On Sun, Feb 06, 2022 at 03:16:24PM +0000, Jonathan Cameron wrote: > On Sat, 5 Feb 2022 20:58:12 +0200 > Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx> wrote: > > On Sat, Feb 05, 2022 at 08:50:31PM +0200, Andy Shevchenko wrote: ... > > https://lore.kernel.org/lkml/20211205190101.26de4a57@jic23-huawei/T/#u > > > > So, the fwnode has a correct implementation, and we may use it here. > > > I wasn't totally sure of the conclusion of that discussion. > a) Fine to just use device_for_each_child_node() for this case and not worry > about it. Yes. As he mentioned the device_for_each_child_node() is implemented correctly from day 1. > b) Worth adding device_for_each_available_child_node() with the same > implementation I believe it's an opposite prospective, i.e. drop of_for_each_available_child_node() and use the of_for_each_child_node() everywhere. > c) (possibly workaround / avoid the issue) Use device_for_each_child_node() > but also check validity (hopefully compiler would remove the check) > in order to act as documentation. Makes no sense because implementation does it already. > I'm fine with any of the above. -- With Best Regards, Andy Shevchenko
