* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 3:17 ` [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter Ridong
@ 2026-07-16 3:25 ` Qi Zheng
2026-07-16 3:58 ` Qi Zheng
2026-07-16 6:06 ` Tao Cui
2 siblings, 0 replies; 9+ messages in thread
From: Qi Zheng @ 2026-07-16 3:25 UTC (permalink / raw)
To: Ridong, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Lorenzo Stoakes, Axel Rasmussen,
Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups, linux-mm,
linux-kernel, Ridong Chen, stable
On 7/16/26 11:17 AM, Ridong wrote:
> From: Ridong Chen <chenridong@xiaomi.com>
>
> sc_swappiness() had two separate definitions depending on
> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
> vm_swappiness, ignoring the proactive_swappiness value passed
> through scan_control. This caused the swappiness parameter
> written to /sys/devices/system/node/nodeX/reclaim to have no
> effect when CONFIG_MEMCG is disabled.
>
> Fix this by consolidating sc_swappiness() into a single definition
> that checks sc->proactive_swappiness first, then falls back to
> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
> and !CONFIG_MEMCG.
>
> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>
> # cat /proc/sys/vm/swappiness
> 60
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 1840
> pgsteal_anon 25
> pgsteal_file 1815
> # echo "64M swappiness=max" > /sys/devices/system/node/node0/reclaim
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 18013
> pgsteal_anon 337
> pgsteal_file 17676
>
> After fix (swappiness=max honored, anon pages reclaimed as expected):
>
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 0
> pgsteal_anon 0
> pgsteal_file 0
> # echo "64M swappiness=max" > /sys/devices/system/node/node0/reclaim
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 16283
> pgsteal_anon 16283
> pgsteal_file 0
>
> Cc: stable@vger.kernel.org # 6.17+
> Fixes: b980077899ea ("mm: introduce per-node proactive reclaim interface")
> Signed-off-by: Ridong Chen <chenridong@xiaomi.com>
> Acked-by: Johannes Weiner <hannes@cmpxchg.org>
> Reviewed-by: Barry Song <baohua@kernel.org>
> ---
> mm/vmscan.c | 19 +++++++------------
> 1 file changed, 7 insertions(+), 12 deletions(-)
Thanks for fixing this!
Acked-by: Qi Zheng <qi.zheng@linux.dev>
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 3:17 ` [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter Ridong
2026-07-16 3:25 ` Qi Zheng
@ 2026-07-16 3:58 ` Qi Zheng
2026-07-16 8:21 ` Ridong Chen
2026-07-16 6:06 ` Tao Cui
2 siblings, 1 reply; 9+ messages in thread
From: Qi Zheng @ 2026-07-16 3:58 UTC (permalink / raw)
To: Ridong, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Lorenzo Stoakes, Axel Rasmussen,
Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups, linux-mm,
linux-kernel, Ridong Chen, stable
Hi Ridong,
On 7/16/26 11:17 AM, Ridong wrote:
> From: Ridong Chen <chenridong@xiaomi.com>
>
> sc_swappiness() had two separate definitions depending on
> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
> vm_swappiness, ignoring the proactive_swappiness value passed
> through scan_control. This caused the swappiness parameter
> written to /sys/devices/system/node/nodeX/reclaim to have no
> effect when CONFIG_MEMCG is disabled.
>
> Fix this by consolidating sc_swappiness() into a single definition
> that checks sc->proactive_swappiness first, then falls back to
> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
> and !CONFIG_MEMCG.
>
> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>
> # cat /proc/sys/vm/swappiness
> 60
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 1840
> pgsteal_anon 25
> pgsteal_file 1815
> # echo "64M swappiness=max" > /sys/devices/system/node/node0/reclaim
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 18013
> pgsteal_anon 337
> pgsteal_file 17676
>
> After fix (swappiness=max honored, anon pages reclaimed as expected):
>
By the way, in get_scan_count(), it seems we still reclaim file pages
without swap space, even if swappiness=max (SWAPPINESS_ANON_ONLY) is
set.
This behavior appears to contradict the semantics of
SWAPPINESS_ANON_ONLY, which probably needs a fix as well.
