alsa-devel.alsa-project.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation
@ 2013-10-22  3:00 Nicolin Chen
  2013-10-22 12:06 ` Lars-Peter Clausen
  0 siblings, 1 reply; 5+ messages in thread
From: Nicolin Chen @ 2013-10-22  3:00 UTC (permalink / raw)
  To: lars, broonie, lgirdwood, tiwai, perex; +Cc: alsa-devel

Now it's quite common that an SoC contains its on-chip internal memory.
By using this memory space for DMA buffer during audio playback/record,
we can shutdown the voltage for external memory to save power.

Thus add new DEV type with iram malloc()/free() and accordingly modify
current default mmap() for the iram circumstance.

Signed-off-by: Nicolin Chen <b42378@freescale.com>
---
Changelog
v5->v6:
 * Dropped remaining OF dependency bacause no need to add them:
   of_node doesn't care about OF; of_get_named_gen_pool() returns NULL if !OF.
v4->v5:
 * Minimized the OF dependency.
v3->v4:
 * Appropriately placed '#ifdef CONFIG_OF'
 * Tested by adding '#undef CONFIG_OF' to modified files, and passed compiling.
v2->v3:
 * Moved iram specific mmap procedure out of ifdef ARCH_HAS_DMA_MMAP_COHERENT
v1->v2:
 * Added of_node check
 * Dropped noops and unused physical addr in snd_free_dev_iram()

---
 include/sound/memalloc.h |  1 +
 sound/core/memalloc.c    | 65 ++++++++++++++++++++++++++++++++++++++++++++++++
 sound/core/pcm_native.c  |  6 +++++
 3 files changed, 72 insertions(+)

diff --git a/include/sound/memalloc.h b/include/sound/memalloc.h
index cf15b82..510aec4 100644
--- a/include/sound/memalloc.h
+++ b/include/sound/memalloc.h
@@ -52,6 +52,7 @@ struct snd_dma_device {
 #else
 #define SNDRV_DMA_TYPE_DEV_SG	SNDRV_DMA_TYPE_DEV /* no SG-buf support */
 #endif
+#define SNDRV_DMA_TYPE_DEV_IRAM		4	/* generic device iram-buffer */
 
 /*
  * info for buffer allocation
diff --git a/sound/core/memalloc.c b/sound/core/memalloc.c
index bdf826f..ee505e9 100644
--- a/sound/core/memalloc.c
+++ b/sound/core/memalloc.c
@@ -30,6 +30,7 @@
 #include <linux/seq_file.h>
 #include <asm/uaccess.h>
 #include <linux/dma-mapping.h>
+#include <linux/genalloc.h>
 #include <linux/moduleparam.h>
 #include <linux/mutex.h>
 #include <sound/memalloc.h>
@@ -157,6 +158,59 @@ static void snd_free_dev_pages(struct device *dev, size_t size, void *ptr,
 	dec_snd_pages(pg);
 	dma_free_coherent(dev, PAGE_SIZE << pg, ptr, dma);
 }
+
+/**
+ * snd_malloc_dev_iram - allocate memory from on-chip internal memory
+ * @dev: DMA device pointer
+ * @size: number of bytes to allocate from the iram
+ * @dma: dma-view physical address
+ *
+ * Return cpu-view address or NULL indicating allocating failure
+ *
+ * This function requires iram phandle provided via of_node
+ */
+void *snd_malloc_dev_iram(struct device *dev, size_t size, dma_addr_t *dma)
+{
+	struct gen_pool *pool = NULL;
+	unsigned long vaddr;
+
+	if (!dev->of_node)
+		return NULL;
+
+	pool = of_get_named_gen_pool(dev->of_node, "iram", 0);
+	if (!pool)
+		return NULL;
+
+	vaddr = gen_pool_alloc(pool, size);
+	if (!vaddr)
+		return NULL;
+
+	*dma = gen_pool_virt_to_phys(pool, vaddr);
+
+	return (void *)vaddr;
+}
+
+/**
+ * snd_free_dev_iram - free allocated specific memory from on-chip internal memory
+ * @dev: DMA device pointer
+ * @size: size in bytes of memory to free
+ * @ptr: cpu-view address returned from snd_malloc_dev_iram
+ *
+ * This function requires iram phandle provided via of_node
+ */
+void snd_free_dev_iram(struct device *dev, size_t size, void *ptr)
+{
+	struct gen_pool *pool = NULL;
+
+	if (!dev->of_node)
+		return;
+
+	pool = of_get_named_gen_pool(dev->of_node, "iram", 0);
+	if (!pool)
+		return;
+
+	gen_pool_free(pool, (unsigned long)ptr, size);
+}
 #endif /* CONFIG_HAS_DMA */
 
 /*
@@ -197,6 +251,14 @@ int snd_dma_alloc_pages(int type, struct device *device, size_t size,
 		dmab->addr = 0;
 		break;
 #ifdef CONFIG_HAS_DMA
+	case SNDRV_DMA_TYPE_DEV_IRAM:
+		dmab->area = snd_malloc_dev_iram(device, size, &dmab->addr);
+		if (dmab->area)
+			break;
+		/* Internal memory might have limited size and no enough space,
+		 * so if we fail to malloc, try to fetch memory traditionally.
+		 */
+		dmab->dev.type = SNDRV_DMA_TYPE_DEV;
 	case SNDRV_DMA_TYPE_DEV:
 		dmab->area = snd_malloc_dev_pages(device, size, &dmab->addr);
 		break;
@@ -269,6 +331,9 @@ void snd_dma_free_pages(struct snd_dma_buffer *dmab)
 		snd_free_pages(dmab->area, dmab->bytes);
 		break;
 #ifdef CONFIG_HAS_DMA
+	case SNDRV_DMA_TYPE_DEV_IRAM:
+		snd_free_dev_iram(dmab->dev.dev, dmab->bytes, dmab->area);
+		break;
 	case SNDRV_DMA_TYPE_DEV:
 		snd_free_dev_pages(dmab->dev.dev, dmab->bytes, dmab->area, dmab->addr);
 		break;
diff --git a/sound/core/pcm_native.c b/sound/core/pcm_native.c
index a68d4c6..513f095 100644
--- a/sound/core/pcm_native.c
+++ b/sound/core/pcm_native.c
@@ -3199,6 +3199,12 @@ int snd_pcm_lib_default_mmap(struct snd_pcm_substream *substream,
 			     struct vm_area_struct *area)
 {
 	area->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
+	if (substream->dma_buffer.dev.type == SNDRV_DMA_TYPE_DEV_IRAM) {
+		area->vm_page_prot = pgprot_writecombine(area->vm_page_prot);
+		return remap_pfn_range(area, area->vm_start,
+				substream->dma_buffer.addr >> PAGE_SHIFT,
+				area->vm_end - area->vm_start, area->vm_page_prot);
+	}
 #ifdef ARCH_HAS_DMA_MMAP_COHERENT
 	if (!substream->ops->page &&
 	    substream->dma_buffer.dev.type == SNDRV_DMA_TYPE_DEV)
-- 
1.8.4

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation
  2013-10-22  3:00 [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation Nicolin Chen
@ 2013-10-22 12:06 ` Lars-Peter Clausen
  2013-10-22 12:49   ` Nicole Otsuka
  0 siblings, 1 reply; 5+ messages in thread
From: Lars-Peter Clausen @ 2013-10-22 12:06 UTC (permalink / raw)
  To: Nicolin Chen; +Cc: tiwai, alsa-devel, broonie, lgirdwood

On 10/22/2013 05:00 AM, Nicolin Chen wrote:
[...]
> +/**
> + * snd_free_dev_iram - free allocated specific memory from on-chip internal memory
> + * @dev: DMA device pointer
> + * @size: size in bytes of memory to free
> + * @ptr: cpu-view address returned from snd_malloc_dev_iram
> + *
> + * This function requires iram phandle provided via of_node
> + */
> +void snd_free_dev_iram(struct device *dev, size_t size, void *ptr)
> +{
> +	struct gen_pool *pool = NULL;
> +
> +	if (!dev->of_node)
> +		return;
> +
> +	pool = of_get_named_gen_pool(dev->of_node, "iram", 0);
> +	if (!pool)
> +		return;

I've had a closer look at the other SNDRV_DMA_TYPE implementations and I
think a better way to handle this is to assign the pool to the
snd_dma_buffer's private_data field and then use it here instead of looking
the pool up again. This will also make it easier to use a non-OF lookup scheme.

> +
> +	gen_pool_free(pool, (unsigned long)ptr, size);
> +}
>  #endif /* CONFIG_HAS_DMA */
>  
>  /*

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation
  2013-10-22 12:06 ` Lars-Peter Clausen
@ 2013-10-22 12:49   ` Nicole Otsuka
  2013-10-22 13:24     ` Takashi Iwai
  0 siblings, 1 reply; 5+ messages in thread
From: Nicole Otsuka @ 2013-10-22 12:49 UTC (permalink / raw)
  To: Lars-Peter Clausen; +Cc: tiwai, alsa-devel, Nicolin Chen, broonie, lgirdwood

On Tue, Oct 22, 2013 at 8:06 PM, Lars-Peter Clausen <lars@metafoo.de> wrote:

> On 10/22/2013 05:00 AM, Nicolin Chen wrote:
> [...]
> > +/**
> > + * snd_free_dev_iram - free allocated specific memory from on-chip
> internal memory
> > + * @dev: DMA device pointer
> > + * @size: size in bytes of memory to free
> > + * @ptr: cpu-view address returned from snd_malloc_dev_iram
> > + *
> > + * This function requires iram phandle provided via of_node
> > + */
> > +void snd_free_dev_iram(struct device *dev, size_t size, void *ptr)
> > +{
> > +     struct gen_pool *pool = NULL;
> > +
> > +     if (!dev->of_node)
> > +             return;
> > +
> > +     pool = of_get_named_gen_pool(dev->of_node, "iram", 0);
> > +     if (!pool)
> > +             return;
>
> I've had a closer look at the other SNDRV_DMA_TYPE implementations and I
> think a better way to handle this is to assign the pool to the
> snd_dma_buffer's private_data field and then use it here instead of looking
> the pool up again. This will also make it easier to use a non-OF lookup
> scheme.
>

> By saving the pool to pdata during the allocating? Hmm..that should be
nicer.
> Just need to change the parameter "struct device *" to "struct gen_pool
*" directly
> I'll see what I can do to refine it. Thank you for the suggestion.
>
> Nicolin Chen


>
> > +
> > +     gen_pool_free(pool, (unsigned long)ptr, size);
> > +}
> >  #endif /* CONFIG_HAS_DMA */
> >
> >  /*
> _______________________________________________
> Alsa-devel mailing list
> Alsa-devel@alsa-project.org
> http://mailman.alsa-project.org/mailman/listinfo/alsa-devel
>

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation
  2013-10-22 12:49   ` Nicole Otsuka
@ 2013-10-22 13:24     ` Takashi Iwai
  2013-10-22 16:28       ` Nicole Otsuka
  0 siblings, 1 reply; 5+ messages in thread
From: Takashi Iwai @ 2013-10-22 13:24 UTC (permalink / raw)
  To: Nicole Otsuka
  Cc: alsa-devel, Lars-Peter Clausen, broonie, Nicolin Chen, lgirdwood

At Tue, 22 Oct 2013 20:49:49 +0800,
Nicole Otsuka wrote:
> 
> On Tue, Oct 22, 2013 at 8:06 PM, Lars-Peter Clausen <lars@metafoo.de> wrote:
> 
> > On 10/22/2013 05:00 AM, Nicolin Chen wrote:
> > [...]
> > > +/**
> > > + * snd_free_dev_iram - free allocated specific memory from on-chip
> > internal memory
> > > + * @dev: DMA device pointer
> > > + * @size: size in bytes of memory to free
> > > + * @ptr: cpu-view address returned from snd_malloc_dev_iram
> > > + *
> > > + * This function requires iram phandle provided via of_node
> > > + */
> > > +void snd_free_dev_iram(struct device *dev, size_t size, void *ptr)
> > > +{
> > > +     struct gen_pool *pool = NULL;
> > > +
> > > +     if (!dev->of_node)
> > > +             return;
> > > +
> > > +     pool = of_get_named_gen_pool(dev->of_node, "iram", 0);
> > > +     if (!pool)
> > > +             return;
> >
> > I've had a closer look at the other SNDRV_DMA_TYPE implementations and I
> > think a better way to handle this is to assign the pool to the
> > snd_dma_buffer's private_data field and then use it here instead of looking
> > the pool up again. This will also make it easier to use a non-OF lookup
> > scheme.
> >
> 
> > By saving the pool to pdata during the allocating? Hmm..that should be
> nicer.
> > Just need to change the parameter "struct device *" to "struct gen_pool
> *" directly
> > I'll see what I can do to refine it. Thank you for the suggestion.

I guess Lars meant to use simply snd_dma_buffer.private_data field for
storing the pool pointer in snd_malloc_dev_iram(), and use this
pointer instead of calling of_get_named_gen_pool() again in
snd_free_dev_iram().


Takashi

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation
  2013-10-22 13:24     ` Takashi Iwai
@ 2013-10-22 16:28       ` Nicole Otsuka
  0 siblings, 0 replies; 5+ messages in thread
From: Nicole Otsuka @ 2013-10-22 16:28 UTC (permalink / raw)
  To: Takashi Iwai
  Cc: alsa-devel, Lars-Peter Clausen, broonie, Nicolin Chen,
	Liam Girdwood

On Tue, Oct 22, 2013 at 9:24 PM, Takashi Iwai <tiwai@suse.de> wrote:

> At Tue, 22 Oct 2013 20:49:49 +0800,
> Nicole Otsuka wrote:
> >
> > On Tue, Oct 22, 2013 at 8:06 PM, Lars-Peter Clausen <lars@metafoo.de>
> wrote:
> >
> > > On 10/22/2013 05:00 AM, Nicolin Chen wrote:
> > > [...]
> > > > +/**
> > > > + * snd_free_dev_iram - free allocated specific memory from on-chip
> > > internal memory
> > > > + * @dev: DMA device pointer
> > > > + * @size: size in bytes of memory to free
> > > > + * @ptr: cpu-view address returned from snd_malloc_dev_iram
> > > > + *
> > > > + * This function requires iram phandle provided via of_node
> > > > + */
> > > > +void snd_free_dev_iram(struct device *dev, size_t size, void *ptr)
> > > > +{
> > > > +     struct gen_pool *pool = NULL;
> > > > +
> > > > +     if (!dev->of_node)
> > > > +             return;
> > > > +
> > > > +     pool = of_get_named_gen_pool(dev->of_node, "iram", 0);
> > > > +     if (!pool)
> > > > +             return;
> > >
> > > I've had a closer look at the other SNDRV_DMA_TYPE implementations and
> I
> > > think a better way to handle this is to assign the pool to the
> > > snd_dma_buffer's private_data field and then use it here instead of
> looking
> > > the pool up again. This will also make it easier to use a non-OF lookup
> > > scheme.
> > >
> >
> > > By saving the pool to pdata during the allocating? Hmm..that should be
> > nicer.
> > > Just need to change the parameter "struct device *" to "struct gen_pool
> > *" directly
> > > I'll see what I can do to refine it. Thank you for the suggestion.
>
> > I guess Lars meant to use simply snd_dma_buffer.private_data field for
> > storing the pool pointer in snd_malloc_dev_iram(), and use this
> > pointer instead of calling of_get_named_gen_pool() again in
> > snd_free_dev_iram().
>
> Oh, thank you for the explain.
I'll try to refine it tomorrow.

Nicolin Chen

>
> > Takashi
>

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2013-10-22 16:28 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-10-22  3:00 [PATCH v6] ALSA: Add SoC on-chip internal memory support for DMA buffer allocation Nicolin Chen
2013-10-22 12:06 ` Lars-Peter Clausen
2013-10-22 12:49   ` Nicole Otsuka
2013-10-22 13:24     ` Takashi Iwai
2013-10-22 16:28       ` Nicole Otsuka

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).