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 S1751969AbcFWLTa (ORCPT <rfc822;w@1wt.eu>);
	Thu, 23 Jun 2016 07:19:30 -0400
Received: from bombadil.infradead.org ([198.137.202.9]:38425 "EHLO
	bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
	with ESMTP id S1751734AbcFWLT2 (ORCPT
	<rfc822;linux-kernel@vger.kernel.org>);
	Thu, 23 Jun 2016 07:19:28 -0400
Date: Thu, 23 Jun 2016 13:19:20 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Yuyang Du <yuyang.du@intel.com>, Ingo Molnar <mingo@kernel.org>,
        linux-kernel <linux-kernel@vger.kernel.org>,
        Mike Galbraith <umgwanakikbuti@gmail.com>,
        Benjamin Segall <bsegall@google.com>, Paul Turner <pjt@google.com>,
        Morten Rasmussen <morten.rasmussen@arm.com>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Matt Fleming <matt@codeblueprint.co.uk>,
        Vincent Guittot <vincent.guittot@linaro.org>
Subject: Re: [PATCH 4/4] sched,fair: Fix PELT integrity for new tasks
Message-ID: <20160623111920.GJ30154@twins.programming.kicks-ass.net>
References: <20160617120136.064100812@infradead.org>
 <20160617120454.150630859@infradead.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20160617120454.150630859@infradead.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 Fri, Jun 17, 2016 at 02:01:40PM +0200, Peter Zijlstra wrote:
> @@ -8219,6 +8254,19 @@ static int cpu_cgroup_can_attach(struct
>  		if (task->sched_class != &fair_sched_class)
>  			return -EINVAL;
>  #endif
> +		/*
> +		 * Serialize against wake_up_new_task() such
> +		 * that if its running, we're sure to observe
> +		 * its full state.
> +		 */
> +		raw_spin_unlock_wait(&task->pi_lock);
> +		/*
> +		 * Avoid calling sched_move_task() before wake_up_new_task()
> +		 * has happened. This would lead to problems with PELT. See
> +		 * XXX.
> +		 */
> +		if (task->state == TASK_NEW)
> +			return -EINVAL;
>  	}
>  	return 0;
>  }

So I think that's backwards; we want:

		if (task->state == TASK_NEW)
			return -EINVAL;
		raw_spin_unlock_wait(&task->pi_lock);

Because failing the attach is 'safe', but if we do not observe NEW we
must be absolutely sure to observe the full wakeup_new.

But since its not critical I'll change it to more obvious code:

		raw_spin_lock_irq(&task->pi_lock);
		if (task->state == TASK_NEW)
			ret = -EINVAL;
		raw_spin_unlock_irq(&task->pi_lock);