mingw(is_msys2_sh): handle forward slashes in the `sh.exe` path, too
[git/debian.git] / fsck.c
blobeea71454701c6dd5f48c60e2b6d9cd25388971c1
1 #define USE_THE_REPOSITORY_VARIABLE
3 #include "git-compat-util.h"
4 #include "date.h"
5 #include "dir.h"
6 #include "hex.h"
7 #include "object-store-ll.h"
8 #include "path.h"
9 #include "repository.h"
10 #include "object.h"
11 #include "attr.h"
12 #include "blob.h"
13 #include "tree.h"
14 #include "tree-walk.h"
15 #include "commit.h"
16 #include "tag.h"
17 #include "fsck.h"
18 #include "refs.h"
19 #include "url.h"
20 #include "utf8.h"
21 #include "oidset.h"
22 #include "packfile.h"
23 #include "submodule-config.h"
24 #include "config.h"
25 #include "help.h"
27 static ssize_t max_tree_entry_len = 4096;
29 #define STR(x) #x
30 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
31 static struct {
32 const char *id_string;
33 const char *downcased;
34 const char *camelcased;
35 enum fsck_msg_type msg_type;
36 } msg_id_info[FSCK_MSG_MAX + 1] = {
37 FOREACH_FSCK_MSG_ID(MSG_ID)
38 { NULL, NULL, NULL, -1 }
40 #undef MSG_ID
41 #undef STR
43 static void prepare_msg_ids(void)
45 int i;
47 if (msg_id_info[0].downcased)
48 return;
50 /* convert id_string to lower case, without underscores. */
51 for (i = 0; i < FSCK_MSG_MAX; i++) {
52 const char *p = msg_id_info[i].id_string;
53 int len = strlen(p);
54 char *q = xmalloc(len);
56 msg_id_info[i].downcased = q;
57 while (*p)
58 if (*p == '_')
59 p++;
60 else
61 *(q)++ = tolower(*(p)++);
62 *q = '\0';
64 p = msg_id_info[i].id_string;
65 q = xmalloc(len);
66 msg_id_info[i].camelcased = q;
67 while (*p) {
68 if (*p == '_') {
69 p++;
70 if (*p)
71 *q++ = *p++;
72 } else {
73 *q++ = tolower(*p++);
76 *q = '\0';
80 static int parse_msg_id(const char *text)
82 int i;
84 prepare_msg_ids();
86 for (i = 0; i < FSCK_MSG_MAX; i++)
87 if (!strcmp(text, msg_id_info[i].downcased))
88 return i;
90 return -1;
93 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
95 int i;
97 prepare_msg_ids();
99 for (i = 0; i < FSCK_MSG_MAX; i++)
100 list_config_item(list, prefix, msg_id_info[i].camelcased);
103 static enum fsck_msg_type fsck_msg_type(enum fsck_msg_id msg_id,
104 struct fsck_options *options)
106 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
108 if (!options->msg_type) {
109 enum fsck_msg_type msg_type = msg_id_info[msg_id].msg_type;
111 if (options->strict && msg_type == FSCK_WARN)
112 msg_type = FSCK_ERROR;
113 return msg_type;
116 return options->msg_type[msg_id];
119 static enum fsck_msg_type parse_msg_type(const char *str)
121 if (!strcmp(str, "error"))
122 return FSCK_ERROR;
123 else if (!strcmp(str, "warn"))
124 return FSCK_WARN;
125 else if (!strcmp(str, "ignore"))
126 return FSCK_IGNORE;
127 else
128 die("Unknown fsck message type: '%s'", str);
131 int is_valid_msg_type(const char *msg_id, const char *msg_type)
133 if (parse_msg_id(msg_id) < 0)
134 return 0;
135 parse_msg_type(msg_type);
136 return 1;
139 void fsck_set_msg_type_from_ids(struct fsck_options *options,
140 enum fsck_msg_id msg_id,
141 enum fsck_msg_type msg_type)
143 if (!options->msg_type) {
144 int i;
145 enum fsck_msg_type *severity;
146 ALLOC_ARRAY(severity, FSCK_MSG_MAX);
147 for (i = 0; i < FSCK_MSG_MAX; i++)
148 severity[i] = fsck_msg_type(i, options);
149 options->msg_type = severity;
152 options->msg_type[msg_id] = msg_type;
155 void fsck_set_msg_type(struct fsck_options *options,
156 const char *msg_id_str, const char *msg_type_str)
158 int msg_id = parse_msg_id(msg_id_str);
159 char *to_free = NULL;
160 enum fsck_msg_type msg_type;
162 if (msg_id < 0)
163 die("Unhandled message id: %s", msg_id_str);
165 if (msg_id == FSCK_MSG_LARGE_PATHNAME) {
166 const char *colon = strchr(msg_type_str, ':');
167 if (colon) {
168 msg_type_str = to_free =
169 xmemdupz(msg_type_str, colon - msg_type_str);
170 colon++;
171 if (!git_parse_ssize_t(colon, &max_tree_entry_len))
172 die("unable to parse max tree entry len: %s", colon);
175 msg_type = parse_msg_type(msg_type_str);
177 if (msg_type != FSCK_ERROR && msg_id_info[msg_id].msg_type == FSCK_FATAL)
178 die("Cannot demote %s to %s", msg_id_str, msg_type_str);
180 fsck_set_msg_type_from_ids(options, msg_id, msg_type);
181 free(to_free);
184 void fsck_set_msg_types(struct fsck_options *options, const char *values)
186 char *buf = xstrdup(values), *to_free = buf;
187 int done = 0;
189 while (!done) {
190 int len = strcspn(buf, " ,|"), equal;
192 done = !buf[len];
193 if (!len) {
194 buf++;
195 continue;
197 buf[len] = '\0';
199 for (equal = 0;
200 equal < len && buf[equal] != '=' && buf[equal] != ':';
201 equal++)
202 buf[equal] = tolower(buf[equal]);
203 buf[equal] = '\0';
205 if (!strcmp(buf, "skiplist")) {
206 if (equal == len)
207 die("skiplist requires a path");
208 oidset_parse_file(&options->skiplist, buf + equal + 1,
209 the_repository->hash_algo);
210 buf += len + 1;
211 continue;
214 if (equal == len)
215 die("Missing '=': '%s'", buf);
217 fsck_set_msg_type(options, buf, buf + equal + 1);
218 buf += len + 1;
220 free(to_free);
223 static int object_on_skiplist(struct fsck_options *opts,
224 const struct object_id *oid)
226 return opts && oid && oidset_contains(&opts->skiplist, oid);
229 __attribute__((format (printf, 5, 6)))
230 static int report(struct fsck_options *options,
231 const struct object_id *oid, enum object_type object_type,
232 enum fsck_msg_id msg_id, const char *fmt, ...)
234 va_list ap;
235 struct strbuf sb = STRBUF_INIT;
236 enum fsck_msg_type msg_type = fsck_msg_type(msg_id, options);
237 int result;
239 if (msg_type == FSCK_IGNORE)
240 return 0;
242 if (object_on_skiplist(options, oid))
243 return 0;
245 if (msg_type == FSCK_FATAL)
246 msg_type = FSCK_ERROR;
247 else if (msg_type == FSCK_INFO)
248 msg_type = FSCK_WARN;
250 prepare_msg_ids();
251 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
253 va_start(ap, fmt);
254 strbuf_vaddf(&sb, fmt, ap);
255 result = options->error_func(options, oid, object_type,
256 msg_type, msg_id, sb.buf);
257 strbuf_release(&sb);
258 va_end(ap);
260 return result;
263 void fsck_enable_object_names(struct fsck_options *options)
265 if (!options->object_names)
266 options->object_names = kh_init_oid_map();
269 const char *fsck_get_object_name(struct fsck_options *options,
270 const struct object_id *oid)
272 khiter_t pos;
273 if (!options->object_names)
274 return NULL;
275 pos = kh_get_oid_map(options->object_names, *oid);
276 if (pos >= kh_end(options->object_names))
277 return NULL;
278 return kh_value(options->object_names, pos);
281 void fsck_put_object_name(struct fsck_options *options,
282 const struct object_id *oid,
283 const char *fmt, ...)
285 va_list ap;
286 struct strbuf buf = STRBUF_INIT;
287 khiter_t pos;
288 int hashret;
290 if (!options->object_names)
291 return;
293 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
294 if (!hashret)
295 return;
296 va_start(ap, fmt);
297 strbuf_vaddf(&buf, fmt, ap);
298 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
299 va_end(ap);
302 const char *fsck_describe_object(struct fsck_options *options,
303 const struct object_id *oid)
305 static struct strbuf bufs[] = {
306 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
308 static int b = 0;
309 struct strbuf *buf;
310 const char *name = fsck_get_object_name(options, oid);
312 buf = bufs + b;
313 b = (b + 1) % ARRAY_SIZE(bufs);
314 strbuf_reset(buf);
315 strbuf_addstr(buf, oid_to_hex(oid));
316 if (name)
317 strbuf_addf(buf, " (%s)", name);
319 return buf->buf;
322 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
324 struct tree_desc desc;
325 struct name_entry entry;
326 int res = 0;
327 const char *name;
329 if (parse_tree(tree))
330 return -1;
332 name = fsck_get_object_name(options, &tree->object.oid);
333 if (init_tree_desc_gently(&desc, &tree->object.oid,
334 tree->buffer, tree->size, 0))
335 return -1;
336 while (tree_entry_gently(&desc, &entry)) {
337 struct object *obj;
338 int result;
340 if (S_ISGITLINK(entry.mode))
341 continue;
343 if (S_ISDIR(entry.mode)) {
344 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
345 if (name && obj)
346 fsck_put_object_name(options, &entry.oid, "%s%s/",
347 name, entry.path);
348 result = options->walk(obj, OBJ_TREE, data, options);
350 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
351 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
352 if (name && obj)
353 fsck_put_object_name(options, &entry.oid, "%s%s",
354 name, entry.path);
