]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
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 | ||
1da177e4 LT |
11 | #include <linux/syscalls.h> |
12 | #include <linux/slab.h> | |
13 | #include <linux/sched.h> | |
99b7db7b NP |
14 | #include <linux/spinlock.h> |
15 | #include <linux/percpu.h> | |
1da177e4 | 16 | #include <linux/init.h> |
15a67dd8 | 17 | #include <linux/kernel.h> |
1da177e4 | 18 | #include <linux/acct.h> |
16f7e0fe | 19 | #include <linux/capability.h> |
3d733633 | 20 | #include <linux/cpumask.h> |
1da177e4 | 21 | #include <linux/module.h> |
f20a9ead | 22 | #include <linux/sysfs.h> |
1da177e4 | 23 | #include <linux/seq_file.h> |
6b3286ed | 24 | #include <linux/mnt_namespace.h> |
1da177e4 | 25 | #include <linux/namei.h> |
b43f3cbd | 26 | #include <linux/nsproxy.h> |
1da177e4 LT |
27 | #include <linux/security.h> |
28 | #include <linux/mount.h> | |
07f3f05c | 29 | #include <linux/ramfs.h> |
13f14b4d | 30 | #include <linux/log2.h> |
73cd49ec | 31 | #include <linux/idr.h> |
5ad4e53b | 32 | #include <linux/fs_struct.h> |
2504c5d6 | 33 | #include <linux/fsnotify.h> |
1da177e4 LT |
34 | #include <asm/uaccess.h> |
35 | #include <asm/unistd.h> | |
07b20889 | 36 | #include "pnode.h" |
948730b0 | 37 | #include "internal.h" |
1da177e4 | 38 | |
13f14b4d ED |
39 | #define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head)) |
40 | #define HASH_SIZE (1UL << HASH_SHIFT) | |
41 | ||
5addc5dd | 42 | static int event; |
73cd49ec | 43 | static DEFINE_IDA(mnt_id_ida); |
719f5d7f | 44 | static DEFINE_IDA(mnt_group_ida); |
99b7db7b | 45 | static DEFINE_SPINLOCK(mnt_id_lock); |
f21f6220 AV |
46 | static int mnt_id_start = 0; |
47 | static int mnt_group_start = 1; | |
1da177e4 | 48 | |
fa3536cc | 49 | static struct list_head *mount_hashtable __read_mostly; |
e18b890b | 50 | static struct kmem_cache *mnt_cache __read_mostly; |
390c6843 | 51 | static struct rw_semaphore namespace_sem; |
1da177e4 | 52 | |
f87fd4c2 | 53 | /* /sys/fs */ |
00d26666 GKH |
54 | struct kobject *fs_kobj; |
55 | EXPORT_SYMBOL_GPL(fs_kobj); | |
f87fd4c2 | 56 | |
99b7db7b NP |
57 | /* |
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 | */ | |
65 | DEFINE_BRLOCK(vfsmount_lock); | |
66 | ||
1da177e4 LT |
67 | static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry) |
68 | { | |
b58fed8b RP |
69 | unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES); |
70 | tmp += ((unsigned long)dentry / L1_CACHE_BYTES); | |
13f14b4d ED |
71 | tmp = tmp + (tmp >> HASH_SHIFT); |
72 | return tmp & (HASH_SIZE - 1); | |
1da177e4 LT |
73 | } |
74 | ||
3d733633 DH |
75 | #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16) |
76 | ||
99b7db7b NP |
77 | /* |
78 | * allocation is serialized by namespace_sem, but we need the spinlock to | |
79 | * serialize with freeing. | |
80 | */ | |
73cd49ec MS |
81 | static int mnt_alloc_id(struct vfsmount *mnt) |
82 | { | |
83 | int res; | |
84 | ||
85 | retry: | |
86 | ida_pre_get(&mnt_id_ida, GFP_KERNEL); | |
99b7db7b | 87 | spin_lock(&mnt_id_lock); |
f21f6220 AV |
88 | 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; | |
99b7db7b | 91 | spin_unlock(&mnt_id_lock); |
73cd49ec MS |
92 | if (res == -EAGAIN) |
93 | goto retry; | |
94 | ||
95 | return res; | |
96 | } | |
97 | ||
98 | static void mnt_free_id(struct vfsmount *mnt) | |
99 | { | |
f21f6220 | 100 | int id = mnt->mnt_id; |
99b7db7b | 101 | spin_lock(&mnt_id_lock); |
f21f6220 AV |
102 | ida_remove(&mnt_id_ida, id); |
103 | if (mnt_id_start > id) | |
104 | mnt_id_start = id; | |
99b7db7b | 105 | spin_unlock(&mnt_id_lock); |
73cd49ec MS |
106 | } |
107 | ||
719f5d7f MS |
108 | /* |
109 | * Allocate a new peer group ID | |
110 | * | |
111 | * mnt_group_ida is protected by namespace_sem | |
112 | */ | |
113 | static int mnt_alloc_group_id(struct vfsmount *mnt) | |
114 | { | |
f21f6220 AV |
115 | int res; |
116 | ||
719f5d7f MS |
117 | if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL)) |
118 | return -ENOMEM; | |
119 | ||
f21f6220 AV |
120 | 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; | |
719f5d7f MS |
127 | } |
128 | ||
129 | /* | |
130 | * Release a peer group ID | |
131 | */ | |
132 | void mnt_release_group_id(struct vfsmount *mnt) | |
133 | { | |
f21f6220 AV |
134 | int id = mnt->mnt_group_id; |
135 | ida_remove(&mnt_group_ida, id); | |
136 | if (mnt_group_start > id) | |
137 | mnt_group_start = id; | |
719f5d7f MS |
138 | mnt->mnt_group_id = 0; |
139 | } | |
140 | ||
b3e19d92 NP |
141 | /* |
142 | * vfsmount lock must be held for read | |
143 | */ | |
144 | static 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 | ||
b3e19d92 NP |
155 | /* |
156 | * vfsmount lock must be held for write | |
157 | */ | |
158 | unsigned int mnt_get_count(struct vfsmount *mnt) | |
159 | { | |
160 | #ifdef CONFIG_SMP | |
f03c6599 | 161 | unsigned int count = 0; |
b3e19d92 NP |
162 | int cpu; |
163 | ||
164 | for_each_possible_cpu(cpu) { | |
165 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count; | |
166 | } | |
167 | ||
168 | return count; | |
169 | #else | |
170 | return mnt->mnt_count; | |
171 | #endif | |
172 | } | |
173 | ||
9d412a43 | 174 | static struct vfsmount *alloc_vfsmnt(const char *name) |
1da177e4 | 175 | { |
7d6fec45 AV |
176 | struct mount *p = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); |
177 | if (p) { | |
178 | struct vfsmount *mnt = &p->mnt; | |
73cd49ec MS |
179 | int err; |
180 | ||
181 | err = mnt_alloc_id(mnt); | |
88b38782 LZ |
182 | if (err) |
183 | goto out_free_cache; | |
184 | ||
185 | if (name) { | |
186 | mnt->mnt_devname = kstrdup(name, GFP_KERNEL); | |
187 | if (!mnt->mnt_devname) | |
188 | goto out_free_id; | |
73cd49ec MS |
189 | } |
190 | ||
b3e19d92 NP |
191 | #ifdef CONFIG_SMP |
192 | mnt->mnt_pcp = alloc_percpu(struct mnt_pcp); | |
193 | if (!mnt->mnt_pcp) | |
194 | goto out_free_devname; | |
195 | ||
f03c6599 | 196 | this_cpu_add(mnt->mnt_pcp->mnt_count, 1); |
b3e19d92 NP |
197 | #else |
198 | mnt->mnt_count = 1; | |
199 | mnt->mnt_writers = 0; | |
200 | #endif | |
201 | ||
1da177e4 LT |
202 | INIT_LIST_HEAD(&mnt->mnt_hash); |
203 | INIT_LIST_HEAD(&mnt->mnt_child); | |
204 | INIT_LIST_HEAD(&mnt->mnt_mounts); | |
205 | INIT_LIST_HEAD(&mnt->mnt_list); | |
55e700b9 | 206 | INIT_LIST_HEAD(&mnt->mnt_expire); |
03e06e68 | 207 | INIT_LIST_HEAD(&mnt->mnt_share); |
a58b0eb8 RP |
208 | INIT_LIST_HEAD(&mnt->mnt_slave_list); |
209 | INIT_LIST_HEAD(&mnt->mnt_slave); | |
2504c5d6 AG |
210 | #ifdef CONFIG_FSNOTIFY |
211 | INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks); | |
d3ef3d73 | 212 | #endif |
1da177e4 | 213 | } |
7d6fec45 | 214 | return &p->mnt; |
88b38782 | 215 | |
d3ef3d73 | 216 | #ifdef CONFIG_SMP |
217 | out_free_devname: | |
7d6fec45 | 218 | kfree(p->mnt.mnt_devname); |
d3ef3d73 | 219 | #endif |
88b38782 | 220 | out_free_id: |
7d6fec45 | 221 | mnt_free_id(&p->mnt); |
88b38782 | 222 | out_free_cache: |
7d6fec45 | 223 | kmem_cache_free(mnt_cache, p); |
88b38782 | 224 | return NULL; |
1da177e4 LT |
225 | } |
226 | ||
3d733633 DH |
227 | /* |
228 | * Most r/o checks on a fs are for operations that take | |
229 | * discrete amounts of time, like a write() or unlink(). | |
230 | * We must keep track of when those operations start | |
231 | * (for permission checks) and when they end, so that | |
232 | * we can determine when writes are able to occur to | |
233 | * a filesystem. | |
234 | */ | |
235 | /* | |
236 | * __mnt_is_readonly: check whether a mount is read-only | |
237 | * @mnt: the mount to check for its write status | |
238 | * | |
239 | * This shouldn't be used directly ouside of the VFS. | |
240 | * It does not guarantee that the filesystem will stay | |
241 | * r/w, just that it is right *now*. This can not and | |
242 | * should not be used in place of IS_RDONLY(inode). | |
243 | * mnt_want/drop_write() will _keep_ the filesystem | |
244 | * r/w. | |
245 | */ | |
246 | int __mnt_is_readonly(struct vfsmount *mnt) | |
247 | { | |
2e4b7fcd DH |
248 | if (mnt->mnt_flags & MNT_READONLY) |
249 | return 1; | |
250 | if (mnt->mnt_sb->s_flags & MS_RDONLY) | |
251 | return 1; | |
252 | return 0; | |
3d733633 DH |
253 | } |
254 | EXPORT_SYMBOL_GPL(__mnt_is_readonly); | |
255 | ||
c6653a83 | 256 | static inline void mnt_inc_writers(struct vfsmount *mnt) |
d3ef3d73 | 257 | { |
258 | #ifdef CONFIG_SMP | |
b3e19d92 | 259 | this_cpu_inc(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 | 260 | #else |
261 | mnt->mnt_writers++; | |
262 | #endif | |
263 | } | |
3d733633 | 264 | |
c6653a83 | 265 | static inline void mnt_dec_writers(struct vfsmount *mnt) |
3d733633 | 266 | { |
d3ef3d73 | 267 | #ifdef CONFIG_SMP |
b3e19d92 | 268 | this_cpu_dec(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 | 269 | #else |
270 | mnt->mnt_writers--; | |
271 | #endif | |
3d733633 | 272 | } |
3d733633 | 273 | |
c6653a83 | 274 | static unsigned int mnt_get_writers(struct vfsmount *mnt) |
3d733633 | 275 | { |
d3ef3d73 | 276 | #ifdef CONFIG_SMP |
277 | unsigned int count = 0; | |
3d733633 | 278 | int cpu; |
3d733633 DH |
279 | |
280 | for_each_possible_cpu(cpu) { | |
b3e19d92 | 281 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers; |
3d733633 | 282 | } |
3d733633 | 283 | |
d3ef3d73 | 284 | return count; |
285 | #else | |
286 | return mnt->mnt_writers; | |
287 | #endif | |
3d733633 DH |
288 | } |
289 | ||
8366025e DH |
290 | /* |
291 | * Most r/o checks on a fs are for operations that take | |
292 | * discrete amounts of time, like a write() or unlink(). | |
293 | * We must keep track of when those operations start | |
294 | * (for permission checks) and when they end, so that | |
295 | * we can determine when writes are able to occur to | |
296 | * a filesystem. | |
297 | */ | |
298 | /** | |
299 | * mnt_want_write - get write access to a mount | |
300 | * @mnt: the mount on which to take a write | |
301 | * | |
302 | * This tells the low-level filesystem that a write is | |
303 | * about to be performed to it, and makes sure that | |
304 | * writes are allowed before returning success. When | |
305 | * the write operation is finished, mnt_drop_write() | |
306 | * must be called. This is effectively a refcount. | |
307 | */ | |
308 | int mnt_want_write(struct vfsmount *mnt) | |
309 | { | |
3d733633 | 310 | int ret = 0; |
3d733633 | 311 | |
d3ef3d73 | 312 | preempt_disable(); |
c6653a83 | 313 | mnt_inc_writers(mnt); |
d3ef3d73 | 314 | /* |
c6653a83 | 315 | * The store to mnt_inc_writers must be visible before we pass |
d3ef3d73 | 316 | * MNT_WRITE_HOLD loop below, so that the slowpath can see our |
317 | * incremented count after it has set MNT_WRITE_HOLD. | |
318 | */ | |
319 | smp_mb(); | |
320 | while (mnt->mnt_flags & MNT_WRITE_HOLD) | |
321 | cpu_relax(); | |
322 | /* | |
323 | * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will | |
324 | * be set to match its requirements. So we must not load that until | |
325 | * MNT_WRITE_HOLD is cleared. | |
326 | */ | |
327 | smp_rmb(); | |
3d733633 | 328 | if (__mnt_is_readonly(mnt)) { |
c6653a83 | 329 | mnt_dec_writers(mnt); |
3d733633 DH |
330 | ret = -EROFS; |
331 | goto out; | |
332 | } | |
3d733633 | 333 | out: |
d3ef3d73 | 334 | preempt_enable(); |
3d733633 | 335 | return ret; |
8366025e DH |
336 | } |
337 | EXPORT_SYMBOL_GPL(mnt_want_write); | |
338 | ||
96029c4e | 339 | /** |
340 | * mnt_clone_write - get write access to a mount | |
341 | * @mnt: the mount on which to take a write | |
342 | * | |
343 | * This is effectively like mnt_want_write, except | |
344 | * it must only be used to take an extra write reference | |
345 | * on a mountpoint that we already know has a write reference | |
346 | * on it. This allows some optimisation. | |
347 | * | |
348 | * After finished, mnt_drop_write must be called as usual to | |
349 | * drop the reference. | |
350 | */ | |
351 | int mnt_clone_write(struct vfsmount *mnt) | |
352 | { | |
353 | /* superblock may be r/o */ | |
354 | if (__mnt_is_readonly(mnt)) | |
355 | return -EROFS; | |
356 | preempt_disable(); | |
c6653a83 | 357 | mnt_inc_writers(mnt); |
96029c4e | 358 | preempt_enable(); |
359 | return 0; | |
360 | } | |
361 | EXPORT_SYMBOL_GPL(mnt_clone_write); | |
362 | ||
363 | /** | |
364 | * mnt_want_write_file - get write access to a file's mount | |
365 | * @file: the file who's mount on which to take a write | |
366 | * | |
367 | * This is like mnt_want_write, but it takes a file and can | |
368 | * do some optimisations if the file is open for write already | |
369 | */ | |
370 | int mnt_want_write_file(struct file *file) | |
371 | { | |
2d8dd38a OH |
372 | struct inode *inode = file->f_dentry->d_inode; |
373 | if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) | |
96029c4e | 374 | return mnt_want_write(file->f_path.mnt); |
375 | else | |
376 | return mnt_clone_write(file->f_path.mnt); | |
377 | } | |
