]> git.ipfire.org Git - thirdparty/git.git/blob - fsck.c
Sync with Git 2.45.1
[thirdparty/git.git] / fsck.c
1 #include "git-compat-util.h"
2 #include "date.h"
3 #include "dir.h"
4 #include "hex.h"
5 #include "object-store-ll.h"
6 #include "path.h"
7 #include "repository.h"
8 #include "object.h"
9 #include "attr.h"
10 #include "blob.h"
11 #include "tree.h"
12 #include "tree-walk.h"
13 #include "commit.h"
14 #include "tag.h"
15 #include "fsck.h"
16 #include "refs.h"
17 #include "url.h"
18 #include "utf8.h"
19 #include "oidset.h"
20 #include "packfile.h"
21 #include "submodule-config.h"
22 #include "config.h"
23 #include "help.h"
24
25 static ssize_t max_tree_entry_len = 4096;
26
27 #define STR(x) #x
28 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
29 static struct {
30 const char *id_string;
31 const char *downcased;
32 const char *camelcased;
33 enum fsck_msg_type msg_type;
34 } msg_id_info[FSCK_MSG_MAX + 1] = {
35 FOREACH_FSCK_MSG_ID(MSG_ID)
36 { NULL, NULL, NULL, -1 }
37 };
38 #undef MSG_ID
39 #undef STR
40
41 static void prepare_msg_ids(void)
42 {
43 int i;
44
45 if (msg_id_info[0].downcased)
46 return;
47
48 /* convert id_string to lower case, without underscores. */
49 for (i = 0; i < FSCK_MSG_MAX; i++) {
50 const char *p = msg_id_info[i].id_string;
51 int len = strlen(p);
52 char *q = xmalloc(len);
53
54 msg_id_info[i].downcased = q;
55 while (*p)
56 if (*p == '_')
57 p++;
58 else
59 *(q)++ = tolower(*(p)++);
60 *q = '\0';
61
62 p = msg_id_info[i].id_string;
63 q = xmalloc(len);
64 msg_id_info[i].camelcased = q;
65 while (*p) {
66 if (*p == '_') {
67 p++;
68 if (*p)
69 *q++ = *p++;
70 } else {
71 *q++ = tolower(*p++);
72 }
73 }
74 *q = '\0';
75 }
76 }
77
78 static int parse_msg_id(const char *text)
79 {
80 int i;
81
82 prepare_msg_ids();
83
84 for (i = 0; i < FSCK_MSG_MAX; i++)
85 if (!strcmp(text, msg_id_info[i].downcased))
86 return i;
87
88 return -1;
89 }
90
91 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
92 {
93 int i;
94
95 prepare_msg_ids();
96
97 for (i = 0; i < FSCK_MSG_MAX; i++)
98 list_config_item(list, prefix, msg_id_info[i].camelcased);
99 }
100
101 static enum fsck_msg_type fsck_msg_type(enum fsck_msg_id msg_id,
102 struct fsck_options *options)
103 {
104 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
105
106 if (!options->msg_type) {
107 enum fsck_msg_type msg_type = msg_id_info[msg_id].msg_type;
108
109 if (options->strict && msg_type == FSCK_WARN)
110 msg_type = FSCK_ERROR;
111 return msg_type;
112 }
113
114 return options->msg_type[msg_id];
115 }
116
117 static enum fsck_msg_type parse_msg_type(const char *str)
118 {
119 if (!strcmp(str, "error"))
120 return FSCK_ERROR;
121 else if (!strcmp(str, "warn"))
122 return FSCK_WARN;
123 else if (!strcmp(str, "ignore"))
124 return FSCK_IGNORE;
125 else
126 die("Unknown fsck message type: '%s'", str);
127 }
128
129 int is_valid_msg_type(const char *msg_id, const char *msg_type)
130 {
131 if (parse_msg_id(msg_id) < 0)
132 return 0;
133 parse_msg_type(msg_type);
134 return 1;
135 }
136
137 void fsck_set_msg_type_from_ids(struct fsck_options *options,
138 enum fsck_msg_id msg_id,
139 enum fsck_msg_type msg_type)
140 {
141 if (!options->msg_type) {
142 int i;
143 enum fsck_msg_type *severity;
144 ALLOC_ARRAY(severity, FSCK_MSG_MAX);
145 for (i = 0; i < FSCK_MSG_MAX; i++)
146 severity[i] = fsck_msg_type(i, options);
147 options->msg_type = severity;
148 }
149
150 options->msg_type[msg_id] = msg_type;
151 }
152
153 void fsck_set_msg_type(struct fsck_options *options,
154 const char *msg_id_str, const char *msg_type_str)
155 {
156 int msg_id = parse_msg_id(msg_id_str);
157 char *to_free = NULL;
158 enum fsck_msg_type msg_type;
159
160 if (msg_id < 0)
161 die("Unhandled message id: %s", msg_id_str);
162
163 if (msg_id == FSCK_MSG_LARGE_PATHNAME) {
164 const char *colon = strchr(msg_type_str, ':');
165 if (colon) {
166 msg_type_str = to_free =
167 xmemdupz(msg_type_str, colon - msg_type_str);
168 colon++;
169 if (!git_parse_ssize_t(colon, &max_tree_entry_len))
170 die("unable to parse max tree entry len: %s", colon);
171 }
172 }
173 msg_type = parse_msg_type(msg_type_str);
174
175 if (msg_type != FSCK_ERROR && msg_id_info[msg_id].msg_type == FSCK_FATAL)
176 die("Cannot demote %s to %s", msg_id_str, msg_type_str);
177
178 fsck_set_msg_type_from_ids(options, msg_id, msg_type);
179 free(to_free);
180 }
181
182 void fsck_set_msg_types(struct fsck_options *options, const char *values)
183 {
184 char *buf = xstrdup(values), *to_free = buf;
185 int done = 0;
186
187 while (!done) {
188 int len = strcspn(buf, " ,|"), equal;
189
190 done = !buf[len];
191 if (!len) {
192 buf++;
193 continue;
194 }
195 buf[len] = '\0';
196
197 for (equal = 0;
198 equal < len && buf[equal] != '=' && buf[equal] != ':';
199 equal++)
200 buf[equal] = tolower(buf[equal]);
201 buf[equal] = '\0';
202
203 if (!strcmp(buf, "skiplist")) {
204 if (equal == len)
205 die("skiplist requires a path");
206 oidset_parse_file(&options->skiplist, buf + equal + 1);
207 buf += len + 1;
208 continue;
209 }
210
211 if (equal == len)
212 die("Missing '=': '%s'", buf);
213
214 fsck_set_msg_type(options, buf, buf + equal + 1);
215 buf += len + 1;
216 }
217 free(to_free);
218 }
219
220 static int object_on_skiplist(struct fsck_options *opts,
221 const struct object_id *oid)
222 {
223 return opts && oid && oidset_contains(&opts->skiplist, oid);
224 }
225
226 __attribute__((format (printf, 5, 6)))
227 static int report(struct fsck_options *options,
228 const struct object_id *oid, enum object_type object_type,
229 enum fsck_msg_id msg_id, const char *fmt, ...)
230 {
231 va_list ap;
232 struct strbuf sb = STRBUF_INIT;
233 enum fsck_msg_type msg_type = fsck_msg_type(msg_id, options);
234 int result;
235
236 if (msg_type == FSCK_IGNORE)
237 return 0;
238
239 if (object_on_skiplist(options, oid))
240 return 0;
241
242 if (msg_type == FSCK_FATAL)
243 msg_type = FSCK_ERROR;
244 else if (msg_type == FSCK_INFO)
245 msg_type = FSCK_WARN;
246
247 prepare_msg_ids();
248 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
249
250 va_start(ap, fmt);
251 strbuf_vaddf(&sb, fmt, ap);
252 result = options->error_func(options, oid, object_type,
253 msg_type, msg_id, sb.buf);
254 strbuf_release(&sb);
255 va_end(ap);
256
257 return result;
258 }
259
260 void fsck_enable_object_names(struct fsck_options *options)
261 {
262 if (!options->object_names)
263 options->object_names = kh_init_oid_map();
264 }
265
266 const char *fsck_get_object_name(struct fsck_options *options,
267 const struct object_id *oid)
268 {
269 khiter_t pos;
270 if (!options->object_names)
271 return NULL;
272 pos = kh_get_oid_map(options->object_names, *oid);
273 if (pos >= kh_end(options->object_names))
274 return NULL;
275 return kh_value(options->object_names, pos);
276 }
277
278 void fsck_put_object_name(struct fsck_options *options,
279 const struct object_id *oid,
280 const char *fmt, ...)
281 {
282 va_list ap;
283 struct strbuf buf = STRBUF_INIT;
284 khiter_t pos;
285 int hashret;
286
287 if (!options->object_names)
288 return;
289
290 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
291 if (!hashret)
292 return;
293 va_start(ap, fmt);
294 strbuf_vaddf(&buf, fmt, ap);
295 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
296 va_end(ap);
297 }
298
299 const char *fsck_describe_object(struct fsck_options *options,
300 const struct object_id *oid)
301 {
302 static struct strbuf bufs[] = {
303 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
304 };
305 static int b = 0;
306 struct strbuf *buf;
307 const char *name = fsck_get_object_name(options, oid);
308
309 buf = bufs + b;
310 b = (b + 1) % ARRAY_SIZE(bufs);
311 strbuf_reset(buf);
312 strbuf_addstr(buf, oid_to_hex(oid));
313 if (name)
314 strbuf_addf(buf, " (%s)", name);
315
316 return buf->buf;
317 }
318
319 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
320 {
321 struct tree_desc desc;
322 struct name_entry entry;
323 int res = 0;
324 const char *name;
325
326 if (parse_tree(tree))
327 return -1;
328
329 name = fsck_get_object_name(options, &tree->object.oid);
330 if (init_tree_desc_gently(&desc, &tree->object.oid,
331 tree->buffer, tree->size, 0))
332 return -1;
333 while (tree_entry_gently(&desc, &entry)) {
334 struct object *obj;
335 int result;
336
337 if (S_ISGITLINK(entry.mode))
338 continue;
339
340 if (S_ISDIR(entry.mode)) {
341 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
342 if (name && obj)
343 fsck_put_object_name(options, &entry.oid, "%s%s/",
344 name, entry.path);
345 result = options->walk(obj, OBJ_TREE, data, options);
346 }
347 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
348 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
349 if (name && obj)
350 fsck_put_object_name(options, &entry.oid, "%s%s",
351 name, entry.path);
352 result = options->walk(obj, OBJ_BLOB, data, options);
353 }
354 else {
355 result = error("in tree %s: entry %s has bad mode %.6o",
356 fsck_describe_object(options, &tree->object.oid),
357 entry.path, entry.mode);
358 }
359 if (result < 0)
360 return result;
361 if (!res)
362 res = result;
363 }
364 return res;
365 }
366
367 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
368 {
369 int counter = 0, generation = 0, name_prefix_len = 0;
370 struct commit_list *parents;
371 int res;
372 int result;
373 const char *name;
374
375 if (repo_parse_commit(the_repository, commit))
376 return -1;
377
