2 * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License as
6 * published by the Free Software Foundation; either version 2 of
7 * the License, or (at your option) any later version.
9 * The code this is based on carried the following copyright notice:
11 * (C) Copyright 2001-2006
12 * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
13 * Re-worked by Ben Greear <greearb@candelatech.com>
16 #include <linux/kernel.h>
17 #include <linux/types.h>
18 #include <linux/module.h>
19 #include <linux/init.h>
20 #include <linux/errno.h>
21 #include <linux/slab.h>
22 #include <linux/string.h>
23 #include <linux/rculist.h>
24 #include <linux/notifier.h>
25 #include <linux/netdevice.h>
26 #include <linux/etherdevice.h>
27 #include <linux/ethtool.h>
28 #include <linux/if_arp.h>
29 #include <linux/if_vlan.h>
30 #include <linux/if_link.h>
31 #include <linux/if_macvlan.h>
32 #include <linux/hash.h>
33 #include <net/rtnetlink.h>
36 #define MACVLAN_HASH_SIZE (1 << BITS_PER_BYTE)
39 struct net_device *dev;
40 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE];
41 struct list_head vlans;
47 static void macvlan_port_destroy(struct net_device *dev);
49 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
51 return rcu_dereference(dev->rx_handler_data);
54 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
56 return rtnl_dereference(dev->rx_handler_data);
59 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
61 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
62 const unsigned char *addr)
64 struct macvlan_dev *vlan;
66 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[addr[5]], hlist) {
67 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
73 static void macvlan_hash_add(struct macvlan_dev *vlan)
75 struct macvlan_port *port = vlan->port;
76 const unsigned char *addr = vlan->dev->dev_addr;
78 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
81 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
83 hlist_del_rcu(&vlan->hlist);
88 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
89 const unsigned char *addr)
91 macvlan_hash_del(vlan, true);
92 /* Now that we are unhashed it is safe to change the device
93 * address without confusing packet delivery.
95 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
96 macvlan_hash_add(vlan);
99 static int macvlan_addr_busy(const struct macvlan_port *port,
100 const unsigned char *addr)
102 /* Test to see if the specified multicast address is
103 * currently in use by the underlying device or
106 if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
109 if (macvlan_hash_lookup(port, addr))
116 static int macvlan_broadcast_one(struct sk_buff *skb,
117 const struct macvlan_dev *vlan,
118 const struct ethhdr *eth, bool local)
120 struct net_device *dev = vlan->dev;
123 return dev_forward_skb(dev, skb);
126 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
127 skb->pkt_type = PACKET_BROADCAST;
129 skb->pkt_type = PACKET_MULTICAST;
131 return netif_rx(skb);
134 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
136 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
140 static unsigned int mc_hash(const struct macvlan_dev *vlan,
141 const unsigned char *addr)
143 u32 val = __get_unaligned_cpu32(addr + 2);
145 val ^= macvlan_hash_mix(vlan);
146 return hash_32(val, MACVLAN_MC_FILTER_BITS);
149 static void macvlan_broadcast(struct sk_buff *skb,
150 const struct macvlan_port *port,
151 struct net_device *src,
152 enum macvlan_mode mode)
154 const struct ethhdr *eth = eth_hdr(skb);
155 const struct macvlan_dev *vlan;
156 struct sk_buff *nskb;
161 if (skb->protocol == htons(ETH_P_PAUSE))
164 for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
165 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
166 if (vlan->dev == src || !(vlan->mode & mode))
169 hash = mc_hash(vlan, eth->h_dest);
170 if (!test_bit(hash, vlan->mc_filter))
174 nskb = skb_clone(skb, GFP_ATOMIC);
176 err = macvlan_broadcast_one(
178 mode == MACVLAN_MODE_BRIDGE);
179 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
180 err == NET_RX_SUCCESS, 1);
185 /* called under rcu_read_lock() from netif_receive_skb */
186 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
188 struct macvlan_port *port;
189 struct sk_buff *skb = *pskb;
190 const struct ethhdr *eth = eth_hdr(skb);
191 const struct macvlan_dev *vlan;
192 const struct macvlan_dev *src;
193 struct net_device *dev;
194 unsigned int len = 0;
195 int ret = NET_RX_DROP;
197 port = macvlan_port_get_rcu(skb->dev);
198 if (is_multicast_ether_addr(eth->h_dest)) {
199 skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
201 return RX_HANDLER_CONSUMED;
203 src = macvlan_hash_lookup(port, eth->h_source);
205 /* frame comes from an external address */
206 macvlan_broadcast(skb, port, NULL,
207 MACVLAN_MODE_PRIVATE |
209 MACVLAN_MODE_PASSTHRU|
210 MACVLAN_MODE_BRIDGE);
211 else if (src->mode == MACVLAN_MODE_VEPA)
212 /* flood to everyone except source */
213 macvlan_broadcast(skb, port, src->dev,
215 MACVLAN_MODE_BRIDGE);
216 else if (src->mode == MACVLAN_MODE_BRIDGE)
218 * flood only to VEPA ports, bridge ports
219 * already saw the frame on the way out.
221 macvlan_broadcast(skb, port, src->dev,
224 /* forward to original port. */
226 ret = macvlan_broadcast_one(skb, vlan, eth, 0);
230 return RX_HANDLER_PASS;
234 vlan = list_first_or_null_rcu(&port->vlans,
235 struct macvlan_dev, list);
237 vlan = macvlan_hash_lookup(port, eth->h_dest);
239 return RX_HANDLER_PASS;
242 if (unlikely(!(dev->flags & IFF_UP))) {
244 return RX_HANDLER_CONSUMED;
246 len = skb->len + ETH_HLEN;
247 skb = skb_share_check(skb, GFP_ATOMIC);
252 skb->pkt_type = PACKET_HOST;
257 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
258 return RX_HANDLER_CONSUMED;
261 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
263 const struct macvlan_dev *vlan = netdev_priv(dev);
264 const struct macvlan_port *port = vlan->port;
265 const struct macvlan_dev *dest;
267 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
268 const struct ethhdr *eth = (void *)skb->data;
270 /* send to other bridge ports directly */
271 if (is_multicast_ether_addr(eth->h_dest)) {
272 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
276 dest = macvlan_hash_lookup(port, eth->h_dest);
277 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
278 /* send to lowerdev first for its network taps */
279 dev_forward_skb(vlan->lowerdev, skb);
281 return NET_XMIT_SUCCESS;
286 skb->dev = vlan->lowerdev;
287 return dev_queue_xmit(skb);
290 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
291 struct net_device *dev)
293 unsigned int len = skb->len;
295 const struct macvlan_dev *vlan = netdev_priv(dev);
297 if (vlan->fwd_priv) {
298 skb->dev = vlan->lowerdev;
299 ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
301 ret = macvlan_queue_xmit(skb, dev);
304 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
305 struct vlan_pcpu_stats *pcpu_stats;
307 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
308 u64_stats_update_begin(&pcpu_stats->syncp);
309 pcpu_stats->tx_packets++;
310 pcpu_stats->tx_bytes += len;
311 u64_stats_update_end(&pcpu_stats->syncp);
313 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
318 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
319 unsigned short type, const void *daddr,
320 const void *saddr, unsigned len)
322 const struct macvlan_dev *vlan = netdev_priv(dev);
323 struct net_device *lowerdev = vlan->lowerdev;
325 return dev_hard_header(skb, lowerdev, type, daddr,
326 saddr ? : dev->dev_addr, len);
329 static const struct header_ops macvlan_hard_header_ops = {
330 .create = macvlan_hard_header,
331 .rebuild = eth_rebuild_header,
332 .parse = eth_header_parse,
333 .cache = eth_header_cache,
334 .cache_update = eth_header_cache_update,
337 static struct rtnl_link_ops macvlan_link_ops;
339 static int macvlan_open(struct net_device *dev)
341 struct macvlan_dev *vlan = netdev_priv(dev);
342 struct net_device *lowerdev = vlan->lowerdev;
345 if (vlan->port->passthru) {
346 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
347 err = dev_set_promiscuity(lowerdev, 1);
354 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD &&
355 dev->rtnl_link_ops == &macvlan_link_ops) {
357 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
359 /* If we get a NULL pointer back, or if we get an error
360 * then we should just fall through to the non accelerated path
362 if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
363 vlan->fwd_priv = NULL;
369 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
372 err = dev_uc_add(lowerdev, dev->dev_addr);
375 if (dev->flags & IFF_ALLMULTI) {
376 err = dev_set_allmulti(lowerdev, 1);
382 macvlan_hash_add(vlan);
386 dev_uc_del(lowerdev, dev->dev_addr);
388 if (vlan->fwd_priv) {
389 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
391 vlan->fwd_priv = NULL;
396 static int macvlan_stop(struct net_device *dev)
398 struct macvlan_dev *vlan = netdev_priv(dev);
399 struct net_device *lowerdev = vlan->lowerdev;
401 if (vlan->fwd_priv) {
402 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
404 vlan->fwd_priv = NULL;
408 dev_uc_unsync(lowerdev, dev);
409 dev_mc_unsync(lowerdev, dev);
411 if (vlan->port->passthru) {
412 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
413 dev_set_promiscuity(lowerdev, -1);
417 if (dev->flags & IFF_ALLMULTI)
418 dev_set_allmulti(lowerdev, -1);
420 dev_uc_del(lowerdev, dev->dev_addr);
423 macvlan_hash_del(vlan, !dev->dismantle);
427 static int macvlan_set_mac_address(struct net_device *dev, void *p)
429 struct macvlan_dev *vlan = netdev_priv(dev);
430 struct net_device *lowerdev = vlan->lowerdev;
431 struct sockaddr *addr = p;
434 if (!is_valid_ether_addr(addr->sa_data))
435 return -EADDRNOTAVAIL;
437 if (!(dev->flags & IFF_UP)) {
438 /* Just copy in the new address */
439 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
441 /* Rehash and update the device filters */
442 if (macvlan_addr_busy(vlan->port, addr->sa_data))
445 err = dev_uc_add(lowerdev, addr->sa_data);
449 dev_uc_del(lowerdev, dev->dev_addr);
451 macvlan_hash_change_addr(vlan, addr->sa_data);
456 static void macvlan_change_rx_flags(struct net_device *dev, int change)
458 struct macvlan_dev *vlan = netdev_priv(dev);
459 struct net_device *lowerdev = vlan->lowerdev;
461 if (dev->flags & IFF_UP) {
462 if (change & IFF_ALLMULTI)
463 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
467 static void macvlan_set_mac_lists(struct net_device *dev)
469 struct macvlan_dev *vlan = netdev_priv(dev);
471 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
472 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
474 struct netdev_hw_addr *ha;
475 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
477 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
478 netdev_for_each_mc_addr(ha, dev) {
479 __set_bit(mc_hash(vlan, ha->addr), filter);
482 __set_bit(mc_hash(vlan, dev->broadcast), filter);
484 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
486 dev_uc_sync(vlan->lowerdev, dev);
487 dev_mc_sync(vlan->lowerdev, dev);
490 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
492 struct macvlan_dev *vlan = netdev_priv(dev);
494 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
501 * macvlan network devices have devices nesting below it and are a special
502 * "super class" of normal network devices; split their locks off into a
503 * separate class since they always nest.
505 static struct lock_class_key macvlan_netdev_xmit_lock_key;
506 static struct lock_class_key macvlan_netdev_addr_lock_key;
508 #define ALWAYS_ON_FEATURES \
509 (NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX)
511 #define MACVLAN_FEATURES \
512 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
513 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
514 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
515 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
517 #define MACVLAN_STATE_MASK \
518 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
520 static int macvlan_get_nest_level(struct net_device *dev)
522 return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
525 static void macvlan_set_lockdep_class_one(struct net_device *dev,
526 struct netdev_queue *txq,
529 lockdep_set_class(&txq->_xmit_lock,
530 &macvlan_netdev_xmit_lock_key);
533 static void macvlan_set_lockdep_class(struct net_device *dev)
535 lockdep_set_class_and_subclass(&dev->addr_list_lock,
536 &macvlan_netdev_addr_lock_key,
537 macvlan_get_nest_level(dev));
538 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
541 static int macvlan_init(struct net_device *dev)
543 struct macvlan_dev *vlan = netdev_priv(dev);
544 const struct net_device *lowerdev = vlan->lowerdev;
547 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
548 (lowerdev->state & MACVLAN_STATE_MASK);
549 dev->features = lowerdev->features & MACVLAN_FEATURES;
550 dev->features |= ALWAYS_ON_FEATURES;
551 dev->vlan_features = lowerdev->vlan_features & MACVLAN_FEATURES;
552 dev->gso_max_size = lowerdev->gso_max_size;
553 dev->iflink = lowerdev->ifindex;
554 dev->hard_header_len = lowerdev->hard_header_len;
556 macvlan_set_lockdep_class(dev);
558 vlan->pcpu_stats = alloc_percpu(struct vlan_pcpu_stats);
559 if (!vlan->pcpu_stats)
562 for_each_possible_cpu(i) {
563 struct vlan_pcpu_stats *mvlstats;
564 mvlstats = per_cpu_ptr(vlan->pcpu_stats, i);
565 u64_stats_init(&mvlstats->syncp);
571 static void macvlan_uninit(struct net_device *dev)
573 struct macvlan_dev *vlan = netdev_priv(dev);
574 struct macvlan_port *port = vlan->port;
576 free_percpu(vlan->pcpu_stats);
580 macvlan_port_destroy(port->dev);
583 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
584 struct rtnl_link_stats64 *stats)
586 struct macvlan_dev *vlan = netdev_priv(dev);
588 if (vlan->pcpu_stats) {
589 struct vlan_pcpu_stats *p;
590 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
591 u32 rx_errors = 0, tx_dropped = 0;
595 for_each_possible_cpu(i) {
596 p = per_cpu_ptr(vlan->pcpu_stats, i);
598 start = u64_stats_fetch_begin_bh(&p->syncp);
599 rx_packets = p->rx_packets;
600 rx_bytes = p->rx_bytes;
601 rx_multicast = p->rx_multicast;
602 tx_packets = p->tx_packets;
603 tx_bytes = p->tx_bytes;
604 } while (u64_stats_fetch_retry_bh(&p->syncp, start));
606 stats->rx_packets += rx_packets;
607 stats->rx_bytes += rx_bytes;
608 stats->multicast += rx_multicast;
609 stats->tx_packets += tx_packets;
610 stats->tx_bytes += tx_bytes;
611 /* rx_errors & tx_dropped are u32, updated
612 * without syncp protection.
614 rx_errors += p->rx_errors;
615 tx_dropped += p->tx_dropped;
617 stats->rx_errors = rx_errors;
618 stats->rx_dropped = rx_errors;
619 stats->tx_dropped = tx_dropped;
624 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
625 __be16 proto, u16 vid)
627 struct macvlan_dev *vlan = netdev_priv(dev);
628 struct net_device *lowerdev = vlan->lowerdev;
630 return vlan_vid_add(lowerdev, proto, vid);
633 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
634 __be16 proto, u16 vid)
636 struct macvlan_dev *vlan = netdev_priv(dev);
637 struct net_device *lowerdev = vlan->lowerdev;
639 vlan_vid_del(lowerdev, proto, vid);
643 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
644 struct net_device *dev,
645 const unsigned char *addr,
648 struct macvlan_dev *vlan = netdev_priv(dev);
651 if (!vlan->port->passthru)
654 if (flags & NLM_F_REPLACE)
657 if (is_unicast_ether_addr(addr))
658 err = dev_uc_add_excl(dev, addr);
659 else if (is_multicast_ether_addr(addr))
660 err = dev_mc_add_excl(dev, addr);
665 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
666 struct net_device *dev,
667 const unsigned char *addr)
669 struct macvlan_dev *vlan = netdev_priv(dev);
672 if (!vlan->port->passthru)
675 if (is_unicast_ether_addr(addr))
676 err = dev_uc_del(dev, addr);
677 else if (is_multicast_ether_addr(addr))
678 err = dev_mc_del(dev, addr);
683 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
684 struct ethtool_drvinfo *drvinfo)
686 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
687 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
690 static int macvlan_ethtool_get_settings(struct net_device *dev,
691 struct ethtool_cmd *cmd)
693 const struct macvlan_dev *vlan = netdev_priv(dev);
695 return __ethtool_get_settings(vlan->lowerdev, cmd);
698 static netdev_features_t macvlan_fix_features(struct net_device *dev,
699 netdev_features_t features)
701 struct macvlan_dev *vlan = netdev_priv(dev);
702 netdev_features_t mask;
704 features |= NETIF_F_ALL_FOR_ALL;
705 features &= (vlan->set_features | ~MACVLAN_FEATURES);
708 features = netdev_increment_features(vlan->lowerdev->features,
711 features |= ALWAYS_ON_FEATURES;
712 features &= ~NETIF_F_NETNS_LOCAL;
717 static const struct ethtool_ops macvlan_ethtool_ops = {
718 .get_link = ethtool_op_get_link,
719 .get_settings = macvlan_ethtool_get_settings,
720 .get_drvinfo = macvlan_ethtool_get_drvinfo,
723 static const struct net_device_ops macvlan_netdev_ops = {
724 .ndo_init = macvlan_init,
725 .ndo_uninit = macvlan_uninit,
726 .ndo_open = macvlan_open,
727 .ndo_stop = macvlan_stop,
728 .ndo_start_xmit = macvlan_start_xmit,
729 .ndo_change_mtu = macvlan_change_mtu,
730 .ndo_fix_features = macvlan_fix_features,
731 .ndo_change_rx_flags = macvlan_change_rx_flags,
732 .ndo_set_mac_address = macvlan_set_mac_address,
