2 .\" Hey Emacs! This file is -*- nroff -*- source.
4 .\" Copyright (C) 2006 Michael Kerrisk <mtk.manpages@gmail.com>
6 .\" Permission is granted to make and distribute verbatim copies of this
7 .\" manual provided the copyright notice and this permission notice are
8 .\" preserved on all copies.
10 .\" Permission is granted to copy and distribute modified versions of this
11 .\" manual under the conditions for verbatim copying, provided that the
12 .\" entire resulting derived work is distributed under the terms of a
13 .\" permission notice identical to this one.
15 .\" Since the Linux kernel and libraries are constantly changing, this
16 .\" manual page may be incorrect or out-of-date. The author(s) assume no
17 .\" responsibility for errors or omissions, or for damages resulting from
18 .\" the use of the information contained herein. The author(s) may not
19 .\" have taken the same level of care in the production of this manual,
20 .\" which is licensed free of charge, as they might when working
23 .\" Formatted or processed versions of this manual, if unaccompanied by
24 .\" the source, must acknowledge the copyright and authors of this work.
26 .\" FIXME 2.6.36 adds IN_EXCL_UNLINK
28 .TH INOTIFY 7 2010-09-27 "Linux" "Linux Programmer's Manual"
30 inotify \- monitoring file system events
34 API provides a mechanism for monitoring file system events.
35 Inotify can be used to monitor individual files,
36 or to monitor directories.
37 When a directory is monitored, inotify will return events
38 for the directory itself, and for files inside the directory.
40 The following system calls are used with this API:
43 .BR inotify_init1 (2)),
44 .BR inotify_add_watch (2),
45 .BR inotify_rm_watch (2),
51 creates an inotify instance and returns a file descriptor
52 referring to the inotify instance.
57 but provides some extra functionality.
59 .BR inotify_add_watch (2)
60 manipulates the "watch list" associated with an inotify instance.
61 Each item ("watch") in the watch list specifies the pathname of
63 along with some set of events that the kernel should monitor for the
64 file referred to by that pathname.
65 .BR inotify_add_watch (2)
66 either creates a new watch item, or modifies an existing watch.
67 Each watch has a unique "watch descriptor", an integer
69 .BR inotify_add_watch (2)
70 when the watch is created.
72 .BR inotify_rm_watch (2)
73 removes an item from an inotify watch list.
75 When all file descriptors referring to an inotify
76 instance have been closed,
77 the underlying object and its resources are
78 freed for reuse by the kernel;
79 all associated watches are automatically freed.
81 To determine what events have occurred, an application
83 from the inotify file descriptor.
84 If no events have so far occurred, then,
85 assuming a blocking file descriptor,
87 will block until at least one event occurs
88 (unless interrupted by a signal,
89 in which case the call fails with the error
96 returns a buffer containing one or more of the following structures:
100 struct inotify_event {
101 int wd; /* Watch descriptor */
102 .\" FIXME . The type of the 'wd' field should probably be "int32_t".
103 .\" I submitted a patch to fix this. See the LKML thread
104 .\" "[patch] Fix type errors in inotify interfaces", 18 Nov 2008
105 .\" Glibc bug filed: http://sources.redhat.com/bugzilla/show_bug.cgi?id=7040
106 uint32_t mask; /* Mask of events */
107 uint32_t cookie; /* Unique cookie associating related
108 events (for rename(2)) */
109 uint32_t len; /* Size of \fIname\fP field */
110 char name[]; /* Optional null-terminated name */
116 identifies the watch for which this event occurs.
117 It is one of the watch descriptors returned by a previous call to
118 .BR inotify_add_watch (2).
121 contains bits that describe the event that occurred (see below).
124 is a unique integer that connects related events.
125 Currently this is only used for rename events, and
126 allows the resulting pair of
130 events to be connected by the application.
134 field is only present when an event is returned
135 for a file inside a watched directory;
136 it identifies the file pathname relative to the watched directory.
137 This pathname is null-terminated,
138 and may include further null bytes to align subsequent reads to a
139 suitable address boundary.
143 field counts all of the bytes in
145 including the null bytes;
149 .IR "sizeof(inotify_event)+len" .
151 The behavior when the buffer given to
153 is too small to return information about the next event depends
154 on the kernel version: in kernels before 2.6.21,
156 returns 0; since kernel 2.6.21,
162 .BR inotify_add_watch (2)
168 structure returned when
170 an inotify file descriptor are both bit masks identifying
172 The following bits can be specified in
175 .BR inotify_add_watch (2)
176 and may be returned in the
185 File was accessed (read) (*).
188 Metadata changed, e.g., permissions, timestamps, extended attributes,
189 link count (since Linux 2.6.25), UID, GID, etc. (*).
