mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Minchan Kim <minchan.kim@gmail.com>
To: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Andrea Righi <arighi@develer.com>,
	Vivek Goyal <vgoyal@redhat.com>,
	David Rientjes <rientjes@google.com>,
	Balbir Singh <balbir@linux.vnet.ibm.com>,
	Suleiman Souhlal <suleiman@google.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	containers@lists.linux-foundation.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/2] memcg: dirty pages accounting and limiting  infrastructure
Date: Fri, 26 Feb 2010 15:35:11 +0900	[thread overview]
Message-ID: <28c262361002252235s4c2a29d0yb46a6d7a7a7a1fdf@mail.gmail.com> (raw)
In-Reply-To: <20100226151506.c78b4312.kamezawa.hiroyu@jp.fujitsu.com>

On Fri, Feb 26, 2010 at 3:15 PM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@jp.fujitsu.com> wrote:
> On Fri, 26 Feb 2010 14:53:39 +0900
> Minchan Kim <minchan.kim@gmail.com> wrote:
>
>> On Fri, Feb 26, 2010 at 2:01 PM, KAMEZAWA Hiroyuki
>> <kamezawa.hiroyu@jp.fujitsu.com> wrote:
>> > Hi,
>> >
>> > On Fri, 26 Feb 2010 13:50:04 +0900
>> > Minchan Kim <minchan.kim@gmail.com> wrote:
>> >
>> >> > Hm ? I don't read the whole thread but can_attach() is called under
>> >> > cgroup_mutex(). So, it doesn't need to use RCU.
>> >>
>> >> Vivek mentioned memcg is protected by RCU if I understand his intention right.
>> >> So I commented that without enough knowledge of memcg.
>> >> After your comment, I dive into the code.
>> >>
>> >> Just out of curiosity.
>> >>
>> >> Really, memcg is protected by RCU?
>> > yes. All cgroup subsystem is protected by RCU.
>> >
>> >> I think most of RCU around memcg is for protecting task_struct and
>> >> cgroup_subsys_state.
>> >> The memcg is protected by cgroup_mutex as you mentioned.
>> >> Am I missing something?
>> >
>> > There are several levels of protections.
>> >
>> > cgroup subsystem's ->destroy() call back is finally called by
>> >
>> > As this.
>> >
>> >  768                 synchronize_rcu();
>> >  769
>> >  770                 mutex_lock(&cgroup_mutex);
>> >  771                 /*
>> >  772                  * Release the subsystem state objects.
>> >  773                  */
>> >  774                 for_each_subsys(cgrp->root, ss)
>> >  775                         ss->destroy(ss, cgrp);
>> >  776
>> >  777                 cgrp->root->number_of_cgroups--;
>> >  778                 mutex_unlock(&cgroup_mutex);
>> >
>> > Before here,
>> >        - there are no tasks under this cgroup (cgroup's refcnt is 0)
>> >          && cgroup is marked as REMOVED.
>> >
>> > Then, this access
>> >        rcu_read_lock();
>> >        mem = mem_cgroup_from_task(task);
>> >        if (css_tryget(mem->css))   <===============checks cgroup refcnt
>>
>> If it it, do we always need css_tryget after mem_cgroup_from_task
>> without cgroup_mutex to make sure css is vaild?
>>
> On a case by cases.
>
>> But I found several cases that don't call css_tryget
>>
>> 1. mm_match_cgroup
>> It's used by page_referenced_xxx. so we I think we don't grab
>> cgroup_mutex at that time.
>>
> yes. but all check are done under RCU. And this function never
> access contents of memory which pointer holds.
> And, please conider the whole context.
>
>        mem_cgroup_try_charge()
>                mem_cout =..... (refcnt +1)
>                ....
>                try_to_free_mem_cgroup_pages(mem_cont)
>                ....
>                shrink_xxx_list()
>                ....
>                        page_referenced_anon(page, mem_cont)
>                                mm_match_cgroup(mm, mem_cont)
>
> Then, this mem_cont (2nd argument to mm_match_cgroup) is always valid.
>        rcu_read_lock();
>        memcg = mem_cgroup_from_task(rcu_dereference(mm->ownder));
>        rcu_read_unlock();
>        return memcg != mem_cont;
>
> Here,
>        1. mem_cont is never reused. (because refcnt+1)
>        2. we don't access memcg's contents.
>
> I think even rcu_read_lock()/unlock() is unnecessary.
>
>
>
>> 2. mem_cgroup_oom_called
>> I think in here we don't grab cgroup_mutex, too.
>>
> In OOM-killer context, memcg which causes OOM has refcnt +1.
> Then, not necessary.
>
>
>> I guess some design would cover that problems.
> Maybe.
>
>> Could you tell me if you don't mind?
>> Sorry for bothering you.
>>
>
> In my point of view, the most terrible porblem is heavy cost of
> css_tryget() when you run multi-thread heavy program.
> So, I want to see some inovation, but haven't find yet.
>
> I admit this RCU+refcnt is tend to be hard to review. But it's costly
> operation to take refcnt and it's worth to be handled in the best
> usage of each logics, on a case by cases for now.
>

Thanks for always kind explanation, Kame.

> Thanks,
> -Kame
>
>
>



-- 
Kind regards,
Minchan Kim

  reply	other threads:[~2010-02-26  6:35 UTC|newest]

Thread overview: 56+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-02-21 15:18 [RFC] [PATCH 0/2] memcg: per cgroup dirty limit Andrea Righi
2010-02-21 15:18 ` [PATCH 1/2] memcg: dirty pages accounting and limiting infrastructure Andrea Righi
2010-02-21 21:28   ` David Rientjes
2010-02-21 22:17     ` Andrea Righi
2010-02-22 18:07       ` Vivek Goyal
2010-02-23 11:58         ` Andrea Righi
2010-02-25 15:36           ` Minchan Kim
2010-02-26  0:23             ` KAMEZAWA Hiroyuki
2010-02-26  4:50               ` Minchan Kim
2010-02-26  5:01                 ` KAMEZAWA Hiroyuki
2010-02-26  5:53                   ` Minchan Kim
2010-02-26  6:15                     ` KAMEZAWA Hiroyuki
2010-02-26  6:35                       ` Minchan Kim [this message]
2010-02-22  0:22   ` KAMEZAWA Hiroyuki
2010-02-22 18:00     ` Andrea Righi
2010-02-22 21:21       ` David Rientjes
2010-02-22 19:31     ` Vivek Goyal
2010-02-23  9:58       ` Andrea Righi
2010-02-22 15:58   ` Vivek Goyal
2010-02-22 17:29     ` Balbir Singh
2010-02-23  9:26     ` Andrea Righi
2010-02-22 16:14   ` Balbir Singh
2010-02-23  9:28     ` Andrea Righi
2010-02-24  0:09       ` KAMEZAWA Hiroyuki
2010-02-21 15:18 ` [PATCH 2/2] memcg: dirty pages instrumentation Andrea Righi
2010-02-21 21:38   ` David Rientjes
2010-02-21 22:33     ` Andrea Righi
2010-02-22  0:32   ` KAMEZAWA Hiroyuki
2010-02-22 17:57     ` Andrea Righi
2010-02-22 16:52   ` Vivek Goyal
2010-02-23  9:40     ` Andrea Righi
2010-02-23  9:45       ` Andrea Righi
2010-02-23 19:56       ` Vivek Goyal
2010-02-23 22:22         ` David Rientjes
2010-02-25 14:34           ` Andrea Righi
2010-02-26  0:14             ` KAMEZAWA Hiroyuki
2010-02-22 18:20   ` Peter Zijlstra
2010-02-23  9:46     ` Andrea Righi
2010-02-23 21:29   ` Vivek Goyal
2010-02-25 15:12     ` Andrea Righi
2010-02-26 21:48       ` Vivek Goyal
2010-02-26 22:21         ` Andrea Righi
2010-02-26 22:28           ` Vivek Goyal
2010-03-01  0:47         ` KAMEZAWA Hiroyuki
2010-02-21 23:48 ` [RFC] [PATCH 0/2] memcg: per cgroup dirty limit KAMEZAWA Hiroyuki
2010-02-22 14:27 ` Vivek Goyal
2010-02-22 17:36   ` Balbir Singh
2010-02-22 17:58     ` Vivek Goyal
2010-02-23  0:07       ` KAMEZAWA Hiroyuki
2010-02-23 15:12         ` Vivek Goyal
2010-02-24  0:19           ` KAMEZAWA Hiroyuki
2010-02-22 18:12   ` Andrea Righi
2010-02-22 18:29     ` Vivek Goyal
2010-02-22 21:15       ` David Rientjes
2010-02-23  9:55       ` Andrea Righi
2010-02-23 20:01         ` Vivek Goyal

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=28c262361002252235s4c2a29d0yb46a6d7a7a7a1fdf@mail.gmail.com \
    --to=minchan.kim@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=arighi@develer.com \
    --cc=balbir@linux.vnet.ibm.com \
    --cc=containers@lists.linux-foundation.org \
    --cc=kamezawa.hiroyu@jp.fujitsu.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=rientjes@google.com \
    --cc=suleiman@google.com \
    --cc=vgoyal@redhat.com \
    /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

all inboxes | Powered by JetHome®