Re: [RFC PATCH v4 01/15] target/arm: Move 64-bit TCG CPUs into tcg/

[Richard Henderson <richard.henderson@linaro.org>]

On 1/19/23 09:07, Fabiano Rosas wrote:
> Richard Henderson <richard.henderson@linaro.org> writes:
> 
>> On 1/19/23 03:54, Fabiano Rosas wrote:
>>> Move the 64-bit CPUs that are TCG-only:
>>> - cortex-a35
>>> - cortex-a55
>>> - cortex-a72
>>> - cortex-a76
>>> - a64fx
>>> - neoverse-n1
>>>
>>> Keep the CPUs that can be used with KVM:
>>> - cortex-a57
>>> - cortex-a53
>>> - max
>>> - host
>>
>> All of those cpus can be used with kvm, if and only if you have matching hardware.  There
>> is no rationale for considering any of them differently.
> 
> But is that allowed by QEMU today? If so I must be missing something. I
> see that kvm_arch_init_vcpu looks at cpu->kvm_target, which is only set
> at kvm_arm_set_cpu_features_from_host, called from aarch64_host_initfn.
> 
>      if (cpu->kvm_target == QEMU_KVM_ARM_TARGET_NONE ||
>          !object_dynamic_cast(OBJECT(cpu), TYPE_AARCH64_CPU)) {
>          error_report("KVM is not supported for this guest CPU type");
>          return -EINVAL;
>      }

Hmm, no, you're right.  It could guess that some of those could be marked 
KVM_ARM_TARGET_GENERIC_V8, but there's probably no point, since -cpu {host,max} are the 
only reasonable options to use with kvm, practically.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~