Fixed broken git help -w when installing from RPM
[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
192a6be2 38const signed char hexval_table[256] = {
e49521b5
JH
39 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
40 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
41 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
42 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
43 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
44 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
45 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
46 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
47 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
48 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
49 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
50 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
51 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
52 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
53 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
54 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
55 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
56 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
57 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
58 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
59 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
60 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
61 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
62 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
63 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
64 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
65 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
66 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
67 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
68 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
69 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
70 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
71};
0fcfd160
LT
72
73int get_sha1_hex(const char *hex, unsigned char *sha1)
74{
75 int i;
76 for (i = 0; i < 20; i++) {
77 unsigned int val = (hexval(hex[0]) << 4) | hexval(hex[1]);
78 if (val & ~0xff)
79 return -1;
80 *sha1++ = val;
81 hex += 2;
82 }
83 return 0;
84}
85
8385abfd
JS
86static inline int offset_1st_component(const char *path)
87{
88 if (has_dos_drive_prefix(path))
89 return 2 + (path[2] == '/');
90 return *path == '/';
91}
92
b2cb9425
JH
93int safe_create_leading_directories(char *path)
94{
8385abfd 95 char *pos = path + offset_1st_component(path);
67d42212
JR
96 struct stat st;
97
b2cb9425
JH
98 while (pos) {
99 pos = strchr(pos, '/');
100 if (!pos)
101 break;
102 *pos = 0;
67d42212
JR
103 if (!stat(path, &st)) {
104 /* path exists */
105 if (!S_ISDIR(st.st_mode)) {
b2cb9425 106 *pos = '/';
67d42212 107 return -3;
b2cb9425 108 }
457f06d6 109 }
67d42212
JR
110 else if (mkdir(path, 0777)) {
111 *pos = '/';
112 return -1;
113 }
457f06d6
JS
114 else if (adjust_shared_perm(path)) {
115 *pos = '/';
116 return -2;
117 }
b2cb9425
JH
118 *pos++ = '/';
119 }
120 return 0;
121}
723c31fe 122
8e21d63b
JK
123int safe_create_leading_directories_const(const char *path)
124{
125 /* path points to cache entries, so xstrdup before messing with it */
126 char *buf = xstrdup(path);
127 int result = safe_create_leading_directories(buf);
128 free(buf);
129 return result;
130}
131
3bfaf018 132char *sha1_to_hex(const unsigned char *sha1)
0fcfd160 133{
dcb3450f
LT
134 static int bufno;
135 static char hexbuffer[4][50];
0fcfd160 136 static const char hex[] = "0123456789abcdef";
dcb3450f 137 char *buffer = hexbuffer[3 & ++bufno], *buf = buffer;
0fcfd160
LT
138 int i;
139
140 for (i = 0; i < 20; i++) {
141 unsigned int val = *sha1++;
142 *buf++ = hex[val >> 4];
143 *buf++ = hex[val & 0xf];
144 }
1e80e044
JS
145 *buf = '\0';
146
0fcfd160
LT
147 return buffer;
148}
149
ace1534d
JH
150static void fill_sha1_path(char *pathbuf, const unsigned char *sha1)
151{
152 int i;
153 for (i = 0; i < 20; i++) {
154 static char hex[] = "0123456789abcdef";
155 unsigned int val = sha1[i];
156 char *pos = pathbuf + i*2 + (i > 0);
157 *pos++ = hex[val >> 4];
158 *pos = hex[val & 0xf];
159 }
160}
161
0fcfd160
LT
162/*
163 * NOTE! This returns a statically allocated buffer, so you have to be
790296fd 164 * careful about using it. Do an "xstrdup()" if you need to save the
0fcfd160 165 * filename.
ace1534d
JH
166 *
167 * Also note that this returns the location for creating. Reading
168 * SHA1 file can happen from any alternate directory listed in the
d19938ab 169 * DB_ENVIRONMENT environment variable if it is not found in
ace1534d 170 * the primary object database.
0fcfd160
LT
171 */
172char *sha1_file_name(const unsigned char *sha1)
173{
0fcfd160
LT
174 static char *name, *base;
175
176 if (!base) {
d19938ab 177 const char *sha1_file_directory = get_object_directory();
0fcfd160 178 int len = strlen(sha1_file_directory);
812666c8 179 base = xmalloc(len + 60);
0fcfd160
LT
180 memcpy(base, sha1_file_directory, len);
181 memset(base+len, 0, 60);
182 base[len] = '/';
183 base[len+3] = '/';
184 name = base + len + 1;
185 }
ace1534d 186 fill_sha1_path(name, sha1);
0fcfd160
LT
187 return base;
188}
189
633f43e1 190static char *sha1_get_pack_name(const unsigned char *sha1,
6eec46bd 191 char **name, char **base, const char *which)
bf592c50
DB
192{
193 static const char hex[] = "0123456789abcdef";
633f43e1 194 char *buf;
bf592c50
DB
195 int i;
196
633f43e1 197 if (!*base) {
bf592c50
DB
198 const char *sha1_file_directory = get_object_directory();
199 int len = strlen(sha1_file_directory);
633f43e1 200 *base = xmalloc(len + 60);
6eec46bd
JH
201 sprintf(*base, "%s/pack/pack-1234567890123456789012345678901234567890.%s",
202 sha1_file_directory, which);
633f43e1 203 *name = *base + len + 11;
bf592c50
DB
204 }
205
633f43e1 206 buf = *name;
bf592c50
DB
207
208 for (i = 0; i < 20; i++) {
209 unsigned int val = *sha1++;
210 *buf++ = hex[val >> 4];
211 *buf++ = hex[val & 0xf];
212 }
a6080a0a 213
633f43e1 214 return *base;
bf592c50
DB
215}
216
633f43e1 217char *sha1_pack_name(const unsigned char *sha1)
bf592c50 218{
633f43e1 219 static char *name, *base;
bf592c50 220
6eec46bd 221 return sha1_get_pack_name(sha1, &name, &base, "pack");
633f43e1 222}
bf592c50 223
633f43e1
HO
224char *sha1_pack_index_name(const unsigned char *sha1)
225{
226 static char *name, *base;
a6080a0a 227
6eec46bd 228 return sha1_get_pack_name(sha1, &name, &base, "idx");
bf592c50
DB
229}
230
d5a63b99
JH
231struct alternate_object_database *alt_odb_list;
232static struct alternate_object_database **alt_odb_tail;
ace1534d 233
c2f493a4
MW
234static void read_info_alternates(const char * alternates, int depth);
235
ddd5d056
JH
236/*
237 * Prepare alternate object database registry.
d5a63b99
JH
238 *
239 * The variable alt_odb_list points at the list of struct
240 * alternate_object_database. The elements on this list come from
241 * non-empty elements from colon separated ALTERNATE_DB_ENVIRONMENT
242 * environment variable, and $GIT_OBJECT_DIRECTORY/info/alternates,
1494e038
JH
243 * whose contents is similar to that environment variable but can be
244 * LF separated. Its base points at a statically allocated buffer that
d5a63b99
JH
245 * contains "/the/directory/corresponding/to/.git/objects/...", while
246 * its name points just after the slash at the end of ".git/objects/"
247 * in the example above, and has enough space to hold 40-byte hex
248 * SHA1, an extra slash for the first level indirection, and the
249 * terminating NUL.
ddd5d056 250 */
c2f493a4 251static int link_alt_odb_entry(const char * entry, int len, const char * relative_base, int depth)
ace1534d 252{
c2f493a4 253 struct stat st;
1494e038 254 const char *objdir = get_object_directory();
c2f493a4
MW
255 struct alternate_object_database *ent;
256 struct alternate_object_database *alt;
257 /* 43 = 40-byte + 2 '/' + terminating NUL */
258 int pfxlen = len;
259 int entlen = pfxlen + 43;
ccfd3e99 260 int base_len = -1;
d5a63b99 261
85dadc38 262 if (!is_absolute_path(entry) && relative_base) {
c2f493a4
MW
263 /* Relative alt-odb */
264 if (base_len < 0)
265 base_len = strlen(relative_base) + 1;
266 entlen += base_len;
267 pfxlen += base_len;
268 }
269 ent = xmalloc(sizeof(*ent) + entlen);
270
85dadc38 271 if (!is_absolute_path(entry) && relative_base) {
c2f493a4
MW
272 memcpy(ent->base, relative_base, base_len - 1);
273 ent->base[base_len - 1] = '/';
274 memcpy(ent->base + base_len, entry, len);
275 }
276 else
277 memcpy(ent->base, entry, pfxlen);
278
279 ent->name = ent->base + pfxlen + 1;
280 ent->base[pfxlen + 3] = '/';
281 ent->base[pfxlen] = ent->base[entlen-1] = 0;
282
283 /* Detect cases where alternate disappeared */
284 if (stat(ent->base, &st) || !S_ISDIR(st.st_mode)) {
285 error("object directory %s does not exist; "
286 "check .git/objects/info/alternates.",
287 ent->base);
288 free(ent);
289 return -1;
290 }
291
292 /* Prevent the common mistake of listing the same
293 * thing twice, or object directory itself.
