blob: b8a30928d0c1c5cfef12ba5adae50eb87bdfe089 [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 Viro15169fe2011-11-25 00:50:41 -050088 res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt_id);
Al Virof21f6222009-06-24 03:12:00 -040089 if (!res)
Al Viro15169fe2011-11-25 00:50:41 -050090 mnt_id_start = 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 Viro15169fe2011-11-25 00:50:41 -0500100 int id = 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 Viro15169fe2011-11-25 00:50:41 -0500122 &mnt->mnt_group_id);
Al Virof21f6222009-06-24 03:12:00 -0400123 if (!res)
Al Viro15169fe2011-11-25 00:50:41 -0500124 mnt_group_start = 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 Viro15169fe2011-11-25 00:50:41 -0500134 int id = 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 Viro15169fe2011-11-25 00:50:41 -0500138 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 */
Al Viro83adc752011-11-24 22:37:54 -0500144static inline void mnt_add_count(struct mount *mnt, int n)
Nick Pigginb3e19d92011-01-07 17:50:11 +1100145{
146#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500147 this_cpu_add(mnt->mnt_pcp->mnt_count, n);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100148#else
149 preempt_disable();
Al Viro68e8a9f2011-11-24 22:53:09 -0500150 mnt->mnt_count += n;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100151 preempt_enable();
152#endif
153}
154
Nick Pigginb3e19d92011-01-07 17:50:11 +1100155/*
156 * vfsmount lock must be held for write
157 */
Al Viro83adc752011-11-24 22:37:54 -0500158unsigned int mnt_get_count(struct mount *mnt)
Nick Pigginb3e19d92011-01-07 17:50:11 +1100159{
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) {
Al Viro68e8a9f2011-11-24 22:53:09 -0500165 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100166 }
167
168 return count;
169#else
Al Viro68e8a9f2011-11-24 22:53:09 -0500170 return mnt->mnt_count;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100171#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) {
Al Viro52ba1622011-11-25 02:25:17 -0500186 p->mnt_devname = kstrdup(name, GFP_KERNEL);
187 if (!p->mnt_devname)
Li Zefan88b38782008-07-21 18:06:36 +0800188 goto out_free_id;
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100189 }
190
Nick Pigginb3e19d92011-01-07 17:50:11 +1100191#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500192 p->mnt_pcp = alloc_percpu(struct mnt_pcp);
193 if (!p->mnt_pcp)
Nick Pigginb3e19d92011-01-07 17:50:11 +1100194 goto out_free_devname;
195
Al Viro68e8a9f2011-11-24 22:53:09 -0500196 this_cpu_add(p->mnt_pcp->mnt_count, 1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100197#else
Al Viro68e8a9f2011-11-24 22:53:09 -0500198 p->mnt_count = 1;
199 p->mnt_writers = 0;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100200#endif
201
Al Viro1b8e5562011-11-24 21:01:32 -0500202 INIT_LIST_HEAD(&p->mnt_hash);
Al Viro6b41d532011-11-24 23:24:33 -0500203 INIT_LIST_HEAD(&p->mnt_child);
204 INIT_LIST_HEAD(&p->mnt_mounts);
Al Viro1a4eeaf2011-11-25 02:19:55 -0500205 INIT_LIST_HEAD(&p->mnt_list);
Al Viro6776db3d2011-11-25 00:22:05 -0500206 INIT_LIST_HEAD(&p->mnt_expire);
207 INIT_LIST_HEAD(&p->mnt_share);
208 INIT_LIST_HEAD(&p->mnt_slave_list);
209 INIT_LIST_HEAD(&p->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 Viro52ba1622011-11-25 02:25:17 -0500218 kfree(p->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
Al Viro83adc752011-11-24 22:37:54 -0500256static inline void mnt_inc_writers(struct mount *mnt)
Dave Hansen3d733632008-02-15 14:37:59 -0800257{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000258#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500259 this_cpu_inc(mnt->mnt_pcp->mnt_writers);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000260#else
Al Viro68e8a9f2011-11-24 22:53:09 -0500261 mnt->mnt_writers++;
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000262#endif
Dave Hansen3d733632008-02-15 14:37:59 -0800263}
Dave Hansen3d733632008-02-15 14:37:59 -0800264
Al Viro83adc752011-11-24 22:37:54 -0500265static inline void mnt_dec_writers(struct mount *mnt)
Dave Hansen3d733632008-02-15 14:37:59 -0800266{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000267#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500268 this_cpu_dec(mnt->mnt_pcp->mnt_writers);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000269#else
Al Viro68e8a9f2011-11-24 22:53:09 -0500270 mnt->mnt_writers--;
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000271#endif
272}
273
Al Viro83adc752011-11-24 22:37:54 -0500274static unsigned int mnt_get_writers(struct mount *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) {
Al Viro68e8a9f2011-11-24 22:53:09 -0500281 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
Al Viro83adc752011-11-24 22:37:54 -0500300 * @m: the mount on which to take a write
Dave Hansen83660252008-02-15 14:37:30 -0800301 *
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 */
Al Viro83adc752011-11-24 22:37:54 -0500308int mnt_want_write(struct vfsmount *m)
Dave Hansen83660252008-02-15 14:37:30 -0800309{
Al Viro83adc752011-11-24 22:37:54 -0500310 struct mount *mnt = real_mount(m);
Dave Hansen3d733632008-02-15 14:37:59 -0800311 int ret = 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800312
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000313 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100314 mnt_inc_writers(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000315 /*
Nick Pigginc6653a82011-01-07 17:50:10 +1100316 * The store to mnt_inc_writers must be visible before we pass
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000317 * MNT_WRITE_HOLD loop below, so that the slowpath can see our
318 * incremented count after it has set MNT_WRITE_HOLD.
319 */
320 smp_mb();
Al Viro83adc752011-11-24 22:37:54 -0500321 while (mnt->mnt.mnt_flags & MNT_WRITE_HOLD)
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000322 cpu_relax();
323 /*
324 * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will
325 * be set to match its requirements. So we must not load that until
326 * MNT_WRITE_HOLD is cleared.
327 */
328 smp_rmb();
Al Viro83adc752011-11-24 22:37:54 -0500329 if (__mnt_is_readonly(m)) {
Nick Pigginc6653a82011-01-07 17:50:10 +1100330 mnt_dec_writers(mnt);
Dave Hansen3d733632008-02-15 14:37:59 -0800331 ret = -EROFS;
332 goto out;
333 }
Dave Hansen3d733632008-02-15 14:37:59 -0800334out:
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000335 preempt_enable();
Dave Hansen3d733632008-02-15 14:37:59 -0800336 return ret;
Dave Hansen83660252008-02-15 14:37:30 -0800337}
338EXPORT_SYMBOL_GPL(mnt_want_write);
339
340/**
npiggin@suse.de96029c42009-04-26 20:25:55 +1000341 * mnt_clone_write - get write access to a mount
342 * @mnt: the mount on which to take a write
343 *
344 * This is effectively like mnt_want_write, except
345 * it must only be used to take an extra write reference
346 * on a mountpoint that we already know has a write reference
347 * on it. This allows some optimisation.
348 *
349 * After finished, mnt_drop_write must be called as usual to
350 * drop the reference.
351 */
352int mnt_clone_write(struct vfsmount *mnt)
353{
354 /* superblock may be r/o */
355 if (__mnt_is_readonly(mnt))
356 return -EROFS;
357 preempt_disable();
Al Viro83adc752011-11-24 22:37:54 -0500358 mnt_inc_writers(real_mount(mnt));
npiggin@suse.de96029c42009-04-26 20:25:55 +1000359 preempt_enable();
360 return 0;
361}
362EXPORT_SYMBOL_GPL(mnt_clone_write);
363
364/**
365 * mnt_want_write_file - get write access to a file's mount
366 * @file: the file who's mount on which to take a write
367 *
368 * This is like mnt_want_write, but it takes a file and can
369 * do some optimisations if the file is open for write already
370 */
371int mnt_want_write_file(struct file *file)
372{
OGAWA Hirofumi2d8dd382009-08-06 15:07:39 -0700373 struct inode *inode = file->f_dentry->d_inode;
374 if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode))
npiggin@suse.de96029c42009-04-26 20:25:55 +1000375 return mnt_want_write(file->f_path.mnt);
376 else
377 return mnt_clone_write(file->f_path.mnt);
378}
379EXPORT_SYMBOL_GPL(mnt_want_write_file);
380
381/**
Dave Hansen83660252008-02-15 14:37:30 -0800382 * mnt_drop_write - give up write access to a mount
383 * @mnt: the mount on which to give up write access
384 *
385 * Tells the low-level filesystem that we are done
386 * performing writes to it. Must be matched with
387 * mnt_want_write() call above.
388 */
389void mnt_drop_write(struct vfsmount *mnt)
390{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000391 preempt_disable();
Al Viro83adc752011-11-24 22:37:54 -0500392 mnt_dec_writers(real_mount(mnt));
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000393 preempt_enable();
Dave Hansen83660252008-02-15 14:37:30 -0800394}
395EXPORT_SYMBOL_GPL(mnt_drop_write);
396
Al Viro2a79f172011-12-09 08:06:57 -0500397void mnt_drop_write_file(struct file *file)
398{
399 mnt_drop_write(file->f_path.mnt);
400}
401EXPORT_SYMBOL(mnt_drop_write_file);
402
Al Viro83adc752011-11-24 22:37:54 -0500403static int mnt_make_readonly(struct mount *mnt)
Dave Hansen83660252008-02-15 14:37:30 -0800404{
Dave Hansen3d733632008-02-15 14:37:59 -0800405 int ret = 0;
406
Nick Piggin99b7db72010-08-18 04:37:39 +1000407 br_write_lock(vfsmount_lock);
Al Viro83adc752011-11-24 22:37:54 -0500408 mnt->mnt.mnt_flags |= MNT_WRITE_HOLD;
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000409 /*
410 * After storing MNT_WRITE_HOLD, we'll read the counters. This store
411 * should be visible before we do.
412 */
413 smp_mb();
414
415 /*
416 * With writers on hold, if this value is zero, then there are
417 * definitely no active writers (although held writers may subsequently
418 * increment the count, they'll have to wait, and decrement it after
419 * seeing MNT_READONLY).
420 *
421 * It is OK to have counter incremented on one CPU and decremented on
422 * another: the sum will add up correctly. The danger would be when we
423 * sum up each counter, if we read a counter before it is incremented,
424 * but then read another CPU's count which it has been subsequently
425 * decremented from -- we would see more decrements than we should.
426 * MNT_WRITE_HOLD protects against this scenario, because
427 * mnt_want_write first increments count, then smp_mb, then spins on
428 * MNT_WRITE_HOLD, so it can't be decremented by another CPU while
429 * we're counting up here.
430 */
Nick Pigginc6653a82011-01-07 17:50:10 +1100431 if (mnt_get_writers(mnt) > 0)
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000432 ret = -EBUSY;
433 else
Al Viro83adc752011-11-24 22:37:54 -0500434 mnt->mnt.mnt_flags |= MNT_READONLY;
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000435 /*
436 * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers
437 * that become unheld will see MNT_READONLY.
438 */
439 smp_wmb();
Al Viro83adc752011-11-24 22:37:54 -0500440 mnt->mnt.mnt_flags &= ~MNT_WRITE_HOLD;
Nick Piggin99b7db72010-08-18 04:37:39 +1000441 br_write_unlock(vfsmount_lock);
Dave Hansen3d733632008-02-15 14:37:59 -0800442 return ret;
Dave Hansen83660252008-02-15 14:37:30 -0800443}
Dave Hansen83660252008-02-15 14:37:30 -0800444
Al Viro83adc752011-11-24 22:37:54 -0500445static void __mnt_unmake_readonly(struct mount *mnt)
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800446{
Nick Piggin99b7db72010-08-18 04:37:39 +1000447 br_write_lock(vfsmount_lock);
Al Viro83adc752011-11-24 22:37:54 -0500448 mnt->mnt.mnt_flags &= ~MNT_READONLY;
Nick Piggin99b7db72010-08-18 04:37:39 +1000449 br_write_unlock(vfsmount_lock);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800450}
451
Al Virob105e272011-11-24 20:38:33 -0500452static void free_vfsmnt(struct mount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453{
Al Viro52ba1622011-11-25 02:25:17 -0500454 kfree(mnt->mnt_devname);
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100455 mnt_free_id(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000456#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500457 free_percpu(mnt->mnt_pcp);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000458#endif
Al Virob105e272011-11-24 20:38:33 -0500459 kmem_cache_free(mnt_cache, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460}
461
462/*
Ram Paia05964f2005-11-07 17:20:17 -0500463 * find the first or last mount at @dentry on vfsmount @mnt depending on
464 * @dir. If @dir is set return the first mount else return the last mount.
Nick Piggin99b7db72010-08-18 04:37:39 +1000465 * vfsmount_lock must be held for read or write.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 */
Al Viroc7105362011-11-24 18:22:03 -0500467struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry,
Ram Paia05964f2005-11-07 17:20:17 -0500468 int dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469{
Ram Paib58fed82005-11-07 17:16:09 -0500470 struct list_head *head = mount_hashtable + hash(mnt, dentry);
471 struct list_head *tmp = head;
Al Viroc7105362011-11-24 18:22:03 -0500472 struct mount *p, *found = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 for (;;) {
Ram Paia05964f2005-11-07 17:20:17 -0500475 tmp = dir ? tmp->next : tmp->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 p = NULL;
477 if (tmp == head)
478 break;
Al Viro1b8e5562011-11-24 21:01:32 -0500479 p = list_entry(tmp, struct mount, mnt_hash);
Al Viroa73324d2011-11-24 22:25:07 -0500480 if (&p->mnt_parent->mnt == mnt && p->mnt_mountpoint == dentry) {
Ram Paia05964f2005-11-07 17:20:17 -0500481 found = p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 break;
483 }
484 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 return found;
486}
487
Ram Paia05964f2005-11-07 17:20:17 -0500488/*
489 * lookup_mnt increments the ref count before returning
490 * the vfsmount struct.
491 */
Al Viro1c755af2009-04-18 14:06:57 -0400492struct vfsmount *lookup_mnt(struct path *path)
Ram Paia05964f2005-11-07 17:20:17 -0500493{
Al Viroc7105362011-11-24 18:22:03 -0500494 struct mount *child_mnt;
Nick Piggin99b7db72010-08-18 04:37:39 +1000495
496 br_read_lock(vfsmount_lock);
Al Viroc7105362011-11-24 18:22:03 -0500497 child_mnt = __lookup_mnt(path->mnt, path->dentry, 1);
498 if (child_mnt) {
499 mnt_add_count(child_mnt, 1);
500 br_read_unlock(vfsmount_lock);
501 return &child_mnt->mnt;
502 } else {
503 br_read_unlock(vfsmount_lock);
504 return NULL;
505 }
Ram Paia05964f2005-11-07 17:20:17 -0500506}
507
Al Viro143c8c92011-11-25 00:46:35 -0500508static inline int check_mnt(struct mount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800510 return mnt->mnt_ns == current->nsproxy->mnt_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511}
512
Nick Piggin99b7db72010-08-18 04:37:39 +1000513/*
514 * vfsmount lock must be held for write
515 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800516static void touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500517{
518 if (ns) {
519 ns->event = ++event;
520 wake_up_interruptible(&ns->poll);
521 }
522}
523
Nick Piggin99b7db72010-08-18 04:37:39 +1000524/*
525 * vfsmount lock must be held for write
526 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800527static void __touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500528{
529 if (ns && ns->event != event) {
530 ns->event = event;
531 wake_up_interruptible(&ns->poll);
532 }
533}
534
Nick Piggin99b7db72010-08-18 04:37:39 +1000535/*
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100536 * Clear dentry's mounted state if it has no remaining mounts.
537 * vfsmount_lock must be held for write.
