1 #define USE_THE_INDEX_COMPATIBILITY_MACROS
4 #include "repository.h"
12 #include "cache-tree.h"
13 #include "tree-walk.h"
15 #include "parse-options.h"
18 #include "streaming.h"
21 #include "object-store.h"
22 #include "run-command.h"
25 #define REACHABLE 0x0001
27 #define HAS_OBJ 0x0004
28 /* This flag is set if something points to this object. */
33 static int show_unreachable;
34 static int include_reflogs = 1;
35 static int check_full = 1;
36 static int connectivity_only;
37 static int check_strict;
38 static int keep_cache_objects;
39 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
40 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
41 static int errors_found;
42 static int write_lost_and_found;
44 static int show_progress = -1;
45 static int show_dangling = 1;
46 static int name_objects;
47 #define ERROR_OBJECT 01
48 #define ERROR_REACHABLE 02
50 #define ERROR_REFS 010
51 #define ERROR_COMMIT_GRAPH 020
52 #define ERROR_MULTI_PACK_INDEX 040
54 static const char *describe_object(const struct object_id *oid)
56 return fsck_describe_object(&fsck_walk_options, oid);
59 static const char *printable_type(const struct object_id *oid,
60 enum object_type type)
65 type = oid_object_info(the_repository, oid, NULL);
67 ret = type_name(type);
74 static int fsck_config(const char *var, const char *value, void *cb)
76 return fsck_config_internal(var, value, cb, &fsck_obj_options);
79 static int objerror(struct object *obj, const char *err)
81 errors_found |= ERROR_OBJECT;
82 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
83 fprintf_ln(stderr, _("error in %s %s: %s"),
84 printable_type(&obj->oid, obj->type),
85 describe_object(&obj->oid), err);
89 static int fsck_error_func(struct fsck_options *o,
90 const struct object_id *oid,
91 enum object_type object_type,
92 int msg_type, const char *message)
96 /* TRANSLATORS: e.g. warning in tree 01bfda: <more explanation> */
97 fprintf_ln(stderr, _("warning in %s %s: %s"),
98 printable_type(oid, object_type),
99 describe_object(oid), message);
102 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
103 fprintf_ln(stderr, _("error in %s %s: %s"),
104 printable_type(oid, object_type),
105 describe_object(oid), message);
108 BUG("%d (FSCK_IGNORE?) should never trigger this callback",
113 static struct object_array pending;
115 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
117 struct object *parent = data;
120 * The only case data is NULL or type is OBJ_ANY is when
121 * mark_object_reachable() calls us. All the callers of
122 * that function has non-NULL obj hence ...
125 /* ... these references to parent->fld are safe here */
126 printf_ln(_("broken link from %7s %s"),
127 printable_type(&parent->oid, parent->type),
128 describe_object(&parent->oid));
129 printf_ln(_("broken link from %7s %s"),
130 (type == OBJ_ANY ? _("unknown") : type_name(type)),
132 errors_found |= ERROR_REACHABLE;
136 if (type != OBJ_ANY && obj->type != type)
137 /* ... and the reference to parent is safe here */
138 objerror(parent, _("wrong object type in link"));
140 if (obj->flags & REACHABLE)
142 obj->flags |= REACHABLE;
144 if (is_promisor_object(&obj->oid))
146 * Further recursion does not need to be performed on this
147 * object since it is a promisor object (so it does not need to
148 * be added to "pending").
152 if (!(obj->flags & HAS_OBJ)) {
153 if (parent && !has_object(the_repository, &obj->oid, 1)) {
154 printf_ln(_("broken link from %7s %s\n"
156 printable_type(&parent->oid, parent->type),
157 describe_object(&parent->oid),
158 printable_type(&obj->oid, obj->type),
159 describe_object(&obj->oid));
160 errors_found |= ERROR_REACHABLE;
165 add_object_array(obj, NULL, &pending);
169 static void mark_object_reachable(struct object *obj)
171 mark_object(obj, OBJ_ANY, NULL, NULL);
174 static int traverse_one_object(struct object *obj)
176 int result = fsck_walk(obj, obj, &fsck_walk_options);
178 if (obj->type == OBJ_TREE) {
179 struct tree *tree = (struct tree *)obj;
180 free_tree_buffer(tree);
185 static int traverse_reachable(void)
187 struct progress *progress = NULL;
191 progress = start_delayed_progress(_("Checking connectivity"), 0);
193 result |= traverse_one_object(object_array_pop(&pending));
194 display_progress(progress, ++nr);
196 stop_progress(&progress);
200 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
208 static void mark_unreachable_referents(const struct object_id *oid)
210 struct fsck_options options = FSCK_OPTIONS_DEFAULT;
211 struct object *obj = lookup_object(the_repository, oid);
213 if (!obj || !(obj->flags & HAS_OBJ))
214 return; /* not part of our original set */
215 if (obj->flags & REACHABLE)
216 return; /* reachable objects already traversed */
219 * Avoid passing OBJ_NONE to fsck_walk, which will parse the object
220 * (and we want to avoid parsing blobs).
