test-ctype: check EOF
[git/gitster.git] / fsck.c
blob2b18717ee805bc73aa50943731ac9a7212303d18
1 #include "cache.h"
2 #include "object-store.h"
3 #include "repository.h"
4 #include "object.h"
5 #include "attr.h"
6 #include "blob.h"
7 #include "tree.h"
8 #include "tree-walk.h"
9 #include "commit.h"
10 #include "tag.h"
11 #include "fsck.h"
12 #include "refs.h"
13 #include "url.h"
14 #include "utf8.h"
15 #include "decorate.h"
16 #include "oidset.h"
17 #include "packfile.h"
18 #include "submodule-config.h"
19 #include "config.h"
20 #include "credential.h"
21 #include "help.h"
23 #define STR(x) #x
24 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
25 static struct {
26 const char *id_string;
27 const char *downcased;
28 const char *camelcased;
29 enum fsck_msg_type msg_type;
30 } msg_id_info[FSCK_MSG_MAX + 1] = {
31 FOREACH_FSCK_MSG_ID(MSG_ID)
32 { NULL, NULL, NULL, -1 }
34 #undef MSG_ID
35 #undef STR
37 static void prepare_msg_ids(void)
39 int i;
41 if (msg_id_info[0].downcased)
42 return;
44 /* convert id_string to lower case, without underscores. */
45 for (i = 0; i < FSCK_MSG_MAX; i++) {
46 const char *p = msg_id_info[i].id_string;
47 int len = strlen(p);
48 char *q = xmalloc(len);
50 msg_id_info[i].downcased = q;
51 while (*p)
52 if (*p == '_')
53 p++;
54 else
55 *(q)++ = tolower(*(p)++);
56 *q = '\0';
58 p = msg_id_info[i].id_string;
59 q = xmalloc(len);
60 msg_id_info[i].camelcased = q;
61 while (*p) {
62 if (*p == '_') {
63 p++;
64 if (*p)
65 *q++ = *p++;
66 } else {
67 *q++ = tolower(*p++);
70 *q = '\0';
74 static int parse_msg_id(const char *text)
76 int i;
78 prepare_msg_ids();
80 for (i = 0; i < FSCK_MSG_MAX; i++)
81 if (!strcmp(text, msg_id_info[i].downcased))
82 return i;
84 return -1;
87 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
89 int i;
91 prepare_msg_ids();
93 for (i = 0; i < FSCK_MSG_MAX; i++)
94 list_config_item(list, prefix, msg_id_info[i].camelcased);
97 static enum fsck_msg_type fsck_msg_type(enum fsck_msg_id msg_id,
98 struct fsck_options *options)
100 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
102 if (!options->msg_type) {
103 enum fsck_msg_type msg_type = msg_id_info[msg_id].msg_type;
105 if (options->strict && msg_type == FSCK_WARN)
106 msg_type = FSCK_ERROR;
107 return msg_type;
110 return options->msg_type[msg_id];
113 static enum fsck_msg_type parse_msg_type(const char *str)
115 if (!strcmp(str, "error"))
116 return FSCK_ERROR;
117 else if (!strcmp(str, "warn"))
118 return FSCK_WARN;
119 else if (!strcmp(str, "ignore"))
120 return FSCK_IGNORE;
121 else
122 die("Unknown fsck message type: '%s'", str);
125 int is_valid_msg_type(const char *msg_id, const char *msg_type)
127 if (parse_msg_id(msg_id) < 0)
128 return 0;
129 parse_msg_type(msg_type);
130 return 1;
133 void fsck_set_msg_type_from_ids(struct fsck_options *options,
134 enum fsck_msg_id msg_id,
135 enum fsck_msg_type msg_type)
137 if (!options->msg_type) {
138 int i;
139 enum fsck_msg_type *severity;
140 ALLOC_ARRAY(severity, FSCK_MSG_MAX);
141 for (i = 0; i < FSCK_MSG_MAX; i++)
142 severity[i] = fsck_msg_type(i, options);
143 options->msg_type = severity;
146 options->msg_type[msg_id] = msg_type;
149 void fsck_set_msg_type(struct fsck_options *options,
150 const char *msg_id_str, const char *msg_type_str)
152 int msg_id = parse_msg_id(msg_id_str);
153 enum fsck_msg_type msg_type = parse_msg_type(msg_type_str);
155 if (msg_id < 0)
156 die("Unhandled message id: %s", msg_id_str);
158 if (msg_type != FSCK_ERROR && msg_id_info[msg_id].msg_type == FSCK_FATAL)
159 die("Cannot demote %s to %s", msg_id_str, msg_type_str);
161 fsck_set_msg_type_from_ids(options, msg_id, msg_type);
164 void fsck_set_msg_types(struct fsck_options *options, const char *values)
166 char *buf = xstrdup(values), *to_free = buf;
167 int done = 0;
169 while (!done) {
170 int len = strcspn(buf, " ,|"), equal;
172 done = !buf[len];
173 if (!len) {
174 buf++;
175 continue;
177 buf[len] = '\0';
179 for (equal = 0;
180 equal < len && buf[equal] != '=' && buf[equal] != ':';
181 equal++)
182 buf[equal] = tolower(buf[equal]);
183 buf[equal] = '\0';
185 if (!strcmp(buf, "skiplist")) {
186 if (equal == len)
187 die("skiplist requires a path");
188 oidset_parse_file(&options->skiplist, buf + equal + 1);
189 buf += len + 1;
190 continue;
193 if (equal == len)
194 die("Missing '=': '%s'", buf);
196 fsck_set_msg_type(options, buf, buf + equal + 1);
197 buf += len + 1;
199 free(to_free);
202 static int object_on_skiplist(struct fsck_options *opts,
203 const struct object_id *oid)
205 return opts && oid && oidset_contains(&opts->skiplist, oid);
208 __attribute__((format (printf, 5, 6)))
209 static int report(struct fsck_options *options,
210 const struct object_id *oid, enum object_type object_type,
211 enum fsck_msg_id msg_id, const char *fmt, ...)
