3 #include "repository.h"
11 #include "cache-tree.h"
12 #include "tree-walk.h"
14 #include "parse-options.h"
17 #include "streaming.h"
20 #include "object-store.h"
22 #define REACHABLE 0x0001
24 #define HAS_OBJ 0x0004
25 /* This flag is set if something points to this object. */
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
;
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
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
;
58 strbuf_addstr(&buf
, oid_to_hex(&obj
->oid
));
60 strbuf_addf(&buf
, " (%s)", name
);
65 static const char *printable_type(struct object
*obj
)
69 if (obj
->type
== OBJ_NONE
) {
70 enum object_type type
= oid_object_info(the_repository
,
73 object_as_type(obj
, type
, 0);
76 ret
= type_name(obj
->type
);
83 static int fsck_config(const char *var
, const char *value
, void *cb
)
85 if (strcmp(var
, "fsck.skiplist") == 0) {
87 struct strbuf sb
= STRBUF_INIT
;
89 if (git_config_pathname(&path
, var
, value
))
91 strbuf_addf(&sb
, "skiplist=%s", path
);
93 fsck_set_msg_types(&fsck_obj_options
, sb
.buf
);
98 if (skip_prefix(var
, "fsck.", &var
)) {
99 fsck_set_msg_type(&fsck_obj_options
, var
, value
);
103 return git_default_config(var
, value
, cb
);
106 static void objreport(struct object
*obj
, const char *msg_type
,
109 fprintf(stderr
, "%s in %s %s: %s\n",
110 msg_type
, printable_type(obj
), describe_object(obj
), err
);
113 static int objerror(struct object
*obj
, const char *err
)
115 errors_found
|= ERROR_OBJECT
;
116 objreport(obj
, "error", err
);
120 static int fsck_error_func(struct fsck_options
*o
,
121 struct object
*obj
, int type
, const char *message
)
123 objreport(obj
, (type
== FSCK_WARN
) ? "warning" : "error", message
);
124 return (type
== FSCK_WARN
) ? 0 : 1;
127 static struct object_array pending
;
129 static int mark_object(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
131 struct object
*parent
= data
;
134 * The only case data is NULL or type is OBJ_ANY is when
135 * mark_object_reachable() calls us. All the callers of
136 * that function has non-NULL obj hence ...
139 /* ... these references to parent->fld are safe here */
140 printf("broken link from %7s %s\n",
141 printable_type(parent
), describe_object(parent
));
142 printf("broken link from %7s %s\n",
143 (type
== OBJ_ANY
? "unknown" : type_name(type
)), "unknown");
144 errors_found
|= ERROR_REACHABLE
;
148 if (type
!= OBJ_ANY
&& obj
->type
!= type
)
149 /* ... and the reference to parent is safe here */
150 objerror(parent
, "wrong object type in link");
152 if (obj
->flags
& REACHABLE
)
154 obj
->flags
|= REACHABLE
;
156 if (is_promisor_object(&obj
->oid
))
158 * Further recursion does not need to be performed on this
159 * object since it is a promisor object (so it does not need to
160 * be added to "pending").
164 if (!(obj
->flags
& HAS_OBJ
)) {
165 if (parent
&& !has_object_file(&obj
->oid
)) {
166 printf("broken link from %7s %s\n",
167 printable_type(parent
), describe_object(parent
));
168 printf(" to %7s %s\n",
169 printable_type(obj
), describe_object(obj
));
170 errors_found
|= ERROR_REACHABLE
;
175 add_object_array(obj
, NULL
, &pending
);
179 static void mark_object_reachable(struct object
*obj
)
181 mark_object(obj
, OBJ_ANY
, NULL
, NULL
);
184 static int traverse_one_object(struct object
*obj
)
186 int result
= fsck_walk(obj
, obj
, &fsck_walk_options
);
188 if (obj
->type
== OBJ_TREE
) {
189 struct tree
*tree
= (struct tree
*)obj
;
190 free_tree_buffer(tree
);
195 static int traverse_reachable(void)
197 struct progress
*progress
= NULL
;
201 progress
= start_delayed_progress(_("Checking connectivity"), 0);
203 result
|= traverse_one_object(object_array_pop(&pending
));
204 display_progress(progress
, ++nr
);
206 stop_progress(&progress
);
210 static int mark_used(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
219 * Check a single reachable object
221 static void check_reachable_object(struct object
*obj
)
224 * We obviously want the object to be parsed,
225 * except if it was in a pack-file and we didn't
228 if (!(obj
->flags
& HAS_OBJ
)) {
229 if (is_promisor_object(&obj
->oid
))
231 if (has_object_pack(&obj
->oid
))
232 return; /* it is in pack - forget about it */
233 printf("missing %s %s\n", printable_type(obj
),
234 describe_object(obj
));
235 errors_found
|= ERROR_REACHABLE
;
241 * Check a single unreachable object
243 static void check_unreachable_object(struct object
*obj
)
246 * Missing unreachable object? Ignore it. It's not like
247 * we miss it (since it can't be reached), nor do we want
248 * to complain about it being unreachable (since it does
251 if (!(obj
->flags
& HAS_OBJ
))
255 * Unreachable object that exists? Show it if asked to,
256 * since this is something that is prunable.
