Hi Ming, I really struggly to review this series. A lot of the code and especially code organization is really strange. For example a signiciant chunk of the bio based poll logic is added to blk-mq.c where it obviously does not fit in, and a lot of the patches seems to both change blk-mq logic and add new bio based logic. Also the bio poll logic is split over various patches in a way that is absolutel not obvious to me. Can you try to restructure this a bit: - move everything that cleans up or improves existing logic to the beginning of the series and into separate patches (where that isn't already the case) - keep the actual bio polling logic together where the functionality isn't otherwise useful - maybe add a new block/bio-poll.c file to keep all the code togeher?
