Re: [PATCH 207/493] i2c: remove use of __devinit

From: Jean Delvare <khali@linux-fr.org>
To: Russell King - ARM Linux <linux@arm.linux.org.uk>
Cc: Rudolf Marek <r.marek@assembler.cz>,
	Tony Lindgren <tony@atomide.com>,
	gregkh@linuxfoundation.org, Guan Xuetao <gxt@mprc.pku.edu.cn>,
	Wolfram Sang <w.sang@pengutronix.de>,
	"Mark M. Hoffman" <mhoffman@lightlink.com>,
	Bill Pemberton <wfp5p@virginia.edu>,
	linux-i2c@vger.kernel.org, Ben Dooks <ben-linux@fluff.org>,
	Barry Song <baohua.song@csr.com>, Olof Johansson <olof@lixom.net>,
	linux-omap@vger.kernel.org, linuxppc-dev@lists.ozlabs.org,
	linux-arm-kernel@lists.infradead.org
Subject: Re: [PATCH 207/493] i2c: remove use of __devinit
Date: Tue, 20 Nov 2012 14:37:43 +0100	[thread overview]
Message-ID: <20121120143743.3f54a3cb@endymion.delvare> (raw)
In-Reply-To: <20121120132342.GP3332@n2100.arm.linux.org.uk>

On Tue, 20 Nov 2012 13:23:42 +0000, Russell King - ARM Linux wrote:
> On Tue, Nov 20, 2012 at 09:20:46AM +0100, Jean Delvare wrote:
> > Hi Bill,
> > 
> > On Mon, 19 Nov 2012 13:22:36 -0500, Bill Pemberton wrote:
> > > CONFIG_HOTPLUG is going away as an option so __devinit is no longer
> > > needed.
> > 
> > Can you please point me/us to the discussion explaining the rationale
> > behind this move, and the explanation of what will be done exactly?
> > While I can easily understand that we want to drop CONFIG_HOTPLUG and
> > always enable hot-plug support, I don't see where we are going with
> > removing __devinit annotations and the like.
> 
> It's actually very simple to understand.
> 
> 1. CONFIG_HOTPLUG is going away; it's already defined to always be 'Y'.
> 2. This means that the the devinit sections will not be discarded anymore.
> 3. As a result, there's no point the devinit sections existing anymore.
> 4. As there's no devinit sections, the __devinit marker is entirely
>    redundant and useless.

Ah, yes, very simple indeed. Not sure how I managed to not understand
it earlier today. Thanks for explaining.

> The reason this is being done is because the benefit to cost ratio of this
> is far too high; it's well proven that people constantly get these markings
> wrong, and with most kernels having had hotplug enabled anyway, it's not
> providing much in the way of space saving benefit over the number of section
> conflicts it causes.  So, it's been decided a few years ago that this is

Yes, I completely agree.

> going to happen, with that justification, and it's been accepted by 300
> odd kernel developers in at least one kernel summit when it was talked

Probably that was one I didn't attend to, as I can't remember this
discussion.

> about... and it's been mentioned on mailing lists several times.

Maybe LKML, which I don't read. So I wasn't aware of the plan before
seeing Bill's patches.

-- 
Jean Delvare