xref: /openbmc/linux/drivers/net/macvlan.c (revision cdfce539)
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 	if (!skb)
122 		return NET_RX_DROP;
123 
124 	if (local)
125 		return vlan->forward(dev, skb);
126 
127 	skb->dev = dev;
128 	if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
129 		skb->pkt_type = PACKET_BROADCAST;
130 	else
131 		skb->pkt_type = PACKET_MULTICAST;
132 
133 	return vlan->receive(skb);
134 }
135 
136 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
137 {
138 	return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
139 }
140 
141 
142 static unsigned int mc_hash(const struct macvlan_dev *vlan,
143 			    const unsigned char *addr)
144 {
145 	u32 val = __get_unaligned_cpu32(addr + 2);
146 
147 	val ^= macvlan_hash_mix(vlan);
148 	return hash_32(val, MACVLAN_MC_FILTER_BITS);
149 }
150 
151 static void macvlan_broadcast(struct sk_buff *skb,
152 			      const struct macvlan_port *port,
153 			      struct net_device *src,
154 			      enum macvlan_mode mode)
155 {
156 	const struct ethhdr *eth = eth_hdr(skb);
157 	const struct macvlan_dev *vlan;
158 	struct sk_buff *nskb;
159 	unsigned int i;
160 	int err;
161 	unsigned int hash;
162 
163 	if (skb->protocol == htons(ETH_P_PAUSE))
164 		return;
165 
166 	for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
167 		hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
168 			if (vlan->dev == src || !(vlan->mode & mode))
169 				continue;
170 
171 			hash = mc_hash(vlan, eth->h_dest);
172 			if (!test_bit(hash, vlan->mc_filter))
173 				continue;
174 			nskb = skb_clone(skb, GFP_ATOMIC);
175 			err = macvlan_broadcast_one(nskb, vlan, eth,
176 					 mode == MACVLAN_MODE_BRIDGE);
177 			macvlan_count_rx(vlan, skb->len + ETH_HLEN,
178 					 err == NET_RX_SUCCESS, 1);
179 		}
180 	}
181 }
182 
183 /* called under rcu_read_lock() from netif_receive_skb */
184 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
185 {
186 	struct macvlan_port *port;
187 	struct sk_buff *skb = *pskb;
188 	const struct ethhdr *eth = eth_hdr(skb);
189 	const struct macvlan_dev *vlan;
190 	const struct macvlan_dev *src;
191 	struct net_device *dev;
192 	unsigned int len = 0;
193 	int ret = NET_RX_DROP;
194 
195 	port = macvlan_port_get_rcu(skb->dev);
196 	if (is_multicast_ether_addr(eth->h_dest)) {
197 		skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
198 		if (!skb)
199 			return RX_HANDLER_CONSUMED;
200 		eth = eth_hdr(skb);
201 		src = macvlan_hash_lookup(port, eth->h_source);
202 		if (!src)
203 			/* frame comes from an external address */
204 			macvlan_broadcast(skb, port, NULL,
205 					  MACVLAN_MODE_PRIVATE |
206 					  MACVLAN_MODE_VEPA    |
207 					  MACVLAN_MODE_PASSTHRU|
208 					  MACVLAN_MODE_BRIDGE);
209 		else if (src->mode == MACVLAN_MODE_VEPA)
210 			/* flood to everyone except source */
211 			macvlan_broadcast(skb, port, src->dev,
212 					  MACVLAN_MODE_VEPA |
213 					  MACVLAN_MODE_BRIDGE);
214 		else if (src->mode == MACVLAN_MODE_BRIDGE)
215 			/*
216 			 * flood only to VEPA ports, bridge ports
217 			 * already saw the frame on the way out.
218 			 */
219 			macvlan_broadcast(skb, port, src->dev,
220 					  MACVLAN_MODE_VEPA);
221 		else {
222 			/* forward to original port. */
223 			vlan = src;
224 			ret = macvlan_broadcast_one(skb, vlan, eth, 0);
225 			goto out;
226 		}
227 
228 		return RX_HANDLER_PASS;
229 	}
230 
231 	if (port->passthru)
232 		vlan = list_first_or_null_rcu(&port->vlans,
233 					      struct macvlan_dev, list);
234 	else
235 		vlan = macvlan_hash_lookup(port, eth->h_dest);
236 	if (vlan == NULL)
237 		return RX_HANDLER_PASS;
238 
239 	dev = vlan->dev;
240 	if (unlikely(!(dev->flags & IFF_UP))) {
241 		kfree_skb(skb);
242 		return RX_HANDLER_CONSUMED;
243 	}
244 	len = skb->len + ETH_HLEN;
245 	skb = skb_share_check(skb, GFP_ATOMIC);
246 	if (!skb)
247 		goto out;
248 
249 	skb->dev = dev;
250 	skb->pkt_type = PACKET_HOST;
251 
252 	ret = vlan->receive(skb);
253 
254 out:
255 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
256 	return RX_HANDLER_CONSUMED;
257 }
258 
259 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
260 {
261 	const struct macvlan_dev *vlan = netdev_priv(dev);
262 	const struct macvlan_port *port = vlan->port;
263 	const struct macvlan_dev *dest;
264 	__u8 ip_summed = skb->ip_summed;
265 
266 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
267 		const struct ethhdr *eth = (void *)skb->data;
268 		skb->ip_summed = CHECKSUM_UNNECESSARY;
269 
270 		/* send to other bridge ports directly */
271 		if (is_multicast_ether_addr(eth->h_dest)) {
272 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
273 			goto xmit_world;
274 		}
275 
276 		dest = macvlan_hash_lookup(port, eth->h_dest);
277 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
278 			/* send to lowerdev first for its network taps */
279 			dev_forward_skb(vlan->lowerdev, skb);
280 
281 			return NET_XMIT_SUCCESS;
282 		}
283 	}
284 
285 xmit_world:
286 	skb->ip_summed = ip_summed;
287 	skb->dev = vlan->lowerdev;
288 	return dev_queue_xmit(skb);
289 }
290 
291 netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
292 			       struct net_device *dev)
293 {
294 	unsigned int len = skb->len;
295 	int ret;
296 	const struct macvlan_dev *vlan = netdev_priv(dev);
297 
298 	ret = macvlan_queue_xmit(skb, dev);
299 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
300 		struct macvlan_pcpu_stats *pcpu_stats;
301 
302 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
303 		u64_stats_update_begin(&pcpu_stats->syncp);
304 		pcpu_stats->tx_packets++;
305 		pcpu_stats->tx_bytes += len;
306 		u64_stats_update_end(&pcpu_stats->syncp);
307 	} else {
308 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
309 	}
310 	return ret;
311 }
312 EXPORT_SYMBOL_GPL(macvlan_start_xmit);
313 
314 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
315 			       unsigned short type, const void *daddr,
316 			       const void *saddr, unsigned len)
317 {
318 	const struct macvlan_dev *vlan = netdev_priv(dev);
319 	struct net_device *lowerdev = vlan->lowerdev;
320 
321 	return dev_hard_header(skb, lowerdev, type, daddr,
322 			       saddr ? : dev->dev_addr, len);
323 }
324 
325 static const struct header_ops macvlan_hard_header_ops = {
326 	.create  	= macvlan_hard_header,
327 	.rebuild	= eth_rebuild_header,
328 	.parse		= eth_header_parse,
329 	.cache		= eth_header_cache,
330 	.cache_update	= eth_header_cache_update,
331 };
332 
333 static int macvlan_open(struct net_device *dev)
334 {
335 	struct macvlan_dev *vlan = netdev_priv(dev);
336 	struct net_device *lowerdev = vlan->lowerdev;
337 	int err;
338 
