From mboxrd@z Thu Jan  1 00:00:00 1970
From: broonie@opensource.wolfsonmicro.com (Mark Brown)
Date: Wed, 16 Mar 2011 10:53:47 +0000
Subject: [PATCH] ARM: S3C64XX: Fix section mismatch from cpufreq init
In-Reply-To: <20110316101259.GH13316@pengutronix.de>
References: <1299769093-14516-1-git-send-email-broonie@opensource.wolfsonmicro.com>
	<20110316083811.GG13316@pengutronix.de>
	<20110316094959.GB2688@opensource.wolfsonmicro.com>
	<20110316101259.GH13316@pengutronix.de>
Message-ID: <20110316105346.GA14125@opensource.wolfsonmicro.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Wed, Mar 16, 2011 at 11:13:00AM +0100, Uwe Kleine-K?nig wrote:
> On Wed, Mar 16, 2011 at 09:49:59AM +0000, Mark Brown wrote:

> > That's what I said above, pretty much.  Ideally we'd have a way of
> > annotating the call sites so that the linker infrastructure rather than
> > having to faff about splitting the structures like this.

> You can mark the driver struct using __ref. But IMHO not saving a
> reference to a function only used once is cleaner.

It's pretty annoying to have to define two structs for the affected
drivers (it's not like cpufreq is the only affected subsystem here) -
there's good reasons why we use things like __devexit_p() rather than
split structs for common cases like this.