On 10/29/18 8:25 AM, Johannes Thumshirn wrote: > On 29/10/18 15:23, Jens Axboe wrote: >> On 10/29/18 5:00 AM, Johannes Thumshirn wrote: >>> Hi Jens, >>> >>> On 25/10/18 23:10, Jens Axboe wrote: >>> [...] >>>> @@ -1487,8 +1485,6 @@ int blkcg_activate_policy(struct request_queue *q, >>>> out_bypass_end: >>>> if (q->mq_ops) >>>> blk_mq_unfreeze_queue(q); >>>> - else >>>> - blk_queue_bypass_end(q); >>> >>>> >>>> >>> >>> Now that we only have mq, do we still need all these checks for q->mq_ops? >> >> We need it for the cases where we can be passing in a stacked driver >> or an mq driver, or for cases where we explicitly register mq attributes, >> and that sort of thing. >> > > Ah ok. Guess we'll have to live with it then. At this point the check should be equivalent to queue_is_rq_based(), fwiw. -- Jens Axboe
