[PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[Jaewon Kim]

reclaim_clean_pages_from_list() decreases NR_ISOLATED_FILE by returned
value from shrink_page_list(). But mlocked pages in the isolated
clean_pages page list would be removed from the list but not counted as
nr_reclaimed. Fix this miscounting by returning the number of mlocked
pages and count it.

Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
---
 mm/vmscan.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 5e8eadd..5837695 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -849,6 +849,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 				      unsigned long *ret_nr_congested,
 				      unsigned long *ret_nr_writeback,
 				      unsigned long *ret_nr_immediate,
+				      unsigned long *ret_nr_mlocked,
 				      bool force_reclaim)
 {
 	LIST_HEAD(ret_pages);
@@ -1158,6 +1159,7 @@ cull_mlocked:
 			try_to_free_swap(page);
 		unlock_page(page);
 		putback_lru_page(page);
+		(*ret_nr_mlocked)++;
 		continue;
 
 activate_locked:
@@ -1197,6 +1199,7 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 		.may_unmap = 1,
 	};
 	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
+	unsigned long nr_mlocked = 0;
 	struct page *page, *next;
 	LIST_HEAD(clean_pages);
 
@@ -1210,8 +1213,10 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 
 	ret = shrink_page_list(&clean_pages, zone, &sc,
 			TTU_UNMAP|TTU_IGNORE_ACCESS,
-			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
+			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5,
+			&nr_mlocked, true);
 	list_splice(&clean_pages, page_list);
+	ret += nr_mlocked;
 	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
 	return ret;
 }
@@ -1523,6 +1528,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
 	unsigned long nr_unqueued_dirty = 0;
 	unsigned long nr_writeback = 0;
 	unsigned long nr_immediate = 0;
+	unsigned long nr_mlocked = 0;
 	isolate_mode_t isolate_mode = 0;
 	int file = is_file_lru(lru);
 	struct zone *zone = lruvec_zone(lruvec);
@@ -1565,7 +1571,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
 
 	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
 				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
-				&nr_writeback, &nr_immediate,
+				&nr_writeback, &nr_immediate, &nr_mlocked,
 				false);
 
 	spin_lock_irq(&zone->lru_lock);
-- 
1.9.1

Re: [PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[Minchan Kim]

Hello,

On Mon, Aug 03, 2015 at 07:18:27PM +0900, Jaewon Kim wrote:
> reclaim_clean_pages_from_list() decreases NR_ISOLATED_FILE by returned
> value from shrink_page_list(). But mlocked pages in the isolated
> clean_pages page list would be removed from the list but not counted as
> nr_reclaimed. Fix this miscounting by returning the number of mlocked
> pages and count it.

If there are pages not able to reclaim, VM try to migrate it and
have to handle the stat in migrate_pages.
If migrate_pages fails again, putback-fiends should handle it.

Is there anyting I am missing now?

Thanks.

> 
> Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
> ---
>  mm/vmscan.c | 10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 5e8eadd..5837695 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -849,6 +849,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
>  				      unsigned long *ret_nr_congested,
>  				      unsigned long *ret_nr_writeback,
>  				      unsigned long *ret_nr_immediate,
> +				      unsigned long *ret_nr_mlocked,
>  				      bool force_reclaim)
>  {
>  	LIST_HEAD(ret_pages);
> @@ -1158,6 +1159,7 @@ cull_mlocked:
>  			try_to_free_swap(page);
>  		unlock_page(page);
>  		putback_lru_page(page);
> +		(*ret_nr_mlocked)++;
>  		continue;
>  
>  activate_locked:
> @@ -1197,6 +1199,7 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
>  		.may_unmap = 1,
>  	};
>  	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
> +	unsigned long nr_mlocked = 0;
>  	struct page *page, *next;
>  	LIST_HEAD(clean_pages);
>  
> @@ -1210,8 +1213,10 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
>  
>  	ret = shrink_page_list(&clean_pages, zone, &sc,
>  			TTU_UNMAP|TTU_IGNORE_ACCESS,
> -			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
> +			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5,
> +			&nr_mlocked, true);
>  	list_splice(&clean_pages, page_list);
> +	ret += nr_mlocked;
>  	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
>  	return ret;
>  }
> @@ -1523,6 +1528,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>  	unsigned long nr_unqueued_dirty = 0;
>  	unsigned long nr_writeback = 0;
>  	unsigned long nr_immediate = 0;
> +	unsigned long nr_mlocked = 0;
>  	isolate_mode_t isolate_mode = 0;
>  	int file = is_file_lru(lru);
>  	struct zone *zone = lruvec_zone(lruvec);
> @@ -1565,7 +1571,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>  
>  	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
>  				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
> -				&nr_writeback, &nr_immediate,
> +				&nr_writeback, &nr_immediate, &nr_mlocked,
>  				false);
>  
>  	spin_lock_irq(&zone->lru_lock);
> -- 
> 1.9.1
> 

