Loading...
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
19void 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}
36EXPORT_SYMBOL_GPL(ipvlan_count_rx);
37
38#if IS_ENABLED(CONFIG_IPV6)
39static u8 ipvlan_get_v6_hash(const void *iaddr)
40{
41 const struct in6_addr *ip6_addr = iaddr;
42
43 return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
44 IPVLAN_HASH_MASK;
45}
46#else
47static u8 ipvlan_get_v6_hash(const void *iaddr)
48{
49 return 0;
50}
51#endif
52
53static u8 ipvlan_get_v4_hash(const void *iaddr)
54{
55 const struct in_addr *ip4_addr = iaddr;
56
57 return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
58 IPVLAN_HASH_MASK;
59}
60
61static bool addr_equal(bool is_v6, struct ipvl_addr *addr, const void *iaddr)
62{
63 if (!is_v6 && addr->atype == IPVL_IPV4) {
64 struct in_addr *i4addr = (struct in_addr *)iaddr;
65
66 return addr->ip4addr.s_addr == i4addr->s_addr;
67#if IS_ENABLED(CONFIG_IPV6)
68 } else if (is_v6 && addr->atype == IPVL_IPV6) {
69 struct in6_addr *i6addr = (struct in6_addr *)iaddr;
70
71 return ipv6_addr_equal(&addr->ip6addr, i6addr);
72#endif
73 }
74
75 return false;
76}
77
78static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
79 const void *iaddr, bool is_v6)
80{
81 struct ipvl_addr *addr;
82 u8 hash;
83
84 hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
85 ipvlan_get_v4_hash(iaddr);
86 hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode)
87 if (addr_equal(is_v6, addr, iaddr))
88 return addr;
89 return NULL;
90}
91
92void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
93{
94 struct ipvl_port *port = ipvlan->port;
95 u8 hash;
96
97 hash = (addr->atype == IPVL_IPV6) ?
98 ipvlan_get_v6_hash(&addr->ip6addr) :
99 ipvlan_get_v4_hash(&addr->ip4addr);
100 if (hlist_unhashed(&addr->hlnode))
101 hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
102}
103
104void ipvlan_ht_addr_del(struct ipvl_addr *addr)
105{
106 hlist_del_init_rcu(&addr->hlnode);
107}
108
109struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
110 const void *iaddr, bool is_v6)
111{
112 struct ipvl_addr *addr, *ret = NULL;
113
114 rcu_read_lock();
115 list_for_each_entry_rcu(addr, &ipvlan->addrs, anode) {
116 if (addr_equal(is_v6, addr, iaddr)) {
117 ret = addr;
118 break;
119 }
120 }
121 rcu_read_unlock();
122 return ret;
123}
124
125bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
126{
127 struct ipvl_dev *ipvlan;
128 bool ret = false;
129
130 rcu_read_lock();
131 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
132 if (ipvlan_find_addr(ipvlan, iaddr, is_v6)) {
133 ret = true;
134 break;
135 }
136 }
137 rcu_read_unlock();
138 return ret;
139}
140
141static void *ipvlan_get_L3_hdr(struct ipvl_port *port, struct sk_buff *skb, int *type)
142{
143 void *lyr3h = NULL;
144
145 switch (skb->protocol) {
146 case htons(ETH_P_ARP): {
147 struct arphdr *arph;
148
149 if (unlikely(!pskb_may_pull(skb, arp_hdr_len(port->dev))))
150 return NULL;
151
152 arph = arp_hdr(skb);
153 *type = IPVL_ARP;
154 lyr3h = arph;
155 break;
156 }
157 case htons(ETH_P_IP): {
158 u32 pktlen;
159 struct iphdr *ip4h;
160
161 if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
162 return NULL;
163
164 ip4h = ip_hdr(skb);
165 pktlen = ntohs(ip4h->tot_len);
166 if (ip4h->ihl < 5 || ip4h->version != 4)
167 return NULL;
168 if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
169 return NULL;
170
171 *type = IPVL_IPV4;
172 lyr3h = ip4h;
173 break;
174 }
175#if IS_ENABLED(CONFIG_IPV6)
176 case htons(ETH_P_IPV6): {
177 struct ipv6hdr *ip6h;
178
179 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
180 return NULL;
181
182 ip6h = ipv6_hdr(skb);
183 if (ip6h->version != 6)
184 return NULL;
185
186 *type = IPVL_IPV6;
187 lyr3h = ip6h;
188 /* Only Neighbour Solicitation pkts need different treatment */
189 if (ipv6_addr_any(&ip6h->saddr) &&
190 ip6h->nexthdr == NEXTHDR_ICMP) {
191 struct icmp6hdr *icmph;
192
193 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph))))
194 return NULL;
195
196 ip6h = ipv6_hdr(skb);
197 icmph = (struct icmp6hdr *)(ip6h + 1);
198
199 if (icmph->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
200 /* Need to access the ipv6 address in body */
201 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph)
202 + sizeof(struct in6_addr))))
203 return NULL;
204
205 ip6h = ipv6_hdr(skb);
206 icmph = (struct icmp6hdr *)(ip6h + 1);
207 }
208
209 *type = IPVL_ICMPV6;
210 lyr3h = icmph;
211 }
212 break;
213 }
214#endif
215 default:
216 return NULL;
217 }
218
219 return lyr3h;
220}
221
222unsigned int ipvlan_mac_hash(const unsigned char *addr)
223{
224 u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
225 ipvlan_jhash_secret);
226
227 return hash & IPVLAN_MAC_FILTER_MASK;
228}
229
230void ipvlan_process_multicast(struct work_struct *work)
231{
232 struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
233 struct ethhdr *ethh;
234 struct ipvl_dev *ipvlan;
235 struct sk_buff *skb, *nskb;
236 struct sk_buff_head list;
237 unsigned int len;
238 unsigned int mac_hash;
239 int ret;
240 u8 pkt_type;
241 bool tx_pkt;
242
243 __skb_queue_head_init(&list);
244
245 spin_lock_bh(&port->backlog.lock);
246 skb_queue_splice_tail_init(&port->backlog, &list);
247 spin_unlock_bh(&port->backlog.lock);
248
249 while ((skb = __skb_dequeue(&list)) != NULL) {
250 struct net_device *dev = skb->dev;
251 bool consumed = false;
252
253 ethh = eth_hdr(skb);
254 tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
255 mac_hash = ipvlan_mac_hash(ethh->h_dest);
256
257 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
258 pkt_type = PACKET_BROADCAST;
259 else
260 pkt_type = PACKET_MULTICAST;
261
262 rcu_read_lock();
263 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
264 if (tx_pkt && (ipvlan->dev == skb->dev))
265 continue;
266 if (!test_bit(mac_hash, ipvlan->mac_filters))
267 continue;
268 if (!(ipvlan->dev->flags & IFF_UP))
269 continue;
270 ret = NET_RX_DROP;
271 len = skb->len + ETH_HLEN;
272 nskb = skb_clone(skb, GFP_ATOMIC);
273 local_bh_disable();
274 if (nskb) {
275 consumed = true;
276 nskb->pkt_type = pkt_type;
277 nskb->dev = ipvlan->dev;
278 if (tx_pkt)
279 ret = dev_forward_skb(ipvlan->dev, nskb);
280 else
281 ret = netif_rx(nskb);
282 }
283 ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
284 local_bh_enable();
285 }
286 rcu_read_unlock();
287
288 if (tx_pkt) {
289 /* If the packet originated here, send it out. */
290 skb->dev = port->dev;
291 skb->pkt_type = pkt_type;
292 dev_queue_xmit(skb);
293 } else {
294 if (consumed)
295 consume_skb(skb);
296 else
297 kfree_skb(skb);
298 }
299 if (dev)
300 dev_put(dev);
301 }
302}
303
304static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
305{
306 bool xnet = true;
307
308 if (dev)
309 xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
310
311 skb_scrub_packet(skb, xnet);
312 if (dev)
313 skb->dev = dev;
314}
315
316static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
317 bool local)
318{
319 struct ipvl_dev *ipvlan = addr->master;
320 struct net_device *dev = ipvlan->dev;
321 unsigned int len;
322 rx_handler_result_t ret = RX_HANDLER_CONSUMED;
323 bool success = false;
324 struct sk_buff *skb = *pskb;
325
326 len = skb->len + ETH_HLEN;
327 /* Only packets exchanged between two local slaves need to have
328 * device-up check as well as skb-share check.
