3 * Copyright (C) 2011 Novell Inc.
5 * This program is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 as published by
7 * the Free Software Foundation.
10 #include <uapi/linux/magic.h>
12 #include <linux/namei.h>
13 #include <linux/xattr.h>
14 #include <linux/mount.h>
15 #include <linux/parser.h>
16 #include <linux/module.h>
17 #include <linux/statfs.h>
18 #include <linux/seq_file.h>
19 #include <linux/posix_acl_xattr.h>
20 #include "overlayfs.h"
22 MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>");
23 MODULE_DESCRIPTION("Overlay filesystem");
24 MODULE_LICENSE("GPL");
29 #define OVL_MAX_STACK 500
31 static bool ovl_redirect_dir_def
= IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_DIR
);
32 module_param_named(redirect_dir
, ovl_redirect_dir_def
, bool, 0644);
33 MODULE_PARM_DESC(ovl_redirect_dir_def
,
34 "Default to on or off for the redirect_dir feature");
36 static bool ovl_redirect_always_follow
=
37 IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW
);
38 module_param_named(redirect_always_follow
, ovl_redirect_always_follow
,
40 MODULE_PARM_DESC(ovl_redirect_always_follow
,
41 "Follow redirects even if redirect_dir feature is turned off");
43 static bool ovl_index_def
= IS_ENABLED(CONFIG_OVERLAY_FS_INDEX
);
44 module_param_named(index
, ovl_index_def
, bool, 0644);
45 MODULE_PARM_DESC(ovl_index_def
,
46 "Default to on or off for the inodes index feature");
48 static void ovl_entry_stack_free(struct ovl_entry
*oe
)
52 for (i
= 0; i
< oe
->numlower
; i
++)
53 dput(oe
->lowerstack
[i
].dentry
);
56 static void ovl_dentry_release(struct dentry
*dentry
)
58 struct ovl_entry
*oe
= dentry
->d_fsdata
;
61 ovl_entry_stack_free(oe
);
66 static int ovl_check_append_only(struct inode
*inode
, int flag
)
69 * This test was moot in vfs may_open() because overlay inode does
70 * not have the S_APPEND flag, so re-check on real upper inode
72 if (IS_APPEND(inode
)) {
73 if ((flag
& O_ACCMODE
) != O_RDONLY
&& !(flag
& O_APPEND
))
82 static struct dentry
*ovl_d_real(struct dentry
*dentry
,
83 const struct inode
*inode
,
84 unsigned int open_flags
, unsigned int flags
)
89 if (flags
& D_REAL_UPPER
)
90 return ovl_dentry_upper(dentry
);
92 if (!d_is_reg(dentry
)) {
93 if (!inode
|| inode
== d_inode(dentry
))
99 err
= ovl_open_maybe_copy_up(dentry
, open_flags
);
104 real
= ovl_dentry_upper(dentry
);
105 if (real
&& (!inode
|| inode
== d_inode(real
))) {
107 err
= ovl_check_append_only(d_inode(real
), open_flags
);
114 real
= ovl_dentry_lower(dentry
);
118 /* Handle recursion */
119 real
= d_real(real
, inode
, open_flags
, 0);
121 if (!inode
|| inode
== d_inode(real
))
124 WARN(1, "ovl_d_real(%pd4, %s:%lu): real dentry not found\n", dentry
,
125 inode
? inode
->i_sb
->s_id
: "NULL", inode
? inode
->i_ino
: 0);
129 static int ovl_dentry_revalidate(struct dentry
*dentry
, unsigned int flags
)
131 struct ovl_entry
*oe
= dentry
->d_fsdata
;
135 for (i
= 0; i
< oe
->numlower
; i
++) {
136 struct dentry
*d
= oe
->lowerstack
[i
].dentry
;
138 if (d
->d_flags
& DCACHE_OP_REVALIDATE
) {
139 ret
= d
->d_op
->d_revalidate(d
, flags
);
143 if (!(flags
& LOOKUP_RCU
))
152 static int ovl_dentry_weak_revalidate(struct dentry
*dentry
, unsigned int flags
)
154 struct ovl_entry
*oe
= dentry
->d_fsdata
;
158 for (i
= 0; i
< oe
->numlower
; i
++) {
159 struct dentry
*d
= oe
->lowerstack
[i
].dentry
;
161 if (d
->d_flags
& DCACHE_OP_WEAK_REVALIDATE
) {
162 ret
= d
->d_op
->d_weak_revalidate(d
, flags
);
170 static const struct dentry_operations ovl_dentry_operations
= {
171 .d_release
= ovl_dentry_release
,
172 .d_real
= ovl_d_real
,
175 static const struct dentry_operations ovl_reval_dentry_operations
= {
176 .d_release
= ovl_dentry_release
,
177 .d_real
= ovl_d_real
,
178 .d_revalidate
= ovl_dentry_revalidate
,
179 .d_weak_revalidate
= ovl_dentry_weak_revalidate
,
182 static struct kmem_cache
*ovl_inode_cachep
;
184 static struct inode
*ovl_alloc_inode(struct super_block
*sb
)
186 struct ovl_inode
*oi
= kmem_cache_alloc(ovl_inode_cachep
, GFP_KERNEL
);
195 oi
->__upperdentry
= NULL
;
197 mutex_init(&oi
->lock
);
199 return &oi
->vfs_inode
;
202 static void ovl_i_callback(struct rcu_head
*head
)
204 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
206 kmem_cache_free(ovl_inode_cachep
, OVL_I(inode
));
209 static void ovl_destroy_inode(struct inode
*inode
)
211 struct ovl_inode
*oi
= OVL_I(inode
);
213 dput(oi
->__upperdentry
);
215 ovl_dir_cache_free(inode
);
216 mutex_destroy(&oi
->lock
);
218 call_rcu(&inode
->i_rcu
, ovl_i_callback
);
221 static void ovl_free_fs(struct ovl_fs
*ofs
)
227 if (ofs
->workdir_locked
)
228 ovl_inuse_unlock(ofs
->workbasedir
);
229 dput(ofs
->workbasedir
);
230 if (ofs
->upperdir_locked
)
231 ovl_inuse_unlock(ofs
->upper_mnt
->mnt_root
);
232 mntput(ofs
->upper_mnt
);
233 for (i
= 0; i
< ofs
->numlower
; i
++) {
234 mntput(ofs
->lower_layers
[i
].mnt
);
235 free_anon_bdev(ofs
->lower_layers
[i
].pseudo_dev
);
237 kfree(ofs
->lower_layers
);
239 kfree(ofs
->config
.lowerdir
);
240 kfree(ofs
->config
.upperdir
);
241 kfree(ofs
->config
.workdir
);
242 kfree(ofs
->config
.redirect_mode
);
243 if (ofs
->creator_cred
)
244 put_cred(ofs
->creator_cred
);
248 static void ovl_put_super(struct super_block
*sb
)
250 struct ovl_fs
*ofs
= sb
->s_fs_info
;
255 /* Sync real dirty inodes in upper filesystem (if it exists) */
256 static int ovl_sync_fs(struct super_block
*sb
, int wait
)
258 struct ovl_fs
*ofs
= sb
->s_fs_info
;
259 struct super_block
*upper_sb
;
266 * If this is a sync(2) call or an emergency sync, all the super blocks
267 * will be iterated, including upper_sb, so no need to do anything.
269 * If this is a syncfs(2) call, then we do need to call
270 * sync_filesystem() on upper_sb, but enough if we do it when being
271 * called with wait == 1.
276 upper_sb
= ofs
->upper_mnt
->mnt_sb
;
278 down_read(&upper_sb
->s_umount
);
279 ret
= sync_filesystem(upper_sb
);
280 up_read(&upper_sb
->s_umount
);
287 * @sb: The overlayfs super block
288 * @buf: The struct kstatfs to fill in with stats
290 * Get the filesystem statistics. As writes always target the upper layer
291 * filesystem pass the statfs to the upper filesystem (if it exists)
293 static int ovl_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
295 struct ovl_fs
*ofs
= dentry
->d_sb
->s_fs_info
;
296 struct dentry
*root_dentry
= dentry
->d_sb
->s_root
;
300 ovl_path_real(root_dentry
, &path
);
302 err
= vfs_statfs(&path
, buf
);
304 buf
->f_namelen
= ofs
->namelen
;
305 buf
->f_type
= OVERLAYFS_SUPER_MAGIC
;
311 /* Will this overlay be forced to mount/remount ro? */
312 static bool ovl_force_readonly(struct ovl_fs
*ofs
)
314 return (!ofs
->upper_mnt
|| !ofs
->workdir
);
317 static const char *ovl_redirect_mode_def(void)
319 return ovl_redirect_dir_def
? "on" : "off";
325 * Prints the mount options for a given superblock.
326 * Returns zero; does not fail.
