IBM Support

EEH injection fails on NVMe drives without host location code

Flashes (Alerts)


Abstract

On PowerPC systems that use VFIO PCI pass-through devices, error injection is performed from the host that uses the errinjct tool, which targets devices that are based on their location code. The injected error is then detected and handled on the guest side.
However, certain PCI devices, particularly those devices that are connected through NVMe backplane configurations, do not display a valid location code or identifiable device name on the host. This situation prevents host-side error injection for such devices. Also, guest-initiated error injection for VFIO pass-through devices is not yet supported, limiting end-to-end Enhanced Error Handling (EEH) validation from within the guest environment.

Content

Linux Releases Affected

SUSE Linux Enterprise Server (SLES 16.0) and all supported PowerPC platforms that use VFIO PCI pass-through devices.

IBM Systems Affected

All IBM PowerPC (pseries) systems that use VFIO PCI pass-through devices, particularly those with devices that are connected through backplane slots that do not have location code mapping on the host.

Symptoms

In current PowerPC virtualization environments, VFIO pass-through devices rely on host-based error injection that uses the errinjct tool. This tool requires a valid location code to identify and target the device for EEH testing.

For PCI devices that are connected through a NVMe backplane, the host does not display a unique location code or recognizable identifier. As a result, these devices cannot be used for host-initiated EEH error injection.

The guest-side error injection mechanisms for VFIO pass-through devices are not implemented. This limitation prevents comprehensive testing and validation of EEH recovery paths that are initiated from within the guest operating system.

Workaround

No direct methods exist to perform error injection on VFIO pass-through devices without a valid location code.

To validate EEH recovery, use devices that expose valid location codes and are recognized by the host for testing and verification.

Fix Outlook

The fix for this issue will be included in a later release.

I/O devices impacted

EEH testing and validation are affected for all PCI devices that use VFIO for user-space pass-through. These devices include:

  • GPUs
  • Network interface cards (NICs)
  • Storage controllers (such as NVMe drives)
  • Any other hardware bound to vfio-pci for direct guest or user-space assignment

[{"Type":"MASTER","Line of Business":{"code":"LOB68","label":"Power HW"},"Business Unit":{"code":"BU070","label":"IBM Infrastructure"},"Product":{"code":"SGMV168","label":"IBM Support for SUSE Linux Enterprise Server"},"ARM Category":[{"code":"a8m0z000000GnlCAAS","label":"SUSE Linux Enterprise Server"}],"Platform":[{"code":"PF025","label":"Platform Independent"}],"Version":"11.0.0;12.0.0;15.0.0"}]

Document Information

Modified date:
10 November 2025

UID

ibm17247833