bonding: API to create/enslave/release

Although it has been possible to create bonding devices, enslave and
release using the regular link API. The added API simplifies usage
and hides some of the compatibility logic.

F.e. enslave() and release() will both verify that the master assignment
has in fact been changed and return -NLE_OPNOTSUPP if it did not.

Also the API will make sure to use RTM_NEWLINK or RTM_SETLINK depending
on what is availble.

Examples are provided in src/ as nl-link-enslave.c and nl-link-release.c

1 files changed, 176 insertions, 0 deletions

diff --git a/lib/route/link/bonding.c b/lib/route/link/bonding.c
index 176be27..5788f69 100644
--- a/lib/route/link/bonding.c
+++ b/lib/route/link/bonding.c
@@ -24,6 +24,182 @@
 #include <netlink/netlink.h>
 #include <netlink/route/link/api.h>
 
+/**
+ * Create a new kernel bonding device
+ * @arg sock		netlink socket
+ * @arg name		name of bonding device or NULL
+ * @arg opts		bonding options (currently unused)
+ *
+ * Creates a new bonding device in the kernel. If no name is
+ * provided, the kernel will automatically pick a name of the
+ * form "type%d" (e.g. bond0, vlan1, etc.)
+ *
+ * The \a opts argument is currently unused. In the future, it
+ * may be used to carry additional bonding options to be set
+ * when creating the bonding device.
+ *
+ * @note When letting the kernel assign a name, it will become
+ *       difficult to retrieve the interface afterwards because
+ *       you have to guess the name the kernel has chosen. It is
+ *       therefore not recommended to not provide a device name.
+ *
+ * @see rtnl_link_bond_enslave()
+ * @see rtnl_link_bond_release()
+ *
+ * @return 0 on success or a negative error code
+ */
+int rtnl_link_bond_add(struct nl_sock *sock, const char *name,
+		       struct rtnl_link *opts)
+{
+	struct rtnl_link *link;
+	int err;
+
+	if (!(link = rtnl_link_alloc()))
+		return -NLE_NOMEM;
+
+	if (!name) {
+		if (opts)
+			name = rtnl_link_get_name(opts);
+
+		if (!name)
+			return -NLE_MISSING_ATTR;
+	}
+
+	if ((err = rtnl_link_set_type(link, "bond")) < 0)
+		goto errout;
+	
+
+		rtnl_link_set_name(link, name);
+
+	err = rtnl_link_add(sock, link, NLM_F_CREATE);
+errout:
+	rtnl_link_put(link);
+
+	return err;
+}
+
+/**
+ * Add a link to a bond (enslave)
+ * @arg sock		netlink socket
+ * @arg master		ifindex of bonding master
+ * @arg slave		ifindex of slave link to add to bond
+ *
+ * This function is identical to rtnl_link_bond_enslave() except that
+ * it takes interface indices instead of rtnl_link objcets.
+ *
+ * @see rtnl_link_bond_enslave()
+ *
+ * @return 0 on success or a negative error code.
+ */
+int rtnl_link_bond_enslave_ifindex(struct nl_sock *sock, int master,
+				   int slave)
+{
+	struct rtnl_link *link;
+	int err;
+
+	if (!(link = rtnl_link_alloc()))
+		return -NLE_NOMEM;
+
+	if ((err = rtnl_link_set_type(link, "bond")) < 0)
+		goto errout;
+
+	rtnl_link_set_ifindex(link, slave);
+	rtnl_link_set_master(link, master);
+	
+	if ((err = rtnl_link_change(sock, link, link, 0)) < 0)
+		goto errout;
+
+	rtnl_link_put(link);
+
+	/*
+	 * Due to the kernel not signaling whether this opertion is
+	 * supported or not, we will retrieve the attribute to see  if the
+	 * request was successful. If the master assigned remains unchanged
+	 * we will return NLE_OPNOTSUPP to allow performing backwards
+	 * compatibility of some sort.
+	 */
+	if ((err = rtnl_link_get_kernel(sock, slave, NULL, &link)) < 0)
+		return err;
+
+	if (rtnl_link_get_master(link) != master)
+		err = -NLE_OPNOTSUPP;
+
+errout:
+	rtnl_link_put(link);
+
+	return err;
+}
+
+/**
+ * Add a link to a bond (enslave)
+ * @arg sock		netlink socket
+ * @arg master		bonding master
+ * @arg slave		slave link to add to bond
+ *
+ * Constructs a RTM_NEWLINK or RTM_SETLINK message adding the slave to
+ * the master and sends the request via the specified netlink socket.
+ *
+ * @note The feature of enslaving/releasing via netlink has only been added
+ *       recently to the kernel (Feb 2011). Also, the kernel does not signal
+ *       if the operation is not supported. Therefore this function will
+ *       verify if the master assignment has changed and will return
+ *       -NLE_OPNOTSUPP if it did not.
+ *
+ * @see rtnl_link_bond_enslave_ifindex()
+ * @see rtnl_link_bond_release()
+ *
+ * @return 0 on success or a negative error code.
+ */
+int rtnl_link_bond_enslave(struct nl_sock *sock, struct rtnl_link *master,
+			   struct rtnl_link *slave)
+{
+	return rtnl_link_bond_enslave_ifindex(sock,
+				rtnl_link_get_ifindex(master),
+				rtnl_link_get_ifindex(slave));
+}
+
+/**
+ * Release a link from a bond
+ * @arg sock		netlink socket
+ * @arg slave		slave link to be released
+ *
+ * This function is identical to rtnl_link_bond_release() except that
+ * it takes an interface index instead of a rtnl_link object.
+ *
+ * @see rtnl_link_bond_release()
+ *
+ * @return 0 on success or a negative error code.
+ */
+int rtnl_link_bond_release_ifindex(struct nl_sock *sock, int slave)
+{
+	return rtnl_link_bond_enslave_ifindex(sock, 0, slave);
+}
+
+/**
+ * Release a link from a bond
+ * @arg sock		netlink socket
+ * @arg slave		slave link to be released
+ *
+ * Constructs a RTM_NEWLINK or RTM_SETLINK message releasing the slave from
+ * its master and sends the request via the specified netlink socket.
+ *
+ * @note The feature of enslaving/releasing via netlink has only been added
+ *       recently to the kernel (Feb 2011). Also, the kernel does not signal
+ *       if the operation is not supported. Therefore this function will
+ *       verify if the master assignment has changed and will return
+ *       -NLE_OPNOTSUPP if it did not.
+ *
+ * @see rtnl_link_bond_release_ifindex()
+ * @see rtnl_link_bond_enslave()
+ *
+ * @return 0 on success or a negative error code.
+ */
+int rtnl_link_bond_release(struct nl_sock *sock, struct rtnl_link *slave)
+{
+	return rtnl_link_bond_release_ifindex(sock,
+				rtnl_link_get_ifindex(slave));
+}
+
 static struct rtnl_link_info_ops bonding_info_ops = {
 	.io_name		= "bond",
 };