fsck.c: refactor fsck_msg_type() to limit scope of "int msg_type"
[alt-git.git] / fsck.c
blob8614ee2c2a0713501a86e28276d9539f126d502c
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 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
172 if (!options->msg_type) {
173 int msg_type = msg_id_info[msg_id].msg_type;
175 if (options->strict && msg_type == FSCK_WARN)
176 msg_type = FSCK_ERROR;
177 return msg_type;
180 return options->msg_type[msg_id];
183 static int parse_msg_type(const char *str)
185 if (!strcmp(str, "error"))
186 return FSCK_ERROR;
187 else if (!strcmp(str, "warn"))
188 return FSCK_WARN;
189 else if (!strcmp(str, "ignore"))
190 return FSCK_IGNORE;
191 else
192 die("Unknown fsck message type: '%s'", str);
195 int is_valid_msg_type(const char *msg_id, const char *msg_type)
197 if (parse_msg_id(msg_id) < 0)
198 return 0;
199 parse_msg_type(msg_type);
200 return 1;
203 void fsck_set_msg_type(struct fsck_options *options,
204 const char *msg_id_str, const char *msg_type_str)
206 int msg_id = parse_msg_id(msg_id_str), msg_type;
208 if (msg_id < 0)
209 die("Unhandled message id: %s", msg_id_str);
210 msg_type = parse_msg_type(msg_type_str);
212 if (msg_type != FSCK_ERROR && msg_id_info[msg_id].msg_type == FSCK_FATAL)
213 die("Cannot demote %s to %s", msg_id_str, msg_type_str);
215 if (!options->msg_type) {
216 int i;
217 int *severity;
218 ALLOC_ARRAY(severity, FSCK_MSG_MAX);
219 for (i = 0; i < FSCK_MSG_MAX; i++)
220 severity[i] = fsck_msg_type(i, options);
221 options->msg_type = severity;
224 options->msg_type[msg_id] = msg_type;
227 void fsck_set_msg_types(struct fsck_options *options, const char *values)
229 char *buf = xstrdup(values), *to_free = buf;
230 int done = 0;
232 while (!done) {
233 int len = strcspn(buf, " ,|"), equal;
235 done = !buf[len];
236 if (!len) {
237 buf++;
238 continue;
240 buf[len] = '\0';
242 for (equal = 0;
243 equal < len && buf[equal] != '=' && buf[equal] != ':';
244 equal++)
245 buf[equal] = tolower(buf[equal]);
246 buf[equal] = '\0';
248 if (!strcmp(buf, "skiplist")) {
249 if (equal == len)
250 die("skiplist requires a path");
251 oidset_parse_file(&options->skiplist, buf + equal + 1);
252 buf += len + 1;
253 continue;
256 if (equal == len)
257 die("Missing '=': '%s'", buf);
259 fsck_set_msg_type(options, buf, buf + equal + 1);
260 buf += len + 1;
262 free(to_free);
265 static int object_on_skiplist(struct fsck_options *opts,
266 const struct object_id *oid)
268 return opts && oid && oidset_contains(&opts->skiplist, oid);
271 __attribute__((format (printf, 5, 6)))
272 static int report(struct fsck_options *options,
273 const struct object_id *oid, enum object_type object_type,
274 enum fsck_msg_id msg_id, const char *fmt, ...)
