]> git.ipfire.org Git - thirdparty/git.git/blob - fsck.c
fsck.c: rename remaining fsck_msg_id "id" to "msg_id"
[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_str, const char *msg_type_str)
207 {
208 int msg_id = parse_msg_id(msg_id_str), msg_type;
209
210 if (msg_id < 0)
211 die("Unhandled message id: %s", msg_id_str);
212 msg_type = parse_msg_type(msg_type_str);
213
214 if (msg_type != FSCK_ERROR && msg_id_info[msg_id].msg_type == FSCK_FATAL)
215 die("Cannot demote %s to %s", msg_id_str, msg_type_str);
216
217 if (!options->msg_type) {
218 int i;
219 int *severity;
220 ALLOC_ARRAY(severity, FSCK_MSG_MAX);
221 for (i = 0; i < FSCK_MSG_MAX; i++)
222 severity[i] = fsck_msg_type(i, options);
223 options->msg_type = severity;
224 }
225
226 options->msg_type[msg_id] = msg_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 int object_on_skiplist(struct fsck_options *opts,
268 const struct object_id *oid)
269 {
270 return opts && oid && oidset_contains(&opts->skiplist, oid);
271 }
272
273 __attribute__((format (printf, 5, 6)))
274 static int report(struct fsck_options *options,
275 const struct object_id *oid, enum object_type object_type,
276 enum fsck_msg_id msg_id, const char *fmt, ...)
277 {
278 va_list ap;
279 struct strbuf sb = STRBUF_INIT;
280 int msg_type = fsck_msg_type(msg_id, options), result;
281
282 if (msg_type == FSCK_IGNORE)
283 return 0;
284
285 if (object_on_skiplist(options, oid))
286 return 0;
287
288 if (msg_type == FSCK_FATAL)
289 msg_type = FSCK_ERROR;
290 else if (msg_type == FSCK_INFO)
291 msg_type = FSCK_WARN;
292
293 prepare_msg_ids();
294 strbuf_addf(&sb, "%s: ", msg_id_info[msg_id].camelcased);
295
296 va_start(ap, fmt);
297 strbuf_vaddf(&sb, fmt, ap);
298 result = options->error_func(options, oid, object_type,
299 msg_type, sb.buf);
300 strbuf_release(&sb);
301 va_end(ap);
302
303 return result;
304 }
305
306 void fsck_enable_object_names(struct fsck_options *options)
307 {
308 if (!options->object_names)
309 options->object_names = kh_init_oid_map();
310 }
311
312 const char *fsck_get_object_name(struct fsck_options *options,
313 const struct object_id *oid)
314 {
315 khiter_t pos;
316 if (!options->object_names)
317 return NULL;
318 pos = kh_get_oid_map(options->object_names, *oid);
319 if (pos >= kh_end(options->object_names))
320 return NULL;
321 return kh_value(options->object_names, pos);
322 }
323
324 void fsck_put_object_name(struct fsck_options *options,
325 const struct object_id *oid,
326 const char *fmt, ...)
327 {
328 va_list ap;
329 struct strbuf buf = STRBUF_INIT;
330 khiter_t pos;
331 int hashret;
332
333 if (!options->object_names)
334 return;
335
336 pos = kh_put_oid_map(options->object_names, *oid, &hashret);
337 if (!hashret)
338 return;
339 va_start(ap, fmt);
340 strbuf_vaddf(&buf, fmt, ap);
341 kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
342 va_end(ap);
343 }
344
345 const char *fsck_describe_object(struct fsck_options *options,
346 const struct object_id *oid)
347 {
348 static struct strbuf bufs[] = {
349 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
350 };
351 static int b = 0;
352 struct strbuf *buf;
353 const char *name = fsck_get_object_name(options, oid);
354
355 buf = bufs + b;
356 b = (b + 1) % ARRAY_SIZE(bufs);
357 strbuf_reset(buf);
358 strbuf_addstr(buf, oid_to_hex(oid));
359 if (name)
360 strbuf_addf(buf, " (%s)", name);
361
362 return buf->buf;
363 }
364
365 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
366 {
367 struct tree_desc desc;
368 struct name_entry entry;
369 int res = 0;
370 const char *name;
371
372 if (parse_tree(tree))
373 return -1;
374
375 name = fsck_get_object_name(options, &tree->object.oid);
376 if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
377 return -1;
378 while (tree_entry_gently(&desc, &entry)) {
379 struct object *obj;
380 int result;
381
382 if (S_ISGITLINK(entry.mode))
383 continue;
384
385 if (S_ISDIR(entry.mode)) {
386 obj = (struct object *)lookup_tree(the_repository, &entry.oid);
387 if (name && obj)
388 fsck_put_object_name(options, &entry.oid, "%s%s/",
389 name, entry.path);
390 result = options->walk(obj, OBJ_TREE, data, options);
391 }
392 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
393 obj = (struct object *)lookup_blob(the_repository, &entry.oid);
394 if (name && obj)
395 fsck_put_object_name(options, &entry.oid, "%s%s",
396 name, entry.path);
397 result = options->walk(obj, OBJ_BLOB, data, options);
398 }
399 else {
400 result = error("in tree %s: entry %s has bad mode %.6o",
401 fsck_describe_object(options, &tree->object.oid),
402 entry.path, entry.mode);
403 }
404 if (result < 0)
405 return result;
406 if (!res)
407 res = result;
408 }
409 return res;
410 }
411
412 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
413 {
414 int counter = 0, generation = 0, name_prefix_len = 0;
415 struct commit_list *parents;
416 int res;
417 int result;
418 const char *name;
419
420 if (parse_commit(commit))
421 return -1;
422
423 name = fsck_get_object_name(options, &commit->object.oid);
424 if (name)
425 fsck_put_object_name(options, get_commit_tree_oid(commit),
426 "%s:", name);
427
428 result = options->walk((struct object *)get_commit_tree(commit),
429 OBJ_TREE, data, options);
430 if (result < 0)
431 return result;
432 res = result;
433
434 parents = commit->parents;
435 if (name && parents) {
436 int len = strlen(name), power;
437
438 if (len && name[len - 1] == '^') {
439 generation = 1;
440 name_prefix_len = len - 1;
441 }
442 else { /* parse ~<generation> suffix */
443 for (generation = 0, power = 1;
444 len && isdigit(name[len - 1]);
445 power *= 10)
446 generation += power * (name[--len] - '0');
447 if (power > 1 && len && name[len - 1] == '~')
448 name_prefix_len = len - 1;
449 else {
450 /* Maybe a non-first parent, e.g. HEAD^2 */
451 generation = 0;
452 name_prefix_len = len;
453 }
454 }
455 }
456
457 while (parents) {
458 if (name) {
459 struct object_id *oid = &parents->item->object.oid;
460
461 if (counter++)
462 fsck_put_object_name(options, oid, "%s^%d",
463 name, counter);
464 else if (generation > 0)
465 fsck_put_object_name(options, oid, "%.*s~%d",
466 name_prefix_len, name,
467 generation + 1);
468 else
469 fsck_put_object_name(options, oid, "%s^", name);
470 }
471 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
472 if (result < 0)
473 return result;
474 if (!res)
475 res = result;
476 parents = parents->next;
477 }
478 return res;
479 }
480
481 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
482 {
483 const char *name = fsck_get_object_name(options, &tag->object.oid);
484
485 if (parse_tag(tag))
486 return -1;
487 if (name)
488 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
489 return options->walk(tag->tagged, OBJ_ANY, data, options);
490 }
491
492 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
493 {
494 if (!obj)
495 return -1;
496
497 if (obj->type == OBJ_NONE)
498 parse_object(the_repository, &obj->oid);
499
500 switch (obj->type) {
501 case OBJ_BLOB:
502 return 0;
503 case OBJ_TREE:
504 return fsck_walk_tree((struct tree *)obj, data, options);
505 case OBJ_COMMIT:
506 return fsck_walk_commit((struct commit *)obj, data, options);
507 case OBJ_TAG:
508 return fsck_walk_tag((struct tag *)obj, data, options);
509 default:
510 error("Unknown object type for %s",
511 fsck_describe_object(options, &obj->oid));
512 return -1;
513 }
514 }
515
516 struct name_stack {
517 const char **names;
518 size_t nr, alloc;
519 };
520
521 static void name_stack_push(struct name_stack *stack, const char *name)
522 {
523 ALLOC_GROW(stack->names, stack->nr + 1, stack->alloc);
524 stack->names[stack->nr++] = name;
525 }
526
527 static const char *name_stack_pop(struct name_stack *stack)
528 {
529 return stack->nr ? stack->names[--stack->nr] : NULL;
530 }
531
532 static void name_stack_clear(struct name_stack *stack)
533 {
534 FREE_AND_NULL(stack->names);
535 stack->nr = stack->alloc = 0;
536 }
537
538 /*
539 * The entries in a tree are ordered in the _path_ order,
540 * which means that a directory entry is ordered by adding
541 * a slash to the end of it.
