Fwd: [PATCH] ALSA: hda/generic: Add mic autoswitch support for dyn_adc_switch mode

[Zhang Heng]

-------- 转发的消息 --------
主题: 	Re: [PATCH] ALSA: hda/generic: Add mic autoswitch support for 
dyn_adc_switch mode
日期: 	Wed, 6 May 2026 14:19:14 +0800
发件人: 	Zhang Heng <zhangheng@xxxxxxxxxx>
收件人: 	Takashi Iwai <tiwai@xxxxxxx>




在 2026/5/5 0:25, Takashi Iwai 写道:
> On Sun, 03 May 2026 13:45:12 +0200,
> Zhang Heng wrote:
> > When auto_mic is not available but dyn_adc_switch mode is enabled
> > (e.g., on laptops with both front and rear mic jacks), this patch
> > enables automatic microphone switching based on jack detection.
> >
> > The patch includes three changes:
> >
> > 1. In check_dyn_adc_switch(): Register jack detect callbacks for
> > all input pins when dyn_adc_switch is enabled and auto_mic is
> > not available.
> >
> > 2. In call_mic_autoswitch(): Add handling for dyn_adc_switch mode
> > to check imux_pins[] for jack presence, searching from back to
> > front (last inserted wins).
> >
> > 3. In mux_select(): Notify Capture Source controls after switching
> > to sync with user-space (PulseAudio/PipeWire).
> >
> > Problem description:
> > On Ubuntu 20.04 (with older PulseAudio/PipeWire):
> > - Front mic is unplugged
> > - Plug in rear mic
> > - Jack event is reported correctly
> > - Volume control (PulseAudio/PipeWire) recognizes the rear mic
> > - But alsamixer does NOT switch to rear mic
> > - Codec also does NOT perform the switch
> >
> > The root cause is that in dyn_adc_switch mode without auto_mic,
> > the jack detect callback was not properly set up to trigger the
> > mic autoswitch. The call_mic_autoswitch() function only calls
> > mic_autoswitch_hook or snd_hda_gen_mic_autoswitch(), which rely
> > on auto_mic being enabled.
> >
> > Additionally, after mux_select() performs the switch, user-space
> > (PulseAudio/PipeWire) may not be aware of the path change,
> > causing Capture Switch to show 'off'.
> >
> > This patch fixes both issues by:
> > 1. Registering jack detect callbacks for all input pins in
> > dyn_adc_switch mode
> > 2. Notifying Capture Source controls after switching
> >
> > Tested on SN6186 codec with Ubuntu 20.04 and 25.10.
> >
> > Question to the community: Is this approach correct? Should additional
> > changes be made to handle mute state preservation, or is this purely
> > a user-space issue that requires updating PulseAudio/PipeWire?
> >
> > Testing and feedback are welcome.
> I think the basic idea is OK. We can treat the auto-mic with dynamic
> ADC switches, too.
>
> But it's not clear what's the actual intent in your code changes:
>
>
> > Signed-off-by: Zhang Heng <zhangheng@xxxxxxxxxx>
> > ---
> > sound/hda/codecs/generic.c | 82 ++++++++++++++++++++++++++++++++++++--
> > 1 file changed, 79 insertions(+), 3 deletions(-)
> >
> > diff --git a/sound/hda/codecs/generic.c b/sound/hda/codecs/generic.c
> > index 660a9f2c0ded..c536de10d8b8 100644
> > --- a/sound/hda/codecs/generic.c
> > +++ b/sound/hda/codecs/generic.c
> > @@ -27,6 +27,10 @@
> > #include "hda_beep.h"
> > #include "generic.h"
> > +/* Forward declaration */
> > +static void call_mic_autoswitch(struct hda_codec *codec,
> > + struct hda_jack_callback *jack);
> > +
> > /**
> > * snd_hda_gen_spec_init - initialize hda_gen_spec struct
> > @@ -3238,6 +3242,19 @@ static int check_dyn_adc_switch(struct 
> > hda_codec *codec)
> > if (!spec->dyn_adc_switch && spec->multi_cap_vol)
> > spec->num_adc_nids = 1;
> > + /* Enable mic autoswitch for dyn_adc_switch mode when auto_mic is 
> > not available */
> > + if (!spec->auto_mic && imux->num_items > 1) {
> > + int i;
> > + for (i = 0; i < imux->num_items; i++) {
> > + hda_nid_t pin = spec->imux_pins[i];
> > + if (!is_jack_detectable(codec, pin))
> > + continue;
> > + snd_hda_jack_detect_enable_callback(codec, pin,
> > + call_mic_autoswitch);
> > + }
> > + codec_dbg(codec, "Enable mic autoswitch for input sources\n");
> > + }
> IMO, this should be rather put at auto_mic_check_imux() instead,
> something like:
>
> @@ -4789,13 +4789,15 @@ static bool auto_mic_check_imux(struct 
> hda_codec *codec)
> const struct hda_input_mux *imux;
> int i;
> - imux = &spec->input_mux;
> - for (i = 0; i < spec->am_num_entries; i++) {
> - spec->am_entry[i].idx =
> - find_idx_in_nid_list(spec->am_entry[i].pin,
> - spec->imux_pins, imux->num_items);
> - if (spec->am_entry[i].idx < 0)
> - return false; /* no corresponding imux */
> + if (!spec->dyn_adc_switch) {
> + imux = &spec->input_mux;
> + for (i = 0; i < spec->am_num_entries; i++) {
> + spec->am_entry[i].idx =
> + find_idx_in_nid_list(spec->am_entry[i].pin,
> + spec->imux_pins, imux->num_items);
> + if (spec->am_entry[i].idx < 0)
> + return false; /* no corresponding imux */
> + }
> }
> /* we don't need the jack detection for the first pin */
In fact, auto_ic_check_imux cannot be executed on desktop computers at 
all. You can take a look at the check_outo_ic_availability section
for (int i = 0; i < cfg->num_inputs; i++) {
    hda_nid_t nid = cfg->inputs[i].pin;
    unsigned int attr;
    attr = snd_hda_codec_get_pincfg(codec, nid);
    attr = snd_hda_get_input_pin_attr(attr);
    if (types & (1 << attr))        For desktop computers, the rear mic 
and line are an attr, which will be directly returned here.
        return 0;  /* already occupied */
}

> > return 0;
> > }
> > @@ -4107,9 +4124,38 @@ static int mux_select(struct hda_codec *codec, 
> > unsigned int adc_idx,
> > path = get_input_path(codec, adc_idx, idx);
> > if (!path)
> > return 0;
> > - if (path->active)
> > - return 0;
> > - snd_hda_activate_path(codec, path, true, false);
> > + if (path->active) {
> > + codec_err(codec, "[MUX] mux_select: path already active, skip 
> > activate but still notify\n");
> > + } else {
> > + snd_hda_activate_path(codec, path, true, false);
> > + }
> > +
> > + /* Notify Input Source / Capture Source controls that the path has 
> > changed */
> > + if (!spec->auto_mic && spec->input_mux.num_items > 1) {
> > + struct snd_card *card = codec->card;
> > + struct snd_kcontrol *kctl;
> > + struct snd_ctl_elem_id id;
> > + int notified = 0;
> > +
> > + codec_err(codec, "[MUX] mux_select: notifying controls, 
> > num_items=%d\n", spec->input_mux.num_items);
> > +
> > + if (card) {
> > + list_for_each_entry(kctl, &card->controls, list) {
> > + if (kctl->id.iface != SNDRV_CTL_ELEM_IFACE_MIXER)
> > + continue;
> > + codec_err(codec, "[MUX] mux_select: found control '%s'\n", 
> > kctl->id.name);
> > + if (strncmp(kctl->id.name, "Capture Source", 14) == 0) {
> > + id = kctl->id;
> > + snd_ctl_notify(card, SNDRV_CTL_EVENT_MASK_VALUE, &id);
> > + codec_err(codec, "[MUX] mux_select: notified '%s'\n", kctl->id.name);
> > + notified = 1;
> > + }
> > + }
> > + }
> > + if (!notified)
> > + codec_err(codec, "[MUX] mux_select: no Input/Capture Source control 
> > found!\n");
> > + }
> > +
> Hmm, I can't follow the logic here. What does this code change try to
> achieve...?

If I open Alsamixer first without adding snd_ctl_notify, the ADC 
switching caused

by inserting and unplugging mic will not notify Alsamixer in real time, 
resulting

in asynchronous information and still retaining the state when it was 
opened.


I will create another patch for this notification later to prevent 
confusion in understanding.

> > if (spec->cap_sync_hook)
> > spec->cap_sync_hook(codec, NULL, NULL);
> > path_power_down_sync(codec, old_path);
> > @@ -4602,6 +4648,36 @@ static void call_mic_autoswitch(struct 
> > hda_codec *codec,
> > struct hda_jack_callback *jack)
> > {
> > struct hda_gen_spec *spec = codec->spec;
> > + const struct hda_input_mux *imux;
> > + int i;
> > +
> > + /* Handle dyn_adc_switch mode - use imux_pins[] */
> > + if (!spec->auto_mic && spec->dyn_adc_switch) {
> > + imux = &spec->input_mux;
> > + if (imux->num_items <= 1)
> > + goto fallback;
> > +
> > + codec_err(codec, "[AUTO_MIC] call_mic_autoswitch: dyn_adc_switch 
> > mode, checking %d inputs\n", imux->num_items);
> > +
> > + /* Search from back to front (last inserted wins) */
> > + for (i = imux->num_items - 1; i >= 0; i--) {
> > + hda_nid_t pin = spec->imux_pins[i];
> > + int det = is_jack_detectable(codec, pin);
> > + int state = snd_hda_jack_detect_state(codec, pin);
> > + codec_err(codec, "[AUTO_MIC] call_mic_autoswitch: imux[%d] 
> > pin=0x%02x, det=%d, state=%s\n",
> > + i, pin, det, state == HDA_JACK_PRESENT ? "PRESENT" : "ABSENT");
> > + if (det && state == HDA_JACK_PRESENT) {
> > + codec_err(codec, "[AUTO_MIC] call_mic_autoswitch: switching to imux 
> > %d\n", i);
> > + mux_select(codec, 0, i);
> > + return;
> > + }
> > + }
> > + /* No jack present, keep current selection (don't switch to imux 0) */
> > + codec_err(codec, "[AUTO_MIC] call_mic_autoswitch: no jack present, 
> > keeping current\n");
> > + return;
> > + }
> > +
> > +fallback:
> > if (spec->mic_autoswitch_hook)
> > spec->mic_autoswitch_hook(codec, jack);
> > else
> Ditto. Something too complex here.
> mux_select() should handle the dynamic ADC switching, per se.
>
> And, with the auto-mic mode, the driver shouldn't expose the Capture
> Source mixer control. That said, enabling the auto-mic would have a
> clear downside for the user of PA/PW, too.

I have added a lot of debugging information here, which will be 
simplified when submitting v2 later.