On Wed, Apr 28, 2010 at 03:05:43AM +0200, Andrea Arcangeli wrote: > 1) adjusting page->index atomically with the pte updates inside pt > lock (while it moves from one pte to another) actually no need of this at all! of course the dst vma will have vma->vm_pgoff adjusted instead... never mind. So I don't see a problem there. I think this is very special of how exec.c abuses move_page_tables by passing vma as src and dst, when it obviously cannot be indexed in two anon-vmas because there's a single vma and a single vma->anon_vma, and src and dst obviously cannot have two different vm_pgoff again because there's a single vma and there can't be two different vma->vm_pgoff. So I'm very hopeful do_mremap is already fully safe... -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>