10 #include "cache-tree.h"
11 #include "tree-walk.h"
13 #include "parse-options.h"
16 #include "streaming.h"
20 #define REACHABLE 0x0001
22 #define HAS_OBJ 0x0004
23 /* This flag is set if something points to this object. */
28 static int show_unreachable
;
29 static int include_reflogs
= 1;
30 static int check_full
= 1;
31 static int connectivity_only
;
32 static int check_strict
;
33 static int keep_cache_objects
;
34 static struct fsck_options fsck_walk_options
= FSCK_OPTIONS_DEFAULT
;
35 static struct fsck_options fsck_obj_options
= FSCK_OPTIONS_DEFAULT
;
36 static struct object_id head_oid
;
37 static const char *head_points_at
;
38 static int errors_found
;
39 static int write_lost_and_found
;
41 static int show_progress
= -1;
42 static int show_dangling
= 1;
43 static int name_objects
;
44 #define ERROR_OBJECT 01
45 #define ERROR_REACHABLE 02
47 #define ERROR_REFS 010
49 static const char *describe_object(struct object
*obj
)
51 static struct strbuf buf
= STRBUF_INIT
;
52 char *name
= name_objects
?
53 lookup_decoration(fsck_walk_options
.object_names
, obj
) : NULL
;
56 strbuf_addstr(&buf
, oid_to_hex(&obj
->oid
));
58 strbuf_addf(&buf
, " (%s)", name
);
63 static const char *printable_type(struct object
*obj
)
67 if (obj
->type
== OBJ_NONE
) {
68 enum object_type type
= sha1_object_info(obj
->oid
.hash
, NULL
);
70 object_as_type(obj
, type
, 0);
73 ret
= type_name(obj
->type
);
80 static int fsck_config(const char *var
, const char *value
, void *cb
)
82 if (strcmp(var
, "fsck.skiplist") == 0) {
84 struct strbuf sb
= STRBUF_INIT
;
86 if (git_config_pathname(&path
, var
, value
))
88 strbuf_addf(&sb
, "skiplist=%s", path
);
90 fsck_set_msg_types(&fsck_obj_options
, sb
.buf
);
95 if (skip_prefix(var
, "fsck.", &var
)) {
96 fsck_set_msg_type(&fsck_obj_options
, var
, value
);
100 return git_default_config(var
, value
, cb
);
103 static void objreport(struct object
*obj
, const char *msg_type
,
106 fprintf(stderr
, "%s in %s %s: %s\n",
107 msg_type
, printable_type(obj
), describe_object(obj
), err
);
110 static int objerror(struct object
*obj
, const char *err
)
112 errors_found
|= ERROR_OBJECT
;
113 objreport(obj
, "error", err
);
117 static int fsck_error_func(struct fsck_options
*o
,
118 struct object
*obj
, int type
, const char *message
)
120 objreport(obj
, (type
== FSCK_WARN
) ? "warning" : "error", message
);
121 return (type
== FSCK_WARN
) ? 0 : 1;
124 static struct object_array pending
;
126 static int mark_object(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
128 struct object
*parent
= data
;
131 * The only case data is NULL or type is OBJ_ANY is when
132 * mark_object_reachable() calls us. All the callers of
133 * that function has non-NULL obj hence ...
136 /* ... these references to parent->fld are safe here */
137 printf("broken link from %7s %s\n",
138 printable_type(parent
), describe_object(parent
));
139 printf("broken link from %7s %s\n",
140 (type
== OBJ_ANY
? "unknown" : type_name(type
)), "unknown");
141 errors_found
|= ERROR_REACHABLE
;
145 if (type
!= OBJ_ANY
&& obj
->type
!= type
)
146 /* ... and the reference to parent is safe here */
147 objerror(parent
, "wrong object type in link");
149 if (obj
->flags
& REACHABLE
)
151 obj
->flags
|= REACHABLE
;
153 if (is_promisor_object(&obj
->oid
))
155 * Further recursion does not need to be performed on this
156 * object since it is a promisor object (so it does not need to
157 * be added to "pending").