378 name = fsck_get_object_name(options, &commit->object.oid);
379 if (name)
380 fsck_put_object_name(options, get_commit_tree_oid(commit),
381 "%s:", name);
382
383 result = options->walk((struct object *) repo_get_commit_tree(the_repository, commit),
384 OBJ_TREE, data, options);
385 if (result < 0)
386 return result;
387 res = result;
388
389 parents = commit->parents;
390 if (name && parents) {
391 int len = strlen(name), power;
392
393 if (len && name[len - 1] == '^') {
394 generation = 1;
395 name_prefix_len = len - 1;
396 }
397 else { /* parse ~<generation> suffix */
398 for (generation = 0, power = 1;
399 len && isdigit(name[len - 1]);
400 power *= 10)
401 generation += power * (name[--len] - '0');
402 if (power > 1 && len && name[len - 1] == '~')
403 name_prefix_len = len - 1;
404 else {
405 /* Maybe a non-first parent, e.g. HEAD^2 */
406 generation = 0;
407 name_prefix_len = len;
408 }
409 }
410 }
411
412 while (parents) {
413 if (name) {
414 struct object_id *oid = &parents->item->object.oid;
415
416 if (counter++)
417 fsck_put_object_name(options, oid, "%s^%d",
418 name, counter);
419 else if (generation > 0)
420 fsck_put_object_name(options, oid, "%.*s~%d",
421 name_prefix_len, name,
422 generation + 1);
423 else
424 fsck_put_object_name(options, oid, "%s^", name);
425 }
426 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
427 if (result < 0)
428 return result;
429 if (!res)
430 res = result;
431 parents = parents->next;
432 }
433 return res;
434 }
435
436 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
437 {
438 const char *name = fsck_get_object_name(options, &tag->object.oid);
439
440 if (parse_tag(tag))
441 return -1;
442 if (name)
443 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
444 return options->walk(tag->tagged, OBJ_ANY, data, options);
445 }
446
447 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
448 {
449 if (!obj)
450 return -1;
451
452 if (obj->type == OBJ_NONE)
453 parse_object(the_repository, &obj->oid);
454
455 switch (obj->type) {
456 case OBJ_BLOB:
457 return 0;
458 case OBJ_TREE:
459 return fsck_walk_tree((struct tree *)obj, data, options);
460 case OBJ_COMMIT:
461 return fsck_walk_commit((struct commit *)obj, data, options);
462 case OBJ_TAG:
463 return fsck_walk_tag((struct tag *)obj, data, options);
464 default:
465 error("Unknown object type for %s",
466 fsck_describe_object(options, &obj->oid));
467 return -1;
468 }
469 }
470
471 struct name_stack {
472 const char **names;
473 size_t nr, alloc;
474 };
475
476 static void name_stack_push(struct name_stack *stack, const char *name)
477 {
478 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
479 stack->names[stack->nr++] = name;
480 }
481
482 static const char *name_stack_pop(struct name_stack *stack)
483 {
484 return stack->nr ? stack->names[--stack->nr] : NULL;
485 }
486
487 static void name_stack_clear(struct name_stack *stack)
488 {
489 FREE_AND_NULL(stack->names);
490 stack->nr = stack->alloc = 0;
491 }
492
493 /*
494 * The entries in a tree are ordered in the _path_ order,
495 * which means that a directory entry is ordered by adding
496 * a slash to the end of it.
497 *
498 * So a directory called "a" is ordered _after_ a file
499 * called "a.c", because "a/" sorts after "a.c".
500 */
501 #define TREE_UNORDERED (-1)
502 #define TREE_HAS_DUPS (-2)
503
504 static int is_less_than_slash(unsigned char c)
505 {
506 return '\0' < c && c < '/';
507 }
508
509 static int verify_ordered(unsigned mode1, const char *name1,
510 unsigned mode2, const char *name2,
511 struct name_stack *candidates)
512 {
513 int len1 = strlen(name1);
514 int len2 = strlen(name2);
515 int len = len1 < len2 ? len1 : len2;
516 unsigned char c1, c2;
517 int cmp;
518
519 cmp = memcmp(name1, name2, len);
520 if (cmp < 0)
521 return 0;
522 if (cmp > 0)
523 return TREE_UNORDERED;
524
525 /*
526 * Ok, the first <len> characters are the same.
527 * Now we need to order the next one, but turn
528 * a '\0' into a '/' for a directory entry.
529 */
530 c1 = name1[len];
531 c2 = name2[len];
532 if (!c1 && !c2)
533 /*
534 * git-write-tree used to write out a nonsense tree that has
535 * entries with the same name, one blob and one tree. Make
536 * sure we do not have duplicate entries.
537 */
538 return TREE_HAS_DUPS;
539 if (!c1 && S_ISDIR(mode1))
540 c1 = '/';
541 if (!c2 && S_ISDIR(mode2))
542 c2 = '/';
543
544 /*
545 * There can be non-consecutive duplicates due to the implicitly
546 * added slash, e.g.:
547 *
548 * foo
549 * foo.bar
550 * foo.bar.baz
551 * foo.bar/
552 * foo/
553 *
554 * Record non-directory candidates (like "foo" and "foo.bar" in
555 * the example) on a stack and check directory candidates (like
556 * foo/" and "foo.bar/") against that stack.
