inotify.7: Add text on dealing with rename() events

[thirdparty/man-pages.git] / man7 / inotify.7

'\" t
.\" Copyright (C) 2006, 2014 Michael Kerrisk <mtk.manpages@gmail.com>
.\"
.\" %%%LICENSE_START(VERBATIM)
.\" Permission is granted to make and distribute verbatim copies of this
.\" manual provided the copyright notice and this permission notice are
.\" preserved on all copies.
.\"
.\" Permission is granted to copy and distribute modified versions of this
.\" manual under the conditions for verbatim copying, provided that the
.\" entire resulting derived work is distributed under the terms of a
.\" permission notice identical to this one.
.\"
.\" Since the Linux kernel and libraries are constantly changing, this
.\" manual page may be incorrect or out-of-date.  The author(s) assume no
.\" responsibility for errors or omissions, or for damages resulting from
.\" the use of the information contained herein.  The author(s) may not
.\" have taken the same level of care in the production of this manual,
.\" which is licensed free of charge, as they might when working
.\" professionally.
.\"
.\" Formatted or processed versions of this manual, if unaccompanied by
.\" the source, must acknowledge the copyright and authors of this work.
.\" %%%LICENSE_END
.\"
.TH INOTIFY 7 2014-04-01 "Linux" "Linux Programmer's Manual"
.SH NAME
inotify \- monitoring filesystem events
.SH DESCRIPTION
The
.I inotify
API provides a mechanism for monitoring filesystem events.
Inotify can be used to monitor individual files,
or to monitor directories.
When a directory is monitored, inotify will return events
for the directory itself, and for files inside the directory.

The following system calls are used with this API:
.IP * 3
.BR inotify_init (2)
creates an inotify instance and returns a file descriptor
referring to the inotify instance.
The more recent
.BR inotify_init1 (2)
is like
.BR inotify_init (2),
but has a
.IR flags
argument that provides access to some extra functionality.
.IP *
.BR inotify_add_watch (2)
manipulates the "watch list" associated with an inotify instance.
Each item ("watch") in the watch list specifies the pathname of
a file or directory,
along with some set of events that the kernel should monitor for the
file referred to by that pathname.
.BR inotify_add_watch (2)
either creates a new watch item, or modifies an existing watch.
Each watch has a unique "watch descriptor", an integer
returned by
.BR inotify_add_watch (2)
when the watch is created.
.IP *
When events occur for monitored files and directories,
those events are made available to the application as structured data that
can be read from the inotify file descriptor using
.BR read (2)
(see below).
.IP *
.BR inotify_rm_watch (2)
removes an item from an inotify watch list.
.IP *
When all file descriptors referring to an inotify
instance have been closed (using
.BR close (2)),
the underlying object and its resources are
freed for reuse by the kernel;
all associated watches are automatically freed.
.SS Reading events from an inotify file descriptor
To determine what events have occurred, an application
.BR read (2)s
from the inotify file descriptor.
If no events have so far occurred, then,
assuming a blocking file descriptor,
.BR read (2)
will block until at least one event occurs
(unless interrupted by a signal,
in which case the call fails with the error
.BR EINTR ;
see
.BR signal (7)).

Each successful
.BR read (2)
returns a buffer containing one or more of the following structures:
.in +4n
.nf

struct inotify_event {
    int      wd;       /* Watch descriptor */
.\" FIXME . The type of the 'wd' field should probably be "int32_t".
.\" I submitted a patch to fix this.  See the LKML thread
.\" "[patch] Fix type errors in inotify interfaces", 18 Nov 2008
.\" Glibc bug filed: http://sources.redhat.com/bugzilla/show_bug.cgi?id=7040
    uint32_t mask;     /* Mask of events */
    uint32_t cookie;   /* Unique cookie associating related
                          events (for rename(2)) */
    uint32_t len;      /* Size of \fIname\fP field */
    char     name[];   /* Optional null-terminated name */
};
.fi
.in

.I wd
identifies the watch for which this event occurs.
It is one of the watch descriptors returned by a previous call to
.BR inotify_add_watch (2).

.I mask
contains bits that describe the event that occurred (see below).

.I cookie
is a unique integer that connects related events.
Currently this is used only for rename events, and
allows the resulting pair of
.B IN_MOVED_FROM
and
.B IN_MOVED_TO
events to be connected by the application.
For all other event types,
.I cookie
is set to 0.

The
.I name
field is present only when an event is returned
for a file inside a watched directory;
it identifies the file pathname relative to the watched directory.
This pathname is null-terminated,
and may include further null bytes (\(aq\\0\(aq) to align subsequent reads to a
suitable address boundary.

The
.I len
field counts all of the bytes in
.IR name ,
including the null bytes;
the length of each
.I inotify_event
structure is thus
.IR "sizeof(struct inotify_event)+len" .

The behavior when the buffer given to
.BR read (2)
is too small to return information about the next event depends
on the kernel version: in kernels before 2.6.21,
.BR read (2)
returns 0; since kernel 2.6.21,
.BR read (2)
fails with the error
.BR EINVAL .
Specifying a buffer of size

    sizeof(struct inotify_event) + NAME_MAX + 1

will be sufficient to read at least one event.
.SS inotify events
The
.BR inotify_add_watch (2)
.I mask
argument and the
.I mask
field of the
.I inotify_event
structure returned when
.BR read (2)ing
an inotify file descriptor are both bit masks identifying
inotify events.
The following bits can be specified in
.I mask
when calling
.BR inotify_add_watch (2)
and may be returned in the
.I mask
field returned by
.BR read (2):
.RS 4
.TP
.BR IN_ACCESS " (*)"
File was accessed (e.g.,
.BR read (2),
.BR execve (2)).
.TP
.BR IN_ATTRIB " (*)"
Metadata changed\(emfor example, permissions (e.g.,
.BR chmod (2)),
timestamps (e.g.,
.BR utimensat (2)),
extended attributes
.RB ( setxattr (2)),
link count (since Linux 2.6.25; e.g.,
for the target of
.BR link (2)
and for
.BR unlink (2)),
and user/group ID (e.g.,
.BR chown (2)).
.TP
.BR IN_CLOSE_WRITE " (*)"
File opened for writing was closed.
.TP
.BR IN_CLOSE_NOWRITE " (*)"
File not opened for writing was closed.
.TP
.BR IN_CREATE " (*)"
File/directory created in watched directory (e.g.,
.BR open (2)
.BR O_CREAT ,
.BR mkdir (2),
.BR link (2),
.BR symlink (2),
.BR bind (2)
on a UNIX domain socket).
.TP
.BR IN_DELETE " (*)"
File/directory deleted from watched directory.
.TP
.B IN_DELETE_SELF
Watched file/directory was itself deleted.
(This event also occurs if an object is moved to another filesystem,
since
.BR mv (1)
in effect copies the file to the other filesystem and
then deletes it from the original filesystem.)
In addition, an
.B IN_IGNORED
event will subsequently be generated for the watch descriptor.
.TP
.BR IN_MODIFY " (*)"
File was modified (e.g.,
.BR write (2),
.BR truncate (2)).
.TP
.B IN_MOVE_SELF
Watched file/directory was itself moved.
.TP
.BR IN_MOVED_FROM " (*)"
Generated for the directory containing the old filename
when a file is renamed.
.TP
.BR IN_MOVED_TO " (*)"
Generated for the directory containing the new filename
when a file is renamed.
.TP
.BR IN_OPEN " (*)"
File was opened.
.RE
.PP
When monitoring a directory,
the events marked with an asterisk (*) above can occur for
files in the directory, in which case the
.I name
field in the returned
.I inotify_event
structure identifies the name of the file within the directory.
.PP
The
.B IN_ALL_EVENTS
macro is defined as a bit mask of all of the above events.
This macro can be used as the
.I mask
argument when calling
.BR inotify_add_watch (2).

