]>
Commit | Line | Data |
---|---|---|
8bc9a0c7 LT |
1 | /* |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | */ | |
4aab5b46 | 6 | #define NO_THE_INDEX_COMPATIBILITY_MACROS |
e83c5163 | 7 | #include "cache.h" |
b2141fc1 | 8 | #include "config.h" |
cffbfad5 EN |
9 | #include "diff.h" |
10 | #include "diffcore.h" | |
f6ecc62d | 11 | #include "tempfile.h" |
697cc8ef | 12 | #include "lockfile.h" |
bad68ec9 | 13 | #include "cache-tree.h" |
f35a6d3b | 14 | #include "refs.h" |
d616813d | 15 | #include "dir.h" |
cbd53a21 | 16 | #include "object-store.h" |
041aee31 JH |
17 | #include "tree.h" |
18 | #include "commit.h" | |
39425819 | 19 | #include "blob.h" |
cfc5789a | 20 | #include "resolve-undo.h" |
6c9cd161 JH |
21 | #include "strbuf.h" |
22 | #include "varint.h" | |
5fc2fc8f | 23 | #include "split-index.h" |
a42643aa | 24 | #include "utf8.h" |
883e248b | 25 | #include "fsmonitor.h" |
abb4bb83 | 26 | #include "thread-utils.h" |
bad68ec9 | 27 | |
b60e188c TG |
28 | /* Mask for the name length in ce_flags in the on-disk index */ |
29 | ||
30 | #define CE_NAMEMASK (0x0fff) | |
31 | ||
bad68ec9 JH |
32 | /* Index extensions. |
33 | * | |
34 | * The first letter should be 'A'..'Z' for extensions that are not | |
35 | * necessary for a correct operation (i.e. optimization data). | |
36 | * When new extensions are added that _needs_ to be understood in | |
37 | * order to correctly interpret the index file, pick character that | |
38 | * is outside the range, to cause the reader to abort. | |
39 | */ | |
40 | ||
41 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
42 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 43 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 44 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 45 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 46 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 47 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 48 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
5fc2fc8f NTND |
49 | |
50 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 51 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 52 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 53 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 54 | |
8e72d675 JM |
55 | |
56 | /* | |
57 | * This is an estimate of the pathname length in the index. We use | |
58 | * this for V4 index files to guess the un-deltafied size of the index | |
59 | * in memory because of pathname deltafication. This is not required | |
60 | * for V2/V3 index formats because their pathnames are not compressed. | |
61 | * If the initial amount of memory set aside is not sufficient, the | |
62 | * mem pool will allocate extra memory. | |
63 | */ | |
64 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
65 | ||
66 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) | |
67 | { | |
68 | struct cache_entry *ce; | |
69 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
70 | ce->mem_pool_allocated = 1; | |
71 | return ce; | |
72 | } | |
73 | ||
74 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
75 | { | |
76 | struct cache_entry * ce; | |
77 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
78 | ce->mem_pool_allocated = 1; | |
79 | return ce; | |
80 | } | |
81 | ||
82 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
83 | { | |
84 | struct mem_pool **pool_ptr; | |
85 | ||
86 | if (istate->split_index && istate->split_index->base) | |
87 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
88 | else | |
89 | pool_ptr = &istate->ce_mem_pool; | |
90 | ||
91 | if (!*pool_ptr) | |
92 | mem_pool_init(pool_ptr, 0); | |
93 | ||
94 | return *pool_ptr; | |
95 | } | |
96 | ||
228e94f9 | 97 | struct index_state the_index; |
626f35c8 | 98 | static const char *alternate_index_output; |
8fd2cb40 | 99 | |
9cb76b8c JH |
100 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
101 | { | |
102 | istate->cache[nr] = ce; | |
96872bc2 | 103 | add_name_hash(istate, ce); |
9cb76b8c JH |
104 | } |
105 | ||
cf558704 LT |
106 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
107 | { | |
108 | struct cache_entry *old = istate->cache[nr]; | |
109 | ||
078a58e8 | 110 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 111 | remove_name_hash(istate, old); |
a849735b | 112 | discard_cache_entry(old); |
0e267b7a | 113 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 114 | set_index_entry(istate, nr, ce); |
078a58e8 | 115 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 116 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 117 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
118 | } |
119 | ||
81dc2307 PB |
120 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
121 | { | |
285c2e25 | 122 | struct cache_entry *old_entry = istate->cache[nr], *new_entry; |
81dc2307 PB |
123 | int namelen = strlen(new_name); |
124 | ||
a849735b | 125 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
126 | copy_cache_entry(new_entry, old_entry); |
127 | new_entry->ce_flags &= ~CE_HASHED; | |
128 | new_entry->ce_namelen = namelen; | |
129 | new_entry->index = 0; | |
130 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 131 | |
285c2e25 BW |
132 | cache_tree_invalidate_path(istate, old_entry->name); |
133 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 134 | remove_index_entry_at(istate, nr); |
285c2e25 | 135 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); |
81dc2307 PB |
136 | } |
137 | ||
c21d39d7 MH |
138 | void fill_stat_data(struct stat_data *sd, struct stat *st) |
139 | { | |
140 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
141 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
142 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
143 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
144 | sd->sd_dev = st->st_dev; | |
145 | sd->sd_ino = st->st_ino; | |
146 | sd->sd_uid = st->st_uid; | |
147 | sd->sd_gid = st->st_gid; | |
148 | sd->sd_size = st->st_size; | |
149 | } | |
150 | ||
151 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
152 | { | |
153 | int changed = 0; | |
154 | ||
155 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
156 | changed |= MTIME_CHANGED; | |
157 | if (trust_ctime && check_stat && | |
158 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
159 | changed |= CTIME_CHANGED; | |
160 | ||
161 | #ifdef USE_NSEC | |
162 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
163 | changed |= MTIME_CHANGED; | |
164 | if (trust_ctime && check_stat && | |
165 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
166 | changed |= CTIME_CHANGED; | |
167 | #endif | |
168 | ||
169 | if (check_stat) { | |
170 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
171 | sd->sd_gid != (unsigned int) st->st_gid) | |
172 | changed |= OWNER_CHANGED; | |
173 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
174 | changed |= INODE_CHANGED; | |
175 | } | |
176 | ||
177 | #ifdef USE_STDEV | |
178 | /* | |
179 | * st_dev breaks on network filesystems where different | |
180 | * clients will have different views of what "device" | |
181 | * the filesystem is on | |
182 | */ | |
183 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
184 | changed |= INODE_CHANGED; | |
185 | #endif | |
186 | ||
187 | if (sd->sd_size != (unsigned int) st->st_size) | |
188 | changed |= DATA_CHANGED; | |
189 | ||
190 | return changed; | |
191 | } | |
192 | ||
415e96c8 JH |
193 | /* |
194 | * This only updates the "non-critical" parts of the directory | |
195 | * cache, ie the parts that aren't tracked by GIT, and only used | |
196 | * to validate the cache. | |
197 | */ | |
198 | void fill_stat_cache_info(struct cache_entry *ce, struct stat *st) | |
199 | { | |
c21d39d7 | 200 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
201 | |
202 | if (assume_unchanged) | |
7a51ed66 | 203 | ce->ce_flags |= CE_VALID; |
eadb5831 | 204 | |
883e248b | 205 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 206 | ce_mark_uptodate(ce); |
883e248b BP |
207 | mark_fsmonitor_valid(ce); |
208 | } | |
415e96c8 JH |
209 | } |
210 | ||
58bf2a4c NTND |
211 | static int ce_compare_data(struct index_state *istate, |
212 | const struct cache_entry *ce, | |
213 | struct stat *st) | |
29e4d363 JH |
214 | { |
215 | int match = -1; | |
1b8ac5ea | 216 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
217 | |
218 | if (fd >= 0) { | |
bebfecb9 | 219 | struct object_id oid; |
58bf2a4c | 220 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 221 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 222 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
223 | } |
224 | return match; | |
225 | } | |
226 | ||
21a6b9fa | 227 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
228 | { |
229 | int match = -1; | |
29e4d363 JH |
230 | void *buffer; |
231 | unsigned long size; | |
21666f1a | 232 | enum object_type type; |
a60272b3 | 233 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 234 | |
a60272b3 | 235 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 236 | return -1; |
a60272b3 | 237 | |
b4f5aca4 | 238 | buffer = read_object_file(&ce->oid, &type, &size); |
a60272b3 LT |
239 | if (buffer) { |
240 | if (size == sb.len) | |
241 | match = memcmp(buffer, sb.buf, size); | |
242 | free(buffer); | |
29e4d363 | 243 | } |
a60272b3 | 244 | strbuf_release(&sb); |
29e4d363 JH |
245 | return match; |
246 | } | |
247 | ||
21a6b9fa | 248 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 249 | { |
1053fe82 | 250 | struct object_id oid; |
f35a6d3b LT |
251 | |
252 | /* | |
253 | * We don't actually require that the .git directory | |
302b9282 | 254 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
255 | * might even be missing (in case nobody populated that |
256 | * sub-project). | |
257 | * | |
258 | * If so, we consider it always to match. | |
259 | */ | |
a98e6101 | 260 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 261 | return 0; |
6a29d7b7 | 262 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
263 | } |
264 | ||
58bf2a4c NTND |
265 | static int ce_modified_check_fs(struct index_state *istate, |
266 | const struct cache_entry *ce, | |
267 | struct stat *st) | |
29e4d363 JH |
268 | { |
269 | switch (st->st_mode & S_IFMT) { | |
270 | case S_IFREG: | |
58bf2a4c | 271 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
272 | return DATA_CHANGED; |
273 | break; | |
274 | case S_IFLNK: | |
dc49cd76 | 275 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
276 | return DATA_CHANGED; |
277 | break; | |
a8ee75bc | 278 | case S_IFDIR: |
7a51ed66 | 279 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 280 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 281 | /* else fallthrough */ |
29e4d363 JH |
282 | default: |
283 | return TYPE_CHANGED; | |
284 | } | |
285 | return 0; | |
286 | } | |
287 | ||
21a6b9fa | 288 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
289 | { |
290 | unsigned int changed = 0; | |
291 | ||
7a51ed66 LT |
292 | if (ce->ce_flags & CE_REMOVE) |
293 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
294 | ||
295 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
296 | case S_IFREG: |
297 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
298 | /* We consider only the owner x bit to be relevant for |
299 | * "mode changes" | |
300 | */ | |
301 | if (trust_executable_bit && | |
7a51ed66 | 302 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 303 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
304 | break; |
305 | case S_IFLNK: | |
78a8d641 JS |
306 | if (!S_ISLNK(st->st_mode) && |
307 | (has_symlinks || !S_ISREG(st->st_mode))) | |
308 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 309 | break; |
302b9282 | 310 | case S_IFGITLINK: |
c70115b4 | 311 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
312 | if (!S_ISDIR(st->st_mode)) |
313 | changed |= TYPE_CHANGED; | |
314 | else if (ce_compare_gitlink(ce)) | |
315 | changed |= DATA_CHANGED; | |
a8ee75bc | 316 | return changed; |
8ae0a8c5 | 317 | default: |
7a51ed66 | 318 | die("internal error: ce_mode is %o", ce->ce_mode); |
8ae0a8c5 | 319 | } |
2cb45e95 | 320 | |
c21d39d7 | 321 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 322 | |
f49c2c22 | 323 | /* Racily smudged entry? */ |
c21d39d7 | 324 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 325 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
326 | changed |= DATA_CHANGED; |
327 | } | |
328 | ||
407c8eb0 JH |
329 | return changed; |
330 | } | |
331 | ||
2bb4cda1 NTND |
332 | static int is_racy_stat(const struct index_state *istate, |
333 | const struct stat_data *sd) | |
6d91da6d | 334 | { |
2bb4cda1 | 335 | return (istate->timestamp.sec && |
fba2f38a KB |
336 | #ifdef USE_NSEC |
337 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
338 | (istate->timestamp.sec < sd->sd_mtime.sec || |
339 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
340 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 341 | #else |
2bb4cda1 | 342 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 343 | #endif |
2bb4cda1 NTND |
344 | ); |
345 | } | |
346 | ||
347 | static int is_racy_timestamp(const struct index_state *istate, | |
348 | const struct cache_entry *ce) | |
349 | { | |
350 | return (!S_ISGITLINK(ce->ce_mode) && | |
351 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
352 | } |
353 | ||
ed4efab1 NTND |
354 | int match_stat_data_racy(const struct index_state *istate, |
355 | const struct stat_data *sd, struct stat *st) | |
356 | { | |
357 | if (is_racy_stat(istate, sd)) | |
358 | return MTIME_CHANGED; | |
359 | return match_stat_data(sd, st); | |
6d91da6d JH |
360 | } |
361 | ||
883e248b | 362 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 363 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 364 | unsigned int options) |
407c8eb0 | 365 | { |
5f73076c | 366 | unsigned int changed; |
4bd5b7da | 367 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 368 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 369 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 370 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 371 | |
883e248b BP |
372 | if (!ignore_fsmonitor) |
373 | refresh_fsmonitor(istate); | |
5f73076c JH |
374 | /* |
375 | * If it's marked as always valid in the index, it's | |
376 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
377 | * |
378 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 379 | */ |
56cac48c NTND |
380 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
381 | return 0; | |
7a51ed66 | 382 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 383 | return 0; |
883e248b BP |
384 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
385 | return 0; | |
5f73076c | 386 | |
331fcb59 JH |
387 | /* |
388 | * Intent-to-add entries have not been added, so the index entry | |
389 | * by definition never matches what is in the work tree until it | |
390 | * actually gets added. | |
391 | */ | |
895ff3b2 | 392 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
393 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
394 | ||
5f73076c | 395 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 396 | |
29e4d363 JH |
397 | /* |
398 | * Within 1 second of this sequence: | |
399 | * echo xyzzy >file && git-update-index --add file | |
400 | * running this command: | |
401 | * echo frotz >file | |
402 | * would give a falsely clean cache entry. The mtime and | |
403 | * length match the cache, and other stat fields do not change. | |
404 | * | |
405 | * We could detect this at update-index time (the cache entry | |
406 | * being registered/updated records the same time as "now") | |
407 | * and delay the return from git-update-index, but that would | |
408 | * effectively mean we can make at most one commit per second, | |
409 | * which is not acceptable. Instead, we check cache entries | |
410 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 411 | * carefully than others. |
29e4d363 | 412 | */ |
6d91da6d | 413 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
414 | if (assume_racy_is_modified) |
415 | changed |= DATA_CHANGED; | |
416 | else | |
58bf2a4c | 417 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 418 | } |
b0391890 | 419 | |
29e4d363 | 420 | return changed; |
b0391890 JH |
421 | } |
422 | ||
883e248b | 423 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
424 | const struct cache_entry *ce, |
425 | struct stat *st, unsigned int options) | |
b0391890 | 426 | { |
29e4d363 | 427 | int changed, changed_fs; |
4bd5b7da JH |
428 | |
429 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
430 | if (!changed) |
431 | return 0; | |
b0391890 JH |
432 | /* |
433 | * If the mode or type has changed, there's no point in trying | |
434 | * to refresh the entry - it's not going to match | |
435 | */ | |
436 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
437 | return changed; | |
438 | ||