339 	if (vlan->port->passthru) {
340 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
341 			dev_set_promiscuity(lowerdev, 1);
342 		goto hash_add;
343 	}
344 
345 	err = -EBUSY;
346 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
347 		goto out;
348 
349 	err = dev_uc_add(lowerdev, dev->dev_addr);
350 	if (err < 0)
351 		goto out;
352 	if (dev->flags & IFF_ALLMULTI) {
353 		err = dev_set_allmulti(lowerdev, 1);
354 		if (err < 0)
355 			goto del_unicast;
356 	}
357 
358 hash_add:
359 	macvlan_hash_add(vlan);
360 	return 0;
361 
362 del_unicast:
363 	dev_uc_del(lowerdev, dev->dev_addr);
364 out:
365 	return err;
366 }
367 
368 static int macvlan_stop(struct net_device *dev)
369 {
370 	struct macvlan_dev *vlan = netdev_priv(dev);
371 	struct net_device *lowerdev = vlan->lowerdev;
372 
373 	dev_uc_unsync(lowerdev, dev);
374 	dev_mc_unsync(lowerdev, dev);
375 
376 	if (vlan->port->passthru) {
377 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
378 			dev_set_promiscuity(lowerdev, -1);
379 		goto hash_del;
380 	}
381 
382 	if (dev->flags & IFF_ALLMULTI)
383 		dev_set_allmulti(lowerdev, -1);
384 
385 	dev_uc_del(lowerdev, dev->dev_addr);
386 
387 hash_del:
388 	macvlan_hash_del(vlan, !dev->dismantle);
389 	return 0;
390 }
391 
392 static int macvlan_set_mac_address(struct net_device *dev, void *p)
393 {
394 	struct macvlan_dev *vlan = netdev_priv(dev);
395 	struct net_device *lowerdev = vlan->lowerdev;
396 	struct sockaddr *addr = p;
397 	int err;
398 
399 	if (!is_valid_ether_addr(addr->sa_data))
400 		return -EADDRNOTAVAIL;
401 
402 	if (!(dev->flags & IFF_UP)) {
403 		/* Just copy in the new address */
404 		memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
405 	} else {
406 		/* Rehash and update the device filters */
407 		if (macvlan_addr_busy(vlan->port, addr->sa_data))
408 			return -EBUSY;
409 
410 		err = dev_uc_add(lowerdev, addr->sa_data);
411 		if (err)
412 			return err;
413 
414 		dev_uc_del(lowerdev, dev->dev_addr);
415 
416 		macvlan_hash_change_addr(vlan, addr->sa_data);
417 	}
418 	return 0;
419 }
420 
421 static void macvlan_change_rx_flags(struct net_device *dev, int change)
422 {
423 	struct macvlan_dev *vlan = netdev_priv(dev);
424 	struct net_device *lowerdev = vlan->lowerdev;
425 
426 	if (change & IFF_ALLMULTI)
427 		dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
428 }
429 
430 static void macvlan_set_mac_lists(struct net_device *dev)
431 {
432 	struct macvlan_dev *vlan = netdev_priv(dev);
433 
434 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
435 		bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
436 	} else {
437 		struct netdev_hw_addr *ha;
438 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
439 
440 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
441 		netdev_for_each_mc_addr(ha, dev) {
442 			__set_bit(mc_hash(vlan, ha->addr), filter);
443 		}
444 
445 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
446 
447 		bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
448 	}
449 	dev_uc_sync(vlan->lowerdev, dev);
450 	dev_mc_sync(vlan->lowerdev, dev);
451 }
452 
453 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
454 {
455 	struct macvlan_dev *vlan = netdev_priv(dev);
456 
457 	if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
458 		return -EINVAL;
459 	dev->mtu = new_mtu;
460 	return 0;
461 }
462 
463 /*
464  * macvlan network devices have devices nesting below it and are a special
465  * "super class" of normal network devices; split their locks off into a
466  * separate class since they always nest.