329 */
330 if (local) {
331 if (unlikely(!(dev->flags & IFF_UP))) {
332 kfree_skb(skb);
333 goto out;
334 }
335
336 skb = skb_share_check(skb, GFP_ATOMIC);
337 if (!skb)
338 goto out;
339
340 *pskb = skb;
341 }
342
343 if (local) {
344 skb->pkt_type = PACKET_HOST;
345 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
346 success = true;
347 } else {
348 skb->dev = dev;
349 ret = RX_HANDLER_ANOTHER;
350 success = true;
351 }
352
353out:
354 ipvlan_count_rx(ipvlan, len, success, false);
355 return ret;
356}
357
358static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
359 void *lyr3h, int addr_type,
360 bool use_dest)
361{
362 struct ipvl_addr *addr = NULL;
363
364 switch (addr_type) {
365#if IS_ENABLED(CONFIG_IPV6)
366 case IPVL_IPV6: {
367 struct ipv6hdr *ip6h;
368 struct in6_addr *i6addr;
369
370 ip6h = (struct ipv6hdr *)lyr3h;
371 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
372 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
373 break;
374 }
375 case IPVL_ICMPV6: {
376 struct nd_msg *ndmh;
377 struct in6_addr *i6addr;
378
379 /* Make sure that the NeighborSolicitation ICMPv6 packets
380 * are handled to avoid DAD issue.
381 */
382 ndmh = (struct nd_msg *)lyr3h;
383 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
384 i6addr = &ndmh->target;
385 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
386 }
387 break;
388 }
389#endif
390 case IPVL_IPV4: {
391 struct iphdr *ip4h;
392 __be32 *i4addr;
393
394 ip4h = (struct iphdr *)lyr3h;
395 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
396 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
397 break;
398 }
399 case IPVL_ARP: {
400 struct arphdr *arph;
401 unsigned char *arp_ptr;
402 __be32 dip;
403
404 arph = (struct arphdr *)lyr3h;
405 arp_ptr = (unsigned char *)(arph + 1);
406 if (use_dest)
407 arp_ptr += (2 * port->dev->addr_len) + 4;
408 else
409 arp_ptr += port->dev->addr_len;
410
411 memcpy(&dip, arp_ptr, 4);
412 addr = ipvlan_ht_addr_lookup(port, &dip, false);
413 break;
414 }
415 }
416
417 return addr;
418}
419
420static int ipvlan_process_v4_outbound(struct sk_buff *skb)
421{
422 const struct iphdr *ip4h = ip_hdr(skb);
423 struct net_device *dev = skb->dev;
424 struct net *net = dev_net(dev);
425 struct rtable *rt;
426 int err, ret = NET_XMIT_DROP;
427 struct flowi4 fl4 = {
428 .flowi4_oif = dev->ifindex,
429 .flowi4_tos = RT_TOS(ip4h->tos),
430 .flowi4_flags = FLOWI_FLAG_ANYSRC,
431 .flowi4_mark = skb->mark,
432 .daddr = ip4h->daddr,
433 .saddr = ip4h->saddr,
434 };
435
436 rt = ip_route_output_flow(net, &fl4, NULL);
437 if (IS_ERR(rt))
438 goto err;
439
440 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
441 ip_rt_put(rt);
442 goto err;
443 }
444 skb_dst_set(skb, &rt->dst);
445 err = ip_local_out(net, skb->sk, skb);
446 if (unlikely(net_xmit_eval(err)))
447 dev->stats.tx_errors++;
448 else
449 ret = NET_XMIT_SUCCESS;
450 goto out;
451err:
452 dev->stats.tx_errors++;
453 kfree_skb(skb);
454out:
455 return ret;
456}
457
458#if IS_ENABLED(CONFIG_IPV6)
459static int ipvlan_process_v6_outbound(struct sk_buff *skb)
460{
461 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
462 struct net_device *dev = skb->dev;
463 struct net *net = dev_net(dev);
464 struct dst_entry *dst;
465 int err, ret = NET_XMIT_DROP;
466 struct flowi6 fl6 = {
467 .flowi6_oif = dev->ifindex,
468 .daddr = ip6h->daddr,
469 .saddr = ip6h->saddr,
470 .flowi6_flags = FLOWI_FLAG_ANYSRC,
471 .flowlabel = ip6_flowinfo(ip6h),
472 .flowi6_mark = skb->mark,
473 .flowi6_proto = ip6h->nexthdr,
474 };
475
476 dst = ip6_route_output(net, NULL, &fl6);
477 if (dst->error) {
478 ret = dst->error;
479 dst_release(dst);
480 goto err;
481 }
482 skb_dst_set(skb, dst);
483 err = ip6_local_out(net, skb->sk, skb);
484 if (unlikely(net_xmit_eval(err)))
485 dev->stats.tx_errors++;
486 else
487 ret = NET_XMIT_SUCCESS;
488 goto out;
489err:
490 dev->stats.tx_errors++;
491 kfree_skb(skb);
492out:
493 return ret;
494}
495#else
496static int ipvlan_process_v6_outbound(struct sk_buff *skb)
497{
498 return NET_XMIT_DROP;
499}
500#endif
501
502static int ipvlan_process_outbound(struct sk_buff *skb)
503{
504 struct ethhdr *ethh = eth_hdr(skb);
505 int ret = NET_XMIT_DROP;
506
507 /* In this mode we dont care about multicast and broadcast traffic */
508 if (is_multicast_ether_addr(ethh->h_dest)) {
509 pr_debug_ratelimited("Dropped {multi|broad}cast of type=[%x]\n",
510 ntohs(skb->protocol));
511 kfree_skb(skb);
512 goto out;
513 }
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 skb_pull(skb, sizeof(*ethh));
521 skb->mac_header = (typeof(skb->mac_header))~0U;
522 skb_reset_network_header(skb);
523 }
524
525 if (skb->protocol == htons(ETH_P_IPV6))
526 ret = ipvlan_process_v6_outbound(skb);
527 else if (skb->protocol == htons(ETH_P_IP))
528 ret = ipvlan_process_v4_outbound(skb);
529 else {
530 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
531 ntohs(skb->protocol));
532 kfree_skb(skb);
533 }
534out:
535 return ret;
536}
537
538static void ipvlan_multicast_enqueue(struct ipvl_port *port,
539 struct sk_buff *skb, bool tx_pkt)
540{
541 if (skb->protocol == htons(ETH_P_PAUSE)) {
542 kfree_skb(skb);
543 return;
544 }
545
546 /* Record that the deferred packet is from TX or RX path. By
547 * looking at mac-addresses on packet will lead to erronus decisions.
548 * (This would be true for a loopback-mode on master device or a
549 * hair-pin mode of the switch.)
550 */
551 IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
552
553 spin_lock(&port->backlog.lock);
554 if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
555 if (skb->dev)
556 dev_hold(skb->dev);
557 __skb_queue_tail(&port->backlog, skb);
558 spin_unlock(&port->backlog.lock);
559 schedule_work(&port->wq);
560 } else {
561 spin_unlock(&port->backlog.lock);
562 atomic_long_inc(&skb->dev->rx_dropped);
563 kfree_skb(skb);
564 }
565}
566
567static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
568{
569 const struct ipvl_dev *ipvlan = netdev_priv(dev);
570 void *lyr3h;
571 struct ipvl_addr *addr;
572 int addr_type;
573
574 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
575 if (!lyr3h)
576 goto out;
577
578 if (!ipvlan_is_vepa(ipvlan->port)) {
579 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
580 if (addr) {
581 if (ipvlan_is_private(ipvlan->port)) {
582 consume_skb(skb);
583 return NET_XMIT_DROP;
584 }
585 return ipvlan_rcv_frame(addr, &skb, true);
586 }
587 }
588out:
589 ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
590 return ipvlan_process_outbound(skb);
591}
592
593static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
594{
595 const struct ipvl_dev *ipvlan = netdev_priv(dev);
596 struct ethhdr *eth = eth_hdr(skb);
597 struct ipvl_addr *addr;
598 void *lyr3h;
599 int addr_type;
600
601 if (!ipvlan_is_vepa(ipvlan->port) &&
602 ether_addr_equal(eth->h_dest, eth->h_source)) {
603 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
604 if (lyr3h) {
605 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
606 if (addr) {
607 if (ipvlan_is_private(ipvlan->port)) {
608 consume_skb(skb);
609 return NET_XMIT_DROP;
610 }
611 return ipvlan_rcv_frame(addr, &skb, true);
612 }
613 }
614 skb = skb_share_check(skb, GFP_ATOMIC);
615 if (!skb)
616 return NET_XMIT_DROP;
617
618 /* Packet definitely does not belong to any of the
619 * virtual devices, but the dest is local. So forward
620 * the skb for the main-dev. At the RX side we just return
621 * RX_PASS for it to be processed further on the stack.