c70115b4 JH |
439 | /* |
440 | * Immediately after read-tree or update-index --cacheinfo, | |
441 | * the length field is zero, as we have never even read the | |
442 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
443 | * returned by ie_match_stat() which in turn was returned by | |
444 | * ce_match_stat_basic() to signal that the filesize of the | |
445 | * blob changed. We have to actually go to the filesystem to | |
446 | * see if the contents match, and if so, should answer "unchanged". | |
447 | * | |
448 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
449 | * already has checked the actual HEAD from the filesystem in the | |
450 | * subproject. If ie_match_stat() already said it is different, | |
451 | * then we know it is. | |
b0391890 | 452 | */ |
c70115b4 | 453 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 454 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
455 | return changed; |
456 | ||
58bf2a4c | 457 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
458 | if (changed_fs) |
459 | return changed | changed_fs; | |
b0391890 JH |
460 | return 0; |
461 | } | |
462 | ||
958ba6c9 LT |
463 | int base_name_compare(const char *name1, int len1, int mode1, |
464 | const char *name2, int len2, int mode2) | |
465 | { | |
466 | unsigned char c1, c2; | |
467 | int len = len1 < len2 ? len1 : len2; | |
468 | int cmp; | |
469 | ||
470 | cmp = memcmp(name1, name2, len); | |
471 | if (cmp) | |
472 | return cmp; | |
473 | c1 = name1[len]; | |
474 | c2 = name2[len]; | |
1833a925 | 475 | if (!c1 && S_ISDIR(mode1)) |
958ba6c9 | 476 | c1 = '/'; |
1833a925 | 477 | if (!c2 && S_ISDIR(mode2)) |
958ba6c9 LT |
478 | c2 = '/'; |
479 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
480 | } | |
481 | ||
0ab9e1e8 LT |
482 | /* |
483 | * df_name_compare() is identical to base_name_compare(), except it | |
484 | * compares conflicting directory/file entries as equal. Note that | |
485 | * while a directory name compares as equal to a regular file, they | |
486 | * then individually compare _differently_ to a filename that has | |
487 | * a dot after the basename (because '\0' < '.' < '/'). | |
488 | * | |
489 | * This is used by routines that want to traverse the git namespace | |
490 | * but then handle conflicting entries together when possible. | |
491 | */ | |
492 | int df_name_compare(const char *name1, int len1, int mode1, | |
493 | const char *name2, int len2, int mode2) | |
494 | { | |
495 | int len = len1 < len2 ? len1 : len2, cmp; | |
496 | unsigned char c1, c2; | |
497 | ||
498 | cmp = memcmp(name1, name2, len); | |
499 | if (cmp) | |
500 | return cmp; | |
501 | /* Directories and files compare equal (same length, same name) */ | |
502 | if (len1 == len2) | |
503 | return 0; | |
504 | c1 = name1[len]; | |
505 | if (!c1 && S_ISDIR(mode1)) | |
506 | c1 = '/'; | |
507 | c2 = name2[len]; | |
508 | if (!c2 && S_ISDIR(mode2)) | |
509 | c2 = '/'; | |
510 | if (c1 == '/' && !c2) | |
511 | return 0; | |
512 | if (c2 == '/' && !c1) | |
513 | return 0; | |
514 | return c1 - c2; | |
515 | } | |
516 | ||
ccdd4a0f | 517 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) |
eb38c22f | 518 | { |
ccdd4a0f JM |
519 | size_t min_len = (len1 < len2) ? len1 : len2; |
520 | int cmp = memcmp(name1, name2, min_len); | |
eb38c22f LT |
521 | if (cmp) |
522 | return cmp; | |
523 | if (len1 < len2) | |
524 | return -1; | |
525 | if (len1 > len2) | |
526 | return 1; | |
ccdd4a0f JM |
527 | return 0; |
528 | } | |
529 | ||
530 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
531 | { | |
532 | int cmp; | |
533 | ||
534 | cmp = name_compare(name1, len1, name2, len2); | |
535 | if (cmp) | |
536 | return cmp; | |
5f73076c | 537 | |
b60e188c | 538 | if (stage1 < stage2) |
95fd5bf8 | 539 | return -1; |
b60e188c | 540 | if (stage1 > stage2) |
95fd5bf8 | 541 | return 1; |
eb38c22f LT |
542 | return 0; |
543 | } | |
544 | ||
357e9c69 | 545 | static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage) |
eb38c22f LT |
546 | { |
547 | int first, last; | |
548 | ||
549 | first = 0; | |
4aab5b46 | 550 | last = istate->cache_nr; |
eb38c22f LT |
551 | while (last > first) { |
552 | int next = (last + first) >> 1; | |
4aab5b46 | 553 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 554 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 555 | if (!cmp) |
76e7f4ec | 556 | return next; |
eb38c22f LT |
557 | if (cmp < 0) { |
558 | last = next; | |
559 | continue; | |
560 | } | |
561 | first = next+1; | |
562 | } | |
76e7f4ec | 563 | return -first-1; |
eb38c22f LT |
564 | } |
565 | ||
b60e188c TG |
566 | int index_name_pos(const struct index_state *istate, const char *name, int namelen) |
567 | { | |
568 | return index_name_stage_pos(istate, name, namelen, 0); | |
569 | } | |
570 | ||
4aab5b46 | 571 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 572 | { |
cf558704 LT |
573 | struct cache_entry *ce = istate->cache[pos]; |
574 | ||
cfc5789a | 575 | record_resolve_undo(istate, ce); |
2092678c | 576 | remove_name_hash(istate, ce); |
045113a5 | 577 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 578 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
579 | istate->cache_nr--; |
580 | if (pos >= istate->cache_nr) | |
7b937ca3 | 581 | return 0; |
f331ab9d RS |
582 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
583 | istate->cache_nr - pos); | |
7b937ca3 LT |
584 | return 1; |
585 | } | |
586 | ||
36419c8e | 587 | /* |
98e023de | 588 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
589 | * CE_REMOVE is set in ce_flags. This is much more effective than |
590 | * calling remove_index_entry_at() for each entry to be removed. | |
591 | */ | |
592 | void remove_marked_cache_entries(struct index_state *istate) | |
593 | { | |
594 | struct cache_entry **ce_array = istate->cache; | |
595 | unsigned int i, j; | |
596 | ||
597 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 598 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
2092678c | 599 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 600 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 601 | } |
36419c8e KB |
602 | else |
603 | ce_array[j++] = ce_array[i]; | |
604 | } | |
ad837d9e NTND |
605 | if (j == istate->cache_nr) |
606 | return; | |
e636a7b4 | 607 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
608 | istate->cache_nr = j; |
609 | } | |
610 | ||
4aab5b46 | 611 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 612 | { |
4aab5b46 | 613 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
614 | if (pos < 0) |
615 | pos = -pos-1; | |
a5400efe | 616 | cache_tree_invalidate_path(istate, path); |
e931371a | 617 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
618 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
619 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
620 | return 0; |
621 | } | |
622 | ||
20314271 JS |
623 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
624 | { | |
625 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
626 | } | |
627 | ||
628 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
629 | const char *path, int namelen) | |
630 | { | |
631 | int pos = index_name_pos(istate, path, namelen); | |
632 | struct cache_entry *ce; | |
633 | ||
634 | if (pos >= 0) | |
635 | return pos; | |
636 | ||
637 | /* maybe unmerged? */ | |
638 | pos = -1 - pos; | |
639 | if (pos >= istate->cache_nr || | |
640 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
641 | return -1; | |
642 | ||
643 | /* order of preference: stage 2, 1, 3 */ | |
644 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
645 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
646 | !compare_name(ce, path, namelen)) | |
647 | pos++; | |
648 | return pos; | |
649 | } | |
650 | ||
1102952b LT |
651 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
652 | { | |
653 | int len = ce_namelen(ce); | |
654 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
655 | } | |
656 | ||
657 | /* | |
658 | * If we add a filename that aliases in the cache, we will use the | |
659 | * name that we already have - but we don't want to update the same | |
660 | * alias twice, because that implies that there were actually two | |
661 | * different files with aliasing names! | |
662 | * | |
663 | * So we use the CE_ADDED flag to verify that the alias was an old | |
664 | * one before we accept it as | |
665 | */ | |
045113a5 NTND |
666 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
667 | struct cache_entry *ce, | |
668 | struct cache_entry *alias) | |
1102952b LT |
669 | { |
670 | int len; | |
285c2e25 | 671 | struct cache_entry *new_entry; |
1102952b LT |
672 | |
673 | if (alias->ce_flags & CE_ADDED) | |
674 | die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name); | |
675 | ||
676 | /* Ok, create the new entry using the name of the existing alias */ | |
677 | len = ce_namelen(alias); | |
a849735b | 678 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
679 | memcpy(new_entry->name, alias->name, len); |
680 | copy_cache_entry(new_entry, ce); | |
045113a5 | 681 | save_or_free_index_entry(istate, ce); |
285c2e25 | 682 | return new_entry; |
1102952b LT |
683 | } |
684 | ||
b4b313f9 | 685 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 686 | { |
a09c985e PO |
687 | struct object_id oid; |
688 | if (write_object_file("", 0, blob_type, &oid)) | |
39425819 | 689 | die("cannot create an empty blob in the object database"); |
a09c985e | 690 | oidcpy(&ce->oid, &oid); |
39425819 JH |
691 | } |
692 | ||
610d55af | 693 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 694 | { |
a849735b | 695 | int namelen, was_same; |
d177cab0 | 696 | mode_t st_mode = st->st_mode; |
9472935d | 697 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 698 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
699 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
700 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
701 | int intent_only = flags & ADD_CACHE_INTENT; |
702 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
703 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
9472935d TB |
704 | int newflags = HASH_WRITE_OBJECT; |
705 | ||
706 | if (flags & HASH_RENORMALIZE) | |
707 | newflags |= HASH_RENORMALIZE; | |
11be42a4 | 708 | |
d177cab0 | 709 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
960b8ad1 | 710 | return error("%s: can only add regular files, symbolic links or git-directories", path); |
11be42a4 JS |
711 | |
712 | namelen = strlen(path); | |
d177cab0 | 713 | if (S_ISDIR(st_mode)) { |
09595258 LT |
714 | while (namelen && path[namelen-1] == '/') |
715 | namelen--; | |
716 | } | |
a849735b | 717 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 718 | memcpy(ce->name, path, namelen); |
b60e188c | 719 | ce->ce_namelen = namelen; |
39425819 JH |
720 | if (!intent_only) |
721 | fill_stat_cache_info(ce, st); | |
388b2acd JH |
722 | else |
723 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 724 | |
610d55af TG |
725 | |
726 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 727 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 728 | } else { |
78a8d641 JS |
729 | /* If there is an existing entry, pick the mode bits and type |
730 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 731 | */ |
185c975f | 732 | struct cache_entry *ent; |
20314271 | 733 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 734 | |
4aab5b46 | 735 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 736 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
737 | } |
738 | ||
dc1ae704 JJ |
739 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
740 | * case already exists within the Git repository. If it does, ensure the directory | |
741 | * case of the file being added to the repository matches (is folded into) the existing | |
742 | * entry's directory case. | |
743 | */ | |
744 | if (ignore_case) { | |
41284eb0 | 745 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 746 | } |
9472935d TB |
747 | if (!(flags & HASH_RENORMALIZE)) { |
748 | alias = index_file_exists(istate, ce->name, | |
749 | ce_namelen(ce), ignore_case); | |
750 | if (alias && | |
751 | !ce_stage(alias) && | |
752 | !ie_match_stat(istate, alias, st, ce_option)) { | |
753 | /* Nothing changed, really */ | |
754 | if (!S_ISGITLINK(alias->ce_mode)) | |
755 | ce_mark_uptodate(alias); | |
756 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 757 | |
a849735b | 758 | discard_cache_entry(ce); |
9472935d TB |
759 | return 0; |
760 | } | |
0781b8a9 | 761 | } |
39425819 | 762 | if (!intent_only) { |
58bf2a4c | 763 | if (index_path(istate, &ce->oid, path, st, newflags)) { |
a849735b | 764 | discard_cache_entry(ce); |
39425819 | 765 | return error("unable to index file %s", path); |
2d9426b0 | 766 | } |
39425819 | 767 | } else |
b4b313f9 | 768 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 769 | |
1102952b | 770 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 771 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 772 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 773 | |
3bf0dd1f | 774 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
775 | was_same = (alias && |
776 | !ce_stage(alias) && | |
4a7e27e9 | 777 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
778 | ce->ce_mode == alias->ce_mode); |
779 | ||
780 | if (pretend) | |
a849735b | 781 | discard_cache_entry(ce); |
067178ed | 782 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 783 | discard_cache_entry(ce); |
067178ed JH |
784 | return error("unable to add %s to index", path); |
785 | } | |
38ed1d89 | 786 | if (verbose && !was_same) |
11be42a4 | 787 | printf("add '%s'\n", path); |
11be42a4 JS |
788 | return 0; |
789 | } | |
790 | ||
610d55af | 791 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
792 | { |
793 | struct stat st; | |
794 | if (lstat(path, &st)) | |
d824cbba | 795 | die_errno("unable to stat '%s'", path); |
610d55af | 796 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
797 | } |
798 | ||
a849735b JM |
799 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
800 | { | |
8e72d675 | 801 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
802 | } |
803 | ||
804 | struct cache_entry *make_empty_transient_cache_entry(size_t len) | |
805 | { | |
806 | return xcalloc(1, cache_entry_size(len)); | |
807 | } | |
808 | ||
809 | struct cache_entry *make_cache_entry(struct index_state *istate, | |
810 | unsigned int mode, | |
825ed4d9 JM |
811 | const struct object_id *oid, |
812 | const char *path, | |
813 | int stage, | |
814 | unsigned int refresh_options) | |
6640f881 | 815 | { |
bc1c2caa | 816 | struct cache_entry *ce, *ret; |
a849735b | 817 | int len; |
6640f881 | 818 | |
10ecfa76 | 819 | if (!verify_path(path, mode)) { |
7e7abea9 | 820 | error("Invalid path '%s'", path); |
6640f881 | 821 | return NULL; |
7e7abea9 | 822 | } |
6640f881 CR |
823 | |
824 | len = strlen(path); | |
a849735b | 825 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 826 | |
825ed4d9 | 827 | oidcpy(&ce->oid, oid); |
6640f881 | 828 | memcpy(ce->name, path, len); |
b60e188c TG |
829 | ce->ce_flags = create_ce_flags(stage); |
830 | ce->ce_namelen = len; | |
6640f881 CR |
831 | ce->ce_mode = create_ce_mode(mode); |
832 | ||
d7b665c3 | 833 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 834 | if (ret != ce) |
a849735b | 835 | discard_cache_entry(ce); |
915e44c6 | 836 | return ret; |
6640f881 CR |
837 | } |
838 | ||
a849735b JM |
839 | struct cache_entry *make_transient_cache_entry(unsigned int mode, const struct object_id *oid, |
840 | const char *path, int stage) | |
841 | { | |
842 | struct cache_entry *ce; | |
843 | int len; | |
844 | ||
845 | if (!verify_path(path, mode)) { | |
846 | error("Invalid path '%s'", path); | |
847 | return NULL; | |
848 | } | |
849 | ||
850 | len = strlen(path); | |
851 | ce = make_empty_transient_cache_entry(len); | |
852 | ||
853 | oidcpy(&ce->oid, oid); | |
854 | memcpy(ce->name, path, len); | |
855 | ce->ce_flags = create_ce_flags(stage); | |
856 | ce->ce_namelen = len; | |
857 | ce->ce_mode = create_ce_mode(mode); | |
858 | ||
859 | return ce; | |
860 | } | |
861 | ||
d9d70966 TG |
862 | /* |
863 | * Chmod an index entry with either +x or -x. | |
864 | * | |
865 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
866 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
867 | * otherwise. | |
868 | */ | |
869 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
870 | char flip) | |
871 | { | |
872 | if (!S_ISREG(ce->ce_mode)) | |
873 | return -1; | |
874 | switch (flip) { | |
875 | case '+': | |
876 | ce->ce_mode |= 0111; | |
877 | break; | |
878 | case '-': | |
879 | ce->ce_mode &= ~0111; | |
880 | break; | |
881 | default: | |
882 | return -2; | |
883 | } | |
884 | cache_tree_invalidate_path(istate, ce->name); | |
885 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 886 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
