Re: Cyrix 6x86MX and Centaur C6 CPUs in 2.1.102

Linus Torvalds (torvalds@transmeta.com)
Tue, 19 May 1998 15:56:59 -0700 (PDT)


On Tue, 19 May 1998, Trevor Johnson wrote:
> >
> > Anybody want to do something like this?
>
> Here's my attempt. This patch:
>
> - moves the X86_FEATURE_ macros from arch/i386/mm/init.c to
> include/asm-i386/processor.h
> - defines X86_FEATURE_BADTSC, X86_FEATURE_FCMOV, X86_FEATURE_MMX,
> X86_FEATURE_OSFXSR, X86_FEATURE_AMD3D, and several unused, numbered
> capabilities
> - adds a badtsc capability and tsc_bug bug to /proc/cpuinfo

Ok, this looks otherwise ok, but the badtsc capability mustn't be in the
same word as the other capabilities, because that will result in problems
in the future when intel adds new capability masks in their hardware.

So I'd really like to see a "anti-capability" mask. So you'd have

X86_FEATURE_FPU 1 /* The chip as a FPU */
X86_FEATURE_VME 2 /* Virtual x86 mode extensions */
X86_FEATURE_DE 4 /* Debug extensions */
...

X86_BUG_WP 1 /* WP doesn't work in supervisor mode */
X86_BUG_HLT 2 /* "hlt" can hang */
X86_BUG_IRQ13 4 /* exception 16 doesn't work */
X86_BUG_FDIV 8 /* fdiv is broken */
X86_BUG_POPAD 16 /* popad can crash the machine */
X86_BUG_F00F 32 /* F0 0F can hang the machine */
X86_BUG_TSC 64 /* TSC doesn't increment reliably */
...

and have separate words for these flags (right now the "bug" information
is spread out in various flags rather than concentrated in one word).

Would you be willing to re-do the patch with something like the above?

Linus

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu