On Sat, May 30, 2020 at 3:32 AM Wolfram Sang <wsa@xxxxxxxxxxxxx> wrote: > > > > + addr = reg & 0x3FFFFFFFU; > > + snprintf(unit_addr, sizeof(unit_addr), "%x", addr); > > Hmm, this hardcoded value will not work if we ever need to add another > bit. I hope this will never happen, though. I had this concern and requested the first time this was submitted (and abandoned) to just mask out the top byte. However, Joel's version of this fix[1] does some actual checks on 10-bit addressing, so I've dropped that request. > > + if ((reg & (1U << 31)) && addr > 0x3ff) > > Same here with bit 31. I haven't checked DTC but can't we import the > header with the defines into the project? Or is this then a circular > dependency? Easier to just duplicate the define here which Joel's patches do. Rob [1] https://www.spinics.net/lists/devicetree-compiler/msg03196.html
