AMD has confirmed a report from researcher Matt Dillon of DragonFly BSD that there's a bug in some AMD processor families. Dillon discovered that running a very specific sequence of consecutive back-to-back pops and (near) return instructions can create a condition where the process incorrectly updates the stack pointer, leading to a segmentation fault.
The bug was reproduced on an Opteron 6168 and Phenom II X4 820 processor, but its unknown how many other chips are affected. It seems it's just ordinary errata, the bug is quite specific, it does not lead to data corruption and it's hard to reproduce:
We exchanged a few emails to try to come up with a good test case.
Owing to the difficulty of reproducing the bug I constructed a
fully bootable DFly operating system & test case USB image and
verified that the bug was present on my test box using that image.
AMD was then able to reproduce the bug using that image on their own
machines. Over the last few months they have been working through
the possibilities and today emailed me the confirmation that it was,
indeed, a cpu bug.