[PATCH v2] arm/arm64: KVM: Properly account for guest CPU time

[Christian Borntraeger]

Am 01.06.2015 um 15:35 schrieb Christoffer Dall:
> On Mon, Jun 01, 2015 at 11:21:19AM +0200, Christian Borntraeger wrote:
>> Am 01.06.2015 um 11:08 schrieb Christoffer Dall:
>>
>>>>>
>>>>> Second, looking at the ppc and mips code, they seem to also call
>>>>> kvm_guest_exit() before enabling interrupts, so I don't understand how
>>>>> guest CPU time accounting works on those architectures.
>>>>
>>>> Not an expert here, but I assume mips has the same logic as arm so if your
>>>> patch is right for arm its probably also for mips.
>>>>
>>>> powerpc looks similar to what s390 does (not using the tick, instead it uses
>>>> a hw-timer) so this should be fine.
>>>>
>>> I wonder if we can simply enable HAVE_VIRT_CPU_ACCOUNTING_GEN and get
>>> this for free which would avoid the need for this patch?
>>
>> Asssuming that HAVE_VIRT_CPU_ACCOUNTING_GEN behaves similar to 
>> HAVE_VIRT_CPU_ACCOUNTING on s390/power in respect to not rely on ticks
>> - yes it might work out. Can you give it a try?
>>
> Adding HAVE_VIRT_CPU_ACCOUNTING_GEN to arch/arm64/Kconfig works, but has
> no effect unless you also enable CONFIG_NO_HZ_FULL, so that hardly feels
> like a fix since it would be a shame to force users to use this config
> option to report CPU usage correctly.
> 
> I'm not entirely sure what the history and meaning behind these configs
> are, so maybe there is an entirely different rework needed here.  It
> seems logical that you could simply sample the counter at entry/exit of
> the guest, but if there is nowhere to store this data without
> NO_HZ_FULL+VIRT_CPU_ACCOUNTING_GEN then I guess that would be why?

Given Paolos response that irq_disable/enable is faster than save/restore
at least on x86 your v2 patch might actually be the right thing to do.

Christian