lkml.org 
[lkml]   [2016]   [Mar]   [28]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
From
Subject[PATCH v2 1/2] dma/iommu: Add pgsize_bitmap confirmation in __iommu_dma_alloc_pages
Date
Currently __iommu_dma_alloc_pages assumes that all the IOMMU support
the granule of PAGE_SIZE. It call alloc_page to try allocating memory
in the last time. Fortunately the mininum pagesize in all the
current IOMMU is SZ_4K, so this works well.

But there may be a case in which the mininum granule of IOMMU may be
larger than PAGE_SIZE, then it will abort as the IOMMU cann't map the
discontinuous memory within a granule. For example, the pgsize_bitmap
of the IOMMU only has SZ_16K while the PAGE_SIZE is SZ_4K, then we
have to prepare SZ_16K continuous memory at least for each a granule
iommu mapping.

Signed-off-by: Yong Wu <yong.wu@mediatek.com>

---
v2:
-rebase on v4.6-rc1.
-add a new patch([1/2] add pgsize_bitmap) here.

drivers/iommu/dma-iommu.c | 22 +++++++++++++---------
1 file changed, 13 insertions(+), 9 deletions(-)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 72d6182..75ce71e 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -190,11 +190,13 @@ static void __iommu_dma_free_pages(struct page **pages, int count)
kvfree(pages);
}

-static struct page **__iommu_dma_alloc_pages(unsigned int count, gfp_t gfp)
+static struct page **__iommu_dma_alloc_pages(unsigned int count, gfp_t gfp,
+ unsigned long pgsize_bitmap)
{
struct page **pages;
unsigned int i = 0, array_size = count * sizeof(*pages);
- unsigned int order = MAX_ORDER;
+ int min_order = get_order(1 << __ffs(pgsize_bitmap));
+ int order = MAX_ORDER;

if (array_size <= PAGE_SIZE)
pages = kzalloc(array_size, GFP_KERNEL);
@@ -213,13 +215,16 @@ static struct page **__iommu_dma_alloc_pages(unsigned int count, gfp_t gfp)
/*
* Higher-order allocations are a convenience rather
* than a necessity, hence using __GFP_NORETRY until
- * falling back to single-page allocations.
+ * falling back to min size allocations.
*/
- for (order = min_t(unsigned int, order, __fls(count));
- order > 0; order--) {
- page = alloc_pages(gfp | __GFP_NORETRY, order);
+ for (order = min_t(int, order, __fls(count));
+ order >= min_order; order--) {
+ page = alloc_pages((order == min_order) ? gfp :
+ gfp | __GFP_NORETRY, order);
if (!page)
continue;
+ if (!order)
+ break;
if (PageCompound(page)) {
if (!split_huge_page(page))
break;
@@ -229,8 +234,6 @@ static struct page **__iommu_dma_alloc_pages(unsigned int count, gfp_t gfp)
break;
}
}
- if (!page)
- page = alloc_page(gfp);
if (!page) {
__iommu_dma_free_pages(pages, i);
return NULL;
@@ -292,7 +295,8 @@ struct page **iommu_dma_alloc(struct device *dev, size_t size,

*handle = DMA_ERROR_CODE;

- pages = __iommu_dma_alloc_pages(count, gfp);
+ pages = __iommu_dma_alloc_pages(count, gfp,
+ domain->ops->pgsize_bitmap);
if (!pages)
return NULL;

--
1.8.1.1.dirty
\
 
 \ /
  Last update: 2016-03-28 09:01    [W:0.078 / U:0.340 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site