xref: /freebsd/share/man/man4/aio.4 (revision 8ddb146abcdf061be9f2c0db7e391697dafad85c)
1.\"-
2.\" Copyright (c) 2002 Dag-Erling Coïdan Smørgrav
3.\" All rights reserved.
4.\"
5.\" Redistribution and use in source and binary forms, with or without
6.\" modification, are permitted provided that the following conditions
7.\" are met:
8.\" 1. Redistributions of source code must retain the above copyright
9.\"    notice, this list of conditions and the following disclaimer.
10.\" 2. Redistributions in binary form must reproduce the above copyright
11.\"    notice, this list of conditions and the following disclaimer in the
12.\"    documentation and/or other materials provided with the distribution.
13.\" 3. The name of the author may not be used to endorse or promote products
14.\"    derived from this software without specific prior written permission.
15.\"
16.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26.\" SUCH DAMAGE.
27.\"
28.\" $FreeBSD$
29.\"
30.Dd January 2, 2021
31.Dt AIO 4
32.Os
33.Sh NAME
34.Nm aio
35.Nd asynchronous I/O
36.Sh DESCRIPTION
37The
38.Nm
39facility provides system calls for asynchronous I/O.
40Asynchronous I/O operations are not completed synchronously by the
41calling thread.
42Instead, the calling thread invokes one system call to request an
43asynchronous I/O operation.
44The status of a completed request is retrieved later via a separate
45system call.
46.Pp
47Asynchronous I/O operations on some file descriptor types may block an
48AIO daemon indefinitely resulting in process and/or system hangs.
49Operations on these file descriptor types are considered
50.Dq unsafe
51and disabled by default.
52They can be enabled by setting
53the
54.Va vfs.aio.enable_unsafe
55sysctl node to a non-zero value.
56.Pp
57Asynchronous I/O operations on sockets,
58raw disk devices,
59and regular files on local filesystems do not block
60indefinitely and are always enabled.
61.Pp
62The
63.Nm
64facility uses kernel processes
65(also known as AIO daemons)
66to service most asynchronous I/O requests.
67These processes are grouped into pools containing a variable number of
68processes.
69Each pool will add or remove processes to the pool based on load.
70Pools can be configured by sysctl nodes that define the minimum
71and maximum number of processes as well as the amount of time an idle
72process will wait before exiting.
73.Pp
74One pool of AIO daemons is used to service asynchronous I/O requests for
75sockets.
76These processes are named
77.Dq soaiod<N> .
78The following sysctl nodes are used with this pool:
79.Bl -tag -width indent
80.It Va kern.ipc.aio.num_procs
81The current number of processes in the pool.
82.It Va kern.ipc.aio.target_procs
83The minimum number of processes that should be present in the pool.
84.It Va kern.ipc.aio.max_procs
85The maximum number of processes permitted in the pool.
86.It Va kern.ipc.aio.lifetime
87The amount of time a process is permitted to idle in clock ticks.
88If a process is idle for this amount of time and there are more processes
89in the pool than the target minimum,
90the process will exit.
91.El
92.Pp
93A second pool of AIO daemons is used to service all other asynchronous I/O
94requests except for I/O requests to raw disks.
95These processes are named
96.Dq aiod<N> .
97The following sysctl nodes are used with this pool:
98.Bl -tag -width indent
99.It Va vfs.aio.num_aio_procs
100The current number of processes in the pool.
101.It Va vfs.aio.target_aio_procs
102The minimum number of processes that should be present in the pool.
103.It Va vfs.aio.max_aio_procs
104The maximum number of processes permitted in the pool.
105.It Va vfs.aio.aiod_lifetime
106The amount of time a process is permitted to idle in clock ticks.
107If a process is idle for this amount of time and there are more processes
108in the pool than the target minimum,
109the process will exit.
110.El
111.Pp
112Asynchronous I/O requests for raw disks are queued directly to the disk
113device layer after temporarily wiring the user pages associated with the
114request.
115These requests are not serviced by any of the AIO daemon pools.
116.Pp
117Several limits on the number of asynchronous I/O requests are imposed both
118system-wide and per-process.
119These limits are configured via the following sysctls:
120.Bl -tag -width indent
121.It Va vfs.aio.max_buf_aio
122The maximum number of queued asynchronous I/O requests for raw disks permitted
123for a single process.
124Asynchronous I/O requests that have completed but whose status has not been
125retrieved via
126.Xr aio_return 2
127or
128.Xr aio_waitcomplete 2
129are not counted against this limit.
130.It Va vfs.aio.num_buf_aio
131The number of queued asynchronous I/O requests for raw disks system-wide.
132.It Va vfs.aio.max_aio_queue_per_proc
133The maximum number of asynchronous I/O requests for a single process
134serviced concurrently by the default AIO daemon pool.
135.It Va vfs.aio.max_aio_per_proc
136The maximum number of outstanding asynchronous I/O requests permitted for a
137single process.
138This includes requests that have not been serviced,
139requests currently being serviced,
140and requests that have completed but whose status has not been retrieved via
141.Xr aio_return 2
142or
143.Xr aio_waitcomplete 2 .
144.It Va vfs.aio.num_queue_count
145The number of outstanding asynchronous I/O requests system-wide.
146.It Va vfs.aio.max_aio_queue
147The maximum number of outstanding asynchronous I/O requests permitted
148system-wide.
149.El
150.Pp
151Asynchronous I/O control buffers should be zeroed before initializing
152individual fields.
153This ensures all fields are initialized.
154.Pp
155All asynchronous I/O control buffers contain a
156.Vt sigevent
157structure in the
158.Va aio_sigevent
159field which can be used to request notification when an operation completes.
160.Pp
161For
162.Dv SIGEV_KEVENT
163notifications,
164the
165.Va sigevent
166.Ap
167s
168.Va sigev_notify_kqueue
169field should contain the descriptor of the kqueue that the event should be attached
170to, its
171.Va sigev_notify_kevent_flags
172field may contain
173.Dv EV_ONESHOT ,
174.Dv EV_CLEAR , and/or
175.Dv EV_DISPATCH , and its
176.Va sigev_notify
177field should be set to
178.Dv SIGEV_KEVENT .
179The posted kevent will contain:
180.Bl -column ".Va filter"
181.It Sy Member Ta Sy Value
182.It Va ident Ta asynchronous I/O control buffer pointer
183.It Va filter Ta Dv EVFILT_AIO
184.It Va flags Ta Dv EV_EOF
185.It Va udata Ta
186value stored in
187.Va aio_sigevent.sigev_value
188.El
189.Pp
190For
191.Dv SIGEV_SIGNO
192and
193.Dv SIGEV_THREAD_ID
194notifications,
195the information for the queued signal will include
196.Dv SI_ASYNCIO
197in the
198.Va si_code
199field and the value stored in
200.Va sigevent.sigev_value
201in the
202.Va si_value
203field.
204.Pp
205For
206.Dv SIGEV_THREAD
207notifications,
208the value stored in
209.Va aio_sigevent.sigev_value
210is passed to the
211.Va aio_sigevent.sigev_notify_function
212as described in
213.Xr sigevent 3 .
214.Sh SEE ALSO
215.Xr aio_cancel 2 ,
216.Xr aio_error 2 ,
217.Xr aio_read 2 ,
218.Xr aio_readv 2 ,
219.Xr aio_return 2 ,
220.Xr aio_suspend 2 ,
221.Xr aio_waitcomplete 2 ,
222.Xr aio_write 2 ,
223.Xr aio_writev 2 ,
224.Xr lio_listio 2 ,
225.Xr sigevent 3 ,
226.Xr sysctl 8
227.Sh HISTORY
228The
229.Nm
230facility appeared as a kernel option in
231.Fx 3.0 .
232The
233.Nm
234kernel module appeared in
235.Fx 5.0 .
236The
237.Nm
238facility was integrated into all kernels in
239.Fx 11.0 .
240