887 | istate->cache_changed |= CE_ENTRY_CHANGED; |
888 | ||
889 | return 0; | |
890 | } | |
891 | ||
9c5e6c80 | 892 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
893 | { |
894 | int len = ce_namelen(a); | |
895 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
896 | } | |
897 | ||
8dcf39c4 LT |
898 | /* |
899 | * We fundamentally don't like some paths: we don't want | |
900 | * dot or dot-dot anywhere, and for obvious reasons don't | |
901 | * want to recurse into ".git" either. | |
902 | * | |
903 | * Also, we don't want double slashes or slashes at the | |
904 | * end that can make pathnames ambiguous. | |
905 | */ | |
10ecfa76 | 906 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
907 | { |
908 | /* | |
909 | * The first character was '.', but that | |
910 | * has already been discarded, we now test | |
911 | * the rest. | |
912 | */ | |
e0f530ff | 913 | |
8dcf39c4 | 914 | /* "." is not allowed */ |
e0f530ff | 915 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
916 | return 0; |
917 | ||
e0f530ff | 918 | switch (*rest) { |
8dcf39c4 | 919 | /* |
641084b6 JK |
920 | * ".git" followed by NUL or slash is bad. Note that we match |
921 | * case-insensitively here, even if ignore_case is not set. | |
922 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
923 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
924 | * |
925 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
926 | * case-insensitively). | |
8dcf39c4 LT |
927 | */ |
928 | case 'g': | |
cc2fc7c2 JK |
929 | case 'G': |
930 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 931 | break; |
cc2fc7c2 | 932 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 933 | break; |
e19e5e66 JK |
934 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
935 | return 0; | |
10ecfa76 JK |
936 | if (S_ISLNK(mode)) { |
937 | rest += 3; | |
938 | if (skip_iprefix(rest, "modules", &rest) && | |
939 | (*rest == '\0' || is_dir_sep(*rest))) | |
940 | return 0; | |
941 | } | |
e19e5e66 | 942 | break; |
8dcf39c4 | 943 | case '.': |
e0f530ff | 944 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
945 | return 0; |
946 | } | |
947 | return 1; | |
948 | } | |
949 | ||
10ecfa76 | 950 | int verify_path(const char *path, unsigned mode) |
8dcf39c4 LT |
951 | { |
952 | char c; | |
953 | ||
56948cb6 EFL |
954 | if (has_dos_drive_prefix(path)) |
955 | return 0; | |
956 | ||
8dcf39c4 LT |
957 | goto inside; |
958 | for (;;) { | |
959 | if (!c) | |
960 | return 1; | |
56948cb6 | 961 | if (is_dir_sep(c)) { |
8dcf39c4 | 962 | inside: |
10ecfa76 JK |
963 | if (protect_hfs) { |
964 | if (is_hfs_dotgit(path)) | |
965 | return 0; | |
966 | if (S_ISLNK(mode)) { | |
967 | if (is_hfs_dotgitmodules(path)) | |
968 | return 0; | |
969 | } | |
970 | } | |
971 | if (protect_ntfs) { | |
972 | if (is_ntfs_dotgit(path)) | |
973 | return 0; | |
974 | if (S_ISLNK(mode)) { | |
975 | if (is_ntfs_dotgitmodules(path)) | |
976 | return 0; | |
977 | } | |
978 | } | |
979 | ||
8dcf39c4 | 980 | c = *path++; |
10ecfa76 | 981 | if ((c == '.' && !verify_dotfile(path, mode)) || |
3bdf09c7 JH |
982 | is_dir_sep(c) || c == '\0') |
983 | return 0; | |
8dcf39c4 LT |
984 | } |
985 | c = *path++; | |
986 | } | |
987 | } | |
988 | ||
12676608 LT |
989 | /* |
990 | * Do we have another file that has the beginning components being a | |
991 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 992 | */ |
4aab5b46 JH |
993 | static int has_file_name(struct index_state *istate, |
994 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 995 | { |
12676608 LT |
996 | int retval = 0; |
997 | int len = ce_namelen(ce); | |
b155725d | 998 | int stage = ce_stage(ce); |
12676608 | 999 | const char *name = ce->name; |
0f1e4f04 | 1000 | |
4aab5b46 JH |
1001 | while (pos < istate->cache_nr) { |
1002 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1003 | |
12676608 | 1004 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1005 | break; |
12676608 LT |
1006 | if (memcmp(name, p->name, len)) |
1007 | break; | |
b155725d JH |
1008 | if (ce_stage(p) != stage) |
1009 | continue; | |
12676608 LT |
1010 | if (p->name[len] != '/') |
1011 | continue; | |
7a51ed66 | 1012 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1013 | continue; |
12676608 LT |
1014 | retval = -1; |
1015 | if (!ok_to_replace) | |
1016 | break; | |
4aab5b46 | 1017 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1018 | } |
12676608 LT |
1019 | return retval; |
1020 | } | |
0f1e4f04 | 1021 | |
a6db3fbb JH |
1022 | |
1023 | /* | |
1024 | * Like strcmp(), but also return the offset of the first change. | |
1025 | * If strings are equal, return the length. | |
1026 | */ | |
1027 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1028 | { | |
1029 | size_t k; | |
1030 | ||
1031 | if (!first_change) | |
1032 | return strcmp(s1, s2); | |
1033 | ||
1034 | for (k = 0; s1[k] == s2[k]; k++) | |
1035 | if (s1[k] == '\0') | |
1036 | break; | |
1037 | ||
1038 | *first_change = k; | |
1039 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1040 | } | |
1041 | ||
12676608 LT |
1042 | /* |
1043 | * Do we have another file with a pathname that is a proper | |
1044 | * subset of the name we're trying to add? | |
06b6d81b JH |
1045 | * |
1046 | * That is, is there another file in the index with a path | |
1047 | * that matches a sub-directory in the given entry? | |
12676608 | 1048 | */ |
4aab5b46 JH |
1049 | static int has_dir_name(struct index_state *istate, |
1050 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1051 | { |
1052 | int retval = 0; | |
b155725d | 1053 | int stage = ce_stage(ce); |
12676608 LT |
1054 | const char *name = ce->name; |
1055 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1056 | size_t len_eq_last; |
1057 | int cmp_last = 0; | |
1058 | ||
1059 | /* | |
1060 | * We are frequently called during an iteration on a sorted | |
1061 | * list of pathnames and while building a new index. Therefore, | |
1062 | * there is a high probability that this entry will eventually | |
1063 | * be appended to the index, rather than inserted in the middle. | |
1064 | * If we can confirm that, we can avoid binary searches on the | |
1065 | * components of the pathname. | |
1066 | * | |
1067 | * Compare the entry's full path with the last path in the index. | |
1068 | */ | |
1069 | if (istate->cache_nr > 0) { | |
1070 | cmp_last = strcmp_offset(name, | |
1071 | istate->cache[istate->cache_nr - 1]->name, | |
1072 | &len_eq_last); | |
1073 | if (cmp_last > 0) { | |
1074 | if (len_eq_last == 0) { | |
1075 | /* | |
1076 | * The entry sorts AFTER the last one in the | |
1077 | * index and their paths have no common prefix, | |
1078 | * so there cannot be a F/D conflict. | |
1079 | */ | |
1080 | return retval; | |
1081 | } else { | |
1082 | /* | |
1083 | * The entry sorts AFTER the last one in the | |
1084 | * index, but has a common prefix. Fall through | |
1085 | * to the loop below to disect the entry's path | |
1086 | * and see where the difference is. | |
1087 | */ | |
1088 | } | |
1089 | } else if (cmp_last == 0) { | |
1090 | /* | |
1091 | * The entry exactly matches the last one in the | |
1092 | * index, but because of multiple stage and CE_REMOVE | |
1093 | * items, we fall through and let the regular search | |
1094 | * code handle it. | |
1095 | */ | |
1096 | } | |
1097 | } | |
0f1e4f04 | 1098 | |
12676608 | 1099 | for (;;) { |
b986df5c | 1100 | size_t len; |
0f1e4f04 | 1101 | |
12676608 LT |
1102 | for (;;) { |
1103 | if (*--slash == '/') | |
1104 | break; | |
1105 | if (slash <= ce->name) | |
1106 | return retval; | |
1107 | } | |
1108 | len = slash - name; | |
0f1e4f04 | 1109 | |
b986df5c JH |
1110 | if (cmp_last > 0) { |
1111 | /* | |
1112 | * (len + 1) is a directory boundary (including | |
1113 | * the trailing slash). And since the loop is | |
1114 | * decrementing "slash", the first iteration is | |
1115 | * the longest directory prefix; subsequent | |
1116 | * iterations consider parent directories. | |
1117 | */ | |
1118 | ||
1119 | if (len + 1 <= len_eq_last) { | |
1120 | /* | |
1121 | * The directory prefix (including the trailing | |
1122 | * slash) also appears as a prefix in the last | |
1123 | * entry, so the remainder cannot collide (because | |
1124 | * strcmp said the whole path was greater). | |
1125 | * | |
1126 | * EQ: last: xxx/A | |
1127 | * this: xxx/B | |
1128 | * | |
1129 | * LT: last: xxx/file_A | |
1130 | * this: xxx/file_B | |
1131 | */ | |
1132 | return retval; | |
1133 | } | |
1134 | ||
1135 | if (len > len_eq_last) { | |
1136 | /* | |
1137 | * This part of the directory prefix (excluding | |
1138 | * the trailing slash) is longer than the known | |
1139 | * equal portions, so this sub-directory cannot | |
1140 | * collide with a file. | |
1141 | * | |
1142 | * GT: last: xxxA | |
1143 | * this: xxxB/file | |
1144 | */ | |
1145 | return retval; | |
1146 | } | |
1147 | ||
1148 | if (istate->cache_nr > 0 && | |
1149 | ce_namelen(istate->cache[istate->cache_nr - 1]) > len) { | |
1150 | /* | |
1151 | * The directory prefix lines up with part of | |
1152 | * a longer file or directory name, but sorts | |
1153 | * after it, so this sub-directory cannot | |
1154 | * collide with a file. | |
1155 | * | |
1156 | * last: xxx/yy-file (because '-' sorts before '/') | |
1157 | * this: xxx/yy/abc | |
1158 | */ | |
1159 | return retval; | |
1160 | } | |
1161 | ||
1162 | /* | |
1163 | * This is a possible collision. Fall through and | |
1164 | * let the regular search code handle it. | |
1165 | * | |
1166 | * last: xxx | |
1167 | * this: xxx/file | |
1168 | */ | |
1169 | } | |
1170 | ||
b60e188c | 1171 | pos = index_name_stage_pos(istate, name, len, stage); |
12676608 | 1172 | if (pos >= 0) { |
21cd8d00 JH |
1173 | /* |
1174 | * Found one, but not so fast. This could | |
1175 | * be a marker that says "I was here, but | |
1176 | * I am being removed". Such an entry is | |
1177 | * not a part of the resulting tree, and | |
1178 | * it is Ok to have a directory at the same | |
1179 | * path. | |
1180 | */ | |
077c48df | 1181 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1182 | retval = -1; |
1183 | if (!ok_to_replace) | |
1184 | break; | |
4aab5b46 | 1185 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1186 | continue; |
1187 | } | |
12676608 | 1188 | } |
21cd8d00 JH |
1189 | else |
1190 | pos = -pos-1; | |
12676608 LT |
1191 | |
1192 | /* | |
1193 | * Trivial optimization: if we find an entry that | |
1194 | * already matches the sub-directory, then we know | |
b155725d | 1195 | * we're ok, and we can exit. |
12676608 | 1196 | */ |
4aab5b46 JH |
1197 | while (pos < istate->cache_nr) { |
1198 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1199 | if ((ce_namelen(p) <= len) || |
1200 | (p->name[len] != '/') || | |
1201 | memcmp(p->name, name, len)) | |
1202 | break; /* not our subdirectory */ | |
077c48df JH |
1203 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1204 | /* | |
1205 | * p is at the same stage as our entry, and | |
b155725d JH |
1206 | * is a subdirectory of what we are looking |
1207 | * at, so we cannot have conflicts at our | |
1208 | * level or anything shorter. | |
1209 | */ | |
1210 | return retval; | |
1211 | pos++; | |
192268c1 | 1212 | } |
0f1e4f04 | 1213 | } |
12676608 LT |
1214 | return retval; |
1215 | } | |
1216 | ||
1217 | /* We may be in a situation where we already have path/file and path | |
1218 | * is being added, or we already have path and path/file is being | |
1219 | * added. Either one would result in a nonsense tree that has path | |
1220 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1221 | * |
12676608 LT |
1222 | * If ok-to-replace is specified, we remove the conflicting entries |
1223 | * from the cache so the caller should recompute the insert position. | |
1224 | * When this happens, we return non-zero. | |
1225 | */ | |
4aab5b46 JH |
1226 | static int check_file_directory_conflict(struct index_state *istate, |
1227 | const struct cache_entry *ce, | |
1228 | int pos, int ok_to_replace) | |
12676608 | 1229 | { |
21cd8d00 JH |
1230 | int retval; |
1231 | ||
1232 | /* | |
1233 | * When ce is an "I am going away" entry, we allow it to be added | |
1234 | */ | |
7a51ed66 | 1235 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1236 | return 0; |
1237 | ||
12676608 LT |
1238 | /* |
1239 | * We check if the path is a sub-path of a subsequent pathname | |
1240 | * first, since removing those will not change the position | |
21cd8d00 | 1241 | * in the array. |
12676608 | 1242 | */ |
4aab5b46 | 1243 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1244 | |
12676608 LT |
1245 | /* |
1246 | * Then check if the path might have a clashing sub-directory | |
1247 | * before it. | |
1248 | */ | |
4aab5b46 | 1249 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1250 | } |
1251 | ||
af3785dc | 1252 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1253 | { |
1254 | int pos; | |
192268c1 JH |
1255 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1256 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1257 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1258 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1259 | |
ce7c614b NTND |
1260 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1261 | cache_tree_invalidate_path(istate, ce->name); | |
e5494631 JH |
1262 | |
1263 | /* | |
1264 | * If this entry's path sorts after the last entry in the index, | |
1265 | * we can avoid searching for it. | |
1266 | */ | |
1267 | if (istate->cache_nr > 0 && | |
1268 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
1269 | pos = -istate->cache_nr - 1; | |
1270 | else | |
1271 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); | |
197ee8c9 | 1272 | |
3e09cdfd | 1273 | /* existing match? Just replace it. */ |
76e7f4ec | 1274 | if (pos >= 0) { |
39425819 JH |
1275 | if (!new_only) |
1276 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1277 | return 0; |
1278 | } | |
76e7f4ec | 1279 | pos = -pos-1; |
197ee8c9 | 1280 | |
ffcc9ba7 NTND |
1281 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1282 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1283 | |
7b937ca3 LT |
1284 | /* |
1285 | * Inserting a merged entry ("stage 0") into the index | |
1286 | * will always replace all non-merged entries.. | |
1287 | */ | |
4aab5b46 JH |
1288 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1289 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1290 | ok_to_add = 1; |
4aab5b46 | 1291 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1292 | break; |
1293 | } | |
1294 | } | |
1295 | ||
121481ab LT |
1296 | if (!ok_to_add) |
1297 | return -1; | |
10ecfa76 | 1298 | if (!verify_path(ce->name, ce->ce_mode)) |
7e7abea9 | 1299 | return error("Invalid path '%s'", ce->name); |
121481ab | 1300 | |
3e09cdfd | 1301 | if (!skip_df_check && |
4aab5b46 | 1302 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1303 | if (!ok_to_replace) |
4aab5b46 JH |
1304 | return error("'%s' appears as both a file and as a directory", |
1305 | ce->name); | |
b60e188c | 1306 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); |
192268c1 JH |
1307 | pos = -pos-1; |
1308 | } | |
af3785dc JH |
1309 | return pos + 1; |
1310 | } | |
1311 | ||
1312 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1313 | { | |
1314 | int pos; | |
1315 | ||
1316 | if (option & ADD_CACHE_JUST_APPEND) | |
1317 | pos = istate->cache_nr; | |
1318 | else { | |
1319 | int ret; | |
1320 | ret = add_index_entry_with_check(istate, ce, option); | |
1321 | if (ret <= 0) | |
1322 | return ret; | |
1323 | pos = ret - 1; | |
1324 | } | |
0f1e4f04 | 1325 | |
197ee8c9 | 1326 | /* Make sure the array is big enough .. */ |
999f5660 | 1327 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1328 | |
1329 | /* Add it in.. */ | |
4aab5b46 | 1330 | istate->cache_nr++; |
af3785dc | 1331 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1332 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1333 | istate->cache_nr - pos - 1); | |
cf558704 | 1334 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1335 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1336 | return 0; |
1337 | } | |
1338 | ||
405e5b2f LT |
1339 | /* |
1340 | * "refresh" does not calculate a new sha1 file or bring the | |
1341 | * cache up-to-date for mode/content changes. But what it | |
1342 | * _does_ do is to "re-match" the stat information of a file | |
1343 | * with the cache, so that you can refresh the cache for a | |
1344 | * file that hasn't been changed but where the stat entry is | |
1345 | * out of date. | |
1346 | * | |
1347 | * For example, you'd want to do this after doing a "git-read-tree", | |
1348 | * to link up the stat cache details with the proper files. | |
1349 | */ | |
4aab5b46 | 1350 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1351 | struct cache_entry *ce, |
d05e6970 JK |
1352 | unsigned int options, int *err, |
1353 | int *changed_ret) | |
405e5b2f LT |
1354 | { |
1355 | struct stat st; | |
1356 | struct cache_entry *updated; | |
a849735b | 1357 | int changed; |
25762726 | 1358 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1359 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1360 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1361 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1362 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1363 | |
25762726 | 1364 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1365 | return ce; |
1366 | ||
883e248b BP |
1367 | if (!ignore_fsmonitor) |
1368 | refresh_fsmonitor(istate); | |
aa9349d4 | 1369 | /* |
56cac48c NTND |
1370 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1371 | * that the change to the work tree does not matter and told | |
1372 | * us not to worry. | |
aa9349d4 | 1373 | */ |
56cac48c NTND |
1374 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1375 | ce_mark_uptodate(ce); | |
1376 | return ce; | |
1377 | } | |
aa9349d4 MSO |
1378 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1379 | ce_mark_uptodate(ce); | |
1380 | return ce; | |
1381 | } | |
883e248b BP |
1382 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1383 | ce_mark_uptodate(ce); | |
1384 | return ce; | |
1385 | } | |
aa9349d4 | 1386 | |
ccad42d4 RS |
1387 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1388 | if (ignore_missing) | |
1389 | return ce; | |
1390 | if (err) | |
1391 | *err = ENOENT; | |
1392 | return NULL; | |
1393 | } | |
1394 | ||
8fd2cb40 | 1395 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1396 | if (ignore_missing && errno == ENOENT) |
1397 | return ce; | |
ec0cc704 JH |
1398 | if (err) |
1399 | *err = errno; | |
8fd2cb40 JS |
1400 | return NULL; |
1401 | } | |
405e5b2f | 1402 | |
4bd5b7da | 1403 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1404 | if (changed_ret) |
1405 | *changed_ret = changed; | |
405e5b2f | 1406 | if (!changed) { |
4bd5b7da JH |
1407 | /* |
1408 | * The path is unchanged. If we were told to ignore | |
1409 | * valid bit, then we did the actual stat check and | |
1410 | * found that the entry is unmodified. If the entry | |
1411 | * is not marked VALID, this is the place to mark it | |
1412 | * valid again, under "assume unchanged" mode. | |
1413 | */ | |
1414 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1415 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1416 | ; /* mark this one VALID again */ |
eadb5831 JH |
1417 | else { |
1418 | /* | |
1419 | * We do not mark the index itself "modified" | |
1420 | * because CE_UPTODATE flag is in-core only; | |
1421 | * we are not going to write this change out. | |
1422 | */ | |
883e248b | 1423 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1424 | ce_mark_uptodate(ce); |
883e248b BP |
1425 | mark_fsmonitor_valid(ce); |
1426 | } | |
8fd2cb40 | 1427 | return ce; |
eadb5831 | 1428 | } |
405e5b2f LT |
1429 | } |
1430 | ||
4bd5b7da | 1431 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1432 | if (err) |
1433 | *err = EINVAL; | |
8fd2cb40 JS |
1434 | return NULL; |
1435 | } | |
405e5b2f | 1436 | |
a849735b | 1437 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1438 | copy_cache_entry(updated, ce); |
1439 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
405e5b2f | 1440 | fill_stat_cache_info(updated, &st); |
4bd5b7da JH |
1441 | /* |
1442 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1443 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1444 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1445 | * automatically, which is not really what we want. | |
405e5b2f | 1446 | */ |
4bd5b7da | 1447 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1448 | !(ce->ce_flags & CE_VALID)) |
1449 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1450 | |
e636a7b4 | 1451 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1452 | return updated; |
1453 | } | |
1454 | ||
3deffc52 | 1455 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1456 | int * first, const char *header_msg) |
3deffc52 MM |
1457 | { |
1458 | if (in_porcelain && *first && header_msg) { | |
1459 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1460 | *first = 0; |
3deffc52 MM |
1461 | } |
1462 | printf(fmt, name); | |
1463 | } | |
1464 | ||
9b2d6149 NTND |
1465 | int refresh_index(struct index_state *istate, unsigned int flags, |
1466 | const struct pathspec *pathspec, | |
046613c5 | 1467 | char *seen, const char *header_msg) |
405e5b2f LT |
1468 | { |
1469 | int i; | |
1470 | int has_errors = 0; | |
1471 | int really = (flags & REFRESH_REALLY) != 0; | |
1472 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1473 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1474 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1475 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
3deffc52 MM |
1476 | int first = 1; |
1477 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1478 | unsigned int options = (CE_MATCH_REFRESH | |
1479 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1480 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1481 | const char *modified_fmt; |
73b7eae6 JK |
1482 | const char *deleted_fmt; |
1483 | const char *typechange_fmt; | |
1484 | const char *added_fmt; | |
4bd4e730 | 1485 | const char *unmerged_fmt; |
405e5b2f | 1486 | |
c46c406a | 1487 | trace_performance_enter(); |
4bd4e730 | 1488 | modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n"); |
73b7eae6 JK |
1489 | deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n"); |
1490 | typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n"); | |
1491 | added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n"); | |
4bd4e730 | 1492 | unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n"); |
4aab5b46 | 1493 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1494 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1495 | int cache_errno = 0; |
73b7eae6 | 1496 | int changed = 0; |
3d1f148c | 1497 | int filtered = 0; |
ec0cc704 | 1498 | |
4aab5b46 | 1499 | ce = istate->cache[i]; |
5fdeacb0 JS |
1500 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1501 | continue; | |
1502 | ||
d7b665c3 | 1503 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1504 | filtered = 1; |
1505 | ||
405e5b2f | 1506 | if (ce_stage(ce)) { |
4aab5b46 JH |
1507 | while ((i < istate->cache_nr) && |
1508 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1509 | i++; |
1510 | i--; | |
1511 | if (allow_unmerged) | |
1512 | continue; | |
3d1f148c JH |
1513 | if (!filtered) |
1514 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1515 | &first, header_msg); | |
405e5b2f LT |
1516 | has_errors = 1; |
1517 | continue; | |
1518 | } | |
1519 | ||
3d1f148c | 1520 | if (filtered) |
d616813d AJ |
1521 | continue; |
1522 | ||
285c2e25 BW |
1523 | new_entry = refresh_cache_ent(istate, ce, options, &cache_errno, &changed); |
1524 | if (new_entry == ce) | |
405e5b2f | 1525 | continue; |
285c2e25 | 1526 | if (!new_entry) { |
73b7eae6 JK |
1527 | const char *fmt; |
1528 | ||
8fd2cb40 | 1529 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1530 | /* If we are doing --really-refresh that |
1531 | * means the index is not valid anymore. | |
1532 | */ | |
7a51ed66 | 1533 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1534 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1535 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1536 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1537 | } |
1538 | if (quiet) | |
1539 | continue; | |
73b7eae6 JK |
1540 | |
1541 | if (cache_errno == ENOENT) | |
1542 | fmt = deleted_fmt; | |
895ff3b2 | 1543 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1544 | fmt = added_fmt; /* must be before other checks */ |
1545 | else if (changed & TYPE_CHANGED) | |
1546 | fmt = typechange_fmt; | |
1547 | else | |
1548 | fmt = modified_fmt; | |
1549 | show_file(fmt, | |
1550 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1551 | has_errors = 1; |
1552 | continue; | |
1553 | } | |
cf558704 | 1554 | |
285c2e25 | 1555 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1556 | } |
c46c406a | 1557 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1558 | return has_errors; |
1559 | } | |
1560 | ||
768d7965 JM |
1561 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1562 | struct cache_entry *ce, | |
1563 | unsigned int options) | |
ec0cc704 | 1564 | { |
768d7965 | 1565 | return refresh_cache_ent(istate, ce, options, NULL, NULL); |
ec0cc704 JH |
1566 | } |
1567 | ||
db3b313c JH |
1568 | |
1569 | /***************************************************************** | |
1570 | * Index File I/O | |
1571 | *****************************************************************/ | |
1572 | ||
9d227781 JH |
1573 | #define INDEX_FORMAT_DEFAULT 3 |
1574 | ||
136347d7 TG |
1575 | static unsigned int get_index_format_default(void) |
1576 | { | |
1577 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 | 1578 | char *endp; |
b27a5720 | 1579 | int value; |
3c09d684 TG |
1580 | unsigned int version = INDEX_FORMAT_DEFAULT; |
1581 | ||
136347d7 | 1582 | if (!envversion) { |
b27a5720 TA |
1583 | if (!git_config_get_int("index.version", &value)) |
1584 | version = value; | |
3c09d684 TG |
1585 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1586 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1587 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1588 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1589 | } |
1590 | return version; | |
1591 | } | |
3c09d684 TG |
1592 | |
1593 | version = strtoul(envversion, &endp, 10); | |
1594 | if (*endp || | |
1595 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1596 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1597 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1598 | version = INDEX_FORMAT_DEFAULT; | |
1599 | } | |
1600 | return version; | |
136347d7 TG |
1601 | } |
1602 | ||
db3b313c JH |
1603 | /* |
1604 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1605 | * Again - this is just a (very strong in practice) heuristic that | |
1606 | * the inode hasn't changed. | |
1607 | * | |
1608 | * We save the fields in big-endian order to allow using the | |
1609 | * index file over NFS transparently. | |
1610 | */ | |
1611 | struct ondisk_cache_entry { | |
1612 | struct cache_time ctime; | |
1613 | struct cache_time mtime; | |
7800c1eb TG |
1614 | uint32_t dev; |
1615 | uint32_t ino; | |
1616 | uint32_t mode; | |
1617 | uint32_t uid; | |
1618 | uint32_t gid; | |
1619 | uint32_t size; | |
db3b313c | 1620 | unsigned char sha1[20]; |
7800c1eb | 1621 | uint16_t flags; |
db3b313c JH |
1622 | char name[FLEX_ARRAY]; /* more */ |
1623 | }; | |
1624 | ||
1625 | /* | |
1626 | * This struct is used when CE_EXTENDED bit is 1 | |
1627 | * The struct must match ondisk_cache_entry exactly from | |
1628 | * ctime till flags | |
1629 | */ | |
1630 | struct ondisk_cache_entry_extended { | |
1631 | struct cache_time ctime; | |
1632 | struct cache_time mtime; | |
7800c1eb TG |
1633 | uint32_t dev; |
1634 | uint32_t ino; | |
1635 | uint32_t mode; | |
1636 | uint32_t uid; | |
1637 | uint32_t gid; | |
1638 | uint32_t size; | |
db3b313c | 1639 | unsigned char sha1[20]; |
7800c1eb TG |
1640 | uint16_t flags; |
1641 | uint16_t flags2; | |
db3b313c JH |
1642 | char name[FLEX_ARRAY]; /* more */ |
1643 | }; | |
1644 | ||
6c9cd161 | 1645 | /* These are only used for v3 or lower */ |
ce012deb | 1646 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
db3b313c JH |
1647 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,name) + (len) + 8) & ~7) |
1648 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) | |
1649 | #define ondisk_cache_entry_extended_size(len) align_flex_name(ondisk_cache_entry_extended,len) | |
1650 | #define ondisk_ce_size(ce) (((ce)->ce_flags & CE_EXTENDED) ? \ | |
1651 | ondisk_cache_entry_extended_size(ce_namelen(ce)) : \ | |
1652 | ondisk_cache_entry_size(ce_namelen(ce))) | |
1653 | ||
a33fc72f JH |
1654 | /* Allow fsck to force verification of the index checksum. */ |
1655 | int verify_index_checksum; | |
1656 | ||
00ec50e5 BP |
1657 | /* Allow fsck to force verification of the cache entry order. */ |
1658 | int verify_ce_order; | |
1659 | ||
371ed0de | 1660 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1661 | { |
aab61359 | 1662 | git_hash_ctx c; |
1663 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1664 | int hdr_version; |
e83c5163 | 1665 | |
ccc4feb5 | 1666 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
e83c5163 | 1667 | return error("bad signature"); |
0136bac9 | 1668 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1669 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
0136bac9 | 1670 | return error("bad index version %d", hdr_version); |
a33fc72f JH |
1671 | |
1672 | if (!verify_index_checksum) | |
1673 | return 0; | |
1674 | ||
aab61359 | 1675 | the_hash_algo->init_fn(&c); |
1676 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1677 | the_hash_algo->final_fn(hash, &c); | |
67947c34 | 1678 | if (!hasheq(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz)) |
ca9be054 | 1679 | return error("bad index file sha1 signature"); |
e83c5163 LT |
1680 | return 0; |
1681 | } | |
1682 | ||
4aab5b46 | 1683 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1684 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1685 | { |
1686 | switch (CACHE_EXT(ext)) { | |
1687 | case CACHE_EXT_TREE: | |
4aab5b46 | 1688 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1689 | break; |
cfc5789a JH |
1690 | case CACHE_EXT_RESOLVE_UNDO: |
1691 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1692 | break; | |
5fc2fc8f NTND |
1693 | case CACHE_EXT_LINK: |
1694 | if (read_link_extension(istate, data, sz)) | |
1695 | return -1; | |
1696 | break; | |
f9e6c649 NTND |
1697 | case CACHE_EXT_UNTRACKED: |
1698 | istate->untracked = read_untracked_extension(data, sz); | |
1699 | break; | |
883e248b BP |
1700 | case CACHE_EXT_FSMONITOR: |
1701 | read_fsmonitor_extension(istate, data, sz); | |
1702 | break; | |
3b1d9e04 | 1703 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1704 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1705 | /* already handled in do_read_index() */ |
1706 | break; | |
bad68ec9 JH |
1707 | default: |
1708 | if (*ext < 'A' || 'Z' < *ext) | |
1709 | return error("index uses %.4s extension, which we do not understand", | |
1710 | ext); | |
1711 | fprintf(stderr, "ignoring %.4s extension\n", ext); | |
1712 | break; | |
1713 | } | |
1714 | return 0; | |
1715 | } | |
1716 | ||
b3e83cc7 | 1717 | int hold_locked_index(struct lock_file *lk, int lock_flags) |
216aab1e | 1718 | { |
b3e83cc7 | 1719 | return hold_lock_file_for_update(lk, get_index_file(), lock_flags); |
216aab1e MH |
1720 | } |
1721 | ||
4aab5b46 | 1722 | int read_index(struct index_state *istate) |
8fd2cb40 | 1723 | { |
a125a223 | 1724 | return read_index_from(istate, get_index_file(), get_git_dir()); |
8fd2cb40 JS |
1725 | } |
1726 | ||
77ff1127 BP |
1727 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1728 | unsigned int version, | |
a849735b | 1729 | struct ondisk_cache_entry *ondisk, |
6c9cd161 | 1730 | unsigned long *ent_size, |
252d079c | 1731 | const struct cache_entry *previous_ce) |
7a51ed66 | 1732 | { |
debed2a6 | 1733 | struct cache_entry *ce; |
7fec10b7 | 1734 | size_t len; |
06aaaa0b | 1735 | const char *name; |
debed2a6 | 1736 | unsigned int flags; |
252d079c NTND |
1737 | size_t copy_len; |
1738 | /* | |
1739 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1740 | * sense to only store the differences in later entries. In the v4 | |
1741 | * on-disk format of the index, each on-disk cache entry stores the | |
1742 | * number of bytes to be stripped from the end of the previous name, | |
1743 | * and the bytes to append to the result, to come up with its name. | |
1744 | */ | |
77ff1127 | 1745 | int expand_name_field = version == 4; |
7fec10b7 | 1746 | |
7a51ed66 | 1747 | /* On-disk flags are just 16 bits */ |
c3d8da57 | 1748 | flags = get_be16(&ondisk->flags); |
debed2a6 | 1749 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1750 | |
debed2a6 | 1751 | if (flags & CE_EXTENDED) { |
06aaaa0b NTND |
1752 | struct ondisk_cache_entry_extended *ondisk2; |
1753 | int extended_flags; | |
1754 | ondisk2 = (struct ondisk_cache_entry_extended *)ondisk; | |
c3d8da57 | 1755 | extended_flags = get_be16(&ondisk2->flags2) << 16; |
06aaaa0b NTND |
1756 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1757 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
1758 | die("Unknown index entry format %08x", extended_flags); | |
debed2a6 | 1759 | flags |= extended_flags; |
06aaaa0b NTND |
1760 | name = ondisk2->name; |
1761 | } | |
1762 | else | |
1763 | name = ondisk->name; | |
