Merge branch 'sb/packfiles-in-repository'
[git.git] / builtin / fsck.c
blob087360a6757c607c4b38cce5fafad3bbb0822fc0
1 #include "builtin.h"
2 #include "cache.h"
3 #include "repository.h"
4 #include "config.h"
5 #include "commit.h"
6 #include "tree.h"
7 #include "blob.h"
8 #include "tag.h"
9 #include "refs.h"
10 #include "pack.h"
11 #include "cache-tree.h"
12 #include "tree-walk.h"
13 #include "fsck.h"
14 #include "parse-options.h"
15 #include "dir.h"
16 #include "progress.h"
17 #include "streaming.h"
18 #include "decorate.h"
19 #include "packfile.h"
20 #include "object-store.h"
22 #define REACHABLE 0x0001
23 #define SEEN 0x0002
24 #define HAS_OBJ 0x0004
25 /* This flag is set if something points to this object. */
26 #define USED 0x0008
28 static int show_root;
29 static int show_tags;
30 static int show_unreachable;
31 static int include_reflogs = 1;
32 static int check_full = 1;
33 static int connectivity_only;
34 static int check_strict;
35 static int keep_cache_objects;
36 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
37 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
38 static struct object_id head_oid;
39 static const char *head_points_at;
40 static int errors_found;
41 static int write_lost_and_found;
42 static int verbose;
43 static int show_progress = -1;
44 static int show_dangling = 1;
45 static int name_objects;
46 #define ERROR_OBJECT 01
47 #define ERROR_REACHABLE 02
48 #define ERROR_PACK 04
49 #define ERROR_REFS 010
51 static const char *describe_object(struct object *obj)
53 static struct strbuf buf = STRBUF_INIT;
54 char *name = name_objects ?
55 lookup_decoration(fsck_walk_options.object_names, obj) : NULL;
57 strbuf_reset(&buf);
58 strbuf_addstr(&buf, oid_to_hex(&obj->oid));
59 if (name)
60 strbuf_addf(&buf, " (%s)", name);
62 return buf.buf;
65 static const char *printable_type(struct object *obj)
67 const char *ret;
69 if (obj->type == OBJ_NONE) {
70 enum object_type type = oid_object_info(&obj->oid, NULL);
71 if (type > 0)
72 object_as_type(obj, type, 0);
75 ret = type_name(obj->type);
76 if (!ret)
77 ret = "unknown";
79 return ret;
82 static int fsck_config(const char *var, const char *value, void *cb)
84 if (strcmp(var, "fsck.skiplist") == 0) {
85 const char *path;
86 struct strbuf sb = STRBUF_INIT;
88 if (git_config_pathname(&path, var, value))
89 return 1;
90 strbuf_addf(&sb, "skiplist=%s", path);
91 free((char *)path);
92 fsck_set_msg_types(&fsck_obj_options, sb.buf);
93 strbuf_release(&sb);
94 return 0;
97 if (skip_prefix(var, "fsck.", &var)) {
98 fsck_set_msg_type(&fsck_obj_options, var, value);
99 return 0;
102 return git_default_config(var, value, cb);
105 static void objreport(struct object *obj, const char *msg_type,
106 const char *err)
108 fprintf(stderr, "%s in %s %s: %s\n",
109 msg_type, printable_type(obj), describe_object(obj), err);
112 static int objerror(struct object *obj, const char *err)
114 errors_found |= ERROR_OBJECT;
115 objreport(obj, "error", err);
116 return -1;
119 static int fsck_error_func(struct fsck_options *o,
120 struct object *obj, int type, const char *message)
122 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
123 return (type == FSCK_WARN) ? 0 : 1;
126 static struct object_array pending;
128 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
130 struct object *parent = data;
133 * The only case data is NULL or type is OBJ_ANY is when
134 * mark_object_reachable() calls us. All the callers of
135 * that function has non-NULL obj hence ...
137 if (!obj) {
138 /* ... these references to parent->fld are safe here */
139 printf("broken link from %7s %s\n",
140 printable_type(parent), describe_object(parent));
141 printf("broken link from %7s %s\n",
142 (type == OBJ_ANY ? "unknown" : type_name(type)), "unknown");
143 errors_found |= ERROR_REACHABLE;
144 return 1;
147 if (type != OBJ_ANY && obj->type != type)
148 /* ... and the reference to parent is safe here */
149 objerror(parent, "wrong object type in link");
151 if (obj->flags & REACHABLE)
152 return 0;
153 obj->flags |= REACHABLE;
155 if (is_promisor_object(&obj->oid))
157 * Further recursion does not need to be performed on this
158 * object since it is a promisor object (so it does not need to
159 * be added to "pending").
