]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - share/man/man9/netisr.9
Remove $FreeBSD$: two-line nroff pattern
[FreeBSD/FreeBSD.git] / share / man / man9 / netisr.9
1 .\"
2 .\" Copyright (c) 2009 Robert N. M. Watson
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(s), this list of conditions and the following disclaimer as
10 .\"    the first lines of this file unmodified other than the possible
11 .\"    addition of one or more copyright notices.
12 .\" 2. Redistributions in binary form must reproduce the above copyright
13 .\"    notice(s), this list of conditions and the following disclaimer in the
14 .\"    documentation and/or other materials provided with the distribution.
15 .\"
16 .\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY
17 .\" EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
18 .\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
19 .\" DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY
20 .\" DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21 .\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
22 .\" SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
23 .\" 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 SUCH
26 .\" DAMAGE.
27 .\"
28 .Dd April 12, 2023
29 .Dt NETISR 9
30 .Os
31 .Sh NAME
32 .Nm netisr
33 .Nd Kernel network dispatch service
34 .Sh SYNOPSIS
35 .In net/netisr.h
36 .Ft void
37 .Fn netisr_register "const struct netisr_handler *nhp"
38 .Ft void
39 .Fn netisr_unregister "const struct netisr_handler *nhp"
40 .Ft int
41 .Fn netisr_dispatch "u_int proto" "struct mbuf *m"
42 .Ft int
43 .Fn netisr_dispatch_src "u_int proto" "uintptr_t source" "struct mbuf *m"
44 .Ft int
45 .Fn netisr_queue "u_int proto" "struct mbuf *m"
46 .Ft int
47 .Fn netisr_queue_src "u_int proto" "uintptr_t source" "struct mbuf *m"
48 .Ft void
49 .Fn netisr_clearqdrops "const struct netisr_handler *nhp"
50 .Ft void
51 .Fn netisr_getqdrops "const struct netisr_handler *nhp" "uint64_t *qdropsp"
52 .Ft void
53 .Fn netisr_getqlimit "const struct netisr_handler *nhp" "u_int *qlimitp"
54 .Ft int
55 .Fn netisr_setqlimit "const struct netisr_handler *nhp" "u_int qlimit"
56 .Ft u_int
57 .Fn netisr_default_flow2cpu "u_int flowid"
58 .Ft u_int
59 .Fn netisr_get_cpucount "void"
60 .Ft u_int
61 .Fn netisr_get_cpuid "u_int cpunumber"
62 .Pp
63 With optional virtual network stack support enabled via the following kernel
64 compile option:
65 .Bd -ragged -offset indent
66 .Cd "options VIMAGE"
67 .Ed
68 .Ft void
69 .Fn netisr_register_vnet "const struct netisr_handler *nhp"
70 .Ft void
71 .Fn netisr_unregister_vnet "const struct netisr_handler *nhp"
72 .Sh DESCRIPTION
73 The
74 .Nm
75 kernel interface suite allows device drivers (and other packet sources) to
76 direct packets to protocols for directly dispatched or deferred processing.
77 Protocol registration and work stream statistics may be monitored using
78 .Xr netstat 1 .
79 .Ss Protocol registration
80 Protocols register and unregister handlers using
81 .Fn netisr_register
82 and
83 .Fn netisr_unregister ,
84 and may also manage queue limits and statistics using the
85 .Fn netisr_clearqdrops ,
86 .Fn netisr_getqdrops ,
87 .Fn netisr_getqlimit ,
88 and
89 .Fn netisr_setqlimit .
90 .Pp
91 In case of VIMAGE kernels each virtual network stack (vnet), that is not the
92 default base system network stack, calls
93 .Fn netisr_register_vnet
94 and
95 .Fn netisr_unregister_vnet
96 to enable or disable packet processing by the
97 .Nm
98 for each protocol.
99 Disabling will also purge any outstanding packet from the protocol queue.
100 .Pp
101 .Nm
102 supports multi-processor execution of handlers, and relies on a combination
103 of source ordering and protocol-specific ordering and work-placement
104 policies to decide how to distribute work across one or more worker
105 threads.
106 Registering protocols will declare one of three policies:
107 .Bl -tag -width NETISR_POLICY_SOURCE
108 .It Dv NETISR_POLICY_SOURCE
109 .Nm
110 should maintain source ordering without advice from the protocol.
