packed_ref_store: new struct
[git/debian.git] / builtin / fsck.c
blobcb2ba6cd1be46635ca8416a7d3f2b006f964190b
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 #include "decorate.h"
18 #define REACHABLE 0x0001
19 #define SEEN 0x0002
20 #define HAS_OBJ 0x0004
22 static int show_root;
23 static int show_tags;
24 static int show_unreachable;
25 static int include_reflogs = 1;
26 static int check_full = 1;
27 static int connectivity_only;
28 static int check_strict;
29 static int keep_cache_objects;
30 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
31 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
32 static struct object_id head_oid;
33 static const char *head_points_at;
34 static int errors_found;
35 static int write_lost_and_found;
36 static int verbose;
37 static int show_progress = -1;
38 static int show_dangling = 1;
39 static int name_objects;
40 #define ERROR_OBJECT 01
41 #define ERROR_REACHABLE 02
42 #define ERROR_PACK 04
43 #define ERROR_REFS 010
45 static const char *describe_object(struct object *obj)
47 static struct strbuf buf = STRBUF_INIT;
48 char *name = name_objects ?
49 lookup_decoration(fsck_walk_options.object_names, obj) : NULL;
51 strbuf_reset(&buf);
52 strbuf_addstr(&buf, oid_to_hex(&obj->oid));
53 if (name)
54 strbuf_addf(&buf, " (%s)", name);
56 return buf.buf;
59 static const char *printable_type(struct object *obj)
61 const char *ret;
63 if (obj->type == OBJ_NONE) {
64 enum object_type type = sha1_object_info(obj->oid.hash, NULL);
65 if (type > 0)
66 object_as_type(obj, type, 0);
69 ret = typename(obj->type);
70 if (!ret)
71 ret = "unknown";
73 return ret;
76 static int fsck_config(const char *var, const char *value, void *cb)
78 if (strcmp(var, "fsck.skiplist") == 0) {
79 const char *path;
80 struct strbuf sb = STRBUF_INIT;
82 if (git_config_pathname(&path, var, value))
83 return 1;
84 strbuf_addf(&sb, "skiplist=%s", path);
85 free((char *)path);
86 fsck_set_msg_types(&fsck_obj_options, sb.buf);
87 strbuf_release(&sb);
88 return 0;
91 if (skip_prefix(var, "fsck.", &var)) {
92 fsck_set_msg_type(&fsck_obj_options, var, value);
93 return 0;
96 return git_default_config(var, value, cb);
99 static void objreport(struct object *obj, const char *msg_type,
100 const char *err)
102 fprintf(stderr, "%s in %s %s: %s\n",
103 msg_type, printable_type(obj), describe_object(obj), err);
106 static int objerror(struct object *obj, const char *err)
108 errors_found |= ERROR_OBJECT;
109 objreport(obj, "error", err);
110 return -1;
113 static int fsck_error_func(struct fsck_options *o,
114 struct object *obj, int type, const char *message)
116 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
117 return (type == FSCK_WARN) ? 0 : 1;
120 static struct object_array pending;
122 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
124 struct object *parent = data;
127 * The only case data is NULL or type is OBJ_ANY is when
128 * mark_object_reachable() calls us. All the callers of
129 * that function has non-NULL obj hence ...
