On Sat, Aug 15, 2015 at 9:00 AM, Christoph Hellwig <hch@xxxxxx> wrote: > On Sat, Aug 15, 2015 at 08:44:27AM -0700, Dan Williams wrote: >> That said, while we don't need special accessors we do need guarantees >> that anything that has written to a persistent memory address has done >> so in a way that wmb_pmem() is able to flush it. It's more of a "I've >> audited this code path for wmb_pmem() compatibility so use this api to >> write to pmem." > > I'm more worried about things where don't just do plain loads and stores > to a pmem region but DMA, which will end up as a nightmare of casts. > > But we can wait and see how that evolves in the end.. It's already not possible to do something like dma_map_single() on an ioremap()'d address, so there currently are't any __iomem/DMA collisions. As long as DMA setup is relative to a physical address resource I think we're ok. Making sure a DMA is both complete and persistent though is a different problem. -- To unsubscribe from this list: send the line "unsubscribe linux-doc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