161 return 0;
163 if (!(obj->flags & HAS_OBJ)) {
164 if (parent && !has_object_file(&obj->oid)) {
165 printf("broken link from %7s %s\n",
166 printable_type(parent), describe_object(parent));
167 printf(" to %7s %s\n",
168 printable_type(obj), describe_object(obj));
169 errors_found |= ERROR_REACHABLE;
171 return 1;
174 add_object_array(obj, NULL, &pending);
175 return 0;
178 static void mark_object_reachable(struct object *obj)
180 mark_object(obj, OBJ_ANY, NULL, NULL);
183 static int traverse_one_object(struct object *obj)
185 int result = fsck_walk(obj, obj, &fsck_walk_options);
187 if (obj->type == OBJ_TREE) {
188 struct tree *tree = (struct tree *)obj;
189 free_tree_buffer(tree);
191 return result;
194 static int traverse_reachable(void)
196 struct progress *progress = NULL;
197 unsigned int nr = 0;
198 int result = 0;
199 if (show_progress)
200 progress = start_delayed_progress(_("Checking connectivity"), 0);
201 while (pending.nr) {
202 result |= traverse_one_object(object_array_pop(&pending));
203 display_progress(progress, ++nr);
205 stop_progress(&progress);
206 return !!result;
209 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
211 if (!obj)
212 return 1;
213 obj->flags |= USED;
214 return 0;
218 * Check a single reachable object
220 static void check_reachable_object(struct object *obj)
223 * We obviously want the object to be parsed,
224 * except if it was in a pack-file and we didn't
225 * do a full fsck
227 if (!(obj->flags & HAS_OBJ)) {
228 if (is_promisor_object(&obj->oid))
229 return;
230 if (has_sha1_pack(obj->oid.hash))
231 return; /* it is in pack - forget about it */
232 printf("missing %s %s\n", printable_type(obj),
233 describe_object(obj));
234 errors_found |= ERROR_REACHABLE;
235 return;
240 * Check a single unreachable object
242 static void check_unreachable_object(struct object *obj)
245 * Missing unreachable object? Ignore it. It's not like
246 * we miss it (since it can't be reached), nor do we want
247 * to complain about it being unreachable (since it does
248 * not exist).
250 if (!(obj->flags & HAS_OBJ))
251 return;
254 * Unreachable object that exists? Show it if asked to,
255 * since this is something that is prunable.
257 if (show_unreachable) {
258 printf("unreachable %s %s\n", printable_type(obj),
259 describe_object(obj));
260 return;
264 * "!USED" means that nothing at all points to it, including
265 * other unreachable objects. In other words, it's the "tip"
266 * of some set of unreachable objects, usually a commit that
267 * got dropped.
269 * Such starting points are more interesting than some random
270 * set of unreachable objects, so we show them even if the user
271 * hasn't asked for _all_ unreachable objects. If you have
272 * deleted a branch by mistake, this is a prime candidate to
273 * start looking at, for example.
275 if (!(obj->flags & USED)) {
276 if (show_dangling)
277 printf("dangling %s %s\n", printable_type(obj),
278 describe_object(obj));
279 if (write_lost_and_found) {
280 char *filename = git_pathdup("lost-found/%s/%s",
281 obj->type == OBJ_COMMIT ? "commit" : "other",
282 describe_object(obj));
283 FILE *f;
285 if (safe_create_leading_directories_const(filename)) {
286 error("Could not create lost-found");
287 free(filename);
288 return;
290 f = xfopen(filename, "w");
291 if (obj->type == OBJ_BLOB) {
292 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
293 die_errno("Could not write '%s'", filename);
294 } else
295 fprintf(f, "%s\n", describe_object(obj));
296 if (fclose(f))
297 die_errno("Could not finish '%s'",
298 filename);
299 free(filename);
301 return;
305 * Otherwise? It's there, it's unreachable, and some other unreachable
306 * object points to it. Ignore it - it's not interesting, and we showed
307 * all the interesting cases above.
