Merge branch 'ta/docfix-index-format-tech'
[git/mjg.git] / builtin / fsck.c
blobf4b87e9b33710e5ae67ec8886aad5259fdb0177f
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"
17 #define REACHABLE 0x0001
18 #define SEEN 0x0002
19 #define HAS_OBJ 0x0004
21 static int show_root;
22 static int show_tags;
23 static int show_unreachable;
24 static int include_reflogs = 1;
25 static int check_full = 1;
26 static int connectivity_only;
27 static int check_strict;
28 static int keep_cache_objects;
29 static struct fsck_options fsck_walk_options = FSCK_OPTIONS_DEFAULT;
30 static struct fsck_options fsck_obj_options = FSCK_OPTIONS_DEFAULT;
31 static struct object_id head_oid;
32 static const char *head_points_at;
33 static int errors_found;
34 static int write_lost_and_found;
35 static int verbose;
36 static int show_progress = -1;
37 static int show_dangling = 1;
38 #define ERROR_OBJECT 01
39 #define ERROR_REACHABLE 02
40 #define ERROR_PACK 04
42 #ifdef NO_D_INO_IN_DIRENT
43 #define SORT_DIRENT 0
44 #define DIRENT_SORT_HINT(de) 0
45 #else
46 #define SORT_DIRENT 1
47 #define DIRENT_SORT_HINT(de) ((de)->d_ino)
48 #endif
50 static int fsck_config(const char *var, const char *value, void *cb)
52 if (strcmp(var, "fsck.skiplist") == 0) {
53 const char *path;
54 struct strbuf sb = STRBUF_INIT;
56 if (git_config_pathname(&path, var, value))
57 return 1;
58 strbuf_addf(&sb, "skiplist=%s", path);
59 free((char *)path);
60 fsck_set_msg_types(&fsck_obj_options, sb.buf);
61 strbuf_release(&sb);
62 return 0;
65 if (skip_prefix(var, "fsck.", &var)) {
66 fsck_set_msg_type(&fsck_obj_options, var, value);
67 return 0;
70 return git_default_config(var, value, cb);
73 static void objreport(struct object *obj, const char *msg_type,
74 const char *err)
76 fprintf(stderr, "%s in %s %s: %s\n",
77 msg_type, typename(obj->type), sha1_to_hex(obj->sha1), err);
80 static int objerror(struct object *obj, const char *err)
82 errors_found |= ERROR_OBJECT;
83 objreport(obj, "error", err);
84 return -1;
87 static int fsck_error_func(struct object *obj, int type, const char *message)
89 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
90 return (type == FSCK_WARN) ? 0 : 1;
93 static struct object_array pending;
95 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
97 struct object *parent = data;
100 * The only case data is NULL or type is OBJ_ANY is when
101 * mark_object_reachable() calls us. All the callers of
102 * that function has non-NULL obj hence ...
104 if (!obj) {
105 /* ... these references to parent->fld are safe here */
106 printf("broken link from %7s %s\n",
107 typename(parent->type), sha1_to_hex(parent->sha1));
108 printf("broken link from %7s %s\n",
109 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
110 errors_found |= ERROR_REACHABLE;
111 return 1;
114 if (type != OBJ_ANY && obj->type != type)
115 /* ... and the reference to parent is safe here */
116 objerror(parent, "wrong object type in link");
118 if (obj->flags & REACHABLE)
119 return 0;
120 obj->flags |= REACHABLE;
121 if (!(obj->flags & HAS_OBJ)) {
122 if (parent && !has_sha1_file(obj->sha1)) {
123 printf("broken link from %7s %s\n",
124 typename(parent->type), sha1_to_hex(parent->sha1));
125 printf(" to %7s %s\n",
126 typename(obj->type), sha1_to_hex(obj->sha1));
127 errors_found |= ERROR_REACHABLE;
129 return 1;
132 add_object_array(obj, NULL, &pending);
133 return 0;
136 static void mark_object_reachable(struct object *obj)
138 mark_object(obj, OBJ_ANY, NULL, NULL);
