Hi Chiawei, On Mon, 14 Dec 2020, at 13:14, ChiaWei Wang wrote: > Hi Andrew & Rob, > > Do you have any suggestion on this patch? Rob hasn't responded, but I think it will be easier to get an Ack out of him if we do a v2 of the binding so we're not breaking backwards-compatibility with the current definition. Concretely: - compatible: One of: "aspeed,ast2400-lpc", "simple-mfd" "aspeed,ast2500-lpc", "simple-mfd" Becomes something like: - compatible: One of: "aspeed,ast2400-lpc-v2", "simple-mfd" "aspeed,ast2500-lpc-v2", "simple-mfd" We can convert the in-tree devicetrees, immediately drop support for the current binding in the drivers, and _only_ support v2 of the binding going forward. That way your patches stay largely the same, the binding isn't hamstrung as it is currently, and we're not trying to maintain code to support the current binding definition - but we're also not pretending that old devicetrees will work with newer kernels that only support the new binding definition (which is the problem with your current patch series). How does that sound? As to how to implement this, I think we'll need to add some of_device_is_compatible() checks in the relevant drivers to make sure that they're using the new LPC binding, such as in drivers/pinctrl/aspeed/pinctrl-aspeed-g5.c before we fetch the regmap on line 2657. Sorry that this is dragging out a bit (and for the mess I made). Cheers, Andrew