]>
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> |
174cd4b1 | 17 | #include <linux/sched/signal.h> |
a1014f10 EP |
18 | |
19 | #include <asm/ioctls.h> | |
11637e4b | 20 | |
c63181e6 | 21 | #include "../../mount.h" |
be77196b | 22 | #include "../fdinfo.h" |
7053aee2 | 23 | #include "fanotify.h" |
c63181e6 | 24 | |
2529a0df | 25 | #define FANOTIFY_DEFAULT_MAX_EVENTS 16384 |
e7099d8a | 26 | #define FANOTIFY_DEFAULT_MAX_MARKS 8192 |
4afeff85 | 27 | #define FANOTIFY_DEFAULT_MAX_LISTENERS 128 |
2529a0df | 28 | |
48149e9d HS |
29 | /* |
30 | * All flags that may be specified in parameter event_f_flags of fanotify_init. | |
31 | * | |
32 | * Internal and external open flags are stored together in field f_flags of | |
33 | * struct file. Only external open flags shall be allowed in event_f_flags. | |
34 | * Internal flags like FMODE_NONOTIFY, FMODE_EXEC, FMODE_NOCMTIME shall be | |
35 | * excluded. | |
36 | */ | |
37 | #define FANOTIFY_INIT_ALL_EVENT_F_BITS ( \ | |
38 | O_ACCMODE | O_APPEND | O_NONBLOCK | \ | |
39 | __O_SYNC | O_DSYNC | O_CLOEXEC | \ | |
40 | O_LARGEFILE | O_NOATIME ) | |
41 | ||
33d3dfff | 42 | extern const struct fsnotify_ops fanotify_fsnotify_ops; |
11637e4b | 43 | |
054c636e | 44 | struct kmem_cache *fanotify_mark_cache __read_mostly; |
7053aee2 | 45 | struct kmem_cache *fanotify_event_cachep __read_mostly; |
f083441b | 46 | struct kmem_cache *fanotify_perm_event_cachep __read_mostly; |
2a3edf86 | 47 | |
a1014f10 EP |
48 | /* |
49 | * Get an fsnotify notification event if one exists and is small | |
50 | * enough to fit in "count". Return an error pointer if the count | |
51 | * is not large enough. | |
52 | * | |
c21dbe20 | 53 | * Called with the group->notification_lock held. |
a1014f10 EP |
54 | */ |
55 | static struct fsnotify_event *get_one_event(struct fsnotify_group *group, | |
56 | size_t count) | |
57 | { | |
ed272640 | 58 | assert_spin_locked(&group->notification_lock); |
a1014f10 EP |
59 | |
60 | pr_debug("%s: group=%p count=%zd\n", __func__, group, count); | |
61 | ||
62 | if (fsnotify_notify_queue_is_empty(group)) | |
63 | return NULL; | |
64 | ||
65 | if (FAN_EVENT_METADATA_LEN > count) | |
66 | return ERR_PTR(-EINVAL); | |
67 | ||
c21dbe20 | 68 | /* held the notification_lock the whole time, so this is the |
a1014f10 | 69 | * same event we peeked above */ |
8ba8fa91 | 70 | return fsnotify_remove_first_event(group); |
a1014f10 EP |
71 | } |
72 | ||
352e3b24 | 73 | static int create_fd(struct fsnotify_group *group, |
7053aee2 JK |
74 | struct fanotify_event_info *event, |
75 | struct file **file) | |
a1014f10 EP |
76 | { |
77 | int client_fd; | |
a1014f10 EP |
78 | struct file *new_file; |
79 | ||
22aa425d | 80 | pr_debug("%s: group=%p event=%p\n", __func__, group, event); |
a1014f10 | 81 | |
0b37e097 | 82 | client_fd = get_unused_fd_flags(group->fanotify_data.f_flags); |
a1014f10 EP |
83 | if (client_fd < 0) |
84 | return client_fd; | |
85 | ||
a1014f10 EP |
86 | /* |
87 | * we need a new file handle for the userspace program so it can read even if it was | |
88 | * originally opened O_WRONLY. | |
89 | */ | |
a1014f10 EP |
90 | /* it's possible this event was an overflow event. in that case dentry and mnt |
91 | * are NULL; That's fine, just don't call dentry open */ | |
765927b2 AV |
92 | if (event->path.dentry && event->path.mnt) |
93 | new_file = dentry_open(&event->path, | |
80af2588 | 94 | group->fanotify_data.f_flags | FMODE_NONOTIFY, |
a1014f10 EP |
95 | current_cred()); |
96 | else | |
97 | new_file = ERR_PTR(-EOVERFLOW); | |
98 | if (IS_ERR(new_file)) { | |
99 | /* | |
100 | * we still send an event even if we can't open the file. this | |
101 | * can happen when say tasks are gone and we try to open their | |
102 | * /proc files or we try to open a WRONLY file like in sysfs | |
103 | * we just send the errno to userspace since there isn't much | |
104 | * else we can do. | |
105 | */ | |
106 | put_unused_fd(client_fd); | |
107 | client_fd = PTR_ERR(new_file); | |
108 | } else { | |
352e3b24 | 109 | *file = new_file; |
a1014f10 EP |
110 | } |
111 | ||
22aa425d | 112 | return client_fd; |
a1014f10 EP |
113 | } |
114 | ||
ecf6f5e7 | 115 | static int fill_event_metadata(struct fsnotify_group *group, |
7053aee2 JK |
116 | struct fanotify_event_metadata *metadata, |
117 | struct fsnotify_event *fsn_event, | |
118 | struct file **file) | |
a1014f10 | 119 | { |
fdbf3cee | 120 | int ret = 0; |
7053aee2 | 121 | struct fanotify_event_info *event; |
fdbf3cee | 122 | |
a1014f10 | 123 | pr_debug("%s: group=%p metadata=%p event=%p\n", __func__, |
7053aee2 | 124 | group, metadata, fsn_event); |
a1014f10 | 125 | |
352e3b24 | 126 | *file = NULL; |
7053aee2 | 127 | event = container_of(fsn_event, struct fanotify_event_info, fse); |
a1014f10 | 128 | metadata->event_len = FAN_EVENT_METADATA_LEN; |
7d131623 | 129 | metadata->metadata_len = FAN_EVENT_METADATA_LEN; |
a1014f10 | 130 | metadata->vers = FANOTIFY_METADATA_VERSION; |
de1e0c40 | 131 | metadata->reserved = 0; |
7053aee2 | 132 | metadata->mask = fsn_event->mask & FAN_ALL_OUTGOING_EVENTS; |
32c32632 | 133 | metadata->pid = pid_vnr(event->tgid); |
7053aee2 | 134 | if (unlikely(fsn_event->mask & FAN_Q_OVERFLOW)) |
fdbf3cee LS |
135 | metadata->fd = FAN_NOFD; |
136 | else { | |
352e3b24 | 137 | metadata->fd = create_fd(group, event, file); |
fdbf3cee LS |
138 | if (metadata->fd < 0) |
139 | ret = metadata->fd; | |
140 | } | |
a1014f10 | 141 | |
fdbf3cee | 142 | return ret; |
a1014f10 EP |
143 | } |
144 | ||
b2d87909 | 145 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
f083441b JK |
146 | static struct fanotify_perm_event_info *dequeue_event( |
147 | struct fsnotify_group *group, int fd) | |
b2d87909 | 148 | { |
f083441b | 149 | struct fanotify_perm_event_info *event, *return_e = NULL; |
b2d87909 | 150 | |
073f6552 | 151 | spin_lock(&group->notification_lock); |
f083441b JK |
152 | list_for_each_entry(event, &group->fanotify_data.access_list, |
153 | fae.fse.list) { | |
154 | if (event->fd != fd) | |
b2d87909 EP |
155 | continue; |
156 | ||
f083441b JK |
157 | list_del_init(&event->fae.fse.list); |
158 | return_e = event; | |
b2d87909 EP |
159 | break; |
160 | } | |
073f6552 | 161 | spin_unlock(&group->notification_lock); |
b2d87909 | 162 | |
f083441b | 163 | pr_debug("%s: found return_re=%p\n", __func__, return_e); |
b2d87909 | 164 | |
f083441b | 165 | return return_e; |
b2d87909 EP |
166 | } |
167 | ||
168 | static int process_access_response(struct fsnotify_group *group, | |
169 | struct fanotify_response *response_struct) | |
170 | { | |
f083441b JK |
171 | struct fanotify_perm_event_info *event; |
172 | int fd = response_struct->fd; | |
173 | int response = response_struct->response; | |
b2d87909 EP |
174 | |
175 | pr_debug("%s: group=%p fd=%d response=%d\n", __func__, group, | |
176 | fd, response); | |
177 | /* | |
178 | * make sure the response is valid, if invalid we do nothing and either | |
25985edc | 179 | * userspace can send a valid response or we will clean it up after the |
b2d87909 EP |
180 | * timeout |
181 | */ | |
182 | switch (response) { | |
183 | case FAN_ALLOW: | |
184 | case FAN_DENY: | |
185 | break; | |
186 | default: | |
187 | return -EINVAL; | |
188 | } | |
189 | ||
190 | if (fd < 0) | |
191 | return -EINVAL; | |
192 | ||
f083441b JK |
193 | event = dequeue_event(group, fd); |
194 | if (!event) | |
b2d87909 EP |
195 | return -ENOENT; |
196 | ||
f083441b | 197 | event->response = response; |
b2d87909 EP |
198 | wake_up(&group->fanotify_data.access_waitq); |
199 | ||
b2d87909 EP |
200 | return 0; |
201 | } | |
b2d87909 EP |
202 | #endif |
203 | ||
a1014f10 EP |
204 | static ssize_t copy_event_to_user(struct fsnotify_group *group, |
205 | struct fsnotify_event *event, | |
206 | char __user *buf) | |
207 | { | |
208 | struct fanotify_event_metadata fanotify_event_metadata; | |
352e3b24 | 209 | struct file *f; |
b2d87909 | 210 | int fd, ret; |
a1014f10 EP |
211 | |
212 | pr_debug("%s: group=%p event=%p\n", __func__, group, event); | |
213 | ||
352e3b24 | 214 | ret = fill_event_metadata(group, &fanotify_event_metadata, event, &f); |
ecf6f5e7 | 215 | if (ret < 0) |
d507816b | 216 | return ret; |
b2d87909 | 217 | |
fdbf3cee | 218 | fd = fanotify_event_metadata.fd; |
b2d87909 | 219 | ret = -EFAULT; |
7d131623 EP |
220 | if (copy_to_user(buf, &fanotify_event_metadata, |
221 | fanotify_event_metadata.event_len)) | |
352e3b24 AV |
222 | goto out_close_fd; |
223 | ||
f083441b | 224 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
d507816b JK |
225 | if (event->mask & FAN_ALL_PERM_EVENTS) |
226 | FANOTIFY_PE(event)->fd = fd; | |
f083441b | 227 | #endif |
a1014f10 | 228 | |
3587b1b0 AV |
229 | if (fd != FAN_NOFD) |
230 | fd_install(fd, f); | |
7d131623 | 231 | return fanotify_event_metadata.event_len; |
b2d87909 | 232 | |
b2d87909 | 233 | out_close_fd: |
352e3b24 AV |
234 | if (fd != FAN_NOFD) { |
235 | put_unused_fd(fd); | |
236 | fput(f); | |
237 | } | |
b2d87909 | 238 | return ret; |
a1014f10 EP |
239 | } |
240 | ||
241 | /* intofiy userspace file descriptor functions */ | |
242 | static unsigned int fanotify_poll(struct file *file, poll_table *wait) | |
243 | { | |
244 | struct fsnotify_group *group = file->private_data; | |
245 | int ret = 0; | |
246 | ||
247 | poll_wait(file, &group->notification_waitq, wait); | |
c21dbe20 | 248 | spin_lock(&group->notification_lock); |
a1014f10 EP |
249 | if (!fsnotify_notify_queue_is_empty(group)) |
250 | ret = POLLIN | POLLRDNORM; | |
c21dbe20 | 251 | spin_unlock(&group->notification_lock); |
a1014f10 EP |
252 | |
253 | return ret; | |
254 | } | |
255 | ||
256 | static ssize_t fanotify_read(struct file *file, char __user *buf, | |
257 | size_t count, loff_t *pos) | |
258 | { | |
259 | struct fsnotify_group *group; | |
260 | struct fsnotify_event *kevent; | |
261 | char __user *start; | |
262 | int ret; | |
536ebe9c | 263 | DEFINE_WAIT_FUNC(wait, woken_wake_function); |
a1014f10 EP |
264 | |
265 | start = buf; | |
266 | group = file->private_data; | |
267 | ||
268 | pr_debug("%s: group=%p\n", __func__, group); | |
269 | ||
536ebe9c | 270 | add_wait_queue(&group->notification_waitq, &wait); |
a1014f10 | 271 | while (1) { |
c21dbe20 | 272 | spin_lock(&group->notification_lock); |
a1014f10 | 273 | kevent = get_one_event(group, count); |
c21dbe20 | 274 | spin_unlock(&group->notification_lock); |
a1014f10 | 275 | |
d8aaab4f | 276 | if (IS_ERR(kevent)) { |
a1014f10 | 277 | ret = PTR_ERR(kevent); |
d8aaab4f JK |
278 | break; |
279 | } | |
280 | ||
281 | if (!kevent) { | |
282 | ret = -EAGAIN; | |
283 | if (file->f_flags & O_NONBLOCK) | |
a1014f10 | 284 | break; |
d8aaab4f JK |
285 | |
286 | ret = -ERESTARTSYS; | |
287 | if (signal_pending(current)) | |
288 | break; | |
289 | ||
290 | if (start != buf) | |
a1014f10 | 291 | break; |
536ebe9c PZ |
292 | |
293 | wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT); | |
a1014f10 EP |
294 | continue; |
295 | } | |
296 | ||
d8aaab4f | 297 | ret = copy_event_to_user(group, kevent, buf); |
4ff33aaf AG |
298 | if (unlikely(ret == -EOPENSTALE)) { |
299 | /* | |
300 | * We cannot report events with stale fd so drop it. | |
301 | * Setting ret to 0 will continue the event loop and | |
302 | * do the right thing if there are no more events to | |
303 | * read (i.e. return bytes read, -EAGAIN or wait). | |
304 | */ | |
305 | ret = 0; | |
306 | } | |
307 | ||
d8aaab4f JK |
308 | /* |
309 | * Permission events get queued to wait for response. Other | |
310 | * events can be destroyed now. | |
311 | */ | |
d507816b | 312 | if (!(kevent->mask & FAN_ALL_PERM_EVENTS)) { |
d8aaab4f | 313 | fsnotify_destroy_event(group, kevent); |
d507816b JK |
314 | } else { |
315 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS | |
4ff33aaf | 316 | if (ret <= 0) { |
d507816b JK |
317 | FANOTIFY_PE(kevent)->response = FAN_DENY; |
318 | wake_up(&group->fanotify_data.access_waitq); | |
4ff33aaf AG |
319 | } else { |
320 | spin_lock(&group->notification_lock); | |
321 | list_add_tail(&kevent->list, | |
322 | &group->fanotify_data.access_list); | |
323 | spin_unlock(&group->notification_lock); | |
d507816b | 324 | } |
d507816b JK |
325 | #endif |
326 | } | |
4ff33aaf AG |
327 | if (ret < 0) |
328 | break; | |
d8aaab4f JK |
329 | buf += ret; |
330 | count -= ret; | |
a1014f10 | 331 | } |
536ebe9c | 332 | remove_wait_queue(&group->notification_waitq, &wait); |
a1014f10 | 333 | |
a1014f10 EP |
334 | if (start != buf && ret != -EFAULT) |
335 | ret = buf - start; | |
336 | return ret; | |
337 | } | |
338 | ||
b2d87909 EP |
339 | static ssize_t fanotify_write(struct file *file, const char __user *buf, size_t count, loff_t *pos) |
340 | { | |
341 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS | |
342 | struct fanotify_response response = { .fd = -1, .response = -1 }; | |
343 | struct fsnotify_group *group; | |
344 | int ret; | |
345 | ||
346 | group = file->private_data; | |
347 | ||
348 | if (count > sizeof(response)) | |
349 | count = sizeof(response); | |
350 | ||
351 | pr_debug("%s: group=%p count=%zu\n", __func__, group, count); | |
352 | ||
353 | if (copy_from_user(&response, buf, count)) | |
354 | return -EFAULT; | |
355 | ||
356 | ret = process_access_response(group, &response); | |
357 | if (ret < 0) | |
358 | count = ret; | |
359 | ||
360 | return count; | |
361 | #else | |
362 | return -EINVAL; | |
363 | #endif | |
364 | } | |
365 | ||
52c923dd EP |
366 | static int fanotify_release(struct inode *ignored, struct file *file) |
367 | { | |
368 | struct fsnotify_group *group = file->private_data; | |
52c923dd | 369 | |
2eebf582 | 370 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
f083441b | 371 | struct fanotify_perm_event_info *event, *next; |
96d41019 | 372 | struct fsnotify_event *fsn_event; |
19ba54f4 | 373 | |
5838d444 | 374 | /* |
96d41019 JK |
375 | * Stop new events from arriving in the notification queue. since |
376 | * userspace cannot use fanotify fd anymore, no event can enter or | |
377 | * leave access_list by now either. | |
5838d444 | 378 | */ |
96d41019 | 379 | fsnotify_group_stop_queueing(group); |
2eebf582 | 380 | |
96d41019 JK |
381 | /* |
382 | * Process all permission events on access_list and notification queue | |
383 | * and simulate reply from userspace. | |
384 | */ | |
073f6552 | 385 | spin_lock(&group->notification_lock); |
f083441b JK |
386 | list_for_each_entry_safe(event, next, &group->fanotify_data.access_list, |
387 | fae.fse.list) { | |
388 | pr_debug("%s: found group=%p event=%p\n", __func__, group, | |
389 | event); | |
2eebf582 | 390 | |
f083441b JK |
391 | list_del_init(&event->fae.fse.list); |
392 | event->response = FAN_ALLOW; | |
2eebf582 | 393 | } |
2eebf582 | 394 | |
5838d444 | 395 | /* |
96d41019 JK |
396 | * Destroy all non-permission events. For permission events just |
397 | * dequeue them and set the response. They will be freed once the | |
398 | * response is consumed and fanotify_get_response() returns. | |
5838d444 | 399 | */ |
96d41019 JK |
400 | while (!fsnotify_notify_queue_is_empty(group)) { |
401 | fsn_event = fsnotify_remove_first_event(group); | |
1404ff3c | 402 | if (!(fsn_event->mask & FAN_ALL_PERM_EVENTS)) { |
c21dbe20 | 403 | spin_unlock(&group->notification_lock); |
96d41019 | 404 | fsnotify_destroy_event(group, fsn_event); |
c21dbe20 | 405 | spin_lock(&group->notification_lock); |
1404ff3c | 406 | } else |
96d41019 JK |
407 | FANOTIFY_PE(fsn_event)->response = FAN_ALLOW; |
408 | } | |
c21dbe20 | 409 | spin_unlock(&group->notification_lock); |
96d41019 JK |
410 | |
411 | /* Response for all permission events it set, wakeup waiters */ | |
2eebf582 EP |
412 | wake_up(&group->fanotify_data.access_waitq); |
413 | #endif | |
0a6b6bd5 | 414 | |
52c923dd | 415 | /* matches the fanotify_init->fsnotify_alloc_group */ |
d8153d4d | 416 | fsnotify_destroy_group(group); |
52c923dd EP |
417 | |
418 | return 0; | |
419 | } | |
420 | ||
a1014f10 EP |
421 | static long fanotify_ioctl(struct file *file, unsigned int cmd, unsigned long arg) |
422 | { | |
423 | struct fsnotify_group *group; | |
7053aee2 | 424 | struct fsnotify_event *fsn_event; |
a1014f10 EP |
425 | void __user *p; |
426 | int ret = -ENOTTY; | |
427 | size_t send_len = 0; | |
428 | ||
429 | group = file->private_data; | |
430 | ||
431 | p = (void __user *) arg; | |
432 | ||
433 | switch (cmd) { | |
434 | case FIONREAD: | |
c21dbe20 | 435 | spin_lock(&group->notification_lock); |
7053aee2 | 436 | list_for_each_entry(fsn_event, &group->notification_list, list) |
a1014f10 | 437 | send_len += FAN_EVENT_METADATA_LEN; |
c21dbe20 | 438 | spin_unlock(&group->notification_lock); |
a1014f10 EP |
439 | ret = put_user(send_len, (int __user *) p); |
440 | break; | |
441 | } | |
442 | ||
443 | return ret; | |
444 | } | |
445 | ||
52c923dd | 446 | static const struct file_operations fanotify_fops = { |
be77196b | 447 | .show_fdinfo = fanotify_show_fdinfo, |
a1014f10 EP |
448 | .poll = fanotify_poll, |
449 | .read = fanotify_read, | |
b2d87909 | 450 | .write = fanotify_write, |
52c923dd EP |
451 | .fasync = NULL, |
452 | .release = fanotify_release, | |
a1014f10 EP |
453 | .unlocked_ioctl = fanotify_ioctl, |
454 | .compat_ioctl = fanotify_ioctl, | |
6038f373 | 455 | .llseek = noop_llseek, |
52c923dd EP |
456 | }; |
457 | ||
2a3edf86 EP |
458 | static int fanotify_find_path(int dfd, const char __user *filename, |
459 | struct path *path, unsigned int flags) | |
460 | { | |
461 | int ret; | |
462 | ||
463 | pr_debug("%s: dfd=%d filename=%p flags=%x\n", __func__, | |
464 | dfd, filename, flags); | |
465 | ||
466 | if (filename == NULL) { | |
2903ff01 | 467 | struct fd f = fdget(dfd); |
2a3edf86 EP |
468 | |
469 | ret = -EBADF; | |
2903ff01 | 470 | if (!f.file) |
2a3edf86 EP |
471 | goto out; |
472 | ||
473 | ret = -ENOTDIR; | |
474 | if ((flags & FAN_MARK_ONLYDIR) && | |
496ad9aa | 475 | !(S_ISDIR(file_inode(f.file)->i_mode))) { |
2903ff01 | 476 | fdput(f); |
2a3edf86 EP |
477 | goto out; |
478 | } | |
479 | ||
2903ff01 | 480 | *path = f.file->f_path; |
2a3edf86 | 481 | path_get(path); |
2903ff01 | 482 | fdput(f); |
2a3edf86 EP |
483 | } else { |
484 | unsigned int lookup_flags = 0; | |
