This patchset adds a trivial helper function, is_buffer_dma_capable()
and cleans up some IOMMU implementations. This function is to to see
if a memory region is DMA-capable or not with the arguments, the
dma_mask (or coherent_dma_mask) of a device and the address and size
of a memory region. It's useful for IOMMUs that don't do virtual
mappings at all times.

The patch for swiotlb is a bug fix (the rest are just
cleanups). Currently, swiotlb doesn't take account of the size of a
memory region to see if if the memory region is DMA-capable.

=
 arch/x86/kernel/pci-gart_64.c |   16 +++-------------
 arch/x86/kernel/pci-nommu.c   |   10 ++++++----
 include/linux/dma-mapping.h   |    5 +++++
 lib/swiotlb.c                 |   15 ++++++++-------
 4 files changed, 22 insertions(+), 24 deletions(-)


--

[ message continues ]

is_buffer_dma_capable helper function is to see if a memory region is
DMA-capable or not. The arugments are the dma_mask (or
coherent_dma_mask) of a device and the address and size of a memory
region.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Joerg Roedel <joerg.roedel@amd.com>
---
 include/linux/dma-mapping.h |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)

diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index 952e0f8..6ed50c1 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -48,6 +48,11 @@ static inline int is_device_dma_capable(struct device *dev)
 	return dev->dma_mask != NULL && *dev->dma_mask != DMA_MASK_NONE;
 }
 
+static inline int is_buffer_dma_capable(u64 mask, dma_addr_t addr, size_t size)
+{
+	return addr + size <= mask;
+}
+
 #ifdef CONFIG_HAS_DMA
 #include <asm/dma-mapping.h>
 #else
-- 
1.5.4.2

--

[ message continues ]

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/kernel/pci-gart_64.c |   16 +++-------------
 1 files changed, 3 insertions(+), 13 deletions(-)

diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index 0b99d4a..1b0c412 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -214,24 +214,14 @@ static void iommu_full(struct device *dev, size_t size, int dir)
 static inline int
 need_iommu(struct device *dev, unsigned long addr, size_t size)
 {
-	u64 mask = *dev->dma_mask;
-	int high = addr + size > mask;
-	int mmu = high;
-
-	if (force_iommu)
-		mmu = 1;
-
-	return mmu;
+	return force_iommu ||
+		!is_buffer_dma_capable(*dev->dma_mask, addr, size);
 }
 
 static inline int
 nonforced_iommu(struct device *dev, unsigned long addr, size_t size)
 {
-	u64 mask = *dev->dma_mask;
-	int high = addr + size > mask;
-	int mmu = high;
-
-	return mmu;
+	return !is_buffer_dma_capable(*dev->dma_mask, addr, size);
 }
 
 /* Map a single continuous physical area into the IOMMU.
-- 
1.5.4.2

--

[ message continues ]

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
---
 arch/x86/kernel/pci-nommu.c |   10 ++++++----
 1 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kernel/pci-nommu.c b/arch/x86/kernel/pci-nommu.c
index 8e398b5..1c1c98a 100644
--- a/arch/x86/kernel/pci-nommu.c
+++ b/arch/x86/kernel/pci-nommu.c
@@ -14,7 +14,7 @@
 static int
 check_addr(char *name, struct device *hwdev, dma_addr_t bus, size_t size)
 {
-	if (hwdev && bus + size > *hwdev->dma_mask) {
+	if (hwdev && !is_buffer_dma_capable(*hwdev->dma_mask, bus, size)) {
 		if (*hwdev->dma_mask >= DMA_32BIT_MASK)
 			printk(KERN_ERR
 			    "nommu_%s: overflow %Lx+%zu of device mask %Lx\n",
@@ -79,6 +79,7 @@ nommu_alloc_coherent(struct device *hwdev, size_t size,
 	unsigned long dma_mask;
 	int node;
 	struct page *page;
+	dma_addr_t addr;
 
 	dma_mask = dma_alloc_coherent_mask(hwdev, gfp);
 
@@ -90,14 +91,15 @@ again:
 	if (!page)
 		return NULL;
 
-	if ((page_to_phys(page) + size > dma_mask) && !(gfp & GFP_DMA)) {
+	addr = page_to_phys(page);
+	if (!is_buffer_dma_capable(dma_mask, addr, size) && !(gfp & GFP_DMA)) {
 		free_pages((unsigned long)page_address(page), get_order(size));
 		gfp |= GFP_DMA;
 		goto again;
 	}
 
-	*dma_addr = page_to_phys(page);
-	if (check_addr("alloc_coherent", hwdev, *dma_addr, size)) {
+	if (check_addr("alloc_coherent", hwdev, addr, size)) {
+		*dma_addr = addr;
 		flush_write_buffers();
 		return page_address(page);
 	}
-- 
1.5.4.2

