1 #include "git-compat-util.h"
5 #include "object-store-ll.h"
7 #include "repository.h"
12 #include "tree-walk.h"
21 #include "submodule-config.h"
25 static ssize_t max_tree_entry_len
= 4096;
28 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
30 const char *id_string
;
31 const char *downcased
;
32 const char *camelcased
;
33 enum fsck_msg_type msg_type
;
34 } msg_id_info
[FSCK_MSG_MAX
+ 1] = {
35 FOREACH_FSCK_MSG_ID(MSG_ID
)
36 { NULL
, NULL
, NULL
, -1 }
41 static void prepare_msg_ids(void)
45 if (msg_id_info
[0].downcased
)
48 /* convert id_string to lower case, without underscores. */
49 for (i
= 0; i
< FSCK_MSG_MAX
; i
++) {
50 const char *p
= msg_id_info
[i
].id_string
;
52 char *q
= xmalloc(len
);
54 msg_id_info
[i
].downcased
= q
;
59 *(q
)++ = tolower(*(p
)++);
62 p
= msg_id_info
[i
].id_string
;
64 msg_id_info
[i
].camelcased
= q
;
78 static int parse_msg_id(const char *text
)
84 for (i
= 0; i
< FSCK_MSG_MAX
; i
++)
85 if (!strcmp(text
, msg_id_info
[i
].downcased
))
91 void list_config_fsck_msg_ids(struct string_list
*list
, const char *prefix
)
97 for (i
= 0; i
< FSCK_MSG_MAX
; i
++)
98 list_config_item(list
, prefix
, msg_id_info
[i
].camelcased
);
101 static enum fsck_msg_type
fsck_msg_type(enum fsck_msg_id msg_id
,
102 struct fsck_options
*options
)
104 assert(msg_id
>= 0 && msg_id
< FSCK_MSG_MAX
);
106 if (!options
->msg_type
) {
107 enum fsck_msg_type msg_type
= msg_id_info
[msg_id
].msg_type
;
109 if (options
->strict
&& msg_type
== FSCK_WARN
)
110 msg_type
= FSCK_ERROR
;
114 return options
->msg_type
[msg_id
];
117 static enum fsck_msg_type
parse_msg_type(const char *str
)
119 if (!strcmp(str
, "error"))
121 else if (!strcmp(str
, "warn"))
123 else if (!strcmp(str
, "ignore"))
126 die("Unknown fsck message type: '%s'", str
);
129 int is_valid_msg_type(const char *msg_id
, const char *msg_type
)
131 if (parse_msg_id(msg_id
) < 0)
133 parse_msg_type(msg_type
);
137 void fsck_set_msg_type_from_ids(struct fsck_options
*options
,
138 enum fsck_msg_id msg_id
,
139 enum fsck_msg_type msg_type
)
141 if (!options
->msg_type
) {
143 enum fsck_msg_type
*severity
;
144 ALLOC_ARRAY(severity
, FSCK_MSG_MAX
);
145 for (i
= 0; i
< FSCK_MSG_MAX
; i
++)
146 severity
[i
] = fsck_msg_type(i
, options
);
147 options
->msg_type
= severity
;
150 options
->msg_type
[msg_id
] = msg_type
;
153 void fsck_set_msg_type(struct fsck_options
*options
,
154 const char *msg_id_str
, const char *msg_type_str
)
156 int msg_id
= parse_msg_id(msg_id_str
);
157 char *to_free
= NULL
;
158 enum fsck_msg_type msg_type
;
161 die("Unhandled message id: %s", msg_id_str
);
163 if (msg_id
== FSCK_MSG_LARGE_PATHNAME
) {
164 const char *colon
= strchr(msg_type_str
, ':');
166 msg_type_str
= to_free
=
167 xmemdupz(msg_type_str
, colon
- msg_type_str
);
169 if (!git_parse_ssize_t(colon
, &max_tree_entry_len
))
170 die("unable to parse max tree entry len: %s", colon
);
173 msg_type
= parse_msg_type(msg_type_str
);
175 if (msg_type
!= FSCK_ERROR
&& msg_id_info
[msg_id
].msg_type
== FSCK_FATAL
)
176 die("Cannot demote %s to %s", msg_id_str
, msg_type_str
);
178 fsck_set_msg_type_from_ids(options
, msg_id
, msg_type
);
182 void fsck_set_msg_types(struct fsck_options
*options
, const char *values
)
184 char *buf
= xstrdup(values
), *to_free
= buf
;
188 int len
= strcspn(buf
, " ,|"), equal
;
198 equal
< len
&& buf
[equal
] != '=' && buf
[equal
] != ':';
200 buf
[equal
] = tolower(buf
[equal
]);
203 if (!strcmp(buf
, "skiplist")) {
205 die("skiplist requires a path");
206 oidset_parse_file(&options
->skiplist
, buf
+ equal
+ 1);
212 die("Missing '=': '%s'", buf
);
214 fsck_set_msg_type(options
, buf
, buf
+ equal
+ 1);
220 static int object_on_skiplist(struct fsck_options
*opts
,
221 const struct object_id
*oid
)
223 return opts
&& oid
&& oidset_contains(&opts
->skiplist
, oid
);
226 __attribute__((format (printf
, 5, 6)))
227 static int report(struct fsck_options
*options
,
228 const struct object_id
*oid
, enum object_type object_type
,
229 enum fsck_msg_id msg_id
, const char *fmt
, ...)
232 struct strbuf sb
= STRBUF_INIT
;
233 enum fsck_msg_type msg_type
= fsck_msg_type(msg_id
, options
);
236 if (msg_type
== FSCK_IGNORE
)
239 if (object_on_skiplist(options
, oid
))
242 if (msg_type
== FSCK_FATAL
)
243 msg_type
= FSCK_ERROR
;
244 else if (msg_type
== FSCK_INFO
)
245 msg_type
= FSCK_WARN
;
248 strbuf_addf(&sb
, "%s: ", msg_id_info
[msg_id
].camelcased
);
251 strbuf_vaddf(&sb
, fmt
, ap
);
252 result
= options
->error_func(options
, oid
, object_type
,
253 msg_type
, msg_id
, sb
.buf
);
260 void fsck_enable_object_names(struct fsck_options
*options
)
262 if (!options
->object_names
)
263 options
->object_names
= kh_init_oid_map();
266 const char *fsck_get_object_name(struct fsck_options
*options
,
267 const struct object_id
*oid
)
270 if (!options
->object_names
)
272 pos
= kh_get_oid_map(options
->object_names
, *oid
);
273 if (pos
>= kh_end(options
->object_names
))
275 return kh_value(options
->object_names
, pos
);
278 void fsck_put_object_name(struct fsck_options
*options
,
279 const struct object_id
*oid
,
280 const char *fmt
, ...)
283 struct strbuf buf
= STRBUF_INIT
;
287 if (!options
->object_names
)
290 pos
= kh_put_oid_map(options
->object_names
, *oid
, &hashret
);
294 strbuf_vaddf(&buf
, fmt
, ap
);
295 kh_value(options
->object_names
, pos
) = strbuf_detach(&buf
, NULL
);
299 const char *fsck_describe_object(struct fsck_options
*options
,
300 const struct object_id
*oid
)
302 static struct strbuf bufs
[] = {
303 STRBUF_INIT
, STRBUF_INIT
, STRBUF_INIT
, STRBUF_INIT
307 const char *name
= fsck_get_object_name(options
, oid
);
310 b
= (b
+ 1) % ARRAY_SIZE(bufs
);
312 strbuf_addstr(buf
, oid_to_hex(oid
));
314 strbuf_addf(buf
, " (%s)", name
);
319 static int fsck_walk_tree(struct tree
*tree
, void *data
, struct fsck_options
*options
)
321 struct tree_desc desc
;
322 struct name_entry entry
;
326 if (parse_tree(tree
))
329 name
= fsck_get_object_name(options
, &tree
->object
.oid
);
330 if (init_tree_desc_gently(&desc
, &tree
->object
.oid
,
331 tree
->buffer
, tree
->size
, 0))
333 while (tree_entry_gently(&desc
, &entry
)) {
337 if (S_ISGITLINK(entry
.mode
))
340 if (S_ISDIR(entry
.mode
)) {
341 obj
= (struct object
*)lookup_tree(the_repository
, &entry
.oid
);
343 fsck_put_object_name(options
, &entry
.oid
, "%s%s/",
345 result
= options
->walk(obj
, OBJ_TREE
, data
, options
);
347 else if (S_ISREG(entry
.mode
) || S_ISLNK(entry
.mode
)) {
348 obj
= (struct object
*)lookup_blob(the_repository
, &entry
.oid
);
350 fsck_put_object_name(options
, &entry
.oid
, "%s%s",
352 result
= options
->walk(obj
, OBJ_BLOB
, data
, options
);
355 result
= error("in tree %s: entry %s has bad mode %.6o",
356 fsck_describe_object(options
, &tree
->object
.oid
),
357 entry
.path
, entry
.mode
);
367 static int fsck_walk_commit(struct commit
*commit
, void *data
, struct fsck_options
*options
)
369 int counter
= 0, generation
= 0, name_prefix_len
= 0;
370 struct commit_list
*parents
;
375 if (repo_parse_commit(the_repository
, commit
))
378 name
= fsck_get_object_name(options
, &commit
->object
.oid
);
380 fsck_put_object_name(options
, get_commit_tree_oid(commit
),
383 result
= options
->walk((struct object
*) repo_get_commit_tree(the_repository
, commit
),
384 OBJ_TREE
, data
, options
);
389 parents
= commit
->parents
;
390 if (name
&& parents
) {
391 int len
= strlen(name
), power
;
393 if (len
&& name
[len
- 1] == '^') {
395 name_prefix_len
= len
- 1;
397 else { /* parse ~<generation> suffix */
398 for (generation
= 0, power
= 1;
399 len
&& isdigit(name
[len
- 1]);
401 generation
+= power
* (name
[--len
] - '0');
402 if (power
> 1 && len
&& name
[len
- 1] == '~')
403 name_prefix_len
= len
- 1;
405 /* Maybe a non-first parent, e.g. HEAD^2 */
407 name_prefix_len
= len
;
414 struct object_id
*oid
= &parents
->item
->object
.oid
;
417 fsck_put_object_name(options
, oid
, "%s^%d",
419 else if (generation
> 0)
420 fsck_put_object_name(options
, oid
, "%.*s~%d",
421 name_prefix_len
, name
,
424 fsck_put_object_name(options
, oid
, "%s^", name
);
426 result
= options
->walk((struct object
*)parents
->item
, OBJ_COMMIT
, data
, options
);
431 parents
= parents
->next
;
436 static int fsck_walk_tag(struct tag
*tag
, void *data
, struct fsck_options
*options
)
438 const char *name
= fsck_get_object_name(options
, &tag
->object
.oid
);
443 fsck_put_object_name(options
, &tag
->tagged
->oid
, "%s", name
);
444 return options
->walk(tag
->tagged
, OBJ_ANY
, data
, options
);
447 int fsck_walk(struct object
*obj
, void *data
, struct fsck_options
*options
)
452 if (obj
->type
== OBJ_NONE
)
453 parse_object(the_repository
, &obj
->oid
);
459 return fsck_walk_tree((struct tree
*)obj
, data
, options
);
461 return fsck_walk_commit((struct commit
*)obj
, data
, options
);
463 return fsck_walk_tag((struct tag
*)obj
, data
, options
);
465 error("Unknown object type for %s",
466 fsck_describe_object(options
, &obj
->oid
));
476 static void name_stack_push(struct name_stack
*stack
, const char *name
)
478 ALLOC_GROW(stack
->names
, stack
->nr
+ 1, stack
->alloc
);
479 stack
->names
[stack
->nr
++] = name
;
482 static const char *name_stack_pop(struct name_stack
*stack
)
484 return stack
->nr
? stack
->names
[--stack
->nr
] : NULL
;
487 static void name_stack_clear(struct name_stack
*stack
)
489 FREE_AND_NULL(stack
->names
);
490 stack
->nr
= stack
->alloc
= 0;
494 * The entries in a tree are ordered in the _path_ order,
495 * which means that a directory entry is ordered by adding
496 * a slash to the end of it.
498 * So a directory called "a" is ordered _after_ a file
499 * called "a.c", because "a/" sorts after "a.c".
501 #define TREE_UNORDERED (-1)
502 #define TREE_HAS_DUPS (-2)
504 static int is_less_than_slash(unsigned char c
)
506 return '\0' < c
&& c
< '/';
509 static int verify_ordered(unsigned mode1
, const char *name1
,
510 unsigned mode2
, const char *name2
,
511 struct name_stack
*candidates
)
513 int len1
= strlen(name1
);
514 int len2
= strlen(name2
);
515 int len
= len1
< len2
? len1
: len2
;
516 unsigned char c1
, c2
;
519 cmp
= memcmp(name1
, name2
, len
);
523 return TREE_UNORDERED
;
526 * Ok, the first <len> characters are the same.
527 * Now we need to order the next one, but turn
528 * a '\0' into a '/' for a directory entry.
534 * git-write-tree used to write out a nonsense tree that has
535 * entries with the same name, one blob and one tree. Make
536 * sure we do not have duplicate entries.
538 return TREE_HAS_DUPS
;
539 if (!c1
&& S_ISDIR(mode1
))
541 if (!c2
&& S_ISDIR(mode2
))
545 * There can be non-consecutive duplicates due to the implicitly
554 * Record non-directory candidates (like "foo" and "foo.bar" in
555 * the example) on a stack and check directory candidates (like
556 * foo/" and "foo.bar/") against that stack.
