Re: [PATCH 1/2] dma-mapping: prepare dma_map_ops to conversion to physical address

[Jason Gunthorpe <jgg@ziepe.ca>]

On Thu, Jul 17, 2025 at 09:47:09AM +0300, Leon Romanovsky wrote:

> @@ -183,7 +187,11 @@ dma_addr_t dma_map_phys(struct device *dev, phys_addr_t phys, size_t size,
>  		 * All platforms which implement .map_page() don't support
>  		 * non-struct page backed addresses.
>  		 */
> -		addr = ops->map_page(dev, page, offset, size, dir, attrs);
> +		if (ops->map_phys)
> +			addr = ops->map_phys(dev, phys, size, dir, attrs);
> +		if (ops->map_page)
> +			addr = ops->map_page(dev, page, offset, size, dir,
> +					     attrs);

I would expect 'else if'

> @@ -213,8 +221,16 @@ void dma_unmap_phys(struct device *dev, dma_addr_t addr, size_t size,
>  		dma_direct_unmap_phys(dev, addr, size, dir, attrs);
>  	else if (use_dma_iommu(dev))
>  		iommu_dma_unmap_phys(dev, addr, size, dir, attrs);
> -	else
> -		ops->unmap_page(dev, addr, size, dir, attrs);
> +	else {
> +		if (IS_ENABLED(CONFIG_DMA_API_DEBUG))
> +			/* We shouldn't have both functions */
> +			WARN_ON_ONCE(ops->unmap_page && ops->unmap_phys);
> +
> +		if (ops->unmap_phys)
> +			ops->unmap_phys(dev, addr, size, dir, attrs);
> +		if (ops->unmap_page)
> +			ops->unmap_page(dev, addr, size, dir, attrs);

Here too

What is the practical difference between unmap_phys and unmap_page?

Do we expect implementations of map/unmap phys to never touch the
struct page? I see a few cases of dma_ops that are doing this, eg for
cache flushing so they won't be converted?

If not, how are the ops supposed to know if a KVA is available or not
for the phys argument?

Jason