Re: [PATCH v3 15/21] mm: handle lruvec relocks on lumpy reclaim

[KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>]

On Thu, 23 Feb 2012 17:52:52 +0400
Konstantin Khlebnikov <khlebnikov@openvz.org> wrote:

> Prepare for lock splitting in lumly reclaim logic.
> Now move_active_pages_to_lru() and putback_inactive_pages()
> can put pages into different lruvecs.
> 
> * relock book before SetPageLRU()

lruvec ?

> * update reclaim_stat pointer after relocks
> * return currently locked lruvec
> 
> Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
> ---
>  mm/vmscan.c |   45 +++++++++++++++++++++++++++++++++------------
>  1 files changed, 33 insertions(+), 12 deletions(-)
> 
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index a3941d1..6eeeb4b 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -1114,6 +1114,7 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
>  		unsigned long *nr_scanned, struct scan_control *sc,
>  		isolate_mode_t mode, int active, int file)
>  {
> +	struct lruvec *cursor_lruvec = lruvec;
>  	struct list_head *src;
>  	unsigned long nr_taken = 0;
>  	unsigned long nr_lumpy_taken = 0;
> @@ -1197,14 +1198,17 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
>  			    !PageSwapCache(cursor_page))
>  				break;
>  
> +			/* Switch cursor_lruvec lock for lumpy isolate */
> +			if (!__lock_page_lruvec_irq(&cursor_lruvec,
> +						    cursor_page))
> +				continue;
> +
>  			if (__isolate_lru_page(cursor_page, mode, file) == 0) {
>  				unsigned int isolated_pages;
> -				struct lruvec *cursor_lruvec;
>  				int cursor_lru = page_lru(cursor_page);
>  
>  				list_move(&cursor_page->lru, dst);
>  				isolated_pages = hpage_nr_pages(cursor_page);
> -				cursor_lruvec = page_lruvec(cursor_page);
>  				cursor_lruvec->pages_count[cursor_lru] -=
>  								isolated_pages;
>  				VM_BUG_ON((long)cursor_lruvec->
> @@ -1235,6 +1239,9 @@ static unsigned long isolate_lru_pages(unsigned long nr_to_scan,
>  			}
>  		}
>  
> +		/* Restore original lruvec lock */
> +		cursor_lruvec = __relock_page_lruvec(cursor_lruvec, page);
> +
>  		/* If we break out of the loop above, lumpy reclaim failed */
>  		if (pfn < end_pfn)
>  			nr_lumpy_failed++;
> @@ -1325,7 +1332,10 @@ static int too_many_isolated(struct zone *zone, int file,
>  	return isolated > inactive;
>  }
>  
> -static noinline_for_stack void
> +/*
> + * Returns currently locked lruvec
> + */
> +static noinline_for_stack struct lruvec *
>  putback_inactive_pages(struct lruvec *lruvec,
>  		       struct list_head *page_list)
>  {
> @@ -1347,10 +1357,13 @@ putback_inactive_pages(struct lruvec *lruvec,
>  			lock_lruvec_irq(lruvec);
>  			continue;
>  		}
> +
> +		lruvec = __relock_page_lruvec(lruvec, page);
> +		reclaim_stat = &lruvec->reclaim_stat;
> +
>  		SetPageLRU(page);
>  		lru = page_lru(page);
>  
> -		lruvec = page_lruvec(page);
>  		add_page_to_lru_list(lruvec, page, lru);
>  		if (is_active_lru(lru)) {
>  			int file = is_file_lru(lru);
> @@ -1375,6 +1388,8 @@ putback_inactive_pages(struct lruvec *lruvec,
>  	 * To save our caller's stack, now use input list for pages to free.
>  	 */
>  	list_splice(&pages_to_free, page_list);
> +
> +	return lruvec;
>  }
>  
>  static noinline_for_stack void
> @@ -1544,7 +1559,7 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>  		__count_vm_events(KSWAPD_STEAL, nr_reclaimed);
>  	__count_zone_vm_events(PGSTEAL, zone, nr_reclaimed);
>  
> -	putback_inactive_pages(lruvec, &page_list);
> +	lruvec = putback_inactive_pages(lruvec, &page_list);
>  
>  	__mod_zone_page_state(zone, NR_ISOLATED_ANON, -nr_anon);
>  	__mod_zone_page_state(zone, NR_ISOLATED_FILE, -nr_file);
> @@ -1603,12 +1618,15 @@ shrink_inactive_list(unsigned long nr_to_scan, struct lruvec *lruvec,
>   *
>   * The downside is that we have to touch page->_count against each page.
>   * But we had to alter page->flags anyway.
> + *
> + * Returns currently locked lruvec
>   */
>  
> -static void move_active_pages_to_lru(struct lruvec *lruvec,
> -				     struct list_head *list,
> -				     struct list_head *pages_to_free,
> -				     enum lru_list lru)
> +static struct lruvec *
> +move_active_pages_to_lru(struct lruvec *lruvec,
> +			 struct list_head *list,
> +			 struct list_head *pages_to_free,
> +			 enum lru_list lru)
>  {
>  	unsigned long pgmoved = 0;
>  	struct page *page;
> @@ -1630,10 +1648,11 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
>  
>  		page = lru_to_page(list);
>  
> +		lruvec = __relock_page_lruvec(lruvec, page);
> +
>  		VM_BUG_ON(PageLRU(page));
>  		SetPageLRU(page);
>  
> -		lruvec = page_lruvec(page);
>  		list_move(&page->lru, &lruvec->pages_lru[lru]);
>  		numpages = hpage_nr_pages(page);
>  		lruvec->pages_count[lru] += numpages;
> @@ -1655,6 +1674,8 @@ static void move_active_pages_to_lru(struct lruvec *lruvec,
>  	__mod_zone_page_state(lruvec_zone(lruvec), NR_LRU_BASE + lru, pgmoved);
>  	if (!is_active_lru(lru))
>  		__count_vm_events(PGDEACTIVATE, pgmoved);
> +
> +	return lruvec;
>  }
>  
>  static void shrink_active_list(unsigned long nr_to_scan,
> @@ -1744,9 +1765,9 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  	 */
>  	reclaim_stat->recent_rotated[file] += nr_rotated;
>  
> -	move_active_pages_to_lru(lruvec, &l_active, &l_hold,
> +	lruvec = move_active_pages_to_lru(lruvec, &l_active, &l_hold,
>  						LRU_ACTIVE + file * LRU_FILE);
> -	move_active_pages_to_lru(lruvec, &l_inactive, &l_hold,
> +	lruvec = move_active_pages_to_lru(lruvec, &l_inactive, &l_hold,
>  						LRU_BASE   + file * LRU_FILE);
>  	__mod_zone_page_state(zone, NR_ISOLATED_ANON + file, -nr_taken);
>  	unlock_lruvec_irq(lruvec);
> 

Hmm...could you add comments to each function as
"The caller should _lock_ lruvec before calling this functions.
 This function returns a lruvec with _locked_. It may be different from passed one.
 And The callser should unlock lruvec"


Thanks,
-Kame




--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>