stk1160: Let the driver setup the device's internal AC97 codec

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

Signed-off-by Marcel Hasler <[email protected]>

---
 drivers/media/usb/stk1160/stk1160-ac97.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

drivers/media/usb/stk1160/stk1160-ac97.c +5 -0

index 95648ac..60327af 100644
@@ -28,6 +28,11 @@
 #include "stk1160.h"
 #include "stk1160-reg.h"
 
+static u8 gain = 8;
+
+module_param(gain, byte, 0444);
+MODULE_PARM_DESC(gain, "Set capture gain level if AC97 codec is available (0-15, default: 8)");
+
 static void stk1160_write_ac97(struct stk1160 *dev, u16 reg, u16 value)
 {
 	/* Set codec register address */

@@ -136,7 +141,10 @@ void stk1160_ac97_setup(struct stk1160 *dev)
 	stk1160_write_ac97(dev, 0x16, 0x0808); /* Aux volume */

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Either way, if you still think this solution is "crap", feel free to modify the patches in any way you see fit. I've wasted too much time on this already, and since I'm not being paid for it, I don't intend to put any more effort into this.

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Best regards
Marcel

2016-12-05 22:06 GMT+01:00 Marcel Hasler <[email protected]>:

Quoted 120 lines

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

Quoted 6 lines

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Best regards
Marcel

One more thing...

The driver currently uses a lot of "magic numbers", both for the AC97 register addresses as well as the STK1160 register contents. That makes it a bit difficult to read unless you happen to have the datasheet open. Would it maybe be better to add defines for those, especially if we're going to poll individual bits?

I usually prefer
that approach myself. Would you put the defines for the AC97 chip registers into stk1160-reg.h or keep them in stk1160-ac97.c since they're only used there?

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Best regards
Marcel

Em Mon, 5 Dec 2016 22:06:59 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 113 lines

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

Quoted 6 lines

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

Sounds ok. My experience with AC97 on em28xx is that, as new devices were added, the delay needed for AC97 varied on some of those new devices. That's why checking if AC97 is ready before writing was added to its code.

Quoted 5 lines

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Quoted 2 lines

Best regards
Marcel

Thanks,
Mauro

Em Mon, 5 Dec 2016 22:06:59 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 113 lines

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

Quoted 6 lines

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

Sounds ok. My experience with AC97 on em28xx is that, as new devices were added, the delay needed for AC97 varied on some of those new devices. That's why checking if AC97 is ready before writing was added to its code.

Quoted 5 lines

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Quoted 2 lines

Best regards
Marcel

Thanks,
Mauro

Sorry about the broken formatting. Here's the diff once more:

drivers/media/usb/stk1160/stk1160-ac97.c +4 -0

+#include <linux/delay.h>
 
 #include "stk1160.h"
 #include "stk1160-reg.h"
 
+static int stk1160_ac97_wait_transfer_complete(struct stk1160 *dev)
+{
+       unsigned long timeout = jiffies + msecs_to_jiffies(STK1160_AC97_TIMEOUT);

drivers/media/usb/stk1160/stk1160-ac97.c +16 -8

+       }
+
+       stk1160_err("AC97 transfer took too long, this should never happen!");
+       return -EBUSY;
+}
+
 static void stk1160_write_ac97(struct stk1160 *dev, u16 reg, u16 value)
 {
        /* Set codec register address */

@@ -37,11 +56,11 @@ static void stk1160_write_ac97(struct stk1160 *dev, u16 reg, u16 value)
        stk1160_write_reg(dev, STK1160_AC97_CMD, value & 0xff);
        stk1160_write_reg(dev, STK1160_AC97_CMD + 1, (value & 0xff00) >> 8);
 
-       /*
-        * Set command write bit to initiate write operation.
-        * The bit will be cleared when transfer is done.
-        */
+       /* Set command write bit to initiate write operation */
        stk1160_write_reg(dev, STK1160_AC97CTL_0, 0x8c);
+
+       /* Wait for command write bit to be cleared */
+       stk1160_ac97_wait_transfer_complete(dev);
 }
 
 #ifdef DEBUG

@@ -53,12 +72,14 @@ static u16 stk1160_read_ac97(struct stk1160 *dev, u16 reg)
        /* Set codec register address */
        stk1160_write_reg(dev, STK1160_AC97_ADDR, reg);
 
-       /*
-        * Set command read bit to initiate read operation.
-        * The bit will be cleared when transfer is done.
-        */
+       /* Set command read bit to initiate read operation */
        stk1160_write_reg(dev, STK1160_AC97CTL_0, 0x8b);
 
+       /* Wait for command read bit to be cleared */
+       if (stk1160_ac97_wait_transfer_complete(dev) < 0) {
+               return 0;
+       }
+
        /* Retrieve register value */
        stk1160_read_reg(dev, STK1160_AC97_CMD, &vall);
        stk1160_read_reg(dev, STK1160_AC97_CMD + 1, &valh);

drivers/media/usb/stk1160/stk1160-reg.h +2 -0

index 296a9e7..7b08a3c 100644
@@ -122,6 +122,8 @@
 /* AC97 Audio Control */
 #define STK1160_AC97CTL_0              0x500
 #define STK1160_AC97CTL_1              0x504
+#define  STK1160_AC97CTL_0_CR          BIT(1)
+#define  STK1160_AC97CTL_0_CW          BIT(2)
 
 /* Use [0:6] bits of register 0x504 to set codec command address */
 #define STK1160_AC97_ADDR              0x504

drivers/media/usb/stk1160/stk1160.h +2 -0

index e85e12e..acd1c81 100644
@@ -50,6 +50,8 @@
 #define STK1160_MAX_INPUT 4
 #define STK1160_SVIDEO_INPUT 4
 
+#define STK1160_AC97_TIMEOUT 50
+
 #define STK1160_I2C_TIMEOUT 100
 
 

Quoted 138 lines · Mauro Carvalho Chehab

Em Mon, 5 Dec 2016 22:06:59 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 113 lines

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

Quoted 6 lines

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

Sounds ok. My experience with AC97 on em28xx is that, as new devices were added, the delay needed for AC97 varied on some of those new devices. That's why checking if AC97 is ready before writing was added to its code.

Quoted 7 lines

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Best regards
Marcel

Quoted 2 lines

Thanks,
Mauro

Marcel

Em Sun, 11 Dec 2016 13:20:06 +0100
[email protected] escreveu:

Quoted 1 line

Sorry about the broken formatting. Here's the diff once more:

The patch itself looks ok. Just a few comments.

Quoted 9 lines

drivers/media/usb/stk1160/stk1160-ac97.c +0 -0

index 95648ac..708792b 100644
@@ -23,11 +23,30 @@
  *
  */
 

drivers/media/usb/stk1160/stk1160-ac97.c line 26

Quoted 1 line

drivers/media/usb/stk1160/stk1160-ac97.c +0 -1

-#include <linux/module.h>

This change seems to be unrelated.

Quoted 14 lines

drivers/media/usb/stk1160/stk1160-ac97.c +10 -0

+#include <linux/delay.h>
 
 #include "stk1160.h"
 #include "stk1160-reg.h"
 
+static int stk1160_ac97_wait_transfer_complete(struct stk1160 *dev)
+{
+       unsigned long timeout = jiffies + msecs_to_jiffies(STK1160_AC97_TIMEOUT);
+       u8 value;
+
+       /* Wait for AC97 transfer to complete */
+       while (time_is_after_jiffies(timeout)) {
+               stk1160_read_reg(dev, STK1160_AC97CTL_0, &value);
+

drivers/media/usb/stk1160/stk1160-ac97.c line 43

Quoted 4 lines

drivers/media/usb/stk1160/stk1160-ac97.c +4 -0

+               if (!(value & (STK1160_AC97CTL_0_CR | STK1160_AC97CTL_0_CW)))
+                       return 0;
+
+               msleep(1);

It will likely sleep ~10ms. Maybe you likely need to use usleep_range(). Please read:
Documentation/timers/timers-howto.txt

Quoted 213 lines

drivers/media/usb/stk1160/stk1160-ac97.c +16 -8

+       }
+
+       stk1160_err("AC97 transfer took too long, this should never happen!");
+       return -EBUSY;
+}
+
 static void stk1160_write_ac97(struct stk1160 *dev, u16 reg, u16 value)
 {
        /* Set codec register address */

