New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Intel GPU: Video-related kernel errors, GPU hang (only mouse works) #2836

Closed
marmarek opened this Issue May 31, 2017 · 3 comments

Comments

Projects
None yet
2 participants
@marmarek
Member

marmarek commented May 31, 2017

Qubes OS version (e.g., R3.2):

R4.0
Xen 4.8.1, linux 4.9.29-17


Expected behavior:

No errors, no hangs

Actual behavior:

A lot of errors like:

[drm:gen8_irq_hander [i915]] *ERROR* Fault errors on pipe A 0x00000100

And similarly in Xen messages:

[VT-D]DMAR:[DMA Read] Request device [0000:00:02.0] fault addr 29375f2000, iommu reg = ffff82c0009f2000
[VT-D]DMAR: reason 06 - PTE Read access is not set

Steps to reproduce the behavior:

Boot Qubes 4.0 on Broadwell machine - in this case both Librem 13v1 and Librem 15v2.

General notes:

This appears to be fixed by iommu=no-igfx Xen option. Since we don't have GUI VM yet, it doesn't have much security implications. Maybe we should enable it by default?

@marmarek marmarek added this to the Release 4.0 milestone May 31, 2017

@marmarek

This comment has been minimized.

Show comment
Hide comment
@marmarek

marmarek Jun 2, 2017

Member

@rootkovska what do you think about enabling iommu=no-igfx by default?

Member

marmarek commented Jun 2, 2017

@rootkovska what do you think about enabling iommu=no-igfx by default?

@rootkovska

This comment has been minimized.

Show comment
Hide comment
@rootkovska

rootkovska Jun 2, 2017

Member

I agree it might be justified for 4.0, but we should make sure we won't forget to enable IOMMU for the GPU back in 4.1 when we (hopefully) will get a GPU-passthrough-based GUI domain. Perhaps creating a high-priority ticket for 4.1 milestone should be enough?

Member

rootkovska commented Jun 2, 2017

I agree it might be justified for 4.0, but we should make sure we won't forget to enable IOMMU for the GPU back in 4.1 when we (hopefully) will get a GPU-passthrough-based GUI domain. Perhaps creating a high-priority ticket for 4.1 milestone should be enough?

@marmarek

This comment has been minimized.

Show comment
Hide comment
Member

marmarek commented Jun 2, 2017

Ok. #2841

@qubesos-bot qubesos-bot referenced this issue in QubesOS/updates-status Jul 5, 2017

Closed

installer-qubes-os v25.20.9-5-anaconda (r4.0) #117

fepitre added a commit to fepitre/qubes-installer-qubes-os that referenced this issue Dec 29, 2017

anaconda: disable iommu for IGFX
Many Intel processors (and BIOSes) have invalid IOMMU configuration for
IGFX, which cause multiple problems - from screen glitches, to system
hang.
Since IGFX currently is still in dom0 (isn't isolated from other system
components), disabling IOMMU for it doesn't lower overall security.
When GUI domain will be implemented, we need to re-enable IOMMU here and
hope hardware manufacturers will fix it in the meantime.

Fixes QubesOS/qubes-issues#2836

fepitre added a commit to fepitre/qubes-installer-qubes-os that referenced this issue Dec 30, 2017

anaconda: disable iommu for IGFX
Many Intel processors (and BIOSes) have invalid IOMMU configuration for
IGFX, which cause multiple problems - from screen glitches, to system
hang.
Since IGFX currently is still in dom0 (isn't isolated from other system
components), disabling IOMMU for it doesn't lower overall security.
When GUI domain will be implemented, we need to re-enable IOMMU here and
hope hardware manufacturers will fix it in the meantime.

Fixes QubesOS/qubes-issues#2836
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment