Linux Audio

Check our new training course

Loading...
v6.8
  1// SPDX-License-Identifier: GPL-2.0-or-later
  2/* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
 
 
 
 
 
 
  3 */
  4
  5#include "ipvlan.h"
  6
  7static u32 ipvlan_jhash_secret __read_mostly;
  8
  9void ipvlan_init_secret(void)
 10{
 11	net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
 12}
 13
 14void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
 15			    unsigned int len, bool success, bool mcast)
 16{
 
 
 
 17	if (likely(success)) {
 18		struct ipvl_pcpu_stats *pcptr;
 19
 20		pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
 21		u64_stats_update_begin(&pcptr->syncp);
 22		u64_stats_inc(&pcptr->rx_pkts);
 23		u64_stats_add(&pcptr->rx_bytes, len);
 24		if (mcast)
 25			u64_stats_inc(&pcptr->rx_mcast);
 26		u64_stats_update_end(&pcptr->syncp);
 27	} else {
 28		this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
 29	}
 30}
 31EXPORT_SYMBOL_GPL(ipvlan_count_rx);
 32
 33#if IS_ENABLED(CONFIG_IPV6)
 34static u8 ipvlan_get_v6_hash(const void *iaddr)
 35{
 36	const struct in6_addr *ip6_addr = iaddr;
 37
 38	return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
 39	       IPVLAN_HASH_MASK;
 40}
 41#else
 42static u8 ipvlan_get_v6_hash(const void *iaddr)
 43{
 44	return 0;
 45}
 46#endif
 47
 48static u8 ipvlan_get_v4_hash(const void *iaddr)
 49{
 50	const struct in_addr *ip4_addr = iaddr;
 51
 52	return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
 53	       IPVLAN_HASH_MASK;
 54}
 55
 56static bool addr_equal(bool is_v6, struct ipvl_addr *addr, const void *iaddr)
 57{
 58	if (!is_v6 && addr->atype == IPVL_IPV4) {
 59		struct in_addr *i4addr = (struct in_addr *)iaddr;
 60
 61		return addr->ip4addr.s_addr == i4addr->s_addr;
 62#if IS_ENABLED(CONFIG_IPV6)
 63	} else if (is_v6 && addr->atype == IPVL_IPV6) {
 64		struct in6_addr *i6addr = (struct in6_addr *)iaddr;
 65
 66		return ipv6_addr_equal(&addr->ip6addr, i6addr);
 67#endif
 68	}
 69
 70	return false;
 71}
 72
 73static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
 74					       const void *iaddr, bool is_v6)
 75{
 76	struct ipvl_addr *addr;
 77	u8 hash;
 78
 79	hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
 80	       ipvlan_get_v4_hash(iaddr);
 81	hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode)
 82		if (addr_equal(is_v6, addr, iaddr))
 
 
 
 
 
 83			return addr;
 
 84	return NULL;
 85}
 86
 87void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
 88{
 89	struct ipvl_port *port = ipvlan->port;
 90	u8 hash;
 91
 92	hash = (addr->atype == IPVL_IPV6) ?
 93	       ipvlan_get_v6_hash(&addr->ip6addr) :
 94	       ipvlan_get_v4_hash(&addr->ip4addr);
 95	if (hlist_unhashed(&addr->hlnode))
 96		hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
 97}
 98
 99void ipvlan_ht_addr_del(struct ipvl_addr *addr)
100{
101	hlist_del_init_rcu(&addr->hlnode);
102}
103
104struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
105				   const void *iaddr, bool is_v6)
106{
107	struct ipvl_addr *addr, *ret = NULL;
108
109	rcu_read_lock();
110	list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) {
111		if (addr_equal(is_v6, addr, iaddr)) {
112			ret = addr;
113			break;
114		}
115	}
116	rcu_read_unlock();
117	return ret;
118}
119
120bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
121{
122	struct ipvl_dev *ipvlan;
123	bool ret = false;
124
125	rcu_read_lock();
126	list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
127		if (ipvlan_find_addr(ipvlan, iaddr, is_v6)) {
128			ret = true;
129			break;
130		}
131	}
132	rcu_read_unlock();
133	return ret;
134}
135
136void *ipvlan_get_L3_hdr(struct ipvl_port *port, struct sk_buff *skb, int *type)
137{
138	void *lyr3h = NULL;
139
140	switch (skb->protocol) {
141	case htons(ETH_P_ARP): {
142		struct arphdr *arph;
143
144		if (unlikely(!pskb_may_pull(skb, arp_hdr_len(port->dev))))
145			return NULL;
146
147		arph = arp_hdr(skb);
148		*type = IPVL_ARP;
149		lyr3h = arph;
150		break;
151	}
152	case htons(ETH_P_IP): {
153		u32 pktlen;
154		struct iphdr *ip4h;
155
156		if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
157			return NULL;
158
159		ip4h = ip_hdr(skb);
160		pktlen = skb_ip_totlen(skb);
161		if (ip4h->ihl < 5 || ip4h->version != 4)
162			return NULL;
163		if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
164			return NULL;
165
166		*type = IPVL_IPV4;
167		lyr3h = ip4h;
168		break;
169	}
170#if IS_ENABLED(CONFIG_IPV6)
171	case htons(ETH_P_IPV6): {
172		struct ipv6hdr *ip6h;
173
174		if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
175			return NULL;
176
177		ip6h = ipv6_hdr(skb);
178		if (ip6h->version != 6)
179			return NULL;
180
181		*type = IPVL_IPV6;
182		lyr3h = ip6h;
183		/* Only Neighbour Solicitation pkts need different treatment */
184		if (ipv6_addr_any(&ip6h->saddr) &&
185		    ip6h->nexthdr == NEXTHDR_ICMP) {
186			struct icmp6hdr	*icmph;
187
188			if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph))))
189				return NULL;
190
191			ip6h = ipv6_hdr(skb);
192			icmph = (struct icmp6hdr *)(ip6h + 1);
193
194			if (icmph->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
195				/* Need to access the ipv6 address in body */
196				if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph)
197						+ sizeof(struct in6_addr))))
198					return NULL;
199
200				ip6h = ipv6_hdr(skb);
201				icmph = (struct icmp6hdr *)(ip6h + 1);
202			}
203
204			*type = IPVL_ICMPV6;
205			lyr3h = icmph;
206		}
207		break;
208	}
209#endif
210	default:
211		return NULL;
212	}
213
214	return lyr3h;
215}
216
217unsigned int ipvlan_mac_hash(const unsigned char *addr)
218{
219	u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
220			       ipvlan_jhash_secret);
221
222	return hash & IPVLAN_MAC_FILTER_MASK;
223}
224
225void ipvlan_process_multicast(struct work_struct *work)
226{
227	struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
228	struct ethhdr *ethh;
229	struct ipvl_dev *ipvlan;
230	struct sk_buff *skb, *nskb;
231	struct sk_buff_head list;
232	unsigned int len;
233	unsigned int mac_hash;
234	int ret;
235	u8 pkt_type;
236	bool tx_pkt;
237
238	__skb_queue_head_init(&list);
239
240	spin_lock_bh(&port->backlog.lock);
241	skb_queue_splice_tail_init(&port->backlog, &list);
242	spin_unlock_bh(&port->backlog.lock);
243
244	while ((skb = __skb_dequeue(&list)) != NULL) {
245		struct net_device *dev = skb->dev;
246		bool consumed = false;
247
248		ethh = eth_hdr(skb);
249		tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
250		mac_hash = ipvlan_mac_hash(ethh->h_dest);
251
252		if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
253			pkt_type = PACKET_BROADCAST;
254		else
255			pkt_type = PACKET_MULTICAST;
256
257		rcu_read_lock();
258		list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
259			if (tx_pkt && (ipvlan->dev == skb->dev))
260				continue;
261			if (!test_bit(mac_hash, ipvlan->mac_filters))
262				continue;
263			if (!(ipvlan->dev->flags & IFF_UP))
264				continue;
265			ret = NET_RX_DROP;
266			len = skb->len + ETH_HLEN;
267			nskb = skb_clone(skb, GFP_ATOMIC);
268			local_bh_disable();
269			if (nskb) {
270				consumed = true;
271				nskb->pkt_type = pkt_type;
272				nskb->dev = ipvlan->dev;
273				if (tx_pkt)
274					ret = dev_forward_skb(ipvlan->dev, nskb);
275				else
276					ret = netif_rx(nskb);
277			}
278			ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
279			local_bh_enable();
280		}
281		rcu_read_unlock();
282
283		if (tx_pkt) {
284			/* If the packet originated here, send it out. */
285			skb->dev = port->dev;
286			skb->pkt_type = pkt_type;
287			dev_queue_xmit(skb);
288		} else {
289			if (consumed)
290				consume_skb(skb);
291			else
292				kfree_skb(skb);
293		}
294		dev_put(dev);
295		cond_resched();
296	}
297}
298
299static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
300{
301	bool xnet = true;
302
303	if (dev)
304		xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
305
306	skb_scrub_packet(skb, xnet);
307	if (dev)
308		skb->dev = dev;
309}
310
311static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
312			    bool local)
313{
314	struct ipvl_dev *ipvlan = addr->master;
315	struct net_device *dev = ipvlan->dev;
316	unsigned int len;
317	rx_handler_result_t ret = RX_HANDLER_CONSUMED;
318	bool success = false;
319	struct sk_buff *skb = *pskb;
320
321	len = skb->len + ETH_HLEN;
322	/* Only packets exchanged between two local slaves need to have
323	 * device-up check as well as skb-share check.
324	 */
325	if (local) {
326		if (unlikely(!(dev->flags & IFF_UP))) {
327			kfree_skb(skb);
328			goto out;
329		}
330
331		skb = skb_share_check(skb, GFP_ATOMIC);
332		if (!skb)
333			goto out;
334
335		*pskb = skb;
336	}
 
337
338	if (local) {
339		skb->pkt_type = PACKET_HOST;
340		if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
341			success = true;
342	} else {
343		skb->dev = dev;
344		ret = RX_HANDLER_ANOTHER;
345		success = true;
346	}
347
348out:
349	ipvlan_count_rx(ipvlan, len, success, false);
350	return ret;
351}
352
353struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port, void *lyr3h,
354				     int addr_type, bool use_dest)
 
355{
356	struct ipvl_addr *addr = NULL;
357
358	switch (addr_type) {
359#if IS_ENABLED(CONFIG_IPV6)
360	case IPVL_IPV6: {
361		struct ipv6hdr *ip6h;
362		struct in6_addr *i6addr;
363
364		ip6h = (struct ipv6hdr *)lyr3h;
365		i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
366		addr = ipvlan_ht_addr_lookup(port, i6addr, true);
367		break;
368	}
369	case IPVL_ICMPV6: {
370		struct nd_msg *ndmh;
371		struct in6_addr *i6addr;
372
373		/* Make sure that the NeighborSolicitation ICMPv6 packets
374		 * are handled to avoid DAD issue.
375		 */
376		ndmh = (struct nd_msg *)lyr3h;
377		if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
378			i6addr = &ndmh->target;
379			addr = ipvlan_ht_addr_lookup(port, i6addr, true);
380		}
381		break;
382	}
383#endif
384	case IPVL_IPV4: {
385		struct iphdr *ip4h;
386		__be32 *i4addr;
387
388		ip4h = (struct iphdr *)lyr3h;
389		i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
390		addr = ipvlan_ht_addr_lookup(port, i4addr, false);
391		break;
392	}
393	case IPVL_ARP: {
394		struct arphdr *arph;
395		unsigned char *arp_ptr;
396		__be32 dip;
397
398		arph = (struct arphdr *)lyr3h;
399		arp_ptr = (unsigned char *)(arph + 1);
400		if (use_dest)
401			arp_ptr += (2 * port->dev->addr_len) + 4;
402		else
403			arp_ptr += port->dev->addr_len;
404
405		memcpy(&dip, arp_ptr, 4);
406		addr = ipvlan_ht_addr_lookup(port, &dip, false);
407		break;
408	}
409	}
410
411	return addr;
412}
413
414static noinline_for_stack int ipvlan_process_v4_outbound(struct sk_buff *skb)
415{
416	const struct iphdr *ip4h = ip_hdr(skb);
417	struct net_device *dev = skb->dev;
418	struct net *net = dev_net(dev);
419	struct rtable *rt;
420	int err, ret = NET_XMIT_DROP;
421	struct flowi4 fl4 = {
422		.flowi4_oif = dev->ifindex,
423		.flowi4_tos = RT_TOS(ip4h->tos),
424		.flowi4_flags = FLOWI_FLAG_ANYSRC,
425		.flowi4_mark = skb->mark,
426		.daddr = ip4h->daddr,
427		.saddr = ip4h->saddr,
428	};
429
430	rt = ip_route_output_flow(net, &fl4, NULL);
431	if (IS_ERR(rt))
432		goto err;
433
434	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
435		ip_rt_put(rt);
436		goto err;
437	}
438	skb_dst_set(skb, &rt->dst);
439
440	memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
441
442	err = ip_local_out(net, skb->sk, skb);
443	if (unlikely(net_xmit_eval(err)))
444		DEV_STATS_INC(dev, tx_errors);
445	else
446		ret = NET_XMIT_SUCCESS;
447	goto out;
448err:
449	DEV_STATS_INC(dev, tx_errors);
450	kfree_skb(skb);
451out:
452	return ret;
453}
454
455#if IS_ENABLED(CONFIG_IPV6)
456
457static noinline_for_stack int
458ipvlan_route_v6_outbound(struct net_device *dev, struct sk_buff *skb)
459{
460	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
 
 
 
 
461	struct flowi6 fl6 = {
462		.flowi6_oif = dev->ifindex,
463		.daddr = ip6h->daddr,
464		.saddr = ip6h->saddr,
465		.flowi6_flags = FLOWI_FLAG_ANYSRC,
466		.flowlabel = ip6_flowinfo(ip6h),
467		.flowi6_mark = skb->mark,
468		.flowi6_proto = ip6h->nexthdr,
469	};
470	struct dst_entry *dst;
471	int err;
472
473	dst = ip6_route_output(dev_net(dev), NULL, &fl6);
474	err = dst->error;
475	if (err) {
476		dst_release(dst);
477		return err;
478	}
479	skb_dst_set(skb, dst);
480	return 0;
481}
482
483static int ipvlan_process_v6_outbound(struct sk_buff *skb)
484{
485	struct net_device *dev = skb->dev;
486	int err, ret = NET_XMIT_DROP;
487
488	err = ipvlan_route_v6_outbound(dev, skb);
489	if (unlikely(err)) {
490		DEV_STATS_INC(dev, tx_errors);
491		kfree_skb(skb);
492		return err;
493	}
494
495	memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
496
497	err = ip6_local_out(dev_net(dev), skb->sk, skb);
498	if (unlikely(net_xmit_eval(err)))
499		DEV_STATS_INC(dev, tx_errors);
500	else
501		ret = NET_XMIT_SUCCESS;
 
 
 
 
 
502	return ret;
503}
504#else
505static int ipvlan_process_v6_outbound(struct sk_buff *skb)
506{
507	return NET_XMIT_DROP;
508}
509#endif
510
511static int ipvlan_process_outbound(struct sk_buff *skb)
512{
 
513	int ret = NET_XMIT_DROP;
514
 
 
 
 
 
 
 
 
515	/* The ipvlan is a pseudo-L2 device, so the packets that we receive
516	 * will have L2; which need to discarded and processed further
517	 * in the net-ns of the main-device.
518	 */
519	if (skb_mac_header_was_set(skb)) {
520		/* In this mode we dont care about
521		 * multicast and broadcast traffic */
522		struct ethhdr *ethh = eth_hdr(skb);
523
524		if (is_multicast_ether_addr(ethh->h_dest)) {
525			pr_debug_ratelimited(
526				"Dropped {multi|broad}cast of type=[%x]\n",
527				ntohs(skb->protocol));
528			kfree_skb(skb);
529			goto out;
530		}
531
532		skb_pull(skb, sizeof(*ethh));
533		skb->mac_header = (typeof(skb->mac_header))~0U;
534		skb_reset_network_header(skb);
535	}
536
537	if (skb->protocol == htons(ETH_P_IPV6))
538		ret = ipvlan_process_v6_outbound(skb);
539	else if (skb->protocol == htons(ETH_P_IP))
540		ret = ipvlan_process_v4_outbound(skb);
541	else {
542		pr_warn_ratelimited("Dropped outbound packet type=%x\n",
543				    ntohs(skb->protocol));
544		kfree_skb(skb);
545	}
546out:
547	return ret;
548}
549
550static void ipvlan_multicast_enqueue(struct ipvl_port *port,
551				     struct sk_buff *skb, bool tx_pkt)
552{
553	if (skb->protocol == htons(ETH_P_PAUSE)) {
554		kfree_skb(skb);
555		return;
556	}
557
558	/* Record that the deferred packet is from TX or RX path. By
559	 * looking at mac-addresses on packet will lead to erronus decisions.
560	 * (This would be true for a loopback-mode on master device or a
561	 * hair-pin mode of the switch.)
562	 */
563	IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
564
565	spin_lock(&port->backlog.lock);
566	if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
567		dev_hold(skb->dev);
 
568		__skb_queue_tail(&port->backlog, skb);
569		spin_unlock(&port->backlog.lock);
570		schedule_work(&port->wq);
571	} else {
572		spin_unlock(&port->backlog.lock);
573		dev_core_stats_rx_dropped_inc(skb->dev);
574		kfree_skb(skb);
575	}
576}
577
578static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
579{
580	const struct ipvl_dev *ipvlan = netdev_priv(dev);
581	void *lyr3h;
582	struct ipvl_addr *addr;
583	int addr_type;
584
585	lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
586	if (!lyr3h)
587		goto out;
588
589	if (!ipvlan_is_vepa(ipvlan->port)) {
590		addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
591		if (addr) {
592			if (ipvlan_is_private(ipvlan->port)) {
593				consume_skb(skb);
594				return NET_XMIT_DROP;
595			}
596			ipvlan_rcv_frame(addr, &skb, true);
597			return NET_XMIT_SUCCESS;
598		}
599	}
600out:
601	ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
602	return ipvlan_process_outbound(skb);
603}
604
605static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
606{
607	const struct ipvl_dev *ipvlan = netdev_priv(dev);
608	struct ethhdr *eth = skb_eth_hdr(skb);
609	struct ipvl_addr *addr;
610	void *lyr3h;
611	int addr_type;
612
613	if (!ipvlan_is_vepa(ipvlan->port) &&
614	    ether_addr_equal(eth->h_dest, eth->h_source)) {
615		lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
616		if (lyr3h) {
617			addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
618			if (addr) {
619				if (ipvlan_is_private(ipvlan->port)) {
620					consume_skb(skb);
621					return NET_XMIT_DROP;
622				}
623				ipvlan_rcv_frame(addr, &skb, true);
624				return NET_XMIT_SUCCESS;
625			}
626		}
627		skb = skb_share_check(skb, GFP_ATOMIC);
628		if (!skb)
629			return NET_XMIT_DROP;
630
631		/* Packet definitely does not belong to any of the
632		 * virtual devices, but the dest is local. So forward
633		 * the skb for the main-dev. At the RX side we just return
634		 * RX_PASS for it to be processed further on the stack.
635		 */
636		dev_forward_skb(ipvlan->phy_dev, skb);
637		return NET_XMIT_SUCCESS;
638
639	} else if (is_multicast_ether_addr(eth->h_dest)) {
640		skb_reset_mac_header(skb);
641		ipvlan_skb_crossing_ns(skb, NULL);
642		ipvlan_multicast_enqueue(ipvlan->port, skb, true);
643		return NET_XMIT_SUCCESS;
644	}
645
646	skb->dev = ipvlan->phy_dev;
647	return dev_queue_xmit(skb);
648}
649
650int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
651{
652	struct ipvl_dev *ipvlan = netdev_priv(dev);
653	struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
654
655	if (!port)
656		goto out;
657
658	if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
659		goto out;
660
661	switch(port->mode) {
662	case IPVLAN_MODE_L2:
663		return ipvlan_xmit_mode_l2(skb, dev);
664	case IPVLAN_MODE_L3:
665#ifdef CONFIG_IPVLAN_L3S
666	case IPVLAN_MODE_L3S:
667#endif
668		return ipvlan_xmit_mode_l3(skb, dev);
669	}
670
671	/* Should not reach here */
672	WARN_ONCE(true, "%s called for mode = [%x]\n", __func__, port->mode);
 
673out:
674	kfree_skb(skb);
675	return NET_XMIT_DROP;
676}
677
678static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
679{
680	struct ethhdr *eth = eth_hdr(skb);
681	struct ipvl_addr *addr;
682	void *lyr3h;
683	int addr_type;
684
685	if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
686		lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
687		if (!lyr3h)
688			return true;
689
690		addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
691		if (addr)
692			return false;
693	}
694
695	return true;
696}
697
698static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
699						 struct ipvl_port *port)
700{
701	void *lyr3h;
702	int addr_type;
703	struct ipvl_addr *addr;
704	struct sk_buff *skb = *pskb;
705	rx_handler_result_t ret = RX_HANDLER_PASS;
706
707	lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
708	if (!lyr3h)
709		goto out;
710
711	addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
712	if (addr)
713		ret = ipvlan_rcv_frame(addr, pskb, false);
714
715out:
716	return ret;
717}
718
719static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
720						 struct ipvl_port *port)
721{
722	struct sk_buff *skb = *pskb;
723	struct ethhdr *eth = eth_hdr(skb);
724	rx_handler_result_t ret = RX_HANDLER_PASS;
 
 
725
726	if (is_multicast_ether_addr(eth->h_dest)) {
727		if (ipvlan_external_frame(skb, port)) {
728			struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
729
730			/* External frames are queued for device local
731			 * distribution, but a copy is given to master
732			 * straight away to avoid sending duplicates later
733			 * when work-queue processes this frame. This is
734			 * achieved by returning RX_HANDLER_PASS.
735			 */
736			if (nskb) {
737				ipvlan_skb_crossing_ns(nskb, NULL);
738				ipvlan_multicast_enqueue(port, nskb, false);
739			}
740		}
741	} else {
742		/* Perform like l3 mode for non-multicast packet */
743		ret = ipvlan_handle_mode_l3(pskb, port);
 
 
 
 
 
 
 
744	}
745
746	return ret;
747}
748
749rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
750{
751	struct sk_buff *skb = *pskb;
752	struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
753
754	if (!port)
755		return RX_HANDLER_PASS;
756
757	switch (port->mode) {
758	case IPVLAN_MODE_L2:
759		return ipvlan_handle_mode_l2(pskb, port);
760	case IPVLAN_MODE_L3:
761		return ipvlan_handle_mode_l3(pskb, port);
762#ifdef CONFIG_IPVLAN_L3S
763	case IPVLAN_MODE_L3S:
764		return RX_HANDLER_PASS;
765#endif
766	}
767
768	/* Should not reach here */
769	WARN_ONCE(true, "%s called for mode = [%x]\n", __func__, port->mode);
 
