On 12/31/2013 03:16 PM, Mark Brown wrote:

On Fri, Dec 20, 2013 at 12:38:27PM +0200, Jyri Sarha wrote:

quoted

+static int evm_startup(struct snd_pcm_substream *substream)
+{
+	struct snd_soc_pcm_runtime *rtd = substream->private_data;
+	struct snd_soc_card *soc_card = rtd->codec->card;
+	struct clk *mclk = ((struct snd_soc_card_drvdata_davinci *)
+			    snd_soc_card_get_drvdata(soc_card))->mclk;

Why do you need to cast away void?  Ths indicates something is going
wrong here though I can't see what.

I'll fix that.

quoted

+	mclk = of_clk_get_by_name(np, "ti,codec-clock");
+	if (PTR_ERR(mclk) == -EPROBE_DEFER) {
+		return -EPROBE_DEFER;
+	} else if (IS_ERR(mclk)) {
+		dev_dbg(&pdev->dev, "Codec clock not found.\n");
+		mclk = NULL;
+	}

The driver will unconditionally enable and disable the clock which I'd
not expect to work well if we got an error, I'd expect either NULL checks
on use or a fixed clock to be registered from code in the case where
we're using the old binding.

In the drivers/clk/clk.c the clk == NULL is always checked before using 
the pointer. However, adding NULL checks would save couple of 
lock-unlock cycles. I'll add them.

I'd also expect to see devm_clk_get() used here, with the standard
clock-names based lookup from DT.

I'll fix that.

Best regards,
Jyri