From mboxrd@z Thu Jan  1 00:00:00 1970
From: John Fastabend <john.r.fastabend@intel.com>
Subject: Re: [net-next  1/2] vxlan: Notify drivers for listening UDP port
 changes
Date: Wed, 28 Aug 2013 12:34:09 -0700
Message-ID: <521E50B1.4010102@intel.com>
References: <1377665218-6760-1-git-send-email-jeffrey.t.kirsher@intel.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: davem@davemloft.net,
	Joseph Gasparakis <joseph.gasparakis@intel.com>,
	netdev@vger.kernel.org, gospo@redhat.com, sassmann@redhat.com,
	Stephen Hemminger <stephen@networkplumber.org>
To: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Return-path: <netdev-owner@vger.kernel.org>
Received: from mga01.intel.com ([192.55.52.88]:53937 "EHLO mga01.intel.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1754418Ab3H1Te2 (ORCPT <rfc822;netdev@vger.kernel.org>);
	Wed, 28 Aug 2013 15:34:28 -0400
In-Reply-To: <1377665218-6760-1-git-send-email-jeffrey.t.kirsher@intel.com>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

On 8/27/2013 9:46 PM, Jeff Kirsher wrote:
> From: Joseph Gasparakis <joseph.gasparakis@intel.com>
>
> This patch adds two more ndo ops: ndo_add_rx_vxlan_port() and
> ndo_del_rx_vxlan_port().
>
> Drivers can get notifications through the above functions about changes
> of the UDP listening port of VXLAN. Also, when physical ports come up,
> now they can call vxlan_get_rx_port() in order to obtain the port number(s)
> of the existing VXLAN interface in case they already up before them.
>
> This information about the listening UDP port would be used for VXLAN
> related offloads.

[...]

>   /* Add new entry to forwarding table -- assumes lock held */
>   static int vxlan_fdb_create(struct vxlan_dev *vxlan,
>   			    const u8 *mac, __be32 ip,
> @@ -797,13 +823,15 @@ static void vxlan_sock_hold(struct vxlan_sock *vs)
>
>   void vxlan_sock_release(struct vxlan_sock *vs)
>   {
> -	struct vxlan_net *vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
> +	struct net *net = sock_net(vs->sock->sk);
> +	struct vxlan_net *vn = net_generic(net, vxlan_net_id);
>
>   	if (!atomic_dec_and_test(&vs->refcnt))
>   		return;
>
>   	spin_lock(&vn->sock_lock);
>   	hlist_del_rcu(&vs->hlist);
> +	vxlan_notify_del_rx_port(net, inet_sk(vs->sock->sk)->inet_sport);
>   	spin_unlock(&vn->sock_lock);

Both the del and add port are protected by sock_lock serializing the
operations.

>
>   	queue_work(vxlan_wq, &vs->del_work);
> @@ -1543,6 +1571,28 @@ static struct device_type vxlan_type = {
>   	.name = "vxlan",
>   };
>
> +/* Calls the ndo_add_vxlan_port of the caller in order to
> + * supply the listening VXLAN udp ports.
> + */
> +void vxlan_get_rx_port(struct net_device *dev)
> +{
> +	struct vxlan_sock *vs;
> +	struct net *net = dev_net(dev);
> +	u16 port;
> +	int i;
> +
> +	if (!dev || !dev->netdev_ops || !dev->netdev_ops->ndo_add_vxlan_port)
> +		return;
> +
> +	for (i = 0; i < PORT_HASH_SIZE; ++i) {
> +		hlist_for_each_entry_rcu(vs, vs_head(net, i), hlist) {
> +			port = htons(inet_sk(vs->sock->sk)->inet_sport);
> +			dev->netdev_ops->ndo_add_vxlan_port(dev, port);

However this list walk occurs without the sock_lock. Looks like you
could delete a port and then subsequently add it here if you had really
"good" timing.

Then it would be deleted from the vxlan list but pushed into hardware.
Probably not a terrible scenario but it would waste hardware resources.

I suspect you want to lock this list traversal with the sock_lock as
well.

> +		}
> +	}
> +}
> +EXPORT_SYMBOL_GPL(vxlan_get_rx_port);