1 // SPDX-License-Identifier: GPL-2.0-only
4 * Copyright (C) 2011 Novell Inc.
7 #include <uapi/linux/magic.h>
9 #include <linux/namei.h>
10 #include <linux/xattr.h>
11 #include <linux/mount.h>
12 #include <linux/parser.h>
13 #include <linux/module.h>
14 #include <linux/statfs.h>
15 #include <linux/seq_file.h>
16 #include <linux/posix_acl_xattr.h>
17 #include <linux/exportfs.h>
18 #include "overlayfs.h"
20 MODULE_AUTHOR("Miklos Szeredi <miklos@szeredi.hu>");
21 MODULE_DESCRIPTION("Overlay filesystem");
22 MODULE_LICENSE("GPL");
27 #define OVL_MAX_STACK 500
29 static bool ovl_redirect_dir_def = IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_DIR);
30 module_param_named(redirect_dir, ovl_redirect_dir_def, bool, 0644);
31 MODULE_PARM_DESC(redirect_dir,
32 "Default to on or off for the redirect_dir feature");
34 static bool ovl_redirect_always_follow =
35 IS_ENABLED(CONFIG_OVERLAY_FS_REDIRECT_ALWAYS_FOLLOW);
36 module_param_named(redirect_always_follow, ovl_redirect_always_follow,
38 MODULE_PARM_DESC(redirect_always_follow,
39 "Follow redirects even if redirect_dir feature is turned off");
41 static bool ovl_index_def = IS_ENABLED(CONFIG_OVERLAY_FS_INDEX);
42 module_param_named(index, ovl_index_def, bool, 0644);
43 MODULE_PARM_DESC(index,
44 "Default to on or off for the inodes index feature");
46 static bool ovl_nfs_export_def = IS_ENABLED(CONFIG_OVERLAY_FS_NFS_EXPORT);
47 module_param_named(nfs_export, ovl_nfs_export_def, bool, 0644);
48 MODULE_PARM_DESC(nfs_export,
49 "Default to on or off for the NFS export feature");
51 static bool ovl_xino_auto_def = IS_ENABLED(CONFIG_OVERLAY_FS_XINO_AUTO);
52 module_param_named(xino_auto, ovl_xino_auto_def, bool, 0644);
53 MODULE_PARM_DESC(xino_auto,
54 "Auto enable xino feature");
56 static void ovl_entry_stack_free(struct ovl_entry *oe)
60 for (i = 0; i < oe->numlower; i++)
61 dput(oe->lowerstack[i].dentry);
64 static bool ovl_metacopy_def = IS_ENABLED(CONFIG_OVERLAY_FS_METACOPY);
65 module_param_named(metacopy, ovl_metacopy_def, bool, 0644);
66 MODULE_PARM_DESC(metacopy,
67 "Default to on or off for the metadata only copy up feature");
69 static void ovl_dentry_release(struct dentry *dentry)
71 struct ovl_entry *oe = dentry->d_fsdata;
74 ovl_entry_stack_free(oe);
79 static struct dentry *ovl_d_real(struct dentry *dentry,
80 const struct inode *inode)
82 struct dentry *real = NULL, *lower;
84 /* It's an overlay file */
85 if (inode && d_inode(dentry) == inode)
88 if (!d_is_reg(dentry)) {
89 if (!inode || inode == d_inode(dentry))
94 real = ovl_dentry_upper(dentry);
95 if (real && (inode == d_inode(real)))
98 if (real && !inode && ovl_has_upperdata(d_inode(dentry)))
101 lower = ovl_dentry_lowerdata(dentry);
106 /* Handle recursion */
107 real = d_real(real, inode);
109 if (!inode || inode == d_inode(real))
112 WARN(1, "%s(%pd4, %s:%lu): real dentry (%p/%lu) not found\n",
113 __func__, dentry, inode ? inode->i_sb->s_id : "NULL",
114 inode ? inode->i_ino : 0, real,
115 real && d_inode(real) ? d_inode(real)->i_ino : 0);
119 static int ovl_revalidate_real(struct dentry *d, unsigned int flags, bool weak)
124 if (d->d_flags & DCACHE_OP_WEAK_REVALIDATE)
125 ret = d->d_op->d_weak_revalidate(d, flags);
126 } else if (d->d_flags & DCACHE_OP_REVALIDATE) {
127 ret = d->d_op->d_revalidate(d, flags);
129 if (!(flags & LOOKUP_RCU))
137 static int ovl_dentry_revalidate_common(struct dentry *dentry,
138 unsigned int flags, bool weak)
140 struct ovl_entry *oe = dentry->d_fsdata;
141 struct dentry *upper;
145 upper = ovl_dentry_upper(dentry);
147 ret = ovl_revalidate_real(upper, flags, weak);
149 for (i = 0; ret > 0 && i < oe->numlower; i++) {
150 ret = ovl_revalidate_real(oe->lowerstack[i].dentry, flags,
156 static int ovl_dentry_revalidate(struct dentry *dentry, unsigned int flags)
158 return ovl_dentry_revalidate_common(dentry, flags, false);
161 static int ovl_dentry_weak_revalidate(struct dentry *dentry, unsigned int flags)
163 return ovl_dentry_revalidate_common(dentry, flags, true);
166 static const struct dentry_operations ovl_dentry_operations = {
167 .d_release = ovl_dentry_release,
168 .d_real = ovl_d_real,
169 .d_revalidate = ovl_dentry_revalidate,
170 .d_weak_revalidate = ovl_dentry_weak_revalidate,
173 static struct kmem_cache *ovl_inode_cachep;
175 static struct inode *ovl_alloc_inode(struct super_block *sb)
177 struct ovl_inode *oi = kmem_cache_alloc(ovl_inode_cachep, GFP_KERNEL);
186 oi->__upperdentry = NULL;
188 oi->lowerdata = NULL;
189 mutex_init(&oi->lock);
191 return &oi->vfs_inode;
194 static void ovl_free_inode(struct inode *inode)
196 struct ovl_inode *oi = OVL_I(inode);
199 mutex_destroy(&oi->lock);
200 kmem_cache_free(ovl_inode_cachep, oi);
203 static void ovl_destroy_inode(struct inode *inode)
205 struct ovl_inode *oi = OVL_I(inode);
207 dput(oi->__upperdentry);
209 if (S_ISDIR(inode->i_mode))
210 ovl_dir_cache_free(inode);
215 static void ovl_free_fs(struct ovl_fs *ofs)
217 struct vfsmount **mounts;
220 iput(ofs->workbasedir_trap);
221 iput(ofs->indexdir_trap);
222 iput(ofs->workdir_trap);
226 if (ofs->workdir_locked)
227 ovl_inuse_unlock(ofs->workbasedir);
228 dput(ofs->workbasedir);
229 if (ofs->upperdir_locked)
230 ovl_inuse_unlock(ovl_upper_mnt(ofs)->mnt_root);
232 /* Hack! Reuse ofs->layers as a vfsmount array before freeing it */
233 mounts = (struct vfsmount **) ofs->layers;
234 for (i = 0; i < ofs->numlayer; i++) {
235 iput(ofs->layers[i].trap);
236 mounts[i] = ofs->layers[i].mnt;
238 kern_unmount_array(mounts, ofs->numlayer);
240 for (i = 0; i < ofs->numfs; i++)
241 free_anon_bdev(ofs->fs[i].pseudo_dev);
244 kfree(ofs->config.lowerdir);
245 kfree(ofs->config.upperdir);
246 kfree(ofs->config.workdir);
247 kfree(ofs->config.redirect_mode);
248 if (ofs->creator_cred)
249 put_cred(ofs->creator_cred);
253 static void ovl_put_super(struct super_block *sb)
255 struct ovl_fs *ofs = sb->s_fs_info;
260 /* Sync real dirty inodes in upper filesystem (if it exists) */
261 static int ovl_sync_fs(struct super_block *sb, int wait)
263 struct ovl_fs *ofs = sb->s_fs_info;
264 struct super_block *upper_sb;
267 ret = ovl_sync_status(ofs);
269 * We have to always set the err, because the return value isn't
270 * checked in syncfs, and instead indirectly return an error via
271 * the sb's writeback errseq, which VFS inspects after this call.
