Description of problem:
dma_map_ops.alloc_coherent should be using the coherent_dma_mask, intel_alloc_coherent()
is not.  intel_alloc_coherent() calls __intel_map_single() passing in the correct mask,
hwdev->coherent_dma_mask, but then __intel_map_single() doesn't use it... it ends up 
using the pdev->dma_mask:


    static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
                                         size_t size, int dir, u64 dma_mask)
    {
    ...txt deleted...
            iova = intel_alloc_iova(hwdev, domain, dma_to_mm_pfn(size),
                                    pdev->dma_mask);
                                    ^^^^^^^^^^^^^^

I *think* this call to intel_alloc_iova() should be using the "dma_mask"
passed into it.  We should look into why the code is using pdev->dma_mask
rather than the hwdev->coherent_dma_mask on a dma_map_ops.alloc_coherent()
call other dma_map_ops alloc_coherent routines do.

...

Hopefully final version submitted:

http://marc.info/?l=linux-kernel&m=130220593715364&w=4

Subject: Intel pci: Use coherent DMA mask when requested

    The __intel_map_single function is not honoring the passed in
    DMA mask.  This results in not using the coherent DMA mask when
    called from intel_alloc_coherent().

Signed-off-by: Mike Travis <travis>
Reviewed-by: Mike Habeck <habeck>


Version-Release number of selected component (if applicable):


How reproducible:


Steps to Reproduce:
1.
2.
3.
  
Actual results:


Expected results:


Additional info: