aic(4): Remove ISA attachment.
[dragonfly.git] / share / man / man4 / bridge.4
1 .\" Copyright 2001 Wasabi Systems, Inc.
2 .\" All rights reserved.
3 .\"
4 .\" Written by Jason R. Thorpe for Wasabi Systems, Inc.
5 .\" Spanning tree modifications by Matthew Dillon
6 .\"
7 .\" Redistribution and use in source and binary forms, with or without
8 .\" modification, are permitted provided that the following conditions
9 .\" are met:
10 .\" 1. Redistributions of source code must retain the above copyright
11 .\"    notice, this list of conditions and the following disclaimer.
12 .\" 2. Redistributions in binary form must reproduce the above copyright
13 .\"    notice, this list of conditions and the following disclaimer in the
14 .\"    documentation and/or other materials provided with the distribution.
15 .\" 3. All advertising materials mentioning features or use of this software
16 .\"    must display the following acknowledgement:
17 .\"     This product includes software developed for the NetBSD Project by
18 .\"     Wasabi Systems, Inc.
19 .\" 4. The name of Wasabi Systems, Inc. may not be used to endorse
20 .\"    or promote products derived from this software without specific prior
21 .\"    written permission.
22 .\"
23 .\" THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
24 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
25 .\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
26 .\" PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
27 .\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
28 .\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
29 .\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
30 .\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
31 .\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
32 .\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
33 .\" POSSIBILITY OF SUCH DAMAGE.
34 .\"
35 .\"
36 .Dd February 22, 2011
37 .Dt BRIDGE 4
38 .Os
39 .Sh NAME
40 .Nm bridge
41 .Nd network bridge device
42 .Sh SYNOPSIS
43 .Cd "pseudo-device bridge"
44 .Sh DESCRIPTION
45 The
46 .Nm
47 driver creates a logical link between two or more IEEE 802 networks
48 that use the same (or
49 .Dq similar enough )
50 framing format.
51 For example, it is possible to bridge Ethernet and 802.11 networks together,
52 but it is not possible to bridge Ethernet and Token Ring together.
53 .Pp
54 To use
55 .Nm ,
56 the administrator must first create the interface and configure
57 the bridge parameters.
58 The bridge is created using the
59 .Xr ifconfig 8
60 .Cm create
61 subcommand.
62 See the
63 .Xr ifconfig 8
64 manual page for further information on configuring bridges.
65 .Pp
66 A bridge can be used to provide several services, such as a simple
67 802.11-to-Ethernet bridge for wireless hosts, and traffic isolation.
68 .Pp
69 A bridge works like a hub, forwarding traffic from one interface
70 to another.
71 Multicast and broadcast packets are always forwarded to all
72 interfaces that are part of the bridge.
73 For unicast traffic, the bridge learns which MAC addresses are associated
74 with which interfaces and will forward the traffic selectively.
75 .Pp
76 The bridge operates in a safe mode by default, setting the MAC source in
77 the link header on outgoing packets to the outgoing interface MAC.
78 This reduces the chance that the layer-2 switching in your switches
79 will become confused.
80 .Pp
81 The bridge supports various special features via
82 .Cm link
83 options.
84 .Bl -tag -width indent
85 .It Cm link0
86 The link0 option enables transparent bridging mode.
87 The bridge will make every effort to retain the ethernet header
88 when forwarding packets between interfaces, making the bridging
89 function work more like a hardware bridge device.
90 .It Cm link1
91 The link1 option enables keepalive transmission and automatically
92 places a member into a special blocked mode if no keepalive reception
93 occurs.
94 If either sides of the link uses this option then both sides must use
95 this option.
96 This option is impemented by sending CFG updates on the hello interval
97 to the remote.
98 The link is considered lost after 10 intervals (typically 20 seconds).
99 .It Cm link2
100 The link2 option enables channel bonding (see also ifbondweight).
101 All member interfaces with the same mac address are considered to
102 be in a bonding group.
103 When something like
104 .Xr tap 4
105 is used, you can manually control or copy the mac to create bonding groups.
106 When interface bonding is enabled normally blocked interfaces belonging
107 to the same bonding group as an active forwarding interface will be
108 changed to the bonding state.
109 Both sides of link the member represents must operate in bonding mode
110 for this to work, otherwise the remote end may decide to throw away
111 half your packets.
