On 24 Jun 2021, at 04:23, Mark Jones mailto:mpj@pdx.edu> wrote:
On the verification side, I've long been interested in the question of how we protect against covert channels resulting from hardware state components that are undocumented, or that don't work as intended, or that are not captured in a formal model. [There was a news item in the last few weeks about a (seemingly harmless) flaw of this kind in the Apple M1 processor, which is probably why this was on my mind again.]
You’ll need a full reset of microarchitectural state. I’m working on a paper with ETH colleagues (follow-up from https://ts.data61.csiro.au/publications/csiro_full_text/Wistoff_SGBH_21.pdf) which shows that a comprehensive reset is feasible and cheap.
Gernot
