linux-kernel - Re: [RFC 2/5] memcg: rework mem_cgroup

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: Windows password security audit tool. GUI, reports in PDF.

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

Date:	Thu, 15 Nov 2012 16:12:55 +0100
From:	Michal Hocko <mhocko@...e.cz>
To:	Tejun Heo <htejun@...il.com>
Cc:	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>,
	Johannes Weiner <hannes@...xchg.org>,
	Ying Han <yinghan@...gle.com>,
	Glauber Costa <glommer@...allels.com>
Subject: Re: [RFC 2/5] memcg: rework mem_cgroup_iter to use cgroup iterators

On Thu 15-11-12 06:47:32, Tejun Heo wrote:
> Hello, Michal.
> 
> On Thu, Nov 15, 2012 at 10:51:03AM +0100, Michal Hocko wrote:
> > > I'm a bit confused.  Why would that make any difference?  Shouldn't it
> > > be just able to test the condition and continue?
> > 
> > Ohh, I misunderstood your proposal. So what you are suggesting is
> > to put all the logic we have in mem_cgroup_iter inside what you call
> > reclaim here + mem_cgroup_iter_break inside the loop, right?
> > 
> > I do not see how this would help us much. mem_cgroup_iter is not the
> > nicest piece of code but it handles quite a complex requirements that we
> > have currently (css reference count, multiple reclaimers racing). So I
> > would rather keep it this way. Further simplifications are welcome of
> > course.
> > 
> > Is there any reason why you are not happy about direct using of
> > cgroup_next_descendant_pre?
> 
> Because I'd like to consider the next functions as implementation
> detail, and having interations structred as loops tend to read better
> and less error-prone.  e.g. when you use next functions directly, it's
> way easier to circumvent locking requirements in a way which isn't
> very obvious. 

The whole point behind mem_cgroup_iter is to hide all the complexity
behind memcg iteration. Memcg code either use for_each_mem_cgroup_tree
for !reclaim case and mem_cgroup_iter otherwise.

> So, unless it messes up the code too much (and I can't see why it
> would), I'd much prefer if memcg used for_each_*() macros.

As I said this would mean that the current mem_cgroup_iter code would
have to be inverted which doesn't simplify the code much. I'd rather
hide all the grossy details inside the memcg iterator.
Or am I still missing your suggestion?
-- 
Michal Hocko
SUSE Labs
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/