Ath works as module, not inkernel

[Jason Hellenthal]

On Mon, Aug 01, 2011 at 10:08:48AM +0800, Adrian Chadd wrote:
> Hi,
> 
> On 1 August 2011 08:56, richo <richo at psych0tik.net> wrote:
> > Hi -wireless,
> >
> > A while ago I posted with some issues with ath on 9.0-CURRENT, which were
> > fixed by a patch someone posted to the list.
> >
> > A few weeks later, ath broke again and neither a vanilla kernel nor one with
> > the patch worked, complaining of errors attaching HAL. Googling the terms
> > turned up only results from 2008.
> >
> > I didn't have time to look into it so I ignored it, figuring that it would
> > be
> > fixed in due course and that I'd pull when I had time.
> 
> Well, if noone reported the issue and noone chases the issue up, it's
> not likely to get magically fixed by itself. :)
> 
> > Over the weekend I spent some time on the problem, and it appeared that
> > building ath support into the kernel was the issue, my config with
> >
> >> device ath
> >> device ath_pci
> >> device ath_hal
> >
> > (Apologies if that's not spelled correctly, I'm not actually in front of the
> > machine)
> >
> > Doesn't load, however if I removed MODULES_OVERRIDE from make.conf and try
> > to
> > load those modules, I'm told that it's already in the kernel.
> >
> > I also built GENERIC from the config in SVN and no luck.
> >
> > Finally, I removed those device lines from my config, built with all modules
> > and enabled load_ath and load_ath_pci in loader.conf and wireless now works
> > flawlessly.
> >
> > I'm now curious about why- and what I can do to know how to chase this up
> > myself next time.
> 
> Well, can you please post boot messages, including the HAL load
> failure information?
> 
> This isn't the first time I've heard about strange issues like this;
> it almost always boils down to something strange in PCI/ACPI handling.
> 

Should be noted .... does this happen with a generic kernel ? The
default ATH setup works perfectly and I can attest to it working just
fine built in here.