On Tue, Mar 08, 2016 at 11:43:36AM +0900, Chanho Min wrote:
> The use of phys_to_page is wrong in commit cb06ff102e2d ("ARM: PL011:
> Add support for Rx DMA buffer polling."), beacase DMA address is not
> physical address. Also, phys_to_page is not declared in some architecture.
> The use of virt_to_page is wrong as well because dma_alloc_coherent
> implementaion are allowed to returned remapped memory. So I use sg_table
> instead of scatterlist and change to dma_get_sgtable() implementaion.
dma_get_sgtable() is also broken.
> @@ -344,17 +344,23 @@ static int pl011_sgbuf_init(struct dma_chan *chan,
> struct pl011_sgbuf *sg,
> enum dma_data_direction dir)
> {
> dma_addr_t dma_addr;
> + int ret;
>
> sg->buf = dma_alloc_coherent(chan->device->dev,
> PL011_DMA_BUFFER_SIZE, &dma_addr, GFP_KERNEL);
> if (!sg->buf)
> return -ENOMEM;
sg->buf can be a mapped address, which virt_to_page() will return invalid
results:
int dma_common_get_sgtable(struct device *dev, struct sg_table *sgt,
void *cpu_addr, dma_addr_t handle, size_t size)
{
struct page *page = virt_to_page(cpu_addr);
int ret;
ret = sg_alloc_table(sgt, 1, GFP_KERNEL);
if (unlikely(ret))
return ret;
sg_set_page(sgt->sgl, page, PAGE_ALIGN(size), 0);
return 0;
}
EXPORT_SYMBOL(dma_common_get_sgtable);
I've no idea who is shoveling crap like this into the kernel, but it's
_really_ far from good that such broken abstractions are being merged
as generic code.
--
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.
