?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'random_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso...
[linux-2.6.git] / drivers / net / macvlan.c
blob9bf46bd19b87bf947fd79e547bdde626c8a12097
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_vlan.h>
30 #include <linux/if_link.h>
31 #include <linux/if_macvlan.h>
32 #include <linux/hash.h>
33 #include <net/rtnetlink.h>
34 #include <net/xfrm.h>
35
36 #define MACVLAN_HASH_SIZE (1 << BITS_PER_BYTE)
37
38 struct macvlan_port {
39 struct net_device *dev;
40 struct hlist_head vlan_hash[MACVLAN_HASH_SIZE];
41 struct list_head vlans;
42 struct rcu_head rcu;
43 bool passthru;
44 int count;
45 };
46
47 static void macvlan_port_destroy(struct net_device *dev);
48
49 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
50 {
51 return rcu_dereference(dev->rx_handler_data);
52 }
53
54 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
55 {
56 return rtnl_dereference(dev->rx_handler_data);
57 }
58
59 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
60
61 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
62 const unsigned char *addr)
63 {
64 struct macvlan_dev *vlan;
65
66 hlist_for_each_entry_rcu(vlan, &port->vlan_hash[addr[5]], hlist) {
67 if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
68 return vlan;
69 }
70 return NULL;
71 }
72
73 static void macvlan_hash_add(struct macvlan_dev *vlan)
74 {
75 struct macvlan_port *port = vlan->port;
76 const unsigned char *addr = vlan->dev->dev_addr;
77
78 hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[addr[5]]);
79 }
80
81 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
82 {
83 hlist_del_rcu(&vlan->hlist);
84 if (sync)
85 synchronize_rcu();
86 }
87
88 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
89 const unsigned char *addr)
90 {
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.
94 */
95 memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
96 macvlan_hash_add(vlan);
97 }
98
99 static int macvlan_addr_busy(const struct macvlan_port *port,
100 const unsigned char *addr)
101 {
102 /* Test to see if the specified multicast address is
103 * currently in use by the underlying device or
104 * another macvlan.
105 */
106 if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
107 return 1;
108
109 if (macvlan_hash_lookup(port, addr))
110 return 1;
111
112 return 0;
113 }
114
115
116 static int macvlan_broadcast_one(struct sk_buff *skb,
117 const struct macvlan_dev *vlan,
118 const struct ethhdr *eth, bool local)
119 {
120 struct net_device *dev = vlan->dev;
121
122 if (local)
123 return vlan->forward(dev, skb);
124
125 skb->dev = dev;
126 if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
127 skb->pkt_type = PACKET_BROADCAST;
128 else
129 skb->pkt_type = PACKET_MULTICAST;
130
131 return vlan->receive(skb);
132 }
133
134 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
135 {
136 return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
137 }
138
139
140 static unsigned int mc_hash(const struct macvlan_dev *vlan,
141 const unsigned char *addr)
142 {
143 u32 val = __get_unaligned_cpu32(addr + 2);
144
145 val ^= macvlan_hash_mix(vlan);
146 return hash_32(val, MACVLAN_MC_FILTER_BITS);
147 }
148
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)
153 {
154 const struct ethhdr *eth = eth_hdr(skb);
155 const struct macvlan_dev *vlan;
156 struct sk_buff *nskb;
157 unsigned int i;
158 int err;
159 unsigned int hash;
160
161 if (skb->protocol == htons(ETH_P_PAUSE))
162 return;
163
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))
167 continue;
168
169 hash = mc_hash(vlan, eth->h_dest);
170 if (!test_bit(hash, vlan->mc_filter))
171 continue;
172
173 err = NET_RX_DROP;
174 nskb = skb_clone(skb, GFP_ATOMIC);
175 if (likely(nskb))
176 err = macvlan_broadcast_one(
177 nskb, vlan, eth,
178 mode == MACVLAN_MODE_BRIDGE);
179 macvlan_count_rx(vlan, skb->len + ETH_HLEN,
180 err == NET_RX_SUCCESS, 1);
181 }
182 }
183 }
184
185 /* called under rcu_read_lock() from netif_receive_skb */
186 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
187 {
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;
196
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);
200 if (!skb)
201 return RX_HANDLER_CONSUMED;
202 eth = eth_hdr(skb);
203 src = macvlan_hash_lookup(port, eth->h_source);
204 if (!src)
205 /* frame comes from an external address */
206 macvlan_broadcast(skb, port, NULL,
207 MACVLAN_MODE_PRIVATE |
208 MACVLAN_MODE_VEPA |
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,
214 MACVLAN_MODE_VEPA |
215 MACVLAN_MODE_BRIDGE);
216 else if (src->mode == MACVLAN_MODE_BRIDGE)
217 /*
218 * flood only to VEPA ports, bridge ports
219 * already saw the frame on the way out.
220 */
221 macvlan_broadcast(skb, port, src->dev,
222 MACVLAN_MODE_VEPA);
223 else {
224 /* forward to original port. */
225 vlan = src;
226 ret = macvlan_broadcast_one(skb, vlan, eth, 0);
227 goto out;
228 }
229
230 return RX_HANDLER_PASS;
231 }
232
233 if (port->passthru)
234 vlan = list_first_or_null_rcu(&port->vlans,
235 struct macvlan_dev, list);
236 else
237 vlan = macvlan_hash_lookup(port, eth->h_dest);
238 if (vlan == NULL)
239 return RX_HANDLER_PASS;
240
241 dev = vlan->dev;
242 if (unlikely(!(dev->flags & IFF_UP))) {
243 kfree_skb(skb);
244 return RX_HANDLER_CONSUMED;
245 }
246 len = skb->len + ETH_HLEN;
247 skb = skb_share_check(skb, GFP_ATOMIC);
248 if (!skb)
249 goto out;
250
251 skb->dev = dev;
252 skb->pkt_type = PACKET_HOST;
253
254 ret = vlan->receive(skb);
255
256 out:
257 macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
258 return RX_HANDLER_CONSUMED;
259 }
260
261 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
262 {
263 const struct macvlan_dev *vlan = netdev_priv(dev);
264 const struct macvlan_port *port = vlan->port;
265 const struct macvlan_dev *dest;
266 __u8 ip_summed = skb->ip_summed;
267
268 if (vlan->mode == MACVLAN_MODE_BRIDGE) {
269 const struct ethhdr *eth = (void *)skb->data;
270 skb->ip_summed = CHECKSUM_UNNECESSARY;
271
272 /* send to other bridge ports directly */
273 if (is_multicast_ether_addr(eth->h_dest)) {
274 macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
275 goto xmit_world;
276 }
277
278 dest = macvlan_hash_lookup(port, eth->h_dest);
279 if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
280 /* send to lowerdev first for its network taps */
281 dev_forward_skb(vlan->lowerdev, skb);
282
283 return NET_XMIT_SUCCESS;
284 }
285 }
286
287 xmit_world:
288 skb->ip_summed = ip_summed;
289 skb->dev = vlan->lowerdev;
290 return dev_queue_xmit(skb);
291 }
292
293 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
294 struct net_device *dev)
295 {
296 unsigned int len = skb->len;
297 int ret;
298 const struct macvlan_dev *vlan = netdev_priv(dev);
299
300 ret = macvlan_queue_xmit(skb, dev);
301 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
302 struct macvlan_pcpu_stats *pcpu_stats;
303
304 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
305 u64_stats_update_begin(&pcpu_stats->syncp);
306 pcpu_stats->tx_packets++;
307 pcpu_stats->tx_bytes += len;
308 u64_stats_update_end(&pcpu_stats->syncp);
309 } else {
310 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
311 }
312 return ret;
313 }
314 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
315
316 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
317 unsigned short type, const void *daddr,
318 const void *saddr, unsigned len)
319 {
320 const struct macvlan_dev *vlan = netdev_priv(dev);
321 struct net_device *lowerdev = vlan->lowerdev;
322
323 return dev_hard_header(skb, lowerdev, type, daddr,
324 saddr ? : dev->dev_addr, len);
325 }
326
327 static const struct header_ops macvlan_hard_header_ops = {
328 .create = macvlan_hard_header,
329 .rebuild = eth_rebuild_header,
330 .parse = eth_header_parse,
331 .cache = eth_header_cache,
332 .cache_update = eth_header_cache_update,
333 };
334
335 static int macvlan_open(struct net_device *dev)
336 {
337 struct macvlan_dev *vlan = netdev_priv(dev);
338 struct net_device *lowerdev = vlan->lowerdev;
339 int err;
340
341 if (vlan->port->passthru) {
