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 | 960 | /* Don't allow unprivileged users to change mount flags */ |
187985d9 EB |
961 | if (flag & CL_UNPRIVILEGED) { |
962 | mnt->mnt.mnt_flags |= MNT_LOCK_ATIME; | |
963 | ||
964 | if (mnt->mnt.mnt_flags & MNT_READONLY) | |
965 | mnt->mnt.mnt_flags |= MNT_LOCK_READONLY; | |
966 | ||
967 | if (mnt->mnt.mnt_flags & MNT_NODEV) | |
968 | mnt->mnt.mnt_flags |= MNT_LOCK_NODEV; | |
969 | ||
970 | if (mnt->mnt.mnt_flags & MNT_NOSUID) | |
971 | mnt->mnt.mnt_flags |= MNT_LOCK_NOSUID; | |
972 | ||
973 | if (mnt->mnt.mnt_flags & MNT_NOEXEC) | |
974 | mnt->mnt.mnt_flags |= MNT_LOCK_NOEXEC; | |
975 | } | |
132c94e3 | 976 | |
be34d1a3 DH |
977 | atomic_inc(&sb->s_active); |
978 | mnt->mnt.mnt_sb = sb; | |
979 | mnt->mnt.mnt_root = dget(root); | |
980 | mnt->mnt_mountpoint = mnt->mnt.mnt_root; | |
981 | mnt->mnt_parent = mnt; | |
982 | br_write_lock(&vfsmount_lock); | |
983 | list_add_tail(&mnt->mnt_instance, &sb->s_mounts); | |
984 | br_write_unlock(&vfsmount_lock); | |
985 | ||
7a472ef4 EB |
986 | if ((flag & CL_SLAVE) || |
987 | ((flag & CL_SHARED_TO_SLAVE) && IS_MNT_SHARED(old))) { | |
be34d1a3 DH |
988 | list_add(&mnt->mnt_slave, &old->mnt_slave_list); |
989 | mnt->mnt_master = old; | |
990 | CLEAR_MNT_SHARED(mnt); | |
991 | } else if (!(flag & CL_PRIVATE)) { | |
992 | if ((flag & CL_MAKE_SHARED) || IS_MNT_SHARED(old)) | |
993 | list_add(&mnt->mnt_share, &old->mnt_share); | |
994 | if (IS_MNT_SLAVE(old)) | |
995 | list_add(&mnt->mnt_slave, &old->mnt_slave); | |
996 | mnt->mnt_master = old->mnt_master; | |
997 | } | |
998 | if (flag & CL_MAKE_SHARED) | |
999 | set_mnt_shared(mnt); | |
1000 | ||
1001 | /* stick the duplicate mount on the same expiry list | |
1002 | * as the original if that was on one */ | |
1003 | if (flag & CL_EXPIRE) { | |
1004 | if (!list_empty(&old->mnt_expire)) | |
1005 | list_add(&mnt->mnt_expire, &old->mnt_expire); | |
1006 | } | |
1007 | ||
cb338d06 | 1008 | return mnt; |
719f5d7f MS |
1009 | |
1010 | out_free: | |
1011 | free_vfsmnt(mnt); | |
be34d1a3 | 1012 | return ERR_PTR(err); |
1da177e4 LT |
1013 | } |
1014 | ||
83adc753 | 1015 | static inline void mntfree(struct mount *mnt) |
1da177e4 | 1016 | { |
83adc753 AV |
1017 | struct vfsmount *m = &mnt->mnt; |
1018 | struct super_block *sb = m->mnt_sb; | |
b3e19d92 | 1019 | |
3d733633 DH |
1020 | /* |
1021 | * This probably indicates that somebody messed | |
1022 | * up a mnt_want/drop_write() pair. If this | |
1023 | * happens, the filesystem was probably unable | |
1024 | * to make r/w->r/o transitions. | |
1025 | */ | |
d3ef3d73 | 1026 | /* |
b3e19d92 NP |
1027 | * The locking used to deal with mnt_count decrement provides barriers, |
1028 | * so mnt_get_writers() below is safe. | |
d3ef3d73 | 1029 | */ |
c6653a83 | 1030 | WARN_ON(mnt_get_writers(mnt)); |
83adc753 AV |
1031 | fsnotify_vfsmount_delete(m); |
1032 | dput(m->mnt_root); | |
1033 | free_vfsmnt(mnt); | |
1da177e4 LT |
1034 | deactivate_super(sb); |
1035 | } | |
1036 | ||
900148dc | 1037 | static void mntput_no_expire(struct mount *mnt) |
b3e19d92 | 1038 | { |
b3e19d92 | 1039 | put_again: |
f03c6599 | 1040 | #ifdef CONFIG_SMP |
962830df | 1041 | br_read_lock(&vfsmount_lock); |
f7a99c5b AV |
1042 | if (likely(mnt->mnt_ns)) { |
1043 | /* shouldn't be the last one */ | |
aa9c0e07 | 1044 | mnt_add_count(mnt, -1); |
962830df | 1045 | br_read_unlock(&vfsmount_lock); |
f03c6599 | 1046 | return; |
b3e19d92 | 1047 | } |
962830df | 1048 | br_read_unlock(&vfsmount_lock); |
b3e19d92 | 1049 | |
962830df | 1050 | br_write_lock(&vfsmount_lock); |
aa9c0e07 | 1051 | mnt_add_count(mnt, -1); |
b3e19d92 | 1052 | if (mnt_get_count(mnt)) { |
962830df | 1053 | br_write_unlock(&vfsmount_lock); |
99b7db7b NP |
1054 | return; |
1055 | } | |
b3e19d92 | 1056 | #else |
aa9c0e07 | 1057 | mnt_add_count(mnt, -1); |
b3e19d92 | 1058 | if (likely(mnt_get_count(mnt))) |
99b7db7b | 1059 | return; |
962830df | 1060 | br_write_lock(&vfsmount_lock); |
f03c6599 | 1061 | #endif |
863d684f AV |
1062 | if (unlikely(mnt->mnt_pinned)) { |
1063 | mnt_add_count(mnt, mnt->mnt_pinned + 1); | |
1064 | mnt->mnt_pinned = 0; | |
962830df | 1065 | br_write_unlock(&vfsmount_lock); |
900148dc | 1066 | acct_auto_close_mnt(&mnt->mnt); |
b3e19d92 | 1067 | goto put_again; |
7b7b1ace | 1068 | } |
962830df | 1069 | |
39f7c4db | 1070 | list_del(&mnt->mnt_instance); |
962830df | 1071 | br_write_unlock(&vfsmount_lock); |
b3e19d92 NP |
1072 | mntfree(mnt); |
1073 | } | |
b3e19d92 NP |
1074 | |
1075 | void mntput(struct vfsmount *mnt) | |
1076 | { | |
1077 | if (mnt) { | |
863d684f | 1078 | struct mount *m = real_mount(mnt); |
b3e19d92 | 1079 | /* avoid cacheline pingpong, hope gcc doesn't get "smart" */ |
863d684f AV |
1080 | if (unlikely(m->mnt_expiry_mark)) |
1081 | m->mnt_expiry_mark = 0; | |
1082 | mntput_no_expire(m); | |
b3e19d92 NP |
1083 | } |
1084 | } | |
1085 | EXPORT_SYMBOL(mntput); | |
1086 | ||
1087 | struct vfsmount *mntget(struct vfsmount *mnt) | |
1088 | { | |
1089 | if (mnt) | |
83adc753 | 1090 | mnt_add_count(real_mount(mnt), 1); |
b3e19d92 NP |
1091 | return mnt; |
1092 | } | |
1093 | EXPORT_SYMBOL(mntget); | |
1094 | ||
7b7b1ace AV |
1095 | void mnt_pin(struct vfsmount *mnt) |
1096 | { | |
962830df | 1097 | br_write_lock(&vfsmount_lock); |
863d684f | 1098 | real_mount(mnt)->mnt_pinned++; |
962830df | 1099 | br_write_unlock(&vfsmount_lock); |
7b7b1ace | 1100 | } |
7b7b1ace AV |
1101 | EXPORT_SYMBOL(mnt_pin); |
1102 | ||
863d684f | 1103 | void mnt_unpin(struct vfsmount *m) |
7b7b1ace | 1104 | { |
863d684f | 1105 | struct mount *mnt = real_mount(m); |
962830df | 1106 | br_write_lock(&vfsmount_lock); |
7b7b1ace | 1107 | if (mnt->mnt_pinned) { |
863d684f | 1108 | mnt_add_count(mnt, 1); |
7b7b1ace AV |
1109 | mnt->mnt_pinned--; |
1110 | } | |
962830df | 1111 | br_write_unlock(&vfsmount_lock); |
7b7b1ace | 1112 | } |
7b7b1ace | 1113 | EXPORT_SYMBOL(mnt_unpin); |
1da177e4 | 1114 | |
b3b304a2 MS |
1115 | static inline void mangle(struct seq_file *m, const char *s) |
1116 | { | |
1117 | seq_escape(m, s, " \t\n\\"); | |
1118 | } | |
1119 | ||
1120 | /* | |
1121 | * Simple .show_options callback for filesystems which don't want to | |
1122 | * implement more complex mount option showing. | |
1123 | * | |
1124 | * See also save_mount_options(). | |
1125 | */ | |
34c80b1d | 1126 | int generic_show_options(struct seq_file *m, struct dentry *root) |
b3b304a2 | 1127 | { |
2a32cebd AV |
1128 | const char *options; |
1129 | ||
1130 | rcu_read_lock(); | |
34c80b1d | 1131 | options = rcu_dereference(root->d_sb->s_options); |
b3b304a2 MS |
1132 | |
1133 | if (options != NULL && options[0]) { | |
1134 | seq_putc(m, ','); | |
1135 | mangle(m, options); | |
1136 | } | |
2a32cebd | 1137 | rcu_read_unlock(); |
b3b304a2 MS |
1138 | |
1139 | return 0; | |
1140 | } | |
1141 | EXPORT_SYMBOL(generic_show_options); | |
1142 | ||
1143 | /* | |
1144 | * If filesystem uses generic_show_options(), this function should be | |
1145 | * called from the fill_super() callback. | |
1146 | * | |
1147 | * The .remount_fs callback usually needs to be handled in a special | |
1148 | * way, to make sure, that previous options are not overwritten if the | |
1149 | * remount fails. | |
1150 | * | |
1151 | * Also note, that if the filesystem's .remount_fs function doesn't | |
1152 | * reset all options to their default value, but changes only newly | |
1153 | * given options, then the displayed options will not reflect reality | |
1154 | * any more. | |
1155 | */ | |
1156 | void save_mount_options(struct super_block *sb, char *options) | |
1157 | { | |
2a32cebd AV |
1158 | BUG_ON(sb->s_options); |
1159 | rcu_assign_pointer(sb->s_options, kstrdup(options, GFP_KERNEL)); | |
b3b304a2 MS |
1160 | } |
1161 | EXPORT_SYMBOL(save_mount_options); | |
1162 | ||
2a32cebd AV |
1163 | void replace_mount_options(struct super_block *sb, char *options) |
1164 | { | |
1165 | char *old = sb->s_options; | |
1166 | rcu_assign_pointer(sb->s_options, options); | |
1167 | if (old) { | |
1168 | synchronize_rcu(); | |
1169 | kfree(old); | |
1170 | } | |
1171 | } | |
1172 | EXPORT_SYMBOL(replace_mount_options); | |
1173 | ||
a1a2c409 | 1174 | #ifdef CONFIG_PROC_FS |
0226f492 | 1175 | /* iterator; we want it to have access to namespace_sem, thus here... */ |
1da177e4 LT |
1176 | static void *m_start(struct seq_file *m, loff_t *pos) |
1177 | { | |
6ce6e24e | 1178 | struct proc_mounts *p = proc_mounts(m); |
1da177e4 | 1179 | |
390c6843 | 1180 | down_read(&namespace_sem); |
a1a2c409 | 1181 | return seq_list_start(&p->ns->list, *pos); |
1da177e4 LT |
1182 | } |
1183 | ||
1184 | static void *m_next(struct seq_file *m, void *v, loff_t *pos) | |
1185 | { | |
6ce6e24e | 1186 | struct proc_mounts *p = proc_mounts(m); |
b0765fb8 | 1187 | |
a1a2c409 | 1188 | return seq_list_next(v, &p->ns->list, pos); |
1da177e4 LT |
1189 | } |
1190 | ||
1191 | static void m_stop(struct seq_file *m, void *v) | |
1192 | { | |
390c6843 | 1193 | up_read(&namespace_sem); |
1da177e4 LT |
1194 | } |
1195 | ||
0226f492 | 1196 | static int m_show(struct seq_file *m, void *v) |
2d4d4864 | 1197 | { |
6ce6e24e | 1198 | struct proc_mounts *p = proc_mounts(m); |
1a4eeaf2 | 1199 | struct mount *r = list_entry(v, struct mount, mnt_list); |
0226f492 | 1200 | return p->show(m, &r->mnt); |
1da177e4 LT |
1201 | } |
1202 | ||
a1a2c409 | 1203 | const struct seq_operations mounts_op = { |
1da177e4 LT |
1204 | .start = m_start, |
1205 | .next = m_next, | |
1206 | .stop = m_stop, | |
0226f492 | 1207 | .show = m_show, |
b4629fe2 | 1208 | }; |
a1a2c409 | 1209 | #endif /* CONFIG_PROC_FS */ |
b4629fe2 | 1210 | |
1da177e4 LT |
1211 | /** |
1212 | * may_umount_tree - check if a mount tree is busy | |
1213 | * @mnt: root of mount tree | |
1214 | * | |
1215 | * This is called to check if a tree of mounts has any | |
1216 | * open files, pwds, chroots or sub mounts that are | |
1217 | * busy. | |
1218 | */ | |
909b0a88 | 1219 | int may_umount_tree(struct vfsmount *m) |
1da177e4 | 1220 | { |
909b0a88 | 1221 | struct mount *mnt = real_mount(m); |
36341f64 RP |
1222 | int actual_refs = 0; |
1223 | int minimum_refs = 0; | |
315fc83e | 1224 | struct mount *p; |
909b0a88 | 1225 | BUG_ON(!m); |
1da177e4 | 1226 | |
b3e19d92 | 1227 | /* write lock needed for mnt_get_count */ |
962830df | 1228 | br_write_lock(&vfsmount_lock); |
909b0a88 | 1229 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
83adc753 | 1230 | actual_refs += mnt_get_count(p); |
1da177e4 | 1231 | minimum_refs += 2; |
1da177e4 | 1232 | } |
962830df | 1233 | br_write_unlock(&vfsmount_lock); |
