]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - drivers/net/macvlan.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
net: Add layer 2 hardware acceleration operations for macvlan devices
[mirror_ubuntu-artful-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_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 if (vlan->fwd_priv) {
301 skb->dev = vlan->lowerdev;
302 ret = dev_hard_start_xmit(skb, skb->dev, NULL, vlan->fwd_priv);
303 } else {
304 ret = macvlan_queue_xmit(skb, dev);
305 }
306
307 if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
308 struct macvlan_pcpu_stats *pcpu_stats;
309
310 pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
311 u64_stats_update_begin(&pcpu_stats->syncp);
312 pcpu_stats->tx_packets++;
313 pcpu_stats->tx_bytes += len;
314 u64_stats_update_end(&pcpu_stats->syncp);
315 } else {
316 this_cpu_inc(vlan->pcpu_stats->tx_dropped);
317 }
318 return ret;
319 }
320 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
321
322 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
323 unsigned short type, const void *daddr,
324 const void *saddr, unsigned len)
325 {
326 const struct macvlan_dev *vlan = netdev_priv(dev);
327 struct net_device *lowerdev = vlan->lowerdev;
328
329 return dev_hard_header(skb, lowerdev, type, daddr,
330 saddr ? : dev->dev_addr, len);
331 }
332
333 static const struct header_ops macvlan_hard_header_ops = {
334 .create = macvlan_hard_header,
335 .rebuild = eth_rebuild_header,
336 .parse = eth_header_parse,
337 .cache = eth_header_cache,
338 .cache_update = eth_header_cache_update,
339 };
340
341 static int macvlan_open(struct net_device *dev)
342 {
343 struct macvlan_dev *vlan = netdev_priv(dev);
344 struct net_device *lowerdev = vlan->lowerdev;
345 int err;
346
347 if (vlan->port->passthru) {
348 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
349 err = dev_set_promiscuity(lowerdev, 1);
350 if (err < 0)
351 goto out;
352 }
353 goto hash_add;
354 }
355
356 if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD) {
357 vlan->fwd_priv =
358 lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
359
360 /* If we get a NULL pointer back, or if we get an error
361 * then we should just fall through to the non accelerated path
362 */
363 if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
364 vlan->fwd_priv = NULL;
365 } else {
366 dev->features &= ~NETIF_F_LLTX;
367 return 0;
368 }
369 }
370
371 err = -EBUSY;
372 if (macvlan_addr_busy(vlan->port, dev->dev_addr))
373 goto out;
374
375 err = dev_uc_add(lowerdev, dev->dev_addr);
376 if (err < 0)
377 goto out;
378 if (dev->flags & IFF_ALLMULTI) {
379 err = dev_set_allmulti(lowerdev, 1);
380 if (err < 0)
381 goto del_unicast;
382 }
383
384 hash_add:
385 macvlan_hash_add(vlan);
386 return 0;
387
388 del_unicast:
389 dev_uc_del(lowerdev, dev->dev_addr);
390 out:
391 if (vlan->fwd_priv) {
392 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
393 vlan->fwd_priv);
394 vlan->fwd_priv = NULL;
395 }
396 return err;
397 }
398
399 static int macvlan_stop(struct net_device *dev)
400 {
401 struct macvlan_dev *vlan = netdev_priv(dev);
402 struct net_device *lowerdev = vlan->lowerdev;
403
404 if (vlan->fwd_priv) {
405 lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
406 vlan->fwd_priv);
407 vlan->fwd_priv = NULL;
408 return 0;
409 }
410
411 dev_uc_unsync(lowerdev, dev);
412 dev_mc_unsync(lowerdev, dev);
413
414 if (vlan->port->passthru) {
415 if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
416 dev_set_promiscuity(lowerdev, -1);
417 goto hash_del;
418 }
419
420 if (dev->flags & IFF_ALLMULTI)
421 dev_set_allmulti(lowerdev, -1);
422
423 dev_uc_del(lowerdev, dev->dev_addr);
424
425 hash_del:
426 macvlan_hash_del(vlan, !dev->dismantle);
427 return 0;
428 }
429
430 static int macvlan_set_mac_address(struct net_device *dev, void *p)
431 {
432 struct macvlan_dev *vlan = netdev_priv(dev);
433 struct net_device *lowerdev = vlan->lowerdev;
434 struct sockaddr *addr = p;
435 int err;
436
437 if (!is_valid_ether_addr(addr->sa_data))
438 return -EADDRNOTAVAIL;
439
440 if (!(dev->flags & IFF_UP)) {
441 /* Just copy in the new address */
442 memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
443 } else {
444 /* Rehash and update the device filters */
445 if (macvlan_addr_busy(vlan->port, addr->sa_data))
446 return -EBUSY;
447
448 err = dev_uc_add(lowerdev, addr->sa_data);
449 if (err)
450 return err;
451
