Date: Wed, 4 Nov 2015 10:11:40 -0800
From: Davidlohr Bueso <dave@stgolabs.net>
To: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: linux-kernel@vger.kernel.org, George Spelvin <linux@horizon.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Peter Zijlstra <peterz@infradead.org>,
        Manfred Spraul <manfred@colorfullife.com>,
        Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH v2] ipc/msg: Implement lockless pipelined wakeups
Message-ID: <20151104181140.GB15994@linux-uzut.site>
References: <1446563009-9076-1-git-send-email-bigeasy@linutronix.de>
 <20151103173021.GE1707@linux-uzut.site>
 <56391547.2040105@linutronix.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Disposition: inline
In-Reply-To: <56391547.2040105@linutronix.de>
User-Agent: Mutt/1.5.21 (2010-09-15)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 829
Lines: 21

On Tue, 03 Nov 2015, Sebastian Andrzej Siewior wrote:

>In pipelined_receive() (mqueue) there is the wake_q_add() with the
>implicit cmpxchg barrier. The matching barrier pairing should be in
>wq_sleep() but there is none. Why is it okay to have none there and I
>need one here?

So the pairing in mqueue is done in wq_sleep() as we unlock info->lock,
so there's an implied barrier there.

Also, considering Peter's suggestion to wrap around READ/WRITE_ONCE
for sysv queues (specially since you got rid of the volatile tag),
it seems we also need them for mqueues.

Thanks,
Davidlohr
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/