Re: [xen-unstable test] 57852: regressions - FAIL

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 08/06/15 13:16, Ian Campbell wrote:
> On Mon, 2015-06-08 at 10:27 +0100, Ian Campbell wrote:
>> On Mon, 2015-06-08 at 10:15 +0100, Jan Beulich wrote:
>>>>>> On 08.06.15 at 10:53, <ian.campbell@citrix.com> wrote:
>>>> That's 6/14 (43%) failure rate on fiano0 and 2/10 (20%) on fiano1. Which
>>>> differs form the apparent xen-unstable failure rate. But I wouldn't take
>>>> this as evidence that the two systems differ significantly, despite how
>>>> the unstable results looked at first glance.
>>> So we can basically rule out just one of the hosts being the culprit;
>>> it's either both or our software. Considering that (again at the
>>> example of the recent 4.2 flight) the guest is apparently waiting for
>>> a timer (or other) interrupt (on a HLT instruction), this is very likely
>>> interrupt delivery related, yet (as said before, albeit wrongly for
>>> 4.3) 4.2 doesn't have APICV support yet (4.3 only lack the option
>>> to disable it), so it can't be that (alone).
>>>
>>> Looking at the hardware - are fiano[01], in terms of CPU and
>>> chipset, perhaps the newest or oldest in the pool? (I'm trying to
>>> make myself a picture of what debugging options we have.)
>> I don't know much about the hardware in the pool other than what can be
>> gathered from the serial and dmesg logs.
>>
>> http://logs.test-lab.xenproject.org/osstest/logs/58028/test-amd64-amd64-xl-qemuu-win7-amd64/info.html
>>
>> From the serial log and this:
>>
>> Jun  6 12:09:27.089020 (XEN) VMX: Supported advanced features:
>> Jun  6 12:09:27.089052 (XEN)  - APIC MMIO access virtualisation
>> Jun  6 12:09:27.097051 (XEN)  - APIC TPR shadow
>> Jun  6 12:09:27.097088 (XEN)  - Extended Page Tables (EPT)
>> Jun  6 12:09:27.097118 (XEN)  - Virtual-Processor Identifiers (VPID)
>> Jun  6 12:09:27.105066 (XEN)  - Virtual NMI
>> Jun  6 12:09:27.105100 (XEN)  - MSR direct-access bitmap
>> Jun  6 12:09:27.105130 (XEN)  - Unrestricted Guest
> Running with no-apicv seems to have disabled these three:
>
>> Jun  6 12:09:27.113269 (XEN)  - APIC Register Virtualization
>> Jun  6 12:09:27.113290 (XEN)  - Virtual Interrupt Delivery
>> Jun  6 12:09:27.113328 (XEN)  - Posted Interrupt Processing
> Is that expected?

Yes - The first is APICV itself, and the further two are dependent features.

~Andrew

>
> The adhoc run passed, but that's not statistically significant.
>
> Ian.
>