On Tue, Dec 17, 2013 at 10:50:02PM +0000, Mark Brown wrote: > On Thu, Dec 12, 2013 at 05:59:28PM +0800, Nicolin Chen wrote: > > > + mask = WM8962_MIXINL_TO_HPMIXL_MASK | WM8962_MIXINR_TO_HPMIXL_MASK | > > + WM8962_IN4L_TO_HPMIXL_MASK | WM8962_IN4R_TO_HPMIXL_MASK; > > + bypass |= snd_soc_read(codec, WM8962_HEADPHONE_MIXER_1) & mask; > > + bypass |= snd_soc_read(codec, WM8962_HEADPHONE_MIXER_2) & mask; > > + bypass |= snd_soc_read(codec, WM8962_SPEAKER_MIXER_1) & mask; > > + bypass |= snd_soc_read(codec, WM8962_SPEAKER_MIXER_2) & mask; > > + > > + /* Don't diable FLL if running multi-substreams or analogue bypass */ > > + if (codec_dai->active != 1 || bypass) > > + return 0; > > I don't think this works with the power down delay we do on playback - > the DAI will go inactive when closed but we'll still have the CODEC > active and using its clocks until the power down time has elapsed if > it's a playback DAI. Trying to reclock the device while active is at > best risky, even if it's muted. > > I do think refcounting from both here and the bias level changes is > going to be the most robust thing, that'd also avoid the need to peer > into the CODEC register map.
I've tried count reference way to handle FLL enabler/disabler here before I sent this version. But the result shows the FLL would be never disabled in hw_free() because the refcount is accumulated to 2, one from hw_params() and the other from set_bias_level(PREPARE), which just made this patch meaningless to me. So the reclocking with bypass checking seems to be the last resort I can figure out right here as the playback flow for 'aplay -Dhw:0 44k16bit.wav 48k24bit.wav' does need to reprogram the FLL during CODEC active. Thank you, Nicolin Chen -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
