ipvlan_core.c 18 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774
  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. #include "ipvlan.h"
  10. static u32 ipvlan_jhash_secret __read_mostly;
  11. void ipvlan_init_secret(void)
  12. {
  13. net_get_random_once(&ipvlan_jhash_secret, sizeof(ipvlan_jhash_secret));
  14. }
  15. static void ipvlan_count_rx(const struct ipvl_dev *ipvlan,
  16. unsigned int len, bool success, bool mcast)
  17. {
  18. if (!ipvlan)
  19. return;
  20. if (likely(success)) {
  21. struct ipvl_pcpu_stats *pcptr;
  22. pcptr = this_cpu_ptr(ipvlan->pcpu_stats);
  23. u64_stats_update_begin(&pcptr->syncp);
  24. pcptr->rx_pkts++;
  25. pcptr->rx_bytes += len;
  26. if (mcast)
  27. pcptr->rx_mcast++;
  28. u64_stats_update_end(&pcptr->syncp);
  29. } else {
  30. this_cpu_inc(ipvlan->pcpu_stats->rx_errs);
  31. }
  32. }
  33. static u8 ipvlan_get_v6_hash(const void *iaddr)
  34. {
  35. const struct in6_addr *ip6_addr = iaddr;
  36. return __ipv6_addr_jhash(ip6_addr, ipvlan_jhash_secret) &
  37. IPVLAN_HASH_MASK;
  38. }
  39. static u8 ipvlan_get_v4_hash(const void *iaddr)
  40. {
  41. const struct in_addr *ip4_addr = iaddr;
  42. return jhash_1word(ip4_addr->s_addr, ipvlan_jhash_secret) &
  43. IPVLAN_HASH_MASK;
  44. }
  45. static struct ipvl_addr *ipvlan_ht_addr_lookup(const struct ipvl_port *port,
  46. const void *iaddr, bool is_v6)
  47. {
  48. struct ipvl_addr *addr;
  49. u8 hash;
  50. hash = is_v6 ? ipvlan_get_v6_hash(iaddr) :
  51. ipvlan_get_v4_hash(iaddr);
  52. hlist_for_each_entry_rcu(addr, &port->hlhead[hash], hlnode) {
  53. if (is_v6 && addr->atype == IPVL_IPV6 &&
  54. ipv6_addr_equal(&addr->ip6addr, iaddr))
  55. return addr;
  56. else if (!is_v6 && addr->atype == IPVL_IPV4 &&
  57. addr->ip4addr.s_addr ==
  58. ((struct in_addr *)iaddr)->s_addr)
  59. return addr;
  60. }
  61. return NULL;
  62. }
  63. void ipvlan_ht_addr_add(struct ipvl_dev *ipvlan, struct ipvl_addr *addr)
  64. {
  65. struct ipvl_port *port = ipvlan->port;
  66. u8 hash;
  67. hash = (addr->atype == IPVL_IPV6) ?
