xref: /openbmc/linux/drivers/net/macvlan.c (revision 62dbe83015ed15b307eb22b89568dd41091dda76)
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(skb, IP_DEFRAG_MACVLAN);
416 		if (!skb)
417 			return RX_HANDLER_CONSUMED;
418 		eth = eth_hdr(skb);
419 		macvlan_forward_source(skb, port, eth->h_source);
420 		src = macvlan_hash_lookup(port, eth->h_source);
421 		if (src && src->mode != MACVLAN_MODE_VEPA &&
422 		    src->mode != MACVLAN_MODE_BRIDGE) {
423 			/* forward to original port. */
424 			vlan = src;
425 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
426 			      netif_rx(skb);
427 			handle_res = RX_HANDLER_CONSUMED;
428 			goto out;
429 		}
430 
431 		MACVLAN_SKB_CB(skb)->src = src;
432 		macvlan_broadcast_enqueue(port, skb);
433 
434 		return RX_HANDLER_PASS;
435 	}
436 
437 	macvlan_forward_source(skb, port, eth->h_source);
438 	if (port->passthru)
439 		vlan = list_first_or_null_rcu(&port->vlans,
440 					      struct macvlan_dev, list);
441 	else
442 		vlan = macvlan_hash_lookup(port, eth->h_dest);
443 	if (vlan == NULL)
444 		return RX_HANDLER_PASS;
445 
446 	dev = vlan->dev;
447 	if (unlikely(!(dev->flags & IFF_UP))) {
448 		kfree_skb(skb);
449 		return RX_HANDLER_CONSUMED;
450 	}
451 	len = skb->len + ETH_HLEN;
452 	skb = skb_share_check(skb, GFP_ATOMIC);
453 	if (!skb) {
454 		ret = NET_RX_DROP;
455 		handle_res = RX_HANDLER_CONSUMED;
456 		goto out;
457 	}
458 
459 	skb->dev = dev;
460 	skb->pkt_type = PACKET_HOST;
461 
462 	ret = NET_RX_SUCCESS;
463 	handle_res = RX_HANDLER_ANOTHER;
464 out:
465 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
466 	return handle_res;
467 }
468 
469 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
470 {
471 	const struct macvlan_dev *vlan = netdev_priv(dev);
472 	const struct macvlan_port *port = vlan->port;
473 	const struct macvlan_dev *dest;
474 
475 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
476 		const struct ethhdr *eth = (void *)skb->data;
477 
478 		/* send to other bridge ports directly */
479 		if (is_multicast_ether_addr(eth->h_dest)) {
480 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
481 			goto xmit_world;
482 		}
483 
484 		dest = macvlan_hash_lookup(port, eth->h_dest);
485 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
486 			/* send to lowerdev first for its network taps */
487 			dev_forward_skb(vlan->lowerdev, skb);
488 
489 			return NET_XMIT_SUCCESS;
490 		}
491 	}
492 
493 xmit_world:
494 	skb->dev = vlan->lowerdev;
495 	return dev_queue_xmit(skb);
496 }
497 
498 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
499 {
500 #ifdef CONFIG_NET_POLL_CONTROLLER
501 	if (vlan->netpoll)
502 		netpoll_send_skb(vlan->netpoll, skb);
503 #else
504 	BUG();
505 #endif
506 	return NETDEV_TX_OK;
507 }
508 
509 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
510 				      struct net_device *dev)
511 {
512 	unsigned int len = skb->len;
513 	int ret;
514 	struct macvlan_dev *vlan = netdev_priv(dev);
515 
516 	if (unlikely(netpoll_tx_running(dev)))
517 		return macvlan_netpoll_send_skb(vlan, skb);
518 
519 	if (vlan->fwd_priv) {
520 		skb->dev = vlan->lowerdev;
521 		ret = dev_queue_xmit_accel(skb, vlan->fwd_priv);
522 	} else {
523 		ret = macvlan_queue_xmit(skb, dev);
524 	}
525 
526 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
527 		struct vlan_pcpu_stats *pcpu_stats;
528 
529 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
530 		u64_stats_update_begin(&pcpu_stats->syncp);
531 		pcpu_stats->tx_packets++;
532 		pcpu_stats->tx_bytes += len;
533 		u64_stats_update_end(&pcpu_stats->syncp);
534 	} else {
535 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
536 	}
537 	return ret;
538 }
539 
540 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
541 			       unsigned short type, const void *daddr,
542 			       const void *saddr, unsigned len)
543 {
544 	const struct macvlan_dev *vlan = netdev_priv(dev);
545 	struct net_device *lowerdev = vlan->lowerdev;
546 
547 	return dev_hard_header(skb, lowerdev, type, daddr,
548 			       saddr ? : dev->dev_addr, len);
549 }
550 
551 static const struct header_ops macvlan_hard_header_ops = {
552 	.create  	= macvlan_hard_header,
553 	.rebuild	= eth_rebuild_header,
554 	.parse		= eth_header_parse,
555 	.cache		= eth_header_cache,
556 	.cache_update	= eth_header_cache_update,
557 };
558 
559 static struct rtnl_link_ops macvlan_link_ops;
560 
561 static int macvlan_open(struct net_device *dev)
562 {
563 	struct macvlan_dev *vlan = netdev_priv(dev);
564 	struct net_device *lowerdev = vlan->lowerdev;
565 	int err;
566 
567 	if (vlan->port->passthru) {
568 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