294 */
295 for (alt = alt_odb_list; alt; alt = alt->next) {
296 if (!memcmp(ent->base, alt->base, pfxlen)) {
297 free(ent);
298 return -1;
299 }
300 }
301 if (!memcmp(ent->base, objdir, pfxlen)) {
302 free(ent);
303 return -1;
304 }
305
306 /* add the alternate entry */
307 *alt_odb_tail = ent;
308 alt_odb_tail = &(ent->next);
309 ent->next = NULL;
310
311 /* recursively add alternates */
312 read_info_alternates(ent->base, depth + 1);
313
314 ent->base[pfxlen] = '/';
315
316 return 0;
317}
318
319static void link_alt_odb_entries(const char *alt, const char *ep, int sep,
320 const char *relative_base, int depth)
321{
322 const char *cp, *last;
323
324 if (depth > 5) {
325 error("%s: ignoring alternate object stores, nesting too deep.",
326 relative_base);
327 return;
328 }
329
d5a63b99 330 last = alt;
9577e7e3
JH
331 while (last < ep) {
332 cp = last;
333 if (cp < ep && *cp == '#') {
334 while (cp < ep && *cp != sep)
335 cp++;
336 last = cp + 1;
337 continue;
338 }
c2f493a4
MW
339 while (cp < ep && *cp != sep)
340 cp++;
d5a63b99 341 if (last != cp) {
85dadc38 342 if (!is_absolute_path(last) && depth) {
c2f493a4
MW
343 error("%s: ignoring relative alternate object store %s",
344 relative_base, last);
345 } else {
346 link_alt_odb_entry(last, cp - last,
347 relative_base, depth);
1494e038 348 }
d5a63b99 349 }
9577e7e3 350 while (cp < ep && *cp == sep)
d5a63b99
JH
351 cp++;
352 last = cp;
9577e7e3 353 }
d5a63b99
JH
354}
355
c2f493a4 356static void read_info_alternates(const char * relative_base, int depth)
d5a63b99 357{
9577e7e3 358 char *map;
dc49cd76 359 size_t mapsz;
d5a63b99 360 struct stat st;
9cb18f56
JM
361 const char alt_file_name[] = "info/alternates";
362 /* Given that relative_base is no longer than PATH_MAX,
363 ensure that "path" has enough space to append "/", the
364 file name, "info/alternates", and a trailing NUL. */
365 char path[PATH_MAX + 1 + sizeof alt_file_name];
c2f493a4 366 int fd;
d5a63b99 367
9cb18f56 368 sprintf(path, "%s/%s", relative_base, alt_file_name);
d5a63b99
JH
369 fd = open(path, O_RDONLY);
370 if (fd < 0)
371 return;
372 if (fstat(fd, &st) || (st.st_size == 0)) {
373 close(fd);
9a217f2a 374 return;
ace1534d 375 }
dc49cd76
SP
376 mapsz = xsize_t(st.st_size);
377 map = xmmap(NULL, mapsz, PROT_READ, MAP_PRIVATE, fd, 0);
d5a63b99 378 close(fd);
d5a63b99 379
dc49cd76 380 link_alt_odb_entries(map, map + mapsz, '\n', relative_base, depth);
c2f493a4 381
dc49cd76 382 munmap(map, mapsz);
ace1534d
JH
383}
384
bef70b22
DB
385void add_to_alternates_file(const char *reference)
386{
387 struct lock_file *lock = xcalloc(1, sizeof(struct lock_file));
acd3b9ec 388 int fd = hold_lock_file_for_append(lock, git_path("objects/info/alternates"), LOCK_DIE_ON_ERROR);
bef70b22
DB
389 char *alt = mkpath("%s/objects\n", reference);
390 write_or_die(fd, alt, strlen(alt));
391 if (commit_lock_file(lock))
392 die("could not close alternates file");
393 if (alt_odb_tail)
394 link_alt_odb_entries(alt, alt + strlen(alt), '\n', NULL, 0);
395}
396
c2f493a4
MW
397void prepare_alt_odb(void)
398{
554fe20d 399 const char *alt;
c2f493a4 400
7dc24aa5
SP
401 if (alt_odb_tail)
402 return;
403
c2f493a4
MW
404 alt = getenv(ALTERNATE_DB_ENVIRONMENT);
405 if (!alt) alt = "";
406
c2f493a4 407 alt_odb_tail = &alt_odb_list;
80ba074f 408 link_alt_odb_entries(alt, alt + strlen(alt), PATH_SEP, NULL, 0);
c2f493a4
MW
409
410 read_info_alternates(get_object_directory(), 0);
411}
412
0f4dc14a 413static int has_loose_object_local(const unsigned char *sha1)
ace1534d 414{
ace1534d 415 char *name = sha1_file_name(sha1);
0f4dc14a
BC
416 return !access(name, F_OK);
417}
ace1534d 418
0f4dc14a
BC
419int has_loose_object_nonlocal(const unsigned char *sha1)
420{
421 struct alternate_object_database *alt;
9a217f2a 422 prepare_alt_odb();
d5a63b99 423 for (alt = alt_odb_list; alt; alt = alt->next) {
0f4dc14a 424 fill_sha1_path(alt->name, sha1);
c529d75a
LT
425 if (!access(alt->base, F_OK))
426 return 1;
ace1534d 427 }
c529d75a 428 return 0;
ace1534d
JH
429}
430
0f4dc14a
BC
431static int has_loose_object(const unsigned char *sha1)
432{
433 return has_loose_object_local(sha1) ||
434 has_loose_object_nonlocal(sha1);
435}
436
60bb8b14 437static unsigned int pack_used_ctr;
a53128b6
SP
438static unsigned int pack_mmap_calls;
439static unsigned int peak_pack_open_windows;
440static unsigned int pack_open_windows;
441static size_t peak_pack_mapped;
60bb8b14 442static size_t pack_mapped;
9a217f2a 443struct packed_git *packed_git;
1f688557 444
b79d18c9 445void pack_report(void)
a53128b6
SP
446{
447 fprintf(stderr,
e05db0fd
PR
448 "pack_report: getpagesize() = %10" SZ_FMT "\n"
449 "pack_report: core.packedGitWindowSize = %10" SZ_FMT "\n"
450 "pack_report: core.packedGitLimit = %10" SZ_FMT "\n",
9e42d6a1
SP
451 sz_fmt(getpagesize()),
452 sz_fmt(packed_git_window_size),
453 sz_fmt(packed_git_limit));
a53128b6
SP
454 fprintf(stderr,
455 "pack_report: pack_used_ctr = %10u\n"
456 "pack_report: pack_mmap_calls = %10u\n"
457 "pack_report: pack_open_windows = %10u / %10u\n"
e05db0fd
PR
458 "pack_report: pack_mapped = "
459 "%10" SZ_FMT " / %10" SZ_FMT "\n",
a53128b6
SP
460 pack_used_ctr,
461 pack_mmap_calls,
462 pack_open_windows, peak_pack_open_windows,
9e42d6a1 463 sz_fmt(pack_mapped), sz_fmt(peak_pack_mapped));
a53128b6
SP
464}
465
42873078 466static int check_packed_git_idx(const char *path, struct packed_git *p)
1f688557
JH
467{
468 void *idx_map;
42873078 469 struct pack_idx_header *hdr;
dc49cd76 470 size_t idx_size;
74e34e1f 471 uint32_t version, nr, i, *index;
e93ec6f9 472 int fd = open(path, O_RDONLY);
1f688557 473 struct stat st;
42873078 474
1f688557
JH
475 if (fd < 0)
476 return -1;
477 if (fstat(fd, &st)) {
478 close(fd);
479 return -1;
480 }
dc49cd76 481 idx_size = xsize_t(st.st_size);
2d88451b
SP
482 if (idx_size < 4 * 256 + 20 + 20) {
483 close(fd);
484 return error("index file %s is too small", path);
485 }
c4712e45 486 idx_map = xmmap(NULL, idx_size, PROT_READ, MAP_PRIVATE, fd, 0);
1f688557 487 close(fd);
1f688557 488
42873078
NP
489 hdr = idx_map;
490 if (hdr->idx_signature == htonl(PACK_IDX_SIGNATURE)) {
74e34e1f
NP
491 version = ntohl(hdr->idx_version);
492 if (version < 2 || version > 2) {
493 munmap(idx_map, idx_size);
6e1c2344 494 return error("index file %s is version %"PRIu32
74e34e1f
NP
495 " and is not supported by this binary"
496 " (try upgrading GIT to a newer version)",
497 path, version);
498 }
499 } else
500 version = 1;
df1b059d 501
1f688557 502 nr = 0;
42873078 503 index = idx_map;
74e34e1f
NP
504 if (version > 1)
505 index += 2; /* skip index header */
1f688557 506 for (i = 0; i < 256; i++) {
326bf396 507 uint32_t n = ntohl(index[i]);
2d88451b
SP
508 if (n < nr) {
509 munmap(idx_map, idx_size);
df1b059d 510 return error("non-monotonic index %s", path);
2d88451b 511 }
1f688557
JH
512 nr = n;
513 }
514
74e34e1f
NP
515 if (version == 1) {
516 /*
517 * Total size:
518 * - 256 index entries 4 bytes each
519 * - 24-byte entries * nr (20-byte sha1 + 4-byte offset)
520 * - 20-byte SHA1 of the packfile
521 * - 20-byte SHA1 file checksum
522 */
523 if (idx_size != 4*256 + nr * 24 + 20 + 20) {
524 munmap(idx_map, idx_size);
eef427a0 525 return error("wrong index v1 file size in %s", path);
74e34e1f
NP
526 }
527 } else if (version == 2) {
528 /*
529 * Minimum size:
530 * - 8 bytes of header
531 * - 256 index entries 4 bytes each
532 * - 20-byte sha1 entry * nr
533 * - 4-byte crc entry * nr
534 * - 4-byte offset entry * nr
535 * - 20-byte SHA1 of the packfile
536 * - 20-byte SHA1 file checksum
537 * And after the 4-byte offset table might be a
538 * variable sized table containing 8-byte entries
539 * for offsets larger than 2^31.
540 */
541 unsigned long min_size = 8 + 4*256 + nr*(20 + 4 + 4) + 20 + 20;
1164f1e4
LT
542 unsigned long max_size = min_size;
543 if (nr)
544 max_size += (nr - 1)*8;
545 if (idx_size < min_size || idx_size > max_size) {
74e34e1f 546 munmap(idx_map, idx_size);
eef427a0 547 return error("wrong index v2 file size in %s", path);
74e34e1f 548 }
7109c889
JH
549 if (idx_size != min_size &&
550 /*
551 * make sure we can deal with large pack offsets.
552 * 31-bit signed offset won't be enough, neither
553 * 32-bit unsigned one will be.
554 */
555 (sizeof(off_t) <= 4)) {
556 munmap(idx_map, idx_size);
557 return error("pack too large for current definition of off_t in %s", path);
74e34e1f 558 }
2d88451b 559 }
1f688557 560
74e34e1f 561 p->index_version = version;
42873078
NP
562 p->index_data = idx_map;
563 p->index_size = idx_size;
57059091 564 p->num_objects = nr;
1f688557
JH
565 return 0;
566}
567
bc8e478a 568int open_pack_index(struct packed_git *p)
d079837e
SP
569{
570 char *idx_name;
571 int ret;
572
573 if (p->index_data)
574 return 0;
575
576 idx_name = xstrdup(p->pack_name);
577 strcpy(idx_name + strlen(idx_name) - strlen(".pack"), ".idx");
578 ret = check_packed_git_idx(idx_name, p);
579 free(idx_name);
580 return ret;
581}
582
11daf39b
SP
583static void scan_windows(struct packed_git *p,
584 struct packed_git **lru_p,
585 struct pack_window **lru_w,
586 struct pack_window **lru_l)
1f688557 587{
11daf39b
SP
588 struct pack_window *w, *w_l;
589
590 for (w_l = NULL, w = p->windows; w; w = w->next) {
591 if (!w->inuse_cnt) {
592 if (!*lru_w || w->last_used < (*lru_w)->last_used) {
593 *lru_p = p;
594 *lru_w = w;
595 *lru_l = w_l;
54044bf8 596 }
54044bf8 597 }
11daf39b 598 w_l = w;
f9253394 599 }
11daf39b
SP
600}
601
d1efefa4 602static int unuse_one_window(struct packed_git *current, int keep_fd)
11daf39b
SP
603{
604 struct packed_git *p, *lru_p = NULL;
605 struct pack_window *lru_w = NULL, *lru_l = NULL;
606
607 if (current)
608 scan_windows(current, &lru_p, &lru_w, &lru_l);
609 for (p = packed_git; p; p = p->next)
610 scan_windows(p, &lru_p, &lru_w, &lru_l);
54044bf8
SP
611 if (lru_p) {
612 munmap(lru_w->base, lru_w->len);
613 pack_mapped -= lru_w->len;
614 if (lru_l)
615 lru_l->next = lru_w->next;
616 else {
617 lru_p->windows = lru_w->next;
d1efefa4 618 if (!lru_p->windows && lru_p->pack_fd != keep_fd) {
54044bf8
SP
619 close(lru_p->pack_fd);
620 lru_p->pack_fd = -1;
621 }
622 }
623 free(lru_w);
a53128b6 624 pack_open_windows--;
54044bf8
SP
625 return 1;
626 }
627 return 0;
f9253394
JH
628}
629
d1efefa4 630void release_pack_memory(size_t need, int fd)
97bfeb34
SP
631{
632 size_t cur = pack_mapped;
d1efefa4 633 while (need >= (cur - pack_mapped) && unuse_one_window(NULL, fd))
97bfeb34
SP
634 ; /* nothing */
635}
636
c9ced051
SP
637void close_pack_windows(struct packed_git *p)
638{
639 while (p->windows) {
640 struct pack_window *w = p->windows;
641
642 if (w->inuse_cnt)
643 die("pack '%s' still has open windows to it",
644 p->pack_name);
645 munmap(w->base, w->len);
646 pack_mapped -= w->len;
647 pack_open_windows--;
648 p->windows = w->next;
649 free(w);
650 }
651}
652
03e79c88 653void unuse_pack(struct pack_window **w_cursor)
f9253394 654{
03e79c88
SP
655 struct pack_window *w = *w_cursor;
656 if (w) {
657 w->inuse_cnt--;
658 *w_cursor = NULL;
659 }
1f688557
JH
660}
661
c74faea1
NP
662/*
663 * This is used by git-repack in case a newly created pack happens to
664 * contain the same set of objects as an existing one. In that case
665 * the resulting file might be different even if its name would be the
666 * same. It is best to close any reference to the old pack before it is
667 * replaced on disk. Of course no index pointers nor windows for given pack
668 * must subsist at this point. If ever objects from this pack are requested
669 * again, the new version of the pack will be reinitialized through
670 * reprepare_packed_git().
671 */
672void free_pack_by_name(const char *pack_name)
673{
674 struct packed_git *p, **pp = &packed_git;
675
676 while (*pp) {
677 p = *pp;
678 if (strcmp(pack_name, p->pack_name) == 0) {
679 close_pack_windows(p);
680 if (p->pack_fd != -1)
681 close(p->pack_fd);
682 if (p->index_data)
683 munmap((void *)p->index_data, p->index_size);
684 free(p->bad_object_sha1);
685 *pp = p->next;
686 free(p);
687 return;
688 }
689 pp = &p->next;
690 }
691}
692
3cf8b462
SP
693/*
694 * Do not call this directly as this leaks p->pack_fd on error return;
695 * call open_packed_git() instead.
696 */
697static int open_packed_git_1(struct packed_git *p)
1f688557 698{
9bc879c1
SP
699 struct stat st;
700 struct pack_header hdr;
701 unsigned char sha1[20];
702 unsigned char *idx_sha1;
2c039da8 703 long fd_flag;
9bc879c1 704
d079837e
SP
705 if (!p->index_data && open_pack_index(p))
706 return error("packfile %s index unavailable", p->pack_name);
707
9bc879c1
SP
708 p->pack_fd = open(p->pack_name, O_RDONLY);
709 if (p->pack_fd < 0 || fstat(p->pack_fd, &st))
072db278 710 return -1;
9bc879c1
SP
711
712 /* If we created the struct before we had the pack we lack size. */
bf592c50 713 if (!p->pack_size) {
bf592c50 714 if (!S_ISREG(st.st_mode))
072db278 715 return error("packfile %s not a regular file", p->pack_name);
bf592c50 716 p->pack_size = st.st_size;
9bc879c1 717 } else if (p->pack_size != st.st_size)
072db278 718 return error("packfile %s size changed", p->pack_name);
9bc879c1 719
2c039da8
JH
720 /* We leave these file descriptors open with sliding mmap;
721 * there is no point keeping them open across exec(), though.
