diff options
Diffstat (limited to 'fs/notify/inotify/inotify_user.c')
-rw-r--r-- | fs/notify/inotify/inotify_user.c | 778 |
1 files changed, 778 insertions, 0 deletions
diff --git a/fs/notify/inotify/inotify_user.c b/fs/notify/inotify/inotify_user.c new file mode 100644 index 000000000000..400f8064a548 --- /dev/null +++ b/fs/notify/inotify/inotify_user.c | |||
@@ -0,0 +1,778 @@ | |||
1 | /* | ||
2 | * fs/inotify_user.c - inotify support for userspace | ||
3 | * | ||
4 | * Authors: | ||
5 | * John McCutchan <ttb@tentacle.dhs.org> | ||
6 | * Robert Love <rml@novell.com> | ||
7 | * | ||
8 | * Copyright (C) 2005 John McCutchan | ||
9 | * Copyright 2006 Hewlett-Packard Development Company, L.P. | ||
10 | * | ||
11 | * This program is free software; you can redistribute it and/or modify it | ||
12 | * under the terms of the GNU General Public License as published by the | ||
13 | * Free Software Foundation; either version 2, or (at your option) any | ||
14 | * later version. | ||
15 | * | ||
16 | * This program is distributed in the hope that it will be useful, but | ||
17 | * WITHOUT ANY WARRANTY; without even the implied warranty of | ||
18 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
19 | * General Public License for more details. | ||
20 | */ | ||
21 | |||
22 | #include <linux/kernel.h> | ||
23 | #include <linux/sched.h> | ||
24 | #include <linux/slab.h> | ||
25 | #include <linux/fs.h> | ||
26 | #include <linux/file.h> | ||
27 | #include <linux/mount.h> | ||
28 | #include <linux/namei.h> | ||
29 | #include <linux/poll.h> | ||
30 | #include <linux/init.h> | ||
31 | #include <linux/list.h> | ||
32 | #include <linux/inotify.h> | ||
33 | #include <linux/syscalls.h> | ||
34 | #include <linux/magic.h> | ||
35 | |||
36 | #include <asm/ioctls.h> | ||
37 | |||
38 | static struct kmem_cache *watch_cachep __read_mostly; | ||
39 | static struct kmem_cache *event_cachep __read_mostly; | ||
40 | |||
41 | static struct vfsmount *inotify_mnt __read_mostly; | ||
42 | |||
43 | /* these are configurable via /proc/sys/fs/inotify/ */ | ||
44 | static int inotify_max_user_instances __read_mostly; | ||
45 | static int inotify_max_user_watches __read_mostly; | ||
46 | static int inotify_max_queued_events __read_mostly; | ||
47 | |||
48 | /* | ||
49 | * Lock ordering: | ||
50 | * | ||
51 | * inotify_dev->up_mutex (ensures we don't re-add the same watch) | ||
52 | * inode->inotify_mutex (protects inode's watch list) | ||
53 | * inotify_handle->mutex (protects inotify_handle's watch list) | ||
54 | * inotify_dev->ev_mutex (protects device's event queue) | ||
55 | */ | ||
56 | |||
57 | /* | ||
58 | * Lifetimes of the main data structures: | ||
59 | * | ||
60 | * inotify_device: Lifetime is managed by reference count, from | ||
61 | * sys_inotify_init() until release. Additional references can bump the count | ||
62 | * via get_inotify_dev() and drop the count via put_inotify_dev(). | ||
63 | * | ||
64 | * inotify_user_watch: Lifetime is from create_watch() to the receipt of an | ||
65 | * IN_IGNORED event from inotify, or when using IN_ONESHOT, to receipt of the | ||
66 | * first event, or to inotify_destroy(). | ||
67 | */ | ||
68 | |||
69 | /* | ||
70 | * struct inotify_device - represents an inotify instance | ||
71 | * | ||
72 | * This structure is protected by the mutex 'mutex'. | ||
73 | */ | ||
