aboutsummaryrefslogtreecommitdiffstats
path: root/fs/notify/fanotify
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2010-08-10 14:39:13 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2010-08-10 14:39:13 -0400
commit8c8946f509a494769a8c602b5ed189df01917d39 (patch)
treedfd96bd6ca5ea6803c6d77f65ba37e04f78b2d3b /fs/notify/fanotify
parent5f248c9c251c60af3403902b26e08de43964ea0b (diff)
parent1968f5eed54ce47bde488fd9a450912e4a2d7138 (diff)
Merge branch 'for-linus' of git://git.infradead.org/users/eparis/notify
* 'for-linus' of git://git.infradead.org/users/eparis/notify: (132 commits) fanotify: use both marks when possible fsnotify: pass both the vfsmount mark and inode mark fsnotify: walk the inode and vfsmount lists simultaneously fsnotify: rework ignored mark flushing fsnotify: remove global fsnotify groups lists fsnotify: remove group->mask fsnotify: remove the global masks fsnotify: cleanup should_send_event fanotify: use the mark in handler functions audit: use the mark in handler functions dnotify: use the mark in handler functions inotify: use the mark in handler functions fsnotify: send fsnotify_mark to groups in event handling functions fsnotify: Exchange list heads instead of moving elements fsnotify: srcu to protect read side of inode and vfsmount locks fsnotify: use an explicit flag to indicate fsnotify_destroy_mark has been called fsnotify: use _rcu functions for mark list traversal fsnotify: place marks on object in order of group memory address vfs/fsnotify: fsnotify_close can delay the final work in fput fsnotify: store struct file not struct path ... Fix up trivial delete/modify conflict in fs/notify/inotify/inotify.c.
Diffstat (limited to 'fs/notify/fanotify')
-rw-r--r--fs/notify/fanotify/Kconfig26
-rw-r--r--fs/notify/fanotify/Makefile1
-rw-r--r--fs/notify/fanotify/fanotify.c212
-rw-r--r--fs/notify/fanotify/fanotify_user.c760
4 files changed, 999 insertions, 0 deletions
diff --git a/fs/notify/fanotify/Kconfig b/fs/notify/fanotify/Kconfig
new file mode 100644
index 000000000000..3ac36b7bf6b9
--- /dev/null
+++ b/fs/notify/fanotify/Kconfig
@@ -0,0 +1,26 @@
1config FANOTIFY
2 bool "Filesystem wide access notification"
3 select FSNOTIFY
4 select ANON_INODES
5 default n
6 ---help---
7 Say Y here to enable fanotify suport. fanotify is a file access
8 notification system which differs from inotify in that it sends
9 and open file descriptor to the userspace listener along with
10 the event.
11
12 If unsure, say Y.
13
14config FANOTIFY_ACCESS_PERMISSIONS
15 bool "fanotify permissions checking"
16 depends on FANOTIFY
17 depends on SECURITY
18 default n
19 ---help---
20 Say Y here is you want fanotify listeners to be able to make permissions
21 decisions concerning filesystem events. This is used by some fanotify
22 listeners which need to scan files before allowing the system access to
23 use those files. This is used by some anti-malware vendors and by some
24 hierarchical storage managent systems.
25
26 If unsure, say N.
diff --git a/fs/notify/fanotify/Makefile b/fs/notify/fanotify/Makefile
new file mode 100644
index 000000000000..0999213e7e6e
--- /dev/null
+++ b/fs/notify/fanotify/Makefile
@@ -0,0 +1 @@
obj-$(CONFIG_FANOTIFY) += fanotify.o fanotify_user.o
diff --git a/fs/notify/fanotify/fanotify.c b/fs/notify/fanotify/fanotify.c
new file mode 100644
index 000000000000..eb8f73c9c131
--- /dev/null
+++ b/fs/notify/fanotify/fanotify.c
@@ -0,0 +1,212 @@
1#include <linux/fanotify.h>
2#include <linux/fdtable.h>
3#include <linux/fsnotify_backend.h>
4#include <linux/init.h>
5#include <linux/jiffies.h>
6#include <linux/kernel.h> /* UINT_MAX */
7#include <linux/mount.h>
8#include <linux/sched.h>
9#include <linux/types.h>
10#include <linux/wait.h>
11
12static bool should_merge(struct fsnotify_event *old, struct fsnotify_event *new)
13{
14 pr_debug("%s: old=%p new=%p\n", __func__, old, new);
15
16 if (old->to_tell == new->to_tell &&
17 old->data_type == new->data_type &&
18 old->tgid == new->tgid) {
19 switch (old->data_type) {
20 case (FSNOTIFY_EVENT_FILE):
21 if ((old->file->f_path.mnt == new->file->f_path.mnt) &&
22 (old->file->f_path.dentry == new->file->f_path.dentry))
23 return true;
24 case (FSNOTIFY_EVENT_NONE):
25 return true;
26 default:
27 BUG();
28 };
29 }
30 return false;
31}
32
33/* and the list better be locked by something too! */
34static struct fsnotify_event *fanotify_merge(struct list_head *list,
35 struct fsnotify_event *event)
36{
37 struct fsnotify_event_holder *test_holder;
38 struct fsnotify_event *test_event = NULL;
39 struct fsnotify_event *new_event;
40
41 pr_debug("%s: list=%p event=%p\n", __func__, list, event);
42
43
44 list_for_each_entry_reverse(test_holder, list, event_list) {
45 if (should_merge(test_holder->event, event)) {
46 test_event = test_holder->event;
47 break;
48 }
49 }
50
51 if (!test_event)
52 return NULL;
53
54 fsnotify_get_event(test_event);
55
56 /* if they are exactly the same we are done */
57 if (test_event->mask == event->mask)
58 return test_event;
59
60 /*
61 * if the refcnt == 2 this is the only queue
62 * for this event and so we can update the mask
63 * in place.
64 */
65 if (atomic_read(&test_event->refcnt) == 2) {
66 test_event->mask |= event->mask;
67 return test_event;
68 }
69
70 new_event = fsnotify_clone_event(test_event);
71
72 /* done with test_event */
73 fsnotify_put_event(test_event);
74
75 /* couldn't allocate memory, merge was not possible */
76 if (unlikely(!new_event))
77 return ERR_PTR(-ENOMEM);
78
79 /* build new event and replace it on the list */
80 new_event->mask = (test_event->mask | event->mask);
81 fsnotify_replace_event(test_holder, new_event);
82
83 /* we hold a reference on new_event from clone_event */
84 return new_event;
85}
86
87#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
88static int fanotify_get_response_from_access(struct fsnotify_group *group,
89 struct fsnotify_event *event)
90{
91 int ret;
92
93 pr_debug("%s: group=%p event=%p\n", __func__, group, event);
94
95 wait_event(group->fanotify_data.access_waitq, event->response);
96
97 /* userspace responded, convert to something usable */
98 spin_lock(&event->lock);
99 switch (event->response) {
100 case FAN_ALLOW:
101 ret = 0;
102 break;
103 case FAN_DENY:
104 default:
105 ret = -EPERM;
106 }
107 event->response = 0;
108 spin_unlock(&event->lock);
109
110 pr_debug("%s: group=%p event=%p about to return ret=%d\n", __func__,
111 group, event, ret);
112
113 return ret;
114}
115#endif
116
117static int fanotify_handle_event(struct fsnotify_group *group,
118 struct fsnotify_mark *inode_mark,
119 struct fsnotify_mark *fanotify_mark,
120 struct fsnotify_event *event)
121{
122 int ret = 0;
123 struct fsnotify_event *notify_event = NULL;
124
125 BUILD_BUG_ON(FAN_ACCESS != FS_ACCESS);
126 BUILD_BUG_ON(FAN_MODIFY != FS_MODIFY);
127 BUILD_BUG_ON(FAN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
128 BUILD_BUG_ON(FAN_CLOSE_WRITE != FS_CLOSE_WRITE);
129 BUILD_BUG_ON(FAN_OPEN != FS_OPEN);
130 BUILD_BUG_ON(FAN_EVENT_ON_CHILD != FS_EVENT_ON_CHILD);
131 BUILD_BUG_ON(FAN_Q_OVERFLOW != FS_Q_OVERFLOW);
132 BUILD_BUG_ON(FAN_OPEN_PERM != FS_OPEN_PERM);
133 BUILD_BUG_ON(FAN_ACCESS_PERM != FS_ACCESS_PERM);
134
135 pr_debug("%s: group=%p event=%p\n", __func__, group, event);
136
137 notify_event = fsnotify_add_notify_event(group, event, NULL, fanotify_merge);
138 if (IS_ERR(notify_event))
139 return PTR_ERR(notify_event);
140
141#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
142 if (event->mask & FAN_ALL_PERM_EVENTS) {
143 /* if we merged we need to wait on the new event */
144 if (notify_event)
145 event = notify_event;
146 ret = fanotify_get_response_from_access(group, event);
147 }
148#endif
149
150 if (notify_event)
151 fsnotify_put_event(notify_event);
152
153 return ret;
154}
155
156static bool fanotify_should_send_event(struct fsnotify_group *group,
157 struct inode *to_tell,
158 struct fsnotify_mark *inode_mark,
159 struct fsnotify_mark *vfsmnt_mark,
160 __u32 event_mask, void *data, int data_type)
161{
162 __u32 marks_mask, marks_ignored_mask;
163
164 pr_debug("%s: group=%p to_tell=%p inode_mark=%p vfsmnt_mark=%p "
165 "mask=%x data=%p data_type=%d\n", __func__, group, to_tell,
166 inode_mark, vfsmnt_mark, event_mask, data, data_type);
167
168 pr_debug("%s: group=%p vfsmount_mark=%p inode_mark=%p mask=%x\n",
169 __func__, group, vfsmnt_mark, inode_mark, event_mask);
170
171 /* sorry, fanotify only gives a damn about files and dirs */
172 if (!S_ISREG(to_tell->i_mode) &&
173 !S_ISDIR(to_tell->i_mode))
174 return false;
175
176 /* if we don't have enough info to send an event to userspace say no */
177 if (data_type != FSNOTIFY_EVENT_FILE)
178 return false;
179
180 if (inode_mark && vfsmnt_mark) {
181 marks_mask = (vfsmnt_mark->mask | inode_mark->mask);
182 marks_ignored_mask = (vfsmnt_mark->ignored_mask | inode_mark->ignored_mask);
183 } else if (inode_mark) {
184 /*
185 * if the event is for a child and this inode doesn't care about
186 * events on the child, don't send it!