276 va_list ap;
277 struct strbuf sb = STRBUF_INIT;
278 int msg_type = fsck_msg_type(msg_id, options), result;
280 if (msg_type == FSCK_IGNORE)
281 return 0;
283 if (object_on_skiplist(options, oid))
284 return 0;
286 if (msg_type == FSCK_FATAL)
287 msg_type = FSCK_ERROR;
288 else if (msg_type == FSCK_INFO)
289 msg_type = FSCK_WARN;
291 prepare_msg_ids();
292 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
294 va_start(ap, fmt);
295 strbuf_vaddf(&sb, fmt, ap);
296 result = options->error_func(options, oid, object_type,
297 msg_type, sb.buf);
298 strbuf_release(&sb);
299 va_end(ap);
301 return result;
304 void fsck_enable_object_names(struct fsck_options *options)
306 if (!options->object_names)
307 options->object_names = kh_init_oid_map();
310 const char *fsck_get_object_name(struct fsck_options *options,
311 const struct object_id *oid)
313 khiter_t pos;
314 if (!options->object_names)
315 return NULL;
316 pos = kh_get_oid_map(options->object_names, *oid);
317 if (pos >= kh_end(options->object_names))
318 return NULL;
319 return kh_value(options->object_names, pos);
322 void fsck_put_object_name(struct fsck_options *options,
323 const struct object_id *oid,
324 const char *fmt, ...)
326 va_list ap;
327 struct strbuf buf = STRBUF_INIT;
328 khiter_t pos;
329 int hashret;
331 if (!options->object_names)
332 return;
334 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
335 if (!hashret)
336 return;
337 va_start(ap, fmt);
338 strbuf_vaddf(&buf, fmt, ap);
339 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
340 va_end(ap);
343 const char *fsck_describe_object(struct fsck_options *options,
344 const struct object_id *oid)
346 static struct strbuf bufs[] = {
347 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
349 static int b = 0;
350 struct strbuf *buf;
351 const char *name = fsck_get_object_name(options, oid);
353 buf = bufs + b;
354 b = (b + 1) % ARRAY_SIZE(bufs);
355 strbuf_reset(buf);
356 strbuf_addstr(buf, oid_to_hex(oid));
357 if (name)
358 strbuf_addf(buf, " (%s)", name);
360 return buf->buf;
363 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
365 struct tree_desc desc;
366 struct name_entry entry;
367 int res = 0;
368 const char *name;
370 if (parse_tree(tree))
371 return -1;
373 name = fsck_get_object_name(options, &tree->object.oid);
374 if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
375 return -1;
376 while (tree_entry_gently(&desc, &entry)) {
377 struct object *obj;
378 int result;
380 if (S_ISGITLINK(entry.mode))
381 continue;
383 if (S_ISDIR(entry.mode)) {
384 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
385 if (name && obj)
386 fsck_put_object_name(options, &entry.oid, "%s%s/",
387 name, entry.path);
388 result = options->walk(obj, OBJ_TREE, data, options);
390 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
391 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
392 if (name && obj)
393 fsck_put_object_name(options, &entry.oid, "%s%s",
394 name, entry.path);
395 result = options->walk(obj, OBJ_BLOB, data, options);
397 else {
398 result = error("in tree %s: entry %s has bad mode %.6o",
399 fsck_describe_object(options, &tree->object.oid),
400 entry.path, entry.mode);
402 if (result < 0)
403 return result;
404 if (!res)
405 res = result;
407 return res;
410 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
412 int counter = 0, generation = 0, name_prefix_len = 0;
413 struct commit_list *parents;
414 int res;
415 int result;
416 const char *name;
418 if (parse_commit(commit))
419 return -1;
421 name = fsck_get_object_name(options, &commit->object.oid);
422 if (name)
423 fsck_put_object_name(options, get_commit_tree_oid(commit),
424 "%s:", name);
426 result = options->walk((struct object *)get_commit_tree(commit),
427 OBJ_TREE, data, options);
428 if (result < 0)
429 return result;
430 res = result;
432 parents = commit->parents;
433 if (name && parents) {
434 int len = strlen(name), power;
436 if (len && name[len - 1] == '^') {
