Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6 into lvs-next-2.6
[linux-2.6] / net / ipv4 / netfilter.c
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         const struct iphdr *iph = ip_hdr(skb);
16         struct rtable *rt;
17         struct flowi fl = {};
18         struct dst_entry *odst;
19         unsigned int hh_len;
20         unsigned int type;
21
22         type = inet_addr_type(&init_net, iph->saddr);
23         if (skb->sk && inet_sk(skb->sk)->transparent)
24                 type = RTN_LOCAL;
25         if (addr_type == RTN_UNSPEC)
26                 addr_type = type;
27
28         /* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
29          * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
30          */
31         if (addr_type == RTN_LOCAL) {
32                 fl.nl_u.ip4_u.daddr = iph->daddr;
33                 if (type == RTN_LOCAL)
34                         fl.nl_u.ip4_u.saddr = iph->saddr;
35                 fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
36                 fl.oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
37                 fl.mark = skb->mark;
38                 fl.flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
39                 if (ip_route_output_key(&init_net, &rt, &fl) != 0)
40                         return -1;
41
42                 /* Drop old route. */
43                 dst_release(skb->dst);
44                 skb->dst = &rt->u.dst;
45         } else {
46                 /* non-local src, find valid iif to satisfy
47                  * rp-filter when calling ip_route_input. */
48                 fl.nl_u.ip4_u.daddr = iph->saddr;
49                 if (ip_route_output_key(&init_net, &rt, &fl) != 0)
50                         return -1;
51
52                 odst = skb->dst;
53                 if (ip_route_input(skb, iph->daddr, iph->saddr,
54                                    RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
55                         dst_release(&rt->u.dst);
56                         return -1;
57                 }
58                 dst_release(&rt->u.dst);
59                 dst_release(odst);
60         }
61
62         if (skb->dst->error)
63                 return -1;
64
65 #ifdef CONFIG_XFRM
66         if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
67             xfrm_decode_session(skb, &fl, AF_INET) == 0)
68                 if (xfrm_lookup(&skb->dst, &fl, skb->sk, 0))
69                         return -1;
70 #endif
71
72         /* Change in oif may mean change in hh_len. */
73         hh_len = skb->dst->dev->hard_header_len;
74         if (skb_headroom(skb) < hh_len &&
75             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
76                 return -1;
77
78         return 0;
79 }
80 EXPORT_SYMBOL(ip_route_me_harder);
81
82 #ifdef CONFIG_XFRM
83 int ip_xfrm_me_harder(struct sk_buff *skb)
84 {
85         struct flowi fl;
86         unsigned int hh_len;
87         struct dst_entry *dst;
88
89         if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
90                 return 0;
91         if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
92                 return -1;
93
94         dst = skb->dst;
95         if (dst->xfrm)
96                 dst = ((struct xfrm_dst *)dst)->route;
97         dst_hold(dst);
98
99         if (xfrm_lookup(&dst, &fl, skb->sk, 0) < 0)
100                 return -1;
101
102         dst_release(skb->dst);
103         skb->dst = dst;
104
105         /* Change in oif may mean change in hh_len. */
106         hh_len = skb->dst->dev->hard_header_len;
107         if (skb_headroom(skb) < hh_len &&
108             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
109                 return -1;
110         return 0;
111 }
112 EXPORT_SYMBOL(ip_xfrm_me_harder);
113 #endif
114
115 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
116 EXPORT_SYMBOL(ip_nat_decode_session);
117
118 /*
119  * Extra routing may needed on local out, as the QUEUE target never
120  * returns control to the table.
121  */
122
123 struct ip_rt_info {
124         __be32 daddr;
125         __be32 saddr;
126         u_int8_t tos;
127 };
128
129 static void nf_ip_saveroute(const struct sk_buff *skb,
130                             struct nf_queue_entry *entry)
131 {
132         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
133
134         if (entry->hook == NF_INET_LOCAL_OUT) {
135                 const struct iphdr *iph = ip_hdr(skb);
136
137                 rt_info->tos = iph->tos;
138                 rt_info->daddr = iph->daddr;
139                 rt_info->saddr = iph->saddr;
140         }
141 }
142
143 static int nf_ip_reroute(struct sk_buff *skb,
144                          const struct nf_queue_entry *entry)
145 {
146         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
147
148         if (entry->hook == NF_INET_LOCAL_OUT) {
149                 const struct iphdr *iph = ip_hdr(skb);
150
151                 if (!(iph->tos == rt_info->tos
152                       && iph->daddr == rt_info->daddr
153                       && iph->saddr == rt_info->saddr))
154                         return ip_route_me_harder(skb, RTN_UNSPEC);
155         }
156         return 0;
157 }
158
159 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
160                             unsigned int dataoff, u_int8_t protocol)
161 {
162         const struct iphdr *iph = ip_hdr(skb);
163         __sum16 csum = 0;
164
165         switch (skb->ip_summed) {
166         case CHECKSUM_COMPLETE:
167                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
168                         break;
169                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
170                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
171                                        skb->len - dataoff, protocol,
172                                        skb->csum)) {
173                         skb->ip_summed = CHECKSUM_UNNECESSARY;
174                         break;
175                 }
176                 /* fall through */
177         case CHECKSUM_NONE:
178                 if (protocol == 0)
179                         skb->csum = 0;
180                 else
181                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
182                                                        skb->len - dataoff,
183                                                        protocol, 0);
184                 csum = __skb_checksum_complete(skb);
185         }
186         return csum;
187 }
188 EXPORT_SYMBOL(nf_ip_checksum);
189
190 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
191                                       unsigned int dataoff, unsigned int len,
192                                       u_int8_t protocol)
193 {
194         const struct iphdr *iph = ip_hdr(skb);
195         __sum16 csum = 0;
196
197         switch (skb->ip_summed) {
198         case CHECKSUM_COMPLETE:
199                 if (len == skb->len - dataoff)
200                         return nf_ip_checksum(skb, hook, dataoff, protocol);
201                 /* fall through */
202         case CHECKSUM_NONE:
203                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
204                                                skb->len - dataoff, 0);
205                 skb->ip_summed = CHECKSUM_NONE;
206                 csum = __skb_checksum_complete_head(skb, dataoff + len);
207                 if (!csum)
208                         skb->ip_summed = CHECKSUM_UNNECESSARY;
209         }
210         return csum;
211 }
212
213 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
214 {
215         return ip_route_output_key(&init_net, (struct rtable **)dst, fl);
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", .ctl_name = CTL_NET, },
244         { .procname = "ipv4", .ctl_name = NET_IPV4, },
245         { .procname = "netfilter", .ctl_name = NET_IPV4_NETFILTER, },
246         { }
247 };
248 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
249 #endif /* CONFIG_SYSCTL */