On Wed, Dec 19, 2018 at 10:49:08PM +0530, Maruthi Srinivas Bayyavarapu wrote:

This looks mostly good, a few small things but all pretty trivial:

> +	switch (bits_per_sample) {
> +	case 8:
> +		val |= (BIT_DEPTH_8 << AUD_CTRL_DATA_WIDTH_SHIFT);
> +		break;
> +	case 16:
> +		val |= (BIT_DEPTH_16 << AUD_CTRL_DATA_WIDTH_SHIFT);
> +		break;
> +	case 20:
> +		val |= (BIT_DEPTH_20 << AUD_CTRL_DATA_WIDTH_SHIFT);
> +		break;
> +	case 24:
> +		val |= (BIT_DEPTH_24 << AUD_CTRL_DATA_WIDTH_SHIFT);
> +		break;
> +	case 32:
> +		val |= (BIT_DEPTH_32 << AUD_CTRL_DATA_WIDTH_SHIFT);
> +		break;
> +	}

It'd be better to have a default case returning an error for unsupported
sample sizes.

> +		ret = devm_request_irq(dev, aud_drv_data->mm2s_irq,
> +				       xlnx_mm2s_irq_handler, 0,
> +				       "xlnx_formatter_pcm_mm2s_irq", dev);
> +		if (ret) {
> +			dev_err(dev, "xlnx audio mm2s irq request failed\n");
> +			goto clk_err;
> +		}
> +		ret = xlnx_formatter_pcm_reset(aud_drv_data->mmio +
> +					       XLNX_MM2S_OFFSET);
> +		if (ret) {
> +			dev_err(dev, "audio formatter reset failed\n");
> +			goto clk_err;
> +		}
> +		xlnx_formatter_disable_irqs(aud_drv_data->mmio +
> +					    XLNX_MM2S_OFFSET,
> +					    SNDRV_PCM_STREAM_PLAYBACK);

This requests the interrupt before we've reset and quiesced the hardware
meaning the interrupt could fire immediately on request with the
hardware in an unknown state; it'd be better to reset the hardware first
then request the interrupt to minimize the risk that something goes
wrong with the handler.

> +	dev_info(dev, "pcm platform device registered\n");
> +	return 0;

This is just adding noise to the log.