Hopefully the last batch before -rc1
[git.git] / fsck.c
blobe3030f3b358045cfad4f4cdcc64d529677bda124
1 #include "cache.h"
2 #include "object-store.h"
3 #include "repository.h"
4 #include "object.h"
5 #include "blob.h"
6 #include "tree.h"
7 #include "tree-walk.h"
8 #include "commit.h"
9 #include "tag.h"
10 #include "fsck.h"
11 #include "refs.h"
12 #include "url.h"
13 #include "utf8.h"
14 #include "decorate.h"
15 #include "oidset.h"
16 #include "packfile.h"
17 #include "submodule-config.h"
18 #include "config.h"
19 #include "credential.h"
20 #include "help.h"
22 static struct oidset gitmodules_found = OIDSET_INIT;
23 static struct oidset gitmodules_done = OIDSET_INIT;
25 #define FSCK_FATAL -1
26 #define FSCK_INFO -2
28 #define FOREACH_MSG_ID(FUNC) \
29 /* fatal errors */ \
30 FUNC(NUL_IN_HEADER, FATAL) \
31 FUNC(UNTERMINATED_HEADER, FATAL) \
32 /* errors */ \
33 FUNC(BAD_DATE, ERROR) \
34 FUNC(BAD_DATE_OVERFLOW, ERROR) \
35 FUNC(BAD_EMAIL, ERROR) \
36 FUNC(BAD_NAME, ERROR) \
37 FUNC(BAD_OBJECT_SHA1, ERROR) \
38 FUNC(BAD_PARENT_SHA1, ERROR) \
39 FUNC(BAD_TAG_OBJECT, ERROR) \
40 FUNC(BAD_TIMEZONE, ERROR) \
41 FUNC(BAD_TREE, ERROR) \
42 FUNC(BAD_TREE_SHA1, ERROR) \
43 FUNC(BAD_TYPE, ERROR) \
44 FUNC(DUPLICATE_ENTRIES, ERROR) \
45 FUNC(MISSING_AUTHOR, ERROR) \
46 FUNC(MISSING_COMMITTER, ERROR) \
47 FUNC(MISSING_EMAIL, ERROR) \
48 FUNC(MISSING_NAME_BEFORE_EMAIL, ERROR) \
49 FUNC(MISSING_OBJECT, ERROR) \
50 FUNC(MISSING_SPACE_BEFORE_DATE, ERROR) \
51 FUNC(MISSING_SPACE_BEFORE_EMAIL, ERROR) \
52 FUNC(MISSING_TAG, ERROR) \
53 FUNC(MISSING_TAG_ENTRY, ERROR) \
54 FUNC(MISSING_TREE, ERROR) \
55 FUNC(MISSING_TREE_OBJECT, ERROR) \
56 FUNC(MISSING_TYPE, ERROR) \
57 FUNC(MISSING_TYPE_ENTRY, ERROR) \
58 FUNC(MULTIPLE_AUTHORS, ERROR) \
59 FUNC(TREE_NOT_SORTED, ERROR) \
60 FUNC(UNKNOWN_TYPE, ERROR) \
61 FUNC(ZERO_PADDED_DATE, ERROR) \
62 FUNC(GITMODULES_MISSING, ERROR) \
63 FUNC(GITMODULES_BLOB, ERROR) \
64 FUNC(GITMODULES_LARGE, ERROR) \
65 FUNC(GITMODULES_NAME, ERROR) \
66 FUNC(GITMODULES_SYMLINK, ERROR) \
67 FUNC(GITMODULES_URL, ERROR) \
68 FUNC(GITMODULES_PATH, ERROR) \
69 FUNC(GITMODULES_UPDATE, ERROR) \
70 /* warnings */ \
71 FUNC(BAD_FILEMODE, WARN) \
72 FUNC(EMPTY_NAME, WARN) \
73 FUNC(FULL_PATHNAME, WARN) \
74 FUNC(HAS_DOT, WARN) \
75 FUNC(HAS_DOTDOT, WARN) \
76 FUNC(HAS_DOTGIT, WARN) \
77 FUNC(NULL_SHA1, WARN) \
78 FUNC(ZERO_PADDED_FILEMODE, WARN) \
79 FUNC(NUL_IN_COMMIT, WARN) \
80 /* infos (reported as warnings, but ignored by default) */ \
81 FUNC(GITMODULES_PARSE, INFO) \
82 FUNC(BAD_TAG_NAME, INFO) \
83 FUNC(MISSING_TAGGER_ENTRY, INFO) \
84 /* ignored (elevated when requested) */ \
85 FUNC(EXTRA_HEADER_ENTRY, IGNORE)
87 #define MSG_ID(id, msg_type) FSCK_MSG_##id,
88 enum fsck_msg_id {
89 FOREACH_MSG_ID(MSG_ID)
90 FSCK_MSG_MAX
92 #undef MSG_ID
94 #define STR(x) #x
95 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
96 static struct {
97 const char *id_string;
98 const char *downcased;
99 const char *camelcased;
100 int msg_type;
101 } msg_id_info[FSCK_MSG_MAX + 1] = {
102 FOREACH_MSG_ID(MSG_ID)
103 { NULL, NULL, NULL, -1 }
105 #undef MSG_ID
107 static void prepare_msg_ids(void)
109 int i;
111 if (msg_id_info[0].downcased)
112 return;
114 /* convert id_string to lower case, without underscores. */
115 for (i = 0; i < FSCK_MSG_MAX; i++) {
116 const char *p = msg_id_info[i].id_string;
117 int len = strlen(p);
118 char *q = xmalloc(len);
120 msg_id_info[i].downcased = q;
121 while (*p)
122 if (*p == '_')
123 p++;
124 else
125 *(q)++ = tolower(*(p)++);
126 *q = '\0';
128 p = msg_id_info[i].id_string;
129 q = xmalloc(len);
130 msg_id_info[i].camelcased = q;
131 while (*p) {
132 if (*p == '_') {
133 p++;
134 if (*p)
135 *q++ = *p++;
136 } else {
137 *q++ = tolower(*p++);
140 *q = '\0';
144 static int parse_msg_id(const char *text)
146 int i;
148 prepare_msg_ids();
150 for (i = 0; i < FSCK_MSG_MAX; i++)
151 if (!strcmp(text, msg_id_info[i].downcased))
152 return i;
154 return -1;
157 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
159 int i;
161 prepare_msg_ids();
163 for (i = 0; i < FSCK_MSG_MAX; i++)
164 list_config_item(list, prefix, msg_id_info[i].camelcased);
167 static int fsck_msg_type(enum fsck_msg_id msg_id,
168 struct fsck_options *options)
170 int msg_type;
172 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
174 if (options->msg_type)
175 msg_type = options->msg_type[msg_id];
176 else {
177 msg_type = msg_id_info[msg_id].msg_type;
178 if (options->strict && msg_type == FSCK_WARN)
179 msg_type = FSCK_ERROR;
182 return msg_type;
185 static int parse_msg_type(const char *str)
187 if (!strcmp(str, "error"))
188 return FSCK_ERROR;
189 else if (!strcmp(str, "warn"))
190 return FSCK_WARN;
191 else if (!strcmp(str, "ignore"))
192 return FSCK_IGNORE;
193 else
194 die("Unknown fsck message type: '%s'", str);
197 int is_valid_msg_type(const char *msg_id, const char *msg_type)
199 if (parse_msg_id(msg_id) < 0)
200 return 0;
201 parse_msg_type(msg_type);
202 return 1;
205 void fsck_set_msg_type(struct fsck_options *options,
206 const char *msg_id, const char *msg_type)
208 int id = parse_msg_id(msg_id), type;
210 if (id < 0)
211 die("Unhandled message id: %s", msg_id);
212 type = parse_msg_type(msg_type);
214 if (type != FSCK_ERROR && msg_id_info[id].msg_type == FSCK_FATAL)
215 die("Cannot demote %s to %s", msg_id, msg_type);
217 if (!options->msg_type) {
218 int i;
219 int *msg_type;
220 ALLOC_ARRAY(msg_type, FSCK_MSG_MAX);
221 for (i = 0; i < FSCK_MSG_MAX; i++)
222 msg_type[i] = fsck_msg_type(i, options);
223 options->msg_type = msg_type;
226 options->msg_type[id] = type;
229 void fsck_set_msg_types(struct fsck_options *options, const char *values)
231 char *buf = xstrdup(values), *to_free = buf;
232 int done = 0;
234 while (!done) {
235 int len = strcspn(buf, " ,|"), equal;
237 done = !buf[len];
238 if (!len) {
239 buf++;
240 continue;
242 buf[len] = '\0';
244 for (equal = 0;
245 equal < len && buf[equal] != '=' && buf[equal] != ':';
246 equal++)
247 buf[equal] = tolower(buf[equal]);
248 buf[equal] = '\0';
250 if (!strcmp(buf, "skiplist")) {
251 if (equal == len)
252 die("skiplist requires a path");
253 oidset_parse_file(&options->skiplist, buf + equal + 1);
254 buf += len + 1;
255 continue;
258 if (equal == len)
259 die("Missing '=': '%s'", buf);
261 fsck_set_msg_type(options, buf, buf + equal + 1);
262 buf += len + 1;
264 free(to_free);
267 static void append_msg_id(struct strbuf *sb, const char *msg_id)
269 for (;;) {
270 char c = *(msg_id)++;
272 if (!c)
273 break;
274 if (c != '_')
275 strbuf_addch(sb, tolower(c));
276 else {
277 assert(*msg_id);
278 strbuf_addch(sb, *(msg_id)++);
282 strbuf_addstr(sb, ": ");
285 static int object_on_skiplist(struct fsck_options *opts,
286 const struct object_id *oid)
288 return opts && oid && oidset_contains(&opts->skiplist, oid);
291 __attribute__((format (printf, 5, 6)))
292 static int report(struct fsck_options *options,
293 const struct object_id *oid, enum object_type object_type,
294 enum fsck_msg_id id, const char *fmt, ...)
296 va_list ap;
297 struct strbuf sb = STRBUF_INIT;
298 int msg_type = fsck_msg_type(id, options), result;
300 if (msg_type == FSCK_IGNORE)
301 return 0;
303 if (object_on_skiplist(options, oid))
304 return 0;
306 if (msg_type == FSCK_FATAL)
307 msg_type = FSCK_ERROR;
308 else if (msg_type == FSCK_INFO)
309 msg_type = FSCK_WARN;
311 append_msg_id(&sb, msg_id_info[id].id_string);
313 va_start(ap, fmt);
314 strbuf_vaddf(&sb, fmt, ap);
315 result = options->error_func(options, oid, object_type,
316 msg_type, sb.buf);
317 strbuf_release(&sb);
318 va_end(ap);
320 return result;
323 void fsck_enable_object_names(struct fsck_options *options)
325 if (!options->object_names)
326 options->object_names = kh_init_oid_map();
329 const char *fsck_get_object_name(struct fsck_options *options,
330 const struct object_id *oid)
332 khiter_t pos;
333 if (!options->object_names)
334 return NULL;
335 pos = kh_get_oid_map(options->object_names, *oid);
336 if (pos >= kh_end(options->object_names))
337 return NULL;
338 return kh_value(options->object_names, pos);
341 void fsck_put_object_name(struct fsck_options *options,
342 const struct object_id *oid,
343 const char *fmt, ...)
345 va_list ap;
346 struct strbuf buf = STRBUF_INIT;
347 khiter_t pos;
348 int hashret;
350 if (!options->object_names)
351 return;
353 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
354 if (!hashret)
355 return;
356 va_start(ap, fmt);
357 strbuf_vaddf(&buf, fmt, ap);
358 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
359 va_end(ap);
362 const char *fsck_describe_object(struct fsck_options *options,
363 const struct object_id *oid)
365 static struct strbuf bufs[] = {
366 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
368 static int b = 0;
369 struct strbuf *buf;
370 const char *name = fsck_get_object_name(options, oid);
372 buf = bufs + b;
373 b = (b + 1) % ARRAY_SIZE(bufs);
374 strbuf_reset(buf);
375 strbuf_addstr(buf, oid_to_hex(oid));
376 if (name)
377 strbuf_addf(buf, " (%s)", name);
379 return buf->buf;
382 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
384 struct tree_desc desc;
385 struct name_entry entry;
386 int res = 0;
387 const char *name;
389 if (parse_tree(tree))
390 return -1;
392 name = fsck_get_object_name(options, &tree->object.oid);
393 if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
394 return -1;
395 while (tree_entry_gently(&desc, &entry)) {
396 struct object *obj;
397 int result;
399 if (S_ISGITLINK(entry.mode))
400 continue;
402 if (S_ISDIR(entry.mode)) {
403 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
404 if (name && obj)
405 fsck_put_object_name(options, &entry.oid, "%s%s/",
406 name, entry.path);
407 result = options->walk(obj, OBJ_TREE, data, options);
409 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
410 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
411 if (name && obj)
412 fsck_put_object_name(options, &entry.oid, "%s%s",
413 name, entry.path);
414 result = options->walk(obj, OBJ_BLOB, data, options);
416 else {
417 result = error("in tree %s: entry %s has bad mode %.6o",
418 fsck_describe_object(options, &tree->object.oid),
419 entry.path, entry.mode);
421 if (result < 0)
422 return result;
423 if (!res)
424 res = result;
426 return res;
429 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
431 int counter = 0, generation = 0, name_prefix_len = 0;
432 struct commit_list *parents;
433 int res;
434 int result;
435 const char *name;
437 if (parse_commit(commit))
438 return -1;
440 name = fsck_get_object_name(options, &commit->object.oid);
441 if (name)
442 fsck_put_object_name(options, get_commit_tree_oid(commit),
443 "%s:", name);
445 result = options->walk((struct object *)get_commit_tree(commit),
446 OBJ_TREE, data, options);
447 if (result < 0)
448 return result;
449 res = result;
451 parents = commit->parents;
452 if (name && parents) {
453 int len = strlen(name), power;
455 if (len && name[len - 1] == '^') {
456 generation = 1;
