xref: /linux/net/ipv4/netfilter.c (revision 60e13231561b3a4c5269bfa1ef6c0569ad6f28ec)
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 <net/route.h>
9 #include <net/xfrm.h>
10 #include <net/ip.h>
11 #include <net/netfilter/nf_queue.h>
12 
13 /* route_me_harder function, used by iptable_nat, iptable_mangle + ip_queue */
14 int ip_route_me_harder(struct sk_buff *skb, unsigned addr_type)
15 {
16 	struct net *net = dev_net(skb_dst(skb)->dev);
17 	const struct iphdr *iph = ip_hdr(skb);
18 	struct rtable *rt;
19 	struct flowi4 fl4 = {};
20 	__be32 saddr = iph->saddr;
21 	__u8 flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
22 	unsigned int hh_len;
23 
24 	if (addr_type == RTN_UNSPEC)
25 		addr_type = inet_addr_type(net, saddr);
26 	if (addr_type == RTN_LOCAL || addr_type == RTN_UNICAST)
27 		flags |= FLOWI_FLAG_ANYSRC;
28 	else
29 		saddr = 0;
30 
31 	/* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
32 	 * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
33 	 */
34 	fl4.daddr = iph->daddr;
35 	fl4.saddr = saddr;
36 	fl4.flowi4_tos = RT_TOS(iph->tos);
37 	fl4.flowi4_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
38 	fl4.flowi4_mark = skb->mark;
39 	fl4.flowi4_flags = flags;
40 	rt = ip_route_output_key(net, &fl4);
41 	if (IS_ERR(rt))
42 		return -1;
43 
44 	/* Drop old route. */
45 	skb_dst_drop(skb);
46 	skb_dst_set(skb, &rt->dst);
47 
48 	if (skb_dst(skb)->error)
49 		return -1;
50 
51 #ifdef CONFIG_XFRM
52 	if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
53 	    xfrm_decode_session(skb, flowi4_to_flowi(&fl4), AF_INET) == 0) {
54 		struct dst_entry *dst = skb_dst(skb);
55 		skb_dst_set(skb, NULL);
56 		dst = xfrm_lookup(net, dst, flowi4_to_flowi(&fl4), skb->sk, 0);
57 		if (IS_ERR(dst))
58 			return -1;
59 		skb_dst_set(skb, dst);
60 	}
61 #endif
62 
63 	/* Change in oif may mean change in hh_len. */
64 	hh_len = skb_dst(skb)->dev->hard_header_len;
65 	if (skb_headroom(skb) < hh_len &&
66 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
67 		return -1;
68 
69 	return 0;
70 }
71 EXPORT_SYMBOL(ip_route_me_harder);
72 
73 #ifdef CONFIG_XFRM
74 int ip_xfrm_me_harder(struct sk_buff *skb)
75 {
76 	struct flowi fl;
77 	unsigned int hh_len;
78 	struct dst_entry *dst;
79 
80 	if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
81 		return 0;
82 	if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
83 		return -1;
84 
85 	dst = skb_dst(skb);
86 	if (dst->xfrm)
87 		dst = ((struct xfrm_dst *)dst)->route;
88 	dst_hold(dst);
89 
90 	dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0);
91 	if (IS_ERR(dst))
92 		return -1;
93 
94 	skb_dst_drop(skb);
95 	skb_dst_set(skb, dst);
96 
97 	/* Change in oif may mean change in hh_len. */
98 	hh_len = skb_dst(skb)->dev->hard_header_len;
99 	if (skb_headroom(skb) < hh_len &&
100 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
101 		return -1;
102 	return 0;
103 }
104 EXPORT_SYMBOL(ip_xfrm_me_harder);
105 #endif
106 
107 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
108 EXPORT_SYMBOL(ip_nat_decode_session);
109 
110 /*
111  * Extra routing may needed on local out, as the QUEUE target never
112  * returns control to the table.
113  */
114 
115 struct ip_rt_info {
116 	__be32 daddr;
117 	__be32 saddr;
118 	u_int8_t tos;
119 	u_int32_t mark;
120 };
121 
122 static void nf_ip_saveroute(const struct sk_buff *skb,
123 			    struct nf_queue_entry *entry)
124 {
125 	struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
126 
127 	if (entry->hook == NF_INET_LOCAL_OUT) {
128 		const struct iphdr *iph = ip_hdr(skb);
129 
130 		rt_info->tos = iph->tos;
131 		rt_info->daddr = iph->daddr;
132 		rt_info->saddr = iph->saddr;
133 		rt_info->mark = skb->mark;
134 	}
135 }
136 
137 static int nf_ip_reroute(struct sk_buff *skb,
138 			 const struct nf_queue_entry *entry)
139 {
140 	const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
141 
142 	if (entry->hook == NF_INET_LOCAL_OUT) {
143 		const struct iphdr *iph = ip_hdr(skb);
144 
145 		if (!(iph->tos == rt_info->tos &&
146 		      skb->mark == rt_info->mark &&
147 		      iph->daddr == rt_info->daddr &&
148 		      iph->saddr == rt_info->saddr))
149 			return ip_route_me_harder(skb, RTN_UNSPEC);
150 	}
151 	return 0;
152 }
153 
154 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
155 			    unsigned int dataoff, u_int8_t protocol)
156 {
157 	const struct iphdr *iph = ip_hdr(skb);
158 	__sum16 csum = 0;
159 
160 	switch (skb->ip_summed) {
161 	case CHECKSUM_COMPLETE:
162 		if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
163 			break;
164 		if ((protocol == 0 && !csum_fold(skb->csum)) ||
165 		    !csum_tcpudp_magic(iph->saddr, iph->daddr,
166 				       skb->len - dataoff, protocol,
167 				       skb->csum)) {
168 			skb->ip_summed = CHECKSUM_UNNECESSARY;
169 			break;
170 		}
171 		/* fall through */
172 	case CHECKSUM_NONE:
173 		if (protocol == 0)
174 			skb->csum = 0;
175 		else
176 			skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
177 						       skb->len - dataoff,
178 						       protocol, 0);
179 		csum = __skb_checksum_complete(skb);
180 	}
181 	return csum;
182 }
183 EXPORT_SYMBOL(nf_ip_checksum);
184 
185 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
186 				      unsigned int dataoff, unsigned int len,
187 				      u_int8_t protocol)
188 {
189 	const struct iphdr *iph = ip_hdr(skb);
190 	__sum16 csum = 0;
191 
192 	switch (skb->ip_summed) {
193 	case CHECKSUM_COMPLETE:
194 		if (len == skb->len - dataoff)
195 			return nf_ip_checksum(skb, hook, dataoff, protocol);
196 		/* fall through */
197 	case CHECKSUM_NONE:
198 		skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
199 					       skb->len - dataoff, 0);
200 		skb->ip_summed = CHECKSUM_NONE;
201 		return __skb_checksum_complete_head(skb, dataoff + len);
202 	}
203 	return csum;
204 }
205 
206 static int nf_ip_route(struct net *net, struct dst_entry **dst,
207 		       struct flowi *fl, bool strict __always_unused)
208 {
209 	struct rtable *rt = ip_route_output_key(net, &fl->u.ip4);
210 	if (IS_ERR(rt))
211 		return PTR_ERR(rt);
212 	*dst = &rt->dst;
213 	return 0;
214 }
215 
216 static const struct nf_afinfo nf_ip_afinfo = {
217 	.family			= AF_INET,
218 	.checksum		= nf_ip_checksum,
219 	.checksum_partial	= nf_ip_checksum_partial,
220 	.route			= nf_ip_route,
221 	.saveroute		= nf_ip_saveroute,
222 	.reroute		= nf_ip_reroute,
223 	.route_key_size		= sizeof(struct ip_rt_info),
224 };
225 
226 static int ipv4_netfilter_init(void)
227 {
228 	return nf_register_afinfo(&nf_ip_afinfo);
229 }
230 
231 static void ipv4_netfilter_fini(void)
232 {
233 	nf_unregister_afinfo(&nf_ip_afinfo);
234 }
235 
236 module_init(ipv4_netfilter_init);
237 module_exit(ipv4_netfilter_fini);
238 
239 #ifdef CONFIG_SYSCTL
240 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
241 	{ .procname = "net", },
242 	{ .procname = "ipv4", },
243 	{ .procname = "netfilter", },
244 	{ }
245 };
246 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
247 #endif /* CONFIG_SYSCTL */
248