74 | struct inotify_device { | ||
75 | wait_queue_head_t wq; /* wait queue for i/o */ | ||
76 | struct mutex ev_mutex; /* protects event queue */ | ||
77 | struct mutex up_mutex; /* synchronizes watch updates */ | ||
78 | struct list_head events; /* list of queued events */ | ||
79 | struct user_struct *user; /* user who opened this dev */ | ||
80 | struct inotify_handle *ih; /* inotify handle */ | ||
81 | struct fasync_struct *fa; /* async notification */ | ||
82 | atomic_t count; /* reference count */ | ||
83 | unsigned int queue_size; /* size of the queue (bytes) */ | ||
84 | unsigned int event_count; /* number of pending events */ | ||
85 | unsigned int max_events; /* maximum number of events */ | ||
86 | }; | ||
87 | |||
88 | /* | ||
89 | * struct inotify_kernel_event - An inotify event, originating from a watch and | ||
90 | * queued for user-space. A list of these is attached to each instance of the | ||
91 | * device. In read(), this list is walked and all events that can fit in the | ||
92 | * buffer are returned. | ||
93 | * | ||
94 | * Protected by dev->ev_mutex of the device in which we are queued. | ||
95 | */ | ||
96 | struct inotify_kernel_event { | ||
97 | struct inotify_event event; /* the user-space event */ | ||
98 | struct list_head list; /* entry in inotify_device's list */ | ||
99 | char *name; /* filename, if any */ | ||
100 | }; | ||
101 | |||
102 | /* | ||
103 | * struct inotify_user_watch - our version of an inotify_watch, we add | ||
104 | * a reference to the associated inotify_device. | ||
105 | */ | ||
106 | struct inotify_user_watch { | ||
107 | struct inotify_device *dev; /* associated device */ | ||
108 | struct inotify_watch wdata; /* inotify watch data */ | ||
109 | }; | ||
110 | |||
111 | #ifdef CONFIG_SYSCTL | ||
112 | |||
113 | #include <linux/sysctl.h> | ||
114 | |||
115 | static int zero; | ||
116 | |||
117 | ctl_table inotify_table[] = { | ||
118 | { | ||
119 | .ctl_name = INOTIFY_MAX_USER_INSTANCES, | ||
120 | .procname = "max_user_instances", | ||
121 | .data = &inotify_max_user_instances, | ||
122 | .maxlen = sizeof(int), | ||
123 | .mode = 0644, | ||
124 | .proc_handler = &proc_dointvec_minmax, | ||
125 | .strategy = &sysctl_intvec, | ||
126 | .extra1 = &zero, | ||
127 | }, | ||
128 | { | ||
129 | .ctl_name = INOTIFY_MAX_USER_WATCHES, | ||
130 | .procname = "max_user_watches", | ||
131 | .data = &inotify_max_user_watches, | ||
132 | .maxlen = sizeof(int), | ||
133 | .mode = 0644, | ||
134 | .proc_handler = &proc_dointvec_minmax, | ||
135 | .strategy = &sysctl_intvec, | ||
136 | .extra1 = &zero, | ||
137 | }, | ||
138 | { | ||
139 | .ctl_name = INOTIFY_MAX_QUEUED_EVENTS, | ||
140 | .procname = "max_queued_events", | ||
141 | .data = &inotify_max_queued_events, | ||
142 | .maxlen = sizeof(int), | ||
143 | .mode = 0644, | ||
144 | .proc_handler = &proc_dointvec_minmax, | ||
145 | .strategy = &sysctl_intvec, | ||
146 | .extra1 = &zero | ||
147 | }, | ||
148 | { .ctl_name = 0 } | ||
149 | }; | ||
150 | #endif /* CONFIG_SYSCTL */ | ||
151 | |||
152 | static inline void get_inotify_dev(struct inotify_device *dev) | ||
153 | { | ||
154 | atomic_inc(&dev->count); | ||
155 | } | ||
156 | |||
157 | static inline void put_inotify_dev(struct inotify_device *dev) | ||
158 | { | ||
159 | if (atomic_dec_and_test(&dev->count)) { | ||
160 | atomic_dec(&dev->user->inotify_devs); | ||