131 if (!obj) {
132 /* ... these references to parent->fld are safe here */
133 printf("broken link from %7s %s\n",
134 printable_type(parent), describe_object(parent));
135 printf("broken link from %7s %s\n",
136 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
137 errors_found |= ERROR_REACHABLE;
138 return 1;
141 if (type != OBJ_ANY && obj->type != type)
142 /* ... and the reference to parent is safe here */
143 objerror(parent, "wrong object type in link");
145 if (obj->flags & REACHABLE)
146 return 0;
147 obj->flags |= REACHABLE;
148 if (!(obj->flags & HAS_OBJ)) {
149 if (parent && !has_object_file(&obj->oid)) {
150 printf("broken link from %7s %s\n",
151 printable_type(parent), describe_object(parent));
152 printf(" to %7s %s\n",
153 printable_type(obj), describe_object(obj));
154 errors_found |= ERROR_REACHABLE;
156 return 1;
159 add_object_array(obj, NULL, &pending);
160 return 0;
163 static void mark_object_reachable(struct object *obj)
165 mark_object(obj, OBJ_ANY, NULL, NULL);
168 static int traverse_one_object(struct object *obj)
170 int result;
171 struct tree *tree = NULL;
173 if (obj->type == OBJ_TREE) {
174 tree = (struct tree *)obj;
175 if (parse_tree(tree) < 0)
176 return 1; /* error already displayed */
178 result = fsck_walk(obj, obj, &fsck_walk_options);
179 if (tree)
180 free_tree_buffer(tree);
181 return result;
184 static int traverse_reachable(void)
186 struct progress *progress = NULL;
187 unsigned int nr = 0;
188 int result = 0;
189 if (show_progress)
190 progress = start_progress_delay(_("Checking connectivity"), 0, 0, 2);
191 while (pending.nr) {
192 struct object_array_entry *entry;
193 struct object *obj;
195 entry = pending.objects + --pending.nr;
196 obj = entry->item;
197 result |= traverse_one_object(obj);
198 display_progress(progress, ++nr);
200 stop_progress(&progress);
201 return !!result;
204 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
206 if (!obj)
207 return 1;
208 obj->used = 1;
209 return 0;
213 * Check a single reachable object
215 static void check_reachable_object(struct object *obj)
218 * We obviously want the object to be parsed,
219 * except if it was in a pack-file and we didn't
220 * do a full fsck
222 if (!(obj->flags & HAS_OBJ)) {
223 if (has_sha1_pack(obj->oid.hash))
224 return; /* it is in pack - forget about it */
225 printf("missing %s %s\n", printable_type(obj),
226 describe_object(obj));
227 errors_found |= ERROR_REACHABLE;
228 return;
233 * Check a single unreachable object
235 static void check_unreachable_object(struct object *obj)
238 * Missing unreachable object? Ignore it. It's not like
239 * we miss it (since it can't be reached), nor do we want
240 * to complain about it being unreachable (since it does
241 * not exist).
243 if (!(obj->flags & HAS_OBJ))
244 return;
247 * Unreachable object that exists? Show it if asked to,
248 * since this is something that is prunable.
250 if (show_unreachable) {
251 printf("unreachable %s %s\n", printable_type(obj),
252 describe_object(obj));
253 return;
257 * "!used" means that nothing at all points to it, including
258 * other unreachable objects. In other words, it's the "tip"
259 * of some set of unreachable objects, usually a commit that
260 * got dropped.
262 * Such starting points are more interesting than some random
263 * set of unreachable objects, so we show them even if the user
264 * hasn't asked for _all_ unreachable objects. If you have
265 * deleted a branch by mistake, this is a prime candidate to
266 * start looking at, for example.
