Merge branch 'jk/drop-hg-to-git'
[git/gitster.git] / fsck.c
blob78af29d26459e1d392d8a64ff72dd5fed575daff
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"
25 static ssize_t max_tree_entry_len = 4096;
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 }
38 #undef MSG_ID
39 #undef STR
41 static void prepare_msg_ids(void)
43 int i;
45 if (msg_id_info[0].downcased)
46 return;
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);
54 msg_id_info[i].downcased = q;
55 while (*p)
56 if (*p == '_')
57 p++;
58 else
59 *(q)++ = tolower(*(p)++);
60 *q = '\0';
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++);
74 *q = '\0';
78 static int parse_msg_id(const char *text)
80 int i;
82 prepare_msg_ids();
84 for (i = 0; i < FSCK_MSG_MAX; i++)
85 if (!strcmp(text, msg_id_info[i].downcased))
86 return i;
88 return -1;
91 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
93 int i;
95 prepare_msg_ids();
97 for (i = 0; i < FSCK_MSG_MAX; i++)
98 list_config_item(list, prefix, msg_id_info[i].camelcased);
101 static enum fsck_msg_type fsck_msg_type(enum fsck_msg_id msg_id,
102 struct fsck_options *options)
104 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
106 if (!options->msg_type) {
107 enum fsck_msg_type msg_type = msg_id_info[msg_id].msg_type;
109 if (options->strict && msg_type == FSCK_WARN)
110 msg_type = FSCK_ERROR;
111 return msg_type;
114 return options->msg_type[msg_id];
117 static enum fsck_msg_type parse_msg_type(const char *str)
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);
129 int is_valid_msg_type(const char *msg_id, const char *msg_type)
131 if (parse_msg_id(msg_id) < 0)
132 return 0;
133 parse_msg_type(msg_type);
134 return 1;
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)
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;
150 options->msg_type[msg_id] = msg_type;
153 void fsck_set_msg_type(struct fsck_options *options,
154 const char *msg_id_str, const char *msg_type_str)
156 int msg_id = parse_msg_id(msg_id_str);
157 char *to_free = NULL;
158 enum fsck_msg_type msg_type;
160 if (msg_id < 0)
161 die("Unhandled message id: %s", msg_id_str);
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);
173 msg_type = parse_msg_type(msg_type_str);
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);
178 fsck_set_msg_type_from_ids(options, msg_id, msg_type);
179 free(to_free);
182 void fsck_set_msg_types(struct fsck_options *options, const char *values)
184 char *buf = xstrdup(values), *to_free = buf;
185 int done = 0;
187 while (!done) {
188 int len = strcspn(buf, " ,|"), equal;
190 done = !buf[len];
191 if (!len) {
192 buf++;
193 continue;
195 buf[len] = '\0';
197 for (equal = 0;
198 equal < len && buf[equal] != '=' && buf[equal] != ':';
199 equal++)
200 buf[equal] = tolower(buf[equal]);
201 buf[equal] = '\0';
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;
211 if (equal == len)
212 die("Missing '=': '%s'", buf);
214 fsck_set_msg_type(options, buf, buf + equal + 1);
215 buf += len + 1;
217 free(to_free);
220 static int object_on_skiplist(struct fsck_options *opts,
221 const struct object_id *oid)
223 return opts && oid && oidset_contains(&opts->skiplist, oid);
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, ...)
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;
236 if (msg_type == FSCK_IGNORE)
237 return 0;
239 if (object_on_skiplist(options, oid))
240 return 0;
242 if (msg_type == FSCK_FATAL)
243 msg_type = FSCK_ERROR;
244 else if (msg_type == FSCK_INFO)
245 msg_type = FSCK_WARN;
247 prepare_msg_ids();
248 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
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);
257 return result;
260 void fsck_enable_object_names(struct fsck_options *options)
262 if (!options->object_names)
263 options->object_names = kh_init_oid_map();
266 const char *fsck_get_object_name(struct fsck_options *options,
267 const struct object_id *oid)
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);
278 void fsck_put_object_name(struct fsck_options *options,
279 const struct object_id *oid,
280 const char *fmt, ...)
