On Tue, Mar 19, 2019 at 11:52:32PM +0000, Jethro Beekman wrote:
> > And IMO encl.{c,h} belongs in the "driver" code, but to let the driver be
> > loadable it got shoved into the core subsystem. All of that code is
> > specific to running enclaves in the host, i.e. it shouldn't exist if I
> > compile out the driver entirely (in a future world where I want the core
> > SGX subsystem for virtualization purposes).
>
> Your argument here is "something that belongs in the driver isn't, therefore
> we shouldn't have a loadable driver". This seems backwards to me. Instead,
> we should see what interface would be needed so that this stuff *can* be in
> the driver.
Yes, all the code that needs to be in there to deal with enclaves is
already there in v19. Only the ioctl's live in the driver.
The way I see it the driver part was wrong before and now the split
is way than before.
/Jarkko
