[PATCH v2] media: vb2-dc: skip CPU sync in map/unmap dma_buf

[PATCH v2] media: vb2-dc: skip CPU sync in map/unmap dma_buf

[Lucas Stach]

This is rougly equivalent to ca0e68e21aae (drm/prime: skip CPU sync
in map/unmap dma_buf). The contig memory allocated is already device
coherent memory, so there is no point in doing a CPU sync when
mapping it to another deevice. Also most importers currently cache
the mapping so the CPU sync would only happen on the first import,
so we are better off with not pretending to do a cache synchronization
at all.

This gets rid of a lot of CPU overhead in uses where those dma-bufs
are regularily imported and detached again, like Weston is currently
doing in the DRM compositor.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
---
v2: Add comments why it is safe to skip the CPU sync.
---
 .../common/videobuf2/videobuf2-dma-contig.c   | 23 +++++++++++++------
 1 file changed, 16 insertions(+), 7 deletions(-)

diff --git a/drivers/media/common/videobuf2/videobuf2-dma-contig.c b/drivers/media/common/videobuf2/videobuf2-dma-contig.c
index ecbef266130b..1b8f86366290 100644
--- a/drivers/media/common/videobuf2/videobuf2-dma-contig.c
+++ b/drivers/media/common/videobuf2/videobuf2-dma-contig.c
@@ -267,8 +267,14 @@ static void vb2_dc_dmabuf_ops_detach(struct dma_buf *dbuf,
 
 	/* release the scatterlist cache */
 	if (attach->dma_dir != DMA_NONE)
-		dma_unmap_sg(db_attach->dev, sgt->sgl, sgt->orig_nents,
-			attach->dma_dir);
+		/*
+		 * Cache sync can be skipped here, as the vb2_dc memory is
+		 * allocated from device coherent memory, which means the
+		 * memory locations do not require any explicit cache
+		 * maintenance prior or after being used by the device.
+		 */
+		dma_unmap_sg_attrs(db_attach->dev, sgt->sgl, sgt->orig_nents,
+			attach->dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
 	sg_free_table(sgt);
 	kfree(attach);
 	db_attach->priv = NULL;
@@ -293,14 +299,17 @@ static struct sg_table *vb2_dc_dmabuf_ops_map(
 
 	/* release any previous cache */
 	if (attach->dma_dir != DMA_NONE) {
-		dma_unmap_sg(db_attach->dev, sgt->sgl, sgt->orig_nents,
-			attach->dma_dir);
+		dma_unmap_sg_attrs(db_attach->dev, sgt->sgl, sgt->orig_nents,
+			attach->dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
 		attach->dma_dir = DMA_NONE;
 	}
 
-	/* mapping to the client with new direction */
-	sgt->nents = dma_map_sg(db_attach->dev, sgt->sgl, sgt->orig_nents,
-				dma_dir);
+	/*
+	 * mapping to the client with new direction, no cache sync
+	 * required see comment in vb2_dc_dmabuf_ops_detach()
+	 */
+	sgt->nents = dma_map_sg_attrs(db_attach->dev, sgt->sgl, sgt->orig_nents,
+				dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
 	if (!sgt->nents) {
 		pr_err("failed to map scatterlist\n");
 		mutex_unlock(lock);
-- 
2.20.1

Re: [PATCH v2] media: vb2-dc: skip CPU sync in map/unmap dma_buf

[Marek Szyprowski]

Hi Lucas,

On 2019-07-08 15:07, Lucas Stach wrote:
> This is rougly equivalent to ca0e68e21aae (drm/prime: skip CPU sync
> in map/unmap dma_buf). The contig memory allocated is already device
> coherent memory, so there is no point in doing a CPU sync when
> mapping it to another deevice. Also most importers currently cache
> the mapping so the CPU sync would only happen on the first import,
> so we are better off with not pretending to do a cache synchronization
> at all.
>
> This gets rid of a lot of CPU overhead in uses where those dma-bufs
> are regularily imported and detached again, like Weston is currently
> doing in the DRM compositor.
>
> Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Acked-by: Marek Szyprowski <m.szyprowski@samsung.com>
> ---
> v2: Add comments why it is safe to skip the CPU sync.
> ---
>   .../common/videobuf2/videobuf2-dma-contig.c   | 23 +++++++++++++------
>   1 file changed, 16 insertions(+), 7 deletions(-)
>
> diff --git a/drivers/media/common/videobuf2/videobuf2-dma-contig.c b/drivers/media/common/videobuf2/videobuf2-dma-contig.c
> index ecbef266130b..1b8f86366290 100644
> --- a/drivers/media/common/videobuf2/videobuf2-dma-contig.c
> +++ b/drivers/media/common/videobuf2/videobuf2-dma-contig.c
> @@ -267,8 +267,14 @@ static void vb2_dc_dmabuf_ops_detach(struct dma_buf *dbuf,
>   
>   	/* release the scatterlist cache */
>   	if (attach->dma_dir != DMA_NONE)
> -		dma_unmap_sg(db_attach->dev, sgt->sgl, sgt->orig_nents,
> -			attach->dma_dir);
> +		/*
> +		 * Cache sync can be skipped here, as the vb2_dc memory is
> +		 * allocated from device coherent memory, which means the
> +		 * memory locations do not require any explicit cache
> +		 * maintenance prior or after being used by the device.
> +		 */
> +		dma_unmap_sg_attrs(db_attach->dev, sgt->sgl, sgt->orig_nents,
> +			attach->dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
>   	sg_free_table(sgt);
>   	kfree(attach);
>   	db_attach->priv = NULL;
> @@ -293,14 +299,17 @@ static struct sg_table *vb2_dc_dmabuf_ops_map(
>   
>   	/* release any previous cache */
>   	if (attach->dma_dir != DMA_NONE) {
> -		dma_unmap_sg(db_attach->dev, sgt->sgl, sgt->orig_nents,
> -			attach->dma_dir);
> +		dma_unmap_sg_attrs(db_attach->dev, sgt->sgl, sgt->orig_nents,
> +			attach->dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
>   		attach->dma_dir = DMA_NONE;
>   	}
>   
> -	/* mapping to the client with new direction */
> -	sgt->nents = dma_map_sg(db_attach->dev, sgt->sgl, sgt->orig_nents,
> -				dma_dir);
> +	/*
> +	 * mapping to the client with new direction, no cache sync
> +	 * required see comment in vb2_dc_dmabuf_ops_detach()
> +	 */
> +	sgt->nents = dma_map_sg_attrs(db_attach->dev, sgt->sgl, sgt->orig_nents,
> +				dma_dir, DMA_ATTR_SKIP_CPU_SYNC);
>   	if (!sgt->nents) {
>   		pr_err("failed to map scatterlist\n");
>   		mutex_unlock(lock);

Best regards
-- 
Marek Szyprowski, PhD
Samsung R&D Institute Poland