282 va_list ap;
283 struct strbuf buf = STRBUF_INIT;
284 khiter_t pos;
285 int hashret;
287 if (!options->object_names)
288 return;
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);
299 const char *fsck_describe_object(struct fsck_options *options,
300 const struct object_id *oid)
302 static struct strbuf bufs[] = {
303 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
305 static int b = 0;
306 struct strbuf *buf;
307 const char *name = fsck_get_object_name(options, oid);
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);
316 return buf->buf;
319 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
321 struct tree_desc desc;
322 struct name_entry entry;
323 int res = 0;
324 const char *name;
326 if (parse_tree(tree))
327 return -1;
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;
337 if (S_ISGITLINK(entry.mode))
338 continue;
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);
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);
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);
359 if (result < 0)
360 return result;
361 if (!res)
362 res = result;
364 return res;
367 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
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;
375 if (repo_parse_commit(the_repository, commit))
376 return -1;
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);
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;
389 parents = commit->parents;
390 if (name && parents) {
391 int len = strlen(name), power;
393 if (len && name[len - 1] == '^') {
394 generation = 1;
395 name_prefix_len = len - 1;
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;
412 while (parents) {
413 if (name) {
414 struct object_id *oid = &parents->item->object.oid;
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);
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;
433 return res;
436 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
438 const char *name = fsck_get_object_name(options, &tag->object.oid);
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);
447 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
449 if (!obj)
450 return -1;
452 if (obj->type == OBJ_NONE)
453 parse_object(the_repository, &obj->oid);
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;
471 struct name_stack {
472 const char **names;
473 size_t nr, alloc;
476 static void name_stack_push(struct name_stack *stack, const char *name)
478 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
479 stack->names[stack->nr++] = name;
482 static const char *name_stack_pop(struct name_stack *stack)
484 return stack->nr ? stack->names[--stack->nr] : NULL;
487 static void name_stack_clear(struct name_stack *stack)
489 FREE_AND_NULL(stack->names);
490 stack->nr = stack->alloc = 0;
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.
498 * So a directory called "a" is ordered _after_ a file
499 * called "a.c", because "a/" sorts after "a.c".
501 #define TREE_UNORDERED (-1)
502 #define TREE_HAS_DUPS (-2)
504 static int is_less_than_slash(unsigned char c)
506 return '\0' < c && c < '/';
509 static int verify_ordered(unsigned mode1, const char *name1,
510 unsigned mode2, const char *name2,
511 struct name_stack *candidates)
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;
519 cmp = memcmp(name1, name2, len);
520 if (cmp < 0)
521 return 0;
522 if (cmp > 0)
523 return TREE_UNORDERED;
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.
530 c1 = name1[len];
531 c2 = name2[len];
532 if (!c1 && !c2)
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.
538 return TREE_HAS_DUPS;
539 if (!c1 && S_ISDIR(mode1))
540 c1 = '/';
541 if (!c2 && S_ISDIR(mode2))
542 c2 = '/';
545 * There can be non-consecutive duplicates due to the implicitly
546 * added slash, e.g.:
548 * foo
549 * foo.bar
550 * foo.bar.baz
551 * foo.bar/
552 * foo/
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.