538 */
Al Viroaa0a4cf2011-11-24 19:31:36 -0500539static void dentry_reset_mounted(struct dentry *dentry)
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100540{
541 unsigned u;
542
543 for (u = 0; u < HASH_SIZE; u++) {
Al Virod5e50f72011-11-24 20:58:57 -0500544 struct mount *p;
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100545
Al Viro1b8e5562011-11-24 21:01:32 -0500546 list_for_each_entry(p, &mount_hashtable[u], mnt_hash) {
Al Viroa73324d2011-11-24 22:25:07 -0500547 if (p->mnt_mountpoint == dentry)
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100548 return;
549 }
550 }
551 spin_lock(&dentry->d_lock);
552 dentry->d_flags &= ~DCACHE_MOUNTED;
553 spin_unlock(&dentry->d_lock);
554}
555
556/*
Nick Piggin99b7db72010-08-18 04:37:39 +1000557 * vfsmount lock must be held for write
558 */
Al Viro419148d2011-11-24 19:41:16 -0500559static void detach_mnt(struct mount *mnt, struct path *old_path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560{
Al Viroa73324d2011-11-24 22:25:07 -0500561 old_path->dentry = mnt->mnt_mountpoint;
Al Viro0714a532011-11-24 22:19:58 -0500562 old_path->mnt = &mnt->mnt_parent->mnt;
563 mnt->mnt_parent = mnt;
Al Viroa73324d2011-11-24 22:25:07 -0500564 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
Al Viro6b41d532011-11-24 23:24:33 -0500565 list_del_init(&mnt->mnt_child);
Al Viro1b8e5562011-11-24 21:01:32 -0500566 list_del_init(&mnt->mnt_hash);
Al Viroaa0a4cf2011-11-24 19:31:36 -0500567 dentry_reset_mounted(old_path->dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568}
569
Nick Piggin99b7db72010-08-18 04:37:39 +1000570/*
571 * vfsmount lock must be held for write
572 */
Al Viro14cf1fa2011-11-25 00:01:17 -0500573void mnt_set_mountpoint(struct mount *mnt, struct dentry *dentry,
Al Viro44d964d62011-11-24 21:28:22 -0500574 struct mount *child_mnt)
Ram Paib90fa9a2005-11-07 17:19:50 -0500575{
Al Viro14cf1fa2011-11-25 00:01:17 -0500576 child_mnt->mnt_parent = real_mount(mntget(&mnt->mnt));
Al Viroa73324d2011-11-24 22:25:07 -0500577 child_mnt->mnt_mountpoint = dget(dentry);
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100578 spin_lock(&dentry->d_lock);
579 dentry->d_flags |= DCACHE_MOUNTED;
580 spin_unlock(&dentry->d_lock);
Ram Paib90fa9a2005-11-07 17:19:50 -0500581}
582
Nick Piggin99b7db72010-08-18 04:37:39 +1000583/*
584 * vfsmount lock must be held for write
585 */
Al Viro419148d2011-11-24 19:41:16 -0500586static void attach_mnt(struct mount *mnt, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587{
Al Viro14cf1fa2011-11-25 00:01:17 -0500588 mnt_set_mountpoint(real_mount(path->mnt), path->dentry, mnt);
Al Viro1b8e5562011-11-24 21:01:32 -0500589 list_add_tail(&mnt->mnt_hash, mount_hashtable +
Al Viro1a390682008-03-21 20:48:19 -0400590 hash(path->mnt, path->dentry));
Al Viro6b41d532011-11-24 23:24:33 -0500591 list_add_tail(&mnt->mnt_child, &real_mount(path->mnt)->mnt_mounts);
Ram Paib90fa9a2005-11-07 17:19:50 -0500592}
593
Al Viro83adc752011-11-24 22:37:54 -0500594static inline void __mnt_make_longterm(struct mount *mnt)
Al Viro7e3d0eb02011-01-16 16:32:11 -0500595{
596#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500597 atomic_inc(&mnt->mnt_longterm);
Al Viro7e3d0eb02011-01-16 16:32:11 -0500598#endif
599}
600
601/* needs vfsmount lock for write */
Al Viro83adc752011-11-24 22:37:54 -0500602static inline void __mnt_make_shortterm(struct mount *mnt)
Al Viro7e3d0eb02011-01-16 16:32:11 -0500603{
604#ifdef CONFIG_SMP
Al Viro68e8a9f2011-11-24 22:53:09 -0500605 atomic_dec(&mnt->mnt_longterm);
Al Viro7e3d0eb02011-01-16 16:32:11 -0500606#endif
607}
608
Ram Paib90fa9a2005-11-07 17:19:50 -0500609/*
Nick Piggin99b7db72010-08-18 04:37:39 +1000610 * vfsmount lock must be held for write
Ram Paib90fa9a2005-11-07 17:19:50 -0500611 */
Al Viro4b2619a2011-11-24 19:47:15 -0500612static void commit_tree(struct mount *mnt)
Ram Paib90fa9a2005-11-07 17:19:50 -0500613{
Al Viro0714a532011-11-24 22:19:58 -0500614 struct mount *parent = mnt->mnt_parent;
Al Viro83adc752011-11-24 22:37:54 -0500615 struct mount *m;
Ram Paib90fa9a2005-11-07 17:19:50 -0500616 LIST_HEAD(head);
Al Viro143c8c92011-11-25 00:46:35 -0500617 struct mnt_namespace *n = parent->mnt_ns;
Ram Paib90fa9a2005-11-07 17:19:50 -0500618
Al Viro0714a532011-11-24 22:19:58 -0500619 BUG_ON(parent == mnt);
Ram Paib90fa9a2005-11-07 17:19:50 -0500620
Al Viro1a4eeaf2011-11-25 02:19:55 -0500621 list_add_tail(&head, &mnt->mnt_list);
622 list_for_each_entry(m, &head, mnt_list) {
Al Viro143c8c92011-11-25 00:46:35 -0500623 m->mnt_ns = n;
Al Viro7e3d0eb02011-01-16 16:32:11 -0500624 __mnt_make_longterm(m);
Al Virof03c6592011-01-14 22:30:21 -0500625 }
626
Ram Paib90fa9a2005-11-07 17:19:50 -0500627 list_splice(&head, n->list.prev);
628
Al Viro1b8e5562011-11-24 21:01:32 -0500629 list_add_tail(&mnt->mnt_hash, mount_hashtable +
Al Viroa73324d2011-11-24 22:25:07 -0500630 hash(&parent->mnt, mnt->mnt_mountpoint));
Al Viro6b41d532011-11-24 23:24:33 -0500631 list_add_tail(&mnt->mnt_child, &parent->mnt_mounts);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800632 touch_mnt_namespace(n);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633}
634
Al Viro315fc832011-11-24 18:57:30 -0500635static struct mount *next_mnt(struct mount *p, struct vfsmount *root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636{
Al Viro6b41d532011-11-24 23:24:33 -0500637 struct list_head *next = p->mnt_mounts.next;
638 if (next == &p->mnt_mounts) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 while (1) {
Al Viro315fc832011-11-24 18:57:30 -0500640 if (&p->mnt == root)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 return NULL;
Al Viro6b41d532011-11-24 23:24:33 -0500642 next = p->mnt_child.next;
643 if (next != &p->mnt_parent->mnt_mounts)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644 break;
Al Viro0714a532011-11-24 22:19:58 -0500645 p = p->mnt_parent;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 }
647 }
Al Viro6b41d532011-11-24 23:24:33 -0500648 return list_entry(next, struct mount, mnt_child);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649}
650
Al Viro315fc832011-11-24 18:57:30 -0500651static struct mount *skip_mnt_tree(struct mount *p)
Ram Pai9676f0c2005-11-07 17:21:20 -0500652{
Al Viro6b41d532011-11-24 23:24:33 -0500653 struct list_head *prev = p->mnt_mounts.prev;
654 while (prev != &p->mnt_mounts) {
655 p = list_entry(prev, struct mount, mnt_child);
656 prev = p->mnt_mounts.prev;
Ram Pai9676f0c2005-11-07 17:21:20 -0500657 }
658 return p;
659}
660
Al Viro9d412a42011-03-17 22:08:28 -0400661struct vfsmount *
662vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data)
663{
Al Virob105e272011-11-24 20:38:33 -0500664 struct mount *mnt;
Al Viro9d412a42011-03-17 22:08:28 -0400665 struct dentry *root;
666
667 if (!type)
668 return ERR_PTR(-ENODEV);
669
670 mnt = alloc_vfsmnt(name);
671 if (!mnt)
672 return ERR_PTR(-ENOMEM);
673
674 if (flags & MS_KERNMOUNT)
Al Virob105e272011-11-24 20:38:33 -0500675 mnt->mnt.mnt_flags = MNT_INTERNAL;
Al Viro9d412a42011-03-17 22:08:28 -0400676
677 root = mount_fs(type, flags, name, data);
678 if (IS_ERR(root)) {
679 free_vfsmnt(mnt);
680 return ERR_CAST(root);
681 }
682
Al Virob105e272011-11-24 20:38:33 -0500683 mnt->mnt.mnt_root = root;
684 mnt->mnt.mnt_sb = root->d_sb;
Al Viroa73324d2011-11-24 22:25:07 -0500685 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
Al Viro0714a532011-11-24 22:19:58 -0500686 mnt->mnt_parent = mnt;
Al Virob105e272011-11-24 20:38:33 -0500687 return &mnt->mnt;
Al Viro9d412a42011-03-17 22:08:28 -0400688}
689EXPORT_SYMBOL_GPL(vfs_kern_mount);
690
Al Viro87129cc2011-11-24 21:24:27 -0500691static struct mount *clone_mnt(struct mount *old, struct dentry *root,
Ram Pai36341f62005-11-07 17:17:22 -0500692 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693{
Al Viro87129cc2011-11-24 21:24:27 -0500694 struct super_block *sb = old->mnt.mnt_sb;
Al Viro52ba1622011-11-25 02:25:17 -0500695 struct mount *mnt = alloc_vfsmnt(old->mnt_devname);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
697 if (mnt) {
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100698 if (flag & (CL_SLAVE | CL_PRIVATE))
Al Viro15169fe2011-11-25 00:50:41 -0500699 mnt->mnt_group_id = 0; /* not a peer of original */
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100700 else
Al Viro15169fe2011-11-25 00:50:41 -0500701 mnt->mnt_group_id = old->mnt_group_id;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100702
Al Viro15169fe2011-11-25 00:50:41 -0500703 if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) {
Al Virob105e272011-11-24 20:38:33 -0500704 int err = mnt_alloc_group_id(mnt);
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100705 if (err)
706 goto out_free;
707 }
708
Al Viro87129cc2011-11-24 21:24:27 -0500709 mnt->mnt.mnt_flags = old->mnt.mnt_flags & ~MNT_WRITE_HOLD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710 atomic_inc(&sb->s_active);
Al Virob105e272011-11-24 20:38:33 -0500711 mnt->mnt.mnt_sb = sb;
712 mnt->mnt.mnt_root = dget(root);
Al Viroa73324d2011-11-24 22:25:07 -0500713 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
Al Viro0714a532011-11-24 22:19:58 -0500714 mnt->mnt_parent = mnt;
Ram Paib90fa9a2005-11-07 17:19:50 -0500715
Ram Pai5afe0022005-11-07 17:21:01 -0500716 if (flag & CL_SLAVE) {
Al Viro6776db3d2011-11-25 00:22:05 -0500717 list_add(&mnt->mnt_slave, &old->mnt_slave_list);
Al Viro32301922011-11-25 00:10:28 -0500718 mnt->mnt_master = old;
Al Virofc7be132011-11-25 01:05:37 -0500719 CLEAR_MNT_SHARED(mnt);
Al Viro8aec0802007-06-07 12:20:32 -0400720 } else if (!(flag & CL_PRIVATE)) {
Al Virofc7be132011-11-25 01:05:37 -0500721 if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old))
Al Viro6776db3d2011-11-25 00:22:05 -0500722 list_add(&mnt->mnt_share, &old->mnt_share);
Al Virod10e8de2011-11-25 00:07:16 -0500723 if (IS_MNT_SLAVE(old))
Al Viro6776db3d2011-11-25 00:22:05 -0500724 list_add(&mnt->mnt_slave, &old->mnt_slave);
Al Virod10e8de2011-11-25 00:07:16 -0500725 mnt->mnt_master = old->mnt_master;
Ram Pai5afe0022005-11-07 17:21:01 -0500726 }
Ram Paib90fa9a2005-11-07 17:19:50 -0500727 if (flag & CL_MAKE_SHARED)
Al Viro0f0afb12011-11-24 20:43:10 -0500728 set_mnt_shared(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729
730 /* stick the duplicate mount on the same expiry list
731 * as the original if that was on one */
Ram Pai36341f62005-11-07 17:17:22 -0500732 if (flag & CL_EXPIRE) {
Al Viro6776db3d2011-11-25 00:22:05 -0500733 if (!list_empty(&old->mnt_expire))
734 list_add(&mnt->mnt_expire, &old->mnt_expire);
Ram Pai36341f62005-11-07 17:17:22 -0500735 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 }
Al Virocb338d02011-11-24 20:55:08 -0500737 return mnt;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100738
739 out_free:
740 free_vfsmnt(mnt);
741 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742}
743
Al Viro83adc752011-11-24 22:37:54 -0500744static inline void mntfree(struct mount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745{
Al Viro83adc752011-11-24 22:37:54 -0500746 struct vfsmount *m = &mnt->mnt;
747 struct super_block *sb = m->mnt_sb;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100748
Dave Hansen3d733632008-02-15 14:37:59 -0800749 /*
Dave Hansen3d733632008-02-15 14:37:59 -0800750 * This probably indicates that somebody messed
751 * up a mnt_want/drop_write() pair. If this
752 * happens, the filesystem was probably unable
753 * to make r/w->r/o transitions.
754 */
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000755 /*
Nick Pigginb3e19d92011-01-07 17:50:11 +1100756 * The locking used to deal with mnt_count decrement provides barriers,
757 * so mnt_get_writers() below is safe.
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000758 */
Nick Pigginc6653a82011-01-07 17:50:10 +1100759 WARN_ON(mnt_get_writers(mnt));
Al Viro83adc752011-11-24 22:37:54 -0500760 fsnotify_vfsmount_delete(m);
761 dput(m->mnt_root);
762 free_vfsmnt(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 deactivate_super(sb);
764}
765
Al Viro900148d2011-11-25 00:33:11 -0500766static void mntput_no_expire(struct mount *mnt)
Al Viro7b7b1ac2005-11-07 17:13:39 -0500767{
Nick Pigginb3e19d92011-01-07 17:50:11 +1100768put_again:
Al Virof03c6592011-01-14 22:30:21 -0500769#ifdef CONFIG_SMP
770 br_read_lock(vfsmount_lock);
Al Viro68e8a9f2011-11-24 22:53:09 -0500771 if (likely(atomic_read(&mnt->mnt_longterm))) {
Al Viroaa9c0e02011-11-23 19:04:52 -0500772 mnt_add_count(mnt, -1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100773 br_read_unlock(vfsmount_lock);
Al Virof03c6592011-01-14 22:30:21 -0500774 return;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100775 }
Al Virof03c6592011-01-14 22:30:21 -0500776 br_read_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100777
778 br_write_lock(vfsmount_lock);
Al Viroaa9c0e02011-11-23 19:04:52 -0500779 mnt_add_count(mnt, -1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100780 if (mnt_get_count(mnt)) {
781 br_write_unlock(vfsmount_lock);
782 return;
783 }
Nick Pigginb3e19d92011-01-07 17:50:11 +1100784#else
Al Viroaa9c0e02011-11-23 19:04:52 -0500785 mnt_add_count(mnt, -1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100786 if (likely(mnt_get_count(mnt)))
Nick Piggin99b7db72010-08-18 04:37:39 +1000787 return;
788 br_write_lock(vfsmount_lock);
Al Virof03c6592011-01-14 22:30:21 -0500789#endif
Al Viro863d6842011-11-25 00:57:42 -0500790 if (unlikely(mnt->mnt_pinned)) {
791 mnt_add_count(mnt, mnt->mnt_pinned + 1);
792 mnt->mnt_pinned = 0;
Nick Piggin99b7db72010-08-18 04:37:39 +1000793 br_write_unlock(vfsmount_lock);
Al Viro900148d2011-11-25 00:33:11 -0500794 acct_auto_close_mnt(&mnt->mnt);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100795 goto put_again;
Al Viro7b7b1ac2005-11-07 17:13:39 -0500796 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000797 br_write_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100798 mntfree(mnt);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500799}
Nick Pigginb3e19d92011-01-07 17:50:11 +1100800
801void mntput(struct vfsmount *mnt)
802{
803 if (mnt) {
Al Viro863d6842011-11-25 00:57:42 -0500804 struct mount *m = real_mount(mnt);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100805 /* avoid cacheline pingpong, hope gcc doesn't get "smart" */
Al Viro863d6842011-11-25 00:57:42 -0500806 if (unlikely(m->mnt_expiry_mark))
807 m->mnt_expiry_mark = 0;
808 mntput_no_expire(m);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100809 }
810}
811EXPORT_SYMBOL(mntput);
812
813struct vfsmount *mntget(struct vfsmount *mnt)
814{
815 if (mnt)
Al Viro83adc752011-11-24 22:37:54 -0500816 mnt_add_count(real_mount(mnt), 1);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100817 return mnt;
818}
819EXPORT_SYMBOL(mntget);
820
Al Viro7b7b1ac2005-11-07 17:13:39 -0500821void mnt_pin(struct vfsmount *mnt)
822{
Nick Piggin99b7db72010-08-18 04:37:39 +1000823 br_write_lock(vfsmount_lock);
Al Viro863d6842011-11-25 00:57:42 -0500824 real_mount(mnt)->mnt_pinned++;
Nick Piggin99b7db72010-08-18 04:37:39 +1000825 br_write_unlock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500826}
Al Viro7b7b1ac2005-11-07 17:13:39 -0500827EXPORT_SYMBOL(mnt_pin);
828
Al Viro863d6842011-11-25 00:57:42 -0500829void mnt_unpin(struct vfsmount *m)
Al Viro7b7b1ac2005-11-07 17:13:39 -0500830{
Al Viro863d6842011-11-25 00:57:42 -0500831 struct mount *mnt = real_mount(m);
Nick Piggin99b7db72010-08-18 04:37:39 +1000832 br_write_lock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500833 if (mnt->mnt_pinned) {
Al Viro863d6842011-11-25 00:57:42 -0500834 mnt_add_count(mnt, 1);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500835 mnt->mnt_pinned--;
836 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000837 br_write_unlock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500838}
Al Viro7b7b1ac2005-11-07 17:13:39 -0500839EXPORT_SYMBOL(mnt_unpin);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800841static inline void mangle(struct seq_file *m, const char *s)
842{
843 seq_escape(m, s, " \t\n\\");
844}
845
846/*
847 * Simple .show_options callback for filesystems which don't want to
848 * implement more complex mount option showing.
