1 #include "git-compat-util.h"
5 #include "object-store-ll.h"
7 #include "repository.h"
12 #include "tree-walk.h"
22 #include "submodule-config.h"
24 #include "credential.h"
27 static ssize_t max_tree_entry_len
= 4096;
30 #define MSG_ID(id, msg_type) { STR(id), NULL, NULL, FSCK_##msg_type },
32 const char *id_string
;
33 const char *downcased
;
34 const char *camelcased
;
35 enum fsck_msg_type msg_type
;
36 } msg_id_info
[FSCK_MSG_MAX
+ 1] = {
37 FOREACH_FSCK_MSG_ID(MSG_ID
)
38 { NULL
, NULL
, NULL
, -1 }
43 static void prepare_msg_ids(void)
47 if (msg_id_info
[0].downcased
)
50 /* convert id_string to lower case, without underscores. */
51 for (i
= 0; i
< FSCK_MSG_MAX
; i
++) {
52 const char *p
= msg_id_info
[i
].id_string
;
54 char *q
= xmalloc(len
);
56 msg_id_info
[i
].downcased
= q
;
61 *(q
)++ = tolower(*(p
)++);
64 p
= msg_id_info
[i
].id_string
;
66 msg_id_info
[i
].camelcased
= q
;
80 static int parse_msg_id(const char *text
)
86 for (i
= 0; i
< FSCK_MSG_MAX
; i
++)
87 if (!strcmp(text
, msg_id_info
[i
].downcased
))
93 void list_config_fsck_msg_ids(struct string_list
*list
, const char *prefix
)
99 for (i
= 0; i
< FSCK_MSG_MAX
; i
++)
100 list_config_item(list
, prefix
, msg_id_info
[i
].camelcased
);
103 static enum fsck_msg_type
fsck_msg_type(enum fsck_msg_id msg_id
,
104 struct fsck_options
*options
)
106 assert(msg_id
>= 0 && msg_id
< FSCK_MSG_MAX
);
108 if (!options
->msg_type
) {
109 enum fsck_msg_type msg_type
= msg_id_info
[msg_id
].msg_type
;
111 if (options
->strict
&& msg_type
== FSCK_WARN
)
112 msg_type
= FSCK_ERROR
;
116 return options
->msg_type
[msg_id
];
119 static enum fsck_msg_type
parse_msg_type(const char *str
)
121 if (!strcmp(str
, "error"))
123 else if (!strcmp(str
, "warn"))
125 else if (!strcmp(str
, "ignore"))
128 die("Unknown fsck message type: '%s'", str
);
131 int is_valid_msg_type(const char *msg_id
, const char *msg_type
)
133 if (parse_msg_id(msg_id
) < 0)
135 parse_msg_type(msg_type
);
139 void fsck_set_msg_type_from_ids(struct fsck_options
*options
,
140 enum fsck_msg_id msg_id
,
141 enum fsck_msg_type msg_type
)
143 if (!options
->msg_type
) {
145 enum fsck_msg_type
*severity
;
146 ALLOC_ARRAY(severity
, FSCK_MSG_MAX
);
147 for (i
= 0; i
< FSCK_MSG_MAX
; i
++)
148 severity
[i
] = fsck_msg_type(i
, options
);
149 options
->msg_type
= severity
;
152 options
->msg_type
[msg_id
] = msg_type
;
155 void fsck_set_msg_type(struct fsck_options
*options
,
156 const char *msg_id_str
, const char *msg_type_str
)
158 int msg_id
= parse_msg_id(msg_id_str
);
159 char *to_free
= NULL
;
160 enum fsck_msg_type msg_type
;
163 die("Unhandled message id: %s", msg_id_str
);
165 if (msg_id
== FSCK_MSG_LARGE_PATHNAME
) {
166 const char *colon
= strchr(msg_type_str
, ':');
168 msg_type_str
= to_free
=
169 xmemdupz(msg_type_str
, colon
- msg_type_str
);
171 if (!git_parse_ssize_t(colon
, &max_tree_entry_len
))
172 die("unable to parse max tree entry len: %s", colon
);
175 msg_type
= parse_msg_type(msg_type_str
);
177 if (msg_type
!= FSCK_ERROR
&& msg_id_info
[msg_id
].msg_type
== FSCK_FATAL
)
178 die("Cannot demote %s to %s", msg_id_str
, msg_type_str
);
180 fsck_set_msg_type_from_ids(options
, msg_id
, msg_type
);
184 void fsck_set_msg_types(struct fsck_options
*options
, const char *values
)
186 char *buf
= xstrdup(values
), *to_free
= buf
;
190 int len
= strcspn(buf
, " ,|"), equal
;
200 equal
< len
&& buf
[equal
] != '=' && buf
[equal
] != ':';
202 buf
[equal
] = tolower(buf
[equal
]);
205 if (!strcmp(buf
, "skiplist")) {
207 die("skiplist requires a path");
208 oidset_parse_file(&options
->skiplist
, buf
+ equal
+ 1);
214 die("Missing '=': '%s'", buf
);
216 fsck_set_msg_type(options
, buf
, buf
+ equal
+ 1);
222 static int object_on_skiplist(struct fsck_options
*opts
,
223 const struct object_id
*oid
)
225 return opts
&& oid
&& oidset_contains(&opts
->skiplist
, oid
);
228 __attribute__((format (printf
, 5, 6)))
229 static int report(struct fsck_options
*options
,
230 const struct object_id
*oid
, enum object_type object_type
,
231 enum fsck_msg_id msg_id
, const char *fmt
, ...)
234 struct strbuf sb
= STRBUF_INIT
;
235 enum fsck_msg_type msg_type
= fsck_msg_type(msg_id
, options
);
238 if (msg_type
== FSCK_IGNORE
)
241 if (object_on_skiplist(options
, oid
))
244 if (msg_type
== FSCK_FATAL
)
245 msg_type
= FSCK_ERROR
;
246 else if (msg_type
== FSCK_INFO
)
247 msg_type
= FSCK_WARN
;
250 strbuf_addf(&sb
, "%s: ", msg_id_info
[msg_id
].camelcased
);
253 strbuf_vaddf(&sb
, fmt
, ap
);
254 result
= options
->error_func(options
, oid
, object_type
,
255 msg_type
, msg_id
, sb
.buf
);
262 void fsck_enable_object_names(struct fsck_options
*options
)
264 if (!options
->object_names
)
265 options
->object_names
= kh_init_oid_map();
268 const char *fsck_get_object_name(struct fsck_options
*options
,
269 const struct object_id
*oid
)
272 if (!options
->object_names
)
274 pos
= kh_get_oid_map(options
->object_names
, *oid
);
275 if (pos
>= kh_end(options
->object_names
))
277 return kh_value(options
->object_names
, pos
);
280 void fsck_put_object_name(struct fsck_options
*options
,
281 const struct object_id
*oid
,
282 const char *fmt
, ...)
285 struct strbuf buf
= STRBUF_INIT
;
289 if (!options
->object_names
)
292 pos
= kh_put_oid_map(options
->object_names
, *oid
, &hashret
);
296 strbuf_vaddf(&buf
, fmt
, ap
);
297 kh_value(options
->object_names
, pos
) = strbuf_detach(&buf
, NULL
);
301 const char *fsck_describe_object(struct fsck_options
*options
,
302 const struct object_id
*oid
)
304 static struct strbuf bufs
[] = {
305 STRBUF_INIT
, STRBUF_INIT
, STRBUF_INIT
, STRBUF_INIT
309 const char *name
= fsck_get_object_name(options
, oid
);
312 b
= (b
+ 1) % ARRAY_SIZE(bufs
);
314 strbuf_addstr(buf
, oid_to_hex(oid
));
316 strbuf_addf(buf
, " (%s)", name
);
321 static int fsck_walk_tree(struct tree
*tree
, void *data
, struct fsck_options
*options
)
323 struct tree_desc desc
;
324 struct name_entry entry
;
328 if (parse_tree(tree
))
331 name
= fsck_get_object_name(options
, &tree
->object
.oid
);
332 if (init_tree_desc_gently(&desc
, tree
->buffer
, tree
->size
, 0))
334 while (tree_entry_gently(&desc
, &entry
)) {
338 if (S_ISGITLINK(entry
.mode
))
341 if (S_ISDIR(entry
.mode
)) {
342 obj
= (struct object
*)lookup_tree(the_repository
, &entry
.oid
);
344 fsck_put_object_name(options
, &entry
.oid
, "%s%s/",
346 result
= options
->walk(obj
, OBJ_TREE
, data
, options
);
348 else if (S_ISREG(entry
.mode
) || S_ISLNK(entry
.mode
)) {
349 obj
= (struct object
*)lookup_blob(the_repository
, &entry
.oid
);
351 fsck_put_object_name(options
, &entry
.oid
, "%s%s",
353 result
= options
->walk(obj
, OBJ_BLOB
, data
, options
);
356 result
= error("in tree %s: entry %s has bad mode %.6o",
357 fsck_describe_object(options
, &tree
->object
.oid
),
358 entry
.path
, entry
.mode
);
368 static int fsck_walk_commit(struct commit
*commit
, void *data
, struct fsck_options
*options
)
370 int counter
= 0, generation
= 0, name_prefix_len
= 0;
371 struct commit_list
*parents
;
376 if (repo_parse_commit(the_repository
, commit
))
379 name
= fsck_get_object_name(options
, &commit
->object
.oid
);
381 fsck_put_object_name(options
, get_commit_tree_oid(commit
),
384 result
= options
->walk((struct object
*) repo_get_commit_tree(the_repository
, commit
),
385 OBJ_TREE
, data
, options
);
390 parents
= commit
->parents
;
391 if (name
&& parents
) {
392 int len
= strlen(name
), power
;
394 if (len
&& name
[len
- 1] == '^') {
396 name_prefix_len
= len
- 1;
398 else { /* parse ~<generation> suffix */
399 for (generation
= 0, power
= 1;
400 len
&& isdigit(name
[len
- 1]);
402 generation
+= power
* (name
[--len
] - '0');
403 if (power
> 1 && len
&& name
[len
- 1] == '~')
404 name_prefix_len
= len
- 1;
406 /* Maybe a non-first parent, e.g. HEAD^2 */
408 name_prefix_len
= len
;
415 struct object_id
*oid
= &parents
->item
->object
.oid
;
418 fsck_put_object_name(options
, oid
, "%s^%d",
420 else if (generation
> 0)
421 fsck_put_object_name(options
, oid
, "%.*s~%d",
422 name_prefix_len
, name
,
425 fsck_put_object_name(options
, oid
, "%s^", name
);
427 result
= options
->walk((struct object
*)parents
->item
, OBJ_COMMIT
, data
, options
);
432 parents
= parents
->next
;
437 static int fsck_walk_tag(struct tag
*tag
, void *data
, struct fsck_options
*options
)
439 const char *name
= fsck_get_object_name(options
, &tag
->object
.oid
);
444 fsck_put_object_name(options
, &tag
->tagged
->oid
, "%s", name
);
445 return options
->walk(tag
->tagged
, OBJ_ANY
, data
, options
);
448 int fsck_walk(struct object
*obj
, void *data
, struct fsck_options
*options
)
453 if (obj
->type
== OBJ_NONE
)
454 parse_object(the_repository
, &obj
->oid
);
460 return fsck_walk_tree((struct tree
*)obj
, data
, options
);
462 return fsck_walk_commit((struct commit
*)obj
, data
, options
);
464 return fsck_walk_tag((struct tag
*)obj
, data
, options
);
466 error("Unknown object type for %s",
467 fsck_describe_object(options
, &obj
->oid
));
477 static void name_stack_push(struct name_stack
*stack
, const char *name
)
479 ALLOC_GROW(stack
->names
, stack
->nr
+ 1, stack
->alloc
);
480 stack
->names
[stack
->nr
++] = name
;
483 static const char *name_stack_pop(struct name_stack
*stack
)
485 return stack
->nr
? stack
->names
[--stack
->nr
] : NULL
;
488 static void name_stack_clear(struct name_stack
*stack
)
490 FREE_AND_NULL(stack
->names
);
491 stack
->nr
= stack
->alloc
= 0;
495 * The entries in a tree are ordered in the _path_ order,
496 * which means that a directory entry is ordered by adding
497 * a slash to the end of it.
499 * So a directory called "a" is ordered _after_ a file
500 * called "a.c", because "a/" sorts after "a.c".
502 #define TREE_UNORDERED (-1)
503 #define TREE_HAS_DUPS (-2)
505 static int is_less_than_slash(unsigned char c
)
507 return '\0' < c
&& c
< '/';
510 static int verify_ordered(unsigned mode1
, const char *name1
,
511 unsigned mode2
, const char *name2
,
512 struct name_stack
*candidates
)
514 int len1
= strlen(name1
);
515 int len2
= strlen(name2
);
516 int len
= len1
< len2
? len1
: len2
;
517 unsigned char c1
, c2
;
520 cmp
= memcmp(name1
, name2
, len
);
524 return TREE_UNORDERED
;
527 * Ok, the first <len> characters are the same.
528 * Now we need to order the next one, but turn
529 * a '\0' into a '/' for a directory entry.
535 * git-write-tree used to write out a nonsense tree that has
536 * entries with the same name, one blob and one tree. Make
537 * sure we do not have duplicate entries.
539 return TREE_HAS_DUPS
;
540 if (!c1
&& S_ISDIR(mode1
))
542 if (!c2
&& S_ISDIR(mode2
))
546 * There can be non-consecutive duplicates due to the implicitly
555 * Record non-directory candidates (like "foo" and "foo.bar" in
556 * the example) on a stack and check directory candidates (like
557 * foo/" and "foo.bar/") against that stack.
