On Wednesday 13 June 2001  6:24 am, Matt Thomas wrote:
> At 12:52 AM 6/13/2001 +0100, Chris Gilbert wrote:
> >On Tuesday 12 June 2001  6:10 pm, Matt Thomas wrote:
> > > Module Name:  syssrc
> > > Committed By: matt
> > > Date:         Tue Jun 12 17:10:26 UTC 2001
> > >
> > > Modified Files:
> > >       syssrc/sys/arch/arm/conf: files.footbridge
> > >       syssrc/sys/arch/arm/footbridge: footbridge_pci.c
> > >       syssrc/sys/arch/netwinder/conf: files.netwinder
> > >
> > > Log Message:
> > > Move netwinder_machdep.c to files.netwinder.  Eliminate opt_netwinder.h
> > > and use #ifdef netwinder insead.  XXX ebsa285_machdep.c should move to
> > > the cats directory and files.cats
> >
> >I was debating this.  It's a matter of whether we want to treat EBSA285 as
> > a seperate arch dir, or as part of cats.  cats is really an EBSA285 with
> > a few extras.  IE the 2 are near identical.  Should I just make an
> > EBSA285 kernel in the cats dir?
>
> How hard is it to get a GENERIC kernel what will boot off either?  One
> thing we are trying to do is to have a single GENERIC config that will work
> for all supported under a given port.  There has to be a way to detect
> whether you are being run on a CAT .vs. a EBSA285.

We can tell from the bootloader magic number.  As each bootloader passes a 
different magic number, IE BT_MAGIC_NUMBER_EBSA or BT_MAGIC_NUMBER_CATS, so 
we can set a global based on that.  The other alternative is that CATS has a 
pci bus and ebsa doesn't, so the existance of that could be used, but that 
would require more work as things like pmap having allocated memory into the 
ISA intersection area into a different free pool should really be undone.

It does mean the one or two places where it makes a difference will need 
tweaking, but that's not hard.

Does anyone have an EBSA they could test changes on?

Cheers,
Chris