722 */
723 fd_flag = fcntl(p->pack_fd, F_GETFD, 0);
724 if (fd_flag < 0)
072db278 725 return error("cannot determine file descriptor flags");
2c039da8
JH
726 fd_flag |= FD_CLOEXEC;
727 if (fcntl(p->pack_fd, F_SETFD, fd_flag) == -1)
072db278 728 return error("cannot set FD_CLOEXEC");
2c039da8 729
9bc879c1 730 /* Verify we recognize this pack file format. */
e6e2bd62 731 if (read_in_full(p->pack_fd, &hdr, sizeof(hdr)) != sizeof(hdr))
072db278 732 return error("file %s is far too short to be a packfile", p->pack_name);
9bc879c1 733 if (hdr.hdr_signature != htonl(PACK_SIGNATURE))
072db278 734 return error("file %s is not a GIT packfile", p->pack_name);
9bc879c1 735 if (!pack_version_ok(hdr.hdr_version))
6e1c2344
RJ
736 return error("packfile %s is version %"PRIu32" and not"
737 " supported (try upgrading GIT to a newer version)",
9bc879c1
SP
738 p->pack_name, ntohl(hdr.hdr_version));
739
740 /* Verify the pack matches its index. */
57059091 741 if (p->num_objects != ntohl(hdr.hdr_entries))
6e1c2344
RJ
742 return error("packfile %s claims to have %"PRIu32" objects"
743 " while index indicates %"PRIu32" objects",
57059091
NP
744 p->pack_name, ntohl(hdr.hdr_entries),
745 p->num_objects);
9bc879c1 746 if (lseek(p->pack_fd, p->pack_size - sizeof(sha1), SEEK_SET) == -1)
072db278 747 return error("end of packfile %s is unavailable", p->pack_name);
e6e2bd62 748 if (read_in_full(p->pack_fd, sha1, sizeof(sha1)) != sizeof(sha1))
072db278 749 return error("packfile %s signature is unavailable", p->pack_name);
42873078 750 idx_sha1 = ((unsigned char *)p->index_data) + p->index_size - 40;
9bc879c1 751 if (hashcmp(sha1, idx_sha1))
072db278
SP
752 return error("packfile %s does not match index", p->pack_name);
753 return 0;
9bc879c1
SP
754}
755
3cf8b462
SP
756static int open_packed_git(struct packed_git *p)
757{
758 if (!open_packed_git_1(p))
759 return 0;
760 if (p->pack_fd != -1) {
761 close(p->pack_fd);
762 p->pack_fd = -1;
763 }
764 return -1;
765}
766
c4001d92 767static int in_window(struct pack_window *win, off_t offset)
60bb8b14
SP
768{
769 /* We must promise at least 20 bytes (one hash) after the
770 * offset is available from this window, otherwise the offset
771 * is not actually in this window and a different window (which
772 * has that one hash excess) must be used. This is to support
773 * the object header and delta base parsing routines below.
774 */
775 off_t win_off = win->offset;
776 return win_off <= offset
777 && (offset + 20) <= (win_off + win->len);
778}
779
03e79c88
SP
780unsigned char* use_pack(struct packed_git *p,
781 struct pack_window **w_cursor,
c4001d92 782 off_t offset,
03e79c88 783 unsigned int *left)
9bc879c1 784{
60bb8b14 785 struct pack_window *win = *w_cursor;
03e79c88 786
072db278
SP
787 if (p->pack_fd == -1 && open_packed_git(p))
788 die("packfile %s cannot be accessed", p->pack_name);
60bb8b14
SP
789
790 /* Since packfiles end in a hash of their content and its
791 * pointless to ask for an offset into the middle of that
792 * hash, and the in_window function above wouldn't match
793 * don't allow an offset too close to the end of the file.
794 */
795 if (offset > (p->pack_size - 20))
796 die("offset beyond end of packfile (truncated pack?)");
797
798 if (!win || !in_window(win, offset)) {
799 if (win)
800 win->inuse_cnt--;
801 for (win = p->windows; win; win = win->next) {
802 if (in_window(win, offset))
803 break;
804 }
805 if (!win) {
78a28df9 806 size_t window_align = packed_git_window_size / 2;
dc49cd76 807 off_t len;
60bb8b14 808 win = xcalloc(1, sizeof(*win));
78a28df9 809 win->offset = (offset / window_align) * window_align;
dc49cd76
SP
810 len = p->pack_size - win->offset;
811 if (len > packed_git_window_size)
812 len = packed_git_window_size;
813 win->len = (size_t)len;
60bb8b14 814 pack_mapped += win->len;
11daf39b 815 while (packed_git_limit < pack_mapped
d1efefa4 816 && unuse_one_window(p, p->pack_fd))
60bb8b14 817 ; /* nothing */
c4712e45 818 win->base = xmmap(NULL, win->len,
60bb8b14
SP
819 PROT_READ, MAP_PRIVATE,
820 p->pack_fd, win->offset);
821 if (win->base == MAP_FAILED)
73b4e4be
SP
822 die("packfile %s cannot be mapped: %s",
823 p->pack_name,
824 strerror(errno));
a53128b6
SP
825 pack_mmap_calls++;
826 pack_open_windows++;
827 if (pack_mapped > peak_pack_mapped)
828 peak_pack_mapped = pack_mapped;
829 if (pack_open_windows > peak_pack_open_windows)
830 peak_pack_open_windows = pack_open_windows;
60bb8b14
SP
831 win->next = p->windows;
832 p->windows = win;
833 }
1f688557 834 }
03e79c88
SP
835 if (win != *w_cursor) {
836 win->last_used = pack_used_ctr++;
837 win->inuse_cnt++;
838 *w_cursor = win;
839 }
60bb8b14 840 offset -= win->offset;
03e79c88 841 if (left)
dc49cd76 842 *left = win->len - xsize_t(offset);
03e79c88 843 return win->base + offset;
1f688557
JH
844}
845
27d69a46
NP
846static struct packed_git *alloc_packed_git(int extra)
847{
848 struct packed_git *p = xmalloc(sizeof(*p) + extra);
849 memset(p, 0, sizeof(*p));
850 p->pack_fd = -1;
851 return p;
852}
853
42873078 854struct packed_git *add_packed_git(const char *path, int path_len, int local)
1f688557
JH
855{
856 struct stat st;
27d69a46 857 struct packed_git *p = alloc_packed_git(path_len + 2);
1f688557 858
42873078
NP
859 /*
860 * Make sure a corresponding .pack file exists and that
861 * the index looks sane.
862 */
863 path_len -= strlen(".idx");
27d69a46
NP
864 if (path_len < 1) {
865 free(p);
1f688557 866 return NULL;
27d69a46 867 }
42873078 868 memcpy(p->pack_name, path, path_len);
8d25931d
BC
869
870 strcpy(p->pack_name + path_len, ".keep");
871 if (!access(p->pack_name, F_OK))
872 p->pack_keep = 1;
873
42873078 874 strcpy(p->pack_name + path_len, ".pack");
d079837e 875 if (stat(p->pack_name, &st) || !S_ISREG(st.st_mode)) {
42873078 876 free(p);
1f688557
JH
877 return NULL;
878 }
42873078 879
1f688557
JH
880 /* ok, it looks sane as far as we can check without
881 * actually mapping the pack file.
882 */
1f688557 883 p->pack_size = st.st_size;
9d835df2 884 p->pack_local = local;
b867092f 885 p->mtime = st.st_mtime;
42873078
NP
886 if (path_len < 40 || get_sha1_hex(path + path_len - 40, p->sha1))
887 hashclr(p->sha1);
1f688557
JH
888 return p;
889}
890
bf592c50 891struct packed_git *parse_pack_index(unsigned char *sha1)
c508df5e 892{
79c6dca4 893 const char *idx_path = sha1_pack_index_name(sha1);
42873078 894 const char *path = sha1_pack_name(sha1);
27d69a46 895 struct packed_git *p = alloc_packed_git(strlen(path) + 1);
bf592c50 896
27d69a46
NP
897 strcpy(p->pack_name, path);
898 hashcpy(p->sha1, sha1);
42873078
NP
899 if (check_packed_git_idx(idx_path, p)) {
900 free(p);
bf592c50 901 return NULL;
42873078 902 }
bf592c50 903
bf592c50
DB
904 return p;
905}
906
907void install_packed_git(struct packed_git *pack)
908{
909 pack->next = packed_git;
910 packed_git = pack;
911}
912
9d835df2 913static void prepare_packed_git_one(char *objdir, int local)
1f688557 914{
9cb18f56
JM
915 /* Ensure that this buffer is large enough so that we can
916 append "/pack/" without clobbering the stack even if
917 strlen(objdir) were PATH_MAX. */
918 char path[PATH_MAX + 1 + 4 + 1 + 1];
1f688557
JH
919 int len;
920 DIR *dir;
921 struct dirent *de;
922
923 sprintf(path, "%s/pack", objdir);
924 len = strlen(path);
925 dir = opendir(path);
b5b16990 926 if (!dir) {
26125f6b 927 if (errno != ENOENT)
bd2afde8 928 error("unable to open object pack directory: %s: %s",
26125f6b 929 path, strerror(errno));
1f688557 930 return;
b5b16990 931 }
1f688557
JH
932 path[len++] = '/';
933 while ((de = readdir(dir)) != NULL) {
934 int namelen = strlen(de->d_name);
935 struct packed_git *p;
936
5bb1cda5 937 if (!has_extension(de->d_name, ".idx"))
1f688557
JH
938 continue;
939
9cb18f56
JM
940 if (len + namelen + 1 > sizeof(path))
941 continue;
942
54a15a8d 943 /* Don't reopen a pack we already have. */
1f688557 944 strcpy(path + len, de->d_name);
86f7780c
JK
945 for (p = packed_git; p; p = p->next) {
946 if (!memcmp(path, p->pack_name, len + namelen - 4))
947 break;
948 }
949 if (p)
950 continue;
54a15a8d
SP
951 /* See if it really is a valid .idx file with corresponding
952 * .pack file that we can map.
953 */
9d835df2 954 p = add_packed_git(path, len + namelen, local);
1f688557
JH
955 if (!p)
956 continue;
625e9421 957 install_packed_git(p);
1f688557 958 }
5b35bcd5 959 closedir(dir);
1f688557
JH
960}
961
b867092f
JH
962static int sort_pack(const void *a_, const void *b_)
963{
964 struct packed_git *a = *((struct packed_git **)a_);
965 struct packed_git *b = *((struct packed_git **)b_);
966 int st;
967
968 /*
969 * Local packs tend to contain objects specific to our
970 * variant of the project than remote ones. In addition,
971 * remote ones could be on a network mounted filesystem.
972 * Favor local ones for these reasons.
973 */
974 st = a->pack_local - b->pack_local;
975 if (st)
976 return -st;
977
978 /*
979 * Younger packs tend to contain more recent objects,
980 * and more recent objects tend to get accessed more
981 * often.