213 va_list ap;
214 struct strbuf sb = STRBUF_INIT;
215 enum fsck_msg_type msg_type = fsck_msg_type(msg_id, options);
216 int result;
218 if (msg_type == FSCK_IGNORE)
219 return 0;
221 if (object_on_skiplist(options, oid))
222 return 0;
224 if (msg_type == FSCK_FATAL)
225 msg_type = FSCK_ERROR;
226 else if (msg_type == FSCK_INFO)
227 msg_type = FSCK_WARN;
229 prepare_msg_ids();
230 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
232 va_start(ap, fmt);
233 strbuf_vaddf(&sb, fmt, ap);
234 result = options->error_func(options, oid, object_type,
235 msg_type, msg_id, sb.buf);
236 strbuf_release(&sb);
237 va_end(ap);
239 return result;
242 void fsck_enable_object_names(struct fsck_options *options)
244 if (!options->object_names)
245 options->object_names = kh_init_oid_map();
248 const char *fsck_get_object_name(struct fsck_options *options,
249 const struct object_id *oid)
251 khiter_t pos;
252 if (!options->object_names)
253 return NULL;
254 pos = kh_get_oid_map(options->object_names, *oid);
255 if (pos >= kh_end(options->object_names))
256 return NULL;
257 return kh_value(options->object_names, pos);
260 void fsck_put_object_name(struct fsck_options *options,
261 const struct object_id *oid,
262 const char *fmt, ...)
264 va_list ap;
265 struct strbuf buf = STRBUF_INIT;
266 khiter_t pos;
267 int hashret;
269 if (!options->object_names)
270 return;
272 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
273 if (!hashret)
274 return;
275 va_start(ap, fmt);
276 strbuf_vaddf(&buf, fmt, ap);
277 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
278 va_end(ap);
281 const char *fsck_describe_object(struct fsck_options *options,
282 const struct object_id *oid)
284 static struct strbuf bufs[] = {
285 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
287 static int b = 0;
288 struct strbuf *buf;
289 const char *name = fsck_get_object_name(options, oid);
291 buf = bufs + b;
292 b = (b + 1) % ARRAY_SIZE(bufs);
293 strbuf_reset(buf);
294 strbuf_addstr(buf, oid_to_hex(oid));
295 if (name)
296 strbuf_addf(buf, " (%s)", name);
298 return buf->buf;
301 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
303 struct tree_desc desc;
304 struct name_entry entry;
305 int res = 0;
306 const char *name;
308 if (parse_tree(tree))
309 return -1;
311 name = fsck_get_object_name(options, &tree->object.oid);
312 if (init_tree_desc_gently(&desc, tree->buffer, tree->size, 0))
313 return -1;
314 while (tree_entry_gently(&desc, &entry)) {
315 struct object *obj;
316 int result;
318 if (S_ISGITLINK(entry.mode))
319 continue;
321 if (S_ISDIR(entry.mode)) {
322 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
323 if (name && obj)
324 fsck_put_object_name(options, &entry.oid, "%s%s/",
325 name, entry.path);
326 result = options->walk(obj, OBJ_TREE, data, options);
328 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
329 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
330 if (name && obj)
331 fsck_put_object_name(options, &entry.oid, "%s%s",
332 name, entry.path);
333 result = options->walk(obj, OBJ_BLOB, data, options);
335 else {
336 result = error("in tree %s: entry %s has bad mode %.6o",
337 fsck_describe_object(options, &tree->object.oid),
338 entry.path, entry.mode);
340 if (result < 0)
341 return result;
342 if (!res)
343 res = result;
345 return res;
348 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
350 int counter = 0, generation = 0, name_prefix_len = 0;
351 struct commit_list *parents;
352 int res;
353 int result;
354 const char *name;
356 if (parse_commit(commit))
357 return -1;
359 name = fsck_get_object_name(options, &commit->object.oid);
360 if (name)
361 fsck_put_object_name(options, get_commit_tree_oid(commit),
362 "%s:", name);
364 result = options->walk((struct object *)get_commit_tree(commit),
365 OBJ_TREE, data, options);
366 if (result < 0)
367 return result;
368 res = result;
370 parents = commit->parents;
371 if (name && parents) {
372 int len = strlen(name), power;
374 if (len && name[len - 1] == '^') {
