> Il giorno 16 giu 2017, alle ore 18:15, Christoph Hellwig <hch@xxxxxx> ha scritto:
>
> This patch makes sure we always allocate requests in the core blk-mq
> code and use a common prepare_request method to initialize them for
> both mq I/O schedulers. For Kyber and additional limit_depth method
> is added that is called before allocating the request.
>
> Also because none of the intializations can really fail the new method
> does not return an error - instead the bfq finish method is hardened
> to deal with the no-IOC case.
>
> Last but not least this removes the abuse of RQF_QUEUE by the blk-mq
> scheduling code as RQF_ELFPRIV is all that is needed now.
>
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
> block/bfq-iosched.c | 19 ++++++++++++-------
> block/blk-mq.c | 22 ++++++----------------
> block/kyber-iosched.c | 23 +++++++++++------------
> include/linux/elevator.h | 4 ++--
> 4 files changed, 31 insertions(+), 37 deletions(-)
>
> diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
> index f037b005faa1..60d32700f104 100644
> --- a/block/bfq-iosched.c
> +++ b/block/bfq-iosched.c
> @@ -4292,8 +4292,14 @@ static void bfq_put_rq_priv_body(struct bfq_queue *bfqq)
>
> static void bfq_finish_request(struct request *rq)
> {
> - struct bfq_queue *bfqq = RQ_BFQQ(rq);
> - struct bfq_data *bfqd = bfqq->bfqd;
> + struct bfq_queue *bfqq;
> + struct bfq_data *bfqd;
> +
> + if (!rq->elv.icq)
> + return;
> +
If this is a rq dispatched from a bfqq (or even a request still in the
scheduler), then just exiting here will break bfq state seriously.
However, I guess that this case can never occur.
If this is a rq dispatched from the bfqd->dispatch list, then exiting
here should only unbalance the bfqd->rq_in_driver counter.
I guess that in both cases there wouldn't be problems for the missing
reset of the content of rq->elv.priv (if the function terminates
here).
If I can be of further help in some way, I'm willing to help.
Thanks,
Paolo
> + bfqq = RQ_BFQQ(rq);
> + bfqd = bfqq->bfqd;
>
> if (rq->rq_flags & RQF_STARTED)
> bfqg_stats_update_completion(bfqq_group(bfqq),
> @@ -4394,9 +4400,9 @@ static struct bfq_queue *bfq_get_bfqq_handle_split(struct bfq_data *bfqd,
> /*
> * Allocate bfq data structures associated with this request.
> */
> -static int bfq_get_rq_private(struct request_queue *q, struct request *rq,
> - struct bio *bio)
> +static void bfq_prepare_request(struct request *rq, struct bio *bio)
> {
> + struct request_queue *q = rq->q;
> struct bfq_data *bfqd = q->elevator->elevator_data;
> struct bfq_io_cq *bic;
> const int is_sync = rq_is_sync(rq);
> @@ -4405,7 +4411,7 @@ static int bfq_get_rq_private(struct request_queue *q, struct request *rq,
> bool split = false;
>
> if (!rq->elv.icq)
> - return 1;
> + return;
> bic = icq_to_bic(rq->elv.icq);
>
> spin_lock_irq(&bfqd->lock);
> @@ -4466,7 +4472,6 @@ static int bfq_get_rq_private(struct request_queue *q, struct request *rq,
> bfq_handle_burst(bfqd, bfqq);
>
> spin_unlock_irq(&bfqd->lock);
> - return 0;
> }
>
> static void bfq_idle_slice_timer_body(struct bfq_queue *bfqq)
> @@ -4945,7 +4950,7 @@ static struct elv_fs_entry bfq_attrs[] = {
>
> static struct elevator_type iosched_bfq_mq = {
> .ops.mq = {
> - .get_rq_priv = bfq_get_rq_private,
> + .prepare_request = bfq_prepare_request,
> .finish_request = bfq_finish_request,
> .exit_icq = bfq_exit_icq,
> .insert_requests = bfq_insert_requests,
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 2f380ab7a603..81d05c19d4b3 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -298,16 +298,11 @@ static struct request *blk_mq_get_request(struct request_queue *q,
> * Flush requests are special and go directly to the
> * dispatch list.
