[V5 PATCH 25/26] usb: otg: mv_otg: add device tree support

[Mark Rutland]

On Thu, Jan 24, 2013 at 06:38:49AM +0000, Chao Xie wrote:
> All blocks are removed. Add the device tree support for otg.

As with mv_udc, this binding should be documented. Please Cc
devicetree-discuss when you have one.

> 
> Signed-off-by: Chao Xie <chao.xie at marvell.com>
> ---
>  drivers/usb/otg/mv_otg.c |  128 +++++++++++++++++++++++++++++++++++++--------
>  drivers/usb/otg/mv_otg.h |    6 +-
>  2 files changed, 108 insertions(+), 26 deletions(-)
> 
> diff --git a/drivers/usb/otg/mv_otg.c b/drivers/usb/otg/mv_otg.c
> index 379df92..3aa7fdc 100644
> --- a/drivers/usb/otg/mv_otg.c
> +++ b/drivers/usb/otg/mv_otg.c
> @@ -17,6 +17,7 @@
>  #include <linux/device.h>
>  #include <linux/proc_fs.h>
>  #include <linux/clk.h>
> +#include <linux/of.h>
>  #include <linux/workqueue.h>
>  #include <linux/platform_device.h>
>  
> @@ -333,7 +334,7 @@ static void mv_otg_update_inputs(struct mv_otg *mvotg)
>  	else
>  		otg_ctrl->id = !!(otgsc & OTGSC_STS_USB_ID);
>  
> -	if (mvotg->pdata->otg_force_a_bus_req && !otg_ctrl->id)
> +	if (mvotg->otg_force_a_bus_req && !otg_ctrl->id)
>  		otg_ctrl->a_bus_req = 1;
>  
>  	otg_ctrl->a_sess_vld = !!(otgsc & OTGSC_STS_A_SESSION_VALID);
> @@ -690,21 +691,69 @@ int mv_otg_remove(struct platform_device *pdev)
>  	return 0;
>  }
>  
> +static int mv_otg_parse_dt(struct platform_device *pdev,
> +				struct mv_otg *mvotg)
> +{
> +	struct device_node *np = pdev->dev.of_node;
> +	unsigned int clks_num;
> +	unsigned int val;
> +	int i, ret;
> +	const char *clk_name;
> +
> +	if (!np)
> +		return 1;
> +
> +	clks_num = of_property_count_strings(np, "clocks");
> +	if (clks_num < 0)
> +		return clks_num;
> +
> +	mvotg->clk = devm_kzalloc(&pdev->dev,
> +		sizeof(struct clk *) * clks_num, GFP_KERNEL);
> +	if (mvotg->clk == NULL)
> +		return -ENOMEM;
> +
> +	for (i = 0; i < clks_num; i++) {
> +		ret = of_property_read_string_index(np, "clocks", i,
> +			&clk_name);
> +		if (ret)
> +			return ret;
> +		mvotg->clk[i] = devm_clk_get(&pdev->dev, clk_name);
> +		if (IS_ERR(mvotg->clk[i]))
> +			return PTR_ERR(mvotg->clk[i]);
> +	}

Again, it seems a shame there's no devm_of_clk_get.

> +
> +	mvotg->clknum = clks_num;
> +
> +	ret = of_property_read_u32(np, "extern_attr", &mvotg->extern_attr);
> +	if (ret)
> +		return ret;
> +
> +	ret = of_property_read_u32(np, "mode", &mvotg->mode);
> +	if (ret)
> +		return ret;

I *really* don't like this, for the same reasons I stated in reply to the
mv_udc patch.

> +
> +	ret = of_property_read_u32(np, "force_a_bus_req", &val);
> +	if (ret)
> +		return ret;
> +	mvotg->otg_force_a_bus_req = !!val;

In devicetree, the typical way to handle a boolean case like this would be to
have a valueless property. If present, the property is true, if not present
it's considered to default to false:

device at 4000 {
	compatible = "manufacturer,some-device";
	reg = <0x4000, 0x1000>;
	manufacturer,boolean-property; /* no value, true if present */
};

Properties which may only have a meaning on one manufacturer's devices are also
typically prefixed with the manufacturer similarly to compatible strings, e.g.
"mrvl,force-a-bus-req".

