I'd like to sollicit comments on the following patches againse v3.0: Patch 1 introduces new accessors for page->_count. I believe it is actually ready for upstream submission. Patch 2 introduces a new page count lock. Using it, the interaction between __split_huge_page_refcount and put_compound_page is much simplified. I would like to get comments about this part. One known issue is that my implementation of the page count lock does not currently provide the required memory barrier semantics on non-x86 architectures. This could however be remedied if we decide to go ahead with the idea. Patch 3 demonstrates my motivation for this patch series: in my pre-THP implementation of idle page tracking, I was able to use get_page_unless_zero in a way that __split_huge_page_refcount made unsafe. Building on top of patch 2, I can make the required operation safe again. If patch 2 was to be rejected, I would like to get suggestions about alternative approaches to implement the get_first_page_unless_zero() operation described here. Michel Lespinasse (3): mm: Replace naked page->_count accesses with accessor functions mm: page count lock mm: get_first_page_unless_zero() arch/powerpc/mm/gup.c | 4 +- arch/powerpc/platforms/512x/mpc512x_shared.c | 5 +- arch/x86/mm/gup.c | 6 +- drivers/net/niu.c | 4 +- include/linux/mm.h | 38 ++++++-- include/linux/pagemap.h | 10 ++- mm/huge_memory.c | 25 +++--- mm/internal.h | 97 ++++++++++++++++++++- mm/memory_hotplug.c | 4 +- mm/swap.c | 122 +++++++++++--------------- 10 files changed, 205 insertions(+), 110 deletions(-) -- 1.7.3.1 -- 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/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>