[PATCH 0/1] KVM: more halt polling

[PATCH 0/1] KVM: more halt polling

[Christian Borntraeger]

Paolo, Radim,

looks like I messed up. While performance showed variant 2 to be
better (very aggressive shrinking) I send you variant 1 last week.

Please consider this addon patch to improve the halt polling for
s390 even further. (variant 1 is still better than the situation
before, so not a big issue, but still).



Christian Borntraeger (1):
  KVM: shrink halt polling even more for invalid wakeups

 virt/kvm/kvm_main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

-- 
2.5.5

[PATCH 1/1] KVM: shrink halt polling even more for invalid wakeups

[Christian Borntraeger]

commit 3491caf2755e ("KVM: halt_polling: provide a way to qualify
 wakeups during poll") added more aggressive shrinking of the
polling interval if the wakeup did not match some criteria. This
still allows to keep polling enabled if the polling time was
smaller that the current max poll time (block_ns <= vcpu->halt_poll_ns).
Performance measurement shows that even more aggressive shrinking
(shrink polling on any invalid wakeup) reduces absolute and relative
(to the workload) CPU usage even further.

Cc: David Matlack <dmatlack@google.com>
Cc: Wanpeng Li <kernellwp@gmail.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
CC: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
---
 virt/kvm/kvm_main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index fbd7698..dec1ed2 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2055,12 +2055,13 @@ void kvm_vcpu_block(struct kvm_vcpu *vcpu)
 out:
 	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);
 
-	if (halt_poll_ns) {
+	if (!vcpu_valid_wakeup(vcpu))
+		 shrink_halt_poll_ns(vcpu);
+	else if (halt_poll_ns) {
 		if (block_ns <= vcpu->halt_poll_ns)
 			;
 		/* we had a long block, shrink polling */
-		else if (!vcpu_valid_wakeup(vcpu) ||
-			(vcpu->halt_poll_ns && block_ns > halt_poll_ns))
+		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
 			shrink_halt_poll_ns(vcpu);
 		/* we had a short halt and our poll time is too small */
 		else if (vcpu->halt_poll_ns < halt_poll_ns &&
-- 
2.5.5

Re: [PATCH 1/1] KVM: shrink halt polling even more for invalid wakeups

[Christian Borntraeger]

On 05/17/2016 10:49 AM, Christian Borntraeger wrote:
> commit 3491caf2755e ("KVM: halt_polling: provide a way to qualify
>  wakeups during poll") added more aggressive shrinking of the
> polling interval if the wakeup did not match some criteria. This
> still allows to keep polling enabled if the polling time was
> smaller that the current max poll time (block_ns <= vcpu->halt_poll_ns).
> Performance measurement shows that even more aggressive shrinking
> (shrink polling on any invalid wakeup) reduces absolute and relative
> (to the workload) CPU usage even further.
> 
> Cc: David Matlack <dmatlack@google.com>
> Cc: Wanpeng Li <kernellwp@gmail.com>
> Cc: Radim Krčmář <rkrcmar@redhat.com>
> CC: Paolo Bonzini <pbonzini@redhat.com>
> CC: Cornelia Huck <cornelia.huck@de.ibm.com>
> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> ---
>  virt/kvm/kvm_main.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
> 
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index fbd7698..dec1ed2 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -2055,12 +2055,13 @@ void kvm_vcpu_block(struct kvm_vcpu *vcpu)
>  out:
>  	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);
> 
> -	if (halt_poll_ns) {
> +	if (!vcpu_valid_wakeup(vcpu))
> +		 shrink_halt_poll_ns(vcpu);
		^ hmpf. Small whitespace error. I can resend if wanted.

> +	else if (halt_poll_ns) {
>  		if (block_ns <= vcpu->halt_poll_ns)
>  			;
>  		/* we had a long block, shrink polling */
> -		else if (!vcpu_valid_wakeup(vcpu) ||
> -			(vcpu->halt_poll_ns && block_ns > halt_poll_ns))
> +		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
>  			shrink_halt_poll_ns(vcpu);
>  		/* we had a short halt and our poll time is too small */
>  		else if (vcpu->halt_poll_ns < halt_poll_ns &&
> 

Re: [PATCH 1/1] KVM: shrink halt polling even more for invalid wakeups

[Paolo Bonzini]

On 17/05/2016 15:54, Christian Borntraeger wrote:
>> > -	if (halt_poll_ns) {
>> > +	if (!vcpu_valid_wakeup(vcpu))
>> > +		 shrink_halt_poll_ns(vcpu);
> 		^ hmpf. Small whitespace error. I can resend if wanted.

No need.

Paolo

>> > 

Re: [PATCH 1/1] KVM: shrink halt polling even more for invalid wakeups

[Cornelia Huck]

On Tue, 17 May 2016 10:49:22 +0200
Christian Borntraeger <borntraeger@de.ibm.com> wrote:

> commit 3491caf2755e ("KVM: halt_polling: provide a way to qualify
>  wakeups during poll") added more aggressive shrinking of the
> polling interval if the wakeup did not match some criteria. This
> still allows to keep polling enabled if the polling time was
> smaller that the current max poll time (block_ns <= vcpu->halt_poll_ns).
> Performance measurement shows that even more aggressive shrinking
> (shrink polling on any invalid wakeup) reduces absolute and relative
> (to the workload) CPU usage even further.
> 
> Cc: David Matlack <dmatlack@google.com>
> Cc: Wanpeng Li <kernellwp@gmail.com>
> Cc: Radim Krčmář <rkrcmar@redhat.com>
> CC: Paolo Bonzini <pbonzini@redhat.com>
> CC: Cornelia Huck <cornelia.huck@de.ibm.com>
> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
> ---
>  virt/kvm/kvm_main.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)

Acked-by: Cornelia Huck <cornelia.huck@de.ibm.com>