]>
Commit | Line | Data |
---|---|---|
33d3dfff | 1 | #include <linux/fanotify.h> |
11637e4b | 2 | #include <linux/fcntl.h> |
2a3edf86 | 3 | #include <linux/file.h> |
11637e4b | 4 | #include <linux/fs.h> |
52c923dd | 5 | #include <linux/anon_inodes.h> |
11637e4b | 6 | #include <linux/fsnotify_backend.h> |
2a3edf86 | 7 | #include <linux/init.h> |
a1014f10 | 8 | #include <linux/mount.h> |
2a3edf86 | 9 | #include <linux/namei.h> |
a1014f10 | 10 | #include <linux/poll.h> |
11637e4b EP |
11 | #include <linux/security.h> |
12 | #include <linux/syscalls.h> | |
e4e047a2 | 13 | #include <linux/slab.h> |
2a3edf86 | 14 | #include <linux/types.h> |
a1014f10 | 15 | #include <linux/uaccess.h> |
91c2e0bc | 16 | #include <linux/compat.h> |
a1014f10 EP |
17 | |
18 | #include <asm/ioctls.h> | |
11637e4b | 19 | |
c63181e6 | 20 | #include "../../mount.h" |
be77196b | 21 | #include "../fdinfo.h" |
7053aee2 | 22 | #include "fanotify.h" |
c63181e6 | 23 | |
2529a0df | 24 | #define FANOTIFY_DEFAULT_MAX_EVENTS 16384 |
e7099d8a | 25 | #define FANOTIFY_DEFAULT_MAX_MARKS 8192 |
4afeff85 | 26 | #define FANOTIFY_DEFAULT_MAX_LISTENERS 128 |
2529a0df | 27 | |
33d3dfff | 28 | extern const struct fsnotify_ops fanotify_fsnotify_ops; |
11637e4b | 29 | |
2a3edf86 | 30 | static struct kmem_cache *fanotify_mark_cache __read_mostly; |
7053aee2 | 31 | struct kmem_cache *fanotify_event_cachep __read_mostly; |
f083441b | 32 | struct kmem_cache *fanotify_perm_event_cachep __read_mostly; |
2a3edf86 | 33 | |
a1014f10 EP |
34 | /* |
35 | * Get an fsnotify notification event if one exists and is small | |
36 | * enough to fit in "count". Return an error pointer if the count | |
37 | * is not large enough. | |
38 | * | |
39 | * Called with the group->notification_mutex held. | |
40 | */ | |
41 | static struct fsnotify_event *get_one_event(struct fsnotify_group *group, | |
42 | size_t count) | |
43 | { | |
44 | BUG_ON(!mutex_is_locked(&group->notification_mutex)); | |
45 | ||
46 | pr_debug("%s: group=%p count=%zd\n", __func__, group, count); | |
47 | ||
48 | if (fsnotify_notify_queue_is_empty(group)) | |
49 | return NULL; | |
50 | ||
51 | if (FAN_EVENT_METADATA_LEN > count) | |
52 | return ERR_PTR(-EINVAL); | |
53 | ||
54 | /* held the notification_mutex the whole time, so this is the | |
55 | * same event we peeked above */ | |
56 | return fsnotify_remove_notify_event(group); | |
57 | } | |
58 | ||
352e3b24 | 59 | static int create_fd(struct fsnotify_group *group, |
7053aee2 JK |
60 | struct fanotify_event_info *event, |
61 | struct file **file) | |
a1014f10 EP |
62 | { |
63 | int client_fd; | |
a1014f10 EP |
64 | struct file *new_file; |
65 | ||
22aa425d | 66 | pr_debug("%s: group=%p event=%p\n", __func__, group, event); |
a1014f10 EP |
67 | |
68 | client_fd = get_unused_fd(); | |
69 | if (client_fd < 0) | |
70 | return client_fd; | |
71 | ||
a1014f10 EP |
72 | /* |
73 | * we need a new file handle for the userspace program so it can read even if it was | |
74 | * originally opened O_WRONLY. | |
75 | */ | |
a1014f10 EP |
76 | /* it's possible this event was an overflow event. in that case dentry and mnt |
77 | * are NULL; That's fine, just don't call dentry open */ | |
765927b2 AV |
78 | if (event->path.dentry && event->path.mnt) |
79 | new_file = dentry_open(&event->path, | |
80af2588 | 80 | group->fanotify_data.f_flags | FMODE_NONOTIFY, |
a1014f10 EP |
81 | current_cred()); |
82 | else | |
83 | new_file = ERR_PTR(-EOVERFLOW); | |
84 | if (IS_ERR(new_file)) { | |
85 | /* | |
86 | * we still send an event even if we can't open the file. this | |
87 | * can happen when say tasks are gone and we try to open their | |
88 | * /proc files or we try to open a WRONLY file like in sysfs | |
89 | * we just send the errno to userspace since there isn't much | |
90 | * else we can do. | |
91 | */ | |
92 | put_unused_fd(client_fd); | |
93 | client_fd = PTR_ERR(new_file); | |
94 | } else { | |
352e3b24 | 95 | *file = new_file; |
a1014f10 EP |
96 | } |
97 | ||
22aa425d | 98 | return client_fd; |
a1014f10 EP |
99 | } |
100 | ||
ecf6f5e7 | 101 | static int fill_event_metadata(struct fsnotify_group *group, |
7053aee2 JK |
102 | struct fanotify_event_metadata *metadata, |
103 | struct fsnotify_event *fsn_event, | |
104 | struct file **file) | |
a1014f10 | 105 | { |
fdbf3cee | 106 | int ret = 0; |
7053aee2 | 107 | struct fanotify_event_info *event; |
fdbf3cee | 108 | |
a1014f10 | 109 | pr_debug("%s: group=%p metadata=%p event=%p\n", __func__, |
7053aee2 | 110 | group, metadata, fsn_event); |
a1014f10 | 111 | |
352e3b24 | 112 | *file = NULL; |
7053aee2 | 113 | event = container_of(fsn_event, struct fanotify_event_info, fse); |
a1014f10 | 114 | metadata->event_len = FAN_EVENT_METADATA_LEN; |
7d131623 | 115 | metadata->metadata_len = FAN_EVENT_METADATA_LEN; |
a1014f10 | 116 | metadata->vers = FANOTIFY_METADATA_VERSION; |
de1e0c40 | 117 | metadata->reserved = 0; |
7053aee2 | 118 | metadata->mask = fsn_event->mask & FAN_ALL_OUTGOING_EVENTS; |
32c32632 | 119 | metadata->pid = pid_vnr(event->tgid); |
7053aee2 | 120 | if (unlikely(fsn_event->mask & FAN_Q_OVERFLOW)) |
fdbf3cee LS |
121 | metadata->fd = FAN_NOFD; |
122 | else { | |
352e3b24 | 123 | metadata->fd = create_fd(group, event, file); |
fdbf3cee LS |
124 | if (metadata->fd < 0) |
125 | ret = metadata->fd; | |
126 | } | |
a1014f10 | 127 | |
fdbf3cee | 128 | return ret; |
a1014f10 EP |
129 | } |
130 | ||
b2d87909 | 131 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
f083441b JK |
132 | static struct fanotify_perm_event_info *dequeue_event( |
133 | struct fsnotify_group *group, int fd) | |
b2d87909 | 134 | { |
f083441b | 135 | struct fanotify_perm_event_info *event, *return_e = NULL; |
b2d87909 | 136 | |
9573f793 | 137 | spin_lock(&group->fanotify_data.access_lock); |
f083441b JK |
138 | list_for_each_entry(event, &group->fanotify_data.access_list, |
139 | fae.fse.list) { | |
140 | if (event->fd != fd) | |
b2d87909 EP |
141 | continue; |
142 | ||
f083441b JK |
143 | list_del_init(&event->fae.fse.list); |
144 | return_e = event; | |
b2d87909 EP |
145 | break; |
146 | } | |
9573f793 | 147 | spin_unlock(&group->fanotify_data.access_lock); |
b2d87909 | 148 | |
f083441b | 149 | pr_debug("%s: found return_re=%p\n", __func__, return_e); |
b2d87909 | 150 | |
f083441b | 151 | return return_e; |
b2d87909 EP |
152 | } |
153 | ||
154 | static int process_access_response(struct fsnotify_group *group, | |
155 | struct fanotify_response *response_struct) | |
156 | { | |
f083441b JK |
157 | struct fanotify_perm_event_info *event; |
158 | int fd = response_struct->fd; | |
159 | int response = response_struct->response; | |
b2d87909 EP |
160 | |
161 | pr_debug("%s: group=%p fd=%d response=%d\n", __func__, group, | |
162 | fd, response); | |
163 | /* | |
164 | * make sure the response is valid, if invalid we do nothing and either | |
25985edc | 165 | * userspace can send a valid response or we will clean it up after the |
b2d87909 EP |
166 | * timeout |
167 | */ | |
168 | switch (response) { | |
169 | case FAN_ALLOW: | |
170 | case FAN_DENY: | |
171 | break; | |
172 | default: | |
173 | return -EINVAL; | |
174 | } | |
175 | ||
176 | if (fd < 0) | |
177 | return -EINVAL; | |
178 | ||
f083441b JK |
179 | event = dequeue_event(group, fd); |
180 | if (!event) | |
b2d87909 EP |
181 | return -ENOENT; |
182 | ||
f083441b | 183 | event->response = response; |
b2d87909 EP |
184 | wake_up(&group->fanotify_data.access_waitq); |
185 | ||
b2d87909 EP |
186 | return 0; |
187 | } | |
b2d87909 EP |
188 | #endif |
189 | ||
a1014f10 EP |
190 | static ssize_t copy_event_to_user(struct fsnotify_group *group, |
191 | struct fsnotify_event *event, | |
192 | char __user *buf) | |
193 | { | |
194 | struct fanotify_event_metadata fanotify_event_metadata; | |
352e3b24 | 195 | struct file *f; |
b2d87909 | 196 | int fd, ret; |
a1014f10 EP |
197 | |
198 | pr_debug("%s: group=%p event=%p\n", __func__, group, event); | |
199 | ||
352e3b24 | 200 | ret = fill_event_metadata(group, &fanotify_event_metadata, event, &f); |
ecf6f5e7 | 201 | if (ret < 0) |
d507816b | 202 | return ret; |
b2d87909 | 203 | |
fdbf3cee | 204 | fd = fanotify_event_metadata.fd; |
b2d87909 | 205 | ret = -EFAULT; |
7d131623 EP |
206 | if (copy_to_user(buf, &fanotify_event_metadata, |
207 | fanotify_event_metadata.event_len)) | |
352e3b24 AV |
208 | goto out_close_fd; |
209 | ||
f083441b | 210 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
