xref: /openbmc/linux/drivers/net/macvlan.c (revision b2765275)
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_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			flags;
50 	int			count;
51 	struct hlist_head	vlan_source_hash[MACVLAN_HASH_SIZE];
52 	DECLARE_BITMAP(mc_filter, MACVLAN_MC_FILTER_SZ);
53 	unsigned char           perm_addr[ETH_ALEN];
54 };
55 
56 struct macvlan_source_entry {
57 	struct hlist_node	hlist;
58 	struct macvlan_dev	*vlan;
59 	unsigned char		addr[6+2] __aligned(sizeof(u16));
60 	struct rcu_head		rcu;
61 };
62 
63 struct macvlan_skb_cb {
64 	const struct macvlan_dev *src;
65 };
66 
67 #define MACVLAN_SKB_CB(__skb) ((struct macvlan_skb_cb *)&((__skb)->cb[0]))
68 
69 static void macvlan_port_destroy(struct net_device *dev);
70 
71 static inline bool macvlan_passthru(const struct macvlan_port *port)
72 {
73 	return port->flags & MACVLAN_F_PASSTHRU;
74 }
75 
76 static inline void macvlan_set_passthru(struct macvlan_port *port)
77 {
78 	port->flags |= MACVLAN_F_PASSTHRU;
79 }
80 
81 static inline bool macvlan_addr_change(const struct macvlan_port *port)
82 {
83 	return port->flags & MACVLAN_F_ADDRCHANGE;
84 }
85 
86 static inline void macvlan_set_addr_change(struct macvlan_port *port)
87 {
88 	port->flags |= MACVLAN_F_ADDRCHANGE;
89 }
90 
91 static inline void macvlan_clear_addr_change(struct macvlan_port *port)
92 {
93 	port->flags &= ~MACVLAN_F_ADDRCHANGE;
94 }
95 
96 /* Hash Ethernet address */
97 static u32 macvlan_eth_hash(const unsigned char *addr)
98 {
99 	u64 value = get_unaligned((u64 *)addr);
100 
101 	/* only want 6 bytes */
102 #ifdef __BIG_ENDIAN
103 	value >>= 16;
104 #else
105 	value <<= 16;
106 #endif
107 	return hash_64(value, MACVLAN_HASH_BITS);
108 }
109 
110 static struct macvlan_port *macvlan_port_get_rcu(const struct net_device *dev)
111 {
112 	return rcu_dereference(dev->rx_handler_data);
113 }
114 
115 static struct macvlan_port *macvlan_port_get_rtnl(const struct net_device *dev)
116 {
117 	return rtnl_dereference(dev->rx_handler_data);
118 }
119 
120 static struct macvlan_dev *macvlan_hash_lookup(const struct macvlan_port *port,
121 					       const unsigned char *addr)
122 {
123 	struct macvlan_dev *vlan;
124 	u32 idx = macvlan_eth_hash(addr);
125 
126 	hlist_for_each_entry_rcu(vlan, &port->vlan_hash[idx], hlist) {
127 		if (ether_addr_equal_64bits(vlan->dev->dev_addr, addr))
128 			return vlan;
129 	}
130 	return NULL;
131 }
132 
133 static struct macvlan_source_entry *macvlan_hash_lookup_source(
134 	const struct macvlan_dev *vlan,
135 	const unsigned char *addr)
136 {
137 	struct macvlan_source_entry *entry;
138 	u32 idx = macvlan_eth_hash(addr);
139 	struct hlist_head *h = &vlan->port->vlan_source_hash[idx];
140 
141 	hlist_for_each_entry_rcu(entry, h, hlist) {
142 		if (ether_addr_equal_64bits(entry->addr, addr) &&
143 		    entry->vlan == vlan)
144 			return entry;
145 	}
146 	return NULL;
147 }
148 
149 static int macvlan_hash_add_source(struct macvlan_dev *vlan,
150 				   const unsigned char *addr)
151 {
152 	struct macvlan_port *port = vlan->port;
153 	struct macvlan_source_entry *entry;
154 	struct hlist_head *h;
155 
156 	entry = macvlan_hash_lookup_source(vlan, addr);
157 	if (entry)
158 		return 0;
159 
160 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
161 	if (!entry)
162 		return -ENOMEM;
163 
164 	ether_addr_copy(entry->addr, addr);
165 	entry->vlan = vlan;
166 	h = &port->vlan_source_hash[macvlan_eth_hash(addr)];
167 	hlist_add_head_rcu(&entry->hlist, h);
168 	vlan->macaddr_count++;
169 
170 	return 0;
171 }
172 
173 static void macvlan_hash_add(struct macvlan_dev *vlan)
174 {
175 	struct macvlan_port *port = vlan->port;
176 	const unsigned char *addr = vlan->dev->dev_addr;
177 	u32 idx = macvlan_eth_hash(addr);
178 
179 	hlist_add_head_rcu(&vlan->hlist, &port->vlan_hash[idx]);
180 }
181 
182 static void macvlan_hash_del_source(struct macvlan_source_entry *entry)
183 {
184 	hlist_del_rcu(&entry->hlist);
185 	kfree_rcu(entry, rcu);
186 }
187 
188 static void macvlan_hash_del(struct macvlan_dev *vlan, bool sync)
189 {
190 	hlist_del_rcu(&vlan->hlist);
191 	if (sync)
192 		synchronize_rcu();
193 }
194 
195 static void macvlan_hash_change_addr(struct macvlan_dev *vlan,
196 					const unsigned char *addr)
197 {
198 	macvlan_hash_del(vlan, true);
199 	/* Now that we are unhashed it is safe to change the device
200 	 * address without confusing packet delivery.
201 	 */
202 	memcpy(vlan->dev->dev_addr, addr, ETH_ALEN);
203 	macvlan_hash_add(vlan);
204 }
205 
206 static bool macvlan_addr_busy(const struct macvlan_port *port,
207 			      const unsigned char *addr)
208 {
209 	/* Test to see if the specified address is
210 	 * currently in use by the underlying device or
211 	 * another macvlan.
