Re: [PATCH v1 13/16] kvm: arm64: Configure VTCR per VM

[Suzuki K Poulose <Suzuki.Poulose@arm.com>]

Hi Christoffer,

On 08/02/18 18:04, Christoffer Dall wrote:
> On Tue, Jan 09, 2018 at 07:04:08PM +0000, Suzuki K Poulose wrote:
>> We set VTCR_EL2 very early during the stage2 init and don't
>> touch it ever. This is fine as we had a fixed IPA size. This
>> patch changes the behavior to set the VTCR for a given VM,
>> depending on its stage2 table. The common configuration for
>> VTCR is still performed during the early init. But the SL0
>> and T0SZ are programmed for each VM and is cleared once we
>> exit the VM.
>>
>> Cc: Marc Zyngier <marc.zyngier@arm.com>
>> Cc: Christoffer Dall <cdall@linaro.org>
>> Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>

>> diff --git a/arch/arm64/kvm/hyp/switch.c b/arch/arm64/kvm/hyp/switch.c
>> index f7c651f3a8c0..523471f0af7b 100644
>> --- a/arch/arm64/kvm/hyp/switch.c
>> +++ b/arch/arm64/kvm/hyp/switch.c
>> @@ -157,11 +157,20 @@ static void __hyp_text __deactivate_traps(struct kvm_vcpu *vcpu)
>>   static void __hyp_text __activate_vm(struct kvm_vcpu *vcpu)
>>   {
>>    struct kvm *kvm = kern_hyp_va(vcpu->kvm);
>> + u64 vtcr = read_sysreg(vtcr_el2);
>> +
>> + vtcr &= ~VTCR_EL2_PRIVATE_MASK;
>> + vtcr |= VTCR_EL2_SL0(stage2_pt_levels(kvm)) |
>> + VTCR_EL2_T0SZ(kvm_phys_shift(kvm));
>> + write_sysreg(vtcr, vtcr_el2);
> 
> If we're writing VTCR_EL2 on each entry, do we really need to read the
> value back first and calculate things on every entry to the VM?  It
> seems to me we should be able to compute the vtcr_el2 and store it on
> struct kvm, and simply restore that per-VM value upon entering the VM?

I took a look at this and we need to do this to make sure we retain the
Hardware update of Access flags for stage2 (VTCR_EL2_HA) bits on the CPUs
that has it and it is safe to run a mix of CPUs with and without the feature.

>>    write_sysreg(kvm->arch.vttbr, vttbr_el2);
>>   }
>>
>>   static void __hyp_text __deactivate_vm(struct kvm_vcpu *vcpu)
>>   {
>> + u64 vtcr = read_sysreg(vtcr_el2) & ~VTCR_EL2_PRIVATE_MASK;
>> +
>> + write_sysreg(vtcr, vtcr_el2);
> 
> Why do we need to care about restoring VTCR when returning to the host?

Yes, this can be skipped.

Cheers
Suzuki