485 | ||
486 | if (!(flags & FAN_MARK_DONT_FOLLOW)) | |
487 | lookup_flags |= LOOKUP_FOLLOW; | |
488 | if (flags & FAN_MARK_ONLYDIR) | |
489 | lookup_flags |= LOOKUP_DIRECTORY; | |
490 | ||
491 | ret = user_path_at(dfd, filename, lookup_flags, path); | |
492 | if (ret) | |
493 | goto out; | |
494 | } | |
495 | ||
496 | /* you can only watch an inode if you have read permissions on it */ | |
497 | ret = inode_permission(path->dentry->d_inode, MAY_READ); | |
498 | if (ret) | |
499 | path_put(path); | |
500 | out: | |
501 | return ret; | |
502 | } | |
503 | ||
b9e4e3bd EP |
504 | static __u32 fanotify_mark_remove_from_mask(struct fsnotify_mark *fsn_mark, |
505 | __u32 mask, | |
6dfbd149 LS |
506 | unsigned int flags, |
507 | int *destroy) | |
088b09b0 | 508 | { |
d2c1874c | 509 | __u32 oldmask = 0; |
088b09b0 AG |
510 | |
511 | spin_lock(&fsn_mark->lock); | |
b9e4e3bd | 512 | if (!(flags & FAN_MARK_IGNORED_MASK)) { |
66ba93c0 LS |
513 | __u32 tmask = fsn_mark->mask & ~mask; |
514 | ||
515 | if (flags & FAN_MARK_ONDIR) | |
516 | tmask &= ~FAN_ONDIR; | |
517 | ||
b9e4e3bd | 518 | oldmask = fsn_mark->mask; |
66d2b81b | 519 | fsn_mark->mask = tmask; |
b9e4e3bd | 520 | } else { |
d2c1874c | 521 | __u32 tmask = fsn_mark->ignored_mask & ~mask; |
66ba93c0 LS |
522 | if (flags & FAN_MARK_ONDIR) |
523 | tmask &= ~FAN_ONDIR; | |
66d2b81b | 524 | fsn_mark->ignored_mask = tmask; |
b9e4e3bd | 525 | } |
a118449a | 526 | *destroy = !(fsn_mark->mask | fsn_mark->ignored_mask); |
088b09b0 AG |
527 | spin_unlock(&fsn_mark->lock); |
528 | ||
088b09b0 AG |
529 | return mask & oldmask; |
530 | } | |
531 | ||
f3640192 | 532 | static int fanotify_remove_vfsmount_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
533 | struct vfsmount *mnt, __u32 mask, |
534 | unsigned int flags) | |
88826276 EP |
535 | { |
536 | struct fsnotify_mark *fsn_mark = NULL; | |
088b09b0 | 537 | __u32 removed; |
6dfbd149 | 538 | int destroy_mark; |
88826276 | 539 | |
7b18527c | 540 | mutex_lock(&group->mark_mutex); |
b1362edf JK |
541 | fsn_mark = fsnotify_find_mark(&real_mount(mnt)->mnt_fsnotify_marks, |
542 | group); | |
7b18527c LS |
543 | if (!fsn_mark) { |
544 | mutex_unlock(&group->mark_mutex); | |
f3640192 | 545 | return -ENOENT; |
7b18527c | 546 | } |
88826276 | 547 | |
6dfbd149 LS |
548 | removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags, |
549 | &destroy_mark); | |
c9747640 | 550 | if (removed & real_mount(mnt)->mnt_fsnotify_mask) |
8920d273 | 551 | fsnotify_recalc_mask(real_mount(mnt)->mnt_fsnotify_marks); |
6dfbd149 | 552 | if (destroy_mark) |
4712e722 | 553 | fsnotify_detach_mark(fsn_mark); |
7b18527c | 554 | mutex_unlock(&group->mark_mutex); |
4712e722 JK |
555 | if (destroy_mark) |
556 | fsnotify_free_mark(fsn_mark); | |
6dfbd149 | 557 | |
f3640192 | 558 | fsnotify_put_mark(fsn_mark); |
f3640192 AG |
559 | return 0; |
560 | } | |
2a3edf86 | 561 | |
f3640192 | 562 | static int fanotify_remove_inode_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
563 | struct inode *inode, __u32 mask, |
564 | unsigned int flags) | |
f3640192 AG |
565 | { |
566 | struct fsnotify_mark *fsn_mark = NULL; | |
567 | __u32 removed; | |
6dfbd149 | 568 | int destroy_mark; |
f3640192 | 569 | |
7b18527c | 570 | mutex_lock(&group->mark_mutex); |
b1362edf | 571 | fsn_mark = fsnotify_find_mark(&inode->i_fsnotify_marks, group); |
7b18527c LS |
572 | if (!fsn_mark) { |
573 | mutex_unlock(&group->mark_mutex); | |
88826276 | 574 | return -ENOENT; |
7b18527c | 575 | } |
88826276 | 576 | |
6dfbd149 LS |
577 | removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags, |
578 | &destroy_mark); | |
c9747640 | 579 | if (removed & inode->i_fsnotify_mask) |
8920d273 | 580 | fsnotify_recalc_mask(inode->i_fsnotify_marks); |
6dfbd149 | 581 | if (destroy_mark) |
4712e722 | 582 | fsnotify_detach_mark(fsn_mark); |
7b18527c | 583 | mutex_unlock(&group->mark_mutex); |
4712e722 JK |
584 | if (destroy_mark) |
585 | fsnotify_free_mark(fsn_mark); | |
7b18527c | 586 | |
b1362edf | 587 | /* matches the fsnotify_find_mark() */ |
2a3edf86 | 588 | fsnotify_put_mark(fsn_mark); |
088b09b0 | 589 | |
2a3edf86 EP |
590 | return 0; |
591 | } | |
592 | ||
b9e4e3bd EP |
593 | static __u32 fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark, |
594 | __u32 mask, | |
595 | unsigned int flags) | |
912ee394 | 596 | { |
192ca4d1 | 597 | __u32 oldmask = -1; |
912ee394 AG |
598 | |
599 | spin_lock(&fsn_mark->lock); | |
b9e4e3bd | 600 | if (!(flags & FAN_MARK_IGNORED_MASK)) { |
66ba93c0 LS |
