lkml.org 
[lkml]   [1999]   [Jul]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Linux 2.2.11pre2 proposed patch
Steve Dodd wrote:
>
> On Tue, Jul 20, 1999 at 12:41:40PM +0100, BOSZORMENYI Zoltan wrote:
>
> > Well, then will you please remove the f00f bugfix as well
> > and develop a user space tool to switch it on?
> > (The sarcastic taste was intended. :-))
> > IMHO the processor bug workarounds should be in the
> > the kernel.
>
> I personally feel that all the (reasonably small) CPU workarounds should be
> in the kernel, but I seem to have lost that battle. The idea that all the

Maybe Alan Cox can be convinced this time with _proper_ information.
(At least it seemed so in his mail.)

> distributions are going to include start-up scripts which grep /proc/cpuinfo
> and run set6x86 accordingly is ludicrous wishful thinking IMHO.

Well, I think that the proper approach is that everything that controls
processor behaviour belongs to the kernel. Just an example: someone
complained to me that he set ARRs with set6x86 and /proc/mtrr did not
reflect the changes...

> To address your first point, I think the f00f bugfix has to be kernel space.
> It seems to mess with mm things (does this indicate how little I understand
> about it? :)

It changes the interrupt descriptor table layout, as far as I understand
it.
And I said it only for fairness. :-)

> > And if the bugfix itself causes the file corruption then
> > no matter where you enable the bugfix and which tool you use...
>
> Note that there are two different techniques for working around the bug:
>
> - set the NO_LOCK bit in CR<whatever>; can cause problems in obscure
> circumstances (Alan mentioned a graphics / accelerator card problem)

To repeat myself: this crept into linux-2.2.8, ...

> - do some undocumented magic provided by (or reverse engineered from)
> Cyrix, that seems to modify the CPU behaviour when executing the
> offending instruction, making to do a noop afterwards. I don't know
> of any problems caused by this.

... and this is already in 2.2.10. Don't you guys read the code others
write?
:-(

Best regards,
Zoltan Boszormenyi

--
Microsoft: It's where you don't want to go today.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:53    [W:0.037 / U:0.048 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site