> -----Original Message-----
> From: Konrad Rzeszutek Wilk [mailto:konrad.w...@oracle.com]
> Sent: Tuesday, September 15, 2015 2:40 PM
> To: William Davis
> Cc: Bjorn Helgaas; Alex Williamson; Joerg Roedel;
> iommu@lists.linux-foundation.org; linux-
> p...@vger.kernel.org; Mark Hounschell; David S. Miller; Jonathan Corbet;
> Terence Ripperda; John
> Hubbard; Jerome Glisse
> Subject: Re: [PATCH 11/22] swiotlb: Add map_peer_resource stub
>
> On Tue, Sep 15, 2015 at 12:10:56PM -0500, Will Davis wrote:
> > Add swiotlb stub of the 'map_peer_resource' DMA operation.
> >
> > Signed-off-by: Will Davis <wda...@nvidia.com>
> > ---
> > include/linux/swiotlb.h | 8 ++++++++
> > lib/swiotlb.c | 17 +++++++++++++++++
> > 2 files changed, 25 insertions(+)
> >
> > diff --git a/include/linux/swiotlb.h b/include/linux/swiotlb.h
> > index e7a018e..872463a 100644
> > --- a/include/linux/swiotlb.h
> > +++ b/include/linux/swiotlb.h
> > @@ -87,6 +87,14 @@ swiotlb_unmap_sg_attrs(struct device *hwdev, struct
> > scatterlist *sgl,
> > int nelems, enum dma_data_direction dir,
> > struct dma_attrs *attrs);
> >
> > +#ifdef CONFIG_HAS_DMA_P2P
> > +extern dma_peer_addr_t
> > +swiotlb_map_peer_resource(struct device *hwdev, struct device *hwpeer,
> > + struct resource *res, unsigned long offset,
> > + size_t size, enum dma_data_direction dir,
> > + struct dma_attrs *attrs);
> > +#endif
> > +
> > extern void
> > swiotlb_sync_single_for_cpu(struct device *hwdev, dma_addr_t dev_addr,
> > size_t size, enum dma_data_direction dir);
> > diff --git a/lib/swiotlb.c b/lib/swiotlb.c
> > index 76f29ec..61522a7 100644
> > --- a/lib/swiotlb.c
> > +++ b/lib/swiotlb.c
> > @@ -945,6 +945,23 @@ swiotlb_unmap_sg(struct device *hwdev, struct
> > scatterlist *sgl, int nelems,
> > }
> > EXPORT_SYMBOL(swiotlb_unmap_sg);
> >
> > +#ifdef CONFIG_HAS_DMA_P2P
> > +/*
> > + * Maps a single region from a peer device's resource for DMA streaming.
> > This
> > + * is currently unimplemented because there's no bounce buffer abstraction
> > for
> > + * peer mappings.
> > + */
> > +dma_peer_addr_t
> > +swiotlb_map_peer_resource(struct device *hwdev, struct device *hwpeer,
> > + struct resource *res, unsigned long offset,
> > + size_t size, enum dma_data_direction dir,
> > + struct dma_attrs *attrs)
> > +{
> > + return phys_to_dma(hwdev, io_tlb_overflow_buffer);
>
> That seems to be implemented :-)
>
> Did you mean to return DMA_ERROR_CODE instead?
I didn't see DMA_ERROR_CODE used in lib/swiotlb.c, and
swiotlb_dma_mapping_error is implemented as:
return (dma_addr == phys_to_dma(hwdev, io_tlb_overflow_buffer));
So I figured that phys_to_dma(hwdev, io_tlb_overflow_buffer) was intended
to be used as DMA_ERROR_CODE instead.
Thanks,
Will
> > +}
> > +EXPORT_SYMBOL(swiotlb_map_peer_resource);
> > +#endif
> > +
> > /*
> > * Make physical memory consistent for a set of streaming mode DMA
> > translations
> > * after a transfer.
> > --
> > 2.5.1
> >
--
nvpublic
_______________________________________________
iommu mailing list
iommu@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/iommu