355 result = options->walk(obj, OBJ_BLOB, data, options);
357 else {
358 result = error("in tree %s: entry %s has bad mode %.6o",
359 fsck_describe_object(options, &tree->object.oid),
360 entry.path, entry.mode);
362 if (result < 0)
363 return result;
364 if (!res)
365 res = result;
367 return res;
370 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
372 int counter = 0, generation = 0, name_prefix_len = 0;
373 struct commit_list *parents;
374 int res;
375 int result;
376 const char *name;
378 if (repo_parse_commit(the_repository, commit))
379 return -1;
381 name = fsck_get_object_name(options, &commit->object.oid);
382 if (name)
383 fsck_put_object_name(options, get_commit_tree_oid(commit),
384 "%s:", name);
386 result = options->walk((struct object *) repo_get_commit_tree(the_repository, commit),
387 OBJ_TREE, data, options);
388 if (result < 0)
389 return result;
390 res = result;
392 parents = commit->parents;
393 if (name && parents) {
394 int len = strlen(name), power;
396 if (len && name[len - 1] == '^') {
397 generation = 1;
398 name_prefix_len = len - 1;
400 else { /* parse ~<generation> suffix */
401 for (generation = 0, power = 1;
402 len && isdigit(name[len - 1]);
403 power *= 10)
404 generation += power * (name[--len] - '0');
405 if (power > 1 && len && name[len - 1] == '~')
406 name_prefix_len = len - 1;
407 else {
408 /* Maybe a non-first parent, e.g. HEAD^2 */
409 generation = 0;
410 name_prefix_len = len;
415 while (parents) {
416 if (name) {
417 struct object_id *oid = &parents->item->object.oid;
419 if (counter++)
420 fsck_put_object_name(options, oid, "%s^%d",
421 name, counter);
422 else if (generation > 0)
423 fsck_put_object_name(options, oid, "%.*s~%d",
424 name_prefix_len, name,
425 generation + 1);
426 else
427 fsck_put_object_name(options, oid, "%s^", name);
429 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
430 if (result < 0)
431 return result;
432 if (!res)
433 res = result;
434 parents = parents->next;
436 return res;
439 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
441 const char *name = fsck_get_object_name(options, &tag->object.oid);
443 if (parse_tag(tag))
444 return -1;
445 if (name)
446 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
447 return options->walk(tag->tagged, OBJ_ANY, data, options);
450 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
452 if (!obj)
453 return -1;
455 if (obj->type == OBJ_NONE)
456 parse_object(the_repository, &obj->oid);
458 switch (obj->type) {
459 case OBJ_BLOB:
460 return 0;
461 case OBJ_TREE:
462 return fsck_walk_tree((struct tree *)obj, data, options);
463 case OBJ_COMMIT:
464 return fsck_walk_commit((struct commit *)obj, data, options);
465 case OBJ_TAG:
466 return fsck_walk_tag((struct tag *)obj, data, options);
467 default:
468 error("Unknown object type for %s",
469 fsck_describe_object(options, &obj->oid));
470 return -1;
474 struct name_stack {
475 const char **names;
476 size_t nr, alloc;
479 static void name_stack_push(struct name_stack *stack, const char *name)
481 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
482 stack->names[stack->nr++] = name;
485 static const char *name_stack_pop(struct name_stack *stack)
487 return stack->nr ? stack->names[--stack->nr] : NULL;
490 static void name_stack_clear(struct name_stack *stack)
492 FREE_AND_NULL(stack->names);
493 stack->nr = stack->alloc = 0;
497 * The entries in a tree are ordered in the _path_ order,
498 * which means that a directory entry is ordered by adding
499 * a slash to the end of it.
501 * So a directory called "a" is ordered _after_ a file
502 * called "a.c", because "a/" sorts after "a.c".
504 #define TREE_UNORDERED (-1)
505 #define TREE_HAS_DUPS (-2)
507 static int is_less_than_slash(unsigned char c)
509 return '\0' < c && c < '/';
512 static int verify_ordered(unsigned mode1, const char *name1,
513 unsigned mode2, const char *name2,
514 struct name_stack *candidates)
516 int len1 = strlen(name1);
517 int len2 = strlen(name2);
518 int len = len1 < len2 ? len1 : len2;
519 unsigned char c1, c2;
520 int cmp;
522 cmp = memcmp(name1, name2, len);
523 if (cmp < 0)
524 return 0;
525 if (cmp > 0)
526 return TREE_UNORDERED;
529 * Ok, the first <len> characters are the same.
530 * Now we need to order the next one, but turn
531 * a '\0' into a '/' for a directory entry.
533 c1 = name1[len];
534 c2 = name2[len];
535 if (!c1 && !c2)
537 * git-write-tree used to write out a nonsense tree that has
538 * entries with the same name, one blob and one tree. Make
539 * sure we do not have duplicate entries.
541 return TREE_HAS_DUPS;
542 if (!c1 && S_ISDIR(mode1))
543 c1 = '/';
544 if (!c2 && S_ISDIR(mode2))
545 c2 = '/';
548 * There can be non-consecutive duplicates due to the implicitly
549 * added slash, e.g.:
551 * foo
552 * foo.bar
553 * foo.bar.baz
554 * foo.bar/
555 * foo/
557 * Record non-directory candidates (like "foo" and "foo.bar" in
558 * the example) on a stack and check directory candidates (like
559 * foo/" and "foo.bar/") against that stack.
561 if (!c1 && is_less_than_slash(c2)) {
562 name_stack_push(candidates, name1);
563 } else if (c2 == '/' && is_less_than_slash(c1)) {
564 for (;;) {
565 const char *p;
566 const char *f_name = name_stack_pop(candidates);
568 if (!f_name)
569 break;
570 if (!skip_prefix(name2, f_name, &p))
571 continue;
572 if (!*p)
573 return TREE_HAS_DUPS;
574 if (is_less_than_slash(*p)) {
575 name_stack_push(candidates, f_name);
576 break;
581 return c1 < c2 ? 0 : TREE_UNORDERED;
584 static int fsck_tree(const struct object_id *tree_oid,
585 const char *buffer, unsigned long size,
586 struct fsck_options *options)
588 int retval = 0;
589 int has_null_sha1 = 0;
590 int has_full_path = 0;
591 int has_empty_name = 0;
592 int has_dot = 0;
593 int has_dotdot = 0;
594 int has_dotgit = 0;
595 int has_zero_pad = 0;
596 int has_bad_modes = 0;
597 int has_dup_entries = 0;
598 int not_properly_sorted = 0;
599 int has_large_name = 0;
600 struct tree_desc desc;
601 unsigned o_mode;
602 const char *o_name;
603 struct name_stack df_dup_candidates = { NULL };
605 if (init_tree_desc_gently(&desc, tree_oid, buffer, size,
606 TREE_DESC_RAW_MODES)) {
607 retval += report(options, tree_oid, OBJ_TREE,
608 FSCK_MSG_BAD_TREE,
609 "cannot be parsed as a tree");
610 return retval;
613 o_mode = 0;
614 o_name = NULL;
616 while (desc.size) {
617 unsigned short mode;
618 const char *name, *backslash;
619 const struct object_id *entry_oid;
621 entry_oid = tree_entry_extract(&desc, &name, &mode);
623 has_null_sha1 |= is_null_oid(entry_oid);
624 has_full_path |= !!strchr(name, '/');
625 has_empty_name |= !*name;
626 has_dot |= !strcmp(name, ".");
627 has_dotdot |= !strcmp(name, "..");
628 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
629 has_zero_pad |= *(char *)desc.buffer == '0';
630 has_large_name |= tree_entry_len(&desc.entry) > max_tree_entry_len;
632 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
633 if (!S_ISLNK(mode))
634 oidset_insert(&options->gitmodules_found,
635 entry_oid);
636 else
637 retval += report(options,
638 tree_oid, OBJ_TREE,
639 FSCK_MSG_GITMODULES_SYMLINK,
640 ".gitmodules is a symbolic link");
643 if (is_hfs_dotgitattributes(name) || is_ntfs_dotgitattributes(name)) {
644 if (!S_ISLNK(mode))
645 oidset_insert(&options->gitattributes_found,
646 entry_oid);
647 else
648 retval += report(options, tree_oid, OBJ_TREE,
649 FSCK_MSG_GITATTRIBUTES_SYMLINK,
650 ".gitattributes is a symlink");
653 if (S_ISLNK(mode)) {
654 if (is_hfs_dotgitignore(name) ||
655 is_ntfs_dotgitignore(name))
656 retval += report(options, tree_oid, OBJ_TREE,
657 FSCK_MSG_GITIGNORE_SYMLINK,
658 ".gitignore is a symlink");
659 if (is_hfs_dotmailmap(name) ||
660 is_ntfs_dotmailmap(name))
661 retval += report(options, tree_oid, OBJ_TREE,
662 FSCK_MSG_MAILMAP_SYMLINK,
663 ".mailmap is a symlink");
666 if ((backslash = strchr(name, '\\'))) {
667 while (backslash) {
668 backslash++;
669 has_dotgit |= is_ntfs_dotgit(backslash);
670 if (is_ntfs_dotgitmodules(backslash)) {
671 if (!S_ISLNK(mode))
672 oidset_insert(&options->gitmodules_found,
673 entry_oid);
674 else
675 retval += report(options, tree_oid, OBJ_TREE,
676 FSCK_MSG_GITMODULES_SYMLINK,
677 ".gitmodules is a symbolic link");
679 backslash = strchr(backslash, '\\');
683 if (update_tree_entry_gently(&desc)) {
684 retval += report(options, tree_oid, OBJ_TREE,
685 FSCK_MSG_BAD_TREE,
686 "cannot be parsed as a tree");
687 break;
690 switch (mode) {
692 * Standard modes..
694 case S_IFREG | 0755:
695 case S_IFREG | 0644:
696 case S_IFLNK:
697 case S_IFDIR:
698 case S_IFGITLINK:
699 break;
701 * This is nonstandard, but we had a few of these
702 * early on when we honored the full set of mode
703 * bits..
705 case S_IFREG | 0664:
706 if (!options->strict)
707 break;
708 /* fallthrough */
709 default:
710 has_bad_modes = 1;
713 if (o_name) {
714 switch (verify_ordered(o_mode, o_name, mode, name,
715 &df_dup_candidates)) {
716 case TREE_UNORDERED:
717 not_properly_sorted = 1;
718 break;
719 case TREE_HAS_DUPS:
720 has_dup_entries = 1;
721 break;
722 default:
723 break;
727 o_mode = mode;
728 o_name = name;
731 name_stack_clear(&df_dup_candidates);
733 if (has_null_sha1)
734 retval += report(options, tree_oid, OBJ_TREE,
735 FSCK_MSG_NULL_SHA1,
736 "contains entries pointing to null sha1");
737 if (has_full_path)
738 retval += report(options, tree_oid, OBJ_TREE,
739 FSCK_MSG_FULL_PATHNAME,
740 "contains full pathnames");
741 if (has_empty_name)
742 retval += report(options, tree_oid, OBJ_TREE,
743 FSCK_MSG_EMPTY_NAME,
744 "contains empty pathname");
745 if (has_dot)
746 retval += report(options, tree_oid, OBJ_TREE,
747 FSCK_MSG_HAS_DOT,
748 "contains '.'");
749 if (has_dotdot)
750 retval += report(options, tree_oid, OBJ_TREE,
751 FSCK_MSG_HAS_DOTDOT,
752 "contains '..'");
753 if (has_dotgit)
754 retval += report(options, tree_oid, OBJ_TREE,
755 FSCK_MSG_HAS_DOTGIT,
756 "contains '.git'");
757 if (has_zero_pad)
758 retval += report(options, tree_oid, OBJ_TREE,
759 FSCK_MSG_ZERO_PADDED_FILEMODE,
760 "contains zero-padded file modes");
761 if (has_bad_modes)
762 retval += report(options, tree_oid, OBJ_TREE,
763 FSCK_MSG_BAD_FILEMODE,
764 "contains bad file modes");
765 if (has_dup_entries)
766 retval += report(options, tree_oid, OBJ_TREE,
767 FSCK_MSG_DUPLICATE_ENTRIES,
768 "contains duplicate file entries");
769 if (not_properly_sorted)
770 retval += report(options, tree_oid, OBJ_TREE,
771 FSCK_MSG_TREE_NOT_SORTED,
772 "not properly sorted");
773 if (has_large_name)
774 retval += report(options, tree_oid, OBJ_TREE,
775 FSCK_MSG_LARGE_PATHNAME,
776 "contains excessively large pathname");
777 return retval;
781 * Confirm that the headers of a commit or tag object end in a reasonable way,
782 * either with the usual "\n\n" separator, or at least with a trailing newline
783 * on the final header line.
785 * This property is important for the memory safety of our callers. It allows
786 * them to scan the buffer linewise without constantly checking the remaining
787 * size as long as:
789 * - they check that there are bytes left in the buffer at the start of any
790 * line (i.e., that the last newline they saw was not the final one we
791 * found here)
793 * - any intra-line scanning they do will stop at a newline, which will worst
794 * case hit the newline we found here as the end-of-header. This makes it
795 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
797 static int verify_headers(const void *data, unsigned long size,
798 const struct object_id *oid, enum object_type type,
799 struct fsck_options *options)
801 const char *buffer = (const char *)data;
802 unsigned long i;
804 for (i = 0; i < size; i++) {
805 switch (buffer[i]) {
806 case '\0':
807 return report(options, oid, type,
808 FSCK_MSG_NUL_IN_HEADER,
809 "unterminated header: NUL at offset %ld", i);
810 case '\n':
811 if (i + 1 < size && buffer[i + 1] == '\n')
812 return 0;
817 * We did not find double-LF that separates the header
818 * and the body. Not having a body is not a crime but
819 * we do want to see the terminating LF for the last header
820 * line.
822 if (size && buffer[size - 1] == '\n')
823 return 0;
825 return report(options, oid, type,
826 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
829 static int fsck_ident(const char **ident,
830 const struct object_id *oid, enum object_type type,
831 struct fsck_options *options)
833 const char *p = *ident;
834 char *end;
836 *ident = strchrnul(*ident, '\n');
837 if (**ident == '\n')
838 (*ident)++;
840 if (*p == '<')
841 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
842 p += strcspn(p, "<>\n");
843 if (*p == '>')
844 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
845 if (*p != '<')
846 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
847 if (p[-1] != ' ')
848 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
849 p++;
850 p += strcspn(p, "<>\n");
851 if (*p != '>')
852 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
853 p++;
854 if (*p != ' ')
855 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
856 p++;
858 * Our timestamp parser is based on the C strto*() functions, which
859 * will happily eat whitespace, including the newline that is supposed
860 * to prevent us walking past the end of the buffer. So do our own
861 * scan, skipping linear whitespace but not newlines, and then
862 * confirming we found a digit. We _could_ be even more strict here,
863 * as we really expect only a single space, but since we have
864 * traditionally allowed extra whitespace, we'll continue to do so.
866 while (*p == ' ' || *p == '\t')
867 p++;
868 if (!isdigit(*p))
869 return report(options, oid, type, FSCK_MSG_BAD_DATE,
870 "invalid author/committer line - bad date");
871 if (*p == '0' && p[1] != ' ')
872 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
873 if (date_overflows(parse_timestamp(p, &end, 10)))
874 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
875 if ((end == p || *end != ' '))
876 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
877 p = end + 1;
878 if ((*p != '+' && *p != '-') ||
879 !isdigit(p[1]) ||
880 !isdigit(p[2]) ||
881 !isdigit(p[3]) ||
882 !isdigit(p[4]) ||
883 (p[5] != '\n'))
884 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
885 p += 6;
886 return 0;
889 static int fsck_commit(const struct object_id *oid,
890 const char *buffer, unsigned long size,
891 struct fsck_options *options)
893 struct object_id tree_oid, parent_oid;
894 unsigned author_count;
895 int err;
896 const char *buffer_begin = buffer;
897 const char *buffer_end = buffer + size;
898 const char *p;
901 * We _must_ stop parsing immediately if this reports failure, as the
902 * memory safety of the rest of the function depends on it. See the
903 * comment above the definition of verify_headers() for more details.
905 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
906 return -1;
908 if (buffer >= buffer_end || !skip_prefix(buffer, "tree ", &buffer))
909 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
910 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
911 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
912 if (err)
913 return err;
915 buffer = p + 1;
916 while (buffer < buffer_end && skip_prefix(buffer, "parent ", &buffer)) {
917 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
918 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
919 if (err)
920 return err;
922 buffer = p + 1;
924 author_count = 0;
925 while (buffer < buffer_end && skip_prefix(buffer, "author ", &buffer)) {
926 author_count++;
927 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
928 if (err)
929 return err;
931 if (author_count < 1)
932 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
933 else if (author_count > 1)
934 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
935 if (err)
936 return err;
937 if (buffer >= buffer_end || !skip_prefix(buffer, "committer ", &buffer))
938 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
939 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
940 if (err)
941 return err;
942 if (memchr(buffer_begin, '\0', size)) {
943 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
944 "NUL byte in the commit object body");
945 if (err)
946 return err;
948 return 0;
951 static int fsck_tag(const struct object_id *oid, const char *buffer,
952 unsigned long size, struct fsck_options *options)
954 struct object_id tagged_oid;
955 int tagged_type;
956 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
957 &tagged_type);
960 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
961 unsigned long size, struct fsck_options *options,
962 struct object_id *tagged_oid,
963 int *tagged_type)
965 int ret = 0;
966 char *eol;
967 struct strbuf sb = STRBUF_INIT;
968 const char *buffer_end = buffer + size;
969 const char *p;
972 * We _must_ stop parsing immediately if this reports failure, as the
973 * memory safety of the rest of the function depends on it. See the
974 * comment above the definition of verify_headers() for more details.
976 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
977 if (ret)
978 goto done;
980 if (buffer >= buffer_end || !skip_prefix(buffer, "object ", &buffer)) {
981 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
982 goto done;
984 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
985 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
986 if (ret)
987 goto done;
989 buffer = p + 1;
991 if (buffer >= buffer_end || !skip_prefix(buffer, "type ", &buffer)) {
992 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
993 goto done;
995 eol = memchr(buffer, '\n', buffer_end - buffer);
996 if (!eol) {
997 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
998 goto done;
1000 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
1001 if (*tagged_type < 0)
1002 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
1003 if (ret)
1004 goto done;
1005 buffer = eol + 1;
1007 if (buffer >= buffer_end || !skip_prefix(buffer, "tag ", &buffer)) {
1008 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
1009 goto done;
1011 eol = memchr(buffer, '\n', buffer_end - buffer);
1012 if (!eol) {
1013 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
1014 goto done;
1016 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
1017 if (check_refname_format(sb.buf, 0)) {
1018 ret = report(options, oid, OBJ_TAG,
1019 FSCK_MSG_BAD_TAG_NAME,
1020 "invalid 'tag' name: %.*s",
1021 (int)(eol - buffer), buffer);
1022 if (ret)
1023 goto done;
1025 buffer = eol + 1;
1027 if (buffer >= buffer_end || !skip_prefix(buffer, "tagger ", &buffer)) {
1028 /* early tags do not contain 'tagger' lines; warn only */
1029 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
1030 if (ret)
1031 goto done;
1033 else
1034 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
1036 if (buffer < buffer_end && !starts_with(buffer, "\n")) {
1038 * The verify_headers() check will allow
1039 * e.g. "[...]tagger <tagger>\nsome
1040 * garbage\n\nmessage" to pass, thinking "some
1041 * garbage" could be a custom header. E.g. "mktag"
1042 * doesn't want any unknown headers.
1044 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1045 if (ret)
1046 goto done;
1049 done:
1050 strbuf_release(&sb);
1051 return ret;
1054 struct fsck_gitmodules_data {
1055 const struct object_id *oid;
1056 struct fsck_options *options;
1057 int ret;
1060 static int fsck_gitmodules_fn(const char *var, const char *value,
1061 const struct config_context *ctx UNUSED,
1062 void *vdata)
1064 struct fsck_gitmodules_data *data = vdata;
1065 const char *subsection, *key;
1066 size_t subsection_len;
1067 char *name;
1069 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1070 !subsection)
1071 return 0;
1073 name = xmemdupz(subsection, subsection_len);
1074 if (check_submodule_name(name) < 0)
1075 data->ret |= report(data->options,
1076 data->oid, OBJ_BLOB,
1077 FSCK_MSG_GITMODULES_NAME,
1078 "disallowed submodule name: %s",
1079 name);
1080 if (!strcmp(key, "url") && value &&
1081 check_submodule_url(value) < 0)
1082 data->ret |= report(data->options,
1083 data->oid, OBJ_BLOB,
1084 FSCK_MSG_GITMODULES_URL,
1085 "disallowed submodule url: %s",
1086 value);
1087 if (!strcmp(key, "path") && value &&
1088 looks_like_command_line_option(value))
1089 data->ret |= report(data->options,
1090 data->oid, OBJ_BLOB,
1091 FSCK_MSG_GITMODULES_PATH,
1092 "disallowed submodule path: %s",
1093 value);
1094 if (!strcmp(key, "update") && value &&
1095 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1096 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1097 FSCK_MSG_GITMODULES_UPDATE,
1098 "disallowed submodule update setting: %s",
1099 value);
1100 free(name);
1102 return 0;
1105 static int fsck_blob(const struct object_id *oid, const char *buf,
1106 unsigned long size, struct fsck_options *options)
1108 int ret = 0;
1110 if (object_on_skiplist(options, oid))
1111 return 0;
1113 if (oidset_contains(&options->gitmodules_found, oid)) {
1114 struct config_options config_opts = { 0 };
1115 struct fsck_gitmodules_data data;
1117 oidset_insert(&options->gitmodules_done, oid);
1119 if (!buf) {
1121 * A missing buffer here is a sign that the caller found the
1122 * blob too gigantic to load into memory. Let's just consider
1123 * that an error.
1125 return report(options, oid, OBJ_BLOB,
1126 FSCK_MSG_GITMODULES_LARGE,
1127 ".gitmodules too large to parse");
1130 data.oid = oid;
1131 data.options = options;
1132 data.ret = 0;
1133 config_opts.error_action = CONFIG_ERROR_SILENT;
1134 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1135 ".gitmodules", buf, size, &data,
1136 CONFIG_SCOPE_UNKNOWN, &config_opts))
1137 data.ret |= report(options, oid, OBJ_BLOB,
1138 FSCK_MSG_GITMODULES_PARSE,
1139 "could not parse gitmodules blob");
1140 ret |= data.ret;
1143 if (oidset_contains(&options->gitattributes_found, oid)) {
1144 const char *ptr;
1146 oidset_insert(&options->gitattributes_done, oid);
1148 if (!buf || size > ATTR_MAX_FILE_SIZE) {
1150 * A missing buffer here is a sign that the caller found the
1151 * blob too gigantic to load into memory. Let's just consider
1152 * that an error.
