I don't think we can use the intel patch in any serious way. It can be
tripped up by SMP too easily. It can fault fetching the instruction that
it wants to decode which may in turn fault fetching the instruction... etc
when some smartarse mprotects the page as the other cpu faults on it and
they can certainly modify the instruction.
I think we are indeed better off taking the simple approach so that int3
does the right thing and nothing clever is attempted. If we have to do the
long clever decode we can shortcut that by testing current is vm86 or
has an ldt early on and skipping the complex case if so.
Alan