On Wed, Jan 08, 2020 at 02:35:33PM +0100, Wolfram Sang wrote: > > > >> I.e. if say the adv748x had in DT defined aliases at 0x08, 0x09, > > >> 0x0A..., but not yet probed (thus no device is listening at these > > >> addresses) ... and then a max9286 came along and asked for 'any' spare > > >> address with this call, would it be given 0x08 first? > > You have a point here. Ancillary addresses are not blocked until the > driver probes, this is true. I wonder now if we should handle multiple > addresses in i2c-core-of.c somehow, too? It does block the first <reg> > entry, but not all. > > > The core 'could' parse all reg entries, and conclude that any extended > > entries within a device node are aliases as well, which should be > > reserved, but I don't think it could know if the device is actually > > going to be enabled by a driver (well, it could look it up). > > We could argue that if it is described in DT, it should be blocked in > any case, or? That seems fair to me. > > But it would also have to traverse any i2c-muxes too! > > I probably need a second thought about muxes as well. -- Regards, Laurent Pinchart