Thanks,
Qi
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 3:58 ` Qi Zheng
@ 2026-07-16 8:21 ` Ridong Chen
2026-07-16 8:42 ` Qi Zheng
0 siblings, 1 reply; 9+ messages in thread
From: Ridong Chen @ 2026-07-16 8:21 UTC (permalink / raw)
To: Qi Zheng, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Lorenzo Stoakes, Axel Rasmussen,
Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups, linux-mm,
linux-kernel, Ridong Chen, stable
On 7/16/2026 11:58 AM, Qi Zheng wrote:
> Hi Ridong,
>
> On 7/16/26 11:17 AM, Ridong wrote:
>> From: Ridong Chen <chenridong@xiaomi.com>
>>
>> sc_swappiness() had two separate definitions depending on
>> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
>> vm_swappiness, ignoring the proactive_swappiness value passed
>> through scan_control. This caused the swappiness parameter
>> written to /sys/devices/system/node/nodeX/reclaim to have no
>> effect when CONFIG_MEMCG is disabled.
>>
>> Fix this by consolidating sc_swappiness() into a single definition
>> that checks sc->proactive_swappiness first, then falls back to
>> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
>> and !CONFIG_MEMCG.
>>
>> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>>
>> # cat /proc/sys/vm/swappiness
>> 60
>> # cat /proc/vmstat | grep pgsteal
>> pgsteal_kswapd 0
>> pgsteal_direct 0
>> pgsteal_khugepaged 0
>> pgsteal_proactive 1840
>> pgsteal_anon 25
>> pgsteal_file 1815
>> # echo "64M swappiness=max" > /sys/devices/system/node/node0/reclaim
>> # cat /proc/vmstat | grep pgsteal
>> pgsteal_kswapd 0
>> pgsteal_direct 0
>> pgsteal_khugepaged 0
>> pgsteal_proactive 18013
>> pgsteal_anon 337
>> pgsteal_file 17676
>>
>> After fix (swappiness=max honored, anon pages reclaimed as expected):
>>
>
> By the way, in get_scan_count(), it seems we still reclaim file pages
> without swap space, even if swappiness=max (SWAPPINESS_ANON_ONLY) is
> set.
>
> This behavior appears to contradict the semantics of
> SWAPPINESS_ANON_ONLY, which probably needs a fix as well.
>
Thanks, Qi.
You're right. Currently, when can_reclaim_anon_pages() returns false,
the reclaim logic falls back to scanning file folios even if swappiness
== SWAPPINESS_ANON_ONLY is set:
```
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
unsigned long *nr)
{
....
/* If we have no swap space, do not bother scanning anon folios. */
if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat->node_id, sc)) {
scan_balance = SCAN_FILE;
goto out;
}
...
/* Proactive reclaim initiated by userspace for anonymous memory only */
if (swappiness == SWAPPINESS_ANON_ONLY) {
WARN_ON_ONCE(!sc->proactive);
scan_balance = SCAN_ANON;
goto out;
}
...
}
```
The !can_reclaim_anon_pages() check above takes precedence over the
SWAPPINESS_ANON_ONLY branch, making the latter effectively unreachable
when anonymous pages are deemed non-reclaimable.
To fix this, could we move the SWAPPINESS_ANON_ONLY check up before the
can_reclaim_anon_pages() check?
```
static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
unsigned long *nr)
{
...
/* Proactive reclaim initiated by userspace for anonymous memory only */
if (swappiness == SWAPPINESS_ANON_ONLY) {
WARN_ON_ONCE(!sc->proactive);
scan_balance = SCAN_ANON;
goto out;
}
....
/* If we have no swap space, do not bother scanning anon folios. */
if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat->node_id, sc)) {
scan_balance = SCAN_FILE;
goto out;
}
...
