]>
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 | ||
155 | static inline void mnt_set_count(struct vfsmount *mnt, int n) | |
156 | { | |
157 | #ifdef CONFIG_SMP | |
158 | this_cpu_write(mnt->mnt_pcp->mnt_count, n); | |
159 | #else | |
160 | mnt->mnt_count = n; | |
161 | #endif | |
162 | } | |
163 | ||
164 | /* | |
165 | * vfsmount lock must be held for read | |
166 | */ | |
167 | static inline void mnt_inc_count(struct vfsmount *mnt) | |
168 | { | |
169 | mnt_add_count(mnt, 1); | |
170 | } | |
171 | ||
172 | /* | |
173 | * vfsmount lock must be held for read | |
174 | */ | |
175 | static inline void mnt_dec_count(struct vfsmount *mnt) | |
176 | { | |
177 | mnt_add_count(mnt, -1); | |
178 | } | |
179 | ||
180 | /* | |
181 | * vfsmount lock must be held for write | |
182 | */ | |
183 | unsigned int mnt_get_count(struct vfsmount *mnt) | |
184 | { | |
185 | #ifdef CONFIG_SMP | |
f03c6599 | 186 | unsigned int count = 0; |
b3e19d92 NP |
187 | int cpu; |
188 | ||
189 | for_each_possible_cpu(cpu) { | |
190 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count; | |
191 | } | |
192 | ||
193 | return count; | |
194 | #else | |
195 | return mnt->mnt_count; | |
196 | #endif | |
197 | } | |
198 | ||
1da177e4 LT |
199 | struct vfsmount *alloc_vfsmnt(const char *name) |
200 | { | |
c3762229 | 201 | struct vfsmount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); |
1da177e4 | 202 | if (mnt) { |
73cd49ec MS |
203 | int err; |
204 | ||
205 | err = mnt_alloc_id(mnt); | |
88b38782 LZ |
206 | if (err) |
207 | goto out_free_cache; | |
208 | ||
209 | if (name) { | |
210 | mnt->mnt_devname = kstrdup(name, GFP_KERNEL); | |
211 | if (!mnt->mnt_devname) | |
212 | goto out_free_id; | |
73cd49ec MS |
213 | } |
214 | ||
b3e19d92 NP |
215 | #ifdef CONFIG_SMP |
216 | mnt->mnt_pcp = alloc_percpu(struct mnt_pcp); | |
217 | if (!mnt->mnt_pcp) | |
218 | goto out_free_devname; | |
219 | ||
f03c6599 | 220 | this_cpu_add(mnt->mnt_pcp->mnt_count, 1); |
b3e19d92 NP |
221 | #else |
222 | mnt->mnt_count = 1; | |
223 | mnt->mnt_writers = 0; | |
224 | #endif | |
225 | ||
1da177e4 LT |
226 | INIT_LIST_HEAD(&mnt->mnt_hash); |
227 | INIT_LIST_HEAD(&mnt->mnt_child); | |
228 | INIT_LIST_HEAD(&mnt->mnt_mounts); | |
229 | INIT_LIST_HEAD(&mnt->mnt_list); | |
55e700b9 | 230 | INIT_LIST_HEAD(&mnt->mnt_expire); |
03e06e68 | 231 | INIT_LIST_HEAD(&mnt->mnt_share); |
a58b0eb8 RP |
232 | INIT_LIST_HEAD(&mnt->mnt_slave_list); |
233 | INIT_LIST_HEAD(&mnt->mnt_slave); | |
2504c5d6 AG |
234 | #ifdef CONFIG_FSNOTIFY |
235 | INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks); | |
d3ef3d73 | 236 | #endif |
1da177e4 LT |
237 | } |
238 | return mnt; | |
88b38782 | 239 | |
d3ef3d73 NP |
240 | #ifdef CONFIG_SMP |
241 | out_free_devname: | |
242 | kfree(mnt->mnt_devname); | |
243 | #endif | |
88b38782 LZ |
244 | out_free_id: |
245 | mnt_free_id(mnt); | |
246 | out_free_cache: | |
247 | kmem_cache_free(mnt_cache, mnt); | |
248 | return NULL; | |
1da177e4 LT |
249 | } |
250 | ||
3d733633 DH |
251 | /* |
252 | * Most r/o checks on a fs are for operations that take | |
253 | * discrete amounts of time, like a write() or unlink(). | |
254 | * We must keep track of when those operations start | |
255 | * (for permission checks) and when they end, so that | |
256 | * we can determine when writes are able to occur to | |
257 | * a filesystem. | |
258 | */ | |
259 | /* | |
260 | * __mnt_is_readonly: check whether a mount is read-only | |
261 | * @mnt: the mount to check for its write status | |
262 | * | |
263 | * This shouldn't be used directly ouside of the VFS. | |
264 | * It does not guarantee that the filesystem will stay | |
265 | * r/w, just that it is right *now*. This can not and | |
266 | * should not be used in place of IS_RDONLY(inode). | |
267 | * mnt_want/drop_write() will _keep_ the filesystem | |
268 | * r/w. | |
269 | */ | |
270 | int __mnt_is_readonly(struct vfsmount *mnt) | |
271 | { | |
2e4b7fcd DH |
272 | if (mnt->mnt_flags & MNT_READONLY) |
273 | return 1; | |
274 | if (mnt->mnt_sb->s_flags & MS_RDONLY) | |
275 | return 1; | |
276 | return 0; | |
3d733633 DH |
277 | } |
278 | EXPORT_SYMBOL_GPL(__mnt_is_readonly); | |
279 | ||
c6653a83 | 280 | static inline void mnt_inc_writers(struct vfsmount *mnt) |
d3ef3d73 NP |
281 | { |
282 | #ifdef CONFIG_SMP | |
b3e19d92 | 283 | this_cpu_inc(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 NP |
284 | #else |
285 | mnt->mnt_writers++; | |
286 | #endif | |
287 | } | |
3d733633 | 288 | |
c6653a83 | 289 | static inline void mnt_dec_writers(struct vfsmount *mnt) |
3d733633 | 290 | { |
d3ef3d73 | 291 | #ifdef CONFIG_SMP |
b3e19d92 | 292 | this_cpu_dec(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 NP |
293 | #else |
294 | mnt->mnt_writers--; | |
295 | #endif | |
3d733633 | 296 | } |
3d733633 | 297 | |
c6653a83 | 298 | static unsigned int mnt_get_writers(struct vfsmount *mnt) |
3d733633 | 299 | { |
d3ef3d73 NP |
300 | #ifdef CONFIG_SMP |
301 | unsigned int count = 0; | |
3d733633 | 302 | int cpu; |
3d733633 DH |
303 | |
304 | for_each_possible_cpu(cpu) { | |
b3e19d92 | 305 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers; |
3d733633 | 306 | } |
3d733633 | 307 | |
d3ef3d73 NP |
308 | return count; |
309 | #else | |
310 | return mnt->mnt_writers; | |
311 | #endif | |
3d733633 DH |
312 | } |
313 | ||
8366025e DH |
314 | /* |
315 | * Most r/o checks on a fs are for operations that take | |
316 | * discrete amounts of time, like a write() or unlink(). | |
317 | * We must keep track of when those operations start | |
318 | * (for permission checks) and when they end, so that | |
319 | * we can determine when writes are able to occur to | |
320 | * a filesystem. | |
321 | */ | |
322 | /** | |
323 | * mnt_want_write - get write access to a mount | |
324 | * @mnt: the mount on which to take a write | |
325 | * | |
326 | * This tells the low-level filesystem that a write is | |
327 | * about to be performed to it, and makes sure that | |
328 | * writes are allowed before returning success. When | |
329 | * the write operation is finished, mnt_drop_write() | |
330 | * must be called. This is effectively a refcount. | |
331 | */ | |
332 | int mnt_want_write(struct vfsmount *mnt) | |
333 | { | |
3d733633 | 334 | int ret = 0; |
3d733633 | 335 | |
d3ef3d73 | 336 | preempt_disable(); |
c6653a83 | 337 | mnt_inc_writers(mnt); |
d3ef3d73 | 338 | /* |
c6653a83 | 339 | * The store to mnt_inc_writers must be visible before we pass |
d3ef3d73 NP |
340 | * MNT_WRITE_HOLD loop below, so that the slowpath can see our |
341 | * incremented count after it has set MNT_WRITE_HOLD. | |
342 | */ | |
343 | smp_mb(); | |
344 | while (mnt->mnt_flags & MNT_WRITE_HOLD) | |
345 | cpu_relax(); | |
346 | /* | |
347 | * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will | |
348 | * be set to match its requirements. So we must not load that until | |
349 | * MNT_WRITE_HOLD is cleared. | |
350 | */ | |
351 | smp_rmb(); | |
3d733633 | 352 | if (__mnt_is_readonly(mnt)) { |
c6653a83 | 353 | mnt_dec_writers(mnt); |
3d733633 DH |
354 | ret = -EROFS; |
355 | goto out; | |
356 | } | |
3d733633 | 357 | out: |
d3ef3d73 | 358 | preempt_enable(); |
3d733633 | 359 | return ret; |
8366025e DH |
360 | } |
361 | EXPORT_SYMBOL_GPL(mnt_want_write); | |
362 | ||
96029c4e NP |
363 | /** |
364 | * mnt_clone_write - get write access to a mount | |
365 | * @mnt: the mount on which to take a write | |
366 | * | |
367 | * This is effectively like mnt_want_write, except | |
368 | * it must only be used to take an extra write reference | |
369 | * on a mountpoint that we already know has a write reference | |
370 | * on it. This allows some optimisation. | |
371 | * | |
372 | * After finished, mnt_drop_write must be called as usual to | |
373 | * drop the reference. | |
374 | */ | |
375 | int mnt_clone_write(struct vfsmount *mnt) | |
376 | { | |
377 | /* superblock may be r/o */ | |
378 | if (__mnt_is_readonly(mnt)) | |
379 | return -EROFS; | |
380 | preempt_disable(); | |
c6653a83 | 381 | mnt_inc_writers(mnt); |
96029c4e NP |
382 | preempt_enable(); |
383 | return 0; | |
384 | } | |
385 | EXPORT_SYMBOL_GPL(mnt_clone_write); | |
386 | ||
387 | /** | |
388 | * mnt_want_write_file - get write access to a file's mount | |
389 | * @file: the file who's mount on which to take a write | |
390 | * | |
391 | * This is like mnt_want_write, but it takes a file and can | |
392 | * do some optimisations if the file is open for write already | |
393 | */ | |
394 | int mnt_want_write_file(struct file *file) | |
395 | { | |
2d8dd38a OH |
396 | struct inode *inode = file->f_dentry->d_inode; |
397 | if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) | |
96029c4e NP |
398 | return mnt_want_write(file->f_path.mnt); |
399 | else | |
400 | return mnt_clone_write(file->f_path.mnt); | |
401 | } | |
402 | EXPORT_SYMBOL_GPL(mnt_want_write_file); | |
403 | ||
8366025e DH |
404 | /** |
405 | * mnt_drop_write - give up write access to a mount | |
406 | * @mnt: the mount on which to give up write access | |
407 | * | |
408 | * Tells the low-level filesystem that we are done | |
409 | * performing writes to it. Must be matched with | |
410 | * mnt_want_write() call above. | |
411 | */ | |
412 | void mnt_drop_write(struct vfsmount *mnt) | |
413 | { | |
d3ef3d73 | 414 | preempt_disable(); |
c6653a83 | 415 | mnt_dec_writers(mnt); |
d3ef3d73 | 416 | preempt_enable(); |
8366025e DH |
417 | } |
418 | EXPORT_SYMBOL_GPL(mnt_drop_write); | |
419 | ||
2e4b7fcd | 420 | static int mnt_make_readonly(struct vfsmount *mnt) |
8366025e | 421 | { |
3d733633 DH |
422 | int ret = 0; |
423 | ||
99b7db7b | 424 | br_write_lock(vfsmount_lock); |
d3ef3d73 | 425 | mnt->mnt_flags |= MNT_WRITE_HOLD; |
3d733633 | 426 | /* |
d3ef3d73 NP |
427 | * After storing MNT_WRITE_HOLD, we'll read the counters. This store |
428 | * should be visible before we do. | |
3d733633 | 429 | */ |
d3ef3d73 NP |
430 | smp_mb(); |
431 | ||
3d733633 | 432 | /* |
d3ef3d73 NP |
433 | * With writers on hold, if this value is zero, then there are |
434 | * definitely no active writers (although held writers may subsequently | |
435 | * increment the count, they'll have to wait, and decrement it after | |
436 | * seeing MNT_READONLY). | |
437 | * | |
438 | * It is OK to have counter incremented on one CPU and decremented on | |
439 | * another: the sum will add up correctly. The danger would be when we | |
440 | * sum up each counter, if we read a counter before it is incremented, | |
441 | * but then read another CPU's count which it has been subsequently | |
442 | * decremented from -- we would see more decrements than we should. | |
443 | * MNT_WRITE_HOLD protects against this scenario, because | |
444 | * mnt_want_write first increments count, then smp_mb, then spins on | |
445 | * MNT_WRITE_HOLD, so it can't be decremented by another CPU while | |
446 | * we're counting up here. | |
3d733633 | 447 | */ |
c6653a83 | 448 | if (mnt_get_writers(mnt) > 0) |
d3ef3d73 NP |
449 | ret = -EBUSY; |
450 | else | |
2e4b7fcd | 451 | mnt->mnt_flags |= MNT_READONLY; |
d3ef3d73 NP |
452 | /* |
453 | * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers | |
454 | * that become unheld will see MNT_READONLY. | |
455 | */ | |
456 | smp_wmb(); | |
457 | mnt->mnt_flags &= ~MNT_WRITE_HOLD; | |
99b7db7b | 458 | br_write_unlock(vfsmount_lock); |
3d733633 | 459 | return ret; |
8366025e | 460 | } |
8366025e | 461 | |
2e4b7fcd DH |
462 | static void __mnt_unmake_readonly(struct vfsmount *mnt) |
463 | { | |
99b7db7b | 464 | br_write_lock(vfsmount_lock); |
2e4b7fcd | 465 | mnt->mnt_flags &= ~MNT_READONLY; |
99b7db7b | 466 | br_write_unlock(vfsmount_lock); |
2e4b7fcd DH |
467 | } |
468 | ||
a3ec947c | 469 | void simple_set_mnt(struct vfsmount *mnt, struct super_block *sb) |
454e2398 DH |
470 | { |
471 | mnt->mnt_sb = sb; | |
472 | mnt->mnt_root = dget(sb->s_root); | |
454e2398 DH |
473 | } |
474 | ||
475 | EXPORT_SYMBOL(simple_set_mnt); | |
476 | ||
1da177e4 LT |
477 | void free_vfsmnt(struct vfsmount *mnt) |
478 | { | |
479 | kfree(mnt->mnt_devname); | |
73cd49ec | 480 | mnt_free_id(mnt); |
d3ef3d73 | 481 | #ifdef CONFIG_SMP |
b3e19d92 | 482 | free_percpu(mnt->mnt_pcp); |
d3ef3d73 | 483 | #endif |
1da177e4 LT |
484 | kmem_cache_free(mnt_cache, mnt); |
485 | } | |
486 | ||
487 | /* | |
a05964f3 RP |
488 | * find the first or last mount at @dentry on vfsmount @mnt depending on |
489 | * @dir. If @dir is set return the first mount else return the last mount. | |
99b7db7b | 490 | * vfsmount_lock must be held for read or write. |
1da177e4 | 491 | */ |
a05964f3 RP |
492 | struct vfsmount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, |
493 | int dir) | |
1da177e4 | 494 | { |
b58fed8b RP |
495 | struct list_head *head = mount_hashtable + hash(mnt, dentry); |
496 | struct list_head *tmp = head; | |
1da177e4 LT |
497 | struct vfsmount *p, *found = NULL; |
498 | ||
1da177e4 | 499 | for (;;) { |
a05964f3 | 500 | tmp = dir ? tmp->next : tmp->prev; |
1da177e4 LT |
501 | p = NULL; |
502 | if (tmp == head) | |
503 | break; | |
504 | p = list_entry(tmp, struct vfsmount, mnt_hash); | |
505 | if (p->mnt_parent == mnt && p->mnt_mountpoint == dentry) { | |
a05964f3 | 506 | found = p; |
1da177e4 LT |
507 | break; |
508 | } | |
509 | } | |
1da177e4 LT |
510 | return found; |
511 | } | |
512 | ||
a05964f3 RP |
513 | /* |
514 | * lookup_mnt increments the ref count before returning | |
515 | * the vfsmount struct. | |
516 | */ | |
1c755af4 | 517 | struct vfsmount *lookup_mnt(struct path *path) |
a05964f3 RP |
518 | { |
519 | struct vfsmount *child_mnt; | |
99b7db7b NP |
520 | |
521 | br_read_lock(vfsmount_lock); | |
1c755af4 | 522 | if ((child_mnt = __lookup_mnt(path->mnt, path->dentry, 1))) |
a05964f3 | 523 | mntget(child_mnt); |
99b7db7b | 524 | br_read_unlock(vfsmount_lock); |
a05964f3 RP |
525 | return child_mnt; |
526 | } | |
527 | ||
1da177e4 LT |
528 | static inline int check_mnt(struct vfsmount *mnt) |
529 | { | |
6b3286ed | 530 | return mnt->mnt_ns == current->nsproxy->mnt_ns; |
1da177e4 LT |
531 | } |
532 | ||
99b7db7b NP |
533 | /* |
534 | * vfsmount lock must be held for write | |
535 | */ | |
6b3286ed | 536 | static void touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
537 | { |
538 | if (ns) { | |
539 | ns->event = ++event; | |
540 | wake_up_interruptible(&ns->poll); | |
541 | } | |
542 | } | |
543 | ||
99b7db7b NP |
544 | /* |
545 | * vfsmount lock must be held for write | |
546 | */ | |
6b3286ed | 547 | static void __touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
548 | { |
549 | if (ns && ns->event != event) { | |
550 | ns->event = event; | |
551 | wake_up_interruptible(&ns->poll); | |
552 | } | |
553 | } | |
554 | ||
5f57cbcc NP |
555 | /* |
556 | * Clear dentry's mounted state if it has no remaining mounts. | |
557 | * vfsmount_lock must be held for write. | |
558 | */ | |
559 | static void dentry_reset_mounted(struct vfsmount *mnt, struct dentry *dentry) | |
560 | { | |
561 | unsigned u; | |
562 | ||
563 | for (u = 0; u < HASH_SIZE; u++) { | |
564 | struct vfsmount *p; | |
565 | ||
566 | list_for_each_entry(p, &mount_hashtable[u], mnt_hash) { | |
567 | if (p->mnt_mountpoint == dentry) | |
568 | return; | |
569 | } | |
570 | } | |
571 | spin_lock(&dentry->d_lock); | |
572 | dentry->d_flags &= ~DCACHE_MOUNTED; | |
573 | spin_unlock(&dentry->d_lock); | |
574 | } | |
575 | ||
99b7db7b NP |
576 | /* |
577 | * vfsmount lock must be held for write | |
578 | */ | |
1a390689 | 579 | static void detach_mnt(struct vfsmount *mnt, struct path *old_path) |
1da177e4 | 580 | { |
1a390689 AV |
581 | old_path->dentry = mnt->mnt_mountpoint; |
582 | old_path->mnt = mnt->mnt_parent; | |
1da177e4 LT |
583 | mnt->mnt_parent = mnt; |
584 | mnt->mnt_mountpoint = mnt->mnt_root; | |
585 | list_del_init(&mnt->mnt_child); | |
586 | list_del_init(&mnt->mnt_hash); | |
5f57cbcc | 587 | dentry_reset_mounted(old_path->mnt, old_path->dentry); |
1da177e4 LT |
588 | } |
589 | ||
99b7db7b NP |
590 | /* |
591 | * vfsmount lock must be held for write | |
592 | */ | |
b90fa9ae RP |
593 | void mnt_set_mountpoint(struct vfsmount *mnt, struct dentry *dentry, |
594 | struct vfsmount *child_mnt) | |
595 | { | |
596 | child_mnt->mnt_parent = mntget(mnt); | |
597 | child_mnt->mnt_mountpoint = dget(dentry); | |
5f57cbcc NP |
598 | spin_lock(&dentry->d_lock); |
599 | dentry->d_flags |= DCACHE_MOUNTED; | |
600 | spin_unlock(&dentry->d_lock); | |
b90fa9ae RP |
601 | } |
602 | ||
99b7db7b NP |
603 | /* |
604 | * vfsmount lock must be held for write | |
605 | */ | |
1a390689 | 606 | static void attach_mnt(struct vfsmount *mnt, struct path *path) |
1da177e4 | 607 | { |
1a390689 | 608 | mnt_set_mountpoint(path->mnt, path->dentry, mnt); |
b90fa9ae | 609 | list_add_tail(&mnt->mnt_hash, mount_hashtable + |
1a390689 AV |
610 | hash(path->mnt, path->dentry)); |
611 | list_add_tail(&mnt->mnt_child, &path->mnt->mnt_mounts); | |
b90fa9ae RP |
612 | } |
613 | ||
7e3d0eb0 AV |
614 | static inline void __mnt_make_longterm(struct vfsmount *mnt) |
615 | { | |
616 | #ifdef CONFIG_SMP | |
617 | atomic_inc(&mnt->mnt_longterm); | |
618 | #endif | |
619 | } | |
620 | ||
621 | /* needs vfsmount lock for write */ | |
622 | static inline void __mnt_make_shortterm(struct vfsmount *mnt) | |
623 | { | |
624 | #ifdef CONFIG_SMP | |
625 | atomic_dec(&mnt->mnt_longterm); | |
626 | #endif | |
627 | } | |
628 | ||
b90fa9ae | 629 | /* |
99b7db7b | 630 | * vfsmount lock must be held for write |
b90fa9ae RP |
631 | */ |
632 | static void commit_tree(struct vfsmount *mnt) | |
633 | { | |
634 | struct vfsmount *parent = mnt->mnt_parent; | |
635 | struct vfsmount *m; | |
636 | LIST_HEAD(head); | |
6b3286ed | 637 | struct mnt_namespace *n = parent->mnt_ns; |
b90fa9ae RP |
638 | |
639 | BUG_ON(parent == mnt); | |
640 | ||
641 | list_add_tail(&head, &mnt->mnt_list); | |
f03c6599 | 642 | list_for_each_entry(m, &head, mnt_list) { |
6b3286ed | 643 | m->mnt_ns = n; |
7e3d0eb0 | 644 | __mnt_make_longterm(m); |
f03c6599 AV |
645 | } |
646 | ||
b90fa9ae RP |
647 | list_splice(&head, n->list.prev); |
648 | ||
649 | list_add_tail(&mnt->mnt_hash, mount_hashtable + | |
650 | hash(parent, mnt->mnt_mountpoint)); | |
651 | list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); | |
6b3286ed | 652 | touch_mnt_namespace(n); |
1da177e4 LT |
653 | } |
654 | ||
655 | static struct vfsmount *next_mnt(struct vfsmount *p, struct vfsmount *root) | |
656 | { | |
657 | struct list_head *next = p->mnt_mounts.next; | |
658 | if (next == &p->mnt_mounts) { | |
659 | while (1) { | |
660 | if (p == root) | |
661 | return NULL; | |
662 | next = p->mnt_child.next; | |
663 | if (next != &p->mnt_parent->mnt_mounts) | |
664 | break; | |
665 | p = p->mnt_parent; | |
666 | } | |
667 | } | |
668 | return list_entry(next, struct vfsmount, mnt_child); | |
669 | } | |
670 | ||
9676f0c6 RP |
671 | static struct vfsmount *skip_mnt_tree(struct vfsmount *p) |
672 | { | |
673 | struct list_head *prev = p->mnt_mounts.prev; | |
674 | while (prev != &p->mnt_mounts) { | |
675 | p = list_entry(prev, struct vfsmount, mnt_child); | |
676 | prev = p->mnt_mounts.prev; | |
677 | } | |
678 | return p; | |
679 | } | |
680 | ||
36341f64 RP |
681 | static struct vfsmount *clone_mnt(struct vfsmount *old, struct dentry *root, |
682 | int flag) | |
1da177e4 LT |
683 | { |
684 | struct super_block *sb = old->mnt_sb; | |
685 | struct vfsmount *mnt = alloc_vfsmnt(old->mnt_devname); | |
686 | ||
687 | if (mnt) { | |
719f5d7f MS |
688 | if (flag & (CL_SLAVE | CL_PRIVATE)) |
689 | mnt->mnt_group_id = 0; /* not a peer of original */ | |
690 | else | |
691 | mnt->mnt_group_id = old->mnt_group_id; | |
692 | ||
693 | if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { | |
694 | int err = mnt_alloc_group_id(mnt); | |
695 | if (err) | |
696 | goto out_free; | |
697 | } | |
698 | ||
be1a16a0 | 699 | mnt->mnt_flags = old->mnt_flags & ~MNT_WRITE_HOLD; |
1da177e4 LT |
700 | atomic_inc(&sb->s_active); |
701 | mnt->mnt_sb = sb; | |
702 | mnt->mnt_root = dget(root); | |
703 | mnt->mnt_mountpoint = mnt->mnt_root; | |
704 | mnt->mnt_parent = mnt; | |
b90fa9ae | 705 | |
5afe0022 RP |
706 | if (flag & CL_SLAVE) { |
707 | list_add(&mnt->mnt_slave, &old->mnt_slave_list); | |
708 | mnt->mnt_master = old; | |
709 | CLEAR_MNT_SHARED(mnt); | |
8aec0809 | 710 | } else if (!(flag & CL_PRIVATE)) { |
796a6b52 | 711 | if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) |
5afe0022 RP |
712 | list_add(&mnt->mnt_share, &old->mnt_share); |
713 | if (IS_MNT_SLAVE(old)) | |
714 | list_add(&mnt->mnt_slave, &old->mnt_slave); | |
715 | mnt->mnt_master = old->mnt_master; | |
716 | } | |
b90fa9ae RP |
717 | if (flag & CL_MAKE_SHARED) |
718 | set_mnt_shared(mnt); | |
1da177e4 LT |
719 | |
720 | /* stick the duplicate mount on the same expiry list | |
721 | * as the original if that was on one */ | |
36341f64 | 722 | if (flag & CL_EXPIRE) { |
36341f64 RP |
723 | if (!list_empty(&old->mnt_expire)) |
724 | list_add(&mnt->mnt_expire, &old->mnt_expire); | |
36341f64 | 725 | } |
1da177e4 LT |
726 | } |
727 | return mnt; | |
719f5d7f MS |
728 | |
729 | out_free: | |
730 | free_vfsmnt(mnt); | |
731 | return NULL; | |
1da177e4 LT |
732 | } |
733 | ||
b3e19d92 | 734 | static inline void mntfree(struct vfsmount *mnt) |
1da177e4 LT |
735 | { |
736 | struct super_block *sb = mnt->mnt_sb; | |
b3e19d92 | 737 | |
3d733633 DH |
738 | /* |
739 | * This probably indicates that somebody messed | |
740 | * up a mnt_want/drop_write() pair. If this | |
741 | * happens, the filesystem was probably unable | |
742 | * to make r/w->r/o transitions. | |
743 | */ | |
d3ef3d73 | 744 | /* |
b3e19d92 NP |
745 | * The locking used to deal with mnt_count decrement provides barriers, |
746 | * so mnt_get_writers() below is safe. | |
d3ef3d73 | 747 | */ |
c6653a83 | 748 | WARN_ON(mnt_get_writers(mnt)); |
ca9c726e | 749 | fsnotify_vfsmount_delete(mnt); |
1da177e4 LT |
750 | dput(mnt->mnt_root); |
751 | free_vfsmnt(mnt); | |
752 | deactivate_super(sb); | |
753 | } | |
754 | ||
f03c6599 | 755 | static void mntput_no_expire(struct vfsmount *mnt) |
b3e19d92 | 756 | { |
b3e19d92 | 757 | put_again: |
f03c6599 AV |
758 | #ifdef CONFIG_SMP |
759 | br_read_lock(vfsmount_lock); | |
760 | if (likely(atomic_read(&mnt->mnt_longterm))) { | |
761 | mnt_dec_count(mnt); | |
b3e19d92 | 762 | br_read_unlock(vfsmount_lock); |
f03c6599 | 763 | return; |
b3e19d92 | 764 | } |
f03c6599 | 765 | br_read_unlock(vfsmount_lock); |
b3e19d92 | 766 | |
99b7db7b | 767 | br_write_lock(vfsmount_lock); |
f03c6599 | 768 | mnt_dec_count(mnt); |
b3e19d92 | 769 | if (mnt_get_count(mnt)) { |
99b7db7b NP |
770 | br_write_unlock(vfsmount_lock); |
771 | return; | |
772 | } | |
b3e19d92 | 773 | #else |
b3e19d92 NP |
774 | mnt_dec_count(mnt); |
775 | if (likely(mnt_get_count(mnt))) | |
99b7db7b | 776 | return; |
b3e19d92 | 777 | br_write_lock(vfsmount_lock); |
f03c6599 | 778 | #endif |
b3e19d92 NP |
779 | if (unlikely(mnt->mnt_pinned)) { |
780 | mnt_add_count(mnt, mnt->mnt_pinned + 1); | |
781 | mnt->mnt_pinned = 0; | |
782 | br_write_unlock(vfsmount_lock); | |
783 | acct_auto_close_mnt(mnt); | |
784 | goto put_again; | |
7b7b1ace | 785 | } |
99b7db7b | 786 | br_write_unlock(vfsmount_lock); |
b3e19d92 NP |
787 | mntfree(mnt); |
788 | } | |
b3e19d92 NP |
789 | |
790 | void mntput(struct vfsmount *mnt) | |
791 | { | |
792 | if (mnt) { | |
793 | /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ | |
794 | if (unlikely(mnt->mnt_expiry_mark)) | |
795 | mnt->mnt_expiry_mark = 0; | |
f03c6599 | 796 | mntput_no_expire(mnt); |
b3e19d92 NP |
797 | } |
798 | } | |
799 | EXPORT_SYMBOL(mntput); | |
800 | ||
801 | struct vfsmount *mntget(struct vfsmount *mnt) | |
802 | { | |
803 | if (mnt) | |
