On Thu, 9 May 2002, Matt Thomas wrote:

> At 02:37 AM 5/9/2002, MAEKAWA Masahide wrote:
> >Here is a proposal for a framework to support the auto-generation of
> >block/character device switch tables by config(8).
>
> My only suggestion is to not have separate cdevsw/bdevsw structs but
> merge them into a struct devsw but still have separate bdevsw and
> cdevsw tables.  This would also that devsw_add only needs to be
> called by a driver but its parameters would need to change to include
> both CHR and BLK major nos.  There should also be defines for meaning
> no block major number (-1 seems natural).

I definitely agree devsw_add should do both at once. Otherwise how else do
we update chr2blk and blk2chr for lkms? :-)

Do we want to unify bdevsw and cdevsw? As MAEKAWA-san points out, that
should be a seperate thread.

If we decide we should do it, this change could well be the right time to
do it. Since we're doing a major rototill of the devsw code, we can do it
all at once.

Take care,

Bill