Zachary Amsden wrote:
> We need to have a discussion and spec out the exact steps to be taken
> during init / start-of-day.
Yes. Since it's not really viable to have hypervisor-specific
sub-arches and still be able to do boot-time paravirt interface
selection, it means we need to look at parallel structure which allows
the different pieces of hypervisor-specific code to co-exist, and that
will look very similar to pieces of the existing subarch code.
As a starting point, perhaps we should have a mach-paravirt subarch,
which can encapsulate the stuff we need. It could also be the template
for getting rid of the existing subarch stuff altogether. Which should
make you happy...
What are the proposals on the table for start of day? Two entrypoints
for native and hypervisor, an entrypoint for each hypervisor, or
something else?
J
