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"
21 #include "run-command.h"
23 #define REACHABLE 0x0001
25 #define HAS_OBJ 0x0004
26 /* This flag is set if something points to this object. */
31 static int show_unreachable
;
32 static int include_reflogs
= 1;
33 static int check_full
= 1;
34 static int connectivity_only
;
35 static int check_strict
;
36 static int keep_cache_objects
;
37 static struct fsck_options fsck_walk_options
= FSCK_OPTIONS_DEFAULT
;
38 static struct fsck_options fsck_obj_options
= FSCK_OPTIONS_DEFAULT
;
39 static struct object_id head_oid
;
40 static const char *head_points_at
;
41 static int errors_found
;
42 static int write_lost_and_found
;
44 static int show_progress
= -1;
45 static int show_dangling
= 1;
46 static int name_objects
;
47 #define ERROR_OBJECT 01
48 #define ERROR_REACHABLE 02
50 #define ERROR_REFS 010
51 #define ERROR_COMMIT_GRAPH 020
53 static const char *describe_object(struct object
*obj
)
55 static struct strbuf bufs
[] = {
56 STRBUF_INIT
, STRBUF_INIT
, STRBUF_INIT
, STRBUF_INIT
63 name
= lookup_decoration(fsck_walk_options
.object_names
, obj
);
66 b
= (b
+ 1) % ARRAY_SIZE(bufs
);
68 strbuf_addstr(buf
, oid_to_hex(&obj
->oid
));
70 strbuf_addf(buf
, " (%s)", name
);
75 static const char *printable_type(struct object
*obj
)
79 if (obj
->type
== OBJ_NONE
) {
80 enum object_type type
= oid_object_info(the_repository
,
83 object_as_type(the_repository
, obj
, type
, 0);
86 ret
= type_name(obj
->type
);
93 static int fsck_config(const char *var
, const char *value
, void *cb
)
95 if (strcmp(var
, "fsck.skiplist") == 0) {
97 struct strbuf sb
= STRBUF_INIT
;
99 if (git_config_pathname(&path
, var
, value
))
101 strbuf_addf(&sb
, "skiplist=%s", path
);
103 fsck_set_msg_types(&fsck_obj_options
, sb
.buf
);
108 if (skip_prefix(var
, "fsck.", &var
)) {
109 fsck_set_msg_type(&fsck_obj_options
, var
, value
);
113 return git_default_config(var
, value
, cb
);
116 static int objerror(struct object
*obj
, const char *err
)
118 errors_found
|= ERROR_OBJECT
;
119 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
120 fprintf_ln(stderr
, _("error in %s %s: %s"),
121 printable_type(obj
), describe_object(obj
), err
);
125 static int fsck_error_func(struct fsck_options
*o
,
126 struct object
*obj
, int type
, const char *message
)
130 /* TRANSLATORS: e.g. warning in tree 01bfda: <more explanation> */
131 fprintf_ln(stderr
, _("warning in %s %s: %s"),
132 printable_type(obj
), describe_object(obj
), message
);
135 /* TRANSLATORS: e.g. error in tree 01bfda: <more explanation> */
136 fprintf_ln(stderr
, _("error in %s %s: %s"),
137 printable_type(obj
), describe_object(obj
), message
);
140 BUG("%d (FSCK_IGNORE?) should never trigger this callback", type
);
144 static struct object_array pending
;
146 static int mark_object(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
148 struct object
*parent
= data
;
151 * The only case data is NULL or type is OBJ_ANY is when
152 * mark_object_reachable() calls us. All the callers of
153 * that function has non-NULL obj hence ...
156 /* ... these references to parent->fld are safe here */
157 printf_ln(_("broken link from %7s %s"),
158 printable_type(parent
), describe_object(parent
));
159 printf_ln(_("broken link from %7s %s"),
160 (type
== OBJ_ANY
? _("unknown") : type_name(type
)),
162 errors_found
|= ERROR_REACHABLE
;
166 if (type
!= OBJ_ANY
&& obj
->type
!= type
)
167 /* ... and the reference to parent is safe here */
168 objerror(parent
, _("wrong object type in link"));
170 if (obj
->flags
& REACHABLE
)
172 obj
->flags
|= REACHABLE
;
174 if (is_promisor_object(&obj
->oid
))
176 * Further recursion does not need to be performed on this
177 * object since it is a promisor object (so it does not need to
178 * be added to "pending").
