xref: /linux/drivers/net/macvlan.c (revision 7f8998c7aef3ac9c5f3f2943e083dfa6302e90d0)
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 <linux/workqueue.h>
34 #include <net/rtnetlink.h>
35 #include <net/xfrm.h>
36 #include <linux/netpoll.h>
37 
38 #define MACVLAN_HASH_BITS	8
39 #define MACVLAN_HASH_SIZE	(1<<MACVLAN_HASH_BITS)
40 #define MACVLAN_BC_QUEUE_LEN	1000
41 
42 struct macvlan_port {
43 	struct net_device	*dev;
44 	struct hlist_head	vlan_hash[MACVLAN_HASH_SIZE];
45 	struct list_head	vlans;
46 	struct rcu_head		rcu;
47 	struct sk_buff_head	bc_queue;
48 	struct work_struct	bc_work;
49 	bool 			passthru;
50 	int			count;
51 	struct hlist_head	vlan_source_hash[MACVLAN_HASH_SIZE];
52 };
53 
54 struct macvlan_source_entry {
55 	struct hlist_node	hlist;
56 	struct macvlan_dev	*vlan;
57 	unsigned char		addr[6+2] __aligned(sizeof(u16));
58 	struct rcu_head		rcu;
59 };
60 
61 struct macvlan_skb_cb {
62 	const struct macvlan_dev *src;
63 };
64 
65 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
66 
67 static void macvlan_port_destroy(struct net_device *dev);
68 
69 /* Hash Ethernet address */
70 static u32 macvlan_eth_hash(const unsigned char *addr)
71 {
72 	u64 value = get_unaligned((u64 *)addr);
73 
74 	/* only want 6 bytes */
75 #ifdef __BIG_ENDIAN
76 	value >>= 16;
77 #else
78 	value <<= 16;
79 #endif
80 	return hash_64(value, MACVLAN_HASH_BITS);
81 }
82 
83 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
84 {
85 	return rcu_dereference(dev->rx_handler_data);
86 }
87 
88 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
89 {
90 	return rtnl_dereference(dev->rx_handler_data);
91 }
92 
93 #define macvlan_port_exists(dev) (dev->priv_flags & IFF_MACVLAN_PORT)
94 
95 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
96 					       const unsigned char *addr)
97 {
98 	struct macvlan_dev *vlan;
99 	u32 idx = macvlan_eth_hash(addr);
100 
101 	hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) {
102 		if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
103 			return vlan;
104 	}
105 	return NULL;
106 }
107 
108 static struct macvlan_source_entry *macvlan_hash_lookup_source(
109 	const struct macvlan_dev *vlan,
110 	const unsigned char *addr)
111 {
112 	struct macvlan_source_entry *entry;
113 	u32 idx = macvlan_eth_hash(addr);
114 	struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
115 
116 	hlist_for_each_entry_rcu(entry, h, hlist) {
117 		if (ether_addr_equal_64bits(entry->addr, addr) &&
118 		    entry->vlan == vlan)
119 			return entry;
120 	}
121 	return NULL;
122 }
123 
124 static int macvlan_hash_add_source(struct macvlan_dev *vlan,
125 				   const unsigned char *addr)
126 {
127 	struct macvlan_port *port = vlan->port;
128 	struct macvlan_source_entry *entry;
129 	struct hlist_head *h;
130 
131 	entry = macvlan_hash_lookup_source(vlan, addr);
132 	if (entry)
133 		return 0;
134 
135 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
136 	if (!entry)
137 		return -ENOMEM;
138 
139 	ether_addr_copy(entry->addr, addr);
140 	entry->vlan = vlan;
141 	h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
142 	hlist_add_head_rcu(&entry->hlist, h);
143 	vlan->macaddr_count++;
144 
145 	return 0;
146 }
147 
148 static void macvlan_hash_add(struct macvlan_dev *vlan)
149 {
150 	struct macvlan_port *port = vlan->port;
151 	const unsigned char *addr = vlan->dev->dev_addr;
152 	u32 idx = macvlan_eth_hash(addr);
153 
154 	hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
155 }
156 
157 static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
158 {
159 	hlist_del_rcu(&entry->hlist);
160 	kfree_rcu(entry, rcu);
161 }
162 
163 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
164 {
165 	hlist_del_rcu(&vlan->hlist);
166 	if (sync)
167 		synchronize_rcu();
168 }
169 
170 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
171 					const unsigned char *addr)
172 {
173 	macvlan_hash_del(vlan, true);
174 	/* Now that we are unhashed it is safe to change the device
175 	 * address without confusing packet delivery.
176 	 */
177 	memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
178 	macvlan_hash_add(vlan);
179 }
180 
181 static int macvlan_addr_busy(const struct macvlan_port *port,
182 				const unsigned char *addr)
183 {
184 	/* Test to see if the specified multicast address is
185 	 * currently in use by the underlying device or
186 	 * another macvlan.
