1 // SPDX-License-Identifier: GPL-2.0-only
4 * Copyright (C) 2011 Novell Inc.
7 #include <uapi/linux/magic.h>
9 #include <linux/namei.h>
10 #include <linux/xattr.h>
11 #include <linux/mount.h>
12 #include <linux/parser.h>
13 #include <linux/module.h>
14 #include <linux/statfs.h>
15 #include <linux/seq_file.h>
16 #include <linux/posix_acl_xattr.h>
17 #include <linux/exportfs.h>
18 #include "overlayfs.h"
20 MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>");
21 MODULE_DESCRIPTION("Overlay filesystem");
22 MODULE_LICENSE("GPL");
27 #define OVL_MAX_STACK 500
29 static bool ovl_redirect_dir_def
= IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_DIR
);
30 module_param_named(redirect_dir
, ovl_redirect_dir_def
, bool, 0644);
31 MODULE_PARM_DESC(redirect_dir
,
32 "Default to on or off for the redirect_dir feature");
34 static bool ovl_redirect_always_follow
=
35 IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW
);
36 module_param_named(redirect_always_follow
, ovl_redirect_always_follow
,
38 MODULE_PARM_DESC(redirect_always_follow
,
39 "Follow redirects even if redirect_dir feature is turned off");
41 static bool ovl_index_def
= IS_ENABLED(CONFIG_OVERLAY_FS_INDEX
);
42 module_param_named(index
, ovl_index_def
, bool, 0644);
43 MODULE_PARM_DESC(index
,
44 "Default to on or off for the inodes index feature");
46 static bool ovl_nfs_export_def
= IS_ENABLED(CONFIG_OVERLAY_FS_NFS_EXPORT
);
47 module_param_named(nfs_export
, ovl_nfs_export_def
, bool, 0644);
48 MODULE_PARM_DESC(nfs_export
,
49 "Default to on or off for the NFS export feature");
51 static bool ovl_xino_auto_def
= IS_ENABLED(CONFIG_OVERLAY_FS_XINO_AUTO
);
52 module_param_named(xino_auto
, ovl_xino_auto_def
, bool, 0644);
53 MODULE_PARM_DESC(xino_auto
,
54 "Auto enable xino feature");
56 static void ovl_entry_stack_free(struct ovl_entry
*oe
)
60 for (i
= 0; i
< oe
->numlower
; i
++)
61 dput(oe
->lowerstack
[i
].dentry
);
64 static bool ovl_metacopy_def
= IS_ENABLED(CONFIG_OVERLAY_FS_METACOPY
);
65 module_param_named(metacopy
, ovl_metacopy_def
, bool, 0644);
66 MODULE_PARM_DESC(metacopy
,
67 "Default to on or off for the metadata only copy up feature");
69 static void ovl_dentry_release(struct dentry
*dentry
)
71 struct ovl_entry
*oe
= dentry
->d_fsdata
;
74 ovl_entry_stack_free(oe
);
79 static struct dentry
*ovl_d_real(struct dentry
*dentry
,
80 const struct inode
*inode
)
82 struct dentry
*real
= NULL
, *lower
;
84 /* It's an overlay file */
85 if (inode
&& d_inode(dentry
) == inode
)
88 if (!d_is_reg(dentry
)) {
89 if (!inode
|| inode
== d_inode(dentry
))
94 real
= ovl_dentry_upper(dentry
);
95 if (real
&& (inode
== d_inode(real
)))
98 if (real
&& !inode
&& ovl_has_upperdata(d_inode(dentry
)))
101 lower
= ovl_dentry_lowerdata(dentry
);
106 /* Handle recursion */
107 real
= d_real(real
, inode
);
109 if (!inode
|| inode
== d_inode(real
))
112 WARN(1, "%s(%pd4, %s:%lu): real dentry (%p/%lu) not found\n",
113 __func__
, dentry
, inode
? inode
->i_sb
->s_id
: "NULL",
114 inode
? inode
->i_ino
: 0, real
,
115 real
&& d_inode(real
) ? d_inode(real
)->i_ino
: 0);
119 static int ovl_revalidate_real(struct dentry
*d
, unsigned int flags
, bool weak
)
124 if (d
->d_flags
& DCACHE_OP_WEAK_REVALIDATE
)
125 ret
= d
->d_op
->d_weak_revalidate(d
, flags
);
126 } else if (d
->d_flags
& DCACHE_OP_REVALIDATE
) {
127 ret
= d
->d_op
->d_revalidate(d
, flags
);
129 if (!(flags
& LOOKUP_RCU
))
137 static int ovl_dentry_revalidate_common(struct dentry
*dentry
,
138 unsigned int flags
, bool weak
)
140 struct ovl_entry
*oe
= dentry
->d_fsdata
;
141 struct dentry
*upper
;
145 upper
= ovl_dentry_upper(dentry
);
147 ret
= ovl_revalidate_real(upper
, flags
, weak
);
149 for (i
= 0; ret
> 0 && i
< oe
->numlower
; i
++) {
150 ret
= ovl_revalidate_real(oe
->lowerstack
[i
].dentry
, flags
,
156 static int ovl_dentry_revalidate(struct dentry
*dentry
, unsigned int flags
)
158 return ovl_dentry_revalidate_common(dentry
, flags
, false);
161 static int ovl_dentry_weak_revalidate(struct dentry
*dentry
, unsigned int flags
)
163 return ovl_dentry_revalidate_common(dentry
, flags
, true);
166 static const struct dentry_operations ovl_dentry_operations
= {
167 .d_release
= ovl_dentry_release
,
168 .d_real
= ovl_d_real
,
169 .d_revalidate
= ovl_dentry_revalidate
,
170 .d_weak_revalidate
= ovl_dentry_weak_revalidate
,
173 static struct kmem_cache
*ovl_inode_cachep
;
175 static struct inode
*ovl_alloc_inode(struct super_block
*sb
)
177 struct ovl_inode
*oi
= kmem_cache_alloc(ovl_inode_cachep
, GFP_KERNEL
);
186 oi
->__upperdentry
= NULL
;
188 oi
->lowerdata
= NULL
;
189 mutex_init(&oi
->lock
);
191 return &oi
->vfs_inode
;
194 static void ovl_free_inode(struct inode
*inode
)
196 struct ovl_inode
*oi
= OVL_I(inode
);
199 mutex_destroy(&oi
->lock
);
200 kmem_cache_free(ovl_inode_cachep
, oi
);
203 static void ovl_destroy_inode(struct inode
*inode
)
205 struct ovl_inode
*oi
= OVL_I(inode
);
207 dput(oi
->__upperdentry
);
209 if (S_ISDIR(inode
->i_mode
))
210 ovl_dir_cache_free(inode
);
215 static void ovl_free_fs(struct ovl_fs
*ofs
)
217 struct vfsmount
**mounts
;
220 iput(ofs
->workbasedir_trap
);
221 iput(ofs
->indexdir_trap
);
222 iput(ofs
->workdir_trap
);
226 if (ofs
->workdir_locked
)
227 ovl_inuse_unlock(ofs
->workbasedir
);
228 dput(ofs
->workbasedir
);
229 if (ofs
->upperdir_locked
)
230 ovl_inuse_unlock(ovl_upper_mnt(ofs
)->mnt_root
);
232 /* Hack! Reuse ofs->layers as a vfsmount array before freeing it */
233 mounts
= (struct vfsmount
**) ofs
->layers
;
234 for (i
= 0; i
< ofs
->numlayer
; i
++) {
235 iput(ofs
->layers
[i
].trap
);
236 mounts
[i
] = ofs
->layers
[i
].mnt
;
238 kern_unmount_array(mounts
, ofs
->numlayer
);
240 for (i
= 0; i
< ofs
->numfs
; i
++)
241 free_anon_bdev(ofs
->fs
[i
].pseudo_dev
);
244 kfree(ofs
->config
.lowerdir
);
245 kfree(ofs
->config
.upperdir
);
246 kfree(ofs
->config
.workdir
);
247 kfree(ofs
->config
.redirect_mode
);
248 if (ofs
->creator_cred
)
249 put_cred(ofs
->creator_cred
);
253 static void ovl_put_super(struct super_block
*sb
)
255 struct ovl_fs
*ofs
= sb
->s_fs_info
;
260 /* Sync real dirty inodes in upper filesystem (if it exists) */
261 static int ovl_sync_fs(struct super_block
*sb
, int wait
)
263 struct ovl_fs
*ofs
= sb
->s_fs_info
;
264 struct super_block
*upper_sb
;
267 ret
= ovl_sync_status(ofs
);
269 * We have to always set the err, because the return value isn't
270 * checked in syncfs, and instead indirectly return an error via
271 * the sb's writeback errseq, which VFS inspects after this call.
274 errseq_set(&sb
->s_wb_err
, -EIO
);
282 * Not called for sync(2) call or an emergency sync (SB_I_SKIP_SYNC).
283 * All the super blocks will be iterated, including upper_sb.
285 * If this is a syncfs(2) call, then we do need to call
286 * sync_filesystem() on upper_sb, but enough if we do it when being
287 * called with wait == 1.
292 upper_sb
= ovl_upper_mnt(ofs
)->mnt_sb
;
294 down_read(&upper_sb
->s_umount
);
295 ret
= sync_filesystem(upper_sb
);
296 up_read(&upper_sb
->s_umount
);
303 * @sb: The overlayfs super block
304 * @buf: The struct kstatfs to fill in with stats
306 * Get the filesystem statistics. As writes always target the upper layer
307 * filesystem pass the statfs to the upper filesystem (if it exists)
309 static int ovl_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
311 struct ovl_fs
*ofs
= dentry
->d_sb
->s_fs_info
;
312 struct dentry
*root_dentry
= dentry
->d_sb
->s_root
;
316 ovl_path_real(root_dentry
, &path
);
318 err
= vfs_statfs(&path
, buf
);
320 buf
->f_namelen
= ofs
->namelen
;
321 buf
->f_type
= OVERLAYFS_SUPER_MAGIC
;
327 /* Will this overlay be forced to mount/remount ro? */
328 static bool ovl_force_readonly(struct ovl_fs
*ofs
)
330 return (!ovl_upper_mnt(ofs
) || !ofs
->workdir
);
333 static const char *ovl_redirect_mode_def(void)
335 return ovl_redirect_dir_def
? "on" : "off";
338 static const char * const ovl_xino_str
[] = {
344 static inline int ovl_xino_def(void)
346 return ovl_xino_auto_def
? OVL_XINO_AUTO
: OVL_XINO_OFF
;
352 * Prints the mount options for a given superblock.
353 * Returns zero; does not fail.
355 static int ovl_show_options(struct seq_file
*m
, struct dentry
*dentry
)
357 struct super_block
*sb
= dentry
->d_sb
;
358 struct ovl_fs
*ofs
= sb
->s_fs_info
;
360 seq_show_option(m
, "lowerdir", ofs
->config
.lowerdir
);
361 if (ofs
->config
.upperdir
) {
362 seq_show_option(m
, "upperdir", ofs
->config
.upperdir
);
363 seq_show_option(m
, "workdir", ofs
->config
.workdir
);
365 if (ofs
->config
.default_permissions
)
366 seq_puts(m
, ",default_permissions");
367 if (strcmp(ofs
->config
.redirect_mode
, ovl_redirect_mode_def()) != 0)
368 seq_printf(m
, ",redirect_dir=%s", ofs
->config
.redirect_mode
);
369 if (ofs
->config
.index
!= ovl_index_def
)
370 seq_printf(m
, ",index=%s", ofs
->config
.index
? "on" : "off");
371 if (!ofs
->config
.uuid
)
372 seq_puts(m
, ",uuid=off");
373 if (ofs
->config
.nfs_export
!= ovl_nfs_export_def
)
374 seq_printf(m
, ",nfs_export=%s", ofs
->config
.nfs_export
?
376 if (ofs
->config
.xino
!= ovl_xino_def() && !ovl_same_fs(sb
))
377 seq_printf(m
, ",xino=%s", ovl_xino_str
[ofs
->config
.xino
]);
378 if (ofs
->config
.metacopy
!= ovl_metacopy_def
)
379 seq_printf(m
, ",metacopy=%s",
380 ofs
->config
.metacopy
? "on" : "off");
381 if (ofs
->config
.ovl_volatile
)
382 seq_puts(m
, ",volatile");
383 if (ofs
->config
.userxattr
)
384 seq_puts(m
, ",userxattr");
388 static int ovl_remount(struct super_block
*sb
, int *flags
, char *data
)
390 struct ovl_fs
*ofs
= sb
->s_fs_info
;
391 struct super_block
*upper_sb
;
394 if (!(*flags
& SB_RDONLY
) && ovl_force_readonly(ofs
))
397 if (*flags
& SB_RDONLY
&& !sb_rdonly(sb
)) {
398 upper_sb
= ovl_upper_mnt(ofs
)->mnt_sb
;
399 if (ovl_should_sync(ofs
)) {
400 down_read(&upper_sb
->s_umount
);
401 ret
= sync_filesystem(upper_sb
);
402 up_read(&upper_sb
->s_umount
);
409 static const struct super_operations ovl_super_operations
= {
410 .alloc_inode
= ovl_alloc_inode
,
411 .free_inode
= ovl_free_inode
,
412 .destroy_inode
= ovl_destroy_inode
,
413 .drop_inode
= generic_delete_inode
,
414 .put_super
= ovl_put_super
,
415 .sync_fs
= ovl_sync_fs
,
416 .statfs
= ovl_statfs
,
417 .show_options
= ovl_show_options
,
418 .remount_fs
= ovl_remount
,
425 OPT_DEFAULT_PERMISSIONS
,
443 static const match_table_t ovl_tokens
= {
444 {OPT_LOWERDIR
, "lowerdir=%s"},
445 {OPT_UPPERDIR
, "upperdir=%s"},
446 {OPT_WORKDIR
, "workdir=%s"},
447 {OPT_DEFAULT_PERMISSIONS
, "default_permissions"},
448 {OPT_REDIRECT_DIR
, "redirect_dir=%s"},
449 {OPT_INDEX_ON
, "index=on"},
450 {OPT_INDEX_OFF
, "index=off"},
451 {OPT_USERXATTR
, "userxattr"},
452 {OPT_UUID_ON
, "uuid=on"},
453 {OPT_UUID_OFF
, "uuid=off"},
454 {OPT_NFS_EXPORT_ON
, "nfs_export=on"},
455 {OPT_NFS_EXPORT_OFF
, "nfs_export=off"},
456 {OPT_XINO_ON
, "xino=on"},
457 {OPT_XINO_OFF
, "xino=off"},
458 {OPT_XINO_AUTO
, "xino=auto"},
459 {OPT_METACOPY_ON
, "metacopy=on"},
460 {OPT_METACOPY_OFF
, "metacopy=off"},
461 {OPT_VOLATILE
, "volatile"},
465 static char *ovl_next_opt(char **s
)
473 for (p
= sbegin
; *p
; p
++) {
478 } else if (*p
== ',') {
488 static int ovl_parse_redirect_mode(struct ovl_config
*config
, const char *mode
)
490 if (strcmp(mode
, "on") == 0) {
491 config
->redirect_dir
= true;
493 * Does not make sense to have redirect creation without
494 * redirect following.
496 config
->redirect_follow
= true;
497 } else if (strcmp(mode
, "follow") == 0) {
498 config
->redirect_follow
= true;
499 } else if (strcmp(mode
, "off") == 0) {
500 if (ovl_redirect_always_follow
)
501 config
->redirect_follow
= true;
502 } else if (strcmp(mode
, "nofollow") != 0) {
503 pr_err("bad mount option \"redirect_dir=%s\"\n",
511 static int ovl_parse_opt(char *opt
, struct ovl_config
*config
)
515 bool metacopy_opt
= false, redirect_opt
= false;
516 bool nfs_export_opt
= false, index_opt
= false;
518 config
->redirect_mode
= kstrdup(ovl_redirect_mode_def(), GFP_KERNEL
);
519 if (!config
->redirect_mode
)
522 while ((p
= ovl_next_opt(&opt
)) != NULL
) {
524 substring_t args
[MAX_OPT_ARGS
];
529 token
= match_token(p
, ovl_tokens
, args
);
532 kfree(config
->upperdir
);
533 config
->upperdir
= match_strdup(&args
[0]);
534 if (!config
->upperdir
)
539 kfree(config
->lowerdir
);
540 config
->lowerdir
= match_strdup(&args
[0]);
541 if (!config
->lowerdir
)
546 kfree(config
->workdir
);
547 config
->workdir
= match_strdup(&args
[0]);
548 if (!config
->workdir
)
552 case OPT_DEFAULT_PERMISSIONS
:
553 config
->default_permissions
= true;
556 case OPT_REDIRECT_DIR
:
557 kfree(config
->redirect_mode
);
558 config
->redirect_mode
= match_strdup(&args
[0]);
559 if (!config
->redirect_mode
)
565 config
->index
= true;
570 config
->index
= false;
579 config
->uuid
= false;
582 case OPT_NFS_EXPORT_ON
:
583 config
->nfs_export
= true;
584 nfs_export_opt
= true;
587 case OPT_NFS_EXPORT_OFF
:
588 config
->nfs_export
= false;
589 nfs_export_opt
= true;
593 config
->xino
= OVL_XINO_ON
;
597 config
->xino
= OVL_XINO_OFF
;
601 config
->xino
= OVL_XINO_AUTO
;
604 case OPT_METACOPY_ON
:
605 config
->metacopy
= true;
609 case OPT_METACOPY_OFF
:
610 config
->metacopy
= false;
615 config
->ovl_volatile
= true;
619 config
->userxattr
= true;
623 pr_err("unrecognized mount option \"%s\" or missing value\n",
629 /* Workdir/index are useless in non-upper mount */
630 if (!config
->upperdir
) {
631 if (config
->workdir
) {
632 pr_info("option \"workdir=%s\" is useless in a non-upper mount, ignore\n",
634 kfree(config
->workdir
);
635 config
->workdir
= NULL
;
637 if (config
->index
&& index_opt
) {
638 pr_info("option \"index=on\" is useless in a non-upper mount, ignore\n");
641 config
->index
= false;
644 if (!config
->upperdir
&& config
->ovl_volatile
) {
645 pr_info("option \"volatile\" is meaningless in a non-upper mount, ignoring it.\n");
646 config
->ovl_volatile
= false;
649 err
= ovl_parse_redirect_mode(config
, config
->redirect_mode
);
654 * This is to make the logic below simpler. It doesn't make any other
655 * difference, since config->redirect_dir is only used for upper.
657 if (!config
->upperdir
&& config
->redirect_follow
)
658 config
->redirect_dir
= true;
660 /* Resolve metacopy -> redirect_dir dependency */
661 if (config
->metacopy
&& !config
->redirect_dir
) {
662 if (metacopy_opt
&& redirect_opt
) {
663 pr_err("conflicting options: metacopy=on,redirect_dir=%s\n",
664 config
->redirect_mode
);
669 * There was an explicit redirect_dir=... that resulted
672 pr_info("disabling metacopy due to redirect_dir=%s\n",
673 config
->redirect_mode
);
674 config
->metacopy
= false;
676 /* Automatically enable redirect otherwise. */
677 config
->redirect_follow
= config
->redirect_dir
= true;
681 /* Resolve nfs_export -> index dependency */
682 if (config
->nfs_export
&& !config
->index
) {
683 if (!config
->upperdir
&& config
->redirect_follow
) {
684 pr_info("NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n");
685 config
->nfs_export
= false;
686 } else if (nfs_export_opt
&& index_opt
) {
687 pr_err("conflicting options: nfs_export=on,index=off\n");
689 } else if (index_opt
) {
691 * There was an explicit index=off that resulted
694 pr_info("disabling nfs_export due to index=off\n");
695 config
->nfs_export
= false;
697 /* Automatically enable index otherwise. */
698 config
->index
= true;
702 /* Resolve nfs_export -> !metacopy dependency */
703 if (config
->nfs_export
&& config
->metacopy
) {
704 if (nfs_export_opt
&& metacopy_opt
) {
705 pr_err("conflicting options: nfs_export=on,metacopy=on\n");
710 * There was an explicit metacopy=on that resulted
713 pr_info("disabling nfs_export due to metacopy=on\n");
714 config
->nfs_export
= false;
717 * There was an explicit nfs_export=on that resulted
720 pr_info("disabling metacopy due to nfs_export=on\n");
721 config
->metacopy
= false;
726 /* Resolve userxattr -> !redirect && !metacopy dependency */
727 if (config
->userxattr
) {
728 if (config
->redirect_follow
&& redirect_opt
) {
729 pr_err("conflicting options: userxattr,redirect_dir=%s\n",
730 config
->redirect_mode
);
733 if (config
->metacopy
&& metacopy_opt
) {
734 pr_err("conflicting options: userxattr,metacopy=on\n");
738 * Silently disable default setting of redirect and metacopy.
739 * This shall be the default in the future as well: these
740 * options must be explicitly enabled if used together with
743 config
->redirect_dir
= config
->redirect_follow
= false;
744 config
->metacopy
= false;
750 #define OVL_WORKDIR_NAME "work"
751 #define OVL_INDEXDIR_NAME "index"
753 static struct dentry
*ovl_workdir_create(struct ovl_fs
*ofs
,
754 const char *name
, bool persist
)
756 struct inode
*dir
= ofs
->workbasedir
->d_inode
;
757 struct vfsmount
*mnt
= ovl_upper_mnt(ofs
);
760 bool retried
= false;
762 inode_lock_nested(dir
, I_MUTEX_PARENT
);
764 work
= lookup_one_len(name
, ofs
->workbasedir
, strlen(name
));
767 struct iattr attr
= {
768 .ia_valid
= ATTR_MODE
,
769 .ia_mode
= S_IFDIR
| 0,
781 err
= ovl_workdir_cleanup(dir
, mnt
, work
, 0);
783 if (err
== -EINVAL
) {
790 err
= ovl_mkdir_real(dir
, &work
, attr
.ia_mode
);
794 /* Weird filesystem returning with hashed negative (kernfs)? */
796 if (d_really_is_negative(work
))
800 * Try to remove POSIX ACL xattrs from workdir. We are good if:
802 * a) success (there was a POSIX ACL xattr and was removed)
803 * b) -ENODATA (there was no POSIX ACL xattr)
804 * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported)
806 * There are various other error values that could effectively
807 * mean that the xattr doesn't exist (e.g. -ERANGE is returned
808 * if the xattr name is too long), but the set of filesystems
809 * allowed as upper are limited to "normal" ones, where checking
810 * for the above two errors is sufficient.
812 err
= vfs_removexattr(&init_user_ns
, work
,
813 XATTR_NAME_POSIX_ACL_DEFAULT
);
814 if (err
&& err
!= -ENODATA
&& err
!= -EOPNOTSUPP
)
817 err
= vfs_removexattr(&init_user_ns
, work
,
818 XATTR_NAME_POSIX_ACL_ACCESS
);
819 if (err
&& err
!= -ENODATA
&& err
!= -EOPNOTSUPP
)
822 /* Clear any inherited mode bits */
823 inode_lock(work
->d_inode
);
824 err
= notify_change(&init_user_ns
, work
, &attr
, NULL
);
825 inode_unlock(work
->d_inode
);
839 pr_warn("failed to create directory %s/%s (errno: %i); mounting read-only\n",
840 ofs
->config
.workdir
, name
, -err
);
845 static void ovl_unescape(char *s
)
858 static int ovl_mount_dir_noesc(const char *name
, struct path
*path
)
863 pr_err("empty lowerdir\n");
866 err
= kern_path(name
, LOOKUP_FOLLOW
, path
);
868 pr_err("failed to resolve '%s': %i\n", name
, err
);
872 if (ovl_dentry_weird(path
->dentry
)) {
873 pr_err("filesystem on '%s' not supported\n", name
);
876 if (mnt_user_ns(path
->mnt
) != &init_user_ns
) {
877 pr_err("idmapped layers are currently not supported\n");
880 if (!d_is_dir(path
->dentry
)) {
881 pr_err("'%s' not a directory\n", name
);
892 static int ovl_mount_dir(const char *name
, struct path
*path
)
895 char *tmp
= kstrdup(name
, GFP_KERNEL
);
899 err
= ovl_mount_dir_noesc(tmp
, path
);
901 if (!err
&& (path
->dentry
->d_flags
& DCACHE_OP_REAL
&&
902 path
->dentry
->d_sb
->s_magic
!= SHIFTFS_MAGIC
)) {
903 pr_err("filesystem on '%s' not supported as upperdir\n",
913 static int ovl_check_namelen(struct path
*path
, struct ovl_fs
*ofs
,
916 struct kstatfs statfs
;
917 int err
= vfs_statfs(path
, &statfs
);
920 pr_err("statfs failed on '%s'\n", name
);
922 ofs
->namelen
= max(ofs
->namelen
, statfs
.f_namelen
);
927 static int ovl_lower_dir(const char *name
, struct path
*path
,
928 struct ovl_fs
*ofs
, int *stack_depth
)
933 err
= ovl_mount_dir_noesc(name
, path
);
937 err
= ovl_check_namelen(path
, ofs
, name
);
941 *stack_depth
= max(*stack_depth
, path
->mnt
->mnt_sb
->s_stack_depth
);
944 * The inodes index feature and NFS export need to encode and decode
945 * file handles, so they require that all layers support them.
