> This fixes the regression of the
> 6fee48cd330c68332f9712bc968d934a1a84a32a commit.
>
> After the above commit, pci_set_consistent_dma_mask doesn't work for
> dmabounce devices. It converts ARM to use the generic implementation
> of pci_set_consistent_dma_mask.
>
> The root cause that ARM's CONFIG_DMABOUNCE uses dev->coherent_dma_mask
> for the dma mask of a device _AND_ a bus. dev->coherent_dma_mask is
> supposed to represent the dma mask of a device.
>
> The proper solution is that device and bus has the own dma mask
> respectively (POWERPC does the similar). But the fix must go to stable
> trees so this simple (and hacky a bit) patch works better for now.
>
> For further information:
http://lkml.org/lkml/2010/8/10/272
>
> Reported-by: Krzysztof Halasa <khc@pm.waw.pl>
> Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
> Tested-by: Krzysztof Halasa <khc@pm.waw.pl>
> Cc:
stable@kernel.org
> ---
> arch/arm/mm/dma-mapping.c | 5 +++++
> 1 files changed, 5 insertions(+), 0 deletions(-)
>
> diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c
> index c704eed..2a3fc2e 100644
> --- a/arch/arm/mm/dma-mapping.c
> +++ b/arch/arm/mm/dma-mapping.c
> @@ -77,6 +77,11 @@ static struct page *__dma_alloc_buffer(struct device *dev, size_t size, gfp_t gf
> if (mask < 0xffffffffULL)
> gfp |= GFP_DMA;
>
> +#ifdef CONFIG_DMABOUNCE
> + if (dev->archdata.dmabounce)
> + gfp |= GFP_DMA;
> +#endif
> +
> page = alloc_pages(gfp, order);
> if (!page)
> return NULL;
