blob: 3ddfd9046c449199a8f75ffcaf54761dde37fa73 [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 Howells07f3f05c2006-09-30 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 */
Miklos Szeredi73cd49e2008-03-26 22:11:34 +010081static int mnt_alloc_id(struct vfsmount *mnt)
82{
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 Virof21f6222009-06-24 03:12:00 -040088 res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt_id);
89 if (!res)
90 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
98static void mnt_free_id(struct vfsmount *mnt)
99{
Al Virof21f6222009-06-24 03:12:00 -0400100 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 */
113static int mnt_alloc_group_id(struct vfsmount *mnt)
114{
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,
122 &mnt->mnt_group_id);
123 if (!res)
124 mnt_group_start = mnt->mnt_group_id + 1;
125
126 return res;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100127}
128
129/*
130 * Release a peer group ID
131 */
132void mnt_release_group_id(struct vfsmount *mnt)
133{
Al Virof21f6222009-06-24 03:12:00 -0400134 int id = mnt->mnt_group_id;
135 ida_remove(&mnt_group_ida, id);
136 if (mnt_group_start > id)
137 mnt_group_start = id;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100138 mnt->mnt_group_id = 0;
139}
140
Nick Pigginb3e19d92011-01-07 17:50:11 +1100141/*
142 * vfsmount lock must be held for read
143 */
144static inline void mnt_add_count(struct vfsmount *mnt, int n)
145{
146#ifdef CONFIG_SMP
147 this_cpu_add(mnt->mnt_pcp->mnt_count, n);
148#else
149 preempt_disable();
150 mnt->mnt_count += n;
151 preempt_enable();
152#endif
153}
154
155static inline void mnt_set_count(struct vfsmount *mnt, int n)
156{
157#ifdef CONFIG_SMP
158 this_cpu_write(mnt->mnt_pcp->mnt_count, n);
159#else
160 mnt->mnt_count = n;
161#endif
162}
163
164/*
165 * vfsmount lock must be held for read
166 */
167static inline void mnt_inc_count(struct vfsmount *mnt)
168{
169 mnt_add_count(mnt, 1);
170}
171
172/*
173 * vfsmount lock must be held for read
174 */
175static inline void mnt_dec_count(struct vfsmount *mnt)
176{
177 mnt_add_count(mnt, -1);
178}
179
180/*
181 * vfsmount lock must be held for write
182 */
183unsigned int mnt_get_count(struct vfsmount *mnt)
184{
185#ifdef CONFIG_SMP
186 unsigned int count = atomic_read(&mnt->mnt_longrefs);
187 int cpu;
188
189 for_each_possible_cpu(cpu) {
190 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count;
191 }
192
193 return count;
194#else
195 return mnt->mnt_count;
196#endif
197}
198
Linus Torvalds1da177e2005-04-16 15:20:36 -0700199struct vfsmount *alloc_vfsmnt(const char *name)
200{
Robert P. J. Dayc3762222007-02-10 01:45:03 -0800201 struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202 if (mnt) {
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100203 int err;
204
205 err = mnt_alloc_id(mnt);
Li Zefan88b38782008-07-21 18:06:36 +0800206 if (err)
207 goto out_free_cache;
208
209 if (name) {
210 mnt->mnt_devname = kstrdup(name, GFP_KERNEL);
211 if (!mnt->mnt_devname)
212 goto out_free_id;
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100213 }
214
Nick Pigginb3e19d92011-01-07 17:50:11 +1100215#ifdef CONFIG_SMP
216 mnt->mnt_pcp = alloc_percpu(struct mnt_pcp);
217 if (!mnt->mnt_pcp)
218 goto out_free_devname;
219
220 atomic_set(&mnt->mnt_longrefs, 1);
221#else
222 mnt->mnt_count = 1;
223 mnt->mnt_writers = 0;
224#endif
225
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226 INIT_LIST_HEAD(&mnt->mnt_hash);
227 INIT_LIST_HEAD(&mnt->mnt_child);
228 INIT_LIST_HEAD(&mnt->mnt_mounts);
229 INIT_LIST_HEAD(&mnt->mnt_list);
Miklos Szeredi55e700b2005-07-07 17:57:30 -0700230 INIT_LIST_HEAD(&mnt->mnt_expire);
Ram Pai03e06e62005-11-07 17:19:33 -0500231 INIT_LIST_HEAD(&mnt->mnt_share);
Ram Paia58b0eb2005-11-07 17:20:48 -0500232 INIT_LIST_HEAD(&mnt->mnt_slave_list);
233 INIT_LIST_HEAD(&mnt->mnt_slave);
Andreas Gruenbacher2504c5d2009-12-17 21:24:27 -0500234#ifdef CONFIG_FSNOTIFY
235 INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks);
236#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 }
238 return mnt;
Li Zefan88b38782008-07-21 18:06:36 +0800239
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000240#ifdef CONFIG_SMP
241out_free_devname:
242 kfree(mnt->mnt_devname);
243#endif
Li Zefan88b38782008-07-21 18:06:36 +0800244out_free_id:
245 mnt_free_id(mnt);
246out_free_cache:
247 kmem_cache_free(mnt_cache, mnt);
248 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249}
250
Dave Hansen83660252008-02-15 14:37:30 -0800251/*
252 * Most r/o checks on a fs are for operations that take
253 * discrete amounts of time, like a write() or unlink().
254 * We must keep track of when those operations start
255 * (for permission checks) and when they end, so that
256 * we can determine when writes are able to occur to
257 * a filesystem.
258 */
Dave Hansen3d733632008-02-15 14:37:59 -0800259/*
260 * __mnt_is_readonly: check whether a mount is read-only
261 * @mnt: the mount to check for its write status
262 *
263 * This shouldn't be used directly ouside of the VFS.
264 * It does not guarantee that the filesystem will stay
265 * r/w, just that it is right *now*. This can not and
266 * should not be used in place of IS_RDONLY(inode).
267 * mnt_want/drop_write() will _keep_ the filesystem
268 * r/w.
269 */
270int __mnt_is_readonly(struct vfsmount *mnt)
271{
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800272 if (mnt->mnt_flags & MNT_READONLY)
273 return 1;
274 if (mnt->mnt_sb->s_flags & MS_RDONLY)
275 return 1;
276 return 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800277}
278EXPORT_SYMBOL_GPL(__mnt_is_readonly);
279
Nick Pigginc6653a82011-01-07 17:50:10 +1100280static inline void mnt_inc_writers(struct vfsmount *mnt)
Dave Hansen3d733632008-02-15 14:37:59 -0800281{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000282#ifdef CONFIG_SMP
Nick Pigginb3e19d92011-01-07 17:50:11 +1100283 this_cpu_inc(mnt->mnt_pcp->mnt_writers);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000284#else
285 mnt->mnt_writers++;
286#endif
Dave Hansen3d733632008-02-15 14:37:59 -0800287}
Dave Hansen3d733632008-02-15 14:37:59 -0800288
Nick Pigginc6653a82011-01-07 17:50:10 +1100289static inline void mnt_dec_writers(struct vfsmount *mnt)
Dave Hansen3d733632008-02-15 14:37:59 -0800290{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000291#ifdef CONFIG_SMP
Nick Pigginb3e19d92011-01-07 17:50:11 +1100292 this_cpu_dec(mnt->mnt_pcp->mnt_writers);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000293#else
294 mnt->mnt_writers--;
295#endif
296}
297
Nick Pigginc6653a82011-01-07 17:50:10 +1100298static unsigned int mnt_get_writers(struct vfsmount *mnt)
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000299{
300#ifdef CONFIG_SMP
301 unsigned int count = 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800302 int cpu;
Dave Hansen3d733632008-02-15 14:37:59 -0800303
304 for_each_possible_cpu(cpu) {
Nick Pigginb3e19d92011-01-07 17:50:11 +1100305 count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers;
Dave Hansen3d733632008-02-15 14:37:59 -0800306 }
Dave Hansen3d733632008-02-15 14:37:59 -0800307
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000308 return count;
309#else
310 return mnt->mnt_writers;
311#endif
Dave Hansen3d733632008-02-15 14:37:59 -0800312}
313
314/*
315 * Most r/o checks on a fs are for operations that take
316 * discrete amounts of time, like a write() or unlink().
317 * We must keep track of when those operations start
318 * (for permission checks) and when they end, so that
319 * we can determine when writes are able to occur to
320 * a filesystem.
321 */
Dave Hansen83660252008-02-15 14:37:30 -0800322/**
323 * mnt_want_write - get write access to a mount
324 * @mnt: the mount on which to take a write
325 *
326 * This tells the low-level filesystem that a write is
327 * about to be performed to it, and makes sure that
328 * writes are allowed before returning success. When
329 * the write operation is finished, mnt_drop_write()
330 * must be called. This is effectively a refcount.
331 */
332int mnt_want_write(struct vfsmount *mnt)
333{
Dave Hansen3d733632008-02-15 14:37:59 -0800334 int ret = 0;
Dave Hansen3d733632008-02-15 14:37:59 -0800335
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000336 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100337 mnt_inc_writers(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000338 /*
Nick Pigginc6653a82011-01-07 17:50:10 +1100339 * The store to mnt_inc_writers must be visible before we pass
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000340 * MNT_WRITE_HOLD loop below, so that the slowpath can see our
341 * incremented count after it has set MNT_WRITE_HOLD.
342 */
343 smp_mb();
344 while (mnt->mnt_flags & MNT_WRITE_HOLD)
345 cpu_relax();
346 /*
347 * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will
348 * be set to match its requirements. So we must not load that until
349 * MNT_WRITE_HOLD is cleared.
350 */
351 smp_rmb();
Dave Hansen3d733632008-02-15 14:37:59 -0800352 if (__mnt_is_readonly(mnt)) {
Nick Pigginc6653a82011-01-07 17:50:10 +1100353 mnt_dec_writers(mnt);
Dave Hansen3d733632008-02-15 14:37:59 -0800354 ret = -EROFS;
355 goto out;
356 }
Dave Hansen3d733632008-02-15 14:37:59 -0800357out:
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000358 preempt_enable();
Dave Hansen3d733632008-02-15 14:37:59 -0800359 return ret;
Dave Hansen83660252008-02-15 14:37:30 -0800360}
361EXPORT_SYMBOL_GPL(mnt_want_write);
362
363/**
npiggin@suse.de96029c42009-04-26 20:25:55 +1000364 * mnt_clone_write - get write access to a mount
365 * @mnt: the mount on which to take a write
366 *
367 * This is effectively like mnt_want_write, except
368 * it must only be used to take an extra write reference
369 * on a mountpoint that we already know has a write reference
370 * on it. This allows some optimisation.
371 *
372 * After finished, mnt_drop_write must be called as usual to
373 * drop the reference.
374 */
375int mnt_clone_write(struct vfsmount *mnt)
376{
377 /* superblock may be r/o */
378 if (__mnt_is_readonly(mnt))
379 return -EROFS;
380 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100381 mnt_inc_writers(mnt);
npiggin@suse.de96029c42009-04-26 20:25:55 +1000382 preempt_enable();
383 return 0;
384}
385EXPORT_SYMBOL_GPL(mnt_clone_write);
386
387/**
388 * mnt_want_write_file - get write access to a file's mount
389 * @file: the file who's mount on which to take a write
390 *
391 * This is like mnt_want_write, but it takes a file and can
392 * do some optimisations if the file is open for write already
393 */
394int mnt_want_write_file(struct file *file)
395{
OGAWA Hirofumi2d8dd382009-08-06 15:07:39 -0700396 struct inode *inode = file->f_dentry->d_inode;
397 if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode))
npiggin@suse.de96029c42009-04-26 20:25:55 +1000398 return mnt_want_write(file->f_path.mnt);
399 else
400 return mnt_clone_write(file->f_path.mnt);
401}
402EXPORT_SYMBOL_GPL(mnt_want_write_file);
403
404/**
Dave Hansen83660252008-02-15 14:37:30 -0800405 * mnt_drop_write - give up write access to a mount
406 * @mnt: the mount on which to give up write access
407 *
408 * Tells the low-level filesystem that we are done
409 * performing writes to it. Must be matched with
410 * mnt_want_write() call above.
411 */
412void mnt_drop_write(struct vfsmount *mnt)
413{
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000414 preempt_disable();
Nick Pigginc6653a82011-01-07 17:50:10 +1100415 mnt_dec_writers(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000416 preempt_enable();
Dave Hansen83660252008-02-15 14:37:30 -0800417}
418EXPORT_SYMBOL_GPL(mnt_drop_write);
419
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800420static int mnt_make_readonly(struct vfsmount *mnt)
Dave Hansen83660252008-02-15 14:37:30 -0800421{
Dave Hansen3d733632008-02-15 14:37:59 -0800422 int ret = 0;
423
Nick Piggin99b7db72010-08-18 04:37:39 +1000424 br_write_lock(vfsmount_lock);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000425 mnt->mnt_flags |= MNT_WRITE_HOLD;
426 /*
427 * After storing MNT_WRITE_HOLD, we'll read the counters. This store
428 * should be visible before we do.
429 */
430 smp_mb();
431
432 /*
433 * With writers on hold, if this value is zero, then there are
434 * definitely no active writers (although held writers may subsequently
435 * increment the count, they'll have to wait, and decrement it after
436 * seeing MNT_READONLY).
437 *
438 * It is OK to have counter incremented on one CPU and decremented on
439 * another: the sum will add up correctly. The danger would be when we
440 * sum up each counter, if we read a counter before it is incremented,
441 * but then read another CPU's count which it has been subsequently
442 * decremented from -- we would see more decrements than we should.
443 * MNT_WRITE_HOLD protects against this scenario, because
444 * mnt_want_write first increments count, then smp_mb, then spins on
445 * MNT_WRITE_HOLD, so it can't be decremented by another CPU while
446 * we're counting up here.
