blob: 63ced21c12dc2a2f212786c3b5e2d25b48437bf6 [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>
14#include <linux/smp_lock.h>
15#include <linux/init.h>
Randy Dunlap15a67dd2006-09-29 01:58:57 -070016#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/quotaops.h>
18#include <linux/acct.h>
Randy Dunlap16f7e0f2006-01-11 12:17:46 -080019#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include <linux/module.h>
Andrew Mortonf20a9ea2006-08-14 22:43:23 -070021#include <linux/sysfs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/seq_file.h>
Kirill Korotaev6b3286e2006-12-08 02:37:56 -080023#include <linux/mnt_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/namei.h>
25#include <linux/security.h>
26#include <linux/mount.h>
David Howells07f3f052006-09-30 20:52:18 +020027#include <linux/ramfs.h>
Eric Dumazet13f14b42008-02-06 01:37:57 -080028#include <linux/log2.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <asm/uaccess.h>
30#include <asm/unistd.h>
Ram Pai07b20882005-11-07 17:19:07 -050031#include "pnode.h"
Adrian Bunk948730b2007-07-15 23:41:25 -070032#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070033
Eric Dumazet13f14b42008-02-06 01:37:57 -080034#define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head))
35#define HASH_SIZE (1UL << HASH_SHIFT)
36
Linus Torvalds1da177e2005-04-16 15:20:36 -070037/* spinlock for vfsmount related operations, inplace of dcache_lock */
Al Viro5addc5d2005-11-07 17:15:49 -050038__cacheline_aligned_in_smp DEFINE_SPINLOCK(vfsmount_lock);
39
40static int event;
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Eric Dumazetfa3536c2006-03-26 01:37:24 -080042static struct list_head *mount_hashtable __read_mostly;
Christoph Lametere18b8902006-12-06 20:33:20 -080043static struct kmem_cache *mnt_cache __read_mostly;
Ram Pai390c6842005-11-07 17:17:51 -050044static struct rw_semaphore namespace_sem;
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
Miklos Szeredif87fd4c2006-01-16 22:14:23 -080046/* /sys/fs */
Greg Kroah-Hartman00d26662007-10-29 14:17:23 -060047struct kobject *fs_kobj;
48EXPORT_SYMBOL_GPL(fs_kobj);
Miklos Szeredif87fd4c2006-01-16 22:14:23 -080049
Linus Torvalds1da177e2005-04-16 15:20:36 -070050static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry)
51{
Ram Paib58fed82005-11-07 17:16:09 -050052 unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES);
53 tmp += ((unsigned long)dentry / L1_CACHE_BYTES);
Eric Dumazet13f14b42008-02-06 01:37:57 -080054 tmp = tmp + (tmp >> HASH_SHIFT);
55 return tmp & (HASH_SIZE - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -070056}
57
58struct vfsmount *alloc_vfsmnt(const char *name)
59{
Robert P. J. Dayc3762222007-02-10 01:45:03 -080060 struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 if (mnt) {
Ram Paib58fed82005-11-07 17:16:09 -050062 atomic_set(&mnt->mnt_count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 INIT_LIST_HEAD(&mnt->mnt_hash);
64 INIT_LIST_HEAD(&mnt->mnt_child);
65 INIT_LIST_HEAD(&mnt->mnt_mounts);
66 INIT_LIST_HEAD(&mnt->mnt_list);
Miklos Szeredi55e700b2005-07-07 17:57:30 -070067 INIT_LIST_HEAD(&mnt->mnt_expire);
Ram Pai03e06e62005-11-07 17:19:33 -050068 INIT_LIST_HEAD(&mnt->mnt_share);
Ram Paia58b0eb2005-11-07 17:20:48 -050069 INIT_LIST_HEAD(&mnt->mnt_slave_list);
70 INIT_LIST_HEAD(&mnt->mnt_slave);
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 if (name) {
Ram Paib58fed82005-11-07 17:16:09 -050072 int size = strlen(name) + 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -070073 char *newname = kmalloc(size, GFP_KERNEL);
74 if (newname) {
75 memcpy(newname, name, size);
76 mnt->mnt_devname = newname;
77 }
78 }
79 }
80 return mnt;
81}
82
David Howells454e2392006-06-23 02:02:57 -070083int simple_set_mnt(struct vfsmount *mnt, struct super_block *sb)
84{
85 mnt->mnt_sb = sb;
86 mnt->mnt_root = dget(sb->s_root);
87 return 0;
88}
89
90EXPORT_SYMBOL(simple_set_mnt);
91
Linus Torvalds1da177e2005-04-16 15:20:36 -070092void free_vfsmnt(struct vfsmount *mnt)
93{
94 kfree(mnt->mnt_devname);
95 kmem_cache_free(mnt_cache, mnt);
96}
97
98/*
Ram Paia05964f2005-11-07 17:20:17 -050099 * find the first or last mount at @dentry on vfsmount @mnt depending on
100 * @dir. If @dir is set return the first mount else return the last mount.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 */
Ram Paia05964f2005-11-07 17:20:17 -0500102struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry,
103 int dir)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104{
Ram Paib58fed82005-11-07 17:16:09 -0500105 struct list_head *head = mount_hashtable + hash(mnt, dentry);
106 struct list_head *tmp = head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 struct vfsmount *p, *found = NULL;
108
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 for (;;) {
Ram Paia05964f2005-11-07 17:20:17 -0500110 tmp = dir ? tmp->next : tmp->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 p = NULL;
112 if (tmp == head)
113 break;
114 p = list_entry(tmp, struct vfsmount, mnt_hash);
115 if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) {
Ram Paia05964f2005-11-07 17:20:17 -0500116 found = p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 break;
118 }
119 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120 return found;
121}
122
Ram Paia05964f2005-11-07 17:20:17 -0500123/*
124 * lookup_mnt increments the ref count before returning
125 * the vfsmount struct.
126 */
127struct vfsmount *lookup_mnt(struct vfsmount *mnt, struct dentry *dentry)
128{
129 struct vfsmount *child_mnt;
130 spin_lock(&vfsmount_lock);
131 if ((child_mnt = __lookup_mnt(mnt, dentry, 1)))
132 mntget(child_mnt);
133 spin_unlock(&vfsmount_lock);
134 return child_mnt;
135}
136
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137static inline int check_mnt(struct vfsmount *mnt)
138{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800139 return mnt->mnt_ns == current->nsproxy->mnt_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140}
141
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800142static void touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500143{
144 if (ns) {
145 ns->event = ++event;
146 wake_up_interruptible(&ns->poll);
147 }
148}
149
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800150static void __touch_mnt_namespace(struct mnt_namespace *ns)
Al Viro5addc5d2005-11-07 17:15:49 -0500151{
152 if (ns && ns->event != event) {
153 ns->event = event;
154 wake_up_interruptible(&ns->poll);
155 }
156}
157
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158static void detach_mnt(struct vfsmount *mnt, struct nameidata *old_nd)
159{
160 old_nd->dentry = mnt->mnt_mountpoint;
161 old_nd->mnt = mnt->mnt_parent;
162 mnt->mnt_parent = mnt;
163 mnt->mnt_mountpoint = mnt->mnt_root;
164 list_del_init(&mnt->mnt_child);
165 list_del_init(&mnt->mnt_hash);
166 old_nd->dentry->d_mounted--;
167}
168
Ram Paib90fa9a2005-11-07 17:19:50 -0500169void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry,
170 struct vfsmount *child_mnt)
171{
172 child_mnt->mnt_parent = mntget(mnt);
173 child_mnt->mnt_mountpoint = dget(dentry);
174 dentry->d_mounted++;
175}
176
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177static void attach_mnt(struct vfsmount *mnt, struct nameidata *nd)
178{
Ram Paib90fa9a2005-11-07 17:19:50 -0500179 mnt_set_mountpoint(nd->mnt, nd->dentry, mnt);
180 list_add_tail(&mnt->mnt_hash, mount_hashtable +
181 hash(nd->mnt, nd->dentry));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 list_add_tail(&mnt->mnt_child, &nd->mnt->mnt_mounts);
Ram Paib90fa9a2005-11-07 17:19:50 -0500183}
184
185/*
186 * the caller must hold vfsmount_lock
187 */
188static void commit_tree(struct vfsmount *mnt)
189{
190 struct vfsmount *parent = mnt->mnt_parent;
191 struct vfsmount *m;
192 LIST_HEAD(head);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800193 struct mnt_namespace *n = parent->mnt_ns;
Ram Paib90fa9a2005-11-07 17:19:50 -0500194
195 BUG_ON(parent == mnt);
196
197 list_add_tail(&head, &mnt->mnt_list);
198 list_for_each_entry(m, &head, mnt_list)
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800199 m->mnt_ns = n;
Ram Paib90fa9a2005-11-07 17:19:50 -0500200 list_splice(&head, n->list.prev);
201
202 list_add_tail(&mnt->mnt_hash, mount_hashtable +
203 hash(parent, mnt->mnt_mountpoint));
204 list_add_tail(&mnt->mnt_child, &parent->mnt_mounts);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800205 touch_mnt_namespace(n);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206}
207
208static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root)
209{
210 struct list_head *next = p->mnt_mounts.next;
211 if (next == &p->mnt_mounts) {
212 while (1) {
213 if (p == root)
214 return NULL;
215 next = p->mnt_child.next;
216 if (next != &p->mnt_parent->mnt_mounts)
217 break;
218 p = p->mnt_parent;
219 }
220 }
221 return list_entry(next, struct vfsmount, mnt_child);