542 *
543 * So a directory called "a" is ordered _after_ a file
544 * called "a.c", because "a/" sorts after "a.c".
545 */
546 #define TREE_UNORDERED (-1)
547 #define TREE_HAS_DUPS (-2)
548
549 static int is_less_than_slash(unsigned char c)
550 {
551 return '\0' < c && c < '/';
552 }
553
554 static int verify_ordered(unsigned mode1, const char *name1,
555 unsigned mode2, const char *name2,
556 struct name_stack *candidates)
557 {
558 int len1 = strlen(name1);
559 int len2 = strlen(name2);
560 int len = len1 < len2 ? len1 : len2;
561 unsigned char c1, c2;
562 int cmp;
563
564 cmp = memcmp(name1, name2, len);
565 if (cmp < 0)
566 return 0;
567 if (cmp > 0)
568 return TREE_UNORDERED;
569
570 /*
571 * Ok, the first <len> characters are the same.
572 * Now we need to order the next one, but turn
573 * a '\0' into a '/' for a directory entry.
574 */
575 c1 = name1[len];
576 c2 = name2[len];
577 if (!c1 && !c2)
578 /*
579 * git-write-tree used to write out a nonsense tree that has
580 * entries with the same name, one blob and one tree. Make
581 * sure we do not have duplicate entries.
582 */
583 return TREE_HAS_DUPS;
584 if (!c1 && S_ISDIR(mode1))
585 c1 = '/';
586 if (!c2 && S_ISDIR(mode2))
587 c2 = '/';
588
589 /*
590 * There can be non-consecutive duplicates due to the implicitly
591 * added slash, e.g.:
592 *
593 * foo
594 * foo.bar
595 * foo.bar.baz
596 * foo.bar/
597 * foo/
598 *
599 * Record non-directory candidates (like "foo" and "foo.bar" in
600 * the example) on a stack and check directory candidates (like
601 * foo/" and "foo.bar/") against that stack.
602 */
603 if (!c1 && is_less_than_slash(c2)) {
604 name_stack_push(candidates, name1);
605 } else if (c2 == '/' && is_less_than_slash(c1)) {
606 for (;;) {
607 const char *p;
608 const char *f_name = name_stack_pop(candidates);
609
610 if (!f_name)
611 break;
612 if (!skip_prefix(name2, f_name, &p))
613 continue;
614 if (!*p)
615 return TREE_HAS_DUPS;
616 if (is_less_than_slash(*p)) {
617 name_stack_push(candidates, f_name);
618 break;
619 }
620 }
621 }
622
623 return c1 < c2 ? 0 : TREE_UNORDERED;
624 }
625
626 static int fsck_tree(const struct object_id *oid,
627 const char *buffer, unsigned long size,
628 struct fsck_options *options)
629 {
630 int retval = 0;
631 int has_null_sha1 = 0;
632 int has_full_path = 0;
633 int has_empty_name = 0;
634 int has_dot = 0;
635 int has_dotdot = 0;
636 int has_dotgit = 0;
637 int has_zero_pad = 0;
638 int has_bad_modes = 0;
639 int has_dup_entries = 0;
640 int not_properly_sorted = 0;
641 struct tree_desc desc;
642 unsigned o_mode;
643 const char *o_name;
644 struct name_stack df_dup_candidates = { NULL };
645
646 if (init_tree_desc_gently(&desc, buffer, size)) {
647 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
648 return retval;
649 }
650
651 o_mode = 0;
652 o_name = NULL;
653
654 while (desc.size) {
655 unsigned short mode;
656 const char *name, *backslash;
657 const struct object_id *oid;
658
659 oid = tree_entry_extract(&desc, &name, &mode);
660
661 has_null_sha1 |= is_null_oid(oid);
662 has_full_path |= !!strchr(name, '/');
663 has_empty_name |= !*name;
664 has_dot |= !strcmp(name, ".");
665 has_dotdot |= !strcmp(name, "..");
666 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
667 has_zero_pad |= *(char *)desc.buffer == '0';
668
669 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
670 if (!S_ISLNK(mode))