1da177e4 LT |
1234 | |
1235 | if (actual_refs > minimum_refs) | |
e3474a8e | 1236 | return 0; |
1da177e4 | 1237 | |
e3474a8e | 1238 | return 1; |
1da177e4 LT |
1239 | } |
1240 | ||
1241 | EXPORT_SYMBOL(may_umount_tree); | |
1242 | ||
1243 | /** | |
1244 | * may_umount - check if a mount point is busy | |
1245 | * @mnt: root of mount | |
1246 | * | |
1247 | * This is called to check if a mount point has any | |
1248 | * open files, pwds, chroots or sub mounts. If the | |
1249 | * mount has sub mounts this will return busy | |
1250 | * regardless of whether the sub mounts are busy. | |
1251 | * | |
1252 | * Doesn't take quota and stuff into account. IOW, in some cases it will | |
1253 | * give false negatives. The main reason why it's here is that we need | |
1254 | * a non-destructive way to look for easily umountable filesystems. | |
1255 | */ | |
1256 | int may_umount(struct vfsmount *mnt) | |
1257 | { | |
e3474a8e | 1258 | int ret = 1; |
8ad08d8a | 1259 | down_read(&namespace_sem); |
962830df | 1260 | br_write_lock(&vfsmount_lock); |
1ab59738 | 1261 | if (propagate_mount_busy(real_mount(mnt), 2)) |
e3474a8e | 1262 | ret = 0; |
962830df | 1263 | br_write_unlock(&vfsmount_lock); |
8ad08d8a | 1264 | up_read(&namespace_sem); |
a05964f3 | 1265 | return ret; |
1da177e4 LT |
1266 | } |
1267 | ||
1268 | EXPORT_SYMBOL(may_umount); | |
1269 | ||
e3197d83 AV |
1270 | static LIST_HEAD(unmounted); /* protected by namespace_sem */ |
1271 | ||
97216be0 | 1272 | static void namespace_unlock(void) |
70fbcdf4 | 1273 | { |
d5e50f74 | 1274 | struct mount *mnt; |
97216be0 AV |
1275 | LIST_HEAD(head); |
1276 | ||
1277 | if (likely(list_empty(&unmounted))) { | |
1278 | up_write(&namespace_sem); | |
1279 | return; | |
1280 | } | |
1281 | ||
1282 | list_splice_init(&unmounted, &head); | |
1283 | up_write(&namespace_sem); | |
1284 | ||
1285 | while (!list_empty(&head)) { | |
1286 | mnt = list_first_entry(&head, struct mount, mnt_hash); | |
1b8e5564 | 1287 | list_del_init(&mnt->mnt_hash); |
676da58d | 1288 | if (mnt_has_parent(mnt)) { |
70fbcdf4 | 1289 | struct dentry *dentry; |
863d684f | 1290 | struct mount *m; |
99b7db7b | 1291 | |
962830df | 1292 | br_write_lock(&vfsmount_lock); |
a73324da | 1293 | dentry = mnt->mnt_mountpoint; |
863d684f | 1294 | m = mnt->mnt_parent; |
a73324da | 1295 | mnt->mnt_mountpoint = mnt->mnt.mnt_root; |
0714a533 | 1296 | mnt->mnt_parent = mnt; |
7c4b93d8 | 1297 | m->mnt_ghosts--; |
962830df | 1298 | br_write_unlock(&vfsmount_lock); |
70fbcdf4 | 1299 | dput(dentry); |
863d684f | 1300 | mntput(&m->mnt); |
70fbcdf4 | 1301 | } |
d5e50f74 | 1302 | mntput(&mnt->mnt); |
70fbcdf4 RP |
1303 | } |
1304 | } | |
1305 | ||
97216be0 | 1306 | static inline void namespace_lock(void) |
e3197d83 | 1307 | { |
97216be0 | 1308 | down_write(&namespace_sem); |
e3197d83 AV |
1309 | } |
1310 | ||
99b7db7b NP |
1311 | /* |
1312 | * vfsmount lock must be held for write | |
1313 | * namespace_sem must be held for write | |
1314 | */ | |
328e6d90 | 1315 | void umount_tree(struct mount *mnt, int propagate) |
1da177e4 | 1316 | { |
7b8a53fd | 1317 | LIST_HEAD(tmp_list); |
315fc83e | 1318 | struct mount *p; |
1da177e4 | 1319 | |
909b0a88 | 1320 | for (p = mnt; p; p = next_mnt(p, mnt)) |
1b8e5564 | 1321 | list_move(&p->mnt_hash, &tmp_list); |
1da177e4 | 1322 | |
a05964f3 | 1323 | if (propagate) |
7b8a53fd | 1324 | propagate_umount(&tmp_list); |
a05964f3 | 1325 | |
1b8e5564 | 1326 | list_for_each_entry(p, &tmp_list, mnt_hash) { |
6776db3d | 1327 | list_del_init(&p->mnt_expire); |
1a4eeaf2 | 1328 | list_del_init(&p->mnt_list); |
143c8c91 AV |
1329 | __touch_mnt_namespace(p->mnt_ns); |
1330 | p->mnt_ns = NULL; | |
6b41d536 | 1331 | list_del_init(&p->mnt_child); |
676da58d | 1332 | if (mnt_has_parent(p)) { |
863d684f | 1333 | p->mnt_parent->mnt_ghosts++; |
84d17192 AV |
1334 | put_mountpoint(p->mnt_mp); |
1335 | p->mnt_mp = NULL; | |
7c4b93d8 | 1336 | } |
0f0afb1d | 1337 | change_mnt_propagation(p, MS_PRIVATE); |
1da177e4 | 1338 | } |
328e6d90 | 1339 | list_splice(&tmp_list, &unmounted); |
1da177e4 LT |
1340 | } |
1341 | ||
b54b9be7 | 1342 | static void shrink_submounts(struct mount *mnt); |
c35038be | 1343 | |
1ab59738 | 1344 | static int do_umount(struct mount *mnt, int flags) |
1da177e4 | 1345 | { |
1ab59738 | 1346 | struct super_block *sb = mnt->mnt.mnt_sb; |
1da177e4 LT |
1347 | int retval; |
1348 | ||
1ab59738 | 1349 | retval = security_sb_umount(&mnt->mnt, flags); |
1da177e4 LT |
1350 | if (retval) |
1351 | return retval; | |
1352 | ||
1353 | /* | |
1354 | * Allow userspace to request a mountpoint be expired rather than | |
1355 | * unmounting unconditionally. Unmount only happens if: | |
1356 | * (1) the mark is already set (the mark is cleared by mntput()) | |
1357 | * (2) the usage count == 1 [parent vfsmount] + 1 [sys_umount] | |
1358 | */ | |
1359 | if (flags & MNT_EXPIRE) { | |
1ab59738 | 1360 | if (&mnt->mnt == current->fs->root.mnt || |
1da177e4 LT |
1361 | flags & (MNT_FORCE | MNT_DETACH)) |
1362 | return -EINVAL; | |
1363 | ||
b3e19d92 NP |
1364 | /* |
1365 | * probably don't strictly need the lock here if we examined | |
1366 | * all race cases, but it's a slowpath. | |
1367 | */ | |
962830df | 1368 | br_write_lock(&vfsmount_lock); |
83adc753 | 1369 | if (mnt_get_count(mnt) != 2) { |
962830df | 1370 | br_write_unlock(&vfsmount_lock); |
1da177e4 | 1371 | return -EBUSY; |
b3e19d92 | 1372 | } |
962830df | 1373 | br_write_unlock(&vfsmount_lock); |
1da177e4 | 1374 | |
863d684f | 1375 | if (!xchg(&mnt->mnt_expiry_mark, 1)) |
1da177e4 LT |
1376 | return -EAGAIN; |
1377 | } | |
1378 | ||
1379 | /* | |
1380 | * If we may have to abort operations to get out of this | |
1381 | * mount, and they will themselves hold resources we must | |
1382 | * allow the fs to do things. In the Unix tradition of | |
1383 | * 'Gee thats tricky lets do it in userspace' the umount_begin | |
1384 | * might fail to complete on the first run through as other tasks | |
1385 | * must return, and the like. Thats for the mount program to worry | |
1386 | * about for the moment. | |
1387 | */ | |
1388 | ||
42faad99 | 1389 | if (flags & MNT_FORCE && sb->s_op->umount_begin) { |
42faad99 | 1390 | sb->s_op->umount_begin(sb); |
42faad99 | 1391 | } |
1da177e4 LT |
1392 | |
1393 | /* | |
1394 | * No sense to grab the lock for this test, but test itself looks | |
1395 | * somewhat bogus. Suggestions for better replacement? | |
1396 | * Ho-hum... In principle, we might treat that as umount + switch | |
1397 | * to rootfs. GC would eventually take care of the old vfsmount. | |
1398 | * Actually it makes sense, especially if rootfs would contain a | |
1399 | * /reboot - static binary that would close all descriptors and | |
1400 | * call reboot(9). Then init(8) could umount root and exec /reboot. | |
1401 | */ | |
1ab59738 | 1402 | if (&mnt->mnt == current->fs->root.mnt && !(flags & MNT_DETACH)) { |
1da177e4 LT |
1403 | /* |
1404 | * Special case for "unmounting" root ... | |
1405 | * we just try to remount it readonly. | |
1406 | */ | |
a7dbb3e3 AL |
1407 | if (!capable(CAP_SYS_ADMIN)) |
1408 | return -EPERM; | |
1da177e4 | 1409 | down_write(&sb->s_umount); |
4aa98cf7 | 1410 | if (!(sb->s_flags & MS_RDONLY)) |
1da177e4 | 1411 | retval = do_remount_sb(sb, MS_RDONLY, NULL, 0); |
1da177e4 LT |
1412 | up_write(&sb->s_umount); |
1413 | return retval; | |
1414 | } | |
1415 | ||
97216be0 | 1416 | namespace_lock(); |
962830df | 1417 | br_write_lock(&vfsmount_lock); |
5addc5dd | 1418 | event++; |
1da177e4 | 1419 | |
c35038be | 1420 | if (!(flags & MNT_DETACH)) |
b54b9be7 | 1421 | shrink_submounts(mnt); |
c35038be | 1422 | |
1da177e4 | 1423 | retval = -EBUSY; |
a05964f3 | 1424 | if (flags & MNT_DETACH || !propagate_mount_busy(mnt, 2)) { |
1a4eeaf2 | 1425 | if (!list_empty(&mnt->mnt_list)) |
328e6d90 | 1426 | umount_tree(mnt, 1); |
1da177e4 LT |
1427 | retval = 0; |
1428 | } | |
962830df | 1429 | br_write_unlock(&vfsmount_lock); |
e3197d83 | 1430 | namespace_unlock(); |
1da177e4 LT |
1431 | return retval; |
1432 | } | |
1433 | ||
9b40bc90 AV |
1434 | /* |
1435 | * Is the caller allowed to modify his namespace? | |
1436 | */ | |
1437 | static inline bool may_mount(void) | |
1438 | { | |
1439 | return ns_capable(current->nsproxy->mnt_ns->user_ns, CAP_SYS_ADMIN); | |
1440 | } | |
1441 | ||
1da177e4 LT |
1442 | /* |
1443 | * Now umount can handle mount points as well as block devices. | |
1444 | * This is important for filesystems which use unnamed block devices. | |
1445 | * | |
1446 | * We now support a flag for forced unmount like the other 'big iron' | |
1447 | * unixes. Our API is identical to OSF/1 to avoid making a mess of AMD | |
1448 | */ | |
1449 | ||
bdc480e3 | 1450 | SYSCALL_DEFINE2(umount, char __user *, name, int, flags) |
1da177e4 | 1451 | { |
2d8f3038 | 1452 | struct path path; |
900148dc | 1453 | struct mount *mnt; |
1da177e4 | 1454 | int retval; |
db1f05bb | 1455 | int lookup_flags = 0; |
6fa3eb70 S |
1456 | #ifdef UMOUNT_LOG |
1457 | int total_value =0; | |
1458 | #endif | |
db1f05bb MS |
1459 | if (flags & ~(MNT_FORCE | MNT_DETACH | MNT_EXPIRE | UMOUNT_NOFOLLOW)) |
1460 | return -EINVAL; | |
1461 | ||
9b40bc90 AV |
1462 | if (!may_mount()) |
1463 | return -EPERM; | |
1464 | ||
db1f05bb MS |
1465 | if (!(flags & UMOUNT_NOFOLLOW)) |
1466 | lookup_flags |= LOOKUP_FOLLOW; | |
1467 | ||
1468 | retval = user_path_at(AT_FDCWD, name, lookup_flags, &path); | |
1da177e4 LT |
1469 | if (retval) |
1470 | goto out; | |
900148dc | 1471 | mnt = real_mount(path.mnt); |
1da177e4 | 1472 | retval = -EINVAL; |
2d8f3038 | 1473 | if (path.dentry != path.mnt->mnt_root) |
1da177e4 | 1474 | goto dput_and_out; |
143c8c91 | 1475 | if (!check_mnt(mnt)) |
1da177e4 | 1476 | goto dput_and_out; |
c65d3b05 EB |
1477 | retval = -EPERM; |
1478 | if (flags & MNT_FORCE && !capable(CAP_SYS_ADMIN)) | |
1479 | goto dput_and_out; | |
1da177e4 | 1480 | |
900148dc | 1481 | retval = do_umount(mnt, flags); |
6fa3eb70 S |
1482 | #ifdef UMOUNT_LOG |
1483 | { | |
1484 | printk("Ahsin do_umount retval=%d \n",retval); | |
1485 | //do_umount success: 0, do_umount busy: -16 | |
1486 | //if do_umount fail, need to dump the link list here | |
1487 | ||
1488 | if(retval) | |
1489 | printk("Ahsin do_umount fail; mnt_get_count=%d mnt->mnt_devname=%s\n",mnt_get_count(mnt),mnt->mnt_devname); | |
1490 | else | |
1491 | printk("Ahsin do_umount success; mnt_get_count=%d mnt->mnt_devname=%s\n",mnt_get_count(mnt),mnt->mnt_devname); | |
1492 | ||
1493 | // print linked list | |
1494 | spin_lock(&mnt_id_lock); | |
1495 | ref_current = ref_head; | |
1496 | while(ref_current != NULL) | |
1497 | { | |
1498 | total_value = total_value + ref_current->count; | |
1499 | ||
1500 | if (ref_current->count) | |
1501 | printk("Ahsin PID= %d, Name = %s, Count= %d \n", ref_current->pid, ref_current->name, ref_current->count); | |
1502 | ref_current = ref_current->next; | |
1503 | } | |
1504 | spin_unlock(&mnt_id_lock); | |
1505 | printk("Ahsin total_value=%d \n",total_value); | |
1506 | } | |
1507 | #endif | |
1da177e4 | 1508 | dput_and_out: |
429731b1 | 1509 | /* we mustn't call path_put() as that would clear mnt_expiry_mark */ |
2d8f3038 | 1510 | dput(path.dentry); |
900148dc | 1511 | mntput_no_expire(mnt); |
1da177e4 LT |
1512 | out: |
1513 | return retval; | |
1514 | } | |
1515 | ||
1516 | #ifdef __ARCH_WANT_SYS_OLDUMOUNT | |
1517 | ||
1518 | /* | |
b58fed8b | 1519 | * The 2.0 compatible umount. No flags. |
1da177e4 | 1520 | */ |
bdc480e3 | 1521 | SYSCALL_DEFINE1(oldumount, char __user *, name) |
1da177e4 | 1522 | { |
b58fed8b | 1523 | return sys_umount(name, 0); |
1da177e4 LT |
1524 | } |
1525 | ||
1526 | #endif | |
1527 | ||
8823c079 EB |
1528 | static bool mnt_ns_loop(struct path *path) |
1529 | { | |
1530 | /* Could bind mounting the mount namespace inode cause a | |
1531 | * mount namespace loop? | |
1532 | */ | |
1533 | struct inode *inode = path->dentry->d_inode; | |
0bb80f24 | 1534 | struct proc_ns *ei; |
8823c079 EB |
1535 | struct mnt_namespace *mnt_ns; |
1536 | ||
1537 | if (!proc_ns_inode(inode)) | |
1538 | return false; | |
1539 | ||
0bb80f24 | 1540 | ei = get_proc_ns(inode); |
8823c079 EB |
1541 | if (ei->ns_ops != &mntns_operations) |
1542 | return false; | |
1543 | ||
1544 | mnt_ns = ei->ns; | |
1545 | return current->nsproxy->mnt_ns->seq >= mnt_ns->seq; | |
1546 | } | |
1547 | ||
87129cc0 | 1548 | struct mount *copy_tree(struct mount *mnt, struct dentry *dentry, |
36341f64 | 1549 | int flag) |
1da177e4 | 1550 | { |
84d17192 | 1551 | struct mount *res, *p, *q, *r, *parent; |
1da177e4 | 1552 | |
fc7be130 | 1553 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(mnt)) |
be34d1a3 | 1554 | return ERR_PTR(-EINVAL); |
9676f0c6 | 1555 | |
36341f64 | 1556 | res = q = clone_mnt(mnt, dentry, flag); |
be34d1a3 DH |
1557 | if (IS_ERR(q)) |
1558 | return q; | |
1559 | ||
a73324da | 1560 | q->mnt_mountpoint = mnt->mnt_mountpoint; |
1da177e4 LT |
1561 | |
1562 | p = mnt; | |
6b41d536 | 1563 | list_for_each_entry(r, &mnt->mnt_mounts, mnt_child) { |
315fc83e | 1564 | struct mount *s; |
7ec02ef1 | 1565 | if (!is_subdir(r->mnt_mountpoint, dentry)) |
1da177e4 LT |
1566 | continue; |
1567 | ||
909b0a88 | 1568 | for (s = r; s; s = next_mnt(s, r)) { |
fc7be130 | 1569 | if (!(flag & CL_COPY_ALL) && IS_MNT_UNBINDABLE(s)) { |
9676f0c6 RP |
1570 | s = skip_mnt_tree(s); |
1571 | continue; | |
1572 | } | |
0714a533 AV |
1573 | while (p != s->mnt_parent) { |
1574 | p = p->mnt_parent; | |
1575 | q = q->mnt_parent; | |
1da177e4 | 1576 | } |
87129cc0 | 1577 | p = s; |
84d17192 | 1578 | parent = q; |
87129cc0 | 1579 | q = clone_mnt(p, p->mnt.mnt_root, flag); |
be34d1a3 DH |
1580 | if (IS_ERR(q)) |
1581 | goto out; | |
962830df | 1582 | br_write_lock(&vfsmount_lock); |
1a4eeaf2 | 1583 | list_add_tail(&q->mnt_list, &res->mnt_list); |
84d17192 | 1584 | attach_mnt(q, parent, p->mnt_mp); |
962830df | 1585 | br_write_unlock(&vfsmount_lock); |
1da177e4 LT |
1586 | } |
1587 | } | |
1588 | return res; | |
be34d1a3 | 1589 | out: |
1da177e4 | 1590 | if (res) { |
962830df | 1591 | br_write_lock(&vfsmount_lock); |
328e6d90 | 1592 | umount_tree(res, 0); |
962830df | 1593 | br_write_unlock(&vfsmount_lock); |
1da177e4 | 1594 | } |
be34d1a3 | 1595 | return q; |
1da177e4 LT |
1596 | } |
1597 | ||
be34d1a3 DH |
1598 | /* Caller should check returned pointer for errors */ |
1599 | ||
589ff870 | 1600 | struct vfsmount *collect_mounts(struct path *path) |
8aec0809 | 1601 | { |
cb338d06 | 1602 | struct mount *tree; |
97216be0 | 1603 | namespace_lock(); |
87129cc0 AV |
1604 | tree = copy_tree(real_mount(path->mnt), path->dentry, |
1605 | CL_COPY_ALL | CL_PRIVATE); | |
328e6d90 | 1606 | namespace_unlock(); |
be34d1a3 | 1607 | if (IS_ERR(tree)) |
2bee1e0f | 1608 | return ERR_CAST(tree); |
be34d1a3 | 1609 | return &tree->mnt; |
8aec0809 AV |
1610 | } |
1611 | ||
1612 | void drop_collected_mounts(struct vfsmount *mnt) | |
1613 | { | |
97216be0 | 1614 | namespace_lock(); |
962830df | 1615 | br_write_lock(&vfsmount_lock); |
328e6d90 | 1616 | umount_tree(real_mount(mnt), 0); |
962830df | 1617 | br_write_unlock(&vfsmount_lock); |
3ab6abee | 1618 | namespace_unlock(); |
8aec0809 AV |
1619 | } |
1620 | ||
1f707137 AV |
1621 | int iterate_mounts(int (*f)(struct vfsmount *, void *), void *arg, |
1622 | struct vfsmount *root) | |
1623 | { | |
1a4eeaf2 | 1624 | struct mount *mnt; |
1f707137 AV |
1625 | int res = f(root, arg); |
1626 | if (res) | |
1627 | return res; | |
1a4eeaf2 AV |
1628 | list_for_each_entry(mnt, &real_mount(root)->mnt_list, mnt_list) { |
1629 | res = f(&mnt->mnt, arg); | |
1f707137 AV |
1630 | if (res) |
1631 | return res; | |
1632 | } | |
1633 | return 0; | |
1634 | } | |
1635 | ||
4b8b21f4 | 1636 | static void cleanup_group_ids(struct mount *mnt, struct mount *end) |
719f5d7f | 1637 | { |
315fc83e | 1638 | struct mount *p; |
719f5d7f | 1639 | |
909b0a88 | 1640 | for (p = mnt; p != end; p = next_mnt(p, mnt)) { |
fc7be130 | 1641 | if (p->mnt_group_id && !IS_MNT_SHARED(p)) |
4b8b21f4 | 1642 | mnt_release_group_id(p); |
719f5d7f MS |
1643 | } |
1644 | } | |
1645 | ||
4b8b21f4 | 1646 | static int invent_group_ids(struct mount *mnt, bool recurse) |
719f5d7f | 1647 | { |
315fc83e | 1648 | struct mount *p; |
719f5d7f | 1649 | |
909b0a88 | 1650 | for (p = mnt; p; p = recurse ? next_mnt(p, mnt) : NULL) { |
fc7be130 | 1651 | if (!p->mnt_group_id && !IS_MNT_SHARED(p)) { |
4b8b21f4 | 1652 | int err = mnt_alloc_group_id(p); |
719f5d7f | 1653 | if (err) { |
4b8b21f4 | 1654 | cleanup_group_ids(mnt, p); |
719f5d7f MS |
1655 | return err; |
1656 | } | |
1657 | } | |
1658 | } | |
1659 | ||
1660 | return 0; | |
1661 | } | |
1662 | ||
b90fa9ae RP |
1663 | /* |
1664 | * @source_mnt : mount tree to be attached | |
21444403 RP |
1665 | * @nd : place the mount tree @source_mnt is attached |
1666 | * @parent_nd : if non-null, detach the source_mnt from its parent and | |
1667 | * store the parent mount and mountpoint dentry. | |
1668 | * (done when source_mnt is moved) | |
b90fa9ae RP |
1669 | * |
1670 | * NOTE: in the table below explains the semantics when a source mount | |
1671 | * of a given type is attached to a destination mount of a given type. | |
9676f0c6 RP |
1672 | * --------------------------------------------------------------------------- |
1673 | * | BIND MOUNT OPERATION | | |
1674 | * |************************************************************************** | |
1675 | * | source-->| shared | private | slave | unbindable | | |
1676 | * | dest | | | | | | |
1677 | * | | | | | | | | |
1678 | * | v | | | | | | |
1679 | * |************************************************************************** | |
1680 | * | shared | shared (++) | shared (+) | shared(+++)| invalid | | |
1681 | * | | | | | | | |
1682 | * |non-shared| shared (+) | private | slave (*) | invalid | | |
1683 | * *************************************************************************** | |
b90fa9ae RP |
1684 | * A bind operation clones the source mount and mounts the clone on the |
1685 | * destination mount. | |
1686 | * | |
1687 | * (++) the cloned mount is propagated to all the mounts in the propagation | |
1688 | * tree of the destination mount and the cloned mount is added to | |
1689 | * the peer group of the source mount. | |
1690 | * (+) the cloned mount is created under the destination mount and is marked | |
1691 | * as shared. The cloned mount is added to the peer group of the source | |
1692 | * mount. | |
5afe0022 RP |
1693 | * (+++) the mount is propagated to all the mounts in the propagation tree |
1694 | * of the destination mount and the cloned mount is made slave | |
1695 | * of the same master as that of the source mount. The cloned mount | |
1696 | * is marked as 'shared and slave'. | |
1697 | * (*) the cloned mount is made a slave of the same master as that of the | |
1698 | * source mount. | |
1699 | * | |
9676f0c6 RP |
1700 | * --------------------------------------------------------------------------- |
1701 | * | MOVE MOUNT OPERATION | | |
1702 | * |************************************************************************** | |
1703 | * | source-->| shared | private | slave | unbindable | | |
1704 | * | dest | | | | | | |
1705 | * | | | | | | | | |
1706 | * | v | | | | | | |
1707 | * |************************************************************************** | |
1708 | * | shared | shared (+) | shared (+) | shared(+++) | invalid | | |
1709 | * | | | | | | | |
1710 | * |non-shared| shared (+*) | private | slave (*) | unbindable | | |
1711 | * *************************************************************************** | |
5afe0022 RP |
1712 | * |
1713 | * (+) the mount is moved to the destination. And is then propagated to | |
1714 | * all the mounts in the propagation tree of the destination mount. | |
21444403 | 1715 | * (+*) the mount is moved to the destination. |
5afe0022 RP |
1716 | * (+++) the mount is moved to the destination and is then propagated to |
1717 | * all the mounts belonging to the destination mount's propagation tree. | |
1718 | * the mount is marked as 'shared and slave'. | |
1719 | * (*) the mount continues to be a slave at the new location. | |
b90fa9ae RP |
1720 | * |
1721 | * if the source mount is a tree, the operations explained above is | |
1722 | * applied to each mount in the tree. | |
1723 | * Must be called without spinlocks held, since this function can sleep | |
1724 | * in allocations. | |
1725 | */ | |
0fb54e50 | 1726 | static int attach_recursive_mnt(struct mount *source_mnt, |
84d17192 AV |
1727 | struct mount *dest_mnt, |
1728 | struct mountpoint *dest_mp, | |
1729 | struct path *parent_path) | |
b90fa9ae RP |
1730 | { |
1731 | LIST_HEAD(tree_list); | |
315fc83e | 1732 | struct mount *child, *p; |
719f5d7f | 1733 | int err; |
b90fa9ae | 1734 | |
fc7be130 | 1735 | if (IS_MNT_SHARED(dest_mnt)) { |
0fb54e50 | 1736 | err = invent_group_ids(source_mnt, true); |
719f5d7f MS |
1737 | if (err) |
1738 | goto out; | |
1739 | } | |
84d17192 | 1740 | err = propagate_mnt(dest_mnt, dest_mp, source_mnt, &tree_list); |
719f5d7f MS |
1741 | if (err) |
1742 | goto out_cleanup_ids; | |
b90fa9ae | 1743 | |
962830df | 1744 | br_write_lock(&vfsmount_lock); |
df1a1ad2 | 1745 | |
fc7be130 | 1746 | if (IS_MNT_SHARED(dest_mnt)) { |
909b0a88 | 1747 | for (p = source_mnt; p; p = next_mnt(p, source_mnt)) |
0f0afb1d | 1748 | set_mnt_shared(p); |
b90fa9ae | 1749 | } |
1a390689 | 1750 | if (parent_path) { |
0fb54e50 | 1751 | detach_mnt(source_mnt, parent_path); |
84d17192 | 1752 | attach_mnt(source_mnt, dest_mnt, dest_mp); |
143c8c91 | 1753 | touch_mnt_namespace(source_mnt->mnt_ns); |
21444403 | 1754 | } else { |
84d17192 | 1755 | mnt_set_mountpoint(dest_mnt, dest_mp, source_mnt); |
0fb54e50 | 1756 | commit_tree(source_mnt); |
21444403 | 1757 | } |
b90fa9ae | 1758 | |
1b8e5564 AV |
1759 | list_for_each_entry_safe(child, p, &tree_list, mnt_hash) { |
1760 | list_del_init(&child->mnt_hash); | |
4b2619a5 | 1761 | commit_tree(child); |
b90fa9ae | 1762 | } |
962830df | 1763 | br_write_unlock(&vfsmount_lock); |
99b7db7b | 1764 | |
b90fa9ae | 1765 | return 0; |
719f5d7f MS |
1766 | |
1767 | out_cleanup_ids: | |
fc7be130 | 1768 | if (IS_MNT_SHARED(dest_mnt)) |
0fb54e50 | 1769 | cleanup_group_ids(source_mnt, NULL); |
719f5d7f MS |
1770 | out: |
1771 | return err; | |
b90fa9ae RP |
1772 | } |
1773 | ||
84d17192 | 1774 | static struct mountpoint *lock_mount(struct path *path) |
b12cea91 AV |
1775 | { |
1776 | struct vfsmount *mnt; | |
84d17192 | 1777 | struct dentry *dentry = path->dentry; |
b12cea91 | 1778 | retry: |
84d17192 AV |
1779 | mutex_lock(&dentry->d_inode->i_mutex); |
1780 | if (unlikely(cant_mount(dentry))) { | |
1781 | mutex_unlock(&dentry->d_inode->i_mutex); | |
1782 | return ERR_PTR(-ENOENT); | |
b12cea91 | 1783 | } |
97216be0 | 1784 | namespace_lock(); |
b12cea91 | 1785 | mnt = lookup_mnt(path); |
84d17192 AV |
1786 | if (likely(!mnt)) { |
1787 | struct mountpoint *mp = new_mountpoint(dentry); | |
1788 | if (IS_ERR(mp)) { | |
97216be0 | 1789 | namespace_unlock(); |
84d17192 AV |
1790 | mutex_unlock(&dentry->d_inode->i_mutex); |
1791 | return mp; | |
1792 | } | |
1793 | return mp; | |
1794 | } | |
97216be0 | 1795 | namespace_unlock(); |
b12cea91 AV |
1796 | mutex_unlock(&path->dentry->d_inode->i_mutex); |
1797 | path_put(path); | |
1798 | path->mnt = mnt; | |
84d17192 | 1799 | dentry = path->dentry = dget(mnt->mnt_root); |
b12cea91 AV |
1800 | goto retry; |
1801 | } | |
1802 | ||
84d17192 | 1803 | static void unlock_mount(struct mountpoint *where) |
b12cea91 | 1804 | { |
84d17192 AV |
1805 | struct dentry *dentry = where->m_dentry; |
1806 | put_mountpoint(where); | |
328e6d90 | 1807 | namespace_unlock(); |
84d17192 | 1808 | mutex_unlock(&dentry->d_inode->i_mutex); |
b12cea91 AV |
1809 | } |
1810 | ||
84d17192 | 1811 | static int graft_tree(struct mount *mnt, struct mount *p, struct mountpoint *mp) |
1da177e4 | 1812 | { |
95bc5f25 | 1813 | if (mnt->mnt.mnt_sb->s_flags & MS_NOUSER) |
1da177e4 LT |
1814 | return -EINVAL; |
1815 | ||
84d17192 | 1816 | if (S_ISDIR(mp->m_dentry->d_inode->i_mode) != |
95bc5f25 | 1817 | S_ISDIR(mnt->mnt.mnt_root->d_inode->i_mode)) |
1da177e4 LT |
1818 | return -ENOTDIR; |
1819 | ||
84d17192 | 1820 | return attach_recursive_mnt(mnt, p, mp, NULL); |
1da177e4 LT |
1821 | } |
1822 | ||
7a2e8a8f VA |
1823 | /* |
1824 | * Sanity check the flags to change_mnt_propagation. | |
1825 | */ | |
1826 | ||
1827 | static int flags_to_propagation_type(int flags) | |
1828 | { | |
7c6e984d | 1829 | int type = flags & ~(MS_REC | MS_SILENT); |
7a2e8a8f VA |
1830 | |
1831 | /* Fail if any non-propagation flags are set */ | |
1832 | if (type & ~(MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) | |
1833 | return 0; | |
1834 | /* Only one propagation flag should be set */ | |
1835 | if (!is_power_of_2(type)) | |
1836 | return 0; | |
1837 | return type; | |
1838 | } | |
1839 | ||
07b20889 RP |
1840 | /* |
1841 | * recursively change the type of the mountpoint. | |
1842 | */ | |
0a0d8a46 | 1843 | static int do_change_type(struct path *path, int flag) |
07b20889 | 1844 | { |
315fc83e | 1845 | struct mount *m; |
4b8b21f4 | 1846 | struct mount *mnt = real_mount(path->mnt); |
07b20889 | 1847 | int recurse = flag & MS_REC; |
7a2e8a8f | 1848 | int type; |
719f5d7f | 1849 | int err = 0; |
07b20889 | 1850 | |
2d92ab3c | 1851 | if (path->dentry != path->mnt->mnt_root) |
07b20889 RP |
1852 | return -EINVAL; |
1853 | ||
7a2e8a8f VA |
1854 | type = flags_to_propagation_type(flag); |
1855 | if (!type) | |
1856 | return -EINVAL; | |
1857 | ||
97216be0 | 1858 | namespace_lock(); |
719f5d7f MS |
1859 | if (type == MS_SHARED) { |
1860 | err = invent_group_ids(mnt, recurse); | |
1861 | if (err) | |
1862 | goto out_unlock; | |
1863 | } | |
1864 | ||
962830df | 1865 | br_write_lock(&vfsmount_lock); |
909b0a88 | 1866 | for (m = mnt; m; m = (recurse ? next_mnt(m, mnt) : NULL)) |
0f0afb1d | 1867 | change_mnt_propagation(m, type); |
962830df | 1868 | br_write_unlock(&vfsmount_lock); |
719f5d7f MS |
1869 | |
1870 | out_unlock: | |
97216be0 | 1871 | namespace_unlock(); |
719f5d7f | 1872 | return err; |
07b20889 RP |
1873 | } |
1874 | ||
1da177e4 LT |
1875 | /* |
1876 | * do loopback mount. | |
1877 | */ | |
808d4e3c | 1878 | static int do_loopback(struct path *path, const char *old_name, |
2dafe1c4 | 1879 | int recurse) |
1da177e4 | 1880 | { |
2d92ab3c | 1881 | struct path old_path; |
84d17192 AV |
1882 | struct mount *mnt = NULL, *old, *parent; |
1883 | struct mountpoint *mp; | |
57eccb83 | 1884 | int err; |
1da177e4 LT |
1885 | if (!old_name || !*old_name) |
1886 | return -EINVAL; | |
815d405c | 1887 | err = kern_path(old_name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &old_path); |
1da177e4 LT |
1888 | if (err) |
1889 | return err; | |
1890 | ||
8823c079 EB |
1891 | err = -EINVAL; |
1892 | if (mnt_ns_loop(&old_path)) | |
1893 | goto out; | |
1894 | ||
84d17192 AV |
1895 | mp = lock_mount(path); |
1896 | err = PTR_ERR(mp); | |
1897 | if (IS_ERR(mp)) | |
b12cea91 AV |
1898 | goto out; |
1899 | ||
87129cc0 | 1900 | old = real_mount(old_path.mnt); |
84d17192 | 1901 | parent = real_mount(path->mnt); |
87129cc0 | 1902 | |
1da177e4 | 1903 | err = -EINVAL; |
fc7be130 | 1904 | if (IS_MNT_UNBINDABLE(old)) |
b12cea91 | 1905 | goto out2; |
9676f0c6 | 1906 | |
84d17192 | 1907 | if (!check_mnt(parent) || !check_mnt(old)) |
b12cea91 | 1908 | goto out2; |
1da177e4 | 1909 | |
ccd48bc7 | 1910 | if (recurse) |
87129cc0 | 1911 | mnt = copy_tree(old, old_path.dentry, 0); |
ccd48bc7 | 1912 | else |
87129cc0 | 1913 | mnt = clone_mnt(old, old_path.dentry, 0); |
ccd48bc7 | 1914 | |
be34d1a3 DH |
1915 | if (IS_ERR(mnt)) { |
1916 | err = PTR_ERR(mnt); | |
e9c5d8a5 | 1917 | goto out2; |
be34d1a3 | 1918 | } |
ccd48bc7 | 1919 | |
84d17192 | 1920 | err = graft_tree(mnt, parent, mp); |
ccd48bc7 | 1921 | if (err) { |
962830df | 1922 | br_write_lock(&vfsmount_lock); |
328e6d90 | 1923 | umount_tree(mnt, 0); |
962830df | 1924 | br_write_unlock(&vfsmount_lock); |
5b83d2c5 | 1925 | } |
b12cea91 | 1926 | out2: |
84d17192 | 1927 | unlock_mount(mp); |
ccd48bc7 | 1928 | out: |
2d92ab3c | 1929 | path_put(&old_path); |
1da177e4 LT |
1930 | return err; |
1931 | } | |
1932 | ||
2e4b7fcd DH |
1933 | static int change_mount_flags(struct vfsmount *mnt, int ms_flags) |
1934 | { | |
1935 | int error = 0; | |
1936 | int readonly_request = 0; | |
1937 | ||
1938 | if (ms_flags & MS_RDONLY) | |
1939 | readonly_request = 1; | |
1940 | if (readonly_request == __mnt_is_readonly(mnt)) | |
1941 | return 0; | |
1942 | ||
1943 | if (readonly_request) | |
83adc753 | 1944 | error = mnt_make_readonly(real_mount(mnt)); |
2e4b7fcd | 1945 | else |
83adc753 | 1946 | __mnt_unmake_readonly(real_mount(mnt)); |
2e4b7fcd DH |
1947 | return error; |
1948 | } | |
1949 | ||
1da177e4 LT |
1950 | /* |
1951 | * change filesystem flags. dir should be a physical root of filesystem. | |
1952 | * If you've mounted a non-root directory somewhere and want to do remount | |
1953 | * on it - tough luck. | |
1954 | */ | |
0a0d8a46 | 1955 | static int do_remount(struct path *path, int flags, int mnt_flags, |
1da177e4 LT |
1956 | void *data) |
1957 | { | |
1958 | int err; | |
2d92ab3c | 1959 | struct super_block *sb = path->mnt->mnt_sb; |
143c8c91 | 1960 | struct mount *mnt = real_mount(path->mnt); |
1da177e4 | 1961 | |
143c8c91 | 1962 | if (!check_mnt(mnt)) |
1da177e4 LT |
1963 | return -EINVAL; |
1964 | ||
2d92ab3c | 1965 | if (path->dentry != path->mnt->mnt_root) |
1da177e4 LT |
1966 | return -EINVAL; |
1967 | ||
81d4c13e EB |
1968 | /* Don't allow changing of locked mnt flags. |
1969 | * | |
1970 | * No locks need to be held here while testing the various | |
1971 | * MNT_LOCK flags because those flags can never be cleared | |
1972 | * once they are set. | |
1973 | */ | |
1974 | if ((mnt->mnt.mnt_flags & MNT_LOCK_READONLY) && | |
1975 | !(mnt_flags & MNT_READONLY)) { | |
1976 | return -EPERM; | |
1977 | } | |
187985d9 EB |
1978 | if ((mnt->mnt.mnt_flags & MNT_LOCK_NODEV) && |
1979 | !(mnt_flags & MNT_NODEV)) { | |
4be461b1 EB |
1980 | /* Was the nodev implicitly added in mount? */ |
1981 | if ((mnt->mnt_ns->user_ns != &init_user_ns) && | |
1982 | !(sb->s_type->fs_flags & FS_USERNS_DEV_MOUNT)) { | |
1983 | mnt_flags |= MNT_NODEV; | |
1984 | } else { | |
1985 | return -EPERM; | |
1986 | } | |
187985d9 EB |
1987 | } |
1988 | if ((mnt->mnt.mnt_flags & MNT_LOCK_NOSUID) && | |
1989 | !(mnt_flags & MNT_NOSUID)) { | |
1990 | return -EPERM; | |
1991 | } | |
1992 | if ((mnt->mnt.mnt_flags & MNT_LOCK_NOEXEC) && | |
1993 | !(mnt_flags & MNT_NOEXEC)) { | |
1994 | return -EPERM; | |
1995 | } | |
1996 | if ((mnt->mnt.mnt_flags & MNT_LOCK_ATIME) && | |
1997 | ((mnt->mnt.mnt_flags & MNT_ATIME_MASK) != (mnt_flags & MNT_ATIME_MASK))) { | |
1998 | return -EPERM; | |
1999 | } | |
2000 | ||
ff36fe2c EP |
2001 | err = security_sb_remount(sb, data); |
2002 | if (err) | |
2003 | return err; | |
2004 | ||
1da177e4 | 2005 | down_write(&sb->s_umount); |
2e4b7fcd | 2006 | if (flags & MS_BIND) |
2d92ab3c | 2007 | err = change_mount_flags(path->mnt, flags); |
57eccb83 AV |
2008 | else if (!capable(CAP_SYS_ADMIN)) |
2009 | err = -EPERM; | |
4aa98cf7 | 2010 | else |
2e4b7fcd | 2011 | err = do_remount_sb(sb, flags, data, 0); |
7b43a79f | 2012 | if (!err) { |
962830df | 2013 | br_write_lock(&vfsmount_lock); |
8c30f227 | 2014 | mnt_flags |= mnt->mnt.mnt_flags & ~MNT_USER_SETTABLE_MASK; |
143c8c91 | 2015 | mnt->mnt.mnt_flags = mnt_flags; |
962830df | 2016 | br_write_unlock(&vfsmount_lock); |
7b43a79f | 2017 | } |
1da177e4 | 2018 | up_write(&sb->s_umount); |
0e55a7cc | 2019 | if (!err) { |
962830df | 2020 | br_write_lock(&vfsmount_lock); |
