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 | 11 | #include <linux/syscalls.h> |
d10577a8 | 12 | #include <linux/export.h> |
16f7e0fe | 13 | #include <linux/capability.h> |
6b3286ed | 14 | #include <linux/mnt_namespace.h> |
771b1371 | 15 | #include <linux/user_namespace.h> |
1da177e4 LT |
16 | #include <linux/namei.h> |
17 | #include <linux/security.h> | |
73cd49ec | 18 | #include <linux/idr.h> |
d10577a8 AV |
19 | #include <linux/acct.h> /* acct_auto_close_mnt */ |
20 | #include <linux/ramfs.h> /* init_rootfs */ | |
21 | #include <linux/fs_struct.h> /* get_fs_root et.al. */ | |
22 | #include <linux/fsnotify.h> /* fsnotify_vfsmount_delete */ | |
23 | #include <linux/uaccess.h> | |
6fa3eb70 S |
24 | //#define UMOUNT_LOG //enable kernel layer unmount log when unmount fail |
25 | #ifdef UMOUNT_LOG | |
26 | #include <linux/types.h> | |
27 | #endif | |
0bb80f24 | 28 | #include <linux/proc_ns.h> |
20b4fb48 | 29 | #include <linux/magic.h> |
07b20889 | 30 | #include "pnode.h" |
948730b0 | 31 | #include "internal.h" |
1da177e4 | 32 | |
13f14b4d ED |
33 | #define HASH_SHIFT ilog2(PAGE_SIZE / sizeof(struct list_head)) |
34 | #define HASH_SIZE (1UL << HASH_SHIFT) | |
35 | ||
5addc5dd | 36 | static int event; |
73cd49ec | 37 | static DEFINE_IDA(mnt_id_ida); |
719f5d7f | 38 | static DEFINE_IDA(mnt_group_ida); |
99b7db7b | 39 | static DEFINE_SPINLOCK(mnt_id_lock); |
f21f6220 AV |
40 | static int mnt_id_start = 0; |
41 | static int mnt_group_start = 1; | |
1da177e4 | 42 | |
fa3536cc | 43 | static struct list_head *mount_hashtable __read_mostly; |
84d17192 | 44 | static struct list_head *mountpoint_hashtable __read_mostly; |
e18b890b | 45 | static struct kmem_cache *mnt_cache __read_mostly; |
390c6843 | 46 | static struct rw_semaphore namespace_sem; |
1da177e4 | 47 | |
f87fd4c2 | 48 | /* /sys/fs */ |
00d26666 GKH |
49 | struct kobject *fs_kobj; |
50 | EXPORT_SYMBOL_GPL(fs_kobj); | |
f87fd4c2 | 51 | |
99b7db7b NP |
52 | /* |
53 | * vfsmount lock may be taken for read to prevent changes to the | |
54 | * vfsmount hash, ie. during mountpoint lookups or walking back | |
55 | * up the tree. | |
56 | * | |
57 | * It should be taken for write in all cases where the vfsmount | |
58 | * tree or hash is modified or when a vfsmount structure is modified. | |
59 | */ | |
60 | DEFINE_BRLOCK(vfsmount_lock); | |
61 | ||
1da177e4 LT |
62 | static inline unsigned long hash(struct vfsmount *mnt, struct dentry *dentry) |
63 | { | |
b58fed8b RP |
64 | unsigned long tmp = ((unsigned long)mnt / L1_CACHE_BYTES); |
65 | tmp += ((unsigned long)dentry / L1_CACHE_BYTES); | |
13f14b4d ED |
66 | tmp = tmp + (tmp >> HASH_SHIFT); |
67 | return tmp & (HASH_SIZE - 1); | |
1da177e4 LT |
68 | } |
69 | ||
3d733633 DH |
70 | #define MNT_WRITER_UNDERFLOW_LIMIT -(1<<16) |
71 | ||
99b7db7b NP |
72 | /* |
73 | * allocation is serialized by namespace_sem, but we need the spinlock to | |
74 | * serialize with freeing. | |
75 | */ | |
b105e270 | 76 | static int mnt_alloc_id(struct mount *mnt) |
73cd49ec MS |
77 | { |
78 | int res; | |
79 | ||
80 | retry: | |
81 | ida_pre_get(&mnt_id_ida, GFP_KERNEL); | |
99b7db7b | 82 | spin_lock(&mnt_id_lock); |
15169fe7 | 83 | res = ida_get_new_above(&mnt_id_ida, mnt_id_start, &mnt->mnt_id); |
f21f6220 | 84 | if (!res) |
15169fe7 | 85 | mnt_id_start = mnt->mnt_id + 1; |
99b7db7b | 86 | spin_unlock(&mnt_id_lock); |
73cd49ec MS |
87 | if (res == -EAGAIN) |
88 | goto retry; | |
89 | ||
90 | return res; | |
91 | } | |
92 | ||
b105e270 | 93 | static void mnt_free_id(struct mount *mnt) |
73cd49ec | 94 | { |
15169fe7 | 95 | int id = mnt->mnt_id; |
99b7db7b | 96 | spin_lock(&mnt_id_lock); |
f21f6220 AV |
97 | ida_remove(&mnt_id_ida, id); |
98 | if (mnt_id_start > id) | |
99 | mnt_id_start = id; | |
99b7db7b | 100 | spin_unlock(&mnt_id_lock); |
73cd49ec MS |
101 | } |
102 | ||
719f5d7f MS |
103 | /* |
104 | * Allocate a new peer group ID | |
105 | * | |
106 | * mnt_group_ida is protected by namespace_sem | |
107 | */ | |
4b8b21f4 | 108 | static int mnt_alloc_group_id(struct mount *mnt) |
719f5d7f | 109 | { |
f21f6220 AV |
110 | int res; |
111 | ||
719f5d7f MS |
112 | if (!ida_pre_get(&mnt_group_ida, GFP_KERNEL)) |
113 | return -ENOMEM; | |
114 | ||
f21f6220 AV |
115 | res = ida_get_new_above(&mnt_group_ida, |
116 | mnt_group_start, | |
15169fe7 | 117 | &mnt->mnt_group_id); |
f21f6220 | 118 | if (!res) |
15169fe7 | 119 | mnt_group_start = mnt->mnt_group_id + 1; |
f21f6220 AV |
120 | |
121 | return res; | |
719f5d7f MS |
122 | } |
123 | ||
124 | /* | |
125 | * Release a peer group ID | |
126 | */ | |
4b8b21f4 | 127 | void mnt_release_group_id(struct mount *mnt) |
719f5d7f | 128 | { |
15169fe7 | 129 | int id = mnt->mnt_group_id; |
f21f6220 AV |
130 | ida_remove(&mnt_group_ida, id); |
131 | if (mnt_group_start > id) | |
132 | mnt_group_start = id; | |
15169fe7 | 133 | mnt->mnt_group_id = 0; |
719f5d7f MS |
134 | } |
135 | ||
b3e19d92 NP |
136 | /* |
137 | * vfsmount lock must be held for read | |
138 | */ | |
6fa3eb70 S |
139 | #ifdef UMOUNT_LOG |
140 | #define UMOUNT_Partition "/emmc@usrdata" | |
141 | struct record_ref_count{ | |
142 | pid_t pid; | |
143 | char name[TASK_COMM_LEN]; | |
144 | int count; | |
145 | struct record_ref_count *next; | |
146 | }; | |
147 | ||
148 | struct record_ref_count *ref_head = NULL; | |
149 | struct record_ref_count *ref_current = NULL; | |
150 | struct record_ref_count *ref_prev = NULL; | |
151 | int s_total_count = 0; | |
152 | #endif | |
83adc753 | 153 | static inline void mnt_add_count(struct mount *mnt, int n) |
b3e19d92 | 154 | { |
6fa3eb70 S |
155 | #ifdef UMOUNT_LOG |
156 | int print_link_list=0; | |
157 | #ifndef CONFIG_SMP | |
158 | preempt_disable(); | |
159 | #endif | |
160 | ||
161 | if (strcmp(UMOUNT_Partition,mnt->mnt_devname)==0) | |
162 | { | |
163 | //if (strcmp("mobile_log_d",current->comm)!=0) | |
164 | { | |
165 | //if (current->pid < 100) //((current->pid < 70) && (current->pid > 60)) | |
166 | { | |
167 | //printk("Ahsin n=%d current->pid=%d name=%s \n",n,current->pid,current->comm); | |
168 | spin_lock(&mnt_id_lock); | |
169 | if (ref_head == NULL) //linked list head (start) | |
170 | { | |
171 | printk("Ahsin link list init mnt_get_count=%d \n",mnt_get_count(mnt)); | |
172 | ||
173 | ref_current = kmalloc(sizeof(struct record_ref_count), GFP_KERNEL); | |
174 | if (ref_current == NULL) | |
175 | printk("Ahsin can't allocate memory for ref_current /n"); | |
176 | ||
177 | ref_current->next = NULL; | |
178 | ref_current->pid = current->pid; | |
179 | strncpy(ref_current->name, current->comm, TASK_COMM_LEN -1); | |
180 | ref_current->name[TASK_COMM_LEN -1] = '\0'; | |
181 | ref_current->count = n; | |
182 | s_total_count = s_total_count + n; | |
183 | ref_head = ref_current; | |
184 | ||
185 | printk("Ahsin ref_head == NULL pid=%d name=%s counter=%d n=%d \n",ref_current->pid,ref_current->name,ref_current->count,n); | |
186 | } | |
187 | else //check exist first and then add linked list or modify counter | |
188 | { | |
189 | ref_prev = ref_head; | |
190 | while(ref_prev != NULL) | |
191 | { | |
192 | //printk("Ahsin PID= %d, Name= %s, Count= %d n=%d current->pid=%d \n", ref_prev->pid, ref_prev->name, ref_prev->count,n,current->pid); | |
193 | if (strcmp(ref_prev->name,current->comm)==0) //(ref_prev->pid==current->pid)//exist and find, modify counter | |
194 | { | |
195 | ref_prev->count = ref_prev->count + n; | |
196 | s_total_count = s_total_count + n; | |
197 | //printk("Ahsin (ref_prev->name,current->comm) pid=%d name=%s counter=%d n=%d \n",ref_prev->pid,ref_prev->name,ref_prev->count,n); | |
198 | break; | |
199 | } | |
200 | else | |
201 | { | |
202 | if (ref_prev->next != NULL) | |
203 | ref_prev = ref_prev->next; | |
204 | else | |
205 | { // end of link list | |
206 | ref_current = kmalloc(sizeof(struct record_ref_count), GFP_KERNEL); | |
207 | if (ref_current == NULL) | |
208 | printk("Ahsin can't allocate memory for ref_prev /n"); | |
209 | ||
210 | ref_current->next = NULL; | |
211 | ref_current->pid = current->pid; | |
212 | strncpy(ref_current->name, current->comm, TASK_COMM_LEN -1); | |
213 | ref_current->name[TASK_COMM_LEN -1] = '\0'; | |
214 | ref_current->count = n; | |
215 | s_total_count = s_total_count + n; | |
216 | ref_prev->next = ref_current; | |
217 | //printk("Ahsin new node(end of link list) pid=%d name=%s counter=%d n=%d \n",ref_current->pid,ref_current->name,ref_current->count,n); | |
218 | break; | |
219 | ||
220 | } | |
221 | } | |
222 | } | |
223 | } | |
224 | spin_unlock(&mnt_id_lock); | |
225 | } | |
226 | } | |
227 | } | |
228 | ||
229 | #ifndef CONFIG_SMP | |
230 | preempt_enable(); | |
231 | #endif | |
232 | #endif | |
b3e19d92 | 233 | #ifdef CONFIG_SMP |
68e8a9fe | 234 | this_cpu_add(mnt->mnt_pcp->mnt_count, n); |
6fa3eb70 S |
235 | #ifdef UMOUNT_LOG |
236 | #if 0 | |
237 | if (strcmp(UMOUNT_Partition,mnt->mnt_devname)==0) | |
238 | { | |
239 | if (strcmp("mobile_log_d",current->comm)!=0) | |
240 | { | |
241 | printk("Ahsin s_total_count=%d mnt_get_count=%d n=%d current->pid=%d \n",s_total_count,mnt_get_count(mnt),n,current->pid); | |
242 | //if (current->pid < 100) | |
243 | { | |
244 | // print linked list | |
245 | spin_lock(&mnt_id_lock); | |
246 | ref_current = ref_head; | |
247 | while(ref_current != NULL) | |
248 | { | |
249 | if (ref_current->count) | |
250 | { | |
251 | print_link_list = print_link_list + ref_current->count; | |
252 | //printk("Ahsin PID= %d, Name = %s, Count= %d \n", ref_current->pid, ref_current->name, ref_current->count); | |
253 | } | |
254 | ref_current = ref_current->next; | |
255 | } | |
256 | spin_unlock(&mnt_id_lock); | |
257 | printk("Ahsin print_link_list=%d \n",print_link_list); | |
258 | } | |
259 | } | |
260 | } | |
261 | #endif | |
262 | #endif | |
b3e19d92 NP |
263 | #else |
264 | preempt_disable(); | |
68e8a9fe | 265 | mnt->mnt_count += n; |
b3e19d92 NP |
266 | preempt_enable(); |
267 | #endif | |
268 | } | |
269 | ||
b3e19d92 NP |
270 | /* |
271 | * vfsmount lock must be held for write | |
272 | */ | |
83adc753 | 273 | unsigned int mnt_get_count(struct mount *mnt) |
b3e19d92 NP |
274 | { |
275 | #ifdef CONFIG_SMP | |
f03c6599 | 276 | unsigned int count = 0; |
b3e19d92 NP |
277 | int cpu; |
278 | ||
279 | for_each_possible_cpu(cpu) { | |
68e8a9fe | 280 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_count; |
b3e19d92 NP |
281 | } |
282 | ||
283 | return count; | |
284 | #else | |
68e8a9fe | 285 | return mnt->mnt_count; |
b3e19d92 NP |
286 | #endif |
287 | } | |
288 | ||
b105e270 | 289 | static struct mount *alloc_vfsmnt(const char *name) |
1da177e4 | 290 | { |
c63181e6 AV |
291 | struct mount *mnt = kmem_cache_zalloc(mnt_cache, GFP_KERNEL); |
292 | if (mnt) { | |
73cd49ec MS |
293 | int err; |
294 | ||
c63181e6 | 295 | err = mnt_alloc_id(mnt); |
88b38782 LZ |
296 | if (err) |
297 | goto out_free_cache; | |
298 | ||
299 | if (name) { | |
c63181e6 AV |
300 | mnt->mnt_devname = kstrdup(name, GFP_KERNEL); |
301 | if (!mnt->mnt_devname) | |
88b38782 | 302 | goto out_free_id; |
73cd49ec MS |
303 | } |
304 | ||
b3e19d92 | 305 | #ifdef CONFIG_SMP |
c63181e6 AV |
306 | mnt->mnt_pcp = alloc_percpu(struct mnt_pcp); |
307 | if (!mnt->mnt_pcp) | |
b3e19d92 NP |
308 | goto out_free_devname; |
309 | ||
6fa3eb70 S |
310 | #ifdef UMOUNT_LOG |
311 | if (strcmp(UMOUNT_Partition,mnt->mnt_devname)==0) | |
312 | { | |
313 | printk("Ahsin alloc_vfsmnt current->pid=%d name=%s \n",current->pid,current->comm); | |
314 | } | |
315 | #endif | |
c63181e6 | 316 | this_cpu_add(mnt->mnt_pcp->mnt_count, 1); |
b3e19d92 | 317 | #else |
c63181e6 AV |
318 | mnt->mnt_count = 1; |
319 | mnt->mnt_writers = 0; | |
b3e19d92 NP |
320 | #endif |
321 | ||
c63181e6 AV |
322 | INIT_LIST_HEAD(&mnt->mnt_hash); |
323 | INIT_LIST_HEAD(&mnt->mnt_child); | |
324 | INIT_LIST_HEAD(&mnt->mnt_mounts); | |
325 | INIT_LIST_HEAD(&mnt->mnt_list); | |
326 | INIT_LIST_HEAD(&mnt->mnt_expire); | |
327 | INIT_LIST_HEAD(&mnt->mnt_share); | |
328 | INIT_LIST_HEAD(&mnt->mnt_slave_list); | |
329 | INIT_LIST_HEAD(&mnt->mnt_slave); | |
2504c5d6 AG |
330 | #ifdef CONFIG_FSNOTIFY |
331 | INIT_HLIST_HEAD(&mnt->mnt_fsnotify_marks); | |
d3ef3d73 | 332 | #endif |
1da177e4 | 333 | } |
c63181e6 | 334 | return mnt; |
88b38782 | 335 | |
d3ef3d73 | 336 | #ifdef CONFIG_SMP |
337 | out_free_devname: | |
c63181e6 | 338 | kfree(mnt->mnt_devname); |
d3ef3d73 | 339 | #endif |
88b38782 | 340 | out_free_id: |
c63181e6 | 341 | mnt_free_id(mnt); |
88b38782 | 342 | out_free_cache: |
c63181e6 | 343 | kmem_cache_free(mnt_cache, mnt); |
88b38782 | 344 | return NULL; |
1da177e4 LT |
345 | } |
346 | ||
3d733633 DH |
347 | /* |
348 | * Most r/o checks on a fs are for operations that take | |
349 | * discrete amounts of time, like a write() or unlink(). | |
350 | * We must keep track of when those operations start | |
351 | * (for permission checks) and when they end, so that | |
352 | * we can determine when writes are able to occur to | |
353 | * a filesystem. | |
354 | */ | |
355 | /* | |
356 | * __mnt_is_readonly: check whether a mount is read-only | |
357 | * @mnt: the mount to check for its write status | |
358 | * | |
359 | * This shouldn't be used directly ouside of the VFS. | |
360 | * It does not guarantee that the filesystem will stay | |
361 | * r/w, just that it is right *now*. This can not and | |
362 | * should not be used in place of IS_RDONLY(inode). | |
363 | * mnt_want/drop_write() will _keep_ the filesystem | |
364 | * r/w. | |
365 | */ | |
366 | int __mnt_is_readonly(struct vfsmount *mnt) | |
367 | { | |
2e4b7fcd DH |
368 | if (mnt->mnt_flags & MNT_READONLY) |
369 | return 1; | |
370 | if (mnt->mnt_sb->s_flags & MS_RDONLY) | |
371 | return 1; | |
372 | return 0; | |
3d733633 DH |
373 | } |
374 | EXPORT_SYMBOL_GPL(__mnt_is_readonly); | |
375 | ||
83adc753 | 376 | static inline void mnt_inc_writers(struct mount *mnt) |
d3ef3d73 | 377 | { |
378 | #ifdef CONFIG_SMP | |
68e8a9fe | 379 | this_cpu_inc(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 | 380 | #else |
68e8a9fe | 381 | mnt->mnt_writers++; |
d3ef3d73 | 382 | #endif |
383 | } | |
3d733633 | 384 | |
83adc753 | 385 | static inline void mnt_dec_writers(struct mount *mnt) |
3d733633 | 386 | { |
d3ef3d73 | 387 | #ifdef CONFIG_SMP |
68e8a9fe | 388 | this_cpu_dec(mnt->mnt_pcp->mnt_writers); |
d3ef3d73 | 389 | #else |
68e8a9fe | 390 | mnt->mnt_writers--; |
d3ef3d73 | 391 | #endif |
3d733633 | 392 | } |
3d733633 | 393 | |
83adc753 | 394 | static unsigned int mnt_get_writers(struct mount *mnt) |
3d733633 | 395 | { |
d3ef3d73 | 396 | #ifdef CONFIG_SMP |
397 | unsigned int count = 0; | |
3d733633 | 398 | int cpu; |
3d733633 DH |
399 | |
400 | for_each_possible_cpu(cpu) { | |
68e8a9fe | 401 | count += per_cpu_ptr(mnt->mnt_pcp, cpu)->mnt_writers; |
