All of lore.kernel.org
 help / color / mirror / Atom feed
From: Michal Hocko <mhocko@suse.cz>
To: Tejun Heo <tj@kernel.org>
Cc: Hugh Dickins <hughd@google.com>,
	Filipe Brandenburger <filbranden@google.com>,
	Li Zefan <lizefan@huawei.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Greg Thelen <gthelen@google.com>,
	Michel Lespinasse <walken@google.com>,
	Markus Blank-Burian <burian@muenster.de>,
	Shawn Bohrer <shawn.bohrer@gmail.com>,
	cgroups@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-mm@kvack.org
Subject: Re: [PATCH] cgroup: use an ordered workqueue for cgroup destruction
Date: Fri, 7 Feb 2014 16:28:49 +0100	[thread overview]
Message-ID: <20140207152849.GF5121@dhcp22.suse.cz> (raw)
In-Reply-To: <20140207151341.GB3304@htj.dyndns.org>

On Fri 07-02-14 10:13:41, Tejun Heo wrote:
> Hello, Michal.
> 
> On Fri, Feb 07, 2014 at 03:37:40PM +0100, Michal Hocko wrote:
> > Hmm, this is a bit tricky. We cannot use memcg iterators to reach
> > children because css_tryget would fail on them. We can use cgroup
> > iterators instead, alright, and reparent pages from leafs but this all
> > sounds like a lot of complications.
> 
> Hmmm... I think we're talking past each other here.  Why would the
> parent need to reach down to the children?  Just bail out if it can't
> make things down to zero and let the child when it finishes its own
> cleaning walk up the tree propagating changes.  ->parent is always
> accessible.  Would that be complicated too?

This would be basically the option #2 bellow.

> > Another option would be weakening css_offline reparenting and do not
> > insist on having 0 charges. We want to get rid of as many charges as
> > possible but do not need to have all of them gone
> > (http://marc.info/?l=linux-kernel&m=139161412932193&w=2). The last part
> > would be reparenting to the upmost parent which is still online.
> > 
> > I guess this is implementable but I would prefer Hugh's fix for now and
> > for stable.
> 
> Yeah, for -stable, I think Hugh's patch is good but I really don't
> want to keep it long term.

Based on our recent discussion regarding css_offline semantic we want to
do some changes in that area. I thought we would simply update comments
but considering this report css_offline needs some changes as well. I
will look at it. The idea is to split mem_cgroup_reparent_charges into
two parts. The core one which drains LRUs and would be called from
mem_cgroup_css_offline and one which loops until all charges are gone
for mem_cgroup_css_free. mem_cgroup_move_parent will need an update as
well. It would have to go up the hierarchy to the first alive parent.

-- 
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>

WARNING: multiple messages have this Message-ID (diff)
From: Michal Hocko <mhocko@suse.cz>
To: Tejun Heo <tj@kernel.org>
Cc: Hugh Dickins <hughd@google.com>,
	Filipe Brandenburger <filbranden@google.com>,
	Li Zefan <lizefan@huawei.com>,
	Andrew Morton <akpm@linux-foundation.org>,
	Johannes Weiner <hannes@cmpxchg.org>,
	Greg Thelen <gthelen@google.com>,
	Michel Lespinasse <walken@google.com>,
	Markus Blank-Burian <burian@muenster.de>,
	Shawn Bohrer <shawn.bohrer@gmail.com>,
	cgroups@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-mm@kvack.org
Subject: Re: [PATCH] cgroup: use an ordered workqueue for cgroup destruction
Date: Fri, 7 Feb 2014 16:28:49 +0100	[thread overview]
Message-ID: <20140207152849.GF5121@dhcp22.suse.cz> (raw)
In-Reply-To: <20140207151341.GB3304@htj.dyndns.org>

On Fri 07-02-14 10:13:41, Tejun Heo wrote:
> Hello, Michal.
> 
> On Fri, Feb 07, 2014 at 03:37:40PM +0100, Michal Hocko wrote:
> > Hmm, this is a bit tricky. We cannot use memcg iterators to reach
> > children because css_tryget would fail on them. We can use cgroup
> > iterators instead, alright, and reparent pages from leafs but this all
> > sounds like a lot of complications.
> 
> Hmmm... I think we're talking past each other here.  Why would the
> parent need to reach down to the children?  Just bail out if it can't
> make things down to zero and let the child when it finishes its own
> cleaning walk up the tree propagating changes.  ->parent is always
> accessible.  Would that be complicated too?

This would be basically the option #2 bellow.

> > Another option would be weakening css_offline reparenting and do not
> > insist on having 0 charges. We want to get rid of as many charges as
> > possible but do not need to have all of them gone
> > (http://marc.info/?l=linux-kernel&m=139161412932193&w=2). The last part
> > would be reparenting to the upmost parent which is still online.
> > 
> > I guess this is implementable but I would prefer Hugh's fix for now and
> > for stable.
> 
> Yeah, for -stable, I think Hugh's patch is good but I really don't
> want to keep it long term.

Based on our recent discussion regarding css_offline semantic we want to
do some changes in that area. I thought we would simply update comments
but considering this report css_offline needs some changes as well. I
will look at it. The idea is to split mem_cgroup_reparent_charges into
two parts. The core one which drains LRUs and would be called from
mem_cgroup_css_offline and one which loops until all charges are gone
for mem_cgroup_css_free. mem_cgroup_move_parent will need an update as
well. It would have to go up the hierarchy to the first alive parent.

-- 
Michal Hocko
SUSE Labs

  reply	other threads:[~2014-02-07 15:28 UTC|newest]

Thread overview: 43+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-02-06 23:56 [PATCH] cgroup: use an ordered workqueue for cgroup destruction Hugh Dickins
2014-02-06 23:56 ` Hugh Dickins
2014-02-06 23:56 ` Hugh Dickins
2014-02-07 14:04 ` Tejun Heo
2014-02-07 14:04   ` Tejun Heo
2014-02-07 14:37   ` Michal Hocko
2014-02-07 14:37     ` Michal Hocko
     [not found]     ` <20140207143740.GD5121-2MMpYkNvuYDjFM9bn6wA6Q@public.gmane.org>
2014-02-07 15:13       ` Tejun Heo
2014-02-07 15:13         ` Tejun Heo
2014-02-07 15:13         ` Tejun Heo
2014-02-07 15:28         ` Michal Hocko [this message]
2014-02-07 15:28           ` Michal Hocko
     [not found]   ` <20140207140402.GA3304-Gd/HAXX7CRxy/B6EtB590w@public.gmane.org>
2014-02-07 20:20     ` Hugh Dickins
2014-02-07 20:20       ` Hugh Dickins
2014-02-07 20:20       ` Hugh Dickins
2014-02-07 20:35       ` Tejun Heo
2014-02-07 20:35         ` Tejun Heo
2014-02-07 21:06         ` Hugh Dickins
2014-02-07 21:06           ` Hugh Dickins
2014-02-07 15:21 ` Tejun Heo
2014-02-07 15:21   ` Tejun Heo
     [not found] ` <alpine.LSU.2.11.1402061541560.31342-fupSdm12i1nKWymIFiNcPA@public.gmane.org>
2014-02-07 13:45   ` Michal Hocko
2014-02-07 13:45     ` Michal Hocko
2014-02-07 13:45     ` Michal Hocko
2014-02-07 16:43   ` Johannes Weiner
2014-02-07 16:43     ` Johannes Weiner
2014-02-07 16:43     ` Johannes Weiner
     [not found]     ` <20140207164321.GE6963-druUgvl0LCNAfugRpC6u6w@public.gmane.org>
2014-02-10 15:46       ` Michal Hocko
2014-02-10 15:46         ` Michal Hocko
2014-02-10 15:46         ` Michal Hocko
2014-02-12 22:59       ` Hugh Dickins
2014-02-12 22:59         ` Hugh Dickins
2014-02-12 22:59         ` Hugh Dickins
     [not found]         ` <alpine.LSU.2.11.1402121417230.5029-fupSdm12i1nKWymIFiNcPA@public.gmane.org>
2014-02-12 23:06           ` [PATCH 2/2] cgroup: bring back kill_cnt to order css destruction Hugh Dickins
2014-02-12 23:06             ` Hugh Dickins
2014-02-12 23:06             ` Hugh Dickins
2014-02-13  0:28             ` Tejun Heo
2014-02-13  0:28               ` Tejun Heo
2014-02-13  0:38               ` Hugh Dickins
2014-02-13  0:38                 ` Hugh Dickins
2014-02-13  0:09         ` [PATCH] Revert "cgroup: use an ordered workqueue for cgroup destruction" Tejun Heo
2014-02-13  0:09           ` Tejun Heo
2014-02-13  0:09           ` Tejun Heo

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=20140207152849.GF5121@dhcp22.suse.cz \
    --to=mhocko@suse.cz \
    --cc=akpm@linux-foundation.org \
    --cc=burian@muenster.de \
    --cc=cgroups@vger.kernel.org \
    --cc=filbranden@google.com \
    --cc=gthelen@google.com \
    --cc=hannes@cmpxchg.org \
    --cc=hughd@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=lizefan@huawei.com \
    --cc=shawn.bohrer@gmail.com \
    --cc=tj@kernel.org \
    --cc=walken@google.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.