On Fri, May 28, 2021 at 12:02:54PM +0200, Pavel Machek wrote: > On Mon 2021-05-17 10:30:08, Andy Shevchenko wrote: > > On Mon, May 10, 2021 at 12:50:17PM +0300, Andy Shevchenko wrote: > > > When analyzing the current state of affairs with fwnode reference counting > > > I found that a lot of core doesn't take it right. Here is a bunch of > > > corresponding fixes against LED drivers. > > > > > > The series includes some cleanups and a few other fixes grouped by a driver. > > > > > > First two patches are taking care of -ENOTSUPP error code too prevent its > > > appearance in the user space. > > > > Pavel, any comments on this bug fix series? > > I took these: Thanks! What branch/tree should I rebase the rest on? > For the "remove depends on OF"... I'd preffer not to take those. We > don't need to ask the user for configurations that never happen. What do you mean by this? ACPI is quite a good configuration to make use of it on the corresponding platforms. By default any discrete LED driver (in hardware term here) IC should be considered independent from the type of the platform description. Do you agree? If so, it means that dropping OF dependency is a right thing to do to allow users of those ICs to be happy even on ACPI based platforms. Note, entire IIO subsystem is a good example of this activity. All the sensors can be used now in ACPI environment without explicit requirement to have an ACPI ID, although it's highly recommended to acquire for the real products (not DIY ones). -- With Best Regards, Andy Shevchenko
