ESXi Running Qedentv Driver Suffers PSOD with Event #PF Exception 14 in World <ID>:ensLBMain

Products

• VMware vSphere ESXi

Issue/Introduction

PSOD event occurs in the ESXi host with below logs in the vmkernel-zdump.log

cpu8:2101868)Backtrace for current CPU #8, worldID=2101868, fp=0x3
cpu8:2101868)0x4539b6f9bab0:[0x4200269912ee]qedentv_multictx_remove_rx_rule@(qedentv)#<None>+0x266 stack: 0x0, 0xffffffff00000000, 0x10, 0x450100000000, 0x4200277f06a9
cpu8:2101868)0x4539b6f9bb60:[0x4200269922e6]qedentv_remove_rx_filter@(qedentv)#<None>+0x157 stack: 0x420043800080, 0x2ca54c, 0x4539e009f680, 0x4539b6f9f140, 0xbf879af37fb04
cpu8:2101868)0x4539b6f9bbd0:[0x4200277a047d]EnsNetQRemoveFilters@(nsxt-ens-20115690)#<None>+0xb2 stack: 0x4501400ad9e0, 0x4501400ada00, 0x4501401fd210, 0x4200277a56ae, 0xa00000006
cpu8:2101868)ESC[45mESC[33;1mVMware ESXi 7.0.3 [Releasebuild-23794027 x86_64]ESC[0m
#PF Exception 14 in world <ID>:ensLBMain IP 0x4200269912ee addr 0x1

Enhanced Datapath(ENS) mode is enabled in the host transport node switch.

Host has below adapter running Qedentv driver 3.#.#.# version.

ThinkSystem Marvell QL41232 10/25GbE SFP28 2-Port PCIe Ethernet Adapter

1077:8070

1077:0059

Environment

VMware ESXi

VMware NSX-T 

Cause

The firmware assert happened due to the TSO constraint violation in the ENS data path and PSOD was triggered due to missing recovery code in ENS mode.

Resolution

The above-mentioned adapters are Lenovo customized qedentv NIC devices, Lenovo and Marvell no longer provide support for this device for ESXi 7.x. But the support is to be extended for ESX 8.x for ethernet functions (no iSCSI and FCoE support).

Option 1: Temporary workaround is to disable the ENS mode.

Refer: https://techdocs.broadcom.com/us/en/vmware-cis/nsx/nsxt-dc/3-2/installation-guide/transport-zones-and-transport-nodes/enhanced-datapath-1.html

Option 2: For ESX 7.x and lower versions

Upgrade ESXi version to 8.x, and find a new Qedentv driver that of the version is higher than the one that has the fix:

ESXi 8.0: Version 3.71.30.0 (Nov 17, 2022)

Option 3: Replace the card with a compatible NIC for the respective OEM of the server, which may be from a vendor other than QLogic(Marvell).