transfer.unpackLimit: fetch/receive.unpackLimit takes precedence
[alt-git.git] / fsck.c
blob3261ef9ec28928974c6b555c05752511e63e484a
1 #include "git-compat-util.h"
2 #include "alloc.h"
3 #include "date.h"
4 #include "hex.h"
5 #include "object-store.h"
6 #include "repository.h"
7 #include "object.h"
8 #include "attr.h"
9 #include "blob.h"
10 #include "tree.h"
11 #include "tree-walk.h"
12 #include "commit.h"
13 #include "tag.h"
14 #include "fsck.h"
15 #include "refs.h"
16 #include "url.h"
17 #include "utf8.h"
18 #include "decorate.h"
19 #include "oidset.h"
20 #include "packfile.h"
21 #include "submodule-config.h"
22 #include "config.h"
23 #include "credential.h"
24 #include "help.h"
26 #define STR(x) #x
27 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
28 static struct {
29 const char *id_string;
30 const char *downcased;
31 const char *camelcased;
32 enum fsck_msg_type msg_type;
33 } msg_id_info[FSCK_MSG_MAX + 1] = {
34 FOREACH_FSCK_MSG_ID(MSG_ID)
35 { NULL, NULL, NULL, -1 }
37 #undef MSG_ID
38 #undef STR
40 static void prepare_msg_ids(void)
42 int i;
44 if (msg_id_info[0].downcased)
45 return;
47 /* convert id_string to lower case, without underscores. */
48 for (i = 0; i < FSCK_MSG_MAX; i++) {
49 const char *p = msg_id_info[i].id_string;
50 int len = strlen(p);
51 char *q = xmalloc(len);
53 msg_id_info[i].downcased = q;
54 while (*p)
55 if (*p == '_')
56 p++;
57 else
58 *(q)++ = tolower(*(p)++);
59 *q = '\0';
61 p = msg_id_info[i].id_string;
62 q = xmalloc(len);
63 msg_id_info[i].camelcased = q;
64 while (*p) {
65 if (*p == '_') {
66 p++;
67 if (*p)
68 *q++ = *p++;
69 } else {
70 *q++ = tolower(*p++);
73 *q = '\0';
77 static int parse_msg_id(const char *text)
79 int i;
81 prepare_msg_ids();
83 for (i = 0; i < FSCK_MSG_MAX; i++)
84 if (!strcmp(text, msg_id_info[i].downcased))
85 return i;
87 return -1;
90 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
92 int i;
94 prepare_msg_ids();
96 for (i = 0; i < FSCK_MSG_MAX; i++)
97 list_config_item(list, prefix, msg_id_info[i].camelcased);
100 static enum fsck_msg_type fsck_msg_type(enum fsck_msg_id msg_id,
101 struct fsck_options *options)
103 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
105 if (!options->msg_type) {
106 enum fsck_msg_type msg_type = msg_id_info[msg_id].msg_type;
108 if (options->strict && msg_type == FSCK_WARN)
109 msg_type = FSCK_ERROR;
110 return msg_type;
113 return options->msg_type[msg_id];
116 static enum fsck_msg_type parse_msg_type(const char *str)
118 if (!strcmp(str, "error"))
119 return FSCK_ERROR;
120 else if (!strcmp(str, "warn"))
121 return FSCK_WARN;
122 else if (!strcmp(str, "ignore"))
123 return FSCK_IGNORE;
124 else
125 die("Unknown fsck message type: '%s'", str);
128 int is_valid_msg_type(const char *msg_id, const char *msg_type)
130 if (parse_msg_id(msg_id) < 0)
131 return 0;
132 parse_msg_type(msg_type);
133 return 1;
136 void fsck_set_msg_type_from_ids(struct fsck_options *options,
137 enum fsck_msg_id msg_id,
138 enum fsck_msg_type msg_type)
140 if (!options->msg_type) {
141 int i;
142 enum fsck_msg_type *severity;
143 ALLOC_ARRAY(severity, FSCK_MSG_MAX);
144 for (i = 0; i < FSCK_MSG_MAX; i++)
145 severity[i] = fsck_msg_type(i, options);
146 options->msg_type = severity;
149 options->msg_type[msg_id] = msg_type;
152 void fsck_set_msg_type(struct fsck_options *options,
153 const char *msg_id_str, const char *msg_type_str)
155 int msg_id = parse_msg_id(msg_id_str);
156 enum fsck_msg_type msg_type = parse_msg_type(msg_type_str);
158 if (msg_id < 0)
159 die("Unhandled message id: %s", msg_id_str);
161 if (msg_type != FSCK_ERROR && msg_id_info[msg_id].msg_type == FSCK_FATAL)
162 die("Cannot demote %s to %s", msg_id_str, msg_type_str);
164 fsck_set_msg_type_from_ids(options, msg_id, msg_type);
167 void fsck_set_msg_types(struct fsck_options *options, const char *values)
169 char *buf = xstrdup(values), *to_free = buf;
170 int done = 0;
172 while (!done) {
173 int len = strcspn(buf, " ,|"), equal;
175 done = !buf[len];
176 if (!len) {
177 buf++;
178 continue;
180 buf[len] = '\0';
182 for (equal = 0;
183 equal < len && buf[equal] != '=' && buf[equal] != ':';
184 equal++)
185 buf[equal] = tolower(buf[equal]);
186 buf[equal] = '\0';
188 if (!strcmp(buf, "skiplist")) {
189 if (equal == len)
190 die("skiplist requires a path");
191 oidset_parse_file(&options->skiplist, buf + equal + 1);
192 buf += len + 1;
193 continue;
196 if (equal == len)
197 die("Missing '=': '%s'", buf);
199 fsck_set_msg_type(options, buf, buf + equal + 1);
200 buf += len + 1;
202 free(to_free);
205 static int object_on_skiplist(struct fsck_options *opts,
206 const struct object_id *oid)
208 return opts && oid && oidset_contains(&opts->skiplist, oid);
211 __attribute__((format (printf, 5, 6)))
212 static int report(struct fsck_options *options,
213 const struct object_id *oid, enum object_type object_type,
214 enum fsck_msg_id msg_id, const char *fmt, ...)
216 va_list ap;
217 struct strbuf sb = STRBUF_INIT;
218 enum fsck_msg_type msg_type = fsck_msg_type(msg_id, options);
219 int result;
221 if (msg_type == FSCK_IGNORE)
222 return 0;
224 if (object_on_skiplist(options, oid))
225 return 0;
227 if (msg_type == FSCK_FATAL)
228 msg_type = FSCK_ERROR;
229 else if (msg_type == FSCK_INFO)
230 msg_type = FSCK_WARN;
232 prepare_msg_ids();
233 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
235 va_start(ap, fmt);
236 strbuf_vaddf(&sb, fmt, ap);
237 result = options->error_func(options, oid, object_type,
238 msg_type, msg_id, sb.buf);
239 strbuf_release(&sb);
240 va_end(ap);
242 return result;
245 void fsck_enable_object_names(struct fsck_options *options)
247 if (!options->object_names)
248 options->object_names = kh_init_oid_map();
251 const char *fsck_get_object_name(struct fsck_options *options,
252 const struct object_id *oid)
254 khiter_t pos;
255 if (!options->object_names)
256 return NULL;
257 pos = kh_get_oid_map(options->object_names, *oid);
258 if (pos >= kh_end(options->object_names))
259 return NULL;
260 return kh_value(options->object_names, pos);
263 void fsck_put_object_name(struct fsck_options *options,
264 const struct object_id *oid,
265 const char *fmt, ...)
267 va_list ap;
268 struct strbuf buf = STRBUF_INIT;
269 khiter_t pos;
270 int hashret;
272 if (!options->object_names)
273 return;
275 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
276 if (!hashret)
277 return;
278 va_start(ap, fmt);
279 strbuf_vaddf(&buf, fmt, ap);
280 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
281 va_end(ap);
284 const char *fsck_describe_object(struct fsck_options *options,
285 const struct object_id *oid)
287 static struct strbuf bufs[] = {
288 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
290 static int b = 0;
291 struct strbuf *buf;
292 const char *name = fsck_get_object_name(options, oid);
294 buf = bufs + b;
295 b = (b + 1) % ARRAY_SIZE(bufs);
296 strbuf_reset(buf);
297 strbuf_addstr(buf, oid_to_hex(oid));
298 if (name)
299 strbuf_addf(buf, " (%s)", name);
301 return buf->buf;
304 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
306 struct tree_desc desc;
307 struct name_entry entry;
308 int res = 0;
309 const char *name;
311 if (parse_tree(tree))
312 return -1;
314 name = fsck_get_object_name(options, &tree->object.oid);
315 if (init_tree_desc_gently(&desc, tree->buffer, tree->size, 0))
316 return -1;
317 while (tree_entry_gently(&desc, &entry)) {
318 struct object *obj;
319 int result;
321 if (S_ISGITLINK(entry.mode))
322 continue;
324 if (S_ISDIR(entry.mode)) {
325 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
326 if (name && obj)
327 fsck_put_object_name(options, &entry.oid, "%s%s/",
328 name, entry.path);
329 result = options->walk(obj, OBJ_TREE, data, options);
331 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
332 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
333 if (name && obj)
334 fsck_put_object_name(options, &entry.oid, "%s%s",
335 name, entry.path);
336 result = options->walk(obj, OBJ_BLOB, data, options);
338 else {
339 result = error("in tree %s: entry %s has bad mode %.6o",
340 fsck_describe_object(options, &tree->object.oid),
341 entry.path, entry.mode);
343 if (result < 0)
344 return result;
345 if (!res)
346 res = result;
348 return res;
351 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
353 int counter = 0, generation = 0, name_prefix_len = 0;
354 struct commit_list *parents;
355 int res;
356 int result;
357 const char *name;
359 if (repo_parse_commit(the_repository, commit))
360 return -1;
362 name = fsck_get_object_name(options, &commit->object.oid);
363 if (name)
364 fsck_put_object_name(options, get_commit_tree_oid(commit),
365 "%s:", name);
367 result = options->walk((struct object *) repo_get_commit_tree(the_repository, commit),
368 OBJ_TREE, data, options);
369 if (result < 0)
370 return result;
371 res = result;
373 parents = commit->parents;
374 if (name && parents) {
375 int len = strlen(name), power;
377 if (len && name[len - 1] == '^') {
378 generation = 1;
379 name_prefix_len = len - 1;
381 else { /* parse ~<generation> suffix */
382 for (generation = 0, power = 1;
383 len && isdigit(name[len - 1]);
384 power *= 10)
385 generation += power * (name[--len] - '0');
386 if (power > 1 && len && name[len - 1] == '~')
387 name_prefix_len = len - 1;
388 else {
389 /* Maybe a non-first parent, e.g. HEAD^2 */
390 generation = 0;
391 name_prefix_len = len;
396 while (parents) {
397 if (name) {
398 struct object_id *oid = &parents->item->object.oid;
400 if (counter++)
401 fsck_put_object_name(options, oid, "%s^%d",
402 name, counter);
403 else if (generation > 0)
404 fsck_put_object_name(options, oid, "%.*s~%d",
405 name_prefix_len, name,
406 generation + 1);
407 else
408 fsck_put_object_name(options, oid, "%s^", name);
410 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
411 if (result < 0)
412 return result;
413 if (!res)
414 res = result;
415 parents = parents->next;
417 return res;
420 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
422 const char *name = fsck_get_object_name(options, &tag->object.oid);
424 if (parse_tag(tag))
425 return -1;
426 if (name)
427 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
428 return options->walk(tag->tagged, OBJ_ANY, data, options);
431 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
433 if (!obj)
434 return -1;
436 if (obj->type == OBJ_NONE)
437 parse_object(the_repository, &obj->oid);
439 switch (obj->type) {
440 case OBJ_BLOB:
441 return 0;
442 case OBJ_TREE:
443 return fsck_walk_tree((struct tree *)obj, data, options);
444 case OBJ_COMMIT:
445 return fsck_walk_commit((struct commit *)obj, data, options);
446 case OBJ_TAG:
447 return fsck_walk_tag((struct tag *)obj, data, options);
448 default:
449 error("Unknown object type for %s",
450 fsck_describe_object(options, &obj->oid));
451 return -1;
455 struct name_stack {
456 const char **names;
457 size_t nr, alloc;
460 static void name_stack_push(struct name_stack *stack, const char *name)
462 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
463 stack->names[stack->nr++] = name;
466 static const char *name_stack_pop(struct name_stack *stack)
468 return stack->nr ? stack->names[--stack->nr] : NULL;
471 static void name_stack_clear(struct name_stack *stack)
473 FREE_AND_NULL(stack->names);
474 stack->nr = stack->alloc = 0;
478 * The entries in a tree are ordered in the _path_ order,
479 * which means that a directory entry is ordered by adding
480 * a slash to the end of it.
482 * So a directory called "a" is ordered _after_ a file
483 * called "a.c", because "a/" sorts after "a.c".
485 #define TREE_UNORDERED (-1)
486 #define TREE_HAS_DUPS (-2)
488 static int is_less_than_slash(unsigned char c)
490 return '\0' < c && c < '/';
493 static int verify_ordered(unsigned mode1, const char *name1,
494 unsigned mode2, const char *name2,
495 struct name_stack *candidates)
497 int len1 = strlen(name1);
498 int len2 = strlen(name2);
499 int len = len1 < len2 ? len1 : len2;
500 unsigned char c1, c2;
501 int cmp;
503 cmp = memcmp(name1, name2, len);
504 if (cmp < 0)
505 return 0;
506 if (cmp > 0)
507 return TREE_UNORDERED;
510 * Ok, the first <len> characters are the same.
511 * Now we need to order the next one, but turn
512 * a '\0' into a '/' for a directory entry.
514 c1 = name1[len];
515 c2 = name2[len];
516 if (!c1 && !c2)
518 * git-write-tree used to write out a nonsense tree that has
519 * entries with the same name, one blob and one tree. Make
520 * sure we do not have duplicate entries.
522 return TREE_HAS_DUPS;
523 if (!c1 && S_ISDIR(mode1))
524 c1 = '/';
525 if (!c2 && S_ISDIR(mode2))
526 c2 = '/';
529 * There can be non-consecutive duplicates due to the implicitly
530 * added slash, e.g.:
532 * foo
533 * foo.bar
534 * foo.bar.baz
535 * foo.bar/
536 * foo/
538 * Record non-directory candidates (like "foo" and "foo.bar" in
539 * the example) on a stack and check directory candidates (like
540 * foo/" and "foo.bar/") against that stack.