447 */
Nick Pigginc6653a82011-01-07 17:50:10 +1100448 if (mnt_get_writers(mnt) > 0)
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000449 ret = -EBUSY;
450 else
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800451 mnt->mnt_flags |= MNT_READONLY;
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000452 /*
453 * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers
454 * that become unheld will see MNT_READONLY.
455 */
456 smp_wmb();
457 mnt->mnt_flags &= ~MNT_WRITE_HOLD;
Nick Piggin99b7db72010-08-18 04:37:39 +1000458 br_write_unlock(vfsmount_lock);
Dave Hansen3d733632008-02-15 14:37:59 -0800459 return ret;
Dave Hansen83660252008-02-15 14:37:30 -0800460}
Dave Hansen83660252008-02-15 14:37:30 -0800461
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800462static void __mnt_unmake_readonly(struct vfsmount *mnt)
463{
Nick Piggin99b7db72010-08-18 04:37:39 +1000464 br_write_lock(vfsmount_lock);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800465 mnt->mnt_flags &= ~MNT_READONLY;
Nick Piggin99b7db72010-08-18 04:37:39 +1000466 br_write_unlock(vfsmount_lock);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -0800467}
468
Sukadev Bhattiprolua3ec9472009-03-04 12:06:34 -0800469void simple_set_mnt(struct vfsmount *mnt, struct super_block *sb)
David Howells454e2392006-06-23 02:02:57 -0700470{
471 mnt->mnt_sb = sb;
472 mnt->mnt_root = dget(sb->s_root);
David Howells454e2392006-06-23 02:02:57 -0700473}
474
475EXPORT_SYMBOL(simple_set_mnt);
476
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477void free_vfsmnt(struct vfsmount *mnt)
478{
479 kfree(mnt->mnt_devname);
Miklos Szeredi73cd49e2008-03-26 22:11:34 +0100480 mnt_free_id(mnt);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000481#ifdef CONFIG_SMP
Nick Pigginb3e19d92011-01-07 17:50:11 +1100482 free_percpu(mnt->mnt_pcp);
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000483#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 kmem_cache_free(mnt_cache, mnt);
485}
486
487/*
Ram Paia05964f2005-11-07 17:20:17 -0500488 * find the first or last mount at @dentry on vfsmount @mnt depending on
489 * @dir. If @dir is set return the first mount else return the last mount.
Nick Piggin99b7db72010-08-18 04:37:39 +1000490 * vfsmount_lock must be held for read or write.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 */
Ram Paia05964f2005-11-07 17:20:17 -0500492struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry,
493 int dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494{
Ram Paib58fed82005-11-07 17:16:09 -0500495 struct list_head *head = mount_hashtable + hash(mnt, dentry);
496 struct list_head *tmp = head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 struct vfsmount *p, *found = NULL;
498
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 for (;;) {
Ram Paia05964f2005-11-07 17:20:17 -0500500 tmp = dir ? tmp->next : tmp->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 p = NULL;
502 if (tmp == head)
503 break;
504 p = list_entry(tmp, struct vfsmount, mnt_hash);
505 if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) {
Ram Paia05964f2005-11-07 17:20:17 -0500506 found = p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 break;
508 }
509 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 return found;
511}
512
Ram Paia05964f2005-11-07 17:20:17 -0500513/*
514 * lookup_mnt increments the ref count before returning
515 * the vfsmount struct.
516 */
Al Viro1c755af2009-04-18 14:06:57 -0400517struct vfsmount *lookup_mnt(struct path *path)
Ram Paia05964f2005-11-07 17:20:17 -0500518{
519 struct vfsmount *child_mnt;
Nick Piggin99b7db72010-08-18 04:37:39 +1000520
521 br_read_lock(vfsmount_lock);
Al Viro1c755af2009-04-18 14:06:57 -0400522 if ((child_mnt = __lookup_mnt(path->mnt, path->dentry, 1)))
Ram Paia05964f2005-11-07 17:20:17 -0500523 mntget(child_mnt);
Nick Piggin99b7db72010-08-18 04:37:39 +1000524 br_read_unlock(vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -0500525 return child_mnt;
526}
527
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528static inline int check_mnt(struct vfsmount *mnt)
529{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800530 return mnt->mnt_ns == current->nsproxy->mnt_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531}
532
Nick Piggin99b7db72010-08-18 04:37:39 +1000533/*
534 * vfsmount lock must be held for write
535 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800536static void touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500537{
538 if (ns) {
539 ns->event = ++event;
540 wake_up_interruptible(&ns->poll);
541 }
542}
543
Nick Piggin99b7db72010-08-18 04:37:39 +1000544/*
545 * vfsmount lock must be held for write
546 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800547static void __touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500548{
549 if (ns && ns->event != event) {
550 ns->event = event;
551 wake_up_interruptible(&ns->poll);
552 }
553}
554
Nick Piggin99b7db72010-08-18 04:37:39 +1000555/*
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100556 * Clear dentry's mounted state if it has no remaining mounts.
557 * vfsmount_lock must be held for write.
558 */
559static void dentry_reset_mounted(struct vfsmount *mnt, struct dentry *dentry)
560{
561 unsigned u;
562
563 for (u = 0; u < HASH_SIZE; u++) {
564 struct vfsmount *p;
565
566 list_for_each_entry(p, &mount_hashtable[u], mnt_hash) {
567 if (p->mnt_mountpoint == dentry)
568 return;
569 }
570 }
571 spin_lock(&dentry->d_lock);
572 dentry->d_flags &= ~DCACHE_MOUNTED;
573 spin_unlock(&dentry->d_lock);
574}
575
576/*
Nick Piggin99b7db72010-08-18 04:37:39 +1000577 * vfsmount lock must be held for write
578 */
Al Viro1a390682008-03-21 20:48:19 -0400579static void detach_mnt(struct vfsmount *mnt, struct path *old_path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
Al Viro1a390682008-03-21 20:48:19 -0400581 old_path->dentry = mnt->mnt_mountpoint;
582 old_path->mnt = mnt->mnt_parent;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583 mnt->mnt_parent = mnt;
584 mnt->mnt_mountpoint = mnt->mnt_root;
585 list_del_init(&mnt->mnt_child);
586 list_del_init(&mnt->mnt_hash);
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100587 dentry_reset_mounted(old_path->mnt, old_path->dentry);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588}
589
Nick Piggin99b7db72010-08-18 04:37:39 +1000590/*
591 * vfsmount lock must be held for write
592 */
Ram Paib90fa9a2005-11-07 17:19:50 -0500593void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry,
594 struct vfsmount *child_mnt)
595{
596 child_mnt->mnt_parent = mntget(mnt);
597 child_mnt->mnt_mountpoint = dget(dentry);
Nick Piggin5f57cbc2011-01-07 17:49:54 +1100598 spin_lock(&dentry->d_lock);
599 dentry->d_flags |= DCACHE_MOUNTED;
600 spin_unlock(&dentry->d_lock);
Ram Paib90fa9a2005-11-07 17:19:50 -0500601}
602
Nick Piggin99b7db72010-08-18 04:37:39 +1000603/*
604 * vfsmount lock must be held for write
605 */
Al Viro1a390682008-03-21 20:48:19 -0400606static void attach_mnt(struct vfsmount *mnt, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607{
Al Viro1a390682008-03-21 20:48:19 -0400608 mnt_set_mountpoint(path->mnt, path->dentry, mnt);
Ram Paib90fa9a2005-11-07 17:19:50 -0500609 list_add_tail(&mnt->mnt_hash, mount_hashtable +
Al Viro1a390682008-03-21 20:48:19 -0400610 hash(path->mnt, path->dentry));
611 list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts);
Ram Paib90fa9a2005-11-07 17:19:50 -0500612}
613
614/*
Nick Piggin99b7db72010-08-18 04:37:39 +1000615 * vfsmount lock must be held for write
Ram Paib90fa9a2005-11-07 17:19:50 -0500616 */
617static void commit_tree(struct vfsmount *mnt)
618{
619 struct vfsmount *parent = mnt->mnt_parent;
620 struct vfsmount *m;
621 LIST_HEAD(head);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800622 struct mnt_namespace *n = parent->mnt_ns;
Ram Paib90fa9a2005-11-07 17:19:50 -0500623
624 BUG_ON(parent == mnt);
625
626 list_add_tail(&head, &mnt->mnt_list);
627 list_for_each_entry(m, &head, mnt_list)
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800628 m->mnt_ns = n;
Ram Paib90fa9a2005-11-07 17:19:50 -0500629 list_splice(&head, n->list.prev);
630
631 list_add_tail(&mnt->mnt_hash, mount_hashtable +
632 hash(parent, mnt->mnt_mountpoint));
633 list_add_tail(&mnt->mnt_child, &parent->mnt_mounts);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800634 touch_mnt_namespace(n);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635}
636
637static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root)
638{
639 struct list_head *next = p->mnt_mounts.next;
640 if (next == &p->mnt_mounts) {
641 while (1) {
642 if (p == root)
643 return NULL;
644 next = p->mnt_child.next;
645 if (next != &p->mnt_parent->mnt_mounts)
646 break;
647 p = p->mnt_parent;
648 }
649 }
650 return list_entry(next, struct vfsmount, mnt_child);
651}
652
Ram Pai9676f0c2005-11-07 17:21:20 -0500653static struct vfsmount *skip_mnt_tree(struct vfsmount *p)
654{
655 struct list_head *prev = p->mnt_mounts.prev;
656 while (prev != &p->mnt_mounts) {
657 p = list_entry(prev, struct vfsmount, mnt_child);
658 prev = p->mnt_mounts.prev;
659 }
660 return p;
661}
662
Ram Pai36341f62005-11-07 17:17:22 -0500663static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root,
664 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665{
666 struct super_block *sb = old->mnt_sb;
667 struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname);
668
669 if (mnt) {
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100670 if (flag & (CL_SLAVE | CL_PRIVATE))
671 mnt->mnt_group_id = 0; /* not a peer of original */
672 else
673 mnt->mnt_group_id = old->mnt_group_id;
674
675 if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) {
676 int err = mnt_alloc_group_id(mnt);
677 if (err)
678 goto out_free;
679 }
680
Miklos Szeredibe1a16a2010-10-05 12:31:09 +0200681 mnt->mnt_flags = old->mnt_flags & ~MNT_WRITE_HOLD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 atomic_inc(&sb->s_active);
683 mnt->mnt_sb = sb;
684 mnt->mnt_root = dget(root);
685 mnt->mnt_mountpoint = mnt->mnt_root;
686 mnt->mnt_parent = mnt;
Ram Paib90fa9a2005-11-07 17:19:50 -0500687
Ram Pai5afe0022005-11-07 17:21:01 -0500688 if (flag & CL_SLAVE) {
689 list_add(&mnt->mnt_slave, &old->mnt_slave_list);
690 mnt->mnt_master = old;
691 CLEAR_MNT_SHARED(mnt);
Al Viro8aec0802007-06-07 12:20:32 -0400692 } else if (!(flag & CL_PRIVATE)) {
Al Viro796a6b52010-01-16 13:28:47 -0500693 if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old))
Ram Pai5afe0022005-11-07 17:21:01 -0500694 list_add(&mnt->mnt_share, &old->mnt_share);
695 if (IS_MNT_SLAVE(old))
696 list_add(&mnt->mnt_slave, &old->mnt_slave);
697 mnt->mnt_master = old->mnt_master;
698 }
Ram Paib90fa9a2005-11-07 17:19:50 -0500699 if (flag & CL_MAKE_SHARED)
700 set_mnt_shared(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
702 /* stick the duplicate mount on the same expiry list
703 * as the original if that was on one */
Ram Pai36341f62005-11-07 17:17:22 -0500704 if (flag & CL_EXPIRE) {
Ram Pai36341f62005-11-07 17:17:22 -0500705 if (!list_empty(&old->mnt_expire))
706 list_add(&mnt->mnt_expire, &old->mnt_expire);
Ram Pai36341f62005-11-07 17:17:22 -0500707 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 }
709 return mnt;
Miklos Szeredi719f5d72008-03-27 13:06:23 +0100710
711 out_free:
712 free_vfsmnt(mnt);
713 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714}
715
Nick Pigginb3e19d92011-01-07 17:50:11 +1100716static inline void mntfree(struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717{
718 struct super_block *sb = mnt->mnt_sb;
Nick Pigginb3e19d92011-01-07 17:50:11 +1100719
Dave Hansen3d733632008-02-15 14:37:59 -0800720 /*
Dave Hansen3d733632008-02-15 14:37:59 -0800721 * This probably indicates that somebody messed
722 * up a mnt_want/drop_write() pair. If this
723 * happens, the filesystem was probably unable
724 * to make r/w->r/o transitions.
