Hi Laurent, On Sun, Mar 13, 2022 at 04:16:44PM +0200, Laurent Pinchart wrote: > Hi Sakari, > > On Sun, Mar 13, 2022 at 03:38:34PM +0200, Sakari Ailus wrote: > > On Sun, Mar 13, 2022 at 03:01:52PM +0200, Laurent Pinchart wrote: > > > On Fri, Mar 11, 2022 at 03:32:26PM +0200, Sakari Ailus wrote: > > > > On Fri, Mar 11, 2022 at 03:20:55PM +0200, Laurent Pinchart wrote: > > > > > On Fri, Mar 11, 2022 at 03:15:54PM +0200, Sakari Ailus wrote: > > > > > > On Fri, Mar 11, 2022 at 02:30:09PM +0200, Laurent Pinchart wrote: > > > > > > > On Fri, Mar 11, 2022 at 02:23:53PM +0200, Sakari Ailus wrote: > > > > > > > > On Fri, Mar 11, 2022 at 08:12:59PM +0900, Paul Elder wrote: > > > > > > > > > Switch to using runtime PM for power management. > > > > > > > > > > > > > > > > > > Signed-off-by: Paul Elder <paul.elder@xxxxxxxxxxxxxxxx> > > > > > > > > > > > > > > > > > > --- > > > > > > > > > Changes in v2: > > > > > > > > > - replace manual tracking of power status with pm_runtime_get_if_in_use > > > > > > > > > - power on the sensor before reading the checking the chip id > > > > > > > > > - add dependency on PM to Kconfig > > > > > > > > > --- > > > > > > > > > drivers/media/i2c/Kconfig | 1 + > > > > > > > > > drivers/media/i2c/ov5640.c | 112 ++++++++++++++++++++++--------------- > > > > > > > > > 2 files changed, 67 insertions(+), 46 deletions(-) > > > > > > > > > > > > > > > > > > diff --git a/drivers/media/i2c/Kconfig b/drivers/media/i2c/Kconfig > > > > > > > > > index e7194c1be4d2..97c3611d9304 100644 > > > > > > > > > --- a/drivers/media/i2c/Kconfig > > > > > > > > > +++ b/drivers/media/i2c/Kconfig > > > > > > > > > @@ -1025,6 +1025,7 @@ config VIDEO_OV5640 > > > > > > > > > tristate "OmniVision OV5640 sensor support" > > > > > > > > > depends on OF > > > > > > > > > depends on GPIOLIB && VIDEO_V4L2 && I2C > > > > > > > > > + depends on PM > > > > > > > > > > > > > > > > I think this is not needed as the sensor is powered on explicitly in probe. > > > > > > > > > > > > > > > > You should similarly power it off explicitly in remove, set the runtime PM > > > > > > > > status suspended and disable runtime PM. See e.g. imx319 driver for an > > > > > > > > example. It doesn't have resume callback but that doesn't really matter --- > > > > > > > > it's just ACPI-only. > > > > > > > > > > > > > > Do we want to continue supporting !PM ? Does it have any real use case > > > > > > > when dealing with camera sensors ? > > > > > > > > > > > > Probably not much. > > > > > > > > > > > > The changes I proposed are not eve related on runtime PM. Hence the > > > > > > question here is whether there should be a dependency to CONFIG_PM or not, > > > > > > and as there's no technical reason to have it, it should be omitted. > > > > > > > > > > But if there's no real use case for !PM, wouldn't we be better off > > > > > depending on PM and simplifying the probe functions instead ? > > > > > > > > What would change in the probe function if runtime PM was required by the > > > > driver? > > > > > > We wouldn't need the complicated dance of calling > > > > > > ret = ov5640_set_power(sensor, true); > > > if (ret) > > > goto free_ctrls; > > > > > > pm_runtime_set_active(dev); > > > pm_runtime_enable(dev); > > > pm_runtime_get(dev); > > > > pm_runtime_get() is redundant here. > > > > > > > > but could write it as > > > > > > pm_runtime_enable(dev); > > > pm_runtime_resume_and_get(dev); > > > > You'll need put here, too. > > Yes, after reading the version register (or doing any other harware > access). Actually the full code would be > > > pm_runtime_enable(dev); > pm_runtime_resume_and_get(dev); > > /* Hardware access */ > > pm_runtime_set_autosuspend_delay(dev, 1000); > pm_runtime_use_autosuspend(dev); > pm_runtime_put_autosuspend(dev); > > (plus error handling). > > If the probe function doesn't need to access the hardware, then > the above becomes > > pm_runtime_enable(dev); > pm_runtime_set_autosuspend_delay(dev, 1000); > pm_runtime_use_autosuspend(dev); > > instead of having to power up the device just in case !PM. > > > Also the latter only works on DT-based systems so it's not an option for > > most of the drivers. > > How so, what's wrong with the above for ACPI-based system ? I²C devices are already powered on for probe on ACPI based systems. -- Regards, Sakari Ailus
