From mboxrd@z Thu Jan  1 00:00:00 1970
From: arnd@arndb.de (Arnd Bergmann)
Date: Fri, 17 May 2013 15:17:31 +0200
Subject: [PATCH v3 2/4] ARM: imx: add initial support for MVF600
In-Reply-To: <20130517130603.GJ8607@S2101-09.ap.freescale.net>
References: <1368684648-20826-1-git-send-email-b35083@freescale.com>
 <201305171429.52646.arnd@arndb.de>
 <20130517130603.GJ8607@S2101-09.ap.freescale.net>
Message-ID: <4003326.AbobOukrYT@wuerfel>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Friday 17 May 2013 21:06:05 Shawn Guo wrote:
> On Fri, May 17, 2013 at 02:29:52PM +0200, Arnd Bergmann wrote:
> > > > Actually I think you should move that driver to drivers/clk and use
> > > > of_clk_init(NULL) to initialize it.
> > > > 
> > > The mvf600 clock driver uses a lot of base clk support from mach-imx,
> > > and can not be moved into drivers/clk as a single driver.  Right now, in
> > > IMX clock drivers, we call of_clk_init() to only register fixed rate
> > > clocks, since all the other clocks are not represented in device tree.
> > 
> > What are your plans for this in the long run?
> > 
> We can move all the IMX clock drivers into drivers/clk at some point
> when necessary.  But I do not have a plan to register all the clocks
> by merely calling of_clk_init(), because doing that would mean we have
> to represent all these clocks in device tree.  For imx6q example, it's
> about 200 ~ 300 nodes addition to DTB.  Device tree maintainers are
> against to the idea.  They are perfectly fine with having clock driver
> in kernel to represent/register these SoC internal clocks to clk
> framework.

Can't we move the driver to drivers/clk and have it initialized through
of_clk_init() without representing all clocks in the DT?

For all I can tell, CLK_OF_DECLARE() requires only the clock provider
to be described in DT, but not the actual clocks.

	Arnd