Re: 2.1.63 - testing Pentium bug workaround..

Linus Torvalds (torvalds@transmeta.com)
Wed, 12 Nov 1997 22:23:40 -0800 (PST)


On Thu, 13 Nov 1997, Frank Sweetser wrote:
>
> > [ pre-2.0.32-3 Pentium fix test ]
>
> gets a few lines past 'calibrating delay loop', then reboots the
> machine....

Ahh, forget about the pre-3 patch for now. Looking at it it has two
distinct problems:

- 2.0.x interrupt installation mucks directly with the IDT. Ugly, and I
had fixed that a long time ago in 2.1.x and had forgotten all about it.
This is a problem because the pentium-lock-up-patch creates a shadow
of the IDT, and now the interrupts won't be installed in the shadow.

- 2.0.x has a different kernel virtual mapping than 2.1.x, and needs
various "correction factors" to fix it up. This one is fairly easy
to fix.

Anyway, don't use pre-3 for now, it obviously needs more thought.

(I'm impressed by how well 2.1.x was able to integrate the patch - it's
very clean in 2.1. That's a good sign, I guess, and tends to indicate that
the fixes done since 2.0.x have been "good" fixes. Now when I look at the
2.0.x problems, I get bad vibes. I feel sorry for MS programmers, because
I'd bet they have even worse problems when trying to fix this..)

Linus