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 <linux/exportfs.h>
21 #include "overlayfs.h"
23 MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>");
24 MODULE_DESCRIPTION("Overlay filesystem");
25 MODULE_LICENSE("GPL");
30 #define OVL_MAX_STACK 500
32 static bool ovl_redirect_dir_def = IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_DIR);
33 module_param_named(redirect_dir, ovl_redirect_dir_def, bool, 0644);
34 MODULE_PARM_DESC(ovl_redirect_dir_def,
35 "Default to on or off for the redirect_dir feature");
37 static bool ovl_redirect_always_follow =
38 IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW);
39 module_param_named(redirect_always_follow, ovl_redirect_always_follow,
41 MODULE_PARM_DESC(ovl_redirect_always_follow,
42 "Follow redirects even if redirect_dir feature is turned off");
44 static bool ovl_index_def = IS_ENABLED(CONFIG_OVERLAY_FS_INDEX);
45 module_param_named(index, ovl_index_def, bool, 0644);
46 MODULE_PARM_DESC(ovl_index_def,
47 "Default to on or off for the inodes index feature");
49 static bool ovl_nfs_export_def = IS_ENABLED(CONFIG_OVERLAY_FS_NFS_EXPORT);
50 module_param_named(nfs_export, ovl_nfs_export_def, bool, 0644);
51 MODULE_PARM_DESC(ovl_nfs_export_def,
52 "Default to on or off for the NFS export feature");
54 static bool ovl_xino_auto_def = IS_ENABLED(CONFIG_OVERLAY_FS_XINO_AUTO);
55 module_param_named(xino_auto, ovl_xino_auto_def, bool, 0644);
56 MODULE_PARM_DESC(ovl_xino_auto_def,
57 "Auto enable xino feature");
59 static void ovl_entry_stack_free(struct ovl_entry *oe)
63 for (i = 0; i < oe->numlower; i++)
64 dput(oe->lowerstack[i].dentry);
67 static bool ovl_metacopy_def = IS_ENABLED(CONFIG_OVERLAY_FS_METACOPY);
68 module_param_named(metacopy, ovl_metacopy_def, bool, 0644);
69 MODULE_PARM_DESC(ovl_metacopy_def,
70 "Default to on or off for the metadata only copy up feature");
72 static void ovl_dentry_release(struct dentry *dentry)
74 struct ovl_entry *oe = dentry->d_fsdata;
77 ovl_entry_stack_free(oe);
82 static struct dentry *ovl_d_real(struct dentry *dentry,
83 const struct inode *inode)
87 /* It's an overlay file */
88 if (inode && d_inode(dentry) == inode)
91 if (!d_is_reg(dentry)) {
92 if (!inode || inode == d_inode(dentry))
97 real = ovl_dentry_upper(dentry);
98 if (real && (!inode || inode == d_inode(real)))
101 real = ovl_dentry_lower(dentry);
105 /* Handle recursion */
106 real = d_real(real, inode);
108 if (!inode || inode == d_inode(real))
111 WARN(1, "ovl_d_real(%pd4, %s:%lu): real dentry not found\n", dentry,
112 inode ? inode->i_sb->s_id : "NULL", inode ? inode->i_ino : 0);
116 static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags)
118 struct ovl_entry *oe = dentry->d_fsdata;
122 for (i = 0; i < oe->numlower; i++) {
123 struct dentry *d = oe->lowerstack[i].dentry;
125 if (d->d_flags & DCACHE_OP_REVALIDATE) {
126 ret = d->d_op->d_revalidate(d, flags);
130 if (!(flags & LOOKUP_RCU))
139 static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags)
141 struct ovl_entry *oe = dentry->d_fsdata;
145 for (i = 0; i < oe->numlower; i++) {
146 struct dentry *d = oe->lowerstack[i].dentry;
148 if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) {
149 ret = d->d_op->d_weak_revalidate(d, flags);
157 static const struct dentry_operations ovl_dentry_operations = {
158 .d_release = ovl_dentry_release,
159 .d_real = ovl_d_real,
162 static const struct dentry_operations ovl_reval_dentry_operations = {
163 .d_release = ovl_dentry_release,
164 .d_real = ovl_d_real,
165 .d_revalidate = ovl_dentry_revalidate,
166 .d_weak_revalidate = ovl_dentry_weak_revalidate,
169 static struct kmem_cache *ovl_inode_cachep;
171 static struct inode *ovl_alloc_inode(struct super_block *sb)
173 struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL);
