> And there is a regression! Good that I didn't push out before > double-checking. No one noticed that this breaks registering child > devices because of_i2c_register_devices() doesn't have a pointer to work > with anymore? Well, sorry, I forgot an important detail. There is no regression because most drivers still populate adap->dev.of_data with the node pointer of their parent. I experimentally removed this from my driver under test motivated by this comment from the commit in the Fixes: tag: "Linking it to the device node of the parent device is wrong, as it leads to 2 devices sharing the same device node, which is bad practice," But removing this bad practice from I2C core is more work. I wonder now if we are in some inconsistent in-between state if I apply this patch as is?
Attachment:
signature.asc
Description: PGP signature