542 if (!c1 && is_less_than_slash(c2)) {
543 name_stack_push(candidates, name1);
544 } else if (c2 == '/' && is_less_than_slash(c1)) {
545 for (;;) {
546 const char *p;
547 const char *f_name = name_stack_pop(candidates);
549 if (!f_name)
550 break;
551 if (!skip_prefix(name2, f_name, &p))
552 continue;
553 if (!*p)
554 return TREE_HAS_DUPS;
555 if (is_less_than_slash(*p)) {
556 name_stack_push(candidates, f_name);
557 break;
562 return c1 < c2 ? 0 : TREE_UNORDERED;
565 static int fsck_tree(const struct object_id *tree_oid,
566 const char *buffer, unsigned long size,
567 struct fsck_options *options)
569 int retval = 0;
570 int has_null_sha1 = 0;
571 int has_full_path = 0;
572 int has_empty_name = 0;
573 int has_dot = 0;
574 int has_dotdot = 0;
575 int has_dotgit = 0;
576 int has_zero_pad = 0;
577 int has_bad_modes = 0;
578 int has_dup_entries = 0;
579 int not_properly_sorted = 0;
580 struct tree_desc desc;
581 unsigned o_mode;
582 const char *o_name;
583 struct name_stack df_dup_candidates = { NULL };
585 if (init_tree_desc_gently(&desc, buffer, size, TREE_DESC_RAW_MODES)) {
586 retval += report(options, tree_oid, OBJ_TREE,
587 FSCK_MSG_BAD_TREE,
588 "cannot be parsed as a tree");
589 return retval;
592 o_mode = 0;
593 o_name = NULL;
595 while (desc.size) {
596 unsigned short mode;
597 const char *name, *backslash;
598 const struct object_id *entry_oid;
600 entry_oid = tree_entry_extract(&desc, &name, &mode);
602 has_null_sha1 |= is_null_oid(entry_oid);
603 has_full_path |= !!strchr(name, '/');
604 has_empty_name |= !*name;
605 has_dot |= !strcmp(name, ".");
606 has_dotdot |= !strcmp(name, "..");
607 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
608 has_zero_pad |= *(char *)desc.buffer == '0';
610 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
611 if (!S_ISLNK(mode))
612 oidset_insert(&options->gitmodules_found,
613 entry_oid);
614 else
615 retval += report(options,
616 tree_oid, OBJ_TREE,
617 FSCK_MSG_GITMODULES_SYMLINK,
618 ".gitmodules is a symbolic link");
621 if (is_hfs_dotgitattributes(name) || is_ntfs_dotgitattributes(name)) {
622 if (!S_ISLNK(mode))
623 oidset_insert(&options->gitattributes_found,
624 entry_oid);
625 else
626 retval += report(options, tree_oid, OBJ_TREE,
627 FSCK_MSG_GITATTRIBUTES_SYMLINK,
628 ".gitattributes is a symlink");
631 if (S_ISLNK(mode)) {
632 if (is_hfs_dotgitignore(name) ||
633 is_ntfs_dotgitignore(name))
634 retval += report(options, tree_oid, OBJ_TREE,
635 FSCK_MSG_GITIGNORE_SYMLINK,
636 ".gitignore is a symlink");
637 if (is_hfs_dotmailmap(name) ||
638 is_ntfs_dotmailmap(name))
639 retval += report(options, tree_oid, OBJ_TREE,
640 FSCK_MSG_MAILMAP_SYMLINK,
641 ".mailmap is a symlink");
644 if ((backslash = strchr(name, '\\'))) {
645 while (backslash) {
646 backslash++;
647 has_dotgit |= is_ntfs_dotgit(backslash);
648 if (is_ntfs_dotgitmodules(backslash)) {
649 if (!S_ISLNK(mode))
650 oidset_insert(&options->gitmodules_found,
651 entry_oid);
652 else
653 retval += report(options, tree_oid, OBJ_TREE,
654 FSCK_MSG_GITMODULES_SYMLINK,
655 ".gitmodules is a symbolic link");
657 backslash = strchr(backslash, '\\');
661 if (update_tree_entry_gently(&desc)) {
662 retval += report(options, tree_oid, OBJ_TREE,
663 FSCK_MSG_BAD_TREE,
664 "cannot be parsed as a tree");
665 break;
668 switch (mode) {
670 * Standard modes..
672 case S_IFREG | 0755:
673 case S_IFREG | 0644:
674 case S_IFLNK:
675 case S_IFDIR:
676 case S_IFGITLINK:
677 break;
679 * This is nonstandard, but we had a few of these
680 * early on when we honored the full set of mode
681 * bits..