187 	 */
188 	if (ether_addr_equal_64bits(port->dev->dev_addr, addr))
189 		return 1;
190 
191 	if (macvlan_hash_lookup(port, addr))
192 		return 1;
193 
194 	return 0;
195 }
196 
197 
198 static int macvlan_broadcast_one(struct sk_buff *skb,
199 				 const struct macvlan_dev *vlan,
200 				 const struct ethhdr *eth, bool local)
201 {
202 	struct net_device *dev = vlan->dev;
203 
204 	if (local)
205 		return __dev_forward_skb(dev, skb);
206 
207 	skb->dev = dev;
208 	if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
209 		skb->pkt_type = PACKET_BROADCAST;
210 	else
211 		skb->pkt_type = PACKET_MULTICAST;
212 
213 	return 0;
214 }
215 
216 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
217 {
218 	return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
219 }
220 
221 
222 static unsigned int mc_hash(const struct macvlan_dev *vlan,
223 			    const unsigned char *addr)
224 {
225 	u32 val = __get_unaligned_cpu32(addr + 2);
226 
227 	val ^= macvlan_hash_mix(vlan);
228 	return hash_32(val, MACVLAN_MC_FILTER_BITS);
229 }
230 
231 static void macvlan_broadcast(struct sk_buff *skb,
232 			      const struct macvlan_port *port,
233 			      struct net_device *src,
234 			      enum macvlan_mode mode)
235 {
236 	const struct ethhdr *eth = eth_hdr(skb);
237 	const struct macvlan_dev *vlan;
238 	struct sk_buff *nskb;
239 	unsigned int i;
240 	int err;
241 	unsigned int hash;
242 
243 	if (skb->protocol == htons(ETH_P_PAUSE))
244 		return;
245 
246 	for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
247 		hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
248 			if (vlan->dev == src || !(vlan->mode & mode))
249 				continue;
250 
251 			hash = mc_hash(vlan, eth->h_dest);
252 			if (!test_bit(hash, vlan->mc_filter))
253 				continue;
254 
255 			err = NET_RX_DROP;
256 			nskb = skb_clone(skb, GFP_ATOMIC);
257 			if (likely(nskb))
258 				err = macvlan_broadcast_one(
259 					nskb, vlan, eth,
260 					mode == MACVLAN_MODE_BRIDGE) ?:
261 				      netif_rx_ni(nskb);
262 			macvlan_count_rx(vlan, skb->len + ETH_HLEN,
263 					 err == NET_RX_SUCCESS, 1);
264 		}
265 	}
266 }
267 
268 static void macvlan_process_broadcast(struct work_struct *w)
269 {
270 	struct macvlan_port *port = container_of(w, struct macvlan_port,
271 						 bc_work);
272 	struct sk_buff *skb;
273 	struct sk_buff_head list;
274 
275 	skb_queue_head_init(&list);
276 
277 	spin_lock_bh(&port->bc_queue.lock);
278 	skb_queue_splice_tail_init(&port->bc_queue, &list);
279 	spin_unlock_bh(&port->bc_queue.lock);
280 
281 	while ((skb = __skb_dequeue(&list))) {
282 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
283 
284 		rcu_read_lock();
285 
286 		if (!src)
287 			/* frame comes from an external address */
288 			macvlan_broadcast(skb, port, NULL,
289 					  MACVLAN_MODE_PRIVATE |
290 					  MACVLAN_MODE_VEPA    |
291 					  MACVLAN_MODE_PASSTHRU|
292 					  MACVLAN_MODE_BRIDGE);
293 		else if (src->mode == MACVLAN_MODE_VEPA)
294 			/* flood to everyone except source */
295 			macvlan_broadcast(skb, port, src->dev,
296 					  MACVLAN_MODE_VEPA |
297 					  MACVLAN_MODE_BRIDGE);
298 		else
299 			/*
300 			 * flood only to VEPA ports, bridge ports
301 			 * already saw the frame on the way out.
302 			 */
303 			macvlan_broadcast(skb, port, src->dev,
304 					  MACVLAN_MODE_VEPA);
305 
306 		rcu_read_unlock();
307 
308 		kfree_skb(skb);
309 	}
310 }
311 
312 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
313 				      struct sk_buff *skb)
314 {
315 	struct sk_buff *nskb;
316 	int err = -ENOMEM;
317 
318 	nskb = skb_clone(skb, GFP_ATOMIC);
319 	if (!nskb)
320 		goto err;
321 
322 	spin_lock(&port->bc_queue.lock);
323 	if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) {
324 		__skb_queue_tail(&port->bc_queue, nskb);
325 		err = 0;
326 	}
327 	spin_unlock(&port->bc_queue.lock);
328 
329 	if (err)
330 		goto free_nskb;
331 
332 	schedule_work(&port->bc_work);
333 	return;
334 
335 free_nskb:
336 	kfree_skb(nskb);
337 err:
338 	atomic_long_inc(&skb->dev->rx_dropped);
339 }
340 
341 static void macvlan_flush_sources(struct macvlan_port *port,
342 				  struct macvlan_dev *vlan)
343 {
344 	int i;
345 
346 	for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
347 		struct hlist_node *h, *n;
348 
349 		hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) {
350 			struct macvlan_source_entry *entry;
351 
352 			entry = hlist_entry(h, struct macvlan_source_entry,
353 					    hlist);
354 			if (entry->vlan == vlan)
355 				macvlan_hash_del_source(entry);
356 		}
357 	}
358 	vlan->macaddr_count = 0;
359 }
360 
361 static void macvlan_forward_source_one(struct sk_buff *skb,
362 				       struct macvlan_dev *vlan)
363 {
364 	struct sk_buff *nskb;
365 	struct net_device *dev;
366 	int len;
367 	int ret;
368 
369 	dev = vlan->dev;
370 	if (unlikely(!(dev->flags & IFF_UP)))
371 		return;
372 
373 	nskb = skb_clone(skb, GFP_ATOMIC);
374 	if (!nskb)
375 		return;
376 
377 	len = nskb->len + ETH_HLEN;
378 	nskb->dev = dev;
379 	nskb->pkt_type = PACKET_HOST;
380 
381 	ret = netif_rx(nskb);
382 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
383 }
384 
385 static void macvlan_forward_source(struct sk_buff *skb,
386 				   struct macvlan_port *port,
387 				   const unsigned char *addr)
388 {
389 	struct macvlan_source_entry *entry;
390 	u32 idx = macvlan_eth_hash(addr);
391 	struct hlist_head *h = &port->vlan_source_hash[idx];
392 
393 	hlist_for_each_entry_rcu(entry, h, hlist) {
394 		if (ether_addr_equal_64bits(entry->addr, addr))
395 			if (entry->vlan->dev->flags & IFF_UP)
396 				macvlan_forward_source_one(skb, entry->vlan);
397 	}
398 }
399 
400 /* called under rcu_read_lock() from netif_receive_skb */
401 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
402 {
403 	struct macvlan_port *port;
404 	struct sk_buff *skb = *pskb;
405 	const struct ethhdr *eth = eth_hdr(skb);
406 	const struct macvlan_dev *vlan;
407 	const struct macvlan_dev *src;
408 	struct net_device *dev;
409 	unsigned int len = 0;
410 	int ret = NET_RX_DROP;
411 
412 	port = macvlan_port_get_rcu(skb->dev);
413 	if (is_multicast_ether_addr(eth->h_dest)) {
414 		skb = ip_check_defrag(skb, IP_DEFRAG_MACVLAN);
415 		if (!skb)
416 			return RX_HANDLER_CONSUMED;
