3 #include "repository.h"
11 #include "cache-tree.h"
12 #include "tree-walk.h"
14 #include "parse-options.h"
17 #include "streaming.h"
20 #include "object-store.h"
21 #include "run-command.h"
24 #define REACHABLE 0x0001
26 #define HAS_OBJ 0x0004
27 /* This flag is set if something points to this object. */
32 static int show_unreachable;
33 static int include_reflogs = 1;
34 static int check_full = 1;
35 static int connectivity_only;
36 static int check_strict;
37 static int keep_cache_objects;
38 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
39 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
40 static int errors_found;
41 static int write_lost_and_found;
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
49 #define ERROR_REFS 010
50 #define ERROR_COMMIT_GRAPH 020
52 static const char *describe_object(struct object *obj)
54 static struct strbuf bufs[] = {
55 STRBUF_INIT, STRBUF_INIT, STRBUF_INIT, STRBUF_INIT
62 name = lookup_decoration(fsck_walk_options.object_names, obj);
65 b = (b + 1) % ARRAY_SIZE(bufs);
67 strbuf_addstr(buf, oid_to_hex(&obj->oid));
69 strbuf_addf(buf, " (%s)", name);
74 static const char *printable_type(struct object *obj)
78 if (obj->type == OBJ_NONE) {
79 enum object_type type = oid_object_info(the_repository,
82 object_as_type(the_repository, obj, type, 0);
85 ret = type_name(obj->type);
92 static int fsck_config(const char *var, const char *value, void *cb)
94 if (strcmp(var, "fsck.skiplist") == 0) {
96 struct strbuf sb = STRBUF_INIT;
98 if (git_config_pathname(&path, var, value))
100 strbuf_addf(&sb, "skiplist=%s", path);
102 fsck_set_msg_types(&fsck_obj_options, sb.buf);
107 if (skip_prefix(var, "fsck.", &var)) {
108 fsck_set_msg_type(&fsck_obj_options, var, value);
112 return git_default_config(var, value, cb);
115 static int objerror(struct object *obj, const char *err)
117 errors_found |= ERROR_OBJECT;
118 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
119 fprintf_ln(stderr, _("error in %s %s: %s"),
120 printable_type(obj), describe_object(obj), err);
124 static int fsck_error_func(struct fsck_options *o,
125 struct object *obj, int type, const char *message)
129 /* TRANSLATORS: e.g. warning in tree 01bfda: <more explanation> */
130 fprintf_ln(stderr, _("warning in %s %s: %s"),
131 printable_type(obj), describe_object(obj), message);
134 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
135 fprintf_ln(stderr, _("error in %s %s: %s"),
136 printable_type(obj), describe_object(obj), message);
139 BUG("%d (FSCK_IGNORE?) should never trigger this callback", type);
143 static struct object_array pending;
145 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
147 struct object *parent = data;
150 * The only case data is NULL or type is OBJ_ANY is when
151 * mark_object_reachable() calls us. All the callers of
152 * that function has non-NULL obj hence ...
155 /* ... these references to parent->fld are safe here */
156 printf_ln(_("broken link from %7s %s"),
157 printable_type(parent), describe_object(parent));
158 printf_ln(_("broken link from %7s %s"),
159 (type == OBJ_ANY ? _("unknown") : type_name(type)),
161 errors_found |= ERROR_REACHABLE;
165 if (type != OBJ_ANY && obj->type != type)
166 /* ... and the reference to parent is safe here */
167 objerror(parent, _("wrong object type in link"));
169 if (obj->flags & REACHABLE)
171 obj->flags |= REACHABLE;
173 if (is_promisor_object(&obj->oid))
175 * Further recursion does not need to be performed on this
176 * object since it is a promisor object (so it does not need to
177 * be added to "pending").
181 if (!(obj->flags & HAS_OBJ)) {
182 if (parent && !has_object_file(&obj->oid)) {
183 printf_ln(_("broken link from %7s %s\n"
185 printable_type(parent),
186 describe_object(parent),
188 describe_object(obj));
189 errors_found |= ERROR_REACHABLE;
194 add_object_array(obj, NULL, &pending);
198 static void mark_object_reachable(struct object *obj)
200 mark_object(obj, OBJ_ANY, NULL, NULL);
203 static int traverse_one_object(struct object *obj)
205 int result = fsck_walk(obj, obj, &fsck_walk_options);
207 if (obj->type == OBJ_TREE) {
208 struct tree *tree = (struct tree *)obj;
209 free_tree_buffer(tree);
214 static int traverse_reachable(void)
216 struct progress *progress = NULL;
220 progress = start_delayed_progress(_("Checking connectivity"), 0);
222 result |= traverse_one_object(object_array_pop(&pending));
223 display_progress(progress, ++nr);
225 stop_progress(&progress);
229 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
238 * Check a single reachable object
240 static void check_reachable_object(struct object *obj)
243 * We obviously want the object to be parsed,
244 * except if it was in a pack-file and we didn't
247 if (!(obj->flags & HAS_OBJ)) {
248 if (is_promisor_object(&obj->oid))
250 if (has_object_pack(&obj->oid))
251 return; /* it is in pack - forget about it */
252 printf_ln(_("missing %s %s"), printable_type(obj),
253 describe_object(obj));
254 errors_found |= ERROR_REACHABLE;
260 * Check a single unreachable object
262 static void check_unreachable_object(struct object *obj)
265 * Missing unreachable object? Ignore it. It's not like
266 * we miss it (since it can't be reached), nor do we want
267 * to complain about it being unreachable (since it does
270 if (!(obj->flags & HAS_OBJ))
274 * Unreachable object that exists? Show it if asked to,
275 * since this is something that is prunable.
277 if (show_unreachable) {
278 printf_ln(_("unreachable %s %s"), printable_type(obj),
279 describe_object(obj));
284 * "!USED" means that nothing at all points to it, including
285 * other unreachable objects. In other words, it's the "tip"
286 * of some set of unreachable objects, usually a commit that
289 * Such starting points are more interesting than some random
290 * set of unreachable objects, so we show them even if the user
291 * hasn't asked for _all_ unreachable objects. If you have
292 * deleted a branch by mistake, this is a prime candidate to
293 * start looking at, for example.
295 if (!(obj->flags & USED)) {
297 printf_ln(_("dangling %s %s"), printable_type(obj),
298 describe_object(obj));
299 if (write_lost_and_found) {
300 char *filename = git_pathdup("lost-found/%s/%s",
301 obj->type == OBJ_COMMIT ? "commit" : "other",
302 describe_object(obj));
305 if (safe_create_leading_directories_const(filename)) {
306 error(_("could not create lost-found"));
310 f = xfopen(filename, "w");
311 if (obj->type == OBJ_BLOB) {
312 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
313 die_errno(_("could not write '%s'"), filename);
315 fprintf(f, "%s\n", describe_object(obj));
317 die_errno(_("could not finish '%s'"),
325 * Otherwise? It's there, it's unreachable, and some other unreachable
326 * object points to it. Ignore it - it's not interesting, and we showed
327 * all the interesting cases above.
331 static void check_object(struct object *obj)
334 fprintf_ln(stderr, _("Checking %s"), describe_object(obj));
336 if (obj->flags & REACHABLE)
337 check_reachable_object(obj);
339 check_unreachable_object(obj);
342 static void check_connectivity(void)
346 /* Traverse the pending reachable objects */
347 traverse_reachable();
349 /* Look up all the requirements, warn about missing objects.. */
350 max = get_max_object_index();
352 fprintf_ln(stderr, _("Checking connectivity (%d objects)"), max);
354 for (i = 0; i < max; i++) {
355 struct object *obj = get_indexed_object(i);
362 static int fsck_obj(struct object *obj, void *buffer, unsigned long size)
366 if (obj->flags & SEEN)
371 fprintf_ln(stderr, _("Checking %s %s"),
372 printable_type(obj), describe_object(obj));
374 if (fsck_walk(obj, NULL, &fsck_obj_options))
375 objerror(obj, _("broken links"));
376 err = fsck_object(obj, buffer, size, &fsck_obj_options);
380 if (obj->type == OBJ_COMMIT) {
381 struct commit *commit = (struct commit *) obj;
383 if (!commit->parents && show_root)
384 printf_ln(_("root %s"),
385 describe_object(&commit->object));
388 if (obj->type == OBJ_TAG) {
389 struct tag *tag = (struct tag *) obj;
391 if (show_tags && tag->tagged) {
392 printf_ln(_("tagged %s %s (%s) in %s"),
393 printable_type(tag->tagged),
394 describe_object(tag->tagged),
396 describe_object(&tag->object));
401 if (obj->type == OBJ_TREE)
402 free_tree_buffer((struct tree *)obj);
403 if (obj->type == OBJ_COMMIT)
404 free_commit_buffer(the_repository->parsed_objects,
405 (struct commit *)obj);
409 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
410 unsigned long size, void *buffer, int *eaten)
413 * Note, buffer may be NULL if type is OBJ_BLOB. See
414 * verify_packfile(), data_valid variable for details.
