Required kernel parameter `iommu=pt` for Dell EMC's AMD Rome based servers equipped with Broadcom 5720 LOM

Solution In Progress - Updated -

Environment

  • Red Hat Enterprise Linux 7
  • Dell EMC PowerEdge servers with AMD EPYC CPUs and Broadcom 5720 LOM

Issue

  • tg3 kdump failing via NFS on Dell EMC server with AMD Rome.
  • iommu=pt required for kdump over NFS on AMD Rome systems

Resolution

As a workaround, customers can add iommu=pt to the kernel command line. For reference, iommu=pt is the default on RHEL 8, and has positive performance implications for many workloads.

Root Cause

A bug has been identified in the tg3 driver, it is likely a race/timing issue with the system kexec-ing, and the IOMMU still being used by a device for active DMA.

Red Hat Engineering is working with Broadcom, Dell, and upstream to find a permanent solution.

This solution is part of Red Hat’s fast-track publication program, providing a huge library of solutions that Red Hat engineers have created while supporting our customers. To give you the knowledge you need the instant it becomes available, these articles may be presented in a raw and unedited form.

Comments