On Wed, Oct 23, 2013 at 02:13:24PM +0800, Shawn Guo wrote:
> > +
> > +enum imx_phy_type {
> > + IMX6Q_USB_PHY,
> > + IMX6SL_USB_PHY,
> > + IMX23_USB_PHY,
> > +};
> > +
> > struct mxs_phy {
> > struct usb_phy phy;
> > struct clk *clk;
> > + enum imx_phy_type devtype;
> > };
> >
> > -#define to_mxs_phy(p) container_of((p), struct mxs_phy, phy)
> > +static inline int is_mx6q_phy(struct mxs_phy *data)
> > +{
> > + return data->devtype == IMX6Q_USB_PHY;
> > +}
> > +
> > +static inline int is_mx6sl_phy(struct mxs_phy *data)
> > +{
> > + return data->devtype == IMX6SL_USB_PHY;
> > +}
> > +
> > +static inline int is_mx23_phy(struct mxs_phy *data)
> > +{
> > + return data->devtype == IMX23_USB_PHY;
> > +}
> > +
> > +static struct platform_device_id imx_phy_devtype[] = {
> > + {
> > + .name = "usb-phy-imx6q",
> > + .driver_data = IMX6Q_USB_PHY,
> > + }, {
> > + .name = "usb-phy-imx6sl",
> > + .driver_data = IMX6SL_USB_PHY,
> > + }, {
> > + .name = "usb-phy-imx23",
> > + .driver_data = IMX23_USB_PHY,
> > + }, {
> > + /* sentinel */
> > + }
> > +};
>
> I know many imx device drivers have this platform_device_id table, but
> that's because they need to support both non-DT and DT probe. Since
> this driver supports DT probe only, we can save this table by passing
> imx_phy_type value through of_device_id.data directly.
How about compare compatible string directly at probe?
if (of_device_is_compatible(np, "fsl,imx6q-usbphy"))
mxs_phy->devtype = IMX6Q_USB_PHY;
else if ((of_device_is_compatible(np, "fsl,imx6sl-usbphy"))
mxs_phy->devtype = IMX6SL_USB_PHY;
else if (...)
...;
I don't know how to passing imx_phy_type value through below table
directly? imx_phy_type is a enum variable, not a arrary.
>
> > +static const struct of_device_id mxs_phy_dt_ids[] = {
> > + { .compatible = "fsl,imx6q-usbphy", .data = &imx_phy_devtype[IMX6Q_USB_PHY], },
> > + { .compatible = "fsl,imx6sl-usbphy", .data = &imx_phy_devtype[IMX6SL_USB_PHY], },
> > + { .compatible = "fsl,imx23-usbphy", .data = &imx_phy_devtype[IMX23_USB_PHY], },
> > + { /* sentinel */ }
> > +};
> > +MODULE_DEVICE_TABLE(of, mxs_phy_dt_ids);
> >
> > static int mxs_phy_hw_init(struct mxs_phy *mxs_phy)
> > {
> > @@ -131,6 +177,14 @@ static int mxs_phy_probe(struct platform_device *pdev)
> > struct clk *clk;
> > struct mxs_phy *mxs_phy;
> > int ret;
> > + const struct of_device_id *of_id =
> > + of_match_device(mxs_phy_dt_ids, &pdev->dev);
> > +
> > + /* This driver is DT-only version now */
> > + if (!of_id)
> > + return -ENXIO;
>
> Since it's DT-only, I'm not sure you will run into the case that
> mxs_phy_probe() is called with a NULL of_id. The check looks
> unnecessary to me.
Will change.
>
> > +
> > + pdev->id_entry = of_id->data;
>
> Some imx device drivers did the same thing, but we should keep
> pdev->id_entry immutable. The removal of that platform_device_id table
> will help save this.
Surely we can delete this line if no platform_device_id table.
--
Best Regards,
Peter Chen
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
