Re: [PATCH] x86/entry/64: randomize kernel stack offset upon syscall

[Edgecombe, Rick P]

On Fri, 2019-04-26 at 12:33 +0100, Reshetova, Elena wrote:
> 1) rdtsc or variations based on it (David proposed some CRC-based variants for
> > example)
Hi,

Could we repeatedly measure the time for a short syscall on a quiet system and
estimate the entropy we get from this? In some scenarios the attacker might have
less control of the timing as well.

Since this is a statistical defense, assuming the argument can be made that
there is at least some randomness in the timer, and could at least be out of the
control of an attacker sometimes, I wonder if this feature could be valuable
before the search for a faster stronger random number generator completes.

Could that be a way forward for now?

Thanks,

Rick