683 case S_IFREG | 0664:
684 if (!options->strict)
685 break;
686 /* fallthrough */
687 default:
688 has_bad_modes = 1;
691 if (o_name) {
692 switch (verify_ordered(o_mode, o_name, mode, name,
693 &df_dup_candidates)) {
694 case TREE_UNORDERED:
695 not_properly_sorted = 1;
696 break;
697 case TREE_HAS_DUPS:
698 has_dup_entries = 1;
699 break;
700 default:
701 break;
705 o_mode = mode;
706 o_name = name;
709 name_stack_clear(&df_dup_candidates);
711 if (has_null_sha1)
712 retval += report(options, tree_oid, OBJ_TREE,
713 FSCK_MSG_NULL_SHA1,
714 "contains entries pointing to null sha1");
715 if (has_full_path)
716 retval += report(options, tree_oid, OBJ_TREE,
717 FSCK_MSG_FULL_PATHNAME,
718 "contains full pathnames");
719 if (has_empty_name)
720 retval += report(options, tree_oid, OBJ_TREE,
721 FSCK_MSG_EMPTY_NAME,
722 "contains empty pathname");
723 if (has_dot)
724 retval += report(options, tree_oid, OBJ_TREE,
725 FSCK_MSG_HAS_DOT,
726 "contains '.'");
727 if (has_dotdot)
728 retval += report(options, tree_oid, OBJ_TREE,
729 FSCK_MSG_HAS_DOTDOT,
730 "contains '..'");
731 if (has_dotgit)
732 retval += report(options, tree_oid, OBJ_TREE,
733 FSCK_MSG_HAS_DOTGIT,
734 "contains '.git'");
735 if (has_zero_pad)
736 retval += report(options, tree_oid, OBJ_TREE,
737 FSCK_MSG_ZERO_PADDED_FILEMODE,
738 "contains zero-padded file modes");
739 if (has_bad_modes)
740 retval += report(options, tree_oid, OBJ_TREE,
741 FSCK_MSG_BAD_FILEMODE,
742 "contains bad file modes");
743 if (has_dup_entries)
744 retval += report(options, tree_oid, OBJ_TREE,
745 FSCK_MSG_DUPLICATE_ENTRIES,
746 "contains duplicate file entries");
747 if (not_properly_sorted)
748 retval += report(options, tree_oid, OBJ_TREE,
749 FSCK_MSG_TREE_NOT_SORTED,
750 "not properly sorted");
751 return retval;
755 * Confirm that the headers of a commit or tag object end in a reasonable way,
756 * either with the usual "\n\n" separator, or at least with a trailing newline
757 * on the final header line.
759 * This property is important for the memory safety of our callers. It allows
760 * them to scan the buffer linewise without constantly checking the remaining
761 * size as long as:
763 * - they check that there are bytes left in the buffer at the start of any
764 * line (i.e., that the last newline they saw was not the final one we
765 * found here)
767 * - any intra-line scanning they do will stop at a newline, which will worst
768 * case hit the newline we found here as the end-of-header. This makes it
769 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
771 static int verify_headers(const void *data, unsigned long size,
772 const struct object_id *oid, enum object_type type,
773 struct fsck_options *options)
775 const char *buffer = (const char *)data;
776 unsigned long i;
778 for (i = 0; i < size; i++) {
779 switch (buffer[i]) {
780 case '\0':
781 return report(options, oid, type,
782 FSCK_MSG_NUL_IN_HEADER,
783 "unterminated header: NUL at offset %ld", i);
784 case '\n':
785 if (i + 1 < size && buffer[i + 1] == '\n')
786 return 0;
791 * We did not find double-LF that separates the header
792 * and the body. Not having a body is not a crime but
793 * we do want to see the terminating LF for the last header
794 * line.
796 if (size && buffer[size - 1] == '\n')
797 return 0;
799 return report(options, oid, type,
800 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
803 static int fsck_ident(const char **ident,
804 const struct object_id *oid, enum object_type type,
805 struct fsck_options *options)
807 const char *p = *ident;
808 char *end;
810 *ident = strchrnul(*ident, '\n');
811 if (**ident == '\n')
812 (*ident)++;
814 if (*p == '<')
815 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
816 p += strcspn(p, "<>\n");
817 if (*p == '>')
818 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
819 if (*p != '<')
820 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
821 if (p[-1] != ' ')
822 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
823 p++;
824 p += strcspn(p, "<>\n");
825 if (*p != '>')
826 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
827 p++;
828 if (*p != ' ')
829 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
830 p++;
832 * Our timestamp parser is based on the C strto*() functions, which
833 * will happily eat whitespace, including the newline that is supposed
834 * to prevent us walking past the end of the buffer. So do our own
835 * scan, skipping linear whitespace but not newlines, and then
836 * confirming we found a digit. We _could_ be even more strict here,
837 * as we really expect only a single space, but since we have
838 * traditionally allowed extra whitespace, we'll continue to do so.
840 while (*p == ' ' || *p == '\t')
841 p++;
842 if (!isdigit(*p))
843 return report(options, oid, type, FSCK_MSG_BAD_DATE,
844 "invalid author/committer line - bad date");
845 if (*p == '0' && p[1] != ' ')
846 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
847 if (date_overflows(parse_timestamp(p, &end, 10)))
848 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
849 if ((end == p || *end != ' '))
850 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
851 p = end + 1;
852 if ((*p != '+' && *p != '-') ||
853 !isdigit(p[1]) ||
854 !isdigit(p[2]) ||
855 !isdigit(p[3]) ||
856 !isdigit(p[4]) ||
857 (p[5] != '\n'))
858 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
859 p += 6;
860 return 0;
863 static int fsck_commit(const struct object_id *oid,
864 const char *buffer, unsigned long size,
865 struct fsck_options *options)
867 struct object_id tree_oid, parent_oid;
868 unsigned author_count;
869 int err;
870 const char *buffer_begin = buffer;
871 const char *buffer_end = buffer + size;
872 const char *p;
875 * We _must_ stop parsing immediately if this reports failure, as the
876 * memory safety of the rest of the function depends on it. See the
877 * comment above the definition of verify_headers() for more details.