467  */
468 static struct lock_class_key macvlan_netdev_xmit_lock_key;
469 static struct lock_class_key macvlan_netdev_addr_lock_key;
470 
471 #define MACVLAN_FEATURES \
472 	(NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
473 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
474 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
475 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
476 
477 #define MACVLAN_STATE_MASK \
478 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
479 
480 static void macvlan_set_lockdep_class_one(struct net_device *dev,
481 					  struct netdev_queue *txq,
482 					  void *_unused)
483 {
484 	lockdep_set_class(&txq->_xmit_lock,
485 			  &macvlan_netdev_xmit_lock_key);
486 }
487 
488 static void macvlan_set_lockdep_class(struct net_device *dev)
489 {
490 	lockdep_set_class(&dev->addr_list_lock,
491 			  &macvlan_netdev_addr_lock_key);
492 	netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
493 }
494 
495 static int macvlan_init(struct net_device *dev)
496 {
497 	struct macvlan_dev *vlan = netdev_priv(dev);
498 	const struct net_device *lowerdev = vlan->lowerdev;
499 
500 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
501 				  (lowerdev->state & MACVLAN_STATE_MASK);
502 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
503 	dev->features		|= NETIF_F_LLTX;
504 	dev->gso_max_size	= lowerdev->gso_max_size;
505 	dev->iflink		= lowerdev->ifindex;
506 	dev->hard_header_len	= lowerdev->hard_header_len;
507 
508 	macvlan_set_lockdep_class(dev);
509 
510 	vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
511 	if (!vlan->pcpu_stats)
512 		return -ENOMEM;
513 
514 	return 0;
515 }
516 
517 static void macvlan_uninit(struct net_device *dev)
518 {
519 	struct macvlan_dev *vlan = netdev_priv(dev);
520 	struct macvlan_port *port = vlan->port;
521 
522 	free_percpu(vlan->pcpu_stats);
523 
524 	port->count -= 1;
525 	if (!port->count)
526 		macvlan_port_destroy(port->dev);
527 }
528 
529 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
530 							 struct rtnl_link_stats64 *stats)
531 {
532 	struct macvlan_dev *vlan = netdev_priv(dev);
533 
534 	if (vlan->pcpu_stats) {
535 		struct macvlan_pcpu_stats *p;
536 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
537 		u32 rx_errors = 0, tx_dropped = 0;
538 		unsigned int start;
539 		int i;
540 
541 		for_each_possible_cpu(i) {
542 			p = per_cpu_ptr(vlan->pcpu_stats, i);
543 			do {
544 				start = u64_stats_fetch_begin_bh(&p->syncp);
545 				rx_packets	= p->rx_packets;
546 				rx_bytes	= p->rx_bytes;
547 				rx_multicast	= p->rx_multicast;
548 				tx_packets	= p->tx_packets;
549 				tx_bytes	= p->tx_bytes;
550 			} while (u64_stats_fetch_retry_bh(&p->syncp, start));
551 
552 			stats->rx_packets	+= rx_packets;
553 			stats->rx_bytes		+= rx_bytes;
554 			stats->multicast	+= rx_multicast;
555 			stats->tx_packets	+= tx_packets;
556 			stats->tx_bytes		+= tx_bytes;
557 			/* rx_errors & tx_dropped are u32, updated
558 			 * without syncp protection.
559 			 */
560 			rx_errors	+= p->rx_errors;
561 			tx_dropped	+= p->tx_dropped;
562 		}
563 		stats->rx_errors	= rx_errors;
564 		stats->rx_dropped	= rx_errors;
565 		stats->tx_dropped	= tx_dropped;
566 	}
567 	return stats;
568 }
569 
570 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
571 				   __be16 proto, u16 vid)
572 {
573 	struct macvlan_dev *vlan = netdev_priv(dev);
574 	struct net_device *lowerdev = vlan->lowerdev;
575 
576 	return vlan_vid_add(lowerdev, proto, vid);
577 }
578 
579 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
580 				    __be16 proto, u16 vid)
581 {
582 	struct macvlan_dev *vlan = netdev_priv(dev);
583 	struct net_device *lowerdev = vlan->lowerdev;
584 
585 	vlan_vid_del(lowerdev, proto, vid);
586 	return 0;
587 }
588 
589 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
590 			   struct net_device *dev,
591 			   const unsigned char *addr,
592 			   u16 flags)
593 {
594 	struct macvlan_dev *vlan = netdev_priv(dev);
595 	int err = -EINVAL;
596 
597 	if (!vlan->port->passthru)
598 		return -EOPNOTSUPP;
599 
600 	if (is_unicast_ether_addr(addr))
601 		err = dev_uc_add_excl(dev, addr);
602 	else if (is_multicast_ether_addr(addr))
603 		err = dev_mc_add_excl(dev, addr);
604 
605 	return err;
606 }
607 
608 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
609 			   struct net_device *dev,
610 			   const unsigned char *addr)
611 {
612 	struct macvlan_dev *vlan = netdev_priv(dev);
613 	int err = -EINVAL;
614 
615 	if (!vlan->port->passthru)
