Failed to set cache compression type

Failed to set cache compression type

[Fabio Estevam]

Hi,

I am booting linux-next 20140602 on a mx6qsabresd and I get the following probe
error:

wm8962 0-001a: customer id 0 revision D
wm8962 0-001a: ASoC: Failed to set cache compression type: -22
imx-wm8962 sound: snd_soc_register_card failed (-22)
imx-wm8962: probe of sound failed with error -22

What is the proper way to fix it?

Thanks,

Fabio Estevam

Re: Failed to set cache compression type

[Jarkko Nikula]

On 06/02/2014 04:43 PM, Fabio Estevam wrote:
> Hi,
>
> I am booting linux-next 20140602 on a mx6qsabresd and I get the following probe
> error:
>
> wm8962 0-001a: customer id 0 revision D
> wm8962 0-001a: ASoC: Failed to set cache compression type: -22
> imx-wm8962 sound: snd_soc_register_card failed (-22)
> imx-wm8962: probe of sound failed with error -22
>
> What is the proper way to fix it?
>
I was looking at this too and it was introduced by the commit 
b59dce53ef61 ("ASoC: cache: Fix possible ZERO_SIZE_PTR pointer 
dereferencing error.").

If I understood commit 657254714ad2 ("ASoC: io: Remove support for ASoC 
cache in conjunction with regmap") right should we remove remaining 
soc-cache use? All codec drivers seems to include regmap.h although some 
of them are still calling soc_cache_sync().

-- 
Jarkko

Re: Failed to set cache compression type

[Lars-Peter Clausen]

On 06/02/2014 04:02 PM, Jarkko Nikula wrote:
> On 06/02/2014 04:43 PM, Fabio Estevam wrote:
>> Hi,
>>
>> I am booting linux-next 20140602 on a mx6qsabresd and I get the following
>> probe
>> error:
>>
>> wm8962 0-001a: customer id 0 revision D
>> wm8962 0-001a: ASoC: Failed to set cache compression type: -22
>> imx-wm8962 sound: snd_soc_register_card failed (-22)
>> imx-wm8962: probe of sound failed with error -22
>>
>> What is the proper way to fix it?
>>
> I was looking at this too and it was introduced by the commit b59dce53ef61
> ("ASoC: cache: Fix possible ZERO_SIZE_PTR pointer dereferencing error.").

I think snd_soc_cache_init() should return 0 rather than -EINVAL if reg_size 
is 0.

>
> If I understood commit 657254714ad2 ("ASoC: io: Remove support for ASoC
> cache in conjunction with regmap") right should we remove remaining
> soc-cache use? All codec drivers seems to include regmap.h although some of
> them are still calling soc_cache_sync().

Yes, ASoC level support for caching is going to be removed, hopefully sooner 
than later. If a driver still uses soc_cache_sync() even though it has 
switched to regmap for IO that's a bug.

- Lars

Re: Failed to set cache compression type

[Fabio Estevam]

On Mon, Jun 2, 2014 at 11:10 AM, Lars-Peter Clausen <lars@metafoo.de> wrote:

> I think snd_soc_cache_init() should return 0 rather than -EINVAL if reg_size
> is 0.

Yes, this fixes the issue:

--- a/sound/soc/soc-cache.c
+++ b/sound/soc/soc-cache.c
@@ -73,7 +73,7 @@ int snd_soc_cache_init(struct snd_soc_codec *codec)
        reg_size = codec_drv->reg_cache_size * codec_drv->reg_word_size;

        if (!reg_size)
-               return -EINVAL;
+               return 0;

        mutex_init(&codec->cache_rw_mutex);

Re: Failed to set cache compression type

[Mark Brown]

[-- Attachment #1.1: Type: text/plain, Size: 403 bytes --]

On Mon, Jun 02, 2014 at 05:02:50PM +0300, Jarkko Nikula wrote:

> If I understood commit 657254714ad2 ("ASoC: io: Remove support for ASoC
> cache in conjunction with regmap") right should we remove remaining
> soc-cache use? All codec drivers seems to include regmap.h although some of
> them are still calling soc_cache_sync().

Anything using both regmap and the ASoC cache is definitely broken, yes.

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: Failed to set cache compression type

[Mark Brown]

[-- Attachment #1.1: Type: text/plain, Size: 766 bytes --]

On Mon, Jun 02, 2014 at 05:02:50PM +0300, Jarkko Nikula wrote:

> I was looking at this too and it was introduced by the commit b59dce53ef61
> ("ASoC: cache: Fix possible ZERO_SIZE_PTR pointer dereferencing error.").

I've just fixed that now, the error checking isn't right - it should be
returning 0 not -EINVAL.

> If I understood commit 657254714ad2 ("ASoC: io: Remove support for ASoC
> cache in conjunction with regmap") right should we remove remaining
> soc-cache use? All codec drivers seems to include regmap.h although some of
> them are still calling soc_cache_sync().

There's still the AC'97 stuff at least, plus the ab8500 though if that
ends up blocking anything I'm just inclined to reapply the conversion
patch and let someone fix it if they care.

[-- Attachment #1.2: Digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]