[PATCH] [ALSA] mpc8610: Add mmap support

[PATCH] [ALSA] mpc8610: Add mmap support

[Timur Tabi]

Enable mmap support in the MPC8610 ASoC driver.  The driver can use ALSA's
default mmap functionality, it was just not enabled previously.

Signed-off-by: Timur Tabi <timur@freescale.com>
---
 sound/soc/fsl/fsl_dma.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/sound/soc/fsl/fsl_dma.c b/sound/soc/fsl/fsl_dma.c
index 2173203..652514f 100644
--- a/sound/soc/fsl/fsl_dma.c
+++ b/sound/soc/fsl/fsl_dma.c
@@ -136,7 +136,9 @@ struct fsl_dma_private {
  */
 static const struct snd_pcm_hardware fsl_dma_hardware = {
 
-	.info   		= SNDRV_PCM_INFO_INTERLEAVED,
+	.info   		= SNDRV_PCM_INFO_INTERLEAVED |
+				  SNDRV_PCM_INFO_MMAP |
+				  SNDRV_PCM_INFO_MMAP_VALID,
 	.formats		= FSLDMA_PCM_FORMATS,
 	.rates  		= FSLDMA_PCM_RATES,
 	.rate_min       	= 5512,
-- 
1.5.2.4

Re: [PATCH] [ALSA] mpc8610: Add mmap support

[Alexander E. Patrakov]

Timur Tabi wrote:
> Enable mmap support in the MPC8610 ASoC driver.  The driver can use ALSA's
> default mmap functionality, it was just not enabled previously.
> 
> Signed-off-by: Timur Tabi <timur@freescale.com>

[sorry, I may be completely wrong below about the .page callback purpose - in 
this case, just confirm that the testcase works]

Hm, the "struct snd_pcm_ops fsl_dma_ops" doesn't contain the .page callback. 
With saa7134-alsa, this led to the "device claims to support mmap but actually 
doesn't work" bug, see the following subthread:

http://mailman.alsa-project.org/pipermail/alsa-devel/2007-October/003773.html

IOW: have you actually tested that the following commands work (X,Y refer to the 
FSL chip)?

aplay -M -D hw:X,Y sample.wav
arecord -f cd -M -D:X,Y capture.wav

> ---
>  sound/soc/fsl/fsl_dma.c |    4 +++-
>  1 files changed, 3 insertions(+), 1 deletions(-)
> 
> diff --git a/sound/soc/fsl/fsl_dma.c b/sound/soc/fsl/fsl_dma.c
> index 2173203..652514f 100644
> --- a/sound/soc/fsl/fsl_dma.c
> +++ b/sound/soc/fsl/fsl_dma.c
> @@ -136,7 +136,9 @@ struct fsl_dma_private {
>   */
>  static const struct snd_pcm_hardware fsl_dma_hardware = {
>  
> -	.info   		= SNDRV_PCM_INFO_INTERLEAVED,
> +	.info   		= SNDRV_PCM_INFO_INTERLEAVED |
> +				  SNDRV_PCM_INFO_MMAP |
> +				  SNDRV_PCM_INFO_MMAP_VALID,
>  	.formats		= FSLDMA_PCM_FORMATS,
>  	.rates  		= FSLDMA_PCM_RATES,
>  	.rate_min       	= 5512,

-- 
Alexander E. Patrakov

Re: [PATCH] [ALSA] mpc8610: Add mmap support

[Timur Tabi]

Alexander E. Patrakov wrote:
> Timur Tabi wrote:
>> Enable mmap support in the MPC8610 ASoC driver.  The driver can use 
>> ALSA's
>> default mmap functionality, it was just not enabled previously.
>>
>> Signed-off-by: Timur Tabi <timur@freescale.com>
> 
> [sorry, I may be completely wrong below about the .page callback purpose 
> - in this case, just confirm that the testcase works]
> 
> Hm, the "struct snd_pcm_ops fsl_dma_ops" doesn't contain the .page 
> callback. With saa7134-alsa, this led to the "device claims to support 
> mmap but actually doesn't work" bug, see the following subthread:

Isn't that an ARM chip?  The default MMAP handler in ALSA doesn't support ARM, 
but it works for PowerPC.

> http://mailman.alsa-project.org/pipermail/alsa-devel/2007-October/003773.html 
> 
> 
> IOW: have you actually tested that the following commands work (X,Y 
> refer to the FSL chip)?
> 
> aplay -M -D hw:X,Y sample.wav
> arecord -f cd -M -D:X,Y capture.wav

I haven't tried record, but I did try playback, and it worked.  Without this 
patch, adding -M gives me this error:

aplay: set_params:852: Access type not available

After applying this patch, playback works fine.