111 .Nm
112 will ignore any flow IDs present on
113 .Vt mbuf
114 headers for the purposes of work placement.
115 .It Dv NETISR_POLICY_FLOW
116 .Nm
117 should maintain flow ordering as defined by the
118 .Vt mbuf
119 header flow ID field.
120 If the protocol implements
121 .Va nh_m2flow ,
122 then
123 .Nm
124 will query the protocol in the event that the
125 .Vt mbuf
126 doesn't have a flow ID, falling back on source ordering.
127 .It NETISR_POLICY_CPU
128 .Nm
129 will entirely delegate all work placement decisions to the protocol,
130 querying
131 .Va nh_m2cpuid
132 for each packet.
133 .El
134 .Pp
135 Registration is declared using
136 .Vt "struct netisr_handler" ,
137 whose fields are defined as follows:
138 .Bl -tag -width "netisr_handler_t nh_handler"
139 .It Vt "const char *" Va nh_name
140 Unique character string name of the protocol, which may be included in
141 .Xr sysctl 3
142 MIB names, so should not contain whitespace.
143 .It Vt netisr_handler_t Va nh_handler
144 Protocol handler function that will be invoked on each packet received for
145 the protocol.
146 .It Vt netisr_m2flow_t Va nh_m2flow
147 Optional protocol function to generate a flow ID and set a valid
148 hashtype for packets that enter the
149 .Nm
150 with
151 .Dv M_HASHTYPE_GET(m)
152 equal to
153 .Dv M_HASHTYPE_NONE .
154 Will be used only with
155 .Dv NETISR_POLICY_FLOW .
156 .It Vt netisr_m2cpuid_t Va nh_m2cpuid
157 Protocol function to determine what CPU a packet should be processed on.
158 Will be used only with
159 .Dv NETISR_POLICY_CPU .
160 .It Vt netisr_drainedcpu_t Va nh_drainedcpu
161 Optional callback function that will be invoked when a per-CPU queue
162 was drained.
163 It will never fire for directly dispatched packets.
164 Unless fully understood, this special-purpose function should not be used.
165 .\" In case you intend to use this please send 50 chocolate bars to each
166 .\" of rwatson and bz and wait for an answer.
167 .It Vt u_int Va nh_proto
168 Protocol number used by both protocols to identify themselves to
169 .Nm ,
170 and by packet sources to select what handler will be used to process
171 packets.
172 A table of supported protocol numbers appears below.
173 For implementation reasons, protocol numbers great than 15 are currently
174 unsupported.
175 .It Vt u_int Va nh_qlimit
176 The maximum per-CPU queue depth for the protocol; due to internal
177 implementation details, the effective queue depth may be as much as twice
178 this number.
179 .It Vt u_int Va nh_policy
180 The ordering and work placement policy for the protocol, as described
181 earlier.
182 .El
183 .Ss Packet source interface
184 Packet sources, such as network interfaces, may request protocol processing
185 using the
186 .Fn netisr_dispatch
187 and
188 .Fn netisr_queue
189 interfaces.
190 Both accept a protocol number and
191 .Vt mbuf
192 argument, but while
193 .Fn netisr_queue
194 will always execute the protocol handler asynchronously in a deferred
195 context,
196 .Fn netisr_dispatch
197 will optionally direct dispatch if permitted by global and per-protocol
198 policy.
199 .Pp
200 In order to provide additional load balancing and flow information,
201 packet sources may also specify an opaque source identifier, which in
202 practice might be a network interface number or socket pointer, using
203 the
204 .Fn netisr_dispatch_src
205 and
206 .Fn netisr_queue_src
207 variants.
208 .Ss Protocol number constants
209 The follow protocol numbers are currently defined:
210 .Bl -tag -width NETISR_ROUTE
211 .It Dv NETISR_IP
212 IPv4
213 .It Dv NETISR_IGMP
214 IGMPv3 loopback
215 .It Dv NETISR_ROUTE
216 Routing socket loopback
217 .It Dv NETISR_ARP
218 ARP
219 .It Dv NETISR_IPV6
220 IPv6
221 .El
222 .Sh AUTHORS
223 This manual page and the
224 .Nm
225 implementation were written by
226 .An Robert N. M. Watson .