569 			err = dev_set_promiscuity(lowerdev, 1);
570 			if (err < 0)
571 				goto out;
572 		}
573 		goto hash_add;
574 	}
575 
576 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD &&
577 	    dev->rtnl_link_ops == &macvlan_link_ops) {
578 		vlan->fwd_priv =
579 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
580 
581 		/* If we get a NULL pointer back, or if we get an error
582 		 * then we should just fall through to the non accelerated path
583 		 */
584 		if (IS_ERR_OR_NULL(vlan->fwd_priv)) {
585 			vlan->fwd_priv = NULL;
586 		} else
587 			return 0;
588 	}
589 
590 	err = -EBUSY;
591 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
592 		goto out;
593 
594 	err = dev_uc_add(lowerdev, dev->dev_addr);
595 	if (err < 0)
596 		goto out;
597 	if (dev->flags & IFF_ALLMULTI) {
598 		err = dev_set_allmulti(lowerdev, 1);
599 		if (err < 0)
600 			goto del_unicast;
601 	}
602 
603 hash_add:
604 	macvlan_hash_add(vlan);
605 	return 0;
606 
607 del_unicast:
608 	dev_uc_del(lowerdev, dev->dev_addr);
609 out:
610 	if (vlan->fwd_priv) {
611 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
612 							   vlan->fwd_priv);
613 		vlan->fwd_priv = NULL;
614 	}
615 	return err;
616 }
617 
618 static int macvlan_stop(struct net_device *dev)
619 {
620 	struct macvlan_dev *vlan = netdev_priv(dev);
621 	struct net_device *lowerdev = vlan->lowerdev;
622 
623 	if (vlan->fwd_priv) {
624 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
625 							   vlan->fwd_priv);
626 		vlan->fwd_priv = NULL;
627 		return 0;
628 	}
629 
630 	dev_uc_unsync(lowerdev, dev);
631 	dev_mc_unsync(lowerdev, dev);
632 
633 	if (vlan->port->passthru) {
634 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
635 			dev_set_promiscuity(lowerdev, -1);
636 		goto hash_del;
637 	}
638 
639 	if (dev->flags & IFF_ALLMULTI)
640 		dev_set_allmulti(lowerdev, -1);
641 
642 	dev_uc_del(lowerdev, dev->dev_addr);
643 
644 hash_del:
645 	macvlan_hash_del(vlan, !dev->dismantle);
646 	return 0;
647 }
648 
649 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
650 {
651 	struct macvlan_dev *vlan = netdev_priv(dev);
652 	struct net_device *lowerdev = vlan->lowerdev;
653 	int err;
654 
655 	if (!(dev->flags & IFF_UP)) {
656 		/* Just copy in the new address */
657 		ether_addr_copy(dev->dev_addr, addr);
658 	} else {
659 		/* Rehash and update the device filters */
660 		if (macvlan_addr_busy(vlan->port, addr))
661 			return -EBUSY;
662 
663 		if (!vlan->port->passthru) {
664 			err = dev_uc_add(lowerdev, addr);
665 			if (err)
666 				return err;
667 
668 			dev_uc_del(lowerdev, dev->dev_addr);
669 		}
670 
671 		macvlan_hash_change_addr(vlan, addr);
672 	}
673 	return 0;
674 }
675 
676 static int macvlan_set_mac_address(struct net_device *dev, void *p)
677 {
678 	struct macvlan_dev *vlan = netdev_priv(dev);
679 	struct sockaddr *addr = p;
680 
681 	if (!is_valid_ether_addr(addr->sa_data))
682 		return -EADDRNOTAVAIL;
683 
684 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
685 		dev_set_mac_address(vlan->lowerdev, addr);
686 		return 0;
687 	}
688 
689 	return macvlan_sync_address(dev, addr->sa_data);
690 }
691 
692 static void macvlan_change_rx_flags(struct net_device *dev, int change)
693 {
694 	struct macvlan_dev *vlan = netdev_priv(dev);
695 	struct net_device *lowerdev = vlan->lowerdev;
696 
697 	if (dev->flags & IFF_UP) {
698 		if (change & IFF_ALLMULTI)
699 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
700 	}
701 }
702 
703 static void macvlan_set_mac_lists(struct net_device *dev)
704 {
705 	struct macvlan_dev *vlan = netdev_priv(dev);
706 
707 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
708 		bitmap_fill(vlan->mc_filter, MACVLAN_MC_FILTER_SZ);
709 	} else {
710 		struct netdev_hw_addr *ha;
711 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
712 
713 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
714 		netdev_for_each_mc_addr(ha, dev) {
715 			__set_bit(mc_hash(vlan, ha->addr), filter);
716 		}
717 
718 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
719 
720 		bitmap_copy(vlan->mc_filter, filter, MACVLAN_MC_FILTER_SZ);
721 	}
722 	dev_uc_sync(vlan->lowerdev, dev);
723 	dev_mc_sync(vlan->lowerdev, dev);
724 }
725 
726 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
727 {
728 	struct macvlan_dev *vlan = netdev_priv(dev);
729 
730 	if (new_mtu < 68 || vlan->lowerdev->mtu < new_mtu)
731 		return -EINVAL;
732 	dev->mtu = new_mtu;
733 	return 0;
734 }
735 
736 /*
737  * macvlan network devices have devices nesting below it and are a special
738  * "super class" of normal network devices; split their locks off into a
739  * separate class since they always nest.
