xref: /openbmc/linux/drivers/net/macvlan.c (revision 80ecbd24)
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 			err = dev_set_promiscuity(lowerdev, 1);
342 			if (err < 0)
343 				goto out;
344 		}
345 		goto hash_add;
346 	}
347 
348 	err = -EBUSY;
349 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
350 		goto out;
351 
352 	err = dev_uc_add(lowerdev, dev->dev_addr);
353 	if (err < 0)
354 		goto out;
355 	if (dev->flags & IFF_ALLMULTI) {
356 		err = dev_set_allmulti(lowerdev, 1);
357 		if (err < 0)
358 			goto del_unicast;
359 	}
360 
361 hash_add:
362 	macvlan_hash_add(vlan);
363 	return 0;
364 
365 del_unicast:
366 	dev_uc_del(lowerdev, dev->dev_addr);
367 out:
368 	return err;
369 }
370 
371 static int macvlan_stop(struct net_device *dev)
372 {
373 	struct macvlan_dev *vlan = netdev_priv(dev);
374 	struct net_device *lowerdev = vlan->lowerdev;
375 
376 	dev_uc_unsync(lowerdev, dev);
377 	dev_mc_unsync(lowerdev, dev);
378 
379 	if (vlan->port->passthru) {
380 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
381 			dev_set_promiscuity(lowerdev, -1);
382 		goto hash_del;
383 	}
384 
385 	if (dev->flags & IFF_ALLMULTI)
386 		dev_set_allmulti(lowerdev, -1);
387 
388 	dev_uc_del(lowerdev, dev->dev_addr);
389 
390 hash_del:
391 	macvlan_hash_del(vlan, !dev->dismantle);
392 	return 0;
393 }
394 
395 static int macvlan_set_mac_address(struct net_device *dev, void *p)
396 {
397 	struct macvlan_dev *vlan = netdev_priv(dev);
398 	struct net_device *lowerdev = vlan->lowerdev;
399 	struct sockaddr *addr = p;
400 	int err;
401 
402 	if (!is_valid_ether_addr(addr->sa_data))
403 		return -EADDRNOTAVAIL;
404 
405 	if (!(dev->flags & IFF_UP)) {
406 		/* Just copy in the new address */
407 		memcpy(dev->dev_addr, addr->sa_data, ETH_ALEN);
408 	} else {
409 		/* Rehash and update the device filters */
410 		if (macvlan_addr_busy(vlan->port, addr->sa_data))
411 			return -EBUSY;
412 
413 		err = dev_uc_add(lowerdev, addr->sa_data);
414 		if (err)
415 			return err;
416 
417 		dev_uc_del(lowerdev, dev->dev_addr);
418 
419 		macvlan_hash_change_addr(vlan, addr->sa_data);
420 	}
421 	return 0;
422 }
423 
424 static void macvlan_change_rx_flags(struct net_device *dev, int change)
425 {
426 	struct macvlan_dev *vlan = netdev_priv(dev);
427 	struct net_device *lowerdev = vlan->lowerdev;
428 
429 	if (change & IFF_ALLMULTI)
430 		dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
431 }
432 
433 static void macvlan_set_mac_lists(struct net_device *dev)
434 {
435 	struct macvlan_dev *vlan = netdev_priv(dev);
436 
437 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
438 		bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
439 	} else {
440 		struct netdev_hw_addr *ha;
441 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
442 
443 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
444 		netdev_for_each_mc_addr(ha, dev) {
445 			__set_bit(mc_hash(vlan, ha->addr), filter);
446 		}
447 
448 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
449 
450 		bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
451 	}
452 	dev_uc_sync(vlan->lowerdev, dev);
453 	dev_mc_sync(vlan->lowerdev, dev);
454 }
455 
456 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
457 {
458 	struct macvlan_dev *vlan = netdev_priv(dev);
459 
460 	if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
461 		return -EINVAL;
462 	dev->mtu = new_mtu;
463 	return 0;
464 }
465 
466 /*
467  * macvlan network devices have devices nesting below it and are a special
468  * "super class" of normal network devices; split their locks off into a
469  * separate class since they always nest.
470  */
471 static struct lock_class_key macvlan_netdev_xmit_lock_key;
472 static struct lock_class_key macvlan_netdev_addr_lock_key;
473 
474 #define MACVLAN_FEATURES \
475 	(NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
476 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
477 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
478 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
479 
480 #define MACVLAN_STATE_MASK \
481 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
482 
483 static void macvlan_set_lockdep_class_one(struct net_device *dev,
484 					  struct netdev_queue *txq,
485 					  void *_unused)
486 {
487 	lockdep_set_class(&txq->_xmit_lock,
488 			  &macvlan_netdev_xmit_lock_key);
489 }
490 
491 static void macvlan_set_lockdep_class(struct net_device *dev)
492 {
493 	lockdep_set_class(&dev->addr_list_lock,
494 			  &macvlan_netdev_addr_lock_key);
495 	netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
496 }
497 
498 static int macvlan_init(struct net_device *dev)
499 {
500 	struct macvlan_dev *vlan = netdev_priv(dev);
501 	const struct net_device *lowerdev = vlan->lowerdev;
502 
503 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
504 				  (lowerdev->state & MACVLAN_STATE_MASK);
505 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
506 	dev->features		|= NETIF_F_LLTX;
507 	dev->gso_max_size	= lowerdev->gso_max_size;
508 	dev->iflink		= lowerdev->ifindex;
509 	dev->hard_header_len	= lowerdev->hard_header_len;
510 
511 	macvlan_set_lockdep_class(dev);
512 
513 	vlan->pcpu_stats = alloc_percpu(struct macvlan_pcpu_stats);
514 	if (!vlan->pcpu_stats)
515 		return -ENOMEM;
516 
517 	return 0;
518 }
519 
520 static void macvlan_uninit(struct net_device *dev)
521 {
522 	struct macvlan_dev *vlan = netdev_priv(dev);
523 	struct macvlan_port *port = vlan->port;
524 
525 	free_percpu(vlan->pcpu_stats);
526 
527 	port->count -= 1;
528 	if (!port->count)
529 		macvlan_port_destroy(port->dev);
530 }
531 
532 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
533 							 struct rtnl_link_stats64 *stats)
534 {
535 	struct macvlan_dev *vlan = netdev_priv(dev);
536 
537 	if (vlan->pcpu_stats) {
538 		struct macvlan_pcpu_stats *p;
539 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
540 		u32 rx_errors = 0, tx_dropped = 0;
541 		unsigned int start;
542 		int i;
543 
544 		for_each_possible_cpu(i) {
545 			p = per_cpu_ptr(vlan->pcpu_stats, i);
546 			do {
547 				start = u64_stats_fetch_begin_bh(&p->syncp);
548 				rx_packets	= p->rx_packets;
549 				rx_bytes	= p->rx_bytes;
550 				rx_multicast	= p->rx_multicast;
551 				tx_packets	= p->tx_packets;
552 				tx_bytes	= p->tx_bytes;
553 			} while (u64_stats_fetch_retry_bh(&p->syncp, start));
554 
555 			stats->rx_packets	+= rx_packets;
556 			stats->rx_bytes		+= rx_bytes;
557 			stats->multicast	+= rx_multicast;
558 			stats->tx_packets	+= tx_packets;
559 			stats->tx_bytes		+= tx_bytes;
560 			/* rx_errors & tx_dropped are u32, updated
561 			 * without syncp protection.
562 			 */
563 			rx_errors	+= p->rx_errors;
564 			tx_dropped	+= p->tx_dropped;
565 		}
566 		stats->rx_errors	= rx_errors;
567 		stats->rx_dropped	= rx_errors;
568 		stats->tx_dropped	= tx_dropped;
569 	}
570 	return stats;
571 }
572 
573 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
574 				   __be16 proto, u16 vid)
575 {
576 	struct macvlan_dev *vlan = netdev_priv(dev);
577 	struct net_device *lowerdev = vlan->lowerdev;
578 
579 	return vlan_vid_add(lowerdev, proto, vid);
580 }
581 
582 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
583 				    __be16 proto, u16 vid)
584 {
585 	struct macvlan_dev *vlan = netdev_priv(dev);
586 	struct net_device *lowerdev = vlan->lowerdev;
587 
588 	vlan_vid_del(lowerdev, proto, vid);
589 	return 0;
590 }
591 
592 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
593 			   struct net_device *dev,
594 			   const unsigned char *addr,
595 			   u16 flags)
596 {
597 	struct macvlan_dev *vlan = netdev_priv(dev);
598 	int err = -EINVAL;
599 
600 	if (!vlan->port->passthru)
601 		return -EOPNOTSUPP;
602 
603 	if (is_unicast_ether_addr(addr))
604 		err = dev_uc_add_excl(dev, addr);
605 	else if (is_multicast_ether_addr(addr))
606 		err = dev_mc_add_excl(dev, addr);
607 
608 	return err;
609 }
610 
611 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
612 			   struct net_device *dev,
613 			   const unsigned char *addr)
614 {
615 	struct macvlan_dev *vlan = netdev_priv(dev);
616 	int err = -EINVAL;
617 
618 	if (!vlan->port->passthru)
619 		return -EOPNOTSUPP;
620 
621 	if (is_unicast_ether_addr(addr))
622 		err = dev_uc_del(dev, addr);
623 	else if (is_multicast_ether_addr(addr))
624 		err = dev_mc_del(dev, addr);
625 
626 	return err;
627 }
628 
629 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
630 					struct ethtool_drvinfo *drvinfo)
631 {
632 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
633 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
634 }
635 
636 static int macvlan_ethtool_get_settings(struct net_device *dev,
637 					struct ethtool_cmd *cmd)
638 {
639 	const struct macvlan_dev *vlan = netdev_priv(dev);
640 
641 	return __ethtool_get_settings(vlan->lowerdev, cmd);
642 }
643 
644 static netdev_features_t macvlan_fix_features(struct net_device *dev,
645 					      netdev_features_t features)
646 {
647 	struct macvlan_dev *vlan = netdev_priv(dev);
648 
649 	return features & (vlan->set_features | ~MACVLAN_FEATURES);
650 }
651 
652 static const struct ethtool_ops macvlan_ethtool_ops = {
653 	.get_link		= ethtool_op_get_link,
654 	.get_settings		= macvlan_ethtool_get_settings,
655 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
656 };
657 
658 static const struct net_device_ops macvlan_netdev_ops = {
659 	.ndo_init		= macvlan_init,
660 	.ndo_uninit		= macvlan_uninit,
661 	.ndo_open		= macvlan_open,
662 	.ndo_stop		= macvlan_stop,
663 	.ndo_start_xmit		= macvlan_start_xmit,
664 	.ndo_change_mtu		= macvlan_change_mtu,
665 	.ndo_fix_features	= macvlan_fix_features,
666 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
667 	.ndo_set_mac_address	= macvlan_set_mac_address,
668 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
669 	.ndo_get_stats64	= macvlan_dev_get_stats64,
670 	.ndo_validate_addr	= eth_validate_addr,
671 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
672 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
673 	.ndo_fdb_add		= macvlan_fdb_add,
674 	.ndo_fdb_del		= macvlan_fdb_del,
675 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
676 };
677 
678 void macvlan_common_setup(struct net_device *dev)
679 {
680 	ether_setup(dev);
681 
682 	dev->priv_flags	       &= ~(IFF_XMIT_DST_RELEASE | IFF_TX_SKB_SHARING);
683 	dev->priv_flags	       |= IFF_UNICAST_FLT;
684 	dev->netdev_ops		= &macvlan_netdev_ops;
685 	dev->destructor		= free_netdev;
686 	dev->header_ops		= &macvlan_hard_header_ops,
687 	dev->ethtool_ops	= &macvlan_ethtool_ops;
688 }
689 EXPORT_SYMBOL_GPL(macvlan_common_setup);
690 
691 static void macvlan_setup(struct net_device *dev)
692 {
693 	macvlan_common_setup(dev);
694 	dev->tx_queue_len	= 0;
695 }
696 
697 static int macvlan_port_create(struct net_device *dev)
698 {
699 	struct macvlan_port *port;
700 	unsigned int i;
701 	int err;
702 
703 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
704 		return -EINVAL;
705 
706 	port = kzalloc(sizeof(*port), GFP_KERNEL);
707 	if (port == NULL)
708 		return -ENOMEM;
709 
710 	port->passthru = false;
711 	port->dev = dev;
712 	INIT_LIST_HEAD(&port->vlans);
713 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
714 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
715 
716 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
717 	if (err)
718 		kfree(port);
719 	else
720 		dev->priv_flags |= IFF_MACVLAN_PORT;
721 	return err;
722 }
723 
724 static void macvlan_port_destroy(struct net_device *dev)
725 {
726 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
727 
728 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
729 	netdev_rx_handler_unregister(dev);
730 	kfree_rcu(port, rcu);
731 }
732 
733 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
734 {
735 	if (tb[IFLA_ADDRESS]) {
736 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
737 			return -EINVAL;
738 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
739 			return -EADDRNOTAVAIL;
740 	}
741 
742 	if (data && data[IFLA_MACVLAN_FLAGS] &&
743 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
744 		return -EINVAL;
745 
746 	if (data && data[IFLA_MACVLAN_MODE]) {
747 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
748 		case MACVLAN_MODE_PRIVATE:
749 		case MACVLAN_MODE_VEPA:
750 		case MACVLAN_MODE_BRIDGE:
751 		case MACVLAN_MODE_PASSTHRU:
752 			break;
753 		default:
754 			return -EINVAL;
755 		}
756 	}
757 	return 0;
758 }
759 
760 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
761 			   struct nlattr *tb[], struct nlattr *data[],
762 			   int (*receive)(struct sk_buff *skb),
763 			   int (*forward)(struct net_device *dev,
764 					  struct sk_buff *skb))
765 {
766 	struct macvlan_dev *vlan = netdev_priv(dev);
767 	struct macvlan_port *port;
768 	struct net_device *lowerdev;
769 	int err;
770 
771 	if (!tb[IFLA_LINK])
772 		return -EINVAL;
773 
774 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
775 	if (lowerdev == NULL)
776 		return -ENODEV;
777 
778 	/* When creating macvlans on top of other macvlans - use
779 	 * the real device as the lowerdev.
780 	 */
781 	if (lowerdev->rtnl_link_ops == dev->rtnl_link_ops) {
782 		struct macvlan_dev *lowervlan = netdev_priv(lowerdev);
783 		lowerdev = lowervlan->lowerdev;
784 	}
785 
786 	if (!tb[IFLA_MTU])
787 		dev->mtu = lowerdev->mtu;
788 	else if (dev->mtu > lowerdev->mtu)
789 		return -EINVAL;
790 
791 	if (!tb[IFLA_ADDRESS])
792 		eth_hw_addr_random(dev);
793 
794 	if (!macvlan_port_exists(lowerdev)) {
795 		err = macvlan_port_create(lowerdev);
796 		if (err < 0)
797 			return err;
798 	}
799 	port = macvlan_port_get_rtnl(lowerdev);
800 
801 	/* Only 1 macvlan device can be created in passthru mode */
802 	if (port->passthru)
803 		return -EINVAL;
804 
805 	vlan->lowerdev = lowerdev;
806 	vlan->dev      = dev;
807 	vlan->port     = port;
808 	vlan->receive  = receive;
809 	vlan->forward  = forward;
810 	vlan->set_features = MACVLAN_FEATURES;
811 
812 	vlan->mode     = MACVLAN_MODE_VEPA;
813 	if (data && data[IFLA_MACVLAN_MODE])
814 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
815 
816 	if (data && data[IFLA_MACVLAN_FLAGS])
817 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
818 
819 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
820 		if (port->count)
821 			return -EINVAL;
822 		port->passthru = true;
823 		memcpy(dev->dev_addr, lowerdev->dev_addr, ETH_ALEN);
824 	}
825 
826 	err = netdev_upper_dev_link(lowerdev, dev);
827 	if (err)
828 		goto destroy_port;
829 
830 	port->count += 1;
831 	err = register_netdevice(dev);
832 	if (err < 0)
833 		goto upper_dev_unlink;
834 
835 	list_add_tail_rcu(&vlan->list, &port->vlans);
836 	netif_stacked_transfer_operstate(lowerdev, dev);
837 
838 	return 0;
839 
840 upper_dev_unlink:
841 	netdev_upper_dev_unlink(lowerdev, dev);
842 destroy_port:
843 	port->count -= 1;
844 	if (!port->count)
845 		macvlan_port_destroy(lowerdev);
846 
847 	return err;
848 }
849 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
850 
851 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
852 			   struct nlattr *tb[], struct nlattr *data[])
853 {
854 	return macvlan_common_newlink(src_net, dev, tb, data,
855 				      netif_rx,
856 				      dev_forward_skb);
857 }
858 
859 void macvlan_dellink(struct net_device *dev, struct list_head *head)
860 {
861 	struct macvlan_dev *vlan = netdev_priv(dev);
862 
863 	list_del_rcu(&vlan->list);
864 	unregister_netdevice_queue(dev, head);
865 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
866 }
867 EXPORT_SYMBOL_GPL(macvlan_dellink);
868 
869 static int macvlan_changelink(struct net_device *dev,
870 		struct nlattr *tb[], struct nlattr *data[])
871 {
872 	struct macvlan_dev *vlan = netdev_priv(dev);
873 	enum macvlan_mode mode;
874 	bool set_mode = false;
875 
876 	/* Validate mode, but don't set yet: setting flags may fail. */
877 	if (data && data[IFLA_MACVLAN_MODE]) {
878 		set_mode = true;
879 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
880 		/* Passthrough mode can't be set or cleared dynamically */
881 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
882 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
883 			return -EINVAL;
884 	}
885 
886 	if (data && data[IFLA_MACVLAN_FLAGS]) {
887 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
888 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
889 		if (vlan->port->passthru && promisc) {
890 			int err;
891 
892 			if (flags & MACVLAN_FLAG_NOPROMISC)
893 				err = dev_set_promiscuity(vlan->lowerdev, -1);
894 			else
895 				err = dev_set_promiscuity(vlan->lowerdev, 1);
896 			if (err < 0)
897 				return err;
898 		}
899 		vlan->flags = flags;
900 	}
901 	if (set_mode)
902 		vlan->mode = mode;
903 	return 0;
904 }
905 
906 static size_t macvlan_get_size(const struct net_device *dev)