Re: [PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[Jaewon Kim]

On 2015년 08월 03일 21:27, Minchan Kim wrote:
> Hello,
> 
> On Mon, Aug 03, 2015 at 07:18:27PM +0900, Jaewon Kim wrote:
>> reclaim_clean_pages_from_list() decreases NR_ISOLATED_FILE by returned
>> value from shrink_page_list(). But mlocked pages in the isolated
>> clean_pages page list would be removed from the list but not counted as
>> nr_reclaimed. Fix this miscounting by returning the number of mlocked
>> pages and count it.
> 
> If there are pages not able to reclaim, VM try to migrate it and
> have to handle the stat in migrate_pages.
> If migrate_pages fails again, putback-fiends should handle it.
> 
> Is there anyting I am missing now?
> 
> Thanks.
> 
Hello

Only pages in cc->migratepages will be handled by migrate_pages or
putback_movable_pages, and NR_ISOLATED_FILE will be counted properly.
However mlocked pages will not be put back into cc->migratepages,
and also not be counted in NR_ISOLATED_FILE because putback_lru_page
in shrink_page_list does not increase NR_ISOLATED_FILE.
The current reclaim_clean_pages_from_list assumes that shrink_page_list
returns number of pages removed from the candidate list.

i.e)
isolate_migratepages_range    : NR_ISOLATED_FILE += 10
reclaim_clean_pages_from_list : NR_ISOLATED_FILE -= 5 (1 mlocked page)
migrate_pages                 : NR_ISOLATED_FILE -=4
=> NR_ISOLATED_FILE increased by 1

Thank you.
>>
>> Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
>> ---
>>  mm/vmscan.c | 10 ++++++++--
>>  1 file changed, 8 insertions(+), 2 deletions(-)
>>
>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>> index 5e8eadd..5837695 100644
>> --- a/mm/vmscan.c
>> +++ b/mm/vmscan.c
>> @@ -849,6 +849,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
>>  				      unsigned long *ret_nr_congested,
>>  				      unsigned long *ret_nr_writeback,
>>  				      unsigned long *ret_nr_immediate,
>> +				      unsigned long *ret_nr_mlocked,
>>  				      bool force_reclaim)
>>  {
>>  	LIST_HEAD(ret_pages);
>> @@ -1158,6 +1159,7 @@ cull_mlocked:
>>  			try_to_free_swap(page);
>>  		unlock_page(page);
>>  		putback_lru_page(page);
>> +		(*ret_nr_mlocked)++;
>>  		continue;
>>  
>>  activate_locked:
>> @@ -1197,6 +1199,7 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
>>  		.may_unmap = 1,
>>  	};
>>  	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
>> +	unsigned long nr_mlocked = 0;
>>  	struct page *page, *next;
>>  	LIST_HEAD(clean_pages);
>>  
>> @@ -1210,8 +1213,10 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
>>  
>>  	ret = shrink_page_list(&clean_pages, zone, &sc,
>>  			TTU_UNMAP|TTU_IGNORE_ACCESS,
>> -			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
>> +			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5,
>> +			&nr_mlocked, true);
>>  	list_splice(&clean_pages, page_list);
>> +	ret += nr_mlocked;
>>  	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
>>  	return ret;
>>  }
>> @@ -1523,6 +1528,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>>  	unsigned long nr_unqueued_dirty = 0;
>>  	unsigned long nr_writeback = 0;
>>  	unsigned long nr_immediate = 0;
>> +	unsigned long nr_mlocked = 0;
>>  	isolate_mode_t isolate_mode = 0;
>>  	int file = is_file_lru(lru);
>>  	struct zone *zone = lruvec_zone(lruvec);
>> @@ -1565,7 +1571,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>>  
>>  	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
>>  				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
>> -				&nr_writeback, &nr_immediate,
>> +				&nr_writeback, &nr_immediate, &nr_mlocked,
>>  				false);
>>  
>>  	spin_lock_irq(&zone->lru_lock);
>> -- 
>> 1.9.1
>>
> 

Re: [PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[Minchan Kim]

On Mon, Aug 03, 2015 at 11:55:46PM +0900, Jaewon Kim wrote:
> 
> 
> On 2015년 08월 03일 21:27, Minchan Kim wrote:
> > Hello,
> > 
> > On Mon, Aug 03, 2015 at 07:18:27PM +0900, Jaewon Kim wrote:
> >> reclaim_clean_pages_from_list() decreases NR_ISOLATED_FILE by returned
> >> value from shrink_page_list(). But mlocked pages in the isolated
> >> clean_pages page list would be removed from the list but not counted as
> >> nr_reclaimed. Fix this miscounting by returning the number of mlocked
> >> pages and count it.
> > 
> > If there are pages not able to reclaim, VM try to migrate it and
> > have to handle the stat in migrate_pages.
> > If migrate_pages fails again, putback-fiends should handle it.
> > 
> > Is there anyting I am missing now?
> > 
> > Thanks.
> > 
> Hello
> 
> Only pages in cc->migratepages will be handled by migrate_pages or
> putback_movable_pages, and NR_ISOLATED_FILE will be counted properly.
> However mlocked pages will not be put back into cc->migratepages,
> and also not be counted in NR_ISOLATED_FILE because putback_lru_page
> in shrink_page_list does not increase NR_ISOLATED_FILE.
> The current reclaim_clean_pages_from_list assumes that shrink_page_list
> returns number of pages removed from the candidate list.
> 
> i.e)
> isolate_migratepages_range    : NR_ISOLATED_FILE += 10
> reclaim_clean_pages_from_list : NR_ISOLATED_FILE -= 5 (1 mlocked page)
> migrate_pages                 : NR_ISOLATED_FILE -=4
> => NR_ISOLATED_FILE increased by 1

Thanks for the clarity.

I think the problem is shrink_page_list is awkard. It put back to
unevictable pages instantly instead of passing it to caller while
it relies on caller for non-reclaimed-non-unevictable page's putback.

I think we can make it consistent so that shrink_page_list could
return non-reclaimed pages via page_list and caller can handle it.
As a bonus, it could try to migrate mlocked pages without retrial.

> 
> Thank you.
> >>
> >> Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
> >> ---
> >>  mm/vmscan.c | 10 ++++++++--
> >>  1 file changed, 8 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/mm/vmscan.c b/mm/vmscan.c
> >> index 5e8eadd..5837695 100644
> >> --- a/mm/vmscan.c
> >> +++ b/mm/vmscan.c
> >> @@ -849,6 +849,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
> >>  				      unsigned long *ret_nr_congested,
> >>  				      unsigned long *ret_nr_writeback,
> >>  				      unsigned long *ret_nr_immediate,
> >> +				      unsigned long *ret_nr_mlocked,
> >>  				      bool force_reclaim)
> >>  {
> >>  	LIST_HEAD(ret_pages);
> >> @@ -1158,6 +1159,7 @@ cull_mlocked:
> >>  			try_to_free_swap(page);
> >>  		unlock_page(page);
> >>  		putback_lru_page(page);
> >> +		(*ret_nr_mlocked)++;
> >>  		continue;
> >>  
> >>  activate_locked:
> >> @@ -1197,6 +1199,7 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
> >>  		.may_unmap = 1,
> >>  	};
> >>  	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
> >> +	unsigned long nr_mlocked = 0;
> >>  	struct page *page, *next;
> >>  	LIST_HEAD(clean_pages);
> >>  
> >> @@ -1210,8 +1213,10 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
> >>  
> >>  	ret = shrink_page_list(&clean_pages, zone, &sc,
> >>  			TTU_UNMAP|TTU_IGNORE_ACCESS,
> >> -			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
> >> +			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5,
> >> +			&nr_mlocked, true);
> >>  	list_splice(&clean_pages, page_list);
> >> +	ret += nr_mlocked;
> >>  	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
> >>  	return ret;
> >>  }
> >> @@ -1523,6 +1528,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
> >>  	unsigned long nr_unqueued_dirty = 0;
> >>  	unsigned long nr_writeback = 0;
> >>  	unsigned long nr_immediate = 0;
> >> +	unsigned long nr_mlocked = 0;
> >>  	isolate_mode_t isolate_mode = 0;
> >>  	int file = is_file_lru(lru);
> >>  	struct zone *zone = lruvec_zone(lruvec);
> >> @@ -1565,7 +1571,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
> >>  
> >>  	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
> >>  				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
> >> -				&nr_writeback, &nr_immediate,
> >> +				&nr_writeback, &nr_immediate, &nr_mlocked,
> >>  				false);
> >>  
> >>  	spin_lock_irq(&zone->lru_lock);
> >> -- 
> >> 1.9.1
> >>
> > 

-- 
Kind regards,
Minchan Kim

Re: [PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[Jaewon Kim]

On 2015년 08월 04일 00:33, Minchan Kim wrote:
> On Mon, Aug 03, 2015 at 11:55:46PM +0900, Jaewon Kim wrote:
>>
>>
>> On 2015년 08월 03일 21:27, Minchan Kim wrote:
>>> Hello,
>>>
>>> On Mon, Aug 03, 2015 at 07:18:27PM +0900, Jaewon Kim wrote:
>>>> reclaim_clean_pages_from_list() decreases NR_ISOLATED_FILE by returned
>>>> value from shrink_page_list(). But mlocked pages in the isolated
>>>> clean_pages page list would be removed from the list but not counted as
>>>> nr_reclaimed. Fix this miscounting by returning the number of mlocked
>>>> pages and count it.
>>>
>>> If there are pages not able to reclaim, VM try to migrate it and
>>> have to handle the stat in migrate_pages.
>>> If migrate_pages fails again, putback-fiends should handle it.
>>>
>>> Is there anyting I am missing now?
>>>
>>> Thanks.
>>>
>> Hello
>>
>> Only pages in cc->migratepages will be handled by migrate_pages or
>> putback_movable_pages, and NR_ISOLATED_FILE will be counted properly.
>> However mlocked pages will not be put back into cc->migratepages,
>> and also not be counted in NR_ISOLATED_FILE because putback_lru_page
>> in shrink_page_list does not increase NR_ISOLATED_FILE.
>> The current reclaim_clean_pages_from_list assumes that shrink_page_list
>> returns number of pages removed from the candidate list.
>>
>> i.e)
>> isolate_migratepages_range    : NR_ISOLATED_FILE += 10
>> reclaim_clean_pages_from_list : NR_ISOLATED_FILE -= 5 (1 mlocked page)
>> migrate_pages                 : NR_ISOLATED_FILE -=4
>> => NR_ISOLATED_FILE increased by 1
> 
> Thanks for the clarity.
> 
> I think the problem is shrink_page_list is awkard. It put back to
> unevictable pages instantly instead of passing it to caller while
> it relies on caller for non-reclaimed-non-unevictable page's putback.
> 
> I think we can make it consistent so that shrink_page_list could
> return non-reclaimed pages via page_list and caller can handle it.
> As a bonus, it could try to migrate mlocked pages without retrial.
> 
>>
>> Thank you.

To make clear do you mean changing shrink_page_list like this rather than
previous my suggestion?

@@ -1157,7 +1157,7 @@ cull_mlocked:
                if (PageSwapCache(page))
                        try_to_free_swap(page);
                unlock_page(page);
-               putback_lru_page(page);
+               list_add(&page->lru, &ret_pages);
                continue;

Thank you.


>>>>
>>>> Signed-off-by: Jaewon Kim <jaewon31.kim@samsung.com>
>>>> ---
>>>>  mm/vmscan.c | 10 ++++++++--
>>>>  1 file changed, 8 insertions(+), 2 deletions(-)
>>>>
>>>> diff --git a/mm/vmscan.c b/mm/vmscan.c
>>>> index 5e8eadd..5837695 100644
>>>> --- a/mm/vmscan.c
>>>> +++ b/mm/vmscan.c
>>>> @@ -849,6 +849,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
>>>>  				      unsigned long *ret_nr_congested,
>>>>  				      unsigned long *ret_nr_writeback,
>>>>  				      unsigned long *ret_nr_immediate,
>>>> +				      unsigned long *ret_nr_mlocked,
>>>>  				      bool force_reclaim)
>>>>  {
>>>>  	LIST_HEAD(ret_pages);
>>>> @@ -1158,6 +1159,7 @@ cull_mlocked:
>>>>  			try_to_free_swap(page);
>>>>  		unlock_page(page);
>>>>  		putback_lru_page(page);
>>>> +		(*ret_nr_mlocked)++;
>>>>  		continue;
>>>>  
>>>>  activate_locked:
>>>> @@ -1197,6 +1199,7 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
>>>>  		.may_unmap = 1,
>>>>  	};
>>>>  	unsigned long ret, dummy1, dummy2, dummy3, dummy4, dummy5;
>>>> +	unsigned long nr_mlocked = 0;
>>>>  	struct page *page, *next;
>>>>  	LIST_HEAD(clean_pages);
>>>>  
>>>> @@ -1210,8 +1213,10 @@ unsigned long reclaim_clean_pages_from_list(struct zone *zone,
>>>>  
>>>>  	ret = shrink_page_list(&clean_pages, zone, &sc,
>>>>  			TTU_UNMAP|TTU_IGNORE_ACCESS,
>>>> -			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5, true);
>>>> +			&dummy1, &dummy2, &dummy3, &dummy4, &dummy5,
>>>> +			&nr_mlocked, true);
>>>>  	list_splice(&clean_pages, page_list);
>>>> +	ret += nr_mlocked;
>>>>  	mod_zone_page_state(zone, NR_ISOLATED_FILE, -ret);
>>>>  	return ret;
>>>>  }
>>>> @@ -1523,6 +1528,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>>>>  	unsigned long nr_unqueued_dirty = 0;
>>>>  	unsigned long nr_writeback = 0;
>>>>  	unsigned long nr_immediate = 0;
>>>> +	unsigned long nr_mlocked = 0;
>>>>  	isolate_mode_t isolate_mode = 0;
>>>>  	int file = is_file_lru(lru);
>>>>  	struct zone *zone = lruvec_zone(lruvec);
>>>> @@ -1565,7 +1571,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>>>>  
>>>>  	nr_reclaimed = shrink_page_list(&page_list, zone, sc, TTU_UNMAP,
>>>>  				&nr_dirty, &nr_unqueued_dirty, &nr_congested,
>>>> -				&nr_writeback, &nr_immediate,
>>>> +				&nr_writeback, &nr_immediate, &nr_mlocked,
>>>>  				false);
>>>>  
>>>>  	spin_lock_irq(&zone->lru_lock);
>>>> -- 
>>>> 1.9.1
>>>>
>>>
> 