222}
223
Ram Pai9676f0c2005-11-07 17:21:20 -0500224static struct vfsmount *skip_mnt_tree(struct vfsmount *p)
225{
226 struct list_head *prev = p->mnt_mounts.prev;
227 while (prev != &p->mnt_mounts) {
228 p = list_entry(prev, struct vfsmount, mnt_child);
229 prev = p->mnt_mounts.prev;
230 }
231 return p;
232}
233
Ram Pai36341f62005-11-07 17:17:22 -0500234static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root,
235 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236{
237 struct super_block *sb = old->mnt_sb;
238 struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname);
239
240 if (mnt) {
241 mnt->mnt_flags = old->mnt_flags;
242 atomic_inc(&sb->s_active);
243 mnt->mnt_sb = sb;
244 mnt->mnt_root = dget(root);
245 mnt->mnt_mountpoint = mnt->mnt_root;
246 mnt->mnt_parent = mnt;
Ram Paib90fa9a2005-11-07 17:19:50 -0500247
Ram Pai5afe0022005-11-07 17:21:01 -0500248 if (flag & CL_SLAVE) {
249 list_add(&mnt->mnt_slave, &old->mnt_slave_list);
250 mnt->mnt_master = old;
251 CLEAR_MNT_SHARED(mnt);
Al Viro8aec0802007-06-07 12:20:32 -0400252 } else if (!(flag & CL_PRIVATE)) {
Ram Pai5afe0022005-11-07 17:21:01 -0500253 if ((flag & CL_PROPAGATION) || IS_MNT_SHARED(old))
254 list_add(&mnt->mnt_share, &old->mnt_share);
255 if (IS_MNT_SLAVE(old))
256 list_add(&mnt->mnt_slave, &old->mnt_slave);
257 mnt->mnt_master = old->mnt_master;
258 }
Ram Paib90fa9a2005-11-07 17:19:50 -0500259 if (flag & CL_MAKE_SHARED)
260 set_mnt_shared(mnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
262 /* stick the duplicate mount on the same expiry list
263 * as the original if that was on one */
Ram Pai36341f62005-11-07 17:17:22 -0500264 if (flag & CL_EXPIRE) {
265 spin_lock(&vfsmount_lock);
266 if (!list_empty(&old->mnt_expire))
267 list_add(&mnt->mnt_expire, &old->mnt_expire);
268 spin_unlock(&vfsmount_lock);
269 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 }
271 return mnt;
272}
273
Al Viro7b7b1ac2005-11-07 17:13:39 -0500274static inline void __mntput(struct vfsmount *mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275{
276 struct super_block *sb = mnt->mnt_sb;
277 dput(mnt->mnt_root);
278 free_vfsmnt(mnt);
279 deactivate_super(sb);
280}
281
Al Viro7b7b1ac2005-11-07 17:13:39 -0500282void mntput_no_expire(struct vfsmount *mnt)
283{
284repeat:
285 if (atomic_dec_and_lock(&mnt->mnt_count, &vfsmount_lock)) {
286 if (likely(!mnt->mnt_pinned)) {
287 spin_unlock(&vfsmount_lock);
288 __mntput(mnt);
289 return;
290 }
291 atomic_add(mnt->mnt_pinned + 1, &mnt->mnt_count);
292 mnt->mnt_pinned = 0;
293 spin_unlock(&vfsmount_lock);
294 acct_auto_close_mnt(mnt);
295 security_sb_umount_close(mnt);
296 goto repeat;
297 }
298}
299
300EXPORT_SYMBOL(mntput_no_expire);
301
302void mnt_pin(struct vfsmount *mnt)
303{
304 spin_lock(&vfsmount_lock);
305 mnt->mnt_pinned++;
306 spin_unlock(&vfsmount_lock);
307}
308
309EXPORT_SYMBOL(mnt_pin);
310
311void mnt_unpin(struct vfsmount *mnt)
312{
313 spin_lock(&vfsmount_lock);
314 if (mnt->mnt_pinned) {
315 atomic_inc(&mnt->mnt_count);
316 mnt->mnt_pinned--;
317 }
318 spin_unlock(&vfsmount_lock);
319}
320
321EXPORT_SYMBOL(mnt_unpin);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322
Miklos Szeredib3b304a2008-02-08 04:21:35 -0800323static inline void mangle(struct seq_file *m, const char *s)
324{
325 seq_escape(m, s, " \t\n\\");
326}
327
328/*
329 * Simple .show_options callback for filesystems which don't want to
330 * implement more complex mount option showing.
331 *
332 * See also save_mount_options().
333 */
334int generic_show_options(struct seq_file *m, struct vfsmount *mnt)
335{
336 const char *options = mnt->mnt_sb->s_options;
337
338 if (options != NULL && options[0]) {
339 seq_putc(m, ',');
340 mangle(m, options);
341 }
342
343 return 0;
344}
345EXPORT_SYMBOL(generic_show_options);
346
347/*
348 * If filesystem uses generic_show_options(), this function should be
349 * called from the fill_super() callback.
350 *
351 * The .remount_fs callback usually needs to be handled in a special
352 * way, to make sure, that previous options are not overwritten if the
353 * remount fails.
354 *
355 * Also note, that if the filesystem's .remount_fs function doesn't
356 * reset all options to their default value, but changes only newly
357 * given options, then the displayed options will not reflect reality
358 * any more.
359 */
360void save_mount_options(struct super_block *sb, char *options)
361{
362 kfree(sb->s_options);
363 sb->s_options = kstrdup(options, GFP_KERNEL);
364}
365EXPORT_SYMBOL(save_mount_options);
366
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367/* iterator */
368static void *m_start(struct seq_file *m, loff_t *pos)
369{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800370 struct mnt_namespace *n = m->private;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371
Ram Pai390c6842005-11-07 17:17:51 -0500372 down_read(&namespace_sem);
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700373 return seq_list_start(&n->list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374}
375
376static void *m_next(struct seq_file *m, void *v, loff_t *pos)
377{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800378 struct mnt_namespace *n = m->private;
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700379
380 return seq_list_next(v, &n->list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381}
382
383static void m_stop(struct seq_file *m, void *v)
384{
Ram Pai390c6842005-11-07 17:17:51 -0500385 up_read(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386}
387
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388static int show_vfsmnt(struct seq_file *m, void *v)
389{
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700390 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 int err = 0;
392 static struct proc_fs_info {
393 int flag;
394 char *str;
395 } fs_info[] = {
396 { MS_SYNCHRONOUS, ",sync" },
397 { MS_DIRSYNC, ",dirsync" },
398 { MS_MANDLOCK, ",mand" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 { 0, NULL }
400 };
401 static struct proc_fs_info mnt_info[] = {
402 { MNT_NOSUID, ",nosuid" },
403 { MNT_NODEV, ",nodev" },
404 { MNT_NOEXEC, ",noexec" },
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -0800405 { MNT_NOATIME, ",noatime" },
406 { MNT_NODIRATIME, ",nodiratime" },
Valerie Henson47ae32d2006-12-13 00:34:34 -0800407 { MNT_RELATIME, ",relatime" },
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 { 0, NULL }
409 };
410 struct proc_fs_info *fs_infop;
411
412 mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none");
413 seq_putc(m, ' ');
414 seq_path(m, mnt, mnt->mnt_root, " \t\n\\");
415 seq_putc(m, ' ');
416 mangle(m, mnt->mnt_sb->s_type->name);
Miklos Szeredi79c0b2d2007-05-08 00:25:43 -0700417 if (mnt->mnt_sb->s_subtype && mnt->mnt_sb->s_subtype[0]) {
418 seq_putc(m, '.');
419 mangle(m, mnt->mnt_sb->s_subtype);
420 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 seq_puts(m, mnt->mnt_sb->s_flags & MS_RDONLY ? " ro" : " rw");
422 for (fs_infop = fs_info; fs_infop->flag; fs_infop++) {
423 if (mnt->mnt_sb->s_flags & fs_infop->flag)
424 seq_puts(m, fs_infop->str);
425 }
426 for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) {
427 if (mnt->mnt_flags & fs_infop->flag)
428 seq_puts(m, fs_infop->str);
429 }
430 if (mnt->mnt_sb->s_op->show_options)
431 err = mnt->mnt_sb->s_op->show_options(m, mnt);
432 seq_puts(m, " 0 0\n");
433 return err;
434}
435
436struct seq_operations mounts_op = {
437 .start = m_start,
438 .next = m_next,
439 .stop = m_stop,
440 .show = show_vfsmnt
441};
442
Chuck Leverb4629fe2006-03-20 13:44:12 -0500443static int show_vfsstat(struct seq_file *m, void *v)
444{
Pavel Emelianovb0765fb2007-07-15 23:39:55 -0700445 struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list);
Chuck Leverb4629fe2006-03-20 13:44:12 -0500446 int err = 0;
447
448 /* device */
449 if (mnt->mnt_devname) {
450 seq_puts(m, "device ");
451 mangle(m, mnt->mnt_devname);
452 } else
453 seq_puts(m, "no device");
454
455 /* mount point */
456 seq_puts(m, " mounted on ");
457 seq_path(m, mnt, mnt->mnt_root, " \t\n\\");
458 seq_putc(m, ' ');
459
460 /* file system type */
461 seq_puts(m, "with fstype ");
462 mangle(m, mnt->mnt_sb->s_type->name);
463
464 /* optional statistics */
465 if (mnt->mnt_sb->s_op->show_stats) {
466 seq_putc(m, ' ');
467 err = mnt->mnt_sb->s_op->show_stats(m, mnt);
468 }
469
470 seq_putc(m, '\n');
471 return err;
472}
473
474struct seq_operations mountstats_op = {
475 .start = m_start,
476 .next = m_next,
477 .stop = m_stop,
478 .show = show_vfsstat,
479};
480
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481/**
482 * may_umount_tree - check if a mount tree is busy
483 * @mnt: root of mount tree
484 *
485 * This is called to check if a tree of mounts has any
486 * open files, pwds, chroots or sub mounts that are
487 * busy.
488 */
489int may_umount_tree(struct vfsmount *mnt)
490{
Ram Pai36341f62005-11-07 17:17:22 -0500491 int actual_refs = 0;
492 int minimum_refs = 0;
493 struct vfsmount *p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494
495 spin_lock(&vfsmount_lock);
Ram Pai36341f62005-11-07 17:17:22 -0500496 for (p = mnt; p; p = next_mnt(p, mnt)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 actual_refs += atomic_read(&p->mnt_count);
498 minimum_refs += 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 }
500 spin_unlock(&vfsmount_lock);
501
502 if (actual_refs > minimum_refs)
Ian Kente3474a82006-03-27 01:14:51 -0800503 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Ian Kente3474a82006-03-27 01:14:51 -0800505 return 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506}
507
508EXPORT_SYMBOL(may_umount_tree);
509
510/**
511 * may_umount - check if a mount point is busy
512 * @mnt: root of mount
513 *
514 * This is called to check if a mount point has any
515 * open files, pwds, chroots or sub mounts. If the
516 * mount has sub mounts this will return busy
517 * regardless of whether the sub mounts are busy.
518 *
519 * Doesn't take quota and stuff into account. IOW, in some cases it will
520 * give false negatives. The main reason why it's here is that we need
521 * a non-destructive way to look for easily umountable filesystems.
522 */
523int may_umount(struct vfsmount *mnt)
524{
Ian Kente3474a82006-03-27 01:14:51 -0800525 int ret = 1;
Ram Paia05964f2005-11-07 17:20:17 -0500526 spin_lock(&vfsmount_lock);
527 if (propagate_mount_busy(mnt, 2))
Ian Kente3474a82006-03-27 01:14:51 -0800528 ret = 0;
Ram Paia05964f2005-11-07 17:20:17 -0500529 spin_unlock(&vfsmount_lock);
530 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531}
532
533EXPORT_SYMBOL(may_umount);
534
Ram Paib90fa9a2005-11-07 17:19:50 -0500535void release_mounts(struct list_head *head)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536{
Ram Pai70fbcdf2005-11-07 17:17:04 -0500537 struct vfsmount *mnt;
Miklos Szeredibf066c72006-01-08 01:03:19 -0800538 while (!list_empty(head)) {
Pavel Emelianovb5e61812007-05-08 00:30:19 -0700539 mnt = list_first_entry(head, struct vfsmount, mnt_hash);
Ram Pai70fbcdf2005-11-07 17:17:04 -0500540 list_del_init(&mnt->mnt_hash);
541 if (mnt->mnt_parent != mnt) {
542 struct dentry *dentry;
543 struct vfsmount *m;
544 spin_lock(&vfsmount_lock);
545 dentry = mnt->mnt_mountpoint;
546 m = mnt->mnt_parent;
547 mnt->mnt_mountpoint = mnt->mnt_root;
548 mnt->mnt_parent = mnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 spin_unlock(&vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -0500550 dput(dentry);
551 mntput(m);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552 }
553 mntput(mnt);
Ram Pai70fbcdf2005-11-07 17:17:04 -0500554 }
555}
556
Ram Paia05964f2005-11-07 17:20:17 -0500557void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill)
Ram Pai70fbcdf2005-11-07 17:17:04 -0500558{
559 struct vfsmount *p;
560
Akinobu Mita1bfba4e2006-06-26 00:24:40 -0700561 for (p = mnt; p; p = next_mnt(p, mnt))
562 list_move(&p->mnt_hash, kill);
Ram Pai70fbcdf2005-11-07 17:17:04 -0500563
Ram Paia05964f2005-11-07 17:20:17 -0500564 if (propagate)
565 propagate_umount(kill);
566
Ram Pai70fbcdf2005-11-07 17:17:04 -0500567 list_for_each_entry(p, kill, mnt_hash) {
568 list_del_init(&p->mnt_expire);
569 list_del_init(&p->mnt_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800570 __touch_mnt_namespace(p->mnt_ns);
571 p->mnt_ns = NULL;
Ram Pai70fbcdf2005-11-07 17:17:04 -0500572 list_del_init(&p->mnt_child);
573 if (p->mnt_parent != p)
Al Virof30ac312006-02-01 07:53:21 -0500574 p->mnt_mountpoint->d_mounted--;
Ram Paia05964f2005-11-07 17:20:17 -0500575 change_mnt_propagation(p, MS_PRIVATE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 }
577}
578
579static int do_umount(struct vfsmount *mnt, int flags)
580{
Ram Paib58fed82005-11-07 17:16:09 -0500581 struct super_block *sb = mnt->mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 int retval;
Ram Pai70fbcdf2005-11-07 17:17:04 -0500583 LIST_HEAD(umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
585 retval = security_sb_umount(mnt, flags);
586 if (retval)
587 return retval;
588
589 /*
590 * Allow userspace to request a mountpoint be expired rather than
591 * unmounting unconditionally. Unmount only happens if:
592 * (1) the mark is already set (the mark is cleared by mntput())
593 * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount]
594 */
595 if (flags & MNT_EXPIRE) {
596 if (mnt == current->fs->rootmnt ||
597 flags & (MNT_FORCE | MNT_DETACH))
598 return -EINVAL;
599
600 if (atomic_read(&mnt->mnt_count) != 2)
601 return -EBUSY;
602
603 if (!xchg(&mnt->mnt_expiry_mark, 1))
604 return -EAGAIN;
605 }
606
607 /*
608 * If we may have to abort operations to get out of this
609 * mount, and they will themselves hold resources we must
610 * allow the fs to do things. In the Unix tradition of
611 * 'Gee thats tricky lets do it in userspace' the umount_begin
612 * might fail to complete on the first run through as other tasks
613 * must return, and the like. Thats for the mount program to worry
614 * about for the moment.
615 */
616
617 lock_kernel();
Trond Myklebust8b512d92006-06-09 09:34:18 -0400618 if (sb->s_op->umount_begin)
619 sb->s_op->umount_begin(mnt, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 unlock_kernel();
621
622 /*
623 * No sense to grab the lock for this test, but test itself looks
624 * somewhat bogus. Suggestions for better replacement?
625 * Ho-hum... In principle, we might treat that as umount + switch
626 * to rootfs. GC would eventually take care of the old vfsmount.
627 * Actually it makes sense, especially if rootfs would contain a
628 * /reboot - static binary that would close all descriptors and
629 * call reboot(9). Then init(8) could umount root and exec /reboot.
630 */
631 if (mnt == current->fs->rootmnt && !(flags & MNT_DETACH)) {
632 /*
633 * Special case for "unmounting" root ...
634 * we just try to remount it readonly.
635 */
636 down_write(&sb->s_umount);
637 if (!(sb->s_flags & MS_RDONLY)) {
638 lock_kernel();
639 DQUOT_OFF(sb);
640 retval = do_remount_sb(sb, MS_RDONLY, NULL, 0);
641 unlock_kernel();
642 }
643 up_write(&sb->s_umount);
644 return retval;
645 }
646
Ram Pai390c6842005-11-07 17:17:51 -0500647 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 spin_lock(&vfsmount_lock);
Al Viro5addc5d2005-11-07 17:15:49 -0500649 event++;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 retval = -EBUSY;
Ram Paia05964f2005-11-07 17:20:17 -0500652 if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 if (!list_empty(&mnt->mnt_list))
Ram Paia05964f2005-11-07 17:20:17 -0500654 umount_tree(mnt, 1, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 retval = 0;
656 }
657 spin_unlock(&vfsmount_lock);
658 if (retval)
659 security_sb_umount_busy(mnt);
Ram Pai390c6842005-11-07 17:17:51 -0500660 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -0500661 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 return retval;
663}
664
665/*
666 * Now umount can handle mount points as well as block devices.
667 * This is important for filesystems which use unnamed block devices.
668 *
669 * We now support a flag for forced unmount like the other 'big iron'
670 * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD
671 */
672
673asmlinkage long sys_umount(char __user * name, int flags)
674{
675 struct nameidata nd;
676 int retval;
677
678 retval = __user_walk(name, LOOKUP_FOLLOW, &nd);
679 if (retval)
680 goto out;
681 retval = -EINVAL;
682 if (nd.dentry != nd.mnt->mnt_root)
683 goto dput_and_out;
684 if (!check_mnt(nd.mnt))
685 goto dput_and_out;
686
687 retval = -EPERM;
688 if (!capable(CAP_SYS_ADMIN))
689 goto dput_and_out;
690
691 retval = do_umount(nd.mnt, flags);
692dput_and_out:
693 path_release_on_umount(&nd);
694out:
695 return retval;
696}
697
698#ifdef __ARCH_WANT_SYS_OLDUMOUNT
699
700/*
Ram Paib58fed82005-11-07 17:16:09 -0500701 * The 2.0 compatible umount. No flags.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703asmlinkage long sys_oldumount(char __user * name)
704{
Ram Paib58fed82005-11-07 17:16:09 -0500705 return sys_umount(name, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706}
707
708#endif
709
710static int mount_is_safe(struct nameidata *nd)
711{
712 if (capable(CAP_SYS_ADMIN))
713 return 0;
714 return -EPERM;
715#ifdef notyet
716 if (S_ISLNK(nd->dentry->d_inode->i_mode))
717 return -EPERM;
718 if (nd->dentry->d_inode->i_mode & S_ISVTX) {
719 if (current->uid != nd->dentry->d_inode->i_uid)
720 return -EPERM;
721 }
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800722 if (vfs_permission(nd, MAY_WRITE))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 return -EPERM;
724 return 0;
725#endif
726}
727
Ram Paib58fed82005-11-07 17:16:09 -0500728static int lives_below_in_same_fs(struct dentry *d, struct dentry *dentry)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729{
730 while (1) {
731 if (d == dentry)
732 return 1;
733 if (d == NULL || d == d->d_parent)
734 return 0;
735 d = d->d_parent;
736 }
737}
738
Ram Paib90fa9a2005-11-07 17:19:50 -0500739struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry,
Ram Pai36341f62005-11-07 17:17:22 -0500740 int flag)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741{
742 struct vfsmount *res, *p, *q, *r, *s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 struct nameidata nd;
744
Ram Pai9676f0c2005-11-07 17:21:20 -0500745 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt))
746 return NULL;
747
Ram Pai36341f62005-11-07 17:17:22 -0500748 res = q = clone_mnt(mnt, dentry, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 if (!q)
750 goto Enomem;
751 q->mnt_mountpoint = mnt->mnt_mountpoint;
752
753 p = mnt;
Domen Puncerfdadd652005-09-10 00:27:07 -0700754 list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 if (!lives_below_in_same_fs(r->mnt_mountpoint, dentry))
756 continue;
757
758 for (s = r; s; s = next_mnt(s, r)) {
Ram Pai9676f0c2005-11-07 17:21:20 -0500759 if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) {
760 s = skip_mnt_tree(s);
761 continue;
762 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763 while (p != s->mnt_parent) {
764 p = p->mnt_parent;
765 q = q->mnt_parent;
766 }
767 p = s;
768 nd.mnt = q;
769 nd.dentry = p->mnt_mountpoint;
Ram Pai36341f62005-11-07 17:17:22 -0500770 q = clone_mnt(p, p->mnt_root, flag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 if (!q)
772 goto Enomem;
773 spin_lock(&vfsmount_lock);
774 list_add_tail(&q->mnt_list, &res->mnt_list);
775 attach_mnt(q, &nd);
776 spin_unlock(&vfsmount_lock);
777 }
778 }
779 return res;
Ram Paib58fed82005-11-07 17:16:09 -0500780Enomem:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781 if (res) {
Ram Pai70fbcdf2005-11-07 17:17:04 -0500782 LIST_HEAD(umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700783 spin_lock(&vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -0500784 umount_tree(res, 0, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 spin_unlock(&vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -0500786 release_mounts(&umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 }
788 return NULL;
789}
790
Al Viro8aec0802007-06-07 12:20:32 -0400791struct vfsmount *collect_mounts(struct vfsmount *mnt, struct dentry *dentry)
792{
793 struct vfsmount *tree;
794 down_read(&namespace_sem);
795 tree = copy_tree(mnt, dentry, CL_COPY_ALL | CL_PRIVATE);
796 up_read(&namespace_sem);
797 return tree;
798}
799
800void drop_collected_mounts(struct vfsmount *mnt)
801{
802 LIST_HEAD(umount_list);
803 down_read(&namespace_sem);
804 spin_lock(&vfsmount_lock);
805 umount_tree(mnt, 0, &umount_list);
806 spin_unlock(&vfsmount_lock);
807 up_read(&namespace_sem);
808 release_mounts(&umount_list);
809}
810
Ram Paib90fa9a2005-11-07 17:19:50 -0500811/*
812 * @source_mnt : mount tree to be attached
Ram Pai21444402005-11-07 17:20:03 -0500813 * @nd : place the mount tree @source_mnt is attached
814 * @parent_nd : if non-null, detach the source_mnt from its parent and
815 * store the parent mount and mountpoint dentry.
