netfilter.c 6.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265
  1. /*
  2. * IPv6 specific functions of netfilter core
  3. *
  4. * Rusty Russell (C) 2000 -- This code is GPL.
  5. * Patrick McHardy (C) 2006-2012
  6. */
  7. #include <linux/kernel.h>
  8. #include <linux/init.h>
  9. #include <linux/ipv6.h>
  10. #include <linux/netfilter.h>
  11. #include <linux/netfilter_ipv6.h>
  12. #include <linux/export.h>
  13. #include <net/addrconf.h>
  14. #include <net/dst.h>
  15. #include <net/ipv6.h>
  16. #include <net/ip6_route.h>
  17. #include <net/xfrm.h>
  18. #include <net/netfilter/nf_queue.h>
  19. #include <net/netfilter/nf_conntrack_bridge.h>
  20. #include <net/netfilter/ipv6/nf_defrag_ipv6.h>
  21. #include "../bridge/br_private.h"
  22. int ip6_route_me_harder(struct net *net, struct sock *sk_partial, struct sk_buff *skb)
  23. {
  24. const struct ipv6hdr *iph = ipv6_hdr(skb);
  25. struct sock *sk = sk_to_full_sk(sk_partial);
  26. unsigned int hh_len;
  27. struct dst_entry *dst;
  28. int strict = (ipv6_addr_type(&iph->daddr) &
  29. (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
  30. struct flowi6 fl6 = {
  31. .flowi6_oif = sk && sk->sk_bound_dev_if ? sk->sk_bound_dev_if :
  32. strict ? skb_dst(skb)->dev->ifindex : 0,
  33. .flowi6_mark = skb->mark,
  34. .flowi6_uid = sock_net_uid(net, sk),
  35. .daddr = iph->daddr,
  36. .saddr = iph->saddr,
  37. };
  38. int err;
  39. dst = ip6_route_output(net, sk, &fl6);
  40. err = dst->error;
  41. if (err) {
  42. IP6_INC_STATS(net, ip6_dst_idev(dst), IPSTATS_MIB_OUTNOROUTES);
  43. net_dbg_ratelimited("ip6_route_me_harder: No more route\n");
  44. dst_release(dst);
  45. return err;
  46. }
  47. /* Drop old route. */
  48. skb_dst_drop(skb);
  49. skb_dst_set(skb, dst);
  50. #ifdef CONFIG_XFRM
  51. if (!(IP6CB(skb)->flags & IP6SKB_XFRM_TRANSFORMED) &&
  52. xfrm_decode_session(skb, flowi6_to_flowi(&fl6), AF_INET6) == 0) {
  53. skb_dst_set(skb, NULL);
  54. dst = xfrm_lookup(net, dst, flowi6_to_flowi(&fl6), sk, 0);
  55. if (IS_ERR(dst))
  56. return PTR_ERR(dst);
  57. skb_dst_set(skb, dst);
  58. }
  59. #endif
  60. /* Change in oif may mean change in hh_len. */
  61. hh_len = skb_dst(skb)->dev->hard_header_len;
  62. if (skb_headroom(skb) < hh_len &&
  63. pskb_expand_head(skb, HH_DATA_ALIGN(hh_len - skb_headroom(skb)),
  64. 0, GFP_ATOMIC))
  65. return -ENOMEM;
  66. return 0;
  67. }
  68. EXPORT_SYMBOL(ip6_route_me_harder);
  69. static int nf_ip6_reroute(struct sk_buff *skb,
  70. const struct nf_queue_entry *entry)
  71. {
  72. struct ip6_rt_info *rt_info = nf_queue_entry_reroute(entry);
  73. if (entry->state.hook == NF_INET_LOCAL_OUT) {
  74. const struct ipv6hdr *iph = ipv6_hdr(skb);
  75. if (!ipv6_addr_equal(&iph->daddr, &rt_info->daddr) ||
  76. !ipv6_addr_equal(&iph->saddr, &rt_info->saddr) ||
  77. skb->mark != rt_info->mark)
  78. return ip6_route_me_harder(entry->state.net, entry->state.sk, skb);
  79. }
  80. return 0;
  81. }
  82. int __nf_ip6_route(struct net *net, struct dst_entry **dst,
  83. struct flowi *fl, bool strict)
  84. {
  85. static const struct ipv6_pinfo fake_pinfo;
  86. static const struct inet_sock fake_sk = {
  87. /* makes ip6_route_output set RT6_LOOKUP_F_IFACE: */
  88. .sk.sk_bound_dev_if = 1,
  89. .pinet6 = (struct ipv6_pinfo *) &fake_pinfo,
  90. };
  91. const void *sk = strict ? &fake_sk : NULL;
  92. struct dst_entry *result;
  93. int err;
  94. result = ip6_route_output(net, sk, &fl->u.ip6);
  95. err = result->error;
  96. if (err)
  97. dst_release(result);
  98. else
  99. *dst = result;
  100. return err;
  101. }
  102. EXPORT_SYMBOL_GPL(__nf_ip6_route);
  103. int br_ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
  104. struct nf_bridge_frag_data *data,
  105. int (*output)(struct net *, struct sock *sk,
  106. const struct nf_bridge_frag_data *data,
  107. struct sk_buff *))
