1 /* Copyright (c) 2014 Mahesh Bandewar <maheshb@google.com>
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.
12 static u32 ipvlan_jhash_secret __read_mostly;
14 void ipvlan_init_secret(void)
16 net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
19 static void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
20 unsigned int len, bool success, bool mcast)
25 if (likely(success)) {
26 struct ipvl_pcpu_stats *pcptr;
28 pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
29 u64_stats_update_begin(&pcptr->syncp);
31 pcptr->rx_bytes += len;
34 u64_stats_update_end(&pcptr->syncp);
36 this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
40 static u8 ipvlan_get_v6_hash(const void *iaddr)
42 const struct in6_addr *ip6_addr = iaddr;
44 return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
48 static u8 ipvlan_get_v4_hash(const void *iaddr)
50 const struct in_addr *ip4_addr = iaddr;
52 return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
56 struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
57 const void *iaddr, bool is_v6)
59 struct ipvl_addr *addr;
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))
68 else if (!is_v6 && addr->atype == IPVL_IPV4 &&
69 addr->ip4addr.s_addr ==
70 ((struct in_addr *)iaddr)->s_addr)
76 void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
78 struct ipvl_port *port = ipvlan->port;
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]);
88 void ipvlan_ht_addr_del(struct ipvl_addr *addr)
90 hlist_del_init_rcu(&addr->hlnode);
93 struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
94 const void *iaddr, bool is_v6)
96 struct ipvl_addr *addr;
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))
108 bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
110 struct ipvl_dev *ipvlan;
114 list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
115 if (ipvlan_find_addr(ipvlan, iaddr, is_v6))
121 static void *ipvlan_get_L3_hdr(struct sk_buff *skb, int *type)
125 switch (skb->protocol) {
126 case htons(ETH_P_ARP): {
129 if (unlikely(!pskb_may_pull(skb, sizeof(*arph))))
137 case htons(ETH_P_IP): {
141 if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
145 pktlen = ntohs(ip4h->tot_len);
146 if (ip4h->ihl < 5 || ip4h->version != 4)
148 if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
155 case htons(ETH_P_IPV6): {
156 struct ipv6hdr *ip6h;
158 if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
161 ip6h = ipv6_hdr(skb);
162 if (ip6h->version != 6)
167 /* Only Neighbour Solicitation pkts need different treatment */
168 if (ipv6_addr_any(&ip6h->saddr) &&
169 ip6h->nexthdr == NEXTHDR_ICMP) {
182 unsigned int ipvlan_mac_hash(const unsigned char *addr)
184 u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
185 ipvlan_jhash_secret);
187 return hash & IPVLAN_MAC_FILTER_MASK;
190 void ipvlan_process_multicast(struct work_struct *work)
192 struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
194 struct ipvl_dev *ipvlan;
195 struct sk_buff *skb, *nskb;
196 struct sk_buff_head list;
198 unsigned int mac_hash;
203 __skb_queue_head_init(&list);
205 spin_lock_bh(&port->backlog.lock);
206 skb_queue_splice_tail_init(&port->backlog, &list);
207 spin_unlock_bh(&port->backlog.lock);
209 while ((skb = __skb_dequeue(&list)) != NULL) {
211 hlocal = ether_addr_equal(ethh->h_source, port->dev->dev_addr);
212 mac_hash = ipvlan_mac_hash(ethh->h_dest);
214 if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
215 pkt_type = PACKET_BROADCAST;
217 pkt_type = PACKET_MULTICAST;
221 list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
222 if (hlocal && (ipvlan->dev == skb->dev)) {
226 if (!test_bit(mac_hash, ipvlan->mac_filters))
230 len = skb->len + ETH_HLEN;
231 nskb = skb_clone(skb, GFP_ATOMIC);
235 nskb->pkt_type = pkt_type;
236 nskb->dev = ipvlan->dev;
238 ret = dev_forward_skb(ipvlan->dev, nskb);
240 ret = netif_rx(nskb);
242 ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
247 /* If the packet originated here, send it out. */
248 skb->dev = port->dev;
249 skb->pkt_type = pkt_type;
257 static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff *skb,
260 struct ipvl_dev *ipvlan = addr->master;
261 struct net_device *dev = ipvlan->dev;
263 rx_handler_result_t ret = RX_HANDLER_CONSUMED;
264 bool success = false;
266 len = skb->len + ETH_HLEN;
267 if (unlikely(!(dev->flags & IFF_UP))) {
272 skb = skb_share_check(skb, GFP_ATOMIC);
277 skb->pkt_type = PACKET_HOST;
280 if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
283 ret = RX_HANDLER_ANOTHER;
288 ipvlan_count_rx(ipvlan, len, success, false);
292 static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
293 void *lyr3h, int addr_type,
296 struct ipvl_addr *addr = NULL;
298 if (addr_type == IPVL_IPV6) {
299 struct ipv6hdr *ip6h;
300 struct in6_addr *i6addr;
302 ip6h = (struct ipv6hdr *)lyr3h;
303 i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
304 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
305 } else if (addr_type == IPVL_ICMPV6) {
307 struct in6_addr *i6addr;
309 /* Make sure that the NeighborSolicitation ICMPv6 packets
310 * are handled to avoid DAD issue.
