Merge branch 'for-linus' of git://git.infradead.org/users/eparis/selinux into for...
[pandora-kernel.git] / drivers / net / macvlan.c
1 /*
2  * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
3  *
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.
8  *
9  * The code this is based on carried the following copyright notice:
10  * ---
11  * (C) Copyright 2001-2006
12  * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
13  * Re-worked by Ben Greear <greearb@candelatech.com>
14  * ---
15  */
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_link.h>
30 #include <linux/if_macvlan.h>
31 #include <net/rtnetlink.h>
32 #include <net/xfrm.h>
33
34 #define MACVLAN_HASH_SIZE       (1 << BITS_PER_BYTE)
35
36 struct macvlan_port {
37         struct net_device       *dev;
38         struct hlist_head       vlan_hash[MACVLAN_HASH_SIZE];
39         struct list_head        vlans;
40         struct rcu_head         rcu;
41         bool                    passthru;
42         int                     count;
43 };
44
45 static void macvlan_port_destroy(struct net_device *dev);
46
47 #define macvlan_port_get_rcu(dev) \
48         ((struct macvlan_port *) rcu_dereference(dev->rx_handler_data))
49 #define macvlan_port_get(dev) ((struct macvlan_port *) dev->rx_handler_data)
50 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
51
52 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
53                                                const unsigned char *addr)
54 {
55         struct macvlan_dev *vlan;
56         struct hlist_node *n;
57
58         hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[addr[5]], hlist) {
59                 if (!compare_ether_addr_64bits(vlan->dev->dev_addr, addr))
60                         return vlan;
61         }
62         return NULL;
63 }
64
65 static void macvlan_hash_add(struct macvlan_dev *vlan)
66 {
67         struct macvlan_port *port = vlan->port;
68         const unsigned char *addr = vlan->dev->dev_addr;
69
70         hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
71 }
72
73 static void macvlan_hash_del(struct macvlan_dev *vlan)
74 {
75         hlist_del_rcu(&vlan->hlist);
76         synchronize_rcu();
77 }
78
79 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
80                                         const unsigned char *addr)
81 {
82         macvlan_hash_del(vlan);
83         /* Now that we are unhashed it is safe to change the device
84          * address without confusing packet delivery.
85          */
86         memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
87         macvlan_hash_add(vlan);
88 }
89
90 static int macvlan_addr_busy(const struct macvlan_port *port,
91                                 const unsigned char *addr)
92 {
93         /* Test to see if the specified multicast address is
94          * currently in use by the underlying device or
95          * another macvlan.
96          */
97         if (!compare_ether_addr_64bits(port->dev->dev_addr, addr))
98                 return 1;
99
100         if (macvlan_hash_lookup(port, addr))
101                 return 1;
102
103         return 0;
104 }
105
106
107 static int macvlan_broadcast_one(struct sk_buff *skb,
108                                  const struct macvlan_dev *vlan,
109                                  const struct ethhdr *eth, bool local)
110 {
111         struct net_device *dev = vlan->dev;
112         if (!skb)
113                 return NET_RX_DROP;
114
115         if (local)
116                 return vlan->forward(dev, skb);
117
118         skb->dev = dev;
119         if (!compare_ether_addr_64bits(eth->h_dest,
120                                        dev->broadcast))
121                 skb->pkt_type = PACKET_BROADCAST;
122         else
123                 skb->pkt_type = PACKET_MULTICAST;
124
125         return vlan->receive(skb);
126 }
127
128 static void macvlan_broadcast(struct sk_buff *skb,
129                               const struct macvlan_port *port,
130                               struct net_device *src,
131                               enum macvlan_mode mode)
132 {
133         const struct ethhdr *eth = eth_hdr(skb);
134         const struct macvlan_dev *vlan;
135         struct hlist_node *n;
136         struct sk_buff *nskb;
137         unsigned int i;
138         int err;
139
140         if (skb->protocol == htons(ETH_P_PAUSE))
141                 return;
142
143         for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
144                 hlist_for_each_entry_rcu(vlan, n, &port->vlan_hash[i], hlist) {
145                         if (vlan->dev == src || !(vlan->mode & mode))
146                                 continue;
147
148                         nskb = skb_clone(skb, GFP_ATOMIC);
149                         err = macvlan_broadcast_one(nskb, vlan, eth,