112 .El
113 .Pp
114 If your network becomes glitchy, with long pauses in tcp sessions, then
115 transparent bridging mode is likely the cause.  This mode should only be
116 used when you are bridging networks with devices that do MAC-based security
117 or firewalling (for example, the supremely braindead at&t uverse router),
118 or which impose severe limitations on MAC:IP assignments.
119 .Pp
120 If member interfaces constantly enter a 'blocked (link1)' state then the
121 other end of those interfaces is not implementing the link1 keepalive.
122 Both sides must implement the keepalive.
123 .Pp
124 If you get an enormous amount of packet loss and are using link2-based
125 bonding, then the other side of those member interfaces are probably
126 not implementing link2-based bonding.
127 .Pp
128 The
129 .Nm
130 driver implements the IEEE 802.1D Spanning Tree protocol (STP).
131 Spanning Tree is used to detect and remove loops in a network topology.
132 .Pp
133 Packet filtering can be used with any firewall package that hooks in via the
134 .Xr pfil 9
135 framework.
136 When filtering is enabled, bridged packets will pass through the filter
137 inbound on the originating interface, on the bridge interface and outbound on
138 the appropriate interfaces.
139 Either stage can be disabled, this behaviour can be controlled using
140 .Xr sysctl 8 :
141 Set
142 .Va net.link.bridge.pfil_member
143 to
144 .Li 1
145 to enable filtering on the incoming and outgoing member interfaces
146 and set
147 .Va net.link.bridge.pfil_bridge
148 to
149 .Li 1
150 to enable filtering on the bridge interface.
151 .Pp
152 ARP and REVARP packets are forwarded without being filtered and others
153 that are not IP nor IPv6 packets are not forwarded when filtering is
154 enabled.
155 .Pp
156 Note that packets to and from the bridging host will be seen by the
157 filter on the interface with the appropriate address configured as well
158 as on the interface on which the packet arrives or departs.
159 .Pp
160 The MTU of the first member interface to be added is used as the bridge MTU,
161 all additional members are required to have exactly the same value.
162 .Sh EXTRA FEATURES
163 .Dx
164 implements two additional features to make spanning tree operation more
165 resilient.
166 .Pp
167 Specifying
168 .Cm link0
169 on the bridge interface places the bridge in transparent bridging mode.
170 The bridge will make every attempt to retain the original source MAC in
171 the ethernet link header.
172 .Pp
173 Specifying
174 .Cm link1
175 on the bridge interface forces the bridge to generate a 802.11d CFG
176 message on every hello interval for all interfaces participating
177 in the STP protocol.
178 Normally CFG messages are only generated by the root bridge interface
179 or during topology changes.
180 In addition the bridge code expects to receive 802.11d frames from
181 all interface participating in the STP protocol.
182 .Pp
183 An interface which fails to receive a 802.11d frame within 10 times
184 the hello interval (usually 20 seconds) automatically goes into
185 l1blocking mode, which can be observed in the ifconfig output for
186 the bridge.  This removes the interface from consideration and the
187 bridge code automatically routes around it.
188 .Pp
189 Using
190 .Cm link0
191 and
192 .Cm link1
193 together between two
194 .Dx
195 boxes allows you to maintain multiple parallel vpns between those
196 boxes via different networks (if you happen to be on more than one
197 with internet access).
198 Use separate openvpn instances and tap devices for each vpn link
199 to accomplish this, placing them in the same bridge interface on
200 the two endpoints.
201 The tap devices do not need any IP configuration when bridged and
202 can be assigned the same ether MAC (in fact they have to be
203 if you want the failover to work nicely).
204 .Sh SEE ALSO
205 .Xr pf 4 ,
206 .Xr ifconfig 8
207 .Sh HISTORY
208 The
209 .Nm
210 driver first appeared in
211 .Ox 2.5
212 and found its way into
213 .Dx 1.3 .
214 Transparent bridging (link0) was added in
215 .Dx 2.9
216 in 2011.
217 .Sh AUTHORS
218 .An -nosplit
219 The
220 .Nm
221 driver was originally written by
222 .An Jason L. Wright
223 .Aq jason@thought.net
224 as part of an undergraduate independent study at the University of
225 North Carolina at Greensboro.
226 .Pp
227 This version of the
228 .Nm
229 driver has been heavily modified from the original version by
230 .An Jason R. Thorpe
231 .Aq thorpej@wasabisystems.com .
232 .Sh BUGS
233 The
234 .Nm
235 driver currently supports only Ethernet and Ethernet-like (e.g. 802.11)
236 network devices, with exactly the same interface MTU size as the bridge device.