770	kfree_skb(skb);
771	return RX_HANDLER_CONSUMED;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
772}
v4.10.11
 
  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
 12static u32 ipvlan_jhash_secret __read_mostly;
 13
 14void ipvlan_init_secret(void)
 15{
 16	net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
 17}
 18
 19static void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
 20			    unsigned int len, bool success, bool mcast)
 21{
 22	if (!ipvlan)
 23		return;
 24
 25	if (likely(success)) {
 26		struct ipvl_pcpu_stats *pcptr;
 27
 28		pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
 29		u64_stats_update_begin(&pcptr->syncp);
 30		pcptr->rx_pkts++;
 31		pcptr->rx_bytes += len;
 32		if (mcast)
 33			pcptr->rx_mcast++;
 34		u64_stats_update_end(&pcptr->syncp);
 35	} else {
 36		this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
 37	}
 38}
 
 39
 
 40static u8 ipvlan_get_v6_hash(const void *iaddr)
 41{
 42	const struct in6_addr *ip6_addr = iaddr;
 43
 44	return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
 45	       IPVLAN_HASH_MASK;
 46}
 
 
 
 
 
 
 47
 48static u8 ipvlan_get_v4_hash(const void *iaddr)
 49{
 50	const struct in_addr *ip4_addr = iaddr;
 51
 52	return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
 53	       IPVLAN_HASH_MASK;
 54}
 55
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 56static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
 57					       const void *iaddr, bool is_v6)
 58{
 59	struct ipvl_addr *addr;
 60	u8 hash;
 61
 62	hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
 63	       ipvlan_get_v4_hash(iaddr);
 64	hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode) {
 65		if (is_v6 && addr->atype == IPVL_IPV6 &&
 66		    ipv6_addr_equal(&addr->ip6addr, iaddr))
 67			return addr;
 68		else if (!is_v6 && addr->atype == IPVL_IPV4 &&
 69			 addr->ip4addr.s_addr ==
 70				((struct in_addr *)iaddr)->s_addr)
 71			return addr;
 72	}
 73	return NULL;
 74}
 75
 76void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
 77{
 78	struct ipvl_port *port = ipvlan->port;
 79	u8 hash;
 80
 81	hash = (addr->atype == IPVL_IPV6) ?
 82	       ipvlan_get_v6_hash(&addr->ip6addr) :
 83	       ipvlan_get_v4_hash(&addr->ip4addr);
 84	if (hlist_unhashed(&addr->hlnode))
 85		hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
 86}
 87
 88void ipvlan_ht_addr_del(struct ipvl_addr *addr)
 89{
 90	hlist_del_init_rcu(&addr->hlnode);
 91}
 92
 93struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
 94				   const void *iaddr, bool is_v6)
 95{
 96	struct ipvl_addr *addr;
 97
 98	list_for_each_entry(addr, &ipvlan->addrs, anode) {
 99		if ((is_v6 && addr->atype == IPVL_IPV6 &&
100		    ipv6_addr_equal(&addr->ip6addr, iaddr)) ||
101		    (!is_v6 && addr->atype == IPVL_IPV4 &&
102		    addr->ip4addr.s_addr == ((struct in_addr *)iaddr)->s_addr))
103			return addr;
104	}
105	return NULL;
 
106}
107
108bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
109{
110	struct ipvl_dev *ipvlan;
 
111
112	ASSERT_RTNL();
113
114	list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
115		if (ipvlan_find_addr(ipvlan, iaddr, is_v6))
116			return true;
 
117	}
118	return false;
 
119}
120
121static void *ipvlan_get_L3_hdr(struct sk_buff *skb, int *type)
122{
123	void *lyr3h = NULL;
124
125	switch (skb->protocol) {
126	case htons(ETH_P_ARP): {
127		struct arphdr *arph;
128
129		if (unlikely(!pskb_may_pull(skb, sizeof(*arph))))
130			return NULL;
131
132		arph = arp_hdr(skb);
133		*type = IPVL_ARP;
134		lyr3h = arph;
135		break;
136	}
137	case htons(ETH_P_IP): {
138		u32 pktlen;
139		struct iphdr *ip4h;
140
141		if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
142			return NULL;
143
144		ip4h = ip_hdr(skb);
145		pktlen = ntohs(ip4h->tot_len);
146		if (ip4h->ihl < 5 || ip4h->version != 4)
147			return NULL;
148		if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
149			return NULL;
150
151		*type = IPVL_IPV4;
152		lyr3h = ip4h;
153		break;
154	}
 
155	case htons(ETH_P_IPV6): {
156		struct ipv6hdr *ip6h;
157
158		if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
159			return NULL;
160
161		ip6h = ipv6_hdr(skb);
162		if (ip6h->version != 6)
163			return NULL;
164
165		*type = IPVL_IPV6;
166		lyr3h = ip6h;
167		/* Only Neighbour Solicitation pkts need different treatment */
168		if (ipv6_addr_any(&ip6h->saddr) &&
169		    ip6h->nexthdr == NEXTHDR_ICMP) {
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
170			*type = IPVL_ICMPV6;
171			lyr3h = ip6h + 1;
172		}
173		break;
174	}
 
