[coyotos-dev] Updates to spec

[Jonathan S. Shapiro]

I have updated chapter 3 to provide a bunch of explicit case analysis
for run-in and run-out, and added some discussion in chapter 5 about the
theory of operation on the invocation block.

There is a remaining open issue: how is a fault handler able to resume
its victim? This is not currently answered by the spec, because I want
to talk it through with Jonathan Adams, but the intended answer is that
the message delivered to the fault handler includes a capability. This
is a special type of process capability known as a fault capability. It
conveys the authority to resume the process. This capability is
synthesized within the kernel during fault delivery and delivered to the
fault handler via the FCRB.

For this and other reasons, the FCRB needs to hold a single capability
payload argument c0. This is added to FCRB in the current document.

Sam: with a bit of back and forth through the spec this *may* now
provide enough information to figure out how all the cases of fault
handling work.


shap