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 void ovl_dentry_release(struct dentry *dentry)
69 struct ovl_entry *oe = dentry->d_fsdata;
72 ovl_entry_stack_free(oe);
77 static int ovl_check_append_only(struct inode *inode, int flag)
80 * This test was moot in vfs may_open() because overlay inode does
81 * not have the S_APPEND flag, so re-check on real upper inode
83 if (IS_APPEND(inode)) {
84 if ((flag & O_ACCMODE) != O_RDONLY && !(flag & O_APPEND))
93 static struct dentry *ovl_d_real(struct dentry *dentry,
94 const struct inode *inode,
95 unsigned int open_flags, unsigned int flags)
100 if (flags & D_REAL_UPPER)
101 return ovl_dentry_upper(dentry);
103 if (!d_is_reg(dentry)) {
104 if (!inode || inode == d_inode(dentry))
110 err = ovl_open_maybe_copy_up(dentry, open_flags);
115 real = ovl_dentry_upper(dentry);
116 if (real && (!inode || inode == d_inode(real))) {
118 err = ovl_check_append_only(d_inode(real), open_flags);
125 real = ovl_dentry_lower(dentry);
129 /* Handle recursion */
130 real = d_real(real, inode, open_flags, 0);
132 if (!inode || inode == d_inode(real))
135 WARN(1, "ovl_d_real(%pd4, %s:%lu): real dentry not found\n", dentry,
136 inode ? inode->i_sb->s_id : "NULL", inode ? inode->i_ino : 0);
140 static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags)
142 struct ovl_entry *oe = dentry->d_fsdata;
146 for (i = 0; i < oe->numlower; i++) {
147 struct dentry *d = oe->lowerstack[i].dentry;
149 if (d->d_flags & DCACHE_OP_REVALIDATE) {
150 ret = d->d_op->d_revalidate(d, flags);
154 if (!(flags & LOOKUP_RCU))
163 static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags)
165 struct ovl_entry *oe = dentry->d_fsdata;
169 for (i = 0; i < oe->numlower; i++) {
170 struct dentry *d = oe->lowerstack[i].dentry;
172 if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) {
173 ret = d->d_op->d_weak_revalidate(d, flags);
181 static const struct dentry_operations ovl_dentry_operations = {
182 .d_release = ovl_dentry_release,
183 .d_real = ovl_d_real,
186 static const struct dentry_operations ovl_reval_dentry_operations = {
187 .d_release = ovl_dentry_release,
188 .d_real = ovl_d_real,
189 .d_revalidate = ovl_dentry_revalidate,
190 .d_weak_revalidate = ovl_dentry_weak_revalidate,
193 static struct kmem_cache *ovl_inode_cachep;
195 static struct inode *ovl_alloc_inode(struct super_block *sb)
197 struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL);
206 oi->__upperdentry = NULL;
208 mutex_init(&oi->lock);
210 return &oi->vfs_inode;
213 static void ovl_i_callback(struct rcu_head *head)
215 struct inode *inode = container_of(head, struct inode, i_rcu);
217 kmem_cache_free(ovl_inode_cachep, OVL_I(inode));
220 static void ovl_destroy_inode(struct inode *inode)
222 struct ovl_inode *oi = OVL_I(inode);
224 dput(oi->__upperdentry);
227 ovl_dir_cache_free(inode);
228 mutex_destroy(&oi->lock);
230 call_rcu(&inode->i_rcu, ovl_i_callback);
233 static void ovl_free_fs(struct ovl_fs *ofs)
239 if (ofs->workdir_locked)
240 ovl_inuse_unlock(ofs->workbasedir);
241 dput(ofs->workbasedir);
242 if (ofs->upperdir_locked)
243 ovl_inuse_unlock(ofs->upper_mnt->mnt_root);
244 mntput(ofs->upper_mnt);
245 for (i = 0; i < ofs->numlower; i++)
246 mntput(ofs->lower_layers[i].mnt);
247 for (i = 0; i < ofs->numlowerfs; i++)
248 free_anon_bdev(ofs->lower_fs[i].pseudo_dev);
249 kfree(ofs->lower_layers);
250 kfree(ofs->lower_fs);
252 kfree(ofs->config.lowerdir);
253 kfree(ofs->config.upperdir);
254 kfree(ofs->config.workdir);
255 kfree(ofs->config.redirect_mode);
256 if (ofs->creator_cred)
257 put_cred(ofs->creator_cred);
261 static void ovl_put_super(struct super_block *sb)
263 struct ovl_fs *ofs = sb->s_fs_info;
268 /* Sync real dirty inodes in upper filesystem (if it exists) */
269 static int ovl_sync_fs(struct super_block *sb, int wait)
271 struct ovl_fs *ofs = sb->s_fs_info;
272 struct super_block *upper_sb;
279 * If this is a sync(2) call or an emergency sync, all the super blocks
280 * will be iterated, including upper_sb, so no need to do anything.
282 * If this is a syncfs(2) call, then we do need to call
283 * sync_filesystem() on upper_sb, but enough if we do it when being
284 * called with wait == 1.
289 upper_sb = ofs->upper_mnt->mnt_sb;
291 down_read(&upper_sb->s_umount);
292 ret = sync_filesystem(upper_sb);
293 up_read(&upper_sb->s_umount);
300 * @sb: The overlayfs super block
301 * @buf: The struct kstatfs to fill in with stats
303 * Get the filesystem statistics. As writes always target the upper layer
304 * filesystem pass the statfs to the upper filesystem (if it exists)
306 static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf)
308 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
309 struct dentry *root_dentry = dentry->d_sb->s_root;
313 ovl_path_real(root_dentry, &path);
315 err = vfs_statfs(&path, buf);
317 buf->f_namelen = ofs->namelen;
318 buf->f_type = OVERLAYFS_SUPER_MAGIC;
324 /* Will this overlay be forced to mount/remount ro? */
325 static bool ovl_force_readonly(struct ovl_fs *ofs)
327 return (!ofs->upper_mnt || !ofs->workdir);
330 static const char *ovl_redirect_mode_def(void)
332 return ovl_redirect_dir_def ? "on" : "off";
341 static const char * const ovl_xino_str[] = {
347 static inline int ovl_xino_def(void)
349 return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF;
355 * Prints the mount options for a given superblock.
356 * Returns zero; does not fail.
358 static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
360 struct super_block *sb = dentry->d_sb;
361 struct ovl_fs *ofs = sb->s_fs_info;
363 seq_show_option(m, "lowerdir", ofs->config.lowerdir);
364 if (ofs->config.upperdir) {
365 seq_show_option(m, "upperdir", ofs->config.upperdir);
366 seq_show_option(m, "workdir", ofs->config.workdir);
368 if (ofs->config.default_permissions)
369 seq_puts(m, ",default_permissions");
370 if (strcmp(ofs->config.redirect_mode, ovl_redirect_mode_def()) != 0)
371 seq_printf(m, ",redirect_dir=%s", ofs->config.redirect_mode);
372 if (ofs->config.index != ovl_index_def)
373 seq_printf(m, ",index=%s", ofs->config.index ? "on" : "off");
374 if (ofs->config.nfs_export != ovl_nfs_export_def)
375 seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ?
377 if (ofs->config.xino != ovl_xino_def())
378 seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]);
382 static int ovl_remount(struct super_block *sb, int *flags, char *data)
384 struct ovl_fs *ofs = sb->s_fs_info;
386 if (!(*flags & SB_RDONLY) && ovl_force_readonly(ofs))
392 static const struct super_operations ovl_super_operations = {
393 .alloc_inode = ovl_alloc_inode,
394 .destroy_inode = ovl_destroy_inode,
395 .drop_inode = generic_delete_inode,
396 .put_super = ovl_put_super,
397 .sync_fs = ovl_sync_fs,
398 .statfs = ovl_statfs,
399 .show_options = ovl_show_options,
400 .remount_fs = ovl_remount,
407 OPT_DEFAULT_PERMISSIONS,
419 static const match_table_t ovl_tokens = {
420 {OPT_LOWERDIR, "lowerdir=%s"},
421 {OPT_UPPERDIR, "upperdir=%s"},
422 {OPT_WORKDIR, "workdir=%s"},
423 {OPT_DEFAULT_PERMISSIONS, "default_permissions"},
424 {OPT_REDIRECT_DIR, "redirect_dir=%s"},
425 {OPT_INDEX_ON, "index=on"},
426 {OPT_INDEX_OFF, "index=off"},
427 {OPT_NFS_EXPORT_ON, "nfs_export=on"},
428 {OPT_NFS_EXPORT_OFF, "nfs_export=off"},
429 {OPT_XINO_ON, "xino=on"},
430 {OPT_XINO_OFF, "xino=off"},
431 {OPT_XINO_AUTO, "xino=auto"},
435 static char *ovl_next_opt(char **s)
443 for (p = sbegin; *p; p++) {
448 } else if (*p == ',') {
458 static int ovl_parse_redirect_mode(struct ovl_config *config, const char *mode)
460 if (strcmp(mode, "on") == 0) {
461 config->redirect_dir = true;
463 * Does not make sense to have redirect creation without
464 * redirect following.
466 config->redirect_follow = true;
467 } else if (strcmp(mode, "follow") == 0) {
468 config->redirect_follow = true;
469 } else if (strcmp(mode, "off") == 0) {
470 if (ovl_redirect_always_follow)
471 config->redirect_follow = true;
472 } else if (strcmp(mode, "nofollow") != 0) {
473 pr_err("overlayfs: bad mount option \"redirect_dir=%s\"\n",
481 static int ovl_parse_opt(char *opt, struct ovl_config *config)
485 config->redirect_mode = kstrdup(ovl_redirect_mode_def(), GFP_KERNEL);
486 if (!config->redirect_mode)
489 while ((p = ovl_next_opt(&opt)) != NULL) {
491 substring_t args[MAX_OPT_ARGS];
496 token = match_token(p, ovl_tokens, args);
499 kfree(config->upperdir);
500 config->upperdir = match_strdup(&args[0]);
501 if (!config->upperdir)
506 kfree(config->lowerdir);
507 config->lowerdir = match_strdup(&args[0]);
508 if (!config->lowerdir)
513 kfree(config->workdir);
514 config->workdir = match_strdup(&args[0]);
515 if (!config->workdir)
519 case OPT_DEFAULT_PERMISSIONS:
520 config->default_permissions = true;
523 case OPT_REDIRECT_DIR:
524 kfree(config->redirect_mode);
525 config->redirect_mode = match_strdup(&args[0]);
526 if (!config->redirect_mode)
531 config->index = true;
535 config->index = false;
538 case OPT_NFS_EXPORT_ON:
539 config->nfs_export = true;
542 case OPT_NFS_EXPORT_OFF:
543 config->nfs_export = false;
547 config->xino = OVL_XINO_ON;
551 config->xino = OVL_XINO_OFF;
555 config->xino = OVL_XINO_AUTO;
559 pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p);
564 /* Workdir is useless in non-upper mount */
565 if (!config->upperdir && config->workdir) {
566 pr_info("overlayfs: option \"workdir=%s\" is useless in a non-upper mount, ignore\n",
568 kfree(config->workdir);
569 config->workdir = NULL;
572 return ovl_parse_redirect_mode(config, config->redirect_mode);
575 #define OVL_WORKDIR_NAME "work"
576 #define OVL_INDEXDIR_NAME "index"
578 static struct dentry *ovl_workdir_create(struct ovl_fs *ofs,
579 const char *name, bool persist)
581 struct inode *dir = ofs->workbasedir->d_inode;
582 struct vfsmount *mnt = ofs->upper_mnt;
585 bool retried = false;
588 inode_lock_nested(dir, I_MUTEX_PARENT);
592 work = lookup_one_len(name, ofs->workbasedir, strlen(name));
595 struct iattr attr = {
596 .ia_valid = ATTR_MODE,
597 .ia_mode = S_IFDIR | 0,
609 ovl_workdir_cleanup(dir, mnt, work, 0);
614 work = ovl_create_real(dir, work, OVL_CATTR(attr.ia_mode));
620 * Try to remove POSIX ACL xattrs from workdir. We are good if:
622 * a) success (there was a POSIX ACL xattr and was removed)
623 * b) -ENODATA (there was no POSIX ACL xattr)
624 * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported)
626 * There are various other error values that could effectively
627 * mean that the xattr doesn't exist (e.g. -ERANGE is returned
628 * if the xattr name is too long), but the set of filesystems
629 * allowed as upper are limited to "normal" ones, where checking
630 * for the above two errors is sufficient.