Re: [PATCH] vmscan: reclaim_clean_pages_from_list() must count mlocked pages

[Minchan Kim]

Hello,

On Tue, Aug 04, 2015 at 12:46:57AM +0900, Jaewon Kim wrote:
> 
> 
> On 2015년 08월 04일 00:33, Minchan Kim wrote:
> > On Mon, Aug 03, 2015 at 11:55:46PM +0900, Jaewon Kim wrote:
> >>
> >>
> >> On 2015년 08월 03일 21:27, Minchan Kim wrote:
> >>> Hello,
> >>>
> >>> On Mon, Aug 03, 2015 at 07:18:27PM +0900, Jaewon Kim wrote:
> >>>> reclaim_clean_pages_from_list() decreases NR_ISOLATED_FILE by returned
> >>>> value from shrink_page_list(). But mlocked pages in the isolated
> >>>> clean_pages page list would be removed from the list but not counted as
> >>>> nr_reclaimed. Fix this miscounting by returning the number of mlocked
> >>>> pages and count it.
> >>>
> >>> If there are pages not able to reclaim, VM try to migrate it and
> >>> have to handle the stat in migrate_pages.
> >>> If migrate_pages fails again, putback-fiends should handle it.
> >>>
> >>> Is there anyting I am missing now?
> >>>
> >>> Thanks.
> >>>
> >> Hello
> >>
> >> Only pages in cc->migratepages will be handled by migrate_pages or
> >> putback_movable_pages, and NR_ISOLATED_FILE will be counted properly.
> >> However mlocked pages will not be put back into cc->migratepages,
> >> and also not be counted in NR_ISOLATED_FILE because putback_lru_page
> >> in shrink_page_list does not increase NR_ISOLATED_FILE.
> >> The current reclaim_clean_pages_from_list assumes that shrink_page_list
> >> returns number of pages removed from the candidate list.
> >>
> >> i.e)
> >> isolate_migratepages_range    : NR_ISOLATED_FILE += 10
> >> reclaim_clean_pages_from_list : NR_ISOLATED_FILE -= 5 (1 mlocked page)
> >> migrate_pages                 : NR_ISOLATED_FILE -=4
> >> => NR_ISOLATED_FILE increased by 1
> > 
> > Thanks for the clarity.
> > 
> > I think the problem is shrink_page_list is awkard. It put back to
> > unevictable pages instantly instead of passing it to caller while
> > it relies on caller for non-reclaimed-non-unevictable page's putback.
> > 
> > I think we can make it consistent so that shrink_page_list could
> > return non-reclaimed pages via page_list and caller can handle it.
> > As a bonus, it could try to migrate mlocked pages without retrial.
> > 
> >>
> >> Thank you.
> 
> To make clear do you mean changing shrink_page_list like this rather than
> previous my suggestion?
> 
> @@ -1157,7 +1157,7 @@ cull_mlocked:
>                 if (PageSwapCache(page))
>                         try_to_free_swap(page);
>                 unlock_page(page);
> -               putback_lru_page(page);
> +               list_add(&page->lru, &ret_pages);
>                 continue;

Yes. That's what I said.

Thanks.

-- 
Kind regards,
Minchan Kim