143c8c91 | 2021 | touch_mnt_namespace(mnt->mnt_ns); |
962830df | 2022 | br_write_unlock(&vfsmount_lock); |
0e55a7cc | 2023 | } |
1da177e4 LT |
2024 | return err; |
2025 | } | |
2026 | ||
cbbe362c | 2027 | static inline int tree_contains_unbindable(struct mount *mnt) |
9676f0c6 | 2028 | { |
315fc83e | 2029 | struct mount *p; |
909b0a88 | 2030 | for (p = mnt; p; p = next_mnt(p, mnt)) { |
fc7be130 | 2031 | if (IS_MNT_UNBINDABLE(p)) |
9676f0c6 RP |
2032 | return 1; |
2033 | } | |
2034 | return 0; | |
2035 | } | |
2036 | ||
808d4e3c | 2037 | static int do_move_mount(struct path *path, const char *old_name) |
1da177e4 | 2038 | { |
2d92ab3c | 2039 | struct path old_path, parent_path; |
676da58d | 2040 | struct mount *p; |
0fb54e50 | 2041 | struct mount *old; |
84d17192 | 2042 | struct mountpoint *mp; |
57eccb83 | 2043 | int err; |
1da177e4 LT |
2044 | if (!old_name || !*old_name) |
2045 | return -EINVAL; | |
2d92ab3c | 2046 | err = kern_path(old_name, LOOKUP_FOLLOW, &old_path); |
1da177e4 LT |
2047 | if (err) |
2048 | return err; | |
2049 | ||
84d17192 AV |
2050 | mp = lock_mount(path); |
2051 | err = PTR_ERR(mp); | |
2052 | if (IS_ERR(mp)) | |
cc53ce53 DH |
2053 | goto out; |
2054 | ||
143c8c91 | 2055 | old = real_mount(old_path.mnt); |
fc7be130 | 2056 | p = real_mount(path->mnt); |
143c8c91 | 2057 | |
1da177e4 | 2058 | err = -EINVAL; |
fc7be130 | 2059 | if (!check_mnt(p) || !check_mnt(old)) |
1da177e4 LT |
2060 | goto out1; |
2061 | ||
1da177e4 | 2062 | err = -EINVAL; |
2d92ab3c | 2063 | if (old_path.dentry != old_path.mnt->mnt_root) |
21444403 | 2064 | goto out1; |
1da177e4 | 2065 | |
676da58d | 2066 | if (!mnt_has_parent(old)) |
21444403 | 2067 | goto out1; |
1da177e4 | 2068 | |
2d92ab3c AV |
2069 | if (S_ISDIR(path->dentry->d_inode->i_mode) != |
2070 | S_ISDIR(old_path.dentry->d_inode->i_mode)) | |
21444403 RP |
2071 | goto out1; |
2072 | /* | |
2073 | * Don't move a mount residing in a shared parent. | |
2074 | */ | |
fc7be130 | 2075 | if (IS_MNT_SHARED(old->mnt_parent)) |
21444403 | 2076 | goto out1; |
9676f0c6 RP |
2077 | /* |
2078 | * Don't move a mount tree containing unbindable mounts to a destination | |
2079 | * mount which is shared. | |
2080 | */ | |
fc7be130 | 2081 | if (IS_MNT_SHARED(p) && tree_contains_unbindable(old)) |
9676f0c6 | 2082 | goto out1; |
1da177e4 | 2083 | err = -ELOOP; |
fc7be130 | 2084 | for (; mnt_has_parent(p); p = p->mnt_parent) |
676da58d | 2085 | if (p == old) |
21444403 | 2086 | goto out1; |
1da177e4 | 2087 | |
84d17192 | 2088 | err = attach_recursive_mnt(old, real_mount(path->mnt), mp, &parent_path); |
4ac91378 | 2089 | if (err) |
21444403 | 2090 | goto out1; |
1da177e4 LT |
2091 | |
2092 | /* if the mount is moved, it should no longer be expire | |
2093 | * automatically */ | |
6776db3d | 2094 | list_del_init(&old->mnt_expire); |
1da177e4 | 2095 | out1: |
84d17192 | 2096 | unlock_mount(mp); |
1da177e4 | 2097 | out: |
1da177e4 | 2098 | if (!err) |
1a390689 | 2099 | path_put(&parent_path); |
2d92ab3c | 2100 | path_put(&old_path); |
1da177e4 LT |
2101 | return err; |
2102 | } | |
2103 | ||
9d412a43 AV |
2104 | static struct vfsmount *fs_set_subtype(struct vfsmount *mnt, const char *fstype) |
2105 | { | |
2106 | int err; | |
2107 | const char *subtype = strchr(fstype, '.'); | |
2108 | if (subtype) { | |
2109 | subtype++; | |
2110 | err = -EINVAL; | |
2111 | if (!subtype[0]) | |
2112 | goto err; | |
2113 | } else | |
2114 | subtype = ""; | |
2115 | ||
2116 | mnt->mnt_sb->s_subtype = kstrdup(subtype, GFP_KERNEL); | |
2117 | err = -ENOMEM; | |
2118 | if (!mnt->mnt_sb->s_subtype) | |
2119 | goto err; | |
2120 | return mnt; | |
2121 | ||
2122 | err: | |
2123 | mntput(mnt); | |
2124 | return ERR_PTR(err); | |
2125 | } | |
2126 | ||
9d412a43 AV |
2127 | /* |
2128 | * add a mount into a namespace's mount tree | |
2129 | */ | |
95bc5f25 | 2130 | static int do_add_mount(struct mount *newmnt, struct path *path, int mnt_flags) |
9d412a43 | 2131 | { |
84d17192 AV |
2132 | struct mountpoint *mp; |
2133 | struct mount *parent; | |
9d412a43 AV |
2134 | int err; |
2135 | ||
2136 | mnt_flags &= ~(MNT_SHARED | MNT_WRITE_HOLD | MNT_INTERNAL); | |
2137 | ||
84d17192 AV |
2138 | mp = lock_mount(path); |
2139 | if (IS_ERR(mp)) | |
2140 | return PTR_ERR(mp); | |
9d412a43 | 2141 | |
84d17192 | 2142 | parent = real_mount(path->mnt); |
9d412a43 | 2143 | err = -EINVAL; |
84d17192 | 2144 | if (unlikely(!check_mnt(parent))) { |
156cacb1 AV |
2145 | /* that's acceptable only for automounts done in private ns */ |
2146 | if (!(mnt_flags & MNT_SHRINKABLE)) | |
2147 | goto unlock; | |
2148 | /* ... and for those we'd better have mountpoint still alive */ | |
84d17192 | 2149 | if (!parent->mnt_ns) |
156cacb1 AV |
2150 | goto unlock; |
2151 | } | |
9d412a43 AV |
2152 | |
2153 | /* Refuse the same filesystem on the same mount point */ | |
2154 | err = -EBUSY; | |
95bc5f25 | 2155 | if (path->mnt->mnt_sb == newmnt->mnt.mnt_sb && |
9d412a43 AV |
2156 | path->mnt->mnt_root == path->dentry) |
2157 | goto unlock; | |
2158 | ||
2159 | err = -EINVAL; | |
95bc5f25 | 2160 | if (S_ISLNK(newmnt->mnt.mnt_root->d_inode->i_mode)) |
9d412a43 AV |
2161 | goto unlock; |
2162 | ||
95bc5f25 | 2163 | newmnt->mnt.mnt_flags = mnt_flags; |
84d17192 | 2164 | err = graft_tree(newmnt, parent, mp); |
9d412a43 AV |
2165 | |
2166 | unlock: | |
84d17192 | 2167 | unlock_mount(mp); |
9d412a43 AV |
2168 | return err; |
2169 | } | |
b1e75df4 | 2170 | |
1da177e4 LT |
2171 | /* |
2172 | * create a new mount for userspace and request it to be added into the | |
2173 | * namespace's tree | |
2174 | */ | |
0c55cfc4 | 2175 | static int do_new_mount(struct path *path, const char *fstype, int flags, |
808d4e3c | 2176 | int mnt_flags, const char *name, void *data) |
1da177e4 | 2177 | { |
0c55cfc4 | 2178 | struct file_system_type *type; |
9b40bc90 | 2179 | struct user_namespace *user_ns = current->nsproxy->mnt_ns->user_ns; |
1da177e4 | 2180 | struct vfsmount *mnt; |
15f9a3f3 | 2181 | int err; |
1da177e4 | 2182 | |
0c55cfc4 | 2183 | if (!fstype) |
1da177e4 LT |
2184 | return -EINVAL; |
2185 | ||
0c55cfc4 EB |
2186 | type = get_fs_type(fstype); |
2187 | if (!type) | |
2188 | return -ENODEV; | |
2189 | ||
2190 | if (user_ns != &init_user_ns) { | |
2191 | if (!(type->fs_flags & FS_USERNS_MOUNT)) { | |
2192 | put_filesystem(type); | |
2193 | return -EPERM; | |
2194 | } | |
2195 | /* Only in special cases allow devices from mounts | |
2196 | * created outside the initial user namespace. | |
2197 | */ | |
2198 | if (!(type->fs_flags & FS_USERNS_DEV_MOUNT)) { | |
2199 | flags |= MS_NODEV; | |
187985d9 | 2200 | mnt_flags |= MNT_NODEV | MNT_LOCK_NODEV; |
0c55cfc4 EB |
2201 | } |
2202 | } | |
2203 | ||
2204 | mnt = vfs_kern_mount(type, flags, name, data); | |
2205 | if (!IS_ERR(mnt) && (type->fs_flags & FS_HAS_SUBTYPE) && | |
2206 | !mnt->mnt_sb->s_subtype) | |
2207 | mnt = fs_set_subtype(mnt, fstype); | |
2208 | ||
2209 | put_filesystem(type); | |
1da177e4 LT |
2210 | if (IS_ERR(mnt)) |
2211 | return PTR_ERR(mnt); | |
2212 | ||
95bc5f25 | 2213 | err = do_add_mount(real_mount(mnt), path, mnt_flags); |
15f9a3f3 AV |
2214 | if (err) |
2215 | mntput(mnt); | |
2216 | return err; | |
1da177e4 LT |
2217 | } |
2218 | ||
19a167af AV |
2219 | int finish_automount(struct vfsmount *m, struct path *path) |
2220 | { | |
6776db3d | 2221 | struct mount *mnt = real_mount(m); |
19a167af AV |
2222 | int err; |
2223 | /* The new mount record should have at least 2 refs to prevent it being | |
2224 | * expired before we get a chance to add it | |
2225 | */ | |
6776db3d | 2226 | BUG_ON(mnt_get_count(mnt) < 2); |
19a167af AV |
2227 | |
2228 | if (m->mnt_sb == path->mnt->mnt_sb && | |
2229 | m->mnt_root == path->dentry) { | |
b1e75df4 AV |
2230 | err = -ELOOP; |
2231 | goto fail; | |
19a167af AV |
2232 | } |
2233 | ||
95bc5f25 | 2234 | err = do_add_mount(mnt, path, path->mnt->mnt_flags | MNT_SHRINKABLE); |
b1e75df4 AV |
2235 | if (!err) |
2236 | return 0; | |
2237 | fail: | |
2238 | /* remove m from any expiration list it may be on */ | |
6776db3d | 2239 | if (!list_empty(&mnt->mnt_expire)) { |
97216be0 | 2240 | namespace_lock(); |
962830df | 2241 | br_write_lock(&vfsmount_lock); |
6776db3d | 2242 | list_del_init(&mnt->mnt_expire); |
962830df | 2243 | br_write_unlock(&vfsmount_lock); |
97216be0 | 2244 | namespace_unlock(); |
19a167af | 2245 | } |
b1e75df4 AV |
2246 | mntput(m); |
2247 | mntput(m); | |
19a167af AV |
2248 | return err; |
2249 | } | |
2250 | ||
ea5b778a DH |
2251 | /** |
2252 | * mnt_set_expiry - Put a mount on an expiration list | |
2253 | * @mnt: The mount to list. | |
2254 | * @expiry_list: The list to add the mount to. | |
2255 | */ | |
2256 | void mnt_set_expiry(struct vfsmount *mnt, struct list_head *expiry_list) | |
2257 | { | |
97216be0 | 2258 | namespace_lock(); |
962830df | 2259 | br_write_lock(&vfsmount_lock); |
ea5b778a | 2260 | |
6776db3d | 2261 | list_add_tail(&real_mount(mnt)->mnt_expire, expiry_list); |
ea5b778a | 2262 | |
962830df | 2263 | br_write_unlock(&vfsmount_lock); |
97216be0 | 2264 | namespace_unlock(); |
ea5b778a DH |
2265 | } |
2266 | EXPORT_SYMBOL(mnt_set_expiry); | |
2267 | ||
1da177e4 LT |
2268 | /* |
2269 | * process a list of expirable mountpoints with the intent of discarding any | |
2270 | * mountpoints that aren't in use and haven't been touched since last we came | |
2271 | * here | |
2272 | */ | |
2273 | void mark_mounts_for_expiry(struct list_head *mounts) | |
2274 | { | |
761d5c38 | 2275 | struct mount *mnt, *next; |
1da177e4 LT |
2276 | LIST_HEAD(graveyard); |
2277 | ||
2278 | if (list_empty(mounts)) | |
2279 | return; | |
2280 | ||
97216be0 | 2281 | namespace_lock(); |
962830df | 2282 | br_write_lock(&vfsmount_lock); |
1da177e4 LT |
2283 | |
2284 | /* extract from the expiration list every vfsmount that matches the | |
2285 | * following criteria: | |
2286 | * - only referenced by its parent vfsmount | |
2287 | * - still marked for expiry (marked on the last call here; marks are | |
2288 | * cleared by mntput()) | |
2289 | */ | |
6776db3d | 2290 | list_for_each_entry_safe(mnt, next, mounts, mnt_expire) { |
863d684f | 2291 | if (!xchg(&mnt->mnt_expiry_mark, 1) || |
1ab59738 | 2292 | propagate_mount_busy(mnt, 1)) |
1da177e4 | 2293 | continue; |
6776db3d | 2294 | list_move(&mnt->mnt_expire, &graveyard); |
1da177e4 | 2295 | } |
bcc5c7d2 | 2296 | while (!list_empty(&graveyard)) { |
6776db3d | 2297 | mnt = list_first_entry(&graveyard, struct mount, mnt_expire); |
143c8c91 | 2298 | touch_mnt_namespace(mnt->mnt_ns); |
328e6d90 | 2299 | umount_tree(mnt, 1); |
bcc5c7d2 | 2300 | } |