671 oidset_insert(&gitmodules_found, oid);
672 else
673 retval += report(options,
674 oid, OBJ_TREE,
675 FSCK_MSG_GITMODULES_SYMLINK,
676 ".gitmodules is a symbolic link");
677 }
678
679 if ((backslash = strchr(name, '\\'))) {
680 while (backslash) {
681 backslash++;
682 has_dotgit |= is_ntfs_dotgit(backslash);
683 if (is_ntfs_dotgitmodules(backslash)) {
684 if (!S_ISLNK(mode))
685 oidset_insert(&gitmodules_found, oid);
686 else
687 retval += report(options, oid, OBJ_TREE,
688 FSCK_MSG_GITMODULES_SYMLINK,
689 ".gitmodules is a symbolic link");
690 }
691 backslash = strchr(backslash, '\\');
692 }
693 }
694
695 if (update_tree_entry_gently(&desc)) {
696 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
697 break;
698 }
699
700 switch (mode) {
701 /*
702 * Standard modes..
703 */
704 case S_IFREG | 0755:
705 case S_IFREG | 0644:
706 case S_IFLNK:
707 case S_IFDIR:
708 case S_IFGITLINK:
709 break;
710 /*
711 * This is nonstandard, but we had a few of these
712 * early on when we honored the full set of mode
713 * bits..
714 */
715 case S_IFREG | 0664:
716 if (!options->strict)
717 break;
718 /* fallthrough */
719 default:
720 has_bad_modes = 1;
721 }
722
723 if (o_name) {
724 switch (verify_ordered(o_mode, o_name, mode, name,
725 &df_dup_candidates)) {
726 case TREE_UNORDERED:
727 not_properly_sorted = 1;
728 break;
729 case TREE_HAS_DUPS:
730 has_dup_entries = 1;
731 break;
732 default:
733 break;
734 }
735 }
736
737 o_mode = mode;
738 o_name = name;
739 }
740
741 name_stack_clear(&df_dup_candidates);
742
743 if (has_null_sha1)
744 retval += report(options, oid, OBJ_TREE, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
745 if (has_full_path)
746 retval += report(options, oid, OBJ_TREE, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
747 if (has_empty_name)
748 retval += report(options, oid, OBJ_TREE, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
749 if (has_dot)
750 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOT, "contains '.'");
751 if (has_dotdot)
752 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTDOT, "contains '..'");
753 if (has_dotgit)
754 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
755 if (has_zero_pad)
756 retval += report(options, oid, OBJ_TREE, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
757 if (has_bad_modes)
758 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
759 if (has_dup_entries)
760 retval += report(options, oid, OBJ_TREE, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
761 if (not_properly_sorted)
762 retval += report(options, oid, OBJ_TREE, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
763 return retval;
764 }
765
766 static int verify_headers(const void *data, unsigned long size,
767 const struct object_id *oid, enum object_type type,
768 struct fsck_options *options)
769 {
770 const char *buffer = (const char *)data;
771 unsigned long i;
772
773 for (i = 0; i < size; i++) {
774 switch (buffer[i]) {
775 case '\0':
776 return report(options, oid, type,
777 FSCK_MSG_NUL_IN_HEADER,
778 "unterminated header: NUL at offset %ld", i);
779 case '\n':
780 if (i + 1 < size && buffer[i + 1] == '\n')
781 return 0;
782 }
783 }
784
785 /*
786 * We did not find double-LF that separates the header
787 * and the body. Not having a body is not a crime but
788 * we do want to see the terminating LF for the last header
789 * line.
