fsck: use for_each_loose_file_in_objdir
[git/git.git] / builtin / fsck.c
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
17 #define REACHABLE 0x0001
18 #define SEEN 0x0002
19 #define HAS_OBJ 0x0004
20
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
42 static int fsck_config(const char *var, const char *value, void *cb)
43 {
44 if (strcmp(var, "fsck.skiplist") == 0) {
45 const char *path;
46 struct strbuf sb = STRBUF_INIT;
47
48 if (git_config_pathname(&path, var, value))
49 return 1;
50 strbuf_addf(&sb, "skiplist=%s", path);
51 free((char *)path);
52 fsck_set_msg_types(&fsck_obj_options, sb.buf);
53 strbuf_release(&sb);
54 return 0;
55 }
56
57 if (skip_prefix(var, "fsck.", &var)) {
58 fsck_set_msg_type(&fsck_obj_options, var, value);
59 return 0;
60 }
61
62 return git_default_config(var, value, cb);
63 }
64
65 static void objreport(struct object *obj, const char *msg_type,
66 const char *err)
67 {
68 fprintf(stderr, "%s in %s %s: %s\n",
69 msg_type, typename(obj->type), sha1_to_hex(obj->sha1), err);
70 }
71
72 static int objerror(struct object *obj, const char *err)
73 {
74 errors_found |= ERROR_OBJECT;
75 objreport(obj, "error", err);
76 return -1;
77 }
78
79 static int fsck_error_func(struct object *obj, int type, const char *message)
80 {
81 objreport(obj, (type == FSCK_WARN) ? "warning" : "error", message);
82 return (type == FSCK_WARN) ? 0 : 1;
83 }
84
85 static struct object_array pending;
86
87 static int mark_object(struct object *obj, int type, void *data, struct fsck_options *options)
88 {
89 struct object *parent = data;
90
91 /*
92 * The only case data is NULL or type is OBJ_ANY is when
93 * mark_object_reachable() calls us. All the callers of
94 * that function has non-NULL obj hence ...
95 */
96 if (!obj) {
97 /* ... these references to parent->fld are safe here */
98 printf("broken link from %7s %s\n",
99 typename(parent->type), sha1_to_hex(parent->sha1));
100 printf("broken link from %7s %s\n",
101 (type == OBJ_ANY ? "unknown" : typename(type)), "unknown");
102 errors_found |= ERROR_REACHABLE;
103 return 1;
104 }
105
106 if (type != OBJ_ANY && obj->type != type)
107 /* ... and the reference to parent is safe here */
108 objerror(parent, "wrong object type in link");
109
110 if (obj->flags & REACHABLE)
111 return 0;
112 obj->flags |= REACHABLE;
113 if (!(obj->flags & HAS_OBJ)) {
114 if (parent && !has_sha1_file(obj->sha1)) {
115 printf("broken link from %7s %s\n",
116 typename(parent->type), sha1_to_hex(parent->sha1));
117 printf(" to %7s %s\n",
118 typename(obj->type), sha1_to_hex(obj->sha1));
119 errors_found |= ERROR_REACHABLE;
120 }
121 return 1;
122 }
123
124 add_object_array(obj, NULL, &pending);
125 return 0;
126 }
127
128 static void mark_object_reachable(struct object *obj)
129 {
130 mark_object(obj, OBJ_ANY, NULL, NULL);
131 }
132
133 static int traverse_one_object(struct object *obj)
134 {
135 int result;
136 struct tree *tree = NULL;
137
138 if (obj->type == OBJ_TREE) {
139 tree = (struct tree *)obj;
140 if (parse_tree(tree) < 0)
141 return 1; /* error already displayed */
142 }
143 result = fsck_walk(obj, obj, &fsck_walk_options);
144 if (tree)
145 free_tree_buffer(tree);
146 return result;
147 }
148
149 static int traverse_reachable(void)
150 {
151 struct progress *progress = NULL;
152 unsigned int nr = 0;
153 int result = 0;
154 if (show_progress)
155 progress = start_progress_delay(_("Checking connectivity"), 0, 0, 2);
156 while (pending.nr) {
157 struct object_array_entry *entry;
158 struct object *obj;
159
160 entry = pending.objects + --pending.nr;
161 obj = entry->item;
162 result |= traverse_one_object(obj);
163 display_progress(progress, ++nr);
164 }
165 stop_progress(&progress);
166 return !!result;
167 }
168
169 static int mark_used(struct object *obj, int type, void *data, struct fsck_options *options)
170 {
171 if (!obj)
172 return 1;
173 obj->used = 1;
174 return 0;
175 }
176
177 /*
178 * Check a single reachable object
179 */
180 static void check_reachable_object(struct object *obj)
181 {
182 /*
183 * We obviously want the object to be parsed,
184 * except if it was in a pack-file and we didn't
185 * do a full fsck
186 */
187 if (!(obj->flags & HAS_OBJ)) {
188 if (has_sha1_pack(obj->sha1))
189 return; /* it is in pack - forget about it */
190 if (connectivity_only && has_sha1_file(obj->sha1))
191 return;
192 printf("missing %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
193 errors_found |= ERROR_REACHABLE;
194 return;
195 }
196 }
197
198 /*
199 * Check a single unreachable object
200 */
201 static void check_unreachable_object(struct object *obj)
202 {
203 /*
204 * Missing unreachable object? Ignore it. It's not like
205 * we miss it (since it can't be reached), nor do we want
206 * to complain about it being unreachable (since it does
207 * not exist).
