Hi Torsten, On 2026-10-15T16:39:07, Torsten Duwe <[email protected]> wrote:
> Solution A seems to be the least problematic one; it enhances U-Boot's > capabilities. > > Solution B is more like the Linux kernel does it, but without allocating its > own node, > which makes it look a bit hacky. I agree with Tom - solution A looks best. The appeal of B is that any later code which asks the dynamic device for its ofnode gets a plausible answer, but that answer is a lie: dev_has_ofnode() would then return true on a device whose properties, compatible string, reg etc. actually belong to an ancestor bridge. Any future caller that assumes 'ofnode present => describes this device' would quietly misread, and the breakage would be far from this file. Keeping the fake-ofnode trick out of the core and localising the walk to device_get_dma_constraints() seems much safer. > Solution C would be for an ideal world [TM] ;) > Nonetheless we could at least for the RPi5 craft a DT overlay and send it > upstream. An RPi5 overlay would be nice to have regardless, but it does not help the general case where U-Boot enumerates past the last described node, so it is not a substitute for A. I would land A now and treat the overlay as an independent clean-up. Regards, Simon
