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
42static 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
48static 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
60static int ip_forward_finish(struct sock *sk, 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	skb_sender_cpu_clear(skb);
71	return dst_output_sk(sk, skb);
72}
73
74int ip_forward(struct sk_buff *skb)
75{
76	u32 mtu;
77	struct iphdr *iph;	/* Our header */
78	struct rtable *rt;	/* Route we use */
79	struct ip_options *opt	= &(IPCB(skb)->opt);
80
81	/* that should never happen */
82	if (skb->pkt_type != PACKET_HOST)
83		goto drop;
84
85	if (unlikely(skb->sk))
86		goto drop;
87
88	if (skb_warn_if_lro(skb))
89		goto drop;
90
91	if (!xfrm4_policy_check(NULL, XFRM_POLICY_FWD, skb))
92		goto drop;
93
94	if (IPCB(skb)->opt.router_alert && ip_call_ra_chain(skb))
95		return NET_RX_SUCCESS;
96
97	skb_forward_csum(skb);
98
99	/*
100	 *	According to the RFC, we must first decrease the TTL field. If
101	 *	that reaches zero, we must reply an ICMP control message telling
102	 *	that the packet's lifetime expired.
103	 */
104	if (ip_hdr(skb)->ttl <= 1)
105		goto too_many_hops;
106
107	if (!xfrm4_route_forward(skb))
108		goto drop;
109
110	rt = skb_rtable(skb);
111
112	if (opt->is_strictroute && rt->rt_uses_gateway)
113		goto sr_failed;
114
115	IPCB(skb)->flags |= IPSKB_FORWARDED;
116	mtu = ip_dst_mtu_maybe_forward(&rt->dst, true);
117	if (!ip_may_fragment(skb) && ip_exceeds_mtu(skb, mtu)) {
118		IP_INC_STATS(dev_net(rt->dst.dev), IPSTATS_MIB_FRAGFAILS);
119		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
120			  htonl(mtu));
121		goto drop;
122	}
123
124	/* We are about to mangle packet. Copy it! */
125	if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
126		goto drop;
127	iph = ip_hdr(skb);
128
129	/* Decrease ttl after skb cow done */
130	ip_decrease_ttl(iph);
131
132	/*
133	 *	We now generate an ICMP HOST REDIRECT giving the route
134	 *	we calculated.
135	 */
136	if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
137	    !skb_sec_path(skb))
138		ip_rt_send_redirect(skb);
139
140	skb->priority = rt_tos2priority(iph->tos);
141
142	return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, NULL, skb,
143		       skb->dev, rt->dst.dev, ip_forward_finish);
144
145sr_failed:
146	/*
147	 *	Strict routing permits no gatewaying
148	 */
149	 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0);
150	 goto drop;
151
152too_many_hops:
153	/* Tell the sender its packet died... */
154	IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_INHDRERRORS);
155	icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0);
156drop:
157	kfree_skb(skb);
158	return NET_RX_DROP;
159}
160