Intel® Network Adapter Driver for PCIe* 40 Gigabit Ethernet Network Connections under Linux*

18026
3/9/2020

Introduction

This download record includes the i40e Linux* base driver version 2.11.21 for the 700 series devices.

Available Downloads

  • Linux*
  • Size: 567.4 KB
  • SHA1: 2359E0EB5C0BA1D68488044EADCA54316D0D8C89

Detailed Description

Overview

The driver in this release is compatible with the 700 series devices for controller-based Intel® Ethernet Network Connections.

What's new?

  • Fix minimized amount of busy-waiting during AQ send on Linux
  • Fix compilation issue on latest 5.5.x kernels
  • Fix sysfs for VF promiscuous mode
  • Fix link speed reported after dis/en VF via link_state
  • Fix proper loopback value after iavf driver reload
  • Fix default value of allow_untagged flag to true
  • Fix removal of PF to VF mirror rules
  • Fix displaying queue statistics shown by ethtool
  • Fix reload VF driver after configured vlan if trunk was set
  • Fix get/set trunk when port VLAN is enabled
  • Fix link up/down procedure
  • Fix printing of actual link speed for 2.5 and 5 Gbps
  • Fix VF failed during set MAC on VF with set mirroring
  • Fix check mirror configuration
  • Fix opcode msg on unsuccessful VF MAC change
  • Fix incorrect flags for egress mirror
  • Fix general protection fault during remove VLAN from VF
  • Fix wrong error OK message during remove VF
  • Fix restore mirror config after iavf reload
  • Fix VF failed during changing MAC on VF with VLAN mirror
  • Fix PF wait reset loop, make it more reliable
  • Fix VF to VF ingress/egress mirroring
  • Fix defect with adding/deleting ingress/egress mirror
  • Fix for reading lpi duration statistics from FW
  • Fix compilation under RHEL 8.1
  • Add support for UDP source cloud filters
  • Add preliminary VF-D 1.5 specification support
  • Add of sysfs for max_tx_rate
  • Add build time choise of pf->vf speed reporting protocol
  • Add log entry when a NIC enters recovery mode after reboot
  • Add Enable/Disable DCB on X710-T*L
  • Add support F/W API version up to 1.10
  • Add missing 0 length checks to a couple RSS opcodes
  • Add missing AQ command buffer for sriov/loopback setup
  • Add error handling to the i40e_set_vf_port_vlan function
  • Add omit TC creation if first queue is not power of two
  • Add support for text sending error msg during adding filter for VF
  • Add MSI-X checks during creating tc queue maps

And some minor bugfixes.

Known limitations:

Incorrect reporting of link speed on older/DPDK VF drivers
----------------------------------------------------------
For i40evf/iavf drivers older than 3.9.3 and all DPDK i40e VF drivers, this driver needs to be built with extra parameter (make CFLAGS_MODULE='-DNO_VF_CAP_ADV_LINK_SPEED') in order to enable correct reporting of the link speed for VFs. This will have side effect of incorrect reporting link speeds of 2,5 and 5 Gbps.

Limited support of 802.1ad VLAN traffic
---------------------------------------
For packets tagged per 802.1ad standard (QinQ) certain SR-IOV Hypervisor Management Interface features like vlan_mirror and trunk will not work.

VFs inability to capture packets directed to another VF
-------------------------------------------------------
Virtual functions will not be able to capture packets with destination MAC of another VF unless vf-true-promisc-support private flag is set.

vlan_anti_spoof doesn't affect ability to send untagged packets
---------------------------------------------------------------
When vlan_anti_spoof option is set using SR-IOV Hypervisor Management Interface, Virtual Functions will be still able to send packets without VLAN tags.

Limited capabilities of viewing MAC address list
------------------------------------------------
Because of the buffer size limit (kernel limitation) it is possible to add more MAC addresses to the mac_list then it is possible to display using #cat.
Therefore it is not recommended to add more than 180 MAC addresses.

Maximum queue number per VF
---------------------------
num_queues feature of SR-IOV Hypervisor Management Interface has no protection against overutilization of the available queues among VFs.

 

See the readme for installation instructions, how to identify your adapter, command line parameters, additional features, and known issues.