]> git.ipfire.org Git - thirdparty/git.git/blob - fsck.c
Merge branch 'rs/receive-pack-remove-find-header'
[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->buffer, tree->size, 0))
331 return -1;
332 while (tree_entry_gently(&desc, &entry)) {
333 struct object *obj;
334 int result;
335
336 if (S_ISGITLINK(entry.mode))
337 continue;
338
339 if (S_ISDIR(entry.mode)) {
340 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
341 if (name && obj)
342 fsck_put_object_name(options, &entry.oid, "%s%s/",
343 name, entry.path);
344 result = options->walk(obj, OBJ_TREE, data, options);
345 }
346 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
347 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
348 if (name && obj)
349 fsck_put_object_name(options, &entry.oid, "%s%s",
350 name, entry.path);
351 result = options->walk(obj, OBJ_BLOB, data, options);
352 }
353 else {
354 result = error("in tree %s: entry %s has bad mode %.6o",
355 fsck_describe_object(options, &tree->object.oid),
356 entry.path, entry.mode);
357 }
358 if (result < 0)
359 return result;
360 if (!res)
361 res = result;
362 }
363 return res;
364 }
365
366 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
367 {
368 int counter = 0, generation = 0, name_prefix_len = 0;
369 struct commit_list *parents;
370 int res;
371 int result;
372 const char *name;
373
374 if (repo_parse_commit(the_repository, commit))
375 return -1;
376
377 name = fsck_get_object_name(options, &commit->object.oid);
378 if (name)
379 fsck_put_object_name(options, get_commit_tree_oid(commit),
380 "%s:", name);
381
382 result = options->walk((struct object *) repo_get_commit_tree(the_repository, commit),
383 OBJ_TREE, data, options);
384 if (result < 0)
385 return result;
386 res = result;
387
388 parents = commit->parents;
389 if (name && parents) {
390 int len = strlen(name), power;
391
392 if (len && name[len - 1] == '^') {
393 generation = 1;
394 name_prefix_len = len - 1;
395 }
396 else { /* parse ~<generation> suffix */
397 for (generation = 0, power = 1;
398 len && isdigit(name[len - 1]);
399 power *= 10)
400 generation += power * (name[--len] - '0');
401 if (power > 1 && len && name[len - 1] == '~')
402 name_prefix_len = len - 1;
403 else {
404 /* Maybe a non-first parent, e.g. HEAD^2 */
405 generation = 0;
406 name_prefix_len = len;
407 }
408 }
409 }
410
411 while (parents) {
412 if (name) {
413 struct object_id *oid = &parents->item->object.oid;
414
415 if (counter++)
416 fsck_put_object_name(options, oid, "%s^%d",
417 name, counter);
418 else if (generation > 0)
419 fsck_put_object_name(options, oid, "%.*s~%d",
420 name_prefix_len, name,
421 generation + 1);
422 else
423 fsck_put_object_name(options, oid, "%s^", name);
424 }
425 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
426 if (result < 0)
427 return result;
428 if (!res)
429 res = result;
430 parents = parents->next;
431 }
432 return res;
433 }
434
435 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
436 {
437 const char *name = fsck_get_object_name(options, &tag->object.oid);
438
439 if (parse_tag(tag))
440 return -1;
441 if (name)
442 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
443 return options->walk(tag->tagged, OBJ_ANY, data, options);
444 }
445
446 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
447 {
448 if (!obj)
449 return -1;
450
451 if (obj->type == OBJ_NONE)
452 parse_object(the_repository, &obj->oid);
453
454 switch (obj->type) {
455 case OBJ_BLOB:
456 return 0;
457 case OBJ_TREE:
458 return fsck_walk_tree((struct tree *)obj, data, options);
459 case OBJ_COMMIT:
460 return fsck_walk_commit((struct commit *)obj, data, options);
461 case OBJ_TAG:
462 return fsck_walk_tag((struct tag *)obj, data, options);
463 default:
464 error("Unknown object type for %s",
465 fsck_describe_object(options, &obj->oid));
466 return -1;
467 }
468 }
469
470 struct name_stack {
471 const char **names;
472 size_t nr, alloc;
473 };
474
475 static void name_stack_push(struct name_stack *stack, const char *name)
476 {
477 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
478 stack->names[stack->nr++] = name;
479 }
480
481 static const char *name_stack_pop(struct name_stack *stack)
482 {
483 return stack->nr ? stack->names[--stack->nr] : NULL;
484 }
485
486 static void name_stack_clear(struct name_stack *stack)
487 {
488 FREE_AND_NULL(stack->names);
489 stack->nr = stack->alloc = 0;
490 }
491
492 /*
493 * The entries in a tree are ordered in the _path_ order,
494 * which means that a directory entry is ordered by adding
495 * a slash to the end of it.
