~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/notify/inotify/inotify_user.c

Version: ~ [ linux-5.1-rc5 ] ~ [ linux-5.0.7 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.34 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.111 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.168 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.178 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.138 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.65 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.39.4 ] ~ [ linux-2.6.38.8 ] ~ [ linux-2.6.37.6 ] ~ [ linux-2.6.36.4 ] ~ [ linux-2.6.35.14 ] ~ [ linux-2.6.34.15 ] ~ [ linux-2.6.33.20 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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  * Copyright (C) 2009 Eric Paris <Red Hat Inc>
 12  * inotify was largely rewriten to make use of the fsnotify infrastructure
 13  *
 14  * This program is free software; you can redistribute it and/or modify it
 15  * under the terms of the GNU General Public License as published by the
 16  * Free Software Foundation; either version 2, or (at your option) any
 17  * later version.
 18  *
 19  * This program is distributed in the hope that it will be useful, but
 20  * WITHOUT ANY WARRANTY; without even the implied warranty of
 21  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 22  * General Public License for more details.
 23  */
 24 
 25 #include <linux/file.h>
 26 #include <linux/fs.h> /* struct inode */
 27 #include <linux/fsnotify_backend.h>
 28 #include <linux/idr.h>
 29 #include <linux/init.h> /* fs_initcall */
 30 #include <linux/inotify.h>
 31 #include <linux/kernel.h> /* roundup() */
 32 #include <linux/namei.h> /* LOOKUP_FOLLOW */
 33 #include <linux/sched/signal.h>
 34 #include <linux/slab.h> /* struct kmem_cache */
 35 #include <linux/syscalls.h>
 36 #include <linux/types.h>
 37 #include <linux/anon_inodes.h>
 38 #include <linux/uaccess.h>
 39 #include <linux/poll.h>
 40 #include <linux/wait.h>
 41 #include <linux/memcontrol.h>
 42 
 43 #include "inotify.h"
 44 #include "../fdinfo.h"
 45 
 46 #include <asm/ioctls.h>
 47 
 48 /* configurable via /proc/sys/fs/inotify/ */
 49 static int inotify_max_queued_events __read_mostly;
 50 
 51 struct kmem_cache *inotify_inode_mark_cachep __read_mostly;
 52 
 53 #ifdef CONFIG_SYSCTL
 54 
 55 #include <linux/sysctl.h>
 56 
 57 static int zero;
 58 
 59 struct ctl_table inotify_table[] = {
 60         {
 61                 .procname       = "max_user_instances",
 62                 .data           = &init_user_ns.ucount_max[UCOUNT_INOTIFY_INSTANCES],
 63                 .maxlen         = sizeof(int),
 64                 .mode           = 0644,
 65                 .proc_handler   = proc_dointvec_minmax,
 66                 .extra1         = &zero,
 67         },
 68         {
 69                 .procname       = "max_user_watches",
 70                 .data           = &init_user_ns.ucount_max[UCOUNT_INOTIFY_WATCHES],
 71                 .maxlen         = sizeof(int),
 72                 .mode           = 0644,
 73                 .proc_handler   = proc_dointvec_minmax,
 74                 .extra1         = &zero,
 75         },
 76         {
 77                 .procname       = "max_queued_events",
 78                 .data           = &inotify_max_queued_events,
 79                 .maxlen         = sizeof(int),
 80                 .mode           = 0644,
 81                 .proc_handler   = proc_dointvec_minmax,
 82                 .extra1         = &zero
 83         },
 84         { }
 85 };
 86 #endif /* CONFIG_SYSCTL */
 87 
 88 static inline __u32 inotify_arg_to_mask(u32 arg)
 89 {
 90         __u32 mask;
 91 
 92         /*
 93          * everything should accept their own ignored, cares about children,
 94          * and should receive events when the inode is unmounted
 95          */
 96         mask = (FS_IN_IGNORED | FS_EVENT_ON_CHILD | FS_UNMOUNT);
 97 
 98         /* mask off the flags used to open the fd */
 99         mask |= (arg & (IN_ALL_EVENTS | IN_ONESHOT | IN_EXCL_UNLINK));
100 
101         return mask;
102 }
103 
104 static inline u32 inotify_mask_to_arg(__u32 mask)
105 {
106         return mask & (IN_ALL_EVENTS | IN_ISDIR | IN_UNMOUNT | IN_IGNORED |
107                        IN_Q_OVERFLOW);
108 }
109 
110 /* intofiy userspace file descriptor functions */
111 static __poll_t inotify_poll(struct file *file, poll_table *wait)
112 {
113         struct fsnotify_group *group = file->private_data;
114         __poll_t ret = 0;
115 
116         poll_wait(file, &group->notification_waitq, wait);
117         spin_lock(&group->notification_lock);
118         if (!fsnotify_notify_queue_is_empty(group))
119                 ret = EPOLLIN | EPOLLRDNORM;
120         spin_unlock(&group->notification_lock);
121 
122         return ret;
123 }
124 
125 static int round_event_name_len(struct fsnotify_event *fsn_event)
126 {
127         struct inotify_event_info *event;
128 
129         event = INOTIFY_E(fsn_event);
130         if (!event->name_len)
131                 return 0;
132         return roundup(event->name_len + 1, sizeof(struct inotify_event));
133 }
134 
135 /*
136  * Get an inotify_kernel_event if one exists and is small
137  * enough to fit in "count". Return an error pointer if
138  * not large enough.
