Re: [TECH TOPIC] Improving kernel security & integrity by generalizing ad-hoc safety mechanisms

["Chuck Wolber" <chuck@wolber.net>]

On Wed, Sep 9, 2026, at 7:44 AM, Gabriele Monaco wrote:

[...]

> Whether RV can be made free of interference to fit the functional
> safety use cases is still an open problem, though.

I think it fits, but not in the way it is being proposed.

It is less about detection and more about what happens when RV detects a
non-comforming state. You can only handle runtime non-conformance in a
"safe" way if you know *why* you are non-comformant in the first place.

This is why an ER doctor is at a significant disadvantage if they have
no knowledge of what got you there in the first place. Attempting to
remedy the wrong thing often kills the patient.

My understanding of Igor's fences proposal is that it attempts to bound
the state space rather than detect non-conformance.

If we stop using the word "safety" and replace it with the more correct
and far less glamorous phrase "deterministic behavior", then pruning the
state space becomes an obvious first choice.

Detection is in there too, but you have to have a plan for what to do
when you detect something. And that only becomes feasible when the state
space is manageable.

..Ch:W..