xref: /linux/drivers/net/macvlan.c (revision b7019ac550eb3916f34d79db583e9b7ea2524afa)
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 }
339 
340 static void macvlan_broadcast_enqueue(struct macvlan_port *port,
341 				      const struct macvlan_dev *src,
342 				      struct sk_buff *skb)
343 {
344 	struct sk_buff *nskb;
345 	int err = -ENOMEM;
346 
347 	nskb = skb_clone(skb, GFP_ATOMIC);
348 	if (!nskb)
349 		goto err;
350 
351 	MACVLAN_SKB_CB(nskb)->src = src;
352 
353 	spin_lock(&port->bc_queue.lock);
354 	if (skb_queue_len(&port->bc_queue) < MACVLAN_BC_QUEUE_LEN) {
355 		if (src)
356 			dev_hold(src->dev);
357 		__skb_queue_tail(&port->bc_queue, nskb);
358 		err = 0;
359 	}
360 	spin_unlock(&port->bc_queue.lock);
361 
362 	if (err)
363 		goto free_nskb;
364 
365 	schedule_work(&port->bc_work);
366 	return;
367 
368 free_nskb:
369 	kfree_skb(nskb);
370 err:
371 	atomic_long_inc(&skb->dev->rx_dropped);
372 }
373 
374 static void macvlan_flush_sources(struct macvlan_port *port,
375 				  struct macvlan_dev *vlan)
376 {
377 	int i;
378 
379 	for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
380 		struct hlist_node *h, *n;
381 
382 		hlist_for_each_safe(h, n, &port->vlan_source_hash[i]) {
383 			struct macvlan_source_entry *entry;
384 
385 			entry = hlist_entry(h, struct macvlan_source_entry,
386 					    hlist);
387 			if (entry->vlan == vlan)
388 				macvlan_hash_del_source(entry);
389 		}
390 	}
391 	vlan->macaddr_count = 0;
392 }
393 
394 static void macvlan_forward_source_one(struct sk_buff *skb,
395 				       struct macvlan_dev *vlan)
396 {
397 	struct sk_buff *nskb;
398 	struct net_device *dev;
399 	int len;
400 	int ret;
401 
402 	dev = vlan->dev;
403 	if (unlikely(!(dev->flags & IFF_UP)))
404 		return;
405 
406 	nskb = skb_clone(skb, GFP_ATOMIC);
407 	if (!nskb)
408 		return;
409 
410 	len = nskb->len + ETH_HLEN;
411 	nskb->dev = dev;
412 
413 	if (ether_addr_equal_64bits(eth_hdr(skb)->h_dest, dev->dev_addr))
414 		nskb->pkt_type = PACKET_HOST;
415 
416 	ret = netif_rx(nskb);
417 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
418 }
419 
420 static void macvlan_forward_source(struct sk_buff *skb,
421 				   struct macvlan_port *port,
422 				   const unsigned char *addr)
423 {
424 	struct macvlan_source_entry *entry;
425 	u32 idx = macvlan_eth_hash(addr);
426 	struct hlist_head *h = &port->vlan_source_hash[idx];
427 
428 	hlist_for_each_entry_rcu(entry, h, hlist) {
429 		if (ether_addr_equal_64bits(entry->addr, addr))
430 			macvlan_forward_source_one(skb, entry->vlan);
431 	}
432 }
433 
434 /* called under rcu_read_lock() from netif_receive_skb */
435 static rx_handler_result_t macvlan_handle_frame(struct sk_buff **pskb)
436 {
437 	struct macvlan_port *port;
438 	struct sk_buff *skb = *pskb;
439 	const struct ethhdr *eth = eth_hdr(skb);
440 	const struct macvlan_dev *vlan;
441 	const struct macvlan_dev *src;
442 	struct net_device *dev;
443 	unsigned int len = 0;
444 	int ret;
445 	rx_handler_result_t handle_res;
446 
447 	port = macvlan_port_get_rcu(skb->dev);
448 	if (is_multicast_ether_addr(eth->h_dest)) {
449 		unsigned int hash;
450 
451 		skb = ip_check_defrag(dev_net(skb->dev), skb, IP_DEFRAG_MACVLAN);
452 		if (!skb)
453 			return RX_HANDLER_CONSUMED;
454 		*pskb = skb;
455 		eth = eth_hdr(skb);
456 		macvlan_forward_source(skb, port, eth->h_source);
457 		src = macvlan_hash_lookup(port, eth->h_source);
458 		if (src && src->mode != MACVLAN_MODE_VEPA &&
459 		    src->mode != MACVLAN_MODE_BRIDGE) {
460 			/* forward to original port. */
461 			vlan = src;
462 			ret = macvlan_broadcast_one(skb, vlan, eth, 0) ?:
463 			      netif_rx(skb);
464 			handle_res = RX_HANDLER_CONSUMED;
465 			goto out;
466 		}
467 
468 		hash = mc_hash(NULL, eth->h_dest);
469 		if (test_bit(hash, port->mc_filter))
470 			macvlan_broadcast_enqueue(port, src, skb);
471 
472 		return RX_HANDLER_PASS;
473 	}
474 
475 	macvlan_forward_source(skb, port, eth->h_source);
476 	if (macvlan_passthru(port))
477 		vlan = list_first_or_null_rcu(&port->vlans,
478 					      struct macvlan_dev, list);
479 	else
480 		vlan = macvlan_hash_lookup(port, eth->h_dest);
481 	if (!vlan || vlan->mode == MACVLAN_MODE_SOURCE)
482 		return RX_HANDLER_PASS;
483 
484 	dev = vlan->dev;
485 	if (unlikely(!(dev->flags & IFF_UP))) {
486 		kfree_skb(skb);
487 		return RX_HANDLER_CONSUMED;
488 	}
489 	len = skb->len + ETH_HLEN;
490 	skb = skb_share_check(skb, GFP_ATOMIC);
491 	if (!skb) {
492 		ret = NET_RX_DROP;
493 		handle_res = RX_HANDLER_CONSUMED;
494 		goto out;
495 	}
496 
497 	*pskb = skb;
498 	skb->dev = dev;
499 	skb->pkt_type = PACKET_HOST;
500 
501 	ret = NET_RX_SUCCESS;
502 	handle_res = RX_HANDLER_ANOTHER;
503 out:
504 	macvlan_count_rx(vlan, len, ret == NET_RX_SUCCESS, false);
505 	return handle_res;
506 }
507 
508 static int macvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
509 {
510 	const struct macvlan_dev *vlan = netdev_priv(dev);
511 	const struct macvlan_port *port = vlan->port;
512 	const struct macvlan_dev *dest;
513 
514 	if (vlan->mode == MACVLAN_MODE_BRIDGE) {
515 		const struct ethhdr *eth = (void *)skb->data;
516 
517 		/* send to other bridge ports directly */
518 		if (is_multicast_ether_addr(eth->h_dest)) {
519 			macvlan_broadcast(skb, port, dev, MACVLAN_MODE_BRIDGE);
520 			goto xmit_world;
521 		}
522 
523 		dest = macvlan_hash_lookup(port, eth->h_dest);
524 		if (dest && dest->mode == MACVLAN_MODE_BRIDGE) {
525 			/* send to lowerdev first for its network taps */
526 			dev_forward_skb(vlan->lowerdev, skb);
527 