982 */
983 if (a->mtime < b->mtime)
984 return 1;
985 else if (a->mtime == b->mtime)
986 return 0;
987 return -1;
988}
989
990static void rearrange_packed_git(void)
991{
992 struct packed_git **ary, *p;
993 int i, n;
994
995 for (n = 0, p = packed_git; p; p = p->next)
996 n++;
997 if (n < 2)
998 return;
999
1000 /* prepare an array of packed_git for easier sorting */
1001 ary = xcalloc(n, sizeof(struct packed_git *));
1002 for (n = 0, p = packed_git; p; p = p->next)
1003 ary[n++] = p;
1004
1005 qsort(ary, n, sizeof(struct packed_git *), sort_pack);
1006
1007 /* link them back again */
1008 for (i = 0; i < n - 1; i++)
1009 ary[i]->next = ary[i + 1];
1010 ary[n - 1]->next = NULL;
1011 packed_git = ary[0];
1012
1013 free(ary);
1014}
1015
637cdd9d 1016static int prepare_packed_git_run_once = 0;
9a217f2a 1017void prepare_packed_git(void)
1f688557 1018{
d5a63b99 1019 struct alternate_object_database *alt;
1f688557 1020
637cdd9d 1021 if (prepare_packed_git_run_once)
1f688557 1022 return;
9d835df2 1023 prepare_packed_git_one(get_object_directory(), 1);
9a217f2a 1024 prepare_alt_odb();
d5a63b99 1025 for (alt = alt_odb_list; alt; alt = alt->next) {
1494e038 1026 alt->name[-1] = 0;
9d835df2 1027 prepare_packed_git_one(alt->base, 0);
1494e038 1028 alt->name[-1] = '/';
1f688557 1029 }
b867092f 1030 rearrange_packed_git();
637cdd9d
JK
1031 prepare_packed_git_run_once = 1;
1032}
1033
fc04c412 1034void reprepare_packed_git(void)
637cdd9d 1035{
4b480c67 1036 discard_revindex();
637cdd9d
JK
1037 prepare_packed_git_run_once = 0;
1038 prepare_packed_git();
1f688557
JH
1039}
1040
8eca0b47
NP
1041static void mark_bad_packed_object(struct packed_git *p,
1042 const unsigned char *sha1)
1043{
1044 unsigned i;
1045 for (i = 0; i < p->num_bad_objects; i++)
1046 if (!hashcmp(sha1, p->bad_object_sha1 + 20 * i))
1047 return;
1048 p->bad_object_sha1 = xrealloc(p->bad_object_sha1, 20 * (p->num_bad_objects + 1));
1049 hashcpy(p->bad_object_sha1 + 20 * p->num_bad_objects, sha1);
1050 p->num_bad_objects++;
1051}
1052
ac939109
NP
1053static int has_packed_and_bad(const unsigned char *sha1)
1054{
1055 struct packed_git *p;
1056 unsigned i;
1057
1058 for (p = packed_git; p; p = p->next)
1059 for (i = 0; i < p->num_bad_objects; i++)
1060 if (!hashcmp(sha1, p->bad_object_sha1 + 20 * i))
1061 return 1;
1062 return 0;
1063}
1064
5d6ccf5c 1065int check_sha1_signature(const unsigned char *sha1, void *map, unsigned long size, const char *type)
0fcfd160
LT
1066{
1067 unsigned char real_sha1[20];
7cfb5f36 1068 hash_sha1_file(map, size, type, real_sha1);
a89fccd2 1069 return hashcmp(sha1, real_sha1) ? -1 : 0;
0fcfd160
LT
1070}
1071
44d1c19e
LT
1072static int git_open_noatime(const char *name)
1073{
1074 static int sha1_file_open_flag = O_NOATIME;
1075 int fd = open(name, O_RDONLY | sha1_file_open_flag);
1076
1077 /* Might the failure be due to O_NOATIME? */
1078 if (fd < 0 && errno != ENOENT && sha1_file_open_flag) {
1079 fd = open(name, O_RDONLY);
1080 if (fd >= 0)
1081 sha1_file_open_flag = 0;
1082 }
1083 return fd;
1084}
1085
1086static int open_sha1_file(const unsigned char *sha1)
1087{
1088 int fd;
1089 char *name = sha1_file_name(sha1);
1090 struct alternate_object_database *alt;
1091
1092 fd = git_open_noatime(name);
1093 if (fd >= 0)
1094 return fd;
1095
1096 prepare_alt_odb();
1097 errno = ENOENT;
1098 for (alt = alt_odb_list; alt; alt = alt->next) {
1099 name = alt->name;
1100 fill_sha1_path(name, sha1);
1101 fd = git_open_noatime(alt->base);
1102 if (fd >= 0)
1103 return fd;
1104 }
1105 return -1;
1106}
1107
4175e9e3 1108static void *map_sha1_file(const unsigned char *sha1, unsigned long *size)
0fcfd160 1109{
0fcfd160 1110 void *map;
144bde78 1111 int fd;
ace1534d 1112
44d1c19e
LT
1113 fd = open_sha1_file(sha1);
1114 map = NULL;
1115 if (fd >= 0) {
1116 struct stat st;
0fcfd160 1117
44d1c19e
LT
1118 if (!fstat(fd, &st)) {
1119 *size = xsize_t(st.st_size);
1120 map = xmmap(NULL, *size, PROT_READ, MAP_PRIVATE, fd, 0);
144bde78 1121 }
44d1c19e 1122 close(fd);
0fcfd160 1123 }
0fcfd160
LT
1124 return map;
1125}
1126
726f852b 1127static int legacy_loose_object(unsigned char *map)
bb6b8e4f
JH
1128{
1129 unsigned int word;
1130
1131 /*
1132 * Is it a zlib-compressed buffer? If so, the first byte
1133 * must be 0x78 (15-bit window size, deflated), and the
1134 * first 16-bit word is evenly divisible by 31
1135 */
1136 word = (map[0] << 8) + map[1];
1137 if (map[0] == 0x78 && !(word % 31))
1138 return 1;
1139 else
1140 return 0;
1141}
1142
72518e9c 1143unsigned long unpack_object_header_gently(const unsigned char *buf, unsigned long len, enum object_type *type, unsigned long *sizep)
c4483576 1144{
ad1ed5ee 1145 unsigned shift;
93821bd9 1146 unsigned char c;
93821bd9 1147 unsigned long size;
ad1ed5ee
JH
1148 unsigned long used = 0;
1149
1150 c = buf[used++];
1151 *type = (c >> 4) & 7;
1152 size = c & 15;
1153 shift = 4;
1154 while (c & 0x80) {
1155 if (len <= used)
1156 return 0;
1157 if (sizeof(long) * 8 <= shift)
1158 return 0;
1159 c = buf[used++];
1160 size += (c & 0x7f) << shift;
1161 shift += 7;
1162 }
1163 *sizep = size;
1164 return used;
1165}
1166
1167static int unpack_sha1_header(z_stream *stream, unsigned char *map, unsigned long mapsize, void *buffer, unsigned long bufsiz)
1168{
1169 unsigned long size, used;
1170 static const char valid_loose_object_type[8] = {
1171 0, /* OBJ_EXT */
1172 1, 1, 1, 1, /* "commit", "tree", "blob", "tag" */
1173 0, /* "delta" and others are invalid in a loose object */
93821bd9 1174 };
ad1ed5ee 1175 enum object_type type;
93821bd9 1176
c4483576
LT
1177 /* Get the data stream */
1178 memset(stream, 0, sizeof(*stream));
1179 stream->next_in = map;
1180 stream->avail_in = mapsize;
1181 stream->next_out = buffer;
93821bd9
LT
1182 stream->avail_out = bufsiz;
1183
bb6b8e4f 1184 if (legacy_loose_object(map)) {
93821bd9
LT
1185 inflateInit(stream);
1186 return inflate(stream, 0);
1187 }
1188
726f852b
NP
1189
1190 /*
1191 * There used to be a second loose object header format which
1192 * was meant to mimic the in-pack format, allowing for direct
1193 * copy of the object data. This format turned up not to be
1194 * really worth it and we don't write it any longer. But we
1195 * can still read it.
1196 */
ad1ed5ee
JH
1197 used = unpack_object_header_gently(map, mapsize, &type, &size);
1198 if (!used || !valid_loose_object_type[type])
93821bd9 1199 return -1;
ad1ed5ee
JH
1200 map += used;
1201 mapsize -= used;
93821bd9
LT
1202
1203 /* Set up the stream for the rest.. */
1204 stream->next_in = map;
1205 stream->avail_in = mapsize;
c4483576 1206 inflateInit(stream);
93821bd9
LT
1207
1208 /* And generate the fake traditional header */
ad1ed5ee 1209 stream->total_out = 1 + snprintf(buffer, bufsiz, "%s %lu",
df843662 1210 typename(type), size);
93821bd9 1211 return 0;
c4483576
LT
1212}
1213
7efbff75 1214static void *unpack_sha1_rest(z_stream *stream, void *buffer, unsigned long size, const unsigned char *sha1)
5180cacc
LT
1215{
1216 int bytes = strlen(buffer) + 1;
d565b341 1217 unsigned char *buf = xmalloc(1+size);
93821bd9 1218 unsigned long n;
7efbff75 1219 int status = Z_OK;
5180cacc 1220
93821bd9
LT
1221 n = stream->total_out - bytes;
1222 if (n > size)
1223 n = size;
1224 memcpy(buf, (char *) buffer + bytes, n);
1225 bytes = n;
456cdf6e
LT
1226 if (bytes <= size) {
1227 /*
1228 * The above condition must be (bytes <= size), not
1229 * (bytes < size). In other words, even though we
1230 * expect no more output and set avail_out to zer0,
1231 * the input zlib stream may have bytes that express
1232 * "this concludes the stream", and we *do* want to
1233 * eat that input.
1234 *
1235 * Otherwise we would not be able to test that we
1236 * consumed all the input to reach the expected size;
1237 * we also want to check that zlib tells us that all
1238 * went well with status == Z_STREAM_END at the end.
1239 */
5180cacc
LT
1240 stream->next_out = buf + bytes;
1241 stream->avail_out = size - bytes;
7efbff75
JH
1242 while (status == Z_OK)
1243 status = inflate(stream, Z_FINISH);
5180cacc
LT
1244 }
1245 buf[size] = 0;
456cdf6e 1246 if (status == Z_STREAM_END && !stream->avail_in) {
7efbff75
JH
1247 inflateEnd(stream);
1248 return buf;
1249 }
1250
1251 if (status < 0)
1252 error("corrupt loose object '%s'", sha1_to_hex(sha1));
1253 else if (stream->avail_in)
1254 error("garbage at end of loose object '%s'",
1255 sha1_to_hex(sha1));
1256 free(buf);
1257 return NULL;
5180cacc
LT
1258}
1259
1260/*
1261 * We used to just use "sscanf()", but that's actually way
1262 * too permissive for what we want to check. So do an anal
1263 * object header parse by hand.
1264 */
21666f1a 1265static int parse_sha1_header(const char *hdr, unsigned long *sizep)
5180cacc 1266{
21666f1a 1267 char type[10];
5180cacc
LT
1268 int i;
1269 unsigned long size;
1270
1271 /*
a6080a0a 1272 * The type can be at most ten bytes (including the
5180cacc 1273 * terminating '\0' that we add), and is followed by
21666f1a 1274 * a space.
5180cacc 1275 */
21666f1a 1276 i = 0;
5180cacc
LT
1277 for (;;) {
1278 char c = *hdr++;
1279 if (c == ' ')
1280 break;
21666f1a
NP
1281 type[i++] = c;
1282 if (i >= sizeof(type))
5180cacc 1283 return -1;
5180cacc 1284 }
21666f1a 1285 type[i] = 0;
5180cacc
LT
1286
1287 /*
1288 * The length must follow immediately, and be in canonical
1289 * decimal format (ie "010" is not valid).
1290 */
1291 size = *hdr++ - '0';
1292 if (size > 9)
1293 return -1;
1294 if (size) {
1295 for (;;) {
1296 unsigned long c = *hdr - '0';
1297 if (c > 9)
1298 break;
1299 hdr++;
1300 size = size * 10 + c;
1301 }
1302 }
1303 *sizep = size;
1304
1305 /*
1306 * The length must be followed by a zero byte
1307 */
21666f1a 1308 return *hdr ? -1 : type_from_string(type);
5180cacc
LT
1309}
1310
7efbff75 1311static void *unpack_sha1_file(void *map, unsigned long mapsize, enum object_type *type, unsigned long *size, const unsigned char *sha1)
0fcfd160 1312{
5180cacc 1313 int ret;
0fcfd160 1314 z_stream stream;
5180cacc 1315 char hdr[8192];
0fcfd160 1316
5180cacc 1317 ret = unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr));
21666f1a 1318 if (ret < Z_OK || (*type = parse_sha1_header(hdr, size)) < 0)
0fcfd160
LT
1319 return NULL;
1320
7efbff75 1321 return unpack_sha1_rest(&stream, hdr, *size, sha1);
0fcfd160
LT
1322}
1323
54dab52a
NP
1324unsigned long get_size_from_delta(struct packed_git *p,
1325 struct pack_window **w_curs,
1326 off_t curpos)
1327{
1328 const unsigned char *data;
1329 unsigned char delta_head[20], *in;
1330 z_stream stream;
1331 int st;
1332
1333 memset(&stream, 0, sizeof(stream));
1334 stream.next_out = delta_head;
1335 stream.avail_out = sizeof(delta_head);
1336
1337 inflateInit(&stream);
1338 do {
1339 in = use_pack(p, w_curs, curpos, &stream.avail_in);
1340 stream.next_in = in;
1341 st = inflate(&stream, Z_FINISH);
1342 curpos += stream.next_in - in;
1343 } while ((st == Z_OK || st == Z_BUF_ERROR) &&
1344 stream.total_out < sizeof(delta_head));
1345 inflateEnd(&stream);
1346 if ((st != Z_STREAM_END) && stream.total_out != sizeof(delta_head))
1347 die("delta data unpack-initial failed");
1348
1349 /* Examine the initial part of the delta to figure out
1350 * the result size.
