xref: /openbmc/linux/drivers/net/macvlan.c (revision 8c749ce9)
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, true);
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, false);
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;
411 	rx_handler_result_t handle_res;
412 
413 	port = macvlan_port_get_rcu(skb->dev);
414 	if (is_multicast_ether_addr(eth->h_dest)) {
415 		skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
416 		if (!skb)
417 			return RX_HANDLER_CONSUMED;
418 		*pskb = skb;
419 		eth = eth_hdr(skb);
420 		macvlan_forward_source(skb, port, eth->h_source);
421 		src = macvlan_hash_lookup(port, eth->h_source);
422 		if (src && src->mode != MACVLAN_MODE_VEPA &&
423 		    src->mode != MACVLAN_MODE_BRIDGE) {
424 			/* forward to original port. */
425 			vlan = src;
426 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
427 			      netif_rx(skb);
428 			handle_res = RX_HANDLER_CONSUMED;
429 			goto out;
430 		}
431 
432 		MACVLAN_SKB_CB(skb)->src = src;
433 		macvlan_broadcast_enqueue(port, skb);
434 
435 		return RX_HANDLER_PASS;
436 	}
437 
438 	macvlan_forward_source(skb, port, eth->h_source);
439 	if (port->passthru)
440 		vlan = list_first_or_null_rcu(&port->vlans,
441 					      struct macvlan_dev, list);
442 	else
443 		vlan = macvlan_hash_lookup(port, eth->h_dest);
444 	if (vlan == NULL)
445 		return RX_HANDLER_PASS;
446 
447 	dev = vlan->dev;
448 	if (unlikely(!(dev->flags & IFF_UP))) {
449 		kfree_skb(skb);
450 		return RX_HANDLER_CONSUMED;
451 	}
452 	len = skb->len + ETH_HLEN;
453 	skb = skb_share_check(skb, GFP_ATOMIC);
454 	if (!skb) {
455 		ret = NET_RX_DROP;
456 		handle_res = RX_HANDLER_CONSUMED;
457 		goto out;
458 	}
459 
460 	*pskb = skb;
461 	skb->dev = dev;
462 	skb->pkt_type = PACKET_HOST;
463 
464 	ret = NET_RX_SUCCESS;
465 	handle_res = RX_HANDLER_ANOTHER;
466 out:
467 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
468 	return handle_res;
469 }
470 
471 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
472 {
473 	const struct macvlan_dev *vlan = netdev_priv(dev);
474 	const struct macvlan_port *port = vlan->port;
475 	const struct macvlan_dev *dest;
476 
477 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
478 		const struct ethhdr *eth = (void *)skb->data;
479 
480 		/* send to other bridge ports directly */
481 		if (is_multicast_ether_addr(eth->h_dest)) {
482 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
483 			goto xmit_world;
484 		}
485 
486 		dest = macvlan_hash_lookup(port, eth->h_dest);
487 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
488 			/* send to lowerdev first for its network taps */
489 			dev_forward_skb(vlan->lowerdev, skb);
490 
491 			return NET_XMIT_SUCCESS;
492 		}
493 	}
494 
495 xmit_world:
496 	skb->dev = vlan->lowerdev;
497 	return dev_queue_xmit(skb);
498 }
499 
500 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
501 {
502 #ifdef CONFIG_NET_POLL_CONTROLLER
503 	if (vlan->netpoll)
504 		netpoll_send_skb(vlan->netpoll, skb);
505 #else
506 	BUG();
507 #endif
508 	return NETDEV_TX_OK;
509 }
510 
511 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
512 				      struct net_device *dev)
513 {
514 	unsigned int len = skb->len;
515 	int ret;
516 	struct macvlan_dev *vlan = netdev_priv(dev);
517 
518 	if (unlikely(netpoll_tx_running(dev)))
519 		return macvlan_netpoll_send_skb(vlan, skb);
520 
521 	if (vlan->fwd_priv) {
522 		skb->dev = vlan->lowerdev;
523 		ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
524 	} else {
525 		ret = macvlan_queue_xmit(skb, dev);
526 	}
527 
528 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
529 		struct vlan_pcpu_stats *pcpu_stats;
530 
531 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
532 		u64_stats_update_begin(&pcpu_stats->syncp);
533 		pcpu_stats->tx_packets++;
534 		pcpu_stats->tx_bytes += len;
535 		u64_stats_update_end(&pcpu_stats->syncp);
536 	} else {
537 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
538 	}
539 	return ret;
540 }
541 
542 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
543 			       unsigned short type, const void *daddr,
544 			       const void *saddr, unsigned len)
545 {
546 	const struct macvlan_dev *vlan = netdev_priv(dev);
547 	struct net_device *lowerdev = vlan->lowerdev;
548 
549 	return dev_hard_header(skb, lowerdev, type, daddr,
550 			       saddr ? : dev->dev_addr, len);
551 }
552 
553 static const struct header_ops macvlan_hard_header_ops = {
554 	.create  	= macvlan_hard_header,
555 	.parse		= eth_header_parse,
556 	.cache		= eth_header_cache,
557 	.cache_update	= eth_header_cache_update,
558 };
559 
560 static struct rtnl_link_ops macvlan_link_ops;
561 
562 static int macvlan_open(struct net_device *dev)
563 {
564 	struct macvlan_dev *vlan = netdev_priv(dev);
565 	struct net_device *lowerdev = vlan->lowerdev;
566 	int err;
567 
568 	if (vlan->port->passthru) {
569 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
570 			err = dev_set_promiscuity(lowerdev, 1);
571 			if (err < 0)
572 				goto out;
573 		}
574 		goto hash_add;
575 	}
576 
577 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD &&
578 	    dev->rtnl_link_ops == &macvlan_link_ops) {
579 		vlan->fwd_priv =
580 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
581 
582 		/* If we get a NULL pointer back, or if we get an error
583 		 * then we should just fall through to the non accelerated path
584 		 */
585 		if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
586 			vlan->fwd_priv = NULL;
587 		} else
588 			return 0;
589 	}
590 
591 	err = -EBUSY;
592 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
593 		goto out;
594 
595 	err = dev_uc_add(lowerdev, dev->dev_addr);
596 	if (err < 0)
597 		goto out;
598 	if (dev->flags & IFF_ALLMULTI) {
599 		err = dev_set_allmulti(lowerdev, 1);
600 		if (err < 0)
601 			goto del_unicast;
602 	}
603 
604 	if (dev->flags & IFF_PROMISC) {
605 		err = dev_set_promiscuity(lowerdev, 1);
606 		if (err < 0)
607 			goto clear_multi;
608 	}
609 
610 hash_add:
611 	macvlan_hash_add(vlan);
612 	return 0;
613 
614 clear_multi:
615 	dev_set_allmulti(lowerdev, -1);
616 del_unicast:
617 	dev_uc_del(lowerdev, dev->dev_addr);
618 out:
619 	if (vlan->fwd_priv) {
620 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
621 							   vlan->fwd_priv);
622 		vlan->fwd_priv = NULL;
623 	}
624 	return err;
625 }
626 
627 static int macvlan_stop(struct net_device *dev)
628 {
629 	struct macvlan_dev *vlan = netdev_priv(dev);
630 	struct net_device *lowerdev = vlan->lowerdev;
631 
632 	if (vlan->fwd_priv) {
633 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
634 							   vlan->fwd_priv);
635 		vlan->fwd_priv = NULL;
636 		return 0;
637 	}
638 
639 	dev_uc_unsync(lowerdev, dev);
640 	dev_mc_unsync(lowerdev, dev);
641 
642 	if (vlan->port->passthru) {
643 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
644 			dev_set_promiscuity(lowerdev, -1);
645 		goto hash_del;
646 	}
647 
648 	if (dev->flags & IFF_ALLMULTI)
649 		dev_set_allmulti(lowerdev, -1);
650 
651 	if (dev->flags & IFF_PROMISC)
652 		dev_set_promiscuity(lowerdev, -1);
653 
654 	dev_uc_del(lowerdev, dev->dev_addr);
655 
656 hash_del:
657 	macvlan_hash_del(vlan, !dev->dismantle);
658 	return 0;
659 }
660 
661 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
662 {
663 	struct macvlan_dev *vlan = netdev_priv(dev);
664 	struct net_device *lowerdev = vlan->lowerdev;
665 	int err;
666 
667 	if (!(dev->flags & IFF_UP)) {
668 		/* Just copy in the new address */
669 		ether_addr_copy(dev->dev_addr, addr);
670 	} else {
671 		/* Rehash and update the device filters */
672 		if (macvlan_addr_busy(vlan->port, addr))
673 			return -EBUSY;
674 
675 		if (!vlan->port->passthru) {
676 			err = dev_uc_add(lowerdev, addr);
677 			if (err)
678 				return err;
679 
680 			dev_uc_del(lowerdev, dev->dev_addr);
681 		}
682 
683 		macvlan_hash_change_addr(vlan, addr);
684 	}
685 	return 0;
686 }
687 
688 static int macvlan_set_mac_address(struct net_device *dev, void *p)
689 {
690 	struct macvlan_dev *vlan = netdev_priv(dev);
691 	struct sockaddr *addr = p;
692 
693 	if (!is_valid_ether_addr(addr->sa_data))
694 		return -EADDRNOTAVAIL;
695 
696 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
697 		dev_set_mac_address(vlan->lowerdev, addr);
698 		return 0;
699 	}
700 
701 	return macvlan_sync_address(dev, addr->sa_data);
702 }
703 
704 static void macvlan_change_rx_flags(struct net_device *dev, int change)
705 {
706 	struct macvlan_dev *vlan = netdev_priv(dev);
707 	struct net_device *lowerdev = vlan->lowerdev;
708 
709 	if (dev->flags & IFF_UP) {
710 		if (change & IFF_ALLMULTI)
711 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
712 		if (change & IFF_PROMISC)
713 			dev_set_promiscuity(lowerdev,
714 					    dev->flags & IFF_PROMISC ? 1 : -1);
715 
716 	}
717 }
718 
719 static void macvlan_set_mac_lists(struct net_device *dev)
720 {
721 	struct macvlan_dev *vlan = netdev_priv(dev);
722 
723 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
724 		bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
725 	} else {
726 		struct netdev_hw_addr *ha;
727 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
728 
729 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
730 		netdev_for_each_mc_addr(ha, dev) {
731 			__set_bit(mc_hash(vlan, ha->addr), filter);
732 		}
733 
734 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
735 
736 		bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
737 	}
738 	dev_uc_sync(vlan->lowerdev, dev);
739 	dev_mc_sync(vlan->lowerdev, dev);
740 }
741 
742 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
743 {
744 	struct macvlan_dev *vlan = netdev_priv(dev);
745 
746 	if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
747 		return -EINVAL;
748 	dev->mtu = new_mtu;
749 	return 0;
750 }
751 
752 /*
753  * macvlan network devices have devices nesting below it and are a special
754  * "super class" of normal network devices; split their locks off into a
755  * separate class since they always nest.