--

[ message continues ]

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Tony Luck <tony.luck@intel.com>
---
 lib/swiotlb.c |   15 ++++++++-------
 1 files changed, 8 insertions(+), 7 deletions(-)

diff --git a/lib/swiotlb.c b/lib/swiotlb.c
index b5f5d11..240a67c 100644
--- a/lib/swiotlb.c
+++ b/lib/swiotlb.c
@@ -274,13 +274,13 @@ cleanup1:
 }
 
 static int
-address_needs_mapping(struct device *hwdev, dma_addr_t addr)
+address_needs_mapping(struct device *hwdev, dma_addr_t addr, size_t size)
 {
 	dma_addr_t mask = 0xffffffff;
 	/* If the device has a mask, use it, otherwise default to 32 bits */
 	if (hwdev && hwdev->dma_mask)
 		mask = *hwdev->dma_mask;
-	return (addr & ~mask) != 0;
+	return !is_buffer_dma_capable(mask, addr, size);
 }
 
 static int is_swiotlb_buffer(char *addr)
@@ -473,7 +473,7 @@ swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 	int order = get_order(size);
 
 	ret = (void *)__get_free_pages(flags, order);
-	if (ret && address_needs_mapping(hwdev, virt_to_bus(ret))) {
+	if (ret && address_needs_mapping(hwdev, virt_to_bus(ret), size)) {
 		/*
 		 * The allocated memory isn't reachable by the device.
 		 * Fall back on swiotlb_map_single().
@@ -497,7 +497,7 @@ swiotlb_alloc_coherent(struct device *hwdev, size_t size,
 	dev_addr = virt_to_bus(ret);
 
 	/* Confirm address can be DMA'd by device */
-	if (address_needs_mapping(hwdev, dev_addr)) {
+	if (address_needs_mapping(hwdev, dev_addr, size)) {
 		printk("hwdev DMA mask = 0x%016Lx, dev_addr = 0x%016Lx\n",
 		       (unsigned long long)*hwdev->dma_mask,
 		       (unsigned long long)dev_addr);
@@ -561,7 +561,7 @@ swiotlb_map_single_attrs(struct device *hwdev, void *ptr, size_t size,
 	 * we can safely return the device addr and not worry about bounce
 	 * buffering it.
 	 */
-	if (!address_needs_mapping(hwdev, dev_addr) && !swiotlb_force)
+	if (!address_needs_mapping(hwdev, dev_addr, size) && !swiotlb_force)
 		return dev_addr;
 
 	/*
@@ -578,7 +578,7 @@ swiotlb_map_single_attrs(struct ...
[ message continues ]

+static inline int is_buffer_dma_capable(u64 mask, dma_addr_t addr, size_t size)
+{
+       return addr + size <= mask;
+}

Do we care about wrap-around (e.g. addr=0xffffffffffffffff size=2)?

-Tony
--

[ message continues ]

I don't think so. The current code does not care about it and if it
happens its a bug in the iommu implementation.

Joerg

-- 
           |           AMD Saxony Limited Liability Company & Co. KG
 Operating |         Wilschdorfer Landstr. 101, 01109 Dresden, Germany
 System    |                  Register Court Dresden: HRA 4896
 Research  |              General Partner authorized to represent:
 Center    |             AMD Saxony LLC (Wilmington, Delaware, US)
           | General Manager of AMD Saxony LLC: Dr. Hans-R. Deppe, Thomas McCoy

--

[ message continues ]

On Tue, 9 Sep 2008 09:17:05 -0700

We could do something like:

      return addr < mask && addr + size <= mask;


But such wrap-around can happen only with a buggy IOMMU
implementation. A driver will does DMA from 0xffffffffffffffff to
0x1. Even if this function does the right thing in such case, we will
be dead?
--

[ message continues ]

Good idea. This patchset increases the readability of the software
dma-api implementations.

Acked-by: Joerg Roedel <joerg.roedel@amd.com>

-- 
           |           AMD Saxony Limited Liability Company & Co. KG
 Operating |         Wilschdorfer Landstr. 101, 01109 Dresden, Germany
 System    |                  Register Court Dresden: HRA 4896
 Research  |              General Partner authorized to represent:
 Center    |             AMD Saxony LLC (Wilmington, Delaware, US)
           | General Manager of AMD Saxony LLC: Dr. Hans-R. Deppe, Thomas McCoy

--

[ message continues ]

applied to tip/x86/iommu:

 2797982: swiotlb: convert swiotlb to use is_buffer_dma_capable helper function
 49fbf4e: x86: convert pci-nommu to use is_buffer_dma_capable helper function
 ac4ff65: x86: convert gart to use is_buffer_dma_capable helper function
 636dc67: add is_buffer_dma_capable helper function

thanks!

	Ingo
--

[ message continues ]