557 */
558 if (!c1 && is_less_than_slash(c2)) {
559 name_stack_push(candidates, name1);
560 } else if (c2 == '/' && is_less_than_slash(c1)) {
561 for (;;) {
562 const char *p;
563 const char *f_name = name_stack_pop(candidates);
564
565 if (!f_name)
566 break;
567 if (!skip_prefix(name2, f_name, &p))
568 continue;
569 if (!*p)
570 return TREE_HAS_DUPS;
571 if (is_less_than_slash(*p)) {
572 name_stack_push(candidates, f_name);
573 break;
574 }
575 }
576 }
577
578 return c1 < c2 ? 0 : TREE_UNORDERED;
579 }
580
581 static int fsck_tree(const struct object_id *tree_oid,
582 const char *buffer, unsigned long size,
583 struct fsck_options *options)
584 {
585 int retval = 0;
586 int has_null_sha1 = 0;
587 int has_full_path = 0;
588 int has_empty_name = 0;
589 int has_dot = 0;
590 int has_dotdot = 0;
591 int has_dotgit = 0;
592 int has_zero_pad = 0;
593 int has_bad_modes = 0;
594 int has_dup_entries = 0;
595 int not_properly_sorted = 0;
596 int has_large_name = 0;
597 struct tree_desc desc;
598 unsigned o_mode;
599 const char *o_name;
600 struct name_stack df_dup_candidates = { NULL };
601
602 if (init_tree_desc_gently(&desc, tree_oid, buffer, size,
603 TREE_DESC_RAW_MODES)) {
604 retval += report(options, tree_oid, OBJ_TREE,
605 FSCK_MSG_BAD_TREE,
606 "cannot be parsed as a tree");
607 return retval;
608 }
609
610 o_mode = 0;
611 o_name = NULL;
612
613 while (desc.size) {
614 unsigned short mode;
615 const char *name, *backslash;
616 const struct object_id *entry_oid;
617
618 entry_oid = tree_entry_extract(&desc, &name, &mode);
619
620 has_null_sha1 |= is_null_oid(entry_oid);
621 has_full_path |= !!strchr(name, '/');
622 has_empty_name |= !*name;
623 has_dot |= !strcmp(name, ".");
624 has_dotdot |= !strcmp(name, "..");
625 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
626 has_zero_pad |= *(char *)desc.buffer == '0';
627 has_large_name |= tree_entry_len(&desc.entry) > max_tree_entry_len;
628
629 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
630 if (!S_ISLNK(mode))
631 oidset_insert(&options->gitmodules_found,
632 entry_oid);
633 else
634 retval += report(options,
635 tree_oid, OBJ_TREE,
636 FSCK_MSG_GITMODULES_SYMLINK,
637 ".gitmodules is a symbolic link");
638 }
639
640 if (is_hfs_dotgitattributes(name) || is_ntfs_dotgitattributes(name)) {
641 if (!S_ISLNK(mode))
642 oidset_insert(&options->gitattributes_found,
643 entry_oid);
644 else
645 retval += report(options, tree_oid, OBJ_TREE,
646 FSCK_MSG_GITATTRIBUTES_SYMLINK,
647 ".gitattributes is a symlink");
648 }
649
650 if (S_ISLNK(mode)) {
651 if (is_hfs_dotgitignore(name) ||
652 is_ntfs_dotgitignore(name))
653 retval += report(options, tree_oid, OBJ_TREE,
654 FSCK_MSG_GITIGNORE_SYMLINK,
655 ".gitignore is a symlink");
656 if (is_hfs_dotmailmap(name) ||
657 is_ntfs_dotmailmap(name))
658 retval += report(options, tree_oid, OBJ_TREE,
659 FSCK_MSG_MAILMAP_SYMLINK,
660 ".mailmap is a symlink");
661 oidset_insert(&options->symlink_targets_found,
662 entry_oid);
663 }
664
665 if ((backslash = strchr(name, '\\'))) {
666 while (backslash) {
667 backslash++;
668 has_dotgit |= is_ntfs_dotgit(backslash);
669 if (is_ntfs_dotgitmodules(backslash)) {
670 if (!S_ISLNK(mode))
671 oidset_insert(&options->gitmodules_found,
672 entry_oid);
673 else
674 retval += report(options, tree_oid, OBJ_TREE,
675 FSCK_MSG_GITMODULES_SYMLINK,
676 ".gitmodules is a symbolic link");
677 }
678 backslash = strchr(backslash, '\\');
679 }
680 }
681
682 if (update_tree_entry_gently(&desc)) {
683 retval += report(options, tree_oid, OBJ_TREE,
684 FSCK_MSG_BAD_TREE,
685 "cannot be parsed as a tree");
686 break;
687 }
688
689 switch (mode) {
690 /*
691 * Standard modes..
692 */
693 case S_IFREG | 0755:
694 case S_IFREG | 0644:
695 case S_IFLNK:
696 case S_IFDIR:
697 case S_IFGITLINK:
698 break;
699 /*
700 * This is nonstandard, but we had a few of these
701 * early on when we honored the full set of mode
702 * bits..