378 | EXPORT_SYMBOL_GPL(mnt_want_write_file); | |
379 | ||
8366025e DH |
380 | /** |
381 | * mnt_drop_write - give up write access to a mount | |
382 | * @mnt: the mount on which to give up write access | |
383 | * | |
384 | * Tells the low-level filesystem that we are done | |
385 | * performing writes to it. Must be matched with | |
386 | * mnt_want_write() call above. | |
387 | */ | |
388 | void mnt_drop_write(struct vfsmount *mnt) | |
389 | { | |
d3ef3d73 | 390 | preempt_disable(); |
c6653a83 | 391 | mnt_dec_writers(mnt); |
d3ef3d73 | 392 | preempt_enable(); |
8366025e DH |
393 | } |
394 | EXPORT_SYMBOL_GPL(mnt_drop_write); | |
395 | ||
2a79f17e AV |
396 | void mnt_drop_write_file(struct file *file) |
397 | { | |
398 | mnt_drop_write(file->f_path.mnt); | |
399 | } | |
400 | EXPORT_SYMBOL(mnt_drop_write_file); | |
401 | ||
2e4b7fcd | 402 | static int mnt_make_readonly(struct vfsmount *mnt) |
8366025e | 403 | { |
3d733633 DH |
404 | int ret = 0; |
405 | ||
99b7db7b | 406 | br_write_lock(vfsmount_lock); |
d3ef3d73 | 407 | mnt->mnt_flags |= MNT_WRITE_HOLD; |
3d733633 | 408 | /* |
d3ef3d73 | 409 | * After storing MNT_WRITE_HOLD, we'll read the counters. This store |
410 | * should be visible before we do. | |
3d733633 | 411 | */ |
d3ef3d73 | 412 | smp_mb(); |
413 | ||
3d733633 | 414 | /* |
d3ef3d73 | 415 | * With writers on hold, if this value is zero, then there are |
416 | * definitely no active writers (although held writers may subsequently | |
417 | * increment the count, they'll have to wait, and decrement it after | |
418 | * seeing MNT_READONLY). | |
419 | * | |
420 | * It is OK to have counter incremented on one CPU and decremented on | |
421 | * another: the sum will add up correctly. The danger would be when we | |
422 | * sum up each counter, if we read a counter before it is incremented, | |
423 | * but then read another CPU's count which it has been subsequently | |
424 | * decremented from -- we would see more decrements than we should. | |
425 | * MNT_WRITE_HOLD protects against this scenario, because | |
426 | * mnt_want_write first increments count, then smp_mb, then spins on | |
427 | * MNT_WRITE_HOLD, so it can't be decremented by another CPU while | |
428 | * we're counting up here. | |
3d733633 | 429 | */ |
c6653a83 | 430 | if (mnt_get_writers(mnt) > 0) |
d3ef3d73 | 431 | ret = -EBUSY; |
432 | else | |
2e4b7fcd | 433 | mnt->mnt_flags |= MNT_READONLY; |
d3ef3d73 | 434 | /* |
435 | * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers | |
436 | * that become unheld will see MNT_READONLY. | |
437 | */ | |
438 | smp_wmb(); | |
439 | mnt->mnt_flags &= ~MNT_WRITE_HOLD; | |
99b7db7b | 440 | br_write_unlock(vfsmount_lock); |
3d733633 | 441 | return ret; |
8366025e | 442 | } |
8366025e | 443 | |
2e4b7fcd DH |
444 | static void __mnt_unmake_readonly(struct vfsmount *mnt) |
445 | { | |
99b7db7b | 446 | br_write_lock(vfsmount_lock); |
2e4b7fcd | 447 | mnt->mnt_flags &= ~MNT_READONLY; |
99b7db7b | 448 | br_write_unlock(vfsmount_lock); |
2e4b7fcd DH |
449 | } |
450 | ||
9d412a43 | 451 | static void free_vfsmnt(struct vfsmount *mnt) |
1da177e4 | 452 | { |
7d6fec45 | 453 | struct mount *p = real_mount(mnt); |
1da177e4 | 454 | kfree(mnt->mnt_devname); |
73cd49ec | 455 | mnt_free_id(mnt); |
d3ef3d73 | 456 | #ifdef CONFIG_SMP |
b3e19d92 | 457 | free_percpu(mnt->mnt_pcp); |
d3ef3d73 | 458 | #endif |
7d6fec45 | 459 | kmem_cache_free(mnt_cache, p); |
1da177e4 LT |
460 | } |
461 | ||
462 | /* | |
a05964f3 RP |
463 | * find the first or last mount at @dentry on vfsmount @mnt depending on |
464 | * @dir. If @dir is set return the first mount else return the last mount. | |
99b7db7b | 465 | * vfsmount_lock must be held for read or write. |
1da177e4 | 466 | */ |
c7105365 | 467 | struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, |
a05964f3 | 468 | int dir) |
1da177e4 | 469 | { |
b58fed8b RP |
470 | struct list_head *head = mount_hashtable + hash(mnt, dentry); |
471 | struct list_head *tmp = head; | |
c7105365 | 472 | struct mount *p, *found = NULL; |
1da177e4 | 473 | |
1da177e4 | 474 | for (;;) { |
a05964f3 | 475 | tmp = dir ? tmp->next : tmp->prev; |
1da177e4 LT |
476 | p = NULL; |
477 | if (tmp == head) | |
478 | break; | |
c7105365 AV |
479 | p = list_entry(tmp, struct mount, mnt.mnt_hash); |
480 | if (p->mnt.mnt_parent == mnt && p->mnt.mnt_mountpoint == dentry) { | |
a05964f3 | 481 | found = p; |
1da177e4 LT |
482 | break; |
483 | } | |
484 | } | |
1da177e4 LT |
485 | return found; |
486 | } | |
487 | ||
a05964f3 RP |
488 | /* |
489 | * lookup_mnt increments the ref count before returning | |
490 | * the vfsmount struct. | |
491 | */ | |
1c755af4 | 492 | struct vfsmount *lookup_mnt(struct path *path) |
a05964f3 | 493 | { |
c7105365 | 494 | struct mount *child_mnt; |
99b7db7b NP |
495 | |
496 | br_read_lock(vfsmount_lock); | |
c7105365 AV |
497 | child_mnt = __lookup_mnt(path->mnt, path->dentry, 1); |
498 | if (child_mnt) { | |
499 | mnt_add_count(child_mnt, 1); | |
500 | br_read_unlock(vfsmount_lock); | |
501 | return &child_mnt->mnt; | |
502 | } else { | |
503 | br_read_unlock(vfsmount_lock); | |
504 | return NULL; | |
505 | } | |
a05964f3 RP |
506 | } |
507 | ||
1da177e4 LT |
508 | static inline int check_mnt(struct vfsmount *mnt) |
509 | { | |
6b3286ed | 510 | return mnt->mnt_ns == current->nsproxy->mnt_ns; |
1da177e4 LT |
511 | } |
512 | ||
99b7db7b NP |
513 | /* |
514 | * vfsmount lock must be held for write | |
515 | */ | |
6b3286ed | 516 | static void touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
517 | { |
518 | if (ns) { | |
519 | ns->event = ++event; | |
520 | wake_up_interruptible(&ns->poll); | |
521 | } | |
522 | } | |
523 | ||
99b7db7b NP |
524 | /* |
525 | * vfsmount lock must be held for write | |
526 | */ | |
6b3286ed | 527 | static void __touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
528 | { |
529 | if (ns && ns->event != event) { | |
530 | ns->event = event; | |
531 | wake_up_interruptible(&ns->poll); | |
532 | } | |
533 | } | |
534 | ||
5f57cbcc NP |
535 | /* |
536 | * Clear dentry's mounted state if it has no remaining mounts. | |
537 | * vfsmount_lock must be held for write. | |
538 | */ | |
aa0a4cf0 | 539 | static void dentry_reset_mounted(struct dentry *dentry) |
5f57cbcc NP |
540 | { |
541 | unsigned u; | |
542 | ||
543 | for (u = 0; u < HASH_SIZE; u++) { | |
544 | struct vfsmount *p; | |
545 | ||
546 | list_for_each_entry(p, &mount_hashtable[u], mnt_hash) { | |
547 | if (p->mnt_mountpoint == dentry) | |
548 | return; | |
549 | } | |
550 | } | |
551 | spin_lock(&dentry->d_lock); | |
552 | dentry->d_flags &= ~DCACHE_MOUNTED; | |
553 | spin_unlock(&dentry->d_lock); | |
554 | } | |
555 | ||
99b7db7b NP |
556 | /* |
557 | * vfsmount lock must be held for write | |
558 | */ | |
1a390689 | 559 | static void detach_mnt(struct vfsmount *mnt, struct path *old_path) |
1da177e4 | 560 | { |
1a390689 AV |
561 | old_path->dentry = mnt->mnt_mountpoint; |
562 | old_path->mnt = mnt->mnt_parent; | |
1da177e4 LT |
563 | mnt->mnt_parent = mnt; |
564 | mnt->mnt_mountpoint = mnt->mnt_root; | |
565 | list_del_init(&mnt->mnt_child); | |
566 | list_del_init(&mnt->mnt_hash); | |
aa0a4cf0 | 567 | dentry_reset_mounted(old_path->dentry); |
1da177e4 LT |
568 | } |
569 | ||
99b7db7b NP |
570 | /* |
571 | * vfsmount lock must be held for write | |
572 | */ | |
b90fa9ae RP |
573 | void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, |
574 | struct vfsmount *child_mnt) | |
575 | { | |
576 | child_mnt->mnt_parent = mntget(mnt); | |
577 | child_mnt->mnt_mountpoint = dget(dentry); | |
5f57cbcc NP |
578 | spin_lock(&dentry->d_lock); |
579 | dentry->d_flags |= DCACHE_MOUNTED; | |
580 | spin_unlock(&dentry->d_lock); | |
b90fa9ae RP |
581 | } |
582 | ||
99b7db7b NP |
583 | /* |
584 | * vfsmount lock must be held for write | |
585 | */ | |
1a390689 | 586 | static void attach_mnt(struct vfsmount *mnt, struct path *path) |
1da177e4 | 587 | { |
1a390689 | 588 | mnt_set_mountpoint(path->mnt, path->dentry, mnt); |
b90fa9ae | 589 | list_add_tail(&mnt->mnt_hash, mount_hashtable + |
1a390689 AV |
590 | hash(path->mnt, path->dentry)); |
591 | list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); | |
b90fa9ae RP |
592 | } |
593 | ||
7e3d0eb0 AV |
594 | static inline void __mnt_make_longterm(struct vfsmount *mnt) |
595 | { | |
596 | #ifdef CONFIG_SMP | |
597 | atomic_inc(&mnt->mnt_longterm); | |
598 | #endif | |
599 | } | |
600 | ||
601 | /* needs vfsmount lock for write */ | |
602 | static inline void __mnt_make_shortterm(struct vfsmount *mnt) | |
603 | { | |
604 | #ifdef CONFIG_SMP | |
605 | atomic_dec(&mnt->mnt_longterm); | |
606 | #endif | |
607 | } | |
608 | ||
b90fa9ae | 609 | /* |
99b7db7b | 610 | * vfsmount lock must be held for write |
b90fa9ae RP |
611 | */ |
612 | static void commit_tree(struct vfsmount *mnt) | |
613 | { | |
614 | struct vfsmount *parent = mnt->mnt_parent; | |
615 | struct vfsmount *m; | |
616 | LIST_HEAD(head); | |
6b3286ed | 617 | struct mnt_namespace *n = parent->mnt_ns; |
b90fa9ae RP |
618 | |
619 | BUG_ON(parent == mnt); | |
620 | ||
621 | list_add_tail(&head, &mnt->mnt_list); | |
f03c6599 | 622 | list_for_each_entry(m, &head, mnt_list) { |
6b3286ed | 623 | m->mnt_ns = n; |
7e3d0eb0 | 624 | __mnt_make_longterm(m); |
f03c6599 AV |
625 | } |
626 | ||
b90fa9ae RP |
627 | list_splice(&head, n->list.prev); |
628 | ||
629 | list_add_tail(&mnt->mnt_hash, mount_hashtable + | |
630 | hash(parent, mnt->mnt_mountpoint)); | |
631 | list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); | |
6b3286ed | 632 | touch_mnt_namespace(n); |
1da177e4 LT |
633 | } |
634 | ||
635 | static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) | |
636 | { | |
637 | struct list_head *next = p->mnt_mounts.next; | |
638 | if (next == &p->mnt_mounts) { | |
639 | while (1) { | |
640 | if (p == root) | |
641 | return NULL; | |
642 | next = p->mnt_child.next; | |
643 | if (next != &p->mnt_parent->mnt_mounts) | |
644 | break; | |
645 | p = p->mnt_parent; | |
646 | } | |
647 | } | |
648 | return list_entry(next, struct vfsmount, mnt_child); | |
649 | } | |
650 | ||
9676f0c6 RP |
651 | static struct vfsmount *skip_mnt_tree(struct vfsmount *p) |
652 | { | |
653 | struct list_head *prev = p->mnt_mounts.prev; | |
654 | while (prev != &p->mnt_mounts) { | |
655 | p = list_entry(prev, struct vfsmount, mnt_child); | |
656 | prev = p->mnt_mounts.prev; | |
657 | } | |
658 | return p; | |
659 | } | |
660 | ||
9d412a43 AV |
661 | struct vfsmount * |
662 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) | |
663 | { | |
664 | struct vfsmount *mnt; | |
665 | struct dentry *root; | |
666 | ||
667 | if (!type) | |
668 | return ERR_PTR(-ENODEV); | |
669 | ||
670 | mnt = alloc_vfsmnt(name); | |
671 | if (!mnt) | |
672 | return ERR_PTR(-ENOMEM); | |
673 | ||
674 | if (flags & MS_KERNMOUNT) | |
675 | mnt->mnt_flags = MNT_INTERNAL; | |
676 | ||
677 | root = mount_fs(type, flags, name, data); | |
678 | if (IS_ERR(root)) { | |
679 | free_vfsmnt(mnt); | |
680 | return ERR_CAST(root); | |
681 | } | |
682 | ||
683 | mnt->mnt_root = root; | |
684 | mnt->mnt_sb = root->d_sb; | |
685 | mnt->mnt_mountpoint = mnt->mnt_root; | |
686 | mnt->mnt_parent = mnt; | |
687 | return mnt; | |
688 | } | |
689 | EXPORT_SYMBOL_GPL(vfs_kern_mount); | |
690 | ||
36341f64 RP |
691 | static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, |
692 | int flag) | |
1da177e4 LT |
693 | { |
694 | struct super_block *sb = old->mnt_sb; | |
695 | struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); | |
696 | ||
697 | if (mnt) { | |
719f5d7f MS |
698 | if (flag & (CL_SLAVE | CL_PRIVATE)) |
699 | mnt->mnt_group_id = 0; /* not a peer of original */ | |
700 | else | |
701 | mnt->mnt_group_id = old->mnt_group_id; | |
702 | ||
703 | if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { | |
704 | int err = mnt_alloc_group_id(mnt); | |
705 | if (err) | |
706 | goto out_free; | |
707 | } | |
708 | ||
be1a16a0 | 709 | mnt->mnt_flags = old->mnt_flags & ~MNT_WRITE_HOLD; |
1da177e4 LT |
710 | atomic_inc(&sb->s_active); |
711 | mnt->mnt_sb = sb; | |
712 | mnt->mnt_root = dget(root); | |
713 | mnt->mnt_mountpoint = mnt->mnt_root; | |
714 | mnt->mnt_parent = mnt; | |
b90fa9ae | 715 | |
5afe0022 RP |
716 | if (flag & CL_SLAVE) { |
717 | list_add(&mnt->mnt_slave, &old->mnt_slave_list); | |
718 | mnt->mnt_master = old; | |
719 | CLEAR_MNT_SHARED(mnt); | |
8aec0809 | 720 | } else if (!(flag & CL_PRIVATE)) { |
796a6b52 | 721 | if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) |
5afe0022 RP |
722 | list_add(&mnt->mnt_share, &old->mnt_share); |
723 | if (IS_MNT_SLAVE(old)) | |
724 | list_add(&mnt->mnt_slave, &old->mnt_slave); | |
725 | mnt->mnt_master = old->mnt_master; | |
726 | } | |
b90fa9ae RP |
727 | if (flag & CL_MAKE_SHARED) |
728 | set_mnt_shared(mnt); | |
1da177e4 LT |
729 | |
730 | /* stick the duplicate mount on the same expiry list | |
731 | * as the original if that was on one */ | |
36341f64 | 732 | if (flag & CL_EXPIRE) { |
36341f64 RP |
733 | if (!list_empty(&old->mnt_expire)) |
734 | list_add(&mnt->mnt_expire, &old->mnt_expire); | |
36341f64 | 735 | } |
1da177e4 LT |
736 | } |
737 | return mnt; | |
719f5d7f MS |
738 | |
739 | out_free: | |
740 | free_vfsmnt(mnt); | |
741 | return NULL; | |
1da177e4 LT |
742 | } |
743 | ||
b3e19d92 | 744 | static inline void mntfree(struct vfsmount *mnt) |
1da177e4 LT |
745 | { |
746 | struct super_block *sb = mnt->mnt_sb; | |
b3e19d92 | 747 | |
3d733633 DH |
748 | /* |
749 | * This probably indicates that somebody messed | |
750 | * up a mnt_want/drop_write() pair. If this | |
751 | * happens, the filesystem was probably unable | |
752 | * to make r/w->r/o transitions. | |
753 | */ | |
d3ef3d73 | 754 | /* |
b3e19d92 NP |
755 | * The locking used to deal with mnt_count decrement provides barriers, |
756 | * so mnt_get_writers() below is safe. | |
d3ef3d73 | 757 | */ |
c6653a83 | 758 | WARN_ON(mnt_get_writers(mnt)); |
ca9c726e | 759 | fsnotify_vfsmount_delete(mnt); |
1da177e4 LT |
760 | dput(mnt->mnt_root); |
761 | free_vfsmnt(mnt); | |
762 | deactivate_super(sb); | |
763 | } | |
764 | ||
f03c6599 | 765 | static void mntput_no_expire(struct vfsmount *mnt) |
b3e19d92 | 766 | { |
b3e19d92 | 767 | put_again: |
f03c6599 AV |
768 | #ifdef CONFIG_SMP |
769 | br_read_lock(vfsmount_lock); | |
770 | if (likely(atomic_read(&mnt->mnt_longterm))) { | |
aa9c0e07 | 771 | mnt_add_count(mnt, -1); |
b3e19d92 | 772 | br_read_unlock(vfsmount_lock); |
f03c6599 | 773 | return; |
b3e19d92 | 774 | } |
f03c6599 | 775 | br_read_unlock(vfsmount_lock); |
b3e19d92 | 776 | |
99b7db7b | 777 | br_write_lock(vfsmount_lock); |
aa9c0e07 | 778 | mnt_add_count(mnt, -1); |
b3e19d92 | 779 | if (mnt_get_count(mnt)) { |
99b7db7b NP |
780 | br_write_unlock(vfsmount_lock); |
781 | return; | |
782 | } | |
b3e19d92 | 783 | #else |
aa9c0e07 | 784 | mnt_add_count(mnt, -1); |
b3e19d92 | 785 | if (likely(mnt_get_count(mnt))) |
99b7db7b | 786 | return; |
b3e19d92 | 787 | br_write_lock(vfsmount_lock); |
f03c6599 | 788 | #endif |
b3e19d92 NP |
789 | if (unlikely(mnt->mnt_pinned)) { |
790 | mnt_add_count(mnt, mnt->mnt_pinned + 1); | |
791 | mnt->mnt_pinned = 0; | |
792 | br_write_unlock(vfsmount_lock); | |
793 | acct_auto_close_mnt(mnt); | |
794 | goto put_again; | |
7b7b1ace | 795 | } |
99b7db7b | 796 | br_write_unlock(vfsmount_lock); |
b3e19d92 NP |
797 | mntfree(mnt); |
798 | } | |
b3e19d92 NP |
799 | |
800 | void mntput(struct vfsmount *mnt) | |
801 | { | |
802 | if (mnt) { | |
803 | /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ | |
804 | if (unlikely(mnt->mnt_expiry_mark)) | |
805 | mnt->mnt_expiry_mark = 0; | |
f03c6599 | 806 | mntput_no_expire(mnt); |
b3e19d92 NP |
807 | } |
808 | } | |
809 | EXPORT_SYMBOL(mntput); | |
810 | ||
811 | struct vfsmount *mntget(struct vfsmount *mnt) | |
812 | { | |
813 | if (mnt) | |
aa9c0e07 | 814 | mnt_add_count(mnt, 1); |
b3e19d92 NP |
815 | return mnt; |
816 | } | |
817 | EXPORT_SYMBOL(mntget); | |
818 | ||
7b7b1ace AV |
819 | void mnt_pin(struct vfsmount *mnt) |
820 | { | |
99b7db7b | 821 | br_write_lock(vfsmount_lock); |
7b7b1ace | 822 | mnt->mnt_pinned++; |
99b7db7b | 823 | br_write_unlock(vfsmount_lock); |
7b7b1ace | 824 | } |
7b7b1ace AV |
825 | EXPORT_SYMBOL(mnt_pin); |
826 | ||
827 | void mnt_unpin(struct vfsmount *mnt) | |
828 | { | |
99b7db7b | 829 | br_write_lock(vfsmount_lock); |
7b7b1ace | 830 | if (mnt->mnt_pinned) { |
aa9c0e07 | 831 | mnt_add_count(mnt, 1); |
7b7b1ace AV |
832 | mnt->mnt_pinned--; |
833 | } | |
99b7db7b | 834 | br_write_unlock(vfsmount_lock); |
7b7b1ace | 835 | } |
7b7b1ace | 836 | EXPORT_SYMBOL(mnt_unpin); |
1da177e4 | 837 | |
b3b304a2 MS |
838 | static inline void mangle(struct seq_file *m, const char *s) |
839 | { | |
840 | seq_escape(m, s, " \t\n\\"); | |
841 | } | |
842 | ||
843 | /* | |
844 | * Simple .show_options callback for filesystems which don't want to | |
845 | * implement more complex mount option showing. | |
846 | * | |
847 | * See also save_mount_options(). | |
848 | */ | |
849 | int generic_show_options(struct seq_file *m, struct vfsmount *mnt) | |
850 | { | |
2a32cebd AV |
851 | const char *options; |
852 | ||
853 | rcu_read_lock(); | |
854 | options = rcu_dereference(mnt->mnt_sb->s_options); | |
b3b304a2 MS |
855 | |
856 | if (options != NULL && options[0]) { | |
857 | seq_putc(m, ','); | |
858 | mangle(m, options); | |
859 | } | |
2a32cebd | 860 | rcu_read_unlock(); |
b3b304a2 MS |
861 | |
862 | return 0; | |
863 | } | |
864 | EXPORT_SYMBOL(generic_show_options); | |
865 | ||
866 | /* | |
867 | * If filesystem uses generic_show_options(), this function should be | |
868 | * called from the fill_super() callback. | |
869 | * | |
870 | * The .remount_fs callback usually needs to be handled in a special | |
871 | * way, to make sure, that previous options are not overwritten if the | |
872 | * remount fails. | |
873 | * | |
874 | * Also note, that if the filesystem's .remount_fs function doesn't | |
875 | * reset all options to their default value, but changes only newly | |
876 | * given options, then the displayed options will not reflect reality | |
877 | * any more. | |
878 | */ | |
879 | void save_mount_options(struct super_block *sb, char *options) | |
880 | { | |
2a32cebd AV |
881 | BUG_ON(sb->s_options); |
882 | rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); | |
b3b304a2 MS |
883 | } |
884 | EXPORT_SYMBOL(save_mount_options); | |
885 | ||
2a32cebd AV |
886 | void replace_mount_options(struct super_block *sb, char *options) |
887 | { | |
888 | char *old = sb->s_options; | |
889 | rcu_assign_pointer(sb->s_options, options); | |
890 | if (old) { | |
891 | synchronize_rcu(); | |
892 | kfree(old); | |
893 | } | |
894 | } | |
895 | EXPORT_SYMBOL(replace_mount_options); | |
896 | ||
a1a2c409 | 897 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
898 | /* iterator */ |
899 | static void *m_start(struct seq_file *m, loff_t *pos) | |
900 | { | |
a1a2c409 | 901 | struct proc_mounts *p = m->private; |
1da177e4 | 902 | |
390c6843 | 903 | down_read(&namespace_sem); |
a1a2c409 | 904 | return seq_list_start(&p->ns->list, *pos); |
1da177e4 LT |
905 | } |
906 | ||
907 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
908 | { | |
a1a2c409 | 909 | struct proc_mounts *p = m->private; |
b0765fb8 | 910 | |
a1a2c409 | 911 | return seq_list_next(v, &p->ns->list, pos); |
1da177e4 LT |
912 | } |
913 | ||
914 | static void m_stop(struct seq_file *m, void *v) | |
915 | { | |
390c6843 | 916 | up_read(&namespace_sem); |
1da177e4 LT |
917 | } |
918 | ||
9f5596af AV |
919 | int mnt_had_events(struct proc_mounts *p) |
920 | { | |
921 | struct mnt_namespace *ns = p->ns; | |
922 | int res = 0; | |
923 | ||
99b7db7b | 924 | br_read_lock(vfsmount_lock); |
f1514638 KS |
925 | if (p->m.poll_event != ns->event) { |
926 | p->m.poll_event = ns->event; | |
9f5596af AV |
927 | res = 1; |
928 | } | |
99b7db7b | 929 | br_read_unlock(vfsmount_lock); |
9f5596af AV |
930 | |
931 | return res; | |
932 | } | |
933 | ||
2d4d4864 RP |
934 | struct proc_fs_info { |
935 | int flag; | |
936 | const char *str; | |
937 | }; | |
938 | ||
2069f457 | 939 | static int show_sb_opts(struct seq_file *m, struct super_block *sb) |
1da177e4 | 940 | { |
2d4d4864 | 941 | static const struct proc_fs_info fs_info[] = { |
1da177e4 LT |
942 | { MS_SYNCHRONOUS, ",sync" }, |
943 | { MS_DIRSYNC, ",dirsync" }, | |
944 | { MS_MANDLOCK, ",mand" }, | |
1da177e4 LT |
945 | { 0, NULL } |
946 | }; | |
2d4d4864 RP |
947 | const struct proc_fs_info *fs_infop; |
948 | ||
949 | for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { | |
950 | if (sb->s_flags & fs_infop->flag) | |
951 | seq_puts(m, fs_infop->str); | |
952 | } | |
2069f457 EP |
953 | |
954 | return security_sb_show_options(m, sb); | |
2d4d4864 RP |
955 | } |
956 | ||
957 | static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) | |
958 | { | |
959 | static const struct proc_fs_info mnt_info[] = { | |
1da177e4 LT |
960 | { MNT_NOSUID, ",nosuid" }, |
961 | { MNT_NODEV, ",nodev" }, | |
962 | { MNT_NOEXEC, ",noexec" }, | |
fc33a7bb CH |
963 | { MNT_NOATIME, ",noatime" }, |
964 | { MNT_NODIRATIME, ",nodiratime" }, | |
47ae32d6 | 965 | { MNT_RELATIME, ",relatime" }, |
1da177e4 LT |
966 | { 0, NULL } |
967 | }; | |
2d4d4864 RP |
968 | const struct proc_fs_info *fs_infop; |
969 | ||
970 | for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { | |
971 | if (mnt->mnt_flags & fs_infop->flag) | |
972 | seq_puts(m, fs_infop->str); | |
973 | } | |
974 | } | |
975 | ||
976 | static void show_type(struct seq_file *m, struct super_block *sb) | |
977 | { | |
978 | mangle(m, sb->s_type->name); | |
979 | if (sb->s_subtype && sb->s_subtype[0]) { | |
980 | seq_putc(m, '.'); | |
981 | mangle(m, sb->s_subtype); | |
982 | } | |
983 | } | |
984 | ||
985 | static int show_vfsmnt(struct seq_file *m, void *v) | |
986 | { | |
987 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); | |
988 | int err = 0; | |
c32c2f63 | 989 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; |
1da177e4 | 990 | |
c7f404b4 AV |
991 | if (mnt->mnt_sb->s_op->show_devname) { |
992 | err = mnt->mnt_sb->s_op->show_devname(m, mnt); | |
993 | if (err) | |
994 | goto out; | |
995 | } else { | |
996 | mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); | |
997 | } | |
1da177e4 | 998 | seq_putc(m, ' '); |
c32c2f63 | 999 | seq_path(m, &mnt_path, " \t\n\\"); |
1da177e4 | 1000 | seq_putc(m, ' '); |
2d4d4864 | 1001 | show_type(m, mnt->mnt_sb); |
2e4b7fcd | 1002 | seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); |
2069f457 EP |
1003 | err = show_sb_opts(m, mnt->mnt_sb); |
1004 | if (err) | |
1005 | goto out; | |
2d4d4864 | 1006 | show_mnt_opts(m, mnt); |
1da177e4 LT |
1007 | if (mnt->mnt_sb->s_op->show_options) |
1008 | err = mnt->mnt_sb->s_op->show_options(m, mnt); | |
1009 | seq_puts(m, " 0 0\n"); | |
2069f457 | 1010 | out: |
1da177e4 LT |
1011 | return err; |
1012 | } | |
1013 | ||
a1a2c409 | 1014 | const struct seq_operations mounts_op = { |
1da177e4 LT |
1015 | .start = m_start, |
1016 | .next = m_next, | |
1017 | .stop = m_stop, | |
1018 | .show = show_vfsmnt | |
1019 | }; | |
1020 | ||
2d4d4864 RP |
1021 | static int show_mountinfo(struct seq_file *m, void *v) |
1022 | { | |
1023 | struct proc_mounts *p = m->private; | |
1024 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); | |
1025 | struct super_block *sb = mnt->mnt_sb; | |
1026 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; | |
1027 | struct path root = p->root; | |
1028 | int err = 0; | |
1029 | ||
1030 | seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, | |
1031 | MAJOR(sb->s_dev), MINOR(sb->s_dev)); | |
c7f404b4 AV |
1032 | if (sb->s_op->show_path) |
1033 | err = sb->s_op->show_path(m, mnt); | |
1034 | else | |
1035 | seq_dentry(m, mnt->mnt_root, " \t\n\\"); | |
1036 | if (err) | |
1037 | goto out; | |
2d4d4864 | 1038 | seq_putc(m, ' '); |
02125a82 AV |
1039 | |
1040 | /* mountpoints outside of chroot jail will give SEQ_SKIP on this */ | |
1041 | err = seq_path_root(m, &mnt_path, &root, " \t\n\\"); | |
1042 | if (err) | |
1043 | goto out; | |
1044 | ||
2d4d4864 RP |
1045 | seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); |
1046 | show_mnt_opts(m, mnt); | |
1047 | ||
1048 | /* Tagged fields ("foo:X" or "bar") */ | |
1049 | if (IS_MNT_SHARED(mnt)) | |
1050 | seq_printf(m, " shared:%i", mnt->mnt_group_id); | |
97e7e0f7 MS |
1051 | if (IS_MNT_SLAVE(mnt)) { |
1052 | int master = mnt->mnt_master->mnt_group_id; | |
1053 | int dom = get_dominating_id(mnt, &p->root); | |
1054 | seq_printf(m, " master:%i", master); | |
1055 | if (dom && dom != master) | |
1056 | seq_printf(m, " propagate_from:%i", dom); | |
1057 | } | |
2d4d4864 RP |
1058 | if (IS_MNT_UNBINDABLE(mnt)) |
1059 | seq_puts(m, " unbindable"); | |
1060 | ||
1061 | /* Filesystem specific data */ | |
1062 | seq_puts(m, " - "); | |
1063 | show_type(m, sb); | |
1064 | seq_putc(m, ' '); | |
c7f404b4 AV |
1065 | if (sb->s_op->show_devname) |
1066 | err = sb->s_op->show_devname(m, mnt); | |
1067 | else | |
1068 | mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); | |
1069 | if (err) | |
1070 | goto out; | |
2d4d4864 | 1071 | seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); |
2069f457 EP |
1072 | err = show_sb_opts(m, sb); |
1073 | if (err) | |
1074 | goto out; | |
2d4d4864 RP |
1075 | if (sb->s_op->show_options) |
1076 | err = sb->s_op->show_options(m, mnt); | |
1077 | seq_putc(m, '\n'); | |
2069f457 | 1078 | out: |
2d4d4864 RP |
1079 | return err; |
1080 | } | |
1081 | ||
1082 | const struct seq_operations mountinfo_op = { | |
1083 | .start = m_start, | |
1084 | .next = m_next, | |
1085 | .stop = m_stop, | |
1086 | .show = show_mountinfo, | |
