tile pci: enable IOMMU to support DMA for legacy devices

This change uses the TRIO IOMMU to map the PCI DMA space and physical
memory at different addresses.  We also now use the dma_mapping_ops
to provide support for non-PCI DMA, PCIe DMA (64-bit) and legacy PCI
DMA (32-bit).  We use the kernel's software I/O TLB framework
(i.e. bounce buffers) for the legacy 32-bit PCI device support since
there are a limited number of TLB entries in the IOMMU and it is
non-trivial to handle indexing, searching, matching, etc.  For 32-bit
devices the performance impact of bounce buffers should not be a concern.

Signed-off-by: Chris Metcalf <cmetcalf@tilera.com>

1 files changed, 98 insertions, 48 deletions

diff --git a/arch/tile/include/asm/dma-mapping.h b/arch/tile/include/asm/dma-mapping.h
index eaa06d175b39..4b6247d1a315 100644
--- a/arch/tile/include/asm/dma-mapping.h
+++ b/arch/tile/include/asm/dma-mapping.h
@@ -20,69 +20,80 @@
 #include <linux/cache.h>
 #include <linux/io.h>
 
-/*
- * Note that on x86 and powerpc, there is a "struct dma_mapping_ops"
- * that is used for all the DMA operations.  For now, we don't have an
- * equivalent on tile, because we only have a single way of doing DMA.
- * (Tilera bug 7994 to use dma_mapping_ops.)
- */
+extern struct dma_map_ops *tile_dma_map_ops;
+extern struct dma_map_ops *gx_pci_dma_map_ops;
+extern struct dma_map_ops *gx_legacy_pci_dma_map_ops;
+
+static inline struct dma_map_ops *get_dma_ops(struct device *dev)
+{
+	if (dev && dev->archdata.dma_ops)
+		return dev->archdata.dma_ops;
+	else
+		return tile_dma_map_ops;
+}
+
+static inline dma_addr_t get_dma_offset(struct device *dev)
+{
+	return dev->archdata.dma_offset;
+}
+
+static inline void set_dma_offset(struct device *dev, dma_addr_t off)
+{
+	dev->archdata.dma_offset = off;
+}
 
-#define dma_alloc_noncoherent(d, s, h, f) dma_alloc_coherent(d, s, h, f)
-#define dma_free_noncoherent(d, s, v, h) dma_free_coherent(d, s, v, h)
-
-extern dma_addr_t dma_map_single(struct device *dev, void *ptr, size_t size,
-			  enum dma_data_direction);
-extern void dma_unmap_single(struct device *dev, dma_addr_t dma_addr,
-			     size_t size, enum dma_data_direction);
-extern int dma_map_sg(struct device *dev, struct scatterlist *sg, int nents,
-	       enum dma_data_direction);
-extern void dma_unmap_sg(struct device *dev, struct scatterlist *sg,
-			 int nhwentries, enum dma_data_direction);
-extern dma_addr_t dma_map_page(struct device *dev, struct page *page,
-			       unsigned long offset, size_t size,
-			       enum dma_data_direction);
-extern void dma_unmap_page(struct device *dev, dma_addr_t dma_address,
-			   size_t size, enum dma_data_direction);
-extern void dma_sync_sg_for_cpu(struct device *dev, struct scatterlist *sg,
-				int nelems, enum dma_data_direction);
-extern void dma_sync_sg_for_device(struct device *dev, struct scatterlist *sg,
-				   int nelems, enum dma_data_direction);
-
-
-void *dma_alloc_coherent(struct device *dev, size_t size,
-			   dma_addr_t *dma_handle, gfp_t flag);
-
-void dma_free_coherent(struct device *dev, size_t size,
-			 void *vaddr, dma_addr_t dma_handle);
-
-extern void dma_sync_single_for_cpu(struct device *, dma_addr_t, size_t,
-				    enum dma_data_direction);
-extern void dma_sync_single_for_device(struct device *, dma_addr_t,
-				       size_t, enum dma_data_direction);
-extern void dma_sync_single_range_for_cpu(struct device *, dma_addr_t,
-					  unsigned long offset, size_t,
-					  enum dma_data_direction);
-extern void dma_sync_single_range_for_device(struct device *, dma_addr_t,
-					     unsigned long offset, size_t,
-					     enum dma_data_direction);
-extern void dma_cache_sync(struct device *dev, void *vaddr, size_t,
-			   enum dma_data_direction);
+static inline dma_addr_t phys_to_dma(struct device *dev, phys_addr_t paddr)
+{
+	return paddr + get_dma_offset(dev);
+}
+
+static inline phys_addr_t dma_to_phys(struct device *dev, dma_addr_t daddr)
+{
+	return daddr - get_dma_offset(dev);
+}
+
+static inline void dma_mark_clean(void *addr, size_t size) {}
+
+#include <asm-generic/dma-mapping-common.h>
+
+static inline void set_dma_ops(struct device *dev, struct dma_map_ops *ops)
+{
+	dev->archdata.dma_ops = ops;
+}
+
+static inline bool dma_capable(struct device *dev, dma_addr_t addr, size_t size)
+{
+	if (!dev->dma_mask)
+		return 0;
+
+	return addr + size - 1 <= *dev->dma_mask;
+}
 