328 static int ovl_show_options(struct seq_file
*m
, struct dentry
*dentry
)
330 struct super_block
*sb
= dentry
->d_sb
;
331 struct ovl_fs
*ofs
= sb
->s_fs_info
;
333 seq_show_option(m
, "lowerdir", ofs
->config
.lowerdir
);
334 if (ofs
->config
.upperdir
) {
335 seq_show_option(m
, "upperdir", ofs
->config
.upperdir
);
336 seq_show_option(m
, "workdir", ofs
->config
.workdir
);
338 if (ofs
->config
.default_permissions
)
339 seq_puts(m
, ",default_permissions");
340 if (strcmp(ofs
->config
.redirect_mode
, ovl_redirect_mode_def()) != 0)
341 seq_printf(m
, ",redirect_dir=%s", ofs
->config
.redirect_mode
);
342 if (ofs
->config
.index
!= ovl_index_def
)
343 seq_printf(m
, ",index=%s", ofs
->config
.index
? "on" : "off");
347 static int ovl_remount(struct super_block
*sb
, int *flags
, char *data
)
349 struct ovl_fs
*ofs
= sb
->s_fs_info
;
351 if (!(*flags
& SB_RDONLY
) && ovl_force_readonly(ofs
))
357 static const struct super_operations ovl_super_operations
= {
358 .alloc_inode
= ovl_alloc_inode
,
359 .destroy_inode
= ovl_destroy_inode
,
360 .drop_inode
= generic_delete_inode
,
361 .put_super
= ovl_put_super
,
362 .sync_fs
= ovl_sync_fs
,
363 .statfs
= ovl_statfs
,
364 .show_options
= ovl_show_options
,
365 .remount_fs
= ovl_remount
,
372 OPT_DEFAULT_PERMISSIONS
,
379 static const match_table_t ovl_tokens
= {
380 {OPT_LOWERDIR
, "lowerdir=%s"},
381 {OPT_UPPERDIR
, "upperdir=%s"},
382 {OPT_WORKDIR
, "workdir=%s"},
383 {OPT_DEFAULT_PERMISSIONS
, "default_permissions"},
384 {OPT_REDIRECT_DIR
, "redirect_dir=%s"},
385 {OPT_INDEX_ON
, "index=on"},
386 {OPT_INDEX_OFF
, "index=off"},
390 static char *ovl_next_opt(char **s
)
398 for (p
= sbegin
; *p
; p
++) {
403 } else if (*p
== ',') {
413 static int ovl_parse_redirect_mode(struct ovl_config
*config
, const char *mode
)
415 if (strcmp(mode
, "on") == 0) {
416 config
->redirect_dir
= true;
418 * Does not make sense to have redirect creation without
419 * redirect following.
421 config
->redirect_follow
= true;
422 } else if (strcmp(mode
, "follow") == 0) {
423 config
->redirect_follow
= true;
424 } else if (strcmp(mode
, "off") == 0) {
425 if (ovl_redirect_always_follow
)
426 config
->redirect_follow
= true;
427 } else if (strcmp(mode
, "nofollow") != 0) {
428 pr_err("overlayfs: bad mount option \"redirect_dir=%s\"\n",
436 static int ovl_parse_opt(char *opt
, struct ovl_config
*config
)
440 config
->redirect_mode
= kstrdup(ovl_redirect_mode_def(), GFP_KERNEL
);
441 if (!config
->redirect_mode
)
444 while ((p
= ovl_next_opt(&opt
)) != NULL
) {
446 substring_t args
[MAX_OPT_ARGS
];
451 token
= match_token(p
, ovl_tokens
, args
);
454 kfree(config
->upperdir
);
455 config
->upperdir
= match_strdup(&args
[0]);
456 if (!config
->upperdir
)
461 kfree(config
->lowerdir
);
462 config
->lowerdir
= match_strdup(&args
[0]);
463 if (!config
->lowerdir
)
468 kfree(config
->workdir
);
469 config
->workdir
= match_strdup(&args
[0]);
470 if (!config
->workdir
)
474 case OPT_DEFAULT_PERMISSIONS
:
475 config
->default_permissions
= true;
478 case OPT_REDIRECT_DIR
:
479 kfree(config
->redirect_mode
);
480 config
->redirect_mode
= match_strdup(&args
[0]);
481 if (!config
->redirect_mode
)
486 config
->index
= true;
490 config
->index
= false;
494 pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p
);
499 /* Workdir is useless in non-upper mount */
500 if (!config
->upperdir
&& config
->workdir
) {