d507816b JK |
211 | if (event->mask & FAN_ALL_PERM_EVENTS) |
212 | FANOTIFY_PE(event)->fd = fd; | |
f083441b | 213 | #endif |
a1014f10 | 214 | |
3587b1b0 AV |
215 | if (fd != FAN_NOFD) |
216 | fd_install(fd, f); | |
7d131623 | 217 | return fanotify_event_metadata.event_len; |
b2d87909 | 218 | |
b2d87909 | 219 | out_close_fd: |
352e3b24 AV |
220 | if (fd != FAN_NOFD) { |
221 | put_unused_fd(fd); | |
222 | fput(f); | |
223 | } | |
b2d87909 | 224 | return ret; |
a1014f10 EP |
225 | } |
226 | ||
227 | /* intofiy userspace file descriptor functions */ | |
228 | static unsigned int fanotify_poll(struct file *file, poll_table *wait) | |
229 | { | |
230 | struct fsnotify_group *group = file->private_data; | |
231 | int ret = 0; | |
232 | ||
233 | poll_wait(file, &group->notification_waitq, wait); | |
234 | mutex_lock(&group->notification_mutex); | |
235 | if (!fsnotify_notify_queue_is_empty(group)) | |
236 | ret = POLLIN | POLLRDNORM; | |
237 | mutex_unlock(&group->notification_mutex); | |
238 | ||
239 | return ret; | |
240 | } | |
241 | ||
242 | static ssize_t fanotify_read(struct file *file, char __user *buf, | |
243 | size_t count, loff_t *pos) | |
244 | { | |
245 | struct fsnotify_group *group; | |
246 | struct fsnotify_event *kevent; | |
247 | char __user *start; | |
248 | int ret; | |
249 | DEFINE_WAIT(wait); | |
250 | ||
251 | start = buf; | |
252 | group = file->private_data; | |
253 | ||
254 | pr_debug("%s: group=%p\n", __func__, group); | |
255 | ||
256 | while (1) { | |
257 | prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE); | |
258 | ||
259 | mutex_lock(&group->notification_mutex); | |
260 | kevent = get_one_event(group, count); | |
261 | mutex_unlock(&group->notification_mutex); | |
262 | ||
d8aaab4f | 263 | if (IS_ERR(kevent)) { |
a1014f10 | 264 | ret = PTR_ERR(kevent); |
d8aaab4f JK |
265 | break; |
266 | } | |
267 | ||
268 | if (!kevent) { | |
269 | ret = -EAGAIN; | |
270 | if (file->f_flags & O_NONBLOCK) | |
a1014f10 | 271 | break; |
d8aaab4f JK |
272 | |
273 | ret = -ERESTARTSYS; | |
274 | if (signal_pending(current)) | |
275 | break; | |
276 | ||
277 | if (start != buf) | |
a1014f10 | 278 | break; |
d8aaab4f | 279 | schedule(); |
a1014f10 EP |
280 | continue; |
281 | } | |
282 | ||
d8aaab4f JK |
283 | ret = copy_event_to_user(group, kevent, buf); |
284 | /* | |
285 | * Permission events get queued to wait for response. Other | |
286 | * events can be destroyed now. | |
287 | */ | |
d507816b | 288 | if (!(kevent->mask & FAN_ALL_PERM_EVENTS)) { |
d8aaab4f | 289 | fsnotify_destroy_event(group, kevent); |
d507816b JK |
290 | if (ret < 0) |
291 | break; | |
292 | } else { | |
293 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS | |
294 | if (ret < 0) { | |
295 | FANOTIFY_PE(kevent)->response = FAN_DENY; | |
296 | wake_up(&group->fanotify_data.access_waitq); | |
297 | break; | |
298 | } | |
299 | spin_lock(&group->fanotify_data.access_lock); | |
300 | list_add_tail(&kevent->list, | |
301 | &group->fanotify_data.access_list); | |
302 | spin_unlock(&group->fanotify_data.access_lock); | |
303 | #endif | |
304 | } | |
d8aaab4f JK |
305 | buf += ret; |
306 | count -= ret; | |
a1014f10 EP |
307 | } |
308 | ||
309 | finish_wait(&group->notification_waitq, &wait); | |
310 | if (start != buf && ret != -EFAULT) | |
311 | ret = buf - start; | |
312 | return ret; | |
313 | } | |
314 | ||
b2d87909 EP |
315 | static ssize_t fanotify_write(struct file *file, const char __user *buf, size_t count, loff_t *pos) |
316 | { | |
317 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS | |
318 | struct fanotify_response response = { .fd = -1, .response = -1 }; | |
319 | struct fsnotify_group *group; | |
320 | int ret; | |
321 | ||
322 | group = file->private_data; | |
323 | ||
324 | if (count > sizeof(response)) | |
325 | count = sizeof(response); | |
326 | ||
327 | pr_debug("%s: group=%p count=%zu\n", __func__, group, count); | |
328 | ||
329 | if (copy_from_user(&response, buf, count)) | |
330 | return -EFAULT; | |
331 | ||
332 | ret = process_access_response(group, &response); | |
333 | if (ret < 0) | |
334 | count = ret; | |
335 | ||
336 | return count; | |
337 | #else | |
338 | return -EINVAL; | |
