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 struct dentry *ovl_d_real(struct dentry *dentry,
78 const struct inode *inode)
82 /* It's an overlay file */
83 if (inode && d_inode(dentry) == inode)
86 if (!d_is_reg(dentry)) {
87 if (!inode || inode == d_inode(dentry))
92 real = ovl_dentry_upper(dentry);
93 if (real && (!inode || inode == d_inode(real)))
96 real = ovl_dentry_lower(dentry);
100 /* Handle recursion */
101 real = d_real(real, inode);
103 if (!inode || inode == d_inode(real))
106 WARN(1, "ovl_d_real(%pd4, %s:%lu): real dentry not found\n", dentry,
107 inode ? inode->i_sb->s_id : "NULL", inode ? inode->i_ino : 0);
111 static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags)
113 struct ovl_entry *oe = dentry->d_fsdata;
117 for (i = 0; i < oe->numlower; i++) {
118 struct dentry *d = oe->lowerstack[i].dentry;
120 if (d->d_flags & DCACHE_OP_REVALIDATE) {
121 ret = d->d_op->d_revalidate(d, flags);
125 if (!(flags & LOOKUP_RCU))
134 static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags)
136 struct ovl_entry *oe = dentry->d_fsdata;
140 for (i = 0; i < oe->numlower; i++) {
141 struct dentry *d = oe->lowerstack[i].dentry;
143 if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE) {
144 ret = d->d_op->d_weak_revalidate(d, flags);
152 static const struct dentry_operations ovl_dentry_operations = {
153 .d_release = ovl_dentry_release,
154 .d_real = ovl_d_real,
157 static const struct dentry_operations ovl_reval_dentry_operations = {
158 .d_release = ovl_dentry_release,
159 .d_real = ovl_d_real,
160 .d_revalidate = ovl_dentry_revalidate,
161 .d_weak_revalidate = ovl_dentry_weak_revalidate,
164 static struct kmem_cache *ovl_inode_cachep;
166 static struct inode *ovl_alloc_inode(struct super_block *sb)
168 struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL);
177 oi->__upperdentry = NULL;
179 mutex_init(&oi->lock);
181 return &oi->vfs_inode;
184 static void ovl_i_callback(struct rcu_head *head)
186 struct inode *inode = container_of(head, struct inode, i_rcu);
188 kmem_cache_free(ovl_inode_cachep, OVL_I(inode));
191 static void ovl_destroy_inode(struct inode *inode)
193 struct ovl_inode *oi = OVL_I(inode);
195 dput(oi->__upperdentry);
198 ovl_dir_cache_free(inode);
199 mutex_destroy(&oi->lock);
201 call_rcu(&inode->i_rcu, ovl_i_callback);
204 static void ovl_free_fs(struct ovl_fs *ofs)
210 if (ofs->workdir_locked)
211 ovl_inuse_unlock(ofs->workbasedir);
212 dput(ofs->workbasedir);
213 if (ofs->upperdir_locked)
214 ovl_inuse_unlock(ofs->upper_mnt->mnt_root);
215 mntput(ofs->upper_mnt);
216 for (i = 0; i < ofs->numlower; i++)
217 mntput(ofs->lower_layers[i].mnt);
218 for (i = 0; i < ofs->numlowerfs; i++)
219 free_anon_bdev(ofs->lower_fs[i].pseudo_dev);
220 kfree(ofs->lower_layers);
221 kfree(ofs->lower_fs);
223 kfree(ofs->config.lowerdir);
224 kfree(ofs->config.upperdir);
225 kfree(ofs->config.workdir);
226 kfree(ofs->config.redirect_mode);
227 if (ofs->creator_cred)
228 put_cred(ofs->creator_cred);
232 static void ovl_put_super(struct super_block *sb)
234 struct ovl_fs *ofs = sb->s_fs_info;
239 /* Sync real dirty inodes in upper filesystem (if it exists) */
240 static int ovl_sync_fs(struct super_block *sb, int wait)
242 struct ovl_fs *ofs = sb->s_fs_info;
243 struct super_block *upper_sb;
250 * If this is a sync(2) call or an emergency sync, all the super blocks
251 * will be iterated, including upper_sb, so no need to do anything.
253 * If this is a syncfs(2) call, then we do need to call
254 * sync_filesystem() on upper_sb, but enough if we do it when being
255 * called with wait == 1.
260 upper_sb = ofs->upper_mnt->mnt_sb;
262 down_read(&upper_sb->s_umount);
263 ret = sync_filesystem(upper_sb);
264 up_read(&upper_sb->s_umount);
271 * @sb: The overlayfs super block
272 * @buf: The struct kstatfs to fill in with stats
274 * Get the filesystem statistics. As writes always target the upper layer
275 * filesystem pass the statfs to the upper filesystem (if it exists)
277 static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf)
279 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
280 struct dentry *root_dentry = dentry->d_sb->s_root;
284 ovl_path_real(root_dentry, &path);
286 err = vfs_statfs(&path, buf);
288 buf->f_namelen = ofs->namelen;
289 buf->f_type = OVERLAYFS_SUPER_MAGIC;
295 /* Will this overlay be forced to mount/remount ro? */
296 static bool ovl_force_readonly(struct ovl_fs *ofs)
298 return (!ofs->upper_mnt || !ofs->workdir);
301 static const char *ovl_redirect_mode_def(void)
303 return ovl_redirect_dir_def ? "on" : "off";
312 static const char * const ovl_xino_str[] = {
318 static inline int ovl_xino_def(void)
320 return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF;
326 * Prints the mount options for a given superblock.
327 * Returns zero; does not fail.
329 static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
331 struct super_block *sb = dentry->d_sb;
332 struct ovl_fs *ofs = sb->s_fs_info;
334 seq_show_option(m, "lowerdir", ofs->config.lowerdir);
335 if (ofs->config.upperdir) {
336 seq_show_option(m, "upperdir", ofs->config.upperdir);
337 seq_show_option(m, "workdir", ofs->config.workdir);
339 if (ofs->config.default_permissions)
340 seq_puts(m, ",default_permissions");
341 if (strcmp(ofs->config.redirect_mode, ovl_redirect_mode_def()) != 0)
342 seq_printf(m, ",redirect_dir=%s", ofs->config.redirect_mode);
343 if (ofs->config.index != ovl_index_def)
344 seq_printf(m, ",index=%s", ofs->config.index ? "on" : "off");
345 if (ofs->config.nfs_export != ovl_nfs_export_def)
346 seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ?
348 if (ofs->config.xino != ovl_xino_def())
349 seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]);
353 static int ovl_remount(struct super_block *sb, int *flags, char *data)
355 struct ovl_fs *ofs = sb->s_fs_info;
357 if (!(*flags & SB_RDONLY) && ovl_force_readonly(ofs))
363 static const struct super_operations ovl_super_operations = {
364 .alloc_inode = ovl_alloc_inode,
365 .destroy_inode = ovl_destroy_inode,
366 .drop_inode = generic_delete_inode,
367 .put_super = ovl_put_super,
368 .sync_fs = ovl_sync_fs,
369 .statfs = ovl_statfs,
370 .show_options = ovl_show_options,
371 .remount_fs = ovl_remount,
378 OPT_DEFAULT_PERMISSIONS,
390 static const match_table_t ovl_tokens = {
391 {OPT_LOWERDIR, "lowerdir=%s"},
392 {OPT_UPPERDIR, "upperdir=%s"},
393 {OPT_WORKDIR, "workdir=%s"},
394 {OPT_DEFAULT_PERMISSIONS, "default_permissions"},
395 {OPT_REDIRECT_DIR, "redirect_dir=%s"},
396 {OPT_INDEX_ON, "index=on"},
397 {OPT_INDEX_OFF, "index=off"},
398 {OPT_NFS_EXPORT_ON, "nfs_export=on"},
399 {OPT_NFS_EXPORT_OFF, "nfs_export=off"},
400 {OPT_XINO_ON, "xino=on"},
401 {OPT_XINO_OFF, "xino=off"},
402 {OPT_XINO_AUTO, "xino=auto"},
406 static char *ovl_next_opt(char **s)
414 for (p = sbegin; *p; p++) {
419 } else if (*p == ',') {
429 static int ovl_parse_redirect_mode(struct ovl_config *config, const char *mode)
431 if (strcmp(mode, "on") == 0) {
432 config->redirect_dir = true;
434 * Does not make sense to have redirect creation without
435 * redirect following.
437 config->redirect_follow = true;
438 } else if (strcmp(mode, "follow") == 0) {
439 config->redirect_follow = true;
440 } else if (strcmp(mode, "off") == 0) {
441 if (ovl_redirect_always_follow)
442 config->redirect_follow = true;
443 } else if (strcmp(mode, "nofollow") != 0) {
444 pr_err("overlayfs: bad mount option \"redirect_dir=%s\"\n",
452 static int ovl_parse_opt(char *opt, struct ovl_config *config)
456 config->redirect_mode = kstrdup(ovl_redirect_mode_def(), GFP_KERNEL);
457 if (!config->redirect_mode)
460 while ((p = ovl_next_opt(&opt)) != NULL) {
462 substring_t args[MAX_OPT_ARGS];
467 token = match_token(p, ovl_tokens, args);
470 kfree(config->upperdir);
471 config->upperdir = match_strdup(&args[0]);
472 if (!config->upperdir)
477 kfree(config->lowerdir);
478 config->lowerdir = match_strdup(&args[0]);
479 if (!config->lowerdir)
484 kfree(config->workdir);
485 config->workdir = match_strdup(&args[0]);
486 if (!config->workdir)
490 case OPT_DEFAULT_PERMISSIONS:
491 config->default_permissions = true;
494 case OPT_REDIRECT_DIR:
495 kfree(config->redirect_mode);
496 config->redirect_mode = match_strdup(&args[0]);
497 if (!config->redirect_mode)
502 config->index = true;
506 config->index = false;
509 case OPT_NFS_EXPORT_ON:
510 config->nfs_export = true;
513 case OPT_NFS_EXPORT_OFF:
514 config->nfs_export = false;
518 config->xino = OVL_XINO_ON;
522 config->xino = OVL_XINO_OFF;
526 config->xino = OVL_XINO_AUTO;
530 pr_err("overlayfs: unrecognized mount option \"%s\" or missing value\n", p);
535 /* Workdir is useless in non-upper mount */
536 if (!config->upperdir && config->workdir) {
537 pr_info("overlayfs: option \"workdir=%s\" is useless in a non-upper mount, ignore\n",
539 kfree(config->workdir);
540 config->workdir = NULL;
543 return ovl_parse_redirect_mode(config, config->redirect_mode);
546 #define OVL_WORKDIR_NAME "work"
547 #define OVL_INDEXDIR_NAME "index"
549 static struct dentry *ovl_workdir_create(struct ovl_fs *ofs,
550 const char *name, bool persist)
552 struct inode *dir = ofs->workbasedir->d_inode;
553 struct vfsmount *mnt = ofs->upper_mnt;
556 bool retried = false;
559 inode_lock_nested(dir, I_MUTEX_PARENT);
563 work = lookup_one_len(name, ofs->workbasedir, strlen(name));
566 struct iattr attr = {
567 .ia_valid = ATTR_MODE,
568 .ia_mode = S_IFDIR | 0,
580 ovl_workdir_cleanup(dir, mnt, work, 0);
585 work = ovl_create_real(dir, work, OVL_CATTR(attr.ia_mode));
591 * Try to remove POSIX ACL xattrs from workdir. We are good if:
593 * a) success (there was a POSIX ACL xattr and was removed)
594 * b) -ENODATA (there was no POSIX ACL xattr)
595 * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported)
597 * There are various other error values that could effectively
598 * mean that the xattr doesn't exist (e.g. -ERANGE is returned
599 * if the xattr name is too long), but the set of filesystems
600 * allowed as upper are limited to "normal" ones, where checking
601 * for the above two errors is sufficient.
603 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT);
604 if (err && err != -ENODATA && err != -EOPNOTSUPP)
607 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS);
608 if (err && err != -ENODATA && err != -EOPNOTSUPP)
611 /* Clear any inherited mode bits */
612 inode_lock(work->d_inode);
613 err = notify_change(work, &attr, NULL);
614 inode_unlock(work->d_inode);
630 pr_warn("overlayfs: failed to create directory %s/%s (errno: %i); mounting read-only\n",
631 ofs->config.workdir, name, -err);
636 static void ovl_unescape(char *s)
649 static int ovl_mount_dir_noesc(const char *name, struct path *path)
654 pr_err("overlayfs: empty lowerdir\n");
657 err = kern_path(name, LOOKUP_FOLLOW, path);
659 pr_err("overlayfs: failed to resolve '%s': %i\n", name, err);
663 if (ovl_dentry_weird(path->dentry)) {
664 pr_err("overlayfs: filesystem on '%s' not supported\n", name);
667 if (!d_is_dir(path->dentry)) {
668 pr_err("overlayfs: '%s' not a directory\n", name);
679 static int ovl_mount_dir(const char *name, struct path *path)
682 char *tmp = kstrdup(name, GFP_KERNEL);
686 err = ovl_mount_dir_noesc(tmp, path);
689 if (ovl_dentry_remote(path->dentry)) {
690 pr_err("overlayfs: filesystem on '%s' not supported as upperdir\n",
700 static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs,
703 struct kstatfs statfs;
704 int err = vfs_statfs(path, &statfs);
707 pr_err("overlayfs: statfs failed on '%s'\n", name);
709 ofs->namelen = max(ofs->namelen, statfs.f_namelen);
714 static int ovl_lower_dir(const char *name, struct path *path,
715 struct ovl_fs *ofs, int *stack_depth, bool *remote)
720 err = ovl_mount_dir_noesc(name, path);
724 err = ovl_check_namelen(path, ofs, name);
728 *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth);
730 if (ovl_dentry_remote(path->dentry))
734 * The inodes index feature and NFS export need to encode and decode
735 * file handles, so they require that all layers support them.
