blob: c11b99af53cfb4928b20bc3a56995800201e1cd6 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/namespace.c
3 *
4 * (C) Copyright Al Viro 2000, 2001
5 * Released under GPL v2.
6 *
7 * Based on code from fs/super.c, copyright Linus Torvalds and others.
8 * Heavily rewritten.
9 */
10
Linus Torvalds1da177e2005-04-16 15:20:36 -070011#include <linux/syscalls.h>
12#include <linux/slab.h>
13#include <linux/sched.h>
Nick Piggin99b7db72010-08-18 04:37:39 +100014#include <linux/spinlock.h>
15#include <linux/percpu.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070016#include <linux/init.h>
Randy Dunlap15a67dd2006-09-29 01:58:57 -070017#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/acct.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080019#include <linux/capability.h>
Dave Hansen3d733632008-02-15 14:37:59 -080020#include <linux/cpumask.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/module.h>
Andrew Mortonf20a9ea2006-08-14 22:43:23 -070022#include <linux/sysfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023#include <linux/seq_file.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080024#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/namei.h>
Alexey Dobriyanb43f3cb2009-07-08 01:54:37 +040026#include <linux/nsproxy.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/security.h>
28#include <linux/mount.h>
David Howells07f3f052006-09-30 20:52:18 +020029#include <linux/ramfs.h>
Eric Dumazet13f14b42008-02-06 01:37:57 -080030#include <linux/log2.h>
Miklos Szeredi73cd49e2008-03-26 22:11:34 +010031#include <linux/idr.h>
Al Viro5ad4e532009-03-29 19:50:06 -040032#include <linux/fs_struct.h>
Andreas Gruenbacher2504c5d2009-12-17 21:24:27 -050033#include <linux/fsnotify.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <asm/uaccess.h>
35#include <asm/unistd.h>
Ram Pai07b20882005-11-07 17:19:07 -050036#include "pnode.h"
Adrian Bunk948730b2007-07-15 23:41:25 -070037#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
Eric Dumazet13f14b42008-02-06 01:37:57 -080039#define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head))
40#define HASH_SIZE (1UL << HASH_SHIFT)
41
Al Viro5addc5d2005-11-07 17:15:49 -050042static int event;
Miklos Szeredi73cd49e2008-03-26 22:11:34 +010043static DEFINE_IDA(mnt_id_ida);
Miklos Szeredi719f5d72008-03-27 13:06:23 +010044static DEFINE_IDA(mnt_group_ida);
Nick Piggin99b7db72010-08-18 04:37:39 +100045static DEFINE_SPINLOCK(mnt_id_lock);
Al Virof21f6222009-06-24 03:12:00 -040046static int mnt_id_start = 0;
47static int mnt_group_start = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -070048
Eric Dumazetfa3536c2006-03-26 01:37:24 -080049static struct list_head *mount_hashtable __read_mostly;
Christoph Lametere18b8902006-12-06 20:33:20 -080050static struct kmem_cache *mnt_cache __read_mostly;
Ram Pai390c6842005-11-07 17:17:51 -050051static struct rw_semaphore namespace_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070052
Miklos Szeredif87fd4c2006-01-16 22:14:23 -080053/* /sys/fs */
Greg Kroah-Hartman00d26662007-10-29 14:17:23 -060054struct kobject *fs_kobj;
55EXPORT_SYMBOL_GPL(fs_kobj);
Miklos Szeredif87fd4c2006-01-16 22:14:23 -080056
Nick Piggin99b7db72010-08-18 04:37:39 +100057/*
58 * vfsmount lock may be taken for read to prevent changes to the
59 * vfsmount hash, ie. during mountpoint lookups or walking back
60 * up the tree.
61 *
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.
64 */
65DEFINE_BRLOCK(vfsmount_lock);
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry)
68{
Ram Paib58fed82005-11-07 17:16:09 -050069 unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES);
70 tmp += ((unsigned long)dentry / L1_CACHE_BYTES);
Eric Dumazet13f14b42008-02-06 01:37:57 -080071 tmp = tmp + (tmp >> HASH_SHIFT);
72 return tmp & (HASH_SIZE - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073}
74
Dave Hansen3d733632008-02-15 14:37:59 -080075#define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16)
76
Nick Piggin99b7db72010-08-18 04:37:39 +100077/*
78 * allocation is serialized by namespace_sem, but we need the spinlock to
79 * serialize with freeing.
80 */
Al Virob105e272011-11-24 20:38:33 -050081static int mnt_alloc_id(struct mount *mnt)
Miklos Szeredi73cd49e2008-03-26 22:11:34 +010082{
83 int res;
84
85retry:
86 ida_pre_get(&mnt_id_ida, GFP_KERNEL);
Nick Piggin99b7db72010-08-18 04:37:39 +100087 spin_lock(&mnt_id_lock);
Al Virob105e272011-11-24 20:38:33 -050088 res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt.mnt_id);
Al Virof21f6222009-06-24 03:12:00 -040089 if (!res)
Al Virob105e272011-11-24 20:38:33 -050090 mnt_id_start = mnt->mnt.mnt_id + 1;
Nick Piggin99b7db72010-08-18 04:37:39 +100091 spin_unlock(&mnt_id_lock);
Miklos Szeredi73cd49e2008-03-26 22:11:34 +010092 if (res == -EAGAIN)
93 goto retry;
94
95 return res;
96}
97
Al Virob105e272011-11-24 20:38:33 -050098static void mnt_free_id(struct mount *mnt)
Miklos Szeredi73cd49e2008-03-26 22:11:34 +010099{
Al Virob105e272011-11-24 20:38:33 -0500100 int id = mnt->mnt.mnt_id;
Nick Piggin99b7db72010-08-18 04:37:39 +1000101 spin_lock(&mnt_id_lock);
Al Virof21f6222009-06-24 03:12:00 -0400102 ida_remove(&mnt_id_ida, id);
103 if (mnt_id_start > id)
104 mnt_id_start = id;
Nick Piggin99b7db72010-08-18 04:37:39 +1000105 spin_unlock(&mnt_id_lock);
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100106}
107
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100108/*
109 * Allocate a new peer group ID
110 *
111 * mnt_group_ida is protected by namespace_sem
112 */
Al Viro4b8b21f2011-11-24 19:54:23 -0500113static int mnt_alloc_group_id(struct mount *mnt)
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100114{
Al Virof21f6222009-06-24 03:12:00 -0400115 int res;
116
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100117 if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL))
118 return -ENOMEM;
119
Al Virof21f6222009-06-24 03:12:00 -0400120 res = ida_get_new_above(&mnt_group_ida,
121 mnt_group_start,
Al Viro4b8b21f2011-11-24 19:54:23 -0500122 &mnt->mnt.mnt_group_id);
Al Virof21f6222009-06-24 03:12:00 -0400123 if (!res)
Al Viro4b8b21f2011-11-24 19:54:23 -0500124 mnt_group_start = mnt->mnt.mnt_group_id + 1;
Al Virof21f6222009-06-24 03:12:00 -0400125
126 return res;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100127}
128
129/*
130 * Release a peer group ID
131 */
Al Viro4b8b21f2011-11-24 19:54:23 -0500132void mnt_release_group_id(struct mount *mnt)
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100133{
Al Viro4b8b21f2011-11-24 19:54:23 -0500134 int id = mnt->mnt.mnt_group_id;
Al Virof21f6222009-06-24 03:12:00 -0400135 ida_remove(&mnt_group_ida, id);
136 if (mnt_group_start > id)
137 mnt_group_start = id;
Al Viro4b8b21f2011-11-24 19:54:23 -0500138 mnt->mnt.mnt_group_id = 0;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100139}
140
Nick Pigginb3e19d92011-01-07 17:50:11 +1100141/*
142 * vfsmount lock must be held for read
143 */
144static inline void mnt_add_count(struct vfsmount *mnt, int n)
145{
146#ifdef CONFIG_SMP
147 this_cpu_add(mnt->mnt_pcp->mnt_count, n);
148#else
149 preempt_disable();
150 mnt->mnt_count += n;
151 preempt_enable();
152#endif
153}
154
Nick Pigginb3e19d92011-01-07 17:50:11 +1100155/*
156 * vfsmount lock must be held for write
157 */
158unsigned int mnt_get_count(struct vfsmount *mnt)
159{
160#ifdef CONFIG_SMP
Al Virof03c6592011-01-14 22:30:21 -0500161 unsigned int count = 0;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100162 int cpu;
163
164 for_each_possible_cpu(cpu) {
165 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count;
166 }
167
168 return count;
169#else
170 return mnt->mnt_count;
171#endif
172}
173
Al Virob105e272011-11-24 20:38:33 -0500174static struct mount *alloc_vfsmnt(const char *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175{
Al Viro7d6fec42011-11-23 12:14:10 -0500176 struct mount *p = kmem_cache_zalloc(mnt_cache, GFP_KERNEL);
177 if (p) {
178 struct vfsmount *mnt = &p->mnt;
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100179 int err;
180
Al Virob105e272011-11-24 20:38:33 -0500181 err = mnt_alloc_id(p);
Li Zefan88b38782008-07-21 18:06:36 +0800182 if (err)
183 goto out_free_cache;
184
185 if (name) {
186 mnt->mnt_devname = kstrdup(name, GFP_KERNEL);
187 if (!mnt->mnt_devname)
188 goto out_free_id;
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100189 }
190
Nick Pigginb3e19d92011-01-07 17:50:11 +1100191#ifdef CONFIG_SMP
192 mnt->mnt_pcp = alloc_percpu(struct mnt_pcp);
193 if (!mnt->mnt_pcp)
194 goto out_free_devname;
195
Al Virof03c6592011-01-14 22:30:21 -0500196 this_cpu_add(mnt->mnt_pcp->mnt_count, 1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100197#else
198 mnt->mnt_count = 1;
199 mnt->mnt_writers = 0;
200#endif
201
Al Viro1b8e5562011-11-24 21:01:32 -0500202 INIT_LIST_HEAD(&p->mnt_hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203 INIT_LIST_HEAD(&mnt->mnt_child);
204 INIT_LIST_HEAD(&mnt->mnt_mounts);
205 INIT_LIST_HEAD(&mnt->mnt_list);
Miklos Szeredi55e700b2005-07-07 17:57:30 -0700206 INIT_LIST_HEAD(&mnt->mnt_expire);
Ram Pai03e06e62005-11-07 17:19:33 -0500207 INIT_LIST_HEAD(&mnt->mnt_share);
Ram Paia58b0eb2005-11-07 17:20:48 -0500208 INIT_LIST_HEAD(&mnt->mnt_slave_list);
209 INIT_LIST_HEAD(&mnt->mnt_slave);
Andreas Gruenbacher2504c5d2009-12-17 21:24:27 -0500210#ifdef CONFIG_FSNOTIFY
211 INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks);
212#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213 }
Al Virob105e272011-11-24 20:38:33 -0500214 return p;
Li Zefan88b38782008-07-21 18:06:36 +0800215
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000216#ifdef CONFIG_SMP
217out_free_devname:
Al Viro7d6fec42011-11-23 12:14:10 -0500218 kfree(p->mnt.mnt_devname);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000219#endif
Li Zefan88b38782008-07-21 18:06:36 +0800220out_free_id:
Al Virob105e272011-11-24 20:38:33 -0500221 mnt_free_id(p);
Li Zefan88b38782008-07-21 18:06:36 +0800222out_free_cache:
Al Viro7d6fec42011-11-23 12:14:10 -0500223 kmem_cache_free(mnt_cache, p);
Li Zefan88b38782008-07-21 18:06:36 +0800224 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700225}
226
Dave Hansen83660252008-02-15 14:37:30 -0800227/*
228 * Most r/o checks on a fs are for operations that take
229 * discrete amounts of time, like a write() or unlink().
230 * We must keep track of when those operations start
231 * (for permission checks) and when they end, so that
232 * we can determine when writes are able to occur to
233 * a filesystem.
234 */
Dave Hansen3d733632008-02-15 14:37:59 -0800235/*
236 * __mnt_is_readonly: check whether a mount is read-only
237 * @mnt: the mount to check for its write status
238 *
239 * This shouldn't be used directly ouside of the VFS.
240 * It does not guarantee that the filesystem will stay
241 * r/w, just that it is right *now*. This can not and
242 * should not be used in place of IS_RDONLY(inode).
243 * mnt_want/drop_write() will _keep_ the filesystem
244 * r/w.
245 */
246int __mnt_is_readonly(struct vfsmount *mnt)
247{
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800248 if (mnt->mnt_flags & MNT_READONLY)
249 return 1;
250 if (mnt->mnt_sb->s_flags & MS_RDONLY)
251 return 1;
252 return 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800253}
254EXPORT_SYMBOL_GPL(__mnt_is_readonly);
255
Nick Pigginc6653a82011-01-07 17:50:10 +1100256static inline void mnt_inc_writers(struct vfsmount *mnt)
Dave Hansen3d733632008-02-15 14:37:59 -0800257{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000258#ifdef CONFIG_SMP
Nick Pigginb3e19d92011-01-07 17:50:11 +1100259 this_cpu_inc(mnt->mnt_pcp->mnt_writers);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000260#else
261 mnt->mnt_writers++;
262#endif
Dave Hansen3d733632008-02-15 14:37:59 -0800263}
Dave Hansen3d733632008-02-15 14:37:59 -0800264
Nick Pigginc6653a82011-01-07 17:50:10 +1100265static inline void mnt_dec_writers(struct vfsmount *mnt)
Dave Hansen3d733632008-02-15 14:37:59 -0800266{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000267#ifdef CONFIG_SMP
Nick Pigginb3e19d92011-01-07 17:50:11 +1100268 this_cpu_dec(mnt->mnt_pcp->mnt_writers);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000269#else
270 mnt->mnt_writers--;
271#endif
272}
273
Nick Pigginc6653a82011-01-07 17:50:10 +1100274static unsigned int mnt_get_writers(struct vfsmount *mnt)
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000275{
276#ifdef CONFIG_SMP
277 unsigned int count = 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800278 int cpu;
Dave Hansen3d733632008-02-15 14:37:59 -0800279
280 for_each_possible_cpu(cpu) {
Nick Pigginb3e19d92011-01-07 17:50:11 +1100281 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers;
Dave Hansen3d733632008-02-15 14:37:59 -0800282 }
Dave Hansen3d733632008-02-15 14:37:59 -0800283
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000284 return count;
285#else
286 return mnt->mnt_writers;
287#endif
Dave Hansen3d733632008-02-15 14:37:59 -0800288}
289
290/*
291 * Most r/o checks on a fs are for operations that take
292 * discrete amounts of time, like a write() or unlink().
293 * We must keep track of when those operations start
294 * (for permission checks) and when they end, so that
295 * we can determine when writes are able to occur to
296 * a filesystem.
297 */
Dave Hansen83660252008-02-15 14:37:30 -0800298/**
299 * mnt_want_write - get write access to a mount
300 * @mnt: the mount on which to take a write
301 *
302 * This tells the low-level filesystem that a write is
303 * about to be performed to it, and makes sure that
304 * writes are allowed before returning success. When
305 * the write operation is finished, mnt_drop_write()
306 * must be called. This is effectively a refcount.
307 */
308int mnt_want_write(struct vfsmount *mnt)
309{
Dave Hansen3d733632008-02-15 14:37:59 -0800310 int ret = 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800311
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000312 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100313 mnt_inc_writers(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000314 /*
Nick Pigginc6653a82011-01-07 17:50:10 +1100315 * The store to mnt_inc_writers must be visible before we pass
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000316 * MNT_WRITE_HOLD loop below, so that the slowpath can see our
317 * incremented count after it has set MNT_WRITE_HOLD.
318 */
319 smp_mb();
320 while (mnt->mnt_flags & MNT_WRITE_HOLD)
321 cpu_relax();
322 /*
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.
326 */
327 smp_rmb();
Dave Hansen3d733632008-02-15 14:37:59 -0800328 if (__mnt_is_readonly(mnt)) {
Nick Pigginc6653a82011-01-07 17:50:10 +1100329 mnt_dec_writers(mnt);
Dave Hansen3d733632008-02-15 14:37:59 -0800330 ret = -EROFS;
331 goto out;
332 }
Dave Hansen3d733632008-02-15 14:37:59 -0800333out:
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000334 preempt_enable();
Dave Hansen3d733632008-02-15 14:37:59 -0800335 return ret;
Dave Hansen83660252008-02-15 14:37:30 -0800336}
337EXPORT_SYMBOL_GPL(mnt_want_write);
338
339/**
npiggin@suse.de96029c42009-04-26 20:25:55 +1000340 * mnt_clone_write - get write access to a mount
341 * @mnt: the mount on which to take a write
342 *
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.
347 *
348 * After finished, mnt_drop_write must be called as usual to
349 * drop the reference.
