lkml.org 
[lkml]   [2008]   [Aug]   [17]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH]: PCI: GART iommu alignment fixes [v2]

* FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> wrote:

> The above commit log of Prarit's patch is completely wrong (so I wrote
> this patch). To avoid misunderstanding, can you apply this patch with
> a proper description like this:

sure, i've created the commit below - is that fine?

Ingo

--------------->
From 3f18931be15856db01a216995670412e4421f3c9 Mon Sep 17 00:00:00 2001
From: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Date: Mon, 18 Aug 2008 00:36:18 +0900
Subject: [PATCH] x86 gart: allocate size-aligned address for alloc_coherent, v2

This patch changes GART IOMMU to return a size aligned address wrt
dma_alloc_coherent, as DMA-mapping.txt defines:

The cpu return address and the DMA bus master address are both
guaranteed to be aligned to the smallest PAGE_SIZE order which
is greater than or equal to the requested size. This invariant
exists (for example) to guarantee that if you allocate a chunk
which is smaller than or equal to 64 kilobytes, the extent of the
buffer you receive will not cross a 64K boundary.

But it is very unlikely that this matters. As DMA-mapping.txt
explains, This invariant is to avoid the boundary problem (such as
64K). Now the majority of IOMMUs including GART (except for Intel
IOMMU) don't allocate a buffer that crosses a 64K boundary wrt all the
DMA mapping interfaces (dma_alloc_coherent, dma_map_sg, and
dma_map_single) because of segment_boundary_mask in struct
device_dma_parameters.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
arch/x86/kernel/pci-gart_64.c | 25 ++++++++++++++++---------
1 files changed, 16 insertions(+), 9 deletions(-)

diff --git a/arch/x86/kernel/pci-gart_64.c b/arch/x86/kernel/pci-gart_64.c
index cdab678..4d8efb0 100644
--- a/arch/x86/kernel/pci-gart_64.c
+++ b/arch/x86/kernel/pci-gart_64.c
@@ -82,7 +82,8 @@ AGPEXTERN __u32 *agp_gatt_table;
static unsigned long next_bit; /* protected by iommu_bitmap_lock */
static int need_flush; /* global flush state. set for each gart wrap */

-static unsigned long alloc_iommu(struct device *dev, int size)
+static unsigned long alloc_iommu(struct device *dev, int size,
+ unsigned long align_mask)
{
unsigned long offset, flags;
unsigned long boundary_size;
@@ -95,11 +96,12 @@ static unsigned long alloc_iommu(struct device *dev, int size)

spin_lock_irqsave(&iommu_bitmap_lock, flags);
offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, next_bit,
- size, base_index, boundary_size, 0);
+ size, base_index, boundary_size, align_mask);
if (offset == -1) {
need_flush = 1;
offset = iommu_area_alloc(iommu_gart_bitmap, iommu_pages, 0,
- size, base_index, boundary_size, 0);
+ size, base_index, boundary_size,
+ align_mask);
}
if (offset != -1) {
next_bit = offset+size;
@@ -236,10 +238,10 @@ nonforced_iommu(struct device *dev, unsigned long addr, size_t size)
* Caller needs to check if the iommu is needed and flush.
*/
static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem,
- size_t size, int dir)
+ size_t size, int dir, unsigned long align_mask)
{
unsigned long npages = iommu_num_pages(phys_mem, size);
- unsigned long iommu_page = alloc_iommu(dev, npages);
+ unsigned long iommu_page = alloc_iommu(dev, npages, align_mask);
int i;

if (iommu_page == -1) {
@@ -262,7 +264,11 @@ static dma_addr_t dma_map_area(struct device *dev, dma_addr_t phys_mem,
static dma_addr_t
gart_map_simple(struct device *dev, phys_addr_t paddr, size_t size, int dir)
{
- dma_addr_t map = dma_map_area(dev, paddr, size, dir);
+ dma_addr_t map;
+ unsigned long align_mask;
+
+ align_mask = (1UL << get_order(size)) - 1;
+ map = dma_map_area(dev, paddr, size, dir, align_mask);

flush_gart();

@@ -281,7 +287,8 @@ gart_map_single(struct device *dev, phys_addr_t paddr, size_t size, int dir)
if (!need_iommu(dev, paddr, size))
return paddr;

- bus = gart_map_simple(dev, paddr, size, dir);
+ bus = dma_map_area(dev, paddr, size, dir, 0);
+ flush_gart();

return bus;
}
@@ -340,7 +347,7 @@ static int dma_map_sg_nonforce(struct device *dev, struct scatterlist *sg,
unsigned long addr = sg_phys(s);

if (nonforced_iommu(dev, addr, s->length)) {
- addr = dma_map_area(dev, addr, s->length, dir);
+ addr = dma_map_area(dev, addr, s->length, dir, 0);
if (addr == bad_dma_address) {
if (i > 0)
gart_unmap_sg(dev, sg, i, dir);
@@ -362,7 +369,7 @@ static int __dma_map_cont(struct device *dev, struct scatterlist *start,
int nelems, struct scatterlist *sout,
unsigned long pages)
{
- unsigned long iommu_start = alloc_iommu(dev, pages);
+ unsigned long iommu_start = alloc_iommu(dev, pages, 0);
unsigned long iommu_page = iommu_start;
struct scatterlist *s;
int i;

\
 
 \ /
  Last update: 2008-08-17 17:45    [W:1.601 / U:0.356 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site