Re: [RFC PATCH 1/4] x86, cpu: Expand cpufeature facility to includecpu bugs

From: H. Peter Anvin
Date: Wed Feb 13 2013 - 18:24:47 EST

On 02/11/2013 03:18 AM, Borislav Petkov wrote:
> +#define __BUG_CHECK_BIT(bit) \
> +({ \
> + WARN_ON(bit >> 5 < NCAPINTS); \
> + bit; \
> +})
> +

Do we need this? Either way, if we do, I would suggest doing something

if (__builtin_constant_p(bit))

... and flag bad_bug_number as a compile time error, since the vast
majority of the time the bit number will be constant.

However, I don't think it is necessary. In order for this to ever
trigger someone must have known they were testing for a bug, and yet not
used the X86_BUG_ macros, which seems very unlikely.

> #define X86_BUG_F00F (NCAPINTS*32+ 0) /* Intel F00F bug */
> #define X86_BUG_FDIV (NCAPINTS*32+ 1) /* FPU FDIV bug */
> +#define X86_BUG_COMA (NCAPINTS*32+ 2) /* Cyrix 6x86 coma */

Just to make it a bit cleaner once we have more than one word of bug
tests, I would suggest macroizing this:

#define X86_BUG(x) (NCAPINTS*32 + (x))

... and then just ...

#define X86_BUG_F00F X86_BUG(0)
#define X86_BUG_FDIV X86_BUG(1)

... and so on.

The only reason we *don't* do that with the features is that they tend
to come chunkwise in the form of CPUID words.


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at