[keyme]

Perhaps it is no longer realistic to push such a huge changeset into linux anymore. Could this be solved with some hypervisor layer? That is, a hypervisor doing most of the work (in rust) and a small support layer upstreamed into the kernel? Of course, no actual virtualization is even necessary. Just some kind of ABI to the kernel running underneath.

[gpm]

There was a proposal to do something like what you suggest in redox: https://www.redox-os.org/news/rsoc-2023-eny-1/

I'm not sure how much work has actually been put into it though.