Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Sat, 3 Feb 2018 17:10:06 -0500 (EST)
From:   Alan Stern <stern@rowland.harvard.edu>
To:     "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
cc:     Ingo Molnar <mingo@kernel.org>, <linux-kernel@vger.kernel.org>,
        <linux-arch@vger.kernel.org>, <parri.andrea@gmail.com>,
        <j.alglave@ucl.ac.uk>, <luc.maranget@inria.fr>,
        <boqun.feng@gmail.com>, <will.deacon@arm.com>,
        <peterz@infradead.org>, <npiggin@gmail.com>, <dhowells@redhat.com>,
        <elena.reshetova@intel.com>, <mhocko@suse.com>, <akiyks@gmail.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Peter Zijlstra <a.p.zijlstra@chello.nl>,
        Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [GIT PULL tools] Linux kernel memory model
In-Reply-To: <20180203084809.GE3617@linux.vnet.ibm.com>
Message-ID: <Pine.LNX.4.44L0.1802031708380.29515-100000@netrider.rowland.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sat, 3 Feb 2018, Paul E. McKenney wrote:

> Please see below for an initial patch to this effect.  This activity
> proved to be more productive than expected for these tests, which certainly
> supports our assertion that locking needs more testing...
> 
> MP+polocks.litmus
> MP+porevlocks.litmus
> 
> 	These are allowed by the current model, which surprised me a bit,
> 	given that even powerpc would forbid them.  Is the rationale
> 	that a lock-savvy compiler could pull accesses into the lock's
> 	critical section and then reorder those accesses?  Or does this
> 	constitute a bug in our model of locking?
> 
> 	(And these were allowed when I wrote recipes.txt, embarrassingly
> 	enough...)
> 
> Z6.0+pooncelock+poonceLock+pombonce.litmus
> 
> 	This was forbidden when I wrote recipes.txt, but now is allowed.
> 	The header comment for smp_mb__after_spinlock() makes it pretty
> 	clear that it must be forbidden.  So this one is a bug in our
> 	model of locking.

I just tried testing these under the most recent version of herd, and 
all three were forbidden.

Alan