437 generation = 1;
438 name_prefix_len = len - 1;
440 else { /* parse ~<generation> suffix */
441 for (generation = 0, power = 1;
442 len && isdigit(name[len - 1]);
443 power *= 10)
444 generation += power * (name[--len] - '0');
445 if (power > 1 && len && name[len - 1] == '~')
446 name_prefix_len = len - 1;
447 else {
448 /* Maybe a non-first parent, e.g. HEAD^2 */
449 generation = 0;
450 name_prefix_len = len;
455 while (parents) {
456 if (name) {
457 struct object_id *oid = &parents->item->object.oid;
459 if (counter++)
460 fsck_put_object_name(options, oid, "%s^%d",
461 name, counter);
462 else if (generation > 0)
463 fsck_put_object_name(options, oid, "%.*s~%d",
464 name_prefix_len, name,
465 generation + 1);
466 else
467 fsck_put_object_name(options, oid, "%s^", name);
469 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
470 if (result < 0)
471 return result;
472 if (!res)
473 res = result;
474 parents = parents->next;
476 return res;
479 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
481 const char *name = fsck_get_object_name(options, &tag->object.oid);
483 if (parse_tag(tag))
484 return -1;
485 if (name)
486 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
487 return options->walk(tag->tagged, OBJ_ANY, data, options);
490 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
492 if (!obj)
493 return -1;
495 if (obj->type == OBJ_NONE)
496 parse_object(the_repository, &obj->oid);
498 switch (obj->type) {
499 case OBJ_BLOB:
500 return 0;
501 case OBJ_TREE:
502 return fsck_walk_tree((struct tree *)obj, data, options);
503 case OBJ_COMMIT:
504 return fsck_walk_commit((struct commit *)obj, data, options);
505 case OBJ_TAG:
506 return fsck_walk_tag((struct tag *)obj, data, options);
507 default:
508 error("Unknown object type for %s",
509 fsck_describe_object(options, &obj->oid));
510 return -1;
514 struct name_stack {
515 const char **names;
516 size_t nr, alloc;
519 static void name_stack_push(struct name_stack *stack, const char *name)
521 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
522 stack->names[stack->nr++] = name;
525 static const char *name_stack_pop(struct name_stack *stack)
527 return stack->nr ? stack->names[--stack->nr] : NULL;
530 static void name_stack_clear(struct name_stack *stack)
532 FREE_AND_NULL(stack->names);
533 stack->nr = stack->alloc = 0;
537 * The entries in a tree are ordered in the _path_ order,
538 * which means that a directory entry is ordered by adding
539 * a slash to the end of it.
541 * So a directory called "a" is ordered _after_ a file
542 * called "a.c", because "a/" sorts after "a.c".
544 #define TREE_UNORDERED (-1)
545 #define TREE_HAS_DUPS (-2)
547 static int is_less_than_slash(unsigned char c)
549 return '\0' < c && c < '/';
552 static int verify_ordered(unsigned mode1, const char *name1,
553 unsigned mode2, const char *name2,
554 struct name_stack *candidates)
556 int len1 = strlen(name1);
557 int len2 = strlen(name2);
558 int len = len1 < len2 ? len1 : len2;
559 unsigned char c1, c2;
560 int cmp;
562 cmp = memcmp(name1, name2, len);
563 if (cmp < 0)
564 return 0;
565 if (cmp > 0)
566 return TREE_UNORDERED;
569 * Ok, the first <len> characters are the same.
570 * Now we need to order the next one, but turn
571 * a '\0' into a '/' for a directory entry.
573 c1 = name1[len];
574 c2 = name2[len];
575 if (!c1 && !c2)
577 * git-write-tree used to write out a nonsense tree that has
578 * entries with the same name, one blob and one tree. Make
579 * sure we do not have duplicate entries.
581 return TREE_HAS_DUPS;
582 if (!c1 && S_ISDIR(mode1))
583 c1 = '/';
584 if (!c2 && S_ISDIR(mode2))
585 c2 = '/';
588 * There can be non-consecutive duplicates due to the implicitly
589 * added slash, e.g.:
591 * foo
592 * foo.bar
593 * foo.bar.baz
594 * foo.bar/
595 * foo/
597 * Record non-directory candidates (like "foo" and "foo.bar" in
598 * the example) on a stack and check directory candidates (like
599 * foo/" and "foo.bar/") against that stack.
601 if (!c1 && is_less_than_slash(c2)) {
602 name_stack_push(candidates, name1);
603 } else if (c2 == '/' && is_less_than_slash(c1)) {
604 for (;;) {
605 const char *p;
606 const char *f_name = name_stack_pop(candidates);
608 if (!f_name)
609 break;
610 if (!skip_prefix(name2, f_name, &p))
611 continue;
612 if (!*p)
613 return TREE_HAS_DUPS;
614 if (is_less_than_slash(*p)) {
615 name_stack_push(candidates, f_name);
616 break;
621 return c1 < c2 ? 0 : TREE_UNORDERED;
624 static int fsck_tree(const struct object_id *oid,
625 const char *buffer, unsigned long size,
626 struct fsck_options *options)
628 int retval = 0;
629 int has_null_sha1 = 0;
630 int has_full_path = 0;
631 int has_empty_name = 0;
632 int has_dot = 0;
633 int has_dotdot = 0;
634 int has_dotgit = 0;
635 int has_zero_pad = 0;
636 int has_bad_modes = 0;
637 int has_dup_entries = 0;
638 int not_properly_sorted = 0;
639 struct tree_desc desc;
640 unsigned o_mode;
641 const char *o_name;
642 struct name_stack df_dup_candidates = { NULL };
644 if (init_tree_desc_gently(&desc, buffer, size)) {
645 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
646 return retval;
649 o_mode = 0;
650 o_name = NULL;
652 while (desc.size) {
653 unsigned short mode;
654 const char *name, *backslash;
655 const struct object_id *oid;
657 oid = tree_entry_extract(&desc, &name, &mode);
659 has_null_sha1 |= is_null_oid(oid);
660 has_full_path |= !!strchr(name, '/');
661 has_empty_name |= !*name;
662 has_dot |= !strcmp(name, ".");
663 has_dotdot |= !strcmp(name, "..");
664 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
665 has_zero_pad |= *(char *)desc.buffer == '0';
667 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
668 if (!S_ISLNK(mode))
669 oidset_insert(&gitmodules_found, oid);
670 else
671 retval += report(options,
672 oid, OBJ_TREE,
673 FSCK_MSG_GITMODULES_SYMLINK,
674 ".gitmodules is a symbolic link");
677 if ((backslash = strchr(name, '\\'))) {
678 while (backslash) {
679 backslash++;
680 has_dotgit |= is_ntfs_dotgit(backslash);
681 if (is_ntfs_dotgitmodules(backslash)) {
682 if (!S_ISLNK(mode))
683 oidset_insert(&gitmodules_found, oid);
684 else
685 retval += report(options, oid, OBJ_TREE,
686 FSCK_MSG_GITMODULES_SYMLINK,
687 ".gitmodules is a symbolic link");
689 backslash = strchr(backslash, '\\');
693 if (update_tree_entry_gently(&desc)) {
694 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
695 break;
698 switch (mode) {
700 * Standard modes..
702 case S_IFREG | 0755:
703 case S_IFREG | 0644:
704 case S_IFLNK:
705 case S_IFDIR:
706 case S_IFGITLINK:
707 break;
709 * This is nonstandard, but we had a few of these
710 * early on when we honored the full set of mode
711 * bits..