559 if (!c1
&& is_less_than_slash(c2
)) {
560 name_stack_push(candidates
, name1
);
561 } else if (c2
== '/' && is_less_than_slash(c1
)) {
564 const char *f_name
= name_stack_pop(candidates
);
568 if (!skip_prefix(name2
, f_name
, &p
))
571 return TREE_HAS_DUPS
;
572 if (is_less_than_slash(*p
)) {
573 name_stack_push(candidates
, f_name
);
579 return c1
< c2
? 0 : TREE_UNORDERED
;
582 static int fsck_tree(const struct object_id
*tree_oid
,
583 const char *buffer
, unsigned long size
,
584 struct fsck_options
*options
)
587 int has_null_sha1
= 0;
588 int has_full_path
= 0;
589 int has_empty_name
= 0;
593 int has_zero_pad
= 0;
594 int has_bad_modes
= 0;
595 int has_dup_entries
= 0;
596 int not_properly_sorted
= 0;
597 int has_large_name
= 0;
598 struct tree_desc desc
;
601 struct name_stack df_dup_candidates
= { NULL
};
603 if (init_tree_desc_gently(&desc
, buffer
, size
, 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");
663 if ((backslash
= strchr(name
, '\\'))) {
666 has_dotgit
|= is_ntfs_dotgit(backslash
);
667 if (is_ntfs_dotgitmodules(backslash
)) {
669 oidset_insert(&options
->gitmodules_found
,
672 retval
+= report(options
, tree_oid
, OBJ_TREE
,
673 FSCK_MSG_GITMODULES_SYMLINK
,
674 ".gitmodules is a symbolic link");
676 backslash
= strchr(backslash
, '\\');
680 if (update_tree_entry_gently(&desc
)) {
681 retval
+= report(options
, tree_oid
, OBJ_TREE
,
683 "cannot be parsed as a tree");
698 * This is nonstandard, but we had a few of these
699 * early on when we honored the full set of mode
703 if (!options
->strict
)
711 switch (verify_ordered(o_mode
, o_name
, mode
, name
,
712 &df_dup_candidates
)) {
714 not_properly_sorted
= 1;
728 name_stack_clear(&df_dup_candidates
);
731 retval
+= report(options
, tree_oid
, OBJ_TREE
,
733 "contains entries pointing to null sha1");
735 retval
+= report(options
, tree_oid
, OBJ_TREE
,
736 FSCK_MSG_FULL_PATHNAME
,
737 "contains full pathnames");
739 retval
+= report(options
, tree_oid
, OBJ_TREE
,
741 "contains empty pathname");
743 retval
+= report(options
, tree_oid
, OBJ_TREE
,
747 retval
+= report(options
, tree_oid
, OBJ_TREE
,
751 retval
+= report(options
, tree_oid
, OBJ_TREE
,
755 retval
+= report(options
, tree_oid
, OBJ_TREE
,
756 FSCK_MSG_ZERO_PADDED_FILEMODE
,
757 "contains zero-padded file modes");
759 retval
+= report(options
, tree_oid
, OBJ_TREE
,
760 FSCK_MSG_BAD_FILEMODE
,
761 "contains bad file modes");
763 retval
+= report(options
, tree_oid
, OBJ_TREE
,
764 FSCK_MSG_DUPLICATE_ENTRIES
,
765 "contains duplicate file entries");
766 if (not_properly_sorted
)
767 retval
+= report(options
, tree_oid
, OBJ_TREE
,
768 FSCK_MSG_TREE_NOT_SORTED
,
769 "not properly sorted");
771 retval
+= report(options
, tree_oid
, OBJ_TREE
,
772 FSCK_MSG_LARGE_PATHNAME
,
773 "contains excessively large pathname");
778 * Confirm that the headers of a commit or tag object end in a reasonable way,
779 * either with the usual "\n\n" separator, or at least with a trailing newline
780 * on the final header line.
782 * This property is important for the memory safety of our callers. It allows
783 * them to scan the buffer linewise without constantly checking the remaining
786 * - they check that there are bytes left in the buffer at the start of any
787 * line (i.e., that the last newline they saw was not the final one we
790 * - any intra-line scanning they do will stop at a newline, which will worst
791 * case hit the newline we found here as the end-of-header. This makes it
792 * OK for them to use helpers like parse_oid_hex(), or even skip_prefix().
794 static int verify_headers(const void *data
, unsigned long size
,
795 const struct object_id
*oid
, enum object_type type
,
796 struct fsck_options
*options
)
798 const char *buffer
= (const char *)data
;
801 for (i
= 0; i
< size
; i
++) {
804 return report(options
, oid
, type
,
805 FSCK_MSG_NUL_IN_HEADER
,
806 "unterminated header: NUL at offset %ld", i
);
808 if (i
+ 1 < size
&& buffer
[i
+ 1] == '\n')
814 * We did not find double-LF that separates the header
815 * and the body. Not having a body is not a crime but
816 * we do want to see the terminating LF for the last header
819 if (size
&& buffer
[size
- 1] == '\n')
822 return report(options
, oid
, type
,
823 FSCK_MSG_UNTERMINATED_HEADER
, "unterminated header");
826 static int fsck_ident(const char **ident
,
827 const struct object_id
*oid
, enum object_type type
,
828 struct fsck_options
*options
)
830 const char *p
= *ident
;
833 *ident
= strchrnul(*ident
, '\n');
838 return report(options
, oid
, type
, FSCK_MSG_MISSING_NAME_BEFORE_EMAIL
, "invalid author/committer line - missing space before email");
839 p
+= strcspn(p
, "<>\n");
841 return report(options
, oid
, type
, FSCK_MSG_BAD_NAME
, "invalid author/committer line - bad name");
843 return report(options
, oid
, type
, FSCK_MSG_MISSING_EMAIL
, "invalid author/committer line - missing email");
845 return report(options
, oid
, type
, FSCK_MSG_MISSING_SPACE_BEFORE_EMAIL
, "invalid author/committer line - missing space before email");
847 p
+= strcspn(p
, "<>\n");
849 return report(options
, oid
, type
, FSCK_MSG_BAD_EMAIL
, "invalid author/committer line - bad email");
852 return report(options
, oid
, type
, FSCK_MSG_MISSING_SPACE_BEFORE_DATE
, "invalid author/committer line - missing space before date");
855 * Our timestamp parser is based on the C strto*() functions, which
856 * will happily eat whitespace, including the newline that is supposed
857 * to prevent us walking past the end of the buffer. So do our own
858 * scan, skipping linear whitespace but not newlines, and then
859 * confirming we found a digit. We _could_ be even more strict here,
860 * as we really expect only a single space, but since we have
861 * traditionally allowed extra whitespace, we'll continue to do so.
