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, 2022 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. 2802 instructs the hardware to both backpressure the pipeline and drop frames. 281.It Va hw.cxgbe.pause_settings 282PAUSE frame settings. 283Bit 0 is rx_pause, bit 1 is tx_pause, bit 2 is pause_autoneg. 284rx_pause = 1 instructs the hardware to heed incoming PAUSE frames, 0 instructs 285it to ignore them. 286tx_pause = 1 allows the hardware to emit PAUSE frames when its receive FIFO 287reaches a high threshold, 0 prohibits the hardware from emitting PAUSE frames. 288pause_autoneg = 1 overrides the rx_pause and tx_pause bits and instructs the 289hardware to negotiate PAUSE settings with the link peer. 290The default is 7 (all three = 1). 291This tunable establishes the default PAUSE settings for all ports. 292Settings can be displayed and controlled on a per-port basis via the 293dev.<port>.X.pause_settings sysctl. 294.It Va hw.cxgbe.fec 295Forward Error Correction settings. 296-1 (default) means driver should automatically pick a value. 2970 disables FEC. 298Finer grained control can be achieved by setting individual bits. 299Bit 0 enables RS FEC, bit 1 enables BASE-R FEC (aka Firecode FEC), bit 3002 enables NO FEC, and bit 6 enables the FEC that is recommended by the 301transceiver/cable that is plugged in. 302These bits can be set together in any combination. 303This tunable establishes the default FEC settings for all ports. 304Settings can be controlled on a per-port basis via the 305dev.<port>.X.requested_fec sysctl. 306The FEC in use on the link is available in dev.<port>.X.link_fec when 307the link is up. 308.It Va hw.cxgbe.autoneg 309Link autonegotiation settings. 310This tunable establishes the default autonegotiation settings for all ports. 311Settings can be displayed and controlled on a per-port basis via the 312dev.<port>.X.autoneg sysctl. 3130 disables autonegotiation. 3141 enables autonegotiation. 315The default is -1 which lets the driver pick a value. 316dev.<port>.X.autoneg is -1 for port and module combinations that do not support 317autonegotiation. 318.It Va hw.cxgbe.buffer_packing 319Allow the hardware to deliver multiple frames in the same receive buffer 320opportunistically. 321The default is -1 which lets the driver decide. 3220 or 1 explicitly disable or enable this feature. 323.It Va hw.cxgbe.largest_rx_cluster 324.It Va hw.cxgbe.safest_rx_cluster 325Sizes of rx clusters. 326Each of these must be set to one of the sizes available 327(usually 2048, 4096, 9216, and 16384) and largest_rx_cluster must be greater 328than or equal to safest_rx_cluster. 329The defaults are 16384 and 4096 respectively. 330The driver never attempts to allocate a receive buffer larger than 331largest_rx_cluster and falls back to allocating buffers of 332safest_rx_cluster size if an allocation larger than safest_rx_cluster fails. 333Note that largest_rx_cluster merely establishes a ceiling -- the driver is 334allowed to allocate buffers of smaller sizes. 335.It Va hw.cxgbe.config_file 336Select a pre-packaged device configuration file. 337A configuration file contains a recipe for partitioning and configuring the 338hardware resources on the card. 339This tunable is for specialized applications only and should not be used in 340normal operation. 341The configuration profile currently in use is available in the dev.<nexus>.X.cf 342and dev.<nexus>.X.cfcsum sysctls. 343.It Va hw.cxgbe.linkcaps_allowed 344.It Va hw.cxgbe.niccaps_allowed 345.It Va hw.cxgbe.toecaps_allowed 346.It Va hw.cxgbe.rdmacaps_allowed 347.It Va hw.cxgbe.iscsicaps_allowed 348.It Va hw.cxgbe.fcoecaps_allowed 349Disallowing capabilities provides a hint to the driver and firmware to not 350reserve hardware resources for that feature. 351Each of these is a bit field with a bit for each sub-capability within the 352capability. 353This tunable is for specialized applications only and should not be used in 354normal operation. 355The capabilities for which hardware resources have been reserved are listed in 356dev.<nexus>.X.*caps sysctls. 357.It Va hw.cxgbe.tx_vm_wr 358Setting this to 1 instructs the driver to use VM work requests to transmit data. 359This lets PF interfaces transmit frames to VF interfaces over the internal 360switch in the ASIC. 361Note that the 362.Xr cxgbev 4 363VF driver always uses VM work requests and is not affected by this tunable. 364The default value is 0 and should be changed only if PF and VF interfaces need 365to communicate with each other. 366Different interfaces can be assigned different values using the 367dev.<port>.X.tx_vm_wr sysctl when the interface is administratively down. 368.It Va hw.cxgbe.attack_filter 369Set to 1 to enable the "attack filter". 370Default is 0. 371The attack filter will drop an incoming frame if any of these conditions is 372true: src ip/ip6 == dst ip/ip6; tcp and src/dst ip is not unicast; src/dst ip is 373loopback (127.x.y.z); src ip6 is not unicast; src/dst ip6 is loopback (::1/128) 374or unspecified (::/128); tcp and src/dst ip6 is mcast (ff00::/8). 375This facility is available on T4 and T5 based cards only. 376.It Va hw.cxgbe.drop_ip_fragments 377Set to 1 to drop all incoming IP fragments. 378Default is 0. 379Note that this drops valid frames. 380.It Va hw.cxgbe.drop_pkts_with_l2_errors 381Set to 1 to drop incoming frames with Layer 2 length or checksum errors. 382Default is 1. 383.It Va hw.cxgbe.drop_pkts_with_l3_errors 384Set to 1 to drop incoming frames with IP version, length, or checksum errors. 385The IP checksum is validated for TCP or UDP packets only. 386Default is 0. 387.It Va hw.cxgbe.drop_pkts_with_l4_errors 388Set to 1 to drop incoming frames with Layer 4 (TCP or UDP) length, 389checksum, or other errors. 390Default is 0. 391.El 392.Sh SUPPORT 393For general information and support, 394go to the Chelsio support website at: 395.Pa http://www.chelsio.com/ . 396.Pp 397If an issue is identified with this driver with a supported adapter, 398email all the specific information related to the issue to 399.Aq Mt support@chelsio.com . 400.Sh SEE ALSO 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