849 *
850 * See also save_mount_options().
851 */
852int generic_show_options(struct seq_file *m, struct vfsmount *mnt)
853{
Al Viro2a32ceb2009-05-08 16:05:57 -0400854 const char *options;
855
856 rcu_read_lock();
857 options = rcu_dereference(mnt->mnt_sb->s_options);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800858
859 if (options != NULL && options[0]) {
860 seq_putc(m, ',');
861 mangle(m, options);
862 }
Al Viro2a32ceb2009-05-08 16:05:57 -0400863 rcu_read_unlock();
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800864
865 return 0;
866}
867EXPORT_SYMBOL(generic_show_options);
868
869/*
870 * If filesystem uses generic_show_options(), this function should be
871 * called from the fill_super() callback.
872 *
873 * The .remount_fs callback usually needs to be handled in a special
874 * way, to make sure, that previous options are not overwritten if the
875 * remount fails.
876 *
877 * Also note, that if the filesystem's .remount_fs function doesn't
878 * reset all options to their default value, but changes only newly
879 * given options, then the displayed options will not reflect reality
880 * any more.
881 */
882void save_mount_options(struct super_block *sb, char *options)
883{
Al Viro2a32ceb2009-05-08 16:05:57 -0400884 BUG_ON(sb->s_options);
885 rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL));
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800886}
887EXPORT_SYMBOL(save_mount_options);
888
Al Viro2a32ceb2009-05-08 16:05:57 -0400889void replace_mount_options(struct super_block *sb, char *options)
890{
891 char *old = sb->s_options;
892 rcu_assign_pointer(sb->s_options, options);
893 if (old) {
894 synchronize_rcu();
895 kfree(old);
896 }
897}
898EXPORT_SYMBOL(replace_mount_options);
899
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100900#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901/* iterator */
902static void *m_start(struct seq_file *m, loff_t *pos)
903{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100904 struct proc_mounts *p = m->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905
Ram Pai390c6842005-11-07 17:17:51 -0500906 down_read(&namespace_sem);
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100907 return seq_list_start(&p->ns->list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908}
909
910static void *m_next(struct seq_file *m, void *v, loff_t *pos)
911{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100912 struct proc_mounts *p = m->private;
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700913
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100914 return seq_list_next(v, &p->ns->list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916
917static void m_stop(struct seq_file *m, void *v)
918{
Ram Pai390c6842005-11-07 17:17:51 -0500919 up_read(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920}
921
Al Viro9f5596a2010-02-05 00:40:25 -0500922int mnt_had_events(struct proc_mounts *p)
923{
924 struct mnt_namespace *ns = p->ns;
925 int res = 0;
926
Nick Piggin99b7db72010-08-18 04:37:39 +1000927 br_read_lock(vfsmount_lock);
Kay Sieversf1514632011-07-12 20:48:39 +0200928 if (p->m.poll_event != ns->event) {
929 p->m.poll_event = ns->event;
Al Viro9f5596a2010-02-05 00:40:25 -0500930 res = 1;
931 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000932 br_read_unlock(vfsmount_lock);
Al Viro9f5596a2010-02-05 00:40:25 -0500933
934 return res;
935}
936
Ram Pai2d4d4862008-03-27 13:06:25 +0100937struct proc_fs_info {
938 int flag;
939 const char *str;
940};
941
Eric Paris2069f452008-07-04 09:47:13 +1000942static int show_sb_opts(struct seq_file *m, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943{
Ram Pai2d4d4862008-03-27 13:06:25 +0100944 static const struct proc_fs_info fs_info[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 { MS_SYNCHRONOUS, ",sync" },
946 { MS_DIRSYNC, ",dirsync" },
947 { MS_MANDLOCK, ",mand" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 { 0, NULL }
949 };
Ram Pai2d4d4862008-03-27 13:06:25 +0100950 const struct proc_fs_info *fs_infop;
951
952 for (fs_infop = fs_info; fs_infop->flag; fs_infop++) {
953 if (sb->s_flags & fs_infop->flag)
954 seq_puts(m, fs_infop->str);
955 }
Eric Paris2069f452008-07-04 09:47:13 +1000956
957 return security_sb_show_options(m, sb);
Ram Pai2d4d4862008-03-27 13:06:25 +0100958}
959
960static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt)
961{
962 static const struct proc_fs_info mnt_info[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 { MNT_NOSUID, ",nosuid" },
964 { MNT_NODEV, ",nodev" },
965 { MNT_NOEXEC, ",noexec" },
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -0800966 { MNT_NOATIME, ",noatime" },
967 { MNT_NODIRATIME, ",nodiratime" },
Valerie Henson47ae32d2006-12-13 00:34:34 -0800968 { MNT_RELATIME, ",relatime" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 { 0, NULL }
970 };
Ram Pai2d4d4862008-03-27 13:06:25 +0100971 const struct proc_fs_info *fs_infop;
972
973 for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) {
974 if (mnt->mnt_flags & fs_infop->flag)
975 seq_puts(m, fs_infop->str);
976 }
977}
978
979static void show_type(struct seq_file *m, struct super_block *sb)
980{
981 mangle(m, sb->s_type->name);
982 if (sb->s_subtype && sb->s_subtype[0]) {
983 seq_putc(m, '.');
984 mangle(m, sb->s_subtype);
985 }
986}
987
988static int show_vfsmnt(struct seq_file *m, void *v)
989{
Al Viro1a4eeaf2011-11-25 02:19:55 -0500990 struct mount *r = list_entry(v, struct mount, mnt_list);
991 struct vfsmount *mnt = &r->mnt;
Ram Pai2d4d4862008-03-27 13:06:25 +0100992 int err = 0;
Jan Blunckc32c2f62008-02-14 19:38:43 -0800993 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
Al Viroc7f404b2011-03-16 06:59:40 -0400995 if (mnt->mnt_sb->s_op->show_devname) {
996 err = mnt->mnt_sb->s_op->show_devname(m, mnt);
997 if (err)
998 goto out;
999 } else {
Al Viro52ba1622011-11-25 02:25:17 -05001000 mangle(m, r->mnt_devname ? r->mnt_devname : "none");
Al Viroc7f404b2011-03-16 06:59:40 -04001001 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 seq_putc(m, ' ');
Jan Blunckc32c2f62008-02-14 19:38:43 -08001003 seq_path(m, &mnt_path, " \t\n\\");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 seq_putc(m, ' ');
Ram Pai2d4d4862008-03-27 13:06:25 +01001005 show_type(m, mnt->mnt_sb);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001006 seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw");
Eric Paris2069f452008-07-04 09:47:13 +10001007 err = show_sb_opts(m, mnt->mnt_sb);
1008 if (err)
1009 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001010 show_mnt_opts(m, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011 if (mnt->mnt_sb->s_op->show_options)
1012 err = mnt->mnt_sb->s_op->show_options(m, mnt);
1013 seq_puts(m, " 0 0\n");
Eric Paris2069f452008-07-04 09:47:13 +10001014out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 return err;
1016}
1017
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001018const struct seq_operations mounts_op = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 .start = m_start,
1020 .next = m_next,
1021 .stop = m_stop,
1022 .show = show_vfsmnt
1023};
1024
Ram Pai2d4d4862008-03-27 13:06:25 +01001025static int show_mountinfo(struct seq_file *m, void *v)
1026{
1027 struct proc_mounts *p = m->private;
Al Viro1a4eeaf2011-11-25 02:19:55 -05001028 struct mount *r = list_entry(v, struct mount, mnt_list);
1029 struct vfsmount *mnt = &r->mnt;
Ram Pai2d4d4862008-03-27 13:06:25 +01001030 struct super_block *sb = mnt->mnt_sb;
1031 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
1032 struct path root = p->root;
1033 int err = 0;
1034
Al Viro15169fe2011-11-25 00:50:41 -05001035 seq_printf(m, "%i %i %u:%u ", r->mnt_id, r->mnt_parent->mnt_id,
Ram Pai2d4d4862008-03-27 13:06:25 +01001036 MAJOR(sb->s_dev), MINOR(sb->s_dev));
Al Viroc7f404b2011-03-16 06:59:40 -04001037 if (sb->s_op->show_path)
1038 err = sb->s_op->show_path(m, mnt);
1039 else
1040 seq_dentry(m, mnt->mnt_root, " \t\n\\");
1041 if (err)
1042 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001043 seq_putc(m, ' ');
Al Viro02125a82011-12-05 08:43:34 -05001044
1045 /* mountpoints outside of chroot jail will give SEQ_SKIP on this */
1046 err = seq_path_root(m, &mnt_path, &root, " \t\n\\");
1047 if (err)
1048 goto out;
1049
Ram Pai2d4d4862008-03-27 13:06:25 +01001050 seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw");
1051 show_mnt_opts(m, mnt);
1052
1053 /* Tagged fields ("foo:X" or "bar") */
Al Virofc7be132011-11-25 01:05:37 -05001054 if (IS_MNT_SHARED(r))
Al Viro15169fe2011-11-25 00:50:41 -05001055 seq_printf(m, " shared:%i", r->mnt_group_id);
Al Virod10e8de2011-11-25 00:07:16 -05001056 if (IS_MNT_SLAVE(r)) {
Al Viro15169fe2011-11-25 00:50:41 -05001057 int master = r->mnt_master->mnt_group_id;
Al Viro6fc78712011-11-24 23:35:54 -05001058 int dom = get_dominating_id(r, &p->root);
Miklos Szeredi97e7e0f2008-03-27 13:06:26 +01001059 seq_printf(m, " master:%i", master);
1060 if (dom && dom != master)
1061 seq_printf(m, " propagate_from:%i", dom);
1062 }
Al Virofc7be132011-11-25 01:05:37 -05001063 if (IS_MNT_UNBINDABLE(r))
Ram Pai2d4d4862008-03-27 13:06:25 +01001064 seq_puts(m, " unbindable");
1065
1066 /* Filesystem specific data */
1067 seq_puts(m, " - ");
1068 show_type(m, sb);
1069 seq_putc(m, ' ');
Al Viroc7f404b2011-03-16 06:59:40 -04001070 if (sb->s_op->show_devname)
1071 err = sb->s_op->show_devname(m, mnt);
1072 else
Al Viro52ba1622011-11-25 02:25:17 -05001073 mangle(m, r->mnt_devname ? r->mnt_devname : "none");
Al Viroc7f404b2011-03-16 06:59:40 -04001074 if (err)
1075 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001076 seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw");
Eric Paris2069f452008-07-04 09:47:13 +10001077 err = show_sb_opts(m, sb);
1078 if (err)
1079 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001080 if (sb->s_op->show_options)
1081 err = sb->s_op->show_options(m, mnt);
1082 seq_putc(m, '\n');
Eric Paris2069f452008-07-04 09:47:13 +10001083out:
Ram Pai2d4d4862008-03-27 13:06:25 +01001084 return err;
1085}
1086
1087const struct seq_operations mountinfo_op = {
1088 .start = m_start,
1089 .next = m_next,
1090 .stop = m_stop,
1091 .show = show_mountinfo,
1092};
1093
Chuck Leverb4629fe2006-03-20 13:44:12 -05001094static int show_vfsstat(struct seq_file *m, void *v)
1095{
Al Viro1a4eeaf2011-11-25 02:19:55 -05001096 struct mount *r = list_entry(v, struct mount, mnt_list);
1097 struct vfsmount *mnt = &r->mnt;
Jan Blunckc32c2f62008-02-14 19:38:43 -08001098 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
Chuck Leverb4629fe2006-03-20 13:44:12 -05001099 int err = 0;
1100
1101 /* device */
Al Viroc7f404b2011-03-16 06:59:40 -04001102 if (mnt->mnt_sb->s_op->show_devname) {
Bryan Schumakera877ee02011-10-07 13:41:15 -04001103 seq_puts(m, "device ");
Al Viroc7f404b2011-03-16 06:59:40 -04001104 err = mnt->mnt_sb->s_op->show_devname(m, mnt);
1105 } else {
Al Viro52ba1622011-11-25 02:25:17 -05001106 if (r->mnt_devname) {
Al Viroc7f404b2011-03-16 06:59:40 -04001107 seq_puts(m, "device ");
Al Viro52ba1622011-11-25 02:25:17 -05001108 mangle(m, r->mnt_devname);
Al Viroc7f404b2011-03-16 06:59:40 -04001109 } else
1110 seq_puts(m, "no device");
1111 }
Chuck Leverb4629fe2006-03-20 13:44:12 -05001112
1113 /* mount point */
1114 seq_puts(m, " mounted on ");
Jan Blunckc32c2f62008-02-14 19:38:43 -08001115 seq_path(m, &mnt_path, " \t\n\\");
Chuck Leverb4629fe2006-03-20 13:44:12 -05001116 seq_putc(m, ' ');
1117
1118 /* file system type */
1119 seq_puts(m, "with fstype ");
Ram Pai2d4d4862008-03-27 13:06:25 +01001120 show_type(m, mnt->mnt_sb);
Chuck Leverb4629fe2006-03-20 13:44:12 -05001121
1122 /* optional statistics */
1123 if (mnt->mnt_sb->s_op->show_stats) {
1124 seq_putc(m, ' ');
Al Viroc7f404b2011-03-16 06:59:40 -04001125 if (!err)
1126 err = mnt->mnt_sb->s_op->show_stats(m, mnt);
Chuck Leverb4629fe2006-03-20 13:44:12 -05001127 }
1128
1129 seq_putc(m, '\n');
1130 return err;
1131}
1132
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001133const struct seq_operations mountstats_op = {
Chuck Leverb4629fe2006-03-20 13:44:12 -05001134 .start = m_start,
1135 .next = m_next,
1136 .stop = m_stop,
1137 .show = show_vfsstat,
1138};
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001139#endif /* CONFIG_PROC_FS */
Chuck Leverb4629fe2006-03-20 13:44:12 -05001140
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141/**
1142 * may_umount_tree - check if a mount tree is busy
1143 * @mnt: root of mount tree
1144 *
1145 * This is called to check if a tree of mounts has any
1146 * open files, pwds, chroots or sub mounts that are
1147 * busy.
1148 */
1149int may_umount_tree(struct vfsmount *mnt)
1150{
Ram Pai36341f62005-11-07 17:17:22 -05001151 int actual_refs = 0;
1152 int minimum_refs = 0;
Al Viro315fc832011-11-24 18:57:30 -05001153 struct mount *p;
1154 BUG_ON(!mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155
Nick Pigginb3e19d92011-01-07 17:50:11 +11001156 /* write lock needed for mnt_get_count */
1157 br_write_lock(vfsmount_lock);
Al Viro315fc832011-11-24 18:57:30 -05001158 for (p = real_mount(mnt); p; p = next_mnt(p, mnt)) {
Al Viro83adc752011-11-24 22:37:54 -05001159 actual_refs += mnt_get_count(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 minimum_refs += 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 }
Nick Pigginb3e19d92011-01-07 17:50:11 +11001162 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164 if (actual_refs > minimum_refs)
Ian Kente3474a82006-03-27 01:14:51 -08001165 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
Ian Kente3474a82006-03-27 01:14:51 -08001167 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168}
1169
1170EXPORT_SYMBOL(may_umount_tree);
1171
1172/**
1173 * may_umount - check if a mount point is busy
1174 * @mnt: root of mount
1175 *
1176 * This is called to check if a mount point has any
1177 * open files, pwds, chroots or sub mounts. If the
1178 * mount has sub mounts this will return busy
1179 * regardless of whether the sub mounts are busy.
1180 *
1181 * Doesn't take quota and stuff into account. IOW, in some cases it will
1182 * give false negatives. The main reason why it's here is that we need
1183 * a non-destructive way to look for easily umountable filesystems.
