Re: 2.1.63 - testing Pentium bug workaround..

Morten Welinder (terra@diku.dk)
Fri, 14 Nov 1997 03:59:38 +0100


Alan Cox (alan@lxorguk.ukuu.org.uk) writes

> Looking at other errata cmpxchg8b is also obviously "magic" because it
> has another errata when storing the 8 bytes split across a page boundary
> which as it happens isnt a bad one. The only other 8 byte stores appear
> to be FPU and MMX - neither of which you can lock

sgdt/sidt might do. (Well, it's _six_ bytes, not eight.)

pusha/pushad might. (8+ bytes, of course.)

btc/bts/btr/cmpxchg _misaligned_

Morten
terra@diku.dk