Re: [PATCH] Mutilated form of Andi Kleen's AMD prefetch errata patch

[Dave Jones <davej@redhat.com>]

On Tue, Sep 30, 2003 at 08:38:14AM +0100, Jamie Lokier wrote:
 > diff -urN --exclude-from=dontdiff orig-2.6.0-test6/arch/i386/Kconfig dual-2.6.0-test6/arch/i386/Kconfig
 > --- orig-2.6.0-test6/arch/i386/Kconfig	2003-09-30 05:39:33.467103692 +0100
 > +++ dual-2.6.0-test6/arch/i386/Kconfig	2003-09-30 06:05:19.868623767 +0100
 > @@ -397,6 +397,12 @@
 >  	depends on MWINCHIP3D || MWINCHIP2 || MWINCHIPC6
 >  	default y
 >  
 > +config X86_PREFETCH_FIXUP
 > +	bool
 > +	depends on MK7 || MK8
 > +	default y
 > +
 > +
 >  config HPET_TIMER
 >  	bool "HPET Timer Support"
 >  	help
 > diff -urN --exclude-from=dontdiff orig-2.6.0-test6/arch/i386/kernel/cpu/common.c dual-2.6.0-test6/arch/i386/kernel/cpu/common.c
 > --- orig-2.6.0-test6/arch/i386/kernel/cpu/common.c	2003-09-30 05:39:33.871035696 +0100
 > +++ dual-2.6.0-test6/arch/i386/kernel/cpu/common.c	2003-09-30 06:03:46.222851345 +0100
 > @@ -327,6 +327,17 @@
 >  	 * we do "generic changes."
 >  	 */
 >  
 > +	/* Prefetch works ok? */
 > +#ifndef CONFIG_X86_PREFETCH_FIXUP
 > +	if (c->x86_vendor != X86_VENDOR_AMD || c->x86 < 6)
 > +#endif
 > +	{
 > +		if (cpu_has(c, X86_FEATURE_XMM))
 > +			set_bit(X86_FEATURE_XMM_PREFETCH, c->x86_capability);
 > +		if (cpu_has(c, X86_FEATURE_3DNOW))
 > +			set_bit(X86_FEATURE_3DNOW_PREFETCH, c->x86_capability);
 > +	}
 > +
 >  	/* TSC disabled? */
 >  	if ( tsc_disable )

This looks to be completely gratuitous. Why disable it when we have the
ability to work around it ?

		Dave

-- 
 Dave Jones     http://www.codemonkey.org.uk