737 fh_type = ovl_can_decode_fh(path->dentry->d_sb);
738 if ((ofs->config.nfs_export ||
739 (ofs->config.index && ofs->config.upperdir)) && !fh_type) {
740 ofs->config.index = false;
741 ofs->config.nfs_export = false;
742 pr_warn("overlayfs: fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n",
746 /* Check if lower fs has 32bit inode numbers */
747 if (fh_type != FILEID_INO32_GEN)
758 /* Workdir should not be subdir of upperdir and vice versa */
759 static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir)
763 if (workdir != upperdir) {
764 ok = (lock_rename(workdir, upperdir) == NULL);
765 unlock_rename(workdir, upperdir);
770 static unsigned int ovl_split_lowerdirs(char *str)
772 unsigned int ctr = 1;
775 for (s = d = str;; s++, d++) {
778 } else if (*s == ':') {
790 static int __maybe_unused
791 ovl_posix_acl_xattr_get(const struct xattr_handler *handler,
792 struct dentry *dentry, struct inode *inode,
793 const char *name, void *buffer, size_t size)
795 return ovl_xattr_get(dentry, inode, handler->name, buffer, size);
798 static int __maybe_unused
799 ovl_posix_acl_xattr_set(const struct xattr_handler *handler,
800 struct dentry *dentry, struct inode *inode,
801 const char *name, const void *value,
802 size_t size, int flags)
804 struct dentry *workdir = ovl_workdir(dentry);
805 struct inode *realinode = ovl_inode_real(inode);
806 struct posix_acl *acl = NULL;
809 /* Check that everything is OK before copy-up */
811 acl = posix_acl_from_xattr(&init_user_ns, value, size);
816 if (!IS_POSIXACL(d_inode(workdir)))
817 goto out_acl_release;
818 if (!realinode->i_op->set_acl)
819 goto out_acl_release;
820 if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) {
821 err = acl ? -EACCES : 0;
822 goto out_acl_release;
825 if (!inode_owner_or_capable(inode))
826 goto out_acl_release;
828 posix_acl_release(acl);
831 * Check if sgid bit needs to be cleared (actual setacl operation will
832 * be done with mounter's capabilities and so that won't do it for us).
