On 11/12/2013 08:35 AM, Li Xiubo wrote:

quoted

quoted

quoted

quoted

+static int fsl_sai_probe(struct platform_device *pdev) {

[...]

quoted

+
+	sai->dma_params_rx.addr = res->start + SAI_RDR;
+	sai->dma_params_rx.maxburst = 6;
+	index = of_property_match_string(np, "dma-names", "rx");
+	ret = of_parse_phandle_with_args(np, "dmas", "#dma-cells",

index,

quoted

quoted

quoted

+				&dma_args);
+	if (ret)
+		return ret;
+	sai->dma_params_rx.slave_id = dma_args.args[1];
+
+	sai->dma_params_tx.addr = res->start + SAI_TDR;
+	sai->dma_params_tx.maxburst = 6;
+	index = of_property_match_string(np, "dma-names", "tx");
+	ret = of_parse_phandle_with_args(np, "dmas", "#dma-cells",

index,

quoted

quoted

quoted

+				&dma_args);
+	if (ret)
+		return ret;
+	sai->dma_params_tx.slave_id = dma_args.args[1];

The driver should not have to manually parse the dma devicetree
properties, this is something that should be handled by the dma
engine driver.

What do you think about the DMA slave_id ?
I have been noticed by one colleague that this should be parsed here,
which is from your opinions ?

Sure slave_id can be parsed here, but IMO it should be programmed via the
dma_slave_confog into the respective channel

Actually, these are parsed for cpu_dai->playback_dma_data and cpu_dai->capture_dma_data dynamically, whose type is struct dma_slave_config.

And now I must parse them here, because the platform eDMA driver's newest version will check and use the slave_ids to select and configure the eDMA channels via dma_device->device_control().

Parsing them here is a layering violation. The format of the DMA specifier 
depends on the DMA controller. A DMA slave should not make any assumptions 
about how the specifier looks like, it should not even look at them. You should 
fix the DMA controller driver to work without slave_id in the devicetree case.

- Lars