[PATCH v3 11/11] nvme: Fix a race condition

[Christoph Hellwig]

Hi Bart,

this looks great!

Reviewed-by: Christoph Hellwig <hch at lst.de>

Some minor nitpicks below:

>  void nvme_requeue_req(struct request *req)
>  {
> +	blk_mq_requeue_request(req, true);
>  }
>  EXPORT_SYMBOL_GPL(nvme_requeue_req);

Please just remove the nvme_requeue_req wrapper.

>  
> @@ -2074,11 +2068,14 @@ EXPORT_SYMBOL_GPL(nvme_kill_queues);
>  void nvme_stop_queues(struct nvme_ctrl *ctrl)
>  {
>  	struct nvme_ns *ns;
> +	struct request_queue *q;
>  
>  	mutex_lock(&ctrl->namespaces_mutex);
>  	list_for_each_entry(ns, &ctrl->namespaces, list) {
> +		q = ns->queue;
> +		blk_mq_cancel_requeue_work(q);
> +		blk_mq_stop_hw_queues(q);
> +		blk_mq_quiesce_queue(q);
>  	}

I'd keep the q declaration in the minimal scope, e.g.

	list_for_each_entry(ns, &ctrl->namespaces, list) {
		struct request_queue *q = ns->queue;

		blk_mq_cancel_requeue_work(q);
		blk_mq_stop_hw_queues(q);
		blk_mq_quiesce_queue(q);
	}