417 		eth = eth_hdr(skb);
418 		macvlan_forward_source(skb, port, eth->h_source);
419 		src = macvlan_hash_lookup(port, eth->h_source);
420 		if (src && src->mode != MACVLAN_MODE_VEPA &&
421 		    src->mode != MACVLAN_MODE_BRIDGE) {
422 			/* forward to original port. */
423 			vlan = src;
424 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
425 			      netif_rx(skb);
426 			goto out;
427 		}
428 
429 		MACVLAN_SKB_CB(skb)->src = src;
430 		macvlan_broadcast_enqueue(port, skb);
431 
432 		return RX_HANDLER_PASS;
433 	}
434 
435 	macvlan_forward_source(skb, port, eth->h_source);
436 	if (port->passthru)
437 		vlan = list_first_or_null_rcu(&port->vlans,
438 					      struct macvlan_dev, list);
439 	else
440 		vlan = macvlan_hash_lookup(port, eth->h_dest);
441 	if (vlan == NULL)
442 		return RX_HANDLER_PASS;
443 
444 	dev = vlan->dev;
445 	if (unlikely(!(dev->flags & IFF_UP))) {
446 		kfree_skb(skb);
447 		return RX_HANDLER_CONSUMED;
448 	}
449 	len = skb->len + ETH_HLEN;
450 	skb = skb_share_check(skb, GFP_ATOMIC);
451 	if (!skb)
452 		goto out;
453 
454 	skb->dev = dev;
455 	skb->pkt_type = PACKET_HOST;
456 
457 	ret = netif_rx(skb);
458 
459 out:
460 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, 0);
461 	return RX_HANDLER_CONSUMED;
462 }
463 
464 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
465 {
466 	const struct macvlan_dev *vlan = netdev_priv(dev);
467 	const struct macvlan_port *port = vlan->port;
468 	const struct macvlan_dev *dest;
469 
470 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
471 		const struct ethhdr *eth = (void *)skb->data;
472 
473 		/* send to other bridge ports directly */
474 		if (is_multicast_ether_addr(eth->h_dest)) {
475 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
476 			goto xmit_world;
477 		}
478 
479 		dest = macvlan_hash_lookup(port, eth->h_dest);
480 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
481 			/* send to lowerdev first for its network taps */
482 			dev_forward_skb(vlan->lowerdev, skb);
483 
484 			return NET_XMIT_SUCCESS;
485 		}
486 	}
487 
488 xmit_world:
489 	skb->dev = vlan->lowerdev;
490 	return dev_queue_xmit(skb);
491 }
492 
493 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
494 {
495 #ifdef CONFIG_NET_POLL_CONTROLLER
496 	if (vlan->netpoll)
497 		netpoll_send_skb(vlan->netpoll, skb);
498 #else
499 	BUG();
500 #endif
501 	return NETDEV_TX_OK;
502 }
503 
504 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
505 				      struct net_device *dev)
506 {
507 	unsigned int len = skb->len;
508 	int ret;
509 	struct macvlan_dev *vlan = netdev_priv(dev);
510 
511 	if (unlikely(netpoll_tx_running(dev)))
512 		return macvlan_netpoll_send_skb(vlan, skb);
513 
514 	if (vlan->fwd_priv) {
515 		skb->dev = vlan->lowerdev;
516 		ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
517 	} else {
518 		ret = macvlan_queue_xmit(skb, dev);
519 	}
520 
521 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
522 		struct vlan_pcpu_stats *pcpu_stats;
523 
524 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
525 		u64_stats_update_begin(&pcpu_stats->syncp);
526 		pcpu_stats->tx_packets++;
527 		pcpu_stats->tx_bytes += len;
528 		u64_stats_update_end(&pcpu_stats->syncp);
529 	} else {
530 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
531 	}
532 	return ret;
533 }
534 
535 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
536 			       unsigned short type, const void *daddr,
537 			       const void *saddr, unsigned len)
538 {
539 	const struct macvlan_dev *vlan = netdev_priv(dev);
540 	struct net_device *lowerdev = vlan->lowerdev;
541 
542 	return dev_hard_header(skb, lowerdev, type, daddr,
543 			       saddr ? : dev->dev_addr, len);
544 }
545 
546 static const struct header_ops macvlan_hard_header_ops = {
547 	.create  	= macvlan_hard_header,
548 	.rebuild	= eth_rebuild_header,
549 	.parse		= eth_header_parse,
550 	.cache		= eth_header_cache,
551 	.cache_update	= eth_header_cache_update,
552 };
553 
554 static struct rtnl_link_ops macvlan_link_ops;
555 
556 static int macvlan_open(struct net_device *dev)
557 {
558 	struct macvlan_dev *vlan = netdev_priv(dev);
559 	struct net_device *lowerdev = vlan->lowerdev;
560 	int err;
561 
562 	if (vlan->port->passthru) {
563 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
564 			err = dev_set_promiscuity(lowerdev, 1);
565 			if (err < 0)
566 				goto out;
567 		}
568 		goto hash_add;
569 	}
570 
571 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD &&
572 	    dev->rtnl_link_ops == &macvlan_link_ops) {
573 		vlan->fwd_priv =
574 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
575 
576 		/* If we get a NULL pointer back, or if we get an error
577 		 * then we should just fall through to the non accelerated path
578 		 */
579 		if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
580 			vlan->fwd_priv = NULL;
581 		} else
582 			return 0;
583 	}
584 
585 	err = -EBUSY;
586 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
587 		goto out;
588 
589 	err = dev_uc_add(lowerdev, dev->dev_addr);
590 	if (err < 0)
591 		goto out;
592 	if (dev->flags & IFF_ALLMULTI) {
593 		err = dev_set_allmulti(lowerdev, 1);
594 		if (err < 0)
595 			goto del_unicast;
596 	}
597 
598 hash_add:
599 	macvlan_hash_add(vlan);
600 	return 0;
601 
602 del_unicast:
603 	dev_uc_del(lowerdev, dev->dev_addr);
604 out:
605 	if (vlan->fwd_priv) {
606 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
607 							   vlan->fwd_priv);
608 		vlan->fwd_priv = NULL;
609 	}
610 	return err;
611 }
612 
613 static int macvlan_stop(struct net_device *dev)
614 {
615 	struct macvlan_dev *vlan = netdev_priv(dev);
616 	struct net_device *lowerdev = vlan->lowerdev;
617 
618 	if (vlan->fwd_priv) {
619 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
620 							   vlan->fwd_priv);
621 		vlan->fwd_priv = NULL;
622 		return 0;
623 	}
624 
625 	dev_uc_unsync(lowerdev, dev);
626 	dev_mc_unsync(lowerdev, dev);
627 
628 	if (vlan->port->passthru) {
629 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
630 			dev_set_promiscuity(lowerdev, -1);
631 		goto hash_del;
632 	}
633 
634 	if (dev->flags & IFF_ALLMULTI)
635 		dev_set_allmulti(lowerdev, -1);
636 
637 	dev_uc_del(lowerdev, dev->dev_addr);
638 
639 hash_del:
640 	macvlan_hash_del(vlan, !dev->dismantle);
641 	return 0;
642 }
643 
644 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
645 {
646 	struct macvlan_dev *vlan = netdev_priv(dev);
647 	struct net_device *lowerdev = vlan->lowerdev;
648 	int err;
649 
650 	if (!(dev->flags & IFF_UP)) {
651 		/* Just copy in the new address */
652 		ether_addr_copy(dev->dev_addr, addr);
653 	} else {
654 		/* Rehash and update the device filters */
655 		if (macvlan_addr_busy(vlan->port, addr))
656 			return -EBUSY;
657 
658 		if (!vlan->port->passthru) {
659 			err = dev_uc_add(lowerdev, addr);
660 			if (err)
661 				return err;
662 
663 			dev_uc_del(lowerdev, dev->dev_addr);
664 		}
665 
666 		macvlan_hash_change_addr(vlan, addr);
667 	}
668 	return 0;
669 }
670 
671 static int macvlan_set_mac_address(struct net_device *dev, void *p)
672 {
673 	struct macvlan_dev *vlan = netdev_priv(dev);
674 	struct sockaddr *addr = p;
675 
676 	if (!is_valid_ether_addr(addr->sa_data))
677 		return -EADDRNOTAVAIL;
678 
679 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
680 		dev_set_mac_address(vlan->lowerdev, addr);
681 		return 0;
682 	}
683 
684 	return macvlan_sync_address(dev, addr->sa_data);
685 }
686 
687 static void macvlan_change_rx_flags(struct net_device *dev, int change)
688 {
689 	struct macvlan_dev *vlan = netdev_priv(dev);
690 	struct net_device *lowerdev = vlan->lowerdev;
691 
692 	if (dev->flags & IFF_UP) {
693 		if (change & IFF_ALLMULTI)
694 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
695 	}
696 }
697 
698 static void macvlan_set_mac_lists(struct net_device *dev)
699 {
700 	struct macvlan_dev *vlan = netdev_priv(dev);
701 
702 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
703 		bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
704 	} else {
705 		struct netdev_hw_addr *ha;
706 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
707 
708 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
709 		netdev_for_each_mc_addr(ha, dev) {
710 			__set_bit(mc_hash(vlan, ha->addr), filter);
711 		}
712 
713 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
714 
715 		bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
716 	}
717 	dev_uc_sync(vlan->lowerdev, dev);
718 	dev_mc_sync(vlan->lowerdev, dev);
719 }
720 
721 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
722 {
723 	struct macvlan_dev *vlan = netdev_priv(dev);
724 
725 	if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
726 		return -EINVAL;
727 	dev->mtu = new_mtu;
728 	return 0;
729 }
730 
731 /*
732  * macvlan network devices have devices nesting below it and are a special
733  * "super class" of normal network devices; split their locks off into a
734  * separate class since they always nest.
735  */
736 static struct lock_class_key macvlan_netdev_xmit_lock_key;
737 static struct lock_class_key macvlan_netdev_addr_lock_key;
738 
739 #define ALWAYS_ON_FEATURES \
740 	(NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX)
741 
742 #define MACVLAN_FEATURES \
743 	(NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
744 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_GSO_ROBUST | \
745 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
746 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
747 
748 #define MACVLAN_STATE_MASK \
749 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
750 
751 static int macvlan_get_nest_level(struct net_device *dev)
752 {
753 	return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
754 }
755 
756 static void macvlan_set_lockdep_class_one(struct net_device *dev,
757 					  struct netdev_queue *txq,
758 					  void *_unused)
759 {
760 	lockdep_set_class(&txq->_xmit_lock,
761 			  &macvlan_netdev_xmit_lock_key);
762 }
763 
764 static void macvlan_set_lockdep_class(struct net_device *dev)
765 {
766 	lockdep_set_class_and_subclass(&dev->addr_list_lock,
767 				       &macvlan_netdev_addr_lock_key,
768 				       macvlan_get_nest_level(dev));
769 	netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
770 }
771 
772 static int macvlan_init(struct net_device *dev)
773 {
774 	struct macvlan_dev *vlan = netdev_priv(dev);
775 	const struct net_device *lowerdev = vlan->lowerdev;
776 
777 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
778 				  (lowerdev->state & MACVLAN_STATE_MASK);
779 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
780 	dev->features		|= ALWAYS_ON_FEATURES;
781 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
782 	dev->gso_max_size	= lowerdev->gso_max_size;
783 	dev->iflink		= lowerdev->ifindex;
784 	dev->hard_header_len	= lowerdev->hard_header_len;
785 
786 	macvlan_set_lockdep_class(dev);
787 
788 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
789 	if (!vlan->pcpu_stats)
790 		return -ENOMEM;
791 
792 	return 0;
793 }
794 
795 static void macvlan_uninit(struct net_device *dev)
796 {
797 	struct macvlan_dev *vlan = netdev_priv(dev);
798 	struct macvlan_port *port = vlan->port;
799 
800 	free_percpu(vlan->pcpu_stats);
801 
802 	macvlan_flush_sources(port, vlan);
803 	port->count -= 1;
804 	if (!port->count)
805 		macvlan_port_destroy(port->dev);
806 }
807 
808 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
809 							 struct rtnl_link_stats64 *stats)
810 {
811 	struct macvlan_dev *vlan = netdev_priv(dev);
812 
813 	if (vlan->pcpu_stats) {
814 		struct vlan_pcpu_stats *p;
815 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
816 		u32 rx_errors = 0, tx_dropped = 0;
817 		unsigned int start;
818 		int i;
819 
820 		for_each_possible_cpu(i) {
821 			p = per_cpu_ptr(vlan->pcpu_stats, i);
822 			do {
823 				start = u64_stats_fetch_begin_irq(&p->syncp);
824 				rx_packets	= p->rx_packets;
825 				rx_bytes	= p->rx_bytes;
826 				rx_multicast	= p->rx_multicast;
827 				tx_packets	= p->tx_packets;
828 				tx_bytes	= p->tx_bytes;
829 			} while (u64_stats_fetch_retry_irq(&p->syncp, start));
830 
831 			stats->rx_packets	+= rx_packets;
832 			stats->rx_bytes		+= rx_bytes;
833 			stats->multicast	+= rx_multicast;
834 			stats->tx_packets	+= tx_packets;
835 			stats->tx_bytes		+= tx_bytes;
836 			/* rx_errors & tx_dropped are u32, updated
837 			 * without syncp protection.
838 			 */
839 			rx_errors	+= p->rx_errors;
840 			tx_dropped	+= p->tx_dropped;
841 		}
842 		stats->rx_errors	= rx_errors;
843 		stats->rx_dropped	= rx_errors;
844 		stats->tx_dropped	= tx_dropped;
845 	}
846 	return stats;
847 }
848 
849 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
850 				   __be16 proto, u16 vid)
851 {
852 	struct macvlan_dev *vlan = netdev_priv(dev);
853 	struct net_device *lowerdev = vlan->lowerdev;
854 
855 	return vlan_vid_add(lowerdev, proto, vid);
856 }
857 
858 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
859 				    __be16 proto, u16 vid)
860 {
861 	struct macvlan_dev *vlan = netdev_priv(dev);
862 	struct net_device *lowerdev = vlan->lowerdev;
863 
864 	vlan_vid_del(lowerdev, proto, vid);
865 	return 0;
866 }
867 
868 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
869 			   struct net_device *dev,
870 			   const unsigned char *addr,
871 			   u16 flags)
872 {
873 	struct macvlan_dev *vlan = netdev_priv(dev);
874 	int err = -EINVAL;
875 
876 	/* Support unicast filter only on passthru devices.
877 	 * Multicast filter should be allowed on all devices.
878 	 */
879 	if (!vlan->port->passthru && is_unicast_ether_addr(addr))
880 		return -EOPNOTSUPP;
881 
882 	if (flags & NLM_F_REPLACE)
883 		return -EOPNOTSUPP;
884 
885 	if (is_unicast_ether_addr(addr))
886 		err = dev_uc_add_excl(dev, addr);
887 	else if (is_multicast_ether_addr(addr))
888 		err = dev_mc_add_excl(dev, addr);
889 
890 	return err;
891 }
892 
893 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
894 			   struct net_device *dev,
895 			   const unsigned char *addr)
896 {
897 	struct macvlan_dev *vlan = netdev_priv(dev);
898 	int err = -EINVAL;
899 
900 	/* Support unicast filter only on passthru devices.
901 	 * Multicast filter should be allowed on all devices.
902 	 */
903 	if (!vlan->port->passthru && is_unicast_ether_addr(addr))
904 		return -EOPNOTSUPP;
905 
906 	if (is_unicast_ether_addr(addr))
907 		err = dev_uc_del(dev, addr);
908 	else if (is_multicast_ether_addr(addr))
909 		err = dev_mc_del(dev, addr);
910 
911 	return err;
912 }
913 
914 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
915 					struct ethtool_drvinfo *drvinfo)
916 {
917 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
918 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
919 }
920 
921 static int macvlan_ethtool_get_settings(struct net_device *dev,
922 					struct ethtool_cmd *cmd)
923 {
924 	const struct macvlan_dev *vlan = netdev_priv(dev);
925 
926 	return __ethtool_get_settings(vlan->lowerdev, cmd);
927 }
928 
929 static netdev_features_t macvlan_fix_features(struct net_device *dev,
930 					      netdev_features_t features)
931 {
932 	struct macvlan_dev *vlan = netdev_priv(dev);
933 	netdev_features_t mask;
934 
935 	features |= NETIF_F_ALL_FOR_ALL;
936 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
937 	mask = features;
938 
939 	features = netdev_increment_features(vlan->lowerdev->features,
940 					     features,
941 					     mask);
942 	features |= ALWAYS_ON_FEATURES;
943 	features &= ~NETIF_F_NETNS_LOCAL;
944 
945 	return features;
946 }
947 
948 #ifdef CONFIG_NET_POLL_CONTROLLER
949 static void macvlan_dev_poll_controller(struct net_device *dev)
950 {
951 	return;
952 }
953 
954 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
955 {
956 	struct macvlan_dev *vlan = netdev_priv(dev);
957 	struct net_device *real_dev = vlan->lowerdev;
958 	struct netpoll *netpoll;
959 	int err = 0;
960 
961 	netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
962 	err = -ENOMEM;
963 	if (!netpoll)
964 		goto out;
965 
966 	err = __netpoll_setup(netpoll, real_dev);
967 	if (err) {
968 		kfree(netpoll);
969 		goto out;
970 	}
971 
972 	vlan->netpoll = netpoll;
973 
974 out:
975 	return err;
976 }
977 
978 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
979 {
980 	struct macvlan_dev *vlan = netdev_priv(dev);
981 	struct netpoll *netpoll = vlan->netpoll;
982 
983 	if (!netpoll)
984 		return;
985 
986 	vlan->netpoll = NULL;
987 
988 	__netpoll_free_async(netpoll);
989 }
990 #endif	/* CONFIG_NET_POLL_CONTROLLER */
991 
992 static const struct ethtool_ops macvlan_ethtool_ops = {
993 	.get_link		= ethtool_op_get_link,
994 	.get_settings		= macvlan_ethtool_get_settings,
995 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
996 };
997 
998 static const struct net_device_ops macvlan_netdev_ops = {
999 	.ndo_init		= macvlan_init,
1000 	.ndo_uninit		= macvlan_uninit,
1001 	.ndo_open		= macvlan_open,
1002 	.ndo_stop		= macvlan_stop,
1003 	.ndo_start_xmit		= macvlan_start_xmit,
1004 	.ndo_change_mtu		= macvlan_change_mtu,
1005 	.ndo_fix_features	= macvlan_fix_features,
1006 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1007 	.ndo_set_mac_address	= macvlan_set_mac_address,
1008 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1009 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1010 	.ndo_validate_addr	= eth_validate_addr,
1011 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1012 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1013 	.ndo_fdb_add		= macvlan_fdb_add,
1014 	.ndo_fdb_del		= macvlan_fdb_del,
1015 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1016 	.ndo_get_lock_subclass  = macvlan_get_nest_level,
1017 #ifdef CONFIG_NET_POLL_CONTROLLER
1018 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1019 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1020 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1021 #endif
1022 };
1023 
1024 void macvlan_common_setup(struct net_device *dev)
1025 {
1026 	ether_setup(dev);
1027 
1028 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1029 	netif_keep_dst(dev);