863 while (*p
== ' ' || *p
== '\t')
866 return report(options
, oid
, type
, FSCK_MSG_BAD_DATE
,
867 "invalid author/committer line - bad date");
868 if (*p
== '0' && p
[1] != ' ')
869 return report(options
, oid
, type
, FSCK_MSG_ZERO_PADDED_DATE
, "invalid author/committer line - zero-padded date");
870 if (date_overflows(parse_timestamp(p
, &end
, 10)))
871 return report(options
, oid
, type
, FSCK_MSG_BAD_DATE_OVERFLOW
, "invalid author/committer line - date causes integer overflow");
872 if ((end
== p
|| *end
!= ' '))
873 return report(options
, oid
, type
, FSCK_MSG_BAD_DATE
, "invalid author/committer line - bad date");
875 if ((*p
!= '+' && *p
!= '-') ||
881 return report(options
, oid
, type
, FSCK_MSG_BAD_TIMEZONE
, "invalid author/committer line - bad time zone");
886 static int fsck_commit(const struct object_id
*oid
,
887 const char *buffer
, unsigned long size
,
888 struct fsck_options
*options
)
890 struct object_id tree_oid
, parent_oid
;
891 unsigned author_count
;
893 const char *buffer_begin
= buffer
;
894 const char *buffer_end
= buffer
+ size
;
898 * We _must_ stop parsing immediately if this reports failure, as the
899 * memory safety of the rest of the function depends on it. See the
900 * comment above the definition of verify_headers() for more details.
902 if (verify_headers(buffer
, size
, oid
, OBJ_COMMIT
, options
))
905 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "tree ", &buffer
))
906 return report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MISSING_TREE
, "invalid format - expected 'tree' line");
907 if (parse_oid_hex(buffer
, &tree_oid
, &p
) || *p
!= '\n') {
908 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_BAD_TREE_SHA1
, "invalid 'tree' line format - bad sha1");
913 while (buffer
< buffer_end
&& skip_prefix(buffer
, "parent ", &buffer
)) {
914 if (parse_oid_hex(buffer
, &parent_oid
, &p
) || *p
!= '\n') {
915 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_BAD_PARENT_SHA1
, "invalid 'parent' line format - bad sha1");
922 while (buffer
< buffer_end
&& skip_prefix(buffer
, "author ", &buffer
)) {
924 err
= fsck_ident(&buffer
, oid
, OBJ_COMMIT
, options
);
928 if (author_count
< 1)
929 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MISSING_AUTHOR
, "invalid format - expected 'author' line");
930 else if (author_count
> 1)
931 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MULTIPLE_AUTHORS
, "invalid format - multiple 'author' lines");
934 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "committer ", &buffer
))
935 return report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_MISSING_COMMITTER
, "invalid format - expected 'committer' line");
936 err
= fsck_ident(&buffer
, oid
, OBJ_COMMIT
, options
);
939 if (memchr(buffer_begin
, '\0', size
)) {
940 err
= report(options
, oid
, OBJ_COMMIT
, FSCK_MSG_NUL_IN_COMMIT
,
941 "NUL byte in the commit object body");
948 static int fsck_tag(const struct object_id
*oid
, const char *buffer
,
949 unsigned long size
, struct fsck_options
*options
)
951 struct object_id tagged_oid
;
953 return fsck_tag_standalone(oid
, buffer
, size
, options
, &tagged_oid
,
957 int fsck_tag_standalone(const struct object_id
*oid
, const char *buffer
,
958 unsigned long size
, struct fsck_options
*options
,
959 struct object_id
*tagged_oid
,
964 struct strbuf sb
= STRBUF_INIT
;
965 const char *buffer_end
= buffer
+ size
;
969 * We _must_ stop parsing immediately if this reports failure, as the
970 * memory safety of the rest of the function depends on it. See the
971 * comment above the definition of verify_headers() for more details.
973 ret
= verify_headers(buffer
, size
, oid
, OBJ_TAG
, options
);
977 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "object ", &buffer
)) {
978 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_OBJECT
, "invalid format - expected 'object' line");
981 if (parse_oid_hex(buffer
, tagged_oid
, &p
) || *p
!= '\n') {
982 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_BAD_OBJECT_SHA1
, "invalid 'object' line format - bad sha1");
988 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "type ", &buffer
)) {
989 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TYPE_ENTRY
, "invalid format - expected 'type' line");
992 eol
= memchr(buffer
, '\n', buffer_end
- buffer
);
994 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TYPE
, "invalid format - unexpected end after 'type' line");
997 *tagged_type
= type_from_string_gently(buffer
, eol
- buffer
, 1);
998 if (*tagged_type
< 0)
999 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_BAD_TYPE
, "invalid 'type' value");
1004 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "tag ", &buffer
)) {
1005 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TAG_ENTRY
, "invalid format - expected 'tag' line");
1008 eol
= memchr(buffer
, '\n', buffer_end
- buffer
);
1010 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TAG
, "invalid format - unexpected end after 'type' line");
1013 strbuf_addf(&sb
, "refs/tags/%.*s", (int)(eol
- buffer
), buffer
);
1014 if (check_refname_format(sb
.buf
, 0)) {
1015 ret
= report(options
, oid
, OBJ_TAG
,
1016 FSCK_MSG_BAD_TAG_NAME
,
1017 "invalid 'tag' name: %.*s",
1018 (int)(eol
- buffer
), buffer
);
1024 if (buffer
>= buffer_end
|| !skip_prefix(buffer
, "tagger ", &buffer
)) {
1025 /* early tags do not contain 'tagger' lines; warn only */
1026 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_MISSING_TAGGER_ENTRY
, "invalid format - expected 'tagger' line");
1031 ret
= fsck_ident(&buffer
, oid
, OBJ_TAG
, options
);
1033 if (buffer
< buffer_end
&& !starts_with(buffer
, "\n")) {
1035 * The verify_headers() check will allow
1036 * e.g. "[...]tagger <tagger>\nsome
1037 * garbage\n\nmessage" to pass, thinking "some
1038 * garbage" could be a custom header. E.g. "mktag"
1039 * doesn't want any unknown headers.
1041 ret
= report(options
, oid
, OBJ_TAG
, FSCK_MSG_EXTRA_HEADER_ENTRY
, "invalid format - extra header(s) after 'tagger'");
1047 strbuf_release(&sb
);
1051 static int starts_with_dot_slash(const char *const path
)
1053 return path_match_flags(path
, PATH_MATCH_STARTS_WITH_DOT_SLASH
|
1054 PATH_MATCH_XPLATFORM
);
1057 static int starts_with_dot_dot_slash(const char *const path
)
1059 return path_match_flags(path
, PATH_MATCH_STARTS_WITH_DOT_DOT_SLASH
|
1060 PATH_MATCH_XPLATFORM
);
1063 static int submodule_url_is_relative(const char *url
)
1065 return starts_with_dot_slash(url
) || starts_with_dot_dot_slash(url
);
1069 * Count directory components that a relative submodule URL should chop
1070 * from the remote_url it is to be resolved against.
1072 * In other words, this counts "../" components at the start of a
1075 * Returns the number of directory components to chop and writes a
1076 * pointer to the next character of url after all leading "./" and
1077 * "../" components to out.
1079 static int count_leading_dotdots(const char *url
, const char **out
)
1083 if (starts_with_dot_dot_slash(url
)) {
1085 url
+= strlen("../");
1088 if (starts_with_dot_slash(url
)) {
1089 url
+= strlen("./");
1097 * Check whether a transport is implemented by git-remote-curl.
1099 * If it is, returns 1 and writes the URL that would be passed to
1100 * git-remote-curl to the "out" parameter.
1102 * Otherwise, returns 0 and leaves "out" untouched.
1105 * http::https://example.com/repo.git -> 1, https://example.com/repo.git
1106 * https://example.com/repo.git -> 1, https://example.com/repo.git
1107 * git://example.com/repo.git -> 0
1109 * This is for use in checking for previously exploitable bugs that
1110 * required a submodule URL to be passed to git-remote-curl.
1112 static int url_to_curl_url(const char *url
, const char **out
)
1115 * We don't need to check for case-aliases, "http.exe", and so
1116 * on because in the default configuration, is_transport_allowed
1117 * prevents URLs with those schemes from being cloned
1120 if (skip_prefix(url
, "http::", out
) ||
1121 skip_prefix(url
, "https::", out
) ||
1122 skip_prefix(url
, "ftp::", out
) ||
1123 skip_prefix(url
, "ftps::", out
))
1125 if (starts_with(url
, "http://") ||
1126 starts_with(url
, "https://") ||
1127 starts_with(url
, "ftp://") ||
1128 starts_with(url
, "ftps://")) {
1135 static int check_submodule_url(const char *url
)
1137 const char *curl_url
;
1139 if (looks_like_command_line_option(url
))
1142 if (submodule_url_is_relative(url
) || starts_with(url
, "git://")) {
1148 * This could be appended to an http URL and url-decoded;
1149 * check for malicious characters.