342 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
343 err = dev_set_promiscuity(lowerdev, 1);
344 if (err < 0)
345 goto out;
346 }
347 goto hash_add;
348 }
349
350 err = -EBUSY;
351 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
352 goto out;
353
354 err = dev_uc_add(lowerdev, dev->dev_addr);
355 if (err < 0)
356 goto out;
357 if (dev->flags & IFF_ALLMULTI) {
358 err = dev_set_allmulti(lowerdev, 1);
359 if (err < 0)
360 goto del_unicast;
361 }
362
363 hash_add:
364 macvlan_hash_add(vlan);
365 return 0;
366
367 del_unicast:
368 dev_uc_del(lowerdev, dev->dev_addr);
369 out:
370 return err;
371 }
372
373 static int macvlan_stop(struct net_device *dev)
374 {
375 struct macvlan_dev *vlan = netdev_priv(dev);
376 struct net_device *lowerdev = vlan->lowerdev;
377
378 dev_uc_unsync(lowerdev, dev);
379 dev_mc_unsync(lowerdev, dev);
380
381 if (vlan->port->passthru) {
382 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
383 dev_set_promiscuity(lowerdev, -1);
384 goto hash_del;
385 }
386
387 if (dev->flags & IFF_ALLMULTI)
388 dev_set_allmulti(lowerdev, -1);
389
390 dev_uc_del(lowerdev, dev->dev_addr);
391
392 hash_del:
393 macvlan_hash_del(vlan, !dev->dismantle);
394 return 0;
395 }
396
397 static int macvlan_set_mac_address(struct net_device *dev, void *p)
398 {
399 struct macvlan_dev *vlan = netdev_priv(dev);
400 struct net_device *lowerdev = vlan->lowerdev;
401 struct sockaddr *addr = p;
402 int err;
403
404 if (!is_valid_ether_addr(addr->sa_data))
405 return -EADDRNOTAVAIL;
406
407 if (!(dev->flags & IFF_UP)) {
408 /* Just copy in the new address */
409 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
410 } else {
411 /* Rehash and update the device filters */
412 if (macvlan_addr_busy(vlan->port, addr->sa_data))
413 return -EBUSY;
414
415 err = dev_uc_add(lowerdev, addr->sa_data);
416 if (err)
417 return err;
418
419 dev_uc_del(lowerdev, dev->dev_addr);
420
421 macvlan_hash_change_addr(vlan, addr->sa_data);
422 }
423 return 0;
424 }
425
426 static void macvlan_change_rx_flags(struct net_device *dev, int change)
427 {
428 struct macvlan_dev *vlan = netdev_priv(dev);
429 struct net_device *lowerdev = vlan->lowerdev;
430
431 if (change & IFF_ALLMULTI)
432 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
433 }
434
435 static void macvlan_set_mac_lists(struct net_device *dev)
436 {
437 struct macvlan_dev *vlan = netdev_priv(dev);
438
439 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
440 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
441 } else {
442 struct netdev_hw_addr *ha;
443 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
444
445 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
446 netdev_for_each_mc_addr(ha, dev) {
447 __set_bit(mc_hash(vlan, ha->addr), filter);
448 }
449
450 __set_bit(mc_hash(vlan, dev->broadcast), filter);
451
452 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
453 }
454 dev_uc_sync(vlan->lowerdev, dev);
455 dev_mc_sync(vlan->lowerdev, dev);
456 }
457
458 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
459 {
460 struct macvlan_dev *vlan = netdev_priv(dev);
461
462 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
463 return -EINVAL;
464 dev->mtu = new_mtu;
465 return 0;
466 }
467
468 /*
469 * macvlan network devices have devices nesting below it and are a special
470 * "super class" of normal network devices; split their locks off into a
471 * separate class since they always nest.
