xref: /freebsd/share/man/man4/cxgbe.4 (revision 6580f5c38dd5b01aeeaed16b370f1a12423437f0)
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 November 10, 2022
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.El
77.Pp
78Loader tunables with the hw.cxgbe prefix apply to all cards.
79The driver provides sysctl MIBs for both ports and parent devices using
80the names above.
81For example, a T5 adapter provides port MIBs under dev.cxl and
82adapter-wide MIBs under dev.t5nex.
83References to sysctl MIBs in the remainder of this page use
84dev.<port> for port MIBs and dev.<nexus> for adapter-wide MIBs.
85.Pp
86For more information on configuring this device, see
87.Xr ifconfig 8 .
88.Sh HARDWARE
89The
90.Nm
91driver supports 100Gb and 25Gb Ethernet adapters based on the T6 ASIC:
92.Pp
93.Bl -bullet -compact
94.It
95Chelsio T6225-CR
96.It
97Chelsio T6225-SO-CR
98.It
99Chelsio T62100-LP-CR
100.It
101Chelsio T62100-SO-CR
102.It
103Chelsio T62100-CR
104.El
105.Pp
106The
107.Nm
108driver supports 40Gb, 10Gb and 1Gb Ethernet adapters based on the T5 ASIC:
109.Pp
110.Bl -bullet -compact
111.It
112Chelsio T580-CR
113.It
114Chelsio T580-LP-CR
115.It
116Chelsio T580-LP-SO-CR
117.It
118Chelsio T560-CR
119.It
120Chelsio T540-CR
121.It
122Chelsio T540-LP-CR
123.It
124Chelsio T522-CR
125.It
126Chelsio T520-LL-CR
127.It
128Chelsio T520-CR
129.It
130Chelsio T520-SO
131.It
132Chelsio T520-BT
133.It
134Chelsio T504-BT
135.El
136.Pp
137The
138.Nm
139driver supports 10Gb and 1Gb Ethernet adapters based on the T4 ASIC:
140.Pp
141.Bl -bullet -compact
142.It
143Chelsio T420-CR
144.It
145Chelsio T422-CR
146.It
147Chelsio T440-CR
148.It
149Chelsio T420-BCH
150.It
151Chelsio T440-BCH
152.It
153Chelsio T440-CH
154.It
155Chelsio T420-SO
156.It
157Chelsio T420-CX
158.It
159Chelsio T420-BT
160.It
161Chelsio T404-BT
162.El
163.Sh LOADER TUNABLES
164Tunables can be set at the
165.Xr loader 8
166prompt before booting the kernel or stored in
167.Xr loader.conf 5 .
168There are multiple tunables that control the number of queues of various
169types.
170A negative value for such a tunable instructs the driver to create
171up to that many queues if there are enough CPU cores available.
172.Bl -tag -width indent
173.It Va hw.cxgbe.ntxq
174Number of NIC tx queues used for a port.
175The default is 16 or the number
176of CPU cores in the system, whichever is less.
177.It Va hw.cxgbe.nrxq
178Number of NIC rx queues used for a port.
179The default is 8 or the number
180of CPU cores in the system, whichever is less.
181.It Va hw.cxgbe.nofldtxq
182Number of TOE tx queues used for a port.
183The default is 8 or the
184number of CPU cores in the system, whichever is less.
185.It Va hw.cxgbe.nofldrxq
186Number of TOE rx queues used for a port.
187The default is 2 or the
188number of CPU cores in the system, whichever is less.
189.It Va hw.cxgbe.num_vis
190Number of virtual interfaces (VIs) created for each port.
191Each virtual interface creates a separate network interface.
192The first virtual interface on each port is required and represents
193the primary network interface on the port.
194Additional virtual interfaces on a port are named using the Virtual Interface
195name from the table above.
196Additional virtual interfaces use a single pair of queues
197for rx and tx as well an additional pair of queues for TOE rx and tx.
198The default is 1.
199.It Va hw.cxgbe.holdoff_timer_idx
200.It Va hw.cxgbe.holdoff_timer_idx_ofld
201Timer index value used to delay interrupts.
202The holdoff timer list has the values 1, 5, 10, 50, 100, and 200
203by default (all values are in microseconds) and the index selects a
204value from this list.
205holdoff_timer_idx_ofld applies to queues used for TOE rx.
206The default value is 1 which means the timer value is 5us.
207Different interfaces can be assigned different values at any time via the
208dev.<port>.X.holdoff_tmr_idx and dev.<port>.X.holdoff_tmr_idx_ofld sysctls.
209.It Va hw.cxgbe.holdoff_pktc_idx
210.It Va hw.cxgbe.holdoff_pktc_idx_ofld
211Packet-count index value used to delay interrupts.
212The packet-count list has the values 1, 8, 16, and 32 by default,
213and the index selects a value from this list.
214holdoff_pktc_idx_ofld applies to queues used for TOE rx.
215The default value is -1 which means packet counting is disabled and interrupts
216are generated based solely on the holdoff timer value.
217Different interfaces can be assigned different values via the
218dev.<port>.X.holdoff_pktc_idx and dev.<port>.X.holdoff_pktc_idx_ofld sysctls.
219These sysctls work only when the interface has never been marked up (as done by
220ifconfig up).
221.It Va hw.cxgbe.qsize_txq
222Number of entries in a transmit queue's descriptor ring.
223A buf_ring of the same size is also allocated for additional
224software queuing.
225See
226.Xr ifnet 9 .
227The default value is 1024.
228Different interfaces can be assigned different values via the
229dev.<port>.X.qsize_txq sysctl.
230This sysctl works only when the interface has never been marked up (as done by
231ifconfig up).
232.It Va hw.cxgbe.qsize_rxq
233Number of entries in a receive queue's descriptor ring.
234The default value is 1024.
235Different interfaces can be assigned different values via the
236dev.<port>.X.qsize_rxq sysctl.
237This sysctl works only when the interface has never been marked up (as done by
238ifconfig up).
239.It Va hw.cxgbe.interrupt_types
240Permitted interrupt types.
241Bit 0 represents INTx (line interrupts), bit 1 MSI, and bit 2 MSI-X.
242The default is 7 (all allowed).
243The driver selects the best possible type out of the allowed types.
244.It Va hw.cxgbe.pcie_relaxed_ordering
245PCIe Relaxed Ordering.
246-1 indicates the driver should determine whether to enable or disable PCIe RO.
2470 disables PCIe RO.
2481 enables PCIe RO.
2492 indicates the driver should not modify the PCIe RO setting.
250The default is -1.
251.It Va hw.cxgbe.fw_install
2520 prohibits the driver from installing a firmware on the card.
2531 allows the driver to install a new firmware if internal driver
254heuristics indicate that the new firmware is preferable to the one
255already on the card.
2562 instructs the driver to always install the new firmware on the card as
257long as it is compatible with the driver and is a different version than
258the one already on the card.
259The default is 1.
260.It Va hw.cxgbe.fl_pktshift
261Number of padding bytes inserted before the beginning of an Ethernet
262frame in the receive buffer.
263The default value is 0.
264A value of 2 would ensure that the Ethernet payload (usually the IP header)
265is at a 4 byte aligned address.
2660-7 are all valid values.
267.It Va hw.cxgbe.fl_pad
268A non-zero value ensures that writes from the hardware to a receive buffer are
269padded up to the specified boundary.
270The default is -1 which lets the driver pick a pad boundary.
2710 disables trailer padding completely.
272.It Va hw.cxgbe.cong_drop
273Controls the hardware response to congestion.
274-1 disables congestion feedback and is not recommended.
2750 instructs the hardware to backpressure its pipeline on congestion.
276This usually results in the port emitting PAUSE frames.
2771 instructs the hardware to drop frames destined for congested queues.
2782 instructs the hardware to both backpressure the pipeline and drop frames.
279.It Va hw.cxgbe.pause_settings
280PAUSE frame settings.
281Bit 0 is rx_pause, bit 1 is tx_pause, bit 2 is pause_autoneg.
282rx_pause = 1 instructs the hardware to heed incoming PAUSE frames, 0 instructs
283it to ignore them.
284tx_pause = 1 allows the hardware to emit PAUSE frames when its receive FIFO
285reaches a high threshold, 0 prohibits the hardware from emitting PAUSE frames.
286pause_autoneg = 1 overrides the rx_pause and tx_pause bits and instructs the
287hardware to negotiate PAUSE settings with the link peer.
288The default is 7 (all three = 1).
289This tunable establishes the default PAUSE settings for all ports.
290Settings can be displayed and controlled on a per-port basis via the
291dev.<port>.X.pause_settings sysctl.
292.It Va hw.cxgbe.fec
293Forward Error Correction settings.
294-1 (default) means driver should automatically pick a value.
2950 disables FEC.
296Finer grained control can be achieved by setting individual bits.
297Bit 0 enables RS FEC, bit 1 enables BASE-R FEC (aka Firecode FEC), bit
2982 enables NO FEC, and bit 6 enables the FEC that is recommended by the
299transceiver/cable that is plugged in.
300These bits can be set together in any combination.
301This tunable establishes the default FEC settings for all ports.
302Settings can be controlled on a per-port basis via the
303dev.<port>.X.requested_fec sysctl.
304The FEC in use on the link is available in dev.<port>.X.link_fec when
305the link is up.
306.It Va hw.cxgbe.autoneg
307Link autonegotiation settings.
308This tunable establishes the default autonegotiation settings for all ports.
309Settings can be displayed and controlled on a per-port basis via the
310dev.<port>.X.autoneg sysctl.
3110 disables autonegotiation.
3121 enables autonegotiation.
313The default is -1 which lets the driver pick a value.
314dev.<port>.X.autoneg is -1 for port and module combinations that do not support
315autonegotiation.
