4 * (C) Copyright Al Viro 2000, 2001
5 * Released under GPL v2.
7 * Based on code from fs/super.c, copyright Linus Torvalds and others.
11 #include <linux/syscalls.h>
12 #include <linux/slab.h>
13 #include <linux/sched.h>
14 #include <linux/spinlock.h>
15 #include <linux/percpu.h>
16 #include <linux/init.h>
17 #include <linux/kernel.h>
18 #include <linux/acct.h>
19 #include <linux/capability.h>
20 #include <linux/cpumask.h>
21 #include <linux/module.h>
22 #include <linux/sysfs.h>
23 #include <linux/seq_file.h>
24 #include <linux/mnt_namespace.h>
25 #include <linux/namei.h>
26 #include <linux/nsproxy.h>
27 #include <linux/security.h>
28 #include <linux/mount.h>
29 #include <linux/ramfs.h>
30 #include <linux/log2.h>
31 #include <linux/idr.h>
32 #include <linux/fs_struct.h>
33 #include <linux/fsnotify.h>
34 #include <asm/uaccess.h>
35 #include <asm/unistd.h>
39 #define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head))
40 #define HASH_SIZE (1UL << HASH_SHIFT)
43 static DEFINE_IDA(mnt_id_ida);
44 static DEFINE_IDA(mnt_group_ida);
45 static DEFINE_SPINLOCK(mnt_id_lock);
46 static int mnt_id_start = 0;
47 static int mnt_group_start = 1;
49 static struct list_head *mount_hashtable __read_mostly;
50 static struct kmem_cache *mnt_cache __read_mostly;
51 static struct rw_semaphore namespace_sem;
54 struct kobject *fs_kobj;
55 EXPORT_SYMBOL_GPL(fs_kobj);
58 * vfsmount lock may be taken for read to prevent changes to the
59 * vfsmount hash, ie. during mountpoint lookups or walking back
62 * It should be taken for write in all cases where the vfsmount
63 * tree or hash is modified or when a vfsmount structure is modified.
65 DEFINE_BRLOCK(vfsmount_lock);
67 static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry)
69 unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES);
70 tmp += ((unsigned long)dentry / L1_CACHE_BYTES);
71 tmp = tmp + (tmp >> HASH_SHIFT);
72 return tmp & (HASH_SIZE - 1);
75 #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16)
78 * allocation is serialized by namespace_sem, but we need the spinlock to
79 * serialize with freeing.
81 static int mnt_alloc_id(struct mount *mnt)
86 ida_pre_get(&mnt_id_ida, GFP_KERNEL);
87 spin_lock(&mnt_id_lock);
88 res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt_id);
90 mnt_id_start = mnt->mnt_id + 1;
91 spin_unlock(&mnt_id_lock);
98 static void mnt_free_id(struct mount *mnt)
100 int id = mnt->mnt_id;
101 spin_lock(&mnt_id_lock);
102 ida_remove(&mnt_id_ida, id);
103 if (mnt_id_start > id)
105 spin_unlock(&mnt_id_lock);
109 * Allocate a new peer group ID
111 * mnt_group_ida is protected by namespace_sem
113 static int mnt_alloc_group_id(struct mount *mnt)
117 if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL))
120 res = ida_get_new_above(&mnt_group_ida,
124 mnt_group_start = mnt->mnt_group_id + 1;
130 * Release a peer group ID
132 void mnt_release_group_id(struct mount *mnt)
134 int id = mnt->mnt_group_id;
135 ida_remove(&mnt_group_ida, id);
136 if (mnt_group_start > id)
137 mnt_group_start = id;
138 mnt->mnt_group_id = 0;
142 * vfsmount lock must be held for read
144 static inline void mnt_add_count(struct mount *mnt, int n)
147 this_cpu_add(mnt->mnt_pcp->mnt_count, n);
156 * vfsmount lock must be held for write
158 unsigned int mnt_get_count(struct mount *mnt)
161 unsigned int count = 0;
164 for_each_possible_cpu(cpu) {
165 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count;
170 return mnt->mnt_count;
174 static struct mount *alloc_vfsmnt(const char *name)
176 struct mount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL);
180 err = mnt_alloc_id(mnt);
185 mnt->mnt_devname = kstrdup(name, GFP_KERNEL);
186 if (!mnt->mnt_devname)
191 mnt->mnt_pcp = alloc_percpu(struct mnt_pcp);
193 goto out_free_devname;
195 this_cpu_add(mnt->mnt_pcp->mnt_count, 1);
198 mnt->mnt_writers = 0;
201 INIT_LIST_HEAD(&mnt->mnt_hash);
202 INIT_LIST_HEAD(&mnt->mnt_child);
203 INIT_LIST_HEAD(&mnt->mnt_mounts);
204 INIT_LIST_HEAD(&mnt->mnt_list);
205 INIT_LIST_HEAD(&mnt->mnt_expire);
206 INIT_LIST_HEAD(&mnt->mnt_share);
207 INIT_LIST_HEAD(&mnt->mnt_slave_list);
208 INIT_LIST_HEAD(&mnt->mnt_slave);
209 #ifdef CONFIG_FSNOTIFY
210 INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks);
217 kfree(mnt->mnt_devname);
222 kmem_cache_free(mnt_cache, mnt);
227 * Most r/o checks on a fs are for operations that take
228 * discrete amounts of time, like a write() or unlink().
229 * We must keep track of when those operations start
230 * (for permission checks) and when they end, so that
231 * we can determine when writes are able to occur to
235 * __mnt_is_readonly: check whether a mount is read-only
236 * @mnt: the mount to check for its write status
238 * This shouldn't be used directly ouside of the VFS.
239 * It does not guarantee that the filesystem will stay
240 * r/w, just that it is right *now*. This can not and
241 * should not be used in place of IS_RDONLY(inode).
242 * mnt_want/drop_write() will _keep_ the filesystem
245 int __mnt_is_readonly(struct vfsmount *mnt)
247 if (mnt->mnt_flags & MNT_READONLY)
249 if (mnt->mnt_sb->s_flags & MS_RDONLY)
253 EXPORT_SYMBOL_GPL(__mnt_is_readonly);
255 static inline void mnt_inc_writers(struct mount *mnt)
258 this_cpu_inc(mnt->mnt_pcp->mnt_writers);
264 static inline void mnt_dec_writers(struct mount *mnt)
267 this_cpu_dec(mnt->mnt_pcp->mnt_writers);
273 static unsigned int mnt_get_writers(struct mount *mnt)
276 unsigned int count = 0;
279 for_each_possible_cpu(cpu) {
280 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers;
285 return mnt->mnt_writers;
290 * Most r/o checks on a fs are for operations that take
291 * discrete amounts of time, like a write() or unlink().
292 * We must keep track of when those operations start
293 * (for permission checks) and when they end, so that
294 * we can determine when writes are able to occur to
298 * mnt_want_write - get write access to a mount
299 * @m: the mount on which to take a write
301 * This tells the low-level filesystem that a write is
302 * about to be performed to it, and makes sure that
303 * writes are allowed before returning success. When
304 * the write operation is finished, mnt_drop_write()
305 * must be called. This is effectively a refcount.
307 int mnt_want_write(struct vfsmount *m)
309 struct mount *mnt = real_mount(m);
313 mnt_inc_writers(mnt);
315 * The store to mnt_inc_writers must be visible before we pass
316 * MNT_WRITE_HOLD loop below, so that the slowpath can see our
317 * incremented count after it has set MNT_WRITE_HOLD.
320 while (mnt->mnt.mnt_flags & MNT_WRITE_HOLD)
323 * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will
324 * be set to match its requirements. So we must not load that until
325 * MNT_WRITE_HOLD is cleared.
328 if (__mnt_is_readonly(m)) {
329 mnt_dec_writers(mnt);
337 EXPORT_SYMBOL_GPL(mnt_want_write);
340 * mnt_clone_write - get write access to a mount
341 * @mnt: the mount on which to take a write
343 * This is effectively like mnt_want_write, except
344 * it must only be used to take an extra write reference
345 * on a mountpoint that we already know has a write reference
346 * on it. This allows some optimisation.
348 * After finished, mnt_drop_write must be called as usual to
349 * drop the reference.
351 int mnt_clone_write(struct vfsmount *mnt)
353 /* superblock may be r/o */
354 if (__mnt_is_readonly(mnt))
357 mnt_inc_writers(real_mount(mnt));
361 EXPORT_SYMBOL_GPL(mnt_clone_write);
364 * mnt_want_write_file - get write access to a file's mount
365 * @file: the file who's mount on which to take a write
367 * This is like mnt_want_write, but it takes a file and can
368 * do some optimisations if the file is open for write already
370 int mnt_want_write_file(struct file *file)
372 struct inode *inode = file->f_dentry->d_inode;
373 if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode))
374 return mnt_want_write(file->f_path.mnt);
376 return mnt_clone_write(file->f_path.mnt);
378 EXPORT_SYMBOL_GPL(mnt_want_write_file);
381 * mnt_drop_write - give up write access to a mount
382 * @mnt: the mount on which to give up write access
384 * Tells the low-level filesystem that we are done
385 * performing writes to it. Must be matched with
386 * mnt_want_write() call above.
388 void mnt_drop_write(struct vfsmount *mnt)
391 mnt_dec_writers(real_mount(mnt));
394 EXPORT_SYMBOL_GPL(mnt_drop_write);
396 void mnt_drop_write_file(struct file *file)
398 mnt_drop_write(file->f_path.mnt);
400 EXPORT_SYMBOL(mnt_drop_write_file);
402 static int mnt_make_readonly(struct mount *mnt)
406 br_write_lock(vfsmount_lock);
407 mnt->mnt.mnt_flags |= MNT_WRITE_HOLD;
409 * After storing MNT_WRITE_HOLD, we'll read the counters. This store
410 * should be visible before we do.
415 * With writers on hold, if this value is zero, then there are
416 * definitely no active writers (although held writers may subsequently
417 * increment the count, they'll have to wait, and decrement it after
418 * seeing MNT_READONLY).
420 * It is OK to have counter incremented on one CPU and decremented on
421 * another: the sum will add up correctly. The danger would be when we
422 * sum up each counter, if we read a counter before it is incremented,
423 * but then read another CPU's count which it has been subsequently
424 * decremented from -- we would see more decrements than we should.
425 * MNT_WRITE_HOLD protects against this scenario, because
426 * mnt_want_write first increments count, then smp_mb, then spins on
427 * MNT_WRITE_HOLD, so it can't be decremented by another CPU while
428 * we're counting up here.
430 if (mnt_get_writers(mnt) > 0)
433 mnt->mnt.mnt_flags |= MNT_READONLY;
435 * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers
436 * that become unheld will see MNT_READONLY.
439 mnt->mnt.mnt_flags &= ~MNT_WRITE_HOLD;
440 br_write_unlock(vfsmount_lock);
444 static void __mnt_unmake_readonly(struct mount *mnt)
446 br_write_lock(vfsmount_lock);
447 mnt->mnt.mnt_flags &= ~MNT_READONLY;
448 br_write_unlock(vfsmount_lock);
451 static void free_vfsmnt(struct mount *mnt)
453 kfree(mnt->mnt_devname);
456 free_percpu(mnt->mnt_pcp);
458 kmem_cache_free(mnt_cache, mnt);
462 * find the first or last mount at @dentry on vfsmount @mnt depending on
463 * @dir. If @dir is set return the first mount else return the last mount.
464 * vfsmount_lock must be held for read or write.
466 struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry,
469 struct list_head *head = mount_hashtable + hash(mnt, dentry);
470 struct list_head *tmp = head;
471 struct mount *p, *found = NULL;
474 tmp = dir ? tmp->next : tmp->prev;
478 p = list_entry(tmp, struct mount, mnt_hash);
479 if (&p->mnt_parent->mnt == mnt && p->mnt_mountpoint == dentry) {
488 * lookup_mnt increments the ref count before returning
489 * the vfsmount struct.
491 struct vfsmount *lookup_mnt(struct path *path)
493 struct mount *child_mnt;
495 br_read_lock(vfsmount_lock);
496 child_mnt = __lookup_mnt(path->mnt, path->dentry, 1);
498 mnt_add_count(child_mnt, 1);
499 br_read_unlock(vfsmount_lock);
500 return &child_mnt->mnt;
502 br_read_unlock(vfsmount_lock);
507 static inline int check_mnt(struct mount *mnt)
509 return mnt->mnt_ns == current->nsproxy->mnt_ns;
513 * vfsmount lock must be held for write
515 static void touch_mnt_namespace(struct mnt_namespace *ns)
519 wake_up_interruptible(&ns->poll);
524 * vfsmount lock must be held for write
526 static void __touch_mnt_namespace(struct mnt_namespace *ns)
528 if (ns && ns->event != event) {
530 wake_up_interruptible(&ns->poll);
535 * Clear dentry's mounted state if it has no remaining mounts.
536 * vfsmount_lock must be held for write.
538 static void dentry_reset_mounted(struct dentry *dentry)
542 for (u = 0; u < HASH_SIZE; u++) {
545 list_for_each_entry(p, &mount_hashtable[u], mnt_hash) {
546 if (p->mnt_mountpoint == dentry)
550 spin_lock(&dentry->d_lock);
551 dentry->d_flags &= ~DCACHE_MOUNTED;
552 spin_unlock(&dentry->d_lock);
556 * vfsmount lock must be held for write
558 static void detach_mnt(struct mount *mnt, struct path *old_path)
560 old_path->dentry = mnt->mnt_mountpoint;
561 old_path->mnt = &mnt->mnt_parent->mnt;
562 mnt->mnt_parent = mnt;
563 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
564 list_del_init(&mnt->mnt_child);
565 list_del_init(&mnt->mnt_hash);
566 dentry_reset_mounted(old_path->dentry);
570 * vfsmount lock must be held for write
572 void mnt_set_mountpoint(struct mount *mnt, struct dentry *dentry,
573 struct mount *child_mnt)
575 child_mnt->mnt_parent = real_mount(mntget(&mnt->mnt));
576 child_mnt->mnt_mountpoint = dget(dentry);
577 spin_lock(&dentry->d_lock);
578 dentry->d_flags |= DCACHE_MOUNTED;
579 spin_unlock(&dentry->d_lock);
583 * vfsmount lock must be held for write
585 static void attach_mnt(struct mount *mnt, struct path *path)
587 mnt_set_mountpoint(real_mount(path->mnt), path->dentry, mnt);
588 list_add_tail(&mnt->mnt_hash, mount_hashtable +
589 hash(path->mnt, path->dentry));
590 list_add_tail(&mnt->mnt_child, &real_mount(path->mnt)->mnt_mounts);
593 static inline void __mnt_make_longterm(struct mount *mnt)
596 atomic_inc(&mnt->mnt_longterm);
600 /* needs vfsmount lock for write */
601 static inline void __mnt_make_shortterm(struct mount *mnt)
604 atomic_dec(&mnt->mnt_longterm);
609 * vfsmount lock must be held for write
611 static void commit_tree(struct mount *mnt)
613 struct mount *parent = mnt->mnt_parent;
616 struct mnt_namespace *n = parent->mnt_ns;
618 BUG_ON(parent == mnt);
620 list_add_tail(&head, &mnt->mnt_list);
621 list_for_each_entry(m, &head, mnt_list) {
623 __mnt_make_longterm(m);
626 list_splice(&head, n->list.prev);
628 list_add_tail(&mnt->mnt_hash, mount_hashtable +
629 hash(&parent->mnt, mnt->mnt_mountpoint));
630 list_add_tail(&mnt->mnt_child, &parent->mnt_mounts);
631 touch_mnt_namespace(n);
634 static struct mount *next_mnt(struct mount *p, struct vfsmount *root)
636 struct list_head *next = p->mnt_mounts.next;
637 if (next == &p->mnt_mounts) {
641 next = p->mnt_child.next;
642 if (next != &p->mnt_parent->mnt_mounts)
647 return list_entry(next, struct mount, mnt_child);
650 static struct mount *skip_mnt_tree(struct mount *p)
652 struct list_head *prev = p->mnt_mounts.prev;
653 while (prev != &p->mnt_mounts) {
654 p = list_entry(prev, struct mount, mnt_child);
655 prev = p->mnt_mounts.prev;
661 vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
667 return ERR_PTR(-ENODEV);
669 mnt = alloc_vfsmnt(name);
671 return ERR_PTR(-ENOMEM);
673 if (flags & MS_KERNMOUNT)
674 mnt->mnt.mnt_flags = MNT_INTERNAL;
676 root = mount_fs(type, flags, name, data);
679 return ERR_CAST(root);
682 mnt->mnt.mnt_root = root;
683 mnt->mnt.mnt_sb = root->d_sb;
684 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
685 mnt->mnt_parent = mnt;
688 EXPORT_SYMBOL_GPL(vfs_kern_mount);
690 static struct mount *clone_mnt(struct mount *old, struct dentry *root,
693 struct super_block *sb = old->mnt.mnt_sb;
694 struct mount *mnt = alloc_vfsmnt(old->mnt_devname);
697 if (flag & (CL_SLAVE | CL_PRIVATE))
698 mnt->mnt_group_id = 0; /* not a peer of original */
700 mnt->mnt_group_id = old->mnt_group_id;
702 if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) {
703 int err = mnt_alloc_group_id(mnt);
708 mnt->mnt.mnt_flags = old->mnt.mnt_flags & ~MNT_WRITE_HOLD;
709 atomic_inc(&sb->s_active);
710 mnt->mnt.mnt_sb = sb;
711 mnt->mnt.mnt_root = dget(root);
712 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
713 mnt->mnt_parent = mnt;
715 if (flag & CL_SLAVE) {
716 list_add(&mnt->mnt_slave, &old->mnt_slave_list);
717 mnt->mnt_master = old;
718 CLEAR_MNT_SHARED(mnt);
719 } else if (!(flag & CL_PRIVATE)) {
720 if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old))
721 list_add(&mnt->mnt_share, &old->mnt_share);
722 if (IS_MNT_SLAVE(old))
723 list_add(&mnt->mnt_slave, &old->mnt_slave);
724 mnt->mnt_master = old->mnt_master;
726 if (flag & CL_MAKE_SHARED)
729 /* stick the duplicate mount on the same expiry list
730 * as the original if that was on one */
731 if (flag & CL_EXPIRE) {
732 if (!list_empty(&old->mnt_expire))
733 list_add(&mnt->mnt_expire, &old->mnt_expire);
743 static inline void mntfree(struct mount *mnt)
745 struct vfsmount *m = &mnt->mnt;
746 struct super_block *sb = m->mnt_sb;
749 * This probably indicates that somebody messed
750 * up a mnt_want/drop_write() pair. If this
751 * happens, the filesystem was probably unable
752 * to make r/w->r/o transitions.