528 			return NET_XMIT_SUCCESS;
529 		}
530 	}
531 xmit_world:
532 	skb->dev = vlan->lowerdev;
533 	return dev_queue_xmit_accel(skb,
534 				    netdev_get_sb_channel(dev) ? dev : NULL);
535 }
536 
537 static inline netdev_tx_t macvlan_netpoll_send_skb(struct macvlan_dev *vlan, struct sk_buff *skb)
538 {
539 #ifdef CONFIG_NET_POLL_CONTROLLER
540 	if (vlan->netpoll)
541 		netpoll_send_skb(vlan->netpoll, skb);
542 #else
543 	BUG();
544 #endif
545 	return NETDEV_TX_OK;
546 }
547 
548 static netdev_tx_t macvlan_start_xmit(struct sk_buff *skb,
549 				      struct net_device *dev)
550 {
551 	struct macvlan_dev *vlan = netdev_priv(dev);
552 	unsigned int len = skb->len;
553 	int ret;
554 
555 	if (unlikely(netpoll_tx_running(dev)))
556 		return macvlan_netpoll_send_skb(vlan, skb);
557 
558 	ret = macvlan_queue_xmit(skb, dev);
559 
560 	if (likely(ret == NET_XMIT_SUCCESS || ret == NET_XMIT_CN)) {
561 		struct vlan_pcpu_stats *pcpu_stats;
562 
563 		pcpu_stats = this_cpu_ptr(vlan->pcpu_stats);
564 		u64_stats_update_begin(&pcpu_stats->syncp);
565 		pcpu_stats->tx_packets++;
566 		pcpu_stats->tx_bytes += len;
567 		u64_stats_update_end(&pcpu_stats->syncp);
568 	} else {
569 		this_cpu_inc(vlan->pcpu_stats->tx_dropped);
570 	}
571 	return ret;
572 }
573 
574 static int macvlan_hard_header(struct sk_buff *skb, struct net_device *dev,
575 			       unsigned short type, const void *daddr,
576 			       const void *saddr, unsigned len)
577 {
578 	const struct macvlan_dev *vlan = netdev_priv(dev);
579 	struct net_device *lowerdev = vlan->lowerdev;
580 
581 	return dev_hard_header(skb, lowerdev, type, daddr,
582 			       saddr ? : dev->dev_addr, len);
583 }
584 
585 static const struct header_ops macvlan_hard_header_ops = {
586 	.create  	= macvlan_hard_header,
587 	.parse		= eth_header_parse,
588 	.cache		= eth_header_cache,
589 	.cache_update	= eth_header_cache_update,
590 };
591 
592 static int macvlan_open(struct net_device *dev)
593 {
594 	struct macvlan_dev *vlan = netdev_priv(dev);
595 	struct net_device *lowerdev = vlan->lowerdev;
596 	int err;
597 
598 	if (macvlan_passthru(vlan->port)) {
599 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC)) {
600 			err = dev_set_promiscuity(lowerdev, 1);
601 			if (err < 0)
602 				goto out;
603 		}
604 		goto hash_add;
605 	}
606 
607 	err = -EADDRINUSE;
608 	if (macvlan_addr_busy(vlan->port, dev->dev_addr))
609 		goto out;
610 
611 	/* Attempt to populate accel_priv which is used to offload the L2
612 	 * forwarding requests for unicast packets.
613 	 */
614 	if (lowerdev->features & NETIF_F_HW_L2FW_DOFFLOAD)
615 		vlan->accel_priv =
616 		      lowerdev->netdev_ops->ndo_dfwd_add_station(lowerdev, dev);
617 
618 	/* If earlier attempt to offload failed, or accel_priv is not
619 	 * populated we must add the unicast address to the lower device.
620 	 */
621 	if (IS_ERR_OR_NULL(vlan->accel_priv)) {
622 		vlan->accel_priv = NULL;
623 		err = dev_uc_add(lowerdev, dev->dev_addr);
624 		if (err < 0)
625 			goto out;
626 	}
627 
628 	if (dev->flags & IFF_ALLMULTI) {
629 		err = dev_set_allmulti(lowerdev, 1);
630 		if (err < 0)
631 			goto del_unicast;
632 	}
633 
634 	if (dev->flags & IFF_PROMISC) {
635 		err = dev_set_promiscuity(lowerdev, 1);
636 		if (err < 0)
637 			goto clear_multi;
638 	}
639 
640 hash_add:
641 	macvlan_hash_add(vlan);
642 	return 0;
643 
644 clear_multi:
645 	if (dev->flags & IFF_ALLMULTI)
646 		dev_set_allmulti(lowerdev, -1);
647 del_unicast:
648 	if (vlan->accel_priv) {
649 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
650 							   vlan->accel_priv);
651 		vlan->accel_priv = NULL;
652 	} else {
653 		dev_uc_del(lowerdev, dev->dev_addr);
654 	}
655 out:
656 	return err;
657 }
658 
659 static int macvlan_stop(struct net_device *dev)
660 {
661 	struct macvlan_dev *vlan = netdev_priv(dev);
662 	struct net_device *lowerdev = vlan->lowerdev;
663 
664 	if (vlan->accel_priv) {
665 		lowerdev->netdev_ops->ndo_dfwd_del_station(lowerdev,
666 							   vlan->accel_priv);
667 		vlan->accel_priv = NULL;
668 	}
669 
670 	dev_uc_unsync(lowerdev, dev);
671 	dev_mc_unsync(lowerdev, dev);
672 
673 	if (macvlan_passthru(vlan->port)) {
674 		if (!(vlan->flags & MACVLAN_FLAG_NOPROMISC))
675 			dev_set_promiscuity(lowerdev, -1);
676 		goto hash_del;
677 	}
678 
679 	if (dev->flags & IFF_ALLMULTI)
680 		dev_set_allmulti(lowerdev, -1);
681 
682 	if (dev->flags & IFF_PROMISC)
683 		dev_set_promiscuity(lowerdev, -1);
684 
685 	dev_uc_del(lowerdev, dev->dev_addr);
686 
687 hash_del:
688 	macvlan_hash_del(vlan, !dev->dismantle);
689 	return 0;
690 }
691 
692 static int macvlan_sync_address(struct net_device *dev, unsigned char *addr)
693 {
694 	struct macvlan_dev *vlan = netdev_priv(dev);
695 	struct net_device *lowerdev = vlan->lowerdev;
696 	struct macvlan_port *port = vlan->port;
697 	int err;
698 
699 	if (!(dev->flags & IFF_UP)) {
700 		/* Just copy in the new address */
701 		ether_addr_copy(dev->dev_addr, addr);
702 	} else {
703 		/* Rehash and update the device filters */
704 		if (macvlan_addr_busy(vlan->port, addr))
705 			return -EADDRINUSE;
706 
707 		if (!macvlan_passthru(port)) {
708 			err = dev_uc_add(lowerdev, addr);
709 			if (err)
710 				return err;
711 
712 			dev_uc_del(lowerdev, dev->dev_addr);
713 		}
714 
715 		macvlan_hash_change_addr(vlan, addr);
716 	}
717 	if (macvlan_passthru(port) && !macvlan_addr_change(port)) {
718 		/* Since addr_change isn't set, we are here due to lower
719 		 * device change.  Save the lower-dev address so we can
720 		 * restore it later.
721 		 */
722 		ether_addr_copy(vlan->port->perm_addr,
723 				lowerdev->dev_addr);
724 	}
725 	macvlan_clear_addr_change(port);
726 	return 0;
727 }
728 
729 static int macvlan_set_mac_address(struct net_device *dev, void *p)
730 {
731 	struct macvlan_dev *vlan = netdev_priv(dev);
732 	struct sockaddr *addr = p;
733 
734 	if (!is_valid_ether_addr(addr->sa_data))
735 		return -EADDRNOTAVAIL;
736 
737 	/* If the addresses are the same, this is a no-op */
738 	if (ether_addr_equal(dev->dev_addr, addr->sa_data))
739 		return 0;
740 
741 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
742 		macvlan_set_addr_change(vlan->port);
743 		return dev_set_mac_address(vlan->lowerdev, addr, NULL);
744 	}
745 
746 	if (macvlan_addr_busy(vlan->port, addr->sa_data))
747 		return -EADDRINUSE;
748 
749 	return macvlan_sync_address(dev, addr->sa_data);
750 }
751 
752 static void macvlan_change_rx_flags(struct net_device *dev, int change)
753 {
754 	struct macvlan_dev *vlan = netdev_priv(dev);
755 	struct net_device *lowerdev = vlan->lowerdev;
756 
757 	if (dev->flags & IFF_UP) {
758 		if (change & IFF_ALLMULTI)
759 			dev_set_allmulti(lowerdev, dev->flags & IFF_ALLMULTI ? 1 : -1);
760 		if (change & IFF_PROMISC)
761 			dev_set_promiscuity(lowerdev,
762 					    dev->flags & IFF_PROMISC ? 1 : -1);
763 
764 	}
765 }
766 
767 static void macvlan_compute_filter(unsigned long *mc_filter,
768 				   struct net_device *dev,
769 				   struct macvlan_dev *vlan)
770 {
771 	if (dev->flags & (IFF_PROMISC | IFF_ALLMULTI)) {
772 		bitmap_fill(mc_filter, MACVLAN_MC_FILTER_SZ);
773 	} else {
774 		struct netdev_hw_addr *ha;
775 		DECLARE_BITMAP(filter, MACVLAN_MC_FILTER_SZ);
776 
777 		bitmap_zero(filter, MACVLAN_MC_FILTER_SZ);
778 		netdev_for_each_mc_addr(ha, dev) {
779 			__set_bit(mc_hash(vlan, ha->addr), filter);
780 		}
781 
782 		__set_bit(mc_hash(vlan, dev->broadcast), filter);
783 
784 		bitmap_copy(mc_filter, filter, MACVLAN_MC_FILTER_SZ);
785 	}
786 }
787 
788 static void macvlan_set_mac_lists(struct net_device *dev)
789 {
790 	struct macvlan_dev *vlan = netdev_priv(dev);
791 
792 	macvlan_compute_filter(vlan->mc_filter, dev, vlan);
793 
794 	dev_uc_sync(vlan->lowerdev, dev);
795 	dev_mc_sync(vlan->lowerdev, dev);
796 
797 	/* This is slightly inaccurate as we're including the subscription
798 	 * list of vlan->lowerdev too.
799 	 *
800 	 * Bug alert: This only works if everyone has the same broadcast
801 	 * address as lowerdev.  As soon as someone changes theirs this
802 	 * will break.
803 	 *
804 	 * However, this is already broken as when you change your broadcast
805 	 * address we don't get called.
806 	 *
807 	 * The solution is to maintain a list of broadcast addresses like
808 	 * we do for uc/mc, if you care.
809 	 */
810 	macvlan_compute_filter(vlan->port->mc_filter, vlan->lowerdev, NULL);
811 }
812 
813 static int macvlan_change_mtu(struct net_device *dev, int new_mtu)
814 {
815 	struct macvlan_dev *vlan = netdev_priv(dev);
816 
817 	if (vlan->lowerdev->mtu < new_mtu)
818 		return -EINVAL;
819 	dev->mtu = new_mtu;
820 	return 0;
821 }
822 
823 static int macvlan_do_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
824 {
825 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
826 	const struct net_device_ops *ops = real_dev->netdev_ops;
827 	struct ifreq ifrr;
828 	int err = -EOPNOTSUPP;
829 
830 	strncpy(ifrr.ifr_name, real_dev->name, IFNAMSIZ);
831 	ifrr.ifr_ifru = ifr->ifr_ifru;
832 
833 	switch (cmd) {
834 	case SIOCSHWTSTAMP:
835 		if (!net_eq(dev_net(dev), &init_net))
836 			break;
837 		/* fall through */
838 	case SIOCGHWTSTAMP:
839 		if (netif_device_present(real_dev) && ops->ndo_do_ioctl)
840 			err = ops->ndo_do_ioctl(real_dev, &ifrr, cmd);
841 		break;
842 	}
843 
844 	if (!err)
845 		ifr->ifr_ifru = ifrr.ifr_ifru;
846 
847 	return err;
848 }
849 
850 /*
851  * macvlan network devices have devices nesting below it and are a special
852  * "super class" of normal network devices; split their locks off into a
853  * separate class since they always nest.
854  */
855 static struct lock_class_key macvlan_netdev_addr_lock_key;
856 
857 #define ALWAYS_ON_OFFLOADS \
858 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE | \
859 	 NETIF_F_GSO_ROBUST | NETIF_F_GSO_ENCAP_ALL)
860 
861 #define ALWAYS_ON_FEATURES (ALWAYS_ON_OFFLOADS | NETIF_F_LLTX)
862 
863 #define MACVLAN_FEATURES \
864 	(NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_HIGHDMA | NETIF_F_FRAGLIST | \
865 	 NETIF_F_GSO | NETIF_F_TSO | NETIF_F_LRO | \
866 	 NETIF_F_TSO_ECN | NETIF_F_TSO6 | NETIF_F_GRO | NETIF_F_RXCSUM | \
867 	 NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER)
868 
869 #define MACVLAN_STATE_MASK \
870 	((1<<__LINK_STATE_NOCARRIER) | (1<<__LINK_STATE_DORMANT))
871 
872 static int macvlan_get_nest_level(struct net_device *dev)
873 {
874 	return ((struct macvlan_dev *)netdev_priv(dev))->nest_level;
875 }
876 
877 static void macvlan_set_lockdep_class(struct net_device *dev)
878 {
879 	netdev_lockdep_set_classes(dev);
880 	lockdep_set_class_and_subclass(&dev->addr_list_lock,
881 				       &macvlan_netdev_addr_lock_key,
882 				       macvlan_get_nest_level(dev));
883 }
884 
885 static int macvlan_init(struct net_device *dev)
886 {
887 	struct macvlan_dev *vlan = netdev_priv(dev);
888 	const struct net_device *lowerdev = vlan->lowerdev;
889 	struct macvlan_port *port = vlan->port;
890 
891 	dev->state		= (dev->state & ~MACVLAN_STATE_MASK) |
892 				  (lowerdev->state & MACVLAN_STATE_MASK);
893 	dev->features 		= lowerdev->features & MACVLAN_FEATURES;
894 	dev->features		|= ALWAYS_ON_FEATURES;
895 	dev->hw_features	|= NETIF_F_LRO;
896 	dev->vlan_features	= lowerdev->vlan_features & MACVLAN_FEATURES;
897 	dev->vlan_features	|= ALWAYS_ON_OFFLOADS;
898 	dev->hw_enc_features    |= dev->features;
899 	dev->gso_max_size	= lowerdev->gso_max_size;
900 	dev->gso_max_segs	= lowerdev->gso_max_segs;
901 	dev->hard_header_len	= lowerdev->hard_header_len;
902 
903 	macvlan_set_lockdep_class(dev);
904 
905 	vlan->pcpu_stats = netdev_alloc_pcpu_stats(struct vlan_pcpu_stats);
906 	if (!vlan->pcpu_stats)
907 		return -ENOMEM;
908 
909 	port->count += 1;
910 
911 	return 0;
912 }
913 
914 static void macvlan_uninit(struct net_device *dev)
915 {
916 	struct macvlan_dev *vlan = netdev_priv(dev);
917 	struct macvlan_port *port = vlan->port;
918 
919 	free_percpu(vlan->pcpu_stats);
920 
921 	macvlan_flush_sources(port, vlan);
922 	port->count -= 1;
923 	if (!port->count)
924 		macvlan_port_destroy(port->dev);
925 }
926 
927 static void macvlan_dev_get_stats64(struct net_device *dev,
928 				    struct rtnl_link_stats64 *stats)
929 {
930 	struct macvlan_dev *vlan = netdev_priv(dev);
931 
932 	if (vlan->pcpu_stats) {
933 		struct vlan_pcpu_stats *p;
934 		u64 rx_packets, rx_bytes, rx_multicast, tx_packets, tx_bytes;
935 		u32 rx_errors = 0, tx_dropped = 0;
936 		unsigned int start;
937 		int i;
938 
939 		for_each_possible_cpu(i) {
940 			p = per_cpu_ptr(vlan->pcpu_stats, i);
941 			do {
942 				start = u64_stats_fetch_begin_irq(&p->syncp);
943 				rx_packets	= p->rx_packets;
944 				rx_bytes	= p->rx_bytes;
945 				rx_multicast	= p->rx_multicast;
946 				tx_packets	= p->tx_packets;
947 				tx_bytes	= p->tx_bytes;
948 			} while (u64_stats_fetch_retry_irq(&p->syncp, start));
949 
950 			stats->rx_packets	+= rx_packets;
951 			stats->rx_bytes		+= rx_bytes;
952 			stats->multicast	+= rx_multicast;
953 			stats->tx_packets	+= tx_packets;
954 			stats->tx_bytes		+= tx_bytes;
955 			/* rx_errors & tx_dropped are u32, updated
956 			 * without syncp protection.