161 if (!(obj
->flags
& HAS_OBJ
)) {
162 if (parent
&& !has_object_file(&obj
->oid
)) {
163 printf("broken link from %7s %s\n",
164 printable_type(parent
), describe_object(parent
));
165 printf(" to %7s %s\n",
166 printable_type(obj
), describe_object(obj
));
167 errors_found
|= ERROR_REACHABLE
;
172 add_object_array(obj
, NULL
, &pending
);
176 static void mark_object_reachable(struct object
*obj
)
178 mark_object(obj
, OBJ_ANY
, NULL
, NULL
);
181 static int traverse_one_object(struct object
*obj
)
183 int result
= fsck_walk(obj
, obj
, &fsck_walk_options
);
185 if (obj
->type
== OBJ_TREE
) {
186 struct tree
*tree
= (struct tree
*)obj
;
187 free_tree_buffer(tree
);
192 static int traverse_reachable(void)
194 struct progress
*progress
= NULL
;
198 progress
= start_delayed_progress(_("Checking connectivity"), 0);
200 result
|= traverse_one_object(object_array_pop(&pending
));
201 display_progress(progress
, ++nr
);
203 stop_progress(&progress
);
207 static int mark_used(struct object
*obj
, int type
, void *data
, struct fsck_options
*options
)
216 * Check a single reachable object
218 static void check_reachable_object(struct object
*obj
)
221 * We obviously want the object to be parsed,
222 * except if it was in a pack-file and we didn't
225 if (!(obj
->flags
& HAS_OBJ
)) {
226 if (is_promisor_object(&obj
->oid
))
228 if (has_sha1_pack(obj
->oid
.hash
))
229 return; /* it is in pack - forget about it */
230 printf("missing %s %s\n", printable_type(obj
),
231 describe_object(obj
));
232 errors_found
|= ERROR_REACHABLE
;
238 * Check a single unreachable object
240 static void check_unreachable_object(struct object
*obj
)
243 * Missing unreachable object? Ignore it. It's not like
244 * we miss it (since it can't be reached), nor do we want
245 * to complain about it being unreachable (since it does
248 if (!(obj
->flags
& HAS_OBJ
))
252 * Unreachable object that exists? Show it if asked to,
253 * since this is something that is prunable.
255 if (show_unreachable
) {
256 printf("unreachable %s %s\n", printable_type(obj
),
257 describe_object(obj
));
262 * "!USED" means that nothing at all points to it, including
263 * other unreachable objects. In other words, it's the "tip"
264 * of some set of unreachable objects, usually a commit that
267 * Such starting points are more interesting than some random
268 * set of unreachable objects, so we show them even if the user
269 * hasn't asked for _all_ unreachable objects. If you have
270 * deleted a branch by mistake, this is a prime candidate to
271 * start looking at, for example.
273 if (!(obj
->flags
& USED
)) {
275 printf("dangling %s %s\n", printable_type(obj
),
276 describe_object(obj
));
277 if (write_lost_and_found
) {
278 char *filename
= git_pathdup("lost-found/%s/%s",
279 obj
->type
== OBJ_COMMIT
? "commit" : "other",
280 describe_object(obj
));
283 if (safe_create_leading_directories_const(filename
)) {
284 error("Could not create lost-found");
288 f
= xfopen(filename
, "w");
289 if (obj
->type
== OBJ_BLOB
) {
290 if (stream_blob_to_fd(fileno(f
), &obj
->oid
, NULL
, 1))
291 die_errno("Could not write '%s'", filename
);
293 fprintf(f
, "%s\n", describe_object(obj
));
295 die_errno("Could not finish '%s'",
303 * Otherwise? It's there, it's unreachable, and some other unreachable
304 * object points to it. Ignore it - it's not interesting, and we showed
305 * all the interesting cases above.
309 static void check_object(struct object
*obj
)
312 fprintf(stderr
, "Checking %s\n", describe_object(obj
));
314 if (obj
->flags
& REACHABLE
)
315 check_reachable_object(obj
);
317 check_unreachable_object(obj
);
320 static void check_connectivity(void)
324 /* Traverse the pending reachable objects */
325 traverse_reachable();
327 /* Look up all the requirements, warn about missing objects.. */
328 max
= get_max_object_index();
330 fprintf(stderr
, "Checking connectivity (%d objects)\n", max
);
332 for (i
= 0; i
< max
; i
++) {
333 struct object
*obj
= get_indexed_object(i
);
340 static int fsck_obj(struct object
*obj
, void *buffer
, unsigned long size
)
344 if (obj
->flags
& SEEN
)
349 fprintf(stderr
, "Checking %s %s\n",
350 printable_type(obj
), describe_object(obj
));
352 if (fsck_walk(obj
, NULL
, &fsck_obj_options
))
353 objerror(obj
, "broken links");
354 err
= fsck_object(obj
, buffer
, size
, &fsck_obj_options
);
358 if (obj
->type
== OBJ_COMMIT
) {
359 struct commit
*commit
= (struct commit
*) obj
;
361 if (!commit
->parents
&& show_root
)
362 printf("root %s\n", describe_object(&commit
->object
));
365 if (obj
->type
== OBJ_TAG
) {
366 struct tag
*tag
= (struct tag
*) obj
;
368 if (show_tags
&& tag
->tagged
) {
369 printf("tagged %s %s", printable_type(tag
->tagged
),
370 describe_object(tag
->tagged
));
371 printf(" (%s) in %s\n", tag
->tag
,
372 describe_object(&tag
->object
));
377 if (obj
->type
== OBJ_TREE
)
378 free_tree_buffer((struct tree
*)obj
);
379 if (obj
->type
== OBJ_COMMIT
)
380 free_commit_buffer((struct commit
*)obj
);
384 static int fsck_obj_buffer(const struct object_id
*oid
, enum object_type type
,
385 unsigned long size
, void *buffer
, int *eaten
)
388 * Note, buffer may be NULL if type is OBJ_BLOB. See
389 * verify_packfile(), data_valid variable for details.