  68. ipvlan_get_v6_hash(&addr->ip6addr) :
  69. ipvlan_get_v4_hash(&addr->ip4addr);
  70. if (hlist_unhashed(&addr->hlnode))
  71. hlist_add_head_rcu(&addr->hlnode, &port->hlhead[hash]);
  72. }
  73. void ipvlan_ht_addr_del(struct ipvl_addr *addr)
  74. {
  75. hlist_del_init_rcu(&addr->hlnode);
  76. }
  77. struct ipvl_addr *ipvlan_find_addr(const struct ipvl_dev *ipvlan,
  78. const void *iaddr, bool is_v6)
  79. {
  80. struct ipvl_addr *addr;
  81. list_for_each_entry(addr, &ipvlan->addrs, anode) {
  82. if ((is_v6 && addr->atype == IPVL_IPV6 &&
  83. ipv6_addr_equal(&addr->ip6addr, iaddr)) ||
  84. (!is_v6 && addr->atype == IPVL_IPV4 &&
  85. addr->ip4addr.s_addr == ((struct in_addr *)iaddr)->s_addr))
  86. return addr;
  87. }
  88. return NULL;
  89. }
  90. bool ipvlan_addr_busy(struct ipvl_port *port, void *iaddr, bool is_v6)
  91. {
  92. struct ipvl_dev *ipvlan;
  93. ASSERT_RTNL();
  94. list_for_each_entry(ipvlan, &port->ipvlans, pnode) {
  95. if (ipvlan_find_addr(ipvlan, iaddr, is_v6))
  96. return true;
  97. }
  98. return false;
  99. }
  100. static void *ipvlan_get_L3_hdr(struct sk_buff *skb, int *type)
  101. {
  102. void *lyr3h = NULL;
  103. switch (skb->protocol) {
  104. case htons(ETH_P_ARP): {
  105. struct arphdr *arph;
  106. if (unlikely(!pskb_may_pull(skb, sizeof(*arph))))
  107. return NULL;
  108. arph = arp_hdr(skb);
  109. *type = IPVL_ARP;
  110. lyr3h = arph;
  111. break;
  112. }
  113. case htons(ETH_P_IP): {
  114. u32 pktlen;
  115. struct iphdr *ip4h;
  116. if (unlikely(!pskb_may_pull(skb, sizeof(*ip4h))))
  117. return NULL;
  118. ip4h = ip_hdr(skb);
  119. pktlen = ntohs(ip4h->tot_len);
  120. if (ip4h->ihl < 5 || ip4h->version != 4)
  121. return NULL;
  122. if (skb->len < pktlen || pktlen < (ip4h->ihl * 4))
  123. return NULL;
  124. *type = IPVL_IPV4;
  125. lyr3h = ip4h;
  126. break;
  127. }
  128. case htons(ETH_P_IPV6): {
  129. struct ipv6hdr *ip6h;
  130. if (unlikely(!pskb_may_pull(skb, sizeof(*ip6h))))
  131. return NULL;
  132. ip6h = ipv6_hdr(skb);
  133. if (ip6h->version != 6)
  134. return NULL;
  135. *type = IPVL_IPV6;
  136. lyr3h = ip6h;
  137. /* Only Neighbour Solicitation pkts need different treatment */
  138. if (ipv6_addr_any(&ip6h->saddr) &&
  139. ip6h->nexthdr == NEXTHDR_ICMP) {
  140. *type = IPVL_ICMPV6;
  141. lyr3h = ip6h + 1;
  142. }
  143. break;
  144. }
  145. default:
  146. return NULL;
  147. }
  148. return lyr3h;
  149. }
  150. unsigned int ipvlan_mac_hash(const unsigned char *addr)
  151. {
  152. u32 hash = jhash_1word(__get_unaligned_cpu32(addr+2),
  153. ipvlan_jhash_secret);
  154. return hash & IPVLAN_MAC_FILTER_MASK;
  155. }
  156. void ipvlan_process_multicast(struct work_struct *work)
  157. {
  158. struct ipvl_port *port = container_of(work, struct ipvl_port, wq);
  159. struct ethhdr *ethh;
  160. struct ipvl_dev *ipvlan;
  161. struct sk_buff *skb, *nskb;
  162. struct sk_buff_head list;
  163. unsigned int len;
  164. unsigned int mac_hash;
  165. int ret;
  166. u8 pkt_type;
  167. bool hlocal, dlocal;
  168. __skb_queue_head_init(&list);
  169. spin_lock_bh(&port->backlog.lock);
  170. skb_queue_splice_tail_init(&port->backlog, &list);
  171. spin_unlock_bh(&port->backlog.lock);
  172. while ((skb = __skb_dequeue(&list)) != NULL) {
  173. ethh = eth_hdr(skb);
  174. hlocal = ether_addr_equal(ethh->h_source, port->dev->dev_addr);
  175. mac_hash = ipvlan_mac_hash(ethh->h_dest);
  176. if (ether_addr_equal(ethh->h_dest, port->dev->broadcast))
  177. pkt_type = PACKET_BROADCAST;
  178. else
  179. pkt_type = PACKET_MULTICAST;
  180. dlocal = false;
  181. rcu_read_lock();
  182. list_for_each_entry_rcu(ipvlan, &port->ipvlans, pnode) {
  183. if (hlocal && (ipvlan->dev == skb->dev)) {
  184. dlocal = true;
  185. continue;
  186. }
  187. if (!test_bit(mac_hash, ipvlan->mac_filters))
  188. continue;
  189. ret = NET_RX_DROP;
  190. len = skb->len + ETH_HLEN;
  191. nskb = skb_clone(skb, GFP_ATOMIC);
  192. if (!nskb)
  193. goto acct;
  194. nskb->pkt_type = pkt_type;
  195. nskb->dev = ipvlan->dev;
  196. if (hlocal)
  197. ret = dev_forward_skb(ipvlan->dev, nskb);
  198. else
  199. ret = netif_rx(nskb);
  200. acct:
  201. ipvlan_count_rx(ipvlan, len, ret == NET_RX_SUCCESS, true);
  202. }
  203. rcu_read_unlock();
  204. if (dlocal) {
  205. /* If the packet originated here, send it out. */
  206. skb->dev = port->dev;
  207. skb->pkt_type = pkt_type;
  208. dev_queue_xmit(skb);
  209. } else {
  210. kfree_skb(skb);
  211. }
  212. }
  213. }
  214. static void ipvlan_skb_crossing_ns(struct sk_buff *skb, struct net_device *dev)
  215. {
  216. bool xnet = true;
  217. if (dev)
  218. xnet = !net_eq(dev_net(skb->dev), dev_net(dev));
  219. skb_scrub_packet(skb, xnet);
  220. if (dev)
  221. skb->dev = dev;
  222. }
  223. static int ipvlan_rcv_frame(struct ipvl_addr *addr, struct sk_buff **pskb,
  224. bool local)
  225. {
  226. struct ipvl_dev *ipvlan = addr->master;
  227. struct net_device *dev = ipvlan->dev;
  228. unsigned int len;
  229. rx_handler_result_t ret = RX_HANDLER_CONSUMED;
  230. bool success = false;
  231. struct sk_buff *skb = *pskb;
  232. len = skb->len + ETH_HLEN;
  233. /* Only packets exchanged between two local slaves need to have
  234. * device-up check as well as skb-share check.
  235. */
  236. if (local) {
  237. if (unlikely(!(dev->flags & IFF_UP))) {
  238. kfree_skb(skb);
  239. goto out;
  240. }
  241. skb = skb_share_check(skb, GFP_ATOMIC);
  242. if (!skb)
  243. goto out;
  244. *pskb = skb;
  245. }
  246. ipvlan_skb_crossing_ns(skb, dev);
  247. if (local) {
  248. skb->pkt_type = PACKET_HOST;
  249. if (dev_forward_skb(ipvlan->dev, skb) == NET_RX_SUCCESS)
  250. success = true;
  251. } else {
  252. if (!ether_addr_equal_64bits(eth_hdr(skb)->h_dest,
  253. ipvlan->phy_dev->dev_addr))
  254. skb->pkt_type = PACKET_OTHERHOST;
  255. ret = RX_HANDLER_ANOTHER;
  256. success = true;
  257. }
  258. out:
  259. ipvlan_count_rx(ipvlan, len, success, false);
  260. return ret;
  261. }
  262. static struct ipvl_addr *ipvlan_addr_lookup(struct ipvl_port *port,
  263. void *lyr3h, int addr_type,
  264. bool use_dest)
  265. {
  266. struct ipvl_addr *addr = NULL;
  267. if (addr_type == IPVL_IPV6) {
  268. struct ipv6hdr *ip6h;
  269. struct in6_addr *i6addr;
  270. ip6h = (struct ipv6hdr *)lyr3h;
  271. i6addr = use_dest ? &ip6h->daddr : &ip6h->saddr;
  272. addr = ipvlan_ht_addr_lookup(port, i6addr, true);
  273. } else if (addr_type == IPVL_ICMPV6) {
  274. struct nd_msg *ndmh;
  275. struct in6_addr *i6addr;
  276. /* Make sure that the NeighborSolicitation ICMPv6 packets
  277. * are handled to avoid DAD issue.