834 if (unlikely(inode->i_mode & S_ISGID) &&
835 handler->flags == ACL_TYPE_ACCESS &&
836 !in_group_p(inode->i_gid) &&
837 !capable_wrt_inode_uidgid(inode, CAP_FSETID)) {
838 struct iattr iattr = { .ia_valid = ATTR_KILL_SGID };
840 err = ovl_setattr(dentry, &iattr);
845 err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags);
847 ovl_copyattr(ovl_inode_real(inode), inode);
852 posix_acl_release(acl);
856 static int ovl_own_xattr_get(const struct xattr_handler *handler,
857 struct dentry *dentry, struct inode *inode,
858 const char *name, void *buffer, size_t size)
863 static int ovl_own_xattr_set(const struct xattr_handler *handler,
864 struct dentry *dentry, struct inode *inode,
865 const char *name, const void *value,
866 size_t size, int flags)
871 static int ovl_other_xattr_get(const struct xattr_handler *handler,
872 struct dentry *dentry, struct inode *inode,
873 const char *name, void *buffer, size_t size)
875 return ovl_xattr_get(dentry, inode, name, buffer, size);
878 static int ovl_other_xattr_set(const struct xattr_handler *handler,
879 struct dentry *dentry, struct inode *inode,
880 const char *name, const void *value,
881 size_t size, int flags)
883 return ovl_xattr_set(dentry, inode, name, value, size, flags);
886 static const struct xattr_handler __maybe_unused
887 ovl_posix_acl_access_xattr_handler = {
888 .name = XATTR_NAME_POSIX_ACL_ACCESS,
889 .flags = ACL_TYPE_ACCESS,
890 .get = ovl_posix_acl_xattr_get,
891 .set = ovl_posix_acl_xattr_set,
894 static const struct xattr_handler __maybe_unused
895 ovl_posix_acl_default_xattr_handler = {
896 .name = XATTR_NAME_POSIX_ACL_DEFAULT,
897 .flags = ACL_TYPE_DEFAULT,
898 .get = ovl_posix_acl_xattr_get,
899 .set = ovl_posix_acl_xattr_set,
902 static const struct xattr_handler ovl_own_xattr_handler = {
903 .prefix = OVL_XATTR_PREFIX,
904 .get = ovl_own_xattr_get,
905 .set = ovl_own_xattr_set,
908 static const struct xattr_handler ovl_other_xattr_handler = {
909 .prefix = "", /* catch all */
910 .get = ovl_other_xattr_get,
911 .set = ovl_other_xattr_set,
914 static const struct xattr_handler *ovl_xattr_handlers[] = {
915 #ifdef CONFIG_FS_POSIX_ACL
916 &ovl_posix_acl_access_xattr_handler,
917 &ovl_posix_acl_default_xattr_handler,
919 &ovl_own_xattr_handler,
920 &ovl_other_xattr_handler,
924 static int ovl_get_upper(struct ovl_fs *ofs, struct path *upperpath)
926 struct vfsmount *upper_mnt;
929 err = ovl_mount_dir(ofs->config.upperdir, upperpath);
933 /* Upper fs should not be r/o */
934 if (sb_rdonly(upperpath->mnt->mnt_sb)) {
935 pr_err("overlayfs: upper fs is r/o, try multi-lower layers mount\n");
940 err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir);
945 if (ovl_inuse_trylock(upperpath->dentry)) {
946 ofs->upperdir_locked = true;
947 } else if (ofs->config.index) {
948 pr_err("overlayfs: upperdir is in-use by another mount, mount with '-o index=off' to override exclusive upperdir protection.\n");
951 pr_warn("overlayfs: upperdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
954 upper_mnt = clone_private_mount(upperpath);
955 err = PTR_ERR(upper_mnt);
956 if (IS_ERR(upper_mnt)) {
957 pr_err("overlayfs: failed to clone upperpath\n");
961 /* Don't inherit atime flags */
962 upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME);
963 ofs->upper_mnt = upper_mnt;
969 static int ovl_make_workdir(struct ovl_fs *ofs, struct path *workpath)
971 struct vfsmount *mnt = ofs->upper_mnt;
976 err = mnt_want_write(mnt);
980 ofs->workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false);
985 * Upper should support d_type, else whiteouts are visible. Given
986 * workdir and upper are on same fs, we can do iterate_dir() on
987 * workdir. This check requires successful creation of workdir in
990 err = ovl_check_d_type_supported(workpath);
995 * We allowed this configuration and don't want to break users over
996 * kernel upgrade. So warn instead of erroring out.
999 pr_warn("overlayfs: upper fs needs to support d_type.\n");
1001 /* Check if upper/work fs supports O_TMPFILE */
1002 temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0);
1003 ofs->tmpfile = !IS_ERR(temp);
1007 pr_warn("overlayfs: upper fs does not support tmpfile.\n");
1010 * Check if upper/work fs supports trusted.overlay.* xattr
1012 err = ovl_do_setxattr(ofs->workdir, OVL_XATTR_OPAQUE, "0", 1, 0);
1014 ofs->noxattr = true;
1015 ofs->config.index = false;
1016 pr_warn("overlayfs: upper fs does not support xattr, falling back to index=off.\n");
1019 vfs_removexattr(ofs->workdir, OVL_XATTR_OPAQUE);
1022 /* Check if upper/work fs supports file handles */
1023 fh_type = ovl_can_decode_fh(ofs->workdir->d_sb);
1024 if (ofs->config.index && !fh_type) {
1025 ofs->config.index = false;
1026 pr_warn("overlayfs: upper fs does not support file handles, falling back to index=off.\n");
1029 /* Check if upper fs has 32bit inode numbers */
1030 if (fh_type != FILEID_INO32_GEN)
1033 /* NFS export of r/w mount depends on index */
1034 if (ofs->config.nfs_export && !ofs->config.index) {
1035 pr_warn("overlayfs: NFS export requires \"index=on\", falling back to nfs_export=off.\n");
1036 ofs->config.nfs_export = false;
1040 mnt_drop_write(mnt);
1044 static int ovl_get_workdir(struct ovl_fs *ofs, struct path *upperpath)
1047 struct path workpath = { };
1049 err = ovl_mount_dir(ofs->config.workdir, &workpath);
1054 if (upperpath->mnt != workpath.mnt) {
1055 pr_err("overlayfs: workdir and upperdir must reside under the same mount\n");
1058 if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) {
1059 pr_err("overlayfs: workdir and upperdir must be separate subtrees\n");
1064 if (ovl_inuse_trylock(workpath.dentry)) {
1065 ofs->workdir_locked = true;
1066 } else if (ofs->config.index) {
1067 pr_err("overlayfs: workdir is in-use by another mount, mount with '-o index=off' to override exclusive workdir protection.\n");
1070 pr_warn("overlayfs: workdir is in-use by another mount, accessing files from both mounts will result in undefined behavior.\n");
1073 ofs->workbasedir = dget(workpath.dentry);
1074 err = ovl_make_workdir(ofs, &workpath);
1080 path_put(&workpath);
1085 static int ovl_get_indexdir(struct ovl_fs *ofs, struct ovl_entry *oe,
1086 struct path *upperpath)
1088 struct vfsmount *mnt = ofs->upper_mnt;
1091 err = mnt_want_write(mnt);
1095 /* Verify lower root is upper root origin */
1096 err = ovl_verify_origin(upperpath->dentry, oe->lowerstack[0].dentry,
1099 pr_err("overlayfs: failed to verify upper root origin\n");
1103 ofs->indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true);
1104 if (ofs->indexdir) {
1106 * Verify upper root is exclusively associated with index dir.
1107 * Older kernels stored upper fh in "trusted.overlay.origin"
1108 * xattr. If that xattr exists, verify that it is a match to
1109 * upper dir file handle. In any case, verify or set xattr
1110 * "trusted.overlay.upper" to indicate that index may have
1111 * directory entries.
1113 if (ovl_check_origin_xattr(ofs->indexdir)) {
1114 err = ovl_verify_set_fh(ofs->indexdir, OVL_XATTR_ORIGIN,
1115 upperpath->dentry, true, false);
1117 pr_err("overlayfs: failed to verify index dir 'origin' xattr\n");
1119 err = ovl_verify_upper(ofs->indexdir, upperpath->dentry, true);
1121 pr_err("overlayfs: failed to verify index dir 'upper' xattr\n");
1123 /* Cleanup bad/stale/orphan index entries */
1125 err = ovl_indexdir_cleanup(ofs);
1127 if (err || !ofs->indexdir)
1128 pr_warn("overlayfs: try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
1131 mnt_drop_write(mnt);
1135 /* Get a unique fsid for the layer */
1136 static int ovl_get_fsid(struct ovl_fs *ofs, struct super_block *sb)
1142 /* fsid 0 is reserved for upper fs even with non upper overlay */
1143 if (ofs->upper_mnt && ofs->upper_mnt->mnt_sb == sb)
1146 for (i = 0; i < ofs->numlowerfs; i++) {
1147 if (ofs->lower_fs[i].sb == sb)
1151 err = get_anon_bdev(&dev);
1153 pr_err("overlayfs: failed to get anonymous bdev for lowerpath\n");
1157 ofs->lower_fs[ofs->numlowerfs].sb = sb;
1158 ofs->lower_fs[ofs->numlowerfs].pseudo_dev = dev;
1161 return ofs->numlowerfs;
1164 static int ovl_get_lower_layers(struct ovl_fs *ofs, struct path *stack,
1165 unsigned int numlower)
1171 ofs->lower_layers = kcalloc(numlower, sizeof(struct ovl_layer),
1173 if (ofs->lower_layers == NULL)
1176 ofs->lower_fs = kcalloc(numlower, sizeof(struct ovl_sb),
1178 if (ofs->lower_fs == NULL)
1181 for (i = 0; i < numlower; i++) {
1182 struct vfsmount *mnt;
1185 err = fsid = ovl_get_fsid(ofs, stack[i].mnt->mnt_sb);
1189 mnt = clone_private_mount(&stack[i]);
1192 pr_err("overlayfs: failed to clone lowerpath\n");
1197 * Make lower layers R/O. That way fchmod/fchown on lower file
1198 * will fail instead of modifying lower fs.
1200 mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME;
1202 ofs->lower_layers[ofs->numlower].mnt = mnt;
1203 ofs->lower_layers[ofs->numlower].idx = i + 1;
1204 ofs->lower_layers[ofs->numlower].fsid = fsid;
1206 ofs->lower_layers[ofs->numlower].fs =
1207 &ofs->lower_fs[fsid - 1];
1213 * When all layers on same fs, overlay can use real inode numbers.
1214 * With mount option "xino=on", mounter declares that there are enough
1215 * free high bits in underlying fs to hold the unique fsid.
1216 * If overlayfs does encounter underlying inodes using the high xino
1217 * bits reserved for fsid, it emits a warning and uses the original
1220 if (!ofs->numlowerfs || (ofs->numlowerfs == 1 && !ofs->upper_mnt)) {
1222 ofs->config.xino = OVL_XINO_OFF;
1223 } else if (ofs->config.xino == OVL_XINO_ON && !ofs->xino_bits) {
1225 * This is a roundup of number of bits needed for numlowerfs+1
1226 * (i.e. ilog2(numlowerfs+1 - 1) + 1). fsid 0 is reserved for
1227 * upper fs even with non upper overlay.
1229 BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 31);
1230 ofs->xino_bits = ilog2(ofs->numlowerfs) + 1;
1233 if (ofs->xino_bits) {
1234 pr_info("overlayfs: \"xino\" feature enabled using %d upper inode bits.\n",
1243 static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb,
1247 char *lowertmp, *lower;
1248 struct path *stack = NULL;
1249 unsigned int stacklen, numlower = 0, i;
1250 bool remote = false;
1251 struct ovl_entry *oe;
1254 lowertmp = kstrdup(ofs->config.lowerdir, GFP_KERNEL);
1259 stacklen = ovl_split_lowerdirs(lowertmp);
1260 if (stacklen > OVL_MAX_STACK) {
1261 pr_err("overlayfs: too many lower directories, limit is %d\n",
1264 } else if (!ofs->config.upperdir && stacklen == 1) {
1265 pr_err("overlayfs: at least 2 lowerdir are needed while upperdir nonexistent\n");
1267 } else if (!ofs->config.upperdir && ofs->config.nfs_export &&
1268 ofs->config.redirect_follow) {
1269 pr_warn("overlayfs: NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n");
1270 ofs->config.nfs_export = false;
1274 stack = kcalloc(stacklen, sizeof(struct path), GFP_KERNEL);
1280 for (numlower = 0; numlower < stacklen; numlower++) {
1281 err = ovl_lower_dir(lower, &stack[numlower], ofs,
1282 &sb->s_stack_depth, &remote);
1286 lower = strchr(lower, '\0') + 1;
1290 sb->s_stack_depth++;
1291 if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) {
1292 pr_err("overlayfs: maximum fs stacking depth exceeded\n");
1296 err = ovl_get_lower_layers(ofs, stack, numlower);
1301 oe = ovl_alloc_entry(numlower);
1305 for (i = 0; i < numlower; i++) {
1306 oe->lowerstack[i].dentry = dget(stack[i].dentry);
1307 oe->lowerstack[i].layer = &ofs->lower_layers[i];
1311 sb->s_d_op = &ovl_reval_dentry_operations;
1313 sb->s_d_op = &ovl_dentry_operations;
1316 for (i = 0; i < numlower; i++)
1317 path_put(&stack[i]);
1328 static int ovl_fill_super(struct super_block *sb, void *data, int silent)
1330 struct path upperpath = { };
1331 struct dentry *root_dentry;
1332 struct ovl_entry *oe;
1338 ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL);
1342 ofs->creator_cred = cred = prepare_creds();
1346 ofs->config.index = ovl_index_def;
1347 ofs->config.nfs_export = ovl_nfs_export_def;
1348 ofs->config.xino = ovl_xino_def();
1349 err = ovl_parse_opt((char *) data, &ofs->config);
1354 if (!ofs->config.lowerdir) {
1356 pr_err("overlayfs: missing 'lowerdir'\n");
