Update: I2C bus driver TIMEDOUT because of PM autosuspend

I must update my report. It is not PM autosuspend (void calls do nothing, no miracle here). This morning I reproduced the I2C timeout error also with XIIC_PM_TIMEOUT=10000 delay. For several days some other change was misleading me to PM autosuspend conclusion. What looks to me now is that just slight change in driver code (e.g. add or remove dev_info() call) changes the execution timing pattern of driver and this then determine if the problem probability window is hit or not. Probability to hit the problem is very low. The driver timeout error and its result to I2C bus blocked is still a problem. I would still like to let the maintainer aware of this problem.