879 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
880 return -1;
882 if (buffer >= buffer_end || !skip_prefix(buffer, "tree ", &buffer))
883 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
884 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
885 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
886 if (err)
887 return err;
889 buffer = p + 1;
890 while (buffer < buffer_end && skip_prefix(buffer, "parent ", &buffer)) {
891 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
892 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
893 if (err)
894 return err;
896 buffer = p + 1;
898 author_count = 0;
899 while (buffer < buffer_end && skip_prefix(buffer, "author ", &buffer)) {
900 author_count++;
901 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
902 if (err)
903 return err;
905 if (author_count < 1)
906 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
907 else if (author_count > 1)
908 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
909 if (err)
910 return err;
911 if (buffer >= buffer_end || !skip_prefix(buffer, "committer ", &buffer))
912 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
913 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
914 if (err)
915 return err;
916 if (memchr(buffer_begin, '\0', size)) {
917 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
918 "NUL byte in the commit object body");
919 if (err)
920 return err;
922 return 0;
925 static int fsck_tag(const struct object_id *oid, const char *buffer,
926 unsigned long size, struct fsck_options *options)
928 struct object_id tagged_oid;
929 int tagged_type;
930 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
931 &tagged_type);
934 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
935 unsigned long size, struct fsck_options *options,
936 struct object_id *tagged_oid,
937 int *tagged_type)
939 int ret = 0;
940 char *eol;
941 struct strbuf sb = STRBUF_INIT;
942 const char *buffer_end = buffer + size;
943 const char *p;
946 * We _must_ stop parsing immediately if this reports failure, as the
947 * memory safety of the rest of the function depends on it. See the
948 * comment above the definition of verify_headers() for more details.
950 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
951 if (ret)
952 goto done;
954 if (buffer >= buffer_end || !skip_prefix(buffer, "object ", &buffer)) {
955 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
956 goto done;
958 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
959 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
960 if (ret)
961 goto done;
963 buffer = p + 1;
965 if (buffer >= buffer_end || !skip_prefix(buffer, "type ", &buffer)) {
966 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
967 goto done;
969 eol = memchr(buffer, '\n', buffer_end - buffer);
970 if (!eol) {
971 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
972 goto done;
974 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
975 if (*tagged_type < 0)
976 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
977 if (ret)
978 goto done;
979 buffer = eol + 1;
981 if (buffer >= buffer_end || !skip_prefix(buffer, "tag ", &buffer)) {
982 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
983 goto done;
985 eol = memchr(buffer, '\n', buffer_end - buffer);
986 if (!eol) {
987 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
988 goto done;
990 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
991 if (check_refname_format(sb.buf, 0)) {
992 ret = report(options, oid, OBJ_TAG,
993 FSCK_MSG_BAD_TAG_NAME,
994 "invalid 'tag' name: %.*s",
995 (int)(eol - buffer), buffer);
996 if (ret)
997 goto done;
999 buffer = eol + 1;
1001 if (buffer >= buffer_end || !skip_prefix(buffer, "tagger ", &buffer)) {
1002 /* early tags do not contain 'tagger' lines; warn only */
1003 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
1004 if (ret)
1005 goto done;
1007 else
1008 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
1010 if (buffer < buffer_end && !starts_with(buffer, "\n")) {
1012 * The verify_headers() check will allow
1013 * e.g. "[...]tagger <tagger>\nsome
1014 * garbage\n\nmessage" to pass, thinking "some
1015 * garbage" could be a custom header. E.g. "mktag"
1016 * doesn't want any unknown headers.
1018 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1019 if (ret)
1020 goto done;
1023 done:
1024 strbuf_release(&sb);
1025 return ret;
1028 static int starts_with_dot_slash(const char *const path)
1030 return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_SLASH |
1031 PATH_MATCH_XPLATFORM);
1034 static int starts_with_dot_dot_slash(const char *const path)
1036 return path_match_flags(path, PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH |
1037 PATH_MATCH_XPLATFORM);
1040 static int submodule_url_is_relative(const char *url)
1042 return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
1046 * Count directory components that a relative submodule URL should chop
1047 * from the remote_url it is to be resolved against.
1049 * In other words, this counts "../" components at the start of a
1050 * submodule URL.
1052 * Returns the number of directory components to chop and writes a
1053 * pointer to the next character of url after all leading "./" and
1054 * "../" components to out.
1056 static int count_leading_dotdots(const char *url, const char **out)
1058 int result = 0;
1059 while (1) {
1060 if (starts_with_dot_dot_slash(url)) {
1061 result++;
1062 url += strlen("../");
1063 continue;
1065 if (starts_with_dot_slash(url)) {
1066 url += strlen("./");
1067 continue;
1069 *out = url;
1070 return result;
1074 * Check whether a transport is implemented by git-remote-curl.
1076 * If it is, returns 1 and writes the URL that would be passed to
1077 * git-remote-curl to the "out" parameter.
1079 * Otherwise, returns 0 and leaves "out" untouched.
