target crash / host hang with nvme-all.3 branch of nvme-fabrics

[Sagi Grimberg]

> I think nvmet_rdma_delete_ctrl is getting the exlusion vs other calls
> or __nvmet_rdma_queue_disconnect wrong as we rely on a queue that
> is undergoing deletion to not be on any list.

How do we rely on that? __nvmet_rdma_queue_disconnect callers are
responsible for queue_list deletion and queue the release. I don't
see where are we getting it wrong.

  Additionally it also
> check the cntlid instead of the pointer, which would be harmful if
> multiple subsystems have the same cntlid.

That's true, we need to compare pointers...

>
> Does the following patch help?
>
> diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
> index b1c6e5b..9ae65a7 100644
> --- a/drivers/nvme/target/rdma.c
> +++ b/drivers/nvme/target/rdma.c
> @@ -1293,19 +1293,21 @@ static int nvmet_rdma_cm_handler(struct rdma_cm_id *cm_id,
>
>   static void nvmet_rdma_delete_ctrl(struct nvmet_ctrl *ctrl)
>   {
> -	struct nvmet_rdma_queue *queue, *next;
> -	static LIST_HEAD(del_list);
> +	struct nvmet_rdma_queue *queue, *found = NULL;
>
>   	mutex_lock(&nvmet_rdma_queue_mutex);
> -	list_for_each_entry_safe(queue, next,
> -			&nvmet_rdma_queue_list, queue_list) {
> -		if (queue->nvme_sq.ctrl->cntlid == ctrl->cntlid)
> -			list_move_tail(&queue->queue_list, &del_list);
> +	list_for_each_entry(queue, &nvmet_rdma_queue_list, queue_list) {
> +		if (queue->nvme_sq.ctrl == ctrl) {
> +			list_del_init(&queue->queue_list);
> +			found = queue;
> +			break;
> +		}
>   	}
> +
>   	mutex_unlock(&nvmet_rdma_queue_mutex);
>
> -	list_for_each_entry_safe(queue, next, &del_list, queue_list)
> -		nvmet_rdma_queue_disconnect(queue);
> +	if (found)
> +		__nvmet_rdma_queue_disconnect(queue);
>   }
>
>   static int nvmet_rdma_add_port(struct nvmet_port *port)
>

Umm, this looks wrong to me. delete_controller should delete _all_
the ctrl queues (which will usually involve more than 1), what about
all the other queues? what am I missing?