713 case S_IFREG | 0664:
714 if (!options->strict)
715 break;
716 /* fallthrough */
717 default:
718 has_bad_modes = 1;
721 if (o_name) {
722 switch (verify_ordered(o_mode, o_name, mode, name,
723 &df_dup_candidates)) {
724 case TREE_UNORDERED:
725 not_properly_sorted = 1;
726 break;
727 case TREE_HAS_DUPS:
728 has_dup_entries = 1;
729 break;
730 default:
731 break;
735 o_mode = mode;
736 o_name = name;
739 name_stack_clear(&df_dup_candidates);
741 if (has_null_sha1)
742 retval += report(options, oid, OBJ_TREE, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
743 if (has_full_path)
744 retval += report(options, oid, OBJ_TREE, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
745 if (has_empty_name)
746 retval += report(options, oid, OBJ_TREE, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
747 if (has_dot)
748 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOT, "contains '.'");
749 if (has_dotdot)
750 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTDOT, "contains '..'");
751 if (has_dotgit)
752 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
753 if (has_zero_pad)
754 retval += report(options, oid, OBJ_TREE, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
755 if (has_bad_modes)
756 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
757 if (has_dup_entries)
758 retval += report(options, oid, OBJ_TREE, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
759 if (not_properly_sorted)
760 retval += report(options, oid, OBJ_TREE, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
761 return retval;
764 static int verify_headers(const void *data, unsigned long size,
765 const struct object_id *oid, enum object_type type,
766 struct fsck_options *options)
768 const char *buffer = (const char *)data;
769 unsigned long i;
771 for (i = 0; i < size; i++) {
772 switch (buffer[i]) {
773 case '\0':
774 return report(options, oid, type,
775 FSCK_MSG_NUL_IN_HEADER,
776 "unterminated header: NUL at offset %ld", i);
777 case '\n':
778 if (i + 1 < size && buffer[i + 1] == '\n')
779 return 0;
784 * We did not find double-LF that separates the header
785 * and the body. Not having a body is not a crime but
786 * we do want to see the terminating LF for the last header
787 * line.
789 if (size && buffer[size - 1] == '\n')
790 return 0;
792 return report(options, oid, type,
793 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
796 static int fsck_ident(const char **ident,
797 const struct object_id *oid, enum object_type type,
798 struct fsck_options *options)
800 const char *p = *ident;
801 char *end;
803 *ident = strchrnul(*ident, '\n');
804 if (**ident == '\n')
805 (*ident)++;
807 if (*p == '<')
808 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
809 p += strcspn(p, "<>\n");
810 if (*p == '>')
811 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
812 if (*p != '<')
813 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
814 if (p[-1] != ' ')
815 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
816 p++;
817 p += strcspn(p, "<>\n");
818 if (*p != '>')
819 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
820 p++;
821 if (*p != ' ')
822 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
823 p++;
824 if (*p == '0' && p[1] != ' ')
825 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
826 if (date_overflows(parse_timestamp(p, &end, 10)))
827 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
828 if ((end == p || *end != ' '))
829 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
830 p = end + 1;
831 if ((*p != '+' && *p != '-') ||
832 !isdigit(p[1]) ||
833 !isdigit(p[2]) ||
834 !isdigit(p[3]) ||
835 !isdigit(p[4]) ||
836 (p[5] != '\n'))
837 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
838 p += 6;
839 return 0;
842 static int fsck_commit(const struct object_id *oid,
843 const char *buffer, unsigned long size,
844 struct fsck_options *options)
846 struct object_id tree_oid, parent_oid;
847 unsigned author_count;
848 int err;
849 const char *buffer_begin = buffer;
850 const char *p;
852 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
853 return -1;
855 if (!skip_prefix(buffer, "tree ", &buffer))
856 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
857 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
858 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
859 if (err)
860 return err;
862 buffer = p + 1;
863 while (skip_prefix(buffer, "parent ", &buffer)) {
864 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
865 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
866 if (err)
867 return err;
869 buffer = p + 1;
871 author_count = 0;
872 while (skip_prefix(buffer, "author ", &buffer)) {
873 author_count++;
874 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
875 if (err)
876 return err;
878 if (author_count < 1)
879 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
880 else if (author_count > 1)
881 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
882 if (err)
883 return err;
884 if (!skip_prefix(buffer, "committer ", &buffer))
885 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
886 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
887 if (err)
888 return err;
889 if (memchr(buffer_begin, '\0', size)) {
890 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
891 "NUL byte in the commit object body");
892 if (err)
893 return err;
895 return 0;
898 static int fsck_tag(const struct object_id *oid, const char *buffer,
899 unsigned long size, struct fsck_options *options)
901 struct object_id tagged_oid;
902 int tagged_type;
903 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
904 &tagged_type);
907 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
908 unsigned long size, struct fsck_options *options,
909 struct object_id *tagged_oid,
910 int *tagged_type)
912 int ret = 0;
913 char *eol;
914 struct strbuf sb = STRBUF_INIT;
915 const char *p;
917 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
918 if (ret)
919 goto done;
921 if (!skip_prefix(buffer, "object ", &buffer)) {
922 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
923 goto done;
925 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
926 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
927 if (ret)
928 goto done;
930 buffer = p + 1;
932 if (!skip_prefix(buffer, "type ", &buffer)) {
933 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
934 goto done;
936 eol = strchr(buffer, '\n');
937 if (!eol) {
938 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
939 goto done;
941 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
942 if (*tagged_type < 0)
943 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
944 if (ret)
945 goto done;
946 buffer = eol + 1;
948 if (!skip_prefix(buffer, "tag ", &buffer)) {
949 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
950 goto done;
952 eol = strchr(buffer, '\n');
953 if (!eol) {
954 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
955 goto done;
957 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
958 if (check_refname_format(sb.buf, 0)) {
959 ret = report(options, oid, OBJ_TAG,
960 FSCK_MSG_BAD_TAG_NAME,
961 "invalid 'tag' name: %.*s",
962 (int)(eol - buffer), buffer);
963 if (ret)
964 goto done;
966 buffer = eol + 1;
968 if (!skip_prefix(buffer, "tagger ", &buffer)) {
969 /* early tags do not contain 'tagger' lines; warn only */
970 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
971 if (ret)
972 goto done;
974 else
975 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
976 if (!*buffer)
977 goto done;
979 if (!starts_with(buffer, "\n")) {
981 * The verify_headers() check will allow
982 * e.g. "[...]tagger <tagger>\nsome
983 * garbage\n\nmessage" to pass, thinking "some
984 * garbage" could be a custom header. E.g. "mktag"
985 * doesn't want any unknown headers.
987 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
988 if (ret)
989 goto done;
992 done:
993 strbuf_release(&sb);
994 return ret;
998 * Like builtin/submodule--helper.c's starts_with_dot_slash, but without
999 * relying on the platform-dependent is_dir_sep helper.
1001 * This is for use in checking whether a submodule URL is interpreted as
1002 * relative to the current directory on any platform, since \ is a
1003 * directory separator on Windows but not on other platforms.
1005 static int starts_with_dot_slash(const char *str)
1007 return str[0] == '.' && (str[1] == '/' || str[1] == '\\');
1011 * Like starts_with_dot_slash, this is a variant of submodule--helper's
1012 * helper of the same name with the twist that it accepts backslash as a
1013 * directory separator even on non-Windows platforms.
1015 static int starts_with_dot_dot_slash(const char *str)
1017 return str[0] == '.' && starts_with_dot_slash(str + 1);
1020 static int submodule_url_is_relative(const char *url)
1022 return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
1026 * Count directory components that a relative submodule URL should chop
1027 * from the remote_url it is to be resolved against.
1029 * In other words, this counts "../" components at the start of a
1030 * submodule URL.
1032 * Returns the number of directory components to chop and writes a
1033 * pointer to the next character of url after all leading "./" and
1034 * "../" components to out.
1036 static int count_leading_dotdots(const char *url, const char **out)
1038 int result = 0;
1039 while (1) {
1040 if (starts_with_dot_dot_slash(url)) {
1041 result++;
1042 url += strlen("../");
1043 continue;
1045 if (starts_with_dot_slash(url)) {
1046 url += strlen("./");
1047 continue;
1049 *out = url;
1050 return result;
1054 * Check whether a transport is implemented by git-remote-curl.
1056 * If it is, returns 1 and writes the URL that would be passed to
1057 * git-remote-curl to the "out" parameter.
1059 * Otherwise, returns 0 and leaves "out" untouched.
1061 * Examples:
1062 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1063 * https://example.com/repo.git -> 1, https://example.com/repo.git
1064 * git://example.com/repo.git -> 0
1066 * This is for use in checking for previously exploitable bugs that
1067 * required a submodule URL to be passed to git-remote-curl.
