Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1534858 > unrolled thread

[PATCH 0/2] High-order per-cpu cache v6

Back to article view | Back to linux.kernel

Contents

  [PATCH 0/2] High-order per-cpu cache v6 Mel Gorman <mgorman@techsingularity.net> - 2016-12-02 12:40 +0100
    [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in sync if struct page is corrupted Mel Gorman <mgorman@techsingularity.net> - 2016-12-02 12:40 +0100
      Re: [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in  sync if struct page is corrupted Vlastimil Babka <vbabka@suse.cz> - 2016-12-02 13:00 +0100
      Re: [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in  sync if struct page is corrupted Michal Hocko <mhocko@kernel.org> - 2016-12-02 14:20 +0100
      Re: [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in sync if struct page is corrupted "Hillf Danton" <hillf.zj@alibaba-inc.com> - 2016-12-05 04:50 +0100

#1534858 — [PATCH 0/2] High-order per-cpu cache v6

Changelog since v5
o Changelog clarification in patch 1
o Additional comments in patch 2

Changelog since v4
o Avoid pcp->count getting out of sync if struct page gets corrupted

Changelog since v3
o Allow high-order atomic allocations to use reserves

Changelog since v2
o Correct initialisation to avoid -Woverflow warning

The following is two patches that implement a per-cpu cache for high-order
allocations, primarily aimed at SLUB. The first patch is a bug fix that
is technically unrelated but was discovered by review and so batched
together. The second is the patch that implements the high-order pcpu cache.

 include/linux/mmzone.h |  20 +++++++-
 mm/page_alloc.c        | 129 ++++++++++++++++++++++++++++++++-----------------
 2 files changed, 103 insertions(+), 46 deletions(-)

-- 
2.10.2

[toc] | [next] | [standalone]

#1534860 — [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in sync if struct page is corrupted

Vlastimil Babka pointed out that commit 479f854a207c ("mm, page_alloc:
defer debugging checks of pages allocated from the PCP") will allow the
per-cpu list counter to be out of sync with the per-cpu list contents
if a struct page is corrupted.

The consequence is an infinite loop if the per-cpu lists get fully drained
by free_pcppages_bulk because all the lists are empty but the count is
positive. The infinite loop occurs here

                do {
                        batch_free++;
                        if (++migratetype == MIGRATE_PCPTYPES)
                                migratetype = 0;
                        list = &pcp->lists[migratetype];
                } while (list_empty(list));

From a user perspective, it's a bad page warning followed by a soft lockup
with interrupts disabled in free_pcppages_bulk().

This patch keeps the accounting in sync.

Fixes: 479f854a207c ("mm, page_alloc: defer debugging checks of pages allocated from the PCP")
Signed-off-by: Mel Gorman <mgorman@suse.de>
cc: stable@vger.kernel.org [4.7+]
---
 mm/page_alloc.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 6de9440e3ae2..34ada718ef47 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2192,7 +2192,7 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 			unsigned long count, struct list_head *list,
 			int migratetype, bool cold)
 {
-	int i;
+	int i, alloced = 0;
 
 	spin_lock(&zone->lock);
 	for (i = 0; i < count; ++i) {
@@ -2217,13 +2217,21 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
 		else
 			list_add_tail(&page->lru, list);
 		list = &page->lru;
+		alloced++;
 		if (is_migrate_cma(get_pcppage_migratetype(page)))
 			__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
 					      -(1 << order));
 	}
+
+	/*
+	 * i pages were removed from the buddy list even if some leak due
+	 * to check_pcp_refill failing so adjust NR_FREE_PAGES based
+	 * on i. Do not confuse with 'alloced' which is the number of
+	 * pages added to the pcp list.
+	 */
 	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
 	spin_unlock(&zone->lock);
-	return i;
+	return alloced;
 }
 
 #ifdef CONFIG_NUMA
-- 
2.10.2

[toc] | [prev] | [next] | [standalone]

#1534875 — Re: [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in sync if struct page is corrupted

On 12/02/2016 12:29 PM, Mel Gorman wrote:
> Vlastimil Babka pointed out that commit 479f854a207c ("mm, page_alloc:
> defer debugging checks of pages allocated from the PCP") will allow the
> per-cpu list counter to be out of sync with the per-cpu list contents
> if a struct page is corrupted.
>
> The consequence is an infinite loop if the per-cpu lists get fully drained
> by free_pcppages_bulk because all the lists are empty but the count is
> positive. The infinite loop occurs here
>
>                 do {
>                         batch_free++;
>                         if (++migratetype == MIGRATE_PCPTYPES)
>                                 migratetype = 0;
>                         list = &pcp->lists[migratetype];
>                 } while (list_empty(list));
>
> From a user perspective, it's a bad page warning followed by a soft lockup
> with interrupts disabled in free_pcppages_bulk().
>
> This patch keeps the accounting in sync.
>
> Fixes: 479f854a207c ("mm, page_alloc: defer debugging checks of pages allocated from the PCP")
> Signed-off-by: Mel Gorman <mgorman@suse.de>
> cc: stable@vger.kernel.org [4.7+]

Acked-by: Vlastimil Babka <vbabka@suse.cz>

> ---
>  mm/page_alloc.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 6de9440e3ae2..34ada718ef47 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2192,7 +2192,7 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
>  			unsigned long count, struct list_head *list,
>  			int migratetype, bool cold)
>  {
> -	int i;
> +	int i, alloced = 0;
>
>  	spin_lock(&zone->lock);
>  	for (i = 0; i < count; ++i) {
> @@ -2217,13 +2217,21 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
>  		else
>  			list_add_tail(&page->lru, list);
>  		list = &page->lru;
> +		alloced++;
>  		if (is_migrate_cma(get_pcppage_migratetype(page)))
>  			__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
>  					      -(1 << order));
>  	}
> +
> +	/*
> +	 * i pages were removed from the buddy list even if some leak due
> +	 * to check_pcp_refill failing so adjust NR_FREE_PAGES based
> +	 * on i. Do not confuse with 'alloced' which is the number of
> +	 * pages added to the pcp list.
> +	 */
>  	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
>  	spin_unlock(&zone->lock);
> -	return i;
> +	return alloced;
>  }
>
>  #ifdef CONFIG_NUMA
>

[toc] | [prev] | [next] | [standalone]

#1534905 — Re: [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in sync if struct page is corrupted

On Fri 02-12-16 11:29:50, Mel Gorman wrote:
> Vlastimil Babka pointed out that commit 479f854a207c ("mm, page_alloc:
> defer debugging checks of pages allocated from the PCP") will allow the
> per-cpu list counter to be out of sync with the per-cpu list contents
> if a struct page is corrupted.
> 
> The consequence is an infinite loop if the per-cpu lists get fully drained
> by free_pcppages_bulk because all the lists are empty but the count is
> positive. The infinite loop occurs here
> 
>                 do {
>                         batch_free++;
>                         if (++migratetype == MIGRATE_PCPTYPES)
>                                 migratetype = 0;
>                         list = &pcp->lists[migratetype];
>                 } while (list_empty(list));
> 
> >From a user perspective, it's a bad page warning followed by a soft lockup
> with interrupts disabled in free_pcppages_bulk().
> 
> This patch keeps the accounting in sync.
> 
> Fixes: 479f854a207c ("mm, page_alloc: defer debugging checks of pages allocated from the PCP")
> Signed-off-by: Mel Gorman <mgorman@suse.de>
> cc: stable@vger.kernel.org [4.7+]

Thanks for adding the comment it should really make the code more clear.