3d733633 | 402 | } |
3d733633 | 403 | |
d3ef3d73 | 404 | return count; |
405 | #else | |
406 | return mnt->mnt_writers; | |
407 | #endif | |
3d733633 DH |
408 | } |
409 | ||
4ed5e82f MS |
410 | static int mnt_is_readonly(struct vfsmount *mnt) |
411 | { | |
412 | if (mnt->mnt_sb->s_readonly_remount) | |
413 | return 1; | |
414 | /* Order wrt setting s_flags/s_readonly_remount in do_remount() */ | |
415 | smp_rmb(); | |
416 | return __mnt_is_readonly(mnt); | |
417 | } | |
418 | ||
8366025e | 419 | /* |
eb04c282 JK |
420 | * Most r/o & frozen checks on a fs are for operations that take discrete |
421 | * amounts of time, like a write() or unlink(). We must keep track of when | |
422 | * those operations start (for permission checks) and when they end, so that we | |
423 | * can determine when writes are able to occur to a filesystem. | |
8366025e DH |
424 | */ |
425 | /** | |
eb04c282 | 426 | * __mnt_want_write - get write access to a mount without freeze protection |
83adc753 | 427 | * @m: the mount on which to take a write |
8366025e | 428 | * |
eb04c282 JK |
429 | * This tells the low-level filesystem that a write is about to be performed to |
430 | * it, and makes sure that writes are allowed (mnt it read-write) before | |
431 | * returning success. This operation does not protect against filesystem being | |
432 | * frozen. When the write operation is finished, __mnt_drop_write() must be | |
433 | * called. This is effectively a refcount. | |
8366025e | 434 | */ |
eb04c282 | 435 | int __mnt_want_write(struct vfsmount *m) |
8366025e | 436 | { |
83adc753 | 437 | struct mount *mnt = real_mount(m); |
3d733633 | 438 | int ret = 0; |
3d733633 | 439 | |
d3ef3d73 | 440 | preempt_disable(); |
c6653a83 | 441 | mnt_inc_writers(mnt); |
d3ef3d73 | 442 | /* |
c6653a83 | 443 | * The store to mnt_inc_writers must be visible before we pass |
d3ef3d73 | 444 | * MNT_WRITE_HOLD loop below, so that the slowpath can see our |
445 | * incremented count after it has set MNT_WRITE_HOLD. | |
446 | */ | |
447 | smp_mb(); | |
1e75529e | 448 | while (ACCESS_ONCE(mnt->mnt.mnt_flags) & MNT_WRITE_HOLD) |
d3ef3d73 | 449 | cpu_relax(); |
450 | /* | |
451 | * After the slowpath clears MNT_WRITE_HOLD, mnt_is_readonly will | |
452 | * be set to match its requirements. So we must not load that until | |
453 | * MNT_WRITE_HOLD is cleared. | |
454 | */ | |
455 | smp_rmb(); | |
4ed5e82f | 456 | if (mnt_is_readonly(m)) { |
c6653a83 | 457 | mnt_dec_writers(mnt); |
3d733633 | 458 | ret = -EROFS; |
3d733633 | 459 | } |
d3ef3d73 | 460 | preempt_enable(); |
eb04c282 JK |
461 | |
462 | return ret; | |
463 | } | |
464 | ||
465 | /** | |
466 | * mnt_want_write - get write access to a mount | |
467 | * @m: the mount on which to take a write | |
468 | * | |
469 | * This tells the low-level filesystem that a write is about to be performed to | |
470 | * it, and makes sure that writes are allowed (mount is read-write, filesystem | |
471 | * is not frozen) before returning success. When the write operation is | |
472 | * finished, mnt_drop_write() must be called. This is effectively a refcount. | |
473 | */ | |
474 | int mnt_want_write(struct vfsmount *m) | |
475 | { | |
476 | int ret; | |
477 | ||
478 | sb_start_write(m->mnt_sb); | |
479 | ret = __mnt_want_write(m); | |
480 | if (ret) | |
481 | sb_end_write(m->mnt_sb); | |
3d733633 | 482 | return ret; |
8366025e DH |
483 | } |
484 | EXPORT_SYMBOL_GPL(mnt_want_write); | |
485 | ||
96029c4e | 486 | /** |
487 | * mnt_clone_write - get write access to a mount | |
488 | * @mnt: the mount on which to take a write | |
489 | * | |
490 | * This is effectively like mnt_want_write, except | |
491 | * it must only be used to take an extra write reference | |
492 | * on a mountpoint that we already know has a write reference | |
493 | * on it. This allows some optimisation. | |
494 | * | |
495 | * After finished, mnt_drop_write must be called as usual to | |
496 | * drop the reference. | |
497 | */ | |
498 | int mnt_clone_write(struct vfsmount *mnt) | |
499 | { | |
500 | /* superblock may be r/o */ | |
501 | if (__mnt_is_readonly(mnt)) | |
502 | return -EROFS; | |
503 | preempt_disable(); | |
83adc753 | 504 | mnt_inc_writers(real_mount(mnt)); |
96029c4e | 505 | preempt_enable(); |
506 | return 0; | |
507 | } | |
508 | EXPORT_SYMBOL_GPL(mnt_clone_write); | |
509 | ||
510 | /** | |
eb04c282 | 511 | * __mnt_want_write_file - get write access to a file's mount |
96029c4e | 512 | * @file: the file who's mount on which to take a write |
513 | * | |
eb04c282 | 514 | * This is like __mnt_want_write, but it takes a file and can |
96029c4e | 515 | * do some optimisations if the file is open for write already |
516 | */ | |
eb04c282 | 517 | int __mnt_want_write_file(struct file *file) |
96029c4e | 518 | { |
496ad9aa | 519 | struct inode *inode = file_inode(file); |
eb04c282 | 520 | |
2d8dd38a | 521 | if (!(file->f_mode & FMODE_WRITE) || special_file(inode->i_mode)) |
eb04c282 | 522 | return __mnt_want_write(file->f_path.mnt); |
96029c4e | 523 | else |
524 | return mnt_clone_write(file->f_path.mnt); | |
525 | } | |
eb04c282 JK |
526 | |
527 | /** | |
528 | * mnt_want_write_file - get write access to a file's mount | |
529 | * @file: the file who's mount on which to take a write | |
530 | * | |
531 | * This is like mnt_want_write, but it takes a file and can | |
532 | * do some optimisations if the file is open for write already | |
533 | */ | |
534 | int mnt_want_write_file(struct file *file) | |
535 | { | |
536 | int ret; | |
537 | ||
538 | sb_start_write(file->f_path.mnt->mnt_sb); | |
539 | ret = __mnt_want_write_file(file); | |
540 | if (ret) | |
541 | sb_end_write(file->f_path.mnt->mnt_sb); | |
542 | return ret; | |
543 | } | |
96029c4e | 544 | EXPORT_SYMBOL_GPL(mnt_want_write_file); |
545 | ||
8366025e | 546 | /** |
eb04c282 | 547 | * __mnt_drop_write - give up write access to a mount |
8366025e DH |
548 | * @mnt: the mount on which to give up write access |
549 | * | |
550 | * Tells the low-level filesystem that we are done | |
551 | * performing writes to it. Must be matched with | |
eb04c282 | 552 | * __mnt_want_write() call above. |
8366025e | 553 | */ |
eb04c282 | 554 | void __mnt_drop_write(struct vfsmount *mnt) |
8366025e | 555 | { |
d3ef3d73 | 556 | preempt_disable(); |
83adc753 | 557 | mnt_dec_writers(real_mount(mnt)); |
d3ef3d73 | 558 | preempt_enable(); |
8366025e | 559 | } |
eb04c282 JK |
560 | |
561 | /** | |
562 | * mnt_drop_write - give up write access to a mount | |
563 | * @mnt: the mount on which to give up write access | |
564 | * | |
565 | * Tells the low-level filesystem that we are done performing writes to it and | |
566 | * also allows filesystem to be frozen again. Must be matched with | |
567 | * mnt_want_write() call above. | |
568 | */ | |
569 | void mnt_drop_write(struct vfsmount *mnt) | |
570 | { | |
571 | __mnt_drop_write(mnt); | |
572 | sb_end_write(mnt->mnt_sb); | |
573 | } | |
8366025e DH |
574 | EXPORT_SYMBOL_GPL(mnt_drop_write); |
575 | ||
eb04c282 JK |
576 | void __mnt_drop_write_file(struct file *file) |
577 | { | |
578 | __mnt_drop_write(file->f_path.mnt); | |
579 | } | |
580 | ||
2a79f17e AV |
581 | void mnt_drop_write_file(struct file *file) |
582 | { | |
583 | mnt_drop_write(file->f_path.mnt); | |
584 | } | |
585 | EXPORT_SYMBOL(mnt_drop_write_file); | |
586 | ||
83adc753 | 587 | static int mnt_make_readonly(struct mount *mnt) |
8366025e | 588 | { |
3d733633 DH |
589 | int ret = 0; |
590 | ||
962830df | 591 | br_write_lock(&vfsmount_lock); |
83adc753 | 592 | mnt->mnt.mnt_flags |= MNT_WRITE_HOLD; |
3d733633 | 593 | /* |
d3ef3d73 | 594 | * After storing MNT_WRITE_HOLD, we'll read the counters. This store |
595 | * should be visible before we do. | |
3d733633 | 596 | */ |
d3ef3d73 | 597 | smp_mb(); |
598 | ||
3d733633 | 599 | /* |
d3ef3d73 | 600 | * With writers on hold, if this value is zero, then there are |
601 | * definitely no active writers (although held writers may subsequently | |
602 | * increment the count, they'll have to wait, and decrement it after | |
603 | * seeing MNT_READONLY). | |
604 | * | |
605 | * It is OK to have counter incremented on one CPU and decremented on | |
606 | * another: the sum will add up correctly. The danger would be when we | |
607 | * sum up each counter, if we read a counter before it is incremented, | |
608 | * but then read another CPU's count which it has been subsequently | |
609 | * decremented from -- we would see more decrements than we should. | |
610 | * MNT_WRITE_HOLD protects against this scenario, because | |
611 | * mnt_want_write first increments count, then smp_mb, then spins on | |
612 | * MNT_WRITE_HOLD, so it can't be decremented by another CPU while | |
613 | * we're counting up here. | |
3d733633 | 614 | */ |
c6653a83 | 615 | if (mnt_get_writers(mnt) > 0) |
d3ef3d73 | 616 | ret = -EBUSY; |
617 | else | |
83adc753 | 618 | mnt->mnt.mnt_flags |= MNT_READONLY; |
d3ef3d73 | 619 | /* |
620 | * MNT_READONLY must become visible before ~MNT_WRITE_HOLD, so writers | |
621 | * that become unheld will see MNT_READONLY. | |
622 | */ | |
623 | smp_wmb(); | |
83adc753 | 624 | mnt->mnt.mnt_flags &= ~MNT_WRITE_HOLD; |
962830df | 625 | br_write_unlock(&vfsmount_lock); |
3d733633 | 626 | return ret; |
8366025e | 627 | } |
8366025e | 628 | |
83adc753 | 629 | static void __mnt_unmake_readonly(struct mount *mnt) |
2e4b7fcd | 630 | { |
962830df | 631 | br_write_lock(&vfsmount_lock); |
83adc753 | 632 | mnt->mnt.mnt_flags &= ~MNT_READONLY; |
962830df | 633 | br_write_unlock(&vfsmount_lock); |
2e4b7fcd DH |
634 | } |
635 | ||
4ed5e82f MS |
636 | int sb_prepare_remount_readonly(struct super_block *sb) |
637 | { | |
638 | struct mount *mnt; | |
639 | int err = 0; | |
640 | ||
8e8b8796 MS |
641 | /* Racy optimization. Recheck the counter under MNT_WRITE_HOLD */ |
642 | if (atomic_long_read(&sb->s_remove_count)) | |
643 | return -EBUSY; | |
644 | ||
962830df | 645 | br_write_lock(&vfsmount_lock); |
4ed5e82f MS |
646 | list_for_each_entry(mnt, &sb->s_mounts, mnt_instance) { |
647 | if (!(mnt->mnt.mnt_flags & MNT_READONLY)) { | |
648 | mnt->mnt.mnt_flags |= MNT_WRITE_HOLD; | |
649 | smp_mb(); | |
650 | if (mnt_get_writers(mnt) > 0) { | |
651 | err = -EBUSY; | |
652 | break; | |
653 | } | |
654 | } | |
655 | } | |
8e8b8796 MS |
656 | if (!err && atomic_long_read(&sb->s_remove_count)) |
657 | err = -EBUSY; | |
658 | ||
4ed5e82f MS |
659 | if (!err) { |
660 | sb->s_readonly_remount = 1; | |
661 | smp_wmb(); | |
662 | } | |
663 | list_for_each_entry(mnt, &sb->s_mounts, mnt_instance) { | |
664 | if (mnt->mnt.mnt_flags & MNT_WRITE_HOLD) | |
665 | mnt->mnt.mnt_flags &= ~MNT_WRITE_HOLD; | |
666 | } | |
962830df | 667 | br_write_unlock(&vfsmount_lock); |
4ed5e82f MS |
668 | |
669 | return err; | |
670 | } | |
671 | ||
b105e270 | 672 | static void free_vfsmnt(struct mount *mnt) |
1da177e4 | 673 | { |
52ba1621 | 674 | kfree(mnt->mnt_devname); |
73cd49ec | 675 | mnt_free_id(mnt); |
d3ef3d73 | 676 | #ifdef CONFIG_SMP |
68e8a9fe | 677 | free_percpu(mnt->mnt_pcp); |
d3ef3d73 | 678 | #endif |
b105e270 | 679 | kmem_cache_free(mnt_cache, mnt); |
1da177e4 LT |
680 | } |
681 | ||
682 | /* | |
a05964f3 RP |
683 | * find the first or last mount at @dentry on vfsmount @mnt depending on |
684 | * @dir. If @dir is set return the first mount else return the last mount. | |
99b7db7b | 685 | * vfsmount_lock must be held for read or write. |
1da177e4 | 686 | */ |
c7105365 | 687 | struct mount *__lookup_mnt(struct vfsmount *mnt, struct dentry *dentry, |
a05964f3 | 688 | int dir) |
1da177e4 | 689 | { |
b58fed8b RP |
690 | struct list_head *head = mount_hashtable + hash(mnt, dentry); |
691 | struct list_head *tmp = head; | |
c7105365 | 692 | struct mount *p, *found = NULL; |
1da177e4 | 693 | |
1da177e4 | 694 | for (;;) { |
a05964f3 | 695 | tmp = dir ? tmp->next : tmp->prev; |
1da177e4 LT |
696 | p = NULL; |
697 | if (tmp == head) | |
698 | break; | |
1b8e5564 | 699 | p = list_entry(tmp, struct mount, mnt_hash); |
a73324da | 700 | if (&p->mnt_parent->mnt == mnt && p->mnt_mountpoint == dentry) { |
a05964f3 | 701 | found = p; |
1da177e4 LT |
702 | break; |
703 | } | |
704 | } | |
1da177e4 LT |
705 | return found; |
706 | } | |
707 | ||
a05964f3 | 708 | /* |
f015f126 DH |
709 | * lookup_mnt - Return the first child mount mounted at path |
710 | * | |
711 | * "First" means first mounted chronologically. If you create the | |
712 | * following mounts: | |
713 | * | |
714 | * mount /dev/sda1 /mnt | |
715 | * mount /dev/sda2 /mnt | |
716 | * mount /dev/sda3 /mnt | |
717 | * | |
718 | * Then lookup_mnt() on the base /mnt dentry in the root mount will | |
719 | * return successively the root dentry and vfsmount of /dev/sda1, then | |
720 | * /dev/sda2, then /dev/sda3, then NULL. | |
721 | * | |
722 | * lookup_mnt takes a reference to the found vfsmount. | |
a05964f3 | 723 | */ |
1c755af4 | 724 | struct vfsmount *lookup_mnt(struct path *path) |
a05964f3 | 725 | { |
c7105365 | 726 | struct mount *child_mnt; |
99b7db7b | 727 | |
962830df | 728 | br_read_lock(&vfsmount_lock); |
c7105365 AV |
729 | child_mnt = __lookup_mnt(path->mnt, path->dentry, 1); |
730 | if (child_mnt) { | |
731 | mnt_add_count(child_mnt, 1); | |
962830df | 732 | br_read_unlock(&vfsmount_lock); |
c7105365 AV |
733 | return &child_mnt->mnt; |
734 | } else { | |
962830df | 735 | br_read_unlock(&vfsmount_lock); |
c7105365 AV |
736 | return NULL; |
737 | } | |
a05964f3 RP |
738 | } |
739 | ||
84d17192 AV |
740 | static struct mountpoint *new_mountpoint(struct dentry *dentry) |
741 | { | |
742 | struct list_head *chain = mountpoint_hashtable + hash(NULL, dentry); | |
743 | struct mountpoint *mp; | |
744 | ||
745 | list_for_each_entry(mp, chain, m_hash) { | |
746 | if (mp->m_dentry == dentry) { | |
747 | /* might be worth a WARN_ON() */ | |
748 | if (d_unlinked(dentry)) | |
749 | return ERR_PTR(-ENOENT); | |
750 | mp->m_count++; | |
751 | return mp; | |
752 | } | |
753 | } | |
754 | ||
755 | mp = kmalloc(sizeof(struct mountpoint), GFP_KERNEL); | |
756 | if (!mp) | |
757 | return ERR_PTR(-ENOMEM); | |
758 | ||
759 | spin_lock(&dentry->d_lock); | |
760 | if (d_unlinked(dentry)) { | |
761 | spin_unlock(&dentry->d_lock); | |
762 | kfree(mp); | |
763 | return ERR_PTR(-ENOENT); | |
764 | } | |
765 | dentry->d_flags |= DCACHE_MOUNTED; | |
766 | spin_unlock(&dentry->d_lock); | |
767 | mp->m_dentry = dentry; | |
768 | mp->m_count = 1; | |
769 | list_add(&mp->m_hash, chain); | |
770 | return mp; | |
771 | } | |
772 | ||
773 | static void put_mountpoint(struct mountpoint *mp) | |
774 | { | |
775 | if (!--mp->m_count) { | |
776 | struct dentry *dentry = mp->m_dentry; | |
777 | spin_lock(&dentry->d_lock); | |
778 | dentry->d_flags &= ~DCACHE_MOUNTED; | |
779 | spin_unlock(&dentry->d_lock); | |
780 | list_del(&mp->m_hash); | |
781 | kfree(mp); | |
782 | } | |
783 | } | |
784 | ||
143c8c91 | 785 | static inline int check_mnt(struct mount *mnt) |
1da177e4 | 786 | { |
6b3286ed | 787 | return mnt->mnt_ns == current->nsproxy->mnt_ns; |
1da177e4 LT |
788 | } |
789 | ||
99b7db7b NP |
790 | /* |
791 | * vfsmount lock must be held for write | |
792 | */ | |
6b3286ed | 793 | static void touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
794 | { |
795 | if (ns) { | |
796 | ns->event = ++event; | |
797 | wake_up_interruptible(&ns->poll); | |
798 | } | |
799 | } | |
800 | ||
99b7db7b NP |
801 | /* |
802 | * vfsmount lock must be held for write | |
803 | */ | |
6b3286ed | 804 | static void __touch_mnt_namespace(struct mnt_namespace *ns) |
5addc5dd AV |
805 | { |
806 | if (ns && ns->event != event) { | |
807 | ns->event = event; | |
808 | wake_up_interruptible(&ns->poll); | |
809 | } | |
810 | } | |
811 | ||
99b7db7b NP |
812 | /* |
813 | * vfsmount lock must be held for write | |
814 | */ | |
419148da AV |
815 | static void detach_mnt(struct mount *mnt, struct path *old_path) |
816 | { | |
a73324da | 817 | old_path->dentry = mnt->mnt_mountpoint; |
0714a533 AV |
818 | old_path->mnt = &mnt->mnt_parent->mnt; |
819 | mnt->mnt_parent = mnt; | |
a73324da | 820 | mnt->mnt_mountpoint = mnt->mnt.mnt_root; |
6b41d536 | 821 | list_del_init(&mnt->mnt_child); |
1b8e5564 | 822 | list_del_init(&mnt->mnt_hash); |
84d17192 AV |
823 | put_mountpoint(mnt->mnt_mp); |
824 | mnt->mnt_mp = NULL; | |
1da177e4 LT |
825 | } |
826 | ||
99b7db7b NP |
827 | /* |
828 | * vfsmount lock must be held for write | |
829 | */ | |
84d17192 AV |
830 | void mnt_set_mountpoint(struct mount *mnt, |
831 | struct mountpoint *mp, | |
44d964d6 | 832 | struct mount *child_mnt) |
b90fa9ae | 833 | { |
84d17192 | 834 | mp->m_count++; |
3a2393d7 | 835 | mnt_add_count(mnt, 1); /* essentially, that's mntget */ |
84d17192 | 836 | child_mnt->mnt_mountpoint = dget(mp->m_dentry); |
3a2393d7 | 837 | child_mnt->mnt_parent = mnt; |
84d17192 | 838 | child_mnt->mnt_mp = mp; |
b90fa9ae RP |
839 | } |
840 | ||
99b7db7b NP |
841 | /* |
842 | * vfsmount lock must be held for write | |
843 | */ | |
84d17192 AV |
844 | static void attach_mnt(struct mount *mnt, |
845 | struct mount *parent, | |
846 | struct mountpoint *mp) | |
1da177e4 | 847 | { |
84d17192 | 848 | mnt_set_mountpoint(parent, mp, mnt); |
1b8e5564 | 849 | list_add_tail(&mnt->mnt_hash, mount_hashtable + |
84d17192 AV |
850 | hash(&parent->mnt, mp->m_dentry)); |
851 | list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); | |
b90fa9ae RP |
852 | } |
853 | ||
854 | /* | |
99b7db7b | 855 | * vfsmount lock must be held for write |
b90fa9ae | 856 | */ |
4b2619a5 | 857 | static void commit_tree(struct mount *mnt) |
b90fa9ae | 858 | { |
0714a533 | 859 | struct mount *parent = mnt->mnt_parent; |
83adc753 | 860 | struct mount *m; |
b90fa9ae | 861 | LIST_HEAD(head); |
143c8c91 | 862 | struct mnt_namespace *n = parent->mnt_ns; |
b90fa9ae | 863 | |
0714a533 | 864 | BUG_ON(parent == mnt); |
b90fa9ae | 865 | |
1a4eeaf2 | 866 | list_add_tail(&head, &mnt->mnt_list); |
f7a99c5b | 867 | list_for_each_entry(m, &head, mnt_list) |
143c8c91 | 868 | m->mnt_ns = n; |
f03c6599 | 869 | |
b90fa9ae RP |
870 | list_splice(&head, n->list.prev); |
871 | ||
1b8e5564 | 872 | list_add_tail(&mnt->mnt_hash, mount_hashtable + |
a73324da | 873 | hash(&parent->mnt, mnt->mnt_mountpoint)); |
6b41d536 | 874 | list_add_tail(&mnt->mnt_child, &parent->mnt_mounts); |
6b3286ed | 875 | touch_mnt_namespace(n); |
1da177e4 LT |
876 | } |
877 | ||
909b0a88 | 878 | static struct mount *next_mnt(struct mount *p, struct mount *root) |
1da177e4 | 879 | { |
6b41d536 AV |
880 | struct list_head *next = p->mnt_mounts.next; |
881 | if (next == &p->mnt_mounts) { | |
1da177e4 | 882 | while (1) { |
909b0a88 | 883 | if (p == root) |
1da177e4 | 884 | return NULL; |
6b41d536 AV |
885 | next = p->mnt_child.next; |
886 | if (next != &p->mnt_parent->mnt_mounts) | |
1da177e4 | 887 | break; |
0714a533 | 888 | p = p->mnt_parent; |
1da177e4 LT |
889 | } |
890 | } | |
6b41d536 | 891 | return list_entry(next, struct mount, mnt_child); |
1da177e4 LT |
892 | } |
893 | ||
315fc83e | 894 | static struct mount *skip_mnt_tree(struct mount *p) |
9676f0c6 | 895 | { |
6b41d536 AV |
896 | struct list_head *prev = p->mnt_mounts.prev; |
897 | while (prev != &p->mnt_mounts) { | |
898 | p = list_entry(prev, struct mount, mnt_child); | |
899 | prev = p->mnt_mounts.prev; | |
9676f0c6 RP |
900 | } |
901 | return p; | |
902 | } | |
903 | ||
9d412a43 AV |
904 | struct vfsmount * |
905 | vfs_kern_mount(struct file_system_type *type, int flags, const char *name, void *data) | |
906 | { | |
b105e270 | 907 | struct mount *mnt; |
9d412a43 AV |
908 | struct dentry *root; |
909 | ||
910 | if (!type) | |
911 | return ERR_PTR(-ENODEV); | |
912 | ||
913 | mnt = alloc_vfsmnt(name); | |
914 | if (!mnt) | |
915 | return ERR_PTR(-ENOMEM); | |
916 | ||
917 | if (flags & MS_KERNMOUNT) | |
b105e270 | 918 | mnt->mnt.mnt_flags = MNT_INTERNAL; |
9d412a43 AV |
919 | |
920 | root = mount_fs(type, flags, name, data); | |
921 | if (IS_ERR(root)) { | |
922 | free_vfsmnt(mnt); | |
923 | return ERR_CAST(root); | |
924 | } | |
925 | ||
b105e270 AV |
926 | mnt->mnt.mnt_root = root; |
927 | mnt->mnt.mnt_sb = root->d_sb; | |
a73324da | 928 | mnt->mnt_mountpoint = mnt->mnt.mnt_root; |
0714a533 | 929 | mnt->mnt_parent = mnt; |
962830df | 930 | br_write_lock(&vfsmount_lock); |
39f7c4db | 931 | list_add_tail(&mnt->mnt_instance, &root->d_sb->s_mounts); |
962830df | 932 | br_write_unlock(&vfsmount_lock); |
b105e270 | 933 | return &mnt->mnt; |
9d412a43 AV |
934 | } |
935 | EXPORT_SYMBOL_GPL(vfs_kern_mount); | |
936 | ||
87129cc0 | 937 | static struct mount *clone_mnt(struct mount *old, struct dentry *root, |
36341f64 | 938 | int flag) |
1da177e4 | 939 | { |
87129cc0 | 940 | struct super_block *sb = old->mnt.mnt_sb; |
be34d1a3 DH |
941 | struct mount *mnt; |
942 | int err; | |
1da177e4 | 943 | |
be34d1a3 DH |
944 | mnt = alloc_vfsmnt(old->mnt_devname); |
945 | if (!mnt) | |
946 | return ERR_PTR(-ENOMEM); | |
719f5d7f | 947 | |
7a472ef4 | 948 | if (flag & (CL_SLAVE | CL_PRIVATE | CL_SHARED_TO_SLAVE)) |
be34d1a3 DH |
949 | mnt->mnt_group_id = 0; /* not a peer of original */ |
950 | else | |
951 | mnt->mnt_group_id = old->mnt_group_id; | |
b90fa9ae | 952 | |
be34d1a3 DH |
953 | if ((flag & CL_MAKE_SHARED) && !mnt->mnt_group_id) { |
954 | err = mnt_alloc_group_id(mnt); | |
955 | if (err) | |
956 | goto out_free; | |
1da177e4 | 957 | } |
be34d1a3 DH |
958 | |
959 | mnt->mnt.mnt_flags = old->mnt.mnt_flags & ~MNT_WRITE_HOLD; | |
132c94e3 EB |
960 | /* Don't allow unprivileged users to change mount flags */ |
961 | if ((flag & CL_UNPRIVILEGED) && (mnt->mnt.mnt_flags & MNT_READONLY)) | |
962 | mnt->mnt.mnt_flags |= MNT_LOCK_READONLY; | |
963 | ||
be34d1a3 DH |
964 | atomic_inc(&sb->s_active); |
965 | mnt->mnt.mnt_sb = sb; | |
966 | mnt->mnt.mnt_root = dget(root); | |
967 | mnt->mnt_mountpoint = mnt->mnt.mnt_root; | |
968 | mnt->mnt_parent = mnt; | |
969 | br_write_lock(&vfsmount_lock); | |
970 | list_add_tail(&mnt->mnt_instance, &sb->s_mounts); | |
971 | br_write_unlock(&vfsmount_lock); | |
972 | ||
7a472ef4 EB |
973 | if ((flag & CL_SLAVE) || |
974 | ((flag & CL_SHARED_TO_SLAVE) && IS_MNT_SHARED(old))) { | |
be34d1a3 DH |
975 | list_add(&mnt->mnt_slave, &old->mnt_slave_list); |
976 | mnt->mnt_master = old; | |
977 | CLEAR_MNT_SHARED(mnt); | |
978 | } else if (!(flag & CL_PRIVATE)) { | |
979 | if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) | |
980 | list_add(&mnt->mnt_share, &old->mnt_share); | |
981 | if (IS_MNT_SLAVE(old)) | |
982 | list_add(&mnt->mnt_slave, &old->mnt_slave); | |
983 | mnt->mnt_master = old->mnt_master; | |
984 | } | |
985 | if (flag & CL_MAKE_SHARED) | |
986 | set_mnt_shared(mnt); | |
987 | ||
988 | /* stick the duplicate mount on the same expiry list | |
989 | * as the original if that was on one */ | |
990 | if (flag & CL_EXPIRE) { | |
991 | if (!list_empty(&old->mnt_expire)) | |
992 | list_add(&mnt->mnt_expire, &old->mnt_expire); | |
993 | } | |
994 | ||
cb338d06 | 995 | return mnt; |
719f5d7f MS |
996 | |
997 | out_free: | |
998 | free_vfsmnt(mnt); | |
be34d1a3 | 999 | return ERR_PTR(err); |
1da177e4 LT |
1000 | } |
1001 | ||
83adc753 | 1002 | static inline void mntfree(struct mount *mnt) |
1da177e4 | 1003 | { |
83adc753 AV |
1004 | struct vfsmount *m = &mnt->mnt; |
1005 | struct super_block *sb = m->mnt_sb; | |
b3e19d92 | 1006 | |
3d733633 DH |
1007 | /* |
1008 | * This probably indicates that somebody messed | |
1009 | * up a mnt_want/drop_write() pair. If this | |
1010 | * happens, the filesystem was probably unable | |
1011 | * to make r/w->r/o transitions. | |
1012 | */ | |
d3ef3d73 | 1013 | /* |
b3e19d92 NP |
1014 | * The locking used to deal with mnt_count decrement provides barriers, |
1015 | * so mnt_get_writers() below is safe. | |
d3ef3d73 | 1016 | */ |
c6653a83 | 1017 | WARN_ON(mnt_get_writers(mnt)); |
83adc753 AV |
1018 | fsnotify_vfsmount_delete(m); |
1019 | dput(m->mnt_root); | |
1020 | free_vfsmnt(mnt); | |
1da177e4 LT |
1021 | deactivate_super(sb); |
1022 | } | |
1023 | ||
900148dc | 1024 | static void mntput_no_expire(struct mount *mnt) |
b3e19d92 | 1025 | { |
b3e19d92 | 1026 | put_again: |
f03c6599 | 1027 | #ifdef CONFIG_SMP |
962830df | 1028 | br_read_lock(&vfsmount_lock); |
f7a99c5b AV |
1029 | if (likely(mnt->mnt_ns)) { |
1030 | /* shouldn't be the last one */ | |
aa9c0e07 | 1031 | mnt_add_count(mnt, -1); |
962830df | 1032 | br_read_unlock(&vfsmount_lock); |
f03c6599 | 1033 | return; |
b3e19d92 | 1034 | } |
962830df | 1035 | br_read_unlock(&vfsmount_lock); |
b3e19d92 | 1036 | |
962830df | 1037 | br_write_lock(&vfsmount_lock); |
aa9c0e07 | 1038 | mnt_add_count(mnt, -1); |
b3e19d92 | 1039 | if (mnt_get_count(mnt)) { |
962830df | 1040 | br_write_unlock(&vfsmount_lock); |
99b7db7b NP |
1041 | return; |
1042 | } | |
b3e19d92 | 1043 | #else |
aa9c0e07 | 1044 | mnt_add_count(mnt, -1); |
b3e19d92 | 1045 | if (likely(mnt_get_count(mnt))) |
99b7db7b | 1046 | return; |
962830df | 1047 | br_write_lock(&vfsmount_lock); |
f03c6599 | 1048 | #endif |
863d684f AV |
1049 | if (unlikely(mnt->mnt_pinned)) { |
1050 | mnt_add_count(mnt, mnt->mnt_pinned + 1); | |
1051 | mnt->mnt_pinned = 0; | |
962830df | 1052 | br_write_unlock(&vfsmount_lock); |
900148dc | 1053 | acct_auto_close_mnt(&mnt->mnt); |
b3e19d92 | 1054 | goto put_again; |
7b7b1ace | 1055 | } |
962830df | 1056 | |
39f7c4db | 1057 | list_del(&mnt->mnt_instance); |
962830df | 1058 | br_write_unlock(&vfsmount_lock); |
b3e19d92 NP |
1059 | mntfree(mnt); |
1060 | } | |
b3e19d92 NP |
1061 | |
1062 | void mntput(struct vfsmount *mnt) | |
1063 | { | |
1064 | if (mnt) { | |
863d684f | 1065 | struct mount *m = real_mount(mnt); |
b3e19d92 | 1066 | /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ |
863d684f AV |
1067 | if (unlikely(m->mnt_expiry_mark)) |
1068 | m->mnt_expiry_mark = 0; | |
1069 | mntput_no_expire(m); | |
b3e19d92 NP |
1070 | } |
1071 | } | |
1072 | EXPORT_SYMBOL(mntput); | |
1073 | ||
1074 | struct vfsmount *mntget(struct vfsmount *mnt) | |
1075 | { | |
1076 | if (mnt) | |
83adc753 | 1077 | mnt_add_count(real_mount(mnt), 1); |
b3e19d92 NP |
1078 | return mnt; |
1079 | } | |
1080 | EXPORT_SYMBOL(mntget); | |
1081 | ||
7b7b1ace AV |
1082 | void mnt_pin(struct vfsmount *mnt) |
1083 | { | |
962830df | 1084 | br_write_lock(&vfsmount_lock); |
863d684f | 1085 | real_mount(mnt)->mnt_pinned++; |
962830df | 1086 | br_write_unlock(&vfsmount_lock); |
7b7b1ace | 1087 | } |
7b7b1ace AV |
1088 | EXPORT_SYMBOL(mnt_pin); |
1089 | ||
863d684f | 1090 | void mnt_unpin(struct vfsmount *m) |
7b7b1ace | 1091 | { |
863d684f | 1092 | struct mount *mnt = real_mount(m); |
962830df | 1093 | br_write_lock(&vfsmount_lock); |
7b7b1ace | 1094 | if (mnt->mnt_pinned) { |
863d684f | 1095 | mnt_add_count(mnt, 1); |
7b7b1ace AV |
1096 | mnt->mnt_pinned--; |
1097 | } | |
962830df | 1098 | br_write_unlock(&vfsmount_lock); |
7b7b1ace | 1099 | } |
7b7b1ace | 1100 | EXPORT_SYMBOL(mnt_unpin); |
1da177e4 | 1101 | |
b3b304a2 MS |
1102 | static inline void mangle(struct seq_file *m, const char *s) |
1103 | { | |
1104 | seq_escape(m, s, " \t\n\\"); | |
1105 | } | |
1106 | ||
1107 | /* | |
1108 | * Simple .show_options callback for filesystems which don't want to | |
1109 | * implement more complex mount option showing. | |
1110 | * | |
1111 | * See also save_mount_options(). | |
1112 | */ | |
34c80b1d | 1113 | int generic_show_options(struct seq_file *m, struct dentry *root) |
b3b304a2 | 1114 | { |
2a32cebd AV |
1115 | const char *options; |
1116 | ||
1117 | rcu_read_lock(); | |
34c80b1d | 1118 | options = rcu_dereference(root->d_sb->s_options); |
b3b304a2 MS |
1119 | |
1120 | if (options != NULL && options[0]) { | |
1121 | seq_putc(m, ','); | |
1122 | mangle(m, options); | |
1123 | } | |
2a32cebd | 1124 | rcu_read_unlock(); |
b3b304a2 MS |
1125 | |
1126 | return 0; | |
1127 | } | |
1128 | EXPORT_SYMBOL(generic_show_options); | |
1129 | ||
1130 | /* | |
1131 | * If filesystem uses generic_show_options(), this function should be | |
1132 | * called from the fill_super() callback. | |
1133 | * | |
1134 | * The .remount_fs callback usually needs to be handled in a special | |
1135 | * way, to make sure, that previous options are not overwritten if the | |
1136 | * remount fails. | |
1137 | * | |
1138 | * Also note, that if the filesystem's .remount_fs function doesn't | |
1139 | * reset all options to their default value, but changes only newly | |
1140 | * given options, then the displayed options will not reflect reality | |
1141 | * any more. | |
1142 | */ | |
1143 | void save_mount_options(struct super_block *sb, char *options) | |
1144 | { | |
2a32cebd AV |
1145 | BUG_ON(sb->s_options); |
1146 | rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); | |
b3b304a2 MS |
1147 | } |
1148 | EXPORT_SYMBOL(save_mount_options); | |
1149 | ||
2a32cebd AV |
1150 | void replace_mount_options(struct super_block *sb, char *options) |
1151 | { | |
1152 | char *old = sb->s_options; | |
1153 | rcu_assign_pointer(sb->s_options, options); | |
1154 | if (old) { | |
1155 | synchronize_rcu(); | |
1156 | kfree(old); | |
1157 | } | |
1158 | } | |
1159 | EXPORT_SYMBOL(replace_mount_options); | |
1160 | ||
a1a2c409 | 1161 | #ifdef CONFIG_PROC_FS |
0226f492 | 1162 | /* iterator; we want it to have access to namespace_sem, thus here... */ |
1da177e4 LT |
1163 | static void *m_start(struct seq_file *m, loff_t *pos) |
1164 | { | |
6ce6e24e | 1165 | struct proc_mounts *p = proc_mounts(m); |
1da177e4 | 1166 | |
390c6843 | 1167 | down_read(&namespace_sem); |
