Merge branch 'em/test-skip-regex-illseq'
[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 "utf8.h"
13 #include "decorate.h"
14 #include "oidset.h"
15 #include "packfile.h"
16 #include "submodule-config.h"
17 #include "config.h"
18 #include "help.h"
19
20 static struct oidset gitmodules_found = OIDSET_INIT;
21 static struct oidset gitmodules_done = OIDSET_INIT;
22
23 #define FSCK_FATAL -1
24 #define FSCK_INFO -2
25
26 #define FOREACH_MSG_ID(FUNC) \
27         /* fatal errors */ \
28         FUNC(NUL_IN_HEADER, FATAL) \
29         FUNC(UNTERMINATED_HEADER, FATAL) \
30         /* errors */ \
31         FUNC(BAD_DATE, ERROR) \
32         FUNC(BAD_DATE_OVERFLOW, ERROR) \
33         FUNC(BAD_EMAIL, ERROR) \
34         FUNC(BAD_NAME, ERROR) \
35         FUNC(BAD_OBJECT_SHA1, ERROR) \
36         FUNC(BAD_PARENT_SHA1, ERROR) \
37         FUNC(BAD_TAG_OBJECT, ERROR) \
38         FUNC(BAD_TIMEZONE, ERROR) \
39         FUNC(BAD_TREE, ERROR) \
40         FUNC(BAD_TREE_SHA1, ERROR) \
41         FUNC(BAD_TYPE, ERROR) \
42         FUNC(DUPLICATE_ENTRIES, ERROR) \
43         FUNC(MISSING_AUTHOR, ERROR) \
44         FUNC(MISSING_COMMITTER, ERROR) \
45         FUNC(MISSING_EMAIL, ERROR) \
46         FUNC(MISSING_NAME_BEFORE_EMAIL, ERROR) \
47         FUNC(MISSING_OBJECT, ERROR) \
48         FUNC(MISSING_SPACE_BEFORE_DATE, ERROR) \
49         FUNC(MISSING_SPACE_BEFORE_EMAIL, ERROR) \
50         FUNC(MISSING_TAG, ERROR) \
51         FUNC(MISSING_TAG_ENTRY, ERROR) \
52         FUNC(MISSING_TREE, ERROR) \
53         FUNC(MISSING_TREE_OBJECT, ERROR) \
54         FUNC(MISSING_TYPE, ERROR) \
55         FUNC(MISSING_TYPE_ENTRY, ERROR) \
56         FUNC(MULTIPLE_AUTHORS, ERROR) \
57         FUNC(TREE_NOT_SORTED, ERROR) \
58         FUNC(UNKNOWN_TYPE, ERROR) \
59         FUNC(ZERO_PADDED_DATE, ERROR) \
60         FUNC(GITMODULES_MISSING, ERROR) \
61         FUNC(GITMODULES_BLOB, ERROR) \
62         FUNC(GITMODULES_LARGE, ERROR) \
63         FUNC(GITMODULES_NAME, ERROR) \
64         FUNC(GITMODULES_SYMLINK, ERROR) \
65         FUNC(GITMODULES_URL, ERROR) \
66         FUNC(GITMODULES_PATH, ERROR) \
67         FUNC(GITMODULES_UPDATE, ERROR) \
68         /* warnings */ \
69         FUNC(BAD_FILEMODE, WARN) \
70         FUNC(EMPTY_NAME, WARN) \
71         FUNC(FULL_PATHNAME, WARN) \
72         FUNC(HAS_DOT, WARN) \
73         FUNC(HAS_DOTDOT, WARN) \
74         FUNC(HAS_DOTGIT, WARN) \
75         FUNC(NULL_SHA1, WARN) \
76         FUNC(ZERO_PADDED_FILEMODE, WARN) \
77         FUNC(NUL_IN_COMMIT, WARN) \
78         /* infos (reported as warnings, but ignored by default) */ \
79         FUNC(GITMODULES_PARSE, INFO) \
80         FUNC(BAD_TAG_NAME, INFO) \
81         FUNC(MISSING_TAGGER_ENTRY, INFO)
82
83 #define MSG_ID(id, msg_type) FSCK_MSG_##id,
84 enum fsck_msg_id {
85         FOREACH_MSG_ID(MSG_ID)
86         FSCK_MSG_MAX
87 };
88 #undef MSG_ID
89
90 #define STR(x) #x
91 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
92 static struct {
93         const char *id_string;
94         const char *downcased;
95         const char *camelcased;
96         int msg_type;
97 } msg_id_info[FSCK_MSG_MAX + 1] = {
98         FOREACH_MSG_ID(MSG_ID)
99         { NULL, NULL, NULL, -1 }
100 };
101 #undef MSG_ID
102
103 static void prepare_msg_ids(void)
104 {
105         int i;
106
107         if (msg_id_info[0].downcased)
108                 return;
109
110         /* convert id_string to lower case, without underscores. */
111         for (i = 0; i < FSCK_MSG_MAX; i++) {
112                 const char *p = msg_id_info[i].id_string;
113                 int len = strlen(p);
114                 char *q = xmalloc(len);
115
116                 msg_id_info[i].downcased = q;
117                 while (*p)
118                         if (*p == '_')
119                                 p++;
120                         else
121                                 *(q)++ = tolower(*(p)++);
122                 *q = '\0';
123
124                 p = msg_id_info[i].id_string;
125                 q = xmalloc(len);
126                 msg_id_info[i].camelcased = q;
127                 while (*p) {
128                         if (*p == '_') {
129                                 p++;
130                                 if (*p)
131                                         *q++ = *p++;
132                         } else {
133                                 *q++ = tolower(*p++);
134                         }
135                 }
136                 *q = '\0';
137         }
138 }
139
140 static int parse_msg_id(const char *text)
141 {
142         int i;
143
144         prepare_msg_ids();
145
146         for (i = 0; i < FSCK_MSG_MAX; i++)
147                 if (!strcmp(text, msg_id_info[i].downcased))
148                         return i;
149
150         return -1;
151 }
152
153 void list_config_fsck_msg_ids(struct string_list *list, const char *prefix)
154 {
155         int i;
156
157         prepare_msg_ids();
158
159         for (i = 0; i < FSCK_MSG_MAX; i++)
160                 list_config_item(list, prefix, msg_id_info[i].camelcased);
161 }
162
163 static int fsck_msg_type(enum fsck_msg_id msg_id,
164         struct fsck_options *options)
165 {
166         int msg_type;
167
168         assert(msg_id >= 0 && msg_id < FSCK_MSG_MAX);
169
170         if (options->msg_type)
171                 msg_type = options->msg_type[msg_id];
172         else {
173                 msg_type = msg_id_info[msg_id].msg_type;
174                 if (options->strict && msg_type == FSCK_WARN)
175                         msg_type = FSCK_ERROR;
176         }
177
178         return msg_type;
179 }
180
181 static int parse_msg_type(const char *str)
182 {
183         if (!strcmp(str, "error"))
184                 return FSCK_ERROR;
185         else if (!strcmp(str, "warn"))
186                 return FSCK_WARN;
187         else if (!strcmp(str, "ignore"))
188                 return FSCK_IGNORE;
189         else
190                 die("Unknown fsck message type: '%s'", str);
191 }
192
193 int is_valid_msg_type(const char *msg_id, const char *msg_type)
194 {
195         if (parse_msg_id(msg_id) < 0)
196                 return 0;
197         parse_msg_type(msg_type);
198         return 1;
199 }
200
201 void fsck_set_msg_type(struct fsck_options *options,
202                 const char *msg_id, const char *msg_type)
203 {
204         int id = parse_msg_id(msg_id), type;
205
206         if (id < 0)
207                 die("Unhandled message id: %s", msg_id);
208         type = parse_msg_type(msg_type);
209
210         if (type != FSCK_ERROR && msg_id_info[id].msg_type == FSCK_FATAL)
211                 die("Cannot demote %s to %s", msg_id, msg_type);
212
213         if (!options->msg_type) {
214                 int i;
215                 int *msg_type;
216                 ALLOC_ARRAY(msg_type, FSCK_MSG_MAX);
217                 for (i = 0; i < FSCK_MSG_MAX; i++)
218                         msg_type[i] = fsck_msg_type(i, options);
219                 options->msg_type = msg_type;
220         }
221
222         options->msg_type[id] = type;
223 }
224
225 void fsck_set_msg_types(struct fsck_options *options, const char *values)
226 {
227         char *buf = xstrdup(values), *to_free = buf;
228         int done = 0;
229
230         while (!done) {
231                 int len = strcspn(buf, " ,|"), equal;
232
233                 done = !buf[len];
234                 if (!len) {
235                         buf++;
236                         continue;
237                 }
238                 buf[len] = '\0';
239
240                 for (equal = 0;
241                      equal < len && buf[equal] != '=' && buf[equal] != ':';
242                      equal++)
243                         buf[equal] = tolower(buf[equal]);
244                 buf[equal] = '\0';
245
246                 if (!strcmp(buf, "skiplist")) {
247                         if (equal == len)
248                                 die("skiplist requires a path");
249                         oidset_parse_file(&options->skiplist, buf + equal + 1);
250                         buf += len + 1;
251                         continue;
252                 }
253
254                 if (equal == len)
255                         die("Missing '=': '%s'", buf);
256
257                 fsck_set_msg_type(options, buf, buf + equal + 1);
258                 buf += len + 1;
259         }
260         free(to_free);
261 }
262
263 static void append_msg_id(struct strbuf *sb, const char *msg_id)
264 {
265         for (;;) {
266                 char c = *(msg_id)++;
267
268                 if (!c)
269                         break;
270                 if (c != '_')
271                         strbuf_addch(sb, tolower(c));
272                 else {
273                         assert(*msg_id);
274                         strbuf_addch(sb, *(msg_id)++);
275                 }
276         }
277
278         strbuf_addstr(sb, ": ");
279 }
280
281 static int object_on_skiplist(struct fsck_options *opts,
282                               const struct object_id *oid)
283 {
284         return opts && oid && oidset_contains(&opts->skiplist, oid);
285 }
286
287 __attribute__((format (printf, 5, 6)))
288 static int report(struct fsck_options *options,
289                   const struct object_id *oid, enum object_type object_type,
290                   enum fsck_msg_id id, const char *fmt, ...)
291 {
292         va_list ap;
293         struct strbuf sb = STRBUF_INIT;
294         int msg_type = fsck_msg_type(id, options), result;
295
296         if (msg_type == FSCK_IGNORE)
297                 return 0;
298
299         if (object_on_skiplist(options, oid))
300                 return 0;
301
302         if (msg_type == FSCK_FATAL)
303                 msg_type = FSCK_ERROR;
304         else if (msg_type == FSCK_INFO)
305                 msg_type = FSCK_WARN;
306
307         append_msg_id(&sb, msg_id_info[id].id_string);
308
309         va_start(ap, fmt);
310         strbuf_vaddf(&sb, fmt, ap);
311         result = options->error_func(options, oid, object_type,
312                                      msg_type, sb.buf);
313         strbuf_release(&sb);
314         va_end(ap);
315
316         return result;
317 }
318
319 void fsck_enable_object_names(struct fsck_options *options)
320 {
321         if (!options->object_names)
322                 options->object_names = kh_init_oid_map();
323 }
324
325 const char *fsck_get_object_name(struct fsck_options *options,
326                                  const struct object_id *oid)
327 {
328         khiter_t pos;
329         if (!options->object_names)
330                 return NULL;
331         pos = kh_get_oid_map(options->object_names, *oid);
332         if (pos >= kh_end(options->object_names))
333                 return NULL;
334         return kh_value(options->object_names, pos);
335 }
336
337 void fsck_put_object_name(struct fsck_options *options,
338                           const struct object_id *oid,
339                           const char *fmt, ...)
340 {
341         va_list ap;
342         struct strbuf buf = STRBUF_INIT;
343         khiter_t pos;
344         int hashret;
345
346         if (!options->object_names)
347                 return;
348
349         pos = kh_put_oid_map(options->object_names, *oid, &hashret);
350         if (!hashret)
351                 return;
352         va_start(ap, fmt);
353         strbuf_vaddf(&buf, fmt, ap);
354         kh_value(options->object_names, pos) = strbuf_detach(&buf, NULL);
355         va_end(ap);
356 }
357
358 const char *fsck_describe_object(struct fsck_options *options,
359                                  const struct object_id *oid)
360 {
361         static struct strbuf bufs[] = {
362                 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
363         };
364         static int b = 0;
365         struct strbuf *buf;
366         const char *name = fsck_get_object_name(options, oid);
367
368         buf = bufs + b;
369         b = (b + 1) % ARRAY_SIZE(bufs);
370         strbuf_reset(buf);
371         strbuf_addstr(buf, oid_to_hex(oid));
372         if (name)
373                 strbuf_addf(buf, " (%s)", name);
374
375         return buf->buf;
376 }
377
378 static int fsck_walk_tree(struct tree *tree, void *data, struct fsck_options *options)
379 {
380         struct tree_desc desc;
381         struct name_entry entry;
382         int res = 0;
383         const char *name;
384
385         if (parse_tree(tree))
386                 return -1;
387
388         name = fsck_get_object_name(options, &tree->object.oid);
389         if (init_tree_desc_gently(&desc, tree->buffer, tree->size))
390                 return -1;
391         while (tree_entry_gently(&desc, &entry)) {
392                 struct object *obj;
393                 int result;
394
395                 if (S_ISGITLINK(entry.mode))
396                         continue;
397
398                 if (S_ISDIR(entry.mode)) {
399                         obj = (struct object *)lookup_tree(the_repository, &entry.oid);
400                         if (name && obj)
401                                 fsck_put_object_name(options, &entry.oid, "%s%s/",
402                                                      name, entry.path);
403                         result = options->walk(obj, OBJ_TREE, data, options);
404                 }
405                 else if (S_ISREG(entry.mode) || S_ISLNK(entry.mode)) {
406                         obj = (struct object *)lookup_blob(the_repository, &entry.oid);
407                         if (name && obj)
408                                 fsck_put_object_name(options, &entry.oid, "%s%s",
409                                                      name, entry.path);
410                         result = options->walk(obj, OBJ_BLOB, data, options);
411                 }
412                 else {
413                         result = error("in tree %s: entry %s has bad mode %.6o",
414                                        fsck_describe_object(options, &tree->object.oid),
415                                        entry.path, entry.mode);
416                 }
417                 if (result < 0)
418                         return result;
419                 if (!res)
420                         res = result;
421         }
422         return res;
423 }
424
425 static int fsck_walk_commit(struct commit *commit, void *data, struct fsck_options *options)
426 {
427         int counter = 0, generation = 0, name_prefix_len = 0;
428         struct commit_list *parents;
429         int res;
430         int result;
431         const char *name;
432
433         if (parse_commit(commit))
434                 return -1;
435
436         name = fsck_get_object_name(options, &commit->object.oid);
437         if (name)
438                 fsck_put_object_name(options, get_commit_tree_oid(commit),
439                                      "%s:", name);
440
441         result = options->walk((struct object *)get_commit_tree(commit),
442                                OBJ_TREE, data, options);
443         if (result < 0)
444                 return result;
445         res = result;
446
447         parents = commit->parents;
448         if (name && parents) {
449                 int len = strlen(name), power;
450
451                 if (len && name[len - 1] == '^') {
452                         generation = 1;
453                         name_prefix_len = len - 1;
454                 }
455                 else { /* parse ~<generation> suffix */
456                         for (generation = 0, power = 1;
457                              len && isdigit(name[len - 1]);
458                              power *= 10)
459                                 generation += power * (name[--len] - '0');
460                         if (power > 1 && len && name[len - 1] == '~')
461                                 name_prefix_len = len - 1;
462                 }
463         }
464
465         while (parents) {
466                 if (name) {
467                         struct object_id *oid = &parents->item->object.oid;
468
469                         if (counter++)
470                                 fsck_put_object_name(options, oid, "%s^%d",
471                                                      name, counter);
472                         else if (generation > 0)
473                                 fsck_put_object_name(options, oid, "%.*s~%d",
474                                                      name_prefix_len, name,
475                                                      generation + 1);
476                         else
477                                 fsck_put_object_name(options, oid, "%s^", name);
478                 }
479                 result = options->walk((struct object *)parents->item, OBJ_COMMIT, data, options);
480                 if (result < 0)
481                         return result;
482                 if (!res)
483                         res = result;
484                 parents = parents->next;
485         }
486         return res;
487 }
488
489 static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *options)
490 {
491         const char *name = fsck_get_object_name(options, &tag->object.oid);
492
493         if (parse_tag(tag))
494                 return -1;
495         if (name)
496                 fsck_put_object_name(options, &tag->tagged->oid, "%s", name);
497         return options->walk(tag->tagged, OBJ_ANY, data, options);
498 }
499
500 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
501 {
502         if (!obj)
503                 return -1;
504
505         if (obj->type == OBJ_NONE)
506                 parse_object(the_repository, &obj->oid);
507
508         switch (obj->type) {
509         case OBJ_BLOB:
510                 return 0;
511         case OBJ_TREE:
512                 return fsck_walk_tree((struct tree *)obj, data, options);
513         case OBJ_COMMIT:
514                 return fsck_walk_commit((struct commit *)obj, data, options);
515         case OBJ_TAG:
516                 return fsck_walk_tag((struct tag *)obj, data, options);
517         default:
518                 error("Unknown object type for %s",
519                       fsck_describe_object(options, &obj->oid));
520                 return -1;
521         }
522 }
523
524 /*
525  * The entries in a tree are ordered in the _path_ order,
526  * which means that a directory entry is ordered by adding
527  * a slash to the end of it.
528  *
529  * So a directory called "a" is ordered _after_ a file
530  * called "a.c", because "a/" sorts after "a.c".
531  */
532 #define TREE_UNORDERED (-1)
533 #define TREE_HAS_DUPS  (-2)
534
535 static int verify_ordered(unsigned mode1, const char *name1, unsigned mode2, const char *name2)
536 {
537         int len1 = strlen(name1);
538         int len2 = strlen(name2);
539         int len = len1 < len2 ? len1 : len2;
540         unsigned char c1, c2;
541         int cmp;
542
543         cmp = memcmp(name1, name2, len);
544         if (cmp < 0)
545                 return 0;
546         if (cmp > 0)
547                 return TREE_UNORDERED;
548
549         /*
550          * Ok, the first <len> characters are the same.
551          * Now we need to order the next one, but turn
552          * a '\0' into a '/' for a directory entry.
553          */
554         c1 = name1[len];
555         c2 = name2[len];
556         if (!c1 && !c2)
557                 /*
558                  * git-write-tree used to write out a nonsense tree that has
559                  * entries with the same name, one blob and one tree.  Make
560                  * sure we do not have duplicate entries.
