Re: [PATCH 3/5] KVM: arm64: vgic-v3: Erase LPIs from xarray outside of raw spinlocks

[Marc Zyngier <maz@kernel.org>]

On Thu, 04 Sep 2025 07:23:46 +0100,
Oliver Upton <oliver.upton@linux.dev> wrote:
> 
> diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h
> index ed0e96031a65..af224db3cf72 100644
> --- a/include/kvm/arm_vgic.h
> +++ b/include/kvm/arm_vgic.h
> @@ -139,7 +139,11 @@ struct vgic_irq {
>  	bool pending_latch;		/* The pending latch state used to calculate
>  					 * the pending state for both level
>  					 * and edge triggered IRQs. */
> -	bool active;			/* not used for LPIs */
> +	union {
> +		bool active;		/* not used for LPIs */
> +		bool pending_release;	/* LPI pending a release */
> +	};
> +

Err... no. Please don't do that. An activated LPI that hasn't been
EOI'd yet does have the active state in the LR (yes, the original
comment is totally broken, please remove it).

Imagine a case where you'd preempt the guest between the read of IAR
and the write to EOI: pending_release is now set, and that could
result in some funky stuff...

Just add it as a separate field, I don't think anyone will cry over
that.

	M.

-- 
Jazz isn't dead. It just smells funny.