On Fri, Aug 10, 2018 at 01:30:48PM +0100, Kieran Bingham wrote: > Hi Simon, > > On 10/08/18 13:01, Simon Horman wrote: > > On Tue, Aug 07, 2018 at 04:59:33PM +0100, Kieran Bingham wrote: > >> Ensure that the ADV748x device addresses do not conflict, and group them > >> together (visually in i2cdetect) > > > > Hi Kieran, > > > > could you help me out with some pointers on how to correlate this > > with the HW documentation? > > Not easily :) - Except for the 'main' address (0x70), these are not > addresses documented by the datasheet. <by the very nature of the > activity>. The driver supports the DT providing the slave pages to > allocate. One day the I2C framework might allow us to 'request' an > unused page :D > > I performed a scan on the Salvator-X, (i2cdetect -r -y 4) and identified > a region of unused I2C address space to allocate the 12 pages so that > they did not conflict. > > In particular, the address <0x30> which is the default for the CBUS page > conflicts with the default address of the OV10635 used by the GMSL > cameras on the same bus, and so I needed to move that one. > > To make the effects clear, (and the i2cdetect reporting more obvious) I > chose to reassign all of the movable pages so that it is clear they are > from the same device. > > Rather annoyingly it's difficult to map a slave page back to it's driver > due to the fact that it gets registered as a 'dummy' driver, so keeping > the related addresses together is quite valuable. > > As soon as I get some free cycles, I plan to look at being able to map > extra driver information to dummy I2C registrations to make it easier to > identify who owns the address :)
Thanks for the follow-up, that makes things a lot clearer. I see that the corresponding bindings patches have been reviewed. I have now gone ahead and applied this patch for v4.20.
