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
23#include <linux/types.h>
24#include <linux/mm.h>
25#include <linux/skbuff.h>
26#include <linux/ip.h>
27#include <linux/icmp.h>
28#include <linux/netdevice.h>
29#include <linux/slab.h>
30#include <net/sock.h>
31#include <net/ip.h>
32#include <net/tcp.h>
33#include <net/udp.h>
34#include <net/icmp.h>
35#include <linux/tcp.h>
36#include <linux/udp.h>
37#include <linux/netfilter_ipv4.h>
38#include <net/checksum.h>
39#include <linux/route.h>
40#include <net/route.h>
41#include <net/xfrm.h>
42
43static bool ip_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
44{
45 if (skb->len <= mtu)
46 return false;
47
48 if (unlikely((ip_hdr(skb)->frag_off & htons(IP_DF)) == 0))
49 return false;
50
51 /* original fragment exceeds mtu and DF is set */
52 if (unlikely(IPCB(skb)->frag_max_size > mtu))
53 return true;
54
55 if (skb->ignore_df)
56 return false;
57
58 if (skb_is_gso(skb) && skb_gso_validate_network_len(skb, mtu))
59 return false;
60
61 return true;
62}
63
64
65static int ip_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
66{
67 struct ip_options *opt = &(IPCB(skb)->opt);
68
69#ifdef CONFIG_NET_SWITCHDEV
70 if (skb->offload_l3_fwd_mark) {
71 consume_skb(skb);
72 return 0;
73 }
74#endif
75
76 if (unlikely(opt->optlen))
77 ip_forward_options(skb);
78
79 skb_clear_tstamp(skb);
80 return dst_output(net, sk, skb);
81}
82
83int ip_forward(struct sk_buff *skb)
84{
85 u32 mtu;
86 struct iphdr *iph; /* Our header */
87 struct rtable *rt; /* Route we use */
88 struct ip_options *opt = &(IPCB(skb)->opt);
89 struct net *net;
90 SKB_DR(reason);
91
92 /* that should never happen */
93 if (skb->pkt_type != PACKET_HOST)
94 goto drop;
95
96 if (unlikely(skb->sk))
97 goto drop;
98
99 if (skb_warn_if_lro(skb))
100 goto drop;
101
102 if (!xfrm4_policy_check(NULL, dir: XFRM_POLICY_FWD, skb)) {
103 SKB_DR_SET(reason, XFRM_POLICY);
104 goto drop;
105 }
106
107 if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb))
108 return NET_RX_SUCCESS;
109
110 skb_forward_csum(skb);
111 net = dev_net(dev: skb->dev);
112
113 /*
114 * According to the RFC, we must first decrease the TTL field. If
115 * that reaches zero, we must reply an ICMP control message telling
116 * that the packet's lifetime expired.
117 */
118 if (ip_hdr(skb)->ttl <= 1)
119 goto too_many_hops;
120
121 if (!xfrm4_route_forward(skb)) {
122 SKB_DR_SET(reason, XFRM_POLICY);
123 goto drop;
124 }
125
126 rt = skb_rtable(skb);
127
128 if (opt->is_strictroute && rt->rt_uses_gateway)
129 goto sr_failed;
130
131 __IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
132
133 IPCB(skb)->flags |= IPSKB_FORWARDED;
134 mtu = ip_dst_mtu_maybe_forward(dst: &rt->dst, forwarding: true);
135 if (ip_exceeds_mtu(skb, mtu)) {
136 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
137 icmp_send(skb_in: skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
138 htonl(mtu));
139 SKB_DR_SET(reason, PKT_TOO_BIG);
140 goto drop;
141 }
142
143 /* We are about to mangle packet. Copy it! */
144 if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
145 goto drop;
146 iph = ip_hdr(skb);
147
148 /* Decrease ttl after skb cow done */
149 ip_decrease_ttl(iph);
150
151 /*
152 * We now generate an ICMP HOST REDIRECT giving the route
153 * we calculated.
154 */
155 if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
156 !skb_sec_path(skb))
157 ip_rt_send_redirect(skb);
158
159 if (READ_ONCE(net->ipv4.sysctl_ip_fwd_update_priority))
160 skb->priority = rt_tos2priority(tos: iph->tos);
161
162 return NF_HOOK(pf: NFPROTO_IPV4, hook: NF_INET_FORWARD,
163 net, NULL, skb, in: skb->dev, out: rt->dst.dev,
164 okfn: ip_forward_finish);
165
166sr_failed:
167 /*
168 * Strict routing permits no gatewaying
169 */
170 icmp_send(skb_in: skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, info: 0);
171 goto drop;
172
173too_many_hops:
174 /* Tell the sender its packet died... */
175 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
176 icmp_send(skb_in: skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, info: 0);
177 SKB_DR_SET(reason, IP_INHDR);
178drop:
179 kfree_skb_reason(skb, reason);
180 return NET_RX_DROP;
181}
182