268 if (!obj->used) {
269 if (show_dangling)
270 printf("dangling %s %s\n", printable_type(obj),
271 describe_object(obj));
272 if (write_lost_and_found) {
273 char *filename = git_pathdup("lost-found/%s/%s",
274 obj->type == OBJ_COMMIT ? "commit" : "other",
275 describe_object(obj));
276 FILE *f;
278 if (safe_create_leading_directories_const(filename)) {
279 error("Could not create lost-found");
280 free(filename);
281 return;
283 if (!(f = fopen(filename, "w")))
284 die_errno("Could not open '%s'", filename);
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 if (obj->flags & SEEN)
339 return 0;
340 obj->flags |= SEEN;
342 if (verbose)
343 fprintf(stderr, "Checking %s %s\n",
344 printable_type(obj), describe_object(obj));
346 if (fsck_walk(obj, NULL, &fsck_obj_options))
347 objerror(obj, "broken links");
348 if (fsck_object(obj, NULL, 0, &fsck_obj_options))
349 return -1;
351 if (obj->type == OBJ_TREE) {
352 struct tree *item = (struct tree *) obj;
354 free_tree_buffer(item);
357 if (obj->type == OBJ_COMMIT) {
358 struct commit *commit = (struct commit *) obj;
360 free_commit_buffer(commit);
362 if (!commit->parents && show_root)
363 printf("root %s\n", describe_object(&commit->object));
366 if (obj->type == OBJ_TAG) {
367 struct tag *tag = (struct tag *) obj;
369 if (show_tags && tag->tagged) {
370 printf("tagged %s %s", printable_type(tag->tagged),
371 describe_object(tag->tagged));
372 printf(" (%s) in %s\n", tag->tag,
373 describe_object(&tag->object));
377 return 0;
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 = HAS_OBJ;
394 return fsck_obj(obj);
397 static int default_refs;
399 static void fsck_handle_reflog_oid(const char *refname, struct object_id *oid,
400 timestamp_t timestamp)
402 struct object *obj;
404 if (!is_null_oid(oid)) {
405 obj = lookup_object(oid->hash);
406 if (obj && (obj->flags & HAS_OBJ)) {
407 if (timestamp && name_objects)
408 add_decoration(fsck_walk_options.object_names,
409 obj,
410 xstrfmt("%s@{%"PRItime"}", refname, timestamp));
411 obj->used = 1;
412 mark_object_reachable(obj);
413 } else {
414 error("%s: invalid reflog entry %s", refname, oid_to_hex(oid));
415 errors_found |= ERROR_REACHABLE;
420 static int fsck_handle_reflog_ent(struct object_id *ooid, struct object_id *noid,
421 const char *email, timestamp_t timestamp, int tz,
422 const char *message, void *cb_data)
424 const char *refname = cb_data;
426 if (verbose)
427 fprintf(stderr, "Checking reflog %s->%s\n",
428 oid_to_hex(ooid), oid_to_hex(noid));
430 fsck_handle_reflog_oid(refname, ooid, 0);
431 fsck_handle_reflog_oid(refname, noid, timestamp);
432 return 0;
435 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
436 int flag, void *cb_data)
438 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
439 return 0;
442 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
443 int flag, void *cb_data)
445 struct object *obj;
447 obj = parse_object(oid);
448 if (!obj) {
449 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
450 errors_found |= ERROR_REACHABLE;
451 /* We'll continue with the rest despite the error.. */
452 return 0;
454 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
455 error("%s: not a commit", refname);
456 errors_found |= ERROR_REFS;
458 default_refs++;
459 obj->used = 1;
460 if (name_objects)
461 add_decoration(fsck_walk_options.object_names,
462 obj, xstrdup(refname));
463 mark_object_reachable(obj);
465 return 0;
468 static void get_default_heads(void)
470 if (head_points_at && !is_null_oid(&head_oid))
471 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
472 for_each_rawref(fsck_handle_ref, NULL);
473 if (include_reflogs)
474 for_each_reflog(fsck_handle_reflog, NULL);
477 * Not having any default heads isn't really fatal, but
478 * it does mean that "--unreachable" no longer makes any
479 * sense (since in this case everything will obviously
480 * be unreachable by definition.
482 * Showing dangling objects is valid, though (as those
483 * dangling objects are likely lost heads).
485 * So we just print a warning about it, and clear the
486 * "show_unreachable" flag.
488 if (!default_refs) {
489 fprintf(stderr, "notice: No default references\n");
490 show_unreachable = 0;
494 static struct object *parse_loose_object(const struct object_id *oid,
495 const char *path)
497 struct object *obj;
498 void *contents;
499 enum object_type type;
500 unsigned long size;
501 int eaten;
503 if (read_loose_object(path, oid->hash, &type, &size, &contents) < 0)
504 return NULL;
506 if (!contents && type != OBJ_BLOB)
507 die("BUG: read_loose_object streamed a non-blob");
509 obj = parse_object_buffer(oid, type, size, contents, &eaten);
511 if (!eaten)
512 free(contents);
513 return obj;
516 static int fsck_loose(const struct object_id *oid, const char *path, void *data)
518 struct object *obj = parse_loose_object(oid, path);
520 if (!obj) {
521 errors_found |= ERROR_OBJECT;
522 error("%s: object corrupt or missing: %s",
523 oid_to_hex(oid), path);
524 return 0; /* keep checking other objects */
527 obj->flags = HAS_OBJ;
528 if (fsck_obj(obj))
529 errors_found |= ERROR_OBJECT;
530 return 0;
533 static int fsck_cruft(const char *basename, const char *path, void *data)
535 if (!starts_with(basename, "tmp_obj_"))
536 fprintf(stderr, "bad sha1 file: %s\n", path);
537 return 0;
540 static int fsck_subdir(int nr, const char *path, void *progress)
542 display_progress(progress, nr + 1);
543 return 0;
546 static void fsck_object_dir(const char *path)
548 struct progress *progress = NULL;
550 if (verbose)
551 fprintf(stderr, "Checking object directory\n");
553 if (show_progress)
554 progress = start_progress(_("Checking object directories"), 256);
556 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
557 progress);
558 display_progress(progress, 256);
559 stop_progress(&progress);
562 static int fsck_head_link(void)
564 int null_is_error = 0;
566 if (verbose)
567 fprintf(stderr, "Checking HEAD link\n");
569 head_points_at = resolve_ref_unsafe("HEAD", 0, head_oid.hash, NULL);
570 if (!head_points_at) {
571 errors_found |= ERROR_REFS;
572 return error("Invalid HEAD");
574 if (!strcmp(head_points_at, "HEAD"))
575 /* detached HEAD */
576 null_is_error = 1;
577 else if (!starts_with(head_points_at, "refs/heads/")) {
578 errors_found |= ERROR_REFS;
579 return error("HEAD points to something strange (%s)",
580 head_points_at);
582 if (is_null_oid(&head_oid)) {
583 if (null_is_error) {
584 errors_found |= ERROR_REFS;
585 return error("HEAD: detached HEAD points at nothing");
587 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
588 head_points_at + 11);
590 return 0;
593 static int fsck_cache_tree(struct cache_tree *it)
595 int i;
596 int err = 0;
598 if (verbose)
599 fprintf(stderr, "Checking cache tree\n");
601 if (0 <= it->entry_count) {
602 struct object *obj = parse_object(&it->oid);
603 if (!obj) {
604 error("%s: invalid sha1 pointer in cache-tree",
605 oid_to_hex(&it->oid));
606 errors_found |= ERROR_REFS;
607 return 1;
609 obj->used = 1;
610 if (name_objects)
611 add_decoration(fsck_walk_options.object_names,
612 obj, xstrdup(":"));
613 mark_object_reachable(obj);
614 if (obj->type != OBJ_TREE)
615 err |= objerror(obj, "non-tree in cache-tree");
617 for (i = 0; i < it->subtree_nr; i++)
618 err |= fsck_cache_tree(it->down[i]->cache_tree);
619 return err;
622 static void mark_object_for_connectivity(const struct object_id *oid)
624 struct object *obj = lookup_unknown_object(oid->hash);
625 obj->flags |= HAS_OBJ;
628 static int mark_loose_for_connectivity(const struct object_id *oid,
629 const char *path,
630 void *data)
632 mark_object_for_connectivity(oid);
633 return 0;
636 static int mark_packed_for_connectivity(const struct object_id *oid,
637 struct packed_git *pack,
638 uint32_t pos,
639 void *data)
641 mark_object_for_connectivity(oid);
642 return 0;
645 static char const * const fsck_usage[] = {
646 N_("git fsck [<options>] [<object>...]"),
647 NULL
650 static struct option fsck_opts[] = {
651 OPT__VERBOSE(&verbose, N_("be verbose")),
652 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
653 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
654 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
655 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
656 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
657 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
658 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
659 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
660 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
661 OPT_BOOL(0, "lost-found", &write_lost_and_found,
662 N_("write dangling objects in .git/lost-found")),
663 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
664 OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
665 OPT_END(),
668 int cmd_fsck(int argc, const char **argv, const char *prefix)
670 int i, heads;
671 struct alternate_object_database *alt;
673 errors_found = 0;
674 check_replace_refs = 0;
676 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
678 fsck_walk_options.walk = mark_object;
679 fsck_obj_options.walk = mark_used;
680 fsck_obj_options.error_func = fsck_error_func;
681 if (check_strict)
682 fsck_obj_options.strict = 1;
684 if (show_progress == -1)
685 show_progress = isatty(2);
686 if (verbose)
687 show_progress = 0;
689 if (write_lost_and_found) {
690 check_full = 1;
691 include_reflogs = 0;
694 if (name_objects)
695 fsck_walk_options.object_names =
696 xcalloc(1, sizeof(struct decoration));
698 git_config(fsck_config, NULL);
700 fsck_head_link();
701 if (connectivity_only) {
702 for_each_loose_object(mark_loose_for_connectivity, NULL, 0);
703 for_each_packed_object(mark_packed_for_connectivity, NULL, 0);
704 } else {
705 fsck_object_dir(get_object_directory());
707 prepare_alt_odb();
708 for (alt = alt_odb_list; alt; alt = alt->next)
709 fsck_object_dir(alt->path);
711 if (check_full) {
712 struct packed_git *p;
713 uint32_t total = 0, count = 0;
714 struct progress *progress = NULL;
716 prepare_packed_git();
718 if (show_progress) {
719 for (p = packed_git; p; p = p->next) {
720 if (open_pack_index(p))
721 continue;
722 total += p->num_objects;
725 progress = start_progress(_("Checking objects"), total);
727 for (p = packed_git; p; p = p->next) {
728 /* verify gives error messages itself */
729 if (verify_pack(p, fsck_obj_buffer,
730 progress, count))
731 errors_found |= ERROR_PACK;
732 count += p->num_objects;
734 stop_progress(&progress);
738 heads = 0;
739 for (i = 0; i < argc; i++) {
740 const char *arg = argv[i];
741 unsigned char sha1[20];
742 if (!get_sha1(arg, sha1)) {
743 struct object *obj = lookup_object(sha1);
745 if (!obj || !(obj->flags & HAS_OBJ)) {
746 error("%s: object missing", sha1_to_hex(sha1));
747 errors_found |= ERROR_OBJECT;
748 continue;
751 obj->used = 1;
752 if (name_objects)
753 add_decoration(fsck_walk_options.object_names,
754 obj, xstrdup(arg));
755 mark_object_reachable(obj);
756 heads++;
757 continue;
759 error("invalid parameter: expected sha1, got '%s'", arg);
760 errors_found |= ERROR_OBJECT;
764 * If we've not been given any explicit head information, do the
765 * default ones from .git/refs. We also consider the index file
766 * in this case (ie this implies --cache).
768 if (!argc) {
769 get_default_heads();
770 keep_cache_objects = 1;
773 if (keep_cache_objects) {
774 verify_index_checksum = 1;
775 read_cache();
776 for (i = 0; i < active_nr; i++) {
777 unsigned int mode;
778 struct blob *blob;
779 struct object *obj;
781 mode = active_cache[i]->ce_mode;
782 if (S_ISGITLINK(mode))
783 continue;
784 blob = lookup_blob(&active_cache[i]->oid);
785 if (!blob)
786 continue;
787 obj = &blob->object;
788 obj->used = 1;
789 if (name_objects)
790 add_decoration(fsck_walk_options.object_names,
791 obj,
792 xstrfmt(":%s", active_cache[i]->name));
793 mark_object_reachable(obj);
795 if (active_cache_tree)
796 fsck_cache_tree(active_cache_tree);
799 check_connectivity();
800 return errors_found;