350 */
351int mnt_clone_write(struct vfsmount *mnt)
352{
353 /* superblock may be r/o */
354 if (__mnt_is_readonly(mnt))
355 return -EROFS;
356 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100357 mnt_inc_writers(mnt);
npiggin@suse.de96029c42009-04-26 20:25:55 +1000358 preempt_enable();
359 return 0;
360}
361EXPORT_SYMBOL_GPL(mnt_clone_write);
362
363/**
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
366 *
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
369 */
370int mnt_want_write_file(struct file *file)
371{
OGAWA Hirofumi2d8dd382009-08-06 15:07:39 -0700372 struct inode *inode = file->f_dentry->d_inode;
373 if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode))
npiggin@suse.de96029c42009-04-26 20:25:55 +1000374 return mnt_want_write(file->f_path.mnt);
375 else
376 return mnt_clone_write(file->f_path.mnt);
377}
378EXPORT_SYMBOL_GPL(mnt_want_write_file);
379
380/**
Dave Hansen83660252008-02-15 14:37:30 -0800381 * mnt_drop_write - give up write access to a mount
382 * @mnt: the mount on which to give up write access
383 *
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.
387 */
388void mnt_drop_write(struct vfsmount *mnt)
389{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000390 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100391 mnt_dec_writers(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000392 preempt_enable();
Dave Hansen83660252008-02-15 14:37:30 -0800393}
394EXPORT_SYMBOL_GPL(mnt_drop_write);
395
Al Viro2a79f172011-12-09 08:06:57 -0500396void mnt_drop_write_file(struct file *file)
397{
398 mnt_drop_write(file->f_path.mnt);
399}
400EXPORT_SYMBOL(mnt_drop_write_file);
401
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800402static int mnt_make_readonly(struct vfsmount *mnt)
Dave Hansen83660252008-02-15 14:37:30 -0800403{
Dave Hansen3d733632008-02-15 14:37:59 -0800404 int ret = 0;
405
Nick Piggin99b7db72010-08-18 04:37:39 +1000406 br_write_lock(vfsmount_lock);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000407 mnt->mnt_flags |= MNT_WRITE_HOLD;
408 /*
409 * After storing MNT_WRITE_HOLD, we'll read the counters. This store
410 * should be visible before we do.
411 */
412 smp_mb();
413
414 /*
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).
419 *
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.
429 */
Nick Pigginc6653a82011-01-07 17:50:10 +1100430 if (mnt_get_writers(mnt) > 0)
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000431 ret = -EBUSY;
432 else
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800433 mnt->mnt_flags |= MNT_READONLY;
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000434 /*
435 * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers
436 * that become unheld will see MNT_READONLY.
437 */
438 smp_wmb();
439 mnt->mnt_flags &= ~MNT_WRITE_HOLD;
Nick Piggin99b7db72010-08-18 04:37:39 +1000440 br_write_unlock(vfsmount_lock);
Dave Hansen3d733632008-02-15 14:37:59 -0800441 return ret;
Dave Hansen83660252008-02-15 14:37:30 -0800442}
Dave Hansen83660252008-02-15 14:37:30 -0800443
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800444static void __mnt_unmake_readonly(struct vfsmount *mnt)
445{
Nick Piggin99b7db72010-08-18 04:37:39 +1000446 br_write_lock(vfsmount_lock);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800447 mnt->mnt_flags &= ~MNT_READONLY;
Nick Piggin99b7db72010-08-18 04:37:39 +1000448 br_write_unlock(vfsmount_lock);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800449}
450
Al Virob105e272011-11-24 20:38:33 -0500451static void free_vfsmnt(struct mount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452{
Al Virob105e272011-11-24 20:38:33 -0500453 kfree(mnt->mnt.mnt_devname);
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100454 mnt_free_id(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000455#ifdef CONFIG_SMP
Al Virob105e272011-11-24 20:38:33 -0500456 free_percpu(mnt->mnt.mnt_pcp);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000457#endif
Al Virob105e272011-11-24 20:38:33 -0500458 kmem_cache_free(mnt_cache, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459}
460
461/*
Ram Paia05964f2005-11-07 17:20:17 -0500462 * 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.
Nick Piggin99b7db72010-08-18 04:37:39 +1000464 * vfsmount_lock must be held for read or write.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 */
Al Viroc7105362011-11-24 18:22:03 -0500466struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry,
Ram Paia05964f2005-11-07 17:20:17 -0500467 int dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468{
Ram Paib58fed82005-11-07 17:16:09 -0500469 struct list_head *head = mount_hashtable + hash(mnt, dentry);
470 struct list_head *tmp = head;
Al Viroc7105362011-11-24 18:22:03 -0500471 struct mount *p, *found = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 for (;;) {
Ram Paia05964f2005-11-07 17:20:17 -0500474 tmp = dir ? tmp->next : tmp->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 p = NULL;
476 if (tmp == head)
477 break;
Al Viro1b8e5562011-11-24 21:01:32 -0500478 p = list_entry(tmp, struct mount, mnt_hash);
Al Viroc7105362011-11-24 18:22:03 -0500479 if (p->mnt.mnt_parent == mnt && p->mnt.mnt_mountpoint == dentry) {
Ram Paia05964f2005-11-07 17:20:17 -0500480 found = p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 break;
482 }
483 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 return found;
485}
486
Ram Paia05964f2005-11-07 17:20:17 -0500487/*
488 * lookup_mnt increments the ref count before returning
489 * the vfsmount struct.
490 */
Al Viro1c755af2009-04-18 14:06:57 -0400491struct vfsmount *lookup_mnt(struct path *path)
Ram Paia05964f2005-11-07 17:20:17 -0500492{
Al Viroc7105362011-11-24 18:22:03 -0500493 struct mount *child_mnt;
Nick Piggin99b7db72010-08-18 04:37:39 +1000494
495 br_read_lock(vfsmount_lock);
Al Viroc7105362011-11-24 18:22:03 -0500496 child_mnt = __lookup_mnt(path->mnt, path->dentry, 1);
497 if (child_mnt) {
498 mnt_add_count(child_mnt, 1);
499 br_read_unlock(vfsmount_lock);
500 return &child_mnt->mnt;
501 } else {
502 br_read_unlock(vfsmount_lock);
503 return NULL;
504 }
Ram Paia05964f2005-11-07 17:20:17 -0500505}
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507static inline int check_mnt(struct vfsmount *mnt)
508{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800509 return mnt->mnt_ns == current->nsproxy->mnt_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510}
511
Nick Piggin99b7db72010-08-18 04:37:39 +1000512/*
513 * vfsmount lock must be held for write
514 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800515static void touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500516{
517 if (ns) {
518 ns->event = ++event;
519 wake_up_interruptible(&ns->poll);
520 }
521}
522
Nick Piggin99b7db72010-08-18 04:37:39 +1000523/*
524 * vfsmount lock must be held for write
525 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800526static void __touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500527{
528 if (ns && ns->event != event) {
529 ns->event = event;
530 wake_up_interruptible(&ns->poll);
531 }
532}
533
Nick Piggin99b7db72010-08-18 04:37:39 +1000534/*
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100535 * Clear dentry's mounted state if it has no remaining mounts.
536 * vfsmount_lock must be held for write.
537 */
Al Viroaa0a4cf2011-11-24 19:31:36 -0500538static void dentry_reset_mounted(struct dentry *dentry)
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100539{
540 unsigned u;
541
542 for (u = 0; u < HASH_SIZE; u++) {
Al Virod5e50f72011-11-24 20:58:57 -0500543 struct mount *p;
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100544
Al Viro1b8e5562011-11-24 21:01:32 -0500545 list_for_each_entry(p, &mount_hashtable[u], mnt_hash) {
Al Virod5e50f72011-11-24 20:58:57 -0500546 if (p->mnt.mnt_mountpoint == dentry)
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100547 return;
548 }
549 }
550 spin_lock(&dentry->d_lock);
551 dentry->d_flags &= ~DCACHE_MOUNTED;
552 spin_unlock(&dentry->d_lock);
553}
554
555/*
Nick Piggin99b7db72010-08-18 04:37:39 +1000556 * vfsmount lock must be held for write
557 */
Al Viro419148d2011-11-24 19:41:16 -0500558static void detach_mnt(struct mount *mnt, struct path *old_path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559{
Al Viro419148d2011-11-24 19:41:16 -0500560 old_path->dentry = mnt->mnt.mnt_mountpoint;
561 old_path->mnt = mnt->mnt.mnt_parent;
562 mnt->mnt.mnt_parent = &mnt->mnt;
563 mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root;
564 list_del_init(&mnt->mnt.mnt_child);
Al Viro1b8e5562011-11-24 21:01:32 -0500565 list_del_init(&mnt->mnt_hash);
Al Viroaa0a4cf2011-11-24 19:31:36 -0500566 dentry_reset_mounted(old_path->dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567}
568
Nick Piggin99b7db72010-08-18 04:37:39 +1000569/*
570 * vfsmount lock must be held for write
571 */
Ram Paib90fa9a2005-11-07 17:19:50 -0500572void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry,
Al Viro44d964d62011-11-24 21:28:22 -0500573 struct mount *child_mnt)
Ram Paib90fa9a2005-11-07 17:19:50 -0500574{
Al Viro44d964d62011-11-24 21:28:22 -0500575 child_mnt->mnt.mnt_parent = mntget(mnt);
576 child_mnt->mnt.mnt_mountpoint = dget(dentry);
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100577 spin_lock(&dentry->d_lock);
578 dentry->d_flags |= DCACHE_MOUNTED;
579 spin_unlock(&dentry->d_lock);
Ram Paib90fa9a2005-11-07 17:19:50 -0500580}
581
Nick Piggin99b7db72010-08-18 04:37:39 +1000582/*
583 * vfsmount lock must be held for write
584 */
Al Viro419148d2011-11-24 19:41:16 -0500585static void attach_mnt(struct mount *mnt, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586{
Al Viro44d964d62011-11-24 21:28:22 -0500587 mnt_set_mountpoint(path->mnt, path->dentry, mnt);
Al Viro1b8e5562011-11-24 21:01:32 -0500588 list_add_tail(&mnt->mnt_hash, mount_hashtable +
Al Viro1a390682008-03-21 20:48:19 -0400589 hash(path->mnt, path->dentry));
Al Viro419148d2011-11-24 19:41:16 -0500590 list_add_tail(&mnt->mnt.mnt_child, &path->mnt->mnt_mounts);
Ram Paib90fa9a2005-11-07 17:19:50 -0500591}
592
Al Viro7e3d0eb02011-01-16 16:32:11 -0500593static inline void __mnt_make_longterm(struct vfsmount *mnt)
594{
595#ifdef CONFIG_SMP
596 atomic_inc(&mnt->mnt_longterm);
597#endif
598}
599
600/* needs vfsmount lock for write */
601static inline void __mnt_make_shortterm(struct vfsmount *mnt)
602{
603#ifdef CONFIG_SMP
604 atomic_dec(&mnt->mnt_longterm);
605#endif
606}
607
Ram Paib90fa9a2005-11-07 17:19:50 -0500608/*
Nick Piggin99b7db72010-08-18 04:37:39 +1000609 * vfsmount lock must be held for write
Ram Paib90fa9a2005-11-07 17:19:50 -0500610 */
Al Viro4b2619a2011-11-24 19:47:15 -0500611static void commit_tree(struct mount *mnt)
Ram Paib90fa9a2005-11-07 17:19:50 -0500612{
Al Viro4b2619a2011-11-24 19:47:15 -0500613 struct vfsmount *parent = mnt->mnt.mnt_parent;
Ram Paib90fa9a2005-11-07 17:19:50 -0500614 struct vfsmount *m;
615 LIST_HEAD(head);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800616 struct mnt_namespace *n = parent->mnt_ns;
Ram Paib90fa9a2005-11-07 17:19:50 -0500617
Al Viro4b2619a2011-11-24 19:47:15 -0500618 BUG_ON(parent == &mnt->mnt);
Ram Paib90fa9a2005-11-07 17:19:50 -0500619
Al Viro4b2619a2011-11-24 19:47:15 -0500620 list_add_tail(&head, &mnt->mnt.mnt_list);
Al Virof03c6592011-01-14 22:30:21 -0500621 list_for_each_entry(m, &head, mnt_list) {
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800622 m->mnt_ns = n;
Al Viro7e3d0eb02011-01-16 16:32:11 -0500623 __mnt_make_longterm(m);
Al Virof03c6592011-01-14 22:30:21 -0500624 }
625
Ram Paib90fa9a2005-11-07 17:19:50 -0500626 list_splice(&head, n->list.prev);
627
Al Viro1b8e5562011-11-24 21:01:32 -0500628 list_add_tail(&mnt->mnt_hash, mount_hashtable +
Al Viro4b2619a2011-11-24 19:47:15 -0500629 hash(parent, mnt->mnt.mnt_mountpoint));
630 list_add_tail(&mnt->mnt.mnt_child, &parent->mnt_mounts);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800631 touch_mnt_namespace(n);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632}
633
Al Viro315fc832011-11-24 18:57:30 -0500634static struct mount *next_mnt(struct mount *p, struct vfsmount *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635{
Al Viro315fc832011-11-24 18:57:30 -0500636 struct list_head *next = p->mnt.mnt_mounts.next;
637 if (next == &p->mnt.mnt_mounts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 while (1) {
Al Viro315fc832011-11-24 18:57:30 -0500639 if (&p->mnt == root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 return NULL;
Al Viro315fc832011-11-24 18:57:30 -0500641 next = p->mnt.mnt_child.next;
642 if (next != &p->mnt.mnt_parent->mnt_mounts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 break;
Al Viro315fc832011-11-24 18:57:30 -0500644 p = real_mount(p->mnt.mnt_parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645 }
646 }
Al Viro315fc832011-11-24 18:57:30 -0500647 return list_entry(next, struct mount, mnt.mnt_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648}
649
Al Viro315fc832011-11-24 18:57:30 -0500650static struct mount *skip_mnt_tree(struct mount *p)
Ram Pai9676f0c2005-11-07 17:21:20 -0500651{
Al Viro315fc832011-11-24 18:57:30 -0500652 struct list_head *prev = p->mnt.mnt_mounts.prev;
653 while (prev != &p->mnt.mnt_mounts) {
654 p = list_entry(prev, struct mount, mnt.mnt_child);
655 prev = p->mnt.mnt_mounts.prev;
Ram Pai9676f0c2005-11-07 17:21:20 -0500656 }
657 return p;
658}
659
Al Viro9d412a42011-03-17 22:08:28 -0400660struct vfsmount *
661vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
662{
Al Virob105e272011-11-24 20:38:33 -0500663 struct mount *mnt;
Al Viro9d412a42011-03-17 22:08:28 -0400664 struct dentry *root;
665
666 if (!type)
667 return ERR_PTR(-ENODEV);
668
669 mnt = alloc_vfsmnt(name);
670 if (!mnt)
671 return ERR_PTR(-ENOMEM);
672
673 if (flags & MS_KERNMOUNT)
Al Virob105e272011-11-24 20:38:33 -0500674 mnt->mnt.mnt_flags = MNT_INTERNAL;
Al Viro9d412a42011-03-17 22:08:28 -0400675
676 root = mount_fs(type, flags, name, data);
677 if (IS_ERR(root)) {
678 free_vfsmnt(mnt);
679 return ERR_CAST(root);
680 }
681
Al Virob105e272011-11-24 20:38:33 -0500682 mnt->mnt.mnt_root = root;
683 mnt->mnt.mnt_sb = root->d_sb;
684 mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root;
685 mnt->mnt.mnt_parent = &mnt->mnt;
686 return &mnt->mnt;
Al Viro9d412a42011-03-17 22:08:28 -0400687}
688EXPORT_SYMBOL_GPL(vfs_kern_mount);
689
Al Viro87129cc2011-11-24 21:24:27 -0500690static struct mount *clone_mnt(struct mount *old, struct dentry *root,
Ram Pai36341f62005-11-07 17:17:22 -0500691 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692{
Al Viro87129cc2011-11-24 21:24:27 -0500693 struct super_block *sb = old->mnt.mnt_sb;
694 struct mount *mnt = alloc_vfsmnt(old->mnt.mnt_devname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
696 if (mnt) {
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100697 if (flag & (CL_SLAVE | CL_PRIVATE))
Al Virob105e272011-11-24 20:38:33 -0500698 mnt->mnt.mnt_group_id = 0; /* not a peer of original */
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100699 else
Al Viro87129cc2011-11-24 21:24:27 -0500700 mnt->mnt.mnt_group_id = old->mnt.mnt_group_id;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100701
Al Virob105e272011-11-24 20:38:33 -0500702 if ((flag & CL_MAKE_SHARED) && !mnt->mnt.mnt_group_id) {
703 int err = mnt_alloc_group_id(mnt);
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100704 if (err)
705 goto out_free;
706 }
707
Al Viro87129cc2011-11-24 21:24:27 -0500708 mnt->mnt.mnt_flags = old->mnt.mnt_flags & ~MNT_WRITE_HOLD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 atomic_inc(&sb->s_active);
Al Virob105e272011-11-24 20:38:33 -0500710 mnt->mnt.mnt_sb = sb;
711 mnt->mnt.mnt_root = dget(root);
712 mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root;
713 mnt->mnt.mnt_parent = &mnt->mnt;
Ram Paib90fa9a2005-11-07 17:19:50 -0500714
Ram Pai5afe0022005-11-07 17:21:01 -0500715 if (flag & CL_SLAVE) {
Al Viro87129cc2011-11-24 21:24:27 -0500716 list_add(&mnt->mnt.mnt_slave, &old->mnt.mnt_slave_list);
717 mnt->mnt.mnt_master = &old->mnt;
Al Virob105e272011-11-24 20:38:33 -0500718 CLEAR_MNT_SHARED(&mnt->mnt);
Al Viro8aec0802007-06-07 12:20:32 -0400719 } else if (!(flag & CL_PRIVATE)) {
Al Viro87129cc2011-11-24 21:24:27 -0500720 if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(&old->mnt))
721 list_add(&mnt->mnt.mnt_share, &old->mnt.mnt_share);
722 if (IS_MNT_SLAVE(&old->mnt))
723 list_add(&mnt->mnt.mnt_slave, &old->mnt.mnt_slave);
724 mnt->mnt.mnt_master = old->mnt.mnt_master;
Ram Pai5afe0022005-11-07 17:21:01 -0500725 }
Ram Paib90fa9a2005-11-07 17:19:50 -0500726 if (flag & CL_MAKE_SHARED)
Al Viro0f0afb12011-11-24 20:43:10 -0500727 set_mnt_shared(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
729 /* stick the duplicate mount on the same expiry list
730 * as the original if that was on one */
Ram Pai36341f62005-11-07 17:17:22 -0500731 if (flag & CL_EXPIRE) {
Al Viro87129cc2011-11-24 21:24:27 -0500732 if (!list_empty(&old->mnt.mnt_expire))
733 list_add(&mnt->mnt.mnt_expire, &old->mnt.mnt_expire);
Ram Pai36341f62005-11-07 17:17:22 -0500734 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 }
Al Virocb338d02011-11-24 20:55:08 -0500736 return mnt;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100737
738 out_free:
739 free_vfsmnt(mnt);
740 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741}
742
Nick Pigginb3e19d92011-01-07 17:50:11 +1100743static inline void mntfree(struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744{
745 struct super_block *sb = mnt->mnt_sb;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100746
Dave Hansen3d733632008-02-15 14:37:59 -0800747 /*
Dave Hansen3d733632008-02-15 14:37:59 -0800748 * This probably indicates that somebody messed
749 * up a mnt_want/drop_write() pair. If this
750 * happens, the filesystem was probably unable
751 * to make r/w->r/o transitions.
