Hi Adreas, > while writing a driver for a cardbus card which is supposed to use the > bit-banging algorithm I noticed that communication with the I2C slave > (Philips TDA10046) would fail without this patch. It forces SDA to > high for every bit in i2c_inb() instead of once per byte. Can this > patch go into the mainline kernel or will this break other drivers? I > am using Kernel version 2.6.10. There is no reason why this would be necessary. Once SDA is set high on the master's side, the client will be controlling it for the rest of the byte (transferred from client to master). Setting SDA high again on the master's side for each bit shouldn't have any effet, let alone the substantial slowdown. If it changes something for you, then either your implementation of setscl or getsda is broken and does actually change SDA as a side effect, or what really helps is the additional delay, not setting SDA high per se. In the former case, check your code. If you can't figure out what's wrong, post it here and I'll take at look. What's your bus master BTW? In the latter case, you might simply need to lower the bus speed (increase udelay). Thanks, -- Jean Delvare
