Re: [PATCH v3 03/22] KVM: arm64: Define PMI{CNTR,FILTR}_EL0 as undef_access

[Marc Zyngier]

On Thu, 26 Jun 2025 21:04:39 +0100,
Colton Lewis <coltonlewis@xxxxxxxxxx> wrote:
> 
> Because KVM isn't fully prepared to support these yet even though the
> host PMUv3 driver does, define them as undef_access for now.
> 
> Signed-off-by: Colton Lewis <coltonlewis@xxxxxxxxxx>
> ---
>  arch/arm64/kvm/sys_regs.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
> index 76c2f0da821f..99fdbe174202 100644
> --- a/arch/arm64/kvm/sys_regs.c
> +++ b/arch/arm64/kvm/sys_regs.c
> @@ -3092,6 +3092,9 @@ static const struct sys_reg_desc sys_reg_descs[] = {
>  	{ SYS_DESC(SYS_SVCR), undef_access, reset_val, SVCR, 0, .visibility = sme_visibility  },
>  	{ SYS_DESC(SYS_FPMR), undef_access, reset_val, FPMR, 0, .visibility = fp8_visibility },
>  
> +	{ SYS_DESC(SYS_PMICNTR_EL0), undef_access },

$ jq -r --arg FEAT "FEAT_PMUv3_ICNTR" -f ./dumpfeat.jq Features.json
(FEAT_PMUv3_ICNTR --> v8Ap8)
(FEAT_PMUv3_ICNTR --> FEAT_PMUv3p9)
((FEAT_PMUv3_ICNTR && FEAT_AA64EL2) --> FEAT_FGT2)

If you have FEAT_PMUv3_ICNTR, then you have FEAT_FGT2. If you have
FEAT_FGT2, then we already trap and UNDEF PMICNTR_EL0 without any
further handling since 4bc0fe0898406 ("KVM: arm64: Add sanitisation
for FEAT_FGT2 registers").

> +	{ SYS_DESC(SYS_PMICFILTR_EL0), undef_access },

Same thing.
> +
>  	{ PMU_SYS_REG(PMCR_EL0), .access = access_pmcr, .reset = reset_pmcr,
>  	  .reg = PMCR_EL0, .get_user = get_pmcr, .set_user = set_pmcr },
>  	{ PMU_SYS_REG(PMCNTENSET_EL0),

So none of this is actually required.

	M.

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