EMC powerpath Plugin installation results in a PSOD (purple screen of death) event

Products

• VMware vSphere ESXi

Issue/Introduction

Server crash was never observed in prior releases of ESXi. Issue is only observed when the EMC powerpath plugin is installed. 
We also observed this behavior on the NVMe setup but not limited to NVMe.

 

During the Installation of EMC powerpath plugin version 7.1, ESXi server running on version 7.0 U1 and 7.0 U2 crashes with a PSOD (purple screen of death) error.

Environment

VMware vSphere ESXi 7.0.0

Cause

During installation of powerpath plugin version 7.1, ESXi hosts of version 7.0U1C fail with a PSOD.
Cause: lpfc_external_dif_cmpl() checks the response data of an INQUIRY command, but it assume the command's SG array only has one element, i.e. the data buffer is on physical contiguous pages. 
This assumption holds most of the time, but sometimes doesn't.  When INQUIRY command has two SG elements, lpfc_external_dif_cmpl() will the second element represented memory without mapping it, which causes Page Fault.

Server crash was observed when the lpfc driver is at version: 12.8.351.30-1OEM.700.1.0.15843807.

Resolution

PSOD that was triggered during the EMC powerpath VIB installation was due to the existing LPFC driver (12.8.329.0-1OEM.700.1.0.15843807) provided by Broadcom. 

Upgrading to lpfc driver version  12.8.351.30 or above will resolve this issue. 

Additional Information

Impact/Risks:
We will not be able to use the EMC powerpath plugin unless the FC HBA driver is upgraded.