182 if (!(obj
->flags
& HAS_OBJ
)) {
183 if (parent
&& !has_object_file(&obj
->oid
)) {
184 printf_ln(_("broken link from %7s %s\n"
186 printable_type(parent
),
187 describe_object(parent
),
189 describe_object(obj
));
190 errors_found
|= ERROR_REACHABLE
;
195 add_object_array(obj
, NULL
, &pending
);
199 static void mark_object_reachable(struct object
*obj
)
201 mark_object(obj
, OBJ_ANY
, NULL
, NULL
);
204 static int traverse_one_object(struct object
*obj
)
206 int result
= fsck_walk(obj
, obj
, &fsck_walk_options
);
208 if (obj
->type
== OBJ_TREE
) {
209 struct tree
*tree
= (struct tree
*)obj
;
210 free_tree_buffer(tree
);
215 static int traverse_reachable(void)
217 struct progress
*progress
= NULL
;
221 progress
= start_delayed_progress(_("Checking connectivity"), 0);
223 result
|= traverse_one_object(object_array_pop(&pending
));
224 display_progress(progress
, ++nr
);
226 stop_progress(&progress
);
230 static int mark_used(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
239 * Check a single reachable object
241 static void check_reachable_object(struct object
*obj
)
244 * We obviously want the object to be parsed,
245 * except if it was in a pack-file and we didn't
248 if (!(obj
->flags
& HAS_OBJ
)) {
249 if (is_promisor_object(&obj
->oid
))
251 if (has_object_pack(&obj
->oid
))
252 return; /* it is in pack - forget about it */
253 printf_ln(_("missing %s %s"), printable_type(obj
),
254 describe_object(obj
));
255 errors_found
|= ERROR_REACHABLE
;
261 * Check a single unreachable object
263 static void check_unreachable_object(struct object
*obj
)
266 * Missing unreachable object? Ignore it. It's not like
267 * we miss it (since it can't be reached), nor do we want
268 * to complain about it being unreachable (since it does
271 if (!(obj
->flags
& HAS_OBJ
))
275 * Unreachable object that exists? Show it if asked to,
276 * since this is something that is prunable.
278 if (show_unreachable
) {
279 printf_ln(_("unreachable %s %s"), printable_type(obj
),
280 describe_object(obj
));
285 * "!USED" means that nothing at all points to it, including
286 * other unreachable objects. In other words, it's the "tip"
287 * of some set of unreachable objects, usually a commit that
290 * Such starting points are more interesting than some random
291 * set of unreachable objects, so we show them even if the user
292 * hasn't asked for _all_ unreachable objects. If you have
293 * deleted a branch by mistake, this is a prime candidate to
294 * start looking at, for example.
296 if (!(obj
->flags
& USED
)) {
298 printf_ln(_("dangling %s %s"), printable_type(obj
),
299 describe_object(obj
));
300 if (write_lost_and_found
) {
301 char *filename
= git_pathdup("lost-found/%s/%s",
302 obj
->type
== OBJ_COMMIT
? "commit" : "other",
303 describe_object(obj
));
306 if (safe_create_leading_directories_const(filename
)) {
307 error(_("could not create lost-found"));
311 f
= xfopen(filename
, "w");
312 if (obj
->type
== OBJ_BLOB
) {
313 if (stream_blob_to_fd(fileno(f
), &obj
->oid
, NULL
, 1))
314 die_errno(_("could not write '%s'"), filename
);
316 fprintf(f
, "%s\n", describe_object(obj
));
318 die_errno(_("could not finish '%s'"),
326 * Otherwise? It's there, it's unreachable, and some other unreachable
327 * object points to it. Ignore it - it's not interesting, and we showed
328 * all the interesting cases above.
332 static void check_object(struct object
*obj
)
335 fprintf_ln(stderr
, _("Checking %s"), describe_object(obj
));
337 if (obj
->flags
& REACHABLE
)
338 check_reachable_object(obj
);
340 check_unreachable_object(obj
);
343 static void check_connectivity(void)
347 /* Traverse the pending reachable objects */
348 traverse_reachable();
350 /* Look up all the requirements, warn about missing objects.. */
351 max
= get_max_object_index();
353 fprintf_ln(stderr
, _("Checking connectivity (%d objects)"), max
);
355 for (i
= 0; i
< max
; i
++) {
356 struct object
*obj
= get_indexed_object(i
);
363 static int fsck_obj(struct object
*obj
, void *buffer
, unsigned long size
)
367 if (obj
->flags
& SEEN
)
372 fprintf_ln(stderr
, _("Checking %s %s"),
373 printable_type(obj
), describe_object(obj
));
375 if (fsck_walk(obj
, NULL
, &fsck_obj_options
))
376 objerror(obj
, _("broken links"));
377 err
= fsck_object(obj
, buffer
, size
, &fsck_obj_options
);
381 if (obj
->type
== OBJ_COMMIT
) {
382 struct commit
*commit
= (struct commit
*) obj
;
384 if (!commit
->parents
&& show_root
)
385 printf_ln(_("root %s"),
386 describe_object(&commit
->object
));
389 if (obj
->type
== OBJ_TAG
) {
390 struct tag
*tag
= (struct tag
*) obj
;
392 if (show_tags
&& tag
->tagged
) {
393 printf_ln(_("tagged %s %s (%s) in %s"),
394 printable_type(tag
->tagged
),
395 describe_object(tag
->tagged
),
397 describe_object(&tag
->object
));
402 if (obj
->type
== OBJ_TREE
)
403 free_tree_buffer((struct tree
*)obj
);
404 if (obj
->type
== OBJ_COMMIT
)
405 free_commit_buffer((struct commit
*)obj
);
409 static int fsck_obj_buffer(const struct object_id
*oid
, enum object_type type
,
410 unsigned long size
, void *buffer
, int *eaten
)
413 * Note, buffer may be NULL if type is OBJ_BLOB. See
414 * verify_packfile(), data_valid variable for details.
417 obj
= parse_object_buffer(the_repository
, oid
, type
, size
, buffer
,
420 errors_found
|= ERROR_OBJECT
;
421 return error(_("%s: object corrupt or missing"),
424 obj
->flags
&= ~(REACHABLE
| SEEN
);
425 obj
->flags
|= HAS_OBJ
;
426 return fsck_obj(obj
, buffer
, size
);
429 static int default_refs
;
431 static void fsck_handle_reflog_oid(const char *refname
, struct object_id
*oid
,
432 timestamp_t timestamp
)
436 if (!is_null_oid(oid
)) {
437 obj
= lookup_object(the_repository
, oid
->hash
);
438 if (obj
&& (obj
->flags
& HAS_OBJ
)) {
439 if (timestamp
&& name_objects
)
440 add_decoration(fsck_walk_options
.object_names
,
442 xstrfmt("%s@{%"PRItime
"}", refname
, timestamp
));
444 mark_object_reachable(obj
);
445 } else if (!is_promisor_object(oid
)) {
446 error(_("%s: invalid reflog entry %s"),
447 refname
, oid_to_hex(oid
));
448 errors_found
|= ERROR_REACHABLE
;
453 static int fsck_handle_reflog_ent(struct object_id
*ooid
, struct object_id
*noid
,
454 const char *email
, timestamp_t timestamp
, int tz
,
455 const char *message
, void *cb_data
)
457 const char *refname
= cb_data
;
460 fprintf_ln(stderr
, _("Checking reflog %s->%s"),
461 oid_to_hex(ooid
), oid_to_hex(noid
));
463 fsck_handle_reflog_oid(refname
, ooid
, 0);
464 fsck_handle_reflog_oid(refname
, noid
, timestamp
);
468 static int fsck_handle_reflog(const char *logname
, const struct object_id
*oid
,
469 int flag
, void *cb_data
)
471 for_each_reflog_ent(logname
, fsck_handle_reflog_ent
, (void *)logname
);
475 static int fsck_handle_ref(const char *refname
, const struct object_id
*oid
,
476 int flag
, void *cb_data
)
480 obj
= parse_object(the_repository
, oid
);
482 if (is_promisor_object(oid
)) {
484 * Increment default_refs anyway, because this is a
490 error(_("%s: invalid sha1 pointer %s"),
491 refname
, oid_to_hex(oid
));
492 errors_found
|= ERROR_REACHABLE
;
493 /* We'll continue with the rest despite the error.. */
496 if (obj
->type
!= OBJ_COMMIT
&& is_branch(refname
)) {
497 error(_("%s: not a commit"), refname
);
498 errors_found
|= ERROR_REFS
;
503 add_decoration(fsck_walk_options
.object_names
,
504 obj
, xstrdup(refname
));
505 mark_object_reachable(obj
);
510 static void get_default_heads(void)
512 if (head_points_at
&& !is_null_oid(&head_oid
))
513 fsck_handle_ref("HEAD", &head_oid
, 0, NULL
);
514 for_each_rawref(fsck_handle_ref
, NULL
);
516 for_each_reflog(fsck_handle_reflog
, NULL
);
519 * Not having any default heads isn't really fatal, but
520 * it does mean that "--unreachable" no longer makes any
521 * sense (since in this case everything will obviously
522 * be unreachable by definition.
524 * Showing dangling objects is valid, though (as those
525 * dangling objects are likely lost heads).
527 * So we just print a warning about it, and clear the
528 * "show_unreachable" flag.
531 fprintf_ln(stderr
, _("notice: No default references"));
532 show_unreachable
= 0;
536 static int fsck_loose(const struct object_id
*oid
, const char *path
, void *data
)
539 enum object_type type
;
544 if (read_loose_object(path
, oid
, &type
, &size
, &contents
) < 0) {
545 errors_found
|= ERROR_OBJECT
;
546 error(_("%s: object corrupt or missing: %s"),
547 oid_to_hex(oid
), path
);
548 return 0; /* keep checking other objects */
551 if (!contents
&& type
!= OBJ_BLOB
)
552 BUG("read_loose_object streamed a non-blob");
554 obj
= parse_object_buffer(the_repository
, oid
, type
, size
,
558 errors_found
|= ERROR_OBJECT
;
559 error(_("%s: object could not be parsed: %s"),
560 oid_to_hex(oid
), path
);
563 return 0; /* keep checking other objects */
566 obj
->flags
&= ~(REACHABLE
| SEEN
);
567 obj
->flags
|= HAS_OBJ
;
568 if (fsck_obj(obj
, contents
, size
))
569 errors_found
|= ERROR_OBJECT
;
573 return 0; /* keep checking other objects, even if we saw an error */
576 static int fsck_cruft(const char *basename
, const char *path
, void *data
)
578 if (!starts_with(basename
, "tmp_obj_"))
579 fprintf_ln(stderr
, _("bad sha1 file: %s"), path
);
583 static int fsck_subdir(unsigned int nr
, const char *path
, void *progress
)
585 display_progress(progress
, nr
+ 1);
589 static void fsck_object_dir(const char *path
)
591 struct progress
*progress
= NULL
;
594 fprintf_ln(stderr
, _("Checking object directory"));
597 progress
= start_progress(_("Checking object directories"), 256);
599 for_each_loose_file_in_objdir(path
, fsck_loose
, fsck_cruft
, fsck_subdir
,
601 display_progress(progress
, 256);
602 stop_progress(&progress
);
605 static int fsck_head_link(void)
607 int null_is_error
= 0;
610 fprintf_ln(stderr
, _("Checking HEAD link"));
612 head_points_at
= resolve_ref_unsafe("HEAD", 0, &head_oid
, NULL
);
613 if (!head_points_at
) {
614 errors_found
|= ERROR_REFS
;
615 return error(_("invalid HEAD"));
617 if (!strcmp(head_points_at
, "HEAD"))
620 else if (!starts_with(head_points_at
, "refs/heads/")) {
621 errors_found
|= ERROR_REFS
;
622 return error(_("HEAD points to something strange (%s)"),
625 if (is_null_oid(&head_oid
)) {
627 errors_found
|= ERROR_REFS
;
628 return error(_("HEAD: detached HEAD points at nothing"));
630 fprintf_ln(stderr
, _("notice: HEAD points to an unborn branch (%s)"),
631 head_points_at
+ 11);
636 static int fsck_cache_tree(struct cache_tree
*it
)
642 fprintf_ln(stderr
, _("Checking cache tree"));
644 if (0 <= it
->entry_count
) {
645 struct object
*obj
= parse_object(the_repository
, &it
->oid
);
647 error(_("%s: invalid sha1 pointer in cache-tree"),
648 oid_to_hex(&it
->oid
));
649 errors_found
|= ERROR_REFS
;
654 add_decoration(fsck_walk_options
.object_names
,
656 mark_object_reachable(obj
);
657 if (obj
->type
!= OBJ_TREE
)
658 err
|= objerror(obj
, _("non-tree in cache-tree"));
660 for (i
= 0; i
< it
->subtree_nr
; i
++)
661 err
|= fsck_cache_tree(it
->down
[i
]->cache_tree
);
665 static void mark_object_for_connectivity(const struct object_id
*oid
)
667 struct object
*obj
= lookup_unknown_object(oid
->hash
);
668 obj
->flags
|= HAS_OBJ
;
671 static int mark_loose_for_connectivity(const struct object_id
*oid
,
675 mark_object_for_connectivity(oid
);
679 static int mark_packed_for_connectivity(const struct object_id
*oid
,
680 struct packed_git
*pack
,
684 mark_object_for_connectivity(oid
);
688 static char const * const fsck_usage
[] = {
689 N_("git fsck [<options>] [<object>...]"),
693 static struct option fsck_opts
[] = {
694 OPT__VERBOSE(&verbose
, N_("be verbose")),
695 OPT_BOOL(0, "unreachable", &show_unreachable
, N_("show unreachable objects")),
696 OPT_BOOL(0, "dangling", &show_dangling
, N_("show dangling objects")),
697 OPT_BOOL(0, "tags", &show_tags
, N_("report tags")),
698 OPT_BOOL(0, "root", &show_root
, N_("report root nodes")),
699 OPT_BOOL(0, "cache", &keep_cache_objects
, N_("make index objects head nodes")),
700 OPT_BOOL(0, "reflogs", &include_reflogs
, N_("make reflogs head nodes (default)")),
701 OPT_BOOL(0, "full", &check_full
, N_("also consider packs and alternate objects")),
702 OPT_BOOL(0, "connectivity-only", &connectivity_only
, N_("check only connectivity")),
703 OPT_BOOL(0, "strict", &check_strict
, N_("enable more strict checking")),
704 OPT_BOOL(0, "lost-found", &write_lost_and_found
,
705 N_("write dangling objects in .git/lost-found")),
706 OPT_BOOL(0, "progress", &show_progress
, N_("show progress")),
707 OPT_BOOL(0, "name-objects", &name_objects
, N_("show verbose names for reachable objects")),
711 int cmd_fsck(int argc
, const char **argv
, const char *prefix
)
714 struct alternate_object_database
*alt
;
716 /* fsck knows how to handle missing promisor objects */
717 fetch_if_missing
= 0;
720 read_replace_refs
= 0;
722 argc
= parse_options(argc
, argv
, prefix
, fsck_opts
, fsck_usage
, 0);
724 fsck_walk_options
.walk
= mark_object
;
725 fsck_obj_options
.walk
= mark_used
;
726 fsck_obj_options
.error_func
= fsck_error_func
;
728 fsck_obj_options
.strict
= 1;
730 if (show_progress
== -1)
731 show_progress
= isatty(2);
735 if (write_lost_and_found
) {
741 fsck_walk_options
.object_names
=
742 xcalloc(1, sizeof(struct decoration
));
744 git_config(fsck_config
, NULL
);
747 if (connectivity_only
) {
748 for_each_loose_object(mark_loose_for_connectivity
, NULL
, 0);
749 for_each_packed_object(mark_packed_for_connectivity
, NULL
, 0);
751 struct alternate_object_database
*alt_odb_list
;
753 fsck_object_dir(get_object_directory());
755 prepare_alt_odb(the_repository
);
756 alt_odb_list
= the_repository
->objects
->alt_odb_list
;
757 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
)
758 fsck_object_dir(alt
->path
);
761 struct packed_git
*p
;
762 uint32_t total
= 0, count
= 0;
763 struct progress
*progress
= NULL
;
766 for (p
= get_all_packs(the_repository
); p
;
768 if (open_pack_index(p
))
770 total
+= p
->num_objects
;
773 progress
= start_progress(_("Checking objects"), total
);
775 for (p
= get_all_packs(the_repository
); p
;
777 /* verify gives error messages itself */
778 if (verify_pack(p
, fsck_obj_buffer
,
780 errors_found
|= ERROR_PACK
;
781 count
+= p
->num_objects
;
783 stop_progress(&progress
);
786 if (fsck_finish(&fsck_obj_options
))
787 errors_found
|= ERROR_OBJECT
;
790 for (i
= 0; i
< argc
; i
++) {
791 const char *arg
= argv
[i
];
792 struct object_id oid
;
793 if (!get_oid(arg
, &oid
)) {
794 struct object
*obj
= lookup_object(the_repository
,
797 if (!obj
|| !(obj
->flags
& HAS_OBJ
)) {
798 if (is_promisor_object(&oid
))
800 error(_("%s: object missing"), oid_to_hex(&oid
));
801 errors_found
|= ERROR_OBJECT
;
807 add_decoration(fsck_walk_options
.object_names
,
809 mark_object_reachable(obj
);
812 error(_("invalid parameter: expected sha1, got '%s'"), arg
);
813 errors_found
|= ERROR_OBJECT
;
817 * If we've not been given any explicit head information, do the
818 * default ones from .git/refs. We also consider the index file
819 * in this case (ie this implies --cache).
823 keep_cache_objects
= 1;
826 if (keep_cache_objects
) {
827 verify_index_checksum
= 1;
830 for (i
= 0; i
< active_nr
; i
++) {
835 mode
= active_cache
[i
]->ce_mode
;
836 if (S_ISGITLINK(mode
))
838 blob
= lookup_blob(the_repository
,
839 &active_cache
[i
]->oid
);
845 add_decoration(fsck_walk_options
.object_names
,
847 xstrfmt(":%s", active_cache
[i
]->name
));
848 mark_object_reachable(obj
);
850 if (active_cache_tree
)
851 fsck_cache_tree(active_cache_tree
);
854 check_connectivity();
856 if (!git_config_get_bool("core.commitgraph", &i
) && i
) {
857 struct child_process commit_graph_verify
= CHILD_PROCESS_INIT
;
858 const char *verify_argv
[] = { "commit-graph", "verify", NULL
, NULL
, NULL
};
860 commit_graph_verify
.argv
= verify_argv
;
861 commit_graph_verify
.git_cmd
= 1;
862 if (run_command(&commit_graph_verify
))
863 errors_found
|= ERROR_COMMIT_GRAPH
;
865 prepare_alt_odb(the_repository
);
866 for (alt
= the_repository
->objects
->alt_odb_list
; alt
; alt
= alt
->next
) {
867 verify_argv
[2] = "--object-dir";
868 verify_argv
[3] = alt
->path
;
869 if (run_command(&commit_graph_verify
))
870 errors_found
|= ERROR_COMMIT_GRAPH
;
874 if (!git_config_get_bool("core.multipackindex", &i
) && i
) {
875 struct child_process midx_verify
= CHILD_PROCESS_INIT
;
876 const char *midx_argv
[] = { "multi-pack-index", "verify", NULL
, NULL
, NULL
};
878 midx_verify
.argv
= midx_argv
;
879 midx_verify
.git_cmd
= 1;
880 if (run_command(&midx_verify
))
881 errors_found
|= ERROR_COMMIT_GRAPH
;
883 prepare_alt_odb(the_repository
);
884 for (alt
= the_repository
->objects
->alt_odb_list
; alt
; alt
= alt
->next
) {
885 midx_argv
[2] = "--object-dir";
886 midx_argv
[3] = alt
->path
;
887 if (run_command(&midx_verify
))
888 errors_found
|= ERROR_COMMIT_GRAPH
;