616 		return -EOPNOTSUPP;
617 
618 	if (is_unicast_ether_addr(addr))
619 		err = dev_uc_del(dev, addr);
620 	else if (is_multicast_ether_addr(addr))
621 		err = dev_mc_del(dev, addr);
622 
623 	return err;
624 }
625 
626 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
627 					struct ethtool_drvinfo *drvinfo)
628 {
629 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
630 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
631 }
632 
633 static int macvlan_ethtool_get_settings(struct net_device *dev,
634 					struct ethtool_cmd *cmd)
635 {
636 	const struct macvlan_dev *vlan = netdev_priv(dev);
637 
638 	return __ethtool_get_settings(vlan->lowerdev, cmd);
639 }
640 
641 static const struct ethtool_ops macvlan_ethtool_ops = {
642 	.get_link		= ethtool_op_get_link,
643 	.get_settings		= macvlan_ethtool_get_settings,
644 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
645 };
646 
647 static const struct net_device_ops macvlan_netdev_ops = {
648 	.ndo_init		= macvlan_init,
649 	.ndo_uninit		= macvlan_uninit,
650 	.ndo_open		= macvlan_open,
651 	.ndo_stop		= macvlan_stop,
652 	.ndo_start_xmit		= macvlan_start_xmit,
653 	.ndo_change_mtu		= macvlan_change_mtu,
654 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
655 	.ndo_set_mac_address	= macvlan_set_mac_address,
656 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
657 	.ndo_get_stats64	= macvlan_dev_get_stats64,
658 	.ndo_validate_addr	= eth_validate_addr,
659 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
660 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
661 	.ndo_fdb_add		= macvlan_fdb_add,
662 	.ndo_fdb_del		= macvlan_fdb_del,
663 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
664 };
665 
666 void macvlan_common_setup(struct net_device *dev)
667 {
668 	ether_setup(dev);
669 
670 	dev->priv_flags	       &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
671 	dev->priv_flags	       |= IFF_UNICAST_FLT;
672 	dev->netdev_ops		= &macvlan_netdev_ops;
673 	dev->destructor		= free_netdev;
674 	dev->header_ops		= &macvlan_hard_header_ops,
675 	dev->ethtool_ops	= &macvlan_ethtool_ops;
676 }
677 EXPORT_SYMBOL_GPL(macvlan_common_setup);
678 
679 static void macvlan_setup(struct net_device *dev)
680 {
681 	macvlan_common_setup(dev);
682 	dev->tx_queue_len	= 0;
683 }
684 
685 static int macvlan_port_create(struct net_device *dev)
686 {
687 	struct macvlan_port *port;
688 	unsigned int i;
689 	int err;
690 
691 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
692 		return -EINVAL;
693 
694 	port = kzalloc(sizeof(*port), GFP_KERNEL);
695 	if (port == NULL)
696 		return -ENOMEM;
697 
698 	port->passthru = false;
699 	port->dev = dev;
700 	INIT_LIST_HEAD(&port->vlans);
701 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
702 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
703 
704 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
705 	if (err)
706 		kfree(port);
707 	else
708 		dev->priv_flags |= IFF_MACVLAN_PORT;
709 	return err;
710 }
711 
712 static void macvlan_port_destroy(struct net_device *dev)
713 {
714 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
715 
716 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
717 	netdev_rx_handler_unregister(dev);
718 	kfree_rcu(port, rcu);
719 }
720 
721 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
722 {
723 	if (tb[IFLA_ADDRESS]) {
724 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
725 			return -EINVAL;
726 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
727 			return -EADDRNOTAVAIL;
728 	}
729 
730 	if (data && data[IFLA_MACVLAN_MODE]) {
731 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
732 		case MACVLAN_MODE_PRIVATE:
733 		case MACVLAN_MODE_VEPA:
734 		case MACVLAN_MODE_BRIDGE:
735 		case MACVLAN_MODE_PASSTHRU:
736 			break;
737 		default:
738 			return -EINVAL;
739 		}
740 	}
741 	return 0;
742 }
743 
744 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
745 			   struct nlattr *tb[], struct nlattr *data[],
746 			   int (*receive)(struct sk_buff *skb),
747 			   int (*forward)(struct net_device *dev,
748 					  struct sk_buff *skb))
749 {
750 	struct macvlan_dev *vlan = netdev_priv(dev);
751 	struct macvlan_port *port;
752 	struct net_device *lowerdev;
753 	int err;
754 
755 	if (!tb[IFLA_LINK])
756 		return -EINVAL;
757 
758 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
759 	if (lowerdev == NULL)
760 		return -ENODEV;
761 
762 	/* When creating macvlans on top of other macvlans - use
763 	 * the real device as the lowerdev.