804 | mnt_inc_count(mnt); | |
805 | return mnt; | |
806 | } | |
807 | EXPORT_SYMBOL(mntget); | |
808 | ||
7b7b1ace AV |
809 | void mnt_pin(struct vfsmount *mnt) |
810 | { | |
99b7db7b | 811 | br_write_lock(vfsmount_lock); |
7b7b1ace | 812 | mnt->mnt_pinned++; |
99b7db7b | 813 | br_write_unlock(vfsmount_lock); |
7b7b1ace | 814 | } |
7b7b1ace AV |
815 | EXPORT_SYMBOL(mnt_pin); |
816 | ||
817 | void mnt_unpin(struct vfsmount *mnt) | |
818 | { | |
99b7db7b | 819 | br_write_lock(vfsmount_lock); |
7b7b1ace | 820 | if (mnt->mnt_pinned) { |
b3e19d92 | 821 | mnt_inc_count(mnt); |
7b7b1ace AV |
822 | mnt->mnt_pinned--; |
823 | } | |
99b7db7b | 824 | br_write_unlock(vfsmount_lock); |
7b7b1ace | 825 | } |
7b7b1ace | 826 | EXPORT_SYMBOL(mnt_unpin); |
1da177e4 | 827 | |
b3b304a2 MS |
828 | static inline void mangle(struct seq_file *m, const char *s) |
829 | { | |
830 | seq_escape(m, s, " \t\n\\"); | |
831 | } | |
832 | ||
833 | /* | |
834 | * Simple .show_options callback for filesystems which don't want to | |
835 | * implement more complex mount option showing. | |
836 | * | |
837 | * See also save_mount_options(). | |
838 | */ | |
839 | int generic_show_options(struct seq_file *m, struct vfsmount *mnt) | |
840 | { | |
2a32cebd AV |
841 | const char *options; |
842 | ||
843 | rcu_read_lock(); | |
844 | options = rcu_dereference(mnt->mnt_sb->s_options); | |
b3b304a2 MS |
845 | |
846 | if (options != NULL && options[0]) { | |
847 | seq_putc(m, ','); | |
848 | mangle(m, options); | |
849 | } | |
2a32cebd | 850 | rcu_read_unlock(); |
b3b304a2 MS |
851 | |
852 | return 0; | |
853 | } | |
854 | EXPORT_SYMBOL(generic_show_options); | |
855 | ||
856 | /* | |
857 | * If filesystem uses generic_show_options(), this function should be | |
858 | * called from the fill_super() callback. | |
859 | * | |
860 | * The .remount_fs callback usually needs to be handled in a special | |
861 | * way, to make sure, that previous options are not overwritten if the | |
862 | * remount fails. | |
863 | * | |
864 | * Also note, that if the filesystem's .remount_fs function doesn't | |
865 | * reset all options to their default value, but changes only newly | |
866 | * given options, then the displayed options will not reflect reality | |
867 | * any more. | |
868 | */ | |
869 | void save_mount_options(struct super_block *sb, char *options) | |
870 | { | |
2a32cebd AV |
871 | BUG_ON(sb->s_options); |
872 | rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); | |
b3b304a2 MS |
873 | } |
874 | EXPORT_SYMBOL(save_mount_options); | |
875 | ||
2a32cebd AV |
876 | void replace_mount_options(struct super_block *sb, char *options) |
877 | { | |
878 | char *old = sb->s_options; | |
879 | rcu_assign_pointer(sb->s_options, options); | |
880 | if (old) { | |
881 | synchronize_rcu(); | |
882 | kfree(old); | |
883 | } | |
884 | } | |
885 | EXPORT_SYMBOL(replace_mount_options); | |
886 | ||
a1a2c409 | 887 | #ifdef CONFIG_PROC_FS |
1da177e4 LT |
888 | /* iterator */ |
889 | static void *m_start(struct seq_file *m, loff_t *pos) | |
890 | { | |
a1a2c409 | 891 | struct proc_mounts *p = m->private; |
1da177e4 | 892 | |
390c6843 | 893 | down_read(&namespace_sem); |
a1a2c409 | 894 | return seq_list_start(&p->ns->list, *pos); |
1da177e4 LT |
895 | } |
896 | ||
897 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
898 | { | |
a1a2c409 | 899 | struct proc_mounts *p = m->private; |
b0765fb8 | 900 | |
a1a2c409 | 901 | return seq_list_next(v, &p->ns->list, pos); |
1da177e4 LT |
902 | } |
903 | ||
904 | static void m_stop(struct seq_file *m, void *v) | |
905 | { | |
390c6843 | 906 | up_read(&namespace_sem); |
1da177e4 LT |
907 | } |
908 | ||
9f5596af AV |
909 | int mnt_had_events(struct proc_mounts *p) |
910 | { | |
911 | struct mnt_namespace *ns = p->ns; | |
912 | int res = 0; | |
913 | ||
99b7db7b | 914 | br_read_lock(vfsmount_lock); |
9f5596af AV |
915 | if (p->event != ns->event) { |
916 | p->event = ns->event; | |
917 | res = 1; | |
918 | } | |
99b7db7b | 919 | br_read_unlock(vfsmount_lock); |
9f5596af AV |
920 | |
921 | return res; | |
922 | } | |
923 | ||
2d4d4864 RP |
924 | struct proc_fs_info { |
925 | int flag; | |
926 | const char *str; | |
927 | }; | |
928 | ||
2069f457 | 929 | static int show_sb_opts(struct seq_file *m, struct super_block *sb) |
1da177e4 | 930 | { |
2d4d4864 | 931 | static const struct proc_fs_info fs_info[] = { |
1da177e4 LT |
932 | { MS_SYNCHRONOUS, ",sync" }, |
933 | { MS_DIRSYNC, ",dirsync" }, | |
934 | { MS_MANDLOCK, ",mand" }, | |
1da177e4 LT |
935 | { 0, NULL } |
936 | }; | |
2d4d4864 RP |
937 | const struct proc_fs_info *fs_infop; |
938 | ||
939 | for (fs_infop = fs_info; fs_infop->flag; fs_infop++) { | |
940 | if (sb->s_flags & fs_infop->flag) | |
941 | seq_puts(m, fs_infop->str); | |
942 | } | |
2069f457 EP |
943 | |
944 | return security_sb_show_options(m, sb); | |
2d4d4864 RP |
945 | } |
946 | ||
947 | static void show_mnt_opts(struct seq_file *m, struct vfsmount *mnt) | |
948 | { | |
949 | static const struct proc_fs_info mnt_info[] = { | |
1da177e4 LT |
950 | { MNT_NOSUID, ",nosuid" }, |
951 | { MNT_NODEV, ",nodev" }, | |
952 | { MNT_NOEXEC, ",noexec" }, | |
fc33a7bb CH |
953 | { MNT_NOATIME, ",noatime" }, |
954 | { MNT_NODIRATIME, ",nodiratime" }, | |
47ae32d6 | 955 | { MNT_RELATIME, ",relatime" }, |
1da177e4 LT |
956 | { 0, NULL } |
957 | }; | |
2d4d4864 RP |
958 | const struct proc_fs_info *fs_infop; |
959 | ||
960 | for (fs_infop = mnt_info; fs_infop->flag; fs_infop++) { | |
961 | if (mnt->mnt_flags & fs_infop->flag) | |
962 | seq_puts(m, fs_infop->str); | |
963 | } | |
964 | } | |
965 | ||
966 | static void show_type(struct seq_file *m, struct super_block *sb) | |
967 | { | |
968 | mangle(m, sb->s_type->name); | |
969 | if (sb->s_subtype && sb->s_subtype[0]) { | |
970 | seq_putc(m, '.'); | |
971 | mangle(m, sb->s_subtype); | |
972 | } | |
973 | } | |
974 | ||
975 | static int show_vfsmnt(struct seq_file *m, void *v) | |
976 | { | |
977 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); | |
978 | int err = 0; | |
c32c2f63 | 979 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; |
1da177e4 | 980 | |
c7f404b4 AV |
981 | if (mnt->mnt_sb->s_op->show_devname) { |
982 | err = mnt->mnt_sb->s_op->show_devname(m, mnt); | |
983 | if (err) | |
984 | goto out; | |
985 | } else { | |
986 | mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); | |
987 | } | |
1da177e4 | 988 | seq_putc(m, ' '); |
c32c2f63 | 989 | seq_path(m, &mnt_path, " \t\n\\"); |
1da177e4 | 990 | seq_putc(m, ' '); |
2d4d4864 | 991 | show_type(m, mnt->mnt_sb); |
2e4b7fcd | 992 | seq_puts(m, __mnt_is_readonly(mnt) ? " ro" : " rw"); |
2069f457 EP |
993 | err = show_sb_opts(m, mnt->mnt_sb); |
994 | if (err) | |
995 | goto out; | |
2d4d4864 | 996 | show_mnt_opts(m, mnt); |
1da177e4 LT |
997 | if (mnt->mnt_sb->s_op->show_options) |
998 | err = mnt->mnt_sb->s_op->show_options(m, mnt); | |
999 | seq_puts(m, " 0 0\n"); | |
2069f457 | 1000 | out: |
1da177e4 LT |
1001 | return err; |
1002 | } | |
1003 | ||
a1a2c409 | 1004 | const struct seq_operations mounts_op = { |
1da177e4 LT |
1005 | .start = m_start, |
1006 | .next = m_next, | |
1007 | .stop = m_stop, | |
1008 | .show = show_vfsmnt | |
1009 | }; | |
1010 | ||
93f1c20b AK |
1011 | static int uuid_is_nil(u8 *uuid) |
1012 | { | |
1013 | int i; | |
1014 | u8 *cp = (u8 *)uuid; | |
1015 | ||
1016 | for (i = 0; i < 16; i++) { | |
1017 | if (*cp++) | |
1018 | return 0; | |
1019 | } | |
1020 | return 1; | |
1021 | } | |
1022 | ||
2d4d4864 RP |
1023 | static int show_mountinfo(struct seq_file *m, void *v) |
1024 | { | |
1025 | struct proc_mounts *p = m->private; | |
1026 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); | |
1027 | struct super_block *sb = mnt->mnt_sb; | |
1028 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; | |
1029 | struct path root = p->root; | |
1030 | int err = 0; | |
1031 | ||
1032 | seq_printf(m, "%i %i %u:%u ", mnt->mnt_id, mnt->mnt_parent->mnt_id, | |
1033 | MAJOR(sb->s_dev), MINOR(sb->s_dev)); | |
c7f404b4 AV |
1034 | if (sb->s_op->show_path) |
1035 | err = sb->s_op->show_path(m, mnt); | |
1036 | else | |
1037 | seq_dentry(m, mnt->mnt_root, " \t\n\\"); | |
1038 | if (err) | |
1039 | goto out; | |
2d4d4864 RP |
1040 | seq_putc(m, ' '); |
1041 | seq_path_root(m, &mnt_path, &root, " \t\n\\"); | |
1042 | if (root.mnt != p->root.mnt || root.dentry != p->root.dentry) { | |
1043 | /* | |
1044 | * Mountpoint is outside root, discard that one. Ugly, | |
1045 | * but less so than trying to do that in iterator in a | |
1046 | * race-free way (due to renames). | |
1047 | */ | |
1048 | return SEQ_SKIP; | |
1049 | } | |
1050 | seq_puts(m, mnt->mnt_flags & MNT_READONLY ? " ro" : " rw"); | |
1051 | show_mnt_opts(m, mnt); | |
1052 | ||
1053 | /* Tagged fields ("foo:X" or "bar") */ | |
1054 | if (IS_MNT_SHARED(mnt)) | |
1055 | seq_printf(m, " shared:%i", mnt->mnt_group_id); | |
97e7e0f7 MS |
1056 | if (IS_MNT_SLAVE(mnt)) { |
1057 | int master = mnt->mnt_master->mnt_group_id; | |
1058 | int dom = get_dominating_id(mnt, &p->root); | |
1059 | seq_printf(m, " master:%i", master); | |
1060 | if (dom && dom != master) | |
1061 | seq_printf(m, " propagate_from:%i", dom); | |
1062 | } | |
2d4d4864 RP |
1063 | if (IS_MNT_UNBINDABLE(mnt)) |
1064 | seq_puts(m, " unbindable"); | |
1065 | ||
93f1c20b AK |
1066 | if (!uuid_is_nil(mnt->mnt_sb->s_uuid)) |
1067 | /* print the uuid */ | |
1068 | seq_printf(m, " uuid:%pU", mnt->mnt_sb->s_uuid); | |
1069 | ||
2d4d4864 RP |
1070 | /* Filesystem specific data */ |
1071 | seq_puts(m, " - "); | |
1072 | show_type(m, sb); | |
1073 | seq_putc(m, ' '); | |
c7f404b4 AV |
1074 | if (sb->s_op->show_devname) |
1075 | err = sb->s_op->show_devname(m, mnt); | |
1076 | else | |
1077 | mangle(m, mnt->mnt_devname ? mnt->mnt_devname : "none"); | |
1078 | if (err) | |
1079 | goto out; | |
2d4d4864 | 1080 | seq_puts(m, sb->s_flags & MS_RDONLY ? " ro" : " rw"); |
2069f457 EP |
1081 | err = show_sb_opts(m, sb); |
1082 | if (err) | |
1083 | goto out; | |
2d4d4864 RP |
1084 | if (sb->s_op->show_options) |
1085 | err = sb->s_op->show_options(m, mnt); | |
1086 | seq_putc(m, '\n'); | |
2069f457 | 1087 | out: |
2d4d4864 RP |
1088 | return err; |
1089 | } | |
1090 | ||
1091 | const struct seq_operations mountinfo_op = { | |
1092 | .start = m_start, | |
1093 | .next = m_next, | |
1094 | .stop = m_stop, | |
1095 | .show = show_mountinfo, | |
1096 | }; | |
1097 | ||
b4629fe2 CL |
1098 | static int show_vfsstat(struct seq_file *m, void *v) |
1099 | { | |
b0765fb8 | 1100 | struct vfsmount *mnt = list_entry(v, struct vfsmount, mnt_list); |
c32c2f63 | 1101 | struct path mnt_path = { .dentry = mnt->mnt_root, .mnt = mnt }; |
b4629fe2 CL |
1102 | int err = 0; |
1103 | ||
1104 | /* device */ | |
c7f404b4 AV |
1105 | if (mnt->mnt_sb->s_op->show_devname) { |
1106 | err = mnt->mnt_sb->s_op->show_devname(m, mnt); | |
1107 | } else { | |
1108 | if (mnt->mnt_devname) { | |
1109 | seq_puts(m, "device "); | |
1110 | mangle(m, mnt->mnt_devname); | |
1111 | } else | |
1112 | seq_puts(m, "no device"); | |
1113 | } | |
b4629fe2 CL |
1114 | |
1115 | /* mount point */ | |
1116 | seq_puts(m, " mounted on "); | |
c32c2f63 | 1117 | seq_path(m, &mnt_path, " \t\n\\"); |
b4629fe2 CL |
1118 | seq_putc(m, ' '); |
1119 | ||
1120 | /* file system type */ | |
1121 | seq_puts(m, "with fstype "); | |
2d4d4864 | 1122 | show_type(m, mnt->mnt_sb); |
b4629fe2 CL |
1123 | |
1124 | /* optional statistics */ | |
1125 | if (mnt->mnt_sb->s_op->show_stats) { | |
1126 | seq_putc(m, ' '); | |
c7f404b4 AV |
1127 | if (!err) |
1128 | err = mnt->mnt_sb->s_op->show_stats(m, mnt); | |
b4629fe2 CL |
1129 | } |
1130 | ||
1131 | seq_putc(m, '\n'); | |
1132 | return err; | |
1133 | } | |
1134 | ||
a1a2c409 | 1135 | const struct seq_operations mountstats_op = { |
b4629fe2 CL |
1136 | .start = m_start, |
1137 | .next = m_next, | |
1138 | .stop = m_stop, | |
1139 | .show = show_vfsstat, | |
1140 | }; | |
a1a2c409 | 1141 | #endif /* CONFIG_PROC_FS */ |
b4629fe2 | 1142 | |
1da177e4 LT |
1143 | /** |
1144 | * may_umount_tree - check if a mount tree is busy | |
1145 | * @mnt: root of mount tree | |
1146 | * | |
1147 | * This is called to check if a tree of mounts has any | |
1148 | * open files, pwds, chroots or sub mounts that are | |
1149 | * busy. | |
1150 | */ | |
1151 | int may_umount_tree(struct vfsmount *mnt) | |
1152 | { | |
36341f64 RP |
1153 | int actual_refs = 0; |
1154 | int minimum_refs = 0; | |
1155 | struct vfsmount *p; | |
1da177e4 | 1156 | |
b3e19d92 NP |
1157 | /* write lock needed for mnt_get_count */ |
1158 | br_write_lock(vfsmount_lock); | |
36341f64 | 1159 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
b3e19d92 | 1160 | actual_refs += mnt_get_count(p); |
1da177e4 | 1161 | minimum_refs += 2; |
1da177e4 | 1162 | } |
b3e19d92 | 1163 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
1164 | |
1165 | if (actual_refs > minimum_refs) | |
e3474a8e | 1166 | return 0; |
1da177e4 | 1167 | |
e3474a8e | 1168 | return 1; |
1da177e4 LT |
1169 | } |
1170 | ||
1171 | EXPORT_SYMBOL(may_umount_tree); | |
1172 | ||
1173 | /** | |
1174 | * may_umount - check if a mount point is busy | |
1175 | * @mnt: root of mount | |
1176 | * | |
1177 | * This is called to check if a mount point has any | |
1178 | * open files, pwds, chroots or sub mounts. If the | |
1179 | * mount has sub mounts this will return busy | |
1180 | * regardless of whether the sub mounts are busy. | |
1181 | * | |
1182 | * Doesn't take quota and stuff into account. IOW, in some cases it will | |
1183 | * give false negatives. The main reason why it's here is that we need | |
1184 | * a non-destructive way to look for easily umountable filesystems. | |
1185 | */ | |
1186 | int may_umount(struct vfsmount *mnt) | |
1187 | { | |
e3474a8e | 1188 | int ret = 1; |
8ad08d8a | 1189 | down_read(&namespace_sem); |
b3e19d92 | 1190 | br_write_lock(vfsmount_lock); |
a05964f3 | 1191 | if (propagate_mount_busy(mnt, 2)) |
e3474a8e | 1192 | ret = 0; |
b3e19d92 | 1193 | br_write_unlock(vfsmount_lock); |
8ad08d8a | 1194 | up_read(&namespace_sem); |
a05964f3 | 1195 | return ret; |
1da177e4 LT |
1196 | } |
1197 | ||
1198 | EXPORT_SYMBOL(may_umount); | |
1199 | ||
b90fa9ae | 1200 | void release_mounts(struct list_head *head) |
70fbcdf4 RP |
1201 | { |
1202 | struct vfsmount *mnt; | |
bf066c7d | 1203 | while (!list_empty(head)) { |
b5e61818 | 1204 | mnt = list_first_entry(head, struct vfsmount, mnt_hash); |
70fbcdf4 RP |
1205 | list_del_init(&mnt->mnt_hash); |
1206 | if (mnt->mnt_parent != mnt) { | |
1207 | struct dentry *dentry; | |
1208 | struct vfsmount *m; | |
99b7db7b NP |
1209 | |
1210 | br_write_lock(vfsmount_lock); | |
70fbcdf4 RP |
1211 | dentry = mnt->mnt_mountpoint; |
1212 | m = mnt->mnt_parent; | |
1213 | mnt->mnt_mountpoint = mnt->mnt_root; | |
1214 | mnt->mnt_parent = mnt; | |
7c4b93d8 | 1215 | m->mnt_ghosts--; |
99b7db7b | 1216 | br_write_unlock(vfsmount_lock); |
70fbcdf4 RP |
1217 | dput(dentry); |
1218 | mntput(m); | |
1219 | } | |
f03c6599 | 1220 | mntput(mnt); |
70fbcdf4 RP |
1221 | } |
1222 | } | |
1223 | ||
99b7db7b NP |
1224 | /* |
1225 | * vfsmount lock must be held for write | |
1226 | * namespace_sem must be held for write | |
1227 | */ | |
a05964f3 | 1228 | void umount_tree(struct vfsmount *mnt, int propagate, struct list_head *kill) |
1da177e4 | 1229 | { |
7b8a53fd | 1230 | LIST_HEAD(tmp_list); |
1da177e4 | 1231 | struct vfsmount *p; |
1da177e4 | 1232 | |
1bfba4e8 | 1233 | for (p = mnt; p; p = next_mnt(p, mnt)) |
7b8a53fd | 1234 | list_move(&p->mnt_hash, &tmp_list); |
1da177e4 | 1235 | |
a05964f3 | 1236 | if (propagate) |
7b8a53fd | 1237 | propagate_umount(&tmp_list); |
a05964f3 | 1238 | |
7b8a53fd | 1239 | list_for_each_entry(p, &tmp_list, mnt_hash) { |
70fbcdf4 RP |
1240 | list_del_init(&p->mnt_expire); |
1241 | list_del_init(&p->mnt_list); | |
6b3286ed KK |
1242 | __touch_mnt_namespace(p->mnt_ns); |
1243 | p->mnt_ns = NULL; | |
7e3d0eb0 | 1244 | __mnt_make_shortterm(p); |
70fbcdf4 | 1245 | list_del_init(&p->mnt_child); |
7c4b93d8 AV |
1246 | if (p->mnt_parent != p) { |
1247 | p->mnt_parent->mnt_ghosts++; | |
5f57cbcc | 1248 | dentry_reset_mounted(p->mnt_parent, p->mnt_mountpoint); |
7c4b93d8 | 1249 | } |
a05964f3 | 1250 | change_mnt_propagation(p, MS_PRIVATE); |
1da177e4 | 1251 | } |
7b8a53fd | 1252 | list_splice(&tmp_list, kill); |
1da177e4 LT |
1253 | } |
1254 | ||
c35038be AV |
1255 | static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts); |
1256 | ||
1da177e4 LT |
1257 | static int do_umount(struct vfsmount *mnt, int flags) |
1258 | { | |
b58fed8b | 1259 | struct super_block *sb = mnt->mnt_sb; |
1da177e4 | 1260 | int retval; |
70fbcdf4 | 1261 | LIST_HEAD(umount_list); |
1da177e4 LT |
1262 | |
1263 | retval = security_sb_umount(mnt, flags); | |
1264 | if (retval) | |
1265 | return retval; | |
1266 | ||
1267 | /* | |
1268 | * Allow userspace to request a mountpoint be expired rather than | |
1269 | * unmounting unconditionally. Unmount only happens if: | |
1270 | * (1) the mark is already set (the mark is cleared by mntput()) | |
1271 | * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] | |
1272 | */ | |
1273 | if (flags & MNT_EXPIRE) { | |
6ac08c39 | 1274 | if (mnt == current->fs->root.mnt || |
1da177e4 LT |
1275 | flags & (MNT_FORCE | MNT_DETACH)) |
1276 | return -EINVAL; | |
1277 | ||
b3e19d92 NP |
1278 | /* |
1279 | * probably don't strictly need the lock here if we examined | |
1280 | * all race cases, but it's a slowpath. | |
1281 | */ | |
1282 | br_write_lock(vfsmount_lock); | |
1283 | if (mnt_get_count(mnt) != 2) { | |
bf9faa2a | 1284 | br_write_unlock(vfsmount_lock); |
1da177e4 | 1285 | return -EBUSY; |
b3e19d92 NP |
1286 | } |
1287 | br_write_unlock(vfsmount_lock); | |
1da177e4 LT |
1288 | |
1289 | if (!xchg(&mnt->mnt_expiry_mark, 1)) | |
1290 | return -EAGAIN; | |
1291 | } | |
1292 | ||
1293 | /* | |
1294 | * If we may have to abort operations to get out of this | |
1295 | * mount, and they will themselves hold resources we must | |
1296 | * allow the fs to do things. In the Unix tradition of | |
1297 | * 'Gee thats tricky lets do it in userspace' the umount_begin | |
1298 | * might fail to complete on the first run through as other tasks | |
1299 | * must return, and the like. Thats for the mount program to worry | |
1300 | * about for the moment. | |
1301 | */ | |
1302 | ||
42faad99 | 1303 | if (flags & MNT_FORCE && sb->s_op->umount_begin) { |
42faad99 | 1304 | sb->s_op->umount_begin(sb); |
42faad99 | 1305 | } |
1da177e4 LT |
1306 | |
1307 | /* | |
1308 | * No sense to grab the lock for this test, but test itself looks | |
1309 | * somewhat bogus. Suggestions for better replacement? | |
1310 | * Ho-hum... In principle, we might treat that as umount + switch | |
1311 | * to rootfs. GC would eventually take care of the old vfsmount. | |
1312 | * Actually it makes sense, especially if rootfs would contain a | |
1313 | * /reboot - static binary that would close all descriptors and | |
1314 | * call reboot(9). Then init(8) could umount root and exec /reboot. | |
1315 | */ | |
6ac08c39 | 1316 | if (mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { |
1da177e4 LT |
1317 | /* |
1318 | * Special case for "unmounting" root ... | |
1319 | * we just try to remount it readonly. | |
1320 | */ | |
1321 | down_write(&sb->s_umount); | |
4aa98cf7 | 1322 | if (!(sb->s_flags & MS_RDONLY)) |
1da177e4 | 1323 | retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); |
1da177e4 LT |
1324 | up_write(&sb->s_umount); |
1325 | return retval; | |
1326 | } | |
1327 | ||
390c6843 | 1328 | down_write(&namespace_sem); |
99b7db7b | 1329 | br_write_lock(vfsmount_lock); |
5addc5dd | 1330 | event++; |
1da177e4 | 1331 | |
c35038be AV |
1332 | if (!(flags & MNT_DETACH)) |
1333 | shrink_submounts(mnt, &umount_list); | |
1334 | ||
1da177e4 | 1335 | retval = -EBUSY; |
a05964f3 | 1336 | if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { |
1da177e4 | 1337 | if (!list_empty(&mnt->mnt_list)) |
a05964f3 | 1338 | umount_tree(mnt, 1, &umount_list); |
1da177e4 LT |
1339 | retval = 0; |
1340 | } | |
99b7db7b | 1341 | br_write_unlock(vfsmount_lock); |
390c6843 | 1342 | up_write(&namespace_sem); |
70fbcdf4 | 1343 | release_mounts(&umount_list); |
1da177e4 LT |
1344 | return retval; |
1345 | } | |
1346 | ||
1347 | /* | |
1348 | * Now umount can handle mount points as well as block devices. | |
1349 | * This is important for filesystems which use unnamed block devices. | |
1350 | * | |
1351 | * We now support a flag for forced unmount like the other 'big iron' | |
1352 | * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD | |
1353 | */ | |
1354 | ||
bdc480e3 | 1355 | SYSCALL_DEFINE2(umount, char __user *, name, int, flags) |
1da177e4 | 1356 | { |
2d8f3038 | 1357 | struct path path; |
1da177e4 | 1358 | int retval; |
db1f05bb | 1359 | int lookup_flags = 0; |
1da177e4 | 1360 | |
db1f05bb MS |
1361 | if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) |
1362 | return -EINVAL; | |
1363 | ||
1364 | if (!(flags & UMOUNT_NOFOLLOW)) | |
1365 | lookup_flags |= LOOKUP_FOLLOW; | |
1366 | ||
1367 | retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); | |
1da177e4 LT |
1368 | if (retval) |
1369 | goto out; | |
1370 | retval = -EINVAL; | |
2d8f3038 | 1371 | if (path.dentry != path.mnt->mnt_root) |
1da177e4 | 1372 | goto dput_and_out; |
2d8f3038 | 1373 | if (!check_mnt(path.mnt)) |
1da177e4 LT |
1374 | goto dput_and_out; |
1375 | ||
1376 | retval = -EPERM; | |
1377 | if (!capable(CAP_SYS_ADMIN)) | |
1378 | goto dput_and_out; | |
1379 | ||
2d8f3038 | 1380 | retval = do_umount(path.mnt, flags); |
1da177e4 | 1381 | dput_and_out: |
429731b1 | 1382 | /* we mustn't call path_put() as that would clear mnt_expiry_mark */ |
2d8f3038 AV |
1383 | dput(path.dentry); |
1384 | mntput_no_expire(path.mnt); | |
1da177e4 LT |
1385 | out: |
1386 | return retval; | |
1387 | } | |
1388 | ||
1389 | #ifdef __ARCH_WANT_SYS_OLDUMOUNT | |
1390 | ||
1391 | /* | |
b58fed8b | 1392 | * The 2.0 compatible umount. No flags. |
1da177e4 | 1393 | */ |
bdc480e3 | 1394 | SYSCALL_DEFINE1(oldumount, char __user *, name) |
1da177e4 | 1395 | { |
b58fed8b | 1396 | return sys_umount(name, 0); |
1da177e4 LT |
1397 | } |
1398 | ||
1399 | #endif | |
1400 | ||
2d92ab3c | 1401 | static int mount_is_safe(struct path *path) |
1da177e4 LT |
1402 | { |
1403 | if (capable(CAP_SYS_ADMIN)) | |
1404 | return 0; | |
1405 | return -EPERM; | |
1406 | #ifdef notyet | |
2d92ab3c | 1407 | if (S_ISLNK(path->dentry->d_inode->i_mode)) |
1da177e4 | 1408 | return -EPERM; |
2d92ab3c | 1409 | if (path->dentry->d_inode->i_mode & S_ISVTX) { |
da9592ed | 1410 | if (current_uid() != path->dentry->d_inode->i_uid) |
1da177e4 LT |
1411 | return -EPERM; |
1412 | } | |
2d92ab3c | 1413 | if (inode_permission(path->dentry->d_inode, MAY_WRITE)) |
1da177e4 LT |
1414 | return -EPERM; |
1415 | return 0; | |
1416 | #endif | |
1417 | } | |
1418 | ||
b90fa9ae | 1419 | struct vfsmount *copy_tree(struct vfsmount *mnt, struct dentry *dentry, |
36341f64 | 1420 | int flag) |
1da177e4 LT |
1421 | { |
1422 | struct vfsmount *res, *p, *q, *r, *s; | |
1a390689 | 1423 | struct path path; |
1da177e4 | 1424 | |
9676f0c6 RP |
1425 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) |
1426 | return NULL; | |
1427 | ||
36341f64 | 1428 | res = q = clone_mnt(mnt, dentry, flag); |
1da177e4 LT |
1429 | if (!q) |
1430 | goto Enomem; | |
1431 | q->mnt_mountpoint = mnt->mnt_mountpoint; | |
1432 | ||
1433 | p = mnt; | |
fdadd65f | 1434 | list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { |
7ec02ef1 | 1435 | if (!is_subdir(r->mnt_mountpoint, dentry)) |
1da177e4 LT |
1436 | continue; |
1437 | ||
1438 | for (s = r; s; s = next_mnt(s, r)) { | |
9676f0c6 RP |
1439 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { |
1440 | s = skip_mnt_tree(s); | |
1441 | continue; | |
1442 | } | |
1da177e4 LT |
1443 | while (p != s->mnt_parent) { |
1444 | p = p->mnt_parent; | |
1445 | q = q->mnt_parent; | |
1446 | } | |
1447 | p = s; | |
1a390689 AV |
1448 | path.mnt = q; |
1449 | path.dentry = p->mnt_mountpoint; | |
36341f64 | 1450 | q = clone_mnt(p, p->mnt_root, flag); |
1da177e4 LT |
1451 | if (!q) |
1452 | goto Enomem; | |
99b7db7b | 1453 | br_write_lock(vfsmount_lock); |
1da177e4 | 1454 | list_add_tail(&q->mnt_list, &res->mnt_list); |
1a390689 | 1455 | attach_mnt(q, &path); |
99b7db7b | 1456 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
1457 | } |
1458 | } | |
1459 | return res; | |
b58fed8b | 1460 | Enomem: |
1da177e4 | 1461 | if (res) { |
70fbcdf4 | 1462 | LIST_HEAD(umount_list); |
99b7db7b | 1463 | br_write_lock(vfsmount_lock); |
a05964f3 | 1464 | umount_tree(res, 0, &umount_list); |
99b7db7b | 1465 | br_write_unlock(vfsmount_lock); |
70fbcdf4 | 1466 | release_mounts(&umount_list); |
1da177e4 LT |
1467 | } |
1468 | return NULL; | |
1469 | } | |
1470 | ||
589ff870 | 1471 | struct vfsmount *collect_mounts(struct path *path) |
8aec0809 AV |
1472 | { |
1473 | struct vfsmount *tree; | |
1a60a280 | 1474 | down_write(&namespace_sem); |
589ff870 | 1475 | tree = copy_tree(path->mnt, path->dentry, CL_COPY_ALL | CL_PRIVATE); |
1a60a280 | 1476 | up_write(&namespace_sem); |
8aec0809 AV |
1477 | return tree; |
1478 | } | |
1479 | ||
1480 | void drop_collected_mounts(struct vfsmount *mnt) | |
1481 | { | |
1482 | LIST_HEAD(umount_list); | |
1a60a280 | 1483 | down_write(&namespace_sem); |
99b7db7b | 1484 | br_write_lock(vfsmount_lock); |
8aec0809 | 1485 | umount_tree(mnt, 0, &umount_list); |
99b7db7b | 1486 | br_write_unlock(vfsmount_lock); |
1a60a280 | 1487 | up_write(&namespace_sem); |
8aec0809 AV |
1488 | release_mounts(&umount_list); |
1489 | } | |
1490 | ||
1f707137 AV |
1491 | int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, |
1492 | struct vfsmount *root) | |
1493 | { | |
1494 | struct vfsmount *mnt; | |
1495 | int res = f(root, arg); | |
1496 | if (res) | |
1497 | return res; | |
1498 | list_for_each_entry(mnt, &root->mnt_list, mnt_list) { | |
1499 | res = f(mnt, arg); | |
1500 | if (res) | |
1501 | return res; | |
1502 | } | |
1503 | return 0; | |
1504 | } | |
1505 | ||
719f5d7f MS |
1506 | static void cleanup_group_ids(struct vfsmount *mnt, struct vfsmount *end) |
1507 | { | |
1508 | struct vfsmount *p; | |
1509 | ||
1510 | for (p = mnt; p != end; p = next_mnt(p, mnt)) { | |
1511 | if (p->mnt_group_id && !IS_MNT_SHARED(p)) | |
1512 | mnt_release_group_id(p); | |
1513 | } | |
1514 | } | |
1515 | ||
1516 | static int invent_group_ids(struct vfsmount *mnt, bool recurse) | |
1517 | { | |
1518 | struct vfsmount *p; | |
1519 | ||
1520 | for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { | |
1521 | if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { | |
1522 | int err = mnt_alloc_group_id(p); | |
1523 | if (err) { | |
1524 | cleanup_group_ids(mnt, p); | |
1525 | return err; | |
1526 | } | |
1527 | } | |
1528 | } | |
1529 | ||
1530 | return 0; | |
1531 | } | |
1532 | ||
b90fa9ae RP |
1533 | /* |
1534 | * @source_mnt : mount tree to be attached | |
21444403 RP |
1535 | * @nd : place the mount tree @source_mnt is attached |
1536 | * @parent_nd : if non-null, detach the source_mnt from its parent and | |
1537 | * store the parent mount and mountpoint dentry. | |
1538 | * (done when source_mnt is moved) | |
b90fa9ae RP |
1539 | * |
1540 | * NOTE: in the table below explains the semantics when a source mount | |
1541 | * of a given type is attached to a destination mount of a given type. | |
9676f0c6 RP |
1542 | * --------------------------------------------------------------------------- |
1543 | * | BIND MOUNT OPERATION | | |
1544 | * |************************************************************************** | |
1545 | * | source-->| shared | private | slave | unbindable | | |
1546 | * | dest | | | | | | |
1547 | * | | | | | | | | |
1548 | * | v | | | | | | |
1549 | * |************************************************************************** | |
1550 | * | shared | shared (++) | shared (+) | shared(+++)| invalid | | |
1551 | * | | | | | | | |
1552 | * |non-shared| shared (+) | private | slave (*) | invalid | | |
1553 | * *************************************************************************** | |
b90fa9ae RP |
1554 | * A bind operation clones the source mount and mounts the clone on the |
1555 | * destination mount. | |
1556 | * | |
1557 | * (++) the cloned mount is propagated to all the mounts in the propagation | |
1558 | * tree of the destination mount and the cloned mount is added to | |
1559 | * the peer group of the source mount. | |
1560 | * (+) the cloned mount is created under the destination mount and is marked | |
1561 | * as shared. The cloned mount is added to the peer group of the source | |
1562 | * mount. | |
5afe0022 RP |
1563 | * (+++) the mount is propagated to all the mounts in the propagation tree |
1564 | * of the destination mount and the cloned mount is made slave | |
1565 | * of the same master as that of the source mount. The cloned mount | |
1566 | * is marked as 'shared and slave'. | |
1567 | * (*) the cloned mount is made a slave of the same master as that of the | |
1568 | * source mount. | |
1569 | * | |
9676f0c6 RP |
1570 | * --------------------------------------------------------------------------- |
1571 | * | MOVE MOUNT OPERATION | | |
1572 | * |************************************************************************** | |
1573 | * | source-->| shared | private | slave | unbindable | | |
1574 | * | dest | | | | | | |
1575 | * | | | | | | | | |
1576 | * | v | | | | | | |
1577 | * |************************************************************************** | |
1578 | * | shared | shared (+) | shared (+) | shared(+++) | invalid | | |
1579 | * | | | | | | | |
1580 | * |non-shared| shared (+*) | private | slave (*) | unbindable | | |
1581 | * *************************************************************************** | |
5afe0022 RP |
1582 | * |
1583 | * (+) the mount is moved to the destination. And is then propagated to | |
1584 | * all the mounts in the propagation tree of the destination mount. | |
21444403 | 1585 | * (+*) the mount is moved to the destination. |
5afe0022 RP |
1586 | * (+++) the mount is moved to the destination and is then propagated to |
1587 | * all the mounts belonging to the destination mount's propagation tree. | |
1588 | * the mount is marked as 'shared and slave'. | |
1589 | * (*) the mount continues to be a slave at the new location. | |
b90fa9ae RP |
1590 | * |
1591 | * if the source mount is a tree, the operations explained above is | |
1592 | * applied to each mount in the tree. | |
1593 | * Must be called without spinlocks held, since this function can sleep | |
1594 | * in allocations. | |
1595 | */ | |
1596 | static int attach_recursive_mnt(struct vfsmount *source_mnt, | |
1a390689 | 1597 | struct path *path, struct path *parent_path) |
b90fa9ae RP |
1598 | { |
1599 | LIST_HEAD(tree_list); | |
1a390689 AV |
1600 | struct vfsmount *dest_mnt = path->mnt; |
1601 | struct dentry *dest_dentry = path->dentry; | |
b90fa9ae | 1602 | struct vfsmount *child, *p; |
719f5d7f | 1603 | int err; |
b90fa9ae | 1604 | |
719f5d7f MS |
1605 | if (IS_MNT_SHARED(dest_mnt)) { |
1606 | err = invent_group_ids(source_mnt, true); | |
1607 | if (err) | |
1608 | goto out; | |
1609 | } | |
1610 | err = propagate_mnt(dest_mnt, dest_dentry, source_mnt, &tree_list); | |
1611 | if (err) | |
1612 | goto out_cleanup_ids; | |
b90fa9ae | 1613 | |
99b7db7b | 1614 | br_write_lock(vfsmount_lock); |
df1a1ad2 | 1615 | |
b90fa9ae RP |
1616 | if (IS_MNT_SHARED(dest_mnt)) { |
1617 | for (p = source_mnt; p; p = next_mnt(p, source_mnt)) | |
1618 | set_mnt_shared(p); | |
1619 | } | |
1a390689 AV |
1620 | if (parent_path) { |
1621 | detach_mnt(source_mnt, parent_path); | |
1622 | attach_mnt(source_mnt, path); | |
e5d67f07 | 1623 | touch_mnt_namespace(parent_path->mnt->mnt_ns); |
21444403 RP |
1624 | } else { |
1625 | mnt_set_mountpoint(dest_mnt, dest_dentry, source_mnt); | |
1626 | commit_tree(source_mnt); | |
1627 | } | |
b90fa9ae RP |
1628 | |
1629 | list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { | |
1630 | list_del_init(&child->mnt_hash); | |
1631 | commit_tree(child); | |
1632 | } | |
99b7db7b NP |
1633 | br_write_unlock(vfsmount_lock); |
1634 | ||
b90fa9ae | 1635 | return 0; |
719f5d7f MS |
1636 | |
1637 | out_cleanup_ids: | |
1638 | if (IS_MNT_SHARED(dest_mnt)) | |
1639 | cleanup_group_ids(source_mnt, NULL); | |
1640 | out: | |
1641 | return err; | |
b90fa9ae RP |
1642 | } |
1643 | ||
8c3ee42e | 1644 | static int graft_tree(struct vfsmount *mnt, struct path *path) |
1da177e4 LT |
1645 | { |
1646 | int err; | |
1647 | if (mnt->mnt_sb->s_flags & MS_NOUSER) | |
1648 | return -EINVAL; | |
1649 | ||
8c3ee42e | 1650 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
1da177e4 LT |
1651 | S_ISDIR(mnt->mnt_root->d_inode->i_mode)) |
1652 | return -ENOTDIR; | |
1653 | ||
1654 | err = -ENOENT; | |
8c3ee42e | 1655 | mutex_lock(&path->dentry->d_inode->i_mutex); |
d83c49f3 | 1656 | if (cant_mount(path->dentry)) |
1da177e4 LT |
1657 | goto out_unlock; |
1658 | ||
f3da392e | 1659 | if (!d_unlinked(path->dentry)) |
8c3ee42e | 1660 | err = attach_recursive_mnt(mnt, path, NULL); |
1da177e4 | 1661 | out_unlock: |
8c3ee42e | 1662 | mutex_unlock(&path->dentry->d_inode->i_mutex); |
1da177e4 LT |
1663 | return err; |
1664 | } | |
1665 | ||
7a2e8a8f VA |
1666 | /* |
1667 | * Sanity check the flags to change_mnt_propagation. | |
1668 | */ | |
1669 | ||
1670 | static int flags_to_propagation_type(int flags) | |
1671 | { | |
1672 | int type = flags & ~MS_REC; | |
1673 | ||
1674 | /* Fail if any non-propagation flags are set */ | |
1675 | if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) | |
1676 | return 0; | |
1677 | /* Only one propagation flag should be set */ | |
1678 | if (!is_power_of_2(type)) | |
1679 | return 0; | |
1680 | return type; | |
1681 | } | |
1682 | ||
07b20889 RP |
1683 | /* |
1684 | * recursively change the type of the mountpoint. | |
1685 | */ | |
0a0d8a46 | 1686 | static int do_change_type(struct path *path, int flag) |
07b20889 | 1687 | { |
2d92ab3c | 1688 | struct vfsmount *m, *mnt = path->mnt; |
07b20889 | 1689 | int recurse = flag & MS_REC; |
7a2e8a8f | 1690 | int type; |
719f5d7f | 1691 | int err = 0; |
07b20889 | 1692 | |
ee6f9582 MS |
1693 | if (!capable(CAP_SYS_ADMIN)) |
1694 | return -EPERM; | |
1695 | ||
2d92ab3c | 1696 | if (path->dentry != path->mnt->mnt_root) |
07b20889 RP |
1697 | return -EINVAL; |
1698 | ||
7a2e8a8f VA |
1699 | type = flags_to_propagation_type(flag); |
1700 | if (!type) | |
1701 | return -EINVAL; | |
1702 | ||
07b20889 | 1703 | down_write(&namespace_sem); |
719f5d7f MS |
1704 | if (type == MS_SHARED) { |
1705 | err = invent_group_ids(mnt, recurse); | |
1706 | if (err) | |
1707 | goto out_unlock; | |
1708 | } | |
1709 | ||
99b7db7b | 1710 | br_write_lock(vfsmount_lock); |
07b20889 RP |
1711 | for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) |
1712 | change_mnt_propagation(m, type); | |
99b7db7b | 1713 | br_write_unlock(vfsmount_lock); |
719f5d7f MS |
1714 | |
1715 | out_unlock: | |
07b20889 | 1716 | up_write(&namespace_sem); |
719f5d7f | 1717 | return err; |
07b20889 RP |
1718 | } |
1719 | ||
1da177e4 LT |
1720 | /* |
1721 | * do loopback mount. | |
1722 | */ | |
0a0d8a46 | 1723 | static int do_loopback(struct path *path, char *old_name, |
2dafe1c4 | 1724 | int recurse) |
1da177e4 | 1725 | { |
2d92ab3c | 1726 | struct path old_path; |
1da177e4 | 1727 | struct vfsmount *mnt = NULL; |
2d92ab3c | 1728 | int err = mount_is_safe(path); |
1da177e4 LT |
1729 | if (err) |
1730 | return err; | |
1731 | if (!old_name || !*old_name) | |
