1 I40E(7D) Devices I40E(7D) 2 3 NAME 4 i40e - Intel 710/722 Ethernet Device Driver 5 6 SYNOPSIS 7 /dev/net/i40e* 8 9 DESCRIPTION 10 The i40e driver is a GLDv3, multi-threaded, clonable, loadable device 11 driver that supports the Data Link Provider Interface, dlpi(7P). The 12 i40e driver supports the Intel 710 and 722 Ethernet Controller families 13 of networking interface cards which support speeds of 1 GbE, 2.5 GbE, 5 14 GbE, 10 GbE, 25 GbE, and 40 GbE. 15 16 In addition to basic device initialization and the sending and receiving 17 of frames, it supports the following features: 18 19 - Jumbo frames up to 9710 bytes. 20 21 - Promiscuous access via snoop(1M) and dlpi(7P) 22 23 - IPv4 Checksum Offload 24 25 - TCP, UDP, and SCTP checksum offload 26 27 - IPv4 and IPv6 TCP Segmentation offload 28 29 At this time, the i40e driver does not enable the use of energy efficient 30 Ethernet (EEE) or support the use of flow control through hardware pause 31 frames. 32 33 APPLICATION PROGRAMMING INTERFACE 34 For each device supported by the i40e installed in the system, a 35 character-special file will be created. This file supports the Data Link 36 Provider Interface (DLPI) which is documented in dlpi(7P). For most 37 consumers, the use of libdlpi(3LIB), is recommended. 38 39 Each instance is assigned a unique ascending integer identifier. A 40 device which has multiple ports may appear to the system as separate 41 instances. The system does not provide a guarantee on how these will be 42 presented. Using this instance identifier, one can determine the exact 43 character-special file to open. For example, the first instance 44 enumerated in the system, with id 0, would be named i40e0. It exists in 45 the file system at /dev/net/i40e0. 46 47 CONFIGURATION 48 The i40e driver always performs auto-negotiation and depending on the 49 model may negotiate to 40 Gbps, 25 Gbps, 10 Gbps, or 1 Gbps. At this 50 time, the driver requires the use of auto-negotiation. 51 52 The i40e driver is managed by the dladm(1M) utility. dladm(1M) is the 53 preferred interface for setting all properties. While driver.conf(4) 54 based configuration is possible, dladm(1M) is recommended. The i40e 55 driver may be joined into an aggregation based on the link aggregation 56 control protocol (LACP) through dladm(1M). 57 58 PROPERTIES 59 The device supports the following properties which may be tuned through 60 its driver.conf file, /kernel/drv/i40e.conf. Most of these properties 61 cannot be changed after the device has been started. The device is 62 started in response to a DLPI consumer opening the device and binding to 63 it. This happens when an IP interfaces is plumbed or another dlpi(7P) 64 consumer such as snoop(1M) or an LLDP daemon is started. 65 66 Some properties may be tuned at runtime with the dladm(1M) utility. 67 Properties that can be will have the name of the dladm property called 68 out explicitly. 69 70 These properties are not considered stable at this time. They may change 71 and should not be relied on. They are considered Volatile. It is not 72 expected that administrators of the system will have to tune these 73 values. 74 75 default_mtu 76 Minimum: 1500 | Maximum: 9710 | Runtime Property: mtu 77 78 The default_mtu property determines the starting MTU of the 79 various device instances. Note that the device's MTU also 80 determines the upper bound of the MTU of all VNICs created over 81 the device. The default MTU is 1500. 82 83 mr_enable 84 Minimum: 0 | Maximum: 1 85 86 The mr_enable property determines whether or not support for 87 multiple rings is enabled for the device. The default is always 88 to enable them. It is not recommended to to disable them. 89 90 rx_num_groups 91 Minimum: 1 | Maximum: 32 92 93 The rx_num_groups property determines the number of receive mac 94 groups provided by the driver. Each group can handle all unicast 95 traffic for a single MAC address, more groups means more unicast 96 traffic that can be steered by hardware. However, more groups 97 also means more demand for kernel memory. If you are not making 98 heavy use of VNICs, or do not need the efficiency gains of 99 hardware steering, then reducing this number can reduce kernel 100 memory taken by i40e. 101 102 rx_ring_size 103 Minimum: 64 | Maximum: 4096 104 105 The rx_ring_size property determines the number of descriptors 106 that will be used in each receive ring on the card. 107 Administrators should not normally need to tune this value. 108 Hardware requires that the ring size be a multiple of 32. The 109 system will round up the set value to the nearest multiple of 32. 110 111 tx_ring_size 112 Minimum: 64 | Maximum: 4096 113 114 The tx_ring_size property determines the number of descriptors 115 that will be used in each transmit ring on the card. 116 Administrators should not normally need to tune this value. 117 Hardware requires that the ring size be a multiple of 32. The 118 system will round up the set value to the nearest multiple of 32. 119 120 tx_resched_threshold 121 Minimum: 8 | Maximum: Variable 122 123 The tx_resched_threshold property determines the number of 124 descriptors that must be available for a frame to be transmitted. 125 The maximum is variable. It is dependent on the value of the 126 tx_ring_size property. At least eight descriptors must be 127 available for the device to function correctly. 128 129 rx_limit_per_intr 130 Minimum: 16 | Maximum: 4096 131 132 The rx_limit_per_intr property determines the maximum number of 133 packets that will be processed on a given ring during a single 134 interrupt. This is done to try and guarantee some amount of 135 liveness in the system. It is not expected that administrators 136 will have to tune this value. 137 138 tx_hcksum_enable 139 Minimum: 0 | Maximum: 1 140 141 The tx_hcksum_enable property controls whether or not the device 142 enables support for hardware checksumming of outgoing packets. 143 The default is to always enable support for this. Turning it off 144 will increase latency and decrease throughput when transmitting 145 packets, but should be done if a hardware bug is suspected. 146 147 rx_hcksum_enable 148 Minimum: 0 | Maximum: 1 149 150 The rx_hcksum_enable property controls whether or not the device 151 enables support for hardware checksumming of incoming packets. 152 The default is to always enable support for this. Turning it off 153 will increase latency and decrease throughput when receiving 154 packets, but should be done if a hardware bug is suspected. 155 156 rx_dma_threshold 157 Minimum: 0 | Maximum: INT32_MAX | Runtime Property: 158 _rx_dma_threshold 159 160 The rx_dma_threshold indicates the size in bytes of a received 161 frame, including all of its headers, at which the driver should 162 not copy the frame but instead bind DMA memory. By setting this 163 property to its minimum, all frames will be processed with DMA 164 binding. By setting this property to its maximum, all frames 165 will be processed by copying the frame. 166 167 tx_lso_enable 168 Minimum: 0 | Maximum: 1 169 170 The tx_lso_enable property controls whether or not the device 171 enables support for Large Segment Offloand (LSO) when 172 transmitting packets. The default is to always enable support 173 for this. Turning it off will decrease throughput when 174 transmitting packets, but should be done if a hardware bug is 175 suspected. 176 177 ARCHITECTURE 178 The i40e driver is only supported on x86 systems at this time. 179 180 FILES 181 /dev/net/i40e* Per-instance character device. 182 183 /kernel/drv/amd64/i40e Device driver (x86) 184 185 /kernel/drv/i40e.conf Driver configuration file 186 187 SEE ALSO 188 dladm(1M), snoop(1M), driver.conf(4), dlpi(7P) 189 190 illumos January 18, 2021 illumos