Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Jonas Bonn <jonas@norrbonn.se>
To:     nicolas.dichtel@6wind.com, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Cc:     davem@davemloft.net, Jonas Bonn <jonas@norrbonn.se>
Subject: [PATCH v3 3/6] rtnetlink: allow RTM_NEWLINK to act upon interfaces in arbitrary namespaces
Date:   Thu,  7 Nov 2019 14:27:52 +0100
Message-Id: <20191107132755.8517-4-jonas@norrbonn.se>
In-Reply-To: <20191107132755.8517-1-jonas@norrbonn.se>
References: <20191107132755.8517-1-jonas@norrbonn.se>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

RTM_NEWLINK can be used mostly interchangeably with RTM_SETLINK for
modifying device configuration.  As such, this method requires the same
logic as RTM_SETLINK for finding the device to act on.

With this patch, the IFLA_TARGET_NETNSID selects the namespace in which
to search for the interface to act upon.  This allows, for example, to
set the namespace of an interface outside the current namespace by
selecting it with the (IFLA_TARGET_NETNSID,ifi->ifi_index) pair and
specifying the namespace with one of IFLA_NET_NS_[PID|FD].

Since rtnl_newlink branches off into do_setlink, we need to provide the
same backwards compatibility check as we do for RTM_SETLINK:  if the
device is not found in the namespace given by IFLA_TARGET_NETNSID then
we search for it in the current namespace.  If found there, it's
namespace will be changed, as before.

Signed-off-by: Jonas Bonn <jonas@norrbonn.se>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
---
 net/core/rtnetlink.c | 29 +++++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index a21e7d47135b..bcfabda3fc73 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -3021,6 +3021,7 @@ static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	const struct rtnl_link_ops *m_ops = NULL;
 	struct net_device *master_dev = NULL;
 	struct net *net = sock_net(skb->sk);
+	struct net *tgt_net = NULL;
 	const struct rtnl_link_ops *ops;
 	struct nlattr *tb[IFLA_MAX + 1];
 	struct net *dest_net, *link_net;
@@ -3049,6 +3050,15 @@ static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 	else
 		ifname[0] = '\0';
 
+	if (tb[IFLA_TARGET_NETNSID]) {
+		s32 netnsid = nla_get_s32(tb[IFLA_TARGET_NETNSID]);
+
+		tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(skb).sk, netnsid);
+		if (IS_ERR(tgt_net))
+			return PTR_ERR(tgt_net);
+		net = tgt_net;
+	}
+
 	ifm = nlmsg_data(nlh);
 	if (ifm->ifi_index > 0)
 		dev = __dev_get_by_index(net, ifm->ifi_index);
@@ -3059,6 +3069,23 @@ static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 			dev = NULL;
 	}
 
+	/* A hack to preserve kernel<->userspace interface.
+	 * It was previously allowed to pass the IFLA_TARGET_NETNSID
+	 * attribute as a way to _set_ the network namespace.  In this
+	 * case, the device interface was assumed to be in the  _current_
+	 * namespace.
+	 * If the device cannot be found in the target namespace then we
+	 * assume that the request is to set the device in the current
+	 * namespace and thus we attempt to find the device there.
+	 */
+	if (!dev && tgt_net) {
+		net = sock_net(skb->sk);
+		if (ifm->ifi_index > 0)
+			dev = __dev_get_by_index(net, ifm->ifi_index);
+		else if (tb[IFLA_IFNAME])
+			dev = __dev_get_by_name(net, ifname);
+	}
+
 	if (dev) {
 		master_dev = netdev_master_upper_dev_get(dev);
 		if (master_dev)
@@ -3253,6 +3280,8 @@ static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 			goto out_unregister;
 	}
 out:
+	if (tgt_net)
+		put_net(tgt_net);
 	if (link_net)
 		put_net(link_net);
 	put_net(dest_net);
-- 
2.20.1