tcp_bbr: reset full pipe detection on loss recovery undo
[linux-2.6-block.git] / net / ipv4 / ip_forward.c
CommitLineData
b2441318 1// SPDX-License-Identifier: GPL-2.0
1da177e4
LT
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.
e905a9ed 8 *
1da177e4
LT
9 * Authors: see ip.c
10 *
11 * Fixes:
e905a9ed 12 * Many : Split from ip.c , see ip_input.c for
1da177e4 13 * history.
e905a9ed 14 * Dave Gregorich : NULL ip_rt_put fix for multicast
1da177e4
LT
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
1da177e4
LT
23#include <linux/types.h>
24#include <linux/mm.h>
1da177e4
LT
25#include <linux/skbuff.h>
26#include <linux/ip.h>
27#include <linux/icmp.h>
28#include <linux/netdevice.h>
5a0e3ad6 29#include <linux/slab.h>
1da177e4
LT
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
fe6cc55f
FW
43static bool ip_exceeds_mtu(const struct sk_buff *skb, unsigned int mtu)
44{
ca6c5d4a 45 if (skb->len <= mtu)
fe6cc55f
FW
46 return false;
47
cf826244
FW
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
ae7ef81e 58 if (skb_is_gso(skb) && skb_gso_validate_mtu(skb, mtu))
fe6cc55f
FW
59 return false;
60
61 return true;
62}
63
fe6cc55f 64
0c4b51f0 65static int ip_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
1da177e4 66{
5e73ea1a 67 struct ip_options *opt = &(IPCB(skb)->opt);
1da177e4 68
b45386ef 69 __IP_INC_STATS(net, IPSTATS_MIB_OUTFORWDATAGRAMS);
98f61995 70 __IP_ADD_STATS(net, IPSTATS_MIB_OUTOCTETS, skb->len);
1da177e4
LT
71
72 if (unlikely(opt->optlen))
73 ip_forward_options(skb);
74
13206b6b 75 return dst_output(net, sk, skb);
1da177e4
LT
76}
77
78int ip_forward(struct sk_buff *skb)
79{
f87c10a8 80 u32 mtu;
1da177e4
LT
81 struct iphdr *iph; /* Our header */
82 struct rtable *rt; /* Route we use */
5e73ea1a 83 struct ip_options *opt = &(IPCB(skb)->opt);
fcad0ac2 84 struct net *net;
1da177e4 85
d4f2fa6a
DK
86 /* that should never happen */
87 if (skb->pkt_type != PACKET_HOST)
88 goto drop;
89
2ab95749
SP
90 if (unlikely(skb->sk))
91 goto drop;
92
4497b076
BH
93 if (skb_warn_if_lro(skb))
94 goto drop;
95
1da177e4
LT
96 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
35fc92a9 102 skb_forward_csum(skb);
fcad0ac2 103 net = dev_net(skb->dev);
e905a9ed 104
1da177e4
LT
105 /*
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 */
eddc9ec5 110 if (ip_hdr(skb)->ttl <= 1)
e905a9ed 111 goto too_many_hops;
1da177e4
LT
112
113 if (!xfrm4_route_forward(skb))
114 goto drop;
115
511c3f92 116 rt = skb_rtable(skb);
1da177e4 117
155e8336 118 if (opt->is_strictroute && rt->rt_uses_gateway)
1da177e4
LT
119 goto sr_failed;
120
9ee6c5dc 121 IPCB(skb)->flags |= IPSKB_FORWARDED;
f87c10a8 122 mtu = ip_dst_mtu_maybe_forward(&rt->dst, true);
cf826244 123 if (ip_exceeds_mtu(skb, mtu)) {
fcad0ac2 124 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
9af3912e 125 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
f87c10a8 126 htonl(mtu));
9af3912e
JH
127 goto drop;
128 }
129
1da177e4 130 /* We are about to mangle packet. Copy it! */
d8d1f30b 131 if (skb_cow(skb, LL_RESERVED_SPACE(rt->dst.dev)+rt->dst.header_len))
1da177e4 132 goto drop;
eddc9ec5 133 iph = ip_hdr(skb);
1da177e4
LT
134
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 */
df4d9254
HFS
142 if (IPCB(skb)->flags & IPSKB_DOREDIRECT && !opt->srr &&
143 !skb_sec_path(skb))
1da177e4
LT
144 ip_rt_send_redirect(skb);
145
146 skb->priority = rt_tos2priority(iph->tos);
147
29a26a56
EB
148 return NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD,
149 net, NULL, skb, skb->dev, rt->dst.dev,
150 ip_forward_finish);
1da177e4
LT
151
152sr_failed:
e905a9ed 153 /*
1da177e4
LT
154 * Strict routing permits no gatewaying
155 */
e905a9ed
YH
156 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_SR_FAILED, 0);
157 goto drop;
1da177e4
LT
158
159too_many_hops:
e905a9ed 160 /* Tell the sender its packet died... */
b45386ef 161 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
e905a9ed 162 icmp_send(skb, ICMP_TIME_EXCEEDED, ICMP_EXC_TTL, 0);
1da177e4
LT
163drop:
164 kfree_skb(skb);
165 return NET_RX_DROP;
166}