lkml.org 
[lkml]   [2012]   [Jul]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [RFC 3/3] memory-hotplug: bug fix race between isolation and allocation
From
On Wed, Jul 18, 2012 at 11:56 AM, Minchan Kim <minchan@kernel.org> wrote:
> On Wed, Jul 18, 2012 at 11:12:27AM +0800, Bob Liu wrote:
>> On Wed, Jul 18, 2012 at 10:41 AM, Minchan Kim <minchan@kernel.org> wrote:
>> > On Wed, Jul 18, 2012 at 10:12:57AM +0800, Bob Liu wrote:
>> >> On Wed, Jul 18, 2012 at 7:40 AM, Minchan Kim <minchan@kernel.org> wrote:
>> >> > Hi Bob,
>> >> >
>> >> > On Tue, Jul 17, 2012 at 06:13:17PM +0800, Bob Liu wrote:
>> >> >> Hi Minchan,
>> >> >>
>> >> >> On Tue, Jul 17, 2012 at 3:01 PM, Minchan Kim <minchan@kernel.org> wrote:
>> >> >> > Like below, memory-hotplug makes race between page-isolation
>> >> >> > and page-allocation so it can hit BUG_ON in __offline_isolated_pages.
>> >> >> >
>> >> >> > CPU A CPU B
>> >> >> >
>> >> >> > start_isolate_page_range
>> >> >> > set_migratetype_isolate
>> >> >> > spin_lock_irqsave(zone->lock)
>> >> >> >
>> >> >> > free_hot_cold_page(Page A)
>> >> >> > /* without zone->lock */
>> >> >> > migratetype = get_pageblock_migratetype(Page A);
>> >> >> > /*
>> >> >> > * Page could be moved into MIGRATE_MOVABLE
>> >> >> > * of per_cpu_pages
>> >> >> > */
>> >> >> > list_add_tail(&page->lru, &pcp->lists[migratetype]);
>> >> >> >
>> >> >> > set_pageblock_isolate
>> >> >> > move_freepages_block
>> >> >> > drain_all_pages
>> >> >> >
>> >> >> > /* Page A could be in MIGRATE_MOVABLE of free_list. */
>> >> >> >
>> >> >> > check_pages_isolated
>> >> >> > __test_page_isolated_in_pageblock
>> >> >> > /*
>> >> >> > * We can't catch freed page which
>> >> >> > * is free_list[MIGRATE_MOVABLE]
>> >> >> > */
>> >> >> > if (PageBuddy(page A))
>> >> >> > pfn += 1 << page_order(page A);
>> >> >> >
>> >> >> > /* So, Page A could be allocated */
>> >> >> >
>> >> >> > __offline_isolated_pages
>> >> >> > /*
>> >> >> > * BUG_ON hit or offline page
>> >> >> > * which is used by someone
>> >> >> > */
>> >> >> > BUG_ON(!PageBuddy(page A));
>> >> >> >
>> >> >> > Signed-off-by: Minchan Kim <minchan@kernel.org>
>> >> >> > ---
>> >> >> > I found this problem during code review so please confirm it.
>> >> >> > Kame?
>> >> >> >
>> >> >> > mm/page_isolation.c | 5 ++++-
>> >> >> > 1 file changed, 4 insertions(+), 1 deletion(-)
>> >> >> >
>> >> >> > diff --git a/mm/page_isolation.c b/mm/page_isolation.c
>> >> >> > index acf65a7..4699d1f 100644
>> >> >> > --- a/mm/page_isolation.c
>> >> >> > +++ b/mm/page_isolation.c
>> >> >> > @@ -196,8 +196,11 @@ __test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn)
>> >> >> > continue;
>> >> >> > }
>> >> >> > page = pfn_to_page(pfn);
>> >> >> > - if (PageBuddy(page))
>> >> >> > + if (PageBuddy(page)) {
>> >> >> > pfn += 1 << page_order(page);
>> >> >> > + if (get_page_migratetype(page) != MIGRATE_ISOLATE)
>> >> >> > + break;
>> >> >> > + }
>> >> >>
>> >> >> test_page_isolated() already have check
>> >> >> get_pageblock_migratetype(page) != MIGRATE_ISOLATE.
>> >> >>
>> >> >
>> >> > That's why I send a patch.
>> >> > As I describe in description, pageblock migration type of get_page_migratetype(page)
>> >> > is inconsistent with free_list[migrationtype].
>> >> > I mean get_pageblock_migratetype(page) will return MIGRATE_ISOLATE but the page would be
>> >> > in free_list[MIGRATE_MOVABLE] so it could be allocated for someone if that race happens.
>> >> >
>> >>
>> >> Sorry, I'm still not get the situation how this race happens.
>> >>
>> >> set_pageblock_isolate
>> >> move_freepages_block
>> >> drain_all_pages
>> >>
>> >> /* Page A could be in MIGRATE_MOVABLE
>> >> of free_list. */
>> >>
>> >> I think move_freepages_block() will call list_move() to move Page A to
>> >> free_list[MIGRATE_ISOLATE], so this case can't happen?
>> >
>> > move_freepages_block handles only pages in free_list but Page A is on pcp, not free_list.
>> >
>>
>> Got it, then why not just drain pcp pages before move_freepages_block() ?
>
> CPU A CPU B
>
> drain_all_pages
> lock(zone->lock);
> free_hot_cold_page
> MIGRATE_MOVABLE = get_pageblock_migratetype(page);
> list_add(&page->lru, &pcp->lists[migratetype])
> set_pageblock_isolate
> move_free_pages_block
> unlock(zone->lock);
>
> We can't make it atomic.
>
>>
>> And I didn't see the effect by adding the check if
>> (get_page_migratetype(page) != MIGRATE_ISOLATE) for this race.
>> Since set_pageblock_isolate() have been called by CPU A, this check
>> will always false which cause CPU A still consider Page A isolated,
>> then PAGE A still can be allocated by CPU B from pcp.
>
> Please don't confuse get_page_migratetype and get_pageblock_migratetype.
> get_page_migratetype returns migratetype inforamtion of *page* which is
> in free_list while get_pageblock_migratetype returns *pageblock*'s migratetype.
>

Oh, yes i mixed them up.

Last question, it's better to break only before add pfn.
like:

- if (PageBuddy(page))
+ if (PageBuddy(page)) {
+ if (get_page_migratetype(page) != MIGRATE_ISOLATE)
+ break;
pfn += 1 << page_order(page);
+ }

--
Regards,
--Bob


\
 
 \ /
  Last update: 2012-07-18 08:01    [W:0.058 / U:0.288 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site