xref: /openbmc/linux/net/ipv4/netfilter.c (revision 37cf4d1a)
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 = 0;
22 	unsigned int hh_len;
23 
24 	if (!skb->sk && addr_type != RTN_LOCAL) {
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 
33 	/* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
34 	 * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
35 	 */
36 	fl4.daddr = iph->daddr;
37 	fl4.saddr = saddr;
38 	fl4.flowi4_tos = RT_TOS(iph->tos);
39 	fl4.flowi4_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
40 	fl4.flowi4_mark = skb->mark;
41 	fl4.flowi4_flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : flags;
42 	rt = ip_route_output_key(net, &fl4);
43 	if (IS_ERR(rt))
44 		return -1;
45 
46 	/* Drop old route. */
47 	skb_dst_drop(skb);
48 	skb_dst_set(skb, &rt->dst);
49 
50 	if (skb_dst(skb)->error)
51 		return -1;
52 
53 #ifdef CONFIG_XFRM
54 	if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
55 	    xfrm_decode_session(skb, flowi4_to_flowi(&fl4), AF_INET) == 0) {
56 		struct dst_entry *dst = skb_dst(skb);
57 		skb_dst_set(skb, NULL);
58 		dst = xfrm_lookup(net, dst, flowi4_to_flowi(&fl4), skb->sk, 0);
59 		if (IS_ERR(dst))
60 			return -1;
61 		skb_dst_set(skb, dst);
62 	}
63 #endif
64 
65 	/* Change in oif may mean change in hh_len. */
66 	hh_len = skb_dst(skb)->dev->hard_header_len;
67 	if (skb_headroom(skb) < hh_len &&
68 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
69 		return -1;
70 
71 	return 0;
72 }
73 EXPORT_SYMBOL(ip_route_me_harder);
74 
75 #ifdef CONFIG_XFRM
76 int ip_xfrm_me_harder(struct sk_buff *skb)
77 {
78 	struct flowi fl;
79 	unsigned int hh_len;
80 	struct dst_entry *dst;
81 
82 	if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
83 		return 0;
84 	if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
85 		return -1;
86 
87 	dst = skb_dst(skb);
88 	if (dst->xfrm)
89 		dst = ((struct xfrm_dst *)dst)->route;
90 	dst_hold(dst);
91 
92 	dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0);
93 	if (IS_ERR(dst))
94 		return -1;
95 
96 	skb_dst_drop(skb);
97 	skb_dst_set(skb, dst);
98 
99 	/* Change in oif may mean change in hh_len. */
100 	hh_len = skb_dst(skb)->dev->hard_header_len;
101 	if (skb_headroom(skb) < hh_len &&
102 	    pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
103 		return -1;
104 	return 0;
105 }
106 EXPORT_SYMBOL(ip_xfrm_me_harder);
107 #endif
108 
109 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
110 EXPORT_SYMBOL(ip_nat_decode_session);
111 
112 /*
113  * Extra routing may needed on local out, as the QUEUE target never
114  * returns control to the table.
115  */
116 
117 struct ip_rt_info {
118 	__be32 daddr;
119 	__be32 saddr;
120 	u_int8_t tos;
121 	u_int32_t mark;
122 };
123 
124 static void nf_ip_saveroute(const struct sk_buff *skb,
125 			    struct nf_queue_entry *entry)
126 {
127 	struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
128 
129 	if (entry->hook == NF_INET_LOCAL_OUT) {
130 		const struct iphdr *iph = ip_hdr(skb);
131 
132 		rt_info->tos = iph->tos;
133 		rt_info->daddr = iph->daddr;
134 		rt_info->saddr = iph->saddr;
135 		rt_info->mark = skb->mark;
136 	}
137 }
138 
139 static int nf_ip_reroute(struct sk_buff *skb,
140 			 const struct nf_queue_entry *entry)
141 {
142 	const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
143 
144 	if (entry->hook == NF_INET_LOCAL_OUT) {
145 		const struct iphdr *iph = ip_hdr(skb);
146 
147 		if (!(iph->tos == rt_info->tos &&
148 		      skb->mark == rt_info->mark &&
149 		      iph->daddr == rt_info->daddr &&
150 		      iph->saddr == rt_info->saddr))
151 			return ip_route_me_harder(skb, RTN_UNSPEC);
152 	}
153 	return 0;
154 }
155 
156 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
157 			    unsigned int dataoff, u_int8_t protocol)
158 {
159 	const struct iphdr *iph = ip_hdr(skb);
160 	__sum16 csum = 0;
161 
162 	switch (skb->ip_summed) {
163 	case CHECKSUM_COMPLETE:
164 		if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
165 			break;
166 		if ((protocol == 0 && !csum_fold(skb->csum)) ||
167 		    !csum_tcpudp_magic(iph->saddr, iph->daddr,
168 				       skb->len - dataoff, protocol,
169 				       skb->csum)) {
170 			skb->ip_summed = CHECKSUM_UNNECESSARY;
171 			break;
172 		}
173 		/* fall through */
174 	case CHECKSUM_NONE:
175 		if (protocol == 0)
176 			skb->csum = 0;
177 		else
178 			skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
179 						       skb->len - dataoff,
180 						       protocol, 0);
181 		csum = __skb_checksum_complete(skb);
182 	}
183 	return csum;
184 }
185 EXPORT_SYMBOL(nf_ip_checksum);
186 
187 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
188 				      unsigned int dataoff, unsigned int len,
189 				      u_int8_t protocol)
190 {
191 	const struct iphdr *iph = ip_hdr(skb);
192 	__sum16 csum = 0;
193 
194 	switch (skb->ip_summed) {
195 	case CHECKSUM_COMPLETE:
196 		if (len == skb->len - dataoff)
197 			return nf_ip_checksum(skb, hook, dataoff, protocol);
198 		/* fall through */
199 	case CHECKSUM_NONE:
200 		skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
201 					       skb->len - dataoff, 0);
202 		skb->ip_summed = CHECKSUM_NONE;
203 		return __skb_checksum_complete_head(skb, dataoff + len);
204 	}
205 	return csum;
206 }
207 
208 static int nf_ip_route(struct net *net, struct dst_entry **dst,
209 		       struct flowi *fl, bool strict __always_unused)
210 {
211 	struct rtable *rt = ip_route_output_key(net, &fl->u.ip4);
212 	if (IS_ERR(rt))
213 		return PTR_ERR(rt);
214 	*dst = &rt->dst;
215 	return 0;
216 }
217 
218 static const struct nf_afinfo nf_ip_afinfo = {
219 	.family			= AF_INET,
220 	.checksum		= nf_ip_checksum,
221 	.checksum_partial	= nf_ip_checksum_partial,
222 	.route			= nf_ip_route,
223 	.saveroute		= nf_ip_saveroute,
224 	.reroute		= nf_ip_reroute,
225 	.route_key_size		= sizeof(struct ip_rt_info),
226 };
227 
228 static int ipv4_netfilter_init(void)
229 {
230 	return nf_register_afinfo(&nf_ip_afinfo);
231 }
232 
233 static void ipv4_netfilter_fini(void)
234 {
235 	nf_unregister_afinfo(&nf_ip_afinfo);
236 }
237 
238 module_init(ipv4_netfilter_init);
239 module_exit(ipv4_netfilter_fini);
240 
241 #ifdef CONFIG_SYSCTL
242 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
243 	{ .procname = "net", },
244 	{ .procname = "ipv4", },
245 	{ .procname = "netfilter", },
246 	{ }
247 };
248 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
249 #endif /* CONFIG_SYSCTL */
250