Passthrough of Raid Controller to KVM guest results in NMI and unresponsiveness on host
This document (7024342) is provided subject to the disclaimer at the end of this document.
The KVM guests configured with passthrough raid controllers has resulted in NMI and an unresponsive KVM host.
An update of the kernel fixes this issue. Update to the kernel listed below or newer.
A bug in kernel 4.12.14-197.15 caused this to fail in certain cases.
One possible cause of this issue is when the guest driver kicks in and the device is in some low-power state this might cause a PCI error and the NMI.
The below commit, which was included in the newer version of the kernel, provides fixes for Linux to determine power states of PCI devices.
PCI: PM/ACPI: Refresh all stale power state data in pci_pm_complete() (bsc#1149106).
(Upstream commit-id is b51033e06c2ebbad322370f4a35c84488e61b342)
This commit confirms that Linux has the right picture about the power states of PCI devices. It could be the case that Linux' information about the power state of the LSI controller is wrong after the host driver is unloaded and the device attached to VFIO to pass it to the guest.
This Support Knowledgebase provides a valuable tool for SUSE customers and parties interested in our products and solutions to acquire information, ideas and learn from one another. Materials are provided for informational, personal or non-commercial use within your organization and are presented "AS IS" WITHOUT WARRANTY OF ANY KIND.
- Document ID:7024342
- Creation Date: 13-Dec-2019
- Modified Date:23-Apr-2021
- SUSE Linux Enterprise Server
For questions or concerns with the SUSE Knowledgebase please contact: tidfeedback[at]suse.com