If race between isolatation and allocation happens, we could need to move
some freepages to MIGRATE_ISOLATE in __test_page_isolated_in_pageblock().
The current code ignores the zone_freepage accounting after the move,
which cause the zone NR_FREE_PAGES and NR_FREE_CMA_PAGES statistics incorrect.
This patch fixes this rare issue.
Signed-off-by: Weijie Yang <[email protected]>
---
mm/page_isolation.c | 5 ++++-
1 files changed, 4 insertions(+), 1 deletions(-)
diff --git a/mm/page_isolation.c b/mm/page_isolation.c
index 3ddc8b3..15b51de 100644
--- a/mm/page_isolation.c
+++ b/mm/page_isolation.c
@@ -193,12 +193,15 @@ __test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
* is MIGRATE_ISOLATE. Catch it and move the page into
* MIGRATE_ISOLATE list.
*/
- if (get_freepage_migratetype(page) != MIGRATE_ISOLATE) {
+ int migratetype = get_freepage_migratetype(page);
+ if (migratetype != MIGRATE_ISOLATE) {
struct page *end_page;
end_page = page + (1 << page_order(page)) - 1;
move_freepages(page_zone(page), page, end_page,
MIGRATE_ISOLATE);
+ __mod_zone_freepage_state(zone,
+ -(1 << page_order(page)), migratetype);
}
pfn += 1 << page_order(page);
}
--
1.7.0.4
On Thu, Nov 06, 2014 at 04:09:08PM +0800, Weijie Yang wrote:
> If race between isolatation and allocation happens, we could need to move
> some freepages to MIGRATE_ISOLATE in __test_page_isolated_in_pageblock().
> The current code ignores the zone_freepage accounting after the move,
> which cause the zone NR_FREE_PAGES and NR_FREE_CMA_PAGES statistics incorrect.
>
> This patch fixes this rare issue.
Hello,
After "fix freepage count problems in memory isolation" merged, this race
should not happen. I have to remove it in that patchset, but, I
forgot to remove it. Please remove this race handling code completely and
tag with stable. If we don't remove it, there is errornous situation
because get_freepage_migratetype() could return invalid migratetype
although the page is on the correct buddy list. So, we regard
no race situation as race one.
Thanks.
>
> Signed-off-by: Weijie Yang <[email protected]>
> ---
> mm/page_isolation.c | 5 ++++-
> 1 files changed, 4 insertions(+), 1 deletions(-)
>
> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
> index 3ddc8b3..15b51de 100644
> --- a/mm/page_isolation.c
> +++ b/mm/page_isolation.c
> @@ -193,12 +193,15 @@ __test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
> * is MIGRATE_ISOLATE. Catch it and move the page into
> * MIGRATE_ISOLATE list.
> */
> - if (get_freepage_migratetype(page) != MIGRATE_ISOLATE) {
> + int migratetype = get_freepage_migratetype(page);
> + if (migratetype != MIGRATE_ISOLATE) {
> struct page *end_page;
>
> end_page = page + (1 << page_order(page)) - 1;
> move_freepages(page_zone(page), page, end_page,
> MIGRATE_ISOLATE);
> + __mod_zone_freepage_state(zone,
> + -(1 << page_order(page)), migratetype);
> }
> pfn += 1 << page_order(page);
> }
> --
> 1.7.0.4
>
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to [email protected]. For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
On Thu, Nov 6, 2014 at 4:49 PM, Joonsoo Kim <[email protected]> wrote:
> On Thu, Nov 06, 2014 at 04:09:08PM +0800, Weijie Yang wrote:
>> If race between isolatation and allocation happens, we could need to move
>> some freepages to MIGRATE_ISOLATE in __test_page_isolated_in_pageblock().
>> The current code ignores the zone_freepage accounting after the move,
>> which cause the zone NR_FREE_PAGES and NR_FREE_CMA_PAGES statistics incorrect.
>>
>> This patch fixes this rare issue.
>
> Hello,
>
> After "fix freepage count problems in memory isolation" merged, this race
> should not happen. I have to remove it in that patchset, but, I
> forgot to remove it. Please remove this race handling code completely and
> tag with stable. If we don't remove it, there is errornous situation
> because get_freepage_migratetype() could return invalid migratetype
> although the page is on the correct buddy list. So, we regard
> no race situation as race one.
Thanks for your remind, I will read your patch.
> Thanks.
>
>>
>> Signed-off-by: Weijie Yang <[email protected]>
>> ---
>> mm/page_isolation.c | 5 ++++-
>> 1 files changed, 4 insertions(+), 1 deletions(-)
>>
>> diff --git a/mm/page_isolation.c b/mm/page_isolation.c
>> index 3ddc8b3..15b51de 100644
>> --- a/mm/page_isolation.c
>> +++ b/mm/page_isolation.c
>> @@ -193,12 +193,15 @@ __test_page_isolated_in_pageblock(unsigned long pfn, unsigned long end_pfn,
>> * is MIGRATE_ISOLATE. Catch it and move the page into
>> * MIGRATE_ISOLATE list.
>> */
>> - if (get_freepage_migratetype(page) != MIGRATE_ISOLATE) {
>> + int migratetype = get_freepage_migratetype(page);
>> + if (migratetype != MIGRATE_ISOLATE) {
>> struct page *end_page;
>>
>> end_page = page + (1 << page_order(page)) - 1;
>> move_freepages(page_zone(page), page, end_page,
>> MIGRATE_ISOLATE);
>> + __mod_zone_freepage_state(zone,
>> + -(1 << page_order(page)), migratetype);
>> }
>> pfn += 1 << page_order(page);
>> }
>> --
>> 1.7.0.4
>>
>>
>> --
>> To unsubscribe, send a message with 'unsubscribe linux-mm' in
>> the body to [email protected]. For more info on Linux MM,
>> see: http://www.linux-mm.org/ .
>> Don't email: <a href=mailto:"[email protected]"> [email protected] </a>