Imported Upstream version 2.23.3
[platform/upstream/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_GRAFT, ERROR) \
49         FUNC(MISSING_NAME_BEFORE_EMAIL, ERROR) \
50         FUNC(MISSING_OBJECT, ERROR) \
51         FUNC(MISSING_PARENT, ERROR) \
52         FUNC(MISSING_SPACE_BEFORE_DATE, ERROR) \
53         FUNC(MISSING_SPACE_BEFORE_EMAIL, ERROR) \
54         FUNC(MISSING_TAG, ERROR) \
55         FUNC(MISSING_TAG_ENTRY, ERROR) \
56         FUNC(MISSING_TAG_OBJECT, ERROR) \
57         FUNC(MISSING_TREE, ERROR) \
58         FUNC(MISSING_TREE_OBJECT, ERROR) \
59         FUNC(MISSING_TYPE, ERROR) \
60         FUNC(MISSING_TYPE_ENTRY, ERROR) \
61         FUNC(MULTIPLE_AUTHORS, ERROR) \
62         FUNC(TAG_OBJECT_NOT_TAG, ERROR) \
63         FUNC(TREE_NOT_SORTED, ERROR) \
64         FUNC(UNKNOWN_TYPE, ERROR) \
65         FUNC(ZERO_PADDED_DATE, ERROR) \
66         FUNC(GITMODULES_MISSING, ERROR) \
67         FUNC(GITMODULES_BLOB, ERROR) \
68         FUNC(GITMODULES_LARGE, ERROR) \
69         FUNC(GITMODULES_NAME, ERROR) \
70         FUNC(GITMODULES_SYMLINK, ERROR) \
71         FUNC(GITMODULES_URL, ERROR) \
72         FUNC(GITMODULES_PATH, ERROR) \
73         FUNC(GITMODULES_UPDATE, ERROR) \
74         /* warnings */ \
75         FUNC(BAD_FILEMODE, WARN) \
76         FUNC(EMPTY_NAME, WARN) \
77         FUNC(FULL_PATHNAME, WARN) \
78         FUNC(HAS_DOT, WARN) \
79         FUNC(HAS_DOTDOT, WARN) \
80         FUNC(HAS_DOTGIT, WARN) \
81         FUNC(NULL_SHA1, WARN) \
82         FUNC(ZERO_PADDED_FILEMODE, WARN) \
83         FUNC(NUL_IN_COMMIT, WARN) \
84         /* infos (reported as warnings, but ignored by default) */ \
85         FUNC(GITMODULES_PARSE, INFO) \
86         FUNC(BAD_TAG_NAME, INFO) \
87         FUNC(MISSING_TAGGER_ENTRY, INFO)
88
89 #define MSG_ID(id, msg_type) FSCK_MSG_##id,
90 enum fsck_msg_id {
91         FOREACH_MSG_ID(MSG_ID)
92         FSCK_MSG_MAX
93 };
94 #undef MSG_ID
95
96 #define STR(x) #x
97 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
98 static struct {
99         const char *id_string;
100         const char *downcased;
101         const char *camelcased;
102         int msg_type;
103 } msg_id_info[FSCK_MSG_MAX + 1] = {
104         FOREACH_MSG_ID(MSG_ID)
105         { NULL, NULL, NULL, -1 }
106 };
107 #undef MSG_ID
108
109 static void prepare_msg_ids(void)
110 {
111         int i;
112
113         if (msg_id_info[0].downcased)
114                 return;
115
116         /* convert id_string to lower case, without underscores. */
117         for (i = 0; i < FSCK_MSG_MAX; i++) {
118                 const char *p = msg_id_info[i].id_string;
119                 int len = strlen(p);
120                 char *q = xmalloc(len);
121
122                 msg_id_info[i].downcased = q;
123                 while (*p)
124                         if (*p == '_')
125                                 p++;
126                         else
127                                 *(q)++ = tolower(*(p)++);
128                 *q = '\0';
129
130                 p = msg_id_info[i].id_string;
131                 q = xmalloc(len);
132                 msg_id_info[i].camelcased = q;
133                 while (*p) {
134                         if (*p == '_') {
135                                 p++;
136                                 if (*p)
137                                         *q++ = *p++;
138                         } else {
139                                 *q++ = tolower(*p++);
140                         }
141                 }
142                 *q = '\0';
143         }
144 }
145
146 static int parse_msg_id(const char *text)
147 {
148         int i;
149
150         prepare_msg_ids();
151
152         for (i = 0; i < FSCK_MSG_MAX; i++)
153                 if (!strcmp(text, msg_id_info[i].downcased))
154                         return i;
155
156         return -1;
157 }
158
159 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
160 {
161         int i;
162
163         prepare_msg_ids();
164
165         for (i = 0; i < FSCK_MSG_MAX; i++)
166                 list_config_item(list, prefix, msg_id_info[i].camelcased);
167 }
168
169 static int fsck_msg_type(enum fsck_msg_id msg_id,
170         struct fsck_options *options)
171 {
172         int msg_type;
173
174         assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
175
176         if (options->msg_type)
177                 msg_type = options->msg_type[msg_id];
178         else {
179                 msg_type = msg_id_info[msg_id].msg_type;
180                 if (options->strict && msg_type == FSCK_WARN)
181                         msg_type = FSCK_ERROR;
182         }
183
184         return msg_type;
185 }
186
187 static int parse_msg_type(const char *str)
188 {
189         if (!strcmp(str, "error"))
190                 return FSCK_ERROR;
191         else if (!strcmp(str, "warn"))
192                 return FSCK_WARN;
193         else if (!strcmp(str, "ignore"))
194                 return FSCK_IGNORE;
195         else
196                 die("Unknown fsck message type: '%s'", str);
197 }
198
199 int is_valid_msg_type(const char *msg_id, const char *msg_type)
200 {
201         if (parse_msg_id(msg_id) < 0)
202                 return 0;
203         parse_msg_type(msg_type);
204         return 1;
205 }
206
207 void fsck_set_msg_type(struct fsck_options *options,
208                 const char *msg_id, const char *msg_type)
209 {
210         int id = parse_msg_id(msg_id), type;
211
212         if (id < 0)
213                 die("Unhandled message id: %s", msg_id);
214         type = parse_msg_type(msg_type);
215
216         if (type != FSCK_ERROR && msg_id_info[id].msg_type == FSCK_FATAL)
217                 die("Cannot demote %s to %s", msg_id, msg_type);
218
219         if (!options->msg_type) {
220                 int i;
221                 int *msg_type;
222                 ALLOC_ARRAY(msg_type, FSCK_MSG_MAX);
223                 for (i = 0; i < FSCK_MSG_MAX; i++)
224                         msg_type[i] = fsck_msg_type(i, options);
225                 options->msg_type = msg_type;
226         }
227
228         options->msg_type[id] = type;
229 }
230
231 void fsck_set_msg_types(struct fsck_options *options, const char *values)
232 {
233         char *buf = xstrdup(values), *to_free = buf;
234         int done = 0;
235
236         while (!done) {
237                 int len = strcspn(buf, " ,|"), equal;
238
239                 done = !buf[len];
240                 if (!len) {
241                         buf++;
242                         continue;
243                 }
244                 buf[len] = '\0';
245
246                 for (equal = 0;
247                      equal < len && buf[equal] != '=' && buf[equal] != ':';
248                      equal++)
249                         buf[equal] = tolower(buf[equal]);
250                 buf[equal] = '\0';
251
252                 if (!strcmp(buf, "skiplist")) {
253                         if (equal == len)
254                                 die("skiplist requires a path");
255                         oidset_parse_file(&options->skiplist, buf + equal + 1);
256                         buf += len + 1;
257                         continue;
258                 }
259
260                 if (equal == len)
261                         die("Missing '=': '%s'", buf);
262
263                 fsck_set_msg_type(options, buf, buf + equal + 1);
264                 buf += len + 1;
265         }
266         free(to_free);
267 }
268
269 static void append_msg_id(struct strbuf *sb, const char *msg_id)
270 {
271         for (;;) {
272                 char c = *(msg_id)++;
273
274                 if (!c)
275                         break;
276                 if (c != '_')
277                         strbuf_addch(sb, tolower(c));
278                 else {
279                         assert(*msg_id);
280                         strbuf_addch(sb, *(msg_id)++);
281                 }
282         }
283
284         strbuf_addstr(sb, ": ");
285 }
286
287 static int object_on_skiplist(struct fsck_options *opts, struct object *obj)
288 {
289         return opts && obj && oidset_contains(&opts->skiplist, &obj->oid);
290 }
291
292 __attribute__((format (printf, 4, 5)))
293 static int report(struct fsck_options *options, struct object *object,
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, object))
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, object, msg_type, sb.buf);
316         strbuf_release(&sb);
317         va_end(ap);
318
319         return result;
320 }
321
322 static char *get_object_name(struct fsck_options *options, struct object *obj)
323 {
324         if (!options->object_names)
325                 return NULL;
326         return lookup_decoration(options->object_names, obj);
327 }
328
329 static void put_object_name(struct fsck_options *options, struct object *obj,
330         const char *fmt, ...)
331 {
332         va_list ap;
333         struct strbuf buf = STRBUF_INIT;
334         char *existing;
335
336         if (!options->object_names)
337                 return;
338         existing = lookup_decoration(options->object_names, obj);
339         if (existing)
340                 return;
341         va_start(ap, fmt);
342         strbuf_vaddf(&buf, fmt, ap);
343         add_decoration(options->object_names, obj, strbuf_detach(&buf, NULL));
344         va_end(ap);
345 }
346
347 static const char *describe_object(struct fsck_options *o, struct object *obj)
348 {
349         static struct strbuf buf = STRBUF_INIT;
350         char *name;
351
352         strbuf_reset(&buf);
353         strbuf_addstr(&buf, oid_to_hex(&obj->oid));
354         if (o->object_names && (name = lookup_decoration(o->object_names, obj)))
355                 strbuf_addf(&buf, " (%s)", name);
356
357         return buf.buf;
358 }
359
360 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
361 {
362         struct tree_desc desc;
363         struct name_entry entry;
364         int res = 0;
365         const char *name;
366
367         if (parse_tree(tree))
368                 return -1;
369
370         name = get_object_name(options, &tree->object);
371         if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
372                 return -1;
373         while (tree_entry_gently(&desc, &entry)) {
374                 struct object *obj;
375                 int result;
376
377                 if (S_ISGITLINK(entry.mode))
378                         continue;
379
380                 if (S_ISDIR(entry.mode)) {
381                         obj = (struct object *)lookup_tree(the_repository, &entry.oid);
382                         if (name && obj)
383                                 put_object_name(options, obj, "%s%s/", name,
384                                         entry.path);
385                         result = options->walk(obj, OBJ_TREE, data, options);
386                 }
387                 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
388                         obj = (struct object *)lookup_blob(the_repository, &entry.oid);
389                         if (name && obj)
390                                 put_object_name(options, obj, "%s%s", name,
391                                         entry.path);
392                         result = options->walk(obj, OBJ_BLOB, data, options);
393                 }
394                 else {
395                         result = error("in tree %s: entry %s has bad mode %.6o",
396                                         describe_object(options, &tree->object), entry.path, entry.mode);
397                 }
398                 if (result < 0)
399                         return result;
400                 if (!res)
401                         res = result;
402         }
403         return res;
404 }
405
406 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
407 {
408         int counter = 0, generation = 0, name_prefix_len = 0;
409         struct commit_list *parents;
410         int res;
411         int result;
412         const char *name;
413
414         if (parse_commit(commit))
415                 return -1;
416
417         name = get_object_name(options, &commit->object);
418         if (name)
419                 put_object_name(options, &get_commit_tree(commit)->object,
420                                 "%s:", name);
421
422         result = options->walk((struct object *)get_commit_tree(commit),
423                                OBJ_TREE, data, options);
424         if (result < 0)
425                 return result;
426         res = result;
427
428         parents = commit->parents;
429         if (name && parents) {
430                 int len = strlen(name), power;
431
432                 if (len && name[len - 1] == '^') {
433                         generation = 1;
434                         name_prefix_len = len - 1;
435                 }
436                 else { /* parse ~<generation> suffix */
437                         for (generation = 0, power = 1;
438                              len && isdigit(name[len - 1]);
439                              power *= 10)
440                                 generation += power * (name[--len] - '0');
441                         if (power > 1 && len && name[len - 1] == '~')
442                                 name_prefix_len = len - 1;
443                 }
444         }
445
446         while (parents) {
447                 if (name) {
448                         struct object *obj = &parents->item->object;
449
450                         if (counter++)
451                                 put_object_name(options, obj, "%s^%d",
452                                         name, counter);
453                         else if (generation > 0)
454                                 put_object_name(options, obj, "%.*s~%d",
455                                         name_prefix_len, name, generation + 1);
456                         else
457                                 put_object_name(options, obj, "%s^", name);
458                 }
459                 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
460                 if (result < 0)
461                         return result;
462                 if (!res)
463                         res = result;
464                 parents = parents->next;
465         }
466         return res;
467 }
468
469 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
470 {
471         char *name = get_object_name(options, &tag->object);
472
473         if (parse_tag(tag))
474                 return -1;
475         if (name)
476                 put_object_name(options, tag->tagged, "%s", name);
477         return options->walk(tag->tagged, OBJ_ANY, data, options);
478 }
479
480 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
481 {
482         if (!obj)
483                 return -1;
484
485         if (obj->type == OBJ_NONE)
486                 parse_object(the_repository, &obj->oid);
487
488         switch (obj->type) {
489         case OBJ_BLOB:
490                 return 0;
491         case OBJ_TREE:
492                 return fsck_walk_tree((struct tree *)obj, data, options);
493         case OBJ_COMMIT:
494                 return fsck_walk_commit((struct commit *)obj, data, options);
495         case OBJ_TAG:
496                 return fsck_walk_tag((struct tag *)obj, data, options);
497         default:
498                 error("Unknown object type for %s", describe_object(options, obj));
499                 return -1;
500         }
501 }
502
503 /*
504  * The entries in a tree are ordered in the _path_ order,
505  * which means that a directory entry is ordered by adding
506  * a slash to the end of it.
507  *
508  * So a directory called "a" is ordered _after_ a file
509  * called "a.c", because "a/" sorts after "a.c".
510  */
511 #define TREE_UNORDERED (-1)
512 #define TREE_HAS_DUPS  (-2)
513
514 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
515 {
516         int len1 = strlen(name1);
517         int len2 = strlen(name2);
518         int len = len1 < len2 ? len1 : len2;
519         unsigned char c1, c2;
520         int cmp;
521
522         cmp = memcmp(name1, name2, len);
523         if (cmp < 0)
524                 return 0;
525         if (cmp > 0)
526                 return TREE_UNORDERED;
527
528         /*
529          * Ok, the first <len> characters are the same.
530          * Now we need to order the next one, but turn
531          * a '\0' into a '/' for a directory entry.
532          */
533         c1 = name1[len];
534         c2 = name2[len];
535         if (!c1 && !c2)
536                 /*
537                  * git-write-tree used to write out a nonsense tree that has
538                  * entries with the same name, one blob and one tree.  Make
539                  * sure we do not have duplicate entries.
540                  */
541                 return TREE_HAS_DUPS;
542         if (!c1 && S_ISDIR(mode1))
543                 c1 = '/';
544         if (!c2 && S_ISDIR(mode2))
545                 c2 = '/';
546         return c1 < c2 ? 0 : TREE_UNORDERED;
547 }
548
549 static int fsck_tree(struct tree *item, struct fsck_options *options)
550 {
551         int retval = 0;
552         int has_null_sha1 = 0;
553         int has_full_path = 0;
554         int has_empty_name = 0;
555         int has_dot = 0;
556         int has_dotdot = 0;
557         int has_dotgit = 0;
558         int has_zero_pad = 0;
559         int has_bad_modes = 0;
560         int has_dup_entries = 0;
561         int not_properly_sorted = 0;
562         struct tree_desc desc;
563         unsigned o_mode;
564         const char *o_name;
565
566         if (init_tree_desc_gently(&desc, item->buffer, item->size)) {
567                 retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
568                 return retval;
569         }
570
571         o_mode = 0;
572         o_name = NULL;
573
574         while (desc.size) {
575                 unsigned short mode;
576                 const char *name, *backslash;
577                 const struct object_id *oid;
578
579                 oid = tree_entry_extract(&desc, &name, &mode);
580
581                 has_null_sha1 |= is_null_oid(oid);
582                 has_full_path |= !!strchr(name, '/');
583                 has_empty_name |= !*name;
584                 has_dot |= !strcmp(name, ".");
585                 has_dotdot |= !strcmp(name, "..");
586                 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
587                 has_zero_pad |= *(char *)desc.buffer == '0';
588
589                 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
590                         if (!S_ISLNK(mode))
591                                 oidset_insert(&gitmodules_found, oid);
592                         else
593                                 retval += report(options, &item->object,
594                                                  FSCK_MSG_GITMODULES_SYMLINK,
595                                                  ".gitmodules is a symbolic link");
596                 }
597
598                 if ((backslash = strchr(name, '\\'))) {
599                         while (backslash) {
600                                 backslash++;
601                                 has_dotgit |= is_ntfs_dotgit(backslash);
602                                 if (is_ntfs_dotgitmodules(backslash)) {
603                                         if (!S_ISLNK(mode))
604                                                 oidset_insert(&gitmodules_found, oid);
605                                         else
606                                                 retval += report(options, &item->object,
607                                                                  FSCK_MSG_GITMODULES_SYMLINK,
608                                                                  ".gitmodules is a symbolic link");
609                                 }
610                                 backslash = strchr(backslash, '\\');
611                         }
612                 }
613
614                 if (update_tree_entry_gently(&desc)) {
615                         retval += report(options, &item->object, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
616                         break;
617                 }
618
619                 switch (mode) {
620                 /*
621                  * Standard modes..
622                  */
623                 case S_IFREG | 0755:
624                 case S_IFREG | 0644:
625                 case S_IFLNK:
626                 case S_IFDIR:
627                 case S_IFGITLINK:
628                         break;
629                 /*
630                  * This is nonstandard, but we had a few of these
631                  * early on when we honored the full set of mode
632                  * bits..
633                  */
634                 case S_IFREG | 0664:
635                         if (!options->strict)
636                                 break;
637                         /* fallthrough */
638                 default:
639                         has_bad_modes = 1;
640                 }
641
642                 if (o_name) {
643                         switch (verify_ordered(o_mode, o_name, mode, name)) {
644                         case TREE_UNORDERED:
645                                 not_properly_sorted = 1;
646                                 break;
647                         case TREE_HAS_DUPS:
648                                 has_dup_entries = 1;
649                                 break;
650                         default:
651                                 break;
652                         }
653                 }
654
655                 o_mode = mode;
656                 o_name = name;
657         }
658
659         if (has_null_sha1)
660                 retval += report(options, &item->object, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
661         if (has_full_path)
662                 retval += report(options, &item->object, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
663         if (has_empty_name)
664                 retval += report(options, &item->object, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
665         if (has_dot)
666                 retval += report(options, &item->object, FSCK_MSG_HAS_DOT, "contains '.'");
667         if (has_dotdot)
668                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTDOT, "contains '..'");
669         if (has_dotgit)
670                 retval += report(options, &item->object, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
671         if (has_zero_pad)
672                 retval += report(options, &item->object, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
673         if (has_bad_modes)
674                 retval += report(options, &item->object, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
675         if (has_dup_entries)
676                 retval += report(options, &item->object, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
677         if (not_properly_sorted)
678                 retval += report(options, &item->object, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
679         return retval;
680 }
681
682 static int verify_headers(const void *data, unsigned long size,
683                           struct object *obj, struct fsck_options *options)
684 {
685         const char *buffer = (const char *)data;
686         unsigned long i;
687
688         for (i = 0; i < size; i++) {
689                 switch (buffer[i]) {
690                 case '\0':
691                         return report(options, obj,
692                                 FSCK_MSG_NUL_IN_HEADER,
693                                 "unterminated header: NUL at offset %ld", i);
694                 case '\n':
695                         if (i + 1 < size && buffer[i + 1] == '\n')
696                                 return 0;
697                 }
698         }
699
700         /*
701          * We did not find double-LF that separates the header
702          * and the body.  Not having a body is not a crime but
703          * we do want to see the terminating LF for the last header
704          * line.
