Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * INET An implementation of the TCP/IP protocol suite for the LINUX |
| 3 | * operating system. INET is implemented using the BSD Socket |
| 4 | * interface as the means of communication with the user level. |
| 5 | * |
| 6 | * The IP forwarding functionality. |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 7 | * |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 8 | * Authors: see ip.c |
| 9 | * |
| 10 | * Fixes: |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 11 | * Many : Split from ip.c , see ip_input.c for |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 12 | * history. |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 13 | * Dave Gregorich : NULL ip_rt_put fix for multicast |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 14 | * routing. |
| 15 | * Jos Vos : Add call_out_firewall before sending, |
| 16 | * use output device for accounting. |
| 17 | * Jos Vos : Call forward firewall after routing |
| 18 | * (always use output device). |
| 19 | * Mike McLagan : Routing by source |
| 20 | */ |
| 21 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 22 | #include <linux/types.h> |
| 23 | #include <linux/mm.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 24 | #include <linux/skbuff.h> |
| 25 | #include <linux/ip.h> |
| 26 | #include <linux/icmp.h> |
| 27 | #include <linux/netdevice.h> |
Tejun Heo | 5a0e3ad | 2010-03-24 17:04:11 +0900 | [diff] [blame] | 28 | #include <linux/slab.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | |
Patrick McHardy | 861d048 | 2007-10-15 01:48:39 -0700 | [diff] [blame] | 42 | static int ip_forward_finish(struct sk_buff *skb) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 43 | { |
Daniel Baluta | 5e73ea1 | 2012-04-15 01:34:41 +0000 | [diff] [blame] | 44 | struct ip_options *opt = &(IPCB(skb)->opt); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 45 | |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 46 | IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 47 | |
| 48 | if (unlikely(opt->optlen)) |
| 49 | ip_forward_options(skb); |
| 50 | |
| 51 | return dst_output(skb); |
| 52 | } |
| 53 | |
| 54 | int ip_forward(struct sk_buff *skb) |
| 55 | { |
| 56 | struct iphdr *iph; /* Our header */ |
| 57 | struct rtable *rt; /* Route we use */ |
Daniel Baluta | 5e73ea1 | 2012-04-15 01:34:41 +0000 | [diff] [blame] | 58 | struct ip_options *opt = &(IPCB(skb)->opt); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 59 | |
Ben Hutchings | 4497b07 | 2008-06-19 16:22:28 -0700 | [diff] [blame] | 60 | if (skb_warn_if_lro(skb)) |
| 61 | goto drop; |
| 62 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 63 | if (!xfrm4_policy_check(NULL, XFRM_POLICY_FWD, skb)) |
| 64 | goto drop; |
| 65 | |
| 66 | if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb)) |
| 67 | return NET_RX_SUCCESS; |
| 68 | |
| 69 | if (skb->pkt_type != PACKET_HOST) |
| 70 | goto drop; |
| 71 | |
Herbert Xu | 35fc92a | 2007-03-26 23:22:20 -0700 | [diff] [blame] | 72 | skb_forward_csum(skb); |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 73 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 74 | /* |
| 75 | * According to the RFC, we must first decrease the TTL field. If |
| 76 | * that reaches zero, we must reply an ICMP control message telling |
| 77 | * that the packet's lifetime expired. |
| 78 | */ |
Arnaldo Carvalho de Melo | eddc9ec | 2007-04-20 22:47:35 -0700 | [diff] [blame] | 79 | if (ip_hdr(skb)->ttl <= 1) |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 80 | goto too_many_hops; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 81 | |
| 82 | if (!xfrm4_route_forward(skb)) |
| 83 | goto drop; |
| 84 | |
Eric Dumazet | 511c3f9 | 2009-06-02 05:14:27 +0000 | [diff] [blame] | 85 | rt = skb_rtable(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 86 | |
Li Wei | ac8a481 | 2011-11-22 23:33:10 +0000 | [diff] [blame] | 87 | if (opt->is_strictroute && opt->nexthop != rt->rt_gateway) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 88 | goto sr_failed; |
| 89 | |
Changli Gao | d8d1f30 | 2010-06-10 23:31:35 -0700 | [diff] [blame] | 90 | if (unlikely(skb->len > dst_mtu(&rt->dst) && !skb_is_gso(skb) && |
YOSHIFUJI Hideaki | 9c681b4 | 2007-07-19 10:43:47 +0900 | [diff] [blame] | 91 | (ip_hdr(skb)->frag_off & htons(IP_DF))) && !skb->local_df) { |
Changli Gao | d8d1f30 | 2010-06-10 23:31:35 -0700 | [diff] [blame] | 92 | IP_INC_STATS(dev_net(rt->dst.dev), IPSTATS_MIB_FRAGFAILS); |
John Heffner | 9af3912 | 2007-03-25 23:32:29 -0700 | [diff] [blame] | 93 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED, |
Changli Gao | d8d1f30 | 2010-06-10 23:31:35 -0700 | [diff] [blame] | 94 | htonl(dst_mtu(&rt->dst))); |
John Heffner | 9af3912 | 2007-03-25 23:32:29 -0700 | [diff] [blame] | 95 | goto drop; |
| 96 | } |
| 97 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 98 | /* We are about to mangle packet. Copy it! */ |
Changli Gao | d8d1f30 | 2010-06-10 23:31:35 -0700 | [diff] [blame] | 99 | if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 100 | goto drop; |
Arnaldo Carvalho de Melo | eddc9ec | 2007-04-20 22:47:35 -0700 | [diff] [blame] | 101 | iph = ip_hdr(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 102 | |
| 103 | /* Decrease ttl after skb cow done */ |
| 104 | ip_decrease_ttl(iph); |
| 105 | |
| 106 | /* |
| 107 | * We now generate an ICMP HOST REDIRECT giving the route |
| 108 | * we calculated. |
| 109 | */ |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 110 | if (rt->rt_flags&RTCF_DOREDIRECT && !opt->srr && !skb_sec_path(skb)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 111 | ip_rt_send_redirect(skb); |
| 112 | |
| 113 | skb->priority = rt_tos2priority(iph->tos); |
| 114 | |
Jan Engelhardt | 9bbc768 | 2010-03-23 04:07:29 +0100 | [diff] [blame] | 115 | return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, |
Changli Gao | d8d1f30 | 2010-06-10 23:31:35 -0700 | [diff] [blame] | 116 | rt->dst.dev, ip_forward_finish); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 117 | |
| 118 | sr_failed: |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 119 | /* |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 120 | * Strict routing permits no gatewaying |
| 121 | */ |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 122 | icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0); |
| 123 | goto drop; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 124 | |
| 125 | too_many_hops: |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 126 | /* Tell the sender its packet died... */ |
Eric Dumazet | adf3090 | 2009-06-02 05:19:30 +0000 | [diff] [blame] | 127 | IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_INHDRERRORS); |
YOSHIFUJI Hideaki | e905a9e | 2007-02-09 23:24:47 +0900 | [diff] [blame] | 128 | icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 129 | drop: |
| 130 | kfree_skb(skb); |
| 131 | return NET_RX_DROP; |
| 132 | } |