175	default:
176		return NULL;
177	}
178
179	return lyr3h;
180}
181
182unsigned int ipvlan_mac_hash(const unsigned char *addr)
183{
184	u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
185			       ipvlan_jhash_secret);
186
187	return hash & IPVLAN_MAC_FILTER_MASK;
188}
189
190void ipvlan_process_multicast(struct work_struct *work)
191{
192	struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
193	struct ethhdr *ethh;
194	struct ipvl_dev *ipvlan;
195	struct sk_buff *skb, *nskb;
196	struct sk_buff_head list;
197	unsigned int len;
198	unsigned int mac_hash;
199	int ret;
200	u8 pkt_type;
201	bool tx_pkt;
202
203	__skb_queue_head_init(&list);
204
205	spin_lock_bh(&port->backlog.lock);
206	skb_queue_splice_tail_init(&port->backlog, &list);
207	spin_unlock_bh(&port->backlog.lock);
208
209	while ((skb = __skb_dequeue(&list)) != NULL) {
210		struct net_device *dev = skb->dev;
211		bool consumed = false;
212
213		ethh = eth_hdr(skb);
214		tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
215		mac_hash = ipvlan_mac_hash(ethh->h_dest);
216
217		if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
218			pkt_type = PACKET_BROADCAST;
219		else
220			pkt_type = PACKET_MULTICAST;
221
222		rcu_read_lock();
223		list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
224			if (tx_pkt && (ipvlan->dev == skb->dev))
225				continue;
226			if (!test_bit(mac_hash, ipvlan->mac_filters))
227				continue;
228			if (!(ipvlan->dev->flags & IFF_UP))
229				continue;
230			ret = NET_RX_DROP;
231			len = skb->len + ETH_HLEN;
232			nskb = skb_clone(skb, GFP_ATOMIC);
233			local_bh_disable();
234			if (nskb) {
235				consumed = true;
236				nskb->pkt_type = pkt_type;
237				nskb->dev = ipvlan->dev;
238				if (tx_pkt)
239					ret = dev_forward_skb(ipvlan->dev, nskb);
240				else
241					ret = netif_rx(nskb);
242			}
243			ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
244			local_bh_enable();
245		}
246		rcu_read_unlock();
247
248		if (tx_pkt) {
249			/* If the packet originated here, send it out. */
250			skb->dev = port->dev;
251			skb->pkt_type = pkt_type;
252			dev_queue_xmit(skb);
253		} else {
254			if (consumed)
255				consume_skb(skb);
256			else
257				kfree_skb(skb);
258		}
259		if (dev)
260			dev_put(dev);
261	}
262}
263
264static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
265{
266	bool xnet = true;
267
268	if (dev)
269		xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
270
271	skb_scrub_packet(skb, xnet);
272	if (dev)
273		skb->dev = dev;
274}
275
276static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
277			    bool local)
278{
279	struct ipvl_dev *ipvlan = addr->master;
280	struct net_device *dev = ipvlan->dev;
281	unsigned int len;
282	rx_handler_result_t ret = RX_HANDLER_CONSUMED;
283	bool success = false;
284	struct sk_buff *skb = *pskb;
285
286	len = skb->len + ETH_HLEN;
287	/* Only packets exchanged between two local slaves need to have
288	 * device-up check as well as skb-share check.
289	 */
290	if (local) {
291		if (unlikely(!(dev->flags & IFF_UP))) {
292			kfree_skb(skb);
293			goto out;
294		}
295
296		skb = skb_share_check(skb, GFP_ATOMIC);
297		if (!skb)
298			goto out;
299
300		*pskb = skb;
301	}
302	ipvlan_skb_crossing_ns(skb, dev);
303
304	if (local) {
305		skb->pkt_type = PACKET_HOST;
306		if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
307			success = true;
308	} else {
 
309		ret = RX_HANDLER_ANOTHER;
310		success = true;
311	}
312
313out:
314	ipvlan_count_rx(ipvlan, len, success, false);
315	return ret;
316}
317
318static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
319					    void *lyr3h, int addr_type,
320					    bool use_dest)
321{
322	struct ipvl_addr *addr = NULL;
323
324	if (addr_type == IPVL_IPV6) {
 
 
325		struct ipv6hdr *ip6h;
326		struct in6_addr *i6addr;
327
328		ip6h = (struct ipv6hdr *)lyr3h;
329		i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
330		addr = ipvlan_ht_addr_lookup(port, i6addr, true);
331	} else if (addr_type == IPVL_ICMPV6) {
 
 
332		struct nd_msg *ndmh;
333		struct in6_addr *i6addr;
334
335		/* Make sure that the NeighborSolicitation ICMPv6 packets
336		 * are handled to avoid DAD issue.
337		 */
338		ndmh = (struct nd_msg *)lyr3h;
339		if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
340			i6addr = &ndmh->target;
341			addr = ipvlan_ht_addr_lookup(port, i6addr, true);
342		}
343	} else if (addr_type == IPVL_IPV4) {
 
 
 
344		struct iphdr *ip4h;
345		__be32 *i4addr;
346
347		ip4h = (struct iphdr *)lyr3h;
348		i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
349		addr = ipvlan_ht_addr_lookup(port, i4addr, false);
350	} else if (addr_type == IPVL_ARP) {
 
 
351		struct arphdr *arph;
352		unsigned char *arp_ptr;
353		__be32 dip;
354
355		arph = (struct arphdr *)lyr3h;
356		arp_ptr = (unsigned char *)(arph + 1);
357		if (use_dest)
358			arp_ptr += (2 * port->dev->addr_len) + 4;
359		else
360			arp_ptr += port->dev->addr_len;
361
362		memcpy(&dip, arp_ptr, 4);
363		addr = ipvlan_ht_addr_lookup(port, &dip, false);
 
 
364	}
365
366	return addr;
367}
368
369static int ipvlan_process_v4_outbound(struct sk_buff *skb)
370{
371	const struct iphdr *ip4h = ip_hdr(skb);
372	struct net_device *dev = skb->dev;
373	struct net *net = dev_net(dev);
374	struct rtable *rt;
375	int err, ret = NET_XMIT_DROP;
376	struct flowi4 fl4 = {
377		.flowi4_oif = dev->ifindex,
378		.flowi4_tos = RT_TOS(ip4h->tos),
379		.flowi4_flags = FLOWI_FLAG_ANYSRC,
 