222 if (obj->type == OBJ_NONE) {
223 enum object_type type = oid_object_info(the_repository,
226 object_as_type(obj, type, 0);
229 options.walk = mark_used;
230 fsck_walk(obj, NULL, &options);
233 static int mark_loose_unreachable_referents(const struct object_id *oid,
237 mark_unreachable_referents(oid);
241 static int mark_packed_unreachable_referents(const struct object_id *oid,
242 struct packed_git *pack,
246 mark_unreachable_referents(oid);
251 * Check a single reachable object
253 static void check_reachable_object(struct object *obj)
256 * We obviously want the object to be parsed,
257 * except if it was in a pack-file and we didn't
260 if (!(obj->flags & HAS_OBJ)) {
261 if (is_promisor_object(&obj->oid))
263 if (has_object_pack(&obj->oid))
264 return; /* it is in pack - forget about it */
265 printf_ln(_("missing %s %s"),
266 printable_type(&obj->oid, obj->type),
267 describe_object(&obj->oid));
268 errors_found |= ERROR_REACHABLE;
274 * Check a single unreachable object
276 static void check_unreachable_object(struct object *obj)
279 * Missing unreachable object? Ignore it. It's not like
280 * we miss it (since it can't be reached), nor do we want
281 * to complain about it being unreachable (since it does
284 if (!(obj->flags & HAS_OBJ))
288 * Unreachable object that exists? Show it if asked to,
289 * since this is something that is prunable.
291 if (show_unreachable) {
292 printf_ln(_("unreachable %s %s"),
293 printable_type(&obj->oid, obj->type),
294 describe_object(&obj->oid));
299 * "!USED" means that nothing at all points to it, including
300 * other unreachable objects. In other words, it's the "tip"
301 * of some set of unreachable objects, usually a commit that
304 * Such starting points are more interesting than some random
305 * set of unreachable objects, so we show them even if the user
306 * hasn't asked for _all_ unreachable objects. If you have
307 * deleted a branch by mistake, this is a prime candidate to
308 * start looking at, for example.
310 if (!(obj->flags & USED)) {
312 printf_ln(_("dangling %s %s"),
313 printable_type(&obj->oid, obj->type),
314 describe_object(&obj->oid));
315 if (write_lost_and_found) {
316 char *filename = git_pathdup("lost-found/%s/%s",
317 obj->type == OBJ_COMMIT ? "commit" : "other",
318 describe_object(&obj->oid));
321 if (safe_create_leading_directories_const(filename)) {
322 error(_("could not create lost-found"));
326 f = xfopen(filename, "w");
327 if (obj->type == OBJ_BLOB) {
328 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
329 die_errno(_("could not write '%s'"), filename);
331 fprintf(f, "%s\n", describe_object(&obj->oid));
333 die_errno(_("could not finish '%s'"),
341 * Otherwise? It's there, it's unreachable, and some other unreachable
342 * object points to it. Ignore it - it's not interesting, and we showed
343 * all the interesting cases above.
347 static void check_object(struct object *obj)
350 fprintf_ln(stderr, _("Checking %s"), describe_object(&obj->oid));
352 if (obj->flags & REACHABLE)
353 check_reachable_object(obj);
355 check_unreachable_object(obj);
358 static void check_connectivity(void)
362 /* Traverse the pending reachable objects */
363 traverse_reachable();
366 * With --connectivity-only, we won't have actually opened and marked
367 * unreachable objects with USED. Do that now to make --dangling, etc
370 if (connectivity_only && (show_dangling || write_lost_and_found)) {
372 * Even though we already have a "struct object" for each of
373 * these in memory, we must not iterate over the internal
374 * object hash as we do below. Our loop would potentially
375 * resize the hash, making our iteration invalid.
377 * Instead, we'll just go back to the source list of objects,
378 * and ignore any that weren't present in our earlier
381 for_each_loose_object(mark_loose_unreachable_referents, NULL, 0);
382 for_each_packed_object(mark_packed_unreachable_referents, NULL, 0);
385 /* Look up all the requirements, warn about missing objects.. */
386 max = get_max_object_index();
388 fprintf_ln(stderr, _("Checking connectivity (%d objects)"), max);
390 for (i = 0; i < max; i++) {
391 struct object *obj = get_indexed_object(i);
398 static int fsck_obj(struct object *obj, void *buffer, unsigned long size)
402 if (obj->flags & SEEN)
407 fprintf_ln(stderr, _("Checking %s %s"),
408 printable_type(&obj->oid, obj->type),
409 describe_object(&obj->oid));
411 if (fsck_walk(obj, NULL, &fsck_obj_options))
412 objerror(obj, _("broken links"));
413 err = fsck_object(obj, buffer, size, &fsck_obj_options);
417 if (obj->type == OBJ_COMMIT) {
418 struct commit *commit = (struct commit *) obj;
420 if (!commit->parents && show_root)
421 printf_ln(_("root %s"),
422 describe_object(&commit->object.oid));
425 if (obj->type == OBJ_TAG) {
426 struct tag *tag = (struct tag *) obj;
428 if (show_tags && tag->tagged) {
429 printf_ln(_("tagged %s %s (%s) in %s"),
430 printable_type(&tag->tagged->oid, tag->tagged->type),
431 describe_object(&tag->tagged->oid),
433 describe_object(&tag->object.oid));
438 if (obj->type == OBJ_TREE)
439 free_tree_buffer((struct tree *)obj);
440 if (obj->type == OBJ_COMMIT)
441 free_commit_buffer(the_repository->parsed_objects,
442 (struct commit *)obj);
446 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
447 unsigned long size, void *buffer, int *eaten)
450 * Note, buffer may be NULL if type is OBJ_BLOB. See
451 * verify_packfile(), data_valid variable for details.
454 obj = parse_object_buffer(the_repository, oid, type, size, buffer,
457 errors_found |= ERROR_OBJECT;
458 return error(_("%s: object corrupt or missing"),
461 obj->flags &= ~(REACHABLE | SEEN);
462 obj->flags |= HAS_OBJ;
463 return fsck_obj(obj, buffer, size);
466 static int default_refs;
468 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
469 timestamp_t timestamp)
473 if (!is_null_oid(oid)) {
474 obj = lookup_object(the_repository, oid);
475 if (obj && (obj->flags & HAS_OBJ)) {
477 fsck_put_object_name(&fsck_walk_options, oid,
481 mark_object_reachable(obj);
482 } else if (!is_promisor_object(oid)) {
483 error(_("%s: invalid reflog entry %s"),
484 refname, oid_to_hex(oid));
485 errors_found |= ERROR_REACHABLE;
490 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
491 const char *email, timestamp_t timestamp, int tz,
492 const char *message, void *cb_data)
494 const char *refname = cb_data;
497 fprintf_ln(stderr, _("Checking reflog %s->%s"),
498 oid_to_hex(ooid), oid_to_hex(noid));
500 fsck_handle_reflog_oid(refname, ooid, 0);
501 fsck_handle_reflog_oid(refname, noid, timestamp);
505 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
506 int flag, void *cb_data)
508 struct strbuf refname = STRBUF_INIT;
510 strbuf_worktree_ref(cb_data, &refname, logname);
511 for_each_reflog_ent(refname.buf, fsck_handle_reflog_ent, refname.buf);
512 strbuf_release(&refname);
516 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
517 int flag, void *cb_data)
521 obj = parse_object(the_repository, oid);
523 if (is_promisor_object(oid)) {
525 * Increment default_refs anyway, because this is a
531 error(_("%s: invalid sha1 pointer %s"),
532 refname, oid_to_hex(oid));
533 errors_found |= ERROR_REACHABLE;
534 /* We'll continue with the rest despite the error.. */
537 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
538 error(_("%s: not a commit"), refname);
539 errors_found |= ERROR_REFS;
543 fsck_put_object_name(&fsck_walk_options,
545 mark_object_reachable(obj);
550 static int fsck_head_link(const char *head_ref_name,
551 const char **head_points_at,
552 struct object_id *head_oid);
554 static void get_default_heads(void)
556 struct worktree **worktrees, **p;
557 const char *head_points_at;
558 struct object_id head_oid;
560 for_each_rawref(fsck_handle_ref, NULL);
562 worktrees = get_worktrees();
563 for (p = worktrees; *p; p++) {
564 struct worktree *wt = *p;
565 struct strbuf ref = STRBUF_INIT;
567 strbuf_worktree_ref(wt, &ref, "HEAD");
568 fsck_head_link(ref.buf, &head_points_at, &head_oid);
569 if (head_points_at && !is_null_oid(&head_oid))
570 fsck_handle_ref(ref.buf, &head_oid, 0, NULL);
571 strbuf_release(&ref);
574 refs_for_each_reflog(get_worktree_ref_store(wt),
575 fsck_handle_reflog, wt);
577 free_worktrees(worktrees);
580 * Not having any default heads isn't really fatal, but
581 * it does mean that "--unreachable" no longer makes any
582 * sense (since in this case everything will obviously
583 * be unreachable by definition.