139  *
140  * Called with the group->notification_lock held.
141  */
142 static struct fsnotify_event *get_one_event(struct fsnotify_group *group,
143                                             size_t count)
144 {
145         size_t event_size = sizeof(struct inotify_event);
146         struct fsnotify_event *event;
147 
148         if (fsnotify_notify_queue_is_empty(group))
149                 return NULL;
150 
151         event = fsnotify_peek_first_event(group);
152 
153         pr_debug("%s: group=%p event=%p\n", __func__, group, event);
154 
155         event_size += round_event_name_len(event);
156         if (event_size > count)
157                 return ERR_PTR(-EINVAL);
158 
159         /* held the notification_lock the whole time, so this is the
160          * same event we peeked above */
161         fsnotify_remove_first_event(group);
162 
163         return event;
164 }
165 
166 /*
167  * Copy an event to user space, returning how much we copied.
168  *
169  * We already checked that the event size is smaller than the
170  * buffer we had in "get_one_event()" above.
171  */
172 static ssize_t copy_event_to_user(struct fsnotify_group *group,
173                                   struct fsnotify_event *fsn_event,
174                                   char __user *buf)
175 {
176         struct inotify_event inotify_event;
177         struct inotify_event_info *event;
178         size_t event_size = sizeof(struct inotify_event);
179         size_t name_len;
180         size_t pad_name_len;
181 
182         pr_debug("%s: group=%p event=%p\n", __func__, group, fsn_event);
183 
184         event = INOTIFY_E(fsn_event);
185         name_len = event->name_len;
186         /*
187          * round up name length so it is a multiple of event_size
188          * plus an extra byte for the terminating '\0'.
189          */
190         pad_name_len = round_event_name_len(fsn_event);
191         inotify_event.len = pad_name_len;
192         inotify_event.mask = inotify_mask_to_arg(event->mask);
193         inotify_event.wd = event->wd;
194         inotify_event.cookie = event->sync_cookie;
195 
196         /* send the main event */
197         if (copy_to_user(buf, &inotify_event, event_size))
198                 return -EFAULT;
199 
200         buf += event_size;
201 
202         /*
203          * fsnotify only stores the pathname, so here we have to send the pathname
204          * and then pad that pathname out to a multiple of sizeof(inotify_event)
205          * with zeros.
206          */
207         if (pad_name_len) {
208                 /* copy the path name */
209                 if (copy_to_user(buf, event->name, name_len))
210                         return -EFAULT;
211                 buf += name_len;
212 
213                 /* fill userspace with 0's */
214                 if (clear_user(buf, pad_name_len - name_len))
215                         return -EFAULT;
216                 event_size += pad_name_len;
217         }
218 
219         return event_size;
220 }
221 
222 static ssize_t inotify_read(struct file *file, char __user *buf,
223                             size_t count, loff_t *pos)
224 {
225         struct fsnotify_group *group;
226         struct fsnotify_event *kevent;
227         char __user *start;
228         int ret;
229         DEFINE_WAIT_FUNC(wait, woken_wake_function);
230 
231         start = buf;
232         group = file->private_data;
233 
234         add_wait_queue(&group->notification_waitq, &wait);
235         while (1) {
236                 spin_lock(&group->notification_lock);
237                 kevent = get_one_event(group, count);
238                 spin_unlock(&group->notification_lock);
239 
240                 pr_debug("%s: group=%p kevent=%p\n", __func__, group, kevent);
241 
242                 if (kevent) {
243                         ret = PTR_ERR(kevent);
244                         if (IS_ERR(kevent))
245                                 break;
246                         ret = copy_event_to_user(group, kevent, buf);
247                         fsnotify_destroy_event(group, kevent);
248                         if (ret < 0)
249                                 break;
250                         buf += ret;
251                         count -= ret;
252                         continue;
253                 }
254 
255                 ret = -EAGAIN;
256                 if (file->f_flags & O_NONBLOCK)
257                         break;
258                 ret = -ERESTARTSYS;
259                 if (signal_pending(current))
260                         break;
261 
262                 if (start != buf)
263                         break;
264 
265                 wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
266         }
267         remove_wait_queue(&group->notification_waitq, &wait);
268 
269         if (start != buf && ret != -EFAULT)
270                 ret = buf - start;
271         return ret;
272 }
273 
274 static int inotify_release(struct inode *ignored, struct file *file)
275 {
276         struct fsnotify_group *group = file->private_data;
277 
278         pr_debug("%s: group=%p\n", __func__, group);
279 
280         /* free this group, matching get was inotify_init->fsnotify_obtain_group */
281         fsnotify_destroy_group(group);
282 
283         return 0;
284 }
285 
286 static long inotify_ioctl(struct file *file, unsigned int cmd,
287                           unsigned long arg)
288 {
289         struct fsnotify_group *group;
290         struct fsnotify_event *fsn_event;
291         void __user *p;
292         int ret = -ENOTTY;
293         size_t send_len = 0;
294 
295         group = file->private_data;
296         p = (void __user *) arg;
297 
298         pr_debug("%s: group=%p cmd=%u\n", __func__, group, cmd);
299 
300         switch (cmd) {
301         case FIONREAD:
302                 spin_lock(&group->notification_lock);
303                 list_for_each_entry(fsn_event, &group->notification_list,
304                                     list) {
305                         send_len += sizeof(struct inotify_event);
306                         send_len += round_event_name_len(fsn_event);
307                 }
308                 spin_unlock(&group->notification_lock);
309                 ret = put_user(send_len, (int __user *) p);
310                 break;
311 #ifdef CONFIG_CHECKPOINT_RESTORE
312         case INOTIFY_IOC_SETNEXTWD:
313                 ret = -EINVAL;
314                 if (arg >= 1 && arg <= INT_MAX) {
315                         struct inotify_group_private_data *data;
316 
317                         data = &group->inotify_data;
318                         spin_lock(&data->idr_lock);
319                         idr_set_cursor(&data->idr, (unsigned int)arg);
320                         spin_unlock(&data->idr_lock);
321                         ret = 0;
322                 }
323                 break;
324 #endif /* CONFIG_CHECKPOINT_RESTORE */
325         }
326 
327         return ret;
328 }
329 
330 static const struct file_operations inotify_fops = {
331         .show_fdinfo    = inotify_show_fdinfo,
332         .poll           = inotify_poll,
333         .read           = inotify_read,
334         .fasync         = fsnotify_fasync,
335         .release        = inotify_release,
336         .unlocked_ioctl = inotify_ioctl,
337         .compat_ioctl   = inotify_ioctl,
338         .llseek         = noop_llseek,
339 };
340 
341 
342 /*
343  * find_inode - resolve a user-given path to a specific inode
344  */
345 static int inotify_find_inode(const char __user *dirname, struct path *path, unsigned flags)
346 {
347         int error;
348 
