sha1_file: add repository argument to prepare_alt_odb
[git.git] / builtin / fsck.c
blob3ef25fab9712a12edd70700b941dd62203de9d4a
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 = sha1_object_info(obj->oid.hash, NULL);
71 if (type > 0)
72 object_as_type(obj, type, 0);
75 ret = typename(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" : typename(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 return fsck_walk(obj, obj, &fsck_walk_options);
188 static int traverse_reachable(void)
190 struct progress *progress = NULL;
191 unsigned int nr = 0;
192 int result = 0;
193 if (show_progress)
194 progress = start_delayed_progress(_("Checking connectivity"), 0);
195 while (pending.nr) {
196 result |= traverse_one_object(object_array_pop(&pending));
197 display_progress(progress, ++nr);
199 stop_progress(&progress);
200 return !!result;
203 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
205 if (!obj)
206 return 1;
207 obj->flags |= USED;
208 return 0;
212 * Check a single reachable object
214 static void check_reachable_object(struct object *obj)
217 * We obviously want the object to be parsed,
218 * except if it was in a pack-file and we didn't
219 * do a full fsck
221 if (!(obj->flags & HAS_OBJ)) {
222 if (is_promisor_object(&obj->oid))
223 return;
224 if (has_sha1_pack(obj->oid.hash))
225 return; /* it is in pack - forget about it */
226 printf("missing %s %s\n", printable_type(obj),
227 describe_object(obj));
228 errors_found |= ERROR_REACHABLE;
229 return;
234 * Check a single unreachable object
236 static void check_unreachable_object(struct object *obj)
239 * Missing unreachable object? Ignore it. It's not like
240 * we miss it (since it can't be reached), nor do we want
241 * to complain about it being unreachable (since it does
242 * not exist).
244 if (!(obj->flags & HAS_OBJ))
245 return;
248 * Unreachable object that exists? Show it if asked to,
249 * since this is something that is prunable.
251 if (show_unreachable) {
252 printf("unreachable %s %s\n", printable_type(obj),
253 describe_object(obj));
254 return;
258 * "!USED" means that nothing at all points to it, including
259 * other unreachable objects. In other words, it's the "tip"
260 * of some set of unreachable objects, usually a commit that
261 * got dropped.
263 * Such starting points are more interesting than some random
264 * set of unreachable objects, so we show them even if the user
265 * hasn't asked for _all_ unreachable objects. If you have
266 * deleted a branch by mistake, this is a prime candidate to
267 * start looking at, for example.
269 if (!(obj->flags & USED)) {
270 if (show_dangling)
271 printf("dangling %s %s\n", printable_type(obj),
272 describe_object(obj));
273 if (write_lost_and_found) {
274 char *filename = git_pathdup("lost-found/%s/%s",
275 obj->type == OBJ_COMMIT ? "commit" : "other",
276 describe_object(obj));
277 FILE *f;
279 if (safe_create_leading_directories_const(filename)) {
280 error("Could not create lost-found");
281 free(filename);
282 return;
284 f = xfopen(filename, "w");
285 if (obj->type == OBJ_BLOB) {
286 if (stream_blob_to_fd(fileno(f), &obj->oid, NULL, 1))
287 die_errno("Could not write '%s'", filename);
288 } else
289 fprintf(f, "%s\n", describe_object(obj));
290 if (fclose(f))
291 die_errno("Could not finish '%s'",
292 filename);
293 free(filename);
295 return;
299 * Otherwise? It's there, it's unreachable, and some other unreachable
300 * object points to it. Ignore it - it's not interesting, and we showed
301 * all the interesting cases above.
