ESXi PSOD (purple screen of death) with "Unrecoverable error in elxnet ... elxnet_getRxPfragInfo"

Products

• VMware vSphere ESXi

Issue/Introduction

• The ESXi host experiences a Purple Screen of Death (PSOD) with an error message containing "Unrecoverable error in elxnet ... elxnet_getRxPfragInfo"
• This affects hosts utilizing Emulex OneConnect Network Adapters with NetQueue Receive Side Scaling (RSS) enabled.
• The crash is typically triggered during state transitions, specifically:
  • elxnet NIC link up/down events
  • NIC resets
  • Host shutdown or reboot operations

Environment

VMware vSphere ESXi 

Cause

The issue is caused by incorrect handling of RSS queue cleanup by OneConnect Network Adapter firmware.

Resolution

Currently there is no resolution. Future firmware updates of the OneConnect network adapters may resolve the issue.

Workaround:

• Netqueue RSS is disabled by default on recent elxnet drivers for these network adapters. So unless it is manually enabled by changing elxnet's "RSS" module parameter to values greater than 0, the issue would not occur.
• Use "esxcli system module parameters list -m elxnet" to find the current module parameter values of elxnet.