[PATCH v6 15/19] swiotlb-xen: call dma_capable only if dev->dma_mask is allocated

[Rob Herring]

On Wed, Oct 2, 2013 at 12:22 PM, Konrad Rzeszutek Wilk
<konrad.wilk at oracle.com> wrote:
> On Wed, Oct 02, 2013 at 06:13:35PM +0100, Stefano Stabellini wrote:
>> The real issue is that some devices (xgmac, I am looking at you), don't
>> set the dma_mask, even though they are perfectly capable of doing dma.
>
> So this looks like a bug in the drivers. I thought I saw a huge patchset
> by the ARM maintainer that tries to fix the dma_mask and dma_mask_coherent
> bugs? And also fix the drivers to use the dma mask?

I think Russell only fixed incorrect handling of masks, not missing handling.

>>
>> Maybe I should modify the arm implementation of dma_capable (see
>> http://marc.info/?l=linux-kernel&m=138029832007821&w=2) to ignore the
>> dma_mask.
>
> Or fix the 'xgmac'?

There's really some core changes needed to fix this and I'd guess
there are lots of other cases of missing dma_mask. The problem is
platform devices don't create a mask to assign to dma_mask in the
first place. I think the right solution is assigning dma_mask to
&coherent_dma_mask by default as Russell did for AMBA devices. I don't
know if doing that would break anything or not.

Rob

>>
>> On Mon, 30 Sep 2013, Konrad Rzeszutek Wilk wrote:
>> > On Fri, Sep 27, 2013 at 05:10:03PM +0100, Stefano Stabellini wrote:
>> >
>> > Why? I am looking at X86 and IA64 and I see:
>> >
>> >  79         if (!dev->dma_mask)
>> >  80                 return 0;
>> >
>> >
>> > Why not port dma_capable over to ARM?
>> >
>> > > Signed-off-by: Stefano Stabellini <stefano.stabellini at eu.citrix.com>
>> > > ---
>> > >  drivers/xen/swiotlb-xen.c |    6 +++---
>> > >  1 files changed, 3 insertions(+), 3 deletions(-)
>> > >
>> > > diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c
>> > > index 790c2eb..3011736 100644
>> > > --- a/drivers/xen/swiotlb-xen.c
>> > > +++ b/drivers/xen/swiotlb-xen.c
>> > > @@ -512,7 +512,7 @@ dma_addr_t xen_swiotlb_map_page(struct device *dev, struct page *page,
>> > >    * buffering it.
>> > >    */
>> > >   if (!xen_feature(XENFEAT_auto_translated_physmap) &&
>> > > -     dma_capable(dev, dev_addr, size) &&
>> > > +     dev->dma_mask && dma_capable(dev, dev_addr, size) &&
>> > >       !range_straddles_page_boundary(phys, size) && !swiotlb_force)
>> > >           return dev_addr;
>> > >
>> > > @@ -532,7 +532,7 @@ dma_addr_t xen_swiotlb_map_page(struct device *dev, struct page *page,
>> > >   /*
>> > >    * Ensure that the address returned is DMA'ble
>> > >    */
>> > > - if (!dma_capable(dev, dev_addr, size)) {
>> > > + if (dev->dma_mask && !dma_capable(dev, dev_addr, size)) {
>> > >           swiotlb_tbl_unmap_single(dev, map, size, dir);
>> > >           dev_addr = 0;
>> > >   }
>> > > @@ -660,7 +660,7 @@ xen_swiotlb_map_sg_attrs(struct device *hwdev, struct scatterlist *sgl,
>> > >
>> > >           if (swiotlb_force ||
>> > >               xen_feature(XENFEAT_auto_translated_physmap) ||
>> > > -             !dma_capable(hwdev, dev_addr, sg->length) ||
>> > > +             (hwdev->dma_mask && !dma_capable(hwdev, dev_addr, sg->length)) ||
>> > >               range_straddles_page_boundary(paddr, sg->length)) {
>> > >                   /*
>> > >                    * Pass the dma_addr of the first slab in the iotlb buffer as
>> > > --
>> > > 1.7.2.5
>> > >
>> >
>
> _______________________________________________
> linux-arm-kernel mailing list
> linux-arm-kernel at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/linux-arm-kernel