312 ndmh = (struct nd_msg *)lyr3h;
313 if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
314 i6addr = &ndmh->target;
315 addr = ipvlan_ht_addr_lookup(port, i6addr, true);
317 } else if (addr_type == IPVL_IPV4) {
321 ip4h = (struct iphdr *)lyr3h;
322 i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
323 addr = ipvlan_ht_addr_lookup(port, i4addr, false);
324 } else if (addr_type == IPVL_ARP) {
326 unsigned char *arp_ptr;
329 arph = (struct arphdr *)lyr3h;
330 arp_ptr = (unsigned char *)(arph + 1);
332 arp_ptr += (2 * port->dev->addr_len) + 4;
334 arp_ptr += port->dev->addr_len;
336 memcpy(&dip, arp_ptr, 4);
337 addr = ipvlan_ht_addr_lookup(port, &dip, false);
343 static int ipvlan_process_v4_outbound(struct sk_buff *skb)
345 const struct iphdr *ip4h = ip_hdr(skb);
346 struct net_device *dev = skb->dev;
348 int err, ret = NET_XMIT_DROP;
349 struct flowi4 fl4 = {
350 .flowi4_oif = dev_get_iflink(dev),
351 .flowi4_tos = RT_TOS(ip4h->tos),
352 .flowi4_flags = FLOWI_FLAG_ANYSRC,
353 .daddr = ip4h->daddr,
354 .saddr = ip4h->saddr,
357 rt = ip_route_output_flow(dev_net(dev), &fl4, NULL);
361 if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
366 skb_dst_set(skb, &rt->dst);
367 err = ip_local_out(skb);
368 if (unlikely(net_xmit_eval(err)))
369 dev->stats.tx_errors++;
371 ret = NET_XMIT_SUCCESS;
374 dev->stats.tx_errors++;
380 static int ipvlan_process_v6_outbound(struct sk_buff *skb)
382 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
383 struct net_device *dev = skb->dev;
384 struct dst_entry *dst;
385 int err, ret = NET_XMIT_DROP;
386 struct flowi6 fl6 = {
387 .flowi6_iif = skb->dev->ifindex,
388 .daddr = ip6h->daddr,
389 .saddr = ip6h->saddr,
390 .flowi6_flags = FLOWI_FLAG_ANYSRC,
391 .flowlabel = ip6_flowinfo(ip6h),
392 .flowi6_mark = skb->mark,
393 .flowi6_proto = ip6h->nexthdr,
396 dst = ip6_route_output(dev_net(dev), NULL, &fl6);
403 skb_dst_set(skb, dst);
404 err = ip6_local_out(skb);
405 if (unlikely(net_xmit_eval(err)))
406 dev->stats.tx_errors++;
408 ret = NET_XMIT_SUCCESS;
411 dev->stats.tx_errors++;
417 static int ipvlan_process_outbound(struct sk_buff *skb,
418 const struct ipvl_dev *ipvlan)
420 struct ethhdr *ethh = eth_hdr(skb);
421 int ret = NET_XMIT_DROP;
423 /* In this mode we dont care about multicast and broadcast traffic */
424 if (is_multicast_ether_addr(ethh->h_dest)) {
425 pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
426 ntohs(skb->protocol));
431 /* The ipvlan is a pseudo-L2 device, so the packets that we receive
432 * will have L2; which need to discarded and processed further
433 * in the net-ns of the main-device.
