[PATCH] NUC900/audio: add nuc900 audio driver support

Wan ZongShun mcuos.com at gmail.com
Mon May 17 02:26:12 EDT 2010


Dear Mark,

Thanks a lot for your help, and I have other questions needed your help below.

> On Sun, May 16, 2010 at 11:34:01PM +0800, Wan ZongShun wrote:
> 
> This all looks pretty good.  A few comments below but they're all fairly
> minor and should be easy to address.
> 
>> index b1749bc..81d4848 100644
>> --- a/sound/soc/Kconfig
>> +++ b/sound/soc/Kconfig
>> @@ -36,6 +36,7 @@ source "sound/soc/s3c24xx/Kconfig"
>>  source "sound/soc/s6000/Kconfig"
>>  source "sound/soc/sh/Kconfig"
>>  source "sound/soc/txx9/Kconfig"
>> +source "sound/soc/nuc900/Kconfig"
> 
> Please keep the Kconfig and Makefile sorted, this helps avoid merge
> issues.
> 
>> +#define NUC900_AC97_RATES (SNDRV_PCM_RATE_8000 | SNDRV_PCM_RATE_11025 |\
>> +			SNDRV_PCM_RATE_16000 | SNDRV_PCM_RATE_22050 |\
>> +				SNDRV_PCM_RATE_44100 | SNDRV_PCM_RATE_48000)
> 
> SNDRV_PCM_RATE_8000_48000.
> 
>> +	if (!timeout)
>> +		dev_err(nuc900_audio->dev, "AC97 read register time out !\n");
>> +
>> +	val = AUDIO_READ(nuc900_audio->mmio + ACTL_ACOS0) ;
> 
> If the read timed out shouldn't we be returning rather than continuing?
> 
>> +	val = AUDIO_READ(nuc900_audio->mmio + ACTL_ACCON);
>> +	val &= (~AC_C_RES);
>> +	AUDIO_WRITE(nuc900_audio->mmio + ACTL_ACCON, val);
> 
>> +	udelay(1000);
> 
>> +	if (!(AUDIO_READ(nuc900_audio->mmio + ACTL_ACIS0) & CODEC_READY))
>> +		dev_err(nuc900_audio->dev, "AC97 codec cold reset failed!\n");
> 
> What is this actually checking in the hardware?  Not all CODECs enable
> the AC97 link by default after a cold reset, the standard allows them to
> power up in a low power state which will

The bit 'CODEC_READY' indicates the external AC97 audio CODEC is ready,
I just forget to implement the warm reset operation, if need, I will do
it. In addition, Can I add the checking CODEC_READY operation to warm reset 
function?

> On the other hand, given that there's no warm reset operation perhaps
> this isn't a big deal.
> 
>> +static void nuc900_ac97_remove(struct platform_device *pdev,
>> +						struct snd_soc_dai *dai)
>> +{
>> +	struct nuc900_audio *nuc900_audio = nuc900_ac97_data;
>> +
>> +	/* Enable unit clock */
>> +	clk_disable(nuc900_audio->clk);
> 
> Bit rot in the comments here.
> 
>> +/* bit definition of REG_ACTL_CON register */
>> +#define	AUDCLK_EN		0x8000
>> +#define	PFIFO_EN		0x4000
>> +#define	RFIFO_EN		0x2000

I rechecked the latest datasheet, and I find above bits has been removed,
so I delete them here.

> These constants (and the rest) really should be namespaced - they're
> likely to collide with other definitions in, for example, CODEC drivers
> used by machines.
> 
>> +#define	IIS_EN			0x0002
>
> Looks like there's I2S support to come?
> 

Yes,nuc900 supports both IIS and AC97,but my board only enable AC97 support,
what's up here?

>> +static irqreturn_t nuc900_dma_interrupt(int irq, void *dev_id)
>> +{
> 
>> +
>> +	snd_pcm_period_elapsed(substream);
>> +
>> +	return IRQ_HANDLED;
> 
> This is done unconditionally - are you sure there can't be any spurious
> interrupts (eg, error interrupts).  It shouldn't cause any harm, of
> course.
> 

Wan .(Vincent)



More information about the linux-arm-kernel mailing list