756  */
757 static struct lock_class_key macvlan_netdev_xmit_lock_key;
758 static struct lock_class_key macvlan_netdev_addr_lock_key;
759 
760 #define ALWAYS_ON_FEATURES \
761 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX | \
762 	 NETIF_F_GSO_ROBUST)
763 
764 #define MACVLAN_FEATURES \
765 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
766 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_LRO | \
767 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
768 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
769 
770 #define MACVLAN_STATE_MASK \
771 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
772 
773 static int macvlan_get_nest_level(struct net_device *dev)
774 {
775 	return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
776 }
777 
778 static void macvlan_set_lockdep_class_one(struct net_device *dev,
779 					  struct netdev_queue *txq,
780 					  void *_unused)
781 {
782 	lockdep_set_class(&txq->_xmit_lock,
783 			  &macvlan_netdev_xmit_lock_key);
784 }
785 
786 static void macvlan_set_lockdep_class(struct net_device *dev)
787 {
788 	lockdep_set_class_and_subclass(&dev->addr_list_lock,
789 				       &macvlan_netdev_addr_lock_key,
790 				       macvlan_get_nest_level(dev));
791 	netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
792 }
793 
794 static int macvlan_init(struct net_device *dev)
795 {
796 	struct macvlan_dev *vlan = netdev_priv(dev);
797 	const struct net_device *lowerdev = vlan->lowerdev;
798 
799 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
800 				  (lowerdev->state & MACVLAN_STATE_MASK);
801 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
802 	dev->features		|= ALWAYS_ON_FEATURES;
803 	dev->hw_features	|= NETIF_F_LRO;
804 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
805 	dev->gso_max_size	= lowerdev->gso_max_size;
806 	dev->hard_header_len	= lowerdev->hard_header_len;
807 
808 	macvlan_set_lockdep_class(dev);
809 
810 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
811 	if (!vlan->pcpu_stats)
812 		return -ENOMEM;
813 
814 	return 0;
815 }
816 
817 static void macvlan_uninit(struct net_device *dev)
818 {
819 	struct macvlan_dev *vlan = netdev_priv(dev);
820 	struct macvlan_port *port = vlan->port;
821 
822 	free_percpu(vlan->pcpu_stats);
823 
824 	macvlan_flush_sources(port, vlan);
825 	port->count -= 1;
826 	if (!port->count)
827 		macvlan_port_destroy(port->dev);
828 }
829 
830 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
831 							 struct rtnl_link_stats64 *stats)
832 {
833 	struct macvlan_dev *vlan = netdev_priv(dev);
834 
835 	if (vlan->pcpu_stats) {
836 		struct vlan_pcpu_stats *p;
837 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
838 		u32 rx_errors = 0, tx_dropped = 0;
839 		unsigned int start;
840 		int i;
841 
842 		for_each_possible_cpu(i) {
843 			p = per_cpu_ptr(vlan->pcpu_stats, i);
844 			do {
845 				start = u64_stats_fetch_begin_irq(&p->syncp);
846 				rx_packets	= p->rx_packets;
847 				rx_bytes	= p->rx_bytes;
848 				rx_multicast	= p->rx_multicast;
849 				tx_packets	= p->tx_packets;
850 				tx_bytes	= p->tx_bytes;
851 			} while (u64_stats_fetch_retry_irq(&p->syncp, start));
852 
853 			stats->rx_packets	+= rx_packets;
854 			stats->rx_bytes		+= rx_bytes;
855 			stats->multicast	+= rx_multicast;
856 			stats->tx_packets	+= tx_packets;
857 			stats->tx_bytes		+= tx_bytes;
858 			/* rx_errors & tx_dropped are u32, updated
859 			 * without syncp protection.
860 			 */
861 			rx_errors	+= p->rx_errors;
862 			tx_dropped	+= p->tx_dropped;
863 		}
864 		stats->rx_errors	= rx_errors;
865 		stats->rx_dropped	= rx_errors;
866 		stats->tx_dropped	= tx_dropped;
867 	}
868 	return stats;
869 }
870 
871 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
872 				   __be16 proto, u16 vid)
873 {
874 	struct macvlan_dev *vlan = netdev_priv(dev);
875 	struct net_device *lowerdev = vlan->lowerdev;
876 
877 	return vlan_vid_add(lowerdev, proto, vid);
878 }
879 
880 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
881 				    __be16 proto, u16 vid)
882 {
883 	struct macvlan_dev *vlan = netdev_priv(dev);
884 	struct net_device *lowerdev = vlan->lowerdev;
885 
886 	vlan_vid_del(lowerdev, proto, vid);
887 	return 0;
888 }
889 
890 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
891 			   struct net_device *dev,
892 			   const unsigned char *addr, u16 vid,
893 			   u16 flags)
894 {
895 	struct macvlan_dev *vlan = netdev_priv(dev);
896 	int err = -EINVAL;
897 
898 	/* Support unicast filter only on passthru devices.
899 	 * Multicast filter should be allowed on all devices.
900 	 */
901 	if (!vlan->port->passthru && is_unicast_ether_addr(addr))
902 		return -EOPNOTSUPP;
903 
904 	if (flags & NLM_F_REPLACE)
905 		return -EOPNOTSUPP;
906 
907 	if (is_unicast_ether_addr(addr))
908 		err = dev_uc_add_excl(dev, addr);
909 	else if (is_multicast_ether_addr(addr))
910 		err = dev_mc_add_excl(dev, addr);
911 
912 	return err;
913 }
914 
915 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
916 			   struct net_device *dev,
917 			   const unsigned char *addr, u16 vid)
918 {
919 	struct macvlan_dev *vlan = netdev_priv(dev);
920 	int err = -EINVAL;
921 
922 	/* Support unicast filter only on passthru devices.
923 	 * Multicast filter should be allowed on all devices.
924 	 */
925 	if (!vlan->port->passthru && is_unicast_ether_addr(addr))
926 		return -EOPNOTSUPP;
927 
928 	if (is_unicast_ether_addr(addr))
929 		err = dev_uc_del(dev, addr);
930 	else if (is_multicast_ether_addr(addr))
931 		err = dev_mc_del(dev, addr);
932 
933 	return err;
934 }
935 
936 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
937 					struct ethtool_drvinfo *drvinfo)
938 {
939 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
940 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
941 }
942 
943 static int macvlan_ethtool_get_settings(struct net_device *dev,
944 					struct ethtool_cmd *cmd)
945 {
946 	const struct macvlan_dev *vlan = netdev_priv(dev);
947 
948 	return __ethtool_get_settings(vlan->lowerdev, cmd);
949 }
950 
951 static netdev_features_t macvlan_fix_features(struct net_device *dev,
952 					      netdev_features_t features)
953 {
954 	struct macvlan_dev *vlan = netdev_priv(dev);
955 	netdev_features_t lowerdev_features = vlan->lowerdev->features;
956 	netdev_features_t mask;
957 
958 	features |= NETIF_F_ALL_FOR_ALL;
959 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
960 	mask = features;
961 
962 	lowerdev_features &= (features | ~NETIF_F_LRO);
963 	features = netdev_increment_features(lowerdev_features, features, mask);
964 	features |= ALWAYS_ON_FEATURES;
965 	features &= ~NETIF_F_NETNS_LOCAL;
966 
967 	return features;
968 }
969 
970 #ifdef CONFIG_NET_POLL_CONTROLLER
971 static void macvlan_dev_poll_controller(struct net_device *dev)
972 {
973 	return;
974 }
975 
976 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
977 {
978 	struct macvlan_dev *vlan = netdev_priv(dev);
979 	struct net_device *real_dev = vlan->lowerdev;
980 	struct netpoll *netpoll;
981 	int err = 0;
982 
983 	netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
984 	err = -ENOMEM;
985 	if (!netpoll)
986 		goto out;
987 
988 	err = __netpoll_setup(netpoll, real_dev);
989 	if (err) {
990 		kfree(netpoll);
991 		goto out;
992 	}
993 
994 	vlan->netpoll = netpoll;
995 
996 out:
997 	return err;
998 }
999 
1000 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1001 {