601 | __u32 tmask = fsn_mark->mask | mask; |
602 | ||
603 | if (flags & FAN_MARK_ONDIR) | |
604 | tmask |= FAN_ONDIR; | |
605 | ||
b9e4e3bd | 606 | oldmask = fsn_mark->mask; |
66d2b81b | 607 | fsn_mark->mask = tmask; |
b9e4e3bd | 608 | } else { |
192ca4d1 | 609 | __u32 tmask = fsn_mark->ignored_mask | mask; |
66ba93c0 LS |
610 | if (flags & FAN_MARK_ONDIR) |
611 | tmask |= FAN_ONDIR; | |
612 | ||
66d2b81b | 613 | fsn_mark->ignored_mask = tmask; |
c9778a98 EP |
614 | if (flags & FAN_MARK_IGNORED_SURV_MODIFY) |
615 | fsn_mark->flags |= FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY; | |
b9e4e3bd | 616 | } |
912ee394 AG |
617 | spin_unlock(&fsn_mark->lock); |
618 | ||
619 | return mask & ~oldmask; | |
620 | } | |
621 | ||
5e9c070c LS |
622 | static struct fsnotify_mark *fanotify_add_new_mark(struct fsnotify_group *group, |
623 | struct inode *inode, | |
624 | struct vfsmount *mnt) | |
625 | { | |
626 | struct fsnotify_mark *mark; | |
627 | int ret; | |
628 | ||
629 | if (atomic_read(&group->num_marks) > group->fanotify_data.max_marks) | |
630 | return ERR_PTR(-ENOSPC); | |
631 | ||
632 | mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL); | |
633 | if (!mark) | |
634 | return ERR_PTR(-ENOMEM); | |
635 | ||
054c636e | 636 | fsnotify_init_mark(mark, group); |
7b129323 | 637 | ret = fsnotify_add_mark_locked(mark, inode, mnt, 0); |
5e9c070c LS |
638 | if (ret) { |
639 | fsnotify_put_mark(mark); | |
640 | return ERR_PTR(ret); | |
641 | } | |
642 | ||
643 | return mark; | |
644 | } | |
645 | ||
646 | ||
52202dfb | 647 | static int fanotify_add_vfsmount_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
648 | struct vfsmount *mnt, __u32 mask, |
649 | unsigned int flags) | |
2a3edf86 EP |
650 | { |
651 | struct fsnotify_mark *fsn_mark; | |
912ee394 | 652 | __u32 added; |
2a3edf86 | 653 | |
7b18527c | 654 | mutex_lock(&group->mark_mutex); |
b1362edf JK |
655 | fsn_mark = fsnotify_find_mark(&real_mount(mnt)->mnt_fsnotify_marks, |
656 | group); | |
88826276 | 657 | if (!fsn_mark) { |
5e9c070c LS |
658 | fsn_mark = fanotify_add_new_mark(group, NULL, mnt); |
659 | if (IS_ERR(fsn_mark)) { | |
7b18527c | 660 | mutex_unlock(&group->mark_mutex); |
5e9c070c | 661 | return PTR_ERR(fsn_mark); |
7b18527c | 662 | } |
88826276 | 663 | } |
b9e4e3bd | 664 | added = fanotify_mark_add_to_mask(fsn_mark, mask, flags); |
c63181e6 | 665 | if (added & ~real_mount(mnt)->mnt_fsnotify_mask) |
8920d273 | 666 | fsnotify_recalc_mask(real_mount(mnt)->mnt_fsnotify_marks); |
c9747640 | 667 | mutex_unlock(&group->mark_mutex); |
5e9c070c | 668 | |
fa218ab9 | 669 | fsnotify_put_mark(fsn_mark); |
5e9c070c | 670 | return 0; |
88826276 EP |
671 | } |
672 | ||
52202dfb | 673 | static int fanotify_add_inode_mark(struct fsnotify_group *group, |
b9e4e3bd EP |
674 | struct inode *inode, __u32 mask, |
675 | unsigned int flags) | |
88826276 EP |
676 | { |
677 | struct fsnotify_mark *fsn_mark; | |
912ee394 | 678 | __u32 added; |
88826276 EP |
679 | |
680 | pr_debug("%s: group=%p inode=%p\n", __func__, group, inode); | |
2a3edf86 | 681 | |
5322a59f EP |
682 | /* |
683 | * If some other task has this inode open for write we should not add | |
684 | * an ignored mark, unless that ignored mark is supposed to survive | |
685 | * modification changes anyway. | |
686 | */ | |
687 | if ((flags & FAN_MARK_IGNORED_MASK) && | |
688 | !(flags & FAN_MARK_IGNORED_SURV_MODIFY) && | |
689 | (atomic_read(&inode->i_writecount) > 0)) | |
690 | return 0; | |
691 | ||
7b18527c | 692 | mutex_lock(&group->mark_mutex); |
b1362edf | 693 | fsn_mark = fsnotify_find_mark(&inode->i_fsnotify_marks, group); |
2a3edf86 | 694 | if (!fsn_mark) { |
5e9c070c LS |
695 | fsn_mark = fanotify_add_new_mark(group, inode, NULL); |
696 | if (IS_ERR(fsn_mark)) { | |
7b18527c | 697 | mutex_unlock(&group->mark_mutex); |
5e9c070c | 698 | return PTR_ERR(fsn_mark); |
7b18527c | 699 | } |
2a3edf86 | 700 | } |
b9e4e3bd | 701 | added = fanotify_mark_add_to_mask(fsn_mark, mask, flags); |
43709a28 | 702 | if (added & ~inode->i_fsnotify_mask) |
8920d273 | 703 | fsnotify_recalc_mask(inode->i_fsnotify_marks); |
c9747640 | 704 | mutex_unlock(&group->mark_mutex); |
5e9c070c | 705 | |
fa218ab9 | 706 | fsnotify_put_mark(fsn_mark); |
5e9c070c | 707 | return 0; |
88826276 | 708 | } |
2a3edf86 | 709 | |
52c923dd | 710 | /* fanotify syscalls */ |
08ae8938 | 711 | SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags) |
11637e4b | 712 | { |
52c923dd EP |
713 | struct fsnotify_group *group; |
714 | int f_flags, fd; | |
4afeff85 | 715 | struct user_struct *user; |