1081 * Examples:
1082 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1083 * https://example.com/repo.git -> 1, https://example.com/repo.git
1084 * git://example.com/repo.git -> 0
1086 * This is for use in checking for previously exploitable bugs that
1087 * required a submodule URL to be passed to git-remote-curl.
1089 static int url_to_curl_url(const char *url, const char **out)
1092 * We don't need to check for case-aliases, "http.exe", and so
1093 * on because in the default configuration, is_transport_allowed
1094 * prevents URLs with those schemes from being cloned
1095 * automatically.
1097 if (skip_prefix(url, "http::", out) ||
1098 skip_prefix(url, "https::", out) ||
1099 skip_prefix(url, "ftp::", out) ||
1100 skip_prefix(url, "ftps::", out))
1101 return 1;
1102 if (starts_with(url, "http://") ||
1103 starts_with(url, "https://") ||
1104 starts_with(url, "ftp://") ||
1105 starts_with(url, "ftps://")) {
1106 *out = url;
1107 return 1;
1109 return 0;
1112 static int check_submodule_url(const char *url)
1114 const char *curl_url;
1116 if (looks_like_command_line_option(url))
1117 return -1;
1119 if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
1120 char *decoded;
1121 const char *next;
1122 int has_nl;
1125 * This could be appended to an http URL and url-decoded;
1126 * check for malicious characters.
1128 decoded = url_decode(url);
1129 has_nl = !!strchr(decoded, '\n');
1131 free(decoded);
1132 if (has_nl)
1133 return -1;
1136 * URLs which escape their root via "../" can overwrite
1137 * the host field and previous components, resolving to
1138 * URLs like https::example.com/submodule.git and
1139 * https:///example.com/submodule.git that were
1140 * susceptible to CVE-2020-11008.
1142 if (count_leading_dotdots(url, &next) > 0 &&
1143 (*next == ':' || *next == '/'))
1144 return -1;
1147 else if (url_to_curl_url(url, &curl_url)) {
1148 struct credential c = CREDENTIAL_INIT;
1149 int ret = 0;
1150 if (credential_from_url_gently(&c, curl_url, 1) ||
1151 !*c.host)
1152 ret = -1;
1153 credential_clear(&c);
1154 return ret;
1157 return 0;
1160 struct fsck_gitmodules_data {
1161 const struct object_id *oid;
1162 struct fsck_options *options;
1163 int ret;
1166 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1168 struct fsck_gitmodules_data *data = vdata;
1169 const char *subsection, *key;
1170 size_t subsection_len;
1171 char *name;
1173 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1174 !subsection)
1175 return 0;
1177 name = xmemdupz(subsection, subsection_len);
1178 if (check_submodule_name(name) < 0)
1179 data->ret |= report(data->options,
1180 data->oid, OBJ_BLOB,
1181 FSCK_MSG_GITMODULES_NAME,
1182 "disallowed submodule name: %s",
1183 name);
1184 if (!strcmp(key, "url") && value &&
1185 check_submodule_url(value) < 0)
1186 data->ret |= report(data->options,
1187 data->oid, OBJ_BLOB,
1188 FSCK_MSG_GITMODULES_URL,
1189 "disallowed submodule url: %s",
1190 value);
1191 if (!strcmp(key, "path") && value &&
1192 looks_like_command_line_option(value))
1193 data->ret |= report(data->options,
1194 data->oid, OBJ_BLOB,
1195 FSCK_MSG_GITMODULES_PATH,
1196 "disallowed submodule path: %s",
1197 value);
1198 if (!strcmp(key, "update") && value &&
1199 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1200 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1201 FSCK_MSG_GITMODULES_UPDATE,
1202 "disallowed submodule update setting: %s",
1203 value);
1204 free(name);
1206 return 0;
1209 static int fsck_blob(const struct object_id *oid, const char *buf,
1210 unsigned long size, struct fsck_options *options)
1212 int ret = 0;
1214 if (object_on_skiplist(options, oid))
1215 return 0;
1217 if (oidset_contains(&options->gitmodules_found, oid)) {
1218 struct config_options config_opts = { 0 };
1219 struct fsck_gitmodules_data data;
1221 oidset_insert(&options->gitmodules_done, oid);
1223 if (!buf) {
1225 * A missing buffer here is a sign that the caller found the
1226 * blob too gigantic to load into memory. Let's just consider
1227 * that an error.
1229 return report(options, oid, OBJ_BLOB,
1230 FSCK_MSG_GITMODULES_LARGE,
1231 ".gitmodules too large to parse");
1234 data.oid = oid;
1235 data.options = options;
1236 data.ret = 0;
1237 config_opts.error_action = CONFIG_ERROR_SILENT;
1238 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1239 ".gitmodules", buf, size, &data, &config_opts))
1240 data.ret |= report(options, oid, OBJ_BLOB,
1241 FSCK_MSG_GITMODULES_PARSE,
1242 "could not parse gitmodules blob");
1243 ret |= data.ret;
1246 if (oidset_contains(&options->gitattributes_found, oid)) {
1247 const char *ptr;
1249 oidset_insert(&options->gitattributes_done, oid);
1251 if (!buf || size > ATTR_MAX_FILE_SIZE) {
1253 * A missing buffer here is a sign that the caller found the
1254 * blob too gigantic to load into memory. Let's just consider
1255 * that an error.
