On Mon, Mar 8, 2021 at 7:37 PM Rob Herring <r...@kernel.org> wrote: > Can we remove the ones we have already for GPIO?
I think we would get pretty hard pushback if we attempt that. We have all these drivers that utilize it: gpio-clps711x.c: id = of_alias_get_id(np, "gpio"); gpio-mvebu.c: id = of_alias_get_id(pdev->dev.of_node, "gpio"); gpio-mxc.c: port->gc.base = (pdev->id < 0) ? of_alias_get_id(np, "gpio") * 32 : gpio-mxs.c: port->id = of_alias_get_id(np, "gpio"); gpio-vf610.c: gc->base = of_alias_get_id(np, "gpio") * VF610_GPIO_PER_PORT; gpio-zynq.c: chip->base = of_alias_get_id(pdev->dev.of_node, "gpio"); pinctrl-at91.c: int alias_idx = of_alias_get_id(np, "gpio"); pinctrl-st.c: int bank_num = of_alias_get_id(np, "gpio"); samsung/pinctrl-samsung.c: id = of_alias_get_id(node, "pinctrl"); Predictably it is so many bad examples that new driver authors will claim something along the line of "why can't I have a lollipop when all other kids got one". Several of those have this by a claim one way or another that the DT boot need to look like the boardfile boot. Some of these have been migrated from board files so could possible drop this id/base coding. I don't know what the maintainers would say, should we send attack patches? :D At least some kind of motivation would come out of it. Yours, Linus Walleij