703 */
704 case S_IFREG | 0664:
705 if (!options->strict)
706 break;
707 /* fallthrough */
708 default:
709 has_bad_modes = 1;
710 }
711
712 if (o_name) {
713 switch (verify_ordered(o_mode, o_name, mode, name,
714 &df_dup_candidates)) {
715 case TREE_UNORDERED:
716 not_properly_sorted = 1;
717 break;
718 case TREE_HAS_DUPS:
719 has_dup_entries = 1;
720 break;
721 default:
722 break;
723 }
724 }
725
726 o_mode = mode;
727 o_name = name;
728 }
729
730 name_stack_clear(&df_dup_candidates);
731
732 if (has_null_sha1)
733 retval += report(options, tree_oid, OBJ_TREE,
734 FSCK_MSG_NULL_SHA1,
735 "contains entries pointing to null sha1");
736 if (has_full_path)
737 retval += report(options, tree_oid, OBJ_TREE,
738 FSCK_MSG_FULL_PATHNAME,
739 "contains full pathnames");
740 if (has_empty_name)
741 retval += report(options, tree_oid, OBJ_TREE,
742 FSCK_MSG_EMPTY_NAME,
743 "contains empty pathname");
744 if (has_dot)
745 retval += report(options, tree_oid, OBJ_TREE,
746 FSCK_MSG_HAS_DOT,
747 "contains '.'");
748 if (has_dotdot)
749 retval += report(options, tree_oid, OBJ_TREE,
750 FSCK_MSG_HAS_DOTDOT,
751 "contains '..'");
752 if (has_dotgit)
753 retval += report(options, tree_oid, OBJ_TREE,
754 FSCK_MSG_HAS_DOTGIT,
755 "contains '.git'");
756 if (has_zero_pad)
757 retval += report(options, tree_oid, OBJ_TREE,
758 FSCK_MSG_ZERO_PADDED_FILEMODE,
759 "contains zero-padded file modes");
760 if (has_bad_modes)
761 retval += report(options, tree_oid, OBJ_TREE,
762 FSCK_MSG_BAD_FILEMODE,
763 "contains bad file modes");
764 if (has_dup_entries)
765 retval += report(options, tree_oid, OBJ_TREE,
766 FSCK_MSG_DUPLICATE_ENTRIES,
767 "contains duplicate file entries");
768 if (not_properly_sorted)
769 retval += report(options, tree_oid, OBJ_TREE,
770 FSCK_MSG_TREE_NOT_SORTED,
771 "not properly sorted");
772 if (has_large_name)
773 retval += report(options, tree_oid, OBJ_TREE,
774 FSCK_MSG_LARGE_PATHNAME,
775 "contains excessively large pathname");
776 return retval;
777 }
778
779 /*
780 * Confirm that the headers of a commit or tag object end in a reasonable way,
781 * either with the usual "\n\n" separator, or at least with a trailing newline
782 * on the final header line.
783 *
784 * This property is important for the memory safety of our callers. It allows
785 * them to scan the buffer linewise without constantly checking the remaining
786 * size as long as:
787 *
788 * - they check that there are bytes left in the buffer at the start of any
789 * line (i.e., that the last newline they saw was not the final one we
790 * found here)
791 *
792 * - any intra-line scanning they do will stop at a newline, which will worst
793 * case hit the newline we found here as the end-of-header. This makes it
794 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
795 */
796 static int verify_headers(const void *data, unsigned long size,
797 const struct object_id *oid, enum object_type type,
798 struct fsck_options *options)
799 {
800 const char *buffer = (const char *)data;
801 unsigned long i;
802
803 for (i = 0; i < size; i++) {
804 switch (buffer[i]) {
805 case '\0':
806 return report(options, oid, type,
807 FSCK_MSG_NUL_IN_HEADER,
808 "unterminated header: NUL at offset %ld", i);
809 case '\n':
810 if (i + 1 < size && buffer[i + 1] == '\n')
811 return 0;
812 }
813 }
814
815 /*
816 * We did not find double-LF that separates the header
817 * and the body. Not having a body is not a crime but
818 * we do want to see the terminating LF for the last header
819 * line.
820 */
821 if (size && buffer[size - 1] == '\n')
822 return 0;
823
824 return report(options, oid, type,
825 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
826 }
827
828 static int fsck_ident(const char **ident,
829 const struct object_id *oid, enum object_type type,
830 struct fsck_options *options)
831 {
832 const char *p = *ident;
833 char *end;
834
835 *ident = strchrnul(*ident, '\n');
836 if (**ident == '\n')
837 (*ident)++;
838
839 if (*p == '<')
840 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
841 p += strcspn(p, "<>\n");
842 if (*p == '>')
843 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
844 if (*p != '<')
845 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
846 if (p[-1] != ' ')
847 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
848 p++;
849 p += strcspn(p, "<>\n");
850 if (*p != '>')
851 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
852 p++;
853 if (*p != ' ')
854 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
855 p++;
856 /*
857 * Our timestamp parser is based on the C strto*() functions, which
858 * will happily eat whitespace, including the newline that is supposed
859 * to prevent us walking past the end of the buffer. So do our own
860 * scan, skipping linear whitespace but not newlines, and then
861 * confirming we found a digit. We _could_ be even more strict here,
862 * as we really expect only a single space, but since we have
863 * traditionally allowed extra whitespace, we'll continue to do so.