258 if (show_unreachable
) {
259 printf("unreachable %s %s\n", printable_type(obj
),
260 describe_object(obj
));
265 * "!USED" means that nothing at all points to it, including
266 * other unreachable objects. In other words, it's the "tip"
267 * of some set of unreachable objects, usually a commit that
270 * Such starting points are more interesting than some random
271 * set of unreachable objects, so we show them even if the user
272 * hasn't asked for _all_ unreachable objects. If you have
273 * deleted a branch by mistake, this is a prime candidate to
274 * start looking at, for example.
276 if (!(obj
->flags
& USED
)) {
278 printf("dangling %s %s\n", printable_type(obj
),
279 describe_object(obj
));
280 if (write_lost_and_found
) {
281 char *filename
= git_pathdup("lost-found/%s/%s",
282 obj
->type
== OBJ_COMMIT
? "commit" : "other",
283 describe_object(obj
));
286 if (safe_create_leading_directories_const(filename
)) {
287 error("Could not create lost-found");
291 f
= xfopen(filename
, "w");
292 if (obj
->type
== OBJ_BLOB
) {
293 if (stream_blob_to_fd(fileno(f
), &obj
->oid
, NULL
, 1))
294 die_errno("Could not write '%s'", filename
);
296 fprintf(f
, "%s\n", describe_object(obj
));
298 die_errno("Could not finish '%s'",
306 * Otherwise? It's there, it's unreachable, and some other unreachable
307 * object points to it. Ignore it - it's not interesting, and we showed
308 * all the interesting cases above.
312 static void check_object(struct object
*obj
)
315 fprintf(stderr
, "Checking %s\n", describe_object(obj
));
317 if (obj
->flags
& REACHABLE
)
318 check_reachable_object(obj
);
320 check_unreachable_object(obj
);
323 static void check_connectivity(void)
327 /* Traverse the pending reachable objects */
328 traverse_reachable();
330 /* Look up all the requirements, warn about missing objects.. */
331 max
= get_max_object_index();
333 fprintf(stderr
, "Checking connectivity (%d objects)\n", max
);
335 for (i
= 0; i
< max
; i
++) {
336 struct object
*obj
= get_indexed_object(i
);
343 static int fsck_obj(struct object
*obj
, void *buffer
, unsigned long size
)
347 if (obj
->flags
& SEEN
)
352 fprintf(stderr
, "Checking %s %s\n",
353 printable_type(obj
), describe_object(obj
));
355 if (fsck_walk(obj
, NULL
, &fsck_obj_options
))
356 objerror(obj
, "broken links");
357 err
= fsck_object(obj
, buffer
, size
, &fsck_obj_options
);
361 if (obj
->type
== OBJ_COMMIT
) {
362 struct commit
*commit
= (struct commit
*) obj
;
364 if (!commit
->parents
&& show_root
)
365 printf("root %s\n", describe_object(&commit
->object
));
368 if (obj
->type
== OBJ_TAG
) {
369 struct tag
*tag
= (struct tag
*) obj
;
371 if (show_tags
&& tag
->tagged
) {
372 printf("tagged %s %s", printable_type(tag
->tagged
),
373 describe_object(tag
->tagged
));
374 printf(" (%s) in %s\n", tag
->tag
,
375 describe_object(&tag
->object
));
380 if (obj
->type
== OBJ_TREE
)
381 free_tree_buffer((struct tree
*)obj
);
382 if (obj
->type
== OBJ_COMMIT
)
383 free_commit_buffer((struct commit
*)obj
);
387 static int fsck_obj_buffer(const struct object_id
*oid
, enum object_type type
,
388 unsigned long size
, void *buffer
, int *eaten
)
391 * Note, buffer may be NULL if type is OBJ_BLOB. See
392 * verify_packfile(), data_valid variable for details.
