Hello!
On Tue, Aug 13, 2024 at 11:44:00AM +0100, Marc Zyngier wrote:
> Everything is now in place for a guest to "enjoy" FP8 support.
> Expose ID_AA64PFR2_EL1 to both userspace and guests, with the
> explicit restriction of only being able to clear FPMR.
>
> All other features (MTE* at the time of writing) are hidden
> and not writable.
>
> Signed-off-by: Marc Zyngier <maz@xxxxxxxxxx>
> ---
> arch/arm64/kvm/sys_regs.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
> index 51627add0a72..da6d017f24a1 100644
> --- a/arch/arm64/kvm/sys_regs.c
> +++ b/arch/arm64/kvm/sys_regs.c
> @@ -1722,6 +1722,15 @@ static u64 read_sanitised_id_aa64pfr0_el1(struct kvm_vcpu *vcpu,
> return val;
> }
>
> +static u64 read_sanitised_id_aa64pfr2_el1(struct kvm_vcpu *vcpu,
> + const struct sys_reg_desc *rd)
> +{
> + u64 val = read_sanitised_ftr_reg(SYS_ID_AA64PFR2_EL1);
> +
> + /* We only expose FPMR */
> + return val & ID_AA64PFR2_EL1_FPMR;
> +}
Wondering why you're adding this function instead of extending __kvm_read_sanitised_id_reg()?
> +
> #define ID_REG_LIMIT_FIELD_ENUM(val, reg, field, limit) \
> ({ \
> u64 __f_val = FIELD_GET(reg##_##field##_MASK, val); \
> @@ -2381,7 +2390,12 @@ static const struct sys_reg_desc sys_reg_descs[] = {
> ID_AA64PFR0_EL1_AdvSIMD |
> ID_AA64PFR0_EL1_FP), },
> ID_SANITISED(ID_AA64PFR1_EL1),
> - ID_UNALLOCATED(4,2),
> + { SYS_DESC(SYS_ID_AA64PFR2_EL1),
> + .access = access_id_reg,
> + .get_user = get_id_reg,
> + .set_user = set_id_reg,
> + .reset = read_sanitised_id_aa64pfr2_el1,
> + .val = ID_AA64PFR2_EL1_FPMR, },
Then I think this would just be ID_WRITABLE(ID_AA64PFR2_EL1, ID_AA64PFR2_EL1_FPMR).
> ID_UNALLOCATED(4,3),
> ID_WRITABLE(ID_AA64ZFR0_EL1, ~ID_AA64ZFR0_EL1_RES0),
> ID_HIDDEN(ID_AA64SMFR0_EL1),
Thanks,
Joey
