]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/commitdiff
Revert "memcg: cleanup racy sum avoidance code"
authorShakeel Butt <shakeelb@google.com>
Wed, 17 Aug 2022 17:21:39 +0000 (17:21 +0000)
committerStefan Bader <stefan.bader@canonical.com>
Fri, 7 Oct 2022 08:39:34 +0000 (10:39 +0200)
BugLink: https://bugs.launchpad.net/bugs/1991717
commit dbb16df6443c59e8a1ef21c2272fcf387d600ddf upstream.

This reverts commit 96e51ccf1af33e82f429a0d6baebba29c6448d0f.

Recently we started running the kernel with rstat infrastructure on
production traffic and begin to see negative memcg stats values.
Particularly the 'sock' stat is the one which we observed having negative
value.

$ grep "sock " /mnt/memory/job/memory.stat
sock 253952
total_sock 18446744073708724224

Re-run after couple of seconds

$ grep "sock " /mnt/memory/job/memory.stat
sock 253952
total_sock 53248

For now we are only seeing this issue on large machines (256 CPUs) and
only with 'sock' stat.  I think the networking stack increase the stat on
one cpu and decrease it on another cpu much more often.  So, this negative
sock is due to rstat flusher flushing the stats on the CPU that has seen
the decrement of sock but missed the CPU that has increments.  A typical
race condition.

For easy stable backport, revert is the most simple solution.  For long
term solution, I am thinking of two directions.  First is just reduce the
race window by optimizing the rstat flusher.  Second is if the reader sees
a negative stat value, force flush and restart the stat collection.
Basically retry but limited.

Link: https://lkml.kernel.org/r/20220817172139.3141101-1-shakeelb@google.com
Fixes: 96e51ccf1af33e8 ("memcg: cleanup racy sum avoidance code")
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Cc: "Michal Koutný" <mkoutny@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Yosry Ahmed <yosryahmed@google.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: <stable@vger.kernel.org> [5.15]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Stefan Bader <stefan.bader@canonical.com>
include/linux/memcontrol.h

index d35439db047cb22bb6d3ff372bbf1c72af02bd86..4f189b17dafcc1be6545b61b57eb0916b036bbef 100644 (file)
@@ -966,19 +966,30 @@ static inline void mod_memcg_state(struct mem_cgroup *memcg,
 
 static inline unsigned long memcg_page_state(struct mem_cgroup *memcg, int idx)
 {
-       return READ_ONCE(memcg->vmstats.state[idx]);
+       long x = READ_ONCE(memcg->vmstats.state[idx]);
+#ifdef CONFIG_SMP
+       if (x < 0)
+               x = 0;
+#endif
+       return x;
 }
 
 static inline unsigned long lruvec_page_state(struct lruvec *lruvec,
                                              enum node_stat_item idx)
 {
        struct mem_cgroup_per_node *pn;
+       long x;
 
        if (mem_cgroup_disabled())
                return node_page_state(lruvec_pgdat(lruvec), idx);
 
        pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
-       return READ_ONCE(pn->lruvec_stats.state[idx]);
+       x = READ_ONCE(pn->lruvec_stats.state[idx]);
+#ifdef CONFIG_SMP
+       if (x < 0)
+               x = 0;
+#endif
+       return x;
 }
 
 static inline unsigned long lruvec_page_state_local(struct lruvec *lruvec,