1002 	struct macvlan_dev *vlan = netdev_priv(dev);
1003 	struct netpoll *netpoll = vlan->netpoll;
1004 
1005 	if (!netpoll)
1006 		return;
1007 
1008 	vlan->netpoll = NULL;
1009 
1010 	__netpoll_free_async(netpoll);
1011 }
1012 #endif	/* CONFIG_NET_POLL_CONTROLLER */
1013 
1014 static int macvlan_dev_get_iflink(const struct net_device *dev)
1015 {
1016 	struct macvlan_dev *vlan = netdev_priv(dev);
1017 
1018 	return vlan->lowerdev->ifindex;
1019 }
1020 
1021 static const struct ethtool_ops macvlan_ethtool_ops = {
1022 	.get_link		= ethtool_op_get_link,
1023 	.get_settings		= macvlan_ethtool_get_settings,
1024 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
1025 };
1026 
1027 static const struct net_device_ops macvlan_netdev_ops = {
1028 	.ndo_init		= macvlan_init,
1029 	.ndo_uninit		= macvlan_uninit,
1030 	.ndo_open		= macvlan_open,
1031 	.ndo_stop		= macvlan_stop,
1032 	.ndo_start_xmit		= macvlan_start_xmit,
1033 	.ndo_change_mtu		= macvlan_change_mtu,
1034 	.ndo_fix_features	= macvlan_fix_features,
1035 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1036 	.ndo_set_mac_address	= macvlan_set_mac_address,
1037 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1038 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1039 	.ndo_validate_addr	= eth_validate_addr,
1040 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1041 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1042 	.ndo_fdb_add		= macvlan_fdb_add,
1043 	.ndo_fdb_del		= macvlan_fdb_del,
1044 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1045 	.ndo_get_lock_subclass  = macvlan_get_nest_level,
1046 #ifdef CONFIG_NET_POLL_CONTROLLER
1047 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1048 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1049 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1050 #endif
1051 	.ndo_get_iflink		= macvlan_dev_get_iflink,
1052 	.ndo_features_check	= passthru_features_check,
1053 };
1054 
1055 void macvlan_common_setup(struct net_device *dev)
1056 {
1057 	ether_setup(dev);
1058 
1059 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1060 	netif_keep_dst(dev);
1061 	dev->priv_flags	       |= IFF_UNICAST_FLT;
1062 	dev->netdev_ops		= &macvlan_netdev_ops;
1063 	dev->destructor		= free_netdev;
1064 	dev->header_ops		= &macvlan_hard_header_ops;
1065 	dev->ethtool_ops	= &macvlan_ethtool_ops;
1066 }
1067 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1068 
1069 static void macvlan_setup(struct net_device *dev)
1070 {
1071 	macvlan_common_setup(dev);
1072 	dev->tx_queue_len	= 0;
1073 }
1074 
1075 static int macvlan_port_create(struct net_device *dev)
1076 {
1077 	struct macvlan_port *port;
1078 	unsigned int i;
1079 	int err;
1080 
1081 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1082 		return -EINVAL;
1083 
1084 	if (netif_is_ipvlan_port(dev))
1085 		return -EBUSY;
1086 
1087 	port = kzalloc(sizeof(*port), GFP_KERNEL);
1088 	if (port == NULL)
1089 		return -ENOMEM;
1090 
1091 	port->passthru = false;
1092 	port->dev = dev;
1093 	INIT_LIST_HEAD(&port->vlans);
1094 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1095 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
1096 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1097 		INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1098 
1099 	skb_queue_head_init(&port->bc_queue);
1100 	INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1101 
1102 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1103 	if (err)
1104 		kfree(port);
1105 	else
1106 		dev->priv_flags |= IFF_MACVLAN_PORT;
1107 	return err;
1108 }
1109 
1110 static void macvlan_port_destroy(struct net_device *dev)
1111 {
1112 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1113 
1114 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
1115 	netdev_rx_handler_unregister(dev);
1116 
1117 	/* After this point, no packet can schedule bc_work anymore,
1118 	 * but we need to cancel it and purge left skbs if any.
