ip_forward.c 4.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181
  1. // SPDX-License-Identifier: GPL-2.0
  2. /*
  3. * INET An implementation of the TCP/IP protocol suite for the LINUX
  4. * operating system. INET is implemented using the BSD Socket
  5. * interface as the means of communication with the user level.
  6. *
  7. * The IP forwarding functionality.
  8. *
  9. * Authors: see ip.c
  10. *
  11. * Fixes:
  12. * Many : Split from ip.c , see ip_input.c for
  13. * history.
  14. * Dave Gregorich : NULL ip_rt_put fix for multicast
  15. * routing.
  16. * Jos Vos : Add call_out_firewall before sending,
  17. * use output device for accounting.
  18. * Jos Vos : Call forward firewall after routing
  19. * (always use output device).
  20. * Mike McLagan : Routing by source
  21. */
  22. #include <linux/types.h>
  23. #include <linux/mm.h>
  24. #include <linux/skbuff.h>
  25. #include <linux/ip.h>
  26. #include <linux/icmp.h>
  27. #include <linux/netdevice.h>
  28. #include <linux/slab.h>
  29. #include <net/sock.h>
  30. #include <net/ip.h>
  31. #include <net/tcp.h>
  32. #include <net/udp.h>
  33. #include <net/icmp.h>
  34. #include <linux/tcp.h>
  35. #include <linux/udp.h>
  36. #include <linux/netfilter_ipv4.h>
  37. #include <net/checksum.h>
  38. #include <linux/route.h>
  39. #include <net/route.h>
  40. #include <net/xfrm.h>
  41. static bool ip_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
  42. {
  43. if (skb->len <= mtu)
  44. return false;
  45. if (unlikely((ip_hdr(skb)->frag_off & htons(IP_DF)) == 0))
  46. return false;
  47. /* original fragment exceeds mtu and DF is set */
  48. if (unlikely(IPCB(skb)->frag_max_size > mtu))
  49. return true;
  50. if (skb->ignore_df)
  51. return false;
  52. if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
  53. return false;
  54. return true;
  55. }
  56. static int ip_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
  57. {
  58. struct ip_options *opt = &(IPCB(skb)->opt);
  59. #ifdef CONFIG_NET_SWITCHDEV
  60. if (skb->offload_l3_fwd_mark) {
  61. consume_skb(skb);
  62. return 0;
  63. }
  64. #endif
  65. if (unlikely(opt->optlen))
  66. ip_forward_options(skb);
  67. skb_clear_tstamp(skb);
  68. return dst_output(net, sk, skb);
  69. }
  70. int ip_forward(struct sk_buff *skb)
  71. {
  72. u32 mtu;
  73. struct iphdr *iph; /* Our header */
  74. struct rtable *rt; /* Route we use */
  75. struct ip_options *opt = &(IPCB(skb)->opt);
  76. struct net *net;
  77. SKB_DR(reason);
  78. /* that should never happen */
  79. if (skb->pkt_type != PACKET_HOST)
  80. goto drop;
  81. if (unlikely(skb->sk))
  82. goto drop;
  83. if (skb_warn_if_lro(skb))
  84. goto drop;
  85. if (!xfrm4_policy_check(NULL, XFRM_POLICY_FWD, skb)) {
  86. SKB_DR_SET(reason, XFRM_POLICY);
  87. goto drop;
  88. }
  89. if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb))
  90. return NET_RX_SUCCESS;
  91. skb_forward_csum(skb);
  92. net = dev_net(skb->dev);
  93. /*
  94. * According to the RFC, we must first decrease the TTL field. If
  95. * that reaches zero, we must reply an ICMP control message telling
  96. * that the packet's lifetime expired.
  97. */
  98. if (ip_hdr(skb)->ttl <= 1)
  99. goto too_many_hops;
  100. if (!xfrm4_route_forward(skb)) {
  101. SKB_DR_SET(reason, XFRM_POLICY);
  102. goto drop;
  103. }
  104. rt = skb_rtable(skb);
  105. if (opt->is_strictroute && rt->rt_uses_gateway)
  106. goto sr_failed;
  107. __IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
  108. IPCB(skb)->flags |= IPSKB_FORWARDED;
  109. mtu = ip_dst_mtu_maybe_forward(&rt->dst, true);
  110. if (ip_exceeds_mtu(skb, mtu)) {
  111. IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
  112. icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
  113. htonl(mtu));
  114. SKB_DR_SET(reason, PKT_TOO_BIG);
  115. goto drop;
  116. }
  117. /* We are about to mangle packet. Copy it! */
  118. if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
  119. goto drop;
  120. iph = ip_hdr(skb);
  121. /* Decrease ttl after skb cow done */
  122. ip_decrease_ttl(iph);
  123. /*
  124. * We now generate an ICMP HOST REDIRECT giving the route
  125. * we calculated.
  126. */
  127. if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
  128. !skb_sec_path(skb))
  129. ip_rt_send_redirect(skb);
  130. if (READ_ONCE(net->ipv4.sysctl_ip_fwd_update_priority))
  131. skb->priority = rt_tos2priority(iph->tos);
  132. return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
  133. net, NULL, skb, skb->dev, rt->dst.dev,
  134. ip_forward_finish);
  135. sr_failed:
  136. /*
  137. * Strict routing permits no gatewaying
  138. */
  139. icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0);
  140. goto drop;
  141. too_many_hops:
  142. /* Tell the sender its packet died... */
  143. __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
  144. icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0);
  145. SKB_DR_SET(reason, IP_INHDR);
  146. drop:
  147. kfree_skb_reason(skb, reason);
  148. return NET_RX_DROP;
  149. }