1087 | }; | |
1088 | ||
b4629fe2 CL |
1089 | static int show_vfsstat(struct seq_file *m, void *v) |
1090 | { | |
b0765fb8 | 1091 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); |
c32c2f63 | 1092 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; |
b4629fe2 CL |
1093 | int err = 0; |
1094 | ||
1095 | /* device */ | |
c7f404b4 | 1096 | if (mnt->mnt_sb->s_op->show_devname) { |
a877ee03 | 1097 | seq_puts(m, "device "); |
c7f404b4 AV |
1098 | err = mnt->mnt_sb->s_op->show_devname(m, mnt); |
1099 | } else { | |
1100 | if (mnt->mnt_devname) { | |
1101 | seq_puts(m, "device "); | |
1102 | mangle(m, mnt->mnt_devname); | |
1103 | } else | |
1104 | seq_puts(m, "no device"); | |
1105 | } | |
b4629fe2 CL |
1106 | |
1107 | /* mount point */ | |
1108 | seq_puts(m, " mounted on "); | |
c32c2f63 | 1109 | seq_path(m, &mnt_path, " \t\n\\"); |
b4629fe2 CL |
1110 | seq_putc(m, ' '); |
1111 | ||
1112 | /* file system type */ | |
1113 | seq_puts(m, "with fstype "); | |
2d4d4864 | 1114 | show_type(m, mnt->mnt_sb); |
b4629fe2 CL |
1115 | |
1116 | /* optional statistics */ | |
1117 | if (mnt->mnt_sb->s_op->show_stats) { | |
1118 | seq_putc(m, ' '); | |
c7f404b4 AV |
1119 | if (!err) |
1120 | err = mnt->mnt_sb->s_op->show_stats(m, mnt); | |
b4629fe2 CL |
1121 | } |
1122 | ||
1123 | seq_putc(m, '\n'); | |
1124 | return err; | |
1125 | } | |
1126 | ||
a1a2c409 | 1127 | const struct seq_operations mountstats_op = { |
b4629fe2 CL |
1128 | .start = m_start, |
1129 | .next = m_next, | |
1130 | .stop = m_stop, | |
1131 | .show = show_vfsstat, | |
1132 | }; | |
a1a2c409 | 1133 | #endif /* CONFIG_PROC_FS */ |
b4629fe2 | 1134 | |
1da177e4 LT |
1135 | /** |
1136 | * may_umount_tree - check if a mount tree is busy | |
1137 | * @mnt: root of mount tree | |
1138 | * | |
1139 | * This is called to check if a tree of mounts has any | |
1140 | * open files, pwds, chroots or sub mounts that are | |
1141 | * busy. | |
1142 | */ | |
1143 | int may_umount_tree(struct vfsmount *mnt) | |
1144 | { | |
36341f64 RP |
1145 | int actual_refs = 0; |
1146 | int minimum_refs = 0; | |
1147 | struct vfsmount *p; | |
1da177e4 | 1148 | |
b3e19d92 NP |
1149 | /* write lock needed for mnt_get_count */ |
1150 | br_write_lock(vfsmount_lock); | |
36341f64 | 1151 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
b3e19d92 | 1152 | actual_refs += mnt_get_count(p); |
1da177e4 | 1153 | minimum_refs += 2; |
1da177e4 | 1154 | } |
b3e19d92 | 1155 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
1156 | |
1157 | if (actual_refs > minimum_refs) | |
e3474a8e | 1158 | return 0; |
1da177e4 | 1159 | |
e3474a8e | 1160 | return 1; |
1da177e4 LT |
1161 | } |
1162 | ||
1163 | EXPORT_SYMBOL(may_umount_tree); | |
1164 | ||
1165 | /** | |
1166 | * may_umount - check if a mount point is busy | |
1167 | * @mnt: root of mount | |
1168 | * | |
1169 | * This is called to check if a mount point has any | |
1170 | * open files, pwds, chroots or sub mounts. If the | |
1171 | * mount has sub mounts this will return busy | |
1172 | * regardless of whether the sub mounts are busy. | |
1173 | * | |
1174 | * Doesn't take quota and stuff into account. IOW, in some cases it will | |
1175 | * give false negatives. The main reason why it's here is that we need | |
1176 | * a non-destructive way to look for easily umountable filesystems. | |
1177 | */ | |
1178 | int may_umount(struct vfsmount *mnt) | |
1179 | { | |
e3474a8e | 1180 | int ret = 1; |
8ad08d8a | 1181 | down_read(&namespace_sem); |
b3e19d92 | 1182 | br_write_lock(vfsmount_lock); |
a05964f3 | 1183 | if (propagate_mount_busy(mnt, 2)) |
e3474a8e | 1184 | ret = 0; |
b3e19d92 | 1185 | br_write_unlock(vfsmount_lock); |
8ad08d8a | 1186 | up_read(&namespace_sem); |
a05964f3 | 1187 | return ret; |
1da177e4 LT |
1188 | } |
1189 | ||
1190 | EXPORT_SYMBOL(may_umount); | |
1191 | ||
b90fa9ae | 1192 | void release_mounts(struct list_head *head) |
70fbcdf4 RP |
1193 | { |
1194 | struct vfsmount *mnt; | |
bf066c7d | 1195 | while (!list_empty(head)) { |
b5e61818 | 1196 | mnt = list_first_entry(head, struct vfsmount, mnt_hash); |
70fbcdf4 | 1197 | list_del_init(&mnt->mnt_hash); |
b2dba1af | 1198 | if (mnt_has_parent(mnt)) { |
70fbcdf4 RP |
1199 | struct dentry *dentry; |
1200 | struct vfsmount *m; | |
99b7db7b NP |
1201 | |
1202 | br_write_lock(vfsmount_lock); | |
70fbcdf4 RP |
1203 | dentry = mnt->mnt_mountpoint; |
1204 | m = mnt->mnt_parent; | |
1205 | mnt->mnt_mountpoint = mnt->mnt_root; | |
1206 | mnt->mnt_parent = mnt; | |
7c4b93d8 | 1207 | m->mnt_ghosts--; |
99b7db7b | 1208 | br_write_unlock(vfsmount_lock); |
70fbcdf4 RP |
1209 | dput(dentry); |
1210 | mntput(m); | |
1211 | } | |
f03c6599 | 1212 | mntput(mnt); |
70fbcdf4 RP |
1213 | } |
1214 | } | |
1215 | ||
99b7db7b NP |
1216 | /* |
1217 | * vfsmount lock must be held for write | |
1218 | * namespace_sem must be held for write | |
1219 | */ | |
a05964f3 | 1220 | void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) |
1da177e4 | 1221 | { |
7b8a53fd | 1222 | LIST_HEAD(tmp_list); |
1da177e4 | 1223 | struct vfsmount *p; |
1da177e4 | 1224 | |
1bfba4e8 | 1225 | for (p = mnt; p; p = next_mnt(p, mnt)) |
7b8a53fd | 1226 | list_move(&p->mnt_hash, &tmp_list); |
1da177e4 | 1227 | |
a05964f3 | 1228 | if (propagate) |
7b8a53fd | 1229 | propagate_umount(&tmp_list); |
a05964f3 | 1230 | |
7b8a53fd | 1231 | list_for_each_entry(p, &tmp_list, mnt_hash) { |
70fbcdf4 RP |
1232 | list_del_init(&p->mnt_expire); |
1233 | list_del_init(&p->mnt_list); | |
6b3286ed KK |
1234 | __touch_mnt_namespace(p->mnt_ns); |
1235 | p->mnt_ns = NULL; | |
7e3d0eb0 | 1236 | __mnt_make_shortterm(p); |
70fbcdf4 | 1237 | list_del_init(&p->mnt_child); |
b2dba1af | 1238 | if (mnt_has_parent(p)) { |
7c4b93d8 | 1239 | p->mnt_parent->mnt_ghosts++; |
aa0a4cf0 | 1240 | dentry_reset_mounted(p->mnt_mountpoint); |
7c4b93d8 | 1241 | } |
a05964f3 | 1242 | change_mnt_propagation(p, MS_PRIVATE); |
1da177e4 | 1243 | } |
7b8a53fd | 1244 | list_splice(&tmp_list, kill); |
1da177e4 LT |
1245 | } |
1246 | ||
c35038be AV |
1247 | static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); |
1248 | ||
1da177e4 LT |
1249 | static int do_umount(struct vfsmount *mnt, int flags) |
1250 | { | |
b58fed8b | 1251 | struct super_block *sb = mnt->mnt_sb; |
1da177e4 | 1252 | int retval; |
70fbcdf4 | 1253 | LIST_HEAD(umount_list); |
1da177e4 LT |
1254 | |
1255 | retval = security_sb_umount(mnt, flags); | |
1256 | if (retval) | |
1257 | return retval; | |
1258 | ||
1259 | /* | |
1260 | * Allow userspace to request a mountpoint be expired rather than | |
1261 | * unmounting unconditionally. Unmount only happens if: | |
1262 | * (1) the mark is already set (the mark is cleared by mntput()) | |
1263 | * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] | |
1264 | */ | |
1265 | if (flags & MNT_EXPIRE) { | |
6ac08c39 | 1266 | if (mnt == current->fs->root.mnt || |
1da177e4 LT |
1267 | flags & (MNT_FORCE | MNT_DETACH)) |
1268 | return -EINVAL; | |
1269 | ||
b3e19d92 NP |
1270 | /* |
1271 | * probably don't strictly need the lock here if we examined | |
1272 | * all race cases, but it's a slowpath. | |
1273 | */ | |
1274 | br_write_lock(vfsmount_lock); | |
1275 | if (mnt_get_count(mnt) != 2) { | |
bf9faa2a | 1276 | br_write_unlock(vfsmount_lock); |
1da177e4 | 1277 | return -EBUSY; |
b3e19d92 NP |
1278 | } |
1279 | br_write_unlock(vfsmount_lock); | |
1da177e4 LT |
1280 | |
1281 | if (!xchg(&mnt->mnt_expiry_mark, 1)) | |
1282 | return -EAGAIN; | |
1283 | } | |
1284 | ||
1285 | /* | |
1286 | * If we may have to abort operations to get out of this | |
1287 | * mount, and they will themselves hold resources we must | |
1288 | * allow the fs to do things. In the Unix tradition of | |
1289 | * 'Gee thats tricky lets do it in userspace' the umount_begin | |
1290 | * might fail to complete on the first run through as other tasks | |
1291 | * must return, and the like. Thats for the mount program to worry | |
1292 | * about for the moment. | |
1293 | */ | |
1294 | ||
42faad99 | 1295 | if (flags & MNT_FORCE && sb->s_op->umount_begin) { |
42faad99 | 1296 | sb->s_op->umount_begin(sb); |
42faad99 | 1297 | } |
1da177e4 LT |
1298 | |
1299 | /* | |
1300 | * No sense to grab the lock for this test, but test itself looks | |
1301 | * somewhat bogus. Suggestions for better replacement? | |
1302 | * Ho-hum... In principle, we might treat that as umount + switch | |
1303 | * to rootfs. GC would eventually take care of the old vfsmount. | |
1304 | * Actually it makes sense, especially if rootfs would contain a | |
1305 | * /reboot - static binary that would close all descriptors and | |
1306 | * call reboot(9). Then init(8) could umount root and exec /reboot. | |
1307 | */ | |
6ac08c39 | 1308 | if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { |
1da177e4 LT |
1309 | /* |
1310 | * Special case for "unmounting" root ... | |
1311 | * we just try to remount it readonly. | |
1312 | */ | |
1313 | down_write(&sb->s_umount); | |
4aa98cf7 | 1314 | if (!(sb->s_flags & MS_RDONLY)) |
1da177e4 | 1315 | retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); |
1da177e4 LT |
1316 | up_write(&sb->s_umount); |
1317 | return retval; | |
1318 | } | |
1319 | ||
390c6843 | 1320 | down_write(&namespace_sem); |
99b7db7b | 1321 | br_write_lock(vfsmount_lock); |
5addc5dd | 1322 | event++; |
1da177e4 | 1323 | |
c35038be AV |
1324 | if (!(flags & MNT_DETACH)) |
1325 | shrink_submounts(mnt, &umount_list); | |
1326 | ||
1da177e4 | 1327 | retval = -EBUSY; |
a05964f3 | 1328 | if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { |
1da177e4 | 1329 | if (!list_empty(&mnt->mnt_list)) |
a05964f3 | 1330 | umount_tree(mnt, 1, &umount_list); |
1da177e4 LT |
1331 | retval = 0; |
1332 | } | |
99b7db7b | 1333 | br_write_unlock(vfsmount_lock); |
390c6843 | 1334 | up_write(&namespace_sem); |
70fbcdf4 | 1335 | release_mounts(&umount_list); |
1da177e4 LT |
1336 | return retval; |
1337 | } | |
1338 | ||
1339 | /* | |
1340 | * Now umount can handle mount points as well as block devices. | |
1341 | * This is important for filesystems which use unnamed block devices. | |
1342 | * | |
1343 | * We now support a flag for forced unmount like the other 'big iron' | |
1344 | * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD | |
1345 | */ | |
1346 | ||
bdc480e3 | 1347 | SYSCALL_DEFINE2(umount, char __user *, name, int, flags) |
1da177e4 | 1348 | { |
2d8f3038 | 1349 | struct path path; |
1da177e4 | 1350 | int retval; |
db1f05bb | 1351 | int lookup_flags = 0; |
1da177e4 | 1352 | |
db1f05bb MS |
1353 | if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) |
1354 | return -EINVAL; | |
1355 | ||
1356 | if (!(flags & UMOUNT_NOFOLLOW)) | |
1357 | lookup_flags |= LOOKUP_FOLLOW; | |
1358 | ||
1359 | retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); | |
1da177e4 LT |
1360 | if (retval) |
1361 | goto out; | |
1362 | retval = -EINVAL; | |
2d8f3038 | 1363 | if (path.dentry != path.mnt->mnt_root) |
1da177e4 | 1364 | goto dput_and_out; |
2d8f3038 | 1365 | if (!check_mnt(path.mnt)) |
1da177e4 LT |
1366 | goto dput_and_out; |
1367 | ||
1368 | retval = -EPERM; | |
1369 | if (!capable(CAP_SYS_ADMIN)) | |
1370 | goto dput_and_out; | |
1371 | ||
2d8f3038 | 1372 | retval = do_umount(path.mnt, flags); |
1da177e4 | 1373 | dput_and_out: |
429731b1 | 1374 | /* we mustn't call path_put() as that would clear mnt_expiry_mark */ |
2d8f3038 AV |
1375 | dput(path.dentry); |
1376 | mntput_no_expire(path.mnt); | |
1da177e4 LT |
1377 | out: |
1378 | return retval; | |
1379 | } | |
1380 | ||
1381 | #ifdef __ARCH_WANT_SYS_OLDUMOUNT | |
1382 | ||
1383 | /* | |
b58fed8b | 1384 | * The 2.0 compatible umount. No flags. |
1da177e4 | 1385 | */ |
bdc480e3 | 1386 | SYSCALL_DEFINE1(oldumount, char __user *, name) |
1da177e4 | 1387 | { |
b58fed8b | 1388 | return sys_umount(name, 0); |
1da177e4 LT |
1389 | } |
1390 | ||
1391 | #endif | |
1392 | ||
2d92ab3c | 1393 | static int mount_is_safe(struct path *path) |
1da177e4 LT |
1394 | { |
1395 | if (capable(CAP_SYS_ADMIN)) | |
1396 | return 0; | |
1397 | return -EPERM; | |
1398 | #ifdef notyet | |
2d92ab3c | 1399 | if (S_ISLNK(path->dentry->d_inode->i_mode)) |
1da177e4 | 1400 | return -EPERM; |
2d92ab3c | 1401 | if (path->dentry->d_inode->i_mode & S_ISVTX) { |
da9592ed | 1402 | if (current_uid() != path->dentry->d_inode->i_uid) |
1da177e4 LT |
1403 | return -EPERM; |
1404 | } | |
2d92ab3c | 1405 | if (inode_permission(path->dentry->d_inode, MAY_WRITE)) |
1da177e4 LT |
1406 | return -EPERM; |
1407 | return 0; | |
1408 | #endif | |
1409 | } | |
1410 | ||
b90fa9ae | 1411 | struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, |
36341f64 | 1412 | int flag) |
1da177e4 LT |
1413 | { |
1414 | struct vfsmount *res, *p, *q, *r, *s; | |
1a390689 | 1415 | struct path path; |
1da177e4 | 1416 | |
9676f0c6 RP |
1417 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) |