305 static void check_object(struct object *obj)
307 if (verbose)
308 fprintf(stderr, "Checking %s\n", describe_object(obj));
310 if (obj->flags & REACHABLE)
311 check_reachable_object(obj);
312 else
313 check_unreachable_object(obj);
316 static void check_connectivity(void)
318 int i, max;
320 /* Traverse the pending reachable objects */
321 traverse_reachable();
323 /* Look up all the requirements, warn about missing objects.. */
324 max = get_max_object_index();
325 if (verbose)
326 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
328 for (i = 0; i < max; i++) {
329 struct object *obj = get_indexed_object(i);
331 if (obj)
332 check_object(obj);
336 static int fsck_obj(struct object *obj)
338 int err;
340 if (obj->flags & SEEN)
341 return 0;
342 obj->flags |= SEEN;
344 if (verbose)
345 fprintf(stderr, "Checking %s %s\n",
346 printable_type(obj), describe_object(obj));
348 if (fsck_walk(obj, NULL, &fsck_obj_options))
349 objerror(obj, "broken links");
350 err = fsck_object(obj, NULL, 0, &fsck_obj_options);
351 if (err)
352 goto out;
354 if (obj->type == OBJ_COMMIT) {
355 struct commit *commit = (struct commit *) obj;
357 if (!commit->parents && show_root)
358 printf("root %s\n", describe_object(&commit->object));
361 if (obj->type == OBJ_TAG) {
362 struct tag *tag = (struct tag *) obj;
364 if (show_tags && tag->tagged) {
365 printf("tagged %s %s", printable_type(tag->tagged),
366 describe_object(tag->tagged));
367 printf(" (%s) in %s\n", tag->tag,
368 describe_object(&tag->object));
372 out:
373 if (obj->type == OBJ_TREE)
374 free_tree_buffer((struct tree *)obj);
375 if (obj->type == OBJ_COMMIT)
376 free_commit_buffer((struct commit *)obj);
377 return err;
380 static int fsck_obj_buffer(const struct object_id *oid, enum object_type type,
381 unsigned long size, void *buffer, int *eaten)
384 * Note, buffer may be NULL if type is OBJ_BLOB. See
385 * verify_packfile(), data_valid variable for details.
387 struct object *obj;
388 obj = parse_object_buffer(oid, type, size, buffer, eaten);
389 if (!obj) {
390 errors_found |= ERROR_OBJECT;
391 return error("%s: object corrupt or missing", oid_to_hex(oid));
393 obj->flags &= ~(REACHABLE | SEEN);
394 obj->flags |= HAS_OBJ;
395 return fsck_obj(obj);
398 static int default_refs;
400 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
401 timestamp_t timestamp)
403 struct object *obj;
405 if (!is_null_oid(oid)) {
406 obj = lookup_object(oid->hash);
407 if (obj && (obj->flags & HAS_OBJ)) {
408 if (timestamp && name_objects)
409 add_decoration(fsck_walk_options.object_names,
410 obj,
411 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
412 obj->flags |= USED;
413 mark_object_reachable(obj);
414 } else if (!is_promisor_object(oid)) {
415 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
416 errors_found |= ERROR_REACHABLE;
421 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
422 const char *email, timestamp_t timestamp, int tz,
423 const char *message, void *cb_data)
425 const char *refname = cb_data;
427 if (verbose)
428 fprintf(stderr, "Checking reflog %s->%s\n",
429 oid_to_hex(ooid), oid_to_hex(noid));
431 fsck_handle_reflog_oid(refname, ooid, 0);
432 fsck_handle_reflog_oid(refname, noid, timestamp);
433 return 0;
436 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
437 int flag, void *cb_data)
439 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
440 return 0;
443 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
444 int flag, void *cb_data)
446 struct object *obj;
448 obj = parse_object(oid);
449 if (!obj) {
450 if (is_promisor_object(oid)) {
452 * Increment default_refs anyway, because this is a
453 * valid ref.
455 default_refs++;
456 return 0;
458 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
459 errors_found |= ERROR_REACHABLE;
460 /* We'll continue with the rest despite the error.. */
461 return 0;
463 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
464 error("%s: not a commit", refname);
465 errors_found |= ERROR_REFS;
467 default_refs++;
468 obj->flags |= USED;
469 if (name_objects)
470 add_decoration(fsck_walk_options.object_names,
471 obj, xstrdup(refname));
472 mark_object_reachable(obj);
474 return 0;
477 static void get_default_heads(void)
479 if (head_points_at && !is_null_oid(&head_oid))
480 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
481 for_each_rawref(fsck_handle_ref, NULL);
482 if (include_reflogs)
483 for_each_reflog(fsck_handle_reflog, NULL);
486 * Not having any default heads isn't really fatal, but
487 * it does mean that "--unreachable" no longer makes any
488 * sense (since in this case everything will obviously
489 * be unreachable by definition.
491 * Showing dangling objects is valid, though (as those
492 * dangling objects are likely lost heads).
494 * So we just print a warning about it, and clear the
495 * "show_unreachable" flag.