375 generation = 1;
376 name_prefix_len = len - 1;
378 else { /* parse ~<generation> suffix */
379 for (generation = 0, power = 1;
380 len && isdigit(name[len - 1]);
381 power *= 10)
382 generation += power * (name[--len] - '0');
383 if (power > 1 && len && name[len - 1] == '~')
384 name_prefix_len = len - 1;
385 else {
386 /* Maybe a non-first parent, e.g. HEAD^2 */
387 generation = 0;
388 name_prefix_len = len;
393 while (parents) {
394 if (name) {
395 struct object_id *oid = &parents->item->object.oid;
397 if (counter++)
398 fsck_put_object_name(options, oid, "%s^%d",
399 name, counter);
400 else if (generation > 0)
401 fsck_put_object_name(options, oid, "%.*s~%d",
402 name_prefix_len, name,
403 generation + 1);
404 else
405 fsck_put_object_name(options, oid, "%s^", name);
407 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
408 if (result < 0)
409 return result;
410 if (!res)
411 res = result;
412 parents = parents->next;
414 return res;
417 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
419 const char *name = fsck_get_object_name(options, &tag->object.oid);
421 if (parse_tag(tag))
422 return -1;
423 if (name)
424 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
425 return options->walk(tag->tagged, OBJ_ANY, data, options);
428 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
430 if (!obj)
431 return -1;
433 if (obj->type == OBJ_NONE)
434 parse_object(the_repository, &obj->oid);
436 switch (obj->type) {
437 case OBJ_BLOB:
438 return 0;
439 case OBJ_TREE:
440 return fsck_walk_tree((struct tree *)obj, data, options);
441 case OBJ_COMMIT:
442 return fsck_walk_commit((struct commit *)obj, data, options);
443 case OBJ_TAG:
444 return fsck_walk_tag((struct tag *)obj, data, options);
445 default:
446 error("Unknown object type for %s",
447 fsck_describe_object(options, &obj->oid));
448 return -1;
452 struct name_stack {
453 const char **names;
454 size_t nr, alloc;
457 static void name_stack_push(struct name_stack *stack, const char *name)
459 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
460 stack->names[stack->nr++] = name;
463 static const char *name_stack_pop(struct name_stack *stack)
465 return stack->nr ? stack->names[--stack->nr] : NULL;
468 static void name_stack_clear(struct name_stack *stack)
470 FREE_AND_NULL(stack->names);
471 stack->nr = stack->alloc = 0;
475 * The entries in a tree are ordered in the _path_ order,
476 * which means that a directory entry is ordered by adding
477 * a slash to the end of it.
479 * So a directory called "a" is ordered _after_ a file
480 * called "a.c", because "a/" sorts after "a.c".
482 #define TREE_UNORDERED (-1)
483 #define TREE_HAS_DUPS (-2)
485 static int is_less_than_slash(unsigned char c)
487 return '\0' < c && c < '/';
490 static int verify_ordered(unsigned mode1, const char *name1,
491 unsigned mode2, const char *name2,
492 struct name_stack *candidates)
494 int len1 = strlen(name1);
495 int len2 = strlen(name2);
496 int len = len1 < len2 ? len1 : len2;
497 unsigned char c1, c2;
498 int cmp;
500 cmp = memcmp(name1, name2, len);
501 if (cmp < 0)
502 return 0;
503 if (cmp > 0)
504 return TREE_UNORDERED;
507 * Ok, the first <len> characters are the same.
508 * Now we need to order the next one, but turn
509 * a '\0' into a '/' for a directory entry.
511 c1 = name1[len];
512 c2 = name2[len];
513 if (!c1 && !c2)
515 * git-write-tree used to write out a nonsense tree that has
516 * entries with the same name, one blob and one tree. Make
517 * sure we do not have duplicate entries.
519 return TREE_HAS_DUPS;
520 if (!c1 && S_ISDIR(mode1))
521 c1 = '/';
522 if (!c2 && S_ISDIR(mode2))
523 c2 = '/';
526 * There can be non-consecutive duplicates due to the implicitly
527 * added slash, e.g.:
529 * foo
530 * foo.bar
531 * foo.bar.baz
532 * foo.bar/
533 * foo/
535 * Record non-directory candidates (like "foo" and "foo.bar" in
536 * the example) on a stack and check directory candidates (like
537 * foo/" and "foo.bar/") against that stack.
