On 10/13/2015 09:09 AM, Marc Andre wrote: > On 12.10.2015 16:09, Lars-Peter Clausen wrote: >> On 10/12/2015 02:53 PM, Marc Andre wrote: >>> On 12.10.2015 12:26, Lars-Peter Clausen wrote: >>>> On 10/12/2015 10:52 AM, Marc Andre wrote: >>>>> This driver adds support for the Linear LTC26x6, LTC26x7 and LTC26x9 I2C >>>>> DAC chips. >>>> Those look like they are very much register map compatible to what is >>>> supported by the AD5064 driver. It makes sense to add support for them >>>> there >>>> instead of having a separate driver. >>> I see that the interfaces are similar. Not all features of the ADxxxx are >>> supported by the LTC26XX. e.g. it doesn't support the different power down >>> modes. It also doesn't support the configuration register. >>> Other features available to AD which are not available to the LTC are LDAC >>> by software, RESET and CLEAR commands. Those commands are currently not in >>> use by the driver, but in the future, if those commands are used, a >>> separation would have to be done to avoid issues with the LTC. >>> >>> I first also thought that the address / command location is different as the >>> AD5064 sends 4 bytes and "AD5064_ADDR(x)" is "((x) << 20)". By further >>> reviewing I see that this only applies to SPI connected devices, while I2C >>> connected devices have 3 bytes as LTC26xx. >>> >>> I am quite open. I have the tendency to suggest a separate ltc26xx driver, >>> also because it would allow simpler identification of the driver and >>> separate evolution. (Someone would need to know that AD4064 is compatible to >>> LTC26xx) I also think large drivers containing many separations between >>> similar, but not equal devices are more risky to break on changes. No one >>> has all those devices available for test. But I am a new contributor to the >>> Kernel, thus I am interested to listen to the experts... :-) >> It's a subset, but it is a clean subset. The LTC26xx don't have any extra >> functionality not yet supported by the AD5064 driver. >> >> And I'm pretty convinced the chip was purposefully designed to be register >> map compatible, would be a shame not to make use of that. Means we only have >> to fix things in one driver if we find a bug, or e.g. if we want the feature >> of being able to update all channels at the same time. >> > I started testing with the ad5064 driver, but the driver seams to be quite > broken for i2c. Before I fix all of it, I just want to confirm that I didn't > get it wrong: > - ad5064_write_raw() expects the ad5064_write() to return 0 if success. But > ad5064_i2c_write() directly returns the return value of i2c_master_send() > which is the number of bytes sent if successful. I need to add a wrap to > convert the return value of i2c_master_send() to 0 if the correct number of > bytes is written. > - #define AD5064_CHANNEL() sets the shift as follows: .shift = 20 - bits > ad5064_write then applies this shift before sending it to the i2c/spi > command. This assumes that the number of data bits is 20 for all DAC chips. > As written yesterday this seams to be only true for the SPI type chips. The > I2C type chips use 2 bytes (16 bits). > > Do you know if the ad5064 driver has been used for any i2c type device yet? > Was it tested? Yes, it was tested and those two bugs you mentioned were also found, but it looks like somebody forgot to send those fixes upstream. These are the patches, will send them shortly: https://github.com/analogdevicesinc/linux/commit/61281b058cb69102ed1476028eede46e79fb8449 https://github.com/analogdevicesinc/linux/commit/cc424192a5a82e8f5969057beb6d3c25b8c0eced -- To unsubscribe from this list: send the line "unsubscribe linux-iio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