957 			 */
958 			rx_errors	+= p->rx_errors;
959 			tx_dropped	+= p->tx_dropped;
960 		}
961 		stats->rx_errors	= rx_errors;
962 		stats->rx_dropped	= rx_errors;
963 		stats->tx_dropped	= tx_dropped;
964 	}
965 }
966 
967 static int macvlan_vlan_rx_add_vid(struct net_device *dev,
968 				   __be16 proto, u16 vid)
969 {
970 	struct macvlan_dev *vlan = netdev_priv(dev);
971 	struct net_device *lowerdev = vlan->lowerdev;
972 
973 	return vlan_vid_add(lowerdev, proto, vid);
974 }
975 
976 static int macvlan_vlan_rx_kill_vid(struct net_device *dev,
977 				    __be16 proto, u16 vid)
978 {
979 	struct macvlan_dev *vlan = netdev_priv(dev);
980 	struct net_device *lowerdev = vlan->lowerdev;
981 
982 	vlan_vid_del(lowerdev, proto, vid);
983 	return 0;
984 }
985 
986 static int macvlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
987 			   struct net_device *dev,
988 			   const unsigned char *addr, u16 vid,
989 			   u16 flags,
990 			   struct netlink_ext_ack *extack)
991 {
992 	struct macvlan_dev *vlan = netdev_priv(dev);
993 	int err = -EINVAL;
994 
995 	/* Support unicast filter only on passthru devices.
996 	 * Multicast filter should be allowed on all devices.
997 	 */
998 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
999 		return -EOPNOTSUPP;
1000 
1001 	if (flags & NLM_F_REPLACE)
1002 		return -EOPNOTSUPP;
1003 
1004 	if (is_unicast_ether_addr(addr))
1005 		err = dev_uc_add_excl(dev, addr);
1006 	else if (is_multicast_ether_addr(addr))
1007 		err = dev_mc_add_excl(dev, addr);
1008 
1009 	return err;
1010 }
1011 
1012 static int macvlan_fdb_del(struct ndmsg *ndm, struct nlattr *tb[],
1013 			   struct net_device *dev,
1014 			   const unsigned char *addr, u16 vid)
1015 {
1016 	struct macvlan_dev *vlan = netdev_priv(dev);
1017 	int err = -EINVAL;
1018 
1019 	/* Support unicast filter only on passthru devices.
1020 	 * Multicast filter should be allowed on all devices.
1021 	 */
1022 	if (!macvlan_passthru(vlan->port) && is_unicast_ether_addr(addr))
1023 		return -EOPNOTSUPP;
1024 
1025 	if (is_unicast_ether_addr(addr))
1026 		err = dev_uc_del(dev, addr);
1027 	else if (is_multicast_ether_addr(addr))
1028 		err = dev_mc_del(dev, addr);
1029 
1030 	return err;
1031 }
1032 
1033 static void macvlan_ethtool_get_drvinfo(struct net_device *dev,
1034 					struct ethtool_drvinfo *drvinfo)
1035 {
1036 	strlcpy(drvinfo->driver, "macvlan", sizeof(drvinfo->driver));
1037 	strlcpy(drvinfo->version, "0.1", sizeof(drvinfo->version));
1038 }
1039 
1040 static int macvlan_ethtool_get_link_ksettings(struct net_device *dev,
1041 					      struct ethtool_link_ksettings *cmd)
1042 {
1043 	const struct macvlan_dev *vlan = netdev_priv(dev);
1044 
1045 	return __ethtool_get_link_ksettings(vlan->lowerdev, cmd);
1046 }
1047 
1048 static int macvlan_ethtool_get_ts_info(struct net_device *dev,
1049 				       struct ethtool_ts_info *info)
1050 {
1051 	struct net_device *real_dev = macvlan_dev_real_dev(dev);
1052 	const struct ethtool_ops *ops = real_dev->ethtool_ops;
1053 	struct phy_device *phydev = real_dev->phydev;
1054 
1055 	if (phydev && phydev->drv && phydev->drv->ts_info) {
1056 		 return phydev->drv->ts_info(phydev, info);
1057 	} else if (ops->get_ts_info) {
1058 		return ops->get_ts_info(real_dev, info);
1059 	} else {
1060 		info->so_timestamping = SOF_TIMESTAMPING_RX_SOFTWARE |
1061 			SOF_TIMESTAMPING_SOFTWARE;
1062 		info->phc_index = -1;
1063 	}
1064 
1065 	return 0;
1066 }
1067 
1068 static netdev_features_t macvlan_fix_features(struct net_device *dev,
1069 					      netdev_features_t features)
1070 {
1071 	struct macvlan_dev *vlan = netdev_priv(dev);
1072 	netdev_features_t lowerdev_features = vlan->lowerdev->features;
1073 	netdev_features_t mask;
1074 
1075 	features |= NETIF_F_ALL_FOR_ALL;
1076 	features &= (vlan->set_features | ~MACVLAN_FEATURES);
1077 	mask = features;
1078 
1079 	lowerdev_features &= (features | ~NETIF_F_LRO);
1080 	features = netdev_increment_features(lowerdev_features, features, mask);
1081 	features |= ALWAYS_ON_FEATURES;
1082 	features &= (ALWAYS_ON_FEATURES | MACVLAN_FEATURES);
1083 
1084 	return features;
1085 }
1086 
1087 #ifdef CONFIG_NET_POLL_CONTROLLER
1088 static void macvlan_dev_poll_controller(struct net_device *dev)
1089 {
1090 	return;
1091 }
1092 
1093 static int macvlan_dev_netpoll_setup(struct net_device *dev, struct netpoll_info *npinfo)
1094 {
1095 	struct macvlan_dev *vlan = netdev_priv(dev);
1096 	struct net_device *real_dev = vlan->lowerdev;
1097 	struct netpoll *netpoll;
1098 	int err = 0;
1099 
1100 	netpoll = kzalloc(sizeof(*netpoll), GFP_KERNEL);
1101 	err = -ENOMEM;
1102 	if (!netpoll)
1103 		goto out;
1104 
1105 	err = __netpoll_setup(netpoll, real_dev);
1106 	if (err) {