187 */
188 if ((event_mask & FS_EVENT_ON_CHILD) &&
189 !(inode_mark->mask & FS_EVENT_ON_CHILD))
190 return false;
191 marks_mask = inode_mark->mask;
192 marks_ignored_mask = inode_mark->ignored_mask;
193 } else if (vfsmnt_mark) {
194 marks_mask = vfsmnt_mark->mask;
195 marks_ignored_mask = vfsmnt_mark->ignored_mask;
196 } else {
197 BUG();
198 }
199
200 if (event_mask & marks_mask & ~marks_ignored_mask)
201 return true;
202
203 return false;
204}
205
206const struct fsnotify_ops fanotify_fsnotify_ops = {
207 .handle_event = fanotify_handle_event,
208 .should_send_event = fanotify_should_send_event,
209 .free_group_priv = NULL,
210 .free_event_priv = NULL,
211 .freeing_mark = NULL,
212};
diff --git a/fs/notify/fanotify/fanotify_user.c b/fs/notify/fanotify/fanotify_user.c
new file mode 100644
index 000000000000..25a3b4dfcf61
--- /dev/null
+++ b/fs/notify/fanotify/fanotify_user.c
@@ -0,0 +1,760 @@
1#include <linux/fanotify.h>
2#include <linux/fcntl.h>
3#include <linux/file.h>
4#include <linux/fs.h>
5#include <linux/anon_inodes.h>
6#include <linux/fsnotify_backend.h>
7#include <linux/init.h>
8#include <linux/mount.h>
9#include <linux/namei.h>
10#include <linux/poll.h>
11#include <linux/security.h>
12#include <linux/syscalls.h>
13#include <linux/slab.h>
14#include <linux/types.h>
15#include <linux/uaccess.h>
16
17#include <asm/ioctls.h>
18
19extern const struct fsnotify_ops fanotify_fsnotify_ops;
20
21static struct kmem_cache *fanotify_mark_cache __read_mostly;
22static struct kmem_cache *fanotify_response_event_cache __read_mostly;
23
24struct fanotify_response_event {
25 struct list_head list;
26 __s32 fd;
27 struct fsnotify_event *event;
28};
29
30/*
31 * Get an fsnotify notification event if one exists and is small
32 * enough to fit in "count". Return an error pointer if the count
33 * is not large enough.
34 *
35 * Called with the group->notification_mutex held.
36 */
37static struct fsnotify_event *get_one_event(struct fsnotify_group *group,
38 size_t count)
39{
40 BUG_ON(!mutex_is_locked(&group->notification_mutex));
41
42 pr_debug("%s: group=%p count=%zd\n", __func__, group, count);
43
44 if (fsnotify_notify_queue_is_empty(group))
45 return NULL;
46
47 if (FAN_EVENT_METADATA_LEN > count)
48 return ERR_PTR(-EINVAL);
49
50 /* held the notification_mutex the whole time, so this is the
51 * same event we peeked above */
52 return fsnotify_remove_notify_event(group);
53}
54
55static int create_fd(struct fsnotify_group *group, struct fsnotify_event *event)
56{
57 int client_fd;
58 struct dentry *dentry;
59 struct vfsmount *mnt;
60 struct file *new_file;
61
62 pr_debug("%s: group=%p event=%p\n", __func__, group, event);
63
64 client_fd = get_unused_fd();
65 if (client_fd < 0)
66 return client_fd;
67
68 if (event->data_type != FSNOTIFY_EVENT_FILE) {
69 WARN_ON(1);
70 put_unused_fd(client_fd);
71 return -EINVAL;
72 }
73
74 /*
75 * we need a new file handle for the userspace program so it can read even if it was
76 * originally opened O_WRONLY.
77 */
78 dentry = dget(event->file->f_path.dentry);
79 mnt = mntget(event->file->f_path.mnt);
80 /* it's possible this event was an overflow event. in that case dentry and mnt
81 * are NULL; That's fine, just don't call dentry open */
82 if (dentry && mnt)
83 new_file = dentry_open(dentry, mnt,
84 group->fanotify_data.f_flags | FMODE_NONOTIFY,
85 current_cred());
86 else
87 new_file = ERR_PTR(-EOVERFLOW);
88 if (IS_ERR(new_file)) {
89 /*
90 * we still send an event even if we can't open the file. this
91 * can happen when say tasks are gone and we try to open their
92 * /proc files or we try to open a WRONLY file like in sysfs
93 * we just send the errno to userspace since there isn't much
94 * else we can do.
95 */
96 put_unused_fd(client_fd);
97 client_fd = PTR_ERR(new_file);
98 } else {
99 fd_install(client_fd, new_file);
100 }
101
102 return client_fd;
103}
104
105static ssize_t fill_event_metadata(struct fsnotify_group *group,
106 struct fanotify_event_metadata *metadata,
107 struct fsnotify_event *event)
108{
109 pr_debug("%s: group=%p metadata=%p event=%p\n", __func__,
110 group, metadata, event);
111
112 metadata->event_len = FAN_EVENT_METADATA_LEN;
113 metadata->vers = FANOTIFY_METADATA_VERSION;
114 metadata->mask = event->mask & FAN_ALL_OUTGOING_EVENTS;
115 metadata->pid = pid_vnr(event->tgid);
116 metadata->fd = create_fd(group, event);
117
118 return metadata->fd;
119}
120
121#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
122static struct fanotify_response_event *dequeue_re(struct fsnotify_group *group,
123 __s32 fd)
124{
125 struct fanotify_response_event *re, *return_re = NULL;
126
127 mutex_lock(&group->fanotify_data.access_mutex);
128 list_for_each_entry(re, &group->fanotify_data.access_list, list) {
129 if (re->fd != fd)
130 continue;
131
132 list_del_init(&re->list);
133 return_re = re;
134 break;
135 }
136 mutex_unlock(&group->fanotify_data.access_mutex);
137
138 pr_debug("%s: found return_re=%p\n", __func__, return_re);
139
140 return return_re;
141}
142
143static int process_access_response(struct fsnotify_group *group,
144 struct fanotify_response *response_struct)
145{
146 struct fanotify_response_event *re;
147 __s32 fd = response_struct->fd;
148 __u32 response = response_struct->response;
149
150 pr_debug("%s: group=%p fd=%d response=%d\n", __func__, group,
151 fd, response);
152 /*
153 * make sure the response is valid, if invalid we do nothing and either
154 * userspace can send a valid responce or we will clean it up after the
155 * timeout
156 */
157 switch (response) {
158 case FAN_ALLOW:
159 case FAN_DENY:
160 break;
161 default:
162 return -EINVAL;
163 }
164
165 if (fd < 0)
166 return -EINVAL;
167
168 re = dequeue_re(group, fd);
169 if (!re)
170 return -ENOENT;
171
172 re->event->response = response;
173
174 wake_up(&group->fanotify_data.access_waitq);
175
176 kmem_cache_free(fanotify_response_event_cache, re);
177
178 return 0;
179}
180
181static int prepare_for_access_response(struct fsnotify_group *group,
182 struct fsnotify_event *event,
183 __s32 fd)
184{
185 struct fanotify_response_event *re;
186
187 if (!(event->mask & FAN_ALL_PERM_EVENTS))
188 return 0;
189
190 re = kmem_cache_alloc(fanotify_response_event_cache, GFP_KERNEL);
191 if (!re)
192 return -ENOMEM;
193
194 re->event = event;
195 re->fd = fd;
196
197 mutex_lock(&group->fanotify_data.access_mutex);
198 list_add_tail(&re->list, &group->fanotify_data.access_list);
199 mutex_unlock(&group->fanotify_data.access_mutex);
200
201 return 0;
202}
203
204static void remove_access_response(struct fsnotify_group *group,
205 struct fsnotify_event *event,
206 __s32 fd)
207{
