From: Leon Romanovsky <leonro@nvidia.com>

Extend base DMA page API to handle MMIO flow and follow
existing dma_map_resource() implementation to rely on dma_map_direct()
only to take DMA direct path.

Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 kernel/dma/mapping.c | 26 +++++++++++++++++++++-----
 1 file changed, 21 insertions(+), 5 deletions(-)

diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
index 891e1fc3e582..fdabfdaeff1d 100644
--- a/kernel/dma/mapping.c
+++ b/kernel/dma/mapping.c
@@ -158,6 +158,7 @@ dma_addr_t dma_map_page_attrs(struct device *dev, struct page *page,
 {
 	const struct dma_map_ops *ops = get_dma_ops(dev);
 	phys_addr_t phys = page_to_phys(page) + offset;
+	bool is_mmio = attrs & DMA_ATTR_MMIO;
 	dma_addr_t addr;
 
 	BUG_ON(!valid_dma_direction(dir));
@@ -166,14 +167,25 @@ dma_addr_t dma_map_page_attrs(struct device *dev, struct page *page,
 		return DMA_MAPPING_ERROR;
 
 	if (dma_map_direct(dev, ops) ||
-	    arch_dma_map_phys_direct(dev, phys + size))
+	    (!is_mmio && arch_dma_map_phys_direct(dev, phys + size)))
 		addr = dma_direct_map_phys(dev, phys, size, dir, attrs);
 	else if (use_dma_iommu(dev))
 		addr = iommu_dma_map_phys(dev, phys, size, dir, attrs);
-	else
+	else if (is_mmio) {
+		if (!ops->map_resource)
+			return DMA_MAPPING_ERROR;
+
+		addr = ops->map_resource(dev, phys, size, dir, attrs);
+	} else {
+		/*
+		 * The dma_ops API contract for ops->map_page() requires
+		 * kmappable memory, while ops->map_resource() does not.
+		 */
 		addr = ops->map_page(dev, page, offset, size, dir, attrs);
+	}
 
-	kmsan_handle_dma(phys, size, dir);
+	if (!is_mmio)
+		kmsan_handle_dma(phys, size, dir);
 	trace_dma_map_phys(dev, phys, addr, size, dir, attrs);
 	debug_dma_map_phys(dev, phys, size, dir, addr, attrs);
 
@@ -185,14 +197,18 @@ void dma_unmap_page_attrs(struct device *dev, dma_addr_t addr, size_t size,
 		enum dma_data_direction dir, unsigned long attrs)
 {
 	const struct dma_map_ops *ops = get_dma_ops(dev);
+	bool is_mmio = attrs & DMA_ATTR_MMIO;
 
 	BUG_ON(!valid_dma_direction(dir));
 	if (dma_map_direct(dev, ops) ||
-	    arch_dma_unmap_phys_direct(dev, addr + size))
+	    (!is_mmio && arch_dma_unmap_phys_direct(dev, addr + 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
+	else if (is_mmio) {
+		if (ops->unmap_resource)
+			ops->unmap_resource(dev, addr, size, dir, attrs);
+	} else
 		ops->unmap_page(dev, addr, size, dir, attrs);
 	trace_dma_unmap_phys(dev, addr, size, dir, attrs);
 	debug_dma_unmap_phys(dev, addr, size, dir);
-- 
2.50.1

On Tue, Aug 19, 2025 at 08:36:53PM +0300, Leon Romanovsky wrote:
> From: Leon Romanovsky <leonro@nvidia.com>
> 
> Extend base DMA page API to handle MMIO flow and follow
> existing dma_map_resource() implementation to rely on dma_map_direct()
> only to take DMA direct path.

I would reword this a little bit too

dma-mapping: implement DMA_ATTR_MMIO for dma_(un)map_page_attrs()

Make dma_map_page_attrs() and dma_map_page_attrs() respect
DMA_ATTR_MMIO.

DMA_ATR_MMIO makes the functions behave the same as dma_(un)map_resource():
 - No swiotlb is possible
 - Legacy dma_ops arches use ops->map_resource()
 - No kmsan
 - No arch_dma_map_phys_direct()

The prior patches have made the internl funtions called here support
DMA_ATTR_MMIO.