452 dev_uc_del(lowerdev, dev->dev_addr);
453
454 macvlan_hash_change_addr(vlan, addr->sa_data);
455 }
456 return 0;
457 }
458
459 static void macvlan_change_rx_flags(struct net_device *dev, int change)
460 {
461 struct macvlan_dev *vlan = netdev_priv(dev);
462 struct net_device *lowerdev = vlan->lowerdev;
463
464 if (change & IFF_ALLMULTI)
465 dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
466 }
467
468 static void macvlan_set_mac_lists(struct net_device *dev)
469 {
470 struct macvlan_dev *vlan = netdev_priv(dev);
471
472 if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
473 bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
474 } else {
475 struct netdev_hw_addr *ha;
476 DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
477
478 bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
479 netdev_for_each_mc_addr(ha, dev) {
480 __set_bit(mc_hash(vlan, ha->addr), filter);
481 }
482
483 __set_bit(mc_hash(vlan, dev->broadcast), filter);
484
485 bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
486 }
487 dev_uc_sync(vlan->lowerdev, dev);
488 dev_mc_sync(vlan->lowerdev, dev);
489 }
490
491 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
492 {
493 struct macvlan_dev *vlan = netdev_priv(dev);
494
495 if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
496 return -EINVAL;
497 dev->mtu = new_mtu;
498 return 0;
499 }
500
501 /*
502 * macvlan network devices have devices nesting below it and are a special
503 * "super class" of normal network devices; split their locks off into a
504 * separate class since they always nest.
505 */
506 static struct lock_class_key macvlan_netdev_xmit_lock_key;
507 static struct lock_class_key macvlan_netdev_addr_lock_key;
508
509 #define MACVLAN_FEATURES \
510 (NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
511 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
512 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
513 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
514
515 #define MACVLAN_STATE_MASK \
516 ((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
517
518 static void macvlan_set_lockdep_class_one(struct net_device *dev,
519 struct netdev_queue *txq,
520 void *_unused)
521 {
522 lockdep_set_class(&txq->_xmit_lock,
523 &macvlan_netdev_xmit_lock_key);
524 }
525
526 static void macvlan_set_lockdep_class(struct net_device *dev)
527 {
528 lockdep_set_class(&dev->addr_list_lock,
529 &macvlan_netdev_addr_lock_key);
530 netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
531 }
532
533 static int macvlan_init(struct net_device *dev)
534 {
535 struct macvlan_dev *vlan = netdev_priv(dev);
536 const struct net_device *lowerdev = vlan->lowerdev;
537
538 dev->state = (dev->state & ~MACVLAN_STATE_MASK) |
539 (lowerdev->state & MACVLAN_STATE_MASK);
540 dev->features = lowerdev->features & MACVLAN_FEATURES;
541 dev->features |= NETIF_F_LLTX;
542 dev->gso_max_size = lowerdev->gso_max_size;
543 dev->iflink = lowerdev->ifindex;
544 dev->hard_header_len = lowerdev->hard_header_len;
545
546 macvlan_set_lockdep_class(dev);
547
548 vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
549 if (!vlan->pcpu_stats)
550 return -ENOMEM;
551
552 return 0;
553 }
554
555 static void macvlan_uninit(struct net_device *dev)
556 {
557 struct macvlan_dev *vlan = netdev_priv(dev);
558 struct macvlan_port *port = vlan->port;
559
560 free_percpu(vlan->pcpu_stats);
561
562 port->count -= 1;
563 if (!port->count)
564 macvlan_port_destroy(port->dev);
565 }
566
567 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
568 struct rtnl_link_stats64 *stats)
569 {
570 struct macvlan_dev *vlan = netdev_priv(dev);
571
572 if (vlan->pcpu_stats) {
573 struct macvlan_pcpu_stats *p;
574 u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
575 u32 rx_errors = 0, tx_dropped = 0;
576 unsigned int start;
577 int i;
578
579 for_each_possible_cpu(i) {
580 p = per_cpu_ptr(vlan->pcpu_stats, i);
581 do {
582 start = u64_stats_fetch_begin_bh(&p->syncp);
583 rx_packets = p->rx_packets;
584 rx_bytes = p->rx_bytes;
585 rx_multicast = p->rx_multicast;
586 tx_packets = p->tx_packets;
587 tx_bytes = p->tx_bytes;
588 } while (u64_stats_fetch_retry_bh(&p->syncp, start));
589
590 stats->rx_packets += rx_packets;
591 stats->rx_bytes += rx_bytes;
592 stats->multicast += rx_multicast;
593 stats->tx_packets += tx_packets;
594 stats->tx_bytes += tx_bytes;
595 /* rx_errors & tx_dropped are u32, updated
596 * without syncp protection.
