On Sat, Apr 24, 2010 at 01:13:40PM +0200, Andrea Arcangeli wrote: > On Sat, Apr 24, 2010 at 11:52:27AM +0100, Mel Gorman wrote: > > I think you're right. This is a new bug introduced by the anon_vma changes. On > > the plus side, it means we don't have to worry about -stable. > > Correct, no worry about -stable. > > > > vma_adjust already takes the anon_vma->lock and of course I also > > > further verified that trying to apply your snippet to vma_adjust > > > results in immediately deadlock as the very same lock is already taken > > > in my tree as it's the same anon-vma (simpler). > > > > Yes, I expected that. Previously, there was only one anon_vma so if you > > double-take the lock, bad things happen. > > > > > So aa.git will be > > > immune from these bugs for now. > > > > > > > It should be. I expect that's why you have never seen the bugon in > > swapops. > > Correct, I never seen it, and I keep it under very great stress with > swap storms of hugepages, lots of I/O and khugepaged at 100% cpu. > Well, to me this is also good because it shows it's not an existing bug in migration or a new bug introduced by compaction either. Previously I hadn't seen this bug either but until relatively recently, the bulk of the testing was against 2.6.33. > Also keep in mind expand_downwards which also adjusts > vm_start/vm_pgoff the same way (and without mmap_sem write mode). > Will keep it in mind. It's taking the anon_vma lock but once again, there might be more than one anon_vma to worry about and the proper locking still isn't massively clear to me. -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab -- 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>