  278. */
  279. ndmh = (struct nd_msg *)lyr3h;
  280. if (ndmh->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) {
  281. i6addr = &ndmh->target;
  282. addr = ipvlan_ht_addr_lookup(port, i6addr, true);
  283. }
  284. } else if (addr_type == IPVL_IPV4) {
  285. struct iphdr *ip4h;
  286. __be32 *i4addr;
  287. ip4h = (struct iphdr *)lyr3h;
  288. i4addr = use_dest ? &ip4h->daddr : &ip4h->saddr;
  289. addr = ipvlan_ht_addr_lookup(port, i4addr, false);
  290. } else if (addr_type == IPVL_ARP) {
  291. struct arphdr *arph;
  292. unsigned char *arp_ptr;
  293. __be32 dip;
  294. arph = (struct arphdr *)lyr3h;
  295. arp_ptr = (unsigned char *)(arph + 1);
  296. if (use_dest)
  297. arp_ptr += (2 * port->dev->addr_len) + 4;
  298. else
  299. arp_ptr += port->dev->addr_len;
  300. memcpy(&dip, arp_ptr, 4);
  301. addr = ipvlan_ht_addr_lookup(port, &dip, false);
  302. }
  303. return addr;
  304. }
  305. static int ipvlan_process_v4_outbound(struct sk_buff *skb)
  306. {
  307. const struct iphdr *ip4h = ip_hdr(skb);
  308. struct net_device *dev = skb->dev;
  309. struct net *net = dev_net(dev);
  310. struct rtable *rt;
  311. int err, ret = NET_XMIT_DROP;
  312. struct flowi4 fl4 = {
  313. .flowi4_oif = dev->ifindex,
  314. .flowi4_tos = RT_TOS(ip4h->tos),
  315. .flowi4_flags = FLOWI_FLAG_ANYSRC,
  316. .flowi4_mark = skb->mark,
  317. .daddr = ip4h->daddr,
  318. .saddr = ip4h->saddr,
  319. };
  320. rt = ip_route_output_flow(net, &fl4, NULL);
  321. if (IS_ERR(rt))
  322. goto err;
  323. if (rt->rt_type != RTN_UNICAST && rt->rt_type != RTN_LOCAL) {
  324. ip_rt_put(rt);
  325. goto err;
  326. }
  327. skb_dst_set(skb, &rt->dst);
  328. err = ip_local_out(net, skb->sk, skb);
  329. if (unlikely(net_xmit_eval(err)))
  330. dev->stats.tx_errors++;
  331. else
  332. ret = NET_XMIT_SUCCESS;
  333. goto out;
  334. err:
  335. dev->stats.tx_errors++;
  336. kfree_skb(skb);
  337. out:
  338. return ret;
  339. }
  340. static int ipvlan_process_v6_outbound(struct sk_buff *skb)
  341. {
  342. const struct ipv6hdr *ip6h = ipv6_hdr(skb);
  343. struct net_device *dev = skb->dev;
  344. struct net *net = dev_net(dev);
  345. struct dst_entry *dst;
  346. int err, ret = NET_XMIT_DROP;
  347. struct flowi6 fl6 = {
  348. .flowi6_oif = dev->ifindex,
  349. .daddr = ip6h->daddr,
  350. .saddr = ip6h->saddr,
  351. .flowi6_flags = FLOWI_FLAG_ANYSRC,
  352. .flowlabel = ip6_flowinfo(ip6h),
  353. .flowi6_mark = skb->mark,
  354. .flowi6_proto = ip6h->nexthdr,
  355. };
  356. dst = ip6_route_output(net, NULL, &fl6);
  357. if (dst->error) {
  358. ret = dst->error;
  359. dst_release(dst);
  360. goto err;
  361. }
  362. skb_dst_set(skb, dst);
  363. err = ip6_local_out(net, skb->sk, skb);
  364. if (unlikely(net_xmit_eval(err)))
  365. dev->stats.tx_errors++;
  366. else
  367. ret = NET_XMIT_SUCCESS;
  368. goto out;
  369. err:
  370. dev->stats.tx_errors++;
  371. kfree_skb(skb);
  372. out:
  373. return ret;
  374. }
  375. static int ipvlan_process_outbound(struct sk_buff *skb)
  376. {
  377. struct ethhdr *ethh = eth_hdr(skb);
  378. int ret = NET_XMIT_DROP;
  379. /* In this mode we dont care about multicast and broadcast traffic */
  380. if (is_multicast_ether_addr(ethh->h_dest)) {
  381. pr_warn_ratelimited("Dropped {multi|broad}cast of type= [%x]\n",
  382. ntohs(skb->protocol));
  383. kfree_skb(skb);
  384. goto out;
  385. }
  386. /* The ipvlan is a pseudo-L2 device, so the packets that we receive
  387. * will have L2; which need to discarded and processed further
  388. * in the net-ns of the main-device.