1764 | ||
252d079c NTND |
1765 | if (expand_name_field) { |
1766 | const unsigned char *cp = (const unsigned char *)name; | |
1767 | size_t strip_len, previous_len; | |
6c9cd161 | 1768 | |
77ff1127 | 1769 | /* If we're at the begining of a block, ignore the previous name */ |
252d079c | 1770 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1771 | if (previous_ce) { |
1772 | previous_len = previous_ce->ce_namelen; | |
1773 | if (previous_len < strip_len) | |
252d079c | 1774 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1775 | previous_ce->name); |
1776 | copy_len = previous_len - strip_len; | |
1777 | } else { | |
1778 | copy_len = 0; | |
252d079c | 1779 | } |
252d079c NTND |
1780 | name = (const char *)cp; |
1781 | } | |
6c9cd161 | 1782 | |
252d079c NTND |
1783 | if (len == CE_NAMEMASK) { |
1784 | len = strlen(name); | |
1785 | if (expand_name_field) | |
1786 | len += copy_len; | |
1787 | } | |
6c9cd161 | 1788 | |
77ff1127 | 1789 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c NTND |
1790 | |
1791 | ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec); | |
1792 | ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec); | |
1793 | ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec); | |
1794 | ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec); | |
1795 | ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev); | |
1796 | ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino); | |
1797 | ce->ce_mode = get_be32(&ondisk->mode); | |
1798 | ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid); | |
1799 | ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid); | |
1800 | ce->ce_stat_data.sd_size = get_be32(&ondisk->size); | |
1801 | ce->ce_flags = flags & ~CE_NAMEMASK; | |
1802 | ce->ce_namelen = len; | |
1803 | ce->index = 0; | |
1804 | hashcpy(ce->oid.hash, ondisk->sha1); | |
6c9cd161 | 1805 | |
252d079c NTND |
1806 | if (expand_name_field) { |
1807 | if (copy_len) | |
1808 | memcpy(ce->name, previous_ce->name, copy_len); | |
1809 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1810 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1811 | } else { | |
1812 | memcpy(ce->name, name, len + 1); | |
1813 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1814 | } |
debed2a6 | 1815 | return ce; |
cf558704 LT |
1816 | } |
1817 | ||
03f15a79 | 1818 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1819 | { |
03f15a79 TG |
1820 | unsigned int i; |
1821 | ||
00ec50e5 BP |
1822 | if (!verify_ce_order) |
1823 | return; | |
1824 | ||
03f15a79 TG |
1825 | for (i = 1; i < istate->cache_nr; i++) { |
1826 | struct cache_entry *ce = istate->cache[i - 1]; | |
1827 | struct cache_entry *next_ce = istate->cache[i]; | |
1828 | int name_compare = strcmp(ce->name, next_ce->name); | |
1829 | ||
1830 | if (0 < name_compare) | |
1831 | die("unordered stage entries in index"); | |
1832 | if (!name_compare) { | |
1833 | if (!ce_stage(ce)) | |
1834 | die("multiple stage entries for merged file '%s'", | |
1835 | ce->name); | |
1836 | if (ce_stage(ce) > ce_stage(next_ce)) | |
1837 | die("unordered stage entries for '%s'", | |
1838 | ce->name); | |
1839 | } | |
15999d0b JSP |
1840 | } |
1841 | } | |
1842 | ||
435ec090 CC |
1843 | static void tweak_untracked_cache(struct index_state *istate) |
1844 | { | |
1845 | switch (git_config_get_untracked_cache()) { | |
1846 | case -1: /* keep: do nothing */ | |
1847 | break; | |
1848 | case 0: /* false */ | |
1849 | remove_untracked_cache(istate); | |
1850 | break; | |
1851 | case 1: /* true */ | |
1852 | add_untracked_cache(istate); | |
1853 | break; | |
1854 | default: /* unknown value: do nothing */ | |
1855 | break; | |
1856 | } | |
1857 | } | |
1858 | ||
43925312 CC |
1859 | static void tweak_split_index(struct index_state *istate) |
1860 | { | |
1861 | switch (git_config_get_split_index()) { | |
1862 | case -1: /* unset: do nothing */ | |
1863 | break; | |
1864 | case 0: /* false */ | |
1865 | remove_split_index(istate); | |
1866 | break; | |
1867 | case 1: /* true */ | |
1868 | add_split_index(istate); | |
1869 | break; | |
1870 | default: /* unknown value: do nothing */ | |
1871 | break; | |
1872 | } | |
1873 | } | |
1874 | ||
435ec090 CC |
1875 | static void post_read_index_from(struct index_state *istate) |
1876 | { | |
1877 | check_ce_order(istate); | |
1878 | tweak_untracked_cache(istate); | |
43925312 | 1879 | tweak_split_index(istate); |
883e248b | 1880 | tweak_fsmonitor(istate); |
435ec090 CC |
1881 | } |
1882 | ||
8e72d675 JM |
1883 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
1884 | { | |
1885 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
1886 | } | |
1887 | ||
1888 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
1889 | { | |
1890 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
1891 | ||
1892 | /* | |
1893 | * Account for potential alignment differences. | |
1894 | */ | |
1895 | per_entry += align_padding_size(sizeof(struct cache_entry), -sizeof(struct ondisk_cache_entry)); | |
1896 | return ondisk_size + entries * per_entry; | |
1897 | } | |
1898 | ||
3255089a BP |
1899 | struct index_entry_offset |
1900 | { | |
1901 | /* starting byte offset into index file, count of index entries in this block */ | |
1902 | int offset, nr; | |
1903 | }; | |
1904 | ||
1905 | struct index_entry_offset_table | |
1906 | { | |
1907 | int nr; | |
1908 | struct index_entry_offset entries[FLEX_ARRAY]; | |
1909 | }; | |
1910 | ||
1911 | #ifndef NO_PTHREADS | |
1912 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); | |
1913 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
1914 | #endif | |
1915 | ||
3b1d9e04 BP |
1916 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
1917 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
1918 | ||
abb4bb83 BP |
1919 | struct load_index_extensions |
1920 | { | |
1921 | #ifndef NO_PTHREADS | |
1922 | pthread_t pthread; | |
1923 | #endif | |
1924 | struct index_state *istate; | |
1925 | const char *mmap; | |
1926 | size_t mmap_size; | |
1927 | unsigned long src_offset; | |
1928 | }; | |
1929 | ||
1930 | static void *load_index_extensions(void *_data) | |
1931 | { | |
1932 | struct load_index_extensions *p = _data; | |
1933 | unsigned long src_offset = p->src_offset; | |
1934 | ||
1935 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
1936 | /* After an array of active_nr index entries, | |
1937 | * there can be arbitrary number of extended | |
1938 | * sections, each of which is prefixed with | |
1939 | * extension name (4-byte) and section length | |
1940 | * in 4-byte network byte order. | |
1941 | */ | |
1942 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
1943 | if (read_index_extension(p->istate, | |
1944 | p->mmap + src_offset, | |
1945 | p->mmap + src_offset + 8, | |
1946 | extsize) < 0) { | |
1947 | munmap((void *)p->mmap, p->mmap_size); | |
1948 | die(_("index file corrupt")); | |
1949 | } | |
1950 | src_offset += 8; | |
1951 | src_offset += extsize; | |
1952 | } | |
1953 | ||
1954 | return NULL; | |
1955 | } | |
1956 | ||
77ff1127 BP |
1957 | /* |
1958 | * A helper function that will load the specified range of cache entries | |
1959 | * from the memory mapped file and add them to the given index. | |
1960 | */ | |
1961 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
1962 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
1963 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
1964 | { | |
1965 | int i; | |
1966 | unsigned long src_offset = start_offset; | |
1967 | ||
1968 | for (i = offset; i < offset + nr; i++) { | |
1969 | struct ondisk_cache_entry *disk_ce; | |
1970 | struct cache_entry *ce; | |
1971 | unsigned long consumed; | |
1972 | ||
1973 | disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset); | |
1974 | ce = create_from_disk(ce_mem_pool, istate->version, disk_ce, &consumed, previous_ce); | |
1975 | set_index_entry(istate, i, ce); | |
1976 | ||
1977 | src_offset += consumed; | |
1978 | previous_ce = ce; | |
1979 | } | |
1980 | return src_offset - start_offset; | |
1981 | } | |
1982 | ||
1983 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
1984 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
1985 | { | |
1986 | unsigned long consumed; | |
1987 | ||
1988 | if (istate->version == 4) { | |
1989 | mem_pool_init(&istate->ce_mem_pool, | |
1990 | estimate_cache_size_from_compressed(istate->cache_nr)); | |
1991 | } else { | |
1992 | mem_pool_init(&istate->ce_mem_pool, | |
1993 | estimate_cache_size(mmap_size, istate->cache_nr)); | |
1994 | } | |
1995 | ||
1996 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
1997 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
1998 | return consumed; | |
1999 | } | |
2000 | ||
2001 | #ifndef NO_PTHREADS | |
2002 | ||
3255089a BP |
2003 | /* |
2004 | * Mostly randomly chosen maximum thread counts: we | |
2005 | * cap the parallelism to online_cpus() threads, and we want | |
2006 | * to have at least 10000 cache entries per thread for it to | |
2007 | * be worth starting a thread. | |
2008 | */ | |
2009 | ||
2010 | #define THREAD_COST (10000) | |
2011 | ||
77ff1127 BP |
2012 | struct load_cache_entries_thread_data |
2013 | { | |
2014 | pthread_t pthread; | |
2015 | struct index_state *istate; | |
2016 | struct mem_pool *ce_mem_pool; | |
2017 | int offset; | |
2018 | const char *mmap; | |
2019 | struct index_entry_offset_table *ieot; | |
2020 | int ieot_start; /* starting index into the ieot array */ | |
2021 | int ieot_blocks; /* count of ieot entries to process */ | |
2022 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2023 | }; | |
2024 | ||
2025 | /* | |
2026 | * A thread proc to run the load_cache_entries() computation | |
2027 | * across multiple background threads. | |
2028 | */ | |
2029 | static void *load_cache_entries_thread(void *_data) | |
2030 | { | |
2031 | struct load_cache_entries_thread_data *p = _data; | |
2032 | int i; | |
2033 | ||
2034 | /* iterate across all ieot blocks assigned to this thread */ | |
2035 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2036 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2037 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2038 | p->offset += p->ieot->entries[i].nr; | |
2039 | } | |
2040 | return NULL; | |
2041 | } | |
2042 | ||
2043 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
2044 | unsigned long src_offset, int nr_threads, struct index_entry_offset_table *ieot) | |
2045 | { | |
2046 | int i, offset, ieot_blocks, ieot_start, err; | |
2047 | struct load_cache_entries_thread_data *data; | |
2048 | unsigned long consumed = 0; | |
2049 | ||
2050 | /* a little sanity checking */ | |
2051 | if (istate->name_hash_initialized) | |
2052 | BUG("the name hash isn't thread safe"); | |
2053 | ||
2054 | mem_pool_init(&istate->ce_mem_pool, 0); | |
2055 | ||
2056 | /* ensure we have no more threads than we have blocks to process */ | |
2057 | if (nr_threads > ieot->nr) | |
2058 | nr_threads = ieot->nr; | |
2059 | data = xcalloc(nr_threads, sizeof(*data)); | |
2060 | ||
2061 | offset = ieot_start = 0; | |
2062 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2063 | for (i = 0; i < nr_threads; i++) { | |
2064 | struct load_cache_entries_thread_data *p = &data[i]; | |
2065 | int nr, j; | |
2066 | ||
2067 | if (ieot_start + ieot_blocks > ieot->nr) | |
2068 | ieot_blocks = ieot->nr - ieot_start; | |
2069 | ||
2070 | p->istate = istate; | |
2071 | p->offset = offset; | |
2072 | p->mmap = mmap; | |
2073 | p->ieot = ieot; | |
2074 | p->ieot_start = ieot_start; | |
2075 | p->ieot_blocks = ieot_blocks; | |
2076 | ||
2077 | /* create a mem_pool for each thread */ | |
2078 | nr = 0; | |
2079 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2080 | nr += p->ieot->entries[j].nr; | |
2081 | if (istate->version == 4) { | |
2082 | mem_pool_init(&p->ce_mem_pool, | |
2083 | estimate_cache_size_from_compressed(nr)); | |
2084 | } else { | |
2085 | mem_pool_init(&p->ce_mem_pool, | |
2086 | estimate_cache_size(mmap_size, nr)); | |
2087 | } | |
2088 | ||
2089 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2090 | if (err) | |
2091 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2092 | ||
2093 | /* increment by the number of cache entries in the ieot block being processed */ | |
2094 | for (j = 0; j < ieot_blocks; j++) | |
2095 | offset += ieot->entries[ieot_start + j].nr; | |
2096 | ieot_start += ieot_blocks; | |
2097 | } | |
2098 | ||
2099 | for (i = 0; i < nr_threads; i++) { | |
2100 | struct load_cache_entries_thread_data *p = &data[i]; | |
2101 | ||
2102 | err = pthread_join(p->pthread, NULL); | |
2103 | if (err) | |
2104 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2105 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2106 | consumed += p->consumed; | |
2107 | } | |
2108 | ||
2109 | free(data); | |
2110 | ||
2111 | return consumed; | |
2112 | } | |
2113 | #endif | |
2114 | ||
8fd2cb40 | 2115 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2116 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2117 | { |
77ff1127 | 2118 | int fd; |
e83c5163 | 2119 | struct stat st; |
debed2a6 | 2120 | unsigned long src_offset; |
371ed0de BP |
2121 | const struct cache_header *hdr; |
2122 | const char *mmap; | |
7a51ed66 | 2123 | size_t mmap_size; |
abb4bb83 BP |
2124 | struct load_index_extensions p; |
2125 | size_t extension_offset = 0; | |
2126 | #ifndef NO_PTHREADS | |
77ff1127 BP |
2127 | int nr_threads, cpus; |
2128 | struct index_entry_offset_table *ieot = NULL; | |
abb4bb83 | 2129 | #endif |
e83c5163 | 2130 | |
913e0e99 | 2131 | if (istate->initialized) |
4aab5b46 | 2132 | return istate->cache_nr; |
5d1a5c02 | 2133 | |
fba2f38a KB |
2134 | istate->timestamp.sec = 0; |
2135 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2136 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2137 | if (fd < 0) { |
5fc2fc8f | 2138 | if (!must_exist && errno == ENOENT) |
5d1a5c02 | 2139 | return 0; |
5fc2fc8f | 2140 | die_errno("%s: index file open failed", path); |
5d1a5c02 | 2141 | } |
e83c5163 | 2142 | |
3511a377 | 2143 | if (fstat(fd, &st)) |
d824cbba | 2144 | die_errno("cannot stat the open index"); |
3511a377 | 2145 | |
7a51ed66 | 2146 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2147 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
3511a377 LFC |
2148 | die("index file smaller than expected"); |
2149 | ||
a1293ef7 | 2150 | mmap = xmmap(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2151 | if (mmap == MAP_FAILED) |
0721c314 | 2152 | die_errno("unable to map index file"); |
57d84f8d | 2153 | close(fd); |
e83c5163 | 2154 | |
371ed0de | 2155 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2156 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2157 | goto unmap; |
2158 | ||
75691ea3 | 2159 | hashcpy(istate->oid.hash, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2160 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2161 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2162 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
c4aa3167 | 2163 | istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache)); |
913e0e99 | 2164 | istate->initialized = 1; |
7a51ed66 | 2165 | |
abb4bb83 BP |
2166 | p.istate = istate; |
2167 | p.mmap = mmap; | |
2168 | p.mmap_size = mmap_size; | |
6c9cd161 | 2169 | |
7a51ed66 | 2170 | src_offset = sizeof(*hdr); |
4aab5b46 | 2171 | |
abb4bb83 BP |
2172 | #ifndef NO_PTHREADS |
2173 | nr_threads = git_config_get_index_threads(); | |
7a51ed66 | 2174 | |
77ff1127 BP |
2175 | /* TODO: does creating more threads than cores help? */ |
2176 | if (!nr_threads) { | |
2177 | nr_threads = istate->cache_nr / THREAD_COST; | |
2178 | cpus = online_cpus(); | |
2179 | if (nr_threads > cpus) | |
2180 | nr_threads = cpus; | |
e83c5163 | 2181 | } |
abb4bb83 BP |
2182 | |
2183 | if (nr_threads > 1) { | |
2184 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2185 | if (extension_offset) { | |
2186 | int err; | |
2187 | ||
2188 | p.src_offset = extension_offset; | |
2189 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2190 | if (err) | |
2191 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2192 | ||
2193 | nr_threads--; | |
2194 | } | |
2195 | } | |
abb4bb83 | 2196 | |
77ff1127 BP |
2197 | /* |
2198 | * Locate and read the index entry offset table so that we can use it | |
2199 | * to multi-thread the reading of the cache entries. | |
2200 | */ | |
2201 | if (extension_offset && nr_threads > 1) | |
2202 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2203 | ||
2204 | if (ieot) { | |
2205 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, src_offset, nr_threads, ieot); | |
2206 | free(ieot); | |
8e72d675 | 2207 | } else { |
77ff1127 | 2208 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2209 | } |
77ff1127 BP |
2210 | #else |
2211 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); | |
2212 | #endif | |
6c9cd161 | 2213 | |
fba2f38a | 2214 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2215 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2216 | |
abb4bb83 BP |
2217 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
2218 | #ifndef NO_PTHREADS | |
2219 | if (extension_offset) { | |
2220 | int ret = pthread_join(p.pthread, NULL); | |
2221 | if (ret) | |
2222 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
2223 | } | |
2224 | #endif | |
2225 | if (!extension_offset) { | |
2226 | p.src_offset = src_offset; | |
2227 | load_index_extensions(&p); | |
bad68ec9 | 2228 | } |
371ed0de | 2229 | munmap((void *)mmap, mmap_size); |
4aab5b46 | 2230 | return istate->cache_nr; |
e83c5163 LT |
2231 | |
2232 | unmap: | |
371ed0de | 2233 | munmap((void *)mmap, mmap_size); |
5d1a5c02 | 2234 | die("index file corrupt"); |
e83c5163 LT |
2235 | } |
2236 | ||
0d59ffb4 CC |
2237 | /* |
2238 | * Signal that the shared index is used by updating its mtime. | |
2239 | * | |
2240 | * This way, shared index can be removed if they have not been used | |
2241 | * for some time. | |
2242 | */ | |
a125a223 | 2243 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2244 | { |
0d59ffb4 CC |
2245 | if (!check_and_freshen_file(shared_index, 1) && warn) |
2246 | warning("could not freshen shared index '%s'", shared_index); | |
2247 | } | |
2248 | ||
a125a223 TG |
2249 | int read_index_from(struct index_state *istate, const char *path, |
2250 | const char *gitdir) | |
5fc2fc8f NTND |
2251 | { |
2252 | struct split_index *split_index; | |
2253 | int ret; | |
2182abd9 | 2254 | char *base_oid_hex; |
a125a223 | 2255 | char *base_path; |
5fc2fc8f NTND |
2256 | |
2257 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2258 | if (istate->initialized) | |
2259 | return istate->cache_nr; | |
2260 | ||
c46c406a | 2261 | trace_performance_enter(); |
5fc2fc8f | 2262 | ret = do_read_index(istate, path, 0); |
c46c406a | 2263 | trace_performance_leave("read cache %s", path); |
435ec090 | 2264 | |
5fc2fc8f | 2265 | split_index = istate->split_index; |
2182abd9 | 2266 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2267 | post_read_index_from(istate); |
5fc2fc8f | 2268 | return ret; |
03f15a79 | 2269 | } |
5fc2fc8f | 2270 | |
c46c406a | 2271 | trace_performance_enter(); |
5fc2fc8f NTND |
2272 | if (split_index->base) |
2273 | discard_index(split_index->base); | |
2274 | else | |
2275 | split_index->base = xcalloc(1, sizeof(*split_index->base)); | |
de6ae5f9 | 2276 | |
2182abd9 | 2277 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2278 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
de6ae5f9 | 2279 | ret = do_read_index(split_index->base, base_path, 1); |
9001dc2a | 2280 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
5fc2fc8f | 2281 | die("broken index, expect %s in %s, got %s", |
2182abd9 | 2282 | base_oid_hex, base_path, |
75691ea3 | 2283 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2284 | |
a125a223 | 2285 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2286 | merge_base_index(istate); |
435ec090 | 2287 | post_read_index_from(istate); |
a125a223 | 2288 | free(base_path); |
c46c406a | 2289 | trace_performance_leave("read cache %s", base_path); |
5fc2fc8f NTND |
2290 | return ret; |
2291 | } | |
2292 | ||
fa7b3c2f JH |
2293 | int is_index_unborn(struct index_state *istate) |
2294 | { | |
debed2a6 | 2295 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2296 | } |
2297 | ||
4aab5b46 | 2298 | int discard_index(struct index_state *istate) |
6d297f81 | 2299 | { |
8e72d675 JM |
2300 | /* |
2301 | * Cache entries in istate->cache[] should have been allocated | |
2302 | * from the memory pool associated with this index, or from an | |
2303 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2304 | * cache entries. validate_cache_entries can detect when this |
2305 | * assertion does not hold. | |
8e72d675 | 2306 | */ |
8616a2d0 | 2307 | validate_cache_entries(istate); |
debed2a6 | 2308 | |
cfc5789a | 2309 | resolve_undo_clear_index(istate); |
4aab5b46 JH |
2310 | istate->cache_nr = 0; |
2311 | istate->cache_changed = 0; | |
fba2f38a KB |
2312 | istate->timestamp.sec = 0; |
2313 | istate->timestamp.nsec = 0; | |
2092678c | 2314 | free_name_hash(istate); |
4aab5b46 | 2315 | cache_tree_free(&(istate->cache_tree)); |
913e0e99 | 2316 | istate->initialized = 0; |
6a83d902 | 2317 | FREE_AND_NULL(istate->cache); |
a0fc4db0 | 2318 | istate->cache_alloc = 0; |
5fc2fc8f | 2319 | discard_split_index(istate); |
f9e6c649 NTND |
2320 | free_untracked_cache(istate->untracked); |
2321 | istate->untracked = NULL; | |
8e72d675 JM |
2322 | |
2323 | if (istate->ce_mem_pool) { | |
8616a2d0 | 2324 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
8e72d675 JM |
2325 | istate->ce_mem_pool = NULL; |
2326 | } | |
2327 | ||
7a51ed66 | 2328 | return 0; |
6d297f81 JS |
2329 | } |
2330 | ||
8616a2d0 JM |
2331 | /* |
2332 | * Validate the cache entries of this index. | |
2333 | * All cache entries associated with this index | |
2334 | * should have been allocated by the memory pool | |
2335 | * associated with this index, or by a referenced | |
2336 | * split index. | |
2337 | */ | |
2338 | void validate_cache_entries(const struct index_state *istate) | |
2339 | { | |
2340 | int i; | |
2341 | ||
2342 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2343 | return; | |
2344 | ||
2345 | for (i = 0; i < istate->cache_nr; i++) { | |
2346 | if (!istate) { | |
2347 | die("internal error: cache entry is not allocated from expected memory pool"); | |
2348 | } else if (!istate->ce_mem_pool || | |
2349 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2350 | if (!istate->split_index || | |
2351 | !istate->split_index->base || | |
2352 | !istate->split_index->base->ce_mem_pool || | |
2353 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
2354 | die("internal error: cache entry is not allocated from expected memory pool"); | |
2355 | } | |
2356 | } | |
2357 | } | |
2358 | ||
2359 | if (istate->split_index) | |
2360 | validate_cache_entries(istate->split_index->base); | |
2361 | } | |
2362 | ||
d1f128b0 | 2363 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2364 | { |
2365 | int i; | |
2366 | for (i = 0; i < istate->cache_nr; i++) { | |
2367 | if (ce_stage(istate->cache[i])) | |
2368 | return 1; | |
2369 | } | |
2370 | return 0; | |
2371 | } | |
2372 | ||
92a1bf5a | 2373 | int index_has_changes(struct index_state *istate, |
e1f8694f EN |
2374 | struct tree *tree, |
2375 | struct strbuf *sb) | |
cffbfad5 | 2376 | { |
e1f8694f | 2377 | struct object_id cmp; |
cffbfad5 EN |
2378 | int i; |
2379 | ||
1b9fbefb EN |
2380 | if (istate != &the_index) { |
2381 | BUG("index_has_changes cannot yet accept istate != &the_index; do_diff_cache needs updating first."); | |
2382 | } | |
e1f8694f EN |
2383 | if (tree) |
2384 | cmp = tree->object.oid; | |
2385 | if (tree || !get_oid_tree("HEAD", &cmp)) { | |
cffbfad5 EN |
2386 | struct diff_options opt; |
2387 | ||
e6757652 | 2388 | repo_diff_setup(the_repository, &opt); |
cffbfad5 EN |
2389 | opt.flags.exit_with_status = 1; |
2390 | if (!sb) | |
2391 | opt.flags.quick = 1; | |
e1f8694f | 2392 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2393 | diffcore_std(&opt); |
2394 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2395 | if (i) | |
2396 | strbuf_addch(sb, ' '); | |
2397 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2398 | } | |
2399 | diff_flush(&opt); | |
2400 | return opt.flags.has_changes != 0; | |
2401 | } else { | |
1b9fbefb | 2402 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2403 | if (i) |
2404 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2405 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2406 | } |
1b9fbefb | 2407 | return !!istate->cache_nr; |
cffbfad5 EN |
2408 | } |
2409 | } | |
2410 | ||
4990aadc | 2411 | #define WRITE_BUFFER_SIZE 8192 |
bf0f910d | 2412 | static unsigned char write_buffer[WRITE_BUFFER_SIZE]; |
4990aadc LT |
2413 | static unsigned long write_buffer_len; |
2414 | ||
aab61359 | 2415 | static int ce_write_flush(git_hash_ctx *context, int fd) |
6015c28b JH |
2416 | { |
2417 | unsigned int buffered = write_buffer_len; | |
2418 | if (buffered) { | |
aab61359 | 2419 | the_hash_algo->update_fn(context, write_buffer, buffered); |
06f46f23 | 2420 | if (write_in_full(fd, write_buffer, buffered) < 0) |
6015c28b JH |
2421 | return -1; |
2422 | write_buffer_len = 0; | |
2423 | } | |
2424 | return 0; | |
2425 | } | |
2426 | ||
aab61359 | 2427 | static int ce_write(git_hash_ctx *context, int fd, void *data, unsigned int len) |
4990aadc LT |
2428 | { |
2429 | while (len) { | |
2430 | unsigned int buffered = write_buffer_len; | |
2431 | unsigned int partial = WRITE_BUFFER_SIZE - buffered; | |
2432 | if (partial > len) | |
2433 | partial = len; | |
2434 | memcpy(write_buffer + buffered, data, partial); | |
2435 | buffered += partial; | |
2436 | if (buffered == WRITE_BUFFER_SIZE) { | |
6015c28b JH |
2437 | write_buffer_len = buffered; |
2438 | if (ce_write_flush(context, fd)) | |
4990aadc LT |
2439 | return -1; |
2440 | buffered = 0; | |
2441 | } | |
2442 | write_buffer_len = buffered; | |
2443 | len -= partial; | |
1d7f171c | 2444 | data = (char *) data + partial; |
a6080a0a JH |
2445 | } |
2446 | return 0; | |
4990aadc LT |
2447 | } |
2448 | ||
3b1d9e04 BP |
2449 | static int write_index_ext_header(git_hash_ctx *context, git_hash_ctx *eoie_context, |
2450 | int fd, unsigned int ext, unsigned int sz) | |
bad68ec9 JH |
2451 | { |
2452 | ext = htonl(ext); | |
2453 | sz = htonl(sz); | |
3b1d9e04 BP |
2454 | if (eoie_context) { |
2455 | the_hash_algo->update_fn(eoie_context, &ext, 4); | |
2456 | the_hash_algo->update_fn(eoie_context, &sz, 4); | |
2457 | } | |
968a1d65 DR |
2458 | return ((ce_write(context, fd, &ext, 4) < 0) || |
2459 | (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0; | |
bad68ec9 JH |
2460 | } |
2461 | ||
aab61359 | 2462 | static int ce_flush(git_hash_ctx *context, int fd, unsigned char *hash) |
4990aadc LT |
2463 | { |
2464 | unsigned int left = write_buffer_len; | |
ca9be054 | 2465 | |
4990aadc LT |
2466 | if (left) { |
2467 | write_buffer_len = 0; | |
aab61359 | 2468 | the_hash_algo->update_fn(context, write_buffer, left); |
4990aadc | 2469 | } |
ca9be054 | 2470 | |
aab61359 | 2471 | /* Flush first if not enough space for hash signature */ |
2472 | if (left + the_hash_algo->rawsz > WRITE_BUFFER_SIZE) { | |
06f46f23 | 2473 | if (write_in_full(fd, write_buffer, left) < 0) |
2c865d9a QH |
2474 | return -1; |
2475 | left = 0; | |
2476 | } | |
2477 | ||
aab61359 | 2478 | /* Append the hash signature at the end */ |
2479 | the_hash_algo->final_fn(write_buffer + left, context); | |
2480 | hashcpy(hash, write_buffer + left); | |
2481 | left += the_hash_algo->rawsz; | |
06f46f23 | 2482 | return (write_in_full(fd, write_buffer, left) < 0) ? -1 : 0; |
4990aadc LT |
2483 | } |
2484 | ||
58bf2a4c NTND |
2485 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2486 | struct cache_entry *ce) | |
407c8eb0 JH |
2487 | { |
2488 | /* | |
2489 | * The only thing we care about in this function is to smudge the | |
2490 | * falsely clean entry due to touch-update-touch race, so we leave | |
2491 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2492 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2493 | * |
2494 | * Note that this actually does not do much for gitlinks, for | |
2495 | * which ce_match_stat_basic() always goes to the actual | |
2496 | * contents. The caller checks with is_racy_timestamp() which | |
2497 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2498 | */ |
2499 | struct stat st; | |
2500 | ||
2501 | if (lstat(ce->name, &st) < 0) | |
2502 | return; | |
2503 | if (ce_match_stat_basic(ce, &st)) | |
2504 | return; | |
58bf2a4c | 2505 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2506 | /* This is "racily clean"; smudge it. Note that this |
2507 | * is a tricky code. At first glance, it may appear | |
2508 | * that it can break with this sequence: | |
2509 | * | |
2510 | * $ echo xyzzy >frotz | |
2511 | * $ git-update-index --add frotz | |
2512 | * $ : >frotz | |
2513 | * $ sleep 3 | |
2514 | * $ echo filfre >nitfol | |
2515 | * $ git-update-index --add nitfol | |
2516 | * | |
b7e58b17 | 2517 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2518 | * it notices that the entry "frotz" has the same timestamp |
2519 | * as index, and if we were to smudge it by resetting its | |
2520 | * size to zero here, then the object name recorded | |
2521 | * in index is the 6-byte file but the cached stat information | |
2522 | * becomes zero --- which would then match what we would | |
a6080a0a | 2523 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2524 | * |
2525 | * However, the second update-index, before calling | |
2526 | * this function, notices that the cached size is 6 | |
2527 | * bytes and what is on the filesystem is an empty | |
2528 | * file, and never calls us, so the cached size information | |
2529 | * for "frotz" stays 6 which does not match the filesystem. | |
2530 | */ | |
c21d39d7 | 2531 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2532 | } |
2533 | } | |
2534 | ||
f136f7bf | 2535 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2536 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2537 | struct cache_entry *ce) |
7a51ed66 | 2538 | { |
b60e188c TG |
2539 | short flags; |
2540 | ||
c21d39d7 MH |
2541 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2542 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2543 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2544 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2545 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2546 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2547 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2548 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2549 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2550 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
99d1a986 | 2551 | hashcpy(ondisk->sha1, ce->oid.hash); |
b60e188c | 2552 | |
ce51bf09 | 2553 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c TG |
2554 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
2555 | ondisk->flags = htons(flags); | |
06aaaa0b NTND |
2556 | if (ce->ce_flags & CE_EXTENDED) { |
2557 | struct ondisk_cache_entry_extended *ondisk2; | |
2558 | ondisk2 = (struct ondisk_cache_entry_extended *)ondisk; | |
2559 | ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); | |
f136f7bf JH |
2560 | } |
2561 | } | |
2562 | ||
aab61359 | 2563 | static int ce_write_entry(git_hash_ctx *c, int fd, struct cache_entry *ce, |
ce012deb | 2564 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2565 | { |
9d227781 | 2566 | int size; |
f136f7bf | 2567 | int result; |
00a4b035 RJ |
2568 | unsigned int saved_namelen; |
2569 | int stripped_name = 0; | |
ce012deb | 2570 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2571 | |
b3c96fb1 NTND |
2572 | if (ce->ce_flags & CE_STRIP_NAME) { |
2573 | saved_namelen = ce_namelen(ce); | |
2574 | ce->ce_namelen = 0; | |
00a4b035 | 2575 | stripped_name = 1; |
b3c96fb1 NTND |
2576 | } |
2577 | ||
ce012deb KW |
2578 | if (ce->ce_flags & CE_EXTENDED) |
2579 | size = offsetof(struct ondisk_cache_entry_extended, name); | |
2580 | else | |
2581 | size = offsetof(struct ondisk_cache_entry, name); | |
2582 | ||
9d227781 | 2583 | if (!previous_name) { |
ce012deb KW |
2584 | int len = ce_namelen(ce); |
2585 | copy_cache_entry_to_ondisk(ondisk, ce); | |
2586 | result = ce_write(c, fd, ondisk, size); | |
2587 | if (!result) | |
2588 | result = ce_write(c, fd, ce->name, len); | |
2589 | if (!result) | |
2590 | result = ce_write(c, fd, padding, align_padding_size(size, len)); | |
9d227781 JH |
2591 | } else { |
2592 | int common, to_remove, prefix_size; | |
2593 | unsigned char to_remove_vi[16]; | |
2594 | for (common = 0; | |
2595 | (ce->name[common] && | |
2596 | common < previous_name->len && | |
2597 | ce->name[common] == previous_name->buf[common]); | |
2598 | common++) | |
2599 | ; /* still matching */ | |
2600 | to_remove = previous_name->len - common; | |
2601 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2602 | ||
ce012deb KW |
2603 | copy_cache_entry_to_ondisk(ondisk, ce); |
2604 | result = ce_write(c, fd, ondisk, size); | |
2605 | if (!result) | |
2606 | result = ce_write(c, fd, to_remove_vi, prefix_size); | |
2607 | if (!result) | |
0b90b881 TG |
2608 | result = ce_write(c, fd, ce->name + common, ce_namelen(ce) - common); |
2609 | if (!result) | |
2610 | result = ce_write(c, fd, padding, 1); | |
9d227781 JH |
2611 | |
2612 | strbuf_splice(previous_name, common, to_remove, | |
2613 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2614 | } |
00a4b035 | 2615 | if (stripped_name) { |
b3c96fb1 NTND |
2616 | ce->ce_namelen = saved_namelen; |
