ipv4: Use flowi4 in public route lookup interfaces.
[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 flowi4 fl4 = {};
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                 fl4.daddr = iph->daddr;
35                 if (type == RTN_LOCAL)
36                         fl4.saddr = iph->saddr;
37                 fl4.flowi4_tos = RT_TOS(iph->tos);
38                 fl4.flowi4_oif = skb->sk ? skb->sk->sk_bound_dev_if : 0;
39                 fl4.flowi4_mark = skb->mark;
40                 fl4.flowi4_flags = skb->sk ? inet_sk_flowi_flags(skb->sk) : 0;
41                 rt = ip_route_output_key(net, &fl4);
42                 if (IS_ERR(rt))
43                         return -1;
44
45                 /* Drop old route. */
46                 skb_dst_drop(skb);
47                 skb_dst_set(skb, &rt->dst);
48         } else {
49                 /* non-local src, find valid iif to satisfy
50                  * rp-filter when calling ip_route_input. */
51                 fl4.daddr = iph->saddr;
52                 rt = ip_route_output_key(net, &fl4);
53                 if (IS_ERR(rt))
54                         return -1;
55
56                 orefdst = skb->_skb_refdst;
57                 if (ip_route_input(skb, iph->daddr, iph->saddr,
58                                    RT_TOS(iph->tos), rt->dst.dev) != 0) {
59                         dst_release(&rt->dst);
60                         return -1;
61                 }
62                 dst_release(&rt->dst);
63                 refdst_drop(orefdst);
64         }
65
66         if (skb_dst(skb)->error)
67                 return -1;
68
69 #ifdef CONFIG_XFRM
70         if (!(IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED) &&
71             xfrm_decode_session(skb, flowi4_to_flowi(&fl4), AF_INET) == 0) {
72                 struct dst_entry *dst = skb_dst(skb);
73                 skb_dst_set(skb, NULL);
74                 dst = xfrm_lookup(net, dst, flowi4_to_flowi(&fl4), skb->sk, 0);
75                 if (IS_ERR(dst))
76                         return -1;
77                 skb_dst_set(skb, dst);
78         }
79 #endif
80
81         /* Change in oif may mean change in hh_len. */
82         hh_len = skb_dst(skb)->dev->hard_header_len;
83         if (skb_headroom(skb) < hh_len &&
84             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
85                 return -1;
86
87         return 0;
88 }
89 EXPORT_SYMBOL(ip_route_me_harder);
90
91 #ifdef CONFIG_XFRM
92 int ip_xfrm_me_harder(struct sk_buff *skb)
93 {
94         struct flowi fl;
95         unsigned int hh_len;
96         struct dst_entry *dst;
97
98         if (IPCB(skb)->flags & IPSKB_XFRM_TRANSFORMED)
99                 return 0;
100         if (xfrm_decode_session(skb, &fl, AF_INET) < 0)
101                 return -1;
102
103         dst = skb_dst(skb);
104         if (dst->xfrm)
105                 dst = ((struct xfrm_dst *)dst)->route;
106         dst_hold(dst);
107
108         dst = xfrm_lookup(dev_net(dst->dev), dst, &fl, skb->sk, 0);
109         if (IS_ERR(dst))
110                 return -1;
111
112         skb_dst_drop(skb);
113         skb_dst_set(skb, dst);
114
115         /* Change in oif may mean change in hh_len. */
116         hh_len = skb_dst(skb)->dev->hard_header_len;
117         if (skb_headroom(skb) < hh_len &&
118             pskb_expand_head(skb, hh_len - skb_headroom(skb), 0, GFP_ATOMIC))
119                 return -1;
120         return 0;
121 }
122 EXPORT_SYMBOL(ip_xfrm_me_harder);
123 #endif
124
125 void (*ip_nat_decode_session)(struct sk_buff *, struct flowi *);
126 EXPORT_SYMBOL(ip_nat_decode_session);
127
128 /*
129  * Extra routing may needed on local out, as the QUEUE target never
130  * returns control to the table.