1732 | return -EINVAL; | |
2d92ab3c | 1733 | err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); |
1da177e4 LT |
1734 | if (err) |
1735 | return err; | |
1736 | ||
390c6843 | 1737 | down_write(&namespace_sem); |
1da177e4 | 1738 | err = -EINVAL; |
2d92ab3c | 1739 | if (IS_MNT_UNBINDABLE(old_path.mnt)) |
4ac91378 | 1740 | goto out; |
9676f0c6 | 1741 | |
2d92ab3c | 1742 | if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) |
ccd48bc7 | 1743 | goto out; |
1da177e4 | 1744 | |
ccd48bc7 AV |
1745 | err = -ENOMEM; |
1746 | if (recurse) | |
2d92ab3c | 1747 | mnt = copy_tree(old_path.mnt, old_path.dentry, 0); |
ccd48bc7 | 1748 | else |
2d92ab3c | 1749 | mnt = clone_mnt(old_path.mnt, old_path.dentry, 0); |
ccd48bc7 AV |
1750 | |
1751 | if (!mnt) | |
1752 | goto out; | |
1753 | ||
2d92ab3c | 1754 | err = graft_tree(mnt, path); |
ccd48bc7 | 1755 | if (err) { |
70fbcdf4 | 1756 | LIST_HEAD(umount_list); |
99b7db7b NP |
1757 | |
1758 | br_write_lock(vfsmount_lock); | |
a05964f3 | 1759 | umount_tree(mnt, 0, &umount_list); |
99b7db7b | 1760 | br_write_unlock(vfsmount_lock); |
70fbcdf4 | 1761 | release_mounts(&umount_list); |
5b83d2c5 | 1762 | } |
1da177e4 | 1763 | |
ccd48bc7 | 1764 | out: |
390c6843 | 1765 | up_write(&namespace_sem); |
2d92ab3c | 1766 | path_put(&old_path); |
1da177e4 LT |
1767 | return err; |
1768 | } | |
1769 | ||
2e4b7fcd DH |
1770 | static int change_mount_flags(struct vfsmount *mnt, int ms_flags) |
1771 | { | |
1772 | int error = 0; | |
1773 | int readonly_request = 0; | |
1774 | ||
1775 | if (ms_flags & MS_RDONLY) | |
1776 | readonly_request = 1; | |
1777 | if (readonly_request == __mnt_is_readonly(mnt)) | |
1778 | return 0; | |
1779 | ||
1780 | if (readonly_request) | |
1781 | error = mnt_make_readonly(mnt); | |
1782 | else | |
1783 | __mnt_unmake_readonly(mnt); | |
1784 | return error; | |
1785 | } | |
1786 | ||
1da177e4 LT |
1787 | /* |
1788 | * change filesystem flags. dir should be a physical root of filesystem. | |
1789 | * If you've mounted a non-root directory somewhere and want to do remount | |
1790 | * on it - tough luck. | |
1791 | */ | |
0a0d8a46 | 1792 | static int do_remount(struct path *path, int flags, int mnt_flags, |
1da177e4 LT |
1793 | void *data) |
1794 | { | |
1795 | int err; | |
2d92ab3c | 1796 | struct super_block *sb = path->mnt->mnt_sb; |
1da177e4 LT |
1797 | |
1798 | if (!capable(CAP_SYS_ADMIN)) | |
1799 | return -EPERM; | |
1800 | ||
2d92ab3c | 1801 | if (!check_mnt(path->mnt)) |
1da177e4 LT |
1802 | return -EINVAL; |
1803 | ||
2d92ab3c | 1804 | if (path->dentry != path->mnt->mnt_root) |
1da177e4 LT |
1805 | return -EINVAL; |
1806 | ||
1807 | down_write(&sb->s_umount); | |
2e4b7fcd | 1808 | if (flags & MS_BIND) |
2d92ab3c | 1809 | err = change_mount_flags(path->mnt, flags); |
4aa98cf7 | 1810 | else |
2e4b7fcd | 1811 | err = do_remount_sb(sb, flags, data, 0); |
7b43a79f | 1812 | if (!err) { |
99b7db7b | 1813 | br_write_lock(vfsmount_lock); |
495d6c9c | 1814 | mnt_flags |= path->mnt->mnt_flags & MNT_PROPAGATION_MASK; |
2d92ab3c | 1815 | path->mnt->mnt_flags = mnt_flags; |
99b7db7b | 1816 | br_write_unlock(vfsmount_lock); |
7b43a79f | 1817 | } |
1da177e4 | 1818 | up_write(&sb->s_umount); |
0e55a7cc | 1819 | if (!err) { |
99b7db7b | 1820 | br_write_lock(vfsmount_lock); |
0e55a7cc | 1821 | touch_mnt_namespace(path->mnt->mnt_ns); |
99b7db7b | 1822 | br_write_unlock(vfsmount_lock); |
0e55a7cc | 1823 | } |
1da177e4 LT |
1824 | return err; |
1825 | } | |
1826 | ||
9676f0c6 RP |
1827 | static inline int tree_contains_unbindable(struct vfsmount *mnt) |
1828 | { | |
1829 | struct vfsmount *p; | |
1830 | for (p = mnt; p; p = next_mnt(p, mnt)) { | |
1831 | if (IS_MNT_UNBINDABLE(p)) | |
1832 | return 1; | |
1833 | } | |
1834 | return 0; | |
1835 | } | |
1836 | ||
0a0d8a46 | 1837 | static int do_move_mount(struct path *path, char *old_name) |
1da177e4 | 1838 | { |
2d92ab3c | 1839 | struct path old_path, parent_path; |
1da177e4 LT |
1840 | struct vfsmount *p; |
1841 | int err = 0; | |
1842 | if (!capable(CAP_SYS_ADMIN)) | |
1843 | return -EPERM; | |
1844 | if (!old_name || !*old_name) | |
1845 | return -EINVAL; | |
2d92ab3c | 1846 | err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); |
1da177e4 LT |
1847 | if (err) |
1848 | return err; | |
1849 | ||
390c6843 | 1850 | down_write(&namespace_sem); |
cc53ce53 DH |
1851 | err = follow_down(path, true); |
1852 | if (err < 0) | |
1853 | goto out; | |
1854 | ||
1da177e4 | 1855 | err = -EINVAL; |
2d92ab3c | 1856 | if (!check_mnt(path->mnt) || !check_mnt(old_path.mnt)) |
1da177e4 LT |
1857 | goto out; |
1858 | ||
1859 | err = -ENOENT; | |
2d92ab3c | 1860 | mutex_lock(&path->dentry->d_inode->i_mutex); |
d83c49f3 | 1861 | if (cant_mount(path->dentry)) |
1da177e4 LT |
1862 | goto out1; |
1863 | ||
f3da392e | 1864 | if (d_unlinked(path->dentry)) |
21444403 | 1865 | goto out1; |
1da177e4 LT |
1866 | |
1867 | err = -EINVAL; | |
2d92ab3c | 1868 | if (old_path.dentry != old_path.mnt->mnt_root) |
21444403 | 1869 | goto out1; |
1da177e4 | 1870 | |
2d92ab3c | 1871 | if (old_path.mnt == old_path.mnt->mnt_parent) |
21444403 | 1872 | goto out1; |
1da177e4 | 1873 | |
2d92ab3c AV |
1874 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
1875 | S_ISDIR(old_path.dentry->d_inode->i_mode)) | |
21444403 RP |
1876 | goto out1; |
1877 | /* | |
1878 | * Don't move a mount residing in a shared parent. | |
1879 | */ | |
2d92ab3c AV |
1880 | if (old_path.mnt->mnt_parent && |
1881 | IS_MNT_SHARED(old_path.mnt->mnt_parent)) | |
21444403 | 1882 | goto out1; |
9676f0c6 RP |
1883 | /* |
1884 | * Don't move a mount tree containing unbindable mounts to a destination | |
1885 | * mount which is shared. | |
1886 | */ | |
2d92ab3c AV |
1887 | if (IS_MNT_SHARED(path->mnt) && |
1888 | tree_contains_unbindable(old_path.mnt)) | |
9676f0c6 | 1889 | goto out1; |
1da177e4 | 1890 | err = -ELOOP; |
2d92ab3c AV |
1891 | for (p = path->mnt; p->mnt_parent != p; p = p->mnt_parent) |
1892 | if (p == old_path.mnt) | |
21444403 | 1893 | goto out1; |
1da177e4 | 1894 | |
2d92ab3c | 1895 | err = attach_recursive_mnt(old_path.mnt, path, &parent_path); |
4ac91378 | 1896 | if (err) |
21444403 | 1897 | goto out1; |
1da177e4 LT |
1898 | |
1899 | /* if the mount is moved, it should no longer be expire | |
1900 | * automatically */ | |
2d92ab3c | 1901 | list_del_init(&old_path.mnt->mnt_expire); |
1da177e4 | 1902 | out1: |
2d92ab3c | 1903 | mutex_unlock(&path->dentry->d_inode->i_mutex); |
1da177e4 | 1904 | out: |
390c6843 | 1905 | up_write(&namespace_sem); |
1da177e4 | 1906 | if (!err) |
1a390689 | 1907 | path_put(&parent_path); |
2d92ab3c | 1908 | path_put(&old_path); |
1da177e4 LT |
1909 | return err; |
1910 | } | |
1911 | ||
b1e75df4 AV |
1912 | static int do_add_mount(struct vfsmount *, struct path *, int); |
1913 | ||
1da177e4 LT |
1914 | /* |
1915 | * create a new mount for userspace and request it to be added into the | |
1916 | * namespace's tree | |
1917 | */ | |
0a0d8a46 | 1918 | static int do_new_mount(struct path *path, char *type, int flags, |
1da177e4 LT |
1919 | int mnt_flags, char *name, void *data) |
1920 | { | |
1921 | struct vfsmount *mnt; | |
15f9a3f3 | 1922 | int err; |
1da177e4 | 1923 | |
eca6f534 | 1924 | if (!type) |
1da177e4 LT |
1925 | return -EINVAL; |
1926 | ||
1927 | /* we need capabilities... */ | |
1928 | if (!capable(CAP_SYS_ADMIN)) | |
1929 | return -EPERM; | |
1930 | ||
1931 | mnt = do_kern_mount(type, flags, name, data); | |
1932 | if (IS_ERR(mnt)) | |
1933 | return PTR_ERR(mnt); | |
1934 | ||
15f9a3f3 AV |
1935 | err = do_add_mount(mnt, path, mnt_flags); |
1936 | if (err) | |
1937 | mntput(mnt); | |
1938 | return err; | |
1da177e4 LT |
1939 | } |
1940 | ||
19a167af AV |
1941 | int finish_automount(struct vfsmount *m, struct path *path) |
1942 | { | |
1943 | int err; | |
1944 | /* The new mount record should have at least 2 refs to prevent it being | |
1945 | * expired before we get a chance to add it | |
1946 | */ | |
1947 | BUG_ON(mnt_get_count(m) < 2); | |
1948 | ||
1949 | if (m->mnt_sb == path->mnt->mnt_sb && | |
1950 | m->mnt_root == path->dentry) { | |
b1e75df4 AV |
1951 | err = -ELOOP; |
1952 | goto fail; | |
19a167af AV |
1953 | } |
1954 | ||
19a167af | 1955 | err = do_add_mount(m, path, path->mnt->mnt_flags | MNT_SHRINKABLE); |
b1e75df4 AV |
1956 | if (!err) |
1957 | return 0; | |
1958 | fail: | |
1959 | /* remove m from any expiration list it may be on */ | |
1960 | if (!list_empty(&m->mnt_expire)) { | |
1961 | down_write(&namespace_sem); | |
1962 | br_write_lock(vfsmount_lock); | |
1963 | list_del_init(&m->mnt_expire); | |
1964 | br_write_unlock(vfsmount_lock); | |
1965 | up_write(&namespace_sem); | |
19a167af | 1966 | } |
b1e75df4 AV |
1967 | mntput(m); |
1968 | mntput(m); | |
19a167af AV |
1969 | return err; |
1970 | } | |
1971 | ||
1da177e4 LT |
1972 | /* |
1973 | * add a mount into a namespace's mount tree | |
1da177e4 | 1974 | */ |
b1e75df4 | 1975 | static int do_add_mount(struct vfsmount *newmnt, struct path *path, int mnt_flags) |
1da177e4 LT |
1976 | { |
1977 | int err; | |
1978 | ||
8089352a | 1979 | mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); |
27d55f1f | 1980 | |
390c6843 | 1981 | down_write(&namespace_sem); |
1da177e4 | 1982 | /* Something was mounted here while we slept */ |
cc53ce53 DH |
1983 | err = follow_down(path, true); |
1984 | if (err < 0) | |
1985 | goto unlock; | |
1986 | ||
1da177e4 | 1987 | err = -EINVAL; |
dd5cae6e | 1988 | if (!(mnt_flags & MNT_SHRINKABLE) && !check_mnt(path->mnt)) |
1da177e4 LT |
1989 | goto unlock; |
1990 | ||
1991 | /* Refuse the same filesystem on the same mount point */ | |
1992 | err = -EBUSY; | |
8d66bf54 AV |
1993 | if (path->mnt->mnt_sb == newmnt->mnt_sb && |
1994 | path->mnt->mnt_root == path->dentry) | |
1da177e4 LT |
1995 | goto unlock; |
1996 | ||
1997 | err = -EINVAL; | |
1998 | if (S_ISLNK(newmnt->mnt_root->d_inode->i_mode)) | |
1999 | goto unlock; | |
2000 | ||
2001 | newmnt->mnt_flags = mnt_flags; | |
b1e75df4 | 2002 | err = graft_tree(newmnt, path); |
1da177e4 LT |
2003 | |
2004 | unlock: | |
390c6843 | 2005 | up_write(&namespace_sem); |
1da177e4 LT |
2006 | return err; |
2007 | } | |
2008 | ||
ea5b778a DH |
2009 | /** |
2010 | * mnt_set_expiry - Put a mount on an expiration list | |
2011 | * @mnt: The mount to list. | |
2012 | * @expiry_list: The list to add the mount to. | |
2013 | */ | |
2014 | void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) | |
2015 | { | |
2016 | down_write(&namespace_sem); | |
2017 | br_write_lock(vfsmount_lock); | |
2018 | ||
2019 | list_add_tail(&mnt->mnt_expire, expiry_list); | |
2020 | ||
2021 | br_write_unlock(vfsmount_lock); | |
2022 | up_write(&namespace_sem); | |
2023 | } | |
2024 | EXPORT_SYMBOL(mnt_set_expiry); | |
2025 | ||
1da177e4 LT |
2026 | /* |
2027 | * process a list of expirable mountpoints with the intent of discarding any | |
2028 | * mountpoints that aren't in use and haven't been touched since last we came | |
2029 | * here | |
2030 | */ | |
2031 | void mark_mounts_for_expiry(struct list_head *mounts) | |
2032 | { | |
1da177e4 LT |
2033 | struct vfsmount *mnt, *next; |
2034 | LIST_HEAD(graveyard); | |
bcc5c7d2 | 2035 | LIST_HEAD(umounts); |
1da177e4 LT |
2036 | |
2037 | if (list_empty(mounts)) | |
2038 | return; | |
2039 | ||
bcc5c7d2 | 2040 | down_write(&namespace_sem); |
99b7db7b | 2041 | br_write_lock(vfsmount_lock); |
1da177e4 LT |
2042 | |
2043 | /* extract from the expiration list every vfsmount that matches the | |
2044 | * following criteria: | |
2045 | * - only referenced by its parent vfsmount | |
2046 | * - still marked for expiry (marked on the last call here; marks are | |
2047 | * cleared by mntput()) | |
2048 | */ | |
55e700b9 | 2049 | list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { |
1da177e4 | 2050 | if (!xchg(&mnt->mnt_expiry_mark, 1) || |
bcc5c7d2 | 2051 | propagate_mount_busy(mnt, 1)) |
