Re: gpio-ucb1400

[Jean Delvare <jdelvare@suse.de>]

Le lundi 01 avril 2013 à 16:32 +0100, Mark Brown a écrit :
> On Mon, Apr 01, 2013 at 01:06:43PM +0200, Jean Delvare wrote:
> 
> > My point was, the upstream gpio-ucb1400 driver in its current form is
> > unusable (on top of being ugly.) I have no doubt that someone used that
> > driver successfully in some external kernel tree, but the code that was
> > merged upstream is incomplete and no good.
> 
> What the driver is doing is, unfortunately, the best practice for an
> AC'97 connected device like this - we've got an uncomfortable mix of
> an enumerable bus and platform data combined with a subsystem which has
> never had enough love to work in a nice way with the kernel.  The idea
> is that any boards which have the device will call the _set_data()
> function in their board-specific code.

OK but where are these users? I can't see any in the upstream kernel.

> Sadly we're a little short on volunteers to clean up the AC'97
> subsystem so this situation shows little sign of improving.

Well Marek proposed a cleanup which looked good [1], several times
apparently, but it was never applied.

[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2010-October/028656.html

-- 
Jean Delvare
Suse L3