Re: [PATCH 0/3] ASoC: Move pcm writecombine dma buffer allocation to core

[Laxman Dewangan <ldewangan@nvidia.com>]

Hi Takashi,
Thanks for sample code. It helps lot.

On Friday 29 June 2012 05:43 PM, Takashi Iwai wrote:
> At Fri, 29 Jun 2012 15:53:15 +0530,
> Laxman Dewangan wrote:
>> Some of the ARM based soc allocate the writecombine dma buffer for
>> pcm substreams. They have the same codes for managing this buffer.
>> Moving this to the core/pcm files so that they can use that directly.
>>
>> Remove the code from Tegra PCM and use these new library function.
>>
>> This is enabled only for ARM specific and can be extended to other
>> architecture if they support the writecombine dma buffer.
>>
>> This patch is based on detail discussion on patch:
>> [PATCH] ASoC: snd_dmaengine: add common api for pcm_mmap
>> And suggestion from Lars and Takashi.
> Looking through your patch, I think an easier integration is just to
> add writecombine option to memalloc.c which calls
> dma_alloc_writecombine() instead of dma_alloc_coherent().
>
> The addition for mmap is still an open question.  Again, an easier
> option so far looks like just add the call of dma_alloc_writecombine()
> in snd_pcm_lib_default_mmap().  Alternatively, we can create an
> individual mmap pcm_ops as we discussed.
>
> Below is a totally untested patch, but you can imagine what I meant.
>
> After this change, replace with snd_pcm_lib_malloc_pages with
> SNDRV_DMA_TYPE_DEV_WC, and that's all you need in the driver side.

We can not use the snd_pcm_lib_malloc_pages() in the pcm_new callback 
because at this time the substream->runtime is not initialized and it 
leads to the kernel crash.

I used the apis as

int snd_soc_pcm_new_wc_dma_buffer(struct snd_soc_pcm_runtime *rtd,
                 size_t max_bytes)
{
      :::::::::::::
         substream = pcm->streams[SNDRV_PCM_STREAM_PLAYBACK].substream;
         if (substream) {
                 substream->dma_buffer.dev.type = SNDRV_DMA_TYPE_DEV_WC;
                 ret = snd_pcm_lib_malloc_pages(substream, max_bytes);
                 if (ret)
                         goto err;
         }
::::::::
}


int snd_pcm_lib_malloc_pages(struct snd_pcm_substream *substream, size_t 
size)
{
         struct snd_pcm_runtime *runtime;
         struct snd_dma_buffer *dmab = NULL;

         if (PCM_RUNTIME_CHECK(substream))
                 return -EINVAL;

         if (snd_BUG_ON(substream->dma_buffer.dev.type ==
                        SNDRV_DMA_TYPE_UNKNOWN))
                 return -EINVAL;

         runtime = substream->runtime;

         if (runtime->dma_buffer_p) {
---------------Kernel crash at this point ----------

So I used the snd_dma_alloc_pages() from the driver to allocate WC memory.


> +/* allocate the coherent DMA pages */
> +static void *snd_malloc_dev_pages(struct device *dev, size_t size, dma_addr_t *dma)
> +{
> +	return __snd_malloc_dev_pages(dev, size, dma, dma_alloc_coherent);

This does not get compiled in ARM because dma_alloc_coherant is macro 
defined as

#define dma_alloc_coherent(d, s, h, f) dma_alloc_attrs(d, s, h, f, NULL)

static inline void *dma_alloc_attrs(struct device *dev, size_t size,
                                        dma_addr_t *dma_handle, gfp_t flag,
                                        struct dma_attrs *attrs)


I fixed this by
static void *_dma_alloc_coherent(struct device *dev, size_t size,
                 dma_addr_t *dma_handle, gfp_t flag)
{
         return dma_alloc_coherent(dev, size, dma_handle, flag);
}

/* allocate the coherent DMA pages */
static void *snd_malloc_dev_pages(struct device *dev, size_t size, 
dma_addr_t *dma)
{
         return _snd_malloc_dev_pages(dev, size, dma, _dma_alloc_coherent);
}