182 oi->__upperdentry = NULL;
184 mutex_init(&oi->lock);
186 return &oi->vfs_inode;
189 static void ovl_i_callback(struct rcu_head *head)
191 struct inode *inode = container_of(head, struct inode, i_rcu);
193 kmem_cache_free(ovl_inode_cachep, OVL_I(inode));
196 static void ovl_destroy_inode(struct inode *inode)
198 struct ovl_inode *oi = OVL_I(inode);
200 dput(oi->__upperdentry);
203 ovl_dir_cache_free(inode);
204 mutex_destroy(&oi->lock);
206 call_rcu(&inode->i_rcu, ovl_i_callback);
209 static void ovl_free_fs(struct ovl_fs *ofs)
215 if (ofs->workdir_locked)
216 ovl_inuse_unlock(ofs->workbasedir);
217 dput(ofs->workbasedir);
218 if (ofs->upperdir_locked)
219 ovl_inuse_unlock(ofs->upper_mnt->mnt_root);
220 mntput(ofs->upper_mnt);
221 for (i = 0; i < ofs->numlower; i++)
222 mntput(ofs->lower_layers[i].mnt);
223 for (i = 0; i < ofs->numlowerfs; i++)
224 free_anon_bdev(ofs->lower_fs[i].pseudo_dev);
225 kfree(ofs->lower_layers);
226 kfree(ofs->lower_fs);
228 kfree(ofs->config.lowerdir);
229 kfree(ofs->config.upperdir);
230 kfree(ofs->config.workdir);
231 kfree(ofs->config.redirect_mode);
232 if (ofs->creator_cred)
233 put_cred(ofs->creator_cred);
237 static void ovl_put_super(struct super_block *sb)
239 struct ovl_fs *ofs = sb->s_fs_info;
244 /* Sync real dirty inodes in upper filesystem (if it exists) */
245 static int ovl_sync_fs(struct super_block *sb, int wait)
247 struct ovl_fs *ofs = sb->s_fs_info;
248 struct super_block *upper_sb;
255 * If this is a sync(2) call or an emergency sync, all the super blocks
256 * will be iterated, including upper_sb, so no need to do anything.
258 * If this is a syncfs(2) call, then we do need to call
259 * sync_filesystem() on upper_sb, but enough if we do it when being
260 * called with wait == 1.
265 upper_sb = ofs->upper_mnt->mnt_sb;
267 down_read(&upper_sb->s_umount);
268 ret = sync_filesystem(upper_sb);
269 up_read(&upper_sb->s_umount);
276 * @sb: The overlayfs super block
277 * @buf: The struct kstatfs to fill in with stats
279 * Get the filesystem statistics. As writes always target the upper layer
280 * filesystem pass the statfs to the upper filesystem (if it exists)
282 static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf)
284 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
285 struct dentry *root_dentry = dentry->d_sb->s_root;
289 ovl_path_real(root_dentry, &path);
291 err = vfs_statfs(&path, buf);
293 buf->f_namelen = ofs->namelen;
294 buf->f_type = OVERLAYFS_SUPER_MAGIC;
300 /* Will this overlay be forced to mount/remount ro? */
301 static bool ovl_force_readonly(struct ovl_fs *ofs)
303 return (!ofs->upper_mnt || !ofs->workdir);
306 static const char *ovl_redirect_mode_def(void)
308 return ovl_redirect_dir_def ? "on" : "off";
317 static const char * const ovl_xino_str[] = {
323 static inline int ovl_xino_def(void)
325 return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF;
331 * Prints the mount options for a given superblock.
332 * Returns zero; does not fail.
334 static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
336 struct super_block *sb = dentry->d_sb;
337 struct ovl_fs *ofs = sb->s_fs_info;
339 seq_show_option(m, "lowerdir", ofs->config.lowerdir);
340 if (ofs->config.upperdir) {
341 seq_show_option(m, "upperdir", ofs->config.upperdir);
342 seq_show_option(m, "workdir", ofs->config.workdir);
344 if (ofs->config.default_permissions)
345 seq_puts(m, ",default_permissions");
346 if (strcmp(ofs->config.redirect_mode, ovl_redirect_mode_def()) != 0)
347 seq_printf(m, ",redirect_dir=%s", ofs->config.redirect_mode);
348 if (ofs->config.index != ovl_index_def)
349 seq_printf(m, ",index=%s", ofs->config.index ? "on" : "off");
350 if (ofs->config.nfs_export != ovl_nfs_export_def)
351 seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ?
