xref: /netbsd/share/man/man9/softintr.9 (revision 6550d01e)
1.\"	$NetBSD: softintr.9,v 1.20 2009/08/03 23:29:19 rmind Exp $
2.\"
3.\" Copyright (c) 2007, 2008 The NetBSD Foundation, Inc.
4.\" All rights reserved.
5.\"
6.\" This code is derived from software contributed to The NetBSD Foundation
7.\" by Andrew Doran.
8.\"
9.\" Redistribution and use in source and binary forms, with or without
10.\" modification, are permitted provided that the following conditions
11.\" are met:
12.\" 1. Redistributions of source code must retain the above copyright
13.\"    notice, this list of conditions and the following disclaimer.
14.\" 2. Redistributions in binary form must reproduce the above copyright
15.\"    notice, this list of conditions and the following disclaimer in the
16.\"    documentation and/or other materials provided with the distribution.
17.\"
18.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
19.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
20.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
21.\" PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
22.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
28.\" POSSIBILITY OF SUCH DAMAGE.
29.\"
30.\" Copyright (c) 2000 Christopher G. Demetriou.
31.\" All rights reserved.
32.\"
33.\" Redistribution and use in source and binary forms, with or without
34.\" modification, are permitted provided that the following conditions
35.\" are met:
36.\" 1. Redistributions of source code must retain the above copyright
37.\"    notice, this list of conditions and the following disclaimer.
38.\" 2. Redistributions in binary form must reproduce the above copyright
39.\"    notice, this list of conditions and the following disclaimer in the
40.\"    documentation and/or other materials provided with the distribution.
41.\" 3. All advertising materials mentioning features or use of this software
42.\"    must display the following acknowledgement:
43.\"          This product includes software developed for the
44.\"          NetBSD Project.  See http://www.NetBSD.org/ for
45.\"          information about NetBSD.
46.\" 4. The name of the author may not be used to endorse or promote products
47.\"    derived from this software without specific prior written permission.
48.\"
49.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
50.\" IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
51.\" OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
52.\" IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
53.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
54.\" NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
55.\" DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
56.\" THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
57.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
58.\" THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
59.\"
60.\" --(license Id: LICENSE.proto,v 1.1 2000/06/13 21:40:26 cgd Exp )--
61.\"
62.Dd August 3, 2009
63.Dt SOFTINT 9
64.Os
65.Sh NAME
66.Nm softint ,
67.Nm softint_establish ,
68.Nm softint_disestablish ,
69.Nm softint_schedule
70.Nd machine-independent software interrupt framework
71.Sh SYNOPSIS
72.In sys/intr.h
73.Ft void *
74.Fn softint_establish "u_int flags" "void (*func)(void *)" "void *arg"
75.Ft void
76.Fn softint_disestablish "void *cookie"
77.Ft void
78.Fn softint_schedule "void *cookie"
79.Sh DESCRIPTION
80The software interrupt framework is designed to provide
81a generic software interrupt mechanism which can be used any time a
82low-priority callback is needed.
83.Pp
84It allows dynamic registration of software interrupts for loadable
85drivers and protocol stacks, prioritization and fair queueing of software
86interrupts, and allows machine-dependent optimizations to reduce cost.
87.Pp
88Four priority levels are provided.
89In order of priority (lowest to highest) the levels are: clock, bio,
90net, serial.
91The names are symbolic and in isolation do not have any direct
92connection with a particular kind of device activity: they are
93only meant as a guide.
94.Pp
95The four priority levels map directly to scheduler priority
96levels, and where the architecture implements
97.Dq fast
98software interrupts, they also map onto interrupt priorities.
99The interrupt priorities are intended to be hidden from machine
100independent code, which should in general use thread-safe mechanisms
101to synchronize with software interrupts (for example: mutexes).
102.Pp
103Software interrupts run with limited machine context.
104In particular, they do not possess any address space context.
105They should not try to operate on user space addresses, or to use
106virtual memory facilities other than those noted as interrupt
107safe.
108Unlike hardware interrupts, software interrupts do have thread
109context.
110They may block on synchronization objects, sleep, and resume
111execution at a later time.
112.Pp
113Since software interrupts are a limited resource and run with
114higher priority than most other LWPs in the system, all
115block-and-resume activity by a software interrupt must be kept
116short to allow further processing at that level to continue.
117By extension, code running with process context must take care to
118ensure that any lock that may be taken from a software interrupt
119can not be held for more than a short period of time.
120.Pp
121The kernel does not allow software interrupts to use facilities
122or perform actions that are likely to block for a significant
123amount of time.
124This means that it's not valid for a software interrupt to
125sleep on condition variables or to wait for resources to
126become available (for example, memory).
127.Pp
128The following is a brief description of each function in the framework:
129.Bl -tag -width abcxdcc
130.It Fn softint_establish flags func arg
131.Pp
132Register a software interrupt.
133The
134.Fa flags
135value must contain one of the following constants, specifing
136the priority level for the soft interrupt:
137.Pp
138.Dv SOFTINT_CLOCK ,
139.Dv SOFTINT_BIO ,
140.Dv SOFTINT_NET ,
141.Dv SOFTINT_SERIAL
142.Pp
143If the constant
144.Dv SOFTINT_MPSAFE
145is not logically ORed into
146.Fa flags ,
147the global
148.Dv kernel_lock
149will automatically be acquired before the soft interrupt handler
150is called.
151.Pp
152The constant
153.Fa func
154specifies the function to call when the soft interrupt is
155executed.
156The argument
157.Fa arg
158will be passed to this function.
159.Pp
160.Fn softint_establish
161may block in order to allocate memory.
162If successful, it returns a
163.Pf non- Dv NULL
164opaque value to be used as an argument to
165.Fn softint_schedule
166and/or
167.Fn softint_disestablish .
168If for some reason it does not succeed, it returns
169.Dv NULL .
170.It Fn softint_disestablish cookie
171.Pp
172Deallocate a software interrupt previously allocated
173by a call to
174.Fn softint_establish .
175.\" XXX What happens to pending scheduled calls?
176.It Fn softint_schedule cookie
177.Pp
178Schedule a software interrupt previously allocated
179by a call to
180.Fn softint_establish
181to be executed as soon as that software interrupt is unblocked.
182.Fn softint_schedule
183can safely be called multiple times before the
184callback routine is invoked.
185.Pp
186Soft interrupt scheduling is CPU-local.
187A request to dispatch a soft interrupt will only be serviced on
188the same CPU where the request was made.
189The LWPs (light weight processes) dedicated to soft interrupt
190processing are bound to their home CPUs, so if a soft interrupt
191handler sleeps and later resumes, it will always resume on the
192same CPU.
193.Pp
194On a system with multiple processors, multiple instances of
195the same soft interrupt handler can be in flight simultaneously
196(at most one per-CPU).
197.El
198.Sh SEE ALSO
199.Xr callout 9 ,
200.Xr condvar 9 ,
201.Xr kthread 9 ,
202.Xr mutex 9 ,
203.Xr rwlock 9 ,
204.Xr spl 9 ,
205.Xr workqueue 9
206.Sh HISTORY
207The
208.Nx
209machine-independent software interrupt framework was designed in 1997
210and was implemented by one port in
211.Nx 1.3 .
212However, it did not gain wider implementation until
213.Nx 1.5 .
214Between
215.Nx 4.0
216and
217.Nx 5.0
218the framework was re-implemented in a machine-independant way to
219provide software interrupts with thread context.
220