Dne 12. 02. 21 v 10:23 Takashi Iwai napsal(a):
On Thu, 11 Feb 2021 18:53:20 +0100, Jaroslav Kysela wrote:
Dne 11. 02. 21 v 18:15 Takashi Iwai napsal(a):
Jaroslav Kysela (5): ALSA: control - introduce snd_ctl_notify_one() helper ALSA: control - add layer registration routines ALSA: control - add generic LED trigger module as the new control layer ALSA: HDA - remove the custom implementation for the audio LED trigger ALSA: control - add sysfs support to the LED trigger module
One thing I still miss from the picture is how to deal with the case like AMD ACP. It has no mixer control to bundle with the LED trigger. Your idea is to make a (dummy) user element and tie the LED trigger with it?
Yes, the user-space code which guarantee the silence stream should create an user space control with the appropriate LED group access bits. The alsa-lib's softvol PCM plugin can do this silencing for example.
What control would it create? In the case of softvol, it's a volume control that really changes the volume. For the mute LED, it's a control turn on/off the mute? If so, I wonder what makes better than creating it from the kernel driver. (Of course, we can list up like "flexibility", etc, but it has a flip side of "complexity" and "fragility"...)
The current code handles both switch / volume for the marked control (assuming that the minimal value - usually zero - is full mute). And actually, there are snd_pcm_areas_silence() calls in the softvol plugin, so we know that the PCM stream is not passed to the application at this point.
Condition:
if (info.type == SNDRV_CTL_ELEM_TYPE_BOOLEAN || info.type == SNDRV_CTL_ELEM_TYPE_INTEGER) ... value.value.integer.value[i] != info.value.integer.min
The softvol plugin may be extended to add the mute switch control, of course.
Another slight concern is the possible regression: by moving the mute-LED mode enum stuff into the sysfs, user will get incompatibilities after the kernel update. And it's not that trivial to change the sysfs entry as default for each user. It needs some detailed documentation or some temporary workaround (e.g. keep providing the controls for now but warns if the value is changed from the default value via the controls).
I don't think that we have a user space application which is using those controls (Pulseaudio or so..) in an abstract way. I think that it's really minor issue. We should probably concentrate for the main designed purpose (notify about the mute / silent state) and handle those add-on features as an experimental stuff.
I'm sure that there are users of the reverse mic-mute LED ("follow capture" mode); the feature was added because of the explicit request from my colleague, and this mode works no matter whether ALSA native or PA is used.
I see. It looks like a corner case. The proposed sysfs based code is also user space independent. The issue with the HDA code is that it's card based, but the system wide LEDs should not be tied to one sound card. We are seeing that the hardware designers became very creative :-)
Jaroslav