353 if (ofs->config.xino != ovl_xino_def())
354 seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]);
355 if (ofs->config.metacopy != ovl_metacopy_def)
356 seq_printf(m, ",metacopy=%s",
357 ofs->config.metacopy ? "on" : "off");
361 static int ovl_remount(struct super_block *sb, int *flags, char *data)
363 struct ovl_fs *ofs = sb->s_fs_info;
365 if (!(*flags & SB_RDONLY) && ovl_force_readonly(ofs))
371 static const struct super_operations ovl_super_operations = {
372 .alloc_inode = ovl_alloc_inode,
373 .destroy_inode = ovl_destroy_inode,
374 .drop_inode = generic_delete_inode,
375 .put_super = ovl_put_super,
376 .sync_fs = ovl_sync_fs,
377 .statfs = ovl_statfs,
378 .show_options = ovl_show_options,
379 .remount_fs = ovl_remount,
386 OPT_DEFAULT_PERMISSIONS,
400 static const match_table_t ovl_tokens = {
401 {OPT_LOWERDIR, "lowerdir=%s"},
402 {OPT_UPPERDIR, "upperdir=%s"},
403 {OPT_WORKDIR, "workdir=%s"},
404 {OPT_DEFAULT_PERMISSIONS, "default_permissions"},
405 {OPT_REDIRECT_DIR, "redirect_dir=%s"},
406 {OPT_INDEX_ON, "index=on"},
407 {OPT_INDEX_OFF, "index=off"},
408 {OPT_NFS_EXPORT_ON, "nfs_export=on"},
409 {OPT_NFS_EXPORT_OFF, "nfs_export=off"},
410 {OPT_XINO_ON, "xino=on"},
411 {OPT_XINO_OFF, "xino=off"},
412 {OPT_XINO_AUTO, "xino=auto"},
413 {OPT_METACOPY_ON, "metacopy=on"},
414 {OPT_METACOPY_OFF, "metacopy=off"},
418 static char *ovl_next_opt(char **s)
426 for (p = sbegin; *p; p++) {
431 } else if (*p == ',') {
441 static int ovl_parse_redirect_mode(struct ovl_config *config, const char *mode)
443 if (strcmp(mode, "on") == 0) {
444 config->redirect_dir = true;
446 * Does not make sense to have redirect creation without
447 * redirect following.
449 config->redirect_follow = true;
450 } else if (strcmp(mode, "follow") == 0) {
451 config->redirect_follow = true;
452 } else if (strcmp(mode, "off") == 0) {
453 if (ovl_redirect_always_follow)
454 config->redirect_follow = true;
455 } else if (strcmp(mode, "nofollow") != 0) {
456 pr_err("overlayfs: bad mount option \"redirect_dir=%s\"\n",
464 static int ovl_parse_opt(char *opt, struct ovl_config *config)
469 config->redirect_mode = kstrdup(ovl_redirect_mode_def(), GFP_KERNEL);
470 if (!config->redirect_mode)
473 while ((p = ovl_next_opt(&opt)) != NULL) {
475 substring_t args[MAX_OPT_ARGS];
480 token = match_token(p, ovl_tokens, args);
483 kfree(config->upperdir);
484 config->upperdir = match_strdup(&args[0]);
485 if (!config->upperdir)
490 kfree(config->lowerdir);
491 config->lowerdir = match_strdup(&args[0]);
492 if (!config->lowerdir)
497 kfree(config->workdir);
498 config->workdir = match_strdup(&args[0]);
499 if (!config->workdir)
503 case OPT_DEFAULT_PERMISSIONS:
504 config->default_permissions = true;
507 case OPT_REDIRECT_DIR:
508 kfree(config->redirect_mode);
509 config->redirect_mode = match_strdup(&args[0]);
510 if (!config->redirect_mode)
515 config->index = true;
519 config->index = false;
522 case OPT_NFS_EXPORT_ON:
523 config->nfs_export = true;
526 case OPT_NFS_EXPORT_OFF:
527 config->nfs_export = false;
531 config->xino = OVL_XINO_ON;
535 config->xino = OVL_XINO_OFF;
539 config->xino = OVL_XINO_AUTO;
542 case OPT_METACOPY_ON:
543 config->metacopy = true;
546 case OPT_METACOPY_OFF:
547 config->metacopy = false;
551 pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p);
556 /* Workdir is useless in non-upper mount */
557 if (!config->upperdir && config->workdir) {
558 pr_info("overlayfs: option \"workdir=%s\" is useless in a non-upper mount, ignore\n",
560 kfree(config->workdir);
561 config->workdir = NULL;
564 err = ovl_parse_redirect_mode(config, config->redirect_mode);
568 /* metacopy feature with upper requires redirect_dir=on */
569 if (config->upperdir && config->metacopy && !config->redirect_dir) {
570 pr_warn("overlayfs: metadata only copy up requires \"redirect_dir=on\", falling back to metacopy=off.\n");
571 config->metacopy = false;
572 } else if (config->metacopy && !config->redirect_follow) {
573 pr_warn("overlayfs: metadata only copy up requires \"redirect_dir=follow\" on non-upper mount, falling back to metacopy=off.\n");
574 config->metacopy = false;
580 #define OVL_WORKDIR_NAME "work"
581 #define OVL_INDEXDIR_NAME "index"
583 static struct dentry *ovl_workdir_create(struct ovl_fs *ofs,
584 const char *name, bool persist)
586 struct inode *dir = ofs->workbasedir->d_inode;
587 struct vfsmount *mnt = ofs->upper_mnt;
590 bool retried = false;
593 inode_lock_nested(dir, I_MUTEX_PARENT);
597 work = lookup_one_len(name, ofs->workbasedir, strlen(name));
600 struct iattr attr = {
601 .ia_valid = ATTR_MODE,
602 .ia_mode = S_IFDIR | 0,
614 ovl_workdir_cleanup(dir, mnt, work, 0);
619 work = ovl_create_real(dir, work, OVL_CATTR(attr.ia_mode));
625 * Try to remove POSIX ACL xattrs from workdir. We are good if:
627 * a) success (there was a POSIX ACL xattr and was removed)
628 * b) -ENODATA (there was no POSIX ACL xattr)
629 * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported)
631 * There are various other error values that could effectively
632 * mean that the xattr doesn't exist (e.g. -ERANGE is returned
633 * if the xattr name is too long), but the set of filesystems
634 * allowed as upper are limited to "normal" ones, where checking
635 * for the above two errors is sufficient.
