xref: /openbmc/linux/drivers/net/macvlan.c (revision c4c3c32d)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Copyright (c) 2007 Patrick McHardy <kaber@trash.net>
4  *
5  * The code this is based on carried the following copyright notice:
6  * ---
7  * (C) Copyright 2001-2006
8  * Alex Zeffertt, Cambridge Broadband Ltd, ajz@cambridgebroadband.com
9  * Re-worked by Ben Greear <greearb@candelatech.com>
10  * ---
11  */
12 #include <linux/kernel.h>
13 #include <linux/types.h>
14 #include <linux/module.h>
15 #include <linux/init.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/string.h>
19 #include <linux/rculist.h>
20 #include <linux/notifier.h>
21 #include <linux/netdevice.h>
22 #include <linux/etherdevice.h>
23 #include <linux/net_tstamp.h>
24 #include <linux/ethtool.h>
25 #include <linux/if_arp.h>
26 #include <linux/if_vlan.h>
27 #include <linux/if_link.h>
28 #include <linux/if_macvlan.h>
29 #include <linux/hash.h>
30 #include <linux/workqueue.h>
31 #include <net/rtnetlink.h>
32 #include <net/xfrm.h>
33 #include <linux/netpoll.h>
34 #include <linux/phy.h>
35 
36 #define MACVLAN_HASH_BITS	8
37 #define MACVLAN_HASH_SIZE	(1<<MACVLAN_HASH_BITS)
38 #define MACVLAN_DEFAULT_BC_QUEUE_LEN	1000
39 
40 #define MACVLAN_F_PASSTHRU	1
41 #define MACVLAN_F_ADDRCHANGE	2
42 
43 struct macvlan_port {
44 	struct net_device	*dev;
45 	struct hlist_head	vlan_hash[MACVLAN_HASH_SIZE];
46 	struct list_head	vlans;
47 	struct sk_buff_head	bc_queue;
48 	struct work_struct	bc_work;
49 	u32			bc_queue_len_used;
50 	int			bc_cutoff;
51 	u32			flags;
52 	int			count;
53 	struct hlist_head	vlan_source_hash[MACVLAN_HASH_SIZE];
54 	DECLARE_BITMAP(bc_filter, MACVLAN_MC_FILTER_SZ);
55 	DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
56 	unsigned char           perm_addr[ETH_ALEN];
57 };
58 
59 struct macvlan_source_entry {
60 	struct hlist_node	hlist;
61 	struct macvlan_dev	*vlan;
62 	unsigned char		addr[6+2] __aligned(sizeof(u16));
63 	struct rcu_head		rcu;
64 };
65 
66 struct macvlan_skb_cb {
67 	const struct macvlan_dev *src;
68 };
69 
70 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
71 
72 static void macvlan_port_destroy(struct net_device *dev);
73 static void update_port_bc_queue_len(struct macvlan_port *port);
74 
75 static inline bool macvlan_passthru(const struct macvlan_port *port)
76 {
77 	return port->flags & MACVLAN_F_PASSTHRU;
78 }
79 
80 static inline void macvlan_set_passthru(struct macvlan_port *port)
81 {
82 	port->flags |= MACVLAN_F_PASSTHRU;
83 }
84 
85 static inline bool macvlan_addr_change(const struct macvlan_port *port)
86 {
87 	return port->flags & MACVLAN_F_ADDRCHANGE;
88 }
89 
90 static inline void macvlan_set_addr_change(struct macvlan_port *port)
91 {
92 	port->flags |= MACVLAN_F_ADDRCHANGE;
93 }
94 
95 static inline void macvlan_clear_addr_change(struct macvlan_port *port)
96 {
97 	port->flags &= ~MACVLAN_F_ADDRCHANGE;
98 }
99 
100 /* Hash Ethernet address */
101 static u32 macvlan_eth_hash(const unsigned char *addr)
102 {
103 	u64 value = get_unaligned((u64 *)addr);
104 
105 	/* only want 6 bytes */
106 #ifdef __BIG_ENDIAN
107 	value >>= 16;
108 #else
109 	value <<= 16;
110 #endif
111 	return hash_64(value, MACVLAN_HASH_BITS);
112 }
113 
114 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
115 {
116 	return rcu_dereference(dev->rx_handler_data);
117 }
118 
119 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
120 {
121 	return rtnl_dereference(dev->rx_handler_data);
122 }
123 
124 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
125 					       const unsigned char *addr)
126 {
127 	struct macvlan_dev *vlan;
128 	u32 idx = macvlan_eth_hash(addr);
129 
130 	hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist,
131 				 lockdep_rtnl_is_held()) {
132 		if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
133 			return vlan;
134 	}
135 	return NULL;
136 }
137 
138 static struct macvlan_source_entry *macvlan_hash_lookup_source(
139 	const struct macvlan_dev *vlan,
140 	const unsigned char *addr)
141 {
142 	struct macvlan_source_entry *entry;
143 	u32 idx = macvlan_eth_hash(addr);
144 	struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
145 
146 	hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
147 		if (ether_addr_equal_64bits(entry->addr, addr) &&
148 		    entry->vlan == vlan)
149 			return entry;
150 	}
151 	return NULL;
152 }
153 
154 static int macvlan_hash_add_source(struct macvlan_dev *vlan,
155 				   const unsigned char *addr)
156 {
157 	struct macvlan_port *port = vlan->port;
158 	struct macvlan_source_entry *entry;
159 	struct hlist_head *h;
160 
161 	entry = macvlan_hash_lookup_source(vlan, addr);
162 	if (entry)
163 		return 0;
164 
165 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
166 	if (!entry)
167 		return -ENOMEM;
168 
169 	ether_addr_copy(entry->addr, addr);
170 	entry->vlan = vlan;
171 	h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
172 	hlist_add_head_rcu(&entry->hlist, h);
173 	vlan->macaddr_count++;
174 
175 	return 0;
176 }
177 
178 static void macvlan_hash_add(struct macvlan_dev *vlan)
179 {
180 	struct macvlan_port *port = vlan->port;
181 	const unsigned char *addr = vlan->dev->dev_addr;
182 	u32 idx = macvlan_eth_hash(addr);
183 
184 	hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
185 }
186 
187 static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
188 {
189 	hlist_del_rcu(&entry->hlist);
190 	kfree_rcu(entry, rcu);
191 }
192 
193 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
194 {
195 	hlist_del_rcu(&vlan->hlist);
196 	if (sync)
197 		synchronize_rcu();
198 }
199 
200 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
201 					const unsigned char *addr)
202 {
203 	macvlan_hash_del(vlan, true);
204 	/* Now that we are unhashed it is safe to change the device
205 	 * address without confusing packet delivery.
206 	 */
207 	eth_hw_addr_set(vlan->dev, addr);
208 	macvlan_hash_add(vlan);
209 }
210 
211 static bool macvlan_addr_busy(const struct macvlan_port *port,
212 			      const unsigned char *addr)
213 {
214 	/* Test to see if the specified address is
215 	 * currently in use by the underlying device or
216 	 * another macvlan.
217 	 */
218 	if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
219 	    ether_addr_equal_64bits(port->dev->dev_addr, addr))
220 		return true;
221 
222 	if (macvlan_hash_lookup(port, addr))
223 		return true;
224 
225 	return false;
226 }
227 
228 
229 static int macvlan_broadcast_one(struct sk_buff *skb,
230 				 const struct macvlan_dev *vlan,
231 				 const struct ethhdr *eth, bool local)
232 {
233 	struct net_device *dev = vlan->dev;
234 
235 	if (local)
236 		return __dev_forward_skb(dev, skb);
237 
238 	skb->dev = dev;
239 	if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
240 		skb->pkt_type = PACKET_BROADCAST;
241 	else
242 		skb->pkt_type = PACKET_MULTICAST;
243 
244 	return 0;
245 }
246 
247 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
248 {
249 	return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
250 }
251 
252 
253 static unsigned int mc_hash(const struct macvlan_dev *vlan,
254 			    const unsigned char *addr)
255 {
256 	u32 val = __get_unaligned_cpu32(addr + 2);
257 
258 	val ^= macvlan_hash_mix(vlan);
259 	return hash_32(val, MACVLAN_MC_FILTER_BITS);
260 }
261 
262 static void macvlan_broadcast(struct sk_buff *skb,
263 			      const struct macvlan_port *port,
264 			      struct net_device *src,
265 			      enum macvlan_mode mode)
266 {
267 	const struct ethhdr *eth = eth_hdr(skb);
268 	const struct macvlan_dev *vlan;
269 	struct sk_buff *nskb;
270 	unsigned int i;
271 	int err;
272 	unsigned int hash;
273 
274 	if (skb->protocol == htons(ETH_P_PAUSE))
275 		return;
276 
277 	hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) {
278 		if (vlan->dev == src || !(vlan->mode & mode))
279 			continue;
280 
281 		hash = mc_hash(vlan, eth->h_dest);
282 		if (!test_bit(hash, vlan->mc_filter))
283 			continue;
284 
285 		err = NET_RX_DROP;
286 		nskb = skb_clone(skb, GFP_ATOMIC);
287 		if (likely(nskb))
288 			err = macvlan_broadcast_one(nskb, vlan, eth,
289 					mode == MACVLAN_MODE_BRIDGE) ?:
290 			      netif_rx(nskb);
291 		macvlan_count_rx(vlan, skb->len + ETH_HLEN,
292 				 err == NET_RX_SUCCESS, true);
293 	}
294 }
295 
296 static void macvlan_multicast_rx(const struct macvlan_port *port,
297 				 const struct macvlan_dev *src,
298 				 struct sk_buff *skb)
299 {
300 	if (!src)
301 		/* frame comes from an external address */
302 		macvlan_broadcast(skb, port, NULL,
303 				  MACVLAN_MODE_PRIVATE |
304 				  MACVLAN_MODE_VEPA    |
305 				  MACVLAN_MODE_PASSTHRU|
306 				  MACVLAN_MODE_BRIDGE);
307 	else if (src->mode == MACVLAN_MODE_VEPA)
308 		/* flood to everyone except source */
309 		macvlan_broadcast(skb, port, src->dev,
310 				  MACVLAN_MODE_VEPA |
311 				  MACVLAN_MODE_BRIDGE);
312 	else
313 		/*
314 		 * flood only to VEPA ports, bridge ports
315 		 * already saw the frame on the way out.
316 		 */
317 		macvlan_broadcast(skb, port, src->dev,
318 				  MACVLAN_MODE_VEPA);
319 }
320 
321 static void macvlan_process_broadcast(struct work_struct *w)
322 {
323 	struct macvlan_port *port = container_of(w, struct macvlan_port,
324 						 bc_work);
325 	struct sk_buff *skb;
326 	struct sk_buff_head list;
327 
328 	__skb_queue_head_init(&list);
329 
330 	spin_lock_bh(&port->bc_queue.lock);
331 	skb_queue_splice_tail_init(&port->bc_queue, &list);
332 	spin_unlock_bh(&port->bc_queue.lock);
333 
334 	while ((skb = __skb_dequeue(&list))) {
335 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
336 
337 		rcu_read_lock();
338 		macvlan_multicast_rx(port, src, skb);
339 		rcu_read_unlock();
340 
341 		if (src)
342 			dev_put(src->dev);
343 		consume_skb(skb);
344 
345 		cond_resched();
346 	}
347 }
348 
349 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
350 				      const struct macvlan_dev *src,
351 				      struct sk_buff *skb)
352 {
353 	struct sk_buff *nskb;
354 	int err = -ENOMEM;
355 
356 	nskb = skb_clone(skb, GFP_ATOMIC);
357 	if (!nskb)
358 		goto err;
359 
360 	MACVLAN_SKB_CB(nskb)->src = src;
361 
362 	spin_lock(&port->bc_queue.lock);
363 	if (skb_queue_len(&port->bc_queue) < port->bc_queue_len_used) {
364 		if (src)
365 			dev_hold(src->dev);
366 		__skb_queue_tail(&port->bc_queue, nskb);
367 		err = 0;
368 	}
369 	spin_unlock(&port->bc_queue.lock);
370 
371 	queue_work(system_unbound_wq, &port->bc_work);
372 
373 	if (err)
374 		goto free_nskb;
375 
376 	return;
377 
378 free_nskb:
379 	kfree_skb(nskb);
380 err:
381 	dev_core_stats_rx_dropped_inc(skb->dev);
382 }
383 
384 static void macvlan_flush_sources(struct macvlan_port *port,
385 				  struct macvlan_dev *vlan)
386 {
387 	struct macvlan_source_entry *entry;
388 	struct hlist_node *next;
389 	int i;
390 
391 	hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist)
392 		if (entry->vlan == vlan)
393 			macvlan_hash_del_source(entry);
394 
395 	vlan->macaddr_count = 0;
396 }
397 
398 static void macvlan_forward_source_one(struct sk_buff *skb,
399 				       struct macvlan_dev *vlan)
400 {
401 	struct sk_buff *nskb;
402 	struct net_device *dev;
403 	int len;
404 	int ret;
405 
406 	dev = vlan->dev;
407 	if (unlikely(!(dev->flags & IFF_UP)))
408 		return;
409 
410 	nskb = skb_clone(skb, GFP_ATOMIC);
411 	if (!nskb)
412 		return;
413 
414 	len = nskb->len + ETH_HLEN;
415 	nskb->dev = dev;
416 
417 	if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
418 		nskb->pkt_type = PACKET_HOST;
419 
420 	ret = __netif_rx(nskb);
421 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
422 }
423 
424 static bool macvlan_forward_source(struct sk_buff *skb,
425 				   struct macvlan_port *port,
426 				   const unsigned char *addr)
427 {
428 	struct macvlan_source_entry *entry;
429 	u32 idx = macvlan_eth_hash(addr);
430 	struct hlist_head *h = &port->vlan_source_hash[idx];
431 	bool consume = false;
432 
433 	hlist_for_each_entry_rcu(entry, h, hlist) {
434 		if (ether_addr_equal_64bits(entry->addr, addr)) {
435 			if (entry->vlan->flags & MACVLAN_FLAG_NODST)
436 				consume = true;
437 			macvlan_forward_source_one(skb, entry->vlan);
438 		}
439 	}
440 
441 	return consume;
442 }
443 
444 /* called under rcu_read_lock() from netif_receive_skb */
445 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
446 {
447 	struct macvlan_port *port;
448 	struct sk_buff *skb = *pskb;
449 	const struct ethhdr *eth = eth_hdr(skb);
450 	const struct macvlan_dev *vlan;
451 	const struct macvlan_dev *src;
452 	struct net_device *dev;
453 	unsigned int len = 0;
454 	int ret;
455 	rx_handler_result_t handle_res;
456 
457 	/* Packets from dev_loopback_xmit() do not have L2 header, bail out */
458 	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
459 		return RX_HANDLER_PASS;
460 
461 	port = macvlan_port_get_rcu(skb->dev);
462 	if (is_multicast_ether_addr(eth->h_dest)) {
463 		unsigned int hash;
464 
465 		skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
466 		if (!skb)
467 			return RX_HANDLER_CONSUMED;
468 		*pskb = skb;
469 		eth = eth_hdr(skb);
470 		if (macvlan_forward_source(skb, port, eth->h_source)) {
471 			kfree_skb(skb);
472 			return RX_HANDLER_CONSUMED;
473 		}
474 		src = macvlan_hash_lookup(port, eth->h_source);
475 		if (src && src->mode != MACVLAN_MODE_VEPA &&
476 		    src->mode != MACVLAN_MODE_BRIDGE) {
477 			/* forward to original port. */
478 			vlan = src;
479 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
480 			      __netif_rx(skb);
481 			handle_res = RX_HANDLER_CONSUMED;
482 			goto out;
483 		}
484 
485 		hash = mc_hash(NULL, eth->h_dest);