864 */
865 while (*p == ' ' || *p == '\t')
866 p++;
867 if (!isdigit(*p))
868 return report(options, oid, type, FSCK_MSG_BAD_DATE,
869 "invalid author/committer line - bad date");
870 if (*p == '0' && p[1] != ' ')
871 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
872 if (date_overflows(parse_timestamp(p, &end, 10)))
873 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
874 if ((end == p || *end != ' '))
875 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
876 p = end + 1;
877 if ((*p != '+' && *p != '-') ||
878 !isdigit(p[1]) ||
879 !isdigit(p[2]) ||
880 !isdigit(p[3]) ||
881 !isdigit(p[4]) ||
882 (p[5] != '\n'))
883 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
884 p += 6;
885 return 0;
886 }
887
888 static int fsck_commit(const struct object_id *oid,
889 const char *buffer, unsigned long size,
890 struct fsck_options *options)
891 {
892 struct object_id tree_oid, parent_oid;
893 unsigned author_count;
894 int err;
895 const char *buffer_begin = buffer;
896 const char *buffer_end = buffer + size;
897 const char *p;
898
899 /*
900 * We _must_ stop parsing immediately if this reports failure, as the
901 * memory safety of the rest of the function depends on it. See the
902 * comment above the definition of verify_headers() for more details.
903 */
904 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
905 return -1;
906
907 if (buffer >= buffer_end || !skip_prefix(buffer, "tree ", &buffer))
908 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
909 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
910 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
911 if (err)
912 return err;
913 }
914 buffer = p + 1;
915 while (buffer < buffer_end && skip_prefix(buffer, "parent ", &buffer)) {
916 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
917 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
918 if (err)
919 return err;
920 }
921 buffer = p + 1;
922 }
923 author_count = 0;
924 while (buffer < buffer_end && skip_prefix(buffer, "author ", &buffer)) {
925 author_count++;
926 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
927 if (err)
928 return err;
929 }
930 if (author_count < 1)
931 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
932 else if (author_count > 1)
933 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
934 if (err)
935 return err;
936 if (buffer >= buffer_end || !skip_prefix(buffer, "committer ", &buffer))
937 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
938 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
939 if (err)
940 return err;
941 if (memchr(buffer_begin, '\0', size)) {
942 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
943 "NUL byte in the commit object body");
944 if (err)
945 return err;
946 }
947 return 0;
948 }
949
950 static int fsck_tag(const struct object_id *oid, const char *buffer,
951 unsigned long size, struct fsck_options *options)
952 {
953 struct object_id tagged_oid;
954 int tagged_type;
955 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
956 &tagged_type);
957 }
958
959 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
960 unsigned long size, struct fsck_options *options,
961 struct object_id *tagged_oid,
962 int *tagged_type)
963 {
964 int ret = 0;
965 char *eol;
966 struct strbuf sb = STRBUF_INIT;
967 const char *buffer_end = buffer + size;
968 const char *p;
969
970 /*
971 * We _must_ stop parsing immediately if this reports failure, as the
972 * memory safety of the rest of the function depends on it. See the
973 * comment above the definition of verify_headers() for more details.
974 */
975 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
976 if (ret)
977 goto done;
978
979 if (buffer >= buffer_end || !skip_prefix(buffer, "object ", &buffer)) {
980 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
981 goto done;
982 }
983 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
984 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
985 if (ret)
986 goto done;
987 }
988 buffer = p + 1;
989
990 if (buffer >= buffer_end || !skip_prefix(buffer, "type ", &buffer)) {
991 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
992 goto done;
993 }
994 eol = memchr(buffer, '\n', buffer_end - buffer);
995 if (!eol) {
996 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
997 goto done;
998 }
999 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
1000 if (*tagged_type < 0)
1001 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
1002 if (ret)
1003 goto done;
1004 buffer = eol + 1;
1005
1006 if (buffer >= buffer_end || !skip_prefix(buffer, "tag ", &buffer)) {
1007 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
1008 goto done;
1009 }
1010 eol = memchr(buffer, '\n', buffer_end - buffer);
1011 if (!eol) {
1012 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
1013 goto done;
1014 }
1015 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
1016 if (check_refname_format(sb.buf, 0)) {
1017 ret = report(options, oid, OBJ_TAG,
1018 FSCK_MSG_BAD_TAG_NAME,
1019 "invalid 'tag' name: %.*s",
1020 (int)(eol - buffer), buffer);
1021 if (ret)
1022 goto done;
1023 }
1024 buffer = eol + 1;
1025
1026 if (buffer >= buffer_end || !skip_prefix(buffer, "tagger ", &buffer)) {
1027 /* early tags do not contain 'tagger' lines; warn only */
1028 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
1029 if (ret)
1030 goto done;
1031 }
1032 else
1033 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
1034
1035 if (buffer < buffer_end && !starts_with(buffer, "\n")) {
1036 /*
1037 * The verify_headers() check will allow
1038 * e.g. "[...]tagger <tagger>\nsome
1039 * garbage\n\nmessage" to pass, thinking "some
1040 * garbage" could be a custom header. E.g. "mktag"
1041 * doesn't want any unknown headers.