705          */
706         if (size && buffer[size - 1] == '\n')
707                 return 0;
708
709         return report(options, obj,
710                 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
711 }
712
713 static int fsck_ident(const char **ident, struct object *obj, struct fsck_options *options)
714 {
715         const char *p = *ident;
716         char *end;
717
718         *ident = strchrnul(*ident, '\n');
719         if (**ident == '\n')
720                 (*ident)++;
721
722         if (*p == '<')
723                 return report(options, obj, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
724         p += strcspn(p, "<>\n");
725         if (*p == '>')
726                 return report(options, obj, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
727         if (*p != '<')
728                 return report(options, obj, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
729         if (p[-1] != ' ')
730                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
731         p++;
732         p += strcspn(p, "<>\n");
733         if (*p != '>')
734                 return report(options, obj, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
735         p++;
736         if (*p != ' ')
737                 return report(options, obj, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
738         p++;
739         if (*p == '0' && p[1] != ' ')
740                 return report(options, obj, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
741         if (date_overflows(parse_timestamp(p, &end, 10)))
742                 return report(options, obj, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
743         if ((end == p || *end != ' '))
744                 return report(options, obj, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
745         p = end + 1;
746         if ((*p != '+' && *p != '-') ||
747             !isdigit(p[1]) ||
748             !isdigit(p[2]) ||
749             !isdigit(p[3]) ||
750             !isdigit(p[4]) ||
751             (p[5] != '\n'))
752                 return report(options, obj, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
753         p += 6;
754         return 0;
755 }
756
757 static int fsck_commit_buffer(struct commit *commit, const char *buffer,
758         unsigned long size, struct fsck_options *options)
759 {
760         struct object_id tree_oid, oid;
761         struct commit_graft *graft;
762         unsigned parent_count, parent_line_count = 0, author_count;
763         int err;
764         const char *buffer_begin = buffer;
765         const char *p;
766
767         if (verify_headers(buffer, size, &commit->object, options))
768                 return -1;
769
770         if (!skip_prefix(buffer, "tree ", &buffer))
771                 return report(options, &commit->object, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
772         if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
773                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
774                 if (err)
775                         return err;
776         }
777         buffer = p + 1;
778         while (skip_prefix(buffer, "parent ", &buffer)) {
779                 if (parse_oid_hex(buffer, &oid, &p) || *p != '\n') {
780                         err = report(options, &commit->object, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
781                         if (err)
782                                 return err;
783                 }
784                 buffer = p + 1;
785                 parent_line_count++;
786         }
787         graft = lookup_commit_graft(the_repository, &commit->object.oid);
788         parent_count = commit_list_count(commit->parents);
789         if (graft) {
790                 if (graft->nr_parent == -1 && !parent_count)
791                         ; /* shallow commit */
792                 else if (graft->nr_parent != parent_count) {
793                         err = report(options, &commit->object, FSCK_MSG_MISSING_GRAFT, "graft objects missing");
794                         if (err)
795                                 return err;
796                 }
797         } else {
798                 if (parent_count != parent_line_count) {
799                         err = report(options, &commit->object, FSCK_MSG_MISSING_PARENT, "parent objects missing");
800                         if (err)
801                                 return err;
802                 }
803         }
804         author_count = 0;
805         while (skip_prefix(buffer, "author ", &buffer)) {
806                 author_count++;
807                 err = fsck_ident(&buffer, &commit->object, options);
808                 if (err)
809                         return err;
810         }
811         if (author_count < 1)
812                 err = report(options, &commit->object, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
813         else if (author_count > 1)
814                 err = report(options, &commit->object, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
815         if (err)
816                 return err;
817         if (!skip_prefix(buffer, "committer ", &buffer))
818                 return report(options, &commit->object, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
819         err = fsck_ident(&buffer, &commit->object, options);
820         if (err)
821                 return err;
822         if (!get_commit_tree(commit)) {
823                 err = report(options, &commit->object, FSCK_MSG_BAD_TREE, "could not load commit's tree %s", oid_to_hex(&tree_oid));
824                 if (err)
825                         return err;
826         }
827         if (memchr(buffer_begin, '\0', size)) {
828                 err = report(options, &commit->object, FSCK_MSG_NUL_IN_COMMIT,
829                              "NUL byte in the commit object body");
830                 if (err)
831                         return err;
832         }
833         return 0;
834 }
835
836 static int fsck_commit(struct commit *commit, const char *data,
837         unsigned long size, struct fsck_options *options)
838 {
839         const char *buffer = data ?  data : get_commit_buffer(commit, &size);
840         int ret = fsck_commit_buffer(commit, buffer, size, options);
841         if (!data)
842                 unuse_commit_buffer(commit, buffer);
843         return ret;
844 }
845
846 static int fsck_tag_buffer(struct tag *tag, const char *data,
847         unsigned long size, struct fsck_options *options)
848 {
849         struct object_id oid;
850         int ret = 0;
851         const char *buffer;
852         char *to_free = NULL, *eol;
853         struct strbuf sb = STRBUF_INIT;
854         const char *p;
855
856         if (data)
857                 buffer = data;
858         else {
859                 enum object_type type;
860
861                 buffer = to_free =
862                         read_object_file(&tag->object.oid, &type, &size);
863                 if (!buffer)
864                         return report(options, &tag->object,
865                                 FSCK_MSG_MISSING_TAG_OBJECT,
866                                 "cannot read tag object");
867
868                 if (type != OBJ_TAG) {
869                         ret = report(options, &tag->object,
870                                 FSCK_MSG_TAG_OBJECT_NOT_TAG,
871                                 "expected tag got %s",
872                             type_name(type));
873                         goto done;
874                 }
875         }
876
877         ret = verify_headers(buffer, size, &tag->object, options);
878         if (ret)
879                 goto done;
880
881         if (!skip_prefix(buffer, "object ", &buffer)) {
882                 ret = report(options, &tag->object, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
883                 goto done;
884         }
885         if (parse_oid_hex(buffer, &oid, &p) || *p != '\n') {
886                 ret = report(options, &tag->object, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
887                 if (ret)
888                         goto done;
889         }
890         buffer = p + 1;
891
892         if (!skip_prefix(buffer, "type ", &buffer)) {
893                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
894                 goto done;
895         }
896         eol = strchr(buffer, '\n');
897         if (!eol) {
898                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
899                 goto done;
900         }
901         if (type_from_string_gently(buffer, eol - buffer, 1) < 0)
902                 ret = report(options, &tag->object, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
903         if (ret)
904                 goto done;
905         buffer = eol + 1;
906
907         if (!skip_prefix(buffer, "tag ", &buffer)) {
908                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
909                 goto done;
910         }
911         eol = strchr(buffer, '\n');
912         if (!eol) {
913                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
914                 goto done;
915         }
916         strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
917         if (check_refname_format(sb.buf, 0)) {
918                 ret = report(options, &tag->object, FSCK_MSG_BAD_TAG_NAME,
919                            "invalid 'tag' name: %.*s",
920                            (int)(eol - buffer), buffer);
921                 if (ret)
922                         goto done;
923         }
924         buffer = eol + 1;
925
926         if (!skip_prefix(buffer, "tagger ", &buffer)) {
927                 /* early tags do not contain 'tagger' lines; warn only */
928                 ret = report(options, &tag->object, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
929                 if (ret)
930                         goto done;
931         }
932         else
933                 ret = fsck_ident(&buffer, &tag->object, options);
934
935 done:
936         strbuf_release(&sb);
937         free(to_free);
938         return ret;
939 }
940
941 static int fsck_tag(struct tag *tag, const char *data,
942         unsigned long size, struct fsck_options *options)
943 {
944         struct object *tagged = tag->tagged;
945
946         if (!tagged)
947                 return report(options, &tag->object, FSCK_MSG_BAD_TAG_OBJECT, "could not load tagged object");
948
949         return fsck_tag_buffer(tag, data, size, options);
950 }
951
952 /*
953  * Like builtin/submodule--helper.c's starts_with_dot_slash, but without
954  * relying on the platform-dependent is_dir_sep helper.
955  *
956  * This is for use in checking whether a submodule URL is interpreted as
957  * relative to the current directory on any platform, since \ is a
958  * directory separator on Windows but not on other platforms.
959  */
960 static int starts_with_dot_slash(const char *str)
961 {
962         return str[0] == '.' && (str[1] == '/' || str[1] == '\\');
963 }
964
965 /*
966  * Like starts_with_dot_slash, this is a variant of submodule--helper's
967  * helper of the same name with the twist that it accepts backslash as a
968  * directory separator even on non-Windows platforms.
969  */
970 static int starts_with_dot_dot_slash(const char *str)
971 {
972         return str[0] == '.' && starts_with_dot_slash(str + 1);
973 }
974
975 static int submodule_url_is_relative(const char *url)
976 {
977         return starts_with_dot_slash(url) || starts_with_dot_dot_slash(url);
978 }
979
980 /*
981  * Count directory components that a relative submodule URL should chop
982  * from the remote_url it is to be resolved against.
983  *
984  * In other words, this counts "../" components at the start of a
985  * submodule URL.
986  *
987  * Returns the number of directory components to chop and writes a
988  * pointer to the next character of url after all leading "./" and
989  * "../" components to out.
990  */
991 static int count_leading_dotdots(const char *url, const char **out)
992 {
993         int result = 0;
994         while (1) {
995                 if (starts_with_dot_dot_slash(url)) {
996                         result++;
997                         url += strlen("../");
998                         continue;
999                 }
1000                 if (starts_with_dot_slash(url)) {
1001                         url += strlen("./");
1002                         continue;
1003                 }
1004                 *out = url;
1005                 return result;
1006         }
1007 }
1008 /*
1009  * Check whether a transport is implemented by git-remote-curl.
1010  *
1011  * If it is, returns 1 and writes the URL that would be passed to
1012  * git-remote-curl to the "out" parameter.
1013  *
1014  * Otherwise, returns 0 and leaves "out" untouched.
1015  *
1016  * Examples:
1017  *   http::https://example.com/repo.git -> 1, https://example.com/repo.git
1018  *   https://example.com/repo.git -> 1, https://example.com/repo.git
1019  *   git://example.com/repo.git -> 0
1020  *
1021  * This is for use in checking for previously exploitable bugs that
1022  * required a submodule URL to be passed to git-remote-curl.
1023  */
1024 static int url_to_curl_url(const char *url, const char **out)
1025 {
1026         /*
1027          * We don't need to check for case-aliases, "http.exe", and so
1028          * on because in the default configuration, is_transport_allowed
1029          * prevents URLs with those schemes from being cloned
1030          * automatically.
1031          */
1032         if (skip_prefix(url, "http::", out) ||
1033             skip_prefix(url, "https::", out) ||
1034             skip_prefix(url, "ftp::", out) ||
1035             skip_prefix(url, "ftps::", out))
1036                 return 1;
1037         if (starts_with(url, "http://") ||
1038             starts_with(url, "https://") ||
1039             starts_with(url, "ftp://") ||
1040             starts_with(url, "ftps://")) {
1041                 *out = url;
1042                 return 1;
1043         }
1044         return 0;
1045 }
1046
1047 static int check_submodule_url(const char *url)
1048 {
1049         const char *curl_url;
1050
1051         if (looks_like_command_line_option(url))
1052                 return -1;
1053
1054         if (submodule_url_is_relative(url)) {
1055                 char *decoded;
1056                 const char *next;
1057                 int has_nl;
1058
1059                 /*
1060                  * This could be appended to an http URL and url-decoded;
1061                  * check for malicious characters.
1062                  */
1063                 decoded = url_decode(url);
1064                 has_nl = !!strchr(decoded, '\n');
1065
1066                 free(decoded);
1067                 if (has_nl)
1068                         return -1;
1069
1070                 /*
1071                  * URLs which escape their root via "../" can overwrite
1072                  * the host field and previous components, resolving to
1073                  * URLs like https::example.com/submodule.git and
1074                  * https:///example.com/submodule.git that were
1075                  * susceptible to CVE-2020-11008.
1076                  */
1077                 if (count_leading_dotdots(url, &next) > 0 &&
1078                     (*next == ':' || *next == '/'))
1079                         return -1;
1080         }
1081
1082         else if (url_to_curl_url(url, &curl_url)) {
1083                 struct credential c = CREDENTIAL_INIT;
1084                 int ret = 0;
1085                 if (credential_from_url_gently(&c, curl_url, 1) ||
1086                     !*c.host)
1087                         ret = -1;
1088                 credential_clear(&c);
1089                 return ret;
1090         }
1091
1092         return 0;
1093 }
1094
1095 struct fsck_gitmodules_data {
1096         struct object *obj;
1097         struct fsck_options *options;
1098         int ret;
1099 };
1100
1101 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
1102 {
1103         struct fsck_gitmodules_data *data = vdata;
1104         const char *subsection, *key;
1105         int subsection_len;
1106         char *name;
1107
1108         if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
1109             !subsection)
1110                 return 0;
1111
1112         name = xmemdupz(subsection, subsection_len);
1113         if (check_submodule_name(name) < 0)
1114                 data->ret |= report(data->options, data->obj,
1115                                     FSCK_MSG_GITMODULES_NAME,
1116                                     "disallowed submodule name: %s",
1117                                     name);
1118         if (!strcmp(key, "url") && value &&
1119             check_submodule_url(value) < 0)
1120                 data->ret |= report(data->options, data->obj,
1121                                     FSCK_MSG_GITMODULES_URL,
1122                                     "disallowed submodule url: %s",
1123                                     value);
1124         if (!strcmp(key, "path") && value &&
1125             looks_like_command_line_option(value))
1126                 data->ret |= report(data->options, data->obj,
1127                                     FSCK_MSG_GITMODULES_PATH,
1128                                     "disallowed submodule path: %s",
1129                                     value);
1130         if (!strcmp(key, "update") && value &&
1131             parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
1132                 data->ret |= report(data->options, data->obj,
1133                                     FSCK_MSG_GITMODULES_UPDATE,
1134                                     "disallowed submodule update setting: %s",
1135                                     value);
1136         free(name);
1137
1138         return 0;
1139 }
1140
1141 static int fsck_blob(struct blob *blob, const char *buf,
1142                      unsigned long size, struct fsck_options *options)
1143 {
1144         struct fsck_gitmodules_data data;
1145         struct config_options config_opts = { 0 };
1146
1147         if (!oidset_contains(&gitmodules_found, &blob->object.oid))
1148                 return 0;
1149         oidset_insert(&gitmodules_done, &blob->object.oid);
1150
1151         if (object_on_skiplist(options, &blob->object))
1152                 return 0;
1153
1154         if (!buf) {
1155                 /*
1156                  * A missing buffer here is a sign that the caller found the
1157                  * blob too gigantic to load into memory. Let's just consider
1158                  * that an error.