}
```
--
Best regards
Ridong
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 8:21 ` Ridong Chen
@ 2026-07-16 8:42 ` Qi Zheng
2026-07-16 11:17 ` Ridong Chen
0 siblings, 1 reply; 9+ messages in thread
From: Qi Zheng @ 2026-07-16 8:42 UTC (permalink / raw)
To: Ridong Chen, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Lorenzo Stoakes, Axel Rasmussen,
Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups, linux-mm,
linux-kernel, Ridong Chen, stable
On 7/16/26 4:21 PM, Ridong Chen wrote:
>
>
> On 7/16/2026 11:58 AM, Qi Zheng wrote:
>> Hi Ridong,
>>
>> On 7/16/26 11:17 AM, Ridong wrote:
>>> From: Ridong Chen <chenridong@xiaomi.com>
>>>
>>> sc_swappiness() had two separate definitions depending on
>>> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
>>> vm_swappiness, ignoring the proactive_swappiness value passed
>>> through scan_control. This caused the swappiness parameter
>>> written to /sys/devices/system/node/nodeX/reclaim to have no
>>> effect when CONFIG_MEMCG is disabled.
>>>
>>> Fix this by consolidating sc_swappiness() into a single definition
>>> that checks sc->proactive_swappiness first, then falls back to
>>> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
>>> and !CONFIG_MEMCG.
>>>
>>> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>>>
>>> # cat /proc/sys/vm/swappiness
>>> 60
>>> # cat /proc/vmstat | grep pgsteal
>>> pgsteal_kswapd 0
>>> pgsteal_direct 0
>>> pgsteal_khugepaged 0
>>> pgsteal_proactive 1840
>>> pgsteal_anon 25
>>> pgsteal_file 1815
>>> # echo "64M swappiness=max" > /sys/devices/system/node/node0/
>>> reclaim
>>> # cat /proc/vmstat | grep pgsteal
>>> pgsteal_kswapd 0
>>> pgsteal_direct 0
>>> pgsteal_khugepaged 0
>>> pgsteal_proactive 18013
>>> pgsteal_anon 337
>>> pgsteal_file 17676
>>>
>>> After fix (swappiness=max honored, anon pages reclaimed as expected):
>>>
>>
>> By the way, in get_scan_count(), it seems we still reclaim file pages
>> without swap space, even if swappiness=max (SWAPPINESS_ANON_ONLY) is
>> set.
>>
>> This behavior appears to contradict the semantics of
>> SWAPPINESS_ANON_ONLY, which probably needs a fix as well.
>>
> Thanks, Qi.
>
> You're right. Currently, when can_reclaim_anon_pages() returns false,
> the reclaim logic falls back to scanning file folios even if swappiness
> == SWAPPINESS_ANON_ONLY is set:
>
> ```
> static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
> unsigned long *nr)
> {
> ....
> /* If we have no swap space, do not bother scanning anon folios. */
> if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat->node_id,
> sc)) {
> scan_balance = SCAN_FILE;
> goto out;
> }
>
> ...
> /* Proactive reclaim initiated by userspace for anonymous memory
> only */
> if (swappiness == SWAPPINESS_ANON_ONLY) {
> WARN_ON_ONCE(!sc->proactive);
> scan_balance = SCAN_ANON;
> goto out;
> }
> ...
> }
> ```
>
> The !can_reclaim_anon_pages() check above takes precedence over the
> SWAPPINESS_ANON_ONLY branch, making the latter effectively unreachable
> when anonymous pages are deemed non-reclaimable.
>
> To fix this, could we move the SWAPPINESS_ANON_ONLY check up before the
> can_reclaim_anon_pages() check?
>
> ```
> static void get_scan_count(struct lruvec *lruvec, struct scan_control *sc,
> unsigned long *nr)
> {
> ...
> /* Proactive reclaim initiated by userspace for anonymous memory
> only */
> if (swappiness == SWAPPINESS_ANON_ONLY) {
> WARN_ON_ONCE(!sc->proactive);
> scan_balance = SCAN_ANON;
> goto out;
For SWAPPINESS_ANON_ONLY + !can_reclaim_anon_pages, perhaps we should
just return directly instead of goto out, since no pages can be
reclaimed in this situation.
Thanks,
Qi
> }
> ....
> /* If we have no swap space, do not bother scanning anon folios. */
> if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat->node_id,
> sc)) {
> scan_balance = SCAN_FILE;
> goto out;
> }
>
> ...
