From mboxrd@z Thu Jan  1 00:00:00 1970
From: swise@opengridcomputing.com (Steve Wise)
Date: Tue, 30 Aug 2016 10:57:53 -0500
Subject: [PATCH WIP/RFC v2 6/6] nvme-rdma: use ib_client API to detect
 device removal
In-Reply-To: <d0bf0b9b-8f81-4d6b-4534-1fccb146e328@grimberg.me>
References: <cover.1472506505.git.swise@opengridcomputing.com>
 <a78624bcccd2262cf54d8db4aa7da753bd6b499a.1472506505.git.swise@opengridcomputing.com>
 <b0f1740f-bcf7-d508-7a64-223202d6c063@grimberg.me>
 <00e601d202d6$00fc0920$02f41b60$@opengridcomputing.com>
 <d0bf0b9b-8f81-4d6b-4534-1fccb146e328@grimberg.me>
Message-ID: <00eb01d202d7$446f8870$cd4e9950$@opengridcomputing.com>

> 
> >> Will this code work?
> >>
> >> 	mutex_lock(&nvme_rdma_ctrl_mutex);
> >> 	list_for_each_entry(ctrl, &nvme_rdma_ctrl_list, list) {
> >> 		if (ctrl->device->dev != ib_device)
> >> 			continue;
> >>
> >> 		__nvme_rdma_del_ctrl(ctrl);
> >> 	}
> >> 	mutex_unlock(&nvme_rdma_ctrl_mutex);
> >>
> >> 	flush_workqueue(nvme_rdma_wq);
> >
> > I'll try this.  It is cleaner/easier to understand.
> >
> > I'm also moving the controller list to per-device as Christoph recommends.
> 
> I'm still not sure if it's a real must. This will only save us the
> condition (as we'll get the device in the client data)

True, and your logic avoids the unlock/relock.