xref: /linux/drivers/net/macvlan.c (revision 202ab599413c9a66e3e4449886b85c7b21314d50)
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/netdev_lock.h>
32 #include <net/rtnetlink.h>
33 #include <net/xfrm.h>
34 #include <linux/netpoll.h>
35 #include <linux/phy.h>
36 
37 #define MACVLAN_HASH_BITS	8
38 #define MACVLAN_HASH_SIZE	(1<<MACVLAN_HASH_BITS)
39 #define MACVLAN_DEFAULT_BC_QUEUE_LEN	1000
40 
41 #define MACVLAN_F_PASSTHRU	1
42 #define MACVLAN_F_ADDRCHANGE	2
43 
44 struct macvlan_port {
45 	struct net_device	*dev;
46 	struct hlist_head	vlan_hash[MACVLAN_HASH_SIZE];
47 	struct list_head	vlans;
48 	struct sk_buff_head	bc_queue;
49 	struct work_struct	bc_work;
50 	u32			bc_queue_len_used;
51 	int			bc_cutoff;
52 	u32			flags;
53 	int			count;
54 	struct hlist_head	vlan_source_hash[MACVLAN_HASH_SIZE];
55 	DECLARE_BITMAP(bc_filter, MACVLAN_MC_FILTER_SZ);
56 	DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
57 	unsigned char           perm_addr[ETH_ALEN];
58 };
59 
60 struct macvlan_source_entry {
61 	struct hlist_node	hlist;
62 	struct macvlan_dev __rcu *vlan;
63 	unsigned char		addr[6+2] __aligned(sizeof(u16));
64 	struct rcu_head		rcu;
65 };
66 
67 struct macvlan_skb_cb {
68 	const struct macvlan_dev *src;
69 };
70 
71 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
72 
73 static void macvlan_port_destroy(struct net_device *dev);
74 static void update_port_bc_queue_len(struct macvlan_port *port);
75 
76 static inline bool macvlan_passthru(const struct macvlan_port *port)
77 {
78 	return port->flags & MACVLAN_F_PASSTHRU;
79 }
80 
81 static inline void macvlan_set_passthru(struct macvlan_port *port)
82 {
83 	port->flags |= MACVLAN_F_PASSTHRU;
84 }
85 
86 static inline bool macvlan_addr_change(const struct macvlan_port *port)
87 {
88 	return port->flags & MACVLAN_F_ADDRCHANGE;
89 }
90 
91 static inline void macvlan_set_addr_change(struct macvlan_port *port)
92 {
93 	port->flags |= MACVLAN_F_ADDRCHANGE;
94 }
95 
96 static inline void macvlan_clear_addr_change(struct macvlan_port *port)
97 {
98 	port->flags &= ~MACVLAN_F_ADDRCHANGE;
99 }
100 
101 /* Hash Ethernet address */
102 static u32 macvlan_eth_hash(const unsigned char *addr)
103 {
104 	u64 value = get_unaligned((u64 *)addr);
105 
106 	/* only want 6 bytes */
107 #ifdef __BIG_ENDIAN
108 	value >>= 16;
109 #else
110 	value <<= 16;
111 #endif
112 	return hash_64(value, MACVLAN_HASH_BITS);
113 }
114 
115 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
116 {
117 	return rcu_dereference(dev->rx_handler_data);
118 }
119 
120 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
121 {
122 	return rtnl_dereference(dev->rx_handler_data);
123 }
124 
125 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
126 					       const unsigned char *addr)
127 {
128 	struct macvlan_dev *vlan;
129 	u32 idx = macvlan_eth_hash(addr);
130 
131 	hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist,
132 				 lockdep_rtnl_is_held()) {
133 		if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
134 			return vlan;
135 	}
136 	return NULL;
137 }
138 
139 static struct macvlan_source_entry *macvlan_hash_lookup_source(
140 	const struct macvlan_dev *vlan,
141 	const unsigned char *addr)
142 {
143 	struct macvlan_source_entry *entry;
144 	u32 idx = macvlan_eth_hash(addr);
145 	struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
146 
147 	hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
148 		if (ether_addr_equal_64bits(entry->addr, addr) &&
149 		    rcu_access_pointer(entry->vlan) == vlan)
150 			return entry;
151 	}
152 	return NULL;
153 }
154 
155 static int macvlan_hash_add_source(struct macvlan_dev *vlan,
156 				   const unsigned char *addr)
157 {
158 	struct macvlan_port *port = vlan->port;
159 	struct macvlan_source_entry *entry;
160 	struct hlist_head *h;
161 
162 	entry = macvlan_hash_lookup_source(vlan, addr);
163 	if (entry)
164 		return 0;
165 
166 	entry = kmalloc_obj(*entry);
167 	if (!entry)
168 		return -ENOMEM;
169 
170 	ether_addr_copy(entry->addr, addr);
171 	RCU_INIT_POINTER(entry->vlan, vlan);
172 	h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
173 	hlist_add_head_rcu(&entry->hlist, h);
174 	vlan->macaddr_count++;
175 
176 	return 0;
177 }
178 
179 static void macvlan_hash_add(struct macvlan_dev *vlan)
180 {
181 	struct macvlan_port *port = vlan->port;
182 	const unsigned char *addr = vlan->dev->dev_addr;
183 	u32 idx = macvlan_eth_hash(addr);
184 
185 	hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
186 }
187 
188 static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
189 {
190 	RCU_INIT_POINTER(entry->vlan, NULL);
191 	hlist_del_rcu(&entry->hlist);
192 	kfree_rcu(entry, rcu);
193 }
194 
195 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
196 {
197 	hlist_del_rcu(&vlan->hlist);
198 	if (sync)
199 		synchronize_rcu();
200 }
201 
202 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
203 					const unsigned char *addr)
204 {
205 	macvlan_hash_del(vlan, true);
206 	/* Now that we are unhashed it is safe to change the device
207 	 * address without confusing packet delivery.
208 	 */
209 	eth_hw_addr_set(vlan->dev, addr);
210 	macvlan_hash_add(vlan);
211 }
212 
213 static bool macvlan_addr_busy(const struct macvlan_port *port,
214 			      const unsigned char *addr)
215 {
216 	/* Test to see if the specified address is
217 	 * currently in use by the underlying device or
218 	 * another macvlan.