622 */
623 return dev_forward_skb(ipvlan->phy_dev, skb);
624
625 } else if (is_multicast_ether_addr(eth->h_dest)) {
626 ipvlan_skb_crossing_ns(skb, NULL);
627 ipvlan_multicast_enqueue(ipvlan->port, skb, true);
628 return NET_XMIT_SUCCESS;
629 }
630
631 skb->dev = ipvlan->phy_dev;
632 return dev_queue_xmit(skb);
633}
634
635int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
636{
637 struct ipvl_dev *ipvlan = netdev_priv(dev);
638 struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
639
640 if (!port)
641 goto out;
642
643 if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
644 goto out;
645
646 switch(port->mode) {
647 case IPVLAN_MODE_L2:
648 return ipvlan_xmit_mode_l2(skb, dev);
649 case IPVLAN_MODE_L3:
650 case IPVLAN_MODE_L3S:
651 return ipvlan_xmit_mode_l3(skb, dev);
652 }
653
654 /* Should not reach here */
655 WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
656 port->mode);
657out:
658 kfree_skb(skb);
659 return NET_XMIT_DROP;
660}
661
662static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
663{
664 struct ethhdr *eth = eth_hdr(skb);
665 struct ipvl_addr *addr;
666 void *lyr3h;
667 int addr_type;
668
669 if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
670 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
671 if (!lyr3h)
672 return true;
673
674 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
675 if (addr)
676 return false;
677 }
678
679 return true;
680}
681
682static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
683 struct ipvl_port *port)
684{
685 void *lyr3h;
686 int addr_type;
687 struct ipvl_addr *addr;
688 struct sk_buff *skb = *pskb;
689 rx_handler_result_t ret = RX_HANDLER_PASS;
690
691 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
692 if (!lyr3h)
693 goto out;
694
695 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
696 if (addr)
697 ret = ipvlan_rcv_frame(addr, pskb, false);
698
699out:
700 return ret;
701}
702
703static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
704 struct ipvl_port *port)
705{
706 struct sk_buff *skb = *pskb;
707 struct ethhdr *eth = eth_hdr(skb);
708 rx_handler_result_t ret = RX_HANDLER_PASS;
709
710 if (is_multicast_ether_addr(eth->h_dest)) {
711 if (ipvlan_external_frame(skb, port)) {
712 struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
713
714 /* External frames are queued for device local
715 * distribution, but a copy is given to master
716 * straight away to avoid sending duplicates later
717 * when work-queue processes this frame. This is
718 * achieved by returning RX_HANDLER_PASS.
719 */
720 if (nskb) {
721 ipvlan_skb_crossing_ns(nskb, NULL);
722 ipvlan_multicast_enqueue(port, nskb, false);
723 }
724 }
725 } else {
726 /* Perform like l3 mode for non-multicast packet */
727 ret = ipvlan_handle_mode_l3(pskb, port);
728 }
729
730 return ret;
731}
732
733rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
734{
735 struct sk_buff *skb = *pskb;
736 struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
737
738 if (!port)
739 return RX_HANDLER_PASS;
740
741 switch (port->mode) {
742 case IPVLAN_MODE_L2:
743 return ipvlan_handle_mode_l2(pskb, port);
744 case IPVLAN_MODE_L3:
745 return ipvlan_handle_mode_l3(pskb, port);
746 case IPVLAN_MODE_L3S:
747 return RX_HANDLER_PASS;
748 }
749
750 /* Should not reach here */
751 WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
752 port->mode);
753 kfree_skb(skb);
754 return RX_HANDLER_CONSUMED;
755}
756
757static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
758 struct net_device *dev)
759{
760 struct ipvl_addr *addr = NULL;
761 struct ipvl_port *port;
762 void *lyr3h;
763 int addr_type;
764
765 if (!dev || !netif_is_ipvlan_port(dev))
766 goto out;
767
768 port = ipvlan_port_get_rcu(dev);
769 if (!port || port->mode != IPVLAN_MODE_L3S)
770 goto out;
771
772 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
773 if (!lyr3h)
774 goto out;
775
776 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
777out:
778 return addr;
779}
780
781struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
782 u16 proto)
783{
784 struct ipvl_addr *addr;
785 struct net_device *sdev;
786
787 addr = ipvlan_skb_to_addr(skb, dev);
788 if (!addr)
789 goto out;
790
791 sdev = addr->master->dev;
792 switch (proto) {
793 case AF_INET:
794 {
795 int err;
796 struct iphdr *ip4h = ip_hdr(skb);
797
798 err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
799 ip4h->tos, sdev);
800 if (unlikely(err))
801 goto out;
802 break;
803 }
804#if IS_ENABLED(CONFIG_IPV6)
805 case AF_INET6:
806 {
807 struct dst_entry *dst;
808 struct ipv6hdr *ip6h = ipv6_hdr(skb);
809 int flags = RT6_LOOKUP_F_HAS_SADDR;
810 struct flowi6 fl6 = {
811 .flowi6_iif = sdev->ifindex,
812 .daddr = ip6h->daddr,
813 .saddr = ip6h->saddr,
814 .flowlabel = ip6_flowinfo(ip6h),
815 .flowi6_mark = skb->mark,
816 .flowi6_proto = ip6h->nexthdr,
817 };
818
819 skb_dst_drop(skb);
820 dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6,
821 skb, flags);
822 skb_dst_set(skb, dst);
823 break;
824 }
825#endif
826 default:
827 break;
828 }
829
830out:
831 return skb;
832}
833
834unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
835 const struct nf_hook_state *state)
836{
837 struct ipvl_addr *addr;
838 unsigned int len;
839
840 addr = ipvlan_skb_to_addr(skb, skb->dev);
841 if (!addr)
842 goto out;
843
844 skb->dev = addr->master->dev;
845 len = skb->len + ETH_HLEN;
846 ipvlan_count_rx(addr->master, len, true, false);
847out:
848 return NF_ACCEPT;
849}
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 pcptr->rx_pkts++;
23 pcptr->rx_bytes += len;
24 if (mcast)
25 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 = ntohs(ip4h->tot_len);
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 if (dev)
295 dev_put(dev);
296 cond_resched();
297 }
298}
299
300static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
301{
302 bool xnet = true;
303
304 if (dev)
305 xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
306
307 skb_scrub_packet(skb, xnet);
308 if (dev)
309 skb->dev = dev;
310}
311
312static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
313 bool local)
314{
315 struct ipvl_dev *ipvlan = addr->master;
316 struct net_device *dev = ipvlan->dev;
317 unsigned int len;
318 rx_handler_result_t ret = RX_HANDLER_CONSUMED;
319 bool success = false;
320 struct sk_buff *skb = *pskb;
321
322 len = skb->len + ETH_HLEN;
323 /* Only packets exchanged between two local slaves need to have
324 * device-up check as well as skb-share check.
325 */
326 if (local) {
327 if (unlikely(!(dev->flags & IFF_UP))) {
328 kfree_skb(skb);
329 goto out;
330 }
331
332 skb = skb_share_check(skb, GFP_ATOMIC);
333 if (!skb)
334 goto out;
335
336 *pskb = skb;
337 }
338
339 if (local) {
340 skb->pkt_type = PACKET_HOST;
341 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
342 success = true;
343 } else {
344 skb->dev = dev;
345 ret = RX_HANDLER_ANOTHER;
346 success = true;
347 }
348
349out:
350 ipvlan_count_rx(ipvlan, len, success, false);
351 return ret;
352}
353
354struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port, void *lyr3h,
355 int addr_type, bool use_dest)
356{
357 struct ipvl_addr *addr = NULL;
358
359 switch (addr_type) {
360#if IS_ENABLED(CONFIG_IPV6)
361 case IPVL_IPV6: {
362 struct ipv6hdr *ip6h;
363 struct in6_addr *i6addr;
364
365 ip6h = (struct ipv6hdr *)lyr3h;
366 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
367 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
368 break;
369 }
370 case IPVL_ICMPV6: {
371 struct nd_msg *ndmh;
372 struct in6_addr *i6addr;
373
374 /* Make sure that the NeighborSolicitation ICMPv6 packets
375 * are handled to avoid DAD issue.
376 */
377 ndmh = (struct nd_msg *)lyr3h;
378 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
379 i6addr = &ndmh->target;
380 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
381 }
382 break;
383 }
384#endif
385 case IPVL_IPV4: {
386 struct iphdr *ip4h;
387 __be32 *i4addr;
388
389 ip4h = (struct iphdr *)lyr3h;
390 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
391 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
392 break;
393 }
394 case IPVL_ARP: {
395 struct arphdr *arph;
396 unsigned char *arp_ptr;
397 __be32 dip;
398
399 arph = (struct arphdr *)lyr3h;
400 arp_ptr = (unsigned char *)(arph + 1);
401 if (use_dest)
402 arp_ptr += (2 * port->dev->addr_len) + 4;
403 else
404 arp_ptr += port->dev->addr_len;
405
406 memcpy(&dip, arp_ptr, 4);
407 addr = ipvlan_ht_addr_lookup(port, &dip, false);
408 break;
409 }
410 }
411
412 return addr;
413}
414
415static int ipvlan_process_v4_outbound(struct sk_buff *skb)
416{
417 const struct iphdr *ip4h = ip_hdr(skb);
418 struct net_device *dev = skb->dev;
419 struct net *net = dev_net(dev);
420 struct rtable *rt;
421 int err, ret = NET_XMIT_DROP;
422 struct flowi4 fl4 = {
423 .flowi4_oif = dev->ifindex,
424 .flowi4_tos = RT_TOS(ip4h->tos),
425 .flowi4_flags = FLOWI_FLAG_ANYSRC,
426 .flowi4_mark = skb->mark,
427 .daddr = ip4h->daddr,
428 .saddr = ip4h->saddr,
429 };
430
431 rt = ip_route_output_flow(net, &fl4, NULL);
432 if (IS_ERR(rt))
433 goto err;
434
435 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
436 ip_rt_put(rt);
437 goto err;
438 }
439 skb_dst_set(skb, &rt->dst);
440 err = ip_local_out(net, skb->sk, skb);
441 if (unlikely(net_xmit_eval(err)))
442 dev->stats.tx_errors++;
443 else
444 ret = NET_XMIT_SUCCESS;
445 goto out;
446err:
447 dev->stats.tx_errors++;
448 kfree_skb(skb);
449out:
450 return ret;
451}
452
453#if IS_ENABLED(CONFIG_IPV6)
454static int ipvlan_process_v6_outbound(struct sk_buff *skb)
455{
456 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
457 struct net_device *dev = skb->dev;
458 struct net *net = dev_net(dev);
459 struct dst_entry *dst;
460 int err, ret = NET_XMIT_DROP;
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
471 dst = ip6_route_output(net, NULL, &fl6);
472 if (dst->error) {
473 ret = dst->error;
474 dst_release(dst);
475 goto err;
476 }
477 skb_dst_set(skb, dst);
478 err = ip6_local_out(net, skb->sk, skb);
479 if (unlikely(net_xmit_eval(err)))
480 dev->stats.tx_errors++;
481 else
482 ret = NET_XMIT_SUCCESS;
483 goto out;
484err:
485 dev->stats.tx_errors++;
486 kfree_skb(skb);
487out:
488 return ret;
489}
