[comp.os.mach] Issues regarding Mach 3.0 and Motorola 88k

jat@xavax.com (John Tamplin) (02/11/91)

I have just grabbed the newly-freed Mach 3.0 micro-kernel and I am interested
in a MC88100/88200 port of the machine dependent portion.  If anybody has
already done this (and is willing to give it out), I would definitely be
interested.

I have the Usenix Winter 90 tutorials on Mach (including the 68030 pmap stuff),
but I have some questions.

1) TLB invalidations:  since any 88100 can access any 88200, is there any
   reason to postpone invalidates?  It seems that once a processor decides
   to invalidate a range, that cpu could write invalidate commands to all
   88200's which share that memory (after gaining a cmmu semaphore).

2) Kernel address space:  as far as I can tell, there is not any need to
   access the kernel pmap from user mode.  On the 88200, there are separate
   translation trees for supervisor and user modes.  If kernel access from
   user mode is not required, this would save having to copy the kernel
   segment entries into every thread's pmap.

3) software vs hardware pmaps:  I am a bit confused about what is needed
   in this area, as the sample pmap code (68030) returns PMAP_NULL for
   software pmaps (size!=0 to pmap_create).  Can someone explain what is
   required of a software pmap, what it is for, etc?

-- 
John Tamplin						Xavax
jat@xavax.COM						2104 West Ferry Way
...!uunet!xavax!jat					Huntsville, AL 35801