219 	 */
220 	if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
221 	    ether_addr_equal_64bits(port->dev->dev_addr, addr))
222 		return true;
223 
224 	if (macvlan_hash_lookup(port, addr))
225 		return true;
226 
227 	return false;
228 }
229 
230 
231 static int macvlan_broadcast_one(struct sk_buff *skb,
232 				 const struct macvlan_dev *vlan,
233 				 const struct ethhdr *eth, bool local)
234 {
235 	struct net_device *dev = vlan->dev;
236 
237 	if (local)
238 		return __dev_forward_skb(dev, skb);
239 
240 	skb->dev = dev;
241 	if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
242 		skb->pkt_type = PACKET_BROADCAST;
243 	else
244 		skb->pkt_type = PACKET_MULTICAST;
245 
246 	return 0;
247 }
248 
249 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
250 {
251 	return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
252 }
253 
254 
255 static unsigned int mc_hash(const struct macvlan_dev *vlan,
256 			    const unsigned char *addr)
257 {
258 	u32 val = get_unaligned((u32 *)(addr + 2));
259 
260 	val ^= macvlan_hash_mix(vlan);
261 	return hash_32(val, MACVLAN_MC_FILTER_BITS);
262 }
263 
264 static void macvlan_broadcast(struct sk_buff *skb,
265 			      const struct macvlan_port *port,
266 			      struct net_device *src,
267 			      enum macvlan_mode mode)
268 {
269 	const struct ethhdr *eth = eth_hdr(skb);
270 	const struct macvlan_dev *vlan;
271 	struct sk_buff *nskb;
272 	unsigned int i;
273 	int err;
274 	unsigned int hash;
275 
276 	if (skb->protocol == htons(ETH_P_PAUSE))
277 		return;
278 
279 	hash_for_each_rcu(port->vlan_hash, i, vlan, hlist) {
280 		if (vlan->dev == src || !(vlan->mode & mode))
281 			continue;
282 
283 		hash = mc_hash(vlan, eth->h_dest);
284 		if (!test_bit(hash, vlan->mc_filter))
285 			continue;
286 
287 		err = NET_RX_DROP;
288 		nskb = skb_clone(skb, GFP_ATOMIC);
289 		if (likely(nskb))
290 			err = macvlan_broadcast_one(nskb, vlan, eth,
291 					mode == MACVLAN_MODE_BRIDGE) ?:
292 			      netif_rx(nskb);
293 		macvlan_count_rx(vlan, skb->len + ETH_HLEN,
294 				 err == NET_RX_SUCCESS, true);
295 	}
296 }
297 
298 static void macvlan_multicast_rx(const struct macvlan_port *port,
299 				 const struct macvlan_dev *src,
300 				 struct sk_buff *skb)
301 {
302 	if (!src)
303 		/* frame comes from an external address */
304 		macvlan_broadcast(skb, port, NULL,
305 				  MACVLAN_MODE_PRIVATE |
306 				  MACVLAN_MODE_VEPA    |
307 				  MACVLAN_MODE_PASSTHRU|
308 				  MACVLAN_MODE_BRIDGE);
309 	else if (src->mode == MACVLAN_MODE_VEPA)
310 		/* flood to everyone except source */
311 		macvlan_broadcast(skb, port, src->dev,
312 				  MACVLAN_MODE_VEPA |
313 				  MACVLAN_MODE_BRIDGE);
314 	else
315 		/*
316 		 * Flood to VEPA and bridge ports. We cannot distinguish
317 		 * a looped-back locally-originated multicast from one
318 		 * sent by an external source sharing the same source MAC
319 		 * (e.g., VRRP virtual MAC), so deliver to bridge ports
320 		 * as well to ensure correct reception in all cases.
321 		 */
322 		macvlan_broadcast(skb, port, NULL,
323 				  MACVLAN_MODE_VEPA |
324 				  MACVLAN_MODE_BRIDGE);
325 }
326 
327 static void macvlan_process_broadcast(struct work_struct *w)
328 {
329 	struct macvlan_port *port = container_of(w, struct macvlan_port,
330 						 bc_work);
331 	struct sk_buff *skb;
332 	struct sk_buff_head list;
333 
334 	__skb_queue_head_init(&list);
335 
336 	spin_lock_bh(&port->bc_queue.lock);
337 	skb_queue_splice_tail_init(&port->bc_queue, &list);
338 	spin_unlock_bh(&port->bc_queue.lock);
339 
340 	while ((skb = __skb_dequeue(&list))) {
341 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
342 
343 		rcu_read_lock();
344 		macvlan_multicast_rx(port, src, skb);
345 		rcu_read_unlock();
346 
347 		if (src)
348 			dev_put(src->dev);
349 		consume_skb(skb);
350 
351 		cond_resched();
352 	}
353 }
354 
355 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
356 				      const struct macvlan_dev *src,
357 				      struct sk_buff *skb)
358 {
359 	u32 bc_queue_len_used = READ_ONCE(port->bc_queue_len_used);
360 	struct sk_buff *nskb;
361 	int err = -ENOMEM;
362 
363 	if (skb_queue_len_lockless(&port->bc_queue) >= bc_queue_len_used)
364 		goto err;
365 
366 	nskb = skb_clone(skb, GFP_ATOMIC);
367 	if (!nskb)
368 		goto err;
369 
370 	MACVLAN_SKB_CB(nskb)->src = src;
371 
372 	spin_lock(&port->bc_queue.lock);
373 	if (skb_queue_len(&port->bc_queue) < bc_queue_len_used) {
374 		if (src)
375 			dev_hold(src->dev);
376 		__skb_queue_tail(&port->bc_queue, nskb);
377 		err = 0;
378 	}
379 	spin_unlock(&port->bc_queue.lock);
380 
381 	queue_work(system_dfl_wq, &port->bc_work);
382 
383 	if (err)
384 		goto free_nskb;
385 
386 	return;
387 
388 free_nskb:
389 	kfree_skb_reason(nskb, SKB_DROP_REASON_MACVLAN_BROADCAST_BACKLOG);
390 err:
391 	dev_core_stats_rx_dropped_inc(skb->dev);
392 }
393 
394 static void macvlan_flush_sources(struct macvlan_port *port,
395 				  struct macvlan_dev *vlan)
396 {
397 	struct macvlan_source_entry *entry;
398 	struct hlist_node *next;
399 	int i;
400 
401 	hash_for_each_safe(port->vlan_source_hash, i, next, entry, hlist)
402 		if (rcu_access_pointer(entry->vlan) == vlan)
403 			macvlan_hash_del_source(entry);
404 
405 	vlan->macaddr_count = 0;
406 }
407 
408 static void macvlan_forward_source_one(struct sk_buff *skb,
409 				       struct macvlan_dev *vlan)
410 {
411 	struct sk_buff *nskb;
412 	struct net_device *dev;
413 	int len;
414 	int ret;
415 
416 	dev = vlan->dev;
417 	if (unlikely(!(dev->flags & IFF_UP)))
418 		return;
419 
420 	nskb = skb_clone(skb, GFP_ATOMIC);
421 	if (!nskb)
422 		return;
423 
424 	len = nskb->len + ETH_HLEN;
425 	nskb->dev = dev;
426 
427 	if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
428 		nskb->pkt_type = PACKET_HOST;
429 
430 	ret = __netif_rx(nskb);
431 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
432 }
433 
434 static bool macvlan_forward_source(struct sk_buff *skb,
435 				   struct macvlan_port *port,
436 				   const unsigned char *addr)
437 {
438 	struct macvlan_source_entry *entry;
439 	u32 idx = macvlan_eth_hash(addr);
440 	struct hlist_head *h = &port->vlan_source_hash[idx];
441 	bool consume = false;
442 
443 	hlist_for_each_entry_rcu(entry, h, hlist) {
444 		if (ether_addr_equal_64bits(entry->addr, addr)) {
445 			struct macvlan_dev *vlan = rcu_dereference(entry->vlan);
446 
447 			if (!vlan)
448 				continue;
449 
450 			if (vlan->flags & MACVLAN_FLAG_NODST)
451 				consume = true;
452 			macvlan_forward_source_one(skb, vlan);
453 		}
454 	}
455 
456 	return consume;
457 }
458 
459 /* called under rcu_read_lock() from netif_receive_skb */
460 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
461 {
462 	struct macvlan_port *port;
463 	struct sk_buff *skb = *pskb;
464 	const struct ethhdr *eth = eth_hdr(skb);
465 	const struct macvlan_dev *vlan;
466 	const struct macvlan_dev *src;
467 	struct net_device *dev;
468 	unsigned int len = 0;
469 	int ret;
470 	rx_handler_result_t handle_res;
471 
472 	/* Packets from dev_loopback_xmit() do not have L2 header, bail out */
473 	if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
474 		return RX_HANDLER_PASS;
475 
476 	port = macvlan_port_get_rcu(skb->dev);
477 	if (is_multicast_ether_addr(eth->h_dest)) {
478 		unsigned int hash;
479 
480 		skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
481 		if (!skb)
482 			return RX_HANDLER_CONSUMED;
483 		*pskb = skb;
484 		eth = eth_hdr(skb);
485 		if (macvlan_forward_source(skb, port, eth->h_source)) {
486 			kfree_skb(skb);
487 			return RX_HANDLER_CONSUMED;
488 		}
489 		src = macvlan_hash_lookup(port, eth->h_source);
490 		if (src && src->mode != MACVLAN_MODE_VEPA &&
491 		    src->mode != MACVLAN_MODE_BRIDGE) {
492 			/* forward to original port. */
493 			vlan = src;
494 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
495 			      __netif_rx(skb);
496 			handle_res = RX_HANDLER_CONSUMED;
497 			goto out;
498 		}
499 
500 		hash = mc_hash(NULL, eth->h_dest);
501 		if (test_bit(hash, port->bc_filter))
502 			macvlan_broadcast_enqueue(port, src, skb);
503 		else if (test_bit(hash, port->mc_filter))
504 			macvlan_multicast_rx(port, src, skb);
505 
506 		return RX_HANDLER_PASS;
507 	}
508 
509 	if (macvlan_forward_source(skb, port, eth->h_source)) {
510 		kfree_skb(skb);
511 		return RX_HANDLER_CONSUMED;
512 	}
513 	if (macvlan_passthru(port))
514 		vlan = list_first_or_null_rcu(&port->vlans,
515 					      struct macvlan_dev, list);
516 	else
517 		vlan = macvlan_hash_lookup(port, eth->h_dest);
518 	if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
519 		return RX_HANDLER_PASS;
520 
521 	dev = vlan->dev;
522 	if (unlikely(!(dev->flags & IFF_UP))) {
523 		kfree_skb(skb);
524 		return RX_HANDLER_CONSUMED;
525 	}
526 	len = skb->len + ETH_HLEN;
527 	skb = skb_share_check(skb, GFP_ATOMIC);
528 	if (!skb) {
529 		ret = NET_RX_DROP;
530 		handle_res = RX_HANDLER_CONSUMED;
531 		goto out;
532 	}
533 
534 	*pskb = skb;
535 	skb->dev = dev;
536 	skb->pkt_type = PACKET_HOST;
537 
538 	ret = NET_RX_SUCCESS;
539 	handle_res = RX_HANDLER_ANOTHER;
540 out:
541 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
542 	return handle_res;
543 }
544 
545 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
546 {
547 	const struct macvlan_dev *vlan = netdev_priv(dev);
548 	const struct macvlan_port *port = vlan->port;
549 	const struct macvlan_dev *dest;
550 
551 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
552 		const struct ethhdr *eth = skb_eth_hdr(skb);
553 
554 		/* send to other bridge ports directly */
555 		if (is_multicast_ether_addr(eth->h_dest)) {
556 			skb_reset_mac_header(skb);
557 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
558 			goto xmit_world;
559 		}
560 
561 		dest = macvlan_hash_lookup(port, eth->h_dest);
562 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
563 			/* send to lowerdev first for its network taps */
564 			dev_forward_skb(vlan->lowerdev, skb);
565 
566 			return NET_XMIT_SUCCESS;
567 		}
568 	}
569 xmit_world:
570 	skb->dev = vlan->lowerdev;
571 	return dev_queue_xmit_accel(skb,
572 				    netdev_get_sb_channel(dev) ? dev : NULL);
573 }
574 
575 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
576 {
577 #ifdef CONFIG_NET_POLL_CONTROLLER
578 	return netpoll_send_skb(vlan->netpoll, skb);
579 #else
580 	BUG();
581 	return NETDEV_TX_OK;
582 #endif
583 }
584 
585 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
586 				      struct net_device *dev)
587 {
588 	struct macvlan_dev *vlan = netdev_priv(dev);
589 	unsigned int len = skb->len;
590 	int ret;
591 
592 	if (unlikely(netpoll_tx_running(dev)))
593 		return macvlan_netpoll_send_skb(vlan, skb);
594 
595 	ret = macvlan_queue_xmit(skb, dev);
596 
597 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
598 		struct vlan_pcpu_stats *pcpu_stats;
599 
600 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
601 		u64_stats_update_begin(&pcpu_stats->syncp);
602 		u64_stats_inc(&pcpu_stats->tx_packets);
603 		u64_stats_add(&pcpu_stats->tx_bytes, len);
604 		u64_stats_update_end(&pcpu_stats->syncp);
605 	} else {
606 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
607 	}
608 	return ret;
609 }
610 
611 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
612 			       unsigned short type, const void *daddr,
613 			       const void *saddr, unsigned len)
614 {
615 	const struct macvlan_dev *vlan = netdev_priv(dev);
616 	struct net_device *lowerdev = vlan->lowerdev;
617 
618 	return dev_hard_header(skb, lowerdev, type, daddr,
619 			       saddr ? : dev->dev_addr, len);
620 }
621 
622 static const struct header_ops macvlan_hard_header_ops = {
623 	.create  	= macvlan_hard_header,
624 	.parse		= eth_header_parse,
625 	.cache		= eth_header_cache,
626 	.cache_update	= eth_header_cache_update,
627 	.parse_protocol	= eth_header_parse_protocol,
628 };
629 
630 static int macvlan_open(struct net_device *dev)
631 {
632 	struct macvlan_dev *vlan = netdev_priv(dev);
633 	struct net_device *lowerdev = vlan->lowerdev;
634 	int err;
635 
636 	if (macvlan_passthru(vlan->port)) {
637 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
638 			err = dev_set_promiscuity(lowerdev, 1);
639 			if (err < 0)
640 				goto out;
641 		}
642 		goto hash_add;
643 	}
644 
645 	err = -EADDRINUSE;
646 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
647 		goto out;
648 
649 	/* Attempt to populate accel_priv which is used to offload the L2
650 	 * forwarding requests for unicast packets.
651 	 */
652 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
653 		vlan->accel_priv =
654 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
655 
656 	/* If earlier attempt to offload failed, or accel_priv is not
657 	 * populated we must add the unicast address to the lower device.
658 	 */
659 	if (IS_ERR_OR_NULL(vlan->accel_priv)) {
660 		vlan->accel_priv = NULL;
661 		err = dev_uc_add(lowerdev, dev->dev_addr);
662 		if (err < 0)
663 			goto out;
664 	}
665 
666 	if (dev->flags & IFF_ALLMULTI) {
667 		err = dev_set_allmulti(lowerdev, 1);
668 		if (err < 0)
669 			goto del_unicast;
670 	}
671 
672 	if (dev->flags & IFF_PROMISC) {
673 		err = dev_set_promiscuity(lowerdev, 1);
674 		if (err < 0)
675 			goto clear_multi;
676 	}
677 
678 hash_add:
679 	macvlan_hash_add(vlan);
680 	return 0;
681 
682 clear_multi:
683 	if (dev->flags & IFF_ALLMULTI)
684 		dev_set_allmulti(lowerdev, -1);
685 del_unicast:
686 	if (vlan->accel_priv) {
687 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
688 							   vlan->accel_priv);
689 		vlan->accel_priv = NULL;
690 	} else {
691 		dev_uc_del(lowerdev, dev->dev_addr);
692 	}
693 out:
694 	return err;
695 }
696 
697 static int macvlan_stop(struct net_device *dev)
698 {
699 	struct macvlan_dev *vlan = netdev_priv(dev);
700 	struct net_device *lowerdev = vlan->lowerdev;
701 
702 	if (vlan->accel_priv) {
703 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
704 							   vlan->accel_priv);
705 		vlan->accel_priv = NULL;
706 	}
707 
708 	dev_uc_unsync(lowerdev, dev);
709 	dev_mc_unsync(lowerdev, dev);
710 
711 	if (macvlan_passthru(vlan->port)) {
712 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
713 			dev_set_promiscuity(lowerdev, -1);
714 		goto hash_del;
715 	}
716 
717 	if (dev->flags & IFF_ALLMULTI)
718 		dev_set_allmulti(lowerdev, -1);
719 
720 	if (dev->flags & IFF_PROMISC)
721 		dev_set_promiscuity(lowerdev, -1);
722 
723 	dev_uc_del(lowerdev, dev->dev_addr);
724 
725 hash_del:
726 	macvlan_hash_del(vlan, !dev->dismantle);
727 	return 0;
728 }
729 
730 static int macvlan_sync_address(struct net_device *dev,
731 				const unsigned char *addr)
732 {
733 	struct macvlan_dev *vlan = netdev_priv(dev);
734 	struct net_device *lowerdev = vlan->lowerdev;
735 	struct macvlan_port *port = vlan->port;
736 	int err;
737 
738 	if (!(dev->flags & IFF_UP)) {
739 		/* Just copy in the new address */
740 		eth_hw_addr_set(dev, addr);
741 	} else {
742 		/* Rehash and update the device filters */
743 		if (macvlan_addr_busy(vlan->port, addr))
744 			return -EADDRINUSE;
745 
746 		if (!macvlan_passthru(port)) {
747 			err = dev_uc_add(lowerdev, addr);
748 			if (err)
749 				return err;
750 
751 			dev_uc_del(lowerdev, dev->dev_addr);
752 		}
753 
754 		macvlan_hash_change_addr(vlan, addr);
755 	}
756 	if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
757 		/* Since addr_change isn't set, we are here due to lower
758 		 * device change.  Save the lower-dev address so we can
759 		 * restore it later.
760 		 */
761 		ether_addr_copy(vlan->port->perm_addr,
762 				lowerdev->dev_addr);
763 	}
764 	macvlan_clear_addr_change(port);
765 	return 0;
766 }
767 
768 static int macvlan_set_mac_address(struct net_device *dev, void *p)
769 {
770 	struct macvlan_dev *vlan = netdev_priv(dev);
771 	struct sockaddr_storage *addr = p;
772 
773 	if (!is_valid_ether_addr(addr->__data))
774 		return -EADDRNOTAVAIL;
775 
776 	/* If the addresses are the same, this is a no-op */
777 	if (ether_addr_equal(dev->dev_addr, addr->__data))
778 		return 0;
779 
780 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
781 		macvlan_set_addr_change(vlan->port);
782 		return dev_set_mac_address(vlan->lowerdev, addr, NULL);
783 	}
784 
785 	if (macvlan_addr_busy(vlan->port, addr->__data))
786 		return -EADDRINUSE;
787 
788 	return macvlan_sync_address(dev, addr->__data);
789 }
790 
791 static void macvlan_change_rx_flags(struct net_device *dev, int change)
792 {
793 	struct macvlan_dev *vlan = netdev_priv(dev);
794 	struct net_device *lowerdev = vlan->lowerdev;
795 
796 	if (dev->flags & IFF_UP) {
797 		if (change & IFF_ALLMULTI)
798 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
799 		if (!macvlan_passthru(vlan->port) && change & IFF_PROMISC)
800 			dev_set_promiscuity(lowerdev,
801 					    dev->flags & IFF_PROMISC ? 1 : -1);
802 
803 	}
804 }
805 
806 static void macvlan_compute_filter(unsigned long *mc_filter,
807 				   struct net_device *dev,
808 				   struct macvlan_dev *vlan, int cutoff)
809 {
810 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
811 		bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
812 	} else {
813 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
814 		struct netdev_hw_addr *ha;
815 
816 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
817 		netdev_for_each_mc_addr(ha, dev) {
818 			if (!vlan && ha->synced <= cutoff)
819 				continue;
820 
821 			__set_bit(mc_hash(vlan, ha->addr), filter);
822 		}
823 
824 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
825 
826 		bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
827 	}
828 }
829 
830 static void macvlan_recompute_bc_filter(struct macvlan_dev *vlan)
831 {
832 	if (vlan->port->bc_cutoff < 0) {
833 		bitmap_zero(vlan->port->bc_filter, MACVLAN_MC_FILTER_SZ);
834 		return;
835 	}
836 
837 	macvlan_compute_filter(vlan->port->bc_filter, vlan->lowerdev, NULL,
838 			       vlan->port->bc_cutoff);
839 }
840 
841 static void macvlan_set_mac_lists(struct net_device *dev)
842 {
843 	struct macvlan_dev *vlan = netdev_priv(dev);
844 
845 	macvlan_compute_filter(vlan->mc_filter, dev, vlan, 0);
846 
847 	dev_uc_sync(vlan->lowerdev, dev);
848 	dev_mc_sync(vlan->lowerdev, dev);
849 
850 	/* This is slightly inaccurate as we're including the subscription
851 	 * list of vlan->lowerdev too.
852 	 *
853 	 * Bug alert: This only works if everyone has the same broadcast
854 	 * address as lowerdev.  As soon as someone changes theirs this
855 	 * will break.
856 	 *
857 	 * However, this is already broken as when you change your broadcast
858 	 * address we don't get called.
859 	 *
860 	 * The solution is to maintain a list of broadcast addresses like
861 	 * we do for uc/mc, if you care.