150                                          mode == MACVLAN_MODE_BRIDGE);
151                         macvlan_count_rx(vlan, skb->len + ETH_HLEN,
152                                          err == NET_RX_SUCCESS, 1);
153                 }
154         }
155 }
156
157 /* called under rcu_read_lock() from netif_receive_skb */
158 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
159 {
160         struct macvlan_port *port;
161         struct sk_buff *skb = *pskb;
162         const struct ethhdr *eth = eth_hdr(skb);
163         const struct macvlan_dev *vlan;
164         const struct macvlan_dev *src;
165         struct net_device *dev;
166         unsigned int len = 0;
167         int ret = NET_RX_DROP;
168
169         port = macvlan_port_get_rcu(skb->dev);
170         if (is_multicast_ether_addr(eth->h_dest)) {
171                 src = macvlan_hash_lookup(port, eth->h_source);
172                 if (!src)
173                         /* frame comes from an external address */
174                         macvlan_broadcast(skb, port, NULL,
175                                           MACVLAN_MODE_PRIVATE |
176                                           MACVLAN_MODE_VEPA    |
177                                           MACVLAN_MODE_PASSTHRU|
178                                           MACVLAN_MODE_BRIDGE);
179                 else if (src->mode == MACVLAN_MODE_VEPA)
180                         /* flood to everyone except source */
181                         macvlan_broadcast(skb, port, src->dev,
182                                           MACVLAN_MODE_VEPA |
183                                           MACVLAN_MODE_BRIDGE);
184                 else if (src->mode == MACVLAN_MODE_BRIDGE)
185                         /*
186                          * flood only to VEPA ports, bridge ports
187                          * already saw the frame on the way out.
188                          */
189                         macvlan_broadcast(skb, port, src->dev,
190                                           MACVLAN_MODE_VEPA);
191                 return RX_HANDLER_PASS;
192         }
193
194         if (port->passthru)
195                 vlan = list_first_entry(&port->vlans, struct macvlan_dev, list);
196         else
197                 vlan = macvlan_hash_lookup(port, eth->h_dest);
198         if (vlan == NULL)
199                 return RX_HANDLER_PASS;
200
201         dev = vlan->dev;
202         if (unlikely(!(dev->flags & IFF_UP))) {
203                 kfree_skb(skb);
204                 return RX_HANDLER_CONSUMED;
205         }
206         len = skb->len + ETH_HLEN;
207         skb = skb_share_check(skb, GFP_ATOMIC);
208         if (!skb)
209                 goto out;
210
211         skb->dev = dev;
212         skb->pkt_type = PACKET_HOST;
213
214         ret = vlan->receive(skb);
215
216 out:
217         macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
218         return RX_HANDLER_CONSUMED;
219 }
220
221 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
222 {
223         const struct macvlan_dev *vlan = netdev_priv(dev);
224         const struct macvlan_port *port = vlan->port;
225         const struct macvlan_dev *dest;
226         __u8 ip_summed = skb->ip_summed;
227
228         if (vlan->mode == MACVLAN_MODE_BRIDGE) {
229                 const struct ethhdr *eth = (void *)skb->data;
230                 skb->ip_summed = CHECKSUM_UNNECESSARY;
231
232                 /* send to other bridge ports directly */
233                 if (is_multicast_ether_addr(eth->h_dest)) {
234                         macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
235                         goto xmit_world;
236                 }
237
238                 dest = macvlan_hash_lookup(port, eth->h_dest);
239                 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
240                         unsigned int length = skb->len + ETH_HLEN;
241                         int ret = dest->forward(dest->dev, skb);
242                         macvlan_count_rx(dest, length,
243                                          ret == NET_RX_SUCCESS, 0);
244
245                         return NET_XMIT_SUCCESS;
246                 }
247         }
248
249 xmit_world:
250         skb->ip_summed = ip_summed;
251         skb_set_dev(skb, vlan->lowerdev);
252         return dev_queue_xmit(skb);
253 }
254
255 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
256                                struct net_device *dev)
257 {
258         unsigned int len = skb->len;
259         int ret;
260         const struct macvlan_dev *vlan = netdev_priv(dev);
261
262         ret = macvlan_queue_xmit(skb, dev);
263         if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
264                 struct macvlan_pcpu_stats *pcpu_stats;
265
266                 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
267                 u64_stats_update_begin(&pcpu_stats->syncp);
268                 pcpu_stats->tx_packets++;
269                 pcpu_stats->tx_bytes += len;
270                 u64_stats_update_end(&pcpu_stats->syncp);
271         } else {
272                 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
273         }
274         return ret;
275 }
276 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
277
278 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
279                                unsigned short type, const void *daddr,
280                                const void *saddr, unsigned len)
281 {
282         const struct macvlan_dev *vlan = netdev_priv(dev);
283         struct net_device *lowerdev = vlan->lowerdev;
284
285         return dev_hard_header(skb, lowerdev, type, daddr,
286                                saddr ? : dev->dev_addr, len);
287 }
288
289 static const struct header_ops macvlan_hard_header_ops = {
290         .create         = macvlan_hard_header,
291         .rebuild        = eth_rebuild_header,
292         .parse          = eth_header_parse,
293         .cache          = eth_header_cache,
294         .cache_update   = eth_header_cache_update,
295 };
296
297 static int macvlan_open(struct net_device *dev)
298 {
299         struct macvlan_dev *vlan = netdev_priv(dev);
300         struct net_device *lowerdev = vlan->lowerdev;
301         int err;
302
303         if (vlan->port->passthru) {
304                 dev_set_promiscuity(lowerdev, 1);
305                 goto hash_add;
306         }
307
308         err = -EBUSY;
309         if (macvlan_addr_busy(vlan->port, dev->dev_addr))
310                 goto out;
311
312         err = dev_uc_add(lowerdev, dev->dev_addr);
313         if (err < 0)
314                 goto out;
315         if (dev->flags & IFF_ALLMULTI) {
316                 err = dev_set_allmulti(lowerdev, 1);
317                 if (err < 0)
318                         goto del_unicast;
319         }
320
321 hash_add:
322         macvlan_hash_add(vlan);
323         return 0;
324
325 del_unicast:
326         dev_uc_del(lowerdev, dev->dev_addr);
327 out:
328         return err;
329 }
330
331 static int macvlan_stop(struct net_device *dev)
332 {
333         struct macvlan_dev *vlan = netdev_priv(dev);
334         struct net_device *lowerdev = vlan->lowerdev;
335
336         if (vlan->port->passthru) {
337                 dev_set_promiscuity(lowerdev, -1);
338                 goto hash_del;
339         }
340
341         dev_mc_unsync(lowerdev, dev);
342         if (dev->flags & IFF_ALLMULTI)
343                 dev_set_allmulti(lowerdev, -1);
344
345         dev_uc_del(lowerdev, dev->dev_addr);
346
347 hash_del:
348         macvlan_hash_del(vlan);
349         return 0;
350 }
351
352 static int macvlan_set_mac_address(struct net_device *dev, void *p)
353 {
354         struct macvlan_dev *vlan = netdev_priv(dev);
355         struct net_device *lowerdev = vlan->lowerdev;
356         struct sockaddr *addr = p;
357         int err;
358
359         if (!is_valid_ether_addr(addr->sa_data))
360                 return -EADDRNOTAVAIL;
361
362         if (!(dev->flags & IFF_UP)) {
363                 /* Just copy in the new address */
364                 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
365         } else {
366                 /* Rehash and update the device filters */
367                 if (macvlan_addr_busy(vlan->port, addr->sa_data))
368                         return -EBUSY;
369
370                 err = dev_uc_add(lowerdev, addr->sa_data);
371                 if (err)
372                         return err;
373
374                 dev_uc_del(lowerdev, dev->dev_addr);
375
376                 macvlan_hash_change_addr(vlan, addr->sa_data);
377         }
378         return 0;
379 }
380
381 static void macvlan_change_rx_flags(struct net_device *dev, int change)
382 {
383         struct macvlan_dev *vlan = netdev_priv(dev);
384         struct net_device *lowerdev = vlan->lowerdev;
385
386         if (change & IFF_ALLMULTI)
387                 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
388 }
389
390 static void macvlan_set_multicast_list(struct net_device *dev)
391 {
392         struct macvlan_dev *vlan = netdev_priv(dev);
393
394         dev_mc_sync(vlan->lowerdev, dev);
395 }
396
397 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
398 {
399         struct macvlan_dev *vlan = netdev_priv(dev);
400
401         if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
402                 return -EINVAL;
403         dev->mtu = new_mtu;
404         return 0;
405 }
406
407 /*
408  * macvlan network devices have devices nesting below it and are a special
409  * "super class" of normal network devices; split their locks off into a
410  * separate class since they always nest.
411  */
412 static struct lock_class_key macvlan_netdev_xmit_lock_key;
413 static struct lock_class_key macvlan_netdev_addr_lock_key;
414
415 #define MACVLAN_FEATURES \
416         (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
417          NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
418          NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO)
419
420 #define MACVLAN_STATE_MASK \
421         ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
422
423 static void macvlan_set_lockdep_class_one(struct net_device *dev,
424                                           struct netdev_queue *txq,
425                                           void *_unused)
426 {
427         lockdep_set_class(&txq->_xmit_lock,
428                           &macvlan_netdev_xmit_lock_key);
429 }
430
431 static void macvlan_set_lockdep_class(struct net_device *dev)
432 {
433         lockdep_set_class(&dev->addr_list_lock,
434                           &macvlan_netdev_addr_lock_key);
435         netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
436 }
437
438 static int macvlan_init(struct net_device *dev)
439 {
440         struct macvlan_dev *vlan = netdev_priv(dev);
441         const struct net_device *lowerdev = vlan->lowerdev;
442
443         dev->state              = (dev->state & ~MACVLAN_STATE_MASK) |
444                                   (lowerdev->state & MACVLAN_STATE_MASK);
445         dev->features           = lowerdev->features & MACVLAN_FEATURES;
446         dev->features           |= NETIF_F_LLTX;
447         dev->gso_max_size       = lowerdev->gso_max_size;
448         dev->iflink             = lowerdev->ifindex;
449         dev->hard_header_len    = lowerdev->hard_header_len;
450
451         macvlan_set_lockdep_class(dev);
452
453         vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
454         if (!vlan->pcpu_stats)
455                 return -ENOMEM;
456
457         return 0;
458 }
459
460 static void macvlan_uninit(struct net_device *dev)
461 {
462         struct macvlan_dev *vlan = netdev_priv(dev);
463         struct macvlan_port *port = vlan->port;
464
465         free_percpu(vlan->pcpu_stats);
466
467         port->count -= 1;
468         if (!port->count)
469                 macvlan_port_destroy(port->dev);
470 }
471
472 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
473                                                          struct rtnl_link_stats64 *stats)
474 {
475         struct macvlan_dev *vlan = netdev_priv(dev);
476
477         if (vlan->pcpu_stats) {
478                 struct macvlan_pcpu_stats *p;
479                 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
480                 u32 rx_errors = 0, tx_dropped = 0;
481                 unsigned int start;
482                 int i;
483
484                 for_each_possible_cpu(i) {
485                         p = per_cpu_ptr(vlan->pcpu_stats, i);
486                         do {
487                                 start = u64_stats_fetch_begin_bh(&p->syncp);
488                                 rx_packets      = p->rx_packets;
489                                 rx_bytes        = p->rx_bytes;
490                                 rx_multicast    = p->rx_multicast;
491                                 tx_packets      = p->tx_packets;
492                                 tx_bytes        = p->tx_bytes;
493                         } while (u64_stats_fetch_retry_bh(&p->syncp, start));
494
495                         stats->rx_packets       += rx_packets;
496                         stats->rx_bytes         += rx_bytes;
497                         stats->multicast        += rx_multicast;
498                         stats->tx_packets       += tx_packets;
499                         stats->tx_bytes         += tx_bytes;
500                         /* rx_errors & tx_dropped are u32, updated
501                          * without syncp protection.
