Re: [PATCH 1/3] KVM: arm64: Configure HCRX_EL2 dynamically

From: Marc Zyngier <maz@kernel.org>
To: Kristina Martsenko <kristina.martsenko@arm.com>
Cc: kvmarm@lists.linux.dev, linux-arm-kernel@lists.infradead.org,
	Oliver Upton <oliver.upton@linux.dev>,
	James Morse <james.morse@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Zenghui Yu <yuzenghui@huawei.com>,
	Catalin Marinas <catalin.marinas@arm.com>,
	Will Deacon <will@kernel.org>,
	Vladimir Murzin <vladimir.murzin@arm.com>,
	Colton Lewis <coltonlewis@google.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/3] KVM: arm64: Configure HCRX_EL2 dynamically
Date: Mon, 18 Sep 2023 12:10:25 +0100	[thread overview]
Message-ID: <865y47enz2.wl-maz@kernel.org> (raw)
In-Reply-To: <20230915124840.474888-2-kristina.martsenko@arm.com>

Hi Kristina,

On Fri, 15 Sep 2023 13:48:38 +0100,
Kristina Martsenko <kristina.martsenko@arm.com> wrote:
> 
> At the moment the HCRX_EL2 system register is always initialized to
> HCRX_GUEST_FLAGS when running a guest. Instead, choose the configuration
> at vcpu reset time and save it in the vcpu struct, similarly to how
> HCR_EL2 is set up. This will be needed in a subsequent change to
> configure the register based on CPU features detected at runtime.
> 
> Signed-off-by: Kristina Martsenko <kristina.martsenko@arm.com>
> ---
>  arch/arm64/include/asm/kvm_emulate.h    | 5 +++++
>  arch/arm64/include/asm/kvm_host.h       | 1 +
>  arch/arm64/kvm/arm.c                    | 1 +
>  arch/arm64/kvm/hyp/include/hyp/switch.h | 2 +-
>  arch/arm64/kvm/hyp/nvhe/hyp-main.c      | 1 +
>  arch/arm64/kvm/hyp/nvhe/pkvm.c          | 1 +
>  6 files changed, 10 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/include/asm/kvm_emulate.h b/arch/arm64/include/asm/kvm_emulate.h
> index 3d6725ff0bf6..64ea27e6deb1 100644
> --- a/arch/arm64/include/asm/kvm_emulate.h
> +++ b/arch/arm64/include/asm/kvm_emulate.h
> @@ -134,6 +134,11 @@ static inline void vcpu_ptrauth_disable(struct kvm_vcpu *vcpu)
>  	vcpu->arch.hcr_el2 &= ~(HCR_API | HCR_APK);
>  }
>  
> +static inline void vcpu_reset_hcrx(struct kvm_vcpu *vcpu)
> +{
> +	vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS;
> +}
> +
>  static inline unsigned long vcpu_get_vsesr(struct kvm_vcpu *vcpu)
>  {
>  	return vcpu->arch.vsesr_el2;
> diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
> index af06ccb7ee34..2764748756a7 100644
> --- a/arch/arm64/include/asm/kvm_host.h
> +++ b/arch/arm64/include/asm/kvm_host.h
> @@ -487,6 +487,7 @@ struct kvm_vcpu_arch {
>  
>  	/* Values of trap registers for the guest. */
>  	u64 hcr_el2;
> +	u64 hcrx_el2;

Do we really need this extra field? Yes, this is only an extra 64bit,
but they tend to accumulate...

Looking at patch #3, the change is related to this:

 	vcpu->arch.hcrx_el2 = HCRX_GUEST_FLAGS;
+
+	if (cpus_have_final_cap(ARM64_HAS_MOPS)) {
+		vcpu->arch.hcrx_el2 |= HCRX_EL2_MSCEn;
+		vcpu->arch.hcrx_el2 |= HCRX_EL2_MCE2;
+	}

meaning that this is a constant value for a given boot of the host.

At this stage, I'd rather you define HCRX_GUEST_FLAGS as:

#define HCRX_GUEST_FLAGS \
	(HCRX_EL2_SMPME | HCRX_EL2_TCR2En | \
	 cpus_have_final_cap(ARM64_HAS_MOPS) ? \
	 (HCRX_EL2_MSCEn | HCRX_EL2_MCE2) : 0)

and drop the new field altogether, until we have something that
requires dynamic flipping of an HCRX_EL2 field.

Thanks,

	M.

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