[PATCH v5 11/14] sparc: Use physical address DMA mapping

[John Paul Adrian Glaubitz]

Hi Marek,

On Tue, 2026-02-17 at 13:55 +0100, Marek Szyprowski wrote:
> On 17.02.2026 13:27, John Paul Adrian Glaubitz wrote:
> > On Wed, 2025-10-15 at 12:12 +0300, Leon Romanovsky wrote:
> > > From: Leon Romanovsky <leonro at nvidia.com>
> > > 
> > > Convert sparc architecture DMA code to use .map_phys callback.
> > > 
> > > Signed-off-by: Leon Romanovsky <leonro at nvidia.com>
> > > ---
> > >   arch/sparc/kernel/iommu.c     | 30 +++++++++++++++++-----------
> > >   arch/sparc/kernel/pci_sun4v.c | 31 ++++++++++++++++++-----------
> > >   arch/sparc/mm/io-unit.c       | 38 ++++++++++++++++++-----------------
> > >   arch/sparc/mm/iommu.c         | 46 ++++++++++++++++++++++---------------------
> > >   4 files changed, 82 insertions(+), 63 deletions(-)
> > > 
> > > diff --git a/arch/sparc/kernel/iommu.c b/arch/sparc/kernel/iommu.c
> > > index da0363692528..46ef88bc9c26 100644
> > > --- a/arch/sparc/kernel/iommu.c
> > > +++ b/arch/sparc/kernel/iommu.c
> > > @@ -260,26 +260,35 @@ static void dma_4u_free_coherent(struct device *dev, size_t size,
> > >   		free_pages((unsigned long)cpu, order);
> > >   }
> > >   
> > > -static dma_addr_t dma_4u_map_page(struct device *dev, struct page *page,
> > > -				  unsigned long offset, size_t sz,
> > > -				  enum dma_data_direction direction,
> > > +static dma_addr_t dma_4u_map_phys(struct device *dev, phys_addr_t phys,
> > > +				  size_t sz, enum dma_data_direction direction,
> > >   				  unsigned long attrs)
> > >   {
> > >   	struct iommu *iommu;
> > >   	struct strbuf *strbuf;
> > >   	iopte_t *base;
> > >   	unsigned long flags, npages, oaddr;
> > > -	unsigned long i, base_paddr, ctx;
> > > +	unsigned long i, ctx;
> > >   	u32 bus_addr, ret;
> > >   	unsigned long iopte_protection;
> > >   
> > > +	if (unlikely(attrs & DMA_ATTR_MMIO))
> > > +		/*
> > > +		 * This check is included because older versions of the code
> > > +		 * lacked MMIO path support, and my ability to test this path
> > > +		 * is limited. However, from a software technical standpoint,
> > > +		 * there is no restriction, as the following code operates
> > > +		 * solely on physical addresses.
> > > +		 */
> > > +		goto bad_no_ctx;
> > > +
> > >   	iommu = dev->archdata.iommu;
> > >   	strbuf = dev->archdata.stc;
> > >   
> > >   	if (unlikely(direction == DMA_NONE))
> > >   		goto bad_no_ctx;
> > >   
> > > -	oaddr = (unsigned long)(page_address(page) + offset);
> > > +	oaddr = (unsigned long)(phys_to_virt(phys));
> > >   	npages = IO_PAGE_ALIGN(oaddr + sz) - (oaddr & IO_PAGE_MASK);
> > >   	npages >>= IO_PAGE_SHIFT;
> > >   
> > > @@ -296,7 +305,6 @@ static dma_addr_t dma_4u_map_page(struct device *dev, struct page *page,
> > >   	bus_addr = (iommu->tbl.table_map_base +
> > >   		    ((base - iommu->page_table) << IO_PAGE_SHIFT));
> > >   	ret = bus_addr | (oaddr & ~IO_PAGE_MASK);
> > > -	base_paddr = __pa(oaddr & IO_PAGE_MASK);
> > >   	if (strbuf->strbuf_enabled)
> > >   		iopte_protection = IOPTE_STREAMING(ctx);
> > >   	else
> > > @@ -304,8 +312,8 @@ static dma_addr_t dma_4u_map_page(struct device *dev, struct page *page,
> > >   	if (direction != DMA_TO_DEVICE)
> > >   		iopte_protection |= IOPTE_WRITE;
> > >   
> > > -	for (i = 0; i < npages; i++, base++, base_paddr += IO_PAGE_SIZE)
> > > -		iopte_val(*base) = iopte_protection | base_paddr;
> > > +	for (i = 0; i < npages; i++, base++, phys += IO_PAGE_SIZE)
> > > +		iopte_val(*base) = iopte_protection | phys;
> 
> 'base_paddr' was aligned to start of a page before this patch, while 
> 'phys' includes the offset within a page.
> 
> 
> Could You check if the following change fixes the issue?
> 
> diff --git a/arch/sparc/kernel/iommu.c b/arch/sparc/kernel/iommu.c
> index 46ef88bc9c26..edb0ffb21e2f 100644
> --- a/arch/sparc/kernel/iommu.c
> +++ b/arch/sparc/kernel/iommu.c
> @@ -312,6 +312,7 @@ static dma_addr_t dma_4u_map_phys(struct device 
> *dev, phys_addr_t phys,
>          if (direction != DMA_TO_DEVICE)
>                  iopte_protection |= IOPTE_WRITE;
> 
> +       phys -= offset_in_page(phys);
>          for (i = 0; i < npages; i++, base++, phys += IO_PAGE_SIZE)
>                  iopte_val(*base) = iopte_protection | phys;

Thanks, I have forwarded your patch and asked the user to answer here if possible [1].

Adrian

> [1] https://github.com/sparclinux/issues/issues/75#issuecomment-3914791862

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer
`. `'   Physicist
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913