Re: [PATCH] kmemcheck: add hooks for page- and sg-dma-mappings

[Pekka Enberg <penberg@cs.helsinki.fi>]

Vegard Nossum wrote:
> This is needed for page allocator support to prevent false positives
> when accessing pages which are dma-mapped.
> 
> Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
> ---
>  arch/x86/include/asm/dma-mapping.h |    6 ++++++
>  1 files changed, 6 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/x86/include/asm/dma-mapping.h b/arch/x86/include/asm/dma-mapping.h
> index 830bb0e..713a002 100644
> --- a/arch/x86/include/asm/dma-mapping.h
> +++ b/arch/x86/include/asm/dma-mapping.h
> @@ -117,7 +117,12 @@ dma_map_sg(struct device *hwdev, struct scatterlist *sg,
>  {
>  	struct dma_mapping_ops *ops = get_dma_ops(hwdev);
>  
> +	struct scatterlist *s;
> +	int i;
> +
>  	BUG_ON(!valid_dma_direction(direction));
> +	for_each_sg(sg, s, nents, i)
> +		kmemcheck_mark_initialized(sg_virt(s), s->length);
>  	return ops->map_sg(hwdev, sg, nents, direction);
>  }
>  
> @@ -215,6 +220,7 @@ static inline dma_addr_t dma_map_page(struct device *dev, struct page *page,
>  	struct dma_mapping_ops *ops = get_dma_ops(dev);
>  
>  	BUG_ON(!valid_dma_direction(direction));
> +	kmemcheck_mark_initialized(page_address(page) + offset, size);
>  	return ops->map_single(dev, page_to_phys(page) + offset,
>  			       size, direction);
>  }

What's with the new BUG_ON() calls here?