This is also preparation for turning dma_map_resource() into an inline
calling dma_map_phys(DMA_ATTR_MMIO) to consolidate the flows.

> @@ -166,14 +167,25 @@ dma_addr_t dma_map_page_attrs(struct device *dev, struct page *page,
>  		return DMA_MAPPING_ERROR;
>  
>  	if (dma_map_direct(dev, ops) ||
> -	    arch_dma_map_phys_direct(dev, phys + size))
> +	    (!is_mmio && arch_dma_map_phys_direct(dev, phys + size)))
>  		addr = dma_direct_map_phys(dev, phys, size, dir, attrs);

PPC is the only user of arch_dma_map_phys_direct() and it looks like
it should be called on MMIO memory. Seems like another inconsistency
with map_resource. I'd leave it like the above though for this series.

>  	else if (use_dma_iommu(dev))
>  		addr = iommu_dma_map_phys(dev, phys, size, dir, attrs);
> -	else
> +	else if (is_mmio) {
> +		if (!ops->map_resource)
> +			return DMA_MAPPING_ERROR;

Probably written like:

		if (ops->map_resource)
			addr = ops->map_resource(dev, phys, size, dir, attrs);
		else
			addr = DMA_MAPPING_ERROR;

As I think some of the design here is to run the trace even on the
failure path?

Otherwise looks OK

Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>

Jason

On Thu, Aug 28, 2025 at 12:17:30PM -0300, Jason Gunthorpe wrote:
> On Tue, Aug 19, 2025 at 08:36:53PM +0300, Leon Romanovsky wrote:
> > From: Leon Romanovsky <leonro@nvidia.com>
> > 
> > Extend base DMA page API to handle MMIO flow and follow
> > existing dma_map_resource() implementation to rely on dma_map_direct()
> > only to take DMA direct path.
> 
> I would reword this a little bit too
> 
> dma-mapping: implement DMA_ATTR_MMIO for dma_(un)map_page_attrs()
> 
> Make dma_map_page_attrs() and dma_map_page_attrs() respect
> DMA_ATTR_MMIO.
> 
> DMA_ATR_MMIO makes the functions behave the same as dma_(un)map_resource():
>  - No swiotlb is possible
>  - Legacy dma_ops arches use ops->map_resource()
>  - No kmsan
>  - No arch_dma_map_phys_direct()
> 
> The prior patches have made the internl funtions called here support
> DMA_ATTR_MMIO.
> 
> This is also preparation for turning dma_map_resource() into an inline
> calling dma_map_phys(DMA_ATTR_MMIO) to consolidate the flows.
> 
> > @@ -166,14 +167,25 @@ dma_addr_t dma_map_page_attrs(struct device *dev, struct page *page,
> >  		return DMA_MAPPING_ERROR;
> >  
> >  	if (dma_map_direct(dev, ops) ||
> > -	    arch_dma_map_phys_direct(dev, phys + size))
> > +	    (!is_mmio && arch_dma_map_phys_direct(dev, phys + size)))
> >  		addr = dma_direct_map_phys(dev, phys, size, dir, attrs);
> 
> PPC is the only user of arch_dma_map_phys_direct() and it looks like
> it should be called on MMIO memory. Seems like another inconsistency
> with map_resource. I'd leave it like the above though for this series.
> 
> >  	else if (use_dma_iommu(dev))
> >  		addr = iommu_dma_map_phys(dev, phys, size, dir, attrs);
> > -	else
> > +	else if (is_mmio) {
> > +		if (!ops->map_resource)
> > +			return DMA_MAPPING_ERROR;
> 
> Probably written like:
> 
> 		if (ops->map_resource)
> 			addr = ops->map_resource(dev, phys, size, dir, attrs);
> 		else
> 			addr = DMA_MAPPING_ERROR;

I'm big fan of "if (!ops->map_resource)" coding style and prefer to keep it.

> 
> As I think some of the design here is to run the trace even on the
> failure path?

Yes, this is how it worked before.

> 
> Otherwise looks OK
> 
> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
> 
> Jason