457 name_prefix_len = len - 1;
459 else { /* parse ~<generation> suffix */
460 for (generation = 0, power = 1;
461 len && isdigit(name[len - 1]);
462 power *= 10)
463 generation += power * (name[--len] - '0');
464 if (power > 1 && len && name[len - 1] == '~')
465 name_prefix_len = len - 1;
466 else {
467 /* Maybe a non-first parent, e.g. HEAD^2 */
468 generation = 0;
469 name_prefix_len = len;
474 while (parents) {
475 if (name) {
476 struct object_id *oid = &parents->item->object.oid;
478 if (counter++)
479 fsck_put_object_name(options, oid, "%s^%d",
480 name, counter);
481 else if (generation > 0)
482 fsck_put_object_name(options, oid, "%.*s~%d",
483 name_prefix_len, name,
484 generation + 1);
485 else
486 fsck_put_object_name(options, oid, "%s^", name);
488 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
489 if (result < 0)
490 return result;
491 if (!res)
492 res = result;
493 parents = parents->next;
495 return res;
498 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
500 const char *name = fsck_get_object_name(options, &tag->object.oid);
502 if (parse_tag(tag))
503 return -1;
504 if (name)
505 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
506 return options->walk(tag->tagged, OBJ_ANY, data, options);
509 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
511 if (!obj)
512 return -1;
514 if (obj->type == OBJ_NONE)
515 parse_object(the_repository, &obj->oid);
517 switch (obj->type) {
518 case OBJ_BLOB:
519 return 0;
520 case OBJ_TREE:
521 return fsck_walk_tree((struct tree *)obj, data, options);
522 case OBJ_COMMIT:
523 return fsck_walk_commit((struct commit *)obj, data, options);
524 case OBJ_TAG:
525 return fsck_walk_tag((struct tag *)obj, data, options);
526 default:
527 error("Unknown object type for %s",
528 fsck_describe_object(options, &obj->oid));
529 return -1;
533 struct name_stack {
534 const char **names;
535 size_t nr, alloc;
538 static void name_stack_push(struct name_stack *stack, const char *name)
540 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
541 stack->names[stack->nr++] = name;
544 static const char *name_stack_pop(struct name_stack *stack)
546 return stack->nr ? stack->names[--stack->nr] : NULL;
549 static void name_stack_clear(struct name_stack *stack)
551 FREE_AND_NULL(stack->names);
552 stack->nr = stack->alloc = 0;
556 * The entries in a tree are ordered in the _path_ order,
557 * which means that a directory entry is ordered by adding
558 * a slash to the end of it.
560 * So a directory called "a" is ordered _after_ a file
561 * called "a.c", because "a/" sorts after "a.c".
563 #define TREE_UNORDERED (-1)
564 #define TREE_HAS_DUPS (-2)
566 static int is_less_than_slash(unsigned char c)
568 return '\0' < c && c < '/';
571 static int verify_ordered(unsigned mode1, const char *name1,
572 unsigned mode2, const char *name2,
573 struct name_stack *candidates)
575 int len1 = strlen(name1);
576 int len2 = strlen(name2);
577 int len = len1 < len2 ? len1 : len2;
578 unsigned char c1, c2;
579 int cmp;
581 cmp = memcmp(name1, name2, len);
582 if (cmp < 0)
583 return 0;
584 if (cmp > 0)
585 return TREE_UNORDERED;
588 * Ok, the first <len> characters are the same.
589 * Now we need to order the next one, but turn
590 * a '\0' into a '/' for a directory entry.
592 c1 = name1[len];
593 c2 = name2[len];
594 if (!c1 && !c2)
596 * git-write-tree used to write out a nonsense tree that has
597 * entries with the same name, one blob and one tree. Make
598 * sure we do not have duplicate entries.
600 return TREE_HAS_DUPS;
601 if (!c1 && S_ISDIR(mode1))
602 c1 = '/';
603 if (!c2 && S_ISDIR(mode2))
604 c2 = '/';
607 * There can be non-consecutive duplicates due to the implicitly
608 * added slash, e.g.:
610 * foo
611 * foo.bar
612 * foo.bar.baz
613 * foo.bar/
614 * foo/
616 * Record non-directory candidates (like "foo" and "foo.bar" in
617 * the example) on a stack and check directory candidates (like
618 * foo/" and "foo.bar/") against that stack.
620 if (!c1 && is_less_than_slash(c2)) {
621 name_stack_push(candidates, name1);
622 } else if (c2 == '/' && is_less_than_slash(c1)) {
623 for (;;) {
624 const char *p;
625 const char *f_name = name_stack_pop(candidates);
627 if (!f_name)
628 break;
629 if (!skip_prefix(name2, f_name, &p))
630 continue;
631 if (!*p)
632 return TREE_HAS_DUPS;
633 if (is_less_than_slash(*p)) {
634 name_stack_push(candidates, f_name);
635 break;
640 return c1 < c2 ? 0 : TREE_UNORDERED;
643 static int fsck_tree(const struct object_id *oid,
644 const char *buffer, unsigned long size,
645 struct fsck_options *options)
647 int retval = 0;
648 int has_null_sha1 = 0;
649 int has_full_path = 0;
650 int has_empty_name = 0;
651 int has_dot = 0;
652 int has_dotdot = 0;
653 int has_dotgit = 0;
654 int has_zero_pad = 0;
655 int has_bad_modes = 0;
656 int has_dup_entries = 0;
657 int not_properly_sorted = 0;
658 struct tree_desc desc;
659 unsigned o_mode;
660 const char *o_name;
661 struct name_stack df_dup_candidates = { NULL };
663 if (init_tree_desc_gently(&desc, buffer, size)) {
664 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
665 return retval;
668 o_mode = 0;
669 o_name = NULL;
671 while (desc.size) {
672 unsigned short mode;
673 const char *name, *backslash;
674 const struct object_id *oid;
676 oid = tree_entry_extract(&desc, &name, &mode);
678 has_null_sha1 |= is_null_oid(oid);
679 has_full_path |= !!strchr(name, '/');
680 has_empty_name |= !*name;
681 has_dot |= !strcmp(name, ".");
682 has_dotdot |= !strcmp(name, "..");
683 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
684 has_zero_pad |= *(char *)desc.buffer == '0';
686 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
687 if (!S_ISLNK(mode))
688 oidset_insert(&gitmodules_found, oid);
689 else
690 retval += report(options,
691 oid, OBJ_TREE,
692 FSCK_MSG_GITMODULES_SYMLINK,
693 ".gitmodules is a symbolic link");
696 if ((backslash = strchr(name, '\\'))) {
697 while (backslash) {
698 backslash++;
699 has_dotgit |= is_ntfs_dotgit(backslash);
700 if (is_ntfs_dotgitmodules(backslash)) {
701 if (!S_ISLNK(mode))
702 oidset_insert(&gitmodules_found, oid);
703 else
704 retval += report(options, oid, OBJ_TREE,
705 FSCK_MSG_GITMODULES_SYMLINK,
706 ".gitmodules is a symbolic link");
708 backslash = strchr(backslash, '\\');
712 if (update_tree_entry_gently(&desc)) {
713 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
714 break;
717 switch (mode) {
719 * Standard modes..
721 case S_IFREG | 0755:
722 case S_IFREG | 0644:
723 case S_IFLNK:
724 case S_IFDIR:
725 case S_IFGITLINK:
726 break;
728 * This is nonstandard, but we had a few of these
729 * early on when we honored the full set of mode
730 * bits..
732 case S_IFREG | 0664:
733 if (!options->strict)
734 break;
735 /* fallthrough */
736 default:
737 has_bad_modes = 1;
740 if (o_name) {
741 switch (verify_ordered(o_mode, o_name, mode, name,
742 &df_dup_candidates)) {
743 case TREE_UNORDERED:
744 not_properly_sorted = 1;
745 break;
746 case TREE_HAS_DUPS:
747 has_dup_entries = 1;
748 break;
749 default:
750 break;
754 o_mode = mode;
755 o_name = name;
758 name_stack_clear(&df_dup_candidates);
760 if (has_null_sha1)
761 retval += report(options, oid, OBJ_TREE, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
762 if (has_full_path)
763 retval += report(options, oid, OBJ_TREE, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
764 if (has_empty_name)
765 retval += report(options, oid, OBJ_TREE, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
766 if (has_dot)
767 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOT, "contains '.'");
768 if (has_dotdot)
769 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTDOT, "contains '..'");
770 if (has_dotgit)
771 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
772 if (has_zero_pad)
773 retval += report(options, oid, OBJ_TREE, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
774 if (has_bad_modes)
775 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
776 if (has_dup_entries)
777 retval += report(options, oid, OBJ_TREE, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
778 if (not_properly_sorted)
779 retval += report(options, oid, OBJ_TREE, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
780 return retval;
783 static int verify_headers(const void *data, unsigned long size,
784 const struct object_id *oid, enum object_type type,
785 struct fsck_options *options)
787 const char *buffer = (const char *)data;
788 unsigned long i;
790 for (i = 0; i < size; i++) {
791 switch (buffer[i]) {
792 case '\0':
793 return report(options, oid, type,
794 FSCK_MSG_NUL_IN_HEADER,
795 "unterminated header: NUL at offset %ld", i);
796 case '\n':
797 if (i + 1 < size && buffer[i + 1] == '\n')
798 return 0;
803 * We did not find double-LF that separates the header
804 * and the body. Not having a body is not a crime but
805 * we do want to see the terminating LF for the last header
806 * line.
808 if (size && buffer[size - 1] == '\n')
809 return 0;
811 return report(options, oid, type,
812 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
815 static int fsck_ident(const char **ident,
816 const struct object_id *oid, enum object_type type,
817 struct fsck_options *options)
819 const char *p = *ident;
820 char *end;
822 *ident = strchrnul(*ident, '\n');
823 if (**ident == '\n')
824 (*ident)++;
826 if (*p == '<')
827 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
828 p += strcspn(p, "<>\n");
829 if (*p == '>')
830 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
831 if (*p != '<')
832 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
833 if (p[-1] != ' ')
834 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
835 p++;
836 p += strcspn(p, "<>\n");
837 if (*p != '>')
838 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
839 p++;
840 if (*p != ' ')
841 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
842 p++;
843 if (*p == '0' && p[1] != ' ')
844 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
845 if (date_overflows(parse_timestamp(p, &end, 10)))
846 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
847 if ((end == p || *end != ' '))
848 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
849 p = end + 1;
850 if ((*p != '+' && *p != '-') ||
851 !isdigit(p[1]) ||
852 !isdigit(p[2]) ||
853 !isdigit(p[3]) ||
854 !isdigit(p[4]) ||
855 (p[5] != '\n'))
856 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
857 p += 6;
858 return 0;
861 static int fsck_commit(const struct object_id *oid,
862 const char *buffer, unsigned long size,
863 struct fsck_options *options)
865 struct object_id tree_oid, parent_oid;
866 unsigned author_count;
867 int err;
868 const char *buffer_begin = buffer;
869 const char *p;
871 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
872 return -1;
874 if (!skip_prefix(buffer, "tree ", &buffer))
875 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
876 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
877 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
878 if (err)
879 return err;
881 buffer = p + 1;
882 while (skip_prefix(buffer, "parent ", &buffer)) {
883 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
884 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
885 if (err)
886 return err;
888 buffer = p + 1;
890 author_count = 0;
891 while (skip_prefix(buffer, "author ", &buffer)) {
892 author_count++;
893 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
894 if (err)
895 return err;
897 if (author_count < 1)
898 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
899 else if (author_count > 1)
900 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
901 if (err)
902 return err;
903 if (!skip_prefix(buffer, "committer ", &buffer))
904 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
905 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
906 if (err)
907 return err;
908 if (memchr(buffer_begin, '\0', size)) {
909 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
910 "NUL byte in the commit object body");
911 if (err)
912 return err;
914 return 0;
917 static int fsck_tag(const struct object_id *oid, const char *buffer,
918 unsigned long size, struct fsck_options *options)
920 struct object_id tagged_oid;
921 int tagged_type;
922 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
923 &tagged_type);
926 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
927 unsigned long size, struct fsck_options *options,
928 struct object_id *tagged_oid,
929 int *tagged_type)
931 int ret = 0;
932 char *eol;
933 struct strbuf sb = STRBUF_INIT;
934 const char *p;
936 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
937 if (ret)
938 goto done;
940 if (!skip_prefix(buffer, "object ", &buffer)) {
941 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
942 goto done;
944 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
945 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
946 if (ret)
947 goto done;
949 buffer = p + 1;
951 if (!skip_prefix(buffer, "type ", &buffer)) {
952 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
953 goto done;
955 eol = strchr(buffer, '\n');
956 if (!eol) {
957 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
958 goto done;
960 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
961 if (*tagged_type < 0)
962 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
963 if (ret)
964 goto done;
965 buffer = eol + 1;
967 if (!skip_prefix(buffer, "tag ", &buffer)) {
968 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
969 goto done;
971 eol = strchr(buffer, '\n');
972 if (!eol) {
973 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
974 goto done;
976 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
977 if (check_refname_format(sb.buf, 0)) {
978 ret = report(options, oid, OBJ_TAG,
979 FSCK_MSG_BAD_TAG_NAME,
980 "invalid 'tag' name: %.*s",
981 (int)(eol - buffer), buffer);
982 if (ret)
983 goto done;
985 buffer = eol + 1;
987 if (!skip_prefix(buffer, "tagger ", &buffer)) {
988 /* early tags do not contain 'tagger' lines; warn only */
989 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
990 if (ret)
991 goto done;
993 else
994 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
995 if (!*buffer)
996 goto done;
998 if (!starts_with(buffer, "\n")) {
1000 * The verify_headers() check will allow
1001 * e.g. "[...]tagger <tagger>\nsome
1002 * garbage\n\nmessage" to pass, thinking "some
1003 * garbage" could be a custom header. E.g. "mktag"
1004 * doesn't want any unknown headers.
1006 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1007 if (ret)
1008 goto done;
1011 done:
1012 strbuf_release(&sb);
1013 return ret;
1017 * Like builtin/submodule--helper.c's starts_with_dot_slash, but without
1018 * relying on the platform-dependent is_dir_sep helper.
1020 * This is for use in checking whether a submodule URL is interpreted as
1021 * relative to the current directory on any platform, since \ is a
1022 * directory separator on Windows but not on other platforms.
1024 static int starts_with_dot_slash(const char *str)
1026 return str[0] == '.' && (str[1] == '/' || str[1] == '\\');
1030 * Like starts_with_dot_slash, this is a variant of submodule--helper's
1031 * helper of the same name with the twist that it accepts backslash as a
1032 * directory separator even on non-Windows platforms.
1034 static int starts_with_dot_dot_slash(const char *str)
1036 return str[0] == '.' && starts_with_dot_slash(str + 1);
1039 static int submodule_url_is_relative(const char *url)
1041 return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
1045 * Count directory components that a relative submodule URL should chop
1046 * from the remote_url it is to be resolved against.
1048 * In other words, this counts "../" components at the start of a
1049 * submodule URL.
1051 * Returns the number of directory components to chop and writes a
1052 * pointer to the next character of url after all leading "./" and
1053 * "../" components to out.
1055 static int count_leading_dotdots(const char *url, const char **out)
1057 int result = 0;
1058 while (1) {
1059 if (starts_with_dot_dot_slash(url)) {
1060 result++;
1061 url += strlen("../");
1062 continue;
1064 if (starts_with_dot_slash(url)) {
1065 url += strlen("./");
1066 continue;
1068 *out = url;
1069 return result;
1073 * Check whether a transport is implemented by git-remote-curl.
1075 * If it is, returns 1 and writes the URL that would be passed to
1076 * git-remote-curl to the "out" parameter.
1078 * Otherwise, returns 0 and leaves "out" untouched.
1080 * Examples:
1081 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1082 * https://example.com/repo.git -> 1, https://example.com/repo.git
1083 * git://example.com/repo.git -> 0
1085 * This is for use in checking for previously exploitable bugs that
1086 * required a submodule URL to be passed to git-remote-curl.
1088 static int url_to_curl_url(const char *url, const char **out)
1091 * We don't need to check for case-aliases, "http.exe", and so
1092 * on because in the default configuration, is_transport_allowed
1093 * prevents URLs with those schemes from being cloned
1094 * automatically.
1096 if (skip_prefix(url, "http::", out) ||
1097 skip_prefix(url, "https::", out) ||
1098 skip_prefix(url, "ftp::", out) ||
1099 skip_prefix(url, "ftps::", out))
1100 return 1;
1101 if (starts_with(url, "http://") ||
1102 starts_with(url, "https://") ||
1103 starts_with(url, "ftp://") ||
1104 starts_with(url, "ftps://")) {
1105 *out = url;
1106 return 1;
1108 return 0;
1111 static int check_submodule_url(const char *url)
1113 const char *curl_url;
1115 if (looks_like_command_line_option(url))
1116 return -1;
1118 if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
1119 char *decoded;
1120 const char *next;
1121 int has_nl;
1124 * This could be appended to an http URL and url-decoded;
1125 * check for malicious characters.
1127 decoded = url_decode(url);
1128 has_nl = !!strchr(decoded, '\n');
1130 free(decoded);
1131 if (has_nl)
1132 return -1;
1135 * URLs which escape their root via "../" can overwrite
1136 * the host field and previous components, resolving to
1137 * URLs like https::example.com/submodule.git and
1138 * https:///example.com/submodule.git that were
1139 * susceptible to CVE-2020-11008.
1141 if (count_leading_dotdots(url, &next) > 0 &&
1142 (*next == ':' || *next == '/'))
1143 return -1;
1146 else if (url_to_curl_url(url, &curl_url)) {
1147 struct credential c = CREDENTIAL_INIT;
1148 int ret = 0;
1149 if (credential_from_url_gently(&c, curl_url, 1) ||
1150 !*c.host)
1151 ret = -1;
1152 credential_clear(&c);
1153 return ret;
1156 return 0;
1159 struct fsck_gitmodules_data {
1160 const struct object_id *oid;
1161 struct fsck_options *options;
1162 int ret;
1165 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1167 struct fsck_gitmodules_data *data = vdata;
1168 const char *subsection, *key;
1169 size_t subsection_len;
1170 char *name;
1172 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1173 !subsection)
1174 return 0;
1176 name = xmemdupz(subsection, subsection_len);
1177 if (check_submodule_name(name) < 0)
1178 data->ret |= report(data->options,
1179 data->oid, OBJ_BLOB,
1180 FSCK_MSG_GITMODULES_NAME,
1181 "disallowed submodule name: %s",
1182 name);
1183 if (!strcmp(key, "url") && value &&
1184 check_submodule_url(value) < 0)
1185 data->ret |= report(data->options,
1186 data->oid, OBJ_BLOB,
1187 FSCK_MSG_GITMODULES_URL,
1188 "disallowed submodule url: %s",
1189 value);
1190 if (!strcmp(key, "path") && value &&
1191 looks_like_command_line_option(value))
1192 data->ret |= report(data->options,
1193 data->oid, OBJ_BLOB,
1194 FSCK_MSG_GITMODULES_PATH,
1195 "disallowed submodule path: %s",
1196 value);
1197 if (!strcmp(key, "update") && value &&
1198 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1199 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1200 FSCK_MSG_GITMODULES_UPDATE,
1201 "disallowed submodule update setting: %s",
1202 value);
1203 free(name);
1205 return 0;
1208 static int fsck_blob(const struct object_id *oid, const char *buf,
1209 unsigned long size, struct fsck_options *options)
1211 struct fsck_gitmodules_data data;
1212 struct config_options config_opts = { 0 };
1214 if (!oidset_contains(&gitmodules_found, oid))
1215 return 0;
1216 oidset_insert(&gitmodules_done, oid);
1218 if (object_on_skiplist(options, oid))
1219 return 0;
1221 if (!buf) {
1223 * A missing buffer here is a sign that the caller found the
1224 * blob too gigantic to load into memory. Let's just consider
1225 * that an error.