816 * (done when source_mnt is moved)
Ram Paib90fa9a2005-11-07 17:19:50 -0500817 *
818 * NOTE: in the table below explains the semantics when a source mount
819 * of a given type is attached to a destination mount of a given type.
Ram Pai9676f0c2005-11-07 17:21:20 -0500820 * ---------------------------------------------------------------------------
821 * | BIND MOUNT OPERATION |
822 * |**************************************************************************
823 * | source-->| shared | private | slave | unbindable |
824 * | dest | | | | |
825 * | | | | | | |
826 * | v | | | | |
827 * |**************************************************************************
828 * | shared | shared (++) | shared (+) | shared(+++)| invalid |
829 * | | | | | |
830 * |non-shared| shared (+) | private | slave (*) | invalid |
831 * ***************************************************************************
Ram Paib90fa9a2005-11-07 17:19:50 -0500832 * A bind operation clones the source mount and mounts the clone on the
833 * destination mount.
834 *
835 * (++) the cloned mount is propagated to all the mounts in the propagation
836 * tree of the destination mount and the cloned mount is added to
837 * the peer group of the source mount.
838 * (+) the cloned mount is created under the destination mount and is marked
839 * as shared. The cloned mount is added to the peer group of the source
840 * mount.
Ram Pai5afe0022005-11-07 17:21:01 -0500841 * (+++) the mount is propagated to all the mounts in the propagation tree
842 * of the destination mount and the cloned mount is made slave
843 * of the same master as that of the source mount. The cloned mount
844 * is marked as 'shared and slave'.
845 * (*) the cloned mount is made a slave of the same master as that of the
846 * source mount.
847 *
Ram Pai9676f0c2005-11-07 17:21:20 -0500848 * ---------------------------------------------------------------------------
849 * | MOVE MOUNT OPERATION |
850 * |**************************************************************************
851 * | source-->| shared | private | slave | unbindable |
852 * | dest | | | | |
853 * | | | | | | |
854 * | v | | | | |
855 * |**************************************************************************
856 * | shared | shared (+) | shared (+) | shared(+++) | invalid |
857 * | | | | | |
858 * |non-shared| shared (+*) | private | slave (*) | unbindable |
859 * ***************************************************************************
Ram Pai5afe0022005-11-07 17:21:01 -0500860 *
861 * (+) the mount is moved to the destination. And is then propagated to
862 * all the mounts in the propagation tree of the destination mount.
Ram Pai21444402005-11-07 17:20:03 -0500863 * (+*) the mount is moved to the destination.
Ram Pai5afe0022005-11-07 17:21:01 -0500864 * (+++) the mount is moved to the destination and is then propagated to
865 * all the mounts belonging to the destination mount's propagation tree.
866 * the mount is marked as 'shared and slave'.
867 * (*) the mount continues to be a slave at the new location.
Ram Paib90fa9a2005-11-07 17:19:50 -0500868 *
869 * if the source mount is a tree, the operations explained above is
870 * applied to each mount in the tree.
871 * Must be called without spinlocks held, since this function can sleep
872 * in allocations.
873 */
874static int attach_recursive_mnt(struct vfsmount *source_mnt,
Ram Pai21444402005-11-07 17:20:03 -0500875 struct nameidata *nd, struct nameidata *parent_nd)
Ram Paib90fa9a2005-11-07 17:19:50 -0500876{
877 LIST_HEAD(tree_list);
878 struct vfsmount *dest_mnt = nd->mnt;
879 struct dentry *dest_dentry = nd->dentry;
880 struct vfsmount *child, *p;
881
882 if (propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list))
883 return -EINVAL;
884
885 if (IS_MNT_SHARED(dest_mnt)) {
886 for (p = source_mnt; p; p = next_mnt(p, source_mnt))
887 set_mnt_shared(p);
888 }
889
890 spin_lock(&vfsmount_lock);
Ram Pai21444402005-11-07 17:20:03 -0500891 if (parent_nd) {
892 detach_mnt(source_mnt, parent_nd);
893 attach_mnt(source_mnt, nd);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -0800894 touch_mnt_namespace(current->nsproxy->mnt_ns);
Ram Pai21444402005-11-07 17:20:03 -0500895 } else {
896 mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt);
897 commit_tree(source_mnt);
898 }
Ram Paib90fa9a2005-11-07 17:19:50 -0500899
900 list_for_each_entry_safe(child, p, &tree_list, mnt_hash) {
901 list_del_init(&child->mnt_hash);
902 commit_tree(child);
903 }
904 spin_unlock(&vfsmount_lock);
905 return 0;
906}
907
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908static int graft_tree(struct vfsmount *mnt, struct nameidata *nd)
909{
910 int err;
911 if (mnt->mnt_sb->s_flags & MS_NOUSER)
912 return -EINVAL;
913
914 if (S_ISDIR(nd->dentry->d_inode->i_mode) !=
915 S_ISDIR(mnt->mnt_root->d_inode->i_mode))
916 return -ENOTDIR;
917
918 err = -ENOENT;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800919 mutex_lock(&nd->dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 if (IS_DEADDIR(nd->dentry->d_inode))
921 goto out_unlock;
922
923 err = security_sb_check_sb(mnt, nd);
924 if (err)
925 goto out_unlock;
926
927 err = -ENOENT;
Ram Paib90fa9a2005-11-07 17:19:50 -0500928 if (IS_ROOT(nd->dentry) || !d_unhashed(nd->dentry))
Ram Pai21444402005-11-07 17:20:03 -0500929 err = attach_recursive_mnt(mnt, nd, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930out_unlock:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -0800931 mutex_unlock(&nd->dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 if (!err)
933 security_sb_post_addmount(mnt, nd);
934 return err;
935}
936
937/*
Ram Pai07b20882005-11-07 17:19:07 -0500938 * recursively change the type of the mountpoint.
Eric Sandeen2dafe1c2008-02-08 04:22:12 -0800939 * noinline this do_mount helper to save do_mount stack space.
Ram Pai07b20882005-11-07 17:19:07 -0500940 */
Eric Sandeen2dafe1c2008-02-08 04:22:12 -0800941static noinline int do_change_type(struct nameidata *nd, int flag)
Ram Pai07b20882005-11-07 17:19:07 -0500942{
943 struct vfsmount *m, *mnt = nd->mnt;
944 int recurse = flag & MS_REC;
945 int type = flag & ~MS_REC;
946
Miklos Szerediee6f9582007-05-08 00:30:40 -0700947 if (!capable(CAP_SYS_ADMIN))
948 return -EPERM;
949
Ram Pai07b20882005-11-07 17:19:07 -0500950 if (nd->dentry != nd->mnt->mnt_root)
951 return -EINVAL;
952
953 down_write(&namespace_sem);
954 spin_lock(&vfsmount_lock);
955 for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL))
956 change_mnt_propagation(m, type);
957 spin_unlock(&vfsmount_lock);
958 up_write(&namespace_sem);
959 return 0;
960}
961
962/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 * do loopback mount.
