]>
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 | { |
c3762229 | 176 | struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); |
1da177e4 | 177 | if (mnt) { |
73cd49ec MS |
178 | int err; |
179 | ||
180 | err = mnt_alloc_id(mnt); | |
88b38782 LZ |
181 | if (err) |
182 | goto out_free_cache; | |
183 | ||
184 | if (name) { | |
185 | mnt->mnt_devname = kstrdup(name, GFP_KERNEL); | |
186 | if (!mnt->mnt_devname) | |
187 | goto out_free_id; | |
73cd49ec MS |
188 | } |
189 | ||
b3e19d92 NP |
190 | #ifdef CONFIG_SMP |
191 | mnt->mnt_pcp = alloc_percpu(struct mnt_pcp); | |
192 | if (!mnt->mnt_pcp) | |
193 | goto out_free_devname; | |
194 | ||
f03c6599 | 195 | this_cpu_add(mnt->mnt_pcp->mnt_count, 1); |
b3e19d92 NP |
196 | #else |
197 | mnt->mnt_count = 1; | |
198 | mnt->mnt_writers = 0; | |
199 | #endif | |
200 | ||
1da177e4 LT |
201 | INIT_LIST_HEAD(&mnt->mnt_hash); |
202 | INIT_LIST_HEAD(&mnt->mnt_child); | |
203 | INIT_LIST_HEAD(&mnt->mnt_mounts); | |
204 | INIT_LIST_HEAD(&mnt->mnt_list); | |
55e700b9 | 205 | INIT_LIST_HEAD(&mnt->mnt_expire); |
03e06e68 | 206 | INIT_LIST_HEAD(&mnt->mnt_share); |
a58b0eb8 RP |
207 | INIT_LIST_HEAD(&mnt->mnt_slave_list); |
208 | INIT_LIST_HEAD(&mnt->mnt_slave); | |
2504c5d6 AG |
209 | #ifdef CONFIG_FSNOTIFY |
210 | INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks); | |
d3ef3d73 | 211 | #endif |
1da177e4 LT |
212 | } |
213 | return mnt; | |
88b38782 | 214 | |
d3ef3d73 | 215 | #ifdef CONFIG_SMP |
216 | out_free_devname: | |
217 | kfree(mnt->mnt_devname); | |
218 | #endif | |
88b38782 LZ |
219 | out_free_id: |
220 | mnt_free_id(mnt); | |
221 | out_free_cache: | |
222 | kmem_cache_free(mnt_cache, mnt); | |
223 | return NULL; | |
1da177e4 LT |
224 | } |
225 | ||
3d733633 DH |
226 | /* |
227 | * Most r/o checks on a fs are for operations that take | |
228 | * discrete amounts of time, like a write() or unlink(). | |
229 | * We must keep track of when those operations start | |
230 | * (for permission checks) and when they end, so that | |
231 | * we can determine when writes are able to occur to | |
232 | * a filesystem. | |
233 | */ | |
234 | /* | |
235 | * __mnt_is_readonly: check whether a mount is read-only | |
236 | * @mnt: the mount to check for its write status | |
237 | * | |
238 | * This shouldn't be used directly ouside of the VFS. | |
239 | * It does not guarantee that the filesystem will stay | |
240 | * r/w, just that it is right *now*. This can not and | |
241 | * should not be used in place of IS_RDONLY(inode). | |
242 | * mnt_want/drop_write() will _keep_ the filesystem | |
243 | * r/w. | |
244 | */ | |
245 | int __mnt_is_readonly(struct vfsmount *mnt) | |
246 | { | |
2e4b7fcd DH |
247 | if (mnt->mnt_flags & MNT_READONLY) |
248 | return 1; | |
249 | if (mnt->mnt_sb->s_flags & MS_RDONLY) | |
250 | return 1; | |
251 | return 0; | |
3d733633 DH |
252 | } |
253 | EXPORT_SYMBOL_GPL(__mnt_is_readonly); | |
254 | ||
c6653a83 | 255 | static inline void mnt_inc_writers(struct vfsmount *mnt) |
d3ef3d73 | 256 | { |
257 | #ifdef CONFIG_SMP | |
b3e19d92 | 258 | this_cpu_inc(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 | 259 | #else |
260 | mnt->mnt_writers++; | |
261 | #endif | |
262 | } | |
3d733633 | 263 | |
c6653a83 | 264 | static inline void mnt_dec_writers(struct vfsmount *mnt) |
3d733633 | 265 | { |
d3ef3d73 | 266 | #ifdef CONFIG_SMP |
b3e19d92 | 267 | this_cpu_dec(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 | 268 | #else |
269 | mnt->mnt_writers--; | |
270 | #endif | |
3d733633 | 271 | } |
3d733633 | 272 | |
c6653a83 | 273 | static unsigned int mnt_get_writers(struct vfsmount *mnt) |
3d733633 | 274 | { |
d3ef3d73 | 275 | #ifdef CONFIG_SMP |
276 | unsigned int count = 0; | |
3d733633 | 277 | int cpu; |
3d733633 DH |
278 | |
279 | for_each_possible_cpu(cpu) { | |
b3e19d92 | 280 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers; |
3d733633 | 281 | } |
3d733633 | 282 | |
d3ef3d73 | 283 | return count; |
284 | #else | |
285 | return mnt->mnt_writers; | |
286 | #endif | |
3d733633 DH |
287 | } |
288 | ||
8366025e DH |
289 | /* |
290 | * Most r/o checks on a fs are for operations that take | |
291 | * discrete amounts of time, like a write() or unlink(). | |
292 | * We must keep track of when those operations start | |
293 | * (for permission checks) and when they end, so that | |
294 | * we can determine when writes are able to occur to | |
295 | * a filesystem. | |
296 | */ | |
297 | /** | |
298 | * mnt_want_write - get write access to a mount | |
299 | * @mnt: the mount on which to take a write | |
300 | * | |
301 | * This tells the low-level filesystem that a write is | |
302 | * about to be performed to it, and makes sure that | |
303 | * writes are allowed before returning success. When | |
304 | * the write operation is finished, mnt_drop_write() | |
305 | * must be called. This is effectively a refcount. | |
306 | */ | |
307 | int mnt_want_write(struct vfsmount *mnt) | |
308 | { | |
3d733633 | 309 | int ret = 0; |
3d733633 | 310 | |
d3ef3d73 | 311 | preempt_disable(); |
c6653a83 | 312 | mnt_inc_writers(mnt); |
d3ef3d73 | 313 | /* |
c6653a83 | 314 | * The store to mnt_inc_writers must be visible before we pass |
d3ef3d73 | 315 | * MNT_WRITE_HOLD loop below, so that the slowpath can see our |
316 | * incremented count after it has set MNT_WRITE_HOLD. | |
317 | */ | |
318 | smp_mb(); | |
319 | while (mnt->mnt_flags & MNT_WRITE_HOLD) | |
320 | cpu_relax(); | |
321 | /* | |
322 | * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will | |
323 | * be set to match its requirements. So we must not load that until | |
324 | * MNT_WRITE_HOLD is cleared. | |
325 | */ | |
326 | smp_rmb(); | |
3d733633 | 327 | if (__mnt_is_readonly(mnt)) { |
c6653a83 | 328 | mnt_dec_writers(mnt); |
3d733633 DH |
329 | ret = -EROFS; |
330 | goto out; | |
331 | } | |
3d733633 | 332 | out: |
d3ef3d73 | 333 | preempt_enable(); |
3d733633 | 334 | return ret; |
8366025e DH |
335 | } |
336 | EXPORT_SYMBOL_GPL(mnt_want_write); | |
337 | ||
96029c4e | 338 | /** |
339 | * mnt_clone_write - get write access to a mount | |
340 | * @mnt: the mount on which to take a write | |
341 | * | |
342 | * This is effectively like mnt_want_write, except | |
343 | * it must only be used to take an extra write reference | |
344 | * on a mountpoint that we already know has a write reference | |
345 | * on it. This allows some optimisation. | |
346 | * | |
347 | * After finished, mnt_drop_write must be called as usual to | |
348 | * drop the reference. | |
349 | */ | |
350 | int mnt_clone_write(struct vfsmount *mnt) | |
351 | { | |
352 | /* superblock may be r/o */ | |
353 | if (__mnt_is_readonly(mnt)) | |
354 | return -EROFS; | |
355 | preempt_disable(); | |
c6653a83 | 356 | mnt_inc_writers(mnt); |
96029c4e | 357 | preempt_enable(); |
358 | return 0; | |
359 | } | |
360 | EXPORT_SYMBOL_GPL(mnt_clone_write); | |
361 | ||
362 | /** | |
363 | * mnt_want_write_file - get write access to a file's mount | |
364 | * @file: the file who's mount on which to take a write | |
365 | * | |
366 | * This is like mnt_want_write, but it takes a file and can | |
367 | * do some optimisations if the file is open for write already | |
368 | */ | |
369 | int mnt_want_write_file(struct file *file) | |
370 | { | |
2d8dd38a OH |
371 | struct inode *inode = file->f_dentry->d_inode; |
372 | if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) | |
96029c4e | 373 | return mnt_want_write(file->f_path.mnt); |
374 | else | |
375 | return mnt_clone_write(file->f_path.mnt); | |
376 | } | |
377 | EXPORT_SYMBOL_GPL(mnt_want_write_file); | |
378 | ||
8366025e DH |
379 | /** |
380 | * mnt_drop_write - give up write access to a mount | |
381 | * @mnt: the mount on which to give up write access | |
382 | * | |
383 | * Tells the low-level filesystem that we are done | |
384 | * performing writes to it. Must be matched with | |
385 | * mnt_want_write() call above. | |
386 | */ | |
387 | void mnt_drop_write(struct vfsmount *mnt) | |
388 | { | |
d3ef3d73 | 389 | preempt_disable(); |
c6653a83 | 390 | mnt_dec_writers(mnt); |
d3ef3d73 | 391 | preempt_enable(); |
8366025e DH |
392 | } |
393 | EXPORT_SYMBOL_GPL(mnt_drop_write); | |
394 | ||
2e4b7fcd | 395 | static int mnt_make_readonly(struct vfsmount *mnt) |
8366025e | 396 | { |
3d733633 DH |
397 | int ret = 0; |
398 | ||
99b7db7b | 399 | br_write_lock(vfsmount_lock); |
d3ef3d73 | 400 | mnt->mnt_flags |= MNT_WRITE_HOLD; |
3d733633 | 401 | /* |
d3ef3d73 | 402 | * After storing MNT_WRITE_HOLD, we'll read the counters. This store |
403 | * should be visible before we do. | |
3d733633 | 404 | */ |
d3ef3d73 | 405 | smp_mb(); |
406 | ||
3d733633 | 407 | /* |
d3ef3d73 | 408 | * With writers on hold, if this value is zero, then there are |
409 | * definitely no active writers (although held writers may subsequently | |
410 | * increment the count, they'll have to wait, and decrement it after | |
411 | * seeing MNT_READONLY). | |
412 | * | |
413 | * It is OK to have counter incremented on one CPU and decremented on | |
414 | * another: the sum will add up correctly. The danger would be when we | |
415 | * sum up each counter, if we read a counter before it is incremented, | |
416 | * but then read another CPU's count which it has been subsequently | |
417 | * decremented from -- we would see more decrements than we should. | |
418 | * MNT_WRITE_HOLD protects against this scenario, because | |
419 | * mnt_want_write first increments count, then smp_mb, then spins on | |
420 | * MNT_WRITE_HOLD, so it can't be decremented by another CPU while | |
421 | * we're counting up here. | |
3d733633 | 422 | */ |
c6653a83 | 423 | if (mnt_get_writers(mnt) > 0) |
d3ef3d73 | 424 | ret = -EBUSY; |
425 | else | |
2e4b7fcd | 426 | mnt->mnt_flags |= MNT_READONLY; |
d3ef3d73 | 427 | /* |
428 | * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers | |
429 | * that become unheld will see MNT_READONLY. | |
430 | */ | |
431 | smp_wmb(); | |
432 | mnt->mnt_flags &= ~MNT_WRITE_HOLD; | |
99b7db7b | 433 | br_write_unlock(vfsmount_lock); |
3d733633 | 434 | return ret; |
8366025e | 435 | } |
8366025e | 436 | |
2e4b7fcd DH |
437 | static void __mnt_unmake_readonly(struct vfsmount *mnt) |
438 | { | |
99b7db7b | 439 | br_write_lock(vfsmount_lock); |
2e4b7fcd | 440 | mnt->mnt_flags &= ~MNT_READONLY; |
99b7db7b | 441 | br_write_unlock(vfsmount_lock); |
2e4b7fcd DH |
442 | } |
443 | ||
9d412a43 | 444 | static void free_vfsmnt(struct vfsmount *mnt) |
1da177e4 LT |
445 | { |
446 | kfree(mnt->mnt_devname); | |
73cd49ec | 447 | mnt_free_id(mnt); |
d3ef3d73 | 448 | #ifdef CONFIG_SMP |
b3e19d92 | 449 | free_percpu(mnt->mnt_pcp); |
d3ef3d73 | 450 | #endif |
1da177e4 LT |
451 | kmem_cache_free(mnt_cache, mnt); |
452 | } | |
453 | ||
454 | /* | |
a05964f3 RP |
455 | * find the first or last mount at @dentry on vfsmount @mnt depending on |
456 | * @dir. If @dir is set return the first mount else return the last mount. | |
99b7db7b | 457 | * vfsmount_lock must be held for read or write. |
1da177e4 | 458 | */ |
a05964f3 RP |
459 | struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, |
460 | int dir) | |
1da177e4 | 461 | { |
b58fed8b RP |
462 | struct list_head *head = mount_hashtable + hash(mnt, dentry); |
463 | struct list_head *tmp = head; | |
1da177e4 LT |
464 | struct vfsmount *p, *found = NULL; |
465 | ||
1da177e4 | 466 | for (;;) { |
a05964f3 | 467 | tmp = dir ? tmp->next : tmp->prev; |
1da177e4 LT |
468 | p = NULL; |
469 | if (tmp == head) | |
470 | break; | |
471 | p = list_entry(tmp, struct vfsmount, mnt_hash); | |
472 | if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) { | |
a05964f3 | 473 | found = p; |
1da177e4 LT |
474 | break; |
475 | } | |
476 | } | |
1da177e4 LT |
477 | return found; |
478 | } | |
479 | ||
a05964f3 RP |
480 | /* |
481 | * lookup_mnt increments the ref count before returning | |
482 | * the vfsmount struct. | |
483 | */ | |
1c755af4 | 484 | struct vfsmount *lookup_mnt(struct path *path) |
a05964f3 RP |
485 | { |
486 | struct vfsmount *child_mnt; | |
99b7db7b NP |
487 | |
488 | br_read_lock(vfsmount_lock); | |