1119 	 */
1120 	cancel_work_sync(&port->bc_work);
1121 	__skb_queue_purge(&port->bc_queue);
1122 
1123 	kfree_rcu(port, rcu);
1124 }
1125 
1126 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
1127 {
1128 	if (tb[IFLA_ADDRESS]) {
1129 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1130 			return -EINVAL;
1131 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1132 			return -EADDRNOTAVAIL;
1133 	}
1134 
1135 	if (data && data[IFLA_MACVLAN_FLAGS] &&
1136 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1137 		return -EINVAL;
1138 
1139 	if (data && data[IFLA_MACVLAN_MODE]) {
1140 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1141 		case MACVLAN_MODE_PRIVATE:
1142 		case MACVLAN_MODE_VEPA:
1143 		case MACVLAN_MODE_BRIDGE:
1144 		case MACVLAN_MODE_PASSTHRU:
1145 		case MACVLAN_MODE_SOURCE:
1146 			break;
1147 		default:
1148 			return -EINVAL;
1149 		}
1150 	}
1151 
1152 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1153 		switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1154 		case MACVLAN_MACADDR_ADD:
1155 		case MACVLAN_MACADDR_DEL:
1156 		case MACVLAN_MACADDR_FLUSH:
1157 		case MACVLAN_MACADDR_SET:
1158 			break;
1159 		default:
1160 			return -EINVAL;
1161 		}
1162 	}
1163 
1164 	if (data && data[IFLA_MACVLAN_MACADDR]) {
1165 		if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1166 			return -EINVAL;
1167 
1168 		if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1169 			return -EADDRNOTAVAIL;
1170 	}
1171 
1172 	if (data && data[IFLA_MACVLAN_MACADDR_COUNT])
1173 		return -EINVAL;
1174 
1175 	return 0;
1176 }
1177 
1178 /**
1179  * reconfigure list of remote source mac address
1180  * (only for macvlan devices in source mode)
1181  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1182  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1183  */
1184 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1185 				      struct nlattr *data[])
1186 {
1187 	char *addr = NULL;
1188 	int ret, rem, len;
1189 	struct nlattr *nla, *head;
1190 	struct macvlan_source_entry *entry;
1191 
1192 	if (data[IFLA_MACVLAN_MACADDR])
1193 		addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1194 
1195 	if (mode == MACVLAN_MACADDR_ADD) {
1196 		if (!addr)
1197 			return -EINVAL;
1198 
1199 		return macvlan_hash_add_source(vlan, addr);
1200 
1201 	} else if (mode == MACVLAN_MACADDR_DEL) {
1202 		if (!addr)
1203 			return -EINVAL;
1204 
1205 		entry = macvlan_hash_lookup_source(vlan, addr);
1206 		if (entry) {
1207 			macvlan_hash_del_source(entry);
1208 			vlan->macaddr_count--;
1209 		}
1210 	} else if (mode == MACVLAN_MACADDR_FLUSH) {
1211 		macvlan_flush_sources(vlan->port, vlan);
1212 	} else if (mode == MACVLAN_MACADDR_SET) {
1213 		macvlan_flush_sources(vlan->port, vlan);
1214 
1215 		if (addr) {
1216 			ret = macvlan_hash_add_source(vlan, addr);
1217 			if (ret)
1218 				return ret;
1219 		}
1220 
1221 		if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1222 			return 0;
1223 
1224 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1225 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1226 
1227 		nla_for_each_attr(nla, head, len, rem) {
1228 			if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1229 			    nla_len(nla) != ETH_ALEN)
1230 				continue;
1231 
1232 			addr = nla_data(nla);
1233 			ret = macvlan_hash_add_source(vlan, addr);
1234 			if (ret)
1235 				return ret;
1236 		}
1237 	} else {
1238 		return -EINVAL;
1239 	}
1240 
1241 	return 0;
1242 }
1243 
1244 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1245 			   struct nlattr *tb[], struct nlattr *data[])
1246 {
1247 	struct macvlan_dev *vlan = netdev_priv(dev);
1248 	struct macvlan_port *port;
1249 	struct net_device *lowerdev;
1250 	int err;
1251 	int macmode;
1252 
1253 	if (!tb[IFLA_LINK])
1254 		return -EINVAL;
1255 
1256 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1257 	if (lowerdev == NULL)
1258 		return -ENODEV;
1259 
1260 	/* When creating macvlans or macvtaps on top of other macvlans - use
1261 	 * the real device as the lowerdev.
1262 	 */
1263 	if (netif_is_macvlan(lowerdev))
1264 		lowerdev = macvlan_dev_real_dev(lowerdev);
1265 
1266 	if (!tb[IFLA_MTU])
1267 		dev->mtu = lowerdev->mtu;
1268 	else if (dev->mtu > lowerdev->mtu)
1269 		return -EINVAL;
1270 
1271 	if (!tb[IFLA_ADDRESS])
1272 		eth_hw_addr_random(dev);
1273 
1274 	if (!macvlan_port_exists(lowerdev)) {
1275 		err = macvlan_port_create(lowerdev);
1276 		if (err < 0)
1277 			return err;
1278 	}
1279 	port = macvlan_port_get_rtnl(lowerdev);
1280 
1281 	/* Only 1 macvlan device can be created in passthru mode */
1282 	if (port->passthru)
1283 		return -EINVAL;
1284 
1285 	vlan->lowerdev = lowerdev;
1286 	vlan->dev      = dev;
1287 	vlan->port     = port;
1288 	vlan->set_features = MACVLAN_FEATURES;
1289 	vlan->nest_level = dev_get_nest_level(lowerdev, netif_is_macvlan) + 1;
1290 
1291 	vlan->mode     = MACVLAN_MODE_VEPA;
1292 	if (data && data[IFLA_MACVLAN_MODE])