2617 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2618 | } | |
7a51ed66 | 2619 | |
59efba64 | 2620 | return result; |
7a51ed66 LT |
2621 | } |
2622 | ||
426ddeea YM |
2623 | /* |
2624 | * This function verifies if index_state has the correct sha1 of the | |
2625 | * index file. Don't die if we have any other failure, just return 0. | |
2626 | */ | |
2627 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2628 | { | |
2629 | int fd; | |
2630 | ssize_t n; | |
2631 | struct stat st; | |
aab61359 | 2632 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2633 | |
2634 | if (!istate->initialized) | |
2635 | return 0; | |
2636 | ||
2637 | fd = open(path, O_RDONLY); | |
2638 | if (fd < 0) | |
2639 | return 0; | |
2640 | ||
2641 | if (fstat(fd, &st)) | |
2642 | goto out; | |
2643 | ||
aab61359 | 2644 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2645 | goto out; |
2646 | ||
aab61359 | 2647 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2648 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2649 | goto out; |
2650 | ||
67947c34 | 2651 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2652 | goto out; |
2653 | ||
2654 | close(fd); | |
2655 | return 1; | |
2656 | ||
2657 | out: | |
2658 | close(fd); | |
2659 | return 0; | |
2660 | } | |
2661 | ||
2662 | static int verify_index(const struct index_state *istate) | |
2663 | { | |
2664 | return verify_index_from(istate, get_index_file()); | |
2665 | } | |
2666 | ||
483fbe2b JH |
2667 | static int has_racy_timestamp(struct index_state *istate) |
2668 | { | |
2669 | int entries = istate->cache_nr; | |
2670 | int i; | |
2671 | ||
2672 | for (i = 0; i < entries; i++) { | |
2673 | struct cache_entry *ce = istate->cache[i]; | |
2674 | if (is_racy_timestamp(istate, ce)) | |
2675 | return 1; | |
2676 | } | |
2677 | return 0; | |
2678 | } | |
2679 | ||
ccdc4ec3 JH |
2680 | void update_index_if_able(struct index_state *istate, struct lock_file *lockfile) |
2681 | { | |
483fbe2b | 2682 | if ((istate->cache_changed || has_racy_timestamp(istate)) && |
df60cf57 MÅ |
2683 | verify_index(istate)) |
2684 | write_locked_index(istate, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2685 | else |
2686 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2687 | } |
2688 | ||
812d6b00 MÅ |
2689 | /* |
2690 | * On success, `tempfile` is closed. If it is the temporary file | |
2691 | * of a `struct lock_file`, we will therefore effectively perform | |
2692 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2693 | * detail of lockfiles, callers of `do_write_index()` should not | |
2694 | * rely on it. | |
2695 | */ | |
9f41c7a6 | 2696 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
c18b80a0 | 2697 | int strip_extensions) |
197ee8c9 | 2698 | { |
ca54d9ba | 2699 | uint64_t start = getnanotime(); |
9f41c7a6 | 2700 | int newfd = tempfile->fd; |
3b1d9e04 | 2701 | git_hash_ctx c, eoie_c; |
197ee8c9 | 2702 | struct cache_header hdr; |
b50386c7 | 2703 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2704 | struct cache_entry **cache = istate->cache; |
2705 | int entries = istate->cache_nr; | |
e1afca4f | 2706 | struct stat st; |
ce012deb | 2707 | struct ondisk_cache_entry_extended ondisk; |
9d227781 | 2708 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2709 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2710 | off_t offset; |
77ff1127 | 2711 | int ieot_entries = 1; |
3255089a BP |
2712 | struct index_entry_offset_table *ieot = NULL; |
2713 | int nr, nr_threads; | |
025a0709 | 2714 | |
06aaaa0b | 2715 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2716 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2717 | removed++; |
197ee8c9 | 2718 | |
06aaaa0b NTND |
2719 | /* reduce extended entries if possible */ |
2720 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2721 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2722 | extended++; | |
2723 | cache[i]->ce_flags |= CE_EXTENDED; | |
2724 | } | |
2725 | } | |
2726 | ||
d6e3c181 | 2727 | if (!istate->version) { |
136347d7 | 2728 | istate->version = get_index_format_default(); |
4c2db938 | 2729 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) |
d6e3c181 NTND |
2730 | init_split_index(istate); |
2731 | } | |
9d227781 JH |
2732 | |
2733 | /* demote version 3 to version 2 when the latter suffices */ | |
2734 | if (istate->version == 3 || istate->version == 2) | |
2735 | istate->version = extended ? 3 : 2; | |
2736 | ||
2737 | hdr_version = istate->version; | |
2738 | ||
ccc4feb5 | 2739 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2740 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2741 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2742 | |
aab61359 | 2743 | the_hash_algo->init_fn(&c); |
ca9be054 | 2744 | if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0) |
197ee8c9 LT |
2745 | return -1; |
2746 | ||
3255089a BP |
2747 | #ifndef NO_PTHREADS |
2748 | nr_threads = git_config_get_index_threads(); | |
2749 | if (nr_threads != 1) { | |
2750 | int ieot_blocks, cpus; | |
2751 | ||
2752 | /* | |
2753 | * ensure default number of ieot blocks maps evenly to the | |
2754 | * default number of threads that will process them leaving | |
2755 | * room for the thread to load the index extensions. | |
2756 | */ | |
2757 | if (!nr_threads) { | |
2758 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2759 | cpus = online_cpus(); | |
2760 | if (ieot_blocks > cpus - 1) | |
2761 | ieot_blocks = cpus - 1; | |
2762 | } else { | |
2763 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2764 | if (ieot_blocks > istate->cache_nr) |
2765 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2766 | } |
2767 | ||
2768 | /* | |
2769 | * no reason to write out the IEOT extension if we don't | |
2770 | * have enough blocks to utilize multi-threading | |
2771 | */ | |
2772 | if (ieot_blocks > 1) { | |
2773 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2774 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2775 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2776 | } |
2777 | } | |
2778 | #endif | |
2779 | ||
3b1d9e04 | 2780 | offset = lseek(newfd, 0, SEEK_CUR); |
3255089a BP |
2781 | if (offset < 0) { |
2782 | free(ieot); | |
3b1d9e04 | 2783 | return -1; |
3255089a | 2784 | } |
3b1d9e04 | 2785 | offset += write_buffer_len; |
3255089a | 2786 | nr = 0; |
9d227781 | 2787 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2788 | |
197ee8c9 LT |
2789 | for (i = 0; i < entries; i++) { |
2790 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2791 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2792 | continue; |
e06c43c7 | 2793 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2794 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2795 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2796 | static const char msg[] = "cache entry has null sha1: %s"; |
2797 | static int allow = -1; | |
2798 | ||
2799 | if (allow < 0) | |
2800 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2801 | if (allow) | |
2802 | warning(msg, ce->name); | |
2803 | else | |
b50386c7 | 2804 | err = error(msg, ce->name); |
a96d3cc3 JK |
2805 | |
2806 | drop_cache_tree = 1; | |
83bd7437 | 2807 | } |
77ff1127 | 2808 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2809 | ieot->entries[ieot->nr].nr = nr; |
2810 | ieot->entries[ieot->nr].offset = offset; | |
2811 | ieot->nr++; | |
2812 | /* | |
2813 | * If we have a V4 index, set the first byte to an invalid | |
2814 | * character to ensure there is nothing common with the previous | |
2815 | * entry | |
2816 | */ | |
2817 | if (previous_name) | |
2818 | previous_name->buf[0] = 0; | |
2819 | nr = 0; | |
2820 | offset = lseek(newfd, 0, SEEK_CUR); | |
2821 | if (offset < 0) { | |
2822 | free(ieot); | |
2823 | return -1; | |
2824 | } | |
2825 | offset += write_buffer_len; | |
2826 | } | |
ce012deb | 2827 | if (ce_write_entry(&c, newfd, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2828 | err = -1; |
2829 | ||
2830 | if (err) | |
2831 | break; | |
3255089a BP |
2832 | nr++; |
2833 | } | |
2834 | if (ieot && nr) { | |
2835 | ieot->entries[ieot->nr].nr = nr; | |
2836 | ieot->entries[ieot->nr].offset = offset; | |
2837 | ieot->nr++; | |
197ee8c9 | 2838 | } |
9d227781 | 2839 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2840 | |
3255089a BP |
2841 | if (err) { |
2842 | free(ieot); | |
b50386c7 | 2843 | return err; |
3255089a | 2844 | } |
b50386c7 | 2845 | |
bad68ec9 | 2846 | /* Write extension data here */ |
3b1d9e04 | 2847 | offset = lseek(newfd, 0, SEEK_CUR); |
3255089a BP |
2848 | if (offset < 0) { |
2849 | free(ieot); | |
3b1d9e04 | 2850 | return -1; |
3255089a | 2851 | } |
3b1d9e04 BP |
2852 | offset += write_buffer_len; |
2853 | the_hash_algo->init_fn(&eoie_c); | |
2854 | ||
3255089a BP |
2855 | /* |
2856 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
2857 | * can minimize the number of extensions we have to scan through to | |
2858 | * find it during load. Write it out regardless of the | |
2859 | * strip_extensions parameter as we need it when loading the shared | |
2860 | * index. | |
2861 | */ | |
2862 | #ifndef NO_PTHREADS | |
2863 | if (ieot) { | |
2864 | struct strbuf sb = STRBUF_INIT; | |
2865 | ||
2866 | write_ieot_extension(&sb, ieot); | |
2867 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0 | |
2868 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2869 | strbuf_release(&sb); | |
2870 | free(ieot); | |
2871 | if (err) | |
2872 | return -1; | |
2873 | } | |
2874 | #endif | |
2875 | ||
c18b80a0 | 2876 | if (!strip_extensions && istate->split_index) { |
5fc2fc8f NTND |
2877 | struct strbuf sb = STRBUF_INIT; |
2878 | ||
2879 | err = write_link_extension(&sb, istate) < 0 || | |
3b1d9e04 | 2880 | write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_LINK, |
5fc2fc8f NTND |
2881 | sb.len) < 0 || |
2882 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2883 | strbuf_release(&sb); | |
2884 | if (err) | |
2885 | return -1; | |
2886 | } | |
a96d3cc3 | 2887 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { |
f285a2d7 | 2888 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 2889 | |
1dffb8fa | 2890 | cache_tree_write(&sb, istate->cache_tree); |
3b1d9e04 | 2891 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_TREE, sb.len) < 0 |
1dffb8fa PH |
2892 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; |
2893 | strbuf_release(&sb); | |
2894 | if (err) | |
bad68ec9 | 2895 | return -1; |
bad68ec9 | 2896 | } |
c18b80a0 | 2897 | if (!strip_extensions && istate->resolve_undo) { |
cfc5789a JH |
2898 | struct strbuf sb = STRBUF_INIT; |
2899 | ||
2900 | resolve_undo_write(&sb, istate->resolve_undo); | |
3b1d9e04 | 2901 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_RESOLVE_UNDO, |
cfc5789a JH |
2902 | sb.len) < 0 |
2903 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2904 | strbuf_release(&sb); | |
2905 | if (err) | |
2906 | return -1; | |
2907 | } | |
83c094ad NTND |
2908 | if (!strip_extensions && istate->untracked) { |
2909 | struct strbuf sb = STRBUF_INIT; | |
2910 | ||
2911 | write_untracked_extension(&sb, istate->untracked); | |
3b1d9e04 | 2912 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_UNTRACKED, |
83c094ad NTND |
2913 | sb.len) < 0 || |
2914 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2915 | strbuf_release(&sb); | |
2916 | if (err) | |
2917 | return -1; | |
2918 | } | |
883e248b BP |
2919 | if (!strip_extensions && istate->fsmonitor_last_update) { |
2920 | struct strbuf sb = STRBUF_INIT; | |
2921 | ||
2922 | write_fsmonitor_extension(&sb, istate); | |
3b1d9e04 BP |
2923 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_FSMONITOR, sb.len) < 0 |
2924 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2925 | strbuf_release(&sb); | |
2926 | if (err) | |
2927 | return -1; | |
2928 | } | |
2929 | ||
2930 | /* | |
2931 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
2932 | * so that it can be found and processed before all the index entries are | |
2933 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
2934 | * when loading the shared index. | |
2935 | */ | |
2936 | if (offset) { | |
2937 | struct strbuf sb = STRBUF_INIT; | |
2938 | ||
2939 | write_eoie_extension(&sb, &eoie_c, offset); | |
2940 | err = write_index_ext_header(&c, NULL, newfd, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0 | |
883e248b BP |
2941 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; |
2942 | strbuf_release(&sb); | |
2943 | if (err) | |
2944 | return -1; | |
2945 | } | |
e1afca4f | 2946 | |
75691ea3 | 2947 | if (ce_flush(&c, newfd, istate->oid.hash)) |
9f41c7a6 | 2948 | return -1; |
49bd0fc2 JK |
2949 | if (close_tempfile_gently(tempfile)) { |
2950 | error(_("could not close '%s'"), tempfile->filename.buf); | |
49bd0fc2 JK |
2951 | return -1; |
2952 | } | |
9f41c7a6 | 2953 | if (stat(tempfile->filename.buf, &st)) |
e1afca4f | 2954 | return -1; |
5bcf109c KB |
2955 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
2956 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 2957 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
e1afca4f | 2958 | return 0; |
197ee8c9 | 2959 | } |
e46bbcf6 | 2960 | |
626f35c8 NTND |
2961 | void set_alternate_index_output(const char *name) |
2962 | { | |
2963 | alternate_index_output = name; | |
2964 | } | |
2965 | ||
2966 | static int commit_locked_index(struct lock_file *lk) | |
2967 | { | |
751baced MH |
2968 | if (alternate_index_output) |
2969 | return commit_lock_file_to(lk, alternate_index_output); | |
2970 | else | |
626f35c8 | 2971 | return commit_lock_file(lk); |
626f35c8 NTND |
2972 | } |
2973 | ||
03b86647 NTND |
2974 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, |
2975 | unsigned flags) | |
2976 | { | |
076aa2cb | 2977 | int ret = do_write_index(istate, lock->tempfile, 0); |
03b86647 NTND |
2978 | if (ret) |
2979 | return ret; | |
03b86647 NTND |
2980 | if (flags & COMMIT_LOCK) |
2981 | return commit_locked_index(lock); | |
812d6b00 | 2982 | return close_lock_file_gently(lock); |
03b86647 NTND |
2983 | } |
2984 | ||
5fc2fc8f NTND |
2985 | static int write_split_index(struct index_state *istate, |
2986 | struct lock_file *lock, | |
2987 | unsigned flags) | |
2988 | { | |
2989 | int ret; | |
2990 | prepare_to_write_split_index(istate); | |
2991 | ret = do_write_locked_index(istate, lock, flags); | |
2992 | finish_writing_split_index(istate); | |
2993 | return ret; | |
2994 | } | |
2995 | ||
b9683722 CC |
2996 | static const char *shared_index_expire = "2.weeks.ago"; |
2997 | ||
2998 | static unsigned long get_shared_index_expire_date(void) | |
2999 | { | |
3000 | static unsigned long shared_index_expire_date; | |
3001 | static int shared_index_expire_date_prepared; | |
3002 | ||
3003 | if (!shared_index_expire_date_prepared) { | |
3004 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3005 | &shared_index_expire); | |
3006 | shared_index_expire_date = approxidate(shared_index_expire); | |
3007 | shared_index_expire_date_prepared = 1; | |
3008 | } | |
3009 | ||
3010 | return shared_index_expire_date; | |
3011 | } | |
3012 | ||
3013 | static int should_delete_shared_index(const char *shared_index_path) | |
3014 | { | |
3015 | struct stat st; | |
3016 | unsigned long expiration; | |
3017 | ||
3018 | /* Check timestamp */ | |
3019 | expiration = get_shared_index_expire_date(); | |
3020 | if (!expiration) | |
3021 | return 0; | |
3022 | if (stat(shared_index_path, &st)) | |
78bde923 | 3023 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3024 | if (st.st_mtime > expiration) |
3025 | return 0; | |
3026 | ||
3027 | return 1; | |
3028 | } | |
3029 | ||
3030 | static int clean_shared_index_files(const char *current_hex) | |
3031 | { | |
3032 | struct dirent *de; | |
3033 | DIR *dir = opendir(get_git_dir()); | |
3034 | ||
3035 | if (!dir) | |
3036 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3037 | ||
3038 | while ((de = readdir(dir)) != NULL) { | |
3039 | const char *sha1_hex; | |
3040 | const char *shared_index_path; | |
3041 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3042 | continue; | |
3043 | if (!strcmp(sha1_hex, current_hex)) | |
3044 | continue; | |
3045 | shared_index_path = git_path("%s", de->d_name); | |
3046 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3047 | unlink(shared_index_path)) | |
3048 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3049 | } | |
3050 | closedir(dir); | |
3051 | ||
3052 | return 0; | |
3053 | } | |
3054 | ||
a0a96756 | 3055 | static int write_shared_index(struct index_state *istate, |
59f9d2dd | 3056 | struct tempfile **temp) |
c18b80a0 NTND |
3057 | { |
3058 | struct split_index *si = istate->split_index; | |
076aa2cb | 3059 | int ret; |
c18b80a0 | 3060 | |
c18b80a0 | 3061 | move_cache_to_base_index(istate); |
7db2d08c | 3062 | ret = do_write_index(si->base, *temp, 1); |
59f9d2dd | 3063 | if (ret) |
