On Sun, Sep 16, 2012 at 08:44:56PM +0200, Jean Delvare wrote: > On Sun, 16 Sep 2012 09:19:34 -0700, Guenter Roeck wrote: > > On Sun, Sep 16, 2012 at 03:09:01PM +0200, Jean Delvare wrote: > > > On Tue, 11 Sep 2012 13:52:50 -0700, Guenter Roeck wrote: > > > That's a very nice cleanup, but it makes me wonder... > > > > > > Wouldn't it make sense to create the "name" attribute of spi devices at > > > the spi core level, as we do for i2c devices? All spi-based hwmon > > > drivers create the hwmon attributes as the spi device level, as is done > > > for i2c devices, so that would make no difference from a user-space > > > perspective. And this would avoid code redundancy. > > > > > > If we don't want to do that, then let's offer drivers a nicer function > > > to retrieve the spi device name, e.g. spi_dev_name(). That way, the > > > internal implementation can change in the future without having to > > > update all drivers. > > > > Adding a "name" attribute would be great, but it is not used outside hwmon, > > and I am not sure if we want to impose that on other users. > > That's correct, but OTOH we do exactly that for i2c devices and I am > not aware of anybody complaining about that so far. > > > I am obviously fine with spi_dev_name(). > > > > Having said that, modalias is used heavily in SPI drivers, so my patches > > don't really do anything special. > > OK, good to know. This certainly speaks in favor of at least > spi_dev_name(), but we can take your patches as is for the time being, > pending Grant's decision. > That is what I would suggest we should do. Also copying Mark Brown; he is handling much of the SPI changes nowadays. Guenter _______________________________________________ lm-sensors mailing list lm-sensors@xxxxxxxxxxxxxx http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
