xref: /freebsd/share/man/man4/lagg.4 (revision 7899f917b1c0ea178f1d2be0cfb452086d079d23)
1.\"	$OpenBSD: trunk.4,v 1.18 2006/06/09 13:53:34 jmc Exp $
2.\"
3.\" Copyright (c) 2005, 2006 Reyk Floeter <reyk@openbsd.org>
4.\"
5.\" Permission to use, copy, modify, and distribute this software for any
6.\" purpose with or without fee is hereby granted, provided that the above
7.\" copyright notice and this permission notice appear in all copies.
8.\"
9.\" THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10.\" WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11.\" MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12.\" ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13.\" WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14.\" ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15.\" OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
16.\"
17.Dd January 16, 2023
18.Dt LAGG 4
19.Os
20.Sh NAME
21.Nm lagg
22.Nd link aggregation and link failover interface
23.Sh SYNOPSIS
24To compile this driver into the kernel,
25place the following line in your
26kernel configuration file:
27.Bd -ragged -offset indent
28.Cd "device lagg"
29.Ed
30.Pp
31Alternatively, to load the driver as a
32module at boot time, place the following line in
33.Xr loader.conf 5 :
34.Bd -literal -offset indent
35if_lagg_load="YES"
36.Ed
37.Sh DESCRIPTION
38The
39.Nm
40interface allows aggregation of multiple network interfaces as one virtual
41.Nm
42interface for the purpose of providing fault-tolerance and high-speed links.
43.Pp
44Each
45.Nm
46interface is created at runtime using interface cloning.
47This is
48most easily done with the
49.Xr ifconfig 8
50.Cm create
51command or using the
52.Va cloned_interfaces
53variable in
54.Xr rc.conf 5 .
55.Pp
56A
57.Nm
58interface can be created using the
59.Ic ifconfig lagg Ns Ar N Ic create
60command.
61It can use different link aggregation protocols specified
62using the
63.Ic laggproto Ar proto
64option.
65Child interfaces can be added using the
66.Ic laggport Ar child-iface
67option and removed using the
68.Ic -laggport Ar child-iface
69option.
70.Pp
71The driver currently supports the aggregation protocols
72.Ic failover
73(the default),
74.Ic lacp ,
75.Ic loadbalance ,
76.Ic roundrobin ,
77.Ic broadcast ,
78and
79.Ic none .
80The protocols determine which ports are used for outgoing traffic
81and whether a specific port accepts incoming traffic.
82The interface link state is used to validate if the port is active or
83not.
84.Bl -tag -width loadbalance
85.It Ic failover
86Sends traffic only through the active port.
87If the master port becomes unavailable,
88the next active port is used.
89The first interface added is the master port;
90any interfaces added after that are used as failover devices.
91.Pp
92By default, received traffic is only accepted when it is received
93through the active port.
94This constraint can be relaxed by setting the
95.Va net.link.lagg.failover_rx_all
96.Xr sysctl 8
97variable to a nonzero value,
98which is useful for certain bridged network setups.
99.It Ic lacp
100Supports the IEEE 802.1AX (formerly 802.3ad) Link Aggregation Control Protocol
101(LACP) and the Marker Protocol.
102LACP will negotiate a set of aggregable links with the peer in to one or more
103Link Aggregated Groups.
104Each LAG is composed of ports of the same speed, set to full-duplex operation.
105The traffic will be balanced across the ports in the LAG with the greatest
106total speed, in most cases there will only be one LAG which contains all ports.
107In the event of changes in physical connectivity, Link Aggregation will quickly
108converge to a new configuration.
109.It Ic loadbalance
110Balances outgoing traffic across the active ports based on hashed
111protocol header information and accepts incoming traffic from
112any active port.
113This is a static setup and does not negotiate aggregation with the peer or
114exchange frames to monitor the link.
115The hash includes the Ethernet source and destination address, and, if
116available, the VLAN tag, and the IP source and destination address.
117.It Ic roundrobin
118Distributes outgoing traffic using a round-robin scheduler
119through all active ports and accepts incoming traffic from
120any active port.
121Using
122.Ic roundrobin
123mode can cause unordered packet arrival at the client.
124Throughput might be limited as the client performs CPU-intensive packet
125reordering.
126.It Ic broadcast
127Sends frames to all ports of the LAG and receives frames on
128any port of the LAG.
129.It Ic none
130This protocol is intended to do nothing: it disables any traffic without
131disabling the
132.Nm
133interface itself.
134.El
135.Pp
136The MTU of the first interface to be added is used as the lagg MTU.
137All additional interfaces are required to have exactly the same value.
138.Pp
139The
140.Ic loadbalance
141and
142.Ic lacp
143modes will use the RSS hash from the network card if available to avoid
144computing one, this may give poor traffic distribution if the hash is invalid
145or uses less of the protocol header information.
146Local hash computation can be forced per interface by setting the
147.Cm -use_flowid
148.Xr ifconfig 8
149flag.
150The default for new interfaces is set via the
151.Va net.link.lagg.default_use_flowid
152.Xr sysctl 8 .
153.Pp
154When creating a
155.Nm
156interface, the
157.Ic laggtype
158can be specified as either
159.Cm ethernet
160or
161.Cm infiniband .
162If neither is specified then the default is
163.Cm ethernet .
164.Sh EXAMPLES
165Create a link aggregation using LACP with two
166.Xr bge 4
167Gigabit Ethernet interfaces:
168.Bd -literal -offset indent
169# ifconfig bge0 up
170# ifconfig bge1 up
171# ifconfig lagg0 create
172# ifconfig lagg0 laggproto lacp laggport bge0 laggport bge1 \e
173	192.168.1.1 netmask 255.255.255.0
174.Ed
175.Pp
176Create a link aggregation using ROUNDROBIN with two
177.Xr bge 4
178Gigabit Ethernet interfaces and set a stride of 500 packets
179per interface:
180.Bd -literal -offset indent
181# ifconfig bge0 up
182# ifconfig bge1 up
183# ifconfig lagg0 create
184# ifconfig lagg0 laggproto roundrobin laggport bge0 laggport bge1 \e
185	192.168.1.1 netmask 255.255.255.0
186# ifconfig lagg0 rr_limit 500
187.Ed
188.Pp
189The following example uses an active failover interface to set up roaming
190between wired and wireless networks using two network devices.
191Whenever the wired master interface is unplugged, the wireless failover
192device will be used:
193.Bd -literal -offset indent
194# ifconfig em0 ether 00:11:22:33:44:55 up
195# ifconfig wlan0 create wlandev ath0 ssid my_net up
196# ifconfig lagg0 create
197# ifconfig lagg0 laggproto failover laggport em0 laggport wlan0 \e
198	192.168.1.1 netmask 255.255.255.0
199.Ed
200.Pp
201(Note the MAC address of the wired device is forced to match that of the
202wireless device,
203.Sq 00:11:22:33:44:55
204in this example, as some common wireless devices will not allow MAC
205addresses to be changed.)
206.Pp
207The following example shows how to create an infiniband failover interface.
208.Bd -literal -offset indent
209# ifconfig ib0 up
210# ifconfig ib1 up
211# ifconfig lagg0 create laggtype infiniband
212# ifconfig lagg0 laggproto failover laggport ib0 laggport ib1 \e
213	1.1.1.1 netmask 255.255.255.0
214.Ed
215.Pp
216Configure two ethernets for failover with static IP in
217.Pa /etc/rc.conf :
218.Bd -literal -offset indent
219cloned_interfaces="lagg0"
220ifconfig_lagg0="laggproto failover laggport bge0 laggport bge1 \e
221	10.1.29.21/24"
222ifconfig_bge0="up"
223ifconfig_bge1="up"
224.Ed
225.Sh SEE ALSO
226.Xr ng_one2many 4 ,
227.Xr rc.conf 5 ,
228.Xr ifconfig 8 ,
229.Xr sysctl 8
230.Sh HISTORY
231The
232.Nm
233device first appeared in
234.Fx 6.3 .
235.Sh AUTHORS
236.An -nosplit
237The
238.Nm
239driver was written under the name
240.Nm trunk
241by
242.An Reyk Floeter Aq Mt reyk@openbsd.org .
243The LACP implementation was written by
244.An YAMAMOTO Takashi
245for
246.Nx .
247.Sh BUGS
248There is no way to configure LACP administrative variables, including system
249and port priorities.
250The current implementation always performs active-mode LACP and uses 0x8000 as
251system and port priorities.
252