311 static void check_object(struct object *obj)
313 if (verbose)
314 fprintf(stderr, "Checking %s\n", describe_object(obj));
316 if (obj->flags & REACHABLE)
317 check_reachable_object(obj);
318 else
319 check_unreachable_object(obj);
322 static void check_connectivity(void)
324 int i, max;
326 /* Traverse the pending reachable objects */
327 traverse_reachable();
329 /* Look up all the requirements, warn about missing objects.. */
330 max = get_max_object_index();
331 if (verbose)
332 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
334 for (i = 0; i < max; i++) {
335 struct object *obj = get_indexed_object(i);
337 if (obj)
338 check_object(obj);
342 static int fsck_obj(struct object *obj)
344 int err;
346 if (obj->flags & SEEN)
347 return 0;
348 obj->flags |= SEEN;
350 if (verbose)
351 fprintf(stderr, "Checking %s %s\n",
352 printable_type(obj), describe_object(obj));
354 if (fsck_walk(obj, NULL, &fsck_obj_options))
355 objerror(obj, "broken links");
356 err = fsck_object(obj, NULL, 0, &fsck_obj_options);
357 if (err)
358 goto out;
360 if (obj->type == OBJ_COMMIT) {
361 struct commit *commit = (struct commit *) obj;
363 if (!commit->parents && show_root)
364 printf("root %s\n", describe_object(&commit->object));
367 if (obj->type == OBJ_TAG) {
368 struct tag *tag = (struct tag *) obj;
370 if (show_tags && tag->tagged) {
371 printf("tagged %s %s", printable_type(tag->tagged),
372 describe_object(tag->tagged));
373 printf(" (%s) in %s\n", tag->tag,
374 describe_object(&tag->object));
378 out:
379 if (obj->type == OBJ_TREE)
380 free_tree_buffer((struct tree *)obj);
381 if (obj->type == OBJ_COMMIT)
382 free_commit_buffer((struct commit *)obj);
383 return err;
386 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
387 unsigned long size, void *buffer, int *eaten)
390 * Note, buffer may be NULL if type is OBJ_BLOB. See
391 * verify_packfile(), data_valid variable for details.
393 struct object *obj;
394 obj = parse_object_buffer(oid, type, size, buffer, eaten);
395 if (!obj) {
396 errors_found |= ERROR_OBJECT;
397 return error("%s: object corrupt or missing", oid_to_hex(oid));
399 obj->flags &= ~(REACHABLE | SEEN);
400 obj->flags |= HAS_OBJ;
401 return fsck_obj(obj);
404 static int default_refs;
406 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
407 timestamp_t timestamp)
409 struct object *obj;
411 if (!is_null_oid(oid)) {
412 obj = lookup_object(oid->hash);
413 if (obj && (obj->flags & HAS_OBJ)) {
414 if (timestamp && name_objects)
415 add_decoration(fsck_walk_options.object_names,
416 obj,
417 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
418 obj->flags |= USED;
419 mark_object_reachable(obj);
420 } else if (!is_promisor_object(oid)) {
421 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
422 errors_found |= ERROR_REACHABLE;
427 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
428 const char *email, timestamp_t timestamp, int tz,
429 const char *message, void *cb_data)
431 const char *refname = cb_data;
433 if (verbose)
434 fprintf(stderr, "Checking reflog %s->%s\n",
435 oid_to_hex(ooid), oid_to_hex(noid));
437 fsck_handle_reflog_oid(refname, ooid, 0);
438 fsck_handle_reflog_oid(refname, noid, timestamp);
439 return 0;
442 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
443 int flag, void *cb_data)
445 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
446 return 0;
449 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
450 int flag, void *cb_data)
452 struct object *obj;
454 obj = parse_object(oid);
455 if (!obj) {
456 if (is_promisor_object(oid)) {
458 * Increment default_refs anyway, because this is a
459 * valid ref.
461 default_refs++;
462 return 0;
464 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
465 errors_found |= ERROR_REACHABLE;
466 /* We'll continue with the rest despite the error.. */
467 return 0;
469 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
470 error("%s: not a commit", refname);
471 errors_found |= ERROR_REFS;
473 default_refs++;
474 obj->flags |= USED;
475 if (name_objects)
476 add_decoration(fsck_walk_options.object_names,
477 obj, xstrdup(refname));
478 mark_object_reachable(obj);
480 return 0;
483 static void get_default_heads(void)
485 if (head_points_at && !is_null_oid(&head_oid))
486 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
487 for_each_rawref(fsck_handle_ref, NULL);
488 if (include_reflogs)
489 for_each_reflog(fsck_handle_reflog, NULL);
492 * Not having any default heads isn't really fatal, but
493 * it does mean that "--unreachable" no longer makes any
494 * sense (since in this case everything will obviously
495 * be unreachable by definition.
497 * Showing dangling objects is valid, though (as those
498 * dangling objects are likely lost heads).
500 * So we just print a warning about it, and clear the
501 * "show_unreachable" flag.
503 if (!default_refs) {
504 fprintf(stderr, "notice: No default references\n");
505 show_unreachable = 0;
509 static struct object *parse_loose_object(const struct object_id *oid,
510 const char *path)
512 struct object *obj;
513 void *contents;
514 enum object_type type;
515 unsigned long size;
516 int eaten;
518 if (read_loose_object(path, oid, &type, &size, &contents) < 0)
519 return NULL;
521 if (!contents && type != OBJ_BLOB)
522 die("BUG: read_loose_object streamed a non-blob");
524 obj = parse_object_buffer(oid, type, size, contents, &eaten);
526 if (!eaten)
527 free(contents);
528 return obj;
531 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
533 struct object *obj = parse_loose_object(oid, path);
535 if (!obj) {
536 errors_found |= ERROR_OBJECT;
537 error("%s: object corrupt or missing: %s",
538 oid_to_hex(oid), path);
539 return 0; /* keep checking other objects */
542 obj->flags &= ~(REACHABLE | SEEN);
543 obj->flags |= HAS_OBJ;
544 if (fsck_obj(obj))
545 errors_found |= ERROR_OBJECT;
546 return 0;
549 static int fsck_cruft(const char *basename, const char *path, void *data)
551 if (!starts_with(basename, "tmp_obj_"))
552 fprintf(stderr, "bad sha1 file: %s\n", path);
553 return 0;
556 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
558 display_progress(progress, nr + 1);
559 return 0;
562 static void fsck_object_dir(const char *path)
564 struct progress *progress = NULL;
566 if (verbose)
567 fprintf(stderr, "Checking object directory\n");
569 if (show_progress)
570 progress = start_progress(_("Checking object directories"), 256);
572 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
573 progress);
574 display_progress(progress, 256);
575 stop_progress(&progress);
578 static int fsck_head_link(void)
580 int null_is_error = 0;
582 if (verbose)
583 fprintf(stderr, "Checking HEAD link\n");
585 head_points_at = resolve_ref_unsafe("HEAD", 0, &head_oid, NULL);
586 if (!head_points_at) {
587 errors_found |= ERROR_REFS;
588 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 errors_found |= ERROR_REFS;
595 return error("HEAD points to something strange (%s)",
596 head_points_at);
598 if (is_null_oid(&head_oid)) {
599 if (null_is_error) {
600 errors_found |= ERROR_REFS;
601 return error("HEAD: detached HEAD points at nothing");
603 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
604 head_points_at + 11);
606 return 0;
609 static int fsck_cache_tree(struct cache_tree *it)
611 int i;
612 int err = 0;
614 if (verbose)
615 fprintf(stderr, "Checking cache tree\n");
617 if (0 <= it->entry_count) {
618 struct object *obj = parse_object(&it->oid);
619 if (!obj) {
620 error("%s: invalid sha1 pointer in cache-tree",
621 oid_to_hex(&it->oid));
622 errors_found |= ERROR_REFS;
623 return 1;
625 obj->flags |= USED;
626 if (name_objects)
627 add_decoration(fsck_walk_options.object_names,
628 obj, xstrdup(":"));
629 mark_object_reachable(obj);
630 if (obj->type != OBJ_TREE)
631 err |= objerror(obj, "non-tree in cache-tree");
633 for (i = 0; i < it->subtree_nr; i++)
634 err |= fsck_cache_tree(it->down[i]->cache_tree);
635 return err;
638 static void mark_object_for_connectivity(const struct object_id *oid)
640 struct object *obj = lookup_unknown_object(oid->hash);
641 obj->flags |= HAS_OBJ;
644 static int mark_loose_for_connectivity(const struct object_id *oid,
645 const char *path,
646 void *data)
648 mark_object_for_connectivity(oid);
649 return 0;
652 static int mark_packed_for_connectivity(const struct object_id *oid,
653 struct packed_git *pack,
654 uint32_t pos,
655 void *data)
657 mark_object_for_connectivity(oid);
658 return 0;
661 static char const * const fsck_usage[] = {
662 N_("git fsck [<options>] [<object>...]"),
663 NULL
666 static struct option fsck_opts[] = {
667 OPT__VERBOSE(&verbose, N_("be verbose")),
668 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
669 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
