Burt Holzman has noticed that memcg v1 doesn't notify about OOM events via
eventfd anymore.  The reason is that 
29ef680ae7c2 ("memcg, oom: move
out_of_memory back to the charge path") has moved the oom handling back to
the charge path.  While doing so the notification was left behind in
mem_cgroup_oom_synchronize.
Fix the issue by replicating the oom hierarchy locking and the
notification.
Link: http://lkml.kernel.org/r/20181224091107.18354-1-mhocko@kernel.org
Fixes: 29ef680ae7c2 ("memcg, oom: move out_of_memory back to the charge path")
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Burt Holzman <burt@fnal.gov>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com
Cc: <stable@vger.kernel.org>	[4.19+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
 
 
 static enum oom_status mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
 {
+       enum oom_status ret;
+       bool locked;
+
        if (order > PAGE_ALLOC_COSTLY_ORDER)
                return OOM_SKIPPED;
 
                return OOM_ASYNC;
        }
 
+       mem_cgroup_mark_under_oom(memcg);
+
+       locked = mem_cgroup_oom_trylock(memcg);
+
+       if (locked)
+               mem_cgroup_oom_notify(memcg);
+
+       mem_cgroup_unmark_under_oom(memcg);
        if (mem_cgroup_out_of_memory(memcg, mask, order))
-               return OOM_SUCCESS;
+               ret = OOM_SUCCESS;
+       else
+               ret = OOM_FAILED;
 
-       return OOM_FAILED;
+       if (locked)
+               mem_cgroup_oom_unlock(memcg);
+
+       return ret;
 }
 
 /**