[PATCH v2] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[PATCH v2] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Ruan Shiyang]

Goto-san reported confusing pgpromote statistics where the
pgpromote_success count significantly exceeded pgpromote_candidate.

On a system with three nodes (nodes 0-1: DRAM 4GB, node 2: NVDIMM 4GB):
 # Enable demotion only
 echo 1 > /sys/kernel/mm/numa/demotion_enabled
 numactl -m 0-1 memhog -r200 3500M >/dev/null &
 pid=$!
 sleep 2
 numactl memhog -r100 2500M >/dev/null &
 sleep 10
 kill -9 $pid # terminate the 1st memhog
 # Enable promotion
 echo 2 > /proc/sys/kernel/numa_balancing

After a few seconds, we observeed `pgpromote_candidate < pgpromote_success`
$ grep -e pgpromote /proc/vmstat
pgpromote_success 2579
pgpromote_candidate 0

In this scenario, after terminating the first memhog, the conditions for
pgdat_free_space_enough() are quickly met, and triggers promotion.
However, these migrated pages are only counted for in PGPROMOTE_SUCCESS,
not in PGPROMOTE_CANDIDATE.

To solve these confusing statistics, introduce PGPROMOTE_CANDIDATE_NRL to
count the missed promotion pages.  And also, not counting these pages into
PGPROMOTE_CANDIDATE is to avoid changing the existing algorithm or
performance of the promotion rate limit.

Link: https://lkml.kernel.org/r/20250729035101.1601407-1-ruansy.fnst@fujitsu.com
Co-developed-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Ruan Shiyang <ruansy.fnst@fujitsu.com>
Reported-by: Yasunori Gotou (Fujitsu) <y-goto@fujitsu.com>
Suggested-by: Huang Ying <ying.huang@linux.alibaba.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ben Segall <bsegall@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Valentin Schneider <vschneid@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
Changes since v1:
  1. change Li Zhijian from 'Signed-off-by' to 'Co-developed-by' per Vlastimil.
  2. add Acked-by: Vlastimil Babka <vbabka@suse.cz>
---
 include/linux/mmzone.h | 16 +++++++++++++++-
 kernel/sched/fair.c    |  5 +++--
 mm/vmstat.c            |  1 +
 3 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 0c5da9141983..9d3ea9085556 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -234,7 +234,21 @@ enum node_stat_item {
 #endif
 #ifdef CONFIG_NUMA_BALANCING
 	PGPROMOTE_SUCCESS,	/* promote successfully */
-	PGPROMOTE_CANDIDATE,	/* candidate pages to promote */
+	/**
+	 * Candidate pages for promotion based on hint fault latency.  This
+	 * counter is used to control the promotion rate and adjust the hot
+	 * threshold.
+	 */
+	PGPROMOTE_CANDIDATE,
+	/**
+	 * Not rate-limited (NRL) candidate pages for those can be promoted
+	 * without considering hot threshold because of enough free pages in
+	 * fast-tier node.  These promotions bypass the regular hotness checks
+	 * and do NOT influence the promotion rate-limiter or
+	 * threshold-adjustment logic.
+	 * This is for statistics/monitoring purposes.
+	 */
+	PGPROMOTE_CANDIDATE_NRL,
 #endif
 	/* PGDEMOTE_*: pages demoted */
 	PGDEMOTE_KSWAPD,
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index b173a059315c..82c8d804c54c 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1923,11 +1923,13 @@ bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
 		struct pglist_data *pgdat;
 		unsigned long rate_limit;
 		unsigned int latency, th, def_th;
+		long nr = folio_nr_pages(folio);
 
 		pgdat = NODE_DATA(dst_nid);
 		if (pgdat_free_space_enough(pgdat)) {
 			/* workload changed, reset hot threshold */
 			pgdat->nbp_threshold = 0;
+			mod_node_page_state(pgdat, PGPROMOTE_CANDIDATE_NRL, nr);
 			return true;
 		}
 
@@ -1941,8 +1943,7 @@ bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
 		if (latency >= th)
 			return false;
 
-		return !numa_promotion_rate_limit(pgdat, rate_limit,
-						  folio_nr_pages(folio));
+		return !numa_promotion_rate_limit(pgdat, rate_limit, nr);
 	}
 
 	this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid);
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 71cd1ceba191..e74f0b2a1021 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1280,6 +1280,7 @@ const char * const vmstat_text[] = {
 #ifdef CONFIG_NUMA_BALANCING
 	[I(PGPROMOTE_SUCCESS)]			= "pgpromote_success",
 	[I(PGPROMOTE_CANDIDATE)]		= "pgpromote_candidate",
+	[I(PGPROMOTE_CANDIDATE_NRL)]		= "pgpromote_candidate_nrl",
 #endif
 	[I(PGDEMOTE_KSWAPD)]			= "pgdemote_kswapd",
 	[I(PGDEMOTE_DIRECT)]			= "pgdemote_direct",
-- 
2.43.0

Re: [PATCH v2] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Vlastimil Babka]

On 9/1/25 04:05, Ruan Shiyang wrote:
> Goto-san reported confusing pgpromote statistics where the
> pgpromote_success count significantly exceeded pgpromote_candidate.
> 
> On a system with three nodes (nodes 0-1: DRAM 4GB, node 2: NVDIMM 4GB):
>  # Enable demotion only
>  echo 1 > /sys/kernel/mm/numa/demotion_enabled
>  numactl -m 0-1 memhog -r200 3500M >/dev/null &
>  pid=$!
>  sleep 2
>  numactl memhog -r100 2500M >/dev/null &
>  sleep 10
>  kill -9 $pid # terminate the 1st memhog
>  # Enable promotion
>  echo 2 > /proc/sys/kernel/numa_balancing
> 
> After a few seconds, we observeed `pgpromote_candidate < pgpromote_success`
> $ grep -e pgpromote /proc/vmstat
> pgpromote_success 2579
> pgpromote_candidate 0
> 
> In this scenario, after terminating the first memhog, the conditions for
> pgdat_free_space_enough() are quickly met, and triggers promotion.
> However, these migrated pages are only counted for in PGPROMOTE_SUCCESS,
> not in PGPROMOTE_CANDIDATE.
> 
> To solve these confusing statistics, introduce PGPROMOTE_CANDIDATE_NRL to
> count the missed promotion pages.  And also, not counting these pages into
> PGPROMOTE_CANDIDATE is to avoid changing the existing algorithm or
> performance of the promotion rate limit.
> 
> Link: https://lkml.kernel.org/r/20250729035101.1601407-1-ruansy.fnst@fujitsu.com
> Co-developed-by: Li Zhijian <lizhijian@fujitsu.com>
> Signed-off-by: Ruan Shiyang <ruansy.fnst@fujitsu.com>
> Reported-by: Yasunori Gotou (Fujitsu) <y-goto@fujitsu.com>
> Suggested-by: Huang Ying <ying.huang@linux.alibaba.com>
> Acked-by: Vlastimil Babka <vbabka@suse.cz>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Juri Lelli <juri.lelli@redhat.com>
> Cc: Vincent Guittot <vincent.guittot@linaro.org>
> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Ben Segall <bsegall@google.com>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Valentin Schneider <vschneid@redhat.com>
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> Changes since v1:
>   1. change Li Zhijian from 'Signed-off-by' to 'Co-developed-by' per Vlastimil.

Note according to the docs it should be both, Co-developed-by followed by
Signed-off-by.

[PATCH v3] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Ruan Shiyang]

Goto-san reported confusing pgpromote statistics where the
pgpromote_success count significantly exceeded pgpromote_candidate.

On a system with three nodes (nodes 0-1: DRAM 4GB, node 2: NVDIMM 4GB):
 # Enable demotion only
 echo 1 > /sys/kernel/mm/numa/demotion_enabled
 numactl -m 0-1 memhog -r200 3500M >/dev/null &
 pid=$!
 sleep 2
 numactl memhog -r100 2500M >/dev/null &
 sleep 10
 kill -9 $pid # terminate the 1st memhog
 # Enable promotion
 echo 2 > /proc/sys/kernel/numa_balancing

After a few seconds, we observeed `pgpromote_candidate < pgpromote_success`
$ grep -e pgpromote /proc/vmstat
pgpromote_success 2579
pgpromote_candidate 0

In this scenario, after terminating the first memhog, the conditions for
pgdat_free_space_enough() are quickly met, and triggers promotion.
However, these migrated pages are only counted for in PGPROMOTE_SUCCESS,
not in PGPROMOTE_CANDIDATE.

To solve these confusing statistics, introduce PGPROMOTE_CANDIDATE_NRL to
count the missed promotion pages.  And also, not counting these pages into
PGPROMOTE_CANDIDATE is to avoid changing the existing algorithm or
performance of the promotion rate limit.

Link: https://lkml.kernel.org/r/20250729035101.1601407-1-ruansy.fnst@fujitsu.com
Co-developed-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Signed-off-by: Ruan Shiyang <ruansy.fnst@fujitsu.com>
Reported-by: Yasunori Gotou (Fujitsu) <y-goto@fujitsu.com>
Suggested-by: Huang Ying <ying.huang@linux.alibaba.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Ben Segall <bsegall@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Valentin Schneider <vschneid@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---
Changes since v2:
  1. add 'Co-developed-by: Li Zhijian' followed by 'Signed-off-by' per Vlastimil.
---
 include/linux/mmzone.h | 16 +++++++++++++++-
 kernel/sched/fair.c    |  5 +++--
 mm/vmstat.c            |  1 +
 3 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 0c5da9141983..9d3ea9085556 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -234,7 +234,21 @@ enum node_stat_item {
 #endif
 #ifdef CONFIG_NUMA_BALANCING
 	PGPROMOTE_SUCCESS,	/* promote successfully */
-	PGPROMOTE_CANDIDATE,	/* candidate pages to promote */
+	/**
+	 * Candidate pages for promotion based on hint fault latency.  This
+	 * counter is used to control the promotion rate and adjust the hot
+	 * threshold.
+	 */
+	PGPROMOTE_CANDIDATE,
+	/**
+	 * Not rate-limited (NRL) candidate pages for those can be promoted
+	 * without considering hot threshold because of enough free pages in
+	 * fast-tier node.  These promotions bypass the regular hotness checks
+	 * and do NOT influence the promotion rate-limiter or
+	 * threshold-adjustment logic.
+	 * This is for statistics/monitoring purposes.
+	 */
+	PGPROMOTE_CANDIDATE_NRL,
 #endif
 	/* PGDEMOTE_*: pages demoted */
 	PGDEMOTE_KSWAPD,
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index b173a059315c..82c8d804c54c 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -1923,11 +1923,13 @@ bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
 		struct pglist_data *pgdat;
 		unsigned long rate_limit;
 		unsigned int latency, th, def_th;
+		long nr = folio_nr_pages(folio);
 
 		pgdat = NODE_DATA(dst_nid);
 		if (pgdat_free_space_enough(pgdat)) {
 			/* workload changed, reset hot threshold */
 			pgdat->nbp_threshold = 0;
+			mod_node_page_state(pgdat, PGPROMOTE_CANDIDATE_NRL, nr);
 			return true;
 		}
 
@@ -1941,8 +1943,7 @@ bool should_numa_migrate_memory(struct task_struct *p, struct folio *folio,
 		if (latency >= th)
 			return false;
 
-		return !numa_promotion_rate_limit(pgdat, rate_limit,
-						  folio_nr_pages(folio));
+		return !numa_promotion_rate_limit(pgdat, rate_limit, nr);
 	}
 
 	this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid);
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 71cd1ceba191..e74f0b2a1021 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1280,6 +1280,7 @@ const char * const vmstat_text[] = {
 #ifdef CONFIG_NUMA_BALANCING
 	[I(PGPROMOTE_SUCCESS)]			= "pgpromote_success",
 	[I(PGPROMOTE_CANDIDATE)]		= "pgpromote_candidate",
+	[I(PGPROMOTE_CANDIDATE_NRL)]		= "pgpromote_candidate_nrl",
 #endif
 	[I(PGDEMOTE_KSWAPD)]			= "pgdemote_kswapd",
 	[I(PGDEMOTE_DIRECT)]			= "pgdemote_direct",
-- 
2.43.0

Re: [PATCH v3] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Huang, Ying]

Ruan Shiyang <ruansy.fnst@fujitsu.com> writes:

> Goto-san reported confusing pgpromote statistics where the
> pgpromote_success count significantly exceeded pgpromote_candidate.
>
> On a system with three nodes (nodes 0-1: DRAM 4GB, node 2: NVDIMM 4GB):
>  # Enable demotion only
>  echo 1 > /sys/kernel/mm/numa/demotion_enabled
>  numactl -m 0-1 memhog -r200 3500M >/dev/null &
>  pid=$!
>  sleep 2
>  numactl memhog -r100 2500M >/dev/null &
>  sleep 10
>  kill -9 $pid # terminate the 1st memhog
>  # Enable promotion
>  echo 2 > /proc/sys/kernel/numa_balancing
>
> After a few seconds, we observeed `pgpromote_candidate < pgpromote_success`
> $ grep -e pgpromote /proc/vmstat
> pgpromote_success 2579
> pgpromote_candidate 0
>
> In this scenario, after terminating the first memhog, the conditions for
> pgdat_free_space_enough() are quickly met, and triggers promotion.
> However, these migrated pages are only counted for in PGPROMOTE_SUCCESS,
> not in PGPROMOTE_CANDIDATE.
>
> To solve these confusing statistics, introduce PGPROMOTE_CANDIDATE_NRL to
> count the missed promotion pages.  And also, not counting these pages into
> PGPROMOTE_CANDIDATE is to avoid changing the existing algorithm or
> performance of the promotion rate limit.
>
> Link: https://lkml.kernel.org/r/20250729035101.1601407-1-ruansy.fnst@fujitsu.com
> Co-developed-by: Li Zhijian <lizhijian@fujitsu.com>
> Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
> Signed-off-by: Ruan Shiyang <ruansy.fnst@fujitsu.com>
> Reported-by: Yasunori Gotou (Fujitsu) <y-goto@fujitsu.com>
> Suggested-by: Huang Ying <ying.huang@linux.alibaba.com>
> Acked-by: Vlastimil Babka <vbabka@suse.cz>
> Cc: Ingo Molnar <mingo@redhat.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Juri Lelli <juri.lelli@redhat.com>
> Cc: Vincent Guittot <vincent.guittot@linaro.org>
> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Ben Segall <bsegall@google.com>
> Cc: Mel Gorman <mgorman@suse.de>
> Cc: Valentin Schneider <vschneid@redhat.com>
> Cc: <stable@vger.kernel.org>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

LGTM, feel free to add my

Reviewed-by: Huang Ying <ying.huang@linux.alibaba.com>

in the future versions.

[snip]

---
Best Regards,
Huang, Ying

Re: [PATCH v3] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Andrew Morton]

On Mon,  1 Sep 2025 17:01:22 +0800 Ruan Shiyang <ruansy.fnst@fujitsu.com> wrote:

> Goto-san reported confusing pgpromote statistics where the
> pgpromote_success count significantly exceeded pgpromote_candidate.
> 
> On a system with three nodes (nodes 0-1: DRAM 4GB, node 2: NVDIMM 4GB):
>  # Enable demotion only
>  echo 1 > /sys/kernel/mm/numa/demotion_enabled
>  numactl -m 0-1 memhog -r200 3500M >/dev/null &
>  pid=$!
>  sleep 2
>  numactl memhog -r100 2500M >/dev/null &
>  sleep 10
>  kill -9 $pid # terminate the 1st memhog
>  # Enable promotion
>  echo 2 > /proc/sys/kernel/numa_balancing
> 
> After a few seconds, we observeed `pgpromote_candidate < pgpromote_success`
> $ grep -e pgpromote /proc/vmstat
> pgpromote_success 2579
> pgpromote_candidate 0
> 
> In this scenario, after terminating the first memhog, the conditions for
> pgdat_free_space_enough() are quickly met, and triggers promotion.
> However, these migrated pages are only counted for in PGPROMOTE_SUCCESS,
> not in PGPROMOTE_CANDIDATE.
> 
> To solve these confusing statistics, introduce PGPROMOTE_CANDIDATE_NRL to
> count the missed promotion pages.  And also, not counting these pages into
> PGPROMOTE_CANDIDATE is to avoid changing the existing algorithm or
> performance of the promotion rate limit.
> 
> ...
>

It would be good to have a Fixes: here, to tell people how far back to
backport it.

Could be either c6833e10008f or c959924b0dc5 afaict.  I'll go with
c6833e10008f, OK?

Re: [PATCH v3] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Vlastimil Babka]

On 9/1/25 21:59, Andrew Morton wrote:
> On Mon,  1 Sep 2025 17:01:22 +0800 Ruan Shiyang <ruansy.fnst@fujitsu.com> wrote:
> 
>> Goto-san reported confusing pgpromote statistics where the
>> pgpromote_success count significantly exceeded pgpromote_candidate.
>> 
>> On a system with three nodes (nodes 0-1: DRAM 4GB, node 2: NVDIMM 4GB):
>>  # Enable demotion only
>>  echo 1 > /sys/kernel/mm/numa/demotion_enabled
>>  numactl -m 0-1 memhog -r200 3500M >/dev/null &
>>  pid=$!
>>  sleep 2
>>  numactl memhog -r100 2500M >/dev/null &
>>  sleep 10
>>  kill -9 $pid # terminate the 1st memhog
>>  # Enable promotion
>>  echo 2 > /proc/sys/kernel/numa_balancing
>> 
>> After a few seconds, we observeed `pgpromote_candidate < pgpromote_success`
>> $ grep -e pgpromote /proc/vmstat
>> pgpromote_success 2579
>> pgpromote_candidate 0
>> 
>> In this scenario, after terminating the first memhog, the conditions for
>> pgdat_free_space_enough() are quickly met, and triggers promotion.
>> However, these migrated pages are only counted for in PGPROMOTE_SUCCESS,
>> not in PGPROMOTE_CANDIDATE.
>> 
>> To solve these confusing statistics, introduce PGPROMOTE_CANDIDATE_NRL to
>> count the missed promotion pages.  And also, not counting these pages into
>> PGPROMOTE_CANDIDATE is to avoid changing the existing algorithm or
>> performance of the promotion rate limit.
>> 
>> ...
>>
> 
> It would be good to have a Fixes: here, to tell people how far back to
> backport it.
> 
> Could be either c6833e10008f or c959924b0dc5 afaict.  I'll go with
> c6833e10008f, OK?

LGTM as a helpful pointer, but I don't think Cc: stable is necessary for
"admin might be confused" kind of thing if that's there since 6.1 and only
came up now.

Re: [PATCH v3] mm: memory-tiering: fix PGPROMOTE_CANDIDATE counting

[Andrew Morton]

On Mon, 1 Sep 2025 22:34:32 +0200 Vlastimil Babka <vbabka@suse.cz> wrote:

> > Could be either c6833e10008f or c959924b0dc5 afaict.  I'll go with
> > c6833e10008f, OK?
> 
> LGTM as a helpful pointer, but I don't think Cc: stable is necessary for
> "admin might be confused" kind of thing if that's there since 6.1 and only
> came up now.

OK, thanks.