790 */
791 if (size && buffer[size - 1] == '\n')
792 return 0;
793
794 return report(options, oid, type,
795 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
796 }
797
798 static int fsck_ident(const char **ident,
799 const struct object_id *oid, enum object_type type,
800 struct fsck_options *options)
801 {
802 const char *p = *ident;
803 char *end;
804
805 *ident = strchrnul(*ident, '\n');
806 if (**ident == '\n')
807 (*ident)++;
808
809 if (*p == '<')
810 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
811 p += strcspn(p, "<>\n");
812 if (*p == '>')
813 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
814 if (*p != '<')
815 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
816 if (p[-1] != ' ')
817 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
818 p++;
819 p += strcspn(p, "<>\n");
820 if (*p != '>')
821 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
822 p++;
823 if (*p != ' ')
824 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
825 p++;
826 if (*p == '0' && p[1] != ' ')
827 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
828 if (date_overflows(parse_timestamp(p, &end, 10)))
829 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
830 if ((end == p || *end != ' '))
831 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
832 p = end + 1;
833 if ((*p != '+' && *p != '-') ||
834 !isdigit(p[1]) ||
835 !isdigit(p[2]) ||
836 !isdigit(p[3]) ||
837 !isdigit(p[4]) ||
838 (p[5] != '\n'))
839 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
840 p += 6;
841 return 0;
842 }
843
844 static int fsck_commit(const struct object_id *oid,
845 const char *buffer, unsigned long size,
846 struct fsck_options *options)
847 {
848 struct object_id tree_oid, parent_oid;
849 unsigned author_count;
850 int err;
851 const char *buffer_begin = buffer;
852 const char *p;
853
854 if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
855 return -1;
856
857 if (!skip_prefix(buffer, "tree ", &buffer))
858 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
859 if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
860 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
861 if (err)
862 return err;
863 }
864 buffer = p + 1;
865 while (skip_prefix(buffer, "parent ", &buffer)) {
866 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
867 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
868 if (err)
869 return err;
870 }
871 buffer = p + 1;
872 }
873 author_count = 0;
874 while (skip_prefix(buffer, "author ", &buffer)) {
875 author_count++;
876 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
877 if (err)
878 return err;
879 }
880 if (author_count < 1)
881 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
882 else if (author_count > 1)
883 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
884 if (err)
885 return err;
886 if (!skip_prefix(buffer, "committer ", &buffer))
887 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
888 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
889 if (err)
890 return err;
891 if (memchr(buffer_begin, '\0', size)) {
892 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
893 "NUL byte in the commit object body");
894 if (err)
895 return err;
896 }
897 return 0;
898 }
899
900 static int fsck_tag(const struct object_id *oid, const char *buffer,
901 unsigned long size, struct fsck_options *options)
902 {
903 struct object_id tagged_oid;
904 int tagged_type;
905 return fsck_tag_standalone(oid, buffer, size, options, &tagged_oid,
906 &tagged_type);
907 }
908
909 int fsck_tag_standalone(const struct object_id *oid, const char *buffer,
910 unsigned long size, struct fsck_options *options,
911 struct object_id *tagged_oid,
912 int *tagged_type)
913 {
914 int ret = 0;
915 char *eol;
916 struct strbuf sb = STRBUF_INIT;
917 const char *p;
918
919 ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
920 if (ret)
921 goto done;
922
923 if (!skip_prefix(buffer, "object ", &buffer)) {
924 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
925 goto done;
926 }
927 if (parse_oid_hex(buffer, tagged_oid, &p) || *p != '\n') {
928 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
929 if (ret)
930 goto done;
931 }
932 buffer = p + 1;
933
934 if (!skip_prefix(buffer, "type ", &buffer)) {
935 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
936 goto done;
937 }
938 eol = strchr(buffer, '\n');
939 if (!eol) {
940 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
941 goto done;
942 }
943 *tagged_type = type_from_string_gently(buffer, eol - buffer, 1);
944 if (*tagged_type < 0)
945 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
946 if (ret)
947 goto done;
948 buffer = eol + 1;
949
950 if (!skip_prefix(buffer, "tag ", &buffer)) {
951 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
952 goto done;
953 }
954 eol = strchr(buffer, '\n');
955 if (!eol) {
956 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
957 goto done;
958 }
959 strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
960 if (check_refname_format(sb.buf, 0)) {
961 ret = report(options, oid, OBJ_TAG,
962 FSCK_MSG_BAD_TAG_NAME,
963 "invalid 'tag' name: %.*s",
964 (int)(eol - buffer), buffer);
965 if (ret)
966 goto done;
967 }
968 buffer = eol + 1;
969
970 if (!skip_prefix(buffer, "tagger ", &buffer)) {
971 /* early tags do not contain 'tagger' lines; warn only */
972 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
973 if (ret)
974 goto done;
975 }
976 else
977 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
978 if (!*buffer)
979 goto done;
980
981 if (!starts_with(buffer, "\n")) {
982 /*
983 * The verify_headers() check will allow
984 * e.g. "[...]tagger <tagger>\nsome
985 * garbage\n\nmessage" to pass, thinking "some
986 * garbage" could be a custom header. E.g. "mktag"
987 * doesn't want any unknown headers.
988 */
989 ret = report(options, oid, OBJ_TAG, FSCK_MSG_EXTRA_HEADER_ENTRY, "invalid format - extra header(s) after 'tagger'");
990 if (ret)
991 goto done;
992 }
993
994 done:
995 strbuf_release(&sb);
996 return ret;
997 }
998
999 /*
1000 * Like builtin/submodule--helper.c's starts_with_dot_slash, but without
1001 * relying on the platform-dependent is_dir_sep helper.
1002 *
1003 * This is for use in checking whether a submodule URL is interpreted as
1004 * relative to the current directory on any platform, since \ is a
1005 * directory separator on Windows but not on other platforms.
1006 */
1007 static int starts_with_dot_slash(const char *str)
1008 {
1009 return str[0] == '.' && (str[1] == '/' || str[1] == '\\');
1010 }
1011
1012 /*
1013 * Like starts_with_dot_slash, this is a variant of submodule--helper's
1014 * helper of the same name with the twist that it accepts backslash as a
1015 * directory separator even on non-Windows platforms.
1016 */
1017 static int starts_with_dot_dot_slash(const char *str)
1018 {
1019 return str[0] == '.' && starts_with_dot_slash(str + 1);
1020 }
1021
1022 static int submodule_url_is_relative(const char *url)
1023 {
1024 return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
1025 }
1026
1027 /*
1028 * Count directory components that a relative submodule URL should chop
1029 * from the remote_url it is to be resolved against.
1030 *
1031 * In other words, this counts "../" components at the start of a
1032 * submodule URL.
1033 *
1034 * Returns the number of directory components to chop and writes a
1035 * pointer to the next character of url after all leading "./" and
1036 * "../" components to out.
1037 */
1038 static int count_leading_dotdots(const char *url, const char **out)
1039 {
1040 int result = 0;
1041 while (1) {
1042 if (starts_with_dot_dot_slash(url)) {
1043 result++;
1044 url += strlen("../");
1045 continue;
1046 }
1047 if (starts_with_dot_slash(url)) {
1048 url += strlen("./");
1049 continue;
1050 }
1051 *out = url;
1052 return result;
1053 }
1054 }
1055 /*
1056 * Check whether a transport is implemented by git-remote-curl.
1057 *
1058 * If it is, returns 1 and writes the URL that would be passed to
1059 * git-remote-curl to the "out" parameter.
1060 *
1061 * Otherwise, returns 0 and leaves "out" untouched.
1062 *
1063 * Examples:
1064 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1065 * https://example.com/repo.git -> 1, https://example.com/repo.git
1066 * git://example.com/repo.git -> 0
1067 *
1068 * This is for use in checking for previously exploitable bugs that
1069 * required a submodule URL to be passed to git-remote-curl.
1070 */
1071 static int url_to_curl_url(const char *url, const char **out)
1072 {
1073 /*
1074 * We don't need to check for case-aliases, "http.exe", and so
1075 * on because in the default configuration, is_transport_allowed
1076 * prevents URLs with those schemes from being cloned
1077 * automatically.
