Re: [PATCH] ASoC: adau1373: adau1373_hw_params: Silence overflow warning

[Lars-Peter Clausen <lars@metafoo.de>]

On 03/11/2013 04:37 PM, Tim Gardner wrote:
> On 03/11/2013 04:01 AM, Lars-Peter Clausen wrote:
>> On 03/10/2013 06:34 PM, Tim Gardner wrote:
>>> ADAU1373_BCLKDIV_SOURCE is defined as BIT(5) which uses UL constants. On
>>> amd64 the result of the ones complement operator is then truncated to
>>> unsigned int according to the prototype of snd_soc_update_bits(). I think
>>> gcc is correctly warning that the upper 32 bits are lost.
>>>
>>> sound/soc/codecs/adau1373.c: In function 'adau1373_hw_params':
>>> sound/soc/codecs/adau1373.c:940:3: warning: large integer implicitly truncated to unsigned type [-Woverflow]
>>>
>>> gcc version 4.6.3
>>>
>>> Cc: Lars-Peter Clausen <lars@metafoo.de>
>>> Cc: Liam Girdwood <lgirdwood@gmail.com>
>>> Cc: Mark Brown <broonie@opensource.wolfsonmicro.com>
>>> Cc: Jaroslav Kysela <perex@perex.cz>
>>> Cc: Takashi Iwai <tiwai@suse.de>
>>> Cc: device-drivers-devel@blackfin.uclinux.org
>>> Cc: alsa-devel@alsa-project.org
>>> Signed-off-by: Tim Gardner <tim.gardner@canonical.com>
>>> ---
>>>  sound/soc/codecs/adau1373.c |    2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/sound/soc/codecs/adau1373.c b/sound/soc/codecs/adau1373.c
>>> index 068b3ae..56ed788 100644
>>> --- a/sound/soc/codecs/adau1373.c
>>> +++ b/sound/soc/codecs/adau1373.c
>>> @@ -132,7 +132,7 @@ struct adau1373 {
>>>  #define ADAU1373_DAI_FORMAT_I2S		0x2
>>>  #define ADAU1373_DAI_FORMAT_DSP		0x3
>>>  
>>> -#define ADAU1373_BCLKDIV_SOURCE		BIT(5)
>>> +#define ADAU1373_BCLKDIV_SOURCE		(unsigned int)BIT(5)
>>
>> Hm, that's a bit ugly. How about the following instead:
>>
>> diff --git a/sound/soc/codecs/adau1373.c b/sound/soc/codecs/adau1373.c
>> index 068b3ae..1aa10dd 100644
>> --- a/sound/soc/codecs/adau1373.c
>> +++ b/sound/soc/codecs/adau1373.c
>> @@ -133,6 +133,8 @@ struct adau1373 {
>>  #define ADAU1373_DAI_FORMAT_DSP		0x3
>>
>>  #define ADAU1373_BCLKDIV_SOURCE		BIT(5)
>> +#define ADAU1373_BCLKDIV_SR_MASK	(0x07 << 2)
>> +#define ADAU1373_BCLKDIV_BCLK_MASK	0x03
>>  #define ADAU1373_BCLKDIV_32		0x03
>>  #define ADAU1373_BCLKDIV_64		0x02
>>  #define ADAU1373_BCLKDIV_128		0x01
>> @@ -937,7 +939,8 @@ static int adau1373_hw_params(struct snd_pcm_substream
>> *substream,
>>  	adau1373_dai->enable_src = (div != 0);
>>
>>  	snd_soc_update_bits(codec, ADAU1373_BCLKDIV(dai->id),
>> -		~ADAU1373_BCLKDIV_SOURCE, (div << 2) | ADAU1373_BCLKDIV_64);
>> +		ADAU1373_BCLKDIV_SR_MASK | ADAU1373_BCLKDIV_BCLK_MASK,
>> +		(div << 2) | ADAU1373_BCLKDIV_64);
>>
>>  	switch (params_format(params)) {
>>  	case SNDRV_PCM_FORMAT_S16_LE:
>>
>>
> 
> That seems way more complicated then need be. It also uses 2 bits in the
> mask whereas the original code only used 1, e.g., bit 5. Is that correct ?

The BCLKDIV has three fields. The bitclock divider (bit 0-1), the samplerate
(bit 2-4) and the source select (bit 5). Here we want to update the bitclock
divider field and the samplerate field. When I wrote the code I was lazy and
used ~ADAU1373_BCLKDIV_SOURCE as the mask, which for this register is
functional equivalent to ADAU1373_BCLKDIV_SR_MASK | ADAU1373_BCLKDIV_BCLK_MASK.

> 
> How about just open coding the macro thusly:
> 
> #define ADAU1373_BCLKDIV_SOURCE (1<<5) /*BIT(5)*/

But now there is a inconsistency where we sometimes use (1<<x) and sometimes
BIT(x). I prefer my solution.

- Lars