Merge branch 'sa/send-email-smtp-batch-data-limit' into maint
[git.git] / builtin / fsck.c
blobb9a74f0cf6e169073994e466110fd1bde9febba7
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
41 #define ERROR_REFS 010
43 #ifdef NO_D_INO_IN_DIRENT
44 #define SORT_DIRENT 0
45 #define DIRENT_SORT_HINT(de) 0
46 #else
47 #define SORT_DIRENT 1
48 #define DIRENT_SORT_HINT(de) ((de)->d_ino)
49 #endif
51 static int fsck_config(const char *var, const char *value, void *cb)
53 if (strcmp(var, "fsck.skiplist") == 0) {
54 const char *path;
55 struct strbuf sb = STRBUF_INIT;
57 if (git_config_pathname(&path, var, value))
58 return 1;
59 strbuf_addf(&sb, "skiplist=%s", path);
60 free((char *)path);
61 fsck_set_msg_types(&fsck_obj_options, sb.buf);
62 strbuf_release(&sb);
63 return 0;
66 if (skip_prefix(var, "fsck.", &var)) {
67 fsck_set_msg_type(&fsck_obj_options, var, value);
68 return 0;
71 return git_default_config(var, value, cb);
74 static void objreport(struct object *obj, const char *msg_type,
75 const char *err)
77 fprintf(stderr, "%s in %s %s: %s\n",
78 msg_type, typename(obj->type), sha1_to_hex(obj->sha1), err);
81 static int objerror(struct object *obj, const char *err)
83 errors_found |= ERROR_OBJECT;
84 objreport(obj, "error", err);
85 return -1;
88 static int fsck_error_func(struct object *obj, int type, const char *message)
90 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
91 return (type == FSCK_WARN) ? 0 : 1;
94 static struct object_array pending;
96 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
98 struct object *parent = data;
101 * The only case data is NULL or type is OBJ_ANY is when
102 * mark_object_reachable() calls us. All the callers of
103 * that function has non-NULL obj hence ...
105 if (!obj) {
106 /* ... these references to parent->fld are safe here */
107 printf("broken link from %7s %s\n",
108 typename(parent->type), sha1_to_hex(parent->sha1));
109 printf("broken link from %7s %s\n",
110 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
111 errors_found |= ERROR_REACHABLE;
112 return 1;
115 if (type != OBJ_ANY && obj->type != type)
116 /* ... and the reference to parent is safe here */
117 objerror(parent, "wrong object type in link");
119 if (obj->flags & REACHABLE)
120 return 0;
121 obj->flags |= REACHABLE;
122 if (!(obj->flags & HAS_OBJ)) {
123 if (parent && !has_sha1_file(obj->sha1)) {
124 printf("broken link from %7s %s\n",
125 typename(parent->type), sha1_to_hex(parent->sha1));
126 printf(" to %7s %s\n",
127 typename(obj->type), sha1_to_hex(obj->sha1));
128 errors_found |= ERROR_REACHABLE;
130 return 1;
133 add_object_array(obj, NULL, &pending);
134 return 0;
137 static void mark_object_reachable(struct object *obj)
139 mark_object(obj, OBJ_ANY, NULL, NULL);
142 static int traverse_one_object(struct object *obj)
144 int result;
145 struct tree *tree = NULL;
147 if (obj->type == OBJ_TREE) {
148 tree = (struct tree *)obj;
149 if (parse_tree(tree) < 0)
150 return 1; /* error already displayed */
152 result = fsck_walk(obj, obj, &fsck_walk_options);
153 if (tree)
154 free_tree_buffer(tree);
155 return result;
158 static int traverse_reachable(void)
160 struct progress *progress = NULL;
161 unsigned int nr = 0;
162 int result = 0;
163 if (show_progress)
164 progress = start_progress_delay(_("Checking connectivity"), 0, 0, 2);
165 while (pending.nr) {
166 struct object_array_entry *entry;
167 struct object *obj;
169 entry = pending.objects + --pending.nr;
170 obj = entry->item;
171 result |= traverse_one_object(obj);
172 display_progress(progress, ++nr);
174 stop_progress(&progress);
175 return !!result;
178 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
180 if (!obj)
181 return 1;
182 obj->used = 1;
183 return 0;
187 * Check a single reachable object
189 static void check_reachable_object(struct object *obj)
192 * We obviously want the object to be parsed,
193 * except if it was in a pack-file and we didn't
194 * do a full fsck
196 if (!(obj->flags & HAS_OBJ)) {
197 if (has_sha1_pack(obj->sha1))
198 return; /* it is in pack - forget about it */
199 if (connectivity_only && has_sha1_file(obj->sha1))
200 return;
201 printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
202 errors_found |= ERROR_REACHABLE;
203 return;
208 * Check a single unreachable object
210 static void check_unreachable_object(struct object *obj)
213 * Missing unreachable object? Ignore it. It's not like
214 * we miss it (since it can't be reached), nor do we want
215 * to complain about it being unreachable (since it does
216 * not exist).
