[PATCH v6 11/12] ARM/ARM64: KVM: Emulate PSCI v0.2 CPU_SUSPEND

[christoffer.dall@linaro.org (Christoffer Dall)]

On Mon, Mar 31, 2014 at 04:16:12PM +0530, Anup Patel wrote:
> This patch adds emulation of PSCI v0.2 CPU_SUSPEND function call for
> KVM ARM/ARM64. This is a CPU-level function call which can suspend
> current CPU or current CPU cluster. We don't have VCPU clusters in
> KVM so for KVM we simply suspend the current VCPU.
> 
> The CPU_SUSPEND emulation is not tested much because currently there
> is no CPUIDLE driver in Linux kernel that uses PSCI CPU_SUSPEND. The
> PSCI CPU_SUSPEND implementation in ARM64 kernel was tested using a
> Simple CPUIDLE driver which is not published due to unstable DT-bindings
> for PSCI.
> (For more info, http://lwn.net/Articles/574950/)
> 
> Even if we had stable DT-bindings for PSCI and CPUIDLE driver that
> uses PSCI CPU_SUSPEND then still we need to define SUSPEND states
> and WAKEUP events for KVM ARM/ARM64.
> 
> Due to this, we implement CPU_SUSPEND emulation similar to WFI
> (Wait-for-interrupt) emulation.
> 
> Signed-off-by: Anup Patel <anup.patel@linaro.org>
> Signed-off-by: Pranavkumar Sawargaonkar <pranavkumar@linaro.org>
> ---
>  arch/arm/kvm/psci.c |   29 +++++++++++++++++++++++++----
>  1 file changed, 25 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/arm/kvm/psci.c b/arch/arm/kvm/psci.c
> index 85bf896..f414fd3 100644
> --- a/arch/arm/kvm/psci.c
> +++ b/arch/arm/kvm/psci.c
> @@ -52,6 +52,27 @@ static unsigned long psci_affinity_mask(unsigned long affinity_level)
>  	return affinity_mask;
>  }
>  
> +static unsigned long kvm_psci_vcpu_suspend(struct kvm_vcpu *vcpu)
> +{
> +	/*
> +	 * NOTE: Currently, we don't have any wakeup events for KVM
> +	 * so for simplicity we make VCPU suspend emulation same-as
> +	 * WFI (Wait-for-interrupt) emulation.

If you implement it like WFI, we do have wake-up events: Namely
interrupts.

> +	 *
> +	 * To do this we simply update VCPU registers as-per state
> +	 * info provided via r1 - r3 (or x1 - x3) and block the
> +	 * VCPU for irqs.
> +	 */
> +	if (*vcpu_reg(vcpu, 1) & (0x1UL << 16)) {
> +		/* Update return pc and r0 for power-down state. */
> +		*vcpu_pc(vcpu) = *vcpu_reg(vcpu, 2);
> +		*vcpu_reg(vcpu, 0) = *vcpu_reg(vcpu, 3);
> +	}


Hmm, this looks wrong.  This looks like you're respecting the power-down
state request but not resetting the CPU.  What I was saying before was
that if you implement this as kvm_vcpu_block(), just like WFI, then you
need to preserve all state, ignore power-down state requests and treat
them as suspend states, implement them as WFI, and put a big fat comment
here explaining why this is architecturally valid (by referring to the
PSCI 0.2 spec) and what the semantics of doing that is.

-Christoffer


> +	kvm_vcpu_block(vcpu);
> +
> +	return PSCI_RET_SUCCESS;
> +}
> +
>  static void kvm_psci_vcpu_off(struct kvm_vcpu *vcpu)
>  {
>  	vcpu->arch.pause = true;
> @@ -195,6 +216,10 @@ static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
>  		 */
>  		val = 2;
>  		break;
> +	case PSCI_0_2_FN_CPU_SUSPEND:
> +	case PSCI_0_2_FN64_CPU_SUSPEND:
> +		val = kvm_psci_vcpu_suspend(vcpu);
> +		break;
>  	case PSCI_0_2_FN_CPU_OFF:
>  		kvm_psci_vcpu_off(vcpu);
>  		val = PSCI_RET_SUCCESS;
> @@ -232,10 +257,6 @@ static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
>  		val = PSCI_RET_SUCCESS;
>  		ret = 0;
>  		break;
> -	case PSCI_0_2_FN_CPU_SUSPEND:
> -	case PSCI_0_2_FN64_CPU_SUSPEND:
> -		val = PSCI_RET_NOT_SUPPORTED;
> -		break;
>  	default:
>  		return -EINVAL;
>  	}
> -- 
> 1.7.9.5
>