1351 */
1352 data = delta_head;
1353
1354 /* ignore base size */
1355 get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
1356
1357 /* Read the result size */
1358 return get_delta_hdr_size(&data, delta_head+sizeof(delta_head));
1359}
1360
c4001d92 1361static off_t get_delta_base(struct packed_git *p,
03e79c88 1362 struct pack_window **w_curs,
c4001d92 1363 off_t *curpos,
21666f1a 1364 enum object_type type,
c4001d92 1365 off_t delta_obj_offset)
eb32d236 1366{
2b87c45b 1367 unsigned char *base_info = use_pack(p, w_curs, *curpos, NULL);
c4001d92 1368 off_t base_offset;
eb32d236 1369
8d8a4ea5
SP
1370 /* use_pack() assured us we have [base_info, base_info + 20)
1371 * as a range that we can look at without walking off the
1372 * end of the mapped window. Its actually the hash size
1373 * that is assured. An OFS_DELTA longer than the hash size
1374 * is stupid, as then a REF_DELTA would be smaller to store.
1375 */
21666f1a 1376 if (type == OBJ_OFS_DELTA) {
eb32d236
NP
1377 unsigned used = 0;
1378 unsigned char c = base_info[used++];
1379 base_offset = c & 127;
1380 while (c & 128) {
1381 base_offset += 1;
8723f216 1382 if (!base_offset || MSB(base_offset, 7))
8eca0b47 1383 return 0; /* overflow */
eb32d236
NP
1384 c = base_info[used++];
1385 base_offset = (base_offset << 7) + (c & 127);
1386 }
1387 base_offset = delta_obj_offset - base_offset;
1388 if (base_offset >= delta_obj_offset)
8eca0b47 1389 return 0; /* out of bound */
2b87c45b 1390 *curpos += used;
21666f1a 1391 } else if (type == OBJ_REF_DELTA) {
eb32d236
NP
1392 /* The base entry _must_ be in the same pack */
1393 base_offset = find_pack_entry_one(base_info, p);
2b87c45b 1394 *curpos += 20;
eb32d236
NP
1395 } else
1396 die("I am totally screwed");
2b87c45b 1397 return base_offset;
eb32d236
NP
1398}
1399
f3bf9224 1400/* forward declaration for a mutually recursive function */
c4001d92 1401static int packed_object_info(struct packed_git *p, off_t offset,
21666f1a 1402 unsigned long *sizep);
f3bf9224 1403
43057304 1404static int packed_delta_info(struct packed_git *p,
03e79c88 1405 struct pack_window **w_curs,
c4001d92 1406 off_t curpos,
21666f1a 1407 enum object_type type,
c4001d92 1408 off_t obj_offset,
43057304 1409 unsigned long *sizep)
5db47c2b 1410{
c4001d92 1411 off_t base_offset;
f3bf9224 1412
21666f1a
NP
1413 base_offset = get_delta_base(p, w_curs, &curpos, type, obj_offset);
1414 type = packed_object_info(p, base_offset, NULL);
f3bf9224 1415
c62266f3
JH
1416 /* We choose to only get the type of the base object and
1417 * ignore potentially corrupt pack file that expects the delta
1418 * based on a base with a wrong size. This saves tons of
1419 * inflate() calls.
1420 */
54dab52a
NP
1421 if (sizep)
1422 *sizep = get_size_from_delta(p, w_curs, curpos);
21666f1a
NP
1423
1424 return type;
5db47c2b
JH
1425}
1426
2b87c45b
NP
1427static int unpack_object_header(struct packed_git *p,
1428 struct pack_window **w_curs,
c4001d92 1429 off_t *curpos,
2b87c45b 1430 unsigned long *sizep)
a733cb60 1431{
03e79c88
SP
1432 unsigned char *base;
1433 unsigned int left;
ad1ed5ee 1434 unsigned long used;
2b87c45b 1435 enum object_type type;
a733cb60 1436
8d8a4ea5
SP
1437 /* use_pack() assures us we have [base, base + 20) available
1438 * as a range that we can look at at. (Its actually the hash
3dff5379 1439 * size that is assured.) With our object header encoding
8d8a4ea5
SP
1440 * the maximum deflated object size is 2^137, which is just
1441 * insane, so we know won't exceed what we have been given.
1442 */
2b87c45b
NP
1443 base = use_pack(p, w_curs, *curpos, &left);
1444 used = unpack_object_header_gently(base, left, &type, sizep);
ad1ed5ee
JH
1445 if (!used)
1446 die("object offset outside of pack file");
2b87c45b 1447 *curpos += used;
ad1ed5ee 1448
2b87c45b 1449 return type;
a733cb60
LT
1450}
1451
21666f1a 1452const char *packed_object_info_detail(struct packed_git *p,
c4001d92 1453 off_t obj_offset,
21666f1a
NP
1454 unsigned long *size,
1455 unsigned long *store_size,
1456 unsigned int *delta_chain_length,
1457 unsigned char *base_sha1)
ad8c80a5 1458{
03e79c88 1459 struct pack_window *w_curs = NULL;
c4001d92
SP
1460 off_t curpos;
1461 unsigned long dummy;
43057304 1462 unsigned char *next_sha1;
21666f1a 1463 enum object_type type;
70f5d5d3 1464 struct revindex_entry *revidx;
ad8c80a5 1465
43057304 1466 *delta_chain_length = 0;
2b87c45b 1467 curpos = obj_offset;
21666f1a 1468 type = unpack_object_header(p, &w_curs, &curpos, size);
43057304 1469
70f5d5d3
NP
1470 revidx = find_pack_revindex(p, obj_offset);
1471 *store_size = revidx[1].offset - obj_offset;
1472
43057304 1473 for (;;) {
21666f1a 1474 switch (type) {
43057304 1475 default:
08a19d87 1476 die("pack %s contains unknown object type %d",
21666f1a 1477 p->pack_name, type);
43057304
NP
1478 case OBJ_COMMIT:
1479 case OBJ_TREE:
1480 case OBJ_BLOB:
1481 case OBJ_TAG:
03e79c88 1482 unuse_pack(&w_curs);
21666f1a 1483 return typename(type);
eb32d236 1484 case OBJ_OFS_DELTA:
21666f1a 1485 obj_offset = get_delta_base(p, &w_curs, &curpos, type, obj_offset);
8eca0b47
NP
1486 if (!obj_offset)
1487 die("pack %s contains bad delta base reference of type %s",
1488 p->pack_name, typename(type));
eb32d236 1489 if (*delta_chain_length == 0) {
70f5d5d3
NP
1490 revidx = find_pack_revindex(p, obj_offset);
1491 hashcpy(base_sha1, nth_packed_object_sha1(p, revidx->nr));
eb32d236
NP
1492 }
1493 break;
1494 case OBJ_REF_DELTA:
2b87c45b 1495 next_sha1 = use_pack(p, &w_curs, curpos, NULL);
43057304
NP
1496 if (*delta_chain_length == 0)
1497 hashcpy(base_sha1, next_sha1);
2b87c45b 1498 obj_offset = find_pack_entry_one(next_sha1, p);
43057304
NP
1499 break;
1500 }
43057304 1501 (*delta_chain_length)++;
2b87c45b 1502 curpos = obj_offset;
21666f1a 1503 type = unpack_object_header(p, &w_curs, &curpos, &dummy);
ad8c80a5 1504 }
ad8c80a5
JH
1505}
1506
c4001d92 1507static int packed_object_info(struct packed_git *p, off_t obj_offset,
21666f1a 1508 unsigned long *sizep)
1f688557 1509{
03e79c88 1510 struct pack_window *w_curs = NULL;
c4001d92
SP
1511 unsigned long size;
1512 off_t curpos = obj_offset;
21666f1a 1513 enum object_type type;
5db47c2b 1514
21666f1a 1515 type = unpack_object_header(p, &w_curs, &curpos, &size);
5db47c2b 1516
21666f1a 1517 switch (type) {
eb32d236
NP
1518 case OBJ_OFS_DELTA:
1519 case OBJ_REF_DELTA:
21666f1a
NP
1520 type = packed_delta_info(p, &w_curs, curpos,
1521 type, obj_offset, sizep);
2b87c45b 1522 break;
a733cb60 1523 case OBJ_COMMIT:
a733cb60 1524 case OBJ_TREE:
a733cb60 1525 case OBJ_BLOB:
a733cb60 1526 case OBJ_TAG:
2b87c45b
NP
1527 if (sizep)
1528 *sizep = size;
a69d0943 1529 break;
1f688557 1530 default:
08a19d87 1531 die("pack %s contains unknown object type %d",
21666f1a 1532 p->pack_name, type);
1f688557 1533 }
2b87c45b 1534 unuse_pack(&w_curs);
21666f1a 1535 return type;
1f688557
JH
1536}
1537
eb950c19 1538static void *unpack_compressed_entry(struct packed_git *p,
03e79c88 1539 struct pack_window **w_curs,
c4001d92 1540 off_t curpos,
eb950c19 1541 unsigned long size)
de530aaa
SP
1542{
1543 int st;
1544 z_stream stream;
079afb18 1545 unsigned char *buffer, *in;
de530aaa
SP
1546
1547 buffer = xmalloc(size + 1);
1548 buffer[size] = 0;
1549 memset(&stream, 0, sizeof(stream));
de530aaa
SP
1550 stream.next_out = buffer;
1551 stream.avail_out = size;
1552
1553 inflateInit(&stream);
079afb18 1554 do {
2b87c45b 1555 in = use_pack(p, w_curs, curpos, &stream.avail_in);
079afb18
SP
1556 stream.next_in = in;
1557 st = inflate(&stream, Z_FINISH);
2b87c45b 1558 curpos += stream.next_in - in;
079afb18 1559 } while (st == Z_OK || st == Z_BUF_ERROR);
de530aaa
SP
1560 inflateEnd(&stream);
1561 if ((st != Z_STREAM_END) || stream.total_out != size) {
1562 free(buffer);
1563 return NULL;
1564 }
1565
1566 return buffer;
1567}
1568
e5e01619
LT
1569#define MAX_DELTA_CACHE (256)
1570
18bdec11 1571static size_t delta_base_cached;
5e08ecbf
NP
1572
1573static struct delta_base_cache_lru_list {
1574 struct delta_base_cache_lru_list *prev;
1575 struct delta_base_cache_lru_list *next;
1576} delta_base_cache_lru = { &delta_base_cache_lru, &delta_base_cache_lru };
1577
e5e01619 1578static struct delta_base_cache_entry {
5e08ecbf
NP
1579 struct delta_base_cache_lru_list lru;
1580 void *data;
e5e01619
LT
1581 struct packed_git *p;
1582 off_t base_offset;
1583 unsigned long size;
e5e01619
LT
1584 enum object_type type;
1585} delta_base_cache[MAX_DELTA_CACHE];
1586
1587static unsigned long pack_entry_hash(struct packed_git *p, off_t base_offset)
1588{
1589 unsigned long hash;
1590
1591 hash = (unsigned long)p + (unsigned long)base_offset;
1592 hash += (hash >> 8) + (hash >> 16);
3358004a 1593 return hash % MAX_DELTA_CACHE;
e5e01619
LT
1594}
1595
62f255ad 1596static void *cache_or_unpack_entry(struct packed_git *p, off_t base_offset,
a0cba108 1597 unsigned long *base_size, enum object_type *type, int keep_cache)
62f255ad 1598{
e5e01619
LT
1599 void *ret;
1600 unsigned long hash = pack_entry_hash(p, base_offset);
1601 struct delta_base_cache_entry *ent = delta_base_cache + hash;
1602
1603 ret = ent->data;
1604 if (ret && ent->p == p && ent->base_offset == base_offset)
1605 goto found_cache_entry;
62f255ad 1606 return unpack_entry(p, base_offset, type, base_size);
e5e01619
LT
1607
1608found_cache_entry:
18bdec11 1609 if (!keep_cache) {
a0cba108 1610 ent->data = NULL;
5e08ecbf
NP
1611 ent->lru.next->prev = ent->lru.prev;
1612 ent->lru.prev->next = ent->lru.next;
18bdec11 1613 delta_base_cached -= ent->size;
182af834
PH
1614 } else {
1615 ret = xmemdupz(ent->data, ent->size);
a0cba108 1616 }
e5e01619
LT
1617 *type = ent->type;
1618 *base_size = ent->size;
1619 return ret;
62f255ad
LT
1620}
1621
18bdec11
SP
1622static inline void release_delta_base_cache(struct delta_base_cache_entry *ent)
1623{
1624 if (ent->data) {
1625 free(ent->data);
1626 ent->data = NULL;
5e08ecbf
NP
1627 ent->lru.next->prev = ent->lru.prev;
1628 ent->lru.prev->next = ent->lru.next;
18bdec11
SP
1629 delta_base_cached -= ent->size;
1630 }
1631}
1632
62f255ad
LT
1633static void add_delta_base_cache(struct packed_git *p, off_t base_offset,
1634 void *base, unsigned long base_size, enum object_type type)
1635{
5e08ecbf 1636 unsigned long hash = pack_entry_hash(p, base_offset);
e5e01619 1637 struct delta_base_cache_entry *ent = delta_base_cache + hash;
5e08ecbf 1638 struct delta_base_cache_lru_list *lru;
e5e01619 1639
18bdec11
SP
1640 release_delta_base_cache(ent);
1641 delta_base_cached += base_size;
5e08ecbf
NP
1642
1643 for (lru = delta_base_cache_lru.next;
1644 delta_base_cached > delta_base_cache_limit
1645 && lru != &delta_base_cache_lru;
1646 lru = lru->next) {
1647 struct delta_base_cache_entry *f = (void *)lru;
18bdec11
SP
1648 if (f->type == OBJ_BLOB)
1649 release_delta_base_cache(f);
1650 }
5e08ecbf
NP
1651 for (lru = delta_base_cache_lru.next;
1652 delta_base_cached > delta_base_cache_limit
1653 && lru != &delta_base_cache_lru;
1654 lru = lru->next) {
1655 struct delta_base_cache_entry *f = (void *)lru;
1656 release_delta_base_cache(f);
1657 }
18bdec11 1658
e5e01619
LT
1659 ent->p = p;
1660 ent->base_offset = base_offset;
1661 ent->type = type;
1662 ent->data = base;
1663 ent->size = base_size;
5e08ecbf
NP
1664 ent->lru.next = &delta_base_cache_lru;
1665 ent->lru.prev = delta_base_cache_lru.prev;
1666 delta_base_cache_lru.prev->next = &ent->lru;
1667 delta_base_cache_lru.prev = &ent->lru;
62f255ad
LT
1668}
1669
eb950c19 1670static void *unpack_delta_entry(struct packed_git *p,
03e79c88 1671 struct pack_window **w_curs,
c4001d92 1672 off_t curpos,
1f688557 1673 unsigned long delta_size,
c4001d92 1674 off_t obj_offset,
21666f1a 1675 enum object_type *type,
eb950c19 1676 unsigned long *sizep)
1f688557 1677{
7c3e8be3 1678 void *delta_data, *result, *base;
c4001d92
SP
1679 unsigned long base_size;
1680 off_t base_offset;
43057304 1681
21666f1a 1682 base_offset = get_delta_base(p, w_curs, &curpos, *type, obj_offset);
8eca0b47
NP
1683 if (!base_offset) {
1684 error("failed to validate delta base reference "
1685 "at offset %"PRIuMAX" from %s",
1686 (uintmax_t)curpos, p->pack_name);
1687 return NULL;
1688 }
eac12e2d 1689 unuse_pack(w_curs);
a0cba108 1690 base = cache_or_unpack_entry(p, base_offset, &base_size, type, 0);
8eca0b47
NP
1691 if (!base) {
1692 /*
1693 * We're probably in deep shit, but let's try to fetch
1694 * the required base anyway from another pack or loose.