a1a2c409 | 1168 | return seq_list_start(&p->ns->list, *pos); |
1da177e4 LT |
1169 | } |
1170 | ||
1171 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
1172 | { | |
6ce6e24e | 1173 | struct proc_mounts *p = proc_mounts(m); |
b0765fb8 | 1174 | |
a1a2c409 | 1175 | return seq_list_next(v, &p->ns->list, pos); |
1da177e4 LT |
1176 | } |
1177 | ||
1178 | static void m_stop(struct seq_file *m, void *v) | |
1179 | { | |
390c6843 | 1180 | up_read(&namespace_sem); |
1da177e4 LT |
1181 | } |
1182 | ||
0226f492 | 1183 | static int m_show(struct seq_file *m, void *v) |
2d4d4864 | 1184 | { |
6ce6e24e | 1185 | struct proc_mounts *p = proc_mounts(m); |
1a4eeaf2 | 1186 | struct mount *r = list_entry(v, struct mount, mnt_list); |
0226f492 | 1187 | return p->show(m, &r->mnt); |
1da177e4 LT |
1188 | } |
1189 | ||
a1a2c409 | 1190 | const struct seq_operations mounts_op = { |
1da177e4 LT |
1191 | .start = m_start, |
1192 | .next = m_next, | |
1193 | .stop = m_stop, | |
0226f492 | 1194 | .show = m_show, |
b4629fe2 | 1195 | }; |
a1a2c409 | 1196 | #endif /* CONFIG_PROC_FS */ |
b4629fe2 | 1197 | |
1da177e4 LT |
1198 | /** |
1199 | * may_umount_tree - check if a mount tree is busy | |
1200 | * @mnt: root of mount tree | |
1201 | * | |
1202 | * This is called to check if a tree of mounts has any | |
1203 | * open files, pwds, chroots or sub mounts that are | |
1204 | * busy. | |
1205 | */ | |
909b0a88 | 1206 | int may_umount_tree(struct vfsmount *m) |
1da177e4 | 1207 | { |
909b0a88 | 1208 | struct mount *mnt = real_mount(m); |
36341f64 RP |
1209 | int actual_refs = 0; |
1210 | int minimum_refs = 0; | |
315fc83e | 1211 | struct mount *p; |
909b0a88 | 1212 | BUG_ON(!m); |
1da177e4 | 1213 | |
b3e19d92 | 1214 | /* write lock needed for mnt_get_count */ |
962830df | 1215 | br_write_lock(&vfsmount_lock); |
909b0a88 | 1216 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
83adc753 | 1217 | actual_refs += mnt_get_count(p); |
1da177e4 | 1218 | minimum_refs += 2; |
1da177e4 | 1219 | } |
962830df | 1220 | br_write_unlock(&vfsmount_lock); |
1da177e4 LT |
1221 | |
1222 | if (actual_refs > minimum_refs) | |
e3474a8e | 1223 | return 0; |
1da177e4 | 1224 | |
e3474a8e | 1225 | return 1; |
1da177e4 LT |
1226 | } |
1227 | ||
1228 | EXPORT_SYMBOL(may_umount_tree); | |
1229 | ||
1230 | /** | |
1231 | * may_umount - check if a mount point is busy | |
1232 | * @mnt: root of mount | |
1233 | * | |
1234 | * This is called to check if a mount point has any | |
1235 | * open files, pwds, chroots or sub mounts. If the | |
1236 | * mount has sub mounts this will return busy | |
1237 | * regardless of whether the sub mounts are busy. | |
1238 | * | |
1239 | * Doesn't take quota and stuff into account. IOW, in some cases it will | |
1240 | * give false negatives. The main reason why it's here is that we need | |
1241 | * a non-destructive way to look for easily umountable filesystems. | |
1242 | */ | |
1243 | int may_umount(struct vfsmount *mnt) | |
1244 | { | |
e3474a8e | 1245 | int ret = 1; |
8ad08d8a | 1246 | down_read(&namespace_sem); |
962830df | 1247 | br_write_lock(&vfsmount_lock); |
1ab59738 | 1248 | if (propagate_mount_busy(real_mount(mnt), 2)) |
e3474a8e | 1249 | ret = 0; |
962830df | 1250 | br_write_unlock(&vfsmount_lock); |
8ad08d8a | 1251 | up_read(&namespace_sem); |
a05964f3 | 1252 | return ret; |
1da177e4 LT |
1253 | } |
1254 | ||
1255 | EXPORT_SYMBOL(may_umount); | |
1256 | ||
e3197d83 AV |
1257 | static LIST_HEAD(unmounted); /* protected by namespace_sem */ |
1258 | ||
97216be0 | 1259 | static void namespace_unlock(void) |
70fbcdf4 | 1260 | { |
d5e50f74 | 1261 | struct mount *mnt; |
97216be0 AV |
1262 | LIST_HEAD(head); |
1263 | ||
1264 | if (likely(list_empty(&unmounted))) { | |
1265 | up_write(&namespace_sem); | |
1266 | return; | |
1267 | } | |
1268 | ||
1269 | list_splice_init(&unmounted, &head); | |
1270 | up_write(&namespace_sem); | |
1271 | ||
1272 | while (!list_empty(&head)) { | |
1273 | mnt = list_first_entry(&head, struct mount, mnt_hash); | |
1b8e5564 | 1274 | list_del_init(&mnt->mnt_hash); |
676da58d | 1275 | if (mnt_has_parent(mnt)) { |
70fbcdf4 | 1276 | struct dentry *dentry; |
863d684f | 1277 | struct mount *m; |
99b7db7b | 1278 | |
962830df | 1279 | br_write_lock(&vfsmount_lock); |
a73324da | 1280 | dentry = mnt->mnt_mountpoint; |
863d684f | 1281 | m = mnt->mnt_parent; |
a73324da | 1282 | mnt->mnt_mountpoint = mnt->mnt.mnt_root; |
0714a533 | 1283 | mnt->mnt_parent = mnt; |
7c4b93d8 | 1284 | m->mnt_ghosts--; |
962830df | 1285 | br_write_unlock(&vfsmount_lock); |
70fbcdf4 | 1286 | dput(dentry); |
863d684f | 1287 | mntput(&m->mnt); |
70fbcdf4 | 1288 | } |
d5e50f74 | 1289 | mntput(&mnt->mnt); |
70fbcdf4 RP |
1290 | } |
1291 | } | |
1292 | ||
97216be0 | 1293 | static inline void namespace_lock(void) |
e3197d83 | 1294 | { |
97216be0 | 1295 | down_write(&namespace_sem); |
e3197d83 AV |
1296 | } |
1297 | ||
99b7db7b NP |
1298 | /* |
1299 | * vfsmount lock must be held for write | |
1300 | * namespace_sem must be held for write | |
1301 | */ | |
328e6d90 | 1302 | void umount_tree(struct mount *mnt, int propagate) |
1da177e4 | 1303 | { |
7b8a53fd | 1304 | LIST_HEAD(tmp_list); |
315fc83e | 1305 | struct mount *p; |
1da177e4 | 1306 | |
909b0a88 | 1307 | for (p = mnt; p; p = next_mnt(p, mnt)) |
1b8e5564 | 1308 | list_move(&p->mnt_hash, &tmp_list); |
1da177e4 | 1309 | |
a05964f3 | 1310 | if (propagate) |
7b8a53fd | 1311 | propagate_umount(&tmp_list); |
a05964f3 | 1312 | |
1b8e5564 | 1313 | list_for_each_entry(p, &tmp_list, mnt_hash) { |
6776db3d | 1314 | list_del_init(&p->mnt_expire); |
1a4eeaf2 | 1315 | list_del_init(&p->mnt_list); |
143c8c91 AV |
1316 | __touch_mnt_namespace(p->mnt_ns); |
1317 | p->mnt_ns = NULL; | |
6b41d536 | 1318 | list_del_init(&p->mnt_child); |
676da58d | 1319 | if (mnt_has_parent(p)) { |
863d684f | 1320 | p->mnt_parent->mnt_ghosts++; |
84d17192 AV |
1321 | put_mountpoint(p->mnt_mp); |
1322 | p->mnt_mp = NULL; | |
7c4b93d8 | 1323 | } |
0f0afb1d | 1324 | change_mnt_propagation(p, MS_PRIVATE); |
1da177e4 | 1325 | } |
328e6d90 | 1326 | list_splice(&tmp_list, &unmounted); |
1da177e4 LT |
1327 | } |
1328 | ||
b54b9be7 | 1329 | static void shrink_submounts(struct mount *mnt); |
c35038be | 1330 | |
1ab59738 | 1331 | static int do_umount(struct mount *mnt, int flags) |
1da177e4 | 1332 | { |
1ab59738 | 1333 | struct super_block *sb = mnt->mnt.mnt_sb; |
1da177e4 LT |
1334 | int retval; |
1335 | ||
1ab59738 | 1336 | retval = security_sb_umount(&mnt->mnt, flags); |
1da177e4 LT |
1337 | if (retval) |
1338 | return retval; | |
1339 | ||
1340 | /* | |
1341 | * Allow userspace to request a mountpoint be expired rather than | |
1342 | * unmounting unconditionally. Unmount only happens if: | |
1343 | * (1) the mark is already set (the mark is cleared by mntput()) | |
1344 | * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] | |
1345 | */ | |
1346 | if (flags & MNT_EXPIRE) { | |
1ab59738 | 1347 | if (&mnt->mnt == current->fs->root.mnt || |
1da177e4 LT |
1348 | flags & (MNT_FORCE | MNT_DETACH)) |
1349 | return -EINVAL; | |
1350 | ||
b3e19d92 NP |
1351 | /* |
1352 | * probably don't strictly need the lock here if we examined | |
1353 | * all race cases, but it's a slowpath. | |
1354 | */ | |
962830df | 1355 | br_write_lock(&vfsmount_lock); |
83adc753 | 1356 | if (mnt_get_count(mnt) != 2) { |
962830df | 1357 | br_write_unlock(&vfsmount_lock); |
1da177e4 | 1358 | return -EBUSY; |
b3e19d92 | 1359 | } |
962830df | 1360 | br_write_unlock(&vfsmount_lock); |
1da177e4 | 1361 | |
863d684f | 1362 | if (!xchg(&mnt->mnt_expiry_mark, 1)) |
1da177e4 LT |
1363 | return -EAGAIN; |
1364 | } | |
1365 | ||
1366 | /* | |
1367 | * If we may have to abort operations to get out of this | |
1368 | * mount, and they will themselves hold resources we must | |
1369 | * allow the fs to do things. In the Unix tradition of | |
1370 | * 'Gee thats tricky lets do it in userspace' the umount_begin | |
1371 | * might fail to complete on the first run through as other tasks | |
1372 | * must return, and the like. Thats for the mount program to worry | |
1373 | * about for the moment. | |
1374 | */ | |
1375 | ||
42faad99 | 1376 | if (flags & MNT_FORCE && sb->s_op->umount_begin) { |
42faad99 | 1377 | sb->s_op->umount_begin(sb); |
42faad99 | 1378 | } |
1da177e4 LT |
1379 | |
1380 | /* | |
1381 | * No sense to grab the lock for this test, but test itself looks | |
1382 | * somewhat bogus. Suggestions for better replacement? | |
1383 | * Ho-hum... In principle, we might treat that as umount + switch | |
1384 | * to rootfs. GC would eventually take care of the old vfsmount. | |
1385 | * Actually it makes sense, especially if rootfs would contain a | |
1386 | * /reboot - static binary that would close all descriptors and | |
1387 | * call reboot(9). Then init(8) could umount root and exec /reboot. | |
1388 | */ | |
1ab59738 | 1389 | if (&mnt->mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { |
1da177e4 LT |
1390 | /* |
1391 | * Special case for "unmounting" root ... | |
1392 | * we just try to remount it readonly. | |
1393 | */ | |
1394 | down_write(&sb->s_umount); | |
4aa98cf7 | 1395 | if (!(sb->s_flags & MS_RDONLY)) |
1da177e4 | 1396 | retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); |
1da177e4 LT |
1397 | up_write(&sb->s_umount); |
1398 | return retval; | |
1399 | } | |
1400 | ||
97216be0 | 1401 | namespace_lock(); |
962830df | 1402 | br_write_lock(&vfsmount_lock); |
5addc5dd | 1403 | event++; |
1da177e4 | 1404 | |
c35038be | 1405 | if (!(flags & MNT_DETACH)) |
b54b9be7 | 1406 | shrink_submounts(mnt); |
c35038be | 1407 | |
1da177e4 | 1408 | retval = -EBUSY; |
a05964f3 | 1409 | if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { |
1a4eeaf2 | 1410 | if (!list_empty(&mnt->mnt_list)) |
328e6d90 | 1411 | umount_tree(mnt, 1); |
1da177e4 LT |
1412 | retval = 0; |
1413 | } | |
962830df | 1414 | br_write_unlock(&vfsmount_lock); |
e3197d83 | 1415 | namespace_unlock(); |
1da177e4 LT |
1416 | return retval; |
1417 | } | |
1418 | ||
9b40bc90 AV |
1419 | /* |
1420 | * Is the caller allowed to modify his namespace? | |
1421 | */ | |
1422 | static inline bool may_mount(void) | |
1423 | { | |
1424 | return ns_capable(current->nsproxy->mnt_ns->user_ns, CAP_SYS_ADMIN); | |
1425 | } | |
1426 | ||
1da177e4 LT |
1427 | /* |
1428 | * Now umount can handle mount points as well as block devices. | |
1429 | * This is important for filesystems which use unnamed block devices. | |
1430 | * | |
1431 | * We now support a flag for forced unmount like the other 'big iron' | |
1432 | * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD | |
1433 | */ | |
1434 | ||
bdc480e3 | 1435 | SYSCALL_DEFINE2(umount, char __user *, name, int, flags) |
1da177e4 | 1436 | { |
2d8f3038 | 1437 | struct path path; |
900148dc | 1438 | struct mount *mnt; |
1da177e4 | 1439 | int retval; |
db1f05bb | 1440 | int lookup_flags = 0; |
6fa3eb70 S |
1441 | #ifdef UMOUNT_LOG |
1442 | int total_value =0; | |
1443 | #endif | |
db1f05bb MS |
1444 | if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) |
1445 | return -EINVAL; | |
1446 | ||
9b40bc90 AV |
1447 | if (!may_mount()) |
1448 | return -EPERM; | |
1449 | ||
db1f05bb MS |
1450 | if (!(flags & UMOUNT_NOFOLLOW)) |
1451 | lookup_flags |= LOOKUP_FOLLOW; | |
1452 | ||
1453 | retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); | |
1da177e4 LT |
1454 | if (retval) |
1455 | goto out; | |
900148dc | 1456 | mnt = real_mount(path.mnt); |
1da177e4 | 1457 | retval = -EINVAL; |
2d8f3038 | 1458 | if (path.dentry != path.mnt->mnt_root) |
1da177e4 | 1459 | goto dput_and_out; |
143c8c91 | 1460 | if (!check_mnt(mnt)) |
1da177e4 LT |
1461 | goto dput_and_out; |
1462 | ||
900148dc | 1463 | retval = do_umount(mnt, flags); |
6fa3eb70 S |
1464 | #ifdef UMOUNT_LOG |
1465 | { | |
1466 | printk("Ahsin do_umount retval=%d \n",retval); | |
1467 | //do_umount success: 0, do_umount busy: -16 | |
1468 | //if do_umount fail, need to dump the link list here | |
1469 | ||
1470 | if(retval) | |
1471 | printk("Ahsin do_umount fail; mnt_get_count=%d mnt->mnt_devname=%s\n",mnt_get_count(mnt),mnt->mnt_devname); | |
1472 | else | |
1473 | printk("Ahsin do_umount success; mnt_get_count=%d mnt->mnt_devname=%s\n",mnt_get_count(mnt),mnt->mnt_devname); | |
1474 | ||
1475 | // print linked list | |
1476 | spin_lock(&mnt_id_lock); | |
1477 | ref_current = ref_head; | |
1478 | while(ref_current != NULL) | |
1479 | { | |
1480 | total_value = total_value + ref_current->count; | |
1481 | ||
1482 | if (ref_current->count) | |
1483 | printk("Ahsin PID= %d, Name = %s, Count= %d \n", ref_current->pid, ref_current->name, ref_current->count); | |
1484 | ref_current = ref_current->next; | |
1485 | } | |
1486 | spin_unlock(&mnt_id_lock); | |
1487 | printk("Ahsin total_value=%d \n",total_value); | |
1488 | } | |
1489 | #endif | |
1da177e4 | 1490 | dput_and_out: |
429731b1 | 1491 | /* we mustn't call path_put() as that would clear mnt_expiry_mark */ |
2d8f3038 | 1492 | dput(path.dentry); |
900148dc | 1493 | mntput_no_expire(mnt); |
1da177e4 LT |
1494 | out: |
1495 | return retval; | |
1496 | } | |
1497 | ||
1498 | #ifdef __ARCH_WANT_SYS_OLDUMOUNT | |
1499 | ||
1500 | /* | |
b58fed8b | 1501 | * The 2.0 compatible umount. No flags. |
1da177e4 | 1502 | */ |
bdc480e3 | 1503 | SYSCALL_DEFINE1(oldumount, char __user *, name) |
1da177e4 | 1504 | { |
b58fed8b | 1505 | return sys_umount(name, 0); |
1da177e4 LT |
1506 | } |
1507 | ||
1508 | #endif | |
1509 | ||
8823c079 EB |
1510 | static bool mnt_ns_loop(struct path *path) |
1511 | { | |
1512 | /* Could bind mounting the mount namespace inode cause a | |
1513 | * mount namespace loop? | |
1514 | */ | |
1515 | struct inode *inode = path->dentry->d_inode; | |
0bb80f24 | 1516 | struct proc_ns *ei; |
8823c079 EB |
1517 | struct mnt_namespace *mnt_ns; |
1518 | ||
1519 | if (!proc_ns_inode(inode)) | |
1520 | return false; | |
1521 | ||
0bb80f24 | 1522 | ei = get_proc_ns(inode); |
8823c079 EB |
1523 | if (ei->ns_ops != &mntns_operations) |
1524 | return false; | |
1525 | ||
1526 | mnt_ns = ei->ns; | |
1527 | return current->nsproxy->mnt_ns->seq >= mnt_ns->seq; | |
1528 | } | |
1529 | ||
87129cc0 | 1530 | struct mount *copy_tree(struct mount *mnt, struct dentry *dentry, |
36341f64 | 1531 | int flag) |
1da177e4 | 1532 | { |
84d17192 | 1533 | struct mount *res, *p, *q, *r, *parent; |
1da177e4 | 1534 | |
fc7be130 | 1535 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) |
be34d1a3 | 1536 | return ERR_PTR(-EINVAL); |
9676f0c6 | 1537 | |
36341f64 | 1538 | res = q = clone_mnt(mnt, dentry, flag); |
be34d1a3 DH |
1539 | if (IS_ERR(q)) |
1540 | return q; | |
1541 | ||
a73324da | 1542 | q->mnt_mountpoint = mnt->mnt_mountpoint; |
1da177e4 LT |
1543 | |
1544 | p = mnt; | |
6b41d536 | 1545 | list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { |
315fc83e | 1546 | struct mount *s; |
7ec02ef1 | 1547 | if (!is_subdir(r->mnt_mountpoint, dentry)) |
1da177e4 LT |
1548 | continue; |
1549 | ||
909b0a88 | 1550 | for (s = r; s; s = next_mnt(s, r)) { |
fc7be130 | 1551 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { |
9676f0c6 RP |
1552 | s = skip_mnt_tree(s); |
1553 | continue; | |
1554 | } | |
0714a533 AV |
1555 | while (p != s->mnt_parent) { |
1556 | p = p->mnt_parent; | |
1557 | q = q->mnt_parent; | |
1da177e4 | 1558 | } |
87129cc0 | 1559 | p = s; |
84d17192 | 1560 | parent = q; |
87129cc0 | 1561 | q = clone_mnt(p, p->mnt.mnt_root, flag); |
be34d1a3 DH |
1562 | if (IS_ERR(q)) |
1563 | goto out; | |
962830df | 1564 | br_write_lock(&vfsmount_lock); |
1a4eeaf2 | 1565 | list_add_tail(&q->mnt_list, &res->mnt_list); |
84d17192 | 1566 | attach_mnt(q, parent, p->mnt_mp); |
962830df | 1567 | br_write_unlock(&vfsmount_lock); |
1da177e4 LT |
1568 | } |
1569 | } | |
1570 | return res; | |
be34d1a3 | 1571 | out: |
1da177e4 | 1572 | if (res) { |
962830df | 1573 | br_write_lock(&vfsmount_lock); |
328e6d90 | 1574 | umount_tree(res, 0); |
962830df | 1575 | br_write_unlock(&vfsmount_lock); |
1da177e4 | 1576 | } |
be34d1a3 | 1577 | return q; |
1da177e4 LT |
1578 | } |
1579 | ||
be34d1a3 DH |
1580 | /* Caller should check returned pointer for errors */ |
1581 | ||
589ff870 | 1582 | struct vfsmount *collect_mounts(struct path *path) |
8aec0809 | 1583 | { |
cb338d06 | 1584 | struct mount *tree; |
97216be0 | 1585 | namespace_lock(); |
87129cc0 AV |
1586 | tree = copy_tree(real_mount(path->mnt), path->dentry, |
1587 | CL_COPY_ALL | CL_PRIVATE); | |
328e6d90 | 1588 | namespace_unlock(); |
be34d1a3 | 1589 | if (IS_ERR(tree)) |
2bee1e0f | 1590 | return ERR_CAST(tree); |
be34d1a3 | 1591 | return &tree->mnt; |
8aec0809 AV |
1592 | } |
1593 | ||
1594 | void drop_collected_mounts(struct vfsmount *mnt) | |
1595 | { | |
97216be0 | 1596 | namespace_lock(); |
962830df | 1597 | br_write_lock(&vfsmount_lock); |
328e6d90 | 1598 | umount_tree(real_mount(mnt), 0); |
962830df | 1599 | br_write_unlock(&vfsmount_lock); |
3ab6abee | 1600 | namespace_unlock(); |
8aec0809 AV |
1601 | } |
1602 | ||
1f707137 AV |
1603 | int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, |
1604 | struct vfsmount *root) | |
1605 | { | |
1a4eeaf2 | 1606 | struct mount *mnt; |
1f707137 AV |
1607 | int res = f(root, arg); |
1608 | if (res) | |
1609 | return res; | |
1a4eeaf2 AV |
1610 | list_for_each_entry(mnt, &real_mount(root)->mnt_list, mnt_list) { |
1611 | res = f(&mnt->mnt, arg); | |
1f707137 AV |
1612 | if (res) |
1613 | return res; | |
1614 | } | |
1615 | return 0; | |
1616 | } | |
1617 | ||
4b8b21f4 | 1618 | static void cleanup_group_ids(struct mount *mnt, struct mount *end) |
719f5d7f | 1619 | { |
315fc83e | 1620 | struct mount *p; |
719f5d7f | 1621 | |
909b0a88 | 1622 | for (p = mnt; p != end; p = next_mnt(p, mnt)) { |
fc7be130 | 1623 | if (p->mnt_group_id && !IS_MNT_SHARED(p)) |
4b8b21f4 | 1624 | mnt_release_group_id(p); |
719f5d7f MS |
1625 | } |
1626 | } | |
1627 | ||
4b8b21f4 | 1628 | static int invent_group_ids(struct mount *mnt, bool recurse) |
719f5d7f | 1629 | { |
315fc83e | 1630 | struct mount *p; |
719f5d7f | 1631 | |
909b0a88 | 1632 | for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { |
fc7be130 | 1633 | if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { |
4b8b21f4 | 1634 | int err = mnt_alloc_group_id(p); |
719f5d7f | 1635 | if (err) { |
4b8b21f4 | 1636 | cleanup_group_ids(mnt, p); |
719f5d7f MS |
1637 | return err; |
1638 | } | |
1639 | } | |
1640 | } | |
1641 | ||
1642 | return 0; | |
1643 | } | |
1644 | ||
b90fa9ae RP |
1645 | /* |
1646 | * @source_mnt : mount tree to be attached | |
21444403 RP |
1647 | * @nd : place the mount tree @source_mnt is attached |
1648 | * @parent_nd : if non-null, detach the source_mnt from its parent and | |
1649 | * store the parent mount and mountpoint dentry. | |
1650 | * (done when source_mnt is moved) | |
b90fa9ae RP |
1651 | * |
1652 | * NOTE: in the table below explains the semantics when a source mount | |
1653 | * of a given type is attached to a destination mount of a given type. | |
9676f0c6 RP |
1654 | * --------------------------------------------------------------------------- |
1655 | * | BIND MOUNT OPERATION | | |
1656 | * |************************************************************************** | |
1657 | * | source-->| shared | private | slave | unbindable | | |
1658 | * | dest | | | | | | |
1659 | * | | | | | | | | |
1660 | * | v | | | | | | |
1661 | * |************************************************************************** | |
1662 | * | shared | shared (++) | shared (+) | shared(+++)| invalid | | |
1663 | * | | | | | | | |
1664 | * |non-shared| shared (+) | private | slave (*) | invalid | | |
1665 | * *************************************************************************** | |
b90fa9ae RP |
1666 | * A bind operation clones the source mount and mounts the clone on the |
1667 | * destination mount. | |
1668 | * | |
1669 | * (++) the cloned mount is propagated to all the mounts in the propagation | |
1670 | * tree of the destination mount and the cloned mount is added to | |
1671 | * the peer group of the source mount. | |
1672 | * (+) the cloned mount is created under the destination mount and is marked | |
1673 | * as shared. The cloned mount is added to the peer group of the source | |
1674 | * mount. | |
5afe0022 RP |
1675 | * (+++) the mount is propagated to all the mounts in the propagation tree |
1676 | * of the destination mount and the cloned mount is made slave | |
1677 | * of the same master as that of the source mount. The cloned mount | |
1678 | * is marked as 'shared and slave'. | |
1679 | * (*) the cloned mount is made a slave of the same master as that of the | |
1680 | * source mount. | |
1681 | * | |
9676f0c6 RP |
1682 | * --------------------------------------------------------------------------- |
1683 | * | MOVE MOUNT OPERATION | | |
1684 | * |************************************************************************** | |
1685 | * | source-->| shared | private | slave | unbindable | | |
1686 | * | dest | | | | | | |
1687 | * | | | | | | | | |
1688 | * | v | | | | | | |
1689 | * |************************************************************************** | |
1690 | * | shared | shared (+) | shared (+) | shared(+++) | invalid | | |
1691 | * | | | | | | | |
1692 | * |non-shared| shared (+*) | private | slave (*) | unbindable | | |
1693 | * *************************************************************************** | |
5afe0022 RP |
1694 | * |
1695 | * (+) the mount is moved to the destination. And is then propagated to | |
1696 | * all the mounts in the propagation tree of the destination mount. | |
21444403 | 1697 | * (+*) the mount is moved to the destination. |
5afe0022 RP |
1698 | * (+++) the mount is moved to the destination and is then propagated to |
1699 | * all the mounts belonging to the destination mount's propagation tree. | |
1700 | * the mount is marked as 'shared and slave'. | |
1701 | * (*) the mount continues to be a slave at the new location. | |
b90fa9ae RP |
1702 | * |
1703 | * if the source mount is a tree, the operations explained above is | |
1704 | * applied to each mount in the tree. | |
1705 | * Must be called without spinlocks held, since this function can sleep | |
1706 | * in allocations. | |
1707 | */ | |
0fb54e50 | 1708 | static int attach_recursive_mnt(struct mount *source_mnt, |
84d17192 AV |
1709 | struct mount *dest_mnt, |
1710 | struct mountpoint *dest_mp, | |
1711 | struct path *parent_path) | |
b90fa9ae RP |
1712 | { |
1713 | LIST_HEAD(tree_list); | |
315fc83e | 1714 | struct mount *child, *p; |
719f5d7f | 1715 | int err; |
b90fa9ae | 1716 | |
fc7be130 | 1717 | if (IS_MNT_SHARED(dest_mnt)) { |
0fb54e50 | 1718 | err = invent_group_ids(source_mnt, true); |
719f5d7f MS |
1719 | if (err) |
1720 | goto out; | |
1721 | } | |
84d17192 | 1722 | err = propagate_mnt(dest_mnt, dest_mp, source_mnt, &tree_list); |
719f5d7f MS |
1723 | if (err) |
1724 | goto out_cleanup_ids; | |
b90fa9ae | 1725 | |
962830df | 1726 | br_write_lock(&vfsmount_lock); |
df1a1ad2 | 1727 | |
fc7be130 | 1728 | if (IS_MNT_SHARED(dest_mnt)) { |
909b0a88 | 1729 | for (p = source_mnt; p; p = next_mnt(p, source_mnt)) |
0f0afb1d | 1730 | set_mnt_shared(p); |
b90fa9ae | 1731 | } |
1a390689 | 1732 | if (parent_path) { |
0fb54e50 | 1733 | detach_mnt(source_mnt, parent_path); |
84d17192 | 1734 | attach_mnt(source_mnt, dest_mnt, dest_mp); |
143c8c91 | 1735 | touch_mnt_namespace(source_mnt->mnt_ns); |
21444403 | 1736 | } else { |
84d17192 | 1737 | mnt_set_mountpoint(dest_mnt, dest_mp, source_mnt); |
0fb54e50 | 1738 | commit_tree(source_mnt); |
21444403 | 1739 | } |
b90fa9ae | 1740 | |
1b8e5564 AV |
1741 | list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { |
1742 | list_del_init(&child->mnt_hash); | |
4b2619a5 | 1743 | commit_tree(child); |
b90fa9ae | 1744 | } |
962830df | 1745 | br_write_unlock(&vfsmount_lock); |
99b7db7b | 1746 | |
b90fa9ae | 1747 | return 0; |
719f5d7f MS |
1748 | |
1749 | out_cleanup_ids: | |
fc7be130 | 1750 | if (IS_MNT_SHARED(dest_mnt)) |
0fb54e50 | 1751 | cleanup_group_ids(source_mnt, NULL); |
719f5d7f MS |
1752 | out: |
1753 | return err; | |
b90fa9ae RP |
1754 | } |
1755 | ||
84d17192 | 1756 | static struct mountpoint *lock_mount(struct path *path) |
b12cea91 AV |
1757 | { |
1758 | struct vfsmount *mnt; | |
84d17192 | 1759 | struct dentry *dentry = path->dentry; |
b12cea91 | 1760 | retry: |
84d17192 AV |
1761 | mutex_lock(&dentry->d_inode->i_mutex); |
1762 | if (unlikely(cant_mount(dentry))) { | |
1763 | mutex_unlock(&dentry->d_inode->i_mutex); | |
1764 | return ERR_PTR(-ENOENT); | |
b12cea91 | 1765 | } |
97216be0 | 1766 | namespace_lock(); |
b12cea91 | 1767 | mnt = lookup_mnt(path); |
84d17192 AV |
1768 | if (likely(!mnt)) { |
1769 | struct mountpoint *mp = new_mountpoint(dentry); | |
1770 | if (IS_ERR(mp)) { | |
97216be0 | 1771 | namespace_unlock(); |
84d17192 AV |
1772 | mutex_unlock(&dentry->d_inode->i_mutex); |
1773 | return mp; | |
1774 | } | |
1775 | return mp; | |
1776 | } | |
97216be0 | 1777 | namespace_unlock(); |
b12cea91 AV |
1778 | mutex_unlock(&path->dentry->d_inode->i_mutex); |
1779 | path_put(path); | |
1780 | path->mnt = mnt; | |
84d17192 | 1781 | dentry = path->dentry = dget(mnt->mnt_root); |
b12cea91 AV |
1782 | goto retry; |
1783 | } | |
1784 | ||
84d17192 | 1785 | static void unlock_mount(struct mountpoint *where) |
b12cea91 | 1786 | { |
84d17192 AV |
1787 | struct dentry *dentry = where->m_dentry; |
1788 | put_mountpoint(where); | |
328e6d90 | 1789 | namespace_unlock(); |
84d17192 | 1790 | mutex_unlock(&dentry->d_inode->i_mutex); |
b12cea91 AV |
1791 | } |
1792 | ||
84d17192 | 1793 | static int graft_tree(struct mount *mnt, struct mount *p, struct mountpoint *mp) |
1da177e4 | 1794 | { |
95bc5f25 | 1795 | if (mnt->mnt.mnt_sb->s_flags & MS_NOUSER) |
1da177e4 LT |
1796 | return -EINVAL; |
1797 | ||
84d17192 | 1798 | if (S_ISDIR(mp->m_dentry->d_inode->i_mode) != |
95bc5f25 | 1799 | S_ISDIR(mnt->mnt.mnt_root->d_inode->i_mode)) |
1da177e4 LT |
1800 | return -ENOTDIR; |
1801 | ||
84d17192 | 1802 | return attach_recursive_mnt(mnt, p, mp, NULL); |
1da177e4 LT |
1803 | } |
1804 | ||
7a2e8a8f VA |
1805 | /* |
1806 | * Sanity check the flags to change_mnt_propagation. | |
1807 | */ | |
1808 | ||
1809 | static int flags_to_propagation_type(int flags) | |
1810 | { | |
7c6e984d | 1811 | int type = flags & ~(MS_REC | MS_SILENT); |
7a2e8a8f VA |
1812 | |
1813 | /* Fail if any non-propagation flags are set */ | |
1814 | if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) | |
1815 | return 0; | |
1816 | /* Only one propagation flag should be set */ | |
1817 | if (!is_power_of_2(type)) | |
1818 | return 0; | |
1819 | return type; | |
1820 | } | |
1821 | ||
07b20889 RP |
1822 | /* |
1823 | * recursively change the type of the mountpoint. | |
1824 | */ | |
0a0d8a46 | 1825 | static int do_change_type(struct path *path, int flag) |
07b20889 | 1826 | { |
315fc83e | 1827 | struct mount *m; |
4b8b21f4 | 1828 | struct mount *mnt = real_mount(path->mnt); |
07b20889 | 1829 | int recurse = flag & MS_REC; |
7a2e8a8f | 1830 | int type; |
719f5d7f | 1831 | int err = 0; |
07b20889 | 1832 | |
2d92ab3c | 1833 | if (path->dentry != path->mnt->mnt_root) |
07b20889 RP |
1834 | return -EINVAL; |
1835 | ||
7a2e8a8f VA |
1836 | type = flags_to_propagation_type(flag); |
1837 | if (!type) | |
1838 | return -EINVAL; | |
1839 | ||
97216be0 | 1840 | namespace_lock(); |
719f5d7f MS |
1841 | if (type == MS_SHARED) { |
1842 | err = invent_group_ids(mnt, recurse); | |
1843 | if (err) | |
1844 | goto out_unlock; | |
1845 | } | |
1846 | ||
962830df | 1847 | br_write_lock(&vfsmount_lock); |
909b0a88 | 1848 | for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) |
0f0afb1d | 1849 | change_mnt_propagation(m, type); |
962830df | 1850 | br_write_unlock(&vfsmount_lock); |
719f5d7f MS |
1851 | |
1852 | out_unlock: | |
97216be0 | 1853 | namespace_unlock(); |
719f5d7f | 1854 | return err; |
07b20889 RP |
1855 | } |
1856 | ||
1da177e4 LT |
1857 | /* |
1858 | * do loopback mount. | |
1859 | */ | |
808d4e3c | 1860 | static int do_loopback(struct path *path, const char *old_name, |
2dafe1c4 | 1861 | int recurse) |
1da177e4 | 1862 | { |
2d92ab3c | 1863 | struct path old_path; |
84d17192 AV |
1864 | struct mount *mnt = NULL, *old, *parent; |
1865 | struct mountpoint *mp; | |
57eccb83 | 1866 | int err; |
1da177e4 LT |
1867 | if (!old_name || !*old_name) |
1868 | return -EINVAL; | |
815d405c | 1869 | err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); |
1da177e4 LT |
1870 | if (err) |
1871 | return err; | |
1872 | ||
8823c079 EB |
1873 | err = -EINVAL; |
1874 | if (mnt_ns_loop(&old_path)) | |
1875 | goto out; | |
1876 | ||
84d17192 AV |
1877 | mp = lock_mount(path); |
1878 | err = PTR_ERR(mp); | |
1879 | if (IS_ERR(mp)) | |
b12cea91 AV |
1880 | goto out; |
1881 | ||
87129cc0 | 1882 | old = real_mount(old_path.mnt); |
84d17192 | 1883 | parent = real_mount(path->mnt); |
87129cc0 | 1884 | |
1da177e4 | 1885 | err = -EINVAL; |
fc7be130 | 1886 | if (IS_MNT_UNBINDABLE(old)) |
b12cea91 | 1887 | goto out2; |
9676f0c6 | 1888 | |
84d17192 | 1889 | if (!check_mnt(parent) || !check_mnt(old)) |
b12cea91 | 1890 | goto out2; |
1da177e4 | 1891 | |
ccd48bc7 | 1892 | if (recurse) |
87129cc0 | 1893 | mnt = copy_tree(old, old_path.dentry, 0); |
ccd48bc7 | 1894 | else |
87129cc0 | 1895 | mnt = clone_mnt(old, old_path.dentry, 0); |
ccd48bc7 | 1896 | |
be34d1a3 DH |
1897 | if (IS_ERR(mnt)) { |
1898 | err = PTR_ERR(mnt); | |
e9c5d8a5 | 1899 | goto out2; |
be34d1a3 | 1900 | } |
ccd48bc7 | 1901 | |
