Hello Philipp, On 18.08.22 10:19, Philipp Zabel wrote: > Hi Ahmad, > > On Do, 2022-08-18 at 07:19 +0200, Ahmad Fatoum wrote: >> The newly added feature controller framework has two goals: Avoid >> probing device in barebox that aren't indeed available > > This specific wording makes me wonder why this isn't implemented inside > of_device_is_available(). > > This would also take care of other devices querying device node > availability, for example via of_graph_port_is_available(), if that > ever happens. I admit that restricting the barebox-side of feature controllers to device probe can be limiting. For kernel fixup, any node may be disabled, unlike with barebox, where only device probes can be skipped. Placing the feature controller check in of_device_is_available is not trivial. We currently do a single pass through the device tree and create devices that are of_device_is_available(). If we start to look up the feature controller there, we will need to do device creation in multiple passes. While of_probe() can be called multiple times, I think there are quite a few cornercase we will need to take into consideration when doing this and are thus better addressed separately when the need arises. Thanks, Ahmad > > regards > Philipp > -- Pengutronix e.K. | | Steuerwalder Str. 21 | http://www.pengutronix.de/ | 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
