Re: [Linaro-mm-sig] [PATCH 4/4] ARM: dma-mapping: Skip cache maint for invalid page

Hi Marek,

Marek Szyprowski m.szyprowski@samsung.com wrote @ Tue, 28 Aug 2012 23:58:48 +0200:

Hello,

On 8/28/2012 1:43 PM, Hiroshi Doyu wrote:

...

Skip cache maint for invalid page but warn it.

Signed-off-by: Hiroshi Doyu hdoyu@nvidia.com

arch/arm/mm/dma-mapping.c | 9 +++++++++ 1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/arch/arm/mm/dma-mapping.c b/arch/arm/mm/dma-mapping.c index 2621282..0368702 100644 --- a/arch/arm/mm/dma-mapping.c +++ b/arch/arm/mm/dma-mapping.c @@ -1589,6 +1589,9 @@ static void arm_iommu_unmap_page(struct device *dev, dma_addr_t handle, if (!iova) return;

• if (WARN_ON(!pfn_valid(page_to_pfn(page))))

• return;
  

• if (!dma_get_attr(DMA_ATTR_SKIP_CPU_SYNC, attrs)) __dma_page_dev_to_cpu(page, offset, size, dir);

@@ -1607,6 +1610,9 @@ static void arm_iommu_sync_single_for_cpu(struct device *dev, if (!iova) return;

• if (WARN_ON(!pfn_valid(page_to_pfn(page))))

• return;
  

• __dma_page_dev_to_cpu(page, offset, size, dir); }

@@ -1621,6 +1627,9 @@ static void arm_iommu_sync_single_for_device(struct device *dev, if (!iova) return;

• if (WARN_ON(!pfn_valid(page_to_pfn(page))))

• return;
  

• __dma_page_cpu_to_dev(page, offset, size, dir); }

I really wonder what has happened in your system that you require such paranoid checks. All of the above functions are called on the scatter-list which has been first passed to dma_map_sg(). IMHO it makes much more sense to add a pfn_valid() check in dma_map_sg() and return error if the provided scatter list is not valid instead of these WARN_ONs.

Actually one device driver hit Oops at unmap_page() since this function could be called directly from device drivers, and WARN_ON(dump_stack) helped to find out which device driver did something wrong. IIRC, the following functions can be called directly from device drivers?

static inline void dma_unmap_page(struct device *dev, dma_addr_t addr, size_t size, enum dma_data_direction dir); static inline void dma_sync_single_for_cpu(struct device *dev, dma_addr_t addr, size_t size, enum dma_data_direction dir); static inline void dma_sync_single_for_device(struct device *dev, dma_addr_t addr, size_t size, enum dma_data_direction dir);