1c755af4 | 489 | if ((child_mnt = __lookup_mnt(path->mnt, path->dentry, 1))) |
a05964f3 | 490 | mntget(child_mnt); |
99b7db7b | 491 | br_read_unlock(vfsmount_lock); |
a05964f3 RP |
492 | return child_mnt; |
493 | } | |
494 | ||
1da177e4 LT |
495 | static inline int check_mnt(struct vfsmount *mnt) |
496 | { | |
6b3286ed | 497 | return mnt->mnt_ns == current->nsproxy->mnt_ns; |
1da177e4 LT |
498 | } |
499 | ||
99b7db7b NP |
500 | /* |
501 | * vfsmount lock must be held for write | |
502 | */ | |
6b3286ed | 503 | static void touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
504 | { |
505 | if (ns) { | |
506 | ns->event = ++event; | |
507 | wake_up_interruptible(&ns->poll); | |
508 | } | |
509 | } | |
510 | ||
99b7db7b NP |
511 | /* |
512 | * vfsmount lock must be held for write | |
513 | */ | |
6b3286ed | 514 | static void __touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
515 | { |
516 | if (ns && ns->event != event) { | |
517 | ns->event = event; | |
518 | wake_up_interruptible(&ns->poll); | |
519 | } | |
520 | } | |
521 | ||
5f57cbcc NP |
522 | /* |
523 | * Clear dentry's mounted state if it has no remaining mounts. | |
524 | * vfsmount_lock must be held for write. | |
525 | */ | |
526 | static void dentry_reset_mounted(struct vfsmount *mnt, struct dentry *dentry) | |
527 | { | |
528 | unsigned u; | |
529 | ||
530 | for (u = 0; u < HASH_SIZE; u++) { | |
531 | struct vfsmount *p; | |
532 | ||
533 | list_for_each_entry(p, &mount_hashtable[u], mnt_hash) { | |
534 | if (p->mnt_mountpoint == dentry) | |
535 | return; | |
536 | } | |
537 | } | |
538 | spin_lock(&dentry->d_lock); | |
539 | dentry->d_flags &= ~DCACHE_MOUNTED; | |
540 | spin_unlock(&dentry->d_lock); | |
541 | } | |
542 | ||
99b7db7b NP |
543 | /* |
544 | * vfsmount lock must be held for write | |
545 | */ | |
1a390689 | 546 | static void detach_mnt(struct vfsmount *mnt, struct path *old_path) |
1da177e4 | 547 | { |
1a390689 AV |
548 | old_path->dentry = mnt->mnt_mountpoint; |
549 | old_path->mnt = mnt->mnt_parent; | |
1da177e4 LT |
550 | mnt->mnt_parent = mnt; |
551 | mnt->mnt_mountpoint = mnt->mnt_root; | |
552 | list_del_init(&mnt->mnt_child); | |
553 | list_del_init(&mnt->mnt_hash); | |
5f57cbcc | 554 | dentry_reset_mounted(old_path->mnt, old_path->dentry); |
1da177e4 LT |
555 | } |
556 | ||
99b7db7b NP |
557 | /* |
558 | * vfsmount lock must be held for write | |
559 | */ | |
b90fa9ae RP |
560 | void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, |
561 | struct vfsmount *child_mnt) | |
562 | { | |
563 | child_mnt->mnt_parent = mntget(mnt); | |
564 | child_mnt->mnt_mountpoint = dget(dentry); | |
5f57cbcc NP |
565 | spin_lock(&dentry->d_lock); |
566 | dentry->d_flags |= DCACHE_MOUNTED; | |
567 | spin_unlock(&dentry->d_lock); | |
b90fa9ae RP |
568 | } |
569 | ||
99b7db7b NP |
570 | /* |
571 | * vfsmount lock must be held for write | |
572 | */ | |
1a390689 | 573 | static void attach_mnt(struct vfsmount *mnt, struct path *path) |
1da177e4 | 574 | { |
1a390689 | 575 | mnt_set_mountpoint(path->mnt, path->dentry, mnt); |
b90fa9ae | 576 | list_add_tail(&mnt->mnt_hash, mount_hashtable + |
1a390689 AV |
577 | hash(path->mnt, path->dentry)); |
578 | list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); | |
b90fa9ae RP |
579 | } |
580 | ||
7e3d0eb0 AV |
581 | static inline void __mnt_make_longterm(struct vfsmount *mnt) |
582 | { | |
583 | #ifdef CONFIG_SMP | |
584 | atomic_inc(&mnt->mnt_longterm); | |
585 | #endif | |
586 | } | |
587 | ||
588 | /* needs vfsmount lock for write */ | |
589 | static inline void __mnt_make_shortterm(struct vfsmount *mnt) | |
590 | { | |
591 | #ifdef CONFIG_SMP | |
592 | atomic_dec(&mnt->mnt_longterm); | |
593 | #endif | |
594 | } | |
595 | ||
b90fa9ae | 596 | /* |
99b7db7b | 597 | * vfsmount lock must be held for write |
b90fa9ae RP |
598 | */ |
599 | static void commit_tree(struct vfsmount *mnt) | |
600 | { | |
601 | struct vfsmount *parent = mnt->mnt_parent; | |
602 | struct vfsmount *m; | |
603 | LIST_HEAD(head); | |
6b3286ed | 604 | struct mnt_namespace *n = parent->mnt_ns; |
b90fa9ae RP |
605 | |
606 | BUG_ON(parent == mnt); | |
607 | ||
608 | list_add_tail(&head, &mnt->mnt_list); | |
f03c6599 | 609 | list_for_each_entry(m, &head, mnt_list) { |
6b3286ed | 610 | m->mnt_ns = n; |
7e3d0eb0 | 611 | __mnt_make_longterm(m); |
f03c6599 AV |
612 | } |
613 | ||
b90fa9ae RP |
614 | list_splice(&head, n->list.prev); |
615 | ||
616 | list_add_tail(&mnt->mnt_hash, mount_hashtable + | |
617 | hash(parent, mnt->mnt_mountpoint)); | |
618 | list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); | |
6b3286ed | 619 | touch_mnt_namespace(n); |
1da177e4 LT |
620 | } |
621 | ||
622 | static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) | |
623 | { | |
624 | struct list_head *next = p->mnt_mounts.next; | |
625 | if (next == &p->mnt_mounts) { | |
626 | while (1) { | |
627 | if (p == root) | |
628 | return NULL; | |
629 | next = p->mnt_child.next; | |
630 | if (next != &p->mnt_parent->mnt_mounts) | |
631 | break; | |
632 | p = p->mnt_parent; | |
633 | } | |
634 | } | |
635 | return list_entry(next, struct vfsmount, mnt_child); | |
636 | } | |
637 | ||
9676f0c6 RP |
638 | static struct vfsmount *skip_mnt_tree(struct vfsmount *p) |
639 | { | |
640 | struct list_head *prev = p->mnt_mounts.prev; | |
641 | while (prev != &p->mnt_mounts) { | |
642 | p = list_entry(prev, struct vfsmount, mnt_child); | |
643 | prev = p->mnt_mounts.prev; | |
644 | } | |
645 | return p; | |
646 | } | |
647 | ||
9d412a43 AV |
648 | struct vfsmount * |
649 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) | |
650 | { | |
651 | struct vfsmount *mnt; | |
652 | struct dentry *root; | |
653 | ||
654 | if (!type) | |
655 | return ERR_PTR(-ENODEV); | |
656 | ||
657 | mnt = alloc_vfsmnt(name); | |
658 | if (!mnt) | |
659 | return ERR_PTR(-ENOMEM); | |
660 | ||
661 | if (flags & MS_KERNMOUNT) | |
662 | mnt->mnt_flags = MNT_INTERNAL; | |
663 | ||
664 | root = mount_fs(type, flags, name, data); | |
665 | if (IS_ERR(root)) { | |
666 | free_vfsmnt(mnt); | |
667 | return ERR_CAST(root); | |
668 | } | |
669 | ||
670 | mnt->mnt_root = root; | |
671 | mnt->mnt_sb = root->d_sb; | |
672 | mnt->mnt_mountpoint = mnt->mnt_root; | |
673 | mnt->mnt_parent = mnt; | |
674 | return mnt; | |
675 | } | |
676 | EXPORT_SYMBOL_GPL(vfs_kern_mount); | |
677 | ||
36341f64 RP |
678 | static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, |
679 | int flag) | |
1da177e4 LT |
680 | { |
681 | struct super_block *sb = old->mnt_sb; | |
682 | struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); | |
683 | ||
684 | if (mnt) { | |
719f5d7f MS |
685 | if (flag & (CL_SLAVE | CL_PRIVATE)) |
686 | mnt->mnt_group_id = 0; /* not a peer of original */ | |
687 | else | |
688 | mnt->mnt_group_id = old->mnt_group_id; | |
689 | ||
690 | if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { | |
691 | int err = mnt_alloc_group_id(mnt); | |
692 | if (err) | |
693 | goto out_free; | |
694 | } | |
695 | ||
be1a16a0 | 696 | mnt->mnt_flags = old->mnt_flags & ~MNT_WRITE_HOLD; |
1da177e4 LT |
697 | atomic_inc(&sb->s_active); |
698 | mnt->mnt_sb = sb; | |
699 | mnt->mnt_root = dget(root); | |
700 | mnt->mnt_mountpoint = mnt->mnt_root; | |
701 | mnt->mnt_parent = mnt; | |
b90fa9ae | 702 | |
5afe0022 RP |
703 | if (flag & CL_SLAVE) { |
704 | list_add(&mnt->mnt_slave, &old->mnt_slave_list); | |
705 | mnt->mnt_master = old; | |
706 | CLEAR_MNT_SHARED(mnt); | |
8aec0809 | 707 | } else if (!(flag & CL_PRIVATE)) { |
796a6b52 | 708 | if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) |
5afe0022 RP |
709 | list_add(&mnt->mnt_share, &old->mnt_share); |
710 | if (IS_MNT_SLAVE(old)) | |
711 | list_add(&mnt->mnt_slave, &old->mnt_slave); | |
712 | mnt->mnt_master = old->mnt_master; | |
713 | } | |
b90fa9ae RP |
714 | if (flag & CL_MAKE_SHARED) |
715 | set_mnt_shared(mnt); | |
1da177e4 LT |
716 | |
717 | /* stick the duplicate mount on the same expiry list | |
718 | * as the original if that was on one */ | |
36341f64 | 719 | if (flag & CL_EXPIRE) { |
36341f64 RP |
720 | if (!list_empty(&old->mnt_expire)) |
721 | list_add(&mnt->mnt_expire, &old->mnt_expire); | |
36341f64 | 722 | } |
1da177e4 LT |
723 | } |
724 | return mnt; | |
719f5d7f MS |
725 | |
726 | out_free: | |
727 | free_vfsmnt(mnt); | |
728 | return NULL; | |
1da177e4 LT |
729 | } |
730 | ||
b3e19d92 | 731 | static inline void mntfree(struct vfsmount *mnt) |
1da177e4 LT |
732 | { |
733 | struct super_block *sb = mnt->mnt_sb; | |
b3e19d92 | 734 | |
3d733633 DH |
735 | /* |
736 | * This probably indicates that somebody messed | |
737 | * up a mnt_want/drop_write() pair. If this | |
738 | * happens, the filesystem was probably unable | |
739 | * to make r/w->r/o transitions. | |
740 | */ | |
d3ef3d73 | 741 | /* |
b3e19d92 NP |
742 | * The locking used to deal with mnt_count decrement provides barriers, |
743 | * so mnt_get_writers() below is safe. | |
d3ef3d73 | 744 | */ |
c6653a83 | 745 | WARN_ON(mnt_get_writers(mnt)); |
ca9c726e | 746 | fsnotify_vfsmount_delete(mnt); |
1da177e4 LT |
747 | dput(mnt->mnt_root); |
748 | free_vfsmnt(mnt); | |
749 | deactivate_super(sb); | |
750 | } | |
751 | ||
f03c6599 | 752 | static void mntput_no_expire(struct vfsmount *mnt) |
b3e19d92 | 753 | { |
b3e19d92 | 754 | put_again: |
f03c6599 AV |
755 | #ifdef CONFIG_SMP |
756 | br_read_lock(vfsmount_lock); | |
757 | if (likely(atomic_read(&mnt->mnt_longterm))) { | |
aa9c0e07 | 758 | mnt_add_count(mnt, -1); |
b3e19d92 | 759 | br_read_unlock(vfsmount_lock); |
f03c6599 | 760 | return; |
b3e19d92 | 761 | } |
f03c6599 | 762 | br_read_unlock(vfsmount_lock); |
b3e19d92 | 763 | |
99b7db7b | 764 | br_write_lock(vfsmount_lock); |
aa9c0e07 | 765 | mnt_add_count(mnt, -1); |
b3e19d92 | 766 | if (mnt_get_count(mnt)) { |
99b7db7b NP |
767 | br_write_unlock(vfsmount_lock); |
768 | return; | |
769 | } | |
b3e19d92 | 770 | #else |
aa9c0e07 | 771 | mnt_add_count(mnt, -1); |
b3e19d92 | 772 | if (likely(mnt_get_count(mnt))) |
99b7db7b | 773 | return; |
b3e19d92 | 774 | br_write_lock(vfsmount_lock); |
f03c6599 | 775 | #endif |
b3e19d92 NP |
776 | if (unlikely(mnt->mnt_pinned)) { |
777 | mnt_add_count(mnt, mnt->mnt_pinned + 1); | |
778 | mnt->mnt_pinned = 0; | |
779 | br_write_unlock(vfsmount_lock); | |
780 | acct_auto_close_mnt(mnt); | |
781 | goto put_again; | |
7b7b1ace | 782 | } |
99b7db7b | 783 | br_write_unlock(vfsmount_lock); |
b3e19d92 NP |
784 | mntfree(mnt); |
785 | } | |
b3e19d92 NP |
786 | |
787 | void mntput(struct vfsmount *mnt) | |
788 | { | |
789 | if (mnt) { | |
790 | /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ | |
791 | if (unlikely(mnt->mnt_expiry_mark)) | |
792 | mnt->mnt_expiry_mark = 0; | |
f03c6599 | 793 | mntput_no_expire(mnt); |
b3e19d92 NP |
794 | } |
795 | } | |
796 | EXPORT_SYMBOL(mntput); | |
797 | ||
798 | struct vfsmount *mntget(struct vfsmount *mnt) | |
799 | { | |
800 | if (mnt) | |
aa9c0e07 | 801 | mnt_add_count(mnt, 1); |
b3e19d92 NP |
802 | return mnt; |
803 | } | |
804 | EXPORT_SYMBOL(mntget); | |
805 | ||
7b7b1ace AV |
806 | void mnt_pin(struct vfsmount *mnt) |
807 | { | |
99b7db7b | 808 | br_write_lock(vfsmount_lock); |
7b7b1ace | 809 | mnt->mnt_pinned++; |
99b7db7b | 810 | br_write_unlock(vfsmount_lock); |
7b7b1ace | 811 | } |
7b7b1ace AV |
812 | EXPORT_SYMBOL(mnt_pin); |
813 | ||
814 | void mnt_unpin(struct vfsmount *mnt) | |
815 | { | |
99b7db7b | 816 | br_write_lock(vfsmount_lock); |
7b7b1ace | 817 | if (mnt->mnt_pinned) { |
aa9c0e07 | 818 | mnt_add_count(mnt, 1); |
7b7b1ace AV |
819 | mnt->mnt_pinned--; |
820 | } | |
99b7db7b | 821 | br_write_unlock(vfsmount_lock); |
7b7b1ace | 822 | } |
7b7b1ace | 823 | EXPORT_SYMBOL(mnt_unpin); |
1da177e4 | 824 | |
b3b304a2 MS |
825 | static inline void mangle(struct seq_file *m, const char *s) |
826 | { | |
827 | seq_escape(m, s, " \t\n\\"); | |
828 | } | |
829 | ||
830 | /* | |
831 | * Simple .show_options callback for filesystems which don't want to | |
832 | * implement more complex mount option showing. | |
833 | * | |
834 | * See also save_mount_options(). | |
835 | */ | |
836 | int generic_show_options(struct seq_file *m, struct vfsmount *mnt) | |
837 | { | |
2a32cebd AV |
838 | const char *options; |
839 | ||
840 | rcu_read_lock(); | |
841 | options = rcu_dereference(mnt->mnt_sb->s_options); | |
b3b304a2 MS |
842 | |
843 | if (options != NULL && options[0]) { | |
844 | seq_putc(m, ','); | |
845 | mangle(m, options); | |
846 | } | |
2a32cebd | 847 | rcu_read_unlock(); |
b3b304a2 MS |
848 | |
849 | return 0; | |
850 | } | |
851 | EXPORT_SYMBOL(generic_show_options); | |
852 | ||
853 | /* | |
854 | * If filesystem uses generic_show_options(), this function should be | |
855 | * called from the fill_super() callback. | |
856 | * | |
857 | * The .remount_fs callback usually needs to be handled in a special | |
858 | * way, to make sure, that previous options are not overwritten if the | |
859 | * remount fails. | |
860 | * | |
861 | * Also note, that if the filesystem's .remount_fs function doesn't | |
862 | * reset all options to their default value, but changes only newly | |
863 | * given options, then the displayed options will not reflect reality | |
864 | * any more. | |
865 | */ | |
866 | void save_mount_options(struct super_block *sb, char *options) | |
867 | { | |
2a32cebd AV |
868 | BUG_ON(sb->s_options); |
869 | rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); | |
b3b304a2 MS |
870 | } |
871 | EXPORT_SYMBOL(save_mount_options); | |
872 | ||
2a32cebd AV |
873 | void replace_mount_options(struct super_block *sb, char *options) |
874 | { | |
875 | char *old = sb->s_options; | |
876 | rcu_assign_pointer(sb->s_options, options); | |
877 | if (old) { | |
878 | synchronize_rcu(); | |
879 | kfree(old); | |
880 | } | |
881 | } | |
882 | EXPORT_SYMBOL(replace_mount_options); | |
883 | ||
a1a2c409 | 884 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
885 | /* iterator */ |
886 | static void *m_start(struct seq_file *m, loff_t *pos) | |
887 | { | |
a1a2c409 | 888 | struct proc_mounts *p = m->private; |
1da177e4 | 889 | |
390c6843 | 890 | down_read(&namespace_sem); |
a1a2c409 | 891 | return seq_list_start(&p->ns->list, *pos); |
1da177e4 LT |
892 | } |
893 | ||
894 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
895 | { | |
a1a2c409 | 896 | struct proc_mounts *p = m->private; |
b0765fb8 | 897 | |
a1a2c409 | 898 | return seq_list_next(v, &p->ns->list, pos); |
1da177e4 LT |
899 | } |
900 | ||
901 | static void m_stop(struct seq_file *m, void *v) | |
902 | { | |
390c6843 | 903 | up_read(&namespace_sem); |
1da177e4 LT |
904 | } |
905 | ||
9f5596af AV |
906 | int mnt_had_events(struct proc_mounts *p) |
907 | { | |
908 | struct mnt_namespace *ns = p->ns; | |
909 | int res = 0; | |
910 | ||
99b7db7b | 911 | br_read_lock(vfsmount_lock); |
f1514638 KS |
912 | if (p->m.poll_event != ns->event) { |
913 | p->m.poll_event = ns->event; | |
9f5596af AV |
914 | res = 1; |
915 | } | |
99b7db7b | 916 | br_read_unlock(vfsmount_lock); |
9f5596af AV |
917 | |
918 | return res; | |
919 | } | |
920 | ||
2d4d4864 RP |
921 | struct proc_fs_info { |
922 | int flag; | |
923 | const char *str; | |
924 | }; | |
925 | ||
2069f457 | 926 | static int show_sb_opts(struct seq_file *m, struct super_block *sb) |
1da177e4 | 927 | { |
2d4d4864 | 928 | static const struct proc_fs_info fs_info[] = { |
1da177e4 LT |
929 | { MS_SYNCHRONOUS, ",sync" }, |
930 | { MS_DIRSYNC, ",dirsync" }, | |
931 | { MS_MANDLOCK, ",mand" }, | |
1da177e4 LT |
932 | { 0, NULL } |
933 | }; | |
2d4d4864 RP |
934 | const struct proc_fs_info *fs_infop; |
935 | ||
936 | for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { | |
937 | if (sb->s_flags & fs_infop->flag) | |
938 | seq_puts(m, fs_infop->str); | |
939 | } | |
2069f457 EP |
940 | |
941 | return security_sb_show_options(m, sb); | |
2d4d4864 RP |
942 | } |
943 | ||
944 | static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) | |
945 | { | |
946 | static const struct proc_fs_info mnt_info[] = { | |
1da177e4 LT |
947 | { MNT_NOSUID, ",nosuid" }, |
948 | { MNT_NODEV, ",nodev" }, | |
949 | { MNT_NOEXEC, ",noexec" }, | |
fc33a7bb CH |
950 | { MNT_NOATIME, ",noatime" }, |
951 | { MNT_NODIRATIME, ",nodiratime" }, | |
47ae32d6 | 952 | { MNT_RELATIME, ",relatime" }, |
1da177e4 LT |
953 | { 0, NULL } |
954 | }; | |
2d4d4864 RP |
955 | const struct proc_fs_info *fs_infop; |
956 | ||
957 | for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { | |
958 | if (mnt->mnt_flags & fs_infop->flag) | |
959 | seq_puts(m, fs_infop->str); | |
960 | } | |
961 | } | |
962 | ||
963 | static void show_type(struct seq_file *m, struct super_block *sb) | |
964 | { | |
965 | mangle(m, sb->s_type->name); | |
966 | if (sb->s_subtype && sb->s_subtype[0]) { | |
967 | seq_putc(m, '.'); | |
968 | mangle(m, sb->s_subtype); | |
969 | } | |
970 | } | |
971 | ||
972 | static int show_vfsmnt(struct seq_file *m, void *v) | |
973 | { | |
974 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); | |
975 | int err = 0; | |
c32c2f63 | 976 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; |
1da177e4 | 977 | |
c7f404b4 AV |
978 | if (mnt->mnt_sb->s_op->show_devname) { |
979 | err = mnt->mnt_sb->s_op->show_devname(m, mnt); | |
980 | if (err) | |
981 | goto out; | |
982 | } else { | |
983 | mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); | |
984 | } | |
1da177e4 | 985 | seq_putc(m, ' '); |
c32c2f63 | 986 | seq_path(m, &mnt_path, " \t\n\\"); |
1da177e4 | 987 | seq_putc(m, ' '); |
2d4d4864 | 988 | show_type(m, mnt->mnt_sb); |
2e4b7fcd | 989 | seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); |
2069f457 EP |
990 | err = show_sb_opts(m, mnt->mnt_sb); |
991 | if (err) | |
992 | goto out; | |
2d4d4864 | 993 | show_mnt_opts(m, mnt); |
1da177e4 LT |
994 | if (mnt->mnt_sb->s_op->show_options) |
995 | err = mnt->mnt_sb->s_op->show_options(m, mnt); | |
996 | seq_puts(m, " 0 0\n"); | |
2069f457 | 997 | out: |
1da177e4 LT |
998 | return err; |
999 | } | |
1000 | ||
a1a2c409 | 1001 | const struct seq_operations mounts_op = { |
1da177e4 LT |
1002 | .start = m_start, |
1003 | .next = m_next, | |
1004 | .stop = m_stop, | |
1005 | .show = show_vfsmnt | |
1006 | }; | |
1007 | ||
2d4d4864 RP |
1008 | static int show_mountinfo(struct seq_file *m, void *v) |
1009 | { | |
1010 | struct proc_mounts *p = m->private; | |
1011 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); | |
1012 | struct super_block *sb = mnt->mnt_sb; | |
1013 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; | |
1014 | struct path root = p->root; | |
1015 | int err = 0; | |
1016 | ||
1017 | seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, | |
1018 | MAJOR(sb->s_dev), MINOR(sb->s_dev)); | |
c7f404b4 AV |
1019 | if (sb->s_op->show_path) |
1020 | err = sb->s_op->show_path(m, mnt); | |
1021 | else | |
1022 | seq_dentry(m, mnt->mnt_root, " \t\n\\"); | |
1023 | if (err) | |
1024 | goto out; | |
2d4d4864 | 1025 | seq_putc(m, ' '); |
02125a82 AV |
1026 | |
1027 | /* mountpoints outside of chroot jail will give SEQ_SKIP on this */ | |
1028 | err = seq_path_root(m, &mnt_path, &root, " \t\n\\"); | |
1029 | if (err) | |
1030 | goto out; | |
1031 | ||
2d4d4864 RP |
1032 | seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); |
1033 | show_mnt_opts(m, mnt); | |
1034 | ||
1035 | /* Tagged fields ("foo:X" or "bar") */ | |
1036 | if (IS_MNT_SHARED(mnt)) | |
1037 | seq_printf(m, " shared:%i", mnt->mnt_group_id); | |
97e7e0f7 MS |
1038 | if (IS_MNT_SLAVE(mnt)) { |
1039 | int master = mnt->mnt_master->mnt_group_id; | |
1040 | int dom = get_dominating_id(mnt, &p->root); | |
1041 | seq_printf(m, " master:%i", master); | |
1042 | if (dom && dom != master) | |
1043 | seq_printf(m, " propagate_from:%i", dom); | |
1044 | } | |
2d4d4864 RP |
1045 | if (IS_MNT_UNBINDABLE(mnt)) |
1046 | seq_puts(m, " unbindable"); | |
1047 | ||
1048 | /* Filesystem specific data */ | |
1049 | seq_puts(m, " - "); | |
1050 | show_type(m, sb); | |
1051 | seq_putc(m, ' '); | |
c7f404b4 AV |
1052 | if (sb->s_op->show_devname) |
1053 | err = sb->s_op->show_devname(m, mnt); | |
1054 | else | |
1055 | mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); | |
1056 | if (err) | |
1057 | goto out; | |
2d4d4864 | 1058 | seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); |
2069f457 EP |
1059 | err = show_sb_opts(m, sb); |
1060 | if (err) | |
1061 | goto out; | |
2d4d4864 RP |
1062 | if (sb->s_op->show_options) |
1063 | err = sb->s_op->show_options(m, mnt); | |
1064 | seq_putc(m, '\n'); | |
2069f457 | 1065 | out: |
2d4d4864 RP |
1066 | return err; |
1067 | } | |
1068 | ||
1069 | const struct seq_operations mountinfo_op = { | |
1070 | .start = m_start, | |
1071 | .next = m_next, | |
1072 | .stop = m_stop, | |
1073 | .show = show_mountinfo, | |
1074 | }; | |
1075 | ||
b4629fe2 CL |
1076 | static int show_vfsstat(struct seq_file *m, void *v) |
1077 | { | |
b0765fb8 | 1078 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); |
c32c2f63 | 1079 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; |
b4629fe2 CL |
1080 | int err = 0; |
1081 | ||
1082 | /* device */ | |
c7f404b4 | 1083 | if (mnt->mnt_sb->s_op->show_devname) { |
a877ee03 | 1084 | seq_puts(m, "device "); |
c7f404b4 AV |
1085 | err = mnt->mnt_sb->s_op->show_devname(m, mnt); |
1086 | } else { | |
1087 | if (mnt->mnt_devname) { | |
1088 | seq_puts(m, "device "); | |
1089 | mangle(m, mnt->mnt_devname); | |
1090 | } else | |
1091 | seq_puts(m, "no device"); | |
1092 | } | |
b4629fe2 CL |
1093 | |
1094 | /* mount point */ | |
1095 | seq_puts(m, " mounted on "); | |
c32c2f63 | 1096 | seq_path(m, &mnt_path, " \t\n\\"); |
b4629fe2 CL |
1097 | seq_putc(m, ' '); |
1098 | ||
1099 | /* file system type */ | |
1100 | seq_puts(m, "with fstype "); | |
2d4d4864 | 1101 | show_type(m, mnt->mnt_sb); |
b4629fe2 CL |
1102 | |
1103 | /* optional statistics */ | |
1104 | if (mnt->mnt_sb->s_op->show_stats) { | |
1105 | seq_putc(m, ' '); | |
c7f404b4 AV |
1106 | if (!err) |
1107 | err = mnt->mnt_sb->s_op->show_stats(m, mnt); | |
b4629fe2 CL |
1108 | } |
1109 | ||
1110 | seq_putc(m, '\n'); | |
1111 | return err; | |
1112 | } | |
1113 | ||
a1a2c409 | 1114 | const struct seq_operations mountstats_op = { |
b4629fe2 CL |
1115 | .start = m_start, |
1116 | .next = m_next, | |
1117 | .stop = m_stop, | |
1118 | .show = show_vfsstat, | |
1119 | }; | |
a1a2c409 | 1120 | #endif /* CONFIG_PROC_FS */ |
b4629fe2 | 1121 | |
1da177e4 LT |
1122 | /** |
1123 | * may_umount_tree - check if a mount tree is busy | |
1124 | * @mnt: root of mount tree | |
1125 | * | |
1126 | * This is called to check if a tree of mounts has any | |
1127 | * open files, pwds, chroots or sub mounts that are | |
1128 | * busy. | |
1129 | */ | |
1130 | int may_umount_tree(struct vfsmount *mnt) | |
1131 | { | |
36341f64 RP |
1132 | int actual_refs = 0; |
1133 | int minimum_refs = 0; | |
1134 | struct vfsmount *p; | |
1da177e4 | 1135 | |
b3e19d92 NP |
1136 | /* write lock needed for mnt_get_count */ |
1137 | br_write_lock(vfsmount_lock); | |
36341f64 | 1138 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
b3e19d92 | 1139 | actual_refs += mnt_get_count(p); |
1da177e4 | 1140 | minimum_refs += 2; |
1da177e4 | 1141 | } |
b3e19d92 | 1142 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
1143 | |
1144 | if (actual_refs > minimum_refs) | |
e3474a8e | 1145 | return 0; |
1da177e4 | 1146 | |
e3474a8e | 1147 | return 1; |
1da177e4 LT |
1148 | } |
1149 | ||
1150 | EXPORT_SYMBOL(may_umount_tree); | |
1151 | ||
1152 | /** | |
1153 | * may_umount - check if a mount point is busy | |
1154 | * @mnt: root of mount | |
1155 | * | |
1156 | * This is called to check if a mount point has any | |
1157 | * open files, pwds, chroots or sub mounts. If the | |
1158 | * mount has sub mounts this will return busy | |
1159 | * regardless of whether the sub mounts are busy. | |
1160 | * | |
1161 | * Doesn't take quota and stuff into account. IOW, in some cases it will | |
1162 | * give false negatives. The main reason why it's here is that we need | |
1163 | * a non-destructive way to look for easily umountable filesystems. | |
1164 | */ | |
1165 | int may_umount(struct vfsmount *mnt) | |
1166 | { | |
e3474a8e | 1167 | int ret = 1; |
