1/* iptables module for the IPv4 and TCP ECN bits, Version 1.5
2 *
3 * (C) 2002 by Harald Welte <laforge@netfilter.org>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
8*/
9#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10#include <linux/in.h>
11#include <linux/module.h>
12#include <linux/skbuff.h>
13#include <linux/ip.h>
14#include <net/ip.h>
15#include <linux/tcp.h>
16#include <net/checksum.h>
17
18#include <linux/netfilter/x_tables.h>
19#include <linux/netfilter_ipv4/ip_tables.h>
20#include <linux/netfilter_ipv4/ipt_ECN.h>
21
22MODULE_LICENSE("GPL");
23MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
24MODULE_DESCRIPTION("Xtables: Explicit Congestion Notification (ECN) flag modification");
25
26/* set ECT codepoint from IP header.
27 * 	return false if there was an error. */
28static inline bool
29set_ect_ip(struct sk_buff *skb, const struct ipt_ECN_info *einfo)
30{
31	struct iphdr *iph = ip_hdr(skb);
32
33	if ((iph->tos & IPT_ECN_IP_MASK) != (einfo->ip_ect & IPT_ECN_IP_MASK)) {
34		__u8 oldtos;
35		if (!skb_make_writable(skb, sizeof(struct iphdr)))
36			return false;
37		iph = ip_hdr(skb);
38		oldtos = iph->tos;
39		iph->tos &= ~IPT_ECN_IP_MASK;
40		iph->tos |= (einfo->ip_ect & IPT_ECN_IP_MASK);
41		csum_replace2(&iph->check, htons(oldtos), htons(iph->tos));
42	}
43	return true;
44}
45
46/* Return false if there was an error. */
47static inline bool
48set_ect_tcp(struct sk_buff *skb, const struct ipt_ECN_info *einfo)
49{
50	struct tcphdr _tcph, *tcph;
51	__be16 oldval;
52
53	/* Not enough header? */
54	tcph = skb_header_pointer(skb, ip_hdrlen(skb), sizeof(_tcph), &_tcph);
55	if (!tcph)
56		return false;
57
58	if ((!(einfo->operation & IPT_ECN_OP_SET_ECE) ||
59	     tcph->ece == einfo->proto.tcp.ece) &&
60	    (!(einfo->operation & IPT_ECN_OP_SET_CWR) ||
61	     tcph->cwr == einfo->proto.tcp.cwr))
62		return true;
63
64	if (!skb_make_writable(skb, ip_hdrlen(skb) + sizeof(*tcph)))
65		return false;
66	tcph = (void *)ip_hdr(skb) + ip_hdrlen(skb);
67
68	oldval = ((__be16 *)tcph)[6];
69	if (einfo->operation & IPT_ECN_OP_SET_ECE)
70		tcph->ece = einfo->proto.tcp.ece;
71	if (einfo->operation & IPT_ECN_OP_SET_CWR)
72		tcph->cwr = einfo->proto.tcp.cwr;
73
74	inet_proto_csum_replace2(&tcph->check, skb,
75				 oldval, ((__be16 *)tcph)[6], 0);
76	return true;
77}
78
79static unsigned int
80ecn_tg(struct sk_buff *skb, const struct xt_action_param *par)
81{
82	const struct ipt_ECN_info *einfo = par->targinfo;
83
84	if (einfo->operation & IPT_ECN_OP_SET_IP)
85		if (!set_ect_ip(skb, einfo))
86			return NF_DROP;
87
88	if (einfo->operation & (IPT_ECN_OP_SET_ECE | IPT_ECN_OP_SET_CWR) &&
89	    ip_hdr(skb)->protocol == IPPROTO_TCP)
90		if (!set_ect_tcp(skb, einfo))
91			return NF_DROP;
92
93	return XT_CONTINUE;
94}
95
96static int ecn_tg_check(const struct xt_tgchk_param *par)
97{
98	const struct ipt_ECN_info *einfo = par->targinfo;
99	const struct ipt_entry *e = par->entryinfo;
100
101	if (einfo->operation & IPT_ECN_OP_MASK) {
102		pr_info("unsupported ECN operation %x\n", einfo->operation);
103		return -EINVAL;
104	}
105	if (einfo->ip_ect & ~IPT_ECN_IP_MASK) {
106		pr_info("new ECT codepoint %x out of mask\n", einfo->ip_ect);
107		return -EINVAL;
108	}
109	if ((einfo->operation & (IPT_ECN_OP_SET_ECE|IPT_ECN_OP_SET_CWR)) &&
110	    (e->ip.proto != IPPROTO_TCP || (e->ip.invflags & XT_INV_PROTO))) {
111		pr_info("cannot use TCP operations on a non-tcp rule\n");
112		return -EINVAL;
113	}
114	return 0;
115}
116
117static struct xt_target ecn_tg_reg __read_mostly = {
118	.name		= "ECN",
119	.family		= NFPROTO_IPV4,
120	.target		= ecn_tg,
121	.targetsize	= sizeof(struct ipt_ECN_info),
122	.table		= "mangle",
123	.checkentry	= ecn_tg_check,
124	.me		= THIS_MODULE,
125};
126
127static int __init ecn_tg_init(void)
128{
129	return xt_register_target(&ecn_tg_reg);
130}
131
132static void __exit ecn_tg_exit(void)
133{
134	xt_unregister_target(&ecn_tg_reg);
135}
136
137module_init(ecn_tg_init);
138module_exit(ecn_tg_exit);
139