Re: [PATCH v3 05/11] KVM: x86/xen: Retry event channel delivery which races a memslot update

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] kvm/xen: guest hang due to lost/delayed timer interrupt on memslot update
--

commit 8954ef8936e60acd6079cac417ec4cbf08881487
Author: David Woodhouse <dwmw@amazon.co.uk>

KVM: x86/xen: Retry event channel delivery which races a memslot update

This commit modifies event channel delivery to return -EAGAIN when racing
with a memslot update, allowing callers to handle the retry according to
their context.

Closes: https://lore.kernel.org/all/20260920215341.5ECE91F000FF@smtp.kernel.org

> diff --git a/arch/x86/kvm/xen.c b/arch/x86/kvm/xen.c
> index 0ef646cf3990b..6b7532b0e26ab 100644
> --- a/arch/x86/kvm/xen.c
> +++ b/arch/x86/kvm/xen.c
[ ... ]
> @@ -135,7 +136,7 @@ void kvm_xen_inject_timer_irqs(struct kvm_vcpu *vcpu)
>  		e.port = vcpu->arch.xen.timer_virq;
>  		e.priority = KVM_IRQ_ROUTING_XEN_EVTCHN_PRIO_2LEVEL;
>  
> -		rc = kvm_xen_set_evtchn(&e, vcpu->kvm);
> +		rc = kvm_xen_set_evtchn(&e, vcpu->kvm, false);
>  
>  		/*
>  		 * Leave timer_pending set if the memslot is being updated
>  		 * (-EAGAIN); delivery is retried on the next entry pass.
>  		 */
>  		if (rc == -EAGAIN)
>  			return;

[Severity: High]
Does this early return fail to set a KVM request to abort guest entry?

When kvm_xen_set_evtchn() returns -EAGAIN due to a memslot update race,
returning here leaves timer_pending > 0 but KVM's entry path
(vcpu_enter_guest) will see no pending requests or injectable interrupts.

Because the host hrtimer has already expired and will not restart, could this
cause the vCPU to enter the guest without the timer interrupt and run
indefinitely until an unrelated VM-exit occurs?

For context, this caller is at the end of the vcpu_run() loop:

arch/x86/kvm/x86.c:vcpu_run() {
    ...
    if (kvm_cpu_has_pending_timer(vcpu))
        kvm_inject_pending_timer_irqs(vcpu);
    ...
}

If injection fails here without setting a request like
kvm_make_request(KVM_REQ_EVENT, vcpu), won't the next loop iteration call
vcpu_enter_guest() and enter the guest without scheduling an immediate retry?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261016145646.579369-1-dwmw2@infradead.org?part=5