380		.daddr = ip4h->daddr,
381		.saddr = ip4h->saddr,
382	};
383
384	rt = ip_route_output_flow(net, &fl4, NULL);
385	if (IS_ERR(rt))
386		goto err;
387
388	if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
389		ip_rt_put(rt);
390		goto err;
391	}
392	skb_dst_set(skb, &rt->dst);
 
 
 
393	err = ip_local_out(net, skb->sk, skb);
394	if (unlikely(net_xmit_eval(err)))
395		dev->stats.tx_errors++;
396	else
397		ret = NET_XMIT_SUCCESS;
398	goto out;
399err:
400	dev->stats.tx_errors++;
401	kfree_skb(skb);
402out:
403	return ret;
404}
405
406static int ipvlan_process_v6_outbound(struct sk_buff *skb)
 
 
 
407{
408	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
409	struct net_device *dev = skb->dev;
410	struct net *net = dev_net(dev);
411	struct dst_entry *dst;
412	int err, ret = NET_XMIT_DROP;
413	struct flowi6 fl6 = {
414		.flowi6_iif = dev->ifindex,
415		.daddr = ip6h->daddr,
416		.saddr = ip6h->saddr,
417		.flowi6_flags = FLOWI_FLAG_ANYSRC,
418		.flowlabel = ip6_flowinfo(ip6h),
419		.flowi6_mark = skb->mark,
420		.flowi6_proto = ip6h->nexthdr,
421	};
 
 
422
423	dst = ip6_route_output(net, NULL, &fl6);
424	if (dst->error) {
425		ret = dst->error;
426		dst_release(dst);
427		goto err;
428	}
429	skb_dst_set(skb, dst);
430	err = ip6_local_out(net, skb->sk, skb);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
431	if (unlikely(net_xmit_eval(err)))
432		dev->stats.tx_errors++;
433	else
434		ret = NET_XMIT_SUCCESS;
435	goto out;
436err:
437	dev->stats.tx_errors++;
438	kfree_skb(skb);
439out:
440	return ret;
441}
 
 
 
 
 
 
442
443static int ipvlan_process_outbound(struct sk_buff *skb)
444{
445	struct ethhdr *ethh = eth_hdr(skb);
446	int ret = NET_XMIT_DROP;
447
448	/* In this mode we dont care about multicast and broadcast traffic */
449	if (is_multicast_ether_addr(ethh->h_dest)) {
450		pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
451				    ntohs(skb->protocol));
452		kfree_skb(skb);
453		goto out;
454	}
455
456	/* The ipvlan is a pseudo-L2 device, so the packets that we receive
457	 * will have L2; which need to discarded and processed further
458	 * in the net-ns of the main-device.
459	 */
460	if (skb_mac_header_was_set(skb)) {
 
 
 
 
 
 
 
 
 
 
 
 
461		skb_pull(skb, sizeof(*ethh));
462		skb->mac_header = (typeof(skb->mac_header))~0U;
463		skb_reset_network_header(skb);
464	}
465
466	if (skb->protocol == htons(ETH_P_IPV6))
467		ret = ipvlan_process_v6_outbound(skb);
468	else if (skb->protocol == htons(ETH_P_IP))
469		ret = ipvlan_process_v4_outbound(skb);
470	else {
471		pr_warn_ratelimited("Dropped outbound packet type=%x\n",
472				    ntohs(skb->protocol));
473		kfree_skb(skb);
474	}
475out:
476	return ret;
477}
478
479static void ipvlan_multicast_enqueue(struct ipvl_port *port,
480				     struct sk_buff *skb, bool tx_pkt)
481{
482	if (skb->protocol == htons(ETH_P_PAUSE)) {
483		kfree_skb(skb);
484		return;
485	}
486
487	/* Record that the deferred packet is from TX or RX path. By
488	 * looking at mac-addresses on packet will lead to erronus decisions.
489	 * (This would be true for a loopback-mode on master device or a
490	 * hair-pin mode of the switch.)
491	 */
492	IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
493
494	spin_lock(&port->backlog.lock);
495	if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
496		if (skb->dev)
497			dev_hold(skb->dev);
498		__skb_queue_tail(&port->backlog, skb);
499		spin_unlock(&port->backlog.lock);
500		schedule_work(&port->wq);
501	} else {
502		spin_unlock(&port->backlog.lock);
503		atomic_long_inc(&skb->dev->rx_dropped);
504		kfree_skb(skb);
505	}
506}
507
508static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
509{
510	const struct ipvl_dev *ipvlan = netdev_priv(dev);
511	void *lyr3h;
512	struct ipvl_addr *addr;
513	int addr_type;
514
515	lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
516	if (!lyr3h)
517		goto out;
518
519	addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
520	if (addr)
521		return ipvlan_rcv_frame(addr, &skb, true);
522
 
 
 
 
 
 
 
523out:
524	ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
525	return ipvlan_process_outbound(skb);
526}
527
528static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
529{
530	const struct ipvl_dev *ipvlan = netdev_priv(dev);
531	struct ethhdr *eth = eth_hdr(skb);
532	struct ipvl_addr *addr;
533	void *lyr3h;
534	int addr_type;
535
536	if (ether_addr_equal(eth->h_dest, eth->h_source)) {
537		lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
 
538		if (lyr3h) {
539			addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
540			if (addr)
541				return ipvlan_rcv_frame(addr, &skb, true);
 
 
 
 
 
 
542		}
543		skb = skb_share_check(skb, GFP_ATOMIC);
544		if (!skb)
545			return NET_XMIT_DROP;
546
547		/* Packet definitely does not belong to any of the
548		 * virtual devices, but the dest is local. So forward
549		 * the skb for the main-dev. At the RX side we just return
550		 * RX_PASS for it to be processed further on the stack.
551		 */
552		return dev_forward_skb(ipvlan->phy_dev, skb);
 