733 .ndo_set_rx_mode = macvlan_set_mac_lists,
734 .ndo_get_stats64 = macvlan_dev_get_stats64,
735 .ndo_validate_addr = eth_validate_addr,
736 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
737 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
738 .ndo_fdb_add = macvlan_fdb_add,
739 .ndo_fdb_del = macvlan_fdb_del,
740 .ndo_fdb_dump = ndo_dflt_fdb_dump,
741 .ndo_get_lock_subclass = macvlan_get_nest_level,
744 void macvlan_common_setup(struct net_device *dev)
748 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
749 dev->priv_flags |= IFF_UNICAST_FLT;
750 dev->netdev_ops = &macvlan_netdev_ops;
751 dev->destructor = free_netdev;
752 dev->header_ops = &macvlan_hard_header_ops;
753 dev->ethtool_ops = &macvlan_ethtool_ops;
755 EXPORT_SYMBOL_GPL(macvlan_common_setup);
757 static void macvlan_setup(struct net_device *dev)
759 macvlan_common_setup(dev);
760 dev->tx_queue_len = 0;
763 static int macvlan_port_create(struct net_device *dev)
765 struct macvlan_port *port;
769 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
772 port = kzalloc(sizeof(*port), GFP_KERNEL);
776 port->passthru = false;
778 INIT_LIST_HEAD(&port->vlans);
779 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
780 INIT_HLIST_HEAD(&port->vlan_hash[i]);
782 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
786 dev->priv_flags |= IFF_MACVLAN_PORT;
790 static void macvlan_port_destroy(struct net_device *dev)
792 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
794 dev->priv_flags &= ~IFF_MACVLAN_PORT;
795 netdev_rx_handler_unregister(dev);
796 kfree_rcu(port, rcu);
799 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
801 if (tb[IFLA_ADDRESS]) {
802 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
804 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
805 return -EADDRNOTAVAIL;
808 if (data && data[IFLA_MACVLAN_FLAGS] &&
809 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
812 if (data && data[IFLA_MACVLAN_MODE]) {
813 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
814 case MACVLAN_MODE_PRIVATE:
815 case MACVLAN_MODE_VEPA:
816 case MACVLAN_MODE_BRIDGE:
817 case MACVLAN_MODE_PASSTHRU:
826 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
827 struct nlattr *tb[], struct nlattr *data[])
829 struct macvlan_dev *vlan = netdev_priv(dev);
830 struct macvlan_port *port;
831 struct net_device *lowerdev;
837 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
838 if (lowerdev == NULL)
841 /* When creating macvlans or macvtaps on top of other macvlans - use
842 * the real device as the lowerdev.
844 if (netif_is_macvlan(lowerdev))
845 lowerdev = macvlan_dev_real_dev(lowerdev);
848 dev->mtu = lowerdev->mtu;
849 else if (dev->mtu > lowerdev->mtu)
852 if (!tb[IFLA_ADDRESS])
853 eth_hw_addr_random(dev);
855 if (!macvlan_port_exists(lowerdev)) {
856 err = macvlan_port_create(lowerdev);
860 port = macvlan_port_get_rtnl(lowerdev);
862 /* Only 1 macvlan device can be created in passthru mode */
866 vlan->lowerdev = lowerdev;
869 vlan->set_features = MACVLAN_FEATURES;
870 vlan->nest_level = dev_get_nest_level(lowerdev, netif_is_macvlan) + 1;
872 vlan->mode = MACVLAN_MODE_VEPA;
873 if (data && data[IFLA_MACVLAN_MODE])
874 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
876 if (data && data[IFLA_MACVLAN_FLAGS])
877 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
879 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
882 port->passthru = true;
883 eth_hw_addr_inherit(dev, lowerdev);
887 err = register_netdevice(dev);
891 dev->priv_flags |= IFF_MACVLAN;
892 err = netdev_upper_dev_link(lowerdev, dev);
894 goto unregister_netdev;
896 list_add_tail_rcu(&vlan->list, &port->vlans);
897 netif_stacked_transfer_operstate(lowerdev, dev);
902 unregister_netdevice(dev);
906 macvlan_port_destroy(lowerdev);
910 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
912 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
913 struct nlattr *tb[], struct nlattr *data[])
915 return macvlan_common_newlink(src_net, dev, tb, data);
918 void macvlan_dellink(struct net_device *dev, struct list_head *head)
