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 S933266AbcL0TvZ (ORCPT <rfc822;w@1wt.eu>);
        Tue, 27 Dec 2016 14:51:25 -0500
Received: from asavdk4.altibox.net ([109.247.116.15]:35907 "EHLO
        asavdk4.altibox.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1755038AbcL0TvS (ORCPT
        <rfc822;linux-kernel@vger.kernel.org>);
        Tue, 27 Dec 2016 14:51:18 -0500
Date: Tue, 27 Dec 2016 20:41:53 +0100
From: Sam Ravnborg <sam@ravnborg.org>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: LKML <linux-kernel@vger.kernel.org>,
        Sebastian Siewior <bigeasy@linutronix.de>,
        Ingo Molnar <mingo@kernel.org>, Peter Zijlstra <peterz@infradead.org>
Subject: Re: [patch 00/10] cpu/hotplug: Final cleanup
Message-ID: <20161227194153.GA18149@ravnborg.org>
References: <20161221191947.089046759@linutronix.de>
 <20161222175923.GB17658@ravnborg.org>
 <alpine.DEB.2.20.1612271211490.3687@nanos>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <alpine.DEB.2.20.1612271211490.3687@nanos>
User-Agent: Mutt/1.5.21 (2010-09-15)
X-CMAE-Score: 0
X-CMAE-Analysis: v=2.2 cv=eqGd9chX c=1 sm=1 tr=0
        a=Ij76tQDYWdb01v2+RnYW5w==:117 a=Ij76tQDYWdb01v2+RnYW5w==:17
        a=kj9zAlcOel0A:10 a=k9TznIpmSekJfdp_jCsA:9 a=b4s8k4UWueUzy4Nt:21
        a=CjuIK1q_8ugA:10
Sender: linux-kernel-owner@vger.kernel.org
List-ID: <linux-kernel.vger.kernel.org>
X-Mailing-List: linux-kernel@vger.kernel.org

> > Just curious - what is the stage TWO (and onwards) plan?
> > lwn had some coverage of this work but I assume the
> > article is somewhat outdated.
> 
> Stage 2 is to consolidate all the archictecture code, i.e. reduce it to
> really arch specific callbacks and keep all the other things
> (synchronization, state settings etc.) in the core code.
> 
> After that we can do smart things like speed up the boot hotplug process:
> 
> prepare CPU1
> kick CPU1
> 
> prepare CPU2
> kick CPU2
> 
> ....
> 
> prepare CPUN
> kick CPUN
> 
> check CPU1 alive
> bringup CPU1
> 
> ...
> 
> check CPUN alive
> bringup CPUN
> 
> That saves ~200ms per CPU on x86, which we now spend busy waiting for the
> CPU to come alive.
> 
> There is more stuff which we can do in the long run like synchrounosly
> letting CPUs bring up themself after the basic init/sync checks, but that
> needs lot of surgery all over the place.

Thanks for the update - looking forward to follow this great clean-up!

	Sam