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

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

Version: ~ [ linux-5.8-rc5 ] ~ [ linux-5.7.8 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.51 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.132 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.188 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.230 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.230 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ 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-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 
 42 #include "inotify.h"
 43 #include "../fdinfo.h"
 44 
 45 #include <asm/ioctls.h>
 46 
 47 /* configurable via /proc/sys/fs/inotify/ */
 48 static int inotify_max_queued_events __read_mostly;
 49 
 50 struct kmem_cache *inotify_inode_mark_cachep __read_mostly;
 51 
 52 #ifdef CONFIG_SYSCTL
 53 
 54 #include <linux/sysctl.h>
 55 
 56 static int zero;
 57 
 58 struct ctl_table inotify_table[] = {
 59         {
 60                 .procname       = "max_user_instances",
 61                 .data           = &init_user_ns.ucount_max[UCOUNT_INOTIFY_INSTANCES],
 62                 .maxlen         = sizeof(int),
 63                 .mode           = 0644,
 64                 .proc_handler   = proc_dointvec_minmax,
 65                 .extra1         = &zero,
 66         },
 67         {
 68                 .procname       = "max_user_watches",
 69                 .data           = &init_user_ns.ucount_max[UCOUNT_INOTIFY_WATCHES],
 70                 .maxlen         = sizeof(int),
 71                 .mode           = 0644,
 72                 .proc_handler   = proc_dointvec_minmax,
 73                 .extra1         = &zero,
 74         },
 75         {
 76                 .procname       = "max_queued_events",
 77                 .data           = &inotify_max_queued_events,
 78                 .maxlen         = sizeof(int),
 79                 .mode           = 0644,
 80                 .proc_handler   = proc_dointvec_minmax,
 81                 .extra1         = &zero
 82         },
 83         { }
 84 };
 85 #endif /* CONFIG_SYSCTL */
 86 
 87 static inline __u32 inotify_arg_to_mask(u32 arg)
 88 {
 89         __u32 mask;
 90 
 91         /*
 92          * everything should accept their own ignored, cares about children,
 93          * and should receive events when the inode is unmounted
 94          */
 95         mask = (FS_IN_IGNORED | FS_EVENT_ON_CHILD | FS_UNMOUNT);
 96 
 97         /* mask off the flags used to open the fd */
 98         mask |= (arg & (IN_ALL_EVENTS | IN_ONESHOT | IN_EXCL_UNLINK));
 99 
100         return mask;
101 }
102 
103 static inline u32 inotify_mask_to_arg(__u32 mask)
104 {
105         return mask & (IN_ALL_EVENTS | IN_ISDIR | IN_UNMOUNT | IN_IGNORED |
106                        IN_Q_OVERFLOW);
107 }
108 
109 /* intofiy userspace file descriptor functions */
110 static __poll_t inotify_poll(struct file *file, poll_table *wait)
111 {
112         struct fsnotify_group *group = file->private_data;
113         __poll_t ret = 0;
114 
115         poll_wait(file, &group->notification_waitq, wait);
116         spin_lock(&group->notification_lock);
117         if (!fsnotify_notify_queue_is_empty(group))
118                 ret = EPOLLIN | EPOLLRDNORM;
119         spin_unlock(&group->notification_lock);
120 
121         return ret;
122 }
123 
124 static int round_event_name_len(struct fsnotify_event *fsn_event)
125 {
126         struct inotify_event_info *event;
127 
128         event = INOTIFY_E(fsn_event);
129         if (!event->name_len)
130                 return 0;
131         return roundup(event->name_len + 1, sizeof(struct inotify_event));
132 }
133 
134 /*
135  * Get an inotify_kernel_event if one exists and is small
136  * enough to fit in "count". Return an error pointer if
137  * not large enough.
138  *
139  * Called with the group->notification_lock held.