> }
> ```
>
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 8:42 ` Qi Zheng
@ 2026-07-16 11:17 ` Ridong Chen
2026-07-17 2:27 ` Qi Zheng
0 siblings, 1 reply; 9+ messages in thread
From: Ridong Chen @ 2026-07-16 11:17 UTC (permalink / raw)
To: Qi Zheng, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Lorenzo Stoakes, Axel Rasmussen,
Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups, linux-mm,
linux-kernel, Ridong Chen, stable
On 7/16/2026 4:42 PM, Qi Zheng wrote:
>
>
> On 7/16/26 4:21 PM, Ridong Chen wrote:
>>
>>
>> On 7/16/2026 11:58 AM, Qi Zheng wrote:
>>> Hi Ridong,
>>>
>>> On 7/16/26 11:17 AM, Ridong wrote:
>>>> From: Ridong Chen <chenridong@xiaomi.com>
>>>>
>>>> sc_swappiness() had two separate definitions depending on
>>>> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
>>>> vm_swappiness, ignoring the proactive_swappiness value passed
>>>> through scan_control. This caused the swappiness parameter
>>>> written to /sys/devices/system/node/nodeX/reclaim to have no
>>>> effect when CONFIG_MEMCG is disabled.
>>>>
>>>> Fix this by consolidating sc_swappiness() into a single definition
>>>> that checks sc->proactive_swappiness first, then falls back to
>>>> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
>>>> and !CONFIG_MEMCG.
>>>>
>>>> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>>>>
>>>> # cat /proc/sys/vm/swappiness
>>>> 60
>>>> # cat /proc/vmstat | grep pgsteal
>>>> pgsteal_kswapd 0
>>>> pgsteal_direct 0
>>>> pgsteal_khugepaged 0
>>>> pgsteal_proactive 1840
>>>> pgsteal_anon 25
>>>> pgsteal_file 1815
>>>> # echo "64M swappiness=max" > /sys/devices/system/node/node0/
>>>> reclaim
>>>> # cat /proc/vmstat | grep pgsteal
>>>> pgsteal_kswapd 0
>>>> pgsteal_direct 0
>>>> pgsteal_khugepaged 0
>>>> pgsteal_proactive 18013
>>>> pgsteal_anon 337
>>>> pgsteal_file 17676
>>>>
>>>> After fix (swappiness=max honored, anon pages reclaimed as expected):
>>>>
>>>
>>> By the way, in get_scan_count(), it seems we still reclaim file pages
>>> without swap space, even if swappiness=max (SWAPPINESS_ANON_ONLY) is
>>> set.
>>>
>>> This behavior appears to contradict the semantics of
>>> SWAPPINESS_ANON_ONLY, which probably needs a fix as well.
>>>
>> Thanks, Qi.
>>
>> You're right. Currently, when can_reclaim_anon_pages() returns false,
>> the reclaim logic falls back to scanning file folios even if
>> swappiness == SWAPPINESS_ANON_ONLY is set:
>>
>> ```
>> static void get_scan_count(struct lruvec *lruvec, struct scan_control
>> *sc,
>> unsigned long *nr)
>> {
>> ....
>> /* If we have no swap space, do not bother scanning anon folios. */
>> if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat-
>> >node_id, sc)) {
>> scan_balance = SCAN_FILE;
>> goto out;
>> }
>>
>> ...
>> /* Proactive reclaim initiated by userspace for anonymous memory
>> only */
>> if (swappiness == SWAPPINESS_ANON_ONLY) {
>> WARN_ON_ONCE(!sc->proactive);
>> scan_balance = SCAN_ANON;
>> goto out;
>> }
>> ...
>> }
>> ```
>>
>> The !can_reclaim_anon_pages() check above takes precedence over the
>> SWAPPINESS_ANON_ONLY branch, making the latter effectively unreachable
>> when anonymous pages are deemed non-reclaimable.
>>
>> To fix this, could we move the SWAPPINESS_ANON_ONLY check up before
>> the can_reclaim_anon_pages() check?
>>
>> ```
>> static void get_scan_count(struct lruvec *lruvec, struct scan_control
>> *sc,
>> unsigned long *nr)
>> {
>> ...
>> /* Proactive reclaim initiated by userspace for anonymous memory
>> only */
>> if (swappiness == SWAPPINESS_ANON_ONLY) {
>> WARN_ON_ONCE(!sc->proactive);
>> scan_balance = SCAN_ANON;
>> goto out;
>
> For SWAPPINESS_ANON_ONLY + !can_reclaim_anon_pages, perhaps we should
> just return directly instead of goto out, since no pages can be
> reclaimed in this situation.