755 * The locking used to deal with mnt_count decrement provides barriers,
756 * so mnt_get_writers() below is safe.
758 WARN_ON(mnt_get_writers(mnt));
759 fsnotify_vfsmount_delete(m);
762 deactivate_super(sb);
765 static void mntput_no_expire(struct mount *mnt)
769 br_read_lock(vfsmount_lock);
770 if (likely(atomic_read(&mnt->mnt_longterm))) {
771 mnt_add_count(mnt, -1);
772 br_read_unlock(vfsmount_lock);
775 br_read_unlock(vfsmount_lock);
777 br_write_lock(vfsmount_lock);
778 mnt_add_count(mnt, -1);
779 if (mnt_get_count(mnt)) {
780 br_write_unlock(vfsmount_lock);
784 mnt_add_count(mnt, -1);
785 if (likely(mnt_get_count(mnt)))
787 br_write_lock(vfsmount_lock);
789 if (unlikely(mnt->mnt_pinned)) {
790 mnt_add_count(mnt, mnt->mnt_pinned + 1);
792 br_write_unlock(vfsmount_lock);
793 acct_auto_close_mnt(&mnt->mnt);
796 br_write_unlock(vfsmount_lock);
800 void mntput(struct vfsmount *mnt)
803 struct mount *m = real_mount(mnt);
804 /* avoid cacheline pingpong, hope gcc doesn't get "smart" */
805 if (unlikely(m->mnt_expiry_mark))
806 m->mnt_expiry_mark = 0;
810 EXPORT_SYMBOL(mntput);
812 struct vfsmount *mntget(struct vfsmount *mnt)
815 mnt_add_count(real_mount(mnt), 1);
818 EXPORT_SYMBOL(mntget);
820 void mnt_pin(struct vfsmount *mnt)
822 br_write_lock(vfsmount_lock);
823 real_mount(mnt)->mnt_pinned++;
824 br_write_unlock(vfsmount_lock);
826 EXPORT_SYMBOL(mnt_pin);
828 void mnt_unpin(struct vfsmount *m)
830 struct mount *mnt = real_mount(m);
831 br_write_lock(vfsmount_lock);
832 if (mnt->mnt_pinned) {
833 mnt_add_count(mnt, 1);
836 br_write_unlock(vfsmount_lock);
838 EXPORT_SYMBOL(mnt_unpin);
840 static inline void mangle(struct seq_file *m, const char *s)
842 seq_escape(m, s, " \t\n\\");
846 * Simple .show_options callback for filesystems which don't want to
847 * implement more complex mount option showing.
849 * See also save_mount_options().
851 int generic_show_options(struct seq_file *m, struct vfsmount *mnt)
856 options = rcu_dereference(mnt->mnt_sb->s_options);
858 if (options != NULL && options[0]) {
866 EXPORT_SYMBOL(generic_show_options);
869 * If filesystem uses generic_show_options(), this function should be
870 * called from the fill_super() callback.
872 * The .remount_fs callback usually needs to be handled in a special
873 * way, to make sure, that previous options are not overwritten if the
876 * Also note, that if the filesystem's .remount_fs function doesn't
877 * reset all options to their default value, but changes only newly
878 * given options, then the displayed options will not reflect reality
881 void save_mount_options(struct super_block *sb, char *options)
883 BUG_ON(sb->s_options);
884 rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL));
886 EXPORT_SYMBOL(save_mount_options);
888 void replace_mount_options(struct super_block *sb, char *options)
890 char *old = sb->s_options;
891 rcu_assign_pointer(sb->s_options, options);
897 EXPORT_SYMBOL(replace_mount_options);
899 #ifdef CONFIG_PROC_FS
901 static void *m_start(struct seq_file *m, loff_t *pos)
903 struct proc_mounts *p = m->private;
905 down_read(&namespace_sem);
906 return seq_list_start(&p->ns->list, *pos);
909 static void *m_next(struct seq_file *m, void *v, loff_t *pos)
911 struct proc_mounts *p = m->private;
913 return seq_list_next(v, &p->ns->list, pos);
916 static void m_stop(struct seq_file *m, void *v)
918 up_read(&namespace_sem);
921 int mnt_had_events(struct proc_mounts *p)
923 struct mnt_namespace *ns = p->ns;
926 br_read_lock(vfsmount_lock);
927 if (p->m.poll_event != ns->event) {
928 p->m.poll_event = ns->event;
931 br_read_unlock(vfsmount_lock);
936 struct proc_fs_info {
941 static int show_sb_opts(struct seq_file *m, struct super_block *sb)
943 static const struct proc_fs_info fs_info[] = {
944 { MS_SYNCHRONOUS, ",sync" },
945 { MS_DIRSYNC, ",dirsync" },
946 { MS_MANDLOCK, ",mand" },
949 const struct proc_fs_info *fs_infop;
951 for (fs_infop = fs_info; fs_infop->flag; fs_infop++) {
952 if (sb->s_flags & fs_infop->flag)
953 seq_puts(m, fs_infop->str);
956 return security_sb_show_options(m, sb);
959 static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt)
961 static const struct proc_fs_info mnt_info[] = {
962 { MNT_NOSUID, ",nosuid" },
963 { MNT_NODEV, ",nodev" },
964 { MNT_NOEXEC, ",noexec" },
965 { MNT_NOATIME, ",noatime" },
966 { MNT_NODIRATIME, ",nodiratime" },
967 { MNT_RELATIME, ",relatime" },
970 const struct proc_fs_info *fs_infop;
972 for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) {
973 if (mnt->mnt_flags & fs_infop->flag)
974 seq_puts(m, fs_infop->str);
978 static void show_type(struct seq_file *m, struct super_block *sb)
980 mangle(m, sb->s_type->name);
981 if (sb->s_subtype && sb->s_subtype[0]) {
983 mangle(m, sb->s_subtype);
987 static int show_vfsmnt(struct seq_file *m, void *v)
989 struct mount *r = list_entry(v, struct mount, mnt_list);
990 struct vfsmount *mnt = &r->mnt;
992 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
994 if (mnt->mnt_sb->s_op->show_devname) {
995 err = mnt->mnt_sb->s_op->show_devname(m, mnt);
999 mangle(m, r->mnt_devname ? r->mnt_devname : "none");
1002 seq_path(m, &mnt_path, " \t\n\\");
1004 show_type(m, mnt->mnt_sb);
1005 seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw");
1006 err = show_sb_opts(m, mnt->mnt_sb);
1009 show_mnt_opts(m, mnt);
1010 if (mnt->mnt_sb->s_op->show_options)
1011 err = mnt->mnt_sb->s_op->show_options(m, mnt);
1012 seq_puts(m, " 0 0\n");
1017 const struct seq_operations mounts_op = {
1024 static int show_mountinfo(struct seq_file *m, void *v)
1026 struct proc_mounts *p = m->private;
1027 struct mount *r = list_entry(v, struct mount, mnt_list);
1028 struct vfsmount *mnt = &r->mnt;
1029 struct super_block *sb = mnt->mnt_sb;
1030 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
1031 struct path root = p->root;
1034 seq_printf(m, "%i %i %u:%u ", r->mnt_id, r->mnt_parent->mnt_id,
1035 MAJOR(sb->s_dev), MINOR(sb->s_dev));
1036 if (sb->s_op->show_path)
1037 err = sb->s_op->show_path(m, mnt);
1039 seq_dentry(m, mnt->mnt_root, " \t\n\\");
1044 /* mountpoints outside of chroot jail will give SEQ_SKIP on this */
1045 err = seq_path_root(m, &mnt_path, &root, " \t\n\\");
1049 seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw");
1050 show_mnt_opts(m, mnt);
1052 /* Tagged fields ("foo:X" or "bar") */
1053 if (IS_MNT_SHARED(r))
1054 seq_printf(m, " shared:%i", r->mnt_group_id);
1055 if (IS_MNT_SLAVE(r)) {
1056 int master = r->mnt_master->mnt_group_id;
1057 int dom = get_dominating_id(r, &p->root);
1058 seq_printf(m, " master:%i", master);
1059 if (dom && dom != master)
1060 seq_printf(m, " propagate_from:%i", dom);
1062 if (IS_MNT_UNBINDABLE(r))
1063 seq_puts(m, " unbindable");
1065 /* Filesystem specific data */
1069 if (sb->s_op->show_devname)
1070 err = sb->s_op->show_devname(m, mnt);
1072 mangle(m, r->mnt_devname ? r->mnt_devname : "none");
1075 seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw");
1076 err = show_sb_opts(m, sb);
1079 if (sb->s_op->show_options)
1080 err = sb->s_op->show_options(m, mnt);
1086 const struct seq_operations mountinfo_op = {
1090 .show = show_mountinfo,
1093 static int show_vfsstat(struct seq_file *m, void *v)
1095 struct mount *r = list_entry(v, struct mount, mnt_list);
1096 struct vfsmount *mnt = &r->mnt;
1097 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
1101 if (mnt->mnt_sb->s_op->show_devname) {
1102 seq_puts(m, "device ");
1103 err = mnt->mnt_sb->s_op->show_devname(m, mnt);
1105 if (r->mnt_devname) {
1106 seq_puts(m, "device ");
1107 mangle(m, r->mnt_devname);
1109 seq_puts(m, "no device");
1113 seq_puts(m, " mounted on ");
1114 seq_path(m, &mnt_path, " \t\n\\");
1117 /* file system type */
1118 seq_puts(m, "with fstype ");
1119 show_type(m, mnt->mnt_sb);
1121 /* optional statistics */
1122 if (mnt->mnt_sb->s_op->show_stats) {
1125 err = mnt->mnt_sb->s_op->show_stats(m, mnt);
1132 const struct seq_operations mountstats_op = {
1136 .show = show_vfsstat,
1138 #endif /* CONFIG_PROC_FS */
1141 * may_umount_tree - check if a mount tree is busy
1142 * @mnt: root of mount tree
1144 * This is called to check if a tree of mounts has any
1145 * open files, pwds, chroots or sub mounts that are
1148 int may_umount_tree(struct vfsmount *mnt)
1150 int actual_refs = 0;
1151 int minimum_refs = 0;
1155 /* write lock needed for mnt_get_count */
1156 br_write_lock(vfsmount_lock);
1157 for (p = real_mount(mnt); p; p = next_mnt(p, mnt)) {
1158 actual_refs += mnt_get_count(p);
1161 br_write_unlock(vfsmount_lock);
1163 if (actual_refs > minimum_refs)
1169 EXPORT_SYMBOL(may_umount_tree);
1172 * may_umount - check if a mount point is busy
1173 * @mnt: root of mount
1175 * This is called to check if a mount point has any
1176 * open files, pwds, chroots or sub mounts. If the
1177 * mount has sub mounts this will return busy
1178 * regardless of whether the sub mounts are busy.
1180 * Doesn't take quota and stuff into account. IOW, in some cases it will
1181 * give false negatives. The main reason why it's here is that we need
1182 * a non-destructive way to look for easily umountable filesystems.
1184 int may_umount(struct vfsmount *mnt)
1187 down_read(&namespace_sem);
1188 br_write_lock(vfsmount_lock);
1189 if (propagate_mount_busy(real_mount(mnt), 2))
1191 br_write_unlock(vfsmount_lock);
1192 up_read(&namespace_sem);
1196 EXPORT_SYMBOL(may_umount);
1198 void release_mounts(struct list_head *head)
1201 while (!list_empty(head)) {
1202 mnt = list_first_entry(head, struct mount, mnt_hash);
1203 list_del_init(&mnt->mnt_hash);
1204 if (mnt_has_parent(mnt)) {
1205 struct dentry *dentry;
1208 br_write_lock(vfsmount_lock);
1209 dentry = mnt->mnt_mountpoint;
1210 m = mnt->mnt_parent;
1211 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
1212 mnt->mnt_parent = mnt;
1214 br_write_unlock(vfsmount_lock);
1223 * vfsmount lock must be held for write
1224 * namespace_sem must be held for write
1226 void umount_tree(struct mount *mnt, int propagate, struct list_head *kill)
1228 LIST_HEAD(tmp_list);
1231 for (p = mnt; p; p = next_mnt(p, &mnt->mnt))
1232 list_move(&p->mnt_hash, &tmp_list);
1235 propagate_umount(&tmp_list);
1237 list_for_each_entry(p, &tmp_list, mnt_hash) {
1238 list_del_init(&p->mnt_expire);
1239 list_del_init(&p->mnt_list);
1240 __touch_mnt_namespace(p->mnt_ns);
1242 __mnt_make_shortterm(p);
1243 list_del_init(&p->mnt_child);
1244 if (mnt_has_parent(p)) {
1245 p->mnt_parent->mnt_ghosts++;
1246 dentry_reset_mounted(p->mnt_mountpoint);
1248 change_mnt_propagation(p, MS_PRIVATE);
1250 list_splice(&tmp_list, kill);
1253 static void shrink_submounts(struct mount *mnt, struct list_head *umounts);
1255 static int do_umount(struct mount *mnt, int flags)
1257 struct super_block *sb = mnt->mnt.mnt_sb;
1259 LIST_HEAD(umount_list);
1261 retval = security_sb_umount(&mnt->mnt, flags);
1266 * Allow userspace to request a mountpoint be expired rather than
1267 * unmounting unconditionally. Unmount only happens if:
1268 * (1) the mark is already set (the mark is cleared by mntput())
1269 * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount]
1271 if (flags & MNT_EXPIRE) {
1272 if (&mnt->mnt == current->fs->root.mnt ||
1273 flags & (MNT_FORCE | MNT_DETACH))
1277 * probably don't strictly need the lock here if we examined
1278 * all race cases, but it's a slowpath.
1280 br_write_lock(vfsmount_lock);
1281 if (mnt_get_count(mnt) != 2) {
1282 br_write_unlock(vfsmount_lock);
1285 br_write_unlock(vfsmount_lock);
1287 if (!xchg(&mnt->mnt_expiry_mark, 1))
1292 * If we may have to abort operations to get out of this
1293 * mount, and they will themselves hold resources we must
1294 * allow the fs to do things. In the Unix tradition of
1295 * 'Gee thats tricky lets do it in userspace' the umount_begin
1296 * might fail to complete on the first run through as other tasks
1297 * must return, and the like. Thats for the mount program to worry
1298 * about for the moment.
1301 if (flags & MNT_FORCE && sb->s_op->umount_begin) {
1302 sb->s_op->umount_begin(sb);
1306 * No sense to grab the lock for this test, but test itself looks
1307 * somewhat bogus. Suggestions for better replacement?
1308 * Ho-hum... In principle, we might treat that as umount + switch
1309 * to rootfs. GC would eventually take care of the old vfsmount.
1310 * Actually it makes sense, especially if rootfs would contain a
1311 * /reboot - static binary that would close all descriptors and
1312 * call reboot(9). Then init(8) could umount root and exec /reboot.
1314 if (&mnt->mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) {
1316 * Special case for "unmounting" root ...
1317 * we just try to remount it readonly.
1319 down_write(&sb->s_umount);
1320 if (!(sb->s_flags & MS_RDONLY))
1321 retval = do_remount_sb(sb, MS_RDONLY, NULL, 0);
1322 up_write(&sb->s_umount);
1326 down_write(&namespace_sem);
1327 br_write_lock(vfsmount_lock);
1330 if (!(flags & MNT_DETACH))
1331 shrink_submounts(mnt, &umount_list);
1334 if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) {
1335 if (!list_empty(&mnt->mnt_list))
1336 umount_tree(mnt, 1, &umount_list);
1339 br_write_unlock(vfsmount_lock);
1340 up_write(&namespace_sem);
1341 release_mounts(&umount_list);
1346 * Now umount can handle mount points as well as block devices.
1347 * This is important for filesystems which use unnamed block devices.
1349 * We now support a flag for forced unmount like the other 'big iron'
1350 * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD
1353 SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
1358 int lookup_flags = 0;
1360 if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW))
1363 if (!(flags & UMOUNT_NOFOLLOW))
1364 lookup_flags |= LOOKUP_FOLLOW;
1366 retval = user_path_at(AT_FDCWD, name, lookup_flags, &path);
1369 mnt = real_mount(path.mnt);
1371 if (path.dentry != path.mnt->mnt_root)
1373 if (!check_mnt(mnt))
1377 if (!capable(CAP_SYS_ADMIN))
1380 retval = do_umount(mnt, flags);
1382 /* we mustn't call path_put() as that would clear mnt_expiry_mark */
1384 mntput_no_expire(mnt);
1389 #ifdef __ARCH_WANT_SYS_OLDUMOUNT
1392 * The 2.0 compatible umount. No flags.
1394 SYSCALL_DEFINE1(oldumount, char __user *, name)
1396 return sys_umount(name, 0);
1401 static int mount_is_safe(struct path *path)
1403 if (capable(CAP_SYS_ADMIN))
1407 if (S_ISLNK(path->dentry->d_inode->i_mode))
1409 if (path->dentry->d_inode->i_mode & S_ISVTX) {
1410 if (current_uid() != path->dentry->d_inode->i_uid)
1413 if (inode_permission(path->dentry->d_inode, MAY_WRITE))
1419 struct mount *copy_tree(struct mount *mnt, struct dentry *dentry,
1422 struct mount *res, *p, *q, *r;
1425 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt))
1428 res = q = clone_mnt(mnt, dentry, flag);
1431 q->mnt_mountpoint = mnt->mnt_mountpoint;
1434 list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) {
1436 if (!is_subdir(r->mnt_mountpoint, dentry))
1439 for (s = r; s; s = next_mnt(s, &r->mnt)) {
1440 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) {
1441 s = skip_mnt_tree(s);
1444 while (p != s->mnt_parent) {
1450 path.dentry = p->mnt_mountpoint;
1451 q = clone_mnt(p, p->mnt.mnt_root, flag);
1454 br_write_lock(vfsmount_lock);
1455 list_add_tail(&q->mnt_list, &res->mnt_list);
1456 attach_mnt(q, &path);
1457 br_write_unlock(vfsmount_lock);
1463 LIST_HEAD(umount_list);
1464 br_write_lock(vfsmount_lock);
1465 umount_tree(res, 0, &umount_list);
1466 br_write_unlock(vfsmount_lock);
1467 release_mounts(&umount_list);
1472 struct vfsmount *collect_mounts(struct path *path)
1475 down_write(&namespace_sem);
1476 tree = copy_tree(real_mount(path->mnt), path->dentry,
1477 CL_COPY_ALL | CL_PRIVATE);
1478 up_write(&namespace_sem);
1479 return tree ? &tree->mnt : NULL;
1482 void drop_collected_mounts(struct vfsmount *mnt)
1484 LIST_HEAD(umount_list);
1485 down_write(&namespace_sem);
1486 br_write_lock(vfsmount_lock);
1487 umount_tree(real_mount(mnt), 0, &umount_list);
1488 br_write_unlock(vfsmount_lock);
1489 up_write(&namespace_sem);
1490 release_mounts(&umount_list);
1493 int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
1494 struct vfsmount *root)
1497 int res = f(root, arg);
1500 list_for_each_entry(mnt, &real_mount(root)->mnt_list, mnt_list) {
1501 res = f(&mnt->mnt, arg);
1508 static void cleanup_group_ids(struct mount *mnt, struct mount *end)
1512 for (p = mnt; p != end; p = next_mnt(p, &mnt->mnt)) {
1513 if (p->mnt_group_id && !IS_MNT_SHARED(p))
1514 mnt_release_group_id(p);
1518 static int invent_group_ids(struct mount *mnt, bool recurse)
1522 for (p = mnt; p; p = recurse ? next_mnt(p, &mnt->mnt) : NULL) {
1523 if (!p->mnt_group_id && !IS_MNT_SHARED(p)) {
1524 int err = mnt_alloc_group_id(p);
1526 cleanup_group_ids(mnt, p);
1536 * @source_mnt : mount tree to be attached
1537 * @nd : place the mount tree @source_mnt is attached
1538 * @parent_nd : if non-null, detach the source_mnt from its parent and
1539 * store the parent mount and mountpoint dentry.
1540 * (done when source_mnt is moved)
1542 * NOTE: in the table below explains the semantics when a source mount
1543 * of a given type is attached to a destination mount of a given type.
1544 * ---------------------------------------------------------------------------
1545 * | BIND MOUNT OPERATION |
1546 * |**************************************************************************
1547 * | source-->| shared | private | slave | unbindable |
1551 * |**************************************************************************
1552 * | shared | shared (++) | shared (+) | shared(+++)| invalid |
1554 * |non-shared| shared (+) | private | slave (*) | invalid |
1555 * ***************************************************************************
1556 * A bind operation clones the source mount and mounts the clone on the
1557 * destination mount.
1559 * (++) the cloned mount is propagated to all the mounts in the propagation
1560 * tree of the destination mount and the cloned mount is added to
1561 * the peer group of the source mount.
1562 * (+) the cloned mount is created under the destination mount and is marked
1563 * as shared. The cloned mount is added to the peer group of the source
1565 * (+++) the mount is propagated to all the mounts in the propagation tree
1566 * of the destination mount and the cloned mount is made slave
1567 * of the same master as that of the source mount. The cloned mount
1568 * is marked as 'shared and slave'.
1569 * (*) the cloned mount is made a slave of the same master as that of the
1572 * ---------------------------------------------------------------------------
1573 * | MOVE MOUNT OPERATION |
1574 * |**************************************************************************
1575 * | source-->| shared | private | slave | unbindable |
1579 * |**************************************************************************
1580 * | shared | shared (+) | shared (+) | shared(+++) | invalid |
1582 * |non-shared| shared (+*) | private | slave (*) | unbindable |
1583 * ***************************************************************************
1585 * (+) the mount is moved to the destination. And is then propagated to
1586 * all the mounts in the propagation tree of the destination mount.
1587 * (+*) the mount is moved to the destination.
1588 * (+++) the mount is moved to the destination and is then propagated to
1589 * all the mounts belonging to the destination mount's propagation tree.
1590 * the mount is marked as 'shared and slave'.
1591 * (*) the mount continues to be a slave at the new location.
1593 * if the source mount is a tree, the operations explained above is
1594 * applied to each mount in the tree.
1595 * Must be called without spinlocks held, since this function can sleep
1598 static int attach_recursive_mnt(struct mount *source_mnt,
1599 struct path *path, struct path *parent_path)
1601 LIST_HEAD(tree_list);
1602 struct mount *dest_mnt = real_mount(path->mnt);
1603 struct dentry *dest_dentry = path->dentry;
1604 struct mount *child, *p;
1607 if (IS_MNT_SHARED(dest_mnt)) {
1608 err = invent_group_ids(source_mnt, true);
1612 err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list);
1614 goto out_cleanup_ids;
1616 br_write_lock(vfsmount_lock);
1618 if (IS_MNT_SHARED(dest_mnt)) {
1619 for (p = source_mnt; p; p = next_mnt(p, &source_mnt->mnt))
1623 detach_mnt(source_mnt, parent_path);
1624 attach_mnt(source_mnt, path);
1625 touch_mnt_namespace(source_mnt->mnt_ns);
1627 mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt);
1628 commit_tree(source_mnt);
1631 list_for_each_entry_safe(child, p, &tree_list, mnt_hash) {
1632 list_del_init(&child->mnt_hash);
1635 br_write_unlock(vfsmount_lock);
1640 if (IS_MNT_SHARED(dest_mnt))
1641 cleanup_group_ids(source_mnt, NULL);
1646 static int lock_mount(struct path *path)
1648 struct vfsmount *mnt;
1650 mutex_lock(&path->dentry->d_inode->i_mutex);
1651 if (unlikely(cant_mount(path->dentry))) {
1652 mutex_unlock(&path->dentry->d_inode->i_mutex);
1655 down_write(&namespace_sem);
1656 mnt = lookup_mnt(path);
1659 up_write(&namespace_sem);
1660 mutex_unlock(&path->dentry->d_inode->i_mutex);
1663 path->dentry = dget(mnt->mnt_root);
1667 static void unlock_mount(struct path *path)
1669 up_write(&namespace_sem);
1670 mutex_unlock(&path->dentry->d_inode->i_mutex);
1673 static int graft_tree(struct mount *mnt, struct path *path)
1675 if (mnt->mnt.mnt_sb->s_flags & MS_NOUSER)
1678 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
1679 S_ISDIR(mnt->mnt.mnt_root->d_inode->i_mode))
1682 if (d_unlinked(path->dentry))
1685 return attach_recursive_mnt(mnt, path, NULL);
1689 * Sanity check the flags to change_mnt_propagation.
