Re: [media] V4L: videobuf, don't use dma addr as physical

[Benjamin Herrenschmidt]

On Thu, 2011-03-24 at 17:04 +0000, Linux Kernel Mailing List wrote:
> Gitweb:     http://git.kernel.org/linus/35d9f510b67b10338161aba6229d4f55b4000f5b
> Commit:     35d9f510b67b10338161aba6229d4f55b4000f5b
> Parent:     4093a5c4a3f59cba1a085bbf87b6ffdddc5a443d
> Author:     Jiri Slaby <jslaby@xxxxxxx>
> AuthorDate: Mon Feb 28 06:37:02 2011 -0300
> Committer:  Mauro Carvalho Chehab <mchehab@xxxxxxxxxx>
> CommitDate: Tue Mar 22 06:51:54 2011 -0300
> 
>     [media] V4L: videobuf, don't use dma addr as physical
>     
>     mem->dma_handle is a dma address obtained by dma_alloc_coherent which
>     needn't be a physical address in presence of IOMMU, 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.

Sadly that doesn't work on all archs either... some will return
addresses in the "vmalloc" space from dma_alloc_coherent() (due to the
need to create non-cachable mappings there) and you can't pass those
through virt_to_phys() either.

I'm sad that thing got merged at all. This whole videobuf2 looks like an
unreviewed pile of crap to me. It re-invents dma pool management,
re-invents even kref's, etc...

As for mmap'ing DMA, there's an attempt at fixing that done by the Alsa
folks in the form of dma_mmap_coherent(). (See sound/core/pcm_native.c).

Only ARM implements it now. We need to move the

#define ARCH_HAS_DMA_MMAP_COHERENT

Out of the sound code into an ARM header tho, that way you can use it
too, and I shall implement a variant for non-coherent embedded PPCs
while at it (I'll take care of that last part).

Ben.

>     [mchehab@xxxxxxxxxx: Fix compilation breakage due to the lack of a comma]
>     Signed-off-by: Jiri Slaby <jslaby@xxxxxxx>
>     Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
>     Signed-off-by: Mauro Carvalho Chehab <mchehab@xxxxxxxxxx>
> ---
>  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..c4742fc 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);
> --
> To unsubscribe from this list: send the line "unsubscribe git-commits-head" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/