637 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT);
638 if (err && err != -ENODATA && err != -EOPNOTSUPP)
641 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS);
642 if (err && err != -ENODATA && err != -EOPNOTSUPP)
645 /* Clear any inherited mode bits */
646 inode_lock(work->d_inode);
647 err = notify_change(work, &attr, NULL);
648 inode_unlock(work->d_inode);
664 pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n",
665 ofs->config.workdir, name, -err);
670 static void ovl_unescape(char *s)
683 static int ovl_mount_dir_noesc(const char *name, struct path *path)
688 pr_err("overlayfs: empty lowerdir\n");
691 err = kern_path(name, LOOKUP_FOLLOW, path);
693 pr_err("overlayfs: failed to resolve '%s': %i\n", name, err);
697 if (ovl_dentry_weird(path->dentry)) {
698 pr_err("overlayfs: filesystem on '%s' not supported\n", name);
701 if (!d_is_dir(path->dentry)) {
702 pr_err("overlayfs: '%s' not a directory\n", name);
713 static int ovl_mount_dir(const char *name, struct path *path)
716 char *tmp = kstrdup(name, GFP_KERNEL);
720 err = ovl_mount_dir_noesc(tmp, path);
723 if (ovl_dentry_remote(path->dentry)) {
724 pr_err("overlayfs: filesystem on '%s' not supported as upperdir\n",
734 static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs,
737 struct kstatfs statfs;
738 int err = vfs_statfs(path, &statfs);
741 pr_err("overlayfs: statfs failed on '%s'\n", name);
743 ofs->namelen = max(ofs->namelen, statfs.f_namelen);
748 static int ovl_lower_dir(const char *name, struct path *path,
749 struct ovl_fs *ofs, int *stack_depth, bool *remote)
754 err = ovl_mount_dir_noesc(name, path);
758 err = ovl_check_namelen(path, ofs, name);
762 *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth);
764 if (ovl_dentry_remote(path->dentry))
768 * The inodes index feature and NFS export need to encode and decode
769 * file handles, so they require that all layers support them.
771 fh_type = ovl_can_decode_fh(path->dentry->d_sb);
772 if ((ofs->config.nfs_export ||
773 (ofs->config.index && ofs->config.upperdir)) && !fh_type) {
774 ofs->config.index = false;
775 ofs->config.nfs_export = false;
776 pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n",
780 /* Check if lower fs has 32bit inode numbers */
781 if (fh_type != FILEID_INO32_GEN)
792 /* Workdir should not be subdir of upperdir and vice versa */
793 static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir)
797 if (workdir != upperdir) {
798 ok = (lock_rename(workdir, upperdir) == NULL);
799 unlock_rename(workdir, upperdir);
804 static unsigned int ovl_split_lowerdirs(char *str)
806 unsigned int ctr = 1;
809 for (s = d = str;; s++, d++) {
812 } else if (*s == ':') {
824 static int __maybe_unused
825 ovl_posix_acl_xattr_get(const struct xattr_handler *handler,
826 struct dentry *dentry, struct inode *inode,
827 const char *name, void *buffer, size_t size)
829 return ovl_xattr_get(dentry, inode, handler->name, buffer, size);
832 static int __maybe_unused
833 ovl_posix_acl_xattr_set(const struct xattr_handler *handler,
834 struct dentry *dentry, struct inode *inode,
835 const char *name, const void *value,
836 size_t size, int flags)
838 struct dentry *workdir = ovl_workdir(dentry);
839 struct inode *realinode = ovl_inode_real(inode);
840 struct posix_acl *acl = NULL;
843 /* Check that everything is OK before copy-up */
845 acl = posix_acl_from_xattr(&init_user_ns, value, size);
850 if (!IS_POSIXACL(d_inode(workdir)))
851 goto out_acl_release;
852 if (!realinode->i_op->set_acl)
853 goto out_acl_release;
854 if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) {
855 err = acl ? -EACCES : 0;
856 goto out_acl_release;
859 if (!inode_owner_or_capable(inode))
860 goto out_acl_release;
862 posix_acl_release(acl);
865 * Check if sgid bit needs to be cleared (actual setacl operation will
866 * be done with mounter's capabilities and so that won't do it for us).
