Messages in this thread | | | Date | Wed, 22 Feb 2012 13:06:59 -0800 | From | Andrew Morton <> | Subject | Re: [PATCH] mm: hugetlb: bail out unmapping after serving reference page |
| |
On Wed, 22 Feb 2012 20:35:34 +0800 Hillf Danton <dhillf@gmail.com> wrote:
> When unmapping given VM range, we could bail out if a reference page is > supplied and it is unmapped, which is a minor optimization. > > Signed-off-by: Hillf Danton <dhillf@gmail.com> > --- > > --- a/mm/hugetlb.c Wed Feb 22 19:34:12 2012 > +++ b/mm/hugetlb.c Wed Feb 22 19:50:26 2012 > @@ -2280,6 +2280,9 @@ void __unmap_hugepage_range(struct vm_ar > if (pte_dirty(pte)) > set_page_dirty(page); > list_add(&page->lru, &page_list); > + > + if (page == ref_page) > + break; > } > spin_unlock(&mm->page_table_lock); > flush_tlb_range(vma, start, end);
Perhaps add a little comment to this explaining what's going on?
It would be sufficient to do
if (ref_page) break;
This is more efficient, and doesn't make people worry about whether this value of `page' is the same as the one which pte_page(huge_ptep_get()) earlier returned.
Why do we evaluate `page' twice inside that loop anyway? And why do we check for huge_pte_none() twice? It looks all messed up.
| |