xref: /freebsd/share/man/man4/cxgbe.4 (revision ea825d02749f382c3f7e17f28247f20a48733eab)
1.\" Copyright (c) 2011-2016, Chelsio Inc
2.\" All rights reserved.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions are met:
6.\"
7.\" 1. Redistributions of source code must retain the above copyright notice,
8.\"    this list of conditions and the following disclaimer.
9.\"
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.\"
14.\" 3. Neither the name of the Chelsio Inc nor the names of its
15.\"    contributors may be used to endorse or promote products derived from
16.\"    this software without specific prior written permission.
17.\"
18.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
19.\" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21.\" ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
22.\" 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.\" * Other names and brands may be claimed as the property of others.
31.\"
32.\" $FreeBSD$
33.\"
34.Dd May 16, 2017
35.Dt CXGBE 4
36.Os
37.Sh NAME
38.Nm cxgbe
39.Nd "Chelsio T4-, T5-, and T6-based 100Gb, 40Gb, 25Gb, 10Gb, and 1Gb Ethernet adapter driver"
40.Sh SYNOPSIS
41To compile this driver into the kernel,
42place the following lines in your
43kernel configuration file:
44.Bd -ragged -offset indent
45.Cd "device cxgbe"
46.Ed
47.Pp
48To load the driver as a
49module at boot time, place the following lines in
50.Xr loader.conf 5 :
51.Bd -literal -offset indent
52t4fw_cfg_load="YES"
53t5fw_cfg_load="YES"
54t6fw_cfg_load="YES"
55if_cxgbe_load="YES"
56.Ed
57.Sh DESCRIPTION
58The
59.Nm
60driver provides support for PCI Express Ethernet adapters based on
61the Chelsio Terminator 4, Terminator 5, and Terminator 6 ASICs (T4, T5, and T6).
62The driver supports Jumbo Frames, Transmit/Receive checksum offload,
63TCP segmentation offload (TSO), Large Receive Offload (LRO), VLAN
64tag insertion/extraction, VLAN checksum offload, VLAN TSO, and
65Receive Side Steering (RSS).
66For further hardware information and questions related to hardware
67requirements, see
68.Pa http://www.chelsio.com/ .
69.Pp
70The
71.Nm
72driver uses different names for devices based on the associated ASIC:
73.Bl -column -offset indent "ASIC" "Port Name" "Parent Device"
74.It Sy ASIC Ta Sy Port Name Ta Sy Parent Device Ta Sy Virtual Interface
75.It T4 Ta cxgbe Ta t4nex Ta vcxgbe
76.It T5 Ta cxl Ta t5nex Ta vcxl
77.It T6 Ta cc Ta t6nex Ta vcc
78.El
79.Pp
80Loader tunables with the hw.cxgbe prefix apply to all cards.
81The driver provides sysctl MIBs for both ports and parent devices using
82the names above.
83For example, a T5 adapter provides port MIBs under dev.cxl and
84adapter-wide MIBs under dev.t5nex.
85References to sysctl MIBs in the remainder of this page use
86dev.<port> for port MIBs and dev.<nexus> for adapter-wide MIBs.
87.Pp
88For more information on configuring this device, see
89.Xr ifconfig 8 .
