On Mon, Aug 6, 2012 at 9:24 PM, Michal Hocko <mhocko@xxxxxxx> wrote: > On Sat 04-08-12 14:08:31, Hillf Danton wrote: >> The computation of page offset index is incorrect to be used in scanning >> prio tree, as huge page offset is required, and is fixed with well >> defined routine. >> >> Changes from v1 >> o s/linear_page_index/linear_hugepage_index/ for clearer code >> o hp_idx variable added for less change >> >> >> Signed-off-by: Hillf Danton <dhillf@xxxxxxxxx> >> --- >> >> --- a/arch/x86/mm/hugetlbpage.c Fri Aug 3 20:34:58 2012 >> +++ b/arch/x86/mm/hugetlbpage.c Fri Aug 3 20:40:16 2012 >> @@ -62,6 +62,7 @@ static void huge_pmd_share(struct mm_str >> { >> struct vm_area_struct *vma = find_vma(mm, addr); >> struct address_space *mapping = vma->vm_file->f_mapping; >> + pgoff_t hp_idx; >> pgoff_t idx = ((addr - vma->vm_start) >> PAGE_SHIFT) + >> vma->vm_pgoff; > > So we have two indexes now. That is just plain ugly! > Two indexes result in less code change here and no change in page_table_shareable. Plus linear_hugepage_index tells clearly readers that hp_idx and idx are different. Anyway I have no strong opinion to keep page_table_shareable unchanged, but prefer less changes. Thanks, Hillf >> struct prio_tree_iter iter; >> @@ -72,8 +73,10 @@ static void huge_pmd_share(struct mm_str >> if (!vma_shareable(vma, addr)) >> return; >> >> + hp_idx = linear_hugepage_index(vma, addr); >> + >> mutex_lock(&mapping->i_mmap_mutex); >> - vma_prio_tree_foreach(svma, &iter, &mapping->i_mmap, idx, idx) { >> + vma_prio_tree_foreach(svma, &iter, &mapping->i_mmap, hp_idx, hp_idx) { >> if (svma == vma) >> continue; >> >> -- > > -- > Michal Hocko > SUSE Labs -- 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>