Re: [PATCH v2] target/arm: Add raw_writes ops for register whose write induce TLB maintenance

[Peter Maydell]

On Thu, 16 Feb 2023 at 17:06, Eric Auger <eric.auger@redhat.com> wrote:
>
> Hi Peter,
>
> On 2/16/23 14:51, Peter Maydell wrote:
> > On Mon, 13 Feb 2023 at 18:38, Eric Auger <eric.auger@redhat.com> wrote:
> >> Some registers whose 'cooked' writefns induce TLB maintenance do
> >> not have raw_writefn ops defined. If only the writefn ops is set
> >> (ie. no raw_writefn is provided), it is assumed the cooked also
> >> work as the raw one. For those registers it is not obvious the
> >> tlb_flush works on KVM mode so better/safer setting the raw write.
> >>
> >> Signed-off-by: Eric Auger <eric.auger@redhat.com>
> >> Suggested-by: Peter Maydell <peter.maydell@linaro.org>

> >>      { .name = "HCR", .state = ARM_CP_STATE_AA32,
> >>        .type = ARM_CP_ALIAS | ARM_CP_IO,
> >>        .cp = 15, .opc1 = 4, .crn = 1, .crm = 1, .opc2 = 0,
> >>        .access = PL2_RW, .fieldoffset = offsetof(CPUARMState, 
> >> cp15.hcr_el2),
> >> -      .writefn = hcr_writelow },
> >> +      .writefn = hcr_writelow, .raw_writefn = raw_write },
> > This is going to do the wrong thing on big-endian hosts, because
> > the data value is the low 32 bits and the raw_writefn will
> > write that to the whole 64-bit value. We could implement a
> > special purpose raw write for this register which just writes
> > the value to the low 32 bits, I guess.
> >
> > For KVM we could ignore this, though -- the register is EL2 only,
> > and also is ARM_CP_ALIAS, which means we won't try to do anything
> > with it for the KVM<->QEMU state sync or the migration codepaths.
> OK so can I simply revert that change.

Yes, that seems easiest.


> >> @@ -6139,7 +6145,7 @@ static const ARMCPRegInfo el2_v8_cp_reginfo[] = {
> >>        .cp = 15, .opc1 = 4, .crn = 1, .crm = 1, .opc2 = 4,
> >>        .access = PL2_RW,
> >>        .fieldoffset = offsetofhigh32(CPUARMState, cp15.hcr_el2),
> >> -      .writefn = hcr_writehigh },
> >> +      .writefn = hcr_writehigh, .raw_writefn = raw_write },
> > Similarly this would need a special purpose raw write function
> > since we want to only touch the high 32 bits; and again for
> > the KVM case we won't ever be doing a raw access to this register.
>
> same question: can I simply revert that change?

Ditto.

-- PMM