192 File opened for writing was closed (*).
195 File not opened for writing was closed (*).
198 File/directory created in watched directory (*).
201 File/directory deleted from watched directory (*).
204 Watched file/directory was itself deleted.
207 File was modified (*).
210 Watched file/directory was itself moved.
213 File moved out of watched directory (*).
216 File moved into watched directory (*).
223 When monitoring a directory,
224 the events marked with an asterisk (*) above can occur for
225 files in the directory, in which case the
227 field in the returned
229 structure identifies the name of the file within the directory.
233 macro is defined as a bit mask of all of the above events.
234 This macro can be used as the
236 argument when calling
237 .BR inotify_add_watch (2).
239 Two additional convenience macros are
242 IN_MOVED_FROM|IN_MOVED_TO,
246 IN_CLOSE_WRITE|IN_CLOSE_NOWRITE.
248 The following further bits can be specified in
251 .BR inotify_add_watch (2):
256 .BR IN_DONT_FOLLOW " (since Linux 2.6.15)"
257 Don't dereference \fIpathname\fP if it is a symbolic link.
260 Add (OR) events to watch mask for this pathname if
261 it already exists (instead of replacing mask).
264 Monitor \fIpathname\fP for one event, then remove from
267 .BR IN_ONLYDIR " (since Linux 2.6.15)"
268 Only watch \fIpathname\fP if it is a directory.
272 The following bits may be set in the
281 Watch was removed explicitly (\fBinotify_rm_watch\fP(2))
282 or automatically (file was deleted, or file system was unmounted).
285 Subject of this event is a directory.
288 Event queue overflowed (\fIwd\fP is \-1 for this event).
291 File system containing watched object was unmounted.
295 The following interfaces can be used to limit the amount of
296 kernel memory consumed by inotify:
298 .I /proc/sys/fs/inotify/max_queued_events
299 The value in this file is used when an application calls
301 to set an upper limit on the number of events that can be
302 queued to the corresponding inotify instance.
303 Events in excess of this limit are dropped, but an
305 event is always generated.
307 .I /proc/sys/fs/inotify/max_user_instances
308 This specifies an upper limit on the number of inotify instances
309 that can be created per real user ID.
311 .I /proc/sys/fs/inotify/max_user_watches
312 This specifies an upper limit on the number of watches
313 that can be created per real user ID.
315 Inotify was merged into the 2.6.13 Linux kernel.
316 The required library interfaces were added to glibc in version 2.4.
317 .RB ( IN_DONT_FOLLOW ,
321 were only added in version 2.5.)
323 The inotify API is Linux-specific.
325 Inotify file descriptors can be monitored using
330 When an event is available, the file descriptor indicates as readable.
333 signal-driven I/O notification is available for inotify file descriptors;
334 see the discussion of
346 structure (described in
348 that is passed to the signal handler has the following fields set:
350 is set to the inotify file descriptor number;
352 is set to the signal number;
361 If successive output inotify events produced on the
362 inotify file descriptor are identical (same
368 then they are coalesced into a single event if the
369 older event has not yet been read (but see BUGS).
371 The events returned by reading from an inotify file descriptor
372 form an ordered queue.
373 Thus, for example, it is guaranteed that when renaming from
374 one directory to another, events will be produced in the
375 correct order on the inotify file descriptor.
380 returns the number of bytes available to read from an
381 inotify file descriptor.
382 .SS Limitations and caveats
383 Inotify monitoring of directories is not recursive:
384 to monitor subdirectories under a directory,
385 additional watches must be created.
386 This can take a significant amount time for large directory trees.
388 The inotify API provides no information about the user or process that
389 triggered the inotify event.
391 Note that the event queue can overflow.
392 In this case, events are lost.
393 Robust applications should handle the possibility of
394 lost events gracefully.
396 The inotify API identifies affected files by filename.
397 However, by the time an application processes an inotify event,
398 the filename may already have been deleted or renamed.
400 If monitoring an entire directory subtree,
401 and a new subdirectory is created in that tree,
402 be aware that by the time you create a watch for the new subdirectory,
403 new files may already have been created in the subdirectory.
404 Therefore, you may want to scan the contents of the subdirectory
405 immediately after adding the watch.
407 In kernels before 2.6.16, the
412 Before kernel 2.6.25,
413 the kernel code that was intended to coalesce successive identical events
414 (i.e., the two most recent events could potentially be coalesced
415 if the older had not yet been read)
416 instead checked if the most recent event could be coalesced with the
420 .BR inotify_add_watch (2),
421 .BR inotify_init (2),
422 .BR inotify_init1 (2),
423 .BR inotify_rm_watch (2),
426 .IR Documentation/filesystems/inotify.txt .