[PATCH] x86_64: Use function pointers to call DMA mapping functions

AK: I hacked Muli's original patch a lot and there were a lot
of changes - all bugs are probably to blame on me now.
There were also some changes in the fall back behaviour
for swiotlb - in particular it doesn't try to use GFP_DMA
now anymore. Also all DMA mapping operations use the
same core dma_alloc_coherent code with proper fallbacks now.
And various other changes and cleanups.

Known problems: iommu=force swiotlb=force together breaks
                needs more testing.

This patch cleans up x86_64's DMA mapping dispatching code. Right now
we have three possible IOMMU types: AGP GART, swiotlb and nommu, and
in the future we will also have Xen's x86_64 swiotlb and other HW
IOMMUs for x86_64. In order to support all of them cleanly, this
patch:

- introduces a struct dma_mapping_ops with function pointers for each
  of the DMA mapping operations of gart (AMD HW IOMMU), swiotlb
  (software IOMMU) and nommu (no IOMMU).

- gets rid of:

  if (swiotlb)
      return swiotlb_xxx();

- PCI_DMA_BUS_IS_PHYS is now checked against the dma_ops being set
This makes swiotlb faster by avoiding double copying in some cases.

Signed-Off-By: Muli Ben-Yehuda <mulix@mulix.org>
Signed-Off-By: Jon D. Mason <jdmason@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

1 files changed, 74 insertions, 70 deletions

diff --git a/arch/x86_64/kernel/pci-nommu.c b/arch/x86_64/kernel/pci-nommu.c
index 5a981dca87ff..3209adac3272 100644
--- a/arch/x86_64/kernel/pci-nommu.c
+++ b/arch/x86_64/kernel/pci-nommu.c
@@ -6,89 +6,93 @@
 #include <linux/string.h>
 #include <asm/proto.h>
 #include <asm/processor.h>
+#include <asm/dma.h>
 
-int iommu_merge = 0;
-EXPORT_SYMBOL(iommu_merge);
-
-dma_addr_t bad_dma_address;
-EXPORT_SYMBOL(bad_dma_address);
-
-int iommu_bio_merge = 0;
-EXPORT_SYMBOL(iommu_bio_merge);
-
-int iommu_sac_force = 0;
-EXPORT_SYMBOL(iommu_sac_force);
-
-/* 
- * Dummy IO MMU functions
- */
-
-void *dma_alloc_coherent(struct device *hwdev, size_t size,
-                         dma_addr_t *dma_handle, gfp_t gfp)
+static int
+check_addr(char *name, struct device *hwdev, dma_addr_t bus, size_t size)
 {
-        void *ret;
-        u64 mask;
-        int order = get_order(size);
-
-        if (hwdev)
-                mask = hwdev->coherent_dma_mask & *hwdev->dma_mask;
-        else
-                mask = 0xffffffff;
-        for (;;) {
-                ret = (void *)__get_free_pages(gfp, order);
-                if (ret == NULL)
-                        return NULL;
-                *dma_handle = virt_to_bus(ret);
-                if ((*dma_handle & ~mask) == 0)
-                        break;
-                free_pages((unsigned long)ret, order);
-                if (gfp & GFP_DMA)
-                        return NULL;
-                gfp |= GFP_DMA;
+        if (hwdev && bus + size > *hwdev->dma_mask) {
+                printk(KERN_ERR
+                    "nommu_%s: overflow %Lx+%lu of device mask %Lx\n",
+               name, (long long)bus, size, (long long)*hwdev->dma_mask);
+                return 0;
         }
+        return 1;
+}
 
-        memset(ret, 0, size);
-        return ret;
+static dma_addr_t
+nommu_map_single(struct device *hwdev, void *ptr, size_t size,
+               int direction)
+{
+        dma_addr_t bus = virt_to_bus(ptr);
+        if (!check_addr("map_single", hwdev, bus, size))
+                                return bad_dma_address;
+        return bus;
 }
-EXPORT_SYMBOL(dma_alloc_coherent);
 
-void dma_free_coherent(struct device *hwdev, size_t size,
-                         void *vaddr, dma_addr_t dma_handle)
+void nommu_unmap_single(struct device *dev, dma_addr_t addr,size_t size,
+                        int direction)
 {
-        free_pages((unsigned long)vaddr, get_order(size));
 }
-EXPORT_SYMBOL(dma_free_coherent);
 
-int dma_supported(struct device *hwdev, u64 mask)
+/* Map a set of buffers described by scatterlist in streaming
+ * mode for DMA.  This is the scatter-gather version of the
+ * above pci_map_single interface.  Here the scatter gather list
+ * elements are each tagged with the appropriate dma address
+ * and length.  They are obtained via sg_dma_{address,length}(SG).
+ *
+ * NOTE: An implementation may be able to use a smaller number of
+ *       DMA address/length pairs than there are SG table elements.
+ *       (for example via virtual mapping capabilities)
+ *       The routine returns the number of addr/length pairs actually
+ *       used, at most nents.
+ *
+ * Device ownership issues as mentioned above for pci_map_single are
+ * the same here.
+ */
+int nommu_map_sg(struct device *hwdev, struct scatterlist *sg,
+               int nents, int direction)
 {
-        /*
-         * we fall back to GFP_DMA when the mask isn't all 1s,
-         * so we can't guarantee allocations that must be
-         * within a tighter range than GFP_DMA..
-         * RED-PEN this won't work for pci_map_single. Caller has to
-         * use GFP_DMA in the first place.
-         */
-        if (mask < 0x00ffffff)
-                return 0;
+        int i;
 
-        return 1;
-} 
-EXPORT_SYMBOL(dma_supported);
+        BUG_ON(direction == DMA_NONE);
+        for (i = 0; i < nents; i++ ) {
+                struct scatterlist *s = &sg[i];
+                BUG_ON(!s->page);
+                s->dma_address = virt_to_bus(page_address(s->page) +s->offset);
+                if (!check_addr("map_sg", hwdev, s->dma_address, s->length))
+                        return 0;
+                s->dma_length = s->length;
+        }
+        return nents;
+}
 
-int dma_get_cache_alignment(void)
+/* Unmap a set of streaming mode DMA translations.
+ * Again, cpu read rules concerning calls here are the same as for
+ * pci_unmap_single() above.
+ */
+void nommu_unmap_sg(struct device *dev, struct scatterlist *sg,
+                  int nents, int dir)
 {
-        return boot_cpu_data.x86_clflush_size;
 }
-EXPORT_SYMBOL(dma_get_cache_alignment);
 
-static int __init check_ram(void) 
-{ 
-        if (end_pfn >= 0xffffffff>>PAGE_SHIFT) { 
-                printk(
-                KERN_ERR "WARNING more than 4GB of memory but IOMMU not compiled in.\n"
-                KERN_ERR "WARNING 32bit PCI may malfunction.\n");
-        } 
-        return 0;
-} 
-__initcall(check_ram);
+struct dma_mapping_ops nommu_dma_ops = {
+        .map_single = nommu_map_single,
+        .unmap_single = nommu_unmap_single,
+        .map_sg = nommu_map_sg,
+        .unmap_sg = nommu_unmap_sg,
+        .is_phys = 1,
+};
 
+void __init no_iommu_init(void)
+{
+        if (dma_ops)
+                return;
+        printk(KERN_INFO "PCI-DMA: Disabling IOMMU.\n");
+        dma_ops = &nommu_dma_ops;
+        if (end_pfn > MAX_DMA32_PFN) {
+                printk(KERN_ERR
+                       "WARNING more than 4GB of memory but IOMMU disabled.\n"
+                       KERN_ERR "WARNING 32bit PCI may malfunction.\n");
+        }
+}