962830df | 2301 | br_write_unlock(&vfsmount_lock); |
3ab6abee | 2302 | namespace_unlock(); |
5528f911 TM |
2303 | } |
2304 | ||
2305 | EXPORT_SYMBOL_GPL(mark_mounts_for_expiry); | |
2306 | ||
2307 | /* | |
2308 | * Ripoff of 'select_parent()' | |
2309 | * | |
2310 | * search the list of submounts for a given mountpoint, and move any | |
2311 | * shrinkable submounts to the 'graveyard' list. | |
2312 | */ | |
692afc31 | 2313 | static int select_submounts(struct mount *parent, struct list_head *graveyard) |
5528f911 | 2314 | { |
692afc31 | 2315 | struct mount *this_parent = parent; |
5528f911 TM |
2316 | struct list_head *next; |
2317 | int found = 0; | |
2318 | ||
2319 | repeat: | |
6b41d536 | 2320 | next = this_parent->mnt_mounts.next; |
5528f911 | 2321 | resume: |
6b41d536 | 2322 | while (next != &this_parent->mnt_mounts) { |
5528f911 | 2323 | struct list_head *tmp = next; |
6b41d536 | 2324 | struct mount *mnt = list_entry(tmp, struct mount, mnt_child); |
5528f911 TM |
2325 | |
2326 | next = tmp->next; | |
692afc31 | 2327 | if (!(mnt->mnt.mnt_flags & MNT_SHRINKABLE)) |
1da177e4 | 2328 | continue; |
5528f911 TM |
2329 | /* |
2330 | * Descend a level if the d_mounts list is non-empty. | |
2331 | */ | |
6b41d536 | 2332 | if (!list_empty(&mnt->mnt_mounts)) { |
5528f911 TM |
2333 | this_parent = mnt; |
2334 | goto repeat; | |
2335 | } | |
1da177e4 | 2336 | |
1ab59738 | 2337 | if (!propagate_mount_busy(mnt, 1)) { |
6776db3d | 2338 | list_move_tail(&mnt->mnt_expire, graveyard); |
5528f911 TM |
2339 | found++; |
2340 | } | |
1da177e4 | 2341 | } |
5528f911 TM |
2342 | /* |
2343 | * All done at this level ... ascend and resume the search | |
2344 | */ | |
2345 | if (this_parent != parent) { | |
6b41d536 | 2346 | next = this_parent->mnt_child.next; |
0714a533 | 2347 | this_parent = this_parent->mnt_parent; |
5528f911 TM |
2348 | goto resume; |
2349 | } | |
2350 | return found; | |
2351 | } | |
2352 | ||
2353 | /* | |
2354 | * process a list of expirable mountpoints with the intent of discarding any | |
2355 | * submounts of a specific parent mountpoint | |
99b7db7b NP |
2356 | * |
2357 | * vfsmount_lock must be held for write | |
5528f911 | 2358 | */ |
b54b9be7 | 2359 | static void shrink_submounts(struct mount *mnt) |
5528f911 TM |
2360 | { |
2361 | LIST_HEAD(graveyard); | |
761d5c38 | 2362 | struct mount *m; |
5528f911 | 2363 | |
5528f911 | 2364 | /* extract submounts of 'mountpoint' from the expiration list */ |
c35038be | 2365 | while (select_submounts(mnt, &graveyard)) { |
bcc5c7d2 | 2366 | while (!list_empty(&graveyard)) { |
761d5c38 | 2367 | m = list_first_entry(&graveyard, struct mount, |
6776db3d | 2368 | mnt_expire); |
143c8c91 | 2369 | touch_mnt_namespace(m->mnt_ns); |
328e6d90 | 2370 | umount_tree(m, 1); |
bcc5c7d2 AV |
2371 | } |
2372 | } | |
1da177e4 LT |
2373 | } |
2374 | ||
1da177e4 LT |
2375 | /* |
2376 | * Some copy_from_user() implementations do not return the exact number of | |
2377 | * bytes remaining to copy on a fault. But copy_mount_options() requires that. | |
2378 | * Note that this function differs from copy_from_user() in that it will oops | |
2379 | * on bad values of `to', rather than returning a short copy. | |
2380 | */ | |
b58fed8b RP |
2381 | static long exact_copy_from_user(void *to, const void __user * from, |
2382 | unsigned long n) | |
1da177e4 LT |
2383 | { |
2384 | char *t = to; | |
2385 | const char __user *f = from; | |
2386 | char c; | |
2387 | ||
2388 | if (!access_ok(VERIFY_READ, from, n)) | |
2389 | return n; | |
2390 | ||
2391 | while (n) { | |
2392 | if (__get_user(c, f)) { | |
2393 | memset(t, 0, n); | |
2394 | break; | |
2395 | } | |
2396 | *t++ = c; | |
2397 | f++; | |
2398 | n--; | |
2399 | } | |
2400 | return n; | |
2401 | } | |
2402 | ||
b58fed8b | 2403 | int copy_mount_options(const void __user * data, unsigned long *where) |
1da177e4 LT |
2404 | { |
2405 | int i; | |
2406 | unsigned long page; | |
2407 | unsigned long size; | |
b58fed8b | 2408 | |
1da177e4 LT |
2409 | *where = 0; |
2410 | if (!data) | |
2411 | return 0; | |
2412 | ||
2413 | if (!(page = __get_free_page(GFP_KERNEL))) | |
2414 | return -ENOMEM; | |
2415 | ||
2416 | /* We only care that *some* data at the address the user | |
2417 | * gave us is valid. Just in case, we'll zero | |
2418 | * the remainder of the page. | |
2419 | */ | |
2420 | /* copy_from_user cannot cross TASK_SIZE ! */ | |
2421 | size = TASK_SIZE - (unsigned long)data; | |
2422 | if (size > PAGE_SIZE) | |
2423 | size = PAGE_SIZE; | |
2424 | ||
2425 | i = size - exact_copy_from_user((void *)page, data, size); | |
2426 | if (!i) { | |
b58fed8b | 2427 | free_page(page); |
1da177e4 LT |
2428 | return -EFAULT; |
2429 | } | |
2430 | if (i != PAGE_SIZE) | |
2431 | memset((char *)page + i, 0, PAGE_SIZE - i); | |
2432 | *where = page; | |
2433 | return 0; | |
2434 | } | |
2435 | ||
eca6f534 VN |
2436 | int copy_mount_string(const void __user *data, char **where) |
2437 | { | |
2438 | char *tmp; | |
2439 | ||
2440 | if (!data) { | |
2441 | *where = NULL; | |
2442 | return 0; | |
2443 | } | |
2444 | ||
2445 | tmp = strndup_user(data, PAGE_SIZE); | |
2446 | if (IS_ERR(tmp)) | |
2447 | return PTR_ERR(tmp); | |
2448 | ||
2449 | *where = tmp; | |
2450 | return 0; | |
2451 | } | |
2452 | ||
1da177e4 LT |
2453 | /* |
2454 | * Flags is a 32-bit value that allows up to 31 non-fs dependent flags to | |
2455 | * be given to the mount() call (ie: read-only, no-dev, no-suid etc). | |
2456 | * | |
2457 | * data is a (void *) that can point to any structure up to | |
2458 | * PAGE_SIZE-1 bytes, which can contain arbitrary fs-dependent | |
2459 | * information (or be NULL). | |
2460 | * | |
2461 | * Pre-0.97 versions of mount() didn't have a flags word. | |
2462 | * When the flags word was introduced its top half was required | |
2463 | * to have the magic value 0xC0ED, and this remained so until 2.4.0-test9. | |
2464 | * Therefore, if this magic number is present, it carries no information | |
2465 | * and must be discarded. | |
2466 | */ | |
808d4e3c AV |
2467 | long do_mount(const char *dev_name, const char *dir_name, |
2468 | const char *type_page, unsigned long flags, void *data_page) | |
1da177e4 | 2469 | { |
2d92ab3c | 2470 | struct path path; |
1da177e4 LT |
2471 | int retval = 0; |
2472 | int mnt_flags = 0; | |
2473 | ||
2474 | /* Discard magic */ | |
2475 | if ((flags & MS_MGC_MSK) == MS_MGC_VAL) | |
2476 | flags &= ~MS_MGC_MSK; | |
2477 | ||
2478 | /* Basic sanity checks */ | |
2479 | ||
2480 | if (!dir_name || !*dir_name || !memchr(dir_name, 0, PAGE_SIZE)) | |
2481 | return -EINVAL; | |
1da177e4 LT |
2482 | |
2483 | if (data_page) | |
2484 | ((char *)data_page)[PAGE_SIZE - 1] = 0; | |
2485 | ||
a27ab9f2 TH |
2486 | /* ... and get the mountpoint */ |
2487 | retval = kern_path(dir_name, LOOKUP_FOLLOW, &path); | |
2488 | if (retval) | |
2489 | return retval; | |
2490 | ||
2491 | retval = security_sb_mount(dev_name, &path, | |
2492 | type_page, flags, data_page); | |
0d5cadb8 AV |
2493 | if (!retval && !may_mount()) |
2494 | retval = -EPERM; | |
a27ab9f2 TH |
2495 | if (retval) |
2496 | goto dput_out; | |
2497 | ||
613cbe3d AK |
2498 | /* Default to relatime unless overriden */ |
2499 | if (!(flags & MS_NOATIME)) | |
2500 | mnt_flags |= MNT_RELATIME; | |
0a1c01c9 | 2501 | |
1da177e4 LT |
2502 | /* Separate the per-mountpoint flags */ |
2503 | if (flags & MS_NOSUID) | |
2504 | mnt_flags |= MNT_NOSUID; | |
2505 | if (flags & MS_NODEV) | |
2506 | mnt_flags |= MNT_NODEV; | |
2507 | if (flags & MS_NOEXEC) | |
2508 | mnt_flags |= MNT_NOEXEC; | |
fc33a7bb CH |
2509 | if (flags & MS_NOATIME) |
2510 | mnt_flags |= MNT_NOATIME; | |
2511 | if (flags & MS_NODIRATIME) | |
2512 | mnt_flags |= MNT_NODIRATIME; | |
d0adde57 MG |
2513 | if (flags & MS_STRICTATIME) |
2514 | mnt_flags &= ~(MNT_RELATIME | MNT_NOATIME); | |
2e4b7fcd DH |
2515 | if (flags & MS_RDONLY) |
2516 | mnt_flags |= MNT_READONLY; | |
fc33a7bb | 2517 | |
99dd97b8 EB |
2518 | /* The default atime for remount is preservation */ |
2519 | if ((flags & MS_REMOUNT) && | |
2520 | ((flags & (MS_NOATIME | MS_NODIRATIME | MS_RELATIME | | |
2521 | MS_STRICTATIME)) == 0)) { | |
2522 | mnt_flags &= ~MNT_ATIME_MASK; | |
2523 | mnt_flags |= path.mnt->mnt_flags & MNT_ATIME_MASK; | |
2524 | } | |
2525 | ||
7a4dec53 | 2526 | flags &= ~(MS_NOSUID | MS_NOEXEC | MS_NODEV | MS_ACTIVE | MS_BORN | |
d0adde57 MG |
2527 | MS_NOATIME | MS_NODIRATIME | MS_RELATIME| MS_KERNMOUNT | |
2528 | MS_STRICTATIME); | |
1da177e4 | 2529 | |
1da177e4 | 2530 | if (flags & MS_REMOUNT) |
2d92ab3c | 2531 | retval = do_remount(&path, flags & ~MS_REMOUNT, mnt_flags, |
1da177e4 LT |
2532 | data_page); |
2533 | else if (flags & MS_BIND) | |
2d92ab3c | 2534 | retval = do_loopback(&path, dev_name, flags & MS_REC); |
9676f0c6 | 2535 | else if (flags & (MS_SHARED | MS_PRIVATE | MS_SLAVE | MS_UNBINDABLE)) |
2d92ab3c | 2536 | retval = do_change_type(&path, flags); |
1da177e4 | 2537 | else if (flags & MS_MOVE) |
2d92ab3c | 2538 | retval = do_move_mount(&path, dev_name); |
1da177e4 | 2539 | else |
2d92ab3c | 2540 | retval = do_new_mount(&path, type_page, flags, mnt_flags, |
1da177e4 LT |
2541 | dev_name, data_page); |
2542 | dput_out: | |
2d92ab3c | 2543 | path_put(&path); |
1da177e4 LT |
2544 | return retval; |
2545 | } | |
2546 | ||
771b1371 EB |
2547 | static void free_mnt_ns(struct mnt_namespace *ns) |
2548 | { | |
98f842e6 | 2549 | proc_free_inum(ns->proc_inum); |
771b1371 EB |
2550 | put_user_ns(ns->user_ns); |
2551 | kfree(ns); | |
2552 | } | |
2553 | ||
8823c079 EB |
2554 | /* |
2555 | * Assign a sequence number so we can detect when we attempt to bind | |
2556 | * mount a reference to an older mount namespace into the current | |
2557 | * mount namespace, preventing reference counting loops. A 64bit | |
2558 | * number incrementing at 10Ghz will take 12,427 years to wrap which | |
2559 | * is effectively never, so we can ignore the possibility. | |
2560 | */ | |
2561 | static atomic64_t mnt_ns_seq = ATOMIC64_INIT(1); | |
2562 | ||
771b1371 | 2563 | static struct mnt_namespace *alloc_mnt_ns(struct user_namespace *user_ns) |
cf8d2c11 TM |
2564 | { |
2565 | struct mnt_namespace *new_ns; | |
98f842e6 | 2566 | int ret; |
cf8d2c11 TM |
2567 | |
2568 | new_ns = kmalloc(sizeof(struct mnt_namespace), GFP_KERNEL); | |
2569 | if (!new_ns) | |
2570 | return ERR_PTR(-ENOMEM); | |
98f842e6 EB |
2571 | ret = proc_alloc_inum(&new_ns->proc_inum); |
2572 | if (ret) { | |
2573 | kfree(new_ns); | |
2574 | return ERR_PTR(ret); | |
2575 | } | |
8823c079 | 2576 | new_ns->seq = atomic64_add_return(1, &mnt_ns_seq); |