947 fh_type
= ovl_can_decode_fh(path
->dentry
->d_sb
);
948 if ((ofs
->config
.nfs_export
||
949 (ofs
->config
.index
&& ofs
->config
.upperdir
)) && !fh_type
) {
950 ofs
->config
.index
= false;
951 ofs
->config
.nfs_export
= false;
952 pr_warn("fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n",
956 * Decoding origin file handle is required for persistent st_ino.
957 * Without persistent st_ino, xino=auto falls back to xino=off.
959 if (ofs
->config
.xino
== OVL_XINO_AUTO
&&
960 ofs
->config
.upperdir
&& !fh_type
) {
961 ofs
->config
.xino
= OVL_XINO_OFF
;
962 pr_warn("fs on '%s' does not support file handles, falling back to xino=off.\n",
966 /* Check if lower fs has 32bit inode numbers */
967 if (fh_type
!= FILEID_INO32_GEN
)
973 /* Workdir should not be subdir of upperdir and vice versa */
974 static bool ovl_workdir_ok(struct dentry
*workdir
, struct dentry
*upperdir
)
978 if (workdir
!= upperdir
) {
979 ok
= (lock_rename(workdir
, upperdir
) == NULL
);
980 unlock_rename(workdir
, upperdir
);
985 static unsigned int ovl_split_lowerdirs(char *str
)
987 unsigned int ctr
= 1;
990 for (s
= d
= str
;; s
++, d
++) {
993 } else if (*s
== ':') {
1005 static int __maybe_unused
1006 ovl_posix_acl_xattr_get(const struct xattr_handler
*handler
,
1007 struct dentry
*dentry
, struct inode
*inode
,
1008 const char *name
, void *buffer
, size_t size
)
1010 return ovl_xattr_get(dentry
, inode
, handler
->name
, buffer
, size
);
1013 static int __maybe_unused
1014 ovl_posix_acl_xattr_set(const struct xattr_handler
*handler
,
1015 struct user_namespace
*mnt_userns
,
1016 struct dentry
*dentry
, struct inode
*inode
,
1017 const char *name
, const void *value
,
1018 size_t size
, int flags
)
1020 struct dentry
*workdir
= ovl_workdir(dentry
);
1021 struct inode
*realinode
= ovl_inode_real(inode
);
1022 struct posix_acl
*acl
= NULL
;
1025 /* Check that everything is OK before copy-up */
1027 acl
= posix_acl_from_xattr(&init_user_ns
, value
, size
);
1029 return PTR_ERR(acl
);
1032 if (!IS_POSIXACL(d_inode(workdir
)))
1033 goto out_acl_release
;
1034 if (!realinode
->i_op
->set_acl
)
1035 goto out_acl_release
;
1036 if (handler
->flags
== ACL_TYPE_DEFAULT
&& !S_ISDIR(inode
->i_mode
)) {
1037 err
= acl
? -EACCES
: 0;
1038 goto out_acl_release
;
1041 if (!inode_owner_or_capable(&init_user_ns
, inode
))
1042 goto out_acl_release
;
1044 posix_acl_release(acl
);
1047 * Check if sgid bit needs to be cleared (actual setacl operation will
1048 * be done with mounter's capabilities and so that won't do it for us).
1050 if (unlikely(inode
->i_mode
& S_ISGID
) &&
1051 handler
->flags
== ACL_TYPE_ACCESS
&&
1052 !in_group_p(inode
->i_gid
) &&
1053 !capable_wrt_inode_uidgid(&init_user_ns
, inode
, CAP_FSETID
)) {
1054 struct iattr iattr
= { .ia_valid
= ATTR_KILL_SGID
};
1056 err
= ovl_setattr(&init_user_ns
, dentry
, &iattr
);
1061 err
= ovl_xattr_set(dentry
, inode
, handler
->name
, value
, size
, flags
);
1065 posix_acl_release(acl
);
1069 static int ovl_own_xattr_get(const struct xattr_handler
*handler
,
1070 struct dentry
*dentry
, struct inode
*inode
,
1071 const char *name
, void *buffer
, size_t size
)
1076 static int ovl_own_xattr_set(const struct xattr_handler
*handler
,
1077 struct user_namespace
*mnt_userns
,
1078 struct dentry
*dentry
, struct inode
*inode
,
1079 const char *name
, const void *value
,
1080 size_t size
, int flags
)
1085 static int ovl_other_xattr_get(const struct xattr_handler
*handler
,
1086 struct dentry
*dentry
, struct inode
*inode
,
1087 const char *name
, void *buffer
, size_t size
)
1089 return ovl_xattr_get(dentry
, inode
, name
, buffer
, size
);
1092 static int ovl_other_xattr_set(const struct xattr_handler
*handler
,
1093 struct user_namespace
*mnt_userns
,
1094 struct dentry
*dentry
, struct inode
*inode
,
1095 const char *name
, const void *value
,
1096 size_t size
, int flags
)
1098 return ovl_xattr_set(dentry
, inode
, name
, value
, size
, flags
);
1101 static const struct xattr_handler __maybe_unused
1102 ovl_posix_acl_access_xattr_handler
= {
1103 .name
= XATTR_NAME_POSIX_ACL_ACCESS
,
1104 .flags
= ACL_TYPE_ACCESS
,
1105 .get
= ovl_posix_acl_xattr_get
,
1106 .set
= ovl_posix_acl_xattr_set
,
1109 static const struct xattr_handler __maybe_unused
1110 ovl_posix_acl_default_xattr_handler
= {
1111 .name
= XATTR_NAME_POSIX_ACL_DEFAULT
,
1112 .flags
= ACL_TYPE_DEFAULT
,
1113 .get
= ovl_posix_acl_xattr_get
,
1114 .set
= ovl_posix_acl_xattr_set
,
1117 static const struct xattr_handler ovl_own_trusted_xattr_handler
= {
1118 .prefix
= OVL_XATTR_TRUSTED_PREFIX
,
1119 .get
= ovl_own_xattr_get
,
1120 .set
= ovl_own_xattr_set
,
1123 static const struct xattr_handler ovl_own_user_xattr_handler
= {
1124 .prefix
= OVL_XATTR_USER_PREFIX
,
1125 .get
= ovl_own_xattr_get
,
1126 .set
= ovl_own_xattr_set
,
1129 static const struct xattr_handler ovl_other_xattr_handler
= {
1130 .prefix
= "", /* catch all */
1131 .get
= ovl_other_xattr_get
,
1132 .set
= ovl_other_xattr_set
,
1135 static const struct xattr_handler
*ovl_trusted_xattr_handlers
[] = {
1136 #ifdef CONFIG_FS_POSIX_ACL
1137 &ovl_posix_acl_access_xattr_handler
,
1138 &ovl_posix_acl_default_xattr_handler
,
1140 &ovl_own_trusted_xattr_handler
,
1141 &ovl_other_xattr_handler
,
1145 static const struct xattr_handler
*ovl_user_xattr_handlers
[] = {
1146 #ifdef CONFIG_FS_POSIX_ACL
1147 &ovl_posix_acl_access_xattr_handler
,
1148 &ovl_posix_acl_default_xattr_handler
,
1150 &ovl_own_user_xattr_handler
,
1151 &ovl_other_xattr_handler
,
1155 static int ovl_setup_trap(struct super_block
*sb
, struct dentry
*dir
,
1156 struct inode
**ptrap
, const char *name
)
1161 trap
= ovl_get_trap_inode(sb
, dir
);
1162 err
= PTR_ERR_OR_ZERO(trap
);
1165 pr_err("conflicting %s path\n", name
);
1174 * Determine how we treat concurrent use of upperdir/workdir based on the
1175 * index feature. This is papering over mount leaks of container runtimes,
1176 * for example, an old overlay mount is leaked and now its upperdir is
1177 * attempted to be used as a lower layer in a new overlay mount.
1179 static int ovl_report_in_use(struct ovl_fs
*ofs
, const char *name
)
1181 if (ofs
->config
.index
) {
1182 pr_err("%s is in-use as upperdir/workdir of another mount, mount with '-o index=off' to override exclusive upperdir protection.\n",
1186 pr_warn("%s is in-use as upperdir/workdir of another mount, accessing files from both mounts will result in undefined behavior.\n",
1192 static int ovl_get_upper(struct super_block
*sb
, struct ovl_fs
*ofs
,
1193 struct ovl_layer
*upper_layer
, struct path
*upperpath
)
1195 struct vfsmount
*upper_mnt
;
1198 err
= ovl_mount_dir(ofs
->config
.upperdir
, upperpath
);
1202 /* Upperdir path should not be r/o */
1203 if (__mnt_is_readonly(upperpath
->mnt
)) {
1204 pr_err("upper fs is r/o, try multi-lower layers mount\n");
1209 err
= ovl_check_namelen(upperpath
, ofs
, ofs
->config
.upperdir
);
1213 err
= ovl_setup_trap(sb
, upperpath
->dentry
, &upper_layer
->trap
,
1218 upper_mnt
= clone_private_mount(upperpath
);
1219 err
= PTR_ERR(upper_mnt
);
1220 if (IS_ERR(upper_mnt
)) {
1221 pr_err("failed to clone upperpath\n");
1225 /* Don't inherit atime flags */
1226 upper_mnt
->mnt_flags
&= ~(MNT_NOATIME
| MNT_NODIRATIME
| MNT_RELATIME
);
1227 upper_layer
->mnt
= upper_mnt
;
1228 upper_layer
->idx
= 0;
1229 upper_layer
->fsid
= 0;
1232 * Inherit SB_NOSEC flag from upperdir.
1234 * This optimization changes behavior when a security related attribute
1235 * (suid/sgid/security.*) is changed on an underlying layer. This is
1236 * okay because we don't yet have guarantees in that case, but it will
1237 * need careful treatment once we want to honour changes to underlying
1240 if (upper_mnt
->mnt_sb
->s_flags
& SB_NOSEC
)
1241 sb
->s_flags
|= SB_NOSEC
;
1243 if (ovl_inuse_trylock(ovl_upper_mnt(ofs
)->mnt_root
)) {
1244 ofs
->upperdir_locked
= true;
1246 err
= ovl_report_in_use(ofs
, "upperdir");
1257 * Returns 1 if RENAME_WHITEOUT is supported, 0 if not supported and
1258 * negative values if error is encountered.
1260 static int ovl_check_rename_whiteout(struct dentry
*workdir
)
1262 struct inode
*dir
= d_inode(workdir
);
1263 struct dentry
*temp
;
1264 struct dentry
*dest
;
1265 struct dentry
*whiteout
;
1266 struct name_snapshot name
;
1269 inode_lock_nested(dir
, I_MUTEX_PARENT
);
1271 temp
= ovl_create_temp(workdir
, OVL_CATTR(S_IFREG
| 0));
1272 err
= PTR_ERR(temp
);
1276 dest
= ovl_lookup_temp(workdir
);
1277 err
= PTR_ERR(dest
);
1283 /* Name is inline and stable - using snapshot as a copy helper */
1284 take_dentry_name_snapshot(&name
, temp
);
1285 err
= ovl_do_rename(dir
, temp
, dir
, dest
, RENAME_WHITEOUT
);
1292 whiteout
= lookup_one_len(name
.name
.name
, workdir
, name
.name
.len
);
1293 err
= PTR_ERR(whiteout
);
1294 if (IS_ERR(whiteout
))
1297 err
= ovl_is_whiteout(whiteout
);
1299 /* Best effort cleanup of whiteout and temp file */
1301 ovl_cleanup(dir
, whiteout
);
1305 ovl_cleanup(dir
, temp
);
1306 release_dentry_name_snapshot(&name
);
1316 static struct dentry
*ovl_lookup_or_create(struct dentry
*parent
,
1317 const char *name
, umode_t mode
)
1319 size_t len
= strlen(name
);
1320 struct dentry
*child
;
1322 inode_lock_nested(parent
->d_inode
, I_MUTEX_PARENT
);
1323 child
= lookup_one_len(name
, parent
, len
);
1324 if (!IS_ERR(child
) && !child
->d_inode
)
1325 child
= ovl_create_real(parent
->d_inode
, child
,
1327 inode_unlock(parent
->d_inode
);
1334 * Creates $workdir/work/incompat/volatile/dirty file if it is not already
1337 static int ovl_create_volatile_dirty(struct ovl_fs
*ofs
)
1340 struct dentry
*d
= dget(ofs
->workbasedir
);
1341 static const char *const volatile_path
[] = {
1342 OVL_WORKDIR_NAME
, "incompat", "volatile", "dirty"
1344 const char *const *name
= volatile_path
;
1346 for (ctr
= ARRAY_SIZE(volatile_path
); ctr
; ctr
--, name
++) {
1347 d
= ovl_lookup_or_create(d
, *name
, ctr
> 1 ? S_IFDIR
: S_IFREG
);
1355 static int ovl_make_workdir(struct super_block
*sb
, struct ovl_fs
*ofs
,
1356 struct path
*workpath
)
1358 struct vfsmount
*mnt
= ovl_upper_mnt(ofs
);
1359 struct dentry
*temp
, *workdir
;
1360 bool rename_whiteout
;
1365 err
= mnt_want_write(mnt
);
1369 workdir
= ovl_workdir_create(ofs
, OVL_WORKDIR_NAME
, false);
1370 err
= PTR_ERR(workdir
);
1371 if (IS_ERR_OR_NULL(workdir
))
1374 ofs
->workdir
= workdir
;
1376 err
= ovl_setup_trap(sb
, ofs
->workdir
, &ofs
->workdir_trap
, "workdir");
1381 * Upper should support d_type, else whiteouts are visible. Given
1382 * workdir and upper are on same fs, we can do iterate_dir() on
1383 * workdir. This check requires successful creation of workdir in
1386 err
= ovl_check_d_type_supported(workpath
);
1392 pr_warn("upper fs needs to support d_type.\n");
1394 /* Check if upper/work fs supports O_TMPFILE */
1395 temp
= ovl_do_tmpfile(ofs
->workdir
, S_IFREG
| 0);
1396 ofs
->tmpfile
= !IS_ERR(temp
);
1400 pr_warn("upper fs does not support tmpfile.\n");
1403 /* Check if upper/work fs supports RENAME_WHITEOUT */
1404 err
= ovl_check_rename_whiteout(ofs
->workdir
);
1408 rename_whiteout
= err
;
1409 if (!rename_whiteout
)
1410 pr_warn("upper fs does not support RENAME_WHITEOUT.\n");
1413 * Check if upper/work fs supports (trusted|user).overlay.* xattr
1415 err
= ovl_do_setxattr(ofs
, ofs
->workdir
, OVL_XATTR_OPAQUE
, "0", 1);
1417 ofs
->noxattr
= true;
1418 if (ofs
->config
.index
|| ofs
->config
.metacopy
) {
1419 ofs
->config
.index
= false;
1420 ofs
->config
.metacopy
= false;
1421 pr_warn("upper fs does not support xattr, falling back to index=off,metacopy=off.\n");
1424 * xattr support is required for persistent st_ino.
1425 * Without persistent st_ino, xino=auto falls back to xino=off.
1427 if (ofs
->config
.xino
== OVL_XINO_AUTO
) {
1428 ofs
->config
.xino
= OVL_XINO_OFF
;
1429 pr_warn("upper fs does not support xattr, falling back to xino=off.\n");
1433 ovl_do_removexattr(ofs
, ofs
->workdir
, OVL_XATTR_OPAQUE
);
1437 * We allowed sub-optimal upper fs configuration and don't want to break
1438 * users over kernel upgrade, but we never allowed remote upper fs, so
1439 * we can enforce strict requirements for remote upper fs.
1441 if (ovl_dentry_remote(ofs
->workdir
) &&
1442 (!d_type
|| !rename_whiteout
|| ofs
->noxattr
)) {
1443 pr_err("upper fs missing required features.\n");
1449 * For volatile mount, create a incompat/volatile/dirty file to keep
1452 if (ofs
->config
.ovl_volatile
) {
1453 err
= ovl_create_volatile_dirty(ofs
);
1455 pr_err("Failed to create volatile/dirty file.\n");
1460 /* Check if upper/work fs supports file handles */
1461 fh_type
= ovl_can_decode_fh(ofs
->workdir
->d_sb
);
1462 if (ofs
->config
.index
&& !fh_type
) {
1463 ofs
->config
.index
= false;
1464 pr_warn("upper fs does not support file handles, falling back to index=off.\n");
1467 /* Check if upper fs has 32bit inode numbers */
1468 if (fh_type
!= FILEID_INO32_GEN
)
1469 ofs
->xino_mode
= -1;
1471 /* NFS export of r/w mount depends on index */
1472 if (ofs
->config
.nfs_export
&& !ofs
->config
.index
) {
1473 pr_warn("NFS export requires \"index=on\", falling back to nfs_export=off.\n");
1474 ofs
->config
.nfs_export
= false;
1477 mnt_drop_write(mnt
);
1481 static int ovl_get_workdir(struct super_block
*sb
, struct ovl_fs
*ofs
,
1482 struct path
*upperpath
)
1485 struct path workpath
= { };
1487 err
= ovl_mount_dir(ofs
->config
.workdir
, &workpath
);
1492 if (upperpath
->mnt
!= workpath
.mnt
) {
1493 pr_err("workdir and upperdir must reside under the same mount\n");
1496 if (!ovl_workdir_ok(workpath
.dentry
, upperpath
->dentry
)) {
1497 pr_err("workdir and upperdir must be separate subtrees\n");
1501 ofs
->workbasedir
= dget(workpath
.dentry
);
1503 if (ovl_inuse_trylock(ofs
->workbasedir
)) {
1504 ofs
->workdir_locked
= true;
1506 err
= ovl_report_in_use(ofs
, "workdir");
1511 err
= ovl_setup_trap(sb
, ofs
->workbasedir
, &ofs
->workbasedir_trap
,
1516 err
= ovl_make_workdir(sb
, ofs
, &workpath
);
1519 path_put(&workpath
);
1524 static int ovl_get_indexdir(struct super_block
*sb
, struct ovl_fs
*ofs
,
1525 struct ovl_entry
*oe
, struct path
*upperpath
)
1527 struct vfsmount
*mnt
= ovl_upper_mnt(ofs
);
1528 struct dentry
*indexdir
;
1531 err
= mnt_want_write(mnt
);
1535 /* Verify lower root is upper root origin */
1536 err
= ovl_verify_origin(ofs
, upperpath
->dentry
,
1537 oe
->lowerstack
[0].dentry
, true);
1539 pr_err("failed to verify upper root origin\n");
1543 /* index dir will act also as workdir */
1544 iput(ofs
->workdir_trap
);
1545 ofs
->workdir_trap
= NULL
;
1547 ofs
->workdir
= NULL
;
1548 indexdir
= ovl_workdir_create(ofs
, OVL_INDEXDIR_NAME
, true);
1549 if (IS_ERR(indexdir
)) {
1550 err
= PTR_ERR(indexdir
);
1551 } else if (indexdir
) {
1552 ofs
->indexdir
= indexdir
;
1553 ofs
->workdir
= dget(indexdir
);
1555 err
= ovl_setup_trap(sb
, ofs
->indexdir
, &ofs
->indexdir_trap
,
1561 * Verify upper root is exclusively associated with index dir.
1562 * Older kernels stored upper fh in ".overlay.origin"
1563 * xattr. If that xattr exists, verify that it is a match to
1564 * upper dir file handle. In any case, verify or set xattr
1565 * ".overlay.upper" to indicate that index may have
1566 * directory entries.
1568 if (ovl_check_origin_xattr(ofs
, ofs
->indexdir
)) {
1569 err
= ovl_verify_set_fh(ofs
, ofs
->indexdir
,
1571 upperpath
->dentry
, true, false);
1573 pr_err("failed to verify index dir 'origin' xattr\n");
1575 err
= ovl_verify_upper(ofs
, ofs
->indexdir
, upperpath
->dentry
,
1578 pr_err("failed to verify index dir 'upper' xattr\n");
1580 /* Cleanup bad/stale/orphan index entries */
1582 err
= ovl_indexdir_cleanup(ofs
);
1584 if (err
|| !ofs
->indexdir
)
1585 pr_warn("try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
1588 mnt_drop_write(mnt
);
1592 static bool ovl_lower_uuid_ok(struct ovl_fs
*ofs
, const uuid_t
*uuid
)
1596 if (!ofs
->config
.nfs_export
&& !ovl_upper_mnt(ofs
))
1600 * We allow using single lower with null uuid for index and nfs_export
1601 * for example to support those features with single lower squashfs.
1602 * To avoid regressions in setups of overlay with re-formatted lower
1603 * squashfs, do not allow decoding origin with lower null uuid unless
1604 * user opted-in to one of the new features that require following the
1605 * lower inode of non-dir upper.
1607 if (ovl_allow_offline_changes(ofs
) && uuid_is_null(uuid
))
1610 for (i
= 0; i
< ofs
->numfs
; i
++) {
1612 * We use uuid to associate an overlay lower file handle with a
1613 * lower layer, so we can accept lower fs with null uuid as long
1614 * as all lower layers with null uuid are on the same fs.
1615 * if we detect multiple lower fs with the same uuid, we
1616 * disable lower file handle decoding on all of them.
1618 if (ofs
->fs
[i
].is_lower
&&
1619 uuid_equal(&ofs
->fs
[i
].sb
->s_uuid
, uuid
)) {
1620 ofs
->fs
[i
].bad_uuid
= true;
1627 /* Get a unique fsid for the layer */
1628 static int ovl_get_fsid(struct ovl_fs
*ofs
, const struct path
*path
)
1630 struct super_block
*sb
= path
->mnt
->mnt_sb
;
1634 bool bad_uuid
= false;
1637 for (i
= 0; i
< ofs
->numfs
; i
++) {
1638 if (ofs
->fs
[i
].sb
== sb
)
1642 if (!ovl_lower_uuid_ok(ofs
, &sb
->s_uuid
)) {
1644 if (ofs
->config
.xino
== OVL_XINO_AUTO
) {
1645 ofs
->config
.xino
= OVL_XINO_OFF
;
1648 if (ofs
->config
.index
|| ofs
->config
.nfs_export
) {
1649 ofs
->config
.index
= false;
1650 ofs
->config
.nfs_export
= false;
1654 pr_warn("%s uuid detected in lower fs '%pd2', falling back to xino=%s,index=off,nfs_export=off.\n",
1655 uuid_is_null(&sb
->s_uuid
) ? "null" :
1657 path
->dentry
, ovl_xino_str
[ofs
->config
.xino
]);
1661 err
= get_anon_bdev(&dev
);
1663 pr_err("failed to get anonymous bdev for lowerpath\n");
1667 ofs
->fs
[ofs
->numfs
].sb
= sb
;
1668 ofs
->fs
[ofs
->numfs
].pseudo_dev
= dev
;
1669 ofs
->fs
[ofs
->numfs
].bad_uuid
= bad_uuid
;
1671 return ofs
->numfs
++;
1674 static int ovl_get_layers(struct super_block
*sb
, struct ovl_fs
*ofs
,
1675 struct path
*stack
, unsigned int numlower
,
1676 struct ovl_layer
*layers
)
1682 ofs
->fs
= kcalloc(numlower
+ 1, sizeof(struct ovl_sb
), GFP_KERNEL
);
1683 if (ofs
->fs
== NULL
)
1686 /* idx/fsid 0 are reserved for upper fs even with lower only overlay */
1690 * All lower layers that share the same fs as upper layer, use the same
1691 * pseudo_dev as upper layer. Allocate fs[0].pseudo_dev even for lower
1692 * only overlay to simplify ovl_fs_free().