84d17192 | 1902 | err = graft_tree(mnt, parent, mp); |
ccd48bc7 | 1903 | if (err) { |
962830df | 1904 | br_write_lock(&vfsmount_lock); |
328e6d90 | 1905 | umount_tree(mnt, 0); |
962830df | 1906 | br_write_unlock(&vfsmount_lock); |
5b83d2c5 | 1907 | } |
b12cea91 | 1908 | out2: |
84d17192 | 1909 | unlock_mount(mp); |
ccd48bc7 | 1910 | out: |
2d92ab3c | 1911 | path_put(&old_path); |
1da177e4 LT |
1912 | return err; |
1913 | } | |
1914 | ||
2e4b7fcd DH |
1915 | static int change_mount_flags(struct vfsmount *mnt, int ms_flags) |
1916 | { | |
1917 | int error = 0; | |
1918 | int readonly_request = 0; | |
1919 | ||
1920 | if (ms_flags & MS_RDONLY) | |
1921 | readonly_request = 1; | |
1922 | if (readonly_request == __mnt_is_readonly(mnt)) | |
1923 | return 0; | |
1924 | ||
90563b19 EB |
1925 | if (mnt->mnt_flags & MNT_LOCK_READONLY) |
1926 | return -EPERM; | |
1927 | ||
2e4b7fcd | 1928 | if (readonly_request) |
83adc753 | 1929 | error = mnt_make_readonly(real_mount(mnt)); |
2e4b7fcd | 1930 | else |
83adc753 | 1931 | __mnt_unmake_readonly(real_mount(mnt)); |
2e4b7fcd DH |
1932 | return error; |
1933 | } | |
1934 | ||
1da177e4 LT |
1935 | /* |
1936 | * change filesystem flags. dir should be a physical root of filesystem. | |
1937 | * If you've mounted a non-root directory somewhere and want to do remount | |
1938 | * on it - tough luck. | |
1939 | */ | |
0a0d8a46 | 1940 | static int do_remount(struct path *path, int flags, int mnt_flags, |
1da177e4 LT |
1941 | void *data) |
1942 | { | |
1943 | int err; | |
2d92ab3c | 1944 | struct super_block *sb = path->mnt->mnt_sb; |
143c8c91 | 1945 | struct mount *mnt = real_mount(path->mnt); |
1da177e4 | 1946 | |
143c8c91 | 1947 | if (!check_mnt(mnt)) |
1da177e4 LT |
1948 | return -EINVAL; |
1949 | ||
2d92ab3c | 1950 | if (path->dentry != path->mnt->mnt_root) |
1da177e4 LT |
1951 | return -EINVAL; |
1952 | ||
ff36fe2c EP |
1953 | err = security_sb_remount(sb, data); |
1954 | if (err) | |
1955 | return err; | |
1956 | ||
1da177e4 | 1957 | down_write(&sb->s_umount); |
2e4b7fcd | 1958 | if (flags & MS_BIND) |
2d92ab3c | 1959 | err = change_mount_flags(path->mnt, flags); |
57eccb83 AV |
1960 | else if (!capable(CAP_SYS_ADMIN)) |
1961 | err = -EPERM; | |
4aa98cf7 | 1962 | else |
2e4b7fcd | 1963 | err = do_remount_sb(sb, flags, data, 0); |
7b43a79f | 1964 | if (!err) { |
962830df | 1965 | br_write_lock(&vfsmount_lock); |
143c8c91 AV |
1966 | mnt_flags |= mnt->mnt.mnt_flags & MNT_PROPAGATION_MASK; |
1967 | mnt->mnt.mnt_flags = mnt_flags; | |
962830df | 1968 | br_write_unlock(&vfsmount_lock); |
7b43a79f | 1969 | } |
1da177e4 | 1970 | up_write(&sb->s_umount); |
0e55a7cc | 1971 | if (!err) { |
962830df | 1972 | br_write_lock(&vfsmount_lock); |
143c8c91 | 1973 | touch_mnt_namespace(mnt->mnt_ns); |
962830df | 1974 | br_write_unlock(&vfsmount_lock); |
0e55a7cc | 1975 | } |
1da177e4 LT |
1976 | return err; |
1977 | } | |
1978 | ||
cbbe362c | 1979 | static inline int tree_contains_unbindable(struct mount *mnt) |
9676f0c6 | 1980 | { |
315fc83e | 1981 | struct mount *p; |
909b0a88 | 1982 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
fc7be130 | 1983 | if (IS_MNT_UNBINDABLE(p)) |
9676f0c6 RP |
1984 | return 1; |
1985 | } | |
1986 | return 0; | |
1987 | } | |
1988 | ||
808d4e3c | 1989 | static int do_move_mount(struct path *path, const char *old_name) |
1da177e4 | 1990 | { |
2d92ab3c | 1991 | struct path old_path, parent_path; |
676da58d | 1992 | struct mount *p; |
0fb54e50 | 1993 | struct mount *old; |
84d17192 | 1994 | struct mountpoint *mp; |
57eccb83 | 1995 | int err; |
1da177e4 LT |
1996 | if (!old_name || !*old_name) |
1997 | return -EINVAL; | |
2d92ab3c | 1998 | err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); |
1da177e4 LT |
1999 | if (err) |
2000 | return err; | |
2001 | ||
84d17192 AV |
2002 | mp = lock_mount(path); |
2003 | err = PTR_ERR(mp); | |
2004 | if (IS_ERR(mp)) | |
cc53ce53 DH |
2005 | goto out; |
2006 | ||
143c8c91 | 2007 | old = real_mount(old_path.mnt); |
fc7be130 | 2008 | p = real_mount(path->mnt); |
143c8c91 | 2009 | |
1da177e4 | 2010 | err = -EINVAL; |
fc7be130 | 2011 | if (!check_mnt(p) || !check_mnt(old)) |
1da177e4 LT |
2012 | goto out1; |
2013 | ||
1da177e4 | 2014 | err = -EINVAL; |
2d92ab3c | 2015 | if (old_path.dentry != old_path.mnt->mnt_root) |
21444403 | 2016 | goto out1; |
1da177e4 | 2017 | |
676da58d | 2018 | if (!mnt_has_parent(old)) |
21444403 | 2019 | goto out1; |
1da177e4 | 2020 | |
2d92ab3c AV |
2021 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
2022 | S_ISDIR(old_path.dentry->d_inode->i_mode)) | |
21444403 RP |
2023 | goto out1; |
2024 | /* | |
2025 | * Don't move a mount residing in a shared parent. | |
2026 | */ | |
fc7be130 | 2027 | if (IS_MNT_SHARED(old->mnt_parent)) |
21444403 | 2028 | goto out1; |
9676f0c6 RP |
2029 | /* |
2030 | * Don't move a mount tree containing unbindable mounts to a destination | |
2031 | * mount which is shared. | |
2032 | */ | |
fc7be130 | 2033 | if (IS_MNT_SHARED(p) && tree_contains_unbindable(old)) |
9676f0c6 | 2034 | goto out1; |
1da177e4 | 2035 | err = -ELOOP; |
fc7be130 | 2036 | for (; mnt_has_parent(p); p = p->mnt_parent) |
676da58d | 2037 | if (p == old) |
21444403 | 2038 | goto out1; |
1da177e4 | 2039 | |
84d17192 | 2040 | err = attach_recursive_mnt(old, real_mount(path->mnt), mp, &parent_path); |
4ac91378 | 2041 | if (err) |
21444403 | 2042 | goto out1; |
1da177e4 LT |
2043 | |
2044 | /* if the mount is moved, it should no longer be expire | |
2045 | * automatically */ | |
6776db3d | 2046 | list_del_init(&old->mnt_expire); |
1da177e4 | 2047 | out1: |
84d17192 | 2048 | unlock_mount(mp); |
1da177e4 | 2049 | out: |
1da177e4 | 2050 | if (!err) |
1a390689 | 2051 | path_put(&parent_path); |
2d92ab3c | 2052 | path_put(&old_path); |
1da177e4 LT |
2053 | return err; |
2054 | } | |
2055 | ||
9d412a43 AV |
2056 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
2057 | { | |
2058 | int err; | |
2059 | const char *subtype = strchr(fstype, '.'); | |
2060 | if (subtype) { | |
2061 | subtype++; | |
2062 | err = -EINVAL; | |
2063 | if (!subtype[0]) | |
2064 | goto err; | |
2065 | } else | |
2066 | subtype = ""; | |
2067 | ||
2068 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
2069 | err = -ENOMEM; | |
2070 | if (!mnt->mnt_sb->s_subtype) | |
2071 | goto err; | |
2072 | return mnt; | |
2073 | ||
2074 | err: | |
2075 | mntput(mnt); | |
2076 | return ERR_PTR(err); | |
2077 | } | |
2078 | ||
9d412a43 AV |
2079 | /* |
2080 | * add a mount into a namespace's mount tree | |
2081 | */ | |
95bc5f25 | 2082 | static int do_add_mount(struct mount *newmnt, struct path *path, int mnt_flags) |
9d412a43 | 2083 | { |
84d17192 AV |
2084 | struct mountpoint *mp; |
2085 | struct mount *parent; | |
9d412a43 AV |
2086 | int err; |
2087 | ||
2088 | mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); | |
2089 | ||
84d17192 AV |
2090 | mp = lock_mount(path); |
2091 | if (IS_ERR(mp)) | |
2092 | return PTR_ERR(mp); | |
9d412a43 | 2093 | |
84d17192 | 2094 | parent = real_mount(path->mnt); |
9d412a43 | 2095 | err = -EINVAL; |
84d17192 | 2096 | if (unlikely(!check_mnt(parent))) { |
156cacb1 AV |
2097 | /* that's acceptable only for automounts done in private ns */ |
2098 | if (!(mnt_flags & MNT_SHRINKABLE)) | |
2099 | goto unlock; | |
2100 | /* ... and for those we'd better have mountpoint still alive */ | |
84d17192 | 2101 | if (!parent->mnt_ns) |
156cacb1 AV |
2102 | goto unlock; |
2103 | } | |
9d412a43 AV |
2104 | |
2105 | /* Refuse the same filesystem on the same mount point */ | |
2106 | err = -EBUSY; | |
95bc5f25 | 2107 | if (path->mnt->mnt_sb == newmnt->mnt.mnt_sb && |
9d412a43 AV |
2108 | path->mnt->mnt_root == path->dentry) |
2109 | goto unlock; | |
2110 | ||
2111 | err = -EINVAL; | |
95bc5f25 | 2112 | if (S_ISLNK(newmnt->mnt.mnt_root->d_inode->i_mode)) |
9d412a43 AV |
2113 | goto unlock; |
2114 | ||
95bc5f25 | 2115 | newmnt->mnt.mnt_flags = mnt_flags; |
84d17192 | 2116 | err = graft_tree(newmnt, parent, mp); |
9d412a43 AV |
2117 | |
2118 | unlock: | |
84d17192 | 2119 | unlock_mount(mp); |
9d412a43 AV |
2120 | return err; |
2121 | } | |
b1e75df4 | 2122 | |
1da177e4 LT |
2123 | /* |
2124 | * create a new mount for userspace and request it to be added into the | |
2125 | * namespace's tree | |
2126 | */ | |
0c55cfc4 | 2127 | static int do_new_mount(struct path *path, const char *fstype, int flags, |
808d4e3c | 2128 | int mnt_flags, const char *name, void *data) |
1da177e4 | 2129 | { |
0c55cfc4 | 2130 | struct file_system_type *type; |
9b40bc90 | 2131 | struct user_namespace *user_ns = current->nsproxy->mnt_ns->user_ns; |
1da177e4 | 2132 | struct vfsmount *mnt; |
15f9a3f3 | 2133 | int err; |
1da177e4 | 2134 | |
0c55cfc4 | 2135 | if (!fstype) |
1da177e4 LT |
2136 | return -EINVAL; |
2137 | ||
0c55cfc4 EB |
2138 | type = get_fs_type(fstype); |
2139 | if (!type) | |
2140 | return -ENODEV; | |
2141 | ||
2142 | if (user_ns != &init_user_ns) { | |
2143 | if (!(type->fs_flags & FS_USERNS_MOUNT)) { | |
2144 | put_filesystem(type); | |
2145 | return -EPERM; | |
2146 | } | |
2147 | /* Only in special cases allow devices from mounts | |
2148 | * created outside the initial user namespace. | |
2149 | */ | |
2150 | if (!(type->fs_flags & FS_USERNS_DEV_MOUNT)) { | |
2151 | flags |= MS_NODEV; | |
2152 | mnt_flags |= MNT_NODEV; | |
2153 | } | |
2154 | } | |
2155 | ||
2156 | mnt = vfs_kern_mount(type, flags, name, data); | |
2157 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && | |
2158 | !mnt->mnt_sb->s_subtype) | |
2159 | mnt = fs_set_subtype(mnt, fstype); | |
2160 | ||
2161 | put_filesystem(type); | |
1da177e4 LT |
2162 | if (IS_ERR(mnt)) |
2163 | return PTR_ERR(mnt); | |
2164 | ||
95bc5f25 | 2165 | err = do_add_mount(real_mount(mnt), path, mnt_flags); |
15f9a3f3 AV |
2166 | if (err) |
2167 | mntput(mnt); | |
2168 | return err; | |
1da177e4 LT |
2169 | } |
2170 | ||
19a167af AV |
2171 | int finish_automount(struct vfsmount *m, struct path *path) |
2172 | { | |
6776db3d | 2173 | struct mount *mnt = real_mount(m); |
19a167af AV |
2174 | int err; |
2175 | /* The new mount record should have at least 2 refs to prevent it being | |
2176 | * expired before we get a chance to add it | |
2177 | */ | |
6776db3d | 2178 | BUG_ON(mnt_get_count(mnt) < 2); |
19a167af AV |
2179 | |
2180 | if (m->mnt_sb == path->mnt->mnt_sb && | |
2181 | m->mnt_root == path->dentry) { | |
b1e75df4 AV |
2182 | err = -ELOOP; |
2183 | goto fail; | |
19a167af AV |
2184 | } |
2185 | ||
95bc5f25 | 2186 | err = do_add_mount(mnt, path, path->mnt->mnt_flags | MNT_SHRINKABLE); |
b1e75df4 AV |
2187 | if (!err) |
2188 | return 0; | |
2189 | fail: | |
2190 | /* remove m from any expiration list it may be on */ | |
6776db3d | 2191 | if (!list_empty(&mnt->mnt_expire)) { |
97216be0 | 2192 | namespace_lock(); |
962830df | 2193 | br_write_lock(&vfsmount_lock); |
6776db3d | 2194 | list_del_init(&mnt->mnt_expire); |
962830df | 2195 | br_write_unlock(&vfsmount_lock); |
97216be0 | 2196 | namespace_unlock(); |
19a167af | 2197 | } |
b1e75df4 AV |
2198 | mntput(m); |
2199 | mntput(m); | |
19a167af AV |
2200 | return err; |
2201 | } | |
2202 | ||
ea5b778a DH |
2203 | /** |
2204 | * mnt_set_expiry - Put a mount on an expiration list | |
2205 | * @mnt: The mount to list. | |
2206 | * @expiry_list: The list to add the mount to. | |
2207 | */ | |
2208 | void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) | |
2209 | { | |
97216be0 | 2210 | namespace_lock(); |
962830df | 2211 | br_write_lock(&vfsmount_lock); |
ea5b778a | 2212 | |
6776db3d | 2213 | list_add_tail(&real_mount(mnt)->mnt_expire, expiry_list); |
ea5b778a | 2214 | |
962830df | 2215 | br_write_unlock(&vfsmount_lock); |
97216be0 | 2216 | namespace_unlock(); |
ea5b778a DH |
2217 | } |
2218 | EXPORT_SYMBOL(mnt_set_expiry); | |
2219 | ||
1da177e4 LT |
2220 | /* |
2221 | * process a list of expirable mountpoints with the intent of discarding any | |
2222 | * mountpoints that aren't in use and haven't been touched since last we came | |
2223 | * here | |
2224 | */ | |
2225 | void mark_mounts_for_expiry(struct list_head *mounts) | |
2226 | { | |
761d5c38 | 2227 | struct mount *mnt, *next; |
1da177e4 LT |
2228 | LIST_HEAD(graveyard); |
2229 | ||
2230 | if (list_empty(mounts)) | |
2231 | return; | |
2232 | ||
97216be0 | 2233 | namespace_lock(); |
962830df | 2234 | br_write_lock(&vfsmount_lock); |
1da177e4 LT |
2235 | |
2236 | /* extract from the expiration list every vfsmount that matches the | |
2237 | * following criteria: | |
2238 | * - only referenced by its parent vfsmount | |
2239 | * - still marked for expiry (marked on the last call here; marks are | |
2240 | * cleared by mntput()) | |
2241 | */ | |
6776db3d | 2242 | list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { |
863d684f | 2243 | if (!xchg(&mnt->mnt_expiry_mark, 1) || |
1ab59738 | 2244 | propagate_mount_busy(mnt, 1)) |
1da177e4 | 2245 | continue; |
6776db3d | 2246 | list_move(&mnt->mnt_expire, &graveyard); |
1da177e4 | 2247 | } |
bcc5c7d2 | 2248 | while (!list_empty(&graveyard)) { |
6776db3d | 2249 | mnt = list_first_entry(&graveyard, struct mount, mnt_expire); |
143c8c91 | 2250 | touch_mnt_namespace(mnt->mnt_ns); |
328e6d90 | 2251 | umount_tree(mnt, 1); |
bcc5c7d2 | 2252 | } |
962830df | 2253 | br_write_unlock(&vfsmount_lock); |
3ab6abee | 2254 | namespace_unlock(); |
5528f911 TM |
2255 | } |
2256 | ||
2257 | EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); | |
2258 | ||
2259 | /* | |
2260 | * Ripoff of 'select_parent()' | |
2261 | * | |
2262 | * search the list of submounts for a given mountpoint, and move any | |
2263 | * shrinkable submounts to the 'graveyard' list. | |
2264 | */ | |
692afc31 | 2265 | static int select_submounts(struct mount *parent, struct list_head *graveyard) |
5528f911 | 2266 | { |
692afc31 | 2267 | struct mount *this_parent = parent; |
5528f911 TM |
2268 | struct list_head *next; |
2269 | int found = 0; | |
2270 | ||
2271 | repeat: | |
6b41d536 | 2272 | next = this_parent->mnt_mounts.next; |
5528f911 | 2273 | resume: |
6b41d536 | 2274 | while (next != &this_parent->mnt_mounts) { |
5528f911 | 2275 | struct list_head *tmp = next; |
6b41d536 | 2276 | struct mount *mnt = list_entry(tmp, struct mount, mnt_child); |
5528f911 TM |
2277 | |
2278 | next = tmp->next; | |
692afc31 | 2279 | if (!(mnt->mnt.mnt_flags & MNT_SHRINKABLE)) |
1da177e4 | 2280 | continue; |
5528f911 TM |
2281 | /* |
2282 | * Descend a level if the d_mounts list is non-empty. | |
2283 | */ | |
6b41d536 | 2284 | if (!list_empty(&mnt->mnt_mounts)) { |
5528f911 TM |
2285 | this_parent = mnt; |
2286 | goto repeat; | |
2287 | } | |
1da177e4 | 2288 | |
1ab59738 | 2289 | if (!propagate_mount_busy(mnt, 1)) { |
6776db3d | 2290 | list_move_tail(&mnt->mnt_expire, graveyard); |