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);
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;
578 return c1 < c2 ? 0 : TREE_UNORDERED;
581 static int fsck_tree(const struct object_id *tree_oid,
582 const char *buffer, unsigned long size,
583 struct fsck_options *options)
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 };
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;
610 o_mode = 0;
611 o_name = NULL;
613 while (desc.size) {
614 unsigned short mode;
615 const char *name, *backslash;
616 const struct object_id *entry_oid;
618 entry_oid = tree_entry_extract(&desc, &name, &mode);
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;
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");
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");
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");
663 if ((backslash = strchr(name, '\\'))) {
664 while (backslash) {
665 backslash++;
666 has_dotgit |= is_ntfs_dotgit(backslash);
667 if (is_ntfs_dotgitmodules(backslash)) {
668 if (!S_ISLNK(mode))
669 oidset_insert(&options->gitmodules_found,
670 entry_oid);
671 else
672 retval += report(options, tree_oid, OBJ_TREE,
673 FSCK_MSG_GITMODULES_SYMLINK,
674 ".gitmodules is a symbolic link");
676 backslash = strchr(backslash, '\\');
680 if (update_tree_entry_gently(&desc)) {
681 retval += report(options, tree_oid, OBJ_TREE,
682 FSCK_MSG_BAD_TREE,
683 "cannot be parsed as a tree");
684 break;
687 switch (mode) {
689 * Standard modes..
691 case S_IFREG | 0755:
692 case S_IFREG | 0644:
693 case S_IFLNK:
694 case S_IFDIR:
695 case S_IFGITLINK:
696 break;
698 * This is nonstandard, but we had a few of these
699 * early on when we honored the full set of mode
700 * bits..
702 case S_IFREG | 0664:
703 if (!options->strict)
704 break;
705 /* fallthrough */
706 default:
707 has_bad_modes = 1;
710 if (o_name) {
711 switch (verify_ordered(o_mode, o_name, mode, name,
712 &df_dup_candidates)) {
713 case TREE_UNORDERED:
714 not_properly_sorted = 1;
715 break;
716 case TREE_HAS_DUPS:
717 has_dup_entries = 1;
718 break;
719 default:
720 break;
724 o_mode = mode;
725 o_name = name;
728 name_stack_clear(&df_dup_candidates);
730 if (has_null_sha1)
731 retval += report(options, tree_oid, OBJ_TREE,
732 FSCK_MSG_NULL_SHA1,
733 "contains entries pointing to null sha1");
734 if (has_full_path)
735 retval += report(options, tree_oid, OBJ_TREE,
736 FSCK_MSG_FULL_PATHNAME,
737 "contains full pathnames");
738 if (has_empty_name)
739 retval += report(options, tree_oid, OBJ_TREE,
740 FSCK_MSG_EMPTY_NAME,
741 "contains empty pathname");
742 if (has_dot)
743 retval += report(options, tree_oid, OBJ_TREE,
744 FSCK_MSG_HAS_DOT,
745 "contains '.'");
746 if (has_dotdot)
747 retval += report(options, tree_oid, OBJ_TREE,
748 FSCK_MSG_HAS_DOTDOT,
749 "contains '..'");
750 if (has_dotgit)
751 retval += report(options, tree_oid, OBJ_TREE,
752 FSCK_MSG_HAS_DOTGIT,
753 "contains '.git'");
754 if (has_zero_pad)
755 retval += report(options, tree_oid, OBJ_TREE,
756 FSCK_MSG_ZERO_PADDED_FILEMODE,
757 "contains zero-padded file modes");
758 if (has_bad_modes)
759 retval += report(options, tree_oid, OBJ_TREE,
760 FSCK_MSG_BAD_FILEMODE,
761 "contains bad file modes");
762 if (has_dup_entries)
763 retval += report(options, tree_oid, OBJ_TREE,
764 FSCK_MSG_DUPLICATE_ENTRIES,
765 "contains duplicate file entries");
766 if (not_properly_sorted)
767 retval += report(options, tree_oid, OBJ_TREE,
768 FSCK_MSG_TREE_NOT_SORTED,
769 "not properly sorted");
770 if (has_large_name)
771 retval += report(options, tree_oid, OBJ_TREE,
772 FSCK_MSG_LARGE_PATHNAME,
773 "contains excessively large pathname");
774 return retval;
778 * Confirm that the headers of a commit or tag object end in a reasonable way,
779 * either with the usual "\n\n" separator, or at least with a trailing newline
780 * on the final header line.