417 obj = parse_object_buffer(the_repository, oid, type, size, buffer,
420 errors_found |= ERROR_OBJECT;
421 return error(_("%s: object corrupt or missing"),
424 obj->flags &= ~(REACHABLE | SEEN);
425 obj->flags |= HAS_OBJ;
426 return fsck_obj(obj, buffer, size);
429 static int default_refs;
431 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
432 timestamp_t timestamp)
436 if (!is_null_oid(oid)) {
437 obj = lookup_object(the_repository, oid->hash);
438 if (obj && (obj->flags & HAS_OBJ)) {
439 if (timestamp && name_objects)
440 add_decoration(fsck_walk_options.object_names,
442 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
444 mark_object_reachable(obj);
445 } else if (!is_promisor_object(oid)) {
446 error(_("%s: invalid reflog entry %s"),
447 refname, oid_to_hex(oid));
448 errors_found |= ERROR_REACHABLE;
453 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
454 const char *email, timestamp_t timestamp, int tz,
455 const char *message, void *cb_data)
457 const char *refname = cb_data;
460 fprintf_ln(stderr, _("Checking reflog %s->%s"),
461 oid_to_hex(ooid), oid_to_hex(noid));
463 fsck_handle_reflog_oid(refname, ooid, 0);
464 fsck_handle_reflog_oid(refname, noid, timestamp);
468 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
469 int flag, void *cb_data)
471 struct strbuf refname = STRBUF_INIT;
473 strbuf_worktree_ref(cb_data, &refname, logname);
474 for_each_reflog_ent(refname.buf, fsck_handle_reflog_ent, refname.buf);
475 strbuf_release(&refname);
479 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
480 int flag, void *cb_data)
484 obj = parse_object(the_repository, oid);
486 if (is_promisor_object(oid)) {
488 * Increment default_refs anyway, because this is a
494 error(_("%s: invalid sha1 pointer %s"),
495 refname, oid_to_hex(oid));
496 errors_found |= ERROR_REACHABLE;
497 /* We'll continue with the rest despite the error.. */
500 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
501 error(_("%s: not a commit"), refname);
502 errors_found |= ERROR_REFS;
507 add_decoration(fsck_walk_options.object_names,
508 obj, xstrdup(refname));
509 mark_object_reachable(obj);
514 static int fsck_head_link(const char *head_ref_name,
515 const char **head_points_at,
516 struct object_id *head_oid);
518 static void get_default_heads(void)
520 struct worktree **worktrees, **p;
521 const char *head_points_at;
522 struct object_id head_oid;
524 for_each_rawref(fsck_handle_ref, NULL);
526 worktrees = get_worktrees(0);
527 for (p = worktrees; *p; p++) {
528 struct worktree *wt = *p;
529 struct strbuf ref = STRBUF_INIT;
531 strbuf_worktree_ref(wt, &ref, "HEAD");
532 fsck_head_link(ref.buf, &head_points_at, &head_oid);
533 if (head_points_at && !is_null_oid(&head_oid))
534 fsck_handle_ref(ref.buf, &head_oid, 0, NULL);
535 strbuf_release(&ref);
538 refs_for_each_reflog(get_worktree_ref_store(wt),
539 fsck_handle_reflog, wt);
541 free_worktrees(worktrees);
544 * Not having any default heads isn't really fatal, but
545 * it does mean that "--unreachable" no longer makes any
546 * sense (since in this case everything will obviously
547 * be unreachable by definition.
549 * Showing dangling objects is valid, though (as those
550 * dangling objects are likely lost heads).
552 * So we just print a warning about it, and clear the
553 * "show_unreachable" flag.
556 fprintf_ln(stderr, _("notice: No default references"));
557 show_unreachable = 0;
561 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
564 enum object_type type;
569 if (read_loose_object(path, oid, &type, &size, &contents) < 0) {
570 errors_found |= ERROR_OBJECT;
571 error(_("%s: object corrupt or missing: %s"),
572 oid_to_hex(oid), path);
573 return 0; /* keep checking other objects */
576 if (!contents && type != OBJ_BLOB)
577 BUG("read_loose_object streamed a non-blob");
579 obj = parse_object_buffer(the_repository, oid, type, size,
583 errors_found |= ERROR_OBJECT;
584 error(_("%s: object could not be parsed: %s"),
585 oid_to_hex(oid), path);
588 return 0; /* keep checking other objects */
591 obj->flags &= ~(REACHABLE | SEEN);
592 obj->flags |= HAS_OBJ;
593 if (fsck_obj(obj, contents, size))
594 errors_found |= ERROR_OBJECT;
598 return 0; /* keep checking other objects, even if we saw an error */
601 static int fsck_cruft(const char *basename, const char *path, void *data)
603 if (!starts_with(basename, "tmp_obj_"))
604 fprintf_ln(stderr, _("bad sha1 file: %s"), path);
608 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
610 display_progress(progress, nr + 1);
614 static void fsck_object_dir(const char *path)
616 struct progress *progress = NULL;
619 fprintf_ln(stderr, _("Checking object directory"));
622 progress = start_progress(_("Checking object directories"), 256);
624 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
626 display_progress(progress, 256);
627 stop_progress(&progress);
630 static int fsck_head_link(const char *head_ref_name,
631 const char **head_points_at,
632 struct object_id *head_oid)
634 int null_is_error = 0;
637 fprintf_ln(stderr, _("Checking %s link"), head_ref_name);
639 *head_points_at = resolve_ref_unsafe(head_ref_name, 0, head_oid, NULL);
640 if (!*head_points_at) {
641 errors_found |= ERROR_REFS;
642 return error(_("invalid %s"), head_ref_name);
644 if (!strcmp(*head_points_at, head_ref_name))
647 else if (!starts_with(*head_points_at, "refs/heads/")) {
648 errors_found |= ERROR_REFS;
649 return error(_("%s points to something strange (%s)"),
650 head_ref_name, *head_points_at);
652 if (is_null_oid(head_oid)) {
654 errors_found |= ERROR_REFS;
655 return error(_("%s: detached HEAD points at nothing"),
659 _("notice: %s points to an unborn branch (%s)"),
660 head_ref_name, *head_points_at + 11);
665 static int fsck_cache_tree(struct cache_tree *it)
671 fprintf_ln(stderr, _("Checking cache tree"));
673 if (0 <= it->entry_count) {
674 struct object *obj = parse_object(the_repository, &it->oid);
676 error(_("%s: invalid sha1 pointer in cache-tree"),
677 oid_to_hex(&it->oid));
678 errors_found |= ERROR_REFS;
683 add_decoration(fsck_walk_options.object_names,
685 mark_object_reachable(obj);
686 if (obj->type != OBJ_TREE)
687 err |= objerror(obj, _("non-tree in cache-tree"));
689 for (i = 0; i < it->subtree_nr; i++)
690 err |= fsck_cache_tree(it->down[i]->cache_tree);
694 static void mark_object_for_connectivity(const struct object_id *oid)
696 struct object *obj = lookup_unknown_object(oid->hash);
697 obj->flags |= HAS_OBJ;
700 static int mark_loose_for_connectivity(const struct object_id *oid,
704 mark_object_for_connectivity(oid);
708 static int mark_packed_for_connectivity(const struct object_id *oid,
709 struct packed_git *pack,
713 mark_object_for_connectivity(oid);
717 static char const * const fsck_usage[] = {
718 N_("git fsck [<options>] [<object>...]"),
722 static struct option fsck_opts[] = {
723 OPT__VERBOSE(&verbose, N_("be verbose")),
724 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
725 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
726 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
727 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
728 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
729 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
730 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
731 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
732 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
733 OPT_BOOL(0, "lost-found", &write_lost_and_found,
734 N_("write dangling objects in .git/lost-found")),
735 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
736 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
740 int cmd_fsck(int argc, const char **argv, const char *prefix)
743 struct object_directory *odb;
745 /* fsck knows how to handle missing promisor objects */
746 fetch_if_missing = 0;
749 read_replace_refs = 0;
751 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
753 fsck_walk_options.walk = mark_object;
754 fsck_obj_options.walk = mark_used;
755 fsck_obj_options.error_func = fsck_error_func;
757 fsck_obj_options.strict = 1;
759 if (show_progress == -1)
760 show_progress = isatty(2);
764 if (write_lost_and_found) {
770 fsck_walk_options.object_names =
771 xcalloc(1, sizeof(struct decoration));
773 git_config(fsck_config, NULL);
775 if (connectivity_only) {
776 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
777 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
779 prepare_alt_odb(the_repository);
780 for (odb = the_repository->objects->odb; odb; odb = odb->next)
781 fsck_object_dir(odb->path);
784 struct packed_git *p;
785 uint32_t total = 0, count = 0;
786 struct progress *progress = NULL;
789 for (p = get_all_packs(the_repository); p;
791 if (open_pack_index(p))
793 total += p->num_objects;
796 progress = start_progress(_("Checking objects"), total);
798 for (p = get_all_packs(the_repository); p;
800 /* verify gives error messages itself */
801 if (verify_pack(the_repository,
804 errors_found |= ERROR_PACK;
805 count += p->num_objects;
807 stop_progress(&progress);
810 if (fsck_finish(&fsck_obj_options))
811 errors_found |= ERROR_OBJECT;
814 for (i = 0; i < argc; i++) {
815 const char *arg = argv[i];
816 struct object_id oid;
817 if (!get_oid(arg, &oid)) {
818 struct object *obj = lookup_object(the_repository,
821 if (!obj || !(obj->flags & HAS_OBJ)) {
822 if (is_promisor_object(&oid))
824 error(_("%s: object missing"), oid_to_hex(&oid));
825 errors_found |= ERROR_OBJECT;
831 add_decoration(fsck_walk_options.object_names,
833 mark_object_reachable(obj);
836 error(_("invalid parameter: expected sha1, got '%s'"), arg);
837 errors_found |= ERROR_OBJECT;
841 * If we've not been given any explicit head information, do the
842 * default ones from .git/refs. We also consider the index file
843 * in this case (ie this implies --cache).
847 keep_cache_objects = 1;
850 if (keep_cache_objects) {
851 verify_index_checksum = 1;
854 for (i = 0; i < active_nr; i++) {
859 mode = active_cache[i]->ce_mode;
860 if (S_ISGITLINK(mode))
862 blob = lookup_blob(the_repository,
863 &active_cache[i]->oid);
869 add_decoration(fsck_walk_options.object_names,
871 xstrfmt(":%s", active_cache[i]->name));
872 mark_object_reachable(obj);
874 if (active_cache_tree)
875 fsck_cache_tree(active_cache_tree);
878 check_connectivity();
880 if (!git_config_get_bool("core.commitgraph", &i) && i) {
881 struct child_process commit_graph_verify = CHILD_PROCESS_INIT;
882 const char *verify_argv[] = { "commit-graph", "verify", NULL, NULL, NULL };
884 prepare_alt_odb(the_repository);
885 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
886 child_process_init(&commit_graph_verify);
887 commit_graph_verify.argv = verify_argv;
888 commit_graph_verify.git_cmd = 1;
889 verify_argv[2] = "--object-dir";
890 verify_argv[3] = odb->path;
891 if (run_command(&commit_graph_verify))
892 errors_found |= ERROR_COMMIT_GRAPH;
896 if (!git_config_get_bool("core.multipackindex", &i) && i) {
897 struct child_process midx_verify = CHILD_PROCESS_INIT;
898 const char *midx_argv[] = { "multi-pack-index", "verify", NULL, NULL, NULL };
900 prepare_alt_odb(the_repository);
901 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
902 child_process_init(&midx_verify);
903 midx_verify.argv = midx_argv;
904 midx_verify.git_cmd = 1;
905 midx_argv[2] = "--object-dir";
906 midx_argv[3] = odb->path;
907 if (run_command(&midx_verify))
908 errors_found |= ERROR_COMMIT_GRAPH;