Re: [PATCH 19/21] KVM: s390: vsie: Add VSIE shadow stat counters

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The vsie_shadow_scb stat counter overcounts allocations during races, and vsie_shadow_scb_reuse incorrectly counts cache hits rather than resource repurposing.
--

--- Patch [19]: [PATCH 19/21] KVM: s390: vsie: Add VSIE shadow stat counters ---
commit b0e742d7460b00786ba09124150753300ba413ad
Author: Christoph Schlameuss <schlameuss@linux.ibm.com>

KVM: s390: vsie: Add VSIE shadow stat counters

This patch adds new statistics counters to VSIE shadowing to monitor
functionality, specifically tracking the allocations and reuses of SIE
control block and system control area shadows.

> diff --git a/arch/s390/kvm/vsie.c b/arch/s390/kvm/vsie.c
> index c392f7afe130c..80050376eaab1 100644
> --- a/arch/s390/kvm/vsie.c
> +++ b/arch/s390/kvm/vsie.c

[ ... ]

> @@ -2002,6 +2006,7 @@ static struct vsie_page *get_vsie_page_cpu_nr(struct kvm_vcpu *vcpu, struct vsie
>  				vsie_page = vsie_page_new;
>  			}
>  		}
> +		vcpu->kvm->stat.vsie_shadow_scb++;

[Severity: Medium]
Does this increment the vsie_shadow_scb counter even when we lose the race
and subsequently free the newly allocated page? It looks like this might
cause the counter to overcount and increment beyond the configured maximum
value when concurrent allocations occur.

>  	}
>  	if (vsie_page != vsie_page_new) {
>  		if (vsie_page_new)
> @@ -2011,6 +2016,7 @@ static struct vsie_page *get_vsie_page_cpu_nr(struct kvm_vcpu *vcpu, struct vsie
>  		if (!try_get_vsie_page(vsie_page))
>  			return ERR_PTR(-EAGAIN);
>  		vsie_page->vsie_sca = vsie_sca;
> +		vcpu->kvm->stat.vsie_shadow_scb_reuse++;

[Severity: Medium]
Does this increment vsie_shadow_scb_reuse on every cache hit rather than
just on evictions or repurposing? 

The guest-2 path appears to only count actual resource repurposing, whereas
counting every cache hit here will cause the metric to increase continuously
during normal execution.

>  	}
>  	if (vsie_page->scb_gpa != scb_gpa || vsie_page->sca_gpa != vsie_sca->sca_gpa) {
>  		scoped_guard(mutex, &vcpu->kvm->arch.vsie.mutex) {

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260709-vsie-sigpi-v1-0-ea9f12066408@linux.ibm.com?part=19