[PATCHv2,1/3] arm64: Implement custom mmap functions for dma mapping

Commit Message

Laura Abbott March 14, 2014, 7:52 p.m. UTC

The current dma_ops do not specify an mmap function so maping
falls back to the default implementation. There are at least
two issues with using the default implementation:

1) The pgprot is always pgprot_noncached (strongly ordered)
memory even with coherent operations
2) dma_common_mmap calls virt_to_page on the remapped non-coherent
address which leads to invalid memory being mapped.

Fix both these issue by implementing a custom mmap function which
correctly accounts for remapped addresses and sets vm_pg_prot
appropriately.

Signed-off-by: Laura Abbott <[e-mail]>
---
 arch/arm64/mm/dma-mapping.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 44 insertions(+)

Comments

Catalin Marinas March 24, 2014, 10:33 a.m. UTC | #1

On Fri, Mar 14, 2014 at 07:52:23PM +0000, Laura Abbott wrote:
> The current dma_ops do not specify an mmap function so maping
> falls back to the default implementation. There are at least
> two issues with using the default implementation:
> 
> 1) The pgprot is always pgprot_noncached (strongly ordered)
> memory even with coherent operations
> 2) dma_common_mmap calls virt_to_page on the remapped non-coherent
> address which leads to invalid memory being mapped.
> 
> Fix both these issue by implementing a custom mmap function which
> correctly accounts for remapped addresses and sets vm_pg_prot
> appropriately.
> 
> Signed-off-by: Laura Abbott <[e-mail]>

I thought there was still some update needed to this series but it turns
out that patch 2/3 was what I was expecting already, so I merged the
first two patches (with minor changes for s/arm64_/__/ prefix). Patch 3
seems to be folded onto my patch already.

Thanks.

Catalin Marinas March 28, 2014, 10:37 a.m. UTC | #2

On Fri, Mar 28, 2014 at 09:45:08AM +0000, Ritesh Harjani wrote:
> On Sat, Mar 15, 2014 at 1:22 AM, Laura Abbott <[e-mail]> wrote:
> > +static int __dma_common_mmap(struct device *dev, struct vm_area_struct *vma,
> > +                            void *cpu_addr, dma_addr_t dma_addr, size_t size)
> > +{
> > +       int ret = -ENXIO;
> > +       unsigned long nr_vma_pages = (vma->vm_end - vma->vm_start) >>
> > +                                       PAGE_SHIFT;
> > +       unsigned long nr_pages = PAGE_ALIGN(size) >> PAGE_SHIFT;
> > +       unsigned long pfn = dma_to_phys(dev, dma_addr) >> PAGE_SHIFT;
> 
> Why not __phys_to_pfn here ?? just asking, I know there is nothing wrong in
> this too..

Because dma_addr is a DMA address (as seen by the device from the other
side of the bus) rather than a CPU physical address. In many cases they
are the same but not always.

Patch

diff --git a/arch/arm64/mm/dma-mapping.c b/arch/arm64/mm/dma-mapping.c
index e923a5b..0cdd2f6 100644
--- a/arch/arm64/mm/dma-mapping.c
+++ b/arch/arm64/mm/dma-mapping.c
@@ -194,9 +194,52 @@  static void arm64_swiotlb_sync_sg_for_device(struct device *dev,
 			       sg->length, dir);
 }
 
+/* vma->vm_page_prot must be set appropriately before calling this function */
+static int __dma_common_mmap(struct device *dev, struct vm_area_struct *vma,
+			     void *cpu_addr, dma_addr_t dma_addr, size_t size)
+{
+	int ret = -ENXIO;
+	unsigned long nr_vma_pages = (vma->vm_end - vma->vm_start) >>
+					PAGE_SHIFT;
+	unsigned long nr_pages = PAGE_ALIGN(size) >> PAGE_SHIFT;
+	unsigned long pfn = dma_to_phys(dev, dma_addr) >> PAGE_SHIFT;
+	unsigned long off = vma->vm_pgoff;
+
+	if (dma_mmap_from_coherent(dev, vma, cpu_addr, size, &ret))
+		return ret;
+
+	if (off < nr_pages && nr_vma_pages <= (nr_pages - off)) {
+		ret = remap_pfn_range(vma, vma->vm_start,
+				      pfn + off,
+				      vma->vm_end - vma->vm_start,
+				      vma->vm_page_prot);
+	}
+
+	return ret;
+}
+
+static int arm64_swiotlb_mmap_noncoherent(struct device *dev,
+		struct vm_area_struct *vma,
+		void *cpu_addr, dma_addr_t dma_addr, size_t size,
+		struct dma_attrs *attrs)
+{
+	vma->vm_page_prot = pgprot_dmacoherent(vma->vm_page_prot);
+	return __dma_common_mmap(dev, vma, cpu_addr, dma_addr, size);
+}
+
+static int arm64_swiotlb_mmap_coherent(struct device *dev,
+		struct vm_area_struct *vma,
+		void *cpu_addr, dma_addr_t dma_addr, size_t size,
+		struct dma_attrs *attrs)
+{
+	/* Just use whatever page_prot attributes were specified */
+	return __dma_common_mmap(dev, vma, cpu_addr, dma_addr, size);
+}
+
 struct dma_map_ops noncoherent_swiotlb_dma_ops = {
 	.alloc = arm64_swiotlb_alloc_noncoherent,
 	.free = arm64_swiotlb_free_noncoherent,
+	.mmap = arm64_swiotlb_mmap_noncoherent,
 	.map_page = arm64_swiotlb_map_page,
 	.unmap_page = arm64_swiotlb_unmap_page,
 	.map_sg = arm64_swiotlb_map_sg_attrs,
@@ -213,6 +256,7 @@  EXPORT_SYMBOL(noncoherent_swiotlb_dma_ops);
 struct dma_map_ops coherent_swiotlb_dma_ops = {
 	.alloc = arm64_swiotlb_alloc_coherent,
 	.free = arm64_swiotlb_free_coherent,
+	.mmap = arm64_swiotlb_mmap_coherent,
 	.map_page = swiotlb_map_page,
 	.unmap_page = swiotlb_unmap_page,
 	.map_sg = swiotlb_map_sg_attrs,