[v1] mm: bail out from partial cgroup_reclaim inside shrink_lruvec

[RESEND PATCH] mm: bail out from partial cgroup_reclaim inside shrink_lruvec

Posted by zhaoyang.huang 1 month, 2 weeks ago

From: Zhaoyang Huang <zhaoyang.huang@unisoc.com>

Nowadays, ANDROID system replaces madivse with memory.reclaim to implement
user space memory management which desires to reclaim a certain amount of
memcg's memory. However, oversized reclaiming and high latency are observed
as there is no limitation over nr_reclaimed inside try_to_shrink_lruvec
when MGLRU enabled. Besides, this could also affect all none root_reclaim
such as reclaim_high etc.
The commit 'b82b530740b9' ("mm: vmscan: restore incremental cgroup
iteration") introduces sc->memcg_full_walk to limit the walk range of
mem_cgroup_iter. This commit would like to make single memcg's scanning
more precised by judging if nr_reclaimed reached when sc->memcg_full_walk
not set.

Signed-off-by: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
---
 mm/vmscan.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 670fe9fae5ba..03bda1094621 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -4832,8 +4832,8 @@ static bool should_abort_scan(struct lruvec *lruvec, struct scan_control *sc)
 	int i;
 	enum zone_watermarks mark;
 
-	/* don't abort memcg reclaim to ensure fairness */
-	if (!root_reclaim(sc))
+	/* don't abort full walk memcg reclaim to ensure fairness */
+	if (!root_reclaim(sc) && sc->memcg_full_walk)
 		return false;
 
 	if (sc->nr_reclaimed >= max(sc->nr_to_reclaim, compact_gap(sc->order)))
-- 
2.25.1

Re: [RESEND PATCH] mm: bail out from partial cgroup_reclaim inside shrink_lruvec

Posted by T.J. Mercier 1 month, 2 weeks ago

On Mon, Feb 9, 2026 at 9:44 PM zhaoyang.huang <zhaoyang.huang@unisoc.com> wrote:
>
> From: Zhaoyang Huang <zhaoyang.huang@unisoc.com>

Hi Zhaoyang,

> Nowadays, ANDROID system replaces madivse with memory.reclaim to implement
> user space memory management which desires to reclaim a certain amount of
> memcg's memory. However, oversized reclaiming and high latency are observed
> as there is no limitation over nr_reclaimed inside try_to_shrink_lruvec
> when MGLRU enabled. Besides, this could also affect all none root_reclaim
> such as reclaim_high etc.
> The commit 'b82b530740b9' ("mm: vmscan: restore incremental cgroup
> iteration") introduces sc->memcg_full_walk to limit the walk range of
> mem_cgroup_iter. This commit would like to make single memcg's scanning
> more precised by judging if nr_reclaimed reached when sc->memcg_full_walk
> not set.
>
> Signed-off-by: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
> ---
>  mm/vmscan.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/mm/vmscan.c b/mm/vmscan.c
> index 670fe9fae5ba..03bda1094621 100644
> --- a/mm/vmscan.c
> +++ b/mm/vmscan.c
> @@ -4832,8 +4832,8 @@ static bool should_abort_scan(struct lruvec *lruvec, struct scan_control *sc)
>         int i;
>         enum zone_watermarks mark;
>
> -       /* don't abort memcg reclaim to ensure fairness */
> -       if (!root_reclaim(sc))
> +       /* don't abort full walk memcg reclaim to ensure fairness */
> +       if (!root_reclaim(sc) && sc->memcg_full_walk)
>                 return false;

Can't we just get rid of this if (!root_reclaim(sc)) check entirely
now that commit 'b82b530740b9' ("mm: vmscan: restore incremental
cgroup
iteration") provides eventual fairness for the proactive reclaim case?
That wasn't true when this check was added initially.

Thanks,
T.J.

Re: [RESEND PATCH] mm: bail out from partial cgroup_reclaim inside shrink_lruvec

Posted by Zhaoyang Huang 1 month, 2 weeks ago

On Thu, Feb 12, 2026 at 6:13 AM T.J. Mercier <tjmercier@google.com> wrote:
>
> On Mon, Feb 9, 2026 at 9:44 PM zhaoyang.huang <zhaoyang.huang@unisoc.com> wrote:
> >
> > From: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
>
> Hi Zhaoyang,
>
> > Nowadays, ANDROID system replaces madivse with memory.reclaim to implement
> > user space memory management which desires to reclaim a certain amount of
> > memcg's memory. However, oversized reclaiming and high latency are observed
> > as there is no limitation over nr_reclaimed inside try_to_shrink_lruvec
> > when MGLRU enabled. Besides, this could also affect all none root_reclaim
> > such as reclaim_high etc.
> > The commit 'b82b530740b9' ("mm: vmscan: restore incremental cgroup
> > iteration") introduces sc->memcg_full_walk to limit the walk range of
> > mem_cgroup_iter. This commit would like to make single memcg's scanning
> > more precised by judging if nr_reclaimed reached when sc->memcg_full_walk
> > not set.
> >
> > Signed-off-by: Zhaoyang Huang <zhaoyang.huang@unisoc.com>
> > ---
> >  mm/vmscan.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/mm/vmscan.c b/mm/vmscan.c
> > index 670fe9fae5ba..03bda1094621 100644
> > --- a/mm/vmscan.c
> > +++ b/mm/vmscan.c
> > @@ -4832,8 +4832,8 @@ static bool should_abort_scan(struct lruvec *lruvec, struct scan_control *sc)
> >         int i;
> >         enum zone_watermarks mark;
> >
> > -       /* don't abort memcg reclaim to ensure fairness */
> > -       if (!root_reclaim(sc))
> > +       /* don't abort full walk memcg reclaim to ensure fairness */
> > +       if (!root_reclaim(sc) && sc->memcg_full_walk)
> >                 return false;
>
> Can't we just get rid of this if (!root_reclaim(sc)) check entirely
> now that commit 'b82b530740b9' ("mm: vmscan: restore incremental
> cgroup
> iteration") provides eventual fairness for the proactive reclaim case?
> That wasn't true when this check was added initially.
Thanks for the suggestion which works, I will resend the patch.
>
> Thanks,
> T.J.