1257 return report(options, oid, OBJ_BLOB,
1258 FSCK_MSG_GITATTRIBUTES_LARGE,
1259 ".gitattributes too large to parse");
1262 for (ptr = buf; *ptr; ) {
1263 const char *eol = strchrnul(ptr, '\n');
1264 if (eol - ptr >= ATTR_MAX_LINE_LENGTH) {
1265 ret |= report(options, oid, OBJ_BLOB,
1266 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH,
1267 ".gitattributes has too long lines to parse");
1268 break;
1271 ptr = *eol ? eol + 1 : eol;
1275 return ret;
1278 int fsck_object(struct object *obj, void *data, unsigned long size,
1279 struct fsck_options *options)
1281 if (!obj)
1282 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1284 return fsck_buffer(&obj->oid, obj->type, data, size, options);
1287 int fsck_buffer(const struct object_id *oid, enum object_type type,
1288 void *data, unsigned long size,
1289 struct fsck_options *options)
1291 if (type == OBJ_BLOB)
1292 return fsck_blob(oid, data, size, options);
1293 if (type == OBJ_TREE)
1294 return fsck_tree(oid, data, size, options);
1295 if (type == OBJ_COMMIT)
1296 return fsck_commit(oid, data, size, options);
1297 if (type == OBJ_TAG)
1298 return fsck_tag(oid, data, size, options);
1300 return report(options, oid, type,
1301 FSCK_MSG_UNKNOWN_TYPE,
1302 "unknown type '%d' (internal fsck error)",
1303 type);
1306 int fsck_error_function(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_type == FSCK_WARN) {
1314 warning("object %s: %s", fsck_describe_object(o, oid), message);
1315 return 0;
1317 error("object %s: %s", fsck_describe_object(o, oid), message);
1318 return 1;
1321 static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
1322 enum fsck_msg_id msg_missing, enum fsck_msg_id msg_type,
1323 struct fsck_options *options, const char *blob_type)
1325 int ret = 0;
1326 struct oidset_iter iter;
1327 const struct object_id *oid;
1329 oidset_iter_init(blobs_found, &iter);
1330 while ((oid = oidset_iter_next(&iter))) {
1331 enum object_type type;
1332 unsigned long size;
1333 char *buf;
1335 if (oidset_contains(blobs_done, oid))
1336 continue;
1338 buf = repo_read_object_file(the_repository, oid, &type, &size);
1339 if (!buf) {
1340 if (is_promisor_object(oid))
1341 continue;
1342 ret |= report(options,
1343 oid, OBJ_BLOB, msg_missing,
1344 "unable to read %s blob", blob_type);
1345 continue;
1348 if (type == OBJ_BLOB)
1349 ret |= fsck_blob(oid, buf, size, options);
1350 else
1351 ret |= report(options, oid, type, msg_type,
1352 "non-blob found at %s", blob_type);
1353 free(buf);
1356 oidset_clear(blobs_found);
1357 oidset_clear(blobs_done);
1359 return ret;
1362 int fsck_finish(struct fsck_options *options)
1364 int ret = 0;
1366 ret |= fsck_blobs(&options->gitmodules_found, &options->gitmodules_done,
1367 FSCK_MSG_GITMODULES_MISSING, FSCK_MSG_GITMODULES_BLOB,
1368 options, ".gitmodules");
1369 ret |= fsck_blobs(&options->gitattributes_found, &options->gitattributes_done,
1370 FSCK_MSG_GITATTRIBUTES_MISSING, FSCK_MSG_GITATTRIBUTES_BLOB,
1371 options, ".gitattributes");
1373 return ret;
1376 int git_fsck_config(const char *var, const char *value, void *cb)
1378 struct fsck_options *options = cb;
1379 if (strcmp(var, "fsck.skiplist") == 0) {
1380 const char *path;
1381 struct strbuf sb = STRBUF_INIT;
1383 if (git_config_pathname(&path, var, value))
1384 return 1;
1385 strbuf_addf(&sb, "skiplist=%s", path);
1386 free((char *)path);
1387 fsck_set_msg_types(options, sb.buf);
1388 strbuf_release(&sb);
1389 return 0;
1392 if (skip_prefix(var, "fsck.", &var)) {
1393 fsck_set_msg_type(options, var, value);
1394 return 0;
1397 return git_default_config(var, value, cb);
1401 * Custom error callbacks that are used in more than one place.
1404 int fsck_error_cb_print_missing_gitmodules(struct fsck_options *o,
1405 const struct object_id *oid,
1406 enum object_type object_type,
1407 enum fsck_msg_type msg_type,
1408 enum fsck_msg_id msg_id,
1409 const char *message)
1411 if (msg_id == FSCK_MSG_GITMODULES_MISSING) {
1412 puts(oid_to_hex(oid));
1413 return 0;
1415 return fsck_error_function(o, oid, object_type, msg_type, msg_id, message);