On Fri, Aug 13, 2021 at 3:53 PM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote: > On Thu, Aug 12, 2021 at 2:33 PM Andy Shevchenko > <andy.shevchenko@xxxxxxxxx> wrote: > > On Wednesday, August 11, 2021, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote: > >> On Wed, Aug 11, 2021 at 12:48 PM Marek Behún <kabel@xxxxxxxxxx> wrote: > >> > On Wed, 11 Aug 2021 11:57:59 +0200 ... > >> Sure, that can be done later, when an ACPI user appears? > > > > Actually with PRP0001 approach any of compatible driver may be used onACPI platform. So, what you are saying can be interpreted the way “we don’t care about users on ACPI based platforms”. If it is the case, then it should be told explicitly. > > I think you're interpreting too much ;-) > My point is simply: > > >> The dependency on OF was pre-existing, and this series is already > >> at v5. Okay, but we can get rid of it. Why not make it more generic at the same time? Does it make sense? (I believe this is what Marek is asking initially) > If any OF compatible driver can now be used on ACPI platforms, perhaps > this should be handled at the API level? I.e. the distinction between > OF and device properties should be dropped completely, And this is done by device_*() / fwnode_*() APIs. And that's what can be easily done here. > and all drivers > be converted mechanically in one shot, instead of a gradual ad-hoc > conversion being sneaked in through other series like this one? Do you realize that you are asking for something impossible? Moreover, an ad-hoc approach is what we do for plenty of things in the kernel (WRT new APIs, that don't replace old ones immediately). -- With Best Regards, Andy Shevchenko