632 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT);
633 if (err && err != -ENODATA && err != -EOPNOTSUPP)
636 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS);
637 if (err && err != -ENODATA && err != -EOPNOTSUPP)
640 /* Clear any inherited mode bits */
641 inode_lock(work->d_inode);
642 err = notify_change(work, &attr, NULL);
643 inode_unlock(work->d_inode);
659 pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n",
660 ofs->config.workdir, name, -err);
665 static void ovl_unescape(char *s)
678 static int ovl_mount_dir_noesc(const char *name, struct path *path)
683 pr_err("overlayfs: empty lowerdir\n");
686 err = kern_path(name, LOOKUP_FOLLOW, path);
688 pr_err("overlayfs: failed to resolve '%s': %i\n", name, err);
692 if (ovl_dentry_weird(path->dentry)) {
693 pr_err("overlayfs: filesystem on '%s' not supported\n", name);
696 if (!d_is_dir(path->dentry)) {
697 pr_err("overlayfs: '%s' not a directory\n", name);
708 static int ovl_mount_dir(const char *name, struct path *path)
711 char *tmp = kstrdup(name, GFP_KERNEL);
715 err = ovl_mount_dir_noesc(tmp, path);
718 if (ovl_dentry_remote(path->dentry)) {
719 pr_err("overlayfs: filesystem on '%s' not supported as upperdir\n",
729 static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs,
732 struct kstatfs statfs;
733 int err = vfs_statfs(path, &statfs);
736 pr_err("overlayfs: statfs failed on '%s'\n", name);
738 ofs->namelen = max(ofs->namelen, statfs.f_namelen);
743 static int ovl_lower_dir(const char *name, struct path *path,
744 struct ovl_fs *ofs, int *stack_depth, bool *remote)
749 err = ovl_mount_dir_noesc(name, path);
753 err = ovl_check_namelen(path, ofs, name);
757 *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth);
759 if (ovl_dentry_remote(path->dentry))
763 * The inodes index feature and NFS export need to encode and decode
764 * file handles, so they require that all layers support them.
766 fh_type = ovl_can_decode_fh(path->dentry->d_sb);
767 if ((ofs->config.nfs_export ||
768 (ofs->config.index && ofs->config.upperdir)) && !fh_type) {
769 ofs->config.index = false;
770 ofs->config.nfs_export = false;
771 pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n",
775 /* Check if lower fs has 32bit inode numbers */
776 if (fh_type != FILEID_INO32_GEN)
787 /* Workdir should not be subdir of upperdir and vice versa */
788 static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir)
792 if (workdir != upperdir) {
793 ok = (lock_rename(workdir, upperdir) == NULL);
794 unlock_rename(workdir, upperdir);
799 static unsigned int ovl_split_lowerdirs(char *str)
801 unsigned int ctr = 1;
804 for (s = d = str;; s++, d++) {
807 } else if (*s == ':') {
819 static int __maybe_unused
820 ovl_posix_acl_xattr_get(const struct xattr_handler *handler,
821 struct dentry *dentry, struct inode *inode,
822 const char *name, void *buffer, size_t size)
824 return ovl_xattr_get(dentry, inode, handler->name, buffer, size);
827 static int __maybe_unused
828 ovl_posix_acl_xattr_set(const struct xattr_handler *handler,
829 struct dentry *dentry, struct inode *inode,
830 const char *name, const void *value,
831 size_t size, int flags)
833 struct dentry *workdir = ovl_workdir(dentry);
834 struct inode *realinode = ovl_inode_real(inode);
835 struct posix_acl *acl = NULL;
838 /* Check that everything is OK before copy-up */
840 acl = posix_acl_from_xattr(&init_user_ns, value, size);
845 if (!IS_POSIXACL(d_inode(workdir)))
846 goto out_acl_release;
847 if (!realinode->i_op->set_acl)
848 goto out_acl_release;
849 if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) {
850 err = acl ? -EACCES : 0;
851 goto out_acl_release;
854 if (!inode_owner_or_capable(inode))
855 goto out_acl_release;
857 posix_acl_release(acl);
860 * Check if sgid bit needs to be cleared (actual setacl operation will
861 * be done with mounter's capabilities and so that won't do it for us).