208 */
209 if (!obj->parsed)
210 return;
211
212 /*
213 * Unreachable object that exists? Show it if asked to,
214 * since this is something that is prunable.
215 */
216 if (show_unreachable) {
217 printf("unreachable %s %s\n", typename(obj->type), sha1_to_hex(obj->sha1));
218 return;
219 }
220
221 /*
222 * "!used" means that nothing at all points to it, including
223 * other unreachable objects. In other words, it's the "tip"
224 * of some set of unreachable objects, usually a commit that
225 * got dropped.
226 *
227 * Such starting points are more interesting than some random
228 * set of unreachable objects, so we show them even if the user
229 * hasn't asked for _all_ unreachable objects. If you have
230 * deleted a branch by mistake, this is a prime candidate to
231 * start looking at, for example.
232 */
233 if (!obj->used) {
234 if (show_dangling)
235 printf("dangling %s %s\n", typename(obj->type),
236 sha1_to_hex(obj->sha1));
237 if (write_lost_and_found) {
238 char *filename = git_pathdup("lost-found/%s/%s",
239 obj->type == OBJ_COMMIT ? "commit" : "other",
240 sha1_to_hex(obj->sha1));
241 FILE *f;
242
243 if (safe_create_leading_directories_const(filename)) {
244 error("Could not create lost-found");
245 free(filename);
246 return;
247 }
248 if (!(f = fopen(filename, "w")))
249 die_errno("Could not open '%s'", filename);
250 if (obj->type == OBJ_BLOB) {
251 if (stream_blob_to_fd(fileno(f), obj->sha1, NULL, 1))
252 die_errno("Could not write '%s'", filename);
253 } else
254 fprintf(f, "%s\n", sha1_to_hex(obj->sha1));
255 if (fclose(f))
256 die_errno("Could not finish '%s'",
257 filename);
258 free(filename);
259 }
260 return;
261 }
262
263 /*
264 * Otherwise? It's there, it's unreachable, and some other unreachable
265 * object points to it. Ignore it - it's not interesting, and we showed
266 * all the interesting cases above.
