xfrm: Return dst directly from xfrm_lookup()
[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 <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 flowi fl = {};
20         unsigned long orefdst;
21         unsigned int hh_len;
22         unsigned int type;
23
24         type = inet_addr_type(net, iph->saddr);
25         if (skb->sk && inet_sk(skb->sk)->transparent)
26                 type = RTN_LOCAL;
27         if (addr_type == RTN_UNSPEC)
28                 addr_type = type;
29
30         /* some non-standard hacks like ipt_REJECT.c:send_reset() can cause
31          * packets with foreign saddr to appear on the NF_INET_LOCAL_OUT hook.
32          */
33         if (addr_type == RTN_LOCAL) {
34                 fl.fl4_dst = iph->daddr;
35                 if (type == RTN_LOCAL)
36                         fl.fl4_src = iph->saddr;
37                 fl.fl4_tos = RT_TOS(iph->tos);
38                 fl.oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
39                 fl.mark = skb->mark;
40                 fl.flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
41                 if (ip_route_output_key(net, &rt, &fl) != 0)
42                         return -1;
43
44                 /* Drop old route. */
45                 skb_dst_drop(skb);
46                 skb_dst_set(skb, &rt->dst);
47         } else {
48                 /* non-local src, find valid iif to satisfy
49                  * rp-filter when calling ip_route_input. */
50                 fl.fl4_dst = iph->saddr;
51                 if (ip_route_output_key(net, &rt, &fl) != 0)
52                         return -1;
53
54                 orefdst = skb->_skb_refdst;
55                 if (ip_route_input(skb, iph->daddr, iph->saddr,
56                                    RT_TOS(iph->tos), rt->dst.dev) != 0) {
57                         dst_release(&rt->dst);
58                         return -1;
59                 }
60                 dst_release(&rt->dst);
61                 refdst_drop(orefdst);
62         }
63
64         if (skb_dst(skb)->error)
65                 return -1;
66
67 #ifdef CONFIG_XFRM
68         if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
69             xfrm_decode_session(skb, &fl, AF_INET) == 0) {
70                 struct dst_entry *dst = skb_dst(skb);
71                 skb_dst_set(skb, NULL);
72                 dst = xfrm_lookup(net, dst, &fl, skb->sk, 0);
73                 if (IS_ERR(dst))
74                         return -1;
75                 skb_dst_set(skb, dst);
76         }
77 #endif
78
79         /* Change in oif may mean change in hh_len. */
80         hh_len = skb_dst(skb)->dev->hard_header_len;
81         if (skb_headroom(skb) < hh_len &&
82             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
83                 return -1;
84
85         return 0;
86 }
87 EXPORT_SYMBOL(ip_route_me_harder);
88
89 #ifdef CONFIG_XFRM
90 int ip_xfrm_me_harder(struct sk_buff *skb)
91 {
92         struct flowi fl;
93         unsigned int hh_len;
94         struct dst_entry *dst;
95
96         if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
97                 return 0;
98         if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
99                 return -1;
100
101         dst = skb_dst(skb);
102         if (dst->xfrm)
103                 dst = ((struct xfrm_dst *)dst)->route;
104         dst_hold(dst);
105
106         dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0);
107         if (IS_ERR(dst))
108                 return -1;
109
110         skb_dst_drop(skb);
111         skb_dst_set(skb, dst);
112
113         /* Change in oif may mean change in hh_len. */
114         hh_len = skb_dst(skb)->dev->hard_header_len;
115         if (skb_headroom(skb) < hh_len &&
116             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
117                 return -1;
118         return 0;
119 }
120 EXPORT_SYMBOL(ip_xfrm_me_harder);
121 #endif
122
123 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
124 EXPORT_SYMBOL(ip_nat_decode_session);
125
126 /*
127  * Extra routing may needed on local out, as the QUEUE target never
128  * returns control to the table.
129  */
130
131 struct ip_rt_info {
132         __be32 daddr;
133         __be32 saddr;
134         u_int8_t tos;
135         u_int32_t mark;
136 };
137
138 static void nf_ip_saveroute(const struct sk_buff *skb,
139                             struct nf_queue_entry *entry)
140 {
141         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
142
143         if (entry->hook == NF_INET_LOCAL_OUT) {
144                 const struct iphdr *iph = ip_hdr(skb);
145
146                 rt_info->tos = iph->tos;
147                 rt_info->daddr = iph->daddr;
148                 rt_info->saddr = iph->saddr;
149                 rt_info->mark = skb->mark;
150         }
151 }
152
153 static int nf_ip_reroute(struct sk_buff *skb,
154                          const struct nf_queue_entry *entry)
155 {
156         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
157
158         if (entry->hook == NF_INET_LOCAL_OUT) {
159                 const struct iphdr *iph = ip_hdr(skb);
160
161                 if (!(iph->tos == rt_info->tos &&
162                       skb->mark == rt_info->mark &&
163                       iph->daddr == rt_info->daddr &&
164                       iph->saddr == rt_info->saddr))
165                         return ip_route_me_harder(skb, RTN_UNSPEC);
166         }
167         return 0;
168 }
169
170 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
171                             unsigned int dataoff, u_int8_t protocol)
172 {
173         const struct iphdr *iph = ip_hdr(skb);
174         __sum16 csum = 0;
175
176         switch (skb->ip_summed) {
177         case CHECKSUM_COMPLETE:
178                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
179                         break;
180                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
181                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
182                                        skb->len - dataoff, protocol,
183                                        skb->csum)) {
184                         skb->ip_summed = CHECKSUM_UNNECESSARY;
185                         break;
186                 }
187                 /* fall through */
188         case CHECKSUM_NONE:
189                 if (protocol == 0)
190                         skb->csum = 0;
191                 else
192                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
193                                                        skb->len - dataoff,
194                                                        protocol, 0);
195                 csum = __skb_checksum_complete(skb);
196         }
197         return csum;
198 }
199 EXPORT_SYMBOL(nf_ip_checksum);
200
201 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
202                                       unsigned int dataoff, unsigned int len,
203                                       u_int8_t protocol)
204 {
205         const struct iphdr *iph = ip_hdr(skb);
206         __sum16 csum = 0;
207
208         switch (skb->ip_summed) {
209         case CHECKSUM_COMPLETE:
210                 if (len == skb->len - dataoff)
211                         return nf_ip_checksum(skb, hook, dataoff, protocol);
212                 /* fall through */
213         case CHECKSUM_NONE:
214                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
215                                                skb->len - dataoff, 0);
216                 skb->ip_summed = CHECKSUM_NONE;
217                 return __skb_checksum_complete_head(skb, dataoff + len);
218         }
219         return csum;
220 }
221
222 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
223 {
224         return ip_route_output_key(&init_net, (struct rtable **)dst, fl);
225 }
226
227 static const struct nf_afinfo nf_ip_afinfo = {
228         .family                 = AF_INET,
229         .checksum               = nf_ip_checksum,
230         .checksum_partial       = nf_ip_checksum_partial,
231         .route                  = nf_ip_route,
232         .saveroute              = nf_ip_saveroute,
233         .reroute                = nf_ip_reroute,
234         .route_key_size         = sizeof(struct ip_rt_info),
235 };
236
237 static int ipv4_netfilter_init(void)
238 {
239         return nf_register_afinfo(&nf_ip_afinfo);
240 }
241
242 static void ipv4_netfilter_fini(void)
243 {
244         nf_unregister_afinfo(&nf_ip_afinfo);
245 }
246
247 module_init(ipv4_netfilter_init);
248 module_exit(ipv4_netfilter_fini);
249
250 #ifdef CONFIG_SYSCTL
251 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
252         { .procname = "net", },
253         { .procname = "ipv4", },
254         { .procname = "netfilter", },
255         { }
256 };
257 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
258 #endif /* CONFIG_SYSCTL */