863 if (unlikely(inode->i_mode & S_ISGID) &&
864 handler->flags == ACL_TYPE_ACCESS &&
865 !in_group_p(inode->i_gid) &&
866 !capable_wrt_inode_uidgid(inode, CAP_FSETID)) {
867 struct iattr iattr = { .ia_valid = ATTR_KILL_SGID };
869 err = ovl_setattr(dentry, &iattr);
874 err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags);
876 ovl_copyattr(ovl_inode_real(inode), inode);
881 posix_acl_release(acl);
885 static int ovl_own_xattr_get(const struct xattr_handler *handler,
886 struct dentry *dentry, struct inode *inode,
887 const char *name, void *buffer, size_t size)
892 static int ovl_own_xattr_set(const struct xattr_handler *handler,
893 struct dentry *dentry, struct inode *inode,
894 const char *name, const void *value,
895 size_t size, int flags)
900 static int ovl_other_xattr_get(const struct xattr_handler *handler,
901 struct dentry *dentry, struct inode *inode,
902 const char *name, void *buffer, size_t size)
904 return ovl_xattr_get(dentry, inode, name, buffer, size);
907 static int ovl_other_xattr_set(const struct xattr_handler *handler,
908 struct dentry *dentry, struct inode *inode,
909 const char *name, const void *value,
910 size_t size, int flags)
912 return ovl_xattr_set(dentry, inode, name, value, size, flags);
915 static const struct xattr_handler __maybe_unused
916 ovl_posix_acl_access_xattr_handler = {
917 .name = XATTR_NAME_POSIX_ACL_ACCESS,
918 .flags = ACL_TYPE_ACCESS,
919 .get = ovl_posix_acl_xattr_get,
920 .set = ovl_posix_acl_xattr_set,
923 static const struct xattr_handler __maybe_unused
924 ovl_posix_acl_default_xattr_handler = {
925 .name = XATTR_NAME_POSIX_ACL_DEFAULT,
926 .flags = ACL_TYPE_DEFAULT,
927 .get = ovl_posix_acl_xattr_get,
928 .set = ovl_posix_acl_xattr_set,
931 static const struct xattr_handler ovl_own_xattr_handler = {
932 .prefix = OVL_XATTR_PREFIX,
933 .get = ovl_own_xattr_get,
934 .set = ovl_own_xattr_set,
937 static const struct xattr_handler ovl_other_xattr_handler = {
938 .prefix = "", /* catch all */
939 .get = ovl_other_xattr_get,
940 .set = ovl_other_xattr_set,
943 static const struct xattr_handler *ovl_xattr_handlers[] = {
944 #ifdef CONFIG_FS_POSIX_ACL
945 &ovl_posix_acl_access_xattr_handler,
946 &ovl_posix_acl_default_xattr_handler,
948 &ovl_own_xattr_handler,
949 &ovl_other_xattr_handler,
953 static int ovl_get_upper(struct ovl_fs *ofs, struct path *upperpath)
955 struct vfsmount *upper_mnt;
958 err = ovl_mount_dir(ofs->config.upperdir, upperpath);
962 /* Upper fs should not be r/o */
963 if (sb_rdonly(upperpath->mnt->mnt_sb)) {
964 pr_err("overlayfs: upper fs is r/o, try multi-lower layers mount\n");
969 err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir);
974 if (ovl_inuse_trylock(upperpath->dentry)) {
975 ofs->upperdir_locked = true;
976 } else if (ofs->config.index) {
977 pr_err("overlayfs: upperdir is in-use by another mount, mount with '-o index=off' to override exclusive upperdir protection.\n");
980 pr_warn("overlayfs: upperdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
983 upper_mnt = clone_private_mount(upperpath);
984 err = PTR_ERR(upper_mnt);
985 if (IS_ERR(upper_mnt)) {
986 pr_err("overlayfs: failed to clone upperpath\n");
990 /* Don't inherit atime flags */
991 upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME);
992 ofs->upper_mnt = upper_mnt;
998 static int ovl_make_workdir(struct ovl_fs *ofs, struct path *workpath)
1000 struct vfsmount *mnt = ofs->upper_mnt;
1001 struct dentry *temp;
1005 err = mnt_want_write(mnt);
1009 ofs->workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false);
1014 * Upper should support d_type, else whiteouts are visible. Given
1015 * workdir and upper are on same fs, we can do iterate_dir() on
1016 * workdir. This check requires successful creation of workdir in
1019 err = ovl_check_d_type_supported(workpath);
1024 * We allowed this configuration and don't want to break users over
1025 * kernel upgrade. So warn instead of erroring out.