> */
> - if (!op_is_flush(op) && e->type->ops.mq.get_request) {
> - rq = e->type->ops.mq.get_request(q, op, data);
> - if (rq)
> - rq->rq_flags |= RQF_QUEUED;
> - goto allocated;
> - }
> + if (!op_is_flush(op) && e->type->ops.mq.limit_depth)
> + e->type->ops.mq.limit_depth(op, data);
> }
>
> rq = __blk_mq_alloc_request(data, op);
> -allocated:
> if (!rq) {
> blk_queue_exit(q);
> return NULL;
> @@ -315,17 +310,12 @@ static struct request *blk_mq_get_request(struct request_queue *q,
>
> if (!op_is_flush(op)) {
> rq->elv.icq = NULL;
> - if (e && e->type->ops.mq.get_rq_priv) {
> + if (e && e->type->ops.mq.prepare_request) {
> if (e->type->icq_cache && rq_ioc(bio))
> blk_mq_sched_assign_ioc(rq, bio);
>
> - if (e->type->ops.mq.get_rq_priv(q, rq, bio)) {
> - if (rq->elv.icq)
> - put_io_context(rq->elv.icq->ioc);
> - rq->elv.icq = NULL;
> - } else {
> - rq->rq_flags |= RQF_ELVPRIV;
> - }
> + e->type->ops.mq.prepare_request(rq, bio);
> + rq->rq_flags |= RQF_ELVPRIV;
> }
> }
> data->hctx->queued++;
> @@ -413,7 +403,7 @@ void blk_mq_free_request(struct request *rq)
> struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(q, ctx->cpu);
> const int sched_tag = rq->internal_tag;
>
> - if (rq->rq_flags & (RQF_ELVPRIV | RQF_QUEUED)) {
> + if (rq->rq_flags & RQF_ELVPRIV) {
> if (e && e->type->ops.mq.finish_request)
> e->type->ops.mq.finish_request(rq);
> if (rq->elv.icq) {
> diff --git a/block/kyber-iosched.c b/block/kyber-iosched.c
> index 2557b399f0a8..a9f6fd3fab8e 100644
> --- a/block/kyber-iosched.c
> +++ b/block/kyber-iosched.c
> @@ -426,24 +426,22 @@ static void rq_clear_domain_token(struct kyber_queue_data *kqd,
> }
> }
>
> -static struct request *kyber_get_request(struct request_queue *q,
> - unsigned int op,
> - struct blk_mq_alloc_data *data)
> +static void kyber_limit_depth(unsigned int op, struct blk_mq_alloc_data *data)
> {
> - struct kyber_queue_data *kqd = q->elevator->elevator_data;
> - struct request *rq;
> -
> /*
> * We use the scheduler tags as per-hardware queue queueing tokens.
> * Async requests can be limited at this stage.
> */
> - if (!op_is_sync(op))
> + if (!op_is_sync(op)) {
> + struct kyber_queue_data *kqd = data->q->elevator->elevator_data;
> +
> data->shallow_depth = kqd->async_depth;
> + }
> +}
>
> - rq = __blk_mq_alloc_request(data, op);
> - if (rq)
> - rq_set_domain_token(rq, -1);
> - return rq;
> +static void kyber_prepare_request(struct request *rq, struct bio *bio)
> +{
> + rq_set_domain_token(rq, -1);
> }
>
> static void kyber_finish_request(struct request *rq)
> @@ -813,7 +811,8 @@ static struct elevator_type kyber_sched = {
> .exit_sched = kyber_exit_sched,
> .init_hctx = kyber_init_hctx,
> .exit_hctx = kyber_exit_hctx,
> - .get_request = kyber_get_request,
> + .limit_depth = kyber_limit_depth,
> + .prepare_request = kyber_prepare_request,
> .finish_request = kyber_finish_request,
> .completed_request = kyber_completed_request,
> .dispatch_request = kyber_dispatch_request,
> diff --git a/include/linux/elevator.h b/include/linux/elevator.h
> index 4acea351d43f..5bc8f8682a3e 100644
> --- a/include/linux/elevator.h
> +++ b/include/linux/elevator.h
> @@ -104,7 +104,8 @@ struct elevator_mq_ops {
> int (*request_merge)(struct request_queue *q, struct request **, struct bio *);
> void (*request_merged)(struct request_queue *, struct request *, enum elv_merge);
> void (*requests_merged)(struct request_queue *, struct request *, struct request *);
> - struct request *(*get_request)(struct request_queue *, unsigned int, struct blk_mq_alloc_data *);
> + void (*limit_depth)(unsigned int, struct blk_mq_alloc_data *);
> + void (*prepare_request)(struct request *, struct bio *bio);
> void (*finish_request)(struct request *);
> void (*insert_requests)(struct blk_mq_hw_ctx *, struct list_head *, bool);
> struct request *(*dispatch_request)(struct blk_mq_hw_ctx *);
> @@ -114,7 +115,6 @@ struct elevator_mq_ops {
> void (*requeue_request)(struct request *);
> struct request *(*former_request)(struct request_queue *, struct request *);
> struct request *(*next_request)(struct request_queue *, struct request *);
> - int (*get_rq_priv)(struct request_queue *, struct request *, struct bio *);
> void (*init_icq)(struct io_cq *);
> void (*exit_icq)(struct io_cq *);
> };
> --
> 2.11.0
>