725 */
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000726 /*
Nick Pigginb3e19d92011-01-07 17:50:11 +1100727 * The locking used to deal with mnt_count decrement provides barriers,
728 * so mnt_get_writers() below is safe.
npiggin@suse.ded3ef3d72009-04-26 20:25:54 +1000729 */
Nick Pigginc6653a82011-01-07 17:50:10 +1100730 WARN_ON(mnt_get_writers(mnt));
Andreas Gruenbacherca9c7262009-12-17 21:24:27 -0500731 fsnotify_vfsmount_delete(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 dput(mnt->mnt_root);
733 free_vfsmnt(mnt);
734 deactivate_super(sb);
735}
736
Nick Pigginb3e19d92011-01-07 17:50:11 +1100737#ifdef CONFIG_SMP
738static inline void __mntput(struct vfsmount *mnt, int longrefs)
Al Viro7b7b1ac2005-11-07 17:13:39 -0500739{
Nick Pigginb3e19d92011-01-07 17:50:11 +1100740 if (!longrefs) {
741put_again:
742 br_read_lock(vfsmount_lock);
743 if (likely(atomic_read(&mnt->mnt_longrefs))) {
744 mnt_dec_count(mnt);
745 br_read_unlock(vfsmount_lock);
746 return;
747 }
748 br_read_unlock(vfsmount_lock);
749 } else {
750 BUG_ON(!atomic_read(&mnt->mnt_longrefs));
751 if (atomic_add_unless(&mnt->mnt_longrefs, -1, 1))
752 return;
753 }
754
755 br_write_lock(vfsmount_lock);
756 if (!longrefs)
757 mnt_dec_count(mnt);
758 else
759 atomic_dec(&mnt->mnt_longrefs);
760 if (mnt_get_count(mnt)) {
761 br_write_unlock(vfsmount_lock);
762 return;
763 }
764 if (unlikely(mnt->mnt_pinned)) {
765 mnt_add_count(mnt, mnt->mnt_pinned + 1);
766 mnt->mnt_pinned = 0;
767 br_write_unlock(vfsmount_lock);
768 acct_auto_close_mnt(mnt);
769 goto put_again;
770 }
771 br_write_unlock(vfsmount_lock);
772 mntfree(mnt);
773}
774#else
775static inline void __mntput(struct vfsmount *mnt, int longrefs)
776{
777put_again:
778 mnt_dec_count(mnt);
779 if (likely(mnt_get_count(mnt)))
Nick Piggin99b7db72010-08-18 04:37:39 +1000780 return;
781 br_write_lock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100782 if (unlikely(mnt->mnt_pinned)) {
783 mnt_add_count(mnt, mnt->mnt_pinned + 1);
784 mnt->mnt_pinned = 0;
Nick Piggin99b7db72010-08-18 04:37:39 +1000785 br_write_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100786 acct_auto_close_mnt(mnt);
787 goto put_again;
Al Viro7b7b1ac2005-11-07 17:13:39 -0500788 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000789 br_write_unlock(vfsmount_lock);
Nick Pigginb3e19d92011-01-07 17:50:11 +1100790 mntfree(mnt);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500791}
Nick Pigginb3e19d92011-01-07 17:50:11 +1100792#endif
793
794static void mntput_no_expire(struct vfsmount *mnt)
795{
796 __mntput(mnt, 0);
797}
798
799void mntput(struct vfsmount *mnt)
800{
801 if (mnt) {
802 /* avoid cacheline pingpong, hope gcc doesn't get "smart" */
803 if (unlikely(mnt->mnt_expiry_mark))
804 mnt->mnt_expiry_mark = 0;
805 __mntput(mnt, 0);
806 }
807}
808EXPORT_SYMBOL(mntput);
809
810struct vfsmount *mntget(struct vfsmount *mnt)
811{
812 if (mnt)
813 mnt_inc_count(mnt);
814 return mnt;
815}
816EXPORT_SYMBOL(mntget);
817
818void mntput_long(struct vfsmount *mnt)
819{
820#ifdef CONFIG_SMP
821 if (mnt) {
822 /* avoid cacheline pingpong, hope gcc doesn't get "smart" */
823 if (unlikely(mnt->mnt_expiry_mark))
824 mnt->mnt_expiry_mark = 0;
825 __mntput(mnt, 1);
826 }
827#else
828 mntput(mnt);
829#endif
830}
831EXPORT_SYMBOL(mntput_long);
832
833struct vfsmount *mntget_long(struct vfsmount *mnt)
834{
835#ifdef CONFIG_SMP
836 if (mnt)
837 atomic_inc(&mnt->mnt_longrefs);
838 return mnt;
839#else
840 return mntget(mnt);
841#endif
842}
843EXPORT_SYMBOL(mntget_long);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500844
845void mnt_pin(struct vfsmount *mnt)
846{
Nick Piggin99b7db72010-08-18 04:37:39 +1000847 br_write_lock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500848 mnt->mnt_pinned++;
Nick Piggin99b7db72010-08-18 04:37:39 +1000849 br_write_unlock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500850}
Al Viro7b7b1ac2005-11-07 17:13:39 -0500851EXPORT_SYMBOL(mnt_pin);
852
853void mnt_unpin(struct vfsmount *mnt)
854{
Nick Piggin99b7db72010-08-18 04:37:39 +1000855 br_write_lock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500856 if (mnt->mnt_pinned) {
Nick Pigginb3e19d92011-01-07 17:50:11 +1100857 mnt_inc_count(mnt);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500858 mnt->mnt_pinned--;
859 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000860 br_write_unlock(vfsmount_lock);
Al Viro7b7b1ac2005-11-07 17:13:39 -0500861}
Al Viro7b7b1ac2005-11-07 17:13:39 -0500862EXPORT_SYMBOL(mnt_unpin);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800864static inline void mangle(struct seq_file *m, const char *s)
865{
866 seq_escape(m, s, " \t\n\\");
867}
868
869/*
870 * Simple .show_options callback for filesystems which don't want to
871 * implement more complex mount option showing.
872 *
873 * See also save_mount_options().
874 */
875int generic_show_options(struct seq_file *m, struct vfsmount *mnt)
876{
Al Viro2a32ceb2009-05-08 16:05:57 -0400877 const char *options;
878
879 rcu_read_lock();
880 options = rcu_dereference(mnt->mnt_sb->s_options);
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800881
882 if (options != NULL && options[0]) {
883 seq_putc(m, ',');
884 mangle(m, options);
885 }
Al Viro2a32ceb2009-05-08 16:05:57 -0400886 rcu_read_unlock();
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800887
888 return 0;
889}
890EXPORT_SYMBOL(generic_show_options);
891
892/*
893 * If filesystem uses generic_show_options(), this function should be
894 * called from the fill_super() callback.
895 *
896 * The .remount_fs callback usually needs to be handled in a special
897 * way, to make sure, that previous options are not overwritten if the
898 * remount fails.
899 *
900 * Also note, that if the filesystem's .remount_fs function doesn't
901 * reset all options to their default value, but changes only newly
902 * given options, then the displayed options will not reflect reality
903 * any more.
904 */
905void save_mount_options(struct super_block *sb, char *options)
906{
Al Viro2a32ceb2009-05-08 16:05:57 -0400907 BUG_ON(sb->s_options);
908 rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL));
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800909}
910EXPORT_SYMBOL(save_mount_options);
911
Al Viro2a32ceb2009-05-08 16:05:57 -0400912void replace_mount_options(struct super_block *sb, char *options)
913{
914 char *old = sb->s_options;
915 rcu_assign_pointer(sb->s_options, options);
916 if (old) {
917 synchronize_rcu();
918 kfree(old);
919 }
920}
921EXPORT_SYMBOL(replace_mount_options);
922
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100923#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924/* iterator */
925static void *m_start(struct seq_file *m, loff_t *pos)
926{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100927 struct proc_mounts *p = m->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
Ram Pai390c6842005-11-07 17:17:51 -0500929 down_read(&namespace_sem);
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100930 return seq_list_start(&p->ns->list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931}
932
933static void *m_next(struct seq_file *m, void *v, loff_t *pos)
934{
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100935 struct proc_mounts *p = m->private;
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700936
Miklos Szeredia1a2c402008-03-27 13:06:24 +0100937 return seq_list_next(v, &p->ns->list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938}
939
940static void m_stop(struct seq_file *m, void *v)
941{
Ram Pai390c6842005-11-07 17:17:51 -0500942 up_read(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943}
944
Al Viro9f5596a2010-02-05 00:40:25 -0500945int mnt_had_events(struct proc_mounts *p)
946{
947 struct mnt_namespace *ns = p->ns;
948 int res = 0;
949
Nick Piggin99b7db72010-08-18 04:37:39 +1000950 br_read_lock(vfsmount_lock);
Al Viro9f5596a2010-02-05 00:40:25 -0500951 if (p->event != ns->event) {
952 p->event = ns->event;
953 res = 1;
954 }
Nick Piggin99b7db72010-08-18 04:37:39 +1000955 br_read_unlock(vfsmount_lock);
Al Viro9f5596a2010-02-05 00:40:25 -0500956
957 return res;
958}
959
Ram Pai2d4d4862008-03-27 13:06:25 +0100960struct proc_fs_info {
961 int flag;
962 const char *str;
963};
964
Eric Paris2069f452008-07-04 09:47:13 +1000965static int show_sb_opts(struct seq_file *m, struct super_block *sb)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966{
Ram Pai2d4d4862008-03-27 13:06:25 +0100967 static const struct proc_fs_info fs_info[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 { MS_SYNCHRONOUS, ",sync" },
969 { MS_DIRSYNC, ",dirsync" },
970 { MS_MANDLOCK, ",mand" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 { 0, NULL }
972 };
Ram Pai2d4d4862008-03-27 13:06:25 +0100973 const struct proc_fs_info *fs_infop;
974
975 for (fs_infop = fs_info; fs_infop->flag; fs_infop++) {
976 if (sb->s_flags & fs_infop->flag)
977 seq_puts(m, fs_infop->str);
978 }
Eric Paris2069f452008-07-04 09:47:13 +1000979
980 return security_sb_show_options(m, sb);
Ram Pai2d4d4862008-03-27 13:06:25 +0100981}
982
983static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt)
984{
985 static const struct proc_fs_info mnt_info[] = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 { MNT_NOSUID, ",nosuid" },
987 { MNT_NODEV, ",nodev" },
988 { MNT_NOEXEC, ",noexec" },
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -0800989 { MNT_NOATIME, ",noatime" },
990 { MNT_NODIRATIME, ",nodiratime" },
Valerie Henson47ae32d2006-12-13 00:34:34 -0800991 { MNT_RELATIME, ",relatime" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 { 0, NULL }
993 };
Ram Pai2d4d4862008-03-27 13:06:25 +0100994 const struct proc_fs_info *fs_infop;
995
996 for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) {
997 if (mnt->mnt_flags & fs_infop->flag)
998 seq_puts(m, fs_infop->str);
999 }
1000}
1001
1002static void show_type(struct seq_file *m, struct super_block *sb)
1003{
1004 mangle(m, sb->s_type->name);
1005 if (sb->s_subtype && sb->s_subtype[0]) {
1006 seq_putc(m, '.');
1007 mangle(m, sb->s_subtype);
1008 }
1009}
1010
1011static int show_vfsmnt(struct seq_file *m, void *v)
1012{
1013 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
1014 int err = 0;
Jan Blunckc32c2f62008-02-14 19:38:43 -08001015 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016
1017 mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none");
1018 seq_putc(m, ' ');
Jan Blunckc32c2f62008-02-14 19:38:43 -08001019 seq_path(m, &mnt_path, " \t\n\\");
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 seq_putc(m, ' ');
Ram Pai2d4d4862008-03-27 13:06:25 +01001021 show_type(m, mnt->mnt_sb);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001022 seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw");
Eric Paris2069f452008-07-04 09:47:13 +10001023 err = show_sb_opts(m, mnt->mnt_sb);
1024 if (err)
1025 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001026 show_mnt_opts(m, mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 if (mnt->mnt_sb->s_op->show_options)
1028 err = mnt->mnt_sb->s_op->show_options(m, mnt);
1029 seq_puts(m, " 0 0\n");
Eric Paris2069f452008-07-04 09:47:13 +10001030out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 return err;
1032}
1033
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001034const struct seq_operations mounts_op = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 .start = m_start,
1036 .next = m_next,
1037 .stop = m_stop,
1038 .show = show_vfsmnt
1039};
1040
Ram Pai2d4d4862008-03-27 13:06:25 +01001041static int show_mountinfo(struct seq_file *m, void *v)
1042{
1043 struct proc_mounts *p = m->private;
1044 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
1045 struct super_block *sb = mnt->mnt_sb;
1046 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
1047 struct path root = p->root;
1048 int err = 0;
1049
1050 seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id,
1051 MAJOR(sb->s_dev), MINOR(sb->s_dev));
1052 seq_dentry(m, mnt->mnt_root, " \t\n\\");
1053 seq_putc(m, ' ');
1054 seq_path_root(m, &mnt_path, &root, " \t\n\\");
1055 if (root.mnt != p->root.mnt || root.dentry != p->root.dentry) {
1056 /*
1057 * Mountpoint is outside root, discard that one. Ugly,
1058 * but less so than trying to do that in iterator in a
1059 * race-free way (due to renames).