-- 
Timur Tabi
Linux kernel developer at Freescale

Re: [PATCH] [ALSA] mpc8610: Add mmap support

[Takashi Iwai]

At Thu, 17 Jan 2008 10:31:37 -0600,
Timur Tabi wrote:
> 
> Alexander E. Patrakov wrote:
> > Timur Tabi wrote:
> >> Enable mmap support in the MPC8610 ASoC driver.  The driver can use 
> >> ALSA's
> >> default mmap functionality, it was just not enabled previously.
> >>
> >> Signed-off-by: Timur Tabi <timur@freescale.com>
> > 
> > [sorry, I may be completely wrong below about the .page callback purpose 
> > - in this case, just confirm that the testcase works]
> > 
> > Hm, the "struct snd_pcm_ops fsl_dma_ops" doesn't contain the .page 
> > callback. With saa7134-alsa, this led to the "device claims to support 
> > mmap but actually doesn't work" bug, see the following subthread:
> 
> Isn't that an ARM chip?  The default MMAP handler in ALSA doesn't support ARM, 
> but it works for PowerPC.
> 
> > http://mailman.alsa-project.org/pipermail/alsa-devel/2007-October/003773.html 
> > 
> > 
> > IOW: have you actually tested that the following commands work (X,Y 
> > refer to the FSL chip)?
> > 
> > aplay -M -D hw:X,Y sample.wav
> > arecord -f cd -M -D:X,Y capture.wav
> 
> I haven't tried record, but I did try playback, and it worked.  Without this 
> patch, adding -M gives me this error:
> 
> aplay: set_params:852: Access type not available
> 
> After applying this patch, playback works fine.

Great, applied to ALSA tree now.

Thanks.

Takashi

Re: [PATCH] [ALSA] mpc8610: Add mmap support

[Alexander E. Patrakov]

Timur Tabi wrote:
> Alexander E. Patrakov wrote:
>> Hm, the "struct snd_pcm_ops fsl_dma_ops" doesn't contain the .page 
>> callback. With saa7134-alsa, this led to the "device claims to support 
>> mmap but actually doesn't work" bug, see the following subthread:
> 
> Isn't that an ARM chip?  The default MMAP handler in ALSA doesn't support ARM, 
> but it works for PowerPC.

saa7134 is a Philips chip that you'll find in many TV tuners.

As for the .page logic, my question was indeed stupid. I should have read the 
snd_pcm_mmap_data_nopage() function and notice that, by default, it wants to do 
this:

vaddr = runtime->dma_area + offset;
page = virt_to_page(vaddr);

and saa7134 wants to use vmalloc_to_page() instead of virt_to_page() due to the 
way it allocates the DMA buffer.


> I haven't tried record, but I did try playback, and it worked.  Without this 
> patch, adding -M gives me this error:
> 
> aplay: set_params:852: Access type not available
> 
> After applying this patch, playback works fine.

Good!

-- 
Alexander E. Patrakov

Re: [PATCH] [ALSA] mpc8610: Add mmap support

[Trent Piepho]

On Thu, 17 Jan 2008, Timur Tabi wrote:
> Alexander E. Patrakov wrote:
> > Timur Tabi wrote:
> >> Enable mmap support in the MPC8610 ASoC driver.  The driver can use
> >> ALSA's
> >> default mmap functionality, it was just not enabled previously.
> >>
> >> Signed-off-by: Timur Tabi <timur@freescale.com>
> >
> > [sorry, I may be completely wrong below about the .page callback purpose
> > - in this case, just confirm that the testcase works]
> >
> > Hm, the "struct snd_pcm_ops fsl_dma_ops" doesn't contain the .page
> > callback. With saa7134-alsa, this led to the "device claims to support
> > mmap but actually doesn't work" bug, see the following subthread:
>
> Isn't that an ARM chip?  The default MMAP handler in ALSA doesn't support ARM,
> but it works for PowerPC.

Alexander is talking about the same thing I was in my previous message to
you.  saa7134, like cx88-alsa, uses vmalloc instead of ALSA to allocate the
DMA buffer, and so a special page callback is needed.

> I haven't tried record, but I did try playback, and it worked.  Without this
> patch, adding -M gives me this error:
>
> aplay: set_params:852: Access type not available
>
> After applying this patch, playback works fine.

The key thing to know is that you will not get an error message if your
page callback is wrong!  Everything will appear to work.  You just won't
hear any sound.

Re: [PATCH] [ALSA] mpc8610: Add mmap support

[Timur Tabi]

Trent Piepho wrote:

> The key thing to know is that you will not get an error message if your
> page callback is wrong!  Everything will appear to work.  You just won't
> hear any sound.

When I said playback works fine, I was wearing my headphones. :-)

-- 
Timur Tabi
Linux kernel developer at Freescale