218 if (!obj->parsed)
219 return;
222 * Unreachable object that exists? Show it if asked to,
223 * since this is something that is prunable.
225 if (show_unreachable) {
226 printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
227 return;
231 * "!used" means that nothing at all points to it, including
232 * other unreachable objects. In other words, it's the "tip"
233 * of some set of unreachable objects, usually a commit that
234 * got dropped.
236 * Such starting points are more interesting than some random
237 * set of unreachable objects, so we show them even if the user
238 * hasn't asked for _all_ unreachable objects. If you have
239 * deleted a branch by mistake, this is a prime candidate to
240 * start looking at, for example.
242 if (!obj->used) {
243 if (show_dangling)
244 printf("dangling %s %s\n", typename(obj->type),
245 sha1_to_hex(obj->sha1));
246 if (write_lost_and_found) {
247 char *filename = git_pathdup("lost-found/%s/%s",
248 obj->type == OBJ_COMMIT ? "commit" : "other",
249 sha1_to_hex(obj->sha1));
250 FILE *f;
252 if (safe_create_leading_directories_const(filename)) {
253 error("Could not create lost-found");
254 free(filename);
255 return;
257 if (!(f = fopen(filename, "w")))
258 die_errno("Could not open '%s'", filename);
259 if (obj->type == OBJ_BLOB) {
260 if (stream_blob_to_fd(fileno(f), obj->sha1, NULL, 1))
261 die_errno("Could not write '%s'", filename);
262 } else
263 fprintf(f, "%s\n", sha1_to_hex(obj->sha1));
264 if (fclose(f))
265 die_errno("Could not finish '%s'",
266 filename);
267 free(filename);
269 return;
273 * Otherwise? It's there, it's unreachable, and some other unreachable
274 * object points to it. Ignore it - it's not interesting, and we showed
275 * all the interesting cases above.
279 static void check_object(struct object *obj)
281 if (verbose)
282 fprintf(stderr, "Checking %s\n", sha1_to_hex(obj->sha1));
284 if (obj->flags & REACHABLE)
285 check_reachable_object(obj);
286 else
287 check_unreachable_object(obj);
290 static void check_connectivity(void)
292 int i, max;
294 /* Traverse the pending reachable objects */
295 traverse_reachable();
297 /* Look up all the requirements, warn about missing objects.. */
298 max = get_max_object_index();
299 if (verbose)
300 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
302 for (i = 0; i < max; i++) {
303 struct object *obj = get_indexed_object(i);
305 if (obj)
306 check_object(obj);
310 static int fsck_obj(struct object *obj)
312 if (obj->flags & SEEN)
313 return 0;
314 obj->flags |= SEEN;
316 if (verbose)
317 fprintf(stderr, "Checking %s %s\n",
318 typename(obj->type), sha1_to_hex(obj->sha1));
320 if (fsck_walk(obj, NULL, &fsck_obj_options))
321 objerror(obj, "broken links");
322 if (fsck_object(obj, NULL, 0, &fsck_obj_options))
323 return -1;
325 if (obj->type == OBJ_TREE) {
326 struct tree *item = (struct tree *) obj;
328 free_tree_buffer(item);
331 if (obj->type == OBJ_COMMIT) {
332 struct commit *commit = (struct commit *) obj;
334 free_commit_buffer(commit);
336 if (!commit->parents && show_root)
337 printf("root %s\n", sha1_to_hex(commit->object.sha1));
340 if (obj->type == OBJ_TAG) {
341 struct tag *tag = (struct tag *) obj;
343 if (show_tags && tag->tagged) {
344 printf("tagged %s %s", typename(tag->tagged->type), sha1_to_hex(tag->tagged->sha1));
345 printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
349 return 0;
352 static int fsck_sha1(const unsigned char *sha1)
354 struct object *obj = parse_object(sha1);
355 if (!obj) {
356 errors_found |= ERROR_OBJECT;
357 return error("%s: object corrupt or missing",
358 sha1_to_hex(sha1));
360 obj->flags |= HAS_OBJ;
361 return fsck_obj(obj);
364 static int fsck_obj_buffer(const unsigned char *sha1, enum object_type type,
365 unsigned long size, void *buffer, int *eaten)
367 struct object *obj;
368 obj = parse_object_buffer(sha1, type, size, buffer, eaten);
369 if (!obj) {
370 errors_found |= ERROR_OBJECT;
371 return error("%s: object corrupt or missing", sha1_to_hex(sha1));
373 obj->flags = HAS_OBJ;
374 return fsck_obj(obj);
378 * This is the sorting chunk size: make it reasonably
379 * big so that we can sort well..
381 #define MAX_SHA1_ENTRIES (1024)
383 struct sha1_entry {
384 unsigned long ino;
385 unsigned char sha1[20];
388 static struct {
389 unsigned long nr;
390 struct sha1_entry *entry[MAX_SHA1_ENTRIES];
391 } sha1_list;
393 static int ino_compare(const void *_a, const void *_b)
395 const struct sha1_entry *a = _a, *b = _b;
396 unsigned long ino1 = a->ino, ino2 = b->ino;
397 return ino1 < ino2 ? -1 : ino1 > ino2 ? 1 : 0;
400 static void fsck_sha1_list(void)
402 int i, nr = sha1_list.nr;
404 if (SORT_DIRENT)
405 qsort(sha1_list.entry, nr,
406 sizeof(struct sha1_entry *), ino_compare);
407 for (i = 0; i < nr; i++) {
408 struct sha1_entry *entry = sha1_list.entry[i];
409 unsigned char *sha1 = entry->sha1;
411 sha1_list.entry[i] = NULL;
412 if (fsck_sha1(sha1))
413 errors_found |= ERROR_OBJECT;
414 free(entry);
416 sha1_list.nr = 0;
419 static void add_sha1_list(unsigned char *sha1, unsigned long ino)
421 struct sha1_entry *entry = xmalloc(sizeof(*entry));
422 int nr;
424 entry->ino = ino;
425 hashcpy(entry->sha1, sha1);
426 nr = sha1_list.nr;
427 if (nr == MAX_SHA1_ENTRIES) {
428 fsck_sha1_list();
429 nr = 0;
431 sha1_list.entry[nr] = entry;
432 sha1_list.nr = ++nr;
435 static inline int is_loose_object_file(struct dirent *de,
436 char *name, unsigned char *sha1)
438 if (strlen(de->d_name) != 38)
439 return 0;
440 memcpy(name + 2, de->d_name, 39);
441 return !get_sha1_hex(name, sha1);
444 static void fsck_dir(int i, char *path)
446 DIR *dir = opendir(path);
447 struct dirent *de;
448 char name[100];
450 if (!dir)
451 return;
453 if (verbose)
454 fprintf(stderr, "Checking directory %s\n", path);
456 sprintf(name, "%02x", i);
457 while ((de = readdir(dir)) != NULL) {
458 unsigned char sha1[20];
460 if (is_dot_or_dotdot(de->d_name))
461 continue;
462 if (is_loose_object_file(de, name, sha1)) {
463 add_sha1_list(sha1, DIRENT_SORT_HINT(de));
464 continue;
466 if (starts_with(de->d_name, "tmp_obj_"))
467 continue;
468 fprintf(stderr, "bad sha1 file: %s/%s\n", path, de->d_name);
470 closedir(dir);
473 static int default_refs;
475 static void fsck_handle_reflog_sha1(const char *refname, unsigned char *sha1)
477 struct object *obj;
479 if (!is_null_sha1(sha1)) {
480 obj = lookup_object(sha1);
481 if (obj) {
482 obj->used = 1;
483 mark_object_reachable(obj);
484 } else {
485 error("%s: invalid reflog entry %s", refname, sha1_to_hex(sha1));
486 errors_found |= ERROR_REACHABLE;
491 static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
492 const char *email, unsigned long timestamp, int tz,
493 const char *message, void *cb_data)
495 const char *refname = cb_data;
497 if (verbose)
498 fprintf(stderr, "Checking reflog %s->%s\n",
499 sha1_to_hex(osha1), sha1_to_hex(nsha1));
501 fsck_handle_reflog_sha1(refname, osha1);
502 fsck_handle_reflog_sha1(refname, nsha1);
503 return 0;
506 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
507 int flag, void *cb_data)
509 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
510 return 0;
513 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
514 int flag, void *cb_data)
516 struct object *obj;
518 obj = parse_object(oid->hash);
519 if (!obj) {
520 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
521 errors_found |= ERROR_REACHABLE;
522 /* We'll continue with the rest despite the error.. */
523 return 0;
525 if (obj->type != OBJ_COMMIT && is_branch(refname)) {
526 error("%s: not a commit", refname);
527 errors_found |= ERROR_REFS;
529 default_refs++;
530 obj->used = 1;
531 mark_object_reachable(obj);
533 return 0;
536 static void get_default_heads(void)
538 if (head_points_at && !is_null_oid(&head_oid))
539 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
540 for_each_rawref(fsck_handle_ref, NULL);
541 if (include_reflogs)
542 for_each_reflog(fsck_handle_reflog, NULL);
545 * Not having any default heads isn't really fatal, but
546 * it does mean that "--unreachable" no longer makes any
547 * sense (since in this case everything will obviously
548 * be unreachable by definition.
550 * Showing dangling objects is valid, though (as those
551 * dangling objects are likely lost heads).
553 * So we just print a warning about it, and clear the
554 * "show_unreachable" flag.
556 if (!default_refs) {
557 fprintf(stderr, "notice: No default references\n");
558 show_unreachable = 0;
562 static void fsck_object_dir(const char *path)
564 int i;
565 struct progress *progress = NULL;
567 if (verbose)
568 fprintf(stderr, "Checking object directory\n");
570 if (show_progress)
571 progress = start_progress(_("Checking object directories"), 256);
572 for (i = 0; i < 256; i++) {
573 static char dir[4096];
574 sprintf(dir, "%s/%02x", path, i);
575 fsck_dir(i, dir);
576 display_progress(progress, i+1);
578 stop_progress(&progress);
579 fsck_sha1_list();
582 static int fsck_head_link(void)
584 int flag;
585 int null_is_error = 0;
587 if (verbose)
588 fprintf(stderr, "Checking HEAD link\n");
590 head_points_at = resolve_ref_unsafe("HEAD", 0, head_oid.hash, &flag);
591 if (!head_points_at) {
592 errors_found |= ERROR_REFS;
593 return error("Invalid HEAD");
595 if (!strcmp(head_points_at, "HEAD"))
596 /* detached HEAD */
597 null_is_error = 1;
598 else if (!starts_with(head_points_at, "refs/heads/")) {
599 errors_found |= ERROR_REFS;
600 return error("HEAD points to something strange (%s)",
601 head_points_at);
603 if (is_null_oid(&head_oid)) {
604 if (null_is_error) {
605 errors_found |= ERROR_REFS;
606 return error("HEAD: detached HEAD points at nothing");
608 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
609 head_points_at + 11);
611 return 0;
614 static int fsck_cache_tree(struct cache_tree *it)
616 int i;
617 int err = 0;
619 if (verbose)
620 fprintf(stderr, "Checking cache tree\n");
622 if (0 <= it->entry_count) {
623 struct object *obj = parse_object(it->sha1);
624 if (!obj) {
625 error("%s: invalid sha1 pointer in cache-tree",
626 sha1_to_hex(it->sha1));
627 errors_found |= ERROR_REFS;
628 return 1;
630 obj->used = 1;
631 mark_object_reachable(obj);
632 if (obj->type != OBJ_TREE)
633 err |= objerror(obj, "non-tree in cache-tree");
635 for (i = 0; i < it->subtree_nr; i++)
636 err |= fsck_cache_tree(it->down[i]->cache_tree);
637 return err;
640 static char const * const fsck_usage[] = {
641 N_("git fsck [<options>] [<object>...]"),
642 NULL
645 static struct option fsck_opts[] = {
646 OPT__VERBOSE(&verbose, N_("be verbose")),
647 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
648 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
649 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
650 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
651 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
652 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
653 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
654 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
655 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
656 OPT_BOOL(0, "lost-found", &write_lost_and_found,
657 N_("write dangling objects in .git/lost-found")),
658 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
659 OPT_END(),
662 int cmd_fsck(int argc, const char **argv, const char *prefix)
664 int i, heads;
665 struct alternate_object_database *alt;
667 errors_found = 0;
668 check_replace_refs = 0;
670 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
672 fsck_walk_options.walk = mark_object;
673 fsck_obj_options.walk = mark_used;
674 fsck_obj_options.error_func = fsck_error_func;
675 if (check_strict)
676 fsck_obj_options.strict = 1;
678 if (show_progress == -1)
679 show_progress = isatty(2);
680 if (verbose)
681 show_progress = 0;
683 if (write_lost_and_found) {
684 check_full = 1;
685 include_reflogs = 0;
688 git_config(fsck_config, NULL);
690 fsck_head_link();
691 if (!connectivity_only)
692 fsck_object_dir(get_object_directory());
694 prepare_alt_odb();
695 for (alt = alt_odb_list; alt; alt = alt->next) {
696 char namebuf[PATH_MAX];
697 int namelen = alt->name - alt->base;
698 memcpy(namebuf, alt->base, namelen);
699 namebuf[namelen - 1] = 0;
700 fsck_object_dir(namebuf);
703 if (check_full) {
704 struct packed_git *p;
705 uint32_t total = 0, count = 0;
706 struct progress *progress = NULL;
708 prepare_packed_git();
710 if (show_progress) {
711 for (p = packed_git; p; p = p->next) {
712 if (open_pack_index(p))
713 continue;
714 total += p->num_objects;
717 progress = start_progress(_("Checking objects"), total);
719 for (p = packed_git; p; p = p->next) {
720 /* verify gives error messages itself */
721 if (verify_pack(p, fsck_obj_buffer,
722 progress, count))
723 errors_found |= ERROR_PACK;
724 count += p->num_objects;
726 stop_progress(&progress);
729 heads = 0;
730 for (i = 0; i < argc; i++) {
731 const char *arg = argv[i];
732 unsigned char sha1[20];
733 if (!get_sha1(arg, sha1)) {
734 struct object *obj = lookup_object(sha1);
736 /* Error is printed by lookup_object(). */
737 if (!obj)
738 continue;
740 obj->used = 1;
741 mark_object_reachable(obj);
742 heads++;
743 continue;
745 error("invalid parameter: expected sha1, got '%s'", arg);
749 * If we've not been given any explicit head information, do the
750 * default ones from .git/refs. We also consider the index file
751 * in this case (ie this implies --cache).
753 if (!heads) {
754 get_default_heads();
755 keep_cache_objects = 1;
758 if (keep_cache_objects) {
759 read_cache();
760 for (i = 0; i < active_nr; i++) {
761 unsigned int mode;
762 struct blob *blob;
763 struct object *obj;
765 mode = active_cache[i]->ce_mode;
766 if (S_ISGITLINK(mode))
767 continue;
768 blob = lookup_blob(active_cache[i]->sha1);
769 if (!blob)
770 continue;
771 obj = &blob->object;
772 obj->used = 1;
773 mark_object_reachable(obj);
775 if (active_cache_tree)
776 fsck_cache_tree(active_cache_tree);
779 check_connectivity();
780 return errors_found;