212 	 */
213 	if (!macvlan_passthru(port) && !macvlan_addr_change(port) &&
214 	    ether_addr_equal_64bits(port->dev->dev_addr, addr))
215 		return true;
216 
217 	if (macvlan_hash_lookup(port, addr))
218 		return true;
219 
220 	return false;
221 }
222 
223 
224 static int macvlan_broadcast_one(struct sk_buff *skb,
225 				 const struct macvlan_dev *vlan,
226 				 const struct ethhdr *eth, bool local)
227 {
228 	struct net_device *dev = vlan->dev;
229 
230 	if (local)
231 		return __dev_forward_skb(dev, skb);
232 
233 	skb->dev = dev;
234 	if (ether_addr_equal_64bits(eth->h_dest, dev->broadcast))
235 		skb->pkt_type = PACKET_BROADCAST;
236 	else
237 		skb->pkt_type = PACKET_MULTICAST;
238 
239 	return 0;
240 }
241 
242 static u32 macvlan_hash_mix(const struct macvlan_dev *vlan)
243 {
244 	return (u32)(((unsigned long)vlan) >> L1_CACHE_SHIFT);
245 }
246 
247 
248 static unsigned int mc_hash(const struct macvlan_dev *vlan,
249 			    const unsigned char *addr)
250 {
251 	u32 val = __get_unaligned_cpu32(addr + 2);
252 
253 	val ^= macvlan_hash_mix(vlan);
254 	return hash_32(val, MACVLAN_MC_FILTER_BITS);
255 }
256 
257 static void macvlan_broadcast(struct sk_buff *skb,
258 			      const struct macvlan_port *port,
259 			      struct net_device *src,
260 			      enum macvlan_mode mode)
261 {
262 	const struct ethhdr *eth = eth_hdr(skb);
263 	const struct macvlan_dev *vlan;
264 	struct sk_buff *nskb;
265 	unsigned int i;
266 	int err;
267 	unsigned int hash;
268 
269 	if (skb->protocol == htons(ETH_P_PAUSE))
270 		return;
271 
272 	for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
273 		hlist_for_each_entry_rcu(vlan, &port->vlan_hash[i], hlist) {
274 			if (vlan->dev == src || !(vlan->mode & mode))
275 				continue;
276 
277 			hash = mc_hash(vlan, eth->h_dest);
278 			if (!test_bit(hash, vlan->mc_filter))
279 				continue;
280 
281 			err = NET_RX_DROP;
282 			nskb = skb_clone(skb, GFP_ATOMIC);
283 			if (likely(nskb))
284 				err = macvlan_broadcast_one(
285 					nskb, vlan, eth,
286 					mode == MACVLAN_MODE_BRIDGE) ?:
287 				      netif_rx_ni(nskb);
288 			macvlan_count_rx(vlan, skb->len + ETH_HLEN,
289 					 err == NET_RX_SUCCESS, true);
290 		}
291 	}
292 }
293 
294 static void macvlan_process_broadcast(struct work_struct *w)
295 {
296 	struct macvlan_port *port = container_of(w, struct macvlan_port,
297 						 bc_work);
298 	struct sk_buff *skb;
299 	struct sk_buff_head list;
300 
301 	__skb_queue_head_init(&list);
302 
303 	spin_lock_bh(&port->bc_queue.lock);
304 	skb_queue_splice_tail_init(&port->bc_queue, &list);
305 	spin_unlock_bh(&port->bc_queue.lock);
306 
307 	while ((skb = __skb_dequeue(&list))) {
308 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
309 
310 		rcu_read_lock();
311 
312 		if (!src)
313 			/* frame comes from an external address */
314 			macvlan_broadcast(skb, port, NULL,
315 					  MACVLAN_MODE_PRIVATE |
316 					  MACVLAN_MODE_VEPA    |
317 					  MACVLAN_MODE_PASSTHRU|
318 					  MACVLAN_MODE_BRIDGE);
319 		else if (src->mode == MACVLAN_MODE_VEPA)
320 			/* flood to everyone except source */
321 			macvlan_broadcast(skb, port, src->dev,
322 					  MACVLAN_MODE_VEPA |
323 					  MACVLAN_MODE_BRIDGE);
324 		else
325 			/*
326 			 * flood only to VEPA ports, bridge ports
327 			 * already saw the frame on the way out.
328 			 */
329 			macvlan_broadcast(skb, port, src->dev,
330 					  MACVLAN_MODE_VEPA);
331 
332 		rcu_read_unlock();
333 
334 		if (src)
335 			dev_put(src->dev);
336 		consume_skb(skb);
337 
338 		cond_resched();
339 	}
340 }
341 
342 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
343 				      const struct macvlan_dev *src,
344 				      struct sk_buff *skb)
345 {
346 	struct sk_buff *nskb;
347 	int err = -ENOMEM;
348 
349 	nskb = skb_clone(skb, GFP_ATOMIC);
350 	if (!nskb)
351 		goto err;
352 
353 	MACVLAN_SKB_CB(nskb)->src = src;
354 
355 	spin_lock(&port->bc_queue.lock);
356 	if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) {
357 		if (src)
358 			dev_hold(src->dev);
359 		__skb_queue_tail(&port->bc_queue, nskb);
360 		err = 0;
361 	}
362 	spin_unlock(&port->bc_queue.lock);
363 
364 	schedule_work(&port->bc_work);
365 
366 	if (err)
367 		goto free_nskb;
368 
369 	return;
370 
371 free_nskb:
372 	kfree_skb(nskb);
373 err:
374 	atomic_long_inc(&skb->dev->rx_dropped);
375 }
376 
377 static void macvlan_flush_sources(struct macvlan_port *port,
378 				  struct macvlan_dev *vlan)
379 {
380 	int i;
381 
382 	for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
383 		struct hlist_node *h, *n;
384 
385 		hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) {
386 			struct macvlan_source_entry *entry;
387 
388 			entry = hlist_entry(h, struct macvlan_source_entry,
389 					    hlist);
390 			if (entry->vlan == vlan)
391 				macvlan_hash_del_source(entry);
392 		}
393 	}
394 	vlan->macaddr_count = 0;
395 }
396 
397 static void macvlan_forward_source_one(struct sk_buff *skb,
398 				       struct macvlan_dev *vlan)
399 {
400 	struct sk_buff *nskb;
401 	struct net_device *dev;
402 	int len;
403 	int ret;
404 
405 	dev = vlan->dev;
406 	if (unlikely(!(dev->flags & IFF_UP)))
407 		return;
408 
409 	nskb = skb_clone(skb, GFP_ATOMIC);
410 	if (!nskb)
411 		return;
412 
413 	len = nskb->len + ETH_HLEN;
414 	nskb->dev = dev;
415 
416 	if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
417 		nskb->pkt_type = PACKET_HOST;
418 
419 	ret = netif_rx(nskb);
420 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
421 }
422 
423 static void macvlan_forward_source(struct sk_buff *skb,
424 				   struct macvlan_port *port,
425 				   const unsigned char *addr)
426 {
427 	struct macvlan_source_entry *entry;
428 	u32 idx = macvlan_eth_hash(addr);
429 	struct hlist_head *h = &port->vlan_source_hash[idx];
430 
431 	hlist_for_each_entry_rcu(entry, h, hlist) {
432 		if (ether_addr_equal_64bits(entry->addr, addr))
433 			macvlan_forward_source_one(skb, entry->vlan);
434 	}
435 }
436 
437 /* called under rcu_read_lock() from netif_receive_skb */
438 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
439 {
440 	struct macvlan_port *port;
441 	struct sk_buff *skb = *pskb;
442 	const struct ethhdr *eth = eth_hdr(skb);
443 	const struct macvlan_dev *vlan;
444 	const struct macvlan_dev *src;
445 	struct net_device *dev;
446 	unsigned int len = 0;
447 	int ret;
448 	rx_handler_result_t handle_res;
449 
450 	port = macvlan_port_get_rcu(skb->dev);
451 	if (is_multicast_ether_addr(eth->h_dest)) {
452 		unsigned int hash;
453 
454 		skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
455 		if (!skb)
456 			return RX_HANDLER_CONSUMED;
457 		*pskb = skb;
458 		eth = eth_hdr(skb);
459 		macvlan_forward_source(skb, port, eth->h_source);
460 		src = macvlan_hash_lookup(port, eth->h_source);
461 		if (src && src->mode != MACVLAN_MODE_VEPA &&
462 		    src->mode != MACVLAN_MODE_BRIDGE) {
463 			/* forward to original port. */
464 			vlan = src;
465 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
466 			      netif_rx(skb);
467 			handle_res = RX_HANDLER_CONSUMED;
468 			goto out;
469 		}
470 
471 		hash = mc_hash(NULL, eth->h_dest);
472 		if (test_bit(hash, port->mc_filter))
473 			macvlan_broadcast_enqueue(port, src, skb);
474 
475 		return RX_HANDLER_PASS;
476 	}
477 
478 	macvlan_forward_source(skb, port, eth->h_source);
479 	if (macvlan_passthru(port))
480 		vlan = list_first_or_null_rcu(&port->vlans,
481 					      struct macvlan_dev, list);
482 	else
483 		vlan = macvlan_hash_lookup(port, eth->h_dest);
484 	if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
485 		return RX_HANDLER_PASS;
486 
487 	dev = vlan->dev;
488 	if (unlikely(!(dev->flags & IFF_UP))) {
489 		kfree_skb(skb);
490 		return RX_HANDLER_CONSUMED;
491 	}
492 	len = skb->len + ETH_HLEN;
493 	skb = skb_share_check(skb, GFP_ATOMIC);
494 	if (!skb) {
495 		ret = NET_RX_DROP;
496 		handle_res = RX_HANDLER_CONSUMED;
497 		goto out;
498 	}
499 
500 	*pskb = skb;
501 	skb->dev = dev;
502 	skb->pkt_type = PACKET_HOST;
503 
504 	ret = NET_RX_SUCCESS;
505 	handle_res = RX_HANDLER_ANOTHER;
506 out:
507 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
508 	return handle_res;
509 }
510 
511 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
512 {
513 	const struct macvlan_dev *vlan = netdev_priv(dev);
514 	const struct macvlan_port *port = vlan->port;
515 	const struct macvlan_dev *dest;
516 
517 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
518 		const struct ethhdr *eth = skb_eth_hdr(skb);
519 
520 		/* send to other bridge ports directly */
521 		if (is_multicast_ether_addr(eth->h_dest)) {
522 			skb_reset_mac_header(skb);
523 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
524 			goto xmit_world;
525 		}
526 
527 		dest = macvlan_hash_lookup(port, eth->h_dest);
528 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
529 			/* send to lowerdev first for its network taps */
530 			dev_forward_skb(vlan->lowerdev, skb);
531 
532 			return NET_XMIT_SUCCESS;
533 		}
534 	}
535 xmit_world:
536 	skb->dev = vlan->lowerdev;
537 	return dev_queue_xmit_accel(skb,
538 				    netdev_get_sb_channel(dev) ? dev : NULL);
539 }
540 
541 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
542 {
543 #ifdef CONFIG_NET_POLL_CONTROLLER
544 	if (vlan->netpoll)
545 		netpoll_send_skb(vlan->netpoll, skb);
546 #else
547 	BUG();
548 #endif
549 	return NETDEV_TX_OK;
550 }
551 
552 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
553 				      struct net_device *dev)
554 {
555 	struct macvlan_dev *vlan = netdev_priv(dev);
556 	unsigned int len = skb->len;
557 	int ret;
558 
559 	if (unlikely(netpoll_tx_running(dev)))
560 		return macvlan_netpoll_send_skb(vlan, skb);
561 
562 	ret = macvlan_queue_xmit(skb, dev);
563 
564 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
565 		struct vlan_pcpu_stats *pcpu_stats;
566 
567 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
568 		u64_stats_update_begin(&pcpu_stats->syncp);
569 		pcpu_stats->tx_packets++;
570 		pcpu_stats->tx_bytes += len;
571 		u64_stats_update_end(&pcpu_stats->syncp);
572 	} else {
573 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
574 	}
575 	return ret;
576 }
577 
578 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
579 			       unsigned short type, const void *daddr,
580 			       const void *saddr, unsigned len)
581 {
582 	const struct macvlan_dev *vlan = netdev_priv(dev);
583 	struct net_device *lowerdev = vlan->lowerdev;
584 
585 	return dev_hard_header(skb, lowerdev, type, daddr,
586 			       saddr ? : dev->dev_addr, len);
587 }
588 
589 static const struct header_ops macvlan_hard_header_ops = {
590 	.create  	= macvlan_hard_header,
591 	.parse		= eth_header_parse,
592 	.cache		= eth_header_cache,
593 	.cache_update	= eth_header_cache_update,
594 };
595 
596 static int macvlan_open(struct net_device *dev)
597 {
598 	struct macvlan_dev *vlan = netdev_priv(dev);
599 	struct net_device *lowerdev = vlan->lowerdev;
600 	int err;
601 
602 	if (macvlan_passthru(vlan->port)) {
603 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
604 			err = dev_set_promiscuity(lowerdev, 1);
605 			if (err < 0)
606 				goto out;
607 		}
608 		goto hash_add;
609 	}
610 
611 	err = -EADDRINUSE;
612 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
613 		goto out;
614 
615 	/* Attempt to populate accel_priv which is used to offload the L2
616 	 * forwarding requests for unicast packets.
617 	 */
618 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
619 		vlan->accel_priv =
620 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
621 
622 	/* If earlier attempt to offload failed, or accel_priv is not
623 	 * populated we must add the unicast address to the lower device.
624 	 */
625 	if (IS_ERR_OR_NULL(vlan->accel_priv)) {
626 		vlan->accel_priv = NULL;
627 		err = dev_uc_add(lowerdev, dev->dev_addr);
628 		if (err < 0)
629 			goto out;
630 	}
631 
632 	if (dev->flags & IFF_ALLMULTI) {
633 		err = dev_set_allmulti(lowerdev, 1);
634 		if (err < 0)
635 			goto del_unicast;
636 	}
637 
638 	if (dev->flags & IFF_PROMISC) {
639 		err = dev_set_promiscuity(lowerdev, 1);
640 		if (err < 0)
641 			goto clear_multi;
642 	}
643 
644 hash_add:
645 	macvlan_hash_add(vlan);
646 	return 0;
647 
648 clear_multi:
649 	if (dev->flags & IFF_ALLMULTI)
650 		dev_set_allmulti(lowerdev, -1);
651 del_unicast:
652 	if (vlan->accel_priv) {
653 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
654 							   vlan->accel_priv);
655 		vlan->accel_priv = NULL;
656 	} else {
657 		dev_uc_del(lowerdev, dev->dev_addr);
658 	}
659 out:
660 	return err;
661 }
662 
663 static int macvlan_stop(struct net_device *dev)
664 {
665 	struct macvlan_dev *vlan = netdev_priv(dev);
666 	struct net_device *lowerdev = vlan->lowerdev;
667 
668 	if (vlan->accel_priv) {
669 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
670 							   vlan->accel_priv);
671 		vlan->accel_priv = NULL;
672 	}
673 
674 	dev_uc_unsync(lowerdev, dev);
675 	dev_mc_unsync(lowerdev, dev);
676 
677 	if (macvlan_passthru(vlan->port)) {
678 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
679 			dev_set_promiscuity(lowerdev, -1);
680 		goto hash_del;
681 	}
682 
683 	if (dev->flags & IFF_ALLMULTI)
684 		dev_set_allmulti(lowerdev, -1);
685 
686 	if (dev->flags & IFF_PROMISC)
687 		dev_set_promiscuity(lowerdev, -1);
688 
689 	dev_uc_del(lowerdev, dev->dev_addr);
690 
691 hash_del:
692 	macvlan_hash_del(vlan, !dev->dismantle);
693 	return 0;
694 }
695 
696 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
697 {
698 	struct macvlan_dev *vlan = netdev_priv(dev);
699 	struct net_device *lowerdev = vlan->lowerdev;
700 	struct macvlan_port *port = vlan->port;
701 	int err;
702 
703 	if (!(dev->flags & IFF_UP)) {
704 		/* Just copy in the new address */
705 		ether_addr_copy(dev->dev_addr, addr);
706 	} else {
707 		/* Rehash and update the device filters */
708 		if (macvlan_addr_busy(vlan->port, addr))
709 			return -EADDRINUSE;
710 
711 		if (!macvlan_passthru(port)) {
712 			err = dev_uc_add(lowerdev, addr);
713 			if (err)
714 				return err;
715 
716 			dev_uc_del(lowerdev, dev->dev_addr);
717 		}
718 
719 		macvlan_hash_change_addr(vlan, addr);
720 	}
721 	if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
722 		/* Since addr_change isn't set, we are here due to lower
723 		 * device change.  Save the lower-dev address so we can
724 		 * restore it later.
725 		 */
726 		ether_addr_copy(vlan->port->perm_addr,
727 				lowerdev->dev_addr);
728 	}
729 	macvlan_clear_addr_change(port);
730 	return 0;
731 }
732 
733 static int macvlan_set_mac_address(struct net_device *dev, void *p)
734 {
735 	struct macvlan_dev *vlan = netdev_priv(dev);
736 	struct sockaddr *addr = p;
737 
738 	if (!is_valid_ether_addr(addr->sa_data))
739 		return -EADDRNOTAVAIL;
740 
741 	/* If the addresses are the same, this is a no-op */
742 	if (ether_addr_equal(dev->dev_addr, addr->sa_data))
743 		return 0;
744 
745 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
746 		macvlan_set_addr_change(vlan->port);
747 		return dev_set_mac_address(vlan->lowerdev, addr, NULL);
748 	}
749 
750 	if (macvlan_addr_busy(vlan->port, addr->sa_data))
751 		return -EADDRINUSE;
752 
753 	return macvlan_sync_address(dev, addr->sa_data);
754 }
755 
756 static void macvlan_change_rx_flags(struct net_device *dev, int change)
757 {
758 	struct macvlan_dev *vlan = netdev_priv(dev);
759 	struct net_device *lowerdev = vlan->lowerdev;
760 
761 	if (dev->flags & IFF_UP) {
762 		if (change & IFF_ALLMULTI)
763 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
764 		if (change & IFF_PROMISC)
765 			dev_set_promiscuity(lowerdev,
766 					    dev->flags & IFF_PROMISC ? 1 : -1);
767 
768 	}
769 }
770 
771 static void macvlan_compute_filter(unsigned long *mc_filter,
772 				   struct net_device *dev,
773 				   struct macvlan_dev *vlan)
774 {
775 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
776 		bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
777 	} else {
778 		struct netdev_hw_addr *ha;
779 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
780 
781 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
782 		netdev_for_each_mc_addr(ha, dev) {
783 			__set_bit(mc_hash(vlan, ha->addr), filter);
784 		}
785 
786 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
787 
788 		bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
789 	}
790 }
791 
792 static void macvlan_set_mac_lists(struct net_device *dev)
793 {
794 	struct macvlan_dev *vlan = netdev_priv(dev);
795 
796 	macvlan_compute_filter(vlan->mc_filter, dev, vlan);
797 
798 	dev_uc_sync(vlan->lowerdev, dev);
799 	dev_mc_sync(vlan->lowerdev, dev);
800 
801 	/* This is slightly inaccurate as we're including the subscription
802 	 * list of vlan->lowerdev too.
803 	 *
804 	 * Bug alert: This only works if everyone has the same broadcast
805 	 * address as lowerdev.  As soon as someone changes theirs this
806 	 * will break.
807 	 *
808 	 * However, this is already broken as when you change your broadcast
809 	 * address we don't get called.
810 	 *
811 	 * The solution is to maintain a list of broadcast addresses like
812 	 * we do for uc/mc, if you care.
813 	 */
814 	macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
815 }
816 
817 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
818 {
819 	struct macvlan_dev *vlan = netdev_priv(dev);
820 
821 	if (vlan->lowerdev->mtu < new_mtu)
822 		return -EINVAL;
823 	dev->mtu = new_mtu;
824 	return 0;
825 }
826 
827 static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
828 {
829 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
830 	const struct net_device_ops *ops = real_dev->netdev_ops;
831 	struct ifreq ifrr;
832 	int err = -EOPNOTSUPP;
833 
834 	strscpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
835 	ifrr.ifr_ifru = ifr->ifr_ifru;
836 
837 	switch (cmd) {
838 	case SIOCSHWTSTAMP:
839 		if (!net_eq(dev_net(dev), &init_net))
840 			break;
841 		/* fall through */
842 	case SIOCGHWTSTAMP:
843 		if (netif_device_present(real_dev) && ops->ndo_do_ioctl)
844 			err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd);
845 		break;
846 	}
847 
848 	if (!err)
849 		ifr->ifr_ifru = ifrr.ifr_ifru;
850 
851 	return err;
852 }
853 
854 /*
855  * macvlan network devices have devices nesting below it and are a special
856  * "super class" of normal network devices; split their locks off into a
857  * separate class since they always nest.
858  */
859 #define ALWAYS_ON_OFFLOADS \
860 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
861 	 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
862 
863 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
864 
865 #define MACVLAN_FEATURES \
866 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
867 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
868 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
869 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
870 
871 #define MACVLAN_STATE_MASK \
872 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
873 
874 static int macvlan_init(struct net_device *dev)
875 {
876 	struct macvlan_dev *vlan = netdev_priv(dev);
877 	const struct net_device *lowerdev = vlan->lowerdev;
878 	struct macvlan_port *port = vlan->port;
879 
880 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
881 				  (lowerdev->state & MACVLAN_STATE_MASK);
882 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
883 	dev->features		|= ALWAYS_ON_FEATURES;
884 	dev->hw_features	|= NETIF_F_LRO;
885 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
886 	dev->vlan_features	|= ALWAYS_ON_OFFLOADS;
887 	dev->hw_enc_features    |= dev->features;
888 	dev->gso_max_size	= lowerdev->gso_max_size;
889 	dev->gso_max_segs	= lowerdev->gso_max_segs;
890 	dev->hard_header_len	= lowerdev->hard_header_len;
891 
892 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
893 	if (!vlan->pcpu_stats)
894 		return -ENOMEM;
895 
896 	port->count += 1;
897 
898 	return 0;
899 }
900 
901 static void macvlan_uninit(struct net_device *dev)
902 {
903 	struct macvlan_dev *vlan = netdev_priv(dev);
904 	struct macvlan_port *port = vlan->port;
905 
906 	free_percpu(vlan->pcpu_stats);
907 
908 	macvlan_flush_sources(port, vlan);
909 	port->count -= 1;
910 	if (!port->count)
911 		macvlan_port_destroy(port->dev);
912 }
913 
914 static void macvlan_dev_get_stats64(struct net_device *dev,
915 				    struct rtnl_link_stats64 *stats)
916 {
917 	struct macvlan_dev *vlan = netdev_priv(dev);
918 
919 	if (vlan->pcpu_stats) {
920 		struct vlan_pcpu_stats *p;
921 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
922 		u32 rx_errors = 0, tx_dropped = 0;
923 		unsigned int start;
924 		int i;
925 
926 		for_each_possible_cpu(i) {
927 			p = per_cpu_ptr(vlan->pcpu_stats, i);
928 			do {
929 				start = u64_stats_fetch_begin_irq(&p->syncp);
930 				rx_packets	= p->rx_packets;
931 				rx_bytes	= p->rx_bytes;
932 				rx_multicast	= p->rx_multicast;
933 				tx_packets	= p->tx_packets;
934 				tx_bytes	= p->tx_bytes;
935 			} while (u64_stats_fetch_retry_irq(&p->syncp, start));
936 
937 			stats->rx_packets	+= rx_packets;
938 			stats->rx_bytes		+= rx_bytes;
939 			stats->multicast	+= rx_multicast;
940 			stats->tx_packets	+= tx_packets;
941 			stats->tx_bytes		+= tx_bytes;
942 			/* rx_errors & tx_dropped are u32, updated
943 			 * without syncp protection.
944 			 */
945 			rx_errors	+= p->rx_errors;
946 			tx_dropped	+= p->tx_dropped;
947 		}
948 		stats->rx_errors	= rx_errors;
949 		stats->rx_dropped	= rx_errors;
950 		stats->tx_dropped	= tx_dropped;
951 	}
952 }
953 
954 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
955 				   __be16 proto, u16 vid)
956 {
957 	struct macvlan_dev *vlan = netdev_priv(dev);
958 	struct net_device *lowerdev = vlan->lowerdev;
959 
960 	return vlan_vid_add(lowerdev, proto, vid);
961 }
962 
963 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
964 				    __be16 proto, u16 vid)
965 {
966 	struct macvlan_dev *vlan = netdev_priv(dev);
967 	struct net_device *lowerdev = vlan->lowerdev;
968 
969 	vlan_vid_del(lowerdev, proto, vid);
970 	return 0;
971 }
972 
973 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
974 			   struct net_device *dev,
975 			   const unsigned char *addr, u16 vid,
976 			   u16 flags,
977 			   struct netlink_ext_ack *extack)
978 {
979 	struct macvlan_dev *vlan = netdev_priv(dev);
980 	int err = -EINVAL;
981 
982 	/* Support unicast filter only on passthru devices.
983 	 * Multicast filter should be allowed on all devices.
984 	 */
985 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
986 		return -EOPNOTSUPP;
987 
988 	if (flags & NLM_F_REPLACE)
989 		return -EOPNOTSUPP;
990 
991 	if (is_unicast_ether_addr(addr))
992 		err = dev_uc_add_excl(dev, addr);
993 	else if (is_multicast_ether_addr(addr))
994 		err = dev_mc_add_excl(dev, addr);
995 
996 	return err;
997 }
998 
999 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1000 			   struct net_device *dev,
1001 			   const unsigned char *addr, u16 vid)
1002 {
1003 	struct macvlan_dev *vlan = netdev_priv(dev);
1004 	int err = -EINVAL;
1005 
1006 	/* Support unicast filter only on passthru devices.
1007 	 * Multicast filter should be allowed on all devices.
1008 	 */
1009 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1010 		return -EOPNOTSUPP;
1011 
1012 	if (is_unicast_ether_addr(addr))
1013 		err = dev_uc_del(dev, addr);
1014 	else if (is_multicast_ether_addr(addr))
1015 		err = dev_mc_del(dev, addr);
1016 
1017 	return err;
1018 }
1019 
1020 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1021 					struct ethtool_drvinfo *drvinfo)
1022 {
1023 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1024 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1025 }
1026 
1027 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1028 					      struct ethtool_link_ksettings *cmd)
1029 {
1030 	const struct macvlan_dev *vlan = netdev_priv(dev);
1031 
1032 	return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1033 }
1034 
1035 static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1036 				       struct ethtool_ts_info *info)
1037 {
1038 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
1039 	const struct ethtool_ops *ops = real_dev->ethtool_ops;
1040 	struct phy_device *phydev = real_dev->phydev;
1041 
1042 	if (phy_has_tsinfo(phydev)) {
1043 		return phy_ts_info(phydev, info);
1044 	} else if (ops->get_ts_info) {
1045 		return ops->get_ts_info(real_dev, info);
1046 	} else {
1047 		info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1048 			SOF_TIMESTAMPING_SOFTWARE;
1049 		info->phc_index = -1;
1050 	}
1051 
1052 	return 0;
1053 }
1054 
1055 static netdev_features_t macvlan_fix_features(struct net_device *dev,
1056 					      netdev_features_t features)
1057 {
1058 	struct macvlan_dev *vlan = netdev_priv(dev);
1059 	netdev_features_t lowerdev_features = vlan->lowerdev->features;
1060 	netdev_features_t mask;
1061 
1062 	features |= NETIF_F_ALL_FOR_ALL;
1063 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
1064 	mask = features;
1065 
1066 	lowerdev_features &= (features | ~NETIF_F_LRO);
1067 	features = netdev_increment_features(lowerdev_features, features, mask);
1068 	features |= ALWAYS_ON_FEATURES;
1069 	features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1070 
1071 	return features;
1072 }
1073 
1074 #ifdef CONFIG_NET_POLL_CONTROLLER
1075 static void macvlan_dev_poll_controller(struct net_device *dev)
1076 {
1077 	return;
1078 }
1079 
1080 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1081 {
1082 	struct macvlan_dev *vlan = netdev_priv(dev);
1083 	struct net_device *real_dev = vlan->lowerdev;
1084 	struct netpoll *netpoll;
1085 	int err = 0;
1086 
1087 	netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1088 	err = -ENOMEM;
1089 	if (!netpoll)
1090 		goto out;
1091 
1092 	err = __netpoll_setup(netpoll, real_dev);
1093 	if (err) {
1094 		kfree(netpoll);
1095 		goto out;
1096 	}
1097 
1098 	vlan->netpoll = netpoll;
1099 
1100 out:
1101 	return err;
1102 }
1103 
1104 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1105 {
1106 	struct macvlan_dev *vlan = netdev_priv(dev);
1107 	struct netpoll *netpoll = vlan->netpoll;
1108 
1109 	if (!netpoll)
1110 		return;
1111 
1112 	vlan->netpoll = NULL;
1113 
1114 	__netpoll_free(netpoll);
1115 }
1116 #endif	/* CONFIG_NET_POLL_CONTROLLER */
1117 
1118 static int macvlan_dev_get_iflink(const struct net_device *dev)
1119 {
1120 	struct macvlan_dev *vlan = netdev_priv(dev);
1121 
1122 	return vlan->lowerdev->ifindex;
1123 }
1124 
1125 static const struct ethtool_ops macvlan_ethtool_ops = {
1126 	.get_link		= ethtool_op_get_link,
1127 	.get_link_ksettings	= macvlan_ethtool_get_link_ksettings,
1128 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
1129 	.get_ts_info		= macvlan_ethtool_get_ts_info,
1130 };
1131 
1132 static const struct net_device_ops macvlan_netdev_ops = {
1133 	.ndo_init		= macvlan_init,
1134 	.ndo_uninit		= macvlan_uninit,
1135 	.ndo_open		= macvlan_open,
1136 	.ndo_stop		= macvlan_stop,
1137 	.ndo_start_xmit		= macvlan_start_xmit,
1138 	.ndo_change_mtu		= macvlan_change_mtu,
1139 	.ndo_do_ioctl		= macvlan_do_ioctl,
1140 	.ndo_fix_features	= macvlan_fix_features,
1141 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1142 	.ndo_set_mac_address	= macvlan_set_mac_address,
1143 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1144 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1145 	.ndo_validate_addr	= eth_validate_addr,
1146 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1147 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1148 	.ndo_fdb_add		= macvlan_fdb_add,
1149 	.ndo_fdb_del		= macvlan_fdb_del,
1150 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1151 #ifdef CONFIG_NET_POLL_CONTROLLER
1152 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1153 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1154 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1155 #endif
1156 	.ndo_get_iflink		= macvlan_dev_get_iflink,
1157 	.ndo_features_check	= passthru_features_check,
1158 	.ndo_change_proto_down  = dev_change_proto_down_generic,
1159 };
1160 
1161 void macvlan_common_setup(struct net_device *dev)
1162 {
1163 	ether_setup(dev);
1164 
1165 	dev->min_mtu		= 0;
1166 	dev->max_mtu		= ETH_MAX_MTU;
1167 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1168 	netif_keep_dst(dev);
1169 	dev->priv_flags	       |= IFF_UNICAST_FLT;
1170 	dev->netdev_ops		= &macvlan_netdev_ops;
1171 	dev->needs_free_netdev	= true;
1172 	dev->header_ops		= &macvlan_hard_header_ops;
1173 	dev->ethtool_ops	= &macvlan_ethtool_ops;
1174 }
1175 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1176 
1177 static void macvlan_setup(struct net_device *dev)
1178 {
1179 	macvlan_common_setup(dev);
1180 	dev->priv_flags |= IFF_NO_QUEUE;
1181 }
1182 
1183 static int macvlan_port_create(struct net_device *dev)
1184 {
1185 	struct macvlan_port *port;
1186 	unsigned int i;
1187 	int err;
1188 
1189 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1190 		return -EINVAL;
1191 
1192 	if (netdev_is_rx_handler_busy(dev))
1193 		return -EBUSY;
1194 
1195 	port = kzalloc(sizeof(*port), GFP_KERNEL);
1196 	if (port == NULL)
1197 		return -ENOMEM;
1198 
1199 	port->dev = dev;
1200 	ether_addr_copy(port->perm_addr, dev->dev_addr);
1201 	INIT_LIST_HEAD(&port->vlans);
1202 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1203 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
1204 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1205 		INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1206 
1207 	skb_queue_head_init(&port->bc_queue);
1208 	INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1209 
1210 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1211 	if (err)
1212 		kfree(port);
1213 	else
1214 		dev->priv_flags |= IFF_MACVLAN_PORT;
1215 	return err;
1216 }
1217 
1218 static void macvlan_port_destroy(struct net_device *dev)
1219 {
1220 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1221 	struct sk_buff *skb;
1222 
1223 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
1224 	netdev_rx_handler_unregister(dev);
1225 
1226 	/* After this point, no packet can schedule bc_work anymore,
1227 	 * but we need to cancel it and purge left skbs if any.
1228 	 */
1229 	cancel_work_sync(&port->bc_work);
1230 
1231 	while ((skb = __skb_dequeue(&port->bc_queue))) {
1232 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1233 
1234 		if (src)
1235 			dev_put(src->dev);
1236 
1237 		kfree_skb(skb);
1238 	}
1239 
1240 	/* If the lower device address has been changed by passthru
1241 	 * macvlan, put it back.
1242 	 */
1243 	if (macvlan_passthru(port) &&
1244 	    !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1245 		struct sockaddr sa;
1246 
1247 		sa.sa_family = port->dev->type;
1248 		memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1249 		dev_set_mac_address(port->dev, &sa, NULL);
1250 	}
1251 
1252 	kfree(port);
1253 }
1254 
1255 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1256 			    struct netlink_ext_ack *extack)
1257 {
1258 	if (tb[IFLA_ADDRESS]) {
1259 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1260 			return -EINVAL;
1261 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1262 			return -EADDRNOTAVAIL;
1263 	}
1264 
1265 	if (!data)
1266 		return 0;
1267 
1268 	if (data[IFLA_MACVLAN_FLAGS] &&
1269 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1270 		return -EINVAL;
1271 
1272 	if (data[IFLA_MACVLAN_MODE]) {
1273 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1274 		case MACVLAN_MODE_PRIVATE:
1275 		case MACVLAN_MODE_VEPA:
1276 		case MACVLAN_MODE_BRIDGE:
1277 		case MACVLAN_MODE_PASSTHRU:
1278 		case MACVLAN_MODE_SOURCE:
1279 			break;
1280 		default:
1281 			return -EINVAL;
1282 		}
1283 	}
1284 
1285 	if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1286 		switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1287 		case MACVLAN_MACADDR_ADD:
1288 		case MACVLAN_MACADDR_DEL:
1289 		case MACVLAN_MACADDR_FLUSH:
1290 		case MACVLAN_MACADDR_SET:
1291 			break;
1292 		default:
1293 			return -EINVAL;
1294 		}
1295 	}
1296 
1297 	if (data[IFLA_MACVLAN_MACADDR]) {
1298 		if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1299 			return -EINVAL;
1300 
1301 		if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1302 			return -EADDRNOTAVAIL;
1303 	}
1304 
1305 	if (data[IFLA_MACVLAN_MACADDR_COUNT])
1306 		return -EINVAL;
1307 
1308 	return 0;
1309 }
1310 
1311 /**
1312  * reconfigure list of remote source mac address
1313  * (only for macvlan devices in source mode)
1314  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1315  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1316  */
1317 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1318 				      struct nlattr *data[])
1319 {
1320 	char *addr = NULL;
1321 	int ret, rem, len;
1322 	struct nlattr *nla, *head;
1323 	struct macvlan_source_entry *entry;
1324 
1325 	if (data[IFLA_MACVLAN_MACADDR])
1326 		addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1327 
1328 	if (mode == MACVLAN_MACADDR_ADD) {
1329 		if (!addr)
1330 			return -EINVAL;
1331 
1332 		return macvlan_hash_add_source(vlan, addr);
1333 
1334 	} else if (mode == MACVLAN_MACADDR_DEL) {
1335 		if (!addr)
1336 			return -EINVAL;
1337 
1338 		entry = macvlan_hash_lookup_source(vlan, addr);
1339 		if (entry) {
1340 			macvlan_hash_del_source(entry);
1341 			vlan->macaddr_count--;
1342 		}
1343 	} else if (mode == MACVLAN_MACADDR_FLUSH) {
1344 		macvlan_flush_sources(vlan->port, vlan);
1345 	} else if (mode == MACVLAN_MACADDR_SET) {
1346 		macvlan_flush_sources(vlan->port, vlan);
1347 
1348 		if (addr) {
1349 			ret = macvlan_hash_add_source(vlan, addr);
1350 			if (ret)
1351 				return ret;
1352 		}
1353 
1354 		if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1355 			return 0;
1356 
1357 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1358 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1359 
1360 		nla_for_each_attr(nla, head, len, rem) {
1361 			if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1362 			    nla_len(nla) != ETH_ALEN)
1363 				continue;
1364 
1365 			addr = nla_data(nla);
1366 			ret = macvlan_hash_add_source(vlan, addr);
1367 			if (ret)
1368 				return ret;
1369 		}
1370 	} else {
1371 		return -EINVAL;
1372 	}
1373 
1374 	return 0;
1375 }
1376 
1377 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1378 			   struct nlattr *tb[], struct nlattr *data[],
1379 			   struct netlink_ext_ack *extack)
1380 {
1381 	struct macvlan_dev *vlan = netdev_priv(dev);
1382 	struct macvlan_port *port;
1383 	struct net_device *lowerdev;
1384 	int err;
1385 	int macmode;
1386 	bool create = false;
1387 
1388 	if (!tb[IFLA_LINK])
1389 		return -EINVAL;
1390 
1391 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1392 	if (lowerdev == NULL)
1393 		return -ENODEV;
1394 
1395 	/* When creating macvlans or macvtaps on top of other macvlans - use
1396 	 * the real device as the lowerdev.
1397 	 */
1398 	if (netif_is_macvlan(lowerdev))
1399 		lowerdev = macvlan_dev_real_dev(lowerdev);
1400 
1401 	if (!tb[IFLA_MTU])
1402 		dev->mtu = lowerdev->mtu;
1403 	else if (dev->mtu > lowerdev->mtu)
1404 		return -EINVAL;
1405 
1406 	/* MTU range: 68 - lowerdev->max_mtu */
1407 	dev->min_mtu = ETH_MIN_MTU;
1408 	dev->max_mtu = lowerdev->max_mtu;
1409 
1410 	if (!tb[IFLA_ADDRESS])
1411 		eth_hw_addr_random(dev);
1412 
1413 	if (!netif_is_macvlan_port(lowerdev)) {
1414 		err = macvlan_port_create(lowerdev);
1415 		if (err < 0)
1416 			return err;
1417 		create = true;
1418 	}
1419 	port = macvlan_port_get_rtnl(lowerdev);
1420 
1421 	/* Only 1 macvlan device can be created in passthru mode */
1422 	if (macvlan_passthru(port)) {
1423 		/* The macvlan port must be not created this time,
1424 		 * still goto destroy_macvlan_port for readability.
1425 		 */
1426 		err = -EINVAL;
1427 		goto destroy_macvlan_port;
1428 	}
1429 
1430 	vlan->lowerdev = lowerdev;
1431 	vlan->dev      = dev;
1432 	vlan->port     = port;
1433 	vlan->set_features = MACVLAN_FEATURES;
1434 
1435 	vlan->mode     = MACVLAN_MODE_VEPA;
1436 	if (data && data[IFLA_MACVLAN_MODE])
1437 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1438 
1439 	if (data && data[IFLA_MACVLAN_FLAGS])
1440 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1441 
1442 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1443 		if (port->count) {
1444 			err = -EINVAL;
1445 			goto destroy_macvlan_port;
1446 		}
1447 		macvlan_set_passthru(port);
1448 		eth_hw_addr_inherit(dev, lowerdev);
1449 	}
1450 
1451 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1452 		if (vlan->mode != MACVLAN_MODE_SOURCE) {
1453 			err = -EINVAL;
1454 			goto destroy_macvlan_port;
1455 		}
1456 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1457 		err = macvlan_changelink_sources(vlan, macmode, data);
1458 		if (err)
1459 			goto destroy_macvlan_port;
1460 	}
1461 
1462 	err = register_netdevice(dev);
1463 	if (err < 0)
1464 		goto destroy_macvlan_port;
1465 
1466 	dev->priv_flags |= IFF_MACVLAN;
1467 	err = netdev_upper_dev_link(lowerdev, dev, extack);
1468 	if (err)
1469 		goto unregister_netdev;
1470 
1471 	list_add_tail_rcu(&vlan->list, &port->vlans);
1472 	netif_stacked_transfer_operstate(lowerdev, dev);
1473 	linkwatch_fire_event(dev);
1474 
1475 	return 0;
1476 
1477 unregister_netdev:
1478 	/* macvlan_uninit would free the macvlan port */
1479 	unregister_netdevice(dev);
1480 	return err;
1481 destroy_macvlan_port:
1482 	/* the macvlan port may be freed by macvlan_uninit when fail to register.
1483 	 * so we destroy the macvlan port only when it's valid.
1484 	 */
1485 	if (create && macvlan_port_get_rtnl(lowerdev))
1486 		macvlan_port_destroy(port->dev);
1487 	return err;
1488 }
1489 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1490 
1491 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1492 			   struct nlattr *tb[], struct nlattr *data[],
1493 			   struct netlink_ext_ack *extack)
1494 {
1495 	return macvlan_common_newlink(src_net, dev, tb, data, extack);
1496 }
1497 
1498 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1499 {
1500 	struct macvlan_dev *vlan = netdev_priv(dev);
1501 
1502 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1503 		macvlan_flush_sources(vlan->port, vlan);
1504 	list_del_rcu(&vlan->list);
1505 	unregister_netdevice_queue(dev, head);
1506 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1507 }
1508 EXPORT_SYMBOL_GPL(macvlan_dellink);
1509 
1510 static int macvlan_changelink(struct net_device *dev,
1511 			      struct nlattr *tb[], struct nlattr *data[],
1512 			      struct netlink_ext_ack *extack)
1513 {
1514 	struct macvlan_dev *vlan = netdev_priv(dev);
1515 	enum macvlan_mode mode;
1516 	bool set_mode = false;
1517 	enum macvlan_macaddr_mode macmode;
1518 	int ret;
1519 
1520 	/* Validate mode, but don't set yet: setting flags may fail. */
1521 	if (data && data[IFLA_MACVLAN_MODE]) {
1522 		set_mode = true;
1523 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1524 		/* Passthrough mode can't be set or cleared dynamically */
1525 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1526 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1527 			return -EINVAL;
1528 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1529 		    vlan->mode != mode)
1530 			macvlan_flush_sources(vlan->port, vlan);
1531 	}
1532 
1533 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1534 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1535 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1536 		if (macvlan_passthru(vlan->port) && promisc) {
1537 			int err;
1538 
1539 			if (flags & MACVLAN_FLAG_NOPROMISC)
1540 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1541 			else
1542 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1543 			if (err < 0)
1544 				return err;
1545 		}
1546 		vlan->flags = flags;
1547 	}
1548 	if (set_mode)
1549 		vlan->mode = mode;
1550 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1551 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1552 			return -EINVAL;
1553 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1554 		ret = macvlan_changelink_sources(vlan, macmode, data);
1555 		if (ret)
1556 			return ret;
1557 	}
1558 	return 0;
1559 }
1560 
1561 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1562 {
1563 	if (vlan->macaddr_count == 0)
1564 		return 0;
1565 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1566 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1567 }
1568 
1569 static size_t macvlan_get_size(const struct net_device *dev)
1570 {
1571 	struct macvlan_dev *vlan = netdev_priv(dev);
1572 
1573 	return (0
1574 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1575 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1576 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1577 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1578 		);
1579 }
1580 
1581 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1582 				     const struct macvlan_dev *vlan,
1583 				     const int i)
1584 {
1585 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1586 	struct macvlan_source_entry *entry;
1587 
1588 	hlist_for_each_entry_rcu(entry, h, hlist) {
1589 		if (entry->vlan != vlan)
1590 			continue;
1591 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1592 			return 1;
1593 	}
1594 	return 0;
1595 }
1596 
1597 static int macvlan_fill_info(struct sk_buff *skb,
1598 				const struct net_device *dev)
1599 {
1600 	struct macvlan_dev *vlan = netdev_priv(dev);
1601 	int i;
1602 	struct nlattr *nest;
1603 
1604 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1605 		goto nla_put_failure;
1606 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1607 		goto nla_put_failure;
1608 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1609 		goto nla_put_failure;
1610 	if (vlan->macaddr_count > 0) {
1611 		nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1612 		if (nest == NULL)
1613 			goto nla_put_failure;
1614 
1615 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1616 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1617 				goto nla_put_failure;
1618 		}
1619 		nla_nest_end(skb, nest);
1620 	}
1621 	return 0;
1622 
1623 nla_put_failure:
1624 	return -EMSGSIZE;
1625 }
1626 
1627 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1628 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1629 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1630 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1631 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1632 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1633 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1634 };
1635 
1636 int macvlan_link_register(struct rtnl_link_ops *ops)
1637 {
1638 	/* common fields */
1639 	ops->validate		= macvlan_validate;
1640 	ops->maxtype		= IFLA_MACVLAN_MAX;
1641 	ops->policy		= macvlan_policy;
1642 	ops->changelink		= macvlan_changelink;
1643 	ops->get_size		= macvlan_get_size;
1644 	ops->fill_info		= macvlan_fill_info;
1645 
1646 	return rtnl_link_register(ops);
1647 };
1648 EXPORT_SYMBOL_GPL(macvlan_link_register);
1649 
1650 static struct net *macvlan_get_link_net(const struct net_device *dev)
1651 {
1652 	return dev_net(macvlan_dev_real_dev(dev));
1653 }
1654 
1655 static struct rtnl_link_ops macvlan_link_ops = {
1656 	.kind		= "macvlan",
1657 	.setup		= macvlan_setup,
1658 	.newlink	= macvlan_newlink,
1659 	.dellink	= macvlan_dellink,
1660 	.get_link_net	= macvlan_get_link_net,
1661 	.priv_size      = sizeof(struct macvlan_dev),
1662 };
1663 
1664 static int macvlan_device_event(struct notifier_block *unused,
1665 				unsigned long event, void *ptr)
1666 {
1667 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1668 	struct macvlan_dev *vlan, *next;
1669 	struct macvlan_port *port;
1670 	LIST_HEAD(list_kill);
1671 
1672 	if (!netif_is_macvlan_port(dev))
1673 		return NOTIFY_DONE;
1674 
1675 	port = macvlan_port_get_rtnl(dev);
1676 
1677 	switch (event) {
1678 	case NETDEV_UP:
1679 	case NETDEV_DOWN:
1680 	case NETDEV_CHANGE:
1681 		list_for_each_entry(vlan, &port->vlans, list)
1682 			netif_stacked_transfer_operstate(vlan->lowerdev,
1683 							 vlan->dev);
1684 		break;
1685 	case NETDEV_FEAT_CHANGE:
1686 		list_for_each_entry(vlan, &port->vlans, list) {
1687 			vlan->dev->gso_max_size = dev->gso_max_size;
1688 			vlan->dev->gso_max_segs = dev->gso_max_segs;
1689 			netdev_update_features(vlan->dev);
1690 		}
1691 		break;
1692 	case NETDEV_CHANGEMTU:
1693 		list_for_each_entry(vlan, &port->vlans, list) {
1694 			if (vlan->dev->mtu <= dev->mtu)
1695 				continue;
1696 			dev_set_mtu(vlan->dev, dev->mtu);
1697 		}
1698 		break;
1699 	case NETDEV_CHANGEADDR:
1700 		if (!macvlan_passthru(port))
1701 			return NOTIFY_DONE;
1702 
1703 		vlan = list_first_entry_or_null(&port->vlans,
1704 						struct macvlan_dev,
1705 						list);
1706 
1707 		if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1708 			return NOTIFY_BAD;
1709 
1710 		break;
1711 	case NETDEV_UNREGISTER:
1712 		/* twiddle thumbs on netns device moves */
1713 		if (dev->reg_state != NETREG_UNREGISTERING)
1714 			break;
1715 
1716 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1717 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1718 		unregister_netdevice_many(&list_kill);
1719 		break;
1720 	case NETDEV_PRE_TYPE_CHANGE:
1721 		/* Forbid underlaying device to change its type. */
1722 		return NOTIFY_BAD;
1723 
1724 	case NETDEV_NOTIFY_PEERS:
1725 	case NETDEV_BONDING_FAILOVER:
1726 	case NETDEV_RESEND_IGMP:
1727 		/* Propagate to all vlans */
1728 		list_for_each_entry(vlan, &port->vlans, list)
1729 			call_netdevice_notifiers(event, vlan->dev);
1730 	}
1731 	return NOTIFY_DONE;
1732 }
1733 
1734 static struct notifier_block macvlan_notifier_block __read_mostly = {
1735 	.notifier_call	= macvlan_device_event,
1736 };
1737 
1738 static int __init macvlan_init_module(void)
1739 {
1740 	int err;
1741 
1742 	register_netdevice_notifier(&macvlan_notifier_block);
1743 
1744 	err = macvlan_link_register(&macvlan_link_ops);
1745 	if (err < 0)
1746 		goto err1;
1747 	return 0;
1748 err1:
1749 	unregister_netdevice_notifier(&macvlan_notifier_block);
1750 	return err;
1751 }
1752 
1753 static void __exit macvlan_cleanup_module(void)
1754 {
1755 	rtnl_link_unregister(&macvlan_link_ops);
1756 	unregister_netdevice_notifier(&macvlan_notifier_block);
1757 }
1758 
1759 module_init(macvlan_init_module);
1760 module_exit(macvlan_cleanup_module);
1761 
1762 MODULE_LICENSE("GPL");
1763 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1764 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1765 MODULE_ALIAS_RTNL_LINK("macvlan");
1766