553
554	} else if (is_multicast_ether_addr(eth->h_dest)) {
 
555		ipvlan_skb_crossing_ns(skb, NULL);
556		ipvlan_multicast_enqueue(ipvlan->port, skb, true);
557		return NET_XMIT_SUCCESS;
558	}
559
560	ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
561	return dev_queue_xmit(skb);
562}
563
564int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
565{
566	struct ipvl_dev *ipvlan = netdev_priv(dev);
567	struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
568
569	if (!port)
570		goto out;
571
572	if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
573		goto out;
574
575	switch(port->mode) {
576	case IPVLAN_MODE_L2:
577		return ipvlan_xmit_mode_l2(skb, dev);
578	case IPVLAN_MODE_L3:
 
579	case IPVLAN_MODE_L3S:
 
580		return ipvlan_xmit_mode_l3(skb, dev);
581	}
582
583	/* Should not reach here */
584	WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
585			  port->mode);
586out:
587	kfree_skb(skb);
588	return NET_XMIT_DROP;
589}
590
591static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
592{
593	struct ethhdr *eth = eth_hdr(skb);
594	struct ipvl_addr *addr;
595	void *lyr3h;
596	int addr_type;
597
598	if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
599		lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
600		if (!lyr3h)
601			return true;
602
603		addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
604		if (addr)
605			return false;
606	}
607
608	return true;
609}
610
611static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
612						 struct ipvl_port *port)
613{
614	void *lyr3h;
615	int addr_type;
616	struct ipvl_addr *addr;
617	struct sk_buff *skb = *pskb;
618	rx_handler_result_t ret = RX_HANDLER_PASS;
619
620	lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
621	if (!lyr3h)
622		goto out;
623
624	addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
625	if (addr)
626		ret = ipvlan_rcv_frame(addr, pskb, false);
627
628out:
629	return ret;
630}
631
632static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
633						 struct ipvl_port *port)
634{
635	struct sk_buff *skb = *pskb;
636	struct ethhdr *eth = eth_hdr(skb);
637	rx_handler_result_t ret = RX_HANDLER_PASS;
638	void *lyr3h;
639	int addr_type;
640
641	if (is_multicast_ether_addr(eth->h_dest)) {
642		if (ipvlan_external_frame(skb, port)) {
643			struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
644
645			/* External frames are queued for device local
646			 * distribution, but a copy is given to master
647			 * straight away to avoid sending duplicates later
648			 * when work-queue processes this frame. This is
649			 * achieved by returning RX_HANDLER_PASS.
650			 */
651			if (nskb) {
652				ipvlan_skb_crossing_ns(nskb, NULL);
653				ipvlan_multicast_enqueue(port, nskb, false);
654			}
655		}
656	} else {
657		struct ipvl_addr *addr;
658
659		lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
660		if (!lyr3h)
661			return ret;
662
663		addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
664		if (addr)
665			ret = ipvlan_rcv_frame(addr, pskb, false);
666	}
667
668	return ret;
669}
670
671rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
672{
673	struct sk_buff *skb = *pskb;
674	struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
675
676	if (!port)
677		return RX_HANDLER_PASS;
678
679	switch (port->mode) {
680	case IPVLAN_MODE_L2:
681		return ipvlan_handle_mode_l2(pskb, port);
682	case IPVLAN_MODE_L3:
683		return ipvlan_handle_mode_l3(pskb, port);
 
684	case IPVLAN_MODE_L3S:
685		return RX_HANDLER_PASS;
 
686	}
687
688	/* Should not reach here */
689	WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
690			  port->mode);
691	kfree_skb(skb);
692	return RX_HANDLER_CONSUMED;
693}
694
695static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
696					    struct net_device *dev)
697{
698	struct ipvl_addr *addr = NULL;
699	struct ipvl_port *port;
700	void *lyr3h;
701	int addr_type;
702
703	if (!dev || !netif_is_ipvlan_port(dev))
704		goto out;
705
706	port = ipvlan_port_get_rcu(dev);
707	if (!port || port->mode != IPVLAN_MODE_L3S)
708		goto out;
709
710	lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
711	if (!lyr3h)
712		goto out;
713
714	addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
715out:
716	return addr;
717}
718
719struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
720			      u16 proto)
721{
722	struct ipvl_addr *addr;
723	struct net_device *sdev;
724
725	addr = ipvlan_skb_to_addr(skb, dev);
726	if (!addr)
727		goto out;
728
729	sdev = addr->master->dev;
730	switch (proto) {
731	case AF_INET:
732	{
733		int err;
734		struct iphdr *ip4h = ip_hdr(skb);
735
736		err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
737					   ip4h->tos, sdev);
738		if (unlikely(err))
739			goto out;
740		break;
741	}
742	case AF_INET6:
743	{
744		struct dst_entry *dst;
745		struct ipv6hdr *ip6h = ipv6_hdr(skb);
746		int flags = RT6_LOOKUP_F_HAS_SADDR;
747		struct flowi6 fl6 = {
748			.flowi6_iif   = sdev->ifindex,
749			.daddr        = ip6h->daddr,
750			.saddr        = ip6h->saddr,
751			.flowlabel    = ip6_flowinfo(ip6h),
752			.flowi6_mark  = skb->mark,
753			.flowi6_proto = ip6h->nexthdr,
754		};
755
756		skb_dst_drop(skb);
757		dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6, flags);
758		skb_dst_set(skb, dst);
759		break;
760	}
761	default:
762		break;
763	}
764
765out:
766	return skb;
767}
768
769unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
770			     const struct nf_hook_state *state)
771{
772	struct ipvl_addr *addr;
773	unsigned int len;
774
775	addr = ipvlan_skb_to_addr(skb, skb->dev);
776	if (!addr)
777		goto out;
778
779	skb->dev = addr->master->dev;
780	len = skb->len + ETH_HLEN;
781	ipvlan_count_rx(addr->master, len, true, false);
782out:
783	return NF_ACCEPT;
784}