@@ -37,11 +56,11 @@ static void stk1160_write_ac97(struct stk1160 *dev, u16 reg, u16 value)
        stk1160_write_reg(dev, STK1160_AC97_CMD, value & 0xff);
        stk1160_write_reg(dev, STK1160_AC97_CMD + 1, (value & 0xff00) >> 8);
 
-       /*
-        * Set command write bit to initiate write operation.
-        * The bit will be cleared when transfer is done.
-        */
+       /* Set command write bit to initiate write operation */
        stk1160_write_reg(dev, STK1160_AC97CTL_0, 0x8c);
+
+       /* Wait for command write bit to be cleared */
+       stk1160_ac97_wait_transfer_complete(dev);
 }
 
 #ifdef DEBUG

@@ -53,12 +72,14 @@ static u16 stk1160_read_ac97(struct stk1160 *dev, u16 reg)
        /* Set codec register address */
        stk1160_write_reg(dev, STK1160_AC97_ADDR, reg);
 
-       /*
-        * Set command read bit to initiate read operation.
-        * The bit will be cleared when transfer is done.
-        */
+       /* Set command read bit to initiate read operation */
        stk1160_write_reg(dev, STK1160_AC97CTL_0, 0x8b);
 
+       /* Wait for command read bit to be cleared */
+       if (stk1160_ac97_wait_transfer_complete(dev) < 0) {
+               return 0;
+       }
+
        /* Retrieve register value */
        stk1160_read_reg(dev, STK1160_AC97_CMD, &vall);
        stk1160_read_reg(dev, STK1160_AC97_CMD + 1, &valh);

drivers/media/usb/stk1160/stk1160-reg.h +2 -0

index 296a9e7..7b08a3c 100644
@@ -122,6 +122,8 @@
 /* AC97 Audio Control */
 #define STK1160_AC97CTL_0              0x500
 #define STK1160_AC97CTL_1              0x504
+#define  STK1160_AC97CTL_0_CR          BIT(1)
+#define  STK1160_AC97CTL_0_CW          BIT(2)
 
 /* Use [0:6] bits of register 0x504 to set codec command address */
 #define STK1160_AC97_ADDR              0x504

drivers/media/usb/stk1160/stk1160.h +2 -0

index e85e12e..acd1c81 100644
@@ -50,6 +50,8 @@
 #define STK1160_MAX_INPUT 4
 #define STK1160_SVIDEO_INPUT 4
 
+#define STK1160_AC97_TIMEOUT 50
+
 #define STK1160_I2C_TIMEOUT 100
 
 

Quoted 140 lines · Mauro Carvalho Chehab

Em Mon, 5 Dec 2016 22:06:59 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 113 lines

Hello

2016-12-05 16:38 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 93 lines

Quoted 82 lines · Mauro Carvalho Chehab

Em Sun, 4 Dec 2016 15:25:25 -0300
Ezequiel Garcia <[email protected]> escreveu:

Quoted 79 lines · Marcel Hasler

Quoted 49 lines

Hello

2016-12-03 21:46 GMT+01:00 Ezequiel Garcia <[email protected]>:

Quoted 29 lines

Quoted 10 lines · Mauro Carvalho Chehab

Em Sun, 27 Nov 2016 12:11:48 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 3 lines

Allow setting a custom record gain for the internal AC97 codec (if available). This can be a value between 0 and 15, 8 is the default and should be suitable for most users. The Windows driver also sets this to 8 without any possibility for changing it.

The problem of removing the mixer is that you need this kind of crap to setup the volumes on a non-standard way.

Right, that's a good point.

Quoted 9 lines

NACK.

Instead, keep the alsa mixer. The way other drivers do (for example, em28xx) is that they configure the mixer when an input is selected, increasing the volume of the active audio channel to 100% and muting the other audio channels. Yet, as the alsa mixer is exported, users can change the mixer settings in runtime using some alsa (or pa) mixer application.

Yeah, the AC97 mixer we are currently leveraging exposes many controls that have no meaning in this device, so removing that still looks like an improvement.

Quoted 8 lines

I guess the proper way is creating our own mixer (not using snd_ac97_mixer) exposing only the record gain knob.

