Matt Thomas wrote:
> At 08:32 PM 1/11/98 -0500, Todd Vierling wrote:
> >Skip the diffs.  I have some thinking to do on some of the kernel source
> >first, but hopefully I can have something decent looking tomorrow anyway.
> 
> I'm going for a semi-radical change when we do this:
> 
> Can we change bdevsw/cdevsw to be struct Xdevsw *Xdevsw[]; (ie. indirect)
> and let drivers register (ie. Xdevsw[PTY_MAJOR] = &pty_bdevsw;).  (or, even
> better, have cdevsw_add or bdevsw_add to do the dirty work).
> 
> This would doing LKM's and reduce or eliminate the need for conf.c.

I don't think that has to be done at the same time, since it's only an
internal thing and largely orthogonal to what Todd's going to be doing
already.  While it wouldn't be too hard to implement, and I think
that at least some parts of it are good things to do (and have
precendent in e.g. BSDI if i recall correctly, and could be a start at
doing a devfs, and... 8-), I'd like the first pass of the new
interface to show as with minimal changes to the existing system.

Could you propose this again, say, a week or two after we switch to
using all 32 bits of dev_t?  8-)

Or do you think it must be done as part of the first pass?



thanks,

chris