xref: /freebsd/share/man/man4/cxgbe.4 (revision 5bb3134a8c21cb87b30e135ef168483f0333dabb)
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 November 10, 2021
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, VXLAN checksum
65offload, VXLAN TSO, and Receive 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.ntxq
176Number of NIC tx queues used for a port.
177The default is 16 or the number
178of CPU cores in the system, whichever is less.
179.It Va hw.cxgbe.nrxq
180Number of NIC rx queues used for a port.
181The default is 8 or the number
182of CPU cores in the system, whichever is less.
183.It Va hw.cxgbe.nofldtxq
184Number of TOE tx queues used for a port.
185The default is 8 or the
186number of CPU cores in the system, whichever is less.
187.It Va hw.cxgbe.nofldrxq
188Number of TOE rx queues used for a port.
189The default is 2 or the
190number of CPU cores in the system, whichever is less.
191.It Va hw.cxgbe.num_vis
192Number of virtual interfaces (VIs) created for each port.
193Each virtual interface creates a separate network interface.
194The first virtual interface on each port is required and represents
195the primary network interface on the port.
196Additional virtual interfaces on a port are named using the Virtual Interface
197name from the table above.
198Additional virtual interfaces use a single pair of queues
199for rx and tx as well an additional pair of queues for TOE rx and tx.
200The default is 1.
201.It Va hw.cxgbe.holdoff_timer_idx
202.It Va hw.cxgbe.holdoff_timer_idx_ofld
203Timer index value used to delay interrupts.
204The holdoff timer list has the values 1, 5, 10, 50, 100, and 200
205by default (all values are in microseconds) and the index selects a
206value from this list.
207holdoff_timer_idx_ofld applies to queues used for TOE rx.
208The default value is 1 which means the timer value is 5us.
209Different interfaces can be assigned different values at any time via the
210dev.<port>.X.holdoff_tmr_idx and dev.<port>.X.holdoff_tmr_idx_ofld sysctls.
211.It Va hw.cxgbe.holdoff_pktc_idx
212.It Va hw.cxgbe.holdoff_pktc_idx_ofld
213Packet-count index value used to delay interrupts.
214The packet-count list has the values 1, 8, 16, and 32 by default,
215and the index selects a value from this list.
216holdoff_pktc_idx_ofld applies to queues used for TOE rx.
217The default value is -1 which means packet counting is disabled and interrupts
218are generated based solely on the holdoff timer value.
219Different interfaces can be assigned different values via the
220dev.<port>.X.holdoff_pktc_idx and dev.<port>.X.holdoff_pktc_idx_ofld sysctls.
221These sysctls work only when the interface has never been marked up (as done by
222ifconfig up).
223.It Va hw.cxgbe.qsize_txq
224Number of entries in a transmit queue's descriptor ring.
225A buf_ring of the same size is also allocated for additional
226software queuing.
227See
228.Xr ifnet 9 .
229The default value is 1024.
230Different interfaces can be assigned different values via the
231dev.<port>.X.qsize_txq sysctl.
232This sysctl works only when the interface has never been marked up (as done by
233ifconfig up).
234.It Va hw.cxgbe.qsize_rxq
235Number of entries in a receive queue's descriptor ring.
236The default value is 1024.
237Different interfaces can be assigned different values via the
238dev.<port>.X.qsize_rxq sysctl.
239This sysctl works only when the interface has never been marked up (as done by
240ifconfig up).
241.It Va hw.cxgbe.interrupt_types
242Permitted interrupt types.
243Bit 0 represents INTx (line interrupts), bit 1 MSI, and bit 2 MSI-X.
244The default is 7 (all allowed).
245The driver selects the best possible type out of the allowed types.
246.It Va hw.cxgbe.pcie_relaxed_ordering
247PCIe Relaxed Ordering.
248-1 indicates the driver should determine whether to enable or disable PCIe RO.
2490 disables PCIe RO.
2501 enables PCIe RO.
2512 indicates the driver should not modify the PCIe RO setting.
252The default is -1.
253.It Va hw.cxgbe.fw_install
2540 prohibits the driver from installing a firmware on the card.
2551 allows the driver to install a new firmware if internal driver
256heuristics indicate that the new firmware is preferable to the one
257already on the card.
2582 instructs the driver to always install the new firmware on the card as
259long as it is compatible with the driver and is a different version than
260the one already on the card.
261The default is 1.
262.It Va hw.cxgbe.fl_pktshift
263Number of padding bytes inserted before the beginning of an Ethernet
264frame in the receive buffer.
265The default value is 0.
266A value of of 2 would ensure that the Ethernet payload (usually the IP header)
267is at a 4 byte aligned address.
2680-7 are all valid values.
269.It Va hw.cxgbe.fl_pad
270A non-zero value ensures that writes from the hardware to a receive buffer are
271padded up to the specified boundary.
272The default is -1 which lets the driver pick a pad boundary.
2730 disables trailer padding completely.
274.It Va hw.cxgbe.cong_drop
275Controls the hardware response to congestion.
276-1 disables congestion feedback and is not recommended.
2770 instructs the hardware to backpressure its pipeline on congestion.
278This usually results in the port emitting PAUSE frames.
2791 instructs the hardware to drop frames destined for congested queues.
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 altq 4 ,
401.Xr arp 4 ,
402.Xr ccr 4 ,
403.Xr cxgb 4 ,
404.Xr cxgbev 4 ,
405.Xr netintro 4 ,
406.Xr ng_ether 4 ,
407.Xr ifconfig 8
408.Sh HISTORY
409The
410.Nm
411device driver first appeared in
412.Fx 9.0 .
413Support for T5 cards first appeared in
414.Fx 9.2
415and
416.Fx 10.0 .
417Support for T6 cards first appeared in
418.Fx 11.1
419and
420.Fx 12.0 .
421.Sh AUTHORS
422.An -nosplit
423The
424.Nm
425driver was written by
426.An Navdeep Parhar Aq Mt np@FreeBSD.org .
427