539 if (!c1 && is_less_than_slash(c2)) {
540 name_stack_push(candidates, name1);
541 } else if (c2 == '/' && is_less_than_slash(c1)) {
542 for (;;) {
543 const char *p;
544 const char *f_name = name_stack_pop(candidates);
546 if (!f_name)
547 break;
548 if (!skip_prefix(name2, f_name, &p))
549 continue;
550 if (!*p)
551 return TREE_HAS_DUPS;
552 if (is_less_than_slash(*p)) {
553 name_stack_push(candidates, f_name);
554 break;
559 return c1 < c2 ? 0 : TREE_UNORDERED;
562 static int fsck_tree(const struct object_id *tree_oid,
563 const char *buffer, unsigned long size,
564 struct fsck_options *options)
566 int retval = 0;
567 int has_null_sha1 = 0;
568 int has_full_path = 0;
569 int has_empty_name = 0;
570 int has_dot = 0;
571 int has_dotdot = 0;
572 int has_dotgit = 0;
573 int has_zero_pad = 0;
574 int has_bad_modes = 0;
575 int has_dup_entries = 0;
576 int not_properly_sorted = 0;
577 struct tree_desc desc;
578 unsigned o_mode;
579 const char *o_name;
580 struct name_stack df_dup_candidates = { NULL };
582 if (init_tree_desc_gently(&desc, buffer, size, TREE_DESC_RAW_MODES)) {
583 retval += report(options, tree_oid, OBJ_TREE,
584 FSCK_MSG_BAD_TREE,
585 "cannot be parsed as a tree");
586 return retval;
589 o_mode = 0;
590 o_name = NULL;
592 while (desc.size) {
593 unsigned short mode;
594 const char *name, *backslash;
595 const struct object_id *entry_oid;
597 entry_oid = tree_entry_extract(&desc, &name, &mode);
599 has_null_sha1 |= is_null_oid(entry_oid);
600 has_full_path |= !!strchr(name, '/');
601 has_empty_name |= !*name;
602 has_dot |= !strcmp(name, ".");
603 has_dotdot |= !strcmp(name, "..");
604 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
605 has_zero_pad |= *(char *)desc.buffer == '0';
607 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
608 if (!S_ISLNK(mode))
609 oidset_insert(&options->gitmodules_found,
610 entry_oid);
611 else
612 retval += report(options,
613 tree_oid, OBJ_TREE,
614 FSCK_MSG_GITMODULES_SYMLINK,
615 ".gitmodules is a symbolic link");
618 if (is_hfs_dotgitattributes(name) || is_ntfs_dotgitattributes(name)) {
619 if (!S_ISLNK(mode))
620 oidset_insert(&options->gitattributes_found,
621 entry_oid);
622 else
623 retval += report(options, tree_oid, OBJ_TREE,
624 FSCK_MSG_GITATTRIBUTES_SYMLINK,
625 ".gitattributes is a symlink");
628 if (S_ISLNK(mode)) {
629 if (is_hfs_dotgitignore(name) ||
630 is_ntfs_dotgitignore(name))
631 retval += report(options, tree_oid, OBJ_TREE,
632 FSCK_MSG_GITIGNORE_SYMLINK,
633 ".gitignore is a symlink");
634 if (is_hfs_dotmailmap(name) ||
635 is_ntfs_dotmailmap(name))
636 retval += report(options, tree_oid, OBJ_TREE,
637 FSCK_MSG_MAILMAP_SYMLINK,
638 ".mailmap is a symlink");
641 if ((backslash = strchr(name, '\\'))) {
642 while (backslash) {
643 backslash++;
644 has_dotgit |= is_ntfs_dotgit(backslash);
645 if (is_ntfs_dotgitmodules(backslash)) {
646 if (!S_ISLNK(mode))
647 oidset_insert(&options->gitmodules_found,
648 entry_oid);
649 else
650 retval += report(options, tree_oid, OBJ_TREE,
651 FSCK_MSG_GITMODULES_SYMLINK,
652 ".gitmodules is a symbolic link");
654 backslash = strchr(backslash, '\\');
658 if (update_tree_entry_gently(&desc)) {
659 retval += report(options, tree_oid, OBJ_TREE,
660 FSCK_MSG_BAD_TREE,
661 "cannot be parsed as a tree");
662 break;
665 switch (mode) {
667 * Standard modes..
669 case S_IFREG | 0755:
670 case S_IFREG | 0644:
671 case S_IFLNK:
672 case S_IFDIR:
673 case S_IFGITLINK:
674 break;
676 * This is nonstandard, but we had a few of these
677 * early on when we honored the full set of mode
678 * bits..
680 case S_IFREG | 0664:
681 if (!options->strict)
682 break;
683 /* fallthrough */
684 default:
685 has_bad_modes = 1;
688 if (o_name) {
689 switch (verify_ordered(o_mode, o_name, mode, name,
690 &df_dup_candidates)) {
691 case TREE_UNORDERED:
692 not_properly_sorted = 1;
693 break;
694 case TREE_HAS_DUPS:
695 has_dup_entries = 1;
696 break;
697 default:
698 break;
702 o_mode = mode;
703 o_name = name;
706 name_stack_clear(&df_dup_candidates);
708 if (has_null_sha1)
709 retval += report(options, tree_oid, OBJ_TREE,
710 FSCK_MSG_NULL_SHA1,
711 "contains entries pointing to null sha1");
712 if (has_full_path)
713 retval += report(options, tree_oid, OBJ_TREE,
714 FSCK_MSG_FULL_PATHNAME,
715 "contains full pathnames");
716 if (has_empty_name)
717 retval += report(options, tree_oid, OBJ_TREE,
718 FSCK_MSG_EMPTY_NAME,
719 "contains empty pathname");
720 if (has_dot)
721 retval += report(options, tree_oid, OBJ_TREE,
722 FSCK_MSG_HAS_DOT,
723 "contains '.'");
724 if (has_dotdot)
725 retval += report(options, tree_oid, OBJ_TREE,
726 FSCK_MSG_HAS_DOTDOT,
727 "contains '..'");
728 if (has_dotgit)
729 retval += report(options, tree_oid, OBJ_TREE,
730 FSCK_MSG_HAS_DOTGIT,
731 "contains '.git'");
732 if (has_zero_pad)
733 retval += report(options, tree_oid, OBJ_TREE,
734 FSCK_MSG_ZERO_PADDED_FILEMODE,
735 "contains zero-padded file modes");
736 if (has_bad_modes)
737 retval += report(options, tree_oid, OBJ_TREE,
738 FSCK_MSG_BAD_FILEMODE,
739 "contains bad file modes");
740 if (has_dup_entries)
741 retval += report(options, tree_oid, OBJ_TREE,
742 FSCK_MSG_DUPLICATE_ENTRIES,
743 "contains duplicate file entries");
744 if (not_properly_sorted)
745 retval += report(options, tree_oid, OBJ_TREE,
746 FSCK_MSG_TREE_NOT_SORTED,
747 "not properly sorted");
748 return retval;
752 * Confirm that the headers of a commit or tag object end in a reasonable way,
753 * either with the usual "\n\n" separator, or at least with a trailing newline
754 * on the final header line.
756 * This property is important for the memory safety of our callers. It allows
757 * them to scan the buffer linewise without constantly checking the remaining
758 * size as long as:
760 * - they check that there are bytes left in the buffer at the start of any
761 * line (i.e., that the last newline they saw was not the final one we
762 * found here)
764 * - any intra-line scanning they do will stop at a newline, which will worst
765 * case hit the newline we found here as the end-of-header. This makes it
766 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
768 static int verify_headers(const void *data, unsigned long size,
769 const struct object_id *oid, enum object_type type,
770 struct fsck_options *options)
772 const char *buffer = (const char *)data;
773 unsigned long i;
775 for (i = 0; i < size; i++) {
776 switch (buffer[i]) {
777 case '\0':
778 return report(options, oid, type,
779 FSCK_MSG_NUL_IN_HEADER,
780 "unterminated header: NUL at offset %ld", i);
781 case '\n':
782 if (i + 1 < size && buffer[i + 1] == '\n')
783 return 0;
788 * We did not find double-LF that separates the header
789 * and the body. Not having a body is not a crime but
790 * we do want to see the terminating LF for the last header
791 * line.
793 if (size && buffer[size - 1] == '\n')
794 return 0;
796 return report(options, oid, type,
797 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
800 static int fsck_ident(const char **ident,
801 const struct object_id *oid, enum object_type type,
802 struct fsck_options *options)
804 const char *p = *ident;
805 char *end;
807 *ident = strchrnul(*ident, '\n');
808 if (**ident == '\n')
809 (*ident)++;
811 if (*p == '<')
812 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
813 p += strcspn(p, "<>\n");
814 if (*p == '>')
815 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
816 if (*p != '<')
817 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
818 if (p[-1] != ' ')
819 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
820 p++;
821 p += strcspn(p, "<>\n");
822 if (*p != '>')
823 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
824 p++;
825 if (*p != ' ')
826 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
827 p++;
829 * Our timestamp parser is based on the C strto*() functions, which
830 * will happily eat whitespace, including the newline that is supposed
831 * to prevent us walking past the end of the buffer. So do our own
832 * scan, skipping linear whitespace but not newlines, and then
833 * confirming we found a digit. We _could_ be even more strict here,
834 * as we really expect only a single space, but since we have
835 * traditionally allowed extra whitespace, we'll continue to do so.
837 while (*p == ' ' || *p == '\t')
838 p++;
839 if (!isdigit(*p))
840 return report(options, oid, type, FSCK_MSG_BAD_DATE,
841 "invalid author/committer line - bad date");
842 if (*p == '0' && p[1] != ' ')
843 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
844 if (date_overflows(parse_timestamp(p, &end, 10)))
845 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
846 if ((end == p || *end != ' '))
847 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
848 p = end + 1;
849 if ((*p != '+' && *p != '-') ||
850 !isdigit(p[1]) ||
851 !isdigit(p[2]) ||
852 !isdigit(p[3]) ||
853 !isdigit(p[4]) ||
854 (p[5] != '\n'))
855 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
856 p += 6;
857 return 0;
860 static int fsck_commit(const struct object_id *oid,
861 const char *buffer, unsigned long size,
862 struct fsck_options *options)
864 struct object_id tree_oid, parent_oid;
865 unsigned author_count;
866 int err;
867 const char *buffer_begin = buffer;
868 const char *buffer_end = buffer + size;
869 const char *p;
872 * We _must_ stop parsing immediately if this reports failure, as the
873 * memory safety of the rest of the function depends on it. See the
874 * comment above the definition of verify_headers() for more details.
