2022-05-14 at 16:26, Wolfram Sang wrote: > On Tue, Apr 12, 2022 at 10:50:45AM +0200, Vincent Whitchurch wrote: >> When drivers with ->detect callbacks are loaded, the I2C core does a >> bunch of transactions to try to probe for these devices, regardless of >> whether they are specified in the devicetree or not. (This only happens >> on I2C controllers whose drivers enable the I2C_CLASS* flags, but this >> is the case for generic drivers like i2c-gpio.) >> >> These kinds of transactions are unnecessary on systems where the >> devicetree specifies all the devices on the I2C bus, so add a property >> to indicate that the devicetree description of the hardware is complete >> and thus allow this discovery to be disabled. > > Hmm, I don't think the name is fitting. "no-detect" is the desired > behaviour but a proper description is more like "bus-complete" or > something? > > That aside, I am not sure we should handle this at DT level. Maybe we > should better change the GPIO driver to not populate a class if we have > a firmware node? We also have the somewhat related address translation case (which I still need to look at). [Adding Luca to Cc] https://lore.kernel.org/lkml/20220206115939.3091265-1-luca@xxxxxxxxxxxxxxxx/ If a bus is "bus-complete", then address translation could use any unused address instead of from an explicit list of addresses. I.e. the "i2c-alias-pool" in the binding in patch 4/6 of that series could be made optional if the bus is "bus-complete". Not sure how much value there is in that? Cheers, Peter