267 */
268 }
269
270 static void check_object(struct object *obj)
271 {
272 if (verbose)
273 fprintf(stderr, "Checking %s\n", sha1_to_hex(obj->sha1));
274
275 if (obj->flags & REACHABLE)
276 check_reachable_object(obj);
277 else
278 check_unreachable_object(obj);
279 }
280
281 static void check_connectivity(void)
282 {
283 int i, max;
284
285 /* Traverse the pending reachable objects */
286 traverse_reachable();
287
288 /* Look up all the requirements, warn about missing objects.. */
289 max = get_max_object_index();
290 if (verbose)
291 fprintf(stderr, "Checking connectivity (%d objects)\n", max);
292
293 for (i = 0; i < max; i++) {
294 struct object *obj = get_indexed_object(i);
295
296 if (obj)
297 check_object(obj);
298 }
299 }
300
301 static int fsck_obj(struct object *obj)
302 {
303 if (obj->flags & SEEN)
304 return 0;
305 obj->flags |= SEEN;
306
307 if (verbose)
308 fprintf(stderr, "Checking %s %s\n",
309 typename(obj->type), sha1_to_hex(obj->sha1));
310
311 if (fsck_walk(obj, NULL, &fsck_obj_options))
312 objerror(obj, "broken links");
313 if (fsck_object(obj, NULL, 0, &fsck_obj_options))
314 return -1;
315
316 if (obj->type == OBJ_TREE) {
317 struct tree *item = (struct tree *) obj;
318
319 free_tree_buffer(item);
320 }
321
322 if (obj->type == OBJ_COMMIT) {
323 struct commit *commit = (struct commit *) obj;
324
325 free_commit_buffer(commit);
326
327 if (!commit->parents && show_root)
328 printf("root %s\n", sha1_to_hex(commit->object.sha1));
329 }
330
331 if (obj->type == OBJ_TAG) {
332 struct tag *tag = (struct tag *) obj;
333
334 if (show_tags && tag->tagged) {
335 printf("tagged %s %s", typename(tag->tagged->type), sha1_to_hex(tag->tagged->sha1));
336 printf(" (%s) in %s\n", tag->tag, sha1_to_hex(tag->object.sha1));
337 }
338 }
339
340 return 0;
341 }
342
343 static int fsck_sha1(const unsigned char *sha1)
344 {
345 struct object *obj = parse_object(sha1);
346 if (!obj) {
347 errors_found |= ERROR_OBJECT;
348 return error("%s: object corrupt or missing",
349 sha1_to_hex(sha1));
350 }
351 obj->flags |= HAS_OBJ;
352 return fsck_obj(obj);
353 }
354
355 static int fsck_obj_buffer(const unsigned char *sha1, enum object_type type,
356 unsigned long size, void *buffer, int *eaten)
357 {
358 struct object *obj;
359 obj = parse_object_buffer(sha1, type, size, buffer, eaten);
360 if (!obj) {
361 errors_found |= ERROR_OBJECT;
362 return error("%s: object corrupt or missing", sha1_to_hex(sha1));
363 }
364 obj->flags = HAS_OBJ;
365 return fsck_obj(obj);
366 }
367
368 static int default_refs;
369
370 static void fsck_handle_reflog_sha1(const char *refname, unsigned char *sha1)
371 {
372 struct object *obj;
373
374 if (!is_null_sha1(sha1)) {
375 obj = lookup_object(sha1);
376 if (obj) {
377 obj->used = 1;
378 mark_object_reachable(obj);
379 } else {
380 error("%s: invalid reflog entry %s", refname, sha1_to_hex(sha1));
381 errors_found |= ERROR_REACHABLE;
382 }
383 }
384 }
385
386 static int fsck_handle_reflog_ent(unsigned char *osha1, unsigned char *nsha1,
387 const char *email, unsigned long timestamp, int tz,
388 const char *message, void *cb_data)
389 {
390 const char *refname = cb_data;
391
392 if (verbose)
393 fprintf(stderr, "Checking reflog %s->%s\n",
394 sha1_to_hex(osha1), sha1_to_hex(nsha1));
395
396 fsck_handle_reflog_sha1(refname, osha1);
397 fsck_handle_reflog_sha1(refname, nsha1);
398 return 0;
399 }
400
401 static int fsck_handle_reflog(const char *logname, const struct object_id *oid,
402 int flag, void *cb_data)
403 {
404 for_each_reflog_ent(logname, fsck_handle_reflog_ent, (void *)logname);
405 return 0;
406 }
407
408 static int fsck_handle_ref(const char *refname, const struct object_id *oid,
409 int flag, void *cb_data)
410 {
411 struct object *obj;
412
413 obj = parse_object(oid->hash);
414 if (!obj) {
415 error("%s: invalid sha1 pointer %s", refname, oid_to_hex(oid));
416 errors_found |= ERROR_REACHABLE;
417 /* We'll continue with the rest despite the error.. */
418 return 0;
419 }
420 if (obj->type != OBJ_COMMIT && is_branch(refname))
421 error("%s: not a commit", refname);
422 default_refs++;
423 obj->used = 1;
424 mark_object_reachable(obj);
425
426 return 0;
427 }
428
429 static void get_default_heads(void)
430 {
431 if (head_points_at && !is_null_oid(&head_oid))
432 fsck_handle_ref("HEAD", &head_oid, 0, NULL);
433 for_each_rawref(fsck_handle_ref, NULL);
434 if (include_reflogs)
435 for_each_reflog(fsck_handle_reflog, NULL);
436
437 /*
438 * Not having any default heads isn't really fatal, but
439 * it does mean that "--unreachable" no longer makes any
440 * sense (since in this case everything will obviously
441 * be unreachable by definition.
