Re: [PATCH 2/2] selftests: KVM: Fix 'asm-operand-width' warnings in steal_time.c

[Paolo Bonzini <pbonzini@redhat.com>]

On 21/09/21 09:19, Andrew Jones wrote:
> On Tue, Sep 21, 2021 at 01:01:20AM +0000, Oliver Upton wrote:
>> Building steal_time.c for arm64 with clang throws the following:
>>
>>>> steal_time.c:130:22: error: value size does not match register size specified by the constraint and modifier [-Werror,-Wasm-operand-widths]
>>            : "=r" (ret) : "r" (func), "r" (arg) :
>>                                ^
>>>> steal_time.c:130:34: error: value size does not match register size specified by the constraint and modifier [-Werror,-Wasm-operand-widths]
>>            : "=r" (ret) : "r" (func), "r" (arg) :
>>                                            ^
>>
>> Silence by casting operands to 64 bits.
>>
>> Signed-off-by: Oliver Upton <oupton@google.com>
>> ---
>>   tools/testing/selftests/kvm/steal_time.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/tools/testing/selftests/kvm/steal_time.c b/tools/testing/selftests/kvm/steal_time.c
>> index ecec30865a74..eb75b31122c5 100644
>> --- a/tools/testing/selftests/kvm/steal_time.c
>> +++ b/tools/testing/selftests/kvm/steal_time.c
>> @@ -127,7 +127,7 @@ static int64_t smccc(uint32_t func, uint32_t arg)
>>   		"mov	x1, %2\n"
>>   		"hvc	#0\n"
>>   		"mov	%0, x0\n"
>> -	: "=r" (ret) : "r" (func), "r" (arg) :
>> +	: "=r" (ret) : "r" ((uint64_t)func), "r" ((uint64_t)arg) :
> 
> Actually, I think I'd rather fix this smccc implementation to match the
> spec, which I think should be done like this
> 
> diff --git a/tools/testing/selftests/kvm/steal_time.c b/tools/testing/selftests/kvm/steal_time.c
> index ecec30865a74..7da957259ce4 100644
> --- a/tools/testing/selftests/kvm/steal_time.c
> +++ b/tools/testing/selftests/kvm/steal_time.c
> @@ -118,12 +118,12 @@ struct st_time {
>          uint64_t st_time;
>   };
>   
> -static int64_t smccc(uint32_t func, uint32_t arg)
> +static int64_t smccc(uint32_t func, uint64_t arg)
>   {
>          unsigned long ret;
>   
>          asm volatile(
> -               "mov    x0, %1\n"
> +               "mov    w0, %w1\n"
>                  "mov    x1, %2\n"
>                  "hvc    #0\n"
>                  "mov    %0, x0\n"
> 

Agreed, can you send out a patch?  Thanks,

Paolo