Received-SPF: pass (google.com: domain of linux-kernel+bounces-145714-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Precedence: bulk
MIME-Version: 1.0
References: <20240415111123924s9IbQkgHF8S4yZv4su8LI@zte.com.cn> <CAGsJ_4y5+3jKx78-TWh_3jWoXQW4mmx=uKHbBSHWpvMj1eYW7A@mail.gmail.com>
In-Reply-To: <CAGsJ_4y5+3jKx78-TWh_3jWoXQW4mmx=uKHbBSHWpvMj1eYW7A@mail.gmail.com>
From: Yang Shi <shy828301@gmail.com>
Date: Mon, 15 Apr 2024 11:31:55 -0700
Message-ID: <CAHbLzkqwHiSb4e=soerK1qtA7FoJz8PEPXiCmnFx8i1Bhvp2-g@mail.gmail.com>
Subject: Re: [PATCH] mm: thp: makes the memcg THP deferred split shrinker
 aware of node_id
To: Barry Song <21cnbao@gmail.com>
Cc: xu.xin16@zte.com.cn, yang.shi@linux.alibaba.com, v-songbaohua@oppo.com, 
	mhocko@kernel.org, ryan.roberts@arm.com, david@redhat.com, 
	roman.gushchin@linux.dev, shakeel.butt@linux.dev, muchun.song@linux.dev, 
	akpm@linux-foundation.org, cgroups@vger.kernel.org, linux-mm@kvack.org, 
	linux-kernel@vger.kernel.org, ran.xiaokai@zte.com.cn, yang.yang29@zte.com.cn, 
	lu.zhongjun@zte.com.cn
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Sun, Apr 14, 2024 at 8:30=E2=80=AFPM Barry Song <21cnbao@gmail.com> wrot=
e:
>
> On Mon, Apr 15, 2024 at 3:11=E2=80=AFPM <xu.xin16@zte.com.cn> wrote:
> >
> > From: Ran Xiaokai <ran.xiaokai@zte.com.cn>
> >
> > Since commit 87eaceb3faa5 ("mm: thp: make deferred split shrinker
> > memcg aware"), the THP deferred split queue is per memcg but not
> > per mem_cgroup_per_node. This has two aspects of impact:
> >
> > Impact1: for kswapd reclaim
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> >   kswapd
> >     balance_pgdat
> >       kswapd_shrink_node
> >         shrink_node(pgdat, sc);
> >           shrink_node_memcgs(pgdat, sc);
> >             shrink_slab(sc->gfp_mask, pgdat->node_id, memcg...);
> >   the parameter "pgdat->node_id"  does not take effectct for
> >   THP deferred_split_shrinker, as the deferred_split_queue of
> >   specified memcg is not for a certain numa node but for all the nodes.
> >   We want to makes the memcg THP deferred split shrinker aware of
> >   node_id.
> >
> > Impact2: thp-deferred_split shrinker debugfs interface
> > =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> >    for the "count" file:
> >    <cgroup inode id> <objects on node 0> <objects on node 1>
> >      the output is acctually the sum of all numa nodes.
> >    for the "scan" file:
> >    <cgroup inode id> <numa id> <number of objects to scan>
> >      Also the "numa id" input does not take effect here.
> >
> > This patch makes memcg deferred_split_queue per mem_cgroup_per_node
> > so try to conform to semantic logic.

I used to have a similar patch before,
https://lore.kernel.org/linux-mm/1569968203-64647-1-git-send-email-yang.shi=
@linux.alibaba.com/

But it was somehow lost in discussion.

I have no objection to this patch. However, I was thinking about using
list_lru for deferred split queue, but I didn't have time to look
deeper. Maybe we should try now?

>
> This seems to be a correct fix to me,  + Yang Shi, the original author of
> commit 87eaceb3faa5.
>
> >
> > Reviewed-by: Lu Zhongjun <lu.zhongjun@zte.com.cn>
> > Signed-off-by: Ran Xiaokai <ran.xiaokai@zte.com.cn>
> > Cc: xu xin <xu.xin16@zte.com.cn>
> > Cc: Yang Yang <yang.yang29@zte.com.cn>
> > ---
> >  include/linux/memcontrol.h |  7 +++----
> >  mm/huge_memory.c           |  6 +++---
> >  mm/memcontrol.c            | 11 +++++------
> >  3 files changed, 11 insertions(+), 13 deletions(-)
> >
> > diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
> > index 394fd0a887ae..7282861d5a5d 100644
> > --- a/include/linux/memcontrol.h
> > +++ b/include/linux/memcontrol.h
> > @@ -130,6 +130,9 @@ struct mem_cgroup_per_node {
> >         bool                    on_tree;
> >         struct mem_cgroup       *memcg;         /* Back pointer, we can=
not */
> >                                                 /* use container_of    =
    */
> > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > +       struct deferred_split deferred_split_queue;
> > +#endif
> >  };
> >
> >  struct mem_cgroup_threshold {
> > @@ -327,10 +330,6 @@ struct mem_cgroup {
> >         struct list_head event_list;
> >         spinlock_t event_list_lock;
> >
> > -#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > -       struct deferred_split deferred_split_queue;
> > -#endif
> > -
> >  #ifdef CONFIG_LRU_GEN_WALKS_MMU
> >         /* per-memcg mm_struct list */
> >         struct lru_gen_mm_list mm_list;
> > diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> > index 9859aa4f7553..338d071070a6 100644
> > --- a/mm/huge_memory.c
> > +++ b/mm/huge_memory.c
> > @@ -774,7 +774,7 @@ struct deferred_split *get_deferred_split_queue(str=
uct folio *folio)
> >         struct pglist_data *pgdat =3D NODE_DATA(folio_nid(folio));
> >
> >         if (memcg)
> > -               return &memcg->deferred_split_queue;
> > +               return &memcg->nodeinfo[pgdat->node_id]->deferred_split=
_queue;
> >         else
> >                 return &pgdat->deferred_split_queue;
> >  }
> > @@ -3305,7 +3305,7 @@ static unsigned long deferred_split_count(struct =
shrinker *shrink,
> >
> >  #ifdef CONFIG_MEMCG
> >         if (sc->memcg)
> > -               ds_queue =3D &sc->memcg->deferred_split_queue;
> > +               ds_queue =3D &sc->memcg->nodeinfo[sc->nid]->deferred_sp=
lit_queue;
> >  #endif
> >         return READ_ONCE(ds_queue->split_queue_len);
> >  }
> > @@ -3322,7 +3322,7 @@ static unsigned long deferred_split_scan(struct s=
hrinker *shrink,
> >
> >  #ifdef CONFIG_MEMCG
> >         if (sc->memcg)
> > -               ds_queue =3D &sc->memcg->deferred_split_queue;
> > +               ds_queue =3D &sc->memcg->nodeinfo[sc->nid]->deferred_sp=
lit_queue;
> >  #endif
> >
> >         spin_lock_irqsave(&ds_queue->split_queue_lock, flags);
> > diff --git a/mm/memcontrol.c b/mm/memcontrol.c
> > index fabce2b50c69..cdf9f5fa3b8e 100644
> > --- a/mm/memcontrol.c
> > +++ b/mm/memcontrol.c
> > @@ -5445,7 +5445,11 @@ static int alloc_mem_cgroup_per_node_info(struct=
 mem_cgroup *memcg, int node)
> >                 kfree(pn);
> >                 return 1;
> >         }
> > -
> > +#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > +       spin_lock_init(&pn->deferred_split_queue.split_queue_lock);
> > +       INIT_LIST_HEAD(&pn->deferred_split_queue.split_queue);
> > +       pn->deferred_split_queue.split_queue_len =3D 0;
> > +#endif
> >         lruvec_init(&pn->lruvec);
> >         pn->memcg =3D memcg;
> >
> > @@ -5545,11 +5549,6 @@ static struct mem_cgroup *mem_cgroup_alloc(struc=
t mem_cgroup *parent)
> >         for (i =3D 0; i < MEMCG_CGWB_FRN_CNT; i++)
> >                 memcg->cgwb_frn[i].done =3D
> >                         __WB_COMPLETION_INIT(&memcg_cgwb_frn_waitq);
> > -#endif
> > -#ifdef CONFIG_TRANSPARENT_HUGEPAGE
> > -       spin_lock_init(&memcg->deferred_split_queue.split_queue_lock);
> > -       INIT_LIST_HEAD(&memcg->deferred_split_queue.split_queue);
> > -       memcg->deferred_split_queue.split_queue_len =3D 0;
> >  #endif
> >         lru_gen_init_memcg(memcg);
> >         return memcg;
> > --
> > 2.15.2
> >
>
> Thanks
> Barry
>