Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next-2.6
[linux-2.6.git] / 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         struct net *net = dev_net(skb_dst(skb)->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                 skb_dst_drop(skb);
45                 skb_dst_set(skb, &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(skb);
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(skb)->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                 struct dst_entry *dst = skb_dst(skb);
70                 skb_dst_set(skb, NULL);
71                 if (xfrm_lookup(net, &dst, &fl, skb->sk, 0))
72                         return -1;
73                 skb_dst_set(skb, dst);
74         }
75 #endif
76
77         /* Change in oif may mean change in hh_len. */
78         hh_len = skb_dst(skb)->dev->hard_header_len;
79         if (skb_headroom(skb) < hh_len &&
80             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
81                 return -1;
82
83         return 0;
84 }
85 EXPORT_SYMBOL(ip_route_me_harder);
86
87 #ifdef CONFIG_XFRM
88 int ip_xfrm_me_harder(struct sk_buff *skb)
89 {
90         struct flowi fl;
91         unsigned int hh_len;
92         struct dst_entry *dst;
93
94         if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
95                 return 0;
96         if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
97                 return -1;
98
99         dst = skb_dst(skb);
100         if (dst->xfrm)
101                 dst = ((struct xfrm_dst *)dst)->route;
102         dst_hold(dst);
103
104         if (xfrm_lookup(dev_net(dst->dev), &dst, &fl, skb->sk, 0) < 0)
105                 return -1;
106
107         skb_dst_drop(skb);
108         skb_dst_set(skb, dst);
109
110         /* Change in oif may mean change in hh_len. */
111         hh_len = skb_dst(skb)->dev->hard_header_len;
112         if (skb_headroom(skb) < hh_len &&
113             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
114                 return -1;
115         return 0;
116 }
117 EXPORT_SYMBOL(ip_xfrm_me_harder);
118 #endif
119
120 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
121 EXPORT_SYMBOL(ip_nat_decode_session);
122
123 /*
124  * Extra routing may needed on local out, as the QUEUE target never
125  * returns control to the table.
126  */
127
128 struct ip_rt_info {
129         __be32 daddr;
130         __be32 saddr;
131         u_int8_t tos;
132         u_int32_t mark;
133 };
134
135 static void nf_ip_saveroute(const struct sk_buff *skb,
136                             struct nf_queue_entry *entry)
137 {
138         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
139
140         if (entry->hook == NF_INET_LOCAL_OUT) {
141                 const struct iphdr *iph = ip_hdr(skb);
142
143                 rt_info->tos = iph->tos;
144                 rt_info->daddr = iph->daddr;
145                 rt_info->saddr = iph->saddr;
146                 rt_info->mark = skb->mark;
147         }
148 }
149
150 static int nf_ip_reroute(struct sk_buff *skb,
151                          const struct nf_queue_entry *entry)
152 {
153         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
154
155         if (entry->hook == NF_INET_LOCAL_OUT) {
156                 const struct iphdr *iph = ip_hdr(skb);
157
158                 if (!(iph->tos == rt_info->tos &&
159                       skb->mark == rt_info->mark &&
160                       iph->daddr == rt_info->daddr &&
161                       iph->saddr == rt_info->saddr))
162                         return ip_route_me_harder(skb, RTN_UNSPEC);
163         }
164         return 0;
165 }
166
167 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
168                             unsigned int dataoff, u_int8_t protocol)
169 {
170         const struct iphdr *iph = ip_hdr(skb);
171         __sum16 csum = 0;
172
173         switch (skb->ip_summed) {
174         case CHECKSUM_COMPLETE:
175                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
176                         break;
177                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
178                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
179                                        skb->len - dataoff, protocol,
180                                        skb->csum)) {
181                         skb->ip_summed = CHECKSUM_UNNECESSARY;
182                         break;
183                 }
184                 /* fall through */
185         case CHECKSUM_NONE:
186                 if (protocol == 0)
187                         skb->csum = 0;
188                 else
189                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
190                                                        skb->len - dataoff,
191                                                        protocol, 0);
192                 csum = __skb_checksum_complete(skb);
193         }
194         return csum;
195 }
196 EXPORT_SYMBOL(nf_ip_checksum);
197
198 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
199                                       unsigned int dataoff, unsigned int len,
200                                       u_int8_t protocol)
201 {
202         const struct iphdr *iph = ip_hdr(skb);
203         __sum16 csum = 0;
204
205         switch (skb->ip_summed) {
206         case CHECKSUM_COMPLETE:
207                 if (len == skb->len - dataoff)
208                         return nf_ip_checksum(skb, hook, dataoff, protocol);
209                 /* fall through */
210         case CHECKSUM_NONE:
211                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
212                                                skb->len - dataoff, 0);
213                 skb->ip_summed = CHECKSUM_NONE;
214                 csum = __skb_checksum_complete_head(skb, dataoff + len);
215                 if (!csum)
216                         skb->ip_summed = CHECKSUM_UNNECESSARY;
217         }
218         return csum;
219 }
220
221 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
222 {
223         return ip_route_output_key(&init_net, (struct rtable **)dst, fl);
224 }
225
226 static const struct nf_afinfo nf_ip_afinfo = {
227         .family                 = AF_INET,
228         .checksum               = nf_ip_checksum,
229         .checksum_partial       = nf_ip_checksum_partial,
230         .route                  = nf_ip_route,
231         .saveroute              = nf_ip_saveroute,
232         .reroute                = nf_ip_reroute,
233         .route_key_size         = sizeof(struct ip_rt_info),
234 };
235
236 static int ipv4_netfilter_init(void)
237 {
238         return nf_register_afinfo(&nf_ip_afinfo);
239 }
240
241 static void ipv4_netfilter_fini(void)
242 {
243         nf_unregister_afinfo(&nf_ip_afinfo);
244 }
245
246 module_init(ipv4_netfilter_init);
247 module_exit(ipv4_netfilter_fini);
248
249 #ifdef CONFIG_SYSCTL
250 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
251         { .procname = "net", },
252         { .procname = "ipv4", },
253         { .procname = "netfilter", },
254         { }
255 };
256 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
257 #endif /* CONFIG_SYSCTL */