linux-kernel - Re: [patch 00/10] memcg naturalization -rc4

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]

Date:	Tue, 4 Oct 2011 09:47:23 +0200
From:	Johannes Weiner <jweiner@...hat.com>
To:	Andrew Morton <akpm00@...il.com>
Cc:	KAMEZAWA Hiroyuki <kamezawa.hiroyu@...fujitsu.com>,
	Michal Hocko <mhocko@...e.cz>,
	"Kirill A. Shutemov" <kirill@...temov.name>,
	Daisuke Nishimura <nishimura@....nes.nec.co.jp>,
	Balbir Singh <bsingharora@...il.com>,
	Ying Han <yinghan@...gle.com>,
	Greg Thelen <gthelen@...gle.com>,
	Michel Lespinasse <walken@...gle.com>,
	Rik van Riel <riel@...hat.com>,
	Minchan Kim <minchan.kim@...il.com>,
	Christoph Hellwig <hch@...radead.org>,
	Hugh Dickins <hughd@...gle.com>, linux-mm@...ck.org,
	linux-kernel@...r.kernel.org
Subject: Re: [patch 00/10] memcg naturalization -rc4

On Mon, Oct 03, 2011 at 04:11:49PM -0700, Andrew Morton wrote:
> On Thu, 29 Sep 2011 23:00:54 +0200
> Johannes Weiner <jweiner@...hat.com> wrote:
> 
> > this is the fourth revision of the memory cgroup naturalization
> > series.
> 
> The patchset removes 20 lines from include/linux/*.h and removes
> exactly zero lines from mm/*.c.  Freaky.

It adds 42 lines more comments than it deletes.

The diffstat looked better when this series included the soft limit
reclaim rework, which depends on global reclaim doing hierarchy walks.
I plan to do this next, it deletes ~500 lines.

> If we were ever brave/stupid emough to make
> CONFIG_CGROUP_MEM_RES_CTLR=y unconditional, how much could we simplify
> mm/?

There will always be a remaining part that is only of interest to
people with memory cgroups, but that doesn't mean we can't shrink this
part to an adequate size.

> We are adding bits of overhead to the  CONFIG_CGROUP_MEM_RES_CTLR=n case
> all over the place.  This patchset actually decreases the size of allnoconfig
> mm/built-in.o by 1/700th.

Most of the memcg code should be completely optimized away with =n,
except for some on-stack data structures that have a struct mem_cgroup
pointer.

In the meantime, major distros started to =y per default and people
are complaining that memcg functions show up in the profiles of their
non-memcg workload.  This one worries me more.

> A "struct mem_cgroup" sometimes gets called "mem", sometimes "memcg",
> sometimes "mem_cont".  Any more candidates?  Is there any logic to
> this?

I used memcg throughout except for two patches that I fixed up.  I
don't think there is any reason to keep them different, so I'll send a
fix to rename the remaining ones to memcg.

> Anyway...  it all looks pretty sensible to me, but the timing (at
> -rc8!) is terrible.  Please keep this material maintained for -rc1, OK?

Thanks, and yeah, the timing is ambitious, I hoped that the deferred
release and merge window could make it possible.

I'll keep it uptodate.
--
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/