> put this kind of thing.   Of course, if there is some timing constraint
> which means it has to be done as a kernel module, my next preference
> would be a module which loaded, did its dirty deed, and then unloaded,
> again to avoid locking down unpageable kernel memory unnecessarily, and
> also making it easier to update the user-mode program to handle
> additional motherboards without forcing people to use development
> kernels.

What about people who really, really don't want modules enabled?  Boot-time
autodetection (module load time for people who use modules) with __initfuncs
gives you pretty much all the advantages, and lets people choose between
modules and builtin.

> (This is why I think PCMCIA support is best distributed
> outside the kernel, even though it can be a pain.)

PCMCIA is the only reason I use modules on any of the dozens of Linux systems 
I look after.  I would really like an _option_ to not use modules and instead
use __initfuncs (if a PCMCIA machine is only ever used in a fixed config).  If
you move boot-time config to run-time then you need to move things out of
__initfuncs in the non-modular case.. which isn't good :-(

David (who personally would prefer to see PCMCIA in the standard kernel, and 
       loose the array of different licenses covering it, but now doesn't
       know if it will ever happen since PCMCIA has moved to MPL).


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/