558 if (!c1
&& is_less_than_slash(c2
)) {
559 name_stack_push(candidates
, name1
);
560 } else if (c2
== '/' && is_less_than_slash(c1
)) {
563 const char *f_name
= name_stack_pop(candidates
);
567 if (!skip_prefix(name2
, f_name
, &p
))
570 return TREE_HAS_DUPS
;
571 if (is_less_than_slash(*p
)) {
572 name_stack_push(candidates
, f_name
);
578 return c1
< c2
? 0 : TREE_UNORDERED
;
581 static int fsck_tree(const struct object_id
*tree_oid
,
582 const char *buffer
, unsigned long size
,
583 struct fsck_options
*options
)
586 int has_null_sha1
= 0;
587 int has_full_path
= 0;
588 int has_empty_name
= 0;
592 int has_zero_pad
= 0;
593 int has_bad_modes
= 0;
594 int has_dup_entries
= 0;
595 int not_properly_sorted
= 0;
596 int has_large_name
= 0;
597 struct tree_desc desc
;
600 struct name_stack df_dup_candidates
= { NULL
};
602 if (init_tree_desc_gently(&desc
, tree_oid
, buffer
, size
,
603 TREE_DESC_RAW_MODES
)) {
604 retval
+= report(options
, tree_oid
, OBJ_TREE
,
606 "cannot be parsed as a tree");
615 const char *name
, *backslash
;
616 const struct object_id
*entry_oid
;
618 entry_oid
= tree_entry_extract(&desc
, &name
, &mode
);
620 has_null_sha1
|= is_null_oid(entry_oid
);
621 has_full_path
|= !!strchr(name
, '/');
622 has_empty_name
|= !*name
;
623 has_dot
|= !strcmp(name
, ".");
624 has_dotdot
|= !strcmp(name
, "..");
625 has_dotgit
|= is_hfs_dotgit(name
) || is_ntfs_dotgit(name
);
626 has_zero_pad
|= *(char *)desc
.buffer
== '0';
627 has_large_name
|= tree_entry_len(&desc
.entry
) > max_tree_entry_len
;
629 if (is_hfs_dotgitmodules(name
) || is_ntfs_dotgitmodules(name
)) {
631 oidset_insert(&options
->gitmodules_found
,
634 retval
+= report(options
,
636 FSCK_MSG_GITMODULES_SYMLINK
,
637 ".gitmodules is a symbolic link");
640 if (is_hfs_dotgitattributes(name
) || is_ntfs_dotgitattributes(name
)) {
642 oidset_insert(&options
->gitattributes_found
,
645 retval
+= report(options
, tree_oid
, OBJ_TREE
,
646 FSCK_MSG_GITATTRIBUTES_SYMLINK
,
647 ".gitattributes is a symlink");
651 if (is_hfs_dotgitignore(name
) ||
652 is_ntfs_dotgitignore(name
))
653 retval
+= report(options
, tree_oid
, OBJ_TREE
,
654 FSCK_MSG_GITIGNORE_SYMLINK
,
655 ".gitignore is a symlink");
656 if (is_hfs_dotmailmap(name
) ||
657 is_ntfs_dotmailmap(name
))
658 retval
+= report(options
, tree_oid
, OBJ_TREE
,
659 FSCK_MSG_MAILMAP_SYMLINK
,
660 ".mailmap is a symlink");
661 oidset_insert(&options
->symlink_targets_found
,
665 if ((backslash
= strchr(name
, '\\'))) {
668 has_dotgit
|= is_ntfs_dotgit(backslash
);
669 if (is_ntfs_dotgitmodules(backslash
)) {
671 oidset_insert(&options
->gitmodules_found
,
674 retval
+= report(options
, tree_oid
, OBJ_TREE
,
675 FSCK_MSG_GITMODULES_SYMLINK
,
676 ".gitmodules is a symbolic link");
678 backslash
= strchr(backslash
, '\\');
682 if (update_tree_entry_gently(&desc
)) {
683 retval
+= report(options
, tree_oid
, OBJ_TREE
,
685 "cannot be parsed as a tree");
700 * This is nonstandard, but we had a few of these
701 * early on when we honored the full set of mode
705 if (!options
->strict
)
713 switch (verify_ordered(o_mode
, o_name
, mode
, name
,
714 &df_dup_candidates
)) {
716 not_properly_sorted
= 1;
730 name_stack_clear(&df_dup_candidates
);
733 retval
+= report(options
, tree_oid
, OBJ_TREE
,
735 "contains entries pointing to null sha1");
737 retval
+= report(options
, tree_oid
, OBJ_TREE
,
738 FSCK_MSG_FULL_PATHNAME
,
739 "contains full pathnames");
741 retval
+= report(options
, tree_oid
, OBJ_TREE
,
743 "contains empty pathname");
745 retval
+= report(options
, tree_oid
, OBJ_TREE
,
749 retval
+= report(options
, tree_oid
, OBJ_TREE
,
753 retval
+= report(options
, tree_oid
, OBJ_TREE
,
757 retval
+= report(options
, tree_oid
, OBJ_TREE
,
758 FSCK_MSG_ZERO_PADDED_FILEMODE
,
759 "contains zero-padded file modes");
761 retval
+= report(options
, tree_oid
, OBJ_TREE
,
762 FSCK_MSG_BAD_FILEMODE
,
763 "contains bad file modes");
765 retval
+= report(options
, tree_oid
, OBJ_TREE
,
766 FSCK_MSG_DUPLICATE_ENTRIES
,
767 "contains duplicate file entries");
768 if (not_properly_sorted
)
769 retval
+= report(options
, tree_oid
, OBJ_TREE
,
770 FSCK_MSG_TREE_NOT_SORTED
,
771 "not properly sorted");
773 retval
+= report(options
, tree_oid
, OBJ_TREE
,
774 FSCK_MSG_LARGE_PATHNAME
,
775 "contains excessively large pathname");
780 * Confirm that the headers of a commit or tag object end in a reasonable way,
781 * either with the usual "\n\n" separator, or at least with a trailing newline
782 * on the final header line.
784 * This property is important for the memory safety of our callers. It allows
785 * them to scan the buffer linewise without constantly checking the remaining
788 * - they check that there are bytes left in the buffer at the start of any
789 * line (i.e., that the last newline they saw was not the final one we
792 * - any intra-line scanning they do will stop at a newline, which will worst
793 * case hit the newline we found here as the end-of-header. This makes it
794 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
796 static int verify_headers(const void *data
, unsigned long size
,
797 const struct object_id
*oid
, enum object_type type
,
798 struct fsck_options
*options
)
800 const char *buffer
= (const char *)data
;
803 for (i
= 0; i
< size
; i
++) {
806 return report(options
, oid
, type
,
807 FSCK_MSG_NUL_IN_HEADER
,
808 "unterminated header: NUL at offset %ld", i
);
810 if (i
+ 1 < size
&& buffer
[i
+ 1] == '\n')
816 * We did not find double-LF that separates the header
817 * and the body. Not having a body is not a crime but
818 * we do want to see the terminating LF for the last header
821 if (size
&& buffer
[size
- 1] == '\n')
824 return report(options
, oid
, type
,
825 FSCK_MSG_UNTERMINATED_HEADER
, "unterminated header");
828 static int fsck_ident(const char **ident
,
829 const struct object_id
*oid
, enum object_type type
,
830 struct fsck_options
*options
)
832 const char *p
= *ident
;
835 *ident
= strchrnul(*ident
, '\n');
840 return report(options
, oid
, type
, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL
, "invalid author/committer line - missing space before email");
841 p
+= strcspn(p
, "<>\n");
843 return report(options
, oid
, type
, FSCK_MSG_BAD_NAME
, "invalid author/committer line - bad name");
845 return report(options
, oid
, type
, FSCK_MSG_MISSING_EMAIL
, "invalid author/committer line - missing email");
847 return report(options
, oid
, type
, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL
, "invalid author/committer line - missing space before email");
849 p
+= strcspn(p
, "<>\n");
851 return report(options
, oid
, type
, FSCK_MSG_BAD_EMAIL
, "invalid author/committer line - bad email");
854 return report(options
, oid
, type
, FSCK_MSG_MISSING_SPACE_BEFORE_DATE
, "invalid author/committer line - missing space before date");
857 * Our timestamp parser is based on the C strto*() functions, which
858 * will happily eat whitespace, including the newline that is supposed
859 * to prevent us walking past the end of the buffer. So do our own
860 * scan, skipping linear whitespace but not newlines, and then
861 * confirming we found a digit. We _could_ be even more strict here,
862 * as we really expect only a single space, but since we have
863 * traditionally allowed extra whitespace, we'll continue to do so.
865 while (*p
== ' ' || *p
== '\t')
868 return report(options
, oid
, type
, FSCK_MSG_BAD_DATE
,
869 "invalid author/committer line - bad date");
870 if (*p
== '0' && p
[1] != ' ')
871 return report(options
, oid
, type
, FSCK_MSG_ZERO_PADDED_DATE
, "invalid author/committer line - zero-padded date");
872 if (date_overflows(parse_timestamp(p
, &end
, 10)))
873 return report(options
, oid
, type
, FSCK_MSG_BAD_DATE_OVERFLOW
, "invalid author/committer line - date causes integer overflow");
874 if ((end
== p
|| *end
!= ' '))
875 return report(options
, oid
, type
, FSCK_MSG_BAD_DATE
, "invalid author/committer line - bad date");
877 if ((*p
!= '+' && *p
!= '-') ||
883 return report(options
, oid
, type
, FSCK_MSG_BAD_TIMEZONE
, "invalid author/committer line - bad time zone");
888 static int fsck_commit(const struct object_id
*oid
,
889 const char *buffer
, unsigned long size
,
890 struct fsck_options
*options
)
892 struct object_id tree_oid
, parent_oid
;
893 unsigned author_count
;
895 const char *buffer_begin
= buffer
;
896 const char *buffer_end
= buffer
+ size
;
900 * We _must_ stop parsing immediately if this reports failure, as the
901 * memory safety of the rest of the function depends on it. See the
902 * comment above the definition of verify_headers() for more details.
904 if (verify_headers(buffer
, size
, oid
, OBJ_COMMIT
, options
))
907 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "tree ", &buffer
))
908 return report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MISSING_TREE
, "invalid format - expected 'tree' line");
909 if (parse_oid_hex(buffer
, &tree_oid
, &p
) || *p
!= '\n') {
910 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_BAD_TREE_SHA1
, "invalid 'tree' line format - bad sha1");
915 while (buffer
< buffer_end
&& skip_prefix(buffer
, "parent ", &buffer
)) {
916 if (parse_oid_hex(buffer
, &parent_oid
, &p
) || *p
!= '\n') {
917 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_BAD_PARENT_SHA1
, "invalid 'parent' line format - bad sha1");
924 while (buffer
< buffer_end
&& skip_prefix(buffer
, "author ", &buffer
)) {
926 err
= fsck_ident(&buffer
, oid
, OBJ_COMMIT
, options
);
930 if (author_count
< 1)
931 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MISSING_AUTHOR
, "invalid format - expected 'author' line");
932 else if (author_count
> 1)
933 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MULTIPLE_AUTHORS
, "invalid format - multiple 'author' lines");
936 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "committer ", &buffer
))
937 return report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MISSING_COMMITTER
, "invalid format - expected 'committer' line");
938 err
= fsck_ident(&buffer
, oid
, OBJ_COMMIT
, options
);
941 if (memchr(buffer_begin
, '\0', size
)) {
942 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_NUL_IN_COMMIT
,
943 "NUL byte in the commit object body");
950 static int fsck_tag(const struct object_id
*oid
, const char *buffer
,
951 unsigned long size
, struct fsck_options
*options
)
953 struct object_id tagged_oid
;
955 return fsck_tag_standalone(oid
, buffer
, size
, options
, &tagged_oid
,
959 int fsck_tag_standalone(const struct object_id
*oid
, const char *buffer
,
960 unsigned long size
, struct fsck_options
*options
,
961 struct object_id
*tagged_oid
,
966 struct strbuf sb
= STRBUF_INIT
;
967 const char *buffer_end
= buffer
+ size
;
971 * We _must_ stop parsing immediately if this reports failure, as the
972 * memory safety of the rest of the function depends on it. See the
973 * comment above the definition of verify_headers() for more details.
975 ret
= verify_headers(buffer
, size
, oid
, OBJ_TAG
, options
);
979 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "object ", &buffer
)) {
980 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_OBJECT
, "invalid format - expected 'object' line");
983 if (parse_oid_hex(buffer
, tagged_oid
, &p
) || *p
!= '\n') {
984 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_BAD_OBJECT_SHA1
, "invalid 'object' line format - bad sha1");
990 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "type ", &buffer
)) {
991 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TYPE_ENTRY
, "invalid format - expected 'type' line");
994 eol
= memchr(buffer
, '\n', buffer_end
- buffer
);
996 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TYPE
, "invalid format - unexpected end after 'type' line");
999 *tagged_type
= type_from_string_gently(buffer
, eol
- buffer
, 1);
1000 if (*tagged_type
< 0)
1001 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_BAD_TYPE
, "invalid 'type' value");
1006 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "tag ", &buffer
)) {
1007 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TAG_ENTRY
, "invalid format - expected 'tag' line");
1010 eol
= memchr(buffer
, '\n', buffer_end
- buffer
);
1012 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TAG
, "invalid format - unexpected end after 'type' line");
1015 strbuf_addf(&sb
, "refs/tags/%.*s", (int)(eol
- buffer
), buffer
);
1016 if (check_refname_format(sb
.buf
, 0)) {
1017 ret
= report(options
, oid
, OBJ_TAG
,
1018 FSCK_MSG_BAD_TAG_NAME
,
1019 "invalid 'tag' name: %.*s",
1020 (int)(eol
- buffer
), buffer
);
1026 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "tagger ", &buffer
)) {
1027 /* early tags do not contain 'tagger' lines; warn only */
1028 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TAGGER_ENTRY
, "invalid format - expected 'tagger' line");
1033 ret
= fsck_ident(&buffer
, oid
, OBJ_TAG
, options
);
1035 if (buffer
< buffer_end
&& !starts_with(buffer
, "\n")) {
1037 * The verify_headers() check will allow
1038 * e.g. "[...]tagger <tagger>\nsome
1039 * garbage\n\nmessage" to pass, thinking "some
1040 * garbage" could be a custom header. E.g. "mktag"
1041 * doesn't want any unknown headers.
