ESXi PSOD occurs with PsaNvmeAddControllerInt when an NVMe device fails to be added due to insufficient system's low memory

Products

• VMware vSphere ESXi 8.0

Issue/Introduction

ESXi host may fail with a PSOD citing PsaNvmeAddControllerInt in the backtrace :

@BlueScreen: #PF Exception 14 in world #######:vmknvmeGener IP ############## addr 0x#########
PTEs:0x#########;0x0;
Code start: 0x############ VMK uptime: ##:##:##:##.###
[0x############]PsaNvmeAddControllerInt@vmkernel#nover+0x4ca stack: 0x############
[0x############]vmk_PsaNvmeScanAndClaimPaths@vmkernel#nover+0x154 stack: 0x############
[0x############][email protected]#0.0.0.1+0x101 stack: 0x############
[0x############][email protected]#0.0.0.1+0x1b stack: 0x############
[0x############][email protected]#0.0.0.1+0x10 stack: 0x############
[0x############]HelperQueueFunc@vmkernel#nover+0x300 stack: 0x############
[0x############]CpuSched_StartWorld@vmkernel#nover+0xbf stack: 0x############
[0x############]Debug_IsInitialized@vmkernel#nover+0xc stack: 0x############

The following 'Out of memory' message is recorded in the memory dump file's log file just before the PSOD :

WARNING: NvmeDiscover: 1079: PagePool_Alloc failed with status Out of memory while trying to allocate for size #######

 

Environment

VMware vSphere ESXi 8.0

Cause

When the system attempts to add an NVMe controller following a PCIe hot-plug insert event, a PSOD may occur if the system's low memory (the area of physical memory addresses less than 4GB) is exhausted and memory allocation fails.

Resolution

Currently, there is no resolution for this issue.

Broadcom engineering team is working on the fix.