Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
In-Reply-To: <1534317915-5041-2-git-send-email-jianchao.w.wang@oracle.com>
References: <1534317915-5041-1-git-send-email-jianchao.w.wang@oracle.com> <1534317915-5041-2-git-send-email-jianchao.w.wang@oracle.com>
From:   Ming Lei <tom.leiming@gmail.com>
Date:   Wed, 15 Aug 2018 19:32:54 +0800
Message-ID: <CACVXFVO+NoU2YcuhchYOzAbxu9CfY7VCE5moRrH47kWmX_JD7A@mail.gmail.com>
Subject: Re: [PATCH 1/2] blk-mq: init hctx sched after update cpu &
 nr_hw_queues mapping
To:     Jianchao Wang <jianchao.w.wang@oracle.com>
Cc:     Jens Axboe <axboe@kernel.dk>,
        Bart Van Assche <bart.vanassche@wdc.com>,
        Keith Busch <keith.busch@linux.intel.com>,
        linux-block <linux-block@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Aug 15, 2018 at 3:25 PM, Jianchao Wang
<jianchao.w.wang@oracle.com> wrote:
> Kyber depends on the mapping between cpu and nr_hw_queues. When
> update nr_hw_queues, elevator_type->ops.mq.init_hctx will be
> invoked before the mapping is adapted correctly, this would cause
> terrible result. A simply way to fix this is switch the io scheduler
> to none before update the nr_hw_queues, and then get it back after
> update nr_hw_queues. To achieve this, we add a new member elv_type
> in request_queue to save the original elevator and adapt and export
> elevator_switch_mq.
>
> Signed-off-by: Jianchao Wang <jianchao.w.wang@oracle.com>
> ---
>  block/blk-mq.c         | 37 +++++++++++++++++++++++++++++--------
>  block/blk.h            |  2 ++
>  block/elevator.c       | 20 ++++++++++++--------
>  include/linux/blkdev.h |  3 +++
>  4 files changed, 46 insertions(+), 16 deletions(-)
>
> diff --git a/block/blk-mq.c b/block/blk-mq.c
> index 5efd789..89904cc 100644
> --- a/block/blk-mq.c
> +++ b/block/blk-mq.c
> @@ -112,6 +112,7 @@ void blk_mq_in_flight(struct request_queue *q, struct hd_struct *part,
>         struct mq_inflight mi = { .part = part, .inflight = inflight, };
>
>         inflight[0] = inflight[1] = 0;
> +

Not necessary to do that.

>         blk_mq_queue_tag_busy_iter(q, blk_mq_check_inflight, &mi);
>  }
>
> @@ -2147,8 +2148,6 @@ static void blk_mq_exit_hctx(struct request_queue *q,
>         if (set->ops->exit_request)
>                 set->ops->exit_request(set, hctx->fq->flush_rq, hctx_idx);
>
> -       blk_mq_sched_exit_hctx(q, hctx, hctx_idx);
> -
>         if (set->ops->exit_hctx)
>                 set->ops->exit_hctx(hctx, hctx_idx);
>
> @@ -2216,12 +2215,9 @@ static int blk_mq_init_hctx(struct request_queue *q,
>             set->ops->init_hctx(hctx, set->driver_data, hctx_idx))
>                 goto free_bitmap;
>
> -       if (blk_mq_sched_init_hctx(q, hctx, hctx_idx))
> -               goto exit_hctx;
> -
>         hctx->fq = blk_alloc_flush_queue(q, hctx->numa_node, set->cmd_size);
>         if (!hctx->fq)
> -               goto sched_exit_hctx;
> +               goto exit_hctx;
>
>         if (blk_mq_init_request(set, hctx->fq->flush_rq, hctx_idx, node))
>                 goto free_fq;
> @@ -2235,8 +2231,6 @@ static int blk_mq_init_hctx(struct request_queue *q,
>
>   free_fq:
>         kfree(hctx->fq);
> - sched_exit_hctx:
> -       blk_mq_sched_exit_hctx(q, hctx, hctx_idx);

Seems both blk_mq_sched_init_hctx() and blk_mq_sched_exit_hctx() may be
removed now.

>   exit_hctx:
>         if (set->ops->exit_hctx)
>                 set->ops->exit_hctx(hctx, hctx_idx);
> @@ -2913,6 +2907,25 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
>         list_for_each_entry(q, &set->tag_list, tag_set_list)
>                 blk_mq_freeze_queue(q);
>
> +       /*
> +        * switch io scheduler to NULL to clean up the data in it.
> +        * will get it back after update mapping between cpu and hw queues.
> +        */
> +       list_for_each_entry(q, &set->tag_list, tag_set_list) {
> +               if (!q->elevator) {
> +                       q->elv_type = NULL;
> +                       continue;
> +               }
> +               q->elv_type = q->elevator->type;
> +               mutex_lock(&q->sysfs_lock);
> +               /*
> +                * elevator_release will put it.
> +                */
> +               __module_get(q->elv_type->elevator_owner);
> +               elevator_switch_mq(q, NULL);
> +               mutex_unlock(&q->sysfs_lock);
> +       }
> +
>         set->nr_hw_queues = nr_hw_queues;
>         blk_mq_update_queue_map(set);
>         list_for_each_entry(q, &set->tag_list, tag_set_list) {
> @@ -2920,6 +2933,14 @@ static void __blk_mq_update_nr_hw_queues(struct blk_mq_tag_set *set,
>                 blk_mq_queue_reinit(q);
>         }
>
> +       list_for_each_entry(q, &set->tag_list, tag_set_list) {
> +               if (!q->elv_type)
> +                       continue;
> +
> +               mutex_lock(&q->sysfs_lock);
> +               elevator_switch_mq(q, q->elv_type);
> +               mutex_unlock(&q->sysfs_lock);
> +       }

BFQ defines .init_hctx() too, so seems this generic approach is correct way for
this issue.

thanks,
Ming Lei