1042 */
1043 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1044 if (ret)
1045 goto done;
1046 }
1047
1048 done:
1049 strbuf_release(&sb);
1050 return ret;
1051 }
1052
1053 struct fsck_gitmodules_data {
1054 const struct object_id *oid;
1055 struct fsck_options *options;
1056 int ret;
1057 };
1058
1059 static int fsck_gitmodules_fn(const char *var, const char *value,
1060 const struct config_context *ctx UNUSED,
1061 void *vdata)
1062 {
1063 struct fsck_gitmodules_data *data = vdata;
1064 const char *subsection, *key;
1065 size_t subsection_len;
1066 char *name;
1067
1068 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1069 !subsection)
1070 return 0;
1071
1072 name = xmemdupz(subsection, subsection_len);
1073 if (check_submodule_name(name) < 0)
1074 data->ret |= report(data->options,
1075 data->oid, OBJ_BLOB,
1076 FSCK_MSG_GITMODULES_NAME,
1077 "disallowed submodule name: %s",
1078 name);
1079 if (!strcmp(key, "url") && value &&
1080 check_submodule_url(value) < 0)
1081 data->ret |= report(data->options,
1082 data->oid, OBJ_BLOB,
1083 FSCK_MSG_GITMODULES_URL,
1084 "disallowed submodule url: %s",
1085 value);
1086 if (!strcmp(key, "path") && value &&
1087 looks_like_command_line_option(value))
1088 data->ret |= report(data->options,
1089 data->oid, OBJ_BLOB,
1090 FSCK_MSG_GITMODULES_PATH,
1091 "disallowed submodule path: %s",
1092 value);
1093 if (!strcmp(key, "update") && value &&
1094 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1095 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1096 FSCK_MSG_GITMODULES_UPDATE,
1097 "disallowed submodule update setting: %s",
1098 value);
1099 free(name);
1100
1101 return 0;
1102 }
1103
1104 static int fsck_blob(const struct object_id *oid, const char *buf,
1105 unsigned long size, struct fsck_options *options)
1106 {
1107 int ret = 0;
1108
1109 if (object_on_skiplist(options, oid))
1110 return 0;
1111
1112 if (oidset_contains(&options->gitmodules_found, oid)) {
1113 struct config_options config_opts = { 0 };
1114 struct fsck_gitmodules_data data;
1115
1116 oidset_insert(&options->gitmodules_done, oid);
1117
1118 if (!buf) {
1119 /*
1120 * A missing buffer here is a sign that the caller found the
1121 * blob too gigantic to load into memory. Let's just consider
1122 * that an error.
1123 */
1124 return report(options, oid, OBJ_BLOB,
1125 FSCK_MSG_GITMODULES_LARGE,
1126 ".gitmodules too large to parse");
1127 }
1128
1129 data.oid = oid;
1130 data.options = options;
1131 data.ret = 0;
1132 config_opts.error_action = CONFIG_ERROR_SILENT;
1133 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1134 ".gitmodules", buf, size, &data,
1135 CONFIG_SCOPE_UNKNOWN, &config_opts))
1136 data.ret |= report(options, oid, OBJ_BLOB,
1137 FSCK_MSG_GITMODULES_PARSE,
1138 "could not parse gitmodules blob");
1139 ret |= data.ret;
1140 }
1141
1142 if (oidset_contains(&options->gitattributes_found, oid)) {
1143 const char *ptr;
1144
1145 oidset_insert(&options->gitattributes_done, oid);
1146
1147 if (!buf || size > ATTR_MAX_FILE_SIZE) {
1148 /*
1149 * A missing buffer here is a sign that the caller found the
1150 * blob too gigantic to load into memory. Let's just consider
1151 * that an error.
1152 */
1153 return report(options, oid, OBJ_BLOB,
1154 FSCK_MSG_GITATTRIBUTES_LARGE,
1155 ".gitattributes too large to parse");
1156 }
1157
1158 for (ptr = buf; *ptr; ) {
1159 const char *eol = strchrnul(ptr, '\n');
1160 if (eol - ptr >= ATTR_MAX_LINE_LENGTH) {
1161 ret |= report(options, oid, OBJ_BLOB,
1162 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH,
1163 ".gitattributes has too long lines to parse");
1164 break;
1165 }
1166
1167 ptr = *eol ? eol + 1 : eol;
1168 }
1169 }
1170
1171 if (oidset_contains(&options->symlink_targets_found, oid)) {
1172 const char *ptr = buf;
1173 const struct object_id *reported = NULL;
1174
1175 oidset_insert(&options->symlink_targets_done, oid);
1176
1177 if (!buf || size > PATH_MAX) {
1178 /*
1179 * A missing buffer here is a sign that the caller found the
1180 * blob too gigantic to load into memory. Let's just consider
1181 * that an error.
