br_arp_nd_proxy.c 11 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485
  1. // SPDX-License-Identifier: GPL-2.0-or-later
  2. /*
  3. * Handle bridge arp/nd proxy/suppress
  4. *
  5. * Copyright (C) 2017 Cumulus Networks
  6. * Copyright (c) 2017 Roopa Prabhu <roopa@cumulusnetworks.com>
  7. *
  8. * Authors:
  9. * Roopa Prabhu <roopa@cumulusnetworks.com>
  10. */
  11. #include <linux/kernel.h>
  12. #include <linux/netdevice.h>
  13. #include <linux/etherdevice.h>
  14. #include <linux/neighbour.h>
  15. #include <net/arp.h>
  16. #include <linux/if_vlan.h>
  17. #include <linux/inetdevice.h>
  18. #include <net/addrconf.h>
  19. #include <net/ipv6_stubs.h>
  20. #if IS_ENABLED(CONFIG_IPV6)
  21. #include <net/ip6_checksum.h>
  22. #endif
  23. #include "br_private.h"
  24. void br_recalculate_neigh_suppress_enabled(struct net_bridge *br)
  25. {
  26. struct net_bridge_port *p;
  27. bool neigh_suppress = false;
  28. list_for_each_entry(p, &br->port_list, list) {
  29. if (p->flags & BR_NEIGH_SUPPRESS) {
  30. neigh_suppress = true;
  31. break;
  32. }
  33. }
  34. br_opt_toggle(br, BROPT_NEIGH_SUPPRESS_ENABLED, neigh_suppress);
  35. }
  36. #if IS_ENABLED(CONFIG_INET)
  37. static void br_arp_send(struct net_bridge *br, struct net_bridge_port *p,
  38. struct net_device *dev, __be32 dest_ip, __be32 src_ip,
  39. const unsigned char *dest_hw,
  40. const unsigned char *src_hw,
  41. const unsigned char *target_hw,
  42. __be16 vlan_proto, u16 vlan_tci)
  43. {
  44. struct net_bridge_vlan_group *vg;
  45. struct sk_buff *skb;
  46. u16 pvid;
  47. netdev_dbg(dev, "arp send dev %s dst %pI4 dst_hw %pM src %pI4 src_hw %pM\n",
  48. dev->name, &dest_ip, dest_hw, &src_ip, src_hw);
  49. if (!vlan_tci) {
  50. arp_send(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
  51. dest_hw, src_hw, target_hw);
  52. return;
  53. }
  54. skb = arp_create(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
  55. dest_hw, src_hw, target_hw);
  56. if (!skb)
  57. return;
  58. if (p)
  59. vg = nbp_vlan_group_rcu(p);
  60. else
  61. vg = br_vlan_group_rcu(br);
  62. pvid = br_get_pvid(vg);
  63. if (pvid == (vlan_tci & VLAN_VID_MASK))
  64. vlan_tci = 0;
  65. if (vlan_tci)
  66. __vlan_hwaccel_put_tag(skb, vlan_proto, vlan_tci);
  67. if (p) {
  68. arp_xmit(skb);
  69. } else {
  70. skb_reset_mac_header(skb);
  71. __skb_pull(skb, skb_network_offset(skb));
  72. skb->ip_summed = CHECKSUM_UNNECESSARY;
  73. skb->pkt_type = PACKET_HOST;
  74. netif_rx_ni(skb);
  75. }
  76. }
  77. static int br_chk_addr_ip(struct net_device *dev,
  78. struct netdev_nested_priv *priv)
  79. {
  80. __be32 ip = *(__be32 *)priv->data;
  81. struct in_device *in_dev;
  82. __be32 addr = 0;
  83. in_dev = __in_dev_get_rcu(dev);
  84. if (in_dev)
  85. addr = inet_confirm_addr(dev_net(dev), in_dev, 0, ip,
  86. RT_SCOPE_HOST);
  87. if (addr == ip)
  88. return 1;
  89. return 0;
  90. }
  91. static bool br_is_local_ip(struct net_device *dev, __be32 ip)
  92. {
  93. struct netdev_nested_priv priv = {
  94. .data = (void *)&ip,
  95. };
  96. if (br_chk_addr_ip(dev, &priv))
  97. return true;
  98. /* check if ip is configured on upper dev */
  99. if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip, &priv))
  100. return true;
  101. return false;
  102. }
  103. void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br,
  104. u16 vid, struct net_bridge_port *p)
  105. {
  106. struct net_device *dev = br->dev;
  107. struct net_device *vlandev = dev;
  108. struct neighbour *n;
  109. struct arphdr *parp;
  110. u8 *arpptr, *sha;
  111. __be32 sip, tip;
  112. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
  113. if ((dev->flags & IFF_NOARP) ||
  114. !pskb_may_pull(skb, arp_hdr_len(dev)))
  115. return;
  116. parp = arp_hdr(skb);
  117. if (parp->ar_pro != htons(ETH_P_IP) ||
  118. parp->ar_hln != dev->addr_len ||
  119. parp->ar_pln != 4)
  120. return;
  121. arpptr = (u8 *)parp + sizeof(struct arphdr);
  122. sha = arpptr;
  123. arpptr += dev->addr_len; /* sha */
  124. memcpy(&sip, arpptr, sizeof(sip));
  125. arpptr += sizeof(sip);
  126. arpptr += dev->addr_len; /* tha */
  127. memcpy(&tip, arpptr, sizeof(tip));
  128. if (ipv4_is_loopback(tip) ||
  129. ipv4_is_multicast(tip))
  130. return;
  131. if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
  132. if (p && (p->flags & BR_NEIGH_SUPPRESS))
  133. return;
  134. if (parp->ar_op != htons(ARPOP_RREQUEST) &&
  135. parp->ar_op != htons(ARPOP_RREPLY) &&
  136. (ipv4_is_zeronet(sip) || sip == tip)) {
  137. /* prevent flooding to neigh suppress ports */
  138. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  139. return;
  140. }
  141. }
  142. if (parp->ar_op != htons(ARPOP_REQUEST))
  143. return;
  144. if (vid != 0) {
  145. vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
  146. vid);
  147. if (!vlandev)
  148. return;
  149. }
  150. if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
  151. br_is_local_ip(vlandev, tip)) {
  152. /* its our local ip, so don't proxy reply
  153. * and don't forward to neigh suppress ports
  154. */
  155. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  156. return;
  157. }
  158. n = neigh_lookup(&arp_tbl, &tip, vlandev);
  159. if (n) {
  160. struct net_bridge_fdb_entry *f;
  161. if (!(n->nud_state & NUD_VALID)) {
  162. neigh_release(n);
  163. return;
  164. }
  165. f = br_fdb_find_rcu(br, n->ha, vid);
  166. if (f) {
  167. bool replied = false;
  168. if ((p && (p->flags & BR_PROXYARP)) ||
  169. (f->dst && (f->dst->flags & (BR_PROXYARP_WIFI |
  170. BR_NEIGH_SUPPRESS)))) {
  171. if (!vid)
  172. br_arp_send(br, p, skb->dev, sip, tip,
  173. sha, n->ha, sha, 0, 0);
  174. else
  175. br_arp_send(br, p, skb->dev, sip, tip,
  176. sha, n->ha, sha,
  177. skb->vlan_proto,
  178. skb_vlan_tag_get(skb));
  179. replied = true;
  180. }
  181. /* If we have replied or as long as we know the
  182. * mac, indicate to arp replied
  183. */
  184. if (replied ||
  185. br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
  186. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  187. }
  188. neigh_release(n);
  189. }
  190. }
  191. #endif
  192. #if IS_ENABLED(CONFIG_IPV6)
  193. struct nd_msg *br_is_nd_neigh_msg(struct sk_buff *skb, struct nd_msg *msg)
  194. {
  195. struct nd_msg *m;
  196. m = skb_header_pointer(skb, skb_network_offset(skb) +
  197. sizeof(struct ipv6hdr), sizeof(*msg), msg);
  198. if (!m)
  199. return NULL;
  200. if (m->icmph.icmp6_code != 0 ||
  201. (m->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION &&
  202. m->icmph.icmp6_type != NDISC_NEIGHBOUR_ADVERTISEMENT))
  203. return NULL;
  204. return m;
  205. }
  206. static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
  207. struct sk_buff *request, struct neighbour *n,
