1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
3 */
4
5 #include <net/flow.h>
6 #include <net/ip.h>
7
8 #include "ipvlan.h"
9
10 static u32 ipvlan_jhash_secret __read_mostly;
11
ipvlan_init_secret(void)12 void ipvlan_init_secret(void)
13 {
14 net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
15 }
16
ipvlan_count_rx(const struct ipvl_dev * ipvlan,unsigned int len,bool success,bool mcast)17 void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
18 unsigned int len, bool success, bool mcast)
19 {
20 if (likely(success)) {
21 struct ipvl_pcpu_stats *pcptr;
22
23 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
24 u64_stats_update_begin(&pcptr->syncp);
25 u64_stats_inc(&pcptr->rx_pkts);
26 u64_stats_add(&pcptr->rx_bytes, len);
27 if (mcast)
28 u64_stats_inc(&pcptr->rx_mcast);
29 u64_stats_update_end(&pcptr->syncp);
30 } else {
31 this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
32 }
33 }
34 EXPORT_SYMBOL_GPL(ipvlan_count_rx);
35
36 #if IS_ENABLED(CONFIG_IPV6)
ipvlan_get_v6_hash(const void * iaddr)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 #else
ipvlan_get_v6_hash(const void * iaddr)45 static u8 ipvlan_get_v6_hash(const void *iaddr)
46 {
47 return 0;
48 }
49 #endif
50
ipvlan_get_v4_hash(const void * iaddr)51 static u8 ipvlan_get_v4_hash(const void *iaddr)
52 {
53 const struct in_addr *ip4_addr = iaddr;
54
55 return jhash_1word((__force u32)ip4_addr->s_addr, ipvlan_jhash_secret) &
56 IPVLAN_HASH_MASK;
57 }
58
addr_equal(bool is_v6,struct ipvl_addr * addr,const void * iaddr)59 static bool addr_equal(bool is_v6, struct ipvl_addr *addr, const void *iaddr)
60 {
61 if (!is_v6 && addr->atype == IPVL_IPV4) {
62 struct in_addr *i4addr = (struct in_addr *)iaddr;
63
64 return addr->ip4addr.s_addr == i4addr->s_addr;
65 #if IS_ENABLED(CONFIG_IPV6)
66 } else if (is_v6 && addr->atype == IPVL_IPV6) {
67 struct in6_addr *i6addr = (struct in6_addr *)iaddr;
68
69 return ipv6_addr_equal(&addr->ip6addr, i6addr);
70 #endif
71 }
72
73 return false;
74 }
75
ipvlan_ht_addr_lookup(const struct ipvl_port * port,const void * iaddr,bool is_v6)76 static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
77 const void *iaddr, bool is_v6)
78 {
79 struct ipvl_addr *addr;
80 u8 hash;
81
82 hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
83 ipvlan_get_v4_hash(iaddr);
84 hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode)
85 if (addr_equal(is_v6, addr, iaddr))
86 return addr;
87 return NULL;
88 }
89
ipvlan_ht_addr_add(struct ipvl_dev * ipvlan,struct ipvl_addr * addr)90 void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
91 {
92 struct ipvl_port *port = ipvlan->port;
93 u8 hash;
94
95 hash = (addr->atype == IPVL_IPV6) ?