1154 return report(options, oid, OBJ_BLOB,
1155 FSCK_MSG_GITATTRIBUTES_LARGE,
1156 ".gitattributes too large to parse");
1159 for (ptr = buf; *ptr; ) {
1160 const char *eol = strchrnul(ptr, '\n');
1161 if (eol - ptr >= ATTR_MAX_LINE_LENGTH) {
1162 ret |= report(options, oid, OBJ_BLOB,
1163 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH,
1164 ".gitattributes has too long lines to parse");
1165 break;
1168 ptr = *eol ? eol + 1 : eol;
1172 return ret;
1175 int fsck_object(struct object *obj, void *data, unsigned long size,
1176 struct fsck_options *options)
1178 if (!obj)
1179 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1181 return fsck_buffer(&obj->oid, obj->type, data, size, options);
1184 int fsck_buffer(const struct object_id *oid, enum object_type type,
1185 const void *data, unsigned long size,
1186 struct fsck_options *options)
1188 if (type == OBJ_BLOB)
1189 return fsck_blob(oid, data, size, options);
1190 if (type == OBJ_TREE)
1191 return fsck_tree(oid, data, size, options);
1192 if (type == OBJ_COMMIT)
1193 return fsck_commit(oid, data, size, options);
1194 if (type == OBJ_TAG)
1195 return fsck_tag(oid, data, size, options);
1197 return report(options, oid, type,
1198 FSCK_MSG_UNKNOWN_TYPE,
1199 "unknown type '%d' (internal fsck error)",
1200 type);
1203 int fsck_error_function(struct fsck_options *o,
1204 const struct object_id *oid,
1205 enum object_type object_type UNUSED,
1206 enum fsck_msg_type msg_type,
1207 enum fsck_msg_id msg_id UNUSED,
1208 const char *message)
1210 if (msg_type == FSCK_WARN) {
1211 warning("object %s: %s", fsck_describe_object(o, oid), message);
1212 return 0;
1214 error("object %s: %s", fsck_describe_object(o, oid), message);
1215 return 1;
1218 static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
1219 enum fsck_msg_id msg_missing, enum fsck_msg_id msg_type,
1220 struct fsck_options *options, const char *blob_type)
1222 int ret = 0;
1223 struct oidset_iter iter;
1224 const struct object_id *oid;
1226 oidset_iter_init(blobs_found, &iter);
1227 while ((oid = oidset_iter_next(&iter))) {
1228 enum object_type type;
1229 unsigned long size;
1230 char *buf;
1232 if (oidset_contains(blobs_done, oid))
1233 continue;
1235 buf = repo_read_object_file(the_repository, oid, &type, &size);
1236 if (!buf) {
1237 if (is_promisor_object(oid))
1238 continue;
1239 ret |= report(options,
1240 oid, OBJ_BLOB, msg_missing,
1241 "unable to read %s blob", blob_type);
1242 continue;
1245 if (type == OBJ_BLOB)
1246 ret |= fsck_blob(oid, buf, size, options);
1247 else
1248 ret |= report(options, oid, type, msg_type,
1249 "non-blob found at %s", blob_type);
1250 free(buf);
1253 oidset_clear(blobs_found);
1254 oidset_clear(blobs_done);
1256 return ret;
1259 int fsck_finish(struct fsck_options *options)
1261 int ret = 0;
1263 ret |= fsck_blobs(&options->gitmodules_found, &options->gitmodules_done,
1264 FSCK_MSG_GITMODULES_MISSING, FSCK_MSG_GITMODULES_BLOB,
1265 options, ".gitmodules");
1266 ret |= fsck_blobs(&options->gitattributes_found, &options->gitattributes_done,
1267 FSCK_MSG_GITATTRIBUTES_MISSING, FSCK_MSG_GITATTRIBUTES_BLOB,
1268 options, ".gitattributes");
1270 return ret;
1273 int git_fsck_config(const char *var, const char *value,
1274 const struct config_context *ctx, void *cb)
1276 struct fsck_options *options = cb;
1277 const char *msg_id;
1279 if (strcmp(var, "fsck.skiplist") == 0) {
1280 char *path;
1281 struct strbuf sb = STRBUF_INIT;
1283 if (git_config_pathname(&path, var, value))
1284 return 1;
1285 strbuf_addf(&sb, "skiplist=%s", path);
1286 free(path);
1287 fsck_set_msg_types(options, sb.buf);
1288 strbuf_release(&sb);
1289 return 0;
1292 if (skip_prefix(var, "fsck.", &msg_id)) {
1293 if (!value)
1294 return config_error_nonbool(var);
1295 fsck_set_msg_type(options, msg_id, value);
1296 return 0;
1299 return git_default_config(var, value, ctx, cb);
1303 * Custom error callbacks that are used in more than one place.
1306 int fsck_error_cb_print_missing_gitmodules(struct fsck_options *o,
1307 const struct object_id *oid,
1308 enum object_type object_type,
1309 enum fsck_msg_type msg_type,
1310 enum fsck_msg_id msg_id,
1311 const char *message)
1313 if (msg_id == FSCK_MSG_GITMODULES_MISSING) {
1314 puts(oid_to_hex(oid));
1315 return 0;
1317 return fsck_error_function(o, oid, object_type, msg_type, msg_id, message);