linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Roman Gushchin <guroan@gmail.com>
To: linux-mm@kvack.org, kernel-team@fb.com
Cc: linux-kernel@vger.kernel.org, Tejun Heo <tj@kernel.org>,
	Rik van Riel <riel@surriel.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Michal Hocko <mhocko@kernel.org>, Roman Gushchin <guro@fb.com>
Subject: [PATCH v2 4/6] mm: release per-node memcg percpu data prematurely
Date: Tue, 12 Mar 2019 15:34:01 -0700	[thread overview]
Message-ID: <20190312223404.28665-5-guro@fb.com> (raw)
In-Reply-To: <20190312223404.28665-1-guro@fb.com>

Similar to memcg-level statistics, per-node data isn't expected
to be hot after cgroup removal. Switching over to atomics and
prematurely releasing percpu data helps to reduce the memory
footprint of dying cgroups.

Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
---
 include/linux/memcontrol.h |  1 +
 mm/memcontrol.c            | 24 +++++++++++++++++++++++-
 2 files changed, 24 insertions(+), 1 deletion(-)

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index 569337514230..f296693d102b 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -127,6 +127,7 @@ struct mem_cgroup_per_node {
 	struct lruvec		lruvec;
 
 	struct lruvec_stat __rcu /* __percpu */ *lruvec_stat_cpu;
+	struct lruvec_stat __percpu *lruvec_stat_cpu_offlined;
 	atomic_long_t		lruvec_stat[NR_VM_NODE_STAT_ITEMS];
 
 	unsigned long		lru_zone_size[MAX_NR_ZONES][NR_LRU_LISTS];
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index efd5bc131a38..1b5fe826d6d0 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -4460,7 +4460,7 @@ static void free_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
 	if (!pn)
 		return;
 
-	free_percpu(pn->lruvec_stat_cpu);
+	WARN_ON_ONCE(pn->lruvec_stat_cpu != NULL);
 	kfree(pn);
 }
 
@@ -4616,7 +4616,17 @@ static int mem_cgroup_css_online(struct cgroup_subsys_state *css)
 static void percpu_rcu_free(struct rcu_head *rcu)
 {
 	struct mem_cgroup *memcg = container_of(rcu, struct mem_cgroup, rcu);
+	int node;
+
+	for_each_node(node) {
+		struct mem_cgroup_per_node *pn = memcg->nodeinfo[node];
 
+		if (!pn)
+			continue;
+
+		free_percpu(pn->lruvec_stat_cpu_offlined);
+		WARN_ON_ONCE(pn->lruvec_stat_cpu != NULL);
+	}
 	free_percpu(memcg->vmstats_percpu_offlined);
 	WARN_ON_ONCE(memcg->vmstats_percpu);
 
@@ -4625,6 +4635,18 @@ static void percpu_rcu_free(struct rcu_head *rcu)
 
 static void mem_cgroup_offline_percpu(struct mem_cgroup *memcg)
 {
+	int node;
+
+	for_each_node(node) {
+		struct mem_cgroup_per_node *pn = memcg->nodeinfo[node];
+
+		if (!pn)
+			continue;
+
+		pn->lruvec_stat_cpu_offlined = (struct lruvec_stat __percpu *)
+			rcu_dereference(pn->lruvec_stat_cpu);
+		rcu_assign_pointer(pn->lruvec_stat_cpu, NULL);
+	}
 	memcg->vmstats_percpu_offlined = (struct memcg_vmstats_percpu __percpu*)
 		rcu_dereference(memcg->vmstats_percpu);
 	rcu_assign_pointer(memcg->vmstats_percpu, NULL);
-- 
2.20.1


  parent reply	other threads:[~2019-03-12 22:34 UTC|newest]

Thread overview: 12+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-03-12 22:33 [PATCH v2 0/6] mm: reduce the memory footprint of dying memory cgroups Roman Gushchin
2019-03-12 22:33 ` [PATCH v2 1/6] mm: prepare to premature release of memcg->vmstats_percpu Roman Gushchin
2019-03-12 22:33 ` [PATCH v2 2/6] mm: prepare to premature release of per-node lruvec_stat_cpu Roman Gushchin
2019-03-12 22:34 ` [PATCH v2 3/6] mm: release memcg percpu data prematurely Roman Gushchin
2019-03-12 22:34 ` Roman Gushchin [this message]
2019-03-12 22:34 ` [PATCH v2 5/6] mm: flush memcg percpu stats and events before releasing Roman Gushchin
2019-03-13 16:00   ` Johannes Weiner
2019-03-13 18:23     ` Roman Gushchin
2019-03-12 22:34 ` [PATCH 5/5] mm: spill " Roman Gushchin
2019-03-12 22:34 ` [PATCH v2 6/6] mm: refactor memcg_hotplug_cpu_dead() to use memcg_flush_offline_percpu() Roman Gushchin
2019-03-13 16:07   ` Johannes Weiner
2019-03-13 18:23     ` Roman Gushchin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190312223404.28665-5-guro@fb.com \
    --to=guroan@gmail.com \
    --cc=guro@fb.com \
    --cc=hannes@cmpxchg.org \
    --cc=kernel-team@fb.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mhocko@kernel.org \
    --cc=riel@surriel.com \
    --cc=tj@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).