442 *
443 * Showing dangling objects is valid, though (as those
444 * dangling objects are likely lost heads).
445 *
446 * So we just print a warning about it, and clear the
447 * "show_unreachable" flag.
448 */
449 if (!default_refs) {
450 fprintf(stderr, "notice: No default references\n");
451 show_unreachable = 0;
452 }
453 }
454
455 static int fsck_loose(const unsigned char *sha1, const char *path, void *data)
456 {
457 if (fsck_sha1(sha1))
458 errors_found |= ERROR_OBJECT;
459 return 0;
460 }
461
462 static int fsck_cruft(const char *basename, const char *path, void *data)
463 {
464 if (!starts_with(basename, "tmp_obj_"))
465 fprintf(stderr, "bad sha1 file: %s\n", path);
466 return 0;
467 }
468
469 static int fsck_subdir(int nr, const char *path, void *progress)
470 {
471 display_progress(progress, nr + 1);
472 return 0;
473 }
474
475 static void fsck_object_dir(const char *path)
476 {
477 struct progress *progress = NULL;
478
479 if (verbose)
480 fprintf(stderr, "Checking object directory\n");
481
482 if (show_progress)
483 progress = start_progress(_("Checking object directories"), 256);
484
485 for_each_loose_file_in_objdir(path, fsck_loose, fsck_cruft, fsck_subdir,
486 progress);
487 display_progress(progress, 256);
488 stop_progress(&progress);
489 }
490
491 static int fsck_head_link(void)
492 {
493 int flag;
494 int null_is_error = 0;
495
496 if (verbose)
497 fprintf(stderr, "Checking HEAD link\n");
498
499 head_points_at = resolve_ref_unsafe("HEAD", 0, head_oid.hash, &flag);
500 if (!head_points_at)
501 return error("Invalid HEAD");
502 if (!strcmp(head_points_at, "HEAD"))
503 /* detached HEAD */
504 null_is_error = 1;
505 else if (!starts_with(head_points_at, "refs/heads/"))
506 return error("HEAD points to something strange (%s)",
507 head_points_at);
508 if (is_null_oid(&head_oid)) {
509 if (null_is_error)
510 return error("HEAD: detached HEAD points at nothing");
511 fprintf(stderr, "notice: HEAD points to an unborn branch (%s)\n",
512 head_points_at + 11);
513 }
514 return 0;
515 }
516
517 static int fsck_cache_tree(struct cache_tree *it)
518 {
519 int i;
520 int err = 0;
521
522 if (verbose)
523 fprintf(stderr, "Checking cache tree\n");
524
525 if (0 <= it->entry_count) {
526 struct object *obj = parse_object(it->sha1);
527 if (!obj) {
528 error("%s: invalid sha1 pointer in cache-tree",
529 sha1_to_hex(it->sha1));
530 return 1;
531 }
532 obj->used = 1;
533 mark_object_reachable(obj);
534 if (obj->type != OBJ_TREE)
535 err |= objerror(obj, "non-tree in cache-tree");
536 }
537 for (i = 0; i < it->subtree_nr; i++)
538 err |= fsck_cache_tree(it->down[i]->cache_tree);
539 return err;
540 }
541
542 static char const * const fsck_usage[] = {
543 N_("git fsck [<options>] [<object>...]"),
544 NULL
545 };
546
547 static struct option fsck_opts[] = {
548 OPT__VERBOSE(&verbose, N_("be verbose")),
549 OPT_BOOL(0, "unreachable", &show_unreachable, N_("show unreachable objects")),
550 OPT_BOOL(0, "dangling", &show_dangling, N_("show dangling objects")),
551 OPT_BOOL(0, "tags", &show_tags, N_("report tags")),
552 OPT_BOOL(0, "root", &show_root, N_("report root nodes")),
553 OPT_BOOL(0, "cache", &keep_cache_objects, N_("make index objects head nodes")),
554 OPT_BOOL(0, "reflogs", &include_reflogs, N_("make reflogs head nodes (default)")),
555 OPT_BOOL(0, "full", &check_full, N_("also consider packs and alternate objects")),
