Re: [PATCH 3/6] KVM: X86: Make tsc_delta calculation a function of guest tsc

[Zachary Amsden <zamsden@redhat.com>]

>
> On 2011-04-19 08:46, Roedel, Joerg wrote:
>    
>> On Mon, Apr 18, 2011 at 08:02:35PM -0400, Zachary Amsden wrote:
>>      
>>>> On Sat, Apr 16, 2011 at 06:09:17PM +0200, Jan Kiszka wrote:
>>>>
>>>>          
>>>>> On 2011-03-25 09:44, Joerg Roedel wrote:
>>>>>
>>>>>            
>>>>
>>>>          
>>>>>> +		tsc_delta = !vcpu->arch.last_guest_tsc ? 0 :
>>>>>> +			     tsc - vcpu->arch.last_guest_tsc;
>>>>>>
>>>>>>              
>>>>
>>>>          
>>>>> This patch appears to cause troubles to Linux guests on TSC clocksource
>>>>> and APIC highres timer. The first boot after qemu start is always fine,
>>>>> but after a reboot the guest timer appears to fire incorrectly or even
>>>>> not at all.
>>>>>
>>>>> Was this patch tested with a guest reboot scenario as well? Does it
>>>>> account for the TSC being reset to 0 on reboot?
>>>>>
>>>>>            
>>>> Hmm, probably the last_guest_tsc is not updated correctly in this
>>>> scenario. I will have a look tomorrow.
>>>>
>>>> 	Joerg
>>>>
>>>>
>>>>          
>>> To avoid this problem, when the TSC is reset, the overshoot protection
>>> where last_guest_tsc is used is specifically disabled:
>>>
>>>           /* Reset of TSC must disable overshoot protection below */
>>>           vcpu->arch.hv_clock.tsc_timestamp = 0;
>>>           vcpu->arch.last_tsc_write = data;
>>>           vcpu->arch.last_tsc_nsec = ns;
>>>
>>> You can probably use the same test - last_guest_tsc is only valid if
>>> tsc_timestamp above != 0.
>>>        
>> Yes, this should also work. But the other way we get rid of
>> last_host_tsc which is also a good thing :)
>>      
> That reminds me: I think we still have the bug that KVM overeagerly
> declares the host's TSC unstable after resume from S3 because the TSC
> appears to go backward when KVM checks. Or should this have been fixed now?
>
>    

That should have been fixed... let me check if the patch has made it 
upstream.