[PATCH] mm:fix duplicate accounting of free pages in should_reclaim_retry()

[PATCH] mm:fix duplicate accounting of free pages in should_reclaim_retry()

[liuqiqi]

From: liuqiqi <liuqiqi@kylinos.cn>

In the zone_reclaimable_pages() function, if the page counts for
NR_ZONE_INACTIVE_FILE, NR_ZONE_ACTIVE_FILE, NR_ZONE_INACTIVE_ANON,
and NR_ZONE_ACTIVE_ANON are all zero,
the function returns the number of free pages as the result.

In this case, when should_reclaim_retry() calculates reclaimable pages,
it will inadvertently double-count the free pages in its accounting.

static inline bool
should_reclaim_retry(gfp_t gfp_mask, unsigned order,
                     struct alloc_context *ac, int alloc_flags,
                     bool did_some_progress, int *no_progress_loops)
{
        ...
                available = reclaimable = zone_reclaimable_pages(zone);
                available += zone_page_state_snapshot(zone, NR_FREE_PAGES);

Signed-off-by: liuqiqi <liuqiqi@kylinos.cn>
---
 mm/vmscan.c | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 34410d24dc15..a9aaefdba7a2 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -393,14 +393,7 @@ unsigned long zone_reclaimable_pages(struct zone *zone)
 	if (can_reclaim_anon_pages(NULL, zone_to_nid(zone), NULL))
 		nr += zone_page_state_snapshot(zone, NR_ZONE_INACTIVE_ANON) +
 			zone_page_state_snapshot(zone, NR_ZONE_ACTIVE_ANON);
-	/*
-	 * If there are no reclaimable file-backed or anonymous pages,
-	 * ensure zones with sufficient free pages are not skipped.
-	 * This prevents zones like DMA32 from being ignored in reclaim
-	 * scenarios where they can still help alleviate memory pressure.
-	 */
-	if (nr == 0)
-		nr = zone_page_state_snapshot(zone, NR_FREE_PAGES);
+
 	return nr;
 }
 
@@ -6417,7 +6410,7 @@ static bool allow_direct_reclaim(pg_data_t *pgdat)
 		return true;
 
 	for_each_managed_zone_pgdat(zone, pgdat, i, ZONE_NORMAL) {
-		if (!zone_reclaimable_pages(zone))
+		if (!zone_reclaimable_pages(zone) && zone_page_state_snapshot(zone, NR_FREE_PAGES))
 			continue;
 
 		pfmemalloc_reserve += min_wmark_pages(zone);
-- 
2.25.1

Re: [PATCH] mm:fix duplicate accounting of free pages in should_reclaim_retry()

[Andrew Morton]

On Tue, 12 Aug 2025 15:02:10 +0800 liuqiqi@kylinos.cn wrote:

> From: liuqiqi <liuqiqi@kylinos.cn>
> 
> In the zone_reclaimable_pages() function, if the page counts for
> NR_ZONE_INACTIVE_FILE, NR_ZONE_ACTIVE_FILE, NR_ZONE_INACTIVE_ANON,
> and NR_ZONE_ACTIVE_ANON are all zero,
> the function returns the number of free pages as the result.
> 
> In this case, when should_reclaim_retry() calculates reclaimable pages,
> it will inadvertently double-count the free pages in its accounting.
> 
> static inline bool
> should_reclaim_retry(gfp_t gfp_mask, unsigned order,
>                      struct alloc_context *ac, int alloc_flags,
>                      bool did_some_progress, int *no_progress_loops)
> {
>         ...
>                 available = reclaimable = zone_reclaimable_pages(zone);
>                 available += zone_page_state_snapshot(zone, NR_FREE_PAGES);

Thanks.  Does this have any significant runtime effects?

> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -393,14 +393,7 @@ unsigned long zone_reclaimable_pages(struct zone *zone)
>  	if (can_reclaim_anon_pages(NULL, zone_to_nid(zone), NULL))
>  		nr += zone_page_state_snapshot(zone, NR_ZONE_INACTIVE_ANON) +
>  			zone_page_state_snapshot(zone, NR_ZONE_ACTIVE_ANON);
> -	/*
> -	 * If there are no reclaimable file-backed or anonymous pages,
> -	 * ensure zones with sufficient free pages are not skipped.
> -	 * This prevents zones like DMA32 from being ignored in reclaim
> -	 * scenarios where they can still help alleviate memory pressure.
> -	 */
> -	if (nr == 0)
> -		nr = zone_page_state_snapshot(zone, NR_FREE_PAGES);
> +
>  	return nr;
>  }
>  
> @@ -6417,7 +6410,7 @@ static bool allow_direct_reclaim(pg_data_t *pgdat)
>  		return true;
>  
>  	for_each_managed_zone_pgdat(zone, pgdat, i, ZONE_NORMAL) {
> -		if (!zone_reclaimable_pages(zone))
> +		if (!zone_reclaimable_pages(zone) && zone_page_state_snapshot(zone, NR_FREE_PAGES))
>  			continue;
>  
>  		pfmemalloc_reserve += min_wmark_pages(zone);

Re: [PATCH] mm:fix duplicate accounting of free pages in should_reclaim_retry()

[Ye Liu]

在 2025/8/12 15:02, liuqiqi@kylinos.cn 写道:
> From: liuqiqi <liuqiqi@kylinos.cn>
>
> In the zone_reclaimable_pages() function, if the page counts for
> NR_ZONE_INACTIVE_FILE, NR_ZONE_ACTIVE_FILE, NR_ZONE_INACTIVE_ANON,
> and NR_ZONE_ACTIVE_ANON are all zero,
> the function returns the number of free pages as the result.
>
> In this case, when should_reclaim_retry() calculates reclaimable pages,
> it will inadvertently double-count the free pages in its accounting.
>
> static inline bool
> should_reclaim_retry(gfp_t gfp_mask, unsigned order,
>                      struct alloc_context *ac, int alloc_flags,
>                      bool did_some_progress, int *no_progress_loops)
> {
>         ...
>                 available = reclaimable = zone_reclaimable_pages(zone);
>                 available += zone_page_state_snapshot(zone, NR_FREE_PAGES);
A "fixes" should be added here.                                                                                 
Fixes: 6aaced5abd32 ("mm: vmscan: account for free pages to prevent infinite Loop in throttle_direct_reclaim()")
Reviewed-by: Ye Liu <liuye@kylinos.cn>
> Signed-off-by: liuqiqi <liuqiqi@kylinos.cn>
> ---
>  mm/vmscan.c | 11 ++---------
>  1 file changed, 2 insertions(+), 9 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 34410d24dc15..a9aaefdba7a2 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -393,14 +393,7 @@ unsigned long zone_reclaimable_pages(struct zone *zone)
>  	if (can_reclaim_anon_pages(NULL, zone_to_nid(zone), NULL))
>  		nr += zone_page_state_snapshot(zone, NR_ZONE_INACTIVE_ANON) +
>  			zone_page_state_snapshot(zone, NR_ZONE_ACTIVE_ANON);
> -	/*
> -	 * If there are no reclaimable file-backed or anonymous pages,
> -	 * ensure zones with sufficient free pages are not skipped.
> -	 * This prevents zones like DMA32 from being ignored in reclaim
> -	 * scenarios where they can still help alleviate memory pressure.
> -	 */
> -	if (nr == 0)
> -		nr = zone_page_state_snapshot(zone, NR_FREE_PAGES);
> +
>  	return nr;
>  }
>  
> @@ -6417,7 +6410,7 @@ static bool allow_direct_reclaim(pg_data_t *pgdat)
>  		return true;
>  
>  	for_each_managed_zone_pgdat(zone, pgdat, i, ZONE_NORMAL) {
> -		if (!zone_reclaimable_pages(zone))
> +		if (!zone_reclaimable_pages(zone) && zone_page_state_snapshot(zone, NR_FREE_PAGES))
>  			continue;
>  
>  		pfmemalloc_reserve += min_wmark_pages(zone);

-- 
Thanks,
Ye Liu

Re: [PATCH] mm:fix duplicate accounting of free pages in should_reclaim_retry()

[liuqiqi]

Duplicate accounting of free pages in should_reclaim_retry() effects:
The number of retry in the __alloc_pages_slowpath() function has increased.
The execution time of the kswapd process has increased.

static inline struct page * 
__alloc_pages_slowpath(gfp_t gfp_mask, unsigned int order,
                                                struct alloc_context *ac)
{
	......
retry:
	/*
	 * Deal with possible cpuset update races or zonelist updates to avoid
	 * infinite retries.
	 */
	......
        if (should_reclaim_retry(gfp_mask, order, ac, alloc_flags,
                                 did_some_progress > 0, &no_progress_loops))
                goto retry;

The test program: continuously allocates 1k-sized memory through kmalloc(); 
counts the number of retry and the execution time of the kswapd process;
the test results also confirm this.

> Thanks. Does this have any significant runtime effects?

> In the zone_reclaimable_pages() function, if the page counts for
> NR_ZONE_INACTIVE_FILE, NR_ZONE_ACTIVE_FILE, NR_ZONE_INACTIVE_ANON,
> and NR_ZONE_ACTIVE_ANON are all zero,
> the function returns the number of free pages as the result.
>
> In this case, when should_reclaim_retry() calculates reclaimable pages,
> it will inadvertently double-count the free pages in its accounting.
>
> static inline bool
> should_reclaim_retry(gfp_t gfp_mask, unsigned order,
>                      struct alloc_context *ac, int alloc_flags,
>                      bool did_some_progress, int *no_progress_loops)
> {
>         ...
>                 available = reclaimable = zone_reclaimable_pages(zone);
>                 available += zone_page_state_snapshot(zone, NR_FREE_PAGES);

https://lore.kernel.org/all/20250817120016.8dcc091c5b7114d6993a29ae@linux-foundation.org/

---
Best Regards