96 ipvlan_get_v6_hash(&addr->ip6addr) :
97 ipvlan_get_v4_hash(&addr->ip4addr);
98 if (hlist_unhashed(&addr->hlnode))
99 hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
100 }
101
ipvlan_ht_addr_del(struct ipvl_addr * addr)102 void ipvlan_ht_addr_del(struct ipvl_addr *addr)
103 {
104 hlist_del_init_rcu(&addr->hlnode);
105 }
106
ipvlan_find_addr(const struct ipvl_dev * ipvlan,const void * iaddr,bool is_v6)107 struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
108 const void *iaddr, bool is_v6)
109 {
110 struct ipvl_addr *addr;
111
112 assert_spin_locked(&ipvlan->port->addrs_lock);
113
114 list_for_each_entry(addr, &ipvlan->addrs, anode) {
115 if (addr_equal(is_v6, addr, iaddr))
116 return addr;
117 }
118 return NULL;
119 }
120
ipvlan_addr_busy(struct ipvl_port * port,void * iaddr,bool is_v6)121 bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
122 {
123 struct ipvl_dev *ipvlan;
124 bool ret = false;
125
126 rcu_read_lock();
127 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
128 if (ipvlan_find_addr(ipvlan, iaddr, is_v6)) {
129 ret = true;
130 break;
131 }
132 }
133 rcu_read_unlock();
134 return ret;
135 }
136
ipvlan_get_L3_hdr(struct ipvl_port * port,struct sk_buff * skb,int * type)137 void *ipvlan_get_L3_hdr(struct ipvl_port *port, struct sk_buff *skb, int *type)
138 {
139 void *lyr3h = NULL;
140
141 switch (skb->protocol) {
142 case htons(ETH_P_ARP): {
143 struct arphdr *arph;
144
145 if (unlikely(!pskb_may_pull(skb, arp_hdr_len(port->dev))))
146 return NULL;
147
148 arph = arp_hdr(skb);
149 *type = IPVL_ARP;
150 lyr3h = arph;
151 break;
152 }
153 case htons(ETH_P_IP): {
154 u32 pktlen;
155 struct iphdr *ip4h;
156
157 if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
158 return NULL;
159
160 ip4h = ip_hdr(skb);
161 pktlen = skb_ip_totlen(skb);
162 if (ip4h->ihl < 5 || ip4h->version != 4)
163 return NULL;
164 if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
165 return NULL;
166
167 *type = IPVL_IPV4;
168 lyr3h = ip4h;
169 break;
170 }
171 #if IS_ENABLED(CONFIG_IPV6)
172 case htons(ETH_P_IPV6): {
173 struct ipv6hdr *ip6h;
174
175 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
176 return NULL;
177
178 ip6h = ipv6_hdr(skb);
179 if (ip6h->version != 6)
180 return NULL;
181
182 *type = IPVL_IPV6;
183 lyr3h = ip6h;
184 /* Only Neighbour Solicitation pkts need different treatment */
185 if (ipv6_addr_any(&ip6h->saddr) &&
186 ip6h->nexthdr == NEXTHDR_ICMP) {
187 struct icmp6hdr *icmph;
188
189 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph))))
190 return NULL;
191
192 ip6h = ipv6_hdr(skb);
193 icmph = (struct icmp6hdr *)(ip6h + 1);
194
195 if (icmph->icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
196 /* Need to access the ipv6 address in body */
197 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h) + sizeof(*icmph)
198 + sizeof(struct in6_addr))))
199 return NULL;
200
201 ip6h = ipv6_hdr(skb);
202 icmph = (struct icmp6hdr *)(ip6h + 1);
203 }
204
205 *type = IPVL_ICMPV6;
206 lyr3h = icmph;
207 }
208 break;
209 }
210 #endif
211 default:
212 return NULL;
213 }
214
215 return lyr3h;
216 }
217
ipvlan_mac_hash(const unsigned char * addr)218 unsigned int ipvlan_mac_hash(const unsigned char *addr)
219 {
220 u32 hash = jhash_1word(get_unaligned((u32 *)(addr + 2)),
221 ipvlan_jhash_secret);
222
223 return hash & IPVLAN_MAC_FILTER_MASK;
224 }
225
ipvlan_process_multicast(struct work_struct * work)226 void ipvlan_process_multicast(struct work_struct *work)
227 {
228 struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
229 struct ethhdr *ethh;
230 struct ipvl_dev *ipvlan;
231 struct sk_buff *skb, *nskb;
232 struct sk_buff_head list;
233 unsigned int len;
234 unsigned int mac_hash;
235 int ret;
236 u8 pkt_type;
237 bool tx_pkt;
238
239 __skb_queue_head_init(&list);
240
241 spin_lock_bh(&port->backlog.lock);
242 skb_queue_splice_tail_init(&port->backlog, &list);
243 spin_unlock_bh(&port->backlog.lock);
244
245 while ((skb = __skb_dequeue(&list)) != NULL) {
246 struct net_device *dev = skb->dev;
247 bool consumed = false;
248
249 ethh = eth_hdr(skb);
250 tx_pkt = IPVL_SKB_CB(skb)->tx_pkt;
251 mac_hash = ipvlan_mac_hash(ethh->h_dest);
252
253 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
254 pkt_type = PACKET_BROADCAST;
255 else
256 pkt_type = PACKET_MULTICAST;
257
258 rcu_read_lock();
259 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