c18b80a0 | 3064 | return ret; |
7db2d08c | 3065 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3066 | if (ret) { |
7db2d08c | 3067 | error("cannot fix permission bits on %s", get_tempfile_path(*temp)); |
df801f3f CC |
3068 | return ret; |
3069 | } | |
7db2d08c | 3070 | ret = rename_tempfile(temp, |
75691ea3 | 3071 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3072 | if (!ret) { |
75691ea3 | 3073 | oidcpy(&si->base_oid, &si->base->oid); |
3074 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3075 | } |
3076 | ||
c18b80a0 NTND |
3077 | return ret; |
3078 | } | |
3079 | ||
e6a1dd77 CC |
3080 | static const int default_max_percent_split_change = 20; |
3081 | ||
3082 | static int too_many_not_shared_entries(struct index_state *istate) | |
3083 | { | |
3084 | int i, not_shared = 0; | |
3085 | int max_split = git_config_get_max_percent_split_change(); | |
3086 | ||
3087 | switch (max_split) { | |
3088 | case -1: | |
3089 | /* not or badly configured: use the default value */ | |
3090 | max_split = default_max_percent_split_change; | |
3091 | break; | |
3092 | case 0: | |
3093 | return 1; /* 0% means always write a new shared index */ | |
3094 | case 100: | |
3095 | return 0; /* 100% means never write a new shared index */ | |
3096 | default: | |
3097 | break; /* just use the configured value */ | |
3098 | } | |
3099 | ||
3100 | /* Count not shared entries */ | |
3101 | for (i = 0; i < istate->cache_nr; i++) { | |
3102 | struct cache_entry *ce = istate->cache[i]; | |
3103 | if (!ce->index) | |
3104 | not_shared++; | |
3105 | } | |
3106 | ||
3107 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3108 | } | |
3109 | ||
03b86647 NTND |
3110 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3111 | unsigned flags) | |
3112 | { | |
0d59ffb4 | 3113 | int new_shared_index, ret; |
5fc2fc8f NTND |
3114 | struct split_index *si = istate->split_index; |
3115 | ||
4592e608 NTND |
3116 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
3117 | cache_tree_verify(istate); | |
3118 | ||
61000814 MÅ |
3119 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3120 | if (flags & COMMIT_LOCK) | |
3121 | rollback_lock_file(lock); | |
3122 | return 0; | |
3123 | } | |
3124 | ||
3bd28eb2 AV |
3125 | if (istate->fsmonitor_last_update) |
3126 | fill_fsmonitor_bitmap(istate); | |
3127 | ||
5165dd59 NTND |
3128 | if (!si || alternate_index_output || |
3129 | (istate->cache_changed & ~EXTMASK)) { | |
5fc2fc8f | 3130 | if (si) |
2182abd9 | 3131 | oidclr(&si->base_oid); |
df60cf57 MÅ |
3132 | ret = do_write_locked_index(istate, lock, flags); |
3133 | goto out; | |
5fc2fc8f NTND |
3134 | } |
3135 | ||
4c2db938 | 3136 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) { |
2182abd9 | 3137 | int v = si->base_oid.hash[0]; |
d6e3c181 NTND |
3138 | if ((v & 15) < 6) |
3139 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3140 | } | |
e6a1dd77 CC |
3141 | if (too_many_not_shared_entries(istate)) |
3142 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3143 | |
3144 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3145 | ||
3146 | if (new_shared_index) { | |
59f9d2dd NTND |
3147 | struct tempfile *temp; |
3148 | int saved_errno; | |
3149 | ||
3150 | temp = mks_tempfile(git_path("sharedindex_XXXXXX")); | |
3151 | if (!temp) { | |
2182abd9 | 3152 | oidclr(&si->base_oid); |
59f9d2dd | 3153 | ret = do_write_locked_index(istate, lock, flags); |
ef5b3a6c NTND |
3154 | goto out; |
3155 | } | |
3156 | ret = write_shared_index(istate, &temp); | |
59f9d2dd NTND |
3157 | |
3158 | saved_errno = errno; | |
3159 | if (is_tempfile_active(temp)) | |
3160 | delete_tempfile(&temp); | |
3161 | errno = saved_errno; | |
3162 | ||
c18b80a0 | 3163 | if (ret) |
df60cf57 | 3164 | goto out; |
c18b80a0 NTND |
3165 | } |
3166 | ||
0d59ffb4 CC |
3167 | ret = write_split_index(istate, lock, flags); |
3168 | ||
3169 | /* Freshen the shared index only if the split-index was written */ | |
a125a223 TG |
3170 | if (!ret && !new_shared_index) { |
3171 | const char *shared_index = git_path("sharedindex.%s", | |
2182abd9 | 3172 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3173 | freshen_shared_index(shared_index, 1); |
3174 | } | |
0d59ffb4 | 3175 | |
df60cf57 MÅ |
3176 | out: |
3177 | if (flags & COMMIT_LOCK) | |
3178 | rollback_lock_file(lock); | |
0d59ffb4 | 3179 | return ret; |
03b86647 NTND |
3180 | } |
3181 | ||
e46bbcf6 MV |
3182 | /* |
3183 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3184 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3185 | * resulting in a path appearing as both a file and a directory in the | |
3186 | * index; the caller is responsible to clear out the extra entries | |
3187 | * before writing the index to a tree). Returns true if the index is | |
3188 | * unmerged. Callers who want to refuse to work from an unmerged | |
3189 | * state can call this and check its return value, instead of calling | |
3190 | * read_cache(). | |
e46bbcf6 MV |
3191 | */ |
3192 | int read_index_unmerged(struct index_state *istate) | |
3193 | { | |
3194 | int i; | |
d1a43f2a | 3195 | int unmerged = 0; |
e46bbcf6 MV |
3196 | |
3197 | read_index(istate); | |
e46bbcf6 MV |
3198 | for (i = 0; i < istate->cache_nr; i++) { |
3199 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3200 | struct cache_entry *new_ce; |
a849735b | 3201 | int len; |
d1a43f2a JH |
3202 | |
3203 | if (!ce_stage(ce)) | |
e46bbcf6 | 3204 | continue; |
d1a43f2a | 3205 | unmerged = 1; |
68c4f6a5 | 3206 | len = ce_namelen(ce); |
a849735b | 3207 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3208 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3209 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3210 | new_ce->ce_namelen = len; | |
d1a43f2a | 3211 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3212 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
d1a43f2a | 3213 | return error("%s: cannot drop to stage #0", |
5699d17e | 3214 | new_ce->name); |
e46bbcf6 | 3215 | } |
d1a43f2a | 3216 | return unmerged; |
e46bbcf6 | 3217 | } |
041aee31 | 3218 | |
98fa4738 JK |
3219 | /* |
3220 | * Returns 1 if the path is an "other" path with respect to | |
3221 | * the index; that is, the path is not mentioned in the index at all, | |
3222 | * either as a file, a directory with some files in the index, | |
3223 | * or as an unmerged entry. | |
3224 | * | |
3225 | * We helpfully remove a trailing "/" from directories so that | |
3226 | * the output of read_directory can be used as-is. | |
3227 | */ | |
3228 | int index_name_is_other(const struct index_state *istate, const char *name, | |
3229 | int namelen) | |
3230 | { | |
3231 | int pos; | |
3232 | if (namelen && name[namelen - 1] == '/') | |
3233 | namelen--; | |
3234 | pos = index_name_pos(istate, name, namelen); | |
3235 | if (0 <= pos) | |
3236 | return 0; /* exact match */ | |
3237 | pos = -pos - 1; | |
3238 | if (pos < istate->cache_nr) { | |
3239 | struct cache_entry *ce = istate->cache[pos]; | |
3240 | if (ce_namelen(ce) == namelen && | |
3241 | !memcmp(ce->name, name, namelen)) | |
3242 | return 0; /* Yup, this one exists unmerged */ | |
3243 | } | |
3244 | return 1; | |
3245 | } | |
29fb37b2 | 3246 | |
87542508 BW |
3247 | void *read_blob_data_from_index(const struct index_state *istate, |
3248 | const char *path, unsigned long *size) | |
29fb37b2 LF |
3249 | { |
3250 | int pos, len; | |
3251 | unsigned long sz; | |
3252 | enum object_type type; | |
3253 | void *data; | |
3254 | ||
3255 | len = strlen(path); | |
3256 | pos = index_name_pos(istate, path, len); | |
3257 | if (pos < 0) { | |
3258 | /* | |
3259 | * We might be in the middle of a merge, in which | |
3260 | * case we would read stage #2 (ours). | |
3261 | */ | |
3262 | int i; | |
3263 | for (i = -pos - 1; | |
3264 | (pos < 0 && i < istate->cache_nr && | |
3265 | !strcmp(istate->cache[i]->name, path)); | |
3266 | i++) | |
3267 | if (ce_stage(istate->cache[i]) == 2) | |
3268 | pos = i; | |
3269 | } | |
3270 | if (pos < 0) | |
3271 | return NULL; | |
b4f5aca4 | 3272 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); |
29fb37b2 LF |
3273 | if (!data || type != OBJ_BLOB) { |
3274 | free(data); | |
3275 | return NULL; | |
3276 | } | |
ff366825 LF |
3277 | if (size) |
3278 | *size = sz; | |
29fb37b2 LF |
3279 | return data; |
3280 | } | |
38612532 MH |
3281 | |
3282 | void stat_validity_clear(struct stat_validity *sv) | |
3283 | { | |
6a83d902 | 3284 | FREE_AND_NULL(sv->sd); |
38612532 MH |
3285 | } |
3286 | ||
3287 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
3288 | { | |
3289 | struct stat st; | |
3290 | ||
3291 | if (stat(path, &st) < 0) | |
3292 | return sv->sd == NULL; | |
3293 | if (!sv->sd) | |
3294 | return 0; | |
3295 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
3296 | } | |
3297 | ||
3298 | void stat_validity_update(struct stat_validity *sv, int fd) | |
3299 | { | |
3300 | struct stat st; | |
3301 | ||
3302 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
3303 | stat_validity_clear(sv); | |
3304 | else { | |
3305 | if (!sv->sd) | |
3306 | sv->sd = xcalloc(1, sizeof(struct stat_data)); | |
3307 | fill_stat_data(sv->sd, &st); | |
3308 | } | |
3309 | } | |
edf3b905 DT |
3310 | |
3311 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
3312 | { | |
3313 | dst->untracked = src->untracked; | |
3314 | src->untracked = NULL; | |
836ef2b6 NTND |
3315 | dst->cache_tree = src->cache_tree; |
3316 | src->cache_tree = NULL; | |
edf3b905 | 3317 | } |
a849735b | 3318 | |
8e72d675 JM |
3319 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3320 | struct index_state *istate) | |
3321 | { | |
3322 | unsigned int size = ce_size(ce); | |
3323 | int mem_pool_allocated; | |
3324 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3325 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3326 | ||
3327 | memcpy(new_entry, ce, size); | |
3328 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3329 | return new_entry; | |
3330 | } | |
3331 | ||
a849735b JM |
3332 | void discard_cache_entry(struct cache_entry *ce) |
3333 | { | |
8616a2d0 JM |
3334 | if (ce && should_validate_cache_entries()) |
3335 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3336 | ||
8e72d675 JM |
3337 | if (ce && ce->mem_pool_allocated) |
3338 | return; | |
3339 | ||
a849735b JM |
3340 | free(ce); |
3341 | } | |
8616a2d0 JM |
3342 | |
3343 | int should_validate_cache_entries(void) | |
3344 | { | |
3345 | static int validate_index_cache_entries = -1; | |
3346 | ||
3347 | if (validate_index_cache_entries < 0) { | |
3348 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3349 | validate_index_cache_entries = 1; | |
3350 | else | |
3351 | validate_index_cache_entries = 0; | |
3352 | } | |
3353 | ||
3354 | return validate_index_cache_entries; | |
3355 | } | |
3b1d9e04 BP |
3356 | |
3357 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3358 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3359 | ||
3360 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3361 | { | |
3362 | /* | |
3363 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3364 | * so that it can be found by scanning backwards from the EOF. | |
3365 | * | |
3366 | * "EOIE" | |
3367 | * <4-byte length> | |
3368 | * <4-byte offset> | |
3369 | * <20-byte hash> | |
3370 | */ | |
3371 | const char *index, *eoie; | |
3372 | uint32_t extsize; | |
3373 | size_t offset, src_offset; | |
3374 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3375 | git_hash_ctx c; | |
3376 | ||
3377 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3378 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3379 | return 0; | |
3380 | ||
3381 | /* validate the extension signature */ | |
3382 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3383 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3384 | return 0; | |
3385 | index += sizeof(uint32_t); | |
3386 | ||
3387 | /* validate the extension size */ | |
3388 | extsize = get_be32(index); | |
3389 | if (extsize != EOIE_SIZE) | |
3390 | return 0; | |
3391 | index += sizeof(uint32_t); | |
3392 | ||
3393 | /* | |
3394 | * Validate the offset we're going to look for the first extension | |
3395 | * signature is after the index header and before the eoie extension. | |
3396 | */ | |
3397 | offset = get_be32(index); | |
3398 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3399 | return 0; | |
3400 | if (mmap + offset >= eoie) | |
3401 | return 0; | |
3402 | index += sizeof(uint32_t); | |
3403 | ||
3404 | /* | |
3405 | * The hash is computed over extension types and their sizes (but not | |
3406 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3407 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3408 | * then the hash would be: | |
3409 | * | |
3410 | * SHA-1("TREE" + <binary representation of N> + | |
3411 | * "REUC" + <binary representation of M>) | |
3412 | */ | |
3413 | src_offset = offset; | |
3414 | the_hash_algo->init_fn(&c); | |
3415 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3416 | /* After an array of active_nr index entries, | |
3417 | * there can be arbitrary number of extended | |
3418 | * sections, each of which is prefixed with | |
3419 | * extension name (4-byte) and section length | |
3420 | * in 4-byte network byte order. | |
3421 | */ | |
3422 | uint32_t extsize; | |
3423 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3424 | extsize = ntohl(extsize); | |
3425 | ||
3426 | /* verify the extension size isn't so large it will wrap around */ | |
3427 | if (src_offset + 8 + extsize < src_offset) | |
3428 | return 0; | |
3429 | ||
3430 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3431 | ||
3432 | src_offset += 8; | |
3433 | src_offset += extsize; | |
3434 | } | |
3435 | the_hash_algo->final_fn(hash, &c); | |
3436 | if (!hasheq(hash, (const unsigned char *)index)) | |
3437 | return 0; | |
3438 | ||
3439 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3440 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3441 | return 0; | |
3442 | ||
3443 | return offset; | |
3444 | } | |
3445 | ||
3446 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3447 | { | |
3448 | uint32_t buffer; | |
3449 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3450 | ||
3451 | /* offset */ | |
3452 | put_be32(&buffer, offset); | |
3453 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3454 | ||
3455 | /* hash */ | |
3456 | the_hash_algo->final_fn(hash, eoie_context); | |
3457 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3458 | } | |
3255089a BP |
3459 | |
3460 | #ifndef NO_PTHREADS | |
3461 | #define IEOT_VERSION (1) | |
3462 | ||
3463 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3464 | { | |
3465 | const char *index = NULL; | |
3466 | uint32_t extsize, ext_version; | |
3467 | struct index_entry_offset_table *ieot; | |
3468 | int i, nr; | |
3469 | ||
3470 | /* find the IEOT extension */ | |
3471 | if (!offset) | |
3472 | return NULL; | |
3473 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3474 | extsize = get_be32(mmap + offset + 4); | |
3475 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3476 | index = mmap + offset + 4 + 4; | |
3477 | break; | |
3478 | } | |
3479 | offset += 8; | |
3480 | offset += extsize; | |
3481 | } | |
3482 | if (!index) | |
3483 | return NULL; | |
3484 | ||
3485 | /* validate the version is IEOT_VERSION */ | |
3486 | ext_version = get_be32(index); | |
3487 | if (ext_version != IEOT_VERSION) { | |
3488 | error("invalid IEOT version %d", ext_version); | |
3489 | return NULL; | |
3490 | } | |
3491 | index += sizeof(uint32_t); | |
3492 | ||
3493 | /* extension size - version bytes / bytes per entry */ | |
3494 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3495 | if (!nr) { | |
3496 | error("invalid number of IEOT entries %d", nr); | |
3497 | return NULL; | |
3498 | } | |
3499 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3500 | + (nr * sizeof(struct index_entry_offset))); | |
3501 | ieot->nr = nr; | |
3502 | for (i = 0; i < nr; i++) { | |
3503 | ieot->entries[i].offset = get_be32(index); | |
3504 | index += sizeof(uint32_t); | |
3505 | ieot->entries[i].nr = get_be32(index); | |
3506 | index += sizeof(uint32_t); | |
3507 | } | |
3508 | ||
3509 | return ieot; | |
3510 | } | |
3511 | ||
3512 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3513 | { | |
3514 | uint32_t buffer; | |
3515 | int i; | |
3516 | ||
3517 | /* version */ | |
3518 | put_be32(&buffer, IEOT_VERSION); | |
3519 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3520 | ||
3521 | /* ieot */ | |
3522 | for (i = 0; i < ieot->nr; i++) { | |
3523 | ||
3524 | /* offset */ | |
3525 | put_be32(&buffer, ieot->entries[i].offset); | |
3526 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3527 | ||
3528 | /* count */ | |
3529 | put_be32(&buffer, ieot->entries[i].nr); | |
3530 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3531 | } | |
3532 | } | |
3533 | #endif |