On Thu, 2 Jun 2022 15:45:27 +0200 Andy Shevchenko <andy.shevchenko@xxxxxxxxx> wrote: > On Thu, Jun 2, 2022 at 2:08 PM Lars-Peter Clausen <lars@xxxxxxxxxx> wrote: > > On 6/2/22 13:42, Andy Shevchenko wrote: > > ... > > > > - priv->adc_clk = devm_clk_register(&indio_dev->dev, &priv->clk_gate.hw); > > > + priv->adc_clk = devm_clk_register(idev, &priv->clk_gate.hw); > > > > You are not changing anything here. > > The scope of patch is supposed not to change the current behaviour :-) > > > But we shouldn't be devm'ing on the > > IIO device. It will get freed eventually, but only when the last > > reference to the iio device has been dropped, which might be long after > > the platform device has been removed. devm'ing should happen on the > > platform_device's device. Might be worth fixing. > > Thanks for confirming my suspicions (as I mentioned to Martin, using > an IIO device there feels wrong). > I will add another patch to v3. > I thought that the iio_dev ends up holding a reference to the platform device dev so the parent would only be released if the child had already been, but I may well be wrong on that. Either way I don't mind it being tidied up. Jonathan
