[NET]: Add macvlan driver
authorPatrick McHardy <kaber@trash.net>
Sun, 15 Jul 2007 01:55:06 +0000 (18:55 -0700)
committerDavid S. Miller <davem@davemloft.net>
Sun, 15 Jul 2007 01:55:06 +0000 (18:55 -0700)
Add macvlan driver, which allows to create virtual ethernet devices
based on MAC address.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
MAINTAINERS
drivers/net/Kconfig
drivers/net/Makefile
drivers/net/macvlan.c [new file with mode: 0644]
include/linux/if_macvlan.h [new file with mode: 0644]
include/linux/netdevice.h
net/core/dev.c

index 845fbf4478b3adc2b972617892db8fadb8177fd2..360eb581953b658ae9ba6947b95c081866f5fd78 100644 (file)
@@ -2330,6 +2330,12 @@ W:       http://linuxwireless.org/
 T:     git kernel.org:/pub/scm/linux/kernel/git/jbenc/mac80211.git
 S:     Maintained
 
+MACVLAN DRIVER
+P:     Patrick McHardy
+M:     kaber@trash.net
+L:     netdev@vger.kernel.org
+S:     Maintained
+
 MARVELL YUKON / SYSKONNECT DRIVER
 P:     Mirko Lindner
 M:     mlindner@syskonnect.de
index ba314adf68b86ecf856a6ebecada8c2ef019feb9..d17d64eb7065e251e7b6f09f6060a9671b271e03 100644 (file)
@@ -82,6 +82,16 @@ config BONDING
          To compile this driver as a module, choose M here: the module
          will be called bonding.
 
+config MACVLAN
+       tristate "MAC-VLAN support (EXPERIMENTAL)"
+       depends on EXPERIMENTAL
+       ---help---
+         This allows one to create virtual interfaces that map packets to
+         or from specific MAC addresses to a particular interface.
+
+         To compile this driver as a module, choose M here: the module
+         will be called macvlan.
+
 config EQUALIZER
        tristate "EQL (serial line load balancing) support"
        ---help---
index a2241e6e145765a896d90eb91cdbe8b9b54b53f0..c26b8674213cf33ac6c15fbd7a489b08ef44f550 100644 (file)
@@ -128,6 +128,7 @@ obj-$(CONFIG_SLHC) += slhc.o
 
 obj-$(CONFIG_DUMMY) += dummy.o
 obj-$(CONFIG_IFB) += ifb.o
+obj-$(CONFIG_MACVLAN) += macvlan.o
 obj-$(CONFIG_DE600) += de600.o
 obj-$(CONFIG_DE620) += de620.o
 obj-$(CONFIG_LANCE) += lance.o
diff --git a/drivers/net/macvlan.c b/drivers/net/macvlan.c
new file mode 100644 (file)
index 0000000..dc74d00
--- /dev/null
@@ -0,0 +1,496 @@
+/*
+ * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * The code this is based on carried the following copyright notice:
+ * ---
+ * (C) Copyright 2001-2006
+ * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
+ * Re-worked by Ben Greear <greearb@candelatech.com>
+ * ---
+ */
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/errno.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/list.h>
+#include <linux/notifier.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/ethtool.h>
+#include <linux/if_arp.h>
+#include <linux/if_link.h>
+#include <linux/if_macvlan.h>
+#include <net/rtnetlink.h>
+
+#define MACVLAN_HASH_SIZE      (1 << BITS_PER_BYTE)
+
+struct macvlan_port {
+       struct net_device       *dev;
+       struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
+       struct list_head        vlans;
+};
+
+struct macvlan_dev {
+       struct net_device       *dev;
+       struct list_head        list;
+       struct hlist_node       hlist;
+       struct macvlan_port     *port;
+       struct net_device       *lowerdev;
+};
+
+
+static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
+                                              const unsigned char *addr)
+{
+       struct macvlan_dev *vlan;
+       struct hlist_node *n;
+
+       hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[addr[5]], hlist) {
+               if (!compare_ether_addr(vlan->dev->dev_addr, addr))
+                       return vlan;
+       }
+       return NULL;
+}
+
+static void macvlan_broadcast(struct sk_buff *skb,
+                             const struct macvlan_port *port)
+{
+       const struct ethhdr *eth = eth_hdr(skb);
+       const struct macvlan_dev *vlan;
+       struct hlist_node *n;
+       struct net_device *dev;
+       struct sk_buff *nskb;
+       unsigned int i;
+
+       for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
+               hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[i], hlist) {
+                       dev = vlan->dev;
+                       if (unlikely(!(dev->flags & IFF_UP)))
+                               continue;
+
+                       nskb = skb_clone(skb, GFP_ATOMIC);
+                       if (nskb == NULL) {
+                               dev->stats.rx_errors++;
+                               dev->stats.rx_dropped++;
+                               continue;
+                       }
+
+                       dev->stats.rx_bytes += skb->len + ETH_HLEN;
+                       dev->stats.rx_packets++;
+                       dev->stats.multicast++;
+                       dev->last_rx = jiffies;
+
+                       nskb->dev = dev;
+                       if (!compare_ether_addr(eth->h_dest, dev->broadcast))
+                               nskb->pkt_type = PACKET_BROADCAST;
+                       else
+                               nskb->pkt_type = PACKET_MULTICAST;
+
+                       netif_rx(nskb);
+               }
+       }
+}
+
+/* called under rcu_read_lock() from netif_receive_skb */
+static struct sk_buff *macvlan_handle_frame(struct sk_buff *skb)
+{
+       const struct ethhdr *eth = eth_hdr(skb);
+       const struct macvlan_port *port;
+       const struct macvlan_dev *vlan;
+       struct net_device *dev;
+
+       port = rcu_dereference(skb->dev->macvlan_port);
+       if (port == NULL)
+               return skb;
+
+       if (is_multicast_ether_addr(eth->h_dest)) {
+               macvlan_broadcast(skb, port);
+               return skb;
+       }
+
+       vlan = macvlan_hash_lookup(port, eth->h_dest);
+       if (vlan == NULL)
+               return skb;
+
+       dev = vlan->dev;
+       if (unlikely(!(dev->flags & IFF_UP))) {
+               kfree_skb(skb);
+               return NULL;
+       }
+
+       skb = skb_share_check(skb, GFP_ATOMIC);
+       if (skb == NULL) {
+               dev->stats.rx_errors++;
+               dev->stats.rx_dropped++;
+               return NULL;
+       }
+
+       dev->stats.rx_bytes += skb->len + ETH_HLEN;
+       dev->stats.rx_packets++;
+       dev->last_rx = jiffies;
+
+       skb->dev = dev;
+       skb->pkt_type = PACKET_HOST;
+
+       netif_rx(skb);
+       return NULL;
+}
+
+static int macvlan_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+       const struct macvlan_dev *vlan = netdev_priv(dev);
+       unsigned int len = skb->len;
+       int ret;
+
+       skb->dev = vlan->lowerdev;
+       ret = dev_queue_xmit(skb);
+
+       if (likely(ret == NET_XMIT_SUCCESS)) {
+               dev->stats.tx_packets++;
+               dev->stats.tx_bytes += len;
+       } else {
+               dev->stats.tx_errors++;
+               dev->stats.tx_aborted_errors++;
+       }
+       return NETDEV_TX_OK;
+}
+
+static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
+                              unsigned short type, void *daddr, void *saddr,
+                              unsigned len)
+{
+       const struct macvlan_dev *vlan = netdev_priv(dev);
+       struct net_device *lowerdev = vlan->lowerdev;
+
+       return lowerdev->hard_header(skb, lowerdev, type, daddr,
+                                    saddr ? : dev->dev_addr, len);
+}
+
+static int macvlan_open(struct net_device *dev)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       struct macvlan_port *port = vlan->port;
+       struct net_device *lowerdev = vlan->lowerdev;
+       int err;
+
+       err = dev_unicast_add(lowerdev, dev->dev_addr, ETH_ALEN);
+       if (err < 0)
+               return err;
+       if (dev->flags & IFF_ALLMULTI)
+               dev_set_allmulti(lowerdev, 1);
+
+       hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[dev->dev_addr[5]]);
+       return 0;
+}
+
+static int macvlan_stop(struct net_device *dev)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       struct net_device *lowerdev = vlan->lowerdev;
+
+       dev_mc_unsync(lowerdev, dev);
+       if (dev->flags & IFF_ALLMULTI)
+               dev_set_allmulti(lowerdev, -1);
+
+       dev_unicast_delete(lowerdev, dev->dev_addr, ETH_ALEN);
+
+       hlist_del_rcu(&vlan->hlist);
+       synchronize_rcu();
+       return 0;
+}
+
+static void macvlan_change_rx_flags(struct net_device *dev, int change)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       struct net_device *lowerdev = vlan->lowerdev;
+
+       if (change & IFF_ALLMULTI)
+               dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
+}
+
+static void macvlan_set_multicast_list(struct net_device *dev)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+
+       dev_mc_sync(vlan->lowerdev, dev);
+}
+
+static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+
+       if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
+               return -EINVAL;
+       dev->mtu = new_mtu;
+       return 0;
+}
+
+/*
+ * macvlan network devices have devices nesting below it and are a special
+ * "super class" of normal network devices; split their locks off into a
+ * separate class since they always nest.
+ */
+static struct lock_class_key macvlan_netdev_xmit_lock_key;
+
+#define MACVLAN_FEATURES \
+       (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
+        NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
+        NETIF_F_TSO_ECN | NETIF_F_TSO6)
+
+#define MACVLAN_STATE_MASK \
+       ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
+
+static int macvlan_init(struct net_device *dev)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       const struct net_device *lowerdev = vlan->lowerdev;
+
+       dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
+                                 (lowerdev->state & MACVLAN_STATE_MASK);
+       dev->features           = lowerdev->features & MACVLAN_FEATURES;
+       dev->iflink             = lowerdev->ifindex;
+
+       lockdep_set_class(&dev->_xmit_lock, &macvlan_netdev_xmit_lock_key);
+       return 0;
+}
+
+static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
+                                       struct ethtool_drvinfo *drvinfo)
+{
+       snprintf(drvinfo->driver, 32, "macvlan");
+       snprintf(drvinfo->version, 32, "0.1");
+}
+
+static u32 macvlan_ethtool_get_rx_csum(struct net_device *dev)
+{
+       const struct macvlan_dev *vlan = netdev_priv(dev);
+       struct net_device *lowerdev = vlan->lowerdev;
+
+       if (lowerdev->ethtool_ops->get_rx_csum == NULL)
+               return 0;
+       return lowerdev->ethtool_ops->get_rx_csum(lowerdev);
+}
+
+static const struct ethtool_ops macvlan_ethtool_ops = {
+       .get_link               = ethtool_op_get_link,
+       .get_rx_csum            = macvlan_ethtool_get_rx_csum,
+       .get_tx_csum            = ethtool_op_get_tx_csum,
+       .get_tso                = ethtool_op_get_tso,
+       .get_ufo                = ethtool_op_get_ufo,
+       .get_sg                 = ethtool_op_get_sg,
+       .get_drvinfo            = macvlan_ethtool_get_drvinfo,
+};
+
+static void macvlan_setup(struct net_device *dev)
+{
+       ether_setup(dev);
+
+       dev->init               = macvlan_init;
+       dev->open               = macvlan_open;
+       dev->stop               = macvlan_stop;
+       dev->change_mtu         = macvlan_change_mtu;
+       dev->change_rx_flags    = macvlan_change_rx_flags;
+       dev->set_multicast_list = macvlan_set_multicast_list;
+       dev->hard_header        = macvlan_hard_header;
+       dev->hard_start_xmit    = macvlan_hard_start_xmit;
+       dev->destructor         = free_netdev;
+       dev->ethtool_ops        = &macvlan_ethtool_ops;
+       dev->tx_queue_len       = 0;
+}
+
+static int macvlan_port_create(struct net_device *dev)
+{
+       struct macvlan_port *port;
+       unsigned int i;
+
+       if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
+               return -EINVAL;
+
+       port = kzalloc(sizeof(*port), GFP_KERNEL);
+       if (port == NULL)
+               return -ENOMEM;
+
+       port->dev = dev;
+       INIT_LIST_HEAD(&port->vlans);
+       for (i = 0; i < MACVLAN_HASH_SIZE; i++)
+               INIT_HLIST_HEAD(&port->vlan_hash[i]);
+       rcu_assign_pointer(dev->macvlan_port, port);
+       return 0;
+}
+
+static void macvlan_port_destroy(struct net_device *dev)
+{
+       struct macvlan_port *port = dev->macvlan_port;
+
+       rcu_assign_pointer(dev->macvlan_port, NULL);
+       synchronize_rcu();
+       kfree(port);
+}
+
+static void macvlan_transfer_operstate(struct net_device *dev)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       const struct net_device *lowerdev = vlan->lowerdev;