339 | #endif | |
340 | } | |
341 | ||
52c923dd EP |
342 | static int fanotify_release(struct inode *ignored, struct file *file) |
343 | { | |
344 | struct fsnotify_group *group = file->private_data; | |
52c923dd | 345 | |
2eebf582 | 346 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
f083441b | 347 | struct fanotify_perm_event_info *event, *next; |
19ba54f4 | 348 | |
9573f793 | 349 | spin_lock(&group->fanotify_data.access_lock); |
2eebf582 | 350 | |
09e5f14e | 351 | atomic_inc(&group->fanotify_data.bypass_perm); |
2eebf582 | 352 | |
f083441b JK |
353 | list_for_each_entry_safe(event, next, &group->fanotify_data.access_list, |
354 | fae.fse.list) { | |
355 | pr_debug("%s: found group=%p event=%p\n", __func__, group, | |
356 | event); | |
2eebf582 | 357 | |
f083441b JK |
358 | list_del_init(&event->fae.fse.list); |
359 | event->response = FAN_ALLOW; | |
2eebf582 | 360 | } |
9573f793 | 361 | spin_unlock(&group->fanotify_data.access_lock); |
2eebf582 EP |
362 | |
363 | wake_up(&group->fanotify_data.access_waitq); | |
364 | #endif | |
0a6b6bd5 | 365 | |
52c923dd | 366 | /* matches the fanotify_init->fsnotify_alloc_group */ |
d8153d4d | 367 | fsnotify_destroy_group(group); |
52c923dd EP |
368 | |
369 | return 0; | |
370 | } | |
371 | ||
a1014f10 EP |
372 | static long fanotify_ioctl(struct file *file, unsigned int cmd, unsigned long arg) |
373 | { | |
374 | struct fsnotify_group *group; | |
7053aee2 | 375 | struct fsnotify_event *fsn_event; |
a1014f10 EP |
376 | void __user *p; |
377 | int ret = -ENOTTY; | |
378 | size_t send_len = 0; | |
379 | ||
380 | group = file->private_data; | |
381 | ||
382 | p = (void __user *) arg; | |
383 | ||
384 | switch (cmd) { | |
385 | case FIONREAD: | |
386 | mutex_lock(&group->notification_mutex); | |
7053aee2 | 387 | list_for_each_entry(fsn_event, &group->notification_list, list) |
a1014f10 EP |
388 | send_len += FAN_EVENT_METADATA_LEN; |
389 | mutex_unlock(&group->notification_mutex); | |
390 | ret = put_user(send_len, (int __user *) p); | |
391 | break; | |
392 | } | |
393 | ||
394 | return ret; | |
395 | } | |
396 | ||
52c923dd | 397 | static const struct file_operations fanotify_fops = { |
be77196b | 398 | .show_fdinfo = fanotify_show_fdinfo, |
a1014f10 EP |
399 | .poll = fanotify_poll, |
400 | .read = fanotify_read, | |
b2d87909 | 401 | .write = fanotify_write, |
52c923dd EP |
402 | .fasync = NULL, |
403 | .release = fanotify_release, | |
a1014f10 EP |
404 | .unlocked_ioctl = fanotify_ioctl, |
405 | .compat_ioctl = fanotify_ioctl, | |
6038f373 | 406 | .llseek = noop_llseek, |
52c923dd EP |
407 | }; |
408 | ||
2a3edf86 EP |
409 | static void fanotify_free_mark(struct fsnotify_mark *fsn_mark) |
410 | { | |
411 | kmem_cache_free(fanotify_mark_cache, fsn_mark); | |
412 | } | |
413 | ||
414 | static int fanotify_find_path(int dfd, const char __user *filename, | |
415 | struct path *path, unsigned int flags) | |
416 | { | |
417 | int ret; | |
418 | ||
419 | pr_debug("%s: dfd=%d filename=%p flags=%x\n", __func__, | |
420 | dfd, filename, flags); | |
421 | ||
422 | if (filename == NULL) { | |
2903ff01 | 423 | struct fd f = fdget(dfd); |
2a3edf86 EP |
424 | |
425 | ret = -EBADF; | |
2903ff01 | 426 | if (!f.file) |
2a3edf86 EP |
427 | goto out; |
428 | ||
429 | ret = -ENOTDIR; | |
430 | if ((flags & FAN_MARK_ONLYDIR) && | |
496ad9aa | 431 | !(S_ISDIR(file_inode(f.file)->i_mode))) { |
2903ff01 | 432 | fdput(f); |
2a3edf86 EP |
433 | goto out; |
434 | } | |
435 | ||
2903ff01 | 436 | *path = f.file->f_path; |
2a3edf86 | 437 | path_get(path); |
2903ff01 | 438 | fdput(f); |
2a3edf86 EP |
439 | } else { |
440 | unsigned int lookup_flags = 0; | |
441 | ||
442 | if (!(flags & FAN_MARK_DONT_FOLLOW)) | |
443 | lookup_flags |= LOOKUP_FOLLOW; | |
444 | if (flags & FAN_MARK_ONLYDIR) | |
445 | lookup_flags |= LOOKUP_DIRECTORY; | |
446 | ||
447 | ret = user_path_at(dfd, filename, lookup_flags, path); | |
448 | if (ret) | |
449 | goto out; | |
450 | } | |
451 | ||
452 | /* you can only watch an inode if you have read permissions on it */ | |
453 | ret = inode_permission(path->dentry->d_inode, MAY_READ); | |
454 | if (ret) | |
455 | path_put(path); | |
456 | out: | |
457 | return ret; | |
458 | } | |
459 | ||
b9e4e3bd EP |
460 | static __u32 fanotify_mark_remove_from_mask(struct fsnotify_mark *fsn_mark, |
461 | __u32 mask, | |
6dfbd149 LS |
462 | unsigned int flags, |
463 | int *destroy) | |
088b09b0 AG |
464 | { |
465 | __u32 oldmask; | |
466 | ||
467 | spin_lock(&fsn_mark->lock); | |
b9e4e3bd EP |
468 | if (!(flags & FAN_MARK_IGNORED_MASK)) { |
469 | oldmask = fsn_mark->mask; | |
470 | fsnotify_set_mark_mask_locked(fsn_mark, (oldmask & ~mask)); | |
471 | } else { | |
472 | oldmask = fsn_mark->ignored_mask; | |
473 | fsnotify_set_mark_ignored_mask_locked(fsn_mark, (oldmask & ~mask)); | |
474 | } | |
088b09b0 AG |
475 | spin_unlock(&fsn_mark->lock); |
476 | ||
6dfbd149 | 477 | *destroy = !(oldmask & ~mask); |
088b09b0 AG |
478 | |
479 | return mask & oldmask; | |
480 | } | |
481 | ||
f3640192 | 482 | static int fanotify_remove_vfsmount_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
483 | struct vfsmount *mnt, __u32 mask, |
484 | unsigned int flags) | |
88826276 EP |
485 | { |
486 | struct fsnotify_mark *fsn_mark = NULL; | |
088b09b0 | 487 | __u32 removed; |
6dfbd149 | 488 | int destroy_mark; |
88826276 | 489 | |
7b18527c | 490 | mutex_lock(&group->mark_mutex); |
f3640192 | 491 | fsn_mark = fsnotify_find_vfsmount_mark(group, mnt); |
7b18527c LS |
492 | if (!fsn_mark) { |
493 | mutex_unlock(&group->mark_mutex); | |
f3640192 | 494 | return -ENOENT; |
7b18527c | 495 | } |
88826276 | 496 | |
6dfbd149 LS |
497 | removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags, |
498 | &destroy_mark); | |
499 | if (destroy_mark) | |
7b18527c LS |
500 | fsnotify_destroy_mark_locked(fsn_mark, group); |
501 | mutex_unlock(&group->mark_mutex); | |
6dfbd149 | 502 | |
f3640192 | 503 | fsnotify_put_mark(fsn_mark); |
c63181e6 | 504 | if (removed & real_mount(mnt)->mnt_fsnotify_mask) |
f3640192 AG |
505 | fsnotify_recalc_vfsmount_mask(mnt); |
506 | ||
507 | return 0; | |
508 | } | |
2a3edf86 | 509 | |
f3640192 | 510 | static int fanotify_remove_inode_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
511 | struct inode *inode, __u32 mask, |
512 | unsigned int flags) | |
f3640192 AG |
513 | { |
514 | struct fsnotify_mark *fsn_mark = NULL; | |
515 | __u32 removed; | |
6dfbd149 | 516 | int destroy_mark; |
f3640192 | 517 | |
7b18527c | 518 | mutex_lock(&group->mark_mutex); |
f3640192 | 519 | fsn_mark = fsnotify_find_inode_mark(group, inode); |
7b18527c LS |
520 | if (!fsn_mark) { |
521 | mutex_unlock(&group->mark_mutex); | |
88826276 | 522 | return -ENOENT; |
7b18527c | 523 | } |
88826276 | 524 | |
6dfbd149 LS |
525 | removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags, |
526 | &destroy_mark); | |
527 | if (destroy_mark) | |
7b18527c LS |
528 | fsnotify_destroy_mark_locked(fsn_mark, group); |
529 | mutex_unlock(&group->mark_mutex); | |
530 | ||
5444e298 | 531 | /* matches the fsnotify_find_inode_mark() */ |
2a3edf86 | 532 | fsnotify_put_mark(fsn_mark); |
f3640192 AG |
533 | if (removed & inode->i_fsnotify_mask) |
534 | fsnotify_recalc_inode_mask(inode); | |
088b09b0 | 535 | |
2a3edf86 EP |
536 | return 0; |
537 | } | |
538 | ||
b9e4e3bd EP |
539 | static __u32 fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark, |
540 | __u32 mask, | |
541 | unsigned int flags) | |
912ee394 | 542 | { |
192ca4d1 | 543 | __u32 oldmask = -1; |
912ee394 AG |
544 | |
545 | spin_lock(&fsn_mark->lock); | |
b9e4e3bd EP |
546 | if (!(flags & FAN_MARK_IGNORED_MASK)) { |
547 | oldmask = fsn_mark->mask; | |
548 | fsnotify_set_mark_mask_locked(fsn_mark, (oldmask | mask)); | |
549 | } else { | |
192ca4d1 EP |
550 | __u32 tmask = fsn_mark->ignored_mask | mask; |
551 | fsnotify_set_mark_ignored_mask_locked(fsn_mark, tmask); | |
c9778a98 EP |
552 | if (flags & FAN_MARK_IGNORED_SURV_MODIFY) |
553 | fsn_mark->flags |= FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY; | |
b9e4e3bd | 554 | } |
8fcd6528 EP |
555 | |
556 | if (!(flags & FAN_MARK_ONDIR)) { | |
557 | __u32 tmask = fsn_mark->ignored_mask | FAN_ONDIR; | |
558 | fsnotify_set_mark_ignored_mask_locked(fsn_mark, tmask); | |
559 | } | |
560 | ||
912ee394 AG |
561 | spin_unlock(&fsn_mark->lock); |
562 | ||
563 | return mask & ~oldmask; | |
564 | } | |
565 | ||
5e9c070c LS |
566 | static struct fsnotify_mark *fanotify_add_new_mark(struct fsnotify_group *group, |
567 | struct inode *inode, | |
568 | struct vfsmount *mnt) | |
569 | { | |
570 | struct fsnotify_mark *mark; | |
571 | int ret; | |
572 | ||
573 | if (atomic_read(&group->num_marks) > group->fanotify_data.max_marks) | |
574 | return ERR_PTR(-ENOSPC); | |
575 | ||
576 | mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL); | |
577 | if (!mark) | |
578 | return ERR_PTR(-ENOMEM); | |
579 | ||
580 | fsnotify_init_mark(mark, fanotify_free_mark); | |
581 | ret = fsnotify_add_mark_locked(mark, group, inode, mnt, 0); | |
582 | if (ret) { | |
583 | fsnotify_put_mark(mark); | |
584 | return ERR_PTR(ret); | |
585 | } | |
586 | ||
587 | return mark; | |
588 | } | |
589 | ||
590 | ||
52202dfb | 591 | static int fanotify_add_vfsmount_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
592 | struct vfsmount *mnt, __u32 mask, |
593 | unsigned int flags) | |
2a3edf86 EP |
594 | { |
595 | struct fsnotify_mark *fsn_mark; | |
912ee394 | 596 | __u32 added; |
2a3edf86 | 597 | |
7b18527c | 598 | mutex_lock(&group->mark_mutex); |
88826276 EP |
599 | fsn_mark = fsnotify_find_vfsmount_mark(group, mnt); |
600 | if (!fsn_mark) { | |
5e9c070c LS |
601 | fsn_mark = fanotify_add_new_mark(group, NULL, mnt); |
602 | if (IS_ERR(fsn_mark)) { | |
7b18527c | 603 | mutex_unlock(&group->mark_mutex); |
5e9c070c | 604 | return PTR_ERR(fsn_mark); |
7b18527c | 605 | } |
88826276 | 606 | } |
b9e4e3bd | 607 | added = fanotify_mark_add_to_mask(fsn_mark, mask, flags); |
7b18527c | 608 | mutex_unlock(&group->mark_mutex); |
fa218ab9 | 609 | |
c63181e6 | 610 | if (added & ~real_mount(mnt)->mnt_fsnotify_mask) |
43709a28 | 611 | fsnotify_recalc_vfsmount_mask(mnt); |
5e9c070c | 612 | |
fa218ab9 | 613 | fsnotify_put_mark(fsn_mark); |
5e9c070c | 614 | return 0; |
88826276 EP |
615 | } |
616 | ||
52202dfb | 617 | static int fanotify_add_inode_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
618 | struct inode *inode, __u32 mask, |
619 | unsigned int flags) | |
88826276 EP |
620 | { |
621 | struct fsnotify_mark *fsn_mark; | |
912ee394 | 622 | __u32 added; |
88826276 EP |
623 | |
624 | pr_debug("%s: group=%p inode=%p\n", __func__, group, inode); | |
2a3edf86 | 625 | |
5322a59f EP |
626 | /* |
627 | * If some other task has this inode open for write we should not add | |
628 | * an ignored mark, unless that ignored mark is supposed to survive | |
629 | * modification changes anyway. | |
630 | */ | |
631 | if ((flags & FAN_MARK_IGNORED_MASK) && | |
632 | !(flags & FAN_MARK_IGNORED_SURV_MODIFY) && | |
633 | (atomic_read(&inode->i_writecount) > 0)) | |
634 | return 0; | |
635 | ||
7b18527c | 636 | mutex_lock(&group->mark_mutex); |
5444e298 | 637 | fsn_mark = fsnotify_find_inode_mark(group, inode); |
2a3edf86 | 638 | if (!fsn_mark) { |
5e9c070c LS |
639 | fsn_mark = fanotify_add_new_mark(group, inode, NULL); |
640 | if (IS_ERR(fsn_mark)) { | |
7b18527c | 641 | mutex_unlock(&group->mark_mutex); |
5e9c070c | 642 | return PTR_ERR(fsn_mark); |
7b18527c | 643 | } |
2a3edf86 | 644 | } |
b9e4e3bd | 645 | added = fanotify_mark_add_to_mask(fsn_mark, mask, flags); |
7b18527c | 646 | mutex_unlock(&group->mark_mutex); |
fa218ab9 | 647 | |
43709a28 EP |
648 | if (added & ~inode->i_fsnotify_mask) |
649 | fsnotify_recalc_inode_mask(inode); | |
5e9c070c | 650 | |
fa218ab9 | 651 | fsnotify_put_mark(fsn_mark); |
5e9c070c | 652 | return 0; |
88826276 | 653 | } |
2a3edf86 | 654 | |
52c923dd | 655 | /* fanotify syscalls */ |
08ae8938 | 656 | SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags) |
11637e4b | 657 | { |
52c923dd EP |
658 | struct fsnotify_group *group; |
659 | int f_flags, fd; | |
4afeff85 | 660 | struct user_struct *user; |
ff57cd58 | 661 | struct fanotify_event_info *oevent; |
52c923dd | 662 | |
08ae8938 EP |
663 | pr_debug("%s: flags=%d event_f_flags=%d\n", |
664 | __func__, flags, event_f_flags); | |
52c923dd | 665 | |
52c923dd | 666 | if (!capable(CAP_SYS_ADMIN)) |
a2f13ad0 | 667 | return -EPERM; |
52c923dd EP |
668 | |
669 | if (flags & ~FAN_ALL_INIT_FLAGS) | |
670 | return -EINVAL; | |
671 | ||
4afeff85 EP |
672 | user = get_current_user(); |