1693 * is_lower will be set if upper fs is shared with a lower layer.
1695 err
= get_anon_bdev(&ofs
->fs
[0].pseudo_dev
);
1697 pr_err("failed to get anonymous bdev for upper fs\n");
1701 if (ovl_upper_mnt(ofs
)) {
1702 ofs
->fs
[0].sb
= ovl_upper_mnt(ofs
)->mnt_sb
;
1703 ofs
->fs
[0].is_lower
= false;
1706 for (i
= 0; i
< numlower
; i
++) {
1707 struct vfsmount
*mnt
;
1711 err
= fsid
= ovl_get_fsid(ofs
, &stack
[i
]);
1716 * Check if lower root conflicts with this overlay layers before
1717 * checking if it is in-use as upperdir/workdir of "another"
1718 * mount, because we do not bother to check in ovl_is_inuse() if
1719 * the upperdir/workdir is in fact in-use by our
1722 err
= ovl_setup_trap(sb
, stack
[i
].dentry
, &trap
, "lowerdir");
1726 if (ovl_is_inuse(stack
[i
].dentry
)) {
1727 err
= ovl_report_in_use(ofs
, "lowerdir");
1734 mnt
= clone_private_mount(&stack
[i
]);
1737 pr_err("failed to clone lowerpath\n");
1743 * Make lower layers R/O. That way fchmod/fchown on lower file
1744 * will fail instead of modifying lower fs.
1746 mnt
->mnt_flags
|= MNT_READONLY
| MNT_NOATIME
;
1748 layers
[ofs
->numlayer
].trap
= trap
;
1749 layers
[ofs
->numlayer
].mnt
= mnt
;
1750 layers
[ofs
->numlayer
].idx
= ofs
->numlayer
;
1751 layers
[ofs
->numlayer
].fsid
= fsid
;
1752 layers
[ofs
->numlayer
].fs
= &ofs
->fs
[fsid
];
1754 ofs
->fs
[fsid
].is_lower
= true;
1758 * When all layers on same fs, overlay can use real inode numbers.
1759 * With mount option "xino=<on|auto>", mounter declares that there are
1760 * enough free high bits in underlying fs to hold the unique fsid.
1761 * If overlayfs does encounter underlying inodes using the high xino
1762 * bits reserved for fsid, it emits a warning and uses the original
1763 * inode number or a non persistent inode number allocated from a
1766 if (ofs
->numfs
- !ovl_upper_mnt(ofs
) == 1) {
1767 if (ofs
->config
.xino
== OVL_XINO_ON
)
1768 pr_info("\"xino=on\" is useless with all layers on same fs, ignore.\n");
1770 } else if (ofs
->config
.xino
== OVL_XINO_OFF
) {
1771 ofs
->xino_mode
= -1;
1772 } else if (ofs
->xino_mode
< 0) {
1774 * This is a roundup of number of bits needed for encoding
1775 * fsid, where fsid 0 is reserved for upper fs (even with
1776 * lower only overlay) +1 extra bit is reserved for the non
1777 * persistent inode number range that is used for resolving
1778 * xino lower bits overflow.
1780 BUILD_BUG_ON(ilog2(OVL_MAX_STACK
) > 30);
1781 ofs
->xino_mode
= ilog2(ofs
->numfs
- 1) + 2;
1784 if (ofs
->xino_mode
> 0) {
1785 pr_info("\"xino\" feature enabled using %d upper inode bits.\n",
1794 static struct ovl_entry
*ovl_get_lowerstack(struct super_block
*sb
,
1795 const char *lower
, unsigned int numlower
,
1796 struct ovl_fs
*ofs
, struct ovl_layer
*layers
)
1799 struct path
*stack
= NULL
;
1801 struct ovl_entry
*oe
;
1803 if (!ofs
->config
.upperdir
&& numlower
== 1) {
1804 pr_err("at least 2 lowerdir are needed while upperdir nonexistent\n");
1805 return ERR_PTR(-EINVAL
);
1808 stack
= kcalloc(numlower
, sizeof(struct path
), GFP_KERNEL
);
1810 return ERR_PTR(-ENOMEM
);
1813 for (i
= 0; i
< numlower
; i
++) {
1814 err
= ovl_lower_dir(lower
, &stack
[i
], ofs
, &sb
->s_stack_depth
);
1818 lower
= strchr(lower
, '\0') + 1;
1822 sb
->s_stack_depth
++;
1823 if (sb
->s_stack_depth
> FILESYSTEM_MAX_STACK_DEPTH
) {
1824 pr_err("maximum fs stacking depth exceeded\n");
1828 err
= ovl_get_layers(sb
, ofs
, stack
, numlower
, layers
);
1833 oe
= ovl_alloc_entry(numlower
);
1837 for (i
= 0; i
< numlower
; i
++) {
1838 oe
->lowerstack
[i
].dentry
= dget(stack
[i
].dentry
);
1839 oe
->lowerstack
[i
].layer
= &ofs
->layers
[i
+1];
1843 for (i
= 0; i
< numlower
; i
++)
1844 path_put(&stack
[i
]);
1855 * Check if this layer root is a descendant of:
1856 * - another layer of this overlayfs instance
1857 * - upper/work dir of any overlayfs instance
1859 static int ovl_check_layer(struct super_block
*sb
, struct ovl_fs
*ofs
,
1860 struct dentry
*dentry
, const char *name
,
1863 struct dentry
*next
= dentry
, *parent
;
1869 parent
= dget_parent(next
);
1871 /* Walk back ancestors to root (inclusive) looking for traps */
1872 while (!err
&& parent
!= next
) {
1873 if (is_lower
&& ovl_lookup_trap_inode(sb
, parent
)) {
1875 pr_err("overlapping %s path\n", name
);
1876 } else if (ovl_is_inuse(parent
)) {
1877 err
= ovl_report_in_use(ofs
, name
);
1880 parent
= dget_parent(next
);
1890 * Check if any of the layers or work dirs overlap.
1892 static int ovl_check_overlapping_layers(struct super_block
*sb
,
1897 if (ovl_upper_mnt(ofs
)) {
1898 err
= ovl_check_layer(sb
, ofs
, ovl_upper_mnt(ofs
)->mnt_root
,
1904 * Checking workbasedir avoids hitting ovl_is_inuse(parent) of
1905 * this instance and covers overlapping work and index dirs,
1906 * unless work or index dir have been moved since created inside
1907 * workbasedir. In that case, we already have their traps in
1908 * inode cache and we will catch that case on lookup.