496 *
497 * So a directory called "a" is ordered _after_ a file
498 * called "a.c", because "a/" sorts after "a.c".
499 */
500 #define TREE_UNORDERED (-1)
501 #define TREE_HAS_DUPS (-2)
502
503 static int is_less_than_slash(unsigned char c)
504 {
505 return '\0' < c && c < '/';
506 }
507
508 static int verify_ordered(unsigned mode1, const char *name1,
509 unsigned mode2, const char *name2,
510 struct name_stack *candidates)
511 {
512 int len1 = strlen(name1);
513 int len2 = strlen(name2);
514 int len = len1 < len2 ? len1 : len2;
515 unsigned char c1, c2;
516 int cmp;
517
518 cmp = memcmp(name1, name2, len);
519 if (cmp < 0)
520 return 0;
521 if (cmp > 0)
522 return TREE_UNORDERED;
523
524 /*
525 * Ok, the first <len> characters are the same.
526 * Now we need to order the next one, but turn
527 * a '\0' into a '/' for a directory entry.
528 */
529 c1 = name1[len];
530 c2 = name2[len];
531 if (!c1 && !c2)
532 /*
533 * git-write-tree used to write out a nonsense tree that has
534 * entries with the same name, one blob and one tree. Make
535 * sure we do not have duplicate entries.
536 */
537 return TREE_HAS_DUPS;
538 if (!c1 && S_ISDIR(mode1))
539 c1 = '/';
540 if (!c2 && S_ISDIR(mode2))
541 c2 = '/';
542
543 /*
544 * There can be non-consecutive duplicates due to the implicitly
545 * added slash, e.g.:
546 *
547 * foo
548 * foo.bar
549 * foo.bar.baz
550 * foo.bar/
551 * foo/
552 *
553 * Record non-directory candidates (like "foo" and "foo.bar" in
554 * the example) on a stack and check directory candidates (like
555 * foo/" and "foo.bar/") against that stack.
556 */
557 if (!c1 && is_less_than_slash(c2)) {
558 name_stack_push(candidates, name1);
559 } else if (c2 == '/' && is_less_than_slash(c1)) {
560 for (;;) {
561 const char *p;
562 const char *f_name = name_stack_pop(candidates);
563
564 if (!f_name)
565 break;
566 if (!skip_prefix(name2, f_name, &p))
567 continue;
568 if (!*p)
569 return TREE_HAS_DUPS;
570 if (is_less_than_slash(*p)) {
571 name_stack_push(candidates, f_name);
572 break;
573 }
574 }
575 }
576
577 return c1 < c2 ? 0 : TREE_UNORDERED;
578 }
579
580 static int fsck_tree(const struct object_id *tree_oid,
581 const char *buffer, unsigned long size,
582 struct fsck_options *options)
583 {
584 int retval = 0;
585 int has_null_sha1 = 0;
586 int has_full_path = 0;
587 int has_empty_name = 0;
588 int has_dot = 0;
589 int has_dotdot = 0;
590 int has_dotgit = 0;
591 int has_zero_pad = 0;
592 int has_bad_modes = 0;
593 int has_dup_entries = 0;
594 int not_properly_sorted = 0;
595 int has_large_name = 0;
596 struct tree_desc desc;
597 unsigned o_mode;
598 const char *o_name;
599 struct name_stack df_dup_candidates = { NULL };
600