5528f911 TM |
2291 | found++; |
2292 | } | |
1da177e4 | 2293 | } |
5528f911 TM |
2294 | /* |
2295 | * All done at this level ... ascend and resume the search | |
2296 | */ | |
2297 | if (this_parent != parent) { | |
6b41d536 | 2298 | next = this_parent->mnt_child.next; |
0714a533 | 2299 | this_parent = this_parent->mnt_parent; |
5528f911 TM |
2300 | goto resume; |
2301 | } | |
2302 | return found; | |
2303 | } | |
2304 | ||
2305 | /* | |
2306 | * process a list of expirable mountpoints with the intent of discarding any | |
2307 | * submounts of a specific parent mountpoint | |
99b7db7b NP |
2308 | * |
2309 | * vfsmount_lock must be held for write | |
5528f911 | 2310 | */ |
b54b9be7 | 2311 | static void shrink_submounts(struct mount *mnt) |
5528f911 TM |
2312 | { |
2313 | LIST_HEAD(graveyard); | |
761d5c38 | 2314 | struct mount *m; |
5528f911 | 2315 | |
5528f911 | 2316 | /* extract submounts of 'mountpoint' from the expiration list */ |
c35038be | 2317 | while (select_submounts(mnt, &graveyard)) { |
bcc5c7d2 | 2318 | while (!list_empty(&graveyard)) { |
761d5c38 | 2319 | m = list_first_entry(&graveyard, struct mount, |
6776db3d | 2320 | mnt_expire); |
143c8c91 | 2321 | touch_mnt_namespace(m->mnt_ns); |
328e6d90 | 2322 | umount_tree(m, 1); |
bcc5c7d2 AV |
2323 | } |
2324 | } | |
1da177e4 LT |
2325 | } |
2326 | ||
1da177e4 LT |
2327 | /* |
2328 | * Some copy_from_user() implementations do not return the exact number of | |
2329 | * bytes remaining to copy on a fault. But copy_mount_options() requires that. | |
2330 | * Note that this function differs from copy_from_user() in that it will oops | |
2331 | * on bad values of `to', rather than returning a short copy. | |
2332 | */ | |
b58fed8b RP |
2333 | static long exact_copy_from_user(void *to, const void __user * from, |
2334 | unsigned long n) | |
1da177e4 LT |
2335 | { |
2336 | char *t = to; | |
2337 | const char __user *f = from; | |
2338 | char c; | |
2339 | ||
2340 | if (!access_ok(VERIFY_READ, from, n)) | |
2341 | return n; | |
2342 | ||
2343 | while (n) { | |
2344 | if (__get_user(c, f)) { | |
2345 | memset(t, 0, n); | |
2346 | break; | |
2347 | } | |
2348 | *t++ = c; | |
2349 | f++; | |
2350 | n--; | |
2351 | } | |
2352 | return n; | |
2353 | } | |
2354 | ||
b58fed8b | 2355 | int copy_mount_options(const void __user * data, unsigned long *where) |
1da177e4 LT |
2356 | { |
2357 | int i; | |
2358 | unsigned long page; | |
2359 | unsigned long size; | |
b58fed8b | 2360 | |
1da177e4 LT |
2361 | *where = 0; |
2362 | if (!data) | |
2363 | return 0; | |
2364 | ||
2365 | if (!(page = __get_free_page(GFP_KERNEL))) | |
2366 | return -ENOMEM; | |
2367 | ||
2368 | /* We only care that *some* data at the address the user | |
2369 | * gave us is valid. Just in case, we'll zero | |
2370 | * the remainder of the page. | |
2371 | */ | |
2372 | /* copy_from_user cannot cross TASK_SIZE ! */ | |
2373 | size = TASK_SIZE - (unsigned long)data; | |
2374 | if (size > PAGE_SIZE) | |
2375 | size = PAGE_SIZE; | |
2376 | ||
2377 | i = size - exact_copy_from_user((void *)page, data, size); | |
2378 | if (!i) { | |
b58fed8b | 2379 | free_page(page); |
1da177e4 LT |
2380 | return -EFAULT; |
2381 | } | |
2382 | if (i != PAGE_SIZE) | |
2383 | memset((char *)page + i, 0, PAGE_SIZE - i); | |
2384 | *where = page; | |
2385 | return 0; | |
2386 | } | |
2387 | ||
eca6f534 VN |
2388 | int copy_mount_string(const void __user *data, char **where) |
2389 | { | |
2390 | char *tmp; | |
2391 | ||
2392 | if (!data) { | |
2393 | *where = NULL; | |
2394 | return 0; | |
2395 | } | |
2396 | ||
2397 | tmp = strndup_user(data, PAGE_SIZE); | |
2398 | if (IS_ERR(tmp)) | |
2399 | return PTR_ERR(tmp); | |
2400 | ||
2401 | *where = tmp; | |
2402 | return 0; | |
2403 | } | |
2404 | ||
1da177e4 LT |
2405 | /* |
2406 | * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to | |
2407 | * be given to the mount() call (ie: read-only, no-dev, no-suid etc). | |
2408 | * | |
2409 | * data is a (void *) that can point to any structure up to | |
2410 | * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent | |
2411 | * information (or be NULL). | |
2412 | * | |
2413 | * Pre-0.97 versions of mount() didn't have a flags word. | |
2414 | * When the flags word was introduced its top half was required | |
2415 | * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. | |
2416 | * Therefore, if this magic number is present, it carries no information | |
2417 | * and must be discarded. | |
2418 | */ | |
808d4e3c AV |
2419 | long do_mount(const char *dev_name, const char *dir_name, |
2420 | const char *type_page, unsigned long flags, void *data_page) | |
1da177e4 | 2421 | { |
2d92ab3c | 2422 | struct path path; |
1da177e4 LT |
2423 | int retval = 0; |
2424 | int mnt_flags = 0; | |
2425 | ||
2426 | /* Discard magic */ | |
2427 | if ((flags & MS_MGC_MSK) == MS_MGC_VAL) | |
2428 | flags &= ~MS_MGC_MSK; | |
2429 | ||
2430 | /* Basic sanity checks */ | |
2431 | ||
2432 | if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) | |
2433 | return -EINVAL; | |
1da177e4 LT |
2434 | |
2435 | if (data_page) | |
2436 | ((char *)data_page)[PAGE_SIZE - 1] = 0; | |
2437 | ||
a27ab9f2 TH |
2438 | /* ... and get the mountpoint */ |
2439 | retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); | |
2440 | if (retval) | |
2441 | return retval; | |
2442 | ||
2443 | retval = security_sb_mount(dev_name, &path, | |
2444 | type_page, flags, data_page); | |
0d5cadb8 AV |
2445 | if (!retval && !may_mount()) |
2446 | retval = -EPERM; | |
a27ab9f2 TH |
2447 | if (retval) |
2448 | goto dput_out; | |
2449 | ||
613cbe3d AK |
2450 | /* Default to relatime unless overriden */ |
2451 | if (!(flags & MS_NOATIME)) | |
2452 | mnt_flags |= MNT_RELATIME; | |
0a1c01c9 | 2453 | |
1da177e4 LT |
2454 | /* Separate the per-mountpoint flags */ |
2455 | if (flags & MS_NOSUID) | |
2456 | mnt_flags |= MNT_NOSUID; | |
2457 | if (flags & MS_NODEV) | |
2458 | mnt_flags |= MNT_NODEV; | |
2459 | if (flags & MS_NOEXEC) | |
2460 | mnt_flags |= MNT_NOEXEC; | |
fc33a7bb CH |
2461 | if (flags & MS_NOATIME) |
2462 | mnt_flags |= MNT_NOATIME; | |
2463 | if (flags & MS_NODIRATIME) | |
2464 | mnt_flags |= MNT_NODIRATIME; | |
d0adde57 MG |
2465 | if (flags & MS_STRICTATIME) |
2466 | mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); | |
2e4b7fcd DH |
2467 | if (flags & MS_RDONLY) |
2468 | mnt_flags |= MNT_READONLY; | |
fc33a7bb | 2469 | |
7a4dec53 | 2470 | flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | |
d0adde57 MG |
2471 | MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | |
2472 | MS_STRICTATIME); | |
1da177e4 | 2473 | |
1da177e4 | 2474 | if (flags & MS_REMOUNT) |
2d92ab3c | 2475 | retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, |
1da177e4 LT |
2476 | data_page); |
2477 | else if (flags & MS_BIND) | |
2d92ab3c | 2478 | retval = do_loopback(&path, dev_name, flags & MS_REC); |
9676f0c6 | 2479 | else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) |
2d92ab3c | 2480 | retval = do_change_type(&path, flags); |
1da177e4 | 2481 | else if (flags & MS_MOVE) |
2d92ab3c | 2482 | retval = do_move_mount(&path, dev_name); |
1da177e4 | 2483 | else |
2d92ab3c | 2484 | retval = do_new_mount(&path, type_page, flags, mnt_flags, |
1da177e4 LT |
2485 | dev_name, data_page); |
2486 | dput_out: | |
2d92ab3c | 2487 | path_put(&path); |
1da177e4 LT |
2488 | return retval; |
2489 | } | |
2490 | ||
771b1371 EB |
2491 | static void free_mnt_ns(struct mnt_namespace *ns) |
2492 | { | |
98f842e6 | 2493 | proc_free_inum(ns->proc_inum); |
771b1371 EB |
2494 | put_user_ns(ns->user_ns); |
2495 | kfree(ns); | |
2496 | } | |
2497 | ||
8823c079 EB |
2498 | /* |
2499 | * Assign a sequence number so we can detect when we attempt to bind | |
2500 | * mount a reference to an older mount namespace into the current | |
2501 | * mount namespace, preventing reference counting loops. A 64bit | |
2502 | * number incrementing at 10Ghz will take 12,427 years to wrap which | |
2503 | * is effectively never, so we can ignore the possibility. | |
2504 | */ | |
2505 | static atomic64_t mnt_ns_seq = ATOMIC64_INIT(1); | |
2506 | ||
771b1371 | 2507 | static struct mnt_namespace *alloc_mnt_ns(struct user_namespace *user_ns) |
cf8d2c11 TM |
2508 | { |
2509 | struct mnt_namespace *new_ns; | |
98f842e6 | 2510 | int ret; |
cf8d2c11 TM |
2511 | |
2512 | new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); | |
2513 | if (!new_ns) | |
2514 | return ERR_PTR(-ENOMEM); | |
98f842e6 EB |
2515 | ret = proc_alloc_inum(&new_ns->proc_inum); |
2516 | if (ret) { | |
2517 | kfree(new_ns); | |
2518 | return ERR_PTR(ret); | |
2519 | } | |
8823c079 | 2520 | new_ns->seq = atomic64_add_return(1, &mnt_ns_seq); |
cf8d2c11 TM |
2521 | atomic_set(&new_ns->count, 1); |
2522 | new_ns->root = NULL; | |
2523 | INIT_LIST_HEAD(&new_ns->list); | |
2524 | init_waitqueue_head(&new_ns->poll); | |
2525 | new_ns->event = 0; | |
771b1371 | 2526 | new_ns->user_ns = get_user_ns(user_ns); |
cf8d2c11 TM |
2527 | return new_ns; |
2528 | } | |
2529 | ||
741a2951 JD |
2530 | /* |
2531 | * Allocate a new namespace structure and populate it with contents | |
2532 | * copied from the namespace of the passed in task structure. | |
2533 | */ | |
e3222c4e | 2534 | static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, |
771b1371 | 2535 | struct user_namespace *user_ns, struct fs_struct *fs) |
1da177e4 | 2536 | { |
6b3286ed | 2537 | struct mnt_namespace *new_ns; |
7f2da1e7 | 2538 | struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; |
315fc83e | 2539 | struct mount *p, *q; |
be08d6d2 | 2540 | struct mount *old = mnt_ns->root; |
cb338d06 | 2541 | struct mount *new; |
7a472ef4 | 2542 | int copy_flags; |
1da177e4 | 2543 | |
771b1371 | 2544 | new_ns = alloc_mnt_ns(user_ns); |
cf8d2c11 TM |
2545 | if (IS_ERR(new_ns)) |
2546 | return new_ns; | |
1da177e4 | 2547 | |
97216be0 | 2548 | namespace_lock(); |
1da177e4 | 2549 | /* First pass: copy the tree topology */ |
7a472ef4 EB |
2550 | copy_flags = CL_COPY_ALL | CL_EXPIRE; |
2551 | if (user_ns != mnt_ns->user_ns) | |
132c94e3 | 2552 | copy_flags |= CL_SHARED_TO_SLAVE | CL_UNPRIVILEGED; |
7a472ef4 | 2553 | new = copy_tree(old, old->mnt.mnt_root, copy_flags); |
be34d1a3 | 2554 | if (IS_ERR(new)) { |
328e6d90 | 2555 | namespace_unlock(); |
771b1371 | 2556 | free_mnt_ns(new_ns); |
be34d1a3 | 2557 | return ERR_CAST(new); |
1da177e4 | 2558 | } |
be08d6d2 | 2559 | new_ns->root = new; |
962830df | 2560 | br_write_lock(&vfsmount_lock); |
1a4eeaf2 | 2561 | list_add_tail(&new_ns->list, &new->mnt_list); |
962830df | 2562 | br_write_unlock(&vfsmount_lock); |
1da177e4 LT |
2563 | |
2564 | /* | |
2565 | * Second pass: switch the tsk->fs->* elements and mark new vfsmounts | |
2566 | * as belonging to new namespace. We have already acquired a private | |
2567 | * fs_struct, so tsk->fs->lock is not needed. | |
2568 | */ | |
909b0a88 | 2569 | p = old; |
cb338d06 | 2570 | q = new; |
1da177e4 | 2571 | while (p) { |
143c8c91 | 2572 | q->mnt_ns = new_ns; |
1da177e4 | 2573 | if (fs) { |
315fc83e AV |
2574 | if (&p->mnt == fs->root.mnt) { |
2575 | fs->root.mnt = mntget(&q->mnt); | |
315fc83e | 2576 | rootmnt = &p->mnt; |
1da177e4 | 2577 | } |
315fc83e AV |
2578 | if (&p->mnt == fs->pwd.mnt) { |
2579 | fs->pwd.mnt = mntget(&q->mnt); | |
315fc83e | 2580 | pwdmnt = &p->mnt; |
1da177e4 | 2581 | } |
1da177e4 | 2582 | } |
909b0a88 AV |
2583 | p = next_mnt(p, old); |
2584 | q = next_mnt(q, new); | |
1da177e4 | 2585 | } |
328e6d90 | 2586 | namespace_unlock(); |
1da177e4 | 2587 | |
1da177e4 | 2588 | if (rootmnt) |
f03c6599 | 2589 | mntput(rootmnt); |
1da177e4 | 2590 | if (pwdmnt) |
f03c6599 | 2591 | mntput(pwdmnt); |
1da177e4 | 2592 | |
741a2951 JD |
2593 | return new_ns; |
2594 | } | |
2595 | ||
213dd266 | 2596 | struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, |
771b1371 | 2597 | struct user_namespace *user_ns, struct fs_struct *new_fs) |
741a2951 | 2598 | { |
6b3286ed | 2599 | struct mnt_namespace *new_ns; |
741a2951 | 2600 | |
e3222c4e | 2601 | BUG_ON(!ns); |
6b3286ed | 2602 | get_mnt_ns(ns); |
741a2951 JD |
2603 | |
2604 | if (!(flags & CLONE_NEWNS)) | |
e3222c4e | 2605 | return ns; |
741a2951 | 2606 | |
771b1371 | 2607 | new_ns = dup_mnt_ns(ns, user_ns, new_fs); |
741a2951 | 2608 | |
6b3286ed | 2609 | put_mnt_ns(ns); |
e3222c4e | 2610 | return new_ns; |
1da177e4 LT |
2611 | } |
2612 | ||
cf8d2c11 TM |
2613 | /** |
2614 | * create_mnt_ns - creates a private namespace and adds a root filesystem | |
2615 | * @mnt: pointer to the new root filesystem mountpoint | |
2616 | */ | |
1a4eeaf2 | 2617 | static struct mnt_namespace *create_mnt_ns(struct vfsmount *m) |
cf8d2c11 | 2618 | { |
771b1371 | 2619 | struct mnt_namespace *new_ns = alloc_mnt_ns(&init_user_ns); |
cf8d2c11 | 2620 | if (!IS_ERR(new_ns)) { |
1a4eeaf2 AV |
2621 | struct mount *mnt = real_mount(m); |
2622 | mnt->mnt_ns = new_ns; | |
be08d6d2 | 2623 | new_ns->root = mnt; |
b1983cd8 | 2624 | list_add(&mnt->mnt_list, &new_ns->list); |
c1334495 | 2625 | } else { |
1a4eeaf2 | 2626 | mntput(m); |
cf8d2c11 TM |
2627 | } |
2628 | return new_ns; | |
2629 | } | |
cf8d2c11 | 2630 | |
ea441d11 AV |
2631 | struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) |
2632 | { | |
2633 | struct mnt_namespace *ns; | |
d31da0f0 | 2634 | struct super_block *s; |
ea441d11 AV |
2635 | struct path path; |
2636 | int err; | |
2637 | ||
2638 | ns = create_mnt_ns(mnt); | |
2639 | if (IS_ERR(ns)) | |
2640 | return ERR_CAST(ns); | |
2641 | ||
2642 | err = vfs_path_lookup(mnt->mnt_root, mnt, | |
2643 | name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); | |
2644 | ||
2645 | put_mnt_ns(ns); | |
2646 | ||
2647 | if (err) | |
2648 | return ERR_PTR(err); | |
2649 | ||
2650 | /* trade a vfsmount reference for active sb one */ | |
d31da0f0 AV |
2651 | s = path.mnt->mnt_sb; |
2652 | atomic_inc(&s->s_active); | |
ea441d11 AV |
2653 | mntput(path.mnt); |
2654 | /* lock the sucker */ | |
d31da0f0 | 2655 | down_write(&s->s_umount); |
ea441d11 AV |
2656 | /* ... and return the root of (sub)tree on it */ |
2657 | return path.dentry; | |
2658 | } | |
2659 | EXPORT_SYMBOL(mount_subtree); | |
2660 | ||
bdc480e3 HC |
2661 | SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, |
2662 | char __user *, type, unsigned long, flags, void __user *, data) | |
1da177e4 | 2663 | { |
eca6f534 VN |
2664 | int ret; |
2665 | char *kernel_type; | |
91a27b2a | 2666 | struct filename *kernel_dir; |
eca6f534 | 2667 | char *kernel_dev; |
1da177e4 | 2668 | unsigned long data_page; |
1da177e4 | 2669 | |
eca6f534 VN |
2670 | ret = copy_mount_string(type, &kernel_type); |
2671 | if (ret < 0) | |
2672 | goto out_type; | |
1da177e4 | 2673 | |
eca6f534 VN |
2674 | kernel_dir = getname(dir_name); |
2675 | if (IS_ERR(kernel_dir)) { | |