752 */
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000753 /*
Nick Pigginb3e19d92011-01-07 17:50:11 +1100754 * The locking used to deal with mnt_count decrement provides barriers,
755 * so mnt_get_writers() below is safe.
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000756 */
Nick Pigginc6653a82011-01-07 17:50:10 +1100757 WARN_ON(mnt_get_writers(mnt));
Andreas Gruenbacherca9c7262009-12-17 21:24:27 -0500758 fsnotify_vfsmount_delete(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 dput(mnt->mnt_root);
Al Virob105e272011-11-24 20:38:33 -0500760 free_vfsmnt(real_mount(mnt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 deactivate_super(sb);
762}
763
Al Virof03c6592011-01-14 22:30:21 -0500764static void mntput_no_expire(struct vfsmount *mnt)
Al Viro7b7b1ac2005-11-07 17:13:39 -0500765{
Nick Pigginb3e19d92011-01-07 17:50:11 +1100766put_again:
Al Virof03c6592011-01-14 22:30:21 -0500767#ifdef CONFIG_SMP
768 br_read_lock(vfsmount_lock);
769 if (likely(atomic_read(&mnt->mnt_longterm))) {
Al Viroaa9c0e02011-11-23 19:04:52 -0500770 mnt_add_count(mnt, -1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100771 br_read_unlock(vfsmount_lock);
Al Virof03c6592011-01-14 22:30:21 -0500772 return;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100773 }
Al Virof03c6592011-01-14 22:30:21 -0500774 br_read_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100775
776 br_write_lock(vfsmount_lock);
Al Viroaa9c0e02011-11-23 19:04:52 -0500777 mnt_add_count(mnt, -1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100778 if (mnt_get_count(mnt)) {
779 br_write_unlock(vfsmount_lock);
780 return;
781 }
Nick Pigginb3e19d92011-01-07 17:50:11 +1100782#else
Al Viroaa9c0e02011-11-23 19:04:52 -0500783 mnt_add_count(mnt, -1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100784 if (likely(mnt_get_count(mnt)))
Nick Piggin99b7db72010-08-18 04:37:39 +1000785 return;
786 br_write_lock(vfsmount_lock);
Al Virof03c6592011-01-14 22:30:21 -0500787#endif
Nick Pigginb3e19d92011-01-07 17:50:11 +1100788 if (unlikely(mnt->mnt_pinned)) {
789 mnt_add_count(mnt, mnt->mnt_pinned + 1);
790 mnt->mnt_pinned = 0;
Nick Piggin99b7db72010-08-18 04:37:39 +1000791 br_write_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100792 acct_auto_close_mnt(mnt);
793 goto put_again;
Al Viro7b7b1ac2005-11-07 17:13:39 -0500794 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000795 br_write_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100796 mntfree(mnt);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500797}
Nick Pigginb3e19d92011-01-07 17:50:11 +1100798
799void mntput(struct vfsmount *mnt)
800{
801 if (mnt) {
802 /* avoid cacheline pingpong, hope gcc doesn't get "smart" */
803 if (unlikely(mnt->mnt_expiry_mark))
804 mnt->mnt_expiry_mark = 0;
Al Virof03c6592011-01-14 22:30:21 -0500805 mntput_no_expire(mnt);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100806 }
807}
808EXPORT_SYMBOL(mntput);
809
810struct vfsmount *mntget(struct vfsmount *mnt)
811{
812 if (mnt)
Al Viroaa9c0e02011-11-23 19:04:52 -0500813 mnt_add_count(mnt, 1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100814 return mnt;
815}
816EXPORT_SYMBOL(mntget);
817
Al Viro7b7b1ac2005-11-07 17:13:39 -0500818void mnt_pin(struct vfsmount *mnt)
819{
Nick Piggin99b7db72010-08-18 04:37:39 +1000820 br_write_lock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500821 mnt->mnt_pinned++;
Nick Piggin99b7db72010-08-18 04:37:39 +1000822 br_write_unlock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500823}
Al Viro7b7b1ac2005-11-07 17:13:39 -0500824EXPORT_SYMBOL(mnt_pin);
825
826void mnt_unpin(struct vfsmount *mnt)
827{
Nick Piggin99b7db72010-08-18 04:37:39 +1000828 br_write_lock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500829 if (mnt->mnt_pinned) {
Al Viroaa9c0e02011-11-23 19:04:52 -0500830 mnt_add_count(mnt, 1);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500831 mnt->mnt_pinned--;
832 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000833 br_write_unlock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500834}
Al Viro7b7b1ac2005-11-07 17:13:39 -0500835EXPORT_SYMBOL(mnt_unpin);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800837static inline void mangle(struct seq_file *m, const char *s)
838{
839 seq_escape(m, s, " \t\n\\");
840}
841
842/*
843 * Simple .show_options callback for filesystems which don't want to
844 * implement more complex mount option showing.
845 *
846 * See also save_mount_options().
847 */
848int generic_show_options(struct seq_file *m, struct vfsmount *mnt)
849{
Al Viro2a32ceb2009-05-08 16:05:57 -0400850 const char *options;
851
852 rcu_read_lock();
853 options = rcu_dereference(mnt->mnt_sb->s_options);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800854
855 if (options != NULL && options[0]) {
856 seq_putc(m, ',');
857 mangle(m, options);
858 }
Al Viro2a32ceb2009-05-08 16:05:57 -0400859 rcu_read_unlock();
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800860
861 return 0;
862}
863EXPORT_SYMBOL(generic_show_options);
864
865/*
866 * If filesystem uses generic_show_options(), this function should be
867 * called from the fill_super() callback.
868 *
869 * The .remount_fs callback usually needs to be handled in a special
870 * way, to make sure, that previous options are not overwritten if the
871 * remount fails.
872 *
873 * Also note, that if the filesystem's .remount_fs function doesn't
874 * reset all options to their default value, but changes only newly
875 * given options, then the displayed options will not reflect reality
876 * any more.
877 */
878void save_mount_options(struct super_block *sb, char *options)
879{
Al Viro2a32ceb2009-05-08 16:05:57 -0400880 BUG_ON(sb->s_options);
881 rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL));
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800882}
883EXPORT_SYMBOL(save_mount_options);
884
Al Viro2a32ceb2009-05-08 16:05:57 -0400885void replace_mount_options(struct super_block *sb, char *options)
886{
887 char *old = sb->s_options;
888 rcu_assign_pointer(sb->s_options, options);
889 if (old) {
890 synchronize_rcu();
891 kfree(old);
892 }
893}
894EXPORT_SYMBOL(replace_mount_options);
895
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100896#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897/* iterator */
898static void *m_start(struct seq_file *m, loff_t *pos)
899{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100900 struct proc_mounts *p = m->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901
Ram Pai390c6842005-11-07 17:17:51 -0500902 down_read(&namespace_sem);
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100903 return seq_list_start(&p->ns->list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904}
905
906static void *m_next(struct seq_file *m, void *v, loff_t *pos)
907{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100908 struct proc_mounts *p = m->private;
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700909
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100910 return seq_list_next(v, &p->ns->list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911}
912
913static void m_stop(struct seq_file *m, void *v)
914{
Ram Pai390c6842005-11-07 17:17:51 -0500915 up_read(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
Al Viro9f5596a2010-02-05 00:40:25 -0500918int mnt_had_events(struct proc_mounts *p)
919{
920 struct mnt_namespace *ns = p->ns;
921 int res = 0;
922
Nick Piggin99b7db72010-08-18 04:37:39 +1000923 br_read_lock(vfsmount_lock);
Kay Sieversf1514632011-07-12 20:48:39 +0200924 if (p->m.poll_event != ns->event) {
925 p->m.poll_event = ns->event;
Al Viro9f5596a2010-02-05 00:40:25 -0500926 res = 1;
927 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000928 br_read_unlock(vfsmount_lock);
Al Viro9f5596a2010-02-05 00:40:25 -0500929
930 return res;
931}
932
Ram Pai2d4d4862008-03-27 13:06:25 +0100933struct proc_fs_info {
934 int flag;
935 const char *str;
936};
937
Eric Paris2069f452008-07-04 09:47:13 +1000938static int show_sb_opts(struct seq_file *m, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939{
Ram Pai2d4d4862008-03-27 13:06:25 +0100940 static const struct proc_fs_info fs_info[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941 { MS_SYNCHRONOUS, ",sync" },
942 { MS_DIRSYNC, ",dirsync" },
943 { MS_MANDLOCK, ",mand" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 { 0, NULL }
945 };
Ram Pai2d4d4862008-03-27 13:06:25 +0100946 const struct proc_fs_info *fs_infop;
947
948 for (fs_infop = fs_info; fs_infop->flag; fs_infop++) {
949 if (sb->s_flags & fs_infop->flag)
950 seq_puts(m, fs_infop->str);
951 }
Eric Paris2069f452008-07-04 09:47:13 +1000952
953 return security_sb_show_options(m, sb);
Ram Pai2d4d4862008-03-27 13:06:25 +0100954}
955
956static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt)
957{
958 static const struct proc_fs_info mnt_info[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 { MNT_NOSUID, ",nosuid" },
960 { MNT_NODEV, ",nodev" },
961 { MNT_NOEXEC, ",noexec" },
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -0800962 { MNT_NOATIME, ",noatime" },
963 { MNT_NODIRATIME, ",nodiratime" },
Valerie Henson47ae32d2006-12-13 00:34:34 -0800964 { MNT_RELATIME, ",relatime" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 { 0, NULL }
966 };
Ram Pai2d4d4862008-03-27 13:06:25 +0100967 const struct proc_fs_info *fs_infop;
968
969 for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) {
970 if (mnt->mnt_flags & fs_infop->flag)
971 seq_puts(m, fs_infop->str);
972 }
973}
974
975static void show_type(struct seq_file *m, struct super_block *sb)
976{
977 mangle(m, sb->s_type->name);
978 if (sb->s_subtype && sb->s_subtype[0]) {
979 seq_putc(m, '.');
980 mangle(m, sb->s_subtype);
981 }
982}
983
984static int show_vfsmnt(struct seq_file *m, void *v)
985{
986 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
987 int err = 0;
Jan Blunckc32c2f62008-02-14 19:38:43 -0800988 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989
Al Viroc7f404b2011-03-16 06:59:40 -0400990 if (mnt->mnt_sb->s_op->show_devname) {
991 err = mnt->mnt_sb->s_op->show_devname(m, mnt);
992 if (err)
993 goto out;
994 } else {
995 mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none");
996 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 seq_putc(m, ' ');
Jan Blunckc32c2f62008-02-14 19:38:43 -0800998 seq_path(m, &mnt_path, " \t\n\\");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999 seq_putc(m, ' ');
Ram Pai2d4d4862008-03-27 13:06:25 +01001000 show_type(m, mnt->mnt_sb);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001001 seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw");
Eric Paris2069f452008-07-04 09:47:13 +10001002 err = show_sb_opts(m, mnt->mnt_sb);
1003 if (err)
1004 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001005 show_mnt_opts(m, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 if (mnt->mnt_sb->s_op->show_options)
1007 err = mnt->mnt_sb->s_op->show_options(m, mnt);
1008 seq_puts(m, " 0 0\n");
Eric Paris2069f452008-07-04 09:47:13 +10001009out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 return err;
1011}
1012
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001013const struct seq_operations mounts_op = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 .start = m_start,
1015 .next = m_next,
1016 .stop = m_stop,
1017 .show = show_vfsmnt
1018};
1019
Ram Pai2d4d4862008-03-27 13:06:25 +01001020static int show_mountinfo(struct seq_file *m, void *v)
1021{
1022 struct proc_mounts *p = m->private;
1023 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
1024 struct super_block *sb = mnt->mnt_sb;
1025 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
1026 struct path root = p->root;
1027 int err = 0;
1028
1029 seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id,
1030 MAJOR(sb->s_dev), MINOR(sb->s_dev));
Al Viroc7f404b2011-03-16 06:59:40 -04001031 if (sb->s_op->show_path)
1032 err = sb->s_op->show_path(m, mnt);
1033 else
1034 seq_dentry(m, mnt->mnt_root, " \t\n\\");
1035 if (err)
1036 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001037 seq_putc(m, ' ');
Al Viro02125a82011-12-05 08:43:34 -05001038
1039 /* mountpoints outside of chroot jail will give SEQ_SKIP on this */
1040 err = seq_path_root(m, &mnt_path, &root, " \t\n\\");
1041 if (err)
1042 goto out;
1043
Ram Pai2d4d4862008-03-27 13:06:25 +01001044 seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw");
1045 show_mnt_opts(m, mnt);
1046
1047 /* Tagged fields ("foo:X" or "bar") */
1048 if (IS_MNT_SHARED(mnt))
1049 seq_printf(m, " shared:%i", mnt->mnt_group_id);
Miklos Szeredi97e7e0f2008-03-27 13:06:26 +01001050 if (IS_MNT_SLAVE(mnt)) {
1051 int master = mnt->mnt_master->mnt_group_id;
1052 int dom = get_dominating_id(mnt, &p->root);
1053 seq_printf(m, " master:%i", master);
1054 if (dom && dom != master)
1055 seq_printf(m, " propagate_from:%i", dom);
1056 }
Ram Pai2d4d4862008-03-27 13:06:25 +01001057 if (IS_MNT_UNBINDABLE(mnt))
1058 seq_puts(m, " unbindable");
1059
1060 /* Filesystem specific data */
1061 seq_puts(m, " - ");
1062 show_type(m, sb);
1063 seq_putc(m, ' ');
Al Viroc7f404b2011-03-16 06:59:40 -04001064 if (sb->s_op->show_devname)
1065 err = sb->s_op->show_devname(m, mnt);
1066 else
1067 mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none");
1068 if (err)
1069 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001070 seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw");
Eric Paris2069f452008-07-04 09:47:13 +10001071 err = show_sb_opts(m, sb);
1072 if (err)
1073 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001074 if (sb->s_op->show_options)
1075 err = sb->s_op->show_options(m, mnt);
1076 seq_putc(m, '\n');
Eric Paris2069f452008-07-04 09:47:13 +10001077out:
Ram Pai2d4d4862008-03-27 13:06:25 +01001078 return err;
1079}
1080
1081const struct seq_operations mountinfo_op = {
1082 .start = m_start,
1083 .next = m_next,
1084 .stop = m_stop,
1085 .show = show_mountinfo,
1086};
1087
Chuck Leverb4629fe2006-03-20 13:44:12 -05001088static int show_vfsstat(struct seq_file *m, void *v)
1089{
Pavel Emelianovb0765fb2007-07-15 23:39:55 -07001090 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
Jan Blunckc32c2f62008-02-14 19:38:43 -08001091 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
Chuck Leverb4629fe2006-03-20 13:44:12 -05001092 int err = 0;
1093
1094 /* device */
Al Viroc7f404b2011-03-16 06:59:40 -04001095 if (mnt->mnt_sb->s_op->show_devname) {
Bryan Schumakera877ee02011-10-07 13:41:15 -04001096 seq_puts(m, "device ");
Al Viroc7f404b2011-03-16 06:59:40 -04001097 err = mnt->mnt_sb->s_op->show_devname(m, mnt);
1098 } else {
1099 if (mnt->mnt_devname) {
1100 seq_puts(m, "device ");
1101 mangle(m, mnt->mnt_devname);
1102 } else
1103 seq_puts(m, "no device");
1104 }
Chuck Leverb4629fe2006-03-20 13:44:12 -05001105
1106 /* mount point */
1107 seq_puts(m, " mounted on ");
Jan Blunckc32c2f62008-02-14 19:38:43 -08001108 seq_path(m, &mnt_path, " \t\n\\");
Chuck Leverb4629fe2006-03-20 13:44:12 -05001109 seq_putc(m, ' ');
1110
1111 /* file system type */
1112 seq_puts(m, "with fstype ");
Ram Pai2d4d4862008-03-27 13:06:25 +01001113 show_type(m, mnt->mnt_sb);
Chuck Leverb4629fe2006-03-20 13:44:12 -05001114
1115 /* optional statistics */
1116 if (mnt->mnt_sb->s_op->show_stats) {
1117 seq_putc(m, ' ');
Al Viroc7f404b2011-03-16 06:59:40 -04001118 if (!err)
1119 err = mnt->mnt_sb->s_op->show_stats(m, mnt);
Chuck Leverb4629fe2006-03-20 13:44:12 -05001120 }
1121
1122 seq_putc(m, '\n');
1123 return err;
1124}
1125
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001126const struct seq_operations mountstats_op = {
Chuck Leverb4629fe2006-03-20 13:44:12 -05001127 .start = m_start,
1128 .next = m_next,
1129 .stop = m_stop,
1130 .show = show_vfsstat,
1131};
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001132#endif /* CONFIG_PROC_FS */
Chuck Leverb4629fe2006-03-20 13:44:12 -05001133
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134/**
1135 * may_umount_tree - check if a mount tree is busy
1136 * @mnt: root of mount tree
1137 *
1138 * This is called to check if a tree of mounts has any
1139 * open files, pwds, chroots or sub mounts that are
1140 * busy.