597 */
598 rx_errors += p->rx_errors;
599 tx_dropped += p->tx_dropped;
600 }
601 stats->rx_errors = rx_errors;
602 stats->rx_dropped = rx_errors;
603 stats->tx_dropped = tx_dropped;
604 }
605 return stats;
606 }
607
608 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
609 __be16 proto, u16 vid)
610 {
611 struct macvlan_dev *vlan = netdev_priv(dev);
612 struct net_device *lowerdev = vlan->lowerdev;
613
614 return vlan_vid_add(lowerdev, proto, vid);
615 }
616
617 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
618 __be16 proto, u16 vid)
619 {
620 struct macvlan_dev *vlan = netdev_priv(dev);
621 struct net_device *lowerdev = vlan->lowerdev;
622
623 vlan_vid_del(lowerdev, proto, vid);
624 return 0;
625 }
626
627 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
628 struct net_device *dev,
629 const unsigned char *addr,
630 u16 flags)
631 {
632 struct macvlan_dev *vlan = netdev_priv(dev);
633 int err = -EINVAL;
634
635 if (!vlan->port->passthru)
636 return -EOPNOTSUPP;
637
638 if (flags & NLM_F_REPLACE)
639 return -EOPNOTSUPP;
640
641 if (is_unicast_ether_addr(addr))
642 err = dev_uc_add_excl(dev, addr);
643 else if (is_multicast_ether_addr(addr))
644 err = dev_mc_add_excl(dev, addr);
645
646 return err;
647 }
648
649 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
650 struct net_device *dev,
651 const unsigned char *addr)
652 {
653 struct macvlan_dev *vlan = netdev_priv(dev);
654 int err = -EINVAL;
655
656 if (!vlan->port->passthru)
657 return -EOPNOTSUPP;
658
659 if (is_unicast_ether_addr(addr))
660 err = dev_uc_del(dev, addr);
661 else if (is_multicast_ether_addr(addr))
662 err = dev_mc_del(dev, addr);
663
664 return err;
665 }
666
667 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
668 struct ethtool_drvinfo *drvinfo)
669 {
670 strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
671 strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
672 }
673
674 static int macvlan_ethtool_get_settings(struct net_device *dev,
675 struct ethtool_cmd *cmd)
676 {
677 const struct macvlan_dev *vlan = netdev_priv(dev);
678
679 return __ethtool_get_settings(vlan->lowerdev, cmd);
680 }
681
682 static netdev_features_t macvlan_fix_features(struct net_device *dev,
683 netdev_features_t features)
684 {
685 struct macvlan_dev *vlan = netdev_priv(dev);
686
687 return features & (vlan->set_features | ~MACVLAN_FEATURES);
688 }
689
690 static const struct ethtool_ops macvlan_ethtool_ops = {
691 .get_link = ethtool_op_get_link,
692 .get_settings = macvlan_ethtool_get_settings,
693 .get_drvinfo = macvlan_ethtool_get_drvinfo,
694 };
695
696 static const struct net_device_ops macvlan_netdev_ops = {
697 .ndo_init = macvlan_init,
698 .ndo_uninit = macvlan_uninit,
699 .ndo_open = macvlan_open,
700 .ndo_stop = macvlan_stop,
701 .ndo_start_xmit = macvlan_start_xmit,
702 .ndo_change_mtu = macvlan_change_mtu,
703 .ndo_fix_features = macvlan_fix_features,
704 .ndo_change_rx_flags = macvlan_change_rx_flags,
705 .ndo_set_mac_address = macvlan_set_mac_address,
706 .ndo_set_rx_mode = macvlan_set_mac_lists,
707 .ndo_get_stats64 = macvlan_dev_get_stats64,
708 .ndo_validate_addr = eth_validate_addr,
709 .ndo_vlan_rx_add_vid = macvlan_vlan_rx_add_vid,
710 .ndo_vlan_rx_kill_vid = macvlan_vlan_rx_kill_vid,
711 .ndo_fdb_add = macvlan_fdb_add,
712 .ndo_fdb_del = macvlan_fdb_del,
713 .ndo_fdb_dump = ndo_dflt_fdb_dump,
714 };
715
716 void macvlan_common_setup(struct net_device *dev)
717 {
718 ether_setup(dev);
719
720 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
721 dev->priv_flags |= IFF_UNICAST_FLT;
722 dev->netdev_ops = &macvlan_netdev_ops;
723 dev->destructor = free_netdev;
724 dev->header_ops = &macvlan_hard_header_ops;
725 dev->ethtool_ops = &macvlan_ethtool_ops;
726 }
727 EXPORT_SYMBOL_GPL(macvlan_common_setup);
728
729 static void macvlan_setup(struct net_device *dev)
730 {
731 macvlan_common_setup(dev);
732 dev->tx_queue_len = 0;
733 }
734
735 static int macvlan_port_create(struct net_device *dev)
736 {
737 struct macvlan_port *port;
738 unsigned int i;
739 int err;
740
741 if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
742 return -EINVAL;
743
744 port = kzalloc(sizeof(*port), GFP_KERNEL);
745 if (port == NULL)
746 return -ENOMEM;
747
748 port->passthru = false;
749 port->dev = dev;
750 INIT_LIST_HEAD(&port->vlans);
751 for (i = 0; i < MACVLAN_HASH_SIZE; i++)
752 INIT_HLIST_HEAD(&port->vlan_hash[i]);
753
754 err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
755 if (err)
756 kfree(port);
757 else
758 dev->priv_flags |= IFF_MACVLAN_PORT;
759 return err;
760 }
761
762 static void macvlan_port_destroy(struct net_device *dev)
763 {
764 struct macvlan_port *port = macvlan_port_get_rtnl(dev);
765
766 dev->priv_flags &= ~IFF_MACVLAN_PORT;
767 netdev_rx_handler_unregister(dev);
768 kfree_rcu(port, rcu);
769 }
770
771 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
772 {
773 if (tb[IFLA_ADDRESS]) {
774 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
775 return -EINVAL;
776 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
777 return -EADDRNOTAVAIL;
778 }
779
780 if (data && data[IFLA_MACVLAN_FLAGS] &&
781 nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
782 return -EINVAL;
783
784 if (data && data[IFLA_MACVLAN_MODE]) {
785 switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
786 case MACVLAN_MODE_PRIVATE:
787 case MACVLAN_MODE_VEPA:
788 case MACVLAN_MODE_BRIDGE:
789 case MACVLAN_MODE_PASSTHRU:
790 break;
791 default:
792 return -EINVAL;
793 }
794 }
795 return 0;
796 }
797
798 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
799 struct nlattr *tb[], struct nlattr *data[],
800 int (*receive)(struct sk_buff *skb),
801 int (*forward)(struct net_device *dev,
802 struct sk_buff *skb))
803 {
804 struct macvlan_dev *vlan = netdev_priv(dev);
805 struct macvlan_port *port;
806 struct net_device *lowerdev;
807 int err;
808
809 if (!tb[IFLA_LINK])
810 return -EINVAL;
811
812 lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
813 if (lowerdev == NULL)
814 return -ENODEV;
815
816 /* When creating macvlans on top of other macvlans - use
817 * the real device as the lowerdev.
818 */
819 if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
820 struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
821 lowerdev = lowervlan->lowerdev;
822 }
823
824 if (!tb[IFLA_MTU])
825 dev->mtu = lowerdev->mtu;
826 else if (dev->mtu > lowerdev->mtu)
827 return -EINVAL;
828
829 if (!tb[IFLA_ADDRESS])
830 eth_hw_addr_random(dev);
831
832 if (!macvlan_port_exists(lowerdev)) {
833 err = macvlan_port_create(lowerdev);
834 if (err < 0)
835 return err;
836 }
837 port = macvlan_port_get_rtnl(lowerdev);
838
839 /* Only 1 macvlan device can be created in passthru mode */
840 if (port->passthru)
841 return -EINVAL;
842
843 vlan->lowerdev = lowerdev;
844 vlan->dev = dev;
845 vlan->port = port;
846 vlan->receive = receive;
847 vlan->forward = forward;
848 vlan->set_features = MACVLAN_FEATURES;
849
850 vlan->mode = MACVLAN_MODE_VEPA;
851 if (data && data[IFLA_MACVLAN_MODE])
852 vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
853
854 if (data && data[IFLA_MACVLAN_FLAGS])
855 vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
856
857 if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
858 if (port->count)
859 return -EINVAL;
860 port->passthru = true;
861 eth_hw_addr_inherit(dev, lowerdev);
862 }
863
864 port->count += 1;
865 err = register_netdevice(dev);
866 if (err < 0)
867 goto destroy_port;
868
869 dev->priv_flags |= IFF_MACVLAN;
870 err = netdev_upper_dev_link(lowerdev, dev);
871 if (err)
872 goto destroy_port;
873
874
875 list_add_tail_rcu(&vlan->list, &port->vlans);
876 netif_stacked_transfer_operstate(lowerdev, dev);
877
878 return 0;
879
880 destroy_port:
881 port->count -= 1;
882 if (!port->count)
883 macvlan_port_destroy(lowerdev);
884
885 return err;
886 }
887 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
888
889 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
890 struct nlattr *tb[], struct nlattr *data[])
891 {
892 return macvlan_common_newlink(src_net, dev, tb, data,
893 netif_rx,
894 dev_forward_skb);
895 }
896
897 void macvlan_dellink(struct net_device *dev, struct list_head *head)
898 {
899 struct macvlan_dev *vlan = netdev_priv(dev);
900
901 list_del_rcu(&vlan->list);
902 unregister_netdevice_queue(dev, head);
903 netdev_upper_dev_unlink(vlan->lowerdev, dev);
904 }
905 EXPORT_SYMBOL_GPL(macvlan_dellink);
906
907 static int macvlan_changelink(struct net_device *dev,
908 struct nlattr *tb[], struct nlattr *data[])
909 {
910 struct macvlan_dev *vlan = netdev_priv(dev);
911 enum macvlan_mode mode;
912 bool set_mode = false;
913
914 /* Validate mode, but don't set yet: setting flags may fail. */
915 if (data && data[IFLA_MACVLAN_MODE]) {
916 set_mode = true;
917 mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
918 /* Passthrough mode can't be set or cleared dynamically */
919 if ((mode == MACVLAN_MODE_PASSTHRU) !=
920 (vlan->mode == MACVLAN_MODE_PASSTHRU))
921 return -EINVAL;
922 }
923
924 if (data && data[IFLA_MACVLAN_FLAGS]) {
925 __u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
926 bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
927 if (vlan->port->passthru && promisc) {
928 int err;
929
930 if (flags & MACVLAN_FLAG_NOPROMISC)
931 err = dev_set_promiscuity(vlan->lowerdev, -1);
932 else
933 err = dev_set_promiscuity(vlan->lowerdev, 1);
934 if (err < 0)
935 return err;
936 }
937 vlan->flags = flags;
938 }
939 if (set_mode)
940 vlan->mode = mode;
941 return 0;
942 }
943
944 static size_t macvlan_get_size(const struct net_device *dev)
945 {
946 return (0
947 + nla_total_size(4) /* IFLA_MACVLAN_MODE */
948 + nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
949 );
950 }
951
952 static int macvlan_fill_info(struct sk_buff *skb,
953 const struct net_device *dev)
954 {
955 struct macvlan_dev *vlan = netdev_priv(dev);
956
957 if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
958 goto nla_put_failure;
959 if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
960 goto nla_put_failure;
961 return 0;
962
963 nla_put_failure:
964 return -EMSGSIZE;
965 }
966
967 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
968 [IFLA_MACVLAN_MODE] = { .type = NLA_U32 },
969 [IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
970 };
971
972 int macvlan_link_register(struct rtnl_link_ops *ops)
973 {
974 /* common fields */
975 ops->priv_size = sizeof(struct macvlan_dev);
976 ops->validate = macvlan_validate;
977 ops->maxtype = IFLA_MACVLAN_MAX;
978 ops->policy = macvlan_policy;
979 ops->changelink = macvlan_changelink;
980 ops->get_size = macvlan_get_size;
981 ops->fill_info = macvlan_fill_info;
982
983 return rtnl_link_register(ops);
984 };
985 EXPORT_SYMBOL_GPL(macvlan_link_register);
986
987 static struct rtnl_link_ops macvlan_link_ops = {
988 .kind = "macvlan",
989 .setup = macvlan_setup,
990 .newlink = macvlan_newlink,
991 .dellink = macvlan_dellink,
992 };
993
994 static int macvlan_device_event(struct notifier_block *unused,
995 unsigned long event, void *ptr)
996 {
997 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
998 struct macvlan_dev *vlan, *next;
999 struct macvlan_port *port;
1000 LIST_HEAD(list_kill);
1001
1002 if (!macvlan_port_exists(dev))
1003 return NOTIFY_DONE;
1004
1005 port = macvlan_port_get_rtnl(dev);
1006
1007 switch (event) {
1008 case NETDEV_CHANGE:
1009 list_for_each_entry(vlan, &port->vlans, list)
1010 netif_stacked_transfer_operstate(vlan->lowerdev,
1011 vlan->dev);
1012 break;
1013 case NETDEV_FEAT_CHANGE:
1014 list_for_each_entry(vlan, &port->vlans, list) {
1015 vlan->dev->features = dev->features & MACVLAN_FEATURES;
1016 vlan->dev->gso_max_size = dev->gso_max_size;
1017 netdev_features_change(vlan->dev);
1018 }
1019 break;
1020 case NETDEV_UNREGISTER:
1021 /* twiddle thumbs on netns device moves */
1022 if (dev->reg_state != NETREG_UNREGISTERING)
1023 break;
1024
1025 list_for_each_entry_safe(vlan, next, &port->vlans, list)
1026 vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1027 unregister_netdevice_many(&list_kill);
1028 list_del(&list_kill);
1029 break;
1030 case NETDEV_PRE_TYPE_CHANGE:
1031 /* Forbid underlaying device to change its type. */
1032 return NOTIFY_BAD;
1033 }
1034 return NOTIFY_DONE;
1035 }
1036
1037 static struct notifier_block macvlan_notifier_block __read_mostly = {
1038 .notifier_call = macvlan_device_event,
1039 };
1040
1041 static int __init macvlan_init_module(void)
1042 {
1043 int err;
1044
1045 register_netdevice_notifier(&macvlan_notifier_block);
1046
1047 err = macvlan_link_register(&macvlan_link_ops);
1048 if (err < 0)
1049 goto err1;
1050 return 0;
1051 err1:
1052 unregister_netdevice_notifier(&macvlan_notifier_block);
1053 return err;
1054 }
1055
1056 static void __exit macvlan_cleanup_module(void)
1057 {
1058 rtnl_link_unregister(&macvlan_link_ops);
1059 unregister_netdevice_notifier(&macvlan_notifier_block);
1060 }
1061
1062 module_init(macvlan_init_module);
1063 module_exit(macvlan_cleanup_module);
1064
1065 MODULE_LICENSE("GPL");
1066 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1067 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1068 MODULE_ALIAS_RTNL_LINK("macvlan");
Ubuntu Artful kernel mirror