Re: [PATCH -tip] x86: cpu_debug.c prepare report if files are inappropriate or CPU is not supported

[Jaswinder Singh Rajput <jaswinder@kernel.org>]

On Tue, 2009-04-14 at 18:50 +0200, Ingo Molnar wrote:
> * Jaswinder Singh Rajput <jaswinder@kernel.org> wrote:
> 
> > +		if (!per_cpu(cpu_modelflag, cpu))
> 
> hm, on a second look - the whole cpu_model / cpu_modelflag 
> business in arch/x86/kernel/cpu/cpu_debug.c looks 
> over-complicated and broken. You encode it into a 
> 'modelflag':
> 
>                 per_cpu(cpu_model, cpu) = ((cpui->x86_vendor << 16) |
>                                            (cpui->x86 << 8) |
>                                            (cpui->x86_model));
> 
> just to decode it later on:
> 
>         flag = per_cpu(cpu_model, cpu);
> 
>         switch (flag >> 16) {
> 
> That does not make much sense. Please use a proper 
> boot_cpu_data.x86_vendor switch() statement, ok?
> 

I am using flags for each cpu, in case there are different CPU in the
sockets:

struct cpuinfo_x86 *cpui;
cpui = &cpu_data(cpu);

Do you still think that boot_cpu_data.x86_vendor is better option in
case for multiple CPUs.

--
JSR