fsck: don't fsck alternates for connectivity-only check
[git/git.git] / builtin / fsck.c
1 #include "builtin.h"
2 #include "cache.h"
3 #include "commit.h"
4 #include "tree.h"
5 #include "blob.h"
6 #include "tag.h"
7 #include "refs.h"
8 #include "pack.h"
9 #include "cache-tree.h"
10 #include "tree-walk.h"
11 #include "fsck.h"
12 #include "parse-options.h"
13 #include "dir.h"
14 #include "progress.h"
15 #include "streaming.h"
16
17 #define REACHABLE 0x0001
18 #define SEEN 0x0002
19 #define HAS_OBJ 0x0004
20
21 static int show_root;
22 static int show_tags;
23 static int show_unreachable;
24 static int include_reflogs = 1;
25 static int check_full = 1;
26 static int connectivity_only;
27 static int check_strict;
28 static int keep_cache_objects;
29 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
30 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
31 static struct object_id head_oid;
32 static const char *head_points_at;
33 static int errors_found;
34 static int write_lost_and_found;
35 static int verbose;
36 static int show_progress = -1;
37 static int show_dangling = 1;
38 #define ERROR_OBJECT 01
39 #define ERROR_REACHABLE 02
40 #define ERROR_PACK 04
41
42 #ifdef NO_D_INO_IN_DIRENT
43 #define SORT_DIRENT 0
44 #define DIRENT_SORT_HINT(de) 0
45 #else
46 #define SORT_DIRENT 1
47 #define DIRENT_SORT_HINT(de) ((de)->d_ino)
48 #endif
49
50 static int fsck_config(const char *var, const char *value, void *cb)
51 {
52 if (strcmp(var, "fsck.skiplist") == 0) {
53 const char *path;
54 struct strbuf sb = STRBUF_INIT;
55
56 if (git_config_pathname(&path, var, value))
57 return 1;
58 strbuf_addf(&sb, "skiplist=%s", path);
59 free((char *)path);
60 fsck_set_msg_types(&fsck_obj_options, sb.buf);
61 strbuf_release(&sb);
62 return 0;
63 }
64
65 if (skip_prefix(var, "fsck.", &var)) {
66 fsck_set_msg_type(&fsck_obj_options, var, value);
67 return 0;
68 }
69
70 return git_default_config(var, value, cb);
71 }
72
73 static void objreport(struct object *obj, const char *msg_type,
74 const char *err)
75 {
76 fprintf(stderr, "%s in %s %s: %s\n",
77 msg_type, typename(obj->type), sha1_to_hex(obj->sha1), err);
78 }
79
80 static int objerror(struct object *obj, const char *err)
81 {
82 errors_found |= ERROR_OBJECT;
83 objreport(obj, "error", err);
84 return -1;
85 }
86
87 static int fsck_error_func(struct object *obj, int type, const char *message)
88 {
89 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
90 return (type == FSCK_WARN) ? 0 : 1;
91 }
92
93 static struct object_array pending;
94
95 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
96 {
97 struct object *parent = data;
98
99 /*
100 * The only case data is NULL or type is OBJ_ANY is when
101 * mark_object_reachable() calls us. All the callers of
102 * that function has non-NULL obj hence ...
103 */
104 if (!obj) {
105 /* ... these references to parent->fld are safe here */
106 printf("broken link from %7s %s\n",
107 typename(parent->type), sha1_to_hex(parent->sha1));
108 printf("broken link from %7s %s\n",
109 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
110 errors_found |= ERROR_REACHABLE;
111 return 1;
112 }
113
114 if (type != OBJ_ANY && obj->type != type)
115 /* ... and the reference to parent is safe here */
116 objerror(parent, "wrong object type in link");
117
118 if (obj->flags & REACHABLE)
119 return 0;
120 obj->flags |= REACHABLE;
121 if (!(obj->flags & HAS_OBJ)) {
122 if (parent && !has_sha1_file(obj->sha1)) {
123 printf("broken link from %7s %s\n",
124 typename(parent->type), sha1_to_hex(parent->sha1));
125 printf(" to %7s %s\n",
126 typename(obj->type), sha1_to_hex(obj->sha1));
127 errors_found |= ERROR_REACHABLE;
128 }
129 return 1;
130 }
131
132 add_object_array(obj, NULL, &pending);
133 return 0;
134 }
135
136 static void mark_object_reachable(struct object *obj)
137 {