  389. */
  390. if (skb_mac_header_was_set(skb)) {
  391. skb_pull(skb, sizeof(*ethh));
  392. skb->mac_header = (typeof(skb->mac_header))~0U;
  393. skb_reset_network_header(skb);
  394. }
  395. if (skb->protocol == htons(ETH_P_IPV6))
  396. ret = ipvlan_process_v6_outbound(skb);
  397. else if (skb->protocol == htons(ETH_P_IP))
  398. ret = ipvlan_process_v4_outbound(skb);
  399. else {
  400. pr_warn_ratelimited("Dropped outbound packet type=%x\n",
  401. ntohs(skb->protocol));
  402. kfree_skb(skb);
  403. }
  404. out:
  405. return ret;
  406. }
  407. static void ipvlan_multicast_enqueue(struct ipvl_port *port,
  408. struct sk_buff *skb)
  409. {
  410. if (skb->protocol == htons(ETH_P_PAUSE)) {
  411. kfree_skb(skb);
  412. return;
  413. }
  414. spin_lock(&port->backlog.lock);
  415. if (skb_queue_len(&port->backlog) < IPVLAN_QBACKLOG_LIMIT) {
  416. __skb_queue_tail(&port->backlog, skb);
  417. spin_unlock(&port->backlog.lock);
  418. schedule_work(&port->wq);
  419. } else {
  420. spin_unlock(&port->backlog.lock);
  421. atomic_long_inc(&skb->dev->rx_dropped);
  422. kfree_skb(skb);
  423. }
  424. }
  425. static int ipvlan_xmit_mode_l3(struct sk_buff *skb, struct net_device *dev)
  426. {
  427. const struct ipvl_dev *ipvlan = netdev_priv(dev);
  428. void *lyr3h;
  429. struct ipvl_addr *addr;
  430. int addr_type;
  431. lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
  432. if (!lyr3h)
  433. goto out;
  434. addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
  435. if (addr)
  436. return ipvlan_rcv_frame(addr, &skb, true);
  437. out:
  438. ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
  439. return ipvlan_process_outbound(skb);
  440. }
  441. static int ipvlan_xmit_mode_l2(struct sk_buff *skb, struct net_device *dev)
  442. {
  443. const struct ipvl_dev *ipvlan = netdev_priv(dev);
  444. struct ethhdr *eth = eth_hdr(skb);
  445. struct ipvl_addr *addr;
  446. void *lyr3h;
  447. int addr_type;
  448. if (ether_addr_equal(eth->h_dest, eth->h_source)) {
  449. lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
  450. if (lyr3h) {
  451. addr = ipvlan_addr_lookup(ipvlan->port, lyr3h, addr_type, true);
  452. if (addr)
  453. return ipvlan_rcv_frame(addr, &skb, true);
  454. }
  455. skb = skb_share_check(skb, GFP_ATOMIC);
  456. if (!skb)
  457. return NET_XMIT_DROP;
  458. /* Packet definitely does not belong to any of the
  459. * virtual devices, but the dest is local. So forward
  460. * the skb for the main-dev. At the RX side we just return
  461. * RX_PASS for it to be processed further on the stack.