1182 */
1183 return report(options, oid, OBJ_BLOB,
1184 FSCK_MSG_SYMLINK_TARGET_LENGTH,
1185 "symlink target too long");
1186 }
1187
1188 while (!reported && ptr) {
1189 const char *p = ptr;
1190 char c, *slash = strchrnul(ptr, '/');
1191 char *backslash = memchr(ptr, '\\', slash - ptr);
1192
1193 c = *slash;
1194 *slash = '\0';
1195
1196 while (!reported && backslash) {
1197 *backslash = '\0';
1198 if (is_ntfs_dotgit(p))
1199 ret |= report(options, reported = oid, OBJ_BLOB,
1200 FSCK_MSG_SYMLINK_POINTS_TO_GIT_DIR,
1201 "symlink target points to git dir");
1202 *backslash = '\\';
1203 p = backslash + 1;
1204 backslash = memchr(p, '\\', slash - p);
1205 }
1206 if (!reported && is_ntfs_dotgit(p))
1207 ret |= report(options, reported = oid, OBJ_BLOB,
1208 FSCK_MSG_SYMLINK_POINTS_TO_GIT_DIR,
1209 "symlink target points to git dir");
1210
1211 if (!reported && is_hfs_dotgit(ptr))
1212 ret |= report(options, reported = oid, OBJ_BLOB,
1213 FSCK_MSG_SYMLINK_POINTS_TO_GIT_DIR,
1214 "symlink target points to git dir");
1215
1216 *slash = c;
1217 ptr = c ? slash + 1 : NULL;
1218 }
1219 }
1220
1221 return ret;
1222 }
1223
1224 int fsck_object(struct object *obj, void *data, unsigned long size,
1225 struct fsck_options *options)
1226 {
1227 if (!obj)
1228 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1229
1230 return fsck_buffer(&obj->oid, obj->type, data, size, options);
1231 }
1232
1233 int fsck_buffer(const struct object_id *oid, enum object_type type,
1234 void *data, unsigned long size,
1235 struct fsck_options *options)
1236 {
1237 if (type == OBJ_BLOB)
1238 return fsck_blob(oid, data, size, options);
1239 if (type == OBJ_TREE)
1240 return fsck_tree(oid, data, size, options);
1241 if (type == OBJ_COMMIT)
1242 return fsck_commit(oid, data, size, options);
1243 if (type == OBJ_TAG)
1244 return fsck_tag(oid, data, size, options);
1245
1246 return report(options, oid, type,
1247 FSCK_MSG_UNKNOWN_TYPE,
1248 "unknown type '%d' (internal fsck error)",
1249 type);
1250 }
1251
1252 int fsck_error_function(struct fsck_options *o,
1253 const struct object_id *oid,
1254 enum object_type object_type UNUSED,
1255 enum fsck_msg_type msg_type,
1256 enum fsck_msg_id msg_id UNUSED,
1257 const char *message)
1258 {
1259 if (msg_type == FSCK_WARN) {
1260 warning("object %s: %s", fsck_describe_object(o, oid), message);
1261 return 0;
1262 }
1263 error("object %s: %s", fsck_describe_object(o, oid), message);
1264 return 1;
1265 }
1266
1267 static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
1268 enum fsck_msg_id msg_missing, enum fsck_msg_id msg_type,
1269 struct fsck_options *options, const char *blob_type)
1270 {
1271 int ret = 0;
1272 struct oidset_iter iter;
1273 const struct object_id *oid;
1274
1275 oidset_iter_init(blobs_found, &iter);
1276 while ((oid = oidset_iter_next(&iter))) {
1277 enum object_type type;
1278 unsigned long size;
1279 char *buf;
1280
1281 if (oidset_contains(blobs_done, oid))
1282 continue;
1283
1284 buf = repo_read_object_file(the_repository, oid, &type, &size);
1285 if (!buf) {
1286 if (is_promisor_object(oid))
1287 continue;
1288 ret |= report(options,
1289 oid, OBJ_BLOB, msg_missing,
1290 "unable to read %s blob", blob_type);
1291 continue;
1292 }
1293
1294 if (type == OBJ_BLOB)
1295 ret |= fsck_blob(oid, buf, size, options);
1296 else
1297 ret |= report(options, oid, type, msg_type,
1298 "non-blob found at %s", blob_type);
1299 free(buf);
1300 }
1301
1302 oidset_clear(blobs_found);
1303 oidset_clear(blobs_done);
1304
1305 return ret;
1306 }
1307
1308 int fsck_finish(struct fsck_options *options)
1309 {
1310 int ret = 0;
1311
1312 ret |= fsck_blobs(&options->gitmodules_found, &options->gitmodules_done,
1313 FSCK_MSG_GITMODULES_MISSING, FSCK_MSG_GITMODULES_BLOB,
1314 options, ".gitmodules");
1315 ret |= fsck_blobs(&options->gitattributes_found, &options->gitattributes_done,
1316 FSCK_MSG_GITATTRIBUTES_MISSING, FSCK_MSG_GITATTRIBUTES_BLOB,
1317 options, ".gitattributes");
1318
1319 ret |= fsck_blobs(&options->symlink_targets_found, &options->symlink_targets_done,
1320 FSCK_MSG_SYMLINK_TARGET_MISSING, FSCK_MSG_SYMLINK_TARGET_BLOB,
1321 options, "<symlink-target>");
1322
1323 return ret;
1324 }
1325
1326 int git_fsck_config(const char *var, const char *value,
1327 const struct config_context *ctx, void *cb)
1328 {
1329 struct fsck_options *options = cb;
1330 const char *msg_id;
1331
1332 if (strcmp(var, "fsck.skiplist") == 0) {
1333 const char *path;
1334 struct strbuf sb = STRBUF_INIT;
1335
1336 if (git_config_pathname(&path, var, value))
1337 return 1;
1338 strbuf_addf(&sb, "skiplist=%s", path);
1339 free((char *)path);
1340 fsck_set_msg_types(options, sb.buf);
1341 strbuf_release(&sb);
1342 return 0;
1343 }
1344
1345 if (skip_prefix(var, "fsck.", &msg_id)) {
1346 if (!value)
1347 return config_error_nonbool(var);
1348 fsck_set_msg_type(options, msg_id, value);
1349 return 0;
1350 }
1351
1352 return git_default_config(var, value, ctx, cb);
1353 }
1354
1355 /*
1356 * Custom error callbacks that are used in more than one place.
1357 */
1358
1359 int fsck_error_cb_print_missing_gitmodules(struct fsck_options *o,
1360 const struct object_id *oid,
1361 enum object_type object_type,
1362 enum fsck_msg_type msg_type,
1363 enum fsck_msg_id msg_id,
1364 const char *message)
1365 {
1366 if (msg_id == FSCK_MSG_GITMODULES_MISSING) {
1367 puts(oid_to_hex(oid));
1368 return 0;
1369 }
1370 return fsck_error_function(o, oid, object_type, msg_type, msg_id, message);
1371 }