1151 decoded
= url_decode(url
);
1152 has_nl
= !!strchr(decoded
, '\n');
1159 * URLs which escape their root via "../" can overwrite
1160 * the host field and previous components, resolving to
1161 * URLs like https::example.com/submodule.git and
1162 * https:///example.com/submodule.git that were
1163 * susceptible to CVE-2020-11008.
1165 if (count_leading_dotdots(url
, &next
) > 0 &&
1166 (*next
== ':' || *next
== '/'))
1170 else if (url_to_curl_url(url
, &curl_url
)) {
1171 struct credential c
= CREDENTIAL_INIT
;
1173 if (credential_from_url_gently(&c
, curl_url
, 1) ||
1176 credential_clear(&c
);
1183 struct fsck_gitmodules_data
{
1184 const struct object_id
*oid
;
1185 struct fsck_options
*options
;
1189 static int fsck_gitmodules_fn(const char *var
, const char *value
,
1190 const struct config_context
*ctx UNUSED
,
1193 struct fsck_gitmodules_data
*data
= vdata
;
1194 const char *subsection
, *key
;
1195 size_t subsection_len
;
1198 if (parse_config_key(var
, "submodule", &subsection
, &subsection_len
, &key
) < 0 ||
1202 name
= xmemdupz(subsection
, subsection_len
);
1203 if (check_submodule_name(name
) < 0)
1204 data
->ret
|= report(data
->options
,
1205 data
->oid
, OBJ_BLOB
,
1206 FSCK_MSG_GITMODULES_NAME
,
1207 "disallowed submodule name: %s",
1209 if (!strcmp(key
, "url") && value
&&
1210 check_submodule_url(value
) < 0)
1211 data
->ret
|= report(data
->options
,
1212 data
->oid
, OBJ_BLOB
,
1213 FSCK_MSG_GITMODULES_URL
,
1214 "disallowed submodule url: %s",
1216 if (!strcmp(key
, "path") && value
&&
1217 looks_like_command_line_option(value
))
1218 data
->ret
|= report(data
->options
,
1219 data
->oid
, OBJ_BLOB
,
1220 FSCK_MSG_GITMODULES_PATH
,
1221 "disallowed submodule path: %s",
1223 if (!strcmp(key
, "update") && value
&&
1224 parse_submodule_update_type(value
) == SM_UPDATE_COMMAND
)
1225 data
->ret
|= report(data
->options
, data
->oid
, OBJ_BLOB
,
1226 FSCK_MSG_GITMODULES_UPDATE
,
1227 "disallowed submodule update setting: %s",
1234 static int fsck_blob(const struct object_id
*oid
, const char *buf
,
1235 unsigned long size
, struct fsck_options
*options
)
1239 if (object_on_skiplist(options
, oid
))
1242 if (oidset_contains(&options
->gitmodules_found
, oid
)) {
1243 struct config_options config_opts
= { 0 };
1244 struct fsck_gitmodules_data data
;
1246 oidset_insert(&options
->gitmodules_done
, oid
);
1250 * A missing buffer here is a sign that the caller found the
1251 * blob too gigantic to load into memory. Let's just consider
1254 return report(options
, oid
, OBJ_BLOB
,
1255 FSCK_MSG_GITMODULES_LARGE
,
1256 ".gitmodules too large to parse");
1260 data
.options
= options
;
1262 config_opts
.error_action
= CONFIG_ERROR_SILENT
;
1263 if (git_config_from_mem(fsck_gitmodules_fn
, CONFIG_ORIGIN_BLOB
,
1264 ".gitmodules", buf
, size
, &data
,
1265 CONFIG_SCOPE_UNKNOWN
, &config_opts
))
1266 data
.ret
|= report(options
, oid
, OBJ_BLOB
,
1267 FSCK_MSG_GITMODULES_PARSE
,
1268 "could not parse gitmodules blob");
1272 if (oidset_contains(&options
->gitattributes_found
, oid
)) {
1275 oidset_insert(&options
->gitattributes_done
, oid
);
1277 if (!buf
|| size
> ATTR_MAX_FILE_SIZE
) {
1279 * A missing buffer here is a sign that the caller found the
1280 * blob too gigantic to load into memory. Let's just consider
1283 return report(options
, oid
, OBJ_BLOB
,
1284 FSCK_MSG_GITATTRIBUTES_LARGE
,
1285 ".gitattributes too large to parse");
1288 for (ptr
= buf
; *ptr
; ) {
1289 const char *eol
= strchrnul(ptr
, '\n');
1290 if (eol
- ptr
>= ATTR_MAX_LINE_LENGTH
) {
1291 ret
|= report(options
, oid
, OBJ_BLOB
,
1292 FSCK_MSG_GITATTRIBUTES_LINE_LENGTH
,
1293 ".gitattributes has too long lines to parse");
1297 ptr
= *eol
? eol
+ 1 : eol
;
1304 int fsck_object(struct object
*obj
, void *data
, unsigned long size
,
1305 struct fsck_options
*options
)
1308 return report(options
, NULL
, OBJ_NONE
, FSCK_MSG_BAD_OBJECT_SHA1
, "no valid object to fsck");
1310 return fsck_buffer(&obj
->oid
, obj
->type
, data
, size
, options
);
1313 int fsck_buffer(const struct object_id
*oid
, enum object_type type
,
1314 void *data
, unsigned long size
,
1315 struct fsck_options
*options
)
1317 if (type
== OBJ_BLOB
)
1318 return fsck_blob(oid
, data
, size
, options
);
1319 if (type
== OBJ_TREE
)
1320 return fsck_tree(oid
, data
, size
, options
);
1321 if (type
== OBJ_COMMIT
)
1322 return fsck_commit(oid
, data
, size
, options
);
1323 if (type
== OBJ_TAG
)
1324 return fsck_tag(oid
, data
, size
, options
);
1326 return report(options
, oid
, type
,
1327 FSCK_MSG_UNKNOWN_TYPE
,
1328 "unknown type '%d' (internal fsck error)",
1332 int fsck_error_function(struct fsck_options
*o
,
1333 const struct object_id
*oid
,
1334 enum object_type object_type UNUSED
,
1335 enum fsck_msg_type msg_type
,
1336 enum fsck_msg_id msg_id UNUSED
,
1337 const char *message
)
1339 if (msg_type
== FSCK_WARN
) {
1340 warning("object %s: %s", fsck_describe_object(o
, oid
), message
);
1343 error("object %s: %s", fsck_describe_object(o
, oid
), message
);
1347 static int fsck_blobs(struct oidset
*blobs_found
, struct oidset
*blobs_done
,
1348 enum fsck_msg_id msg_missing
, enum fsck_msg_id msg_type
,
1349 struct fsck_options
*options
, const char *blob_type
)
1352 struct oidset_iter iter
;
1353 const struct object_id
*oid
;
1355 oidset_iter_init(blobs_found
, &iter
);
1356 while ((oid
= oidset_iter_next(&iter
))) {
1357 enum object_type type
;
1361 if (oidset_contains(blobs_done
, oid
))
1364 buf
= repo_read_object_file(the_repository
, oid
, &type
, &size
);
1366 if (is_promisor_object(oid
))
1368 ret
|= report(options
,
1369 oid
, OBJ_BLOB
, msg_missing
,
1370 "unable to read %s blob", blob_type
);
1374 if (type
== OBJ_BLOB
)
1375 ret
|= fsck_blob(oid
, buf
, size
, options
);
1377 ret
|= report(options
, oid
, type
, msg_type
,
1378 "non-blob found at %s", blob_type
);
1382 oidset_clear(blobs_found
);
1383 oidset_clear(blobs_done
);
1388 int fsck_finish(struct fsck_options
*options
)
1392 ret
|= fsck_blobs(&options
->gitmodules_found
, &options
->gitmodules_done
,
1393 FSCK_MSG_GITMODULES_MISSING
, FSCK_MSG_GITMODULES_BLOB
,
1394 options
, ".gitmodules");
1395 ret
|= fsck_blobs(&options
->gitattributes_found
, &options
->gitattributes_done
,
1396 FSCK_MSG_GITATTRIBUTES_MISSING
, FSCK_MSG_GITATTRIBUTES_BLOB
,
1397 options
, ".gitattributes");
1402 int git_fsck_config(const char *var
, const char *value
,
1403 const struct config_context
*ctx
, void *cb
)
1405 struct fsck_options
*options
= cb
;
1408 if (strcmp(var
, "fsck.skiplist") == 0) {
1410 struct strbuf sb
= STRBUF_INIT
;
1412 if (git_config_pathname(&path
, var
, value
))
1414 strbuf_addf(&sb
, "skiplist=%s", path
);
1416 fsck_set_msg_types(options
, sb
.buf
);
1417 strbuf_release(&sb
);
1421 if (skip_prefix(var
, "fsck.", &msg_id
)) {
1423 return config_error_nonbool(var
);
1424 fsck_set_msg_type(options
, msg_id
, value
);
1428 return git_default_config(var
, value
, ctx
, cb
);
1432 * Custom error callbacks that are used in more than one place.
1435 int fsck_error_cb_print_missing_gitmodules(struct fsck_options
*o
,
1436 const struct object_id
*oid
,
1437 enum object_type object_type
,
1438 enum fsck_msg_type msg_type
,
1439 enum fsck_msg_id msg_id
,
1440 const char *message
)
1442 if (msg_id
== FSCK_MSG_GITMODULES_MISSING
) {
1443 puts(oid_to_hex(oid
));
1446 return fsck_error_function(o
, oid
, object_type
, msg_type
, msg_id
, message
);