501 pr_info("overlayfs: option \"workdir=%s\" is useless in a non-upper mount, ignore\n",
503 kfree(config
->workdir
);
504 config
->workdir
= NULL
;
507 return ovl_parse_redirect_mode(config
, config
->redirect_mode
);
510 #define OVL_WORKDIR_NAME "work"
511 #define OVL_INDEXDIR_NAME "index"
513 static struct dentry
*ovl_workdir_create(struct ovl_fs
*ofs
,
514 const char *name
, bool persist
)
516 struct inode
*dir
= ofs
->workbasedir
->d_inode
;
517 struct vfsmount
*mnt
= ofs
->upper_mnt
;
520 bool retried
= false;
523 err
= mnt_want_write(mnt
);
527 inode_lock_nested(dir
, I_MUTEX_PARENT
);
531 work
= lookup_one_len(name
, ofs
->workbasedir
, strlen(name
));
534 struct iattr attr
= {
535 .ia_valid
= ATTR_MODE
,
536 .ia_mode
= S_IFDIR
| 0,
548 ovl_workdir_cleanup(dir
, mnt
, work
, 0);
553 err
= ovl_create_real(dir
, work
,
554 &(struct cattr
){.mode
= S_IFDIR
| 0},
560 * Try to remove POSIX ACL xattrs from workdir. We are good if:
562 * a) success (there was a POSIX ACL xattr and was removed)
563 * b) -ENODATA (there was no POSIX ACL xattr)
564 * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported)
566 * There are various other error values that could effectively
567 * mean that the xattr doesn't exist (e.g. -ERANGE is returned
568 * if the xattr name is too long), but the set of filesystems
569 * allowed as upper are limited to "normal" ones, where checking
570 * for the above two errors is sufficient.
572 err
= vfs_removexattr(work
, XATTR_NAME_POSIX_ACL_DEFAULT
);
573 if (err
&& err
!= -ENODATA
&& err
!= -EOPNOTSUPP
)
576 err
= vfs_removexattr(work
, XATTR_NAME_POSIX_ACL_ACCESS
);
577 if (err
&& err
!= -ENODATA
&& err
!= -EOPNOTSUPP
)
580 /* Clear any inherited mode bits */
581 inode_lock(work
->d_inode
);
582 err
= notify_change(work
, &attr
, NULL
);
583 inode_unlock(work
->d_inode
);
600 pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n",
601 ofs
->config
.workdir
, name
, -err
);
606 static void ovl_unescape(char *s
)
619 static int ovl_mount_dir_noesc(const char *name
, struct path
*path
)
624 pr_err("overlayfs: empty lowerdir\n");
627 err
= kern_path(name
, LOOKUP_FOLLOW
, path
);
629 pr_err("overlayfs: failed to resolve '%s': %i\n", name
, err
);
633 if (ovl_dentry_weird(path
->dentry
)) {
634 pr_err("overlayfs: filesystem on '%s' not supported\n", name
);
637 if (!d_is_dir(path
->dentry
)) {
638 pr_err("overlayfs: '%s' not a directory\n", name
);
649 static int ovl_mount_dir(const char *name
, struct path
*path
)
652 char *tmp
= kstrdup(name
, GFP_KERNEL
);
656 err
= ovl_mount_dir_noesc(tmp
, path
);
659 if (ovl_dentry_remote(path
->dentry
)) {
660 pr_err("overlayfs: filesystem on '%s' not supported as upperdir\n",
670 static int ovl_check_namelen(struct path
*path
, struct ovl_fs
*ofs
,
673 struct kstatfs statfs
;
674 int err
= vfs_statfs(path
, &statfs
);
677 pr_err("overlayfs: statfs failed on '%s'\n", name
);
679 ofs
->namelen
= max(ofs
->namelen
, statfs
.f_namelen
);
684 static int ovl_lower_dir(const char *name
, struct path
*path
,
685 struct ovl_fs
*ofs
, int *stack_depth
, bool *remote
)
689 err
= ovl_mount_dir_noesc(name
, path
);
693 err
= ovl_check_namelen(path
, ofs
, name
);
697 *stack_depth
= max(*stack_depth
, path
->mnt
->mnt_sb
->s_stack_depth
);
699 if (ovl_dentry_remote(path
->dentry
))
703 * The inodes index feature needs to encode and decode file
704 * handles, so it requires that all layers support them.