ff57cd58 | 716 | struct fanotify_event_info *oevent; |
52c923dd | 717 | |
08ae8938 EP |
718 | pr_debug("%s: flags=%d event_f_flags=%d\n", |
719 | __func__, flags, event_f_flags); | |
52c923dd | 720 | |
52c923dd | 721 | if (!capable(CAP_SYS_ADMIN)) |
a2f13ad0 | 722 | return -EPERM; |
52c923dd EP |
723 | |
724 | if (flags & ~FAN_ALL_INIT_FLAGS) | |
725 | return -EINVAL; | |
726 | ||
48149e9d HS |
727 | if (event_f_flags & ~FANOTIFY_INIT_ALL_EVENT_F_BITS) |
728 | return -EINVAL; | |
729 | ||
730 | switch (event_f_flags & O_ACCMODE) { | |
731 | case O_RDONLY: | |
732 | case O_RDWR: | |
733 | case O_WRONLY: | |
734 | break; | |
735 | default: | |
736 | return -EINVAL; | |
737 | } | |
738 | ||
4afeff85 EP |
739 | user = get_current_user(); |
740 | if (atomic_read(&user->fanotify_listeners) > FANOTIFY_DEFAULT_MAX_LISTENERS) { | |
741 | free_uid(user); | |
742 | return -EMFILE; | |
743 | } | |
744 | ||
b2d87909 | 745 | f_flags = O_RDWR | FMODE_NONOTIFY; |
52c923dd EP |
746 | if (flags & FAN_CLOEXEC) |
747 | f_flags |= O_CLOEXEC; | |
748 | if (flags & FAN_NONBLOCK) | |
749 | f_flags |= O_NONBLOCK; | |
750 | ||
751 | /* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */ | |
752 | group = fsnotify_alloc_group(&fanotify_fsnotify_ops); | |
26379198 EP |
753 | if (IS_ERR(group)) { |
754 | free_uid(user); | |
52c923dd | 755 | return PTR_ERR(group); |
26379198 | 756 | } |
52c923dd | 757 | |
4afeff85 EP |
758 | group->fanotify_data.user = user; |
759 | atomic_inc(&user->fanotify_listeners); | |
760 | ||
f083441b | 761 | oevent = fanotify_alloc_event(NULL, FS_Q_OVERFLOW, NULL); |
ff57cd58 JK |
762 | if (unlikely(!oevent)) { |
763 | fd = -ENOMEM; | |
764 | goto out_destroy_group; | |
765 | } | |
766 | group->overflow_event = &oevent->fse; | |
ff57cd58 | 767 | |
1e2ee49f WW |
768 | if (force_o_largefile()) |
769 | event_f_flags |= O_LARGEFILE; | |
80af2588 | 770 | group->fanotify_data.f_flags = event_f_flags; |
9e66e423 | 771 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
9e66e423 EP |
772 | init_waitqueue_head(&group->fanotify_data.access_waitq); |
773 | INIT_LIST_HEAD(&group->fanotify_data.access_list); | |
774 | #endif | |
4231a235 EP |
775 | switch (flags & FAN_ALL_CLASS_BITS) { |
776 | case FAN_CLASS_NOTIF: | |
777 | group->priority = FS_PRIO_0; | |
778 | break; | |
779 | case FAN_CLASS_CONTENT: | |
780 | group->priority = FS_PRIO_1; | |
781 | break; | |
782 | case FAN_CLASS_PRE_CONTENT: | |
783 | group->priority = FS_PRIO_2; | |
784 | break; | |
785 | default: | |
786 | fd = -EINVAL; | |
d8153d4d | 787 | goto out_destroy_group; |
4231a235 | 788 | } |
cb2d429f | 789 | |
5dd03f55 EP |
790 | if (flags & FAN_UNLIMITED_QUEUE) { |
791 | fd = -EPERM; | |
792 | if (!capable(CAP_SYS_ADMIN)) | |
d8153d4d | 793 | goto out_destroy_group; |
5dd03f55 EP |
794 | group->max_events = UINT_MAX; |
795 | } else { | |
796 | group->max_events = FANOTIFY_DEFAULT_MAX_EVENTS; | |
797 | } | |
2529a0df | 798 | |
ac7e22dc EP |
799 | if (flags & FAN_UNLIMITED_MARKS) { |
800 | fd = -EPERM; | |
801 | if (!capable(CAP_SYS_ADMIN)) | |
d8153d4d | 802 | goto out_destroy_group; |
ac7e22dc EP |
803 | group->fanotify_data.max_marks = UINT_MAX; |
804 | } else { | |
805 | group->fanotify_data.max_marks = FANOTIFY_DEFAULT_MAX_MARKS; | |
806 | } | |
e7099d8a | 807 | |
52c923dd EP |
808 | fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags); |
809 | if (fd < 0) | |
d8153d4d | 810 | goto out_destroy_group; |
52c923dd EP |
811 | |
812 | return fd; | |
813 | ||
d8153d4d LS |
814 | out_destroy_group: |
815 | fsnotify_destroy_group(group); | |
52c923dd | 816 | return fd; |
11637e4b | 817 | } |
bbaa4168 | 818 | |
4a0fd5bf AV |
819 | SYSCALL_DEFINE5(fanotify_mark, int, fanotify_fd, unsigned int, flags, |
820 | __u64, mask, int, dfd, | |
821 | const char __user *, pathname) | |
bbaa4168 | 822 | { |
0ff21db9 EP |
823 | struct inode *inode = NULL; |
824 | struct vfsmount *mnt = NULL; | |
2a3edf86 | 825 | struct fsnotify_group *group; |
2903ff01 | 826 | struct fd f; |
2a3edf86 | 827 | struct path path; |
2903ff01 | 828 | int ret; |
2a3edf86 EP |
829 | |
830 | pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n", | |
831 | __func__, fanotify_fd, flags, dfd, pathname, mask); | |
832 | ||
833 | /* we only use the lower 32 bits as of right now. */ | |
834 | if (mask & ((__u64)0xffffffff << 32)) | |
835 | return -EINVAL; | |
836 | ||
88380fe6 AG |
837 | if (flags & ~FAN_ALL_MARK_FLAGS) |
838 | return -EINVAL; | |
4d92604c | 839 | switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE | FAN_MARK_FLUSH)) { |