740  */
741 static struct lock_class_key macvlan_netdev_xmit_lock_key;
742 static struct lock_class_key macvlan_netdev_addr_lock_key;
743 
744 #define ALWAYS_ON_FEATURES \
745 	(NETIF_F_SG | NETIF_F_GEN_CSUM | NETIF_F_GSO_SOFTWARE | NETIF_F_LLTX | \
746 	 NETIF_F_GSO_ROBUST)
747 
748 #define MACVLAN_FEATURES \
749 	(NETIF_F_SG | NETIF_F_ALL_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
750 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_UFO | NETIF_F_LRO | \
751 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
752 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
753 
754 #define MACVLAN_STATE_MASK \
755 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
756 
757 static int macvlan_get_nest_level(struct net_device *dev)
758 {
759 	return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
760 }
761 
762 static void macvlan_set_lockdep_class_one(struct net_device *dev,
763 					  struct netdev_queue *txq,
764 					  void *_unused)
765 {
766 	lockdep_set_class(&txq->_xmit_lock,
767 			  &macvlan_netdev_xmit_lock_key);
768 }
769 
770 static void macvlan_set_lockdep_class(struct net_device *dev)
771 {
772 	lockdep_set_class_and_subclass(&dev->addr_list_lock,
773 				       &macvlan_netdev_addr_lock_key,
774 				       macvlan_get_nest_level(dev));
775 	netdev_for_each_tx_queue(dev, macvlan_set_lockdep_class_one, NULL);
776 }
777 
778 static int macvlan_init(struct net_device *dev)
779 {
780 	struct macvlan_dev *vlan = netdev_priv(dev);
781 	const struct net_device *lowerdev = vlan->lowerdev;
782 
783 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
784 				  (lowerdev->state & MACVLAN_STATE_MASK);
785 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
786 	dev->features		|= ALWAYS_ON_FEATURES;
787 	dev->hw_features	|= NETIF_F_LRO;
788 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
789 	dev->gso_max_size	= lowerdev->gso_max_size;
790 	dev->iflink		= lowerdev->ifindex;
791 	dev->hard_header_len	= lowerdev->hard_header_len;
792 
793 	macvlan_set_lockdep_class(dev);
794 
795 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
796 	if (!vlan->pcpu_stats)
797 		return -ENOMEM;
798 
799 	return 0;
800 }
801 
802 static void macvlan_uninit(struct net_device *dev)
803 {
804 	struct macvlan_dev *vlan = netdev_priv(dev);
805 	struct macvlan_port *port = vlan->port;
806 
807 	free_percpu(vlan->pcpu_stats);
808 
809 	macvlan_flush_sources(port, vlan);
810 	port->count -= 1;
811 	if (!port->count)
812 		macvlan_port_destroy(port->dev);
813 }
814 
815 static struct rtnl_link_stats64 *macvlan_dev_get_stats64(struct net_device *dev,
816 							 struct rtnl_link_stats64 *stats)
817 {
818 	struct macvlan_dev *vlan = netdev_priv(dev);
819 
820 	if (vlan->pcpu_stats) {
821 		struct vlan_pcpu_stats *p;
822 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
823 		u32 rx_errors = 0, tx_dropped = 0;
824 		unsigned int start;
825 		int i;
826 
827 		for_each_possible_cpu(i) {
828 			p = per_cpu_ptr(vlan->pcpu_stats, i);
829 			do {
830 				start = u64_stats_fetch_begin_irq(&p->syncp);
831 				rx_packets	= p->rx_packets;
832 				rx_bytes	= p->rx_bytes;
833 				rx_multicast	= p->rx_multicast;
834 				tx_packets	= p->tx_packets;
835 				tx_bytes	= p->tx_bytes;
836 			} while (u64_stats_fetch_retry_irq(&p->syncp, start));
837 
838 			stats->rx_packets	+= rx_packets;
839 			stats->rx_bytes		+= rx_bytes;
840 			stats->multicast	+= rx_multicast;
841 			stats->tx_packets	+= tx_packets;
842 			stats->tx_bytes		+= tx_bytes;
843 			/* rx_errors & tx_dropped are u32, updated
844 			 * without syncp protection.
