On Wed, Jan 30, 2019 at 11:17 AM William Breathitt Gray <vilhelm.gray@xxxxxxxxx> wrote: > On Tue, Jan 29, 2019 at 05:07:34PM -0800, Andrew Morton wrote: > > Do we have any expectation that these facilities will be used by > > anything other than GPIO? If not then perhaps they should be sited > > within drivers/gpio (presumably as a standalone module) until such a > > need is found? > > I can move it within drivers/gpio since my only use at moment is for > these gpio drivers I maintain. However, moving it to the gpio subsystem > does make it less likely to be seen by authors in other subsystems who > may have use for it -- so there is the chance that this becomes isolated > and untilized only amonst the gpio drivers. That may not be such a bad > thing in the end; I suspect it will be easy to spot if other subsystems > start implementing their own for_each_set_clump (if other subsystems > would even have such a need for it). I would sure prefer to have these in the generic core where this patch puts them. I don't see the problem with that. We have to start generic code from some point. > Linus, if I move the for_each_set_clump macro and related functions to > drivers/gpio, should I move the code into the gpiolib.h and gpiolib.c > files? Or would it be best to implement this as a separate standalone > module, and make a Kconfig dependency for those gpio drivers which > require it? A standalone module feels somewhat overkill for such little > code in my opinion. Yeah. I was however thinking about that we should create a helper library for port-mapped IO along the lines of gpio-mmio.c so we might as well get started with that in that case, just call it gpio-port-mapped-io.c and put these in first or something. Yours, Linus Walleij
