Hi All,
Would like to check whether the dma_map_single_attrs() API, can accept
ptr variable from high memory or not?
If we pass the memory pointer, which is pointing high memory does not
work with virt_to_page API.
10 static inline dma_addr_t dma_map_single_attrs(struct device *dev,
void *ptr,
11 size_t size,
12 enum
dma_data_direction dir,
13 struct dma_attrs
*attrs)
14 {
15 const struct dma_map_ops *ops = get_dma_ops(dev);
16 dma_addr_t addr;
17
18 kmemcheck_mark_initialized(ptr, size);
19 BUG_ON(!valid_dma_direction(dir));
20 addr = ops->map_page(dev, virt_to_page(ptr),
21 (unsigned long)ptr & ~PAGE_MASK, size,
22 dir, attrs);
<Prasad> using the virt_to_page for the high memory will end up with
wrong address.
23 debug_dma_map_page(dev, virt_to_page(ptr),
24 (unsigned long)ptr & ~PAGE_MASK, size,
25 dir, addr, true);
26 return addr;
27 }
The only restriction which is observed that, ptr should point to
physically contiguous memory.
So can we pass high memory mapped address to dma_map_single API or not?
-Thanks, Prasad
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora
Forum,
Linux Foundation Collaborative Project
On Tue, Mar 22, 2016 at 03:53:02PM +0530, Sodagudi Prasad wrote:
>
> Hi All,
>
> Would like to check whether the dma_map_single_attrs() API, can accept ptr
> variable from high memory or not?
It can't. Direct mapped, lowmem only.
No pointers allowed from dma_alloc_coherent(), dma_alloc_writecombine(),
ioremap(), vmalloc(), kmap(), kmap_atomic() or similar function, or
derived from one of these.
--
RMK's Patch system: http://www.arm.linux.org.uk/developer/patches/
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up
according to speedtest.net.