On Wed, 17 Oct 2012 16:32:39 -0500, Rob Herring <[email protected]> wrote: > On 10/17/2012 05:26 AM, Srinivas KANDAGATLA wrote: > > From: Srinivas Kandagatla <[email protected]> > > > > Some functions in of.h have dummy functions and real implementations, > > however few of them have only real implementation. > > I think to be more consistent this patch adds below dummy functions if > > for non CONFIG_OF's. > > > > Fixing these up one by one has annoyed me, but I know Grant has > generally resisted blindly creating empty functions. So I would like > Grant's comments. I'm fine with these except the ones noted below. > > > of_find_node_by_type > > This should not be used except a few places as using device_type is > generally only for true OF. > > > of_device_is_available > > of_find_node_with_property > > Only used in 2 places. > > > of_find_matching_node > > of_find_node_by_path > > of_find_node_by_phandle > > of_get_parent > > of_get_next_parent > > of_get_next_child > > of_get_next_available_child > > of_get_child_by_name > > Already have this one. > > > > > Hopefully this can eliminate the need to ifdef code in the drivers. > > Sometimes we want that for drivers that can be built both for OF and !OF > and you can't optimize away all the code and data.
Right. I've been cautious here for a reason. As much as possible I want to encourage of_* calls to be collected together in a driver, not scattered all over the place. Often the whole block of OF-aware code in a driver can be compiled out if it is grouped together into a function. Also, some of these only make sense in the context of an OF-aware code block and there is no sensible empty version. So, I'm fine with adding the empty versions on a case-by-case basis, but not wholesale. g. _______________________________________________ devicetree-discuss mailing list [email protected] https://lists.ozlabs.org/listinfo/devicetree-discuss
