Tails

#1 Updated by intrigeri 2014-05-07 23:32:36

It would be useful to know what version of libvirt and qemu is being used. IIRC, recent qemu / Linux have KMS support for Cirrus, so this may be relevant.

Also, can this be reproduced outside of the context of the test suite, with the same libvirt/qemu version? The answer to this question probably impacts the priority of this issue.

#2 Updated by intrigeri 2014-05-07 23:49:05

Possibly relevant:

• http://linux-kernel.2935.n7.nabble.com/3-14-rc1-cirrus-driver-problem-qemu-td796638.html (not checkd if it was fixed in 3.14 final)
• http://cgit.freedesktop.org/~agd5f/linux/commit/?h=drm-next-3.14-wip&id=99d4a8ae93ead27b5a88cdbd09dc556fe96ac3a8 (not checked if this commit eventually landed into 3.14 at all)

#3 Updated by anonym 2014-05-08 12:59:48

intrigeri wrote:
> It would be useful to know what version of libvirt and qemu is being used.

All is from wheezy-backports:

• qemu 2.0.0+dfsg-4~bpo70+1

• libvirt* 1.2.1-1~bpo70+1

> IIRC, recent qemu / Linux have KMS support for Cirrus, so this may be relevant.

For the record, disabling KMS doesn’t workaround the issue.

> Also, can this be reproduced outside of the context of the test suite, with the same libvirt/qemu version?

Yes. In one of my own x86_64 VMs all I have to do is switching from my normal “qxl” to “cirrus” and the error appears. Then switching to i686 fixes it.

While re-testing, I found another workaround (i.e. so x86_64 works with cirrus) by adding:

  <cpu mode='custom' match='exact'>
    <model>SandyBridge</model>
    <vendor>Intel</vendor>
  </cpu>

so I guess the real issue is with some CPU feature missing.

> Possibly relevant:
>
> * http://linux-kernel.2935.n7.nabble.com/3-14-rc1-cirrus-driver-problem-qemu-td796638.html (not checkd if it was fixed in 3.14 final)

Perhaps, but I do not get a kernel trace, only a bus error and Xorg trace.

> * http://cgit.freedesktop.org/~agd5f/linux/commit/?h=drm-next-3.14-wip&id=99d4a8ae93ead27b5a88cdbd09dc556fe96ac3a8 (not checked if this commit eventually landed into 3.14 at all)

From the referenced bug it’s said that nomodeset is a workaround, but this doesn’t work for me.

#4 Updated by intrigeri 2014-05-08 14:19:18

Thanks for the info and additional research!

My current take on this is:

• this ticket should be cloned to cover the test suite case on the one hand, and the normal usage OTOH
• as far as the test suite is concerned, you have found two acceptable workarounds, so case (almost) closed
• as far as broader usage of Tails is concerned, I think documenting the workarounds on the known issues page should be just enough.

#5 Updated by anonym 2014-05-08 20:49:33

intrigeri wrote:
> * this ticket should be cloned to cover the test suite case on the one hand, and the normal usage OTOH
> * as far as the test suite is concerned, you have found two acceptable workarounds, so case (almost) closed
> * as far as broader usage of Tails is concerned, I think documenting the workarounds on the known issues page should be just enough.

Luckily virt-manager defaults (at least in Wheezy’s version) to qxl when creating new VMs, so maybe we can just completely drop this without bloating the known issues page furher?

#6 Updated by intrigeri 2014-05-08 20:55:11

> Luckily virt-manager defaults (at least in Wheezy’s version) to qxl when creating new
> VMs, so maybe we can just completely drop this without bloating the known issues
> page furher?

Agreed.

#8 Updated by intrigeri 2014-07-01 11:21:20

Added to known issues, so that regular users know about the workarounds too.