[alsa-devel] Regression with .set_bias_level in pcm512x?
Hi!
I'm working on adding support for DAIFMT_CBM_CFM and DAIFMT_CBM_CFS to pcm512x, and it works nicely in the 3.10 kernel + Atmel-patches were I have developed the code. Now I'm trying to forward-port it and test it on 3.18 + other-Atmel-patches in preparation for submission. However, there seems to be a regression in that the (pre-existing) pcm512x_set_bias_level callback is never called in the 3.18 version.
I see that there has been some core work on set_bias_level, and suspect it has something to do with dapm and components, but I think someone else will find the problem much quicker than I am able to.
I have feebly tried to revert 0bd2ac3dae74ee25c5ea171cb572731c7a89c248 and 68f831c2724ab72c0088471b2ed1dc99e81948ef, but that didn't help. That also seems a bit backwards. So, what needs to be done in the pcm512x driver to have set_bias_level called?
Please help.
Cheers, Peter
On 01/13/2015 11:49 PM, Peter Rosin wrote:
Hi!
I'm working on adding support for DAIFMT_CBM_CFM and DAIFMT_CBM_CFS to pcm512x, and it works nicely in the 3.10 kernel + Atmel-patches were I have developed the code. Now I'm trying to forward-port it and test it on 3.18 + other-Atmel-patches in preparation for submission. However, there seems to be a regression in that the (pre-existing) pcm512x_set_bias_level callback is never called in the 3.18 version.
I see that there has been some core work on set_bias_level, and suspect it has something to do with dapm and components, but I think someone else will find the problem much quicker than I am able to.
I have feebly tried to revert 0bd2ac3dae74ee25c5ea171cb572731c7a89c248 and 68f831c2724ab72c0088471b2ed1dc99e81948ef, but that didn't help. That also seems a bit backwards. So, what needs to be done in the pcm512x driver to have set_bias_level called?
It's unlikely that these commits have anything to do with it.
I think it is more likely that the DAPM state (for whatever reason) never actually changes. Check the state of the DAPM context and all the widgets in /sys/kernel/debug/asoc/your_card/... and compare their state when the system is idle to when it should not be idle.
- Lars
Hi!
Lars-Peter Clausen wrote:
On 01/13/2015 11:49 PM, Peter Rosin wrote:
Hi!
I'm working on adding support for DAIFMT_CBM_CFM and
DAIFMT_CBM_CFS to
pcm512x, and it works nicely in the 3.10 kernel + Atmel-patches were I have developed the code. Now I'm trying to forward-port it and test it on 3.18 + other-Atmel-patches in preparation for submission. However, there seems to be a regression in that the (pre-existing) pcm512x_set_bias_level callback is never called in the 3.18 version.
I see that there has been some core work on set_bias_level, and suspect it has something to do with dapm and components, but I think someone else will find the problem much quicker than I am able to.
I have feebly tried to revert 0bd2ac3dae74ee25c5ea171cb572731c7a89c248 and 68f831c2724ab72c0088471b2ed1dc99e81948ef, but that didn't help. That also seems a bit backwards. So, what needs to be done in the pcm512x driver to have set_bias_level called?
It's unlikely that these commits have anything to do with it.
Agreed, I just wanted to show that I at least did try some things...
I think it is more likely that the DAPM state (for whatever reason) never actually changes. Check the state of the DAPM context and all the widgets in /sys/kernel/debug/asoc/your_card/... and compare their state when the system is idle to when it should not be idle.
There is indeed something strange going on with dapm. When things should be active, I have this (for the non-working 3.18 variant):
$ cat codec:pcm512x.0-004c/dapm/Playback \ codec:pcm512x.0-004c/dapm/DACL \ codec:pcm512x.0-004c/dapm/OUTL \ dapm/DAC1 \ dapm/MIX1 \ dapm/MUX1 \ dapm/OUT1 \ dapm/AIFINL Playback: Off in 1 out 0 stream Playback active out "static" "DACR" out "static" "DACL" DACL: Off in 1 out 0 in "static" "Playback" out "static" "OUTL" OUTL: Off in 1 out 0 in "static" "DACL" out "static" "DAC1" DAC1: Off in 1 out 0 in "static" "OUTL" out "static" "MIX1" MIX1: Off in 1 out 0 in "static" "DAC1" out "Mixer" "MUX1" MUX1: Off in 1 out 0 in "Mixer" "MIX1" out "static" "OUT1" OUT1: Off in 1 out 0 in "static" "MUX1" AIFINL: Off in 0 out 0 stream Playback inactive
When things are inactive, all "in 1" change to "in 0", and the top "stream Playback" change to "inactive".
bias_level of both the card and the codec are "Off", regardless.
For reference, I have attached my current card driver, which is given this:
sound { compatible = "axentia,asoc-linea-tse850";
axentia,model = "TSE850 @ Linea"; axentia,audio-routing = "DAC1", "OUTL", "DAC2", "OUTR";
axentia,ssc-controller = <&ssc0>; axentia,audio-codec = <&pcm5142>;
gpios = <&pioA 9 GPIO_ACTIVE_LOW>, <&pioA 8 GPIO_ACTIVE_LOW>, <&pioA 10 GPIO_ACTIVE_LOW>, <&pioA 11 GPIO_ACTIVE_LOW>;
status = "okay"; };
I suspect that it has something to do with the snd_soc_dapm_add_routes call in linea_pcm512x_init? (I "stole" that piece from omap-abe-twl6040)
Cheers, Peter
On 01/14/2015 10:31 AM, Peter Rosin wrote:
Hi!
Lars-Peter Clausen wrote:
On 01/13/2015 11:49 PM, Peter Rosin wrote:
Hi!
I'm working on adding support for DAIFMT_CBM_CFM and
DAIFMT_CBM_CFS to
pcm512x, and it works nicely in the 3.10 kernel + Atmel-patches were I have developed the code. Now I'm trying to forward-port it and test it on 3.18 + other-Atmel-patches in preparation for submission. However, there seems to be a regression in that the (pre-existing) pcm512x_set_bias_level callback is never called in the 3.18 version.
I see that there has been some core work on set_bias_level, and suspect it has something to do with dapm and components, but I think someone else will find the problem much quicker than I am able to.
I have feebly tried to revert 0bd2ac3dae74ee25c5ea171cb572731c7a89c248 and 68f831c2724ab72c0088471b2ed1dc99e81948ef, but that didn't help. That also seems a bit backwards. So, what needs to be done in the pcm512x driver to have set_bias_level called?
It's unlikely that these commits have anything to do with it.
Agreed, I just wanted to show that I at least did try some things...
I think it is more likely that the DAPM state (for whatever reason) never actually changes. Check the state of the DAPM context and all the widgets in /sys/kernel/debug/asoc/your_card/... and compare their state when the system is idle to when it should not be idle.
There is indeed something strange going on with dapm. When things should be active, I have this (for the non-working 3.18 variant):
$ cat codec:pcm512x.0-004c/dapm/Playback \ codec:pcm512x.0-004c/dapm/DACL \ codec:pcm512x.0-004c/dapm/OUTL \ dapm/DAC1 \ dapm/MIX1 \ dapm/MUX1 \ dapm/OUT1 \ dapm/AIFINL Playback: Off in 1 out 0 stream Playback active out "static" "DACR" out "static" "DACL" DACL: Off in 1 out 0 in "static" "Playback" out "static" "OUTL" OUTL: Off in 1 out 0 in "static" "DACL" out "static" "DAC1" DAC1: Off in 1 out 0 in "static" "OUTL" out "static" "MIX1" MIX1: Off in 1 out 0 in "static" "DAC1" out "Mixer" "MUX1" MUX1: Off in 1 out 0 in "Mixer" "MIX1" out "static" "OUT1" OUT1: Off in 1 out 0 in "static" "MUX1" AIFINL: Off in 0 out 0 stream Playback inactive
When things are inactive, all "in 1" change to "in 0", and the top "stream Playback" change to "inactive".
bias_level of both the card and the codec are "Off", regardless.
For reference, I have attached my current card driver, which is given this:
sound { compatible = "axentia,asoc-linea-tse850";
axentia,model = "TSE850 @ Linea"; axentia,audio-routing = "DAC1", "OUTL", "DAC2", "OUTR"; axentia,ssc-controller = <&ssc0>; axentia,audio-codec = <&pcm5142>; gpios = <&pioA 9 GPIO_ACTIVE_LOW>, <&pioA 8 GPIO_ACTIVE_LOW>, <&pioA 10 GPIO_ACTIVE_LOW>, <&pioA 11 GPIO_ACTIVE_LOW>; status = "okay";};
I suspect that it has something to do with the snd_soc_dapm_add_routes call in linea_pcm512x_init? (I "stole" that piece from omap-abe-twl6040)
There does not seem to be a audio sink in your DAPM graph. You have a source which is the Playback stream, hence a one the "1" for "in". But a widget will only power up if it is on a active path from a source to a sink. Try to add a speaker, lineout or similar widget to the end of your graph.