349         error = user_path_at(AT_FDCWD, dirname, flags, path);
350         if (error)
351                 return error;
352         /* you can only watch an inode if you have read permissions on it */
353         error = inode_permission(path->dentry->d_inode, MAY_READ);
354         if (error)
355                 path_put(path);
356         return error;
357 }
358 
359 static int inotify_add_to_idr(struct idr *idr, spinlock_t *idr_lock,
360                               struct inotify_inode_mark *i_mark)
361 {
362         int ret;
363 
364         idr_preload(GFP_KERNEL);
365         spin_lock(idr_lock);
366 
367         ret = idr_alloc_cyclic(idr, i_mark, 1, 0, GFP_NOWAIT);
368         if (ret >= 0) {
369                 /* we added the mark to the idr, take a reference */
370                 i_mark->wd = ret;
371                 fsnotify_get_mark(&i_mark->fsn_mark);
372         }
373 
374         spin_unlock(idr_lock);
375         idr_preload_end();
376         return ret < 0 ? ret : 0;
377 }
378 
379 static struct inotify_inode_mark *inotify_idr_find_locked(struct fsnotify_group *group,
380                                                                 int wd)
381 {
382         struct idr *idr = &group->inotify_data.idr;
383         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
384         struct inotify_inode_mark *i_mark;
385 
386         assert_spin_locked(idr_lock);
387 
388         i_mark = idr_find(idr, wd);
389         if (i_mark) {
390                 struct fsnotify_mark *fsn_mark = &i_mark->fsn_mark;
391 
392                 fsnotify_get_mark(fsn_mark);
393                 /* One ref for being in the idr, one ref we just took */
394                 BUG_ON(refcount_read(&fsn_mark->refcnt) < 2);
395         }
396 
397         return i_mark;
398 }
399 
400 static struct inotify_inode_mark *inotify_idr_find(struct fsnotify_group *group,
401                                                          int wd)
402 {
403         struct inotify_inode_mark *i_mark;
404         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
405 
406         spin_lock(idr_lock);
407         i_mark = inotify_idr_find_locked(group, wd);
408         spin_unlock(idr_lock);
409 
410         return i_mark;
411 }
412 
413 /*
414  * Remove the mark from the idr (if present) and drop the reference
415  * on the mark because it was in the idr.
416  */
417 static void inotify_remove_from_idr(struct fsnotify_group *group,
418                                     struct inotify_inode_mark *i_mark)
419 {
420         struct idr *idr = &group->inotify_data.idr;
421         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
422         struct inotify_inode_mark *found_i_mark = NULL;
423         int wd;
424 
425         spin_lock(idr_lock);
426         wd = i_mark->wd;
427 
428         /*
429          * does this i_mark think it is in the idr?  we shouldn't get called
430          * if it wasn't....
431          */
432         if (wd == -1) {
433                 WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
434                         __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
435                 goto out;
436         }
437 
438         /* Lets look in the idr to see if we find it */
439         found_i_mark = inotify_idr_find_locked(group, wd);
440         if (unlikely(!found_i_mark)) {
441                 WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
442                         __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
443                 goto out;
444         }
445 
446         /*
447          * We found an mark in the idr at the right wd, but it's
448          * not the mark we were told to remove.  eparis seriously
449          * fucked up somewhere.
450          */
451         if (unlikely(found_i_mark != i_mark)) {
452                 WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p "
453                         "found_i_mark=%p found_i_mark->wd=%d "
454                         "found_i_mark->group=%p\n", __func__, i_mark,
455                         i_mark->wd, i_mark->fsn_mark.group, found_i_mark,
456                         found_i_mark->wd, found_i_mark->fsn_mark.group);
457                 goto out;
458         }
459 
460         /*
461          * One ref for being in the idr
462          * one ref grabbed by inotify_idr_find
463          */
464         if (unlikely(refcount_read(&i_mark->fsn_mark.refcnt) < 2)) {
465                 printk(KERN_ERR "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
466                          __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
467                 /* we can't really recover with bad ref cnting.. */
468                 BUG();
469         }
470 
471         idr_remove(idr, wd);
472         /* Removed from the idr, drop that ref. */
473         fsnotify_put_mark(&i_mark->fsn_mark);
474 out:
475         i_mark->wd = -1;
476         spin_unlock(idr_lock);
477         /* match the ref taken by inotify_idr_find_locked() */
478         if (found_i_mark)
479                 fsnotify_put_mark(&found_i_mark->fsn_mark);
480 }
481 
482 /*
483  * Send IN_IGNORED for this wd, remove this wd from the idr.
