On Fri, May 17, 2024 at 04:49:20PM GMT, Wolfram Sang wrote: > > > + /* > > + * Enforce some basic assumptions this function makes about the > > + * transfer. If this proves insufficient, some more complex logic will > > + * be needed. > > + */ > > + if (num > 2 || (num == 2 && msgs[0].addr != msgs[1].addr)) > > + return -EOPNOTSUPP; > > As you populated 'ad24xx_i2c_adapter_quirks' in the I2C driver, you can > drop this. The I2C core will do the checks for you. > The i2c_xfer function here is also available as a general A2B API, see a2b.h: int a2b_node_i2c_xfer(struct a2b_node *node, struct i2c_msg *msgs, int num); This is used by the beo-shape-node.c driver submitted later in this series to perform a firmware update of a more peculiar A2B hardware. In this case it doesn't factor through the codepath you mention, hence this check. It's conceivable that there will be other such cases in the future as well. ADI for example prescribes a specific EEPROM address where device identification data can be stored with a well-defined format. In the event that the driver should support some kind of device type detection, it will also have to perform some I2C transfers out-of-band like this.
