At Wed, 13 Jun 2012 11:52:46 +0200, David Henningsson wrote:
2012-06-12 18:19, Takashi Iwai skrev:
At Mon, 11 Jun 2012 13:58:43 +0200, David Henningsson wrote:
On 06/08/2012 10:52 AM, Takashi Iwai wrote:
At Thu, 7 Jun 2012 15:52:08 +0200, David Henningsson wrote:
PulseAudio sometimes have difficulties knowing that there is a "Speaker" or "Internal Mic", if they have no individual volume controls or selectors. As a result, only e g "Headphone" might be created for a laptop, but no "Speaker". To help out, create phantom jacks (that are always present, at least for now) for "Speaker", "Internal Mic" etc, in case we detect them. The naming convention is e g "Speaker Phantom Jack".
In order not to pollute the /dev/input namespace with even more devices, these are added to the kcontrols only, not the input devices.
An implementation note: The possibility to override how to actually do jack sensing opens up for more advanced usage of the jack infrastructure in the future (e g letting the headphone jack detection set presence status for both headphone and mic in case of a headset jack, among other things).
Signed-off-by: David Henningssondavid.henningsson@canonical.com
sound/pci/hda/hda_jack.c | 64 ++++++++++++++++++++++++++++++++++------------ sound/pci/hda/hda_jack.h | 5 ++++ 2 files changed, 52 insertions(+), 17 deletions(-)
diff --git a/sound/pci/hda/hda_jack.c b/sound/pci/hda/hda_jack.c index 2dd1c11..c5569f1 100644 --- a/sound/pci/hda/hda_jack.c +++ b/sound/pci/hda/hda_jack.c @@ -133,6 +133,13 @@ static void jack_detect_update(struct hda_codec *codec, } }
+static void jack_detect_always_present(struct hda_codec *codec,
struct hda_jack_tbl *jack)+{
- jack->pin_sense = AC_PINSENSE_PRESENCE;
- jack->jack_dirty = 0;
+}
- /**
- snd_hda_set_dirty_all - Mark all the cached as dirty
@@ -162,7 +169,8 @@ u32 snd_hda_pin_sense(struct hda_codec *codec, hda_nid_t nid) { struct hda_jack_tbl *jack = snd_hda_jack_tbl_get(codec, nid); if (jack) {
jack_detect_update(codec, jack);
if (jack->update_func)jack->update_func(codec, jack);IMO, it's easier to modify jack_detect_update(). Instead of adding update_func pointer, add a new flag phantom_jack (can be a bit field), then in jack_detect_update():
The idea was to open up for other types of jack overrides as well, but if you prefer to keep it simpler while we can, I'm okay with that.
I'm attaching the modified patch.
Thanks, this looks better to me. Is it OK to merge this to 3.6 tree?
Hmm, good question. I guess the answer is, it would be nice to have in 3.5 but not strictly necessary.
With the Ubuntu hat on, the autumn version of Ubuntu is more likely to use 3.5 (but maybe 3.6, it has not yet been decided), and I'm going to need this fix in there, but if necessary I can backport it into the Ubuntu kernel.
With the PulseAudio hat on, as it helps us resolve a problem that hits quite a few machines, it would be good to have it ASAP, but OTOH, since we're currently discussing the PulseAudio release schedule, it's difficult to tell whether we will actually have a PulseAudio version out that uses these new phantom jacks anyway before Linux 3.6 is released.
(and with the kernel hat on, I have to admit it isn't strictly a regression fix, and that the merge window is over...)
OK, I'm wearing the kernel hat on ;) Applied to topic/hda branch now.
thanks,
Takashi