497 if (!default_refs) {
498 fprintf(stderr, "notice: No default references\n");
499 show_unreachable = 0;
503 static struct object *parse_loose_object(const struct object_id *oid,
504 const char *path)
506 struct object *obj;
507 void *contents;
508 enum object_type type;
509 unsigned long size;
510 int eaten;
512 if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
513 return NULL;
515 if (!contents && type != OBJ_BLOB)
516 die("BUG: read_loose_object streamed a non-blob");
518 obj = parse_object_buffer(oid, type, size, contents, &eaten);
520 if (!eaten)
521 free(contents);
522 return obj;
525 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
527 struct object *obj = parse_loose_object(oid, path);
529 if (!obj) {
530 errors_found |= ERROR_OBJECT;
531 error("%s: object corrupt or missing: %s",
532 oid_to_hex(oid), path);
533 return 0; /* keep checking other objects */
536 obj->flags &= ~(REACHABLE | SEEN);
537 obj->flags |= HAS_OBJ;
538 if (fsck_obj(obj))
539 errors_found |= ERROR_OBJECT;
540 return 0;
543 static int fsck_cruft(const char *basename, const char *path, void *data)
545 if (!starts_with(basename, "tmp_obj_"))
546 fprintf(stderr, "bad sha1 file: %s\n", path);
547 return 0;
550 static int fsck_subdir(unsigned int nr, const char *path, void *progress)
552 display_progress(progress, nr + 1);
553 return 0;
556 static void fsck_object_dir(const char *path)
558 struct progress *progress = NULL;
560 if (verbose)
561 fprintf(stderr, "Checking object directory\n");
563 if (show_progress)
564 progress = start_progress(_("Checking object directories"), 256);
566 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
567 progress);
568 display_progress(progress, 256);
569 stop_progress(&progress);
572 static int fsck_head_link(void)
574 int null_is_error = 0;
576 if (verbose)
577 fprintf(stderr, "Checking HEAD link\n");
579 head_points_at = resolve_ref_unsafe("HEAD", 0, &head_oid, NULL);
580 if (!head_points_at) {
581 errors_found |= ERROR_REFS;
582 return error("Invalid HEAD");
584 if (!strcmp(head_points_at, "HEAD"))
585 /* detached HEAD */
586 null_is_error = 1;
587 else if (!starts_with(head_points_at, "refs/heads/")) {
588 errors_found |= ERROR_REFS;
589 return error("HEAD points to something strange (%s)",
590 head_points_at);
592 if (is_null_oid(&head_oid)) {
593 if (null_is_error) {
594 errors_found |= ERROR_REFS;
595 return error("HEAD: detached HEAD points at nothing");
597 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
598 head_points_at + 11);
600 return 0;
603 static int fsck_cache_tree(struct cache_tree *it)
605 int i;
606 int err = 0;
608 if (verbose)
609 fprintf(stderr, "Checking cache tree\n");
611 if (0 <= it->entry_count) {
612 struct object *obj = parse_object(&it->oid);
613 if (!obj) {
614 error("%s: invalid sha1 pointer in cache-tree",
615 oid_to_hex(&it->oid));
616 errors_found |= ERROR_REFS;
617 return 1;
619 obj->flags |= USED;
620 if (name_objects)
621 add_decoration(fsck_walk_options.object_names,
622 obj, xstrdup(":"));
623 mark_object_reachable(obj);
624 if (obj->type != OBJ_TREE)
625 err |= objerror(obj, "non-tree in cache-tree");
627 for (i = 0; i < it->subtree_nr; i++)
628 err |= fsck_cache_tree(it->down[i]->cache_tree);
629 return err;
632 static void mark_object_for_connectivity(const struct object_id *oid)
634 struct object *obj = lookup_unknown_object(oid->hash);
635 obj->flags |= HAS_OBJ;
638 static int mark_loose_for_connectivity(const struct object_id *oid,
639 const char *path,
640 void *data)
642 mark_object_for_connectivity(oid);
643 return 0;
646 static int mark_packed_for_connectivity(const struct object_id *oid,
647 struct packed_git *pack,
648 uint32_t pos,
649 void *data)
651 mark_object_for_connectivity(oid);
652 return 0;
655 static char const * const fsck_usage[] = {
656 N_("git fsck [<options>] [<object>...]"),
657 NULL
660 static struct option fsck_opts[] = {
661 OPT__VERBOSE(&verbose, N_("be verbose")),
662 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
663 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