- Lars
Lars-Peter Clausen wrote:
On 01/14/2015 10:31 AM, Peter Rosin wrote:
Hi!
Lars-Peter Clausen wrote:
On 01/13/2015 11:49 PM, Peter Rosin wrote:
Hi!
I'm working on adding support for DAIFMT_CBM_CFM and
DAIFMT_CBM_CFS to
pcm512x, and it works nicely in the 3.10 kernel + Atmel-patches were I have developed the code. Now I'm trying to forward-port it and test it on 3.18 + other-Atmel-patches in preparation for submission. However, there seems to be a regression in that the (pre-existing) pcm512x_set_bias_level callback is never called in the 3.18 version.
I see that there has been some core work on set_bias_level, and suspect it has something to do with dapm and components, but I think someone else will find the problem much quicker than I am able to.
I have feebly tried to revert 0bd2ac3dae74ee25c5ea171cb572731c7a89c248 and 68f831c2724ab72c0088471b2ed1dc99e81948ef, but that didn't help. That also seems a bit backwards. So, what needs to be done in the pcm512x driver to have set_bias_level called?
It's unlikely that these commits have anything to do with it.
Agreed, I just wanted to show that I at least did try some things...
I think it is more likely that the DAPM state (for whatever reason) never actually changes. Check the state of the DAPM context and all the widgets in /sys/kernel/debug/asoc/your_card/... and compare their state when the system is idle to when it should not be idle.
There is indeed something strange going on with dapm. When things should be active, I have this (for the non-working 3.18 variant):
$ cat codec:pcm512x.0-004c/dapm/Playback \ codec:pcm512x.0-004c/dapm/DACL \ codec:pcm512x.0-004c/dapm/OUTL \ dapm/DAC1 \ dapm/MIX1 \ dapm/MUX1 \ dapm/OUT1 \ dapm/AIFINL Playback: Off in 1 out 0 stream Playback active out "static" "DACR" out "static" "DACL" DACL: Off in 1 out 0 in "static" "Playback" out "static" "OUTL" OUTL: Off in 1 out 0 in "static" "DACL" out "static" "DAC1" DAC1: Off in 1 out 0 in "static" "OUTL" out "static" "MIX1" MIX1: Off in 1 out 0 in "static" "DAC1" out "Mixer" "MUX1" MUX1: Off in 1 out 0 in "Mixer" "MIX1" out "static" "OUT1" OUT1: Off in 1 out 0 in "static" "MUX1" AIFINL: Off in 0 out 0 stream Playback inactive
When things are inactive, all "in 1" change to "in 0", and the top "stream Playback" change to "inactive".
bias_level of both the card and the codec are "Off", regardless.
For reference, I have attached my current card driver, which is given this:
sound { compatible = "axentia,asoc-linea-tse850";
axentia,model = "TSE850 @ Linea"; axentia,audio-routing = "DAC1", "OUTL", "DAC2", "OUTR"; axentia,ssc-controller = <&ssc0>; axentia,audio-codec = <&pcm5142>; gpios = <&pioA 9 GPIO_ACTIVE_LOW>, <&pioA 8 GPIO_ACTIVE_LOW>, <&pioA 10 GPIO_ACTIVE_LOW>, <&pioA 11 GPIO_ACTIVE_LOW>; status = "okay";};
I suspect that it has something to do with the snd_soc_dapm_add_routes call in linea_pcm512x_init? (I "stole" that piece from omap-abe-twl6040)
There does not seem to be a audio sink in your DAPM graph. You have a source which is the Playback stream, hence a one the "1" for "in". But a widget will only power up if it is on a active path from a source to a sink. Try to add a speaker, lineout or similar widget to the end of your graph.
Indeed. I changed "OUT1", "OUT2", "IN1" and "IN2" to be SND_SOC_DAPM_LINEs instead and pcm512x_set_bias_level is now called. Thanks! (and it works too!)
Cheers, Peter
participants (2)
-
Lars-Peter Clausen -
Peter Rosin