1695 * This is costly but should happen only in the presence
1696 * of a corrupted pack, and is better than failing outright.
1697 */
1698 struct revindex_entry *revidx = find_pack_revindex(p, base_offset);
1699 const unsigned char *base_sha1 =
1700 nth_packed_object_sha1(p, revidx->nr);
1701 error("failed to read delta base object %s"
1702 " at offset %"PRIuMAX" from %s",
1703 sha1_to_hex(base_sha1), (uintmax_t)base_offset,
1704 p->pack_name);
1705 mark_bad_packed_object(p, base_sha1);
ac939109 1706 base = read_object(base_sha1, type, &base_size);
8eca0b47
NP
1707 if (!base)
1708 return NULL;
1709 }
67686d95 1710
2b87c45b 1711 delta_data = unpack_compressed_entry(p, w_curs, curpos, delta_size);
8eca0b47
NP
1712 if (!delta_data) {
1713 error("failed to unpack compressed delta "
1714 "at offset %"PRIuMAX" from %s",
1715 (uintmax_t)curpos, p->pack_name);
1716 free(base);
1717 return NULL;
1718 }
1f688557
JH
1719 result = patch_delta(base, base_size,
1720 delta_data, delta_size,
21666f1a 1721 sizep);
1f688557
JH
1722 if (!result)
1723 die("failed to apply delta");
1724 free(delta_data);
62f255ad 1725 add_delta_base_cache(p, base_offset, base, base_size, *type);
1f688557
JH
1726 return result;
1727}
1728
c4001d92 1729void *unpack_entry(struct packed_git *p, off_t obj_offset,
21666f1a 1730 enum object_type *type, unsigned long *sizep)
1f688557 1731{
03e79c88 1732 struct pack_window *w_curs = NULL;
c4001d92 1733 off_t curpos = obj_offset;
21666f1a 1734 void *data;
1f688557 1735
21666f1a
NP
1736 *type = unpack_object_header(p, &w_curs, &curpos, sizep);
1737 switch (*type) {
eb32d236
NP
1738 case OBJ_OFS_DELTA:
1739 case OBJ_REF_DELTA:
21666f1a
NP
1740 data = unpack_delta_entry(p, &w_curs, curpos, *sizep,
1741 obj_offset, type, sizep);
4d703a1a 1742 break;
a733cb60 1743 case OBJ_COMMIT:
a733cb60 1744 case OBJ_TREE:
a733cb60 1745 case OBJ_BLOB:
a733cb60 1746 case OBJ_TAG:
21666f1a 1747 data = unpack_compressed_entry(p, &w_curs, curpos, *sizep);
4d703a1a 1748 break;
1f688557 1749 default:
8eca0b47
NP
1750 data = NULL;
1751 error("unknown object type %i at offset %"PRIuMAX" in %s",
1752 *type, (uintmax_t)obj_offset, p->pack_name);
1f688557 1753 }
03e79c88 1754 unuse_pack(&w_curs);
21666f1a 1755 return data;
1f688557
JH
1756}
1757
d079837e 1758const unsigned char *nth_packed_object_sha1(struct packed_git *p,
d72308e0 1759 uint32_t n)
9a217f2a 1760{
42873078 1761 const unsigned char *index = p->index_data;
d079837e
SP
1762 if (!index) {
1763 if (open_pack_index(p))
1764 return NULL;
1765 index = p->index_data;
1766 }
57059091 1767 if (n >= p->num_objects)
d72308e0 1768 return NULL;
74e34e1f
NP
1769 index += 4 * 256;
1770 if (p->index_version == 1) {
1771 return index + 24 * n + 4;
1772 } else {
1773 index += 8;
1774 return index + 20 * n;
1775 }
1776}
1777
99093238 1778off_t nth_packed_object_offset(const struct packed_git *p, uint32_t n)
74e34e1f
NP
1779{
1780 const unsigned char *index = p->index_data;
1781 index += 4 * 256;
1782 if (p->index_version == 1) {
1783 return ntohl(*((uint32_t *)(index + 24 * n)));
1784 } else {
1785 uint32_t off;
1786 index += 8 + p->num_objects * (20 + 4);
1787 off = ntohl(*((uint32_t *)(index + 4 * n)));
1788 if (!(off & 0x80000000))
1789 return off;
1790 index += p->num_objects * 4 + (off & 0x7fffffff) * 8;
1791 return (((uint64_t)ntohl(*((uint32_t *)(index + 0)))) << 32) |
1792 ntohl(*((uint32_t *)(index + 4)));
1793 }
9a217f2a
JH
1794}
1795
c4001d92 1796off_t find_pack_entry_one(const unsigned char *sha1,
43057304 1797 struct packed_git *p)
1f688557 1798{
42873078 1799 const uint32_t *level1_ofs = p->index_data;
42873078 1800 const unsigned char *index = p->index_data;
628522ec
JH
1801 unsigned hi, lo, stride;
1802 static int use_lookup = -1;
1803 static int debug_lookup = -1;
1804
1805 if (debug_lookup < 0)
1806 debug_lookup = !!getenv("GIT_DEBUG_LOOKUP");
42873078 1807
d079837e
SP
1808 if (!index) {
1809 if (open_pack_index(p))
1810 return 0;
1811 level1_ofs = p->index_data;
1812 index = p->index_data;
1813 }
74e34e1f
NP
1814 if (p->index_version > 1) {
1815 level1_ofs += 2;
1816 index += 8;
1817 }
42873078 1818 index += 4 * 256;
74e34e1f
NP
1819 hi = ntohl(level1_ofs[*sha1]);
1820 lo = ((*sha1 == 0x0) ? 0 : ntohl(level1_ofs[*sha1 - 1]));
628522ec
JH
1821 if (p->index_version > 1) {
1822 stride = 20;
1823 } else {
1824 stride = 24;
1825 index += 4;
1826 }
1827
1828 if (debug_lookup)
6e1c2344 1829 printf("%02x%02x%02x... lo %u hi %u nr %"PRIu32"\n",
628522ec
JH
1830 sha1[0], sha1[1], sha1[2], lo, hi, p->num_objects);
1831
1832 if (use_lookup < 0)
1833 use_lookup = !!getenv("GIT_USE_LOOKUP");
1834 if (use_lookup) {
1835 int pos = sha1_entry_pos(index, stride, 0,
1836 lo, hi, p->num_objects, sha1);
1837 if (pos < 0)
1838 return 0;
1839 return nth_packed_object_offset(p, pos);
1840 }
1f688557
JH
1841
1842 do {
74e34e1f 1843 unsigned mi = (lo + hi) / 2;
628522ec
JH
1844 int cmp = hashcmp(index + mi * stride, sha1);
1845
1846 if (debug_lookup)
1847 printf("lo %u hi %u rg %u mi %u\n",
1848 lo, hi, hi - lo, mi);
43057304 1849 if (!cmp)
74e34e1f 1850 return nth_packed_object_offset(p, mi);
1f688557
JH
1851 if (cmp > 0)
1852 hi = mi;
1853 else
1854 lo = mi+1;
1855 } while (lo < hi);
1856 return 0;
1857}
1858
000dfd3f 1859int matches_pack_name(struct packed_git *p, const char *name)
d4ff6d92
SP
1860{
1861 const char *last_c, *c;
1862
000dfd3f
JH
1863 if (!strcmp(p->pack_name, name))
1864 return 1;
d4ff6d92
SP
1865
1866 for (c = p->pack_name, last_c = c; *c;)
1867 if (*c == '/')
1868 last_c = ++c;
1869 else
1870 ++c;
000dfd3f
JH
1871 if (!strcmp(last_c, name))
1872 return 1;
d4ff6d92 1873
000dfd3f 1874 return 0;
d4ff6d92
SP
1875}
1876
106d710b 1877static int find_pack_entry(const unsigned char *sha1, struct pack_entry *e, const char **ignore_packed)
1f688557 1878{
f7c22cc6 1879 static struct packed_git *last_found = (void *)1;
1f688557 1880 struct packed_git *p;
c4001d92 1881 off_t offset;
43057304 1882
1f688557 1883 prepare_packed_git();
f7c22cc6
NP
1884 if (!packed_git)
1885 return 0;
1886 p = (last_found == (void *)1) ? packed_git : last_found;
1f688557 1887
f7c22cc6 1888 do {
106d710b
JH
1889 if (ignore_packed) {
1890 const char **ig;
1891 for (ig = ignore_packed; *ig; ig++)
000dfd3f 1892 if (matches_pack_name(p, *ig))
106d710b
JH
1893 break;
1894 if (*ig)
f7c22cc6 1895 goto next;
106d710b 1896 }
f7c22cc6 1897
8eca0b47
NP
1898 if (p->num_bad_objects) {
1899 unsigned i;
1900 for (i = 0; i < p->num_bad_objects; i++)
1901 if (!hashcmp(sha1, p->bad_object_sha1 + 20 * i))
1902 goto next;
1903 }
1904
43057304
NP
1905 offset = find_pack_entry_one(sha1, p);
1906 if (offset) {
c715f783
SP
1907 /*
1908 * We are about to tell the caller where they can
1909 * locate the requested object. We better make
1910 * sure the packfile is still here and can be
1911 * accessed before supplying that answer, as
1912 * it may have been deleted since the index
1913 * was loaded!
