Hello, This is against 2.6.34-rc1, no other change compared to #12 other than disabling defrag for anything but khugepaged because current direct reclaim is too slow to run it for hugepages allocations during page faults. We'll enable it again for MADV_HUGEPAGE page faults later with memory compaction when there will be better chance that it's useful CPU work. I assume if this will be merged the "memory compaction core" from Mel will plug nicely on top of this by altering alloc_hugepage() accordingly. Hopefully this is polished enough, the main gripe is left is the #ifdef in the futex code pointed out by Peter, but without knowing the details of gup_fast of whatever new architecture that will be able to mix regular pages and hugepages in the same vma, it's hard to tell what is the cleanest way to abstract that code. Feel free to give a direction on how to change it, if that patch isn't polished enough. http://www.kernel.org/pub/linux/kernel/people/andrea/patches/v2.6/2.6.34-rc1/transparent-hugepage-13/ Thanks, Andrea -- 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>