1910 err
= ovl_check_layer(sb
, ofs
, ofs
->workbasedir
, "workdir",
1916 for (i
= 1; i
< ofs
->numlayer
; i
++) {
1917 err
= ovl_check_layer(sb
, ofs
,
1918 ofs
->layers
[i
].mnt
->mnt_root
,
1927 static struct dentry
*ovl_get_root(struct super_block
*sb
,
1928 struct dentry
*upperdentry
,
1929 struct ovl_entry
*oe
)
1931 struct dentry
*root
;
1932 struct ovl_path
*lowerpath
= &oe
->lowerstack
[0];
1933 unsigned long ino
= d_inode(lowerpath
->dentry
)->i_ino
;
1934 int fsid
= lowerpath
->layer
->fsid
;
1935 struct ovl_inode_params oip
= {
1936 .upperdentry
= upperdentry
,
1937 .lowerpath
= lowerpath
,
1940 root
= d_make_root(ovl_new_inode(sb
, S_IFDIR
, 0));
1944 root
->d_fsdata
= oe
;
1947 /* Root inode uses upper st_ino/i_ino */
1948 ino
= d_inode(upperdentry
)->i_ino
;
1950 ovl_dentry_set_upper_alias(root
);
1951 if (ovl_is_impuredir(sb
, upperdentry
))
1952 ovl_set_flag(OVL_IMPURE
, d_inode(root
));
1955 /* Root is always merge -> can have whiteouts */
1956 ovl_set_flag(OVL_WHITEOUTS
, d_inode(root
));
1957 ovl_dentry_set_flag(OVL_E_CONNECTED
, root
);
1958 ovl_set_upperdata(d_inode(root
));
1959 ovl_inode_init(d_inode(root
), &oip
, ino
, fsid
);
1960 ovl_dentry_update_reval(root
, upperdentry
, DCACHE_OP_WEAK_REVALIDATE
);
1965 static int ovl_fill_super(struct super_block
*sb
, void *data
, int silent
)
1967 struct path upperpath
= { };
1968 struct dentry
*root_dentry
;
1969 struct ovl_entry
*oe
;
1971 struct ovl_layer
*layers
;
1973 char *splitlower
= NULL
;
1974 unsigned int numlower
;
1978 if (WARN_ON(sb
->s_user_ns
!= current_user_ns()))
1981 sb
->s_d_op
= &ovl_dentry_operations
;
1984 ofs
= kzalloc(sizeof(struct ovl_fs
), GFP_KERNEL
);
1989 ofs
->creator_cred
= cred
= prepare_creds();
1993 /* Is there a reason anyone would want not to share whiteouts? */
1994 ofs
->share_whiteout
= true;
1996 ofs
->config
.index
= ovl_index_def
;
1997 ofs
->config
.uuid
= true;
1998 ofs
->config
.nfs_export
= ovl_nfs_export_def
;
1999 ofs
->config
.xino
= ovl_xino_def();
2000 ofs
->config
.metacopy
= ovl_metacopy_def
;
2001 err
= ovl_parse_opt((char *) data
, &ofs
->config
);
2006 if (!ofs
->config
.lowerdir
) {
2008 pr_err("missing 'lowerdir'\n");
2013 splitlower
= kstrdup(ofs
->config
.lowerdir
, GFP_KERNEL
);
2018 numlower
= ovl_split_lowerdirs(splitlower
);
2019 if (numlower
> OVL_MAX_STACK
) {
2020 pr_err("too many lower directories, limit is %d\n",
2026 layers
= kcalloc(numlower
+ 1, sizeof(struct ovl_layer
), GFP_KERNEL
);
2030 ofs
->layers
= layers
;
2031 /* Layer 0 is reserved for upper even if there's no upper */
2034 sb
->s_stack_depth
= 0;
2035 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
2036 atomic_long_set(&ofs
->last_ino
, 1);
2037 /* Assume underlaying fs uses 32bit inodes unless proven otherwise */
2038 if (ofs
->config
.xino
!= OVL_XINO_OFF
) {
2039 ofs
->xino_mode
= BITS_PER_LONG
- 32;
2040 if (!ofs
->xino_mode
) {
2041 pr_warn("xino not supported on 32bit kernel, falling back to xino=off.\n");
2042 ofs
->config
.xino
= OVL_XINO_OFF
;
2046 /* alloc/destroy_inode needed for setting up traps in inode cache */
2047 sb
->s_op
= &ovl_super_operations
;
2049 if (ofs
->config
.upperdir
) {
2050 struct super_block
*upper_sb
;
2053 if (!ofs
->config
.workdir
) {
2054 pr_err("missing 'workdir'\n");
2058 err
= ovl_get_upper(sb
, ofs
, &layers
[0], &upperpath
);
2062 upper_sb
= ovl_upper_mnt(ofs
)->mnt_sb
;
2063 if (!ovl_should_sync(ofs
)) {
2064 ofs
->errseq
= errseq_sample(&upper_sb
->s_wb_err
);
2065 if (errseq_check(&upper_sb
->s_wb_err
, ofs
->errseq
)) {
2067 pr_err("Cannot mount volatile when upperdir has an unseen error. Sync upperdir fs to clear state.\n");
2072 err
= ovl_get_workdir(sb
, ofs
, &upperpath
);
2077 sb
->s_flags
|= SB_RDONLY
;
2079 sb
->s_stack_depth
= upper_sb
->s_stack_depth
;
2080 sb
->s_time_gran
= upper_sb
->s_time_gran
;
2082 oe
= ovl_get_lowerstack(sb
, splitlower
, numlower
, ofs
, layers
);
2087 /* If the upper fs is nonexistent, we mark overlayfs r/o too */
2088 if (!ovl_upper_mnt(ofs
))
2089 sb
->s_flags
|= SB_RDONLY
;
2091 if (!ofs
->config
.uuid
&& ofs
->numfs
> 1) {
2092 pr_warn("The uuid=off requires a single fs for lower and upper, falling back to uuid=on.\n");
2093 ofs
->config
.uuid
= true;
2096 if (!ovl_force_readonly(ofs
) && ofs
->config
.index
) {
2097 err
= ovl_get_indexdir(sb
, ofs
, oe
, &upperpath
);
2101 /* Force r/o mount with no index dir */
2103 sb
->s_flags
|= SB_RDONLY
;
2106 err
= ovl_check_overlapping_layers(sb
, ofs
);
2110 /* Show index=off in /proc/mounts for forced r/o mount */
2111 if (!ofs
->indexdir
) {
2112 ofs
->config
.index
= false;
2113 if (ovl_upper_mnt(ofs
) && ofs
->config
.nfs_export
) {
2114 pr_warn("NFS export requires an index dir, falling back to nfs_export=off.\n");
2115 ofs
->config
.nfs_export
= false;
2119 if (ofs
->config
.metacopy
&& ofs
->config
.nfs_export
) {
2120 pr_warn("NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n");
2121 ofs
->config
.nfs_export
= false;
2124 if (ofs
->config
.nfs_export
)
2125 sb
->s_export_op
= &ovl_export_operations
;
2127 /* Never override disk quota limits or use reserved space */
2128 cap_lower(cred
->cap_effective
, CAP_SYS_RESOURCE
);
2130 sb
->s_magic
= OVERLAYFS_SUPER_MAGIC
;
2131 sb
->s_xattr
= ofs
->config
.userxattr
? ovl_user_xattr_handlers
:
2132 ovl_trusted_xattr_handlers
;
2133 sb
->s_fs_info
= ofs
;
2134 sb
->s_flags
|= SB_POSIXACL
;
2135 sb
->s_iflags
|= SB_I_SKIP_SYNC
;
2138 root_dentry
= ovl_get_root(sb
, upperpath
.dentry
, oe
);
2142 mntput(upperpath
.mnt
);
2145 sb
->s_root
= root_dentry
;
2150 ovl_entry_stack_free(oe
);
2154 path_put(&upperpath
);
2160 static struct dentry
*ovl_mount(struct file_system_type
*fs_type
, int flags
,
2161 const char *dev_name
, void *raw_data
)
2163 return mount_nodev(fs_type
, flags
, raw_data
, ovl_fill_super
);
2166 static struct file_system_type ovl_fs_type
= {
2167 .owner
= THIS_MODULE
,
2169 .fs_flags
= FS_USERNS_MOUNT
,
2171 .kill_sb
= kill_anon_super
,
2173 MODULE_ALIAS_FS("overlay");
2175 static void ovl_inode_init_once(void *foo
)
2177 struct ovl_inode
*oi
= foo
;
2179 inode_init_once(&oi
->vfs_inode
);
2182 static int __init
ovl_init(void)
2186 ovl_inode_cachep
= kmem_cache_create("ovl_inode",
2187 sizeof(struct ovl_inode
), 0,
2188 (SLAB_RECLAIM_ACCOUNT
|
2189 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
2190 ovl_inode_init_once
);
2191 if (ovl_inode_cachep
== NULL
)
2194 err
= ovl_aio_request_cache_init();
2196 err
= register_filesystem(&ovl_fs_type
);
2200 ovl_aio_request_cache_destroy();
2202 kmem_cache_destroy(ovl_inode_cachep
);
2207 static void __exit
ovl_exit(void)
2209 unregister_filesystem(&ovl_fs_type
);
2212 * Make sure all delayed rcu free inodes are flushed before we
2216 kmem_cache_destroy(ovl_inode_cachep
);
2217 ovl_aio_request_cache_destroy();
2220 module_init(ovl_init
);
2221 module_exit(ovl_exit
);