1043 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_EXTRA_HEADER_ENTRY
, "invalid format - extra header(s) after 'tagger'");
1049 strbuf_release(&sb
);
1053 struct fsck_gitmodules_data
{
1054 const struct object_id
*oid
;
1055 struct fsck_options
*options
;
1059 static int fsck_gitmodules_fn(const char *var
, const char *value
,
1060 const struct config_context
*ctx UNUSED
,
1063 struct fsck_gitmodules_data
*data
= vdata
;
1064 const char *subsection
, *key
;
1065 size_t subsection_len
;
1068 if (parse_config_key(var
, "submodule", &subsection
, &subsection_len
, &key
) < 0 ||
1072 name
= xmemdupz(subsection
, subsection_len
);
1073 if (check_submodule_name(name
) < 0)
1074 data
->ret
|= report(data
->options
,
1075 data
->oid
, OBJ_BLOB
,
1076 FSCK_MSG_GITMODULES_NAME
,
1077 "disallowed submodule name: %s",
1079 if (!strcmp(key
, "url") && value
&&
1080 check_submodule_url(value
) < 0)
1081 data
->ret
|= report(data
->options
,
1082 data
->oid
, OBJ_BLOB
,
1083 FSCK_MSG_GITMODULES_URL
,
1084 "disallowed submodule url: %s",
1086 if (!strcmp(key
, "path") && value
&&
1087 looks_like_command_line_option(value
))
1088 data
->ret
|= report(data
->options
,
1089 data
->oid
, OBJ_BLOB
,
1090 FSCK_MSG_GITMODULES_PATH
,
1091 "disallowed submodule path: %s",
1093 if (!strcmp(key
, "update") && value
&&
1094 parse_submodule_update_type(value
) == SM_UPDATE_COMMAND
)
1095 data
->ret
|= report(data
->options
, data
->oid
, OBJ_BLOB
,
1096 FSCK_MSG_GITMODULES_UPDATE
,
1097 "disallowed submodule update setting: %s",
1104 static int fsck_blob(const struct object_id
*oid
, const char *buf
,
1105 unsigned long size
, struct fsck_options
*options
)
1109 if (object_on_skiplist(options
, oid
))
1112 if (oidset_contains(&options
->gitmodules_found
, oid
)) {
1113 struct config_options config_opts
= { 0 };
1114 struct fsck_gitmodules_data data
;
1116 oidset_insert(&options
->gitmodules_done
, oid
);
1120 * A missing buffer here is a sign that the caller found the
1121 * blob too gigantic to load into memory. Let's just consider
1124 return report(options
, oid
, OBJ_BLOB
,
1125 FSCK_MSG_GITMODULES_LARGE
,
1126 ".gitmodules too large to parse");
1130 data
.options
= options
;
1132 config_opts
.error_action
= CONFIG_ERROR_SILENT
;
1133 if (git_config_from_mem(fsck_gitmodules_fn
, CONFIG_ORIGIN_BLOB
,
1134 ".gitmodules", buf
, size
, &data
,
1135 CONFIG_SCOPE_UNKNOWN
, &config_opts
))
1136 data
.ret
|= report(options
, oid
, OBJ_BLOB
,
1137 FSCK_MSG_GITMODULES_PARSE
,
1138 "could not parse gitmodules blob");
1142 if (oidset_contains(&options
->gitattributes_found
, oid
)) {
1145 oidset_insert(&options
->gitattributes_done
, oid
);
1147 if (!buf
|| size
> ATTR_MAX_FILE_SIZE
) {
1149 * A missing buffer here is a sign that the caller found the
1150 * blob too gigantic to load into memory. Let's just consider
1153 return report(options
, oid
, OBJ_BLOB
,
1154 FSCK_MSG_GITATTRIBUTES_LARGE
,
1155 ".gitattributes too large to parse");
1158 for (ptr
= buf
; *ptr
; ) {
1159 const char *eol
= strchrnul(ptr
, '\n');
1160 if (eol
- ptr
>= ATTR_MAX_LINE_LENGTH
) {
1161 ret
|= report(options
, oid
, OBJ_BLOB
,
1162 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH
,
1163 ".gitattributes has too long lines to parse");
1167 ptr
= *eol
? eol
+ 1 : eol
;
1171 if (oidset_contains(&options
->symlink_targets_found
, oid
)) {
1172 const char *ptr
= buf
;
1173 const struct object_id
*reported
= NULL
;
1175 oidset_insert(&options
->symlink_targets_done
, oid
);
1177 if (!buf
|| size
> PATH_MAX
) {
1179 * A missing buffer here is a sign that the caller found the
1180 * blob too gigantic to load into memory. Let's just consider
1183 return report(options
, oid
, OBJ_BLOB
,
1184 FSCK_MSG_SYMLINK_TARGET_LENGTH
,
1185 "symlink target too long");
1188 while (!reported
&& ptr
) {
1189 const char *p
= ptr
;
1190 char c
, *slash
= strchrnul(ptr
, '/');
1191 char *backslash
= memchr(ptr
, '\\', slash
- ptr
);
1196 while (!reported
&& backslash
) {
1198 if (is_ntfs_dotgit(p
))
1199 ret
|= report(options
, reported
= oid
, OBJ_BLOB
,
1200 FSCK_MSG_SYMLINK_POINTS_TO_GIT_DIR
,
1201 "symlink target points to git dir");
1204 backslash
= memchr(p
, '\\', slash
- p
);
1206 if (!reported
&& is_ntfs_dotgit(p
))
1207 ret
