xref: /freebsd/share/man/man4/cxgbe.4 (revision 24e4dcf4ba5e9dedcf89efd358ea3e1fe5867020)
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.Dd September 29, 2025
33.Dt CXGBE 4
34.Os
35.Sh NAME
36.Nm cxgbe
37.Nd "Chelsio T4-, T5-, and T6-based 100Gb, 40Gb, 25Gb, 10Gb, and 1Gb Ethernet adapter driver"
38.Sh SYNOPSIS
39To compile this driver into the kernel,
40place the following lines in your
41kernel configuration file:
42.Bd -ragged -offset indent
43.Cd "device cxgbe"
44.Ed
45.Pp
46To load the driver as a
47module at boot time, place the following lines in
48.Xr loader.conf 5 :
49.Bd -literal -offset indent
50t4fw_cfg_load="YES"
51t5fw_cfg_load="YES"
52t6fw_cfg_load="YES"
53if_cxgbe_load="YES"
54.Ed
55.Sh DESCRIPTION
56The
57.Nm
58driver provides support for PCI Express Ethernet adapters based on
59the Chelsio Terminator 4, Terminator 5, and Terminator 6 ASICs (T4, T5, and T6).
60The driver supports Jumbo Frames, Transmit/Receive checksum offload,
61TCP segmentation offload (TSO), Large Receive Offload (LRO), VLAN
62tag insertion/extraction, VLAN checksum offload, VLAN TSO, VXLAN checksum
63offload, VXLAN TSO, and Receive Side Steering (RSS).
64For further hardware information and questions related to hardware
65requirements, see
66.Pa http://www.chelsio.com/ .
67.Pp
68The
69.Nm
70driver uses different names for devices based on the associated ASIC:
71.Bl -column -offset indent "ASIC" "Port Name" "Parent Device"
72.It Sy ASIC Ta Sy Port Name Ta Sy Parent Device Ta Sy Virtual Interface
73.It T4 Ta cxgbe Ta t4nex Ta vcxgbe
74.It T5 Ta cxl Ta t5nex Ta vcxl
75.It T6 Ta cc Ta t6nex Ta vcc
76.It T7 Ta che Ta chnex Ta vche
77.El
78.Pp
79Loader tunables with the hw.cxgbe prefix apply to all cards.
80The driver provides sysctl MIBs for both ports and parent devices using
81the names above.
82For example, a T5 adapter provides port MIBs under dev.cxl and
83adapter-wide MIBs under dev.t5nex.
84References to sysctl MIBs in the remainder of this page use
85dev.<port> for port MIBs and dev.<nexus> for adapter-wide MIBs.
86.Pp
87For more information on configuring this device, see
88.Xr ifconfig 8 .
89.Sh HARDWARE
90The
91.Nm
92driver supports 100Gb and 25Gb Ethernet adapters based on the T6 ASIC:
93.Pp
94.Bl -bullet -compact
95.It
96Chelsio T6225-CR
97.It
98Chelsio T6225-SO-CR
99.It
100Chelsio T62100-LP-CR
101.It
102Chelsio T62100-SO-CR
103.It
104Chelsio T62100-CR
105.El
106.Pp
107The
108.Nm
109driver supports 40Gb, 10Gb and 1Gb Ethernet adapters based on the T5 ASIC:
110.Pp
111.Bl -bullet -compact
112.It
113Chelsio T580-CR
114.It
115Chelsio T580-LP-CR
116.It
117Chelsio T580-LP-SO-CR
118.It
119Chelsio T560-CR
120.It
121Chelsio T540-CR
122.It
123Chelsio T540-LP-CR
124.It
125Chelsio T522-CR
126.It
127Chelsio T520-LL-CR
128.It
129Chelsio T520-CR
130.It
131Chelsio T520-SO
132.It
133Chelsio T520-BT
134.It
135Chelsio T504-BT
136.El
137.Pp
138The
139.Nm
140driver supports 10Gb and 1Gb Ethernet adapters based on the T4 ASIC:
141.Pp
142.Bl -bullet -compact
143.It
144Chelsio T420-CR
145.It
146Chelsio T422-CR
147.It
148Chelsio T440-CR
149.It
150Chelsio T420-BCH
151.It
152Chelsio T440-BCH
153.It
154Chelsio T440-CH
155.It
156Chelsio T420-SO
157.It
158Chelsio T420-CX
159.It
160Chelsio T420-BT
161.It
162Chelsio T404-BT
163.El
164.Sh LOADER TUNABLES
165Tunables can be set at the
166.Xr loader 8
167prompt before booting the kernel or stored in
168.Xr loader.conf 5 .
169There are multiple tunables that control the number of queues of various
170types.
171A negative value for such a tunable instructs the driver to create
172up to that many queues if there are enough CPU cores available.
173.Bl -tag -width indent
174.It Va hw.cxgbe.ntxq
175Number of NIC tx queues used for a port.
176The default is 16 or the number
177of CPU cores in the system, whichever is less.
178.It Va hw.cxgbe.nrxq
179Number of NIC rx queues used for a port.
180The default is 8 or the number
181of CPU cores in the system, whichever is less.
182.It Va hw.cxgbe.nofldtxq
183Number of TOE tx queues used for a port.
184The default is 8 or the
185number of CPU cores in the system, whichever is less.
186.It Va hw.cxgbe.nofldrxq
187Number of TOE rx queues used for a port.
188The default is 2 or the
189number of CPU cores in the system, whichever is less.
190.It Va hw.cxgbe.num_vis
191Number of virtual interfaces (VIs) created for each port.
192Each virtual interface creates a separate network interface.
193The first virtual interface on each port is required and represents
194the primary network interface on the port.
195Additional virtual interfaces on a port are named using the Virtual Interface
196name from the table above.
197Additional virtual interfaces use a single pair of queues
198for rx and tx as well an additional pair of queues for TOE rx and tx.
199The default is 1.
200.It Va hw.cxgbe.holdoff_timer_idx
201.It Va hw.cxgbe.holdoff_timer_idx_ofld
202Timer index value used to delay interrupts.
203The holdoff timer list has the values 1, 5, 10, 50, 100, and 200
204by default (all values are in microseconds) and the index selects a
205value from this list.
206holdoff_timer_idx_ofld applies to queues used for TOE rx.
207The default value is 1 which means the timer value is 5us.
208Different interfaces can be assigned different values at any time via the
209dev.<port>.X.holdoff_tmr_idx and dev.<port>.X.holdoff_tmr_idx_ofld sysctls.
210.It Va hw.cxgbe.holdoff_pktc_idx
211.It Va hw.cxgbe.holdoff_pktc_idx_ofld
212Packet-count index value used to delay interrupts.
213The packet-count list has the values 1, 8, 16, and 32 by default,
214and the index selects a value from this list.
215holdoff_pktc_idx_ofld applies to queues used for TOE rx.
216The default value is -1 which means packet counting is disabled and interrupts
217are generated based solely on the holdoff timer value.
218Different interfaces can be assigned different values via the
219dev.<port>.X.holdoff_pktc_idx and dev.<port>.X.holdoff_pktc_idx_ofld sysctls.
220These sysctls work only when the interface has never been marked up (as done by
221ifconfig up).
222.It Va hw.cxgbe.qsize_txq
223Number of entries in a transmit queue's descriptor ring.
224A buf_ring of the same size is also allocated for additional
225software queuing.
226See
227.Xr ifnet 9 .
228The default value is 1024.
229Different interfaces can be assigned different values via the
230dev.<port>.X.qsize_txq sysctl.
231This sysctl works only when the interface has never been marked up (as done by
232ifconfig up).
233.It Va hw.cxgbe.qsize_rxq
234Number of entries in a receive queue's descriptor ring.
235The default value is 1024.
236Different interfaces can be assigned different values via the
237dev.<port>.X.qsize_rxq sysctl.
238This sysctl works only when the interface has never been marked up (as done by
239ifconfig up).
240.It Va hw.cxgbe.interrupt_types
241Permitted interrupt types.
242Bit 0 represents INTx (line interrupts), bit 1 MSI, and bit 2 MSI-X.
243The default is 7 (all allowed).
244The driver selects the best possible type out of the allowed types.
245.It Va hw.cxgbe.pcie_relaxed_ordering
246PCIe Relaxed Ordering.
247-1 indicates the driver should determine whether to enable or disable PCIe RO.