472 */
473 static struct lock_class_key macvlan_netdev_xmit_lock_key;
474 static struct lock_class_key macvlan_netdev_addr_lock_key;
475
476 #define MACVLAN_FEATURES \
477 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
478 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
479 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
480 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
481
482 #define MACVLAN_STATE_MASK \
483 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
484
485 static void macvlan_set_lockdep_class_one(struct net_device *dev,
486 struct netdev_queue *txq,
487 void *_unused)
488 {
489 lockdep_set_class(&txq->_xmit_lock,
490 &macvlan_netdev_xmit_lock_key);
491 }
492
493 static void macvlan_set_lockdep_class(struct net_device *dev)
494 {
495 lockdep_set_class(&dev->addr_list_lock,
496 &macvlan_netdev_addr_lock_key);
497 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
498 }
499
500 static int macvlan_init(struct net_device *dev)
501 {
502 struct macvlan_dev *vlan = netdev_priv(dev);
503 const struct net_device *lowerdev = vlan->lowerdev;
504
505 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
506 (lowerdev->state & MACVLAN_STATE_MASK);
507 dev->features = lowerdev->features & MACVLAN_FEATURES;
508 dev->features |= NETIF_F_LLTX;
509 dev->gso_max_size = lowerdev->gso_max_size;
510 dev->iflink = lowerdev->ifindex;
511 dev->hard_header_len = lowerdev->hard_header_len;
512
513 macvlan_set_lockdep_class(dev);
514
515 vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
516 if (!vlan->pcpu_stats)
517 return -ENOMEM;
518
519 return 0;
520 }
521
522 static void macvlan_uninit(struct net_device *dev)
523 {
524 struct macvlan_dev *vlan = netdev_priv(dev);
525 struct macvlan_port *port = vlan->port;
526
527 free_percpu(vlan->pcpu_stats);
528
529 port->count -= 1;
530 if (!port->count)
531 macvlan_port_destroy(port->dev);
532 }
533
534 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
535 struct rtnl_link_stats64 *stats)
536 {
537 struct macvlan_dev *vlan = netdev_priv(dev);
538
539 if (vlan->pcpu_stats) {
540 struct macvlan_pcpu_stats *p;
541 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
542 u32 rx_errors = 0, tx_dropped = 0;
543 unsigned int start;
544 int i;
545
546 for_each_possible_cpu(i) {
547 p = per_cpu_ptr(vlan->pcpu_stats, i);
548 do {
549 start = u64_stats_fetch_begin_bh(&p->syncp);
550 rx_packets = p->rx_packets;
551 rx_bytes = p->rx_bytes;
552 rx_multicast = p->rx_multicast;
553 tx_packets = p->tx_packets;
554 tx_bytes = p->tx_bytes;
555 } while (u64_stats_fetch_retry_bh(&p->syncp, start));
556
557 stats->rx_packets += rx_packets;
558 stats->rx_bytes += rx_bytes;
559 stats->multicast += rx_multicast;
560 stats->tx_packets += tx_packets;
561 stats->tx_bytes += tx_bytes;
562 /* rx_errors & tx_dropped are u32, updated
563 * without syncp protection.
564 */
565 rx_errors += p->rx_errors;
566 tx_dropped += p->tx_dropped;
567 }
568 stats->rx_errors = rx_errors;
569 stats->rx_dropped = rx_errors;
570 stats->tx_dropped = tx_dropped;
571 }
572 return stats;
573 }
574
575 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
576 __be16 proto, u16 vid)
577 {
578 struct macvlan_dev *vlan = netdev_priv(dev);
579 struct net_device *lowerdev = vlan->lowerdev;
580
581 return vlan_vid_add(lowerdev, proto, vid);
582 }
583
584 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
585 __be16 proto, u16 vid)
586 {
587 struct macvlan_dev *vlan = netdev_priv(dev);
588 struct net_device *lowerdev = vlan->lowerdev;
589
590 vlan_vid_del(lowerdev, proto, vid);
591 return 0;
592 }
593
594 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
595 struct net_device *dev,
596 const unsigned char *addr,
597 u16 flags)
598 {
599 struct macvlan_dev *vlan = netdev_priv(dev);
600 int err = -EINVAL;
601
602 if (!vlan->port->passthru)
603 return -EOPNOTSUPP;
604
605 if (flags & NLM_F_REPLACE)
606 return -EOPNOTSUPP;
607
608 if (is_unicast_ether_addr(addr))
609 err = dev_uc_add_excl(dev, addr);
610 else if (is_multicast_ether_addr(addr))
611 err = dev_mc_add_excl(dev, addr);
612
613 return err;
614 }
615
616 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
617 struct net_device *dev,
618 const unsigned char *addr)
619 {
620 struct macvlan_dev *vlan = netdev_priv(dev);
621 int err = -EINVAL;
622
623 if (!vlan->port->passthru)
624 return -EOPNOTSUPP;
625
626 if (is_unicast_ether_addr(addr))
627 err = dev_uc_del(dev, addr);
628 else if (is_multicast_ether_addr(addr))