764 	 */
765 	if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
766 		struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
767 		lowerdev = lowervlan->lowerdev;
768 	}
769 
770 	if (!tb[IFLA_MTU])
771 		dev->mtu = lowerdev->mtu;
772 	else if (dev->mtu > lowerdev->mtu)
773 		return -EINVAL;
774 
775 	if (!tb[IFLA_ADDRESS])
776 		eth_hw_addr_random(dev);
777 
778 	if (!macvlan_port_exists(lowerdev)) {
779 		err = macvlan_port_create(lowerdev);
780 		if (err < 0)
781 			return err;
782 	}
783 	port = macvlan_port_get_rtnl(lowerdev);
784 
785 	/* Only 1 macvlan device can be created in passthru mode */
786 	if (port->passthru)
787 		return -EINVAL;
788 
789 	vlan->lowerdev = lowerdev;
790 	vlan->dev      = dev;
791 	vlan->port     = port;
792 	vlan->receive  = receive;
793 	vlan->forward  = forward;
794 
795 	vlan->mode     = MACVLAN_MODE_VEPA;
796 	if (data && data[IFLA_MACVLAN_MODE])
797 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
798 
799 	if (data && data[IFLA_MACVLAN_FLAGS])
800 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
801 
802 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
803 		if (port->count)
804 			return -EINVAL;
805 		port->passthru = true;
806 		memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN);
807 	}
808 
809 	err = netdev_upper_dev_link(lowerdev, dev);
810 	if (err)
811 		goto destroy_port;
812 
813 	port->count += 1;
814 	err = register_netdevice(dev);
815 	if (err < 0)
816 		goto upper_dev_unlink;
817 
818 	list_add_tail_rcu(&vlan->list, &port->vlans);
819 	netif_stacked_transfer_operstate(lowerdev, dev);
820 
821 	return 0;
822 
823 upper_dev_unlink:
824 	netdev_upper_dev_unlink(lowerdev, dev);
825 destroy_port:
826 	port->count -= 1;
827 	if (!port->count)
828 		macvlan_port_destroy(lowerdev);
829 
830 	return err;
831 }
832 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
833 
834 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
835 			   struct nlattr *tb[], struct nlattr *data[])
836 {
837 	return macvlan_common_newlink(src_net, dev, tb, data,
838 				      netif_rx,
839 				      dev_forward_skb);
840 }
841 
842 void macvlan_dellink(struct net_device *dev, struct list_head *head)
843 {
844 	struct macvlan_dev *vlan = netdev_priv(dev);
845 
846 	list_del_rcu(&vlan->list);
847 	unregister_netdevice_queue(dev, head);
848 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
849 }
850 EXPORT_SYMBOL_GPL(macvlan_dellink);
851 
852 static int macvlan_changelink(struct net_device *dev,
853 		struct nlattr *tb[], struct nlattr *data[])
854 {
855 	struct macvlan_dev *vlan = netdev_priv(dev);
856 	if (data && data[IFLA_MACVLAN_MODE])
857 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
858 	if (data && data[IFLA_MACVLAN_FLAGS]) {
859 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
860 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
861 
862 		if (promisc && (flags & MACVLAN_FLAG_NOPROMISC))
863 			dev_set_promiscuity(vlan->lowerdev, -1);
864 		else if (promisc && !(flags & MACVLAN_FLAG_NOPROMISC))
865 			dev_set_promiscuity(vlan->lowerdev, 1);
866 		vlan->flags = flags;
867 	}
868 	return 0;
869 }
870 
871 static size_t macvlan_get_size(const struct net_device *dev)
872 {
873 	return (0
874 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
875 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
876 		);
877 }
878 
879 static int macvlan_fill_info(struct sk_buff *skb,
880 				const struct net_device *dev)
881 {
882 	struct macvlan_dev *vlan = netdev_priv(dev);
883 
884 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
885 		goto nla_put_failure;
886 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
887 		goto nla_put_failure;
888 	return 0;
889 
890 nla_put_failure:
891 	return -EMSGSIZE;
892 }
893 
894 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
895 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
896 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
897 };
898 
899 int macvlan_link_register(struct rtnl_link_ops *ops)
900 {
901 	/* common fields */
902 	ops->priv_size		= sizeof(struct macvlan_dev);
903 	ops->validate		= macvlan_validate;
904 	ops->maxtype		= IFLA_MACVLAN_MAX;
905 	ops->policy		= macvlan_policy;
906 	ops->changelink		= macvlan_changelink;
907 	ops->get_size		= macvlan_get_size;
908 	ops->fill_info		= macvlan_fill_info;
909 
910 	return rtnl_link_register(ops);
911 };
912 EXPORT_SYMBOL_GPL(macvlan_link_register);
913 
914 static struct rtnl_link_ops macvlan_link_ops = {
915 	.kind		= "macvlan",
916 	.setup		= macvlan_setup,
917 	.newlink	= macvlan_newlink,
918 	.dellink	= macvlan_dellink,
919 };
920 
921 static int macvlan_device_event(struct notifier_block *unused,
922 				unsigned long event, void *ptr)
923 {
924 	struct net_device *dev = ptr;
925 	struct macvlan_dev *vlan, *next;
926 	struct macvlan_port *port;
927 	LIST_HEAD(list_kill);
928 
929 	if (!macvlan_port_exists(dev))
930 		return NOTIFY_DONE;
931 
932 	port = macvlan_port_get_rtnl(dev);
933 
934 	switch (event) {
935 	case NETDEV_CHANGE:
936 		list_for_each_entry(vlan, &port->vlans, list)
937 			netif_stacked_transfer_operstate(vlan->lowerdev,
938 							 vlan->dev);
939 		break;
940 	case NETDEV_FEAT_CHANGE:
941 		list_for_each_entry(vlan, &port->vlans, list) {
942 			vlan->dev->features = dev->features & MACVLAN_FEATURES;
943 			vlan->dev->gso_max_size = dev->gso_max_size;
944 			netdev_features_change(vlan->dev);
945 		}
946 		break;
947 	case NETDEV_UNREGISTER:
948 		/* twiddle thumbs on netns device moves */
949 		if (dev->reg_state != NETREG_UNREGISTERING)
950 			break;
951 
952 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
953 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
954 		unregister_netdevice_many(&list_kill);
955 		list_del(&list_kill);
956 		break;
957 	case NETDEV_PRE_TYPE_CHANGE:
958 		/* Forbid underlaying device to change its type. */
959 		return NOTIFY_BAD;
960 	}
961 	return NOTIFY_DONE;
962 }
963 
964 static struct notifier_block macvlan_notifier_block __read_mostly = {
965 	.notifier_call	= macvlan_device_event,
966 };
967 
968 static int __init macvlan_init_module(void)
969 {
970 	int err;
971 
972 	register_netdevice_notifier(&macvlan_notifier_block);
973 
974 	err = macvlan_link_register(&macvlan_link_ops);
975 	if (err < 0)
976 		goto err1;
977 	return 0;
978 err1:
979 	unregister_netdevice_notifier(&macvlan_notifier_block);
980 	return err;
981 }
982 
983 static void __exit macvlan_cleanup_module(void)
984 {
985 	rtnl_link_unregister(&macvlan_link_ops);
986 	unregister_netdevice_notifier(&macvlan_notifier_block);
987 }
988 
989 module_init(macvlan_init_module);
990 module_exit(macvlan_cleanup_module);
991 
992 MODULE_LICENSE("GPL");
993 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
994 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
995 MODULE_ALIAS_RTNL_LINK("macvlan");
996