From mboxrd@z Thu Jan  1 00:00:00 1970
From: Andrew.Jackson@arm.com (Andrew Jackson)
Date: Thu, 04 Dec 2014 09:42:13 +0000
Subject: [PATCH 4/5] ASoC: dwc: Add devicetree support for Designware I2S
In-Reply-To: <20141203182351.GF7712@sirena.org.uk>
References: <547F3CAC.9050105@arm.com> <20141203182351.GF7712@sirena.org.uk>
Message-ID: <54802C75.9060609@arm.com>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On 12/03/14 18:23, Mark Brown wrote:
> On Wed, Dec 03, 2014 at 04:39:08PM +0000, Andrew Jackson wrote:
> 
>> Convert to driver to use either platform_data or device-tree for configuration
>> of the device.  When using device-tree, the I2S block's configuration is read
>> from the relevant registers: this reduces the amount of information required in
>> the device tree.
> 
> This really needs to be split into two or more patches, there's a whole
> bunch of refactoring to support this DT stuff which should be separate
> from the DT addition itself.  Right now it's hard to tell what each
> individual bit of the code is supposed to be doing, the patch is far too
> large and doing far too many individual things.

I will have look at how it might be split.  The majority of the new code is in reading and processing the device's configuration: I didn't want to change the platform data handling to do that because some of the comments in the driver suggested that there were ST specific changes to the Designware IP.  I wasn't in a position to know whether, if I changed the configuration reading, the driver would still function correctly on the SPEAR platform.

>> +	if (dev->using_pd) {
>> +		ret = dev->i2s_clk_cfg(config);
>> +		if (ret < 0) {
>> +			dev_err(dev->dev, "runtime audio clk config fail\n");
>> +			return ret;
>> +		}
>> +	} else {
>> +		u32	bitclk;
> 
> Having this whole separate path for using platform data feels icky, we
> don't want to have completely separate flows like this.  Checking for
> the callbacks being there is probably fine but just having totally
> separate code paths is a bit icky.

I wasn't very happy either but making the test explicit seemed reasonable at the time.  I'll change the code to test for the presence of the callback instead.

	Andrew