1078 */
1079 if (skip_prefix(url, "http::", out) ||
1080 skip_prefix(url, "https::", out) ||
1081 skip_prefix(url, "ftp::", out) ||
1082 skip_prefix(url, "ftps::", out))
1083 return 1;
1084 if (starts_with(url, "http://") ||
1085 starts_with(url, "https://") ||
1086 starts_with(url, "ftp://") ||
1087 starts_with(url, "ftps://")) {
1088 *out = url;
1089 return 1;
1090 }
1091 return 0;
1092 }
1093
1094 static int check_submodule_url(const char *url)
1095 {
1096 const char *curl_url;
1097
1098 if (looks_like_command_line_option(url))
1099 return -1;
1100
1101 if (submodule_url_is_relative(url) || starts_with(url, "git://")) {
1102 char *decoded;
1103 const char *next;
1104 int has_nl;
1105
1106 /*
1107 * This could be appended to an http URL and url-decoded;
1108 * check for malicious characters.
1109 */
1110 decoded = url_decode(url);
1111 has_nl = !!strchr(decoded, '\n');
1112
1113 free(decoded);
1114 if (has_nl)
1115 return -1;
1116
1117 /*
1118 * URLs which escape their root via "../" can overwrite
1119 * the host field and previous components, resolving to
1120 * URLs like https::example.com/submodule.git and
1121 * https:///example.com/submodule.git that were
1122 * susceptible to CVE-2020-11008.
1123 */
1124 if (count_leading_dotdots(url, &next) > 0 &&
1125 (*next == ':' || *next == '/'))
1126 return -1;
1127 }
1128
1129 else if (url_to_curl_url(url, &curl_url)) {
1130 struct credential c = CREDENTIAL_INIT;
1131 int ret = 0;
1132 if (credential_from_url_gently(&c, curl_url, 1) ||
1133 !*c.host)
1134 ret = -1;
1135 credential_clear(&c);
1136 return ret;
1137 }
1138
1139 return 0;
1140 }
1141
1142 struct fsck_gitmodules_data {
1143 const struct object_id *oid;
1144 struct fsck_options *options;
1145 int ret;
1146 };
1147
1148 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1149 {
1150 struct fsck_gitmodules_data *data = vdata;
1151 const char *subsection, *key;
1152 size_t subsection_len;
1153 char *name;
1154
1155 if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1156 !subsection)
1157 return 0;
1158
1159 name = xmemdupz(subsection, subsection_len);
1160 if (check_submodule_name(name) < 0)
1161 data->ret |= report(data->options,
1162 data->oid, OBJ_BLOB,
1163 FSCK_MSG_GITMODULES_NAME,
1164 "disallowed submodule name: %s",
1165 name);
1166 if (!strcmp(key, "url") && value &&
1167 check_submodule_url(value) < 0)
1168 data->ret |= report(data->options,
1169 data->oid, OBJ_BLOB,
1170 FSCK_MSG_GITMODULES_URL,
1171 "disallowed submodule url: %s",
1172 value);
1173 if (!strcmp(key, "path") && value &&
1174 looks_like_command_line_option(value))
1175 data->ret |= report(data->options,
1176 data->oid, OBJ_BLOB,
1177 FSCK_MSG_GITMODULES_PATH,
1178 "disallowed submodule path: %s",
1179 value);
1180 if (!strcmp(key, "update") && value &&
1181 parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1182 data->ret |= report(data->options, data->oid, OBJ_BLOB,
1183 FSCK_MSG_GITMODULES_UPDATE,
1184 "disallowed submodule update setting: %s",
1185 value);
1186 free(name);
1187
1188 return 0;
1189 }
1190
1191 static int fsck_blob(const struct object_id *oid, const char *buf,
1192 unsigned long size, struct fsck_options *options)
1193 {
1194 struct fsck_gitmodules_data data;
1195 struct config_options config_opts = { 0 };
1196
1197 if (!oidset_contains(&gitmodules_found, oid))
1198 return 0;
1199 oidset_insert(&gitmodules_done, oid);
1200
1201 if (object_on_skiplist(options, oid))
1202 return 0;
1203
1204 if (!buf) {
1205 /*
1206 * A missing buffer here is a sign that the caller found the
1207 * blob too gigantic to load into memory. Let's just consider
1208 * that an error.
