xref: /openbmc/linux/net/ipv4/netfilter.c (revision b627b4ed)
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 <net/route.h>
8 #include <net/xfrm.h>
9 #include <net/ip.h>
10 #include <net/netfilter/nf_queue.h>
11 
12 /* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
13 int ip_route_me_harder(struct sk_buff *skb, unsigned addr_type)
14 {
15 	struct net *net = dev_net(skb->dst->dev);
16 	const struct iphdr *iph = ip_hdr(skb);
17 	struct rtable *rt;
18 	struct flowi fl = {};
19 	struct dst_entry *odst;
20 	unsigned int hh_len;
21 	unsigned int type;
22 
23 	type = inet_addr_type(net, iph->saddr);
24 	if (skb->sk && inet_sk(skb->sk)->transparent)
25 		type = RTN_LOCAL;
26 	if (addr_type == RTN_UNSPEC)
27 		addr_type = type;
28 
29 	/* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
30 	 * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
31 	 */
32 	if (addr_type == RTN_LOCAL) {
33 		fl.nl_u.ip4_u.daddr = iph->daddr;
34 		if (type == RTN_LOCAL)
35 			fl.nl_u.ip4_u.saddr = iph->saddr;
36 		fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
37 		fl.oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
38 		fl.mark = skb->mark;
39 		fl.flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
40 		if (ip_route_output_key(net, &rt, &fl) != 0)
41 			return -1;
42 
43 		/* Drop old route. */
44 		dst_release(skb->dst);
45 		skb->dst = &rt->u.dst;
46 	} else {
47 		/* non-local src, find valid iif to satisfy
48 		 * rp-filter when calling ip_route_input. */
49 		fl.nl_u.ip4_u.daddr = iph->saddr;
50 		if (ip_route_output_key(net, &rt, &fl) != 0)
51 			return -1;
52 
53 		odst = skb->dst;
54 		if (ip_route_input(skb, iph->daddr, iph->saddr,
55 				   RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
56 			dst_release(&rt->u.dst);
57 			return -1;
58 		}
59 		dst_release(&rt->u.dst);
60 		dst_release(odst);
61 	}
62 
63 	if (skb->dst->error)
64 		return -1;
65 
66 #ifdef CONFIG_XFRM
67 	if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
68 	    xfrm_decode_session(skb, &fl, AF_INET) == 0)
69 		if (xfrm_lookup(net, &skb->dst, &fl, skb->sk, 0))
70 			return -1;
71 #endif
72 
73 	/* Change in oif may mean change in hh_len. */
74 	hh_len = skb->dst->dev->hard_header_len;
75 	if (skb_headroom(skb) < hh_len &&
76 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
77 		return -1;
78 
79 	return 0;
80 }
81 EXPORT_SYMBOL(ip_route_me_harder);
82 
83 #ifdef CONFIG_XFRM
84 int ip_xfrm_me_harder(struct sk_buff *skb)
85 {
86 	struct flowi fl;
87 	unsigned int hh_len;
88 	struct dst_entry *dst;
89 
90 	if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
91 		return 0;
92 	if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
93 		return -1;
94 
95 	dst = skb->dst;
96 	if (dst->xfrm)
97 		dst = ((struct xfrm_dst *)dst)->route;
98 	dst_hold(dst);
99 
100 	if (xfrm_lookup(dev_net(dst->dev), &dst, &fl, skb->sk, 0) < 0)
101 		return -1;
102 
103 	dst_release(skb->dst);
104 	skb->dst = dst;
105 
106 	/* Change in oif may mean change in hh_len. */
107 	hh_len = skb->dst->dev->hard_header_len;
108 	if (skb_headroom(skb) < hh_len &&
109 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
110 		return -1;
111 	return 0;
112 }
113 EXPORT_SYMBOL(ip_xfrm_me_harder);
114 #endif
115 
116 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
117 EXPORT_SYMBOL(ip_nat_decode_session);
118 
119 /*
120  * Extra routing may needed on local out, as the QUEUE target never
121  * returns control to the table.
122  */
123 
124 struct ip_rt_info {
125 	__be32 daddr;
126 	__be32 saddr;
127 	u_int8_t tos;
128 	u_int32_t mark;
129 };
130 
131 static void nf_ip_saveroute(const struct sk_buff *skb,
132 			    struct nf_queue_entry *entry)
133 {
134 	struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
135 
136 	if (entry->hook == NF_INET_LOCAL_OUT) {
137 		const struct iphdr *iph = ip_hdr(skb);
138 
139 		rt_info->tos = iph->tos;
140 		rt_info->daddr = iph->daddr;
141 		rt_info->saddr = iph->saddr;
142 		rt_info->mark = skb->mark;
143 	}
144 }
145 
146 static int nf_ip_reroute(struct sk_buff *skb,
147 			 const struct nf_queue_entry *entry)
148 {
149 	const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
150 
151 	if (entry->hook == NF_INET_LOCAL_OUT) {
152 		const struct iphdr *iph = ip_hdr(skb);
153 
154 		if (!(iph->tos == rt_info->tos
155 		      && skb->mark == rt_info->mark
156 		      && iph->daddr == rt_info->daddr
157 		      && iph->saddr == rt_info->saddr))
158 			return ip_route_me_harder(skb, RTN_UNSPEC);
159 	}
160 	return 0;
161 }
162 
163 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
164 			    unsigned int dataoff, u_int8_t protocol)
165 {
166 	const struct iphdr *iph = ip_hdr(skb);
167 	__sum16 csum = 0;
168 
169 	switch (skb->ip_summed) {
170 	case CHECKSUM_COMPLETE:
171 		if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
172 			break;
173 		if ((protocol == 0 && !csum_fold(skb->csum)) ||
174 		    !csum_tcpudp_magic(iph->saddr, iph->daddr,
175 				       skb->len - dataoff, protocol,
176 				       skb->csum)) {
177 			skb->ip_summed = CHECKSUM_UNNECESSARY;
178 			break;
179 		}
180 		/* fall through */
181 	case CHECKSUM_NONE:
182 		if (protocol == 0)
183 			skb->csum = 0;
184 		else
185 			skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
186 						       skb->len - dataoff,
187 						       protocol, 0);
188 		csum = __skb_checksum_complete(skb);
189 	}
190 	return csum;
191 }
192 EXPORT_SYMBOL(nf_ip_checksum);
193 
194 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
195 				      unsigned int dataoff, unsigned int len,
196 				      u_int8_t protocol)
197 {
198 	const struct iphdr *iph = ip_hdr(skb);
199 	__sum16 csum = 0;
200 
201 	switch (skb->ip_summed) {
202 	case CHECKSUM_COMPLETE:
203 		if (len == skb->len - dataoff)
204 			return nf_ip_checksum(skb, hook, dataoff, protocol);
205 		/* fall through */
206 	case CHECKSUM_NONE:
207 		skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
208 					       skb->len - dataoff, 0);
209 		skb->ip_summed = CHECKSUM_NONE;
210 		csum = __skb_checksum_complete_head(skb, dataoff + len);
211 		if (!csum)
212 			skb->ip_summed = CHECKSUM_UNNECESSARY;
213 	}
214 	return csum;
215 }
216 
217 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
218 {
219 	return ip_route_output_key(&init_net, (struct rtable **)dst, fl);
220 }
221 
222 static const struct nf_afinfo nf_ip_afinfo = {
223 	.family			= AF_INET,
224 	.checksum		= nf_ip_checksum,
225 	.checksum_partial	= nf_ip_checksum_partial,
226 	.route			= nf_ip_route,
227 	.saveroute		= nf_ip_saveroute,
228 	.reroute		= nf_ip_reroute,
229 	.route_key_size		= sizeof(struct ip_rt_info),
230 };
231 
232 static int ipv4_netfilter_init(void)
233 {
234 	return nf_register_afinfo(&nf_ip_afinfo);
235 }
236 
237 static void ipv4_netfilter_fini(void)
238 {
239 	nf_unregister_afinfo(&nf_ip_afinfo);
240 }
241 
242 module_init(ipv4_netfilter_init);
243 module_exit(ipv4_netfilter_fini);
244 
245 #ifdef CONFIG_SYSCTL
246 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
247 	{ .procname = "net", .ctl_name = CTL_NET, },
248 	{ .procname = "ipv4", .ctl_name = NET_IPV4, },
249 	{ .procname = "netfilter", .ctl_name = NET_IPV4_NETFILTER, },
250 	{ }
251 };
252 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
253 #endif /* CONFIG_SYSCTL */
254