|= report(options
, reported
= oid
, OBJ_BLOB
,
1208 FSCK_MSG_SYMLINK_POINTS_TO_GIT_DIR
,
1209 "symlink target points to git dir");
1211 if (!reported
&& is_hfs_dotgit(ptr
))
1212 ret
|= report(options
, reported
= oid
, OBJ_BLOB
,
1213 FSCK_MSG_SYMLINK_POINTS_TO_GIT_DIR
,
1214 "symlink target points to git dir");
1217 ptr
= c
? slash
+ 1 : NULL
;
1224 int fsck_object(struct object
*obj
, void *data
, unsigned long size
,
1225 struct fsck_options
*options
)
1228 return report(options
, NULL
, OBJ_NONE
, FSCK_MSG_BAD_OBJECT_SHA1
, "no valid object to fsck");
1230 return fsck_buffer(&obj
->oid
, obj
->type
, data
, size
, options
);
1233 int fsck_buffer(const struct object_id
*oid
, enum object_type type
,
1234 void *data
, unsigned long size
,
1235 struct fsck_options
*options
)
1237 if (type
== OBJ_BLOB
)
1238 return fsck_blob(oid
, data
, size
, options
);
1239 if (type
== OBJ_TREE
)
1240 return fsck_tree(oid
, data
, size
, options
);
1241 if (type
== OBJ_COMMIT
)
1242 return fsck_commit(oid
, data
, size
, options
);
1243 if (type
== OBJ_TAG
)
1244 return fsck_tag(oid
, data
, size
, options
);
1246 return report(options
, oid
, type
,
1247 FSCK_MSG_UNKNOWN_TYPE
,
1248 "unknown type '%d' (internal fsck error)",
1252 int fsck_error_function(struct fsck_options
*o
,
1253 const struct object_id
*oid
,
1254 enum object_type object_type UNUSED
,
1255 enum fsck_msg_type msg_type
,
1256 enum fsck_msg_id msg_id UNUSED
,
1257 const char *message
)
1259 if (msg_type
== FSCK_WARN
) {
1260 warning("object %s: %s", fsck_describe_object(o
, oid
), message
);
1263 error("object %s: %s", fsck_describe_object(o
, oid
), message
);
1267 static int fsck_blobs(struct oidset
*blobs_found
, struct oidset
*blobs_done
,
1268 enum fsck_msg_id msg_missing
, enum fsck_msg_id msg_type
,
1269 struct fsck_options
*options
, const char *blob_type
)
1272 struct oidset_iter iter
;
1273 const struct object_id
*oid
;
1275 oidset_iter_init(blobs_found
, &iter
);
1276 while ((oid
= oidset_iter_next(&iter
))) {
1277 enum object_type type
;
1281 if (oidset_contains(blobs_done
, oid
))
1284 buf
= repo_read_object_file(the_repository
, oid
, &type
, &size
);
1286 if (is_promisor_object(oid
))
1288 ret
|= report(options
,
1289 oid
, OBJ_BLOB
, msg_missing
,
1290 "unable to read %s blob", blob_type
);
1294 if (type
== OBJ_BLOB
)
1295 ret
|= fsck_blob(oid
, buf
, size
, options
);
1297 ret
|= report(options
, oid
, type
, msg_type
,
1298 "non-blob found at %s", blob_type
);
1302 oidset_clear(blobs_found
);
1303 oidset_clear(blobs_done
);
1308 int fsck_finish(struct fsck_options
*options
)
1312 ret
|= fsck_blobs(&options
->gitmodules_found
, &options
->gitmodules_done
,
1313 FSCK_MSG_GITMODULES_MISSING
, FSCK_MSG_GITMODULES_BLOB
,
1314 options
, ".gitmodules");
1315 ret
|= fsck_blobs(&options
->gitattributes_found
, &options
->gitattributes_done
,
1316 FSCK_MSG_GITATTRIBUTES_MISSING
, FSCK_MSG_GITATTRIBUTES_BLOB
,
1317 options
, ".gitattributes");
1319 ret
|= fsck_blobs(&options
->symlink_targets_found
, &options
->symlink_targets_done
,
1320 FSCK_MSG_SYMLINK_TARGET_MISSING
, FSCK_MSG_SYMLINK_TARGET_BLOB
,
1321 options
, "<symlink-target>");
1326 int git_fsck_config(const char *var
, const char *value
,
1327 const struct config_context
*ctx
, void *cb
)
1329 struct fsck_options
*options
= cb
;
1332 if (strcmp(var
, "fsck.skiplist") == 0) {
1334 struct strbuf sb
= STRBUF_INIT
;
1336 if (git_config_pathname(&path
, var
, value
))
1338 strbuf_addf(&sb
, "skiplist=%s", path
);
1340 fsck_set_msg_types(options
, sb
.buf
);
1341 strbuf_release(&sb
);
1345 if (skip_prefix(var
, "fsck.", &msg_id
)) {
1347 return config_error_nonbool(var
);
1348 fsck_set_msg_type(options
, msg_id
, value
);
1352 return git_default_config(var
, value
, ctx
, cb
);
1356 * Custom error callbacks that are used in more than one place.
1359 int fsck_error_cb_print_missing_gitmodules(struct fsck_options
*o
,
1360 const struct object_id
*oid
,
1361 enum object_type object_type
,
1362 enum fsck_msg_type msg_type
,
1363 enum fsck_msg_id msg_id
,
1364 const char *message
)
1366 if (msg_id
== FSCK_MSG_GITMODULES_MISSING
) {
1367 puts(oid_to_hex(oid
));
1370 return fsck_error_function(o
, oid
, object_type
, msg_type
, msg_id
, message
);