  462. */
  463. return dev_forward_skb(ipvlan->phy_dev, skb);
  464. } else if (is_multicast_ether_addr(eth->h_dest)) {
  465. ipvlan_skb_crossing_ns(skb, NULL);
  466. ipvlan_multicast_enqueue(ipvlan->port, skb);
  467. return NET_XMIT_SUCCESS;
  468. }
  469. ipvlan_skb_crossing_ns(skb, ipvlan->phy_dev);
  470. return dev_queue_xmit(skb);
  471. }
  472. int ipvlan_queue_xmit(struct sk_buff *skb, struct net_device *dev)
  473. {
  474. struct ipvl_dev *ipvlan = netdev_priv(dev);
  475. struct ipvl_port *port = ipvlan_port_get_rcu_bh(ipvlan->phy_dev);
  476. if (!port)
  477. goto out;
  478. if (unlikely(!pskb_may_pull(skb, sizeof(struct ethhdr))))
  479. goto out;
  480. switch(port->mode) {
  481. case IPVLAN_MODE_L2:
  482. return ipvlan_xmit_mode_l2(skb, dev);
  483. case IPVLAN_MODE_L3:
  484. case IPVLAN_MODE_L3S:
  485. return ipvlan_xmit_mode_l3(skb, dev);
  486. }
  487. /* Should not reach here */
  488. WARN_ONCE(true, "ipvlan_queue_xmit() called for mode = [%hx]\n",
  489. port->mode);
  490. out:
  491. kfree_skb(skb);
  492. return NET_XMIT_DROP;
  493. }
  494. static bool ipvlan_external_frame(struct sk_buff *skb, struct ipvl_port *port)
  495. {
  496. struct ethhdr *eth = eth_hdr(skb);
  497. struct ipvl_addr *addr;
  498. void *lyr3h;
  499. int addr_type;
  500. if (ether_addr_equal(eth->h_source, skb->dev->dev_addr)) {
  501. lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
  502. if (!lyr3h)
  503. return true;
  504. addr = ipvlan_addr_lookup(port, lyr3h, addr_type, false);
  505. if (addr)
  506. return false;
  507. }
  508. return true;
  509. }
  510. static rx_handler_result_t ipvlan_handle_mode_l3(struct sk_buff **pskb,
  511. struct ipvl_port *port)
  512. {
  513. void *lyr3h;
  514. int addr_type;
  515. struct ipvl_addr *addr;
  516. struct sk_buff *skb = *pskb;
  517. rx_handler_result_t ret = RX_HANDLER_PASS;
  518. lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
  519. if (!lyr3h)
  520. goto out;
  521. addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
  522. if (addr)
  523. ret = ipvlan_rcv_frame(addr, pskb, false);
  524. out:
  525. return ret;
  526. }
  527. static rx_handler_result_t ipvlan_handle_mode_l2(struct sk_buff **pskb,
  528. struct ipvl_port *port)
  529. {
  530. struct sk_buff *skb = *pskb;
  531. struct ethhdr *eth = eth_hdr(skb);
  532. rx_handler_result_t ret = RX_HANDLER_PASS;
  533. void *lyr3h;
  534. int addr_type;
  535. if (is_multicast_ether_addr(eth->h_dest)) {
  536. if (ipvlan_external_frame(skb, port)) {
  537. struct sk_buff *nskb = skb_clone(skb, GFP_ATOMIC);
  538. /* External frames are queued for device local
  539. * distribution, but a copy is given to master
  540. * straight away to avoid sending duplicates later
  541. * when work-queue processes this frame. This is
  542. * achieved by returning RX_HANDLER_PASS.