1060 */
1061 return SEQ_SKIP;
1062 }
1063 seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw");
1064 show_mnt_opts(m, mnt);
1065
1066 /* Tagged fields ("foo:X" or "bar") */
1067 if (IS_MNT_SHARED(mnt))
1068 seq_printf(m, " shared:%i", mnt->mnt_group_id);
Miklos Szeredi97e7e0f2008-03-27 13:06:26 +01001069 if (IS_MNT_SLAVE(mnt)) {
1070 int master = mnt->mnt_master->mnt_group_id;
1071 int dom = get_dominating_id(mnt, &p->root);
1072 seq_printf(m, " master:%i", master);
1073 if (dom && dom != master)
1074 seq_printf(m, " propagate_from:%i", dom);
1075 }
Ram Pai2d4d4862008-03-27 13:06:25 +01001076 if (IS_MNT_UNBINDABLE(mnt))
1077 seq_puts(m, " unbindable");
1078
1079 /* Filesystem specific data */
1080 seq_puts(m, " - ");
1081 show_type(m, sb);
1082 seq_putc(m, ' ');
1083 mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none");
1084 seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw");
Eric Paris2069f452008-07-04 09:47:13 +10001085 err = show_sb_opts(m, sb);
1086 if (err)
1087 goto out;
Ram Pai2d4d4862008-03-27 13:06:25 +01001088 if (sb->s_op->show_options)
1089 err = sb->s_op->show_options(m, mnt);
1090 seq_putc(m, '\n');
Eric Paris2069f452008-07-04 09:47:13 +10001091out:
Ram Pai2d4d4862008-03-27 13:06:25 +01001092 return err;
1093}
1094
1095const struct seq_operations mountinfo_op = {
1096 .start = m_start,
1097 .next = m_next,
1098 .stop = m_stop,
1099 .show = show_mountinfo,
1100};
1101
Chuck Leverb4629fe2006-03-20 13:44:12 -05001102static int show_vfsstat(struct seq_file *m, void *v)
1103{
Pavel Emelianovb0765fb2007-07-15 23:39:55 -07001104 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
Jan Blunckc32c2f62008-02-14 19:38:43 -08001105 struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt };
Chuck Leverb4629fe2006-03-20 13:44:12 -05001106 int err = 0;
1107
1108 /* device */
1109 if (mnt->mnt_devname) {
1110 seq_puts(m, "device ");
1111 mangle(m, mnt->mnt_devname);
1112 } else
1113 seq_puts(m, "no device");
1114
1115 /* mount point */
1116 seq_puts(m, " mounted on ");
Jan Blunckc32c2f62008-02-14 19:38:43 -08001117 seq_path(m, &mnt_path, " \t\n\\");
Chuck Leverb4629fe2006-03-20 13:44:12 -05001118 seq_putc(m, ' ');
1119
1120 /* file system type */
1121 seq_puts(m, "with fstype ");
Ram Pai2d4d4862008-03-27 13:06:25 +01001122 show_type(m, mnt->mnt_sb);
Chuck Leverb4629fe2006-03-20 13:44:12 -05001123
1124 /* optional statistics */
1125 if (mnt->mnt_sb->s_op->show_stats) {
1126 seq_putc(m, ' ');
1127 err = mnt->mnt_sb->s_op->show_stats(m, mnt);
1128 }
1129
1130 seq_putc(m, '\n');
1131 return err;
1132}
1133
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001134const struct seq_operations mountstats_op = {
Chuck Leverb4629fe2006-03-20 13:44:12 -05001135 .start = m_start,
1136 .next = m_next,
1137 .stop = m_stop,
1138 .show = show_vfsstat,
1139};
Miklos Szeredia1a2c402008-03-27 13:06:24 +01001140#endif /* CONFIG_PROC_FS */
Chuck Leverb4629fe2006-03-20 13:44:12 -05001141
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142/**
1143 * may_umount_tree - check if a mount tree is busy
1144 * @mnt: root of mount tree
1145 *
1146 * This is called to check if a tree of mounts has any
1147 * open files, pwds, chroots or sub mounts that are
1148 * busy.
1149 */
1150int may_umount_tree(struct vfsmount *mnt)
1151{
Ram Pai36341f62005-11-07 17:17:22 -05001152 int actual_refs = 0;
1153 int minimum_refs = 0;
1154 struct vfsmount *p;
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);
Ram Pai36341f62005-11-07 17:17:22 -05001158 for (p = mnt; p; p = next_mnt(p, mnt)) {
Nick Pigginb3e19d92011-01-07 17:50:11 +11001159 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);
Ram Paia05964f2005-11-07 17:20:17 -05001190 if (propagate_mount_busy(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{
Ram Pai70fbcdf2005-11-07 17:17:04 -05001201 struct vfsmount *mnt;
Miklos Szeredibf066c72006-01-08 01:03:19 -08001202 while (!list_empty(head)) {
Pavel Emelianovb5e61812007-05-08 00:30:19 -07001203 mnt = list_first_entry(head, struct vfsmount, mnt_hash);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001204 list_del_init(&mnt->mnt_hash);
1205 if (mnt->mnt_parent != mnt) {
1206 struct dentry *dentry;
1207 struct vfsmount *m;
Nick Piggin99b7db72010-08-18 04:37:39 +10001208
1209 br_write_lock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001210 dentry = mnt->mnt_mountpoint;
1211 m = mnt->mnt_parent;
1212 mnt->mnt_mountpoint = mnt->mnt_root;
1213 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);
1217 mntput(m);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 }
Nick Pigginb3e19d92011-01-07 17:50:11 +11001219 mntput_long(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 */
Ram Paia05964f2005-11-07 17:20:17 -05001227void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill)
Ram Pai70fbcdf2005-11-07 17:17:04 -05001228{
1229 struct vfsmount *p;
1230
Akinobu Mita1bfba4e2006-06-26 00:24:40 -07001231 for (p = mnt; p; p = next_mnt(p, mnt))
1232 list_move(&p->mnt_hash, kill);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001233
Ram Paia05964f2005-11-07 17:20:17 -05001234 if (propagate)
1235 propagate_umount(kill);
1236
Ram Pai70fbcdf2005-11-07 17:17:04 -05001237 list_for_each_entry(p, kill, mnt_hash) {
1238 list_del_init(&p->mnt_expire);
1239 list_del_init(&p->mnt_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001240 __touch_mnt_namespace(p->mnt_ns);
1241 p->mnt_ns = NULL;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001242 list_del_init(&p->mnt_child);
Al Viro7c4b93d2008-03-21 23:59:49 -04001243 if (p->mnt_parent != p) {
1244 p->mnt_parent->mnt_ghosts++;
Nick Piggin5f57cbc2011-01-07 17:49:54 +11001245 dentry_reset_mounted(p->mnt_parent, p->mnt_mountpoint);
Al Viro7c4b93d2008-03-21 23:59:49 -04001246 }
Ram Paia05964f2005-11-07 17:20:17 -05001247 change_mnt_propagation(p, MS_PRIVATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 }
1249}
1250
Al Viroc35038b2008-03-22 00:46:23 -04001251static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts);
1252
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253static int do_umount(struct vfsmount *mnt, int flags)
1254{
Ram Paib58fed82005-11-07 17:16:09 -05001255 struct super_block *sb = mnt->mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 int retval;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001257 LIST_HEAD(umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
1259 retval = security_sb_umount(mnt, flags);
1260 if (retval)
1261 return retval;
1262
1263 /*
1264 * Allow userspace to request a mountpoint be expired rather than
1265 * unmounting unconditionally. Unmount only happens if:
1266 * (1) the mark is already set (the mark is cleared by mntput())
1267 * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount]
1268 */
1269 if (flags & MNT_EXPIRE) {
Jan Blunck6ac08c32008-02-14 19:34:38 -08001270 if (mnt == current->fs->root.mnt ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 flags & (MNT_FORCE | MNT_DETACH))
1272 return -EINVAL;
1273
Nick Pigginb3e19d92011-01-07 17:50:11 +11001274 /*
1275 * probably don't strictly need the lock here if we examined
1276 * all race cases, but it's a slowpath.
1277 */
1278 br_write_lock(vfsmount_lock);
1279 if (mnt_get_count(mnt) != 2) {
1280 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 return -EBUSY;
Nick Pigginb3e19d92011-01-07 17:50:11 +11001282 }
1283 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284
1285 if (!xchg(&mnt->mnt_expiry_mark, 1))
1286 return -EAGAIN;
1287 }
1288
1289 /*
1290 * If we may have to abort operations to get out of this
1291 * mount, and they will themselves hold resources we must
1292 * allow the fs to do things. In the Unix tradition of
1293 * 'Gee thats tricky lets do it in userspace' the umount_begin
1294 * might fail to complete on the first run through as other tasks
1295 * must return, and the like. Thats for the mount program to worry
1296 * about for the moment.
1297 */
1298
Al Viro42faad92008-04-24 07:21:56 -04001299 if (flags & MNT_FORCE && sb->s_op->umount_begin) {
Al Viro42faad92008-04-24 07:21:56 -04001300 sb->s_op->umount_begin(sb);
Al Viro42faad92008-04-24 07:21:56 -04001301 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302
1303 /*
1304 * No sense to grab the lock for this test, but test itself looks
1305 * somewhat bogus. Suggestions for better replacement?
1306 * Ho-hum... In principle, we might treat that as umount + switch
1307 * to rootfs. GC would eventually take care of the old vfsmount.
1308 * Actually it makes sense, especially if rootfs would contain a
1309 * /reboot - static binary that would close all descriptors and
1310 * call reboot(9). Then init(8) could umount root and exec /reboot.
1311 */
Jan Blunck6ac08c32008-02-14 19:34:38 -08001312 if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 /*
1314 * Special case for "unmounting" root ...
1315 * we just try to remount it readonly.
1316 */
1317 down_write(&sb->s_umount);
Al Viro4aa98cf2009-05-08 13:36:58 -04001318 if (!(sb->s_flags & MS_RDONLY))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 retval = do_remount_sb(sb, MS_RDONLY, NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320 up_write(&sb->s_umount);
1321 return retval;
1322 }
1323
Ram Pai390c6842005-11-07 17:17:51 -05001324 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001325 br_write_lock(vfsmount_lock);
Al Viro5addc5d2005-11-07 17:15:49 -05001326 event++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
Al Viroc35038b2008-03-22 00:46:23 -04001328 if (!(flags & MNT_DETACH))
1329 shrink_submounts(mnt, &umount_list);
1330
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 retval = -EBUSY;
Ram Paia05964f2005-11-07 17:20:17 -05001332 if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 if (!list_empty(&mnt->mnt_list))
Ram Paia05964f2005-11-07 17:20:17 -05001334 umount_tree(mnt, 1, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 retval = 0;
1336 }
Nick Piggin99b7db72010-08-18 04:37:39 +10001337 br_write_unlock(vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05001338 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001339 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 return retval;
1341}
1342
1343/*
1344 * Now umount can handle mount points as well as block devices.
1345 * This is important for filesystems which use unnamed block devices.
1346 *
1347 * We now support a flag for forced unmount like the other 'big iron'
1348 * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD
1349 */
1350
Heiko Carstensbdc480e2009-01-14 14:14:12 +01001351SYSCALL_DEFINE2(umount, char __user *, name, int, flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352{
Al Viro2d8f3032008-07-22 09:59:21 -04001353 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 int retval;
Miklos Szeredidb1f05b2010-02-10 12:15:53 +01001355 int lookup_flags = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
Miklos Szeredidb1f05b2010-02-10 12:15:53 +01001357 if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW))
1358 return -EINVAL;
1359
1360 if (!(flags & UMOUNT_NOFOLLOW))
1361 lookup_flags |= LOOKUP_FOLLOW;
1362
1363 retval = user_path_at(AT_FDCWD, name, lookup_flags, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 if (retval)
1365 goto out;
1366 retval = -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -04001367 if (path.dentry != path.mnt->mnt_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 goto dput_and_out;
Al Viro2d8f3032008-07-22 09:59:21 -04001369 if (!check_mnt(path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 goto dput_and_out;
1371
1372 retval = -EPERM;
1373 if (!capable(CAP_SYS_ADMIN))
1374 goto dput_and_out;
1375
Al Viro2d8f3032008-07-22 09:59:21 -04001376 retval = do_umount(path.mnt, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377dput_and_out:
Jan Blunck429731b2008-02-14 19:34:31 -08001378 /* we mustn't call path_put() as that would clear mnt_expiry_mark */
Al Viro2d8f3032008-07-22 09:59:21 -04001379 dput(path.dentry);
1380 mntput_no_expire(path.mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381out:
1382 return retval;
1383}
1384
1385#ifdef __ARCH_WANT_SYS_OLDUMOUNT
1386
1387/*
Ram Paib58fed82005-11-07 17:16:09 -05001388 * The 2.0 compatible umount. No flags.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 */
Heiko Carstensbdc480e2009-01-14 14:14:12 +01001390SYSCALL_DEFINE1(oldumount, char __user *, name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391{
Ram Paib58fed82005-11-07 17:16:09 -05001392 return sys_umount(name, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393}
1394
1395#endif
1396
Al Viro2d92ab32008-08-02 00:51:11 -04001397static int mount_is_safe(struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398{
1399 if (capable(CAP_SYS_ADMIN))
1400 return 0;
1401 return -EPERM;
1402#ifdef notyet
Al Viro2d92ab32008-08-02 00:51:11 -04001403 if (S_ISLNK(path->dentry->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404 return -EPERM;
Al Viro2d92ab32008-08-02 00:51:11 -04001405 if (path->dentry->d_inode->i_mode & S_ISVTX) {
David Howellsda9592e2008-11-14 10:39:05 +11001406 if (current_uid() != path->dentry->d_inode->i_uid)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 return -EPERM;
1408 }
Al Viro2d92ab32008-08-02 00:51:11 -04001409 if (inode_permission(path->dentry->d_inode, MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 return -EPERM;
1411 return 0;
1412#endif
1413}
1414
Ram Paib90fa9a2005-11-07 17:19:50 -05001415struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry,
Ram Pai36341f62005-11-07 17:17:22 -05001416 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
1418 struct vfsmount *res, *p, *q, *r, *s;
Al Viro1a390682008-03-21 20:48:19 -04001419 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
Ram Pai9676f0c2005-11-07 17:21:20 -05001421 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt))
1422 return NULL;
1423
Ram Pai36341f62005-11-07 17:17:22 -05001424 res = q = clone_mnt(mnt, dentry, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 if (!q)
1426 goto Enomem;
1427 q->mnt_mountpoint = mnt->mnt_mountpoint;
1428
1429 p = mnt;
Domen Puncerfdadd652005-09-10 00:27:07 -07001430 list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) {
Jan Blunck7ec02ef2008-04-29 00:59:40 -07001431 if (!is_subdir(r->mnt_mountpoint, dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 continue;
1433
1434 for (s = r; s; s = next_mnt(s, r)) {
Ram Pai9676f0c2005-11-07 17:21:20 -05001435 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) {
1436 s = skip_mnt_tree(s);
1437 continue;
1438 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 while (p != s->mnt_parent) {
1440 p = p->mnt_parent;
1441 q = q->mnt_parent;
1442 }
1443 p = s;
Al Viro1a390682008-03-21 20:48:19 -04001444 path.mnt = q;
1445 path.dentry = p->mnt_mountpoint;
Ram Pai36341f62005-11-07 17:17:22 -05001446 q = clone_mnt(p, p->mnt_root, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 if (!q)
1448 goto Enomem;
Nick Piggin99b7db72010-08-18 04:37:39 +10001449 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450 list_add_tail(&q->mnt_list, &res->mnt_list);
Al Viro1a390682008-03-21 20:48:19 -04001451 attach_mnt(q, &path);
Nick Piggin99b7db72010-08-18 04:37:39 +10001452 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 }
1454 }
1455 return res;
Ram Paib58fed82005-11-07 17:16:09 -05001456Enomem:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 if (res) {
Ram Pai70fbcdf2005-11-07 17:17:04 -05001458 LIST_HEAD(umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001459 br_write_lock(vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -05001460 umount_tree(res, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001461 br_write_unlock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001462 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 }
1464 return NULL;
1465}
1466
Al Viro589ff872009-04-18 03:28:19 -04001467struct vfsmount *collect_mounts(struct path *path)
Al Viro8aec0802007-06-07 12:20:32 -04001468{
1469 struct vfsmount *tree;
Al Viro1a60a282008-03-22 16:19:49 -04001470 down_write(&namespace_sem);
Al Viro589ff872009-04-18 03:28:19 -04001471 tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE);
Al Viro1a60a282008-03-22 16:19:49 -04001472 up_write(&namespace_sem);
Al Viro8aec0802007-06-07 12:20:32 -04001473 return tree;
1474}
1475
1476void drop_collected_mounts(struct vfsmount *mnt)
1477{
1478 LIST_HEAD(umount_list);
Al Viro1a60a282008-03-22 16:19:49 -04001479 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001480 br_write_lock(vfsmount_lock);
Al Viro8aec0802007-06-07 12:20:32 -04001481 umount_tree(mnt, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001482 br_write_unlock(vfsmount_lock);
Al Viro1a60a282008-03-22 16:19:49 -04001483 up_write(&namespace_sem);
Al Viro8aec0802007-06-07 12:20:32 -04001484 release_mounts(&umount_list);
1485}
1486
Al Viro1f707132010-01-30 22:51:25 -05001487int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg,
1488 struct vfsmount *root)
1489{
1490 struct vfsmount *mnt;
1491 int res = f(root, arg);
1492 if (res)
1493 return res;
1494 list_for_each_entry(mnt, &root->mnt_list, mnt_list) {
1495 res = f(mnt, arg);
1496 if (res)
1497 return res;
1498 }
1499 return 0;
1500}
1501
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001502static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end)
1503{
1504 struct vfsmount *p;
1505
1506 for (p = mnt; p != end; p = next_mnt(p, mnt)) {
1507 if (p->mnt_group_id && !IS_MNT_SHARED(p))
1508 mnt_release_group_id(p);
1509 }
1510}
1511
1512static int invent_group_ids(struct vfsmount *mnt, bool recurse)
1513{
1514 struct vfsmount *p;
1515
1516 for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) {
1517 if (!p->mnt_group_id && !IS_MNT_SHARED(p)) {
1518 int err = mnt_alloc_group_id(p);
1519 if (err) {
1520 cleanup_group_ids(mnt, p);
1521 return err;
1522 }
1523 }
1524 }
1525
1526 return 0;
1527}
1528
Ram Paib90fa9a2005-11-07 17:19:50 -05001529/*
1530 * @source_mnt : mount tree to be attached
Ram Pai21444402005-11-07 17:20:03 -05001531 * @nd : place the mount tree @source_mnt is attached
1532 * @parent_nd : if non-null, detach the source_mnt from its parent and
1533 * store the parent mount and mountpoint dentry.
1534 * (done when source_mnt is moved)
Ram Paib90fa9a2005-11-07 17:19:50 -05001535 *
1536 * NOTE: in the table below explains the semantics when a source mount
1537 * of a given type is attached to a destination mount of a given type.
Ram Pai9676f0c2005-11-07 17:21:20 -05001538 * ---------------------------------------------------------------------------
1539 * | BIND MOUNT OPERATION |
1540 * |**************************************************************************
1541 * | source-->| shared | private | slave | unbindable |
1542 * | dest | | | | |
1543 * | | | | | | |
1544 * | v | | | | |
1545 * |**************************************************************************
1546 * | shared | shared (++) | shared (+) | shared(+++)| invalid |
1547 * | | | | | |
1548 * |non-shared| shared (+) | private | slave (*) | invalid |
1549 * ***************************************************************************
Ram Paib90fa9a2005-11-07 17:19:50 -05001550 * A bind operation clones the source mount and mounts the clone on the
1551 * destination mount.
1552 *
1553 * (++) the cloned mount is propagated to all the mounts in the propagation
1554 * tree of the destination mount and the cloned mount is added to
1555 * the peer group of the source mount.
1556 * (+) the cloned mount is created under the destination mount and is marked
1557 * as shared. The cloned mount is added to the peer group of the source
1558 * mount.
Ram Pai5afe0022005-11-07 17:21:01 -05001559 * (+++) the mount is propagated to all the mounts in the propagation tree
1560 * of the destination mount and the cloned mount is made slave
1561 * of the same master as that of the source mount. The cloned mount
1562 * is marked as 'shared and slave'.
1563 * (*) the cloned mount is made a slave of the same master as that of the
1564 * source mount.
1565 *
Ram Pai9676f0c2005-11-07 17:21:20 -05001566 * ---------------------------------------------------------------------------
1567 * | MOVE MOUNT OPERATION |
1568 * |**************************************************************************
1569 * | source-->| shared | private | slave | unbindable |
1570 * | dest | | | | |
1571 * | | | | | | |
1572 * | v | | | | |
1573 * |**************************************************************************
1574 * | shared | shared (+) | shared (+) | shared(+++) | invalid |
1575 * | | | | | |
1576 * |non-shared| shared (+*) | private | slave (*) | unbindable |
1577 * ***************************************************************************
Ram Pai5afe0022005-11-07 17:21:01 -05001578 *
1579 * (+) the mount is moved to the destination. And is then propagated to
1580 * all the mounts in the propagation tree of the destination mount.
Ram Pai21444402005-11-07 17:20:03 -05001581 * (+*) the mount is moved to the destination.
Ram Pai5afe0022005-11-07 17:21:01 -05001582 * (+++) the mount is moved to the destination and is then propagated to
1583 * all the mounts belonging to the destination mount's propagation tree.
1584 * the mount is marked as 'shared and slave'.
1585 * (*) the mount continues to be a slave at the new location.
Ram Paib90fa9a2005-11-07 17:19:50 -05001586 *
1587 * if the source mount is a tree, the operations explained above is
1588 * applied to each mount in the tree.
1589 * Must be called without spinlocks held, since this function can sleep
1590 * in allocations.
1591 */
1592static int attach_recursive_mnt(struct vfsmount *source_mnt,
Al Viro1a390682008-03-21 20:48:19 -04001593 struct path *path, struct path *parent_path)
Ram Paib90fa9a2005-11-07 17:19:50 -05001594{
1595 LIST_HEAD(tree_list);
Al Viro1a390682008-03-21 20:48:19 -04001596 struct vfsmount *dest_mnt = path->mnt;
1597 struct dentry *dest_dentry = path->dentry;
Ram Paib90fa9a2005-11-07 17:19:50 -05001598 struct vfsmount *child, *p;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001599 int err;
Ram Paib90fa9a2005-11-07 17:19:50 -05001600
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001601 if (IS_MNT_SHARED(dest_mnt)) {
1602 err = invent_group_ids(source_mnt, true);
1603 if (err)
1604 goto out;
1605 }
1606 err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list);
1607 if (err)
1608 goto out_cleanup_ids;
Ram Paib90fa9a2005-11-07 17:19:50 -05001609
Nick Piggin99b7db72010-08-18 04:37:39 +10001610 br_write_lock(vfsmount_lock);
Al Virodf1a1ad2010-01-16 12:57:40 -05001611
Ram Paib90fa9a2005-11-07 17:19:50 -05001612 if (IS_MNT_SHARED(dest_mnt)) {
1613 for (p = source_mnt; p; p = next_mnt(p, source_mnt))
1614 set_mnt_shared(p);
1615 }
Al Viro1a390682008-03-21 20:48:19 -04001616 if (parent_path) {
1617 detach_mnt(source_mnt, parent_path);
1618 attach_mnt(source_mnt, path);
Al Viroe5d67f02009-04-07 12:15:39 -04001619 touch_mnt_namespace(parent_path->mnt->mnt_ns);
Ram Pai21444402005-11-07 17:20:03 -05001620 } else {
1621 mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt);
1622 commit_tree(source_mnt);
1623 }
Ram Paib90fa9a2005-11-07 17:19:50 -05001624
1625 list_for_each_entry_safe(child, p, &tree_list, mnt_hash) {
1626 list_del_init(&child->mnt_hash);
1627 commit_tree(child);
1628 }
Nick Piggin99b7db72010-08-18 04:37:39 +10001629 br_write_unlock(vfsmount_lock);
1630
Ram Paib90fa9a2005-11-07 17:19:50 -05001631 return 0;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001632
1633 out_cleanup_ids:
1634 if (IS_MNT_SHARED(dest_mnt))
1635 cleanup_group_ids(source_mnt, NULL);
1636 out:
1637 return err;
Ram Paib90fa9a2005-11-07 17:19:50 -05001638}
1639
Al Viro8c3ee422008-03-22 18:00:39 -04001640static int graft_tree(struct vfsmount *mnt, struct path *path)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641{
1642 int err;
1643 if (mnt->mnt_sb->s_flags & MS_NOUSER)
1644 return -EINVAL;
1645
Al Viro8c3ee422008-03-22 18:00:39 -04001646 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 S_ISDIR(mnt->mnt_root->d_inode->i_mode))
1648 return -ENOTDIR;
1649
1650 err = -ENOENT;
Al Viro8c3ee422008-03-22 18:00:39 -04001651 mutex_lock(&path->dentry->d_inode->i_mutex);
Al Virod83c49f2010-04-30 17:17:09 -04001652 if (cant_mount(path->dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653 goto out_unlock;
1654
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04001655 if (!d_unlinked(path->dentry))
Al Viro8c3ee422008-03-22 18:00:39 -04001656 err = attach_recursive_mnt(mnt, path, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657out_unlock:
Al Viro8c3ee422008-03-22 18:00:39 -04001658 mutex_unlock(&path->dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 return err;
1660}
1661
1662/*
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001663 * Sanity check the flags to change_mnt_propagation.
1664 */
1665
1666static int flags_to_propagation_type(int flags)
1667{
1668 int type = flags & ~MS_REC;
1669
1670 /* Fail if any non-propagation flags are set */
1671 if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
1672 return 0;
1673 /* Only one propagation flag should be set */
1674 if (!is_power_of_2(type))
1675 return 0;
1676 return type;
1677}
1678
1679/*
Ram Pai07b20882005-11-07 17:19:07 -05001680 * recursively change the type of the mountpoint.
1681 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001682static int do_change_type(struct path *path, int flag)
Ram Pai07b20882005-11-07 17:19:07 -05001683{
Al Viro2d92ab32008-08-02 00:51:11 -04001684 struct vfsmount *m, *mnt = path->mnt;
Ram Pai07b20882005-11-07 17:19:07 -05001685 int recurse = flag & MS_REC;
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001686 int type;
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001687 int err = 0;
Ram Pai07b20882005-11-07 17:19:07 -05001688
Miklos Szerediee6f9582007-05-08 00:30:40 -07001689 if (!capable(CAP_SYS_ADMIN))
1690 return -EPERM;
1691
Al Viro2d92ab32008-08-02 00:51:11 -04001692 if (path->dentry != path->mnt->mnt_root)
Ram Pai07b20882005-11-07 17:19:07 -05001693 return -EINVAL;
1694
Valerie Aurora7a2e8a82010-08-26 11:07:22 -07001695 type = flags_to_propagation_type(flag);
1696 if (!type)
1697 return -EINVAL;
1698
Ram Pai07b20882005-11-07 17:19:07 -05001699 down_write(&namespace_sem);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001700 if (type == MS_SHARED) {
1701 err = invent_group_ids(mnt, recurse);
1702 if (err)
1703 goto out_unlock;
1704 }
1705
Nick Piggin99b7db72010-08-18 04:37:39 +10001706 br_write_lock(vfsmount_lock);
Ram Pai07b20882005-11-07 17:19:07 -05001707 for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL))
1708 change_mnt_propagation(m, type);
Nick Piggin99b7db72010-08-18 04:37:39 +10001709 br_write_unlock(vfsmount_lock);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001710
1711 out_unlock:
Ram Pai07b20882005-11-07 17:19:07 -05001712 up_write(&namespace_sem);
Miklos Szeredi719f5d72008-03-27 13:06:23 +01001713 return err;
Ram Pai07b20882005-11-07 17:19:07 -05001714}
1715
1716/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 * do loopback mount.
