AMD chips delivering 0xFFFFFFFFFFFFFFFFF after suspend-resume #1
Description
News came out today of a bug in AMD chips where RdRand predictable returns a value or values of 0xFFFFFFFFFFFFFFFF (01) with carry flag==1 after a resume event. This is not good.
The correct thing to do is to put in a check for this combination in a way that doesn't lower entropy on actual random occurrences of 0xFFFFFFFFFFFFFFFF (very rare).
I need more details of the nature of the bug before the right algorithm can be identified.