Hi Marc,
On Thu, Aug 17, 2023 at 8:43 AM Marc Zyngier <maz@xxxxxxxxxx> wrote:
>
> On Mon, 07 Aug 2023 17:22:03 +0100,
> Jing Zhang <jingzhangos@xxxxxxxxxx> wrote:
> >
> > All valid fields in ID_AA64DFR0_EL1 and ID_DFR0_EL1 are writable
> > from usrespace with this change.
>
> nit: userspace
>
Fixed.
> >
> > Signed-off-by: Jing Zhang <jingzhangos@xxxxxxxxxx>
> > ---
> > arch/arm64/kvm/sys_regs.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/kvm/sys_regs.c b/arch/arm64/kvm/sys_regs.c
> > index afade7186675..5f6c2be12e44 100644
> > --- a/arch/arm64/kvm/sys_regs.c
> > +++ b/arch/arm64/kvm/sys_regs.c
> > @@ -2006,7 +2006,7 @@ static const struct sys_reg_desc sys_reg_descs[] = {
> > .set_user = set_id_dfr0_el1,
> > .visibility = aa32_id_visibility,
> > .reset = read_sanitised_id_dfr0_el1,
> > - .val = ID_DFR0_EL1_PerfMon_MASK, },
> > + .val = GENMASK(63, 0), },
>
> For obvious reasons, this cannot be a 64 bit mask...
>
> > ID_HIDDEN(ID_AFR0_EL1),
> > AA32_ID_SANITISED(ID_MMFR0_EL1),
> > AA32_ID_SANITISED(ID_MMFR1_EL1),
> > @@ -2055,7 +2055,7 @@ static const struct sys_reg_desc sys_reg_descs[] = {
> > .get_user = get_id_reg,
> > .set_user = set_id_aa64dfr0_el1,
> > .reset = read_sanitised_id_aa64dfr0_el1,
> > - .val = ID_AA64DFR0_EL1_PMUVer_MASK, },
> > + .val = GENMASK(63, 0), },
>
> What is the actual justification to go from "only the PMU version is
> writable" to "everything is writable"?
>
> Also, what about the RES0 fields?
You're right. We should not mark RES0 fields and those fields KVM hide
from userspace as writable.
So, I think all the other fields can be writable?
>
> Thanks,
>
> M.
>
> --
> Without deviation from the norm, progress is not possible.
>
Thanks,
Jing