1184 */
1185int may_umount(struct vfsmount *mnt)
1186{
Ian Kente3474a82006-03-27 01:14:51 -08001187 int ret = 1;
Al Viro8ad08d82010-01-16 12:56:08 -05001188 down_read(&namespace_sem);
Nick Pigginb3e19d92011-01-07 17:50:11 +11001189 br_write_lock(vfsmount_lock);
Al Viro1ab59732011-11-24 21:35:16 -05001190 if (propagate_mount_busy(real_mount(mnt), 2))
Ian Kente3474a82006-03-27 01:14:51 -08001191 ret = 0;
Nick Pigginb3e19d92011-01-07 17:50:11 +11001192 br_write_unlock(vfsmount_lock);
Al Viro8ad08d82010-01-16 12:56:08 -05001193 up_read(&namespace_sem);
Ram Paia05964f2005-11-07 17:20:17 -05001194 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195}
1196
1197EXPORT_SYMBOL(may_umount);
1198
Ram Paib90fa9a2005-11-07 17:19:50 -05001199void release_mounts(struct list_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200{
Al Virod5e50f72011-11-24 20:58:57 -05001201 struct mount *mnt;
Miklos Szeredibf066c72006-01-08 01:03:19 -08001202 while (!list_empty(head)) {
Al Viro1b8e5562011-11-24 21:01:32 -05001203 mnt = list_first_entry(head, struct mount, mnt_hash);
1204 list_del_init(&mnt->mnt_hash);
Al Viro676da582011-11-24 21:47:05 -05001205 if (mnt_has_parent(mnt)) {
Ram Pai70fbcdf2005-11-07 17:17:04 -05001206 struct dentry *dentry;
Al Viro863d6842011-11-25 00:57:42 -05001207 struct mount *m;
Nick Piggin99b7db72010-08-18 04:37:39 +10001208
1209 br_write_lock(vfsmount_lock);
Al Viroa73324d2011-11-24 22:25:07 -05001210 dentry = mnt->mnt_mountpoint;
Al Viro863d6842011-11-25 00:57:42 -05001211 m = mnt->mnt_parent;
Al Viroa73324d2011-11-24 22:25:07 -05001212 mnt->mnt_mountpoint = mnt->mnt.mnt_root;
Al Viro0714a532011-11-24 22:19:58 -05001213 mnt->mnt_parent = mnt;
Al Viro7c4b93d2008-03-21 23:59:49 -04001214 m->mnt_ghosts--;
Nick Piggin99b7db72010-08-18 04:37:39 +10001215 br_write_unlock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001216 dput(dentry);
Al Viro863d6842011-11-25 00:57:42 -05001217 mntput(&m->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 }
Al Virod5e50f72011-11-24 20:58:57 -05001219 mntput(&mnt->mnt);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001220 }
1221}
1222
Nick Piggin99b7db72010-08-18 04:37:39 +10001223/*
1224 * vfsmount lock must be held for write
1225 * namespace_sem must be held for write
1226 */
Al Viro761d5c32011-11-24 21:07:43 -05001227void umount_tree(struct mount *mnt, int propagate, struct list_head *kill)
Ram Pai70fbcdf2005-11-07 17:17:04 -05001228{
Al Viro7b8a53f2011-01-15 20:08:44 -05001229 LIST_HEAD(tmp_list);
Al Viro315fc832011-11-24 18:57:30 -05001230 struct mount *p;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001231
Al Viro761d5c32011-11-24 21:07:43 -05001232 for (p = mnt; p; p = next_mnt(p, &mnt->mnt))
Al Viro1b8e5562011-11-24 21:01:32 -05001233 list_move(&p->mnt_hash, &tmp_list);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001234
Ram Paia05964f2005-11-07 17:20:17 -05001235 if (propagate)
Al Viro7b8a53f2011-01-15 20:08:44 -05001236 propagate_umount(&tmp_list);
Ram Paia05964f2005-11-07 17:20:17 -05001237
Al Viro1b8e5562011-11-24 21:01:32 -05001238 list_for_each_entry(p, &tmp_list, mnt_hash) {
Al Viro6776db3d2011-11-25 00:22:05 -05001239 list_del_init(&p->mnt_expire);
Al Viro1a4eeaf2011-11-25 02:19:55 -05001240 list_del_init(&p->mnt_list);
Al Viro143c8c92011-11-25 00:46:35 -05001241 __touch_mnt_namespace(p->mnt_ns);
1242 p->mnt_ns = NULL;
Al Viro83adc752011-11-24 22:37:54 -05001243 __mnt_make_shortterm(p);
Al Viro6b41d532011-11-24 23:24:33 -05001244 list_del_init(&p->mnt_child);
Al Viro676da582011-11-24 21:47:05 -05001245 if (mnt_has_parent(p)) {
Al Viro863d6842011-11-25 00:57:42 -05001246 p->mnt_parent->mnt_ghosts++;
Al Viroa73324d2011-11-24 22:25:07 -05001247 dentry_reset_mounted(p->mnt_mountpoint);
Al Viro7c4b93d2008-03-21 23:59:49 -04001248 }
Al Viro0f0afb12011-11-24 20:43:10 -05001249 change_mnt_propagation(p, MS_PRIVATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 }
Al Viro7b8a53f2011-01-15 20:08:44 -05001251 list_splice(&tmp_list, kill);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252}
1253
Al Viro692afc32011-11-24 21:15:14 -05001254static void shrink_submounts(struct mount *mnt, struct list_head *umounts);
Al Viroc35038b2008-03-22 00:46:23 -04001255
Al Viro1ab59732011-11-24 21:35:16 -05001256static int do_umount(struct mount *mnt, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257{
Al Viro1ab59732011-11-24 21:35:16 -05001258 struct super_block *sb = mnt->mnt.mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 int retval;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001260 LIST_HEAD(umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261
Al Viro1ab59732011-11-24 21:35:16 -05001262 retval = security_sb_umount(&mnt->mnt, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 if (retval)
1264 return retval;
1265
1266 /*
1267 * Allow userspace to request a mountpoint be expired rather than
1268 * unmounting unconditionally. Unmount only happens if:
1269 * (1) the mark is already set (the mark is cleared by mntput())
1270 * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount]
1271 */
1272 if (flags & MNT_EXPIRE) {
Al Viro1ab59732011-11-24 21:35:16 -05001273 if (&mnt->mnt == current->fs->root.mnt ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 flags & (MNT_FORCE | MNT_DETACH))
1275 return -EINVAL;
1276
Nick Pigginb3e19d92011-01-07 17:50:11 +11001277 /*
1278 * probably don't strictly need the lock here if we examined
1279 * all race cases, but it's a slowpath.
1280 */
1281 br_write_lock(vfsmount_lock);
Al Viro83adc752011-11-24 22:37:54 -05001282 if (mnt_get_count(mnt) != 2) {
J. R. Okajimabf9faa22011-02-23 16:59:49 +09001283 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 return -EBUSY;
Nick Pigginb3e19d92011-01-07 17:50:11 +11001285 }
1286 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287
Al Viro863d6842011-11-25 00:57:42 -05001288 if (!xchg(&mnt->mnt_expiry_mark, 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 return -EAGAIN;
1290 }
1291
1292 /*
1293 * If we may have to abort operations to get out of this
1294 * mount, and they will themselves hold resources we must
1295 * allow the fs to do things. In the Unix tradition of
1296 * 'Gee thats tricky lets do it in userspace' the umount_begin
1297 * might fail to complete on the first run through as other tasks
1298 * must return, and the like. Thats for the mount program to worry
1299 * about for the moment.
1300 */
1301
Al Viro42faad92008-04-24 07:21:56 -04001302 if (flags & MNT_FORCE && sb->s_op->umount_begin) {
Al Viro42faad92008-04-24 07:21:56 -04001303 sb->s_op->umount_begin(sb);
Al Viro42faad92008-04-24 07:21:56 -04001304 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305
1306 /*
1307 * No sense to grab the lock for this test, but test itself looks
1308 * somewhat bogus. Suggestions for better replacement?
1309 * Ho-hum... In principle, we might treat that as umount + switch
1310 * to rootfs. GC would eventually take care of the old vfsmount.
1311 * Actually it makes sense, especially if rootfs would contain a
1312 * /reboot - static binary that would close all descriptors and
1313 * call reboot(9). Then init(8) could umount root and exec /reboot.
1314 */
Al Viro1ab59732011-11-24 21:35:16 -05001315 if (&mnt->mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 /*
1317 * Special case for "unmounting" root ...
1318 * we just try to remount it readonly.
1319 */
1320 down_write(&sb->s_umount);
Al Viro4aa98cf2009-05-08 13:36:58 -04001321 if (!(sb->s_flags & MS_RDONLY))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 retval = do_remount_sb(sb, MS_RDONLY, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 up_write(&sb->s_umount);
1324 return retval;
1325 }
1326
Ram Pai390c6842005-11-07 17:17:51 -05001327 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001328 br_write_lock(vfsmount_lock);
Al Viro5addc5d2005-11-07 17:15:49 -05001329 event++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330
Al Viroc35038b2008-03-22 00:46:23 -04001331 if (!(flags & MNT_DETACH))
Al Viro1ab59732011-11-24 21:35:16 -05001332 shrink_submounts(mnt, &umount_list);
Al Viroc35038b2008-03-22 00:46:23 -04001333
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 retval = -EBUSY;
Ram Paia05964f2005-11-07 17:20:17 -05001335 if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) {
Al Viro1a4eeaf2011-11-25 02:19:55 -05001336 if (!list_empty(&mnt->mnt_list))
Al Viro1ab59732011-11-24 21:35:16 -05001337 umount_tree(mnt, 1, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 retval = 0;
1339 }
Nick Piggin99b7db72010-08-18 04:37:39 +10001340 br_write_unlock(vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05001341 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001342 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 return retval;
1344}
1345
1346/*
1347 * Now umount can handle mount points as well as block devices.
1348 * This is important for filesystems which use unnamed block devices.
1349 *
1350 * We now support a flag for forced unmount like the other 'big iron'
1351 * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD
1352 */
1353
Heiko Carstensbdc480e2009-01-14 14:14:12 +01001354SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355{
Al Viro2d8f3032008-07-22 09:59:21 -04001356 struct path path;
Al Viro900148d2011-11-25 00:33:11 -05001357 struct mount *mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 int retval;
Miklos Szeredidb1f05b2010-02-10 12:15:53 +01001359 int lookup_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
Miklos Szeredidb1f05b2010-02-10 12:15:53 +01001361 if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW))
1362 return -EINVAL;
1363
1364 if (!(flags & UMOUNT_NOFOLLOW))
1365 lookup_flags |= LOOKUP_FOLLOW;
1366
1367 retval = user_path_at(AT_FDCWD, name, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 if (retval)
1369 goto out;
Al Viro900148d2011-11-25 00:33:11 -05001370 mnt = real_mount(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 retval = -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -04001372 if (path.dentry != path.mnt->mnt_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 goto dput_and_out;
Al Viro143c8c92011-11-25 00:46:35 -05001374 if (!check_mnt(mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 goto dput_and_out;
1376
1377 retval = -EPERM;
1378 if (!capable(CAP_SYS_ADMIN))
1379 goto dput_and_out;
1380
Al Viro900148d2011-11-25 00:33:11 -05001381 retval = do_umount(mnt, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382dput_and_out:
Jan Blunck429731b2008-02-14 19:34:31 -08001383 /* we mustn't call path_put() as that would clear mnt_expiry_mark */
Al Viro2d8f3032008-07-22 09:59:21 -04001384 dput(path.dentry);
Al Viro900148d2011-11-25 00:33:11 -05001385 mntput_no_expire(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386out:
1387 return retval;
1388}
1389
1390#ifdef __ARCH_WANT_SYS_OLDUMOUNT
1391
1392/*
Ram Paib58fed82005-11-07 17:16:09 -05001393 * The 2.0 compatible umount. No flags.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 */
Heiko Carstensbdc480e2009-01-14 14:14:12 +01001395SYSCALL_DEFINE1(oldumount, char __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396{
Ram Paib58fed82005-11-07 17:16:09 -05001397 return sys_umount(name, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398}
1399
1400#endif
1401
Al Viro2d92ab32008-08-02 00:51:11 -04001402static int mount_is_safe(struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403{
1404 if (capable(CAP_SYS_ADMIN))
1405 return 0;
1406 return -EPERM;
1407#ifdef notyet
Al Viro2d92ab32008-08-02 00:51:11 -04001408 if (S_ISLNK(path->dentry->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 return -EPERM;
Al Viro2d92ab32008-08-02 00:51:11 -04001410 if (path->dentry->d_inode->i_mode & S_ISVTX) {
David Howellsda9592e2008-11-14 10:39:05 +11001411 if (current_uid() != path->dentry->d_inode->i_uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 return -EPERM;
1413 }
Al Viro2d92ab32008-08-02 00:51:11 -04001414 if (inode_permission(path->dentry->d_inode, MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 return -EPERM;
1416 return 0;
1417#endif
1418}
1419
Al Viro87129cc2011-11-24 21:24:27 -05001420struct mount *copy_tree(struct mount *mnt, struct dentry *dentry,
Ram Pai36341f62005-11-07 17:17:22 -05001421 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422{
Al Viroa73324d2011-11-24 22:25:07 -05001423 struct mount *res, *p, *q, *r;
Al Viro1a390682008-03-21 20:48:19 -04001424 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425
Al Virofc7be132011-11-25 01:05:37 -05001426 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt))
Ram Pai9676f0c2005-11-07 17:21:20 -05001427 return NULL;
1428
Ram Pai36341f62005-11-07 17:17:22 -05001429 res = q = clone_mnt(mnt, dentry, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 if (!q)
1431 goto Enomem;
Al Viroa73324d2011-11-24 22:25:07 -05001432 q->mnt_mountpoint = mnt->mnt_mountpoint;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
1434 p = mnt;
Al Viro6b41d532011-11-24 23:24:33 -05001435 list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) {
Al Viro315fc832011-11-24 18:57:30 -05001436 struct mount *s;
Jan Blunck7ec02ef2008-04-29 00:59:40 -07001437 if (!is_subdir(r->mnt_mountpoint, dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 continue;
1439
Al Viroa73324d2011-11-24 22:25:07 -05001440 for (s = r; s; s = next_mnt(s, &r->mnt)) {
Al Virofc7be132011-11-25 01:05:37 -05001441 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) {
Ram Pai9676f0c2005-11-07 17:21:20 -05001442 s = skip_mnt_tree(s);
1443 continue;
1444 }
Al Viro0714a532011-11-24 22:19:58 -05001445 while (p != s->mnt_parent) {
1446 p = p->mnt_parent;
1447 q = q->mnt_parent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 }
Al Viro87129cc2011-11-24 21:24:27 -05001449 p = s;
Al Virocb338d02011-11-24 20:55:08 -05001450 path.mnt = &q->mnt;
Al Viroa73324d2011-11-24 22:25:07 -05001451 path.dentry = p->mnt_mountpoint;
Al Viro87129cc2011-11-24 21:24:27 -05001452 q = clone_mnt(p, p->mnt.mnt_root, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 if (!q)
1454 goto Enomem;
Nick Piggin99b7db72010-08-18 04:37:39 +10001455 br_write_lock(vfsmount_lock);
Al Viro1a4eeaf2011-11-25 02:19:55 -05001456 list_add_tail(&q->mnt_list, &res->mnt_list);
Al Virocb338d02011-11-24 20:55:08 -05001457 attach_mnt(q, &path);
Nick Piggin99b7db72010-08-18 04:37:39 +10001458 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001459 }
1460 }
1461 return res;
Ram Paib58fed82005-11-07 17:16:09 -05001462Enomem:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 if (res) {
Ram Pai70fbcdf2005-11-07 17:17:04 -05001464 LIST_HEAD(umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001465 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05001466 umount_tree(res, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001467 br_write_unlock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001468 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469 }
1470 return NULL;
1471}
1472
Al Viro589ff872009-04-18 03:28:19 -04001473struct vfsmount *collect_mounts(struct path *path)
Al Viro8aec0802007-06-07 12:20:32 -04001474{
Al Virocb338d02011-11-24 20:55:08 -05001475 struct mount *tree;
Al Viro1a60a282008-03-22 16:19:49 -04001476 down_write(&namespace_sem);
Al Viro87129cc2011-11-24 21:24:27 -05001477 tree = copy_tree(real_mount(path->mnt), path->dentry,
1478 CL_COPY_ALL | CL_PRIVATE);
Al Viro1a60a282008-03-22 16:19:49 -04001479 up_write(&namespace_sem);
Al Virocb338d02011-11-24 20:55:08 -05001480 return tree ? &tree->mnt : NULL;
Al Viro8aec0802007-06-07 12:20:32 -04001481}
1482
1483void drop_collected_mounts(struct vfsmount *mnt)
1484{
1485 LIST_HEAD(umount_list);
Al Viro1a60a282008-03-22 16:19:49 -04001486 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001487 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05001488 umount_tree(real_mount(mnt), 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001489 br_write_unlock(vfsmount_lock);
Al Viro1a60a282008-03-22 16:19:49 -04001490 up_write(&namespace_sem);
Al Viro8aec0802007-06-07 12:20:32 -04001491 release_mounts(&umount_list);
1492}
1493
Al Viro1f707132010-01-30 22:51:25 -05001494int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
1495 struct vfsmount *root)
1496{
Al Viro1a4eeaf2011-11-25 02:19:55 -05001497 struct mount *mnt;
Al Viro1f707132010-01-30 22:51:25 -05001498 int res = f(root, arg);
1499 if (res)
1500 return res;
Al Viro1a4eeaf2011-11-25 02:19:55 -05001501 list_for_each_entry(mnt, &real_mount(root)->mnt_list, mnt_list) {
1502 res = f(&mnt->mnt, arg);
Al Viro1f707132010-01-30 22:51:25 -05001503 if (res)
1504 return res;
1505 }
1506 return 0;
1507}
1508
Al Viro4b8b21f2011-11-24 19:54:23 -05001509static void cleanup_group_ids(struct mount *mnt, struct mount *end)
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001510{
Al Viro315fc832011-11-24 18:57:30 -05001511 struct mount *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001512
Al Viro4b8b21f2011-11-24 19:54:23 -05001513 for (p = mnt; p != end; p = next_mnt(p, &mnt->mnt)) {
Al Virofc7be132011-11-25 01:05:37 -05001514 if (p->mnt_group_id && !IS_MNT_SHARED(p))
Al Viro4b8b21f2011-11-24 19:54:23 -05001515 mnt_release_group_id(p);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001516 }
1517}
1518
Al Viro4b8b21f2011-11-24 19:54:23 -05001519static int invent_group_ids(struct mount *mnt, bool recurse)
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001520{
Al Viro315fc832011-11-24 18:57:30 -05001521 struct mount *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001522
Al Viro4b8b21f2011-11-24 19:54:23 -05001523 for (p = mnt; p; p = recurse ? next_mnt(p, &mnt->mnt) : NULL) {
Al Virofc7be132011-11-25 01:05:37 -05001524 if (!p->mnt_group_id && !IS_MNT_SHARED(p)) {
Al Viro4b8b21f2011-11-24 19:54:23 -05001525 int err = mnt_alloc_group_id(p);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001526 if (err) {
Al Viro4b8b21f2011-11-24 19:54:23 -05001527 cleanup_group_ids(mnt, p);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001528 return err;
1529 }
1530 }
1531 }
1532
1533 return 0;
1534}
1535
Ram Paib90fa9a2005-11-07 17:19:50 -05001536/*
1537 * @source_mnt : mount tree to be attached
Ram Pai21444402005-11-07 17:20:03 -05001538 * @nd : place the mount tree @source_mnt is attached
1539 * @parent_nd : if non-null, detach the source_mnt from its parent and
1540 * store the parent mount and mountpoint dentry.
