lkml.org 
[lkml]   [2014]   [Jul]   [3]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] rmap: fix pgoff calculation to handle hugepage correctly
On Wed, Jul 02, 2014 at 12:30:57AM -0400, Naoya Horiguchi wrote:
> On Tue, Jul 01, 2014 at 11:15:40PM +0300, Kirill A. Shutemov wrote:
> > On Tue, Jul 01, 2014 at 02:50:21PM -0400, Naoya Horiguchi wrote:
> > > On Tue, Jul 01, 2014 at 09:07:39PM +0300, Kirill A. Shutemov wrote:
> > > > Why do we need this special case for hugetlb page ->index? Why not use
> > > > PAGE_SIZE units there too? Or I miss something?
> > >
> > > hugetlb pages are never split, so we use larger page cache size for
> > > hugetlbfs file (to avoid large sparse page cache tree.)
> >
> > For transparent huge page cache I would like to have native support in
> > page cache radix-tree: since huge pages are always naturally aligned we
> > can create a leaf node for it several (RADIX_TREE_MAP_SHIFT -
> > HPAGE_PMD_ORDER) levels up by tree, which would cover all indexes in the
> > range the huge page represents. This approach should fit hugetlb too. And
> > -1 special case for hugetlb.
> > But I'm not sure when I'll get time to play with this...
>
> So I'm OK that hugetlb page should have ->index in PAGE_CACHE_SIZE
> when transparent huge page is merged. I may try to write patches
> on top of your tree after I've done a few series in my work queue.
>
> In order to fix the current problem, I suggest a page_to_pgoff() as a
> short-term workaround. I found a few other call sites which can call
> on hugepage, so this function help us track such callers.
> The similar function seems to be introduced in your transparent huge
> page cache tree (page_cache_index()). So this function will be finally
> overwritten with it.
>
> Thanks,
> Naoya Horiguchi
> ---
> From: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
> Date: Tue, 1 Jul 2014 21:38:22 -0400
> Subject: [PATCH v2] rmap: fix pgoff calculation to handle hugepage correctly
>
> I triggered VM_BUG_ON() in vma_address() when I try to migrate an anonymous
> hugepage with mbind() in the kernel v3.16-rc3. This is because pgoff's
> calculation in rmap_walk_anon() fails to consider compound_order() only to
> have an incorrect value.
>
> This patch introduces page_to_pgoff(), which gets the page's offset in
> PAGE_CACHE_SIZE. Kirill pointed out that page cache tree should natively
> handle hugepages, and in order to make hugetlbfs fit it, page->index of
> hugetlbfs page should be in PAGE_CACHE_SIZE. This is beyond this patch,
> but page_to_pgoff() contains the point to be fixed in a single function.
>
> Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>

Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>

--
Kirill A. Shutemov


\
 
 \ /
  Last update: 2014-07-03 14:01    [W:0.110 / U:0.016 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site