On 2019-06-08 12:55, Wolfram Sang wrote: > While preparing a refactoring series, I noticed that some drivers use a > complicated way of determining the adapter of a client. The easy way is > to use the intended pointer: client->adapter > > These drivers do: > to_i2c_adapter(client->dev.parent); > > The I2C core populates the parent pointer as: > client->dev.parent = &client->adapter->dev; > > Now take into consideration that > to_i2c_adapter(&adapter->dev); > > is a complicated way of saying 'adapter', then we can even formally > prove that the complicated expression can be simplified by using > client->adapter. > > The conversion was done using a coccinelle script with some manual > indentation fixes applied on top. > > To avoid a brown paper bag mistake, I double checked this on a Renesas > Salvator-XS board (R-Car M3N) and verified both expression result in the > same pointer. Other than that, the series is only build tested. Similar things go on in: drivers/hwmon/lm90.c drivers/leds/leds-is31fl319x.c drivers/of/unittest.c Those have this pattern: struct device *dev = &client->dev; struct i2c_adapter *adapter = to_i2c_adapter(dev->parent); And drivers/rtc/rtc-fm3130.c has a couple of these: tmp = i2c_transfer(to_i2c_adapter(fm3130->client->dev.parent), ...); where fm3130->client is of type "struct i2c_client *" Cheers, Peter