1141 */
1142int may_umount_tree(struct vfsmount *mnt)
1143{
Ram Pai36341f62005-11-07 17:17:22 -05001144 int actual_refs = 0;
1145 int minimum_refs = 0;
Al Viro315fc832011-11-24 18:57:30 -05001146 struct mount *p;
1147 BUG_ON(!mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
Nick Pigginb3e19d92011-01-07 17:50:11 +11001149 /* write lock needed for mnt_get_count */
1150 br_write_lock(vfsmount_lock);
Al Viro315fc832011-11-24 18:57:30 -05001151 for (p = real_mount(mnt); p; p = next_mnt(p, mnt)) {
1152 actual_refs += mnt_get_count(&p->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 minimum_refs += 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 }
Nick Pigginb3e19d92011-01-07 17:50:11 +11001155 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
1157 if (actual_refs > minimum_refs)
Ian Kente3474a82006-03-27 01:14:51 -08001158 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159
Ian Kente3474a82006-03-27 01:14:51 -08001160 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161}
1162
1163EXPORT_SYMBOL(may_umount_tree);
1164
1165/**
1166 * may_umount - check if a mount point is busy
1167 * @mnt: root of mount
1168 *
1169 * This is called to check if a mount point has any
1170 * open files, pwds, chroots or sub mounts. If the
1171 * mount has sub mounts this will return busy
1172 * regardless of whether the sub mounts are busy.
1173 *
1174 * Doesn't take quota and stuff into account. IOW, in some cases it will
1175 * give false negatives. The main reason why it's here is that we need
1176 * a non-destructive way to look for easily umountable filesystems.
1177 */
1178int may_umount(struct vfsmount *mnt)
1179{
Ian Kente3474a82006-03-27 01:14:51 -08001180 int ret = 1;
Al Viro8ad08d82010-01-16 12:56:08 -05001181 down_read(&namespace_sem);
Nick Pigginb3e19d92011-01-07 17:50:11 +11001182 br_write_lock(vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -05001183 if (propagate_mount_busy(mnt, 2))
Ian Kente3474a82006-03-27 01:14:51 -08001184 ret = 0;
Nick Pigginb3e19d92011-01-07 17:50:11 +11001185 br_write_unlock(vfsmount_lock);
Al Viro8ad08d82010-01-16 12:56:08 -05001186 up_read(&namespace_sem);
Ram Paia05964f2005-11-07 17:20:17 -05001187 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188}
1189
1190EXPORT_SYMBOL(may_umount);
1191
Ram Paib90fa9a2005-11-07 17:19:50 -05001192void release_mounts(struct list_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193{
Al Virod5e50f72011-11-24 20:58:57 -05001194 struct mount *mnt;
Miklos Szeredibf066c72006-01-08 01:03:19 -08001195 while (!list_empty(head)) {
Al Viro1b8e5562011-11-24 21:01:32 -05001196 mnt = list_first_entry(head, struct mount, mnt_hash);
1197 list_del_init(&mnt->mnt_hash);
Al Virod5e50f72011-11-24 20:58:57 -05001198 if (mnt_has_parent(&mnt->mnt)) {
Ram Pai70fbcdf2005-11-07 17:17:04 -05001199 struct dentry *dentry;
1200 struct vfsmount *m;
Nick Piggin99b7db72010-08-18 04:37:39 +10001201
1202 br_write_lock(vfsmount_lock);
Al Virod5e50f72011-11-24 20:58:57 -05001203 dentry = mnt->mnt.mnt_mountpoint;
1204 m = mnt->mnt.mnt_parent;
1205 mnt->mnt.mnt_mountpoint = mnt->mnt.mnt_root;
1206 mnt->mnt.mnt_parent = &mnt->mnt;
Al Viro7c4b93d2008-03-21 23:59:49 -04001207 m->mnt_ghosts--;
Nick Piggin99b7db72010-08-18 04:37:39 +10001208 br_write_unlock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001209 dput(dentry);
1210 mntput(m);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 }
Al Virod5e50f72011-11-24 20:58:57 -05001212 mntput(&mnt->mnt);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001213 }
1214}
1215
Nick Piggin99b7db72010-08-18 04:37:39 +10001216/*
1217 * vfsmount lock must be held for write
1218 * namespace_sem must be held for write
1219 */
Al Viro761d5c32011-11-24 21:07:43 -05001220void umount_tree(struct mount *mnt, int propagate, struct list_head *kill)
Ram Pai70fbcdf2005-11-07 17:17:04 -05001221{
Al Viro7b8a53f2011-01-15 20:08:44 -05001222 LIST_HEAD(tmp_list);
Al Viro315fc832011-11-24 18:57:30 -05001223 struct mount *p;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001224
Al Viro761d5c32011-11-24 21:07:43 -05001225 for (p = mnt; p; p = next_mnt(p, &mnt->mnt))
Al Viro1b8e5562011-11-24 21:01:32 -05001226 list_move(&p->mnt_hash, &tmp_list);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001227
Ram Paia05964f2005-11-07 17:20:17 -05001228 if (propagate)
Al Viro7b8a53f2011-01-15 20:08:44 -05001229 propagate_umount(&tmp_list);
Ram Paia05964f2005-11-07 17:20:17 -05001230
Al Viro1b8e5562011-11-24 21:01:32 -05001231 list_for_each_entry(p, &tmp_list, mnt_hash) {
Al Viro315fc832011-11-24 18:57:30 -05001232 list_del_init(&p->mnt.mnt_expire);
1233 list_del_init(&p->mnt.mnt_list);
1234 __touch_mnt_namespace(p->mnt.mnt_ns);
1235 p->mnt.mnt_ns = NULL;
1236 __mnt_make_shortterm(&p->mnt);
1237 list_del_init(&p->mnt.mnt_child);
1238 if (mnt_has_parent(&p->mnt)) {
1239 p->mnt.mnt_parent->mnt_ghosts++;
1240 dentry_reset_mounted(p->mnt.mnt_mountpoint);
Al Viro7c4b93d2008-03-21 23:59:49 -04001241 }
Al Viro0f0afb12011-11-24 20:43:10 -05001242 change_mnt_propagation(p, MS_PRIVATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 }
Al Viro7b8a53f2011-01-15 20:08:44 -05001244 list_splice(&tmp_list, kill);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245}
1246
Al Viro692afc32011-11-24 21:15:14 -05001247static void shrink_submounts(struct mount *mnt, struct list_head *umounts);
Al Viroc35038b2008-03-22 00:46:23 -04001248
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249static int do_umount(struct vfsmount *mnt, int flags)
1250{
Ram Paib58fed82005-11-07 17:16:09 -05001251 struct super_block *sb = mnt->mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 int retval;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001253 LIST_HEAD(umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254
1255 retval = security_sb_umount(mnt, flags);
1256 if (retval)
1257 return retval;
1258
1259 /*
1260 * Allow userspace to request a mountpoint be expired rather than
1261 * unmounting unconditionally. Unmount only happens if:
1262 * (1) the mark is already set (the mark is cleared by mntput())
1263 * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount]
1264 */
1265 if (flags & MNT_EXPIRE) {
Jan Blunck6ac08c32008-02-14 19:34:38 -08001266 if (mnt == current->fs->root.mnt ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 flags & (MNT_FORCE | MNT_DETACH))
1268 return -EINVAL;
1269
Nick Pigginb3e19d92011-01-07 17:50:11 +11001270 /*
1271 * probably don't strictly need the lock here if we examined
1272 * all race cases, but it's a slowpath.
1273 */
1274 br_write_lock(vfsmount_lock);
1275 if (mnt_get_count(mnt) != 2) {
J. R. Okajimabf9faa22011-02-23 16:59:49 +09001276 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 return -EBUSY;
Nick Pigginb3e19d92011-01-07 17:50:11 +11001278 }
1279 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
1281 if (!xchg(&mnt->mnt_expiry_mark, 1))
1282 return -EAGAIN;
1283 }
1284
1285 /*
1286 * If we may have to abort operations to get out of this
1287 * mount, and they will themselves hold resources we must
1288 * allow the fs to do things. In the Unix tradition of
1289 * 'Gee thats tricky lets do it in userspace' the umount_begin
1290 * might fail to complete on the first run through as other tasks
1291 * must return, and the like. Thats for the mount program to worry
1292 * about for the moment.
1293 */
1294
Al Viro42faad92008-04-24 07:21:56 -04001295 if (flags & MNT_FORCE && sb->s_op->umount_begin) {
Al Viro42faad92008-04-24 07:21:56 -04001296 sb->s_op->umount_begin(sb);
Al Viro42faad92008-04-24 07:21:56 -04001297 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298
1299 /*
1300 * No sense to grab the lock for this test, but test itself looks
1301 * somewhat bogus. Suggestions for better replacement?
1302 * Ho-hum... In principle, we might treat that as umount + switch
1303 * to rootfs. GC would eventually take care of the old vfsmount.
1304 * Actually it makes sense, especially if rootfs would contain a
1305 * /reboot - static binary that would close all descriptors and
1306 * call reboot(9). Then init(8) could umount root and exec /reboot.
1307 */
Jan Blunck6ac08c32008-02-14 19:34:38 -08001308 if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 /*
1310 * Special case for "unmounting" root ...
1311 * we just try to remount it readonly.
1312 */
1313 down_write(&sb->s_umount);
Al Viro4aa98cf2009-05-08 13:36:58 -04001314 if (!(sb->s_flags & MS_RDONLY))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 retval = do_remount_sb(sb, MS_RDONLY, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 up_write(&sb->s_umount);
1317 return retval;
1318 }
1319
Ram Pai390c6842005-11-07 17:17:51 -05001320 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001321 br_write_lock(vfsmount_lock);
Al Viro5addc5d2005-11-07 17:15:49 -05001322 event++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323
Al Viroc35038b2008-03-22 00:46:23 -04001324 if (!(flags & MNT_DETACH))
Al Viro692afc32011-11-24 21:15:14 -05001325 shrink_submounts(real_mount(mnt), &umount_list);
Al Viroc35038b2008-03-22 00:46:23 -04001326
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 retval = -EBUSY;
Ram Paia05964f2005-11-07 17:20:17 -05001328 if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 if (!list_empty(&mnt->mnt_list))
Al Viro761d5c32011-11-24 21:07:43 -05001330 umount_tree(real_mount(mnt), 1, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 retval = 0;
1332 }
Nick Piggin99b7db72010-08-18 04:37:39 +10001333 br_write_unlock(vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05001334 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001335 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 return retval;
1337}
1338
1339/*
1340 * Now umount can handle mount points as well as block devices.
1341 * This is important for filesystems which use unnamed block devices.