161 | free_uid(dev->user); | ||
162 | kfree(dev); | ||
163 | } | ||
164 | } | ||
165 | |||
166 | /* | ||
167 | * free_inotify_user_watch - cleans up the watch and its references | ||
168 | */ | ||
169 | static void free_inotify_user_watch(struct inotify_watch *w) | ||
170 | { | ||
171 | struct inotify_user_watch *watch; | ||
172 | struct inotify_device *dev; | ||
173 | |||
174 | watch = container_of(w, struct inotify_user_watch, wdata); | ||
175 | dev = watch->dev; | ||
176 | |||
177 | atomic_dec(&dev->user->inotify_watches); | ||
178 | put_inotify_dev(dev); | ||
179 | kmem_cache_free(watch_cachep, watch); | ||
180 | } | ||
181 | |||
182 | /* | ||
183 | * kernel_event - create a new kernel event with the given parameters | ||
184 | * | ||
185 | * This function can sleep. | ||
186 | */ | ||
187 | static struct inotify_kernel_event * kernel_event(s32 wd, u32 mask, u32 cookie, | ||
188 | const char *name) | ||
189 | { | ||
190 | struct inotify_kernel_event *kevent; | ||
191 | |||
192 | kevent = kmem_cache_alloc(event_cachep, GFP_NOFS); | ||
193 | if (unlikely(!kevent)) | ||
194 | return NULL; | ||
195 | |||
196 | /* we hand this out to user-space, so zero it just in case */ | ||
197 | memset(&kevent->event, 0, sizeof(struct inotify_event)); | ||
198 | |||
199 | kevent->event.wd = wd; | ||
200 | kevent->event.mask = mask; | ||
201 | kevent->event.cookie = cookie; | ||
202 | |||
203 | INIT_LIST_HEAD(&kevent->list); | ||
204 | |||
205 | if (name) { | ||
206 | size_t len, rem, event_size = sizeof(struct inotify_event); | ||
207 | |||
208 | /* | ||
209 | * We need to pad the filename so as to properly align an | ||
210 | * array of inotify_event structures. Because the structure is | ||
211 | * small and the common case is a small filename, we just round | ||
212 | * up to the next multiple of the structure's sizeof. This is | ||
213 | * simple and safe for all architectures. | ||
214 | */ | ||
215 | len = strlen(name) + 1; | ||
216 | rem = event_size - len; | ||
217 | if (len > event_size) { | ||
218 | rem = event_size - (len % event_size); | ||
219 | if (len % event_size == 0) | ||
220 | rem = 0; | ||
221 | } | ||
222 | |||
223 | kevent->name = kmalloc(len + rem, GFP_KERNEL); | ||
224 | if (unlikely(!kevent->name)) { | ||
225 | kmem_cache_free(event_cachep, kevent); | ||
226 | return NULL; | ||
227 | } | ||
228 | memcpy(kevent->name, name, len); | ||
229 | if (rem) | ||
230 | memset(kevent->name + len, 0, rem); | ||
231 | kevent->event.len = len + rem; | ||
232 | } else { | ||
233 | kevent->event.len = 0; | ||
234 | kevent->name = NULL; | ||
235 | } | ||
236 | |||
237 | return kevent; | ||
238 | } | ||
239 | |||
240 | /* | ||
241 | * inotify_dev_get_event - return the next event in the given dev's queue | ||
242 | * | ||
243 | * Caller must hold dev->ev_mutex. | ||
244 | */ | ||
245 | static inline struct inotify_kernel_event * | ||
246 | inotify_dev_get_event(struct inotify_device *dev) | ||
247 | { | ||
248 | return list_entry(dev->events.next, struct inotify_kernel_event, list); | ||
249 | } | ||
250 | |||
251 | /* | ||
252 | * inotify_dev_get_last_event - return the last event in the given dev's queue | ||
253 | * | ||
254 | * Caller must hold dev->ev_mutex. | ||
255 | */ | ||
256 | static inline struct inotify_kernel_event * | ||
257 | inotify_dev_get_last_event(struct inotify_device *dev) | ||