274 errseq_set(&sb->s_wb_err, -EIO);
282 * Not called for sync(2) call or an emergency sync (SB_I_SKIP_SYNC).
283 * All the super blocks will be iterated, including upper_sb.
285 * If this is a syncfs(2) call, then we do need to call
286 * sync_filesystem() on upper_sb, but enough if we do it when being
287 * called with wait == 1.
292 upper_sb = ovl_upper_mnt(ofs)->mnt_sb;
294 down_read(&upper_sb->s_umount);
295 ret = sync_filesystem(upper_sb);
296 up_read(&upper_sb->s_umount);
303 * @sb: The overlayfs super block
304 * @buf: The struct kstatfs to fill in with stats
306 * Get the filesystem statistics. As writes always target the upper layer
307 * filesystem pass the statfs to the upper filesystem (if it exists)
309 static int ovl_statfs(struct dentry *dentry, struct kstatfs *buf)
311 struct ovl_fs *ofs = dentry->d_sb->s_fs_info;
312 struct dentry *root_dentry = dentry->d_sb->s_root;
316 ovl_path_real(root_dentry, &path);
318 err = vfs_statfs(&path, buf);
320 buf->f_namelen = ofs->namelen;
321 buf->f_type = OVERLAYFS_SUPER_MAGIC;
327 /* Will this overlay be forced to mount/remount ro? */
328 static bool ovl_force_readonly(struct ovl_fs *ofs)
330 return (!ovl_upper_mnt(ofs) || !ofs->workdir);
333 static const char *ovl_redirect_mode_def(void)
335 return ovl_redirect_dir_def ? "on" : "off";
338 static const char * const ovl_xino_str[] = {
344 static inline int ovl_xino_def(void)
346 return ovl_xino_auto_def ? OVL_XINO_AUTO : OVL_XINO_OFF;
352 * Prints the mount options for a given superblock.
353 * Returns zero; does not fail.
355 static int ovl_show_options(struct seq_file *m, struct dentry *dentry)
357 struct super_block *sb = dentry->d_sb;
358 struct ovl_fs *ofs = sb->s_fs_info;
360 seq_show_option(m, "lowerdir", ofs->config.lowerdir);
361 if (ofs->config.upperdir) {
362 seq_show_option(m, "upperdir", ofs->config.upperdir);
363 seq_show_option(m, "workdir", ofs->config.workdir);
365 if (ofs->config.default_permissions)
366 seq_puts(m, ",default_permissions");
367 if (strcmp(ofs->config.redirect_mode, ovl_redirect_mode_def()) != 0)
368 seq_printf(m, ",redirect_dir=%s", ofs->config.redirect_mode);
369 if (ofs->config.index != ovl_index_def)
370 seq_printf(m, ",index=%s", ofs->config.index ? "on" : "off");
371 if (!ofs->config.uuid)
372 seq_puts(m, ",uuid=off");
373 if (ofs->config.nfs_export != ovl_nfs_export_def)
374 seq_printf(m, ",nfs_export=%s", ofs->config.nfs_export ?
376 if (ofs->config.xino != ovl_xino_def() && !ovl_same_fs(sb))
377 seq_printf(m, ",xino=%s", ovl_xino_str[ofs->config.xino]);
378 if (ofs->config.metacopy != ovl_metacopy_def)
379 seq_printf(m, ",metacopy=%s",
380 ofs->config.metacopy ? "on" : "off");
381 if (ofs->config.ovl_volatile)
382 seq_puts(m, ",volatile");
386 static int ovl_remount(struct super_block *sb, int *flags, char *data)
388 struct ovl_fs *ofs = sb->s_fs_info;
389 struct super_block *upper_sb;
392 if (!(*flags & SB_RDONLY) && ovl_force_readonly(ofs))
395 if (*flags & SB_RDONLY && !sb_rdonly(sb)) {
396 upper_sb = ovl_upper_mnt(ofs)->mnt_sb;
397 if (ovl_should_sync(ofs)) {
398 down_read(&upper_sb->s_umount);
399 ret = sync_filesystem(upper_sb);
400 up_read(&upper_sb->s_umount);
407 static const struct super_operations ovl_super_operations = {
408 .alloc_inode = ovl_alloc_inode,
409 .free_inode = ovl_free_inode,
410 .destroy_inode = ovl_destroy_inode,
411 .drop_inode = generic_delete_inode,
412 .put_super = ovl_put_super,
413 .sync_fs = ovl_sync_fs,
414 .statfs = ovl_statfs,
415 .show_options = ovl_show_options,
416 .remount_fs = ovl_remount,
423 OPT_DEFAULT_PERMISSIONS,
441 static const match_table_t ovl_tokens = {
442 {OPT_LOWERDIR, "lowerdir=%s"},
443 {OPT_UPPERDIR, "upperdir=%s"},
444 {OPT_WORKDIR, "workdir=%s"},
445 {OPT_DEFAULT_PERMISSIONS, "default_permissions"},
446 {OPT_REDIRECT_DIR, "redirect_dir=%s"},
447 {OPT_INDEX_ON, "index=on"},
448 {OPT_INDEX_OFF, "index=off"},
449 {OPT_USERXATTR, "userxattr"},
450 {OPT_UUID_ON, "uuid=on"},
451 {OPT_UUID_OFF, "uuid=off"},
452 {OPT_NFS_EXPORT_ON, "nfs_export=on"},
453 {OPT_NFS_EXPORT_OFF, "nfs_export=off"},
454 {OPT_XINO_ON, "xino=on"},
455 {OPT_XINO_OFF, "xino=off"},
456 {OPT_XINO_AUTO, "xino=auto"},
457 {OPT_METACOPY_ON, "metacopy=on"},
458 {OPT_METACOPY_OFF, "metacopy=off"},
459 {OPT_VOLATILE, "volatile"},
463 static char *ovl_next_opt(char **s)
471 for (p = sbegin; *p; p++) {
476 } else if (*p == ',') {
486 static int ovl_parse_redirect_mode(struct ovl_config *config, const char *mode)
488 if (strcmp(mode, "on") == 0) {
489 config->redirect_dir = true;
491 * Does not make sense to have redirect creation without
492 * redirect following.
494 config->redirect_follow = true;
495 } else if (strcmp(mode, "follow") == 0) {
496 config->redirect_follow = true;
497 } else if (strcmp(mode, "off") == 0) {
498 if (ovl_redirect_always_follow)
499 config->redirect_follow = true;
500 } else if (strcmp(mode, "nofollow") != 0) {
501 pr_err("bad mount option \"redirect_dir=%s\"\n",
509 static int ovl_parse_opt(char *opt, struct ovl_config *config)
513 bool metacopy_opt = false, redirect_opt = false;
514 bool nfs_export_opt = false, index_opt = false;
516 config->redirect_mode = kstrdup(ovl_redirect_mode_def(), GFP_KERNEL);
517 if (!config->redirect_mode)
520 while ((p = ovl_next_opt(&opt)) != NULL) {
522 substring_t args[MAX_OPT_ARGS];
527 token = match_token(p, ovl_tokens, args);
530 kfree(config->upperdir);
531 config->upperdir = match_strdup(&args[0]);
532 if (!config->upperdir)
537 kfree(config->lowerdir);
538 config->lowerdir = match_strdup(&args[0]);
539 if (!config->lowerdir)
544 kfree(config->workdir);
545 config->workdir = match_strdup(&args[0]);
546 if (!config->workdir)
550 case OPT_DEFAULT_PERMISSIONS:
551 config->default_permissions = true;
554 case OPT_REDIRECT_DIR:
555 kfree(config->redirect_mode);
556 config->redirect_mode = match_strdup(&args[0]);
557 if (!config->redirect_mode)
563 config->index = true;
568 config->index = false;
577 config->uuid = false;
580 case OPT_NFS_EXPORT_ON:
581 config->nfs_export = true;
582 nfs_export_opt = true;
585 case OPT_NFS_EXPORT_OFF:
586 config->nfs_export = false;
587 nfs_export_opt = true;
591 config->xino = OVL_XINO_ON;
595 config->xino = OVL_XINO_OFF;
599 config->xino = OVL_XINO_AUTO;
602 case OPT_METACOPY_ON:
603 config->metacopy = true;
607 case OPT_METACOPY_OFF:
608 config->metacopy = false;
613 config->ovl_volatile = true;
617 config->userxattr = true;
621 pr_err("unrecognized mount option \"%s\" or missing value\n",
627 /* Workdir/index are useless in non-upper mount */
628 if (!config->upperdir) {
629 if (config->workdir) {
630 pr_info("option \"workdir=%s\" is useless in a non-upper mount, ignore\n",
632 kfree(config->workdir);
633 config->workdir = NULL;
635 if (config->index && index_opt) {
636 pr_info("option \"index=on\" is useless in a non-upper mount, ignore\n");
639 config->index = false;
642 if (!config->upperdir && config->ovl_volatile) {
643 pr_info("option \"volatile\" is meaningless in a non-upper mount, ignoring it.\n");
644 config->ovl_volatile = false;
647 err = ovl_parse_redirect_mode(config, config->redirect_mode);
652 * This is to make the logic below simpler. It doesn't make any other
653 * difference, since config->redirect_dir is only used for upper.
655 if (!config->upperdir && config->redirect_follow)
656 config->redirect_dir = true;
658 /* Resolve metacopy -> redirect_dir dependency */
659 if (config->metacopy && !config->redirect_dir) {
660 if (metacopy_opt && redirect_opt) {
661 pr_err("conflicting options: metacopy=on,redirect_dir=%s\n",
662 config->redirect_mode);
667 * There was an explicit redirect_dir=... that resulted
670 pr_info("disabling metacopy due to redirect_dir=%s\n",
671 config->redirect_mode);
672 config->metacopy = false;
674 /* Automatically enable redirect otherwise. */
675 config->redirect_follow = config->redirect_dir = true;
679 /* Resolve nfs_export -> index dependency */
680 if (config->nfs_export && !config->index) {
681 if (!config->upperdir && config->redirect_follow) {
682 pr_info("NFS export requires \"redirect_dir=nofollow\" on non-upper mount, falling back to nfs_export=off.\n");
683 config->nfs_export = false;
684 } else if (nfs_export_opt && index_opt) {
685 pr_err("conflicting options: nfs_export=on,index=off\n");
687 } else if (index_opt) {
689 * There was an explicit index=off that resulted
692 pr_info("disabling nfs_export due to index=off\n");
693 config->nfs_export = false;
695 /* Automatically enable index otherwise. */
696 config->index = true;
700 /* Resolve nfs_export -> !metacopy dependency */
701 if (config->nfs_export && config->metacopy) {
702 if (nfs_export_opt && metacopy_opt) {
703 pr_err("conflicting options: nfs_export=on,metacopy=on\n");
708 * There was an explicit metacopy=on that resulted
711 pr_info("disabling nfs_export due to metacopy=on\n");
712 config->nfs_export = false;
715 * There was an explicit nfs_export=on that resulted
718 pr_info("disabling metacopy due to nfs_export=on\n");
719 config->metacopy = false;
724 /* Resolve userxattr -> !redirect && !metacopy dependency */
725 if (config->userxattr) {
726 if (config->redirect_follow && redirect_opt) {
727 pr_err("conflicting options: userxattr,redirect_dir=%s\n",
728 config->redirect_mode);
731 if (config->metacopy && metacopy_opt) {
732 pr_err("conflicting options: userxattr,metacopy=on\n");
736 * Silently disable default setting of redirect and metacopy.
737 * This shall be the default in the future as well: these
738 * options must be explicitly enabled if used together with
741 config->redirect_dir = config->redirect_follow = false;
742 config->metacopy = false;
748 #define OVL_WORKDIR_NAME "work"
749 #define OVL_INDEXDIR_NAME "index"
751 static struct dentry *ovl_workdir_create(struct ovl_fs *ofs,
752 const char *name, bool persist)
754 struct inode *dir = ofs->workbasedir->d_inode;
755 struct vfsmount *mnt = ovl_upper_mnt(ofs);
758 bool retried = false;
760 inode_lock_nested(dir, I_MUTEX_PARENT);
762 work = lookup_one_len(name, ofs->workbasedir, strlen(name));
765 struct iattr attr = {
766 .ia_valid = ATTR_MODE,
767 .ia_mode = S_IFDIR | 0,
779 err = ovl_workdir_cleanup(dir, mnt, work, 0);
781 if (err == -EINVAL) {
788 work = ovl_create_real(dir, work, OVL_CATTR(attr.ia_mode));
794 * Try to remove POSIX ACL xattrs from workdir. We are good if:
796 * a) success (there was a POSIX ACL xattr and was removed)
797 * b) -ENODATA (there was no POSIX ACL xattr)
798 * c) -EOPNOTSUPP (POSIX ACL xattrs are not supported)
800 * There are various other error values that could effectively
801 * mean that the xattr doesn't exist (e.g. -ERANGE is returned
802 * if the xattr name is too long), but the set of filesystems
803 * allowed as upper are limited to "normal" ones, where checking
804 * for the above two errors is sufficient.
806 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_DEFAULT);
807 if (err && err != -ENODATA && err != -EOPNOTSUPP)
810 err = vfs_removexattr(work, XATTR_NAME_POSIX_ACL_ACCESS);
811 if (err && err != -ENODATA && err != -EOPNOTSUPP)
814 /* Clear any inherited mode bits */
815 inode_lock(work->d_inode);
816 err = notify_change(work, &attr, NULL);
817 inode_unlock(work->d_inode);
831 pr_warn("failed to create directory %s/%s (errno: %i); mounting read-only\n",
832 ofs->config.workdir, name, -err);
837 static void ovl_unescape(char *s)
850 static int ovl_mount_dir_noesc(const char *name, struct path *path)
855 pr_err("empty lowerdir\n");
858 err = kern_path(name, LOOKUP_FOLLOW, path);
860 pr_err("failed to resolve '%s': %i\n", name, err);
864 if (ovl_dentry_weird(path->dentry)) {
865 pr_err("filesystem on '%s' not supported\n", name);
868 if (!d_is_dir(path->dentry)) {
869 pr_err("'%s' not a directory\n", name);
880 static int ovl_mount_dir(const char *name, struct path *path)
883 char *tmp = kstrdup(name, GFP_KERNEL);
887 err = ovl_mount_dir_noesc(tmp, path);
889 if (!err && path->dentry->d_flags & DCACHE_OP_REAL) {
890 pr_err("filesystem on '%s' not supported as upperdir\n",
900 static int ovl_check_namelen(struct path *path, struct ovl_fs *ofs,
903 struct kstatfs statfs;
904 int err = vfs_statfs(path, &statfs);
907 pr_err("statfs failed on '%s'\n", name);
909 ofs->namelen = max(ofs->namelen, statfs.f_namelen);
914 static int ovl_lower_dir(const char *name, struct path *path,
915 struct ovl_fs *ofs, int *stack_depth)
920 err = ovl_mount_dir_noesc(name, path);
924 err = ovl_check_namelen(path, ofs, name);
928 *stack_depth = max(*stack_depth, path->mnt->mnt_sb->s_stack_depth);
931 * The inodes index feature and NFS export need to encode and decode
932 * file handles, so they require that all layers support them.
934 fh_type = ovl_can_decode_fh(path->dentry->d_sb);
935 if ((ofs->config.nfs_export ||
936 (ofs->config.index && ofs->config.upperdir)) && !fh_type) {
937 ofs->config.index = false;
938 ofs->config.nfs_export = false;
939 pr_warn("fs on '%s' does not support file handles, falling back to index=off,nfs_export=off.\n",
943 /* Check if lower fs has 32bit inode numbers */
944 if (fh_type != FILEID_INO32_GEN)
950 /* Workdir should not be subdir of upperdir and vice versa */
951 static bool ovl_workdir_ok(struct dentry *workdir, struct dentry *upperdir)
955 if (workdir != upperdir) {
956 ok = (lock_rename(workdir, upperdir) == NULL);
957 unlock_rename(workdir, upperdir);
962 static unsigned int ovl_split_lowerdirs(char *str)
964 unsigned int ctr = 1;
967 for (s = d = str;; s++, d++) {
970 } else if (*s == ':') {
982 static int __maybe_unused
983 ovl_posix_acl_xattr_get(const struct xattr_handler *handler,
984 struct dentry *dentry, struct inode *inode,
985 const char *name, void *buffer, size_t size)
987 return ovl_xattr_get(dentry, inode, handler->name, buffer, size);
990 static int __maybe_unused
991 ovl_posix_acl_xattr_set(const struct xattr_handler *handler,
992 struct dentry *dentry, struct inode *inode,
993 const char *name, const void *value,
994 size_t size, int flags)
996 struct dentry *workdir = ovl_workdir(dentry);
997 struct inode *realinode = ovl_inode_real(inode);
998 struct posix_acl *acl = NULL;
1001 /* Check that everything is OK before copy-up */
1003 acl = posix_acl_from_xattr(&init_user_ns, value, size);
1005 return PTR_ERR(acl);
1008 if (!IS_POSIXACL(d_inode(workdir)))
1009 goto out_acl_release;
1010 if (!realinode->i_op->set_acl)
1011 goto out_acl_release;
1012 if (handler->flags == ACL_TYPE_DEFAULT && !S_ISDIR(inode->i_mode)) {
1013 err = acl ? -EACCES : 0;
1014 goto out_acl_release;
1017 if (!inode_owner_or_capable(inode))
1018 goto out_acl_release;
1020 posix_acl_release(acl);
1023 * Check if sgid bit needs to be cleared (actual setacl operation will
1024 * be done with mounter's capabilities and so that won't do it for us).
1026 if (unlikely(inode->i_mode & S_ISGID) &&
1027 handler->flags == ACL_TYPE_ACCESS &&
1028 !in_group_p(inode->i_gid) &&
1029 !capable_wrt_inode_uidgid(inode, CAP_FSETID)) {
1030 struct iattr iattr = { .ia_valid = ATTR_KILL_SGID };
1032 err = ovl_setattr(dentry, &iattr);
1037 err = ovl_xattr_set(dentry, inode, handler->name, value, size, flags);
1039 ovl_copyattr(ovl_inode_real(inode), inode);
1044 posix_acl_release(acl);
1048 static int ovl_own_xattr_get(const struct xattr_handler *handler,
1049 struct dentry *dentry, struct inode *inode,
1050 const char *name, void *buffer, size_t size)
1055 static int ovl_own_xattr_set(const struct xattr_handler *handler,
1056 struct dentry *dentry, struct inode *inode,
1057 const char *name, const void *value,
1058 size_t size, int flags)
1063 static int ovl_other_xattr_get(const struct xattr_handler *handler,
1064 struct dentry *dentry, struct inode *inode,
1065 const char *name, void *buffer, size_t size)
1067 return ovl_xattr_get(dentry, inode, name, buffer, size);
1070 static int ovl_other_xattr_set(const struct xattr_handler *handler,
1071 struct dentry *dentry, struct inode *inode,
1072 const char *name, const void *value,
1073 size_t size, int flags)
1075 return ovl_xattr_set(dentry, inode, name, value, size, flags);
1078 static const struct xattr_handler __maybe_unused
1079 ovl_posix_acl_access_xattr_handler = {
1080 .name = XATTR_NAME_POSIX_ACL_ACCESS,
1081 .flags = ACL_TYPE_ACCESS,
1082 .get = ovl_posix_acl_xattr_get,
1083 .set = ovl_posix_acl_xattr_set,
1086 static const struct xattr_handler __maybe_unused
1087 ovl_posix_acl_default_xattr_handler = {
1088 .name = XATTR_NAME_POSIX_ACL_DEFAULT,
1089 .flags = ACL_TYPE_DEFAULT,
1090 .get = ovl_posix_acl_xattr_get,
1091 .set = ovl_posix_acl_xattr_set,
1094 static const struct xattr_handler ovl_own_trusted_xattr_handler = {
1095 .prefix = OVL_XATTR_TRUSTED_PREFIX,
1096 .get = ovl_own_xattr_get,
1097 .set = ovl_own_xattr_set,
1100 static const struct xattr_handler ovl_own_user_xattr_handler = {
1101 .prefix = OVL_XATTR_USER_PREFIX,
1102 .get = ovl_own_xattr_get,
1103 .set = ovl_own_xattr_set,
1106 static const struct xattr_handler ovl_other_xattr_handler = {
1107 .prefix = "", /* catch all */
1108 .get = ovl_other_xattr_get,
1109 .set = ovl_other_xattr_set,
1112 static const struct xattr_handler *ovl_trusted_xattr_handlers[] = {
1113 #ifdef CONFIG_FS_POSIX_ACL
1114 &ovl_posix_acl_access_xattr_handler,
1115 &ovl_posix_acl_default_xattr_handler,
1117 &ovl_own_trusted_xattr_handler,
1118 &ovl_other_xattr_handler,
1122 static const struct xattr_handler *ovl_user_xattr_handlers[] = {
1123 #ifdef CONFIG_FS_POSIX_ACL
1124 &ovl_posix_acl_access_xattr_handler,
1125 &ovl_posix_acl_default_xattr_handler,
1127 &ovl_own_user_xattr_handler,
1128 &ovl_other_xattr_handler,
1132 static int ovl_setup_trap(struct super_block *sb, struct dentry *dir,
1133 struct inode **ptrap, const char *name)
1138 trap = ovl_get_trap_inode(sb, dir);
1139 err = PTR_ERR_OR_ZERO(trap);
1142 pr_err("conflicting %s path\n", name);
1151 * Determine how we treat concurrent use of upperdir/workdir based on the
1152 * index feature. This is papering over mount leaks of container runtimes,
1153 * for example, an old overlay mount is leaked and now its upperdir is
1154 * attempted to be used as a lower layer in a new overlay mount.
1156 static int ovl_report_in_use(struct ovl_fs *ofs, const char *name)
1158 if (ofs->config.index) {
1159 pr_err("%s is in-use as upperdir/workdir of another mount, mount with '-o index=off' to override exclusive upperdir protection.\n",
1163 pr_warn("%s is in-use as upperdir/workdir of another mount, accessing files from both mounts will result in undefined behavior.\n",
1169 static int ovl_get_upper(struct super_block *sb, struct ovl_fs *ofs,
1170 struct ovl_layer *upper_layer, struct path *upperpath)
1172 struct vfsmount *upper_mnt;
1175 err = ovl_mount_dir(ofs->config.upperdir, upperpath);
1179 /* Upper fs should not be r/o */
1180 if (sb_rdonly(upperpath->mnt->mnt_sb)) {
1181 pr_err("upper fs is r/o, try multi-lower layers mount\n");
1186 err = ovl_check_namelen(upperpath, ofs, ofs->config.upperdir);
1190 err = ovl_setup_trap(sb, upperpath->dentry, &upper_layer->trap,
1195 upper_mnt = clone_private_mount(upperpath);
1196 err = PTR_ERR(upper_mnt);
1197 if (IS_ERR(upper_mnt)) {
1198 pr_err("failed to clone upperpath\n");
1202 /* Don't inherit atime flags */
1203 upper_mnt->mnt_flags &= ~(MNT_NOATIME | MNT_NODIRATIME | MNT_RELATIME);
1204 upper_layer->mnt = upper_mnt;
1205 upper_layer->idx = 0;
1206 upper_layer->fsid = 0;
1209 * Inherit SB_NOSEC flag from upperdir.
1211 * This optimization changes behavior when a security related attribute
1212 * (suid/sgid/security.*) is changed on an underlying layer. This is
1213 * okay because we don't yet have guarantees in that case, but it will
1214 * need careful treatment once we want to honour changes to underlying
1217 if (upper_mnt->mnt_sb->s_flags & SB_NOSEC)
1218 sb->s_flags |= SB_NOSEC;
1220 if (ovl_inuse_trylock(ovl_upper_mnt(ofs)->mnt_root)) {
1221 ofs->upperdir_locked = true;
1223 err = ovl_report_in_use(ofs, "upperdir");
1234 * Returns 1 if RENAME_WHITEOUT is supported, 0 if not supported and
1235 * negative values if error is encountered.
1237 static int ovl_check_rename_whiteout(struct dentry *workdir)
1239 struct inode *dir = d_inode(workdir);
1240 struct dentry *temp;
1241 struct dentry *dest;
1242 struct dentry *whiteout;
1243 struct name_snapshot name;
1246 inode_lock_nested(dir, I_MUTEX_PARENT);
1248 temp = ovl_create_temp(workdir, OVL_CATTR(S_IFREG | 0));
1249 err = PTR_ERR(temp);
1253 dest = ovl_lookup_temp(workdir);
1254 err = PTR_ERR(dest);
1260 /* Name is inline and stable - using snapshot as a copy helper */
1261 take_dentry_name_snapshot(&name, temp);
1262 err = ovl_do_rename(dir, temp, dir, dest, RENAME_WHITEOUT);
1269 whiteout = lookup_one_len(name.name.name, workdir, name.name.len);
1270 err = PTR_ERR(whiteout);
1271 if (IS_ERR(whiteout))
1274 err = ovl_is_whiteout(whiteout);
1276 /* Best effort cleanup of whiteout and temp file */
1278 ovl_cleanup(dir, whiteout);
1282 ovl_cleanup(dir, temp);
1283 release_dentry_name_snapshot(&name);
1293 static struct dentry *ovl_lookup_or_create(struct dentry *parent,
1294 const char *name, umode_t mode)
1296 size_t len = strlen(name);
1297 struct dentry *child;
1299 inode_lock_nested(parent->d_inode, I_MUTEX_PARENT);
1300 child = lookup_one_len(name, parent, len);
1301 if (!IS_ERR(child) && !child->d_inode)
1302 child = ovl_create_real(parent->d_inode, child,
1304 inode_unlock(parent->d_inode);
1311 * Creates $workdir/work/incompat/volatile/dirty file if it is not already
1314 static int ovl_create_volatile_dirty(struct ovl_fs *ofs)
1317 struct dentry *d = dget(ofs->workbasedir);
1318 static const char *const volatile_path[] = {
1319 OVL_WORKDIR_NAME, "incompat", "volatile", "dirty"
1321 const char *const *name = volatile_path;
1323 for (ctr = ARRAY_SIZE(volatile_path); ctr; ctr--, name++) {
1324 d = ovl_lookup_or_create(d, *name, ctr > 1 ? S_IFDIR : S_IFREG);
1332 static int ovl_make_workdir(struct super_block *sb, struct ovl_fs *ofs,
1333 struct path *workpath)
1335 struct vfsmount *mnt = ovl_upper_mnt(ofs);
1336 struct dentry *temp, *workdir;
1337 bool rename_whiteout;
1342 err = mnt_want_write(mnt);
1346 workdir = ovl_workdir_create(ofs, OVL_WORKDIR_NAME, false);
1347 err = PTR_ERR(workdir);
1348 if (IS_ERR_OR_NULL(workdir))
1351 ofs->workdir = workdir;
1353 err = ovl_setup_trap(sb, ofs->workdir, &ofs->workdir_trap, "workdir");
1358 * Upper should support d_type, else whiteouts are visible. Given
1359 * workdir and upper are on same fs, we can do iterate_dir() on
1360 * workdir. This check requires successful creation of workdir in
1363 err = ovl_check_d_type_supported(workpath);
1369 pr_warn("upper fs needs to support d_type.\n");
1371 /* Check if upper/work fs supports O_TMPFILE */
1372 temp = ovl_do_tmpfile(ofs->workdir, S_IFREG | 0);
1373 ofs->tmpfile = !IS_ERR(temp);
1377 pr_warn("upper fs does not support tmpfile.\n");
1380 /* Check if upper/work fs supports RENAME_WHITEOUT */
1381 err = ovl_check_rename_whiteout(ofs->workdir);
1385 rename_whiteout = err;
1386 if (!rename_whiteout)
1387 pr_warn("upper fs does not support RENAME_WHITEOUT.\n");
1390 * Check if upper/work fs supports (trusted|user).overlay.* xattr
1392 err = ovl_do_setxattr(ofs, ofs->workdir, OVL_XATTR_OPAQUE, "0", 1);
1394 ofs->noxattr = true;
1395 ofs->config.index = false;
1396 ofs->config.metacopy = false;
1397 pr_warn("upper fs does not support xattr, falling back to index=off and metacopy=off.\n");
1400 ovl_do_removexattr(ofs, ofs->workdir, OVL_XATTR_OPAQUE);
1404 * We allowed sub-optimal upper fs configuration and don't want to break
1405 * users over kernel upgrade, but we never allowed remote upper fs, so
1406 * we can enforce strict requirements for remote upper fs.
