Re: [Openvortex-dev] ac97 codec read-write cleanup in cvs

[Jeff Muizelaar]

On Tue, 17 Jun 2003, Ryan Underwood wrote:

>
> Hi Manuel,
>
> On Mon, Jun 16, 2003 at 05:45:41PM -0400, Manuel Jander wrote:
> >
> > THe AC97 part semes to be necesary for some other switches that go to
> > oblivion for some reason... Ryan may explain that better than me.
>
> Perhaps, perhaps not. :) Basically, the alsa ac97 code "tests" for
> features that may or may not be available in all revisions of a
> particular ac97 codec by writing a value to some register, and then
> reading the value back to see if the change "took".  If it reads back
> the value it wrote, it assumes that feature to be available on the
> user's codec and enabled it.
>
> The vortex binary driver ( and the alsa au88x0 driver up to this point)
> had a problem in which the alsa driver's tests would fail and thus
> lots of nice features would be unavailable.  The test would call
> codec_write and then immediately call codec_read.  However, the
> codec_read was returning the value that was previously in the card at
> that time, and only after a next read or two would the new value
> actually be returned on a read.  I am assuming this is due to PCI posted
> writes, read prefetching, and/or a combination of both.  In any case,
> issuing two codec reads of the same memory in the codec_write function
> itself seems to have cleared up that problem.  (There was also a wrong
> conditional in the codec_read function that would never be true as far
> as I could tell.)  This was problem with the old ReadCodec function too,
> as I spent many hours trying to get my own driver hacks to work and
> wondering why wrong data was being read back from the card, which made
> me give up. :)
>
> > this is a shortcomming of the ALSA code, how about sending them a patch
> > (on the ALSA list) ? THey are very busy with some sort of "hdsp" card
> > right now, but they usually merge them in very quickly if the patch is
> > reasonable.
>
> Well, the changes to the alsa ac97 code were just for some feature
> additions.  The fixes were in the vortex driver itself.  I will send
> them patches, but we should also figure out whether or not the external
> amplifier should be enabled or disabled by default.
>
> > I'll help coding once i get out of my final exam nightmare :)
>
> :) I took a night off a *huge* project i'm working on to hack on the
> driver.  Sad that I couldn't do anything about the confused-streams
> problem, but at least I got a little done.
>
> > > The newest cvs has an adaptation of Ryan's ac97 codec
> > > read-write cleanup. But it doesn't have the hunks for
> > > /sound/pci/ac97/ac97_codec.c . Isn't that necessary
> > > for this new adaptation?
>
> Those are feature additions, at least one is necessary for the 4-speaker
> stereo output.  We can merge them here and have headaches later, or we
> can do without the features for now, try to get them merged upstream,
> and wait for the trickle down.  The other features like 3d stereo
> enhancement, amp power, were already present in the ac97 alsa support
> but were not being enabled due to buggy vortex driver ;)

I would prefer to wait for upstream to merge.

-Jeff