Re: [PATCH AUTOSEL 5.16 12/28] x86/kvm: Don't use pv tlb/ipi/sched_yield if on 1 vCPU

[Paolo Bonzini]

On 3/1/22 21:13, Sasha Levin wrote:
> From: Wanpeng Li <wanpengli@xxxxxxxxxxx>
>
> [ Upstream commit ec756e40e271866f951d77c5e923d8deb6002b15 ]
>
> Inspired by commit 3553ae5690a (x86/kvm: Don't use pvqspinlock code if
> only 1 vCPU), on a VM with only 1 vCPU, there is no need to enable
> pv tlb/ipi/sched_yield and we can save the memory for __pv_cpu_mask.
>
> Signed-off-by: Wanpeng Li <wanpengli@xxxxxxxxxxx>
> Message-Id: <1645171838-2855-1-git-send-email-wanpengli@xxxxxxxxxxx>
> Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
> ---
>   arch/x86/kernel/kvm.c | 9 ++++++---
>   1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
> index 59abbdad7729c..ff3db164e52cb 100644
> --- a/arch/x86/kernel/kvm.c
> +++ b/arch/x86/kernel/kvm.c
> @@ -462,19 +462,22 @@ static bool pv_tlb_flush_supported(void)
>   {
>   	return (kvm_para_has_feature(KVM_FEATURE_PV_TLB_FLUSH) &&
>   		!kvm_para_has_hint(KVM_HINTS_REALTIME) &&
> -		kvm_para_has_feature(KVM_FEATURE_STEAL_TIME));
> +		kvm_para_has_feature(KVM_FEATURE_STEAL_TIME) &&
> +		(num_possible_cpus() != 1));
>   }
>   
>   static bool pv_ipi_supported(void)
>   {
> -	return kvm_para_has_feature(KVM_FEATURE_PV_SEND_IPI);
> +	return (kvm_para_has_feature(KVM_FEATURE_PV_SEND_IPI) &&
> +	       (num_possible_cpus() != 1));
>   }
>   
>   static bool pv_sched_yield_supported(void)
>   {
>   	return (kvm_para_has_feature(KVM_FEATURE_PV_SCHED_YIELD) &&
>   		!kvm_para_has_hint(KVM_HINTS_REALTIME) &&
> -	    kvm_para_has_feature(KVM_FEATURE_STEAL_TIME));
> +	    kvm_para_has_feature(KVM_FEATURE_STEAL_TIME) &&
> +	    (num_possible_cpus() != 1));
>   }
>   
>   #define KVM_IPI_CLUSTER_SIZE	(2 * BITS_PER_LONG)

NACK

Not really necessary.

Paolo