1342 *
1343 * We now support a flag for forced unmount like the other 'big iron'
1344 * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD
1345 */
1346
Heiko Carstensbdc480e2009-01-14 14:14:12 +01001347SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348{
Al Viro2d8f3032008-07-22 09:59:21 -04001349 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 int retval;
Miklos Szeredidb1f05b2010-02-10 12:15:53 +01001351 int lookup_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352
Miklos Szeredidb1f05b2010-02-10 12:15:53 +01001353 if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW))
1354 return -EINVAL;
1355
1356 if (!(flags & UMOUNT_NOFOLLOW))
1357 lookup_flags |= LOOKUP_FOLLOW;
1358
1359 retval = user_path_at(AT_FDCWD, name, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 if (retval)
1361 goto out;
1362 retval = -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -04001363 if (path.dentry != path.mnt->mnt_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 goto dput_and_out;
Al Viro2d8f3032008-07-22 09:59:21 -04001365 if (!check_mnt(path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366 goto dput_and_out;
1367
1368 retval = -EPERM;
1369 if (!capable(CAP_SYS_ADMIN))
1370 goto dput_and_out;
1371
Al Viro2d8f3032008-07-22 09:59:21 -04001372 retval = do_umount(path.mnt, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373dput_and_out:
Jan Blunck429731b2008-02-14 19:34:31 -08001374 /* we mustn't call path_put() as that would clear mnt_expiry_mark */
Al Viro2d8f3032008-07-22 09:59:21 -04001375 dput(path.dentry);
1376 mntput_no_expire(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377out:
1378 return retval;
1379}
1380
1381#ifdef __ARCH_WANT_SYS_OLDUMOUNT
1382
1383/*
Ram Paib58fed82005-11-07 17:16:09 -05001384 * The 2.0 compatible umount. No flags.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 */
Heiko Carstensbdc480e2009-01-14 14:14:12 +01001386SYSCALL_DEFINE1(oldumount, char __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387{
Ram Paib58fed82005-11-07 17:16:09 -05001388 return sys_umount(name, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389}
1390
1391#endif
1392
Al Viro2d92ab32008-08-02 00:51:11 -04001393static int mount_is_safe(struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394{
1395 if (capable(CAP_SYS_ADMIN))
1396 return 0;
1397 return -EPERM;
1398#ifdef notyet
Al Viro2d92ab32008-08-02 00:51:11 -04001399 if (S_ISLNK(path->dentry->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 return -EPERM;
Al Viro2d92ab32008-08-02 00:51:11 -04001401 if (path->dentry->d_inode->i_mode & S_ISVTX) {
David Howellsda9592e2008-11-14 10:39:05 +11001402 if (current_uid() != path->dentry->d_inode->i_uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 return -EPERM;
1404 }
Al Viro2d92ab32008-08-02 00:51:11 -04001405 if (inode_permission(path->dentry->d_inode, MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 return -EPERM;
1407 return 0;
1408#endif
1409}
1410
Al Viro87129cc2011-11-24 21:24:27 -05001411struct mount *copy_tree(struct mount *mnt, struct dentry *dentry,
Ram Pai36341f62005-11-07 17:17:22 -05001412 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413{
Al Viro87129cc2011-11-24 21:24:27 -05001414 struct mount *res, *p, *q;
1415 struct vfsmount *r;
Al Viro1a390682008-03-21 20:48:19 -04001416 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417
Al Viro87129cc2011-11-24 21:24:27 -05001418 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(&mnt->mnt))
Ram Pai9676f0c2005-11-07 17:21:20 -05001419 return NULL;
1420
Ram Pai36341f62005-11-07 17:17:22 -05001421 res = q = clone_mnt(mnt, dentry, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 if (!q)
1423 goto Enomem;
Al Viro87129cc2011-11-24 21:24:27 -05001424 q->mnt.mnt_mountpoint = mnt->mnt.mnt_mountpoint;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
1426 p = mnt;
Al Viro87129cc2011-11-24 21:24:27 -05001427 list_for_each_entry(r, &mnt->mnt.mnt_mounts, mnt_child) {
Al Viro315fc832011-11-24 18:57:30 -05001428 struct mount *s;
Jan Blunck7ec02ef2008-04-29 00:59:40 -07001429 if (!is_subdir(r->mnt_mountpoint, dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 continue;
1431
Al Viro315fc832011-11-24 18:57:30 -05001432 for (s = real_mount(r); s; s = next_mnt(s, r)) {
1433 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(&s->mnt)) {
Ram Pai9676f0c2005-11-07 17:21:20 -05001434 s = skip_mnt_tree(s);
1435 continue;
1436 }
Al Viro87129cc2011-11-24 21:24:27 -05001437 while (p != real_mount(s->mnt.mnt_parent)) {
1438 p = real_mount(p->mnt.mnt_parent);
Al Virocb338d02011-11-24 20:55:08 -05001439 q = real_mount(q->mnt.mnt_parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 }
Al Viro87129cc2011-11-24 21:24:27 -05001441 p = s;
Al Virocb338d02011-11-24 20:55:08 -05001442 path.mnt = &q->mnt;
Al Viro87129cc2011-11-24 21:24:27 -05001443 path.dentry = p->mnt.mnt_mountpoint;
1444 q = clone_mnt(p, p->mnt.mnt_root, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 if (!q)
1446 goto Enomem;
Nick Piggin99b7db72010-08-18 04:37:39 +10001447 br_write_lock(vfsmount_lock);
Al Virocb338d02011-11-24 20:55:08 -05001448 list_add_tail(&q->mnt.mnt_list, &res->mnt.mnt_list);
1449 attach_mnt(q, &path);
Nick Piggin99b7db72010-08-18 04:37:39 +10001450 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451 }
1452 }
1453 return res;
Ram Paib58fed82005-11-07 17:16:09 -05001454Enomem:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 if (res) {
Ram Pai70fbcdf2005-11-07 17:17:04 -05001456 LIST_HEAD(umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001457 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05001458 umount_tree(res, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001459 br_write_unlock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001460 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461 }
1462 return NULL;
1463}
1464
Al Viro589ff872009-04-18 03:28:19 -04001465struct vfsmount *collect_mounts(struct path *path)
Al Viro8aec0802007-06-07 12:20:32 -04001466{
Al Virocb338d02011-11-24 20:55:08 -05001467 struct mount *tree;
Al Viro1a60a282008-03-22 16:19:49 -04001468 down_write(&namespace_sem);
Al Viro87129cc2011-11-24 21:24:27 -05001469 tree = copy_tree(real_mount(path->mnt), path->dentry,
1470 CL_COPY_ALL | CL_PRIVATE);
Al Viro1a60a282008-03-22 16:19:49 -04001471 up_write(&namespace_sem);
Al Virocb338d02011-11-24 20:55:08 -05001472 return tree ? &tree->mnt : NULL;
Al Viro8aec0802007-06-07 12:20:32 -04001473}
1474
1475void drop_collected_mounts(struct vfsmount *mnt)
1476{
1477 LIST_HEAD(umount_list);
Al Viro1a60a282008-03-22 16:19:49 -04001478 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001479 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05001480 umount_tree(real_mount(mnt), 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001481 br_write_unlock(vfsmount_lock);
Al Viro1a60a282008-03-22 16:19:49 -04001482 up_write(&namespace_sem);
Al Viro8aec0802007-06-07 12:20:32 -04001483 release_mounts(&umount_list);
1484}
1485
Al Viro1f707132010-01-30 22:51:25 -05001486int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
1487 struct vfsmount *root)
1488{
1489 struct vfsmount *mnt;
1490 int res = f(root, arg);
1491 if (res)
1492 return res;
1493 list_for_each_entry(mnt, &root->mnt_list, mnt_list) {
1494 res = f(mnt, arg);
1495 if (res)
1496 return res;
1497 }
1498 return 0;
1499}
1500
Al Viro4b8b21f2011-11-24 19:54:23 -05001501static void cleanup_group_ids(struct mount *mnt, struct mount *end)
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001502{
Al Viro315fc832011-11-24 18:57:30 -05001503 struct mount *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001504
Al Viro4b8b21f2011-11-24 19:54:23 -05001505 for (p = mnt; p != end; p = next_mnt(p, &mnt->mnt)) {
Al Viro315fc832011-11-24 18:57:30 -05001506 if (p->mnt.mnt_group_id && !IS_MNT_SHARED(&p->mnt))
Al Viro4b8b21f2011-11-24 19:54:23 -05001507 mnt_release_group_id(p);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001508 }
1509}
1510
Al Viro4b8b21f2011-11-24 19:54:23 -05001511static int invent_group_ids(struct mount *mnt, bool recurse)
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001512{
Al Viro315fc832011-11-24 18:57:30 -05001513 struct mount *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001514
Al Viro4b8b21f2011-11-24 19:54:23 -05001515 for (p = mnt; p; p = recurse ? next_mnt(p, &mnt->mnt) : NULL) {
Al Viro315fc832011-11-24 18:57:30 -05001516 if (!p->mnt.mnt_group_id && !IS_MNT_SHARED(&p->mnt)) {
Al Viro4b8b21f2011-11-24 19:54:23 -05001517 int err = mnt_alloc_group_id(p);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001518 if (err) {
Al Viro4b8b21f2011-11-24 19:54:23 -05001519 cleanup_group_ids(mnt, p);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001520 return err;
1521 }
1522 }
1523 }
1524
1525 return 0;
1526}
1527
Ram Paib90fa9a2005-11-07 17:19:50 -05001528/*
1529 * @source_mnt : mount tree to be attached
Ram Pai21444402005-11-07 17:20:03 -05001530 * @nd : place the mount tree @source_mnt is attached
1531 * @parent_nd : if non-null, detach the source_mnt from its parent and
1532 * store the parent mount and mountpoint dentry.
1533 * (done when source_mnt is moved)
Ram Paib90fa9a2005-11-07 17:19:50 -05001534 *
1535 * NOTE: in the table below explains the semantics when a source mount
1536 * of a given type is attached to a destination mount of a given type.
Ram Pai9676f0c2005-11-07 17:21:20 -05001537 * ---------------------------------------------------------------------------
1538 * | BIND MOUNT OPERATION |
1539 * |**************************************************************************
1540 * | source-->| shared | private | slave | unbindable |
1541 * | dest | | | | |
1542 * | | | | | | |
1543 * | v | | | | |
1544 * |**************************************************************************
1545 * | shared | shared (++) | shared (+) | shared(+++)| invalid |
1546 * | | | | | |
1547 * |non-shared| shared (+) | private | slave (*) | invalid |
1548 * ***************************************************************************
Ram Paib90fa9a2005-11-07 17:19:50 -05001549 * A bind operation clones the source mount and mounts the clone on the
1550 * destination mount.
1551 *
1552 * (++) the cloned mount is propagated to all the mounts in the propagation
1553 * tree of the destination mount and the cloned mount is added to
1554 * the peer group of the source mount.
1555 * (+) the cloned mount is created under the destination mount and is marked
1556 * as shared. The cloned mount is added to the peer group of the source
1557 * mount.
Ram Pai5afe0022005-11-07 17:21:01 -05001558 * (+++) the mount is propagated to all the mounts in the propagation tree
1559 * of the destination mount and the cloned mount is made slave
1560 * of the same master as that of the source mount. The cloned mount
1561 * is marked as 'shared and slave'.
1562 * (*) the cloned mount is made a slave of the same master as that of the
1563 * source mount.
1564 *
Ram Pai9676f0c2005-11-07 17:21:20 -05001565 * ---------------------------------------------------------------------------
1566 * | MOVE MOUNT OPERATION |
1567 * |**************************************************************************
1568 * | source-->| shared | private | slave | unbindable |
1569 * | dest | | | | |
1570 * | | | | | | |
1571 * | v | | | | |
1572 * |**************************************************************************
1573 * | shared | shared (+) | shared (+) | shared(+++) | invalid |
1574 * | | | | | |
1575 * |non-shared| shared (+*) | private | slave (*) | unbindable |
1576 * ***************************************************************************
Ram Pai5afe0022005-11-07 17:21:01 -05001577 *
1578 * (+) the mount is moved to the destination. And is then propagated to
1579 * all the mounts in the propagation tree of the destination mount.
Ram Pai21444402005-11-07 17:20:03 -05001580 * (+*) the mount is moved to the destination.
Ram Pai5afe0022005-11-07 17:21:01 -05001581 * (+++) the mount is moved to the destination and is then propagated to
1582 * all the mounts belonging to the destination mount's propagation tree.
1583 * the mount is marked as 'shared and slave'.
1584 * (*) the mount continues to be a slave at the new location.
Ram Paib90fa9a2005-11-07 17:19:50 -05001585 *
1586 * if the source mount is a tree, the operations explained above is
1587 * applied to each mount in the tree.
1588 * Must be called without spinlocks held, since this function can sleep
1589 * in allocations.
1590 */
Al Viro0fb54e52011-11-24 19:59:16 -05001591static int attach_recursive_mnt(struct mount *source_mnt,
Al Viro1a390682008-03-21 20:48:19 -04001592 struct path *path, struct path *parent_path)
Ram Paib90fa9a2005-11-07 17:19:50 -05001593{
1594 LIST_HEAD(tree_list);
Al Viro1a390682008-03-21 20:48:19 -04001595 struct vfsmount *dest_mnt = path->mnt;
1596 struct dentry *dest_dentry = path->dentry;
Al Viro315fc832011-11-24 18:57:30 -05001597 struct mount *child, *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001598 int err;
Ram Paib90fa9a2005-11-07 17:19:50 -05001599
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001600 if (IS_MNT_SHARED(dest_mnt)) {
Al Viro0fb54e52011-11-24 19:59:16 -05001601 err = invent_group_ids(source_mnt, true);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001602 if (err)
1603 goto out;
1604 }
Al Viro0fb54e52011-11-24 19:59:16 -05001605 err = propagate_mnt(dest_mnt, dest_dentry, &source_mnt->mnt, &tree_list);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001606 if (err)
1607 goto out_cleanup_ids;
Ram Paib90fa9a2005-11-07 17:19:50 -05001608
Nick Piggin99b7db72010-08-18 04:37:39 +10001609 br_write_lock(vfsmount_lock);
Al Virodf1a1ad2010-01-16 12:57:40 -05001610
Ram Paib90fa9a2005-11-07 17:19:50 -05001611 if (IS_MNT_SHARED(dest_mnt)) {
Al Viro0fb54e52011-11-24 19:59:16 -05001612 for (p = source_mnt; p; p = next_mnt(p, &source_mnt->mnt))
Al Viro0f0afb12011-11-24 20:43:10 -05001613 set_mnt_shared(p);
Ram Paib90fa9a2005-11-07 17:19:50 -05001614 }
Al Viro1a390682008-03-21 20:48:19 -04001615 if (parent_path) {
Al Viro0fb54e52011-11-24 19:59:16 -05001616 detach_mnt(source_mnt, parent_path);
1617 attach_mnt(source_mnt, path);
Al Viroe5d67f02009-04-07 12:15:39 -04001618 touch_mnt_namespace(parent_path->mnt->mnt_ns);
Ram Pai21444402005-11-07 17:20:03 -05001619 } else {
Al Viro44d964d62011-11-24 21:28:22 -05001620 mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt);
Al Viro0fb54e52011-11-24 19:59:16 -05001621 commit_tree(source_mnt);
Ram Pai21444402005-11-07 17:20:03 -05001622 }
Ram Paib90fa9a2005-11-07 17:19:50 -05001623
Al Viro1b8e5562011-11-24 21:01:32 -05001624 list_for_each_entry_safe(child, p, &tree_list, mnt_hash) {
1625 list_del_init(&child->mnt_hash);
Al Viro4b2619a2011-11-24 19:47:15 -05001626 commit_tree(child);
Ram Paib90fa9a2005-11-07 17:19:50 -05001627 }
Nick Piggin99b7db72010-08-18 04:37:39 +10001628 br_write_unlock(vfsmount_lock);
1629
Ram Paib90fa9a2005-11-07 17:19:50 -05001630 return 0;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001631
1632 out_cleanup_ids:
1633 if (IS_MNT_SHARED(dest_mnt))
Al Viro0fb54e52011-11-24 19:59:16 -05001634 cleanup_group_ids(source_mnt, NULL);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001635 out:
1636 return err;
Ram Paib90fa9a2005-11-07 17:19:50 -05001637}
1638
Al Virob12cea92011-03-18 08:55:38 -04001639static int lock_mount(struct path *path)
1640{
1641 struct vfsmount *mnt;
1642retry:
1643 mutex_lock(&path->dentry->d_inode->i_mutex);
1644 if (unlikely(cant_mount(path->dentry))) {
1645 mutex_unlock(&path->dentry->d_inode->i_mutex);
1646 return -ENOENT;
1647 }
1648 down_write(&namespace_sem);
1649 mnt = lookup_mnt(path);
1650 if (likely(!mnt))
1651 return 0;
1652 up_write(&namespace_sem);
1653 mutex_unlock(&path->dentry->d_inode->i_mutex);
1654 path_put(path);
1655 path->mnt = mnt;
1656 path->dentry = dget(mnt->mnt_root);
1657 goto retry;
1658}
1659
1660static void unlock_mount(struct path *path)
1661{
1662 up_write(&namespace_sem);
1663 mutex_unlock(&path->dentry->d_inode->i_mutex);
1664}
1665
Al Viro8c3ee422008-03-22 18:00:39 -04001666static int graft_tree(struct vfsmount *mnt, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 if (mnt->mnt_sb->s_flags & MS_NOUSER)
1669 return -EINVAL;
1670
Al Viro8c3ee422008-03-22 18:00:39 -04001671 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 S_ISDIR(mnt->mnt_root->d_inode->i_mode))
1673 return -ENOTDIR;
1674
Al Virob12cea92011-03-18 08:55:38 -04001675 if (d_unlinked(path->dentry))
1676 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Al Viro0fb54e52011-11-24 19:59:16 -05001678 return attach_recursive_mnt(real_mount(mnt), path, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679}
1680
1681/*
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001682 * Sanity check the flags to change_mnt_propagation.
1683 */
1684
1685static int flags_to_propagation_type(int flags)
1686{
Roman Borisov7c6e9842011-05-25 16:26:48 -07001687 int type = flags & ~(MS_REC | MS_SILENT);
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001688
1689 /* Fail if any non-propagation flags are set */
1690 if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
1691 return 0;
1692 /* Only one propagation flag should be set */
1693 if (!is_power_of_2(type))
1694 return 0;
1695 return type;
1696}
1697
1698/*
Ram Pai07b20882005-11-07 17:19:07 -05001699 * recursively change the type of the mountpoint.
1700 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001701static int do_change_type(struct path *path, int flag)
Ram Pai07b20882005-11-07 17:19:07 -05001702{
Al Viro315fc832011-11-24 18:57:30 -05001703 struct mount *m;
Al Viro4b8b21f2011-11-24 19:54:23 -05001704 struct mount *mnt = real_mount(path->mnt);
Ram Pai07b20882005-11-07 17:19:07 -05001705 int recurse = flag & MS_REC;
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001706 int type;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001707 int err = 0;
Ram Pai07b20882005-11-07 17:19:07 -05001708
Miklos Szerediee6f9582007-05-08 00:30:40 -07001709 if (!capable(CAP_SYS_ADMIN))
1710 return -EPERM;
1711
Al Viro2d92ab32008-08-02 00:51:11 -04001712 if (path->dentry != path->mnt->mnt_root)
Ram Pai07b20882005-11-07 17:19:07 -05001713 return -EINVAL;
1714
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001715 type = flags_to_propagation_type(flag);
1716 if (!type)
1717 return -EINVAL;
1718
Ram Pai07b20882005-11-07 17:19:07 -05001719 down_write(&namespace_sem);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001720 if (type == MS_SHARED) {
1721 err = invent_group_ids(mnt, recurse);
1722 if (err)
1723 goto out_unlock;
1724 }
1725
Nick Piggin99b7db72010-08-18 04:37:39 +10001726 br_write_lock(vfsmount_lock);
Al Viro4b8b21f2011-11-24 19:54:23 -05001727 for (m = mnt; m; m = (recurse ? next_mnt(m, &mnt->mnt) : NULL))
Al Viro0f0afb12011-11-24 20:43:10 -05001728 change_mnt_propagation(m, type);
Nick Piggin99b7db72010-08-18 04:37:39 +10001729 br_write_unlock(vfsmount_lock);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001730
1731 out_unlock:
Ram Pai07b20882005-11-07 17:19:07 -05001732 up_write(&namespace_sem);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001733 return err;
Ram Pai07b20882005-11-07 17:19:07 -05001734}
1735
1736/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 * do loopback mount.