1293 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1294 
1295 	if (data && data[IFLA_MACVLAN_FLAGS])
1296 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1297 
1298 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1299 		if (port->count)
1300 			return -EINVAL;
1301 		port->passthru = true;
1302 		eth_hw_addr_inherit(dev, lowerdev);
1303 	}
1304 
1305 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1306 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1307 			return -EINVAL;
1308 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1309 		err = macvlan_changelink_sources(vlan, macmode, data);
1310 		if (err)
1311 			return err;
1312 	}
1313 
1314 	port->count += 1;
1315 	err = register_netdevice(dev);
1316 	if (err < 0)
1317 		goto destroy_port;
1318 
1319 	dev->priv_flags |= IFF_MACVLAN;
1320 	err = netdev_upper_dev_link(lowerdev, dev);
1321 	if (err)
1322 		goto unregister_netdev;
1323 
1324 	list_add_tail_rcu(&vlan->list, &port->vlans);
1325 	netif_stacked_transfer_operstate(lowerdev, dev);
1326 
1327 	return 0;
1328 
1329 unregister_netdev:
1330 	unregister_netdevice(dev);
1331 destroy_port:
1332 	port->count -= 1;
1333 	if (!port->count)
1334 		macvlan_port_destroy(lowerdev);
1335 
1336 	return err;
1337 }
1338 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1339 
1340 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1341 			   struct nlattr *tb[], struct nlattr *data[])
1342 {
1343 	return macvlan_common_newlink(src_net, dev, tb, data);
1344 }
1345 
1346 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1347 {
1348 	struct macvlan_dev *vlan = netdev_priv(dev);
1349 
1350 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1351 		macvlan_flush_sources(vlan->port, vlan);
1352 	list_del_rcu(&vlan->list);
1353 	unregister_netdevice_queue(dev, head);
1354 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1355 }
1356 EXPORT_SYMBOL_GPL(macvlan_dellink);
1357 
1358 static int macvlan_changelink(struct net_device *dev,
1359 		struct nlattr *tb[], struct nlattr *data[])
1360 {
1361 	struct macvlan_dev *vlan = netdev_priv(dev);
1362 	enum macvlan_mode mode;
1363 	bool set_mode = false;
1364 	enum macvlan_macaddr_mode macmode;
1365 	int ret;
1366 
1367 	/* Validate mode, but don't set yet: setting flags may fail. */
1368 	if (data && data[IFLA_MACVLAN_MODE]) {
1369 		set_mode = true;
1370 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1371 		/* Passthrough mode can't be set or cleared dynamically */
1372 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1373 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1374 			return -EINVAL;
1375 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1376 		    vlan->mode != mode)
1377 			macvlan_flush_sources(vlan->port, vlan);
1378 	}
1379 
1380 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1381 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1382 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1383 		if (vlan->port->passthru && promisc) {
1384 			int err;
1385 
1386 			if (flags & MACVLAN_FLAG_NOPROMISC)
1387 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1388 			else
1389 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1390 			if (err < 0)
1391 				return err;
1392 		}
1393 		vlan->flags = flags;
1394 	}
1395 	if (set_mode)
1396 		vlan->mode = mode;
1397 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1398 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1399 			return -EINVAL;
1400 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1401 		ret = macvlan_changelink_sources(vlan, macmode, data);
1402 		if (ret)
1403 			return ret;
1404 	}
1405 	return 0;
1406 }
1407 
1408 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1409 {
1410 	if (vlan->macaddr_count == 0)
1411 		return 0;
1412 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1413 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1414 }
1415 
1416 static size_t macvlan_get_size(const struct net_device *dev)
1417 {
1418 	struct macvlan_dev *vlan = netdev_priv(dev);
1419 
1420 	return (0
1421 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1422 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1423 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1424 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1425 		);
1426 }
1427 
1428 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1429 				     const struct macvlan_dev *vlan,
1430 				     const int i)
1431 {
1432 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1433 	struct macvlan_source_entry *entry;
1434 
1435 	hlist_for_each_entry_rcu(entry, h, hlist) {
1436 		if (entry->vlan != vlan)
1437 			continue;
1438 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1439 			return 1;
1440 	}
1441 	return 0;
1442 }
1443 
1444 static int macvlan_fill_info(struct sk_buff *skb,
1445 				const struct net_device *dev)
1446 {
1447 	struct macvlan_dev *vlan = netdev_priv(dev);
1448 	int i;
1449 	struct nlattr *nest;
1450 
