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 S1753679AbcIBMHV (ORCPT <rfc822;w@1wt.eu>);
        Fri, 2 Sep 2016 08:07:21 -0400
Received: from mx1.redhat.com ([209.132.183.28]:36286 "EHLO mx1.redhat.com"
        rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
        id S1751982AbcIBMHT (ORCPT <rfc822;linux-kernel@vger.kernel.org>);
        Fri, 2 Sep 2016 08:07:19 -0400
Date: Fri, 2 Sep 2016 14:06:43 +0200
From: Oleg Nesterov <oleg@redhat.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>, Al Viro <viro@ZenIV.linux.org.uk>,
        Bart Van Assche <bvanassche@acm.org>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        Neil Brown <neilb@suse.de>, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2/2] sched/wait: avoid abort_exclusive_wait() in
        __wait_on_bit_lock()
Message-ID: <20160902120642.GC26495@redhat.com>
References: <20160826124453.GA28894@redhat.com> <20160826124552.GB28904@redhat.com> <20160901190141.GJ10138@twins.programming.kicks-ass.net> <20160901221743.GJ10168@twins.programming.kicks-ass.net>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20160901221743.GJ10168@twins.programming.kicks-ass.net>
User-Agent: Mutt/1.5.18 (2008-05-17)
X-Greylist: Sender IP whitelisted, not delayed by milter-greylist-4.5.16 (mx1.redhat.com [10.5.110.31]); Fri, 02 Sep 2016 12:07:19 +0000 (UTC)
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

On 09/02, Peter Zijlstra wrote:
>
> FWIW, the way the mutex code avoids this issue is by doing the
> signal_pending test while holding the q->lock, that way its exclusive
> with wakeup.

And __wait_event_interruptible_locked() too.

BTW it is buggy anyway, it needs the

	-	__add_wait_queue_tail(&(wq), &__wait);
	+	if (exclusive)
	+		__add_wait_queue_tail(&(wq), &__wait);
	+	else
	+		__add_wait_queue((&(wq), &__wait);

and in fact it should use __add_wait_queue_exclusive() so that we
can remove another "if (exclusive)" but this is off-topic.

Yes, I considered this option, but to me the addtional finish_wait()
looks simpler.

And, if you agree with this change I will try to change __wait_event()
as well and kill abort_exclusive_wait().

And in this case we certainly do not want to check the "condition" with
q->lock held, because this would mean that "condition" won't be able to
take this lock.

Oleg.