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