Re: [PATCH v2 13/17] KVM: arm64: Pend a protected guest's SError with HCR_EL2.VSE only

[Marc Zyngier <maz@kernel.org>]

On Mon, 07 Sep 2026 07:59:58 +0100,
Fuad Tabba <fuad.tabba@linux.dev> wrote:
> 
> kvm_inject_serror_esr() emulates the exception entry when
> serror_is_masked() reads PSTATE.A clear, but the host has no view of a
> protected guest's PSTATE.A, and an emulated entry never reaches the
> guest: the SError is lost. Set VSE instead and let the guest's own
> PSTATE.A mask it.
> 
> Signed-off-by: Fuad Tabba <fuad.tabba@linux.dev>
> ---
>  arch/arm64/kvm/inject_fault.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/kvm/inject_fault.c b/arch/arm64/kvm/inject_fault.c
> index d6c4fc16f8795..b88098a7e756c 100644
> --- a/arch/arm64/kvm/inject_fault.c
> +++ b/arch/arm64/kvm/inject_fault.c
> @@ -378,8 +378,11 @@ int kvm_inject_serror_esr(struct kvm_vcpu *vcpu, u64 esr)
>  	 *
>  	 * As we're emulating the SError injection we need to explicitly populate
>  	 * ESR_ELx.EC because hardware will not do it on our behalf.
> +	 *
> +	 * The host does not see a protected guest's PSTATE.A: leave the
> +	 * vSError to HCR_EL2.VSE below, which the guest masks itself.
>  	 */
> -	if (!serror_is_masked(vcpu)) {
> +	if (!vcpu_is_protected(vcpu) && !serror_is_masked(vcpu)) {
>  		pend_serror_exception(vcpu);
>  		esr |= FIELD_PREP(ESR_ELx_EC_MASK, ESR_ELx_EC_SERROR) | ESR_ELx_IL;
>  		vcpu_write_sys_reg(vcpu, esr, exception_esr_elx(vcpu));

I can't say I like this very much.

The fundamental premise is that as far as the host is concerned,
PSTATE.A is always set. So why can't we give the illusion of this at
all times?

That'd be a much cleaner approach.

Thanks,

	M.

-- 
Without deviation from the norm, progress is not possible.