Change Log History

Linux Kernel Upstream Release Notes v6.5



Power Limitation

Disabled the option of shutting down the link due to power limitation.

Differentiated Services Code Point (DSCP)

Added support for trusting Differentiated Services Code Point (DSCP) and setting default value for RoCE traffic.

SR-IOV VF Counters

Added a new counter that enables the user to query per Virtual Function counters.

RX Counters

Added the RX out-of-buffer counter to indicate any lack of software receive buffers.


SR-IOV InfiniBand is at beta level.


Enhanced Network Stack (ENS)

Enhanced data path is a networking stack mode, which when configured

provides superior network performance. It is primarily targeted for NFV workloads, which requires the performance benefits provided by this mode. ENS utilizes DPDK Poll Mode driver model and significantly improves packet rate and latency for small message sizes.

This feature is compliant with NSX-T version

Note: This driver version maintains support of all previous Ethernet functionalities, and can operate in both ENS and slow-path network stack mode, based on the DVS configuration.



Removed a VF driver limitation. Now the driver can support the

maximum number of VFs supported by the firmware.

Bug Fixes

See “Bug Fixes History” section.


Explicit Congestion Notification


Explicit Congestion Notification (ECN) is an extension to the Internet

Protocol and to the Transmission Control Protocol. ECN allows end-to- end notification of network congestion without dropping packets.

To configure ECN behavior, download the nmlxcli tool from the Mellanox site.

For further information, refer to the User Manual section Explicit Congestion Notification (ECN).

Dynamic RSS

Improves network performance by allowing OS Load Balancer better

RSS RX queue utilization during heavy traffic of the same type.

For further information, refer to the User Manual section Dynamic RSS.

Multiple RSS Engines

Improves network performance by exposing multiple RSS RX queues to

the hypervisor network stack.

For further information, refer to the User Manual section Multiple RSS Engines.

Packet Capture Utility

Packet Capture utility duplicates all traffic, including RDMA, in its raw

Ethernet form (before stripping) to a dedicated "sniffing" QP, and then passes it to an ESX drop capture point.

It allows gathering of Ethernet and RoCE bidirectional traffic via pktcap- uw and viewing it using regular Ethernet tools, e.g. Wireshark

To enable/disable packet capture, download the nmlxcli tool from the Mellanox site.

For further information, refer to the User Manual section Packet Capture Utility.

SR-IOV max_vfs module

parameter Type Modification

Changed the type of the SR-IOV max_vfs module parameter from a

single integer value to an array of unsigned integers. For further information, refer to the User Manual.

InfiniBand SR-IOV

Enables the creation of InfiniBand virtual functions, allowing the guests

to operate over an InfiniBand fabric.

DCBX Negotiation Support for


PFC port configuration can now be auto-negotiated with switches that

support the DCBX protocol.


Added ESXi CLI support for ESXi 6.7

Adapter Cards

Added support for ConnectX-5/ConnectX-5 Ex adapter cards.

Note: ConnectX-5/ConnectX-5 Ex cards are currently at beta level.

Geneve Stateless Offload

Geneve network protocol is encapsulated into IP frame (L2 tunneling).

Encapsulation is suggested as a means to alter the normal IP routing for datagrams, by delivering them to an intermediate destination that would otherwise not be selected based on the (network part of the) IP Destination Address field in the original IP header.

Remote Direct Memory Access


Remote Direct Memory Access (RDMA) is the remote memory

management capability that allows server-to-server data movement directly between application memory without any CPU involvement. Note: It is recommended to use RoCE with PFC enabled in driver and network switches.

For how to enable PFC in the driver see section Priority Flow Control (PFC) in the User Manual.

Set Link Speed

Enables you to set the link speed to a specific link speed supported by


For further information, see the User Manual section “Set Link Speed”.

Priority Flow Control (PFC)

Applies pause functionality to specific classes of traffic on the Ethernet


For further information, see the User Manual section “Priority Flow Control (PFC)”.


Allows the user to configure multiple hardware queues backing up the

single RX queue. NetQ RSS improves vMotion performance and multiple streams of IPv4/IPv6 TCP/UDP/IPSEC bandwidth over single interface between the Virtual Machines.

For further information, see the User Manual section “NetQ RSS”.

Default Queue RSS (DRSS)

Allows the user to configure multiple hardware queues backing up the

default RX queue. DRSS improves performance for large scale multicast traffic between hypervisors and Virtual Machines interfaces.

For further information, see the User Manual section “Default Queue Receive Side Scaling (DRSS)”.


Single Root IO Virtualization (SR-IOV) is a technology that allows a

physical PCIe device to present itself multiple times through the PCIe bus.

Support for up to 8 ConnectX-4 ports and up to 16 VFs.

For further information, refer to the User Manual

RX/TX Ring Resize

Allows the network administrator to set new RX\TX ring buffer size.

VXLAN Hardware Stateless

Offloads for ConnectX®-4

VXLAN hardware offload enables the traditional offloads to be

performed on the encapsulated traffic.


Enables a host to transmit diagnostic information via the network to a

remote netdump service, which stores it on disk. Network-based coredump collection can be configured in addition to or instead of disk- based coredump collection.


NetQueue is a performance technology in VMware ESXi that

significantly improves performance in Ethernet virtualized environments.


Allows a network administrator to remotely power on a system or to wake

it up from sleep mode

Hardware Offload

  • Large Send Offload (TCP Segmentation Offload)

  • RSS (Device RSS)

Hardware Capabilities

  • Multiple Tx/Rx rings

  • Fixed Pass-Through

  • Single/Dual port

  • MSI-X

Ethernet Network

  • TX/RX checksum

  • Auto moderation and Coalescing

  • VLAN stripping offload

© Copyright 2023, NVIDIA. Last updated on Sep 8, 2023.