395 obj
= parse_object_buffer(oid
, type
, size
, buffer
, eaten
);
397 errors_found
|= ERROR_OBJECT
;
398 return error("%s: object corrupt or missing", oid_to_hex(oid
));
400 obj
->flags
&= ~(REACHABLE
| SEEN
);
401 obj
->flags
|= HAS_OBJ
;
402 return fsck_obj(obj
, buffer
, size
);
405 static int default_refs
;
407 static void fsck_handle_reflog_oid(const char *refname
, struct object_id
*oid
,
408 timestamp_t timestamp
)
412 if (!is_null_oid(oid
)) {
413 obj
= lookup_object(oid
->hash
);
414 if (obj
&& (obj
->flags
& HAS_OBJ
)) {
415 if (timestamp
&& name_objects
)
416 add_decoration(fsck_walk_options
.object_names
,
418 xstrfmt("%s@{%"PRItime
"}", refname
, timestamp
));
420 mark_object_reachable(obj
);
421 } else if (!is_promisor_object(oid
)) {
422 error("%s: invalid reflog entry %s", refname
, oid_to_hex(oid
));
423 errors_found
|= ERROR_REACHABLE
;
428 static int fsck_handle_reflog_ent(struct object_id
*ooid
, struct object_id
*noid
,
429 const char *email
, timestamp_t timestamp
, int tz
,
430 const char *message
, void *cb_data
)
432 const char *refname
= cb_data
;
435 fprintf(stderr
, "Checking reflog %s->%s\n",
436 oid_to_hex(ooid
), oid_to_hex(noid
));
438 fsck_handle_reflog_oid(refname
, ooid
, 0);
439 fsck_handle_reflog_oid(refname
, noid
, timestamp
);
443 static int fsck_handle_reflog(const char *logname
, const struct object_id
*oid
,
444 int flag
, void *cb_data
)
446 for_each_reflog_ent(logname
, fsck_handle_reflog_ent
, (void *)logname
);
450 static int fsck_handle_ref(const char *refname
, const struct object_id
*oid
,
451 int flag
, void *cb_data
)
455 obj
= parse_object(oid
);
457 if (is_promisor_object(oid
)) {
459 * Increment default_refs anyway, because this is a
465 error("%s: invalid sha1 pointer %s", refname
, oid_to_hex(oid
));
466 errors_found
|= ERROR_REACHABLE
;
467 /* We'll continue with the rest despite the error.. */
470 if (obj
->type
!= OBJ_COMMIT
&& is_branch(refname
)) {
471 error("%s: not a commit", refname
);
472 errors_found
|= ERROR_REFS
;
477 add_decoration(fsck_walk_options
.object_names
,
478 obj
, xstrdup(refname
));
479 mark_object_reachable(obj
);
484 static void get_default_heads(void)
486 if (head_points_at
&& !is_null_oid(&head_oid
))
487 fsck_handle_ref("HEAD", &head_oid
, 0, NULL
);
488 for_each_rawref(fsck_handle_ref
, NULL
);
490 for_each_reflog(fsck_handle_reflog
, NULL
);
493 * Not having any default heads isn't really fatal, but
494 * it does mean that "--unreachable" no longer makes any
495 * sense (since in this case everything will obviously
496 * be unreachable by definition.
498 * Showing dangling objects is valid, though (as those
499 * dangling objects are likely lost heads).
501 * So we just print a warning about it, and clear the
502 * "show_unreachable" flag.