1209 */
1210 return report(options, oid, OBJ_BLOB,
1211 FSCK_MSG_GITMODULES_LARGE,
1212 ".gitmodules too large to parse");
1213 }
1214
1215 data.oid = oid;
1216 data.options = options;
1217 data.ret = 0;
1218 config_opts.error_action = CONFIG_ERROR_SILENT;
1219 if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1220 ".gitmodules", buf, size, &data, &config_opts))
1221 data.ret |= report(options, oid, OBJ_BLOB,
1222 FSCK_MSG_GITMODULES_PARSE,
1223 "could not parse gitmodules blob");
1224
1225 return data.ret;
1226 }
1227
1228 int fsck_object(struct object *obj, void *data, unsigned long size,
1229 struct fsck_options *options)
1230 {
1231 if (!obj)
1232 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1233
1234 if (obj->type == OBJ_BLOB)
1235 return fsck_blob(&obj->oid, data, size, options);
1236 if (obj->type == OBJ_TREE)
1237 return fsck_tree(&obj->oid, data, size, options);
1238 if (obj->type == OBJ_COMMIT)
1239 return fsck_commit(&obj->oid, data, size, options);
1240 if (obj->type == OBJ_TAG)
1241 return fsck_tag(&obj->oid, data, size, options);
1242
1243 return report(options, &obj->oid, obj->type,
1244 FSCK_MSG_UNKNOWN_TYPE,
1245 "unknown type '%d' (internal fsck error)",
1246 obj->type);
1247 }
1248
1249 int fsck_error_function(struct fsck_options *o,
1250 const struct object_id *oid,
1251 enum object_type object_type,
1252 int msg_type, const char *message)
1253 {
1254 if (msg_type == FSCK_WARN) {
1255 warning("object %s: %s", fsck_describe_object(o, oid), message);
1256 return 0;
1257 }
1258 error("object %s: %s", fsck_describe_object(o, oid), message);
1259 return 1;
1260 }
1261
1262 void register_found_gitmodules(const struct object_id *oid)
1263 {
1264 oidset_insert(&gitmodules_found, oid);
1265 }
1266
1267 int fsck_finish(struct fsck_options *options)
1268 {
1269 int ret = 0;
1270 struct oidset_iter iter;
1271 const struct object_id *oid;
1272
1273 oidset_iter_init(&gitmodules_found, &iter);
1274 while ((oid = oidset_iter_next(&iter))) {
1275 enum object_type type;
1276 unsigned long size;
1277 char *buf;
1278
1279 if (oidset_contains(&gitmodules_done, oid))
1280 continue;
1281
1282 buf = read_object_file(oid, &type, &size);
1283 if (!buf) {
1284 if (is_promisor_object(oid))
1285 continue;
1286 ret |= report(options,
1287 oid, OBJ_BLOB,
1288 FSCK_MSG_GITMODULES_MISSING,
1289 "unable to read .gitmodules blob");
1290 continue;
1291 }
1292
1293 if (type == OBJ_BLOB)
1294 ret |= fsck_blob(oid, buf, size, options);
1295 else
1296 ret |= report(options,
1297 oid, type,
1298 FSCK_MSG_GITMODULES_BLOB,
1299 "non-blob found at .gitmodules");
1300 free(buf);
1301 }
1302
1303
1304 oidset_clear(&gitmodules_found);
1305 oidset_clear(&gitmodules_done);
1306 return ret;
1307 }
1308
1309 int git_fsck_config(const char *var, const char *value, void *cb)
1310 {
1311 struct fsck_options *options = cb;
1312 if (strcmp(var, "fsck.skiplist") == 0) {
1313 const char *path;
1314 struct strbuf sb = STRBUF_INIT;
1315
1316 if (git_config_pathname(&path, var, value))
1317 return 1;
1318 strbuf_addf(&sb, "skiplist=%s", path);
1319 free((char *)path);
1320 fsck_set_msg_types(options, sb.buf);
1321 strbuf_release(&sb);
1322 return 0;
1323 }
1324
1325 if (skip_prefix(var, "fsck.", &var)) {
1326 fsck_set_msg_type(options, var, value);
1327 return 0;
1328 }
1329
1330 return git_default_config(var, value, cb);
1331 }