1692 static int flags_to_propagation_type(int flags)
1694 int type = flags & ~(MS_REC | MS_SILENT);
1696 /* Fail if any non-propagation flags are set */
1697 if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
1699 /* Only one propagation flag should be set */
1700 if (!is_power_of_2(type))
1706 * recursively change the type of the mountpoint.
1708 static int do_change_type(struct path *path, int flag)
1711 struct mount *mnt = real_mount(path->mnt);
1712 int recurse = flag & MS_REC;
1716 if (!capable(CAP_SYS_ADMIN))
1719 if (path->dentry != path->mnt->mnt_root)
1722 type = flags_to_propagation_type(flag);
1726 down_write(&namespace_sem);
1727 if (type == MS_SHARED) {
1728 err = invent_group_ids(mnt, recurse);
1733 br_write_lock(vfsmount_lock);
1734 for (m = mnt; m; m = (recurse ? next_mnt(m, &mnt->mnt) : NULL))
1735 change_mnt_propagation(m, type);
1736 br_write_unlock(vfsmount_lock);
1739 up_write(&namespace_sem);
1744 * do loopback mount.
1746 static int do_loopback(struct path *path, char *old_name,
1749 LIST_HEAD(umount_list);
1750 struct path old_path;
1751 struct mount *mnt = NULL, *old;
1752 int err = mount_is_safe(path);
1755 if (!old_name || !*old_name)
1757 err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path);
1761 err = lock_mount(path);
1765 old = real_mount(old_path.mnt);
1768 if (IS_MNT_UNBINDABLE(old))
1771 if (!check_mnt(real_mount(path->mnt)) || !check_mnt(old))
1776 mnt = copy_tree(old, old_path.dentry, 0);
1778 mnt = clone_mnt(old, old_path.dentry, 0);
1783 err = graft_tree(mnt, path);
1785 br_write_lock(vfsmount_lock);
1786 umount_tree(mnt, 0, &umount_list);
1787 br_write_unlock(vfsmount_lock);
1791 release_mounts(&umount_list);
1793 path_put(&old_path);
1797 static int change_mount_flags(struct vfsmount *mnt, int ms_flags)
1800 int readonly_request = 0;
1802 if (ms_flags & MS_RDONLY)
1803 readonly_request = 1;
1804 if (readonly_request == __mnt_is_readonly(mnt))
1807 if (readonly_request)
1808 error = mnt_make_readonly(real_mount(mnt));
1810 __mnt_unmake_readonly(real_mount(mnt));
1815 * change filesystem flags. dir should be a physical root of filesystem.
1816 * If you've mounted a non-root directory somewhere and want to do remount
1817 * on it - tough luck.
1819 static int do_remount(struct path *path, int flags, int mnt_flags,
1823 struct super_block *sb = path->mnt->mnt_sb;
1824 struct mount *mnt = real_mount(path->mnt);
1826 if (!capable(CAP_SYS_ADMIN))
1829 if (!check_mnt(mnt))
1832 if (path->dentry != path->mnt->mnt_root)
1835 err = security_sb_remount(sb, data);
1839 down_write(&sb->s_umount);
1840 if (flags & MS_BIND)
1841 err = change_mount_flags(path->mnt, flags);
1843 err = do_remount_sb(sb, flags, data, 0);
1845 br_write_lock(vfsmount_lock);
1846 mnt_flags |= mnt->mnt.mnt_flags & MNT_PROPAGATION_MASK;
1847 mnt->mnt.mnt_flags = mnt_flags;
1848 br_write_unlock(vfsmount_lock);
1850 up_write(&sb->s_umount);
1852 br_write_lock(vfsmount_lock);
1853 touch_mnt_namespace(mnt->mnt_ns);
1854 br_write_unlock(vfsmount_lock);
1859 static inline int tree_contains_unbindable(struct mount *mnt)
1862 for (p = mnt; p; p = next_mnt(p, &mnt->mnt)) {
1863 if (IS_MNT_UNBINDABLE(p))
1869 static int do_move_mount(struct path *path, char *old_name)
1871 struct path old_path, parent_path;
1875 if (!capable(CAP_SYS_ADMIN))
1877 if (!old_name || !*old_name)
1879 err = kern_path(old_name, LOOKUP_FOLLOW, &old_path);
1883 err = lock_mount(path);
1887 old = real_mount(old_path.mnt);
1888 p = real_mount(path->mnt);
1891 if (!check_mnt(p) || !check_mnt(old))
1894 if (d_unlinked(path->dentry))
1898 if (old_path.dentry != old_path.mnt->mnt_root)
1901 if (!mnt_has_parent(old))
1904 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
1905 S_ISDIR(old_path.dentry->d_inode->i_mode))
1908 * Don't move a mount residing in a shared parent.
1910 if (IS_MNT_SHARED(old->mnt_parent))
1913 * Don't move a mount tree containing unbindable mounts to a destination
1914 * mount which is shared.
1916 if (IS_MNT_SHARED(p) && tree_contains_unbindable(old))
1919 for (; mnt_has_parent(p); p = p->mnt_parent)
1923 err = attach_recursive_mnt(old, path, &parent_path);
1927 /* if the mount is moved, it should no longer be expire
1929 list_del_init(&old->mnt_expire);
1934 path_put(&parent_path);
1935 path_put(&old_path);
1939 static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
1942 const char *subtype = strchr(fstype, '.');
1951 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1953 if (!mnt->mnt_sb->s_subtype)
1959 return ERR_PTR(err);
1962 static struct vfsmount *
1963 do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1965 struct file_system_type *type = get_fs_type(fstype);
1966 struct vfsmount *mnt;
1968 return ERR_PTR(-ENODEV);
1969 mnt = vfs_kern_mount(type, flags, name, data);
1970 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1971 !mnt->mnt_sb->s_subtype)
1972 mnt = fs_set_subtype(mnt, fstype);
1973 put_filesystem(type);
1978 * add a mount into a namespace's mount tree
1980 static int do_add_mount(struct mount *newmnt, struct path *path, int mnt_flags)
1984 mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL);
1986 err = lock_mount(path);
1991 if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(real_mount(path->mnt)))
1994 /* Refuse the same filesystem on the same mount point */
1996 if (path->mnt->mnt_sb == newmnt->mnt.mnt_sb &&
1997 path->mnt->mnt_root == path->dentry)
2001 if (S_ISLNK(newmnt->mnt.mnt_root->d_inode->i_mode))
2004 newmnt->mnt.mnt_flags = mnt_flags;
2005 err = graft_tree(newmnt, path);
2013 * create a new mount for userspace and request it to be added into the
2016 static int do_new_mount(struct path *path, char *type, int flags,
2017 int mnt_flags, char *name, void *data)
2019 struct vfsmount *mnt;
2025 /* we need capabilities... */
2026 if (!capable(CAP_SYS_ADMIN))
2029 mnt = do_kern_mount(type, flags, name, data);
2031 return PTR_ERR(mnt);
2033 err = do_add_mount(real_mount(mnt), path, mnt_flags);
2039 int finish_automount(struct vfsmount *m, struct path *path)
2041 struct mount *mnt = real_mount(m);
2043 /* The new mount record should have at least 2 refs to prevent it being
2044 * expired before we get a chance to add it
2046 BUG_ON(mnt_get_count(mnt) < 2);
2048 if (m->mnt_sb == path->mnt->mnt_sb &&
2049 m->mnt_root == path->dentry) {
2054 err = do_add_mount(mnt, path, path->mnt->mnt_flags | MNT_SHRINKABLE);
2058 /* remove m from any expiration list it may be on */
2059 if (!list_empty(&mnt->mnt_expire)) {
2060 down_write(&namespace_sem);
2061 br_write_lock(vfsmount_lock);
2062 list_del_init(&mnt->mnt_expire);
2063 br_write_unlock(vfsmount_lock);
2064 up_write(&namespace_sem);
2072 * mnt_set_expiry - Put a mount on an expiration list
2073 * @mnt: The mount to list.
2074 * @expiry_list: The list to add the mount to.
