linux-kernel - Re: [PATCH 1/2] mm,vmscan: Kill global shrinker lock.

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:   Mon, 20 Nov 2017 10:28:38 -0800
From:   "Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
To:     Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>
Cc:     hch@...radead.org, mhocko@...nel.org, minchan@...nel.org,
        ying.huang@...el.com, mgorman@...hsingularity.net,
        vdavydov.dev@...il.com, hannes@...xchg.org,
        akpm@...ux-foundation.org, shakeelb@...gle.com, gthelen@...gle.com,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/2] mm,vmscan: Kill global shrinker lock.

On Mon, Nov 20, 2017 at 07:56:28PM +0900, Tetsuo Handa wrote:
> Christoph Hellwig wrote:
> > On Mon, Nov 20, 2017 at 10:42:37AM +0100, Michal Hocko wrote:
> > > The patch has been dropped because allnoconfig failed to compile back
> > > then http://lkml.kernel.org/r/CAP=VYLr0rPWi1aeuk4w1On9CYRNmnEWwJgGtaX=wEvGaBURtrg@mail.gmail.com
> > > I have problem to find the follow up discussion though. The main
> > > argument was that SRC is not generally available and so the core
> > > kernel should rely on it.
> > 
> > Paul,
> > 
> > isthere any good reason to not use SRCU in the core kernel and
> > instead try to reimplement it using atomic counters?
> 
> CONFIG_SRCU was added in order to save system size. There are users who run Linux on very
> small systems ( https://www.elinux.org/images/5/52/Status-of-embedded-Linux-2017-09-JJ62.pdf ).
> 
> Also, atomic counters are not mandatory for shrinker case; e.g.
> http://lkml.kernel.org/r/201711161956.EBF57883.QFFMOLOVSOHJFt@I-love.SAKURA.ne.jp .

CONFIG_SRCU was indeed added in order to shrink single-CPU systems.
But many architectures are now requiring SRCU for one reason or another,
in more and more situations.

So I recently implemented a UP-only Tiny SRCU, which is quite a bit
smaller than its scalable counterpart, Tree SRCU:

   text	   data	    bss	    dec	    hex	filename
    983	     64	      0	   1047	    417	/tmp/c/kernel/rcu/srcutiny.o

   text	   data	    bss	    dec	    hex	filename
   6844	    193	      0	   7037	   1b7d	/tmp/b/kernel/rcu/srcutree.o

So perhaps it is time to unconditionally enable SRCU?

							Thanx, Paul