[alsa-devel] sgtl5000 fails after suspend/resume
Hi,
I noticed that sgtl5000 does not work after a suspend/resume cycle:
# aplay /media/a2002011001-e02.wav Playing WAVE '/media/a2002011001-e02.wav' : Signed 16 bit Little Endian, Rate 44100 Hz, Stereo aplay: pcm_write:2051: write error: Input/output error
Tested several 4.16-rc1 (and also previous kernels like 4.4) and all of them failed.
I think it worked in some point in time.
Tested on a imx53qsb and imx6 wandboard.
Does anyone have any suggestions?
Thanks,
Fabio Estevam
On Thu, Feb 15, 2018 at 11:35 AM, Fabio Estevam festevam@gmail.com wrote:
Hi,
I noticed that sgtl5000 does not work after a suspend/resume cycle:
# aplay /media/a2002011001-e02.wav Playing WAVE '/media/a2002011001-e02.wav' : Signed 16 bit Little Endian, Rate 44100 Hz, Stereo aplay: pcm_write:2051: write error: Input/output error
Tested several 4.16-rc1 (and also previous kernels like 4.4) and all of them failed.
I think it worked in some point in time.
Tested on a imx53qsb and imx6 wandboard.
Does anyone have any suggestions?
Reverting the commit below on top of 4.16-rc1 makes sgtl5000 suspend/resume to work again:
commit 8419caa7270291e26f8b34b12b29680586c85d30 Author: Eric Nelson eric@nelint.com Date: Tue Jun 7 01:14:52 2016 +0200
ASoC: sgtl5000: Do not disable regulators in SND_SOC_BIAS_OFF
Disabling the SGTL5000 through regulators would certainly save more power than simply disabling the reference voltages as described in the data sheet, but won't properly restore things on resume.
This driver does not support active regulators. So we simply disable the reference bias currents.
Signed-off-by: Eric Nelson eric@nelint.com Signed-off-by: Clemens Gruber clemens.gruber@pqgruber.com Reviewed-by: Fabio Estevam fabio.estevam@nxp.com Signed-off-by: Mark Brown broonie@kernel.org
Or if I do the change below against linux-next it also works:
--- a/sound/soc/codecs/sgtl5000.c +++ b/sound/soc/codecs/sgtl5000.c @@ -880,8 +880,6 @@ static int sgtl5000_set_bias_level(struct snd_soc_component *component, SGTL5000_REFTOP_POWERUP); break; case SND_SOC_BIAS_OFF: - snd_soc_component_update_bits(component, SGTL5000_CHIP_ANA_POWER, - SGTL5000_REFTOP_POWERUP, 0); break; }
Any suggestions for a proper fix? Eric?
Thanks
On Thu, Feb 15, 2018 at 2:36 PM, Fabio Estevam festevam@gmail.com wrote:
Or if I do the change below against linux-next it also works:
--- a/sound/soc/codecs/sgtl5000.c +++ b/sound/soc/codecs/sgtl5000.c @@ -880,8 +880,6 @@ static int sgtl5000_set_bias_level(struct snd_soc_component *component, SGTL5000_REFTOP_POWERUP); break; case SND_SOC_BIAS_OFF:
snd_soc_component_update_bits(component,SGTL5000_CHIP_ANA_POWER,
SGTL5000_REFTOP_POWERUP, 0); break; }Any suggestions for a proper fix? Eric?
I sent a patch that that removes sgtl5000_set_bias_level() completely and fixes the sgtl5000 suspend/resume issue.
On Thu, Feb 15, 2018 at 02:36:00PM -0200, Fabio Estevam wrote:
Reverting the commit below on top of 4.16-rc1 makes sgtl5000 suspend/resume to work again:
@@ -880,8 +880,6 @@ static int sgtl5000_set_bias_level(struct snd_soc_component *component, SGTL5000_REFTOP_POWERUP); break; case SND_SOC_BIAS_OFF:
snd_soc_component_update_bits(component,SGTL5000_CHIP_ANA_POWER,
SGTL5000_REFTOP_POWERUP, 0); break; }
Any suggestions for a proper fix? Eric?
Does it need a register map resync when powering on again?
Hi Mark,
On Thu, Feb 15, 2018 at 4:16 PM, Mark Brown broonie@kernel.org wrote:
Does it need a register map resync when powering on again?
Thanks for your suggestion! This works:
--- a/sound/soc/codecs/sgtl5000.c +++ b/sound/soc/codecs/sgtl5000.c @@ -871,15 +871,26 @@ static int sgtl5000_pcm_hw_params(struct snd_pcm_substream *substream, static int sgtl5000_set_bias_level(struct snd_soc_component *component, enum snd_soc_bias_level level) { + struct sgtl5000_priv *sgtl = snd_soc_component_get_drvdata(component); + int ret; + switch (level) { case SND_SOC_BIAS_ON: case SND_SOC_BIAS_PREPARE: case SND_SOC_BIAS_STANDBY: + regcache_cache_only(sgtl->regmap, false); + ret = regcache_sync(sgtl->regmap); + if (ret) { + regcache_cache_only(sgtl->regmap, true); + return ret; + } + snd_soc_component_update_bits(component, SGTL5000_CHIP_ANA_POWER, SGTL5000_REFTOP_POWERUP, SGTL5000_REFTOP_POWERUP); break; case SND_SOC_BIAS_OFF: + regcache_cache_only(sgtl->regmap, true); snd_soc_component_update_bits(component, SGTL5000_CHIP_ANA_POWER, SGTL5000_REFTOP_POWERUP, 0); break;
Hi Fabio
On Thu, 2018-02-15 at 14:36 -0200, Fabio Estevam wrote:
On Thu, Feb 15, 2018 at 11:35 AM, Fabio Estevam festevam@gmail.com wrote:
Hi,
I noticed that sgtl5000 does not work after a suspend/resume cycle:
# aplay /media/a2002011001-e02.wav Playing WAVE '/media/a2002011001-e02.wav' : Signed 16 bit Little Endian, Rate 44100 Hz, Stereo aplay: pcm_write:2051: write error: Input/output error
Tested several 4.16-rc1 (and also previous kernels like 4.4) and all of them failed.
I think it worked in some point in time.
Tested on a imx53qsb and imx6 wandboard.
Does anyone have any suggestions?
Reverting the commit below on top of 4.16-rc1 makes sgtl5000 suspend/resume to work again:
commit 8419caa7270291e26f8b34b12b29680586c85d30 Author: Eric Nelson eric@nelint.com Date: Tue Jun 7 01:14:52 2016 +0200
ASoC: sgtl5000: Do not disable regulators in SND_SOC_BIAS_OFF Disabling the SGTL5000 through regulators would certainly save more power than simply disabling the reference voltages as described in the data sheet, but won't properly restore things on resume. This driver does not support active regulators. So we simply disable the reference bias currents. Signed-off-by: Eric Nelson <eric@nelint.com> Signed-off-by: Clemens Gruber <clemens.gruber@pqgruber.com> Reviewed-by: Fabio Estevam <fabio.estevam@nxp.com> Signed-off-by: Mark Brown <broonie@kernel.org>Or if I do the change below against linux-next it also works:
--- a/sound/soc/codecs/sgtl5000.c +++ b/sound/soc/codecs/sgtl5000.c @@ -880,8 +880,6 @@ static int sgtl5000_set_bias_level(struct snd_soc_component *component, SGTL5000_REFTOP_POWERUP); break; case SND_SOC_BIAS_OFF:
snd_soc_component_update_bits(component,SGTL5000_CHIP_ANA_POWER,
SGTL5000_REFTOP_POWERUP, 0); break; }Any suggestions for a proper fix? Eric?
I tested this on a Colibri iMX6DL with the same results. Vanilla 4.16-rc1 does resume but audio is silent after resume. The interrupted aplay process is in the process list and never finishes.
Both your fixes make aplay continue playing sound again until the end of the played sound file.
Strangely if I use our downstream kernel (based on NXP/linux-fslc 4.9-1.0.x-imx which has the 8419caa7270291e26f8b34b12b29680586c85d30 commit) suspend/resume works.
Max
participants (3)
-
Fabio Estevam -
Mark Brown -
Max Krummenacher