258 | { | ||
259 | if (list_empty(&dev->events)) | ||
260 | return NULL; | ||
261 | return list_entry(dev->events.prev, struct inotify_kernel_event, list); | ||
262 | } | ||
263 | |||
264 | /* | ||
265 | * inotify_dev_queue_event - event handler registered with core inotify, adds | ||
266 | * a new event to the given device | ||
267 | * | ||
268 | * Can sleep (calls kernel_event()). | ||
269 | */ | ||
270 | static void inotify_dev_queue_event(struct inotify_watch *w, u32 wd, u32 mask, | ||
271 | u32 cookie, const char *name, | ||
272 | struct inode *ignored) | ||
273 | { | ||
274 | struct inotify_user_watch *watch; | ||
275 | struct inotify_device *dev; | ||
276 | struct inotify_kernel_event *kevent, *last; | ||
277 | |||
278 | watch = container_of(w, struct inotify_user_watch, wdata); | ||
279 | dev = watch->dev; | ||
280 | |||
281 | mutex_lock(&dev->ev_mutex); | ||
282 | |||
283 | /* we can safely put the watch as we don't reference it while | ||
284 | * generating the event | ||
285 | */ | ||
286 | if (mask & IN_IGNORED || w->mask & IN_ONESHOT) | ||
287 | put_inotify_watch(w); /* final put */ | ||
288 | |||
289 | /* coalescing: drop this event if it is a dupe of the previous */ | ||
290 | last = inotify_dev_get_last_event(dev); | ||
291 | if (last && last->event.mask == mask && last->event.wd == wd && | ||
292 | last->event.cookie == cookie) { | ||
293 | const char *lastname = last->name; | ||
294 | |||
295 | if (!name && !lastname) | ||
296 | goto out; | ||
297 | if (name && lastname && !strcmp(lastname, name)) | ||
298 | goto out; | ||
299 | } | ||
300 | |||
301 | /* the queue overflowed and we already sent the Q_OVERFLOW event */ | ||
302 | if (unlikely(dev->event_count > dev->max_events)) | ||
303 | goto out; | ||
304 | |||
305 | /* if the queue overflows, we need to notify user space */ | ||
306 | if (unlikely(dev->event_count == dev->max_events)) | ||
307 | kevent = kernel_event(-1, IN_Q_OVERFLOW, cookie, NULL); | ||
308 | else | ||
309 | kevent = kernel_event(wd, mask, cookie, name); | ||
310 | |||
311 | if (unlikely(!kevent)) | ||
312 | goto out; | ||
313 | |||
314 | /* queue the event and wake up anyone waiting */ | ||
315 | dev->event_count++; | ||
316 | dev->queue_size += sizeof(struct inotify_event) + kevent->event.len; | ||
317 | list_add_tail(&kevent->list, &dev->events); | ||
318 | wake_up_interruptible(&dev->wq); | ||
319 | kill_fasync(&dev->fa, SIGIO, POLL_IN); | ||
320 | |||
321 | out: | ||
322 | mutex_unlock(&dev->ev_mutex); | ||
323 | } | ||
324 | |||
325 | /* | ||
326 | * remove_kevent - cleans up the given kevent | ||
327 | * | ||
328 | * Caller must hold dev->ev_mutex. | ||
329 | */ | ||
330 | static void remove_kevent(struct inotify_device *dev, | ||
331 | struct inotify_kernel_event *kevent) | ||
332 | { | ||
333 | list_del(&kevent->list); | ||
334 | |||
335 | dev->event_count--; | ||
336 | dev->queue_size -= sizeof(struct inotify_event) + kevent->event.len; | ||
337 | } | ||
338 | |||
339 | /* | ||
340 | * free_kevent - frees the given kevent. | ||
341 | */ | ||
342 | static void free_kevent(struct inotify_kernel_event *kevent) | ||
343 | { | ||
344 | kfree(kevent->name); | ||
345 | kmem_cache_free(event_cachep, kevent); | ||
346 | } | ||
347 | |||
348 | /* | ||
349 | * inotify_dev_event_dequeue - destroy an event on the given device | ||
350 | * | ||
