Re: [xen-unstable test] 17613: regressions - FAIL

[Keir Fraser <keir.xen@gmail.com>]

On 11/04/2013 08:54, "Jan Beulich" <JBeulich@suse.com> wrote:

>>>> On 11.04.13 at 01:41, xen.org <ian.jackson@eu.citrix.com> wrote:
>> flight 17613 xen-unstable real [real]
>> http://www.chiark.greenend.org.uk/~xensrcts/logs/17613/
>> 
>> Regressions :-(
>> 
>> Tests which did not succeed and are blocking,
>> including tests which could not be run:
>>  test-amd64-i386-qemut-rhel6hvm-intel  9 guest-start.2     fail REGR. vs.
>> 17610
> 
> So this is one of the BUG_ON()s added by bd9be94 triggering,
> and this is not a false positive. scale_delta() is simply not suitable
> for negative inputs, and the main question is why gtime_to_gtsc()
> caps its input to 0 only for PV domains, or whether (considering
> that this parameter is u64) __update_vcpu_system_time() should
> already cap it to zero. Not to speak of the question how the
> calculation there could end up negative in the first place.

Yes, I think it should be capped where guest time gets set/updated. The time
input to gtime_to_gtsc() should always be non-negative.

 -- Keir

> Stefano, Keir (21445:c1ed00d49534)?
> 
> Jan
>