From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <linux-kernel-owner@vger.kernel.org>
Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand
	id S1751880AbcGNSgP (ORCPT <rfc822;w@1wt.eu>);
	Thu, 14 Jul 2016 14:36:15 -0400
Received: from merlin.infradead.org ([205.233.59.134]:46463 "EHLO
	merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751292AbcGNSgL (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Thu, 14 Jul 2016 14:36:11 -0400
Date: Thu, 14 Jul 2016 20:36:09 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Oleg Nesterov <oleg@redhat.com>
Cc: Tejun Heo <tj@kernel.org>, John Stultz <john.stultz@linaro.org>,
        Ingo Molnar <mingo@redhat.com>, lkml <linux-kernel@vger.kernel.org>,
        Dmitry Shmidt <dimitrysh@google.com>,
        Rom Lemarchand <romlem@google.com>, Colin Cross <ccross@google.com>,
        Todd Kjos <tkjos@google.com>,
        "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Subject: Re: Severe performance regression w/ 4.4+ on Android due to cgroup
 locking changes
Message-ID: <20160714183609.GJ30154@twins.programming.kicks-ass.net>
References: <CALAqxLV2H06M2AxP2kK7pGHjLgQAaGdHB9HJEGZjH_Rouy4WhQ@mail.gmail.com>
 <20160713182102.GJ4065@mtj.duckdns.org>
 <20160713183347.GK4065@mtj.duckdns.org>
 <CALAqxLU76kVm8b5GNnSk25-Xk6D1efiWtwNJbY1th_pcN9Ks=g@mail.gmail.com>
 <20160713201823.GB29670@mtj.duckdns.org>
 <20160713202657.GW30154@twins.programming.kicks-ass.net>
 <20160713203944.GC29670@mtj.duckdns.org>
 <20160713205102.GZ30909@twins.programming.kicks-ass.net>
 <20160714131809.GO30927@twins.programming.kicks-ass.net>
 <20160714180951.GA16454@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20160714180951.GA16454@redhat.com>
User-Agent: Mutt/1.5.23.1 (2014-03-12)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On Thu, Jul 14, 2016 at 08:09:52PM +0200, Oleg Nesterov wrote:
> On 07/14, Peter Zijlstra wrote:
> >
> > The below is a compile tested only first draft so far. I'll go give it
> > some runtime next.
> 
> So I will wait for the new version, but at first glance this matches the
> code I already reviewed in the past (at least, tried hard to review ;)
> and it looks correct.
> 
> Just a couple of almost cosmetic nits, feel free to ignore.

Drat, I just mailed out the patches.. I can do a second version later.

> > --- a/include/linux/percpu-rwsem.h
> > +++ b/include/linux/percpu-rwsem.h
> > @@ -10,29 +10,107 @@
> >  
> >  struct percpu_rw_semaphore {
> >  	struct rcu_sync		rss;
> > -	unsigned int __percpu	*fast_read_ctr;
> > +	unsigned int __percpu	*refcount;
> >  	struct rw_semaphore	rw_sem;
> > -	atomic_t		slow_read_ctr;
> > -	wait_queue_head_t	write_waitq;
> > +	wait_queue_head_t	writer;
> > +	int			state;
>         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> 
> I think that this "int state" and "enum { readers_slow, readers_block }"
> just add a bit of complication/confusion.
> 
> All we need is the simple "bool readers_block" in percpu_rw_semaphore,
> no?

So I detest bool in structures because sizeof(bool) isn't defined.
Obviously an implementation needs to pick a size, but this is typically
architecture ABIs, so sizes can differ between architectures.

But yes, I suppose "int readers_block" will do just fine.

IIRC, earlier version had more states, but that all went away.

> > +void percpu_down_write(struct percpu_rw_semaphore *sem)
> >  {
> > +	down_write(&sem->rw_sem);
> > +
> > +	/* Notify readers to take the slow path. */
> > +	rcu_sync_enter(&sem->rss);
> 
> I'd suggest to call rcu_sync_enter() before down_write(). This can help
> when we wait for another writer which holds this lock.

Hurm, I think I figured that might have issues, but I cannot seem to
think of any just now :-), yes can do.