845 			 */
846 			rx_errors	+= p->rx_errors;
847 			tx_dropped	+= p->tx_dropped;
848 		}
849 		stats->rx_errors	= rx_errors;
850 		stats->rx_dropped	= rx_errors;
851 		stats->tx_dropped	= tx_dropped;
852 	}
853 	return stats;
854 }
855 
856 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
857 				   __be16 proto, u16 vid)
858 {
859 	struct macvlan_dev *vlan = netdev_priv(dev);
860 	struct net_device *lowerdev = vlan->lowerdev;
861 
862 	return vlan_vid_add(lowerdev, proto, vid);
863 }
864 
865 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
866 				    __be16 proto, u16 vid)
867 {
868 	struct macvlan_dev *vlan = netdev_priv(dev);
869 	struct net_device *lowerdev = vlan->lowerdev;
870 
871 	vlan_vid_del(lowerdev, proto, vid);
872 	return 0;
873 }
874 
875 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
876 			   struct net_device *dev,
877 			   const unsigned char *addr, u16 vid,
878 			   u16 flags)
879 {
880 	struct macvlan_dev *vlan = netdev_priv(dev);
881 	int err = -EINVAL;
882 
883 	/* Support unicast filter only on passthru devices.
884 	 * Multicast filter should be allowed on all devices.
885 	 */
886 	if (!vlan->port->passthru && is_unicast_ether_addr(addr))
887 		return -EOPNOTSUPP;
888 
889 	if (flags & NLM_F_REPLACE)
890 		return -EOPNOTSUPP;
891 
892 	if (is_unicast_ether_addr(addr))
893 		err = dev_uc_add_excl(dev, addr);
894 	else if (is_multicast_ether_addr(addr))
895 		err = dev_mc_add_excl(dev, addr);
896 
897 	return err;
898 }
899 
900 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
901 			   struct net_device *dev,
902 			   const unsigned char *addr, u16 vid)
903 {
904 	struct macvlan_dev *vlan = netdev_priv(dev);
905 	int err = -EINVAL;
906 
907 	/* Support unicast filter only on passthru devices.
908 	 * Multicast filter should be allowed on all devices.
909 	 */
910 	if (!vlan->port->passthru && is_unicast_ether_addr(addr))
911 		return -EOPNOTSUPP;
912 
913 	if (is_unicast_ether_addr(addr))
914 		err = dev_uc_del(dev, addr);
915 	else if (is_multicast_ether_addr(addr))
916 		err = dev_mc_del(dev, addr);
917 
918 	return err;
919 }
920 
921 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
922 					struct ethtool_drvinfo *drvinfo)
923 {
924 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
925 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
926 }
927 
928 static int macvlan_ethtool_get_settings(struct net_device *dev,
929 					struct ethtool_cmd *cmd)
930 {
931 	const struct macvlan_dev *vlan = netdev_priv(dev);
932 
933 	return __ethtool_get_settings(vlan->lowerdev, cmd);
934 }
935 
936 static netdev_features_t macvlan_fix_features(struct net_device *dev,
937 					      netdev_features_t features)
938 {
939 	struct macvlan_dev *vlan = netdev_priv(dev);
940 	netdev_features_t lowerdev_features = vlan->lowerdev->features;
941 	netdev_features_t mask;
942 
943 	features |= NETIF_F_ALL_FOR_ALL;
944 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
945 	mask = features;
946 
947 	lowerdev_features &= (features | ~NETIF_F_LRO);
948 	features = netdev_increment_features(lowerdev_features, features, mask);
949 	features |= ALWAYS_ON_FEATURES;
950 	features &= ~NETIF_F_NETNS_LOCAL;
951 
952 	return features;
953 }
954 
955 #ifdef CONFIG_NET_POLL_CONTROLLER
956 static void macvlan_dev_poll_controller(struct net_device *dev)
957 {
958 	return;
959 }
960 
961 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
962 {
963 	struct macvlan_dev *vlan = netdev_priv(dev);
964 	struct net_device *real_dev = vlan->lowerdev;
965 	struct netpoll *netpoll;
966 	int err = 0;
967 
968 	netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
969 	err = -ENOMEM;
970 	if (!netpoll)
971 		goto out;
972 
973 	err = __netpoll_setup(netpoll, real_dev);
974 	if (err) {