Eric Sandeen2dafe1c2008-02-08 04:22:12 -0800964 * noinline this do_mount helper to save do_mount stack space.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 */
Eric Sandeen2dafe1c2008-02-08 04:22:12 -0800966static noinline int do_loopback(struct nameidata *nd, char *old_name,
967 int recurse)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968{
969 struct nameidata old_nd;
970 struct vfsmount *mnt = NULL;
971 int err = mount_is_safe(nd);
972 if (err)
973 return err;
974 if (!old_name || !*old_name)
975 return -EINVAL;
976 err = path_lookup(old_name, LOOKUP_FOLLOW, &old_nd);
977 if (err)
978 return err;
979
Ram Pai390c6842005-11-07 17:17:51 -0500980 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 err = -EINVAL;
Ram Pai9676f0c2005-11-07 17:21:20 -0500982 if (IS_MNT_UNBINDABLE(old_nd.mnt))
983 goto out;
984
Al Viroccd48bc2005-11-07 17:15:04 -0500985 if (!check_mnt(nd->mnt) || !check_mnt(old_nd.mnt))
986 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987
Al Viroccd48bc2005-11-07 17:15:04 -0500988 err = -ENOMEM;
989 if (recurse)
Ram Pai36341f62005-11-07 17:17:22 -0500990 mnt = copy_tree(old_nd.mnt, old_nd.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -0500991 else
Ram Pai36341f62005-11-07 17:17:22 -0500992 mnt = clone_mnt(old_nd.mnt, old_nd.dentry, 0);
Al Viroccd48bc2005-11-07 17:15:04 -0500993
994 if (!mnt)
995 goto out;
996
Al Viroccd48bc2005-11-07 17:15:04 -0500997 err = graft_tree(mnt, nd);
998 if (err) {
Ram Pai70fbcdf2005-11-07 17:17:04 -0500999 LIST_HEAD(umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 spin_lock(&vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -05001001 umount_tree(mnt, 0, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 spin_unlock(&vfsmount_lock);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001003 release_mounts(&umount_list);
Ram Pai5b83d2c2005-11-07 17:16:29 -05001004 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Al Viroccd48bc2005-11-07 17:15:04 -05001006out:
Ram Pai390c6842005-11-07 17:17:51 -05001007 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 path_release(&old_nd);
1009 return err;
1010}
1011
1012/*
1013 * change filesystem flags. dir should be a physical root of filesystem.
1014 * If you've mounted a non-root directory somewhere and want to do remount
1015 * on it - tough luck.
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001016 * noinline this do_mount helper to save do_mount stack space.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 */
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001018static noinline int do_remount(struct nameidata *nd, int flags, int mnt_flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 void *data)
1020{
1021 int err;
Ram Paib58fed82005-11-07 17:16:09 -05001022 struct super_block *sb = nd->mnt->mnt_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023
1024 if (!capable(CAP_SYS_ADMIN))
1025 return -EPERM;
1026
1027 if (!check_mnt(nd->mnt))
1028 return -EINVAL;
1029
1030 if (nd->dentry != nd->mnt->mnt_root)
1031 return -EINVAL;
1032
1033 down_write(&sb->s_umount);
1034 err = do_remount_sb(sb, flags, data, 0);
1035 if (!err)
Ram Paib58fed82005-11-07 17:16:09 -05001036 nd->mnt->mnt_flags = mnt_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 up_write(&sb->s_umount);
1038 if (!err)
1039 security_sb_post_remount(nd->mnt, flags, data);
1040 return err;
1041}
1042
Ram Pai9676f0c2005-11-07 17:21:20 -05001043static inline int tree_contains_unbindable(struct vfsmount *mnt)
1044{
1045 struct vfsmount *p;
1046 for (p = mnt; p; p = next_mnt(p, mnt)) {
1047 if (IS_MNT_UNBINDABLE(p))
1048 return 1;
1049 }
1050 return 0;
1051}
1052
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001053/*
1054 * noinline this do_mount helper to save do_mount stack space.
1055 */
1056static noinline int do_move_mount(struct nameidata *nd, char *old_name)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057{
1058 struct nameidata old_nd, parent_nd;
1059 struct vfsmount *p;
1060 int err = 0;
1061 if (!capable(CAP_SYS_ADMIN))
1062 return -EPERM;
1063 if (!old_name || !*old_name)
1064 return -EINVAL;
1065 err = path_lookup(old_name, LOOKUP_FOLLOW, &old_nd);
1066 if (err)
1067 return err;
1068
Ram Pai390c6842005-11-07 17:17:51 -05001069 down_write(&namespace_sem);
Ram Paib58fed82005-11-07 17:16:09 -05001070 while (d_mountpoint(nd->dentry) && follow_down(&nd->mnt, &nd->dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 ;
1072 err = -EINVAL;
1073 if (!check_mnt(nd->mnt) || !check_mnt(old_nd.mnt))
1074 goto out;
1075
1076 err = -ENOENT;
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001077 mutex_lock(&nd->dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 if (IS_DEADDIR(nd->dentry->d_inode))
1079 goto out1;
1080
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 if (!IS_ROOT(nd->dentry) && d_unhashed(nd->dentry))
Ram Pai21444402005-11-07 17:20:03 -05001082 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083
1084 err = -EINVAL;
1085 if (old_nd.dentry != old_nd.mnt->mnt_root)
Ram Pai21444402005-11-07 17:20:03 -05001086 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087
1088 if (old_nd.mnt == old_nd.mnt->mnt_parent)
Ram Pai21444402005-11-07 17:20:03 -05001089 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001090
1091 if (S_ISDIR(nd->dentry->d_inode->i_mode) !=
1092 S_ISDIR(old_nd.dentry->d_inode->i_mode))
Ram Pai21444402005-11-07 17:20:03 -05001093 goto out1;
1094 /*
1095 * Don't move a mount residing in a shared parent.
1096 */
1097 if (old_nd.mnt->mnt_parent && IS_MNT_SHARED(old_nd.mnt->mnt_parent))
1098 goto out1;
Ram Pai9676f0c2005-11-07 17:21:20 -05001099 /*
1100 * Don't move a mount tree containing unbindable mounts to a destination
1101 * mount which is shared.
1102 */
1103 if (IS_MNT_SHARED(nd->mnt) && tree_contains_unbindable(old_nd.mnt))
1104 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 err = -ELOOP;
Ram Paib58fed82005-11-07 17:16:09 -05001106 for (p = nd->mnt; p->mnt_parent != p; p = p->mnt_parent)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 if (p == old_nd.mnt)
Ram Pai21444402005-11-07 17:20:03 -05001108 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109
Ram Pai21444402005-11-07 17:20:03 -05001110 if ((err = attach_recursive_mnt(old_nd.mnt, nd, &parent_nd)))
1111 goto out1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112
Ram Pai21444402005-11-07 17:20:03 -05001113 spin_lock(&vfsmount_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114 /* if the mount is moved, it should no longer be expire
1115 * automatically */
Miklos Szeredi55e700b2005-07-07 17:57:30 -07001116 list_del_init(&old_nd.mnt->mnt_expire);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 spin_unlock(&vfsmount_lock);
1118out1:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001119 mutex_unlock(&nd->dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120out:
Ram Pai390c6842005-11-07 17:17:51 -05001121 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 if (!err)
1123 path_release(&parent_nd);
1124 path_release(&old_nd);
1125 return err;
1126}
1127
1128/*
1129 * create a new mount for userspace and request it to be added into the
1130 * namespace's tree
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001131 * noinline this do_mount helper to save do_mount stack space.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 */
Eric Sandeen2dafe1c2008-02-08 04:22:12 -08001133static noinline int do_new_mount(struct nameidata *nd, char *type, int flags,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 int mnt_flags, char *name, void *data)
1135{
1136 struct vfsmount *mnt;
1137
1138 if (!type || !memchr(type, 0, PAGE_SIZE))
1139 return -EINVAL;
1140
1141 /* we need capabilities... */
1142 if (!capable(CAP_SYS_ADMIN))
1143 return -EPERM;
1144
1145 mnt = do_kern_mount(type, flags, name, data);
1146 if (IS_ERR(mnt))
1147 return PTR_ERR(mnt);
1148
1149 return do_add_mount(mnt, nd, mnt_flags, NULL);
1150}
1151
1152/*
1153 * add a mount into a namespace's mount tree
1154 * - provide the option of adding the new mount to an expiration list
1155 */
1156int do_add_mount(struct vfsmount *newmnt, struct nameidata *nd,
1157 int mnt_flags, struct list_head *fslist)
1158{
1159 int err;
1160
Ram Pai390c6842005-11-07 17:17:51 -05001161 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 /* Something was mounted here while we slept */
Ram Paib58fed82005-11-07 17:16:09 -05001163 while (d_mountpoint(nd->dentry) && follow_down(&nd->mnt, &nd->dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 ;
1165 err = -EINVAL;
1166 if (!check_mnt(nd->mnt))
1167 goto unlock;
1168
1169 /* Refuse the same filesystem on the same mount point */
1170 err = -EBUSY;
1171 if (nd->mnt->mnt_sb == newmnt->mnt_sb &&
1172 nd->mnt->mnt_root == nd->dentry)
1173 goto unlock;
1174
1175 err = -EINVAL;
1176 if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode))
1177 goto unlock;
1178
1179 newmnt->mnt_flags = mnt_flags;
Ram Pai5b83d2c2005-11-07 17:16:29 -05001180 if ((err = graft_tree(newmnt, nd)))
1181 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
Ram Pai5b83d2c2005-11-07 17:16:29 -05001183 if (fslist) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184 /* add to the specified expiration list */
1185 spin_lock(&vfsmount_lock);
Miklos Szeredi55e700b2005-07-07 17:57:30 -07001186 list_add_tail(&newmnt->mnt_expire, fslist);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 spin_unlock(&vfsmount_lock);
1188 }
Ram Pai390c6842005-11-07 17:17:51 -05001189 up_write(&namespace_sem);