664 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
665 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
666 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
667 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
668 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
669 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
670 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
671 OPT_BOOL(0, "lost-found", &write_lost_and_found,
672 N_("write dangling objects in .git/lost-found")),
673 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
674 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
675 OPT_END(),
678 int cmd_fsck(int argc, const char **argv, const char *prefix)
680 int i;
681 struct alternate_object_database *alt;
683 /* fsck knows how to handle missing promisor objects */
684 fetch_if_missing = 0;
686 errors_found = 0;
687 check_replace_refs = 0;
689 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
691 fsck_walk_options.walk = mark_object;
692 fsck_obj_options.walk = mark_used;
693 fsck_obj_options.error_func = fsck_error_func;
694 if (check_strict)
695 fsck_obj_options.strict = 1;
697 if (show_progress == -1)
698 show_progress = isatty(2);
699 if (verbose)
700 show_progress = 0;
702 if (write_lost_and_found) {
703 check_full = 1;
704 include_reflogs = 0;
707 if (name_objects)
708 fsck_walk_options.object_names =
709 xcalloc(1, sizeof(struct decoration));
711 git_config(fsck_config, NULL);
713 fsck_head_link();
714 if (connectivity_only) {
715 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
716 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
717 } else {
718 struct alternate_object_database *alt_odb_list;
720 fsck_object_dir(get_object_directory());
722 prepare_alt_odb(the_repository);
723 alt_odb_list = the_repository->objects->alt_odb_list;
724 for (alt = alt_odb_list; alt; alt = alt->next)
725 fsck_object_dir(alt->path);
727 if (check_full) {
728 struct packed_git *p;
729 uint32_t total = 0, count = 0;
730 struct progress *progress = NULL;
732 prepare_packed_git();
734 if (show_progress) {
735 for (p = get_packed_git(the_repository); p;
736 p = p->next) {
737 if (open_pack_index(p))
738 continue;
739 total += p->num_objects;
742 progress = start_progress(_("Checking objects"), total);
744 for (p = get_packed_git(the_repository); p;
745 p = p->next) {
746 /* verify gives error messages itself */
747 if (verify_pack(p, fsck_obj_buffer,
748 progress, count))
749 errors_found |= ERROR_PACK;
750 count += p->num_objects;
752 stop_progress(&progress);
756 for (i = 0; i < argc; i++) {
757 const char *arg = argv[i];
758 struct object_id oid;
759 if (!get_oid(arg, &oid)) {
760 struct object *obj = lookup_object(oid.hash);
762 if (!obj || !(obj->flags & HAS_OBJ)) {
763 if (is_promisor_object(&oid))
764 continue;
765 error("%s: object missing", oid_to_hex(&oid));
766 errors_found |= ERROR_OBJECT;
767 continue;
770 obj->flags |= USED;
771 if (name_objects)
772 add_decoration(fsck_walk_options.object_names,
773 obj, xstrdup(arg));
774 mark_object_reachable(obj);
775 continue;
777 error("invalid parameter: expected sha1, got '%s'", arg);
778 errors_found |= ERROR_OBJECT;
782 * If we've not been given any explicit head information, do the
783 * default ones from .git/refs. We also consider the index file
784 * in this case (ie this implies --cache).
786 if (!argc) {
787 get_default_heads();
788 keep_cache_objects = 1;
791 if (keep_cache_objects) {
792 verify_index_checksum = 1;
793 verify_ce_order = 1;
794 read_cache();
795 for (i = 0; i < active_nr; i++) {
796 unsigned int mode;
797 struct blob *blob;
798 struct object *obj;
800 mode = active_cache[i]->ce_mode;
801 if (S_ISGITLINK(mode))
802 continue;
803 blob = lookup_blob(&active_cache[i]->oid);
804 if (!blob)
805 continue;
806 obj = &blob->object;
807 obj->flags |= USED;
808 if (name_objects)
809 add_decoration(fsck_walk_options.object_names,
810 obj,
811 xstrfmt(":%s", active_cache[i]->name));
812 mark_object_reachable(obj);
814 if (active_cache_tree)
815 fsck_cache_tree(active_cache_tree);
818 check_connectivity();
819 return errors_found;