1408 if (ovl_dentry_remote(ofs->workdir) &&
1409 (!d_type || !rename_whiteout || ofs->noxattr)) {
1410 pr_err("upper fs missing required features.\n");
1416 * For volatile mount, create a incompat/volatile/dirty file to keep
1419 if (ofs->config.ovl_volatile) {
1420 err = ovl_create_volatile_dirty(ofs);
1422 pr_err("Failed to create volatile/dirty file.\n");
1427 /* Check if upper/work fs supports file handles */
1428 fh_type = ovl_can_decode_fh(ofs->workdir->d_sb);
1429 if (ofs->config.index && !fh_type) {
1430 ofs->config.index = false;
1431 pr_warn("upper fs does not support file handles, falling back to index=off.\n");
1434 /* Check if upper fs has 32bit inode numbers */
1435 if (fh_type != FILEID_INO32_GEN)
1436 ofs->xino_mode = -1;
1438 /* NFS export of r/w mount depends on index */
1439 if (ofs->config.nfs_export && !ofs->config.index) {
1440 pr_warn("NFS export requires \"index=on\", falling back to nfs_export=off.\n");
1441 ofs->config.nfs_export = false;
1444 mnt_drop_write(mnt);
1448 static int ovl_get_workdir(struct super_block *sb, struct ovl_fs *ofs,
1449 struct path *upperpath)
1452 struct path workpath = { };
1454 err = ovl_mount_dir(ofs->config.workdir, &workpath);
1459 if (upperpath->mnt != workpath.mnt) {
1460 pr_err("workdir and upperdir must reside under the same mount\n");
1463 if (!ovl_workdir_ok(workpath.dentry, upperpath->dentry)) {
1464 pr_err("workdir and upperdir must be separate subtrees\n");
1468 ofs->workbasedir = dget(workpath.dentry);
1470 if (ovl_inuse_trylock(ofs->workbasedir)) {
1471 ofs->workdir_locked = true;
1473 err = ovl_report_in_use(ofs, "workdir");
1478 err = ovl_setup_trap(sb, ofs->workbasedir, &ofs->workbasedir_trap,
1483 err = ovl_make_workdir(sb, ofs, &workpath);
1486 path_put(&workpath);
1491 static int ovl_get_indexdir(struct super_block *sb, struct ovl_fs *ofs,
1492 struct ovl_entry *oe, struct path *upperpath)
1494 struct vfsmount *mnt = ovl_upper_mnt(ofs);
1495 struct dentry *indexdir;
1498 err = mnt_want_write(mnt);
1502 /* Verify lower root is upper root origin */
1503 err = ovl_verify_origin(ofs, upperpath->dentry,
1504 oe->lowerstack[0].dentry, true);
1506 pr_err("failed to verify upper root origin\n");
1510 /* index dir will act also as workdir */
1511 iput(ofs->workdir_trap);
1512 ofs->workdir_trap = NULL;
1514 ofs->workdir = NULL;
1515 indexdir = ovl_workdir_create(ofs, OVL_INDEXDIR_NAME, true);
1516 if (IS_ERR(indexdir)) {
1517 err = PTR_ERR(indexdir);
1518 } else if (indexdir) {
1519 ofs->indexdir = indexdir;
1520 ofs->workdir = dget(indexdir);
1522 err = ovl_setup_trap(sb, ofs->indexdir, &ofs->indexdir_trap,
1528 * Verify upper root is exclusively associated with index dir.
1529 * Older kernels stored upper fh in ".overlay.origin"
1530 * xattr. If that xattr exists, verify that it is a match to
1531 * upper dir file handle. In any case, verify or set xattr
1532 * ".overlay.upper" to indicate that index may have
1533 * directory entries.
1535 if (ovl_check_origin_xattr(ofs, ofs->indexdir)) {
1536 err = ovl_verify_set_fh(ofs, ofs->indexdir,
1538 upperpath->dentry, true, false);
1540 pr_err("failed to verify index dir 'origin' xattr\n");
1542 err = ovl_verify_upper(ofs, ofs->indexdir, upperpath->dentry,
1545 pr_err("failed to verify index dir 'upper' xattr\n");
1547 /* Cleanup bad/stale/orphan index entries */
1549 err = ovl_indexdir_cleanup(ofs);
1551 if (err || !ofs->indexdir)
1552 pr_warn("try deleting index dir or mounting with '-o index=off' to disable inodes index.\n");
1555 mnt_drop_write(mnt);
1559 static bool ovl_lower_uuid_ok(struct ovl_fs *ofs, const uuid_t *uuid)
1563 if (!ofs->config.nfs_export && !ovl_upper_mnt(ofs))
1567 * We allow using single lower with null uuid for index and nfs_export
1568 * for example to support those features with single lower squashfs.
1569 * To avoid regressions in setups of overlay with re-formatted lower
1570 * squashfs, do not allow decoding origin with lower null uuid unless
1571 * user opted-in to one of the new features that require following the
1572 * lower inode of non-dir upper.
1574 if (!ofs->config.index && !ofs->config.metacopy && !ofs->config.xino &&
1578 for (i = 0; i < ofs->numfs; i++) {
1580 * We use uuid to associate an overlay lower file handle with a
1581 * lower layer, so we can accept lower fs with null uuid as long
1582 * as all lower layers with null uuid are on the same fs.
1583 * if we detect multiple lower fs with the same uuid, we
1584 * disable lower file handle decoding on all of them.
1586 if (ofs->fs[i].is_lower &&
1587 uuid_equal(&ofs->fs[i].sb->s_uuid, uuid)) {
1588 ofs->fs[i].bad_uuid = true;
1595 /* Get a unique fsid for the layer */
1596 static int ovl_get_fsid(struct ovl_fs *ofs, const struct path *path)
1598 struct super_block *sb = path->mnt->mnt_sb;
1602 bool bad_uuid = false;
1604 for (i = 0; i < ofs->numfs; i++) {
1605 if (ofs->fs[i].sb == sb)
1609 if (!ovl_lower_uuid_ok(ofs, &sb->s_uuid)) {
1611 if (ofs->config.index || ofs->config.nfs_export) {
1612 ofs->config.index = false;
1613 ofs->config.nfs_export = false;
1614 pr_warn("%s uuid detected in lower fs '%pd2', falling back to index=off,nfs_export=off.\n",
1615 uuid_is_null(&sb->s_uuid) ? "null" :
1621 err = get_anon_bdev(&dev);
1623 pr_err("failed to get anonymous bdev for lowerpath\n");
1627 ofs->fs[ofs->numfs].sb = sb;
1628 ofs->fs[ofs->numfs].pseudo_dev = dev;
1629 ofs->fs[ofs->numfs].bad_uuid = bad_uuid;
1631 return ofs->numfs++;
1634 static int ovl_get_layers(struct super_block *sb, struct ovl_fs *ofs,
1635 struct path *stack, unsigned int numlower,
1636 struct ovl_layer *layers)
1642 ofs->fs = kcalloc(numlower + 1, sizeof(struct ovl_sb), GFP_KERNEL);
1643 if (ofs->fs == NULL)
1646 /* idx/fsid 0 are reserved for upper fs even with lower only overlay */
1650 * All lower layers that share the same fs as upper layer, use the same
1651 * pseudo_dev as upper layer. Allocate fs[0].pseudo_dev even for lower
1652 * only overlay to simplify ovl_fs_free().