208 struct fanotify_response_event *re;
209
210 if (!(event->mask & FAN_ALL_PERM_EVENTS))
211 return;
212
213 re = dequeue_re(group, fd);
214 if (!re)
215 return;
216
217 BUG_ON(re->event != event);
218
219 kmem_cache_free(fanotify_response_event_cache, re);
220
221 return;
222}
223#else
224static int prepare_for_access_response(struct fsnotify_group *group,
225 struct fsnotify_event *event,
226 __s32 fd)
227{
228 return 0;
229}
230
231static void remove_access_response(struct fsnotify_group *group,
232 struct fsnotify_event *event,
233 __s32 fd)
234{
235 return;
236}
237#endif
238
239static ssize_t copy_event_to_user(struct fsnotify_group *group,
240 struct fsnotify_event *event,
241 char __user *buf)
242{
243 struct fanotify_event_metadata fanotify_event_metadata;
244 int fd, ret;
245
246 pr_debug("%s: group=%p event=%p\n", __func__, group, event);
247
248 fd = fill_event_metadata(group, &fanotify_event_metadata, event);
249 if (fd < 0)
250 return fd;
251
252 ret = prepare_for_access_response(group, event, fd);
253 if (ret)
254 goto out_close_fd;
255
256 ret = -EFAULT;
257 if (copy_to_user(buf, &fanotify_event_metadata, FAN_EVENT_METADATA_LEN))
258 goto out_kill_access_response;
259
260 return FAN_EVENT_METADATA_LEN;
261
262out_kill_access_response:
263 remove_access_response(group, event, fd);
264out_close_fd:
265 sys_close(fd);
266 return ret;
267}
268
269/* intofiy userspace file descriptor functions */
270static unsigned int fanotify_poll(struct file *file, poll_table *wait)
271{
272 struct fsnotify_group *group = file->private_data;
273 int ret = 0;
274
275 poll_wait(file, &group->notification_waitq, wait);
276 mutex_lock(&group->notification_mutex);
277 if (!fsnotify_notify_queue_is_empty(group))
278 ret = POLLIN | POLLRDNORM;
279 mutex_unlock(&group->notification_mutex);
280
281 return ret;
282}
283
284static ssize_t fanotify_read(struct file *file, char __user *buf,
285 size_t count, loff_t *pos)
286{
287 struct fsnotify_group *group;
288 struct fsnotify_event *kevent;
289 char __user *start;
290 int ret;
291 DEFINE_WAIT(wait);
292
293 start = buf;
294 group = file->private_data;
295
296 pr_debug("%s: group=%p\n", __func__, group);
297
298 while (1) {
299 prepare_to_wait(&group->notification_waitq, &wait, TASK_INTERRUPTIBLE);
300
301 mutex_lock(&group->notification_mutex);
302 kevent = get_one_event(group, count);
303 mutex_unlock(&group->notification_mutex);
304
305 if (kevent) {
306 ret = PTR_ERR(kevent);
307 if (IS_ERR(kevent))
308 break;
309 ret = copy_event_to_user(group, kevent, buf);
310 fsnotify_put_event(kevent);
311 if (ret < 0)
312 break;
313 buf += ret;
314 count -= ret;
315 continue;
316 }
317
318 ret = -EAGAIN;
319 if (file->f_flags & O_NONBLOCK)
320 break;
321 ret = -EINTR;
322 if (signal_pending(current))
323 break;
324
325 if (start != buf)
326 break;
327
328 schedule();
329 }
330
331 finish_wait(&group->notification_waitq, &wait);
332 if (start != buf && ret != -EFAULT)
333 ret = buf - start;
334 return ret;
335}
336
337static ssize_t fanotify_write(struct file *file, const char __user *buf, size_t count, loff_t *pos)
338{
339#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
340 struct fanotify_response response = { .fd = -1, .response = -1 };
341 struct fsnotify_group *group;
342 int ret;
343
344 group = file->private_data;
345
346 if (count > sizeof(response))
347 count = sizeof(response);
348
349 pr_debug("%s: group=%p count=%zu\n", __func__, group, count);
350
351 if (copy_from_user(&response, buf, count))
352 return -EFAULT;
353
354 ret = process_access_response(group, &response);
355 if (ret < 0)
356 count = ret;
357
358 return count;
359#else