673 | if (atomic_read(&user->fanotify_listeners) > FANOTIFY_DEFAULT_MAX_LISTENERS) { | |
674 | free_uid(user); | |
675 | return -EMFILE; | |
676 | } | |
677 | ||
b2d87909 | 678 | f_flags = O_RDWR | FMODE_NONOTIFY; |
52c923dd EP |
679 | if (flags & FAN_CLOEXEC) |
680 | f_flags |= O_CLOEXEC; | |
681 | if (flags & FAN_NONBLOCK) | |
682 | f_flags |= O_NONBLOCK; | |
683 | ||
684 | /* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */ | |
685 | group = fsnotify_alloc_group(&fanotify_fsnotify_ops); | |
26379198 EP |
686 | if (IS_ERR(group)) { |
687 | free_uid(user); | |
52c923dd | 688 | return PTR_ERR(group); |
26379198 | 689 | } |
52c923dd | 690 | |
4afeff85 EP |
691 | group->fanotify_data.user = user; |
692 | atomic_inc(&user->fanotify_listeners); | |
693 | ||
f083441b | 694 | oevent = fanotify_alloc_event(NULL, FS_Q_OVERFLOW, NULL); |
ff57cd58 JK |
695 | if (unlikely(!oevent)) { |
696 | fd = -ENOMEM; | |
697 | goto out_destroy_group; | |
698 | } | |
699 | group->overflow_event = &oevent->fse; | |
ff57cd58 | 700 | |
1e2ee49f WW |
701 | if (force_o_largefile()) |
702 | event_f_flags |= O_LARGEFILE; | |
80af2588 | 703 | group->fanotify_data.f_flags = event_f_flags; |
9e66e423 | 704 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
9573f793 | 705 | spin_lock_init(&group->fanotify_data.access_lock); |
9e66e423 EP |
706 | init_waitqueue_head(&group->fanotify_data.access_waitq); |
707 | INIT_LIST_HEAD(&group->fanotify_data.access_list); | |
09e5f14e | 708 | atomic_set(&group->fanotify_data.bypass_perm, 0); |
9e66e423 | 709 | #endif |
4231a235 EP |
710 | switch (flags & FAN_ALL_CLASS_BITS) { |
711 | case FAN_CLASS_NOTIF: | |
712 | group->priority = FS_PRIO_0; | |
713 | break; | |
714 | case FAN_CLASS_CONTENT: | |
715 | group->priority = FS_PRIO_1; | |
716 | break; | |
717 | case FAN_CLASS_PRE_CONTENT: | |
718 | group->priority = FS_PRIO_2; | |
719 | break; | |
720 | default: | |
721 | fd = -EINVAL; | |
d8153d4d | 722 | goto out_destroy_group; |
4231a235 | 723 | } |
cb2d429f | 724 | |
5dd03f55 EP |
725 | if (flags & FAN_UNLIMITED_QUEUE) { |
726 | fd = -EPERM; | |
727 | if (!capable(CAP_SYS_ADMIN)) | |
d8153d4d | 728 | goto out_destroy_group; |
5dd03f55 EP |
729 | group->max_events = UINT_MAX; |
730 | } else { | |
731 | group->max_events = FANOTIFY_DEFAULT_MAX_EVENTS; | |
732 | } | |
2529a0df | 733 | |
ac7e22dc EP |
734 | if (flags & FAN_UNLIMITED_MARKS) { |
735 | fd = -EPERM; | |
736 | if (!capable(CAP_SYS_ADMIN)) | |
d8153d4d | 737 | goto out_destroy_group; |
ac7e22dc EP |
738 | group->fanotify_data.max_marks = UINT_MAX; |
739 | } else { | |
740 | group->fanotify_data.max_marks = FANOTIFY_DEFAULT_MAX_MARKS; | |
741 | } | |
e7099d8a | 742 | |
52c923dd EP |
743 | fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags); |
744 | if (fd < 0) | |
d8153d4d | 745 | goto out_destroy_group; |
52c923dd EP |
746 | |
747 | return fd; | |
748 | ||
d8153d4d LS |
749 | out_destroy_group: |
750 | fsnotify_destroy_group(group); | |
52c923dd | 751 | return fd; |
11637e4b | 752 | } |
bbaa4168 | 753 | |
4a0fd5bf AV |
754 | SYSCALL_DEFINE5(fanotify_mark, int, fanotify_fd, unsigned int, flags, |
755 | __u64, mask, int, dfd, | |
756 | const char __user *, pathname) | |
bbaa4168 | 757 | { |
0ff21db9 EP |
758 | struct inode *inode = NULL; |
759 | struct vfsmount *mnt = NULL; | |
2a3edf86 | 760 | struct fsnotify_group *group; |
2903ff01 | 761 | struct fd f; |
2a3edf86 | 762 | struct path path; |
2903ff01 | 763 | int ret; |
2a3edf86 EP |
764 | |
765 | pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n", | |
766 | __func__, fanotify_fd, flags, dfd, pathname, mask); | |
767 | ||
768 | /* we only use the lower 32 bits as of right now. */ | |
769 | if (mask & ((__u64)0xffffffff << 32)) | |
770 | return -EINVAL; | |
771 | ||
88380fe6 AG |
772 | if (flags & ~FAN_ALL_MARK_FLAGS) |
773 | return -EINVAL; | |
4d92604c | 774 | switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE | FAN_MARK_FLUSH)) { |
1734dee4 | 775 | case FAN_MARK_ADD: /* fallthrough */ |
88380fe6 | 776 | case FAN_MARK_REMOVE: |
1734dee4 LS |
777 | if (!mask) |
778 | return -EINVAL; | |
cc299a98 | 779 | break; |
4d92604c | 780 | case FAN_MARK_FLUSH: |
cc299a98 HS |
781 | if (flags & ~(FAN_MARK_MOUNT | FAN_MARK_FLUSH)) |
782 | return -EINVAL; | |
88380fe6 AG |
783 | break; |
784 | default: | |
785 | return -EINVAL; | |
786 | } | |
8fcd6528 EP |
787 | |
788 | if (mask & FAN_ONDIR) { | |
789 | flags |= FAN_MARK_ONDIR; | |
790 | mask &= ~FAN_ONDIR; | |
791 | } | |
792 | ||
b2d87909 EP |
793 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
794 | if (mask & ~(FAN_ALL_EVENTS | FAN_ALL_PERM_EVENTS | FAN_EVENT_ON_CHILD)) | |
795 | #else | |
88380fe6 | 796 | if (mask & ~(FAN_ALL_EVENTS | FAN_EVENT_ON_CHILD)) |
b2d87909 | 797 | #endif |
2a3edf86 EP |
798 | return -EINVAL; |
799 | ||
2903ff01 AV |
800 | f = fdget(fanotify_fd); |
801 | if (unlikely(!f.file)) | |
2a3edf86 EP |
802 | return -EBADF; |
803 | ||
804 | /* verify that this is indeed an fanotify instance */ | |
805 | ret = -EINVAL; | |
2903ff01 | 806 | if (unlikely(f.file->f_op != &fanotify_fops)) |
2a3edf86 | 807 | goto fput_and_out; |
2903ff01 | 808 | group = f.file->private_data; |
4231a235 EP |
809 | |
810 | /* | |
811 | * group->priority == FS_PRIO_0 == FAN_CLASS_NOTIF. These are not | |
812 | * allowed to set permissions events. | |
813 | */ | |
814 | ret = -EINVAL; | |
815 | if (mask & FAN_ALL_PERM_EVENTS && | |
816 | group->priority == FS_PRIO_0) | |
817 | goto fput_and_out; | |
2a3edf86 | 818 | |
0a8dd2db HS |
819 | if (flags & FAN_MARK_FLUSH) { |
820 | ret = 0; | |
821 | if (flags & FAN_MARK_MOUNT) | |
822 | fsnotify_clear_vfsmount_marks_by_group(group); | |
823 | else | |
824 | fsnotify_clear_inode_marks_by_group(group); | |
825 | goto fput_and_out; | |
826 | } | |
827 | ||
2a3edf86 EP |
828 | ret = fanotify_find_path(dfd, pathname, &path, flags); |
829 | if (ret) | |
830 | goto fput_and_out; | |
831 | ||
832 | /* inode held in place by reference to path; group by fget on fd */ | |
eac8e9e8 | 833 | if (!(flags & FAN_MARK_MOUNT)) |
0ff21db9 EP |
834 | inode = path.dentry->d_inode; |
835 | else | |
836 | mnt = path.mnt; | |
2a3edf86 EP |
837 | |
838 | /* create/update an inode mark */ | |
0a8dd2db | 839 | switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE)) { |
c6223f46 | 840 | case FAN_MARK_ADD: |
eac8e9e8 | 841 | if (flags & FAN_MARK_MOUNT) |
b9e4e3bd | 842 | ret = fanotify_add_vfsmount_mark(group, mnt, mask, flags); |
0ff21db9 | 843 | else |
b9e4e3bd | 844 | ret = fanotify_add_inode_mark(group, inode, mask, flags); |
c6223f46 AG |
845 | break; |
846 | case FAN_MARK_REMOVE: | |
f3640192 | 847 | if (flags & FAN_MARK_MOUNT) |
b9e4e3bd | 848 | ret = fanotify_remove_vfsmount_mark(group, mnt, mask, flags); |
f3640192 | 849 | else |
b9e4e3bd | 850 | ret = fanotify_remove_inode_mark(group, inode, mask, flags); |
c6223f46 AG |
851 | break; |
852 | default: | |
853 | ret = -EINVAL; | |
854 | } | |
2a3edf86 EP |
855 | |
856 | path_put(&path); | |
857 | fput_and_out: | |
2903ff01 | 858 | fdput(f); |
2a3edf86 EP |
859 | return ret; |
860 | } | |
861 | ||
91c2e0bc AV |
862 | #ifdef CONFIG_COMPAT |
863 | COMPAT_SYSCALL_DEFINE6(fanotify_mark, | |
864 | int, fanotify_fd, unsigned int, flags, | |
865 | __u32, mask0, __u32, mask1, int, dfd, | |
866 | const char __user *, pathname) | |
867 | { | |
868 | return sys_fanotify_mark(fanotify_fd, flags, | |
869 | #ifdef __BIG_ENDIAN | |
91c2e0bc | 870 | ((__u64)mask0 << 32) | mask1, |
592f6b84 HC |
871 | #else |
872 | ((__u64)mask1 << 32) | mask0, | |
91c2e0bc AV |
873 | #endif |
874 | dfd, pathname); | |
875 | } | |
876 | #endif | |
877 | ||
2a3edf86 | 878 | /* |
ae0e47f0 | 879 | * fanotify_user_setup - Our initialization function. Note that we cannot return |
2a3edf86 EP |
880 | * error because we have compiled-in VFS hooks. So an (unlikely) failure here |
881 | * must result in panic(). | |
882 | */ | |
883 | static int __init fanotify_user_setup(void) | |
884 | { | |
885 | fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC); | |
7053aee2 | 886 | fanotify_event_cachep = KMEM_CACHE(fanotify_event_info, SLAB_PANIC); |
f083441b JK |
887 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
888 | fanotify_perm_event_cachep = KMEM_CACHE(fanotify_perm_event_info, | |
889 | SLAB_PANIC); | |
890 | #endif | |
2a3edf86 EP |
891 | |
892 | return 0; | |
bbaa4168 | 893 | } |
2a3edf86 | 894 | device_initcall(fanotify_user_setup); |