556 OPT_BOOL(0, "connectivity-only", &connectivity_only, N_("check only connectivity")),
557 OPT_BOOL(0, "strict", &check_strict, N_("enable more strict checking")),
558 OPT_BOOL(0, "lost-found", &write_lost_and_found,
559 N_("write dangling objects in .git/lost-found")),
560 OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
561 OPT_END(),
562 };
563
564 int cmd_fsck(int argc, const char **argv, const char *prefix)
565 {
566 int i, heads;
567 struct alternate_object_database *alt;
568
569 errors_found = 0;
570 check_replace_refs = 0;
571
572 argc = parse_options(argc, argv, prefix, fsck_opts, fsck_usage, 0);
573
574 fsck_walk_options.walk = mark_object;
575 fsck_obj_options.walk = mark_used;
576 fsck_obj_options.error_func = fsck_error_func;
577 if (check_strict)
578 fsck_obj_options.strict = 1;
579
580 if (show_progress == -1)
581 show_progress = isatty(2);
582 if (verbose)
583 show_progress = 0;
584
585 if (write_lost_and_found) {
586 check_full = 1;
587 include_reflogs = 0;
588 }
589
590 git_config(fsck_config, NULL);
591
592 fsck_head_link();
593 if (!connectivity_only) {
594 fsck_object_dir(get_object_directory());
595
596 prepare_alt_odb();
597 for (alt = alt_odb_list; alt; alt = alt->next) {
598 /* directory name, minus trailing slash */
599 size_t namelen = alt->name - alt->base - 1;
600 struct strbuf name = STRBUF_INIT;
601 strbuf_add(&name, alt->base, namelen);
602 fsck_object_dir(name.buf);
603 strbuf_release(&name);
604 }
605 }
606
607 if (check_full) {
608 struct packed_git *p;
609 uint32_t total = 0, count = 0;
610 struct progress *progress = NULL;
611
612 prepare_packed_git();
613
614 if (show_progress) {
615 for (p = packed_git; p; p = p->next) {
616 if (open_pack_index(p))
617 continue;
618 total += p->num_objects;
619 }
620
621 progress = start_progress(_("Checking objects"), total);
622 }
623 for (p = packed_git; p; p = p->next) {
624 /* verify gives error messages itself */
625 if (verify_pack(p, fsck_obj_buffer,
626 progress, count))
627 errors_found |= ERROR_PACK;
628 count += p->num_objects;
629 }
630 stop_progress(&progress);
631 }
632
633 heads = 0;
634 for (i = 0; i < argc; i++) {
635 const char *arg = argv[i];
636 unsigned char sha1[20];
637 if (!get_sha1(arg, sha1)) {
638 struct object *obj = lookup_object(sha1);
639
640 /* Error is printed by lookup_object(). */
641 if (!obj)
642 continue;
643
644 obj->used = 1;
645 mark_object_reachable(obj);
646 heads++;
647 continue;
648 }
649 error("invalid parameter: expected sha1, got '%s'", arg);
650 }
651
652 /*
653 * If we've not been given any explicit head information, do the
654 * default ones from .git/refs. We also consider the index file
655 * in this case (ie this implies --cache).
656 */
657 if (!heads) {
658 get_default_heads();
659 keep_cache_objects = 1;
660 }
661
662 if (keep_cache_objects) {
663 read_cache();
664 for (i = 0; i < active_nr; i++) {
665 unsigned int mode;
666 struct blob *blob;
667 struct object *obj;
668
669 mode = active_cache[i]->ce_mode;
670 if (S_ISGITLINK(mode))
671 continue;
672 blob = lookup_blob(active_cache[i]->sha1);
673 if (!blob)
674 continue;
675 obj = &blob->object;
676 obj->used = 1;
677 mark_object_reachable(obj);
678 }
679 if (active_cache_tree)
680 fsck_cache_tree(active_cache_tree);
681 }
682
683 check_connectivity();
684 return errors_found;
685 }