+
+       if (lowerdev->operstate == IF_OPER_DORMANT)
+               netif_dormant_on(dev);
+       else
+               netif_dormant_off(dev);
+
+       if (netif_carrier_ok(lowerdev)) {
+               if (!netif_carrier_ok(dev))
+                       netif_carrier_on(dev);
+       } else {
+               if (netif_carrier_ok(lowerdev))
+                       netif_carrier_off(dev);
+       }
+}
+
+static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
+{
+       if (tb[IFLA_ADDRESS]) {
+               if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
+                       return -EINVAL;
+               if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
+                       return -EADDRNOTAVAIL;
+       }
+       return 0;
+}
+
+static int macvlan_newlink(struct net_device *dev,
+                          struct nlattr *tb[], struct nlattr *data[])
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       struct macvlan_port *port;
+       struct net_device *lowerdev;
+       int err;
+
+       if (!tb[IFLA_LINK])
+               return -EINVAL;
+
+       lowerdev = __dev_get_by_index(nla_get_u32(tb[IFLA_LINK]));
+       if (lowerdev == NULL)
+               return -ENODEV;
+
+       if (!tb[IFLA_MTU])
+               dev->mtu = lowerdev->mtu;
+       else if (dev->mtu > lowerdev->mtu)
+               return -EINVAL;
+
+       if (!tb[IFLA_ADDRESS])
+               random_ether_addr(dev->dev_addr);
+
+       if (lowerdev->macvlan_port == NULL) {
+               err = macvlan_port_create(lowerdev);
+               if (err < 0)
+                       return err;
+       }
+       port = lowerdev->macvlan_port;
+
+       vlan->lowerdev = lowerdev;
+       vlan->dev      = dev;
+       vlan->port     = port;
+
+       err = register_netdevice(dev);
+       if (err < 0)
+               return err;
+
+       list_add_tail(&vlan->list, &port->vlans);
+       macvlan_transfer_operstate(dev);
+       return 0;
+}
+
+static void macvlan_dellink(struct net_device *dev)
+{
+       struct macvlan_dev *vlan = netdev_priv(dev);
+       struct macvlan_port *port = vlan->port;
+
+       list_del(&vlan->list);
+       unregister_netdevice(dev);
+
+       if (list_empty(&port->vlans))
+               macvlan_port_destroy(dev);
+}
+
+static struct rtnl_link_ops macvlan_link_ops __read_mostly = {
+       .kind           = "macvlan",
+       .priv_size      = sizeof(struct macvlan_dev),
+       .setup          = macvlan_setup,
+       .validate       = macvlan_validate,
+       .newlink        = macvlan_newlink,
+       .dellink        = macvlan_dellink,
+};
+
+static int macvlan_device_event(struct notifier_block *unused,
+                               unsigned long event, void *ptr)
+{
+       struct net_device *dev = ptr;
+       struct macvlan_dev *vlan, *next;
+       struct macvlan_port *port;
+
+       port = dev->macvlan_port;
+       if (port == NULL)
+               return NOTIFY_DONE;
+
+       switch (event) {
+       case NETDEV_CHANGE:
+               list_for_each_entry(vlan, &port->vlans, list)
+                       macvlan_transfer_operstate(vlan->dev);
+               break;
+       case NETDEV_FEAT_CHANGE:
+               list_for_each_entry(vlan, &port->vlans, list) {
+                       vlan->dev->features = dev->features & MACVLAN_FEATURES;
+                       netdev_features_change(vlan->dev);
+               }
+               break;
+       case NETDEV_UNREGISTER:
+               list_for_each_entry_safe(vlan, next, &port->vlans, list)
+                       macvlan_dellink(vlan->dev);
+               break;
+       }
+       return NOTIFY_DONE;
+}
+
+static struct notifier_block macvlan_notifier_block __read_mostly = {
+       .notifier_call  = macvlan_device_event,
+};
+
+static int __init macvlan_init_module(void)
+{
+       int err;
+
+       register_netdevice_notifier(&macvlan_notifier_block);
+       macvlan_handle_frame_hook = macvlan_handle_frame;
+
+       err = rtnl_link_register(&macvlan_link_ops);
+       if (err < 0)
+               goto err1;
+       return 0;
+err1:
+       macvlan_handle_frame_hook = macvlan_handle_frame;
+       unregister_netdevice_notifier(&macvlan_notifier_block);
+       return err;
+}
+
+static void __exit macvlan_cleanup_module(void)
+{
+       rtnl_link_unregister(&macvlan_link_ops);
+       macvlan_handle_frame_hook = NULL;
+       unregister_netdevice_notifier(&macvlan_notifier_block);
+}
+
+module_init(macvlan_init_module);
+module_exit(macvlan_cleanup_module);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
+MODULE_DESCRIPTION("Driver for MAC address based VLANs");
+MODULE_ALIAS_RTNL_LINK("macvlan");
diff --git a/include/linux/if_macvlan.h b/include/linux/if_macvlan.h
new file mode 100644 (file)
index 0000000..0d9d7ea
--- /dev/null
@@ -0,0 +1,9 @@
+#ifndef _LINUX_IF_MACVLAN_H
+#define _LINUX_IF_MACVLAN_H
+
+#ifdef __KERNEL__
+
+extern struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *);
+
+#endif /* __KERNEL__ */
+#endif /* _LINUX_IF_MACVLAN_H */
index e5af458ab04ba7632aa431473ae5c4834aab24ca..322b5eae57ddf06192760963e91e1c1e64354873 100644 (file)
@@ -564,6 +564,8 @@ struct net_device
 
        /* bridge stuff */
        struct net_bridge_port  *br_port;
+       /* macvlan */
+       struct macvlan_port     *macvlan_port;
 
        /* class/net/name entry */
        struct device           dev;
index 59ec811d2b54ea618c4e572aaa219cb2f2c61183..13a0d9f6da54c02cbed38292ddef7eaa1d9241a0 100644 (file)
@@ -98,6 +98,7 @@
 #include <linux/seq_file.h>
 #include <linux/stat.h>
 #include <linux/if_bridge.h>
+#include <linux/if_macvlan.h>
 #include <net/dst.h>
 #include <net/pkt_sched.h>
 #include <net/checksum.h>
@@ -1813,6 +1814,28 @@ static inline struct sk_buff *handle_bridge(struct sk_buff *skb,
 #define handle_bridge(skb, pt_prev, ret, orig_dev)     (skb)
 #endif
 
+#if defined(CONFIG_MACVLAN) || defined(CONFIG_MACVLAN_MODULE)
+struct sk_buff *(*macvlan_handle_frame_hook)(struct sk_buff *skb) __read_mostly;
+EXPORT_SYMBOL_GPL(macvlan_handle_frame_hook);
+
+static inline struct sk_buff *handle_macvlan(struct sk_buff *skb,
+                                            struct packet_type **pt_prev,
+                                            int *ret,
+                                            struct net_device *orig_dev)
+{
+       if (skb->dev->macvlan_port == NULL)
+               return skb;
+
+       if (*pt_prev) {
+               *ret = deliver_skb(skb, *pt_prev, orig_dev);
+               *pt_prev = NULL;
+       }
+       return macvlan_handle_frame_hook(skb);
+}
+#else
+#define handle_macvlan(skb, pt_prev, ret, orig_dev)    (skb)
+#endif
+
 #ifdef CONFIG_NET_CLS_ACT
 /* TODO: Maybe we should just force sch_ingress to be compiled in
  * when CONFIG_NET_CLS_ACT is? otherwise some useless instructions
@@ -1918,6 +1941,9 @@ ncls:
 #endif
 
        skb = handle_bridge(skb, &pt_prev, &ret, orig_dev);
+       if (!skb)
+               goto out;
+       skb = handle_macvlan(skb, &pt_prev, &ret, orig_dev);
        if (!skb)
                goto out;