|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
Re: WARNING: Linux Intel Pentium Bug
Tim Newsham (newsham
LAVA.NET)Fri, 7 Nov 1997 12:49:41 -1000
- Messages sorted by: [ date ][ thread ][ subject ][ author ]
- Next message: Kragen \: "Re: Intel Pentium Bug"
- Previous message: George Imburgia: "Re: Intel Pentium Bug"
- In reply to: Kurt Seifried: "Re: WARNING: Linux Intel Pentium Bug"
- Next in thread: Alan Cox: "Re: WARNING: Linux Intel Pentium Bug"
> Ran it on my P166, kaboom, ran it on my new P166MMX, kaboom. I guess that
> would be a yes. This is kind of scary, is there any possibility of a
> kernel patch that can take care of this?
There are checks you could do from the kernel but they would not
be able to reliably detect an arbitrary program which would
cause this instruction to run. In other words, if I saw your
patch, i could easily write a program that gets around it.
quick ideas that have been discussed with others:
- check for the instruction
problem: i can synthesize the instruction at run time
- check to see if the code synthesizes and runs the instruction
problem: you cant do this statically (its undecideable)
and to do it dynamically you'd have to interpret every
instruction you executed (too costly)
- dont allow code to run out of stack
problem: i can synthesize it elsewhere
- dont allow code to run out of anything but the read-only
text segment.
problem: i can change the protection on the text segment
- force text segment to always be read-only
forcing code to run out of text could cause some existing programs
to break.
> -seifried
Tim N.
- Next message: Kragen \: "Re: Intel Pentium Bug"
- Previous message: George Imburgia: "Re: Intel Pentium Bug"
- In reply to: Kurt Seifried: "Re: WARNING: Linux Intel Pentium Bug"
- Next in thread: Alan Cox: "Re: WARNING: Linux Intel Pentium Bug"