138 mark_object(obj, OBJ_ANY, NULL, NULL);
139 }
140
141 static int traverse_one_object(struct object *obj)
142 {
143 int result;
144 struct tree *tree = NULL;
145
146 if (obj->type == OBJ_TREE) {
147 tree = (struct tree *)obj;
148 if (parse_tree(tree) < 0)
149 return 1; /* error already displayed */
150 }
151 result = fsck_walk(obj, obj, &fsck_walk_options);
152 if (tree)
153 free_tree_buffer(tree);
154 return result;
155 }
156
157 static int traverse_reachable(void)
158 {
159 struct progress *progress = NULL;
160 unsigned int nr = 0;
161 int result = 0;
162 if (show_progress)
163 progress = start_progress_delay(_("Checking connectivity"), 0, 0, 2);
164 while (pending.nr) {
165 struct object_array_entry *entry;
166 struct object *obj;
167
168 entry = pending.objects + --pending.nr;
169 obj = entry->item;
170 result |= traverse_one_object(obj);
171 display_progress(progress, ++nr);
172 }
173 stop_progress(&progress);
174 return !!result;
175 }
176
177 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
178 {
179 if (!obj)
180 return 1;
181 obj->used = 1;
182 return 0;
183 }
184
185 /*
186 * Check a single reachable object
187 */
188 static void check_reachable_object(struct object *obj)
189 {
190 /*
191 * We obviously want the object to be parsed,
192 * except if it was in a pack-file and we didn't
193 * do a full fsck
194 */
195 if (!(obj->flags & HAS_OBJ)) {
196 if (has_sha1_pack(obj->sha1))
197 return; /* it is in pack - forget about it */
198 if (connectivity_only && has_sha1_file(obj->sha1))
199 return;
200 printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
201 errors_found |= ERROR_REACHABLE;
202 return;
203 }
204 }
205
206 /*
207 * Check a single unreachable object
208 */
209 static void check_unreachable_object(struct object *obj)
210 {
211 /*
212 * Missing unreachable object? Ignore it. It's not like
213 * we miss it (since it can't be reached), nor do we want
214 * to complain about it being unreachable (since it does
215 * not exist).
216 */
217 if (!obj->parsed)
218 return;
219
220 /*
221 * Unreachable object that exists? Show it if asked to,
222 * since this is something that is prunable.
223 */
224 if (show_unreachable) {
225 printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
226 return;
227 }
228
229 /*
230 * "!used" means that nothing at all points to it, including
231 * other unreachable objects. In other words, it's the "tip"
232 * of some set of unreachable objects, usually a commit that
233 * got dropped.
234 *
235 * Such starting points are more interesting than some random
236 * set of unreachable objects, so we show them even if the user
237 * hasn't asked for _all_ unreachable objects. If you have
238 * deleted a branch by mistake, this is a prime candidate to
239 * start looking at, for example.
240 */
241 if (!obj->used) {
242 if (show_dangling)
243 printf("dangling %s %s\n", typename(obj->type),
244 sha1_to_hex(obj->sha1));
245 if (write_lost_and_found) {
246 char *filename = git_pathdup("lost-found/%s/%s",
247 obj->type == OBJ_COMMIT ? "commit" : "other",
248 sha1_to_hex(obj->sha1));
249 FILE *f;
250
251 if (safe_create_leading_directories_const(filename)) {
252 error("Could not create lost-found");
253 free(filename);
254 return;
255 }
256 if (!(f = fopen(filename, "w")))
257 die_errno("Could not open '%s'", filename);
258 if (obj->type == OBJ_BLOB) {
259 if (stream_blob_to_fd(fileno(f), obj->sha1, NULL, 1))
260 die_errno("Could not write '%s'", filename);
261 } else
262 fprintf(f, "%s\n", sha1_to_hex(obj->sha1));
263 if (fclose(f))
264 die_errno("Could not finish '%s'",
265 filename);
266 free(filename);
267 }
268 return;
269 }
270
271 /*
272 * Otherwise? It's there, it's unreachable, and some other unreachable
273 * object points to it. Ignore it - it's not interesting, and we showed
274 * all the interesting cases above.