  108. {
  109. int frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
  110. ktime_t tstamp = skb->tstamp;
  111. struct ip6_frag_state state;
  112. u8 *prevhdr, nexthdr = 0;
  113. unsigned int mtu, hlen;
  114. int hroom, err = 0;
  115. __be32 frag_id;
  116. err = ip6_find_1stfragopt(skb, &prevhdr);
  117. if (err < 0)
  118. goto blackhole;
  119. hlen = err;
  120. nexthdr = *prevhdr;
  121. mtu = skb->dev->mtu;
  122. if (frag_max_size > mtu ||
  123. frag_max_size < IPV6_MIN_MTU)
  124. goto blackhole;
  125. mtu = frag_max_size;
  126. if (mtu < hlen + sizeof(struct frag_hdr) + 8)
  127. goto blackhole;
  128. mtu -= hlen + sizeof(struct frag_hdr);
  129. frag_id = ipv6_select_ident(net, &ipv6_hdr(skb)->daddr,
  130. &ipv6_hdr(skb)->saddr);
  131. if (skb->ip_summed == CHECKSUM_PARTIAL &&
  132. (err = skb_checksum_help(skb)))
  133. goto blackhole;
  134. hroom = LL_RESERVED_SPACE(skb->dev);
  135. if (skb_has_frag_list(skb)) {
  136. unsigned int first_len = skb_pagelen(skb);
  137. struct ip6_fraglist_iter iter;
  138. struct sk_buff *frag2;
  139. if (first_len - hlen > mtu ||
  140. skb_headroom(skb) < (hroom + sizeof(struct frag_hdr)))
  141. goto blackhole;
  142. if (skb_cloned(skb))
  143. goto slow_path;
  144. skb_walk_frags(skb, frag2) {
  145. if (frag2->len > mtu ||
  146. skb_headroom(frag2) < (hlen + hroom + sizeof(struct frag_hdr)))
  147. goto blackhole;
  148. /* Partially cloned skb? */
  149. if (skb_shared(frag2))
  150. goto slow_path;
  151. }
  152. err = ip6_fraglist_init(skb, hlen, prevhdr, nexthdr, frag_id,
  153. &iter);
  154. if (err < 0)
  155. goto blackhole;
  156. for (;;) {
  157. /* Prepare header of the next frame,
  158. * before previous one went down.
  159. */
  160. if (iter.frag)
  161. ip6_fraglist_prepare(skb, &iter);
  162. skb->tstamp = tstamp;
  163. err = output(net, sk, data, skb);
  164. if (err || !iter.frag)
  165. break;
  166. skb = ip6_fraglist_next(&iter);
  167. }
  168. kfree(iter.tmp_hdr);
  169. if (!err)
  170. return 0;
  171. kfree_skb_list(iter.frag);
  172. return err;
  173. }
  174. slow_path:
  175. /* This is a linearized skbuff, the original geometry is lost for us.
  176. * This may also be a clone skbuff, we could preserve the geometry for
  177. * the copies but probably not worth the effort.
  178. */
  179. ip6_frag_init(skb, hlen, mtu, skb->dev->needed_tailroom,
  180. LL_RESERVED_SPACE(skb->dev), prevhdr, nexthdr, frag_id,
  181. &state);
  182. while (state.left > 0) {
  183. struct sk_buff *skb2;
  184. skb2 = ip6_frag_next(skb, &state);
  185. if (IS_ERR(skb2)) {
  186. err = PTR_ERR(skb2);
  187. goto blackhole;
  188. }
  189. skb2->tstamp = tstamp;
  190. err = output(net, sk, data, skb2);
  191. if (err)
  192. goto blackhole;
  193. }
  194. consume_skb(skb);
  195. return err;
  196. blackhole:
  197. kfree_skb(skb);
  198. return 0;
  199. }
  200. EXPORT_SYMBOL_GPL(br_ip6_fragment);
  201. static const struct nf_ipv6_ops ipv6ops = {
  202. #if IS_MODULE(CONFIG_IPV6)
  203. .chk_addr = ipv6_chk_addr,
  204. .route_me_harder = ip6_route_me_harder,
  205. .dev_get_saddr = ipv6_dev_get_saddr,
  206. .route = __nf_ip6_route,
  207. #if IS_ENABLED(CONFIG_SYN_COOKIES)
  208. .cookie_init_sequence = __cookie_v6_init_sequence,
  209. .cookie_v6_check = __cookie_v6_check,
  210. #endif
  211. #endif
  212. .route_input = ip6_route_input,
  213. .fragment = ip6_fragment,
  214. .reroute = nf_ip6_reroute,
  215. #if IS_MODULE(CONFIG_IPV6)
  216. .br_fragment = br_ip6_fragment,
  217. #endif
  218. };
  219. int __init ipv6_netfilter_init(void)
  220. {
  221. RCU_INIT_POINTER(nf_ipv6_ops, &ipv6ops);
  222. return 0;
  223. }
  224. /* This can be called from inet6_init() on errors, so it cannot
  225. * be marked __exit. -DaveM
  226. */
  227. void ipv6_netfilter_fini(void)
  228. {
  229. RCU_INIT_POINTER(nf_ipv6_ops, NULL);
  230. }