561                  */
562                 return TREE_HAS_DUPS;
563         if (!c1 && S_ISDIR(mode1))
564                 c1 = '/';
565         if (!c2 && S_ISDIR(mode2))
566                 c2 = '/';
567         return c1 < c2 ? 0 : TREE_UNORDERED;
568 }
569
570 static int fsck_tree(const struct object_id *oid,
571                      const char *buffer, unsigned long size,
572                      struct fsck_options *options)
573 {
574         int retval = 0;
575         int has_null_sha1 = 0;
576         int has_full_path = 0;
577         int has_empty_name = 0;
578         int has_dot = 0;
579         int has_dotdot = 0;
580         int has_dotgit = 0;
581         int has_zero_pad = 0;
582         int has_bad_modes = 0;
583         int has_dup_entries = 0;
584         int not_properly_sorted = 0;
585         struct tree_desc desc;
586         unsigned o_mode;
587         const char *o_name;
588
589         if (init_tree_desc_gently(&desc, buffer, size)) {
590                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
591                 return retval;
592         }
593
594         o_mode = 0;
595         o_name = NULL;
596
597         while (desc.size) {
598                 unsigned short mode;
599                 const char *name, *backslash;
600                 const struct object_id *oid;
601
602                 oid = tree_entry_extract(&desc, &name, &mode);
603
604                 has_null_sha1 |= is_null_oid(oid);
605                 has_full_path |= !!strchr(name, '/');
606                 has_empty_name |= !*name;
607                 has_dot |= !strcmp(name, ".");
608                 has_dotdot |= !strcmp(name, "..");
609                 has_dotgit |= is_hfs_dotgit(name) || is_ntfs_dotgit(name);
610                 has_zero_pad |= *(char *)desc.buffer == '0';
611
612                 if (is_hfs_dotgitmodules(name) || is_ntfs_dotgitmodules(name)) {
613                         if (!S_ISLNK(mode))
614                                 oidset_insert(&gitmodules_found, oid);
615                         else
616                                 retval += report(options,
617                                                  oid, OBJ_TREE,
618                                                  FSCK_MSG_GITMODULES_SYMLINK,
619                                                  ".gitmodules is a symbolic link");
620                 }
621
622                 if ((backslash = strchr(name, '\\'))) {
623                         while (backslash) {
624                                 backslash++;
625                                 has_dotgit |= is_ntfs_dotgit(backslash);
626                                 if (is_ntfs_dotgitmodules(backslash)) {
627                                         if (!S_ISLNK(mode))
628                                                 oidset_insert(&gitmodules_found, oid);
629                                         else
630                                                 retval += report(options, oid, OBJ_TREE,
631                                                                  FSCK_MSG_GITMODULES_SYMLINK,
632                                                                  ".gitmodules is a symbolic link");
633                                 }
634                                 backslash = strchr(backslash, '\\');
635                         }
636                 }
637
638                 if (update_tree_entry_gently(&desc)) {
639                         retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_TREE, "cannot be parsed as a tree");
640                         break;
641                 }
642
643                 switch (mode) {
644                 /*
645                  * Standard modes..
646                  */
647                 case S_IFREG | 0755:
648                 case S_IFREG | 0644:
649                 case S_IFLNK:
650                 case S_IFDIR:
651                 case S_IFGITLINK:
652                         break;
653                 /*
654                  * This is nonstandard, but we had a few of these
655                  * early on when we honored the full set of mode
656                  * bits..
657                  */
658                 case S_IFREG | 0664:
659                         if (!options->strict)
660                                 break;
661                         /* fallthrough */
662                 default:
663                         has_bad_modes = 1;
664                 }
665
666                 if (o_name) {
667                         switch (verify_ordered(o_mode, o_name, mode, name)) {
668                         case TREE_UNORDERED:
669                                 not_properly_sorted = 1;
670                                 break;
671                         case TREE_HAS_DUPS:
672                                 has_dup_entries = 1;
673                                 break;
674                         default:
675                                 break;
676                         }
677                 }
678
679                 o_mode = mode;
680                 o_name = name;
681         }
682
683         if (has_null_sha1)
684                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_NULL_SHA1, "contains entries pointing to null sha1");
685         if (has_full_path)
686                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_FULL_PATHNAME, "contains full pathnames");
687         if (has_empty_name)
688                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_EMPTY_NAME, "contains empty pathname");
689         if (has_dot)
690                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOT, "contains '.'");
691         if (has_dotdot)
692                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTDOT, "contains '..'");
693         if (has_dotgit)
694                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_HAS_DOTGIT, "contains '.git'");
695         if (has_zero_pad)
696                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_ZERO_PADDED_FILEMODE, "contains zero-padded file modes");
697         if (has_bad_modes)
698                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_BAD_FILEMODE, "contains bad file modes");
699         if (has_dup_entries)
700                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_DUPLICATE_ENTRIES, "contains duplicate file entries");
701         if (not_properly_sorted)
702                 retval += report(options, oid, OBJ_TREE, FSCK_MSG_TREE_NOT_SORTED, "not properly sorted");
703         return retval;
704 }
705
706 static int verify_headers(const void *data, unsigned long size,
707                           const struct object_id *oid, enum object_type type,
708                           struct fsck_options *options)
709 {
710         const char *buffer = (const char *)data;
711         unsigned long i;
712
713         for (i = 0; i < size; i++) {
714                 switch (buffer[i]) {
715                 case '\0':
716                         return report(options, oid, type,
717                                 FSCK_MSG_NUL_IN_HEADER,
718                                 "unterminated header: NUL at offset %ld", i);
719                 case '\n':
720                         if (i + 1 < size && buffer[i + 1] == '\n')
721                                 return 0;
722                 }
723         }
724
725         /*
726          * We did not find double-LF that separates the header
727          * and the body.  Not having a body is not a crime but
728          * we do want to see the terminating LF for the last header
729          * line.
730          */
731         if (size && buffer[size - 1] == '\n')
732                 return 0;
733
734         return report(options, oid, type,
735                 FSCK_MSG_UNTERMINATED_HEADER, "unterminated header");
736 }
737
738 static int fsck_ident(const char **ident,
739                       const struct object_id *oid, enum object_type type,
740                       struct fsck_options *options)
741 {
742         const char *p = *ident;
743         char *end;
744
745         *ident = strchrnul(*ident, '\n');
746         if (**ident == '\n')
747                 (*ident)++;
748
749         if (*p == '<')
750                 return report(options, oid, type, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
751         p += strcspn(p, "<>\n");
752         if (*p == '>')
753                 return report(options, oid, type, FSCK_MSG_BAD_NAME, "invalid author/committer line - bad name");
754         if (*p != '<')
755                 return report(options, oid, type, FSCK_MSG_MISSING_EMAIL, "invalid author/committer line - missing email");
756         if (p[-1] != ' ')
757                 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL, "invalid author/committer line - missing space before email");
758         p++;
759         p += strcspn(p, "<>\n");
760         if (*p != '>')
761                 return report(options, oid, type, FSCK_MSG_BAD_EMAIL, "invalid author/committer line - bad email");
762         p++;
763         if (*p != ' ')
764                 return report(options, oid, type, FSCK_MSG_MISSING_SPACE_BEFORE_DATE, "invalid author/committer line - missing space before date");
765         p++;
766         if (*p == '0' && p[1] != ' ')
767                 return report(options, oid, type, FSCK_MSG_ZERO_PADDED_DATE, "invalid author/committer line - zero-padded date");
768         if (date_overflows(parse_timestamp(p, &end, 10)))
769                 return report(options, oid, type, FSCK_MSG_BAD_DATE_OVERFLOW, "invalid author/committer line - date causes integer overflow");
770         if ((end == p || *end != ' '))
771                 return report(options, oid, type, FSCK_MSG_BAD_DATE, "invalid author/committer line - bad date");
772         p = end + 1;
773         if ((*p != '+' && *p != '-') ||
774             !isdigit(p[1]) ||
775             !isdigit(p[2]) ||
776             !isdigit(p[3]) ||
777             !isdigit(p[4]) ||
778             (p[5] != '\n'))
779                 return report(options, oid, type, FSCK_MSG_BAD_TIMEZONE, "invalid author/committer line - bad time zone");
780         p += 6;
781         return 0;
782 }
783
784 static int fsck_commit(const struct object_id *oid,
785                        const char *buffer, unsigned long size,
786                        struct fsck_options *options)
787 {
788         struct object_id tree_oid, parent_oid;
789         unsigned author_count;
790         int err;
791         const char *buffer_begin = buffer;
792         const char *p;
793
794         if (verify_headers(buffer, size, oid, OBJ_COMMIT, options))
795                 return -1;
796
797         if (!skip_prefix(buffer, "tree ", &buffer))
798                 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_TREE, "invalid format - expected 'tree' line");
799         if (parse_oid_hex(buffer, &tree_oid, &p) || *p != '\n') {
800                 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_TREE_SHA1, "invalid 'tree' line format - bad sha1");
801                 if (err)
802                         return err;
803         }
804         buffer = p + 1;
805         while (skip_prefix(buffer, "parent ", &buffer)) {
806                 if (parse_oid_hex(buffer, &parent_oid, &p) || *p != '\n') {
807                         err = report(options, oid, OBJ_COMMIT, FSCK_MSG_BAD_PARENT_SHA1, "invalid 'parent' line format - bad sha1");
808                         if (err)
809                                 return err;
810                 }
811                 buffer = p + 1;
812         }
813         author_count = 0;
814         while (skip_prefix(buffer, "author ", &buffer)) {
815                 author_count++;
816                 err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
817                 if (err)
818                         return err;
819         }
820         if (author_count < 1)
821                 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_AUTHOR, "invalid format - expected 'author' line");
822         else if (author_count > 1)
823                 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_MULTIPLE_AUTHORS, "invalid format - multiple 'author' lines");
824         if (err)
825                 return err;
826         if (!skip_prefix(buffer, "committer ", &buffer))
827                 return report(options, oid, OBJ_COMMIT, FSCK_MSG_MISSING_COMMITTER, "invalid format - expected 'committer' line");
828         err = fsck_ident(&buffer, oid, OBJ_COMMIT, options);
829         if (err)
830                 return err;
831         if (memchr(buffer_begin, '\0', size)) {
832                 err = report(options, oid, OBJ_COMMIT, FSCK_MSG_NUL_IN_COMMIT,
833                              "NUL byte in the commit object body");
834                 if (err)
835                         return err;
836         }
837         return 0;
838 }
839
840 static int fsck_tag(const struct object_id *oid, const char *buffer,
841                     unsigned long size, struct fsck_options *options)
842 {
843         struct object_id tagged_oid;
844         int ret = 0;
845         char *eol;
846         struct strbuf sb = STRBUF_INIT;
847         const char *p;
848
849         ret = verify_headers(buffer, size, oid, OBJ_TAG, options);
850         if (ret)
851                 goto done;
852
853         if (!skip_prefix(buffer, "object ", &buffer)) {
854                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_OBJECT, "invalid format - expected 'object' line");
855                 goto done;
856         }
857         if (parse_oid_hex(buffer, &tagged_oid, &p) || *p != '\n') {
858                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_OBJECT_SHA1, "invalid 'object' line format - bad sha1");
859                 if (ret)
860                         goto done;
861         }
862         buffer = p + 1;
863
864         if (!skip_prefix(buffer, "type ", &buffer)) {
865                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE_ENTRY, "invalid format - expected 'type' line");
866                 goto done;
867         }
868         eol = strchr(buffer, '\n');
869         if (!eol) {
870                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TYPE, "invalid format - unexpected end after 'type' line");
871                 goto done;
872         }
873         if (type_from_string_gently(buffer, eol - buffer, 1) < 0)
874                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_BAD_TYPE, "invalid 'type' value");
875         if (ret)
876                 goto done;
877         buffer = eol + 1;
878
879         if (!skip_prefix(buffer, "tag ", &buffer)) {
880                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG_ENTRY, "invalid format - expected 'tag' line");
881                 goto done;
882         }
883         eol = strchr(buffer, '\n');
884         if (!eol) {
885                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAG, "invalid format - unexpected end after 'type' line");
886                 goto done;
887         }
888         strbuf_addf(&sb, "refs/tags/%.*s", (int)(eol - buffer), buffer);
889         if (check_refname_format(sb.buf, 0)) {
890                 ret = report(options, oid, OBJ_TAG,
891                              FSCK_MSG_BAD_TAG_NAME,
892                              "invalid 'tag' name: %.*s",
893                              (int)(eol - buffer), buffer);
894                 if (ret)
895                         goto done;
896         }
897         buffer = eol + 1;
898
899         if (!skip_prefix(buffer, "tagger ", &buffer)) {
900                 /* early tags do not contain 'tagger' lines; warn only */
901                 ret = report(options, oid, OBJ_TAG, FSCK_MSG_MISSING_TAGGER_ENTRY, "invalid format - expected 'tagger' line");
902                 if (ret)
903                         goto done;
904         }
905         else
906                 ret = fsck_ident(&buffer, oid, OBJ_TAG, options);
907
908 done:
909         strbuf_release(&sb);
910         return ret;
911 }
912
913 struct fsck_gitmodules_data {
914         const struct object_id *oid;
915         struct fsck_options *options;
916         int ret;
917 };
918
919 static int fsck_gitmodules_fn(const char *var, const char *value, void *vdata)
920 {
921         struct fsck_gitmodules_data *data = vdata;
922         const char *subsection, *key;
923         int subsection_len;
924         char *name;
925
926         if (parse_config_key(var, "submodule", &subsection, &subsection_len, &key) < 0 ||
927             !subsection)
928                 return 0;
929
930         name = xmemdupz(subsection, subsection_len);
931         if (check_submodule_name(name) < 0)
932                 data->ret |= report(data->options,
933                                     data->oid, OBJ_BLOB,
934                                     FSCK_MSG_GITMODULES_NAME,
935                                     "disallowed submodule name: %s",
936                                     name);
937         if (!strcmp(key, "url") && value &&
938             looks_like_command_line_option(value))
939                 data->ret |= report(data->options,
940                                     data->oid, OBJ_BLOB,
941                                     FSCK_MSG_GITMODULES_URL,
942                                     "disallowed submodule url: %s",
943                                     value);
944         if (!strcmp(key, "path") && value &&
945             looks_like_command_line_option(value))
946                 data->ret |= report(data->options,
947                                     data->oid, OBJ_BLOB,
948                                     FSCK_MSG_GITMODULES_PATH,
949                                     "disallowed submodule path: %s",
950                                     value);
951         if (!strcmp(key, "update") && value &&
952             parse_submodule_update_type(value) == SM_UPDATE_COMMAND)
953                 data->ret |= report(data->options, data->oid, OBJ_BLOB,
954                                     FSCK_MSG_GITMODULES_UPDATE,
955                                     "disallowed submodule update setting: %s",
956                                     value);
957         free(name);
958
959         return 0;
960 }
961
962 static int fsck_blob(const struct object_id *oid, const char *buf,
963                      unsigned long size, struct fsck_options *options)
964 {
965         struct fsck_gitmodules_data data;
966         struct config_options config_opts = { 0 };
967
968         if (!oidset_contains(&gitmodules_found, oid))
969                 return 0;
970         oidset_insert(&gitmodules_done, oid);
971
972         if (object_on_skiplist(options, oid))
973                 return 0;
974
975         if (!buf) {
976                 /*
977                  * A missing buffer here is a sign that the caller found the
978                  * blob too gigantic to load into memory. Let's just consider
979                  * that an error.
