xref: /linux/net/ipv4/netfilter.c (revision f2ee442115c9b6219083c019939a9cc0c9abb2f8)
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 <linux/skbuff.h>
7 #include <linux/gfp.h>
8 #include <linux/export.h>
9 #include <net/route.h>
10 #include <net/xfrm.h>
11 #include <net/ip.h>
12 #include <net/netfilter/nf_queue.h>
13 
14 /* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
15 int ip_route_me_harder(struct sk_buff *skb, unsigned addr_type)
16 {
17 	struct net *net = dev_net(skb_dst(skb)->dev);
18 	const struct iphdr *iph = ip_hdr(skb);
19 	struct rtable *rt;
20 	struct flowi4 fl4 = {};
21 	__be32 saddr = iph->saddr;
22 	__u8 flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
23 	unsigned int hh_len;
24 
25 	if (addr_type == RTN_UNSPEC)
26 		addr_type = inet_addr_type(net, saddr);
27 	if (addr_type == RTN_LOCAL || addr_type == RTN_UNICAST)
28 		flags |= FLOWI_FLAG_ANYSRC;
29 	else
30 		saddr = 0;
31 
32 	/* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
33 	 * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
34 	 */
35 	fl4.daddr = iph->daddr;
36 	fl4.saddr = saddr;
37 	fl4.flowi4_tos = RT_TOS(iph->tos);
38 	fl4.flowi4_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
39 	fl4.flowi4_mark = skb->mark;
40 	fl4.flowi4_flags = flags;
41 	rt = ip_route_output_key(net, &fl4);
42 	if (IS_ERR(rt))
43 		return -1;
44 
45 	/* Drop old route. */
46 	skb_dst_drop(skb);
47 	skb_dst_set(skb, &rt->dst);
48 
49 	if (skb_dst(skb)->error)
50 		return -1;
51 
52 #ifdef CONFIG_XFRM
53 	if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
54 	    xfrm_decode_session(skb, flowi4_to_flowi(&fl4), AF_INET) == 0) {
55 		struct dst_entry *dst = skb_dst(skb);
56 		skb_dst_set(skb, NULL);
57 		dst = xfrm_lookup(net, dst, flowi4_to_flowi(&fl4), skb->sk, 0);
58 		if (IS_ERR(dst))
59 			return -1;
60 		skb_dst_set(skb, dst);
61 	}
62 #endif
63 
64 	/* Change in oif may mean change in hh_len. */
65 	hh_len = skb_dst(skb)->dev->hard_header_len;
66 	if (skb_headroom(skb) < hh_len &&
67 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
68 		return -1;
69 
70 	return 0;
71 }
72 EXPORT_SYMBOL(ip_route_me_harder);
73 
74 #ifdef CONFIG_XFRM
75 int ip_xfrm_me_harder(struct sk_buff *skb)
76 {
77 	struct flowi fl;
78 	unsigned int hh_len;
79 	struct dst_entry *dst;
80 
81 	if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
82 		return 0;
83 	if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
84 		return -1;
85 
86 	dst = skb_dst(skb);
87 	if (dst->xfrm)
88 		dst = ((struct xfrm_dst *)dst)->route;
89 	dst_hold(dst);
90 
91 	dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0);
92 	if (IS_ERR(dst))
93 		return -1;
94 
95 	skb_dst_drop(skb);
96 	skb_dst_set(skb, dst);
97 
98 	/* Change in oif may mean change in hh_len. */
99 	hh_len = skb_dst(skb)->dev->hard_header_len;
100 	if (skb_headroom(skb) < hh_len &&
101 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
102 		return -1;
103 	return 0;
104 }
105 EXPORT_SYMBOL(ip_xfrm_me_harder);
106 #endif
107 
108 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
109 EXPORT_SYMBOL(ip_nat_decode_session);
110 
111 /*
112  * Extra routing may needed on local out, as the QUEUE target never
113  * returns control to the table.