1738 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001739static int do_loopback(struct path *path, char *old_name,
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001740 int recurse)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741{
Al Virob12cea92011-03-18 08:55:38 -04001742 LIST_HEAD(umount_list);
Al Viro2d92ab32008-08-02 00:51:11 -04001743 struct path old_path;
Al Viro87129cc2011-11-24 21:24:27 -05001744 struct mount *mnt = NULL, *old;
Al Viro2d92ab32008-08-02 00:51:11 -04001745 int err = mount_is_safe(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 if (err)
1747 return err;
1748 if (!old_name || !*old_name)
1749 return -EINVAL;
Trond Myklebust815d4052011-09-26 20:36:09 -04001750 err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 if (err)
1752 return err;
1753
Al Virob12cea92011-03-18 08:55:38 -04001754 err = lock_mount(path);
1755 if (err)
Jan Blunck4ac91372008-02-14 19:34:32 -08001756 goto out;
Ram Pai9676f0c2005-11-07 17:21:20 -05001757
Al Viro87129cc2011-11-24 21:24:27 -05001758 old = real_mount(old_path.mnt);
1759
Al Virob12cea92011-03-18 08:55:38 -04001760 err = -EINVAL;
1761 if (IS_MNT_UNBINDABLE(old_path.mnt))
1762 goto out2;
1763
Al Viro2d92ab32008-08-02 00:51:11 -04001764 if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt))
Al Virob12cea92011-03-18 08:55:38 -04001765 goto out2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766
Al Viroccd48bc2005-11-07 17:15:04 -05001767 err = -ENOMEM;
1768 if (recurse)
Al Viro87129cc2011-11-24 21:24:27 -05001769 mnt = copy_tree(old, old_path.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -05001770 else
Al Viro87129cc2011-11-24 21:24:27 -05001771 mnt = clone_mnt(old, old_path.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -05001772
1773 if (!mnt)
Al Virob12cea92011-03-18 08:55:38 -04001774 goto out2;
Al Viroccd48bc2005-11-07 17:15:04 -05001775
Al Virocb338d02011-11-24 20:55:08 -05001776 err = graft_tree(&mnt->mnt, path);
Al Viroccd48bc2005-11-07 17:15:04 -05001777 if (err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001778 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05001779 umount_tree(mnt, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001780 br_write_unlock(vfsmount_lock);
Ram Pai5b83d2c5c2005-11-07 17:16:29 -05001781 }
Al Virob12cea92011-03-18 08:55:38 -04001782out2:
1783 unlock_mount(path);
1784 release_mounts(&umount_list);
Al Viroccd48bc2005-11-07 17:15:04 -05001785out:
Al Viro2d92ab32008-08-02 00:51:11 -04001786 path_put(&old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 return err;
1788}
1789
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001790static int change_mount_flags(struct vfsmount *mnt, int ms_flags)
1791{
1792 int error = 0;
1793 int readonly_request = 0;
1794
1795 if (ms_flags & MS_RDONLY)
1796 readonly_request = 1;
1797 if (readonly_request == __mnt_is_readonly(mnt))
1798 return 0;
1799
1800 if (readonly_request)
1801 error = mnt_make_readonly(mnt);
1802 else
1803 __mnt_unmake_readonly(mnt);
1804 return error;
1805}
1806
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807/*
1808 * change filesystem flags. dir should be a physical root of filesystem.
1809 * If you've mounted a non-root directory somewhere and want to do remount
1810 * on it - tough luck.
1811 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001812static int do_remount(struct path *path, int flags, int mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813 void *data)
1814{
1815 int err;
Al Viro2d92ab32008-08-02 00:51:11 -04001816 struct super_block *sb = path->mnt->mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817
1818 if (!capable(CAP_SYS_ADMIN))
1819 return -EPERM;
1820
Al Viro2d92ab32008-08-02 00:51:11 -04001821 if (!check_mnt(path->mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822 return -EINVAL;
1823
Al Viro2d92ab32008-08-02 00:51:11 -04001824 if (path->dentry != path->mnt->mnt_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 return -EINVAL;
1826
Eric Parisff36fe22011-03-03 16:09:14 -05001827 err = security_sb_remount(sb, data);
1828 if (err)
1829 return err;
1830
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 down_write(&sb->s_umount);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001832 if (flags & MS_BIND)
Al Viro2d92ab32008-08-02 00:51:11 -04001833 err = change_mount_flags(path->mnt, flags);
Al Viro4aa98cf2009-05-08 13:36:58 -04001834 else
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001835 err = do_remount_sb(sb, flags, data, 0);
Al Viro7b43a792010-01-16 13:01:26 -05001836 if (!err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001837 br_write_lock(vfsmount_lock);
Valerie Aurora495d6c92010-01-26 14:20:47 -05001838 mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK;
Al Viro2d92ab32008-08-02 00:51:11 -04001839 path->mnt->mnt_flags = mnt_flags;
Nick Piggin99b7db72010-08-18 04:37:39 +10001840 br_write_unlock(vfsmount_lock);
Al Viro7b43a792010-01-16 13:01:26 -05001841 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842 up_write(&sb->s_umount);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001843 if (!err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001844 br_write_lock(vfsmount_lock);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001845 touch_mnt_namespace(path->mnt->mnt_ns);
Nick Piggin99b7db72010-08-18 04:37:39 +10001846 br_write_unlock(vfsmount_lock);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001847 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 return err;
1849}
1850
Al Virocbbe3622011-11-24 20:01:19 -05001851static inline int tree_contains_unbindable(struct mount *mnt)
Ram Pai9676f0c2005-11-07 17:21:20 -05001852{
Al Viro315fc832011-11-24 18:57:30 -05001853 struct mount *p;
Al Virocbbe3622011-11-24 20:01:19 -05001854 for (p = mnt; p; p = next_mnt(p, &mnt->mnt)) {
Al Viro315fc832011-11-24 18:57:30 -05001855 if (IS_MNT_UNBINDABLE(&p->mnt))
Ram Pai9676f0c2005-11-07 17:21:20 -05001856 return 1;
1857 }
1858 return 0;
1859}
1860
Al Viro0a0d8a42008-08-02 00:55:27 -04001861static int do_move_mount(struct path *path, char *old_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862{
Al Viro2d92ab32008-08-02 00:51:11 -04001863 struct path old_path, parent_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864 struct vfsmount *p;
Al Viro0fb54e52011-11-24 19:59:16 -05001865 struct mount *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 int err = 0;
1867 if (!capable(CAP_SYS_ADMIN))
1868 return -EPERM;
1869 if (!old_name || !*old_name)
1870 return -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001871 err = kern_path(old_name, LOOKUP_FOLLOW, &old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 if (err)
1873 return err;
1874
Al Virob12cea92011-03-18 08:55:38 -04001875 err = lock_mount(path);
David Howellscc53ce52011-01-14 18:45:26 +00001876 if (err < 0)
1877 goto out;
1878
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 err = -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001880 if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 goto out1;
1882
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04001883 if (d_unlinked(path->dentry))
Ram Pai21444402005-11-07 17:20:03 -05001884 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885
1886 err = -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001887 if (old_path.dentry != old_path.mnt->mnt_root)
Ram Pai21444402005-11-07 17:20:03 -05001888 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889
Al Viro0fb54e52011-11-24 19:59:16 -05001890 old = real_mount(old_path.mnt);
1891
Al Virob2dba1a2011-11-23 19:26:23 -05001892 if (!mnt_has_parent(old_path.mnt))
Ram Pai21444402005-11-07 17:20:03 -05001893 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894
Al Viro2d92ab32008-08-02 00:51:11 -04001895 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
1896 S_ISDIR(old_path.dentry->d_inode->i_mode))
Ram Pai21444402005-11-07 17:20:03 -05001897 goto out1;
1898 /*
1899 * Don't move a mount residing in a shared parent.
1900 */
Al Viroafac7cb2011-11-23 19:34:49 -05001901 if (IS_MNT_SHARED(old_path.mnt->mnt_parent))
Ram Pai21444402005-11-07 17:20:03 -05001902 goto out1;
Ram Pai9676f0c2005-11-07 17:21:20 -05001903 /*
1904 * Don't move a mount tree containing unbindable mounts to a destination
1905 * mount which is shared.
1906 */
Al Viro2d92ab32008-08-02 00:51:11 -04001907 if (IS_MNT_SHARED(path->mnt) &&
Al Virocbbe3622011-11-24 20:01:19 -05001908 tree_contains_unbindable(old))
Ram Pai9676f0c2005-11-07 17:21:20 -05001909 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910 err = -ELOOP;
Al Virob2dba1a2011-11-23 19:26:23 -05001911 for (p = path->mnt; mnt_has_parent(p); p = p->mnt_parent)
Al Viro2d92ab32008-08-02 00:51:11 -04001912 if (p == old_path.mnt)
Ram Pai21444402005-11-07 17:20:03 -05001913 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914
Al Viro0fb54e52011-11-24 19:59:16 -05001915 err = attach_recursive_mnt(old, path, &parent_path);
Jan Blunck4ac91372008-02-14 19:34:32 -08001916 if (err)
Ram Pai21444402005-11-07 17:20:03 -05001917 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918
1919 /* if the mount is moved, it should no longer be expire
1920 * automatically */
Al Viro2d92ab32008-08-02 00:51:11 -04001921 list_del_init(&old_path.mnt->mnt_expire);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922out1:
Al Virob12cea92011-03-18 08:55:38 -04001923 unlock_mount(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925 if (!err)
Al Viro1a390682008-03-21 20:48:19 -04001926 path_put(&parent_path);
Al Viro2d92ab32008-08-02 00:51:11 -04001927 path_put(&old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928 return err;
1929}
1930
Al Viro9d412a42011-03-17 22:08:28 -04001931static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
1932{
1933 int err;
1934 const char *subtype = strchr(fstype, '.');
1935 if (subtype) {
1936 subtype++;
1937 err = -EINVAL;
1938 if (!subtype[0])
1939 goto err;
1940 } else
1941 subtype = "";
1942
1943 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1944 err = -ENOMEM;
1945 if (!mnt->mnt_sb->s_subtype)
1946 goto err;
1947 return mnt;
1948
1949 err:
1950 mntput(mnt);
1951 return ERR_PTR(err);
1952}
1953
Al Viro79e801a2011-12-12 23:07:05 -05001954static struct vfsmount *
Al Viro9d412a42011-03-17 22:08:28 -04001955do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1956{
1957 struct file_system_type *type = get_fs_type(fstype);
1958 struct vfsmount *mnt;
1959 if (!type)
1960 return ERR_PTR(-ENODEV);
1961 mnt = vfs_kern_mount(type, flags, name, data);
1962 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1963 !mnt->mnt_sb->s_subtype)
1964 mnt = fs_set_subtype(mnt, fstype);
1965 put_filesystem(type);
1966 return mnt;
1967}
Al Viro9d412a42011-03-17 22:08:28 -04001968
1969/*
1970 * add a mount into a namespace's mount tree
1971 */
1972static int do_add_mount(struct vfsmount *newmnt, struct path *path, int mnt_flags)
1973{
1974 int err;
1975
1976 mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL);
1977
Al Virob12cea92011-03-18 08:55:38 -04001978 err = lock_mount(path);
1979 if (err)
1980 return err;
Al Viro9d412a42011-03-17 22:08:28 -04001981
1982 err = -EINVAL;
1983 if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt))
1984 goto unlock;
1985
1986 /* Refuse the same filesystem on the same mount point */
1987 err = -EBUSY;
1988 if (path->mnt->mnt_sb == newmnt->mnt_sb &&
1989 path->mnt->mnt_root == path->dentry)
1990 goto unlock;
1991
1992 err = -EINVAL;
1993 if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode))
1994 goto unlock;
1995
1996 newmnt->mnt_flags = mnt_flags;
1997 err = graft_tree(newmnt, path);
1998
1999unlock:
Al Virob12cea92011-03-18 08:55:38 -04002000 unlock_mount(path);
Al Viro9d412a42011-03-17 22:08:28 -04002001 return err;
2002}
Al Virob1e75df2011-01-17 01:47:59 -05002003
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004/*
2005 * create a new mount for userspace and request it to be added into the
2006 * namespace's tree
2007 */
Al Viro0a0d8a42008-08-02 00:55:27 -04002008static int do_new_mount(struct path *path, char *type, int flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 int mnt_flags, char *name, void *data)
2010{
2011 struct vfsmount *mnt;
Al Viro15f9a3f2011-01-17 01:41:58 -05002012 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013
Vegard Nossumeca6f532009-09-18 13:05:45 -07002014 if (!type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 return -EINVAL;
2016
2017 /* we need capabilities... */
2018 if (!capable(CAP_SYS_ADMIN))
2019 return -EPERM;
2020
2021 mnt = do_kern_mount(type, flags, name, data);
2022 if (IS_ERR(mnt))
2023 return PTR_ERR(mnt);
2024
Al Viro15f9a3f2011-01-17 01:41:58 -05002025 err = do_add_mount(mnt, path, mnt_flags);
2026 if (err)
2027 mntput(mnt);
2028 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029}
2030
Al Viro19a167a2011-01-17 01:35:23 -05002031int finish_automount(struct vfsmount *m, struct path *path)
2032{
2033 int err;
2034 /* The new mount record should have at least 2 refs to prevent it being
2035 * expired before we get a chance to add it
2036 */
2037 BUG_ON(mnt_get_count(m) < 2);
2038
2039 if (m->mnt_sb == path->mnt->mnt_sb &&
2040 m->mnt_root == path->dentry) {
Al Virob1e75df2011-01-17 01:47:59 -05002041 err = -ELOOP;
2042 goto fail;
Al Viro19a167a2011-01-17 01:35:23 -05002043 }
2044
Al Viro19a167a2011-01-17 01:35:23 -05002045 err = do_add_mount(m, path, path->mnt->mnt_flags | MNT_SHRINKABLE);
Al Virob1e75df2011-01-17 01:47:59 -05002046 if (!err)
2047 return 0;
2048fail:
2049 /* remove m from any expiration list it may be on */
2050 if (!list_empty(&m->mnt_expire)) {
2051 down_write(&namespace_sem);
2052 br_write_lock(vfsmount_lock);
2053 list_del_init(&m->mnt_expire);
2054 br_write_unlock(vfsmount_lock);
2055 up_write(&namespace_sem);
Al Viro19a167a2011-01-17 01:35:23 -05002056 }
Al Virob1e75df2011-01-17 01:47:59 -05002057 mntput(m);
2058 mntput(m);
Al Viro19a167a2011-01-17 01:35:23 -05002059 return err;
2060}
2061
David Howellsea5b7782011-01-14 19:10:03 +00002062/**
2063 * mnt_set_expiry - Put a mount on an expiration list
2064 * @mnt: The mount to list.
2065 * @expiry_list: The list to add the mount to.
