Re: [Xen-devel] 4.9.19 (and others) hang when running as a nested dom0

[George Dunlap <george.dunlap@xxxxxxxxxx>]

On 30/03/17 17:53, Juergen Gross wrote:
> On 30/03/17 18:32, George Dunlap wrote:
>> I normally test CentOS using nested virtualization (mainly because
>> it's easy to restore the disk to a known state).  At some point in the
>> last six months this basically stopped working, as newer kernels (and
>> even backports) stopped booting properly under Xen.
>>
>> I've finally gotten around to building my own kernel outside of the
>> normal RPM / CentOS build system to reproduce the issue.
>>
>> I downloaded the 4.9.19 tarball and compiled under CentOS 6 with the
>> attached config.
>>
>> The console output for the VM when running as a simple domU is
>> attached as native-vm.console.
>>
>> Then console output when running with Xen inside the VM and the Linux
>> kernel as a dom0 inside that guest is attached as L1-dom0.console.
>> (That is, the dom0 for L1, as opposed to the dom0 for L0.  Not clear
>> how the terminology should apply here.)
>>
>> As you can see, native-vm boots to completion, whereas L1 hangs just
>> after deciding to use clocksource as the TSC.
>>
>> The guest appears to be mostly but not completely idle -- as one would
>> expect if dom0 just decided it didn't have anything to do.
>>
>> Any ideas on how to proceed?
> 
> Try upstream commit da72ff5bfcb02c6ac8b169a7cf597a3c8e6c4de1.

That seems to have done the trick, thanks!

 -George


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel