netfilter: netns: fix {ip,6}_route_me_harder() in netns
[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->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(&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(&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 };
129
130 static void nf_ip_saveroute(const struct sk_buff *skb,
131                             struct nf_queue_entry *entry)
132 {
133         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
134
135         if (entry->hook == NF_INET_LOCAL_OUT) {
136                 const struct iphdr *iph = ip_hdr(skb);
137
138                 rt_info->tos = iph->tos;
139                 rt_info->daddr = iph->daddr;
140                 rt_info->saddr = iph->saddr;
141         }
142 }
143
144 static int nf_ip_reroute(struct sk_buff *skb,
145                          const struct nf_queue_entry *entry)
146 {
147         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
148
149         if (entry->hook == NF_INET_LOCAL_OUT) {
150                 const struct iphdr *iph = ip_hdr(skb);
151
152                 if (!(iph->tos == rt_info->tos
153                       && iph->daddr == rt_info->daddr
154                       && iph->saddr == rt_info->saddr))
155                         return ip_route_me_harder(skb, RTN_UNSPEC);
156         }
157         return 0;
158 }
159
160 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
161                             unsigned int dataoff, u_int8_t protocol)
162 {
163         const struct iphdr *iph = ip_hdr(skb);
164         __sum16 csum = 0;
165
166         switch (skb->ip_summed) {
167         case CHECKSUM_COMPLETE:
168                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
169                         break;
170                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
171                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
172                                        skb->len - dataoff, protocol,
173                                        skb->csum)) {
174                         skb->ip_summed = CHECKSUM_UNNECESSARY;
175                         break;
176                 }
177                 /* fall through */
178         case CHECKSUM_NONE:
179                 if (protocol == 0)
180                         skb->csum = 0;
181                 else
182                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
183                                                        skb->len - dataoff,
184                                                        protocol, 0);
185                 csum = __skb_checksum_complete(skb);
186         }
187         return csum;
188 }
189 EXPORT_SYMBOL(nf_ip_checksum);
190
191 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
192                                       unsigned int dataoff, unsigned int len,
193                                       u_int8_t protocol)
194 {
195         const struct iphdr *iph = ip_hdr(skb);
196         __sum16 csum = 0;
197
198         switch (skb->ip_summed) {
199         case CHECKSUM_COMPLETE:
200                 if (len == skb->len - dataoff)
201                         return nf_ip_checksum(skb, hook, dataoff, protocol);
202                 /* fall through */
203         case CHECKSUM_NONE:
204                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
205                                                skb->len - dataoff, 0);
206                 skb->ip_summed = CHECKSUM_NONE;
207                 csum = __skb_checksum_complete_head(skb, dataoff + len);
208                 if (!csum)
209                         skb->ip_summed = CHECKSUM_UNNECESSARY;
210         }
211         return csum;
212 }
213
214 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
215 {
216         return ip_route_output_key(&init_net, (struct rtable **)dst, fl);
217 }
218
219 static const struct nf_afinfo nf_ip_afinfo = {
220         .family                 = AF_INET,
221         .checksum               = nf_ip_checksum,
222         .checksum_partial       = nf_ip_checksum_partial,
223         .route                  = nf_ip_route,
224         .saveroute              = nf_ip_saveroute,
225         .reroute                = nf_ip_reroute,
226         .route_key_size         = sizeof(struct ip_rt_info),
227 };
228
229 static int ipv4_netfilter_init(void)
230 {
231         return nf_register_afinfo(&nf_ip_afinfo);
232 }
233
234 static void ipv4_netfilter_fini(void)
235 {
236         nf_unregister_afinfo(&nf_ip_afinfo);
237 }
238
239 module_init(ipv4_netfilter_init);
240 module_exit(ipv4_netfilter_fini);
241
242 #ifdef CONFIG_SYSCTL
243 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
244         { .procname = "net", .ctl_name = CTL_NET, },
245         { .procname = "ipv4", .ctl_name = NET_IPV4, },
246         { .procname = "netfilter", .ctl_name = NET_IPV4_NETFILTER, },
247         { }
248 };
249 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
250 #endif /* CONFIG_SYSCTL */