On Fri, Oct 20, 2017 at 3:37 AM, Christoph Hellwig <hch@xxxxxx> wrote: > On Thu, Oct 19, 2017 at 06:47:45PM -0400, Jim Quinlan wrote: >> The only way to prevent this is to reserve a single page at the end of >> the first memory region of any pair that are adjacent in physical >> memory. A hack, yes, but I don't see an easier way out of this. Many >> if not most of our boards do not have adjacent regions and would not >> need this. > > dma mappings can be much larger than a single page. For the block > world take a look at __blk_segment_map_sg which does the merging > of contiguous pages into a single SG segment. You'd have to override > BIOVEC_PHYS_MERGEABLE to prevent this from happening in your supported > architectures for the block layer. I am not sure I understand your comment -- the size of the request shouldn't be a factor. Let's look at your example of the DMA request of 3fffff00 to 4000000f (physical memory). Lets say it is for 15 pages. If we block out the last page [0x3ffff000..0x3fffffff] from what is available, there is no 15 page span that can happen across the 0x40000000 boundary. For SG, there can be no merge that connects a page from one region to another region. Can you give an example of the scenario you are thinking of?
