At Wed, 10 Jun 2015 12:19:58 +0800, hwang4 wrote:
On 2015年06月09日 21:26, Hui Wang wrote:
On 06/09/2015 07:51 PM, Takashi Iwai wrote:
At Tue, 09 Jun 2015 17:38:33 +0800, hwang4 wrote:
I am trying to enable the subwoofer speaker on a HP laptop, on this machine, there are two speakers and one headphone, but the BIOS verb only enabled one speaker(nid 0xd) and one headphone(nid 0xb), I need to use quirk in the kernel driver to configure the second speaker (subwoofer speaker, nid 0x10). Under current alsa driver, the headphone will be assigned a dac (nid 0x13) and the 2 speakers will be assigned a dac (nid 0x14), this assignment is not good since 2 speakers share the same dac, this means 2 speakers can't work at the same time to support 4.0/2.1 channels.
On another Dell machine with realtek codec, there are also 2 speakers, 1 headphone and 2 dacs, on this machine, 1 speaker and 1 headphone are assigned 1 dac, and the other speaker is assigned another dac, so there is no problem for this machine to support 4.0/2.1 channels.
Through debugging, I found on Dell machine, the speaker nid only has one connection to dac (hardwired), so when driver assign dac to it, the map_single() can successfully assign the each dac to the 2 speakers respectively. But on that HP machine, the speaker has multiple connections for dac, the map_single() can't work for this machine.
The alsa-info.txt for that HP machine is at http://pastebin.ubuntu.com/11667947/ The alsa-info.txt for that Dell machine is at http://pastebin.ubuntu.com/11667982/
To fix this problem, I can use the quirk in the driver to make the speaker have only one connection for dac or do some change in the hda_generic.c like below, in your opinion, which one is better or probably you have a different idea to fix this problem?
diff --git a/sound/pci/hda/hda_generic.c b/sound/pci/hda/hda_generic.c index ac0db16..8194ff1 100644 --- a/sound/pci/hda/hda_generic.c +++ b/sound/pci/hda/hda_generic.c @@ -1363,6 +1363,7 @@ static int try_assign_dacs(struct hda_codec *codec, int num_outs, dac = try_dac(codec, get_primary_out(codec, i), pin); if (!dac) dac = try_dac(codec, dacs[0], pin); if (!dac) dac = try_dac(codec, get_primary_out(codec, i), pin); if (dac) { @@ -1762,6 +1763,12 @@ static int fill_and_eval_dacs(struct hda_codec *codec, if (err < 0) return err; badness += err;
/* if there are 2 speakers and both of them areassigned to the same dac,
we need to increase the badness for this situation,because in this situation
the 2 speakers can't work together to support4.0/2.1 channels */
if (cfg->speaker_outs == 2 && err <spec->extra_out_badness->no_dac &&
spec->multiout.extra_out_nid[1] == 0)badness += spec->extra_out_badness->no_dac;This doesn't look like the correct place to add badness for the missing surround speaker. This function is called for each path, so it's evaluated not only for speakers.
A cleaner way would be to prepare a different badness table for the speaker, and increase the value for shared_surr. An untested patch is below.
Got it, I will test it soon. Thanks.
Hi Takashi,
Your patch can fix the problem, it works very well.
OK, good to know. I'd like to test a bit more via hda-emu whether this gives any ill effects. So far, this seems fixing a few other machines, too, so it's a good thing to have in general.
Takashi