HEADSUP: Soekris and generic LED/lamp support in FreeBSD-current

[Scott Long]

On Fri, 7 Nov 2003, Mathew Kanner wrote:

> On Nov 07, Scott Long wrote:
> > On Fri, 7 Nov 2003, Alfred Perlstein wrote:
> > > * Simon L. Nielsen <simon at FreeBSD.org> [031107 12:48] wrote:
> > > > On 2003.11.07 08:43:11 -0800, Bruce A. Mah wrote:
> > > >
> > > > > I encourage people interested in writing / importing manpages for these
> > > > > drivers to self-organize on the doc@ list.  Note that even a "stub"
> > > > > manual page with a one-line description of a driver and the
> > > > > corresponding entries from the hardware notes would be of value.
> > > >
> > > > I have converted Bruce's list into a TODO list [1].  If anybody start
> > > > writing one of the missing manual pages, please update the TODO list to
> > > > avoid duplicate work.
> > > >
> > > > [1] http://www.freebsd.org/docproj/current.html#sec4manualpages
> > > > (after next web build)
> > >
> > > On the subject, there needs to be a sound(4) manpage that explains
> > > exactly which module to load for each chipset.  Getting sound on
> > > FreeBSD working isn't so cool.  While having the modules split up
> > > into a dozen or so parts for different chipsets is cool from
> > > an engineering perspective, it's pretty confusing to users.  Can't
> > > we just have a "kldload sound" thing that will load them all?
> >
> > We've had that for years. Loading snd_driver.ko loads all of the sound
> > drivers.  Unloading it then unloads all of the drivers that didn't attach.
> > I thought that it was documented somewhere but apparently it isn't.
>
>
> 	kldunload should unload all driver.  I plan to augment
> /dev/sndstat to show which driver is actually behind the device.

The current semantics are intentional (IIRC from talking with Cameron).
The idea was to make it easy to load the correct sound driver without
having to know specifically which one you needed, but not waste RAM by
loading all of them.  Modifying /dev/sndstat like you said would be
interesting, but changing the unload semantics then makes things less
convenient.

Scott