Change all ia64 machvecs to use the new dma_{un}map_*_attrs()

interfaces. Implement the old dma_{un}map_*() interfaces in

terms of the corresponding new interfaces. For ia64/sn, make

use of one dma attribute, DMA_ATTR_BARRIER.
Signed-off-by: Arthur Kepner <akepner@sgi.com>
--
 arch/ia64/hp/common/hwsw_iommu.c         |   60 ++++++++++++------------

 arch/ia64/hp/common/sba_iommu.c          |   62 ++++++++++++++----------

 arch/ia64/sn/pci/pci_dma.c               |   77 ++++++++++++++++++++-----------

 include/asm-ia64/dma-mapping.h           |   28 +++++++++--

 include/asm-ia64/machvec.h               |   52 ++++++++++++--------

 include/asm-ia64/machvec_hpzx1.h         |   16 +++---

 include/asm-ia64/machvec_hpzx1_swiotlb.h |   16 +++---

 include/asm-ia64/machvec_sn2.h           |   16 +++---

 include/linux/dma-attrs.h                |   48 +++++++++++++++++++

 lib/swiotlb.c                            |   50 ++++++++++++++++----

 10 files changed, 289 insertions(+), 136 deletions(-)
diff --git a/arch/ia64/hp/common/hwsw_iommu.c b/arch/ia64/hp/common/hwsw_iommu.c

index 94e5710..8cedd6c 100644

--- a/arch/ia64/hp/common/hwsw_iommu.c

+++ b/arch/ia64/hp/common/hwsw_iommu.c

@@ -20,10 +20,10 @@

 extern int swiotlb_late_init_with_default_size (size_t size);

 extern ia64_mv_dma_alloc_coherent	swiotlb_alloc_coherent;

 extern ia64_mv_dma_free_coherent	swiotlb_free_coherent;

-extern ia64_mv_dma_map_single		swiotlb_map_single;

-extern ia64_mv_dma_unmap_single		swiotlb_unmap_single;

-extern ia64_mv_dma_map_sg		swiotlb_map_sg;

-extern ia64_mv_dma_unmap_sg		swiotlb_unmap_sg;

+extern ia64_mv_dma_map_single_attrs	swiotlb_map_single_attrs;

+extern ia64_mv_dma_unmap_single_attrs	swiotlb_unmap_single_attrs;

+extern ia64_mv_dma_map_sg_attrs		swiotlb_map_sg_attrs;

+extern ia64_mv_dma_unmap_sg_attrs	swiotlb_unmap_sg_attrs;

 extern ia64_mv_dma_supported		swiotlb_dma_supported;

 extern ia64_mv_dma_mapping_error	swiotlb_dma_mapping_error;
@@ -31,19 +31,19 @@ extern ia64_mv_dma_mapp...
[ message continues ]

In general, the patches look reasonable to me.  Just an observation:
The attribute names (DMA_ATTR_...) are going to have to live somewhere

outside of the #ifdef ARCH_USES_DMA_ATTRS otherwise we'll get compile

failures of drivers using attributes on architectures that don't support

them.
Secondly, DMA_ATTR_BARRIER doesn't quite sound right.  What you're

actually doing is trying to prescribe strict ordering, so shouldn't this

be something like DMA_ATTR_STRICT_ORDERING (and perhaps with a

corresponding DMA_ATTR_RELAXED_ORDERING for the PCIe case).  also,

strike the DMA_ATTR_FOO and DMA_ATTR_GOO since they have no plausible

meaning.
James
--

Yeah, I realized only after sending that I'd forgotten to

remove these. 
--

Arthur
--