1451 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1452 		goto nla_put_failure;
1453 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1454 		goto nla_put_failure;
1455 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1456 		goto nla_put_failure;
1457 	if (vlan->macaddr_count > 0) {
1458 		nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA);
1459 		if (nest == NULL)
1460 			goto nla_put_failure;
1461 
1462 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1463 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1464 				goto nla_put_failure;
1465 		}
1466 		nla_nest_end(skb, nest);
1467 	}
1468 	return 0;
1469 
1470 nla_put_failure:
1471 	return -EMSGSIZE;
1472 }
1473 
1474 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1475 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1476 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1477 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1478 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1479 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1480 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1481 };
1482 
1483 int macvlan_link_register(struct rtnl_link_ops *ops)
1484 {
1485 	/* common fields */
1486 	ops->priv_size		= sizeof(struct macvlan_dev);
1487 	ops->validate		= macvlan_validate;
1488 	ops->maxtype		= IFLA_MACVLAN_MAX;
1489 	ops->policy		= macvlan_policy;
1490 	ops->changelink		= macvlan_changelink;
1491 	ops->get_size		= macvlan_get_size;
1492 	ops->fill_info		= macvlan_fill_info;
1493 
1494 	return rtnl_link_register(ops);
1495 };
1496 EXPORT_SYMBOL_GPL(macvlan_link_register);
1497 
1498 static struct net *macvlan_get_link_net(const struct net_device *dev)
1499 {
1500 	return dev_net(macvlan_dev_real_dev(dev));
1501 }
1502 
1503 static struct rtnl_link_ops macvlan_link_ops = {
1504 	.kind		= "macvlan",
1505 	.setup		= macvlan_setup,
1506 	.newlink	= macvlan_newlink,
1507 	.dellink	= macvlan_dellink,
1508 	.get_link_net	= macvlan_get_link_net,
1509 };
1510 
1511 static int macvlan_device_event(struct notifier_block *unused,
1512 				unsigned long event, void *ptr)
1513 {
1514 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1515 	struct macvlan_dev *vlan, *next;
1516 	struct macvlan_port *port;
1517 	LIST_HEAD(list_kill);
1518 
1519 	if (!macvlan_port_exists(dev))
1520 		return NOTIFY_DONE;
1521 
1522 	port = macvlan_port_get_rtnl(dev);
1523 
1524 	switch (event) {
1525 	case NETDEV_CHANGE:
1526 		list_for_each_entry(vlan, &port->vlans, list)
1527 			netif_stacked_transfer_operstate(vlan->lowerdev,
1528 							 vlan->dev);
1529 		break;
1530 	case NETDEV_FEAT_CHANGE:
1531 		list_for_each_entry(vlan, &port->vlans, list) {
1532 			vlan->dev->gso_max_size = dev->gso_max_size;
1533 			netdev_update_features(vlan->dev);
1534 		}
1535 		break;
1536 	case NETDEV_CHANGEMTU:
1537 		list_for_each_entry(vlan, &port->vlans, list) {
1538 			if (vlan->dev->mtu <= dev->mtu)
1539 				continue;
1540 			dev_set_mtu(vlan->dev, dev->mtu);
1541 		}
1542 		break;
1543 	case NETDEV_CHANGEADDR:
1544 		if (!port->passthru)
1545 			return NOTIFY_DONE;
1546 
1547 		vlan = list_first_entry_or_null(&port->vlans,
1548 						struct macvlan_dev,
1549 						list);
1550 
1551 		if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1552 			return NOTIFY_BAD;
1553 
1554 		break;
1555 	case NETDEV_UNREGISTER:
1556 		/* twiddle thumbs on netns device moves */
1557 		if (dev->reg_state != NETREG_UNREGISTERING)
1558 			break;
1559 
1560 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1561 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1562 		unregister_netdevice_many(&list_kill);
1563 		break;
1564 	case NETDEV_PRE_TYPE_CHANGE:
1565 		/* Forbid underlaying device to change its type. */
1566 		return NOTIFY_BAD;
1567 
1568 	case NETDEV_NOTIFY_PEERS:
1569 	case NETDEV_BONDING_FAILOVER:
1570 	case NETDEV_RESEND_IGMP:
1571 		/* Propagate to all vlans */
1572 		list_for_each_entry(vlan, &port->vlans, list)
1573 			call_netdevice_notifiers(event, vlan->dev);
1574 	}
1575 	return NOTIFY_DONE;
1576 }
1577 
1578 static struct notifier_block macvlan_notifier_block __read_mostly = {
1579 	.notifier_call	= macvlan_device_event,
1580 };
1581 
1582 static int __init macvlan_init_module(void)
1583 {
1584 	int err;
1585 
1586 	register_netdevice_notifier(&macvlan_notifier_block);
1587 
1588 	err = macvlan_link_register(&macvlan_link_ops);
1589 	if (err < 0)
1590 		goto err1;
1591 	return 0;
1592 err1:
1593 	unregister_netdevice_notifier(&macvlan_notifier_block);
1594 	return err;
1595 }
1596 
1597 static void __exit macvlan_cleanup_module(void)
1598 {
1599 	rtnl_link_unregister(&macvlan_link_ops);
1600 	unregister_netdevice_notifier(&macvlan_notifier_block);
1601 }
1602 
1603 module_init(macvlan_init_module);
1604 module_exit(macvlan_cleanup_module);
1605 
1606 MODULE_LICENSE("GPL");
1607 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1608 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1609 MODULE_ALIAS_RTNL_LINK("macvlan");
1610