Re: [PATCH v4 05/16] iommu/dma: rename iommu_dma_*map_page to iommu_dma_*map_phys

[Jason Gunthorpe <jgg@xxxxxxxxxx>]

On Tue, Aug 19, 2025 at 08:36:49PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@xxxxxxxxxx>
> 
> Rename the IOMMU DMA mapping functions to better reflect their actual
> calling convention. The functions iommu_dma_map_page() and
> iommu_dma_unmap_page() are renamed to iommu_dma_map_phys() and
> iommu_dma_unmap_phys() respectively, as they already operate on physical
> addresses rather than page structures.
> 
> The calling convention changes from accepting (struct page *page,
> unsigned long offset) to (phys_addr_t phys), which eliminates the need
> for page-to-physical address conversion within the functions. This
> renaming prepares for the broader DMA API conversion from page-based
> to physical address-based mapping throughout the kernel.
> 
> All callers are updated to pass physical addresses directly, including
> dma_map_page_attrs(), scatterlist mapping functions, and DMA page
> allocation helpers. The change simplifies the code by removing the
> page_to_phys() + offset calculation that was previously done inside
> the IOMMU functions.
> 
> Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxx>
> ---
>  drivers/iommu/dma-iommu.c | 14 ++++++--------
>  include/linux/iommu-dma.h |  7 +++----
>  kernel/dma/mapping.c      |  4 ++--
>  kernel/dma/ops_helpers.c  |  6 +++---
>  4 files changed, 14 insertions(+), 17 deletions(-)

This looks fine

Reviewed-by: Jason Gunthorpe <jgg@xxxxxxxxxx>

But related to other patches..

iommu_dma_map_phys() ends up like this:

        if (dev_use_swiotlb(dev, size, dir) &&
            iova_unaligned(iovad, phys, size)) {
                if (attrs & DMA_ATTR_MMIO)
                        return DMA_MAPPING_ERROR;

                phys = iommu_dma_map_swiotlb(dev, phys, size, dir, attrs);

But attrs is passed all the way down to swiotlb_tbl_map_single(),
maybe the DMA_ATTR_MMIO check should be moved there?

There are a few call chains with this redundancy:

dma_iova_link()
 -> iommu_dma_iova_link_swiotlb
  -> iommu_dma_iova_bounce_and_link
   -> iommu_dma_map_swiotlb
    -> swiotlb_tbl_map_single()

iommu_dma_map_phys() 
   -> iommu_dma_map_swiotlb
    -> swiotlb_tbl_map_single()

dma_direct_map_phys()
 -> swiotlb_map()
  -> swiotlb_tbl_map_single()

It makes alot of sense to put the check for MMIO when
slots[].orig_addr is stored because that is the point where we start
to require a pfn_to_page().

Jason