]> git.ipfire.org Git - thirdparty/git.git/blob - fsck.c
Merge branch 'jk/mailmap-only-at-root'
[thirdparty/git.git] / fsck.c
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"
21
22 static struct oidset gitmodules_found = OIDSET_INIT;
23 static struct oidset gitmodules_done = OIDSET_INIT;
24
25 #define FSCK_FATAL -1
26 #define FSCK_INFO -2
27
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)
86
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
91 };
92 #undef MSG_ID
93
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 }
104 };
105 #undef MSG_ID
106
107 static void prepare_msg_ids(void)
108 {
109 int i;
110
111 if (msg_id_info[0].downcased)
112 return;
113
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);
119
120 msg_id_info[i].downcased = q;
121 while (*p)
122 if (*p == '_')
123 p++;
124 else
125 *(q)++ = tolower(*(p)++);
126 *q = '\0';
127
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++);
138 }
139 }
140 *q = '\0';
141 }
142 }
143
144 static int parse_msg_id(const char *text)
145 {
146 int i;
147
148 prepare_msg_ids();
149
150 for (i = 0; i < FSCK_MSG_MAX; i++)
151 if (!strcmp(text, msg_id_info[i].downcased))
152 return i;
153
154 return -1;
155 }
156
157 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
158 {
159 int i;
160
161 prepare_msg_ids();
162
163 for (i = 0; i < FSCK_MSG_MAX; i++)
164 list_config_item(list, prefix, msg_id_info[i].camelcased);
165 }
166
167 static int fsck_msg_type(enum fsck_msg_id msg_id,
168 struct fsck_options *options)
169 {
170 int msg_type;
171
172 assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
173
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;
180 }
181
182 return msg_type;
183 }
184
185 static int parse_msg_type(const char *str)
186 {
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);
195 }
196
197 int is_valid_msg_type(const char *msg_id, const char *msg_type)
198 {
199 if (parse_msg_id(msg_id) < 0)
200 return 0;
201 parse_msg_type(msg_type);
202 return 1;
203 }
204
205 void fsck_set_msg_type(struct fsck_options *options,
206 const char *msg_id, const char *msg_type)
207 {
208 int id = parse_msg_id(msg_id), type;
209
210 if (id < 0)
211 die("Unhandled message id: %s", msg_id);
212 type = parse_msg_type(msg_type);
213
214 if (type != FSCK_ERROR && msg_id_info[id].msg_type == FSCK_FATAL)
215 die("Cannot demote %s to %s", msg_id, msg_type);
216
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;
224 }
225
226 options->msg_type[id] = type;
227 }
228
229 void fsck_set_msg_types(struct fsck_options *options, const char *values)
230 {
231 char *buf = xstrdup(values), *to_free = buf;
232 int done = 0;
233
234 while (!done) {
235 int len = strcspn(buf, " ,|"), equal;
236
237 done = !buf[len];
238 if (!len) {
239 buf++;
240 continue;
241 }
242 buf[len] = '\0';
243
244 for (equal = 0;
245 equal < len && buf[equal] != '=' && buf[equal] != ':';
246 equal++)
247 buf[equal] = tolower(buf[equal]);
248 buf[equal] = '\0';
249
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;
256 }
257
258 if (equal == len)
259 die("Missing '=': '%s'", buf);
260
261 fsck_set_msg_type(options, buf, buf + equal + 1);
262 buf += len + 1;
263 }
264 free(to_free);
265 }
266
267 static void append_msg_id(struct strbuf *sb, const char *msg_id)
268 {
269 for (;;) {
270 char c = *(msg_id)++;
271
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)++);
279 }
280 }
281
282 strbuf_addstr(sb, ": ");
283 }
284
285 static int object_on_skiplist(struct fsck_options *opts,
286 const struct object_id *oid)
287 {
288 return opts && oid && oidset_contains(&opts->skiplist, oid);
289 }
290
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, ...)