601 if (init_tree_desc_gently(&desc, buffer, size, TREE_DESC_RAW_MODES)) {
602 retval += report(options, tree_oid, OBJ_TREE,
603 FSCK_MSG_BAD_TREE,
604 "cannot be parsed as a tree");
605 return retval;
606 }
607
608 o_mode = 0;
609 o_name = NULL;
610
611 while (desc.size) {
612 unsigned short mode;
613 const char *name, *backslash;
614 const struct object_id *entry_oid;
615
616 entry_oid = tree_entry_extract(&desc, &name, &mode);
617
618 has_null_sha1 |= is_null_oid(entry_oid);
619 has_full_path |= !!strchr(name, '/');
620 has_empty_name |= !*name;
621 has_dot |= !strcmp(name, ".");
622 has_dotdot |= !strcmp(name, "..");
623 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
624 has_zero_pad |= *(char *)desc.buffer == '0';
625 has_large_name |= tree_entry_len(&desc.entry) > max_tree_entry_len;
626
627 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
628 if (!S_ISLNK(mode))
629 oidset_insert(&options->gitmodules_found,
630 entry_oid);
631 else
632 retval += report(options,
633 tree_oid, OBJ_TREE,
634 FSCK_MSG_GITMODULES_SYMLINK,
635 ".gitmodules is a symbolic link");
636 }
637
638 if (is_hfs_dotgitattributes(name) || is_ntfs_dotgitattributes(name)) {
639 if (!S_ISLNK(mode))
640 oidset_insert(&options->gitattributes_found,
641 entry_oid);
642 else
643 retval += report(options, tree_oid, OBJ_TREE,
644 FSCK_MSG_GITATTRIBUTES_SYMLINK,
645 ".gitattributes is a symlink");
646 }
647
648 if (S_ISLNK(mode)) {
649 if (is_hfs_dotgitignore(name) ||
650 is_ntfs_dotgitignore(name))
651 retval += report(options, tree_oid, OBJ_TREE,
652 FSCK_MSG_GITIGNORE_SYMLINK,
653 ".gitignore is a symlink");
654 if (is_hfs_dotmailmap(name) ||
655 is_ntfs_dotmailmap(name))
656 retval += report(options, tree_oid, OBJ_TREE,
657 FSCK_MSG_MAILMAP_SYMLINK,
658 ".mailmap is a symlink");
659 }
660
661 if ((backslash = strchr(name, '\\'))) {
662 while (backslash) {
663 backslash++;
664 has_dotgit |= is_ntfs_dotgit(backslash);
665 if (is_ntfs_dotgitmodules(backslash)) {
666 if (!S_ISLNK(mode))
667 oidset_insert(&options->gitmodules_found,
668 entry_oid);
669 else
670 retval += report(options, tree_oid, OBJ_TREE,
671 FSCK_MSG_GITMODULES_SYMLINK,
672 ".gitmodules is a symbolic link");
673 }
674 backslash = strchr(backslash, '\\');
675 }
676 }
677
678 if (update_tree_entry_gently(&desc)) {
679 retval += report(options, tree_oid, OBJ_TREE,
680 FSCK_MSG_BAD_TREE,
681 "cannot be parsed as a tree");
682 break;
683 }
684
685 switch (mode) {
686 /*
687 * Standard modes..
688 */
689 case S_IFREG | 0755:
690 case S_IFREG | 0644:
691 case S_IFLNK:
692 case S_IFDIR:
693 case S_IFGITLINK:
694 break;
695 /*
696 * This is nonstandard, but we had a few of these
697 * early on when we honored the full set of mode
698 * bits..
