On Wed, Jan 6, 2021 at 3:35 AM Kirill Tkhai [off-list ref] wrote:

On 06.01.2021 01:58, Yang Shi wrote:

quoted

Now shrinker's nr_deferred is per memcg for memcg aware shrinkers, add to parent's
corresponding nr_deferred when memcg offline.

Signed-off-by: Yang Shi <redacted>
---
 include/linux/memcontrol.h |  1 +
 mm/memcontrol.c            |  1 +
 mm/vmscan.c                | 29 +++++++++++++++++++++++++++++
 3 files changed, 31 insertions(+)

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index 5599082df623..d1e52e916cc2 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h

@@ -1586,6 +1586,7 @@ extern int memcg_alloc_shrinker_info(struct mem_cgroup *memcg);
 extern void memcg_free_shrinker_info(struct mem_cgroup *memcg);
 extern void memcg_set_shrinker_bit(struct mem_cgroup *memcg,
                                 int nid, int shrinker_id);
+extern void memcg_reparent_shrinker_deferred(struct mem_cgroup *memcg);
 #else
 #define mem_cgroup_sockets_enabled 0
 static inline void mem_cgroup_sk_alloc(struct sock *sk) { };

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 126f1fd550c8..19e555675582 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c

@@ -5284,6 +5284,7 @@ static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
      page_counter_set_low(&memcg->memory, 0);

      memcg_offline_kmem(memcg);
+     memcg_reparent_shrinker_deferred(memcg);
      wb_memcg_offline(memcg);

      drain_all_stock(memcg);

diff --git a/mm/vmscan.c b/mm/vmscan.c
index d9795fb0f1c5..71056057d26d 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c

@@ -396,6 +396,35 @@ static long set_nr_deferred_memcg(long nr, int nid, struct shrinker *shrinker,
      return atomic_long_add_return(nr, &info->nr_deferred[shrinker->id]);
 }

+void memcg_reparent_shrinker_deferred(struct mem_cgroup *memcg)
+{
+     int i, nid;
+     long nr;
+     struct mem_cgroup *parent;
+     struct memcg_shrinker_info *child_info, *parent_info;
+
+     parent = parent_mem_cgroup(memcg);
+     if (!parent)
+             parent = root_mem_cgroup;
+
+     /* Prevent from concurrent shrinker_info expand */
+     down_read(&shrinker_rwsem);
+     for_each_node(nid) {
+             child_info = rcu_dereference_protected(
+                                     memcg->nodeinfo[nid]->shrinker_info,
+                                     true);
+             parent_info = rcu_dereference_protected(
+                                     parent->nodeinfo[nid]->shrinker_info,
+                                     true);

Simple assignment can't take such lots of space, we have to do something with that.

Number of these

        rcu_dereference_protected(memcg->nodeinfo[nid]->shrinker_info, true)

became too big, and we can't allow every of them takes 3 lines.

We should introduce a short helper to dereferrence this, so we will be able to give
out attention to really difficult logic instead of wasting it on parsing this.

                child_info = memcg_shrinker_info(memcg, nid);
or
                child_info = memcg_shrinker_info_protected(memcg, nid);

Both of them fit in single line.

struct memcg_shrinker_info *memcg_shrinker_info_protected(
                                        struct mem_cgroup *memcg, int nid)
{
        return rcu_dereference_protected(memcg->nodeinfo[nid]->shrinker_info,
                                         lockdep_assert_held(&shrinker_rwsem));
}

Thanks for the suggestion, it makes sense to me. Will incorporate it in v4.

quoted

+             for (i = 0; i < shrinker_nr_max; i++) {
+                     nr = atomic_long_read(&child_info->nr_deferred[i]);
+                     atomic_long_add(nr,
+                                     &parent_info->nr_deferred[i]);

Why new line is here? In case of you merge it up, it will be even shorter then previous line.

Just keep in 80 lines. We could relax it.

quoted

+             }
+     }
+     up_read(&shrinker_rwsem);
+}
+
 static bool cgroup_reclaim(struct scan_control *sc)
 {
      return sc->target_mem_cgroup;