862 	 */
863 	macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL,
864 			       0);
865 	macvlan_recompute_bc_filter(vlan);
866 }
867 
868 static void update_port_bc_cutoff(struct macvlan_dev *vlan, int cutoff)
869 {
870 	if (vlan->port->bc_cutoff == cutoff)
871 		return;
872 
873 	vlan->port->bc_cutoff = cutoff;
874 	macvlan_recompute_bc_filter(vlan);
875 }
876 
877 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
878 {
879 	struct macvlan_dev *vlan = netdev_priv(dev);
880 
881 	if (vlan->lowerdev->mtu < new_mtu)
882 		return -EINVAL;
883 	WRITE_ONCE(dev->mtu, new_mtu);
884 	return 0;
885 }
886 
887 static int macvlan_hwtstamp_get(struct net_device *dev,
888 				struct kernel_hwtstamp_config *cfg)
889 {
890 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
891 
892 	return generic_hwtstamp_get_lower(real_dev, cfg);
893 }
894 
895 static int macvlan_hwtstamp_set(struct net_device *dev,
896 				struct kernel_hwtstamp_config *cfg,
897 				struct netlink_ext_ack *extack)
898 {
899 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
900 
901 	if (!net_eq(dev_net(dev), &init_net))
902 		return -EOPNOTSUPP;
903 
904 	return generic_hwtstamp_set_lower(real_dev, cfg, extack);
905 }
906 
907 /*
908  * macvlan network devices have devices nesting below it and are a special
909  * "super class" of normal network devices; split their locks off into a
910  * separate class since they always nest.
911  */
912 static struct lock_class_key macvlan_netdev_addr_lock_key;
913 
914 #define ALWAYS_ON_OFFLOADS \
915 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
916 	 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
917 
918 #define ALWAYS_ON_FEATURES ALWAYS_ON_OFFLOADS
919 
920 #define MACVLAN_FEATURES \
921 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
922 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
923 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
924 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
925 
926 #define MACVLAN_STATE_MASK \
927 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
928 
929 static void macvlan_set_lockdep_class(struct net_device *dev)
930 {
931 	netdev_lockdep_set_classes(dev);
932 	lockdep_set_class(&dev->addr_list_lock,
933 			  &macvlan_netdev_addr_lock_key);
934 }
935 
936 static int macvlan_init(struct net_device *dev)
937 {
938 	struct macvlan_dev *vlan = netdev_priv(dev);
939 	struct net_device *lowerdev = vlan->lowerdev;
940 	struct macvlan_port *port = vlan->port;
941 
942 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
943 				  (lowerdev->state & MACVLAN_STATE_MASK);
944 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
945 	dev->features		|= ALWAYS_ON_FEATURES;
946 	dev->hw_features	|= NETIF_F_LRO;
947 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
948 	dev->vlan_features	|= ALWAYS_ON_OFFLOADS;
949 	dev->hw_enc_features    |= dev->features;
950 	dev->lltx		= true;
951 	netif_inherit_tso_max(dev, lowerdev);
952 	dev->hard_header_len	= lowerdev->hard_header_len;
953 	macvlan_set_lockdep_class(dev);
954 
955 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
956 	if (!vlan->pcpu_stats)
957 		return -ENOMEM;
958 
959 	port->count += 1;
960 
961 	/* Get macvlan's reference to lowerdev */
962 	netdev_hold(lowerdev, &vlan->dev_tracker, GFP_KERNEL);
963 
964 	return 0;
965 }
966 
967 static void macvlan_uninit(struct net_device *dev)
968 {
969 	struct macvlan_dev *vlan = netdev_priv(dev);
970 	struct macvlan_port *port = vlan->port;
971 
972 	free_percpu(vlan->pcpu_stats);
973 
974 	macvlan_flush_sources(port, vlan);
975 	port->count -= 1;
976 	if (!port->count)
977 		macvlan_port_destroy(port->dev);
978 }
979 
980 static void macvlan_dev_get_stats64(struct net_device *dev,
981 				    struct rtnl_link_stats64 *stats)
982 {
983 	struct macvlan_dev *vlan = netdev_priv(dev);
984 
985 	if (vlan->pcpu_stats) {
986 		struct vlan_pcpu_stats *p;
987 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
988 		u32 rx_errors = 0, tx_dropped = 0;
989 		unsigned int start;
990 		int i;
991 
992 		for_each_possible_cpu(i) {
993 			p = per_cpu_ptr(vlan->pcpu_stats, i);
994 			do {
995 				start = u64_stats_fetch_begin(&p->syncp);
996 				rx_packets	= u64_stats_read(&p->rx_packets);
997 				rx_bytes	= u64_stats_read(&p->rx_bytes);
998 				rx_multicast	= u64_stats_read(&p->rx_multicast);
999 				tx_packets	= u64_stats_read(&p->tx_packets);
1000 				tx_bytes	= u64_stats_read(&p->tx_bytes);
1001 			} while (u64_stats_fetch_retry(&p->syncp, start));
1002 
1003 			stats->rx_packets	+= rx_packets;
1004 			stats->rx_bytes		+= rx_bytes;
1005 			stats->multicast	+= rx_multicast;
1006 			stats->tx_packets	+= tx_packets;
1007 			stats->tx_bytes		+= tx_bytes;
1008 			/* rx_errors & tx_dropped are u32, updated
1009 			 * without syncp protection.
1010 			 */
1011 			rx_errors	+= READ_ONCE(p->rx_errors);
1012 			tx_dropped	+= READ_ONCE(p->tx_dropped);
1013 		}
1014 		stats->rx_errors	= rx_errors;
1015 		stats->rx_dropped	= rx_errors;
1016 		stats->tx_dropped	= tx_dropped;
1017 	}
1018 }
1019 
1020 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
1021 				   __be16 proto, u16 vid)
1022 {
1023 	struct macvlan_dev *vlan = netdev_priv(dev);
1024 	struct net_device *lowerdev = vlan->lowerdev;
1025 
1026 	return vlan_vid_add(lowerdev, proto, vid);
1027 }
1028 
1029 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
1030 				    __be16 proto, u16 vid)
1031 {
1032 	struct macvlan_dev *vlan = netdev_priv(dev);
1033 	struct net_device *lowerdev = vlan->lowerdev;
1034 
1035 	vlan_vid_del(lowerdev, proto, vid);
1036 	return 0;
1037 }
1038 
1039 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1040 			   struct net_device *dev,
1041 			   const unsigned char *addr, u16 vid,
1042 			   u16 flags, bool *notified,
1043 			   struct netlink_ext_ack *extack)
1044 {
1045 	struct macvlan_dev *vlan = netdev_priv(dev);
1046 	int err = -EINVAL;
1047 
1048 	/* Support unicast filter only on passthru devices.
1049 	 * Multicast filter should be allowed on all devices.
1050 	 */
1051 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1052 		return -EOPNOTSUPP;
1053 
1054 	if (flags & NLM_F_REPLACE)
1055 		return -EOPNOTSUPP;
1056 
1057 	if (is_unicast_ether_addr(addr))
1058 		err = dev_uc_add_excl(dev, addr);
1059 	else if (is_multicast_ether_addr(addr))
1060 		err = dev_mc_add_excl(dev, addr);
1061 
1062 	return err;
1063 }
1064 
1065 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1066 			   struct net_device *dev,
1067 			   const unsigned char *addr, u16 vid, bool *notified,
1068 			   struct netlink_ext_ack *extack)
1069 {
1070 	struct macvlan_dev *vlan = netdev_priv(dev);
1071 	int err = -EINVAL;
1072 
1073 	/* Support unicast filter only on passthru devices.
1074 	 * Multicast filter should be allowed on all devices.
