Re: [PATCH 3/5] kvm: Directly account vtime to system on guest switch

[Frederic Weisbecker <fweisbec@gmail.com>]

2012/10/25 Christian Borntraeger <borntraeger@de.ibm.com>:
> On 25/10/12 02:51, Frederic Weisbecker wrote:
>> Switching to or from guest context is done on ioctl context.
>> So by the time we call kvm_guest_enter() or kvm_guest_exit()
>> we know we are not running the idle task.
>>
>> As a result, we can directly account the cputime using
>> vtime_account_system_irqsafe().
>>
>> There are two good reasons to do this:
>>
>> * We avoid some useless checks on guest switch. It optimizes
>> a bit this fast path.
>>
>> * In the case of CONFIG_IRQ_TIME_ACCOUNTING, calling vtime_account()
>> checks for irq time to account. This is pointless since we know
>> we are not in an irq on guest switch. This is wasting cpu cycles
>> for no good reason. vtime_account_system() OTOH is a no-op in
>> this config option.
>>
>> * s390 doesn't disable irqs in its implementation of vtime_account().
>> If vtime_account() in kvm races with an irq, the pending time might
>> be accounted twice. With vtime_account_system_irqsafe() we are protected.
>
> We disable irqs before we call kvm_guest_enter/exit, see kvm-s390.c:
>
>
> [...]
>         if (!kvm_is_ucontrol(vcpu->kvm))
>                 kvm_s390_deliver_pending_interrupts(vcpu);
>         vcpu->arch.sie_block->icptcode = 0;
>         local_irq_disable();
>         kvm_guest_enter();
>         local_irq_enable();
> [...]
>

Ah ok. Hmm I still need to keep it irqsafe for the other archs though,
as it is currently with vtime_account(). So perhaps I can remove your
local_irq_disable there and use vtime_account_system_irqsafe()
instead?

thanks.