On 22-09-19, Sakari Ailus wrote: > On Mon, Sep 19, 2022 at 03:49:04PM +0300, Laurent Pinchart wrote: > > On Mon, Sep 19, 2022 at 12:39:35PM +0000, Sakari Ailus wrote: > > > > + dev_set_drvdata(dev, tc358746); > > > > + pm_runtime_set_autosuspend_delay(dev, 200); > > > > + pm_runtime_use_autosuspend(dev); > > > > + pm_runtime_enable(dev); > > > > + > > > > + err = tc358746_init_hw(tc358746); > > > > > > The driver depends on runtime PM being enabled but does not depend on > > > CONFIG_PM. I'd suggest to power the device on and only then enable runtime > > > PM. See > > > <URL:https://hverkuil.home.xs4all.nl/spec/driver-api/camera-sensor.html#power-management>. > > > > Or simply depend on CONFIG_PM :-) > > The user can still disable runtime PM. You mean by user-space? If I read the runtime.c code correctly in such case the core handles this enabling if one forbid it by increasing the usage-counter and calling the resume callback. So it is powered as you want. To fix the PM Kconfig, I will add "depends on PM". Regards, Marco > I guess few do though. This should be addressed separately from this > driver, it'd be good to be able to deny that. In fact, no-one has ever > probably tested this for a large number of drivers and I guess it exists > just to have a way to disable runtime PM support in drivers that do not > work with it... > > -- > Sakari Ailus >