699 */
700 case S_IFREG | 0664:
701 if (!options->strict)
702 break;
703 /* fallthrough */
704 default:
705 has_bad_modes = 1;
706 }
707
708 if (o_name) {
709 switch (verify_ordered(o_mode, o_name, mode, name,
710 &df_dup_candidates)) {
711 case TREE_UNORDERED:
712 not_properly_sorted = 1;
713 break;
714 case TREE_HAS_DUPS:
715 has_dup_entries = 1;
716 break;
717 default:
718 break;
719 }
720 }
721
722 o_mode = mode;
723 o_name = name;
724 }
725
726 name_stack_clear(&df_dup_candidates);
727
728 if (has_null_sha1)
729 retval += report(options, tree_oid, OBJ_TREE,
730 FSCK_MSG_NULL_SHA1,
731 "contains entries pointing to null sha1");
732 if (has_full_path)
733 retval += report(options, tree_oid, OBJ_TREE,
734 FSCK_MSG_FULL_PATHNAME,
735 "contains full pathnames");
736 if (has_empty_name)
737 retval += report(options, tree_oid, OBJ_TREE,
738 FSCK_MSG_EMPTY_NAME,
739 "contains empty pathname");
740 if (has_dot)
741 retval += report(options, tree_oid, OBJ_TREE,
742 FSCK_MSG_HAS_DOT,
743 "contains '.'");
744 if (has_dotdot)
745 retval += report(options, tree_oid, OBJ_TREE,
746 FSCK_MSG_HAS_DOTDOT,
747 "contains '..'");
748 if (has_dotgit)
749 retval += report(options, tree_oid, OBJ_TREE,
750 FSCK_MSG_HAS_DOTGIT,
751 "contains '.git'");
752 if (has_zero_pad)
753 retval += report(options, tree_oid, OBJ_TREE,
754 FSCK_MSG_ZERO_PADDED_FILEMODE,
755 "contains zero-padded file modes");
756 if (has_bad_modes)
757 retval += report(options, tree_oid, OBJ_TREE,
758 FSCK_MSG_BAD_FILEMODE,
759 "contains bad file modes");
760 if (has_dup_entries)
761 retval += report(options, tree_oid, OBJ_TREE,
762 FSCK_MSG_DUPLICATE_ENTRIES,
763 "contains duplicate file entries");
764 if (not_properly_sorted)
765 retval += report(options, tree_oid, OBJ_TREE,
766 FSCK_MSG_TREE_NOT_SORTED,
767 "not properly sorted");
768 if (has_large_name)
769 retval += report(options, tree_oid, OBJ_TREE,
770 FSCK_MSG_LARGE_PATHNAME,
771 "contains excessively large pathname");
772 return retval;
773 }
774
775 /*
776 * Confirm that the headers of a commit or tag object end in a reasonable way,
777 * either with the usual "\n\n" separator, or at least with a trailing newline
778 * on the final header line.
779 *
780 * This property is important for the memory safety of our callers. It allows
781 * them to scan the buffer linewise without constantly checking the remaining
782 * size as long as:
783 *
784 * - they check that there are bytes left in the buffer at the start of any
785 * line (i.e., that the last newline they saw was not the final one we
786 * found here)
787 *
788 * - any intra-line scanning they do will stop at a newline, which will worst
789 * case hit the newline we found here as the end-of-header. This makes it
790 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
791 */
792 static int verify_headers(const void *data, unsigned long size,
793 const struct object_id *oid, enum object_type type,
794 struct fsck_options *options)
795 {
796 const char *buffer = (const char *)data;
797 unsigned long i;
798
799 for (i = 0; i < size; i++) {
800 switch (buffer[i]) {
801 case '\0':
802 return report(options, oid, type,
803 FSCK_MSG_NUL_IN_HEADER,
804 "unterminated header: NUL at offset %ld", i);
805 case '\n':
806 if (i + 1 < size && buffer[i + 1] == '\n')
807 return 0;
808 }
809 }
810
811 /*
812 * We did not find double-LF that separates the header
813 * and the body. Not having a body is not a crime but
814 * we do want to see the terminating LF for the last header
815 * line.
