From mboxrd@z Thu Jan  1 00:00:00 1970
From: dwalker@codeaurora.org (Daniel Walker)
Date: Thu, 30 Sep 2010 13:40:45 -0700
Subject: [PATCH 2/4] arm: dis-allow hotplug on MSM
In-Reply-To: <20100930201749.GA11660@n2100.arm.linux.org.uk>
References: <1284062289-10914-1-git-send-email-dwalker@codeaurora.org>
	<1284062289-10914-3-git-send-email-dwalker@codeaurora.org>
	<20100930071919.GA6434@ucw.cz>
	<1285865137.14953.4.camel@c-dwalke-linux.qualcomm.com>
	<1285870183.14953.7.camel@c-dwalke-linux.qualcomm.com>
	<20100930195237.GA2174@elf.ucw.cz>
	<1285876546.14953.10.camel@c-dwalke-linux.qualcomm.com>
	<20100930200246.GB2174@elf.ucw.cz>
	<1285877287.14953.13.camel@c-dwalke-linux.qualcomm.com>
	<20100930201749.GA11660@n2100.arm.linux.org.uk>
Message-ID: <1285879245.14953.29.camel@c-dwalke-linux.qualcomm.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Thu, 2010-09-30 at 21:17 +0100, Russell King - ARM Linux wrote:
> On Thu, Sep 30, 2010 at 01:08:07PM -0700, Daniel Walker wrote:
> > That's actually what I had original, Russell offered this as an
> > alternative. Now that I think about the stubs tho, I'm not sure it's
> > that bad an idea. All the stubs would do is not put the core into a
> > lower powermode, but would allow suspend. It wouldn't be very
> > efficient , but suspend would work.
> 
> If you just provide empty stubs, esp. for platform_cpu_die(), then an
> attempt to take a CPU offline will result in it immediately restarting
> back into the kernel - which will probably result in an oops as the
> kernel won't be expecting the CPU to come back.

Ok ..

> If you make them spin, you'll eat power until the system powers them off.
> As things currently stand though, we don't have enough code out of the
> init sections to allow a SMP restart from power-off.  So, S2RAM isn't
> going to work on secondary processors at the moment _anyway_.

When you say "SMP restart" do you mean resume? If I did make them spin
wouldn't suspend and resume work at that point?

> Current CPU hotplug implementations are based around taking CPUs offline
> without powering them off for run-time power saving only.

What I was thinking with the stubs is to fake the offlining. So it's not
power efficient , but suspend still works.

Daniel

-- 

Sent by a consultant of the Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.