xref: /openbmc/linux/drivers/net/ipvlan/ipvlan_core.c (revision e4781421e883340b796da5a724bda7226817990b)
1 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
2  *
3  * This program is free software; you can redistribute it and/or
4  * modify it under the terms of the GNU General Public License as
5  * published by the Free Software Foundation; either version 2 of
6  * the License, or (at your option) any later version.
7  *
8  */
9 
10 #include "ipvlan.h"
11 
12 static u32 ipvlan_jhash_secret __read_mostly;
13 
14 void ipvlan_init_secret(void)
15 {
16 	net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
17 }
18 
19 static void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
20 			    unsigned int len, bool success, bool mcast)
21 {
22 	if (likely(success)) {
23 		struct ipvl_pcpu_stats *pcptr;
24 
25 		pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
26 		u64_stats_update_begin(&pcptr->syncp);
27 		pcptr->rx_pkts++;
28 		pcptr->rx_bytes += len;
29 		if (mcast)
30 			pcptr->rx_mcast++;
31 		u64_stats_update_end(&pcptr->syncp);
32 	} else {
33 		this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
34 	}
35 }
36 
37 static u8 ipvlan_get_v6_hash(const void *iaddr)
38 {
39 	const struct in6_addr *ip6_addr = iaddr;
40 
41 	return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
42 	       IPVLAN_HASH_MASK;
43 }
44 
45 static u8 ipvlan_get_v4_hash(const void *iaddr)
46 {
47 	const struct in_addr *ip4_addr = iaddr;
48 
49 	return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
50 	       IPVLAN_HASH_MASK;
51 }
52 
53 static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
54 					       const void *iaddr, bool is_v6)
55 {
56 	struct ipvl_addr *addr;
57 	u8 hash;
58 
59 	hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
60 	       ipvlan_get_v4_hash(iaddr);
61 	hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode) {
62 		if (is_v6 && addr->atype == IPVL_IPV6 &&
63 		    ipv6_addr_equal(&addr->ip6addr, iaddr))
64 			return addr;
65 		else if (!is_v6 && addr->atype == IPVL_IPV4 &&
66 			 addr->ip4addr.s_addr ==
67 				((struct in_addr *)iaddr)->s_addr)
68 			return addr;
69 	}
70 	return NULL;
71 }
72 
73 void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
74 {
75 	struct ipvl_port *port = ipvlan->port;
76 	u8 hash;
77 
78 	hash = (addr->atype == IPVL_IPV6) ?
79 	       ipvlan_get_v6_hash(&addr->ip6addr) :
80 	       ipvlan_get_v4_hash(&addr->ip4addr);
81 	if (hlist_unhashed(&addr->hlnode))
82 		hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
83 }
84 
85 void ipvlan_ht_addr_del(struct ipvl_addr *addr)
86 {
87 	hlist_del_init_rcu(&addr->hlnode);
88 }
89 
90 struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
91 				   const void *iaddr, bool is_v6)
92 {
93 	struct ipvl_addr *addr;
94 
95 	list_for_each_entry(addr, &ipvlan->addrs, anode) {
96 		if ((is_v6 && addr->atype == IPVL_IPV6 &&
97 		    ipv6_addr_equal(&addr->ip6addr, iaddr)) ||
98 		    (!is_v6 && addr->atype == IPVL_IPV4 &&
99 		    addr->ip4addr.s_addr == ((struct in_addr *)iaddr)->s_addr))
100 			return addr;
101 	}
102 	return NULL;
103 }
104 
105 bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
106 {
107 	struct ipvl_dev *ipvlan;
108 
109 	ASSERT_RTNL();
110 
111 	list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
112 		if (ipvlan_find_addr(ipvlan, iaddr, is_v6))
113 			return true;
114 	}
115 	return false;
116 }
117 
118 static void *ipvlan_get_L3_hdr(struct sk_buff *skb, int *type)
119 {
120 	void *lyr3h = NULL;
121 
122 	switch (skb->protocol) {
123 	case htons(ETH_P_ARP): {
124 		struct arphdr *arph;
125 
126 		if (unlikely(!pskb_may_pull(skb, sizeof(*arph))))
127 			return NULL;
128 
129 		arph = arp_hdr(skb);
130 		*type = IPVL_ARP;
131 		lyr3h = arph;
132 		break;
133 	}
134 	case htons(ETH_P_IP): {
135 		u32 pktlen;
136 		struct iphdr *ip4h;
137 
138 		if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