260 if (tx_pkt && (ipvlan->dev == skb->dev))
261 continue;
262 if (!test_bit(mac_hash, ipvlan->mac_filters))
263 continue;
264 if (!(ipvlan->dev->flags & IFF_UP))
265 continue;
266 ret = NET_RX_DROP;
267 len = skb->len + ETH_HLEN;
268 nskb = skb_clone(skb, GFP_ATOMIC);
269 local_bh_disable();
270 if (nskb) {
271 consumed = true;
272 nskb->pkt_type = pkt_type;
273 nskb->dev = ipvlan->dev;
274 if (tx_pkt)
275 ret = dev_forward_skb(ipvlan->dev, nskb);
276 else
277 ret = netif_rx(nskb);
278 }
279 ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
280 local_bh_enable();
281 }
282 rcu_read_unlock();
283
284 if (tx_pkt) {
285 /* If the packet originated here, send it out. */
286 skb->dev = port->dev;
287 skb->pkt_type = pkt_type;
288 dev_queue_xmit(skb);
289 } else {
290 if (consumed)
291 consume_skb(skb);
292 else
293 kfree_skb(skb);
294 }
295 dev_put(dev);
296 cond_resched();
297 }
298 }
299
ipvlan_skb_crossing_ns(struct sk_buff * skb,struct net_device * dev)300 static 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
ipvlan_rcv_frame(struct ipvl_addr * addr,struct sk_buff ** pskb,bool local)312 static 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
349 out:
350 ipvlan_count_rx(ipvlan, len, success, false);
351 return ret;
352 }
353
ipvlan_addr_lookup(struct ipvl_port * port,void * lyr3h,int addr_type,bool use_dest)354 struct 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
ipvlan_process_v4_outbound(struct sk_buff * skb)415 static noinline_for_stack int ipvlan_process_v4_outbound(struct sk_buff *skb)
416 {
417 struct net_device *dev = skb->dev;
418 struct net *net = dev_net(dev);
419 int err, ret = NET_XMIT_DROP;
420 const struct iphdr *ip4h;
421 struct rtable *rt;
422 struct flowi4 fl4 = {
423 .flowi4_oif = dev->ifindex,
424 .flowi4_flags = FLOWI_FLAG_ANYSRC,
425 .flowi4_mark = skb->mark,
426 };
427
428 if (!pskb_network_may_pull(skb, sizeof(struct iphdr)))
429 goto err;
430
431 ip4h = ip_hdr(skb);
432 fl4.daddr = ip4h->daddr;
433 fl4.saddr = ip4h->saddr;
434 fl4.flowi4_dscp = ip4h_dscp(ip4h);
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
446 memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
447
448 err = ip_local_out(net, NULL, skb);
449 if (unlikely(net_xmit_eval(err)))
450 DEV_STATS_INC(dev, tx_errors);
451 else
452 ret = NET_XMIT_SUCCESS;
453 goto out;
454 err:
455 DEV_STATS_INC(dev, tx_errors);
456 kfree_skb(skb);
457 out:
458 return ret;
459 }
460
461 #if IS_ENABLED(CONFIG_IPV6)
462
463 static noinline_for_stack int
ipvlan_route_v6_outbound(struct net_device * dev,struct sk_buff * skb)464 ipvlan_route_v6_outbound(struct net_device *dev, struct sk_buff *skb)
465 {
466 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
467 struct flowi6 fl6 = {
468 .flowi6_oif = dev->ifindex,
469 .daddr = ip6h->daddr,
470 .saddr = ip6h->saddr,
471 .flowi6_flags = FLOWI_FLAG_ANYSRC,
472 .flowlabel = ip6_flowinfo(ip6h),
473 .flowi6_mark = skb->mark,
474 .flowi6_proto = ip6h->nexthdr,
475 };
476 struct dst_entry *dst;
477 int err;
478
479 dst = ip6_route_output(dev_net(dev), NULL, &fl6);
480 err = dst->error;
481 if (err) {
482 dst_release(dst);
483 return err;
484 }
485 skb_dst_set(skb, dst);
486 return 0;
487 }
488
ipvlan_process_v6_outbound(struct sk_buff * skb)489 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
490 {
491 struct net_device *dev = skb->dev;
492 int err, ret = NET_XMIT_DROP;
493
494 if (!pskb_network_may_pull(skb, sizeof(struct ipv6hdr))) {
495 DEV_STATS_INC(dev, tx_errors);
496 kfree_skb(skb);
497 return ret;
498 }
499
500 err = ipvlan_route_v6_outbound(dev, skb);
501 if (unlikely(err)) {
502 DEV_STATS_INC(dev, tx_errors);
503 kfree_skb(skb);
504 return err;
505 }
506
507 memset(IP6CB(skb), 0, sizeof(*IP6CB(skb)));
508
509 err = ip6_local_out(dev_net(dev), NULL, skb);
510 if (unlikely(net_xmit_eval(err)))
511 DEV_STATS_INC(dev, tx_errors);
512 else
513 ret = NET_XMIT_SUCCESS;
514 return ret;
515 }
516 #else
ipvlan_process_v6_outbound(struct sk_buff * skb)517 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
518 {
519 return NET_XMIT_DROP;
520 }
521 #endif
522
ipvlan_process_outbound(struct sk_buff * skb)523 static int ipvlan_process_outbound(struct sk_buff *skb)
524 {
525 int ret = NET_XMIT_DROP;
526
527 /* The ipvlan is a pseudo-L2 device, so the packets that we receive
528 * will have L2; which need to discarded and processed further
529 * in the net-ns of the main-device.