706 if (ofs
->config
.index
&& !ovl_can_decode_fh(path
->dentry
->d_sb
)) {
707 ofs
->config
.index
= false;
708 pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off.\n", name
);
719 /* Workdir should not be subdir of upperdir and vice versa */
720 static bool ovl_workdir_ok(struct dentry
*workdir
, struct dentry
*upperdir
)
724 if (workdir
!= upperdir
) {
725 ok
= (lock_rename(workdir
, upperdir
) == NULL
);
726 unlock_rename(workdir
, upperdir
);
731 static unsigned int ovl_split_lowerdirs(char *str
)
733 unsigned int ctr
= 1;
736 for (s
= d
= str
;; s
++, d
++) {
739 } else if (*s
== ':') {
751 static int __maybe_unused
752 ovl_posix_acl_xattr_get(const struct xattr_handler
*handler
,
753 struct dentry
*dentry
, struct inode
*inode
,
754 const char *name
, void *buffer
, size_t size
)
756 return ovl_xattr_get(dentry
, inode
, handler
->name
, buffer
, size
);
759 static int __maybe_unused
760 ovl_posix_acl_xattr_set(const struct xattr_handler
*handler
,
761 struct dentry
*dentry
, struct inode
*inode
,
762 const char *name
, const void *value
,
763 size_t size
, int flags
)
765 struct dentry
*workdir
= ovl_workdir(dentry
);
766 struct inode
*realinode
= ovl_inode_real(inode
);
767 struct posix_acl
*acl
= NULL
;
770 /* Check that everything is OK before copy-up */
772 acl
= posix_acl_from_xattr(&init_user_ns
, value
, size
);
777 if (!IS_POSIXACL(d_inode(workdir
)))
778 goto out_acl_release
;
779 if (!realinode
->i_op
->set_acl
)
780 goto out_acl_release
;
781 if (handler
->flags
== ACL_TYPE_DEFAULT
&& !S_ISDIR(inode
->i_mode
)) {
782 err
= acl
? -EACCES
: 0;
783 goto out_acl_release
;
786 if (!inode_owner_or_capable(inode
))
787 goto out_acl_release
;
789 posix_acl_release(acl
);
792 * Check if sgid bit needs to be cleared (actual setacl operation will
793 * be done with mounter's capabilities and so that won't do it for us).
795 if (unlikely(inode
->i_mode
& S_ISGID
) &&
796 handler
->flags
== ACL_TYPE_ACCESS
&&
797 !in_group_p(inode
->i_gid
) &&
798 !capable_wrt_inode_uidgid(inode
, CAP_FSETID
)) {
799 struct iattr iattr
= { .ia_valid
= ATTR_KILL_SGID
};
801 err
= ovl_setattr(dentry
, &iattr
);
806 err
= ovl_xattr_set(dentry
, inode
, handler
->name
, value
, size
, flags
);
808 ovl_copyattr(ovl_inode_real(inode
), inode
);
813 posix_acl_release(acl
);
817 static int ovl_own_xattr_get(const struct xattr_handler
*handler
,
818 struct dentry
*dentry
, struct inode
*inode
,
819 const char *name
, void *buffer
, size_t size
)
824 static int ovl_own_xattr_set(const struct xattr_handler
*handler
,
825 struct dentry
*dentry
, struct inode
*inode
,
826 const char *name
, const void *value
,
827 size_t size
, int flags
)
832 static int ovl_other_xattr_get(const struct xattr_handler
*handler
,
833 struct dentry
*dentry
, struct inode
*inode
,
834 const char *name
, void *buffer
, size_t size
)
836 return ovl_xattr_get(dentry
, inode
, name
, buffer
, size
);
839 static int ovl_other_xattr_set(const struct xattr_handler
*handler
,
840 struct dentry
*dentry
, struct inode
*inode
,
841 const char *name
, const void *value
,
842 size_t size
, int flags
)
844 return ovl_xattr_set(dentry
, inode
, name
, value
, size
, flags
);
847 static const struct xattr_handler __maybe_unused
848 ovl_posix_acl_access_xattr_handler
= {
849 .name
= XATTR_NAME_POSIX_ACL_ACCESS
,
850 .flags
= ACL_TYPE_ACCESS
,
851 .get
= ovl_posix_acl_xattr_get
,
852 .set
= ovl_posix_acl_xattr_set
,
855 static const struct xattr_handler __maybe_unused
856 ovl_posix_acl_default_xattr_handler
= {
857 .name