1914 */
1915 if (p->pack_fd == -1 && open_packed_git(p)) {
1916 error("packfile %s cannot be accessed", p->pack_name);
f7c22cc6 1917 goto next;
c715f783 1918 }
43057304
NP
1919 e->offset = offset;
1920 e->p = p;
1921 hashcpy(e->sha1, sha1);
f7c22cc6 1922 last_found = p;
1f688557 1923 return 1;
43057304 1924 }
f7c22cc6
NP
1925
1926 next:
1927 if (p == last_found)
1928 p = packed_git;
1929 else
1930 p = p->next;
1931 if (p == last_found)
1932 p = p->next;
1933 } while (p);
1f688557
JH
1934 return 0;
1935}
1936
a6080a0a 1937struct packed_git *find_sha1_pack(const unsigned char *sha1,
bf592c50
DB
1938 struct packed_git *packs)
1939{
1940 struct packed_git *p;
bf592c50
DB
1941
1942 for (p = packs; p; p = p->next) {
43057304 1943 if (find_pack_entry_one(sha1, p))
bf592c50
DB
1944 return p;
1945 }
1946 return NULL;
21666f1a 1947
bf592c50
DB
1948}
1949
21666f1a 1950static int sha1_loose_object_info(const unsigned char *sha1, unsigned long *sizep)
65c2e0c3 1951{
36e4d74a 1952 int status;
65c2e0c3
JH
1953 unsigned long mapsize, size;
1954 void *map;
1955 z_stream stream;
d65a16f6 1956 char hdr[32];
65c2e0c3 1957
bb6b8e4f 1958 map = map_sha1_file(sha1, &mapsize);
f0df4ed5
JS
1959 if (!map)
1960 return error("unable to find %s", sha1_to_hex(sha1));
36e4d74a
JH
1961 if (unpack_sha1_header(&stream, map, mapsize, hdr, sizeof(hdr)) < 0)
1962 status = error("unable to unpack %s header",
1963 sha1_to_hex(sha1));
21666f1a 1964 else if ((status = parse_sha1_header(hdr, &size)) < 0)
36e4d74a 1965 status = error("unable to parse %s header", sha1_to_hex(sha1));
21666f1a
NP
1966 else if (sizep)
1967 *sizep = size;
65c2e0c3
JH
1968 inflateEnd(&stream);
1969 munmap(map, mapsize);
1970 return status;
1971}
1972
21666f1a 1973int sha1_object_info(const unsigned char *sha1, unsigned long *sizep)
f0df4ed5 1974{
f0df4ed5 1975 struct pack_entry e;
ddd63e64 1976 int status;
f0df4ed5
JS
1977
1978 if (!find_pack_entry(sha1, &e, NULL)) {
ddd63e64
SG
1979 /* Most likely it's a loose object. */
1980 status = sha1_loose_object_info(sha1, sizep);
1981 if (status >= 0)
1982 return status;
1983
1984 /* Not a loose object; someone else may have just packed it. */
f0df4ed5
JS
1985 reprepare_packed_git();
1986 if (!find_pack_entry(sha1, &e, NULL))
ddd63e64 1987 return status;
f0df4ed5 1988 }
21666f1a 1989 return packed_object_info(e.p, e.offset, sizep);
f0df4ed5
JS
1990}
1991
21666f1a
NP
1992static void *read_packed_sha1(const unsigned char *sha1,
1993 enum object_type *type, unsigned long *size)
1f688557
JH
1994{
1995 struct pack_entry e;
8eca0b47 1996 void *data;
1f688557 1997
8276c007 1998 if (!find_pack_entry(sha1, &e, NULL))
1f688557 1999 return NULL;
8eca0b47
NP
2000 data = cache_or_unpack_entry(e.p, e.offset, size, type, 1);
2001 if (!data) {
2002 /*
2003 * We're probably in deep shit, but let's try to fetch
2004 * the required object anyway from another pack or loose.
2005 * This should happen only in the presence of a corrupted
2006 * pack, and is better than failing outright.
2007 */
2008 error("failed to read object %s at offset %"PRIuMAX" from %s",
2009 sha1_to_hex(sha1), (uintmax_t)e.offset, e.p->pack_name);
2010 mark_bad_packed_object(e.p, sha1);
ac939109 2011 data = read_object(sha1, type, size);
8eca0b47
NP
2012 }
2013 return data;
1f688557
JH
2014}
2015
d66b37bb
JH
2016/*
2017 * This is meant to hold a *small* number of objects that you would
2018 * want read_sha1_file() to be able to return, but yet you do not want
2019 * to write them into the object store (e.g. a browse-only
2020 * application).
2021 */
2022static struct cached_object {
2023 unsigned char sha1[20];
21666f1a 2024 enum object_type type;
d66b37bb
JH
2025 void *buf;
2026 unsigned long size;
2027} *cached_objects;
2028static int cached_object_nr, cached_object_alloc;
2029
346245a1
JK
2030static struct cached_object empty_tree = {
2031 /* empty tree sha1: 4b825dc642cb6eb9a060e54bf8d69288fbee4904 */
2032 "\x4b\x82\x5d\xc6\x42\xcb\x6e\xb9\xa0\x60"
2033 "\xe5\x4b\xf8\xd6\x92\x88\xfb\xee\x49\x04",
2034 OBJ_TREE,
2035 "",
2036 0
2037};
2038
d66b37bb
JH
2039static struct cached_object *find_cached_object(const unsigned char *sha1)
2040{
2041 int i;
2042 struct cached_object *co = cached_objects;
2043
2044 for (i = 0; i < cached_object_nr; i++, co++) {
2045 if (!hashcmp(co->sha1, sha1))
2046 return co;
2047 }
346245a1
JK
2048 if (!hashcmp(sha1, empty_tree.sha1))
2049 return &empty_tree;
d66b37bb
JH
2050 return NULL;
2051}
2052
21666f1a
NP
2053int pretend_sha1_file(void *buf, unsigned long len, enum object_type type,
2054 unsigned char *sha1)
d66b37bb
JH
2055{
2056 struct cached_object *co;
2057
21666f1a 2058 hash_sha1_file(buf, len, typename(type), sha1);
d66b37bb
JH
2059 if (has_sha1_file(sha1) || find_cached_object(sha1))
2060 return 0;
2061 if (cached_object_alloc <= cached_object_nr) {
2062 cached_object_alloc = alloc_nr(cached_object_alloc);
2063 cached_objects = xrealloc(cached_objects,
2064 sizeof(*cached_objects) *
2065 cached_object_alloc);
2066 }
2067 co = &cached_objects[cached_object_nr++];
2068 co->size = len;
21666f1a 2069 co->type = type;
efa13f7b
JH
2070 co->buf = xmalloc(len);
2071 memcpy(co->buf, buf, len);
d66b37bb
JH
2072 hashcpy(co->sha1, sha1);
2073 return 0;
2074}
2075
ac939109
NP
2076void *read_object(const unsigned char *sha1, enum object_type *type,
2077 unsigned long *size)
0fcfd160
LT
2078{
2079 unsigned long mapsize;
2080 void *map, *buf;
d66b37bb
JH
2081 struct cached_object *co;
2082
2083 co = find_cached_object(sha1);
2084 if (co) {
21666f1a 2085 *type = co->type;
d66b37bb 2086 *size = co->size;
182af834 2087 return xmemdupz(co->buf, co->size);
d66b37bb 2088 }
0fcfd160 2089
8276c007
PE
2090 buf = read_packed_sha1(sha1, type, size);
2091 if (buf)
2092 return buf;
bb6b8e4f 2093 map = map_sha1_file(sha1, &mapsize);
0fcfd160 2094 if (map) {
7efbff75 2095 buf = unpack_sha1_file(map, mapsize, type, size, sha1);
0fcfd160
LT
2096 munmap(map, mapsize);
2097 return buf;
2098 }
637cdd9d 2099 reprepare_packed_git();
8276c007 2100 return read_packed_sha1(sha1, type, size);
0fcfd160
LT
2101}
2102
ac939109
NP
2103void *read_sha1_file(const unsigned char *sha1, enum object_type *type,
2104 unsigned long *size)
2105{
2106 void *data = read_object(sha1, type, size);
2107 /* legacy behavior is to die on corrupted objects */
2108 if (!data && (has_loose_object(sha1) || has_packed_and_bad(sha1)))
2109 die("object %s is corrupted", sha1_to_hex(sha1));
2110 return data;
2111}
2112
40469ee9 2113void *read_object_with_reference(const unsigned char *sha1,
21666f1a 2114 const char *required_type_name,
40469ee9
JH
2115 unsigned long *size,
2116 unsigned char *actual_sha1_return)
f4913f91 2117{
21666f1a 2118 enum object_type type, required_type;
f4913f91
JH
2119 void *buffer;
2120 unsigned long isize;
40469ee9 2121 unsigned char actual_sha1[20];
f4913f91 2122
21666f1a 2123 required_type = type_from_string(required_type_name);
e702496e 2124 hashcpy(actual_sha1, sha1);
40469ee9
JH
2125 while (1) {
2126 int ref_length = -1;
2127 const char *ref_type = NULL;
f4913f91 2128
21666f1a 2129 buffer = read_sha1_file(actual_sha1, &type, &isize);
40469ee9
JH
2130 if (!buffer)
2131 return NULL;
21666f1a 2132 if (type == required_type) {
40469ee9
JH
2133 *size = isize;
2134 if (actual_sha1_return)
e702496e 2135 hashcpy(actual_sha1_return, actual_sha1);
40469ee9
JH
2136 return buffer;
2137 }
2138 /* Handle references */
21666f1a 2139 else if (type == OBJ_COMMIT)
40469ee9 2140 ref_type = "tree ";
21666f1a 2141 else if (type == OBJ_TAG)
40469ee9
JH
2142 ref_type = "object ";
2143 else {
2144 free(buffer);
2145 return NULL;
2146 }
2147 ref_length = strlen(ref_type);
f4913f91 2148
50974ec9
MK
2149 if (ref_length + 40 > isize ||
2150 memcmp(buffer, ref_type, ref_length) ||
1d7f171c 2151 get_sha1_hex((char *) buffer + ref_length, actual_sha1)) {
40469ee9
JH
2152 free(buffer);
2153 return NULL;
2154 }
1cf58e72 2155 free(buffer);
40469ee9
JH
2156 /* Now we have the ID of the referred-to object in
2157 * actual_sha1. Check again. */
f4913f91 2158 }
f4913f91
JH
2159}
2160
ce9fbf16 2161static void write_sha1_file_prepare(const void *buf, unsigned long len,
972a9155 2162 const char *type, unsigned char *sha1,
d65a16f6 2163 char *hdr, int *hdrlen)
d410c0f5
JH
2164{
2165 SHA_CTX c;
2166
2167 /* Generate the header */
d65a16f6 2168 *hdrlen = sprintf(hdr, "%s %lu", type, len)+1;
d410c0f5
JH
2169
2170 /* Sha1.. */
2171 SHA1_Init(&c);
2172 SHA1_Update(&c, hdr, *hdrlen);
2173 SHA1_Update(&c, buf, len);
2174 SHA1_Final(sha1, &c);
d410c0f5
JH
2175}
2176
230f1322
LT
2177/*
2178 * Move the just written object into its final resting place
2179 */
839837b9 2180int move_temp_to_file(const char *tmpfile, const char *filename)
230f1322 2181{
e32c0a9c
TR
2182 int ret = 0;
2183 if (link(tmpfile, filename))
2184 ret = errno;
7ebb6fca
LT
2185
2186 /*
2187 * Coda hack - coda doesn't like cross-directory links,
2188 * so we fall back to a rename, which will mean that it
2189 * won't be able to check collisions, but that's not a
2190 * big deal.
2191 *
2192 * The same holds for FAT formatted media.
2193 *
2194 * When this succeeds, we just return 0. We have nothing
2195 * left to unlink.
2196 */
2197 if (ret && ret != EEXIST) {
2198 if (!rename(tmpfile, filename))
230f1322 2199 return 0;
9e48b389 2200 ret = errno;
230f1322
LT
2201 }
2202 unlink(tmpfile);
2203 if (ret) {
2204 if (ret != EEXIST) {
916d081b 2205 return error("unable to write sha1 filename %s: %s\n", filename, strerror(ret));
230f1322
LT
2206 }
2207 /* FIXME!!! Collision check here ? */
2208 }
2209
2210 return 0;
2211}
2212
4d548150
LT
2213static int write_buffer(int fd, const void *buf, size_t len)
2214{
d34cf19b 2215 if (write_in_full(fd, buf, len) < 0)
93822c22 2216 return error("file write error (%s)", strerror(errno));
4d548150
LT
2217 return 0;
2218}
2219
ce9fbf16 2220int hash_sha1_file(const void *buf, unsigned long len, const char *type,
abdc3fc8
RS
2221 unsigned char *sha1)
2222{
d65a16f6 2223 char hdr[32];
abdc3fc8
RS
2224 int hdrlen;
2225 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
2226 return 0;
2227}
2228
e9039dd3
LT
2229/* Finalize a file on disk, and close it. */
2230static void close_sha1_file(int fd)
2231{
aafe9fba
LT
2232 if (fsync_object_files)
2233 fsync_or_die(fd, "sha1 file");
e9039dd3
LT
2234 fchmod(fd, 0444);
2235 if (close(fd) != 0)
2236 die("unable to write sha1 file");
2237}
2238
5723fe7e
LT
2239/* Size of directory component, including the ending '/' */
2240static inline int directory_size(const char *filename)
2241{
2242 const char *s = strrchr(filename, '/');
2243 if (!s)
2244 return 0;
2245 return s - filename + 1;
2246}
2247
2248/*
2249 * This creates a temporary file in the same directory as the final
2250 * 'filename'
2251 *
2252 * We want to avoid cross-directory filename renames, because those
2253 * can have problems on various filesystems (FAT, NFS, Coda).