1075 	 */
1076 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1077 		return -EOPNOTSUPP;
1078 
1079 	if (is_unicast_ether_addr(addr))
1080 		err = dev_uc_del(dev, addr);
1081 	else if (is_multicast_ether_addr(addr))
1082 		err = dev_mc_del(dev, addr);
1083 
1084 	return err;
1085 }
1086 
1087 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1088 					struct ethtool_drvinfo *drvinfo)
1089 {
1090 	strscpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1091 	strscpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1092 }
1093 
1094 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1095 					      struct ethtool_link_ksettings *cmd)
1096 {
1097 	const struct macvlan_dev *vlan = netdev_priv(dev);
1098 
1099 	return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1100 }
1101 
1102 static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1103 				       struct kernel_ethtool_ts_info *info)
1104 {
1105 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
1106 
1107 	return ethtool_get_ts_info_by_layer(real_dev, info);
1108 }
1109 
1110 static netdev_features_t macvlan_fix_features(struct net_device *dev,
1111 					      netdev_features_t features)
1112 {
1113 	struct macvlan_dev *vlan = netdev_priv(dev);
1114 	netdev_features_t lowerdev_features = vlan->lowerdev->features;
1115 	netdev_features_t mask;
1116 
1117 	features |= NETIF_F_ALL_FOR_ALL;
1118 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
1119 	mask = features;
1120 
1121 	lowerdev_features &= (features | ~NETIF_F_LRO);
1122 	features = netdev_increment_features(lowerdev_features, features, mask);
1123 	features |= ALWAYS_ON_FEATURES;
1124 	features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1125 
1126 	return features;
1127 }
1128 
1129 #ifdef CONFIG_NET_POLL_CONTROLLER
1130 static void macvlan_dev_poll_controller(struct net_device *dev)
1131 {
1132 	return;
1133 }
1134 
1135 static int macvlan_dev_netpoll_setup(struct net_device *dev)
1136 {
1137 	struct macvlan_dev *vlan = netdev_priv(dev);
1138 	struct net_device *real_dev = vlan->lowerdev;
1139 	struct netpoll *netpoll;
1140 	int err;
1141 
1142 	netpoll = kzalloc_obj(*netpoll);
1143 	err = -ENOMEM;
1144 	if (!netpoll)
1145 		goto out;
1146 
1147 	err = __netpoll_setup(netpoll, real_dev);
1148 	if (err) {
1149 		kfree(netpoll);
1150 		goto out;
1151 	}
1152 
1153 	vlan->netpoll = netpoll;
1154 
1155 out:
1156 	return err;
1157 }
1158 
1159 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1160 {
1161 	struct macvlan_dev *vlan = netdev_priv(dev);
1162 	struct netpoll *netpoll = vlan->netpoll;
1163 
1164 	if (!netpoll)
1165 		return;
1166 
1167 	vlan->netpoll = NULL;
1168 
1169 	__netpoll_free(netpoll);
1170 }
1171 #endif	/* CONFIG_NET_POLL_CONTROLLER */
1172 
1173 static int macvlan_dev_get_iflink(const struct net_device *dev)
1174 {
1175 	struct macvlan_dev *vlan = netdev_priv(dev);
1176 
1177 	return READ_ONCE(vlan->lowerdev->ifindex);
1178 }
1179 
1180 static const struct ethtool_ops macvlan_ethtool_ops = {
1181 	.get_link		= ethtool_op_get_link,
1182 	.get_link_ksettings	= macvlan_ethtool_get_link_ksettings,
1183 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
1184 	.get_ts_info		= macvlan_ethtool_get_ts_info,
1185 };
1186 
1187 static const struct net_device_ops macvlan_netdev_ops = {
1188 	.ndo_init		= macvlan_init,
1189 	.ndo_uninit		= macvlan_uninit,
1190 	.ndo_open		= macvlan_open,
1191 	.ndo_stop		= macvlan_stop,
1192 	.ndo_start_xmit		= macvlan_start_xmit,
1193 	.ndo_change_mtu		= macvlan_change_mtu,
1194 	.ndo_fix_features	= macvlan_fix_features,
1195 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1196 	.ndo_set_mac_address	= macvlan_set_mac_address,
1197 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1198 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1199 	.ndo_validate_addr	= eth_validate_addr,
1200 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1201 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1202 	.ndo_fdb_add		= macvlan_fdb_add,
1203 	.ndo_fdb_del		= macvlan_fdb_del,
1204 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1205 #ifdef CONFIG_NET_POLL_CONTROLLER
1206 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1207 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1208 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1209 #endif
1210 	.ndo_get_iflink		= macvlan_dev_get_iflink,
1211 	.ndo_features_check	= passthru_features_check,
1212 	.ndo_hwtstamp_get	= macvlan_hwtstamp_get,
1213 	.ndo_hwtstamp_set	= macvlan_hwtstamp_set,
1214 };
1215 
1216 static void macvlan_dev_free(struct net_device *dev)
1217 {
1218 	struct macvlan_dev *vlan = netdev_priv(dev);
1219 
1220 	/* Get rid of the macvlan's reference to lowerdev */
1221 	netdev_put(vlan->lowerdev, &vlan->dev_tracker);
1222 }
1223 
1224 void macvlan_common_setup(struct net_device *dev)
1225 {
1226 	ether_setup(dev);
1227 
1228 	/* ether_setup() has set dev->min_mtu to ETH_MIN_MTU. */
1229 	dev->max_mtu		= ETH_MAX_MTU;
1230 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1231 	netif_keep_dst(dev);
1232 	dev->priv_flags	       |= IFF_UNICAST_FLT;
1233 	dev->change_proto_down	= true;
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_obj(*port);
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_storage ss;
1313 
1314 		ss.ss_family = port->dev->type;
1315 		memcpy(&ss.__data, port->perm_addr, port->dev->addr_len);
1316 		dev_set_mac_address(port->dev, &ss, 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_device *dev,
1459 			   struct rtnl_newlink_params *params,
1460 			   struct netlink_ext_ack *extack)
1461 {
1462 	struct net *link_net = rtnl_newlink_link_net(params);
1463 	struct macvlan_dev *vlan = netdev_priv(dev);
1464 	struct nlattr **data = params->data;
1465 	struct nlattr **tb = params->tb;
1466 	struct net_device *lowerdev;
1467 	struct macvlan_port *port;
1468 	bool create = false;
1469 	int macmode;
1470 	int err;
1471 
1472 	if (!tb[IFLA_LINK])
1473 		return -EINVAL;
1474 
1475 	lowerdev = __dev_get_by_index(link_net, nla_get_u32(tb[IFLA_LINK]));
1476 	if (lowerdev == NULL)
1477 		return -ENODEV;
1478 
1479 	/* When creating macvlans or macvtaps on top of other macvlans - use
1480 	 * the real device as the lowerdev.
1481 	 */
1482 	if (netif_is_macvlan(lowerdev))
1483 		lowerdev = macvlan_dev_real_dev(lowerdev);
1484 
1485 	if (!tb[IFLA_MTU])
1486 		dev->mtu = lowerdev->mtu;
1487 	else if (dev->mtu > lowerdev->mtu)
1488 		return -EINVAL;
1489 
1490 	/* MTU range: 68 - lowerdev->max_mtu */
1491 	dev->min_mtu = ETH_MIN_MTU;
1492 	dev->max_mtu = lowerdev->max_mtu;
1493 
1494 	if (!tb[IFLA_ADDRESS])
1495 		eth_hw_addr_random(dev);
1496 
1497 	if (!netif_is_macvlan_port(lowerdev)) {
1498 		err = macvlan_port_create(lowerdev);
1499 		if (err < 0)
1500 			return err;
1501 		create = true;
1502 	}
1503 	port = macvlan_port_get_rtnl(lowerdev);
1504 
1505 	/* Only 1 macvlan device can be created in passthru mode */
1506 	if (macvlan_passthru(port)) {
1507 		/* The macvlan port must be not created this time,
1508 		 * still goto destroy_macvlan_port for readability.
