> Since the VME framework thread seems to trickle away again,

It's not trickling away. I have re-done the sparc vme probes
and attach routines, and I'm currently figuring out what should
go in sparc/include/bus.h (there isn't one currently). I'm also
looking at drivers that would make easy candidates for `sys/dev/vme';
the `xd' and `xy' disk drivers look like good guinea-pigs.

> struct vmebus_softc{
> 	struct device sc_dev;
> 
> 	vme_chipset_tag_t sc_vct;
> 	bus_dma_tag_t sc_bdt;
> 	bus_cntlrdma_tag_t sc_bcdt;
> 	bus_lock_tag_t sc_bltt;
> 
> 	struct extent *vme32ext, *vme24ext, *vme16ext;
> };

We might want throw in some machine dependent state in here. At least
on the sparc, there's some special address space glue that needs to
be stored somewhere. Or one might do

	struct vme_softc_machdep {
		struct vmebus_softc;
		...
	}


> 
> /* nr of addresses given in the kernel configuration file */
> #define VME_NUMCFRANGES 3
> 
> struct vme_range {
> 	vme_addr_t offset;
> 	vme_size_t size;
> 	vme_am_t am;
> };

What do you propose the actual VME device configuration lines will look like?

-pk