484  */
485 void inotify_ignored_and_remove_idr(struct fsnotify_mark *fsn_mark,
486                                     struct fsnotify_group *group)
487 {
488         struct inotify_inode_mark *i_mark;
489         struct fsnotify_iter_info iter_info = { };
490 
491         fsnotify_iter_set_report_type_mark(&iter_info, FSNOTIFY_OBJ_TYPE_INODE,
492                                            fsn_mark);
493 
494         /* Queue ignore event for the watch */
495         inotify_handle_event(group, NULL, FS_IN_IGNORED, NULL,
496                              FSNOTIFY_EVENT_NONE, NULL, 0, &iter_info);
497 
498         i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
499         /* remove this mark from the idr */
500         inotify_remove_from_idr(group, i_mark);
501 
502         dec_inotify_watches(group->inotify_data.ucounts);
503 }
504 
505 static int inotify_update_existing_watch(struct fsnotify_group *group,
506                                          struct inode *inode,
507                                          u32 arg)
508 {
509         struct fsnotify_mark *fsn_mark;
510         struct inotify_inode_mark *i_mark;
511         __u32 old_mask, new_mask;
512         __u32 mask;
513         int add = (arg & IN_MASK_ADD);
514         int create = (arg & IN_MASK_CREATE);
515         int ret;
516 
517         mask = inotify_arg_to_mask(arg);
518 
519         fsn_mark = fsnotify_find_mark(&inode->i_fsnotify_marks, group);
520         if (!fsn_mark)
521                 return -ENOENT;
522         else if (create) {
523                 ret = -EEXIST;
524                 goto out;
525         }
526 
527         i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
528 
529         spin_lock(&fsn_mark->lock);
530         old_mask = fsn_mark->mask;
531         if (add)
532                 fsn_mark->mask |= mask;
533         else
534                 fsn_mark->mask = mask;
535         new_mask = fsn_mark->mask;
536         spin_unlock(&fsn_mark->lock);
537 
538         if (old_mask != new_mask) {
539                 /* more bits in old than in new? */
540                 int dropped = (old_mask & ~new_mask);
541                 /* more bits in this fsn_mark than the inode's mask? */
542                 int do_inode = (new_mask & ~inode->i_fsnotify_mask);
543 
544                 /* update the inode with this new fsn_mark */
545                 if (dropped || do_inode)
546                         fsnotify_recalc_mask(inode->i_fsnotify_marks);
547 
548         }
549 
550         /* return the wd */
551         ret = i_mark->wd;
552 
553 out:
554         /* match the get from fsnotify_find_mark() */
555         fsnotify_put_mark(fsn_mark);
556 
557         return ret;
558 }
559 
560 static int inotify_new_watch(struct fsnotify_group *group,
561                              struct inode *inode,
562                              u32 arg)
563 {
564         struct inotify_inode_mark *tmp_i_mark;
565         __u32 mask;
566         int ret;
567         struct idr *idr = &group->inotify_data.idr;
568         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
569 
570         mask = inotify_arg_to_mask(arg);
571 
572         tmp_i_mark = kmem_cache_alloc(inotify_inode_mark_cachep, GFP_KERNEL);
573         if (unlikely(!tmp_i_mark))
574                 return -ENOMEM;
575 
576         fsnotify_init_mark(&tmp_i_mark->fsn_mark, group);
577         tmp_i_mark->fsn_mark.mask = mask;
578         tmp_i_mark->wd = -1;
579 
580         ret = inotify_add_to_idr(idr, idr_lock, tmp_i_mark);
581         if (ret)
582                 goto out_err;
583 
584         /* increment the number of watches the user has */
585         if (!inc_inotify_watches(group->inotify_data.ucounts)) {
586                 inotify_remove_from_idr(group, tmp_i_mark);
587                 ret = -ENOSPC;
588                 goto out_err;
589         }
590 
591         /* we are on the idr, now get on the inode */
592         ret = fsnotify_add_inode_mark_locked(&tmp_i_mark->fsn_mark, inode, 0);
593         if (ret) {
594                 /* we failed to get on the inode, get off the idr */
595                 inotify_remove_from_idr(group, tmp_i_mark);
596                 goto out_err;
597         }
598 
599 
600         /* return the watch descriptor for this new mark */
601         ret = tmp_i_mark->wd;
602 
603 out_err:
604         /* match the ref from fsnotify_init_mark() */
605         fsnotify_put_mark(&tmp_i_mark->fsn_mark);
606 
607         return ret;
608 }
609 
610 static int inotify_update_watch(struct fsnotify_group *group, struct inode *inode, u32 arg)
611 {
612         int ret = 0;
613 
614         mutex_lock(&group->mark_mutex);
615         /* try to update and existing watch with the new arg */
616         ret = inotify_update_existing_watch(group, inode, arg);
617         /* no mark present, try to add a new one */
618         if (ret == -ENOENT)
619                 ret = inotify_new_watch(group, inode, arg);
620         mutex_unlock(&group->mark_mutex);
621 
622         return ret;
623 }
624 
625 static struct fsnotify_group *inotify_new_group(unsigned int max_events)
626 {
627         struct fsnotify_group *group;
628         struct inotify_event_info *oevent;
629 
630         group = fsnotify_alloc_group(&inotify_fsnotify_ops);
631         if (IS_ERR(group))
632                 return group;
633 
634         oevent = kmalloc(sizeof(struct inotify_event_info), GFP_KERNEL);
635         if (unlikely(!oevent)) {
636                 fsnotify_destroy_group(group);
637                 return ERR_PTR(-ENOMEM);
638         }
639         group->overflow_event = &oevent->fse;
640         fsnotify_init_event(group->overflow_event, NULL);
641         oevent->mask = FS_Q_OVERFLOW;
642         oevent->wd = -1;
643         oevent->sync_cookie = 0;
644         oevent->name_len = 0;
645 
646         group->max_events = max_events;
647         group->memcg = get_mem_cgroup_from_mm(current->mm);
648 
649         spin_lock_init(&group->inotify_data.idr_lock);
650         idr_init(&group->inotify_data.idr);
651         group->inotify_data.ucounts = inc_ucount(current_user_ns(),
652                                                  current_euid(),
653                                                  UCOUNT_INOTIFY_INSTANCES);
654 
655         if (!group->inotify_data.ucounts) {
656                 fsnotify_destroy_group(group);
657                 return ERR_PTR(-EMFILE);
658         }
659 
660         return group;
661 }
662 
663 
664 /* inotify syscalls */
665 static int do_inotify_init(int flags)
666 {
667         struct fsnotify_group *group;
668         int ret;
669 
670         /* Check the IN_* constants for consistency.  */
671         BUILD_BUG_ON(IN_CLOEXEC != O_CLOEXEC);
672         BUILD_BUG_ON(IN_NONBLOCK != O_NONBLOCK);
673 
674         if (flags & ~(IN_CLOEXEC | IN_NONBLOCK))
675                 return -EINVAL;
676 
677         /* fsnotify_obtain_group took a reference to group, we put this when we kill the file in the end */
678         group = inotify_new_group(inotify_max_queued_events);
679         if (IS_ERR(group))
680                 return PTR_ERR(group);
681 
682         ret = anon_inode_getfd("inotify", &inotify_fops, group,
683                                   O_RDONLY | flags);
684         if (ret < 0)
685                 fsnotify_destroy_group(group);
686 
687         return ret;
688 }
689 
690 SYSCALL_DEFINE1(inotify_init1, int, flags)
691 {
692         return do_inotify_init(flags);
693 }
694 
695 SYSCALL_DEFINE0(inotify_init)
696 {
697         return do_inotify_init(0);
698 }
699 
700 SYSCALL_DEFINE3(inotify_add_watch, int, fd, const char __user *, pathname,
701                 u32, mask)
702 {
703         struct fsnotify_group *group;
704         struct inode *inode;
705         struct path path;
706         struct fd f;
707         int ret;
708         unsigned flags = 0;
709 
710         /*
711          * We share a lot of code with fs/dnotify.  We also share
712          * the bit layout between inotify's IN_* and the fsnotify
713          * FS_*.  This check ensures that only the inotify IN_*
714          * bits get passed in and set in watches/events.
715          */
716         if (unlikely(mask & ~ALL_INOTIFY_BITS))
717                 return -EINVAL;
718         /*
719          * Require at least one valid bit set in the mask.
720          * Without _something_ set, we would have no events to
721          * watch for.