2254 */
2255static int create_tmpfile(char *buffer, size_t bufsiz, const char *filename)
2256{
2257 int fd, dirlen = directory_size(filename);
2258
2259 if (dirlen + 20 > bufsiz) {
2260 errno = ENAMETOOLONG;
2261 return -1;
2262 }
2263 memcpy(buffer, filename, dirlen);
2264 strcpy(buffer + dirlen, "tmp_obj_XXXXXX");
2265 fd = mkstemp(buffer);
65117abc 2266 if (fd < 0 && dirlen && errno == ENOENT) {
5723fe7e 2267 /* Make sure the directory exists */
6ff6af62 2268 memcpy(buffer, filename, dirlen);
5723fe7e 2269 buffer[dirlen-1] = 0;
3bfaf018 2270 if (mkdir(buffer, 0777) || adjust_shared_perm(buffer))
5723fe7e
LT
2271 return -1;
2272
2273 /* Try again */
2274 strcpy(buffer + dirlen - 1, "/tmp_obj_XXXXXX");
2275 fd = mkstemp(buffer);
2276 }
2277 return fd;
2278}
2279
bbac7311
NP
2280static int write_loose_object(const unsigned char *sha1, char *hdr, int hdrlen,
2281 void *buf, unsigned long len, time_t mtime)
0fcfd160 2282{
915308b1
JK
2283 int fd, ret;
2284 size_t size;
bf0f910d 2285 unsigned char *compressed;
0fcfd160 2286 z_stream stream;
706bc531 2287 char *filename;
aac17941 2288 static char tmpfile[PATH_MAX];
a44c9a5e 2289
972a9155 2290 filename = sha1_file_name(sha1);
5723fe7e 2291 fd = create_tmpfile(tmpfile, sizeof(tmpfile), filename);
aac17941 2292 if (fd < 0) {
35243577 2293 if (errno == EACCES)
916d081b
PB
2294 return error("insufficient permission for adding an object to repository database %s\n", get_object_directory());
2295 else
2296 return error("unable to create temporary sha1 filename %s: %s\n", tmpfile, strerror(errno));
aac17941
LT
2297 }
2298
0fcfd160
LT
2299 /* Set it up */
2300 memset(&stream, 0, sizeof(stream));
12f6c308 2301 deflateInit(&stream, zlib_compression_level);
93821bd9 2302 size = 8 + deflateBound(&stream, len+hdrlen);
812666c8 2303 compressed = xmalloc(size);
0fcfd160
LT
2304
2305 /* Compress it */
0fcfd160
LT
2306 stream.next_out = compressed;
2307 stream.avail_out = size;
a44c9a5e
LT
2308
2309 /* First header.. */
d65a16f6 2310 stream.next_in = (unsigned char *)hdr;
a44c9a5e 2311 stream.avail_in = hdrlen;
726f852b
NP
2312 while (deflate(&stream, 0) == Z_OK)
2313 /* nothing */;
a44c9a5e
LT
2314
2315 /* Then the data itself.. */
2316 stream.next_in = buf;
2317 stream.avail_in = len;
ac54c277
LT
2318 ret = deflate(&stream, Z_FINISH);
2319 if (ret != Z_STREAM_END)
2320 die("unable to deflate new object %s (%d)", sha1_to_hex(sha1), ret);
2321
2322 ret = deflateEnd(&stream);
2323 if (ret != Z_OK)
2324 die("deflateEnd on object %s failed (%d)", sha1_to_hex(sha1), ret);
2325
0fcfd160
LT
2326 size = stream.total_out;
2327
4d548150
LT
2328 if (write_buffer(fd, compressed, size) < 0)
2329 die("unable to write sha1 file");
e9039dd3 2330 close_sha1_file(fd);
383f85b7 2331 free(compressed);
0fcfd160 2332
bbac7311
NP
2333 if (mtime) {
2334 struct utimbuf utb;
2335 utb.actime = mtime;
2336 utb.modtime = mtime;
2337 if (utime(tmpfile, &utb) < 0)
2338 warning("failed utime() on %s: %s",
2339 tmpfile, strerror(errno));
2340 }
2341
230f1322 2342 return move_temp_to_file(tmpfile, filename);
0fcfd160 2343}
8237b185 2344
bbac7311
NP
2345int write_sha1_file(void *buf, unsigned long len, const char *type, unsigned char *returnsha1)
2346{
2347 unsigned char sha1[20];
2348 char hdr[32];
2349 int hdrlen;
2350
2351 /* Normally if we have it in the pack then we do not bother writing
2352 * it out into .git/objects/??/?{38} file.
2353 */
2354 write_sha1_file_prepare(buf, len, type, sha1, hdr, &hdrlen);
2355 if (returnsha1)
2356 hashcpy(returnsha1, sha1);
2357 if (has_sha1_file(sha1))
2358 return 0;
2359 return write_loose_object(sha1, hdr, hdrlen, buf, len, 0);
2360}
2361
2362int force_object_loose(const unsigned char *sha1, time_t mtime)
2363{
bbac7311
NP
2364 void *buf;
2365 unsigned long len;
2366 enum object_type type;
2367 char hdr[32];
2368 int hdrlen;
1fb23e65 2369 int ret;
bbac7311 2370
c529d75a 2371 if (has_loose_object(sha1))
bbac7311
NP
2372 return 0;
2373 buf = read_packed_sha1(sha1, &type, &len);
2374 if (!buf)
2375 return error("cannot read sha1_file for %s", sha1_to_hex(sha1));
2376 hdrlen = sprintf(hdr, "%s %lu", typename(type), len) + 1;
1fb23e65
BS
2377 ret = write_loose_object(sha1, hdr, hdrlen, buf, len, mtime);
2378 free(buf);
2379
2380 return ret;
bbac7311
NP
2381}
2382
bf592c50
DB
2383int has_pack_index(const unsigned char *sha1)
2384{
2385 struct stat st;
2386 if (stat(sha1_pack_index_name(sha1), &st))
2387 return 0;
2388 return 1;
2389}
2390
2391int has_pack_file(const unsigned char *sha1)
2392{
2393 struct stat st;
2394 if (stat(sha1_pack_name(sha1), &st))
2395 return 0;
2396 return 1;
2397}
2398
106d710b 2399int has_sha1_pack(const unsigned char *sha1, const char **ignore_packed)
dade09c2
LT
2400{
2401 struct pack_entry e;
106d710b 2402 return find_pack_entry(sha1, &e, ignore_packed);
dade09c2
LT
2403}
2404
8237b185
DB
2405int has_sha1_file(const unsigned char *sha1)
2406{
1f688557
JH
2407 struct pack_entry e;
2408
106d710b 2409 if (find_pack_entry(sha1, &e, NULL))
1f688557 2410 return 1;
c529d75a 2411 return has_loose_object(sha1);
8237b185 2412}
74400e71 2413
e7332f96
BE
2414int index_pipe(unsigned char *sha1, int fd, const char *type, int write_object)
2415{
fd17f5b5 2416 struct strbuf buf;
e7332f96 2417 int ret;
e7332f96 2418
fd17f5b5
PH
2419 strbuf_init(&buf, 0);
2420 if (strbuf_read(&buf, fd, 4096) < 0) {
2421 strbuf_release(&buf);
024510c8
DB
2422 return -1;
2423 }
e7332f96 2424
024510c8 2425 if (!type)
8e440259 2426 type = blob_type;
024510c8 2427 if (write_object)
fd17f5b5 2428 ret = write_sha1_file(buf.buf, buf.len, type, sha1);
abdc3fc8 2429 else
fd17f5b5
PH
2430 ret = hash_sha1_file(buf.buf, buf.len, type, sha1);
2431 strbuf_release(&buf);
2432
024510c8
DB
2433 return ret;
2434}
2435
edaec3fb 2436int index_fd(unsigned char *sha1, int fd, struct stat *st, int write_object,
53bca91a 2437 enum object_type type, const char *path)
74400e71 2438{
dc49cd76 2439 size_t size = xsize_t(st->st_size);
3a55602e 2440 void *buf = NULL;
6c510bee 2441 int ret, re_allocated = 0;
74400e71 2442
74400e71 2443 if (size)
c4712e45 2444 buf = xmmap(NULL, size, PROT_READ, MAP_PRIVATE, fd, 0);
74400e71 2445 close(fd);
74400e71 2446
7672db20 2447 if (!type)
edaec3fb 2448 type = OBJ_BLOB;
6c510bee
LT
2449
2450 /*
2451 * Convert blobs to git internal format
2452 */
d0d8e14d 2453 if ((type == OBJ_BLOB) && S_ISREG(st->st_mode)) {
5ecd293d
PH
2454 struct strbuf nbuf;
2455 strbuf_init(&nbuf, 0);
21e5ad50
SP
2456 if (convert_to_git(path, buf, size, &nbuf,
2457 write_object ? safe_crlf : 0)) {
ac78e548 2458 munmap(buf, size);
b315c5c0 2459 buf = strbuf_detach(&nbuf, &size);
6c510bee
LT
2460 re_allocated = 1;
2461 }
2462 }
2463
7672db20 2464 if (write_object)
edaec3fb 2465 ret = write_sha1_file(buf, size, typename(type), sha1);
abdc3fc8 2466 else
edaec3fb 2467 ret = hash_sha1_file(buf, size, typename(type), sha1);
6c510bee
LT
2468 if (re_allocated) {
2469 free(buf);
2470 return ret;
2471 }
aac17941
LT
2472 if (size)
2473 munmap(buf, size);
2474 return ret;
74400e71 2475}
ec1fcc16
JH
2476
2477int index_path(unsigned char *sha1, const char *path, struct stat *st, int write_object)
2478{
2479 int fd;
2480 char *target;
dc49cd76 2481 size_t len;
ec1fcc16
JH
2482
2483 switch (st->st_mode & S_IFMT) {
2484 case S_IFREG:
2485 fd = open(path, O_RDONLY);
2486 if (fd < 0)
2487 return error("open(\"%s\"): %s", path,
2488 strerror(errno));
53bca91a 2489 if (index_fd(sha1, fd, st, write_object, OBJ_BLOB, path) < 0)
ec1fcc16
JH
2490 return error("%s: failed to insert into database",
2491 path);
2492 break;
2493 case S_IFLNK:
dc49cd76
SP
2494 len = xsize_t(st->st_size);
2495 target = xmalloc(len + 1);
2496 if (readlink(path, target, len + 1) != st->st_size) {
ec1fcc16
JH
2497 char *errstr = strerror(errno);
2498 free(target);
2499 return error("readlink(\"%s\"): %s", path,
2500 errstr);
2501 }
abdc3fc8 2502 if (!write_object)
dc49cd76
SP
2503 hash_sha1_file(target, len, blob_type, sha1);
2504 else if (write_sha1_file(target, len, blob_type, sha1))
ec1fcc16
JH
2505 return error("%s: failed to insert into database",
2506 path);
2507 free(target);
2508 break;
f35a6d3b
LT
2509 case S_IFDIR:
2510 return resolve_gitlink_ref(path, "HEAD", sha1);
ec1fcc16
JH
2511 default:
2512 return error("%s: unsupported file type", path);
2513 }
2514 return 0;
2515}
a69e5429
JH
2516
2517int read_pack_header(int fd, struct pack_header *header)
2518{
c697ad14
HO
2519 if (read_in_full(fd, header, sizeof(*header)) < sizeof(*header))
2520 /* "eof before pack header was fully read" */
2521 return PH_ERROR_EOF;
2522
a69e5429
JH
2523 if (header->hdr_signature != htonl(PACK_SIGNATURE))
2524 /* "protocol error (pack signature mismatch detected)" */
2525 return PH_ERROR_PACK_SIGNATURE;
2526 if (!pack_version_ok(header->hdr_version))
2527 /* "protocol error (pack version unsupported)" */
2528 return PH_ERROR_PROTOCOL;
2529 return 0;
2530}