= XATTR_NAME_POSIX_ACL_DEFAULT
,
858 .flags
= ACL_TYPE_DEFAULT
,
859 .get
= ovl_posix_acl_xattr_get
,
860 .set
= ovl_posix_acl_xattr_set
,
863 static const struct xattr_handler ovl_own_xattr_handler
= {
864 .prefix
= OVL_XATTR_PREFIX
,
865 .get
= ovl_own_xattr_get
,
866 .set
= ovl_own_xattr_set
,
869 static const struct xattr_handler ovl_other_xattr_handler
= {
870 .prefix
= "", /* catch all */
871 .get
= ovl_other_xattr_get
,
872 .set
= ovl_other_xattr_set
,
875 static const struct xattr_handler
*ovl_xattr_handlers
[] = {
876 #ifdef CONFIG_FS_POSIX_ACL
877 &ovl_posix_acl_access_xattr_handler
,
878 &ovl_posix_acl_default_xattr_handler
,
880 &ovl_own_xattr_handler
,
881 &ovl_other_xattr_handler
,
885 static int ovl_get_upper(struct ovl_fs
*ofs
, struct path
*upperpath
)
887 struct vfsmount
*upper_mnt
;
890 err
= ovl_mount_dir(ofs
->config
.upperdir
, upperpath
);
894 /* Upper fs should not be r/o */
895 if (sb_rdonly(upperpath
->mnt
->mnt_sb
)) {
896 pr_err("overlayfs: upper fs is r/o, try multi-lower layers mount\n");
901 err
= ovl_check_namelen(upperpath
, ofs
, ofs
->config
.upperdir
);
906 if (ovl_inuse_trylock(upperpath
->dentry
)) {
907 ofs
->upperdir_locked
= true;
908 } else if (ofs
->config
.index
) {
909 pr_err("overlayfs: upperdir is in-use by another mount, mount with '-o index=off' to override exclusive upperdir protection.\n");
912 pr_warn("overlayfs: upperdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
915 upper_mnt
= clone_private_mount(upperpath
);
916 err
= PTR_ERR(upper_mnt
);
917 if (IS_ERR(upper_mnt
)) {
918 pr_err("overlayfs: failed to clone upperpath\n");
922 /* Don't inherit atime flags */
923 upper_mnt
->mnt_flags
&= ~(MNT_NOATIME
| MNT_NODIRATIME
| MNT_RELATIME
);
924 ofs
->upper_mnt
= upper_mnt
;
930 static int ovl_make_workdir(struct ovl_fs
*ofs
, struct path
*workpath
)
935 ofs
->workdir
= ovl_workdir_create(ofs
, OVL_WORKDIR_NAME
, false);
940 * Upper should support d_type, else whiteouts are visible. Given
941 * workdir and upper are on same fs, we can do iterate_dir() on
942 * workdir. This check requires successful creation of workdir in
945 err
= ovl_check_d_type_supported(workpath
);
950 * We allowed this configuration and don't want to break users over
951 * kernel upgrade. So warn instead of erroring out.
954 pr_warn("overlayfs: upper fs needs to support d_type.\n");
956 /* Check if upper/work fs supports O_TMPFILE */
957 temp
= ovl_do_tmpfile(ofs
->workdir
, S_IFREG
| 0);
958 ofs
->tmpfile
= !IS_ERR(temp
);
962 pr_warn("overlayfs: upper fs does not support tmpfile.\n");
965 * Check if upper/work fs supports trusted.overlay.* xattr
967 err
= ovl_do_setxattr(ofs
->workdir
, OVL_XATTR_OPAQUE
, "0", 1, 0);
970 pr_warn("overlayfs: upper fs does not support xattr.\n");
972 vfs_removexattr(ofs
->workdir
, OVL_XATTR_OPAQUE
);
975 /* Check if upper/work fs supports file handles */
976 if (ofs
->config
.index
&&
977 !ovl_can_decode_fh(ofs
->workdir
->d_sb
)) {
978 ofs
->config
.index
= false;
979 pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n");
985 static int ovl_get_workdir(struct ovl_fs
*ofs
, struct path
*upperpath
)
988 struct path workpath
= { };
990 err
= ovl_mount_dir(ofs
->config
.workdir
, &workpath
);
995 if (upperpath
->mnt
!= workpath
.mnt
) {
996 pr_err("overlayfs: workdir and upperdir must reside under the same mount\n");
999 if (!ovl_workdir_ok(workpath
.dentry
, upperpath
->dentry
)) {
1000 pr_err("overlayfs: workdir and upperdir must be separate subtrees\n");
1005 if (ovl_inuse_trylock(workpath
.dentry
)) {