2480 disables PCIe RO.
2491 enables PCIe RO.
2502 indicates the driver should not modify the PCIe RO setting.
251The default is -1.
252.It Va hw.cxgbe.fw_install
2530 prohibits the driver from installing a firmware on the card.
2541 allows the driver to install a new firmware if internal driver
255heuristics indicate that the new firmware is preferable to the one
256already on the card.
2572 instructs the driver to always install the new firmware on the card as
258long as it is compatible with the driver and is a different version than
259the one already on the card.
260The default is 1.
261.It Va hw.cxgbe.fl_pktshift
262Number of padding bytes inserted before the beginning of an Ethernet
263frame in the receive buffer.
264The default value is 0.
265A value of 2 would ensure that the Ethernet payload (usually the IP header)
266is at a 4 byte aligned address.
2670-7 are all valid values.
268.It Va hw.cxgbe.fl_pad
269A non-zero value ensures that writes from the hardware to a receive buffer are
270padded up to the specified boundary.
271The default is -1 which lets the driver pick a pad boundary.
2720 disables trailer padding completely.
273.It Va hw.cxgbe.cong_drop
274Controls the hardware response to congestion.
275-1 disables congestion feedback and is not recommended.
2760 instructs the hardware to backpressure its pipeline on congestion.
277This usually results in the port emitting PAUSE frames.
2781 instructs the hardware to drop frames destined for congested queues.
2792 instructs the hardware to both backpressure the pipeline and drop frames.
280.It Va hw.cxgbe.pause_settings
281PAUSE frame settings.
282Bit 0 is rx_pause, bit 1 is tx_pause, bit 2 is pause_autoneg.
283rx_pause = 1 instructs the hardware to heed incoming PAUSE frames, 0 instructs
284it to ignore them.
285tx_pause = 1 allows the hardware to emit PAUSE frames when its receive FIFO
286reaches a high threshold, 0 prohibits the hardware from emitting PAUSE frames.
287pause_autoneg = 1 overrides the rx_pause and tx_pause bits and instructs the
288hardware to negotiate PAUSE settings with the link peer.
289The default is 7 (all three = 1).
290This tunable establishes the default PAUSE settings for all ports.
291Settings can be displayed and controlled on a per-port basis via the
292dev.<port>.X.pause_settings sysctl.
293.It Va hw.cxgbe.fec
294Forward Error Correction settings.
295-1 (default) means driver should automatically pick a value.
2960 disables FEC.
297Finer grained control can be achieved by setting individual bits.
298Bit 0 enables RS FEC, bit 1 enables BASE-R FEC (aka Firecode FEC), bit
2992 enables NO FEC, and bit 6 enables the FEC that is recommended by the
300transceiver/cable that is plugged in.
301These bits can be set together in any combination.
302This tunable establishes the default FEC settings for all ports.
303Settings can be controlled on a per-port basis via the
304dev.<port>.X.requested_fec sysctl.
305The FEC in use on the link is available in dev.<port>.X.link_fec when
306the link is up.
307.It Va hw.cxgbe.autoneg
308Link autonegotiation settings.
309This tunable establishes the default autonegotiation settings for all ports.
310Settings can be displayed and controlled on a per-port basis via the
311dev.<port>.X.autoneg sysctl.
3120 disables autonegotiation.
3131 enables autonegotiation.
314The default is -1 which lets the driver pick a value.
315dev.<port>.X.autoneg is -1 for port and module combinations that do not support
316autonegotiation.
317.It Va hw.cxgbe.buffer_packing
318Allow the hardware to deliver multiple frames in the same receive buffer
319opportunistically.
320The default is -1 which lets the driver decide.
3210 or 1 explicitly disable or enable this feature.
322.It Va hw.cxgbe.largest_rx_cluster
323.It Va hw.cxgbe.safest_rx_cluster
324Sizes of rx clusters.
325Each of these must be set to one of the sizes available
326(usually 2048, 4096, 9216, and 16384) and largest_rx_cluster must be greater
327than or equal to safest_rx_cluster.
328The defaults are 16384 and 4096 respectively.
329The driver never attempts to allocate a receive buffer larger than
330largest_rx_cluster and falls back to allocating buffers of
331safest_rx_cluster size if an allocation larger than safest_rx_cluster fails.