140  */
141 static struct fsnotify_event *get_one_event(struct fsnotify_group *group,
142                                             size_t count)
143 {
144         size_t event_size = sizeof(struct inotify_event);
145         struct fsnotify_event *event;
146 
147         if (fsnotify_notify_queue_is_empty(group))
148                 return NULL;
149 
150         event = fsnotify_peek_first_event(group);
151 
152         pr_debug("%s: group=%p event=%p\n", __func__, group, event);
153 
154         event_size += round_event_name_len(event);
155         if (event_size > count)
156                 return ERR_PTR(-EINVAL);
157 
158         /* held the notification_lock the whole time, so this is the
159          * same event we peeked above */
160         fsnotify_remove_first_event(group);
161 
162         return event;
163 }
164 
165 /*
166  * Copy an event to user space, returning how much we copied.
167  *
168  * We already checked that the event size is smaller than the
169  * buffer we had in "get_one_event()" above.
170  */
171 static ssize_t copy_event_to_user(struct fsnotify_group *group,
172                                   struct fsnotify_event *fsn_event,
173                                   char __user *buf)
174 {
175         struct inotify_event inotify_event;
176         struct inotify_event_info *event;
177         size_t event_size = sizeof(struct inotify_event);
178         size_t name_len;
179         size_t pad_name_len;
180 
181         pr_debug("%s: group=%p event=%p\n", __func__, group, fsn_event);
182 
183         event = INOTIFY_E(fsn_event);
184         name_len = event->name_len;
185         /*
186          * round up name length so it is a multiple of event_size
187          * plus an extra byte for the terminating '\0'.
188          */
189         pad_name_len = round_event_name_len(fsn_event);
190         inotify_event.len = pad_name_len;
191         inotify_event.mask = inotify_mask_to_arg(fsn_event->mask);
192         inotify_event.wd = event->wd;
193         inotify_event.cookie = event->sync_cookie;
194 
195         /* send the main event */
196         if (copy_to_user(buf, &inotify_event, event_size))
197                 return -EFAULT;
198 
199         buf += event_size;
200 
201         /*
202          * fsnotify only stores the pathname, so here we have to send the pathname
203          * and then pad that pathname out to a multiple of sizeof(inotify_event)
204          * with zeros.
205          */
206         if (pad_name_len) {
207                 /* copy the path name */
208                 if (copy_to_user(buf, event->name, name_len))
209                         return -EFAULT;
210                 buf += name_len;
211 
212                 /* fill userspace with 0's */
213                 if (clear_user(buf, pad_name_len - name_len))
214                         return -EFAULT;
215                 event_size += pad_name_len;
216         }
217 
218         return event_size;
219 }
220 
221 static ssize_t inotify_read(struct file *file, char __user *buf,
222                             size_t count, loff_t *pos)
223 {
224         struct fsnotify_group *group;
225         struct fsnotify_event *kevent;
226         char __user *start;
227         int ret;
228         DEFINE_WAIT_FUNC(wait, woken_wake_function);
229 
230         start = buf;
231         group = file->private_data;
232 
233         add_wait_queue(&group->notification_waitq, &wait);
234         while (1) {
235                 spin_lock(&group->notification_lock);
236                 kevent = get_one_event(group, count);
237                 spin_unlock(&group->notification_lock);
238 
239                 pr_debug("%s: group=%p kevent=%p\n", __func__, group, kevent);
240 
241                 if (kevent) {
242                         ret = PTR_ERR(kevent);
243                         if (IS_ERR(kevent))
244                                 break;
245                         ret = copy_event_to_user(group, kevent, buf);
246                         fsnotify_destroy_event(group, kevent);
247                         if (ret < 0)
248                                 break;
249                         buf += ret;
250                         count -= ret;
251                         continue;
252                 }
253 
254                 ret = -EAGAIN;
255                 if (file->f_flags & O_NONBLOCK)
256                         break;
257                 ret = -ERESTARTSYS;
258                 if (signal_pending(current))
259                         break;
260 
261                 if (start != buf)
262                         break;
263 
264                 wait_woken(&wait, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
265         }
266         remove_wait_queue(&group->notification_waitq, &wait);
267 
268         if (start != buf && ret != -EFAULT)
269                 ret = buf - start;
270         return ret;
271 }
272 
273 static int inotify_release(struct inode *ignored, struct file *file)
274 {
275         struct fsnotify_group *group = file->private_data;
276 
277         pr_debug("%s: group=%p\n", __func__, group);
278 
279         /* free this group, matching get was inotify_init->fsnotify_obtain_group */
280         fsnotify_destroy_group(group);
281 
282         return 0;
283 }
284 
285 static long inotify_ioctl(struct file *file, unsigned int cmd,
286                           unsigned long arg)
287 {
288         struct fsnotify_group *group;
289         struct fsnotify_event *fsn_event;
290         void __user *p;
291         int ret = -ENOTTY;
292         size_t send_len = 0;
293 
294         group = file->private_data;
295         p = (void __user *) arg;
296 
297         pr_debug("%s: group=%p cmd=%u\n", __func__, group, cmd);
298 
299         switch (cmd) {
300         case FIONREAD:
301                 spin_lock(&group->notification_lock);
302                 list_for_each_entry(fsn_event, &group->notification_list,
303                                     list) {
304                         send_len += sizeof(struct inotify_event);
305                         send_len += round_event_name_len(fsn_event);
306                 }
307                 spin_unlock(&group->notification_lock);
308                 ret = put_user(send_len, (int __user *) p);
309                 break;
310 #ifdef CONFIG_CHECKPOINT_RESTORE
311         case INOTIFY_IOC_SETNEXTWD:
312                 ret = -EINVAL;
313                 if (arg >= 1 && arg <= INT_MAX) {
314                         struct inotify_group_private_data *data;
315 
316                         data = &group->inotify_data;
317                         spin_lock(&data->idr_lock);
318                         idr_set_cursor(&data->idr, (unsigned int)arg);
319                         spin_unlock(&data->idr_lock);
320                         ret = 0;
321                 }
322                 break;
323 #endif /* CONFIG_CHECKPOINT_RESTORE */
324         }
325 
326         return ret;
327 }
328 
329 static const struct file_operations inotify_fops = {
330         .show_fdinfo    = inotify_show_fdinfo,
331         .poll           = inotify_poll,
332         .read           = inotify_read,
333         .fasync         = fsnotify_fasync,
334         .release        = inotify_release,
335         .unlocked_ioctl = inotify_ioctl,
336         .compat_ioctl   = inotify_ioctl,
337         .llseek         = noop_llseek,
338 };
339 
340 
341 /*
342  * find_inode - resolve a user-given path to a specific inode
343  */
344 static int inotify_find_inode(const char __user *dirname, struct path *path, unsigned flags)
345 {
346         int error;
347 
348         error = user_path_at(AT_FDCWD, dirname, flags, path);
349         if (error)
350                 return error;
351         /* you can only watch an inode if you have read permissions on it */
352         error = inode_permission(path->dentry->d_inode, MAY_READ);
353         if (error)
354                 path_put(path);
355         return error;
356 }
357 
358 static int inotify_add_to_idr(struct idr *idr, spinlock_t *idr_lock,
359                               struct inotify_inode_mark *i_mark)
360 {
361         int ret;
362 
363         idr_preload(GFP_KERNEL);
364         spin_lock(idr_lock);
365 
366         ret = idr_alloc_cyclic(idr, i_mark, 1, 0, GFP_NOWAIT);
367         if (ret >= 0) {
368                 /* we added the mark to the idr, take a reference */
369                 i_mark->wd = ret;
370                 fsnotify_get_mark(&i_mark->fsn_mark);
371         }
372 
373         spin_unlock(idr_lock);
374         idr_preload_end();
375         return ret < 0 ? ret : 0;
376 }
377 
378 static struct inotify_inode_mark *inotify_idr_find_locked(struct fsnotify_group *group,
379                                                                 int wd)
380 {
381         struct idr *idr = &group->inotify_data.idr;
382         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
383         struct inotify_inode_mark *i_mark;
384 
385         assert_spin_locked(idr_lock);
386 
387         i_mark = idr_find(idr, wd);
388         if (i_mark) {
389                 struct fsnotify_mark *fsn_mark = &i_mark->fsn_mark;
390 
391                 fsnotify_get_mark(fsn_mark);
392                 /* One ref for being in the idr, one ref we just took */
393                 BUG_ON(refcount_read(&fsn_mark->refcnt) < 2);
394         }
395 
396         return i_mark;
397 }
398 
399 static struct inotify_inode_mark *inotify_idr_find(struct fsnotify_group *group,
400                                                          int wd)
401 {
402         struct inotify_inode_mark *i_mark;
403         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
404 
405         spin_lock(idr_lock);
406         i_mark = inotify_idr_find_locked(group, wd);
407         spin_unlock(idr_lock);
408 
409         return i_mark;
410 }
411 
412 /*
413  * Remove the mark from the idr (if present) and drop the reference
414  * on the mark because it was in the idr.
