1/* IPv4 specific functions of netfilter core */
2#include <linux/kernel.h>
3#include <linux/netfilter.h>
4#include <linux/netfilter_ipv4.h>
5#include <linux/ip.h>
6#include <net/route.h>
7#include <net/xfrm.h>
8#include <net/ip.h>
9
10/* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
11int ip_route_me_harder(struct sk_buff **pskb, unsigned addr_type)
12{
13	const struct iphdr *iph = ip_hdr(*pskb);
14	struct rtable *rt;
15	struct flowi fl = {};
16	struct dst_entry *odst;
17	unsigned int hh_len;
18	unsigned int type;
19
20	type = inet_addr_type(iph->saddr);
21	if (addr_type == RTN_UNSPEC)
22		addr_type = type;
23
24	/* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
25	 * packets with foreign saddr to appear on the NF_IP_LOCAL_OUT hook.
26	 */
27	if (addr_type == RTN_LOCAL) {
28		fl.nl_u.ip4_u.daddr = iph->daddr;
29		if (type == RTN_LOCAL)
30			fl.nl_u.ip4_u.saddr = iph->saddr;
31		fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
32		fl.oif = (*pskb)->sk ? (*pskb)->sk->sk_bound_dev_if : 0;
33		fl.mark = (*pskb)->mark;
34		if (ip_route_output_key(&rt, &fl) != 0)
35			return -1;
36
37		/* Drop old route. */
38		dst_release((*pskb)->dst);
39		(*pskb)->dst = &rt->u.dst;
40	} else {
41		/* non-local src, find valid iif to satisfy
42		 * rp-filter when calling ip_route_input. */
43		fl.nl_u.ip4_u.daddr = iph->saddr;
44		if (ip_route_output_key(&rt, &fl) != 0)
45			return -1;
46
47		odst = (*pskb)->dst;
48		if (ip_route_input(*pskb, iph->daddr, iph->saddr,
49				   RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
50			dst_release(&rt->u.dst);
51			return -1;
52		}
53		dst_release(&rt->u.dst);
54		dst_release(odst);
55	}
56
57	if ((*pskb)->dst->error)
58		return -1;
59
60#ifdef CONFIG_XFRM
61	if (!(IPCB(*pskb)->flags & IPSKB_XFRM_TRANSFORMED) &&
62	    xfrm_decode_session(*pskb, &fl, AF_INET) == 0)
63		if (xfrm_lookup(&(*pskb)->dst, &fl, (*pskb)->sk, 0))
64			return -1;
65#endif
66
67	/* Change in oif may mean change in hh_len. */
68	hh_len = (*pskb)->dst->dev->hard_header_len;
69	if (skb_headroom(*pskb) < hh_len) {
70		struct sk_buff *nskb;
71
72		nskb = skb_realloc_headroom(*pskb, hh_len);
73		if (!nskb)
74			return -1;
75		if ((*pskb)->sk)
76			skb_set_owner_w(nskb, (*pskb)->sk);
77		kfree_skb(*pskb);
78		*pskb = nskb;
79	}
80
81	return 0;
82}
83EXPORT_SYMBOL(ip_route_me_harder);
84
85#ifdef CONFIG_XFRM
86int ip_xfrm_me_harder(struct sk_buff **pskb)
87{
88	struct flowi fl;
89	unsigned int hh_len;
90	struct dst_entry *dst;
91
92	if (IPCB(*pskb)->flags & IPSKB_XFRM_TRANSFORMED)
93		return 0;
94	if (xfrm_decode_session(*pskb, &fl, AF_INET) < 0)
95		return -1;
96
97	dst = (*pskb)->dst;
98	if (dst->xfrm)
99		dst = ((struct xfrm_dst *)dst)->route;
100	dst_hold(dst);
101
102	if (xfrm_lookup(&dst, &fl, (*pskb)->sk, 0) < 0)
103		return -1;
104
105	dst_release((*pskb)->dst);
106	(*pskb)->dst = dst;
107
108	/* Change in oif may mean change in hh_len. */
109	hh_len = (*pskb)->dst->dev->hard_header_len;
110	if (skb_headroom(*pskb) < hh_len) {
111		struct sk_buff *nskb;
112
113		nskb = skb_realloc_headroom(*pskb, hh_len);
114		if (!nskb)
115			return -1;
116		if ((*pskb)->sk)
117			skb_set_owner_w(nskb, (*pskb)->sk);
118		kfree_skb(*pskb);
119		*pskb = nskb;
120	}
121	return 0;
122}
123EXPORT_SYMBOL(ip_xfrm_me_harder);
124#endif
125
126void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
127EXPORT_SYMBOL(ip_nat_decode_session);
128
129/*
130 * Extra routing may needed on local out, as the QUEUE target never
131 * returns control to the table.