1159                  */
1160                 return report(options, &blob->object,
1161                               FSCK_MSG_GITMODULES_LARGE,
1162                               ".gitmodules too large to parse");
1163         }
1164
1165         data.obj = &blob->object;
1166         data.options = options;
1167         data.ret = 0;
1168         config_opts.error_action = CONFIG_ERROR_SILENT;
1169         if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
1170                                 ".gitmodules", buf, size, &data, &config_opts))
1171                 data.ret |= report(options, &blob->object,
1172                                    FSCK_MSG_GITMODULES_PARSE,
1173                                    "could not parse gitmodules blob");
1174
1175         return data.ret;
1176 }
1177
1178 int fsck_object(struct object *obj, void *data, unsigned long size,
1179         struct fsck_options *options)
1180 {
1181         if (!obj)
1182                 return report(options, obj, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1183
1184         if (obj->type == OBJ_BLOB)
1185                 return fsck_blob((struct blob *)obj, data, size, options);
1186         if (obj->type == OBJ_TREE)
1187                 return fsck_tree((struct tree *) obj, options);
1188         if (obj->type == OBJ_COMMIT)
1189                 return fsck_commit((struct commit *) obj, (const char *) data,
1190                         size, options);
1191         if (obj->type == OBJ_TAG)
1192                 return fsck_tag((struct tag *) obj, (const char *) data,
1193                         size, options);
1194
1195         return report(options, obj, FSCK_MSG_UNKNOWN_TYPE, "unknown type '%d' (internal fsck error)",
1196                           obj->type);
1197 }
1198
1199 int fsck_error_function(struct fsck_options *o,
1200         struct object *obj, int msg_type, const char *message)
1201 {
1202         if (msg_type == FSCK_WARN) {
1203                 warning("object %s: %s", describe_object(o, obj), message);
1204                 return 0;
1205         }
1206         error("object %s: %s", describe_object(o, obj), message);
1207         return 1;
1208 }
1209
1210 int fsck_finish(struct fsck_options *options)
1211 {
1212         int ret = 0;
1213         struct oidset_iter iter;
1214         const struct object_id *oid;
1215
1216         oidset_iter_init(&gitmodules_found, &iter);
1217         while ((oid = oidset_iter_next(&iter))) {
1218                 struct blob *blob;
1219                 enum object_type type;
1220                 unsigned long size;
1221                 char *buf;
1222
1223                 if (oidset_contains(&gitmodules_done, oid))
1224                         continue;
1225
1226                 blob = lookup_blob(the_repository, oid);
1227                 if (!blob) {
1228                         struct object *obj = lookup_unknown_object(oid);
1229                         ret |= report(options, obj,
1230                                       FSCK_MSG_GITMODULES_BLOB,
1231                                       "non-blob found at .gitmodules");
1232                         continue;
1233                 }
1234
1235                 buf = read_object_file(oid, &type, &size);
1236                 if (!buf) {
1237                         if (is_promisor_object(&blob->object.oid))
1238                                 continue;
1239                         ret |= report(options, &blob->object,
1240                                       FSCK_MSG_GITMODULES_MISSING,
1241                                       "unable to read .gitmodules blob");
1242                         continue;
1243                 }
1244
1245                 if (type == OBJ_BLOB)
1246                         ret |= fsck_blob(blob, buf, size, options);
1247                 else
1248                         ret |= report(options, &blob->object,
1249                                       FSCK_MSG_GITMODULES_BLOB,
1250                                       "non-blob found at .gitmodules");
1251                 free(buf);
1252         }
1253
1254
1255         oidset_clear(&gitmodules_found);
1256         oidset_clear(&gitmodules_done);
1257         return ret;
1258 }