[PATCH] mm: memcg: stop reclaim when a limit update is superseded

[PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Guopeng Zhang]

From: Guopeng Zhang <zhangguopeng@kylinos.cn>

kernfs serializes file operations only per open file, so separate open
files can update the same memory.high or memory.max file concurrently.
Both handlers store the new limit before synchronous reclaim, but
continue to use the writer's local target in the reclaim loop. If another
writer raises or removes the limit, the first writer can continue
reclaiming toward a stale target.

For memory.max, this can leave the writer looping indefinitely once
reclaim retries are exhausted. The OOM path sees sufficient margin under
the current limit and returns true without killing, while the writer
still compares usage against its stale target and records another OOM
event.

Check the current limit at the start of each reclaim iteration and stop
if it no longer matches the writer's target.

Fixes: 8c8c383c04f6 ("mm: memcontrol: try harder to set a new memory.high")
Fixes: b6e6edcfa405 ("mm: memcontrol: reclaim and OOM kill when shrinking memory.max below usage")
Signed-off-by: Guopeng Zhang <zhangguopeng@kylinos.cn>
---
Reproducer:

Populate a cgroup with anonymous memory and disable swapping. Lower
memory.max from one open file, then restore it to "max" through another
open file after the new limit becomes visible.

Without the patch, the first writer remains blocked and repeatedly
increments the OOM event counter. With the patch, it returns normally.

 mm/memcontrol.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 8319ad8c5c23..638bdc766616 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -4798,6 +4798,9 @@ static ssize_t memory_high_write(struct kernfs_open_file *of,
 		unsigned long nr_pages = page_counter_read(&memcg->memory);
 		unsigned long reclaimed;
 
+		if (high != READ_ONCE(memcg->memory.high))
+			break;
+
 		if (nr_pages <= high)
 			break;
 
@@ -4853,6 +4856,9 @@ static ssize_t memory_max_write(struct kernfs_open_file *of,
 	for (;;) {
 		unsigned long nr_pages = page_counter_read(&memcg->memory);
 
+		if (max != READ_ONCE(memcg->memory.max))
+			break;
+
 		if (nr_pages <= max)
 			break;
 
-- 
2.25.1

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Tao Cui]

在 2026/7/24 10:18, Guopeng Zhang 写道:
> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
> 
> kernfs serializes file operations only per open file, so separate open
> files can update the same memory.high or memory.max file concurrently.
> Both handlers store the new limit before synchronous reclaim, but
> continue to use the writer's local target in the reclaim loop. If another
> writer raises or removes the limit, the first writer can continue
> reclaiming toward a stale target.
> 
> For memory.max, this can leave the writer looping indefinitely once
> reclaim retries are exhausted. The OOM path sees sufficient margin under
> the current limit and returns true without killing, while the writer
> still compares usage against its stale target and records another OOM
> event.
> 
> Check the current limit at the start of each reclaim iteration and stop
> if it no longer matches the writer's target.
> 

Fix looks correct to me.

Acked-by: Tao Cui <cuitao@kylinos.cn>

Nit: the message lumps both paths together, but only memory.max loops
indefinitely. memory.high has no OOM path, so it just spins
MAX_RECLAIM_RETRIES times and breaks on its own. Worth a line to avoid
conflating the severity.

> Fixes: 8c8c383c04f6 ("mm: memcontrol: try harder to set a new memory.high")
> Fixes: b6e6edcfa405 ("mm: memcontrol: reclaim and OOM kill when shrinking memory.max below usage")
> Signed-off-by: Guopeng Zhang <zhangguopeng@kylinos.cn>
> ---
> Reproducer:
> 
> Populate a cgroup with anonymous memory and disable swapping. Lower
> memory.max from one open file, then restore it to "max" through another
> open file after the new limit becomes visible.
> 
> Without the patch, the first writer remains blocked and repeatedly
> increments the OOM event counter. With the patch, it returns normally.
> 
>  mm/memcontrol.c | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> index 8319ad8c5c23..638bdc766616 100644
> --- a/mm/memcontrol.c
> +++ b/mm/memcontrol.c
> @@ -4798,6 +4798,9 @@ static ssize_t memory_high_write(struct kernfs_open_file *of,
>  		unsigned long nr_pages = page_counter_read(&memcg->memory);
>  		unsigned long reclaimed;
>  
> +		if (high != READ_ONCE(memcg->memory.high))
> +			break;
> +
>  		if (nr_pages <= high)
>  			break;
>  
> @@ -4853,6 +4856,9 @@ static ssize_t memory_max_write(struct kernfs_open_file *of,
>  	for (;;) {
>  		unsigned long nr_pages = page_counter_read(&memcg->memory);
>  
> +		if (max != READ_ONCE(memcg->memory.max))
> +			break;
> +
>  		if (nr_pages <= max)
>  			break;
>  