722          */
723         if (unlikely(!(mask & ALL_INOTIFY_BITS)))
724                 return -EINVAL;
725 
726         f = fdget(fd);
727         if (unlikely(!f.file))
728                 return -EBADF;
729 
730         /* IN_MASK_ADD and IN_MASK_CREATE don't make sense together */
731         if (unlikely((mask & IN_MASK_ADD) && (mask & IN_MASK_CREATE))) {
732                 ret = -EINVAL;
733                 goto fput_and_out;
734         }
735 
736         /* verify that this is indeed an inotify instance */
737         if (unlikely(f.file->f_op != &inotify_fops)) {
738                 ret = -EINVAL;
739                 goto fput_and_out;
740         }
741 
742         if (!(mask & IN_DONT_FOLLOW))
743                 flags |= LOOKUP_FOLLOW;
744         if (mask & IN_ONLYDIR)
745                 flags |= LOOKUP_DIRECTORY;
746 
747         ret = inotify_find_inode(pathname, &path, flags);
748         if (ret)
749                 goto fput_and_out;
750 
751         /* inode held in place by reference to path; group by fget on fd */
752         inode = path.dentry->d_inode;
753         group = f.file->private_data;
754 
755         /* create/update an inode mark */
756         ret = inotify_update_watch(group, inode, mask);
757         path_put(&path);
758 fput_and_out:
759         fdput(f);
760         return ret;
761 }
762 
763 SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd)
764 {
765         struct fsnotify_group *group;
766         struct inotify_inode_mark *i_mark;
767         struct fd f;
768         int ret = 0;
769 
770         f = fdget(fd);
771         if (unlikely(!f.file))
772                 return -EBADF;
773 
774         /* verify that this is indeed an inotify instance */
775         ret = -EINVAL;
776         if (unlikely(f.file->f_op != &inotify_fops))
777                 goto out;
778 
779         group = f.file->private_data;
780 
781         ret = -EINVAL;
782         i_mark = inotify_idr_find(group, wd);
783         if (unlikely(!i_mark))
784                 goto out;
785 
786         ret = 0;
787 
788         fsnotify_destroy_mark(&i_mark->fsn_mark, group);
789 
790         /* match ref taken by inotify_idr_find */
791         fsnotify_put_mark(&i_mark->fsn_mark);
792 
793 out:
794         fdput(f);
795         return ret;
796 }
797 
798 /*
799  * inotify_user_setup - Our initialization function.  Note that we cannot return
800  * error because we have compiled-in VFS hooks.  So an (unlikely) failure here
801  * must result in panic().
802  */
803 static int __init inotify_user_setup(void)
804 {
805         BUILD_BUG_ON(IN_ACCESS != FS_ACCESS);
806         BUILD_BUG_ON(IN_MODIFY != FS_MODIFY);
807         BUILD_BUG_ON(IN_ATTRIB != FS_ATTRIB);
808         BUILD_BUG_ON(IN_CLOSE_WRITE != FS_CLOSE_WRITE);
809         BUILD_BUG_ON(IN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
810         BUILD_BUG_ON(IN_OPEN != FS_OPEN);
811         BUILD_BUG_ON(IN_MOVED_FROM != FS_MOVED_FROM);
812         BUILD_BUG_ON(IN_MOVED_TO != FS_MOVED_TO);
813         BUILD_BUG_ON(IN_CREATE != FS_CREATE);
814         BUILD_BUG_ON(IN_DELETE != FS_DELETE);
815         BUILD_BUG_ON(IN_DELETE_SELF != FS_DELETE_SELF);
816         BUILD_BUG_ON(IN_MOVE_SELF != FS_MOVE_SELF);
817         BUILD_BUG_ON(IN_UNMOUNT != FS_UNMOUNT);
818         BUILD_BUG_ON(IN_Q_OVERFLOW != FS_Q_OVERFLOW);
819         BUILD_BUG_ON(IN_IGNORED != FS_IN_IGNORED);
820         BUILD_BUG_ON(IN_EXCL_UNLINK != FS_EXCL_UNLINK);
821         BUILD_BUG_ON(IN_ISDIR != FS_ISDIR);
822         BUILD_BUG_ON(IN_ONESHOT != FS_IN_ONESHOT);
823 
824         BUILD_BUG_ON(HWEIGHT32(ALL_INOTIFY_BITS) != 22);
825 
826         inotify_inode_mark_cachep = KMEM_CACHE(inotify_inode_mark,
827                                                SLAB_PANIC|SLAB_ACCOUNT);
828 
829         inotify_max_queued_events = 16384;
830         init_user_ns.ucount_max[UCOUNT_INOTIFY_INSTANCES] = 128;
831         init_user_ns.ucount_max[UCOUNT_INOTIFY_WATCHES] = 8192;
832 
833         return 0;
834 }
835 fs_initcall(inotify_user_setup);
836 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp