[PATCH 2/3] nvme: multipath: only update ctrl->nr_active when using queue-depth iopolicy

[Uday Shankar]

On Tue, Nov 07, 2023 at 04:23:30PM -0500, Ewan D. Milne wrote:
> The atomic updates of ctrl->nr_active are unnecessary when using
> numa or round-robin iopolicy, so avoid that cost on a per-request basis.
> Clear nr_active when changing iopolicy and do not decrement below zero.
> (This handles changing the iopolicy while requests are in flight.)

Instead of trying to handle a changing iopolicy while requests are in
flight, can we quiesce I/O when we change the iopolicy? That should let
us simplify/speed up the logic in the I/O path a bit
(atomic_dec_if_positive seems to hide a cmpxchg loop on most
architectures, which can be slower than an atomic_dec).

> 
> Signed-off-by: Ewan D. Milne <emilne at redhat.com>
> ---
>  drivers/nvme/host/core.c      |  2 +-
>  drivers/nvme/host/multipath.c | 21 ++++++++++++++++++---
>  drivers/nvme/host/nvme.h      |  2 ++
>  3 files changed, 21 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
> index 75a1b58a7a43..9bc19755be77 100644
> --- a/drivers/nvme/host/core.c
> +++ b/drivers/nvme/host/core.c
> @@ -110,7 +110,7 @@ struct workqueue_struct *nvme_delete_wq;
>  EXPORT_SYMBOL_GPL(nvme_delete_wq);
>  
>  static LIST_HEAD(nvme_subsystems);
> -static DEFINE_MUTEX(nvme_subsystems_lock);
> +DEFINE_MUTEX(nvme_subsystems_lock);
>  
>  static DEFINE_IDA(nvme_instance_ida);
>  static dev_t nvme_ctrl_base_chr_devt;
> diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
> index 4c2690cddef3..e184e7c377bc 100644
> --- a/drivers/nvme/host/multipath.c
> +++ b/drivers/nvme/host/multipath.c
> @@ -133,7 +133,8 @@ void nvme_mpath_start_request(struct request *rq)
>  	if (!blk_queue_io_stat(disk->queue) || blk_rq_is_passthrough(rq))
>  		return;
>  
> -	atomic_inc(&ns->ctrl->nr_active);
> +	if (READ_ONCE(ns->head->subsys->iopolicy) == NVME_IOPOLICY_QD)
> +		atomic_inc(&ns->ctrl->nr_active);
>  	nvme_req(rq)->flags |= NVME_MPATH_IO_STATS;
>  	nvme_req(rq)->start_time = bdev_start_io_acct(disk->part0, req_op(rq),
>  						      jiffies);
> @@ -147,7 +148,8 @@ void nvme_mpath_end_request(struct request *rq)
>  	if (!(nvme_req(rq)->flags & NVME_MPATH_IO_STATS))
>  		return;
>  
> -	atomic_dec(&ns->ctrl->nr_active);
> +	if (READ_ONCE(ns->head->subsys->iopolicy) == NVME_IOPOLICY_QD)
> +		atomic_dec_if_positive(&ns->ctrl->nr_active);
>  	bdev_end_io_acct(ns->head->disk->part0, req_op(rq),
>  			 blk_rq_bytes(rq) >> SECTOR_SHIFT,
>  			 nvme_req(rq)->start_time);
> @@ -848,6 +850,19 @@ static ssize_t nvme_subsys_iopolicy_show(struct device *dev,
>  			  nvme_iopolicy_names[READ_ONCE(subsys->iopolicy)]);
>  }
>  
> +void nvme_subsys_iopolicy_update(struct nvme_subsystem *subsys, int iopolicy)
> +{
> +	struct nvme_ctrl *ctrl;
> +
> +	WRITE_ONCE(subsys->iopolicy, iopolicy);
> +
> +	mutex_lock(&nvme_subsystems_lock);
> +	list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry) {
> +		atomic_set(&ctrl->nr_active, 0);
> +	}
> +	mutex_unlock(&nvme_subsystems_lock);
> +}
> +
>  static ssize_t nvme_subsys_iopolicy_store(struct device *dev,
>  		struct device_attribute *attr, const char *buf, size_t count)
>  {
> @@ -857,7 +872,7 @@ static ssize_t nvme_subsys_iopolicy_store(struct device *dev,
>  
>  	for (i = 0; i < ARRAY_SIZE(nvme_iopolicy_names); i++) {
>  		if (sysfs_streq(buf, nvme_iopolicy_names[i])) {
> -			WRITE_ONCE(subsys->iopolicy, i);
> +			nvme_subsys_iopolicy_update(subsys, i);
>  			return count;
>  		}
>  	}
> diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
> index f0f3fd8b4197..c4469bc38d89 100644
> --- a/drivers/nvme/host/nvme.h
> +++ b/drivers/nvme/host/nvme.h
> @@ -49,6 +49,8 @@ extern struct workqueue_struct *nvme_wq;
>  extern struct workqueue_struct *nvme_reset_wq;
>  extern struct workqueue_struct *nvme_delete_wq;
>  
> +extern struct mutex nvme_subsystems_lock;
> +
>  /*
>   * List of workarounds for devices that required behavior not specified in
>   * the standard.
> -- 
> 2.20.1
> 
>