  208. __be16 vlan_proto, u16 vlan_tci, struct nd_msg *ns)
  209. {
  210. struct net_device *dev = request->dev;
  211. struct net_bridge_vlan_group *vg;
  212. struct sk_buff *reply;
  213. struct nd_msg *na;
  214. struct ipv6hdr *pip6;
  215. int na_olen = 8; /* opt hdr + ETH_ALEN for target */
  216. int ns_olen;
  217. int i, len;
  218. u8 *daddr;
  219. u16 pvid;
  220. if (!dev)
  221. return;
  222. len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
  223. sizeof(*na) + na_olen + dev->needed_tailroom;
  224. reply = alloc_skb(len, GFP_ATOMIC);
  225. if (!reply)
  226. return;
  227. reply->protocol = htons(ETH_P_IPV6);
  228. reply->dev = dev;
  229. skb_reserve(reply, LL_RESERVED_SPACE(dev));
  230. skb_push(reply, sizeof(struct ethhdr));
  231. skb_set_mac_header(reply, 0);
  232. daddr = eth_hdr(request)->h_source;
  233. /* Do we need option processing ? */
  234. ns_olen = request->len - (skb_network_offset(request) +
  235. sizeof(struct ipv6hdr)) - sizeof(*ns);
  236. for (i = 0; i < ns_olen - 1; i += (ns->opt[i + 1] << 3)) {
  237. if (!ns->opt[i + 1]) {
  238. kfree_skb(reply);
  239. return;
  240. }
  241. if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
  242. daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
  243. break;
  244. }
  245. }
  246. /* Ethernet header */
  247. ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
  248. ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
  249. eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
  250. reply->protocol = htons(ETH_P_IPV6);
  251. skb_pull(reply, sizeof(struct ethhdr));
  252. skb_set_network_header(reply, 0);
  253. skb_put(reply, sizeof(struct ipv6hdr));
  254. /* IPv6 header */
  255. pip6 = ipv6_hdr(reply);
  256. memset(pip6, 0, sizeof(struct ipv6hdr));
  257. pip6->version = 6;
  258. pip6->priority = ipv6_hdr(request)->priority;
  259. pip6->nexthdr = IPPROTO_ICMPV6;
  260. pip6->hop_limit = 255;
  261. pip6->daddr = ipv6_hdr(request)->saddr;
  262. pip6->saddr = *(struct in6_addr *)n->primary_key;
  263. skb_pull(reply, sizeof(struct ipv6hdr));
  264. skb_set_transport_header(reply, 0);
  265. na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
  266. /* Neighbor Advertisement */
  267. memset(na, 0, sizeof(*na) + na_olen);
  268. na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
  269. na->icmph.icmp6_router = (n->flags & NTF_ROUTER) ? 1 : 0;
  270. na->icmph.icmp6_override = 1;
  271. na->icmph.icmp6_solicited = 1;
  272. na->target = ns->target;
  273. ether_addr_copy(&na->opt[2], n->ha);
  274. na->opt[0] = ND_OPT_TARGET_LL_ADDR;
  275. na->opt[1] = na_olen >> 3;
  276. na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
  277. &pip6->daddr,
  278. sizeof(*na) + na_olen,
  279. IPPROTO_ICMPV6,
  280. csum_partial(na, sizeof(*na) + na_olen, 0));
  281. pip6->payload_len = htons(sizeof(*na) + na_olen);
  282. skb_push(reply, sizeof(struct ipv6hdr));
  283. skb_push(reply, sizeof(struct ethhdr));
  284. reply->ip_summed = CHECKSUM_UNNECESSARY;
  285. if (p)
  286. vg = nbp_vlan_group_rcu(p);
  287. else
  288. vg = br_vlan_group_rcu(br);
  289. pvid = br_get_pvid(vg);
  290. if (pvid == (vlan_tci & VLAN_VID_MASK))
  291. vlan_tci = 0;
  292. if (vlan_tci)
  293. __vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
  294. netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
  295. dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha);
