xref: /freebsd/share/man/man4/cxgbe.4 (revision 076ad2f836d5f49dc1375f1677335a48fe0d4b82)
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 December 22, 2016
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
218Timer index value used to delay interrupts.
219The holdoff timer list has the values 1, 5, 10, 50, 100, and 200
220by default (all values are in microseconds) and the index selects a
221value from this list.
222The default value is 1 which means the timer value is 5us.
223Different interfaces can be assigned different values at any time via the
224dev.<port>.X.holdoff_tmr_idx sysctl.
225.It Va hw.cxgbe.holdoff_pktc_idx_10G
226.It Va hw.cxgbe.holdoff_pktc_idx_1G
227Packet-count index value used to delay interrupts.
228The packet-count list has the values 1, 8, 16, and 32 by default,
229and the index selects a value from this list.
230The default value is -1 which means packet counting is disabled and interrupts
231are generated based solely on the holdoff timer value.
232Different interfaces can be assigned different values via the
233dev.<port>.X.holdoff_pktc_idx sysctl.
234This sysctl works only when the interface has never been marked up (as done by
235ifconfig up).
236.It Va hw.cxgbe.qsize_txq
237Number of entries in a transmit queue's descriptor ring.
238A buf_ring of the same size is also allocated for additional
239software queuing.
240See
241.Xr ifnet 9 .
242The default value is 1024.
243Different interfaces can be assigned different values via the
244dev.<port>.X.qsize_txq sysctl.
245This sysctl works only when the interface has never been marked up (as done by
246ifconfig up).
247.It Va hw.cxgbe.qsize_rxq
248Number of entries in a receive queue's descriptor ring.
249The default value is 1024.
250Different interfaces can be assigned different values via the
251dev.<port>.X.qsize_rxq sysctl.
252This sysctl works only when the interface has never been marked up (as done by
253ifconfig up).
254.It Va hw.cxgbe.interrupt_types
255Permitted interrupt types.
256Bit 0 represents INTx (line interrupts), bit 1 MSI, and bit 2 MSI-X.
257The default is 7 (all allowed).
258The driver selects the best possible type out of the allowed types.
259.It Va hw.cxgbe.fw_install
2600 prohibits the driver from installing a firmware on the card.
2611 allows the driver to install a new firmware if internal driver
262heuristics indicate that the new firmware is preferable to the one
263already on the card.
2642 instructs the driver to always install the new firmware on the card as
265long as it is compatible with the driver and is a different version than
266the one already on the card.
267The default is 1.
268.It Va hw.cxgbe.fl_pktshift
269Number of padding bytes inserted before the beginning of an Ethernet
270frame in the receive buffer.
271The default value of 2 ensures that the Ethernet payload (usually the IP header)
272is at a 4 byte aligned address.
2730-7 are all valid values.
274.It Va hw.cxgbe.fl_pad
275A non-zero value ensures that writes from the hardware to a receive buffer are
276padded up to the specified boundary.
277The default is -1 which lets the driver pick a pad boundary.
2780 disables trailer padding completely.
279.It Va hw.cxgbe.cong_drop
280Controls the hardware response to congestion.
281-1 disables congestion feedback and is not recommended.
2820 instructs the hardware to backpressure its pipeline on congestion.
283This usually results in the port emitting PAUSE frames.
2841 instructs the hardware to drop frames destined for congested queues.
285.It Va hw.cxgbe.pause_settings
286PAUSE frame settings.
287Bit 0 is rx_pause, bit 1 is tx_pause.
288rx_pause = 1 instructs the hardware to heed incoming PAUSE frames, 0 instructs
289it to ignore them.
290tx_pause = 1 allows the hardware to emit PAUSE frames when its receive FIFO
291reaches a high threshold, 0 prohibits the hardware from emitting PAUSE frames.
292The default is 3 (both rx_pause and tx_pause = 1).
293This tunable establishes the default PAUSE settings for all ports.
294Settings can be displayed and controlled on a per-port basis via the
295dev.<port>.X.pause_settings sysctl.
296.It Va hw.cxgbe.fec
297FEC (Forward Error Correction) settings.
2980 diables FEC.
299Bit 0 enables RS FEC, bit 1 enables BASE-R RS, bit 3 is reserved.
300The default is -1 which lets the driver pick a value.
301This tunable establishes the default FEC settings for all ports.
302Settings can be displayed and controlled on a per-port basis via the
303dev.<port>.X.fec sysctl.
304.It Va hw.cxgbe.autoneg
305Link autonegotiation settings.
306This tunable establishes the default autonegotiation settings for all ports.
307Settings can be displayed and controlled on a per-port basis via the
308dev.<port>.X.autoneg sysctl.
3090 disables autonegotiation.
3101 enables autonegotiation.
311The default is -1 which lets the driver pick a value.
312dev.<port>.X.autoneg is -1 for port and module combinations that do not support
313autonegotiation.
314.It Va hw.cxgbe.buffer_packing
315Allow the hardware to deliver multiple frames in the same receive buffer
316opportunistically.
317The default is -1 which lets the driver decide.
3180 or 1 explicitly disable or enable this feature.
319.It Va hw.cxgbe.allow_mbufs_in_cluster
3201 allows the driver to lay down one or more mbufs within the receive buffer
321opportunistically.
322This is the default.
3230 prohibits the driver from doing so.
324.It Va hw.cxgbe.largest_rx_cluster
325.It Va hw.cxgbe.safest_rx_cluster
326Sizes of rx clusters.
327Each of these must be set to one of the sizes available
328(usually 2048, 4096, 9216, and 16384) and largest_rx_cluster must be greater
329than or equal to safest_rx_cluster.
330The defaults are 16384 and 4096 respectively.
331The driver never attempts to allocate a receive buffer larger than
332largest_rx_cluster and falls back to allocating buffers of
333safest_rx_cluster size if an allocation larger than safest_rx_cluster fails.
334Note that largest_rx_cluster merely establishes a ceiling -- the driver is
335allowed to allocate buffers of smaller sizes.
336.It Va hw.cxgbe.config_file
337Select a pre-packaged device configuration file.
338A configuration file contains a recipe for partitioning and configuring the
339hardware resources on the card.
340This tunable is for specialized applications only and should not be used in
341normal operation.
342The configuration profile currently in use is available in the dev.<nexus>.X.cf
343and dev.<nexus>.X.cfcsum sysctls.
344.It Va hw.cxgbe.linkcaps_allowed
345.It Va hw.cxgbe.niccaps_allowed
346.It Va hw.cxgbe.toecaps_allowed
347.It Va hw.cxgbe.rdmacaps_allowed
348.It Va hw.cxgbe.iscsicaps_allowed
349.It Va hw.cxgbe.fcoecaps_allowed
350Disallowing capabilities provides a hint to the driver and firmware to not
351reserve hardware resources for that feature.
352Each of these is a bit field with a bit for each sub-capability within the
353capability.
354This tunable is for specialized applications only and should not be used in
355normal operation.
356The capabilities for which hardware resources have been reserved are listed in
357dev.<nexus>.X.*caps sysctls.
358.El
359.Sh SUPPORT
360For general information and support,
361go to the Chelsio support website at:
362.Pa http://www.chelsio.com/ .
363.Pp
364If an issue is identified with this driver with a supported adapter,
365email all the specific information related to the issue to
366.Aq Mt support@chelsio.com .
367.Sh SEE ALSO
368.Xr altq 4 ,
369.Xr arp 4 ,
370.Xr cxgb 4 ,
371.Xr cxgbev 4 ,
372.Xr netintro 4 ,
373.Xr ng_ether 4 ,
374.Xr ifconfig 8
375.Sh HISTORY
376The
377.Nm
378device driver first appeared in
379.Fx 9.0 .
380Support for T5 cards first appeared in
381.Fx 9.2
382and
383.Fx 10.0 .
384Support for T6 cards first appeared in
385.Fx 11.1
386and
387.Fx 12.0 .
388.Sh AUTHORS
389.An -nosplit
390The
391.Nm
392driver was written by
393.An Navdeep Parhar Aq Mt np@FreeBSD.org .
394