975 		kfree(netpoll);
976 		goto out;
977 	}
978 
979 	vlan->netpoll = netpoll;
980 
981 out:
982 	return err;
983 }
984 
985 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
986 {
987 	struct macvlan_dev *vlan = netdev_priv(dev);
988 	struct netpoll *netpoll = vlan->netpoll;
989 
990 	if (!netpoll)
991 		return;
992 
993 	vlan->netpoll = NULL;
994 
995 	__netpoll_free_async(netpoll);
996 }
997 #endif	/* CONFIG_NET_POLL_CONTROLLER */
998 
999 static const struct ethtool_ops macvlan_ethtool_ops = {
1000 	.get_link		= ethtool_op_get_link,
1001 	.get_settings		= macvlan_ethtool_get_settings,
1002 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
1003 };
1004 
1005 static const struct net_device_ops macvlan_netdev_ops = {
1006 	.ndo_init		= macvlan_init,
1007 	.ndo_uninit		= macvlan_uninit,
1008 	.ndo_open		= macvlan_open,
1009 	.ndo_stop		= macvlan_stop,
1010 	.ndo_start_xmit		= macvlan_start_xmit,
1011 	.ndo_change_mtu		= macvlan_change_mtu,
1012 	.ndo_fix_features	= macvlan_fix_features,
1013 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1014 	.ndo_set_mac_address	= macvlan_set_mac_address,
1015 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1016 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1017 	.ndo_validate_addr	= eth_validate_addr,
1018 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1019 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1020 	.ndo_fdb_add		= macvlan_fdb_add,
1021 	.ndo_fdb_del		= macvlan_fdb_del,
1022 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1023 	.ndo_get_lock_subclass  = macvlan_get_nest_level,
1024 #ifdef CONFIG_NET_POLL_CONTROLLER
1025 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1026 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1027 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1028 #endif
1029 };
1030 
1031 void macvlan_common_setup(struct net_device *dev)
1032 {
1033 	ether_setup(dev);
1034 
1035 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1036 	netif_keep_dst(dev);
1037 	dev->priv_flags	       |= IFF_UNICAST_FLT;
1038 	dev->netdev_ops		= &macvlan_netdev_ops;
1039 	dev->destructor		= free_netdev;
1040 	dev->header_ops		= &macvlan_hard_header_ops;
1041 	dev->ethtool_ops	= &macvlan_ethtool_ops;
1042 }
1043 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1044 
1045 static void macvlan_setup(struct net_device *dev)
1046 {
1047 	macvlan_common_setup(dev);
1048 	dev->tx_queue_len	= 0;
1049 }
1050 
1051 static int macvlan_port_create(struct net_device *dev)
1052 {
1053 	struct macvlan_port *port;
1054 	unsigned int i;
1055 	int err;
1056 
1057 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1058 		return -EINVAL;
1059 
1060 	port = kzalloc(sizeof(*port), GFP_KERNEL);
1061 	if (port == NULL)
1062 		return -ENOMEM;
1063 
1064 	port->passthru = false;
1065 	port->dev = dev;
1066 	INIT_LIST_HEAD(&port->vlans);
1067 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1068 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
1069 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1070 		INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1071 
1072 	skb_queue_head_init(&port->bc_queue);
1073 	INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1074 
1075 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1076 	if (err)
1077 		kfree(port);
1078 	else
1079 		dev->priv_flags |= IFF_MACVLAN_PORT;
1080 	return err;
1081 }
1082 
1083 static void macvlan_port_destroy(struct net_device *dev)
1084 {
1085 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1086 
1087 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
1088 	netdev_rx_handler_unregister(dev);
1089 
1090 	/* After this point, no packet can schedule bc_work anymore,
1091 	 * but we need to cancel it and purge left skbs if any.