  296. if (p) {
  297. dev_queue_xmit(reply);
  298. } else {
  299. skb_reset_mac_header(reply);
  300. __skb_pull(reply, skb_network_offset(reply));
  301. reply->ip_summed = CHECKSUM_UNNECESSARY;
  302. reply->pkt_type = PACKET_HOST;
  303. netif_rx_ni(reply);
  304. }
  305. }
  306. static int br_chk_addr_ip6(struct net_device *dev,
  307. struct netdev_nested_priv *priv)
  308. {
  309. struct in6_addr *addr = (struct in6_addr *)priv->data;
  310. if (ipv6_chk_addr(dev_net(dev), addr, dev, 0))
  311. return 1;
  312. return 0;
  313. }
  314. static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr)
  315. {
  316. struct netdev_nested_priv priv = {
  317. .data = (void *)addr,
  318. };
  319. if (br_chk_addr_ip6(dev, &priv))
  320. return true;
  321. /* check if ip is configured on upper dev */
  322. if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, &priv))
  323. return true;
  324. return false;
  325. }
  326. void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
  327. u16 vid, struct net_bridge_port *p, struct nd_msg *msg)
  328. {
  329. struct net_device *dev = br->dev;
  330. struct net_device *vlandev = NULL;
  331. struct in6_addr *saddr, *daddr;
  332. struct ipv6hdr *iphdr;
  333. struct neighbour *n;
  334. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
  335. if (p && (p->flags & BR_NEIGH_SUPPRESS))
  336. return;
  337. if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT &&
  338. !msg->icmph.icmp6_solicited) {
  339. /* prevent flooding to neigh suppress ports */
  340. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  341. return;
  342. }
  343. if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
  344. return;
  345. iphdr = ipv6_hdr(skb);
  346. saddr = &iphdr->saddr;
  347. daddr = &iphdr->daddr;
  348. if (ipv6_addr_any(saddr) || !ipv6_addr_cmp(saddr, daddr)) {
  349. /* prevent flooding to neigh suppress ports */
  350. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  351. return;
  352. }
  353. if (vid != 0) {
  354. /* build neigh table lookup on the vlan device */
  355. vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
  356. vid);
  357. if (!vlandev)
  358. return;
  359. } else {
  360. vlandev = dev;
  361. }
  362. if (br_is_local_ip6(vlandev, &msg->target)) {
  363. /* its our own ip, so don't proxy reply
  364. * and don't forward to arp suppress ports
  365. */
  366. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  367. return;
  368. }
  369. n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, vlandev);
  370. if (n) {
  371. struct net_bridge_fdb_entry *f;
  372. if (!(n->nud_state & NUD_VALID)) {
  373. neigh_release(n);
  374. return;
  375. }
  376. f = br_fdb_find_rcu(br, n->ha, vid);
  377. if (f) {
  378. bool replied = false;
  379. if (f->dst && (f->dst->flags & BR_NEIGH_SUPPRESS)) {
  380. if (vid != 0)
  381. br_nd_send(br, p, skb, n,
  382. skb->vlan_proto,
  383. skb_vlan_tag_get(skb), msg);
  384. else
  385. br_nd_send(br, p, skb, n, 0, 0, msg);
  386. replied = true;
  387. }
  388. /* If we have replied or as long as we know the
  389. * mac, indicate to NEIGH_SUPPRESS ports that we
  390. * have replied
  391. */
  392. if (replied ||
  393. br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
  394. BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
  395. }
  396. neigh_release(n);
  397. }
  398. }
  399. #endif