490#else
491static int ipvlan_process_v6_outbound(struct sk_buff *skb)
492{
493 return NET_XMIT_DROP;
494}
495#endif
496
497static int ipvlan_process_outbound(struct sk_buff *skb)
498{
499 struct ethhdr *ethh = eth_hdr(skb);
500 int ret = NET_XMIT_DROP;
501
502 /* The ipvlan is a pseudo-L2 device, so the packets that we receive
503 * will have L2; which need to discarded and processed further
504 * in the net-ns of the main-device.
505 */
506 if (skb_mac_header_was_set(skb)) {
507 /* In this mode we dont care about
508 * multicast and broadcast traffic */
509 if (is_multicast_ether_addr(ethh->h_dest)) {
510 pr_debug_ratelimited(
511 "Dropped {multi|broad}cast of type=[%x]\n",
512 ntohs(skb->protocol));
513 kfree_skb(skb);
514 goto out;
515 }
516
517 skb_pull(skb, sizeof(*ethh));
518 skb->mac_header = (typeof(skb->mac_header))~0U;
519 skb_reset_network_header(skb);
520 }
521
522 if (skb->protocol == htons(ETH_P_IPV6))
523 ret = ipvlan_process_v6_outbound(skb);
524 else if (skb->protocol == htons(ETH_P_IP))
525 ret = ipvlan_process_v4_outbound(skb);
526 else {
527 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
528 ntohs(skb->protocol));
529 kfree_skb(skb);
530 }
531out:
532 return ret;
533}
534
535static void ipvlan_multicast_enqueue(struct ipvl_port *port,
536 struct sk_buff *skb, bool tx_pkt)
537{
538 if (skb->protocol == htons(ETH_P_PAUSE)) {
539 kfree_skb(skb);
540 return;
541 }
542
543 /* Record that the deferred packet is from TX or RX path. By
544 * looking at mac-addresses on packet will lead to erronus decisions.
545 * (This would be true for a loopback-mode on master device or a
546 * hair-pin mode of the switch.)
547 */
548 IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
549
550 spin_lock(&port->backlog.lock);
551 if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
552 if (skb->dev)
553 dev_hold(skb->dev);
554 __skb_queue_tail(&port->backlog, skb);
555 spin_unlock(&port->backlog.lock);
556 schedule_work(&port->wq);
557 } else {
558 spin_unlock(&port->backlog.lock);
559 atomic_long_inc(&skb->dev->rx_dropped);
560 kfree_skb(skb);
561 }
562}
563
564static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
565{
566 const struct ipvl_dev *ipvlan = netdev_priv(dev);
567 void *lyr3h;
568 struct ipvl_addr *addr;
569 int addr_type;
570
571 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
572 if (!lyr3h)
573 goto out;
574
575 if (!ipvlan_is_vepa(ipvlan->port)) {
576 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
577 if (addr) {
578 if (ipvlan_is_private(ipvlan->port)) {
579 consume_skb(skb);
580 return NET_XMIT_DROP;
581 }
582 return ipvlan_rcv_frame(addr, &skb, true);
583 }
584 }
585out:
586 ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
587 return ipvlan_process_outbound(skb);
588}
589
590static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
591{
592 const struct ipvl_dev *ipvlan = netdev_priv(dev);
593 struct ethhdr *eth = eth_hdr(skb);
594 struct ipvl_addr *addr;
595 void *lyr3h;
596 int addr_type;
597
598 if (!ipvlan_is_vepa(ipvlan->port) &&
599 ether_addr_equal(eth->h_dest, eth->h_source)) {
600 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
601 if (lyr3h) {
602 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
603 if (addr) {
604 if (ipvlan_is_private(ipvlan->port)) {
605 consume_skb(skb);
606 return NET_XMIT_DROP;
607 }
608 return ipvlan_rcv_frame(addr, &skb, true);
609 }
610 }
611 skb = skb_share_check(skb, GFP_ATOMIC);
612 if (!skb)
613 return NET_XMIT_DROP;
614
615 /* Packet definitely does not belong to any of the
616 * virtual devices, but the dest is local. So forward
617 * the skb for the main-dev. At the RX side we just return
618 * RX_PASS for it to be processed further on the stack.
619 */
620 return dev_forward_skb(ipvlan->phy_dev, skb);
621
622 } else if (is_multicast_ether_addr(eth->h_dest)) {
623 ipvlan_skb_crossing_ns(skb, NULL);
624 ipvlan_multicast_enqueue(ipvlan->port, skb, true);
625 return NET_XMIT_SUCCESS;
626 }
627
628 skb->dev = ipvlan->phy_dev;
629 return dev_queue_xmit(skb);
630}
631
632int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
633{
634 struct ipvl_dev *ipvlan = netdev_priv(dev);
635 struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
636
637 if (!port)
638 goto out;
639
640 if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
641 goto out;
642
643 switch(port->mode) {
644 case IPVLAN_MODE_L2:
645 return ipvlan_xmit_mode_l2(skb, dev);
646 case IPVLAN_MODE_L3:
647#ifdef CONFIG_IPVLAN_L3S
648 case IPVLAN_MODE_L3S:
649#endif
650 return ipvlan_xmit_mode_l3(skb, dev);
651 }
652
653 /* Should not reach here */
654 WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
655 port->mode);
656out:
657 kfree_skb(skb);
658 return NET_XMIT_DROP;
659}
660
661static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
662{
663 struct ethhdr *eth = eth_hdr(skb);
664 struct ipvl_addr *addr;
665 void *lyr3h;
666 int addr_type;
667
668 if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
669 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
670 if (!lyr3h)
671 return true;
672
673 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
674 if (addr)
675 return false;
676 }
677
678 return true;
679}
680
681static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
682 struct ipvl_port *port)
683{
684 void *lyr3h;
685 int addr_type;
686 struct ipvl_addr *addr;
687 struct sk_buff *skb = *pskb;
688 rx_handler_result_t ret = RX_HANDLER_PASS;
689
690 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
691 if (!lyr3h)
692 goto out;
693
694 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
695 if (addr)
696 ret = ipvlan_rcv_frame(addr, pskb, false);
697
698out:
699 return ret;
700}
701
702static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
703 struct ipvl_port *port)
704{
705 struct sk_buff *skb = *pskb;
706 struct ethhdr *eth = eth_hdr(skb);
707 rx_handler_result_t ret = RX_HANDLER_PASS;
708
709 if (is_multicast_ether_addr(eth->h_dest)) {
710 if (ipvlan_external_frame(skb, port)) {
711 struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
712
713 /* External frames are queued for device local
714 * distribution, but a copy is given to master
715 * straight away to avoid sending duplicates later
716 * when work-queue processes this frame. This is
717 * achieved by returning RX_HANDLER_PASS.
718 */
719 if (nskb) {
720 ipvlan_skb_crossing_ns(nskb, NULL);
721 ipvlan_multicast_enqueue(port, nskb, false);
722 }
723 }
724 } else {
725 /* Perform like l3 mode for non-multicast packet */
726 ret = ipvlan_handle_mode_l3(pskb, port);
727 }
728
729 return ret;
730}
731
732rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
733{
734 struct sk_buff *skb = *pskb;
735 struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
736
737 if (!port)
738 return RX_HANDLER_PASS;
739
740 switch (port->mode) {
741 case IPVLAN_MODE_L2:
742 return ipvlan_handle_mode_l2(pskb, port);
743 case IPVLAN_MODE_L3:
744 return ipvlan_handle_mode_l3(pskb, port);
745#ifdef CONFIG_IPVLAN_L3S
746 case IPVLAN_MODE_L3S:
747 return RX_HANDLER_PASS;
748#endif
749 }
750
751 /* Should not reach here */
752 WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
753 port->mode);
754 kfree_skb(skb);
755 return RX_HANDLER_CONSUMED;
756}