>Correction, i2cset can't be used to reproduce the problem because it uses >i2c-dev.h, not the functions in i2c-core.c. How does it matter? i2c-dev is just the access point, it has no real (I2C/SMBus-related) code, has it? So I guess that the commands will end up in i2c-core, whether they come from a chip driver or i2c-dev. After all, i2c-dev is nothing but a universal, user-space controllable chip driver. >However using admxxxx drivers on a driver with i2c-algo-bit should show the >problem if it exists. I did that several times and can confirm it always worked fine. So, whatever the exact problem is, it's more complex than just "emulated smbus writes don't work". Jean Delvare