8ad08d8a | 1168 | down_read(&namespace_sem); |
b3e19d92 | 1169 | br_write_lock(vfsmount_lock); |
a05964f3 | 1170 | if (propagate_mount_busy(mnt, 2)) |
e3474a8e | 1171 | ret = 0; |
b3e19d92 | 1172 | br_write_unlock(vfsmount_lock); |
8ad08d8a | 1173 | up_read(&namespace_sem); |
a05964f3 | 1174 | return ret; |
1da177e4 LT |
1175 | } |
1176 | ||
1177 | EXPORT_SYMBOL(may_umount); | |
1178 | ||
b90fa9ae | 1179 | void release_mounts(struct list_head *head) |
70fbcdf4 RP |
1180 | { |
1181 | struct vfsmount *mnt; | |
bf066c7d | 1182 | while (!list_empty(head)) { |
b5e61818 | 1183 | mnt = list_first_entry(head, struct vfsmount, mnt_hash); |
70fbcdf4 RP |
1184 | list_del_init(&mnt->mnt_hash); |
1185 | if (mnt->mnt_parent != mnt) { | |
1186 | struct dentry *dentry; | |
1187 | struct vfsmount *m; | |
99b7db7b NP |
1188 | |
1189 | br_write_lock(vfsmount_lock); | |
70fbcdf4 RP |
1190 | dentry = mnt->mnt_mountpoint; |
1191 | m = mnt->mnt_parent; | |
1192 | mnt->mnt_mountpoint = mnt->mnt_root; | |
1193 | mnt->mnt_parent = mnt; | |
7c4b93d8 | 1194 | m->mnt_ghosts--; |
99b7db7b | 1195 | br_write_unlock(vfsmount_lock); |
70fbcdf4 RP |
1196 | dput(dentry); |
1197 | mntput(m); | |
1198 | } | |
f03c6599 | 1199 | mntput(mnt); |
70fbcdf4 RP |
1200 | } |
1201 | } | |
1202 | ||
99b7db7b NP |
1203 | /* |
1204 | * vfsmount lock must be held for write | |
1205 | * namespace_sem must be held for write | |
1206 | */ | |
a05964f3 | 1207 | void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) |
1da177e4 | 1208 | { |
7b8a53fd | 1209 | LIST_HEAD(tmp_list); |
1da177e4 | 1210 | struct vfsmount *p; |
1da177e4 | 1211 | |
1bfba4e8 | 1212 | for (p = mnt; p; p = next_mnt(p, mnt)) |
7b8a53fd | 1213 | list_move(&p->mnt_hash, &tmp_list); |
1da177e4 | 1214 | |
a05964f3 | 1215 | if (propagate) |
7b8a53fd | 1216 | propagate_umount(&tmp_list); |
a05964f3 | 1217 | |
7b8a53fd | 1218 | list_for_each_entry(p, &tmp_list, mnt_hash) { |
70fbcdf4 RP |
1219 | list_del_init(&p->mnt_expire); |
1220 | list_del_init(&p->mnt_list); | |
6b3286ed KK |
1221 | __touch_mnt_namespace(p->mnt_ns); |
1222 | p->mnt_ns = NULL; | |
7e3d0eb0 | 1223 | __mnt_make_shortterm(p); |
70fbcdf4 | 1224 | list_del_init(&p->mnt_child); |
7c4b93d8 AV |
1225 | if (p->mnt_parent != p) { |
1226 | p->mnt_parent->mnt_ghosts++; | |
5f57cbcc | 1227 | dentry_reset_mounted(p->mnt_parent, p->mnt_mountpoint); |
7c4b93d8 | 1228 | } |
a05964f3 | 1229 | change_mnt_propagation(p, MS_PRIVATE); |
1da177e4 | 1230 | } |
7b8a53fd | 1231 | list_splice(&tmp_list, kill); |
1da177e4 LT |
1232 | } |
1233 | ||
c35038be AV |
1234 | static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); |
1235 | ||
1da177e4 LT |
1236 | static int do_umount(struct vfsmount *mnt, int flags) |
1237 | { | |
b58fed8b | 1238 | struct super_block *sb = mnt->mnt_sb; |
1da177e4 | 1239 | int retval; |
70fbcdf4 | 1240 | LIST_HEAD(umount_list); |
1da177e4 LT |
1241 | |
1242 | retval = security_sb_umount(mnt, flags); | |
1243 | if (retval) | |
1244 | return retval; | |
1245 | ||
1246 | /* | |
1247 | * Allow userspace to request a mountpoint be expired rather than | |
1248 | * unmounting unconditionally. Unmount only happens if: | |
1249 | * (1) the mark is already set (the mark is cleared by mntput()) | |
1250 | * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] | |
1251 | */ | |
1252 | if (flags & MNT_EXPIRE) { | |
6ac08c39 | 1253 | if (mnt == current->fs->root.mnt || |
1da177e4 LT |
1254 | flags & (MNT_FORCE | MNT_DETACH)) |
1255 | return -EINVAL; | |
1256 | ||
b3e19d92 NP |
1257 | /* |
1258 | * probably don't strictly need the lock here if we examined | |
1259 | * all race cases, but it's a slowpath. | |
1260 | */ | |
1261 | br_write_lock(vfsmount_lock); | |
1262 | if (mnt_get_count(mnt) != 2) { | |
bf9faa2a | 1263 | br_write_unlock(vfsmount_lock); |
1da177e4 | 1264 | return -EBUSY; |
b3e19d92 NP |
1265 | } |
1266 | br_write_unlock(vfsmount_lock); | |
1da177e4 LT |
1267 | |
1268 | if (!xchg(&mnt->mnt_expiry_mark, 1)) | |
1269 | return -EAGAIN; | |
1270 | } | |
1271 | ||
1272 | /* | |
1273 | * If we may have to abort operations to get out of this | |
1274 | * mount, and they will themselves hold resources we must | |
1275 | * allow the fs to do things. In the Unix tradition of | |
1276 | * 'Gee thats tricky lets do it in userspace' the umount_begin | |
1277 | * might fail to complete on the first run through as other tasks | |
1278 | * must return, and the like. Thats for the mount program to worry | |
1279 | * about for the moment. | |
1280 | */ | |
1281 | ||
42faad99 | 1282 | if (flags & MNT_FORCE && sb->s_op->umount_begin) { |
42faad99 | 1283 | sb->s_op->umount_begin(sb); |
42faad99 | 1284 | } |
1da177e4 LT |
1285 | |
1286 | /* | |
1287 | * No sense to grab the lock for this test, but test itself looks | |
1288 | * somewhat bogus. Suggestions for better replacement? | |
1289 | * Ho-hum... In principle, we might treat that as umount + switch | |
1290 | * to rootfs. GC would eventually take care of the old vfsmount. | |
1291 | * Actually it makes sense, especially if rootfs would contain a | |
1292 | * /reboot - static binary that would close all descriptors and | |
1293 | * call reboot(9). Then init(8) could umount root and exec /reboot. | |
1294 | */ | |
6ac08c39 | 1295 | if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { |
1da177e4 LT |
1296 | /* |
1297 | * Special case for "unmounting" root ... | |
1298 | * we just try to remount it readonly. | |
1299 | */ | |
1300 | down_write(&sb->s_umount); | |
4aa98cf7 | 1301 | if (!(sb->s_flags & MS_RDONLY)) |
1da177e4 | 1302 | retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); |
1da177e4 LT |
1303 | up_write(&sb->s_umount); |
1304 | return retval; | |
1305 | } | |
1306 | ||
390c6843 | 1307 | down_write(&namespace_sem); |
99b7db7b | 1308 | br_write_lock(vfsmount_lock); |
5addc5dd | 1309 | event++; |
1da177e4 | 1310 | |
c35038be AV |
1311 | if (!(flags & MNT_DETACH)) |
1312 | shrink_submounts(mnt, &umount_list); | |
1313 | ||
1da177e4 | 1314 | retval = -EBUSY; |
a05964f3 | 1315 | if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { |
1da177e4 | 1316 | if (!list_empty(&mnt->mnt_list)) |
a05964f3 | 1317 | umount_tree(mnt, 1, &umount_list); |
1da177e4 LT |
1318 | retval = 0; |
1319 | } | |
99b7db7b | 1320 | br_write_unlock(vfsmount_lock); |
390c6843 | 1321 | up_write(&namespace_sem); |
70fbcdf4 | 1322 | release_mounts(&umount_list); |
1da177e4 LT |
1323 | return retval; |
1324 | } | |
1325 | ||
1326 | /* | |
1327 | * Now umount can handle mount points as well as block devices. | |
1328 | * This is important for filesystems which use unnamed block devices. | |
1329 | * | |
1330 | * We now support a flag for forced unmount like the other 'big iron' | |
1331 | * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD | |
1332 | */ | |
1333 | ||
bdc480e3 | 1334 | SYSCALL_DEFINE2(umount, char __user *, name, int, flags) |
1da177e4 | 1335 | { |
2d8f3038 | 1336 | struct path path; |
1da177e4 | 1337 | int retval; |
db1f05bb | 1338 | int lookup_flags = 0; |
1da177e4 | 1339 | |
db1f05bb MS |
1340 | if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) |
1341 | return -EINVAL; | |
1342 | ||
1343 | if (!(flags & UMOUNT_NOFOLLOW)) | |
1344 | lookup_flags |= LOOKUP_FOLLOW; | |
1345 | ||
1346 | retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); | |
1da177e4 LT |
1347 | if (retval) |
1348 | goto out; | |
1349 | retval = -EINVAL; | |
2d8f3038 | 1350 | if (path.dentry != path.mnt->mnt_root) |
1da177e4 | 1351 | goto dput_and_out; |
2d8f3038 | 1352 | if (!check_mnt(path.mnt)) |
1da177e4 LT |
1353 | goto dput_and_out; |
1354 | ||
1355 | retval = -EPERM; | |
1356 | if (!capable(CAP_SYS_ADMIN)) | |
1357 | goto dput_and_out; | |
1358 | ||
2d8f3038 | 1359 | retval = do_umount(path.mnt, flags); |
1da177e4 | 1360 | dput_and_out: |
429731b1 | 1361 | /* we mustn't call path_put() as that would clear mnt_expiry_mark */ |
2d8f3038 AV |
1362 | dput(path.dentry); |
1363 | mntput_no_expire(path.mnt); | |
1da177e4 LT |
1364 | out: |
1365 | return retval; | |
1366 | } | |
1367 | ||
1368 | #ifdef __ARCH_WANT_SYS_OLDUMOUNT | |
1369 | ||
1370 | /* | |
b58fed8b | 1371 | * The 2.0 compatible umount. No flags. |
1da177e4 | 1372 | */ |
bdc480e3 | 1373 | SYSCALL_DEFINE1(oldumount, char __user *, name) |
1da177e4 | 1374 | { |
b58fed8b | 1375 | return sys_umount(name, 0); |
1da177e4 LT |
1376 | } |
1377 | ||
1378 | #endif | |
1379 | ||
2d92ab3c | 1380 | static int mount_is_safe(struct path *path) |
1da177e4 LT |
1381 | { |
1382 | if (capable(CAP_SYS_ADMIN)) | |
1383 | return 0; | |
1384 | return -EPERM; | |
1385 | #ifdef notyet | |
2d92ab3c | 1386 | if (S_ISLNK(path->dentry->d_inode->i_mode)) |
1da177e4 | 1387 | return -EPERM; |
2d92ab3c | 1388 | if (path->dentry->d_inode->i_mode & S_ISVTX) { |
da9592ed | 1389 | if (current_uid() != path->dentry->d_inode->i_uid) |
1da177e4 LT |
1390 | return -EPERM; |
1391 | } | |
2d92ab3c | 1392 | if (inode_permission(path->dentry->d_inode, MAY_WRITE)) |
1da177e4 LT |
1393 | return -EPERM; |
1394 | return 0; | |
1395 | #endif | |
1396 | } | |
1397 | ||
b90fa9ae | 1398 | struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, |
36341f64 | 1399 | int flag) |
1da177e4 LT |
1400 | { |
1401 | struct vfsmount *res, *p, *q, *r, *s; | |
1a390689 | 1402 | struct path path; |
1da177e4 | 1403 | |
9676f0c6 RP |
1404 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) |
1405 | return NULL; | |
1406 | ||
36341f64 | 1407 | res = q = clone_mnt(mnt, dentry, flag); |
1da177e4 LT |
1408 | if (!q) |
1409 | goto Enomem; | |
1410 | q->mnt_mountpoint = mnt->mnt_mountpoint; | |
1411 | ||
1412 | p = mnt; | |
fdadd65f | 1413 | list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { |
7ec02ef1 | 1414 | if (!is_subdir(r->mnt_mountpoint, dentry)) |
1da177e4 LT |
1415 | continue; |
1416 | ||
1417 | for (s = r; s; s = next_mnt(s, r)) { | |
9676f0c6 RP |
1418 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { |
1419 | s = skip_mnt_tree(s); | |
1420 | continue; | |
1421 | } | |
1da177e4 LT |
1422 | while (p != s->mnt_parent) { |
1423 | p = p->mnt_parent; | |
1424 | q = q->mnt_parent; | |
1425 | } | |
1426 | p = s; | |
1a390689 AV |
1427 | path.mnt = q; |
1428 | path.dentry = p->mnt_mountpoint; | |
36341f64 | 1429 | q = clone_mnt(p, p->mnt_root, flag); |
1da177e4 LT |
1430 | if (!q) |
1431 | goto Enomem; | |
99b7db7b | 1432 | br_write_lock(vfsmount_lock); |
1da177e4 | 1433 | list_add_tail(&q->mnt_list, &res->mnt_list); |
1a390689 | 1434 | attach_mnt(q, &path); |
99b7db7b | 1435 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
1436 | } |
1437 | } | |
1438 | return res; | |
b58fed8b | 1439 | Enomem: |
1da177e4 | 1440 | if (res) { |
70fbcdf4 | 1441 | LIST_HEAD(umount_list); |
99b7db7b | 1442 | br_write_lock(vfsmount_lock); |
a05964f3 | 1443 | umount_tree(res, 0, &umount_list); |
99b7db7b | 1444 | br_write_unlock(vfsmount_lock); |
70fbcdf4 | 1445 | release_mounts(&umount_list); |
1da177e4 LT |
1446 | } |
1447 | return NULL; | |
1448 | } | |
1449 | ||
589ff870 | 1450 | struct vfsmount *collect_mounts(struct path *path) |
8aec0809 AV |
1451 | { |
1452 | struct vfsmount *tree; | |
1a60a280 | 1453 | down_write(&namespace_sem); |
589ff870 | 1454 | tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); |
1a60a280 | 1455 | up_write(&namespace_sem); |
8aec0809 AV |
1456 | return tree; |
1457 | } | |
1458 | ||
1459 | void drop_collected_mounts(struct vfsmount *mnt) | |
1460 | { | |
1461 | LIST_HEAD(umount_list); | |
1a60a280 | 1462 | down_write(&namespace_sem); |
99b7db7b | 1463 | br_write_lock(vfsmount_lock); |
8aec0809 | 1464 | umount_tree(mnt, 0, &umount_list); |
99b7db7b | 1465 | br_write_unlock(vfsmount_lock); |
1a60a280 | 1466 | up_write(&namespace_sem); |
8aec0809 AV |
1467 | release_mounts(&umount_list); |
1468 | } | |
1469 | ||
1f707137 AV |
1470 | int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, |
1471 | struct vfsmount *root) | |
1472 | { | |
1473 | struct vfsmount *mnt; | |
1474 | int res = f(root, arg); | |
1475 | if (res) | |
1476 | return res; | |
1477 | list_for_each_entry(mnt, &root->mnt_list, mnt_list) { | |
1478 | res = f(mnt, arg); | |
1479 | if (res) | |
1480 | return res; | |
1481 | } | |
1482 | return 0; | |
1483 | } | |
1484 | ||
719f5d7f MS |
1485 | static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) |
1486 | { | |
1487 | struct vfsmount *p; | |
1488 | ||
1489 | for (p = mnt; p != end; p = next_mnt(p, mnt)) { | |
1490 | if (p->mnt_group_id && !IS_MNT_SHARED(p)) | |
1491 | mnt_release_group_id(p); | |
1492 | } | |
1493 | } | |
1494 | ||
1495 | static int invent_group_ids(struct vfsmount *mnt, bool recurse) | |
1496 | { | |
1497 | struct vfsmount *p; | |
1498 | ||
1499 | for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { | |
1500 | if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { | |
1501 | int err = mnt_alloc_group_id(p); | |
1502 | if (err) { | |
1503 | cleanup_group_ids(mnt, p); | |
1504 | return err; | |
1505 | } | |
1506 | } | |
1507 | } | |
1508 | ||
1509 | return 0; | |
1510 | } | |
1511 | ||
b90fa9ae RP |
1512 | /* |
1513 | * @source_mnt : mount tree to be attached | |
21444403 RP |
1514 | * @nd : place the mount tree @source_mnt is attached |
1515 | * @parent_nd : if non-null, detach the source_mnt from its parent and | |
1516 | * store the parent mount and mountpoint dentry. | |
1517 | * (done when source_mnt is moved) | |
b90fa9ae RP |
1518 | * |
1519 | * NOTE: in the table below explains the semantics when a source mount | |
1520 | * of a given type is attached to a destination mount of a given type. | |
9676f0c6 RP |
1521 | * --------------------------------------------------------------------------- |
1522 | * | BIND MOUNT OPERATION | | |
1523 | * |************************************************************************** | |
1524 | * | source-->| shared | private | slave | unbindable | | |
1525 | * | dest | | | | | | |
1526 | * | | | | | | | | |
1527 | * | v | | | | | | |
1528 | * |************************************************************************** | |
1529 | * | shared | shared (++) | shared (+) | shared(+++)| invalid | | |
1530 | * | | | | | | | |
1531 | * |non-shared| shared (+) | private | slave (*) | invalid | | |
1532 | * *************************************************************************** | |
b90fa9ae RP |
1533 | * A bind operation clones the source mount and mounts the clone on the |
1534 | * destination mount. | |
1535 | * | |
1536 | * (++) the cloned mount is propagated to all the mounts in the propagation | |
1537 | * tree of the destination mount and the cloned mount is added to | |
1538 | * the peer group of the source mount. | |
1539 | * (+) the cloned mount is created under the destination mount and is marked | |
1540 | * as shared. The cloned mount is added to the peer group of the source | |
1541 | * mount. | |
5afe0022 RP |
1542 | * (+++) the mount is propagated to all the mounts in the propagation tree |
1543 | * of the destination mount and the cloned mount is made slave | |
1544 | * of the same master as that of the source mount. The cloned mount | |
1545 | * is marked as 'shared and slave'. | |
1546 | * (*) the cloned mount is made a slave of the same master as that of the | |
1547 | * source mount. | |
1548 | * | |
9676f0c6 RP |
1549 | * --------------------------------------------------------------------------- |
1550 | * | MOVE MOUNT OPERATION | | |
1551 | * |************************************************************************** | |
1552 | * | source-->| shared | private | slave | unbindable | | |
1553 | * | dest | | | | | | |
1554 | * | | | | | | | | |
1555 | * | v | | | | | | |
1556 | * |************************************************************************** | |
1557 | * | shared | shared (+) | shared (+) | shared(+++) | invalid | | |
1558 | * | | | | | | | |
1559 | * |non-shared| shared (+*) | private | slave (*) | unbindable | | |
1560 | * *************************************************************************** | |
5afe0022 RP |
1561 | * |
1562 | * (+) the mount is moved to the destination. And is then propagated to | |
1563 | * all the mounts in the propagation tree of the destination mount. | |
21444403 | 1564 | * (+*) the mount is moved to the destination. |
5afe0022 RP |
1565 | * (+++) the mount is moved to the destination and is then propagated to |
1566 | * all the mounts belonging to the destination mount's propagation tree. | |
1567 | * the mount is marked as 'shared and slave'. | |
1568 | * (*) the mount continues to be a slave at the new location. | |
b90fa9ae RP |
1569 | * |
1570 | * if the source mount is a tree, the operations explained above is | |
1571 | * applied to each mount in the tree. | |
1572 | * Must be called without spinlocks held, since this function can sleep | |
1573 | * in allocations. | |
1574 | */ | |
1575 | static int attach_recursive_mnt(struct vfsmount *source_mnt, | |
1a390689 | 1576 | struct path *path, struct path *parent_path) |
b90fa9ae RP |
1577 | { |
1578 | LIST_HEAD(tree_list); | |
1a390689 AV |
1579 | struct vfsmount *dest_mnt = path->mnt; |
1580 | struct dentry *dest_dentry = path->dentry; | |
b90fa9ae | 1581 | struct vfsmount *child, *p; |
719f5d7f | 1582 | int err; |
b90fa9ae | 1583 | |
719f5d7f MS |
1584 | if (IS_MNT_SHARED(dest_mnt)) { |
1585 | err = invent_group_ids(source_mnt, true); | |
1586 | if (err) | |
1587 | goto out; | |
1588 | } | |
1589 | err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); | |
1590 | if (err) | |
1591 | goto out_cleanup_ids; | |
b90fa9ae | 1592 | |
99b7db7b | 1593 | br_write_lock(vfsmount_lock); |
df1a1ad2 | 1594 | |
b90fa9ae RP |
1595 | if (IS_MNT_SHARED(dest_mnt)) { |
1596 | for (p = source_mnt; p; p = next_mnt(p, source_mnt)) | |
1597 | set_mnt_shared(p); | |
1598 | } | |
1a390689 AV |
1599 | if (parent_path) { |
1600 | detach_mnt(source_mnt, parent_path); | |
1601 | attach_mnt(source_mnt, path); | |
e5d67f07 | 1602 | touch_mnt_namespace(parent_path->mnt->mnt_ns); |
21444403 RP |
1603 | } else { |
1604 | mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); | |
1605 | commit_tree(source_mnt); | |
1606 | } | |
b90fa9ae RP |
1607 | |
1608 | list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { | |
1609 | list_del_init(&child->mnt_hash); | |
1610 | commit_tree(child); | |
1611 | } | |
99b7db7b NP |
1612 | br_write_unlock(vfsmount_lock); |
1613 | ||
b90fa9ae | 1614 | return 0; |
719f5d7f MS |
1615 | |
1616 | out_cleanup_ids: | |
1617 | if (IS_MNT_SHARED(dest_mnt)) | |
1618 | cleanup_group_ids(source_mnt, NULL); | |
1619 | out: | |
1620 | return err; | |
b90fa9ae RP |
1621 | } |
1622 | ||
b12cea91 AV |
1623 | static int lock_mount(struct path *path) |
1624 | { | |
1625 | struct vfsmount *mnt; | |
1626 | retry: | |
1627 | mutex_lock(&path->dentry->d_inode->i_mutex); | |
1628 | if (unlikely(cant_mount(path->dentry))) { | |
1629 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
1630 | return -ENOENT; | |
1631 | } | |
1632 | down_write(&namespace_sem); | |
1633 | mnt = lookup_mnt(path); | |
1634 | if (likely(!mnt)) | |
1635 | return 0; | |
1636 | up_write(&namespace_sem); | |
1637 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
1638 | path_put(path); | |
1639 | path->mnt = mnt; | |
1640 | path->dentry = dget(mnt->mnt_root); | |
1641 | goto retry; | |
1642 | } | |
1643 | ||
1644 | static void unlock_mount(struct path *path) | |
1645 | { | |
1646 | up_write(&namespace_sem); | |
1647 | mutex_unlock(&path->dentry->d_inode->i_mutex); | |
1648 | } | |
1649 | ||
8c3ee42e | 1650 | static int graft_tree(struct vfsmount *mnt, struct path *path) |
1da177e4 | 1651 | { |
1da177e4 LT |
1652 | if (mnt->mnt_sb->s_flags & MS_NOUSER) |
1653 | return -EINVAL; | |
1654 | ||
8c3ee42e | 1655 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
1da177e4 LT |
1656 | S_ISDIR(mnt->mnt_root->d_inode->i_mode)) |
1657 | return -ENOTDIR; | |
1658 | ||
b12cea91 AV |
1659 | if (d_unlinked(path->dentry)) |
1660 | return -ENOENT; | |
1da177e4 | 1661 | |
b12cea91 | 1662 | return attach_recursive_mnt(mnt, path, NULL); |
1da177e4 LT |
1663 | } |
1664 | ||
7a2e8a8f VA |
1665 | /* |
1666 | * Sanity check the flags to change_mnt_propagation. | |
1667 | */ | |
1668 | ||
1669 | static int flags_to_propagation_type(int flags) | |
1670 | { | |
7c6e984d | 1671 | int type = flags & ~(MS_REC | MS_SILENT); |
7a2e8a8f VA |
1672 | |
1673 | /* Fail if any non-propagation flags are set */ | |
1674 | if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) | |
1675 | return 0; | |
1676 | /* Only one propagation flag should be set */ | |
1677 | if (!is_power_of_2(type)) | |
1678 | return 0; | |
1679 | return type; | |
1680 | } | |
1681 | ||
07b20889 RP |
1682 | /* |
1683 | * recursively change the type of the mountpoint. | |
1684 | */ | |
0a0d8a46 | 1685 | static int do_change_type(struct path *path, int flag) |
07b20889 | 1686 | { |
2d92ab3c | 1687 | struct vfsmount *m, *mnt = path->mnt; |
07b20889 | 1688 | int recurse = flag & MS_REC; |
7a2e8a8f | 1689 | int type; |
719f5d7f | 1690 | int err = 0; |
07b20889 | 1691 | |
ee6f9582 MS |
1692 | if (!capable(CAP_SYS_ADMIN)) |
1693 | return -EPERM; | |
1694 | ||
2d92ab3c | 1695 | if (path->dentry != path->mnt->mnt_root) |
07b20889 RP |
1696 | return -EINVAL; |
1697 | ||
7a2e8a8f VA |
1698 | type = flags_to_propagation_type(flag); |
1699 | if (!type) | |
1700 | return -EINVAL; | |
1701 | ||
07b20889 | 1702 | down_write(&namespace_sem); |
719f5d7f MS |
1703 | if (type == MS_SHARED) { |
1704 | err = invent_group_ids(mnt, recurse); | |
1705 | if (err) | |
1706 | goto out_unlock; | |
1707 | } | |
1708 | ||
99b7db7b | 1709 | br_write_lock(vfsmount_lock); |
07b20889 RP |
1710 | for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) |
1711 | change_mnt_propagation(m, type); | |
99b7db7b | 1712 | br_write_unlock(vfsmount_lock); |
719f5d7f MS |
1713 | |
1714 | out_unlock: | |
07b20889 | 1715 | up_write(&namespace_sem); |
719f5d7f | 1716 | return err; |
07b20889 RP |
1717 | } |
1718 | ||
1da177e4 LT |
1719 | /* |
1720 | * do loopback mount. | |
1721 | */ | |
0a0d8a46 | 1722 | static int do_loopback(struct path *path, char *old_name, |
2dafe1c4 | 1723 | int recurse) |
1da177e4 | 1724 | { |
b12cea91 | 1725 | LIST_HEAD(umount_list); |
2d92ab3c | 1726 | struct path old_path; |
1da177e4 | 1727 | struct vfsmount *mnt = NULL; |
2d92ab3c | 1728 | int err = mount_is_safe(path); |
1da177e4 LT |
1729 | if (err) |
1730 | return err; | |
1731 | if (!old_name || !*old_name) | |
1732 | return -EINVAL; | |
815d405c | 1733 | err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); |
1da177e4 LT |
1734 | if (err) |
1735 | return err; | |
1736 | ||
b12cea91 AV |
1737 | err = lock_mount(path); |
1738 | if (err) | |
1739 | goto out; | |
1740 | ||
1da177e4 | 1741 | err = -EINVAL; |
2d92ab3c | 1742 | if (IS_MNT_UNBINDABLE(old_path.mnt)) |
b12cea91 | 1743 | goto out2; |
9676f0c6 | 1744 | |
2d92ab3c | 1745 | if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) |
b12cea91 | 1746 | goto out2; |
1da177e4 | 1747 | |
ccd48bc7 AV |
1748 | err = -ENOMEM; |
1749 | if (recurse) | |
2d92ab3c | 1750 | mnt = copy_tree(old_path.mnt, old_path.dentry, 0); |
ccd48bc7 | 1751 | else |
2d92ab3c | 1752 | mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); |
ccd48bc7 AV |
1753 | |
1754 | if (!mnt) | |
b12cea91 | 1755 | goto out2; |
ccd48bc7 | 1756 | |
2d92ab3c | 1757 | err = graft_tree(mnt, path); |
ccd48bc7 | 1758 | if (err) { |
99b7db7b | 1759 | br_write_lock(vfsmount_lock); |
a05964f3 | 1760 | umount_tree(mnt, 0, &umount_list); |
99b7db7b | 1761 | br_write_unlock(vfsmount_lock); |
5b83d2c5 | 1762 | } |
b12cea91 AV |
1763 | out2: |
1764 | unlock_mount(path); | |
1765 | release_mounts(&umount_list); | |
ccd48bc7 | 1766 | out: |
2d92ab3c | 1767 | path_put(&old_path); |
1da177e4 LT |
1768 | return err; |
1769 | } | |
1770 | ||
2e4b7fcd DH |
1771 | static int change_mount_flags(struct vfsmount *mnt, int ms_flags) |
1772 | { | |
1773 | int error = 0; | |
1774 | int readonly_request = 0; | |
1775 | ||
1776 | if (ms_flags & MS_RDONLY) | |
1777 | readonly_request = 1; | |
1778 | if (readonly_request == __mnt_is_readonly(mnt)) | |
1779 | return 0; | |
1780 | ||
1781 | if (readonly_request) | |
1782 | error = mnt_make_readonly(mnt); | |
1783 | else | |
1784 | __mnt_unmake_readonly(mnt); | |
1785 | return error; | |
1786 | } | |
1787 | ||
1da177e4 LT |
1788 | /* |
1789 | * change filesystem flags. dir should be a physical root of filesystem. | |
1790 | * If you've mounted a non-root directory somewhere and want to do remount | |
1791 | * on it - tough luck. | |
1792 | */ | |
0a0d8a46 | 1793 | static int do_remount(struct path *path, int flags, int mnt_flags, |
1da177e4 LT |
1794 | void *data) |
1795 | { | |
1796 | int err; | |
2d92ab3c | 1797 | struct super_block *sb = path->mnt->mnt_sb; |
1da177e4 LT |
1798 | |
1799 | if (!capable(CAP_SYS_ADMIN)) | |
1800 | return -EPERM; | |
1801 | ||
2d92ab3c | 1802 | if (!check_mnt(path->mnt)) |
1da177e4 LT |
1803 | return -EINVAL; |
1804 | ||
2d92ab3c | 1805 | if (path->dentry != path->mnt->mnt_root) |
1da177e4 LT |
1806 | return -EINVAL; |
1807 | ||
ff36fe2c EP |
1808 | err = security_sb_remount(sb, data); |
1809 | if (err) | |
1810 | return err; | |
1811 | ||
1da177e4 | 1812 | down_write(&sb->s_umount); |
2e4b7fcd | 1813 | if (flags & MS_BIND) |
2d92ab3c | 1814 | err = change_mount_flags(path->mnt, flags); |
4aa98cf7 | 1815 | else |
2e4b7fcd | 1816 | err = do_remount_sb(sb, flags, data, 0); |
7b43a79f | 1817 | if (!err) { |
99b7db7b | 1818 | br_write_lock(vfsmount_lock); |
495d6c9c | 1819 | mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK; |
2d92ab3c | 1820 | path->mnt->mnt_flags = mnt_flags; |
99b7db7b | 1821 | br_write_unlock(vfsmount_lock); |
7b43a79f | 1822 | } |
1da177e4 | 1823 | up_write(&sb->s_umount); |
0e55a7cc | 1824 | if (!err) { |
99b7db7b | 1825 | br_write_lock(vfsmount_lock); |
0e55a7cc | 1826 | touch_mnt_namespace(path->mnt->mnt_ns); |
99b7db7b | 1827 | br_write_unlock(vfsmount_lock); |
0e55a7cc | 1828 | } |
1da177e4 LT |
1829 | return err; |
1830 | } | |
1831 | ||
9676f0c6 RP |
1832 | static inline int tree_contains_unbindable(struct vfsmount *mnt) |
1833 | { | |
1834 | struct vfsmount *p; | |
1835 | for (p = mnt; p; p = next_mnt(p, mnt)) { | |
1836 | if (IS_MNT_UNBINDABLE(p)) | |
1837 | return 1; | |
1838 | } | |
1839 | return 0; | |
1840 | } | |
1841 | ||
0a0d8a46 | 1842 | static int do_move_mount(struct path *path, char *old_name) |
1da177e4 | 1843 | { |
2d92ab3c | 1844 | struct path old_path, parent_path; |
1da177e4 LT |
1845 | struct vfsmount *p; |
1846 | int err = 0; | |
1847 | if (!capable(CAP_SYS_ADMIN)) | |
1848 | return -EPERM; | |
1849 | if (!old_name || !*old_name) | |
1850 | return -EINVAL; | |
2d92ab3c | 1851 | err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); |
1da177e4 LT |
1852 | if (err) |
1853 | return err; | |
1854 | ||
b12cea91 | 1855 | err = lock_mount(path); |
cc53ce53 DH |
1856 | if (err < 0) |
1857 | goto out; | |
1858 | ||
1da177e4 | 1859 | err = -EINVAL; |
2d92ab3c | 1860 | if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) |
1da177e4 LT |
1861 | goto out1; |
1862 | ||
f3da392e | 1863 | if (d_unlinked(path->dentry)) |
21444403 | 1864 | goto out1; |
1da177e4 LT |
1865 | |
1866 | err = -EINVAL; | |
2d92ab3c | 1867 | if (old_path.dentry != old_path.mnt->mnt_root) |
21444403 | 1868 | goto out1; |
1da177e4 | 1869 | |
2d92ab3c | 1870 | if (old_path.mnt == old_path.mnt->mnt_parent) |
21444403 | 1871 | goto out1; |
1da177e4 | 1872 | |
2d92ab3c AV |
1873 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
1874 | S_ISDIR(old_path.dentry->d_inode->i_mode)) | |
21444403 RP |
1875 | goto out1; |
1876 | /* | |
1877 | * Don't move a mount residing in a shared parent. | |
1878 | */ | |
2d92ab3c AV |
1879 | if (old_path.mnt->mnt_parent && |
1880 | IS_MNT_SHARED(old_path.mnt->mnt_parent)) | |
21444403 | 1881 | goto out1; |
9676f0c6 RP |
1882 | /* |
1883 | * Don't move a mount tree containing unbindable mounts to a destination | |
1884 | * mount which is shared. | |
1885 | */ | |
2d92ab3c AV |
1886 | if (IS_MNT_SHARED(path->mnt) && |
1887 | tree_contains_unbindable(old_path.mnt)) | |
9676f0c6 | 1888 | goto out1; |
1da177e4 | 1889 | err = -ELOOP; |
2d92ab3c AV |
1890 | for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent) |
1891 | if (p == old_path.mnt) | |
21444403 | 1892 | goto out1; |
1da177e4 | 1893 | |
2d92ab3c | 1894 | err = attach_recursive_mnt(old_path.mnt, path, &parent_path); |
4ac91378 | 1895 | if (err) |
21444403 | 1896 | goto out1; |
1da177e4 LT |
1897 | |
1898 | /* if the mount is moved, it should no longer be expire | |
1899 | * automatically */ | |
2d92ab3c | 1900 | list_del_init(&old_path.mnt->mnt_expire); |
1da177e4 | 1901 | out1: |
b12cea91 | 1902 | unlock_mount(path); |
1da177e4 | 1903 | out: |
1da177e4 | 1904 | if (!err) |
1a390689 | 1905 | path_put(&parent_path); |
2d92ab3c | 1906 | path_put(&old_path); |
1da177e4 LT |
1907 | return err; |
1908 | } | |
1909 | ||
9d412a43 AV |
1910 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
1911 | { | |
1912 | int err; | |
1913 | const char *subtype = strchr(fstype, '.'); | |
1914 | if (subtype) { | |
1915 | subtype++; | |
1916 | err = -EINVAL; | |
1917 | if (!subtype[0]) | |
1918 | goto err; | |
1919 | } else | |
1920 | subtype = ""; | |
1921 | ||
1922 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
1923 | err = -ENOMEM; | |
1924 | if (!mnt->mnt_sb->s_subtype) | |
1925 | goto err; | |
1926 | return mnt; | |
1927 | ||
1928 | err: | |
1929 | mntput(mnt); | |
1930 | return ERR_PTR(err); | |
1931 | } | |
1932 | ||
1933 | struct vfsmount * | |
1934 | do_kern_mount(const char *fstype, int flags, const char *name, void *data) | |
1935 | { | |
1936 | struct file_system_type *type = get_fs_type(fstype); | |
1937 | struct vfsmount *mnt; | |
1938 | if (!type) | |
1939 | return ERR_PTR(-ENODEV); | |
1940 | mnt = vfs_kern_mount(type, flags, name, data); | |
1941 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && | |
1942 | !mnt->mnt_sb->s_subtype) | |
1943 | mnt = fs_set_subtype(mnt, fstype); | |
1944 | put_filesystem(type); | |
1945 | return mnt; | |
1946 | } | |
1947 | EXPORT_SYMBOL_GPL(do_kern_mount); | |
1948 | ||
1949 | /* | |
1950 | * add a mount into a namespace's mount tree | |
1951 | */ | |
1952 | static int do_add_mount(struct vfsmount *newmnt, struct path *path, int mnt_flags) | |
1953 | { | |
1954 | int err; | |
1955 | ||
1956 | mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); | |
1957 | ||
b12cea91 AV |
1958 | err = lock_mount(path); |
1959 | if (err) | |
1960 | return err; | |
9d412a43 AV |
1961 | |
1962 | err = -EINVAL; | |
1963 | if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) | |
1964 | goto unlock; | |
1965 | ||
1966 | /* Refuse the same filesystem on the same mount point */ | |
1967 | err = -EBUSY; | |
1968 | if (path->mnt->mnt_sb == newmnt->mnt_sb && | |
1969 | path->mnt->mnt_root == path->dentry) | |
1970 | goto unlock; | |
1971 | ||
1972 | err = -EINVAL; | |
1973 | if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) | |
1974 | goto unlock; | |
1975 | ||
1976 | newmnt->mnt_flags = mnt_flags; | |
1977 | err = graft_tree(newmnt, path); | |
1978 | ||
1979 | unlock: | |
b12cea91 | 1980 | unlock_mount(path); |
9d412a43 AV |
1981 | return err; |
1982 | } | |
b1e75df4 | 1983 | |
1da177e4 LT |
1984 | /* |
1985 | * create a new mount for userspace and request it to be added into the | |
1986 | * namespace's tree | |
1987 | */ | |
0a0d8a46 | 1988 | static int do_new_mount(struct path *path, char *type, int flags, |
1da177e4 LT |
1989 | int mnt_flags, char *name, void *data) |
1990 | { | |
1991 | struct vfsmount *mnt; | |
15f9a3f3 | 1992 | int err; |
1da177e4 | 1993 | |
eca6f534 | 1994 | if (!type) |
1da177e4 LT |
1995 | return -EINVAL; |
1996 | ||
1997 | /* we need capabilities... */ | |
1998 | if (!capable(CAP_SYS_ADMIN)) | |
1999 | return -EPERM; | |
2000 | ||
2001 | mnt = do_kern_mount(type, flags, name, data); | |
2002 | if (IS_ERR(mnt)) | |
2003 | return PTR_ERR(mnt); | |
2004 | ||
15f9a3f3 AV |
2005 | err = do_add_mount(mnt, path, mnt_flags); |
2006 | if (err) | |
2007 | mntput(mnt); | |
2008 | return err; | |
1da177e4 LT |
2009 | } |
2010 | ||
19a167af AV |
2011 | int finish_automount(struct vfsmount *m, struct path *path) |
2012 | { | |
2013 | int err; | |
2014 | /* The new mount record should have at least 2 refs to prevent it being | |
2015 | * expired before we get a chance to add it | |
2016 | */ | |
2017 | BUG_ON(mnt_get_count(m) < 2); | |
2018 | ||
2019 | if (m->mnt_sb == path->mnt->mnt_sb && | |
2020 | m->mnt_root == path->dentry) { | |
b1e75df4 AV |
2021 | err = -ELOOP; |
2022 | goto fail; | |
19a167af AV |
2023 | } |
2024 | ||
19a167af | 2025 | err = do_add_mount(m, path, path->mnt->mnt_flags | MNT_SHRINKABLE); |
b1e75df4 AV |
2026 | if (!err) |
2027 | return 0; | |
2028 | fail: | |
2029 | /* remove m from any expiration list it may be on */ | |
2030 | if (!list_empty(&m->mnt_expire)) { | |
2031 | down_write(&namespace_sem); | |
2032 | br_write_lock(vfsmount_lock); | |
2033 | list_del_init(&m->mnt_expire); | |
2034 | br_write_unlock(vfsmount_lock); | |
2035 | up_write(&namespace_sem); | |
19a167af | 2036 | } |
b1e75df4 AV |
2037 | mntput(m); |
2038 | mntput(m); | |
19a167af AV |
2039 | return err; |
2040 | } | |
2041 | ||
ea5b778a DH |
2042 | /** |
2043 | * mnt_set_expiry - Put a mount on an expiration list | |
2044 | * @mnt: The mount to list. | |
2045 | * @expiry_list: The list to add the mount to. | |
2046 | */ | |
2047 | void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) | |
2048 | { | |
2049 | down_write(&namespace_sem); | |
2050 | br_write_lock(vfsmount_lock); | |
2051 | ||
2052 | list_add_tail(&mnt->mnt_expire, expiry_list); | |
2053 | ||
2054 | br_write_unlock(vfsmount_lock); | |
2055 | up_write(&namespace_sem); | |
2056 | } | |
2057 | EXPORT_SYMBOL(mnt_set_expiry); | |
2058 | ||
1da177e4 LT |
2059 | /* |
2060 | * process a list of expirable mountpoints with the intent of discarding any | |
2061 | * mountpoints that aren't in use and haven't been touched since last we came | |
2062 | * here | |
2063 | */ | |
2064 | void mark_mounts_for_expiry(struct list_head *mounts) | |
2065 | { | |
1da177e4 LT |
2066 | struct vfsmount *mnt, *next; |
2067 | LIST_HEAD(graveyard); | |
bcc5c7d2 | 2068 | LIST_HEAD(umounts); |
1da177e4 LT |
2069 | |
2070 | if (list_empty(mounts)) | |
2071 | return; | |
2072 | ||
bcc5c7d2 | 2073 | down_write(&namespace_sem); |
99b7db7b | 2074 | br_write_lock(vfsmount_lock); |
1da177e4 LT |
2075 | |
2076 | /* extract from the expiration list every vfsmount that matches the | |
2077 | * following criteria: | |
2078 | * - only referenced by its parent vfsmount | |
2079 | * - still marked for expiry (marked on the last call here; marks are | |
2080 | * cleared by mntput()) | |
2081 | */ | |
55e700b9 | 2082 | list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { |
1da177e4 | 2083 | if (!xchg(&mnt->mnt_expiry_mark, 1) || |
bcc5c7d2 | 2084 | propagate_mount_busy(mnt, 1)) |
1da177e4 | 2085 | continue; |
55e700b9 | 2086 | list_move(&mnt->mnt_expire, &graveyard); |
1da177e4 | 2087 | } |
bcc5c7d2 AV |
2088 | while (!list_empty(&graveyard)) { |
2089 | mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); | |
2090 | touch_mnt_namespace(mnt->mnt_ns); | |
2091 | umount_tree(mnt, 1, &umounts); | |
2092 | } | |
99b7db7b | 2093 | br_write_unlock(vfsmount_lock); |
bcc5c7d2 AV |
2094 | up_write(&namespace_sem); |
2095 | ||
2096 | release_mounts(&umounts); | |
5528f911 TM |
2097 | } |
2098 | ||
2099 | EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); | |
2100 | ||
2101 | /* | |
2102 | * Ripoff of 'select_parent()' | |
2103 | * | |
2104 | * search the list of submounts for a given mountpoint, and move any | |
2105 | * shrinkable submounts to the 'graveyard' list. | |
2106 | */ | |
2107 | static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) | |
2108 | { | |
2109 | struct vfsmount *this_parent = parent; | |
2110 | struct list_head *next; | |
2111 | int found = 0; | |
2112 | ||
2113 | repeat: | |
2114 | next = this_parent->mnt_mounts.next; | |
2115 | resume: | |
2116 | while (next != &this_parent->mnt_mounts) { | |
2117 | struct list_head *tmp = next; | |
2118 | struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); | |
2119 | ||
2120 | next = tmp->next; | |
2121 | if (!(mnt->mnt_flags & MNT_SHRINKABLE)) | |
1da177e4 | 2122 | continue; |
5528f911 TM |
2123 | /* |
2124 | * Descend a level if the d_mounts list is non-empty. | |
2125 | */ | |
2126 | if (!list_empty(&mnt->mnt_mounts)) { | |
2127 | this_parent = mnt; | |
2128 | goto repeat; | |
2129 | } | |
1da177e4 | 2130 | |
5528f911 | 2131 | if (!propagate_mount_busy(mnt, 1)) { |
5528f911 TM |
2132 | list_move_tail(&mnt->mnt_expire, graveyard); |
2133 | found++; | |
2134 | } | |
1da177e4 | 2135 | } |
5528f911 TM |
2136 | /* |
2137 | * All done at this level ... ascend and resume the search | |
2138 | */ | |
2139 | if (this_parent != parent) { | |
2140 | next = this_parent->mnt_child.next; | |
2141 | this_parent = this_parent->mnt_parent; | |
2142 | goto resume; | |
2143 | } | |
2144 | return found; | |
2145 | } | |
2146 | ||
2147 | /* | |
2148 | * process a list of expirable mountpoints with the intent of discarding any | |
2149 | * submounts of a specific parent mountpoint | |
99b7db7b NP |
2150 | * |
2151 | * vfsmount_lock must be held for write | |
5528f911 | 2152 | */ |
c35038be | 2153 | static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) |
5528f911 TM |
2154 | { |
2155 | LIST_HEAD(graveyard); | |
c35038be | 2156 | struct vfsmount *m; |
5528f911 | 2157 | |
5528f911 | 2158 | /* extract submounts of 'mountpoint' from the expiration list */ |
c35038be | 2159 | while (select_submounts(mnt, &graveyard)) { |
bcc5c7d2 | 2160 | while (!list_empty(&graveyard)) { |
c35038be | 2161 | m = list_first_entry(&graveyard, struct vfsmount, |
bcc5c7d2 | 2162 | mnt_expire); |
afef80b3 EB |
2163 | touch_mnt_namespace(m->mnt_ns); |
2164 | umount_tree(m, 1, umounts); | |
bcc5c7d2 AV |
2165 | } |
2166 | } | |
1da177e4 LT |
2167 | } |
2168 | ||
1da177e4 LT |
2169 | /* |
2170 | * Some copy_from_user() implementations do not return the exact number of | |
2171 | * bytes remaining to copy on a fault. But copy_mount_options() requires that. | |
2172 | * Note that this function differs from copy_from_user() in that it will oops | |
2173 | * on bad values of `to', rather than returning a short copy. | |
2174 | */ | |
b58fed8b RP |
2175 | static long exact_copy_from_user(void *to, const void __user * from, |
2176 | unsigned long n) | |
1da177e4 LT |
2177 | { |
2178 | char *t = to; | |
2179 | const char __user *f = from; | |
2180 | char c; | |
2181 | ||
2182 | if (!access_ok(VERIFY_READ, from, n)) | |
2183 | return n; | |
2184 | ||
2185 | while (n) { | |
2186 | if (__get_user(c, f)) { | |
2187 | memset(t, 0, n); | |
2188 | break; | |
2189 | } | |
2190 | *t++ = c; | |
2191 | f++; | |
2192 | n--; | |
2193 | } | |
2194 | return n; | |
2195 | } | |
2196 | ||
b58fed8b | 2197 | int copy_mount_options(const void __user * data, unsigned long *where) |
1da177e4 LT |
2198 | { |
2199 | int i; | |
2200 | unsigned long page; | |
2201 | unsigned long size; | |
b58fed8b | 2202 | |
1da177e4 LT |
2203 | *where = 0; |
2204 | if (!data) | |
2205 | return 0; | |
2206 | ||
2207 | if (!(page = __get_free_page(GFP_KERNEL))) | |
2208 | return -ENOMEM; | |
2209 | ||
2210 | /* We only care that *some* data at the address the user | |
2211 | * gave us is valid. Just in case, we'll zero | |
2212 | * the remainder of the page. | |
2213 | */ | |
2214 | /* copy_from_user cannot cross TASK_SIZE ! */ | |
2215 | size = TASK_SIZE - (unsigned long)data; | |
2216 | if (size > PAGE_SIZE) | |
2217 | size = PAGE_SIZE; | |
2218 | ||
2219 | i = size - exact_copy_from_user((void *)page, data, size); | |
2220 | if (!i) { | |
b58fed8b | 2221 | free_page(page); |
1da177e4 LT |
2222 | return -EFAULT; |
2223 | } | |
2224 | if (i != PAGE_SIZE) | |
2225 | memset((char *)page + i, 0, PAGE_SIZE - i); | |
2226 | *where = page; | |
2227 | return 0; | |
2228 | } | |
2229 | ||
eca6f534 VN |
2230 | int copy_mount_string(const void __user *data, char **where) |
2231 | { | |
2232 | char *tmp; | |
2233 | ||
2234 | if (!data) { | |
2235 | *where = NULL; | |
2236 | return 0; | |
2237 | } | |
2238 | ||
2239 | tmp = strndup_user(data, PAGE_SIZE); | |
2240 | if (IS_ERR(tmp)) | |
2241 | return PTR_ERR(tmp); | |
2242 | ||
2243 | *where = tmp; | |
2244 | return 0; | |
2245 | } | |
2246 | ||
1da177e4 LT |
2247 | /* |
2248 | * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to | |
2249 | * be given to the mount() call (ie: read-only, no-dev, no-suid etc). | |
2250 | * | |
2251 | * data is a (void *) that can point to any structure up to | |
2252 | * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent | |
2253 | * information (or be NULL). | |
2254 | * | |
2255 | * Pre-0.97 versions of mount() didn't have a flags word. | |
2256 | * When the flags word was introduced its top half was required | |
2257 | * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. | |
2258 | * Therefore, if this magic number is present, it carries no information | |
2259 | * and must be discarded. | |
2260 | */ | |
b58fed8b | 2261 | long do_mount(char *dev_name, char *dir_name, char *type_page, |
1da177e4 LT |
2262 | unsigned long flags, void *data_page) |
2263 | { | |
2d92ab3c | 2264 | struct path path; |
1da177e4 LT |
2265 | int retval = 0; |
2266 | int mnt_flags = 0; | |
2267 | ||
2268 | /* Discard magic */ | |
2269 | if ((flags & MS_MGC_MSK) == MS_MGC_VAL) | |
2270 | flags &= ~MS_MGC_MSK; | |
2271 | ||
2272 | /* Basic sanity checks */ | |
2273 | ||
2274 | if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) | |
2275 | return -EINVAL; | |
1da177e4 LT |
2276 | |
2277 | if (data_page) | |
2278 | ((char *)data_page)[PAGE_SIZE - 1] = 0; | |
2279 | ||
a27ab9f2 TH |
2280 | /* ... and get the mountpoint */ |
2281 | retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); | |
2282 | if (retval) | |
2283 | return retval; | |
2284 | ||
2285 | retval = security_sb_mount(dev_name, &path, | |
2286 | type_page, flags, data_page); | |
2287 | if (retval) | |
2288 | goto dput_out; | |
2289 | ||
613cbe3d AK |
2290 | /* Default to relatime unless overriden */ |
2291 | if (!(flags & MS_NOATIME)) | |
2292 | mnt_flags |= MNT_RELATIME; | |
0a1c01c9 | 2293 | |
1da177e4 LT |
2294 | /* Separate the per-mountpoint flags */ |
2295 | if (flags & MS_NOSUID) | |
2296 | mnt_flags |= MNT_NOSUID; | |
2297 | if (flags & MS_NODEV) | |
2298 | mnt_flags |= MNT_NODEV; | |
2299 | if (flags & MS_NOEXEC) | |
2300 | mnt_flags |= MNT_NOEXEC; | |
fc33a7bb CH |
2301 | if (flags & MS_NOATIME) |
2302 | mnt_flags |= MNT_NOATIME; | |
2303 | if (flags & MS_NODIRATIME) | |
2304 | mnt_flags |= MNT_NODIRATIME; | |
d0adde57 MG |
2305 | if (flags & MS_STRICTATIME) |
2306 | mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); | |
2e4b7fcd DH |
2307 | if (flags & MS_RDONLY) |
2308 | mnt_flags |= MNT_READONLY; | |
fc33a7bb | 2309 | |
7a4dec53 | 2310 | flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | |
d0adde57 MG |
2311 | MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | |
2312 | MS_STRICTATIME); | |
1da177e4 | 2313 | |
1da177e4 | 2314 | if (flags & MS_REMOUNT) |
2d92ab3c | 2315 | retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, |
1da177e4 LT |
2316 | data_page); |
2317 | else if (flags & MS_BIND) | |
2d92ab3c | 2318 | retval = do_loopback(&path, dev_name, flags & MS_REC); |
9676f0c6 | 2319 | else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) |
2d92ab3c | 2320 | retval = do_change_type(&path, flags); |
1da177e4 | 2321 | else if (flags & MS_MOVE) |
2d92ab3c | 2322 | retval = do_move_mount(&path, dev_name); |
1da177e4 | 2323 | else |
2d92ab3c | 2324 | retval = do_new_mount(&path, type_page, flags, mnt_flags, |
1da177e4 LT |
2325 | dev_name, data_page); |
2326 | dput_out: | |
2d92ab3c | 2327 | path_put(&path); |
1da177e4 LT |
2328 | return retval; |
2329 | } | |
2330 | ||
cf8d2c11 TM |
2331 | static struct mnt_namespace *alloc_mnt_ns(void) |
2332 | { | |
2333 | struct mnt_namespace *new_ns; | |
2334 | ||
2335 | new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); | |
2336 | if (!new_ns) | |
2337 | return ERR_PTR(-ENOMEM); | |
2338 | atomic_set(&new_ns->count, 1); | |
2339 | new_ns->root = NULL; | |
2340 | INIT_LIST_HEAD(&new_ns->list); | |
2341 | init_waitqueue_head(&new_ns->poll); | |
2342 | new_ns->event = 0; | |
2343 | return new_ns; | |
2344 | } | |
2345 | ||
f03c6599 AV |
2346 | void mnt_make_longterm(struct vfsmount *mnt) |
2347 | { | |
7e3d0eb0 | 2348 | __mnt_make_longterm(mnt); |
f03c6599 AV |
2349 | } |
2350 | ||
2351 | void mnt_make_shortterm(struct vfsmount *mnt) | |
2352 | { | |
7e3d0eb0 | 2353 | #ifdef CONFIG_SMP |
f03c6599 AV |
2354 | if (atomic_add_unless(&mnt->mnt_longterm, -1, 1)) |
2355 | return; | |
2356 | br_write_lock(vfsmount_lock); | |
2357 | atomic_dec(&mnt->mnt_longterm); | |
2358 | br_write_unlock(vfsmount_lock); | |
7e3d0eb0 | 2359 | #endif |
f03c6599 AV |
2360 | } |
2361 | ||
741a2951 JD |
2362 | /* |
2363 | * Allocate a new namespace structure and populate it with contents | |
2364 | * copied from the namespace of the passed in task structure. | |
2365 | */ | |
e3222c4e | 2366 | static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, |
6b3286ed | 2367 | struct fs_struct *fs) |
1da177e4 | 2368 | { |
6b3286ed | 2369 | struct mnt_namespace *new_ns; |
7f2da1e7 | 2370 | struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; |
1da177e4 LT |
2371 | struct vfsmount *p, *q; |
2372 | ||
cf8d2c11 TM |
2373 | new_ns = alloc_mnt_ns(); |
2374 | if (IS_ERR(new_ns)) | |
2375 | return new_ns; | |
1da177e4 | 2376 | |
390c6843 | 2377 | down_write(&namespace_sem); |
1da177e4 | 2378 | /* First pass: copy the tree topology */ |
6b3286ed | 2379 | new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, |
9676f0c6 | 2380 | CL_COPY_ALL | CL_EXPIRE); |
1da177e4 | 2381 | if (!new_ns->root) { |
390c6843 | 2382 | up_write(&namespace_sem); |
1da177e4 | 2383 | kfree(new_ns); |
5cc4a034 | 2384 | return ERR_PTR(-ENOMEM); |
1da177e4 | 2385 | } |
99b7db7b | 2386 | br_write_lock(vfsmount_lock); |
1da177e4 | 2387 | list_add_tail(&new_ns->list, &new_ns->root->mnt_list); |
99b7db7b | 2388 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
2389 | |
2390 | /* | |
2391 | * Second pass: switch the tsk->fs->* elements and mark new vfsmounts | |
2392 | * as belonging to new namespace. We have already acquired a private | |
2393 | * fs_struct, so tsk->fs->lock is not needed. | |
2394 | */ | |
6b3286ed | 2395 | p = mnt_ns->root; |
1da177e4 LT |
2396 | q = new_ns->root; |
2397 | while (p) { | |
6b3286ed | 2398 | q->mnt_ns = new_ns; |
7e3d0eb0 | 2399 | __mnt_make_longterm(q); |
1da177e4 | 2400 | if (fs) { |
6ac08c39 | 2401 | if (p == fs->root.mnt) { |
f03c6599 | 2402 | fs->root.mnt = mntget(q); |
7e3d0eb0 | 2403 | __mnt_make_longterm(q); |
f03c6599 | 2404 | mnt_make_shortterm(p); |
1da177e4 | 2405 | rootmnt = p; |
1da177e4 | 2406 | } |
6ac08c39 | 2407 | if (p == fs->pwd.mnt) { |
f03c6599 | 2408 | fs->pwd.mnt = mntget(q); |
7e3d0eb0 | 2409 | __mnt_make_longterm(q); |
f03c6599 | 2410 | mnt_make_shortterm(p); |
1da177e4 | 2411 | pwdmnt = p; |
1da177e4 | 2412 | } |
1da177e4 | 2413 | } |
6b3286ed | 2414 | p = next_mnt(p, mnt_ns->root); |
1da177e4 LT |
2415 | q = next_mnt(q, new_ns->root); |
2416 | } | |
390c6843 | 2417 | up_write(&namespace_sem); |
1da177e4 | 2418 | |
1da177e4 | 2419 | if (rootmnt) |
f03c6599 | 2420 | mntput(rootmnt); |
1da177e4 | 2421 | if (pwdmnt) |
f03c6599 | 2422 | mntput(pwdmnt); |
1da177e4 | 2423 | |
741a2951 JD |
2424 | return new_ns; |
2425 | } | |
2426 | ||
213dd266 | 2427 | struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, |
e3222c4e | 2428 | struct fs_struct *new_fs) |
741a2951 | 2429 | { |
6b3286ed | 2430 | struct mnt_namespace *new_ns; |
741a2951 | 2431 | |
e3222c4e | 2432 | BUG_ON(!ns); |
6b3286ed | 2433 | get_mnt_ns(ns); |
741a2951 JD |
2434 | |
2435 | if (!(flags & CLONE_NEWNS)) | |
e3222c4e | 2436 | return ns; |
741a2951 | 2437 | |
e3222c4e | 2438 | new_ns = dup_mnt_ns(ns, new_fs); |
741a2951 | 2439 | |
6b3286ed | 2440 | put_mnt_ns(ns); |
e3222c4e | 2441 | return new_ns; |
1da177e4 LT |
2442 | } |
2443 | ||
cf8d2c11 TM |
2444 | /** |
