Hi Vinod,
On 2016-02-21 20:50:46 +0530, Vinod Koul wrote:
> The slave dmaengine semantics required the client to map dma
> addresses and pass DMA address to dmaengine drivers. While this
> was a convenient notion coming from generic dma offload cases
> where dmaengines are interchangeable and client is not aware of
> which engine to map to.
>
> But in case of slave, we know the dmaengine and always use a
> specific one. Further the IOMMU cases can lead to failure of this
> notion, so make this as physical address and now dmaengine driver
> will do the required mapping.
Thanks! I have tested this patch with my IOMMU series and it works
nicely.
>
> Original-patch-by: Linus Walleij <linus.walleij@xxxxxxxxxx>
> Signed-off-by: Vinod Koul <vinod.koul@xxxxxxxxx>
Tested-by: Niklas Söderlund <niklas.soderlund+renesas@xxxxxxxxxxxx>
> ---
> include/linux/dmaengine.h | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
> index 16a1cad30c33..d85ecd20af50 100644
> --- a/include/linux/dmaengine.h
> +++ b/include/linux/dmaengine.h
> @@ -357,8 +357,8 @@ enum dma_slave_buswidth {
> */
> struct dma_slave_config {
> enum dma_transfer_direction direction;
> - dma_addr_t src_addr;
> - dma_addr_t dst_addr;
> + phys_addr_t src_addr;
> + phys_addr_t dst_addr;
> enum dma_slave_buswidth src_addr_width;
> enum dma_slave_buswidth dst_addr_width;
> u32 src_maxburst;
--
Regards,
Niklas Söderlund