Marcel, what do you think?
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

As I have written before, the recording gain isn't actually meant to be changed by the user. In the official Windows driver this value is hard-coded to 8 and cannot be changed in any way. And there really is no good reason why anyone should need to mess with it in the first place. The default value will give the best results in pretty much all cases and produces approximately the same volume as the internal 8-bit ADC whose gain cannot be changed at all, not even by a driver.

Quoted 11 lines

I had considered writing a mixer but chose not to. If the gain setting is openly exposed to mixer applications, how do you tell the users that the value set by the driver already is the optimal and recommended value and that they shouldn't mess with the controls unless they really have to? By having a module parameter, this setting is practically hidden from the normal user but still is available to power-users if they think they really need it. In the end it's really just a compromise between hiding it completely and exposing it openly. Also, this way the driver guarantees reproducible results, since there's no need to remember the positions of any volume sliders.

Hm, right. I've never changed the record gain, and it's true that it doens't really improve the volume. So, I would be OK with having a module parameter.

On the other side, we are exposing it currently, through the "Capture" mixer control:

Simple mixer control 'Capture',0
Capabilities: cvolume cswitch cswitch-joined
Capture channels: Front Left - Front Right
Limits: Capture 0 - 15
Front Left: Capture 10 [67%] [15.00dB] [on]
Front Right: Capture 8 [53%] [12.00dB] [on]

So, it would be user-friendly to keep the user interface and continue to expose the same knob - even if the default is the optimal, etc.

Quoted 8 lines

Quoted 3 lines

To be completely honest, I don't think any user is really relying on any REC_GAIN / Capture setting, and I'm completely OK with having a mixer control or a module parameter. It doesn't matter.

If you're positive that *all* stk1160 use the ac97 mixer the same way, and that there's no sense on having a mixer for it, then it would be ok to remove it.

Quoted 8 lines

Let's remove it then!

Quoted 4 lines

In such case, then why you need a modprobe parameter to allow setting the record level? If this mixer entry is not used, just set it to zero and be happy with that.

Let's remove the module param too, then.

I'm okay with that.

Quoted 5 lines

Thanks,
--
Ezequiel García, VanguardiaSur
www.vanguardiasur.com.ar

Quoted 6 lines

I'm willing to prepare one final patchset, provided we can agree on and resolve all issues beforehand.

So far the changes would be to remove the module param and to poll STK1160_AC97CTL_0 instead of using a fixed delay. It's probably better to also poll it before writing, although that never caused problems.

Sounds ok. My experience with AC97 on em28xx is that, as new devices were added, the delay needed for AC97 varied on some of those new devices. That's why checking if AC97 is ready before writing was added to its code.

Quoted 7 lines

I'll post some code for review before actually submitting patches. Mauro, is there anything else that you think should be changed? If so, please tell me now. Thanks.

Best regards
Marcel

Thanks,
Mauro

Quoted 1 line

Marcel

Thanks,
Mauro

The STK1160 needs some time to transfer data from the AC97 registers into its own. On some systems reading the chip's own registers to soon will return wrong values. The "proper" way to handle this would be to poll STK1160_AC97CTL_0 after every read or write command until the command bit has been cleared, but this may not be worth the hassle.

Signed-off-by Marcel Hasler <[email protected]>

---
 drivers/media/usb/stk1160/stk1160-ac97.c | 9 +++++++++
 1 file changed, 9 insertions(+)

drivers/media/usb/stk1160/stk1160-ac97.c +1 -0

index 60327af..b39f51b 100644
@@ -23,6 +23,7 @@
  *
  */
 
+#include <linux/delay.h>
 #include <linux/module.h>
 
 #include "stk1160.h"

drivers/media/usb/stk1160/stk1160-ac97.c +3 -0

+	 */
+	usleep_range(20, 40);
+
 
 	/* Retrieve register value */
 	stk1160_read_reg(dev, STK1160_AC97_CMD, &vall);

stk1160_read_reg(dev, STK1160_AC97_CMD + 1, &valh);

Em Sun, 27 Nov 2016 12:12:36 +0100
Marcel Hasler <[email protected]> escreveu:

Quoted 23 lines

The STK1160 needs some time to transfer data from the AC97 registers into its own. On some systems reading the chip's own registers to soon will return wrong values. The "proper" way to handle this would be to poll STK1160_AC97CTL_0 after every read or write command until the command bit has been cleared, but this may not be worth the hassle.

Signed-off-by Marcel Hasler <[email protected]>

---
 drivers/media/usb/stk1160/stk1160-ac97.c | 9 +++++++++
 1 file changed, 9 insertions(+)

drivers/media/usb/stk1160/stk1160-ac97.c +1 -0

index 60327af..b39f51b 100644
@@ -23,6 +23,7 @@
  *
  */
 
+#include <linux/delay.h>
 #include <linux/module.h>
 
 #include "stk1160.h"

drivers/media/usb/stk1160/stk1160-ac97.c static u16 stk1160_read_ac97(struct stk1160 *dev, u16 reg) · line 72

Quoted 8 lines

drivers/media/usb/stk1160/stk1160-ac97.c +5 -0

@@ -64,6 +65,14 @@ static u16 stk1160_read_ac97(struct stk1160 *dev, u16 reg)
       */
      stk1160_write_reg(dev, STK1160_AC97CTL_0, 0x8b);
 
+     /*
+      * Give the chip some time to transfer the data.
+      * The proper way would be to poll STK1160_AC97CTL_0
+      * until the command bit has been cleared, but this
+      * may not be worth the hassle.

Why not? Relying on a fixed amount time is not nice.

Take a look at em28xx_is_ac97_ready() function, at drivers/media/usb/em28xx/em28xx-core.c to see how this could be implemented instead.

Quoted 8 lines

drivers/media/usb/stk1160/stk1160-ac97.c +3 -0

+      */
+     usleep_range(20, 40);
+
 
      /* Retrieve register value */
      stk1160_read_reg(dev, STK1160_AC97_CMD, &vall);

      stk1160_read_reg(dev, STK1160_AC97_CMD + 1, &valh);

Thanks,
Mauro

This patchset is a result of my attempt to fix a bug (https://bugzilla.kernel.org/show_bug.cgi?id=180071) that eventually turned out to be caused by a missing quirk in snd-usb-audio. My idea was to remove the AC97 interface and setup the codec using the same values and in the same order as the Windows driver does, hoping there might be some "magic" sequence that would make the sound work the way it should. Although this didn't help to fix the problem, I found these changes to be useful nevertheless.

IMHO, having all of the AC97 codec's channels exposed to userspace is confusing since most of them have no meaning for this device anyway. Changing these values in alsamixer has either no effect at all or may even reduce the sound quality since it can actually increase the line-in DC offset (slightly).

In addition, having to re-select the correct capture channel everytime the device has been plugged in is annoying. At least on my systems the mixer setup is only saved if the device is plugged in during shutdown/reboot. I also get error messages in my kernel log when I unplug the device because some process (probably the AC97 driver) ist trying to read from the device after it has been removed. Either way the device should work out-of-the-box without the need for the user to manually setup channels.

The first patch in the set therefore removes the 'stk1160-mixer' and lets the driver setup the AC97 codec using the same values as the Windows driver. Although some of the values seem to be defaults I let the driver set them either way, just to be sure.

The second patch adds a check to determine whether the device is strapped to use the internal 8-bit ADC or an external chip. There's currently no check in place to determine whether the device uses AC-link or I2S, but then again I haven't heard of any of these devices actually using an I2S chip. If the device uses the internal ADC the AC97 setup can be skipped. I implemented the check inside stk1160-ac97. It could just as well be in stk1160-core but this way just seemed cleaner. If at some point the need arises to check other power-on strap values, it might make sense to refactor this then.

The third patch adds a new module parameter for setting the record gain manually since the AC97 chip is no longer exposed to userspace. The Windows driver doesn't allow this value to be changed but instead always sets it to 8 (of 15). While this should be fine for most users, some may prefer something higher.

The fourth patch addresses an issue when reading from the AC97 chip too soon, resulting in corrupt data.

Changes from version 2:
* Added copyright notice
* Added defines for POSVA bytes and bits
* Added check for ACDOUT bit to determine whether audio is disabled completely
* Removed info output for gain setting
* Added fourth patch which had been submitted independently before
* Expanded comment on AC97 read delay

Marcel Hasler (4):
stk1160: Remove stk1160-mixer and setup internal AC97 codec automatically.
stk1160: Check whether to use AC97 codec.
stk1160: Add module param for setting the record gain.
stk1160: Give the chip some time to retrieve data from AC97 codec.