1092 	 */
1093 	cancel_work_sync(&port->bc_work);
1094 	__skb_queue_purge(&port->bc_queue);
1095 
1096 	kfree_rcu(port, rcu);
1097 }
1098 
1099 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[])
1100 {
1101 	if (tb[IFLA_ADDRESS]) {
1102 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1103 			return -EINVAL;
1104 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1105 			return -EADDRNOTAVAIL;
1106 	}
1107 
1108 	if (data && data[IFLA_MACVLAN_FLAGS] &&
1109 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1110 		return -EINVAL;
1111 
1112 	if (data && data[IFLA_MACVLAN_MODE]) {
1113 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1114 		case MACVLAN_MODE_PRIVATE:
1115 		case MACVLAN_MODE_VEPA:
1116 		case MACVLAN_MODE_BRIDGE:
1117 		case MACVLAN_MODE_PASSTHRU:
1118 		case MACVLAN_MODE_SOURCE:
1119 			break;
1120 		default:
1121 			return -EINVAL;
1122 		}
1123 	}
1124 
1125 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1126 		switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1127 		case MACVLAN_MACADDR_ADD:
1128 		case MACVLAN_MACADDR_DEL:
1129 		case MACVLAN_MACADDR_FLUSH:
1130 		case MACVLAN_MACADDR_SET:
1131 			break;
1132 		default:
1133 			return -EINVAL;
1134 		}
1135 	}
1136 
1137 	if (data && data[IFLA_MACVLAN_MACADDR]) {
1138 		if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1139 			return -EINVAL;
1140 
1141 		if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1142 			return -EADDRNOTAVAIL;
1143 	}
1144 
1145 	if (data && data[IFLA_MACVLAN_MACADDR_COUNT])
1146 		return -EINVAL;
1147 
1148 	return 0;
1149 }
1150 
1151 /**
1152  * reconfigure list of remote source mac address
1153  * (only for macvlan devices in source mode)
1154  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1155  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1156  */
1157 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1158 				      struct nlattr *data[])
1159 {
1160 	char *addr = NULL;
1161 	int ret, rem, len;
1162 	struct nlattr *nla, *head;
1163 	struct macvlan_source_entry *entry;
1164 
1165 	if (data[IFLA_MACVLAN_MACADDR])
1166 		addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1167 
1168 	if (mode == MACVLAN_MACADDR_ADD) {
1169 		if (!addr)
1170 			return -EINVAL;
1171 
1172 		return macvlan_hash_add_source(vlan, addr);
1173 
1174 	} else if (mode == MACVLAN_MACADDR_DEL) {
1175 		if (!addr)
1176 			return -EINVAL;
1177 
1178 		entry = macvlan_hash_lookup_source(vlan, addr);
1179 		if (entry) {
1180 			macvlan_hash_del_source(entry);
1181 			vlan->macaddr_count--;
1182 		}
1183 	} else if (mode == MACVLAN_MACADDR_FLUSH) {
1184 		macvlan_flush_sources(vlan->port, vlan);
1185 	} else if (mode == MACVLAN_MACADDR_SET) {
1186 		macvlan_flush_sources(vlan->port, vlan);
1187 
1188 		if (addr) {
1189 			ret = macvlan_hash_add_source(vlan, addr);
1190 			if (ret)
1191 				return ret;
1192 		}
1193 
1194 		if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1195 			return 0;
1196 
1197 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1198 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1199 
1200 		nla_for_each_attr(nla, head, len, rem) {
1201 			if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1202 			    nla_len(nla) != ETH_ALEN)
1203 				continue;
1204 
1205 			addr = nla_data(nla);
1206 			ret = macvlan_hash_add_source(vlan, addr);
1207 			if (ret)
1208 				return ret;
1209 		}
1210 	} else {
1211 		return -EINVAL;
1212 	}
1213 
1214 	return 0;
1215 }
1216 
1217 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1218 			   struct nlattr *tb[], struct nlattr *data[])
1219 {
1220 	struct macvlan_dev *vlan = netdev_priv(dev);
1221 	struct macvlan_port *port;
1222 	struct net_device *lowerdev;
1223 	int err;
1224 	int macmode;
1225 
1226 	if (!tb[IFLA_LINK])
1227 		return -EINVAL;
1228 
1229 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1230 	if (lowerdev == NULL)
1231 		return -ENODEV;
1232 
1233 	/* When creating macvlans or macvtaps on top of other macvlans - use
1234 	 * the real device as the lowerdev.
1235 	 */
1236 	if (netif_is_macvlan(lowerdev))
1237 		lowerdev = macvlan_dev_real_dev(lowerdev);
1238 
1239 	if (!tb[IFLA_MTU])
1240 		dev->mtu = lowerdev->mtu;
1241 	else if (dev->mtu > lowerdev->mtu)
1242 		return -EINVAL;
1243 
1244 	if (!tb[IFLA_ADDRESS])
1245 		eth_hw_addr_random(dev);
1246 
1247 	if (!macvlan_port_exists(lowerdev)) {
1248 		err = macvlan_port_create(lowerdev);
1249 		if (err < 0)
1250 			return err;
1251 	}
1252 	port = macvlan_port_get_rtnl(lowerdev);
1253 
1254 	/* Only 1 macvlan device can be created in passthru mode */
1255 	if (port->passthru)
1256 		return -EINVAL;
1257 
1258 	vlan->lowerdev = lowerdev;
1259 	vlan->dev      = dev;
1260 	vlan->port     = port;
1261 	vlan->set_features = MACVLAN_FEATURES;
1262 	vlan->nest_level = dev_get_nest_level(lowerdev, netif_is_macvlan) + 1;
1263 
1264 	vlan->mode     = MACVLAN_MODE_VEPA;