360 return -EINVAL;
361#endif
362}
363
364static int fanotify_release(struct inode *ignored, struct file *file)
365{
366 struct fsnotify_group *group = file->private_data;
367
368 pr_debug("%s: file=%p group=%p\n", __func__, file, group);
369
370 /* matches the fanotify_init->fsnotify_alloc_group */
371 fsnotify_put_group(group);
372
373 return 0;
374}
375
376static long fanotify_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
377{
378 struct fsnotify_group *group;
379 struct fsnotify_event_holder *holder;
380 void __user *p;
381 int ret = -ENOTTY;
382 size_t send_len = 0;
383
384 group = file->private_data;
385
386 p = (void __user *) arg;
387
388 switch (cmd) {
389 case FIONREAD:
390 mutex_lock(&group->notification_mutex);
391 list_for_each_entry(holder, &group->notification_list, event_list)
392 send_len += FAN_EVENT_METADATA_LEN;
393 mutex_unlock(&group->notification_mutex);
394 ret = put_user(send_len, (int __user *) p);
395 break;
396 }
397
398 return ret;
399}
400
401static const struct file_operations fanotify_fops = {
402 .poll = fanotify_poll,
403 .read = fanotify_read,
404 .write = fanotify_write,
405 .fasync = NULL,
406 .release = fanotify_release,
407 .unlocked_ioctl = fanotify_ioctl,
408 .compat_ioctl = fanotify_ioctl,
409};
410
411static void fanotify_free_mark(struct fsnotify_mark *fsn_mark)
412{
413 kmem_cache_free(fanotify_mark_cache, fsn_mark);
414}
415
416static int fanotify_find_path(int dfd, const char __user *filename,
417 struct path *path, unsigned int flags)
418{
419 int ret;
420
421 pr_debug("%s: dfd=%d filename=%p flags=%x\n", __func__,
422 dfd, filename, flags);
423
424 if (filename == NULL) {
425 struct file *file;
426 int fput_needed;
427
428 ret = -EBADF;
429 file = fget_light(dfd, &fput_needed);
430 if (!file)
431 goto out;
432
433 ret = -ENOTDIR;
434 if ((flags & FAN_MARK_ONLYDIR) &&
435 !(S_ISDIR(file->f_path.dentry->d_inode->i_mode))) {
436 fput_light(file, fput_needed);
437 goto out;
438 }
439
440 *path = file->f_path;
441 path_get(path);
442 fput_light(file, fput_needed);
443 } else {
444 unsigned int lookup_flags = 0;
445
446 if (!(flags & FAN_MARK_DONT_FOLLOW))
447 lookup_flags |= LOOKUP_FOLLOW;
448 if (flags & FAN_MARK_ONLYDIR)
449 lookup_flags |= LOOKUP_DIRECTORY;
450
451 ret = user_path_at(dfd, filename, lookup_flags, path);
452 if (ret)
453 goto out;
454 }
455
456 /* you can only watch an inode if you have read permissions on it */
457 ret = inode_permission(path->dentry->d_inode, MAY_READ);
458 if (ret)
459 path_put(path);
460out:
461 return ret;
462}
463
464static __u32 fanotify_mark_remove_from_mask(struct fsnotify_mark *fsn_mark,
465 __u32 mask,
466 unsigned int flags)
467{
468 __u32 oldmask;
469
470 spin_lock(&fsn_mark->lock);
471 if (!(flags & FAN_MARK_IGNORED_MASK)) {
472 oldmask = fsn_mark->mask;
473 fsnotify_set_mark_mask_locked(fsn_mark, (oldmask & ~mask));
474 } else {
475 oldmask = fsn_mark->ignored_mask;
476 fsnotify_set_mark_ignored_mask_locked(fsn_mark, (oldmask & ~mask));
477 }
478 spin_unlock(&fsn_mark->lock);
479
480 if (!(oldmask & ~mask))
481 fsnotify_destroy_mark(fsn_mark);
482
483 return mask & oldmask;
484}
485
486static int fanotify_remove_vfsmount_mark(struct fsnotify_group *group,
487 struct vfsmount *mnt, __u32 mask,
488 unsigned int flags)
489{
490 struct fsnotify_mark *fsn_mark = NULL;
491 __u32 removed;
492
493 fsn_mark = fsnotify_find_vfsmount_mark(group, mnt);
494 if (!fsn_mark)
495 return -ENOENT;
496
497 removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags);
498 fsnotify_put_mark(fsn_mark);
499 if (removed & mnt->mnt_fsnotify_mask)
500 fsnotify_recalc_vfsmount_mask(mnt);
501
502 return 0;