139 			return NULL;
140 
141 		ip4h = ip_hdr(skb);
142 		pktlen = ntohs(ip4h->tot_len);
143 		if (ip4h->ihl < 5 || ip4h->version != 4)
144 			return NULL;
145 		if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
146 			return NULL;
147 
148 		*type = IPVL_IPV4;
149 		lyr3h = ip4h;
150 		break;
151 	}
152 	case htons(ETH_P_IPV6): {
153 		struct ipv6hdr *ip6h;
154 
155 		if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
156 			return NULL;
157 
158 		ip6h = ipv6_hdr(skb);
159 		if (ip6h->version != 6)
160 			return NULL;
161 
162 		*type = IPVL_IPV6;
163 		lyr3h = ip6h;
164 		/* Only Neighbour Solicitation pkts need different treatment */
165 		if (ipv6_addr_any(&ip6h->saddr) &&
166 		    ip6h->nexthdr == NEXTHDR_ICMP) {
167 			*type = IPVL_ICMPV6;
168 			lyr3h = ip6h + 1;
169 		}
170 		break;
171 	}
172 	default:
173 		return NULL;
174 	}
175 
176 	return lyr3h;
177 }
178 
179 unsigned int ipvlan_mac_hash(const unsigned char *addr)
180 {
181 	u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
182 			       ipvlan_jhash_secret);
183 
184 	return hash & IPVLAN_MAC_FILTER_MASK;
185 }
186 
187 void ipvlan_process_multicast(struct work_struct *work)
188 {
189 	struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
190 	struct ethhdr *ethh;
191 	struct ipvl_dev *ipvlan;
192 	struct sk_buff *skb, *nskb;
193 	struct sk_buff_head list;
194 	unsigned int len;
195 	unsigned int mac_hash;
196 	int ret;
197 	u8 pkt_type;
198 	bool tx_pkt;
199 
200 	__skb_queue_head_init(&list);
201 
202 	spin_lock_bh(&port->backlog.lock);
203 	skb_queue_splice_tail_init(&port->backlog, &list);
204 	spin_unlock_bh(&port->backlog.lock);
205 
206 	while ((skb = __skb_dequeue(&list)) != NULL) {
207 		struct net_device *dev = skb->dev;
208 		bool consumed = false;
209 
210 		ethh = eth_hdr(skb);
211 		tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
212 		mac_hash = ipvlan_mac_hash(ethh->h_dest);
213 
214 		if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
215 			pkt_type = PACKET_BROADCAST;
216 		else
217 			pkt_type = PACKET_MULTICAST;
218 
219 		rcu_read_lock();
220 		list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
221 			if (tx_pkt && (ipvlan->dev == skb->dev))
222 				continue;
223 			if (!test_bit(mac_hash, ipvlan->mac_filters))
224 				continue;
225 			if (!(ipvlan->dev->flags & IFF_UP))
226 				continue;
227 			ret = NET_RX_DROP;
228 			len = skb->len + ETH_HLEN;
229 			nskb = skb_clone(skb, GFP_ATOMIC);
230 			local_bh_disable();
231 			if (nskb) {
232 				consumed = true;
233 				nskb->pkt_type = pkt_type;
234 				nskb->dev = ipvlan->dev;
235 				if (tx_pkt)
236 					ret = dev_forward_skb(ipvlan->dev, nskb);
237 				else
238 					ret = netif_rx(nskb);
239 			}
240 			ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
241 			local_bh_enable();
242 		}
243 		rcu_read_unlock();
244 
245 		if (tx_pkt) {
246 			/* If the packet originated here, send it out. */
247 			skb->dev = port->dev;
248 			skb->pkt_type = pkt_type;
249 			dev_queue_xmit(skb);
250 		} else {
251 			if (consumed)
252 				consume_skb(skb);
253 			else
254 				kfree_skb(skb);
255 		}
256 		if (dev)
257 			dev_put(dev);
258 	}
259 }
260 
261 static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
262 {
263 	bool xnet = true;
264 
265 	if (dev)
266 		xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
267 
268 	skb_scrub_packet(skb, xnet);
269 	if (dev)
270 		skb->dev = dev;
271 }
272 
273 static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
274 			    bool local)
275 {
276 	struct ipvl_dev *ipvlan = addr->master;
277 	struct net_device *dev = ipvlan->dev;
278 	unsigned int len;
279 	rx_handler_result_t ret = RX_HANDLER_CONSUMED;
280 	bool success = false;
281 	struct sk_buff *skb = *pskb;
282 
283 	len = skb->len + ETH_HLEN;
284 	/* Only packets exchanged between two local slaves need to have
285 	 * device-up check as well as skb-share check.
286 	 */
287 	if (local) {
288 		if (unlikely(!(dev->flags & IFF_UP))) {
289 			kfree_skb(skb);
290 			goto out;
291 		}
292 
293 		skb = skb_share_check(skb, GFP_ATOMIC);
294 		if (!skb)
295 			goto out;
296 
297 		*pskb = skb;
298 	}
299 	ipvlan_skb_crossing_ns(skb, dev);
300 
301 	if (local) {
302 		skb->pkt_type = PACKET_HOST;
303 		if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
304 			success = true;
305 	} else {
306 		ret = RX_HANDLER_ANOTHER;
307 		success = true;
308 	}
309 
310 out:
311 	ipvlan_count_rx(ipvlan, len, success, false);
312 	return ret;
313 }
314 
315 static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
316 					    void *lyr3h, int addr_type,
317 					    bool use_dest)
318 {
319 	struct ipvl_addr *addr = NULL;
320 
321 	if (addr_type == IPVL_IPV6) {
322 		struct ipv6hdr *ip6h;
323 		struct in6_addr *i6addr;
324 
325 		ip6h = (struct ipv6hdr *)lyr3h;
326 		i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
327 		addr = ipvlan_ht_addr_lookup(port, i6addr, true);
328 	} else if (addr_type == IPVL_ICMPV6) {
329 		struct nd_msg *ndmh;
330 		struct in6_addr *i6addr;
331 
332 		/* Make sure that the NeighborSolicitation ICMPv6 packets
333 		 * are handled to avoid DAD issue.
334 		 */
335 		ndmh = (struct nd_msg *)lyr3h;
336 		if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
337 			i6addr = &ndmh->target;
338 			addr = ipvlan_ht_addr_lookup(port, i6addr, true);
339 		}
340 	} else if (addr_type == IPVL_IPV4) {
341 		struct iphdr *ip4h;
342 		__be32 *i4addr;
343 
344 		ip4h = (struct iphdr *)lyr3h;
345 		i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
346 		addr = ipvlan_ht_addr_lookup(port, i4addr, false);
347 	} else if (addr_type == IPVL_ARP) {
348 		struct arphdr *arph;
349 		unsigned char *arp_ptr;
350 		__be32 dip;
351 
352 		arph = (struct arphdr *)lyr3h;
353 		arp_ptr = (unsigned char *)(arph + 1);
354 		if (use_dest)
355 			arp_ptr += (2 * port->dev->addr_len) + 4;
356 		else
357 			arp_ptr += port->dev->addr_len;
358 
359 		memcpy(&dip, arp_ptr, 4);
360 		addr = ipvlan_ht_addr_lookup(port, &dip, false);
361 	}
362 
363 	return addr;
364 }
365 
366 static int ipvlan_process_v4_outbound(struct sk_buff *skb)
367 {
368 	const struct iphdr *ip4h = ip_hdr(skb);
369 	struct net_device *dev = skb->dev;
370 	struct net *net = dev_net(dev);
371 	struct rtable *rt;
372 	int err, ret = NET_XMIT_DROP;
373 	struct flowi4 fl4 = {
374 		.flowi4_oif = dev->ifindex,
375 		.flowi4_tos = RT_TOS(ip4h->tos),
376 		.flowi4_flags = FLOWI_FLAG_ANYSRC,
377 		.daddr = ip4h->daddr,
378 		.saddr = ip4h->saddr,
379 	};
380 
381 	rt = ip_route_output_flow(net, &fl4, NULL);
382 	if (IS_ERR(rt))
383 		goto err;
384 
385 	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
386 		ip_rt_put(rt);
387 		goto err;
388 	}
389 	skb_dst_set(skb, &rt->dst);
390 	err = ip_local_out(net, skb->sk, skb);
391 	if (unlikely(net_xmit_eval(err)))
392 		dev->stats.tx_errors++;
393 	else
394 		ret = NET_XMIT_SUCCESS;
395 	goto out;
396 err:
397 	dev->stats.tx_errors++;
398 	kfree_skb(skb);
399 out:
400 	return ret;
401 }
402 
403 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
404 {
405 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
406 	struct net_device *dev = skb->dev;
407 	struct net *net = dev_net(dev);
408 	struct dst_entry *dst;
409 	int err, ret = NET_XMIT_DROP;
410 	struct flowi6 fl6 = {
411 		.flowi6_iif = dev->ifindex,
412 		.daddr = ip6h->daddr,
413 		.saddr = ip6h->saddr,
414 		.flowi6_flags = FLOWI_FLAG_ANYSRC,
415 		.flowlabel = ip6_flowinfo(ip6h),
416 		.flowi6_mark = skb->mark,
417 		.flowi6_proto = ip6h->nexthdr,
418 	};
419 
420 	dst = ip6_route_output(net, NULL, &fl6);
421 	if (dst->error) {
422 		ret = dst->error;
423 		dst_release(dst);
424 		goto err;
425 	}
426 	skb_dst_set(skb, dst);
427 	err = ip6_local_out(net, skb->sk, skb);
428 	if (unlikely(net_xmit_eval(err)))
429 		dev->stats.tx_errors++;
430 	else
431 		ret = NET_XMIT_SUCCESS;
432 	goto out;
433 err:
434 	dev->stats.tx_errors++;
435 	kfree_skb(skb);
436 out:
437 	return ret;
438 }
439 
440 static int ipvlan_process_outbound(struct sk_buff *skb)
441 {
442 	struct ethhdr *ethh = eth_hdr(skb);
443 	int ret = NET_XMIT_DROP;
444 
445 	/* In this mode we dont care about multicast and broadcast traffic */
446 	if (is_multicast_ether_addr(ethh->h_dest)) {
447 		pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
448 				    ntohs(skb->protocol));
449 		kfree_skb(skb);
450 		goto out;
451 	}
452 
453 	/* The ipvlan is a pseudo-L2 device, so the packets that we receive
454 	 * will have L2; which need to discarded and processed further
455 	 * in the net-ns of the main-device.
456 	 */
457 	if (skb_mac_header_was_set(skb)) {
458 		skb_pull(skb, sizeof(*ethh));
459 		skb->mac_header = (typeof(skb->mac_header))~0U;
460 		skb_reset_network_header(skb);
461 	}
462 
463 	if (skb->protocol == htons(ETH_P_IPV6))
464 		ret = ipvlan_process_v6_outbound(skb);
465 	else if (skb->protocol == htons(ETH_P_IP))
466 		ret = ipvlan_process_v4_outbound(skb);
467 	else {
468 		pr_warn_ratelimited("Dropped outbound packet type=%x\n",
469 				    ntohs(skb->protocol));
470 		kfree_skb(skb);
471 	}
472 out:
473 	return ret;
474 }
475 
476 static void ipvlan_multicast_enqueue(struct ipvl_port *port,
477 				     struct sk_buff *skb, bool tx_pkt)
478 {
479 	if (skb->protocol == htons(ETH_P_PAUSE)) {
480 		kfree_skb(skb);
481 		return;
482 	}
483 
484 	/* Record that the deferred packet is from TX or RX path. By
485 	 * looking at mac-addresses on packet will lead to erronus decisions.
486 	 * (This would be true for a loopback-mode on master device or a
487 	 * hair-pin mode of the switch.)
488 	 */
489 	IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
490 
491 	spin_lock(&port->backlog.lock);
492 	if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
493 		if (skb->dev)
494 			dev_hold(skb->dev);
495 		__skb_queue_tail(&port->backlog, skb);
496 		spin_unlock(&port->backlog.lock);
497 		schedule_work(&port->wq);
498 	} else {
499 		spin_unlock(&port->backlog.lock);
500 		atomic_long_inc(&skb->dev->rx_dropped);
501 		kfree_skb(skb);
502 	}
503 }
504 
505 static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
506 {
507 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
508 	void *lyr3h;
509 	struct ipvl_addr *addr;
510 	int addr_type;
511 
512 	lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
513 	if (!lyr3h)
514 		goto out;
515 
516 	addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
517 	if (addr)
518 		return ipvlan_rcv_frame(addr, &skb, true);
519 
520 out:
521 	ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
522 	return ipvlan_process_outbound(skb);
523 }
524 
525 static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
526 {
527 	const struct ipvl_dev *ipvlan = netdev_priv(dev);
528 	struct ethhdr *eth = eth_hdr(skb);
529 	struct ipvl_addr *addr;
530 	void *lyr3h;
531 	int addr_type;
532 
533 	if (ether_addr_equal(eth->h_dest, eth->h_source)) {
534 		lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
535 		if (lyr3h) {
536 			addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
537 			if (addr)
538 				return ipvlan_rcv_frame(addr, &skb, true);
539 		}
540 		skb = skb_share_check(skb, GFP_ATOMIC);
541 		if (!skb)
542 			return NET_XMIT_DROP;
543 
544 		/* Packet definitely does not belong to any of the
545 		 * virtual devices, but the dest is local. So forward
546 		 * the skb for the main-dev. At the RX side we just return
547 		 * RX_PASS for it to be processed further on the stack.
548 		 */
549 		return dev_forward_skb(ipvlan->phy_dev, skb);
550 
551 	} else if (is_multicast_ether_addr(eth->h_dest)) {
552 		ipvlan_skb_crossing_ns(skb, NULL);
553 		ipvlan_multicast_enqueue(ipvlan->port, skb, true);
554 		return NET_XMIT_SUCCESS;
555 	}
556 
557 	ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
558 	return dev_queue_xmit(skb);
559 }
560 
561 int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
562 {
563 	struct ipvl_dev *ipvlan = netdev_priv(dev);
564 	struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
565 
566 	if (!port)
567 		goto out;
568 
569 	if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
570 		goto out;
571 
572 	switch(port->mode) {
573 	case IPVLAN_MODE_L2:
574 		return ipvlan_xmit_mode_l2(skb, dev);
575 	case IPVLAN_MODE_L3:
576 	case IPVLAN_MODE_L3S:
577 		return ipvlan_xmit_mode_l3(skb, dev);
578 	}
579 
580 	/* Should not reach here */
581 	WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
582 			  port->mode);
583 out:
584 	kfree_skb(skb);
585 	return NET_XMIT_DROP;
586 }
587 
588 static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
589 {
590 	struct ethhdr *eth = eth_hdr(skb);
591 	struct ipvl_addr *addr;
592 	void *lyr3h;
593 	int addr_type;
594 
595 	if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
596 		lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
597 		if (!lyr3h)
598 			return true;
599 
600 		addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
601 		if (addr)
602 			return false;
603 	}
604 
605 	return true;
606 }
607 
608 static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
609 						 struct ipvl_port *port)
610 {
611 	void *lyr3h;
612 	int addr_type;
613 	struct ipvl_addr *addr;
614 	struct sk_buff *skb = *pskb;
615 	rx_handler_result_t ret = RX_HANDLER_PASS;
616 
617 	lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
618 	if (!lyr3h)
619 		goto out;
620 
621 	addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
622 	if (addr)
623 		ret = ipvlan_rcv_frame(addr, pskb, false);
624 
625 out:
626 	return ret;
627 }
628 
629 static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
630 						 struct ipvl_port *port)
631 {
632 	struct sk_buff *skb = *pskb;
633 	struct ethhdr *eth = eth_hdr(skb);
634 	rx_handler_result_t ret = RX_HANDLER_PASS;
635 	void *lyr3h;
636 	int addr_type;
637 
638 	if (is_multicast_ether_addr(eth->h_dest)) {
639 		if (ipvlan_external_frame(skb, port)) {
640 			struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
641 
642 			/* External frames are queued for device local
643 			 * distribution, but a copy is given to master
644 			 * straight away to avoid sending duplicates later
645 			 * when work-queue processes this frame. This is
646 			 * achieved by returning RX_HANDLER_PASS.
647 			 */
648 			if (nskb) {
649 				ipvlan_skb_crossing_ns(nskb, NULL);
650 				ipvlan_multicast_enqueue(port, nskb, false);
651 			}
652 		}
653 	} else {
654 		struct ipvl_addr *addr;
655 
656 		lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
657 		if (!lyr3h)
658 			return ret;
659 
660 		addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
661 		if (addr)
662 			ret = ipvlan_rcv_frame(addr, pskb, false);
663 	}
664 
665 	return ret;
666 }
667 
668 rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
669 {
670 	struct sk_buff *skb = *pskb;
671 	struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
672 
673 	if (!port)
674 		return RX_HANDLER_PASS;
675 
676 	switch (port->mode) {
677 	case IPVLAN_MODE_L2:
678 		return ipvlan_handle_mode_l2(pskb, port);
679 	case IPVLAN_MODE_L3:
680 		return ipvlan_handle_mode_l3(pskb, port);
681 	case IPVLAN_MODE_L3S:
682 		return RX_HANDLER_PASS;
683 	}
684 
685 	/* Should not reach here */
686 	WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
687 			  port->mode);
688 	kfree_skb(skb);
689 	return RX_HANDLER_CONSUMED;
690 }
691 
692 static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
693 					    struct net_device *dev)
694 {
695 	struct ipvl_addr *addr = NULL;
696 	struct ipvl_port *port;
697 	void *lyr3h;
698 	int addr_type;
699 
700 	if (!dev || !netif_is_ipvlan_port(dev))
701 		goto out;
702 
703 	port = ipvlan_port_get_rcu(dev);
704 	if (!port || port->mode != IPVLAN_MODE_L3S)
705 		goto out;
706 
707 	lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
708 	if (!lyr3h)
709 		goto out;
710 
711 	addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
712 out:
713 	return addr;
714 }
715 
716 struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
717 			      u16 proto)
718 {
719 	struct ipvl_addr *addr;
720 	struct net_device *sdev;
721 
722 	addr = ipvlan_skb_to_addr(skb, dev);
723 	if (!addr)
724 		goto out;
725 
726 	sdev = addr->master->dev;
727 	switch (proto) {
728 	case AF_INET:
729 	{
730 		int err;
731 		struct iphdr *ip4h = ip_hdr(skb);
732 
733 		err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
734 					   ip4h->tos, sdev);
735 		if (unlikely(err))
736 			goto out;
737 		break;
738 	}
739 	case AF_INET6:
740 	{
741 		struct dst_entry *dst;
742 		struct ipv6hdr *ip6h = ipv6_hdr(skb);
743 		int flags = RT6_LOOKUP_F_HAS_SADDR;
744 		struct flowi6 fl6 = {
745 			.flowi6_iif   = sdev->ifindex,
746 			.daddr        = ip6h->daddr,
747 			.saddr        = ip6h->saddr,
748 			.flowlabel    = ip6_flowinfo(ip6h),
749 			.flowi6_mark  = skb->mark,
750 			.flowi6_proto = ip6h->nexthdr,
751 		};
752 
753 		skb_dst_drop(skb);
754 		dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6, flags);
755 		skb_dst_set(skb, dst);
756 		break;
757 	}
758 	default:
759 		break;
760 	}
761 
762 out:
763 	return skb;
764 }
765 
766 unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
767 			     const struct nf_hook_state *state)
768 {
769 	struct ipvl_addr *addr;
770 	unsigned int len;
771 
772 	addr = ipvlan_skb_to_addr(skb, skb->dev);
773 	if (!addr)
774 		goto out;
775 
776 	skb->dev = addr->master->dev;
777 	len = skb->len + ETH_HLEN;
778 	ipvlan_count_rx(addr->master, len, true, false);
779 out:
780 	return NF_ACCEPT;
781 }
782