On Mon, 18 Nov 2019 05:40:52 +0100, Hui Wang wrote: > > > On 2019/11/12 下午8:21, Hui Wang wrote: > > > > On 2019/11/12 上午12:04, Takashi Iwai wrote: > >> On Mon, 11 Nov 2019 16:33:45 +0100, > >> Takashi Iwai wrote: > >>> On Mon, 11 Nov 2019 15:45:02 +0100, > >>> Hui Wang wrote: > > > [snip] > >> On the second thought, I wonder whether eld_valid would be corrected > >> later by the graphics side at all. If yes, it's a timing issue, and > >> it can be corrected with the repolling. > >> > >> A totally untested patch is below. > > > > I will build a testing kernel with this patch and let the bug > > reporter test it. > > > > Thanks, > > > > Hui. > > Hello Takashi, > > Tested the patch, it didn't work. The driver always failed to read > the speaker allocation from snd_hdmi_get_eld_ati(). > > This is the dmesg after adding the patch: > > https://launchpadlibrarian.net/451420819/dmesg (both presence and > eld_valid bits are set, but can't get the speaker_alloc) So it's likely a bug in the graphics driver :) In anyway, it indicates that it's not about eld_valid check itself. The eld_valid was returned correctly together with the monitor_present flag. I guess the system worked casually with your patch to ignore eld_valid because we don't care much about the channel mapping if channels <= 2. IOW, another workaround would be to ignore the error if channels <= 2. But I wonder whether this state persists after this resume moment. Could you check what happens if you unload / reload the HD-audio driver? Does the read of spk_alloc still fail? thanks, Takashi > > Regards, > > Hui. > > > > > > > >> > >> thanks, > >> > >> Takashi > >> > >> --- a/sound/pci/hda/patch_hdmi.c > >> +++ b/sound/pci/hda/patch_hdmi.c > >> @@ -1549,19 +1549,25 @@ static bool > >> hdmi_present_sense_via_verbs(struct hdmi_spec_per_pin *per_pin, > >> do_repoll = true; > >> } > >> - if (do_repoll) > >> + do_repoll |= repoll && eld->eld_valid != eld->monitor_present; > >> + if (do_repoll) { > >> schedule_delayed_work(&per_pin->work, msecs_to_jiffies(300)); > >> - else > >> + ret = false; > >> + } else { > >> update_eld(codec, per_pin, eld); > >> - > >> - ret = !repoll || !eld->monitor_present || eld->eld_valid; > >> + per_pin->repoll_count = 0; > >> + ret = true; > >> + } > >> jack = snd_hda_jack_tbl_get(codec, pin_nid); > >> if (jack) { > >> jack->block_report = !ret; > >> - jack->pin_sense = (eld->monitor_present && eld->eld_valid) ? > >> - AC_PINSENSE_PRESENCE : 0; > >> + if (ret) { > >> + jack->pin_sense = (eld->monitor_present && > >> eld->eld_valid) ? > >> + AC_PINSENSE_PRESENCE : 0; > >> + } > >> } > >> + > >> mutex_unlock(&per_pin->lock); > >> return ret; > >> } > >> >
