[LinuxBIOS] [PATCH] v3: cpu.h improvements

[Uwe Hermann]

See patch.

Uwe.

[Corey Osgood]

Uwe Hermann wrote:
> -static inline struct cpuid_result cpuid(int op)
> +static inline struct cpuid_result cpuid(u32 op)
>  {
> struct cpuid_result result;
> asm volatile(
> "cpuid"
> : "=a" (result.eax),
>   "=b" (result.ebx),
>   "=c" (result.ecx),
>   "=d" (result.edx)
> : "0" (op));
> return result;
> struct cpuid_result r;
> _cpuid(op, &r.eax, &r.ebx, %r.ecx, &r.edx);
> return r;
>
> }
>
> +/**
> Generic CPUID function.
> Clear %ecx since some CPUs (Cyrix MII) do not set or clear %ecx,
> resulting in stale register contents being returned.
> */
>
> +static inline void _cpuid(u32 op, u32 *eax, u32 *ebx, u32 *ecx, u32 *edx)
> +{
> *eax = op;
> *ecx = 0;
> native_cpuid(eax, ebx, ecx, edx);
>
> +}

Is there any reason not to combine these two functions? Maybe it's just
me, but I don't like seeing one function call another, make some menial
changes then call a third, all with the exact same variables, and to
complete one task. And, uh, does this build okay? I don't see a forward
declaration of _cpuid, but it might be in the rest of the file or
somewhere else.

-Corey

[LinuxBIOS] [PATCH] v3: cpu.h improvements (take 2)

[Uwe Hermann]

On Wed, Jun 27, 2007 at 07:06:12AM -0400, Corey Osgood wrote:
> Is there any reason not to combine these two functions? Maybe it's just
> me, but I don't like seeing one function call another, make some menial
> changes then call a third, all with the exact same variables, and to
> complete one task.

Good point. The Linux file is a lot more generic (there's also a
cpuid_count() function which is a bit different and uses the generic
function), but all of this is not needed in LinuxBIOS, I guess.

New patch attached. The code is now even shorter and simpler.

Uwe.

[LinuxBIOS] [PATCH] v3: cpu.h improvements (take 2)

[Stefan Reinauer]

* Uwe Hermann  [070630 17:17]:
> Bring the file cpu.h in sync with the current version of the code in
> the Linux kernel (as far as possible). The code is a lot simpler
> and shorter now.
>
> Also, add cpu_relax() (which is also part of the Linux file) which
> could be useful for busy-loops (but is currently not used).
> Drop the cpu_relax() from spinlock.h as it's not spinlock related in any way.
>
> The code builds, but is otherwise untested.
>
> Signed-off-by: Uwe Hermann

Can you test with qemu?

Otherwise
Acked-by: Stefan Reinauer

[LinuxBIOS] [PATCH] v3: cpu.h improvements (take 2)

[Uwe Hermann]

On Sat, Jun 30, 2007 at 05:30:53PM +0200, Stefan Reinauer wrote:
> Uwe Hermann  [070630 17:17]:
> > Bring the file cpu.h in sync with the current version of the code in
> > the Linux kernel (as far as possible). The code is a lot simpler
> > and shorter now.
> >
> > Also, add cpu_relax() (which is also part of the Linux file) which
> > could be useful for busy-loops (but is currently not used).
> > Drop the cpu_relax() from spinlock.h as it's not spinlock related in any way.
> >
> > The code builds, but is otherwise untested.
> >
> > Signed-off-by: Uwe Hermann
>
> Can you test with qemu?
>
> Otherwise
> Acked-by: Stefan Reinauer

r423.

Tested in QEMU, but that's probably not enough, as we don't use the
cpuid() function at the moment, I think.

Uwe.