On Tue, 2014-04-22 at 21:23 -0700, Linus Torvalds wrote: > > But I'm starting to consider this whole thing to be a 3.16 issue by > now. It wasn't as simple as it looked, and while our old location of > set_page_dirty() is clearly wrong, and DaveH even got a test-case for > it (which I initially doubted would even be possible), I still > seriously doubt that anybody sane who cares about data consistency > will do concurrent unmaps (or MADV_DONTNEED) while another writer is > actively using that mapping. I'm more worried about users of unmap_mapping_ranges() than concurrent munmap(). Things like the DRM playing tricks like swapping a mapping from memory to frame buffer and vice-versa. In any case, I agree with delaying that for 3.16, it's still very unlikely that we hit this in any case that actually matters. Cheers, Ben. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>