1030 	dev->priv_flags	       |= IFF_UNICAST_FLT;
1031 	dev->netdev_ops		= &macvlan_netdev_ops;
1032 	dev->destructor		= free_netdev;
1033 	dev->header_ops		= &macvlan_hard_header_ops;
1034 	dev->ethtool_ops	= &macvlan_ethtool_ops;
1035 }
1036 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1037 
1038 static void macvlan_setup(struct net_device *dev)
1039 {
1040 	macvlan_common_setup(dev);
1041 	dev->tx_queue_len	= 0;
1042 }
1043 
1044 static int macvlan_port_create(struct net_device *dev)
1045 {
1046 	struct macvlan_port *port;
1047 	unsigned int i;
1048 	int err;
1049 
1050 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1051 		return -EINVAL;
1052 
1053 	port = kzalloc(sizeof(*port), GFP_KERNEL);
1054 	if (port == NULL)
1055 		return -ENOMEM;
1056 
1057 	port->passthru = false;
1058 	port->dev = dev;
1059 	INIT_LIST_HEAD(&port->vlans);
1060 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1061 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
1062 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1063 		INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1064 
1065 	skb_queue_head_init(&port->bc_queue);
1066 	INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1067 
1068 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1069 	if (err)
1070 		kfree(port);
1071 	else
1072 		dev->priv_flags |= IFF_MACVLAN_PORT;
1073 	return err;
1074 }
1075 
1076 static void macvlan_port_destroy(struct net_device *dev)
1077 {
1078 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1079 
1080 	cancel_work_sync(&port->bc_work);
1081 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
1082 	netdev_rx_handler_unregister(dev);
1083 	kfree_rcu(port, rcu);
1084 }
1085 
1086 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
1087 {
1088 	if (tb[IFLA_ADDRESS]) {
1089 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1090 			return -EINVAL;
1091 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1092 			return -EADDRNOTAVAIL;
1093 	}
1094 
1095 	if (data && data[IFLA_MACVLAN_FLAGS] &&
1096 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1097 		return -EINVAL;
1098 
1099 	if (data && data[IFLA_MACVLAN_MODE]) {
1100 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1101 		case MACVLAN_MODE_PRIVATE:
1102 		case MACVLAN_MODE_VEPA:
1103 		case MACVLAN_MODE_BRIDGE:
1104 		case MACVLAN_MODE_PASSTHRU:
1105 		case MACVLAN_MODE_SOURCE:
1106 			break;
1107 		default:
1108 			return -EINVAL;
1109 		}
1110 	}
1111 
1112 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1113 		switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1114 		case MACVLAN_MACADDR_ADD:
1115 		case MACVLAN_MACADDR_DEL:
1116 		case MACVLAN_MACADDR_FLUSH:
1117 		case MACVLAN_MACADDR_SET:
1118 			break;
1119 		default:
1120 			return -EINVAL;
1121 		}
1122 	}
1123 
1124 	if (data && data[IFLA_MACVLAN_MACADDR]) {
1125 		if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1126 			return -EINVAL;
1127 
1128 		if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1129 			return -EADDRNOTAVAIL;
1130 	}
1131 
1132 	if (data && data[IFLA_MACVLAN_MACADDR_COUNT])
1133 		return -EINVAL;
1134 
1135 	return 0;
1136 }
1137 
1138 /**
1139  * reconfigure list of remote source mac address
1140  * (only for macvlan devices in source mode)
1141  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1142  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1143  */
1144 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1145 				      struct nlattr *data[])
1146 {
1147 	char *addr = NULL;
1148 	int ret, rem, len;
1149 	struct nlattr *nla, *head;
1150 	struct macvlan_source_entry *entry;
1151 
1152 	if (data[IFLA_MACVLAN_MACADDR])
1153 		addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1154 
1155 	if (mode == MACVLAN_MACADDR_ADD) {
1156 		if (!addr)
1157 			return -EINVAL;
1158 
1159 		return macvlan_hash_add_source(vlan, addr);
1160 
1161 	} else if (mode == MACVLAN_MACADDR_DEL) {
1162 		if (!addr)
1163 			return -EINVAL;
1164 
1165 		entry = macvlan_hash_lookup_source(vlan, addr);
1166 		if (entry) {
1167 			macvlan_hash_del_source(entry);
1168 			vlan->macaddr_count--;
1169 		}
1170 	} else if (mode == MACVLAN_MACADDR_FLUSH) {
1171 		macvlan_flush_sources(vlan->port, vlan);
1172 	} else if (mode == MACVLAN_MACADDR_SET) {
1173 		macvlan_flush_sources(vlan->port, vlan);
1174 
1175 		if (addr) {
1176 			ret = macvlan_hash_add_source(vlan, addr);
1177 			if (ret)
1178 				return ret;
1179 		}
1180 
1181 		if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1182 			return 0;
1183 
1184 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1185 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1186 
1187 		nla_for_each_attr(nla, head, len, rem) {
1188 			if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1189 			    nla_len(nla) != ETH_ALEN)
1190 				continue;
1191 
1192 			addr = nla_data(nla);
1193 			ret = macvlan_hash_add_source(vlan, addr);
1194 			if (ret)
1195 				return ret;
1196 		}
1197 	} else {
1198 		return -EINVAL;
1199 	}
1200 
1201 	return 0;
1202 }
1203 
1204 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1205 			   struct nlattr *tb[], struct nlattr *data[])
1206 {
1207 	struct macvlan_dev *vlan = netdev_priv(dev);
1208 	struct macvlan_port *port;
1209 	struct net_device *lowerdev;
1210 	int err;
1211 	int macmode;
1212 
1213 	if (!tb[IFLA_LINK])
1214 		return -EINVAL;
1215 
1216 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1217 	if (lowerdev == NULL)
1218 		return -ENODEV;
1219 
1220 	/* When creating macvlans or macvtaps on top of other macvlans - use
1221 	 * the real device as the lowerdev.
1222 	 */
1223 	if (netif_is_macvlan(lowerdev))
1224 		lowerdev = macvlan_dev_real_dev(lowerdev);
1225 
1226 	if (!tb[IFLA_MTU])
1227 		dev->mtu = lowerdev->mtu;
1228 	else if (dev->mtu > lowerdev->mtu)
1229 		return -EINVAL;
1230 
1231 	if (!tb[IFLA_ADDRESS])
1232 		eth_hw_addr_random(dev);
1233 
1234 	if (!macvlan_port_exists(lowerdev)) {
1235 		err = macvlan_port_create(lowerdev);
1236 		if (err < 0)
1237 			return err;
1238 	}
1239 	port = macvlan_port_get_rtnl(lowerdev);
1240 
1241 	/* Only 1 macvlan device can be created in passthru mode */
1242 	if (port->passthru)
1243 		return -EINVAL;
1244 
1245 	vlan->lowerdev = lowerdev;
1246 	vlan->dev      = dev;
1247 	vlan->port     = port;
1248 	vlan->set_features = MACVLAN_FEATURES;
1249 	vlan->nest_level = dev_get_nest_level(lowerdev, netif_is_macvlan) + 1;
1250 
1251 	vlan->mode     = MACVLAN_MODE_VEPA;
1252 	if (data && data[IFLA_MACVLAN_MODE])
1253 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1254 
1255 	if (data && data[IFLA_MACVLAN_FLAGS])
1256 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1257 
1258 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1259 		if (port->count)
1260 			return -EINVAL;
1261 		port->passthru = true;
1262 		eth_hw_addr_inherit(dev, lowerdev);
1263 	}
1264 
1265 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1266 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1267 			return -EINVAL;
1268 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1269 		err = macvlan_changelink_sources(vlan, macmode, data);
1270 		if (err)
1271 			return err;
1272 	}
1273 
1274 	port->count += 1;
1275 	err = register_netdevice(dev);
1276 	if (err < 0)
1277 		goto destroy_port;
1278 
1279 	dev->priv_flags |= IFF_MACVLAN;
1280 	err = netdev_upper_dev_link(lowerdev, dev);
1281 	if (err)
1282 		goto unregister_netdev;
1283 
1284 	list_add_tail_rcu(&vlan->list, &port->vlans);
1285 	netif_stacked_transfer_operstate(lowerdev, dev);
1286 
1287 	return 0;
1288 
1289 unregister_netdev:
1290 	unregister_netdevice(dev);
1291 destroy_port:
1292 	port->count -= 1;
1293 	if (!port->count)
1294 		macvlan_port_destroy(lowerdev);
1295 
1296 	return err;
1297 }
1298 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1299 
1300 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1301 			   struct nlattr *tb[], struct nlattr *data[])
1302 {
1303 	return macvlan_common_newlink(src_net, dev, tb, data);
1304 }
1305 
1306 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1307 {
1308 	struct macvlan_dev *vlan = netdev_priv(dev);
1309 
1310 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1311 		macvlan_flush_sources(vlan->port, vlan);
1312 	list_del_rcu(&vlan->list);
1313 	unregister_netdevice_queue(dev, head);
1314 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1315 }
1316 EXPORT_SYMBOL_GPL(macvlan_dellink);
1317 
1318 static int macvlan_changelink(struct net_device *dev,
1319 		struct nlattr *tb[], struct nlattr *data[])
1320 {
1321 	struct macvlan_dev *vlan = netdev_priv(dev);
1322 	enum macvlan_mode mode;
1323 	bool set_mode = false;
1324 	enum macvlan_macaddr_mode macmode;
1325 	int ret;
1326 
1327 	/* Validate mode, but don't set yet: setting flags may fail. */
1328 	if (data && data[IFLA_MACVLAN_MODE]) {
1329 		set_mode = true;
1330 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1331 		/* Passthrough mode can't be set or cleared dynamically */
1332 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1333 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1334 			return -EINVAL;
1335 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1336 		    vlan->mode != mode)
1337 			macvlan_flush_sources(vlan->port, vlan);
1338 	}
1339 
1340 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1341 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1342 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1343 		if (vlan->port->passthru && promisc) {
1344 			int err;
1345 
1346 			if (flags & MACVLAN_FLAG_NOPROMISC)
1347 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1348 			else
1349 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1350 			if (err < 0)
1351 				return err;
1352 		}
1353 		vlan->flags = flags;
1354 	}
1355 	if (set_mode)
1356 		vlan->mode = mode;
1357 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1358 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1359 			return -EINVAL;
1360 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1361 		ret = macvlan_changelink_sources(vlan, macmode, data);
1362 		if (ret)
1363 			return ret;
1364 	}
1365 	return 0;
1366 }
1367 
1368 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1369 {
1370 	if (vlan->macaddr_count == 0)
1371 		return 0;
1372 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1373 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1374 }
1375 
1376 static size_t macvlan_get_size(const struct net_device *dev)
1377 {
1378 	struct macvlan_dev *vlan = netdev_priv(dev);
1379 
1380 	return (0
1381 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1382 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1383 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1384 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1385 		);
1386 }
1387 
1388 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1389 				     const struct macvlan_dev *vlan,
1390 				     const int i)
