AMD cpu bug update -- AMD confirms!

[ejc]

Congratulations!
You spent a heck of a lot of time working on that -- nice to see that
the work paid off.

Eric

On Mon, Mar 5, 2012 at 1:27 PM, Matthew Dillon
<dillon at apollo.backplane.com> wrote:
> Â  Â AMD has confirmed the cpu bug that I analyzed. Â I am going to quote
> Â  Â the first paragraph. Â They gave me some more info and will be updating
> Â  Â their revision guide. Â To avoid any confusion (so they can finalize their
> Â  Â wording for the entry they will be putting in the revision guide) I'm
> Â  Â just including the confirmation here.
>
> Â  Â quote>>
>
> Â  Â "AMD has taken your example and also analyzed the segmentation fault and
> Â  Â the fill_sons_in_loop code. We confirm that you have found an erratum
> Â  Â with som e AMD processor families. The specific compiled version of
> Â  Â the fill_sons_in_loop code, through a very specific sequence of
> Â  Â consecutive back-to-back pops and (near) return instructions, can
> Â  Â create a condition where the process or incorrectly updates the
> Â  Â stack pointer."
>
> Â  Â <<endquote
>
> Â  Â We exchanged a few emails to try to come up with a good test case.
> Â  Â Owing to the difficulty of reproducing the bug I constructed a
> Â  Â fully bootable DFly operating system & test case USB image and
> Â  Â verified that the bug was present on my test box using that image.
> Â  Â AMD was then able to reproduce the bug using that image on their own
> Â  Â machines. Â Over the last few months they have been working through
> Â  Â the possibilities and today emailed me the confirmation that it was,
> Â  Â indeed, a cpu bug.
>
> Â  Â I'm pretty stoked... it isn't every day that a guy like me gets to
> Â  Â find an honest-to-god hardware bug in a major cpu!
>
> Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â -Matt
> Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â Matthew Dillon
> Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â <dillon at backplane.com>