141 static int traverse_one_object(struct object *obj)
143 int result;
144 struct tree *tree = NULL;
146 if (obj->type == OBJ_TREE) {
147 tree = (struct tree *)obj;
148 if (parse_tree(tree) < 0)
149 return 1; /* error already displayed */
151 result = fsck_walk(obj, obj, &fsck_walk_options);
152 if (tree)
153 free_tree_buffer(tree);
154 return result;
157 static int traverse_reachable(void)
159 struct progress *progress = NULL;
160 unsigned int nr = 0;
161 int result = 0;
162 if (show_progress)
163 progress = start_progress_delay(_("Checking connectivity"), 0, 0, 2);
164 while (pending.nr) {
165 struct object_array_entry *entry;
166 struct object *obj;
168 entry = pending.objects + --pending.nr;
169 obj = entry->item;
170 result |= traverse_one_object(obj);
171 display_progress(progress, ++nr);
173 stop_progress(&progress);
174 return !!result;
177 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
179 if (!obj)
180 return 1;
181 obj->used = 1;
182 return 0;
186 * Check a single reachable object
188 static void check_reachable_object(struct object *obj)
191 * We obviously want the object to be parsed,
192 * except if it was in a pack-file and we didn't
193 * do a full fsck
195 if (!(obj->flags & HAS_OBJ)) {
196 if (has_sha1_pack(obj->sha1))
197 return; /* it is in pack - forget about it */
198 if (connectivity_only && has_sha1_file(obj->sha1))
199 return;
200 printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
201 errors_found |= ERROR_REACHABLE;
202 return;
207 * Check a single unreachable object
209 static void check_unreachable_object(struct object *obj)
212 * Missing unreachable object? Ignore it. It's not like
213 * we miss it (since it can't be reached), nor do we want
214 * to complain about it being unreachable (since it does
215 * not exist).
217 if (!obj->parsed)
218 return;
221 * Unreachable object that exists? Show it if asked to,
222 * since this is something that is prunable.
224 if (show_unreachable) {
225 printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
226 return;
230 * "!used" means that nothing at all points to it, including
231 * other unreachable objects. In other words, it's the "tip"
232 * of some set of unreachable objects, usually a commit that
233 * got dropped.
235 * Such starting points are more interesting than some random
236 * set of unreachable objects, so we show them even if the user
237 * hasn't asked for _all_ unreachable objects. If you have
238 * deleted a branch by mistake, this is a prime candidate to
239 * start looking at, for example.
241 if (!obj->used) {
242 if (show_dangling)
243 printf("dangling %s %s\n", typename(obj->type),
244 sha1_to_hex(obj->sha1));
245 if (write_lost_and_found) {
246 const char *filename = git_path("lost-found/%s/%s",
247 obj->type == OBJ_COMMIT ? "commit" : "other",
248 sha1_to_hex(obj->sha1));
249 FILE *f;
251 if (safe_create_leading_directories_const(filename)) {
252 error("Could not create lost-found");
253 return;
255 if (!(f = fopen(filename, "w")))
256 die_errno("Could not open '%s'", filename);
257 if (obj->type == OBJ_BLOB) {
258 if (stream_blob_to_fd(fileno(f), obj->sha1, NULL, 1))
259 die_errno("Could not write '%s'", filename);
260 } else
261 fprintf(f, "%s\n", sha1_to_hex(obj->sha1));
262 if (fclose(f))
263 die_errno("Could not finish '%s'",
264 filename);
266 return;
270 * Otherwise? It's there, it's unreachable, and some other unreachable
271 * object points to it. Ignore it - it's not interesting, and we showed
272 * all the interesting cases above.
276 static void check_object(struct object *obj)
278 if (verbose)
279 fprintf(stderr, "Checking %s\n", sha1_to_hex(obj->sha1));
281 if (obj->flags & REACHABLE)
282 check_reachable_object(obj);
283 else
284 check_unreachable_object(obj);
287 static void check_connectivity(void)
289 int i, max;
291 /* Traverse the pending reachable objects */
292 traverse_reachable();
294 /* Look up all the requirements, warn about missing objects.. */
295 max = get_max_object_index();
296 if (verbose)
297 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
299 for (i = 0; i < max; i++) {
300 struct object *obj = get_indexed_object(i);
302 if (obj)
303 check_object(obj);
307 static int fsck_obj(struct object *obj)
309 if (obj->flags & SEEN)
310 return 0;
311 obj->flags |= SEEN;
313 if (verbose)
314 fprintf(stderr, "Checking %s %s\n",
315 typename(obj->type), sha1_to_hex(obj->sha1));
317 if (fsck_walk(obj, NULL, &fsck_obj_options))
318 objerror(obj, "broken links");
319 if (fsck_object(obj, NULL, 0, &fsck_obj_options))
320 return -1;
322 if (obj->type == OBJ_TREE) {
323 struct tree *item = (struct tree *) obj;
325 free_tree_buffer(item);
328 if (obj->type == OBJ_COMMIT) {
329 struct commit *commit = (struct commit *) obj;
331 free_commit_buffer(commit);
333 if (!commit->parents && show_root)
334 printf("root %s\n", sha1_to_hex(commit->object.sha1));
337 if (obj->type == OBJ_TAG) {
338 struct tag *tag = (struct tag *) obj;
340 if (show_tags && tag->tagged) {
341 printf("tagged %s %s", typename(tag->tagged->type), sha1_to_hex(tag->tagged->sha1));
342 printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
346 return 0;
349 static int fsck_sha1(const unsigned char *sha1)
351 struct object *obj = parse_object(sha1);
352 if (!obj) {
353 errors_found |= ERROR_OBJECT;
354 return error("%s: object corrupt or missing",
355 sha1_to_hex(sha1));
357 obj->flags |= HAS_OBJ;
358 return fsck_obj(obj);
361 static int fsck_obj_buffer(const unsigned char *sha1, enum object_type type,
362 unsigned long size, void *buffer, int *eaten)
364 struct object *obj;
365 obj = parse_object_buffer(sha1, type, size, buffer, eaten);
366 if (!obj) {
367 errors_found |= ERROR_OBJECT;
368 return error("%s: object corrupt or missing", sha1_to_hex(sha1));
370 obj->flags = HAS_OBJ;
371 return fsck_obj(obj);
375 * This is the sorting chunk size: make it reasonably
376 * big so that we can sort well..
378 #define MAX_SHA1_ENTRIES (1024)
380 struct sha1_entry {
381 unsigned long ino;
382 unsigned char sha1[20];
385 static struct {
386 unsigned long nr;
387 struct sha1_entry *entry[MAX_SHA1_ENTRIES];
388 } sha1_list;
390 static int ino_compare(const void *_a, const void *_b)
392 const struct sha1_entry *a = _a, *b = _b;
393 unsigned long ino1 = a->ino, ino2 = b->ino;
394 return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
397 static void fsck_sha1_list(void)
399 int i, nr = sha1_list.nr;
401 if (SORT_DIRENT)
402 qsort(sha1_list.entry, nr,
403 sizeof(struct sha1_entry *), ino_compare);
404 for (i = 0; i < nr; i++) {
405 struct sha1_entry *entry = sha1_list.entry[i];
406 unsigned char *sha1 = entry->sha1;
408 sha1_list.entry[i] = NULL;
409 if (fsck_sha1(sha1))
410 errors_found |= ERROR_OBJECT;
411 free(entry);
413 sha1_list.nr = 0;