2066 */
2067void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list)
2068{
2069 down_write(&namespace_sem);
2070 br_write_lock(vfsmount_lock);
2071
2072 list_add_tail(&mnt->mnt_expire, expiry_list);
2073
2074 br_write_unlock(vfsmount_lock);
2075 up_write(&namespace_sem);
2076}
2077EXPORT_SYMBOL(mnt_set_expiry);
2078
2079/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080 * process a list of expirable mountpoints with the intent of discarding any
2081 * mountpoints that aren't in use and haven't been touched since last we came
2082 * here
2083 */
2084void mark_mounts_for_expiry(struct list_head *mounts)
2085{
Al Viro761d5c32011-11-24 21:07:43 -05002086 struct mount *mnt, *next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087 LIST_HEAD(graveyard);
Al Virobcc5c7d2008-03-22 00:21:53 -04002088 LIST_HEAD(umounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089
2090 if (list_empty(mounts))
2091 return;
2092
Al Virobcc5c7d2008-03-22 00:21:53 -04002093 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10002094 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095
2096 /* extract from the expiration list every vfsmount that matches the
2097 * following criteria:
2098 * - only referenced by its parent vfsmount
2099 * - still marked for expiry (marked on the last call here; marks are
2100 * cleared by mntput())
2101 */
Al Viro761d5c32011-11-24 21:07:43 -05002102 list_for_each_entry_safe(mnt, next, mounts, mnt.mnt_expire) {
2103 if (!xchg(&mnt->mnt.mnt_expiry_mark, 1) ||
2104 propagate_mount_busy(&mnt->mnt, 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105 continue;
Al Viro761d5c32011-11-24 21:07:43 -05002106 list_move(&mnt->mnt.mnt_expire, &graveyard);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 }
Al Virobcc5c7d2008-03-22 00:21:53 -04002108 while (!list_empty(&graveyard)) {
Al Viro761d5c32011-11-24 21:07:43 -05002109 mnt = list_first_entry(&graveyard, struct mount, mnt.mnt_expire);
2110 touch_mnt_namespace(mnt->mnt.mnt_ns);
Al Virobcc5c7d2008-03-22 00:21:53 -04002111 umount_tree(mnt, 1, &umounts);
2112 }
Nick Piggin99b7db72010-08-18 04:37:39 +10002113 br_write_unlock(vfsmount_lock);
Al Virobcc5c7d2008-03-22 00:21:53 -04002114 up_write(&namespace_sem);
2115
2116 release_mounts(&umounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117}
2118
2119EXPORT_SYMBOL_GPL(mark_mounts_for_expiry);
2120
2121/*
Trond Myklebust5528f9112006-06-09 09:34:17 -04002122 * Ripoff of 'select_parent()'
2123 *
2124 * search the list of submounts for a given mountpoint, and move any
2125 * shrinkable submounts to the 'graveyard' list.
2126 */
Al Viro692afc32011-11-24 21:15:14 -05002127static int select_submounts(struct mount *parent, struct list_head *graveyard)
Trond Myklebust5528f9112006-06-09 09:34:17 -04002128{
Al Viro692afc32011-11-24 21:15:14 -05002129 struct mount *this_parent = parent;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002130 struct list_head *next;
2131 int found = 0;
2132
2133repeat:
Al Viro692afc32011-11-24 21:15:14 -05002134 next = this_parent->mnt.mnt_mounts.next;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002135resume:
Al Viro692afc32011-11-24 21:15:14 -05002136 while (next != &this_parent->mnt.mnt_mounts) {
Trond Myklebust5528f9112006-06-09 09:34:17 -04002137 struct list_head *tmp = next;
Al Viro692afc32011-11-24 21:15:14 -05002138 struct mount *mnt = list_entry(tmp, struct mount, mnt.mnt_child);
Trond Myklebust5528f9112006-06-09 09:34:17 -04002139
2140 next = tmp->next;
Al Viro692afc32011-11-24 21:15:14 -05002141 if (!(mnt->mnt.mnt_flags & MNT_SHRINKABLE))
Trond Myklebust5528f9112006-06-09 09:34:17 -04002142 continue;
2143 /*
2144 * Descend a level if the d_mounts list is non-empty.
2145 */
Al Viro692afc32011-11-24 21:15:14 -05002146 if (!list_empty(&mnt->mnt.mnt_mounts)) {
Trond Myklebust5528f9112006-06-09 09:34:17 -04002147 this_parent = mnt;
2148 goto repeat;
2149 }
2150
Al Viro692afc32011-11-24 21:15:14 -05002151 if (!propagate_mount_busy(&mnt->mnt, 1)) {
2152 list_move_tail(&mnt->mnt.mnt_expire, graveyard);
Trond Myklebust5528f9112006-06-09 09:34:17 -04002153 found++;
2154 }
2155 }
2156 /*
2157 * All done at this level ... ascend and resume the search
2158 */
2159 if (this_parent != parent) {
Al Viro692afc32011-11-24 21:15:14 -05002160 next = this_parent->mnt.mnt_child.next;
2161 this_parent = real_mount(this_parent->mnt.mnt_parent);
Trond Myklebust5528f9112006-06-09 09:34:17 -04002162 goto resume;
2163 }
2164 return found;
2165}
2166
2167/*
2168 * process a list of expirable mountpoints with the intent of discarding any
2169 * submounts of a specific parent mountpoint
Nick Piggin99b7db72010-08-18 04:37:39 +10002170 *
2171 * vfsmount_lock must be held for write
Trond Myklebust5528f9112006-06-09 09:34:17 -04002172 */
Al Viro692afc32011-11-24 21:15:14 -05002173static void shrink_submounts(struct mount *mnt, struct list_head *umounts)
Trond Myklebust5528f9112006-06-09 09:34:17 -04002174{
2175 LIST_HEAD(graveyard);
Al Viro761d5c32011-11-24 21:07:43 -05002176 struct mount *m;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002177
Trond Myklebust5528f9112006-06-09 09:34:17 -04002178 /* extract submounts of 'mountpoint' from the expiration list */
Al Viroc35038b2008-03-22 00:46:23 -04002179 while (select_submounts(mnt, &graveyard)) {
Al Virobcc5c7d2008-03-22 00:21:53 -04002180 while (!list_empty(&graveyard)) {
Al Viro761d5c32011-11-24 21:07:43 -05002181 m = list_first_entry(&graveyard, struct mount,
2182 mnt.mnt_expire);
2183 touch_mnt_namespace(m->mnt.mnt_ns);
Eric W. Biedermanafef80b2008-11-12 13:26:54 -08002184 umount_tree(m, 1, umounts);
Al Virobcc5c7d2008-03-22 00:21:53 -04002185 }
2186 }
Trond Myklebust5528f9112006-06-09 09:34:17 -04002187}
2188
Trond Myklebust5528f9112006-06-09 09:34:17 -04002189/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 * Some copy_from_user() implementations do not return the exact number of
2191 * bytes remaining to copy on a fault. But copy_mount_options() requires that.
2192 * Note that this function differs from copy_from_user() in that it will oops
2193 * on bad values of `to', rather than returning a short copy.
2194 */
Ram Paib58fed82005-11-07 17:16:09 -05002195static long exact_copy_from_user(void *to, const void __user * from,
2196 unsigned long n)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197{
2198 char *t = to;
2199 const char __user *f = from;
2200 char c;
2201
2202 if (!access_ok(VERIFY_READ, from, n))
2203 return n;
2204
2205 while (n) {
2206 if (__get_user(c, f)) {
2207 memset(t, 0, n);
2208 break;
2209 }
2210 *t++ = c;
2211 f++;
2212 n--;
2213 }
2214 return n;
2215}
2216
Ram Paib58fed82005-11-07 17:16:09 -05002217int copy_mount_options(const void __user * data, unsigned long *where)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218{
2219 int i;
2220 unsigned long page;
2221 unsigned long size;
Ram Paib58fed82005-11-07 17:16:09 -05002222
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223 *where = 0;
2224 if (!data)
2225 return 0;
2226
2227 if (!(page = __get_free_page(GFP_KERNEL)))
2228 return -ENOMEM;
2229
2230 /* We only care that *some* data at the address the user
2231 * gave us is valid. Just in case, we'll zero
2232 * the remainder of the page.
2233 */
2234 /* copy_from_user cannot cross TASK_SIZE ! */
2235 size = TASK_SIZE - (unsigned long)data;
2236 if (size > PAGE_SIZE)
2237 size = PAGE_SIZE;
2238
2239 i = size - exact_copy_from_user((void *)page, data, size);
2240 if (!i) {
Ram Paib58fed82005-11-07 17:16:09 -05002241 free_page(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 return -EFAULT;
2243 }
2244 if (i != PAGE_SIZE)
2245 memset((char *)page + i, 0, PAGE_SIZE - i);
2246 *where = page;
2247 return 0;
2248}
2249
Vegard Nossumeca6f532009-09-18 13:05:45 -07002250int copy_mount_string(const void __user *data, char **where)
2251{
2252 char *tmp;
2253
2254 if (!data) {
2255 *where = NULL;
2256 return 0;
2257 }
2258
2259 tmp = strndup_user(data, PAGE_SIZE);
2260 if (IS_ERR(tmp))
2261 return PTR_ERR(tmp);
2262
2263 *where = tmp;
2264 return 0;
2265}
2266
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267/*
2268 * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to
2269 * be given to the mount() call (ie: read-only, no-dev, no-suid etc).
2270 *
2271 * data is a (void *) that can point to any structure up to
2272 * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent
2273 * information (or be NULL).
2274 *
2275 * Pre-0.97 versions of mount() didn't have a flags word.
2276 * When the flags word was introduced its top half was required
2277 * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9.
2278 * Therefore, if this magic number is present, it carries no information
2279 * and must be discarded.
2280 */
Ram Paib58fed82005-11-07 17:16:09 -05002281long do_mount(char *dev_name, char *dir_name, char *type_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282 unsigned long flags, void *data_page)
2283{
Al Viro2d92ab32008-08-02 00:51:11 -04002284 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002285 int retval = 0;
2286 int mnt_flags = 0;
2287
2288 /* Discard magic */
2289 if ((flags & MS_MGC_MSK) == MS_MGC_VAL)
2290 flags &= ~MS_MGC_MSK;
2291
2292 /* Basic sanity checks */
2293
2294 if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE))
2295 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296
2297 if (data_page)
2298 ((char *)data_page)[PAGE_SIZE - 1] = 0;
2299
Tetsuo Handaa27ab9f22009-10-04 21:49:49 +09002300 /* ... and get the mountpoint */
2301 retval = kern_path(dir_name, LOOKUP_FOLLOW, &path);
2302 if (retval)
2303 return retval;
2304
2305 retval = security_sb_mount(dev_name, &path,
2306 type_page, flags, data_page);
2307 if (retval)
2308 goto dput_out;
2309
Andi Kleen613cbe32009-04-19 18:40:43 +02002310 /* Default to relatime unless overriden */
2311 if (!(flags & MS_NOATIME))
2312 mnt_flags |= MNT_RELATIME;
Matthew Garrett0a1c01c2009-03-26 17:53:14 +00002313
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314 /* Separate the per-mountpoint flags */
2315 if (flags & MS_NOSUID)
2316 mnt_flags |= MNT_NOSUID;
2317 if (flags & MS_NODEV)
2318 mnt_flags |= MNT_NODEV;
2319 if (flags & MS_NOEXEC)
2320 mnt_flags |= MNT_NOEXEC;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08002321 if (flags & MS_NOATIME)
2322 mnt_flags |= MNT_NOATIME;
2323 if (flags & MS_NODIRATIME)
2324 mnt_flags |= MNT_NODIRATIME;
Matthew Garrettd0adde52009-03-26 17:49:56 +00002325 if (flags & MS_STRICTATIME)
2326 mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08002327 if (flags & MS_RDONLY)
2328 mnt_flags |= MNT_READONLY;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08002329
Al Viro7a4dec52010-08-09 12:05:43 -04002330 flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN |
Matthew Garrettd0adde52009-03-26 17:49:56 +00002331 MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT |
2332 MS_STRICTATIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334 if (flags & MS_REMOUNT)
Al Viro2d92ab32008-08-02 00:51:11 -04002335 retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336 data_page);
2337 else if (flags & MS_BIND)
Al Viro2d92ab32008-08-02 00:51:11 -04002338 retval = do_loopback(&path, dev_name, flags & MS_REC);
Ram Pai9676f0c2005-11-07 17:21:20 -05002339 else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
Al Viro2d92ab32008-08-02 00:51:11 -04002340 retval = do_change_type(&path, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341 else if (flags & MS_MOVE)
Al Viro2d92ab32008-08-02 00:51:11 -04002342 retval = do_move_mount(&path, dev_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 else
Al Viro2d92ab32008-08-02 00:51:11 -04002344 retval = do_new_mount(&path, type_page, flags, mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345 dev_name, data_page);
2346dput_out:
Al Viro2d92ab32008-08-02 00:51:11 -04002347 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 return retval;
2349}
2350
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002351static struct mnt_namespace *alloc_mnt_ns(void)
2352{
2353 struct mnt_namespace *new_ns;
2354
2355 new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL);
2356 if (!new_ns)
2357 return ERR_PTR(-ENOMEM);
2358 atomic_set(&new_ns->count, 1);
2359 new_ns->root = NULL;
2360 INIT_LIST_HEAD(&new_ns->list);
2361 init_waitqueue_head(&new_ns->poll);
2362 new_ns->event = 0;
2363 return new_ns;
2364}
2365
Al Virof03c6592011-01-14 22:30:21 -05002366void mnt_make_longterm(struct vfsmount *mnt)
2367{
Al Viro7e3d0eb02011-01-16 16:32:11 -05002368 __mnt_make_longterm(mnt);
Al Virof03c6592011-01-14 22:30:21 -05002369}
2370
2371void mnt_make_shortterm(struct vfsmount *mnt)
2372{
Al Viro7e3d0eb02011-01-16 16:32:11 -05002373#ifdef CONFIG_SMP
Al Virof03c6592011-01-14 22:30:21 -05002374 if (atomic_add_unless(&mnt->mnt_longterm, -1, 1))
2375 return;
2376 br_write_lock(vfsmount_lock);
2377 atomic_dec(&mnt->mnt_longterm);
2378 br_write_unlock(vfsmount_lock);
Al Viro7e3d0eb02011-01-16 16:32:11 -05002379#endif
Al Virof03c6592011-01-14 22:30:21 -05002380}
2381
JANAK DESAI741a2952006-02-07 12:59:00 -08002382/*
2383 * Allocate a new namespace structure and populate it with contents
2384 * copied from the namespace of the passed in task structure.
2385 */
Badari Pulavartye3222c42007-05-08 00:25:21 -07002386static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002387 struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002389 struct mnt_namespace *new_ns;
Al Viro7f2da1e2008-05-10 20:44:54 -04002390 struct vfsmount *rootmnt = NULL, *pwdmnt = NULL;
Al Viro315fc832011-11-24 18:57:30 -05002391 struct mount *p, *q;
Al Virocb338d02011-11-24 20:55:08 -05002392 struct mount *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002394 new_ns = alloc_mnt_ns();
2395 if (IS_ERR(new_ns))
2396 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397
Ram Pai390c6842005-11-07 17:17:51 -05002398 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399 /* First pass: copy the tree topology */
Al Viro87129cc2011-11-24 21:24:27 -05002400 new = copy_tree(real_mount(mnt_ns->root), mnt_ns->root->mnt_root,
Ram Pai9676f0c2005-11-07 17:21:20 -05002401 CL_COPY_ALL | CL_EXPIRE);
Al Virocb338d02011-11-24 20:55:08 -05002402 if (!new) {
Ram Pai390c6842005-11-07 17:17:51 -05002403 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404 kfree(new_ns);
Julia Lawall5cc4a032008-12-01 14:34:51 -08002405 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406 }
Al Virocb338d02011-11-24 20:55:08 -05002407 new_ns->root = &new->mnt;
Nick Piggin99b7db72010-08-18 04:37:39 +10002408 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409 list_add_tail(&new_ns->list, &new_ns->root->mnt_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10002410 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002411
2412 /*
2413 * Second pass: switch the tsk->fs->* elements and mark new vfsmounts
2414 * as belonging to new namespace. We have already acquired a private
2415 * fs_struct, so tsk->fs->lock is not needed.
