ESXi 8.0 Host PSOD during FastSlab_Alloc and DMA Transfers on vmnic

Products

• VMware vSphere ESXi

Issue/Introduction

• An ESXi 8.0 host experiences a Purple Screen of Death (PSOD). The backtrace typically includes the following functions:
  - `FastSlabAllocMagazineFromBuffer` 
- `FastSlabAllocFreshObj` 
- `PsaStorAllocateFragment` 
- `PsaStorIssueFastPathCommand`
• PSOD Backtrace is similar to 

#0  FastSlabAllocMagazineFromBuffer (count=<synthetic pointer>, flags=FASTSLAB_ALLOC_BUF_MEM, node=1, slab=0x430341306140) at bora/vmkernel/main/fastslab.c:4145
#1  FastSlabAllocFreshObj (flags=FASTSLAB_ALLOC_BUF_MEM, node=1, slab=0x430341306140) at bora/vmkernel/main/fastslab.c:4246
#2  FastSlabCreateObj (slab=slab@entry=0x430341306140, node=1, flags=flags@entry=FASTSLAB_ALLOC_BUF_MEM, keepTrying=keepTrying@entry=0x453a7a299c4f "")
    at bora/vmkernel/main/fastslab.c:4313
#3  0x0000420012735df8 in FastSlabReplenishCPU (slabCpu=slabCpu@entry=0x4200508088c0, count=count@entry=0x420050808918) at bora/vmkernel/main/fastslab.c:4882
#4  0x0000420012736a9b in FastSlabAllocSlow (slabCpu=0x4200508088c0, localMag=0x420050808910, busy=<optimized out>) at bora/vmkernel/main/fastslab.c:5293
#5  0x0000420012bad6f0 in FastSlab_Alloc (cache=<optimized out>) at bora/vmkernel/public/fastslab_public.h:317
#6  PsaStorAllocateFragment (dmaEngine=0x431ed700e740, parentIOFrame=parentIOFrame@entry=0x45da816cac80) at bora/vmkernel/storage/midlayer/storage_path.c:2236
#7  0x0000420012baedb9 in PsaStorCreateFragment (retryable=0x453a7a299dbb "\001", splitReason=PSA_STOR_SPLIT_REASON_MAXXFER_CLONE_BUFFERS, parentFrame=0x45da816cac90, 

 

 

Environment

Esxi 8.0.3

Cause

The issue is caused by physical hardware instability or failure of a network interface card (NIC), specifically affecting DMA (Direct Memory Access) transfers. 

Resolution

1. Physical Reseat:Power down the host and reseat the physical NIC and network cabling associated with the affected vmnic.
2. Hardware Replacement: If the PSOD recurs after a reseat, replace the physical NIC.
3. Validation:Monitor the host vmkernel logs and ensure no further DMA-related allocation failures occur after the hardware change.