907 {
908 	return (0
909 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
910 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
911 		);
912 }
913 
914 static int macvlan_fill_info(struct sk_buff *skb,
915 				const struct net_device *dev)
916 {
917 	struct macvlan_dev *vlan = netdev_priv(dev);
918 
919 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
920 		goto nla_put_failure;
921 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
922 		goto nla_put_failure;
923 	return 0;
924 
925 nla_put_failure:
926 	return -EMSGSIZE;
927 }
928 
929 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
930 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
931 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
932 };
933 
934 int macvlan_link_register(struct rtnl_link_ops *ops)
935 {
936 	/* common fields */
937 	ops->priv_size		= sizeof(struct macvlan_dev);
938 	ops->validate		= macvlan_validate;
939 	ops->maxtype		= IFLA_MACVLAN_MAX;
940 	ops->policy		= macvlan_policy;
941 	ops->changelink		= macvlan_changelink;
942 	ops->get_size		= macvlan_get_size;
943 	ops->fill_info		= macvlan_fill_info;
944 
945 	return rtnl_link_register(ops);
946 };
947 EXPORT_SYMBOL_GPL(macvlan_link_register);
948 
949 static struct rtnl_link_ops macvlan_link_ops = {
950 	.kind		= "macvlan",
951 	.setup		= macvlan_setup,
952 	.newlink	= macvlan_newlink,
953 	.dellink	= macvlan_dellink,
954 };
955 
956 static int macvlan_device_event(struct notifier_block *unused,
957 				unsigned long event, void *ptr)
958 {
959 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
960 	struct macvlan_dev *vlan, *next;
961 	struct macvlan_port *port;
962 	LIST_HEAD(list_kill);
963 
964 	if (!macvlan_port_exists(dev))
965 		return NOTIFY_DONE;
966 
967 	port = macvlan_port_get_rtnl(dev);
968 
969 	switch (event) {
970 	case NETDEV_CHANGE:
971 		list_for_each_entry(vlan, &port->vlans, list)
972 			netif_stacked_transfer_operstate(vlan->lowerdev,
973 							 vlan->dev);
974 		break;
975 	case NETDEV_FEAT_CHANGE:
976 		list_for_each_entry(vlan, &port->vlans, list) {
977 			vlan->dev->features = dev->features & MACVLAN_FEATURES;
978 			vlan->dev->gso_max_size = dev->gso_max_size;
979 			netdev_features_change(vlan->dev);
980 		}
981 		break;
982 	case NETDEV_UNREGISTER:
983 		/* twiddle thumbs on netns device moves */
984 		if (dev->reg_state != NETREG_UNREGISTERING)
985 			break;
986 
987 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
988 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
989 		unregister_netdevice_many(&list_kill);
990 		list_del(&list_kill);
991 		break;
992 	case NETDEV_PRE_TYPE_CHANGE:
993 		/* Forbid underlaying device to change its type. */
994 		return NOTIFY_BAD;
995 	}
996 	return NOTIFY_DONE;
997 }
998 
999 static struct notifier_block macvlan_notifier_block __read_mostly = {
1000 	.notifier_call	= macvlan_device_event,
1001 };
1002 
1003 static int __init macvlan_init_module(void)
1004 {
1005 	int err;
1006 
1007 	register_netdevice_notifier(&macvlan_notifier_block);
1008 
1009 	err = macvlan_link_register(&macvlan_link_ops);
1010 	if (err < 0)
1011 		goto err1;
1012 	return 0;
1013 err1:
1014 	unregister_netdevice_notifier(&macvlan_notifier_block);
1015 	return err;
1016 }
1017 
1018 static void __exit macvlan_cleanup_module(void)
1019 {
1020 	rtnl_link_unregister(&macvlan_link_ops);
1021 	unregister_netdevice_notifier(&macvlan_notifier_block);
1022 }
1023 
1024 module_init(macvlan_init_module);
1025 module_exit(macvlan_cleanup_module);
1026 
1027 MODULE_LICENSE("GPL");
1028 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1029 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1030 MODULE_ALIAS_RTNL_LINK("macvlan");
1031