Thread (21 messages) 21 messages, 8 authors, 2018-08-22

[PATCH 4/4] iommu/dma, numa: Use NUMA aware memory allocations in __iommu_dma_alloc_pages

From: Ganapatrao Kulkarni <hidden>
Date: 2017-09-22 15:44:54
Also in: linux-iommu, linux-mm, lkml

Hi Robin,


On Thu, Sep 21, 2017 at 5:11 PM, Robin Murphy [off-list ref] wrote:
On 21/09/17 09:59, Ganapatrao Kulkarni wrote:
quoted
Change function __iommu_dma_alloc_pages to allocate memory/pages
for dma from respective device numa node.

Signed-off-by: Ganapatrao Kulkarni <redacted>
---
 drivers/iommu/dma-iommu.c | 17 ++++++++++-------
 1 file changed, 10 insertions(+), 7 deletions(-)
diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 9d1cebe..0626b58 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -428,20 +428,21 @@ static void __iommu_dma_free_pages(struct page **pages, int count)
      kvfree(pages);
 }

-static struct page **__iommu_dma_alloc_pages(unsigned int count,
-             unsigned long order_mask, gfp_t gfp)
+static struct page **__iommu_dma_alloc_pages(struct device *dev,
+             unsigned int count, unsigned long order_mask, gfp_t gfp)
 {
      struct page **pages;
      unsigned int i = 0, array_size = count * sizeof(*pages);
+     int numa_node = dev_to_node(dev);

      order_mask &= (2U << MAX_ORDER) - 1;
      if (!order_mask)
              return NULL;

      if (array_size <= PAGE_SIZE)
-             pages = kzalloc(array_size, GFP_KERNEL);
+             pages = kzalloc_node(array_size, GFP_KERNEL, numa_node);
      else
-             pages = vzalloc(array_size);
+             pages = vzalloc_node(array_size, numa_node);
kvzalloc{,_node}() didn't exist when this code was first written, but it
does now - since you're touching it you may as well get rid of the whole
if-else and array_size local.
thanks, i will update in next version.
Further nit: some of the indentation below is a bit messed up.
ok, will fix it.
Robin.
quoted
      if (!pages)
              return NULL;
@@ -462,8 +463,9 @@ static struct page **__iommu_dma_alloc_pages(unsigned int count,
                      unsigned int order = __fls(order_mask);

                      order_size = 1U << order;
-                     page = alloc_pages((order_mask - order_size) ?
-                                        gfp | __GFP_NORETRY : gfp, order);
+                     page = alloc_pages_node(numa_node,
+                                     (order_mask - order_size) ?
+                                gfp | __GFP_NORETRY : gfp, order);
                      if (!page)
                              continue;
                      if (!order)
@@ -548,7 +550,8 @@ struct page **iommu_dma_alloc(struct device *dev, size_t size, gfp_t gfp,
              alloc_sizes = min_size;

      count = PAGE_ALIGN(size) >> PAGE_SHIFT;
-     pages = __iommu_dma_alloc_pages(count, alloc_sizes >> PAGE_SHIFT, gfp);
+     pages = __iommu_dma_alloc_pages(dev, count, alloc_sizes >> PAGE_SHIFT,
+                     gfp);
      if (!pages)
              return NULL;
thanks
Ganapat
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help