On Wed, Sep 04, 2019 at 05:42:42PM +0900, Damien Le Moal wrote: > If the default elevator chosen is mq-deadline, elevator_init_mq() may > return an error if mq-deadline initialization fails, leading to > blk_mq_init_allocated_queue() returning an error, which in turn will > cause the block device initialization to fail and the device not being > exposed. > > Instead of taking such extreme measure, handle mq-deadline > initialization failures in the same manner as when mq-deadline is not > available (no module to load), that is, default to the "none" scheduler. > With this change, elevator_init_mq() return type can be changed to void. > > Signed-off-by: Damien Le Moal <damien.lemoal@xxxxxxx> > Reviewed-by: Johannes Thumshirn <jthumshirn@xxxxxxx> > Reviewed-by: Christoph Hellwig <hch@xxxxxx> > --- > block/blk-mq.c | 8 +------- > block/blk.h | 2 +- > block/elevator.c | 23 ++++++++++++----------- > 3 files changed, 14 insertions(+), 19 deletions(-) > > diff --git a/block/blk-mq.c b/block/blk-mq.c > index 13923630e00a..ee4caf0c0807 100644 > --- a/block/blk-mq.c > +++ b/block/blk-mq.c > @@ -2842,8 +2842,6 @@ static unsigned int nr_hw_queues(struct blk_mq_tag_set *set) > struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, > struct request_queue *q) > { > - int ret = -ENOMEM; > - > /* mark the queue as mq asap */ > q->mq_ops = set->ops; > > @@ -2904,14 +2902,10 @@ struct request_queue *blk_mq_init_allocated_queue(struct blk_mq_tag_set *set, > blk_mq_add_queue_tag_set(set, q); > blk_mq_map_swqueue(q); > > - ret = elevator_init_mq(q); > - if (ret) > - goto err_tag_set; > + elevator_init_mq(q); > > return q; > > -err_tag_set: > - blk_mq_del_queue_tag_set(q); > err_hctxs: > kfree(q->queue_hw_ctx); > q->nr_hw_queues = 0; > diff --git a/block/blk.h b/block/blk.h > index e4619fc5c99a..ed347f7a97b1 100644 > --- a/block/blk.h > +++ b/block/blk.h > @@ -184,7 +184,7 @@ void blk_account_io_done(struct request *req, u64 now); > > void blk_insert_flush(struct request *rq); > > -int elevator_init_mq(struct request_queue *q); > +void elevator_init_mq(struct request_queue *q); > int elevator_switch_mq(struct request_queue *q, > struct elevator_type *new_e); > void __elevator_exit(struct request_queue *, struct elevator_queue *); > diff --git a/block/elevator.c b/block/elevator.c > index 4721834815bb..2944c129760c 100644 > --- a/block/elevator.c > +++ b/block/elevator.c > @@ -628,34 +628,35 @@ static inline bool elv_support_iosched(struct request_queue *q) > > /* > * For blk-mq devices supporting IO scheduling, we default to using mq-deadline, > - * if available, for single queue devices. If deadline isn't available OR we > - * have multiple queues, default to "none". > + * if available, for single queue devices. If deadline isn't available OR > + * deadline initialization fails OR we have multiple queues, default to "none". > */ > -int elevator_init_mq(struct request_queue *q) > +void elevator_init_mq(struct request_queue *q) > { > struct elevator_type *e; > - int err = 0; > + int err; > > if (!elv_support_iosched(q)) > - return 0; > + return; > > if (q->nr_hw_queues != 1) > - return 0; > + return; > > WARN_ON_ONCE(test_bit(QUEUE_FLAG_REGISTERED, &q->queue_flags)); > > if (unlikely(q->elevator)) > - goto out; > + return; > > e = elevator_get(q, "mq-deadline", false); > if (!e) > - goto out; > + return; > > err = blk_mq_init_sched(q, e); > - if (err) > + if (err) { > + pr_warn("\"%s\" elevator initialization failed, " > + "falling back to \"none\"\n", e->elevator_name); > elevator_put(e); > -out: > - return err; > + } > } Looks fine: Reviewed-by: Ming Lei <ming.lei@xxxxxxxxxx> BTW, blk_mq_init_sched()'s failure patch should have restored q->nr_request. And that could be done in another standalone patch. -- Ming