Two additional convenience macros are defined:
.RS 4
.TP
.BR IN_MOVE
Equates to
.BR "IN_MOVED_FROM | IN_MOVED_TO" .
.TP
.BR IN_CLOSE
Equates to
.BR "IN_CLOSE_WRITE | IN_CLOSE_NOWRITE" .
.RE
.PP
The following further bits can be specified in
.I mask
when calling
.BR inotify_add_watch (2):
.RS 4
.TP
.BR IN_DONT_FOLLOW " (since Linux 2.6.15)"
Don't dereference
.I pathname
if it is a symbolic link.
.TP
.BR IN_EXCL_UNLINK " (since Linux 2.6.36)"
.\" commit 8c1934c8d70b22ca8333b216aec6c7d09fdbd6a6
By default, when watching events on the children of a directory,
events are generated for children even after they have been unlinked
from the directory.
This can result in large numbers of uninteresting events for
some applications (e.g., if watching
.IR /tmp ,
in which many applications create temporary files whose
names are immediately unlinked).
Specifying
.B IN_EXCL_UNLINK
changes the default behavior,
so that events are not generated for children after
they have been unlinked from the watched directory.
.TP
.B IN_MASK_ADD
Add (OR) events to watch mask for this pathname if
it already exists (instead of replacing mask).
.TP
.B IN_ONESHOT
Monitor
.I pathname
for one event, then remove from
watch list.
.TP
.BR IN_ONLYDIR " (since Linux 2.6.15)"
Only watch
.I pathname
if it is a directory.
.RE
.PP
The following bits may be set in the
.I mask
field returned by
.BR read (2):
.RS 4
.TP
.B IN_IGNORED
Watch was removed explicitly
.RB ( inotify_rm_watch (2))
or automatically (file was deleted, or filesystem was unmounted).
See also BUGS.
.TP
.B IN_ISDIR
Subject of this event is a directory.
.TP
.B IN_Q_OVERFLOW
Event queue overflowed
.RI ( wd
is \-1 for this event).
.TP
.B IN_UNMOUNT
Filesystem containing watched object was unmounted.
In addition, an
.B IN_IGNORED
event will subsequently be generated for the watch descriptor.
.RE
.SS Examples
Suppose an application is watching the directory
.I dir
and the file
.IR dir/myfile
for all events.
The examples below show some events that will be generated
for these two objects.
.RS 4
.TP
fd = open("dir/myfile", O_RDWR);
Generates
.B IN_OPEN
events for both
.I dir
and
.IR dir/myfile .
.TP
read(fd, buf, count);
Generates
.B IN_ACCESS
events for both
.I dir
and
.IR dir/myfile .
.TP
write(fd, buf, count);
Generates
.B IN_MODIFY
events for both
.I dir
and
.IR dir/myfile .
.TP
fchmod(fd, mode);
Generates
.B IN_ATTRIB
events for both
.I dir
and
.IR dir/myfile .
.TP
close(fd);
Generates
.B IN_CLOSE_WRITE
events for both
.I dir
and
.IR dir/myfile .
.RE
.PP
Suppose an application is watching the directories
.I dir1
and
.IR dir ,
and the file
.IR dir1/myfile .
The following examples show some events that may be generated.
.RS 4
.TP
link("dir/myfile", "dir2/new");
Generates an
.B IN_ATTRIB
event for
.IR myfile
and an
.B IN_CREATE
event for
.IR dir2 .
.TP
rename("dir1/myfile", "dir2/myfile");
Generates an
.B IN_MOVED_FROM
event for
.IR dir1 ,
an
.B IN_MOVED_TO
event for
.IR dir2 ,
and an
.B IN_MOVE_SELF
event for
.IR myfile .
The
.B IN_MOVED_FROM
and
.B IN_MOVED_TO
events will have the same
.I cookie
value.
.RE
.PP
Suppose that
.IR dir1/xx
and
.IR dir2/yy
are (the only) links to the same file, and an application is watching
.IR dir1 ,
.IR dir2 ,
.IR dir1/xx ,
and
.IR dir2/yy .
Executing the following calls in the order given below will generate
the following events:
.RS 4
.TP
unlink("dir2/yy");
Generates
.BR IN_ATTRIB
event for
.IR xx
(because its link count changes)
and an
.B IN_DELETE
event for
.IR dir2 .
.TP
unlink("dir1/xx");
Generates
.BR IN_ATTRIB ,
.BR IN_DELETE_SELF ,
and
.BR IN_IGNORED
events for
.IR xx ,
and an
.BR IN_DELETE
for
.IR dir1 .
.RE
.PP
Suppose an application is watching the directory
.IR dir
and (the empty) directory
.IR dir/subdir .
The following examples show some events that may be generated.
.RS 4
.TP
mkdir("dir/new", mode);
Generates an
.B "IN_CREATE | IN_ISDIR"
event for
.IR dir .
.TP
rmdir("dir/sub");
Generates
.B IN_DELETE_SELF
and
.B IN_IGNORED
events for
.IR subdir ,
and an
.B "IN_DELETE | IN_ISDIR"
event for
.IR dir .
.RE
.SS /proc interfaces
The following interfaces can be used to limit the amount of
kernel memory consumed by inotify:
.TP
.I /proc/sys/fs/inotify/max_queued_events
The value in this file is used when an application calls
.BR inotify_init (2)
to set an upper limit on the number of events that can be
queued to the corresponding inotify instance.
Events in excess of this limit are dropped, but an
.B IN_Q_OVERFLOW
event is always generated.
.TP
.I /proc/sys/fs/inotify/max_user_instances
This specifies an upper limit on the number of inotify instances
that can be created per real user ID.
.TP
.I /proc/sys/fs/inotify/max_user_watches
This specifies an upper limit on the number of watches
that can be created per real user ID.
.SH VERSIONS
Inotify was merged into the 2.6.13 Linux kernel.
The required library interfaces were added to glibc in version 2.4.
.RB ( IN_DONT_FOLLOW ,
.BR IN_MASK_ADD ,
and
.B IN_ONLYDIR
were added in glibc version 2.5.)
.SH CONFORMING TO
The inotify API is Linux-specific.
.SH NOTES
Inotify file descriptors can be monitored using
.BR select (2),
.BR poll (2),
and
.BR epoll (7).
When an event is available, the file descriptor indicates as readable.

