Kirill, thank you for the comment. On Mon, Sep 02, 2013 at 01:53:27PM +0300, Kirill A. Shutemov wrote: > Naoya Horiguchi wrote: > > Thp related code also uses per process mm->page_table_lock now. So making > > it fine-grained can provide better performance. > > > > This patch makes thp support split page table lock which makes us use > > page->ptl of the pages storing "pmd_trans_huge" pmds. > > Hm. So, you use page->ptl only when you deal with thp pages, otherwise > mm->page_table_lock, right? Maybe it's not enough. We use page->ptl for both of thp and normal depending on USE_SPLIT_PTLOCKS. And regardless of USE_SPLIT_PTLOCKS, mm->page_table_lock is still used by other contexts like memory initialization code or driver code for their specific usage. > It looks inconsistent to me. Does it mean we have to take both locks on > split and collapse paths? This patch includes the replacement with page->ptl for split/collapse path. > I'm not sure if it's safe to take only > page->ptl for alloc path. Probably not. Right, it's not safe. > Why not to use new locking for pmd everywhere? So I already do this. Thanks, Naoya Horiguchi -- 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>