868 if (unlikely(inode->i_mode & S_ISGID) &&
869 handler->flags == ACL_TYPE_ACCESS &&
870 !in_group_p(inode->i_gid) &&
871 !capable_wrt_inode_uidgid(inode, CAP_FSETID)) {
872 struct iattr iattr = { .ia_valid = ATTR_KILL_SGID };
874 err = ovl_setattr(dentry, &iattr);
879 err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags);
881 ovl_copyattr(ovl_inode_real(inode), inode);
886 posix_acl_release(acl);
890 static int ovl_own_xattr_get(const struct xattr_handler *handler,
891 struct dentry *dentry, struct inode *inode,
892 const char *name, void *buffer, size_t size)
897 static int ovl_own_xattr_set(const struct xattr_handler *handler,
898 struct dentry *dentry, struct inode *inode,
899 const char *name, const void *value,
900 size_t size, int flags)
905 static int ovl_other_xattr_get(const struct xattr_handler *handler,
906 struct dentry *dentry, struct inode *inode,
907 const char *name, void *buffer, size_t size)
909 return ovl_xattr_get(dentry, inode, name, buffer, size);
912 static int ovl_other_xattr_set(const struct xattr_handler *handler,
913 struct dentry *dentry, struct inode *inode,
914 const char *name, const void *value,
915 size_t size, int flags)
917 return ovl_xattr_set(dentry, inode, name, value, size, flags);
920 static const struct xattr_handler __maybe_unused
921 ovl_posix_acl_access_xattr_handler = {
922 .name = XATTR_NAME_POSIX_ACL_ACCESS,
923 .flags = ACL_TYPE_ACCESS,
924 .get = ovl_posix_acl_xattr_get,
925 .set = ovl_posix_acl_xattr_set,
928 static const struct xattr_handler __maybe_unused
929 ovl_posix_acl_default_xattr_handler = {
930 .name = XATTR_NAME_POSIX_ACL_DEFAULT,
931 .flags = ACL_TYPE_DEFAULT,
932 .get = ovl_posix_acl_xattr_get,
933 .set = ovl_posix_acl_xattr_set,
936 static const struct xattr_handler ovl_own_xattr_handler = {
937 .prefix = OVL_XATTR_PREFIX,
938 .get = ovl_own_xattr_get,
939 .set = ovl_own_xattr_set,
942 static const struct xattr_handler ovl_other_xattr_handler = {
943 .prefix = "", /* catch all */
944 .get = ovl_other_xattr_get,
945 .set = ovl_other_xattr_set,
948 static const struct xattr_handler *ovl_xattr_handlers[] = {
949 #ifdef CONFIG_FS_POSIX_ACL
950 &ovl_posix_acl_access_xattr_handler,
951 &ovl_posix_acl_default_xattr_handler,
953 &ovl_own_xattr_handler,
954 &ovl_other_xattr_handler,
958 static int ovl_get_upper(struct ovl_fs *ofs, struct path *upperpath)
960 struct vfsmount *upper_mnt;
963 err = ovl_mount_dir(ofs->config.upperdir, upperpath);
967 /* Upper fs should not be r/o */
968 if (sb_rdonly(upperpath->mnt->mnt_sb)) {
969 pr_err("overlayfs: upper fs is r/o, try multi-lower layers mount\n");
974 err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir);
979 if (ovl_inuse_trylock(upperpath->dentry)) {
980 ofs->upperdir_locked = true;
981 } else if (ofs->config.index) {
982 pr_err("overlayfs: upperdir is in-use by another mount, mount with '-o index=off' to override exclusive upperdir protection.\n");
985 pr_warn("overlayfs: upperdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
988 upper_mnt = clone_private_mount(upperpath);
989 err = PTR_ERR(upper_mnt);
990 if (IS_ERR(upper_mnt)) {
991 pr_err("overlayfs: failed to clone upperpath\n");
995 /* Don't inherit atime flags */
996 upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME);
997 ofs->upper_mnt = upper_mnt;
1003 static int ovl_make_workdir(struct ovl_fs *ofs, struct path *workpath)
1005 struct vfsmount *mnt = ofs->upper_mnt;
1006 struct dentry *temp;
1010 err = mnt_want_write(mnt);
1014 ofs->workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false);
1019 * Upper should support d_type, else whiteouts are visible. Given
1020 * workdir and upper are on same fs, we can do iterate_dir() on
1021 * workdir. This check requires successful creation of workdir in
1024 err = ovl_check_d_type_supported(workpath);
1029 * We allowed this configuration and don't want to break users over
1030 * kernel upgrade. So warn instead of erroring out.
1033 pr_warn("overlayfs: upper fs needs to support d_type.\n");
1035 /* Check if upper/work fs supports O_TMPFILE */
1036 temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0);
1037 ofs->tmpfile = !IS_ERR(temp);
1041 pr_warn("overlayfs: upper fs does not support tmpfile.\n");
1044 * Check if upper/work fs supports trusted.overlay.* xattr
1046 err = ovl_do_setxattr(ofs->workdir, OVL_XATTR_OPAQUE, "0", 1, 0);
1048 ofs->noxattr = true;
1049 ofs->config.index = false;
1050 ofs->config.metacopy = false;
1051 pr_warn("overlayfs: upper fs does not support xattr, falling back to index=off and metacopy=off.\n");
1054 vfs_removexattr(ofs->workdir, OVL_XATTR_OPAQUE);
1057 /* Check if upper/work fs supports file handles */
1058 fh_type = ovl_can_decode_fh(ofs->workdir->d_sb);
1059 if (ofs->config.index && !fh_type) {
1060 ofs->config.index = false;
1061 pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n");
1064 /* Check if upper fs has 32bit inode numbers */
1065 if (fh_type != FILEID_INO32_GEN)
1068 /* NFS export of r/w mount depends on index */
1069 if (ofs->config.nfs_export && !ofs->config.index) {
1070 pr_warn("overlayfs: NFS export requires \"index=on\", falling back to nfs_export=off.\n");
1071 ofs->config.nfs_export = false;
1074 mnt_drop_write(mnt);
1078 static int ovl_get_workdir(struct ovl_fs *ofs, struct path *upperpath)
1081 struct path workpath = { };
1083 err = ovl_mount_dir(ofs->config.workdir, &workpath);
1088 if (upperpath->mnt != workpath.mnt) {
1089 pr_err("overlayfs: workdir and upperdir must reside under the same mount\n");
1092 if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) {
1093 pr_err("overlayfs: workdir and upperdir must be separate subtrees\n");
1098 if (ovl_inuse_trylock(workpath.dentry)) {
1099 ofs->workdir_locked = true;
1100 } else if (ofs->config.index) {
1101 pr_err("overlayfs: workdir is in-use by another mount, mount with '-o index=off' to override exclusive workdir protection.\n");
1104 pr_warn("overlayfs: workdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
1107 ofs->workbasedir = dget(workpath.dentry);
1108 err = ovl_make_workdir(ofs, &workpath);
1114 path_put(&workpath);
1119 static int ovl_get_indexdir(struct ovl_fs *ofs, struct ovl_entry *oe,
1120 struct path *upperpath)
1122 struct vfsmount *mnt = ofs->upper_mnt;
1125 err = mnt_want_write(mnt);
1129 /* Verify lower root is upper root origin */
1130 err = ovl_verify_origin(upperpath->dentry, oe->lowerstack[0].dentry,
1133 pr_err("overlayfs: failed to verify upper root origin\n");
1137 ofs->indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true);
1138 if (ofs->indexdir) {
1140 * Verify upper root is exclusively associated with index dir.
1141 * Older kernels stored upper fh in "trusted.overlay.origin"
1142 * xattr. If that xattr exists, verify that it is a match to
1143 * upper dir file handle. In any case, verify or set xattr
1144 * "trusted.overlay.upper" to indicate that index may have
1145 * directory entries.
1147 if (ovl_check_origin_xattr(ofs->indexdir)) {
1148 err = ovl_verify_set_fh(ofs->indexdir, OVL_XATTR_ORIGIN,
1149 upperpath->dentry, true, false);
1151 pr_err("overlayfs: failed to verify index dir 'origin' xattr\n");
1153 err = ovl_verify_upper(ofs->indexdir, upperpath->dentry, true);
1155 pr_err("overlayfs: failed to verify index dir 'upper' xattr\n");
1157 /* Cleanup bad/stale/orphan index entries */
1159 err = ovl_indexdir_cleanup(ofs);
1161 if (err || !ofs->indexdir)
1162 pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
1165 mnt_drop_write(mnt);
1169 /* Get a unique fsid for the layer */
1170 static int ovl_get_fsid(struct ovl_fs *ofs, struct super_block *sb)
1176 /* fsid 0 is reserved for upper fs even with non upper overlay */
1177 if (ofs->upper_mnt && ofs->upper_mnt->mnt_sb == sb)
1180 for (i = 0; i < ofs->numlowerfs; i++) {
1181 if (ofs->lower_fs[i].sb == sb)
1185 err = get_anon_bdev(&dev);
1187 pr_err("overlayfs: failed to get anonymous bdev for lowerpath\n");
1191 ofs->lower_fs[ofs->numlowerfs].sb = sb;
1192 ofs->lower_fs[ofs->numlowerfs].pseudo_dev = dev;
1195 return ofs->numlowerfs;
1198 static int ovl_get_lower_layers(struct ovl_fs *ofs, struct path *stack,
1199 unsigned int numlower)
1205 ofs->lower_layers = kcalloc(numlower, sizeof(struct ovl_layer),
1207 if (ofs->lower_layers == NULL)
1210 ofs->lower_fs = kcalloc(numlower, sizeof(struct ovl_sb),
1212 if (ofs->lower_fs == NULL)
1215 for (i = 0; i < numlower; i++) {
1216 struct vfsmount *mnt;
1219 err = fsid = ovl_get_fsid(ofs, stack[i].mnt->mnt_sb);
1223 mnt = clone_private_mount(&stack[i]);
1226 pr_err("overlayfs: failed to clone lowerpath\n");
1231 * Make lower layers R/O. That way fchmod/fchown on lower file
1232 * will fail instead of modifying lower fs.