782 * This property is important for the memory safety of our callers. It allows
783 * them to scan the buffer linewise without constantly checking the remaining
784 * size as long as:
786 * - they check that there are bytes left in the buffer at the start of any
787 * line (i.e., that the last newline they saw was not the final one we
788 * found here)
790 * - any intra-line scanning they do will stop at a newline, which will worst
791 * case hit the newline we found here as the end-of-header. This makes it
792 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
794 static int verify_headers(const void *data, unsigned long size,
795 const struct object_id *oid, enum object_type type,
796 struct fsck_options *options)
798 const char *buffer = (const char *)data;
799 unsigned long i;
801 for (i = 0; i < size; i++) {
802 switch (buffer[i]) {
803 case '\0':
804 return report(options, oid, type,
805 FSCK_MSG_NUL_IN_HEADER,
806 "unterminated header: NUL at offset %ld", i);
807 case '\n':
808 if (i + 1 < size && buffer[i + 1] == '\n')
809 return 0;
814 * We did not find double-LF that separates the header
815 * and the body. Not having a body is not a crime but
816 * we do want to see the terminating LF for the last header
817 * line.
819 if (size && buffer[size - 1] == '\n')
820 return 0;
822 return report(options, oid, type,
823 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
826 static int fsck_ident(const char **ident,
827 const struct object_id *oid, enum object_type type,
828 struct fsck_options *options)
830 const char *p = *ident;
831 char *end;
833 *ident = strchrnul(*ident, '\n');
834 if (**ident == '\n')
835 (*ident)++;
837 if (*p == '<')
838 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
839 p += strcspn(p, "<>\n");
840 if (*p == '>')
841 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
842 if (*p != '<')
843 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
844 if (p[-1] != ' ')
845 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
846 p++;
847 p += strcspn(p, "<>\n");
848 if (*p != '>')
849 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
850 p++;
851 if (*p != ' ')
852 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
853 p++;
855 * Our timestamp parser is based on the C strto*() functions, which
856 * will happily eat whitespace, including the newline that is supposed
857 * to prevent us walking past the end of the buffer. So do our own
858 * scan, skipping linear whitespace but not newlines, and then
859 * confirming we found a digit. We _could_ be even more strict here,
860 * as we really expect only a single space, but since we have
861 * traditionally allowed extra whitespace, we'll continue to do so.
863 while (*p == ' ' || *p == '\t')
864 p++;
865 if (!isdigit(*p))
866 return report(options, oid, type, FSCK_MSG_BAD_DATE,
867 "invalid author/committer line - bad date");
868 if (*p == '0' && p[1] != ' ')
869 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
870 if (date_overflows(parse_timestamp(p, &end, 10)))
871 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
872 if ((end == p || *end != ' '))
873 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
874 p = end + 1;
875 if ((*p != '+' && *p != '-') ||
876 !isdigit(p[1]) ||
877 !isdigit(p[2]) ||
878 !isdigit(p[3]) ||
879 !isdigit(p[4]) ||
880 (p[5] != '\n'))
881 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
882 p += 6;
883 return 0;
886 static int fsck_commit(const struct object_id *oid,
887 const char *buffer, unsigned long size,
888 struct fsck_options *options)
890 struct object_id tree_oid, parent_oid;
891 unsigned author_count;
892 int err;
893 const char *buffer_begin = buffer;
894 const char *buffer_end = buffer + size;
895 const char *p;
898 * We _must_ stop parsing immediately if this reports failure, as the
899 * memory safety of the rest of the function depends on it. See the
900 * comment above the definition of verify_headers() for more details.