1da177e4 | 2052 | continue; |
55e700b9 | 2053 | list_move(&mnt->mnt_expire, &graveyard); |
1da177e4 | 2054 | } |
bcc5c7d2 AV |
2055 | while (!list_empty(&graveyard)) { |
2056 | mnt = list_first_entry(&graveyard, struct vfsmount, mnt_expire); | |
2057 | touch_mnt_namespace(mnt->mnt_ns); | |
2058 | umount_tree(mnt, 1, &umounts); | |
2059 | } | |
99b7db7b | 2060 | br_write_unlock(vfsmount_lock); |
bcc5c7d2 AV |
2061 | up_write(&namespace_sem); |
2062 | ||
2063 | release_mounts(&umounts); | |
5528f911 TM |
2064 | } |
2065 | ||
2066 | EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); | |
2067 | ||
2068 | /* | |
2069 | * Ripoff of 'select_parent()' | |
2070 | * | |
2071 | * search the list of submounts for a given mountpoint, and move any | |
2072 | * shrinkable submounts to the 'graveyard' list. | |
2073 | */ | |
2074 | static int select_submounts(struct vfsmount *parent, struct list_head *graveyard) | |
2075 | { | |
2076 | struct vfsmount *this_parent = parent; | |
2077 | struct list_head *next; | |
2078 | int found = 0; | |
2079 | ||
2080 | repeat: | |
2081 | next = this_parent->mnt_mounts.next; | |
2082 | resume: | |
2083 | while (next != &this_parent->mnt_mounts) { | |
2084 | struct list_head *tmp = next; | |
2085 | struct vfsmount *mnt = list_entry(tmp, struct vfsmount, mnt_child); | |
2086 | ||
2087 | next = tmp->next; | |
2088 | if (!(mnt->mnt_flags & MNT_SHRINKABLE)) | |
1da177e4 | 2089 | continue; |
5528f911 TM |
2090 | /* |
2091 | * Descend a level if the d_mounts list is non-empty. | |
2092 | */ | |
2093 | if (!list_empty(&mnt->mnt_mounts)) { | |
2094 | this_parent = mnt; | |
2095 | goto repeat; | |
2096 | } | |
1da177e4 | 2097 | |
5528f911 | 2098 | if (!propagate_mount_busy(mnt, 1)) { |
5528f911 TM |
2099 | list_move_tail(&mnt->mnt_expire, graveyard); |
2100 | found++; | |
2101 | } | |
1da177e4 | 2102 | } |
5528f911 TM |
2103 | /* |
2104 | * All done at this level ... ascend and resume the search | |
2105 | */ | |
2106 | if (this_parent != parent) { | |
2107 | next = this_parent->mnt_child.next; | |
2108 | this_parent = this_parent->mnt_parent; | |
2109 | goto resume; | |
2110 | } | |
2111 | return found; | |
2112 | } | |
2113 | ||
2114 | /* | |
2115 | * process a list of expirable mountpoints with the intent of discarding any | |
2116 | * submounts of a specific parent mountpoint | |
99b7db7b NP |
2117 | * |
2118 | * vfsmount_lock must be held for write | |
5528f911 | 2119 | */ |
c35038be | 2120 | static void shrink_submounts(struct vfsmount *mnt, struct list_head *umounts) |
5528f911 TM |
2121 | { |
2122 | LIST_HEAD(graveyard); | |
c35038be | 2123 | struct vfsmount *m; |
5528f911 | 2124 | |
5528f911 | 2125 | /* extract submounts of 'mountpoint' from the expiration list */ |
c35038be | 2126 | while (select_submounts(mnt, &graveyard)) { |
bcc5c7d2 | 2127 | while (!list_empty(&graveyard)) { |
c35038be | 2128 | m = list_first_entry(&graveyard, struct vfsmount, |
bcc5c7d2 | 2129 | mnt_expire); |
afef80b3 EB |
2130 | touch_mnt_namespace(m->mnt_ns); |
2131 | umount_tree(m, 1, umounts); | |
bcc5c7d2 AV |
2132 | } |
2133 | } | |
1da177e4 LT |
2134 | } |
2135 | ||
1da177e4 LT |
2136 | /* |
2137 | * Some copy_from_user() implementations do not return the exact number of | |
2138 | * bytes remaining to copy on a fault. But copy_mount_options() requires that. | |
2139 | * Note that this function differs from copy_from_user() in that it will oops | |
2140 | * on bad values of `to', rather than returning a short copy. | |
2141 | */ | |
b58fed8b RP |
2142 | static long exact_copy_from_user(void *to, const void __user * from, |
2143 | unsigned long n) | |
1da177e4 LT |
2144 | { |
2145 | char *t = to; | |
2146 | const char __user *f = from; | |
2147 | char c; | |
2148 | ||
2149 | if (!access_ok(VERIFY_READ, from, n)) | |
2150 | return n; | |
2151 | ||
2152 | while (n) { | |
2153 | if (__get_user(c, f)) { | |
2154 | memset(t, 0, n); | |
2155 | break; | |
2156 | } | |
2157 | *t++ = c; | |
2158 | f++; | |
2159 | n--; | |
2160 | } | |
2161 | return n; | |
2162 | } | |
2163 | ||
b58fed8b | 2164 | int copy_mount_options(const void __user * data, unsigned long *where) |
1da177e4 LT |
2165 | { |
2166 | int i; | |
2167 | unsigned long page; | |
2168 | unsigned long size; | |
b58fed8b | 2169 | |
1da177e4 LT |
2170 | *where = 0; |
2171 | if (!data) | |
2172 | return 0; | |
2173 | ||
2174 | if (!(page = __get_free_page(GFP_KERNEL))) | |
2175 | return -ENOMEM; | |
2176 | ||
2177 | /* We only care that *some* data at the address the user | |
2178 | * gave us is valid. Just in case, we'll zero | |
2179 | * the remainder of the page. | |
2180 | */ | |
2181 | /* copy_from_user cannot cross TASK_SIZE ! */ | |
2182 | size = TASK_SIZE - (unsigned long)data; | |
2183 | if (size > PAGE_SIZE) | |
2184 | size = PAGE_SIZE; | |
2185 | ||
2186 | i = size - exact_copy_from_user((void *)page, data, size); | |
2187 | if (!i) { | |
b58fed8b | 2188 | free_page(page); |
1da177e4 LT |
2189 | return -EFAULT; |
2190 | } | |
2191 | if (i != PAGE_SIZE) | |
2192 | memset((char *)page + i, 0, PAGE_SIZE - i); | |
2193 | *where = page; | |
2194 | return 0; | |
2195 | } | |
2196 | ||
eca6f534 VN |
2197 | int copy_mount_string(const void __user *data, char **where) |
2198 | { | |
2199 | char *tmp; | |
2200 | ||
2201 | if (!data) { | |
2202 | *where = NULL; | |
2203 | return 0; | |
2204 | } | |
2205 | ||
2206 | tmp = strndup_user(data, PAGE_SIZE); | |
2207 | if (IS_ERR(tmp)) | |
2208 | return PTR_ERR(tmp); | |
2209 | ||
2210 | *where = tmp; | |
2211 | return 0; | |
2212 | } | |
2213 | ||
1da177e4 LT |
2214 | /* |
2215 | * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to | |
2216 | * be given to the mount() call (ie: read-only, no-dev, no-suid etc). | |
2217 | * | |
2218 | * data is a (void *) that can point to any structure up to | |
2219 | * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent | |
2220 | * information (or be NULL). | |
2221 | * | |
2222 | * Pre-0.97 versions of mount() didn't have a flags word. | |
2223 | * When the flags word was introduced its top half was required | |
2224 | * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. | |
2225 | * Therefore, if this magic number is present, it carries no information | |
2226 | * and must be discarded. | |
2227 | */ | |
b58fed8b | 2228 | long do_mount(char *dev_name, char *dir_name, char *type_page, |
1da177e4 LT |
2229 | unsigned long flags, void *data_page) |
2230 | { | |
2d92ab3c | 2231 | struct path path; |
1da177e4 LT |
2232 | int retval = 0; |
2233 | int mnt_flags = 0; | |
2234 | ||
2235 | /* Discard magic */ | |
2236 | if ((flags & MS_MGC_MSK) == MS_MGC_VAL) | |
2237 | flags &= ~MS_MGC_MSK; | |
2238 | ||
2239 | /* Basic sanity checks */ | |
2240 | ||
2241 | if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) | |
2242 | return -EINVAL; | |
1da177e4 LT |
2243 | |
2244 | if (data_page) | |
2245 | ((char *)data_page)[PAGE_SIZE - 1] = 0; | |
2246 | ||
a27ab9f2 TH |
2247 | /* ... and get the mountpoint */ |
2248 | retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); | |
2249 | if (retval) | |
2250 | return retval; | |
2251 | ||
2252 | retval = security_sb_mount(dev_name, &path, | |
2253 | type_page, flags, data_page); | |
2254 | if (retval) | |
2255 | goto dput_out; | |
2256 | ||
613cbe3d AK |
2257 | /* Default to relatime unless overriden */ |
2258 | if (!(flags & MS_NOATIME)) | |
2259 | mnt_flags |= MNT_RELATIME; | |
0a1c01c9 | 2260 | |
1da177e4 LT |
2261 | /* Separate the per-mountpoint flags */ |
2262 | if (flags & MS_NOSUID) | |
2263 | mnt_flags |= MNT_NOSUID; | |
2264 | if (flags & MS_NODEV) | |
2265 | mnt_flags |= MNT_NODEV; | |
2266 | if (flags & MS_NOEXEC) | |
2267 | mnt_flags |= MNT_NOEXEC; | |
fc33a7bb CH |
2268 | if (flags & MS_NOATIME) |
2269 | mnt_flags |= MNT_NOATIME; | |
2270 | if (flags & MS_NODIRATIME) | |
2271 | mnt_flags |= MNT_NODIRATIME; | |
d0adde57 MG |
2272 | if (flags & MS_STRICTATIME) |
2273 | mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); | |
2e4b7fcd DH |
2274 | if (flags & MS_RDONLY) |
2275 | mnt_flags |= MNT_READONLY; | |
fc33a7bb | 2276 | |
7a4dec53 | 2277 | flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | |
d0adde57 MG |
2278 | MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | |
2279 | MS_STRICTATIME); | |
1da177e4 | 2280 | |
1da177e4 | 2281 | if (flags & MS_REMOUNT) |
2d92ab3c | 2282 | retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, |
1da177e4 LT |
2283 | data_page); |
2284 | else if (flags & MS_BIND) | |
2d92ab3c | 2285 | retval = do_loopback(&path, dev_name, flags & MS_REC); |
9676f0c6 | 2286 | else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) |
2d92ab3c | 2287 | retval = do_change_type(&path, flags); |
1da177e4 | 2288 | else if (flags & MS_MOVE) |
2d92ab3c | 2289 | retval = do_move_mount(&path, dev_name); |
1da177e4 | 2290 | else |
2d92ab3c | 2291 | retval = do_new_mount(&path, type_page, flags, mnt_flags, |
1da177e4 LT |
2292 | dev_name, data_page); |
2293 | dput_out: | |
2d92ab3c | 2294 | path_put(&path); |
1da177e4 LT |
2295 | return retval; |
2296 | } | |
2297 | ||
cf8d2c11 TM |
2298 | static struct mnt_namespace *alloc_mnt_ns(void) |
2299 | { | |
2300 | struct mnt_namespace *new_ns; | |
2301 | ||
2302 | new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); | |
2303 | if (!new_ns) | |
2304 | return ERR_PTR(-ENOMEM); | |
2305 | atomic_set(&new_ns->count, 1); | |
2306 | new_ns->root = NULL; | |
2307 | INIT_LIST_HEAD(&new_ns->list); | |
2308 | init_waitqueue_head(&new_ns->poll); | |
2309 | new_ns->event = 0; | |
2310 | return new_ns; | |
2311 | } | |
2312 | ||
f03c6599 AV |
2313 | void mnt_make_longterm(struct vfsmount *mnt) |
2314 | { | |
7e3d0eb0 | 2315 | __mnt_make_longterm(mnt); |
f03c6599 AV |
2316 | } |
2317 | ||
2318 | void mnt_make_shortterm(struct vfsmount *mnt) | |
2319 | { | |
7e3d0eb0 | 2320 | #ifdef CONFIG_SMP |
f03c6599 AV |
2321 | if (atomic_add_unless(&mnt->mnt_longterm, -1, 1)) |
2322 | return; | |
2323 | br_write_lock(vfsmount_lock); | |
2324 | atomic_dec(&mnt->mnt_longterm); | |
2325 | br_write_unlock(vfsmount_lock); | |
7e3d0eb0 | 2326 | #endif |
f03c6599 AV |
2327 | } |
2328 | ||
741a2951 JD |
2329 | /* |
2330 | * Allocate a new namespace structure and populate it with contents | |
2331 | * copied from the namespace of the passed in task structure. | |
2332 | */ | |
e3222c4e | 2333 | static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, |
6b3286ed | 2334 | struct fs_struct *fs) |
1da177e4 | 2335 | { |
6b3286ed | 2336 | struct mnt_namespace *new_ns; |
7f2da1e7 | 2337 | struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; |
1da177e4 LT |
2338 | struct vfsmount *p, *q; |
2339 | ||
cf8d2c11 TM |
2340 | new_ns = alloc_mnt_ns(); |
2341 | if (IS_ERR(new_ns)) | |
2342 | return new_ns; | |
1da177e4 | 2343 | |
390c6843 | 2344 | down_write(&namespace_sem); |
1da177e4 | 2345 | /* First pass: copy the tree topology */ |
6b3286ed | 2346 | new_ns->root = copy_tree(mnt_ns->root, mnt_ns->root->mnt_root, |
9676f0c6 | 2347 | CL_COPY_ALL | CL_EXPIRE); |
1da177e4 | 2348 | if (!new_ns->root) { |
390c6843 | 2349 | up_write(&namespace_sem); |
1da177e4 | 2350 | kfree(new_ns); |
5cc4a034 | 2351 | return ERR_PTR(-ENOMEM); |
1da177e4 | 2352 | } |
99b7db7b | 2353 | br_write_lock(vfsmount_lock); |
1da177e4 | 2354 | list_add_tail(&new_ns->list, &new_ns->root->mnt_list); |
99b7db7b | 2355 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
2356 | |
2357 | /* | |
2358 | * Second pass: switch the tsk->fs->* elements and mark new vfsmounts | |
2359 | * as belonging to new namespace. We have already acquired a private | |
2360 | * fs_struct, so tsk->fs->lock is not needed. | |
2361 | */ | |
6b3286ed | 2362 | p = mnt_ns->root; |
1da177e4 LT |
2363 | q = new_ns->root; |
2364 | while (p) { | |
6b3286ed | 2365 | q->mnt_ns = new_ns; |
7e3d0eb0 | 2366 | __mnt_make_longterm(q); |
