From mboxrd@z Thu Jan  1 00:00:00 1970
From: mfuzzey@parkeon.com (Martin Fuzzey)
Date: Thu, 18 Apr 2013 10:36:53 +0200
Subject: [PATCH V2] ARM: i.MX5: Allow DT clock providers
In-Reply-To: <20130418063056.GA20566@S2101-09.ap.freescale.net>
References: <20130415080741.20173.70704.stgit@localhost>
 <20130418063056.GA20566@S2101-09.ap.freescale.net>
Message-ID: <516FB0A5.9050903@parkeon.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 18/04/13 08:30, Shawn Guo wrote:
> +static struct clk * __init mx5_obtain_fixed_clock_from_dt(const char *name)
> +{
> +#ifdef CONFIG_OF
> We have selected USE_OF at the sub-architecture level, so the ifdef
> plays nothing here.
Ah ok so what happens on non DT platforms (looks like Babbage)
Does this mean that there is an empty DT in that case?

I was going by the existing code which already has this #ifdef

>> +	struct of_phandle_args phandle = {0};
>> +	struct clk *clk = ERR_PTR(-ENODEV);
>> +	char *compatible;
>> +
>> +	compatible = kasprintf(GFP_KERNEL, "fsl,imx-%s", name);
> What about finding the node by path as below?
>
> 	path = kasprintf(GFP_KERNEL, "/clocks/%s", name);
> 	phandle.np = of_find_node_by_path(path);
>
> I have to admit that those imx custom compatibles for fixed-clock
> shouldn't necessarily be there at all.  Let's stop spreading the use.
I can't find anything in the DT binding documentation that says the 
clocks have to be
under /clocks  (but then again the fsl,imx- compatible string isn't 
documented either).

My understanding of DT is that the absolute node path is generally 
unimportant with
the exception of a few special nodes like /, /memory, /chosen

But I don't have a strong opinion on this - if the consensus is that 
path based lookup
is better I'll do that.

Regards,

Martin