In article <H7800z.tK@tac.nyc.ny.us>,
	christos@zoulas.com (Christos Zoulas) writes:
> Well, another way to do this, is to just do some accounting in the GDT
> entries and when we run out, steal one from another process (policy to
> choose the victim to be determined). This way we will not have such
> limits, and the system will be able to run many more processes.

Is there a good reason why each process needs a GDT entry at all?
Couldn't the scheduler just use two alternating GDT entries? Each
process could have its GDT entry (only 8 bytes) in the architecture
depent part of the process structure. The scheduler would copy it into
the unused GDT entry, load its number into the task register and
mark the user GDT entry as unused.

	Kind regards

-- 
Matthias Scheler                                  http://scheler.de/~matthias/