416 static void add_sha1_list(unsigned char *sha1, unsigned long ino)
418 struct sha1_entry *entry = xmalloc(sizeof(*entry));
419 int nr;
421 entry->ino = ino;
422 hashcpy(entry->sha1, sha1);
423 nr = sha1_list.nr;
424 if (nr == MAX_SHA1_ENTRIES) {
425 fsck_sha1_list();
426 nr = 0;
428 sha1_list.entry[nr] = entry;
429 sha1_list.nr = ++nr;
432 static inline int is_loose_object_file(struct dirent *de,
433 char *name, unsigned char *sha1)
435 if (strlen(de->d_name) != 38)
436 return 0;
437 memcpy(name + 2, de->d_name, 39);
438 return !get_sha1_hex(name, sha1);
441 static void fsck_dir(int i, char *path)
443 DIR *dir = opendir(path);
444 struct dirent *de;
445 char name[100];
447 if (!dir)
448 return;
450 if (verbose)
451 fprintf(stderr, "Checking directory %s\n", path);
453 sprintf(name, "%02x", i);
454 while ((de = readdir(dir)) != NULL) {
455 unsigned char sha1[20];
457 if (is_dot_or_dotdot(de->d_name))
458 continue;
459 if (is_loose_object_file(de, name, sha1)) {
460 add_sha1_list(sha1, DIRENT_SORT_HINT(de));
461 continue;
463 if (starts_with(de->d_name, "tmp_obj_"))
464 continue;
465 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
467 closedir(dir);
470 static int default_refs;
472 static void fsck_handle_reflog_sha1(const char *refname, unsigned char *sha1)
474 struct object *obj;
476 if (!is_null_sha1(sha1)) {
477 obj = lookup_object(sha1);
478 if (obj) {
479 obj->used = 1;
480 mark_object_reachable(obj);
481 } else {
482 error("%s: invalid reflog entry %s", refname, sha1_to_hex(sha1));
483 errors_found |= ERROR_REACHABLE;
488 static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
489 const char *email, unsigned long timestamp, int tz,
490 const char *message, void *cb_data)
492 const char *refname = cb_data;
494 if (verbose)
495 fprintf(stderr, "Checking reflog %s->%s\n",
496 sha1_to_hex(osha1), sha1_to_hex(nsha1));
498 fsck_handle_reflog_sha1(refname, osha1);
499 fsck_handle_reflog_sha1(refname, nsha1);
500 return 0;
503 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
504 int flag, void *cb_data)
506 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
507 return 0;
510 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
511 int flag, void *cb_data)
513 struct object *obj;
515 obj = parse_object(oid->hash);
516 if (!obj) {
517 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
518 errors_found |= ERROR_REACHABLE;
519 /* We'll continue with the rest despite the error.. */
520 return 0;
522 if (obj->type != OBJ_COMMIT && is_branch(refname))
523 error("%s: not a commit", refname);
524 default_refs++;
525 obj->used = 1;
526 mark_object_reachable(obj);
528 return 0;
531 static void get_default_heads(void)
533 if (head_points_at && !is_null_oid(&head_oid))
534 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
535 for_each_rawref(fsck_handle_ref, NULL);
536 if (include_reflogs)
537 for_each_reflog(fsck_handle_reflog, NULL);
540 * Not having any default heads isn't really fatal, but
541 * it does mean that "--unreachable" no longer makes any
542 * sense (since in this case everything will obviously
543 * be unreachable by definition.
545 * Showing dangling objects is valid, though (as those
546 * dangling objects are likely lost heads).
548 * So we just print a warning about it, and clear the
549 * "show_unreachable" flag.
551 if (!default_refs) {
552 fprintf(stderr, "notice: No default references\n");
553 show_unreachable = 0;
557 static void fsck_object_dir(const char *path)
559 int i;
560 struct progress *progress = NULL;
562 if (verbose)
563 fprintf(stderr, "Checking object directory\n");
565 if (show_progress)
566 progress = start_progress(_("Checking object directories"), 256);
567 for (i = 0; i < 256; i++) {
568 static char dir[4096];
569 sprintf(dir, "%s/%02x", path, i);
570 fsck_dir(i, dir);
571 display_progress(progress, i+1);
573 stop_progress(&progress);
574 fsck_sha1_list();
577 static int fsck_head_link(void)
579 int flag;
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.hash, &flag);
586 if (!head_points_at)
587 return error("Invalid HEAD");
588 if (!strcmp(head_points_at, "HEAD"))
589 /* detached HEAD */
590 null_is_error = 1;
591 else if (!starts_with(head_points_at, "refs/heads/"))
592 return error("HEAD points to something strange (%s)",
593 head_points_at);
594 if (is_null_oid(&head_oid)) {
595 if (null_is_error)
596 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->sha1);
613 if (!obj) {
614 error("%s: invalid sha1 pointer in cache-tree",
615 sha1_to_hex(it->sha1));
616 return 1;
618 obj->used = 1;
619 mark_object_reachable(obj);
620 if (obj->type != OBJ_TREE)
621 err |= objerror(obj, "non-tree in cache-tree");
623 for (i = 0; i < it->subtree_nr; i++)
624 err |= fsck_cache_tree(it->down[i]->cache_tree);
625 return err;
628 static char const * const fsck_usage[] = {
629 N_("git fsck [<options>] [<object>...]"),
630 NULL
633 static struct option fsck_opts[] = {
634 OPT__VERBOSE(&verbose, N_("be verbose")),
635 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
636 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
637 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
638 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
639 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
640 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
641 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
642 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
643 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
644 OPT_BOOL(0, "lost-found", &write_lost_and_found,
645 N_("write dangling objects in .git/lost-found")),
646 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
647 OPT_END(),
650 int cmd_fsck(int argc, const char **argv, const char *prefix)
652 int i, heads;
653 struct alternate_object_database *alt;
655 errors_found = 0;
656 check_replace_refs = 0;
658 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
660 fsck_walk_options.walk = mark_object;
661 fsck_obj_options.walk = mark_used;
662 fsck_obj_options.error_func = fsck_error_func;
663 if (check_strict)
664 fsck_obj_options.strict = 1;
666 if (show_progress == -1)
667 show_progress = isatty(2);
668 if (verbose)
669 show_progress = 0;
671 if (write_lost_and_found) {
672 check_full = 1;
673 include_reflogs = 0;
676 git_config(fsck_config, NULL);
678 fsck_head_link();
679 if (!connectivity_only)
680 fsck_object_dir(get_object_directory());
682 prepare_alt_odb();
683 for (alt = alt_odb_list; alt; alt = alt->next) {
684 char namebuf[PATH_MAX];
685 int namelen = alt->name - alt->base;
686 memcpy(namebuf, alt->base, namelen);
687 namebuf[namelen - 1] = 0;
688 fsck_object_dir(namebuf);
691 if (check_full) {
692 struct packed_git *p;
693 uint32_t total = 0, count = 0;
694 struct progress *progress = NULL;
696 prepare_packed_git();
698 if (show_progress) {
699 for (p = packed_git; p; p = p->next) {
700 if (open_pack_index(p))
701 continue;
702 total += p->num_objects;
705 progress = start_progress(_("Checking objects"), total);
707 for (p = packed_git; p; p = p->next) {
708 /* verify gives error messages itself */
709 if (verify_pack(p, fsck_obj_buffer,
710 progress, count))
711 errors_found |= ERROR_PACK;
712 count += p->num_objects;
714 stop_progress(&progress);
717 heads = 0;
718 for (i = 0; i < argc; i++) {
719 const char *arg = argv[i];
720 unsigned char sha1[20];
721 if (!get_sha1(arg, sha1)) {
722 struct object *obj = lookup_object(sha1);
724 /* Error is printed by lookup_object(). */
725 if (!obj)
726 continue;
728 obj->used = 1;
729 mark_object_reachable(obj);
730 heads++;
731 continue;
733 error("invalid parameter: expected sha1, got '%s'", arg);
737 * If we've not been given any explicit head information, do the
738 * default ones from .git/refs. We also consider the index file
739 * in this case (ie this implies --cache).
741 if (!heads) {
742 get_default_heads();
743 keep_cache_objects = 1;
746 if (keep_cache_objects) {
747 read_cache();
748 for (i = 0; i < active_nr; i++) {
749 unsigned int mode;
750 struct blob *blob;
751 struct object *obj;
753 mode = active_cache[i]->ce_mode;
754 if (S_ISGITLINK(mode))
755 continue;
756 blob = lookup_blob(active_cache[i]->sha1);
757 if (!blob)
758 continue;
759 obj = &blob->object;
760 obj->used = 1;
761 mark_object_reachable(obj);
763 if (active_cache_tree)
764 fsck_cache_tree(active_cache_tree);
767 check_connectivity();
768 return errors_found;