1541 * (done when source_mnt is moved)
Ram Paib90fa9a2005-11-07 17:19:50 -05001542 *
1543 * NOTE: in the table below explains the semantics when a source mount
1544 * of a given type is attached to a destination mount of a given type.
Ram Pai9676f0c2005-11-07 17:21:20 -05001545 * ---------------------------------------------------------------------------
1546 * | BIND MOUNT OPERATION |
1547 * |**************************************************************************
1548 * | source-->| shared | private | slave | unbindable |
1549 * | dest | | | | |
1550 * | | | | | | |
1551 * | v | | | | |
1552 * |**************************************************************************
1553 * | shared | shared (++) | shared (+) | shared(+++)| invalid |
1554 * | | | | | |
1555 * |non-shared| shared (+) | private | slave (*) | invalid |
1556 * ***************************************************************************
Ram Paib90fa9a2005-11-07 17:19:50 -05001557 * A bind operation clones the source mount and mounts the clone on the
1558 * destination mount.
1559 *
1560 * (++) the cloned mount is propagated to all the mounts in the propagation
1561 * tree of the destination mount and the cloned mount is added to
1562 * the peer group of the source mount.
1563 * (+) the cloned mount is created under the destination mount and is marked
1564 * as shared. The cloned mount is added to the peer group of the source
1565 * mount.
Ram Pai5afe0022005-11-07 17:21:01 -05001566 * (+++) the mount is propagated to all the mounts in the propagation tree
1567 * of the destination mount and the cloned mount is made slave
1568 * of the same master as that of the source mount. The cloned mount
1569 * is marked as 'shared and slave'.
1570 * (*) the cloned mount is made a slave of the same master as that of the
1571 * source mount.
1572 *
Ram Pai9676f0c2005-11-07 17:21:20 -05001573 * ---------------------------------------------------------------------------
1574 * | MOVE MOUNT OPERATION |
1575 * |**************************************************************************
1576 * | source-->| shared | private | slave | unbindable |
1577 * | dest | | | | |
1578 * | | | | | | |
1579 * | v | | | | |
1580 * |**************************************************************************
1581 * | shared | shared (+) | shared (+) | shared(+++) | invalid |
1582 * | | | | | |
1583 * |non-shared| shared (+*) | private | slave (*) | unbindable |
1584 * ***************************************************************************
Ram Pai5afe0022005-11-07 17:21:01 -05001585 *
1586 * (+) the mount is moved to the destination. And is then propagated to
1587 * all the mounts in the propagation tree of the destination mount.
Ram Pai21444402005-11-07 17:20:03 -05001588 * (+*) the mount is moved to the destination.
Ram Pai5afe0022005-11-07 17:21:01 -05001589 * (+++) the mount is moved to the destination and is then propagated to
1590 * all the mounts belonging to the destination mount's propagation tree.
1591 * the mount is marked as 'shared and slave'.
1592 * (*) the mount continues to be a slave at the new location.
Ram Paib90fa9a2005-11-07 17:19:50 -05001593 *
1594 * if the source mount is a tree, the operations explained above is
1595 * applied to each mount in the tree.
1596 * Must be called without spinlocks held, since this function can sleep
1597 * in allocations.
1598 */
Al Viro0fb54e52011-11-24 19:59:16 -05001599static int attach_recursive_mnt(struct mount *source_mnt,
Al Viro1a390682008-03-21 20:48:19 -04001600 struct path *path, struct path *parent_path)
Ram Paib90fa9a2005-11-07 17:19:50 -05001601{
1602 LIST_HEAD(tree_list);
Al Viroa8d56d82011-11-24 23:59:29 -05001603 struct mount *dest_mnt = real_mount(path->mnt);
Al Viro1a390682008-03-21 20:48:19 -04001604 struct dentry *dest_dentry = path->dentry;
Al Viro315fc832011-11-24 18:57:30 -05001605 struct mount *child, *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001606 int err;
Ram Paib90fa9a2005-11-07 17:19:50 -05001607
Al Virofc7be132011-11-25 01:05:37 -05001608 if (IS_MNT_SHARED(dest_mnt)) {
Al Viro0fb54e52011-11-24 19:59:16 -05001609 err = invent_group_ids(source_mnt, true);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001610 if (err)
1611 goto out;
1612 }
Al Viroa8d56d82011-11-24 23:59:29 -05001613 err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001614 if (err)
1615 goto out_cleanup_ids;
Ram Paib90fa9a2005-11-07 17:19:50 -05001616
Nick Piggin99b7db72010-08-18 04:37:39 +10001617 br_write_lock(vfsmount_lock);
Al Virodf1a1ad2010-01-16 12:57:40 -05001618
Al Virofc7be132011-11-25 01:05:37 -05001619 if (IS_MNT_SHARED(dest_mnt)) {
Al Viro0fb54e52011-11-24 19:59:16 -05001620 for (p = source_mnt; p; p = next_mnt(p, &source_mnt->mnt))
Al Viro0f0afb12011-11-24 20:43:10 -05001621 set_mnt_shared(p);
Ram Paib90fa9a2005-11-07 17:19:50 -05001622 }
Al Viro1a390682008-03-21 20:48:19 -04001623 if (parent_path) {
Al Viro0fb54e52011-11-24 19:59:16 -05001624 detach_mnt(source_mnt, parent_path);
1625 attach_mnt(source_mnt, path);
Al Viro143c8c92011-11-25 00:46:35 -05001626 touch_mnt_namespace(source_mnt->mnt_ns);
Ram Pai21444402005-11-07 17:20:03 -05001627 } else {
Al Viro14cf1fa2011-11-25 00:01:17 -05001628 mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt);
Al Viro0fb54e52011-11-24 19:59:16 -05001629 commit_tree(source_mnt);
Ram Pai21444402005-11-07 17:20:03 -05001630 }
Ram Paib90fa9a2005-11-07 17:19:50 -05001631
Al Viro1b8e5562011-11-24 21:01:32 -05001632 list_for_each_entry_safe(child, p, &tree_list, mnt_hash) {
1633 list_del_init(&child->mnt_hash);
Al Viro4b2619a2011-11-24 19:47:15 -05001634 commit_tree(child);
Ram Paib90fa9a2005-11-07 17:19:50 -05001635 }
Nick Piggin99b7db72010-08-18 04:37:39 +10001636 br_write_unlock(vfsmount_lock);
1637
Ram Paib90fa9a2005-11-07 17:19:50 -05001638 return 0;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001639
1640 out_cleanup_ids:
Al Virofc7be132011-11-25 01:05:37 -05001641 if (IS_MNT_SHARED(dest_mnt))
Al Viro0fb54e52011-11-24 19:59:16 -05001642 cleanup_group_ids(source_mnt, NULL);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001643 out:
1644 return err;
Ram Paib90fa9a2005-11-07 17:19:50 -05001645}
1646
Al Virob12cea92011-03-18 08:55:38 -04001647static int lock_mount(struct path *path)
1648{
1649 struct vfsmount *mnt;
1650retry:
1651 mutex_lock(&path->dentry->d_inode->i_mutex);
1652 if (unlikely(cant_mount(path->dentry))) {
1653 mutex_unlock(&path->dentry->d_inode->i_mutex);
1654 return -ENOENT;
1655 }
1656 down_write(&namespace_sem);
1657 mnt = lookup_mnt(path);
1658 if (likely(!mnt))
1659 return 0;
1660 up_write(&namespace_sem);
1661 mutex_unlock(&path->dentry->d_inode->i_mutex);
1662 path_put(path);
1663 path->mnt = mnt;
1664 path->dentry = dget(mnt->mnt_root);
1665 goto retry;
1666}
1667
1668static void unlock_mount(struct path *path)
1669{
1670 up_write(&namespace_sem);
1671 mutex_unlock(&path->dentry->d_inode->i_mutex);
1672}
1673
Al Viro95bc5f22011-11-25 00:30:56 -05001674static int graft_tree(struct mount *mnt, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675{
Al Viro95bc5f22011-11-25 00:30:56 -05001676 if (mnt->mnt.mnt_sb->s_flags & MS_NOUSER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 return -EINVAL;
1678
Al Viro8c3ee422008-03-22 18:00:39 -04001679 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
Al Viro95bc5f22011-11-25 00:30:56 -05001680 S_ISDIR(mnt->mnt.mnt_root->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681 return -ENOTDIR;
1682
Al Virob12cea92011-03-18 08:55:38 -04001683 if (d_unlinked(path->dentry))
1684 return -ENOENT;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685
Al Viro95bc5f22011-11-25 00:30:56 -05001686 return attach_recursive_mnt(mnt, path, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687}
1688
1689/*
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001690 * Sanity check the flags to change_mnt_propagation.
1691 */
1692
1693static int flags_to_propagation_type(int flags)
1694{
Roman Borisov7c6e9842011-05-25 16:26:48 -07001695 int type = flags & ~(MS_REC | MS_SILENT);
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001696
1697 /* Fail if any non-propagation flags are set */
1698 if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
1699 return 0;
1700 /* Only one propagation flag should be set */
1701 if (!is_power_of_2(type))
1702 return 0;
1703 return type;
1704}
1705
1706/*
Ram Pai07b20882005-11-07 17:19:07 -05001707 * recursively change the type of the mountpoint.
1708 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001709static int do_change_type(struct path *path, int flag)
Ram Pai07b20882005-11-07 17:19:07 -05001710{
Al Viro315fc832011-11-24 18:57:30 -05001711 struct mount *m;
Al Viro4b8b21f2011-11-24 19:54:23 -05001712 struct mount *mnt = real_mount(path->mnt);
Ram Pai07b20882005-11-07 17:19:07 -05001713 int recurse = flag & MS_REC;
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001714 int type;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001715 int err = 0;
Ram Pai07b20882005-11-07 17:19:07 -05001716
Miklos Szerediee6f9582007-05-08 00:30:40 -07001717 if (!capable(CAP_SYS_ADMIN))
1718 return -EPERM;
1719
Al Viro2d92ab32008-08-02 00:51:11 -04001720 if (path->dentry != path->mnt->mnt_root)
Ram Pai07b20882005-11-07 17:19:07 -05001721 return -EINVAL;
1722
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001723 type = flags_to_propagation_type(flag);
1724 if (!type)
1725 return -EINVAL;
1726
Ram Pai07b20882005-11-07 17:19:07 -05001727 down_write(&namespace_sem);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001728 if (type == MS_SHARED) {
1729 err = invent_group_ids(mnt, recurse);
1730 if (err)
1731 goto out_unlock;
1732 }
1733
Nick Piggin99b7db72010-08-18 04:37:39 +10001734 br_write_lock(vfsmount_lock);
Al Viro4b8b21f2011-11-24 19:54:23 -05001735 for (m = mnt; m; m = (recurse ? next_mnt(m, &mnt->mnt) : NULL))
Al Viro0f0afb12011-11-24 20:43:10 -05001736 change_mnt_propagation(m, type);
Nick Piggin99b7db72010-08-18 04:37:39 +10001737 br_write_unlock(vfsmount_lock);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001738
1739 out_unlock:
Ram Pai07b20882005-11-07 17:19:07 -05001740 up_write(&namespace_sem);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001741 return err;
Ram Pai07b20882005-11-07 17:19:07 -05001742}
1743
1744/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001745 * do loopback mount.
1746 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001747static int do_loopback(struct path *path, char *old_name,
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001748 int recurse)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749{
Al Virob12cea92011-03-18 08:55:38 -04001750 LIST_HEAD(umount_list);
Al Viro2d92ab32008-08-02 00:51:11 -04001751 struct path old_path;
Al Viro87129cc2011-11-24 21:24:27 -05001752 struct mount *mnt = NULL, *old;
Al Viro2d92ab32008-08-02 00:51:11 -04001753 int err = mount_is_safe(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 if (err)
1755 return err;
1756 if (!old_name || !*old_name)
1757 return -EINVAL;
Trond Myklebust815d4052011-09-26 20:36:09 -04001758 err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 if (err)
1760 return err;
1761
Al Virob12cea92011-03-18 08:55:38 -04001762 err = lock_mount(path);
1763 if (err)
Jan Blunck4ac91372008-02-14 19:34:32 -08001764 goto out;
Ram Pai9676f0c2005-11-07 17:21:20 -05001765
Al Viro87129cc2011-11-24 21:24:27 -05001766 old = real_mount(old_path.mnt);
1767
Al Virob12cea92011-03-18 08:55:38 -04001768 err = -EINVAL;
Al Virofc7be132011-11-25 01:05:37 -05001769 if (IS_MNT_UNBINDABLE(old))
Al Virob12cea92011-03-18 08:55:38 -04001770 goto out2;
1771
Al Viro143c8c92011-11-25 00:46:35 -05001772 if (!check_mnt(real_mount(path->mnt)) || !check_mnt(old))
Al Virob12cea92011-03-18 08:55:38 -04001773 goto out2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774
Al Viroccd48bc2005-11-07 17:15:04 -05001775 err = -ENOMEM;
1776 if (recurse)
Al Viro87129cc2011-11-24 21:24:27 -05001777 mnt = copy_tree(old, old_path.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -05001778 else
Al Viro87129cc2011-11-24 21:24:27 -05001779 mnt = clone_mnt(old, old_path.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -05001780
1781 if (!mnt)
Al Virob12cea92011-03-18 08:55:38 -04001782 goto out2;
Al Viroccd48bc2005-11-07 17:15:04 -05001783
Al Viro95bc5f22011-11-25 00:30:56 -05001784 err = graft_tree(mnt, path);
Al Viroccd48bc2005-11-07 17:15:04 -05001785 if (err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001786 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05001787 umount_tree(mnt, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001788 br_write_unlock(vfsmount_lock);
Ram Pai5b83d2c5c2005-11-07 17:16:29 -05001789 }
Al Virob12cea92011-03-18 08:55:38 -04001790out2:
1791 unlock_mount(path);
1792 release_mounts(&umount_list);
Al Viroccd48bc2005-11-07 17:15:04 -05001793out:
Al Viro2d92ab32008-08-02 00:51:11 -04001794 path_put(&old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 return err;
1796}
1797
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001798static int change_mount_flags(struct vfsmount *mnt, int ms_flags)
1799{
1800 int error = 0;
1801 int readonly_request = 0;
1802
1803 if (ms_flags & MS_RDONLY)
1804 readonly_request = 1;
1805 if (readonly_request == __mnt_is_readonly(mnt))
1806 return 0;
1807
1808 if (readonly_request)
Al Viro83adc752011-11-24 22:37:54 -05001809 error = mnt_make_readonly(real_mount(mnt));
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001810 else
Al Viro83adc752011-11-24 22:37:54 -05001811 __mnt_unmake_readonly(real_mount(mnt));
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001812 return error;
1813}
1814
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815/*
1816 * change filesystem flags. dir should be a physical root of filesystem.
1817 * If you've mounted a non-root directory somewhere and want to do remount
1818 * on it - tough luck.
1819 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001820static int do_remount(struct path *path, int flags, int mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 void *data)
1822{
1823 int err;
Al Viro2d92ab32008-08-02 00:51:11 -04001824 struct super_block *sb = path->mnt->mnt_sb;
Al Viro143c8c92011-11-25 00:46:35 -05001825 struct mount *mnt = real_mount(path->mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826
1827 if (!capable(CAP_SYS_ADMIN))
1828 return -EPERM;
1829
Al Viro143c8c92011-11-25 00:46:35 -05001830 if (!check_mnt(mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831 return -EINVAL;
1832
Al Viro2d92ab32008-08-02 00:51:11 -04001833 if (path->dentry != path->mnt->mnt_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 return -EINVAL;
1835
Eric Parisff36fe22011-03-03 16:09:14 -05001836 err = security_sb_remount(sb, data);
1837 if (err)
1838 return err;
1839
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 down_write(&sb->s_umount);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001841 if (flags & MS_BIND)
Al Viro2d92ab32008-08-02 00:51:11 -04001842 err = change_mount_flags(path->mnt, flags);
Al Viro4aa98cf2009-05-08 13:36:58 -04001843 else
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001844 err = do_remount_sb(sb, flags, data, 0);
Al Viro7b43a792010-01-16 13:01:26 -05001845 if (!err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001846 br_write_lock(vfsmount_lock);
Al Viro143c8c92011-11-25 00:46:35 -05001847 mnt_flags |= mnt->mnt.mnt_flags & MNT_PROPAGATION_MASK;
1848 mnt->mnt.mnt_flags = mnt_flags;
Nick Piggin99b7db72010-08-18 04:37:39 +10001849 br_write_unlock(vfsmount_lock);
Al Viro7b43a792010-01-16 13:01:26 -05001850 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 up_write(&sb->s_umount);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001852 if (!err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001853 br_write_lock(vfsmount_lock);
Al Viro143c8c92011-11-25 00:46:35 -05001854 touch_mnt_namespace(mnt->mnt_ns);
Nick Piggin99b7db72010-08-18 04:37:39 +10001855 br_write_unlock(vfsmount_lock);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001856 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 return err;
1858}
1859
Al Virocbbe3622011-11-24 20:01:19 -05001860static inline int tree_contains_unbindable(struct mount *mnt)
Ram Pai9676f0c2005-11-07 17:21:20 -05001861{
Al Viro315fc832011-11-24 18:57:30 -05001862 struct mount *p;
Al Virocbbe3622011-11-24 20:01:19 -05001863 for (p = mnt; p; p = next_mnt(p, &mnt->mnt)) {
Al Virofc7be132011-11-25 01:05:37 -05001864 if (IS_MNT_UNBINDABLE(p))
Ram Pai9676f0c2005-11-07 17:21:20 -05001865 return 1;
1866 }
1867 return 0;
1868}
1869
Al Viro0a0d8a42008-08-02 00:55:27 -04001870static int do_move_mount(struct path *path, char *old_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871{
Al Viro2d92ab32008-08-02 00:51:11 -04001872 struct path old_path, parent_path;
Al Viro676da582011-11-24 21:47:05 -05001873 struct mount *p;
Al Viro0fb54e52011-11-24 19:59:16 -05001874 struct mount *old;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 int err = 0;
1876 if (!capable(CAP_SYS_ADMIN))
1877 return -EPERM;
1878 if (!old_name || !*old_name)
1879 return -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001880 err = kern_path(old_name, LOOKUP_FOLLOW, &old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881 if (err)
1882 return err;
1883
Al Virob12cea92011-03-18 08:55:38 -04001884 err = lock_mount(path);
David Howellscc53ce52011-01-14 18:45:26 +00001885 if (err < 0)
1886 goto out;
1887
Al Viro143c8c92011-11-25 00:46:35 -05001888 old = real_mount(old_path.mnt);
Al Virofc7be132011-11-25 01:05:37 -05001889 p = real_mount(path->mnt);
Al Viro143c8c92011-11-25 00:46:35 -05001890
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 err = -EINVAL;
Al Virofc7be132011-11-25 01:05:37 -05001892 if (!check_mnt(p) || !check_mnt(old))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 goto out1;
1894
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04001895 if (d_unlinked(path->dentry))
Ram Pai21444402005-11-07 17:20:03 -05001896 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897
1898 err = -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001899 if (old_path.dentry != old_path.mnt->mnt_root)
Ram Pai21444402005-11-07 17:20:03 -05001900 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901
Al Viro676da582011-11-24 21:47:05 -05001902 if (!mnt_has_parent(old))
Ram Pai21444402005-11-07 17:20:03 -05001903 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904
Al Viro2d92ab32008-08-02 00:51:11 -04001905 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
1906 S_ISDIR(old_path.dentry->d_inode->i_mode))
Ram Pai21444402005-11-07 17:20:03 -05001907 goto out1;
1908 /*
1909 * Don't move a mount residing in a shared parent.
1910 */
Al Virofc7be132011-11-25 01:05:37 -05001911 if (IS_MNT_SHARED(old->mnt_parent))
Ram Pai21444402005-11-07 17:20:03 -05001912 goto out1;
Ram Pai9676f0c2005-11-07 17:21:20 -05001913 /*
1914 * Don't move a mount tree containing unbindable mounts to a destination
1915 * mount which is shared.
1916 */
Al Virofc7be132011-11-25 01:05:37 -05001917 if (IS_MNT_SHARED(p) && tree_contains_unbindable(old))
Ram Pai9676f0c2005-11-07 17:21:20 -05001918 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 err = -ELOOP;
Al Virofc7be132011-11-25 01:05:37 -05001920 for (; mnt_has_parent(p); p = p->mnt_parent)
Al Viro676da582011-11-24 21:47:05 -05001921 if (p == old)
Ram Pai21444402005-11-07 17:20:03 -05001922 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923
Al Viro0fb54e52011-11-24 19:59:16 -05001924 err = attach_recursive_mnt(old, path, &parent_path);
Jan Blunck4ac91372008-02-14 19:34:32 -08001925 if (err)
Ram Pai21444402005-11-07 17:20:03 -05001926 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927
1928 /* if the mount is moved, it should no longer be expire
1929 * automatically */
Al Viro6776db3d2011-11-25 00:22:05 -05001930 list_del_init(&old->mnt_expire);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931out1:
Al Virob12cea92011-03-18 08:55:38 -04001932 unlock_mount(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 if (!err)
Al Viro1a390682008-03-21 20:48:19 -04001935 path_put(&parent_path);
Al Viro2d92ab32008-08-02 00:51:11 -04001936 path_put(&old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937 return err;
1938}
1939
Al Viro9d412a42011-03-17 22:08:28 -04001940static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype)
1941{
1942 int err;
1943 const char *subtype = strchr(fstype, '.');
1944 if (subtype) {
1945 subtype++;
1946 err = -EINVAL;
1947 if (!subtype[0])
1948 goto err;
1949 } else
1950 subtype = "";
1951
1952 mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL);
1953 err = -ENOMEM;
1954 if (!mnt->mnt_sb->s_subtype)
1955 goto err;
1956 return mnt;
1957
1958 err:
1959 mntput(mnt);
1960 return ERR_PTR(err);
1961}
1962
Al Viro79e801a2011-12-12 23:07:05 -05001963static struct vfsmount *
Al Viro9d412a42011-03-17 22:08:28 -04001964do_kern_mount(const char *fstype, int flags, const char *name, void *data)
1965{
1966 struct file_system_type *type = get_fs_type(fstype);
1967 struct vfsmount *mnt;
1968 if (!type)
1969 return ERR_PTR(-ENODEV);
1970 mnt = vfs_kern_mount(type, flags, name, data);
1971 if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) &&
1972 !mnt->mnt_sb->s_subtype)
1973 mnt = fs_set_subtype(mnt, fstype);
1974 put_filesystem(type);
1975 return mnt;
1976}
Al Viro9d412a42011-03-17 22:08:28 -04001977
1978/*
1979 * add a mount into a namespace's mount tree
1980 */
Al Viro95bc5f22011-11-25 00:30:56 -05001981static int do_add_mount(struct mount *newmnt, struct path *path, int mnt_flags)
Al Viro9d412a42011-03-17 22:08:28 -04001982{
1983 int err;
1984
1985 mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL);
1986
Al Virob12cea92011-03-18 08:55:38 -04001987 err = lock_mount(path);
1988 if (err)
1989 return err;
Al Viro9d412a42011-03-17 22:08:28 -04001990
1991 err = -EINVAL;
Al Viro143c8c92011-11-25 00:46:35 -05001992 if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(real_mount(path->mnt)))
Al Viro9d412a42011-03-17 22:08:28 -04001993 goto unlock;
1994
1995 /* Refuse the same filesystem on the same mount point */
1996 err = -EBUSY;
Al Viro95bc5f22011-11-25 00:30:56 -05001997 if (path->mnt->mnt_sb == newmnt->mnt.mnt_sb &&
Al Viro9d412a42011-03-17 22:08:28 -04001998 path->mnt->mnt_root == path->dentry)
1999 goto unlock;
2000
2001 err = -EINVAL;
Al Viro95bc5f22011-11-25 00:30:56 -05002002 if (S_ISLNK(newmnt->mnt.mnt_root->d_inode->i_mode))
Al Viro9d412a42011-03-17 22:08:28 -04002003 goto unlock;
2004
Al Viro95bc5f22011-11-25 00:30:56 -05002005 newmnt->mnt.mnt_flags = mnt_flags;
Al Viro9d412a42011-03-17 22:08:28 -04002006 err = graft_tree(newmnt, path);
2007
2008unlock:
Al Virob12cea92011-03-18 08:55:38 -04002009 unlock_mount(path);
Al Viro9d412a42011-03-17 22:08:28 -04002010 return err;
2011}
Al Virob1e75df2011-01-17 01:47:59 -05002012
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013/*
2014 * create a new mount for userspace and request it to be added into the
2015 * namespace's tree
2016 */
Al Viro0a0d8a42008-08-02 00:55:27 -04002017static int do_new_mount(struct path *path, char *type, int flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018 int mnt_flags, char *name, void *data)
2019{
2020 struct vfsmount *mnt;
Al Viro15f9a3f2011-01-17 01:41:58 -05002021 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022
Vegard Nossumeca6f532009-09-18 13:05:45 -07002023 if (!type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002024 return -EINVAL;
2025
2026 /* we need capabilities... */
2027 if (!capable(CAP_SYS_ADMIN))
2028 return -EPERM;
2029
2030 mnt = do_kern_mount(type, flags, name, data);
2031 if (IS_ERR(mnt))
2032 return PTR_ERR(mnt);
2033
Al Viro95bc5f22011-11-25 00:30:56 -05002034 err = do_add_mount(real_mount(mnt), path, mnt_flags);
Al Viro15f9a3f2011-01-17 01:41:58 -05002035 if (err)
2036 mntput(mnt);
2037 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038}
2039
Al Viro19a167a2011-01-17 01:35:23 -05002040int finish_automount(struct vfsmount *m, struct path *path)
2041{
Al Viro6776db3d2011-11-25 00:22:05 -05002042 struct mount *mnt = real_mount(m);
Al Viro19a167a2011-01-17 01:35:23 -05002043 int err;
2044 /* The new mount record should have at least 2 refs to prevent it being
2045 * expired before we get a chance to add it
2046 */
Al Viro6776db3d2011-11-25 00:22:05 -05002047 BUG_ON(mnt_get_count(mnt) < 2);
Al Viro19a167a2011-01-17 01:35:23 -05002048
2049 if (m->mnt_sb == path->mnt->mnt_sb &&
2050 m->mnt_root == path->dentry) {
Al Virob1e75df2011-01-17 01:47:59 -05002051 err = -ELOOP;
2052 goto fail;
Al Viro19a167a2011-01-17 01:35:23 -05002053 }
2054
Al Viro95bc5f22011-11-25 00:30:56 -05002055 err = do_add_mount(mnt, path, path->mnt->mnt_flags | MNT_SHRINKABLE);
Al Virob1e75df2011-01-17 01:47:59 -05002056 if (!err)
2057 return 0;
2058fail:
2059 /* remove m from any expiration list it may be on */
Al Viro6776db3d2011-11-25 00:22:05 -05002060 if (!list_empty(&mnt->mnt_expire)) {
Al Virob1e75df2011-01-17 01:47:59 -05002061 down_write(&namespace_sem);
2062 br_write_lock(vfsmount_lock);
Al Viro6776db3d2011-11-25 00:22:05 -05002063 list_del_init(&mnt->mnt_expire);
Al Virob1e75df2011-01-17 01:47:59 -05002064 br_write_unlock(vfsmount_lock);
2065 up_write(&namespace_sem);
Al Viro19a167a2011-01-17 01:35:23 -05002066 }
Al Virob1e75df2011-01-17 01:47:59 -05002067 mntput(m);
2068 mntput(m);
Al Viro19a167a2011-01-17 01:35:23 -05002069 return err;
2070}
2071
David Howellsea5b7782011-01-14 19:10:03 +00002072/**
2073 * mnt_set_expiry - Put a mount on an expiration list
2074 * @mnt: The mount to list.
2075 * @expiry_list: The list to add the mount to.
2076 */
2077void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list)
2078{
2079 down_write(&namespace_sem);
2080 br_write_lock(vfsmount_lock);
2081
Al Viro6776db3d2011-11-25 00:22:05 -05002082 list_add_tail(&real_mount(mnt)->mnt_expire, expiry_list);
David Howellsea5b7782011-01-14 19:10:03 +00002083
2084 br_write_unlock(vfsmount_lock);
2085 up_write(&namespace_sem);
2086}
2087EXPORT_SYMBOL(mnt_set_expiry);
2088
2089/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 * process a list of expirable mountpoints with the intent of discarding any
2091 * mountpoints that aren't in use and haven't been touched since last we came
2092 * here
2093 */
2094void mark_mounts_for_expiry(struct list_head *mounts)
2095{
Al Viro761d5c32011-11-24 21:07:43 -05002096 struct mount *mnt, *next;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 LIST_HEAD(graveyard);
Al Virobcc5c7d2008-03-22 00:21:53 -04002098 LIST_HEAD(umounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099
2100 if (list_empty(mounts))
2101 return;
2102
Al Virobcc5c7d2008-03-22 00:21:53 -04002103 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10002104 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002105
2106 /* extract from the expiration list every vfsmount that matches the
2107 * following criteria:
2108 * - only referenced by its parent vfsmount
2109 * - still marked for expiry (marked on the last call here; marks are
2110 * cleared by mntput())
2111 */
Al Viro6776db3d2011-11-25 00:22:05 -05002112 list_for_each_entry_safe(mnt, next, mounts, mnt_expire) {
Al Viro863d6842011-11-25 00:57:42 -05002113 if (!xchg(&mnt->mnt_expiry_mark, 1) ||
Al Viro1ab59732011-11-24 21:35:16 -05002114 propagate_mount_busy(mnt, 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 continue;
Al Viro6776db3d2011-11-25 00:22:05 -05002116 list_move(&mnt->mnt_expire, &graveyard);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117 }
Al Virobcc5c7d2008-03-22 00:21:53 -04002118 while (!list_empty(&graveyard)) {
Al Viro6776db3d2011-11-25 00:22:05 -05002119 mnt = list_first_entry(&graveyard, struct mount, mnt_expire);
Al Viro143c8c92011-11-25 00:46:35 -05002120 touch_mnt_namespace(mnt->mnt_ns);
Al Virobcc5c7d2008-03-22 00:21:53 -04002121 umount_tree(mnt, 1, &umounts);
2122 }
Nick Piggin99b7db72010-08-18 04:37:39 +10002123 br_write_unlock(vfsmount_lock);
Al Virobcc5c7d2008-03-22 00:21:53 -04002124 up_write(&namespace_sem);
2125
2126 release_mounts(&umounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127}
2128
2129EXPORT_SYMBOL_GPL(mark_mounts_for_expiry);
2130
2131/*
Trond Myklebust5528f9112006-06-09 09:34:17 -04002132 * Ripoff of 'select_parent()'
2133 *
2134 * search the list of submounts for a given mountpoint, and move any
2135 * shrinkable submounts to the 'graveyard' list.