>
Much better.
Would you mind if I send a separate patch to fix this issue?
>
>> }
>> ....
>> /* If we have no swap space, do not bother scanning anon folios. */
>> if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat-
>> >node_id, sc)) {
>> scan_balance = SCAN_FILE;
>> goto out;
>> }
>>
>> ...
>> }
>> ```
>>
>
--
Best regards
Ridong
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 11:17 ` Ridong Chen
@ 2026-07-17 2:27 ` Qi Zheng
0 siblings, 0 replies; 9+ messages in thread
From: Qi Zheng @ 2026-07-17 2:27 UTC (permalink / raw)
To: Ridong Chen, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Lorenzo Stoakes, Axel Rasmussen,
Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups, linux-mm,
linux-kernel, Ridong Chen, stable
On 7/16/26 7:17 PM, Ridong Chen wrote:
>
>
> On 7/16/2026 4:42 PM, Qi Zheng wrote:
>>
>>
>> On 7/16/26 4:21 PM, Ridong Chen wrote:
>>>
>>>
>>> On 7/16/2026 11:58 AM, Qi Zheng wrote:
>>>> Hi Ridong,
>>>>
>>>> On 7/16/26 11:17 AM, Ridong wrote:
>>>>> From: Ridong Chen <chenridong@xiaomi.com>
>>>>>
>>>>> sc_swappiness() had two separate definitions depending on
>>>>> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
>>>>> vm_swappiness, ignoring the proactive_swappiness value passed
>>>>> through scan_control. This caused the swappiness parameter
>>>>> written to /sys/devices/system/node/nodeX/reclaim to have no
>>>>> effect when CONFIG_MEMCG is disabled.
>>>>>
>>>>> Fix this by consolidating sc_swappiness() into a single definition
>>>>> that checks sc->proactive_swappiness first, then falls back to
>>>>> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
>>>>> and !CONFIG_MEMCG.
>>>>>
>>>>> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>>>>>
>>>>> # cat /proc/sys/vm/swappiness
>>>>> 60
>>>>> # cat /proc/vmstat | grep pgsteal
>>>>> pgsteal_kswapd 0
>>>>> pgsteal_direct 0
>>>>> pgsteal_khugepaged 0
>>>>> pgsteal_proactive 1840
>>>>> pgsteal_anon 25
>>>>> pgsteal_file 1815
>>>>> # echo "64M swappiness=max" > /sys/devices/system/node/node0/
>>>>> reclaim
>>>>> # cat /proc/vmstat | grep pgsteal
>>>>> pgsteal_kswapd 0
>>>>> pgsteal_direct 0
>>>>> pgsteal_khugepaged 0
>>>>> pgsteal_proactive 18013
>>>>> pgsteal_anon 337
>>>>> pgsteal_file 17676
>>>>>
>>>>> After fix (swappiness=max honored, anon pages reclaimed as expected):
>>>>>
>>>>
>>>> By the way, in get_scan_count(), it seems we still reclaim file pages
>>>> without swap space, even if swappiness=max (SWAPPINESS_ANON_ONLY) is
>>>> set.
>>>>
>>>> This behavior appears to contradict the semantics of
>>>> SWAPPINESS_ANON_ONLY, which probably needs a fix as well.
>>>>
>>> Thanks, Qi.
>>>
>>> You're right. Currently, when can_reclaim_anon_pages() returns false,
>>> the reclaim logic falls back to scanning file folios even if
>>> swappiness == SWAPPINESS_ANON_ONLY is set:
>>>
>>> ```
>>> static void get_scan_count(struct lruvec *lruvec, struct scan_control
>>> *sc,
>>> unsigned long *nr)
>>> {
>>> ....
>>> /* If we have no swap space, do not bother scanning anon folios. */
>>> if (!sc->may_swap || !can_reclaim_anon_pages(memcg, pgdat-
>>> >node_id, sc)) {
>>> scan_balance = SCAN_FILE;
>>> goto out;
>>> }
>>>
>>> ...