1418 | return NULL; | |
1419 | ||
36341f64 | 1420 | res = q = clone_mnt(mnt, dentry, flag); |
1da177e4 LT |
1421 | if (!q) |
1422 | goto Enomem; | |
1423 | q->mnt_mountpoint = mnt->mnt_mountpoint; | |
1424 | ||
1425 | p = mnt; | |
fdadd65f | 1426 | list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { |
7ec02ef1 | 1427 | if (!is_subdir(r->mnt_mountpoint, dentry)) |
1da177e4 LT |
1428 | continue; |
1429 | ||
1430 | for (s = r; s; s = next_mnt(s, r)) { | |
9676f0c6 RP |
1431 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { |
1432 | s = skip_mnt_tree(s); | |
1433 | continue; | |
1434 | } | |
1da177e4 LT |
1435 | while (p != s->mnt_parent) { |
1436 | p = p->mnt_parent; | |
1437 | q = q->mnt_parent; | |
1438 | } | |
1439 | p = s; | |
1a390689 AV |
1440 | path.mnt = q; |
1441 | path.dentry = p->mnt_mountpoint; | |
36341f64 | 1442 | q = clone_mnt(p, p->mnt_root, flag); |
1da177e4 LT |
1443 | if (!q) |
1444 | goto Enomem; | |
99b7db7b | 1445 | br_write_lock(vfsmount_lock); |
1da177e4 | 1446 | list_add_tail(&q->mnt_list, &res->mnt_list); |
1a390689 | 1447 | attach_mnt(q, &path); |
99b7db7b | 1448 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
1449 | } |
1450 | } | |
1451 | return res; | |
b58fed8b | 1452 | Enomem: |
1da177e4 | 1453 | if (res) { |
70fbcdf4 | 1454 | LIST_HEAD(umount_list); |
99b7db7b | 1455 | br_write_lock(vfsmount_lock); |
a05964f3 | 1456 | umount_tree(res, 0, &umount_list); |
99b7db7b | 1457 | br_write_unlock(vfsmount_lock); |
70fbcdf4 | 1458 | release_mounts(&umount_list); |
1da177e4 LT |
1459 | } |
1460 | return NULL; | |
1461 | } | |
1462 | ||
589ff870 | 1463 | struct vfsmount *collect_mounts(struct path *path) |
8aec0809 AV |
1464 | { |
1465 | struct vfsmount *tree; | |
1a60a280 | 1466 | down_write(&namespace_sem); |
589ff870 | 1467 | tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); |
1a60a280 | 1468 | up_write(&namespace_sem); |
8aec0809 AV |
1469 | return tree; |
1470 | } | |
1471 | ||
1472 | void drop_collected_mounts(struct vfsmount *mnt) | |
1473 | { | |
1474 | LIST_HEAD(umount_list); | |
1a60a280 | 1475 | down_write(&namespace_sem); |
99b7db7b | 1476 | br_write_lock(vfsmount_lock); |
8aec0809 | 1477 | umount_tree(mnt, 0, &umount_list); |
99b7db7b | 1478 | br_write_unlock(vfsmount_lock); |
1a60a280 | 1479 | up_write(&namespace_sem); |
8aec0809 AV |
1480 | release_mounts(&umount_list); |
1481 | } | |
1482 | ||
1f707137 AV |
1483 | int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, |
1484 | struct vfsmount *root) | |
1485 | { | |
1486 | struct vfsmount *mnt; | |
1487 | int res = f(root, arg); | |
1488 | if (res) | |
1489 | return res; | |
1490 | list_for_each_entry(mnt, &root->mnt_list, mnt_list) { | |
1491 | res = f(mnt, arg); | |
1492 | if (res) | |
1493 | return res; | |
1494 | } | |
1495 | return 0; | |
1496 | } | |
1497 | ||
719f5d7f MS |
1498 | static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) |
1499 | { | |
1500 | struct vfsmount *p; | |
1501 | ||
1502 | for (p = mnt; p != end; p = next_mnt(p, mnt)) { | |
1503 | if (p->mnt_group_id && !IS_MNT_SHARED(p)) | |
1504 | mnt_release_group_id(p); | |
1505 | } | |
1506 | } | |
1507 | ||
1508 | static int invent_group_ids(struct vfsmount *mnt, bool recurse) | |
1509 | { | |
1510 | struct vfsmount *p; | |
1511 | ||
1512 | for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { | |
1513 | if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { | |
1514 | int err = mnt_alloc_group_id(p); | |
1515 | if (err) { | |
1516 | cleanup_group_ids(mnt, p); | |
1517 | return err; | |
1518 | } | |
1519 | } | |
1520 | } | |
1521 | ||
1522 | return 0; | |
1523 | } | |
1524 | ||
b90fa9ae RP |
1525 | /* |
1526 | * @source_mnt : mount tree to be attached | |
21444403 RP |
1527 | * @nd : place the mount tree @source_mnt is attached |
1528 | * @parent_nd : if non-null, detach the source_mnt from its parent and | |
1529 | * store the parent mount and mountpoint dentry. | |
1530 | * (done when source_mnt is moved) | |
b90fa9ae RP |
1531 | * |
1532 | * NOTE: in the table below explains the semantics when a source mount | |
1533 | * of a given type is attached to a destination mount of a given type. | |
9676f0c6 RP |
1534 | * --------------------------------------------------------------------------- |
1535 | * | BIND MOUNT OPERATION | | |
1536 | * |************************************************************************** | |
1537 | * | source-->| shared | private | slave | unbindable | | |
1538 | * | dest | | | | | | |
1539 | * | | | | | | | | |
1540 | * | v | | | | | | |
1541 | * |************************************************************************** | |
1542 | * | shared | shared (++) | shared (+) | shared(+++)| invalid | | |
1543 | * | | | | | | | |
1544 | * |non-shared| shared (+) | private | slave (*) | invalid | | |
1545 | * *************************************************************************** | |
b90fa9ae RP |
1546 | * A bind operation clones the source mount and mounts the clone on the |
1547 | * destination mount. | |
1548 | * | |
1549 | * (++) the cloned mount is propagated to all the mounts in the propagation | |
1550 | * tree of the destination mount and the cloned mount is added to | |
1551 | * the peer group of the source mount. | |
1552 | * (+) the cloned mount is created under the destination mount and is marked | |
1553 | * as shared. The cloned mount is added to the peer group of the source | |
1554 | * mount. | |
5afe0022 RP |
1555 | * (+++) the mount is propagated to all the mounts in the propagation tree |
1556 | * of the destination mount and the cloned mount is made slave | |
1557 | * of the same master as that of the source mount. The cloned mount | |
1558 | * is marked as 'shared and slave'. | |
1559 | * (*) the cloned mount is made a slave of the same master as that of the | |
1560 | * source mount. | |
1561 | * | |
9676f0c6 RP |
1562 | * --------------------------------------------------------------------------- |
1563 | * | MOVE MOUNT OPERATION | | |
1564 | * |************************************************************************** | |
1565 | * | source-->| shared | private | slave | unbindable | | |
1566 | * | dest | | | | | | |
1567 | * | | | | | | | | |
1568 | * | v | | | | | | |
1569 | * |************************************************************************** | |
1570 | * | shared | shared (+) | shared (+) | shared(+++) | invalid | | |
1571 | * | | | | | | | |
1572 | * |non-shared| shared (+*) | private | slave (*) | unbindable | | |
1573 | * *************************************************************************** | |
5afe0022 RP |
1574 | * |
1575 | * (+) the mount is moved to the destination. And is then propagated to | |
1576 | * all the mounts in the propagation tree of the destination mount. | |
21444403 | 1577 | * (+*) the mount is moved to the destination. |
5afe0022 RP |
1578 | * (+++) the mount is moved to the destination and is then propagated to |
1579 | * all the mounts belonging to the destination mount's propagation tree. | |
1580 | * the mount is marked as 'shared and slave'. | |
1581 | * (*) the mount continues to be a slave at the new location. | |
b90fa9ae RP |
1582 | * |
1583 | * if the source mount is a tree, the operations explained above is | |
1584 | * applied to each mount in the tree. | |
1585 | * Must be called without spinlocks held, since this function can sleep | |
1586 | * in allocations. | |
1587 | */ | |
1588 | static int attach_recursive_mnt(struct vfsmount *source_mnt, | |
1a390689 | 1589 | struct path *path, struct path *parent_path) |
b90fa9ae RP |
1590 | { |
1591 | LIST_HEAD(tree_list); | |
1a390689 AV |
1592 | struct vfsmount *dest_mnt = path->mnt; |
1593 | struct dentry *dest_dentry = path->dentry; | |
b90fa9ae | 1594 | struct vfsmount *child, *p; |
719f5d7f | 1595 | int err; |
b90fa9ae | 1596 | |
719f5d7f MS |
1597 | if (IS_MNT_SHARED(dest_mnt)) { |
1598 | err = invent_group_ids(source_mnt, true); | |
1599 | if (err) | |
1600 | goto out; | |
1601 | } | |
1602 | err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); | |
1603 | if (err) | |
1604 | goto out_cleanup_ids; | |
b90fa9ae | 1605 | |
99b7db7b | 1606 | br_write_lock(vfsmount_lock); |
df1a1ad2 | 1607 | |
b90fa9ae RP |
1608 | if (IS_MNT_SHARED(dest_mnt)) { |
1609 | for (p = source_mnt; p; p = next_mnt(p, source_mnt)) | |
1610 | set_mnt_shared(p); | |
1611 | } | |
1a390689 AV |
1612 | if (parent_path) { |
1613 | detach_mnt(source_mnt, parent_path); | |
1614 | attach_mnt(source_mnt, path); | |
e5d67f07 | 1615 | touch_mnt_namespace(parent_path->mnt->mnt_ns); |
21444403 RP |
1616 | } else { |
1617 | mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); | |
1618 | commit_tree(source_mnt); | |
1619 | } | |
b90fa9ae RP |
1620 | |
1621 | list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { | |
1622 | list_del_init(&child->mnt_hash); | |
1623 | commit_tree(child); | |
1624 | } | |
99b7db7b NP |
1625 | br_write_unlock(vfsmount_lock); |
1626 | ||
b90fa9ae | 1627 | return 0; |
719f5d7f MS |
1628 | |
1629 | out_cleanup_ids: | |
1630 | if (IS_MNT_SHARED(dest_mnt)) | |
1631 | cleanup_group_ids(source_mnt, NULL); | |
1632 | out: | |
1633 | return err; | |
b90fa9ae RP |
1634 | } |
1635 | ||
b12cea91 AV |
1636 | static int lock_mount(struct path *path) |
1637 | { | |
1638 | struct vfsmount *mnt; | |
1639 | retry: | |
1640 | mutex_lock(&path->dentry->d_inode->i_mutex); | |
1641 | if (unlikely(cant_mount(path->dentry))) { | |
1642 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
1643 | return -ENOENT; | |
1644 | } | |
1645 | down_write(&namespace_sem); | |
1646 | mnt = lookup_mnt(path); | |
1647 | if (likely(!mnt)) | |
1648 | return 0; | |
1649 | up_write(&namespace_sem); | |
1650 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
1651 | path_put(path); | |
1652 | path->mnt = mnt; | |
1653 | path->dentry = dget(mnt->mnt_root); | |
1654 | goto retry; | |
1655 | } | |
1656 | ||
1657 | static void unlock_mount(struct path *path) | |
1658 | { | |
1659 | up_write(&namespace_sem); | |
1660 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
1661 | } | |
1662 | ||
8c3ee42e | 1663 | static int graft_tree(struct vfsmount *mnt, struct path *path) |
1da177e4 | 1664 | { |
1da177e4 LT |
1665 | if (mnt->mnt_sb->s_flags & MS_NOUSER) |
1666 | return -EINVAL; | |
1667 | ||
8c3ee42e | 1668 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
1da177e4 LT |
1669 | S_ISDIR(mnt->mnt_root->d_inode->i_mode)) |
1670 | return -ENOTDIR; | |
1671 | ||
b12cea91 AV |
1672 | if (d_unlinked(path->dentry)) |
1673 | return -ENOENT; | |
1da177e4 | 1674 | |
b12cea91 | 1675 | return attach_recursive_mnt(mnt, path, NULL); |
1da177e4 LT |
1676 | } |
1677 | ||
7a2e8a8f VA |
1678 | /* |
1679 | * Sanity check the flags to change_mnt_propagation. | |
1680 | */ | |
1681 | ||
1682 | static int flags_to_propagation_type(int flags) | |
1683 | { | |
7c6e984d | 1684 | int type = flags & ~(MS_REC | MS_SILENT); |
7a2e8a8f VA |
1685 | |
1686 | /* Fail if any non-propagation flags are set */ | |
1687 | if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) | |
1688 | return 0; | |
1689 | /* Only one propagation flag should be set */ | |
1690 | if (!is_power_of_2(type)) | |
1691 | return 0; | |
1692 | return type; | |
1693 | } | |
1694 | ||
07b20889 RP |
1695 | /* |
1696 | * recursively change the type of the mountpoint. | |
1697 | */ | |
0a0d8a46 | 1698 | static int do_change_type(struct path *path, int flag) |
07b20889 | 1699 | { |
2d92ab3c | 1700 | struct vfsmount *m, *mnt = path->mnt; |
07b20889 | 1701 | int recurse = flag & MS_REC; |
7a2e8a8f | 1702 | int type; |
719f5d7f | 1703 | int err = 0; |
07b20889 | 1704 | |
ee6f9582 MS |
1705 | if (!capable(CAP_SYS_ADMIN)) |
1706 | return -EPERM; | |
1707 | ||
2d92ab3c | 1708 | if (path->dentry != path->mnt->mnt_root) |
07b20889 RP |
1709 | return -EINVAL; |
1710 | ||
7a2e8a8f VA |
1711 | type = flags_to_propagation_type(flag); |
1712 | if (!type) | |
1713 | return -EINVAL; | |
1714 | ||
07b20889 | 1715 | down_write(&namespace_sem); |
719f5d7f MS |
1716 | if (type == MS_SHARED) { |
1717 | err = invent_group_ids(mnt, recurse); | |
1718 | if (err) | |
1719 | goto out_unlock; | |
1720 | } | |
1721 | ||
99b7db7b | 1722 | br_write_lock(vfsmount_lock); |
07b20889 RP |
1723 | for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) |
1724 | change_mnt_propagation(m, type); | |
99b7db7b | 1725 | br_write_unlock(vfsmount_lock); |
719f5d7f MS |
1726 | |
1727 | out_unlock: | |
07b20889 | 1728 | up_write(&namespace_sem); |
719f5d7f | 1729 | return err; |
07b20889 RP |
1730 | } |
1731 | ||
1da177e4 LT |
1732 | /* |
1733 | * do loopback mount. | |
1734 | */ | |
0a0d8a46 | 1735 | static int do_loopback(struct path *path, char *old_name, |
2dafe1c4 | 1736 | int recurse) |
1da177e4 | 1737 | { |
b12cea91 | 1738 | LIST_HEAD(umount_list); |
2d92ab3c | 1739 | struct path old_path; |
1da177e4 | 1740 | struct vfsmount *mnt = NULL; |
2d92ab3c | 1741 | int err = mount_is_safe(path); |
1da177e4 LT |
1742 | if (err) |
1743 | return err; | |
1744 | if (!old_name || !*old_name) | |
1745 | return -EINVAL; | |
815d405c | 1746 | err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); |
1da177e4 LT |
1747 | if (err) |
1748 | return err; | |
1749 | ||
b12cea91 AV |
1750 | err = lock_mount(path); |
1751 | if (err) | |
1752 | goto out; | |
1753 | ||
1da177e4 | 1754 | err = -EINVAL; |
2d92ab3c | 1755 | if (IS_MNT_UNBINDABLE(old_path.mnt)) |
b12cea91 | 1756 | goto out2; |
9676f0c6 | 1757 | |
2d92ab3c | 1758 | if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) |
b12cea91 | 1759 | goto out2; |
1da177e4 | 1760 | |
ccd48bc7 AV |
1761 | err = -ENOMEM; |
1762 | if (recurse) | |
2d92ab3c | 1763 | mnt = copy_tree(old_path.mnt, old_path.dentry, 0); |
ccd48bc7 | 1764 | else |
2d92ab3c | 1765 | mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); |
ccd48bc7 AV |
1766 | |
1767 | if (!mnt) | |
b12cea91 | 1768 | goto out2; |
ccd48bc7 | 1769 | |
2d92ab3c | 1770 | err = graft_tree(mnt, path); |
ccd48bc7 | 1771 | if (err) { |
99b7db7b | 1772 | br_write_lock(vfsmount_lock); |
a05964f3 | 1773 | umount_tree(mnt, 0, &umount_list); |
99b7db7b | 1774 | br_write_unlock(vfsmount_lock); |
5b83d2c5 | 1775 | } |
b12cea91 AV |
1776 | out2: |
1777 | unlock_mount(path); | |
1778 | release_mounts(&umount_list); | |
ccd48bc7 | 1779 | out: |
2d92ab3c | 1780 | path_put(&old_path); |
1da177e4 LT |
1781 | return err; |
1782 | } | |
1783 | ||
2e4b7fcd DH |
1784 | static int change_mount_flags(struct vfsmount *mnt, int ms_flags) |
1785 | { | |
1786 | int error = 0; | |
1787 | int readonly_request = 0; | |
1788 | ||
1789 | if (ms_flags & MS_RDONLY) | |
1790 | readonly_request = 1; | |
1791 | if (readonly_request == __mnt_is_readonly(mnt)) | |
1792 | return 0; | |
1793 | ||
1794 | if (readonly_request) | |
1795 | error = mnt_make_readonly(mnt); | |
1796 | else | |
1797 | __mnt_unmake_readonly(mnt); | |
1798 | return error; | |
1799 | } | |
1800 | ||
1da177e4 LT |
1801 | /* |
1802 | * change filesystem flags. dir should be a physical root of filesystem. | |
1803 | * If you've mounted a non-root directory somewhere and want to do remount | |
1804 | * on it - tough luck. | |
1805 | */ | |
0a0d8a46 | 1806 | static int do_remount(struct path *path, int flags, int mnt_flags, |
1da177e4 LT |
1807 | void *data) |
1808 | { | |
1809 | int err; | |
2d92ab3c | 1810 | struct super_block *sb = path->mnt->mnt_sb; |
1da177e4 LT |
1811 | |
1812 | if (!capable(CAP_SYS_ADMIN)) | |
1813 | return -EPERM; | |
1814 | ||
2d92ab3c | 1815 | if (!check_mnt(path->mnt)) |
1da177e4 LT |
1816 | return -EINVAL; |
1817 | ||
2d92ab3c | 1818 | if (path->dentry != path->mnt->mnt_root) |
1da177e4 LT |
1819 | return -EINVAL; |
1820 | ||
ff36fe2c EP |
1821 | err = security_sb_remount(sb, data); |
1822 | if (err) | |
1823 | return err; | |
1824 | ||
1da177e4 | 1825 | down_write(&sb->s_umount); |
2e4b7fcd | 1826 | if (flags & MS_BIND) |
2d92ab3c | 1827 | err = change_mount_flags(path->mnt, flags); |
4aa98cf7 | 1828 | else |
2e4b7fcd | 1829 | err = do_remount_sb(sb, flags, data, 0); |
7b43a79f | 1830 | if (!err) { |
99b7db7b | 1831 | br_write_lock(vfsmount_lock); |
495d6c9c | 1832 | mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK; |
2d92ab3c | 1833 | path->mnt->mnt_flags = mnt_flags; |
99b7db7b | 1834 | br_write_unlock(vfsmount_lock); |
7b43a79f | 1835 | } |
1da177e4 | 1836 | up_write(&sb->s_umount); |
0e55a7cc | 1837 | if (!err) { |
99b7db7b | 1838 | br_write_lock(vfsmount_lock); |
0e55a7cc | 1839 | touch_mnt_namespace(path->mnt->mnt_ns); |
99b7db7b | 1840 | br_write_unlock(vfsmount_lock); |
0e55a7cc | 1841 | } |
1da177e4 LT |
1842 | return err; |
1843 | } | |
1844 | ||
9676f0c6 RP |
1845 | static inline int tree_contains_unbindable(struct vfsmount *mnt) |
1846 | { | |
1847 | struct vfsmount *p; | |
1848 | for (p = mnt; p; p = next_mnt(p, mnt)) { | |
1849 | if (IS_MNT_UNBINDABLE(p)) | |
1850 | return 1; | |
1851 | } | |
1852 | return 0; | |
1853 | } | |
1854 | ||
0a0d8a46 | 1855 | static int do_move_mount(struct path *path, char *old_name) |
1da177e4 | 1856 | { |
2d92ab3c | 1857 | struct path old_path, parent_path; |
1da177e4 LT |
1858 | struct vfsmount *p; |
1859 | int err = 0; | |
1860 | if (!capable(CAP_SYS_ADMIN)) | |
1861 | return -EPERM; | |
1862 | if (!old_name || !*old_name) | |
1863 | return -EINVAL; | |
2d92ab3c | 1864 | err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); |
1da177e4 LT |
1865 | if (err) |
1866 | return err; | |
1867 | ||
b12cea91 | 1868 | err = lock_mount(path); |
cc53ce53 DH |
1869 | if (err < 0) |
1870 | goto out; | |
1871 | ||
1da177e4 | 1872 | err = -EINVAL; |
2d92ab3c | 1873 | if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) |
1da177e4 LT |
1874 | goto out1; |
1875 | ||
f3da392e | 1876 | if (d_unlinked(path->dentry)) |
21444403 | 1877 | goto out1; |
1da177e4 LT |
1878 | |
1879 | err = -EINVAL; | |
2d92ab3c | 1880 | if (old_path.dentry != old_path.mnt->mnt_root) |
21444403 | 1881 | goto out1; |
1da177e4 | 1882 | |
b2dba1af | 1883 | if (!mnt_has_parent(old_path.mnt)) |
21444403 | 1884 | goto out1; |
1da177e4 | 1885 | |
2d92ab3c AV |
1886 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
1887 | S_ISDIR(old_path.dentry->d_inode->i_mode)) | |
21444403 RP |
1888 | goto out1; |
1889 | /* | |
1890 | * Don't move a mount residing in a shared parent. | |
1891 | */ | |
afac7cba | 1892 | if (IS_MNT_SHARED(old_path.mnt->mnt_parent)) |
21444403 | 1893 | goto out1; |
9676f0c6 RP |
1894 | /* |
1895 | * Don't move a mount tree containing unbindable mounts to a destination | |
1896 | * mount which is shared. | |
1897 | */ | |
2d92ab3c AV |
1898 | if (IS_MNT_SHARED(path->mnt) && |
1899 | tree_contains_unbindable(old_path.mnt)) | |
9676f0c6 | 1900 | goto out1; |
1da177e4 | 1901 | err = -ELOOP; |
b2dba1af | 1902 | for (p = path->mnt; mnt_has_parent(p); p = p->mnt_parent) |
2d92ab3c | 1903 | if (p == old_path.mnt) |
21444403 | 1904 | goto out1; |
1da177e4 | 1905 | |
2d92ab3c | 1906 | err = attach_recursive_mnt(old_path.mnt, path, &parent_path); |
4ac91378 | 1907 | if (err) |
21444403 | 1908 | goto out1; |
1da177e4 LT |
1909 | |
1910 | /* if the mount is moved, it should no longer be expire | |
1911 | * automatically */ | |
2d92ab3c | 1912 | list_del_init(&old_path.mnt->mnt_expire); |
1da177e4 | 1913 | out1: |
b12cea91 | 1914 | unlock_mount(path); |
1da177e4 | 1915 | out: |
1da177e4 | 1916 | if (!err) |
1a390689 | 1917 | path_put(&parent_path); |
2d92ab3c | 1918 | path_put(&old_path); |
1da177e4 LT |
1919 | return err; |
1920 | } | |
1921 | ||
9d412a43 AV |
1922 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
1923 | { | |
1924 | int err; | |
1925 | const char *subtype = strchr(fstype, '.'); | |
1926 | if (subtype) { | |
1927 | subtype++; | |
1928 | err = -EINVAL; | |
1929 | if (!subtype[0]) | |
1930 | goto err; | |
1931 | } else | |
1932 | subtype = ""; | |
1933 | ||
1934 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
1935 | err = -ENOMEM; | |
1936 | if (!mnt->mnt_sb->s_subtype) | |
1937 | goto err; | |
1938 | return mnt; | |
1939 | ||
1940 | err: | |
1941 | mntput(mnt); | |
1942 | return ERR_PTR(err); | |
1943 | } | |
1944 | ||
79e801a9 | 1945 | static struct vfsmount * |
9d412a43 AV |
1946 | do_kern_mount(const char *fstype, int flags, const char *name, void *data) |
1947 | { | |
1948 | struct file_system_type *type = get_fs_type(fstype); | |
1949 | struct vfsmount *mnt; | |
1950 | if (!type) | |
1951 | return ERR_PTR(-ENODEV); | |
1952 | mnt = vfs_kern_mount(type, flags, name, data); | |
1953 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && | |
1954 | !mnt->mnt_sb->s_subtype) | |
1955 | mnt = fs_set_subtype(mnt, fstype); | |
1956 | put_filesystem(type); | |
1957 | return mnt; | |
1958 | } | |
9d412a43 AV |
1959 | |
1960 | /* | |
1961 | * add a mount into a namespace's mount tree | |
1962 | */ | |
1963 | static int do_add_mount(struct vfsmount *newmnt, struct path *path, int mnt_flags) | |
1964 | { | |
1965 | int err; | |
1966 | ||
1967 | mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); | |
1968 | ||
b12cea91 AV |
1969 | err = lock_mount(path); |
1970 | if (err) | |
1971 | return err; | |
9d412a43 AV |
1972 | |
1973 | err = -EINVAL; | |
1974 | if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) | |
1975 | goto unlock; | |
1976 | ||
1977 | /* Refuse the same filesystem on the same mount point */ | |
1978 | err = -EBUSY; | |
1979 | if (path->mnt->mnt_sb == newmnt->mnt_sb && | |
1980 | path->mnt->mnt_root == path->dentry) | |
1981 | goto unlock; | |
1982 | ||
1983 | err = -EINVAL; | |
1984 | if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) | |
1985 | goto unlock; | |
1986 | ||
1987 | newmnt->mnt_flags = mnt_flags; | |
1988 | err = graft_tree(newmnt, path); | |
1989 | ||
1990 | unlock: | |
b12cea91 | 1991 | unlock_mount(path); |
9d412a43 AV |
1992 | return err; |
1993 | } | |
b1e75df4 | 1994 | |
1da177e4 LT |
1995 | /* |
1996 | * create a new mount for userspace and request it to be added into the | |
1997 | * namespace's tree | |
1998 | */ | |
0a0d8a46 | 1999 | static int do_new_mount(struct path *path, char *type, int flags, |
1da177e4 LT |
2000 | int mnt_flags, char *name, void *data) |
2001 | { | |
2002 | struct vfsmount *mnt; | |
15f9a3f3 | 2003 | int err; |
1da177e4 | 2004 | |
eca6f534 | 2005 | if (!type) |
1da177e4 LT |
2006 | return -EINVAL; |
2007 | ||
2008 | /* we need capabilities... */ | |
2009 | if (!capable(CAP_SYS_ADMIN)) | |
2010 | return -EPERM; | |
2011 | ||
2012 | mnt = do_kern_mount(type, flags, name, data); | |
2013 | if (IS_ERR(mnt)) | |
2014 | return PTR_ERR(mnt); | |
2015 | ||
15f9a3f3 AV |
2016 | err = do_add_mount(mnt, path, mnt_flags); |
2017 | if (err) | |
2018 | mntput(mnt); | |
2019 | return err; | |
1da177e4 LT |
2020 | } |
2021 | ||
19a167af AV |
2022 | int finish_automount(struct vfsmount *m, struct path *path) |
2023 | { | |
2024 | int err; | |
2025 | /* The new mount record should have at least 2 refs to prevent it being | |
2026 | * expired before we get a chance to add it | |
2027 | */ | |
2028 | BUG_ON(mnt_get_count(m) < 2); | |
2029 | ||
2030 | if (m->mnt_sb == path->mnt->mnt_sb && | |
2031 | m->mnt_root == path->dentry) { | |
b1e75df4 AV |
2032 | err = -ELOOP; |
2033 | goto fail; | |
19a167af AV |
2034 | } |
2035 | ||
19a167af | 2036 | err = do_add_mount(m, path, path->mnt->mnt_flags | MNT_SHRINKABLE); |
b1e75df4 AV |
2037 | if (!err) |
2038 | return 0; | |
2039 | fail: | |
2040 | /* remove m from any expiration list it may be on */ | |
2041 | if (!list_empty(&m->mnt_expire)) { | |
2042 | down_write(&namespace_sem); | |
2043 | br_write_lock(vfsmount_lock); | |
2044 | list_del_init(&m->mnt_expire); | |
2045 | br_write_unlock(vfsmount_lock); | |
2046 | up_write(&namespace_sem); | |
19a167af | 2047 | } |
b1e75df4 AV |
2048 | mntput(m); |
2049 | mntput(m); | |
19a167af AV |
2050 | return err; |
2051 | } | |
2052 | ||
ea5b778a DH |
2053 | /** |
2054 | * mnt_set_expiry - Put a mount on an expiration list | |
2055 | * @mnt: The mount to list. | |
2056 | * @expiry_list: The list to add the mount to. | |
2057 | */ | |
2058 | void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) | |
2059 | { | |
2060 | down_write(&namespace_sem); | |
2061 | br_write_lock(vfsmount_lock); | |
2062 | ||
2063 | list_add_tail(&mnt->mnt_expire, expiry_list); | |
2064 | ||
2065 | br_write_unlock(vfsmount_lock); | |
2066 | up_write(&namespace_sem); | |
2067 | } | |
2068 | EXPORT_SYMBOL(mnt_set_expiry); | |
2069 | ||
1da177e4 LT |
2070 | /* |
2071 | * process a list of expirable mountpoints with the intent of discarding any | |
2072 | * mountpoints that aren't in use and haven't been touched since last we came | |
2073 | * here | |
2074 | */ | |
2075 | void mark_mounts_for_expiry(struct list_head *mounts) | |
2076 | { | |
1da177e4 LT |
2077 | struct vfsmount *mnt, *next; |
2078 | LIST_HEAD(graveyard); | |
bcc5c7d2 | 2079 | LIST_HEAD(umounts); |
1da177e4 LT |
2080 | |
2081 | if (list_empty(mounts)) | |
2082 | return; | |
2083 | ||
bcc5c7d2 | 2084 | down_write(&namespace_sem); |
99b7db7b | 2085 | br_write_lock(vfsmount_lock); |
1da177e4 LT |
2086 | |
2087 | /* extract from the expiration list every vfsmount that matches the | |
2088 | * following criteria: | |
2089 | * - only referenced by its parent vfsmount | |
2090 | * - still marked for expiry (marked on the last call here; marks are | |
2091 | * cleared by mntput()) | |
2092 | */ | |
55e700b9 | 2093 | list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { |
1da177e4 | 2094 | if (!xchg(&mnt->mnt_expiry_mark, 1) || |
bcc5c7d2 | 2095 | propagate_mount_busy(mnt, 1)) |
1da177e4 | 2096 | continue; |
