At Mon, 8 Jan 2007 10:47:35 -0800,
Andrew Johnson wrote:
>
> Hi All,
>
> It seems that there is a bug in the PXA ALSA drivers which causes an
> earlier section of the circular audio buffer to be played erroneously at
> the end of playback when using aplay.
>
> The bug occurs because runtime->period_size is not an integer multiple
> of runtime->buffer_size. When a section of audio to be copied to the
> buffer is less than a period size, aplay uses runtime->period size to
> pad the section with zeros to ensure that the audio stream is an integer
> multiple of runtime->period_size. This is important as ALSA can only
> stop playback on the period boundary.
>
> Example:
>
> period_size = 32
> buffer_size = 90
> audio_size = 180
>
> aplay pads the audio with zeros to be an integer multiple of period_size
> so:
>
> audio_size(padded) = 192
>
> This will fill the buffer twice with 12-bytes left over. Therefore,
> the first period will have 12 bytes of silence and 20 bytes of data
> from earlier in the audio stream. The ALSA core will not stop until
> the entire buffer is empty, thus the 20-bytes of earlier data will be
> played before stopping. The solution is to add a constraint to force
> the
> runtime->buffer_size to be an integer multiple of runtime->period_size.
> I've appended a patch to do this.
>
> Kind Regards,
>
> Andrew
>
> PATCH FOLLOWS
>
> Signed-off-by: Andrew Johnson <ajohnson <at> intrinsyc.com>
Thanks for the patch. The fix looks logical to me.
However, the patch you embedded is broken (the line folded), and the
change to pxa2xx_pcm_dma_irq() is wrong for recent kernels, too.
(We may need a wrapper or a patch file for 2.6.18 or older...)
Could you fix these and repost the patch, together with a short
changelog to apply?
Takashi
>
> --- a/alsa-kernel/sound/arm/pxa2xx-pcm.c 2007-01-08
> 10:37:34.000000000 -0800
> +++ b/alsa-kernel/sound/arm/pxa2xx-pcm.c 2007-01-05
> 17:59:27.000000000 -0800
> @@ -137,7 +137,7 @@ static int pxa2xx_pcm_trigger(struct snd
> return ret;
> }
>
> -static void pxa2xx_pcm_dma_irq(int dma_ch, void *dev_id)
> +static void pxa2xx_pcm_dma_irq(int dma_ch, void *dev_id, struct pt_regs
> *regs)
> {
> struct snd_pcm_substream *substream = dev_id;
> struct pxa2xx_runtime_data *rtd =
> substream->runtime->private_data;
> @@ -213,6 +213,10 @@ static int pxa2xx_pcm_open(struct snd_pc
> if (ret)
> goto out;
>
> + ret = snd_pcm_hw_constraint_integer(runtime,
> SNDRV_PCM_HW_PARAM_PERIODS);
> + if (ret < 0)
> + goto out;
> +
> ret = -ENOMEM;
> rtd = kmalloc(sizeof(*rtd), GFP_KERNEL);
> if (!rtd)
>
> --- a/alsa-kernel/sound/soc/pxa/pxa2xx-pcm.c 2007-01-08
> 10:37:34.000000000 -0800
> +++ b/alsa-kernel/sound/soc/pxa/pxa2xx-pcm.c 2007-01-05
> 18:10:47.000000000 -0800
> @@ -227,6 +227,11 @@ static int pxa2xx_pcm_open(struct snd_pc
> if (ret)
> goto out;
>
> + ret = snd_pcm_hw_constraint_integer(runtime,
> + SNDRV_PCM_HW_PARAM_PERIODS);
> + if (ret < 0)
> + goto out;
> +
> prtd = kzalloc(sizeof(struct pxa2xx_runtime_data), GFP_KERNEL);
> if (prtd == NULL) {
> ret = -ENOMEM;
>
> -------------------------------------------------------------------------
> Take Surveys. Earn Cash. Influence the Future of IT
> Join SourceForge.net's Techsay panel and you'll get the chance to share your
> opinions on IT & business topics through brief surveys - and earn cash
> http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel@xxxxxxxxxxxxxxxxxxxxx
> https://lists.sourceforge.net/lists/listinfo/alsa-devel
>
-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Alsa-devel mailing list
Alsa-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.sourceforge.net/lists/listinfo/alsa-devel