876 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
877 return -1;
879 if (buffer >= buffer_end || !skip_prefix(buffer, "tree ", &buffer))
880 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
881 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
882 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
883 if (err)
884 return err;
886 buffer = p + 1;
887 while (buffer < buffer_end && skip_prefix(buffer, "parent ", &buffer)) {
888 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
889 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
890 if (err)
891 return err;
893 buffer = p + 1;
895 author_count = 0;
896 while (buffer < buffer_end && skip_prefix(buffer, "author ", &buffer)) {
897 author_count++;
898 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
899 if (err)
900 return err;
902 if (author_count < 1)
903 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
904 else if (author_count > 1)
905 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
906 if (err)
907 return err;
908 if (buffer >= buffer_end || !skip_prefix(buffer, "committer ", &buffer))
909 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
910 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
911 if (err)
912 return err;
913 if (memchr(buffer_begin, '\0', size)) {
914 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
915 "NUL byte in the commit object body");
916 if (err)
917 return err;
919 return 0;
922 static int fsck_tag(const struct object_id *oid, const char *buffer,
923 unsigned long size, struct fsck_options *options)
925 struct object_id tagged_oid;
926 int tagged_type;
927 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
928 &tagged_type);
931 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
932 unsigned long size, struct fsck_options *options,
933 struct object_id *tagged_oid,
934 int *tagged_type)
936 int ret = 0;
937 char *eol;
938 struct strbuf sb = STRBUF_INIT;
939 const char *buffer_end = buffer + size;
940 const char *p;
943 * We _must_ stop parsing immediately if this reports failure, as the
944 * memory safety of the rest of the function depends on it. See the
945 * comment above the definition of verify_headers() for more details.
947 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
948 if (ret)
949 goto done;
951 if (buffer >= buffer_end || !skip_prefix(buffer, "object ", &buffer)) {
952 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
953 goto done;
955 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
956 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
957 if (ret)
958 goto done;
960 buffer = p + 1;
962 if (buffer >= buffer_end || !skip_prefix(buffer, "type ", &buffer)) {
963 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
964 goto done;
966 eol = memchr(buffer, '\n', buffer_end - buffer);
967 if (!eol) {
968 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
969 goto done;
971 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
972 if (*tagged_type < 0)
973 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
974 if (ret)
975 goto done;
976 buffer = eol + 1;
978 if (buffer >= buffer_end || !skip_prefix(buffer, "tag ", &buffer)) {
979 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
980 goto done;
982 eol = memchr(buffer, '\n', buffer_end - buffer);
983 if (!eol) {
984 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
985 goto done;
987 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
988 if (check_refname_format(sb.buf, 0)) {
989 ret = report(options, oid, OBJ_TAG,
990 FSCK_MSG_BAD_TAG_NAME,
991 "invalid 'tag' name: %.*s",
992 (int)(eol - buffer), buffer);
993 if (ret)
994 goto done;
996 buffer = eol + 1;
998 if (buffer >= buffer_end || !skip_prefix(buffer, "tagger ", &buffer)) {
999 /* early tags do not contain 'tagger' lines; warn only */
1000 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
1001 if (ret)
1002 goto done;
1004 else
1005 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
1007 if (buffer < buffer_end && !starts_with(buffer, "\n")) {
1009 * The verify_headers() check will allow
1010 * e.g. "[...]tagger <tagger>\nsome
1011 * garbage\n\nmessage" to pass, thinking "some
1012 * garbage" could be a custom header. E.g. "mktag"
1013 * doesn't want any unknown headers.
1015 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1016 if (ret)
1017 goto done;
1020 done:
1021 strbuf_release(&sb);
1022 return ret;
1025 static int starts_with_dot_slash(const char *const path)
1027 return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_SLASH |
1028 PATH_MATCH_XPLATFORM);
1031 static int starts_with_dot_dot_slash(const char *const path)
1033 return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH |
1034 PATH_MATCH_XPLATFORM);
1037 static int submodule_url_is_relative(const char *url)
1039 return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
1043 * Count directory components that a relative submodule URL should chop
1044 * from the remote_url it is to be resolved against.
1046 * In other words, this counts "../" components at the start of a
1047 * submodule URL.
1049 * Returns the number of directory components to chop and writes a
1050 * pointer to the next character of url after all leading "./" and
1051 * "../" components to out.
1053 static int count_leading_dotdots(const char *url, const char **out)
1055 int result = 0;
1056 while (1) {
1057 if (starts_with_dot_dot_slash(url)) {
1058 result++;
1059 url += strlen("../");
1060 continue;
1062 if (starts_with_dot_slash(url)) {
1063 url += strlen("./");
1064 continue;
1066 *out = url;
1067 return result;
1071 * Check whether a transport is implemented by git-remote-curl.
1073 * If it is, returns 1 and writes the URL that would be passed to
1074 * git-remote-curl to the "out" parameter.
1076 * Otherwise, returns 0 and leaves "out" untouched.