Ram Pai5b83d2c2005-11-07 17:16:29 -05001190 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191
1192unlock:
Ram Pai390c6842005-11-07 17:17:51 -05001193 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194 mntput(newmnt);
1195 return err;
1196}
1197
1198EXPORT_SYMBOL_GPL(do_add_mount);
1199
Ram Pai70fbcdf2005-11-07 17:17:04 -05001200static void expire_mount(struct vfsmount *mnt, struct list_head *mounts,
1201 struct list_head *umounts)
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001202{
1203 spin_lock(&vfsmount_lock);
1204
1205 /*
Miklos Szeredied42c872005-07-07 17:57:26 -07001206 * Check if mount is still attached, if not, let whoever holds it deal
1207 * with the sucker
1208 */
1209 if (mnt->mnt_parent == mnt) {
1210 spin_unlock(&vfsmount_lock);
1211 return;
1212 }
1213
1214 /*
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001215 * Check that it is still dead: the count should now be 2 - as
1216 * contributed by the vfsmount parent and the mntget above
1217 */
Ram Paia05964f2005-11-07 17:20:17 -05001218 if (!propagate_mount_busy(mnt, 2)) {
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001219 /* delete from the namespace */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001220 touch_mnt_namespace(mnt->mnt_ns);
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001221 list_del_init(&mnt->mnt_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001222 mnt->mnt_ns = NULL;
Ram Paia05964f2005-11-07 17:20:17 -05001223 umount_tree(mnt, 1, umounts);
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001224 spin_unlock(&vfsmount_lock);
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001225 } else {
1226 /*
1227 * Someone brought it back to life whilst we didn't have any
1228 * locks held so return it to the expiration list
1229 */
Miklos Szeredi55e700b2005-07-07 17:57:30 -07001230 list_add_tail(&mnt->mnt_expire, mounts);
Miklos Szeredi24ca2af2005-07-07 17:57:25 -07001231 spin_unlock(&vfsmount_lock);
1232 }
1233}
1234
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235/*
Trond Myklebust5528f9112006-06-09 09:34:17 -04001236 * go through the vfsmounts we've just consigned to the graveyard to
1237 * - check that they're still dead
1238 * - delete the vfsmount from the appropriate namespace under lock
1239 * - dispose of the corpse
1240 */
1241static void expire_mount_list(struct list_head *graveyard, struct list_head *mounts)
1242{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001243 struct mnt_namespace *ns;
Trond Myklebust5528f9112006-06-09 09:34:17 -04001244 struct vfsmount *mnt;
1245
1246 while (!list_empty(graveyard)) {
1247 LIST_HEAD(umounts);
Pavel Emelianovb5e61812007-05-08 00:30:19 -07001248 mnt = list_first_entry(graveyard, struct vfsmount, mnt_expire);
Trond Myklebust5528f9112006-06-09 09:34:17 -04001249 list_del_init(&mnt->mnt_expire);
1250
1251 /* don't do anything if the namespace is dead - all the
1252 * vfsmounts from it are going away anyway */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001253 ns = mnt->mnt_ns;
1254 if (!ns || !ns->root)
Trond Myklebust5528f9112006-06-09 09:34:17 -04001255 continue;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001256 get_mnt_ns(ns);
Trond Myklebust5528f9112006-06-09 09:34:17 -04001257
1258 spin_unlock(&vfsmount_lock);
1259 down_write(&namespace_sem);
1260 expire_mount(mnt, mounts, &umounts);
1261 up_write(&namespace_sem);
1262 release_mounts(&umounts);
1263 mntput(mnt);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001264 put_mnt_ns(ns);
Trond Myklebust5528f9112006-06-09 09:34:17 -04001265 spin_lock(&vfsmount_lock);
1266 }
1267}
1268
1269/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 * process a list of expirable mountpoints with the intent of discarding any
1271 * mountpoints that aren't in use and haven't been touched since last we came
1272 * here
1273 */
1274void mark_mounts_for_expiry(struct list_head *mounts)
1275{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 struct vfsmount *mnt, *next;
1277 LIST_HEAD(graveyard);
1278
1279 if (list_empty(mounts))
1280 return;
1281
1282 spin_lock(&vfsmount_lock);
1283
1284 /* extract from the expiration list every vfsmount that matches the
1285 * following criteria:
1286 * - only referenced by its parent vfsmount
1287 * - still marked for expiry (marked on the last call here; marks are
1288 * cleared by mntput())
1289 */
Miklos Szeredi55e700b2005-07-07 17:57:30 -07001290 list_for_each_entry_safe(mnt, next, mounts, mnt_expire) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 if (!xchg(&mnt->mnt_expiry_mark, 1) ||
1292 atomic_read(&mnt->mnt_count) != 1)
1293 continue;
1294
1295 mntget(mnt);
Miklos Szeredi55e700b2005-07-07 17:57:30 -07001296 list_move(&mnt->mnt_expire, &graveyard);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 }
1298
Trond Myklebust5528f9112006-06-09 09:34:17 -04001299 expire_mount_list(&graveyard, mounts);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
1301 spin_unlock(&vfsmount_lock);
1302}
1303
1304EXPORT_SYMBOL_GPL(mark_mounts_for_expiry);
1305
1306/*
Trond Myklebust5528f9112006-06-09 09:34:17 -04001307 * Ripoff of 'select_parent()'
1308 *
1309 * search the list of submounts for a given mountpoint, and move any
1310 * shrinkable submounts to the 'graveyard' list.
1311 */
1312static int select_submounts(struct vfsmount *parent, struct list_head *graveyard)
1313{
1314 struct vfsmount *this_parent = parent;
1315 struct list_head *next;
1316 int found = 0;
1317
1318repeat:
1319 next = this_parent->mnt_mounts.next;
1320resume:
1321 while (next != &this_parent->mnt_mounts) {
1322 struct list_head *tmp = next;
1323 struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child);
1324
1325 next = tmp->next;
1326 if (!(mnt->mnt_flags & MNT_SHRINKABLE))
1327 continue;
1328 /*
1329 * Descend a level if the d_mounts list is non-empty.
1330 */
1331 if (!list_empty(&mnt->mnt_mounts)) {
1332 this_parent = mnt;
1333 goto repeat;
1334 }
1335
1336 if (!propagate_mount_busy(mnt, 1)) {
1337 mntget(mnt);
1338 list_move_tail(&mnt->mnt_expire, graveyard);
1339 found++;
1340 }
1341 }
1342 /*
1343 * All done at this level ... ascend and resume the search
1344 */
1345 if (this_parent != parent) {
1346 next = this_parent->mnt_child.next;
1347 this_parent = this_parent->mnt_parent;
1348 goto resume;
1349 }
1350 return found;
1351}
1352
1353/*
1354 * process a list of expirable mountpoints with the intent of discarding any
1355 * submounts of a specific parent mountpoint
1356 */
1357void shrink_submounts(struct vfsmount *mountpoint, struct list_head *mounts)
1358{
1359 LIST_HEAD(graveyard);
1360 int found;
1361
1362 spin_lock(&vfsmount_lock);
1363
1364 /* extract submounts of 'mountpoint' from the expiration list */
1365 while ((found = select_submounts(mountpoint, &graveyard)) != 0)
1366 expire_mount_list(&graveyard, mounts);
1367
1368 spin_unlock(&vfsmount_lock);
1369}
1370
1371EXPORT_SYMBOL_GPL(shrink_submounts);
1372
1373/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 * Some copy_from_user() implementations do not return the exact number of
1375 * bytes remaining to copy on a fault. But copy_mount_options() requires that.
1376 * Note that this function differs from copy_from_user() in that it will oops
1377 * on bad values of `to', rather than returning a short copy.
1378 */
Ram Paib58fed82005-11-07 17:16:09 -05001379static long exact_copy_from_user(void *to, const void __user * from,
1380 unsigned long n)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381{
1382 char *t = to;
1383 const char __user *f = from;
1384 char c;
1385
1386 if (!access_ok(VERIFY_READ, from, n))
1387 return n;
1388
1389 while (n) {
1390 if (__get_user(c, f)) {
1391 memset(t, 0, n);
1392 break;
1393 }
1394 *t++ = c;
1395 f++;
1396 n--;
1397 }
1398 return n;
1399}
1400
Ram Paib58fed82005-11-07 17:16:09 -05001401int copy_mount_options(const void __user * data, unsigned long *where)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402{
1403 int i;
1404 unsigned long page;
1405 unsigned long size;
Ram Paib58fed82005-11-07 17:16:09 -05001406
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 *where = 0;
1408 if (!data)
1409 return 0;
1410
1411 if (!(page = __get_free_page(GFP_KERNEL)))
1412 return -ENOMEM;
1413
1414 /* We only care that *some* data at the address the user
1415 * gave us is valid. Just in case, we'll zero
1416 * the remainder of the page.
1417 */
1418 /* copy_from_user cannot cross TASK_SIZE ! */
1419 size = TASK_SIZE - (unsigned long)data;
1420 if (size > PAGE_SIZE)
1421 size = PAGE_SIZE;
1422
1423 i = size - exact_copy_from_user((void *)page, data, size);
1424 if (!i) {
Ram Paib58fed82005-11-07 17:16:09 -05001425 free_page(page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 return -EFAULT;
1427 }
1428 if (i != PAGE_SIZE)
1429 memset((char *)page + i, 0, PAGE_SIZE - i);
1430 *where = page;
1431 return 0;
1432}
1433
1434/*
1435 * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to
1436 * be given to the mount() call (ie: read-only, no-dev, no-suid etc).
1437 *
1438 * data is a (void *) that can point to any structure up to
1439 * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent
1440 * information (or be NULL).
1441 *
1442 * Pre-0.97 versions of mount() didn't have a flags word.
1443 * When the flags word was introduced its top half was required
1444 * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9.
1445 * Therefore, if this magic number is present, it carries no information
1446 * and must be discarded.
1447 */
Ram Paib58fed82005-11-07 17:16:09 -05001448long do_mount(char *dev_name, char *dir_name, char *type_page,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 unsigned long flags, void *data_page)
1450{
1451 struct nameidata nd;
1452 int retval = 0;
1453 int mnt_flags = 0;
1454
1455 /* Discard magic */
1456 if ((flags & MS_MGC_MSK) == MS_MGC_VAL)
1457 flags &= ~MS_MGC_MSK;
1458
1459 /* Basic sanity checks */
1460
1461 if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE))
1462 return -EINVAL;
1463 if (dev_name && !memchr(dev_name, 0, PAGE_SIZE))
1464 return -EINVAL;
1465
1466 if (data_page)
1467 ((char *)data_page)[PAGE_SIZE - 1] = 0;
1468
1469 /* Separate the per-mountpoint flags */
1470 if (flags & MS_NOSUID)
1471 mnt_flags |= MNT_NOSUID;
1472 if (flags & MS_NODEV)
1473 mnt_flags |= MNT_NODEV;
1474 if (flags & MS_NOEXEC)
1475 mnt_flags |= MNT_NOEXEC;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001476 if (flags & MS_NOATIME)
1477 mnt_flags |= MNT_NOATIME;
1478 if (flags & MS_NODIRATIME)
1479 mnt_flags |= MNT_NODIRATIME;
Valerie Henson47ae32d2006-12-13 00:34:34 -08001480 if (flags & MS_RELATIME)
1481 mnt_flags |= MNT_RELATIME;
Christoph Hellwigfc33a7b2006-01-09 20:52:17 -08001482
1483 flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE |
Pavel Emelyanov8bf97252007-10-18 23:40:02 -07001484 MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485
1486 /* ... and get the mountpoint */
1487 retval = path_lookup(dir_name, LOOKUP_FOLLOW, &nd);
1488 if (retval)
1489 return retval;
1490
1491 retval = security_sb_mount(dev_name, &nd, type_page, flags, data_page);
1492 if (retval)
1493 goto dput_out;
1494
1495 if (flags & MS_REMOUNT)
1496 retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
1497 data_page);
1498 else if (flags & MS_BIND)
Andrew Mortoneee391a2006-05-15 09:44:30 -07001499 retval = do_loopback(&nd, dev_name, flags & MS_REC);
Ram Pai9676f0c2005-11-07 17:21:20 -05001500 else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE))
Ram Pai07b20882005-11-07 17:19:07 -05001501 retval = do_change_type(&nd, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 else if (flags & MS_MOVE)
1503 retval = do_move_mount(&nd, dev_name);
1504 else
1505 retval = do_new_mount(&nd, type_page, flags, mnt_flags,
1506 dev_name, data_page);
1507dput_out:
1508 path_release(&nd);
1509 return retval;
1510}
1511
JANAK DESAI741a2952006-02-07 12:59:00 -08001512/*
1513 * Allocate a new namespace structure and populate it with contents
1514 * copied from the namespace of the passed in task structure.
1515 */
Badari Pulavartye3222c42007-05-08 00:25:21 -07001516static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns,
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001517 struct fs_struct *fs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001519 struct mnt_namespace *new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 struct vfsmount *rootmnt = NULL, *pwdmnt = NULL, *altrootmnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521 struct vfsmount *p, *q;
1522
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001523 new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 if (!new_ns)
Cedric Le Goater467e9f42007-07-15 23:41:06 -07001525 return ERR_PTR(-ENOMEM);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526
1527 atomic_set(&new_ns->count, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 INIT_LIST_HEAD(&new_ns->list);
Al Viro5addc5d2005-11-07 17:15:49 -05001529 init_waitqueue_head(&new_ns->poll);
1530 new_ns->event = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Ram Pai390c6842005-11-07 17:17:51 -05001532 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 /* First pass: copy the tree topology */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001534 new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root,
Ram Pai9676f0c2005-11-07 17:21:20 -05001535 CL_COPY_ALL | CL_EXPIRE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 if (!new_ns->root) {
Ram Pai390c6842005-11-07 17:17:51 -05001537 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 kfree(new_ns);
Cedric Le Goater467e9f42007-07-15 23:41:06 -07001539 return ERR_PTR(-ENOMEM);;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 }
1541 spin_lock(&vfsmount_lock);
1542 list_add_tail(&new_ns->list, &new_ns->root->mnt_list);
1543 spin_unlock(&vfsmount_lock);
1544
1545 /*
1546 * Second pass: switch the tsk->fs->* elements and mark new vfsmounts
1547 * as belonging to new namespace. We have already acquired a private
1548 * fs_struct, so tsk->fs->lock is not needed.
1549 */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001550 p = mnt_ns->root;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 q = new_ns->root;
1552 while (p) {
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001553 q->mnt_ns = new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 if (fs) {
1555 if (p == fs->rootmnt) {
1556 rootmnt = p;
1557 fs->rootmnt = mntget(q);
1558 }
1559 if (p == fs->pwdmnt) {
1560 pwdmnt = p;
1561 fs->pwdmnt = mntget(q);
1562 }
1563 if (p == fs->altrootmnt) {
1564 altrootmnt = p;
1565 fs->altrootmnt = mntget(q);
1566 }
1567 }
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001568 p = next_mnt(p, mnt_ns->root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 q = next_mnt(q, new_ns->root);
1570 }
Ram Pai390c6842005-11-07 17:17:51 -05001571 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 if (rootmnt)
1574 mntput(rootmnt);
1575 if (pwdmnt)
1576 mntput(pwdmnt);
1577 if (altrootmnt)
1578 mntput(altrootmnt);
1579
JANAK DESAI741a2952006-02-07 12:59:00 -08001580 return new_ns;
1581}
1582
Eric W. Biederman213dd262007-07-15 23:41:15 -07001583struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns,
Badari Pulavartye3222c42007-05-08 00:25:21 -07001584 struct fs_struct *new_fs)
JANAK DESAI741a2952006-02-07 12:59:00 -08001585{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001586 struct mnt_namespace *new_ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08001587
Badari Pulavartye3222c42007-05-08 00:25:21 -07001588 BUG_ON(!ns);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001589 get_mnt_ns(ns);
JANAK DESAI741a2952006-02-07 12:59:00 -08001590
1591 if (!(flags & CLONE_NEWNS))
Badari Pulavartye3222c42007-05-08 00:25:21 -07001592 return ns;
JANAK DESAI741a2952006-02-07 12:59:00 -08001593
Badari Pulavartye3222c42007-05-08 00:25:21 -07001594 new_ns = dup_mnt_ns(ns, new_fs);
JANAK DESAI741a2952006-02-07 12:59:00 -08001595
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001596 put_mnt_ns(ns);
Badari Pulavartye3222c42007-05-08 00:25:21 -07001597 return new_ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598}
1599
1600asmlinkage long sys_mount(char __user * dev_name, char __user * dir_name,
1601 char __user * type, unsigned long flags,
1602 void __user * data)
1603{
1604 int retval;
1605 unsigned long data_page;
1606 unsigned long type_page;
1607 unsigned long dev_page;
1608 char *dir_page;
1609
Ram Paib58fed82005-11-07 17:16:09 -05001610 retval = copy_mount_options(type, &type_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 if (retval < 0)
1612 return retval;
1613
1614 dir_page = getname(dir_name);
1615 retval = PTR_ERR(dir_page);
1616 if (IS_ERR(dir_page))
1617 goto out1;
1618
Ram Paib58fed82005-11-07 17:16:09 -05001619 retval = copy_mount_options(dev_name, &dev_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 if (retval < 0)
1621 goto out2;
1622
Ram Paib58fed82005-11-07 17:16:09 -05001623 retval = copy_mount_options(data, &data_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 if (retval < 0)
1625 goto out3;
1626
1627 lock_kernel();
Ram Paib58fed82005-11-07 17:16:09 -05001628 retval = do_mount((char *)dev_page, dir_page, (char *)type_page,
1629 flags, (void *)data_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 unlock_kernel();
1631 free_page(data_page);
1632
1633out3:
1634 free_page(dev_page);
1635out2:
1636 putname(dir_page);
1637out1:
1638 free_page(type_page);
1639 return retval;
1640}
1641
1642/*
1643 * Replace the fs->{rootmnt,root} with {mnt,dentry}. Put the old values.
1644 * It can block. Requires the big lock held.
1645 */
1646void set_fs_root(struct fs_struct *fs, struct vfsmount *mnt,
1647 struct dentry *dentry)
1648{
1649 struct dentry *old_root;
1650 struct vfsmount *old_rootmnt;
1651 write_lock(&fs->lock);
1652 old_root = fs->root;
1653 old_rootmnt = fs->rootmnt;
1654 fs->rootmnt = mntget(mnt);
1655 fs->root = dget(dentry);
1656 write_unlock(&fs->lock);
1657 if (old_root) {
1658 dput(old_root);
1659 mntput(old_rootmnt);
1660 }
1661}
1662
1663/*
1664 * Replace the fs->{pwdmnt,pwd} with {mnt,dentry}. Put the old values.
1665 * It can block. Requires the big lock held.
1666 */
1667void set_fs_pwd(struct fs_struct *fs, struct vfsmount *mnt,
1668 struct dentry *dentry)
1669{
1670 struct dentry *old_pwd;
1671 struct vfsmount *old_pwdmnt;
1672
1673 write_lock(&fs->lock);
1674 old_pwd = fs->pwd;
1675 old_pwdmnt = fs->pwdmnt;
1676 fs->pwdmnt = mntget(mnt);
1677 fs->pwd = dget(dentry);
1678 write_unlock(&fs->lock);
1679
1680 if (old_pwd) {
1681 dput(old_pwd);
1682 mntput(old_pwdmnt);
1683 }
1684}
1685
1686static void chroot_fs_refs(struct nameidata *old_nd, struct nameidata *new_nd)
1687{
1688 struct task_struct *g, *p;
1689 struct fs_struct *fs;
1690
1691 read_lock(&tasklist_lock);
1692 do_each_thread(g, p) {
1693 task_lock(p);
1694 fs = p->fs;
1695 if (fs) {
1696 atomic_inc(&fs->count);
1697 task_unlock(p);
Ram Paib58fed82005-11-07 17:16:09 -05001698 if (fs->root == old_nd->dentry
1699 && fs->rootmnt == old_nd->mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 set_fs_root(fs, new_nd->mnt, new_nd->dentry);
Ram Paib58fed82005-11-07 17:16:09 -05001701 if (fs->pwd == old_nd->dentry
1702 && fs->pwdmnt == old_nd->mnt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 set_fs_pwd(fs, new_nd->mnt, new_nd->dentry);
1704 put_fs_struct(fs);
1705 } else
1706 task_unlock(p);
1707 } while_each_thread(g, p);
1708 read_unlock(&tasklist_lock);
1709}
1710
1711/*
1712 * pivot_root Semantics:
1713 * Moves the root file system of the current process to the directory put_old,
1714 * makes new_root as the new root file system of the current process, and sets
1715 * root/cwd of all processes which had them on the current root to new_root.