275 */
276 }
277
278 static void check_object(struct object *obj)
279 {
280 if (verbose)
281 fprintf(stderr, "Checking %s\n", sha1_to_hex(obj->sha1));
282
283 if (obj->flags & REACHABLE)
284 check_reachable_object(obj);
285 else
286 check_unreachable_object(obj);
287 }
288
289 static void check_connectivity(void)
290 {
291 int i, max;
292
293 /* Traverse the pending reachable objects */
294 traverse_reachable();
295
296 /* Look up all the requirements, warn about missing objects.. */
297 max = get_max_object_index();
298 if (verbose)
299 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
300
301 for (i = 0; i < max; i++) {
302 struct object *obj = get_indexed_object(i);
303
304 if (obj)
305 check_object(obj);
306 }
307 }
308
309 static int fsck_obj(struct object *obj)
310 {
311 if (obj->flags & SEEN)
312 return 0;
313 obj->flags |= SEEN;
314
315 if (verbose)
316 fprintf(stderr, "Checking %s %s\n",
317 typename(obj->type), sha1_to_hex(obj->sha1));
318
319 if (fsck_walk(obj, NULL, &fsck_obj_options))
320 objerror(obj, "broken links");
321 if (fsck_object(obj, NULL, 0, &fsck_obj_options))
322 return -1;
323
324 if (obj->type == OBJ_TREE) {
325 struct tree *item = (struct tree *) obj;
326
327 free_tree_buffer(item);
328 }
329
330 if (obj->type == OBJ_COMMIT) {
331 struct commit *commit = (struct commit *) obj;
332
333 free_commit_buffer(commit);
334
335 if (!commit->parents && show_root)
336 printf("root %s\n", sha1_to_hex(commit->object.sha1));
337 }
338
339 if (obj->type == OBJ_TAG) {
340 struct tag *tag = (struct tag *) obj;
341
342 if (show_tags && tag->tagged) {
343 printf("tagged %s %s", typename(tag->tagged->type), sha1_to_hex(tag->tagged->sha1));
344 printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
345 }
346 }
347
348 return 0;
349 }
350
351 static int fsck_sha1(const unsigned char *sha1)
352 {
353 struct object *obj = parse_object(sha1);
354 if (!obj) {
355 errors_found |= ERROR_OBJECT;
356 return error("%s: object corrupt or missing",
357 sha1_to_hex(sha1));
358 }
359 obj->flags |= HAS_OBJ;
360 return fsck_obj(obj);
361 }
362
363 static int fsck_obj_buffer(const unsigned char *sha1, enum object_type type,
364 unsigned long size, void *buffer, int *eaten)
365 {
366 struct object *obj;
367 obj = parse_object_buffer(sha1, type, size, buffer, eaten);
368 if (!obj) {
369 errors_found |= ERROR_OBJECT;
370 return error("%s: object corrupt or missing", sha1_to_hex(sha1));
371 }
372 obj->flags = HAS_OBJ;
373 return fsck_obj(obj);
374 }
375
376 /*
377 * This is the sorting chunk size: make it reasonably
378 * big so that we can sort well..
379 */
380 #define MAX_SHA1_ENTRIES (1024)
381
382 struct sha1_entry {
383 unsigned long ino;
384 unsigned char sha1[20];
385 };
386
387 static struct {
388 unsigned long nr;
389 struct sha1_entry *entry[MAX_SHA1_ENTRIES];
390 } sha1_list;
391
392 static int ino_compare(const void *_a, const void *_b)
393 {
394 const struct sha1_entry *a = _a, *b = _b;
395 unsigned long ino1 = a->ino, ino2 = b->ino;
396 return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
397 }
398
399 static void fsck_sha1_list(void)
400 {
401 int i, nr = sha1_list.nr;
402
403 if (SORT_DIRENT)
404 qsort(sha1_list.entry, nr,
405 sizeof(struct sha1_entry *), ino_compare);
406 for (i = 0; i < nr; i++) {
407 struct sha1_entry *entry = sha1_list.entry[i];
408 unsigned char *sha1 = entry->sha1;
409
410 sha1_list.entry[i] = NULL;
411 if (fsck_sha1(sha1))
412 errors_found |= ERROR_OBJECT;
413 free(entry);
414 }
415 sha1_list.nr = 0;
416 }
417
418 static void add_sha1_list(unsigned char *sha1, unsigned long ino)
419 {
420 struct sha1_entry *entry = xmalloc(sizeof(*entry));
421 int nr;
422
423 entry->ino = ino;
424 hashcpy(entry->sha1, sha1);
425 nr = sha1_list.nr;
426 if (nr == MAX_SHA1_ENTRIES) {
427 fsck_sha1_list();
428 nr = 0;
429 }
430 sha1_list.entry[nr] = entry;
431 sha1_list.nr = ++nr;
432 }
433
434 static inline int is_loose_object_file(struct dirent *de,
435 char *name, unsigned char *sha1)
436 {
437 if (strlen(de->d_name) != 38)
438 return 0;
439 memcpy(name + 2, de->d_name, 39);
440 return !get_sha1_hex(name, sha1);
441 }
442
443 static void fsck_dir(int i, char *path)
444 {
445 DIR *dir = opendir(path);
446 struct dirent *de;
447 char name[100];
448
449 if (!dir)
450 return;
451
452 if (verbose)
453 fprintf(stderr, "Checking directory %s\n", path);
454
455 sprintf(name, "%02x", i);
456 while ((de = readdir(dir)) != NULL) {
457 unsigned char sha1[20];
458
459 if (is_dot_or_dotdot(de->d_name))
460 continue;
461 if (is_loose_object_file(de, name, sha1)) {
462 add_sha1_list(sha1, DIRENT_SORT_HINT(de));
463 continue;
464 }
465 if (starts_with(de->d_name, "tmp_obj_"))
466 continue;
467 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
468 }
469 closedir(dir);
470 }
471
472 static int default_refs;
473
474 static void fsck_handle_reflog_sha1(const char *refname, unsigned char *sha1)
475 {
476 struct object *obj;
477
478 if (!is_null_sha1(sha1)) {
479 obj = lookup_object(sha1);
480 if (obj) {
481 obj->used = 1;
482 mark_object_reachable(obj);
483 } else {
484 error("%s: invalid reflog entry %s", refname, sha1_to_hex(sha1));
485 errors_found |= ERROR_REACHABLE;
486 }
487 }
488 }
489
490 static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
491 const char *email, unsigned long timestamp, int tz,
492 const char *message, void *cb_data)
493 {
494 const char *refname = cb_data;
495
496 if (verbose)
497 fprintf(stderr, "Checking reflog %s->%s\n",
498 sha1_to_hex(osha1), sha1_to_hex(nsha1));
499
500 fsck_handle_reflog_sha1(refname, osha1);
501 fsck_handle_reflog_sha1(refname, nsha1);
502 return 0;
503 }
504
505 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
506 int flag, void *cb_data)
507 {
508 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
509 return 0;
510 }
511
512 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
513 int flag, void *cb_data)
514 {
515 struct object *obj;
516
517 obj = parse_object(oid->hash);
518 if (!obj) {
519 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
520 errors_found |= ERROR_REACHABLE;
521 /* We'll continue with the rest despite the error.. */
522 return 0;
523 }
524 if (obj->type != OBJ_COMMIT && is_branch(refname))
525 error("%s: not a commit", refname);
526 default_refs++;
527 obj->used = 1;
528 mark_object_reachable(obj);
529
530 return 0;
531 }
532
533 static void get_default_heads(void)
534 {
535 if (head_points_at && !is_null_oid(&head_oid))
536 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
537 for_each_rawref(fsck_handle_ref, NULL);
538 if (include_reflogs)
539 for_each_reflog(fsck_handle_reflog, NULL);
540
541 /*
542 * Not having any default heads isn't really fatal, but
543 * it does mean that "--unreachable" no longer makes any
544 * sense (since in this case everything will obviously
545 * be unreachable by definition.
546 *
547 * Showing dangling objects is valid, though (as those
548 * dangling objects are likely lost heads).
549 *
550 * So we just print a warning about it, and clear the
551 * "show_unreachable" flag.
552 */
553 if (!default_refs) {
554 fprintf(stderr, "notice: No default references\n");
555 show_unreachable = 0;
556 }
557 }
558
559 static void fsck_object_dir(const char *path)
560 {
561 int i;
562 struct progress *progress = NULL;
563
564 if (verbose)
565 fprintf(stderr, "Checking object directory\n");
566
567 if (show_progress)
568 progress = start_progress(_("Checking object directories"), 256);
569 for (i = 0; i < 256; i++) {
570 static char dir[4096];
571 sprintf(dir, "%s/%02x", path, i);
572 fsck_dir(i, dir);
573 display_progress(progress, i+1);
574 }
575 stop_progress(&progress);
576 fsck_sha1_list();
577 }
578
579 static int fsck_head_link(void)
580 {
581 int flag;
582 int null_is_error = 0;
583
584 if (verbose)
585 fprintf(stderr, "Checking HEAD link\n");
586
587 head_points_at = resolve_ref_unsafe("HEAD", 0, head_oid.hash, &flag);
588 if (!head_points_at)
589 return error("Invalid HEAD");
590 if (!strcmp(head_points_at, "HEAD"))
591 /* detached HEAD */
592 null_is_error = 1;
593 else if (!starts_with(head_points_at, "refs/heads/"))
594 return error("HEAD points to something strange (%s)",
595 head_points_at);
596 if (is_null_oid(&head_oid)) {
597 if (null_is_error)
598 return error("HEAD: detached HEAD points at nothing");
599 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
600 head_points_at + 11);
601 }
602 return 0;
603 }
604
605 static int fsck_cache_tree(struct cache_tree *it)
606 {
607 int i;
608 int err = 0;
609
610 if (verbose)
611 fprintf(stderr, "Checking cache tree\n");