1227 return report(options, oid, OBJ_BLOB,
1228 FSCK_MSG_GITMODULES_LARGE,
1229 ".gitmodules too large to parse");
1232 data.oid = oid;
1233 data.options = options;
1234 data.ret = 0;
1235 config_opts.error_action = CONFIG_ERROR_SILENT;
1236 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1237 ".gitmodules", buf, size, &data, &config_opts))
1238 data.ret |= report(options, oid, OBJ_BLOB,
1239 FSCK_MSG_GITMODULES_PARSE,
1240 "could not parse gitmodules blob");
1242 return data.ret;
1245 int fsck_object(struct object *obj, void *data, unsigned long size,
1246 struct fsck_options *options)
1248 if (!obj)
1249 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1251 if (obj->type == OBJ_BLOB)
1252 return fsck_blob(&obj->oid, data, size, options);
1253 if (obj->type == OBJ_TREE)
1254 return fsck_tree(&obj->oid, data, size, options);
1255 if (obj->type == OBJ_COMMIT)
1256 return fsck_commit(&obj->oid, data, size, options);
1257 if (obj->type == OBJ_TAG)
1258 return fsck_tag(&obj->oid, data, size, options);
1260 return report(options, &obj->oid, obj->type,
1261 FSCK_MSG_UNKNOWN_TYPE,
1262 "unknown type '%d' (internal fsck error)",
1263 obj->type);
1266 int fsck_error_function(struct fsck_options *o,
1267 const struct object_id *oid,
1268 enum object_type object_type,
1269 int msg_type, const char *message)
1271 if (msg_type == FSCK_WARN) {
1272 warning("object %s: %s", fsck_describe_object(o, oid), message);
1273 return 0;
1275 error("object %s: %s", fsck_describe_object(o, oid), message);
1276 return 1;
1279 void register_found_gitmodules(const struct object_id *oid)
1281 oidset_insert(&gitmodules_found, oid);
1284 int fsck_finish(struct fsck_options *options)
1286 int ret = 0;
1287 struct oidset_iter iter;
1288 const struct object_id *oid;
1290 oidset_iter_init(&gitmodules_found, &iter);
1291 while ((oid = oidset_iter_next(&iter))) {
1292 enum object_type type;
1293 unsigned long size;
1294 char *buf;
1296 if (oidset_contains(&gitmodules_done, oid))
1297 continue;
1299 buf = read_object_file(oid, &type, &size);
1300 if (!buf) {
1301 if (is_promisor_object(oid))
1302 continue;
1303 ret |= report(options,
1304 oid, OBJ_BLOB,
1305 FSCK_MSG_GITMODULES_MISSING,
1306 "unable to read .gitmodules blob");
1307 continue;
1310 if (type == OBJ_BLOB)
1311 ret |= fsck_blob(oid, buf, size, options);
1312 else
1313 ret |= report(options,
1314 oid, type,
1315 FSCK_MSG_GITMODULES_BLOB,
1316 "non-blob found at .gitmodules");
1317 free(buf);
1321 oidset_clear(&gitmodules_found);
1322 oidset_clear(&gitmodules_done);
1323 return ret;
1326 int fsck_config_internal(const char *var, const char *value, void *cb,
1327 struct fsck_options *options)
1329 if (strcmp(var, "fsck.skiplist") == 0) {
1330 const char *path;
1331 struct strbuf sb = STRBUF_INIT;
1333 if (git_config_pathname(&path, var, value))
1334 return 1;
1335 strbuf_addf(&sb, "skiplist=%s", path);
1336 free((char *)path);
1337 fsck_set_msg_types(options, sb.buf);
1338 strbuf_release(&sb);
1339 return 0;
1342 if (skip_prefix(var, "fsck.", &var)) {
1343 fsck_set_msg_type(options, var, value);
1344 return 0;
1347 return git_default_config(var, value, cb);