816 */
817 if (size && buffer[size - 1] == '\n')
818 return 0;
819
820 return report(options, oid, type,
821 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
822 }
823
824 static int fsck_ident(const char **ident,
825 const struct object_id *oid, enum object_type type,
826 struct fsck_options *options)
827 {
828 const char *p = *ident;
829 char *end;
830
831 *ident = strchrnul(*ident, '\n');
832 if (**ident == '\n')
833 (*ident)++;
834
835 if (*p == '<')
836 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
837 p += strcspn(p, "<>\n");
838 if (*p == '>')
839 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
840 if (*p != '<')
841 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
842 if (p[-1] != ' ')
843 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
844 p++;
845 p += strcspn(p, "<>\n");
846 if (*p != '>')
847 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
848 p++;
849 if (*p != ' ')
850 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
851 p++;
852 /*
853 * Our timestamp parser is based on the C strto*() functions, which
854 * will happily eat whitespace, including the newline that is supposed
855 * to prevent us walking past the end of the buffer. So do our own
856 * scan, skipping linear whitespace but not newlines, and then
857 * confirming we found a digit. We _could_ be even more strict here,
858 * as we really expect only a single space, but since we have
859 * traditionally allowed extra whitespace, we'll continue to do so.
860 */
861 while (*p == ' ' || *p == '\t')
862 p++;
863 if (!isdigit(*p))
864 return report(options, oid, type, FSCK_MSG_BAD_DATE,
865 "invalid author/committer line - bad date");
866 if (*p == '0' && p[1] != ' ')
867 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
868 if (date_overflows(parse_timestamp(p, &end, 10)))
869 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
870 if ((end == p || *end != ' '))
871 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
872 p = end + 1;
873 if ((*p != '+' && *p != '-') ||
874 !isdigit(p[1]) ||
875 !isdigit(p[2]) ||
876 !isdigit(p[3]) ||
877 !isdigit(p[4]) ||
878 (p[5] != '\n'))
879 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
880 p += 6;
881 return 0;
882 }
883
884 static int fsck_commit(const struct object_id *oid,
885 const char *buffer, unsigned long size,
886 struct fsck_options *options)
887 {
888 struct object_id tree_oid, parent_oid;
889 unsigned author_count;
890 int err;
891 const char *buffer_begin = buffer;
892 const char *buffer_end = buffer + size;
893 const char *p;
894
895 /*
896 * We _must_ stop parsing immediately if this reports failure, as the
897 * memory safety of the rest of the function depends on it. See the
898 * comment above the definition of verify_headers() for more details.
899 */
900 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
901 return -1;
902
903 if (buffer >= buffer_end || !skip_prefix(buffer, "tree ", &buffer))
904 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
905 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
906 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
907 if (err)
908 return err;
909 }
910 buffer = p + 1;
911 while (buffer < buffer_end && skip_prefix(buffer, "parent ", &buffer)) {
912 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
913 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
914 if (err)
915 return err;
916 }
917 buffer = p + 1;
918 }
919 author_count = 0;
920 while (buffer < buffer_end && skip_prefix(buffer, "author ", &buffer)) {
921 author_count++;
922 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
923 if (err)
924 return err;
925 }
926 if (author_count < 1)
927 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
928 else if (author_count > 1)
929 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
930 if (err)
931 return err;
932 if (buffer >= buffer_end || !skip_prefix(buffer, "committer ", &buffer))
933 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
934 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
935 if (err)
936 return err;
937 if (memchr(buffer_begin, '\0', size)) {
938 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
939 "NUL byte in the commit object body");
940 if (err)
941 return err;
942 }
943 return 0;
944 }
945
946 static int fsck_tag(const struct object_id *oid, const char *buffer,
947 unsigned long size, struct fsck_options *options)
948 {
949 struct object_id tagged_oid;
950 int tagged_type;
951 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
952 &tagged_type);
953 }
954
955 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
956 unsigned long size, struct fsck_options *options,
957 struct object_id *tagged_oid,
958 int *tagged_type)
959 {
960 int ret = 0;
961 char *eol;
962 struct strbuf sb = STRBUF_INIT;
963 const char *buffer_end = buffer + size;
964 const char *p;
965
966 /*
967 * We _must_ stop parsing immediately if this reports failure, as the
968 * memory safety of the rest of the function depends on it. See the
969 * comment above the definition of verify_headers() for more details.
970 */
971 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
972 if (ret)
973 goto done;
974
975 if (buffer >= buffer_end || !skip_prefix(buffer, "object ", &buffer)) {
976 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
977 goto done;
978 }
979 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
980 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
981 if (ret)
982 goto done;
983 }
984 buffer = p + 1;
985
986 if (buffer >= buffer_end || !skip_prefix(buffer, "type ", &buffer)) {
987 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
988 goto done;
989 }
990 eol = memchr(buffer, '\n', buffer_end - buffer);
991 if (!eol) {
992 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
993 goto done;
994 }
995 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
996 if (*tagged_type < 0)
997 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
998 if (ret)
999 goto done;
1000 buffer = eol + 1;
1001
1002 if (buffer >= buffer_end || !skip_prefix(buffer, "tag ", &buffer)) {
1003 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
1004 goto done;
1005 }
1006 eol = memchr(buffer, '\n', buffer_end - buffer);
1007 if (!eol) {
1008 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
1009 goto done;
1010 }
1011 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
1012 if (check_refname_format(sb.buf, 0)) {
1013 ret = report(options, oid, OBJ_TAG,
1014 FSCK_MSG_BAD_TAG_NAME,
1015 "invalid 'tag' name: %.*s",
1016 (int)(eol - buffer), buffer);
1017 if (ret)
1018 goto done;
1019 }
1020 buffer = eol + 1;
1021
1022 if (buffer >= buffer_end || !skip_prefix(buffer, "tagger ", &buffer)) {
1023 /* early tags do not contain 'tagger' lines; warn only */
1024 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
1025 if (ret)
1026 goto done;
1027 }
1028 else
1029 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
1030
1031 if (buffer < buffer_end && !starts_with(buffer, "\n")) {
1032 /*
1033 * The verify_headers() check will allow
1034 * e.g. "[...]tagger <tagger>\nsome
1035 * garbage\n\nmessage" to pass, thinking "some
1036 * garbage" could be a custom header. E.g. "mktag"
1037 * doesn't want any unknown headers.
1038 */
1039 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1040 if (ret)
1041 goto done;
1042 }
1043
1044 done:
1045 strbuf_release(&sb);
1046 return ret;
1047 }
1048
1049 struct fsck_gitmodules_data {
1050 const struct object_id *oid;
1051 struct fsck_options *options;
1052 int ret;
1053 };
1054
1055 static int fsck_gitmodules_fn(const char *var, const char *value,
1056 const struct config_context *ctx UNUSED,
1057 void *vdata)
1058 {
1059 struct fsck_gitmodules_data *data = vdata;
1060 const char *subsection, *key;
1061 size_t subsection_len;
1062 char *name;
1063
1064 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1065 !subsection)
1066 return 0;
1067
1068 name = xmemdupz(subsection, subsection_len);
1069 if (check_submodule_name(name) < 0)
1070 data->ret |= report(data->options,
1071 data->oid, OBJ_BLOB,
1072 FSCK_MSG_GITMODULES_NAME,
1073 "disallowed submodule name: %s",
1074 name);
1075 if (!strcmp(key, "url") && value &&
1076 check_submodule_url(value) < 0)
1077 data->ret |= report(data->options,
1078 data->oid, OBJ_BLOB,
1079 FSCK_MSG_GITMODULES_URL,
1080 "disallowed submodule url: %s",
1081 value);
1082 if (!strcmp(key, "path") && value &&
1083 looks_like_command_line_option(value))
1084 data->ret |= report(data->options,
1085 data->oid, OBJ_BLOB,
1086 FSCK_MSG_GITMODULES_PATH,
1087 "disallowed submodule path: %s",
1088 value);
1089 if (!strcmp(key, "update") && value &&
1090 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1091 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1092 FSCK_MSG_GITMODULES_UPDATE,
1093 "disallowed submodule update setting: %s",
1094 value);
1095 free(name);
1096
1097 return 0;
1098 }
1099
1100 static int fsck_blob(const struct object_id *oid, const char *buf,
1101 unsigned long size, struct fsck_options *options)
1102 {
1103 int ret = 0;
1104
1105 if (object_on_skiplist(options, oid))
1106 return 0;
1107
1108 if (oidset_contains(&options->gitmodules_found, oid)) {
1109 struct config_options config_opts = { 0 };
1110 struct fsck_gitmodules_data data;
1111
1112 oidset_insert(&options->gitmodules_done, oid);
1113
1114 if (!buf) {
1115 /*
1116 * A missing buffer here is a sign that the caller found the
1117 * blob too gigantic to load into memory. Let's just consider
1118 * that an error.
1119 */
1120 return report(options, oid, OBJ_BLOB,
1121 FSCK_MSG_GITMODULES_LARGE,
1122 ".gitmodules too large to parse");
1123 }
1124
1125 data.oid = oid;
1126 data.options = options;
1127 data.ret = 0;
1128 config_opts.error_action = CONFIG_ERROR_SILENT;
1129 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1130 ".gitmodules", buf, size, &data,
1131 CONFIG_SCOPE_UNKNOWN, &config_opts))
1132 data.ret |= report(options, oid, OBJ_BLOB,
1133 FSCK_MSG_GITMODULES_PARSE,
1134 "could not parse gitmodules blob");
1135 ret |= data.ret;
1136 }
1137
1138 if (oidset_contains(&options->gitattributes_found, oid)) {
1139 const char *ptr;
1140
1141 oidset_insert(&options->gitattributes_done, oid);
1142
1143 if (!buf || size > ATTR_MAX_FILE_SIZE) {
1144 /*
1145 * A missing buffer here is a sign that the caller found the
1146 * blob too gigantic to load into memory. Let's just consider
1147 * that an error.
1148 */
1149 return report(options, oid, OBJ_BLOB,
1150 FSCK_MSG_GITATTRIBUTES_LARGE,
1151 ".gitattributes too large to parse");
1152 }
1153
1154 for (ptr = buf; *ptr; ) {
1155 const char *eol = strchrnul(ptr, '\n');
1156 if (eol - ptr >= ATTR_MAX_LINE_LENGTH) {
1157 ret |= report(options, oid, OBJ_BLOB,
1158 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH,
1159 ".gitattributes has too long lines to parse");
1160 break;
1161 }
1162
1163 ptr = *eol ? eol + 1 : eol;
1164 }
1165 }
1166
1167 return ret;
1168 }
1169
1170 int fsck_object(struct object *obj, void *data, unsigned long size,
1171 struct fsck_options *options)
1172 {
1173 if (!obj)
1174 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1175
1176 return fsck_buffer(&obj->oid, obj->type, data, size, options);
1177 }
1178
1179 int fsck_buffer(const struct object_id *oid, enum object_type type,
1180 void *data, unsigned long size,
1181 struct fsck_options *options)
1182 {
1183 if (type == OBJ_BLOB)
1184 return fsck_blob(oid, data, size, options);
1185 if (type == OBJ_TREE)
1186 return fsck_tree(oid, data, size, options);
1187 if (type == OBJ_COMMIT)
1188 return fsck_commit(oid, data, size, options);
1189 if (type == OBJ_TAG)
1190 return fsck_tag(oid, data, size, options);
1191
1192 return report(options, oid, type,
1193 FSCK_MSG_UNKNOWN_TYPE,
1194 "unknown type '%d' (internal fsck error)",
1195 type);
1196 }
1197
1198 int fsck_error_function(struct fsck_options *o,
1199 const struct object_id *oid,
1200 enum object_type object_type UNUSED,
1201 enum fsck_msg_type msg_type,
1202 enum fsck_msg_id msg_id UNUSED,
1203 const char *message)
1204 {
1205 if (msg_type == FSCK_WARN) {
1206 warning("object %s: %s", fsck_describe_object(o, oid), message);
1207 return 0;
1208 }
1209 error("object %s: %s", fsck_describe_object(o, oid), message);
1210 return 1;
1211 }
1212