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Guopeng Zhang]

在 2026/7/24 11:32, Tao Cui 写道:
> 
> 
> 在 2026/7/24 10:18, Guopeng Zhang 写道:
>> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
>>
>> kernfs serializes file operations only per open file, so separate open
>> files can update the same memory.high or memory.max file concurrently.
>> Both handlers store the new limit before synchronous reclaim, but
>> continue to use the writer's local target in the reclaim loop. If another
>> writer raises or removes the limit, the first writer can continue
>> reclaiming toward a stale target.
>>
>> For memory.max, this can leave the writer looping indefinitely once
>> reclaim retries are exhausted. The OOM path sees sufficient margin under
>> the current limit and returns true without killing, while the writer
>> still compares usage against its stale target and records another OOM
>> event.
>>
>> Check the current limit at the start of each reclaim iteration and stop
>> if it no longer matches the writer's target.
>>
> 
> Fix looks correct to me.
> 
> Acked-by: Tao Cui <cuitao@kylinos.cn>
> 
> Nit: the message lumps both paths together, but only memory.max loops
> indefinitely. memory.high has no OOM path, so it just spins
> MAX_RECLAIM_RETRIES times and breaks on its own. Worth a line to avoid
> conflating the severity.
> 
Hi,

Thanks for the review and Ack.

The message separates the two cases: the first paragraph describes the
stale-target reclaim behavior common to both, while the "For memory.max"
paragraph describes the OOM-based indefinite loop.

One detail is that memory.high is not limited to MAX_RECLAIM_RETRIES
iterations. The retry counter is decremented only when reclaim makes no
progress:

        if (!reclaimed && !nr_retries--)
                break;

If reclaim continues to make progress while pages are refaulted,
nr_retries is not decremented and the loop can still fail to converge.

Thanks,
Guopeng
>> Fixes: 8c8c383c04f6 ("mm: memcontrol: try harder to set a new memory.high")
>> Fixes: b6e6edcfa405 ("mm: memcontrol: reclaim and OOM kill when shrinking memory.max below usage")
>> Signed-off-by: Guopeng Zhang <zhangguopeng@kylinos.cn>
>> ---
>> Reproducer:
>>
>> Populate a cgroup with anonymous memory and disable swapping. Lower
>> memory.max from one open file, then restore it to "max" through another
>> open file after the new limit becomes visible.
>>
>> Without the patch, the first writer remains blocked and repeatedly
>> increments the OOM event counter. With the patch, it returns normally.
>>
>>  mm/memcontrol.c | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>> index 8319ad8c5c23..638bdc766616 100644
>> --- a/mm/memcontrol.c
>> +++ b/mm/memcontrol.c
>> @@ -4798,6 +4798,9 @@ static ssize_t memory_high_write(struct kernfs_open_file *of,
>>  		unsigned long nr_pages = page_counter_read(&memcg->memory);
>>  		unsigned long reclaimed;
>>  
>> +		if (high != READ_ONCE(memcg->memory.high))
>> +			break;
>> +
>>  		if (nr_pages <= high)
>>  			break;
>>  
>> @@ -4853,6 +4856,9 @@ static ssize_t memory_max_write(struct kernfs_open_file *of,
>>  	for (;;) {
>>  		unsigned long nr_pages = page_counter_read(&memcg->memory);
>>  
>> +		if (max != READ_ONCE(memcg->memory.max))
>> +			break;
>> +
>>  		if (nr_pages <= max)
>>  			break;
>>  
> 

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Andrew Morton]

On Fri, 24 Jul 2026 10:18:05 +0800 Guopeng Zhang <guopeng.zhang@linux.dev> wrote:

> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
> 
> kernfs serializes file operations only per open file, so separate open
> files can update the same memory.high or memory.max file concurrently.
> Both handlers store the new limit before synchronous reclaim, but
> continue to use the writer's local target in the reclaim loop. If another
> writer raises or removes the limit, the first writer can continue
> reclaiming toward a stale target.
> 
> For memory.max, this can leave the writer looping indefinitely once
> reclaim retries are exhausted. The OOM path sees sufficient margin under
> the current limit and returns true without killing, while the writer
> still compares usage against its stale target and records another OOM
> event.
> 
> Check the current limit at the start of each reclaim iteration and stop
> if it no longer matches the writer's target.

Thanks.

> Fixes: 8c8c383c04f6 ("mm: memcontrol: try harder to set a new memory.high")

Dec 2019

> Fixes: b6e6edcfa405 ("mm: memcontrol: reclaim and OOM kill when shrinking memory.max below usage")

Mar 2016

Having two Fixes: targets is confusing.  The Fixes: identifies which
kernel versions should be patched, so how is anyone to figure out
whether their 2018 kernel actually needs the patch?

I assume one of the two hunks is for 2019 and the other is for 2016, so
two separate patches each with a single Fixes: would be more useful.

That being said, both versions are so ancient that it probably doesn't
matter.  In fact the bug seems so obscure that I'm not planning on
adding cc:stable to this anyway.  Maintainers, please advise.

> Reproducer:
> 
> Populate a cgroup with anonymous memory and disable swapping. Lower
> memory.max from one open file, then restore it to "max" through another
> open file after the new limit becomes visible.
> 
> Without the patch, the first writer remains blocked and repeatedly
> increments the OOM event counter. With the patch, it returns normally.

This is nice info - I'll promote it into the permanent changelog.


I'll queue the patch for 7.3-rc1 and shall await maintainer input,
please.

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Guopeng Zhang]

在 2026/7/26 11:42, Andrew Morton 写道:
> On Fri, 24 Jul 2026 10:18:05 +0800 Guopeng Zhang <guopeng.zhang@linux.dev> wrote:
> 
>> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
>>
>> kernfs serializes file operations only per open file, so separate open
>> files can update the same memory.high or memory.max file concurrently.
>> Both handlers store the new limit before synchronous reclaim, but
>> continue to use the writer's local target in the reclaim loop. If another
>> writer raises or removes the limit, the first writer can continue
>> reclaiming toward a stale target.
>>
>> For memory.max, this can leave the writer looping indefinitely once
>> reclaim retries are exhausted. The OOM path sees sufficient margin under
>> the current limit and returns true without killing, while the writer
>> still compares usage against its stale target and records another OOM
>> event.
>>
>> Check the current limit at the start of each reclaim iteration and stop
>> if it no longer matches the writer's target.
> 
> Thanks.
> 
>> Fixes: 8c8c383c04f6 ("mm: memcontrol: try harder to set a new memory.high")
> 
> Dec 2019
> 
>> Fixes: b6e6edcfa405 ("mm: memcontrol: reclaim and OOM kill when shrinking memory.max below usage")
> 
> Mar 2016
> 
> Having two Fixes: targets is confusing.  The Fixes: identifies which
> kernel versions should be patched, so how is anyone to figure out
> whether their 2018 kernel actually needs the patch?
> 
> I assume one of the two hunks is for 2019 and the other is for 2016, so
> two separate patches each with a single Fixes: would be more useful.
> 

Thanks, that makes sense.

For similar cases where separate code paths were introduced by different
commits and have different affected-version ranges, I will split the
fixes into separate patches, each with its own Fixes tag.

> That being said, both versions are so ancient that it probably doesn't
> matter.  In fact the bug seems so obscure that I'm not planning on
> adding cc:stable to this anyway.  Maintainers, please advise.
> 
>> Reproducer:
>>
>> Populate a cgroup with anonymous memory and disable swapping. Lower
>> memory.max from one open file, then restore it to "max" through another
>> open file after the new limit becomes visible.
>>
>> Without the patch, the first writer remains blocked and repeatedly
>> increments the OOM event counter. With the patch, it returns normally.
> 
> This is nice info - I'll promote it into the permanent changelog.
> 
> 
> I'll queue the patch for 7.3-rc1 and shall await maintainer input,
> please.
> 

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Michal Hocko]

On Fri 24-07-26 10:18:05, Guopeng Zhang wrote:
> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
> 
> kernfs serializes file operations only per open file, so separate open
> files can update the same memory.high or memory.max file concurrently.
> Both handlers store the new limit before synchronous reclaim, but
> continue to use the writer's local target in the reclaim loop. If another
> writer raises or removes the limit, the first writer can continue
> reclaiming toward a stale target.
> 
> For memory.max, this can leave the writer looping indefinitely once
> reclaim retries are exhausted. The OOM path sees sufficient margin under
> the current limit and returns true without killing, while the writer
> still compares usage against its stale target and records another OOM
> event.

The current behavior is deliberate as described in b6e6edcfa4056.
What is an actual problem you are trying to fix?
-- 
Michal Hocko
SUSE Labs

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Guopeng Zhang]

在 2026/7/27 16:16, Michal Hocko 写道:
> On Fri 24-07-26 10:18:05, Guopeng Zhang wrote:
>> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
>>
>> kernfs serializes file operations only per open file, so separate open
>> files can update the same memory.high or memory.max file concurrently.
>> Both handlers store the new limit before synchronous reclaim, but
>> continue to use the writer's local target in the reclaim loop. If another
>> writer raises or removes the limit, the first writer can continue
>> reclaiming toward a stale target.
>>
>> For memory.max, this can leave the writer looping indefinitely once
>> reclaim retries are exhausted. The OOM path sees sufficient margin under
>> the current limit and returns true without killing, while the writer
>> still compares usage against its stale target and records another OOM
>> event.
> 
> The current behavior is deliberate as described in b6e6edcfa4056.
> What is an actual problem you are trying to fix?

Thanks for raising this. I agree that the behavior introduced by
b6e6edcfa405 is deliberate while the limit installed by the writer
remains current. The problem occurs when that limit is overwritten by a
later write through another open file.

Writer A stores a low memory.max and enters synchronous reclaim. Writer
B then restores memory.max to "max". A still compares usage against its
local old target, while mem_cgroup_out_of_memory() checks the current
memory.max. Since 1378b37d03e8, the current-margin check in
mem_cgroup_out_of_memory() sees sufficient margin and returns true
without selecting a victim. Once the reclaim retries are exhausted, A
therefore loops indefinitely and increments the oom counter in
memory.events on every iteration.

I reproduced this with a cgroup holding 128 MiB of anonymous memory and
with swapping disabled for the cgroup. Writer A lowered memory.max to
32 MiB. After that value became visible, writer B restored memory.max
to "max" through another open file.

On the unpatched kernel, A remained blocked after B's write, and the oom
counter in memory.events increased from 37333 to 13512861 during the
reproducer's one-second sampling interval. With the patch, the same
reproducer observed A return after B superseded A's limit.

The new check does not change the behavior introduced by b6e6edcfa405
while the writer's target remains the active limit. For the indefinite
loop described above, 1378b37d03e8 appears to be the more precise
Fixes: target for the memory.max hunk. Does that match your reading of
the history?

Thanks,
Guopeng

Re: [PATCH] mm: memcg: stop reclaim when a limit update is superseded

[Michal Hocko]

On Mon 27-07-26 20:59:23, Guopeng Zhang wrote:
> 
> 
> 在 2026/7/27 16:16, Michal Hocko 写道:
> > On Fri 24-07-26 10:18:05, Guopeng Zhang wrote:
> >> From: Guopeng Zhang <zhangguopeng@kylinos.cn>
> >>
> >> kernfs serializes file operations only per open file, so separate open
> >> files can update the same memory.high or memory.max file concurrently.
> >> Both handlers store the new limit before synchronous reclaim, but
> >> continue to use the writer's local target in the reclaim loop. If another
> >> writer raises or removes the limit, the first writer can continue
> >> reclaiming toward a stale target.
> >>
> >> For memory.max, this can leave the writer looping indefinitely once
> >> reclaim retries are exhausted. The OOM path sees sufficient margin under
> >> the current limit and returns true without killing, while the writer
> >> still compares usage against its stale target and records another OOM
> >> event.
> > 
> > The current behavior is deliberate as described in b6e6edcfa4056.
> > What is an actual problem you are trying to fix?
> 
> Thanks for raising this. I agree that the behavior introduced by
> b6e6edcfa405 is deliberate while the limit installed by the writer
> remains current. The problem occurs when that limit is overwritten by a
> later write through another open file.
>
> Writer A stores a low memory.max and enters synchronous reclaim. Writer
> B then restores memory.max to "max". A still compares usage against its
> local old target, while mem_cgroup_out_of_memory() checks the current
> memory.max. Since 1378b37d03e8, the current-margin check in
> mem_cgroup_out_of_memory() sees sufficient margin and returns true
> without selecting a victim. Once the reclaim retries are exhausted, A
> therefore loops indefinitely and increments the oom counter in
> memory.events on every iteration.
> 
> I reproduced this with a cgroup holding 128 MiB of anonymous memory and
> with swapping disabled for the cgroup. Writer A lowered memory.max to
> 32 MiB. After that value became visible, writer B restored memory.max
> to "max" through another open file.

Is this trying to replicate any real workload? One would expect that
writers to limit do some sort of coordination otherwise the exact
behavior is not really well defined.

> On the unpatched kernel, A remained blocked after B's write, and the oom
> counter in memory.events increased from 37333 to 13512861 during the
> reproducer's one-second sampling interval. With the patch, the same
> reproducer observed A return after B superseded A's limit.
> 
> The new check does not change the behavior introduced by b6e6edcfa405
> while the writer's target remains the active limit.  For the indefinite
> loop described above, 1378b37d03e8 appears to be the more precise
> Fixes: target for the memory.max hunk. Does that match your reading of
> the history?

Well, to be really honest I am not really convinced this needs fixing.
And if yes, your patch changes a well established behavior existing
userspace might already depend on. While your described case doesn't
look great it doesn't seem really harmful and the looping task is
killable.
-- 
Michal Hocko
SUSE Labs