blob: da0d7ce85844a9ed0732883ac363ab2de62f3a9b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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 Hideakie905a9e2007-02-09 23:24:47 +09007 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 * Authors: see ip.c
9 *
10 * Fixes:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090011 * Many : Split from ip.c , see ip_input.c for
Linus Torvalds1da177e2005-04-16 15:20:36 -070012 * history.
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +090013 * Dave Gregorich : NULL ip_rt_put fix for multicast
Linus Torvalds1da177e2005-04-16 15:20:36 -070014 * 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 Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/types.h>
23#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <linux/skbuff.h>
25#include <linux/ip.h>
26#include <linux/icmp.h>
27#include <linux/netdevice.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090028#include <linux/slab.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#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
Florian Westphalfe6cc552014-02-13 23:09:12 +010042static bool ip_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
43{
Florian Westphalca6c5d42014-05-04 23:24:31 +020044 if (skb->len <= mtu)
Florian Westphalfe6cc552014-02-13 23:09:12 +010045 return false;
46
Florian Westphalcf826242015-05-22 00:06:40 +020047 if (unlikely((ip_hdr(skb)->frag_off & htons(IP_DF)) == 0))
48 return false;
49
50 /* original fragment exceeds mtu and DF is set */
51 if (unlikely(IPCB(skb)->frag_max_size > mtu))
52 return true;
53
54 if (skb->ignore_df)
55 return false;
56
Florian Westphalfe6cc552014-02-13 23:09:12 +010057 if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu)
58 return false;
59
60 return true;
61}
62
Florian Westphalfe6cc552014-02-13 23:09:12 +010063
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -050064static int ip_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065{
Daniel Baluta5e73ea12012-04-15 01:34:41 +000066 struct ip_options *opt = &(IPCB(skb)->opt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
Eric W. Biedermanf8e1ac72015-09-15 20:03:55 -050068 IP_INC_STATS_BH(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
69 IP_ADD_STATS_BH(net, IPSTATS_MIB_OUTOCTETS, skb->len);
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
71 if (unlikely(opt->optlen))
72 ip_forward_options(skb);
73
Eric Dumazetc29390c2015-03-11 18:42:02 -070074 skb_sender_cpu_clear(skb);
Eric W. Biederman13206b62015-10-07 16:48:35 -050075 return dst_output(net, sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -070076}
77
78int ip_forward(struct sk_buff *skb)
79{
Hannes Frederic Sowaf87c10a2014-01-09 10:01:15 +010080 u32 mtu;
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 struct iphdr *iph; /* Our header */
82 struct rtable *rt; /* Route we use */
Daniel Baluta5e73ea12012-04-15 01:34:41 +000083 struct ip_options *opt = &(IPCB(skb)->opt);
Eric W. Biedermanfcad0ac2015-09-15 20:03:54 -050084 struct net *net;
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
Denis Kirjanovd4f2fa62014-02-13 08:58:56 +040086 /* that should never happen */
87 if (skb->pkt_type != PACKET_HOST)
88 goto drop;
89
Sebastian Pöhn2ab95742015-04-20 09:19:20 +020090 if (unlikely(skb->sk))
91 goto drop;
92
Ben Hutchings4497b072008-06-19 16:22:28 -070093 if (skb_warn_if_lro(skb))
94 goto drop;
95
Linus Torvalds1da177e2005-04-16 15:20:36 -070096 if (!xfrm4_policy_check(NULL, XFRM_POLICY_FWD, skb))
97 goto drop;
98
99 if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb))
100 return NET_RX_SUCCESS;
101
Herbert Xu35fc92a2007-03-26 23:22:20 -0700102 skb_forward_csum(skb);
Eric W. Biedermanfcad0ac2015-09-15 20:03:54 -0500103 net = dev_net(skb->dev);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900104
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 /*
106 * According to the RFC, we must first decrease the TTL field. If
107 * that reaches zero, we must reply an ICMP control message telling
108 * that the packet's lifetime expired.
109 */
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700110 if (ip_hdr(skb)->ttl <= 1)
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900111 goto too_many_hops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
113 if (!xfrm4_route_forward(skb))
114 goto drop;
115
Eric Dumazet511c3f92009-06-02 05:14:27 +0000116 rt = skb_rtable(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
Julian Anastasov155e8332012-10-08 11:41:18 +0000118 if (opt->is_strictroute && rt->rt_uses_gateway)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 goto sr_failed;
120
Hannes Frederic Sowaf87c10a2014-01-09 10:01:15 +0100121 IPCB(skb)->flags |= IPSKB_FORWARDED;
122 mtu = ip_dst_mtu_maybe_forward(&rt->dst, true);
Florian Westphalcf826242015-05-22 00:06:40 +0200123 if (ip_exceeds_mtu(skb, mtu)) {
Eric W. Biedermanfcad0ac2015-09-15 20:03:54 -0500124 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
John Heffner9af39122007-03-25 23:32:29 -0700125 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
Hannes Frederic Sowaf87c10a2014-01-09 10:01:15 +0100126 htonl(mtu));
John Heffner9af39122007-03-25 23:32:29 -0700127 goto drop;
128 }
129
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 /* We are about to mangle packet. Copy it! */
Changli Gaod8d1f302010-06-10 23:31:35 -0700131 if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 goto drop;
Arnaldo Carvalho de Meloeddc9ec2007-04-20 22:47:35 -0700133 iph = ip_hdr(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134
135 /* Decrease ttl after skb cow done */
136 ip_decrease_ttl(iph);
137
138 /*
139 * We now generate an ICMP HOST REDIRECT giving the route
140 * we calculated.
141 */
Hannes Frederic Sowadf4d9252015-01-23 12:01:26 +0100142 if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
143 !skb_sec_path(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144 ip_rt_send_redirect(skb);
145
146 skb->priority = rt_tos2priority(iph->tos);
147
Eric W. Biederman29a26a52015-09-15 20:04:16 -0500148 return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
149 net, NULL, skb, skb->dev, rt->dst.dev,
150 ip_forward_finish);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152sr_failed:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900153 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 * Strict routing permits no gatewaying
155 */
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900156 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0);
157 goto drop;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
159too_many_hops:
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900160 /* Tell the sender its packet died... */
Eric W. Biedermanfcad0ac2015-09-15 20:03:54 -0500161 IP_INC_STATS_BH(net, IPSTATS_MIB_INHDRERRORS);
YOSHIFUJI Hideakie905a9e2007-02-09 23:24:47 +0900162 icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163drop:
164 kfree_skb(skb);
165 return NET_RX_DROP;
166}