2445 | * create_mnt_ns - creates a private namespace and adds a root filesystem | |
2446 | * @mnt: pointer to the new root filesystem mountpoint | |
2447 | */ | |
a2770d86 | 2448 | struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) |
cf8d2c11 TM |
2449 | { |
2450 | struct mnt_namespace *new_ns; | |
2451 | ||
2452 | new_ns = alloc_mnt_ns(); | |
2453 | if (!IS_ERR(new_ns)) { | |
2454 | mnt->mnt_ns = new_ns; | |
7e3d0eb0 | 2455 | __mnt_make_longterm(mnt); |
cf8d2c11 TM |
2456 | new_ns->root = mnt; |
2457 | list_add(&new_ns->list, &new_ns->root->mnt_list); | |
c1334495 AV |
2458 | } else { |
2459 | mntput(mnt); | |
cf8d2c11 TM |
2460 | } |
2461 | return new_ns; | |
2462 | } | |
a2770d86 | 2463 | EXPORT_SYMBOL(create_mnt_ns); |
cf8d2c11 | 2464 | |
ea441d11 AV |
2465 | struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) |
2466 | { | |
2467 | struct mnt_namespace *ns; | |
d31da0f0 | 2468 | struct super_block *s; |
ea441d11 AV |
2469 | struct path path; |
2470 | int err; | |
2471 | ||
2472 | ns = create_mnt_ns(mnt); | |
2473 | if (IS_ERR(ns)) | |
2474 | return ERR_CAST(ns); | |
2475 | ||
2476 | err = vfs_path_lookup(mnt->mnt_root, mnt, | |
2477 | name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); | |
2478 | ||
2479 | put_mnt_ns(ns); | |
2480 | ||
2481 | if (err) | |
2482 | return ERR_PTR(err); | |
2483 | ||
2484 | /* trade a vfsmount reference for active sb one */ | |
d31da0f0 AV |
2485 | s = path.mnt->mnt_sb; |
2486 | atomic_inc(&s->s_active); | |
ea441d11 AV |
2487 | mntput(path.mnt); |
2488 | /* lock the sucker */ | |
d31da0f0 | 2489 | down_write(&s->s_umount); |
ea441d11 AV |
2490 | /* ... and return the root of (sub)tree on it */ |
2491 | return path.dentry; | |
2492 | } | |
2493 | EXPORT_SYMBOL(mount_subtree); | |
2494 | ||
bdc480e3 HC |
2495 | SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, |
2496 | char __user *, type, unsigned long, flags, void __user *, data) | |
1da177e4 | 2497 | { |
eca6f534 VN |
2498 | int ret; |
2499 | char *kernel_type; | |
2500 | char *kernel_dir; | |
2501 | char *kernel_dev; | |
1da177e4 | 2502 | unsigned long data_page; |
1da177e4 | 2503 | |
eca6f534 VN |
2504 | ret = copy_mount_string(type, &kernel_type); |
2505 | if (ret < 0) | |
2506 | goto out_type; | |
1da177e4 | 2507 | |
eca6f534 VN |
2508 | kernel_dir = getname(dir_name); |
2509 | if (IS_ERR(kernel_dir)) { | |
2510 | ret = PTR_ERR(kernel_dir); | |
2511 | goto out_dir; | |
2512 | } | |
1da177e4 | 2513 | |
eca6f534 VN |
2514 | ret = copy_mount_string(dev_name, &kernel_dev); |
2515 | if (ret < 0) | |
2516 | goto out_dev; | |
1da177e4 | 2517 | |
eca6f534 VN |
2518 | ret = copy_mount_options(data, &data_page); |
2519 | if (ret < 0) | |
2520 | goto out_data; | |
1da177e4 | 2521 | |
eca6f534 VN |
2522 | ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags, |
2523 | (void *) data_page); | |
1da177e4 | 2524 | |
eca6f534 VN |
2525 | free_page(data_page); |
2526 | out_data: | |
2527 | kfree(kernel_dev); | |
2528 | out_dev: | |
2529 | putname(kernel_dir); | |
2530 | out_dir: | |
2531 | kfree(kernel_type); | |
2532 | out_type: | |
2533 | return ret; | |
1da177e4 LT |
2534 | } |
2535 | ||
1da177e4 LT |
2536 | /* |
2537 | * pivot_root Semantics: | |
2538 | * Moves the root file system of the current process to the directory put_old, | |
2539 | * makes new_root as the new root file system of the current process, and sets | |
2540 | * root/cwd of all processes which had them on the current root to new_root. | |
2541 | * | |
2542 | * Restrictions: | |
2543 | * The new_root and put_old must be directories, and must not be on the | |
2544 | * same file system as the current process root. The put_old must be | |
2545 | * underneath new_root, i.e. adding a non-zero number of /.. to the string | |
2546 | * pointed to by put_old must yield the same directory as new_root. No other | |
2547 | * file system may be mounted on put_old. After all, new_root is a mountpoint. | |
2548 | * | |
4a0d11fa NB |
2549 | * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. |
2550 | * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives | |
2551 | * in this situation. | |
2552 | * | |
1da177e4 LT |
2553 | * Notes: |
2554 | * - we don't move root/cwd if they are not at the root (reason: if something | |
2555 | * cared enough to change them, it's probably wrong to force them elsewhere) | |
2556 | * - it's okay to pick a root that isn't the root of a file system, e.g. | |
2557 | * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, | |
2558 | * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root | |
2559 | * first. | |
2560 | */ | |
3480b257 HC |
2561 | SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, |
2562 | const char __user *, put_old) | |
1da177e4 LT |
2563 | { |
2564 | struct vfsmount *tmp; | |
2d8f3038 | 2565 | struct path new, old, parent_path, root_parent, root; |
1da177e4 LT |
2566 | int error; |
2567 | ||
2568 | if (!capable(CAP_SYS_ADMIN)) | |
2569 | return -EPERM; | |
2570 | ||
2d8f3038 | 2571 | error = user_path_dir(new_root, &new); |
1da177e4 LT |
2572 | if (error) |
2573 | goto out0; | |
1da177e4 | 2574 | |
2d8f3038 | 2575 | error = user_path_dir(put_old, &old); |
1da177e4 LT |
2576 | if (error) |
2577 | goto out1; | |
2578 | ||
2d8f3038 | 2579 | error = security_sb_pivotroot(&old, &new); |
b12cea91 AV |
2580 | if (error) |
2581 | goto out2; | |
1da177e4 | 2582 | |
f7ad3c6b | 2583 | get_fs_root(current->fs, &root); |
b12cea91 AV |
2584 | error = lock_mount(&old); |
2585 | if (error) | |
2586 | goto out3; | |
2587 | ||
1da177e4 | 2588 | error = -EINVAL; |
2d8f3038 AV |
2589 | if (IS_MNT_SHARED(old.mnt) || |
2590 | IS_MNT_SHARED(new.mnt->mnt_parent) || | |
8c3ee42e | 2591 | IS_MNT_SHARED(root.mnt->mnt_parent)) |
b12cea91 | 2592 | goto out4; |
27cb1572 | 2593 | if (!check_mnt(root.mnt) || !check_mnt(new.mnt)) |
b12cea91 | 2594 | goto out4; |
1da177e4 | 2595 | error = -ENOENT; |
f3da392e | 2596 | if (d_unlinked(new.dentry)) |
b12cea91 | 2597 | goto out4; |
f3da392e | 2598 | if (d_unlinked(old.dentry)) |
b12cea91 | 2599 | goto out4; |
1da177e4 | 2600 | error = -EBUSY; |
2d8f3038 AV |
2601 | if (new.mnt == root.mnt || |
2602 | old.mnt == root.mnt) | |
b12cea91 | 2603 | goto out4; /* loop, on the same file system */ |
1da177e4 | 2604 | error = -EINVAL; |
8c3ee42e | 2605 | if (root.mnt->mnt_root != root.dentry) |
b12cea91 | 2606 | goto out4; /* not a mountpoint */ |
8c3ee42e | 2607 | if (root.mnt->mnt_parent == root.mnt) |
b12cea91 | 2608 | goto out4; /* not attached */ |
2d8f3038 | 2609 | if (new.mnt->mnt_root != new.dentry) |
b12cea91 | 2610 | goto out4; /* not a mountpoint */ |
2d8f3038 | 2611 | if (new.mnt->mnt_parent == new.mnt) |
b12cea91 | 2612 | goto out4; /* not attached */ |
4ac91378 | 2613 | /* make sure we can reach put_old from new_root */ |
2d8f3038 | 2614 | tmp = old.mnt; |
2d8f3038 | 2615 | if (tmp != new.mnt) { |
1da177e4 LT |
2616 | for (;;) { |
2617 | if (tmp->mnt_parent == tmp) | |
b12cea91 | 2618 | goto out4; /* already mounted on put_old */ |
2d8f3038 | 2619 | if (tmp->mnt_parent == new.mnt) |
1da177e4 LT |
2620 | break; |
2621 | tmp = tmp->mnt_parent; | |
2622 | } | |
2d8f3038 | 2623 | if (!is_subdir(tmp->mnt_mountpoint, new.dentry)) |
b12cea91 | 2624 | goto out4; |
2d8f3038 | 2625 | } else if (!is_subdir(old.dentry, new.dentry)) |
b12cea91 | 2626 | goto out4; |
27cb1572 | 2627 | br_write_lock(vfsmount_lock); |
2d8f3038 | 2628 | detach_mnt(new.mnt, &parent_path); |
8c3ee42e | 2629 | detach_mnt(root.mnt, &root_parent); |
4ac91378 | 2630 | /* mount old root on put_old */ |
2d8f3038 | 2631 | attach_mnt(root.mnt, &old); |
4ac91378 | 2632 | /* mount new_root on / */ |
2d8f3038 | 2633 | attach_mnt(new.mnt, &root_parent); |
6b3286ed | 2634 | touch_mnt_namespace(current->nsproxy->mnt_ns); |
99b7db7b | 2635 | br_write_unlock(vfsmount_lock); |
2d8f3038 | 2636 | chroot_fs_refs(&root, &new); |
1da177e4 | 2637 | error = 0; |
b12cea91 AV |
2638 | out4: |
2639 | unlock_mount(&old); | |
2640 | if (!error) { | |
2641 | path_put(&root_parent); | |
2642 | path_put(&parent_path); | |
2643 | } | |
2644 | out3: | |
8c3ee42e | 2645 | path_put(&root); |
b12cea91 | 2646 | out2: |
2d8f3038 | 2647 | path_put(&old); |
1da177e4 | 2648 | out1: |
2d8f3038 | 2649 | path_put(&new); |
1da177e4 | 2650 | out0: |
1da177e4 | 2651 | return error; |
1da177e4 LT |
2652 | } |
2653 | ||
2654 | static void __init init_mount_tree(void) | |
2655 | { | |
2656 | struct vfsmount *mnt; | |
6b3286ed | 2657 | struct mnt_namespace *ns; |
ac748a09 | 2658 | struct path root; |
1da177e4 LT |
2659 | |
2660 | mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); | |
2661 | if (IS_ERR(mnt)) | |
2662 | panic("Can't create rootfs"); | |
b3e19d92 | 2663 | |
3b22edc5 TM |
2664 | ns = create_mnt_ns(mnt); |
2665 | if (IS_ERR(ns)) | |
1da177e4 | 2666 | panic("Can't allocate initial namespace"); |
6b3286ed KK |
2667 | |
2668 | init_task.nsproxy->mnt_ns = ns; | |
2669 | get_mnt_ns(ns); | |
2670 | ||
ac748a09 JB |
2671 | root.mnt = ns->root; |
2672 | root.dentry = ns->root->mnt_root; | |
2673 | ||
2674 | set_fs_pwd(current->fs, &root); | |
2675 | set_fs_root(current->fs, &root); | |
1da177e4 LT |
2676 | } |
2677 | ||
74bf17cf | 2678 | void __init mnt_init(void) |
1da177e4 | 2679 | { |
13f14b4d | 2680 | unsigned u; |
15a67dd8 | 2681 | int err; |
1da177e4 | 2682 | |
390c6843 RP |
2683 | init_rwsem(&namespace_sem); |
2684 | ||
1da177e4 | 2685 | mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount), |
20c2df83 | 2686 | 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); |
1da177e4 | 2687 | |
b58fed8b | 2688 | mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
1da177e4 LT |
2689 | |
2690 | if (!mount_hashtable) | |
2691 | panic("Failed to allocate mount hash table\n"); | |
2692 | ||
80cdc6da | 2693 | printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); |
13f14b4d ED |
2694 | |
2695 | for (u = 0; u < HASH_SIZE; u++) | |
2696 | INIT_LIST_HEAD(&mount_hashtable[u]); | |
1da177e4 | 2697 | |
99b7db7b NP |
2698 | br_lock_init(vfsmount_lock); |
2699 | ||
15a67dd8 RD |
2700 | err = sysfs_init(); |
2701 | if (err) | |
2702 | printk(KERN_WARNING "%s: sysfs_init error: %d\n", | |
8e24eea7 | 2703 | __func__, err); |
00d26666 GKH |
2704 | fs_kobj = kobject_create_and_add("fs", NULL); |
2705 | if (!fs_kobj) | |
8e24eea7 | 2706 | printk(KERN_WARNING "%s: kobj create error\n", __func__); |
1da177e4 LT |
2707 | init_rootfs(); |
2708 | init_mount_tree(); | |
2709 | } | |
2710 | ||
616511d0 | 2711 | void put_mnt_ns(struct mnt_namespace *ns) |
1da177e4 | 2712 | { |
70fbcdf4 | 2713 | LIST_HEAD(umount_list); |
616511d0 | 2714 | |
d498b25a | 2715 | if (!atomic_dec_and_test(&ns->count)) |
616511d0 | 2716 | return; |
390c6843 | 2717 | down_write(&namespace_sem); |
99b7db7b | 2718 | br_write_lock(vfsmount_lock); |
d498b25a | 2719 | umount_tree(ns->root, 0, &umount_list); |
99b7db7b | 2720 | br_write_unlock(vfsmount_lock); |
390c6843 | 2721 | up_write(&namespace_sem); |
70fbcdf4 | 2722 | release_mounts(&umount_list); |
6b3286ed | 2723 | kfree(ns); |
1da177e4 | 2724 | } |
cf8d2c11 | 2725 | EXPORT_SYMBOL(put_mnt_ns); |
9d412a43 AV |
2726 | |
2727 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) | |
2728 | { | |
423e0ab0 TC |
2729 | struct vfsmount *mnt; |
2730 | mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); | |
2731 | if (!IS_ERR(mnt)) { | |
2732 | /* | |
2733 | * it is a longterm mount, don't release mnt until | |
2734 | * we unmount before file sys is unregistered | |
2735 | */ | |
2736 | mnt_make_longterm(mnt); | |
2737 | } | |
2738 | return mnt; | |
9d412a43 AV |
2739 | } |
2740 | EXPORT_SYMBOL_GPL(kern_mount_data); | |
423e0ab0 TC |
2741 | |
2742 | void kern_unmount(struct vfsmount *mnt) | |
2743 | { | |
2744 | /* release long term mount so mount point can be released */ | |
2745 | if (!IS_ERR_OR_NULL(mnt)) { | |
2746 | mnt_make_shortterm(mnt); | |
2747 | mntput(mnt); | |
2748 | } | |
2749 | } | |
2750 | EXPORT_SYMBOL(kern_unmount); | |
02125a82 AV |
2751 | |
2752 | bool our_mnt(struct vfsmount *mnt) | |
2753 | { | |
2754 | return check_mnt(mnt); | |
2755 | } |