1078 * Examples:
1079 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1080 * https://example.com/repo.git -> 1, https://example.com/repo.git
1081 * git://example.com/repo.git -> 0
1083 * This is for use in checking for previously exploitable bugs that
1084 * required a submodule URL to be passed to git-remote-curl.
1086 static int url_to_curl_url(const char *url, const char **out)
1089 * We don't need to check for case-aliases, "http.exe", and so
1090 * on because in the default configuration, is_transport_allowed
1091 * prevents URLs with those schemes from being cloned
1092 * automatically.
1094 if (skip_prefix(url, "http::", out) ||
1095 skip_prefix(url, "https::", out) ||
1096 skip_prefix(url, "ftp::", out) ||
1097 skip_prefix(url, "ftps::", out))
1098 return 1;
1099 if (starts_with(url, "http://") ||
1100 starts_with(url, "https://") ||
1101 starts_with(url, "ftp://") ||
1102 starts_with(url, "ftps://")) {
1103 *out = url;
1104 return 1;
1106 return 0;
1109 static int check_submodule_url(const char *url)
1111 const char *curl_url;
1113 if (looks_like_command_line_option(url))
1114 return -1;
1116 if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
1117 char *decoded;
1118 const char *next;
1119 int has_nl;
1122 * This could be appended to an http URL and url-decoded;
1123 * check for malicious characters.
1125 decoded = url_decode(url);
1126 has_nl = !!strchr(decoded, '\n');
1128 free(decoded);
1129 if (has_nl)
1130 return -1;
1133 * URLs which escape their root via "../" can overwrite
1134 * the host field and previous components, resolving to
1135 * URLs like https::example.com/submodule.git and
1136 * https:///example.com/submodule.git that were
1137 * susceptible to CVE-2020-11008.
1139 if (count_leading_dotdots(url, &next) > 0 &&
1140 (*next == ':' || *next == '/'))
1141 return -1;
1144 else if (url_to_curl_url(url, &curl_url)) {
1145 struct credential c = CREDENTIAL_INIT;
1146 int ret = 0;
1147 if (credential_from_url_gently(&c, curl_url, 1) ||
1148 !*c.host)
1149 ret = -1;
1150 credential_clear(&c);
1151 return ret;
1154 return 0;
1157 struct fsck_gitmodules_data {
1158 const struct object_id *oid;
1159 struct fsck_options *options;
1160 int ret;
1163 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1165 struct fsck_gitmodules_data *data = vdata;
1166 const char *subsection, *key;
1167 size_t subsection_len;
1168 char *name;
1170 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1171 !subsection)
1172 return 0;
1174 name = xmemdupz(subsection, subsection_len);
1175 if (check_submodule_name(name) < 0)
1176 data->ret |= report(data->options,
1177 data->oid, OBJ_BLOB,
1178 FSCK_MSG_GITMODULES_NAME,
1179 "disallowed submodule name: %s",
1180 name);
1181 if (!strcmp(key, "url") && value &&
1182 check_submodule_url(value) < 0)
1183 data->ret |= report(data->options,
1184 data->oid, OBJ_BLOB,
1185 FSCK_MSG_GITMODULES_URL,
1186 "disallowed submodule url: %s",
1187 value);
1188 if (!strcmp(key, "path") && value &&
1189 looks_like_command_line_option(value))
1190 data->ret |= report(data->options,
1191 data->oid, OBJ_BLOB,
1192 FSCK_MSG_GITMODULES_PATH,
1193 "disallowed submodule path: %s",
1194 value);
1195 if (!strcmp(key, "update") && value &&
1196 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1197 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1198 FSCK_MSG_GITMODULES_UPDATE,
1199 "disallowed submodule update setting: %s",
1200 value);
1201 free(name);
1203 return 0;
1206 static int fsck_blob(const struct object_id *oid, const char *buf,
1207 unsigned long size, struct fsck_options *options)
1209 int ret = 0;
1211 if (object_on_skiplist(options, oid))
1212 return 0;
1214 if (oidset_contains(&options->gitmodules_found, oid)) {
1215 struct config_options config_opts = { 0 };
1216 struct fsck_gitmodules_data data;
1218 oidset_insert(&options->gitmodules_done, oid);
1220 if (!buf) {
1222 * A missing buffer here is a sign that the caller found the
1223 * blob too gigantic to load into memory. Let's just consider
1224 * that an error.
1226 return report(options, oid, OBJ_BLOB,
1227 FSCK_MSG_GITMODULES_LARGE,
1228 ".gitmodules too large to parse");
1231 data.oid = oid;
1232 data.options = options;
1233 data.ret = 0;
1234 config_opts.error_action = CONFIG_ERROR_SILENT;
1235 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1236 ".gitmodules", buf, size, &data, &config_opts))
1237 data.ret |= report(options, oid, OBJ_BLOB,
1238 FSCK_MSG_GITMODULES_PARSE,
1239 "could not parse gitmodules blob");
1240 ret |= data.ret;
1243 if (oidset_contains(&options->gitattributes_found, oid)) {
1244 const char *ptr;
1246 oidset_insert(&options->gitattributes_done, oid);
1248 if (!buf || size > ATTR_MAX_FILE_SIZE) {
1250 * A missing buffer here is a sign that the caller found the
1251 * blob too gigantic to load into memory. Let's just consider
1252 * that an error.
