Bill Sommerfeld wrote:
> one additional comment: an emulation may want to put data which may
> need cleanup.  how about adding a destructor function to the emualtion
> ops structure?
> 
> 	void *(*e_free_emuldata) __P((struct proc *));

Probably yes - I think that what Chris said makes sense, so I'm thinking
about adding three new members (I'm open to suggestions on better names):

	/*
	 * Called on exec - commonly, allocates new p_emuldata structure
	 * if appropriate for given emulation.
	 */
 	void *(*e_proc_exec) __P((struct proc *newproc));

	/*
	 * Called on fork1() - either copies the emuldata from parent,
	 * creates new, or does anything appropriate.
	 */
 	void *(*e_proc_fork) __P((struct proc *child, struct proc *parent));

	/*
	 * Called when process exits - commonly, frees any per-process
	 * emulation-specific resources.
	 */
 	void *(*e_proc_exit) __P((struct proc *));

sys_exec() should handle special case where the process would be replaced
by process of different emulation - in that case, e_proc_exit() of
previous emulation is called before e_proc_exec() of new emulation
would be called. Or maybe separate e_proc_free() should be added for this
case, too ?

Hmm, it's getting a bit more complicated than I originally though :(

Some more changes to the emulation subsystem would need to happen
to make emulations loadable via LKM in future - it should not be too hard,
even. But I personally don't feel doing that right now, though
it would save me lot of debugging time.

Jaromir
-- 
Jaromir Dolecek <jdolecek@NetBSD.org>      http://www.ics.muni.cz/~dolecek/
@@@@  Wanna a real operating system ? Go and get NetBSD, damn!  @@@@