1718 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001719static int do_loopback(struct path *path, char *old_name,
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001720 int recurse)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721{
Al Viro2d92ab32008-08-02 00:51:11 -04001722 struct path old_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 struct vfsmount *mnt = NULL;
Al Viro2d92ab32008-08-02 00:51:11 -04001724 int err = mount_is_safe(path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 if (err)
1726 return err;
1727 if (!old_name || !*old_name)
1728 return -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001729 err = kern_path(old_name, LOOKUP_FOLLOW, &old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 if (err)
1731 return err;
1732
Ram Pai390c6842005-11-07 17:17:51 -05001733 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 err = -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001735 if (IS_MNT_UNBINDABLE(old_path.mnt))
Jan Blunck4ac91372008-02-14 19:34:32 -08001736 goto out;
Ram Pai9676f0c2005-11-07 17:21:20 -05001737
Al Viro2d92ab32008-08-02 00:51:11 -04001738 if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt))
Al Viroccd48bc2005-11-07 17:15:04 -05001739 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740
Al Viroccd48bc2005-11-07 17:15:04 -05001741 err = -ENOMEM;
1742 if (recurse)
Al Viro2d92ab32008-08-02 00:51:11 -04001743 mnt = copy_tree(old_path.mnt, old_path.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -05001744 else
Al Viro2d92ab32008-08-02 00:51:11 -04001745 mnt = clone_mnt(old_path.mnt, old_path.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -05001746
1747 if (!mnt)
1748 goto out;
1749
Al Viro2d92ab32008-08-02 00:51:11 -04001750 err = graft_tree(mnt, path);
Al Viroccd48bc2005-11-07 17:15:04 -05001751 if (err) {
Ram Pai70fbcdf2005-11-07 17:17:04 -05001752 LIST_HEAD(umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001753
1754 br_write_lock(vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -05001755 umount_tree(mnt, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10001756 br_write_unlock(vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001757 release_mounts(&umount_list);
Ram Pai5b83d2c5c2005-11-07 17:16:29 -05001758 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759
Al Viroccd48bc2005-11-07 17:15:04 -05001760out:
Ram Pai390c6842005-11-07 17:17:51 -05001761 up_write(&namespace_sem);
Al Viro2d92ab32008-08-02 00:51:11 -04001762 path_put(&old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 return err;
1764}
1765
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001766static int change_mount_flags(struct vfsmount *mnt, int ms_flags)
1767{
1768 int error = 0;
1769 int readonly_request = 0;
1770
1771 if (ms_flags & MS_RDONLY)
1772 readonly_request = 1;
1773 if (readonly_request == __mnt_is_readonly(mnt))
1774 return 0;
1775
1776 if (readonly_request)
1777 error = mnt_make_readonly(mnt);
1778 else
1779 __mnt_unmake_readonly(mnt);
1780 return error;
1781}
1782
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783/*
1784 * change filesystem flags. dir should be a physical root of filesystem.
1785 * If you've mounted a non-root directory somewhere and want to do remount
1786 * on it - tough luck.
1787 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001788static int do_remount(struct path *path, int flags, int mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789 void *data)
1790{
1791 int err;
Al Viro2d92ab32008-08-02 00:51:11 -04001792 struct super_block *sb = path->mnt->mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793
1794 if (!capable(CAP_SYS_ADMIN))
1795 return -EPERM;
1796
Al Viro2d92ab32008-08-02 00:51:11 -04001797 if (!check_mnt(path->mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 return -EINVAL;
1799
Al Viro2d92ab32008-08-02 00:51:11 -04001800 if (path->dentry != path->mnt->mnt_root)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 return -EINVAL;
1802
1803 down_write(&sb->s_umount);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001804 if (flags & MS_BIND)
Al Viro2d92ab32008-08-02 00:51:11 -04001805 err = change_mount_flags(path->mnt, flags);
Al Viro4aa98cf2009-05-08 13:36:58 -04001806 else
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08001807 err = do_remount_sb(sb, flags, data, 0);
Al Viro7b43a792010-01-16 13:01:26 -05001808 if (!err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001809 br_write_lock(vfsmount_lock);
Valerie Aurora495d6c92010-01-26 14:20:47 -05001810 mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK;
Al Viro2d92ab32008-08-02 00:51:11 -04001811 path->mnt->mnt_flags = mnt_flags;
Nick Piggin99b7db72010-08-18 04:37:39 +10001812 br_write_unlock(vfsmount_lock);
Al Viro7b43a792010-01-16 13:01:26 -05001813 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 up_write(&sb->s_umount);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001815 if (!err) {
Nick Piggin99b7db72010-08-18 04:37:39 +10001816 br_write_lock(vfsmount_lock);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001817 touch_mnt_namespace(path->mnt->mnt_ns);
Nick Piggin99b7db72010-08-18 04:37:39 +10001818 br_write_unlock(vfsmount_lock);
Dan Williams0e55a7c2008-09-26 19:01:20 -07001819 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 return err;
1821}
1822
Ram Pai9676f0c2005-11-07 17:21:20 -05001823static inline int tree_contains_unbindable(struct vfsmount *mnt)
1824{
1825 struct vfsmount *p;
1826 for (p = mnt; p; p = next_mnt(p, mnt)) {
1827 if (IS_MNT_UNBINDABLE(p))
1828 return 1;
1829 }
1830 return 0;
1831}
1832
Al Viro0a0d8a42008-08-02 00:55:27 -04001833static int do_move_mount(struct path *path, char *old_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834{
Al Viro2d92ab32008-08-02 00:51:11 -04001835 struct path old_path, parent_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 struct vfsmount *p;
1837 int err = 0;
1838 if (!capable(CAP_SYS_ADMIN))
1839 return -EPERM;
1840 if (!old_name || !*old_name)
1841 return -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001842 err = kern_path(old_name, LOOKUP_FOLLOW, &old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843 if (err)
1844 return err;
1845
Ram Pai390c6842005-11-07 17:17:51 -05001846 down_write(&namespace_sem);
Al Viro2d92ab32008-08-02 00:51:11 -04001847 while (d_mountpoint(path->dentry) &&
Al Viro9393bd02009-04-18 13:58:15 -04001848 follow_down(path))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 ;
1850 err = -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001851 if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 goto out;
1853
1854 err = -ENOENT;
Al Viro2d92ab32008-08-02 00:51:11 -04001855 mutex_lock(&path->dentry->d_inode->i_mutex);
Al Virod83c49f2010-04-30 17:17:09 -04001856 if (cant_mount(path->dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 goto out1;
1858
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04001859 if (d_unlinked(path->dentry))
Ram Pai21444402005-11-07 17:20:03 -05001860 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
1862 err = -EINVAL;
Al Viro2d92ab32008-08-02 00:51:11 -04001863 if (old_path.dentry != old_path.mnt->mnt_root)
Ram Pai21444402005-11-07 17:20:03 -05001864 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865
Al Viro2d92ab32008-08-02 00:51:11 -04001866 if (old_path.mnt == old_path.mnt->mnt_parent)
Ram Pai21444402005-11-07 17:20:03 -05001867 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868
Al Viro2d92ab32008-08-02 00:51:11 -04001869 if (S_ISDIR(path->dentry->d_inode->i_mode) !=
1870 S_ISDIR(old_path.dentry->d_inode->i_mode))
Ram Pai21444402005-11-07 17:20:03 -05001871 goto out1;
1872 /*
1873 * Don't move a mount residing in a shared parent.
1874 */
Al Viro2d92ab32008-08-02 00:51:11 -04001875 if (old_path.mnt->mnt_parent &&
1876 IS_MNT_SHARED(old_path.mnt->mnt_parent))
Ram Pai21444402005-11-07 17:20:03 -05001877 goto out1;
Ram Pai9676f0c2005-11-07 17:21:20 -05001878 /*
1879 * Don't move a mount tree containing unbindable mounts to a destination
1880 * mount which is shared.
1881 */
Al Viro2d92ab32008-08-02 00:51:11 -04001882 if (IS_MNT_SHARED(path->mnt) &&
1883 tree_contains_unbindable(old_path.mnt))
Ram Pai9676f0c2005-11-07 17:21:20 -05001884 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 err = -ELOOP;
Al Viro2d92ab32008-08-02 00:51:11 -04001886 for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent)
1887 if (p == old_path.mnt)
Ram Pai21444402005-11-07 17:20:03 -05001888 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889
Al Viro2d92ab32008-08-02 00:51:11 -04001890 err = attach_recursive_mnt(old_path.mnt, path, &parent_path);
Jan Blunck4ac91372008-02-14 19:34:32 -08001891 if (err)
Ram Pai21444402005-11-07 17:20:03 -05001892 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
1894 /* if the mount is moved, it should no longer be expire
1895 * automatically */
Al Viro2d92ab32008-08-02 00:51:11 -04001896 list_del_init(&old_path.mnt->mnt_expire);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897out1:
Al Viro2d92ab32008-08-02 00:51:11 -04001898 mutex_unlock(&path->dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899out:
Ram Pai390c6842005-11-07 17:17:51 -05001900 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 if (!err)
Al Viro1a390682008-03-21 20:48:19 -04001902 path_put(&parent_path);
Al Viro2d92ab32008-08-02 00:51:11 -04001903 path_put(&old_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 return err;
1905}
1906
1907/*
1908 * create a new mount for userspace and request it to be added into the
1909 * namespace's tree
1910 */
Al Viro0a0d8a42008-08-02 00:55:27 -04001911static int do_new_mount(struct path *path, char *type, int flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912 int mnt_flags, char *name, void *data)
1913{
1914 struct vfsmount *mnt;
1915
Vegard Nossumeca6f532009-09-18 13:05:45 -07001916 if (!type)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 return -EINVAL;
1918
1919 /* we need capabilities... */
1920 if (!capable(CAP_SYS_ADMIN))
1921 return -EPERM;
1922
1923 mnt = do_kern_mount(type, flags, name, data);
1924 if (IS_ERR(mnt))
1925 return PTR_ERR(mnt);
1926
Al Viro2d92ab32008-08-02 00:51:11 -04001927 return do_add_mount(mnt, path, mnt_flags, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928}
1929
1930/*
1931 * add a mount into a namespace's mount tree
1932 * - provide the option of adding the new mount to an expiration list
1933 */
Al Viro8d66bf52008-08-01 09:05:54 -04001934int do_add_mount(struct vfsmount *newmnt, struct path *path,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 int mnt_flags, struct list_head *fslist)
1936{
1937 int err;
1938
Al Viro80893522010-02-05 09:30:46 -05001939 mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL);
Al Viro27d55f12010-01-16 13:07:36 -05001940
Ram Pai390c6842005-11-07 17:17:51 -05001941 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942 /* Something was mounted here while we slept */
Al Viro8d66bf52008-08-01 09:05:54 -04001943 while (d_mountpoint(path->dentry) &&
Al Viro9393bd02009-04-18 13:58:15 -04001944 follow_down(path))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 ;
1946 err = -EINVAL;
Al Virodd5cae62009-04-07 12:21:18 -04001947 if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 goto unlock;
1949
1950 /* Refuse the same filesystem on the same mount point */
1951 err = -EBUSY;
Al Viro8d66bf52008-08-01 09:05:54 -04001952 if (path->mnt->mnt_sb == newmnt->mnt_sb &&
1953 path->mnt->mnt_root == path->dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 goto unlock;
1955
1956 err = -EINVAL;
1957 if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode))
1958 goto unlock;
1959
1960 newmnt->mnt_flags = mnt_flags;
Al Viro8d66bf52008-08-01 09:05:54 -04001961 if ((err = graft_tree(newmnt, path)))
Ram Pai5b83d2c5c2005-11-07 17:16:29 -05001962 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963
Al Viro6758f952008-03-22 16:14:30 -04001964 if (fslist) /* add to the specified expiration list */
Miklos Szeredi55e700b2005-07-07 17:57:30 -07001965 list_add_tail(&newmnt->mnt_expire, fslist);
Al Viro6758f952008-03-22 16:14:30 -04001966
Ram Pai390c6842005-11-07 17:17:51 -05001967 up_write(&namespace_sem);
Ram Pai5b83d2c5c2005-11-07 17:16:29 -05001968 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969
1970unlock:
Ram Pai390c6842005-11-07 17:17:51 -05001971 up_write(&namespace_sem);
Nick Pigginb3e19d92011-01-07 17:50:11 +11001972 mntput_long(newmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973 return err;
1974}
1975
1976EXPORT_SYMBOL_GPL(do_add_mount);
1977
Trond Myklebust5528f9112006-06-09 09:34:17 -04001978/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979 * process a list of expirable mountpoints with the intent of discarding any
1980 * mountpoints that aren't in use and haven't been touched since last we came
1981 * here
1982 */
1983void mark_mounts_for_expiry(struct list_head *mounts)
1984{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985 struct vfsmount *mnt, *next;
1986 LIST_HEAD(graveyard);
Al Virobcc5c7d2008-03-22 00:21:53 -04001987 LIST_HEAD(umounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988
1989 if (list_empty(mounts))
1990 return;
1991
Al Virobcc5c7d2008-03-22 00:21:53 -04001992 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10001993 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994
1995 /* extract from the expiration list every vfsmount that matches the
1996 * following criteria:
1997 * - only referenced by its parent vfsmount
1998 * - still marked for expiry (marked on the last call here; marks are
1999 * cleared by mntput())
2000 */
Miklos Szeredi55e700b2005-07-07 17:57:30 -07002001 list_for_each_entry_safe(mnt, next, mounts, mnt_expire) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002 if (!xchg(&mnt->mnt_expiry_mark, 1) ||
Al Virobcc5c7d2008-03-22 00:21:53 -04002003 propagate_mount_busy(mnt, 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 continue;
Miklos Szeredi55e700b2005-07-07 17:57:30 -07002005 list_move(&mnt->mnt_expire, &graveyard);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 }
Al Virobcc5c7d2008-03-22 00:21:53 -04002007 while (!list_empty(&graveyard)) {
2008 mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire);
2009 touch_mnt_namespace(mnt->mnt_ns);
2010 umount_tree(mnt, 1, &umounts);
2011 }
Nick Piggin99b7db72010-08-18 04:37:39 +10002012 br_write_unlock(vfsmount_lock);
Al Virobcc5c7d2008-03-22 00:21:53 -04002013 up_write(&namespace_sem);
2014
2015 release_mounts(&umounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016}
2017
2018EXPORT_SYMBOL_GPL(mark_mounts_for_expiry);
2019
2020/*
Trond Myklebust5528f9112006-06-09 09:34:17 -04002021 * Ripoff of 'select_parent()'
2022 *
2023 * search the list of submounts for a given mountpoint, and move any
2024 * shrinkable submounts to the 'graveyard' list.