132 */
133
134struct ip_rt_info {
135	__be32 daddr;
136	__be32 saddr;
137	u_int8_t tos;
138};
139
140static void nf_ip_saveroute(const struct sk_buff *skb, struct nf_info *info)
141{
142	struct ip_rt_info *rt_info = nf_info_reroute(info);
143
144	if (info->hook == NF_IP_LOCAL_OUT) {
145		const struct iphdr *iph = ip_hdr(skb);
146
147		rt_info->tos = iph->tos;
148		rt_info->daddr = iph->daddr;
149		rt_info->saddr = iph->saddr;
150	}
151}
152
153static int nf_ip_reroute(struct sk_buff **pskb, const struct nf_info *info)
154{
155	const struct ip_rt_info *rt_info = nf_info_reroute(info);
156
157	if (info->hook == NF_IP_LOCAL_OUT) {
158		const struct iphdr *iph = ip_hdr(*pskb);
159
160		if (!(iph->tos == rt_info->tos
161		      && iph->daddr == rt_info->daddr
162		      && iph->saddr == rt_info->saddr))
163			return ip_route_me_harder(pskb, RTN_UNSPEC);
164	}
165	return 0;
166}
167
168__sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
169			    unsigned int dataoff, u_int8_t protocol)
170{
171	const struct iphdr *iph = ip_hdr(skb);
172	__sum16 csum = 0;
173
174	switch (skb->ip_summed) {
175	case CHECKSUM_COMPLETE:
176		if (hook != NF_IP_PRE_ROUTING && hook != NF_IP_LOCAL_IN)
177			break;
178		if ((protocol == 0 && !csum_fold(skb->csum)) ||
179		    !csum_tcpudp_magic(iph->saddr, iph->daddr,
180				       skb->len - dataoff, protocol,
181				       skb->csum)) {
182			skb->ip_summed = CHECKSUM_UNNECESSARY;
183			break;
184		}
185		/* fall through */
186	case CHECKSUM_NONE:
187		if (protocol == 0)
188			skb->csum = 0;
189		else
190			skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
191						       skb->len - dataoff,
192						       protocol, 0);
193		csum = __skb_checksum_complete(skb);
194	}
195	return csum;
196}
197
198EXPORT_SYMBOL(nf_ip_checksum);
199
200static struct nf_afinfo nf_ip_afinfo = {
201	.family		= AF_INET,
202	.checksum	= nf_ip_checksum,
203	.saveroute	= nf_ip_saveroute,
204	.reroute	= nf_ip_reroute,
205	.route_key_size	= sizeof(struct ip_rt_info),
206};
207
208static int ipv4_netfilter_init(void)
209{
210	return nf_register_afinfo(&nf_ip_afinfo);
211}
212
213static void ipv4_netfilter_fini(void)
214{
215	nf_unregister_afinfo(&nf_ip_afinfo);
216}
217
218module_init(ipv4_netfilter_init);
219module_exit(ipv4_netfilter_fini);
220