1254 return report(options, oid, OBJ_BLOB,
1255 FSCK_MSG_GITATTRIBUTES_LARGE,
1256 ".gitattributes too large to parse");
1259 for (ptr = buf; *ptr; ) {
1260 const char *eol = strchrnul(ptr, '\n');
1261 if (eol - ptr >= ATTR_MAX_LINE_LENGTH) {
1262 ret |= report(options, oid, OBJ_BLOB,
1263 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH,
1264 ".gitattributes has too long lines to parse");
1265 break;
1268 ptr = *eol ? eol + 1 : eol;
1272 return ret;
1275 int fsck_object(struct object *obj, void *data, unsigned long size,
1276 struct fsck_options *options)
1278 if (!obj)
1279 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1281 return fsck_buffer(&obj->oid, obj->type, data, size, options);
1284 int fsck_buffer(const struct object_id *oid, enum object_type type,
1285 void *data, unsigned long size,
1286 struct fsck_options *options)
1288 if (type == OBJ_BLOB)
1289 return fsck_blob(oid, data, size, options);
1290 if (type == OBJ_TREE)
1291 return fsck_tree(oid, data, size, options);
1292 if (type == OBJ_COMMIT)
1293 return fsck_commit(oid, data, size, options);
1294 if (type == OBJ_TAG)
1295 return fsck_tag(oid, data, size, options);
1297 return report(options, oid, type,
1298 FSCK_MSG_UNKNOWN_TYPE,
1299 "unknown type '%d' (internal fsck error)",
1300 type);
1303 int fsck_error_function(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_type == FSCK_WARN) {
1311 warning("object %s: %s", fsck_describe_object(o, oid), message);
1312 return 0;
1314 error("object %s: %s", fsck_describe_object(o, oid), message);
1315 return 1;
1318 static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
1319 enum fsck_msg_id msg_missing, enum fsck_msg_id msg_type,
1320 struct fsck_options *options, const char *blob_type)
1322 int ret = 0;
1323 struct oidset_iter iter;
1324 const struct object_id *oid;
1326 oidset_iter_init(blobs_found, &iter);
1327 while ((oid = oidset_iter_next(&iter))) {
1328 enum object_type type;
1329 unsigned long size;
1330 char *buf;
1332 if (oidset_contains(blobs_done, oid))
1333 continue;
1335 buf = read_object_file(oid, &type, &size);
1336 if (!buf) {
1337 if (is_promisor_object(oid))
1338 continue;
1339 ret |= report(options,
1340 oid, OBJ_BLOB, msg_missing,
1341 "unable to read %s blob", blob_type);
1342 continue;
1345 if (type == OBJ_BLOB)
1346 ret |= fsck_blob(oid, buf, size, options);
1347 else
1348 ret |= report(options, oid, type, msg_type,
1349 "non-blob found at %s", blob_type);
1350 free(buf);
1353 oidset_clear(blobs_found);
1354 oidset_clear(blobs_done);
1356 return ret;
1359 int fsck_finish(struct fsck_options *options)
1361 int ret = 0;
1363 ret |= fsck_blobs(&options->gitmodules_found, &options->gitmodules_done,
1364 FSCK_MSG_GITMODULES_MISSING, FSCK_MSG_GITMODULES_BLOB,
1365 options, ".gitmodules");
1366 ret |= fsck_blobs(&options->gitattributes_found, &options->gitattributes_done,
1367 FSCK_MSG_GITATTRIBUTES_MISSING, FSCK_MSG_GITATTRIBUTES_BLOB,
1368 options, ".gitattributes");
1370 return ret;
1373 int git_fsck_config(const char *var, const char *value, void *cb)
1375 struct fsck_options *options = cb;
1376 if (strcmp(var, "fsck.skiplist") == 0) {
1377 const char *path;
1378 struct strbuf sb = STRBUF_INIT;
1380 if (git_config_pathname(&path, var, value))
1381 return 1;
1382 strbuf_addf(&sb, "skiplist=%s", path);
1383 free((char *)path);
1384 fsck_set_msg_types(options, sb.buf);
1385 strbuf_release(&sb);
1386 return 0;
1389 if (skip_prefix(var, "fsck.", &var)) {
1390 fsck_set_msg_type(options, var, value);
1391 return 0;
1394 return git_default_config(var, value, cb);
1398 * Custom error callbacks that are used in more than one place.
1401 int fsck_error_cb_print_missing_gitmodules(struct fsck_options *o,
1402 const struct object_id *oid,
1403 enum object_type object_type,
1404 enum fsck_msg_type msg_type,
1405 enum fsck_msg_id msg_id,
1406 const char *message)
1408 if (msg_id == FSCK_MSG_GITMODULES_MISSING) {
1409 puts(oid_to_hex(oid));
1410 return 0;
1412 return fsck_error_function(o, oid, object_type, msg_type, msg_id, message);