114  */
115 
116 struct ip_rt_info {
117 	__be32 daddr;
118 	__be32 saddr;
119 	u_int8_t tos;
120 	u_int32_t mark;
121 };
122 
123 static void nf_ip_saveroute(const struct sk_buff *skb,
124 			    struct nf_queue_entry *entry)
125 {
126 	struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
127 
128 	if (entry->hook == NF_INET_LOCAL_OUT) {
129 		const struct iphdr *iph = ip_hdr(skb);
130 
131 		rt_info->tos = iph->tos;
132 		rt_info->daddr = iph->daddr;
133 		rt_info->saddr = iph->saddr;
134 		rt_info->mark = skb->mark;
135 	}
136 }
137 
138 static int nf_ip_reroute(struct sk_buff *skb,
139 			 const struct nf_queue_entry *entry)
140 {
141 	const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
142 
143 	if (entry->hook == NF_INET_LOCAL_OUT) {
144 		const struct iphdr *iph = ip_hdr(skb);
145 
146 		if (!(iph->tos == rt_info->tos &&
147 		      skb->mark == rt_info->mark &&
148 		      iph->daddr == rt_info->daddr &&
149 		      iph->saddr == rt_info->saddr))
150 			return ip_route_me_harder(skb, RTN_UNSPEC);
151 	}
152 	return 0;
153 }
154 
155 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
156 			    unsigned int dataoff, u_int8_t protocol)
157 {
158 	const struct iphdr *iph = ip_hdr(skb);
159 	__sum16 csum = 0;
160 
161 	switch (skb->ip_summed) {
162 	case CHECKSUM_COMPLETE:
163 		if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
164 			break;
165 		if ((protocol == 0 && !csum_fold(skb->csum)) ||
166 		    !csum_tcpudp_magic(iph->saddr, iph->daddr,
167 				       skb->len - dataoff, protocol,
168 				       skb->csum)) {
169 			skb->ip_summed = CHECKSUM_UNNECESSARY;
170 			break;
171 		}
172 		/* fall through */
173 	case CHECKSUM_NONE:
174 		if (protocol == 0)
175 			skb->csum = 0;
176 		else
177 			skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
178 						       skb->len - dataoff,
179 						       protocol, 0);
180 		csum = __skb_checksum_complete(skb);
181 	}
182 	return csum;
183 }
184 EXPORT_SYMBOL(nf_ip_checksum);
185 
186 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
187 				      unsigned int dataoff, unsigned int len,
188 				      u_int8_t protocol)
189 {
190 	const struct iphdr *iph = ip_hdr(skb);
191 	__sum16 csum = 0;
192 
193 	switch (skb->ip_summed) {
194 	case CHECKSUM_COMPLETE:
195 		if (len == skb->len - dataoff)
196 			return nf_ip_checksum(skb, hook, dataoff, protocol);
197 		/* fall through */
198 	case CHECKSUM_NONE:
199 		skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
200 					       skb->len - dataoff, 0);
201 		skb->ip_summed = CHECKSUM_NONE;
202 		return __skb_checksum_complete_head(skb, dataoff + len);
203 	}
204 	return csum;
205 }
206 
207 static int nf_ip_route(struct net *net, struct dst_entry **dst,
208 		       struct flowi *fl, bool strict __always_unused)
209 {
210 	struct rtable *rt = ip_route_output_key(net, &fl->u.ip4);
211 	if (IS_ERR(rt))
212 		return PTR_ERR(rt);
213 	*dst = &rt->dst;
214 	return 0;
215 }
216 
217 static const struct nf_afinfo nf_ip_afinfo = {
218 	.family			= AF_INET,
219 	.checksum		= nf_ip_checksum,
220 	.checksum_partial	= nf_ip_checksum_partial,
221 	.route			= nf_ip_route,
222 	.saveroute		= nf_ip_saveroute,
223 	.reroute		= nf_ip_reroute,
224 	.route_key_size		= sizeof(struct ip_rt_info),
225 };
226 
227 static int ipv4_netfilter_init(void)
228 {
229 	return nf_register_afinfo(&nf_ip_afinfo);
230 }
231 
232 static void ipv4_netfilter_fini(void)
233 {
234 	nf_unregister_afinfo(&nf_ip_afinfo);
235 }
236 
237 module_init(ipv4_netfilter_init);
238 module_exit(ipv4_netfilter_fini);
239 
240 #ifdef CONFIG_SYSCTL
241 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
242 	{ .procname = "net", },
243 	{ .procname = "ipv4", },
244 	{ .procname = "netfilter", },
245 	{ }
246 };
247 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
248 #endif /* CONFIG_SYSCTL */
249