1da177e4 | 2367 | if (fs) { |
6ac08c39 | 2368 | if (p == fs->root.mnt) { |
f03c6599 | 2369 | fs->root.mnt = mntget(q); |
7e3d0eb0 | 2370 | __mnt_make_longterm(q); |
f03c6599 | 2371 | mnt_make_shortterm(p); |
1da177e4 | 2372 | rootmnt = p; |
1da177e4 | 2373 | } |
6ac08c39 | 2374 | if (p == fs->pwd.mnt) { |
f03c6599 | 2375 | fs->pwd.mnt = mntget(q); |
7e3d0eb0 | 2376 | __mnt_make_longterm(q); |
f03c6599 | 2377 | mnt_make_shortterm(p); |
1da177e4 | 2378 | pwdmnt = p; |
1da177e4 | 2379 | } |
1da177e4 | 2380 | } |
6b3286ed | 2381 | p = next_mnt(p, mnt_ns->root); |
1da177e4 LT |
2382 | q = next_mnt(q, new_ns->root); |
2383 | } | |
390c6843 | 2384 | up_write(&namespace_sem); |
1da177e4 | 2385 | |
1da177e4 | 2386 | if (rootmnt) |
f03c6599 | 2387 | mntput(rootmnt); |
1da177e4 | 2388 | if (pwdmnt) |
f03c6599 | 2389 | mntput(pwdmnt); |
1da177e4 | 2390 | |
741a2951 JD |
2391 | return new_ns; |
2392 | } | |
2393 | ||
213dd266 | 2394 | struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, |
e3222c4e | 2395 | struct fs_struct *new_fs) |
741a2951 | 2396 | { |
6b3286ed | 2397 | struct mnt_namespace *new_ns; |
741a2951 | 2398 | |
e3222c4e | 2399 | BUG_ON(!ns); |
6b3286ed | 2400 | get_mnt_ns(ns); |
741a2951 JD |
2401 | |
2402 | if (!(flags & CLONE_NEWNS)) | |
e3222c4e | 2403 | return ns; |
741a2951 | 2404 | |
e3222c4e | 2405 | new_ns = dup_mnt_ns(ns, new_fs); |
741a2951 | 2406 | |
6b3286ed | 2407 | put_mnt_ns(ns); |
e3222c4e | 2408 | return new_ns; |
1da177e4 LT |
2409 | } |
2410 | ||
cf8d2c11 TM |
2411 | /** |
2412 | * create_mnt_ns - creates a private namespace and adds a root filesystem | |
2413 | * @mnt: pointer to the new root filesystem mountpoint | |
2414 | */ | |
a2770d86 | 2415 | struct mnt_namespace *create_mnt_ns(struct vfsmount *mnt) |
cf8d2c11 TM |
2416 | { |
2417 | struct mnt_namespace *new_ns; | |
2418 | ||
2419 | new_ns = alloc_mnt_ns(); | |
2420 | if (!IS_ERR(new_ns)) { | |
2421 | mnt->mnt_ns = new_ns; | |
7e3d0eb0 | 2422 | __mnt_make_longterm(mnt); |
cf8d2c11 TM |
2423 | new_ns->root = mnt; |
2424 | list_add(&new_ns->list, &new_ns->root->mnt_list); | |
2425 | } | |
2426 | return new_ns; | |
2427 | } | |
a2770d86 | 2428 | EXPORT_SYMBOL(create_mnt_ns); |
cf8d2c11 | 2429 | |
bdc480e3 HC |
2430 | SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, |
2431 | char __user *, type, unsigned long, flags, void __user *, data) | |
1da177e4 | 2432 | { |
eca6f534 VN |
2433 | int ret; |
2434 | char *kernel_type; | |
2435 | char *kernel_dir; | |
2436 | char *kernel_dev; | |
1da177e4 | 2437 | unsigned long data_page; |
1da177e4 | 2438 | |
eca6f534 VN |
2439 | ret = copy_mount_string(type, &kernel_type); |
2440 | if (ret < 0) | |
2441 | goto out_type; | |
1da177e4 | 2442 | |
eca6f534 VN |
2443 | kernel_dir = getname(dir_name); |
2444 | if (IS_ERR(kernel_dir)) { | |
2445 | ret = PTR_ERR(kernel_dir); | |
2446 | goto out_dir; | |
2447 | } | |
1da177e4 | 2448 | |
eca6f534 VN |
2449 | ret = copy_mount_string(dev_name, &kernel_dev); |
2450 | if (ret < 0) | |
2451 | goto out_dev; | |
1da177e4 | 2452 | |
eca6f534 VN |
2453 | ret = copy_mount_options(data, &data_page); |
2454 | if (ret < 0) | |
2455 | goto out_data; | |
1da177e4 | 2456 | |
eca6f534 VN |
2457 | ret = do_mount(kernel_dev, kernel_dir, kernel_type, flags, |
2458 | (void *) data_page); | |
1da177e4 | 2459 | |
eca6f534 VN |
2460 | free_page(data_page); |
2461 | out_data: | |
2462 | kfree(kernel_dev); | |
2463 | out_dev: | |
2464 | putname(kernel_dir); | |
2465 | out_dir: | |
2466 | kfree(kernel_type); | |
2467 | out_type: | |
2468 | return ret; | |
1da177e4 LT |
2469 | } |
2470 | ||
1da177e4 LT |
2471 | /* |
2472 | * pivot_root Semantics: | |
2473 | * Moves the root file system of the current process to the directory put_old, | |
2474 | * makes new_root as the new root file system of the current process, and sets | |
2475 | * root/cwd of all processes which had them on the current root to new_root. | |
2476 | * | |
2477 | * Restrictions: | |
2478 | * The new_root and put_old must be directories, and must not be on the | |
2479 | * same file system as the current process root. The put_old must be | |
2480 | * underneath new_root, i.e. adding a non-zero number of /.. to the string | |
2481 | * pointed to by put_old must yield the same directory as new_root. No other | |
2482 | * file system may be mounted on put_old. After all, new_root is a mountpoint. | |
2483 | * | |
4a0d11fa NB |
2484 | * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. |
2485 | * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives | |
2486 | * in this situation. | |
2487 | * | |
1da177e4 LT |
2488 | * Notes: |
2489 | * - we don't move root/cwd if they are not at the root (reason: if something | |
2490 | * cared enough to change them, it's probably wrong to force them elsewhere) | |
2491 | * - it's okay to pick a root that isn't the root of a file system, e.g. | |
2492 | * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, | |
2493 | * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root | |
2494 | * first. | |
2495 | */ | |
3480b257 HC |
2496 | SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, |
2497 | const char __user *, put_old) | |
1da177e4 LT |
2498 | { |
2499 | struct vfsmount *tmp; | |
2d8f3038 | 2500 | struct path new, old, parent_path, root_parent, root; |
1da177e4 LT |
2501 | int error; |
2502 | ||
2503 | if (!capable(CAP_SYS_ADMIN)) | |
2504 | return -EPERM; | |
2505 | ||
2d8f3038 | 2506 | error = user_path_dir(new_root, &new); |
1da177e4 LT |
2507 | if (error) |
2508 | goto out0; | |
2509 | error = -EINVAL; | |
2d8f3038 | 2510 | if (!check_mnt(new.mnt)) |
1da177e4 LT |
2511 | goto out1; |
2512 | ||
2d8f3038 | 2513 | error = user_path_dir(put_old, &old); |
1da177e4 LT |
2514 | if (error) |
2515 | goto out1; | |
2516 | ||
2d8f3038 | 2517 | error = security_sb_pivotroot(&old, &new); |
1da177e4 | 2518 | if (error) { |
2d8f3038 | 2519 | path_put(&old); |
1da177e4 LT |
2520 | goto out1; |
2521 | } | |
2522 | ||
f7ad3c6b | 2523 | get_fs_root(current->fs, &root); |
390c6843 | 2524 | down_write(&namespace_sem); |
2d8f3038 | 2525 | mutex_lock(&old.dentry->d_inode->i_mutex); |
1da177e4 | 2526 | error = -EINVAL; |
2d8f3038 AV |
2527 | if (IS_MNT_SHARED(old.mnt) || |
2528 | IS_MNT_SHARED(new.mnt->mnt_parent) || | |
8c3ee42e | 2529 | IS_MNT_SHARED(root.mnt->mnt_parent)) |
21444403 | 2530 | goto out2; |
8c3ee42e | 2531 | if (!check_mnt(root.mnt)) |
1da177e4 LT |
2532 | goto out2; |
2533 | error = -ENOENT; | |
d83c49f3 | 2534 | if (cant_mount(old.dentry)) |
1da177e4 | 2535 | goto out2; |
f3da392e | 2536 | if (d_unlinked(new.dentry)) |
1da177e4 | 2537 | goto out2; |
f3da392e | 2538 | if (d_unlinked(old.dentry)) |
1da177e4 LT |
2539 | goto out2; |
2540 | error = -EBUSY; | |
2d8f3038 AV |
2541 | if (new.mnt == root.mnt || |
2542 | old.mnt == root.mnt) | |
1da177e4 LT |
2543 | goto out2; /* loop, on the same file system */ |
2544 | error = -EINVAL; | |
8c3ee42e | 2545 | if (root.mnt->mnt_root != root.dentry) |
1da177e4 | 2546 | goto out2; /* not a mountpoint */ |
8c3ee42e | 2547 | if (root.mnt->mnt_parent == root.mnt) |
0bb6fcc1 | 2548 | goto out2; /* not attached */ |
2d8f3038 | 2549 | if (new.mnt->mnt_root != new.dentry) |
1da177e4 | 2550 | goto out2; /* not a mountpoint */ |
2d8f3038 | 2551 | if (new.mnt->mnt_parent == new.mnt) |
0bb6fcc1 | 2552 | goto out2; /* not attached */ |
4ac91378 | 2553 | /* make sure we can reach put_old from new_root */ |
2d8f3038 | 2554 | tmp = old.mnt; |
99b7db7b | 2555 | br_write_lock(vfsmount_lock); |
2d8f3038 | 2556 | if (tmp != new.mnt) { |
1da177e4 LT |
2557 | for (;;) { |
2558 | if (tmp->mnt_parent == tmp) | |
2559 | goto out3; /* already mounted on put_old */ | |
2d8f3038 | 2560 | if (tmp->mnt_parent == new.mnt) |
1da177e4 LT |
2561 | break; |
2562 | tmp = tmp->mnt_parent; | |
2563 | } | |
2d8f3038 | 2564 | if (!is_subdir(tmp->mnt_mountpoint, new.dentry)) |
1da177e4 | 2565 | goto out3; |
2d8f3038 | 2566 | } else if (!is_subdir(old.dentry, new.dentry)) |
1da177e4 | 2567 | goto out3; |
2d8f3038 | 2568 | detach_mnt(new.mnt, &parent_path); |
8c3ee42e | 2569 | detach_mnt(root.mnt, &root_parent); |
4ac91378 | 2570 | /* mount old root on put_old */ |
2d8f3038 | 2571 | attach_mnt(root.mnt, &old); |
4ac91378 | 2572 | /* mount new_root on / */ |
2d8f3038 | 2573 | attach_mnt(new.mnt, &root_parent); |
6b3286ed | 2574 | touch_mnt_namespace(current->nsproxy->mnt_ns); |
99b7db7b | 2575 | br_write_unlock(vfsmount_lock); |
2d8f3038 | 2576 | chroot_fs_refs(&root, &new); |
b3e19d92 | 2577 | |
1da177e4 | 2578 | error = 0; |
1a390689 AV |
2579 | path_put(&root_parent); |
2580 | path_put(&parent_path); | |
1da177e4 | 2581 | out2: |
2d8f3038 | 2582 | mutex_unlock(&old.dentry->d_inode->i_mutex); |
390c6843 | 2583 | up_write(&namespace_sem); |
8c3ee42e | 2584 | path_put(&root); |
2d8f3038 | 2585 | path_put(&old); |
1da177e4 | 2586 | out1: |
2d8f3038 | 2587 | path_put(&new); |
1da177e4 | 2588 | out0: |
1da177e4 LT |
2589 | return error; |
2590 | out3: | |
99b7db7b | 2591 | br_write_unlock(vfsmount_lock); |
1da177e4 LT |
2592 | goto out2; |
2593 | } | |
2594 | ||
2595 | static void __init init_mount_tree(void) | |
2596 | { | |
2597 | struct vfsmount *mnt; | |
6b3286ed | 2598 | struct mnt_namespace *ns; |
ac748a09 | 2599 | struct path root; |
1da177e4 LT |
2600 | |
2601 | mnt = do_kern_mount("rootfs", 0, "rootfs", NULL); | |
2602 | if (IS_ERR(mnt)) | |
2603 | panic("Can't create rootfs"); | |
b3e19d92 | 2604 | |
3b22edc5 TM |
2605 | ns = create_mnt_ns(mnt); |
2606 | if (IS_ERR(ns)) | |
1da177e4 | 2607 | panic("Can't allocate initial namespace"); |
6b3286ed KK |
2608 | |
2609 | init_task.nsproxy->mnt_ns = ns; | |
2610 | get_mnt_ns(ns); | |
2611 | ||
ac748a09 JB |
2612 | root.mnt = ns->root; |
2613 | root.dentry = ns->root->mnt_root; | |
2614 | ||
2615 | set_fs_pwd(current->fs, &root); | |
2616 | set_fs_root(current->fs, &root); | |
1da177e4 LT |
2617 | } |
2618 | ||
74bf17cf | 2619 | void __init mnt_init(void) |
1da177e4 | 2620 | { |
13f14b4d | 2621 | unsigned u; |
15a67dd8 | 2622 | int err; |
1da177e4 | 2623 | |
390c6843 RP |
2624 | init_rwsem(&namespace_sem); |
2625 | ||
1da177e4 | 2626 | mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct vfsmount), |
20c2df83 | 2627 | 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); |
1da177e4 | 2628 | |
b58fed8b | 2629 | mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
1da177e4 LT |
2630 | |
2631 | if (!mount_hashtable) | |
2632 | panic("Failed to allocate mount hash table\n"); | |
2633 | ||
13f14b4d ED |
2634 | printk("Mount-cache hash table entries: %lu\n", HASH_SIZE); |
2635 | ||
2636 | for (u = 0; u < HASH_SIZE; u++) | |
2637 | INIT_LIST_HEAD(&mount_hashtable[u]); | |
1da177e4 | 2638 | |
99b7db7b NP |
2639 | br_lock_init(vfsmount_lock); |
2640 | ||
15a67dd8 RD |
2641 | err = sysfs_init(); |
2642 | if (err) | |
2643 | printk(KERN_WARNING "%s: sysfs_init error: %d\n", | |
8e24eea7 | 2644 | __func__, err); |
00d26666 GKH |
2645 | fs_kobj = kobject_create_and_add("fs", NULL); |
2646 | if (!fs_kobj) | |
8e24eea7 | 2647 | printk(KERN_WARNING "%s: kobj create error\n", __func__); |
1da177e4 LT |
2648 | init_rootfs(); |
2649 | init_mount_tree(); | |
2650 | } | |
2651 | ||
616511d0 | 2652 | void put_mnt_ns(struct mnt_namespace *ns) |
1da177e4 | 2653 | { |
70fbcdf4 | 2654 | LIST_HEAD(umount_list); |
616511d0 | 2655 | |
d498b25a | 2656 | if (!atomic_dec_and_test(&ns->count)) |
616511d0 | 2657 | return; |
390c6843 | 2658 | down_write(&namespace_sem); |
99b7db7b | 2659 | br_write_lock(vfsmount_lock); |
d498b25a | 2660 | umount_tree(ns->root, 0, &umount_list); |
99b7db7b | 2661 | br_write_unlock(vfsmount_lock); |
390c6843 | 2662 | up_write(&namespace_sem); |
70fbcdf4 | 2663 | release_mounts(&umount_list); |
6b3286ed | 2664 | kfree(ns); |
1da177e4 | 2665 | } |
cf8d2c11 | 2666 | EXPORT_SYMBOL(put_mnt_ns); |