392 obj
= parse_object_buffer(oid
, type
, size
, buffer
, eaten
);
394 errors_found
|= ERROR_OBJECT
;
395 return error("%s: object corrupt or missing", oid_to_hex(oid
));
397 obj
->flags
&= ~(REACHABLE
| SEEN
);
398 obj
->flags
|= HAS_OBJ
;
399 return fsck_obj(obj
, buffer
, size
);
402 static int default_refs
;
404 static void fsck_handle_reflog_oid(const char *refname
, struct object_id
*oid
,
405 timestamp_t timestamp
)
409 if (!is_null_oid(oid
)) {
410 obj
= lookup_object(oid
->hash
);
411 if (obj
&& (obj
->flags
& HAS_OBJ
)) {
412 if (timestamp
&& name_objects
)
413 add_decoration(fsck_walk_options
.object_names
,
415 xstrfmt("%s@{%"PRItime
"}", refname
, timestamp
));
417 mark_object_reachable(obj
);
418 } else if (!is_promisor_object(oid
)) {
419 error("%s: invalid reflog entry %s", refname
, oid_to_hex(oid
));
420 errors_found
|= ERROR_REACHABLE
;
425 static int fsck_handle_reflog_ent(struct object_id
*ooid
, struct object_id
*noid
,
426 const char *email
, timestamp_t timestamp
, int tz
,
427 const char *message
, void *cb_data
)
429 const char *refname
= cb_data
;
432 fprintf(stderr
, "Checking reflog %s->%s\n",
433 oid_to_hex(ooid
), oid_to_hex(noid
));
435 fsck_handle_reflog_oid(refname
, ooid
, 0);
436 fsck_handle_reflog_oid(refname
, noid
, timestamp
);
440 static int fsck_handle_reflog(const char *logname
, const struct object_id
*oid
,
441 int flag
, void *cb_data
)
443 for_each_reflog_ent(logname
, fsck_handle_reflog_ent
, (void *)logname
);
447 static int fsck_handle_ref(const char *refname
, const struct object_id
*oid
,
448 int flag
, void *cb_data
)
452 obj
= parse_object(oid
);
454 if (is_promisor_object(oid
)) {
456 * Increment default_refs anyway, because this is a
462 error("%s: invalid sha1 pointer %s", refname
, oid_to_hex(oid
));
463 errors_found
|= ERROR_REACHABLE
;
464 /* We'll continue with the rest despite the error.. */
467 if (obj
->type
!= OBJ_COMMIT
&& is_branch(refname
)) {
468 error("%s: not a commit", refname
);
469 errors_found
|= ERROR_REFS
;
474 add_decoration(fsck_walk_options
.object_names
,
475 obj
, xstrdup(refname
));
476 mark_object_reachable(obj
);
481 static void get_default_heads(void)
483 if (head_points_at
&& !is_null_oid(&head_oid
))
484 fsck_handle_ref("HEAD", &head_oid
, 0, NULL
);
485 for_each_rawref(fsck_handle_ref
, NULL
);
487 for_each_reflog(fsck_handle_reflog
, NULL
);
490 * Not having any default heads isn't really fatal, but
491 * it does mean that "--unreachable" no longer makes any
492 * sense (since in this case everything will obviously
493 * be unreachable by definition.
495 * Showing dangling objects is valid, though (as those
496 * dangling objects are likely lost heads).
498 * So we just print a warning about it, and clear the
499 * "show_unreachable" flag.