612
613 if (0 <= it->entry_count) {
614 struct object *obj = parse_object(it->sha1);
615 if (!obj) {
616 error("%s: invalid sha1 pointer in cache-tree",
617 sha1_to_hex(it->sha1));
618 return 1;
619 }
620 obj->used = 1;
621 mark_object_reachable(obj);
622 if (obj->type != OBJ_TREE)
623 err |= objerror(obj, "non-tree in cache-tree");
624 }
625 for (i = 0; i < it->subtree_nr; i++)
626 err |= fsck_cache_tree(it->down[i]->cache_tree);
627 return err;
628 }
629
630 static char const * const fsck_usage[] = {
631 N_("git fsck [<options>] [<object>...]"),
632 NULL
633 };
634
635 static struct option fsck_opts[] = {
636 OPT__VERBOSE(&verbose, N_("be verbose")),
637 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
638 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
639 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
640 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
641 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
642 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
643 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
644 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
645 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
646 OPT_BOOL(0, "lost-found", &write_lost_and_found,
647 N_("write dangling objects in .git/lost-found")),
648 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
649 OPT_END(),
650 };
651
652 int cmd_fsck(int argc, const char **argv, const char *prefix)
653 {
654 int i, heads;
655 struct alternate_object_database *alt;
656
657 errors_found = 0;
658 check_replace_refs = 0;
659
660 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
661
662 fsck_walk_options.walk = mark_object;
663 fsck_obj_options.walk = mark_used;
664 fsck_obj_options.error_func = fsck_error_func;
665 if (check_strict)
666 fsck_obj_options.strict = 1;
667
668 if (show_progress == -1)
669 show_progress = isatty(2);
670 if (verbose)
671 show_progress = 0;
672
673 if (write_lost_and_found) {
674 check_full = 1;
675 include_reflogs = 0;
676 }
677
678 git_config(fsck_config, NULL);
679
680 fsck_head_link();
681 if (!connectivity_only) {
682 fsck_object_dir(get_object_directory());
683
684 prepare_alt_odb();
685 for (alt = alt_odb_list; alt; alt = alt->next) {
686 char namebuf[PATH_MAX];
687 int namelen = alt->name - alt->base;
688 memcpy(namebuf, alt->base, namelen);
689 namebuf[namelen - 1] = 0;
690 fsck_object_dir(namebuf);
691 }
692 }
693
694 if (check_full) {
695 struct packed_git *p;
696 uint32_t total = 0, count = 0;
697 struct progress *progress = NULL;
698
699 prepare_packed_git();
700
701 if (show_progress) {
702 for (p = packed_git; p; p = p->next) {
703 if (open_pack_index(p))
704 continue;
705 total += p->num_objects;
706 }
707
708 progress = start_progress(_("Checking objects"), total);
709 }
710 for (p = packed_git; p; p = p->next) {
711 /* verify gives error messages itself */
712 if (verify_pack(p, fsck_obj_buffer,
713 progress, count))
714 errors_found |= ERROR_PACK;
715 count += p->num_objects;
716 }
717 stop_progress(&progress);
718 }
719
720 heads = 0;
721 for (i = 0; i < argc; i++) {
722 const char *arg = argv[i];
723 unsigned char sha1[20];
724 if (!get_sha1(arg, sha1)) {
725 struct object *obj = lookup_object(sha1);
726
727 /* Error is printed by lookup_object(). */
728 if (!obj)
729 continue;
730
731 obj->used = 1;
732 mark_object_reachable(obj);
733 heads++;
734 continue;
735 }
736 error("invalid parameter: expected sha1, got '%s'", arg);
737 }
738
739 /*
740 * If we've not been given any explicit head information, do the
741 * default ones from .git/refs. We also consider the index file
742 * in this case (ie this implies --cache).
743 */
744 if (!heads) {
745 get_default_heads();
746 keep_cache_objects = 1;
747 }
748
749 if (keep_cache_objects) {
750 read_cache();
751 for (i = 0; i < active_nr; i++) {
752 unsigned int mode;
753 struct blob *blob;
754 struct object *obj;
755
756 mode = active_cache[i]->ce_mode;
757 if (S_ISGITLINK(mode))
758 continue;
759 blob = lookup_blob(active_cache[i]->sha1);
760 if (!blob)
761 continue;
762 obj = &blob->object;
763 obj->used = 1;
764 mark_object_reachable(obj);
765 }
766 if (active_cache_tree)
767 fsck_cache_tree(active_cache_tree);
768 }
769
770 check_connectivity();
771 return errors_found;
772 }