415  */
416 static void inotify_remove_from_idr(struct fsnotify_group *group,
417                                     struct inotify_inode_mark *i_mark)
418 {
419         struct idr *idr = &group->inotify_data.idr;
420         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
421         struct inotify_inode_mark *found_i_mark = NULL;
422         int wd;
423 
424         spin_lock(idr_lock);
425         wd = i_mark->wd;
426 
427         /*
428          * does this i_mark think it is in the idr?  we shouldn't get called
429          * if it wasn't....
430          */
431         if (wd == -1) {
432                 WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
433                         __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
434                 goto out;
435         }
436 
437         /* Lets look in the idr to see if we find it */
438         found_i_mark = inotify_idr_find_locked(group, wd);
439         if (unlikely(!found_i_mark)) {
440                 WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
441                         __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
442                 goto out;
443         }
444 
445         /*
446          * We found an mark in the idr at the right wd, but it's
447          * not the mark we were told to remove.  eparis seriously
448          * fucked up somewhere.
449          */
450         if (unlikely(found_i_mark != i_mark)) {
451                 WARN_ONCE(1, "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p "
452                         "found_i_mark=%p found_i_mark->wd=%d "
453                         "found_i_mark->group=%p\n", __func__, i_mark,
454                         i_mark->wd, i_mark->fsn_mark.group, found_i_mark,
455                         found_i_mark->wd, found_i_mark->fsn_mark.group);
456                 goto out;
457         }
458 
459         /*
460          * One ref for being in the idr
461          * one ref grabbed by inotify_idr_find
462          */
463         if (unlikely(refcount_read(&i_mark->fsn_mark.refcnt) < 2)) {
464                 printk(KERN_ERR "%s: i_mark=%p i_mark->wd=%d i_mark->group=%p\n",
465                          __func__, i_mark, i_mark->wd, i_mark->fsn_mark.group);
466                 /* we can't really recover with bad ref cnting.. */
467                 BUG();
468         }
469 
470         idr_remove(idr, wd);
471         /* Removed from the idr, drop that ref. */
472         fsnotify_put_mark(&i_mark->fsn_mark);
473 out:
474         i_mark->wd = -1;
475         spin_unlock(idr_lock);
476         /* match the ref taken by inotify_idr_find_locked() */
477         if (found_i_mark)
478                 fsnotify_put_mark(&found_i_mark->fsn_mark);
479 }
480 
481 /*
482  * Send IN_IGNORED for this wd, remove this wd from the idr.