295 {
296 va_list ap;
297 struct strbuf sb = STRBUF_INIT;
298 int msg_type = fsck_msg_type(id, options), result;
299
300 if (msg_type == FSCK_IGNORE)
301 return 0;
302
303 if (object_on_skiplist(options, oid))
304 return 0;
305
306 if (msg_type == FSCK_FATAL)
307 msg_type = FSCK_ERROR;
308 else if (msg_type == FSCK_INFO)
309 msg_type = FSCK_WARN;
310
311 append_msg_id(&sb, msg_id_info[id].id_string);
312
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);
319
320 return result;
321 }
322
323 void fsck_enable_object_names(struct fsck_options *options)
324 {
325 if (!options->object_names)
326 options->object_names = kh_init_oid_map();
327 }
328
329 const char *fsck_get_object_name(struct fsck_options *options,
330 const struct object_id *oid)
331 {
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);
339 }
340
341 void fsck_put_object_name(struct fsck_options *options,
342 const struct object_id *oid,
343 const char *fmt, ...)
344 {
345 va_list ap;
346 struct strbuf buf = STRBUF_INIT;
347 khiter_t pos;
348 int hashret;
349
350 if (!options->object_names)
351 return;
352
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);
360 }
361
362 const char *fsck_describe_object(struct fsck_options *options,
363 const struct object_id *oid)
364 {
365 static struct strbuf bufs[] = {
366 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
367 };
368 static int b = 0;
369 struct strbuf *buf;
370 const char *name = fsck_get_object_name(options, oid);
371
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);
378
379 return buf->buf;
380 }
381
382 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
383 {
384 struct tree_desc desc;
385 struct name_entry entry;
386 int res = 0;
387 const char *name;
388
389 if (parse_tree(tree))
390 return -1;
391
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;
398
399 if (S_ISGITLINK(entry.mode))
400 continue;
401
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);
408 }
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);
415 }
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);
420 }
421 if (result < 0)
422 return result;
423 if (!res)
424 res = result;
425 }
426 return res;
427 }
428
429 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
430 {
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;
436
437 if (parse_commit(commit))
438 return -1;
439
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);
444
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;
450
451 parents = commit->parents;
452 if (name && parents) {
453 int len = strlen(name), power;
454
455 if (len && name[len - 1] == '^') {
456 generation = 1;
457 name_prefix_len = len - 1;
458 }
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 }
467 }
468
469 while (parents) {
470 if (name) {
471 struct object_id *oid = &parents->item->object.oid;
472
473 if (counter++)
474 fsck_put_object_name(options, oid, "%s^%d",
475 name, counter);
476 else if (generation > 0)
477 fsck_put_object_name(options, oid, "%.*s~%d",
478 name_prefix_len, name,
479 generation + 1);
480 else
481 fsck_put_object_name(options, oid, "%s^", name);
482 }
483 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
484 if (result < 0)
485 return result;
486 if (!res)
487 res = result;
488 parents = parents->next;
489 }
490 return res;
491 }
492
493 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
494 {
495 const char *name = fsck_get_object_name(options, &tag->object.oid);
496
497 if (parse_tag(tag))
498 return -1;
499 if (name)
500 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
501 return options->walk(tag->tagged, OBJ_ANY, data, options);
502 }
503
504 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
505 {
506 if (!obj)
507 return -1;
508
509 if (obj->type == OBJ_NONE)
510 parse_object(the_repository, &obj->oid);
511
512 switch (obj->type) {
513 case OBJ_BLOB:
514 return 0;
515 case OBJ_TREE:
516 return fsck_walk_tree((struct tree *)obj, data, options);
517 case OBJ_COMMIT:
518 return fsck_walk_commit((struct commit *)obj, data, options);
519 case OBJ_TAG:
520 return fsck_walk_tag((struct tag *)obj, data, options);
521 default:
522 error("Unknown object type for %s",
523 fsck_describe_object(options, &obj->oid));
524 return -1;
525 }
526 }
527
528 struct name_stack {
529 const char **names;
530 size_t nr, alloc;
531 };
532
533 static void name_stack_push(struct name_stack *stack, const char *name)
534 {
535 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
536 stack->names[stack->nr++] = name;
537 }
538
539 static const char *name_stack_pop(struct name_stack *stack)
540 {
541 return stack->nr ? stack->names[--stack->nr] : NULL;
542 }
543
544 static void name_stack_clear(struct name_stack *stack)
545 {
546 FREE_AND_NULL(stack->names);
547 stack->nr = stack->alloc = 0;
548 }
549
550 /*
551 * The entries in a tree are ordered in the _path_ order,
552 * which means that a directory entry is ordered by adding
553 * a slash to the end of it.