629 err = dev_mc_del(dev, addr);
630
631 return err;
632 }
633
634 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
635 struct ethtool_drvinfo *drvinfo)
636 {
637 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
638 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
639 }
640
641 static int macvlan_ethtool_get_settings(struct net_device *dev,
642 struct ethtool_cmd *cmd)
643 {
644 const struct macvlan_dev *vlan = netdev_priv(dev);
645
646 return __ethtool_get_settings(vlan->lowerdev, cmd);
647 }
648
649 static netdev_features_t macvlan_fix_features(struct net_device *dev,
650 netdev_features_t features)
651 {
652 struct macvlan_dev *vlan = netdev_priv(dev);
653
654 return features & (vlan->set_features | ~MACVLAN_FEATURES);
655 }
656
657 static const struct ethtool_ops macvlan_ethtool_ops = {
658 .get_link = ethtool_op_get_link,
659 .get_settings = macvlan_ethtool_get_settings,
660 .get_drvinfo = macvlan_ethtool_get_drvinfo,
661 };
662
663 static const struct net_device_ops macvlan_netdev_ops = {
664 .ndo_init = macvlan_init,
665 .ndo_uninit = macvlan_uninit,
666 .ndo_open = macvlan_open,
667 .ndo_stop = macvlan_stop,
668 .ndo_start_xmit = macvlan_start_xmit,
669 .ndo_change_mtu = macvlan_change_mtu,
670 .ndo_fix_features = macvlan_fix_features,
671 .ndo_change_rx_flags = macvlan_change_rx_flags,
672 .ndo_set_mac_address = macvlan_set_mac_address,
673 .ndo_set_rx_mode = macvlan_set_mac_lists,
674 .ndo_get_stats64 = macvlan_dev_get_stats64,
675 .ndo_validate_addr = eth_validate_addr,
676 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
677 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
678 .ndo_fdb_add = macvlan_fdb_add,
679 .ndo_fdb_del = macvlan_fdb_del,
680 .ndo_fdb_dump = ndo_dflt_fdb_dump,
681 };
682
683 void macvlan_common_setup(struct net_device *dev)
684 {
685 ether_setup(dev);
686
687 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
688 dev->priv_flags |= IFF_UNICAST_FLT;
689 dev->netdev_ops = &macvlan_netdev_ops;
690 dev->destructor = free_netdev;
691 dev->header_ops = &macvlan_hard_header_ops;
692 dev->ethtool_ops = &macvlan_ethtool_ops;
693 }
694 EXPORT_SYMBOL_GPL(macvlan_common_setup);
695
696 static void macvlan_setup(struct net_device *dev)
697 {
698 macvlan_common_setup(dev);
699 dev->tx_queue_len = 0;
700 }
701
702 static int macvlan_port_create(struct net_device *dev)
703 {
704 struct macvlan_port *port;
705 unsigned int i;
706 int err;
707
708 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
709 return -EINVAL;
710
711 port = kzalloc(sizeof(*port), GFP_KERNEL);
712 if (port == NULL)
713 return -ENOMEM;
714
715 port->passthru = false;
716 port->dev = dev;
717 INIT_LIST_HEAD(&port->vlans);
718 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
719 INIT_HLIST_HEAD(&port->vlan_hash[i]);
720
721 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
722 if (err)
723 kfree(port);
724 else
725 dev->priv_flags |= IFF_MACVLAN_PORT;
726 return err;
727 }
728
729 static void macvlan_port_destroy(struct net_device *dev)
730 {
731 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
732
733 dev->priv_flags &= ~IFF_MACVLAN_PORT;
734 netdev_rx_handler_unregister(dev);
735 kfree_rcu(port, rcu);
736 }
737
738 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
739 {
740 if (tb[IFLA_ADDRESS]) {
741 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
742 return -EINVAL;
743 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
744 return -EADDRNOTAVAIL;
745 }
746
747 if (data && data[IFLA_MACVLAN_FLAGS] &&
748 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
749 return -EINVAL;
750
751 if (data && data[IFLA_MACVLAN_MODE]) {
752 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
753 case MACVLAN_MODE_PRIVATE:
754 case MACVLAN_MODE_VEPA:
755 case MACVLAN_MODE_BRIDGE:
756 case MACVLAN_MODE_PASSTHRU:
757 break;
758 default:
759 return -EINVAL;
760 }
761 }
762 return 0;
763 }
764
765 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
766 struct nlattr *tb[], struct nlattr *data[],
767 int (*receive)(struct sk_buff *skb),
768 int (*forward)(struct net_device *dev,
769 struct sk_buff *skb))
770 {
771 struct macvlan_dev *vlan = netdev_priv(dev);
772 struct macvlan_port *port;
773 struct net_device *lowerdev;
774 int err;
775
776 if (!tb[IFLA_LINK])
777 return -EINVAL;
778
779 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
780 if (lowerdev == NULL)
781 return -ENODEV;
782
783 /* When creating macvlans on top of other macvlans - use
784 * the real device as the lowerdev.
785 */
786 if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
787 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
788 lowerdev = lowervlan->lowerdev;
789 }
790
791 if (!tb[IFLA_MTU])
792 dev->mtu = lowerdev->mtu;
793 else if (dev->mtu > lowerdev->mtu)
794 return -EINVAL;
795
796 if (!tb[IFLA_ADDRESS])
797 eth_hw_addr_random(dev);
798
799 if (!macvlan_port_exists(lowerdev)) {
800 err = macvlan_port_create(lowerdev);
801 if (err < 0)
802 return err;
803 }
804 port = macvlan_port_get_rtnl(lowerdev);
805
806 /* Only 1 macvlan device can be created in passthru mode */
807 if (port->passthru)
808 return -EINVAL;
809
810 vlan->lowerdev = lowerdev;
811 vlan->dev = dev;
812 vlan->port = port;
813 vlan->receive = receive;
814 vlan->forward = forward;
815 vlan->set_features = MACVLAN_FEATURES;
816
817 vlan->mode = MACVLAN_MODE_VEPA;
818 if (data && data[IFLA_MACVLAN_MODE])
819 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
820
821 if (data && data[IFLA_MACVLAN_FLAGS])
822 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
823
824 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
825 if (port->count)
826 return -EINVAL;
827 port->passthru = true;
828 eth_hw_addr_inherit(dev, lowerdev);
829 }
830
831 err = netdev_upper_dev_link(lowerdev, dev);
832 if (err)
833 goto destroy_port;
834
835 port->count += 1;
836 err = register_netdevice(dev);
837 if (err < 0)
838 goto upper_dev_unlink;
839
840 list_add_tail_rcu(&vlan->list, &port->vlans);
841 netif_stacked_transfer_operstate(lowerdev, dev);
842
843 return 0;
844
845 upper_dev_unlink:
846 netdev_upper_dev_unlink(lowerdev, dev);
847 destroy_port:
848 port->count -= 1;
849 if (!port->count)
850 macvlan_port_destroy(lowerdev);
851
852 return err;
853 }
854 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
855
856 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
857 struct nlattr *tb[], struct nlattr *data[])
858 {
859 return macvlan_common_newlink(src_net, dev, tb, data,
860 netif_rx,
861 dev_forward_skb);
862 }
863
864 void macvlan_dellink(struct net_device *dev, struct list_head *head)
865 {
866 struct macvlan_dev *vlan = netdev_priv(dev);
867
868 list_del_rcu(&vlan->list);
869 unregister_netdevice_queue(dev, head);
870 netdev_upper_dev_unlink(vlan->lowerdev, dev);
871 }
872 EXPORT_SYMBOL_GPL(macvlan_dellink);
873
874 static int macvlan_changelink(struct net_device *dev,
875 struct nlattr *tb[], struct nlattr *data[])
876 {
877 struct macvlan_dev *vlan = netdev_priv(dev);
878 enum macvlan_mode mode;
879 bool set_mode = false;
880
881 /* Validate mode, but don't set yet: setting flags may fail. */
882 if (data && data[IFLA_MACVLAN_MODE]) {
883 set_mode = true;
884 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
885 /* Passthrough mode can't be set or cleared dynamically */
886 if ((mode == MACVLAN_MODE_PASSTHRU) !=
887 (vlan->mode == MACVLAN_MODE_PASSTHRU))
888 return -EINVAL;
889 }
890
891 if (data && data[IFLA_MACVLAN_FLAGS]) {
892 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
893 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
894 if (vlan->port->passthru && promisc) {
895 int err;
896
897 if (flags & MACVLAN_FLAG_NOPROMISC)
898 err = dev_set_promiscuity(vlan->lowerdev, -1);
899 else
900 err = dev_set_promiscuity(vlan->lowerdev, 1);
901 if (err < 0)
902 return err;
903 }
904 vlan->flags = flags;
905 }
906 if (set_mode)
907 vlan->mode = mode;
908 return 0;
909 }
910
911 static size_t macvlan_get_size(const struct net_device *dev)
912 {
913 return (0
914 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
915 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
916 );
917 }
918
919 static int macvlan_fill_info(struct sk_buff *skb,
920 const struct net_device *dev)
921 {
922 struct macvlan_dev *vlan = netdev_priv(dev);
923
924 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
925 goto nla_put_failure;
926 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
927 goto nla_put_failure;
928 return 0;
929
930 nla_put_failure:
931 return -EMSGSIZE;
932 }
933
934 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
935 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
936 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
937 };
938
939 int macvlan_link_register(struct rtnl_link_ops *ops)
940 {
941 /* common fields */
942 ops->priv_size = sizeof(struct macvlan_dev);
943 ops->validate = macvlan_validate;
944 ops->maxtype = IFLA_MACVLAN_MAX;
945 ops->policy = macvlan_policy;
946 ops->changelink = macvlan_changelink;
947 ops->get_size = macvlan_get_size;
948 ops->fill_info = macvlan_fill_info;
949
950 return rtnl_link_register(ops);
951 };
952 EXPORT_SYMBOL_GPL(macvlan_link_register);
953
954 static struct rtnl_link_ops macvlan_link_ops = {
955 .kind = "macvlan",
956 .setup = macvlan_setup,
957 .newlink = macvlan_newlink,
958 .dellink = macvlan_dellink,
959 };
960
961 static int macvlan_device_event(struct notifier_block *unused,
962 unsigned long event, void *ptr)
963 {
964 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
965 struct macvlan_dev *vlan, *next;
966 struct macvlan_port *port;
967 LIST_HEAD(list_kill);
968
969 if (!macvlan_port_exists(dev))
970 return NOTIFY_DONE;
971
972 port = macvlan_port_get_rtnl(dev);
973
974 switch (event) {
975 case NETDEV_CHANGE:
976 list_for_each_entry(vlan, &port->vlans, list)
977 netif_stacked_transfer_operstate(vlan->lowerdev,
978 vlan->dev);
979 break;
980 case NETDEV_FEAT_CHANGE:
981 list_for_each_entry(vlan, &port->vlans, list) {
982 vlan->dev->features = dev->features & MACVLAN_FEATURES;
983 vlan->dev->gso_max_size = dev->gso_max_size;
984 netdev_features_change(vlan->dev);
985 }
986 break;
987 case NETDEV_UNREGISTER:
988 /* twiddle thumbs on netns device moves */
989 if (dev->reg_state != NETREG_UNREGISTERING)
990 break;
991
992 list_for_each_entry_safe(vlan, next, &port->vlans, list)
993 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
994 unregister_netdevice_many(&list_kill);
995 list_del(&list_kill);
996 break;
997 case NETDEV_PRE_TYPE_CHANGE:
998 /* Forbid underlaying device to change its type. */
999 return NOTIFY_BAD;
1000 }
1001 return NOTIFY_DONE;
1002 }
1003
1004 static struct notifier_block macvlan_notifier_block __read_mostly = {
1005 .notifier_call = macvlan_device_event,
1006 };
1007
1008 static int __init macvlan_init_module(void)
1009 {
1010 int err;
1011
1012 register_netdevice_notifier(&macvlan_notifier_block);
1013
1014 err = macvlan_link_register(&macvlan_link_ops);
1015 if (err < 0)
1016 goto err1;
1017 return 0;
1018 err1:
1019 unregister_netdevice_notifier(&macvlan_notifier_block);
1020 return err;
1021 }
1022
1023 static void __exit macvlan_cleanup_module(void)
1024 {
1025 rtnl_link_unregister(&macvlan_link_ops);
1026 unregister_netdevice_notifier(&macvlan_notifier_block);
1027 }
1028
1029 module_init(macvlan_init_module);
1030 module_exit(macvlan_cleanup_module);
1031
1032 MODULE_LICENSE("GPL");
1033 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1034 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1035 MODULE_ALIAS_RTNL_LINK("macvlan");
Linus' kernel tree