Re: [PATCHv4 1/5] x86: provide add_sized()

[David Vrabel <david.vrabel@citrix.com>]

On 13/05/15 10:23, Jan Beulich wrote:
>>>> On 11.05.15 at 16:37, <david.vrabel@citrix.com> wrote:
>> @@ -53,6 +67,19 @@ void __bad_atomic_size(void);
>>      }                                                     \
>>  })
>>  
>> +#define add_sized(p, x) ({                                \
>> +    typeof(*(p)) __x = (x);                               \
>> +    unsigned long x_ = (unsigned long)__x;                \
> 
> So is there a particular reason you kept this double type conversion?
> As said earlier, I can see why write_atomic() wants it, but I don't see
> the need here.

I kept it in case typeof (*p) wasn't a numeric type (like a struct { u32
a; }) and needed the cast to unsigned long, although in hindsight I'm
not sure this really makes any sense.

David

>> +    switch ( sizeof(*(p)) )                               \
>> +    {                                                     \
>> +    case 1: add_u8_sized((uint8_t *)(p), x_); break;      \
>> +    case 2: add_u16_sized((uint16_t *)(p), x_); break;    \
>> +    case 4: add_u32_sized((uint32_t *)(p), x_); break;    \
>> +    case 8: add_u64_sized((uint64_t *)(p), x_); break;    \
>> +    default: __bad_atomic_size(); break;                  \
>> +    }                                                     \
>> +})
> 
>