902 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
903 return -1;
905 if (buffer >= buffer_end || !skip_prefix(buffer, "tree ", &buffer))
906 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
907 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
908 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
909 if (err)
910 return err;
912 buffer = p + 1;
913 while (buffer < buffer_end && skip_prefix(buffer, "parent ", &buffer)) {
914 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
915 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
916 if (err)
917 return err;
919 buffer = p + 1;
921 author_count = 0;
922 while (buffer < buffer_end && skip_prefix(buffer, "author ", &buffer)) {
923 author_count++;
924 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
925 if (err)
926 return err;
928 if (author_count < 1)
929 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
930 else if (author_count > 1)
931 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
932 if (err)
933 return err;
934 if (buffer >= buffer_end || !skip_prefix(buffer, "committer ", &buffer))
935 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
936 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
937 if (err)
938 return err;
939 if (memchr(buffer_begin, '\0', size)) {
940 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
941 "NUL byte in the commit object body");
942 if (err)
943 return err;
945 return 0;
948 static int fsck_tag(const struct object_id *oid, const char *buffer,
949 unsigned long size, struct fsck_options *options)
951 struct object_id tagged_oid;
952 int tagged_type;
953 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
954 &tagged_type);
957 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
958 unsigned long size, struct fsck_options *options,
959 struct object_id *tagged_oid,
960 int *tagged_type)
962 int ret = 0;
963 char *eol;
964 struct strbuf sb = STRBUF_INIT;
965 const char *buffer_end = buffer + size;
966 const char *p;
969 * We _must_ stop parsing immediately if this reports failure, as the
970 * memory safety of the rest of the function depends on it. See the
971 * comment above the definition of verify_headers() for more details.
973 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
974 if (ret)
975 goto done;
977 if (buffer >= buffer_end || !skip_prefix(buffer, "object ", &buffer)) {
978 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
979 goto done;
981 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
982 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
983 if (ret)
984 goto done;
986 buffer = p + 1;
988 if (buffer >= buffer_end || !skip_prefix(buffer, "type ", &buffer)) {
989 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
990 goto done;
992 eol = memchr(buffer, '\n', buffer_end - buffer);
993 if (!eol) {
994 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
995 goto done;
997 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
998 if (*tagged_type < 0)
999 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
1000 if (ret)
1001 goto done;
1002 buffer = eol + 1;
1004 if (buffer >= buffer_end || !skip_prefix(buffer, "tag ", &buffer)) {
1005 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
1006 goto done;
1008 eol = memchr(buffer, '\n', buffer_end - buffer);
1009 if (!eol) {
1010 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
1011 goto done;
1013 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
1014 if (check_refname_format(sb.buf, 0)) {
1015 ret = report(options, oid, OBJ_TAG,
1016 FSCK_MSG_BAD_TAG_NAME,
1017 "invalid 'tag' name: %.*s",
1018 (int)(eol - buffer), buffer);
1019 if (ret)
1020 goto done;
1022 buffer = eol + 1;
1024 if (buffer >= buffer_end || !skip_prefix(buffer, "tagger ", &buffer)) {
1025 /* early tags do not contain 'tagger' lines; warn only */
1026 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
1027 if (ret)
1028 goto done;
1030 else
1031 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
1033 if (buffer < buffer_end && !starts_with(buffer, "\n")) {
1035 * The verify_headers() check will allow
1036 * e.g. "[...]tagger <tagger>\nsome
1037 * garbage\n\nmessage" to pass, thinking "some
1038 * garbage" could be a custom header. E.g. "mktag"
1039 * doesn't want any unknown headers.
1041 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1042 if (ret)
1043 goto done;
1046 done:
1047 strbuf_release(&sb);
1048 return ret;
1051 struct fsck_gitmodules_data {
1052 const struct object_id *oid;
1053 struct fsck_options *options;
1054 int ret;
1057 static int fsck_gitmodules_fn(const char *var, const char *value,
1058 const struct config_context *ctx UNUSED,
1059 void *vdata)
1061 struct fsck_gitmodules_data *data = vdata;
1062 const char *subsection, *key;
1063 size_t subsection_len;
1064 char *name;
1066 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1067 !subsection)
1068 return 0;
1070 name = xmemdupz(subsection, subsection_len);
1071 if (check_submodule_name(name) < 0)
1072 data->ret |= report(data->options,
1073 data->oid, OBJ_BLOB,
1074 FSCK_MSG_GITMODULES_NAME,
1075 "disallowed submodule name: %s",
1076 name);
1077 if (!strcmp(key, "url") && value &&
1078 check_submodule_url(value) < 0)
1079 data->ret |= report(data->options,
1080 data->oid, OBJ_BLOB,
1081 FSCK_MSG_GITMODULES_URL,
1082 "disallowed submodule url: %s",
1083 value);
1084 if (!strcmp(key, "path") && value &&
1085 looks_like_command_line_option(value))
1086 data->ret |= report(data->options,
1087 data->oid, OBJ_BLOB,
1088 FSCK_MSG_GITMODULES_PATH,
1089 "disallowed submodule path: %s",
1090 value);
1091 if (!strcmp(key, "update") && value &&
1092 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1093 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1094 FSCK_MSG_GITMODULES_UPDATE,
1095 "disallowed submodule update setting: %s",
1096 value);
1097 free(name);
1099 return 0;
1102 static int fsck_blob(const struct object_id *oid, const char *buf,
1103 unsigned long size, struct fsck_options *options)
1105 int ret = 0;
1107 if (object_on_skiplist(options, oid))
1108 return 0;
1110 if (oidset_contains(&options->gitmodules_found, oid)) {
1111 struct config_options config_opts = { 0 };
1112 struct fsck_gitmodules_data data;
1114 oidset_insert(&options->gitmodules_done, oid);
1116 if (!buf) {
1118 * A missing buffer here is a sign that the caller found the
1119 * blob too gigantic to load into memory. Let's just consider
1120 * that an error.
1122 return report(options, oid, OBJ_BLOB,
1123 FSCK_MSG_GITMODULES_LARGE,
1124 ".gitmodules too large to parse");
1127 data.oid = oid;
1128 data.options = options;
1129 data.ret = 0;
1130 config_opts.error_action = CONFIG_ERROR_SILENT;
1131 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1132 ".gitmodules", buf, size, &data,
1133 CONFIG_SCOPE_UNKNOWN, &config_opts))
1134 data.ret |= report(options, oid, OBJ_BLOB,
1135 FSCK_MSG_GITMODULES_PARSE,
1136 "could not parse gitmodules blob");
1137 ret |= data.ret;
1140 if (oidset_contains(&options->gitattributes_found, oid)) {
1141 const char *ptr;
1143 oidset_insert(&options->gitattributes_done, oid);
1145 if (!buf || size > ATTR_MAX_FILE_SIZE) {
1147 * A missing buffer here is a sign that the caller found the
1148 * blob too gigantic to load into memory. Let's just consider
1149 * that an error.
1151 return report(options, oid, OBJ_BLOB,
1152 FSCK_MSG_GITATTRIBUTES_LARGE,
1153 ".gitattributes too large to parse");
1156 for (ptr = buf; *ptr; ) {
1157 const char *eol = strchrnul(ptr, '\n');
1158 if (eol - ptr >= ATTR_MAX_LINE_LENGTH) {
1159 ret |= report(options, oid, OBJ_BLOB,
1160 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH,
1161 ".gitattributes has too long lines to parse");
1162 break;
1165 ptr = *eol ? eol + 1 : eol;
1169 return ret;
1172 int fsck_object(struct object *obj, void *data, unsigned long size,
1173 struct fsck_options *options)
1175 if (!obj)
1176 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1178 return fsck_buffer(&obj->oid, obj->type, data, size, options);
1181 int fsck_buffer(const struct object_id *oid, enum object_type type,
1182 void *data, unsigned long size,
1183 struct fsck_options *options)
1185 if (type == OBJ_BLOB)
1186 return fsck_blob(oid, data, size, options);
1187 if (type == OBJ_TREE)
1188 return fsck_tree(oid, data, size, options);
1189 if (type == OBJ_COMMIT)
1190 return fsck_commit(oid, data, size, options);
1191 if (type == OBJ_TAG)
1192 return fsck_tag(oid, data, size, options);
1194 return report(options, oid, type,
1195 FSCK_MSG_UNKNOWN_TYPE,
1196 "unknown type '%d' (internal fsck error)",
1197 type);
1200 int fsck_error_function(struct fsck_options *o,
1201 const struct object_id *oid,
1202 enum object_type object_type UNUSED,
1203 enum fsck_msg_type msg_type,
1204 enum fsck_msg_id msg_id UNUSED,
1205 const char *message)
1207 if (msg_type == FSCK_WARN) {
1208 warning("object %s: %s", fsck_describe_object(o, oid), message);
1209 return 0;
1211 error("object %s: %s", fsck_describe_object(o, oid), message);
1212 return 1;
1215 static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
1216 enum fsck_msg_id msg_missing, enum fsck_msg_id msg_type,
1217 struct fsck_options *options, const char *blob_type)
1219 int ret = 0;
1220 struct oidset_iter iter;
1221 const struct object_id *oid;
1223 oidset_iter_init(blobs_found, &iter);
1224 while ((oid = oidset_iter_next(&iter))) {
1225 enum object_type type;
1226 unsigned long size;
1227 char *buf;
1229 if (oidset_contains(blobs_done, oid))
1230 continue;
1232 buf = repo_read_object_file(the_repository, oid, &type, &size);
1233 if (!buf) {
1234 if (is_promisor_object(oid))
1235 continue;
1236 ret |= report(options,
1237 oid, OBJ_BLOB, msg_missing,
1238 "unable to read %s blob", blob_type);
1239 continue;
1242 if (type == OBJ_BLOB)
1243 ret |= fsck_blob(oid, buf, size, options);
1244 else
1245 ret |= report(options, oid, type, msg_type,
1246 "non-blob found at %s", blob_type);
1247 free(buf);
1250 oidset_clear(blobs_found);
1251 oidset_clear(blobs_done);
1253 return ret;
1256 int fsck_finish(struct fsck_options *options)
1258 int ret = 0;
1260 ret |= fsck_blobs(&options->gitmodules_found, &options->gitmodules_done,
1261 FSCK_MSG_GITMODULES_MISSING, FSCK_MSG_GITMODULES_BLOB,
1262 options, ".gitmodules");
1263 ret |= fsck_blobs(&options->gitattributes_found, &options->gitattributes_done,
1264 FSCK_MSG_GITATTRIBUTES_MISSING, FSCK_MSG_GITATTRIBUTES_BLOB,
1265 options, ".gitattributes");
1267 return ret;
1270 int git_fsck_config(const char *var, const char *value,
1271 const struct config_context *ctx, void *cb)
1273 struct fsck_options *options = cb;
1274 const char *msg_id;
1276 if (strcmp(var, "fsck.skiplist") == 0) {
1277 const char *path;
1278 struct strbuf sb = STRBUF_INIT;
1280 if (git_config_pathname(&path, var, value))
1281 return 1;
1282 strbuf_addf(&sb, "skiplist=%s", path);
1283 free((char *)path);
1284 fsck_set_msg_types(options, sb.buf);
1285 strbuf_release(&sb);
1286 return 0;
1289 if (skip_prefix(var, "fsck.", &msg_id)) {
1290 if (!value)
1291 return config_error_nonbool(var);
1292 fsck_set_msg_type(options, msg_id, value);
1293 return 0;
1296 return git_default_config(var, value, ctx, cb);
1300 * Custom error callbacks that are used in more than one place.
1303 int fsck_error_cb_print_missing_gitmodules(struct fsck_options *o,
1304 const struct object_id *oid,
1305 enum object_type object_type,
1306 enum fsck_msg_type msg_type,
1307 enum fsck_msg_id msg_id,
1308 const char *message)
1310 if (msg_id == FSCK_MSG_GITMODULES_MISSING) {
1311 puts(oid_to_hex(oid));
1312 return 0;
1314 return fsck_error_function(o, oid, object_type, msg_type, msg_id, message);