  543. */
  544. if (nskb) {
  545. ipvlan_skb_crossing_ns(nskb, NULL);
  546. ipvlan_multicast_enqueue(port, nskb);
  547. }
  548. }
  549. } else {
  550. struct ipvl_addr *addr;
  551. lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
  552. if (!lyr3h)
  553. return ret;
  554. addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
  555. if (addr)
  556. ret = ipvlan_rcv_frame(addr, pskb, false);
  557. }
  558. return ret;
  559. }
  560. rx_handler_result_t ipvlan_handle_frame(struct sk_buff **pskb)
  561. {
  562. struct sk_buff *skb = *pskb;
  563. struct ipvl_port *port = ipvlan_port_get_rcu(skb->dev);
  564. if (!port)
  565. return RX_HANDLER_PASS;
  566. switch (port->mode) {
  567. case IPVLAN_MODE_L2:
  568. return ipvlan_handle_mode_l2(pskb, port);
  569. case IPVLAN_MODE_L3:
  570. return ipvlan_handle_mode_l3(pskb, port);
  571. case IPVLAN_MODE_L3S:
  572. return RX_HANDLER_PASS;
  573. }
  574. /* Should not reach here */
  575. WARN_ONCE(true, "ipvlan_handle_frame() called for mode = [%hx]\n",
  576. port->mode);
  577. kfree_skb(skb);
  578. return RX_HANDLER_CONSUMED;
  579. }
  580. static struct ipvl_addr *ipvlan_skb_to_addr(struct sk_buff *skb,
  581. struct net_device *dev)
  582. {
  583. struct ipvl_addr *addr = NULL;
  584. struct ipvl_port *port;
  585. void *lyr3h;
  586. int addr_type;
  587. if (!dev || !netif_is_ipvlan_port(dev))
  588. goto out;
  589. port = ipvlan_port_get_rcu(dev);
  590. if (!port || port->mode != IPVLAN_MODE_L3S)
  591. goto out;
  592. lyr3h = ipvlan_get_L3_hdr(skb, &addr_type);
  593. if (!lyr3h)
  594. goto out;
  595. addr = ipvlan_addr_lookup(port, lyr3h, addr_type, true);
  596. out:
  597. return addr;
  598. }
  599. struct sk_buff *ipvlan_l3_rcv(struct net_device *dev, struct sk_buff *skb,
  600. u16 proto)
  601. {
  602. struct ipvl_addr *addr;
  603. struct net_device *sdev;
  604. addr = ipvlan_skb_to_addr(skb, dev);
  605. if (!addr)
  606. goto out;
  607. sdev = addr->master->dev;
  608. switch (proto) {
  609. case AF_INET:
  610. {
  611. int err;
  612. struct iphdr *ip4h = ip_hdr(skb);
  613. err = ip_route_input_noref(skb, ip4h->daddr, ip4h->saddr,
  614. ip4h->tos, sdev);
  615. if (unlikely(err))
  616. goto out;
  617. break;
  618. }
  619. case AF_INET6:
  620. {
  621. struct dst_entry *dst;
  622. struct ipv6hdr *ip6h = ipv6_hdr(skb);
  623. int flags = RT6_LOOKUP_F_HAS_SADDR;
  624. struct flowi6 fl6 = {
  625. .flowi6_iif = sdev->ifindex,
  626. .daddr = ip6h->daddr,
  627. .saddr = ip6h->saddr,
  628. .flowlabel = ip6_flowinfo(ip6h),
  629. .flowi6_mark = skb->mark,
  630. .flowi6_proto = ip6h->nexthdr,
  631. };
  632. skb_dst_drop(skb);
  633. dst = ip6_route_input_lookup(dev_net(sdev), sdev, &fl6, flags);
  634. skb_dst_set(skb, dst);
  635. break;
  636. }
  637. default:
  638. break;
  639. }
  640. out:
  641. return skb;
  642. }
  643. unsigned int ipvlan_nf_input(void *priv, struct sk_buff *skb,
  644. const struct nf_hook_state *state)
  645. {
  646. struct ipvl_addr *addr;
  647. unsigned int len;
  648. addr = ipvlan_skb_to_addr(skb, skb->dev);
  649. if (!addr)
  650. goto out;
  651. skb->dev = addr->master->dev;
  652. len = skb->len + ETH_HLEN;
  653. ipvlan_count_rx(addr->master, len, true, false);
  654. out:
  655. return NF_ACCEPT;
  656. }