2136 */
Al Viro692afc32011-11-24 21:15:14 -05002137static int select_submounts(struct mount *parent, struct list_head *graveyard)
Trond Myklebust5528f9112006-06-09 09:34:17 -04002138{
Al Viro692afc32011-11-24 21:15:14 -05002139 struct mount *this_parent = parent;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002140 struct list_head *next;
2141 int found = 0;
2142
2143repeat:
Al Viro6b41d532011-11-24 23:24:33 -05002144 next = this_parent->mnt_mounts.next;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002145resume:
Al Viro6b41d532011-11-24 23:24:33 -05002146 while (next != &this_parent->mnt_mounts) {
Trond Myklebust5528f9112006-06-09 09:34:17 -04002147 struct list_head *tmp = next;
Al Viro6b41d532011-11-24 23:24:33 -05002148 struct mount *mnt = list_entry(tmp, struct mount, mnt_child);
Trond Myklebust5528f9112006-06-09 09:34:17 -04002149
2150 next = tmp->next;
Al Viro692afc32011-11-24 21:15:14 -05002151 if (!(mnt->mnt.mnt_flags & MNT_SHRINKABLE))
Trond Myklebust5528f9112006-06-09 09:34:17 -04002152 continue;
2153 /*
2154 * Descend a level if the d_mounts list is non-empty.
2155 */
Al Viro6b41d532011-11-24 23:24:33 -05002156 if (!list_empty(&mnt->mnt_mounts)) {
Trond Myklebust5528f9112006-06-09 09:34:17 -04002157 this_parent = mnt;
2158 goto repeat;
2159 }
2160
Al Viro1ab59732011-11-24 21:35:16 -05002161 if (!propagate_mount_busy(mnt, 1)) {
Al Viro6776db3d2011-11-25 00:22:05 -05002162 list_move_tail(&mnt->mnt_expire, graveyard);
Trond Myklebust5528f9112006-06-09 09:34:17 -04002163 found++;
2164 }
2165 }
2166 /*
2167 * All done at this level ... ascend and resume the search
2168 */
2169 if (this_parent != parent) {
Al Viro6b41d532011-11-24 23:24:33 -05002170 next = this_parent->mnt_child.next;
Al Viro0714a532011-11-24 22:19:58 -05002171 this_parent = this_parent->mnt_parent;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002172 goto resume;
2173 }
2174 return found;
2175}
2176
2177/*
2178 * process a list of expirable mountpoints with the intent of discarding any
2179 * submounts of a specific parent mountpoint
Nick Piggin99b7db72010-08-18 04:37:39 +10002180 *
2181 * vfsmount_lock must be held for write
Trond Myklebust5528f9112006-06-09 09:34:17 -04002182 */
Al Viro692afc32011-11-24 21:15:14 -05002183static void shrink_submounts(struct mount *mnt, struct list_head *umounts)
Trond Myklebust5528f9112006-06-09 09:34:17 -04002184{
2185 LIST_HEAD(graveyard);
Al Viro761d5c32011-11-24 21:07:43 -05002186 struct mount *m;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002187
Trond Myklebust5528f9112006-06-09 09:34:17 -04002188 /* extract submounts of 'mountpoint' from the expiration list */
Al Viroc35038b2008-03-22 00:46:23 -04002189 while (select_submounts(mnt, &graveyard)) {
Al Virobcc5c7d2008-03-22 00:21:53 -04002190 while (!list_empty(&graveyard)) {
Al Viro761d5c32011-11-24 21:07:43 -05002191 m = list_first_entry(&graveyard, struct mount,
Al Viro6776db3d2011-11-25 00:22:05 -05002192 mnt_expire);
Al Viro143c8c92011-11-25 00:46:35 -05002193 touch_mnt_namespace(m->mnt_ns);
Eric W. Biedermanafef80b2008-11-12 13:26:54 -08002194 umount_tree(m, 1, umounts);
Al Virobcc5c7d2008-03-22 00:21:53 -04002195 }
2196 }
Trond Myklebust5528f9112006-06-09 09:34:17 -04002197}
2198
Trond Myklebust5528f9112006-06-09 09:34:17 -04002199/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200 * Some copy_from_user() implementations do not return the exact number of
2201 * bytes remaining to copy on a fault. But copy_mount_options() requires that.
2202 * Note that this function differs from copy_from_user() in that it will oops
2203 * on bad values of `to', rather than returning a short copy.
2204 */
Ram Paib58fed82005-11-07 17:16:09 -05002205static long exact_copy_from_user(void *to, const void __user * from,
2206 unsigned long n)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207{
2208 char *t = to;
2209 const char __user *f = from;
2210 char c;
2211
2212 if (!access_ok(VERIFY_READ, from, n))
2213 return n;
2214
2215 while (n) {
2216 if (__get_user(c, f)) {
2217 memset(t, 0, n);
2218 break;
2219 }
2220 *t++ = c;
2221 f++;
2222 n--;
2223 }
2224 return n;
2225}
2226
Ram Paib58fed82005-11-07 17:16:09 -05002227int copy_mount_options(const void __user * data, unsigned long *where)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002228{
2229 int i;
2230 unsigned long page;
2231 unsigned long size;
Ram Paib58fed82005-11-07 17:16:09 -05002232
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 *where = 0;
2234 if (!data)
2235 return 0;
2236
2237 if (!(page = __get_free_page(GFP_KERNEL)))
2238 return -ENOMEM;
2239
2240 /* We only care that *some* data at the address the user
2241 * gave us is valid. Just in case, we'll zero
2242 * the remainder of the page.
2243 */
2244 /* copy_from_user cannot cross TASK_SIZE ! */
2245 size = TASK_SIZE - (unsigned long)data;
2246 if (size > PAGE_SIZE)
2247 size = PAGE_SIZE;
2248
2249 i = size - exact_copy_from_user((void *)page, data, size);
2250 if (!i) {
Ram Paib58fed82005-11-07 17:16:09 -05002251 free_page(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252 return -EFAULT;
2253 }
2254 if (i != PAGE_SIZE)
2255 memset((char *)page + i, 0, PAGE_SIZE - i);
2256 *where = page;
2257 return 0;
2258}
2259
Vegard Nossumeca6f532009-09-18 13:05:45 -07002260int copy_mount_string(const void __user *data, char **where)
2261{
2262 char *tmp;
2263
2264 if (!data) {
2265 *where = NULL;
2266 return 0;
2267 }
2268
2269 tmp = strndup_user(data, PAGE_SIZE);
2270 if (IS_ERR(tmp))
2271 return PTR_ERR(tmp);
2272
2273 *where = tmp;
2274 return 0;
2275}
2276
Linus Torvalds1da177e2005-04-16 15:20:36 -07002277/*
2278 * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to
2279 * be given to the mount() call (ie: read-only, no-dev, no-suid etc).
2280 *
2281 * data is a (void *) that can point to any structure up to
2282 * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent
2283 * information (or be NULL).
2284 *
2285 * Pre-0.97 versions of mount() didn't have a flags word.
2286 * When the flags word was introduced its top half was required
2287 * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9.
2288 * Therefore, if this magic number is present, it carries no information
2289 * and must be discarded.
2290 */
Ram Paib58fed82005-11-07 17:16:09 -05002291long do_mount(char *dev_name, char *dir_name, char *type_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292 unsigned long flags, void *data_page)
2293{
Al Viro2d92ab32008-08-02 00:51:11 -04002294 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295 int retval = 0;
2296 int mnt_flags = 0;
2297
2298 /* Discard magic */
2299 if ((flags & MS_MGC_MSK) == MS_MGC_VAL)
2300 flags &= ~MS_MGC_MSK;
2301
2302 /* Basic sanity checks */
2303
2304 if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE))
2305 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306
2307 if (data_page)
2308 ((char *)data_page)[PAGE_SIZE - 1] = 0;
2309
Tetsuo Handaa27ab9f22009-10-04 21:49:49 +09002310 /* ... and get the mountpoint */
2311 retval = kern_path(dir_name, LOOKUP_FOLLOW, &path);
2312 if (retval)
2313 return retval;
2314
2315 retval = security_sb_mount(dev_name, &path,
2316 type_page, flags, data_page);
2317 if (retval)
2318 goto dput_out;
2319
Andi Kleen613cbe32009-04-19 18:40:43 +02002320 /* Default to relatime unless overriden */
2321 if (!(flags & MS_NOATIME))
2322 mnt_flags |= MNT_RELATIME;
Matthew Garrett0a1c01c2009-03-26 17:53:14 +00002323
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324 /* Separate the per-mountpoint flags */
2325 if (flags & MS_NOSUID)
2326 mnt_flags |= MNT_NOSUID;
2327 if (flags & MS_NODEV)
2328 mnt_flags |= MNT_NODEV;
2329 if (flags & MS_NOEXEC)
2330 mnt_flags |= MNT_NOEXEC;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08002331 if (flags & MS_NOATIME)
2332 mnt_flags |= MNT_NOATIME;
2333 if (flags & MS_NODIRATIME)
2334 mnt_flags |= MNT_NODIRATIME;
Matthew Garrettd0adde52009-03-26 17:49:56 +00002335 if (flags & MS_STRICTATIME)
2336 mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08002337 if (flags & MS_RDONLY)
2338 mnt_flags |= MNT_READONLY;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08002339
Al Viro7a4dec52010-08-09 12:05:43 -04002340 flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN |
Matthew Garrettd0adde52009-03-26 17:49:56 +00002341 MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT |
2342 MS_STRICTATIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 if (flags & MS_REMOUNT)
Al Viro2d92ab32008-08-02 00:51:11 -04002345 retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346 data_page);
2347 else if (flags & MS_BIND)
Al Viro2d92ab32008-08-02 00:51:11 -04002348 retval = do_loopback(&path, dev_name, flags & MS_REC);
Ram Pai9676f0c2005-11-07 17:21:20 -05002349 else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
Al Viro2d92ab32008-08-02 00:51:11 -04002350 retval = do_change_type(&path, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 else if (flags & MS_MOVE)
Al Viro2d92ab32008-08-02 00:51:11 -04002352 retval = do_move_mount(&path, dev_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353 else
Al Viro2d92ab32008-08-02 00:51:11 -04002354 retval = do_new_mount(&path, type_page, flags, mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355 dev_name, data_page);
2356dput_out:
Al Viro2d92ab32008-08-02 00:51:11 -04002357 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 return retval;
2359}
2360
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002361static struct mnt_namespace *alloc_mnt_ns(void)
2362{
2363 struct mnt_namespace *new_ns;
2364
2365 new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL);
2366 if (!new_ns)
2367 return ERR_PTR(-ENOMEM);
2368 atomic_set(&new_ns->count, 1);
2369 new_ns->root = NULL;
2370 INIT_LIST_HEAD(&new_ns->list);
2371 init_waitqueue_head(&new_ns->poll);
2372 new_ns->event = 0;
2373 return new_ns;
2374}
2375
Al Virof03c6592011-01-14 22:30:21 -05002376void mnt_make_longterm(struct vfsmount *mnt)
2377{
Al Viro83adc752011-11-24 22:37:54 -05002378 __mnt_make_longterm(real_mount(mnt));
Al Virof03c6592011-01-14 22:30:21 -05002379}
2380
Al Viro83adc752011-11-24 22:37:54 -05002381void mnt_make_shortterm(struct vfsmount *m)
Al Virof03c6592011-01-14 22:30:21 -05002382{
Al Viro7e3d0eb02011-01-16 16:32:11 -05002383#ifdef CONFIG_SMP
Al Viro83adc752011-11-24 22:37:54 -05002384 struct mount *mnt = real_mount(m);
Al Viro68e8a9f2011-11-24 22:53:09 -05002385 if (atomic_add_unless(&mnt->mnt_longterm, -1, 1))
Al Virof03c6592011-01-14 22:30:21 -05002386 return;
2387 br_write_lock(vfsmount_lock);
Al Viro68e8a9f2011-11-24 22:53:09 -05002388 atomic_dec(&mnt->mnt_longterm);
Al Virof03c6592011-01-14 22:30:21 -05002389 br_write_unlock(vfsmount_lock);
Al Viro7e3d0eb02011-01-16 16:32:11 -05002390#endif
Al Virof03c6592011-01-14 22:30:21 -05002391}
2392
JANAK DESAI741a2952006-02-07 12:59:00 -08002393/*
2394 * Allocate a new namespace structure and populate it with contents
2395 * copied from the namespace of the passed in task structure.
2396 */
Badari Pulavartye3222c42007-05-08 00:25:21 -07002397static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002398 struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002400 struct mnt_namespace *new_ns;
Al Viro7f2da1e2008-05-10 20:44:54 -04002401 struct vfsmount *rootmnt = NULL, *pwdmnt = NULL;
Al Viro315fc832011-11-24 18:57:30 -05002402 struct mount *p, *q;
Al Virocb338d02011-11-24 20:55:08 -05002403 struct mount *new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002405 new_ns = alloc_mnt_ns();
2406 if (IS_ERR(new_ns))
2407 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408
Ram Pai390c6842005-11-07 17:17:51 -05002409 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002410 /* First pass: copy the tree topology */
Al Viro87129cc2011-11-24 21:24:27 -05002411 new = copy_tree(real_mount(mnt_ns->root), mnt_ns->root->mnt_root,
Ram Pai9676f0c2005-11-07 17:21:20 -05002412 CL_COPY_ALL | CL_EXPIRE);
Al Virocb338d02011-11-24 20:55:08 -05002413 if (!new) {
Ram Pai390c6842005-11-07 17:17:51 -05002414 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415 kfree(new_ns);
Julia Lawall5cc4a032008-12-01 14:34:51 -08002416 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002417 }
Al Virocb338d02011-11-24 20:55:08 -05002418 new_ns->root = &new->mnt;
Nick Piggin99b7db72010-08-18 04:37:39 +10002419 br_write_lock(vfsmount_lock);
Al Viro1a4eeaf2011-11-25 02:19:55 -05002420 list_add_tail(&new_ns->list, &new->mnt_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10002421 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422
2423 /*
2424 * Second pass: switch the tsk->fs->* elements and mark new vfsmounts
2425 * as belonging to new namespace. We have already acquired a private
2426 * fs_struct, so tsk->fs->lock is not needed.