2676 | ret = PTR_ERR(kernel_dir); | |
2677 | goto out_dir; | |
2678 | } | |
1da177e4 | 2679 | |
eca6f534 VN |
2680 | ret = copy_mount_string(dev_name, &kernel_dev); |
2681 | if (ret < 0) | |
2682 | goto out_dev; | |
1da177e4 | 2683 | |
eca6f534 VN |
2684 | ret = copy_mount_options(data, &data_page); |
2685 | if (ret < 0) | |
2686 | goto out_data; | |
1da177e4 | 2687 | |
91a27b2a | 2688 | ret = do_mount(kernel_dev, kernel_dir->name, kernel_type, flags, |
eca6f534 | 2689 | (void *) data_page); |
1da177e4 | 2690 | |
eca6f534 VN |
2691 | free_page(data_page); |
2692 | out_data: | |
2693 | kfree(kernel_dev); | |
2694 | out_dev: | |
2695 | putname(kernel_dir); | |
2696 | out_dir: | |
2697 | kfree(kernel_type); | |
2698 | out_type: | |
2699 | return ret; | |
1da177e4 LT |
2700 | } |
2701 | ||
afac7cba AV |
2702 | /* |
2703 | * Return true if path is reachable from root | |
2704 | * | |
2705 | * namespace_sem or vfsmount_lock is held | |
2706 | */ | |
643822b4 | 2707 | bool is_path_reachable(struct mount *mnt, struct dentry *dentry, |
afac7cba AV |
2708 | const struct path *root) |
2709 | { | |
643822b4 | 2710 | while (&mnt->mnt != root->mnt && mnt_has_parent(mnt)) { |
a73324da | 2711 | dentry = mnt->mnt_mountpoint; |
0714a533 | 2712 | mnt = mnt->mnt_parent; |
afac7cba | 2713 | } |
643822b4 | 2714 | return &mnt->mnt == root->mnt && is_subdir(dentry, root->dentry); |
afac7cba AV |
2715 | } |
2716 | ||
2717 | int path_is_under(struct path *path1, struct path *path2) | |
2718 | { | |
2719 | int res; | |
962830df | 2720 | br_read_lock(&vfsmount_lock); |
643822b4 | 2721 | res = is_path_reachable(real_mount(path1->mnt), path1->dentry, path2); |
962830df | 2722 | br_read_unlock(&vfsmount_lock); |
afac7cba AV |
2723 | return res; |
2724 | } | |
2725 | EXPORT_SYMBOL(path_is_under); | |
2726 | ||
1da177e4 LT |
2727 | /* |
2728 | * pivot_root Semantics: | |
2729 | * Moves the root file system of the current process to the directory put_old, | |
2730 | * makes new_root as the new root file system of the current process, and sets | |
2731 | * root/cwd of all processes which had them on the current root to new_root. | |
2732 | * | |
2733 | * Restrictions: | |
2734 | * The new_root and put_old must be directories, and must not be on the | |
2735 | * same file system as the current process root. The put_old must be | |
2736 | * underneath new_root, i.e. adding a non-zero number of /.. to the string | |
2737 | * pointed to by put_old must yield the same directory as new_root. No other | |
2738 | * file system may be mounted on put_old. After all, new_root is a mountpoint. | |
2739 | * | |
4a0d11fa NB |
2740 | * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. |
2741 | * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives | |
2742 | * in this situation. | |
2743 | * | |
1da177e4 LT |
2744 | * Notes: |
2745 | * - we don't move root/cwd if they are not at the root (reason: if something | |
2746 | * cared enough to change them, it's probably wrong to force them elsewhere) | |
2747 | * - it's okay to pick a root that isn't the root of a file system, e.g. | |
2748 | * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, | |
2749 | * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root | |
2750 | * first. | |
2751 | */ | |
3480b257 HC |
2752 | SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, |
2753 | const char __user *, put_old) | |
1da177e4 | 2754 | { |
2d8f3038 | 2755 | struct path new, old, parent_path, root_parent, root; |
84d17192 AV |
2756 | struct mount *new_mnt, *root_mnt, *old_mnt; |
2757 | struct mountpoint *old_mp, *root_mp; | |
1da177e4 LT |
2758 | int error; |
2759 | ||
9b40bc90 | 2760 | if (!may_mount()) |
1da177e4 LT |
2761 | return -EPERM; |
2762 | ||
2d8f3038 | 2763 | error = user_path_dir(new_root, &new); |
1da177e4 LT |
2764 | if (error) |
2765 | goto out0; | |
1da177e4 | 2766 | |
2d8f3038 | 2767 | error = user_path_dir(put_old, &old); |
1da177e4 LT |
2768 | if (error) |
2769 | goto out1; | |
2770 | ||
2d8f3038 | 2771 | error = security_sb_pivotroot(&old, &new); |
b12cea91 AV |
2772 | if (error) |
2773 | goto out2; | |
1da177e4 | 2774 | |
f7ad3c6b | 2775 | get_fs_root(current->fs, &root); |
84d17192 AV |
2776 | old_mp = lock_mount(&old); |
2777 | error = PTR_ERR(old_mp); | |
2778 | if (IS_ERR(old_mp)) | |
b12cea91 AV |
2779 | goto out3; |
2780 | ||
1da177e4 | 2781 | error = -EINVAL; |
419148da AV |
2782 | new_mnt = real_mount(new.mnt); |
2783 | root_mnt = real_mount(root.mnt); | |
84d17192 AV |
2784 | old_mnt = real_mount(old.mnt); |
2785 | if (IS_MNT_SHARED(old_mnt) || | |
fc7be130 AV |
2786 | IS_MNT_SHARED(new_mnt->mnt_parent) || |
2787 | IS_MNT_SHARED(root_mnt->mnt_parent)) | |
b12cea91 | 2788 | goto out4; |
143c8c91 | 2789 | if (!check_mnt(root_mnt) || !check_mnt(new_mnt)) |
b12cea91 | 2790 | goto out4; |
1da177e4 | 2791 | error = -ENOENT; |
f3da392e | 2792 | if (d_unlinked(new.dentry)) |
b12cea91 | 2793 | goto out4; |
1da177e4 | 2794 | error = -EBUSY; |
84d17192 | 2795 | if (new_mnt == root_mnt || old_mnt == root_mnt) |
b12cea91 | 2796 | goto out4; /* loop, on the same file system */ |
1da177e4 | 2797 | error = -EINVAL; |
8c3ee42e | 2798 | if (root.mnt->mnt_root != root.dentry) |
b12cea91 | 2799 | goto out4; /* not a mountpoint */ |
676da58d | 2800 | if (!mnt_has_parent(root_mnt)) |
b12cea91 | 2801 | goto out4; /* not attached */ |
84d17192 | 2802 | root_mp = root_mnt->mnt_mp; |
2d8f3038 | 2803 | if (new.mnt->mnt_root != new.dentry) |
b12cea91 | 2804 | goto out4; /* not a mountpoint */ |
676da58d | 2805 | if (!mnt_has_parent(new_mnt)) |
b12cea91 | 2806 | goto out4; /* not attached */ |
4ac91378 | 2807 | /* make sure we can reach put_old from new_root */ |
84d17192 | 2808 | if (!is_path_reachable(old_mnt, old.dentry, &new)) |
b12cea91 | 2809 | goto out4; |
84d17192 | 2810 | root_mp->m_count++; /* pin it so it won't go away */ |
962830df | 2811 | br_write_lock(&vfsmount_lock); |
419148da AV |
2812 | detach_mnt(new_mnt, &parent_path); |
2813 | detach_mnt(root_mnt, &root_parent); | |
4ac91378 | 2814 | /* mount old root on put_old */ |
84d17192 | 2815 | attach_mnt(root_mnt, old_mnt, old_mp); |
4ac91378 | 2816 | /* mount new_root on / */ |
84d17192 | 2817 | attach_mnt(new_mnt, real_mount(root_parent.mnt), root_mp); |
6b3286ed | 2818 | touch_mnt_namespace(current->nsproxy->mnt_ns); |
962830df | 2819 | br_write_unlock(&vfsmount_lock); |
2d8f3038 | 2820 | chroot_fs_refs(&root, &new); |
84d17192 | 2821 | put_mountpoint(root_mp); |
1da177e4 | 2822 | error = 0; |
b12cea91 | 2823 | out4: |
84d17192 | 2824 | unlock_mount(old_mp); |
b12cea91 AV |
2825 | if (!error) { |
2826 | path_put(&root_parent); | |
2827 | path_put(&parent_path); | |
2828 | } | |
2829 | out3: | |
8c3ee42e | 2830 | path_put(&root); |
b12cea91 | 2831 | out2: |
2d8f3038 | 2832 | path_put(&old); |
1da177e4 | 2833 | out1: |
2d8f3038 | 2834 | path_put(&new); |
1da177e4 | 2835 | out0: |
1da177e4 | 2836 | return error; |
1da177e4 LT |
2837 | } |
2838 | ||
2839 | static void __init init_mount_tree(void) | |
2840 | { | |
2841 | struct vfsmount *mnt; | |
6b3286ed | 2842 | struct mnt_namespace *ns; |
ac748a09 | 2843 | struct path root; |
0c55cfc4 | 2844 | struct file_system_type *type; |
1da177e4 | 2845 | |
0c55cfc4 EB |
2846 | type = get_fs_type("rootfs"); |
2847 | if (!type) | |
2848 | panic("Can't find rootfs type"); | |
2849 | mnt = vfs_kern_mount(type, 0, "rootfs", NULL); | |
2850 | put_filesystem(type); | |
1da177e4 LT |
2851 | if (IS_ERR(mnt)) |
2852 | panic("Can't create rootfs"); | |
b3e19d92 | 2853 | |
3b22edc5 TM |
2854 | ns = create_mnt_ns(mnt); |
2855 | if (IS_ERR(ns)) | |
1da177e4 | 2856 | panic("Can't allocate initial namespace"); |
6b3286ed KK |
2857 | |
2858 | init_task.nsproxy->mnt_ns = ns; | |
2859 | get_mnt_ns(ns); | |
2860 | ||
be08d6d2 AV |
2861 | root.mnt = mnt; |
2862 | root.dentry = mnt->mnt_root; | |
ac748a09 JB |
2863 | |
2864 | set_fs_pwd(current->fs, &root); | |
2865 | set_fs_root(current->fs, &root); | |
1da177e4 LT |
2866 | } |
2867 | ||
74bf17cf | 2868 | void __init mnt_init(void) |
1da177e4 | 2869 | { |
13f14b4d | 2870 | unsigned u; |
15a67dd8 | 2871 | int err; |
1da177e4 | 2872 | |
390c6843 RP |
2873 | init_rwsem(&namespace_sem); |
2874 | ||
7d6fec45 | 2875 | mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount), |
20c2df83 | 2876 | 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); |
1da177e4 | 2877 | |
b58fed8b | 2878 | mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
84d17192 | 2879 | mountpoint_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
1da177e4 | 2880 | |
84d17192 | 2881 | if (!mount_hashtable || !mountpoint_hashtable) |
1da177e4 LT |
2882 | panic("Failed to allocate mount hash table\n"); |
2883 | ||
80cdc6da | 2884 | printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); |
13f14b4d ED |
2885 | |
2886 | for (u = 0; u < HASH_SIZE; u++) | |
2887 | INIT_LIST_HEAD(&mount_hashtable[u]); | |
84d17192 AV |
2888 | for (u = 0; u < HASH_SIZE; u++) |
2889 | INIT_LIST_HEAD(&mountpoint_hashtable[u]); | |
1da177e4 | 2890 | |
962830df | 2891 | br_lock_init(&vfsmount_lock); |
99b7db7b | 2892 | |
15a67dd8 RD |
2893 | err = sysfs_init(); |
2894 | if (err) | |
2895 | printk(KERN_WARNING "%s: sysfs_init error: %d\n", | |
8e24eea7 | 2896 | __func__, err); |
00d26666 GKH |
2897 | fs_kobj = kobject_create_and_add("fs", NULL); |
2898 | if (!fs_kobj) | |
8e24eea7 | 2899 | printk(KERN_WARNING "%s: kobj create error\n", __func__); |
1da177e4 LT |
2900 | init_rootfs(); |
2901 | init_mount_tree(); | |
2902 | } | |
2903 | ||
616511d0 | 2904 | void put_mnt_ns(struct mnt_namespace *ns) |
1da177e4 | 2905 | { |
d498b25a | 2906 | if (!atomic_dec_and_test(&ns->count)) |
616511d0 | 2907 | return; |
97216be0 | 2908 | namespace_lock(); |
962830df | 2909 | br_write_lock(&vfsmount_lock); |
328e6d90 | 2910 | umount_tree(ns->root, 0); |
962830df | 2911 | br_write_unlock(&vfsmount_lock); |
3ab6abee | 2912 | namespace_unlock(); |
771b1371 | 2913 | free_mnt_ns(ns); |
1da177e4 | 2914 | } |
9d412a43 AV |
2915 | |
2916 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) | |
2917 | { | |
423e0ab0 TC |
2918 | struct vfsmount *mnt; |
2919 | mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); | |
2920 | if (!IS_ERR(mnt)) { | |
2921 | /* | |
2922 | * it is a longterm mount, don't release mnt until | |
2923 | * we unmount before file sys is unregistered | |
2924 | */ | |
f7a99c5b | 2925 | real_mount(mnt)->mnt_ns = MNT_NS_INTERNAL; |
423e0ab0 TC |
2926 | } |
2927 | return mnt; | |
9d412a43 AV |
2928 | } |
2929 | EXPORT_SYMBOL_GPL(kern_mount_data); | |
423e0ab0 TC |
2930 | |
2931 | void kern_unmount(struct vfsmount *mnt) | |
2932 | { | |
2933 | /* release long term mount so mount point can be released */ | |
2934 | if (!IS_ERR_OR_NULL(mnt)) { | |
f7a99c5b AV |
2935 | br_write_lock(&vfsmount_lock); |
2936 | real_mount(mnt)->mnt_ns = NULL; | |
2937 | br_write_unlock(&vfsmount_lock); | |
423e0ab0 TC |
2938 | mntput(mnt); |
2939 | } | |
2940 | } | |
2941 | EXPORT_SYMBOL(kern_unmount); | |
02125a82 AV |
2942 | |
2943 | bool our_mnt(struct vfsmount *mnt) | |
2944 | { | |
143c8c91 | 2945 | return check_mnt(real_mount(mnt)); |
02125a82 | 2946 | } |
8823c079 | 2947 | |
3151527e EB |
2948 | bool current_chrooted(void) |
2949 | { | |
2950 | /* Does the current process have a non-standard root */ | |
2951 | struct path ns_root; | |
2952 | struct path fs_root; | |
2953 | bool chrooted; | |
2954 | ||
2955 | /* Find the namespace root */ | |
2956 | ns_root.mnt = ¤t->nsproxy->mnt_ns->root->mnt; | |
2957 | ns_root.dentry = ns_root.mnt->mnt_root; | |
2958 | path_get(&ns_root); | |
2959 | while (d_mountpoint(ns_root.dentry) && follow_down_one(&ns_root)) | |
2960 | ; | |
2961 | ||
2962 | get_fs_root(current->fs, &fs_root); | |
2963 | ||
2964 | chrooted = !path_equal(&fs_root, &ns_root); | |
2965 | ||
2966 | path_put(&fs_root); | |
2967 | path_put(&ns_root); | |
2968 | ||
2969 | return chrooted; | |
2970 | } | |
2971 | ||
87a8ebd6 EB |
2972 | void update_mnt_policy(struct user_namespace *userns) |
2973 | { | |
2974 | struct mnt_namespace *ns = current->nsproxy->mnt_ns; | |
2975 | struct mount *mnt; | |
2976 | ||
2977 | down_read(&namespace_sem); | |
2978 | list_for_each_entry(mnt, &ns->list, mnt_list) { | |
2979 | switch (mnt->mnt.mnt_sb->s_magic) { | |
2980 | case SYSFS_MAGIC: | |
2981 | userns->may_mount_sysfs = true; | |
2982 | break; | |
2983 | case PROC_SUPER_MAGIC: | |
2984 | userns->may_mount_proc = true; | |
2985 | break; | |
2986 | } | |
2987 | if (userns->may_mount_sysfs && userns->may_mount_proc) | |
2988 | break; | |
2989 | } | |
2990 | up_read(&namespace_sem); | |
2991 | } | |
2992 | ||
8823c079 EB |
2993 | static void *mntns_get(struct task_struct *task) |
2994 | { | |
2995 | struct mnt_namespace *ns = NULL; | |
2996 | struct nsproxy *nsproxy; | |
2997 | ||
2998 | rcu_read_lock(); | |
2999 | nsproxy = task_nsproxy(task); | |
3000 | if (nsproxy) { | |
3001 | ns = nsproxy->mnt_ns; | |
3002 | get_mnt_ns(ns); | |
3003 | } | |
3004 | rcu_read_unlock(); | |
3005 | ||
3006 | return ns; | |
3007 | } | |
3008 | ||
3009 | static void mntns_put(void *ns) | |
3010 | { | |
3011 | put_mnt_ns(ns); | |
3012 | } | |
3013 | ||
3014 | static int mntns_install(struct nsproxy *nsproxy, void *ns) | |
3015 | { | |
3016 | struct fs_struct *fs = current->fs; | |
3017 | struct mnt_namespace *mnt_ns = ns; | |
3018 | struct path root; | |
3019 | ||
0c55cfc4 | 3020 | if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) || |
5e4a0847 EB |
3021 | !nsown_capable(CAP_SYS_CHROOT) || |
3022 | !nsown_capable(CAP_SYS_ADMIN)) | |
ae11e0f1 | 3023 | return -EPERM; |
8823c079 EB |
3024 | |
3025 | if (fs->users != 1) | |
3026 | return -EINVAL; | |
3027 | ||
3028 | get_mnt_ns(mnt_ns); | |
3029 | put_mnt_ns(nsproxy->mnt_ns); | |
3030 | nsproxy->mnt_ns = mnt_ns; | |
3031 | ||
3032 | /* Find the root */ | |
3033 | root.mnt = &mnt_ns->root->mnt; | |
3034 | root.dentry = mnt_ns->root->mnt.mnt_root; | |
3035 | path_get(&root); | |
3036 | while(d_mountpoint(root.dentry) && follow_down_one(&root)) | |
3037 | ; | |
3038 | ||
3039 | /* Update the pwd and root */ | |
3040 | set_fs_pwd(fs, &root); | |
3041 | set_fs_root(fs, &root); | |
3042 | ||
3043 | path_put(&root); | |
3044 | return 0; | |
3045 | } | |
3046 | ||
98f842e6 EB |
3047 | static unsigned int mntns_inum(void *ns) |
3048 | { | |
3049 | struct mnt_namespace *mnt_ns = ns; | |
3050 | return mnt_ns->proc_inum; | |
3051 | } | |
3052 | ||
8823c079 EB |
3053 | const struct proc_ns_operations mntns_operations = { |
3054 | .name = "mnt", | |
3055 | .type = CLONE_NEWNS, | |
3056 | .get = mntns_get, | |
3057 | .put = mntns_put, | |
3058 | .install = mntns_install, | |
98f842e6 | 3059 | .inum = mntns_inum, |
8823c079 | 3060 | }; |