On Fri, Sep 7, 2018 at 10:16 AM Chris Brandt <Chris.Brandt@xxxxxxxxxxx> wrote: > > On Thursday, August 30, 2018, Daniel Lezcano wrote: > > > AFAIK no attempt was done to support EPROBE_DEFER with *_OF_DECLARE. > > > IMHO it would be pointless, as it would be much easier to just switch to > > real > > > platform drivers. > > > > May be, may be not. > > > > From your point of view, the change is simple because it touches only a > > single driver. > > > > From my point of view, the change implies a split in the approach while > > I'm trying to unify the drivers little by little and there are hundred > > of them. > > > > It is not the first time we face this situation and Bartosz Golaszewski > > has a similar problem [1]. > > > > We have all the frameworks we need to solve this properly but I would > > like something we can propagate to all drivers (OF and !OF) so we end up > > with unified code. > > > > It is time we clearly state the dependency issues and we find a proper > > way to solve it. > > > On Thursday, August 30, 2018, Bartosz Golaszewski wrote: > > This was my latest proposal for early platform drivers: > > > > https://lkml.org/lkml/2018/5/11/488 > > > > I still intend on continuing this work, I just don't have the time right > > now. > > Daniel, > > So what is your final thought on this? > > The current OSTM driver uses TIMER_OF_DECLARE and that basically means > it will never work with my new SoC. > > For now, can I change the driver to register a standard platform driver > in subsys_initcall like the other Renesas timer drivers? I'm confused how this can even work as an initcall. The whole reason *_OF_DECLARE exists is for things that have to be setup before initcalls. Rob