502                          */
503                         rx_errors       += p->rx_errors;
504                         tx_dropped      += p->tx_dropped;
505                 }
506                 stats->rx_errors        = rx_errors;
507                 stats->rx_dropped       = rx_errors;
508                 stats->tx_dropped       = tx_dropped;
509         }
510         return stats;
511 }
512
513 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
514                                         struct ethtool_drvinfo *drvinfo)
515 {
516         snprintf(drvinfo->driver, 32, "macvlan");
517         snprintf(drvinfo->version, 32, "0.1");
518 }
519
520 static u32 macvlan_ethtool_get_rx_csum(struct net_device *dev)
521 {
522         const struct macvlan_dev *vlan = netdev_priv(dev);
523         return dev_ethtool_get_rx_csum(vlan->lowerdev);
524 }
525
526 static int macvlan_ethtool_get_settings(struct net_device *dev,
527                                         struct ethtool_cmd *cmd)
528 {
529         const struct macvlan_dev *vlan = netdev_priv(dev);
530         return dev_ethtool_get_settings(vlan->lowerdev, cmd);
531 }
532
533 static u32 macvlan_ethtool_get_flags(struct net_device *dev)
534 {
535         const struct macvlan_dev *vlan = netdev_priv(dev);
536         return dev_ethtool_get_flags(vlan->lowerdev);
537 }
538
539 static const struct ethtool_ops macvlan_ethtool_ops = {
540         .get_link               = ethtool_op_get_link,
541         .get_settings           = macvlan_ethtool_get_settings,
542         .get_rx_csum            = macvlan_ethtool_get_rx_csum,
543         .get_drvinfo            = macvlan_ethtool_get_drvinfo,
544         .get_flags              = macvlan_ethtool_get_flags,
545 };
546
547 static const struct net_device_ops macvlan_netdev_ops = {
548         .ndo_init               = macvlan_init,
549         .ndo_uninit             = macvlan_uninit,
550         .ndo_open               = macvlan_open,
551         .ndo_stop               = macvlan_stop,
552         .ndo_start_xmit         = macvlan_start_xmit,
553         .ndo_change_mtu         = macvlan_change_mtu,
554         .ndo_change_rx_flags    = macvlan_change_rx_flags,
555         .ndo_set_mac_address    = macvlan_set_mac_address,
556         .ndo_set_multicast_list = macvlan_set_multicast_list,
557         .ndo_get_stats64        = macvlan_dev_get_stats64,
558         .ndo_validate_addr      = eth_validate_addr,
559 };
560
561 void macvlan_common_setup(struct net_device *dev)
562 {
563         ether_setup(dev);
564
565         dev->priv_flags        &= ~IFF_XMIT_DST_RELEASE;
566         dev->netdev_ops         = &macvlan_netdev_ops;
567         dev->destructor         = free_netdev;
568         dev->header_ops         = &macvlan_hard_header_ops,
569         dev->ethtool_ops        = &macvlan_ethtool_ops;
570 }
571 EXPORT_SYMBOL_GPL(macvlan_common_setup);
572
573 static void macvlan_setup(struct net_device *dev)
574 {
575         macvlan_common_setup(dev);
576         dev->tx_queue_len       = 0;
577 }
578
579 static int macvlan_port_create(struct net_device *dev)
580 {
581         struct macvlan_port *port;
582         unsigned int i;
583         int err;
584
585         if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
586                 return -EINVAL;
587
588         port = kzalloc(sizeof(*port), GFP_KERNEL);
589         if (port == NULL)
590                 return -ENOMEM;
591
592         port->passthru = false;
593         port->dev = dev;
594         INIT_LIST_HEAD(&port->vlans);
595         for (i = 0; i < MACVLAN_HASH_SIZE; i++)
596                 INIT_HLIST_HEAD(&port->vlan_hash[i]);
597
598         err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
599         if (err)
600                 kfree(port);
601
602         dev->priv_flags |= IFF_MACVLAN_PORT;
603         return err;
604 }
605
606 static void macvlan_port_rcu_free(struct rcu_head *head)
607 {
608         struct macvlan_port *port;
609
610         port = container_of(head, struct macvlan_port, rcu);
611         kfree(port);
612 }
613
614 static void macvlan_port_destroy(struct net_device *dev)
615 {
616         struct macvlan_port *port = macvlan_port_get(dev);
617
618         dev->priv_flags &= ~IFF_MACVLAN_PORT;
619         netdev_rx_handler_unregister(dev);
620         call_rcu(&port->rcu, macvlan_port_rcu_free);
621 }
622