2025 */
2026static int select_submounts(struct vfsmount *parent, struct list_head *graveyard)
2027{
2028 struct vfsmount *this_parent = parent;
2029 struct list_head *next;
2030 int found = 0;
2031
2032repeat:
2033 next = this_parent->mnt_mounts.next;
2034resume:
2035 while (next != &this_parent->mnt_mounts) {
2036 struct list_head *tmp = next;
2037 struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child);
2038
2039 next = tmp->next;
2040 if (!(mnt->mnt_flags & MNT_SHRINKABLE))
2041 continue;
2042 /*
2043 * Descend a level if the d_mounts list is non-empty.
2044 */
2045 if (!list_empty(&mnt->mnt_mounts)) {
2046 this_parent = mnt;
2047 goto repeat;
2048 }
2049
2050 if (!propagate_mount_busy(mnt, 1)) {
Trond Myklebust5528f9112006-06-09 09:34:17 -04002051 list_move_tail(&mnt->mnt_expire, graveyard);
2052 found++;
2053 }
2054 }
2055 /*
2056 * All done at this level ... ascend and resume the search
2057 */
2058 if (this_parent != parent) {
2059 next = this_parent->mnt_child.next;
2060 this_parent = this_parent->mnt_parent;
2061 goto resume;
2062 }
2063 return found;
2064}
2065
2066/*
2067 * process a list of expirable mountpoints with the intent of discarding any
2068 * submounts of a specific parent mountpoint
Nick Piggin99b7db72010-08-18 04:37:39 +10002069 *
2070 * vfsmount_lock must be held for write
Trond Myklebust5528f9112006-06-09 09:34:17 -04002071 */
Al Viroc35038b2008-03-22 00:46:23 -04002072static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts)
Trond Myklebust5528f9112006-06-09 09:34:17 -04002073{
2074 LIST_HEAD(graveyard);
Al Viroc35038b2008-03-22 00:46:23 -04002075 struct vfsmount *m;
Trond Myklebust5528f9112006-06-09 09:34:17 -04002076
Trond Myklebust5528f9112006-06-09 09:34:17 -04002077 /* extract submounts of 'mountpoint' from the expiration list */
Al Viroc35038b2008-03-22 00:46:23 -04002078 while (select_submounts(mnt, &graveyard)) {
Al Virobcc5c7d2008-03-22 00:21:53 -04002079 while (!list_empty(&graveyard)) {
Al Viroc35038b2008-03-22 00:46:23 -04002080 m = list_first_entry(&graveyard, struct vfsmount,
Al Virobcc5c7d2008-03-22 00:21:53 -04002081 mnt_expire);
Eric W. Biedermanafef80b2008-11-12 13:26:54 -08002082 touch_mnt_namespace(m->mnt_ns);
2083 umount_tree(m, 1, umounts);
Al Virobcc5c7d2008-03-22 00:21:53 -04002084 }
2085 }
Trond Myklebust5528f9112006-06-09 09:34:17 -04002086}
2087
Trond Myklebust5528f9112006-06-09 09:34:17 -04002088/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089 * Some copy_from_user() implementations do not return the exact number of
2090 * bytes remaining to copy on a fault. But copy_mount_options() requires that.
2091 * Note that this function differs from copy_from_user() in that it will oops
2092 * on bad values of `to', rather than returning a short copy.
2093 */
Ram Paib58fed82005-11-07 17:16:09 -05002094static long exact_copy_from_user(void *to, const void __user * from,
2095 unsigned long n)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096{
2097 char *t = to;
2098 const char __user *f = from;
2099 char c;
2100
2101 if (!access_ok(VERIFY_READ, from, n))
2102 return n;
2103
2104 while (n) {
2105 if (__get_user(c, f)) {
2106 memset(t, 0, n);
2107 break;
2108 }
2109 *t++ = c;
2110 f++;
2111 n--;
2112 }
2113 return n;
2114}
2115
Ram Paib58fed82005-11-07 17:16:09 -05002116int copy_mount_options(const void __user * data, unsigned long *where)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117{
2118 int i;
2119 unsigned long page;
2120 unsigned long size;
Ram Paib58fed82005-11-07 17:16:09 -05002121
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 *where = 0;
2123 if (!data)
2124 return 0;
2125
2126 if (!(page = __get_free_page(GFP_KERNEL)))
2127 return -ENOMEM;
2128
2129 /* We only care that *some* data at the address the user
2130 * gave us is valid. Just in case, we'll zero
2131 * the remainder of the page.
2132 */
2133 /* copy_from_user cannot cross TASK_SIZE ! */
2134 size = TASK_SIZE - (unsigned long)data;
2135 if (size > PAGE_SIZE)
2136 size = PAGE_SIZE;
2137
2138 i = size - exact_copy_from_user((void *)page, data, size);
2139 if (!i) {
Ram Paib58fed82005-11-07 17:16:09 -05002140 free_page(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 return -EFAULT;
2142 }
2143 if (i != PAGE_SIZE)
2144 memset((char *)page + i, 0, PAGE_SIZE - i);
2145 *where = page;
2146 return 0;
2147}
2148
Vegard Nossumeca6f532009-09-18 13:05:45 -07002149int copy_mount_string(const void __user *data, char **where)
2150{
2151 char *tmp;
2152
2153 if (!data) {
2154 *where = NULL;
2155 return 0;
2156 }
2157
2158 tmp = strndup_user(data, PAGE_SIZE);
2159 if (IS_ERR(tmp))
2160 return PTR_ERR(tmp);
2161
2162 *where = tmp;
2163 return 0;
2164}
2165
Linus Torvalds1da177e2005-04-16 15:20:36 -07002166/*
2167 * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to
2168 * be given to the mount() call (ie: read-only, no-dev, no-suid etc).
2169 *
2170 * data is a (void *) that can point to any structure up to
2171 * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent
2172 * information (or be NULL).
2173 *
2174 * Pre-0.97 versions of mount() didn't have a flags word.
2175 * When the flags word was introduced its top half was required
2176 * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9.
2177 * Therefore, if this magic number is present, it carries no information
2178 * and must be discarded.
2179 */
Ram Paib58fed82005-11-07 17:16:09 -05002180long do_mount(char *dev_name, char *dir_name, char *type_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 unsigned long flags, void *data_page)
2182{
Al Viro2d92ab32008-08-02 00:51:11 -04002183 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184 int retval = 0;
2185 int mnt_flags = 0;
2186
2187 /* Discard magic */
2188 if ((flags & MS_MGC_MSK) == MS_MGC_VAL)
2189 flags &= ~MS_MGC_MSK;
2190
2191 /* Basic sanity checks */
2192
2193 if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE))
2194 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195
2196 if (data_page)
2197 ((char *)data_page)[PAGE_SIZE - 1] = 0;
2198
Tetsuo Handaa27ab9f22009-10-04 21:49:49 +09002199 /* ... and get the mountpoint */
2200 retval = kern_path(dir_name, LOOKUP_FOLLOW, &path);
2201 if (retval)
2202 return retval;
2203
2204 retval = security_sb_mount(dev_name, &path,
2205 type_page, flags, data_page);
2206 if (retval)
2207 goto dput_out;
2208
Andi Kleen613cbe32009-04-19 18:40:43 +02002209 /* Default to relatime unless overriden */
2210 if (!(flags & MS_NOATIME))
2211 mnt_flags |= MNT_RELATIME;
Matthew Garrett0a1c01c2009-03-26 17:53:14 +00002212
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 /* Separate the per-mountpoint flags */
2214 if (flags & MS_NOSUID)
2215 mnt_flags |= MNT_NOSUID;
2216 if (flags & MS_NODEV)
2217 mnt_flags |= MNT_NODEV;
2218 if (flags & MS_NOEXEC)
2219 mnt_flags |= MNT_NOEXEC;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08002220 if (flags & MS_NOATIME)
2221 mnt_flags |= MNT_NOATIME;
2222 if (flags & MS_NODIRATIME)
2223 mnt_flags |= MNT_NODIRATIME;
Matthew Garrettd0adde52009-03-26 17:49:56 +00002224 if (flags & MS_STRICTATIME)
2225 mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME);
Dave Hansen2e4b7fc2008-02-15 14:38:00 -08002226 if (flags & MS_RDONLY)
2227 mnt_flags |= MNT_READONLY;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08002228
Al Viro7a4dec52010-08-09 12:05:43 -04002229 flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN |
Matthew Garrettd0adde52009-03-26 17:49:56 +00002230 MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT |
2231 MS_STRICTATIME);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233 if (flags & MS_REMOUNT)
Al Viro2d92ab32008-08-02 00:51:11 -04002234 retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235 data_page);
2236 else if (flags & MS_BIND)
Al Viro2d92ab32008-08-02 00:51:11 -04002237 retval = do_loopback(&path, dev_name, flags & MS_REC);
Ram Pai9676f0c2005-11-07 17:21:20 -05002238 else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
Al Viro2d92ab32008-08-02 00:51:11 -04002239 retval = do_change_type(&path, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240 else if (flags & MS_MOVE)
Al Viro2d92ab32008-08-02 00:51:11 -04002241 retval = do_move_mount(&path, dev_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 else
Al Viro2d92ab32008-08-02 00:51:11 -04002243 retval = do_new_mount(&path, type_page, flags, mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244 dev_name, data_page);
2245dput_out:
Al Viro2d92ab32008-08-02 00:51:11 -04002246 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247 return retval;
2248}
2249
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002250static struct mnt_namespace *alloc_mnt_ns(void)
2251{
2252 struct mnt_namespace *new_ns;
2253
2254 new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL);
2255 if (!new_ns)
2256 return ERR_PTR(-ENOMEM);
2257 atomic_set(&new_ns->count, 1);
2258 new_ns->root = NULL;
2259 INIT_LIST_HEAD(&new_ns->list);
2260 init_waitqueue_head(&new_ns->poll);
2261 new_ns->event = 0;
2262 return new_ns;
2263}
2264
JANAK DESAI741a2952006-02-07 12:59:00 -08002265/*
2266 * Allocate a new namespace structure and populate it with contents
2267 * copied from the namespace of the passed in task structure.
2268 */
Badari Pulavartye3222c42007-05-08 00:25:21 -07002269static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002270 struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002272 struct mnt_namespace *new_ns;
Al Viro7f2da1e2008-05-10 20:44:54 -04002273 struct vfsmount *rootmnt = NULL, *pwdmnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 struct vfsmount *p, *q;
2275
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002276 new_ns = alloc_mnt_ns();
2277 if (IS_ERR(new_ns))
2278 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002279
Ram Pai390c6842005-11-07 17:17:51 -05002280 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281 /* First pass: copy the tree topology */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002282 new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root,
Ram Pai9676f0c2005-11-07 17:21:20 -05002283 CL_COPY_ALL | CL_EXPIRE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 if (!new_ns->root) {
Ram Pai390c6842005-11-07 17:17:51 -05002285 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 kfree(new_ns);
Julia Lawall5cc4a032008-12-01 14:34:51 -08002287 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288 }
Nick Piggin99b7db72010-08-18 04:37:39 +10002289 br_write_lock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290 list_add_tail(&new_ns->list, &new_ns->root->mnt_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10002291 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002292
2293 /*
2294 * Second pass: switch the tsk->fs->* elements and mark new vfsmounts
2295 * as belonging to new namespace. We have already acquired a private
2296 * fs_struct, so tsk->fs->lock is not needed.
2297 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002298 p = mnt_ns->root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299 q = new_ns->root;
2300 while (p) {
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002301 q->mnt_ns = new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302 if (fs) {
Jan Blunck6ac08c32008-02-14 19:34:38 -08002303 if (p == fs->root.mnt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304 rootmnt = p;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002305 fs->root.mnt = mntget_long(q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 }
Jan Blunck6ac08c32008-02-14 19:34:38 -08002307 if (p == fs->pwd.mnt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308 pwdmnt = p;
Nick Pigginb3e19d92011-01-07 17:50:11 +11002309 fs->pwd.mnt = mntget_long(q);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 }
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002312 p = next_mnt(p, mnt_ns->root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313 q = next_mnt(q, new_ns->root);
2314 }
Ram Pai390c6842005-11-07 17:17:51 -05002315 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 if (rootmnt)
Nick Pigginb3e19d92011-01-07 17:50:11 +11002318 mntput_long(rootmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319 if (pwdmnt)
Nick Pigginb3e19d92011-01-07 17:50:11 +11002320 mntput_long(pwdmnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321
JANAK DESAI741a2952006-02-07 12:59:00 -08002322 return new_ns;
2323}
2324
Eric W. Biederman213dd262007-07-15 23:41:15 -07002325struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns,
Badari Pulavartye3222c42007-05-08 00:25:21 -07002326 struct fs_struct *new_fs)
JANAK DESAI741a2952006-02-07 12:59:00 -08002327{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002328 struct mnt_namespace *new_ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08002329
Badari Pulavartye3222c42007-05-08 00:25:21 -07002330 BUG_ON(!ns);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002331 get_mnt_ns(ns);
JANAK DESAI741a2952006-02-07 12:59:00 -08002332
2333 if (!(flags & CLONE_NEWNS))
Badari Pulavartye3222c42007-05-08 00:25:21 -07002334 return ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08002335
Badari Pulavartye3222c42007-05-08 00:25:21 -07002336 new_ns = dup_mnt_ns(ns, new_fs);
JANAK DESAI741a2952006-02-07 12:59:00 -08002337
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002338 put_mnt_ns(ns);
Badari Pulavartye3222c42007-05-08 00:25:21 -07002339 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340}
2341
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002342/**
2343 * create_mnt_ns - creates a private namespace and adds a root filesystem
2344 * @mnt: pointer to the new root filesystem mountpoint
2345 */
Linus Torvaldsa2770d82009-12-17 12:51:05 -08002346struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt)
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002347{
2348 struct mnt_namespace *new_ns;
2349
2350 new_ns = alloc_mnt_ns();
2351 if (!IS_ERR(new_ns)) {
2352 mnt->mnt_ns = new_ns;
2353 new_ns->root = mnt;
2354 list_add(&new_ns->list, &new_ns->root->mnt_list);
2355 }
2356 return new_ns;
2357}
Linus Torvaldsa2770d82009-12-17 12:51:05 -08002358EXPORT_SYMBOL(create_mnt_ns);
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002359
Heiko Carstensbdc480e2009-01-14 14:14:12 +01002360SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name,
2361 char __user *, type, unsigned long, flags, void __user *, data)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362{
Vegard Nossumeca6f532009-09-18 13:05:45 -07002363 int ret;
2364 char *kernel_type;
2365 char *kernel_dir;
2366 char *kernel_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367 unsigned long data_page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002368
Vegard Nossumeca6f532009-09-18 13:05:45 -07002369 ret = copy_mount_string(type, &kernel_type);
2370 if (ret < 0)
2371 goto out_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372
Vegard Nossumeca6f532009-09-18 13:05:45 -07002373 kernel_dir = getname(dir_name);
2374 if (IS_ERR(kernel_dir)) {
2375 ret = PTR_ERR(kernel_dir);
2376 goto out_dir;
2377 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378
Vegard Nossumeca6f532009-09-18 13:05:45 -07002379 ret = copy_mount_string(dev_name, &kernel_dev);
2380 if (ret < 0)
2381 goto out_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382
Vegard Nossumeca6f532009-09-18 13:05:45 -07002383 ret = copy_mount_options(data, &data_page);
2384 if (ret < 0)
2385 goto out_data;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386
Vegard Nossumeca6f532009-09-18 13:05:45 -07002387 ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags,
2388 (void *) data_page);
2389
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 free_page(data_page);
Vegard Nossumeca6f532009-09-18 13:05:45 -07002391out_data:
2392 kfree(kernel_dev);
2393out_dev:
2394 putname(kernel_dir);
2395out_dir:
2396 kfree(kernel_type);
2397out_type:
2398 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399}
2400
2401/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002402 * pivot_root Semantics:
2403 * Moves the root file system of the current process to the directory put_old,
2404 * makes new_root as the new root file system of the current process, and sets
2405 * root/cwd of all processes which had them on the current root to new_root.
2406 *
2407 * Restrictions:
2408 * The new_root and put_old must be directories, and must not be on the
2409 * same file system as the current process root. The put_old must be
2410 * underneath new_root, i.e. adding a non-zero number of /.. to the string
2411 * pointed to by put_old must yield the same directory as new_root. No other
2412 * file system may be mounted on put_old. After all, new_root is a mountpoint.
2413 *
Neil Brown4a0d11f2006-01-08 01:03:18 -08002414 * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem.
2415 * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives
2416 * in this situation.
2417 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418 * Notes:
2419 * - we don't move root/cwd if they are not at the root (reason: if something
2420 * cared enough to change them, it's probably wrong to force them elsewhere)
2421 * - it's okay to pick a root that isn't the root of a file system, e.g.
2422 * /nfs/my_root where /nfs is the mount point. It must be a mountpoint,
2423 * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root
2424 * first.
2425 */
Heiko Carstens3480b252009-01-14 14:14:16 +01002426SYSCALL_DEFINE2(pivot_root, const char __user *, new_root,
2427 const char __user *, put_old)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428{
2429 struct vfsmount *tmp;
Al Viro2d8f3032008-07-22 09:59:21 -04002430 struct path new, old, parent_path, root_parent, root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 int error;
2432
2433 if (!capable(CAP_SYS_ADMIN))
2434 return -EPERM;
2435
Al Viro2d8f3032008-07-22 09:59:21 -04002436 error = user_path_dir(new_root, &new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437 if (error)
2438 goto out0;
2439 error = -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -04002440 if (!check_mnt(new.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002441 goto out1;
2442
Al Viro2d8f3032008-07-22 09:59:21 -04002443 error = user_path_dir(put_old, &old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002444 if (error)
2445 goto out1;
2446
Al Viro2d8f3032008-07-22 09:59:21 -04002447 error = security_sb_pivotroot(&old, &new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448 if (error) {
Al Viro2d8f3032008-07-22 09:59:21 -04002449 path_put(&old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 goto out1;
2451 }
2452
Miklos Szeredif7ad3c62010-08-10 11:41:36 +02002453 get_fs_root(current->fs, &root);
Ram Pai390c6842005-11-07 17:17:51 -05002454 down_write(&namespace_sem);
Al Viro2d8f3032008-07-22 09:59:21 -04002455 mutex_lock(&old.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002456 error = -EINVAL;
Al Viro2d8f3032008-07-22 09:59:21 -04002457 if (IS_MNT_SHARED(old.mnt) ||
2458 IS_MNT_SHARED(new.mnt->mnt_parent) ||
Al Viro8c3ee422008-03-22 18:00:39 -04002459 IS_MNT_SHARED(root.mnt->mnt_parent))
Ram Pai21444402005-11-07 17:20:03 -05002460 goto out2;
Al Viro8c3ee422008-03-22 18:00:39 -04002461 if (!check_mnt(root.mnt))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462 goto out2;
2463 error = -ENOENT;
Al Virod83c49f2010-04-30 17:17:09 -04002464 if (cant_mount(old.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002465 goto out2;
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04002466 if (d_unlinked(new.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002467 goto out2;
Alexey Dobriyanf3da3922009-05-04 03:32:03 +04002468 if (d_unlinked(old.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469 goto out2;
2470 error = -EBUSY;
Al Viro2d8f3032008-07-22 09:59:21 -04002471 if (new.mnt == root.mnt ||
2472 old.mnt == root.mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473 goto out2; /* loop, on the same file system */
2474 error = -EINVAL;
Al Viro8c3ee422008-03-22 18:00:39 -04002475 if (root.mnt->mnt_root != root.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002476 goto out2; /* not a mountpoint */
Al Viro8c3ee422008-03-22 18:00:39 -04002477 if (root.mnt->mnt_parent == root.mnt)
Miklos Szeredi0bb6fcc2005-09-06 15:19:36 -07002478 goto out2; /* not attached */
Al Viro2d8f3032008-07-22 09:59:21 -04002479 if (new.mnt->mnt_root != new.dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480 goto out2; /* not a mountpoint */
Al Viro2d8f3032008-07-22 09:59:21 -04002481 if (new.mnt->mnt_parent == new.mnt)
Miklos Szeredi0bb6fcc2005-09-06 15:19:36 -07002482 goto out2; /* not attached */
Jan Blunck4ac91372008-02-14 19:34:32 -08002483 /* make sure we can reach put_old from new_root */
Al Viro2d8f3032008-07-22 09:59:21 -04002484 tmp = old.mnt;
Nick Piggin99b7db72010-08-18 04:37:39 +10002485 br_write_lock(vfsmount_lock);
Al Viro2d8f3032008-07-22 09:59:21 -04002486 if (tmp != new.mnt) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487 for (;;) {
2488 if (tmp->mnt_parent == tmp)
2489 goto out3; /* already mounted on put_old */
Al Viro2d8f3032008-07-22 09:59:21 -04002490 if (tmp->mnt_parent == new.mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002491 break;
2492 tmp = tmp->mnt_parent;
2493 }
Al Viro2d8f3032008-07-22 09:59:21 -04002494 if (!is_subdir(tmp->mnt_mountpoint, new.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495 goto out3;
Al Viro2d8f3032008-07-22 09:59:21 -04002496 } else if (!is_subdir(old.dentry, new.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002497 goto out3;
Al Viro2d8f3032008-07-22 09:59:21 -04002498 detach_mnt(new.mnt, &parent_path);
Al Viro8c3ee422008-03-22 18:00:39 -04002499 detach_mnt(root.mnt, &root_parent);
Jan Blunck4ac91372008-02-14 19:34:32 -08002500 /* mount old root on put_old */
Al Viro2d8f3032008-07-22 09:59:21 -04002501 attach_mnt(root.mnt, &old);
Jan Blunck4ac91372008-02-14 19:34:32 -08002502 /* mount new_root on / */
Al Viro2d8f3032008-07-22 09:59:21 -04002503 attach_mnt(new.mnt, &root_parent);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002504 touch_mnt_namespace(current->nsproxy->mnt_ns);
Nick Piggin99b7db72010-08-18 04:37:39 +10002505 br_write_unlock(vfsmount_lock);
Al Viro2d8f3032008-07-22 09:59:21 -04002506 chroot_fs_refs(&root, &new);
Nick Pigginb3e19d92011-01-07 17:50:11 +11002507
Linus Torvalds1da177e2005-04-16 15:20:36 -07002508 error = 0;
Al Viro1a390682008-03-21 20:48:19 -04002509 path_put(&root_parent);
2510 path_put(&parent_path);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511out2:
Al Viro2d8f3032008-07-22 09:59:21 -04002512 mutex_unlock(&old.dentry->d_inode->i_mutex);
Ram Pai390c6842005-11-07 17:17:51 -05002513 up_write(&namespace_sem);
Al Viro8c3ee422008-03-22 18:00:39 -04002514 path_put(&root);
Al Viro2d8f3032008-07-22 09:59:21 -04002515 path_put(&old);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516out1:
Al Viro2d8f3032008-07-22 09:59:21 -04002517 path_put(&new);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518out0:
Linus Torvalds1da177e2005-04-16 15:20:36 -07002519 return error;
2520out3:
Nick Piggin99b7db72010-08-18 04:37:39 +10002521 br_write_unlock(vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522 goto out2;
2523}
2524
2525static void __init init_mount_tree(void)
2526{
2527 struct vfsmount *mnt;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002528 struct mnt_namespace *ns;
Jan Blunckac748a02008-02-14 19:34:39 -08002529 struct path root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530
2531 mnt = do_kern_mount("rootfs", 0, "rootfs", NULL);
2532 if (IS_ERR(mnt))
2533 panic("Can't create rootfs");
Nick Pigginb3e19d92011-01-07 17:50:11 +11002534
Trond Myklebust3b22edc2009-06-23 17:29:49 -04002535 ns = create_mnt_ns(mnt);
2536 if (IS_ERR(ns))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537 panic("Can't allocate initial namespace");
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002539 init_task.nsproxy->mnt_ns = ns;
2540 get_mnt_ns(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541
Jan Blunckac748a02008-02-14 19:34:39 -08002542 root.mnt = ns->root;
2543 root.dentry = ns->root->mnt_root;
2544
2545 set_fs_pwd(current->fs, &root);
2546 set_fs_root(current->fs, &root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002547}
2548
Denis Cheng74bf17c2007-10-16 23:26:30 -07002549void __init mnt_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550{
Eric Dumazet13f14b42008-02-06 01:37:57 -08002551 unsigned u;
Randy Dunlap15a67dd2006-09-29 01:58:57 -07002552 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553
Ram Pai390c6842005-11-07 17:17:51 -05002554 init_rwsem(&namespace_sem);
2555
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556 mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount),
Paul Mundt20c2df82007-07-20 10:11:58 +09002557 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558
Ram Paib58fed82005-11-07 17:16:09 -05002559 mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002560
2561 if (!mount_hashtable)
2562 panic("Failed to allocate mount hash table\n");
2563
Eric Dumazet13f14b42008-02-06 01:37:57 -08002564 printk("Mount-cache hash table entries: %lu\n", HASH_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002565
Eric Dumazet13f14b42008-02-06 01:37:57 -08002566 for (u = 0; u < HASH_SIZE; u++)
2567 INIT_LIST_HEAD(&mount_hashtable[u]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002568
Nick Piggin99b7db72010-08-18 04:37:39 +10002569 br_lock_init(vfsmount_lock);
2570
Randy Dunlap15a67dd2006-09-29 01:58:57 -07002571 err = sysfs_init();
2572 if (err)
2573 printk(KERN_WARNING "%s: sysfs_init error: %d\n",
Harvey Harrison8e24eea2008-04-30 00:55:09 -07002574 __func__, err);
Greg Kroah-Hartman00d26662007-10-29 14:17:23 -06002575 fs_kobj = kobject_create_and_add("fs", NULL);
2576 if (!fs_kobj)
Harvey Harrison8e24eea2008-04-30 00:55:09 -07002577 printk(KERN_WARNING "%s: kobj create error\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578 init_rootfs();
2579 init_mount_tree();
2580}
2581
Trond Myklebust616511d2009-06-22 15:09:13 -04002582void put_mnt_ns(struct mnt_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002583{
Ram Pai70fbcdf2005-11-07 17:17:04 -05002584 LIST_HEAD(umount_list);
Trond Myklebust616511d2009-06-22 15:09:13 -04002585
Al Virod498b252010-02-05 02:21:06 -05002586 if (!atomic_dec_and_test(&ns->count))
Trond Myklebust616511d2009-06-22 15:09:13 -04002587 return;
Ram Pai390c6842005-11-07 17:17:51 -05002588 down_write(&namespace_sem);
Nick Piggin99b7db72010-08-18 04:37:39 +10002589 br_write_lock(vfsmount_lock);
Al Virod498b252010-02-05 02:21:06 -05002590 umount_tree(ns->root, 0, &umount_list);
Nick Piggin99b7db72010-08-18 04:37:39 +10002591 br_write_unlock(vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05002592 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05002593 release_mounts(&umount_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08002594 kfree(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002595}
Trond Myklebustcf8d2c12009-06-22 15:09:13 -04002596EXPORT_SYMBOL(put_mnt_ns);