131  */
132
133 struct ip_rt_info {
134         __be32 daddr;
135         __be32 saddr;
136         u_int8_t tos;
137         u_int32_t mark;
138 };
139
140 static void nf_ip_saveroute(const struct sk_buff *skb,
141                             struct nf_queue_entry *entry)
142 {
143         struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
144
145         if (entry->hook == NF_INET_LOCAL_OUT) {
146                 const struct iphdr *iph = ip_hdr(skb);
147
148                 rt_info->tos = iph->tos;
149                 rt_info->daddr = iph->daddr;
150                 rt_info->saddr = iph->saddr;
151                 rt_info->mark = skb->mark;
152         }
153 }
154
155 static int nf_ip_reroute(struct sk_buff *skb,
156                          const struct nf_queue_entry *entry)
157 {
158         const struct ip_rt_info *rt_info = nf_queue_entry_reroute(entry);
159
160         if (entry->hook == NF_INET_LOCAL_OUT) {
161                 const struct iphdr *iph = ip_hdr(skb);
162
163                 if (!(iph->tos == rt_info->tos &&
164                       skb->mark == rt_info->mark &&
165                       iph->daddr == rt_info->daddr &&
166                       iph->saddr == rt_info->saddr))
167                         return ip_route_me_harder(skb, RTN_UNSPEC);
168         }
169         return 0;
170 }
171
172 __sum16 nf_ip_checksum(struct sk_buff *skb, unsigned int hook,
173                             unsigned int dataoff, u_int8_t protocol)
174 {
175         const struct iphdr *iph = ip_hdr(skb);
176         __sum16 csum = 0;
177
178         switch (skb->ip_summed) {
179         case CHECKSUM_COMPLETE:
180                 if (hook != NF_INET_PRE_ROUTING && hook != NF_INET_LOCAL_IN)
181                         break;
182                 if ((protocol == 0 && !csum_fold(skb->csum)) ||
183                     !csum_tcpudp_magic(iph->saddr, iph->daddr,
184                                        skb->len - dataoff, protocol,
185                                        skb->csum)) {
186                         skb->ip_summed = CHECKSUM_UNNECESSARY;
187                         break;
188                 }
189                 /* fall through */
190         case CHECKSUM_NONE:
191                 if (protocol == 0)
192                         skb->csum = 0;
193                 else
194                         skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr,
195                                                        skb->len - dataoff,
196                                                        protocol, 0);
197                 csum = __skb_checksum_complete(skb);
198         }
199         return csum;
200 }
201 EXPORT_SYMBOL(nf_ip_checksum);
202
203 static __sum16 nf_ip_checksum_partial(struct sk_buff *skb, unsigned int hook,
204                                       unsigned int dataoff, unsigned int len,
205                                       u_int8_t protocol)
206 {
207         const struct iphdr *iph = ip_hdr(skb);
208         __sum16 csum = 0;
209
210         switch (skb->ip_summed) {
211         case CHECKSUM_COMPLETE:
212                 if (len == skb->len - dataoff)
213                         return nf_ip_checksum(skb, hook, dataoff, protocol);
214                 /* fall through */
215         case CHECKSUM_NONE:
216                 skb->csum = csum_tcpudp_nofold(iph->saddr, iph->daddr, protocol,
217                                                skb->len - dataoff, 0);
218                 skb->ip_summed = CHECKSUM_NONE;
219                 return __skb_checksum_complete_head(skb, dataoff + len);
220         }
221         return csum;
222 }
223
224 static int nf_ip_route(struct dst_entry **dst, struct flowi *fl)
225 {
226         struct rtable *rt = ip_route_output_key(&init_net, &fl->u.ip4);
227         if (IS_ERR(rt))
228                 return PTR_ERR(rt);
229         *dst = &rt->dst;
230         return 0;
231 }
232
233 static const struct nf_afinfo nf_ip_afinfo = {
234         .family                 = AF_INET,
235         .checksum               = nf_ip_checksum,
236         .checksum_partial       = nf_ip_checksum_partial,
237         .route                  = nf_ip_route,
238         .saveroute              = nf_ip_saveroute,
239         .reroute                = nf_ip_reroute,
240         .route_key_size         = sizeof(struct ip_rt_info),
241 };
242
243 static int ipv4_netfilter_init(void)
244 {
245         return nf_register_afinfo(&nf_ip_afinfo);
246 }
247
248 static void ipv4_netfilter_fini(void)
249 {
250         nf_unregister_afinfo(&nf_ip_afinfo);
251 }
252
253 module_init(ipv4_netfilter_init);
254 module_exit(ipv4_netfilter_fini);
255
256 #ifdef CONFIG_SYSCTL
257 struct ctl_path nf_net_ipv4_netfilter_sysctl_path[] = {
258         { .procname = "net", },
259         { .procname = "ipv4", },
260         { .procname = "netfilter", },
261         { }
262 };
263 EXPORT_SYMBOL_GPL(nf_net_ipv4_netfilter_sysctl_path);
264 #endif /* CONFIG_SYSCTL */