1265 	if (data && data[IFLA_MACVLAN_MODE])
1266 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1267 
1268 	if (data && data[IFLA_MACVLAN_FLAGS])
1269 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1270 
1271 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1272 		if (port->count)
1273 			return -EINVAL;
1274 		port->passthru = true;
1275 		eth_hw_addr_inherit(dev, lowerdev);
1276 	}
1277 
1278 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1279 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1280 			return -EINVAL;
1281 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1282 		err = macvlan_changelink_sources(vlan, macmode, data);
1283 		if (err)
1284 			return err;
1285 	}
1286 
1287 	port->count += 1;
1288 	err = register_netdevice(dev);
1289 	if (err < 0)
1290 		goto destroy_port;
1291 
1292 	dev->priv_flags |= IFF_MACVLAN;
1293 	err = netdev_upper_dev_link(lowerdev, dev);
1294 	if (err)
1295 		goto unregister_netdev;
1296 
1297 	list_add_tail_rcu(&vlan->list, &port->vlans);
1298 	netif_stacked_transfer_operstate(lowerdev, dev);
1299 
1300 	return 0;
1301 
1302 unregister_netdev:
1303 	unregister_netdevice(dev);
1304 destroy_port:
1305 	port->count -= 1;
1306 	if (!port->count)
1307 		macvlan_port_destroy(lowerdev);
1308 
1309 	return err;
1310 }
1311 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1312 
1313 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1314 			   struct nlattr *tb[], struct nlattr *data[])
1315 {
1316 	return macvlan_common_newlink(src_net, dev, tb, data);
1317 }
1318 
1319 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1320 {
1321 	struct macvlan_dev *vlan = netdev_priv(dev);
1322 
1323 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1324 		macvlan_flush_sources(vlan->port, vlan);
1325 	list_del_rcu(&vlan->list);
1326 	unregister_netdevice_queue(dev, head);
1327 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1328 }
1329 EXPORT_SYMBOL_GPL(macvlan_dellink);
1330 
1331 static int macvlan_changelink(struct net_device *dev,
1332 		struct nlattr *tb[], struct nlattr *data[])
1333 {
1334 	struct macvlan_dev *vlan = netdev_priv(dev);
1335 	enum macvlan_mode mode;
1336 	bool set_mode = false;
1337 	enum macvlan_macaddr_mode macmode;
1338 	int ret;
1339 
1340 	/* Validate mode, but don't set yet: setting flags may fail. */
1341 	if (data && data[IFLA_MACVLAN_MODE]) {
1342 		set_mode = true;
1343 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1344 		/* Passthrough mode can't be set or cleared dynamically */
1345 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1346 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1347 			return -EINVAL;
1348 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1349 		    vlan->mode != mode)
1350 			macvlan_flush_sources(vlan->port, vlan);
1351 	}
1352 
1353 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1354 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1355 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1356 		if (vlan->port->passthru && promisc) {
1357 			int err;
1358 
1359 			if (flags & MACVLAN_FLAG_NOPROMISC)
1360 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1361 			else
1362 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1363 			if (err < 0)
1364 				return err;
1365 		}
1366 		vlan->flags = flags;
1367 	}
1368 	if (set_mode)
1369 		vlan->mode = mode;
1370 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1371 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1372 			return -EINVAL;
1373 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1374 		ret = macvlan_changelink_sources(vlan, macmode, data);
1375 		if (ret)
1376 			return ret;
1377 	}
1378 	return 0;
1379 }
1380 
1381 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1382 {
1383 	if (vlan->macaddr_count == 0)
1384 		return 0;
1385 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1386 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1387 }
1388 
1389 static size_t macvlan_get_size(const struct net_device *dev)
1390 {
1391 	struct macvlan_dev *vlan = netdev_priv(dev);
1392 
1393 	return (0
1394 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1395 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1396 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1397 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1398 		);
1399 }
1400 
1401 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1402 				     const struct macvlan_dev *vlan,
1403 				     const int i)
1404 {
1405 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1406 	struct macvlan_source_entry *entry;
1407 
1408 	hlist_for_each_entry_rcu(entry, h, hlist) {
1409 		if (entry->vlan != vlan)
1410 			continue;
1411 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1412 			return 1;
1413 	}
1414 	return 0;
1415 }
1416 
1417 static int macvlan_fill_info(struct sk_buff *skb,
1418 				const struct net_device *dev)
1419 {
1420 	struct macvlan_dev *vlan = netdev_priv(dev);
1421 	int i;
1422 	struct nlattr *nest;
1423 
1424 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1425 		goto nla_put_failure;
1426 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1427 		goto nla_put_failure;
1428 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1429 		goto nla_put_failure;
1430 	if (vlan->macaddr_count > 0) {
1431 		nest = nla_nest_start(skb, IFLA_MACVLAN_MACADDR_DATA);
1432 		if (nest == NULL)
1433 			goto nla_put_failure;
1434 
1435 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1436 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1437 				goto nla_put_failure;
1438 		}
1439 		nla_nest_end(skb, nest);
1440 	}
1441 	return 0;
1442 
1443 nla_put_failure:
1444 	return -EMSGSIZE;
1445 }
1446 
1447 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1448 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1449 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1450 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1451 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1452 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1453 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1454 };
1455 
1456 int macvlan_link_register(struct rtnl_link_ops *ops)
1457 {
1458 	/* common fields */
1459 	ops->priv_size		= sizeof(struct macvlan_dev);
1460 	ops->validate		= macvlan_validate;
1461 	ops->maxtype		= IFLA_MACVLAN_MAX;
1462 	ops->policy		= macvlan_policy;
1463 	ops->changelink		= macvlan_changelink;
1464 	ops->get_size		= macvlan_get_size;
1465 	ops->fill_info		= macvlan_fill_info;
1466 
1467 	return rtnl_link_register(ops);
1468 };
1469 EXPORT_SYMBOL_GPL(macvlan_link_register);
1470 
1471 static struct rtnl_link_ops macvlan_link_ops = {
1472 	.kind		= "macvlan",
1473 	.setup		= macvlan_setup,
1474 	.newlink	= macvlan_newlink,
1475 	.dellink	= macvlan_dellink,
1476 };
1477 
1478 static int macvlan_device_event(struct notifier_block *unused,
1479 				unsigned long event, void *ptr)
1480 {
1481 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1482 	struct macvlan_dev *vlan, *next;
1483 	struct macvlan_port *port;
1484 	LIST_HEAD(list_kill);
1485 
1486 	if (!macvlan_port_exists(dev))
1487 		return NOTIFY_DONE;
1488 
1489 	port = macvlan_port_get_rtnl(dev);
1490 
1491 	switch (event) {
1492 	case NETDEV_CHANGE:
1493 		list_for_each_entry(vlan, &port->vlans, list)
1494 			netif_stacked_transfer_operstate(vlan->lowerdev,
1495 							 vlan->dev);
1496 		break;
1497 	case NETDEV_FEAT_CHANGE:
1498 		list_for_each_entry(vlan, &port->vlans, list) {
1499 			vlan->dev->gso_max_size = dev->gso_max_size;
1500 			netdev_update_features(vlan->dev);
1501 		}
1502 		break;
1503 	case NETDEV_CHANGEMTU:
1504 		list_for_each_entry(vlan, &port->vlans, list) {
1505 			if (vlan->dev->mtu <= dev->mtu)
1506 				continue;
1507 			dev_set_mtu(vlan->dev, dev->mtu);
1508 		}
1509 		break;
1510 	case NETDEV_CHANGEADDR:
1511 		if (!port->passthru)
1512 			return NOTIFY_DONE;
1513 
1514 		vlan = list_first_entry_or_null(&port->vlans,
1515 						struct macvlan_dev,
1516 						list);
1517 
1518 		if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1519 			return NOTIFY_BAD;
1520 
1521 		break;
1522 	case NETDEV_UNREGISTER:
1523 		/* twiddle thumbs on netns device moves */
1524 		if (dev->reg_state != NETREG_UNREGISTERING)
1525 			break;
1526 
1527 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1528 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1529 		unregister_netdevice_many(&list_kill);
1530 		break;
1531 	case NETDEV_PRE_TYPE_CHANGE:
1532 		/* Forbid underlaying device to change its type. */
1533 		return NOTIFY_BAD;
1534 
1535 	case NETDEV_NOTIFY_PEERS:
1536 	case NETDEV_BONDING_FAILOVER:
1537 	case NETDEV_RESEND_IGMP:
1538 		/* Propagate to all vlans */
1539 		list_for_each_entry(vlan, &port->vlans, list)
1540 			call_netdevice_notifiers(event, vlan->dev);
1541 	}
1542 	return NOTIFY_DONE;
1543 }
1544 
1545 static struct notifier_block macvlan_notifier_block __read_mostly = {
1546 	.notifier_call	= macvlan_device_event,
1547 };
1548 
1549 static int __init macvlan_init_module(void)
1550 {
1551 	int err;
1552 
1553 	register_netdevice_notifier(&macvlan_notifier_block);
1554 
1555 	err = macvlan_link_register(&macvlan_link_ops);
1556 	if (err < 0)
1557 		goto err1;
1558 	return 0;
1559 err1:
1560 	unregister_netdevice_notifier(&macvlan_notifier_block);
1561 	return err;
1562 }
1563 
1564 static void __exit macvlan_cleanup_module(void)
1565 {
1566 	rtnl_link_unregister(&macvlan_link_ops);
1567 	unregister_netdevice_notifier(&macvlan_notifier_block);
1568 }
1569 
1570 module_init(macvlan_init_module);
1571 module_exit(macvlan_cleanup_module);
1572 
1573 MODULE_LICENSE("GPL");
1574 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1575 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1576 MODULE_ALIAS_RTNL_LINK("macvlan");
1577