1028 pr_warn("overlayfs: upper fs needs to support d_type.\n");
1030 /* Check if upper/work fs supports O_TMPFILE */
1031 temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0);
1032 ofs->tmpfile = !IS_ERR(temp);
1036 pr_warn("overlayfs: upper fs does not support tmpfile.\n");
1039 * Check if upper/work fs supports trusted.overlay.* xattr
1041 err = ovl_do_setxattr(ofs->workdir, OVL_XATTR_OPAQUE, "0", 1, 0);
1043 ofs->noxattr = true;
1044 ofs->config.index = false;
1045 pr_warn("overlayfs: upper fs does not support xattr, falling back to index=off.\n");
1048 vfs_removexattr(ofs->workdir, OVL_XATTR_OPAQUE);
1051 /* Check if upper/work fs supports file handles */
1052 fh_type = ovl_can_decode_fh(ofs->workdir->d_sb);
1053 if (ofs->config.index && !fh_type) {
1054 ofs->config.index = false;
1055 pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n");
1058 /* Check if upper fs has 32bit inode numbers */
1059 if (fh_type != FILEID_INO32_GEN)
1062 /* NFS export of r/w mount depends on index */
1063 if (ofs->config.nfs_export && !ofs->config.index) {
1064 pr_warn("overlayfs: NFS export requires \"index=on\", falling back to nfs_export=off.\n");
1065 ofs->config.nfs_export = false;
1069 mnt_drop_write(mnt);
1073 static int ovl_get_workdir(struct ovl_fs *ofs, struct path *upperpath)
1076 struct path workpath = { };
1078 err = ovl_mount_dir(ofs->config.workdir, &workpath);
1083 if (upperpath->mnt != workpath.mnt) {
1084 pr_err("overlayfs: workdir and upperdir must reside under the same mount\n");
1087 if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) {
1088 pr_err("overlayfs: workdir and upperdir must be separate subtrees\n");
1093 if (ovl_inuse_trylock(workpath.dentry)) {
1094 ofs->workdir_locked = true;
1095 } else if (ofs->config.index) {
1096 pr_err("overlayfs: workdir is in-use by another mount, mount with '-o index=off' to override exclusive workdir protection.\n");
1099 pr_warn("overlayfs: workdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
1102 ofs->workbasedir = dget(workpath.dentry);
1103 err = ovl_make_workdir(ofs, &workpath);
1109 path_put(&workpath);
1114 static int ovl_get_indexdir(struct ovl_fs *ofs, struct ovl_entry *oe,
1115 struct path *upperpath)
1117 struct vfsmount *mnt = ofs->upper_mnt;
1120 err = mnt_want_write(mnt);
1124 /* Verify lower root is upper root origin */
1125 err = ovl_verify_origin(upperpath->dentry, oe->lowerstack[0].dentry,
1128 pr_err("overlayfs: failed to verify upper root origin\n");
1132 ofs->indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true);
1133 if (ofs->indexdir) {
1135 * Verify upper root is exclusively associated with index dir.
1136 * Older kernels stored upper fh in "trusted.overlay.origin"
1137 * xattr. If that xattr exists, verify that it is a match to
1138 * upper dir file handle. In any case, verify or set xattr
1139 * "trusted.overlay.upper" to indicate that index may have
1140 * directory entries.
1142 if (ovl_check_origin_xattr(ofs->indexdir)) {
1143 err = ovl_verify_set_fh(ofs->indexdir, OVL_XATTR_ORIGIN,
1144 upperpath->dentry, true, false);
1146 pr_err("overlayfs: failed to verify index dir 'origin' xattr\n");
1148 err = ovl_verify_upper(ofs->indexdir, upperpath->dentry, true);
1150 pr_err("overlayfs: failed to verify index dir 'upper' xattr\n");
1152 /* Cleanup bad/stale/orphan index entries */
1154 err = ovl_indexdir_cleanup(ofs);
1156 if (err || !ofs->indexdir)
1157 pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
1160 mnt_drop_write(mnt);
1164 /* Get a unique fsid for the layer */
1165 static int ovl_get_fsid(struct ovl_fs *ofs, struct super_block *sb)
1171 /* fsid 0 is reserved for upper fs even with non upper overlay */
1172 if (ofs->upper_mnt && ofs->upper_mnt->mnt_sb == sb)
1175 for (i = 0; i < ofs->numlowerfs; i++) {
1176 if (ofs->lower_fs[i].sb == sb)
1180 err = get_anon_bdev(&dev);
1182 pr_err("overlayfs: failed to get anonymous bdev for lowerpath\n");
1186 ofs->lower_fs[ofs->numlowerfs].sb = sb;
1187 ofs->lower_fs[ofs->numlowerfs].pseudo_dev = dev;
1190 return ofs->numlowerfs;
1193 static int ovl_get_lower_layers(struct ovl_fs *ofs, struct path *stack,
1194 unsigned int numlower)
1200 ofs->lower_layers = kcalloc(numlower, sizeof(struct ovl_layer),
1202 if (ofs->lower_layers == NULL)
1205 ofs->lower_fs = kcalloc(numlower, sizeof(struct ovl_sb),
1207 if (ofs->lower_fs == NULL)
1210 for (i = 0; i < numlower; i++) {
1211 struct vfsmount *mnt;
1214 err = fsid = ovl_get_fsid(ofs, stack[i].mnt->mnt_sb);
1218 mnt = clone_private_mount(&stack[i]);
1221 pr_err("overlayfs: failed to clone lowerpath\n");
1226 * Make lower layers R/O. That way fchmod/fchown on lower file
1227 * will fail instead of modifying lower fs.
1229 mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME;
1231 ofs->lower_layers[ofs->numlower].mnt = mnt;
1232 ofs->lower_layers[ofs->numlower].idx = i + 1;
1233 ofs->lower_layers[ofs->numlower].fsid = fsid;
1235 ofs->lower_layers[ofs->numlower].fs =
1236 &ofs->lower_fs[fsid - 1];
1242 * When all layers on same fs, overlay can use real inode numbers.
1243 * With mount option "xino=on", mounter declares that there are enough
1244 * free high bits in underlying fs to hold the unique fsid.
1245 * If overlayfs does encounter underlying inodes using the high xino
1246 * bits reserved for fsid, it emits a warning and uses the original
1249 if (!ofs->numlowerfs || (ofs->numlowerfs == 1 && !ofs->upper_mnt)) {
1251 ofs->config.xino = OVL_XINO_OFF;
1252 } else if (ofs->config.xino == OVL_XINO_ON && !ofs->xino_bits) {
1254 * This is a roundup of number of bits needed for numlowerfs+1
1255 * (i.e. ilog2(numlowerfs+1 - 1) + 1). fsid 0 is reserved for
1256 * upper fs even with non upper overlay.
1258 BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 31);
1259 ofs->xino_bits = ilog2(ofs->numlowerfs) + 1;
1262 if (ofs->xino_bits) {
1263 pr_info("overlayfs: \"xino\" feature enabled using %d upper inode bits.\n",
1272 static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb,
1276 char *lowertmp, *lower;
1277 struct path *stack = NULL;
1278 unsigned int stacklen, numlower = 0, i;
1279 bool remote = false;
1280 struct ovl_entry *oe;
1283 lowertmp = kstrdup(ofs->config.lowerdir, GFP_KERNEL);
1288 stacklen = ovl_split_lowerdirs(lowertmp);
1289 if (stacklen > OVL_MAX_STACK) {
1290 pr_err("overlayfs: too many lower directories, limit is %d\n",
1293 } else if (!ofs->config.upperdir && stacklen == 1) {
1294 pr_err("overlayfs: at least 2 lowerdir are needed while upperdir nonexistent\n");
1296 } else if (!ofs->config.upperdir && ofs->config.nfs_export &&
1297 ofs->config.redirect_follow) {
1298 pr_warn("overlayfs: NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n");
1299 ofs->config.nfs_export = false;
1303 stack = kcalloc(stacklen, sizeof(struct path), GFP_KERNEL);
1309 for (numlower = 0; numlower < stacklen; numlower++) {
1310 err = ovl_lower_dir(lower, &stack[numlower], ofs,
1311 &sb->s_stack_depth, &remote);
1315 lower = strchr(lower, '\0') + 1;
1319 sb->s_stack_depth++;
1320 if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) {
1321 pr_err("overlayfs: maximum fs stacking depth exceeded\n");
1325 err = ovl_get_lower_layers(ofs, stack, numlower);
1330 oe = ovl_alloc_entry(numlower);
1334 for (i = 0; i < numlower; i++) {
1335 oe->lowerstack[i].dentry = dget(stack[i].dentry);
1336 oe->lowerstack[i].layer = &ofs->lower_layers[i];
1340 sb->s_d_op = &ovl_reval_dentry_operations;
1342 sb->s_d_op = &ovl_dentry_operations;
1345 for (i = 0; i < numlower; i++)
1346 path_put(&stack[i]);
1357 static int ovl_fill_super(struct super_block *sb, void *data, int silent)
1359 struct path upperpath = { };
1360 struct dentry *root_dentry;
1361 struct ovl_entry *oe;
1367 ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL);
1371 ofs->creator_cred = cred = prepare_creds();
1375 ofs->config.index = ovl_index_def;
1376 ofs->config.nfs_export = ovl_nfs_export_def;
1377 ofs->config.xino = ovl_xino_def();
1378 err = ovl_parse_opt((char *) data, &ofs->config);
1383 if (!ofs->config.lowerdir) {
1385 pr_err("overlayfs: missing 'lowerdir'\n");