2076 void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list)
2078 down_write(&namespace_sem);
2079 br_write_lock(vfsmount_lock);
2081 list_add_tail(&real_mount(mnt)->mnt_expire, expiry_list);
2083 br_write_unlock(vfsmount_lock);
2084 up_write(&namespace_sem);
2086 EXPORT_SYMBOL(mnt_set_expiry);
2089 * process a list of expirable mountpoints with the intent of discarding any
2090 * mountpoints that aren't in use and haven't been touched since last we came
2093 void mark_mounts_for_expiry(struct list_head *mounts)
2095 struct mount *mnt, *next;
2096 LIST_HEAD(graveyard);
2099 if (list_empty(mounts))
2102 down_write(&namespace_sem);
2103 br_write_lock(vfsmount_lock);
2105 /* extract from the expiration list every vfsmount that matches the
2106 * following criteria:
2107 * - only referenced by its parent vfsmount
2108 * - still marked for expiry (marked on the last call here; marks are
2109 * cleared by mntput())
2111 list_for_each_entry_safe(mnt, next, mounts, mnt_expire) {
2112 if (!xchg(&mnt->mnt_expiry_mark, 1) ||
2113 propagate_mount_busy(mnt, 1))
2115 list_move(&mnt->mnt_expire, &graveyard);
2117 while (!list_empty(&graveyard)) {
2118 mnt = list_first_entry(&graveyard, struct mount, mnt_expire);
2119 touch_mnt_namespace(mnt->mnt_ns);
2120 umount_tree(mnt, 1, &umounts);
2122 br_write_unlock(vfsmount_lock);
2123 up_write(&namespace_sem);
2125 release_mounts(&umounts);
2128 EXPORT_SYMBOL_GPL(mark_mounts_for_expiry);
2131 * Ripoff of 'select_parent()'
2133 * search the list of submounts for a given mountpoint, and move any
2134 * shrinkable submounts to the 'graveyard' list.
2136 static int select_submounts(struct mount *parent, struct list_head *graveyard)
2138 struct mount *this_parent = parent;
2139 struct list_head *next;
2143 next = this_parent->mnt_mounts.next;
2145 while (next != &this_parent->mnt_mounts) {
2146 struct list_head *tmp = next;
2147 struct mount *mnt = list_entry(tmp, struct mount, mnt_child);
2150 if (!(mnt->mnt.mnt_flags & MNT_SHRINKABLE))
2153 * Descend a level if the d_mounts list is non-empty.
2155 if (!list_empty(&mnt->mnt_mounts)) {
2160 if (!propagate_mount_busy(mnt, 1)) {
2161 list_move_tail(&mnt->mnt_expire, graveyard);
2166 * All done at this level ... ascend and resume the search
2168 if (this_parent != parent) {
2169 next = this_parent->mnt_child.next;
2170 this_parent = this_parent->mnt_parent;
2177 * process a list of expirable mountpoints with the intent of discarding any
2178 * submounts of a specific parent mountpoint
2180 * vfsmount_lock must be held for write
2182 static void shrink_submounts(struct mount *mnt, struct list_head *umounts)
2184 LIST_HEAD(graveyard);
2187 /* extract submounts of 'mountpoint' from the expiration list */
2188 while (select_submounts(mnt, &graveyard)) {
2189 while (!list_empty(&graveyard)) {
2190 m = list_first_entry(&graveyard, struct mount,
2192 touch_mnt_namespace(m->mnt_ns);
2193 umount_tree(m, 1, umounts);
2199 * Some copy_from_user() implementations do not return the exact number of
2200 * bytes remaining to copy on a fault. But copy_mount_options() requires that.
2201 * Note that this function differs from copy_from_user() in that it will oops
2202 * on bad values of `to', rather than returning a short copy.
2204 static long exact_copy_from_user(void *to, const void __user * from,
2208 const char __user *f = from;
2211 if (!access_ok(VERIFY_READ, from, n))
2215 if (__get_user(c, f)) {
2226 int copy_mount_options(const void __user * data, unsigned long *where)
2236 if (!(page = __get_free_page(GFP_KERNEL)))
2239 /* We only care that *some* data at the address the user
2240 * gave us is valid. Just in case, we'll zero
2241 * the remainder of the page.
2243 /* copy_from_user cannot cross TASK_SIZE ! */
2244 size = TASK_SIZE - (unsigned long)data;
2245 if (size > PAGE_SIZE)
2248 i = size - exact_copy_from_user((void *)page, data, size);
2254 memset((char *)page + i, 0, PAGE_SIZE - i);
2259 int copy_mount_string(const void __user *data, char **where)
2268 tmp = strndup_user(data, PAGE_SIZE);
2270 return PTR_ERR(tmp);
2277 * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to
2278 * be given to the mount() call (ie: read-only, no-dev, no-suid etc).
2280 * data is a (void *) that can point to any structure up to
2281 * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent
2282 * information (or be NULL).
2284 * Pre-0.97 versions of mount() didn't have a flags word.
2285 * When the flags word was introduced its top half was required
2286 * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9.
2287 * Therefore, if this magic number is present, it carries no information
2288 * and must be discarded.
2290 long do_mount(char *dev_name, char *dir_name, char *type_page,
2291 unsigned long flags, void *data_page)
2298 if ((flags & MS_MGC_MSK) == MS_MGC_VAL)
2299 flags &= ~MS_MGC_MSK;
2301 /* Basic sanity checks */
2303 if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE))
2307 ((char *)data_page)[PAGE_SIZE - 1] = 0;
2309 /* ... and get the mountpoint */
2310 retval = kern_path(dir_name, LOOKUP_FOLLOW, &path);
2314 retval = security_sb_mount(dev_name, &path,
2315 type_page, flags, data_page);
2319 /* Default to relatime unless overriden */
2320 if (!(flags & MS_NOATIME))
2321 mnt_flags |= MNT_RELATIME;
2323 /* Separate the per-mountpoint flags */
2324 if (flags & MS_NOSUID)
2325 mnt_flags |= MNT_NOSUID;
2326 if (flags & MS_NODEV)
2327 mnt_flags |= MNT_NODEV;
2328 if (flags & MS_NOEXEC)
2329 mnt_flags |= MNT_NOEXEC;
2330 if (flags & MS_NOATIME)
2331 mnt_flags |= MNT_NOATIME;
2332 if (flags & MS_NODIRATIME)
2333 mnt_flags |= MNT_NODIRATIME;
2334 if (flags & MS_STRICTATIME)
2335 mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME);
2336 if (flags & MS_RDONLY)
2337 mnt_flags |= MNT_READONLY;
2339 flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN |
2340 MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT |
2343 if (flags & MS_REMOUNT)
2344 retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags,
2346 else if (flags & MS_BIND)
2347 retval = do_loopback(&path, dev_name, flags & MS_REC);
2348 else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
2349 retval = do_change_type(&path, flags);
2350 else if (flags & MS_MOVE)
2351 retval = do_move_mount(&path, dev_name);
2353 retval = do_new_mount(&path, type_page, flags, mnt_flags,
2354 dev_name, data_page);
2360 static struct mnt_namespace *alloc_mnt_ns(void)
2362 struct mnt_namespace *new_ns;
2364 new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL);
2366 return ERR_PTR(-ENOMEM);
2367 atomic_set(&new_ns->count, 1);
2368 new_ns->root = NULL;
2369 INIT_LIST_HEAD(&new_ns->list);
2370 init_waitqueue_head(&new_ns->poll);
2375 void mnt_make_longterm(struct vfsmount *mnt)
2377 __mnt_make_longterm(real_mount(mnt));
2380 void mnt_make_shortterm(struct vfsmount *m)
2383 struct mount *mnt = real_mount(m);
2384 if (atomic_add_unless(&mnt->mnt_longterm, -1, 1))
2386 br_write_lock(vfsmount_lock);
2387 atomic_dec(&mnt->mnt_longterm);
2388 br_write_unlock(vfsmount_lock);
2393 * Allocate a new namespace structure and populate it with contents
2394 * copied from the namespace of the passed in task structure.
2396 static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
2397 struct fs_struct *fs)
2399 struct mnt_namespace *new_ns;
2400 struct vfsmount *rootmnt = NULL, *pwdmnt = NULL;
2401 struct mount *p, *q;
2404 new_ns = alloc_mnt_ns();
2408 down_write(&namespace_sem);
2409 /* First pass: copy the tree topology */
2410 new = copy_tree(real_mount(mnt_ns->root), mnt_ns->root->mnt_root,
2411 CL_COPY_ALL | CL_EXPIRE);
2413 up_write(&namespace_sem);
2415 return ERR_PTR(-ENOMEM);
2417 new_ns->root = &new->mnt;
2418 br_write_lock(vfsmount_lock);
2419 list_add_tail(&new_ns->list, &new->mnt_list);
2420 br_write_unlock(vfsmount_lock);
2423 * Second pass: switch the tsk->fs->* elements and mark new vfsmounts
2424 * as belonging to new namespace. We have already acquired a private
2425 * fs_struct, so tsk->fs->lock is not needed.
2427 p = real_mount(mnt_ns->root);
2431 __mnt_make_longterm(q);
2433 if (&p->mnt == fs->root.mnt) {
2434 fs->root.mnt = mntget(&q->mnt);
2435 __mnt_make_longterm(q);
2436 mnt_make_shortterm(&p->mnt);
2439 if (&p->mnt == fs->pwd.mnt) {
2440 fs->pwd.mnt = mntget(&q->mnt);
2441 __mnt_make_longterm(q);
2442 mnt_make_shortterm(&p->mnt);
2446 p = next_mnt(p, mnt_ns->root);
2447 q = next_mnt(q, new_ns->root);
2449 up_write(&namespace_sem);
2459 struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns,
2460 struct fs_struct *new_fs)
2462 struct mnt_namespace *new_ns;
2467 if (!(flags & CLONE_NEWNS))
2470 new_ns = dup_mnt_ns(ns, new_fs);
2477 * create_mnt_ns - creates a private namespace and adds a root filesystem
2478 * @mnt: pointer to the new root filesystem mountpoint
2480 static struct mnt_namespace *create_mnt_ns(struct vfsmount *m)
2482 struct mnt_namespace *new_ns = alloc_mnt_ns();
2483 if (!IS_ERR(new_ns)) {
2484 struct mount *mnt = real_mount(m);
2485 mnt->mnt_ns = new_ns;
2486 __mnt_make_longterm(mnt);
2488 list_add(&new_ns->list, &mnt->mnt_list);
2495 struct dentry *mount_subtree(struct vfsmount *mnt, const char *name)
2497 struct mnt_namespace *ns;
2498 struct super_block *s;
2502 ns = create_mnt_ns(mnt);
2504 return ERR_CAST(ns);
2506 err = vfs_path_lookup(mnt->mnt_root, mnt,
2507 name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path);
2512 return ERR_PTR(err);
2514 /* trade a vfsmount reference for active sb one */
2515 s = path.mnt->mnt_sb;
2516 atomic_inc(&s->s_active);
2518 /* lock the sucker */
2519 down_write(&s->s_umount);
2520 /* ... and return the root of (sub)tree on it */
2523 EXPORT_SYMBOL(mount_subtree);
2525 SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name,
2526 char __user *, type, unsigned long, flags, void __user *, data)
2532 unsigned long data_page;
2534 ret = copy_mount_string(type, &kernel_type);
2538 kernel_dir = getname(dir_name);
2539 if (IS_ERR(kernel_dir)) {
2540 ret = PTR_ERR(kernel_dir);
2544 ret = copy_mount_string(dev_name, &kernel_dev);
2548 ret = copy_mount_options(data, &data_page);
2552 ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags,
2553 (void *) data_page);
2555 free_page(data_page);
2559 putname(kernel_dir);
2567 * Return true if path is reachable from root
2569 * namespace_sem or vfsmount_lock is held
2571 bool is_path_reachable(struct mount *mnt, struct dentry *dentry,
2572 const struct path *root)
2574 while (&mnt->mnt != root->mnt && mnt_has_parent(mnt)) {
2575 dentry = mnt->mnt_mountpoint;
2576 mnt = mnt->mnt_parent;
2578 return &mnt->mnt == root->mnt && is_subdir(dentry, root->dentry);
2581 int path_is_under(struct path *path1, struct path *path2)
2584 br_read_lock(vfsmount_lock);
2585 res = is_path_reachable(real_mount(path1->mnt), path1->dentry, path2);
2586 br_read_unlock(vfsmount_lock);
2589 EXPORT_SYMBOL(path_is_under);
2592 * pivot_root Semantics:
2593 * Moves the root file system of the current process to the directory put_old,
2594 * makes new_root as the new root file system of the current process, and sets
2595 * root/cwd of all processes which had them on the current root to new_root.
2598 * The new_root and put_old must be directories, and must not be on the
2599 * same file system as the current process root. The put_old must be
2600 * underneath new_root, i.e. adding a non-zero number of /.. to the string
2601 * pointed to by put_old must yield the same directory as new_root. No other
2602 * file system may be mounted on put_old. After all, new_root is a mountpoint.
2604 * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem.
2605 * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives
2606 * in this situation.
2609 * - we don't move root/cwd if they are not at the root (reason: if something
2610 * cared enough to change them, it's probably wrong to force them elsewhere)
2611 * - it's okay to pick a root that isn't the root of a file system, e.g.
2612 * /nfs/my_root where /nfs is the mount point. It must be a mountpoint,
2613 * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root
2616 SYSCALL_DEFINE2(pivot_root, const char __user *, new_root,
2617 const char __user *, put_old)
2619 struct path new, old, parent_path, root_parent, root;
2620 struct mount *new_mnt, *root_mnt;
2623 if (!capable(CAP_SYS_ADMIN))
2626 error = user_path_dir(new_root, &new);
2630 error = user_path_dir(put_old, &old);
2634 error = security_sb_pivotroot(&old, &new);
2638 get_fs_root(current->fs, &root);
2639 error = lock_mount(&old);
2644 new_mnt = real_mount(new.mnt);
2645 root_mnt = real_mount(root.mnt);
2646 if (IS_MNT_SHARED(real_mount(old.mnt)) ||
2647 IS_MNT_SHARED(new_mnt->mnt_parent) ||
2648 IS_MNT_SHARED(root_mnt->mnt_parent))
2650 if (!check_mnt(root_mnt) || !check_mnt(new_mnt))
2653 if (d_unlinked(new.dentry))
2655 if (d_unlinked(old.dentry))
2658 if (new.mnt == root.mnt ||
2659 old.mnt == root.mnt)
2660 goto out4; /* loop, on the same file system */
2662 if (root.mnt->mnt_root != root.dentry)
2663 goto out4; /* not a mountpoint */
2664 if (!mnt_has_parent(root_mnt))
2665 goto out4; /* not attached */
2666 if (new.mnt->mnt_root != new.dentry)
2667 goto out4; /* not a mountpoint */
2668 if (!mnt_has_parent(new_mnt))
2669 goto out4; /* not attached */
2670 /* make sure we can reach put_old from new_root */
2671 if (!is_path_reachable(real_mount(old.mnt), old.dentry, &new))
2673 br_write_lock(vfsmount_lock);
2674 detach_mnt(new_mnt, &parent_path);
2675 detach_mnt(root_mnt, &root_parent);
2676 /* mount old root on put_old */
2677 attach_mnt(root_mnt, &old);
2678 /* mount new_root on / */
2679 attach_mnt(new_mnt, &root_parent);
2680 touch_mnt_namespace(current->nsproxy->mnt_ns);
2681 br_write_unlock(vfsmount_lock);
2682 chroot_fs_refs(&root, &new);
2687 path_put(&root_parent);
2688 path_put(&parent_path);
2700 static void __init init_mount_tree(void)
2702 struct vfsmount *mnt;
2703 struct mnt_namespace *ns;
2706 mnt = do_kern_mount("rootfs", 0, "rootfs", NULL);
2708 panic("Can't create rootfs");
2710 ns = create_mnt_ns(mnt);
2712 panic("Can't allocate initial namespace");
2714 init_task.nsproxy->mnt_ns = ns;
2717 root.mnt = ns->root;
2718 root.dentry = ns->root->mnt_root;
2720 set_fs_pwd(current->fs, &root);
2721 set_fs_root(current->fs, &root);
2724 void __init mnt_init(void)
2729 init_rwsem(&namespace_sem);
2731 mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount),
2732 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
2734 mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC);
2736 if (!mount_hashtable)
2737 panic("Failed to allocate mount hash table\n");
2739 printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE);
2741 for (u = 0; u < HASH_SIZE; u++)
2742 INIT_LIST_HEAD(&mount_hashtable[u]);
2744 br_lock_init(vfsmount_lock);
2748 printk(KERN_WARNING "%s: sysfs_init error: %d\n",
2750 fs_kobj = kobject_create_and_add("fs", NULL);
2752 printk(KERN_WARNING "%s: kobj create error\n", __func__);
2757 void put_mnt_ns(struct mnt_namespace *ns)
2759 LIST_HEAD(umount_list);
2761 if (!atomic_dec_and_test(&ns->count))
2763 down_write(&namespace_sem);
2764 br_write_lock(vfsmount_lock);
2765 umount_tree(real_mount(ns->root), 0, &umount_list);
2766 br_write_unlock(vfsmount_lock);
2767 up_write(&namespace_sem);
2768 release_mounts(&umount_list);
2772 struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
2774 struct vfsmount *mnt;
2775 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
2778 * it is a longterm mount, don't release mnt until
2779 * we unmount before file sys is unregistered
2781 mnt_make_longterm(mnt);
2785 EXPORT_SYMBOL_GPL(kern_mount_data);
2787 void kern_unmount(struct vfsmount *mnt)
2789 /* release long term mount so mount point can be released */
2790 if (!IS_ERR_OR_NULL(mnt)) {
2791 mnt_make_shortterm(mnt);
2795 EXPORT_SYMBOL(kern_unmount);
2797 bool our_mnt(struct vfsmount *mnt)
2799 return check_mnt(real_mount(mnt));