There may also be a better name for this property.

> +
> +	ret = of_property_read_u32(np, "disable_clock_gating", &val);
> +	if (ret)
> +		return ret;
> +	mvotg->clock_gating = !val;

You can do the same here, but with the logic inverted.

> +
> +	return 0;
> +}
> +
>  static int mv_otg_probe(struct platform_device *pdev)
>  {
> -	struct mv_usb_platform_data *pdata = pdev->dev.platform_data;
>  	struct mv_otg *mvotg;
>  	struct usb_otg *otg;
>  	struct resource *r;
> -	int retval = 0, clk_i, i;
> +	int retval = 0, i;
>  	size_t size;
>  
> -	if (pdata == NULL) {
> -		dev_err(&pdev->dev, "failed to get platform data\n");
> -		return -ENODEV;
> -	}
> -
> -	size = sizeof(*mvotg) + sizeof(struct clk *) * pdata->clknum;
> +	size = sizeof(*mvotg);
>  	mvotg = devm_kzalloc(&pdev->dev, size, GFP_KERNEL);
>  	if (!mvotg) {
>  		dev_err(&pdev->dev, "failed to allocate memory!\n");
> @@ -718,17 +767,45 @@ static int mv_otg_probe(struct platform_device *pdev)
>  	platform_set_drvdata(pdev, mvotg);
>  
>  	mvotg->pdev = pdev;
> -	mvotg->extern_attr = pdata->extern_attr;
> -	mvotg->pdata = pdata;
>  
> -	mvotg->clknum = pdata->clknum;
> -	for (clk_i = 0; clk_i < mvotg->clknum; clk_i++) {
> -		mvotg->clk[clk_i] = devm_clk_get(&pdev->dev,
> +	retval = mv_otg_parse_dt(pdev, mvotg);
> +	if (retval > 0) {
> +		struct mv_usb_platform_data *pdata = pdev->dev.platform_data;
> +		/* no CONFIG_OF */
> +		int clk_i = 0;
> +
> +		if (pdata == NULL) {
> +			dev_err(&pdev->dev, "missing platform_data\n");
> +			return -ENODEV;
> +		}
> +		mvotg->extern_attr = pdata->extern_attr;
> +		mvotg->mode = pdata->mode;
> +		mvotg->clknum = pdata->clknum;
> +		mvotg->otg_force_a_bus_req = pdata->otg_force_a_bus_req;
> +		if (pdata->disable_otg_clock_gating)
> +			mvotg->clock_gating = 0;
> +
> +		size = sizeof(struct clk *) * mvotg->clknum;
> +		mvotg->clk = devm_kzalloc(&pdev->dev, size, GFP_KERNEL);
> +		if (mvotg->clk == NULL) {
> +			dev_err(&pdev->dev,
> +				"failed to allocate memory for clk\n");
> +			return -ENOMEM;
> +		}
> +
> +		for (clk_i = 0; clk_i < mvotg->clknum; clk_i++) {
> +			mvotg->clk[clk_i] = devm_clk_get(&pdev->dev,
>  						pdata->clkname[clk_i]);
> -		if (IS_ERR(mvotg->clk[clk_i])) {
> -			retval = PTR_ERR(mvotg->clk[clk_i]);
> -			return retval;
> +			if (IS_ERR(mvotg->clk[clk_i])) {
> +				dev_err(&pdev->dev, "failed to get clk %s\n",
> +					pdata->clkname[clk_i]);
> +				retval = PTR_ERR(mvotg->clk[clk_i]);
> +				return retval;

You don't need to go via retval here.

[...]

Thanks,
Mark.