920 struct macvlan_dev *vlan = netdev_priv(dev);
922 list_del_rcu(&vlan->list);
923 unregister_netdevice_queue(dev, head);
924 netdev_upper_dev_unlink(vlan->lowerdev, dev);
926 EXPORT_SYMBOL_GPL(macvlan_dellink);
928 static int macvlan_changelink(struct net_device *dev,
929 struct nlattr *tb[], struct nlattr *data[])
931 struct macvlan_dev *vlan = netdev_priv(dev);
932 enum macvlan_mode mode;
933 bool set_mode = false;
935 /* Validate mode, but don't set yet: setting flags may fail. */
936 if (data && data[IFLA_MACVLAN_MODE]) {
938 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
939 /* Passthrough mode can't be set or cleared dynamically */
940 if ((mode == MACVLAN_MODE_PASSTHRU) !=
941 (vlan->mode == MACVLAN_MODE_PASSTHRU))
945 if (data && data[IFLA_MACVLAN_FLAGS]) {
946 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
947 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
948 if (vlan->port->passthru && promisc) {
951 if (flags & MACVLAN_FLAG_NOPROMISC)
952 err = dev_set_promiscuity(vlan->lowerdev, -1);
954 err = dev_set_promiscuity(vlan->lowerdev, 1);
965 static size_t macvlan_get_size(const struct net_device *dev)
968 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
969 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
973 static int macvlan_fill_info(struct sk_buff *skb,
974 const struct net_device *dev)
976 struct macvlan_dev *vlan = netdev_priv(dev);
978 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
979 goto nla_put_failure;
980 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
981 goto nla_put_failure;
988 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
989 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
990 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
993 int macvlan_link_register(struct rtnl_link_ops *ops)
996 ops->priv_size = sizeof(struct macvlan_dev);
997 ops->validate = macvlan_validate;
998 ops->maxtype = IFLA_MACVLAN_MAX;
999 ops->policy = macvlan_policy;
1000 ops->changelink = macvlan_changelink;
1001 ops->get_size = macvlan_get_size;
1002 ops->fill_info = macvlan_fill_info;
1004 return rtnl_link_register(ops);
1006 EXPORT_SYMBOL_GPL(macvlan_link_register);
1008 static struct rtnl_link_ops macvlan_link_ops = {
1010 .setup = macvlan_setup,
1011 .newlink = macvlan_newlink,
1012 .dellink = macvlan_dellink,
1015 static int macvlan_device_event(struct notifier_block *unused,
1016 unsigned long event, void *ptr)
1018 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1019 struct macvlan_dev *vlan, *next;
1020 struct macvlan_port *port;
1021 LIST_HEAD(list_kill);
1023 if (!macvlan_port_exists(dev))
1026 port = macvlan_port_get_rtnl(dev);
1030 list_for_each_entry(vlan, &port->vlans, list)
1031 netif_stacked_transfer_operstate(vlan->lowerdev,
1034 case NETDEV_FEAT_CHANGE:
1035 list_for_each_entry(vlan, &port->vlans, list) {
1036 vlan->dev->gso_max_size = dev->gso_max_size;
1037 netdev_update_features(vlan->dev);
1040 case NETDEV_UNREGISTER:
1041 /* twiddle thumbs on netns device moves */
1042 if (dev->reg_state != NETREG_UNREGISTERING)
1045 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1046 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1047 unregister_netdevice_many(&list_kill);
1049 case NETDEV_PRE_TYPE_CHANGE:
1050 /* Forbid underlaying device to change its type. */
1056 static struct notifier_block macvlan_notifier_block __read_mostly = {
1057 .notifier_call = macvlan_device_event,
1060 static int __init macvlan_init_module(void)
1064 register_netdevice_notifier(&macvlan_notifier_block);
1066 err = macvlan_link_register(&macvlan_link_ops);
1071 unregister_netdevice_notifier(&macvlan_notifier_block);
1075 static void __exit macvlan_cleanup_module(void)
1077 rtnl_link_unregister(&macvlan_link_ops);
1078 unregister_netdevice_notifier(&macvlan_notifier_block);
1081 module_init(macvlan_init_module);
1082 module_exit(macvlan_cleanup_module);
1084 MODULE_LICENSE("GPL");
1085 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1086 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1087 MODULE_ALIAS_RTNL_LINK("macvlan");