351 | * Caller must hold dev->ev_mutex. | ||
352 | */ | ||
353 | static void inotify_dev_event_dequeue(struct inotify_device *dev) | ||
354 | { | ||
355 | if (!list_empty(&dev->events)) { | ||
356 | struct inotify_kernel_event *kevent; | ||
357 | kevent = inotify_dev_get_event(dev); | ||
358 | remove_kevent(dev, kevent); | ||
359 | free_kevent(kevent); | ||
360 | } | ||
361 | } | ||
362 | |||
363 | /* | ||
364 | * find_inode - resolve a user-given path to a specific inode | ||
365 | */ | ||
366 | static int find_inode(const char __user *dirname, struct path *path, | ||
367 | unsigned flags) | ||
368 | { | ||
369 | int error; | ||
370 | |||
371 | error = user_path_at(AT_FDCWD, dirname, flags, path); | ||
372 | if (error) | ||
373 | return error; | ||
374 | /* you can only watch an inode if you have read permissions on it */ | ||
375 | error = inode_permission(path->dentry->d_inode, MAY_READ); | ||
376 | if (error) | ||
377 | path_put(path); | ||
378 | return error; | ||
379 | } | ||
380 | |||
381 | /* | ||
382 | * create_watch - creates a watch on the given device. | ||
383 | * | ||
384 | * Callers must hold dev->up_mutex. | ||
385 | */ | ||
386 | static int create_watch(struct inotify_device *dev, struct inode *inode, | ||
387 | u32 mask) | ||
388 | { | ||
389 | struct inotify_user_watch *watch; | ||
390 | int ret; | ||
391 | |||
392 | if (atomic_read(&dev->user->inotify_watches) >= | ||
393 | inotify_max_user_watches) | ||
394 | return -ENOSPC; | ||
395 | |||
396 | watch = kmem_cache_alloc(watch_cachep, GFP_KERNEL); | ||
397 | if (unlikely(!watch)) | ||
398 | return -ENOMEM; | ||
399 | |||
400 | /* save a reference to device and bump the count to make it official */ | ||
401 | get_inotify_dev(dev); | ||
402 | watch->dev = dev; | ||
403 | |||
404 | atomic_inc(&dev->user->inotify_watches); | ||
405 | |||
406 | inotify_init_watch(&watch->wdata); | ||
407 | ret = inotify_add_watch(dev->ih, &watch->wdata, inode, mask); | ||
408 | if (ret < 0) | ||
409 | free_inotify_user_watch(&watch->wdata); | ||
410 | |||
411 | return ret; | ||
412 | } | ||
413 | |||
414 | /* Device Interface */ | ||
415 | |||
416 | static unsigned int inotify_poll(struct file *file, poll_table *wait) | ||
417 | { | ||
418 | struct inotify_device *dev = file->private_data; | ||
419 | int ret = 0; | ||
420 | |||
421 | poll_wait(file, &dev->wq, wait); | ||
422 | mutex_lock(&dev->ev_mutex); | ||
423 | if (!list_empty(&dev->events)) | ||
424 | ret = POLLIN | POLLRDNORM; | ||
425 | mutex_unlock(&dev->ev_mutex); | ||
426 | |||
427 | return ret; | ||
428 | } | ||
429 | |||
430 | static ssize_t inotify_read(struct file *file, char __user *buf, | ||
431 | size_t count, loff_t *pos) | ||
432 | { | ||
433 | size_t event_size = sizeof (struct inotify_event); | ||
434 | struct inotify_device *dev; | ||
435 | char __user *start; | ||
436 | int ret; | ||
437 | DEFINE_WAIT(wait); | ||
438 | |||
439 | start = buf; | ||
440 | dev = file->private_data; | ||
441 | |||
442 | while (1) { | ||
443 | |||
444 | prepare_to_wait(&dev->wq, &wait, TASK_INTERRUPTIBLE); | ||
445 | |||
446 | mutex_lock(&dev->ev_mutex); | ||
447 | if (!list_empty(&dev->events)) { | ||
448 | ret = 0; | ||
449 | break; | ||
450 | } | ||
451 | mutex_unlock(&dev->ev_mutex); | ||
452 | |||
453 | if (file->f_flags & O_NONBLOCK) { | ||
454 | ret = -EAGAIN; | ||
455 | break; | ||
456 | } | ||
457 | |||
458 | if (signal_pending(current)) { | ||
459 | ret = -EINTR; | ||
460 | break; | ||
461 | } | ||
462 | |||
463 | schedule(); | ||
464 | } | ||
465 | |||
466 | finish_wait(&dev->wq, &wait); | ||
467 | if (ret) | ||
468 | return ret; | ||
469 | |||
470 | while (1) { | ||
471 | struct inotify_kernel_event *kevent; | ||
472 | |||
473 | ret = buf - start; | ||
474 | if (list_empty(&dev->events)) | ||
475 | break; | ||
476 | |||
477 | kevent = inotify_dev_get_event(dev); | ||
478 | if (event_size + kevent->event.len > count) { | ||
479 | if (ret == 0 && count > 0) { | ||
480 | /* | ||
481 | * could not get a single event because we | ||
482 | * didn't have enough buffer space. | ||
483 | */ | ||
484 | ret = -EINVAL; | ||
485 | } | ||
486 | break; | ||
487 | } | ||
488 | remove_kevent(dev, kevent); | ||
489 | |||
490 | /* | ||
491 | * Must perform the copy_to_user outside the mutex in order | ||
492 | * to avoid a lock order reversal with mmap_sem. | ||
493 | */ | ||
494 | mutex_unlock(&dev->ev_mutex); | ||
495 | |||
496 | if (copy_to_user(buf, &kevent->event, event_size)) { | ||
497 | ret = -EFAULT; | ||
498 | break; | ||
499 | } | ||
500 | buf += event_size; | ||
501 | count -= event_size; | ||
502 | |||
503 | if (kevent->name) { | ||
504 | if (copy_to_user(buf, kevent->name, kevent->event.len)){ | ||
505 | ret = -EFAULT; | ||
506 | break; | ||
507 | } | ||
508 | buf += kevent->event.len; | ||
509 | count -= kevent->event.len; | ||
510 | } | ||
511 | |||
512 | free_kevent(kevent); | ||
513 | |||
514 | mutex_lock(&dev->ev_mutex); | ||
515 | } | ||
516 | mutex_unlock(&dev->ev_mutex); | ||
517 | |||
518 | return ret; | ||
519 | } | ||
520 | |||
521 | static int inotify_fasync(int fd, struct file *file, int on) | ||
522 | { | ||
523 | struct inotify_device *dev = file->private_data; | ||
524 | |||
525 | return fasync_helper(fd, file, on, &dev->fa) >= 0 ? 0 : -EIO; | ||
526 | } | ||
527 | |||
528 | static int inotify_release(struct inode *ignored, struct file *file) | ||
529 | { | ||
530 | struct inotify_device *dev = file->private_data; | ||
531 | |||
532 | inotify_destroy(dev->ih); | ||
533 | |||
534 | /* destroy all of the events on this device */ | ||
535 | mutex_lock(&dev->ev_mutex); | ||
536 | while (!list_empty(&dev->events)) | ||
537 | inotify_dev_event_dequeue(dev); | ||
538 | mutex_unlock(&dev->ev_mutex); | ||
539 | |||
540 | /* free this device: the put matching the get in inotify_init() */ | ||
541 | put_inotify_dev(dev); | ||
542 | |||
543 | return 0; | ||
544 | } | ||
545 | |||
546 | static long inotify_ioctl(struct file *file, unsigned int cmd, | ||
547 | unsigned long arg) | ||
548 | { | ||
549 | struct inotify_device *dev; | ||
550 | void __user *p; | ||
551 | int ret = -ENOTTY; | ||
552 | |||
553 | dev = file->private_data; | ||
554 | p = (void __user *) arg; | ||
555 | |||
556 | switch (cmd) { | ||
557 | case FIONREAD: | ||
558 | ret = put_user(dev->queue_size, (int __user *) p); | ||
559 | break; | ||
560 | } | ||
561 | |||
562 | return ret; | ||
563 | } | ||
564 | |||
565 | static const struct file_operations inotify_fops = { | ||
566 | .poll = inotify_poll, | ||
567 | .read = inotify_read, | ||
568 | .fasync = inotify_fasync, | ||
569 | .release = inotify_release, | ||
570 | .unlocked_ioctl = inotify_ioctl, | ||
571 | .compat_ioctl = inotify_ioctl, | ||
572 | }; | ||
573 | |||
574 | static const struct inotify_operations inotify_user_ops = { | ||
575 | .handle_event = inotify_dev_queue_event, | ||
576 | .destroy_watch = free_inotify_user_watch, | ||
577 | }; | ||
578 | |||
579 | asmlinkage long sys_inotify_init1(int flags) | ||
580 | { | ||
581 | struct inotify_device *dev; | ||
582 | struct inotify_handle *ih; | ||
583 | struct user_struct *user; | ||
584 | struct file *filp; | ||
585 | int fd, ret; | ||
586 | |||
587 | /* Check the IN_* constants for consistency. */ | ||
588 | BUILD_BUG_ON(IN_CLOEXEC != O_CLOEXEC); | ||
589 | BUILD_BUG_ON(IN_NONBLOCK != O_NONBLOCK); | ||
590 | |||
591 | if (flags & ~(IN_CLOEXEC | IN_NONBLOCK)) | ||
592 | return -EINVAL; | ||
593 | |||
594 | fd = get_unused_fd_flags(flags & O_CLOEXEC); | ||
595 | if (fd < 0) | ||
596 | return fd; | ||
597 | |||
598 | filp = get_empty_filp(); | ||
599 | if (!filp) { | ||
600 | ret = -ENFILE; | ||
601 | goto out_put_fd; | ||
602 | } | ||
603 | |||
604 | user = get_current_user(); | ||
605 | if (unlikely(atomic_read(&user->inotify_devs) >= | ||
606 | inotify_max_user_instances)) { | ||
607 | ret = -EMFILE; | ||
608 | goto out_free_uid; | ||
609 | } | ||
610 | |||
611 | dev = kmalloc(sizeof(struct inotify_device), GFP_KERNEL); | ||
612 | if (unlikely(!dev)) { | ||
613 | ret = -ENOMEM; | ||
614 | goto out_free_uid; | ||
615 | } | ||
616 | |||
617 | ih = inotify_init(&inotify_user_ops); | ||
618 | if (IS_ERR(ih)) { | ||
619 | ret = PTR_ERR(ih); | ||
620 | goto out_free_dev; | ||
621 | } | ||
622 | dev->ih = ih; | ||
623 | dev->fa = NULL; | ||
624 | |||
625 | filp->f_op = &inotify_fops; | ||
626 | filp->f_path.mnt = mntget(inotify_mnt); | ||
627 | filp->f_path.dentry = dget(inotify_mnt->mnt_root); | ||
628 | filp->f_mapping = filp->f_path.dentry->d_inode->i_mapping; | ||
629 | filp->f_mode = FMODE_READ; | ||
630 | filp->f_flags = O_RDONLY | (flags & O_NONBLOCK); | ||
631 | filp->private_data = dev; | ||
632 | |||
633 | INIT_LIST_HEAD(&dev->events); | ||
634 | init_waitqueue_head(&dev->wq); | ||
635 | mutex_init(&dev->ev_mutex); | ||
636 | mutex_init(&dev->up_mutex); | ||
637 | dev->event_count = 0; | ||
638 | dev->queue_size = 0; | ||
639 | dev->max_events = inotify_max_queued_events; | ||
640 | dev->user = user; | ||
641 | atomic_set(&dev->count, 0); | ||
642 | |||
643 | get_inotify_dev(dev); | ||
644 | atomic_inc(&user->inotify_devs); | ||
645 | fd_install(fd, filp); | ||
646 | |||
647 | return fd; | ||
648 | out_free_dev: | ||
649 | kfree(dev); | ||
650 | out_free_uid: | ||
651 | free_uid(user); | ||
652 | put_filp(filp); | ||
653 | out_put_fd: | ||
654 | put_unused_fd(fd); | ||
655 | return ret; | ||
656 | } | ||
657 | |||
658 | asmlinkage long sys_inotify_init(void) | ||
659 | { | ||
660 | return sys_inotify_init1(0); | ||
661 | } | ||
662 | |||
663 | asmlinkage long sys_inotify_add_watch(int fd, const char __user *pathname, u32 mask) | ||
664 | { | ||
665 | struct inode *inode; | ||
666 | struct inotify_device *dev; | ||
667 | struct path path; | ||
668 | struct file *filp; | ||
669 | int ret, fput_needed; | ||
670 | unsigned flags = 0; | ||
671 | |||
672 | filp = fget_light(fd, &fput_needed); | ||
673 | if (unlikely(!filp)) | ||
674 | return -EBADF; | ||
675 | |||
676 | /* verify that this is indeed an inotify instance */ | ||
677 | if (unlikely(filp->f_op != &inotify_fops)) { | ||
678 | ret = -EINVAL; | ||
679 | goto fput_and_out; | ||
680 | } | ||
681 | |||
682 | if (!(mask & IN_DONT_FOLLOW)) | ||
683 | flags |= LOOKUP_FOLLOW; | ||
684 | if (mask & IN_ONLYDIR) | ||
685 | flags |= LOOKUP_DIRECTORY; | ||
686 | |||
687 | ret = find_inode(pathname, &path, flags); | ||
688 | if (unlikely(ret)) | ||
689 | goto fput_and_out; | ||
690 | |||
691 | /* inode held in place by reference to path; dev by fget on fd */ | ||
692 | inode = path.dentry->d_inode; | ||
693 | dev = filp->private_data; | ||
694 | |||
695 | mutex_lock(&dev->up_mutex); | ||
696 | ret = inotify_find_update_watch(dev->ih, inode, mask); | ||
697 | if (ret == -ENOENT) | ||
698 | ret = create_watch(dev, inode, mask); | ||
699 | mutex_unlock(&dev->up_mutex); | ||
700 | |||
701 | path_put(&path); | ||
702 | fput_and_out: | ||
703 | fput_light(filp, fput_needed); | ||
704 | return ret; | ||
705 | } | ||
706 | |||
707 | asmlinkage long sys_inotify_rm_watch(int fd, u32 wd) | ||
708 | { | ||
709 | struct file *filp; | ||
710 | struct inotify_device *dev; | ||
711 | int ret, fput_needed; | ||
712 | |||
713 | filp = fget_light(fd, &fput_needed); | ||
714 | if (unlikely(!filp)) | ||
715 | return -EBADF; | ||
716 | |||
717 | /* verify that this is indeed an inotify instance */ | ||
718 | if (unlikely(filp->f_op != &inotify_fops)) { | ||
719 | ret = -EINVAL; | ||
720 | goto out; | ||
721 | } | ||
722 | |||
723 | dev = filp->private_data; | ||
724 | |||
725 | /* we free our watch data when we get IN_IGNORED */ | ||
726 | ret = inotify_rm_wd(dev->ih, wd); | ||
727 | |||
728 | out: | ||
729 | fput_light(filp, fput_needed); | ||
730 | return ret; | ||
731 | } | ||
732 | |||
733 | static int | ||
734 | inotify_get_sb(struct file_system_type *fs_type, int flags, | ||
735 | const char *dev_name, void *data, struct vfsmount *mnt) | ||
736 | { | ||
737 | return get_sb_pseudo(fs_type, "inotify", NULL, | ||
738 | INOTIFYFS_SUPER_MAGIC, mnt); | ||
739 | } | ||
740 | |||
741 | static struct file_system_type inotify_fs_type = { | ||
742 | .name = "inotifyfs", | ||
743 | .get_sb = inotify_get_sb, | ||
744 | .kill_sb = kill_anon_super, | ||
745 | }; | ||
746 | |||
747 | /* | ||
748 | * inotify_user_setup - Our initialization function. Note that we cannnot return | ||
749 | * error because we have compiled-in VFS hooks. So an (unlikely) failure here | ||
750 | * must result in panic(). | ||
751 | */ | ||
752 | static int __init inotify_user_setup(void) | ||
753 | { | ||
754 | int ret; | ||
755 | |||
756 | ret = register_filesystem(&inotify_fs_type); | ||
757 | if (unlikely(ret)) | ||
758 | panic("inotify: register_filesystem returned %d!\n", ret); | ||
759 | |||
760 | inotify_mnt = kern_mount(&inotify_fs_type); | ||
761 | if (IS_ERR(inotify_mnt)) | ||
762 | panic("inotify: kern_mount ret %ld!\n", PTR_ERR(inotify_mnt)); | ||
763 | |||
764 | inotify_max_queued_events = 16384; | ||
765 | inotify_max_user_instances = 128; | ||
766 | inotify_max_user_watches = 8192; | ||
767 | |||
768 | watch_cachep = kmem_cache_create("inotify_watch_cache", | ||
769 | sizeof(struct inotify_user_watch), | ||
770 | 0, SLAB_PANIC, NULL); | ||
771 | event_cachep = kmem_cache_create("inotify_event_cache", | ||
772 | sizeof(struct inotify_kernel_event), | ||
773 | 0, SLAB_PANIC, NULL); | ||
774 | |||
775 | return 0; | ||
776 | } | ||
777 | |||
778 | module_init(inotify_user_setup); | ||