1734dee4 | 840 | case FAN_MARK_ADD: /* fallthrough */ |
88380fe6 | 841 | case FAN_MARK_REMOVE: |
1734dee4 LS |
842 | if (!mask) |
843 | return -EINVAL; | |
cc299a98 | 844 | break; |
4d92604c | 845 | case FAN_MARK_FLUSH: |
cc299a98 HS |
846 | if (flags & ~(FAN_MARK_MOUNT | FAN_MARK_FLUSH)) |
847 | return -EINVAL; | |
88380fe6 AG |
848 | break; |
849 | default: | |
850 | return -EINVAL; | |
851 | } | |
8fcd6528 EP |
852 | |
853 | if (mask & FAN_ONDIR) { | |
854 | flags |= FAN_MARK_ONDIR; | |
855 | mask &= ~FAN_ONDIR; | |
856 | } | |
857 | ||
b2d87909 EP |
858 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
859 | if (mask & ~(FAN_ALL_EVENTS | FAN_ALL_PERM_EVENTS | FAN_EVENT_ON_CHILD)) | |
860 | #else | |
88380fe6 | 861 | if (mask & ~(FAN_ALL_EVENTS | FAN_EVENT_ON_CHILD)) |
b2d87909 | 862 | #endif |
2a3edf86 EP |
863 | return -EINVAL; |
864 | ||
2903ff01 AV |
865 | f = fdget(fanotify_fd); |
866 | if (unlikely(!f.file)) | |
2a3edf86 EP |
867 | return -EBADF; |
868 | ||
869 | /* verify that this is indeed an fanotify instance */ | |
870 | ret = -EINVAL; | |
2903ff01 | 871 | if (unlikely(f.file->f_op != &fanotify_fops)) |
2a3edf86 | 872 | goto fput_and_out; |
2903ff01 | 873 | group = f.file->private_data; |
4231a235 EP |
874 | |
875 | /* | |
876 | * group->priority == FS_PRIO_0 == FAN_CLASS_NOTIF. These are not | |
877 | * allowed to set permissions events. | |
878 | */ | |
879 | ret = -EINVAL; | |
880 | if (mask & FAN_ALL_PERM_EVENTS && | |
881 | group->priority == FS_PRIO_0) | |
882 | goto fput_and_out; | |
2a3edf86 | 883 | |
0a8dd2db HS |
884 | if (flags & FAN_MARK_FLUSH) { |
885 | ret = 0; | |
886 | if (flags & FAN_MARK_MOUNT) | |
887 | fsnotify_clear_vfsmount_marks_by_group(group); | |
888 | else | |
889 | fsnotify_clear_inode_marks_by_group(group); | |
890 | goto fput_and_out; | |
891 | } | |
892 | ||
2a3edf86 EP |
893 | ret = fanotify_find_path(dfd, pathname, &path, flags); |
894 | if (ret) | |
895 | goto fput_and_out; | |
896 | ||
897 | /* inode held in place by reference to path; group by fget on fd */ | |
eac8e9e8 | 898 | if (!(flags & FAN_MARK_MOUNT)) |
0ff21db9 EP |
899 | inode = path.dentry->d_inode; |
900 | else | |
901 | mnt = path.mnt; | |
2a3edf86 EP |
902 | |
903 | /* create/update an inode mark */ | |
0a8dd2db | 904 | switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE)) { |
c6223f46 | 905 | case FAN_MARK_ADD: |
eac8e9e8 | 906 | if (flags & FAN_MARK_MOUNT) |
b9e4e3bd | 907 | ret = fanotify_add_vfsmount_mark(group, mnt, mask, flags); |
0ff21db9 | 908 | else |
b9e4e3bd | 909 | ret = fanotify_add_inode_mark(group, inode, mask, flags); |
c6223f46 AG |
910 | break; |
911 | case FAN_MARK_REMOVE: | |
f3640192 | 912 | if (flags & FAN_MARK_MOUNT) |
b9e4e3bd | 913 | ret = fanotify_remove_vfsmount_mark(group, mnt, mask, flags); |
f3640192 | 914 | else |
b9e4e3bd | 915 | ret = fanotify_remove_inode_mark(group, inode, mask, flags); |
c6223f46 AG |
916 | break; |
917 | default: | |
918 | ret = -EINVAL; | |
919 | } | |
2a3edf86 EP |
920 | |
921 | path_put(&path); | |
922 | fput_and_out: | |
2903ff01 | 923 | fdput(f); |
2a3edf86 EP |
924 | return ret; |
925 | } | |
926 | ||
91c2e0bc AV |
927 | #ifdef CONFIG_COMPAT |
928 | COMPAT_SYSCALL_DEFINE6(fanotify_mark, | |
929 | int, fanotify_fd, unsigned int, flags, | |
930 | __u32, mask0, __u32, mask1, int, dfd, | |
931 | const char __user *, pathname) | |
932 | { | |
933 | return sys_fanotify_mark(fanotify_fd, flags, | |
934 | #ifdef __BIG_ENDIAN | |
91c2e0bc | 935 | ((__u64)mask0 << 32) | mask1, |
592f6b84 HC |
936 | #else |
937 | ((__u64)mask1 << 32) | mask0, | |
91c2e0bc AV |
938 | #endif |
939 | dfd, pathname); | |
940 | } | |
941 | #endif | |
942 | ||
2a3edf86 | 943 | /* |
ae0e47f0 | 944 | * fanotify_user_setup - Our initialization function. Note that we cannot return |
2a3edf86 EP |
945 | * error because we have compiled-in VFS hooks. So an (unlikely) failure here |
946 | * must result in panic(). | |
947 | */ | |
948 | static int __init fanotify_user_setup(void) | |
949 | { | |
950 | fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC); | |
7053aee2 | 951 | fanotify_event_cachep = KMEM_CACHE(fanotify_event_info, SLAB_PANIC); |
f083441b JK |
952 | #ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS |
953 | fanotify_perm_event_cachep = KMEM_CACHE(fanotify_perm_event_info, | |
954 | SLAB_PANIC); | |
955 | #endif | |
2a3edf86 EP |
956 | |
957 | return 0; | |
bbaa4168 | 958 | } |
2a3edf86 | 959 | device_initcall(fanotify_user_setup); |