1107 		kfree(netpoll);
1108 		goto out;
1109 	}
1110 
1111 	vlan->netpoll = netpoll;
1112 
1113 out:
1114 	return err;
1115 }
1116 
1117 static void macvlan_dev_netpoll_cleanup(struct net_device *dev)
1118 {
1119 	struct macvlan_dev *vlan = netdev_priv(dev);
1120 	struct netpoll *netpoll = vlan->netpoll;
1121 
1122 	if (!netpoll)
1123 		return;
1124 
1125 	vlan->netpoll = NULL;
1126 
1127 	__netpoll_free(netpoll);
1128 }
1129 #endif	/* CONFIG_NET_POLL_CONTROLLER */
1130 
1131 static int macvlan_dev_get_iflink(const struct net_device *dev)
1132 {
1133 	struct macvlan_dev *vlan = netdev_priv(dev);
1134 
1135 	return vlan->lowerdev->ifindex;
1136 }
1137 
1138 static const struct ethtool_ops macvlan_ethtool_ops = {
1139 	.get_link		= ethtool_op_get_link,
1140 	.get_link_ksettings	= macvlan_ethtool_get_link_ksettings,
1141 	.get_drvinfo		= macvlan_ethtool_get_drvinfo,
1142 	.get_ts_info		= macvlan_ethtool_get_ts_info,
1143 };
1144 
1145 static const struct net_device_ops macvlan_netdev_ops = {
1146 	.ndo_init		= macvlan_init,
1147 	.ndo_uninit		= macvlan_uninit,
1148 	.ndo_open		= macvlan_open,
1149 	.ndo_stop		= macvlan_stop,
1150 	.ndo_start_xmit		= macvlan_start_xmit,
1151 	.ndo_change_mtu		= macvlan_change_mtu,
1152 	.ndo_do_ioctl		= macvlan_do_ioctl,
1153 	.ndo_fix_features	= macvlan_fix_features,
1154 	.ndo_change_rx_flags	= macvlan_change_rx_flags,
1155 	.ndo_set_mac_address	= macvlan_set_mac_address,
1156 	.ndo_set_rx_mode	= macvlan_set_mac_lists,
1157 	.ndo_get_stats64	= macvlan_dev_get_stats64,
1158 	.ndo_validate_addr	= eth_validate_addr,
1159 	.ndo_vlan_rx_add_vid	= macvlan_vlan_rx_add_vid,
1160 	.ndo_vlan_rx_kill_vid	= macvlan_vlan_rx_kill_vid,
1161 	.ndo_fdb_add		= macvlan_fdb_add,
1162 	.ndo_fdb_del		= macvlan_fdb_del,
1163 	.ndo_fdb_dump		= ndo_dflt_fdb_dump,
1164 	.ndo_get_lock_subclass  = macvlan_get_nest_level,
1165 #ifdef CONFIG_NET_POLL_CONTROLLER
1166 	.ndo_poll_controller	= macvlan_dev_poll_controller,
1167 	.ndo_netpoll_setup	= macvlan_dev_netpoll_setup,
1168 	.ndo_netpoll_cleanup	= macvlan_dev_netpoll_cleanup,
1169 #endif
1170 	.ndo_get_iflink		= macvlan_dev_get_iflink,
1171 	.ndo_features_check	= passthru_features_check,
1172 	.ndo_change_proto_down  = dev_change_proto_down_generic,
1173 };
1174 
1175 void macvlan_common_setup(struct net_device *dev)
1176 {
1177 	ether_setup(dev);
1178 
1179 	dev->min_mtu		= 0;
1180 	dev->max_mtu		= ETH_MAX_MTU;
1181 	dev->priv_flags	       &= ~IFF_TX_SKB_SHARING;
1182 	netif_keep_dst(dev);
1183 	dev->priv_flags	       |= IFF_UNICAST_FLT;
1184 	dev->netdev_ops		= &macvlan_netdev_ops;
1185 	dev->needs_free_netdev	= true;
1186 	dev->header_ops		= &macvlan_hard_header_ops;
1187 	dev->ethtool_ops	= &macvlan_ethtool_ops;
1188 }
1189 EXPORT_SYMBOL_GPL(macvlan_common_setup);
1190 
1191 static void macvlan_setup(struct net_device *dev)
1192 {
1193 	macvlan_common_setup(dev);
1194 	dev->priv_flags |= IFF_NO_QUEUE;
1195 }
1196 
1197 static int macvlan_port_create(struct net_device *dev)
1198 {
1199 	struct macvlan_port *port;
1200 	unsigned int i;
1201 	int err;
1202 
1203 	if (dev->type != ARPHRD_ETHER || dev->flags & IFF_LOOPBACK)
1204 		return -EINVAL;
1205 
1206 	if (netdev_is_rx_handler_busy(dev))
1207 		return -EBUSY;
1208 
1209 	port = kzalloc(sizeof(*port), GFP_KERNEL);
1210 	if (port == NULL)
1211 		return -ENOMEM;
1212 
1213 	port->dev = dev;
1214 	ether_addr_copy(port->perm_addr, dev->dev_addr);
1215 	INIT_LIST_HEAD(&port->vlans);
1216 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1217 		INIT_HLIST_HEAD(&port->vlan_hash[i]);
1218 	for (i = 0; i < MACVLAN_HASH_SIZE; i++)
1219 		INIT_HLIST_HEAD(&port->vlan_source_hash[i]);
1220 
1221 	skb_queue_head_init(&port->bc_queue);
1222 	INIT_WORK(&port->bc_work, macvlan_process_broadcast);
1223 
1224 	err = netdev_rx_handler_register(dev, macvlan_handle_frame, port);
1225 	if (err)
1226 		kfree(port);
1227 	else
1228 		dev->priv_flags |= IFF_MACVLAN_PORT;
1229 	return err;
1230 }
1231 
1232 static void macvlan_port_destroy(struct net_device *dev)
1233 {
1234 	struct macvlan_port *port = macvlan_port_get_rtnl(dev);
1235 	struct sk_buff *skb;
1236 
1237 	dev->priv_flags &= ~IFF_MACVLAN_PORT;
1238 	netdev_rx_handler_unregister(dev);
1239 
1240 	/* After this point, no packet can schedule bc_work anymore,
1241 	 * but we need to cancel it and purge left skbs if any.
1242 	 */
1243 	cancel_work_sync(&port->bc_work);
1244 
1245 	while ((skb = __skb_dequeue(&port->bc_queue))) {
1246 		const struct macvlan_dev *src = MACVLAN_SKB_CB(skb)->src;
1247 
1248 		if (src)
1249 			dev_put(src->dev);
1250 
1251 		kfree_skb(skb);
1252 	}
1253 
1254 	/* If the lower device address has been changed by passthru
1255 	 * macvlan, put it back.
