VT-d, iommu, dmar failing - xen/qubesos troubleshooting - thinkpad e15 gen 2

[Bruno_Schroeder]

Please help troubleshoot/fix. A QubesOS/Xen issue.

Xen boot (xl dmesg) will eventually display: Failed to parse ACPI DMAR. Disabling VT-d). This will disable iommu support.

It's a Lenovo E15 Gen 2. The BIOS is the latest and I already ran the lenovo critical update to it. BIOS settings are also correct.

It's strange because I moved the SSD from one laptop to the other and both are exact same model/gen, the only different is the first was ram 8GB, this is 16GB.

I have added to the grub iommu=1 and tried other parameters, but they are not doing the trick. Its an Intel.

Also on:
https://forum.qubes-os.org/t/vt-d-iommu-dmar-failing-xen-troubleshooting-thinkpad-e15-gen-2/42089

https://unix.stackexchange.com/questions/806401/vt-d-iommu-dmar-failing-xen-qubesos-troubleshooting-thinkpad-e15-gen-2

[poddingue]

Your setup is different enough that folks here might not have hit this exact thing.
The Failed to parse ACPI DMAR. Disabling VT-d line usually points at the laptop firmware's ACPI tables rather than anything Xen is doing wrong, which would explain why the grub iommu=1 tweaks aren't moving the needle.
You've already cross-posted to the QubesOS forum, which is probably where the people who've wrestled with E15 firmware quirks hang out.
If it helps, the XCP-ng docs have a short note on reading the DMAR/IVRS ACPI tables to debug IOMMU issues at https://docs.xcp-ng.org/troubleshooting/log-files#dmarivrs-acpi-tables.
It's XCP-ng-specific, but the Xen-level idea of dumping and inspecting those tables carries over.

[Bruno_Schroeder]

Solved!

Issue was, contradicting my notes "Kernel DMA protection" must be on (and apparently is supported by Xen/Qubes).

Attention is to one must boot the machine twice to "stick". Counter-intuitive.

Four days of struggle are over, awkward. Fun thing was to use acpidump, acpixtract and iasl.

[poddingue]

Good to hear, thanks a lot for your feedback.