Re: f00f workaround not needed
Mikael Pettersson (Mikael.Pettersson@sophia.inria.fr)
Wed, 26 Nov 1997 11:48:16 +0100 (MET)
On Tue Nov 25 1997, Cyril Chaboisseau wrote:
>
> just a word to say that since the bug came up, i tested it with all the
> hardware i have around (especially an old Compaq Deskpro with Pentium
> OverDrive PODP5V83 which was _not_ concerned by the bug)
>
> After upgrading to kernel 2.0.32, i noticed that the workaround was set up
> even though it wasn't necessary since the chip is insensible to the bug
>
> Of course this kernel is still working ok on my machine but is there any
> way to incorporate a check in the further f00f path (new) not to include
> workaround for specific pentium model
I suppose one could compile a list of steppings (chip revisions)
of those P5 chips are aren't vulnerable, and augment the workaround
enabling code accordingly. But this code would require constant
maintenance to be up to date with Intel's stepping lists.
Installing the workaround, executing the lock-up code, and backing out
the workaround if the kernel got a #UD instead of a #PF should also be
doable, but it's potentially messy.
However, since the new workaround (not the one in your vanilla 2.0.32
kernel) is essentially zero-overhead, I don't think we should worry
too much about this.
/Mikael