cf8d2c11 TM |
2577 | atomic_set(&new_ns->count, 1); |
2578 | new_ns->root = NULL; | |
2579 | INIT_LIST_HEAD(&new_ns->list); | |
2580 | init_waitqueue_head(&new_ns->poll); | |
2581 | new_ns->event = 0; | |
771b1371 | 2582 | new_ns->user_ns = get_user_ns(user_ns); |
cf8d2c11 TM |
2583 | return new_ns; |
2584 | } | |
2585 | ||
741a2951 JD |
2586 | /* |
2587 | * Allocate a new namespace structure and populate it with contents | |
2588 | * copied from the namespace of the passed in task structure. | |
2589 | */ | |
e3222c4e | 2590 | static struct mnt_namespace *dup_mnt_ns(struct mnt_namespace *mnt_ns, |
771b1371 | 2591 | struct user_namespace *user_ns, struct fs_struct *fs) |
1da177e4 | 2592 | { |
6b3286ed | 2593 | struct mnt_namespace *new_ns; |
7f2da1e7 | 2594 | struct vfsmount *rootmnt = NULL, *pwdmnt = NULL; |
315fc83e | 2595 | struct mount *p, *q; |
be08d6d2 | 2596 | struct mount *old = mnt_ns->root; |
cb338d06 | 2597 | struct mount *new; |
7a472ef4 | 2598 | int copy_flags; |
1da177e4 | 2599 | |
771b1371 | 2600 | new_ns = alloc_mnt_ns(user_ns); |
cf8d2c11 TM |
2601 | if (IS_ERR(new_ns)) |
2602 | return new_ns; | |
1da177e4 | 2603 | |
97216be0 | 2604 | namespace_lock(); |
1da177e4 | 2605 | /* First pass: copy the tree topology */ |
7a472ef4 EB |
2606 | copy_flags = CL_COPY_ALL | CL_EXPIRE; |
2607 | if (user_ns != mnt_ns->user_ns) | |
132c94e3 | 2608 | copy_flags |= CL_SHARED_TO_SLAVE | CL_UNPRIVILEGED; |
7a472ef4 | 2609 | new = copy_tree(old, old->mnt.mnt_root, copy_flags); |
be34d1a3 | 2610 | if (IS_ERR(new)) { |
328e6d90 | 2611 | namespace_unlock(); |
771b1371 | 2612 | free_mnt_ns(new_ns); |
be34d1a3 | 2613 | return ERR_CAST(new); |
1da177e4 | 2614 | } |
be08d6d2 | 2615 | new_ns->root = new; |
962830df | 2616 | br_write_lock(&vfsmount_lock); |
1a4eeaf2 | 2617 | list_add_tail(&new_ns->list, &new->mnt_list); |
962830df | 2618 | br_write_unlock(&vfsmount_lock); |
1da177e4 LT |
2619 | |
2620 | /* | |
2621 | * Second pass: switch the tsk->fs->* elements and mark new vfsmounts | |
2622 | * as belonging to new namespace. We have already acquired a private | |
2623 | * fs_struct, so tsk->fs->lock is not needed. | |
2624 | */ | |
909b0a88 | 2625 | p = old; |
cb338d06 | 2626 | q = new; |
1da177e4 | 2627 | while (p) { |
143c8c91 | 2628 | q->mnt_ns = new_ns; |
1da177e4 | 2629 | if (fs) { |
315fc83e AV |
2630 | if (&p->mnt == fs->root.mnt) { |
2631 | fs->root.mnt = mntget(&q->mnt); | |
315fc83e | 2632 | rootmnt = &p->mnt; |
1da177e4 | 2633 | } |
315fc83e AV |
2634 | if (&p->mnt == fs->pwd.mnt) { |
2635 | fs->pwd.mnt = mntget(&q->mnt); | |
315fc83e | 2636 | pwdmnt = &p->mnt; |
1da177e4 | 2637 | } |
1da177e4 | 2638 | } |
909b0a88 AV |
2639 | p = next_mnt(p, old); |
2640 | q = next_mnt(q, new); | |
1da177e4 | 2641 | } |
328e6d90 | 2642 | namespace_unlock(); |
1da177e4 | 2643 | |
1da177e4 | 2644 | if (rootmnt) |
f03c6599 | 2645 | mntput(rootmnt); |
1da177e4 | 2646 | if (pwdmnt) |
f03c6599 | 2647 | mntput(pwdmnt); |
1da177e4 | 2648 | |
741a2951 JD |
2649 | return new_ns; |
2650 | } | |
2651 | ||
213dd266 | 2652 | struct mnt_namespace *copy_mnt_ns(unsigned long flags, struct mnt_namespace *ns, |
771b1371 | 2653 | struct user_namespace *user_ns, struct fs_struct *new_fs) |
741a2951 | 2654 | { |
6b3286ed | 2655 | struct mnt_namespace *new_ns; |
741a2951 | 2656 | |
e3222c4e | 2657 | BUG_ON(!ns); |
6b3286ed | 2658 | get_mnt_ns(ns); |
741a2951 JD |
2659 | |
2660 | if (!(flags & CLONE_NEWNS)) | |
e3222c4e | 2661 | return ns; |
741a2951 | 2662 | |
771b1371 | 2663 | new_ns = dup_mnt_ns(ns, user_ns, new_fs); |
741a2951 | 2664 | |
6b3286ed | 2665 | put_mnt_ns(ns); |
e3222c4e | 2666 | return new_ns; |
1da177e4 LT |
2667 | } |
2668 | ||
cf8d2c11 TM |
2669 | /** |
2670 | * create_mnt_ns - creates a private namespace and adds a root filesystem | |
2671 | * @mnt: pointer to the new root filesystem mountpoint | |
2672 | */ | |
1a4eeaf2 | 2673 | static struct mnt_namespace *create_mnt_ns(struct vfsmount *m) |
cf8d2c11 | 2674 | { |
771b1371 | 2675 | struct mnt_namespace *new_ns = alloc_mnt_ns(&init_user_ns); |
cf8d2c11 | 2676 | if (!IS_ERR(new_ns)) { |
1a4eeaf2 AV |
2677 | struct mount *mnt = real_mount(m); |
2678 | mnt->mnt_ns = new_ns; | |
be08d6d2 | 2679 | new_ns->root = mnt; |
b1983cd8 | 2680 | list_add(&mnt->mnt_list, &new_ns->list); |
c1334495 | 2681 | } else { |
1a4eeaf2 | 2682 | mntput(m); |
cf8d2c11 TM |
2683 | } |
2684 | return new_ns; | |
2685 | } | |
cf8d2c11 | 2686 | |
ea441d11 AV |
2687 | struct dentry *mount_subtree(struct vfsmount *mnt, const char *name) |
2688 | { | |
2689 | struct mnt_namespace *ns; | |
d31da0f0 | 2690 | struct super_block *s; |
ea441d11 AV |
2691 | struct path path; |
2692 | int err; | |
2693 | ||
2694 | ns = create_mnt_ns(mnt); | |
2695 | if (IS_ERR(ns)) | |
2696 | return ERR_CAST(ns); | |
2697 | ||
2698 | err = vfs_path_lookup(mnt->mnt_root, mnt, | |
2699 | name, LOOKUP_FOLLOW|LOOKUP_AUTOMOUNT, &path); | |
2700 | ||
2701 | put_mnt_ns(ns); | |
2702 | ||
2703 | if (err) | |
2704 | return ERR_PTR(err); | |
2705 | ||
2706 | /* trade a vfsmount reference for active sb one */ | |
d31da0f0 AV |
2707 | s = path.mnt->mnt_sb; |
2708 | atomic_inc(&s->s_active); | |
ea441d11 AV |
2709 | mntput(path.mnt); |
2710 | /* lock the sucker */ | |
d31da0f0 | 2711 | down_write(&s->s_umount); |
ea441d11 AV |
2712 | /* ... and return the root of (sub)tree on it */ |
2713 | return path.dentry; | |
2714 | } | |
2715 | EXPORT_SYMBOL(mount_subtree); | |
2716 | ||
bdc480e3 HC |
2717 | SYSCALL_DEFINE5(mount, char __user *, dev_name, char __user *, dir_name, |
2718 | char __user *, type, unsigned long, flags, void __user *, data) | |
1da177e4 | 2719 | { |
eca6f534 VN |
2720 | int ret; |
2721 | char *kernel_type; | |
91a27b2a | 2722 | struct filename *kernel_dir; |
eca6f534 | 2723 | char *kernel_dev; |
1da177e4 | 2724 | unsigned long data_page; |
1da177e4 | 2725 | |
eca6f534 VN |
2726 | ret = copy_mount_string(type, &kernel_type); |
2727 | if (ret < 0) | |
2728 | goto out_type; | |
1da177e4 | 2729 | |
eca6f534 VN |
2730 | kernel_dir = getname(dir_name); |
2731 | if (IS_ERR(kernel_dir)) { | |
2732 | ret = PTR_ERR(kernel_dir); | |
2733 | goto out_dir; | |
2734 | } | |
1da177e4 | 2735 | |
eca6f534 VN |
2736 | ret = copy_mount_string(dev_name, &kernel_dev); |
2737 | if (ret < 0) | |
2738 | goto out_dev; | |
1da177e4 | 2739 | |
eca6f534 VN |
2740 | ret = copy_mount_options(data, &data_page); |
2741 | if (ret < 0) | |
2742 | goto out_data; | |
1da177e4 | 2743 | |
91a27b2a | 2744 | ret = do_mount(kernel_dev, kernel_dir->name, kernel_type, flags, |
eca6f534 | 2745 | (void *) data_page); |
1da177e4 | 2746 | |
eca6f534 VN |
2747 | free_page(data_page); |
2748 | out_data: | |
2749 | kfree(kernel_dev); | |
2750 | out_dev: | |
2751 | putname(kernel_dir); | |
2752 | out_dir: | |
2753 | kfree(kernel_type); | |
2754 | out_type: | |
2755 | return ret; | |
1da177e4 LT |
2756 | } |
2757 | ||
afac7cba AV |
2758 | /* |
2759 | * Return true if path is reachable from root | |
2760 | * | |
2761 | * namespace_sem or vfsmount_lock is held | |
2762 | */ | |
643822b4 | 2763 | bool is_path_reachable(struct mount *mnt, struct dentry *dentry, |
afac7cba AV |
2764 | const struct path *root) |
2765 | { | |
643822b4 | 2766 | while (&mnt->mnt != root->mnt && mnt_has_parent(mnt)) { |
a73324da | 2767 | dentry = mnt->mnt_mountpoint; |
0714a533 | 2768 | mnt = mnt->mnt_parent; |
afac7cba | 2769 | } |
643822b4 | 2770 | return &mnt->mnt == root->mnt && is_subdir(dentry, root->dentry); |
afac7cba AV |
2771 | } |
2772 | ||
2773 | int path_is_under(struct path *path1, struct path *path2) | |
2774 | { | |
2775 | int res; | |
962830df | 2776 | br_read_lock(&vfsmount_lock); |
643822b4 | 2777 | res = is_path_reachable(real_mount(path1->mnt), path1->dentry, path2); |
962830df | 2778 | br_read_unlock(&vfsmount_lock); |
afac7cba AV |
2779 | return res; |
2780 | } | |
2781 | EXPORT_SYMBOL(path_is_under); | |
2782 | ||
1da177e4 LT |
2783 | /* |
2784 | * pivot_root Semantics: | |
2785 | * Moves the root file system of the current process to the directory put_old, | |
2786 | * makes new_root as the new root file system of the current process, and sets | |
2787 | * root/cwd of all processes which had them on the current root to new_root. | |
2788 | * | |
2789 | * Restrictions: | |
2790 | * The new_root and put_old must be directories, and must not be on the | |
2791 | * same file system as the current process root. The put_old must be | |
2792 | * underneath new_root, i.e. adding a non-zero number of /.. to the string | |
2793 | * pointed to by put_old must yield the same directory as new_root. No other | |
2794 | * file system may be mounted on put_old. After all, new_root is a mountpoint. | |
2795 | * | |
4a0d11fa NB |
2796 | * Also, the current root cannot be on the 'rootfs' (initial ramfs) filesystem. |
2797 | * See Documentation/filesystems/ramfs-rootfs-initramfs.txt for alternatives | |
2798 | * in this situation. | |
2799 | * | |
1da177e4 LT |
2800 | * Notes: |
2801 | * - we don't move root/cwd if they are not at the root (reason: if something | |
2802 | * cared enough to change them, it's probably wrong to force them elsewhere) | |
2803 | * - it's okay to pick a root that isn't the root of a file system, e.g. | |
2804 | * /nfs/my_root where /nfs is the mount point. It must be a mountpoint, | |
2805 | * though, so you may need to say mount --bind /nfs/my_root /nfs/my_root | |
2806 | * first. | |
2807 | */ | |
3480b257 HC |
2808 | SYSCALL_DEFINE2(pivot_root, const char __user *, new_root, |
2809 | const char __user *, put_old) | |
1da177e4 | 2810 | { |
2d8f3038 | 2811 | struct path new, old, parent_path, root_parent, root; |
84d17192 AV |
2812 | struct mount *new_mnt, *root_mnt, *old_mnt; |
2813 | struct mountpoint *old_mp, *root_mp; | |
1da177e4 LT |
2814 | int error; |
2815 | ||
9b40bc90 | 2816 | if (!may_mount()) |
1da177e4 LT |
2817 | return -EPERM; |
2818 | ||
2d8f3038 | 2819 | error = user_path_dir(new_root, &new); |
1da177e4 LT |
2820 | if (error) |
2821 | goto out0; | |
1da177e4 | 2822 | |
2d8f3038 | 2823 | error = user_path_dir(put_old, &old); |
1da177e4 LT |
2824 | if (error) |
2825 | goto out1; | |
2826 | ||
2d8f3038 | 2827 | error = security_sb_pivotroot(&old, &new); |
b12cea91 AV |
2828 | if (error) |
2829 | goto out2; | |
1da177e4 | 2830 | |
f7ad3c6b | 2831 | get_fs_root(current->fs, &root); |
84d17192 AV |
2832 | old_mp = lock_mount(&old); |
2833 | error = PTR_ERR(old_mp); | |
2834 | if (IS_ERR(old_mp)) | |
b12cea91 AV |
2835 | goto out3; |
2836 | ||
1da177e4 | 2837 | error = -EINVAL; |
419148da AV |
2838 | new_mnt = real_mount(new.mnt); |
2839 | root_mnt = real_mount(root.mnt); | |
84d17192 AV |
2840 | old_mnt = real_mount(old.mnt); |
2841 | if (IS_MNT_SHARED(old_mnt) || | |
fc7be130 AV |
2842 | IS_MNT_SHARED(new_mnt->mnt_parent) || |
2843 | IS_MNT_SHARED(root_mnt->mnt_parent)) | |
b12cea91 | 2844 | goto out4; |
143c8c91 | 2845 | if (!check_mnt(root_mnt) || !check_mnt(new_mnt)) |
b12cea91 | 2846 | goto out4; |
1da177e4 | 2847 | error = -ENOENT; |
f3da392e | 2848 | if (d_unlinked(new.dentry)) |
b12cea91 | 2849 | goto out4; |
1da177e4 | 2850 | error = -EBUSY; |
84d17192 | 2851 | if (new_mnt == root_mnt || old_mnt == root_mnt) |
b12cea91 | 2852 | goto out4; /* loop, on the same file system */ |
1da177e4 | 2853 | error = -EINVAL; |
8c3ee42e | 2854 | if (root.mnt->mnt_root != root.dentry) |
b12cea91 | 2855 | goto out4; /* not a mountpoint */ |
676da58d | 2856 | if (!mnt_has_parent(root_mnt)) |
b12cea91 | 2857 | goto out4; /* not attached */ |
84d17192 | 2858 | root_mp = root_mnt->mnt_mp; |
2d8f3038 | 2859 | if (new.mnt->mnt_root != new.dentry) |
b12cea91 | 2860 | goto out4; /* not a mountpoint */ |
676da58d | 2861 | if (!mnt_has_parent(new_mnt)) |
b12cea91 | 2862 | goto out4; /* not attached */ |
4ac91378 | 2863 | /* make sure we can reach put_old from new_root */ |
84d17192 | 2864 | if (!is_path_reachable(old_mnt, old.dentry, &new)) |
b12cea91 | 2865 | goto out4; |
315a75ea EB |
2866 | /* make certain new is below the root */ |
2867 | if (!is_path_reachable(new_mnt, new.dentry, &root)) | |
2868 | goto out4; | |
84d17192 | 2869 | root_mp->m_count++; /* pin it so it won't go away */ |
962830df | 2870 | br_write_lock(&vfsmount_lock); |
419148da AV |
2871 | detach_mnt(new_mnt, &parent_path); |
2872 | detach_mnt(root_mnt, &root_parent); | |
4ac91378 | 2873 | /* mount old root on put_old */ |
84d17192 | 2874 | attach_mnt(root_mnt, old_mnt, old_mp); |
4ac91378 | 2875 | /* mount new_root on / */ |
84d17192 | 2876 | attach_mnt(new_mnt, real_mount(root_parent.mnt), root_mp); |
6b3286ed | 2877 | touch_mnt_namespace(current->nsproxy->mnt_ns); |
962830df | 2878 | br_write_unlock(&vfsmount_lock); |
2d8f3038 | 2879 | chroot_fs_refs(&root, &new); |
84d17192 | 2880 | put_mountpoint(root_mp); |
1da177e4 | 2881 | error = 0; |
b12cea91 | 2882 | out4: |
84d17192 | 2883 | unlock_mount(old_mp); |
b12cea91 AV |
2884 | if (!error) { |
2885 | path_put(&root_parent); | |
2886 | path_put(&parent_path); | |
2887 | } | |
2888 | out3: | |
8c3ee42e | 2889 | path_put(&root); |
b12cea91 | 2890 | out2: |
2d8f3038 | 2891 | path_put(&old); |
1da177e4 | 2892 | out1: |
2d8f3038 | 2893 | path_put(&new); |
1da177e4 | 2894 | out0: |
1da177e4 | 2895 | return error; |
1da177e4 LT |
2896 | } |
2897 | ||
2898 | static void __init init_mount_tree(void) | |
2899 | { | |
2900 | struct vfsmount *mnt; | |
6b3286ed | 2901 | struct mnt_namespace *ns; |
ac748a09 | 2902 | struct path root; |
0c55cfc4 | 2903 | struct file_system_type *type; |
1da177e4 | 2904 | |
0c55cfc4 EB |
2905 | type = get_fs_type("rootfs"); |
2906 | if (!type) | |
2907 | panic("Can't find rootfs type"); | |
2908 | mnt = vfs_kern_mount(type, 0, "rootfs", NULL); | |
2909 | put_filesystem(type); | |
1da177e4 LT |
2910 | if (IS_ERR(mnt)) |
2911 | panic("Can't create rootfs"); | |
b3e19d92 | 2912 | |
3b22edc5 TM |
2913 | ns = create_mnt_ns(mnt); |
2914 | if (IS_ERR(ns)) | |
1da177e4 | 2915 | panic("Can't allocate initial namespace"); |
6b3286ed KK |
2916 | |
2917 | init_task.nsproxy->mnt_ns = ns; | |
2918 | get_mnt_ns(ns); | |
2919 | ||
be08d6d2 AV |
2920 | root.mnt = mnt; |
2921 | root.dentry = mnt->mnt_root; | |
ac748a09 JB |
2922 | |
2923 | set_fs_pwd(current->fs, &root); | |
2924 | set_fs_root(current->fs, &root); | |
1da177e4 LT |
2925 | } |
2926 | ||
74bf17cf | 2927 | void __init mnt_init(void) |
1da177e4 | 2928 | { |
13f14b4d | 2929 | unsigned u; |
15a67dd8 | 2930 | int err; |
1da177e4 | 2931 | |
390c6843 RP |
2932 | init_rwsem(&namespace_sem); |
2933 | ||
7d6fec45 | 2934 | mnt_cache = kmem_cache_create("mnt_cache", sizeof(struct mount), |
20c2df83 | 2935 | 0, SLAB_HWCACHE_ALIGN | SLAB_PANIC, NULL); |
1da177e4 | 2936 | |
b58fed8b | 2937 | mount_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
84d17192 | 2938 | mountpoint_hashtable = (struct list_head *)__get_free_page(GFP_ATOMIC); |
1da177e4 | 2939 | |
84d17192 | 2940 | if (!mount_hashtable || !mountpoint_hashtable) |
1da177e4 LT |
2941 | panic("Failed to allocate mount hash table\n"); |
2942 | ||
80cdc6da | 2943 | printk(KERN_INFO "Mount-cache hash table entries: %lu\n", HASH_SIZE); |
13f14b4d ED |
2944 | |
2945 | for (u = 0; u < HASH_SIZE; u++) | |
2946 | INIT_LIST_HEAD(&mount_hashtable[u]); | |
84d17192 AV |
2947 | for (u = 0; u < HASH_SIZE; u++) |
2948 | INIT_LIST_HEAD(&mountpoint_hashtable[u]); | |
1da177e4 | 2949 | |
962830df | 2950 | br_lock_init(&vfsmount_lock); |
99b7db7b | 2951 | |
15a67dd8 RD |
2952 | err = sysfs_init(); |
2953 | if (err) | |
2954 | printk(KERN_WARNING "%s: sysfs_init error: %d\n", | |
8e24eea7 | 2955 | __func__, err); |
00d26666 GKH |
2956 | fs_kobj = kobject_create_and_add("fs", NULL); |
2957 | if (!fs_kobj) | |
8e24eea7 | 2958 | printk(KERN_WARNING "%s: kobj create error\n", __func__); |
1da177e4 LT |
2959 | init_rootfs(); |
2960 | init_mount_tree(); | |
2961 | } | |
2962 | ||
616511d0 | 2963 | void put_mnt_ns(struct mnt_namespace *ns) |
1da177e4 | 2964 | { |
d498b25a | 2965 | if (!atomic_dec_and_test(&ns->count)) |
616511d0 | 2966 | return; |
97216be0 | 2967 | namespace_lock(); |
962830df | 2968 | br_write_lock(&vfsmount_lock); |
328e6d90 | 2969 | umount_tree(ns->root, 0); |
962830df | 2970 | br_write_unlock(&vfsmount_lock); |
3ab6abee | 2971 | namespace_unlock(); |
771b1371 | 2972 | free_mnt_ns(ns); |
1da177e4 | 2973 | } |
9d412a43 AV |
2974 | |
2975 | struct vfsmount *kern_mount_data(struct file_system_type *type, void *data) | |
2976 | { | |
423e0ab0 TC |
2977 | struct vfsmount *mnt; |
2978 | mnt = vfs_kern_mount(type, MS_KERNMOUNT, type->name, data); | |
2979 | if (!IS_ERR(mnt)) { | |
2980 | /* | |
2981 | * it is a longterm mount, don't release mnt until | |
2982 | * we unmount before file sys is unregistered | |
2983 | */ | |
f7a99c5b | 2984 | real_mount(mnt)->mnt_ns = MNT_NS_INTERNAL; |
423e0ab0 TC |
2985 | } |
2986 | return mnt; | |
9d412a43 AV |
2987 | } |
2988 | EXPORT_SYMBOL_GPL(kern_mount_data); | |
423e0ab0 TC |
2989 | |
2990 | void kern_unmount(struct vfsmount *mnt) | |
2991 | { | |
2992 | /* release long term mount so mount point can be released */ | |
2993 | if (!IS_ERR_OR_NULL(mnt)) { | |
f7a99c5b AV |
2994 | br_write_lock(&vfsmount_lock); |
2995 | real_mount(mnt)->mnt_ns = NULL; | |
2996 | br_write_unlock(&vfsmount_lock); | |
423e0ab0 TC |
2997 | mntput(mnt); |
2998 | } | |
2999 | } | |
3000 | EXPORT_SYMBOL(kern_unmount); | |
02125a82 AV |
3001 | |
3002 | bool our_mnt(struct vfsmount *mnt) | |
3003 | { | |
143c8c91 | 3004 | return check_mnt(real_mount(mnt)); |
02125a82 | 3005 | } |
8823c079 | 3006 | |
3151527e EB |
3007 | bool current_chrooted(void) |
3008 | { | |
3009 | /* Does the current process have a non-standard root */ | |
3010 | struct path ns_root; | |
3011 | struct path fs_root; | |
3012 | bool chrooted; | |
3013 | ||
3014 | /* Find the namespace root */ | |
3015 | ns_root.mnt = ¤t->nsproxy->mnt_ns->root->mnt; | |
3016 | ns_root.dentry = ns_root.mnt->mnt_root; | |
3017 | path_get(&ns_root); | |
3018 | while (d_mountpoint(ns_root.dentry) && follow_down_one(&ns_root)) | |
3019 | ; | |
3020 | ||
3021 | get_fs_root(current->fs, &fs_root); | |
3022 | ||
3023 | chrooted = !path_equal(&fs_root, &ns_root); | |
3024 | ||
3025 | path_put(&fs_root); | |
3026 | path_put(&ns_root); | |
3027 | ||
3028 | return chrooted; | |
3029 | } | |
3030 | ||
87a8ebd6 EB |
3031 | void update_mnt_policy(struct user_namespace *userns) |
3032 | { | |
3033 | struct mnt_namespace *ns = current->nsproxy->mnt_ns; | |
3034 | struct mount *mnt; | |
3035 | ||
3036 | down_read(&namespace_sem); | |
3037 | list_for_each_entry(mnt, &ns->list, mnt_list) { | |
3038 | switch (mnt->mnt.mnt_sb->s_magic) { | |
3039 | case SYSFS_MAGIC: | |
3040 | userns->may_mount_sysfs = true; | |
3041 | break; | |
3042 | case PROC_SUPER_MAGIC: | |
3043 | userns->may_mount_proc = true; | |
3044 | break; | |
3045 | } | |
3046 | if (userns->may_mount_sysfs && userns->may_mount_proc) | |
3047 | break; | |
3048 | } | |
3049 | up_read(&namespace_sem); | |
3050 | } | |
3051 | ||
8823c079 EB |
3052 | static void *mntns_get(struct task_struct *task) |
3053 | { | |
3054 | struct mnt_namespace *ns = NULL; | |
3055 | struct nsproxy *nsproxy; | |
3056 | ||
3057 | rcu_read_lock(); | |
3058 | nsproxy = task_nsproxy(task); | |
3059 | if (nsproxy) { | |
3060 | ns = nsproxy->mnt_ns; | |
3061 | get_mnt_ns(ns); | |
3062 | } | |
3063 | rcu_read_unlock(); | |
3064 | ||
3065 | return ns; | |
3066 | } | |
3067 | ||
3068 | static void mntns_put(void *ns) | |
3069 | { | |
3070 | put_mnt_ns(ns); | |
3071 | } | |
3072 | ||
3073 | static int mntns_install(struct nsproxy *nsproxy, void *ns) | |
3074 | { | |
3075 | struct fs_struct *fs = current->fs; | |
3076 | struct mnt_namespace *mnt_ns = ns; | |
3077 | struct path root; | |
3078 | ||
0c55cfc4 | 3079 | if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) || |
5e4a0847 EB |
3080 | !nsown_capable(CAP_SYS_CHROOT) || |
3081 | !nsown_capable(CAP_SYS_ADMIN)) | |
ae11e0f1 | 3082 | return -EPERM; |
8823c079 EB |
3083 | |
3084 | if (fs->users != 1) | |
3085 | return -EINVAL; | |
3086 | ||
3087 | get_mnt_ns(mnt_ns); | |
3088 | put_mnt_ns(nsproxy->mnt_ns); | |
3089 | nsproxy->mnt_ns = mnt_ns; | |
3090 | ||
3091 | /* Find the root */ | |
3092 | root.mnt = &mnt_ns->root->mnt; | |
3093 | root.dentry = mnt_ns->root->mnt.mnt_root; | |
3094 | path_get(&root); | |
3095 | while(d_mountpoint(root.dentry) && follow_down_one(&root)) | |
3096 | ; | |
3097 | ||
3098 | /* Update the pwd and root */ | |
3099 | set_fs_pwd(fs, &root); | |
3100 | set_fs_root(fs, &root); | |
3101 | ||
3102 | path_put(&root); | |
3103 | return 0; | |
3104 | } | |
3105 | ||
98f842e6 EB |
3106 | static unsigned int mntns_inum(void *ns) |
3107 | { | |
3108 | struct mnt_namespace *mnt_ns = ns; | |
3109 | return mnt_ns->proc_inum; | |
3110 | } | |
3111 | ||
8823c079 EB |
3112 | const struct proc_ns_operations mntns_operations = { |
3113 | .name = "mnt", | |
3114 | .type = CLONE_NEWNS, | |
3115 | .get = mntns_get, | |
3116 | .put = mntns_put, | |
3117 | .install = mntns_install, | |
98f842e6 | 3118 | .inum = mntns_inum, |
8823c079 | 3119 | }; |