503}
504
505static int fanotify_remove_inode_mark(struct fsnotify_group *group,
506 struct inode *inode, __u32 mask,
507 unsigned int flags)
508{
509 struct fsnotify_mark *fsn_mark = NULL;
510 __u32 removed;
511
512 fsn_mark = fsnotify_find_inode_mark(group, inode);
513 if (!fsn_mark)
514 return -ENOENT;
515
516 removed = fanotify_mark_remove_from_mask(fsn_mark, mask, flags);
517 /* matches the fsnotify_find_inode_mark() */
518 fsnotify_put_mark(fsn_mark);
519 if (removed & inode->i_fsnotify_mask)
520 fsnotify_recalc_inode_mask(inode);
521
522 return 0;
523}
524
525static __u32 fanotify_mark_add_to_mask(struct fsnotify_mark *fsn_mark,
526 __u32 mask,
527 unsigned int flags)
528{
529 __u32 oldmask;
530
531 spin_lock(&fsn_mark->lock);
532 if (!(flags & FAN_MARK_IGNORED_MASK)) {
533 oldmask = fsn_mark->mask;
534 fsnotify_set_mark_mask_locked(fsn_mark, (oldmask | mask));
535 } else {
536 oldmask = fsn_mark->ignored_mask;
537 fsnotify_set_mark_ignored_mask_locked(fsn_mark, (oldmask | mask));
538 if (flags & FAN_MARK_IGNORED_SURV_MODIFY)
539 fsn_mark->flags |= FSNOTIFY_MARK_FLAG_IGNORED_SURV_MODIFY;
540 }
541 spin_unlock(&fsn_mark->lock);
542
543 return mask & ~oldmask;
544}
545
546static int fanotify_add_vfsmount_mark(struct fsnotify_group *group,
547 struct vfsmount *mnt, __u32 mask,
548 unsigned int flags)
549{
550 struct fsnotify_mark *fsn_mark;
551 __u32 added;
552
553 fsn_mark = fsnotify_find_vfsmount_mark(group, mnt);
554 if (!fsn_mark) {
555 int ret;
556
557 fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
558 if (!fsn_mark)
559 return -ENOMEM;
560
561 fsnotify_init_mark(fsn_mark, fanotify_free_mark);
562 ret = fsnotify_add_mark(fsn_mark, group, NULL, mnt, 0);
563 if (ret) {
564 fanotify_free_mark(fsn_mark);
565 return ret;
566 }
567 }
568 added = fanotify_mark_add_to_mask(fsn_mark, mask, flags);
569 fsnotify_put_mark(fsn_mark);
570 if (added & ~mnt->mnt_fsnotify_mask)
571 fsnotify_recalc_vfsmount_mask(mnt);
572
573 return 0;
574}
575
576static int fanotify_add_inode_mark(struct fsnotify_group *group,
577 struct inode *inode, __u32 mask,
578 unsigned int flags)
579{
580 struct fsnotify_mark *fsn_mark;
581 __u32 added;
582
583 pr_debug("%s: group=%p inode=%p\n", __func__, group, inode);
584
585 fsn_mark = fsnotify_find_inode_mark(group, inode);
586 if (!fsn_mark) {
587 int ret;
588
589 fsn_mark = kmem_cache_alloc(fanotify_mark_cache, GFP_KERNEL);
590 if (!fsn_mark)
591 return -ENOMEM;
592
593 fsnotify_init_mark(fsn_mark, fanotify_free_mark);
594 ret = fsnotify_add_mark(fsn_mark, group, inode, NULL, 0);
595 if (ret) {
596 fanotify_free_mark(fsn_mark);
597 return ret;
598 }
599 }
600 added = fanotify_mark_add_to_mask(fsn_mark, mask, flags);
601 fsnotify_put_mark(fsn_mark);
602 if (added & ~inode->i_fsnotify_mask)
603 fsnotify_recalc_inode_mask(inode);
604 return 0;
605}
606
607/* fanotify syscalls */
608SYSCALL_DEFINE2(fanotify_init, unsigned int, flags, unsigned int, event_f_flags)
609{
610 struct fsnotify_group *group;
611 int f_flags, fd;
612
613 pr_debug("%s: flags=%d event_f_flags=%d\n",
614 __func__, flags, event_f_flags);
615
616 if (!capable(CAP_SYS_ADMIN))
617 return -EACCES;
618
619 if (flags & ~FAN_ALL_INIT_FLAGS)
620 return -EINVAL;
621
622 f_flags = O_RDWR | FMODE_NONOTIFY;
623 if (flags & FAN_CLOEXEC)
624 f_flags |= O_CLOEXEC;
625 if (flags & FAN_NONBLOCK)
626 f_flags |= O_NONBLOCK;
627
628 /* fsnotify_alloc_group takes a ref. Dropped in fanotify_release */
629 group = fsnotify_alloc_group(&fanotify_fsnotify_ops);
630 if (IS_ERR(group))
631 return PTR_ERR(group);
632
633 group->fanotify_data.f_flags = event_f_flags;
634#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
635 mutex_init(&group->fanotify_data.access_mutex);
636 init_waitqueue_head(&group->fanotify_data.access_waitq);
637 INIT_LIST_HEAD(&group->fanotify_data.access_list);
638#endif
639
640 fd = anon_inode_getfd("[fanotify]", &fanotify_fops, group, f_flags);
641 if (fd < 0)
642 goto out_put_group;
643
644 return fd;
645
646out_put_group:
647 fsnotify_put_group(group);
648 return fd;
649}
650
651SYSCALL_DEFINE(fanotify_mark)(int fanotify_fd, unsigned int flags,
652 __u64 mask, int dfd,
653 const char __user * pathname)
654{
655 struct inode *inode = NULL;
656 struct vfsmount *mnt = NULL;
657 struct fsnotify_group *group;
658 struct file *filp;
659 struct path path;
660 int ret, fput_needed;
661
662 pr_debug("%s: fanotify_fd=%d flags=%x dfd=%d pathname=%p mask=%llx\n",
663 __func__, fanotify_fd, flags, dfd, pathname, mask);
664
665 /* we only use the lower 32 bits as of right now. */
666 if (mask & ((__u64)0xffffffff << 32))
667 return -EINVAL;
668
669 if (flags & ~FAN_ALL_MARK_FLAGS)
670 return -EINVAL;
671 switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE | FAN_MARK_FLUSH)) {
672 case FAN_MARK_ADD:
673 case FAN_MARK_REMOVE:
674 case FAN_MARK_FLUSH:
675 break;
676 default:
677 return -EINVAL;
678 }
679#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
680 if (mask & ~(FAN_ALL_EVENTS | FAN_ALL_PERM_EVENTS | FAN_EVENT_ON_CHILD))
681#else
682 if (mask & ~(FAN_ALL_EVENTS | FAN_EVENT_ON_CHILD))
683#endif
684 return -EINVAL;
685
686 filp = fget_light(fanotify_fd, &fput_needed);
687 if (unlikely(!filp))
688 return -EBADF;
689
690 /* verify that this is indeed an fanotify instance */
691 ret = -EINVAL;
692 if (unlikely(filp->f_op != &fanotify_fops))
693 goto fput_and_out;
694
695 ret = fanotify_find_path(dfd, pathname, &path, flags);
696 if (ret)
697 goto fput_and_out;
698
699 /* inode held in place by reference to path; group by fget on fd */
700 if (!(flags & FAN_MARK_MOUNT))
701 inode = path.dentry->d_inode;
702 else
703 mnt = path.mnt;
704 group = filp->private_data;
705
706 /* create/update an inode mark */
707 switch (flags & (FAN_MARK_ADD | FAN_MARK_REMOVE | FAN_MARK_FLUSH)) {
708 case FAN_MARK_ADD:
709 if (flags & FAN_MARK_MOUNT)
710 ret = fanotify_add_vfsmount_mark(group, mnt, mask, flags);
711 else
712 ret = fanotify_add_inode_mark(group, inode, mask, flags);
713 break;
714 case FAN_MARK_REMOVE:
715 if (flags & FAN_MARK_MOUNT)
716 ret = fanotify_remove_vfsmount_mark(group, mnt, mask, flags);
717 else
718 ret = fanotify_remove_inode_mark(group, inode, mask, flags);
719 break;
720 case FAN_MARK_FLUSH:
721 if (flags & FAN_MARK_MOUNT)
722 fsnotify_clear_vfsmount_marks_by_group(group);
723 else
724 fsnotify_clear_inode_marks_by_group(group);
725 break;
726 default:
727 ret = -EINVAL;
728 }
729
730 path_put(&path);
731fput_and_out:
732 fput_light(filp, fput_needed);
733 return ret;
734}
735
736#ifdef CONFIG_HAVE_SYSCALL_WRAPPERS
737asmlinkage long SyS_fanotify_mark(long fanotify_fd, long flags, __u64 mask,
738 long dfd, long pathname)
739{
740 return SYSC_fanotify_mark((int) fanotify_fd, (unsigned int) flags,
741 mask, (int) dfd,
742 (const char __user *) pathname);
743}
744SYSCALL_ALIAS(sys_fanotify_mark, SyS_fanotify_mark);
745#endif
746
747/*
748 * fanotify_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 */
752static int __init fanotify_user_setup(void)
753{
754 fanotify_mark_cache = KMEM_CACHE(fsnotify_mark, SLAB_PANIC);
755 fanotify_response_event_cache = KMEM_CACHE(fanotify_response_event,
756 SLAB_PANIC);
757
758 return 0;
759}
760device_initcall(fanotify_user_setup);
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-18 17:18:31 -0500