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