554 *
555 * So a directory called "a" is ordered _after_ a file
556 * called "a.c", because "a/" sorts after "a.c".
557 */
558 #define TREE_UNORDERED (-1)
559 #define TREE_HAS_DUPS (-2)
560
561 static int is_less_than_slash(unsigned char c)
562 {
563 return '\0' < c && c < '/';
564 }
565
566 static int verify_ordered(unsigned mode1, const char *name1,
567 unsigned mode2, const char *name2,
568 struct name_stack *candidates)
569 {
570 int len1 = strlen(name1);
571 int len2 = strlen(name2);
572 int len = len1 < len2 ? len1 : len2;
573 unsigned char c1, c2;
574 int cmp;
575
576 cmp = memcmp(name1, name2, len);
577 if (cmp < 0)
578 return 0;
579 if (cmp > 0)
580 return TREE_UNORDERED;
581
582 /*
583 * Ok, the first <len> characters are the same.
584 * Now we need to order the next one, but turn
585 * a '\0' into a '/' for a directory entry.
586 */
587 c1 = name1[len];
588 c2 = name2[len];
589 if (!c1 && !c2)
590 /*
591 * git-write-tree used to write out a nonsense tree that has
592 * entries with the same name, one blob and one tree. Make
593 * sure we do not have duplicate entries.
594 */
595 return TREE_HAS_DUPS;
596 if (!c1 && S_ISDIR(mode1))
597 c1 = '/';
598 if (!c2 && S_ISDIR(mode2))
599 c2 = '/';
600
601 /*
602 * There can be non-consecutive duplicates due to the implicitly
603 * added slash, e.g.:
604 *
605 * foo
606 * foo.bar
607 * foo.bar.baz
608 * foo.bar/
609 * foo/
610 *
611 * Record non-directory candidates (like "foo" and "foo.bar" in
612 * the example) on a stack and check directory candidates (like
613 * foo/" and "foo.bar/") against that stack.
614 */
615 if (!c1 && is_less_than_slash(c2)) {
616 name_stack_push(candidates, name1);
617 } else if (c2 == '/' && is_less_than_slash(c1)) {
618 for (;;) {
619 const char *p;
620 const char *f_name = name_stack_pop(candidates);
621
622 if (!f_name)
623 break;
624 if (!skip_prefix(name2, f_name, &p))
625 continue;
626 if (!*p)
627 return TREE_HAS_DUPS;
628 if (is_less_than_slash(*p)) {
629 name_stack_push(candidates, f_name);
630 break;
631 }
632 }
633 }
634
635 return c1 < c2 ? 0 : TREE_UNORDERED;
636 }
637
638 static int fsck_tree(const struct object_id *oid,
639 const char *buffer, unsigned long size,
640 struct fsck_options *options)
641 {
642 int retval = 0;
643 int has_null_sha1 = 0;
644 int has_full_path = 0;
645 int has_empty_name = 0;
646 int has_dot = 0;
647 int has_dotdot = 0;
648 int has_dotgit = 0;
649 int has_zero_pad = 0;
650 int has_bad_modes = 0;
651 int has_dup_entries = 0;
652 int not_properly_sorted = 0;
653 struct tree_desc desc;
654 unsigned o_mode;
655 const char *o_name;
656 struct name_stack df_dup_candidates = { NULL };
657
658 if (init_tree_desc_gently(&desc, buffer, size)) {
659 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
660 return retval;
661 }
662
663 o_mode = 0;
664 o_name = NULL;
665
666 while (desc.size) {
667 unsigned short mode;
668 const char *name, *backslash;
669 const struct object_id *oid;
670
671 oid = tree_entry_extract(&desc, &name, &mode);
672
673 has_null_sha1 |= is_null_oid(oid);
674 has_full_path |= !!strchr(name, '/');
675 has_empty_name |= !*name;
676 has_dot |= !strcmp(name, ".");
677 has_dotdot |= !strcmp(name, "..");
678 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
679 has_zero_pad |= *(char *)desc.buffer == '0';
680
681 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
682 if (!S_ISLNK(mode))
683 oidset_insert(&gitmodules_found, oid);
684 else
685 retval += report(options,
686 oid, OBJ_TREE,
687 FSCK_MSG_GITMODULES_SYMLINK,
688 ".gitmodules is a symbolic link");
689 }
690
691 if ((backslash = strchr(name, '\\'))) {
692 while (backslash) {
693 backslash++;
694 has_dotgit |= is_ntfs_dotgit(backslash);
695 if (is_ntfs_dotgitmodules(backslash)) {
696 if (!S_ISLNK(mode))
697 oidset_insert(&gitmodules_found, oid);
698 else
699 retval += report(options, oid, OBJ_TREE,
700 FSCK_MSG_GITMODULES_SYMLINK,
701 ".gitmodules is a symbolic link");
702 }
703 backslash = strchr(backslash, '\\');
704 }
705 }
706
707 if (update_tree_entry_gently(&desc)) {
708 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
709 break;
710 }
711
712 switch (mode) {
713 /*
714 * Standard modes..
715 */
716 case S_IFREG | 0755:
717 case S_IFREG | 0644:
718 case S_IFLNK:
719 case S_IFDIR:
720 case S_IFGITLINK:
721 break;
722 /*
723 * This is nonstandard, but we had a few of these
724 * early on when we honored the full set of mode
725 * bits..
726 */
727 case S_IFREG | 0664:
728 if (!options->strict)
729 break;
730 /* fallthrough */
731 default:
732 has_bad_modes = 1;
733 }
734
735 if (o_name) {
736 switch (verify_ordered(o_mode, o_name, mode, name,
737 &df_dup_candidates)) {
738 case TREE_UNORDERED:
739 not_properly_sorted = 1;
740 break;
741 case TREE_HAS_DUPS:
742 has_dup_entries = 1;
743 break;
744 default:
745 break;
746 }
747 }
748
749 o_mode = mode;
750 o_name = name;
751 }
752
753 name_stack_clear(&df_dup_candidates);
754
755 if (has_null_sha1)
756 retval += report(options, oid, OBJ_TREE, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
757 if (has_full_path)
758 retval += report(options, oid, OBJ_TREE, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
759 if (has_empty_name)
760 retval += report(options, oid, OBJ_TREE, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
761 if (has_dot)
762 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOT, "contains '.'");
763 if (has_dotdot)
764 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTDOT, "contains '..'");
765 if (has_dotgit)
766 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
767 if (has_zero_pad)
768 retval += report(options, oid, OBJ_TREE, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
769 if (has_bad_modes)
770 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
771 if (has_dup_entries)
772 retval += report(options, oid, OBJ_TREE, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
773 if (not_properly_sorted)
774 retval += report(options, oid, OBJ_TREE, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
775 return retval;
776 }
777
778 static int verify_headers(const void *data, unsigned long size,
779 const struct object_id *oid, enum object_type type,
780 struct fsck_options *options)
781 {
782 const char *buffer = (const char *)data;
783 unsigned long i;
784
785 for (i = 0; i < size; i++) {
786 switch (buffer[i]) {
787 case '\0':
788 return report(options, oid, type,
789 FSCK_MSG_NUL_IN_HEADER,
790 "unterminated header: NUL at offset %ld", i);
791 case '\n':
792 if (i + 1 < size && buffer[i + 1] == '\n')
793 return 0;
794 }
795 }
796
797 /*
798 * We did not find double-LF that separates the header
799 * and the body. Not having a body is not a crime but
800 * we do want to see the terminating LF for the last header
801 * line.
802 */
803 if (size && buffer[size - 1] == '\n')
804 return 0;
805
806 return report(options, oid, type,
807 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
808 }
809
810 static int fsck_ident(const char **ident,
811 const struct object_id *oid, enum object_type type,
812 struct fsck_options *options)
813 {
814 const char *p = *ident;
815 char *end;
816
817 *ident = strchrnul(*ident, '\n');
818 if (**ident == '\n')
819 (*ident)++;
820
821 if (*p == '<')
822 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
823 p += strcspn(p, "<>\n");
824 if (*p == '>')
825 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
826 if (*p != '<')
827 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
828 if (p[-1] != ' ')
829 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
830 p++;
831 p += strcspn(p, "<>\n");
832 if (*p != '>')
833 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
834 p++;
835 if (*p != ' ')
836 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
837 p++;
838 if (*p == '0' && p[1] != ' ')
839 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
840 if (date_overflows(parse_timestamp(p, &end, 10)))
841 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
842 if ((end == p || *end != ' '))
843 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
844 p = end + 1;
845 if ((*p != '+' && *p != '-') ||
846 !isdigit(p[1]) ||
847 !isdigit(p[2]) ||
848 !isdigit(p[3]) ||
849 !isdigit(p[4]) ||
850 (p[5] != '\n'))
851 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
852 p += 6;
853 return 0;
854 }
855
856 static int fsck_commit(const struct object_id *oid,
857 const char *buffer, unsigned long size,
858 struct fsck_options *options)
859 {
860 struct object_id tree_oid, parent_oid;
861 unsigned author_count;
862 int err;
863 const char *buffer_begin = buffer;
864 const char *p;
865
866 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
867 return -1;
868
869 if (!skip_prefix(buffer, "tree ", &buffer))
870 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
871 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
872 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
873 if (err)
874 return err;
875 }
876 buffer = p + 1;
877 while (skip_prefix(buffer, "parent ", &buffer)) {
878 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
879 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
880 if (err)
881 return err;
882 }
883 buffer = p + 1;
884 }
885 author_count = 0;
886 while (skip_prefix(buffer, "author ", &buffer)) {
887 author_count++;
888 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
889 if (err)
890 return err;
891 }
892 if (author_count < 1)
893 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
894 else if (author_count > 1)
895 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
896 if (err)
897 return err;
898 if (!skip_prefix(buffer, "committer ", &buffer))
899 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
900 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
901 if (err)
902 return err;
903 if (memchr(buffer_begin, '\0', size)) {
904 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
905 "NUL byte in the commit object body");
906 if (err)
907 return err;
908 }
909 return 0;
910 }
911
912 static int fsck_tag(const struct object_id *oid, const char *buffer,
913 unsigned long size, struct fsck_options *options)
914 {
915 struct object_id tagged_oid;
916 int tagged_type;
917 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
918 &tagged_type);
919 }
920
921 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
922 unsigned long size, struct fsck_options *options,
923 struct object_id *tagged_oid,
924 int *tagged_type)
925 {
926 int ret = 0;
927 char *eol;
928 struct strbuf sb = STRBUF_INIT;
929 const char *p;
930
931 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
932 if (ret)
933 goto done;
934
935 if (!skip_prefix(buffer, "object ", &buffer)) {
936 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
937 goto done;
938 }
939 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
940 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
941 if (ret)
942 goto done;
943 }
944 buffer = p + 1;
945
946 if (!skip_prefix(buffer, "type ", &buffer)) {
947 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
948 goto done;
949 }
950 eol = strchr(buffer, '\n');
951 if (!eol) {
952 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
953 goto done;
954 }
955 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
956 if (*tagged_type < 0)
957 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
958 if (ret)
959 goto done;
960 buffer = eol + 1;
961
962 if (!skip_prefix(buffer, "tag ", &buffer)) {
963 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
964 goto done;
965 }
966 eol = strchr(buffer, '\n');
967 if (!eol) {
968 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
969 goto done;
970 }
971 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
972 if (check_refname_format(sb.buf, 0)) {
973 ret = report(options, oid, OBJ_TAG,
974 FSCK_MSG_BAD_TAG_NAME,
975 "invalid 'tag' name: %.*s",
976 (int)(eol - buffer), buffer);
977 if (ret)
978 goto done;
979 }
980 buffer = eol + 1;
981
982 if (!skip_prefix(buffer, "tagger ", &buffer)) {
983 /* early tags do not contain 'tagger' lines; warn only */
984 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
985 if (ret)
986 goto done;
987 }
988 else
989 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
990 if (!*buffer)
991 goto done;
992
993 if (!starts_with(buffer, "\n")) {
994 /*
995 * The verify_headers() check will allow
996 * e.g. "[...]tagger <tagger>\nsome
997 * garbage\n\nmessage" to pass, thinking "some
998 * garbage" could be a custom header. E.g. "mktag"
999 * doesn't want any unknown headers.
1000 */
1001 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
1002 if (ret)
1003 goto done;
1004 }
1005
1006 done:
1007 strbuf_release(&sb);
1008 return ret;
1009 }
1010
1011 /*
1012 * Like builtin/submodule--helper.c's starts_with_dot_slash, but without
1013 * relying on the platform-dependent is_dir_sep helper.
1014 *
1015 * This is for use in checking whether a submodule URL is interpreted as
1016 * relative to the current directory on any platform, since \ is a
1017 * directory separator on Windows but not on other platforms.
1018 */
1019 static int starts_with_dot_slash(const char *str)
1020 {
1021 return str[0] == '.' && (str[1] == '/' || str[1] == '\\');
1022 }
1023
1024 /*
1025 * Like starts_with_dot_slash, this is a variant of submodule--helper's
1026 * helper of the same name with the twist that it accepts backslash as a
1027 * directory separator even on non-Windows platforms.
1028 */
1029 static int starts_with_dot_dot_slash(const char *str)
1030 {
1031 return str[0] == '.' && starts_with_dot_slash(str + 1);
1032 }
1033
1034 static int submodule_url_is_relative(const char *url)
1035 {
1036 return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
1037 }
1038
1039 /*
1040 * Count directory components that a relative submodule URL should chop
1041 * from the remote_url it is to be resolved against.
1042 *
1043 * In other words, this counts "../" components at the start of a
1044 * submodule URL.
1045 *
1046 * Returns the number of directory components to chop and writes a
1047 * pointer to the next character of url after all leading "./" and
1048 * "../" components to out.
1049 */
1050 static int count_leading_dotdots(const char *url, const char **out)
1051 {
1052 int result = 0;
1053 while (1) {
1054 if (starts_with_dot_dot_slash(url)) {
1055 result++;
1056 url += strlen("../");
1057 continue;
1058 }
1059 if (starts_with_dot_slash(url)) {
1060 url += strlen("./");
1061 continue;
1062 }
1063 *out = url;
1064 return result;
1065 }
1066 }
1067 /*
1068 * Check whether a transport is implemented by git-remote-curl.
1069 *
1070 * If it is, returns 1 and writes the URL that would be passed to
1071 * git-remote-curl to the "out" parameter.
1072 *
1073 * Otherwise, returns 0 and leaves "out" untouched.
1074 *
1075 * Examples:
1076 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1077 * https://example.com/repo.git -> 1, https://example.com/repo.git
1078 * git://example.com/repo.git -> 0
1079 *
1080 * This is for use in checking for previously exploitable bugs that
1081 * required a submodule URL to be passed to git-remote-curl.
1082 */
1083 static int url_to_curl_url(const char *url, const char **out)
1084 {
1085 /*
1086 * We don't need to check for case-aliases, "http.exe", and so
1087 * on because in the default configuration, is_transport_allowed
1088 * prevents URLs with those schemes from being cloned
1089 * automatically.
1090 */
1091 if (skip_prefix(url, "http::", out) ||
1092 skip_prefix(url, "https::", out) ||
1093 skip_prefix(url, "ftp::", out) ||
1094 skip_prefix(url, "ftps::", out))
1095 return 1;
1096 if (starts_with(url, "http://") ||
1097 starts_with(url, "https://") ||
1098 starts_with(url, "ftp://") ||
1099 starts_with(url, "ftps://")) {
1100 *out = url;
1101 return 1;
1102 }
1103 return 0;
1104 }
1105
1106 static int check_submodule_url(const char *url)
1107 {
1108 const char *curl_url;
1109
1110 if (looks_like_command_line_option(url))
1111 return -1;
1112
1113 if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
1114 char *decoded;
1115 const char *next;
1116 int has_nl;
1117
1118 /*
1119 * This could be appended to an http URL and url-decoded;
1120 * check for malicious characters.
1121 */
1122 decoded = url_decode(url);
1123 has_nl = !!strchr(decoded, '\n');
1124
1125 free(decoded);
1126 if (has_nl)
1127 return -1;
1128
1129 /*
1130 * URLs which escape their root via "../" can overwrite
1131 * the host field and previous components, resolving to
1132 * URLs like https::example.com/submodule.git and
1133 * https:///example.com/submodule.git that were
1134 * susceptible to CVE-2020-11008.
1135 */
1136 if (count_leading_dotdots(url, &next) > 0 &&
1137 (*next == ':' || *next == '/'))
1138 return -1;
1139 }
1140
1141 else if (url_to_curl_url(url, &curl_url)) {
1142 struct credential c = CREDENTIAL_INIT;
1143 int ret = 0;
1144 if (credential_from_url_gently(&c, curl_url, 1) ||
1145 !*c.host)
1146 ret = -1;
1147 credential_clear(&c);
1148 return ret;
1149 }
1150
1151 return 0;
1152 }
1153
1154 struct fsck_gitmodules_data {
1155 const struct object_id *oid;
1156 struct fsck_options *options;
1157 int ret;
1158 };
1159
1160 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1161 {
1162 struct fsck_gitmodules_data *data = vdata;
1163 const char *subsection, *key;
1164 size_t subsection_len;
1165 char *name;
1166
1167 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1168 !subsection)
1169 return 0;
1170
1171 name = xmemdupz(subsection, subsection_len);
1172 if (check_submodule_name(name) < 0)
1173 data->ret |= report(data->options,
1174 data->oid, OBJ_BLOB,
1175 FSCK_MSG_GITMODULES_NAME,
1176 "disallowed submodule name: %s",
1177 name);
1178 if (!strcmp(key, "url") && value &&
1179 check_submodule_url(value) < 0)
1180 data->ret |= report(data->options,
1181 data->oid, OBJ_BLOB,
1182 FSCK_MSG_GITMODULES_URL,
1183 "disallowed submodule url: %s",
1184 value);
1185 if (!strcmp(key, "path") && value &&
1186 looks_like_command_line_option(value))
1187 data->ret |= report(data->options,
1188 data->oid, OBJ_BLOB,
1189 FSCK_MSG_GITMODULES_PATH,
1190 "disallowed submodule path: %s",
1191 value);
1192 if (!strcmp(key, "update") && value &&
1193 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1194 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1195 FSCK_MSG_GITMODULES_UPDATE,
1196 "disallowed submodule update setting: %s",
1197 value);
1198 free(name);
1199
1200 return 0;
1201 }
1202
1203 static int fsck_blob(const struct object_id *oid, const char *buf,
1204 unsigned long size, struct fsck_options *options)
1205 {
1206 struct fsck_gitmodules_data data;
1207 struct config_options config_opts = { 0 };
1208
1209 if (!oidset_contains(&gitmodules_found, oid))
1210 return 0;
1211 oidset_insert(&gitmodules_done, oid);
1212
1213 if (object_on_skiplist(options, oid))
1214 return 0;
1215
1216 if (!buf) {
1217 /*
1218 * A missing buffer here is a sign that the caller found the
1219 * blob too gigantic to load into memory. Let's just consider
1220 * that an error.
1221 */
1222 return report(options, oid, OBJ_BLOB,
1223 FSCK_MSG_GITMODULES_LARGE,
1224 ".gitmodules too large to parse");
1225 }
1226
1227 data.oid = oid;
1228 data.options = options;
1229 data.ret = 0;
1230 config_opts.error_action = CONFIG_ERROR_SILENT;
1231 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1232 ".gitmodules", buf, size, &data, &config_opts))
1233 data.ret |= report(options, oid, OBJ_BLOB,
1234 FSCK_MSG_GITMODULES_PARSE,
1235 "could not parse gitmodules blob");
1236
1237 return data.ret;
1238 }
1239
1240 int fsck_object(struct object *obj, void *data, unsigned long size,
1241 struct fsck_options *options)
1242 {
1243 if (!obj)
1244 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1245
1246 if (obj->type == OBJ_BLOB)
1247 return fsck_blob(&obj->oid, data, size, options);
1248 if (obj->type == OBJ_TREE)
1249 return fsck_tree(&obj->oid, data, size, options);
1250 if (obj->type == OBJ_COMMIT)
1251 return fsck_commit(&obj->oid, data, size, options);
1252 if (obj->type == OBJ_TAG)
1253 return fsck_tag(&obj->oid, data, size, options);
1254
1255 return report(options, &obj->oid, obj->type,
1256 FSCK_MSG_UNKNOWN_TYPE,
1257 "unknown type '%d' (internal fsck error)",
1258 obj->type);
1259 }
1260
1261 int fsck_error_function(struct fsck_options *o,
1262 const struct object_id *oid,
1263 enum object_type object_type,
1264 int msg_type, const char *message)
1265 {
1266 if (msg_type == FSCK_WARN) {
1267 warning("object %s: %s", fsck_describe_object(o, oid), message);
1268 return 0;
1269 }
1270 error("object %s: %s", fsck_describe_object(o, oid), message);
1271 return 1;
1272 }
1273
1274 int fsck_finish(struct fsck_options *options)
1275 {
1276 int ret = 0;
1277 struct oidset_iter iter;
1278 const struct object_id *oid;
1279
1280 oidset_iter_init(&gitmodules_found, &iter);
1281 while ((oid = oidset_iter_next(&iter))) {
1282 enum object_type type;
1283 unsigned long size;
1284 char *buf;
1285
1286 if (oidset_contains(&gitmodules_done, oid))
1287 continue;
1288
1289 buf = read_object_file(oid, &type, &size);
1290 if (!buf) {
1291 if (is_promisor_object(oid))
1292 continue;
1293 ret |= report(options,
1294 oid, OBJ_BLOB,
1295 FSCK_MSG_GITMODULES_MISSING,
1296 "unable to read .gitmodules blob");
1297 continue;
1298 }
1299
1300 if (type == OBJ_BLOB)
1301 ret |= fsck_blob(oid, buf, size, options);
1302 else
1303 ret |= report(options,
1304 oid, type,
1305 FSCK_MSG_GITMODULES_BLOB,
1306 "non-blob found at .gitmodules");
1307 free(buf);
1308 }
1309
1310
1311 oidset_clear(&gitmodules_found);
1312 oidset_clear(&gitmodules_done);
1313 return ret;
1314 }
1315
1316 int fsck_config_internal(const char *var, const char *value, void *cb,
1317 struct fsck_options *options)
1318 {
1319 if (strcmp(var, "fsck.skiplist") == 0) {
1320 const char *path;
1321 struct strbuf sb = STRBUF_INIT;
1322
1323 if (git_config_pathname(&path, var, value))
1324 return 1;
1325 strbuf_addf(&sb, "skiplist=%s", path);
1326 free((char *)path);
1327 fsck_set_msg_types(options, sb.buf);
1328 strbuf_release(&sb);
1329 return 0;
1330 }
1331
1332 if (skip_prefix(var, "fsck.", &var)) {
1333 fsck_set_msg_type(options, var, value);
1334 return 0;
1335 }
1336
1337 return git_default_config(var, value, cb);
1338 }