Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <abec13da-c74a-0fd8-74e0-3cdc547e4e08@shopee.com>
Date:   Fri, 16 Dec 2022 09:43:02 +0800
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:102.0)
 Gecko/20100101 Thunderbird/102.6.0
Subject: Re: [PATCH] mm/memcontrol: Skip root memcg in memcg_memory_event_mm
To:     Shakeel Butt <shakeelb@google.com>
Cc:     akpm@linux-foundation.org, roman.gushchin@linux.dev,
        songmuchun@bytedance.com, hannes@cmpxchg.org, vbabka@suse.cz,
        willy@infradead.org, vasily.averin@linux.dev,
        linux-kernel@vger.kernel.org
References: <20221215091907.763801-1-haifeng.xu@shopee.com>
 <20221215181803.ome46pkh6g7qu3t4@google.com>
From:   Haifeng Xu <haifeng.xu@shopee.com>
In-Reply-To: <20221215181803.ome46pkh6g7qu3t4@google.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 2022/12/16 02:18, Shakeel Butt wrote:
> On Thu, Dec 15, 2022 at 09:19:07AM +0000, Haifeng Xu wrote:
>> The memory events aren't supported on root cgroup, so there is no need
>> to account MEMCG_OOM_KILL on root memcg.
>>
> 
> Can you explain the scenario where this is happening and causing issue
> for you?
> 
If the victim selected by oom killer belongs to root memcg, memcg_memory_event_mm
still counts the MEMCG_OOM_KILL event. This behavior is meaningless because the
flag of events/events.local in memory_files is CFTYPE_NOT_ON_ROOT. The root memcg
does not count any memory event.

>> Signed-off-by: Haifeng Xu <haifeng.xu@shopee.com>
>> ---
>>  include/linux/memcontrol.h | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
>> index 567f12323f55..09f75161a3bc 100644
>> --- a/include/linux/memcontrol.h
>> +++ b/include/linux/memcontrol.h
>> @@ -1142,7 +1142,7 @@ static inline void memcg_memory_event_mm(struct mm_struct *mm,
>>  
>>  	rcu_read_lock();
>>  	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
>> -	if (likely(memcg))
>> +	if (likely(memcg && !mem_cgroup_is_root(memcg)))
> 
> Even if we need this additional check, this is not the right place.
> 
>>  		memcg_memory_event(memcg, event);
>>  	rcu_read_unlock();
>>  }
>> -- 
>> 2.25.1
>>