mpls_iptunnel.c 7.8 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305
  1. // SPDX-License-Identifier: GPL-2.0-or-later
  2. /*
  3. * mpls tunnels An implementation mpls tunnels using the light weight tunnel
  4. * infrastructure
  5. *
  6. * Authors: Roopa Prabhu, <roopa@cumulusnetworks.com>
  7. */
  8. #include <linux/types.h>
  9. #include <linux/skbuff.h>
  10. #include <linux/net.h>
  11. #include <linux/module.h>
  12. #include <linux/mpls.h>
  13. #include <linux/vmalloc.h>
  14. #include <net/ip.h>
  15. #include <net/dst.h>
  16. #include <net/lwtunnel.h>
  17. #include <net/netevent.h>
  18. #include <net/netns/generic.h>
  19. #include <net/ip6_fib.h>
  20. #include <net/route.h>
  21. #include <net/mpls_iptunnel.h>
  22. #include <linux/mpls_iptunnel.h>
  23. #include "internal.h"
  24. static const struct nla_policy mpls_iptunnel_policy[MPLS_IPTUNNEL_MAX + 1] = {
  25. [MPLS_IPTUNNEL_DST] = { .len = sizeof(u32) },
  26. [MPLS_IPTUNNEL_TTL] = { .type = NLA_U8 },
  27. };
  28. static unsigned int mpls_encap_size(struct mpls_iptunnel_encap *en)
  29. {
  30. /* The size of the layer 2.5 labels to be added for this route */
  31. return en->labels * sizeof(struct mpls_shim_hdr);
  32. }
  33. static int mpls_xmit(struct sk_buff *skb)
  34. {
  35. struct mpls_iptunnel_encap *tun_encap_info;
  36. struct mpls_shim_hdr *hdr;
  37. struct net_device *out_dev;
  38. unsigned int hh_len;
  39. unsigned int new_header_size;
  40. unsigned int mtu;
  41. struct dst_entry *dst = skb_dst(skb);
  42. struct rtable *rt = NULL;
  43. struct rt6_info *rt6 = NULL;
  44. struct mpls_dev *out_mdev;
  45. struct net *net;
  46. int err = 0;
  47. bool bos;
  48. int i;
  49. unsigned int ttl;
  50. /* Find the output device */
  51. out_dev = dst->dev;
  52. net = dev_net(out_dev);
  53. skb_orphan(skb);
  54. if (!mpls_output_possible(out_dev) ||
  55. !dst->lwtstate || skb_warn_if_lro(skb))
  56. goto drop;
  57. skb_forward_csum(skb);
  58. tun_encap_info = mpls_lwtunnel_encap(dst->lwtstate);
  59. /* Obtain the ttl using the following set of rules.
  60. *
  61. * LWT ttl propagation setting:
  62. * - disabled => use default TTL value from LWT
  63. * - enabled => use TTL value from IPv4/IPv6 header
  64. * - default =>
  65. * Global ttl propagation setting:
  66. * - disabled => use default TTL value from global setting
  67. * - enabled => use TTL value from IPv4/IPv6 header
  68. */
  69. if (dst->ops->family == AF_INET) {
  70. if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
  71. ttl = tun_encap_info->default_ttl;
  72. else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
  73. !net->mpls.ip_ttl_propagate)
  74. ttl = net->mpls.default_ttl;
  75. else
  76. ttl = ip_hdr(skb)->ttl;
  77. rt = (struct rtable *)dst;
  78. } else if (dst->ops->family == AF_INET6) {
  79. if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DISABLED)
  80. ttl = tun_encap_info->default_ttl;
  81. else if (tun_encap_info->ttl_propagate == MPLS_TTL_PROP_DEFAULT &&
  82. !net->mpls.ip_ttl_propagate)
  83. ttl = net->mpls.default_ttl;
  84. else
  85. ttl = ipv6_hdr(skb)->hop_limit;
  86. rt6 = (struct rt6_info *)dst;
  87. } else {
  88. goto drop;
  89. }
  90. /* Verify the destination can hold the packet */
  91. new_header_size = mpls_encap_size(tun_encap_info);
  92. mtu = mpls_dev_mtu(out_dev);
  93. if (mpls_pkt_too_big(skb, mtu - new_header_size))
  94. goto drop;
  95. hh_len = LL_RESERVED_SPACE(out_dev);
  96. if (!out_dev->header_ops)
  97. hh_len = 0;
  98. /* Ensure there is enough space for the headers in the skb */
  99. if (skb_cow(skb, hh_len + new_header_size))
  100. goto drop;
  101. skb_set_inner_protocol(skb, skb->protocol);
  102. skb_reset_inner_network_header(skb);
  103. skb_push(skb, new_header_size);
  104. skb_reset_network_header(skb);
  105. skb->dev = out_dev;
  106. skb->protocol = htons(ETH_P_MPLS_UC);
  107. /* Push the new labels */
  108. hdr = mpls_hdr(skb);
  109. bos = true;
  110. for (i = tun_encap_info->labels - 1; i >= 0; i--) {
  111. hdr[i] = mpls_entry_encode(tun_encap_info->label[i],
  112. ttl, 0, bos);
  113. bos = false;
  114. }
  115. mpls_stats_inc_outucastpkts(out_dev, skb);
  116. if (rt) {
  117. if (rt->rt_gw_family == AF_INET6)
  118. err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt->rt_gw6,
  119. skb);
  120. else
  121. err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt->rt_gw4,
  122. skb);
  123. } else if (rt6) {
  124. if (ipv6_addr_v4mapped(&rt6->rt6i_gateway)) {
  125. /* 6PE (RFC 4798) */
  126. err = neigh_xmit(NEIGH_ARP_TABLE, out_dev, &rt6->rt6i_gateway.s6_addr32[3],
  127. skb);
  128. } else
  129. err = neigh_xmit(NEIGH_ND_TABLE, out_dev, &rt6->rt6i_gateway,
  130. skb);
  131. }
  132. if (err)
  133. net_dbg_ratelimited("%s: packet transmission failed: %d\n",
  134. __func__, err);
  135. return LWTUNNEL_XMIT_DONE;
  136. drop:
  137. out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
  138. if (out_mdev)
  139. MPLS_INC_STATS(out_mdev, tx_errors);
  140. kfree_skb(skb);
  141. return -EINVAL;
  142. }
  143. static int mpls_build_state(struct net *net, struct nlattr *nla,
  144. unsigned int family, const void *cfg,
  145. struct lwtunnel_state **ts,
  146. struct netlink_ext_ack *extack)
  147. {
  148. struct mpls_iptunnel_encap *tun_encap_info;
  149. struct nlattr *tb[MPLS_IPTUNNEL_MAX + 1];
  150. struct lwtunnel_state *newts;
  151. u8 n_labels;
  152. int ret;
  153. ret = nla_parse_nested_deprecated(tb, MPLS_IPTUNNEL_MAX, nla,
  154. mpls_iptunnel_policy, extack);
  155. if (ret < 0)
  156. return ret;
  157. if (!tb[MPLS_IPTUNNEL_DST]) {
  158. NL_SET_ERR_MSG(extack, "MPLS_IPTUNNEL_DST attribute is missing");
  159. return -EINVAL;
  160. }
  161. /* determine number of labels */
  162. if (nla_get_labels(tb[MPLS_IPTUNNEL_DST], MAX_NEW_LABELS,
  163. &n_labels, NULL, extack))
  164. return -EINVAL;
  165. newts = lwtunnel_state_alloc(struct_size(tun_encap_info, label,
  166. n_labels));
  167. if (!newts)
  168. return -ENOMEM;
  169. tun_encap_info = mpls_lwtunnel_encap(newts);
  170. ret = nla_get_labels(tb[MPLS_IPTUNNEL_DST], n_labels,
  171. &tun_encap_info->labels, tun_encap_info->label,
  172. extack);
  173. if (ret)
  174. goto errout;
  175. tun_encap_info->ttl_propagate = MPLS_TTL_PROP_DEFAULT;
  176. if (tb[MPLS_IPTUNNEL_TTL]) {
  177. tun_encap_info->default_ttl = nla_get_u8(tb[MPLS_IPTUNNEL_TTL]);
  178. /* TTL 0 implies propagate from IP header */
  179. tun_encap_info->ttl_propagate = tun_encap_info->default_ttl ?
  180. MPLS_TTL_PROP_DISABLED :
  181. MPLS_TTL_PROP_ENABLED;
  182. }
  183. newts->type = LWTUNNEL_ENCAP_MPLS;
  184. newts->flags |= LWTUNNEL_STATE_XMIT_REDIRECT;
  185. newts->headroom = mpls_encap_size(tun_encap_info);
  186. *ts = newts;
  187. return 0;
  188. errout:
  189. kfree(newts);
  190. *ts = NULL;
  191. return ret;
  192. }
  193. static int mpls_fill_encap_info(struct sk_buff *skb,
  194. struct lwtunnel_state *lwtstate)
  195. {
  196. struct mpls_iptunnel_encap *tun_encap_info;
  197. tun_encap_info = mpls_lwtunnel_encap(lwtstate);
  198. if (nla_put_labels(skb, MPLS_IPTUNNEL_DST, tun_encap_info->labels,
  199. tun_encap_info->label))
  200. goto nla_put_failure;
  201. if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT &&
  202. nla_put_u8(skb, MPLS_IPTUNNEL_TTL, tun_encap_info->default_ttl))
  203. goto nla_put_failure;
  204. return 0;
  205. nla_put_failure:
  206. return -EMSGSIZE;
  207. }
  208. static int mpls_encap_nlsize(struct lwtunnel_state *lwtstate)
  209. {
  210. struct mpls_iptunnel_encap *tun_encap_info;
  211. int nlsize;
  212. tun_encap_info = mpls_lwtunnel_encap(lwtstate);
  213. nlsize = nla_total_size(tun_encap_info->labels * 4);
  214. if (tun_encap_info->ttl_propagate != MPLS_TTL_PROP_DEFAULT)
  215. nlsize += nla_total_size(1);
  216. return nlsize;
  217. }
  218. static int mpls_encap_cmp(struct lwtunnel_state *a, struct lwtunnel_state *b)
  219. {
  220. struct mpls_iptunnel_encap *a_hdr = mpls_lwtunnel_encap(a);
  221. struct mpls_iptunnel_encap *b_hdr = mpls_lwtunnel_encap(b);
  222. int l;
  223. if (a_hdr->labels != b_hdr->labels ||
  224. a_hdr->ttl_propagate != b_hdr->ttl_propagate ||
  225. a_hdr->default_ttl != b_hdr->default_ttl)
  226. return 1;
  227. for (l = 0; l < a_hdr->labels; l++)
  228. if (a_hdr->label[l] != b_hdr->label[l])
  229. return 1;
  230. return 0;
  231. }
  232. static const struct lwtunnel_encap_ops mpls_iptun_ops = {
  233. .build_state = mpls_build_state,
  234. .xmit = mpls_xmit,
  235. .fill_encap = mpls_fill_encap_info,
  236. .get_encap_size = mpls_encap_nlsize,
  237. .cmp_encap = mpls_encap_cmp,
  238. .owner = THIS_MODULE,
  239. };
  240. static int __init mpls_iptunnel_init(void)
  241. {
  242. return lwtunnel_encap_add_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
  243. }
  244. module_init(mpls_iptunnel_init);
  245. static void __exit mpls_iptunnel_exit(void)
  246. {
  247. lwtunnel_encap_del_ops(&mpls_iptun_ops, LWTUNNEL_ENCAP_MPLS);
  248. }
  249. module_exit(mpls_iptunnel_exit);
  250. MODULE_ALIAS_RTNL_LWT(MPLS);
  251. MODULE_SOFTDEP("post: mpls_gso");
  252. MODULE_DESCRIPTION("MultiProtocol Label Switching IP Tunnels");
  253. MODULE_LICENSE("GPL v2");