[alsa-devel] [RFC PATCH 0/2] ALSA: hda - DP MST audio for Jack support

[Takashi Iwai]

On Mon, 09 Nov 2015 16:00:01 +0100,
Yang, Libin wrote:
> 
> Hi Takashi,
> 
> > -----Original Message-----
> > From: Takashi Iwai [mailto:tiwai at suse.de]
> > Sent: Monday, November 09, 2015 3:59 PM
> > To: Libin Yang
> > Cc: alsa-devel at alsa-project.org; mengdong.lin at linux.intel.com; Yang,
> > Libin
> > Subject: Re: [alsa-devel] [RFC PATCH 0/2] ALSA: hda - DP MST audio for
> > Jack support
> > 
> > On Wed, 04 Nov 2015 07:23:14 +0100,
> > Libin Yang wrote:
> > >
> > >
> > > On 11/04/2015 12:49 AM, Takashi Iwai wrote:
> > > > On Tue, 03 Nov 2015 09:42:54 +0100,
> > > > libin.yang at linux.intel.com wrote:
> > > >>
> > > >> From: Libin Yang <libin.yang at linux.intel.com>
> > > >>
> > > >> This is the patch set for Jack support for DP MST audio
> > > >>
> > > >> Libin Yang (2):
> > > >>    ALSA: hda - jack support DP MST audio
> > > >>    ALSA: hda - hdmi audio add dynamically jack binding to pin
> > > >
> > > > I'm somehow confused by this patch series.
> > > > Is the unsolicited event handling itself changed for MST?  I mean, you
> > > > cannot know which MST dev# is given beforehand, while you seem
> > trying
> > > > to assign such a number to the jack object.
> > >
> > > Oh, sorry I forgot to mention that we only operate on device entry 0
> > > so far in the code. It's misleading.
> > >
> > > These patches don't operate on MST audio. The full MST audio driver
> > > support will be implemented in the later patches. The patches help to
> > > prepare supporting MST audio.
> > >
> > > The purpose of the patches is try to dynamically attach the Jack.
> > 
> > Well, only judging from the quick glance, I don't buy this strategy,
> > sorry.  Currently for HDMI/DP, the jack is for PCM, not for pin.
> > Namely, user-space watches the jack notification to judge whether the
> > corresponding PCM is available or not.  You're trying to change this
> > scheme completely.
> 
> The patch is to make sure the jack is bound to PCM statically.
> This patch is needed because in old code,  it will create jack with:
> for (pin_idx = 0; pin_idx < spec->num_pins; pin_idx++) {
> 	...
> 	err = generic_hdmi_build_jack(codec, pin_idx);
> 	...
> }
> Which means the jack is created based on pin number and pin_idx.
> As a pin is bound statically to the PCM, the jack is statically bound to PCM. 
 
OK, so far, so good.

> Now we are using the dynamic pcm bound. If we use the old code to
> create the jack, the jack will be bound to pin. So we need the patch, 
> which will use the new code:
> for (pcm_idx = 0; pcm_idx < spec-> pcms_used; pcm_idx++) {
> 	...
> 	err = generic_hdmi_build_jack(codec, pcm_idx);
> 	...
> }
> which makes sure that the jack is bound to PCM (jack[0] is bound 
> to pcm[0], jack[1] is bound to pcm[1] and so on).
> 
> If there is no monitor, no pin is bound to the PCM. 
> So no pin is bound to the Jack. When there is a monitor connected, 
> the pin is bound to a proper PCM. So we should bind the PCM's jack
> to the pin. 
> 
> For example, jack0 is bound to the PCM 3 (the first jack is bound 
> to the first PCM statically). When a monitor is connected to pin 5, 
> PCM 3 will be assigned to the pin. We need bind jack 0 to pin 5. 
> When monitor is disconnected, we must unbind the jack 0 from pin 5. 
> These operation are done in function 
> snd_hda_jack_bind/ snd_hda_jack_unbind.

Then we shouldn't use the hda_jack.c code at all for HDMI/DP.
For HDMI/DP with the dynamic PCM binding, the only state to check is
its PCM assignment.  It doesn't have to issue the jack state check via
HDA verb at all.  This should make things a lot easier.


Takashi