *dir;
230f1322
LT
2028
2029 if (!link(tmpfile, filename))
2030 return 0;
2031
2032 /*
756aaf4a 2033 * Try to mkdir the last path component if that failed.
230f1322
LT
2034 *
2035 * Re-try the "link()" regardless of whether the mkdir
2036 * succeeds, since a race might mean that somebody
2037 * else succeeded.
2038 */
2039 ret = errno;
756aaf4a
SP
2040 dir = strrchr(filename, '/');
2041 if (dir) {
2042 *dir = 0;
866cae0d 2043 if (!mkdir(filename, 0777) && adjust_shared_perm(filename)) {
91c23e48 2044 *dir = '/';
756aaf4a 2045 return -2;
91c23e48 2046 }
756aaf4a
SP
2047 *dir = '/';
2048 if (!link(tmpfile, filename))
2049 return 0;
2050 ret = errno;
230f1322
LT
2051 }
2052 return ret;
2053}
2054
2055/*
2056 * Move the just written object into its final resting place
2057 */
839837b9 2058int move_temp_to_file(const char *tmpfile, const char *filename)
230f1322
LT
2059{
2060 int ret = link_temp_to_file(tmpfile, filename);
7ebb6fca
LT
2061
2062 /*
2063 * Coda hack - coda doesn't like cross-directory links,
2064 * so we fall back to a rename, which will mean that it
2065 * won't be able to check collisions, but that's not a
2066 * big deal.
2067 *
2068 * The same holds for FAT formatted media.
2069 *
2070 * When this succeeds, we just return 0. We have nothing
2071 * left to unlink.
2072 */
2073 if (ret && ret != EEXIST) {
2074 if (!rename(tmpfile, filename))
230f1322 2075 return 0;
9e48b389 2076 ret = errno;
230f1322
LT
2077 }
2078 unlink(tmpfile);
2079 if (ret) {
2080 if (ret != EEXIST) {
916d081b 2081 return error("unable to write sha1 filename %s: %s\n", filename, strerror(ret));
230f1322
LT
2082 }
2083 /* FIXME!!! Collision check here ? */
2084 }
2085
2086 return 0;
2087}
2088
4d548150
LT
2089static int write_buffer(int fd, const void *buf, size_t len)
2090{
d34cf19b 2091 if (write_in_full(fd, buf, len) < 0)
93822c22 2092 return error("file write error (%s)", strerror(errno));
4d548150
LT
2093 return 0;
2094}
2095
ce9fbf16 2096int hash_sha1_file(const void *buf, unsigned long len, const char *type,
abdc3fc8
RS
2097 unsigned char *sha1)
2098{
d65a16f6 2099 char hdr[32];
abdc3fc8
RS
2100 int hdrlen;
2101 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
2102 return 0;
2103}
2104
bbac7311
NP
2105static int write_loose_object(const unsigned char *sha1, char *hdr, int hdrlen,
2106 void *buf, unsigned long len, time_t mtime)
0fcfd160 2107{
bbac7311 2108 int fd, size, ret;
bf0f910d 2109 unsigned char *compressed;
0fcfd160 2110 z_stream stream;
706bc531 2111 char *filename;
aac17941 2112 static char tmpfile[PATH_MAX];
a44c9a5e 2113
972a9155 2114 filename = sha1_file_name(sha1);
aac17941
LT
2115 fd = open(filename, O_RDONLY);
2116 if (fd >= 0) {
706bc531 2117 /*
aac17941
LT
2118 * FIXME!!! We might do collision checking here, but we'd
2119 * need to uncompress the old file and check it. Later.
706bc531 2120 */
aac17941 2121 close(fd);
706bc531
LT
2122 return 0;
2123 }
2124
aac17941 2125 if (errno != ENOENT) {
916d081b 2126 return error("sha1 file %s: %s\n", filename, strerror(errno));
aac17941
LT
2127 }
2128
0e55181f 2129 snprintf(tmpfile, sizeof(tmpfile), "%s/tmp_obj_XXXXXX", get_object_directory());
ace1534d 2130
aac17941
LT
2131 fd = mkstemp(tmpfile);
2132 if (fd < 0) {
916d081b
PB
2133 if (errno == EPERM)
2134 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
2135 else
2136 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
aac17941
LT
2137 }
2138
0fcfd160
LT
2139 /* Set it up */
2140 memset(&stream, 0, sizeof(stream));
12f6c308 2141 deflateInit(&stream, zlib_compression_level);
93821bd9 2142 size = 8 + deflateBound(&stream, len+hdrlen);
812666c8 2143 compressed = xmalloc(size);
0fcfd160
LT
2144
2145 /* Compress it */
0fcfd160
LT
2146 stream.next_out = compressed;
2147 stream.avail_out = size;
a44c9a5e
LT
2148
2149 /* First header.. */
d65a16f6 2150 stream.next_in = (unsigned char *)hdr;
a44c9a5e 2151 stream.avail_in = hdrlen;
726f852b
NP
2152 while (deflate(&stream, 0) == Z_OK)
2153 /* nothing */;
a44c9a5e
LT
2154
2155 /* Then the data itself.. */
2156 stream.next_in = buf;
2157 stream.avail_in = len;
ac54c277
LT
2158 ret = deflate(&stream, Z_FINISH);
2159 if (ret != Z_STREAM_END)
2160 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1), ret);
2161
2162 ret = deflateEnd(&stream);
2163 if (ret != Z_OK)
2164 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1), ret);
2165
0fcfd160
LT
2166 size = stream.total_out;
2167
4d548150
LT
2168 if (write_buffer(fd, compressed, size) < 0)
2169 die("unable to write sha1 file");
aac17941 2170 fchmod(fd, 0444);
e82973cf
JH
2171 if (close(fd))
2172 die("unable to write sha1 file");
383f85b7 2173 free(compressed);
0fcfd160 2174
bbac7311
NP
2175 if (mtime) {
2176 struct utimbuf utb;
2177 utb.actime = mtime;
2178 utb.modtime = mtime;
2179 if (utime(tmpfile, &utb) < 0)
2180 warning("failed utime() on %s: %s",
2181 tmpfile, strerror(errno));
2182 }
2183
230f1322 2184 return move_temp_to_file(tmpfile, filename);
0fcfd160 2185}
8237b185 2186
bbac7311
NP
2187int write_sha1_file(void *buf, unsigned long len, const char *type, unsigned char *returnsha1)
2188{
2189 unsigned char sha1[20];
2190 char hdr[32];
2191 int hdrlen;
2192
2193 /* Normally if we have it in the pack then we do not bother writing
2194 * it out into .git/objects/??/?{38} file.
2195 */
2196 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
2197 if (returnsha1)
2198 hashcpy(returnsha1, sha1);
2199 if (has_sha1_file(sha1))
2200 return 0;
2201 return write_loose_object(sha1, hdr, hdrlen, buf, len, 0);
2202}
2203
2204int force_object_loose(const unsigned char *sha1, time_t mtime)
2205{
2206 struct stat st;
2207 void *buf;
2208 unsigned long len;
2209 enum object_type type;
2210 char hdr[32];
2211 int hdrlen;
2212
2213 if (find_sha1_file(sha1, &st))
2214 return 0;
2215 buf = read_packed_sha1(sha1, &type, &len);
2216 if (!buf)
2217 return error("cannot read sha1_file for %s", sha1_to_hex(sha1));
2218 hdrlen = sprintf(hdr, "%s %lu", typename(type), len) + 1;
2219 return write_loose_object(sha1, hdr, hdrlen, buf, len, mtime);
2220}
2221
4d548150
LT
2222/*
2223 * We need to unpack and recompress the object for writing
2224 * it out to a different file.
2225 */
2226static void *repack_object(const unsigned char *sha1, unsigned long *objsize)
a5eda52b 2227{
4d548150 2228 size_t size;
a5eda52b 2229 z_stream stream;
4d548150
LT
2230 unsigned char *unpacked;
2231 unsigned long len;
21666f1a 2232 enum object_type type;
d65a16f6 2233 char hdr[32];
4d548150
LT
2234 int hdrlen;
2235 void *buf;
bfc66daf 2236
a9486b02 2237 /* need to unpack and recompress it by itself */
21666f1a 2238 unpacked = read_packed_sha1(sha1, &type, &len);
8276c007
PE
2239 if (!unpacked)
2240 error("cannot read sha1_file for %s", sha1_to_hex(sha1));
a5eda52b 2241
21666f1a 2242 hdrlen = sprintf(hdr, "%s %lu", typename(type), len) + 1;
a5eda52b 2243
4d548150
LT
2244 /* Set it up */
2245 memset(&stream, 0, sizeof(stream));
12f6c308 2246 deflateInit(&stream, zlib_compression_level);
4d548150
LT
2247 size = deflateBound(&stream, len + hdrlen);
2248 buf = xmalloc(size);
a5eda52b 2249
4d548150
LT
2250 /* Compress it */
2251 stream.next_out = buf;
2252 stream.avail_out = size;
a5eda52b 2253
4d548150
LT
2254 /* First header.. */
2255 stream.next_in = (void *)hdr;
2256 stream.avail_in = hdrlen;
2257 while (deflate(&stream, 0) == Z_OK)
2258 /* nothing */;
bfc66daf 2259
4d548150
LT
2260 /* Then the data itself.. */
2261 stream.next_in = unpacked;
2262 stream.avail_in = len;
2263 while (deflate(&stream, Z_FINISH) == Z_OK)
2264 /* nothing */;
2265 deflateEnd(&stream);
2266 free(unpacked);
bfc66daf 2267
4d548150
LT
2268 *objsize = stream.total_out;
2269 return buf;
2270}
2271
2272int write_sha1_to_fd(int fd, const unsigned char *sha1)
2273{
2274 int retval;
2275 unsigned long objsize;
bb6b8e4f 2276 void *buf = map_sha1_file(sha1, &objsize);
4d548150
LT
2277
2278 if (buf) {
2279 retval = write_buffer(fd, buf, objsize);
2280 munmap(buf, objsize);
2281 return retval;
2282 }
2283
2284 buf = repack_object(sha1, &objsize);
2285 retval = write_buffer(fd, buf, objsize);
2286 free(buf);
2287 return retval;
a5eda52b
DB
2288}
2289
70b9829e
DB
2290int write_sha1_from_fd(const unsigned char *sha1, int fd, char *buffer,
2291 size_t bufsize, size_t *bufposn)
8237b185 2292{
230f1322 2293 char tmpfile[PATH_MAX];
8237b185
DB
2294 int local;
2295 z_stream stream;
2296 unsigned char real_sha1[20];
bf0f910d 2297 unsigned char discard[4096];
8237b185
DB
2298 int ret;
2299 SHA_CTX c;
2300
0e55181f 2301 snprintf(tmpfile, sizeof(tmpfile), "%s/tmp_obj_XXXXXX", get_object_directory());
8237b185 2302
230f1322 2303 local = mkstemp(tmpfile);
916d081b
PB
2304 if (local < 0) {
2305 if (errno == EPERM)
2306 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
2307 else
2308 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
2309 }
8237b185
DB
2310
2311 memset(&stream, 0, sizeof(stream));
2312
2313 inflateInit(&stream);
2314
2315 SHA1_Init(&c);
2316
2317 do {
2318 ssize_t size;
70b9829e
DB
2319 if (*bufposn) {
2320 stream.avail_in = *bufposn;
96ad15ae 2321 stream.next_in = (unsigned char *) buffer;
70b9829e
DB
2322 do {
2323 stream.next_out = discard;
2324 stream.avail_out = sizeof(discard);
2325 ret = inflate(&stream, Z_SYNC_FLUSH);
2326 SHA1_Update(&c, discard, sizeof(discard) -
2327 stream.avail_out);
2328 } while (stream.avail_in && ret == Z_OK);
4d548150
LT
2329 if (write_buffer(local, buffer, *bufposn - stream.avail_in) < 0)
2330 die("unable to write sha1 file");
70b9829e
DB
2331 memmove(buffer, buffer + *bufposn - stream.avail_in,
2332 stream.avail_in);
2333 *bufposn = stream.avail_in;
2334 if (ret != Z_OK)
2335 break;
2336 }
93d26e4c 2337 size = xread(fd, buffer + *bufposn, bufsize - *bufposn);
8237b185
DB
2338 if (size <= 0) {
2339 close(local);
230f1322 2340 unlink(tmpfile);
8237b185
DB
2341 if (!size)
2342 return error("Connection closed?");
2343 perror("Reading from connection");
2344 return -1;
2345 }
70b9829e
DB
2346 *bufposn += size;
2347 } while (1);
8237b185
DB
2348 inflateEnd(&stream);
2349
b5b8d814 2350 fchmod(local, 0444);
0d315468
JM
2351 if (close(local) != 0)
2352 die("unable to write sha1 file");
8237b185
DB
2353 SHA1_Final(real_sha1, &c);
2354 if (ret != Z_STREAM_END) {
230f1322 2355 unlink(tmpfile);
8237b185
DB
2356 return error("File %s corrupted", sha1_to_hex(sha1));
2357 }
a89fccd2 2358 if (hashcmp(sha1, real_sha1)) {
230f1322 2359 unlink(tmpfile);
bd2afde8 2360 return error("File %s has bad hash", sha1_to_hex(sha1));
8237b185 2361 }
230f1322
LT
2362
2363 return move_temp_to_file(tmpfile, sha1_file_name(sha1));
8237b185
DB
2364}
2365
bf592c50
DB
2366int has_pack_index(const unsigned char *sha1)
2367{
2368 struct stat st;
2369 if (stat(sha1_pack_index_name(sha1), &st))
2370 return 0;
2371 return 1;
2372}
2373
2374int has_pack_file(const unsigned char *sha1)
2375{
2376 struct stat st;
2377 if (stat(sha1_pack_name(sha1), &st))
2378 return 0;
2379 return 1;
2380}
2381
106d710b 2382int has_sha1_pack(const unsigned char *sha1, const char **ignore_packed)
dade09c2
LT
2383{
2384 struct pack_entry e;
106d710b 2385 return find_pack_entry(sha1, &e, ignore_packed);
dade09c2
LT
2386}
2387
8237b185
DB
2388int has_sha1_file(const unsigned char *sha1)
2389{
8237b185 2390 struct stat st;
1f688557
JH
2391 struct pack_entry e;
2392
106d710b 2393 if (find_pack_entry(sha1, &e, NULL))
1f688557 2394 return 1;
ab90ea5d 2395 return find_sha1_file(sha1, &st) ? 1 : 0;
8237b185 2396}
74400e71 2397
e7332f96
BE
2398int index_pipe(unsigned char *sha1, int fd, const char *type, int write_object)
2399{
fd17f5b5 2400 struct strbuf buf;
e7332f96 2401 int ret;
e7332f96 2402
fd17f5b5
PH
2403 strbuf_init(&buf, 0);
2404 if (strbuf_read(&buf, fd, 4096) < 0) {
2405 strbuf_release(&buf);
024510c8
DB
2406 return -1;
2407 }
e7332f96 2408
024510c8 2409 if (!type)
8e440259 2410 type = blob_type;
024510c8 2411 if (write_object)
fd17f5b5 2412 ret = write_sha1_file(buf.buf, buf.len, type, sha1);
abdc3fc8 2413 else
fd17f5b5
PH
2414 ret = hash_sha1_file(buf.buf, buf.len, type, sha1);
2415 strbuf_release(&buf);
2416
024510c8
DB
2417 return ret;
2418}
2419
edaec3fb 2420int index_fd(unsigned char *sha1, int fd, struct stat *st, int write_object,
53bca91a 2421 enum object_type type, const char *path)
74400e71 2422{
dc49cd76 2423 size_t size = xsize_t(st->st_size);
3a55602e 2424 void *buf = NULL;
6c510bee 2425 int ret, re_allocated = 0;
74400e71 2426
74400e71 2427 if (size)
c4712e45 2428 buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
74400e71 2429 close(fd);
74400e71 2430
7672db20 2431 if (!type)
edaec3fb 2432 type = OBJ_BLOB;
6c510bee
LT
2433
2434 /*
2435 * Convert blobs to git internal format
2436 */
d0d8e14d 2437 if ((type == OBJ_BLOB) && S_ISREG(st->st_mode)) {
5ecd293d
PH
2438 struct strbuf nbuf;
2439 strbuf_init(&nbuf, 0);
21e5ad50
SP
2440 if (convert_to_git(path, buf, size, &nbuf,
2441 write_object ? safe_crlf : 0)) {
ac78e548 2442 munmap(buf, size);
b315c5c0 2443 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
2444 re_allocated = 1;
2445 }
2446 }
2447
7672db20 2448 if (write_object)
edaec3fb 2449 ret = write_sha1_file(buf, size, typename(type), sha1);
abdc3fc8 2450 else
edaec3fb 2451 ret = hash_sha1_file(buf, size, typename(type), sha1);
6c510bee
LT
2452 if (re_allocated) {
2453 free(buf);
2454 return ret;
2455 }
aac17941
LT
2456 if (size)
2457 munmap(buf, size);
2458 return ret;
74400e71 2459}
ec1fcc16
JH
2460
2461int index_path(unsigned char *sha1, const char *path, struct stat *st, int write_object)
2462{
2463 int fd;
2464 char *target;
dc49cd76 2465 size_t len;
ec1fcc16
JH
2466
2467 switch (st->st_mode & S_IFMT) {
2468 case S_IFREG:
2469 fd = open(path, O_RDONLY);
2470 if (fd < 0)
2471 return error("open(\"%s\"): %s", path,
2472 strerror(errno));
53bca91a 2473 if (index_fd(sha1, fd, st, write_object, OBJ_BLOB, path) < 0)
ec1fcc16
JH
2474 return error("%s: failed to insert into database",
2475 path);
2476 break;
2477 case S_IFLNK:
dc49cd76
SP
2478 len = xsize_t(st->st_size);
2479 target = xmalloc(len + 1);
2480 if (readlink(path, target, len + 1) != st->st_size) {
ec1fcc16
JH
2481 char *errstr = strerror(errno);
2482 free(target);
2483 return error("readlink(\"%s\"): %s", path,
2484 errstr);
2485 }
abdc3fc8 2486 if (!write_object)
dc49cd76
SP
2487 hash_sha1_file(target, len, blob_type, sha1);
2488 else if (write_sha1_file(target, len, blob_type, sha1))
ec1fcc16
JH
2489 return error("%s: failed to insert into database",
2490 path);
2491 free(target);
2492 break;
f35a6d3b
LT
2493 case S_IFDIR:
2494 return resolve_gitlink_ref(path, "HEAD", sha1);
ec1fcc16
JH
2495 default:
2496 return error("%s: unsupported file type", path);
2497 }
2498 return 0;
2499}
a69e5429
JH
2500
2501int read_pack_header(int fd, struct pack_header *header)
2502{
c697ad14
HO
2503 if (read_in_full(fd, header, sizeof(*header)) < sizeof(*header))
2504 /* "eof before pack header was fully read" */
2505 return PH_ERROR_EOF;
2506
a69e5429
JH
2507 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2508 /* "protocol error (pack signature mismatch detected)" */
2509 return PH_ERROR_PACK_SIGNATURE;
2510 if (!pack_version_ok(header->hdr_version))
2511 /* "protocol error (pack version unsupported)" */
2512 return PH_ERROR_PROTOCOL;
2513 return 0;
2514}