1653 * is_lower will be set if upper fs is shared with a lower layer.
1655 err = get_anon_bdev(&ofs->fs[0].pseudo_dev);
1657 pr_err("failed to get anonymous bdev for upper fs\n");
1661 if (ovl_upper_mnt(ofs)) {
1662 ofs->fs[0].sb = ovl_upper_mnt(ofs)->mnt_sb;
1663 ofs->fs[0].is_lower = false;
1666 for (i = 0; i < numlower; i++) {
1667 struct vfsmount *mnt;
1671 err = fsid = ovl_get_fsid(ofs, &stack[i]);
1676 * Check if lower root conflicts with this overlay layers before
1677 * checking if it is in-use as upperdir/workdir of "another"
1678 * mount, because we do not bother to check in ovl_is_inuse() if
1679 * the upperdir/workdir is in fact in-use by our
1682 err = ovl_setup_trap(sb, stack[i].dentry, &trap, "lowerdir");
1686 if (ovl_is_inuse(stack[i].dentry)) {
1687 err = ovl_report_in_use(ofs, "lowerdir");
1694 mnt = clone_private_mount(&stack[i]);
1697 pr_err("failed to clone lowerpath\n");
1703 * Make lower layers R/O. That way fchmod/fchown on lower file
1704 * will fail instead of modifying lower fs.
1706 mnt->mnt_flags |= MNT_READONLY | MNT_NOATIME;
1708 layers[ofs->numlayer].trap = trap;
1709 layers[ofs->numlayer].mnt = mnt;
1710 layers[ofs->numlayer].idx = ofs->numlayer;
1711 layers[ofs->numlayer].fsid = fsid;
1712 layers[ofs->numlayer].fs = &ofs->fs[fsid];
1714 ofs->fs[fsid].is_lower = true;
1718 * When all layers on same fs, overlay can use real inode numbers.
1719 * With mount option "xino=<on|auto>", mounter declares that there are
1720 * enough free high bits in underlying fs to hold the unique fsid.
1721 * If overlayfs does encounter underlying inodes using the high xino
1722 * bits reserved for fsid, it emits a warning and uses the original
1723 * inode number or a non persistent inode number allocated from a
1726 if (ofs->numfs - !ovl_upper_mnt(ofs) == 1) {
1727 if (ofs->config.xino == OVL_XINO_ON)
1728 pr_info("\"xino=on\" is useless with all layers on same fs, ignore.\n");
1730 } else if (ofs->config.xino == OVL_XINO_OFF) {
1731 ofs->xino_mode = -1;
1732 } else if (ofs->xino_mode < 0) {
1734 * This is a roundup of number of bits needed for encoding
1735 * fsid, where fsid 0 is reserved for upper fs (even with
1736 * lower only overlay) +1 extra bit is reserved for the non
1737 * persistent inode number range that is used for resolving
1738 * xino lower bits overflow.
1740 BUILD_BUG_ON(ilog2(OVL_MAX_STACK) > 30);
1741 ofs->xino_mode = ilog2(ofs->numfs - 1) + 2;
1744 if (ofs->xino_mode > 0) {
1745 pr_info("\"xino\" feature enabled using %d upper inode bits.\n",
1754 static struct ovl_entry *ovl_get_lowerstack(struct super_block *sb,
1755 const char *lower, unsigned int numlower,
1756 struct ovl_fs *ofs, struct ovl_layer *layers)
1759 struct path *stack = NULL;
1761 struct ovl_entry *oe;
1763 if (!ofs->config.upperdir && numlower == 1) {
1764 pr_err("at least 2 lowerdir are needed while upperdir nonexistent\n");
1765 return ERR_PTR(-EINVAL);
1768 stack = kcalloc(numlower, sizeof(struct path), GFP_KERNEL);
1770 return ERR_PTR(-ENOMEM);
1773 for (i = 0; i < numlower; i++) {
1774 err = ovl_lower_dir(lower, &stack[i], ofs, &sb->s_stack_depth);
1778 lower = strchr(lower, '\0') + 1;
1782 sb->s_stack_depth++;
1783 if (sb->s_stack_depth > FILESYSTEM_MAX_STACK_DEPTH) {
1784 pr_err("maximum fs stacking depth exceeded\n");
1788 err = ovl_get_layers(sb, ofs, stack, numlower, layers);
1793 oe = ovl_alloc_entry(numlower);
1797 for (i = 0; i < numlower; i++) {
1798 oe->lowerstack[i].dentry = dget(stack[i].dentry);
1799 oe->lowerstack[i].layer = &ofs->layers[i+1];
1803 for (i = 0; i < numlower; i++)
1804 path_put(&stack[i]);
1815 * Check if this layer root is a descendant of:
1816 * - another layer of this overlayfs instance
1817 * - upper/work dir of any overlayfs instance
1819 static int ovl_check_layer(struct super_block *sb, struct ovl_fs *ofs,
1820 struct dentry *dentry, const char *name)
1822 struct dentry *next = dentry, *parent;
1828 parent = dget_parent(next);
1830 /* Walk back ancestors to root (inclusive) looking for traps */
1831 while (!err && parent != next) {
1832 if (ovl_lookup_trap_inode(sb, parent)) {
1834 pr_err("overlapping %s path\n", name);
1835 } else if (ovl_is_inuse(parent)) {
1836 err = ovl_report_in_use(ofs, name);
1839 parent = dget_parent(next);
1849 * Check if any of the layers or work dirs overlap.
1851 static int ovl_check_overlapping_layers(struct super_block *sb,
1856 if (ovl_upper_mnt(ofs)) {
1857 err = ovl_check_layer(sb, ofs, ovl_upper_mnt(ofs)->mnt_root,
1863 * Checking workbasedir avoids hitting ovl_is_inuse(parent) of
1864 * this instance and covers overlapping work and index dirs,
1865 * unless work or index dir have been moved since created inside
1866 * workbasedir. In that case, we already have their traps in
1867 * inode cache and we will catch that case on lookup.
1869 err = ovl_check_layer(sb, ofs, ofs->workbasedir, "workdir");
1874 for (i = 1; i < ofs->numlayer; i++) {
1875 err = ovl_check_layer(sb, ofs,
1876 ofs->layers[i].mnt->mnt_root,
1885 static struct dentry *ovl_get_root(struct super_block *sb,
1886 struct dentry *upperdentry,
1887 struct ovl_entry *oe)
1889 struct dentry *root;
1890 struct ovl_path *lowerpath = &oe->lowerstack[0];
1891 unsigned long ino = d_inode(lowerpath->dentry)->i_ino;
1892 int fsid = lowerpath->layer->fsid;
1893 struct ovl_inode_params oip = {
1894 .upperdentry = upperdentry,
1895 .lowerpath = lowerpath,
1898 root = d_make_root(ovl_new_inode(sb, S_IFDIR, 0));
1902 root->d_fsdata = oe;
1905 /* Root inode uses upper st_ino/i_ino */
1906 ino = d_inode(upperdentry)->i_ino;
1908 ovl_dentry_set_upper_alias(root);
1909 if (ovl_is_impuredir(sb, upperdentry))
1910 ovl_set_flag(OVL_IMPURE, d_inode(root));
1913 /* Root is always merge -> can have whiteouts */
1914 ovl_set_flag(OVL_WHITEOUTS, d_inode(root));
1915 ovl_dentry_set_flag(OVL_E_CONNECTED, root);
1916 ovl_set_upperdata(d_inode(root));
1917 ovl_inode_init(d_inode(root), &oip, ino, fsid);
1918 ovl_dentry_update_reval(root, upperdentry, DCACHE_OP_WEAK_REVALIDATE);
1923 static int ovl_fill_super(struct super_block *sb, void *data, int silent)
1925 struct path upperpath = { };
1926 struct dentry *root_dentry;
1927 struct ovl_entry *oe;
1929 struct ovl_layer *layers;
1931 char *splitlower = NULL;
1932 unsigned int numlower;
1936 if (WARN_ON(sb->s_user_ns != current_user_ns()))
1939 sb->s_d_op = &ovl_dentry_operations;
1942 ofs = kzalloc(sizeof(struct ovl_fs), GFP_KERNEL);
1946 ofs->creator_cred = cred = prepare_creds();
1950 /* Is there a reason anyone would want not to share whiteouts? */
1951 ofs->share_whiteout = true;
1953 ofs->config.index = ovl_index_def;
1954 ofs->config.uuid = true;
1955 ofs->config.nfs_export = ovl_nfs_export_def;
1956 ofs->config.xino = ovl_xino_def();
1957 ofs->config.metacopy = ovl_metacopy_def;
1958 err = ovl_parse_opt((char *) data, &ofs->config);
1963 if (!ofs->config.lowerdir) {
1965 pr_err("missing 'lowerdir'\n");
1970 splitlower = kstrdup(ofs->config.lowerdir, GFP_KERNEL);
1974 numlower = ovl_split_lowerdirs(splitlower);
1975 if (numlower > OVL_MAX_STACK) {
1976 pr_err("too many lower directories, limit is %d\n",
1981 layers = kcalloc(numlower + 1, sizeof(struct ovl_layer), GFP_KERNEL);
1985 ofs->layers = layers;
1986 /* Layer 0 is reserved for upper even if there's no upper */
1989 sb->s_stack_depth = 0;
1990 sb->s_maxbytes = MAX_LFS_FILESIZE;
1991 atomic_long_set(&ofs->last_ino, 1);
1992 /* Assume underlaying fs uses 32bit inodes unless proven otherwise */
1993 if (ofs->config.xino != OVL_XINO_OFF) {
1994 ofs->xino_mode = BITS_PER_LONG - 32;
1995 if (!ofs->xino_mode) {
1996 pr_warn("xino not supported on 32bit kernel, falling back to xino=off.\n");
1997 ofs->config.xino = OVL_XINO_OFF;
2001 /* alloc/destroy_inode needed for setting up traps in inode cache */
2002 sb->s_op = &ovl_super_operations;
2004 if (ofs->config.upperdir) {
2005 struct super_block *upper_sb;
2007 if (!ofs->config.workdir) {
2008 pr_err("missing 'workdir'\n");
2012 err = ovl_get_upper(sb, ofs, &layers[0], &upperpath);
2016 upper_sb = ovl_upper_mnt(ofs)->mnt_sb;
2017 if (!ovl_should_sync(ofs)) {
2018 ofs->errseq = errseq_sample(&upper_sb->s_wb_err);
2019 if (errseq_check(&upper_sb->s_wb_err, ofs->errseq)) {
2021 pr_err("Cannot mount volatile when upperdir has an unseen error. Sync upperdir fs to clear state.\n");
2026 err = ovl_get_workdir(sb, ofs, &upperpath);
2031 sb->s_flags |= SB_RDONLY;
2033 sb->s_stack_depth = upper_sb->s_stack_depth;
2034 sb->s_time_gran = upper_sb->s_time_gran;
2036 oe = ovl_get_lowerstack(sb, splitlower, numlower, ofs, layers);
2041 /* If the upper fs is nonexistent, we mark overlayfs r/o too */
2042 if (!ovl_upper_mnt(ofs))
2043 sb->s_flags |= SB_RDONLY;
2045 if (!ofs->config.uuid && ofs->numfs > 1) {
2046 pr_warn("The uuid=off requires a single fs for lower and upper, falling back to uuid=on.\n");
2047 ofs->config.uuid = true;
2050 if (!ovl_force_readonly(ofs) && ofs->config.index) {
2051 err = ovl_get_indexdir(sb, ofs, oe, &upperpath);
2055 /* Force r/o mount with no index dir */
2057 sb->s_flags |= SB_RDONLY;
2060 err = ovl_check_overlapping_layers(sb, ofs);
2064 /* Show index=off in /proc/mounts for forced r/o mount */
2065 if (!ofs->indexdir) {
2066 ofs->config.index = false;
2067 if (ovl_upper_mnt(ofs) && ofs->config.nfs_export) {
2068 pr_warn("NFS export requires an index dir, falling back to nfs_export=off.\n");
2069 ofs->config.nfs_export = false;
2073 if (ofs->config.metacopy && ofs->config.nfs_export) {
2074 pr_warn("NFS export is not supported with metadata only copy up, falling back to nfs_export=off.\n");
2075 ofs->config.nfs_export = false;
2078 if (ofs->config.nfs_export)
2079 sb->s_export_op = &ovl_export_operations;
2081 /* Never override disk quota limits or use reserved space */
2082 cap_lower(cred->cap_effective, CAP_SYS_RESOURCE);
2084 sb->s_magic = OVERLAYFS_SUPER_MAGIC;
2085 sb->s_xattr = ofs->config.userxattr ? ovl_user_xattr_handlers :
2086 ovl_trusted_xattr_handlers;
2087 sb->s_fs_info = ofs;
2088 sb->s_flags |= SB_POSIXACL;
2089 sb->s_iflags |= SB_I_SKIP_SYNC;
2092 root_dentry = ovl_get_root(sb, upperpath.dentry, oe);
2096 mntput(upperpath.mnt);
2099 sb->s_root = root_dentry;
2104 ovl_entry_stack_free(oe);
2108 path_put(&upperpath);
2114 static struct dentry *ovl_mount(struct file_system_type *fs_type, int flags,
2115 const char *dev_name, void *raw_data)
2117 return mount_nodev(fs_type, flags, raw_data, ovl_fill_super);
2120 static struct file_system_type ovl_fs_type = {
2121 .owner = THIS_MODULE,
2123 .fs_flags = FS_USERNS_MOUNT,
2125 .kill_sb = kill_anon_super,
2127 MODULE_ALIAS_FS("overlay");
2129 static void ovl_inode_init_once(void *foo)
2131 struct ovl_inode *oi = foo;
2133 inode_init_once(&oi->vfs_inode);
2136 static int __init ovl_init(void)
2140 ovl_inode_cachep = kmem_cache_create("ovl_inode",
2141 sizeof(struct ovl_inode), 0,
2142 (SLAB_RECLAIM_ACCOUNT|
2143 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
2144 ovl_inode_init_once);
2145 if (ovl_inode_cachep == NULL)
2148 err = ovl_aio_request_cache_init();
2150 err = register_filesystem(&ovl_fs_type);
2154 ovl_aio_request_cache_destroy();
2156 kmem_cache_destroy(ovl_inode_cachep);
2161 static void __exit ovl_exit(void)
2163 unregister_filesystem(&ovl_fs_type);
2166 * Make sure all delayed rcu free inodes are flushed before we
2170 kmem_cache_destroy(ovl_inode_cachep);
2171 ovl_aio_request_cache_destroy();
2174 module_init(ovl_init);
2175 module_exit(ovl_exit);