505 fprintf(stderr
, "notice: No default references\n");
506 show_unreachable
= 0;
510 static int fsck_loose(const struct object_id
*oid
, const char *path
, void *data
)
513 enum object_type type
;
518 if (read_loose_object(path
, oid
, &type
, &size
, &contents
) < 0) {
519 errors_found
|= ERROR_OBJECT
;
520 error("%s: object corrupt or missing: %s",
521 oid_to_hex(oid
), path
);
522 return 0; /* keep checking other objects */
525 if (!contents
&& type
!= OBJ_BLOB
)
526 BUG("read_loose_object streamed a non-blob");
528 obj
= parse_object_buffer(oid
, type
, size
, contents
, &eaten
);
530 errors_found
|= ERROR_OBJECT
;
531 error("%s: object could not be parsed: %s",
532 oid_to_hex(oid
), path
);
535 return 0; /* keep checking other objects */
538 obj
->flags
&= ~(REACHABLE
| SEEN
);
539 obj
->flags
|= HAS_OBJ
;
540 if (fsck_obj(obj
, contents
, size
))
541 errors_found
|= ERROR_OBJECT
;
545 return 0; /* keep checking other objects, even if we saw an error */
548 static int fsck_cruft(const char *basename
, const char *path
, void *data
)
550 if (!starts_with(basename
, "tmp_obj_"))
551 fprintf(stderr
, "bad sha1 file: %s\n", path
);
555 static int fsck_subdir(unsigned int nr
, const char *path
, void *progress
)
557 display_progress(progress
, nr
+ 1);
561 static void fsck_object_dir(const char *path
)
563 struct progress
*progress
= NULL
;
566 fprintf(stderr
, "Checking object directory\n");
569 progress
= start_progress(_("Checking object directories"), 256);
571 for_each_loose_file_in_objdir(path
, fsck_loose
, fsck_cruft
, fsck_subdir
,
573 display_progress(progress
, 256);
574 stop_progress(&progress
);
577 static int fsck_head_link(void)
579 int null_is_error
= 0;
582 fprintf(stderr
, "Checking HEAD link\n");
584 head_points_at
= resolve_ref_unsafe("HEAD", 0, &head_oid
, NULL
);
585 if (!head_points_at
) {
586 errors_found
|= ERROR_REFS
;
587 return error("Invalid HEAD");
589 if (!strcmp(head_points_at
, "HEAD"))
592 else if (!starts_with(head_points_at
, "refs/heads/")) {
593 errors_found
|= ERROR_REFS
;
594 return error("HEAD points to something strange (%s)",
597 if (is_null_oid(&head_oid
)) {
599 errors_found
|= ERROR_REFS
;
600 return error("HEAD: detached HEAD points at nothing");
602 fprintf(stderr
, "notice: HEAD points to an unborn branch (%s)\n",
603 head_points_at
+ 11);
608 static int fsck_cache_tree(struct cache_tree
*it
)
614 fprintf(stderr
, "Checking cache tree\n");
616 if (0 <= it
->entry_count
) {
617 struct object
*obj
= parse_object(&it
->oid
);
619 error("%s: invalid sha1 pointer in cache-tree",
620 oid_to_hex(&it
->oid
));
621 errors_found
|= ERROR_REFS
;
626 add_decoration(fsck_walk_options
.object_names
,
628 mark_object_reachable(obj
);
629 if (obj
->type
!= OBJ_TREE
)
630 err
|= objerror(obj
, "non-tree in cache-tree");
632 for (i
= 0; i
< it
->subtree_nr
; i
++)
633 err
|= fsck_cache_tree(it
->down
[i
]->cache_tree
);
637 static void mark_object_for_connectivity(const struct object_id
*oid
)
639 struct object
*obj
= lookup_unknown_object(oid
->hash
);
640 obj
->flags
|= HAS_OBJ
;
643 static int mark_loose_for_connectivity(const struct object_id
*oid
,
647 mark_object_for_connectivity(oid
);
651 static int mark_packed_for_connectivity(const struct object_id
*oid
,
652 struct packed_git
*pack
,
656 mark_object_for_connectivity(oid
);
660 static char const * const fsck_usage
[] = {
661 N_("git fsck [<options>] [<object>...]"),
665 static struct option fsck_opts
[] = {
666 OPT__VERBOSE(&verbose
, N_("be verbose")),
667 OPT_BOOL(0, "unreachable", &show_unreachable
, N_("show unreachable objects")),
668 OPT_BOOL(0, "dangling", &show_dangling
, N_("show dangling objects")),
669 OPT_BOOL(0, "tags", &show_tags
, N_("report tags")),
670 OPT_BOOL(0, "root", &show_root
, N_("report root nodes")),
671 OPT_BOOL(0, "cache", &keep_cache_objects
, N_("make index objects head nodes")),
672 OPT_BOOL(0, "reflogs", &include_reflogs
, N_("make reflogs head nodes (default)")),
673 OPT_BOOL(0, "full", &check_full
, N_("also consider packs and alternate objects")),
674 OPT_BOOL(0, "connectivity-only", &connectivity_only
, N_("check only connectivity")),
675 OPT_BOOL(0, "strict", &check_strict
, N_("enable more strict checking")),
676 OPT_BOOL(0, "lost-found", &write_lost_and_found
,
677 N_("write dangling objects in .git/lost-found")),
678 OPT_BOOL(0, "progress", &show_progress
, N_("show progress")),
679 OPT_BOOL(0, "name-objects", &name_objects
, N_("show verbose names for reachable objects")),
683 int cmd_fsck(int argc
, const char **argv
, const char *prefix
)
686 struct alternate_object_database
*alt
;
688 /* fsck knows how to handle missing promisor objects */
689 fetch_if_missing
= 0;
692 check_replace_refs
= 0;
694 argc
= parse_options(argc
, argv
, prefix
, fsck_opts
, fsck_usage
, 0);
696 fsck_walk_options
.walk
= mark_object
;
697 fsck_obj_options
.walk
= mark_used
;
698 fsck_obj_options
.error_func
= fsck_error_func
;
700 fsck_obj_options
.strict
= 1;
702 if (show_progress
== -1)
703 show_progress
= isatty(2);
707 if (write_lost_and_found
) {
713 fsck_walk_options
.object_names
=
714 xcalloc(1, sizeof(struct decoration
));
716 git_config(fsck_config
, NULL
);
719 if (connectivity_only
) {
720 for_each_loose_object(mark_loose_for_connectivity
, NULL
, 0);
721 for_each_packed_object(mark_packed_for_connectivity
, NULL
, 0);
723 struct alternate_object_database
*alt_odb_list
;
725 fsck_object_dir(get_object_directory());
727 prepare_alt_odb(the_repository
);
728 alt_odb_list
= the_repository
->objects
->alt_odb_list
;
729 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
)
730 fsck_object_dir(alt
->path
);
733 struct packed_git
*p
;
734 uint32_t total
= 0, count
= 0;
735 struct progress
*progress
= NULL
;
738 for (p
= get_packed_git(the_repository
); p
;
740 if (open_pack_index(p
))
742 total
+= p
->num_objects
;
745 progress
= start_progress(_("Checking objects"), total
);
747 for (p
= get_packed_git(the_repository
); p
;
749 /* verify gives error messages itself */
750 if (verify_pack(p
, fsck_obj_buffer
,
752 errors_found
|= ERROR_PACK
;
753 count
+= p
->num_objects
;
755 stop_progress(&progress
);
758 if (fsck_finish(&fsck_obj_options
))
759 errors_found
|= ERROR_OBJECT
;
762 for (i
= 0; i
< argc
; i
++) {
763 const char *arg
= argv
[i
];
764 struct object_id oid
;
765 if (!get_oid(arg
, &oid
)) {
766 struct object
*obj
= lookup_object(oid
.hash
);
768 if (!obj
|| !(obj
->flags
& HAS_OBJ
)) {
769 if (is_promisor_object(&oid
))
771 error("%s: object missing", oid_to_hex(&oid
));
772 errors_found
|= ERROR_OBJECT
;
778 add_decoration(fsck_walk_options
.object_names
,
780 mark_object_reachable(obj
);
783 error("invalid parameter: expected sha1, got '%s'", arg
);
784 errors_found
|= ERROR_OBJECT
;
788 * If we've not been given any explicit head information, do the
789 * default ones from .git/refs. We also consider the index file
790 * in this case (ie this implies --cache).
794 keep_cache_objects
= 1;
797 if (keep_cache_objects
) {
798 verify_index_checksum
= 1;
801 for (i
= 0; i
< active_nr
; i
++) {
806 mode
= active_cache
[i
]->ce_mode
;
807 if (S_ISGITLINK(mode
))
809 blob
= lookup_blob(&active_cache
[i
]->oid
);
815 add_decoration(fsck_walk_options
.object_names
,
817 xstrfmt(":%s", active_cache
[i
]->name
));
818 mark_object_reachable(obj
);
820 if (active_cache_tree
)
821 fsck_cache_tree(active_cache_tree
);
824 check_connectivity();