332Note that largest_rx_cluster merely establishes a ceiling -- the driver is
333allowed to allocate buffers of smaller sizes.
334.It Va hw.cxgbe.config_file
335Select a pre-packaged device configuration file.
336A configuration file contains a recipe for partitioning and configuring the
337hardware resources on the card.
338This tunable is for specialized applications only and should not be used in
339normal operation.
340The configuration profile currently in use is available in the dev.<nexus>.X.cf
341and dev.<nexus>.X.cfcsum sysctls.
342.It Va hw.cxgbe.linkcaps_allowed
343.It Va hw.cxgbe.niccaps_allowed
344.It Va hw.cxgbe.toecaps_allowed
345.It Va hw.cxgbe.rdmacaps_allowed
346.It Va hw.cxgbe.iscsicaps_allowed
347.It Va hw.cxgbe.fcoecaps_allowed
348Disallowing capabilities provides a hint to the driver and firmware to not
349reserve hardware resources for that feature.
350Each of these is a bit field with a bit for each sub-capability within the
351capability.
352This tunable is for specialized applications only and should not be used in
353normal operation.
354The capabilities for which hardware resources have been reserved are listed in
355dev.<nexus>.X.*caps sysctls.
356.It Va hw.cxgbe.tx_vm_wr
357Setting this to 1 instructs the driver to use VM work requests to transmit data.
358This lets PF interfaces transmit frames to VF interfaces over the internal
359switch in the ASIC.
360Note that the
361.Xr cxgbev 4
362VF driver always uses VM work requests and is not affected by this tunable.
363The default value is 0 and should be changed only if PF and VF interfaces need
364to communicate with each other.
365Different interfaces can be assigned different values using the
366dev.<port>.X.tx_vm_wr sysctl when the interface is administratively down.
367.It Va hw.cxgbe.attack_filter
368Set to 1 to enable the "attack filter".
369Default is 0.
370The attack filter will drop an incoming frame if any of these conditions is
371true: src ip/ip6 == dst ip/ip6; tcp and src/dst ip is not unicast; src/dst ip is
372loopback (127.x.y.z); src ip6 is not unicast; src/dst ip6 is loopback (::1/128)
373or unspecified (::/128); tcp and src/dst ip6 is mcast (ff00::/8).
374This facility is available on T4 and T5 based cards only.
375.It Va hw.cxgbe.drop_ip_fragments
376Set to 1 to drop all incoming IP fragments.
377Default is 0.
378Note that this drops valid frames.
379.It Va hw.cxgbe.drop_pkts_with_l2_errors
380Set to 1 to drop incoming frames with Layer 2 length or checksum errors.
381Default is 1.
382.It Va hw.cxgbe.drop_pkts_with_l3_errors
383Set to 1 to drop incoming frames with IP version, length, or checksum errors.
384The IP checksum is validated for TCP or UDP packets only.
385Default is 0.
386.It Va hw.cxgbe.drop_pkts_with_l4_errors
387Set to 1 to drop incoming frames with Layer 4 (TCP or UDP) length,
388checksum, or other errors.
389Default is 0.
390.El
391.Sh SUPPORT
392For general information and support,
393go to the Chelsio support website at:
394.Pa http://www.chelsio.com/ .
395.Pp
396If an issue is identified with this driver with a supported adapter,
397email all the specific information related to the issue to
398.Aq Mt support@chelsio.com .
399.Sh SEE ALSO
400.Xr arp 4 ,
401.Xr ccr 4 ,
402.Xr cxgb 4 ,
403.Xr cxgbev 4 ,
404.Xr netintro 4 ,
405.Xr ng_ether 4 ,
406.Xr ifconfig 8
407.Sh HISTORY
408The
409.Nm
410device driver first appeared in
411.Fx 9.0 .
412Support for T5 cards first appeared in
413.Fx 9.2
414and
415.Fx 10.0 .
416Support for T6 cards first appeared in
417.Fx 11.1
418and
419.Fx 12.0 .
420Support for T7 cards first appeared in
421.Fx 15.0 .
422.Sh AUTHORS
423.An -nosplit
424The
425.Nm
426driver was written by
427.An Navdeep Parhar Aq Mt np@FreeBSD.org .
428