>>> /* Proactive reclaim initiated by userspace for anonymous memory
>>> only */
>>> if (swappiness == SWAPPINESS_ANON_ONLY) {
>>> WARN_ON_ONCE(!sc->proactive);
>>> scan_balance = SCAN_ANON;
>>> goto out;
>>> }
>>> ...
>>> }
>>> ```
>>>
>>> The !can_reclaim_anon_pages() check above takes precedence over the
>>> SWAPPINESS_ANON_ONLY branch, making the latter effectively
>>> unreachable when anonymous pages are deemed non-reclaimable.
>>>
>>> To fix this, could we move the SWAPPINESS_ANON_ONLY check up before
>>> the can_reclaim_anon_pages() check?
>>>
>>> ```
>>> static void get_scan_count(struct lruvec *lruvec, struct scan_control
>>> *sc,
>>> unsigned long *nr)
>>> {
>>> ...
>>> /* Proactive reclaim initiated by userspace for anonymous memory
>>> only */
>>> if (swappiness == SWAPPINESS_ANON_ONLY) {
>>> WARN_ON_ONCE(!sc->proactive);
>>> scan_balance = SCAN_ANON;
>>> goto out;
>>
>> For SWAPPINESS_ANON_ONLY + !can_reclaim_anon_pages, perhaps we should
>> just return directly instead of goto out, since no pages can be
>> reclaimed in this situation.
>>
>
> Much better.
> Would you mind if I send a separate patch to fix this issue?
>
Sure, thanks for your time on this, and I look forward to your fix!
^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 3:17 ` [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter Ridong
2026-07-16 3:25 ` Qi Zheng
2026-07-16 3:58 ` Qi Zheng
@ 2026-07-16 6:06 ` Tao Cui
2026-07-16 8:23 ` Ridong Chen
2 siblings, 1 reply; 9+ messages in thread
From: Tao Cui @ 2026-07-16 6:06 UTC (permalink / raw)
To: Ridong, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: cui.tao, Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He,
Barry Song, Youngjun Park, David Hildenbrand, Qi Zheng,
Lorenzo Stoakes, Axel Rasmussen, Yuanchu Xie, Wei Xu,
Davidlohr Bueso, cgroups, linux-mm, linux-kernel, Ridong Chen,
stable
Hi Ridong,
Reproduced the bug and verified the fix in a QEMU x86_64 guest with
CONFIG_MEMCG disabled. Global vm.swappiness pinned to 1, then:
echo "64M swappiness=<x>" > /sys/devices/system/node/node0/reclaim
swappiness= | buggy (anon/file) | patched (anon/file)
----------- | -------------------- | --------------------
max | 0 / 16385 | 16292 / 0
200 | 2 / 16384 | 16837 / 0
1 | 0 / 16386 | 0 / 16467
Unpatched ignores the swappiness arg (always file); patched honors it.
Matches the commit message.
Tested-by: Song Hu <husong@kylinos.cn>
Reviewed-by: Song Hu <husong@kylinos.cn>
在 2026/7/16 11:17, Ridong 写道:
> From: Ridong Chen <chenridong@xiaomi.com>
>
> sc_swappiness() had two separate definitions depending on
> CONFIG_MEMCG. The !CONFIG_MEMCG variant simply returned
> vm_swappiness, ignoring the proactive_swappiness value passed
> through scan_control. This caused the swappiness parameter
> written to /sys/devices/system/node/nodeX/reclaim to have no
> effect when CONFIG_MEMCG is disabled.
>
> Fix this by consolidating sc_swappiness() into a single definition
> that checks sc->proactive_swappiness first, then falls back to
> mem_cgroup_swappiness() which already handles both CONFIG_MEMCG
> and !CONFIG_MEMCG.
>
> Before fix (swappiness=max ignored, mostly file pages reclaimed):
>
> # cat /proc/sys/vm/swappiness
> 60
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 1840
> pgsteal_anon 25
> pgsteal_file 1815
> # echo "64M swappiness=max" > /sys/devices/system/node/node0/reclaim
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 18013
> pgsteal_anon 337
> pgsteal_file 17676
>
> After fix (swappiness=max honored, anon pages reclaimed as expected):
>
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 0
> pgsteal_anon 0
> pgsteal_file 0
> # echo "64M swappiness=max" > /sys/devices/system/node/node0/reclaim
> # cat /proc/vmstat | grep pgsteal
> pgsteal_kswapd 0
> pgsteal_direct 0
> pgsteal_khugepaged 0
> pgsteal_proactive 16283
> pgsteal_anon 16283
> pgsteal_file 0
>
> Cc: stable@vger.kernel.org # 6.17+
> Fixes: b980077899ea ("mm: introduce per-node proactive reclaim interface")
> Signed-off-by: Ridong Chen <chenridong@xiaomi.com>
> Acked-by: Johannes Weiner <hannes@cmpxchg.org>
> Reviewed-by: Barry Song <baohua@kernel.org>
> ---
> mm/vmscan.c | 19 +++++++------------
> 1 file changed, 7 insertions(+), 12 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 986dde8e7429..bd71595c8b2b 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -199,6 +199,13 @@ struct scan_control {
> */
> int vm_swappiness = 60;
>
> +static int sc_swappiness(struct scan_control *sc, struct mem_cgroup *memcg)
> +{
> + if (sc->proactive && sc->proactive_swappiness)
> + return *sc->proactive_swappiness;
> + return mem_cgroup_swappiness(memcg);
> +}
> +
> #ifdef CONFIG_MEMCG
>
> /* Returns true for reclaim through cgroup limits or cgroup interfaces. */
> @@ -239,13 +246,6 @@ static bool writeback_throttling_sane(struct scan_control *sc)
> #endif
> return false;
> }
> -
> -static int sc_swappiness(struct scan_control *sc, struct mem_cgroup *memcg)
> -{
> - if (sc->proactive && sc->proactive_swappiness)
> - return *sc->proactive_swappiness;
> - return mem_cgroup_swappiness(memcg);
> -}
> #else
> static bool cgroup_reclaim(struct scan_control *sc)
> {
> @@ -261,11 +261,6 @@ static bool writeback_throttling_sane(struct scan_control *sc)
> {
> return true;
> }
> -
> -static int sc_swappiness(struct scan_control *sc, struct mem_cgroup *memcg)
> -{
> - return READ_ONCE(vm_swappiness);
> -}
> #endif
>
> static void set_task_reclaim_state(struct task_struct *task,
^ permalink raw reply [flat|nested] 9+ messages in thread* Re: [PATCH v2 2/2] mm: vmscan: fix node reclaim ignoring swappiness parameter
2026-07-16 6:06 ` Tao Cui
@ 2026-07-16 8:23 ` Ridong Chen
0 siblings, 0 replies; 9+ messages in thread
From: Ridong Chen @ 2026-07-16 8:23 UTC (permalink / raw)
To: Tao Cui, Johannes Weiner, Michal Hocko, Roman Gushchin,
Shakeel Butt, Andrew Morton, Chris Li, Kairui Song
Cc: Muchun Song, Kemeng Shi, Nhat Pham, Baoquan He, Barry Song,
Youngjun Park, David Hildenbrand, Qi Zheng, Lorenzo Stoakes,
Axel Rasmussen, Yuanchu Xie, Wei Xu, Davidlohr Bueso, cgroups,
linux-mm, linux-kernel, Ridong Chen, stable
On 7/16/2026 2:06 PM, Tao Cui wrote:
> Hi Ridong,
>
> Reproduced the bug and verified the fix in a QEMU x86_64 guest with
> CONFIG_MEMCG disabled. Global vm.swappiness pinned to 1, then:
>
> echo "64M swappiness=<x>" > /sys/devices/system/node/node0/reclaim
>
> swappiness= | buggy (anon/file) | patched (anon/file)
> ----------- | -------------------- | --------------------
> max | 0 / 16385 | 16292 / 0
> 200 | 2 / 16384 | 16837 / 0
> 1 | 0 / 16386 | 0 / 16467
>
> Unpatched ignores the swappiness arg (always file); patched honors it.
> Matches the commit message.
>
> Tested-by: Song Hu <husong@kylinos.cn>
> Reviewed-by: Song Hu <husong@kylinos.cn>
>
Thanks.
--
Best regards
Ridong
^ permalink raw reply [flat|nested] 9+ messages in thread