there's no mechanism for separating kernel and usespace revocation sweeps

[tariqkurd-repo]

sstatus.CRG is a single bit covering the whole system
There is currently no way of separating userspace revocatoin and kernel space revocation.
The proposal is to split the bit into two, and have different CRG settings for user and kernel, both in sstatus.
An alternative would be to ahve a bit in mstatus to control the kernel pages, but then the kerneal would have to call M-mode software to update the generation which may be awkward

[jrtc27]

We have not even prototyped kernel revocation. Adding anything to the spec is premature and requires real prior research effort.