670 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
671 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
672 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
673 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
674 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
675 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
676 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
677 OPT_BOOL(0, "lost-found", &write_lost_and_found,
678 N_("write dangling objects in .git/lost-found")),
679 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
680 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
681 OPT_END(),
684 int cmd_fsck(int argc, const char **argv, const char *prefix)
686 int i;
687 struct alternate_object_database *alt;
689 /* fsck knows how to handle missing promisor objects */
690 fetch_if_missing = 0;
692 errors_found = 0;
693 check_replace_refs = 0;
695 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
697 fsck_walk_options.walk = mark_object;
698 fsck_obj_options.walk = mark_used;
699 fsck_obj_options.error_func = fsck_error_func;
700 if (check_strict)
701 fsck_obj_options.strict = 1;
703 if (show_progress == -1)
704 show_progress = isatty(2);
705 if (verbose)
706 show_progress = 0;
708 if (write_lost_and_found) {
709 check_full = 1;
710 include_reflogs = 0;
713 if (name_objects)
714 fsck_walk_options.object_names =
715 xcalloc(1, sizeof(struct decoration));
717 git_config(fsck_config, NULL);
719 fsck_head_link();
720 if (connectivity_only) {
721 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
722 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
723 } else {
724 struct alternate_object_database *alt_odb_list;
726 fsck_object_dir(get_object_directory());
728 prepare_alt_odb(the_repository);
729 alt_odb_list = the_repository->objects->alt_odb_list;
730 for (alt = alt_odb_list; alt; alt = alt->next)
731 fsck_object_dir(alt->path);
733 if (check_full) {
734 struct packed_git *p;
735 uint32_t total = 0, count = 0;
736 struct progress *progress = NULL;
738 if (show_progress) {
739 for (p = get_packed_git(the_repository); p;
740 p = p->next) {
741 if (open_pack_index(p))
742 continue;
743 total += p->num_objects;
746 progress = start_progress(_("Checking objects"), total);
748 for (p = get_packed_git(the_repository); p;
749 p = p->next) {
750 /* verify gives error messages itself */
751 if (verify_pack(p, fsck_obj_buffer,
752 progress, count))
753 errors_found |= ERROR_PACK;
754 count += p->num_objects;
756 stop_progress(&progress);
760 for (i = 0; i < argc; i++) {
761 const char *arg = argv[i];
762 struct object_id oid;
763 if (!get_oid(arg, &oid)) {
764 struct object *obj = lookup_object(oid.hash);
766 if (!obj || !(obj->flags & HAS_OBJ)) {
767 if (is_promisor_object(&oid))
768 continue;
769 error("%s: object missing", oid_to_hex(&oid));
770 errors_found |= ERROR_OBJECT;
771 continue;
774 obj->flags |= USED;
775 if (name_objects)
776 add_decoration(fsck_walk_options.object_names,
777 obj, xstrdup(arg));
778 mark_object_reachable(obj);
779 continue;
781 error("invalid parameter: expected sha1, got '%s'", arg);
782 errors_found |= ERROR_OBJECT;
786 * If we've not been given any explicit head information, do the
787 * default ones from .git/refs. We also consider the index file
788 * in this case (ie this implies --cache).
790 if (!argc) {
791 get_default_heads();
792 keep_cache_objects = 1;
795 if (keep_cache_objects) {
796 verify_index_checksum = 1;
797 verify_ce_order = 1;
798 read_cache();
799 for (i = 0; i < active_nr; i++) {
800 unsigned int mode;
801 struct blob *blob;
802 struct object *obj;
804 mode = active_cache[i]->ce_mode;
805 if (S_ISGITLINK(mode))
806 continue;
807 blob = lookup_blob(&active_cache[i]->oid);
808 if (!blob)
809 continue;
810 obj = &blob->object;
811 obj->flags |= USED;
812 if (name_objects)
813 add_decoration(fsck_walk_options.object_names,
814 obj,
815 xstrfmt(":%s", active_cache[i]->name));
816 mark_object_reachable(obj);
818 if (active_cache_tree)
819 fsck_cache_tree(active_cache_tree);
822 check_connectivity();
823 return errors_found;