mem->dma_handle is a dma address obtained by dma_alloc_coherent which
needn't be a physical address as a hardware IOMMU can (and most
likely will) return a bus address where physical != bus address. So
ensure we are remapping (remap_pfn_range) the right page in
__videobuf_mmap_mapper by using virt_to_phys(mem->vaddr) and not
mem->dma_handle.
While at it, use PFN_DOWN instead of explicit shift to obtain a frame
number.
This was discovered by a random review of the code when looking for
something completely different. I'm not aware of any bug reports for
this.
However it is a bug because many v4l drivers use this layer and have
no idea whether IOMMU is in the system and running or not.
Signed-off-by: Jiri Slaby <[email protected]>
Cc: Mauro Carvalho Chehab <[email protected]>
Cc: Konrad Rzeszutek Wilk <[email protected]>
---
This is a version with updated changelog.
drivers/media/video/videobuf-dma-contig.c | 2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/drivers/media/video/videobuf-dma-contig.c b/drivers/media/video/videobuf-dma-contig.c
index c969111..19d3e4a 100644
--- a/drivers/media/video/videobuf-dma-contig.c
+++ b/drivers/media/video/videobuf-dma-contig.c
@@ -300,7 +300,7 @@ static int __videobuf_mmap_mapper(struct videobuf_queue *q,
vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
retval = remap_pfn_range(vma, vma->vm_start,
- mem->dma_handle >> PAGE_SHIFT,
+ PFN_DOWN(virt_to_phys(mem->vaddr))
size, vma->vm_page_prot);
if (retval) {
dev_err(q->dev, "mmap: remap failed with error %d. ", retval);
--
1.7.4.1
On 03/01/2011 09:21 AM, Jiri Slaby wrote:
> mem->dma_handle is a dma address obtained by dma_alloc_coherent which
> needn't be a physical address as a hardware IOMMU can (and most
> likely will) return a bus address where physical != bus address. So
> ensure we are remapping (remap_pfn_range) the right page in
> __videobuf_mmap_mapper by using virt_to_phys(mem->vaddr) and not
> mem->dma_handle.
>
> While at it, use PFN_DOWN instead of explicit shift to obtain a frame
> number.
>
> This was discovered by a random review of the code when looking for
> something completely different. I'm not aware of any bug reports for
> this.
>
> However it is a bug because many v4l drivers use this layer and have
> no idea whether IOMMU is in the system and running or not.
>
> Signed-off-by: Jiri Slaby <[email protected]>
> Cc: Mauro Carvalho Chehab <[email protected]>
> Cc: Konrad Rzeszutek Wilk <[email protected]>
Ah, this is rather:
Acked-by: Konrad Rzeszutek Wilk <[email protected]>
> ---
>
> This is a version with updated changelog.
>
> drivers/media/video/videobuf-dma-contig.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/media/video/videobuf-dma-contig.c b/drivers/media/video/videobuf-dma-contig.c
> index c969111..19d3e4a 100644
> --- a/drivers/media/video/videobuf-dma-contig.c
> +++ b/drivers/media/video/videobuf-dma-contig.c
> @@ -300,7 +300,7 @@ static int __videobuf_mmap_mapper(struct videobuf_queue *q,
>
> vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
> retval = remap_pfn_range(vma, vma->vm_start,
> - mem->dma_handle >> PAGE_SHIFT,
> + PFN_DOWN(virt_to_phys(mem->vaddr))
> size, vma->vm_page_prot);
> if (retval) {
> dev_err(q->dev, "mmap: remap failed with error %d. ", retval);
--
js