2427 */
Al Viro315fc832011-11-24 18:57:30 -05002428 p = real_mount(mnt_ns->root);
Al Virocb338d02011-11-24 20:55:08 -05002429 q = new;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430 while (p) {
Al Viro143c8c92011-11-25 00:46:35 -05002431 q->mnt_ns = new_ns;
Al Viro83adc752011-11-24 22:37:54 -05002432 __mnt_make_longterm(q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433 if (fs) {
Al Viro315fc832011-11-24 18:57:30 -05002434 if (&p->mnt == fs->root.mnt) {
2435 fs->root.mnt = mntget(&q->mnt);
Al Viro83adc752011-11-24 22:37:54 -05002436 __mnt_make_longterm(q);
Al Viro315fc832011-11-24 18:57:30 -05002437 mnt_make_shortterm(&p->mnt);
2438 rootmnt = &p->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439 }
Al Viro315fc832011-11-24 18:57:30 -05002440 if (&p->mnt == fs->pwd.mnt) {
2441 fs->pwd.mnt = mntget(&q->mnt);
Al Viro83adc752011-11-24 22:37:54 -05002442 __mnt_make_longterm(q);
Al Viro315fc832011-11-24 18:57:30 -05002443 mnt_make_shortterm(&p->mnt);
2444 pwdmnt = &p->mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446 }
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002447 p = next_mnt(p, mnt_ns->root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 q = next_mnt(q, new_ns->root);
2449 }
Ram Pai390c6842005-11-07 17:17:51 -05002450 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 if (rootmnt)
Al Virof03c6592011-01-14 22:30:21 -05002453 mntput(rootmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454 if (pwdmnt)
Al Virof03c6592011-01-14 22:30:21 -05002455 mntput(pwdmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002456
JANAK DESAI741a2952006-02-07 12:59:00 -08002457 return new_ns;
2458}
2459
Eric W. Biederman213dd262007-07-15 23:41:15 -07002460struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns,
Badari Pulavartye3222c42007-05-08 00:25:21 -07002461 struct fs_struct *new_fs)
JANAK DESAI741a2952006-02-07 12:59:00 -08002462{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002463 struct mnt_namespace *new_ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08002464
Badari Pulavartye3222c42007-05-08 00:25:21 -07002465 BUG_ON(!ns);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002466 get_mnt_ns(ns);
JANAK DESAI741a2952006-02-07 12:59:00 -08002467
2468 if (!(flags & CLONE_NEWNS))
Badari Pulavartye3222c42007-05-08 00:25:21 -07002469 return ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08002470
Badari Pulavartye3222c42007-05-08 00:25:21 -07002471 new_ns = dup_mnt_ns(ns, new_fs);
JANAK DESAI741a2952006-02-07 12:59:00 -08002472
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002473 put_mnt_ns(ns);
Badari Pulavartye3222c42007-05-08 00:25:21 -07002474 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002475}
2476
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002477/**
2478 * create_mnt_ns - creates a private namespace and adds a root filesystem
2479 * @mnt: pointer to the new root filesystem mountpoint
2480 */
Al Viro1a4eeaf2011-11-25 02:19:55 -05002481static struct mnt_namespace *create_mnt_ns(struct vfsmount *m)
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002482{
Al Viro1a4eeaf2011-11-25 02:19:55 -05002483 struct mnt_namespace *new_ns = alloc_mnt_ns();
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002484 if (!IS_ERR(new_ns)) {
Al Viro1a4eeaf2011-11-25 02:19:55 -05002485 struct mount *mnt = real_mount(m);
2486 mnt->mnt_ns = new_ns;
2487 __mnt_make_longterm(mnt);
2488 new_ns->root = m;
2489 list_add(&new_ns->list, &mnt->mnt_list);
Al Viroc1334492011-11-16 16:12:14 -05002490 } else {
Al Viro1a4eeaf2011-11-25 02:19:55 -05002491 mntput(m);
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002492 }
2493 return new_ns;
2494}
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002495
Al Viroea441d12011-11-16 21:43:59 -05002496struct dentry *mount_subtree(struct vfsmount *mnt, const char *name)
2497{
2498 struct mnt_namespace *ns;
Al Virod31da0f2011-11-22 12:31:21 -05002499 struct super_block *s;
Al Viroea441d12011-11-16 21:43:59 -05002500 struct path path;
2501 int err;
2502
2503 ns = create_mnt_ns(mnt);
2504 if (IS_ERR(ns))
2505 return ERR_CAST(ns);
2506
2507 err = vfs_path_lookup(mnt->mnt_root, mnt,
2508 name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path);
2509
2510 put_mnt_ns(ns);
2511
2512 if (err)
2513 return ERR_PTR(err);
2514
2515 /* trade a vfsmount reference for active sb one */
Al Virod31da0f2011-11-22 12:31:21 -05002516 s = path.mnt->mnt_sb;
2517 atomic_inc(&s->s_active);
Al Viroea441d12011-11-16 21:43:59 -05002518 mntput(path.mnt);
2519 /* lock the sucker */
Al Virod31da0f2011-11-22 12:31:21 -05002520 down_write(&s->s_umount);
Al Viroea441d12011-11-16 21:43:59 -05002521 /* ... and return the root of (sub)tree on it */
2522 return path.dentry;
2523}
2524EXPORT_SYMBOL(mount_subtree);
2525
Heiko Carstensbdc480e2009-01-14 14:14:12 +01002526SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name,
2527 char __user *, type, unsigned long, flags, void __user *, data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528{
Vegard Nossumeca6f532009-09-18 13:05:45 -07002529 int ret;
2530 char *kernel_type;
2531 char *kernel_dir;
2532 char *kernel_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002533 unsigned long data_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534
Vegard Nossumeca6f532009-09-18 13:05:45 -07002535 ret = copy_mount_string(type, &kernel_type);
2536 if (ret < 0)
2537 goto out_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
Vegard Nossumeca6f532009-09-18 13:05:45 -07002539 kernel_dir = getname(dir_name);
2540 if (IS_ERR(kernel_dir)) {
2541 ret = PTR_ERR(kernel_dir);
2542 goto out_dir;
2543 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002544
Vegard Nossumeca6f532009-09-18 13:05:45 -07002545 ret = copy_mount_string(dev_name, &kernel_dev);
2546 if (ret < 0)
2547 goto out_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548
Vegard Nossumeca6f532009-09-18 13:05:45 -07002549 ret = copy_mount_options(data, &data_page);
2550 if (ret < 0)
2551 goto out_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002552
Vegard Nossumeca6f532009-09-18 13:05:45 -07002553 ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags,
2554 (void *) data_page);
2555
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556 free_page(data_page);
Vegard Nossumeca6f532009-09-18 13:05:45 -07002557out_data:
2558 kfree(kernel_dev);
2559out_dev:
2560 putname(kernel_dir);
2561out_dir:
2562 kfree(kernel_type);
2563out_type:
2564 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565}
2566
2567/*
Al Viroafac7cb2011-11-23 19:34:49 -05002568 * Return true if path is reachable from root
2569 *
2570 * namespace_sem or vfsmount_lock is held
2571 */
Al Viro643822b2011-11-24 22:00:28 -05002572bool is_path_reachable(struct mount *mnt, struct dentry *dentry,
Al Viroafac7cb2011-11-23 19:34:49 -05002573 const struct path *root)
2574{
Al Viro643822b2011-11-24 22:00:28 -05002575 while (&mnt->mnt != root->mnt && mnt_has_parent(mnt)) {
Al Viroa73324d2011-11-24 22:25:07 -05002576 dentry = mnt->mnt_mountpoint;
Al Viro0714a532011-11-24 22:19:58 -05002577 mnt = mnt->mnt_parent;
Al Viroafac7cb2011-11-23 19:34:49 -05002578 }
Al Viro643822b2011-11-24 22:00:28 -05002579 return &mnt->mnt == root->mnt && is_subdir(dentry, root->dentry);
Al Viroafac7cb2011-11-23 19:34:49 -05002580}
2581
2582int path_is_under(struct path *path1, struct path *path2)
2583{
2584 int res;
2585 br_read_lock(vfsmount_lock);
Al Viro643822b2011-11-24 22:00:28 -05002586 res = is_path_reachable(real_mount(path1->mnt), path1->dentry, path2);
Al Viroafac7cb2011-11-23 19:34:49 -05002587 br_read_unlock(vfsmount_lock);
2588 return res;
2589}
2590EXPORT_SYMBOL(path_is_under);
2591
2592/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002593 * pivot_root Semantics:
2594 * Moves the root file system of the current process to the directory put_old,
2595 * makes new_root as the new root file system of the current process, and sets
2596 * root/cwd of all processes which had them on the current root to new_root.
2597 *
2598 * Restrictions:
2599 * The new_root and put_old must be directories, and must not be on the
2600 * same file system as the current process root. The put_old must be
2601 * underneath new_root, i.e. adding a non-zero number of /.. to the string
2602 * pointed to by put_old must yield the same directory as new_root. No other
2603 * file system may be mounted on put_old. After all, new_root is a mountpoint.
2604 *
Neil Brown4a0d11f2006-01-08 01:03:18 -08002605 * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem.
2606 * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives
2607 * in this situation.
2608 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002609 * Notes:
2610 * - we don't move root/cwd if they are not at the root (reason: if something
2611 * cared enough to change them, it's probably wrong to force them elsewhere)
2612 * - it's okay to pick a root that isn't the root of a file system, e.g.
2613 * /nfs/my_root where /nfs is the mount point. It must be a mountpoint,
2614 * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root
2615 * first.
2616 */
Heiko Carstens3480b252009-01-14 14:14:16 +01002617SYSCALL_DEFINE2(pivot_root, const char __user *, new_root,
2618 const char __user *, put_old)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002619{
Al Viro2d8f3032008-07-22 09:59:21 -04002620 struct path new, old, parent_path, root_parent, root;
Al Viro419148d2011-11-24 19:41:16 -05002621 struct mount *new_mnt, *root_mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002622 int error;
2623
2624 if (!capable(CAP_SYS_ADMIN))
2625 return -EPERM;
2626
Al Viro2d8f3032008-07-22 09:59:21 -04002627 error = user_path_dir(new_root, &new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002628 if (error)
2629 goto out0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002630
Al Viro2d8f3032008-07-22 09:59:21 -04002631 error = user_path_dir(put_old, &old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002632 if (error)
2633 goto out1;
2634
Al Viro2d8f3032008-07-22 09:59:21 -04002635 error = security_sb_pivotroot(&old, &new);
Al Virob12cea92011-03-18 08:55:38 -04002636 if (error)
2637 goto out2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638
Miklos Szeredif7ad3c62010-08-10 11:41:36 +02002639 get_fs_root(current->fs, &root);
Al Virob12cea92011-03-18 08:55:38 -04002640 error = lock_mount(&old);
2641 if (error)
2642 goto out3;
2643
Linus Torvalds1da177e2005-04-16 15:20:36 -07002644 error = -EINVAL;
Al Viro419148d2011-11-24 19:41:16 -05002645 new_mnt = real_mount(new.mnt);
2646 root_mnt = real_mount(root.mnt);
Al Virofc7be132011-11-25 01:05:37 -05002647 if (IS_MNT_SHARED(real_mount(old.mnt)) ||
2648 IS_MNT_SHARED(new_mnt->mnt_parent) ||
2649 IS_MNT_SHARED(root_mnt->mnt_parent))
Al Virob12cea92011-03-18 08:55:38 -04002650 goto out4;
Al Viro143c8c92011-11-25 00:46:35 -05002651 if (!check_mnt(root_mnt) || !check_mnt(new_mnt))
Al Virob12cea92011-03-18 08:55:38 -04002652 goto out4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653 error = -ENOENT;
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04002654 if (d_unlinked(new.dentry))
Al Virob12cea92011-03-18 08:55:38 -04002655 goto out4;
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04002656 if (d_unlinked(old.dentry))
Al Virob12cea92011-03-18 08:55:38 -04002657 goto out4;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658 error = -EBUSY;
Al Viro2d8f3032008-07-22 09:59:21 -04002659 if (new.mnt == root.mnt ||
2660 old.mnt == root.mnt)
Al Virob12cea92011-03-18 08:55:38 -04002661 goto out4; /* loop, on the same file system */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002662 error = -EINVAL;
Al Viro8c3ee422008-03-22 18:00:39 -04002663 if (root.mnt->mnt_root != root.dentry)
Al Virob12cea92011-03-18 08:55:38 -04002664 goto out4; /* not a mountpoint */
Al Viro676da582011-11-24 21:47:05 -05002665 if (!mnt_has_parent(root_mnt))
Al Virob12cea92011-03-18 08:55:38 -04002666 goto out4; /* not attached */
Al Viro2d8f3032008-07-22 09:59:21 -04002667 if (new.mnt->mnt_root != new.dentry)
Al Virob12cea92011-03-18 08:55:38 -04002668 goto out4; /* not a mountpoint */
Al Viro676da582011-11-24 21:47:05 -05002669 if (!mnt_has_parent(new_mnt))
Al Virob12cea92011-03-18 08:55:38 -04002670 goto out4; /* not attached */
Jan Blunck4ac91372008-02-14 19:34:32 -08002671 /* make sure we can reach put_old from new_root */
Al Viro643822b2011-11-24 22:00:28 -05002672 if (!is_path_reachable(real_mount(old.mnt), old.dentry, &new))
Al Virob12cea92011-03-18 08:55:38 -04002673 goto out4;
Al Viro27cb1572011-03-18 08:29:36 -04002674 br_write_lock(vfsmount_lock);
Al Viro419148d2011-11-24 19:41:16 -05002675 detach_mnt(new_mnt, &parent_path);
2676 detach_mnt(root_mnt, &root_parent);
Jan Blunck4ac91372008-02-14 19:34:32 -08002677 /* mount old root on put_old */
Al Viro419148d2011-11-24 19:41:16 -05002678 attach_mnt(root_mnt, &old);
Jan Blunck4ac91372008-02-14 19:34:32 -08002679 /* mount new_root on / */
Al Viro419148d2011-11-24 19:41:16 -05002680 attach_mnt(new_mnt, &root_parent);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002681 touch_mnt_namespace(current->nsproxy->mnt_ns);
Nick Piggin99b7db72010-08-18 04:37:39 +10002682 br_write_unlock(vfsmount_lock);
Al Viro2d8f3032008-07-22 09:59:21 -04002683 chroot_fs_refs(&root, &new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002684 error = 0;
Al Virob12cea92011-03-18 08:55:38 -04002685out4:
2686 unlock_mount(&old);
2687 if (!error) {
2688 path_put(&root_parent);
2689 path_put(&parent_path);
2690 }
2691out3:
Al Viro8c3ee422008-03-22 18:00:39 -04002692 path_put(&root);
Al Virob12cea92011-03-18 08:55:38 -04002693out2:
Al Viro2d8f3032008-07-22 09:59:21 -04002694 path_put(&old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002695out1:
Al Viro2d8f3032008-07-22 09:59:21 -04002696 path_put(&new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697out0:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002698 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002699}
2700
2701static void __init init_mount_tree(void)
2702{
2703 struct vfsmount *mnt;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002704 struct mnt_namespace *ns;
Jan Blunckac748a02008-02-14 19:34:39 -08002705 struct path root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002706
2707 mnt = do_kern_mount("rootfs", 0, "rootfs", NULL);
2708 if (IS_ERR(mnt))
2709 panic("Can't create rootfs");
Nick Pigginb3e19d92011-01-07 17:50:11 +11002710
Trond Myklebust3b22edc2009-06-23 17:29:49 -04002711 ns = create_mnt_ns(mnt);
2712 if (IS_ERR(ns))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713 panic("Can't allocate initial namespace");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002714
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002715 init_task.nsproxy->mnt_ns = ns;
2716 get_mnt_ns(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002717
Jan Blunckac748a02008-02-14 19:34:39 -08002718 root.mnt = ns->root;
2719 root.dentry = ns->root->mnt_root;
2720
2721 set_fs_pwd(current->fs, &root);
2722 set_fs_root(current->fs, &root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002723}
2724
Denis Cheng74bf17c2007-10-16 23:26:30 -07002725void __init mnt_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726{
Eric Dumazet13f14b42008-02-06 01:37:57 -08002727 unsigned u;
Randy Dunlap15a67dd2006-09-29 01:58:57 -07002728 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002729
Ram Pai390c6842005-11-07 17:17:51 -05002730 init_rwsem(&namespace_sem);
2731
Al Viro7d6fec42011-11-23 12:14:10 -05002732 mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount),
Paul Mundt20c2df82007-07-20 10:11:58 +09002733 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002734
Ram Paib58fed82005-11-07 17:16:09 -05002735 mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002736
2737 if (!mount_hashtable)
2738 panic("Failed to allocate mount hash table\n");
2739
Mandeep Singh Baines80cdc6d2011-03-22 16:33:54 -07002740 printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741
Eric Dumazet13f14b42008-02-06 01:37:57 -08002742 for (u = 0; u < HASH_SIZE; u++)
2743 INIT_LIST_HEAD(&mount_hashtable[u]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002744
Nick Piggin99b7db72010-08-18 04:37:39 +10002745 br_lock_init(vfsmount_lock);
2746
Randy Dunlap15a67dd2006-09-29 01:58:57 -07002747 err = sysfs_init();
2748 if (err)
2749 printk(KERN_WARNING "%s: sysfs_init error: %d\n",
Harvey Harrison8e24eea2008-04-30 00:55:09 -07002750 __func__, err);
Greg Kroah-Hartman00d26662007-10-29 14:17:23 -06002751 fs_kobj = kobject_create_and_add("fs", NULL);
2752 if (!fs_kobj)
Harvey Harrison8e24eea2008-04-30 00:55:09 -07002753 printk(KERN_WARNING "%s: kobj create error\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002754 init_rootfs();
2755 init_mount_tree();
2756}
2757
Trond Myklebust616511d2009-06-22 15:09:13 -04002758void put_mnt_ns(struct mnt_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002759{
Ram Pai70fbcdf2005-11-07 17:17:04 -05002760 LIST_HEAD(umount_list);
Trond Myklebust616511d2009-06-22 15:09:13 -04002761
Al Virod498b252010-02-05 02:21:06 -05002762 if (!atomic_dec_and_test(&ns->count))
Trond Myklebust616511d2009-06-22 15:09:13 -04002763 return;
Ram Pai390c6842005-11-07 17:17:51 -05002764 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10002765 br_write_lock(vfsmount_lock);
Al Viro761d5c32011-11-24 21:07:43 -05002766 umount_tree(real_mount(ns->root), 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10002767 br_write_unlock(vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05002768 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05002769 release_mounts(&umount_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002770 kfree(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002771}
Al Viro9d412a42011-03-17 22:08:28 -04002772
2773struct vfsmount *kern_mount_data(struct file_system_type *type, void *data)
2774{
Tim Chen423e0ab2011-07-19 09:32:38 -07002775 struct vfsmount *mnt;
2776 mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data);
2777 if (!IS_ERR(mnt)) {
2778 /*
2779 * it is a longterm mount, don't release mnt until
2780 * we unmount before file sys is unregistered
2781 */
2782 mnt_make_longterm(mnt);
2783 }
2784 return mnt;
Al Viro9d412a42011-03-17 22:08:28 -04002785}
2786EXPORT_SYMBOL_GPL(kern_mount_data);
Tim Chen423e0ab2011-07-19 09:32:38 -07002787
2788void kern_unmount(struct vfsmount *mnt)
2789{
2790 /* release long term mount so mount point can be released */
2791 if (!IS_ERR_OR_NULL(mnt)) {
2792 mnt_make_shortterm(mnt);
2793 mntput(mnt);
2794 }
2795}
2796EXPORT_SYMBOL(kern_unmount);
Al Viro02125a82011-12-05 08:43:34 -05002797
2798bool our_mnt(struct vfsmount *mnt)
2799{
Al Viro143c8c92011-11-25 00:46:35 -05002800 return check_mnt(real_mount(mnt));
Al Viro02125a82011-12-05 08:43:34 -05002801}