linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Michal Hocko <mhocko@suse.cz>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: Glauber Costa <glommer@parallels.com>,
	linux-mm@kvack.org, cgroups@vger.kernel.org,
	linux-kernel@vger.kernel.org, Tejun Heo <tj@kernel.org>,
	Li Zefan <lizefan@huawei.com>,
	Johannes Weiner <hannes@cmpxchg.org>,
	KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>,
	Balbir Singh <bsingharora@gmail.com>
Subject: Re: [PATCH v3 3/6] memcg: Simplify mem_cgroup_force_empty_list error handling
Date: Tue, 30 Oct 2012 11:35:59 +0100	[thread overview]
Message-ID: <20121030103559.GA7394@dhcp22.suse.cz> (raw)
In-Reply-To: <20121029150022.a595b866.akpm@linux-foundation.org>

On Mon 29-10-12 15:00:22, Andrew Morton wrote:
> On Mon, 29 Oct 2012 17:58:45 +0400
> Glauber Costa <glommer@parallels.com> wrote:
> 
> > > + * move charges to its parent or the root cgroup if the group has no
> > > + * parent (aka use_hierarchy==0).
> > > + * Although this might fail (get_page_unless_zero, isolate_lru_page or
> > > + * mem_cgroup_move_account fails) the failure is always temporary and
> > > + * it signals a race with a page removal/uncharge or migration. In the
> > > + * first case the page is on the way out and it will vanish from the LRU
> > > + * on the next attempt and the call should be retried later.
> > > + * Isolation from the LRU fails only if page has been isolated from
> > > + * the LRU since we looked at it and that usually means either global
> > > + * reclaim or migration going on. The page will either get back to the
> > > + * LRU or vanish.
> > 
> > I just wonder for how long can it go in the worst case?
> 
> If the kernel is uniprocessor and the caller is SCHED_FIFO: ad infinitum!

You are right, if the rmdir (resp. echo > force_empty) at SCHED_FIFO
races with put_page (on a shared page) which gets preempted after
put_page_testzero and before __page_cache_release then we are screwed:

						put_page(page)
						  put_page_testzero
						  <preempted and page still on LRU>
mem_cgroup_force_empty_list
  page = list_entry(list->prev, struct page, lru);
  mem_cgroup_move_parent(page)
    get_page_unless_zero <fails>
  cond_resched() <scheduled again>

The race window is really small but it is definitely possible. I am not
happy about this state and it should be probably mentioned in the
patch description but I do not see any way around (except for hacks like
sched_setscheduler for the current which is, ehm...) and still keep
do_not_fail contract here.

Can we consider this as a corner case (it is much easier to kill a
machine with SCHED_FIFO than this anyway) or the concern is really
strong and we should come with a solution before this can get merged?

-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2012-10-30 10:36 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-10-26 11:37 memcg/cgroup: do not fail fail on pre_destroy callbacks Michal Hocko
2012-10-26 11:37 ` [PATCH v3 1/6] memcg: split mem_cgroup_force_empty into reclaiming and reparenting parts Michal Hocko
2012-10-29 13:45   ` Glauber Costa
2012-10-31 16:29   ` Johannes Weiner
2012-10-26 11:37 ` [PATCH v3 2/6] memcg: root_cgroup cannot reach mem_cgroup_move_parent Michal Hocko
2012-10-29 13:48   ` Glauber Costa
2012-10-29 13:52     ` Michal Hocko
2012-10-31 16:31   ` Johannes Weiner
2012-10-26 11:37 ` [PATCH v3 3/6] memcg: Simplify mem_cgroup_force_empty_list error handling Michal Hocko
2012-10-29 13:58   ` Glauber Costa
2012-10-29 14:15     ` Michal Hocko
2012-10-29 15:09       ` Glauber Costa
2012-10-29 22:00     ` Andrew Morton
2012-10-30 10:35       ` Michal Hocko [this message]
2012-10-31 21:30         ` Andrew Morton
2012-11-13 21:10         ` Johannes Weiner
2012-11-14 13:59           ` Michal Hocko
2012-11-14 18:33             ` Johannes Weiner
2012-10-26 11:37 ` [PATCH v3 4/6] cgroups: forbid pre_destroy callback to fail Michal Hocko
2012-10-29 14:04   ` Glauber Costa
2012-10-29 14:06     ` Glauber Costa
2012-10-29 14:17       ` Michal Hocko
2012-11-13 21:13   ` Johannes Weiner
2012-10-26 11:37 ` [PATCH v3 5/6] memcg: make mem_cgroup_reparent_charges non failing Michal Hocko
2012-10-29 14:07   ` Glauber Costa
2012-10-26 11:37 ` [PATCH v3 6/6] hugetlb: do not fail in hugetlb_cgroup_pre_destroy Michal Hocko
2012-10-29 14:08   ` Glauber Costa
2012-10-29 23:26 ` memcg/cgroup: do not fail fail on pre_destroy callbacks Tejun Heo
2012-10-30 23:37   ` Michal Hocko

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=20121030103559.GA7394@dhcp22.suse.cz \
    --to=mhocko@suse.cz \
    --cc=akpm@linux-foundation.org \
    --cc=bsingharora@gmail.com \
    --cc=cgroups@vger.kernel.org \
    --cc=glommer@parallels.com \
    --cc=hannes@cmpxchg.org \
    --cc=kamezawa.hiroyu@jp.fujitsu.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=lizefan@huawei.com \
    --cc=tj@kernel.org \
    /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;
as well as URLs for NNTP newsgroup(s).