530 */
531 if (skb_mac_header_was_set(skb)) {
532 /* In this mode we dont care about
533 * multicast and broadcast traffic */
534 struct ethhdr *ethh = eth_hdr(skb);
535
536 if (is_multicast_ether_addr(ethh->h_dest)) {
537 pr_debug_ratelimited(
538 "Dropped {multi|broad}cast of type=[%x]\n",
539 ntohs(skb->protocol));
540 kfree_skb(skb);
541 goto out;
542 }
543
544 skb_pull(skb, sizeof(*ethh));
545 skb->mac_header = (typeof(skb->mac_header))~0U;
546 skb_reset_network_header(skb);
547 }
548
549 if (skb->protocol == htons(ETH_P_IPV6))
550 ret = ipvlan_process_v6_outbound(skb);
551 else if (skb->protocol == htons(ETH_P_IP))
552 ret = ipvlan_process_v4_outbound(skb);
553 else {
554 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
555 ntohs(skb->protocol));
556 kfree_skb(skb);
557 }
558 out:
559 return ret;
560 }
561
ipvlan_multicast_enqueue(struct ipvl_port * port,struct sk_buff * skb,bool tx_pkt)562 static void ipvlan_multicast_enqueue(struct ipvl_port *port,
563 struct sk_buff *skb, bool tx_pkt)
564 {
565 if (skb->protocol == htons(ETH_P_PAUSE)) {
566 kfree_skb(skb);
567 return;
568 }
569
570 /* Record that the deferred packet is from TX or RX path. By
571 * looking at mac-addresses on packet will lead to erronus decisions.
572 * (This would be true for a loopback-mode on master device or a
573 * hair-pin mode of the switch.)
574 */
575 IPVL_SKB_CB(skb)->tx_pkt = tx_pkt;
576
577 spin_lock(&port->backlog.lock);
578 if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
579 dev_hold(skb->dev);
580 __skb_queue_tail(&port->backlog, skb);
581 spin_unlock(&port->backlog.lock);
582 schedule_work(&port->wq);
583 } else {
584 spin_unlock(&port->backlog.lock);
585 dev_core_stats_rx_dropped_inc(skb->dev);
586 kfree_skb(skb);
587 }
588 }
589
ipvlan_xmit_mode_l3(struct sk_buff * skb,struct net_device * dev)590 static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
591 {
592 const struct ipvl_dev *ipvlan = netdev_priv(dev);
593 void *lyr3h;
594 struct ipvl_addr *addr;
595 int addr_type;
596
597 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
598 if (!lyr3h)
599 goto out;
600
601 if (!ipvlan_is_vepa(ipvlan->port)) {
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 ipvlan_rcv_frame(addr, &skb, true);
609 return NET_XMIT_SUCCESS;
610 }
611 }
612 out:
613 ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
614 return ipvlan_process_outbound(skb);
615 }
616
ipvlan_xmit_mode_l2(struct sk_buff * skb,struct net_device * dev)617 static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
618 {
619 const struct ipvl_dev *ipvlan = netdev_priv(dev);
620 struct ethhdr *eth = skb_eth_hdr(skb);
621 struct ipvl_addr *addr;
622 void *lyr3h;
623 int addr_type;
624
625 if (!ipvlan_is_vepa(ipvlan->port) &&
626 ether_addr_equal(eth->h_dest, eth->h_source)) {
627 lyr3h = ipvlan_get_L3_hdr(ipvlan->port, skb, &addr_type);
628 if (lyr3h) {
629 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
630 if (addr) {
631 if (ipvlan_is_private(ipvlan->port)) {
632 consume_skb(skb);
633 return NET_XMIT_DROP;
634 }
635 ipvlan_rcv_frame(addr, &skb, true);
636 return NET_XMIT_SUCCESS;
637 }
638 }
639 skb = skb_share_check(skb, GFP_ATOMIC);
640 if (!skb)
641 return NET_XMIT_DROP;
642
643 /* Packet definitely does not belong to any of the
644 * virtual devices, but the dest is local. So forward
645 * the skb for the main-dev. At the RX side we just return
646 * RX_PASS for it to be processed further on the stack.