435 if (skb_mac_header_was_set(skb)) {
436 skb_pull(skb, sizeof(*ethh));
437 skb->mac_header = (typeof(skb->mac_header))~0U;
438 skb_reset_network_header(skb);
441 if (skb->protocol == htons(ETH_P_IPV6))
442 ret = ipvlan_process_v6_outbound(skb);
443 else if (skb->protocol == htons(ETH_P_IP))
444 ret = ipvlan_process_v4_outbound(skb);
446 pr_warn_ratelimited("Dropped outbound packet type=%x\n",
447 ntohs(skb->protocol));
454 static void ipvlan_multicast_enqueue(struct ipvl_port *port,
457 if (skb->protocol == htons(ETH_P_PAUSE)) {
462 spin_lock(&port->backlog.lock);
463 if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
464 __skb_queue_tail(&port->backlog, skb);
465 spin_unlock(&port->backlog.lock);
466 schedule_work(&port->wq);
468 spin_unlock(&port->backlog.lock);
469 atomic_long_inc(&skb->dev->rx_dropped);
474 static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
476 const struct ipvl_dev *ipvlan = netdev_priv(dev);
478 struct ipvl_addr *addr;
481 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
485 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
487 return ipvlan_rcv_frame(addr, skb, true);
490 skb->dev = ipvlan->phy_dev;
491 return ipvlan_process_outbound(skb, ipvlan);
494 static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
496 const struct ipvl_dev *ipvlan = netdev_priv(dev);
497 struct ethhdr *eth = eth_hdr(skb);
498 struct ipvl_addr *addr;
502 if (ether_addr_equal(eth->h_dest, eth->h_source)) {
503 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
505 addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
507 return ipvlan_rcv_frame(addr, skb, true);
509 skb = skb_share_check(skb, GFP_ATOMIC);
511 return NET_XMIT_DROP;
513 /* Packet definitely does not belong to any of the
514 * virtual devices, but the dest is local. So forward
515 * the skb for the main-dev. At the RX side we just return
516 * RX_PASS for it to be processed further on the stack.
518 return dev_forward_skb(ipvlan->phy_dev, skb);
520 } else if (is_multicast_ether_addr(eth->h_dest)) {
521 ipvlan_multicast_enqueue(ipvlan->port, skb);
522 return NET_XMIT_SUCCESS;
525 skb->dev = ipvlan->phy_dev;
526 return dev_queue_xmit(skb);
529 int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
531 struct ipvl_dev *ipvlan = netdev_priv(dev);
532 struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
537 if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
542 return ipvlan_xmit_mode_l2(skb, dev);
544 return ipvlan_xmit_mode_l3(skb, dev);
547 /* Should not reach here */
548 WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
552 return NET_XMIT_DROP;
555 static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
557 struct ethhdr *eth = eth_hdr(skb);
558 struct ipvl_addr *addr;
562 if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
563 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
567 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
575 static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
576 struct ipvl_port *port)
580 struct ipvl_addr *addr;
581 struct sk_buff *skb = *pskb;
582 rx_handler_result_t ret = RX_HANDLER_PASS;
584 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
588 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
590 ret = ipvlan_rcv_frame(addr, skb, false);
596 static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
597 struct ipvl_port *port)
599 struct sk_buff *skb = *pskb;
600 struct ethhdr *eth = eth_hdr(skb);
601 rx_handler_result_t ret = RX_HANDLER_PASS;
605 if (is_multicast_ether_addr(eth->h_dest)) {
606 if (ipvlan_external_frame(skb, port)) {
607 struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
609 /* External frames are queued for device local
610 * distribution, but a copy is given to master
611 * straight away to avoid sending duplicates later
612 * when work-queue processes this frame. This is
613 * achieved by returning RX_HANDLER_PASS.
616 ipvlan_multicast_enqueue(port, nskb);
619 struct ipvl_addr *addr;
621 lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
625 addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
627 ret = ipvlan_rcv_frame(addr, skb, false);
633 rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
635 struct sk_buff *skb = *pskb;
636 struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
639 return RX_HANDLER_PASS;
641 switch (port->mode) {
643 return ipvlan_handle_mode_l2(pskb, port);
645 return ipvlan_handle_mode_l3(pskb, port);
648 /* Should not reach here */
649 WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",