Re: [Qemu-devel] [PATCH 4/5] target-arm: Add AArch32 guest support to KVM64

[Alex Bennée]

Greg Bellows <address@hidden> writes:

> Add 32-bit to/from 64-bit register synchronization on register gets and puts.
> Set EL1_32BIT feature flag passed to KVM
>
> Signed-off-by: Greg Bellows <address@hidden>
> ---
>  target-arm/kvm64.c | 21 +++++++++++++++++----
>  1 file changed, 17 insertions(+), 4 deletions(-)
>
> diff --git a/target-arm/kvm64.c b/target-arm/kvm64.c
> index ba16821..0061099 100644
> --- a/target-arm/kvm64.c
> +++ b/target-arm/kvm64.c
> @@ -81,8 +81,7 @@ int kvm_arch_init_vcpu(CPUState *cs)
>      int ret;
>      ARMCPU *cpu = ARM_CPU(cs);
>  
> -    if (cpu->kvm_target == QEMU_KVM_ARM_TARGET_NONE ||
> -        !arm_feature(&cpu->env, ARM_FEATURE_AARCH64)) {
> +    if (cpu->kvm_target == QEMU_KVM_ARM_TARGET_NONE) {
>          fprintf(stderr, "KVM is not supported for this guest CPU type\n");
>          return -EINVAL;
>      }
> @@ -96,6 +95,9 @@ int kvm_arch_init_vcpu(CPUState *cs)
>          cpu->psci_version = 2;
>          cpu->kvm_init_features[0] |= 1 << KVM_ARM_VCPU_PSCI_0_2;
>      }
> +    if (!arm_feature(&cpu->env, ARM_FEATURE_AARCH64)) {
> +        cpu->kvm_init_features[0] |= 1 << KVM_ARM_VCPU_EL1_32BIT;
> +    }
>  
>      /* Do KVM_ARM_VCPU_INIT ioctl */
>      ret = kvm_arm_vcpu_init(cs);
> @@ -133,6 +135,7 @@ int kvm_arch_put_registers(CPUState *cs, int level)
>      ARMCPU *cpu = ARM_CPU(cs);
>      CPUARMState *env = &cpu->env;
>  
> +    aarch64_sync_32_to_64(env);
>      for (i = 0; i < 31; i++) {
>          reg.id = AARCH64_CORE_REG(regs.regs[i]);
>          reg.addr = (uintptr_t) &env->xregs[i];
> @@ -162,7 +165,11 @@ int kvm_arch_put_registers(CPUState *cs, int level)
>      }
>  
>      /* Note that KVM thinks pstate is 64 bit but we use a uint32_t */
> -    val = pstate_read(env);
> +    if (is_a64(env)) {
> +        val = pstate_read(env);
> +    } else {
> +        val = cpsr_read(env);
> +    }

I know why we do this (especially given where my attempt ended up) but
perhaps we could at list have a single state aware accessor so we don't
end up duplicating this test all over the place?

-- 
Alex Bennée