502 fprintf(stderr
, "notice: No default references\n");
503 show_unreachable
= 0;
507 static int fsck_loose(const struct object_id
*oid
, const char *path
, void *data
)
510 enum object_type type
;
515 if (read_loose_object(path
, oid
->hash
, &type
, &size
, &contents
) < 0) {
516 errors_found
|= ERROR_OBJECT
;
517 error("%s: object corrupt or missing: %s",
518 oid_to_hex(oid
), path
);
519 return 0; /* keep checking other objects */
522 if (!contents
&& type
!= OBJ_BLOB
)
523 BUG("read_loose_object streamed a non-blob");
525 obj
= parse_object_buffer(oid
, type
, size
, contents
, &eaten
);
527 errors_found
|= ERROR_OBJECT
;
528 error("%s: object could not be parsed: %s",
529 oid_to_hex(oid
), path
);
532 return 0; /* keep checking other objects */
535 obj
->flags
&= ~(REACHABLE
| SEEN
);
536 obj
->flags
|= HAS_OBJ
;
537 if (fsck_obj(obj
, contents
, size
))
538 errors_found
|= ERROR_OBJECT
;
542 return 0; /* keep checking other objects, even if we saw an error */
545 static int fsck_cruft(const char *basename
, const char *path
, void *data
)
547 if (!starts_with(basename
, "tmp_obj_"))
548 fprintf(stderr
, "bad sha1 file: %s\n", path
);
552 static int fsck_subdir(unsigned int nr
, const char *path
, void *progress
)
554 display_progress(progress
, nr
+ 1);
558 static void fsck_object_dir(const char *path
)
560 struct progress
*progress
= NULL
;
563 fprintf(stderr
, "Checking object directory\n");
566 progress
= start_progress(_("Checking object directories"), 256);
568 for_each_loose_file_in_objdir(path
, fsck_loose
, fsck_cruft
, fsck_subdir
,
570 display_progress(progress
, 256);
571 stop_progress(&progress
);
574 static int fsck_head_link(void)
576 int null_is_error
= 0;
579 fprintf(stderr
, "Checking HEAD link\n");
581 head_points_at
= resolve_ref_unsafe("HEAD", 0, &head_oid
, NULL
);
582 if (!head_points_at
) {
583 errors_found
|= ERROR_REFS
;
584 return error("Invalid HEAD");
586 if (!strcmp(head_points_at
, "HEAD"))
589 else if (!starts_with(head_points_at
, "refs/heads/")) {
590 errors_found
|= ERROR_REFS
;
591 return error("HEAD points to something strange (%s)",
594 if (is_null_oid(&head_oid
)) {
596 errors_found
|= ERROR_REFS
;
597 return error("HEAD: detached HEAD points at nothing");
599 fprintf(stderr
, "notice: HEAD points to an unborn branch (%s)\n",
600 head_points_at
+ 11);
605 static int fsck_cache_tree(struct cache_tree
*it
)
611 fprintf(stderr
, "Checking cache tree\n");
613 if (0 <= it
->entry_count
) {
614 struct object
*obj
= parse_object(&it
->oid
);
616 error("%s: invalid sha1 pointer in cache-tree",
617 oid_to_hex(&it
->oid
));
618 errors_found
|= ERROR_REFS
;
623 add_decoration(fsck_walk_options
.object_names
,
625 mark_object_reachable(obj
);
626 if (obj
->type
!= OBJ_TREE
)
627 err
|= objerror(obj
, "non-tree in cache-tree");
629 for (i
= 0; i
< it
->subtree_nr
; i
++)
630 err
|= fsck_cache_tree(it
->down
[i
]->cache_tree
);
634 static void mark_object_for_connectivity(const struct object_id
*oid
)
636 struct object
*obj
= lookup_unknown_object(oid
->hash
);
637 obj
->flags
|= HAS_OBJ
;
640 static int mark_loose_for_connectivity(const struct object_id
*oid
,
644 mark_object_for_connectivity(oid
);
648 static int mark_packed_for_connectivity(const struct object_id
*oid
,
649 struct packed_git
*pack
,
653 mark_object_for_connectivity(oid
);
657 static char const * const fsck_usage
[] = {
658 N_("git fsck [<options>] [<object>...]"),
662 static struct option fsck_opts
[] = {
663 OPT__VERBOSE(&verbose
, N_("be verbose")),
664 OPT_BOOL(0, "unreachable", &show_unreachable
, N_("show unreachable objects")),
665 OPT_BOOL(0, "dangling", &show_dangling