1716 *
1717 * Restrictions:
1718 * The new_root and put_old must be directories, and must not be on the
1719 * same file system as the current process root. The put_old must be
1720 * underneath new_root, i.e. adding a non-zero number of /.. to the string
1721 * pointed to by put_old must yield the same directory as new_root. No other
1722 * file system may be mounted on put_old. After all, new_root is a mountpoint.
1723 *
Neil Brown4a0d11f2006-01-08 01:03:18 -08001724 * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem.
1725 * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives
1726 * in this situation.
1727 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 * Notes:
1729 * - we don't move root/cwd if they are not at the root (reason: if something
1730 * cared enough to change them, it's probably wrong to force them elsewhere)
1731 * - it's okay to pick a root that isn't the root of a file system, e.g.
1732 * /nfs/my_root where /nfs is the mount point. It must be a mountpoint,
1733 * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root
1734 * first.
1735 */
Ram Paib58fed82005-11-07 17:16:09 -05001736asmlinkage long sys_pivot_root(const char __user * new_root,
1737 const char __user * put_old)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738{
1739 struct vfsmount *tmp;
1740 struct nameidata new_nd, old_nd, parent_nd, root_parent, user_nd;
1741 int error;
1742
1743 if (!capable(CAP_SYS_ADMIN))
1744 return -EPERM;
1745
1746 lock_kernel();
1747
Ram Paib58fed82005-11-07 17:16:09 -05001748 error = __user_walk(new_root, LOOKUP_FOLLOW | LOOKUP_DIRECTORY,
1749 &new_nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 if (error)
1751 goto out0;
1752 error = -EINVAL;
1753 if (!check_mnt(new_nd.mnt))
1754 goto out1;
1755
Ram Paib58fed82005-11-07 17:16:09 -05001756 error = __user_walk(put_old, LOOKUP_FOLLOW | LOOKUP_DIRECTORY, &old_nd);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 if (error)
1758 goto out1;
1759
1760 error = security_sb_pivotroot(&old_nd, &new_nd);
1761 if (error) {
1762 path_release(&old_nd);
1763 goto out1;
1764 }
1765
1766 read_lock(&current->fs->lock);
1767 user_nd.mnt = mntget(current->fs->rootmnt);
1768 user_nd.dentry = dget(current->fs->root);
1769 read_unlock(&current->fs->lock);
Ram Pai390c6842005-11-07 17:17:51 -05001770 down_write(&namespace_sem);
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001771 mutex_lock(&old_nd.dentry->d_inode->i_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 error = -EINVAL;
Ram Pai21444402005-11-07 17:20:03 -05001773 if (IS_MNT_SHARED(old_nd.mnt) ||
1774 IS_MNT_SHARED(new_nd.mnt->mnt_parent) ||
1775 IS_MNT_SHARED(user_nd.mnt->mnt_parent))
1776 goto out2;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 if (!check_mnt(user_nd.mnt))
1778 goto out2;
1779 error = -ENOENT;
1780 if (IS_DEADDIR(new_nd.dentry->d_inode))
1781 goto out2;
1782 if (d_unhashed(new_nd.dentry) && !IS_ROOT(new_nd.dentry))
1783 goto out2;
1784 if (d_unhashed(old_nd.dentry) && !IS_ROOT(old_nd.dentry))
1785 goto out2;
1786 error = -EBUSY;
1787 if (new_nd.mnt == user_nd.mnt || old_nd.mnt == user_nd.mnt)
1788 goto out2; /* loop, on the same file system */
1789 error = -EINVAL;
1790 if (user_nd.mnt->mnt_root != user_nd.dentry)
1791 goto out2; /* not a mountpoint */
Miklos Szeredi0bb6fcc2005-09-06 15:19:36 -07001792 if (user_nd.mnt->mnt_parent == user_nd.mnt)
1793 goto out2; /* not attached */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 if (new_nd.mnt->mnt_root != new_nd.dentry)
1795 goto out2; /* not a mountpoint */
Miklos Szeredi0bb6fcc2005-09-06 15:19:36 -07001796 if (new_nd.mnt->mnt_parent == new_nd.mnt)
1797 goto out2; /* not attached */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 tmp = old_nd.mnt; /* make sure we can reach put_old from new_root */
1799 spin_lock(&vfsmount_lock);
1800 if (tmp != new_nd.mnt) {
1801 for (;;) {
1802 if (tmp->mnt_parent == tmp)
1803 goto out3; /* already mounted on put_old */
1804 if (tmp->mnt_parent == new_nd.mnt)
1805 break;
1806 tmp = tmp->mnt_parent;
1807 }
1808 if (!is_subdir(tmp->mnt_mountpoint, new_nd.dentry))
1809 goto out3;
1810 } else if (!is_subdir(old_nd.dentry, new_nd.dentry))
1811 goto out3;
1812 detach_mnt(new_nd.mnt, &parent_nd);
1813 detach_mnt(user_nd.mnt, &root_parent);
1814 attach_mnt(user_nd.mnt, &old_nd); /* mount old root on put_old */
1815 attach_mnt(new_nd.mnt, &root_parent); /* mount new_root on / */
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001816 touch_mnt_namespace(current->nsproxy->mnt_ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 spin_unlock(&vfsmount_lock);
1818 chroot_fs_refs(&user_nd, &new_nd);
1819 security_sb_post_pivotroot(&user_nd, &new_nd);
1820 error = 0;
1821 path_release(&root_parent);
1822 path_release(&parent_nd);
1823out2:
Jes Sorensen1b1dcc12006-01-09 15:59:24 -08001824 mutex_unlock(&old_nd.dentry->d_inode->i_mutex);
Ram Pai390c6842005-11-07 17:17:51 -05001825 up_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 path_release(&user_nd);
1827 path_release(&old_nd);
1828out1:
1829 path_release(&new_nd);
1830out0:
1831 unlock_kernel();
1832 return error;
1833out3:
1834 spin_unlock(&vfsmount_lock);
1835 goto out2;
1836}
1837
1838static void __init init_mount_tree(void)
1839{
1840 struct vfsmount *mnt;
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001841 struct mnt_namespace *ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842
1843 mnt = do_kern_mount("rootfs", 0, "rootfs", NULL);
1844 if (IS_ERR(mnt))
1845 panic("Can't create rootfs");
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001846 ns = kmalloc(sizeof(*ns), GFP_KERNEL);
1847 if (!ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 panic("Can't allocate initial namespace");
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001849 atomic_set(&ns->count, 1);
1850 INIT_LIST_HEAD(&ns->list);
1851 init_waitqueue_head(&ns->poll);
1852 ns->event = 0;
1853 list_add(&mnt->mnt_list, &ns->list);
1854 ns->root = mnt;
1855 mnt->mnt_ns = ns;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001857 init_task.nsproxy->mnt_ns = ns;
1858 get_mnt_ns(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001860 set_fs_pwd(current->fs, ns->root, ns->root->mnt_root);
1861 set_fs_root(current->fs, ns->root, ns->root->mnt_root);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862}
1863
Denis Cheng74bf17c2007-10-16 23:26:30 -07001864void __init mnt_init(void)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865{
Eric Dumazet13f14b42008-02-06 01:37:57 -08001866 unsigned u;
Randy Dunlap15a67dd2006-09-29 01:58:57 -07001867 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868
Ram Pai390c6842005-11-07 17:17:51 -05001869 init_rwsem(&namespace_sem);
1870
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount),
Paul Mundt20c2df82007-07-20 10:11:58 +09001872 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873
Ram Paib58fed82005-11-07 17:16:09 -05001874 mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875
1876 if (!mount_hashtable)
1877 panic("Failed to allocate mount hash table\n");
1878
Eric Dumazet13f14b42008-02-06 01:37:57 -08001879 printk("Mount-cache hash table entries: %lu\n", HASH_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
Eric Dumazet13f14b42008-02-06 01:37:57 -08001881 for (u = 0; u < HASH_SIZE; u++)
1882 INIT_LIST_HEAD(&mount_hashtable[u]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883
Randy Dunlap15a67dd2006-09-29 01:58:57 -07001884 err = sysfs_init();
1885 if (err)
1886 printk(KERN_WARNING "%s: sysfs_init error: %d\n",
1887 __FUNCTION__, err);
Greg Kroah-Hartman00d26662007-10-29 14:17:23 -06001888 fs_kobj = kobject_create_and_add("fs", NULL);
1889 if (!fs_kobj)
1890 printk(KERN_WARNING "%s: kobj create error\n", __FUNCTION__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891 init_rootfs();
1892 init_mount_tree();
1893}
1894
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001895void __put_mnt_ns(struct mnt_namespace *ns)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896{
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001897 struct vfsmount *root = ns->root;
Ram Pai70fbcdf2005-11-07 17:17:04 -05001898 LIST_HEAD(umount_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001899 ns->root = NULL;
Miklos Szeredi1ce88cf2005-07-07 17:57:24 -07001900 spin_unlock(&vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05001901 down_write(&namespace_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902 spin_lock(&vfsmount_lock);
Ram Paia05964f2005-11-07 17:20:17 -05001903 umount_tree(root, 0, &umount_list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 spin_unlock(&vfsmount_lock);
Ram Pai390c6842005-11-07 17:17:51 -05001905 up_write(&namespace_sem);
Ram Pai70fbcdf2005-11-07 17:17:04 -05001906 release_mounts(&umount_list);
Kirill Korotaev6b3286e2006-12-08 02:37:56 -08001907 kfree(ns);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908}