1069 static int url_to_curl_url(const char *url, const char **out)
1072 * We don't need to check for case-aliases, "http.exe", and so
1073 * on because in the default configuration, is_transport_allowed
1074 * prevents URLs with those schemes from being cloned
1075 * automatically.
1077 if (skip_prefix(url, "http::", out) ||
1078 skip_prefix(url, "https::", out) ||
1079 skip_prefix(url, "ftp::", out) ||
1080 skip_prefix(url, "ftps::", out))
1081 return 1;
1082 if (starts_with(url, "http://") ||
1083 starts_with(url, "https://") ||
1084 starts_with(url, "ftp://") ||
1085 starts_with(url, "ftps://")) {
1086 *out = url;
1087 return 1;
1089 return 0;
1092 static int check_submodule_url(const char *url)
1094 const char *curl_url;
1096 if (looks_like_command_line_option(url))
1097 return -1;
1099 if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
1100 char *decoded;
1101 const char *next;
1102 int has_nl;
1105 * This could be appended to an http URL and url-decoded;
1106 * check for malicious characters.
1108 decoded = url_decode(url);
1109 has_nl = !!strchr(decoded, '\n');
1111 free(decoded);
1112 if (has_nl)
1113 return -1;
1116 * URLs which escape their root via "../" can overwrite
1117 * the host field and previous components, resolving to
1118 * URLs like https::example.com/submodule.git and
1119 * https:///example.com/submodule.git that were
1120 * susceptible to CVE-2020-11008.
1122 if (count_leading_dotdots(url, &next) > 0 &&
1123 (*next == ':' || *next == '/'))
1124 return -1;
1127 else if (url_to_curl_url(url, &curl_url)) {
1128 struct credential c = CREDENTIAL_INIT;
1129 int ret = 0;
1130 if (credential_from_url_gently(&c, curl_url, 1) ||
1131 !*c.host)
1132 ret = -1;
1133 credential_clear(&c);
1134 return ret;
1137 return 0;
1140 struct fsck_gitmodules_data {
1141 const struct object_id *oid;
1142 struct fsck_options *options;
1143 int ret;
1146 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1148 struct fsck_gitmodules_data *data = vdata;
1149 const char *subsection, *key;
1150 size_t subsection_len;
1151 char *name;
1153 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1154 !subsection)
1155 return 0;
1157 name = xmemdupz(subsection, subsection_len);
1158 if (check_submodule_name(name) < 0)
1159 data->ret |= report(data->options,
1160 data->oid, OBJ_BLOB,
1161 FSCK_MSG_GITMODULES_NAME,
1162 "disallowed submodule name: %s",
1163 name);
1164 if (!strcmp(key, "url") && value &&
1165 check_submodule_url(value) < 0)
1166 data->ret |= report(data->options,
1167 data->oid, OBJ_BLOB,
1168 FSCK_MSG_GITMODULES_URL,
1169 "disallowed submodule url: %s",
1170 value);
1171 if (!strcmp(key, "path") && value &&
1172 looks_like_command_line_option(value))
1173 data->ret |= report(data->options,
1174 data->oid, OBJ_BLOB,
1175 FSCK_MSG_GITMODULES_PATH,
1176 "disallowed submodule path: %s",
1177 value);
1178 if (!strcmp(key, "update") && value &&
1179 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1180 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1181 FSCK_MSG_GITMODULES_UPDATE,
1182 "disallowed submodule update setting: %s",
1183 value);
1184 free(name);
1186 return 0;
1189 static int fsck_blob(const struct object_id *oid, const char *buf,
1190 unsigned long size, struct fsck_options *options)
1192 struct fsck_gitmodules_data data;
1193 struct config_options config_opts = { 0 };
1195 if (!oidset_contains(&gitmodules_found, oid))
1196 return 0;
1197 oidset_insert(&gitmodules_done, oid);
1199 if (object_on_skiplist(options, oid))
1200 return 0;
1202 if (!buf) {
1204 * A missing buffer here is a sign that the caller found the
1205 * blob too gigantic to load into memory. Let's just consider
1206 * that an error.
1208 return report(options, oid, OBJ_BLOB,
1209 FSCK_MSG_GITMODULES_LARGE,
1210 ".gitmodules too large to parse");
1213 data.oid = oid;
1214 data.options = options;
1215 data.ret = 0;
1216 config_opts.error_action = CONFIG_ERROR_SILENT;
1217 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1218 ".gitmodules", buf, size, &data, &config_opts))
1219 data.ret |= report(options, oid, OBJ_BLOB,
1220 FSCK_MSG_GITMODULES_PARSE,
1221 "could not parse gitmodules blob");
1223 return data.ret;
1226 int fsck_object(struct object *obj, void *data, unsigned long size,
1227 struct fsck_options *options)
1229 if (!obj)
1230 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1232 if (obj->type == OBJ_BLOB)
1233 return fsck_blob(&obj->oid, data, size, options);
1234 if (obj->type == OBJ_TREE)
1235 return fsck_tree(&obj->oid, data, size, options);
1236 if (obj->type == OBJ_COMMIT)
1237 return fsck_commit(&obj->oid, data, size, options);
1238 if (obj->type == OBJ_TAG)
1239 return fsck_tag(&obj->oid, data, size, options);
1241 return report(options, &obj->oid, obj->type,
1242 FSCK_MSG_UNKNOWN_TYPE,
1243 "unknown type '%d' (internal fsck error)",
1244 obj->type);
1247 int fsck_error_function(struct fsck_options *o,
1248 const struct object_id *oid,
1249 enum object_type object_type,
1250 int msg_type, const char *message)
1252 if (msg_type == FSCK_WARN) {
1253 warning("object %s: %s", fsck_describe_object(o, oid), message);
1254 return 0;
1256 error("object %s: %s", fsck_describe_object(o, oid), message);
1257 return 1;
1260 void register_found_gitmodules(const struct object_id *oid)
1262 oidset_insert(&gitmodules_found, oid);
1265 int fsck_finish(struct fsck_options *options)
1267 int ret = 0;
1268 struct oidset_iter iter;
1269 const struct object_id *oid;
1271 oidset_iter_init(&gitmodules_found, &iter);
1272 while ((oid = oidset_iter_next(&iter))) {
1273 enum object_type type;
1274 unsigned long size;
1275 char *buf;
1277 if (oidset_contains(&gitmodules_done, oid))
1278 continue;
1280 buf = read_object_file(oid, &type, &size);
1281 if (!buf) {
1282 if (is_promisor_object(oid))
1283 continue;
1284 ret |= report(options,
1285 oid, OBJ_BLOB,
1286 FSCK_MSG_GITMODULES_MISSING,
1287 "unable to read .gitmodules blob");
1288 continue;
1291 if (type == OBJ_BLOB)
1292 ret |= fsck_blob(oid, buf, size, options);
1293 else
1294 ret |= report(options,
1295 oid, type,
1296 FSCK_MSG_GITMODULES_BLOB,
1297 "non-blob found at .gitmodules");
1298 free(buf);
1302 oidset_clear(&gitmodules_found);
1303 oidset_clear(&gitmodules_done);
1304 return ret;
1307 int git_fsck_config(const char *var, const char *value, void *cb)
1309 struct fsck_options *options = cb;
1310 if (strcmp(var, "fsck.skiplist") == 0) {
1311 const char *path;
1312 struct strbuf sb = STRBUF_INIT;
1314 if (git_config_pathname(&path, var, value))
1315 return 1;
1316 strbuf_addf(&sb, "skiplist=%s", path);
1317 free((char *)path);
1318 fsck_set_msg_types(options, sb.buf);
1319 strbuf_release(&sb);
1320 return 0;
1323 if (skip_prefix(var, "fsck.", &var)) {
1324 fsck_set_msg_type(options, var, value);
1325 return 0;
1328 return git_default_config(var, value, cb);