Date: Mon, 31 Jul 2017 12:45:38 +0200
From: Peter Zijlstra <peterz@infradead.org>
To: Josef Bacik <josef@toxicpanda.com>
Cc: mingo@redhat.com, linux-kernel@vger.kernel.org,
        umgwanakikbuti@gmail.com, tj@kernel.org, kernel-team@fb.com,
        Josef Bacik <jbacik@fb.com>
Subject: Re: [PATCH 1/7] sched/fair: use reweight_entity to reweight tasks
Message-ID: <20170731104538.2a7n3ovhyegryvy6@hirez.programming.kicks-ass.net>
References: <1500038464-8742-1-git-send-email-josef@toxicpanda.com>
 <1500038464-8742-2-git-send-email-josef@toxicpanda.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1500038464-8742-2-git-send-email-josef@toxicpanda.com>
User-Agent: NeoMutt/20170609 (1.8.3)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 743
Lines: 19

On Fri, Jul 14, 2017 at 01:20:58PM +0000, Josef Bacik wrote:
> From: Josef Bacik <jbacik@fb.com>
> 
> reweight_task only accounts for the load average change in the cfs_rq, but
> doesn't account for the runnable_average change in the cfs_rq.  We need to do
> everything reweight_entity does, and then we just set our inv_weight
> appropriately.

The difference is in the calling convention. If you look at the
callsite:

	set_user_nice()
	  set_load_weight()
	    reweight_task()

You'll see that ->on_rq will always be false. That said, I think you're
right in that we're missing a se->runnable_weight update, because while
__update_load_avg*() doesn't use it (and will in fact (re)set it), there
are other users that could come before that.