1006 ofs
->workdir_locked
= true;
1007 } else if (ofs
->config
.index
) {
1008 pr_err("overlayfs: workdir is in-use by another mount, mount with '-o index=off' to override exclusive workdir protection.\n");
1011 pr_warn("overlayfs: workdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
1014 ofs
->workbasedir
= dget(workpath
.dentry
);
1015 err
= ovl_make_workdir(ofs
, &workpath
);
1021 path_put(&workpath
);
1026 static int ovl_get_indexdir(struct ovl_fs
*ofs
, struct ovl_entry
*oe
,
1027 struct path
*upperpath
)
1031 /* Verify lower root is upper root origin */
1032 err
= ovl_verify_origin(upperpath
->dentry
, oe
->lowerstack
[0].dentry
,
1035 pr_err("overlayfs: failed to verify upper root origin\n");
1039 ofs
->indexdir
= ovl_workdir_create(ofs
, OVL_INDEXDIR_NAME
, true);
1040 if (ofs
->indexdir
) {
1041 /* Verify upper root is index dir origin */
1042 err
= ovl_verify_origin(ofs
->indexdir
, upperpath
->dentry
,
1045 pr_err("overlayfs: failed to verify index dir origin\n");
1047 /* Cleanup bad/stale/orphan index entries */
1049 err
= ovl_indexdir_cleanup(ofs
->indexdir
,
1054 if (err
|| !ofs
->indexdir
)
1055 pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
1061 static int ovl_get_lower_layers(struct ovl_fs
*ofs
, struct path
*stack
,
1062 unsigned int numlower
)
1068 ofs
->lower_layers
= kcalloc(numlower
, sizeof(struct ovl_layer
),
1070 if (ofs
->lower_layers
== NULL
)
1072 for (i
= 0; i
< numlower
; i
++) {
1073 struct vfsmount
*mnt
;
1076 err
= get_anon_bdev(&dev
);
1078 pr_err("overlayfs: failed to get anonymous bdev for lowerpath\n");
1082 mnt
= clone_private_mount(&stack
[i
]);
1085 pr_err("overlayfs: failed to clone lowerpath\n");
1086 free_anon_bdev(dev
);
1090 * Make lower layers R/O. That way fchmod/fchown on lower file
1091 * will fail instead of modifying lower fs.
1093 mnt
->mnt_flags
|= MNT_READONLY
| MNT_NOATIME
;
1095 ofs
->lower_layers
[ofs
->numlower
].mnt
= mnt
;
1096 ofs
->lower_layers
[ofs
->numlower
].pseudo_dev
= dev
;
1099 /* Check if all lower layers are on same sb */
1101 ofs
->same_sb
= mnt
->mnt_sb
;
1102 else if (ofs
->same_sb
!= mnt
->mnt_sb
)
1103 ofs
->same_sb
= NULL
;
1110 static struct ovl_entry
*ovl_get_lowerstack(struct super_block
*sb
,
1114 char *lowertmp
, *lower
;
1115 struct path
*stack
= NULL
;
1116 unsigned int stacklen
, numlower
= 0, i
;
1117 bool remote
= false;
1118 struct ovl_entry
*oe
;
1121 lowertmp
= kstrdup(ofs
->config
.lowerdir
, GFP_KERNEL
);
1126 stacklen
= ovl_split_lowerdirs(lowertmp
);
1127 if (stacklen
> OVL_MAX_STACK
) {
1128 pr_err("overlayfs: too many lower directories, limit is %d\n",
1131 } else if (!ofs
->config
.upperdir
&& stacklen
== 1) {
1132 pr_err("overlayfs: at least 2 lowerdir are needed while upperdir nonexistent\n");
1137 stack
= kcalloc(stacklen
, sizeof(struct path
), GFP_KERNEL
);
1143 for (numlower
= 0; numlower
< stacklen
; numlower
++) {
1144 err
= ovl_lower_dir(lower
, &stack
[numlower
], ofs
,
1145 &sb
->s_stack_depth
, &remote
);
1149 lower
= strchr(lower
, '\0') + 1;
1153 sb
->s_stack_depth
++;
1154 if (sb
->s_stack_depth
> FILESYSTEM_MAX_STACK_DEPTH
) {
1155 pr_err("overlayfs: maximum fs stacking depth exceeded\n");
1159 err
= ovl_get_lower_layers(ofs
, stack
, numlower
);
1164 oe
= ovl_alloc_entry(numlower
);
1168 for (i
= 0; i
< numlower
; i
++) {
1169 oe
->lowerstack
[i
].dentry
= dget(stack
[i
].dentry
);
1170 oe
->lowerstack
[i
].layer
= &ofs
->lower_layers
[i
];
1174 sb
->s_d_op
= &ovl_reval_dentry_operations
;
1176 sb
->s_d_op
= &ovl_dentry_operations
;
1179 for (i
