On Fri, Jun 19, 2015 at 06:50:03PM +0900, Chanwoo Choi wrote: > On Fri, Jun 19, 2015 at 6:13 PM, Charles Keepax > <ckeepax@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > > On Fri, Jun 19, 2015 at 05:39:22PM +0900, Chanwoo Choi wrote: > >> On Fri, Jun 19, 2015 at 5:14 PM, Charles Keepax > >> <ckeepax@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > >> > On Fri, Jun 19, 2015 at 11:36:47AM +0900, Chanwoo Choi wrote: > >> >> Hi Charles, > >> >> > >> >> On Thu, Jun 18, 2015 at 11:43 PM, Charles Keepax > >> >> <ckeepax@xxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > >> >> > Convert to using the newer gpiod interface for the micd_pol_gpio. > >> >> > Although we still carry support for the old gpio interface from pdata. > >> >> > > >> >> > Signed-off-by: Charles Keepax <ckeepax@xxxxxxxxxxxxxxxxxxxxxxxxxxx> > >> >> > --- > >> >> > + mode = GPIOD_OUT_HIGH; > >> >> > + else > >> >> > + mode = GPIOD_OUT_LOW; > >> >> > + > >> >> > + info->micd_pol_gpio = gpiod_get_optional(arizona->dev, > >> >> > + "wlf,micd-pol", > >> >> > + GPIOD_OUT_LOW); > >> >> > >> >> You can use the devm_gpiod_get_optional() to manage the system > >> >> resource automatically. > >> >> > >> > > >> > We can't actually use the devm call here, we need to pass > >> > arizona->dev as that is where the DT will reside, which is the > >> > device for the MFD. But if the devm is attached to the device for > >> > the MFD then it will not clear up when the extcon driver is > >> > unloaded. As such we have to do the put manually. > >> > > >> > I will look at respinning for the other comments. > >> > >> I don't understand. extcon-arizona.c used already following devm_* functions: > >> - devm_kzalloc() > >> - devm_regulator_get() > >> - devm_extcon_dev_*() > >> - devm_input_allocate_device() > >> - devm_gpio_request_one() > > > > Yes but if you look at those all of those are against &pdev->dev > > which is the extcon device. > > > > The gpiod interface expects the device passed to both contain > > the of_node and be used for the devm operations. But we need to > > use the extcon device for the devm operations, but the of_node is > > contained on the MFD device. > > > > So if I do: > > > > devm_gpiod_get_optional(arizona->dev, .... > > > > Then the gpiod won't be released when the extcon device is > > removed. But if I do: > > > > devm_gpiod_get_optional(&pdev->dev, .... > > > > Then it won't be able to find the DT entries. > > I understand the difference between arizona->dev and &pdev->dev. > > But, extcon-arizona.c already get the instance of regulator by using > devm_regulator_get() with &pdev->dev as following: > The devm_regulator_get() can find the DT entry in dts file. > info->micvdd = devm_regulator_get(&pdev-dev, "MICVDD"); > > How did extcon-arizona.c get the instance of regulator throught > devm_regulator_get() in dts file? The regulator API contains a feature called regulator aliases, (see regulator_register_supply_alias) which the MFD driver for Arizona registers a bunch of. So when the regulator lookup is done the regulator core will actually do the lookup on the parent MFD device rather than on the extcon device. Thanks, Charles -- 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