980                  */
981                 return report(options, oid, OBJ_BLOB,
982                               FSCK_MSG_GITMODULES_LARGE,
983                               ".gitmodules too large to parse");
984         }
985
986         data.oid = oid;
987         data.options = options;
988         data.ret = 0;
989         config_opts.error_action = CONFIG_ERROR_SILENT;
990         if (git_config_from_mem(fsck_gitmodules_fn, CONFIG_ORIGIN_BLOB,
991                                 ".gitmodules", buf, size, &data, &config_opts))
992                 data.ret |= report(options, oid, OBJ_BLOB,
993                                    FSCK_MSG_GITMODULES_PARSE,
994                                    "could not parse gitmodules blob");
995
996         return data.ret;
997 }
998
999 int fsck_object(struct object *obj, void *data, unsigned long size,
1000         struct fsck_options *options)
1001 {
1002         if (!obj)
1003                 return report(options, NULL, OBJ_NONE, FSCK_MSG_BAD_OBJECT_SHA1, "no valid object to fsck");
1004
1005         if (obj->type == OBJ_BLOB)
1006                 return fsck_blob(&obj->oid, data, size, options);
1007         if (obj->type == OBJ_TREE)
1008                 return fsck_tree(&obj->oid, data, size, options);
1009         if (obj->type == OBJ_COMMIT)
1010                 return fsck_commit(&obj->oid, data, size, options);
1011         if (obj->type == OBJ_TAG)
1012                 return fsck_tag(&obj->oid, data, size, options);
1013
1014         return report(options, &obj->oid, obj->type,
1015                       FSCK_MSG_UNKNOWN_TYPE,
1016                       "unknown type '%d' (internal fsck error)",
1017                       obj->type);
1018 }
1019
1020 int fsck_error_function(struct fsck_options *o,
1021                         const struct object_id *oid,
1022                         enum object_type object_type,
1023                         int msg_type, const char *message)
1024 {
1025         if (msg_type == FSCK_WARN) {
1026                 warning("object %s: %s", fsck_describe_object(o, oid), message);
1027                 return 0;
1028         }
1029         error("object %s: %s", fsck_describe_object(o, oid), message);
1030         return 1;
1031 }
1032
1033 int fsck_finish(struct fsck_options *options)
1034 {
1035         int ret = 0;
1036         struct oidset_iter iter;
1037         const struct object_id *oid;
1038
1039         oidset_iter_init(&gitmodules_found, &iter);
1040         while ((oid = oidset_iter_next(&iter))) {
1041                 enum object_type type;
1042                 unsigned long size;
1043                 char *buf;
1044
1045                 if (oidset_contains(&gitmodules_done, oid))
1046                         continue;
1047
1048                 buf = read_object_file(oid, &type, &size);
1049                 if (!buf) {
1050                         if (is_promisor_object(oid))
1051                                 continue;
1052                         ret |= report(options,
1053                                       oid, OBJ_BLOB,
1054                                       FSCK_MSG_GITMODULES_MISSING,
1055                                       "unable to read .gitmodules blob");
1056                         continue;
1057                 }
1058
1059                 if (type == OBJ_BLOB)
1060                         ret |= fsck_blob(oid, buf, size, options);
1061                 else
1062                         ret |= report(options,
1063                                       oid, type,
1064                                       FSCK_MSG_GITMODULES_BLOB,
1065                                       "non-blob found at .gitmodules");
1066                 free(buf);
1067         }
1068
1069
1070         oidset_clear(&gitmodules_found);
1071         oidset_clear(&gitmodules_done);
1072         return ret;
1073 }