Acked-by: Michal Hocko <mhocko@suse.com>

> ---
>  mm/page_alloc.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 6de9440e3ae2..34ada718ef47 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2192,7 +2192,7 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
>  			unsigned long count, struct list_head *list,
>  			int migratetype, bool cold)
>  {
> -	int i;
> +	int i, alloced = 0;
>  
>  	spin_lock(&zone->lock);
>  	for (i = 0; i < count; ++i) {
> @@ -2217,13 +2217,21 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
>  		else
>  			list_add_tail(&page->lru, list);
>  		list = &page->lru;
> +		alloced++;
>  		if (is_migrate_cma(get_pcppage_migratetype(page)))
>  			__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
>  					      -(1 << order));
>  	}
> +
> +	/*
> +	 * i pages were removed from the buddy list even if some leak due
> +	 * to check_pcp_refill failing so adjust NR_FREE_PAGES based
> +	 * on i. Do not confuse with 'alloced' which is the number of
> +	 * pages added to the pcp list.
> +	 */
>  	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
>  	spin_unlock(&zone->lock);
> -	return i;
> +	return alloced;
>  }
>  
>  #ifdef CONFIG_NUMA
> -- 
> 2.10.2
> 

-- 
Michal Hocko
SUSE Labs

[toc] | [prev] | [next] | [standalone]

#1535791 — Re: [PATCH 1/2] mm, page_alloc: Keep pcp count and list contents in sync if struct page is corrupted

On Friday, December 02, 2016 7:30 PM Mel Gorman wrote:
> Vlastimil Babka pointed out that commit 479f854a207c ("mm, page_alloc:
> defer debugging checks of pages allocated from the PCP") will allow the
> per-cpu list counter to be out of sync with the per-cpu list contents
> if a struct page is corrupted.
> 
> The consequence is an infinite loop if the per-cpu lists get fully drained
> by free_pcppages_bulk because all the lists are empty but the count is
> positive. The infinite loop occurs here
> 
>                 do {
>                         batch_free++;
>                         if (++migratetype == MIGRATE_PCPTYPES)
>                                 migratetype = 0;
>                         list = &pcp->lists[migratetype];
>                 } while (list_empty(list));
> 
> From a user perspective, it's a bad page warning followed by a soft lockup
> with interrupts disabled in free_pcppages_bulk().
> 
> This patch keeps the accounting in sync.
> 
> Fixes: 479f854a207c ("mm, page_alloc: defer debugging checks of pages allocated from the PCP")
> Signed-off-by: Mel Gorman <mgorman@suse.de>
> cc: stable@vger.kernel.org [4.7+]
> ---
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>

>  mm/page_alloc.c | 12 ++++++++++--
>  1 file changed, 10 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 6de9440e3ae2..34ada718ef47 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -2192,7 +2192,7 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
>  			unsigned long count, struct list_head *list,
>  			int migratetype, bool cold)
>  {
> -	int i;
> +	int i, alloced = 0;
> 
>  	spin_lock(&zone->lock);
>  	for (i = 0; i < count; ++i) {
> @@ -2217,13 +2217,21 @@ static int rmqueue_bulk(struct zone *zone, unsigned int order,
>  		else
>  			list_add_tail(&page->lru, list);
>  		list = &page->lru;
> +		alloced++;
>  		if (is_migrate_cma(get_pcppage_migratetype(page)))
>  			__mod_zone_page_state(zone, NR_FREE_CMA_PAGES,
>  					      -(1 << order));
>  	}
> +
> +	/*
> +	 * i pages were removed from the buddy list even if some leak due
> +	 * to check_pcp_refill failing so adjust NR_FREE_PAGES based
> +	 * on i. Do not confuse with 'alloced' which is the number of
> +	 * pages added to the pcp list.
> +	 */
>  	__mod_zone_page_state(zone, NR_FREE_PAGES, -(i << order));
>  	spin_unlock(&zone->lock);
> -	return i;
> +	return alloced;
>  }
> 
>  #ifdef CONFIG_NUMA
> --
> 2.10.2

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web