623 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
624 {
625         if (tb[IFLA_ADDRESS]) {
626                 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
627                         return -EINVAL;
628                 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
629                         return -EADDRNOTAVAIL;
630         }
631
632         if (data && data[IFLA_MACVLAN_MODE]) {
633                 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
634                 case MACVLAN_MODE_PRIVATE:
635                 case MACVLAN_MODE_VEPA:
636                 case MACVLAN_MODE_BRIDGE:
637                 case MACVLAN_MODE_PASSTHRU:
638                         break;
639                 default:
640                         return -EINVAL;
641                 }
642         }
643         return 0;
644 }
645
646 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
647                            struct nlattr *tb[], struct nlattr *data[],
648                            int (*receive)(struct sk_buff *skb),
649                            int (*forward)(struct net_device *dev,
650                                           struct sk_buff *skb))
651 {
652         struct macvlan_dev *vlan = netdev_priv(dev);
653         struct macvlan_port *port;
654         struct net_device *lowerdev;
655         int err;
656
657         if (!tb[IFLA_LINK])
658                 return -EINVAL;
659
660         lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
661         if (lowerdev == NULL)
662                 return -ENODEV;
663
664         /* When creating macvlans on top of other macvlans - use
665          * the real device as the lowerdev.
666          */
667         if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
668                 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
669                 lowerdev = lowervlan->lowerdev;
670         }
671
672         if (!tb[IFLA_MTU])
673                 dev->mtu = lowerdev->mtu;
674         else if (dev->mtu > lowerdev->mtu)
675                 return -EINVAL;
676
677         if (!tb[IFLA_ADDRESS])
678                 random_ether_addr(dev->dev_addr);
679
680         if (!macvlan_port_exists(lowerdev)) {
681                 err = macvlan_port_create(lowerdev);
682                 if (err < 0)
683                         return err;
684         }
685         port = macvlan_port_get(lowerdev);
686
687         /* Only 1 macvlan device can be created in passthru mode */
688         if (port->passthru)
689                 return -EINVAL;
690
691         vlan->lowerdev = lowerdev;
692         vlan->dev      = dev;
693         vlan->port     = port;
694         vlan->receive  = receive;
695         vlan->forward  = forward;
696
697         vlan->mode     = MACVLAN_MODE_VEPA;
698         if (data && data[IFLA_MACVLAN_MODE])
699                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
700
701         if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
702                 if (port->count)
703                         return -EINVAL;
704                 port->passthru = true;
705                 memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN);
706         }
707
708         port->count += 1;
709         err = register_netdevice(dev);
710         if (err < 0)
711                 goto destroy_port;
712
713         list_add_tail(&vlan->list, &port->vlans);
714         netif_stacked_transfer_operstate(lowerdev, dev);
715
716         return 0;
717
718 destroy_port:
719         port->count -= 1;
720         if (!port->count)
721                 macvlan_port_destroy(lowerdev);
722
723         return err;
724 }
725 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
726
727 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
728                            struct nlattr *tb[], struct nlattr *data[])
729 {
730         return macvlan_common_newlink(src_net, dev, tb, data,
731                                       netif_rx,
732                                       dev_forward_skb);
733 }
734
735 void macvlan_dellink(struct net_device *dev, struct list_head *head)
736 {
737         struct macvlan_dev *vlan = netdev_priv(dev);
738
739         list_del(&vlan->list);
740         unregister_netdevice_queue(dev, head);
741 }
742 EXPORT_SYMBOL_GPL(macvlan_dellink);
743
744 static int macvlan_changelink(struct net_device *dev,
745                 struct nlattr *tb[], struct nlattr *data[])
746 {
747         struct macvlan_dev *vlan = netdev_priv(dev);
748         if (data && data[IFLA_MACVLAN_MODE])
749                 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
750         return 0;
751 }
752
753 static size_t macvlan_get_size(const struct net_device *dev)
754 {
755         return nla_total_size(4);
756 }
757
758 static int macvlan_fill_info(struct sk_buff *skb,
759                                 const struct net_device *dev)
760 {
761         struct macvlan_dev *vlan = netdev_priv(dev);
762
763         NLA_PUT_U32(skb, IFLA_MACVLAN_MODE, vlan->mode);
764         return 0;
765
766 nla_put_failure:
767         return -EMSGSIZE;
768 }
769
770 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
771         [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
772 };
773
774 int macvlan_link_register(struct rtnl_link_ops *ops)
775 {
776         /* common fields */
777         ops->priv_size          = sizeof(struct macvlan_dev);
778         ops->validate           = macvlan_validate;
779         ops->maxtype            = IFLA_MACVLAN_MAX;
780         ops->policy             = macvlan_policy;
781         ops->changelink         = macvlan_changelink;
782         ops->get_size           = macvlan_get_size;
783         ops->fill_info          = macvlan_fill_info;
784
785         return rtnl_link_register(ops);
786 };
787 EXPORT_SYMBOL_GPL(macvlan_link_register);
788
789 static struct rtnl_link_ops macvlan_link_ops = {
790         .kind           = "macvlan",
791         .setup          = macvlan_setup,
792         .newlink        = macvlan_newlink,
793         .dellink        = macvlan_dellink,
794 };
795
796 static int macvlan_device_event(struct notifier_block *unused,
797                                 unsigned long event, void *ptr)
798 {
799         struct net_device *dev = ptr;
800         struct macvlan_dev *vlan, *next;
801         struct macvlan_port *port;
802
803         if (!macvlan_port_exists(dev))
804                 return NOTIFY_DONE;
805
806         port = macvlan_port_get(dev);
807
808         switch (event) {
809         case NETDEV_CHANGE:
810                 list_for_each_entry(vlan, &port->vlans, list)
811                         netif_stacked_transfer_operstate(vlan->lowerdev,
812                                                          vlan->dev);
813                 break;
814         case NETDEV_FEAT_CHANGE:
815                 list_for_each_entry(vlan, &port->vlans, list) {
816                         vlan->dev->features = dev->features & MACVLAN_FEATURES;
817                         vlan->dev->gso_max_size = dev->gso_max_size;
818                         netdev_features_change(vlan->dev);
819                 }
820                 break;
821         case NETDEV_UNREGISTER:
822                 /* twiddle thumbs on netns device moves */
823                 if (dev->reg_state != NETREG_UNREGISTERING)
824                         break;
825
826                 list_for_each_entry_safe(vlan, next, &port->vlans, list)
827                         vlan->dev->rtnl_link_ops->dellink(vlan->dev, NULL);
828                 break;
829         case NETDEV_PRE_TYPE_CHANGE:
830                 /* Forbid underlaying device to change its type. */
831                 return NOTIFY_BAD;
832         }
833         return NOTIFY_DONE;
834 }
835
836 static struct notifier_block macvlan_notifier_block __read_mostly = {
837         .notifier_call  = macvlan_device_event,
838 };
839
840 static int __init macvlan_init_module(void)
841 {
842         int err;
843
844         register_netdevice_notifier(&macvlan_notifier_block);
845
846         err = macvlan_link_register(&macvlan_link_ops);
847         if (err < 0)
848                 goto err1;
849         return 0;
850 err1:
851         unregister_netdevice_notifier(&macvlan_notifier_block);
852         return err;
853 }
854
855 static void __exit macvlan_cleanup_module(void)
856 {
857         rtnl_link_unregister(&macvlan_link_ops);
858         unregister_netdevice_notifier(&macvlan_notifier_block);
859 }
860
861 module_init(macvlan_init_module);
862 module_exit(macvlan_cleanup_module);
863
864 MODULE_LICENSE("GPL");
865 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
866 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
867 MODULE_ALIAS_RTNL_LINK("macvlan");