1389 sb->s_stack_depth = 0;
1390 sb->s_maxbytes = MAX_LFS_FILESIZE;
1391 /* Assume underlaying fs uses 32bit inodes unless proven otherwise */
1392 if (ofs->config.xino != OVL_XINO_OFF)
1393 ofs->xino_bits = BITS_PER_LONG - 32;
1395 if (ofs->config.upperdir) {
1396 if (!ofs->config.workdir) {
1397 pr_err("overlayfs: missing 'workdir'\n");
1401 err = ovl_get_upper(ofs, &upperpath);
1405 err = ovl_get_workdir(ofs, &upperpath);
1410 sb->s_flags |= SB_RDONLY;
1412 sb->s_stack_depth = ofs->upper_mnt->mnt_sb->s_stack_depth;
1413 sb->s_time_gran = ofs->upper_mnt->mnt_sb->s_time_gran;
1416 oe = ovl_get_lowerstack(sb, ofs);
1421 /* If the upper fs is nonexistent, we mark overlayfs r/o too */
1422 if (!ofs->upper_mnt)
1423 sb->s_flags |= SB_RDONLY;
1425 if (!(ovl_force_readonly(ofs)) && ofs->config.index) {
1426 err = ovl_get_indexdir(ofs, oe, &upperpath);
1430 /* Force r/o mount with no index dir */
1431 if (!ofs->indexdir) {
1433 ofs->workdir = NULL;
1434 sb->s_flags |= SB_RDONLY;
1439 /* Show index=off in /proc/mounts for forced r/o mount */
1440 if (!ofs->indexdir) {
1441 ofs->config.index = false;
1442 if (ofs->upper_mnt && ofs->config.nfs_export) {
1443 pr_warn("overlayfs: NFS export requires an index dir, falling back to nfs_export=off.\n");
1444 ofs->config.nfs_export = false;
1448 if (ofs->config.nfs_export)
1449 sb->s_export_op = &ovl_export_operations;
1451 /* Never override disk quota limits or use reserved space */
1452 cap_lower(cred->cap_effective, CAP_SYS_RESOURCE);
1454 sb->s_magic = OVERLAYFS_SUPER_MAGIC;
1455 sb->s_op = &ovl_super_operations;
1456 sb->s_xattr = ovl_xattr_handlers;
1457 sb->s_fs_info = ofs;
1458 sb->s_flags |= SB_POSIXACL | SB_NOREMOTELOCK;
1461 root_dentry = d_make_root(ovl_new_inode(sb, S_IFDIR, 0));
1465 root_dentry->d_fsdata = oe;
1467 mntput(upperpath.mnt);
1468 if (upperpath.dentry) {
1469 ovl_dentry_set_upper_alias(root_dentry);
1470 if (ovl_is_impuredir(upperpath.dentry))
1471 ovl_set_flag(OVL_IMPURE, d_inode(root_dentry));
1474 /* Root is always merge -> can have whiteouts */
1475 ovl_set_flag(OVL_WHITEOUTS, d_inode(root_dentry));
1476 ovl_dentry_set_flag(OVL_E_CONNECTED, root_dentry);
1477 ovl_inode_init(d_inode(root_dentry), upperpath.dentry,
1478 ovl_dentry_lower(root_dentry));
1480 sb->s_root = root_dentry;
1485 ovl_entry_stack_free(oe);
1488 path_put(&upperpath);
1494 static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags,
1495 const char *dev_name, void *raw_data)
1497 return mount_nodev(fs_type, flags, raw_data, ovl_fill_super);
1500 static struct file_system_type ovl_fs_type = {
1501 .owner = THIS_MODULE,
1504 .kill_sb = kill_anon_super,
1506 MODULE_ALIAS_FS("overlay");
1508 static void ovl_inode_init_once(void *foo)
1510 struct ovl_inode *oi = foo;
1512 inode_init_once(&oi->vfs_inode);
1515 static int __init ovl_init(void)
1519 ovl_inode_cachep = kmem_cache_create("ovl_inode",
1520 sizeof(struct ovl_inode), 0,
1521 (SLAB_RECLAIM_ACCOUNT|
1522 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1523 ovl_inode_init_once);
1524 if (ovl_inode_cachep == NULL)
1527 err = register_filesystem(&ovl_fs_type);
1529 kmem_cache_destroy(ovl_inode_cachep);
1534 static void __exit ovl_exit(void)
1536 unregister_filesystem(&ovl_fs_type);
1539 * Make sure all delayed rcu free inodes are flushed before we
1543 kmem_cache_destroy(ovl_inode_cachep);
1547 module_init(ovl_init);
1548 module_exit(ovl_exit);