55e700b9 | 2097 | list_move(&mnt->mnt_expire, &graveyard); |
1da177e4 | 2098 | } |
bcc5c7d2 AV |
2099 | while (!list_empty(&graveyard)) { |
2100 | mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); | |
2101 | touch_mnt_namespace(mnt->mnt_ns); | |
2102 | umount_tree(mnt, 1, &umounts); | |
2103 | } | |
99b7db7b | 2104 | br_write_unlock(vfsmount_lock); |
bcc5c7d2 AV |
2105 | up_write(&namespace_sem); |
2106 | ||
2107 | release_mounts(&umounts); | |
5528f911 TM |
2108 | } |
2109 | ||
2110 | EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); | |
2111 | ||
2112 | /* | |
2113 | * Ripoff of 'select_parent()' | |
2114 | * | |
2115 | * search the list of submounts for a given mountpoint, and move any | |
2116 | * shrinkable submounts to the 'graveyard' list. | |
2117 | */ | |
2118 | static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) | |
2119 | { | |
2120 | struct vfsmount *this_parent = parent; | |
2121 | struct list_head *next; | |
2122 | int found = 0; | |
2123 | ||
2124 | repeat: | |
2125 | next = this_parent->mnt_mounts.next; | |
2126 | resume: | |
2127 | while (next != &this_parent->mnt_mounts) { | |
2128 | struct list_head *tmp = next; | |
2129 | struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); | |
2130 | ||
2131 | next = tmp->next; | |
2132 | if (!(mnt->mnt_flags & MNT_SHRINKABLE)) | |
1da177e4 | 2133 | continue; |
5528f911 TM |
2134 | /* |
2135 | * Descend a level if the d_mounts list is non-empty. | |
2136 | */ | |
2137 | if (!list_empty(&mnt->mnt_mounts)) { | |
2138 | this_parent = mnt; | |
2139 | goto repeat; | |
2140 | } | |
1da177e4 | 2141 | |
5528f911 | 2142 | if (!propagate_mount_busy(mnt, 1)) { |
5528f911 TM |
2143 | list_move_tail(&mnt->mnt_expire, graveyard); |
2144 | found++; | |
2145 | } | |
1da177e4 | 2146 | } |
5528f911 TM |
2147 | /* |
2148 | * All done at this level ... ascend and resume the search | |
2149 | */ | |
2150 | if (this_parent != parent) { | |
2151 | next = this_parent->mnt_child.next; | |
2152 | this_parent = this_parent->mnt_parent; | |
2153 | goto resume; | |
2154 | } | |
2155 | return found; | |
2156 | } | |
2157 | ||
2158 | /* | |
2159 | * process a list of expirable mountpoints with the intent of discarding any | |
2160 | * submounts of a specific parent mountpoint | |
99b7db7b NP |
2161 | * |
2162 | * vfsmount_lock must be held for write | |
5528f911 | 2163 | */ |
c35038be | 2164 | static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) |
5528f911 TM |
2165 | { |
2166 | LIST_HEAD(graveyard); | |
c35038be | 2167 | struct vfsmount *m; |
5528f911 | 2168 | |
5528f911 | 2169 | /* extract submounts of 'mountpoint' from the expiration list */ |
c35038be | 2170 | while (select_submounts(mnt, &graveyard)) { |
bcc5c7d2 | 2171 | while (!list_empty(&graveyard)) { |
c35038be | 2172 | m = list_first_entry(&graveyard, struct vfsmount, |
bcc5c7d2 | 2173 | mnt_expire); |
afef80b3 EB |
2174 | touch_mnt_namespace(m->mnt_ns); |
2175 | umount_tree(m, 1, umounts); | |
bcc5c7d2 AV |
2176 | } |
2177 | } | |
1da177e4 LT |
2178 | } |
2179 | ||
1da177e4 LT |
2180 | /* |
2181 | * Some copy_from_user() implementations do not return the exact number of | |
2182 | * bytes remaining to copy on a fault. But copy_mount_options() requires that. | |
2183 | * Note that this function differs from copy_from_user() in that it will oops | |
2184 | * on bad values of `to', rather than returning a short copy. | |
2185 | */ | |
b58fed8b RP |
2186 | static long exact_copy_from_user(void *to, const void __user * from, |
2187 | unsigned long n) | |
1da177e4 LT |
2188 | { |
2189 | char *t = to; | |
2190 | const char __user *f = from; | |
2191 | char c; | |
2192 | ||
2193 | if (!access_ok(VERIFY_READ, from, n)) | |
2194 | return n; | |
2195 | ||
2196 | while (n) { | |
2197 | if (__get_user(c, f)) { | |
2198 | memset(t, 0, n); | |
2199 | break; | |
2200 | } | |
2201 | *t++ = c; | |
2202 | f++; | |
2203 | n--; | |
2204 | } | |
2205 | return n; | |
2206 | } | |
2207 | ||
b58fed8b | 2208 | int copy_mount_options(const void __user * data, unsigned long *where) |
1da177e4 LT |
2209 | { |
2210 | int i; | |
2211 | unsigned long page; | |
2212 | unsigned long size; | |
b58fed8b | 2213 | |
1da177e4 LT |
2214 | *where = 0; |
2215 | if (!data) | |
2216 | return 0; | |
2217 | ||
2218 | if (!(page = __get_free_page(GFP_KERNEL))) | |
2219 | return -ENOMEM; | |
2220 | ||
2221 | /* We only care that *some* data at the address the user | |
2222 | * gave us is valid. Just in case, we'll zero | |
2223 | * the remainder of the page. | |
2224 | */ | |
2225 | /* copy_from_user cannot cross TASK_SIZE ! */ | |
2226 | size = TASK_SIZE - (unsigned long)data; | |
2227 | if (size > PAGE_SIZE) | |
2228 | size = PAGE_SIZE; | |
2229 | ||
2230 | i = size - exact_copy_from_user((void *)page, data, size); | |
2231 | if (!i) { | |
b58fed8b | 2232 | free_page(page); |
1da177e4 LT |
2233 | return -EFAULT; |
2234 | } | |
2235 | if (i != PAGE_SIZE) | |
2236 | memset((char *)page + i, 0, PAGE_SIZE - i); | |
2237 | *where = page; | |
2238 | return 0; | |
2239 | } | |
2240 | ||
eca6f534 VN |
2241 | int copy_mount_string(const void __user *data, char **where) |
2242 | { | |
2243 | char *tmp; | |
2244 | ||
2245 | if (!data) { | |
2246 | *where = NULL; | |
2247 | return 0; | |
2248 | } | |
2249 | ||
2250 | tmp = strndup_user(data, PAGE_SIZE); | |
2251 | if (IS_ERR(tmp)) | |
2252 | return PTR_ERR(tmp); | |
2253 | ||
2254 | *where = tmp; | |
2255 | return 0; | |
2256 | } | |
2257 | ||
1da177e4 LT |
2258 | /* |
2259 | * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to | |
2260 | * be given to the mount() call (ie: read-only, no-dev, no-suid etc). | |
2261 | * | |
2262 | * data is a (void *) that can point to any structure up to | |
2263 | * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent | |
2264 | * information (or be NULL). | |
2265 | * | |
2266 | * Pre-0.97 versions of mount() didn't have a flags word. | |
2267 | * When the flags word was introduced its top half was required | |
2268 | * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. | |
2269 | * Therefore, if this magic number is present, it carries no information | |
2270 | * and must be discarded. | |
2271 | */ | |
b58fed8b | 2272 | long do_mount(char *dev_name, char *dir_name, char *type_page, |
1da177e4 LT |
2273 | unsigned long flags, void *data_page) |
2274 | { | |
2d92ab3c | 2275 | struct path path; |
1da177e4 LT |
2276 | int retval = 0; |
2277 | int mnt_flags = 0; | |
2278 | ||
2279 | /* Discard magic */ | |
2280 | if ((flags & MS_MGC_MSK) == MS_MGC_VAL) | |
2281 | flags &= ~MS_MGC_MSK; | |
2282 | ||
2283 | /* Basic sanity checks */ | |
2284 | ||
2285 | if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) | |
2286 | return -EINVAL; | |
1da177e4 LT |
2287 | |
2288 | if (data_page) | |
2289 | ((char *)data_page)[PAGE_SIZE - 1] = 0; | |
2290 | ||
a27ab9f2 TH |
2291 | /* ... and get the mountpoint */ |
2292 | retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); | |
2293 | if (retval) | |
2294 | return retval; | |
2295 | ||
2296 | retval = security_sb_mount(dev_name, &path, | |
2297 | type_page, flags, data_page); | |
2298 | if (retval) | |
2299 | goto dput_out; | |
2300 | ||
613cbe3d AK |
2301 | /* Default to relatime unless overriden */ |
2302 | if (!(flags & MS_NOATIME)) | |
2303 | mnt_flags |= MNT_RELATIME; | |
0a1c01c9 | 2304 | |
1da177e4 LT |
2305 | /* Separate the per-mountpoint flags */ |
2306 | if (flags & MS_NOSUID) | |
2307 | mnt_flags |= MNT_NOSUID; | |
2308 | if (flags & MS_NODEV) | |
2309 | mnt_flags |= MNT_NODEV; | |
2310 | if (flags & MS_NOEXEC) | |
2311 | mnt_flags |= MNT_NOEXEC; | |
fc33a7bb CH |
2312 | if (flags & MS_NOATIME) |
2313 | mnt_flags |= MNT_NOATIME; | |
2314 | if (flags & MS_NODIRATIME) | |
2315 | mnt_flags |= MNT_NODIRATIME; | |
d0adde57 MG |
2316 | if (flags & MS_STRICTATIME) |
2317 | mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); | |
2e4b7fcd DH |
2318 | if (flags & MS_RDONLY) |
2319 | mnt_flags |= MNT_READONLY; | |
fc33a7bb | 2320 | |
7a4dec53 | 2321 | flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | |
d0adde57 MG |
2322 | MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | |
2323 | MS_STRICTATIME); | |
1da177e4 | 2324 | |
1da177e4 | 2325 | if (flags & MS_REMOUNT) |
2d92ab3c | 2326 | retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, |
1da177e4 LT |
2327 | data_page); |
2328 | else if (flags & MS_BIND) | |
2d92ab3c | 2329 | retval = do_loopback(&path, dev_name, flags & MS_REC); |
9676f0c6 | 2330 | else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) |
2d92ab3c | 2331 | retval = do_change_type(&path, flags); |
1da177e4 | 2332 | else if (flags & MS_MOVE) |
2d92ab3c | 2333 | retval = do_move_mount(&path, dev_name); |
1da177e4 | 2334 | else |
2d92ab3c | 2335 | retval = do_new_mount(&path, type_page, flags, mnt_flags, |
1da177e4 LT |
2336 | dev_name, data_page); |
2337 | dput_out: | |
2d92ab3c | 2338 | path_put(&path); |
1da177e4 LT |
2339 | return retval; |
2340 | } | |
2341 | ||
cf8d2c11 TM |
2342 | static struct mnt_namespace *alloc_mnt_ns(void) |
2343 | { | |
2344 | struct mnt_namespace *new_ns; | |
2345 | ||
2346 | new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); | |
2347 | if (!new_ns) | |
2348 | return ERR_PTR(-ENOMEM); | |
2349 | atomic_set(&new_ns->count, 1); | |
2350 | new_ns->root = NULL; | |
2351 | INIT_LIST_HEAD(&new_ns->list); | |
2352 | init_waitqueue_head(&new_ns->poll); | |
2353 | new_ns->event = 0; | |
2354 | return new_ns; | |
2355 | } | |
2356 | ||
f03c6599 AV |
2357 | void mnt_make_longterm(struct vfsmount *mnt) |
2358 | { | |
7e3d0eb0 | 2359 | __mnt_make_longterm(mnt); |
f03c6599 AV |
2360 | } |
2361 | ||
2362 | void mnt_make_shortterm(struct vfsmount *mnt) | |
2363 | { | |
7e3d0eb0 | 2364 | #ifdef CONFIG_SMP |
f03c6599 AV |
2365 | if (atomic_add_unless(&mnt->mnt_longterm, -1, 1)) |
2366 | return; | |
2367 | br_write_lock(vfsmount_lock); | |
2368 | atomic_dec(&mnt->mnt_longterm); | |
2369 | br_write_unlock(vfsmount_lock); | |
7e3d0eb0 | 2370 | #endif |
f03c6599 AV |
2371 | } |
2372 | ||
741a2951 JD |
2373 | /* |
2374 | * Allocate a new namespace structure and populate it with contents | |
2375 | * copied from the namespace of the passed in task structure. | |
2376 | */ | |
e3222c4e | 2377 | static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, |
6b3286ed | 2378 | struct fs_struct *fs) |
1da177e4 | 2379 | { |
6b3286ed | 2380 | struct mnt_namespace *new_ns; |
7f2da1e7 | 2381 | struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; |
1da177e4 LT |
2382 | struct vfsmount *p, *q; |
2383 | ||
cf8d2c11 TM |
2384 | new_ns = alloc_mnt_ns(); |
2385 | if (IS_ERR(new_ns)) | |
2386 | return new_ns; | |
1da177e4 | 2387 | |
390c6843 | 2388 | down_write(&namespace_sem); |
1da177e4 | 2389 | /* First pass: copy the tree topology */ |
6b3286ed | 2390 | new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, |
9676f0c6 | 2391 | CL_COPY_ALL | CL_EXPIRE); |
1da177e4 | 2392 | if (!new_ns->root) { |
390c6843 | 2393 | up_write(&namespace_sem); |
1da177e4 | 2394 | kfree(new_ns); |
5cc4a034 | 2395 | return ERR_PTR(-ENOMEM); |
1da177e4 | 2396 | } |
99b7db7b | 2397 | br_write_lock(vfsmount_lock); |
1da177e4 | 2398 | list_add_tail(&new_ns->list, &new_ns->root->mnt_list); |
99b7db7b | 2399 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
2400 | |
2401 | /* | |
2402 | * Second pass: switch the tsk->fs->* elements and mark new vfsmounts | |
2403 | * as belonging to new namespace. We have already acquired a private | |
2404 | * fs_struct, so tsk->fs->lock is not needed. | |
2405 | */ | |
6b3286ed | 2406 | p = mnt_ns->root; |
1da177e4 LT |
2407 | q = new_ns->root; |
2408 | while (p) { | |
6b3286ed | 2409 | q->mnt_ns = new_ns; |
7e3d0eb0 | 2410 | __mnt_make_longterm(q); |
1da177e4 | 2411 | if (fs) { |
6ac08c39 | 2412 | if (p == fs->root.mnt) { |
f03c6599 | 2413 | fs->root.mnt = mntget(q); |
7e3d0eb0 | 2414 | __mnt_make_longterm(q); |
f03c6599 | 2415 | mnt_make_shortterm(p); |
1da177e4 | 2416 | rootmnt = p; |
1da177e4 | 2417 | } |
6ac08c39 | 2418 | if (p == fs->pwd.mnt) { |
f03c6599 | 2419 | fs->pwd.mnt = mntget(q); |
7e3d0eb0 | 2420 | __mnt_make_longterm(q); |
f03c6599 | 2421 | mnt_make_shortterm(p); |
1da177e4 | 2422 | pwdmnt = p; |
1da177e4 | 2423 | } |
1da177e4 | 2424 | } |
6b3286ed | 2425 | p = next_mnt(p, mnt_ns->root); |
1da177e4 LT |
2426 | q = next_mnt(q, new_ns->root); |
2427 | } | |
390c6843 | 2428 | up_write(&namespace_sem); |
1da177e4 | 2429 | |
1da177e4 | 2430 | if (rootmnt) |
f03c6599 | 2431 | mntput(rootmnt); |
1da177e4 | 2432 | if (pwdmnt) |
f03c6599 | 2433 | mntput(pwdmnt); |
1da177e4 | 2434 | |
741a2951 JD |
2435 | return new_ns; |
2436 | } | |
2437 | ||
213dd266 | 2438 | struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, |
e3222c4e | 2439 | struct fs_struct *new_fs) |
741a2951 | 2440 | { |
6b3286ed | 2441 | struct mnt_namespace *new_ns; |
741a2951 | 2442 | |
e3222c4e | 2443 | BUG_ON(!ns); |
6b3286ed | 2444 | get_mnt_ns(ns); |
741a2951 JD |
2445 | |
2446 | if (!(flags & CLONE_NEWNS)) | |
e3222c4e | 2447 | return ns; |
741a2951 | 2448 | |
e3222c4e | 2449 | new_ns = dup_mnt_ns(ns, new_fs); |
741a2951 | 2450 | |
6b3286ed | 2451 | put_mnt_ns(ns); |
e3222c4e | 2452 | return new_ns; |
1da177e4 LT |
2453 | } |
2454 | ||
cf8d2c11 TM |
2455 | /** |
2456 | * create_mnt_ns - creates a private namespace and adds a root filesystem | |
2457 | * @mnt: pointer to the new root filesystem mountpoint | |
2458 | */ | |
6c449c8d | 2459 | static struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) |
cf8d2c11 TM |
2460 | { |
2461 | struct mnt_namespace *new_ns; | |
2462 | ||
2463 | new_ns = alloc_mnt_ns(); | |
2464 | if (!IS_ERR(new_ns)) { | |
2465 | mnt->mnt_ns = new_ns; | |
7e3d0eb0 | 2466 | __mnt_make_longterm(mnt); |
cf8d2c11 TM |
2467 | new_ns->root = mnt; |
2468 | list_add(&new_ns->list, &new_ns->root->mnt_list); | |
c1334495 AV |
2469 | } else { |
2470 | mntput(mnt); | |
cf8d2c11 TM |
2471 | } |
2472 | return new_ns; | |
2473 | } | |
cf8d2c11 | 2474 | |
ea441d11 AV |
2475 | struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) |
2476 | { | |
2477 | struct mnt_namespace *ns; | |
d31da0f0 | 2478 | struct super_block *s; |
ea441d11 AV |
2479 | struct path path; |
2480 | int err; | |
2481 | ||
2482 | ns = create_mnt_ns(mnt); | |
2483 | if (IS_ERR(ns)) | |
2484 | return ERR_CAST(ns); | |
2485 | ||
2486 | err = vfs_path_lookup(mnt->mnt_root, mnt, | |
2487 | name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); | |
2488 | ||
2489 | put_mnt_ns(ns); | |
2490 | ||
2491 | if (err) | |
2492 | return ERR_PTR(err); | |
2493 | ||
2494 | /* trade a vfsmount reference for active sb one */ | |
d31da0f0 AV |
2495 | s = path.mnt->mnt_sb; |
2496 | atomic_inc(&s->s_active); | |
ea441d11 AV |
2497 | mntput(path.mnt); |
2498 | /* lock the sucker */ | |
d31da0f0 | 2499 | down_write(&s->s_umount); |
ea441d11 AV |
2500 | /* ... and return the root of (sub)tree on it */ |
2501 | return path.dentry; | |
2502 | } | |
2503 | EXPORT_SYMBOL(mount_subtree); | |
2504 | ||
bdc480e3 HC |
2505 | SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, |
2506 | char __user *, type, unsigned long, flags, void __user *, data) | |
1da177e4 | 2507 | { |
eca6f534 VN |
2508 | int ret; |
2509 | char *kernel_type; | |
2510 | char *kernel_dir; | |
2511 | char *kernel_dev; | |
1da177e4 | 2512 | unsigned long data_page; |
1da177e4 | 2513 | |
eca6f534 VN |
2514 | ret = copy_mount_string(type, &kernel_type); |
2515 | if (ret < 0) | |
2516 | goto out_type; | |
1da177e4 | 2517 | |
eca6f534 VN |
2518 | kernel_dir = getname(dir_name); |
2519 | if (IS_ERR(kernel_dir)) { | |
2520 | ret = PTR_ERR(kernel_dir); | |
2521 | goto out_dir; | |
2522 | } | |
1da177e4 | 2523 | |
eca6f534 VN |
2524 | ret = copy_mount_string(dev_name, &kernel_dev); |
2525 | if (ret < 0) | |
2526 | goto out_dev; | |
1da177e4 | 2527 | |
eca6f534 VN |
2528 | ret = copy_mount_options(data, &data_page); |
2529 | if (ret < 0) | |
2530 | goto out_data; | |
1da177e4 | 2531 | |
eca6f534 VN |
2532 | ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags, |
2533 | (void *) data_page); | |
1da177e4 | 2534 | |
eca6f534 VN |
2535 | free_page(data_page); |
2536 | out_data: | |
2537 | kfree(kernel_dev); | |
2538 | out_dev: | |
2539 | putname(kernel_dir); | |
2540 | out_dir: | |
2541 | kfree(kernel_type); | |
2542 | out_type: | |
2543 | return ret; | |
1da177e4 LT |
2544 | } |
2545 | ||
afac7cba AV |
2546 | /* |
2547 | * Return true if path is reachable from root | |
2548 | * | |
2549 | * namespace_sem or vfsmount_lock is held | |
2550 | */ | |
2551 | bool is_path_reachable(struct vfsmount *mnt, struct dentry *dentry, | |
2552 | const struct path *root) | |
2553 | { | |
2554 | while (mnt != root->mnt && mnt_has_parent(mnt)) { | |
2555 | dentry = mnt->mnt_mountpoint; | |
2556 | mnt = mnt->mnt_parent; | |
2557 | } | |
2558 | return mnt == root->mnt && is_subdir(dentry, root->dentry); | |
2559 | } | |
2560 | ||
2561 | int path_is_under(struct path *path1, struct path *path2) | |
2562 | { | |
2563 | int res; | |
2564 | br_read_lock(vfsmount_lock); | |
2565 | res = is_path_reachable(path1->mnt, path1->dentry, path2); | |
2566 | br_read_unlock(vfsmount_lock); | |
2567 | return res; | |
2568 | } | |
2569 | EXPORT_SYMBOL(path_is_under); | |
2570 | ||
1da177e4 LT |
2571 | /* |
2572 | * pivot_root Semantics: | |
2573 | * Moves the root file system of the current process to the directory put_old, | |
2574 | * makes new_root as the new root file system of the current process, and sets | |
2575 | * root/cwd of all processes which had them on the current root to new_root. | |
2576 | * | |
2577 | * Restrictions: | |
2578 | * The new_root and put_old must be directories, and must not be on the | |
2579 | * same file system as the current process root. The put_old must be | |
2580 | * underneath new_root, i.e. adding a non-zero number of /.. to the string | |
2581 | * pointed to by put_old must yield the same directory as new_root. No other | |
2582 | * file system may be mounted on put_old. After all, new_root is a mountpoint. | |
2583 | * | |
4a0d11fa NB |
2584 | * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. |
2585 | * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives | |
2586 | * in this situation. | |
2587 | * | |
1da177e4 LT |
2588 | * Notes: |
2589 | * - we don't move root/cwd if they are not at the root (reason: if something | |
2590 | * cared enough to change them, it's probably wrong to force them elsewhere) | |
2591 | * - it's okay to pick a root that isn't the root of a file system, e.g. | |
2592 | * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, | |
2593 | * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root | |
2594 | * first. | |
2595 | */ | |
3480b257 HC |
2596 | SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, |
2597 | const char __user *, put_old) | |
1da177e4 | 2598 | { |
2d8f3038 | 2599 | struct path new, old, parent_path, root_parent, root; |
1da177e4 LT |
2600 | int error; |
2601 | ||
2602 | if (!capable(CAP_SYS_ADMIN)) | |
2603 | return -EPERM; | |
2604 | ||
2d8f3038 | 2605 | error = user_path_dir(new_root, &new); |
1da177e4 LT |
2606 | if (error) |
2607 | goto out0; | |
1da177e4 | 2608 | |
2d8f3038 | 2609 | error = user_path_dir(put_old, &old); |
1da177e4 LT |
2610 | if (error) |
2611 | goto out1; | |
2612 | ||
2d8f3038 | 2613 | error = security_sb_pivotroot(&old, &new); |
b12cea91 AV |
2614 | if (error) |
2615 | goto out2; | |
1da177e4 | 2616 | |
f7ad3c6b | 2617 | get_fs_root(current->fs, &root); |
b12cea91 AV |
2618 | error = lock_mount(&old); |
2619 | if (error) | |
2620 | goto out3; | |
2621 | ||
1da177e4 | 2622 | error = -EINVAL; |
2d8f3038 AV |
2623 | if (IS_MNT_SHARED(old.mnt) || |
2624 | IS_MNT_SHARED(new.mnt->mnt_parent) || | |
8c3ee42e | 2625 | IS_MNT_SHARED(root.mnt->mnt_parent)) |
b12cea91 | 2626 | goto out4; |
27cb1572 | 2627 | if (!check_mnt(root.mnt) || !check_mnt(new.mnt)) |
b12cea91 | 2628 | goto out4; |
1da177e4 | 2629 | error = -ENOENT; |
f3da392e | 2630 | if (d_unlinked(new.dentry)) |
b12cea91 | 2631 | goto out4; |
f3da392e | 2632 | if (d_unlinked(old.dentry)) |
b12cea91 | 2633 | goto out4; |
1da177e4 | 2634 | error = -EBUSY; |
2d8f3038 AV |
2635 | if (new.mnt == root.mnt || |
2636 | old.mnt == root.mnt) | |
b12cea91 | 2637 | goto out4; /* loop, on the same file system */ |
1da177e4 | 2638 | error = -EINVAL; |
8c3ee42e | 2639 | if (root.mnt->mnt_root != root.dentry) |
b12cea91 | 2640 | goto out4; /* not a mountpoint */ |
b2dba1af | 2641 | if (!mnt_has_parent(root.mnt)) |
b12cea91 | 2642 | goto out4; /* not attached */ |
2d8f3038 | 2643 | if (new.mnt->mnt_root != new.dentry) |
b12cea91 | 2644 | goto out4; /* not a mountpoint */ |
b2dba1af | 2645 | if (!mnt_has_parent(new.mnt)) |
b12cea91 | 2646 | goto out4; /* not attached */ |
4ac91378 | 2647 | /* make sure we can reach put_old from new_root */ |
afac7cba | 2648 | if (!is_path_reachable(old.mnt, old.dentry, &new)) |
b12cea91 | 2649 | goto out4; |
27cb1572 | 2650 | br_write_lock(vfsmount_lock); |
2d8f3038 | 2651 | detach_mnt(new.mnt, &parent_path); |
8c3ee42e | 2652 | detach_mnt(root.mnt, &root_parent); |
4ac91378 | 2653 | /* mount old root on put_old */ |
2d8f3038 | 2654 | attach_mnt(root.mnt, &old); |
4ac91378 | 2655 | /* mount new_root on / */ |
2d8f3038 | 2656 | attach_mnt(new.mnt, &root_parent); |
6b3286ed | 2657 | touch_mnt_namespace(current->nsproxy->mnt_ns); |
99b7db7b | 2658 | br_write_unlock(vfsmount_lock); |
2d8f3038 | 2659 | chroot_fs_refs(&root, &new); |
1da177e4 | 2660 | error = 0; |
b12cea91 AV |
2661 | out4: |
2662 | unlock_mount(&old); | |
2663 | if (!error) { | |
2664 | path_put(&root_parent); | |
2665 | path_put(&parent_path); | |
2666 | } | |
2667 | out3: | |
8c3ee42e | 2668 | path_put(&root); |
b12cea91 | 2669 | out2: |
2d8f3038 | 2670 | path_put(&old); |
1da177e4 | 2671 | out1: |
2d8f3038 | 2672 | path_put(&new); |
1da177e4 | 2673 | out0: |
1da177e4 | 2674 | return error; |
1da177e4 LT |
2675 | } |
2676 | ||
2677 | static void __init init_mount_tree(void) | |
2678 | { | |
2679 | struct vfsmount *mnt; | |
6b3286ed | 2680 | struct mnt_namespace *ns; |
ac748a09 | 2681 | struct path root; |
1da177e4 LT |
2682 | |
2683 | mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); | |
2684 | if (IS_ERR(mnt)) | |
2685 | panic("Can't create rootfs"); | |
b3e19d92 | 2686 | |
3b22edc5 TM |
2687 | ns = create_mnt_ns(mnt); |
2688 | if (IS_ERR(ns)) | |
1da177e4 | 2689 | panic("Can't allocate initial namespace"); |
6b3286ed KK |
2690 | |
2691 | init_task.nsproxy->mnt_ns = ns; | |
2692 | get_mnt_ns(ns); | |
2693 | ||
ac748a09 JB |
2694 | root.mnt = ns->root; |
2695 | root.dentry = ns->root->mnt_root; | |
2696 | ||
2697 | set_fs_pwd(current->fs, &root); | |
2698 | set_fs_root(current->fs, &root); | |
1da177e4 LT |
2699 | } |
2700 | ||
74bf17cf | 2701 | void __init mnt_init(void) |
1da177e4 | 2702 | { |
13f14b4d | 2703 | unsigned u; |
15a67dd8 | 2704 | int err; |
1da177e4 | 2705 | |
390c6843 RP |
2706 | init_rwsem(&namespace_sem); |
2707 | ||
7d6fec45 | 2708 | mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount), |
20c2df83 | 2709 | 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); |
1da177e4 | 2710 | |
b58fed8b | 2711 | mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
1da177e4 LT |
2712 | |
2713 | if (!mount_hashtable) | |
2714 | panic("Failed to allocate mount hash table\n"); | |
2715 | ||
80cdc6da | 2716 | printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); |
13f14b4d ED |
2717 | |
2718 | for (u = 0; u < HASH_SIZE; u++) | |
2719 | INIT_LIST_HEAD(&mount_hashtable[u]); | |
1da177e4 | 2720 | |
99b7db7b NP |
2721 | br_lock_init(vfsmount_lock); |
2722 | ||
15a67dd8 RD |
2723 | err = sysfs_init(); |
2724 | if (err) | |
2725 | printk(KERN_WARNING "%s: sysfs_init error: %d\n", | |
8e24eea7 | 2726 | __func__, err); |
00d26666 GKH |
2727 | fs_kobj = kobject_create_and_add("fs", NULL); |
2728 | if (!fs_kobj) | |
8e24eea7 | 2729 | printk(KERN_WARNING "%s: kobj create error\n", __func__); |
1da177e4 LT |
2730 | init_rootfs(); |
2731 | init_mount_tree(); | |
2732 | } | |
2733 | ||
616511d0 | 2734 | void put_mnt_ns(struct mnt_namespace *ns) |
1da177e4 | 2735 | { |
70fbcdf4 | 2736 | LIST_HEAD(umount_list); |
616511d0 | 2737 | |
d498b25a | 2738 | if (!atomic_dec_and_test(&ns->count)) |
616511d0 | 2739 | return; |
390c6843 | 2740 | down_write(&namespace_sem); |
99b7db7b | 2741 | br_write_lock(vfsmount_lock); |
d498b25a | 2742 | umount_tree(ns->root, 0, &umount_list); |
99b7db7b | 2743 | br_write_unlock(vfsmount_lock); |
390c6843 | 2744 | up_write(&namespace_sem); |
70fbcdf4 | 2745 | release_mounts(&umount_list); |
6b3286ed | 2746 | kfree(ns); |
1da177e4 | 2747 | } |
9d412a43 AV |
2748 | |
2749 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) | |
2750 | { | |
423e0ab0 TC |
2751 | struct vfsmount *mnt; |
2752 | mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); | |
2753 | if (!IS_ERR(mnt)) { | |
2754 | /* | |
2755 | * it is a longterm mount, don't release mnt until | |
2756 | * we unmount before file sys is unregistered | |
2757 | */ | |
2758 | mnt_make_longterm(mnt); | |
2759 | } | |
2760 | return mnt; | |
9d412a43 AV |
2761 | } |
2762 | EXPORT_SYMBOL_GPL(kern_mount_data); | |
423e0ab0 TC |
2763 | |
2764 | void kern_unmount(struct vfsmount *mnt) | |
2765 | { | |
2766 | /* release long term mount so mount point can be released */ | |
2767 | if (!IS_ERR_OR_NULL(mnt)) { | |
2768 | mnt_make_shortterm(mnt); | |
2769 | mntput(mnt); | |
2770 | } | |
2771 | } | |
2772 | EXPORT_SYMBOL(kern_unmount); | |
02125a82 AV |
2773 | |
2774 | bool our_mnt(struct vfsmount *mnt) | |
2775 | { | |
2776 | return check_mnt(mnt); | |
2777 | } |