lkml.org 
[lkml]   [2011]   [Jan]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [BUGFIX][PATCH] memcg: fix memory migration of shmem swapcache
From
On Thu, Jan 6, 2011 at 10:09 AM, Daisuke Nishimura
<nishimura@mxp.nes.nec.co.jp> wrote:
> On Wed, 5 Jan 2011 12:58:40 +0100
> Johannes Weiner <hannes@cmpxchg.org> wrote:
>
>> On Wed, Jan 05, 2011 at 01:00:20PM +0900, Daisuke Nishimura wrote:
>> > In current implimentation, mem_cgroup_end_migration() decides whether the page
>> > migration has succeeded or not by checking "oldpage->mapping".
>> >
>> > But if we are tring to migrate a shmem swapcache, the page->mapping of it is
>> > NULL from the begining, so the check would be invalid.
>> > As a result, mem_cgroup_end_migration() assumes the migration has succeeded
>> > even if it's not, so "newpage" would be freed while it's not uncharged.
>> >
>> > This patch fixes it by passing mem_cgroup_end_migration() the result of the
>> > page migration.
>>
>> Are there other users that rely on unused->mapping being NULL after
>> migration?
>>
> As long as I can see, no.
>
>> If so, aren't they prone to misinterpreting this for shmem swapcache
>> as well?
>>
>> If not, wouldn't it be better to remove that page->mapping = NULL from
>> migrate_page_copy() altogether?  I think it's an ugly exception where
>> the outcome of PageAnon() is not meaningful for an LRU page.
>>
> IIUC, oldpage will be freed on success of page migration, so we hit bad_page
> check at freeing the page unless we clear oldpage->mapping,
>
>> To your patch:
>>
>> > --- a/mm/memcontrol.c
>> > +++ b/mm/memcontrol.c
>> > @@ -2856,7 +2856,7 @@ int mem_cgroup_prepare_migration(struct page *page,
>> >
>> >  /* remove redundant charge if migration failed*/
>> >  void mem_cgroup_end_migration(struct mem_cgroup *mem,
>> > -   struct page *oldpage, struct page *newpage)
>> > +   struct page *oldpage, struct page *newpage, int result)
>> >  {
>> >     struct page *used, *unused;
>> >     struct page_cgroup *pc;
>> > @@ -2865,8 +2865,7 @@ void mem_cgroup_end_migration(struct mem_cgroup *mem,
>> >             return;
>> >     /* blocks rmdir() */
>> >     cgroup_exclude_rmdir(&mem->css);
>> > -   /* at migration success, oldpage->mapping is NULL. */
>> > -   if (oldpage->mapping) {
>> > +   if (result) {
>>
>> Since this function does not really need more than a boolean value,
>> wouldn't it make the code more obvious if the parameter was `bool
>> success'?
>>
>>       if (!success) {
>> >             used = oldpage;
>> >             unused = newpage;
>> >     } else {
>>
>> Minor nit, though.  I agree with the patch in general.
>>
> Thank you for your review.
> How about this ?
>
> ===
> From: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
>
> In current implimentation, mem_cgroup_end_migration() decides whether the page
> migration has succeeded or not by checking "oldpage->mapping".
>
> But if we are tring to migrate a shmem swapcache, the page->mapping of it is
> NULL from the begining, so the check would be invalid.
> As a result, mem_cgroup_end_migration() assumes the migration has succeeded
> even if it's not, so "newpage" would be freed while it's not uncharged.
>
> This patch fixes it by passing mem_cgroup_end_migration() the result of the
> page migration.
>
> Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>

Below minor nitpick.

> ---
>  include/linux/memcontrol.h |    5 ++---
>  mm/memcontrol.c            |    5 ++---
>  mm/migrate.c               |    2 +-
>  3 files changed, 5 insertions(+), 7 deletions(-)
>
> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> index 159a076..cc5a8fd 100644
> --- a/include/linux/memcontrol.h
> +++ b/include/linux/memcontrol.h
> @@ -93,7 +93,7 @@ extern int
>  mem_cgroup_prepare_migration(struct page *page,
>        struct page *newpage, struct mem_cgroup **ptr);
>  extern void mem_cgroup_end_migration(struct mem_cgroup *mem,
> -       struct page *oldpage, struct page *newpage);
> +       struct page *oldpage, struct page *newpage, bool success);

The term "success" implies present or future tense.
The event this variable cares about in the past so "succeed" might be
a more appropriate term.
Sorry to be picky about the English but there is an important
distinction here since we don't have any comment about the variable.

Am I being too fussy?
I don't want to bother you since Kame already acked it so I will
depend on your decision.

--
Kind regards,
Minchan Kim
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2011-01-06 03:51    [W:0.072 / U:0.932 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site