2416 */
Al Viro315fc832011-11-24 18:57:30 -05002417 p = real_mount(mnt_ns->root);
Al Virocb338d02011-11-24 20:55:08 -05002418 q = new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419 while (p) {
Al Viro315fc832011-11-24 18:57:30 -05002420 q->mnt.mnt_ns = new_ns;
2421 __mnt_make_longterm(&q->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422 if (fs) {
Al Viro315fc832011-11-24 18:57:30 -05002423 if (&p->mnt == fs->root.mnt) {
2424 fs->root.mnt = mntget(&q->mnt);
2425 __mnt_make_longterm(&q->mnt);
2426 mnt_make_shortterm(&p->mnt);
2427 rootmnt = &p->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428 }
Al Viro315fc832011-11-24 18:57:30 -05002429 if (&p->mnt == fs->pwd.mnt) {
2430 fs->pwd.mnt = mntget(&q->mnt);
2431 __mnt_make_longterm(&q->mnt);
2432 mnt_make_shortterm(&p->mnt);
2433 pwdmnt = &p->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435 }
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002436 p = next_mnt(p, mnt_ns->root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437 q = next_mnt(q, new_ns->root);
2438 }
Ram Pai390c6842005-11-07 17:17:51 -05002439 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 if (rootmnt)
Al Virof03c6592011-01-14 22:30:21 -05002442 mntput(rootmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443 if (pwdmnt)
Al Virof03c6592011-01-14 22:30:21 -05002444 mntput(pwdmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445
JANAK DESAI741a2952006-02-07 12:59:00 -08002446 return new_ns;
2447}
2448
Eric W. Biederman213dd262007-07-15 23:41:15 -07002449struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns,
Badari Pulavartye3222c42007-05-08 00:25:21 -07002450 struct fs_struct *new_fs)
JANAK DESAI741a2952006-02-07 12:59:00 -08002451{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002452 struct mnt_namespace *new_ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08002453
Badari Pulavartye3222c42007-05-08 00:25:21 -07002454 BUG_ON(!ns);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002455 get_mnt_ns(ns);
JANAK DESAI741a2952006-02-07 12:59:00 -08002456
2457 if (!(flags & CLONE_NEWNS))
Badari Pulavartye3222c42007-05-08 00:25:21 -07002458 return ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08002459
Badari Pulavartye3222c42007-05-08 00:25:21 -07002460 new_ns = dup_mnt_ns(ns, new_fs);
JANAK DESAI741a2952006-02-07 12:59:00 -08002461
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002462 put_mnt_ns(ns);
Badari Pulavartye3222c42007-05-08 00:25:21 -07002463 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002464}
2465
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002466/**
2467 * create_mnt_ns - creates a private namespace and adds a root filesystem
2468 * @mnt: pointer to the new root filesystem mountpoint
2469 */
Al Viro6c449c82011-11-16 22:01:36 -05002470static struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt)
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002471{
2472 struct mnt_namespace *new_ns;
2473
2474 new_ns = alloc_mnt_ns();
2475 if (!IS_ERR(new_ns)) {
2476 mnt->mnt_ns = new_ns;
Al Viro7e3d0eb02011-01-16 16:32:11 -05002477 __mnt_make_longterm(mnt);
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002478 new_ns->root = mnt;
2479 list_add(&new_ns->list, &new_ns->root->mnt_list);
Al Viroc1334492011-11-16 16:12:14 -05002480 } else {
2481 mntput(mnt);
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002482 }
2483 return new_ns;
2484}
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002485
Al Viroea441d12011-11-16 21:43:59 -05002486struct dentry *mount_subtree(struct vfsmount *mnt, const char *name)
2487{
2488 struct mnt_namespace *ns;
Al Virod31da0f2011-11-22 12:31:21 -05002489 struct super_block *s;
Al Viroea441d12011-11-16 21:43:59 -05002490 struct path path;
2491 int err;
2492
2493 ns = create_mnt_ns(mnt);
2494 if (IS_ERR(ns))
2495 return ERR_CAST(ns);
2496
2497 err = vfs_path_lookup(mnt->mnt_root, mnt,
2498 name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path);
2499
2500 put_mnt_ns(ns);
2501
2502 if (err)
2503 return ERR_PTR(err);
2504
2505 /* trade a vfsmount reference for active sb one */
Al Virod31da0f2011-11-22 12:31:21 -05002506 s = path.mnt->mnt_sb;
2507 atomic_inc(&s->s_active);
Al Viroea441d12011-11-16 21:43:59 -05002508 mntput(path.mnt);
2509 /* lock the sucker */
Al Virod31da0f2011-11-22 12:31:21 -05002510 down_write(&s->s_umount);
Al Viroea441d12011-11-16 21:43:59 -05002511 /* ... and return the root of (sub)tree on it */
2512 return path.dentry;
2513}
2514EXPORT_SYMBOL(mount_subtree);
2515
Heiko Carstensbdc480e2009-01-14 14:14:12 +01002516SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name,
2517 char __user *, type, unsigned long, flags, void __user *, data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518{
Vegard Nossumeca6f532009-09-18 13:05:45 -07002519 int ret;
2520 char *kernel_type;
2521 char *kernel_dir;
2522 char *kernel_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523 unsigned long data_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002524
Vegard Nossumeca6f532009-09-18 13:05:45 -07002525 ret = copy_mount_string(type, &kernel_type);
2526 if (ret < 0)
2527 goto out_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528
Vegard Nossumeca6f532009-09-18 13:05:45 -07002529 kernel_dir = getname(dir_name);
2530 if (IS_ERR(kernel_dir)) {
2531 ret = PTR_ERR(kernel_dir);
2532 goto out_dir;
2533 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534
Vegard Nossumeca6f532009-09-18 13:05:45 -07002535 ret = copy_mount_string(dev_name, &kernel_dev);
2536 if (ret < 0)
2537 goto out_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
Vegard Nossumeca6f532009-09-18 13:05:45 -07002539 ret = copy_mount_options(data, &data_page);
2540 if (ret < 0)
2541 goto out_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002542
Vegard Nossumeca6f532009-09-18 13:05:45 -07002543 ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags,
2544 (void *) data_page);
2545
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546 free_page(data_page);
Vegard Nossumeca6f532009-09-18 13:05:45 -07002547out_data:
2548 kfree(kernel_dev);
2549out_dev:
2550 putname(kernel_dir);
2551out_dir:
2552 kfree(kernel_type);
2553out_type:
2554 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555}
2556
2557/*
Al Viroafac7cb2011-11-23 19:34:49 -05002558 * Return true if path is reachable from root
2559 *
2560 * namespace_sem or vfsmount_lock is held
2561 */
2562bool is_path_reachable(struct vfsmount *mnt, struct dentry *dentry,
2563 const struct path *root)
2564{
2565 while (mnt != root->mnt && mnt_has_parent(mnt)) {
2566 dentry = mnt->mnt_mountpoint;
2567 mnt = mnt->mnt_parent;
2568 }
2569 return mnt == root->mnt && is_subdir(dentry, root->dentry);
2570}
2571
2572int path_is_under(struct path *path1, struct path *path2)
2573{
2574 int res;
2575 br_read_lock(vfsmount_lock);
2576 res = is_path_reachable(path1->mnt, path1->dentry, path2);
2577 br_read_unlock(vfsmount_lock);
2578 return res;
2579}
2580EXPORT_SYMBOL(path_is_under);
2581
2582/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002583 * pivot_root Semantics:
2584 * Moves the root file system of the current process to the directory put_old,
2585 * makes new_root as the new root file system of the current process, and sets
2586 * root/cwd of all processes which had them on the current root to new_root.
2587 *
2588 * Restrictions:
2589 * The new_root and put_old must be directories, and must not be on the
2590 * same file system as the current process root. The put_old must be
2591 * underneath new_root, i.e. adding a non-zero number of /.. to the string
2592 * pointed to by put_old must yield the same directory as new_root. No other
2593 * file system may be mounted on put_old. After all, new_root is a mountpoint.
2594 *
Neil Brown4a0d11f2006-01-08 01:03:18 -08002595 * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem.
2596 * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives
2597 * in this situation.
2598 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002599 * Notes:
2600 * - we don't move root/cwd if they are not at the root (reason: if something
2601 * cared enough to change them, it's probably wrong to force them elsewhere)
2602 * - it's okay to pick a root that isn't the root of a file system, e.g.
2603 * /nfs/my_root where /nfs is the mount point. It must be a mountpoint,
2604 * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root
2605 * first.
2606 */
Heiko Carstens3480b252009-01-14 14:14:16 +01002607SYSCALL_DEFINE2(pivot_root, const char __user *, new_root,
2608 const char __user *, put_old)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609{
Al Viro2d8f3032008-07-22 09:59:21 -04002610 struct path new, old, parent_path, root_parent, root;
Al Viro419148d2011-11-24 19:41:16 -05002611 struct mount *new_mnt, *root_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612 int error;
2613
2614 if (!capable(CAP_SYS_ADMIN))
2615 return -EPERM;
2616
Al Viro2d8f3032008-07-22 09:59:21 -04002617 error = user_path_dir(new_root, &new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618 if (error)
2619 goto out0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002620
Al Viro2d8f3032008-07-22 09:59:21 -04002621 error = user_path_dir(put_old, &old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622 if (error)
2623 goto out1;
2624
Al Viro2d8f3032008-07-22 09:59:21 -04002625 error = security_sb_pivotroot(&old, &new);
Al Virob12cea92011-03-18 08:55:38 -04002626 if (error)
2627 goto out2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002628
Miklos Szeredif7ad3c62010-08-10 11:41:36 +02002629 get_fs_root(current->fs, &root);
Al Virob12cea92011-03-18 08:55:38 -04002630 error = lock_mount(&old);
2631 if (error)
2632 goto out3;
2633
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634 error = -EINVAL;
Al Viro419148d2011-11-24 19:41:16 -05002635 new_mnt = real_mount(new.mnt);
2636 root_mnt = real_mount(root.mnt);
Al Viro2d8f3032008-07-22 09:59:21 -04002637 if (IS_MNT_SHARED(old.mnt) ||
2638 IS_MNT_SHARED(new.mnt->mnt_parent) ||
Al Viro8c3ee422008-03-22 18:00:39 -04002639 IS_MNT_SHARED(root.mnt->mnt_parent))
Al Virob12cea92011-03-18 08:55:38 -04002640 goto out4;
Al Viro27cb1572011-03-18 08:29:36 -04002641 if (!check_mnt(root.mnt) || !check_mnt(new.mnt))
Al Virob12cea92011-03-18 08:55:38 -04002642 goto out4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643 error = -ENOENT;
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04002644 if (d_unlinked(new.dentry))
Al Virob12cea92011-03-18 08:55:38 -04002645 goto out4;
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04002646 if (d_unlinked(old.dentry))
Al Virob12cea92011-03-18 08:55:38 -04002647 goto out4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002648 error = -EBUSY;
Al Viro2d8f3032008-07-22 09:59:21 -04002649 if (new.mnt == root.mnt ||
2650 old.mnt == root.mnt)
Al Virob12cea92011-03-18 08:55:38 -04002651 goto out4; /* loop, on the same file system */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652 error = -EINVAL;
Al Viro8c3ee422008-03-22 18:00:39 -04002653 if (root.mnt->mnt_root != root.dentry)
Al Virob12cea92011-03-18 08:55:38 -04002654 goto out4; /* not a mountpoint */
Al Virob2dba1a2011-11-23 19:26:23 -05002655 if (!mnt_has_parent(root.mnt))
Al Virob12cea92011-03-18 08:55:38 -04002656 goto out4; /* not attached */
Al Viro2d8f3032008-07-22 09:59:21 -04002657 if (new.mnt->mnt_root != new.dentry)
Al Virob12cea92011-03-18 08:55:38 -04002658 goto out4; /* not a mountpoint */
Al Virob2dba1a2011-11-23 19:26:23 -05002659 if (!mnt_has_parent(new.mnt))
Al Virob12cea92011-03-18 08:55:38 -04002660 goto out4; /* not attached */
Jan Blunck4ac91372008-02-14 19:34:32 -08002661 /* make sure we can reach put_old from new_root */
Al Viroafac7cb2011-11-23 19:34:49 -05002662 if (!is_path_reachable(old.mnt, old.dentry, &new))
Al Virob12cea92011-03-18 08:55:38 -04002663 goto out4;
Al Viro27cb1572011-03-18 08:29:36 -04002664 br_write_lock(vfsmount_lock);
Al Viro419148d2011-11-24 19:41:16 -05002665 detach_mnt(new_mnt, &parent_path);
2666 detach_mnt(root_mnt, &root_parent);
Jan Blunck4ac91372008-02-14 19:34:32 -08002667 /* mount old root on put_old */
Al Viro419148d2011-11-24 19:41:16 -05002668 attach_mnt(root_mnt, &old);
Jan Blunck4ac91372008-02-14 19:34:32 -08002669 /* mount new_root on / */
Al Viro419148d2011-11-24 19:41:16 -05002670 attach_mnt(new_mnt, &root_parent);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002671 touch_mnt_namespace(current->nsproxy->mnt_ns);
Nick Piggin99b7db72010-08-18 04:37:39 +10002672 br_write_unlock(vfsmount_lock);
Al Viro2d8f3032008-07-22 09:59:21 -04002673 chroot_fs_refs(&root, &new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002674 error = 0;
Al Virob12cea92011-03-18 08:55:38 -04002675out4:
2676 unlock_mount(&old);
2677 if (!error) {
2678 path_put(&root_parent);
2679 path_put(&parent_path);
2680 }
2681out3:
Al Viro8c3ee422008-03-22 18:00:39 -04002682 path_put(&root);
Al Virob12cea92011-03-18 08:55:38 -04002683out2:
Al Viro2d8f3032008-07-22 09:59:21 -04002684 path_put(&old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002685out1:
Al Viro2d8f3032008-07-22 09:59:21 -04002686 path_put(&new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002687out0:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002688 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002689}
2690
2691static void __init init_mount_tree(void)
2692{
2693 struct vfsmount *mnt;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002694 struct mnt_namespace *ns;
Jan Blunckac748a02008-02-14 19:34:39 -08002695 struct path root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696
2697 mnt = do_kern_mount("rootfs", 0, "rootfs", NULL);
2698 if (IS_ERR(mnt))
2699 panic("Can't create rootfs");
Nick Pigginb3e19d92011-01-07 17:50:11 +11002700
Trond Myklebust3b22edc2009-06-23 17:29:49 -04002701 ns = create_mnt_ns(mnt);
2702 if (IS_ERR(ns))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002703 panic("Can't allocate initial namespace");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002704
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002705 init_task.nsproxy->mnt_ns = ns;
2706 get_mnt_ns(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002707
Jan Blunckac748a02008-02-14 19:34:39 -08002708 root.mnt = ns->root;
2709 root.dentry = ns->root->mnt_root;
2710
2711 set_fs_pwd(current->fs, &root);
2712 set_fs_root(current->fs, &root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713}
2714
Denis Cheng74bf17c2007-10-16 23:26:30 -07002715void __init mnt_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002716{
Eric Dumazet13f14b42008-02-06 01:37:57 -08002717 unsigned u;
Randy Dunlap15a67dd2006-09-29 01:58:57 -07002718 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002719
Ram Pai390c6842005-11-07 17:17:51 -05002720 init_rwsem(&namespace_sem);
2721
Al Viro7d6fec42011-11-23 12:14:10 -05002722 mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount),
Paul Mundt20c2df82007-07-20 10:11:58 +09002723 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002724
Ram Paib58fed82005-11-07 17:16:09 -05002725 mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726
2727 if (!mount_hashtable)
2728 panic("Failed to allocate mount hash table\n");
2729
Mandeep Singh Baines80cdc6d2011-03-22 16:33:54 -07002730 printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002731
Eric Dumazet13f14b42008-02-06 01:37:57 -08002732 for (u = 0; u < HASH_SIZE; u++)
2733 INIT_LIST_HEAD(&mount_hashtable[u]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734
Nick Piggin99b7db72010-08-18 04:37:39 +10002735 br_lock_init(vfsmount_lock);
2736
Randy Dunlap15a67dd2006-09-29 01:58:57 -07002737 err = sysfs_init();
2738 if (err)
2739 printk(KERN_WARNING "%s: sysfs_init error: %d\n",
Harvey Harrison8e24eea2008-04-30 00:55:09 -07002740 __func__, err);
Greg Kroah-Hartman00d26662007-10-29 14:17:23 -06002741 fs_kobj = kobject_create_and_add("fs", NULL);
2742 if (!fs_kobj)
Harvey Harrison8e24eea2008-04-30 00:55:09 -07002743 printk(KERN_WARNING "%s: kobj create error\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002744 init_rootfs();
2745 init_mount_tree();
2746}
2747
Trond Myklebust616511d2009-06-22 15:09:13 -04002748void put_mnt_ns(struct mnt_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749{
Ram Pai70fbcdf2005-11-07 17:17:04 -05002750 LIST_HEAD(umount_list);
Trond Myklebust616511d2009-06-22 15:09:13 -04002751
Al Virod498b252010-02-05 02:21:06 -05002752 if (!atomic_dec_and_test(&ns->count))
Trond Myklebust616511d2009-06-22 15:09:13 -04002753 return;
Ram Pai390c6842005-11-07 17:17:51 -05002754 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10002755 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05002756 umount_tree(real_mount(ns->root), 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10002757 br_write_unlock(vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05002758 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05002759 release_mounts(&umount_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002760 kfree(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761}
Al Viro9d412a42011-03-17 22:08:28 -04002762
2763struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
2764{
Tim Chen423e0ab2011-07-19 09:32:38 -07002765 struct vfsmount *mnt;
2766 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
2767 if (!IS_ERR(mnt)) {
2768 /*
2769 * it is a longterm mount, don't release mnt until
2770 * we unmount before file sys is unregistered
2771 */
2772 mnt_make_longterm(mnt);
2773 }
2774 return mnt;
Al Viro9d412a42011-03-17 22:08:28 -04002775}
2776EXPORT_SYMBOL_GPL(kern_mount_data);
Tim Chen423e0ab2011-07-19 09:32:38 -07002777
2778void kern_unmount(struct vfsmount *mnt)
2779{
2780 /* release long term mount so mount point can be released */
2781 if (!IS_ERR_OR_NULL(mnt)) {
2782 mnt_make_shortterm(mnt);
2783 mntput(mnt);
2784 }
2785}
2786EXPORT_SYMBOL(kern_unmount);
Al Viro02125a82011-12-05 08:43:34 -05002787
2788bool our_mnt(struct vfsmount *mnt)
2789{
2790 return check_mnt(mnt);
2791}