486 		if (test_bit(hash, port->bc_filter))
487 			macvlan_broadcast_enqueue(port, src, skb);
488 		else if (test_bit(hash, port->mc_filter))
489 			macvlan_multicast_rx(port, src, skb);
490 
491 		return RX_HANDLER_PASS;
492 	}
493 
494 	if (macvlan_forward_source(skb, port, eth->h_source)) {
495 		kfree_skb(skb);
496 		return RX_HANDLER_CONSUMED;
497 	}
498 	if (macvlan_passthru(port))
499 		vlan = list_first_or_null_rcu(&port->vlans,
500 					      struct macvlan_dev, list);
501 	else
502 		vlan = macvlan_hash_lookup(port, eth->h_dest);
503 	if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
504 		return RX_HANDLER_PASS;
505 
506 	dev = vlan->dev;
507 	if (unlikely(!(dev->flags & IFF_UP))) {
508 		kfree_skb(skb);
509 		return RX_HANDLER_CONSUMED;
510 	}
511 	len = skb->len + ETH_HLEN;
512 	skb = skb_share_check(skb, GFP_ATOMIC);
513 	if (!skb) {
514 		ret = NET_RX_DROP;
515 		handle_res = RX_HANDLER_CONSUMED;
516 		goto out;
517 	}
518 
519 	*pskb = skb;
520 	skb->dev = dev;
521 	skb->pkt_type = PACKET_HOST;
522 
523 	ret = NET_RX_SUCCESS;
524 	handle_res = RX_HANDLER_ANOTHER;
525 out:
526 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
527 	return handle_res;
528 }
529 
530 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
531 {
532 	const struct macvlan_dev *vlan = netdev_priv(dev);
533 	const struct macvlan_port *port = vlan->port;
534 	const struct macvlan_dev *dest;
535 
536 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
537 		const struct ethhdr *eth = skb_eth_hdr(skb);
538 
539 		/* send to other bridge ports directly */
540 		if (is_multicast_ether_addr(eth->h_dest)) {
541 			skb_reset_mac_header(skb);
542 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
543 			goto xmit_world;
544 		}
545 
546 		dest = macvlan_hash_lookup(port, eth->h_dest);
547 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
548 			/* send to lowerdev first for its network taps */
549 			dev_forward_skb(vlan->lowerdev, skb);
550 
551 			return NET_XMIT_SUCCESS;
552 		}
553 	}
554 xmit_world:
555 	skb->dev = vlan->lowerdev;
556 	return dev_queue_xmit_accel(skb,
557 				    netdev_get_sb_channel(dev) ? dev : NULL);
558 }
559 
560 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
561 {
562 #ifdef CONFIG_NET_POLL_CONTROLLER
563 	return netpoll_send_skb(vlan->netpoll, skb);
564 #else
565 	BUG();
566 	return NETDEV_TX_OK;
567 #endif
568 }
569 
570 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
571 				      struct net_device *dev)
572 {
573 	struct macvlan_dev *vlan = netdev_priv(dev);
574 	unsigned int len = skb->len;
575 	int ret;
576 
577 	if (unlikely(netpoll_tx_running(dev)))
578 		return macvlan_netpoll_send_skb(vlan, skb);
579 
580 	ret = macvlan_queue_xmit(skb, dev);
581 
582 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
583 		struct vlan_pcpu_stats *pcpu_stats;
584 
585 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
586 		u64_stats_update_begin(&pcpu_stats->syncp);
587 		u64_stats_inc(&pcpu_stats->tx_packets);
588 		u64_stats_add(&pcpu_stats->tx_bytes, len);
589 		u64_stats_update_end(&pcpu_stats->syncp);
590 	} else {
591 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
592 	}
593 	return ret;
594 }
595 
596 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
597 			       unsigned short type, const void *daddr,
598 			       const void *saddr, unsigned len)
599 {
600 	const struct macvlan_dev *vlan = netdev_priv(dev);
601 	struct net_device *lowerdev = vlan->lowerdev;
602 
603 	return dev_hard_header(skb, lowerdev, type, daddr,
604 			       saddr ? : dev->dev_addr, len);
605 }
606 
607 static const struct header_ops macvlan_hard_header_ops = {
608 	.create  	= macvlan_hard_header,
609 	.parse		= eth_header_parse,
610 	.cache		= eth_header_cache,
611 	.cache_update	= eth_header_cache_update,
612 };
613 
614 static int macvlan_open(struct net_device *dev)
615 {
616 	struct macvlan_dev *vlan = netdev_priv(dev);
617 	struct net_device *lowerdev = vlan->lowerdev;
618 	int err;
619 
620 	if (macvlan_passthru(vlan->port)) {
621 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
622 			err = dev_set_promiscuity(lowerdev, 1);
623 			if (err < 0)
624 				goto out;
625 		}
626 		goto hash_add;
627 	}
628 
629 	err = -EADDRINUSE;
630 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
631 		goto out;
632 
633 	/* Attempt to populate accel_priv which is used to offload the L2
634 	 * forwarding requests for unicast packets.
635 	 */
636 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
637 		vlan->accel_priv =
638 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
639 
640 	/* If earlier attempt to offload failed, or accel_priv is not
641 	 * populated we must add the unicast address to the lower device.
642 	 */
643 	if (IS_ERR_OR_NULL(vlan->accel_priv)) {
644 		vlan->accel_priv = NULL;
645 		err = dev_uc_add(lowerdev, dev->dev_addr);
646 		if (err < 0)
647 			goto out;
648 	}
649 
650 	if (dev->flags & IFF_ALLMULTI) {
651 		err = dev_set_allmulti(lowerdev, 1);
652 		if (err < 0)
653 			goto del_unicast;
654 	}
655 
656 	if (dev->flags & IFF_PROMISC) {
657 		err = dev_set_promiscuity(lowerdev, 1);
658 		if (err < 0)
659 			goto clear_multi;
660 	}
661 
662 hash_add:
663 	macvlan_hash_add(vlan);
664 	return 0;
665 
666 clear_multi:
667 	if (dev->flags & IFF_ALLMULTI)
668 		dev_set_allmulti(lowerdev, -1);
669 del_unicast:
670 	if (vlan->accel_priv) {
671 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
672 							   vlan->accel_priv);
673 		vlan->accel_priv = NULL;
674 	} else {
675 		dev_uc_del(lowerdev, dev->dev_addr);
676 	}
677 out:
678 	return err;
679 }
680 
681 static int macvlan_stop(struct net_device *dev)
682 {
683 	struct macvlan_dev *vlan = netdev_priv(dev);
684 	struct net_device *lowerdev = vlan->lowerdev;
685 
686 	if (vlan->accel_priv) {
687 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
688 							   vlan->accel_priv);
689 		vlan->accel_priv = NULL;
690 	}
691 
692 	dev_uc_unsync(lowerdev, dev);
693 	dev_mc_unsync(lowerdev, dev);
694 
695 	if (macvlan_passthru(vlan->port)) {
696 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
697 			dev_set_promiscuity(lowerdev, -1);
698 		goto hash_del;
699 	}
700 
701 	if (dev->flags & IFF_ALLMULTI)
702 		dev_set_allmulti(lowerdev, -1);
703 
704 	if (dev->flags & IFF_PROMISC)
705 		dev_set_promiscuity(lowerdev, -1);
706 
707 	dev_uc_del(lowerdev, dev->dev_addr);
708 
709 hash_del:
710 	macvlan_hash_del(vlan, !dev->dismantle);
711 	return 0;
712 }
713 
714 static int macvlan_sync_address(struct net_device *dev,
715 				const unsigned char *addr)
716 {
717 	struct macvlan_dev *vlan = netdev_priv(dev);
718 	struct net_device *lowerdev = vlan->lowerdev;
719 	struct macvlan_port *port = vlan->port;
720 	int err;
721 
722 	if (!(dev->flags & IFF_UP)) {
723 		/* Just copy in the new address */
724 		eth_hw_addr_set(dev, addr);
725 	} else {
726 		/* Rehash and update the device filters */
727 		if (macvlan_addr_busy(vlan->port, addr))
728 			return -EADDRINUSE;
729 
730 		if (!macvlan_passthru(port)) {
731 			err = dev_uc_add(lowerdev, addr);
732 			if (err)
733 				return err;
734 
735 			dev_uc_del(lowerdev, dev->dev_addr);
736 		}
737 
738 		macvlan_hash_change_addr(vlan, addr);
739 	}
740 	if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
741 		/* Since addr_change isn't set, we are here due to lower
742 		 * device change.  Save the lower-dev address so we can
743 		 * restore it later.
