Re: [PATCH 07/11] x86-64: Simplify constraints for fxsave/fxtstor

[Pekka Enberg <penberg@kernel.org>]

  On 8/30/10 2:44 AM, Brian Gerst wrote:
> On Sun, Aug 29, 2010 at 2:45 PM, Pekka Enberg<penberg@kernel.org>  wrote:
>> On Sat, Aug 28, 2010 at 7:04 PM, Brian Gerst<brgerst@gmail.com>  wrote:
>>> Use the "R" constraint (legacy register) instead of listing all the
>>> possible registers.  Clean up the comments as well.
>>>
>>> Signed-off-by: Brian Gerst<brgerst@gmail.com>
>>> ---
>>>   arch/x86/include/asm/i387.h |   44 ++++++++++++++++--------------------------
>>>   1 files changed, 17 insertions(+), 27 deletions(-)
>>>
>>> diff --git a/arch/x86/include/asm/i387.h b/arch/x86/include/asm/i387.h
>>> index 8b40a83..768fcb2 100644
>>> --- a/arch/x86/include/asm/i387.h
>>> +++ b/arch/x86/include/asm/i387.h
>>> @@ -81,6 +81,7 @@ static inline int fxrstor_checking(struct i387_fxsave_struct *fx)
>>>   {
>>>         int err;
>>>
>>> +       /* See comment in fxsave() below. */
>>>         asm volatile("1:  rex64/fxrstor (%[fx])\n\t"
>>>                      "2:\n"
>>>                      ".section .fixup,\"ax\"\n"
>>> @@ -89,11 +90,7 @@ static inline int fxrstor_checking(struct i387_fxsave_struct *fx)
>>>                      ".previous\n"
>>>                      _ASM_EXTABLE(1b, 3b)
>>>                      : [err] "=r" (err)
>>> -#if 0 /* See comment in fxsave() below. */
>>> -                    : [fx] "r" (fx), "m" (*fx), "0" (0));
>>> -#else
>>> -                    : [fx] "cdaSDb" (fx), "m" (*fx), "0" (0));
>>> -#endif
>>> +                    : [fx] "R" (fx), "m" (*fx), "0" (0));
>> Please correct me if I'm wrong but "legacy registers" also include bp
>> and sp that are not part of the original constraints:
>>
>> http://gcc.gnu.org/onlinedocs/gcc/Machine-Constraints.html
>>
>> So why is it OK to use "R" here?
> There is no constraint to explicitly request %bp (or %sp), but there
> is no reason it could not be used.  The compiler would never choose
> %sp for "R" for the same reason it wouldn't for "r": it's not
> available as a general purpose register.
Yeah, if I try to force GCC to use bp or sp, it will complain that it 
runs out of legacy registers:

error: can't find a register in class `LEGACY_REGS' while reloading `asm'

which makes sense as it knows that sp and bp are used by the function 
that contains the inline asm.

Acked-by: Pekka Enberg <penberg@kernel.org>

             Pekka