 static inline int
 dma_mapping_error(struct device *dev, dma_addr_t dma_addr)
 {
-	return 0;
+	return get_dma_ops(dev)->mapping_error(dev, dma_addr);
 }
 
 static inline int
 dma_supported(struct device *dev, u64 mask)
 {
-	return 1;
+	return get_dma_ops(dev)->dma_supported(dev, mask);
 }
 
 static inline int
 dma_set_mask(struct device *dev, u64 mask)
 {
+	struct dma_map_ops *dma_ops = get_dma_ops(dev);
+
+	/* Handle legacy PCI devices with limited memory addressability. */
+	if ((dma_ops == gx_pci_dma_map_ops) && (mask <= DMA_BIT_MASK(32))) {
+		set_dma_ops(dev, gx_legacy_pci_dma_map_ops);
+		set_dma_offset(dev, 0);
+		if (mask > dev->archdata.max_direct_dma_addr)
+			mask = dev->archdata.max_direct_dma_addr;
+	}
+
 	if (!dev->dma_mask || !dma_supported(dev, mask))
 		return -EIO;
 
@@ -91,4 +102,43 @@ dma_set_mask(struct device *dev, u64 mask)
 	return 0;
 }
 
+static inline void *dma_alloc_attrs(struct device *dev, size_t size,
+				    dma_addr_t *dma_handle, gfp_t flag,
+				    struct dma_attrs *attrs)
+{
+	struct dma_map_ops *dma_ops = get_dma_ops(dev);
+	void *cpu_addr;
+
+	cpu_addr = dma_ops->alloc(dev, size, dma_handle, flag, attrs);
+
+	debug_dma_alloc_coherent(dev, size, *dma_handle, cpu_addr);
+
+	return cpu_addr;
+}
+
+static inline void dma_free_attrs(struct device *dev, size_t size,
+				  void *cpu_addr, dma_addr_t dma_handle,
+				  struct dma_attrs *attrs)
+{
+	struct dma_map_ops *dma_ops = get_dma_ops(dev);
+
+	debug_dma_free_coherent(dev, size, cpu_addr, dma_handle);
+
+	dma_ops->free(dev, size, cpu_addr, dma_handle, attrs);
+}
+
+#define dma_alloc_coherent(d, s, h, f) dma_alloc_attrs(d, s, h, f, NULL)
+#define dma_alloc_noncoherent(d, s, h, f) dma_alloc_attrs(d, s, h, f, NULL)
+#define dma_free_coherent(d, s, v, h) dma_free_attrs(d, s, v, h, NULL)
+#define dma_free_noncoherent(d, s, v, h) dma_free_attrs(d, s, v, h, NULL)
+
+/*
+ * dma_alloc_noncoherent() is #defined to return coherent memory,
+ * so there's no need to do any flushing here.
+ */
+static inline void dma_cache_sync(struct device *dev, void *vaddr, size_t size,
+				  enum dma_data_direction direction)
+{
+}
+
 #endif /* _ASM_TILE_DMA_MAPPING_H */