744 		 */
745 		ether_addr_copy(vlan->port->perm_addr,
746 				lowerdev->dev_addr);
747 	}
748 	macvlan_clear_addr_change(port);
749 	return 0;
750 }
751 
752 static int macvlan_set_mac_address(struct net_device *dev, void *p)
753 {
754 	struct macvlan_dev *vlan = netdev_priv(dev);
755 	struct sockaddr *addr = p;
756 
757 	if (!is_valid_ether_addr(addr->sa_data))
758 		return -EADDRNOTAVAIL;
759 
760 	/* If the addresses are the same, this is a no-op */
761 	if (ether_addr_equal(dev->dev_addr, addr->sa_data))
762 		return 0;
763 
764 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
765 		macvlan_set_addr_change(vlan->port);
766 		return dev_set_mac_address(vlan->lowerdev, addr, NULL);
767 	}
768 
769 	if (macvlan_addr_busy(vlan->port, addr->sa_data))
770 		return -EADDRINUSE;
771 
772 	return macvlan_sync_address(dev, addr->sa_data);
773 }
774 
775 static void macvlan_change_rx_flags(struct net_device *dev, int change)
776 {
777 	struct macvlan_dev *vlan = netdev_priv(dev);
778 	struct net_device *lowerdev = vlan->lowerdev;
779 
780 	if (dev->flags & IFF_UP) {
781 		if (change & IFF_ALLMULTI)
782 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
783 		if (change & IFF_PROMISC)
784 			dev_set_promiscuity(lowerdev,
785 					    dev->flags & IFF_PROMISC ? 1 : -1);
786 
787 	}
788 }
789 
790 static void macvlan_compute_filter(unsigned long *mc_filter,
791 				   struct net_device *dev,
792 				   struct macvlan_dev *vlan, int cutoff)
793 {
794 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
795 		bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
796 	} else {
797 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
798 		struct netdev_hw_addr *ha;
799 
800 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
801 		netdev_for_each_mc_addr(ha, dev) {
802 			if (!vlan && ha->synced <= cutoff)
803 				continue;
804 
805 			__set_bit(mc_hash(vlan, ha->addr), filter);
806 		}
807 
808 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
809 
810 		bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
811 	}
812 }
813 
814 static void macvlan_recompute_bc_filter(struct macvlan_dev *vlan)
815 {
816 	if (vlan->port->bc_cutoff < 0) {
817 		bitmap_zero(vlan->port->bc_filter, MACVLAN_MC_FILTER_SZ);
818 		return;
819 	}
820 
821 	macvlan_compute_filter(vlan->port->bc_filter, vlan->lowerdev, NULL,
822 			       vlan->port->bc_cutoff);
823 }
824 
825 static void macvlan_set_mac_lists(struct net_device *dev)
826 {
827 	struct macvlan_dev *vlan = netdev_priv(dev);
828 
829 	macvlan_compute_filter(vlan->mc_filter, dev, vlan, 0);
830 
831 	dev_uc_sync(vlan->lowerdev, dev);
832 	dev_mc_sync(vlan->lowerdev, dev);
833 
834 	/* This is slightly inaccurate as we're including the subscription
835 	 * list of vlan->lowerdev too.
836 	 *
837 	 * Bug alert: This only works if everyone has the same broadcast
838 	 * address as lowerdev.  As soon as someone changes theirs this
839 	 * will break.
840 	 *
841 	 * However, this is already broken as when you change your broadcast
842 	 * address we don't get called.
843 	 *
844 	 * The solution is to maintain a list of broadcast addresses like
845 	 * we do for uc/mc, if you care.
846 	 */
847 	macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL,
848 			       0);
849 	macvlan_recompute_bc_filter(vlan);
850 }
851 
852 static void update_port_bc_cutoff(struct macvlan_dev *vlan, int cutoff)
853 {
854 	if (vlan->port->bc_cutoff == cutoff)
855 		return;
856 
857 	vlan->port->bc_cutoff = cutoff;
858 	macvlan_recompute_bc_filter(vlan);
859 }
860 
861 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
862 {
863 	struct macvlan_dev *vlan = netdev_priv(dev);
864 
865 	if (vlan->lowerdev->mtu < new_mtu)
866 		return -EINVAL;
867 	dev->mtu = new_mtu;
868 	return 0;
869 }
870 
871 static int macvlan_eth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
872 {
873 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
874 	const struct net_device_ops *ops = real_dev->netdev_ops;
875 	struct ifreq ifrr;
876 	int err = -EOPNOTSUPP;
877 
878 	strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
879 	ifrr.ifr_ifru = ifr->ifr_ifru;
880 
881 	switch (cmd) {
882 	case SIOCSHWTSTAMP:
883 		if (!net_eq(dev_net(dev), &init_net))
884 			break;
885 		fallthrough;
886 	case SIOCGHWTSTAMP:
887 		if (netif_device_present(real_dev) && ops->ndo_eth_ioctl)
888 			err = ops->ndo_eth_ioctl(real_dev, &ifrr, cmd);
889 		break;
890 	}
891 
892 	if (!err)
893 		ifr->ifr_ifru = ifrr.ifr_ifru;
894 
895 	return err;
896 }
897 
898 /*
899  * macvlan network devices have devices nesting below it and are a special
900  * "super class" of normal network devices; split their locks off into a
901  * separate class since they always nest.
902  */
903 static struct lock_class_key macvlan_netdev_addr_lock_key;
904 
905 #define ALWAYS_ON_OFFLOADS \
906 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
907 	 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
908 
909 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
910 
911 #define MACVLAN_FEATURES \
912 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
913 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
914 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
915 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
916 
917 #define MACVLAN_STATE_MASK \
918 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
919 
920 static void macvlan_set_lockdep_class(struct net_device *dev)
921 {
922 	netdev_lockdep_set_classes(dev);
923 	lockdep_set_class(&dev->addr_list_lock,
924 			  &macvlan_netdev_addr_lock_key);
925 }
926 
927 static int macvlan_init(struct net_device *dev)
928 {
929 	struct macvlan_dev *vlan = netdev_priv(dev);
930 	struct net_device *lowerdev = vlan->lowerdev;
931 	struct macvlan_port *port = vlan->port;
932 
933 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
934 				  (lowerdev->state & MACVLAN_STATE_MASK);
935 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
936 	dev->features		|= ALWAYS_ON_FEATURES;
937 	dev->hw_features	|= NETIF_F_LRO;
938 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
939 	dev->vlan_features	|= ALWAYS_ON_OFFLOADS;
940 	dev->hw_enc_features    |= dev->features;
941 	netif_inherit_tso_max(dev, lowerdev);
942 	dev->hard_header_len	= lowerdev->hard_header_len;
943 	macvlan_set_lockdep_class(dev);
944 
945 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
946 	if (!vlan->pcpu_stats)
947 		return -ENOMEM;
948 
949 	port->count += 1;
950 
951 	/* Get macvlan's reference to lowerdev */
952 	netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL);
953 
954 	return 0;
955 }
956 
957 static void macvlan_uninit(struct net_device *dev)
958 {
959 	struct macvlan_dev *vlan = netdev_priv(dev);
960 	struct macvlan_port *port = vlan->port;
961 
962 	free_percpu(vlan->pcpu_stats);
963 
964 	macvlan_flush_sources(port, vlan);
965 	port->count -= 1;
966 	if (!port->count)
967 		macvlan_port_destroy(port->dev);
968 }
969 
970 static void macvlan_dev_get_stats64(struct net_device *dev,
971 				    struct rtnl_link_stats64 *stats)
972 {
973 	struct macvlan_dev *vlan = netdev_priv(dev);
974 
975 	if (vlan->pcpu_stats) {
976 		struct vlan_pcpu_stats *p;
977 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
978 		u32 rx_errors = 0, tx_dropped = 0;
979 		unsigned int start;
980 		int i;
981 
982 		for_each_possible_cpu(i) {
983 			p = per_cpu_ptr(vlan->pcpu_stats, i);
984 			do {
985 				start = u64_stats_fetch_begin(&p->syncp);
986 				rx_packets	= u64_stats_read(&p->rx_packets);
987 				rx_bytes	= u64_stats_read(&p->rx_bytes);
988 				rx_multicast	= u64_stats_read(&p->rx_multicast);
989 				tx_packets	= u64_stats_read(&p->tx_packets);
990 				tx_bytes	= u64_stats_read(&p->tx_bytes);
991 			} while (u64_stats_fetch_retry(&p->syncp, start));
992 
993 			stats->rx_packets	+= rx_packets;
994 			stats->rx_bytes		+= rx_bytes;
995 			stats->multicast	+= rx_multicast;
996 			stats->tx_packets	+= tx_packets;
997 			stats->tx_bytes		+= tx_bytes;
998 			/* rx_errors & tx_dropped are u32, updated
999 			 * without syncp protection.
1000 			 */
1001 			rx_errors	+= READ_ONCE(p->rx_errors);
1002 			tx_dropped	+= READ_ONCE(p->tx_dropped);
1003 		}
1004 		stats->rx_errors	= rx_errors;
1005 		stats->rx_dropped	= rx_errors;
1006 		stats->tx_dropped	= tx_dropped;
1007 	}
1008 }
1009 
1010 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
1011 				   __be16 proto, u16 vid)
1012 {
1013 	struct macvlan_dev *vlan = netdev_priv(dev);
1014 	struct net_device *lowerdev = vlan->lowerdev;
1015 
1016 	return vlan_vid_add(lowerdev, proto, vid);
1017 }
1018 
1019 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
1020 				    __be16 proto, u16 vid)
1021 {
1022 	struct macvlan_dev *vlan = netdev_priv(dev);
1023 	struct net_device *lowerdev = vlan->lowerdev;
1024 
1025 	vlan_vid_del(lowerdev, proto, vid);
1026 	return 0;
1027 }
1028 
1029 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1030 			   struct net_device *dev,
1031 			   const unsigned char *addr, u16 vid,
1032 			   u16 flags,
1033 			   struct netlink_ext_ack *extack)
1034 {
1035 	struct macvlan_dev *vlan = netdev_priv(dev);
1036 	int err = -EINVAL;
1037 
1038 	/* Support unicast filter only on passthru devices.
1039 	 * Multicast filter should be allowed on all devices.
1040 	 */
1041 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1042 		return -EOPNOTSUPP;
1043 
1044 	if (flags & NLM_F_REPLACE)
1045 		return -EOPNOTSUPP;
1046 
1047 	if (is_unicast_ether_addr(addr))
1048 		err = dev_uc_add_excl(dev, addr);
1049 	else if (is_multicast_ether_addr(addr))
1050 		err = dev_mc_add_excl(dev, addr);
1051 
1052 	return err;
1053 }
1054 
1055 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1056 			   struct net_device *dev,
1057 			   const unsigned char *addr, u16 vid,
1058 			   struct netlink_ext_ack *extack)
1059 {
1060 	struct macvlan_dev *vlan = netdev_priv(dev);
1061 	int err = -EINVAL;
1062 
1063 	/* Support unicast filter only on passthru devices.
1064 	 * Multicast filter should be allowed on all devices.
1065 	 */
1066 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1067 		return -EOPNOTSUPP;
1068 
1069 	if (is_unicast_ether_addr(addr))
1070 		err = dev_uc_del(dev, addr);
1071 	else if (is_multicast_ether_addr(addr))
1072 		err = dev_mc_del(dev, addr);
1073 
1074 	return err;
1075 }
1076 
1077 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1078 					struct ethtool_drvinfo *drvinfo)
1079 {
1080 	strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1081 	strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1082 }
1083 
1084 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1085 					      struct ethtool_link_ksettings *cmd)
1086 {
1087 	const struct macvlan_dev *vlan = netdev_priv(dev);
1088 
1089 	return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1090 }
1091 
1092 static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1093 				       struct ethtool_ts_info *info)
1094 {
1095 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
1096 	const struct ethtool_ops *ops = real_dev->ethtool_ops;
1097 	struct phy_device *phydev = real_dev->phydev;
1098 
1099 	if (phy_has_tsinfo(phydev)) {
1100 		return phy_ts_info(phydev, info);
1101 	} else if (ops->get_ts_info) {
1102 		return ops->get_ts_info(real_dev, info);
1103 	} else {
1104 		info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1105 			SOF_TIMESTAMPING_SOFTWARE;
1106 		info->phc_index = -1;
1107 	}
1108 
1109 	return 0;
1110 }
1111 
1112 static netdev_features_t macvlan_fix_features(struct net_device *dev,
1113 					      netdev_features_t features)
1114 {
1115 	struct macvlan_dev *vlan = netdev_priv(dev);
1116 	netdev_features_t lowerdev_features = vlan->lowerdev->features;
1117 	netdev_features_t mask;
1118 
1119 	features |= NETIF_F_ALL_FOR_ALL;
1120 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
1121 	mask = features;
1122 
1123 	lowerdev_features &= (features | ~NETIF_F_LRO);
1124 	features = netdev_increment_features(lowerdev_features, features, mask);
1125 	features |= ALWAYS_ON_FEATURES;
1126 	features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1127 
1128 	return features;
1129 }
1130 
1131 #ifdef CONFIG_NET_POLL_CONTROLLER
1132 static void macvlan_dev_poll_controller(struct net_device *dev)
1133 {
1134 	return;
1135 }
1136 
1137 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1138 {
1139 	struct macvlan_dev *vlan = netdev_priv(dev);
1140 	struct net_device *real_dev = vlan->lowerdev;
1141 	struct netpoll *netpoll;
1142 	int err;
1143 
1144 	netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1145 	err = -ENOMEM;
1146 	if (!netpoll)
1147 		goto out;
1148 
1149 	err = __netpoll_setup(netpoll, real_dev);
1150 	if (err) {
1151 		kfree(netpoll);
1152 		goto out;
1153 	}
1154 
1155 	vlan->netpoll = netpoll;
1156 
1157 out:
1158 	return err;
1159 }
1160 
1161 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1162 {
1163 	struct macvlan_dev *vlan = netdev_priv(dev);
1164 	struct netpoll *netpoll = vlan->netpoll;
1165 
1166 	if (!netpoll)
1167 		return;
1168 
1169 	vlan->netpoll = NULL;
1170 
1171 	__netpoll_free(netpoll);
1172 }
1173 #endif	/* CONFIG_NET_POLL_CONTROLLER */
1174 
1175 static int macvlan_dev_get_iflink(const struct net_device *dev)
1176 {
1177 	struct macvlan_dev *vlan = netdev_priv(dev);
1178 
1179 	return vlan->lowerdev->ifindex;
1180 }
1181 
1182 static const struct ethtool_ops macvlan_ethtool_ops = {
1183 	.get_link		= ethtool_op_get_link,
1184 	.get_link_ksettings	= macvlan_ethtool_get_link_ksettings,
1185 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
1186 	.get_ts_info		= macvlan_ethtool_get_ts_info,
1187 };
1188 
1189 static const struct net_device_ops macvlan_netdev_ops = {
1190 	.ndo_init		= macvlan_init,
1191 	.ndo_uninit		= macvlan_uninit,
1192 	.ndo_open		= macvlan_open,
1193 	.ndo_stop		= macvlan_stop,
1194 	.ndo_start_xmit		= macvlan_start_xmit,
1195 	.ndo_change_mtu		= macvlan_change_mtu,
1196 	.ndo_eth_ioctl		= macvlan_eth_ioctl,
1197 	.ndo_fix_features	= macvlan_fix_features,
1198 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1199 	.ndo_set_mac_address	= macvlan_set_mac_address,
1200 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1201 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1202 	.ndo_validate_addr	= eth_validate_addr,
1203 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1204 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1205 	.ndo_fdb_add		= macvlan_fdb_add,
1206 	.ndo_fdb_del		= macvlan_fdb_del,
1207 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1208 #ifdef CONFIG_NET_POLL_CONTROLLER
1209 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1210 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1211 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1212 #endif
1213 	.ndo_get_iflink		= macvlan_dev_get_iflink,
1214 	.ndo_features_check	= passthru_features_check,
1215 };
1216 
1217 static void macvlan_dev_free(struct net_device *dev)
1218 {
1219 	struct macvlan_dev *vlan = netdev_priv(dev);
1220 
1221 	/* Get rid of the macvlan's reference to lowerdev */
1222 	netdev_put(vlan->lowerdev, &vlan->dev_tracker);
1223 }
1224 
1225 void macvlan_common_setup(struct net_device *dev)
1226 {
1227 	ether_setup(dev);
1228 
1229 	/* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1230 	dev->max_mtu		= ETH_MAX_MTU;
1231 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1232 	netif_keep_dst(dev);
1233 	dev->priv_flags	       |= IFF_UNICAST_FLT | IFF_CHANGE_PROTO_DOWN;
1234 	dev->netdev_ops		= &macvlan_netdev_ops;
1235 	dev->needs_free_netdev	= true;
1236 	dev->priv_destructor	= macvlan_dev_free;
1237 	dev->header_ops		= &macvlan_hard_header_ops;
1238 	dev->ethtool_ops	= &macvlan_ethtool_ops;
1239 }
1240 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1241 
1242 static void macvlan_setup(struct net_device *dev)
1243 {
1244 	macvlan_common_setup(dev);
1245 	dev->priv_flags |= IFF_NO_QUEUE;
1246 }
1247 
1248 static int macvlan_port_create(struct net_device *dev)
1249 {
1250 	struct macvlan_port *port;
1251 	unsigned int i;
1252 	int err;
1253 
1254 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1255 		return -EINVAL;
1256 
1257 	if (netdev_is_rx_handler_busy(dev))
1258 		return -EBUSY;
1259 
1260 	port = kzalloc(sizeof(*port), GFP_KERNEL);
1261 	if (port == NULL)
1262 		return -ENOMEM;
1263 
1264 	port->dev = dev;
1265 	ether_addr_copy(port->perm_addr, dev->dev_addr);
1266 	INIT_LIST_HEAD(&port->vlans);
1267 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1268 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
1269 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1270 		INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1271 
1272 	port->bc_queue_len_used = 0;
1273 	port->bc_cutoff = 1;
1274 	skb_queue_head_init(&port->bc_queue);
1275 	INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1276 
1277 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1278 	if (err)
1279 		kfree(port);
1280 	else
1281 		dev->priv_flags |= IFF_MACVLAN_PORT;
1282 	return err;
1283 }
1284 
1285 static void macvlan_port_destroy(struct net_device *dev)
1286 {
1287 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1288 	struct sk_buff *skb;
1289 
1290 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
1291 	netdev_rx_handler_unregister(dev);
1292 
1293 	/* After this point, no packet can schedule bc_work anymore,
1294 	 * but we need to cancel it and purge left skbs if any.
1295 	 */
1296 	cancel_work_sync(&port->bc_work);
1297 
1298 	while ((skb = __skb_dequeue(&port->bc_queue))) {
1299 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1300 
1301 		if (src)
1302 			dev_put(src->dev);
1303 
1304 		kfree_skb(skb);
1305 	}
1306 
1307 	/* If the lower device address has been changed by passthru
1308 	 * macvlan, put it back.
1309 	 */
1310 	if (macvlan_passthru(port) &&
1311 	    !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1312 		struct sockaddr sa;
1313 
1314 		sa.sa_family = port->dev->type;
1315 		memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1316 		dev_set_mac_address(port->dev, &sa, NULL);
1317 	}
1318 
1319 	kfree(port);
1320 }
1321 
1322 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1323 			    struct netlink_ext_ack *extack)
1324 {
1325 	struct nlattr *nla, *head;
1326 	int rem, len;
1327 
1328 	if (tb[IFLA_ADDRESS]) {
1329 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1330 			return -EINVAL;
1331 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1332 			return -EADDRNOTAVAIL;
1333 	}
1334 
1335 	if (!data)
1336 		return 0;
1337 
1338 	if (data[IFLA_MACVLAN_FLAGS] &&
1339 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~(MACVLAN_FLAG_NOPROMISC |
1340 						      MACVLAN_FLAG_NODST))
1341 		return -EINVAL;
1342 
1343 	if (data[IFLA_MACVLAN_MODE]) {
1344 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1345 		case MACVLAN_MODE_PRIVATE:
1346 		case MACVLAN_MODE_VEPA:
1347 		case MACVLAN_MODE_BRIDGE:
1348 		case MACVLAN_MODE_PASSTHRU:
1349 		case MACVLAN_MODE_SOURCE:
1350 			break;
1351 		default:
1352 			return -EINVAL;
1353 		}
1354 	}
1355 
1356 	if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1357 		switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1358 		case MACVLAN_MACADDR_ADD:
1359 		case MACVLAN_MACADDR_DEL:
1360 		case MACVLAN_MACADDR_FLUSH:
1361 		case MACVLAN_MACADDR_SET:
1362 			break;
1363 		default:
1364 			return -EINVAL;
1365 		}
1366 	}
1367 
1368 	if (data[IFLA_MACVLAN_MACADDR]) {
1369 		if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1370 			return -EINVAL;
1371 
1372 		if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1373 			return -EADDRNOTAVAIL;
1374 	}
1375 
1376 	if (data[IFLA_MACVLAN_MACADDR_DATA]) {
1377 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1378 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1379 
1380 		nla_for_each_attr(nla, head, len, rem) {
1381 			if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1382 			    nla_len(nla) != ETH_ALEN)
1383 				return -EINVAL;
1384 
1385 			if (!is_valid_ether_addr(nla_data(nla)))
1386 				return -EADDRNOTAVAIL;
1387 		}
1388 	}
1389 
1390 	if (data[IFLA_MACVLAN_MACADDR_COUNT])
1391 		return -EINVAL;
1392 
1393 	return 0;
1394 }
1395 
1396 /*
1397  * reconfigure list of remote source mac address
1398  * (only for macvlan devices in source mode)
1399  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1400  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1401  */
1402 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1403 				      struct nlattr *data[])
1404 {
1405 	char *addr = NULL;
1406 	int ret, rem, len;
1407 	struct nlattr *nla, *head;
1408 	struct macvlan_source_entry *entry;
1409 
1410 	if (data[IFLA_MACVLAN_MACADDR])
1411 		addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1412 
1413 	if (mode == MACVLAN_MACADDR_ADD) {
1414 		if (!addr)
1415 			return -EINVAL;
1416 
1417 		return macvlan_hash_add_source(vlan, addr);
1418 
1419 	} else if (mode == MACVLAN_MACADDR_DEL) {
1420 		if (!addr)
1421 			return -EINVAL;
1422 
1423 		entry = macvlan_hash_lookup_source(vlan, addr);
1424 		if (entry) {
1425 			macvlan_hash_del_source(entry);
1426 			vlan->macaddr_count--;
1427 		}
1428 	} else if (mode == MACVLAN_MACADDR_FLUSH) {
1429 		macvlan_flush_sources(vlan->port, vlan);
1430 	} else if (mode == MACVLAN_MACADDR_SET) {
1431 		macvlan_flush_sources(vlan->port, vlan);
1432 
1433 		if (addr) {
1434 			ret = macvlan_hash_add_source(vlan, addr);
1435 			if (ret)
1436 				return ret;
1437 		}
1438 
1439 		if (!data[IFLA_MACVLAN_MACADDR_DATA])
1440 			return 0;
1441 
1442 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1443 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1444 
1445 		nla_for_each_attr(nla, head, len, rem) {
1446 			addr = nla_data(nla);
1447 			ret = macvlan_hash_add_source(vlan, addr);
1448 			if (ret)
1449 				return ret;
1450 		}
1451 	} else {
1452 		return -EINVAL;
1453 	}
1454 
1455 	return 0;
1456 }
1457 
1458 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1459 			   struct nlattr *tb[], struct nlattr *data[],
1460 			   struct netlink_ext_ack *extack)
1461 {
1462 	struct macvlan_dev *vlan = netdev_priv(dev);
1463 	struct macvlan_port *port;
1464 	struct net_device *lowerdev;
1465 	int err;
1466 	int macmode;
1467 	bool create = false;
1468 
1469 	if (!tb[IFLA_LINK])
1470 		return -EINVAL;
1471 
1472 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1473 	if (lowerdev == NULL)
1474 		return -ENODEV;
1475 
1476 	/* When creating macvlans or macvtaps on top of other macvlans - use
1477 	 * the real device as the lowerdev.
1478 	 */
1479 	if (netif_is_macvlan(lowerdev))
1480 		lowerdev = macvlan_dev_real_dev(lowerdev);
1481 
1482 	if (!tb[IFLA_MTU])
1483 		dev->mtu = lowerdev->mtu;
1484 	else if (dev->mtu > lowerdev->mtu)
1485 		return -EINVAL;
1486 
1487 	/* MTU range: 68 - lowerdev->max_mtu */
1488 	dev->min_mtu = ETH_MIN_MTU;
1489 	dev->max_mtu = lowerdev->max_mtu;
1490 
1491 	if (!tb[IFLA_ADDRESS])
1492 		eth_hw_addr_random(dev);
1493 
1494 	if (!netif_is_macvlan_port(lowerdev)) {
1495 		err = macvlan_port_create(lowerdev);
1496 		if (err < 0)
1497 			return err;
1498 		create = true;
1499 	}
1500 	port = macvlan_port_get_rtnl(lowerdev);
1501 
1502 	/* Only 1 macvlan device can be created in passthru mode */
1503 	if (macvlan_passthru(port)) {
1504 		/* The macvlan port must be not created this time,
1505 		 * still goto destroy_macvlan_port for readability.
1506 		 */
1507 		err = -EINVAL;
1508 		goto destroy_macvlan_port;
1509 	}
1510 
1511 	vlan->lowerdev = lowerdev;
1512 	vlan->dev      = dev;
1513 	vlan->port     = port;
1514 	vlan->set_features = MACVLAN_FEATURES;
1515 
1516 	vlan->mode     = MACVLAN_MODE_VEPA;
1517 	if (data && data[IFLA_MACVLAN_MODE])
1518 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1519 
1520 	if (data && data[IFLA_MACVLAN_FLAGS])
1521 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1522 
1523 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1524 		if (port->count) {
1525 			err = -EINVAL;
1526 			goto destroy_macvlan_port;
1527 		}
1528 		macvlan_set_passthru(port);
1529 		eth_hw_addr_inherit(dev, lowerdev);
1530 	}
1531 
1532 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1533 		if (vlan->mode != MACVLAN_MODE_SOURCE) {
1534 			err = -EINVAL;
1535 			goto destroy_macvlan_port;
1536 		}
1537 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1538 		err = macvlan_changelink_sources(vlan, macmode, data);
1539 		if (err)
1540 			goto destroy_macvlan_port;
1541 	}
1542 
1543 	vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1544 	if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1545 		vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1546 
1547 	if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1548 		update_port_bc_cutoff(
1549 			vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1550 
1551 	err = register_netdevice(dev);
1552 	if (err < 0)
1553 		goto destroy_macvlan_port;
1554 
1555 	dev->priv_flags |= IFF_MACVLAN;
1556 	err = netdev_upper_dev_link(lowerdev, dev, extack);
1557 	if (err)
1558 		goto unregister_netdev;
1559 
1560 	list_add_tail_rcu(&vlan->list, &port->vlans);
1561 	update_port_bc_queue_len(vlan->port);
1562 	netif_stacked_transfer_operstate(lowerdev, dev);
1563 	linkwatch_fire_event(dev);
1564 
1565 	return 0;
1566 
1567 unregister_netdev:
1568 	/* macvlan_uninit would free the macvlan port */
1569 	unregister_netdevice(dev);
1570 	return err;
1571 destroy_macvlan_port:
1572 	/* the macvlan port may be freed by macvlan_uninit when fail to register.
1573 	 * so we destroy the macvlan port only when it's valid.
1574 	 */
1575 	if (create && macvlan_port_get_rtnl(lowerdev)) {
1576 		macvlan_flush_sources(port, vlan);
1577 		macvlan_port_destroy(port->dev);
1578 	}
1579 	return err;
1580 }
1581 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1582 
1583 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1584 			   struct nlattr *tb[], struct nlattr *data[],
1585 			   struct netlink_ext_ack *extack)
1586 {
1587 	return macvlan_common_newlink(src_net, dev, tb, data, extack);
1588 }
1589 
1590 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1591 {
1592 	struct macvlan_dev *vlan = netdev_priv(dev);
1593 
1594 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1595 		macvlan_flush_sources(vlan->port, vlan);
1596 	list_del_rcu(&vlan->list);
1597 	update_port_bc_queue_len(vlan->port);
1598 	unregister_netdevice_queue(dev, head);
1599 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1600 }
1601 EXPORT_SYMBOL_GPL(macvlan_dellink);
1602 
1603 static int macvlan_changelink(struct net_device *dev,
1604 			      struct nlattr *tb[], struct nlattr *data[],
1605 			      struct netlink_ext_ack *extack)
1606 {
1607 	struct macvlan_dev *vlan = netdev_priv(dev);
1608 	enum macvlan_mode mode;
1609 	bool set_mode = false;
1610 	enum macvlan_macaddr_mode macmode;
1611 	int ret;
1612 
1613 	/* Validate mode, but don't set yet: setting flags may fail. */
1614 	if (data && data[IFLA_MACVLAN_MODE]) {
1615 		set_mode = true;
1616 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1617 		/* Passthrough mode can't be set or cleared dynamically */
1618 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1619 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1620 			return -EINVAL;
1621 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1622 		    vlan->mode != mode)
1623 			macvlan_flush_sources(vlan->port, vlan);
1624 	}
1625 
1626 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1627 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1628 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1629 		if (macvlan_passthru(vlan->port) && promisc) {
1630 			int err;
1631 
1632 			if (flags & MACVLAN_FLAG_NOPROMISC)
1633 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1634 			else
1635 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1636 			if (err < 0)
1637 				return err;
1638 		}
1639 		vlan->flags = flags;
1640 	}
1641 
1642 	if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1643 		vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1644 		update_port_bc_queue_len(vlan->port);
1645 	}
1646 
1647 	if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1648 		update_port_bc_cutoff(
1649 			vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1650 
1651 	if (set_mode)
1652 		vlan->mode = mode;
1653 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1654 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1655 			return -EINVAL;
1656 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1657 		ret = macvlan_changelink_sources(vlan, macmode, data);
1658 		if (ret)
1659 			return ret;
1660 	}
1661 	return 0;
1662 }
1663 
1664 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1665 {
1666 	if (vlan->macaddr_count == 0)
1667 		return 0;
1668 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1669 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1670 }
1671 
1672 static size_t macvlan_get_size(const struct net_device *dev)
1673 {
1674 	struct macvlan_dev *vlan = netdev_priv(dev);
1675 
1676 	return (0
1677 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1678 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1679 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1680 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1681 		+ nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1682 		+ nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1683 		);
1684 }
1685 
1686 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1687 				     const struct macvlan_dev *vlan,
1688 				     const int i)
1689 {
1690 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1691 	struct macvlan_source_entry *entry;
1692 
1693 	hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1694 		if (entry->vlan != vlan)
1695 			continue;
1696 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1697 			return 1;
1698 	}
1699 	return 0;
1700 }
1701 
1702 static int macvlan_fill_info(struct sk_buff *skb,
1703 				const struct net_device *dev)
1704 {
1705 	struct macvlan_dev *vlan = netdev_priv(dev);
1706 	struct macvlan_port *port = vlan->port;
1707 	int i;
1708 	struct nlattr *nest;
1709 
1710 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1711 		goto nla_put_failure;
1712 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1713 		goto nla_put_failure;
1714 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1715 		goto nla_put_failure;
1716 	if (vlan->macaddr_count > 0) {
1717 		nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1718 		if (nest == NULL)
1719 			goto nla_put_failure;
1720 
1721 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1722 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1723 				goto nla_put_failure;
1724 		}
1725 		nla_nest_end(skb, nest);
1726 	}
1727 	if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1728 		goto nla_put_failure;
1729 	if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED, port->bc_queue_len_used))
1730 		goto nla_put_failure;
1731 	if (port->bc_cutoff != 1 &&
1732 	    nla_put_s32(skb, IFLA_MACVLAN_BC_CUTOFF, port->bc_cutoff))
1733 		goto nla_put_failure;
1734 	return 0;
1735 
1736 nla_put_failure:
1737 	return -EMSGSIZE;
1738 }
1739 
1740 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1741 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1742 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1743 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1744 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1745 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1746 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1747 	[IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1748 	[IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1749 	[IFLA_MACVLAN_BC_CUTOFF] = { .type = NLA_S32 },
1750 };
1751 
1752 int macvlan_link_register(struct rtnl_link_ops *ops)
1753 {
1754 	/* common fields */
1755 	ops->validate		= macvlan_validate;
1756 	ops->maxtype		= IFLA_MACVLAN_MAX;
1757 	ops->policy		= macvlan_policy;
1758 	ops->changelink		= macvlan_changelink;
1759 	ops->get_size		= macvlan_get_size;
1760 	ops->fill_info		= macvlan_fill_info;
1761 
1762 	return rtnl_link_register(ops);
1763 };
1764 EXPORT_SYMBOL_GPL(macvlan_link_register);
1765 
1766 static struct net *macvlan_get_link_net(const struct net_device *dev)
1767 {
1768 	return dev_net(macvlan_dev_real_dev(dev));
1769 }
1770 
1771 static struct rtnl_link_ops macvlan_link_ops = {
1772 	.kind		= "macvlan",
1773 	.setup		= macvlan_setup,
1774 	.newlink	= macvlan_newlink,
1775 	.dellink	= macvlan_dellink,
1776 	.get_link_net	= macvlan_get_link_net,
1777 	.priv_size      = sizeof(struct macvlan_dev),
1778 };
1779 
1780 static void update_port_bc_queue_len(struct macvlan_port *port)
1781 {
1782 	u32 max_bc_queue_len_req = 0;
1783 	struct macvlan_dev *vlan;
1784 
1785 	list_for_each_entry(vlan, &port->vlans, list) {
1786 		if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1787 			max_bc_queue_len_req = vlan->bc_queue_len_req;
1788 	}
1789 	port->bc_queue_len_used = max_bc_queue_len_req;
1790 }
1791 
1792 static int macvlan_device_event(struct notifier_block *unused,
1793 				unsigned long event, void *ptr)
1794 {
1795 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1796 	struct macvlan_dev *vlan, *next;
1797 	struct macvlan_port *port;
1798 	LIST_HEAD(list_kill);
1799 
1800 	if (!netif_is_macvlan_port(dev))
1801 		return NOTIFY_DONE;
1802 
1803 	port = macvlan_port_get_rtnl(dev);
1804 
1805 	switch (event) {
1806 	case NETDEV_UP:
1807 	case NETDEV_DOWN:
1808 	case NETDEV_CHANGE:
1809 		list_for_each_entry(vlan, &port->vlans, list)
1810 			netif_stacked_transfer_operstate(vlan->lowerdev,
1811 							 vlan->dev);
1812 		break;
1813 	case NETDEV_FEAT_CHANGE:
1814 		list_for_each_entry(vlan, &port->vlans, list) {
1815 			netif_inherit_tso_max(vlan->dev, dev);
1816 			netdev_update_features(vlan->dev);
1817 		}
1818 		break;
1819 	case NETDEV_CHANGEMTU:
1820 		list_for_each_entry(vlan, &port->vlans, list) {
1821 			if (vlan->dev->mtu <= dev->mtu)
1822 				continue;
1823 			dev_set_mtu(vlan->dev, dev->mtu);
1824 		}
1825 		break;
1826 	case NETDEV_CHANGEADDR:
1827 		if (!macvlan_passthru(port))
1828 			return NOTIFY_DONE;
1829 
1830 		vlan = list_first_entry_or_null(&port->vlans,
1831 						struct macvlan_dev,
1832 						list);
1833 
1834 		if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1835 			return NOTIFY_BAD;
1836 
1837 		break;
1838 	case NETDEV_UNREGISTER:
1839 		/* twiddle thumbs on netns device moves */
1840 		if (dev->reg_state != NETREG_UNREGISTERING)
1841 			break;
1842 
1843 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1844 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1845 		unregister_netdevice_many(&list_kill);
1846 		break;
1847 	case NETDEV_PRE_TYPE_CHANGE:
1848 		/* Forbid underlying device to change its type. */
1849 		return NOTIFY_BAD;
1850 
1851 	case NETDEV_NOTIFY_PEERS:
1852 	case NETDEV_BONDING_FAILOVER:
1853 	case NETDEV_RESEND_IGMP:
1854 		/* Propagate to all vlans */
1855 		list_for_each_entry(vlan, &port->vlans, list)
1856 			call_netdevice_notifiers(event, vlan->dev);
1857 	}
1858 	return NOTIFY_DONE;
1859 }
1860 
1861 static struct notifier_block macvlan_notifier_block __read_mostly = {
1862 	.notifier_call	= macvlan_device_event,
1863 };
1864 
1865 static int __init macvlan_init_module(void)
1866 {
1867 	int err;
1868 
1869 	register_netdevice_notifier(&macvlan_notifier_block);
1870 
1871 	err = macvlan_link_register(&macvlan_link_ops);
1872 	if (err < 0)
1873 		goto err1;
1874 	return 0;
1875 err1:
1876 	unregister_netdevice_notifier(&macvlan_notifier_block);
1877 	return err;
1878 }
1879 
1880 static void __exit macvlan_cleanup_module(void)
1881 {
1882 	rtnl_link_unregister(&macvlan_link_ops);
1883 	unregister_netdevice_notifier(&macvlan_notifier_block);
1884 }
1885 
1886 module_init(macvlan_init_module);
1887 module_exit(macvlan_cleanup_module);
1888 
1889 MODULE_LICENSE("GPL");
1890 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1891 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1892 MODULE_ALIAS_RTNL_LINK("macvlan");
1893