linux-kernel - Re: [PATCH v2] memcg: fix soft lockup in the OOM process

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <acb8ff3b-8c7a-4e31-a566-35c5427ba951@huaweicloud.com>
Date: Mon, 23 Dec 2024 10:37:49 +0800
From: Chen Ridong <chenridong@...weicloud.com>
To: Michal Hocko <mhocko@...e.com>, Andrew Morton <akpm@...ux-foundation.org>
Cc: hannes@...xchg.org, yosryahmed@...gle.com, roman.gushchin@...ux.dev,
 shakeel.butt@...ux.dev, muchun.song@...ux.dev, davidf@...eo.com,
 vbabka@...e.cz, handai.szj@...bao.com, rientjes@...gle.com,
 kamezawa.hiroyu@...fujitsu.com, linux-mm@...ck.org,
 linux-kernel@...r.kernel.org, cgroups@...r.kernel.org,
 chenridong@...wei.com, wangweiyang2@...wei.com
Subject: Re: [PATCH v2] memcg: fix soft lockup in the OOM process



On 2024/12/23 10:23, Chen Ridong wrote:
> 
> 
> On 2024/12/21 15:28, Michal Hocko wrote:
>> On Fri 20-12-24 14:47:34, Andrew Morton wrote:
>>> On Fri, 20 Dec 2024 10:31:23 +0000 Chen Ridong <chenridong@...weicloud.com> wrote:
>>>
>>>> From: Chen Ridong <chenridong@...wei.com>
>>>>
>>>> A soft lockup issue was found in the product with about 56,000 tasks were
>>>> in the OOM cgroup, it was traversing them when the soft lockup was
>>>> triggered.
>>>>
>>>> ...
>>>>
>>>> This is because thousands of processes are in the OOM cgroup, it takes a
>>>> long time to traverse all of them. As a result, this lead to soft lockup
>>>> in the OOM process.
>>>>
>>>> To fix this issue, call 'cond_resched' in the 'mem_cgroup_scan_tasks'
>>>> function per 1000 iterations. For global OOM, call
>>>> 'touch_softlockup_watchdog' per 1000 iterations to avoid this issue.
>>>>
>>>> ...
>>>>
>>>> --- a/include/linux/oom.h
>>>> +++ b/include/linux/oom.h
>>>> @@ -14,6 +14,13 @@ struct notifier_block;
>>>>  struct mem_cgroup;
>>>>  struct task_struct;
>>>>  
>>>> +/* When it traverses for long time,  to prevent softlockup, call
>>>> + * cond_resched/touch_softlockup_watchdog very 1000 iterations.
>>>> + * The 1000 value  is not exactly right, it's used to mitigate the overhead
>>>> + * of cond_resched/touch_softlockup_watchdog.
>>>> + */
>>>> +#define SOFTLOCKUP_PREVENTION_LIMIT 1000
>>>
>>> If this is to have potentially kernel-wide scope, its name should
>>> identify which subsystem it belongs to.  Maybe OOM_KILL_RESCHED or
>>> something.
>>>
>>> But I'm not sure that this really needs to exist.  Are the two usage
>>> sites particularly related?
>>
>> Yes, I do not think this needs to pretend to be a more generic mechanism
>> to prevent soft lockups. The number of iterations highly depends on the
>> operation itself.
>>
> 
> Thanks， I will update.
> 
>>>
>>>>  enum oom_constraint {
>>>>  	CONSTRAINT_NONE,
>>>>  	CONSTRAINT_CPUSET,
>>>> diff --git a/mm/memcontrol.c b/mm/memcontrol.c
>>>> index 5c373d275e7a..f4c12d6e7b37 100644
>>>> --- a/mm/memcontrol.c
>>>> +++ b/mm/memcontrol.c
>>>> @@ -1161,6 +1161,7 @@ void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
>>>>  {
>>>>  	struct mem_cgroup *iter;
>>>>  	int ret = 0;
>>>> +	int i = 0;
>>>>  
>>>>  	BUG_ON(mem_cgroup_is_root(memcg));
>>>>  
>>>> @@ -1169,8 +1170,11 @@ void mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
>>>>  		struct task_struct *task;
>>>>  
>>>>  		css_task_iter_start(&iter->css, CSS_TASK_ITER_PROCS, &it);
>>>> -		while (!ret && (task = css_task_iter_next(&it)))
>>>> +		while (!ret && (task = css_task_iter_next(&it))) {
>>>>  			ret = fn(task, arg);
>>>> +			if (++i % SOFTLOCKUP_PREVENTION_LIMIT)
>>>
>>> And a modulus operation is somewhat expensive.
>>
>> This is a cold path used during OOM. While we can make it more optimal I
>> doubt it matters in practice so we should aim at readbility. I do not
>> mind either way, I just wanted to note that this is not performance
>> sensitive.
>>
> 
> I think '(++i & 1023)' is much better, I will update.
> Thank you all gays.
> 
So sorry for wrong spelling
Sorry sorry.

Thank you all guys.

Best regards
Ridong

> Best regards
> Ridong
> 
>>>
>>> Perhaps a simple
>>>
>>> 		/* Avoid potential softlockup warning */
>>> 		if ((++i & 1023) == 0)
>>>
>>> at both sites will suffice.  Opinions might vary...
>>>
>>
>