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.
7 *
8 * Authors: see ip.c
9 *
10 * Fixes:
11 * Many : Split from ip.c , see ip_input.c for
12 * history.
13 * Dave Gregorich : NULL ip_rt_put fix for multicast
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
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
ip_may_fragment(const struct sk_buff * skb)42 static bool ip_may_fragment(const struct sk_buff *skb)
43 {
44 return unlikely((ip_hdr(skb)->frag_off & htons(IP_DF)) == 0) ||
45 skb->ignore_df;
46 }
47
ip_exceeds_mtu(const struct sk_buff * skb,unsigned int mtu)48 static bool ip_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
49 {
50 if (skb->len <= mtu)
51 return false;
52
53 if (skb_is_gso(skb) && skb_gso_network_seglen(skb) <= mtu)
54 return false;
55
56 return true;
57 }
58
59
ip_forward_finish(struct sk_buff * skb)60 static int ip_forward_finish(struct sk_buff *skb)
61 {
62 struct ip_options *opt = &(IPCB(skb)->opt);
63
64 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
65 IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
66
67 if (unlikely(opt->optlen))
68 ip_forward_options(skb);
69
70 return dst_output(skb);
71 }
72
ip_forward(struct sk_buff * skb)73 int ip_forward(struct sk_buff *skb)
74 {
75 u32 mtu;
76 struct iphdr *iph; /* Our header */
77 struct rtable *rt; /* Route we use */
78 struct ip_options *opt = &(IPCB(skb)->opt);
79
80 /* that should never happen */
81 if (skb->pkt_type != PACKET_HOST)
82 goto drop;
83
84 if (unlikely(skb->sk))
85 goto drop;
86
87 if (skb_warn_if_lro(skb))
88 goto drop;
89
90 if (!xfrm4_policy_check(NULL, XFRM_POLICY_FWD, skb))
91 goto drop;
92
93 if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb))
94 return NET_RX_SUCCESS;
95
96 skb_forward_csum(skb);
97
98 /*
99 * According to the RFC, we must first decrease the TTL field. If
100 * that reaches zero, we must reply an ICMP control message telling
101 * that the packet's lifetime expired.
102 */
103 if (ip_hdr(skb)->ttl <= 1)
104 goto too_many_hops;
105
106 if (!xfrm4_route_forward(skb))
107 goto drop;
108
109 rt = skb_rtable(skb);
110
111 if (opt->is_strictroute && rt->rt_uses_gateway)
112 goto sr_failed;
113
114 IPCB(skb)->flags |= IPSKB_FORWARDED;
115 mtu = ip_dst_mtu_maybe_forward(&rt->dst, true);
116 if (!ip_may_fragment(skb) && ip_exceeds_mtu(skb, mtu)) {
117 IP_INC_STATS(dev_net(rt->dst.dev), IPSTATS_MIB_FRAGFAILS);
118 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
119 htonl(mtu));
120 goto drop;
121 }
122
123 /* We are about to mangle packet. Copy it! */
124 if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
125 goto drop;
126 iph = ip_hdr(skb);
127
128 /* Decrease ttl after skb cow done */
129 ip_decrease_ttl(iph);
130
131 /*
132 * We now generate an ICMP HOST REDIRECT giving the route
133 * we calculated.
134 */
135 if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
136 !skb_sec_path(skb))
137 ip_rt_send_redirect(skb);
138
139 skb->priority = rt_tos2priority(iph->tos);
140
141 return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev,
142 rt->dst.dev, ip_forward_finish);
143
144 sr_failed:
145 /*
146 * Strict routing permits no gatewaying
147 */
148 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0);
149 goto drop;
150
151 too_many_hops:
152 /* Tell the sender its packet died... */
153 IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_INHDRERRORS);
154 icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0);
155 drop:
156 kfree_skb(skb);
157 return NET_RX_DROP;
158 }
159