1213 static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
1214 enum fsck_msg_id msg_missing, enum fsck_msg_id msg_type,
1215 struct fsck_options *options, const char *blob_type)
1216 {
1217 int ret = 0;
1218 struct oidset_iter iter;
1219 const struct object_id *oid;
1220
1221 oidset_iter_init(blobs_found, &iter);
1222 while ((oid = oidset_iter_next(&iter))) {
1223 enum object_type type;
1224 unsigned long size;
1225 char *buf;
1226
1227 if (oidset_contains(blobs_done, oid))
1228 continue;
1229
1230 buf = repo_read_object_file(the_repository, oid, &type, &size);
1231 if (!buf) {
1232 if (is_promisor_object(oid))
1233 continue;
1234 ret |= report(options,
1235 oid, OBJ_BLOB, msg_missing,
1236 "unable to read %s blob", blob_type);
1237 continue;
1238 }
1239
1240 if (type == OBJ_BLOB)
1241 ret |= fsck_blob(oid, buf, size, options);
1242 else
1243 ret |= report(options, oid, type, msg_type,
1244 "non-blob found at %s", blob_type);
1245 free(buf);
1246 }
1247
1248 oidset_clear(blobs_found);
1249 oidset_clear(blobs_done);
1250
1251 return ret;
1252 }
1253
1254 int fsck_finish(struct fsck_options *options)
1255 {
1256 int ret = 0;
1257
1258 ret |= fsck_blobs(&options->gitmodules_found, &options->gitmodules_done,
1259 FSCK_MSG_GITMODULES_MISSING, FSCK_MSG_GITMODULES_BLOB,
1260 options, ".gitmodules");
1261 ret |= fsck_blobs(&options->gitattributes_found, &options->gitattributes_done,
1262 FSCK_MSG_GITATTRIBUTES_MISSING, FSCK_MSG_GITATTRIBUTES_BLOB,
1263 options, ".gitattributes");
1264
1265 return ret;
1266 }
1267
1268 int git_fsck_config(const char *var, const char *value,
1269 const struct config_context *ctx, void *cb)
1270 {
1271 struct fsck_options *options = cb;
1272 const char *msg_id;
1273
1274 if (strcmp(var, "fsck.skiplist") == 0) {
1275 const char *path;
1276 struct strbuf sb = STRBUF_INIT;
1277
1278 if (git_config_pathname(&path, var, value))
1279 return 1;
1280 strbuf_addf(&sb, "skiplist=%s", path);
1281 free((char *)path);
1282 fsck_set_msg_types(options, sb.buf);
1283 strbuf_release(&sb);
1284 return 0;
1285 }
1286
1287 if (skip_prefix(var, "fsck.", &msg_id)) {
1288 if (!value)
1289 return config_error_nonbool(var);
1290 fsck_set_msg_type(options, msg_id, value);
1291 return 0;
1292 }
1293
1294 return git_default_config(var, value, ctx, cb);
1295 }
1296
1297 /*
1298 * Custom error callbacks that are used in more than one place.
1299 */
1300
1301 int fsck_error_cb_print_missing_gitmodules(struct fsck_options *o,
1302 const struct object_id *oid,
1303 enum object_type object_type,
1304 enum fsck_msg_type msg_type,
1305 enum fsck_msg_id msg_id,
1306 const char *message)
1307 {
1308 if (msg_id == FSCK_MSG_GITMODULES_MISSING) {
1309 puts(oid_to_hex(oid));
1310 return 0;
1311 }
1312 return fsck_error_function(o, oid, object_type, msg_type, msg_id, message);
1313 }