Hi Maxime, On Fri, 2020-04-24 at 17:33 +0200, Maxime Ripard wrote: > The firmware clocks driver was previously probed through a platform_device > created by the firmware driver. > > Since we will now have a node for that clocks driver, we need to create the > device only in the case where there's no node for it already. > > Signed-off-by: Maxime Ripard > --- > drivers/firmware/raspberrypi.c | 17 +++++++++++++++-- > 1 file changed, 15 insertions(+), 2 deletions(-) > > diff --git a/drivers/firmware/raspberrypi.c b/drivers/firmware/raspberrypi.c > index da26a584dca0..1874f41b007c 100644 > --- a/drivers/firmware/raspberrypi.c > +++ b/drivers/firmware/raspberrypi.c > @@ -210,6 +210,15 @@ rpi_register_hwmon_driver(struct device *dev, struct > rpi_firmware *fw) > > static void rpi_register_clk_driver(struct device *dev) > { > + /* > + * Earlier DTs don't have a node for the firmware clocks but > + * rely on us creating a platform device by hand. If we do > + * have a node for the firmware clocks, just bail out here. > + */ > + if (of_get_compatible_child(dev->of_node, > + "raspberrypi,firmware-clocks")) > + return; > + > rpi_clk = platform_device_register_data(dev, "raspberrypi-clk", > -1, NULL, 0); > } > @@ -262,8 +271,12 @@ static int rpi_firmware_remove(struct platform_device > *pdev) > > platform_device_unregister(rpi_hwmon); > rpi_hwmon = NULL; > - platform_device_unregister(rpi_clk); > - rpi_clk = NULL; > + > + if (rpi_clk) { rpi_clk, being a static global variable, will be zeroed by default. So, if you don't register the platform device, rpi_clk is going to be NULL, which you're allowed to feed platform_device_unregister(), so no need to be extra careful. Regards, Nicolas > + platform_device_unregister(rpi_clk); > + rpi_clk = NULL; > + } > + > mbox_free_channel(fw->chan); > > return 0;