Patch for config(1) to auto-generate LKMs

[Hiroyuki Bessho <bsh%grotto.jp@localhost>]

Folks,

   Although it took me much more time than I had expected (mainly 
because I couldn't find enough spare time), finally I made the first 
patch to config(1) to auto-generate LKMs from config(5) with new syntax.

  I put the patch and some files at
    ftp://ftp.netbsd.org/pub/NetBSD/misc/bsh/config_lkm/patch1/

  It only supports LKMs for devices; Pseudo-device support is not
finished, and neither file-systems, syscalls, or options are not
supported.  Some devices have problems to become LKMs as I mentions in
README.txt.  Many devices have not been tested, so some modules may
not even be loadable.  I have tested some device driver LKMs on my
laptop, and some of them worked fine.  Others didn't work because of
various reasons; I think I need to modify kernel source codes to fix
them.  (I'll come up with a list of success/failure later)

  Currently the patched config supports only one "module grouping
rule": one-attribute-in-one-module, which is the good default.  I had
a plan to add an option to config(1) to gather many things into one
module to reduce the number of modules, but it is not implemented yet.

  I changed my mind on one thing from the discussion at this list; I
thought I should put everything in one Makefile, but I switched to
one-makefile-for-one-module way.  I actually implemented the single
Makefile way and it worked for most cases but I couldn't find a good
workaround to handle ath device which has its own extra rules in
Makefile.ath.inc.

  It still has many problems and lots to do, but I believe it is a
good step forward.

  Please give a try and make some comments.

TIA.