1234 mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME;
1236 ofs->lower_layers[ofs->numlower].mnt = mnt;
1237 ofs->lower_layers[ofs->numlower].idx = i + 1;
1238 ofs->lower_layers[ofs->numlower].fsid = fsid;
1240 ofs->lower_layers[ofs->numlower].fs =
1241 &ofs->lower_fs[fsid - 1];
1247 * When all layers on same fs, overlay can use real inode numbers.
1248 * With mount option "xino=on", mounter declares that there are enough
1249 * free high bits in underlying fs to hold the unique fsid.
1250 * If overlayfs does encounter underlying inodes using the high xino
1251 * bits reserved for fsid, it emits a warning and uses the original
1254 if (!ofs->numlowerfs || (ofs->numlowerfs == 1 && !ofs->upper_mnt)) {
1256 ofs->config.xino = OVL_XINO_OFF;
1257 } else if (ofs->config.xino == OVL_XINO_ON && !ofs->xino_bits) {
1259 * This is a roundup of number of bits needed for numlowerfs+1
1260 * (i.e. ilog2(numlowerfs+1 - 1) + 1). fsid 0 is reserved for
1261 * upper fs even with non upper overlay.
1263 BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 31);
1264 ofs->xino_bits = ilog2(ofs->numlowerfs) + 1;
1267 if (ofs->xino_bits) {
1268 pr_info("overlayfs: \"xino\" feature enabled using %d upper inode bits.\n",
1277 static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb,
1281 char *lowertmp, *lower;
1282 struct path *stack = NULL;
1283 unsigned int stacklen, numlower = 0, i;
1284 bool remote = false;
1285 struct ovl_entry *oe;
1288 lowertmp = kstrdup(ofs->config.lowerdir, GFP_KERNEL);
1293 stacklen = ovl_split_lowerdirs(lowertmp);
1294 if (stacklen > OVL_MAX_STACK) {
1295 pr_err("overlayfs: too many lower directories, limit is %d\n",
1298 } else if (!ofs->config.upperdir && stacklen == 1) {
1299 pr_err("overlayfs: at least 2 lowerdir are needed while upperdir nonexistent\n");
1301 } else if (!ofs->config.upperdir && ofs->config.nfs_export &&
1302 ofs->config.redirect_follow) {
1303 pr_warn("overlayfs: NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n");
1304 ofs->config.nfs_export = false;
1308 stack = kcalloc(stacklen, sizeof(struct path), GFP_KERNEL);
1314 for (numlower = 0; numlower < stacklen; numlower++) {
1315 err = ovl_lower_dir(lower, &stack[numlower], ofs,
1316 &sb->s_stack_depth, &remote);
1320 lower = strchr(lower, '\0') + 1;
1324 sb->s_stack_depth++;
1325 if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) {
1326 pr_err("overlayfs: maximum fs stacking depth exceeded\n");
1330 err = ovl_get_lower_layers(ofs, stack, numlower);
1335 oe = ovl_alloc_entry(numlower);
1339 for (i = 0; i < numlower; i++) {
1340 oe->lowerstack[i].dentry = dget(stack[i].dentry);
1341 oe->lowerstack[i].layer = &ofs->lower_layers[i];
1345 sb->s_d_op = &ovl_reval_dentry_operations;
1347 sb->s_d_op = &ovl_dentry_operations;
1350 for (i = 0; i < numlower; i++)
1351 path_put(&stack[i]);
1362 static int ovl_fill_super(struct super_block *sb, void *data, int silent)
1364 struct path upperpath = { };
1365 struct dentry *root_dentry;
1366 struct ovl_entry *oe;
1372 ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL);
1376 ofs->creator_cred = cred = prepare_creds();
1380 ofs->config.index = ovl_index_def;
1381 ofs->config.nfs_export = ovl_nfs_export_def;
1382 ofs->config.xino = ovl_xino_def();
1383 ofs->config.metacopy = ovl_metacopy_def;
1384 err = ovl_parse_opt((char *) data, &ofs->config);
1389 if (!ofs->config.lowerdir) {
1391 pr_err("overlayfs: missing 'lowerdir'\n");
1395 sb->s_stack_depth = 0;
1396 sb->s_maxbytes = MAX_LFS_FILESIZE;
1397 /* Assume underlaying fs uses 32bit inodes unless proven otherwise */
1398 if (ofs->config.xino != OVL_XINO_OFF)
1399 ofs->xino_bits = BITS_PER_LONG - 32;
1401 if (ofs->config.upperdir) {
1402 if (!ofs->config.workdir) {
1403 pr_err("overlayfs: missing 'workdir'\n");
1407 err = ovl_get_upper(ofs, &upperpath);
1411 err = ovl_get_workdir(ofs, &upperpath);
1416 sb->s_flags |= SB_RDONLY;
1418 sb->s_stack_depth = ofs->upper_mnt->mnt_sb->s_stack_depth;
1419 sb->s_time_gran = ofs->upper_mnt->mnt_sb->s_time_gran;
1422 oe = ovl_get_lowerstack(sb, ofs);
1427 /* If the upper fs is nonexistent, we mark overlayfs r/o too */
1428 if (!ofs->upper_mnt)
1429 sb->s_flags |= SB_RDONLY;
1431 if (!(ovl_force_readonly(ofs)) && ofs->config.index) {
1432 err = ovl_get_indexdir(ofs, oe, &upperpath);
1436 /* Force r/o mount with no index dir */
1437 if (!ofs->indexdir) {
1439 ofs->workdir = NULL;
1440 sb->s_flags |= SB_RDONLY;
1445 /* Show index=off in /proc/mounts for forced r/o mount */
1446 if (!ofs->indexdir) {
1447 ofs->config.index = false;
1448 if (ofs->upper_mnt && ofs->config.nfs_export) {
1449 pr_warn("overlayfs: NFS export requires an index dir, falling back to nfs_export=off.\n");
1450 ofs->config.nfs_export = false;
1454 if (ofs->config.metacopy && ofs->config.nfs_export) {
1455 pr_warn("overlayfs: NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n");
1456 ofs->config.nfs_export = false;
1459 if (ofs->config.nfs_export)
1460 sb->s_export_op = &ovl_export_operations;
1462 /* Never override disk quota limits or use reserved space */
1463 cap_lower(cred->cap_effective, CAP_SYS_RESOURCE);
1465 sb->s_magic = OVERLAYFS_SUPER_MAGIC;
1466 sb->s_op = &ovl_super_operations;
1467 sb->s_xattr = ovl_xattr_handlers;
1468 sb->s_fs_info = ofs;
1469 sb->s_flags |= SB_POSIXACL;
1472 root_dentry = d_make_root(ovl_new_inode(sb, S_IFDIR, 0));
1476 root_dentry->d_fsdata = oe;
1478 mntput(upperpath.mnt);
1479 if (upperpath.dentry) {
1480 ovl_dentry_set_upper_alias(root_dentry);
1481 if (ovl_is_impuredir(upperpath.dentry))
1482 ovl_set_flag(OVL_IMPURE, d_inode(root_dentry));
1485 /* Root is always merge -> can have whiteouts */
1486 ovl_set_flag(OVL_WHITEOUTS, d_inode(root_dentry));
1487 ovl_dentry_set_flag(OVL_E_CONNECTED, root_dentry);
1488 ovl_set_upperdata(d_inode(root_dentry));
1489 ovl_inode_init(d_inode(root_dentry), upperpath.dentry,
1490 ovl_dentry_lower(root_dentry));
1492 sb->s_root = root_dentry;
1497 ovl_entry_stack_free(oe);
1500 path_put(&upperpath);
1506 static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags,
1507 const char *dev_name, void *raw_data)
1509 return mount_nodev(fs_type, flags, raw_data, ovl_fill_super);
1512 static struct file_system_type ovl_fs_type = {
1513 .owner = THIS_MODULE,
1516 .kill_sb = kill_anon_super,
1518 MODULE_ALIAS_FS("overlay");
1520 static void ovl_inode_init_once(void *foo)
1522 struct ovl_inode *oi = foo;
1524 inode_init_once(&oi->vfs_inode);
1527 static int __init ovl_init(void)
1531 ovl_inode_cachep = kmem_cache_create("ovl_inode",
1532 sizeof(struct ovl_inode), 0,
1533 (SLAB_RECLAIM_ACCOUNT|
1534 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1535 ovl_inode_init_once);
1536 if (ovl_inode_cachep == NULL)
1539 err = register_filesystem(&ovl_fs_type);
1541 kmem_cache_destroy(ovl_inode_cachep);
1546 static void __exit ovl_exit(void)
1548 unregister_filesystem(&ovl_fs_type);
1551 * Make sure all delayed rcu free inodes are flushed before we
1555 kmem_cache_destroy(ovl_inode_cachep);
1559 module_init(ovl_init);
1560 module_exit(ovl_exit);