Hi Richard and Dave,

On Fri, 10 Nov 2000, Richard Earnshaw wrote:
> > Does anybody else have any more suggestions or comments on the above
> > list?
> 
> I would also suggest adding, under longer term.
> 
>   Make !BtNetBSD untangle the memory map and present a flat, unmanaged 
> memory map to the kernel boot process.  This currently has to be done 
> inside the kernel, which isn't really the best place for it.
> 
> I think this was discussed briefly once before on this list, so you might 
> be able to find something about it in the archives.

I totally agree ... I am currently working out a new sceme of booting
NetBSD on the RiscPC with just that feature ...  if my scheme works i'll
try to work it in the arm26 bootloader...

My initial strategy is swapping pages in a running RiscOS but for some
reason that isnt working yet... but i have good hopes of it succeeding...

My second strategy is killing RiscOS after loading the kernel in a piece
of memory like the old bootloader but then move the kernel around putting
it on its good place in a flat memory space in physical memory (i.e. MMU
off) and then calling the kernel.

I've fiddled a bit with the first strategy and RiscOS doesnt seem to mind
swapping pages around allthough the `OS_FindMemMapEntries' and the
`OS_ReadMemMapEntries' are getting their data from another datastructure
than the real page tables so i'll have to rebuild these functionalities to
get the good positions and then swap the pages directly in the PTs after
copying its contents ofcource... For doing this i also fiddled around with
processormodes and can now switch from Usr26 to Sup32 and back without
RiscOS caring :))

Take care,
Reinoud