316.It Va hw.cxgbe.buffer_packing
317Allow the hardware to deliver multiple frames in the same receive buffer
318opportunistically.
319The default is -1 which lets the driver decide.
3200 or 1 explicitly disable or enable this feature.
321.It Va hw.cxgbe.largest_rx_cluster
322.It Va hw.cxgbe.safest_rx_cluster
323Sizes of rx clusters.
324Each of these must be set to one of the sizes available
325(usually 2048, 4096, 9216, and 16384) and largest_rx_cluster must be greater
326than or equal to safest_rx_cluster.
327The defaults are 16384 and 4096 respectively.
328The driver never attempts to allocate a receive buffer larger than
329largest_rx_cluster and falls back to allocating buffers of
330safest_rx_cluster size if an allocation larger than safest_rx_cluster fails.
331Note that largest_rx_cluster merely establishes a ceiling -- the driver is
332allowed to allocate buffers of smaller sizes.
333.It Va hw.cxgbe.config_file
334Select a pre-packaged device configuration file.
335A configuration file contains a recipe for partitioning and configuring the
336hardware resources on the card.
337This tunable is for specialized applications only and should not be used in
338normal operation.
339The configuration profile currently in use is available in the dev.<nexus>.X.cf
340and dev.<nexus>.X.cfcsum sysctls.
341.It Va hw.cxgbe.linkcaps_allowed
342.It Va hw.cxgbe.niccaps_allowed
343.It Va hw.cxgbe.toecaps_allowed
344.It Va hw.cxgbe.rdmacaps_allowed
345.It Va hw.cxgbe.iscsicaps_allowed
346.It Va hw.cxgbe.fcoecaps_allowed
347Disallowing capabilities provides a hint to the driver and firmware to not
348reserve hardware resources for that feature.
349Each of these is a bit field with a bit for each sub-capability within the
350capability.
351This tunable is for specialized applications only and should not be used in
352normal operation.
353The capabilities for which hardware resources have been reserved are listed in
354dev.<nexus>.X.*caps sysctls.
355.It Va hw.cxgbe.tx_vm_wr
356Setting this to 1 instructs the driver to use VM work requests to transmit data.
357This lets PF interfaces transmit frames to VF interfaces over the internal
358switch in the ASIC.
359Note that the
360.Xr cxgbev 4
361VF driver always uses VM work requests and is not affected by this tunable.
362The default value is 0 and should be changed only if PF and VF interfaces need
363to communicate with each other.
364Different interfaces can be assigned different values using the
365dev.<port>.X.tx_vm_wr sysctl when the interface is administratively down.
366.It Va hw.cxgbe.attack_filter
367Set to 1 to enable the "attack filter".
368Default is 0.
369The attack filter will drop an incoming frame if any of these conditions is
370true: src ip/ip6 == dst ip/ip6; tcp and src/dst ip is not unicast; src/dst ip is
371loopback (127.x.y.z); src ip6 is not unicast; src/dst ip6 is loopback (::1/128)
372or unspecified (::/128); tcp and src/dst ip6 is mcast (ff00::/8).
373This facility is available on T4 and T5 based cards only.
374.It Va hw.cxgbe.drop_ip_fragments
375Set to 1 to drop all incoming IP fragments.
376Default is 0.
377Note that this drops valid frames.
378.It Va hw.cxgbe.drop_pkts_with_l2_errors
379Set to 1 to drop incoming frames with Layer 2 length or checksum errors.
380Default is 1.
381.It Va hw.cxgbe.drop_pkts_with_l3_errors
382Set to 1 to drop incoming frames with IP version, length, or checksum errors.
383The IP checksum is validated for TCP or UDP packets only.
384Default is 0.
385.It Va hw.cxgbe.drop_pkts_with_l4_errors
386Set to 1 to drop incoming frames with Layer 4 (TCP or UDP) length,
387checksum, or other errors.
388Default is 0.
389.El
390.Sh SUPPORT
391For general information and support,
392go to the Chelsio support website at:
393.Pa http://www.chelsio.com/ .
394.Pp
395If an issue is identified with this driver with a supported adapter,
396email all the specific information related to the issue to
397.Aq Mt support@chelsio.com .
398.Sh SEE ALSO
399.Xr arp 4 ,
400.Xr ccr 4 ,
401.Xr cxgb 4 ,
402.Xr cxgbev 4 ,
403.Xr netintro 4 ,
404.Xr ng_ether 4 ,
405.Xr ifconfig 8
406.Sh HISTORY
407The
408.Nm
409device driver first appeared in
410.Fx 9.0 .
411Support for T5 cards first appeared in
412.Fx 9.2
413and
414.Fx 10.0 .
415Support for T6 cards first appeared in
416.Fx 11.1
417and
418.Fx 12.0 .
419.Sh AUTHORS
420.An -nosplit
421The
422.Nm
423driver was written by
424.An Navdeep Parhar Aq Mt np@FreeBSD.org .
425