90.Sh HARDWARE
91The
92.Nm
93driver supports 100Gb and 25Gb Ethernet adapters based on the T6 ASIC:
94.Pp
95.Bl -bullet -compact
96.It
97Chelsio T6225-CR
98.It
99Chelsio T6225-SO-CR
100.It
101Chelsio T62100-LP-CR
102.It
103Chelsio T62100-SO-CR
104.It
105Chelsio T62100-CR
106.El
107.Pp
108The
109.Nm
110driver supports 40Gb, 10Gb and 1Gb Ethernet adapters based on the T5 ASIC:
111.Pp
112.Bl -bullet -compact
113.It
114Chelsio T580-CR
115.It
116Chelsio T580-LP-CR
117.It
118Chelsio T580-LP-SO-CR
119.It
120Chelsio T560-CR
121.It
122Chelsio T540-CR
123.It
124Chelsio T540-LP-CR
125.It
126Chelsio T522-CR
127.It
128Chelsio T520-LL-CR
129.It
130Chelsio T520-CR
131.It
132Chelsio T520-SO
133.It
134Chelsio T520-BT
135.It
136Chelsio T504-BT
137.El
138.Pp
139The
140.Nm
141driver supports 10Gb and 1Gb Ethernet adapters based on the T4 ASIC:
142.Pp
143.Bl -bullet -compact
144.It
145Chelsio T420-CR
146.It
147Chelsio T422-CR
148.It
149Chelsio T440-CR
150.It
151Chelsio T420-BCH
152.It
153Chelsio T440-BCH
154.It
155Chelsio T440-CH
156.It
157Chelsio T420-SO
158.It
159Chelsio T420-CX
160.It
161Chelsio T420-BT
162.It
163Chelsio T404-BT
164.El
165.Sh LOADER TUNABLES
166Tunables can be set at the
167.Xr loader 8
168prompt before booting the kernel or stored in
169.Xr loader.conf 5 .
170There are multiple tunables that control the number of queues of various
171types.
172A negative value for such a tunable instructs the driver to create
173up to that many queues if there are enough CPU cores available.
174.Bl -tag -width indent
175.It Va hw.cxgbe.ntxq10g
176Number of tx queues used for a 10Gb or higher-speed port.
177The default is 16 or the number
178of CPU cores in the system, whichever is less.
179.It Va hw.cxgbe.nrxq10g
180Number of rx queues used for a 10Gb or higher-speed port.
181The default is 8 or the number
182of CPU cores in the system, whichever is less.
183.It Va hw.cxgbe.ntxq1g
184Number of tx queues used for a 1Gb port.
185The default is 4 or the number
186of CPU cores in the system, whichever is less.
187.It Va hw.cxgbe.nrxq1g
188Number of rx queues used for a 1Gb port.
189The default is 2 or the number
190of CPU cores in the system, whichever is less.
191.It Va hw.cxgbe.nofldtxq10g
192Number of TOE tx queues used for a 10Gb or higher-speed port.
193The default is 8 or the
194number of CPU cores in the system, whichever is less.
195.It Va hw.cxgbe.nofldrxq10g
196Number of TOE rx queues used for a 10Gb or higher-speed port.
197The default is 2 or the
198number of CPU cores in the system, whichever is less.
199.It Va hw.cxgbe.nofldtxq1g
200Number of TOE tx queues used for a 1Gb port.
201The default is 2 or the
202number of CPU cores in the system, whichever is less.
203.It Va hw.cxgbe.nofldrxq1g
204Number of TOE rx queues used for a 1Gb port.
205The default is 1.
206.It Va hw.cxgbe.num_vis
207Number of virtual interfaces (VIs) created for each port.
208Each virtual interface creates a separate network interface.
209The first virtual interface on each port is required and represents
210the primary network interface on the port.
211Additional virtual interfaces on a port are named using the Virtual Interface
212name from the table above.
213Additional virtual interfaces use a single pair of queues
214for rx and tx as well an additional pair of queues for TOE rx and tx.
215The default is 1.
216.It Va hw.cxgbe.holdoff_timer_idx_10G
217.It Va hw.cxgbe.holdoff_timer_idx_1G
218.It Va hw.cxgbe.holdoff_timer_idx_ofld
219Timer index value used to delay interrupts.
220The holdoff timer list has the values 1, 5, 10, 50, 100, and 200
221by default (all values are in microseconds) and the index selects a
222value from this list.
223holdoff_timer_idx_ofld applies to queues used for TOE rx.
224The default value is 1 which means the timer value is 5us.
225Different interfaces can be assigned different values at any time via the
226dev.<port>.X.holdoff_tmr_idx and dev.<port>.X.holdoff_tmr_idx_ofld sysctls.
227.It Va hw.cxgbe.holdoff_pktc_idx_10G
228.It Va hw.cxgbe.holdoff_pktc_idx_1G
229.It Va hw.cxgbe.holdoff_pktc_idx_ofld
230Packet-count index value used to delay interrupts.
231The packet-count list has the values 1, 8, 16, and 32 by default,
232and the index selects a value from this list.
233holdoff_pktc_idx_ofld applies to queues used for TOE rx.
234The default value is -1 which means packet counting is disabled and interrupts
235are generated based solely on the holdoff timer value.
236Different interfaces can be assigned different values via the
237dev.<port>.X.holdoff_pktc_idx and dev.<port>.X.holdoff_pktc_idx_ofld sysctls.
238These sysctls work only when the interface has never been marked up (as done by
239ifconfig up).
240.It Va hw.cxgbe.qsize_txq
241Number of entries in a transmit queue's descriptor ring.
242A buf_ring of the same size is also allocated for additional
243software queuing.
244See
245.Xr ifnet 9 .
246The default value is 1024.
247Different interfaces can be assigned different values via the
248dev.<port>.X.qsize_txq sysctl.
249This sysctl works only when the interface has never been marked up (as done by
250ifconfig up).
251.It Va hw.cxgbe.qsize_rxq
252Number of entries in a receive queue's descriptor ring.
253The default value is 1024.
254Different interfaces can be assigned different values via the
255dev.<port>.X.qsize_rxq sysctl.
256This sysctl works only when the interface has never been marked up (as done by
257ifconfig up).
258.It Va hw.cxgbe.interrupt_types
259Permitted interrupt types.
260Bit 0 represents INTx (line interrupts), bit 1 MSI, and bit 2 MSI-X.
261The default is 7 (all allowed).
262The driver selects the best possible type out of the allowed types.
263.It Va hw.cxgbe.fw_install
2640 prohibits the driver from installing a firmware on the card.
2651 allows the driver to install a new firmware if internal driver
266heuristics indicate that the new firmware is preferable to the one
267already on the card.
2682 instructs the driver to always install the new firmware on the card as
269long as it is compatible with the driver and is a different version than
270the one already on the card.
271The default is 1.
272.It Va hw.cxgbe.fl_pktshift
273Number of padding bytes inserted before the beginning of an Ethernet
274frame in the receive buffer.
275The default value of 2 ensures that the Ethernet payload (usually the IP header)
276is at a 4 byte aligned address.
2770-7 are all valid values.
278.It Va hw.cxgbe.fl_pad
279A non-zero value ensures that writes from the hardware to a receive buffer are
280padded up to the specified boundary.
281The default is -1 which lets the driver pick a pad boundary.
2820 disables trailer padding completely.
283.It Va hw.cxgbe.cong_drop
284Controls the hardware response to congestion.
285-1 disables congestion feedback and is not recommended.
2860 instructs the hardware to backpressure its pipeline on congestion.
287This usually results in the port emitting PAUSE frames.
2881 instructs the hardware to drop frames destined for congested queues.
289.It Va hw.cxgbe.pause_settings
290PAUSE frame settings.
291Bit 0 is rx_pause, bit 1 is tx_pause.
292rx_pause = 1 instructs the hardware to heed incoming PAUSE frames, 0 instructs
293it to ignore them.
294tx_pause = 1 allows the hardware to emit PAUSE frames when its receive FIFO
295reaches a high threshold, 0 prohibits the hardware from emitting PAUSE frames.
296The default is 3 (both rx_pause and tx_pause = 1).
297This tunable establishes the default PAUSE settings for all ports.
298Settings can be displayed and controlled on a per-port basis via the
299dev.<port>.X.pause_settings sysctl.
300.It Va hw.cxgbe.fec
301FEC (Forward Error Correction) settings.
3020 diables FEC.
303Bit 0 enables RS FEC, bit 1 enables BASE-R RS, bit 3 is reserved.
304The default is -1 which lets the driver pick a value.
305This tunable establishes the default FEC settings for all ports.
306Settings can be displayed and controlled on a per-port basis via the
307dev.<port>.X.fec sysctl.
308.It Va hw.cxgbe.autoneg
309Link autonegotiation settings.
310This tunable establishes the default autonegotiation settings for all ports.
311Settings can be displayed and controlled on a per-port basis via the
312dev.<port>.X.autoneg sysctl.
3130 disables autonegotiation.
3141 enables autonegotiation.
315The default is -1 which lets the driver pick a value.
316dev.<port>.X.autoneg is -1 for port and module combinations that do not support
317autonegotiation.
318.It Va hw.cxgbe.buffer_packing
319Allow the hardware to deliver multiple frames in the same receive buffer
320opportunistically.
321The default is -1 which lets the driver decide.
3220 or 1 explicitly disable or enable this feature.
323.It Va hw.cxgbe.allow_mbufs_in_cluster
3241 allows the driver to lay down one or more mbufs within the receive buffer
325opportunistically.
326This is the default.
3270 prohibits the driver from doing so.
328.It Va hw.cxgbe.largest_rx_cluster
329.It Va hw.cxgbe.safest_rx_cluster
330Sizes of rx clusters.
331Each of these must be set to one of the sizes available
332(usually 2048, 4096, 9216, and 16384) and largest_rx_cluster must be greater
333than or equal to safest_rx_cluster.
334The defaults are 16384 and 4096 respectively.
335The driver never attempts to allocate a receive buffer larger than
336largest_rx_cluster and falls back to allocating buffers of
337safest_rx_cluster size if an allocation larger than safest_rx_cluster fails.
338Note that largest_rx_cluster merely establishes a ceiling -- the driver is
339allowed to allocate buffers of smaller sizes.
340.It Va hw.cxgbe.config_file
341Select a pre-packaged device configuration file.
342A configuration file contains a recipe for partitioning and configuring the
343hardware resources on the card.
344This tunable is for specialized applications only and should not be used in
345normal operation.
346The configuration profile currently in use is available in the dev.<nexus>.X.cf
347and dev.<nexus>.X.cfcsum sysctls.
348.It Va hw.cxgbe.linkcaps_allowed
349.It Va hw.cxgbe.niccaps_allowed
350.It Va hw.cxgbe.toecaps_allowed
351.It Va hw.cxgbe.rdmacaps_allowed
352.It Va hw.cxgbe.iscsicaps_allowed
353.It Va hw.cxgbe.fcoecaps_allowed
354Disallowing capabilities provides a hint to the driver and firmware to not
355reserve hardware resources for that feature.
356Each of these is a bit field with a bit for each sub-capability within the
357capability.
358This tunable is for specialized applications only and should not be used in
359normal operation.
360The capabilities for which hardware resources have been reserved are listed in
361dev.<nexus>.X.*caps sysctls.
362.El
363.Sh SUPPORT
364For general information and support,
365go to the Chelsio support website at:
366.Pa http://www.chelsio.com/ .
367.Pp
368If an issue is identified with this driver with a supported adapter,
369email all the specific information related to the issue to
370.Aq Mt support@chelsio.com .
371.Sh SEE ALSO
372.Xr altq 4 ,
373.Xr arp 4 ,
374.Xr ccr 4 ,
375.Xr cxgb 4 ,
376.Xr cxgbev 4 ,
377.Xr netintro 4 ,
378.Xr ng_ether 4 ,
379.Xr ifconfig 8
380.Sh HISTORY
381The
382.Nm
383device driver first appeared in
384.Fx 9.0 .
385Support for T5 cards first appeared in
386.Fx 9.2
387and
388.Fx 10.0 .
389Support for T6 cards first appeared in
390.Fx 11.1
391and
392.Fx 12.0 .
393.Sh AUTHORS
394.An -nosplit
395The
396.Nm
397driver was written by
398.An Navdeep Parhar Aq Mt np@FreeBSD.org .
399