483  */
484 void inotify_ignored_and_remove_idr(struct fsnotify_mark *fsn_mark,
485                                     struct fsnotify_group *group)
486 {
487         struct inotify_inode_mark *i_mark;
488         struct fsnotify_iter_info iter_info = { };
489 
490         fsnotify_iter_set_report_type_mark(&iter_info, FSNOTIFY_OBJ_TYPE_INODE,
491                                            fsn_mark);
492 
493         /* Queue ignore event for the watch */
494         inotify_handle_event(group, NULL, FS_IN_IGNORED, NULL,
495                              FSNOTIFY_EVENT_NONE, NULL, 0, &iter_info);
496 
497         i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
498         /* remove this mark from the idr */
499         inotify_remove_from_idr(group, i_mark);
500 
501         dec_inotify_watches(group->inotify_data.ucounts);
502 }
503 
504 static int inotify_update_existing_watch(struct fsnotify_group *group,
505                                          struct inode *inode,
506                                          u32 arg)
507 {
508         struct fsnotify_mark *fsn_mark;
509         struct inotify_inode_mark *i_mark;
510         __u32 old_mask, new_mask;
511         __u32 mask;
512         int add = (arg & IN_MASK_ADD);
513         int ret;
514 
515         mask = inotify_arg_to_mask(arg);
516 
517         fsn_mark = fsnotify_find_mark(&inode->i_fsnotify_marks, group);
518         if (!fsn_mark)
519                 return -ENOENT;
520 
521         i_mark = container_of(fsn_mark, struct inotify_inode_mark, fsn_mark);
522 
523         spin_lock(&fsn_mark->lock);
524         old_mask = fsn_mark->mask;
525         if (add)
526                 fsn_mark->mask |= mask;
527         else
528                 fsn_mark->mask = mask;
529         new_mask = fsn_mark->mask;
530         spin_unlock(&fsn_mark->lock);
531 
532         if (old_mask != new_mask) {
533                 /* more bits in old than in new? */
534                 int dropped = (old_mask & ~new_mask);
535                 /* more bits in this fsn_mark than the inode's mask? */
536                 int do_inode = (new_mask & ~inode->i_fsnotify_mask);
537 
538                 /* update the inode with this new fsn_mark */
539                 if (dropped || do_inode)
540                         fsnotify_recalc_mask(inode->i_fsnotify_marks);
541 
542         }
543 
544         /* return the wd */
545         ret = i_mark->wd;
546 
547         /* match the get from fsnotify_find_mark() */
548         fsnotify_put_mark(fsn_mark);
549 
550         return ret;
551 }
552 
553 static int inotify_new_watch(struct fsnotify_group *group,
554                              struct inode *inode,
555                              u32 arg)
556 {
557         struct inotify_inode_mark *tmp_i_mark;
558         __u32 mask;
559         int ret;
560         struct idr *idr = &group->inotify_data.idr;
561         spinlock_t *idr_lock = &group->inotify_data.idr_lock;
562 
563         mask = inotify_arg_to_mask(arg);
564 
565         tmp_i_mark = kmem_cache_alloc(inotify_inode_mark_cachep, GFP_KERNEL);
566         if (unlikely(!tmp_i_mark))
567                 return -ENOMEM;
568 
569         fsnotify_init_mark(&tmp_i_mark->fsn_mark, group);
570         tmp_i_mark->fsn_mark.mask = mask;
571         tmp_i_mark->wd = -1;
572 
573         ret = inotify_add_to_idr(idr, idr_lock, tmp_i_mark);
574         if (ret)
575                 goto out_err;
576 
577         /* increment the number of watches the user has */
578         if (!inc_inotify_watches(group->inotify_data.ucounts)) {
579                 inotify_remove_from_idr(group, tmp_i_mark);
580                 ret = -ENOSPC;
581                 goto out_err;
582         }
583 
584         /* we are on the idr, now get on the inode */
585         ret = fsnotify_add_inode_mark_locked(&tmp_i_mark->fsn_mark, inode, 0);
586         if (ret) {
587                 /* we failed to get on the inode, get off the idr */
588                 inotify_remove_from_idr(group, tmp_i_mark);
589                 goto out_err;
590         }
591 
592 
593         /* return the watch descriptor for this new mark */
594         ret = tmp_i_mark->wd;
595 
596 out_err:
597         /* match the ref from fsnotify_init_mark() */
598         fsnotify_put_mark(&tmp_i_mark->fsn_mark);
599 
600         return ret;
601 }
602 
603 static int inotify_update_watch(struct fsnotify_group *group, struct inode *inode, u32 arg)
604 {
605         int ret = 0;
606 
607         mutex_lock(&group->mark_mutex);
608         /* try to update and existing watch with the new arg */
609         ret = inotify_update_existing_watch(group, inode, arg);
610         /* no mark present, try to add a new one */
611         if (ret == -ENOENT)
612                 ret = inotify_new_watch(group, inode, arg);
613         mutex_unlock(&group->mark_mutex);
614 
615         return ret;
616 }
617 
618 static struct fsnotify_group *inotify_new_group(unsigned int max_events)
619 {
620         struct fsnotify_group *group;
621         struct inotify_event_info *oevent;
622 
623         group = fsnotify_alloc_group(&inotify_fsnotify_ops);
624         if (IS_ERR(group))
625                 return group;
626 
627         oevent = kmalloc(sizeof(struct inotify_event_info), GFP_KERNEL);
628         if (unlikely(!oevent)) {
629                 fsnotify_destroy_group(group);
630                 return ERR_PTR(-ENOMEM);
631         }
632         group->overflow_event = &oevent->fse;
633         fsnotify_init_event(group->overflow_event, NULL, FS_Q_OVERFLOW);
634         oevent->wd = -1;
635         oevent->sync_cookie = 0;
636         oevent->name_len = 0;
637 
638         group->max_events = max_events;
639 
640         spin_lock_init(&group->inotify_data.idr_lock);
641         idr_init(&group->inotify_data.idr);
642         group->inotify_data.ucounts = inc_ucount(current_user_ns(),
643                                                  current_euid(),
644                                                  UCOUNT_INOTIFY_INSTANCES);
645 
646         if (!group->inotify_data.ucounts) {
647                 fsnotify_destroy_group(group);
648                 return ERR_PTR(-EMFILE);
649         }
650 
651         return group;
652 }
653 
654 
655 /* inotify syscalls */
656 static int do_inotify_init(int flags)
657 {
658         struct fsnotify_group *group;
659         int ret;
660 
661         /* Check the IN_* constants for consistency.  */
662         BUILD_BUG_ON(IN_CLOEXEC != O_CLOEXEC);
663         BUILD_BUG_ON(IN_NONBLOCK != O_NONBLOCK);
664 
665         if (flags & ~(IN_CLOEXEC | IN_NONBLOCK))
666                 return -EINVAL;
667 
668         /* fsnotify_obtain_group took a reference to group, we put this when we kill the file in the end */
669         group = inotify_new_group(inotify_max_queued_events);
670         if (IS_ERR(group))
671                 return PTR_ERR(group);
672 
673         ret = anon_inode_getfd("inotify", &inotify_fops, group,
674                                   O_RDONLY | flags);
675         if (ret < 0)
676                 fsnotify_destroy_group(group);
677 
678         return ret;
679 }
680 
681 SYSCALL_DEFINE1(inotify_init1, int, flags)
682 {
683         return do_inotify_init(flags);
684 }
685 
686 SYSCALL_DEFINE0(inotify_init)
687 {
688         return do_inotify_init(0);
689 }
690 
691 SYSCALL_DEFINE3(inotify_add_watch, int, fd, const char __user *, pathname,
692                 u32, mask)
693 {
694         struct fsnotify_group *group;
695         struct inode *inode;
696         struct path path;
697         struct fd f;
698         int ret;
699         unsigned flags = 0;
700 
701         /*
702          * We share a lot of code with fs/dnotify.  We also share
703          * the bit layout between inotify's IN_* and the fsnotify
704          * FS_*.  This check ensures that only the inotify IN_*
705          * bits get passed in and set in watches/events.
706          */
707         if (unlikely(mask & ~ALL_INOTIFY_BITS))
708                 return -EINVAL;
709         /*
710          * Require at least one valid bit set in the mask.
711          * Without _something_ set, we would have no events to
712          * watch for.
713          */
714         if (unlikely(!(mask & ALL_INOTIFY_BITS)))
715                 return -EINVAL;
716 
717         f = fdget(fd);
718         if (unlikely(!f.file))
719                 return -EBADF;
720 
721         /* verify that this is indeed an inotify instance */
722         if (unlikely(f.file->f_op != &inotify_fops)) {
723                 ret = -EINVAL;
724                 goto fput_and_out;
725         }
726 
727         if (!(mask & IN_DONT_FOLLOW))
728                 flags |= LOOKUP_FOLLOW;
729         if (mask & IN_ONLYDIR)
730                 flags |= LOOKUP_DIRECTORY;
731 
732         ret = inotify_find_inode(pathname, &path, flags);
733         if (ret)
734                 goto fput_and_out;
735 
736         /* inode held in place by reference to path; group by fget on fd */
737         inode = path.dentry->d_inode;
738         group = f.file->private_data;
739 
740         /* create/update an inode mark */
741         ret = inotify_update_watch(group, inode, mask);
742         path_put(&path);
743 fput_and_out:
744         fdput(f);
745         return ret;
746 }
747 
748 SYSCALL_DEFINE2(inotify_rm_watch, int, fd, __s32, wd)
749 {
750         struct fsnotify_group *group;
751         struct inotify_inode_mark *i_mark;
752         struct fd f;
753         int ret = 0;
754 
755         f = fdget(fd);
756         if (unlikely(!f.file))
757                 return -EBADF;
758 
759         /* verify that this is indeed an inotify instance */
760         ret = -EINVAL;
761         if (unlikely(f.file->f_op != &inotify_fops))
762                 goto out;
763 
764         group = f.file->private_data;
765 
766         ret = -EINVAL;
767         i_mark = inotify_idr_find(group, wd);
768         if (unlikely(!i_mark))
769                 goto out;
770 
771         ret = 0;
772 
773         fsnotify_destroy_mark(&i_mark->fsn_mark, group);
774 
775         /* match ref taken by inotify_idr_find */
776         fsnotify_put_mark(&i_mark->fsn_mark);
777 
778 out:
779         fdput(f);
780         return ret;
781 }
782 
783 /*
784  * inotify_user_setup - Our initialization function.  Note that we cannot return
785  * error because we have compiled-in VFS hooks.  So an (unlikely) failure here
786  * must result in panic().
787  */
788 static int __init inotify_user_setup(void)
789 {
790         BUILD_BUG_ON(IN_ACCESS != FS_ACCESS);
791         BUILD_BUG_ON(IN_MODIFY != FS_MODIFY);
792         BUILD_BUG_ON(IN_ATTRIB != FS_ATTRIB);
793         BUILD_BUG_ON(IN_CLOSE_WRITE != FS_CLOSE_WRITE);
794         BUILD_BUG_ON(IN_CLOSE_NOWRITE != FS_CLOSE_NOWRITE);
795         BUILD_BUG_ON(IN_OPEN != FS_OPEN);
796         BUILD_BUG_ON(IN_MOVED_FROM != FS_MOVED_FROM);
797         BUILD_BUG_ON(IN_MOVED_TO != FS_MOVED_TO);
798         BUILD_BUG_ON(IN_CREATE != FS_CREATE);
799         BUILD_BUG_ON(IN_DELETE != FS_DELETE);
800         BUILD_BUG_ON(IN_DELETE_SELF != FS_DELETE_SELF);
801         BUILD_BUG_ON(IN_MOVE_SELF != FS_MOVE_SELF);
802         BUILD_BUG_ON(IN_UNMOUNT != FS_UNMOUNT);
803         BUILD_BUG_ON(IN_Q_OVERFLOW != FS_Q_OVERFLOW);
804         BUILD_BUG_ON(IN_IGNORED != FS_IN_IGNORED);
805         BUILD_BUG_ON(IN_EXCL_UNLINK != FS_EXCL_UNLINK);
806         BUILD_BUG_ON(IN_ISDIR != FS_ISDIR);
807         BUILD_BUG_ON(IN_ONESHOT != FS_IN_ONESHOT);
808 
809         BUG_ON(hweight32(ALL_INOTIFY_BITS) != 21);
810 
811         inotify_inode_mark_cachep = KMEM_CACHE(inotify_inode_mark, SLAB_PANIC);
812 
813         inotify_max_queued_events = 16384;
814         init_user_ns.ucount_max[UCOUNT_INOTIFY_INSTANCES] = 128;
815         init_user_ns.ucount_max[UCOUNT_INOTIFY_WATCHES] = 8192;
816 
817         return 0;
818 }
819 fs_initcall(inotify_user_setup);
820 

~ [ 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