1509 		 */
1510 		err = -EINVAL;
1511 		goto destroy_macvlan_port;
1512 	}
1513 
1514 	vlan->lowerdev = lowerdev;
1515 	vlan->dev      = dev;
1516 	vlan->port     = port;
1517 	vlan->set_features = MACVLAN_FEATURES;
1518 
1519 	vlan->mode     = MACVLAN_MODE_VEPA;
1520 	if (data && data[IFLA_MACVLAN_MODE])
1521 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1522 
1523 	if (data && data[IFLA_MACVLAN_FLAGS])
1524 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1525 
1526 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1527 		if (port->count) {
1528 			err = -EINVAL;
1529 			goto destroy_macvlan_port;
1530 		}
1531 		macvlan_set_passthru(port);
1532 		eth_hw_addr_inherit(dev, lowerdev);
1533 	}
1534 
1535 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1536 		if (vlan->mode != MACVLAN_MODE_SOURCE) {
1537 			err = -EINVAL;
1538 			goto destroy_macvlan_port;
1539 		}
1540 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1541 		err = macvlan_changelink_sources(vlan, macmode, data);
1542 		if (err)
1543 			goto destroy_macvlan_port;
1544 	}
1545 
1546 	vlan->bc_queue_len_req = MACVLAN_DEFAULT_BC_QUEUE_LEN;
1547 	if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN])
1548 		vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1549 
1550 	if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1551 		update_port_bc_cutoff(
1552 			vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1553 
1554 	err = register_netdevice(dev);
1555 	if (err < 0)
1556 		goto destroy_macvlan_port;
1557 
1558 	dev->priv_flags |= IFF_MACVLAN;
1559 	err = netdev_upper_dev_link(lowerdev, dev, extack);
1560 	if (err)
1561 		goto unregister_netdev;
1562 
1563 	list_add_tail_rcu(&vlan->list, &port->vlans);
1564 	update_port_bc_queue_len(vlan->port);
1565 	netif_stacked_transfer_operstate(lowerdev, dev);
1566 	linkwatch_fire_event(dev);
1567 
1568 	return 0;
1569 
1570 unregister_netdev:
1571 	/* macvlan_uninit would free the macvlan port */
1572 	unregister_netdevice(dev);
1573 	return err;
1574 destroy_macvlan_port:
1575 	/* the macvlan port may be freed by macvlan_uninit when fail to register.
1576 	 * so we destroy the macvlan port only when it's valid.
1577 	 */
1578 	if (macvlan_port_get_rtnl(lowerdev)) {
1579 		macvlan_flush_sources(port, vlan);
1580 		if (create)
1581 			macvlan_port_destroy(port->dev);
1582 	}
1583 	/* @dev might have been made visible before an error was detected.
1584 	 * Make sure to observe an RCU grace period before our caller
1585 	 * (rtnl_newlink()) frees it.
1586 	 */
1587 	synchronize_net();
1588 	return err;
1589 }
1590 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1591 
1592 static int macvlan_newlink(struct net_device *dev,
1593 			   struct rtnl_newlink_params *params,
1594 			   struct netlink_ext_ack *extack)
1595 {
1596 	return macvlan_common_newlink(dev, params, extack);
1597 }
1598 
1599 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1600 {
1601 	struct macvlan_dev *vlan = netdev_priv(dev);
1602 
1603 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1604 		macvlan_flush_sources(vlan->port, vlan);
1605 	list_del_rcu(&vlan->list);
1606 	update_port_bc_queue_len(vlan->port);
1607 	unregister_netdevice_queue(dev, head);
1608 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1609 }
1610 EXPORT_SYMBOL_GPL(macvlan_dellink);
1611 
1612 static int macvlan_changelink(struct net_device *dev,
1613 			      struct nlattr *tb[], struct nlattr *data[],
1614 			      struct netlink_ext_ack *extack)
1615 {
1616 	struct macvlan_dev *vlan = netdev_priv(dev);
1617 	enum macvlan_mode mode;
1618 	bool set_mode = false;
1619 	enum macvlan_macaddr_mode macmode;
1620 	int ret;
1621 
1622 	/* Validate mode, but don't set yet: setting flags may fail. */
1623 	if (data && data[IFLA_MACVLAN_MODE]) {
1624 		set_mode = true;
1625 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1626 		/* Passthrough mode can't be set or cleared dynamically */
1627 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1628 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1629 			return -EINVAL;
1630 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1631 		    vlan->mode != mode)
1632 			macvlan_flush_sources(vlan->port, vlan);
1633 	}
1634 
1635 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1636 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1637 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1638 		if (macvlan_passthru(vlan->port) && promisc) {
1639 			int err;
1640 
1641 			if (flags & MACVLAN_FLAG_NOPROMISC)
1642 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1643 			else
1644 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1645 			if (err < 0)
1646 				return err;
1647 		}
1648 		vlan->flags = flags;
1649 	}
1650 
1651 	if (data && data[IFLA_MACVLAN_BC_QUEUE_LEN]) {
1652 		vlan->bc_queue_len_req = nla_get_u32(data[IFLA_MACVLAN_BC_QUEUE_LEN]);
1653 		update_port_bc_queue_len(vlan->port);
1654 	}
1655 
1656 	if (data && data[IFLA_MACVLAN_BC_CUTOFF])
1657 		update_port_bc_cutoff(
1658 			vlan, nla_get_s32(data[IFLA_MACVLAN_BC_CUTOFF]));
1659 
1660 	if (set_mode)
1661 		vlan->mode = mode;
1662 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1663 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1664 			return -EINVAL;
1665 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1666 		ret = macvlan_changelink_sources(vlan, macmode, data);
1667 		if (ret)
1668 			return ret;
1669 	}
1670 	return 0;
1671 }
1672 
1673 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1674 {
1675 	if (vlan->macaddr_count == 0)
1676 		return 0;
1677 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1678 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1679 }
1680 
1681 static size_t macvlan_get_size(const struct net_device *dev)
1682 {
1683 	struct macvlan_dev *vlan = netdev_priv(dev);
1684 
1685 	return (0
1686 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1687 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1688 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1689 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1690 		+ nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN */
1691 		+ nla_total_size(4) /* IFLA_MACVLAN_BC_QUEUE_LEN_USED */
1692 		);
1693 }
1694 
1695 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1696 				     const struct macvlan_dev *vlan,
1697 				     const int i)
1698 {
1699 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1700 	struct macvlan_source_entry *entry;
1701 
1702 	hlist_for_each_entry_rcu(entry, h, hlist, lockdep_rtnl_is_held()) {
1703 		if (rcu_access_pointer(entry->vlan) != vlan)
1704 			continue;
1705 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1706 			return 1;
1707 	}
1708 	return 0;
1709 }
1710 
1711 static int macvlan_fill_info(struct sk_buff *skb,
1712 				const struct net_device *dev)
1713 {
1714 	struct macvlan_dev *vlan = netdev_priv(dev);
1715 	struct macvlan_port *port = vlan->port;
1716 	int i;
1717 	struct nlattr *nest;
1718 
1719 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1720 		goto nla_put_failure;
1721 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1722 		goto nla_put_failure;
1723 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1724 		goto nla_put_failure;
1725 	if (vlan->macaddr_count > 0) {
1726 		nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1727 		if (nest == NULL)
1728 			goto nla_put_failure;
1729 
1730 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1731 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1732 				goto nla_put_failure;
1733 		}
1734 		nla_nest_end(skb, nest);
1735 	}
1736 	if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN, vlan->bc_queue_len_req))
1737 		goto nla_put_failure;
1738 	if (nla_put_u32(skb, IFLA_MACVLAN_BC_QUEUE_LEN_USED,
1739 			READ_ONCE(port->bc_queue_len_used)))
1740 		goto nla_put_failure;
1741 	if (port->bc_cutoff != 1 &&
1742 	    nla_put_s32(skb, IFLA_MACVLAN_BC_CUTOFF, port->bc_cutoff))
1743 		goto nla_put_failure;
1744 	return 0;
1745 
1746 nla_put_failure:
1747 	return -EMSGSIZE;
1748 }
1749 
1750 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1751 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1752 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1753 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1754 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1755 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1756 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1757 	[IFLA_MACVLAN_BC_QUEUE_LEN] = { .type = NLA_U32 },
1758 	[IFLA_MACVLAN_BC_QUEUE_LEN_USED] = { .type = NLA_REJECT },
1759 	[IFLA_MACVLAN_BC_CUTOFF] = { .type = NLA_S32 },
1760 };
1761 
1762 int macvlan_link_register(struct rtnl_link_ops *ops)
1763 {
1764 	/* common fields */
1765 	ops->validate		= macvlan_validate;
1766 	ops->maxtype		= IFLA_MACVLAN_MAX;
1767 	ops->policy		= macvlan_policy;
1768 	ops->changelink		= macvlan_changelink;
1769 	ops->get_size		= macvlan_get_size;
1770 	ops->fill_info		= macvlan_fill_info;
1771 
1772 	return rtnl_link_register(ops);
1773 };
1774 EXPORT_SYMBOL_GPL(macvlan_link_register);
1775 
1776 static struct net *macvlan_get_link_net(const struct net_device *dev)
1777 {
1778 	return dev_net(macvlan_dev_real_dev(dev));
1779 }
1780 
1781 static struct rtnl_link_ops macvlan_link_ops = {
1782 	.kind		= "macvlan",
1783 	.setup		= macvlan_setup,
1784 	.newlink	= macvlan_newlink,
1785 	.dellink	= macvlan_dellink,
1786 	.get_link_net	= macvlan_get_link_net,
1787 	.priv_size      = sizeof(struct macvlan_dev),
1788 };
1789 
1790 static void update_port_bc_queue_len(struct macvlan_port *port)
1791 {
1792 	u32 max_bc_queue_len_req = 0;
1793 	struct macvlan_dev *vlan;
1794 
1795 	list_for_each_entry(vlan, &port->vlans, list) {
1796 		if (vlan->bc_queue_len_req > max_bc_queue_len_req)
1797 			max_bc_queue_len_req = vlan->bc_queue_len_req;
1798 	}
1799 	WRITE_ONCE(port->bc_queue_len_used, max_bc_queue_len_req);
1800 }
1801 
1802 static int macvlan_device_event(struct notifier_block *unused,
1803 				unsigned long event, void *ptr)
1804 {
1805 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1806 	struct macvlan_dev *vlan, *next;
1807 	struct macvlan_port *port;
1808 	LIST_HEAD(list_kill);
1809 
1810 	if (!netif_is_macvlan_port(dev))
1811 		return NOTIFY_DONE;
1812 
1813 	port = macvlan_port_get_rtnl(dev);
1814 
1815 	switch (event) {
1816 	case NETDEV_UP:
1817 	case NETDEV_DOWN:
1818 	case NETDEV_CHANGE:
1819 		list_for_each_entry(vlan, &port->vlans, list)
1820 			netif_stacked_transfer_operstate(vlan->lowerdev,
1821 							 vlan->dev);
1822 		break;
1823 	case NETDEV_FEAT_CHANGE:
1824 		list_for_each_entry(vlan, &port->vlans, list) {
1825 			netif_inherit_tso_max(vlan->dev, dev);
1826 			netdev_update_features(vlan->dev);
1827 		}
1828 		break;
1829 	case NETDEV_CHANGEMTU:
1830 		list_for_each_entry(vlan, &port->vlans, list) {
1831 			if (vlan->dev->mtu <= dev->mtu)
1832 				continue;
1833 			dev_set_mtu(vlan->dev, dev->mtu);
1834 		}
1835 		break;
1836 	case NETDEV_CHANGEADDR:
1837 		if (!macvlan_passthru(port))
1838 			return NOTIFY_DONE;
1839 
1840 		vlan = list_first_entry_or_null(&port->vlans,
1841 						struct macvlan_dev,
1842 						list);
1843 
1844 		if (vlan && macvlan_sync_address(vlan->dev, dev->dev_addr))
1845 			return NOTIFY_BAD;
1846 
1847 		break;
1848 	case NETDEV_UNREGISTER:
1849 		/* twiddle thumbs on netns device moves */
1850 		if (dev->reg_state != NETREG_UNREGISTERING)
1851 			break;
1852 
1853 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1854 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1855 		unregister_netdevice_many(&list_kill);
1856 		break;
1857 	case NETDEV_PRE_TYPE_CHANGE:
1858 		/* Forbid underlying device to change its type. */
1859 		return NOTIFY_BAD;
1860 
1861 	case NETDEV_NOTIFY_PEERS:
1862 	case NETDEV_BONDING_FAILOVER:
1863 	case NETDEV_RESEND_IGMP:
1864 		/* Propagate to all vlans */
1865 		list_for_each_entry(vlan, &port->vlans, list)
1866 			call_netdevice_notifiers(event, vlan->dev);
1867 	}
1868 	return NOTIFY_DONE;
1869 }
1870 
1871 static struct notifier_block macvlan_notifier_block __read_mostly = {
1872 	.notifier_call	= macvlan_device_event,
1873 };
1874 
1875 static int __init macvlan_init_module(void)
1876 {
1877 	int err;
1878 
1879 	register_netdevice_notifier(&macvlan_notifier_block);
1880 
1881 	err = macvlan_link_register(&macvlan_link_ops);
1882 	if (err < 0)
1883 		goto err1;
1884 	return 0;
1885 err1:
1886 	unregister_netdevice_notifier(&macvlan_notifier_block);
1887 	return err;
1888 }
1889 
1890 static void __exit macvlan_cleanup_module(void)
1891 {
1892 	rtnl_link_unregister(&macvlan_link_ops);
1893 	unregister_netdevice_notifier(&macvlan_notifier_block);
1894 }
1895 
1896 module_init(macvlan_init_module);
1897 module_exit(macvlan_cleanup_module);
1898 
1899 MODULE_LICENSE("GPL");
1900 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1901 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1902 MODULE_ALIAS_RTNL_LINK("macvlan");
1903