, N_("show dangling objects")),
666 OPT_BOOL(0, "tags", &show_tags
, N_("report tags")),
667 OPT_BOOL(0, "root", &show_root
, N_("report root nodes")),
668 OPT_BOOL(0, "cache", &keep_cache_objects
, N_("make index objects head nodes")),
669 OPT_BOOL(0, "reflogs", &include_reflogs
, N_("make reflogs head nodes (default)")),
670 OPT_BOOL(0, "full", &check_full
, N_("also consider packs and alternate objects")),
671 OPT_BOOL(0, "connectivity-only", &connectivity_only
, N_("check only connectivity")),
672 OPT_BOOL(0, "strict", &check_strict
, N_("enable more strict checking")),
673 OPT_BOOL(0, "lost-found", &write_lost_and_found
,
674 N_("write dangling objects in .git/lost-found")),
675 OPT_BOOL(0, "progress", &show_progress
, N_("show progress")),
676 OPT_BOOL(0, "name-objects", &name_objects
, N_("show verbose names for reachable objects")),
680 int cmd_fsck(int argc
, const char **argv
, const char *prefix
)
683 struct alternate_object_database
*alt
;
685 /* fsck knows how to handle missing promisor objects */
686 fetch_if_missing
= 0;
689 check_replace_refs
= 0;
691 argc
= parse_options(argc
, argv
, prefix
, fsck_opts
, fsck_usage
, 0);
693 fsck_walk_options
.walk
= mark_object
;
694 fsck_obj_options
.walk
= mark_used
;
695 fsck_obj_options
.error_func
= fsck_error_func
;
697 fsck_obj_options
.strict
= 1;
699 if (show_progress
== -1)
700 show_progress
= isatty(2);
704 if (write_lost_and_found
) {
710 fsck_walk_options
.object_names
=
711 xcalloc(1, sizeof(struct decoration
));
713 git_config(fsck_config
, NULL
);
716 if (connectivity_only
) {
717 for_each_loose_object(mark_loose_for_connectivity
, NULL
, 0);
718 for_each_packed_object(mark_packed_for_connectivity
, NULL
, 0);
720 fsck_object_dir(get_object_directory());
723 for (alt
= alt_odb_list
; alt
; alt
= alt
->next
)
724 fsck_object_dir(alt
->path
);
727 struct packed_git
*p
;
728 uint32_t total
= 0, count
= 0;
729 struct progress
*progress
= NULL
;
731 prepare_packed_git();
734 for (p
= packed_git
; p
; p
= p
->next
) {
735 if (open_pack_index(p
))
737 total
+= p
->num_objects
;
740 progress
= start_progress(_("Checking objects"), total
);
742 for (p
= packed_git
; p
; p
= p
->next
) {
743 /* verify gives error messages itself */
744 if (verify_pack(p
, fsck_obj_buffer
,
746 errors_found
|= ERROR_PACK
;
747 count
+= p
->num_objects
;
749 stop_progress(&progress
);
752 if (fsck_finish(&fsck_obj_options
))
753 errors_found
|= ERROR_OBJECT
;
756 for (i
= 0; i
< argc
; i
++) {
757 const char *arg
= argv
[i
];
758 struct object_id oid
;
759 if (!get_oid(arg
, &oid
)) {
760 struct object
*obj
= lookup_object(oid
.hash
);
762 if (!obj
|| !(obj
->flags
& HAS_OBJ
)) {
763 if (is_promisor_object(&oid
))
765 error("%s: object missing", oid_to_hex(&oid
));
766 errors_found
|= ERROR_OBJECT
;
772 add_decoration(fsck_walk_options
.object_names
,
774 mark_object_reachable(obj
);
777 error("invalid parameter: expected sha1, got '%s'", arg
);
778 errors_found
|= ERROR_OBJECT
;
782 * If we've not been given any explicit head information, do the
783 * default ones from .git/refs. We also consider the index file
784 * in this case (ie this implies --cache).
788 keep_cache_objects
= 1;
791 if (keep_cache_objects
) {
792 verify_index_checksum
= 1;
795 for (i
= 0; i
< active_nr
; i
++) {
800 mode
= active_cache
[i
]->ce_mode
;
801 if (S_ISGITLINK(mode
))
803 blob
= lookup_blob(&active_cache
[i
]->oid
);
809 add_decoration(fsck_walk_options
.object_names
,
811 xstrfmt(":%s", active_cache
[i
]->name
));
812 mark_object_reachable(obj
);
814 if (active_cache_tree
)
815 fsck_cache_tree(active_cache_tree
);
818 check_connectivity();