Subject: Re: removing double mapping of kstack.
To: None <briggs@puma.bevd.blacksburg.va.us, chopps@water.emich.edu>
From: Chris Torek <torek@bsdi.com>
List: port-m68k
Date: 05/08/1995 08:44:51
>Perhaps a return to another function in locore.s that then calls
>a function in trap.c to finalize the fork() (i.e. syscall debug,
>userret(), etc..)

This is pretty much the Right Answer -- a newly forked child should
just start out with a state that does a `finish child return'
`system call' from scratch.  It might even be worth allocating a
syscall number to this, just so that locore's can simply call
syscall() as if from a user trap.  I am not sure if there is
anything you would want to omit from such a call (perhaps the
`cnt.v_syscall++', though one could just have the `finish_child'
call do a `cnt.v_syscall--' to fix it up :-) ).

Chris