647 */
648 dev_forward_skb(ipvlan->phy_dev, skb);
649 return NET_XMIT_SUCCESS;
650
651 } else if (is_multicast_ether_addr(eth->h_dest)) {
652 skb_reset_mac_header(skb);
653 ipvlan_skb_crossing_ns(skb, NULL);
654 ipvlan_multicast_enqueue(ipvlan->port, skb, true);
655 return NET_XMIT_SUCCESS;
656 }
657
658 skb->dev = ipvlan->phy_dev;
659 return dev_queue_xmit(skb);
660 }
661
ipvlan_queue_xmit(struct sk_buff * skb,struct net_device * dev)662 int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
663 {
664 struct ipvl_dev *ipvlan = netdev_priv(dev);
665 struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
666
667 if (!port)
668 goto out;
669
670 if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
671 goto out;
672
673 switch(port->mode) {
674 case IPVLAN_MODE_L2:
675 return ipvlan_xmit_mode_l2(skb, dev);
676 case IPVLAN_MODE_L3:
677 #ifdef CONFIG_IPVLAN_L3S
678 case IPVLAN_MODE_L3S:
679 #endif
680 return ipvlan_xmit_mode_l3(skb, dev);
681 }
682
683 /* Should not reach here */
684 WARN_ONCE(true, "%s called for mode = [%x]\n", __func__, port->mode);
685 out:
686 kfree_skb(skb);
687 return NET_XMIT_DROP;
688 }
689
ipvlan_external_frame(struct sk_buff * skb,struct ipvl_port * port)690 static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
691 {
692 struct ethhdr *eth = eth_hdr(skb);
693 struct ipvl_addr *addr;
694 void *lyr3h;
695 int addr_type;
696
697 if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
698 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
699 if (!lyr3h)
700 return true;
701
702 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
703 if (addr)
704 return false;
705 }
706
707 return true;
708 }
709
ipvlan_handle_mode_l3(struct sk_buff ** pskb,struct ipvl_port * port)710 static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
711 struct ipvl_port *port)
712 {
713 void *lyr3h;
714 int addr_type;
715 struct ipvl_addr *addr;
716 struct sk_buff *skb = *pskb;
717 rx_handler_result_t ret = RX_HANDLER_PASS;
718
719 lyr3h = ipvlan_get_L3_hdr(port, skb, &addr_type);
720 if (!lyr3h)
721 goto out;
722
723 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
724 if (addr)
725 ret = ipvlan_rcv_frame(addr, pskb, false);
726
727 out:
728 return ret;
729 }
730
ipvlan_handle_mode_l2(struct sk_buff ** pskb,struct ipvl_port * port)731 static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
732 struct ipvl_port *port)
733 {
734 struct sk_buff *skb = *pskb;
735 struct ethhdr *eth = eth_hdr(skb);
736 rx_handler_result_t ret = RX_HANDLER_PASS;
737
738 if (unlikely(skb->pkt_type == PACKET_LOOPBACK))
739 return RX_HANDLER_PASS;
740
741 if (is_multicast_ether_addr(eth->h_dest)) {
742 if (ipvlan_external_frame(skb, port)) {
743 struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
744
745 /* External frames are queued for device local
746 * distribution, but a copy is given to master
747 * straight away to avoid sending duplicates later
748 * when work-queue processes this frame. This is
749 * achieved by returning RX_HANDLER_PASS.
750 */
751 if (nskb) {
752 ipvlan_skb_crossing_ns(nskb, NULL);
753 ipvlan_multicast_enqueue(port, nskb, false);
754 }
755 }
756 } else {
757 /* Perform like l3 mode for non-multicast packet */
758 ret = ipvlan_handle_mode_l3(pskb, port);
759 }
760
761 return ret;
762 }
763
ipvlan_handle_frame(struct sk_buff ** pskb)764 rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
765 {
766 struct sk_buff *skb = *pskb;
767 struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
768
769 if (!port)
770 return RX_HANDLER_PASS;
771
772 switch (port->mode) {
773 case IPVLAN_MODE_L2:
774 return ipvlan_handle_mode_l2(pskb, port);
775 case IPVLAN_MODE_L3:
776 return ipvlan_handle_mode_l3(pskb, port);
777 #ifdef CONFIG_IPVLAN_L3S
778 case IPVLAN_MODE_L3S:
779 return RX_HANDLER_PASS;
780 #endif
781 }
782
783 /* Should not reach here */
784 WARN_ONCE(true, "%s called for mode = [%x]\n", __func__, port->mode);
785 kfree_skb(skb);
786 return RX_HANDLER_CONSUMED;
787 }
788