Since Linux 2.6.25,
signal-driven I/O notification is available for inotify file descriptors;
see the discussion of
.B F_SETFL
(for setting the
.B O_ASYNC
flag),
.BR F_SETOWN ,
and
.B F_SETSIG
in
.BR fcntl (2).
The
.I siginfo_t
structure (described in
.BR sigaction (2))
that is passed to the signal handler has the following fields set:
.IR si_fd
is set to the inotify file descriptor number;
.IR si_signo
is set to the signal number;
.IR si_code
is set to
.BR POLL_IN ;
and
.B POLLIN
is set in
.IR si_band .

If successive output inotify events produced on the
inotify file descriptor are identical (same
.IR wd ,
.IR mask ,
.IR cookie ,
and
.IR name ),
then they are coalesced into a single event if the
older event has not yet been read (but see BUGS).
This reduces the amount of kernel memory required for the event queue,
but also means that an application can't use inotify to reliably count
file events.

The events returned by reading from an inotify file descriptor
form an ordered queue.
Thus, for example, it is guaranteed that when renaming from
one directory to another, events will be produced in the
correct order on the inotify file descriptor.

The
.B FIONREAD
.BR ioctl (2)
returns the number of bytes available to read from an
inotify file descriptor.
.SS Limitations and caveats
The inotify applications identifies events via watch descriptors.
It is the application's responsibility to cache a mapping
(if one is needed) between watch descriptors and pathnames.
Be aware that directory renamings may affect multiple cached pathnames.