585 * Showing dangling objects is valid, though (as those
586 * dangling objects are likely lost heads).
588 * So we just print a warning about it, and clear the
589 * "show_unreachable" flag.
592 fprintf_ln(stderr, _("notice: No default references"));
593 show_unreachable = 0;
597 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
600 enum object_type type;
605 if (read_loose_object(path, oid, &type, &size, &contents) < 0) {
606 errors_found |= ERROR_OBJECT;
607 error(_("%s: object corrupt or missing: %s"),
608 oid_to_hex(oid), path);
609 return 0; /* keep checking other objects */
612 if (!contents && type != OBJ_BLOB)
613 BUG("read_loose_object streamed a non-blob");
615 obj = parse_object_buffer(the_repository, oid, type, size,
619 errors_found |= ERROR_OBJECT;
620 error(_("%s: object could not be parsed: %s"),
621 oid_to_hex(oid), path);
624 return 0; /* keep checking other objects */
627 obj->flags &= ~(REACHABLE | SEEN);
628 obj->flags |= HAS_OBJ;
629 if (fsck_obj(obj, contents, size))
630 errors_found |= ERROR_OBJECT;
634 return 0; /* keep checking other objects, even if we saw an error */
637 static int fsck_cruft(const char *basename, const char *path, void *data)
639 if (!starts_with(basename, "tmp_obj_"))
640 fprintf_ln(stderr, _("bad sha1 file: %s"), path);
644 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
646 display_progress(progress, nr + 1);
650 static void fsck_object_dir(const char *path)
652 struct progress *progress = NULL;
655 fprintf_ln(stderr, _("Checking object directory"));
658 progress = start_progress(_("Checking object directories"), 256);
660 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
662 display_progress(progress, 256);
663 stop_progress(&progress);
666 static int fsck_head_link(const char *head_ref_name,
667 const char **head_points_at,
668 struct object_id *head_oid)
670 int null_is_error = 0;
673 fprintf_ln(stderr, _("Checking %s link"), head_ref_name);
675 *head_points_at = resolve_ref_unsafe(head_ref_name, 0, head_oid, NULL);
676 if (!*head_points_at) {
677 errors_found |= ERROR_REFS;
678 return error(_("invalid %s"), head_ref_name);
680 if (!strcmp(*head_points_at, head_ref_name))
683 else if (!starts_with(*head_points_at, "refs/heads/")) {
684 errors_found |= ERROR_REFS;
685 return error(_("%s points to something strange (%s)"),
686 head_ref_name, *head_points_at);
688 if (is_null_oid(head_oid)) {
690 errors_found |= ERROR_REFS;
691 return error(_("%s: detached HEAD points at nothing"),
695 _("notice: %s points to an unborn branch (%s)"),
696 head_ref_name, *head_points_at + 11);
701 static int fsck_cache_tree(struct cache_tree *it)
707 fprintf_ln(stderr, _("Checking cache tree"));
709 if (0 <= it->entry_count) {
710 struct object *obj = parse_object(the_repository, &it->oid);
712 error(_("%s: invalid sha1 pointer in cache-tree"),
713 oid_to_hex(&it->oid));
714 errors_found |= ERROR_REFS;
718 fsck_put_object_name(&fsck_walk_options, &it->oid, ":");
719 mark_object_reachable(obj);
720 if (obj->type != OBJ_TREE)
721 err |= objerror(obj, _("non-tree in cache-tree"));
723 for (i = 0; i < it->subtree_nr; i++)
724 err |= fsck_cache_tree(it->down[i]->cache_tree);
728 static void mark_object_for_connectivity(const struct object_id *oid)
730 struct object *obj = lookup_unknown_object(oid);
731 obj->flags |= HAS_OBJ;
734 static int mark_loose_for_connectivity(const struct object_id *oid,
738 mark_object_for_connectivity(oid);
742 static int mark_packed_for_connectivity(const struct object_id *oid,
743 struct packed_git *pack,
747 mark_object_for_connectivity(oid);
751 static char const * const fsck_usage[] = {
752 N_("git fsck [<options>] [<object>...]"),
756 static struct option fsck_opts[] = {
757 OPT__VERBOSE(&verbose, N_("be verbose")),
758 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
759 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
760 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
761 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
762 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
763 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
764 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
765 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
766 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
767 OPT_BOOL(0, "lost-found", &write_lost_and_found,
768 N_("write dangling objects in .git/lost-found")),
769 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
770 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
774 int cmd_fsck(int argc, const char **argv, const char *prefix)
777 struct object_directory *odb;
779 /* fsck knows how to handle missing promisor objects */
780 fetch_if_missing = 0;
783 read_replace_refs = 0;
785 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
787 fsck_walk_options.walk = mark_object;
788 fsck_obj_options.walk = mark_used;
789 fsck_obj_options.error_func = fsck_error_func;
791 fsck_obj_options.strict = 1;
793 if (show_progress == -1)
794 show_progress = isatty(2);
798 if (write_lost_and_found) {
804 fsck_enable_object_names(&fsck_walk_options);
806 git_config(fsck_config, NULL);
808 if (connectivity_only) {
809 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
810 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
812 prepare_alt_odb(the_repository);
813 for (odb = the_repository->objects->odb; odb; odb = odb->next)
814 fsck_object_dir(odb->path);
817 struct packed_git *p;
818 uint32_t total = 0, count = 0;
819 struct progress *progress = NULL;
822 for (p = get_all_packs(the_repository); p;
824 if (open_pack_index(p))
826 total += p->num_objects;
829 progress = start_progress(_("Checking objects"), total);
831 for (p = get_all_packs(the_repository); p;
833 /* verify gives error messages itself */
834 if (verify_pack(the_repository,
837 errors_found |= ERROR_PACK;
838 count += p->num_objects;
840 stop_progress(&progress);
843 if (fsck_finish(&fsck_obj_options))
844 errors_found |= ERROR_OBJECT;
847 for (i = 0; i < argc; i++) {
848 const char *arg = argv[i];
849 struct object_id oid;
850 if (!get_oid(arg, &oid)) {
851 struct object *obj = lookup_object(the_repository,
854 if (!obj || !(obj->flags & HAS_OBJ)) {
855 if (is_promisor_object(&oid))
857 error(_("%s: object missing"), oid_to_hex(&oid));
858 errors_found |= ERROR_OBJECT;
863 fsck_put_object_name(&fsck_walk_options, &oid,
865 mark_object_reachable(obj);
868 error(_("invalid parameter: expected sha1, got '%s'"), arg);
869 errors_found |= ERROR_OBJECT;
873 * If we've not been given any explicit head information, do the
874 * default ones from .git/refs. We also consider the index file
875 * in this case (ie this implies --cache).
879 keep_cache_objects = 1;
882 if (keep_cache_objects) {
883 verify_index_checksum = 1;
886 for (i = 0; i < active_nr; i++) {
891 mode = active_cache[i]->ce_mode;
892 if (S_ISGITLINK(mode))
894 blob = lookup_blob(the_repository,
895 &active_cache[i]->oid);
900 fsck_put_object_name(&fsck_walk_options, &obj->oid,
901 ":%s", active_cache[i]->name);
902 mark_object_reachable(obj);
904 if (active_cache_tree)
905 fsck_cache_tree(active_cache_tree);
908 check_connectivity();
910 if (!git_config_get_bool("core.commitgraph", &i) && i) {
911 struct child_process commit_graph_verify = CHILD_PROCESS_INIT;
912 const char *verify_argv[] = { "commit-graph", "verify", NULL, NULL, NULL };
914 prepare_alt_odb(the_repository);
915 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
916 child_process_init(&commit_graph_verify);
917 commit_graph_verify.argv = verify_argv;
918 commit_graph_verify.git_cmd = 1;
919 verify_argv[2] = "--object-dir";
920 verify_argv[3] = odb->path;
921 if (run_command(&commit_graph_verify))
922 errors_found |= ERROR_COMMIT_GRAPH;
926 if (!git_config_get_bool("core.multipackindex", &i) && i) {
927 struct child_process midx_verify = CHILD_PROCESS_INIT;
928 const char *midx_argv[] = { "multi-pack-index", "verify", NULL, NULL, NULL };
930 prepare_alt_odb(the_repository);
931 for (odb = the_repository->objects->odb; odb; odb = odb->next) {
932 child_process_init(&midx_verify);
933 midx_verify.argv = midx_argv;
934 midx_verify.git_cmd = 1;
935 midx_argv[2] = "--object-dir";
936 midx_argv[3] = odb->path;
937 if (run_command(&midx_verify))
938 errors_found |= ERROR_MULTI_PACK_INDEX;