1391 {
1392 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1393 	struct macvlan_source_entry *entry;
1394 
1395 	hlist_for_each_entry_rcu(entry, h, hlist) {
1396 		if (entry->vlan != vlan)
1397 			continue;
1398 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1399 			return 1;
1400 	}
1401 	return 0;
1402 }
1403 
1404 static int macvlan_fill_info(struct sk_buff *skb,
1405 				const struct net_device *dev)
1406 {
1407 	struct macvlan_dev *vlan = netdev_priv(dev);
1408 	int i;
1409 	struct nlattr *nest;
1410 
1411 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1412 		goto nla_put_failure;
1413 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1414 		goto nla_put_failure;
1415 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1416 		goto nla_put_failure;
1417 	if (vlan->macaddr_count > 0) {
1418 		nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA);
1419 		if (nest == NULL)
1420 			goto nla_put_failure;
1421 
1422 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1423 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1424 				goto nla_put_failure;
1425 		}
1426 		nla_nest_end(skb, nest);
1427 	}
1428 	return 0;
1429 
1430 nla_put_failure:
1431 	return -EMSGSIZE;
1432 }
1433 
1434 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1435 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1436 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1437 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1438 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1439 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1440 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1441 };
1442 
1443 int macvlan_link_register(struct rtnl_link_ops *ops)
1444 {
1445 	/* common fields */
1446 	ops->priv_size		= sizeof(struct macvlan_dev);
1447 	ops->validate		= macvlan_validate;
1448 	ops->maxtype		= IFLA_MACVLAN_MAX;
1449 	ops->policy		= macvlan_policy;
1450 	ops->changelink		= macvlan_changelink;
1451 	ops->get_size		= macvlan_get_size;
1452 	ops->fill_info		= macvlan_fill_info;
1453 
1454 	return rtnl_link_register(ops);
1455 };
1456 EXPORT_SYMBOL_GPL(macvlan_link_register);
1457 
1458 static struct rtnl_link_ops macvlan_link_ops = {
1459 	.kind		= "macvlan",
1460 	.setup		= macvlan_setup,
1461 	.newlink	= macvlan_newlink,
1462 	.dellink	= macvlan_dellink,
1463 };
1464 
1465 static int macvlan_device_event(struct notifier_block *unused,
1466 				unsigned long event, void *ptr)
1467 {
1468 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1469 	struct macvlan_dev *vlan, *next;
1470 	struct macvlan_port *port;
1471 	LIST_HEAD(list_kill);
1472 
1473 	if (!macvlan_port_exists(dev))
1474 		return NOTIFY_DONE;
1475 
1476 	port = macvlan_port_get_rtnl(dev);
1477 
1478 	switch (event) {
1479 	case NETDEV_CHANGE:
1480 		list_for_each_entry(vlan, &port->vlans, list)
1481 			netif_stacked_transfer_operstate(vlan->lowerdev,
1482 							 vlan->dev);
1483 		break;
1484 	case NETDEV_FEAT_CHANGE:
1485 		list_for_each_entry(vlan, &port->vlans, list) {
1486 			vlan->dev->gso_max_size = dev->gso_max_size;
1487 			netdev_update_features(vlan->dev);
1488 		}
1489 		break;
1490 	case NETDEV_CHANGEMTU:
1491 		list_for_each_entry(vlan, &port->vlans, list) {
1492 			if (vlan->dev->mtu <= dev->mtu)
1493 				continue;
1494 			dev_set_mtu(vlan->dev, dev->mtu);
1495 		}
1496 		break;
1497 	case NETDEV_CHANGEADDR:
1498 		if (!port->passthru)
1499 			return NOTIFY_DONE;
1500 
1501 		vlan = list_first_entry_or_null(&port->vlans,
1502 						struct macvlan_dev,
1503 						list);
1504 
1505 		if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1506 			return NOTIFY_BAD;
1507 
1508 		break;
1509 	case NETDEV_UNREGISTER:
1510 		/* twiddle thumbs on netns device moves */
1511 		if (dev->reg_state != NETREG_UNREGISTERING)
1512 			break;
1513 
1514 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1515 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1516 		unregister_netdevice_many(&list_kill);
1517 		break;
1518 	case NETDEV_PRE_TYPE_CHANGE:
1519 		/* Forbid underlaying device to change its type. */
1520 		return NOTIFY_BAD;
1521 
1522 	case NETDEV_NOTIFY_PEERS:
1523 	case NETDEV_BONDING_FAILOVER:
1524 	case NETDEV_RESEND_IGMP:
1525 		/* Propagate to all vlans */
1526 		list_for_each_entry(vlan, &port->vlans, list)
1527 			call_netdevice_notifiers(event, vlan->dev);
1528 	}
1529 	return NOTIFY_DONE;
1530 }
1531 
1532 static struct notifier_block macvlan_notifier_block __read_mostly = {
1533 	.notifier_call	= macvlan_device_event,
1534 };
1535 
1536 static int __init macvlan_init_module(void)
1537 {
1538 	int err;
1539 
1540 	register_netdevice_notifier(&macvlan_notifier_block);
1541 
1542 	err = macvlan_link_register(&macvlan_link_ops);
1543 	if (err < 0)
1544 		goto err1;
1545 	return 0;
1546 err1:
1547 	unregister_netdevice_notifier(&macvlan_notifier_block);
1548 	return err;
1549 }
1550 
1551 static void __exit macvlan_cleanup_module(void)
1552 {
1553 	rtnl_link_unregister(&macvlan_link_ops);
1554 	unregister_netdevice_notifier(&macvlan_notifier_block);
1555 }
1556 
1557 module_init(macvlan_init_module);
1558 module_exit(macvlan_cleanup_module);
1559 
1560 MODULE_LICENSE("GPL");
1561 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1562 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1563 MODULE_ALIAS_RTNL_LINK("macvlan");
1564