1256 	 */
1257 	if (macvlan_passthru(port) &&
1258 	    !ether_addr_equal(port->dev->dev_addr, port->perm_addr)) {
1259 		struct sockaddr sa;
1260 
1261 		sa.sa_family = port->dev->type;
1262 		memcpy(&sa.sa_data, port->perm_addr, port->dev->addr_len);
1263 		dev_set_mac_address(port->dev, &sa, NULL);
1264 	}
1265 
1266 	kfree(port);
1267 }
1268 
1269 static int macvlan_validate(struct nlattr *tb[], struct nlattr *data[],
1270 			    struct netlink_ext_ack *extack)
1271 {
1272 	if (tb[IFLA_ADDRESS]) {
1273 		if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN)
1274 			return -EINVAL;
1275 		if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS])))
1276 			return -EADDRNOTAVAIL;
1277 	}
1278 
1279 	if (!data)
1280 		return 0;
1281 
1282 	if (data[IFLA_MACVLAN_FLAGS] &&
1283 	    nla_get_u16(data[IFLA_MACVLAN_FLAGS]) & ~MACVLAN_FLAG_NOPROMISC)
1284 		return -EINVAL;
1285 
1286 	if (data[IFLA_MACVLAN_MODE]) {
1287 		switch (nla_get_u32(data[IFLA_MACVLAN_MODE])) {
1288 		case MACVLAN_MODE_PRIVATE:
1289 		case MACVLAN_MODE_VEPA:
1290 		case MACVLAN_MODE_BRIDGE:
1291 		case MACVLAN_MODE_PASSTHRU:
1292 		case MACVLAN_MODE_SOURCE:
1293 			break;
1294 		default:
1295 			return -EINVAL;
1296 		}
1297 	}
1298 
1299 	if (data[IFLA_MACVLAN_MACADDR_MODE]) {
1300 		switch (nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE])) {
1301 		case MACVLAN_MACADDR_ADD:
1302 		case MACVLAN_MACADDR_DEL:
1303 		case MACVLAN_MACADDR_FLUSH:
1304 		case MACVLAN_MACADDR_SET:
1305 			break;
1306 		default:
1307 			return -EINVAL;
1308 		}
1309 	}
1310 
1311 	if (data[IFLA_MACVLAN_MACADDR]) {
1312 		if (nla_len(data[IFLA_MACVLAN_MACADDR]) != ETH_ALEN)
1313 			return -EINVAL;
1314 
1315 		if (!is_valid_ether_addr(nla_data(data[IFLA_MACVLAN_MACADDR])))
1316 			return -EADDRNOTAVAIL;
1317 	}
1318 
1319 	if (data[IFLA_MACVLAN_MACADDR_COUNT])
1320 		return -EINVAL;
1321 
1322 	return 0;
1323 }
1324 
1325 /**
1326  * reconfigure list of remote source mac address
1327  * (only for macvlan devices in source mode)
1328  * Note regarding alignment: all netlink data is aligned to 4 Byte, which
1329  * suffices for both ether_addr_copy and ether_addr_equal_64bits usage.
1330  */
1331 static int macvlan_changelink_sources(struct macvlan_dev *vlan, u32 mode,
1332 				      struct nlattr *data[])
1333 {
1334 	char *addr = NULL;
1335 	int ret, rem, len;
1336 	struct nlattr *nla, *head;
1337 	struct macvlan_source_entry *entry;
1338 
1339 	if (data[IFLA_MACVLAN_MACADDR])
1340 		addr = nla_data(data[IFLA_MACVLAN_MACADDR]);
1341 
1342 	if (mode == MACVLAN_MACADDR_ADD) {
1343 		if (!addr)
1344 			return -EINVAL;
1345 
1346 		return macvlan_hash_add_source(vlan, addr);
1347 
1348 	} else if (mode == MACVLAN_MACADDR_DEL) {
1349 		if (!addr)
1350 			return -EINVAL;
1351 
1352 		entry = macvlan_hash_lookup_source(vlan, addr);
1353 		if (entry) {
1354 			macvlan_hash_del_source(entry);
1355 			vlan->macaddr_count--;
1356 		}
1357 	} else if (mode == MACVLAN_MACADDR_FLUSH) {
1358 		macvlan_flush_sources(vlan->port, vlan);
1359 	} else if (mode == MACVLAN_MACADDR_SET) {
1360 		macvlan_flush_sources(vlan->port, vlan);
1361 
1362 		if (addr) {
1363 			ret = macvlan_hash_add_source(vlan, addr);
1364 			if (ret)
1365 				return ret;
1366 		}
1367 
1368 		if (!data || !data[IFLA_MACVLAN_MACADDR_DATA])
1369 			return 0;
1370 
1371 		head = nla_data(data[IFLA_MACVLAN_MACADDR_DATA]);
1372 		len = nla_len(data[IFLA_MACVLAN_MACADDR_DATA]);
1373 
1374 		nla_for_each_attr(nla, head, len, rem) {
1375 			if (nla_type(nla) != IFLA_MACVLAN_MACADDR ||
1376 			    nla_len(nla) != ETH_ALEN)
1377 				continue;
1378 
1379 			addr = nla_data(nla);
1380 			ret = macvlan_hash_add_source(vlan, addr);
1381 			if (ret)
1382 				return ret;
1383 		}
1384 	} else {
1385 		return -EINVAL;
1386 	}
1387 
1388 	return 0;
1389 }
1390 
1391 int macvlan_common_newlink(struct net *src_net, struct net_device *dev,
1392 			   struct nlattr *tb[], struct nlattr *data[],
1393 			   struct netlink_ext_ack *extack)
1394 {
1395 	struct macvlan_dev *vlan = netdev_priv(dev);
1396 	struct macvlan_port *port;
1397 	struct net_device *lowerdev;
1398 	int err;
1399 	int macmode;
1400 	bool create = false;
1401 
1402 	if (!tb[IFLA_LINK])
1403 		return -EINVAL;
1404 
1405 	lowerdev = __dev_get_by_index(src_net, nla_get_u32(tb[IFLA_LINK]));
1406 	if (lowerdev == NULL)
1407 		return -ENODEV;
1408 
1409 	/* When creating macvlans or macvtaps on top of other macvlans - use
1410 	 * the real device as the lowerdev.
1411 	 */
1412 	if (netif_is_macvlan(lowerdev))
1413 		lowerdev = macvlan_dev_real_dev(lowerdev);
1414 
1415 	if (!tb[IFLA_MTU])
1416 		dev->mtu = lowerdev->mtu;
1417 	else if (dev->mtu > lowerdev->mtu)
1418 		return -EINVAL;
1419 
1420 	/* MTU range: 68 - lowerdev->max_mtu */
1421 	dev->min_mtu = ETH_MIN_MTU;
1422 	dev->max_mtu = lowerdev->max_mtu;
1423 
1424 	if (!tb[IFLA_ADDRESS])
1425 		eth_hw_addr_random(dev);
1426 
1427 	if (!netif_is_macvlan_port(lowerdev)) {
1428 		err = macvlan_port_create(lowerdev);
1429 		if (err < 0)
1430 			return err;
1431 		create = true;
1432 	}
1433 	port = macvlan_port_get_rtnl(lowerdev);
1434 
1435 	/* Only 1 macvlan device can be created in passthru mode */
1436 	if (macvlan_passthru(port)) {
1437 		/* The macvlan port must be not created this time,
1438 		 * still goto destroy_macvlan_port for readability.
1439 		 */
1440 		err = -EINVAL;
1441 		goto destroy_macvlan_port;
1442 	}
1443 
1444 	vlan->lowerdev = lowerdev;
1445 	vlan->dev      = dev;
1446 	vlan->port     = port;
1447 	vlan->set_features = MACVLAN_FEATURES;
1448 	vlan->nest_level = dev_get_nest_level(lowerdev) + 1;
1449 
1450 	vlan->mode     = MACVLAN_MODE_VEPA;
1451 	if (data && data[IFLA_MACVLAN_MODE])
1452 		vlan->mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1453 
1454 	if (data && data[IFLA_MACVLAN_FLAGS])
1455 		vlan->flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1456 
1457 	if (vlan->mode == MACVLAN_MODE_PASSTHRU) {
1458 		if (port->count) {
1459 			err = -EINVAL;
1460 			goto destroy_macvlan_port;
1461 		}
1462 		macvlan_set_passthru(port);
1463 		eth_hw_addr_inherit(dev, lowerdev);
1464 	}
1465 
1466 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1467 		if (vlan->mode != MACVLAN_MODE_SOURCE) {
1468 			err = -EINVAL;
1469 			goto destroy_macvlan_port;
1470 		}
1471 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1472 		err = macvlan_changelink_sources(vlan, macmode, data);
1473 		if (err)
1474 			goto destroy_macvlan_port;
1475 	}
1476 
1477 	err = register_netdevice(dev);
1478 	if (err < 0)
1479 		goto destroy_macvlan_port;
1480 
1481 	dev->priv_flags |= IFF_MACVLAN;
1482 	err = netdev_upper_dev_link(lowerdev, dev, extack);
1483 	if (err)
1484 		goto unregister_netdev;
1485 
1486 	list_add_tail_rcu(&vlan->list, &port->vlans);
1487 	netif_stacked_transfer_operstate(lowerdev, dev);
1488 	linkwatch_fire_event(dev);
1489 
1490 	return 0;
1491 
1492 unregister_netdev:
1493 	/* macvlan_uninit would free the macvlan port */
1494 	unregister_netdevice(dev);
1495 	return err;
1496 destroy_macvlan_port:
1497 	/* the macvlan port may be freed by macvlan_uninit when fail to register.
1498 	 * so we destroy the macvlan port only when it's valid.
1499 	 */
1500 	if (create && macvlan_port_get_rtnl(lowerdev))
1501 		macvlan_port_destroy(port->dev);
1502 	return err;
1503 }
1504 EXPORT_SYMBOL_GPL(macvlan_common_newlink);
1505 
1506 static int macvlan_newlink(struct net *src_net, struct net_device *dev,
1507 			   struct nlattr *tb[], struct nlattr *data[],
1508 			   struct netlink_ext_ack *extack)
1509 {
1510 	return macvlan_common_newlink(src_net, dev, tb, data, extack);
1511 }
1512 
1513 void macvlan_dellink(struct net_device *dev, struct list_head *head)
1514 {
1515 	struct macvlan_dev *vlan = netdev_priv(dev);
1516 
1517 	if (vlan->mode == MACVLAN_MODE_SOURCE)
1518 		macvlan_flush_sources(vlan->port, vlan);
1519 	list_del_rcu(&vlan->list);
1520 	unregister_netdevice_queue(dev, head);
1521 	netdev_upper_dev_unlink(vlan->lowerdev, dev);
1522 }
1523 EXPORT_SYMBOL_GPL(macvlan_dellink);
1524 
1525 static int macvlan_changelink(struct net_device *dev,
1526 			      struct nlattr *tb[], struct nlattr *data[],
1527 			      struct netlink_ext_ack *extack)
1528 {
1529 	struct macvlan_dev *vlan = netdev_priv(dev);
1530 	enum macvlan_mode mode;
1531 	bool set_mode = false;
1532 	enum macvlan_macaddr_mode macmode;
1533 	int ret;
1534 
1535 	/* Validate mode, but don't set yet: setting flags may fail. */
1536 	if (data && data[IFLA_MACVLAN_MODE]) {
1537 		set_mode = true;
1538 		mode = nla_get_u32(data[IFLA_MACVLAN_MODE]);
1539 		/* Passthrough mode can't be set or cleared dynamically */
1540 		if ((mode == MACVLAN_MODE_PASSTHRU) !=
1541 		    (vlan->mode == MACVLAN_MODE_PASSTHRU))
1542 			return -EINVAL;
1543 		if (vlan->mode == MACVLAN_MODE_SOURCE &&
1544 		    vlan->mode != mode)
1545 			macvlan_flush_sources(vlan->port, vlan);
1546 	}
1547 
1548 	if (data && data[IFLA_MACVLAN_FLAGS]) {
1549 		__u16 flags = nla_get_u16(data[IFLA_MACVLAN_FLAGS]);
1550 		bool promisc = (flags ^ vlan->flags) & MACVLAN_FLAG_NOPROMISC;
1551 		if (macvlan_passthru(vlan->port) && promisc) {
1552 			int err;
1553 
1554 			if (flags & MACVLAN_FLAG_NOPROMISC)
1555 				err = dev_set_promiscuity(vlan->lowerdev, -1);
1556 			else
1557 				err = dev_set_promiscuity(vlan->lowerdev, 1);
1558 			if (err < 0)
1559 				return err;
1560 		}
1561 		vlan->flags = flags;
1562 	}
1563 	if (set_mode)
1564 		vlan->mode = mode;
1565 	if (data && data[IFLA_MACVLAN_MACADDR_MODE]) {
1566 		if (vlan->mode != MACVLAN_MODE_SOURCE)
1567 			return -EINVAL;
1568 		macmode = nla_get_u32(data[IFLA_MACVLAN_MACADDR_MODE]);
1569 		ret = macvlan_changelink_sources(vlan, macmode, data);
1570 		if (ret)
1571 			return ret;
1572 	}
1573 	return 0;
1574 }
1575 
1576 static size_t macvlan_get_size_mac(const struct macvlan_dev *vlan)
1577 {
1578 	if (vlan->macaddr_count == 0)
1579 		return 0;
1580 	return nla_total_size(0) /* IFLA_MACVLAN_MACADDR_DATA */
1581 		+ vlan->macaddr_count * nla_total_size(sizeof(u8) * ETH_ALEN);
1582 }
1583 
1584 static size_t macvlan_get_size(const struct net_device *dev)
1585 {
1586 	struct macvlan_dev *vlan = netdev_priv(dev);
1587 
1588 	return (0
1589 		+ nla_total_size(4) /* IFLA_MACVLAN_MODE */
1590 		+ nla_total_size(2) /* IFLA_MACVLAN_FLAGS */
1591 		+ nla_total_size(4) /* IFLA_MACVLAN_MACADDR_COUNT */
1592 		+ macvlan_get_size_mac(vlan) /* IFLA_MACVLAN_MACADDR */
1593 		);
1594 }
1595 
1596 static int macvlan_fill_info_macaddr(struct sk_buff *skb,
1597 				     const struct macvlan_dev *vlan,
1598 				     const int i)
1599 {
1600 	struct hlist_head *h = &vlan->port->vlan_source_hash[i];
1601 	struct macvlan_source_entry *entry;
1602 
1603 	hlist_for_each_entry_rcu(entry, h, hlist) {
1604 		if (entry->vlan != vlan)
1605 			continue;
1606 		if (nla_put(skb, IFLA_MACVLAN_MACADDR, ETH_ALEN, entry->addr))
1607 			return 1;
1608 	}
1609 	return 0;
1610 }
1611 
1612 static int macvlan_fill_info(struct sk_buff *skb,
1613 				const struct net_device *dev)
1614 {
1615 	struct macvlan_dev *vlan = netdev_priv(dev);
1616 	int i;
1617 	struct nlattr *nest;
1618 
1619 	if (nla_put_u32(skb, IFLA_MACVLAN_MODE, vlan->mode))
1620 		goto nla_put_failure;
1621 	if (nla_put_u16(skb, IFLA_MACVLAN_FLAGS, vlan->flags))
1622 		goto nla_put_failure;
1623 	if (nla_put_u32(skb, IFLA_MACVLAN_MACADDR_COUNT, vlan->macaddr_count))
1624 		goto nla_put_failure;
1625 	if (vlan->macaddr_count > 0) {
1626 		nest = nla_nest_start_noflag(skb, IFLA_MACVLAN_MACADDR_DATA);
1627 		if (nest == NULL)
1628 			goto nla_put_failure;
1629 
1630 		for (i = 0; i < MACVLAN_HASH_SIZE; i++) {
1631 			if (macvlan_fill_info_macaddr(skb, vlan, i))
1632 				goto nla_put_failure;
1633 		}
1634 		nla_nest_end(skb, nest);
1635 	}
1636 	return 0;
1637 
1638 nla_put_failure:
1639 	return -EMSGSIZE;
1640 }
1641 
1642 static const struct nla_policy macvlan_policy[IFLA_MACVLAN_MAX + 1] = {
1643 	[IFLA_MACVLAN_MODE]  = { .type = NLA_U32 },
1644 	[IFLA_MACVLAN_FLAGS] = { .type = NLA_U16 },
1645 	[IFLA_MACVLAN_MACADDR_MODE] = { .type = NLA_U32 },
1646 	[IFLA_MACVLAN_MACADDR] = { .type = NLA_BINARY, .len = MAX_ADDR_LEN },
1647 	[IFLA_MACVLAN_MACADDR_DATA] = { .type = NLA_NESTED },
1648 	[IFLA_MACVLAN_MACADDR_COUNT] = { .type = NLA_U32 },
1649 };
1650 
1651 int macvlan_link_register(struct rtnl_link_ops *ops)
1652 {
1653 	/* common fields */
1654 	ops->validate		= macvlan_validate;
1655 	ops->maxtype		= IFLA_MACVLAN_MAX;
1656 	ops->policy		= macvlan_policy;
1657 	ops->changelink		= macvlan_changelink;
1658 	ops->get_size		= macvlan_get_size;
1659 	ops->fill_info		= macvlan_fill_info;
1660 
1661 	return rtnl_link_register(ops);
1662 };
1663 EXPORT_SYMBOL_GPL(macvlan_link_register);
1664 
1665 static struct net *macvlan_get_link_net(const struct net_device *dev)
1666 {
1667 	return dev_net(macvlan_dev_real_dev(dev));
1668 }
1669 
1670 static struct rtnl_link_ops macvlan_link_ops = {
1671 	.kind		= "macvlan",
1672 	.setup		= macvlan_setup,
1673 	.newlink	= macvlan_newlink,
1674 	.dellink	= macvlan_dellink,
1675 	.get_link_net	= macvlan_get_link_net,
1676 	.priv_size      = sizeof(struct macvlan_dev),
1677 };
1678 
1679 static int macvlan_device_event(struct notifier_block *unused,
1680 				unsigned long event, void *ptr)
1681 {
1682 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1683 	struct macvlan_dev *vlan, *next;
1684 	struct macvlan_port *port;
1685 	LIST_HEAD(list_kill);
1686 
1687 	if (!netif_is_macvlan_port(dev))
1688 		return NOTIFY_DONE;
1689 
1690 	port = macvlan_port_get_rtnl(dev);
1691 
1692 	switch (event) {
1693 	case NETDEV_UP:
1694 	case NETDEV_DOWN:
1695 	case NETDEV_CHANGE:
1696 		list_for_each_entry(vlan, &port->vlans, list)
1697 			netif_stacked_transfer_operstate(vlan->lowerdev,
1698 							 vlan->dev);
1699 		break;
1700 	case NETDEV_FEAT_CHANGE:
1701 		list_for_each_entry(vlan, &port->vlans, list) {
1702 			vlan->dev->gso_max_size = dev->gso_max_size;
1703 			vlan->dev->gso_max_segs = dev->gso_max_segs;
1704 			netdev_update_features(vlan->dev);
1705 		}
1706 		break;
1707 	case NETDEV_CHANGEMTU:
1708 		list_for_each_entry(vlan, &port->vlans, list) {
1709 			if (vlan->dev->mtu <= dev->mtu)
1710 				continue;
1711 			dev_set_mtu(vlan->dev, dev->mtu);
1712 		}
1713 		break;
1714 	case NETDEV_CHANGEADDR:
1715 		if (!macvlan_passthru(port))
1716 			return NOTIFY_DONE;
1717 
1718 		vlan = list_first_entry_or_null(&port->vlans,
1719 						struct macvlan_dev,
1720 						list);
1721 
1722 		if (macvlan_sync_address(vlan->dev, dev->dev_addr))
1723 			return NOTIFY_BAD;
1724 
1725 		break;
1726 	case NETDEV_UNREGISTER:
1727 		/* twiddle thumbs on netns device moves */
1728 		if (dev->reg_state != NETREG_UNREGISTERING)
1729 			break;
1730 
1731 		list_for_each_entry_safe(vlan, next, &port->vlans, list)
1732 			vlan->dev->rtnl_link_ops->dellink(vlan->dev, &list_kill);
1733 		unregister_netdevice_many(&list_kill);
1734 		break;
1735 	case NETDEV_PRE_TYPE_CHANGE:
1736 		/* Forbid underlaying device to change its type. */
1737 		return NOTIFY_BAD;
1738 
1739 	case NETDEV_NOTIFY_PEERS:
1740 	case NETDEV_BONDING_FAILOVER:
1741 	case NETDEV_RESEND_IGMP:
1742 		/* Propagate to all vlans */
1743 		list_for_each_entry(vlan, &port->vlans, list)
1744 			call_netdevice_notifiers(event, vlan->dev);
1745 	}
1746 	return NOTIFY_DONE;
1747 }
1748 
1749 static struct notifier_block macvlan_notifier_block __read_mostly = {
1750 	.notifier_call	= macvlan_device_event,
1751 };
1752 
1753 static int __init macvlan_init_module(void)
1754 {
1755 	int err;
1756 
1757 	register_netdevice_notifier(&macvlan_notifier_block);
1758 
1759 	err = macvlan_link_register(&macvlan_link_ops);
1760 	if (err < 0)
1761 		goto err1;
1762 	return 0;
1763 err1:
1764 	unregister_netdevice_notifier(&macvlan_notifier_block);
1765 	return err;
1766 }
1767 
1768 static void __exit macvlan_cleanup_module(void)
1769 {
1770 	rtnl_link_unregister(&macvlan_link_ops);
1771 	unregister_netdevice_notifier(&macvlan_notifier_block);
1772 }
1773 
1774 module_init(macvlan_init_module);
1775 module_exit(macvlan_cleanup_module);
1776 
1777 MODULE_LICENSE("GPL");
1778 MODULE_AUTHOR("Patrick McHardy <kaber@trash.net>");
1779 MODULE_DESCRIPTION("Driver for MAC address based VLANs");
1780 MODULE_ALIAS_RTNL_LINK("macvlan");
1781