Inotify monitoring of directories is not recursive:
to monitor subdirectories under a directory,
additional watches must be created.
This can take a significant amount time for large directory trees.

If monitoring an entire directory subtree,
and a new subdirectory is created in that tree or an existing directory
is renamed into that tree,
be aware that by the time you create a watch for the new subdirectory,
new files (and subdirectories) may already exist inside the subdirectory.
Therefore, you may want to scan the contents of the subdirectory
immediately after adding the watch (and, if desired,
recursively add watches for any subdirectories that it contains).

The inotify API provides no information about the user or process that
triggered the inotify event.
In particular, there is no easy
way for a process that is monitoring events via inotify
to distinguish events that it triggers
itself from those that are triggered by other processes.

Note that the event queue can overflow.
In this case, events are lost.
Robust applications should handle the possibility of
lost events gracefully.
For example, it may be necessary to rebuild part or all of
the application cache.
(One simple, but possibly expensive,
approach is to close the inotify file descriptor, empty the cache,
create a new inotify file descriptor,
and then re-create watches and cache entries
for the objects to be monitored.)

The inotify API identifies affected files by filename.
However, by the time an application processes an inotify event,
the filename may already have been deleted or renamed.
.SS Dealing with rename() events
The
.B IN_MOVED_FROM
and
.B IN_MOVED_TO
events that are generated by
.BR rename (2)
are usually available as consecutive events when reading
from the inotify file descriptor.
However, this is not guaranteed.
If multiple processes are triggering events for monitored objects,
then (on rare occasions) an arbitrary number of
other events may appear between the
.B IN_MOVED_FROM
and
.B IN_MOVED_TO
events.

Matching up the
.B IN_MOVED_FROM
and
.B IN_MOVED_TO
event pair generated by
.BR rename (2)
is thus inherently racy.
(Don't forget that if an object is renamed outside of a monitored directory,
there may not even be an
.BR IN_MOVED_TO
event.)
Heuristic approaches (e.g., assume the events are always consecutive)
can be used to ensure a match in most cases,
but will inevitably miss some cases,
causing the application to perceive the
.B IN_MOVED_FROM
and
.B IN_MOVED_TO
events as being unrelated.
If watch descriptors are destroyed and re-created as a result,
then those watch descriptors will be inconsistent with
the watch descriptors in any pending events.
(Rebuilding the cache and re-creating the inotify file descriptor may
be useful to deal with this scenario.)

Applications should also allow for the possibility that the
.B IN_MOVED_FROM
event was the last event that could fit in the buffer
returned by the current call to
.BR read (2),
and the accompanying
.B IN_MOVED_TO
event might be fetched only on the next
.BR read (2).
.SH BUGS
.\" FIXME kernel commit 611da04f7a31b2208e838be55a42c7a1310ae321
.\" implies that unmount events were buggy 2.6.11 to 2.6.36
.\"
In kernels before 2.6.16, the
.B IN_ONESHOT
.I mask
flag does not work.

As originally designed and implemented, the
.B IN_ONESHOT
flag did not cause an
.B IN_IGNORED
event to be generated when the watch was dropped after one event.
However, as an unintended effect of other changes,
since Linux 2.6.36, an
.B IN_IGNORED
event is generated in this case.

Before kernel 2.6.25,
.\" commit 1c17d18e3775485bf1e0ce79575eb637a94494a2
the kernel code that was intended to coalesce successive identical events
(i.e., the two most recent events could potentially be coalesced
if the older had not yet been read)
instead checked if the most recent event could be coalesced with the
.I oldest
unread event.
.SH SEE ALSO
.BR inotifywait (1),
.BR inotifywatch (1),
.BR inotify_add_watch (2),
.BR inotify_init (2),
.BR inotify_init1 (2),
.BR inotify_rm_watch (2),
.BR read (2),
.BR stat (2)

.IR Documentation/filesystems/inotify.txt
in the Linux kernel source tree