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 S1756123AbcHaHTV (ORCPT <rfc822;w@1wt.eu>);
        Wed, 31 Aug 2016 03:19:21 -0400
Received: from merlin.infradead.org ([205.233.59.134]:44534 "EHLO
        merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1752630AbcHaHTS (ORCPT
        <rfc822;linux-kernel@vger.kernel.org>);
        Wed, 31 Aug 2016 03:19:18 -0400
Date: Wed, 31 Aug 2016 09:18:10 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Oleg Nesterov <oleg@redhat.com>, Balbir Singh <bsingharora@gmail.com>,
        LKML <linux-kernel@vger.kernel.org>,
        Nicholas Piggin <nicholas.piggin@gmail.com>
Subject: Re: [RFC][PATCH] Fix a race between rwsem and the scheduler
Message-ID: <20160831071810.GZ10138@twins.programming.kicks-ass.net>
References: <4050f2ce-1aee-d2aa-39e3-36e995b56252@gmail.com>
 <20160830121937.GQ10138@twins.programming.kicks-ass.net>
 <20160830130426.GA17795@redhat.com>
 <20160830141321.GB2794@worktop>
 <20160830165746.GA29218@redhat.com>
 <20160830183416.GV10138@twins.programming.kicks-ass.net>
 <1472592498.2388.40.camel@kernel.crashing.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1472592498.2388.40.camel@kernel.crashing.org>
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 Wed, Aug 31, 2016 at 07:28:18AM +1000, Benjamin Herrenschmidt wrote:
> It's always been a requirement that if you actually context switch a
> full mb() is implied ...

> On powerpc we have a sync deep in _switch to achieve that.

OK, fair enough. I must've missed it in the x86 switch_to, must be one
of those implied serializing instructions I'm not too familiar with.

> (though that isn't the case if you don't actually
> switch, ie, you are back to RUNNING before you even hit schedule).

Right, which invalidates the claim that schedule() implies a full mb,

> This is necessary so that a process who wakes up on a different CPU sees
> all of its own load/stores.

Don't actually think its needed for that, see the comment from
8643cda549ca4, the scheduler has enough barriers to guarantee
Program-Order for tasks without that.