= 0; i
< numlower
; i
++)
1180 path_put(&stack
[i
]);
1191 static int ovl_fill_super(struct super_block
*sb
, void *data
, int silent
)
1193 struct path upperpath
= { };
1194 struct dentry
*root_dentry
;
1195 struct ovl_entry
*oe
;
1201 ofs
= kzalloc(sizeof(struct ovl_fs
), GFP_KERNEL
);
1205 ofs
->creator_cred
= cred
= prepare_creds();
1209 ofs
->config
.index
= ovl_index_def
;
1210 err
= ovl_parse_opt((char *) data
, &ofs
->config
);
1215 if (!ofs
->config
.lowerdir
) {
1217 pr_err("overlayfs: missing 'lowerdir'\n");
1221 sb
->s_stack_depth
= 0;
1222 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
1223 if (ofs
->config
.upperdir
) {
1224 if (!ofs
->config
.workdir
) {
1225 pr_err("overlayfs: missing 'workdir'\n");
1229 err
= ovl_get_upper(ofs
, &upperpath
);
1233 err
= ovl_get_workdir(ofs
, &upperpath
);
1238 sb
->s_flags
|= SB_RDONLY
;
1240 sb
->s_stack_depth
= ofs
->upper_mnt
->mnt_sb
->s_stack_depth
;
1241 sb
->s_time_gran
= ofs
->upper_mnt
->mnt_sb
->s_time_gran
;
1244 oe
= ovl_get_lowerstack(sb
, ofs
);
1249 /* If the upper fs is nonexistent, we mark overlayfs r/o too */
1250 if (!ofs
->upper_mnt
)
1251 sb
->s_flags
|= SB_RDONLY
;
1252 else if (ofs
->upper_mnt
->mnt_sb
!= ofs
->same_sb
)
1253 ofs
->same_sb
= NULL
;
1255 if (!(ovl_force_readonly(ofs
)) && ofs
->config
.index
) {
1256 err
= ovl_get_indexdir(ofs
, oe
, &upperpath
);
1261 sb
->s_flags
|= SB_RDONLY
;
1264 /* Show index=off/on in /proc/mounts for any of the reasons above */
1266 ofs
->config
.index
= false;
1268 /* Never override disk quota limits or use reserved space */
1269 cap_lower(cred
->cap_effective
, CAP_SYS_RESOURCE
);
1271 sb
->s_magic
= OVERLAYFS_SUPER_MAGIC
;
1272 sb
->s_op
= &ovl_super_operations
;
1273 sb
->s_xattr
= ovl_xattr_handlers
;
1274 sb
->s_fs_info
= ofs
;
1275 sb
->s_flags
|= SB_POSIXACL
| SB_NOREMOTELOCK
;
1278 root_dentry
= d_make_root(ovl_new_inode(sb
, S_IFDIR
, 0));
1282 mntput(upperpath
.mnt
);
1283 if (upperpath
.dentry
) {
1284 oe
->has_upper
= true;
1285 if (ovl_is_impuredir(upperpath
.dentry
))
1286 ovl_set_flag(OVL_IMPURE
, d_inode(root_dentry
));
1289 root_dentry
->d_fsdata
= oe
;
1291 /* Root is always merge -> can have whiteouts */
1292 ovl_set_flag(OVL_WHITEOUTS
, d_inode(root_dentry
));
1293 ovl_inode_init(d_inode(root_dentry
), upperpath
.dentry
,
1294 ovl_dentry_lower(root_dentry
));
1296 sb
->s_root
= root_dentry
;
1301 ovl_entry_stack_free(oe
);
1304 path_put(&upperpath
);
1310 static struct dentry
*ovl_mount(struct file_system_type
*fs_type
, int flags
,
1311 const char *dev_name
, void *raw_data
)
1313 return mount_nodev(fs_type
, flags
, raw_data
, ovl_fill_super
);
1316 static struct file_system_type ovl_fs_type
= {
1317 .owner
= THIS_MODULE
,
1320 .kill_sb
= kill_anon_super
,
1322 MODULE_ALIAS_FS("overlay");
1324 static void ovl_inode_init_once(void *foo
)
1326 struct ovl_inode
*oi
= foo
;
1328 inode_init_once(&oi
->vfs_inode
);
1331 static int __init
ovl_init(void)
1335 ovl_inode_cachep
= kmem_cache_create("ovl_inode",
1336 sizeof(struct ovl_inode
), 0,
1337 (SLAB_RECLAIM_ACCOUNT
|
1338 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
1339 ovl_inode_init_once
);
1340 if (ovl_inode_cachep
== NULL
)
1343 err
= register_filesystem(&ovl_fs_type
);
1345 kmem_cache_destroy(ovl_inode_cachep
);
1350 static void __exit
ovl_exit(void)
1352 unregister_filesystem(&ovl_fs_type
);
1355 * Make sure all delayed rcu free inodes are flushed before we
1359 kmem_cache_destroy(ovl_inode_cachep
);
1363 module_init(ovl_init
);
1364 module_exit(ovl_exit
);