Re: [LINUX:3243] Re: New Intel Pentium bug? (fwd)

ILKER GUNGOR (ilker@frye.com)
Fri, 14 Nov 1997 09:36:15 -0800

Yes, the bug is an illegal opcode bug. No program is supposed to generate
that opcode but somehow some programs (example: some ActiveX programs) do
generate it.
I don't know how they do it.
Essentially, the opcode turns out to be "Lock CmpChg8b % eax" which does
nothing but
put the CPU in an infinite loop, resulting in a crash of the system. This
may not be a
big problem for single user systems (at most you can reset the system), but
imagine what can happen when the computer is an internet server and somehow
someone generates this code. Crash!
By the way, this bug is only present in pentiums, not pentium II or PROs.

Regards,
ILKER GUNGOR
Software Engineer,
FRYE ELECTRONICS, INC.
ilker@frye.com
Phone :(503)620-2722
Fax :(503)639-0128

At 10:10 AM 11/14/97 +0200, you wrote:
>
>
>---------- Forwarded message ----------
>Date: Fri, 14 Nov 1997 12:22:58 +1100
>From: Mark A Brophy <mab@cs.monash.edu.au>
>To: f.debruijn@bhtafe.edu.au
>Cc: melblinux@luv.asn.au
>Subject: Re: New Intel Pentium bug?
>
>On Fri, Nov 14, 1997 at 10:13:45AM +0000, f.debruijn@bhtafe.edu.au wrote:
>> > To: melblinux@luv.asn.au
>> > Subject: New Intel Pentium bug?
>> > From: Nathan Bailey <Nathan.Bailey@cc.monash.edu.au>
>> > Date: Sun, 09 Nov 97 17:29:35 +1100
>>
>> > Anyone heard anything about this?
>> >
>> > ------- Forwarded Message
>> > Date: Sat, 08 Nov 1997 10:45:23 +1100 (EST)
>> > From: Kai OYang <Kai.Oyang@fcit.monash.edu.au>
>> > Subject: WARNING: New Intel Pentium bug...
>> >
>> > [... snip ...]
>> > Recently a new Pentium bug has been floating around in comp.sys.intel.
>> > It's a microcode bug that will hang a machine for all Intel Pentiums,
even
>> > in user mode, requiring a hard reset. I've tested it on
P5-233MMX/P5-120 with
>> > NT, DOS, Win95, Solaris for Intel, all hung solid. As I don't have an
AMD /
>> > Cyrix / P2 machine so don't know if they are also affected.
>> > [... snip ...]
>> > ------- End of Forwarded Message
>> >
>> >
>> Dr.Dobb's Journal August 1997 has an artical " inside the pentuim II
>> math bug by RobertR. Collins -rcollins@x86.org" and included on the
accompanying CD
>> is a programme (FISTBUG.EXE plus source code) to test for
>> the bug on your system.
>
>I think FISTBUG.EXE is actually a different bug. From memory, it's a
>problem with the carry bit being set, or not being set incorrectly in
>some cases. I know it affects the ppro, and I suppose the p2 as well.
>The new pentium bug is an illegal opcode, afaik and it works nicely. 8)
>
> Mark Brophy
>
>