1360 sb->s_stack_depth = 0;
1361 sb->s_maxbytes = MAX_LFS_FILESIZE;
1362 /* Assume underlaying fs uses 32bit inodes unless proven otherwise */
1363 if (ofs->config.xino != OVL_XINO_OFF)
1364 ofs->xino_bits = BITS_PER_LONG - 32;
1366 if (ofs->config.upperdir) {
1367 if (!ofs->config.workdir) {
1368 pr_err("overlayfs: missing 'workdir'\n");
1372 err = ovl_get_upper(ofs, &upperpath);
1376 err = ovl_get_workdir(ofs, &upperpath);
1381 sb->s_flags |= SB_RDONLY;
1383 sb->s_stack_depth = ofs->upper_mnt->mnt_sb->s_stack_depth;
1384 sb->s_time_gran = ofs->upper_mnt->mnt_sb->s_time_gran;
1387 oe = ovl_get_lowerstack(sb, ofs);
1392 /* If the upper fs is nonexistent, we mark overlayfs r/o too */
1393 if (!ofs->upper_mnt)
1394 sb->s_flags |= SB_RDONLY;
1396 if (!(ovl_force_readonly(ofs)) && ofs->config.index) {
1397 err = ovl_get_indexdir(ofs, oe, &upperpath);
1401 /* Force r/o mount with no index dir */
1402 if (!ofs->indexdir) {
1404 ofs->workdir = NULL;
1405 sb->s_flags |= SB_RDONLY;
1410 /* Show index=off in /proc/mounts for forced r/o mount */
1411 if (!ofs->indexdir) {
1412 ofs->config.index = false;
1413 if (ofs->upper_mnt && ofs->config.nfs_export) {
1414 pr_warn("overlayfs: NFS export requires an index dir, falling back to nfs_export=off.\n");
1415 ofs->config.nfs_export = false;
1419 if (ofs->config.nfs_export)
1420 sb->s_export_op = &ovl_export_operations;
1422 /* Never override disk quota limits or use reserved space */
1423 cap_lower(cred->cap_effective, CAP_SYS_RESOURCE);
1425 sb->s_magic = OVERLAYFS_SUPER_MAGIC;
1426 sb->s_op = &ovl_super_operations;
1427 sb->s_xattr = ovl_xattr_handlers;
1428 sb->s_fs_info = ofs;
1429 sb->s_flags |= SB_POSIXACL;
1432 root_dentry = d_make_root(ovl_new_inode(sb, S_IFDIR, 0));
1436 root_dentry->d_fsdata = oe;
1438 mntput(upperpath.mnt);
1439 if (upperpath.dentry) {
1440 ovl_dentry_set_upper_alias(root_dentry);
1441 if (ovl_is_impuredir(upperpath.dentry))
1442 ovl_set_flag(OVL_IMPURE, d_inode(root_dentry));
1445 /* Root is always merge -> can have whiteouts */
1446 ovl_set_flag(OVL_WHITEOUTS, d_inode(root_dentry));
1447 ovl_dentry_set_flag(OVL_E_CONNECTED, root_dentry);
1448 ovl_inode_init(d_inode(root_dentry), upperpath.dentry,
1449 ovl_dentry_lower(root_dentry));
1451 sb->s_root = root_dentry;
1456 ovl_entry_stack_free(oe);
1459 path_put(&upperpath);
1465 static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags,
1466 const char *dev_name, void *raw_data)
1468 return mount_nodev(fs_type, flags, raw_data, ovl_fill_super);
1471 static struct file_system_type ovl_fs_type = {
1472 .owner = THIS_MODULE,
1475 .kill_sb = kill_anon_super,
1477 MODULE_ALIAS_FS("overlay");
1479 static void ovl_inode_init_once(void *foo)
1481 struct ovl_inode *oi = foo;
1483 inode_init_once(&oi->vfs_inode);
1486 static int __init ovl_init(void)
1490 ovl_inode_cachep = kmem_cache_create("ovl_inode",
1491 sizeof(struct ovl_inode), 0,
1492 (SLAB_RECLAIM_ACCOUNT|
1493 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1494 ovl_inode_init_once);
1495 if (ovl_inode_cachep == NULL)
1498 err = register_filesystem(&ovl_fs_type);
1500 kmem_cache_destroy(ovl_inode_cachep);
1505 static void __exit ovl_exit(void)
1507 unregister_filesystem(&ovl_fs_type);
1510 * Make sure all delayed rcu free inodes are flushed before we
1514 kmem_cache_destroy(ovl_inode_cachep);
1518 module_init(ovl_init);
1519 module_exit(ovl_exit);