vserver 2.0 rc7
[linux-2.6.git] / net / ipv4 / netfilter / ipt_REJECT.c
1 /*
2  * This is a module which is used for rejecting packets.
3  * Added support for customized reject packets (Jozsef Kadlecsik).
4  * Added support for ICMP type-3-code-13 (Maciej Soltysiak). [RFC 1812]
5  */
6
7 /* (C) 1999-2001 Paul `Rusty' Russell
8  * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
9  *
10  * This program is free software; you can redistribute it and/or modify
11  * it under the terms of the GNU General Public License version 2 as
12  * published by the Free Software Foundation.
13  */
14
15 #include <linux/config.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
18 #include <linux/ip.h>
19 #include <linux/udp.h>
20 #include <linux/icmp.h>
21 #include <net/icmp.h>
22 #include <net/ip.h>
23 #include <net/tcp.h>
24 #include <net/route.h>
25 #include <net/dst.h>
26 #include <linux/netfilter_ipv4/ip_tables.h>
27 #include <linux/netfilter_ipv4/ipt_REJECT.h>
28 #ifdef CONFIG_BRIDGE_NETFILTER
29 #include <linux/netfilter_bridge.h>
30 #endif
31
32 MODULE_LICENSE("GPL");
33 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
34 MODULE_DESCRIPTION("iptables REJECT target module");
35
36 #if 0
37 #define DEBUGP printk
38 #else
39 #define DEBUGP(format, args...)
40 #endif
41
42 static inline struct rtable *route_reverse(struct sk_buff *skb, 
43                                            struct tcphdr *tcph, int hook)
44 {
45         struct iphdr *iph = skb->nh.iph;
46         struct dst_entry *odst;
47         struct flowi fl = {};
48         struct rtable *rt;
49
50         /* We don't require ip forwarding to be enabled to be able to
51          * send a RST reply for bridged traffic. */
52         if (hook != NF_IP_FORWARD
53 #ifdef CONFIG_BRIDGE_NETFILTER
54             || (skb->nf_bridge && skb->nf_bridge->mask & BRNF_BRIDGED)
55 #endif
56            ) {
57                 fl.nl_u.ip4_u.daddr = iph->saddr;
58                 if (hook == NF_IP_LOCAL_IN)
59                         fl.nl_u.ip4_u.saddr = iph->daddr;
60                 fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
61
62                 if (ip_route_output_key(&rt, &fl) != 0)
63                         return NULL;
64         } else {
65                 /* non-local src, find valid iif to satisfy
66                  * rp-filter when calling ip_route_input. */
67                 fl.nl_u.ip4_u.daddr = iph->daddr;
68                 if (ip_route_output_key(&rt, &fl) != 0)
69                         return NULL;
70
71                 odst = skb->dst;
72                 if (ip_route_input(skb, iph->saddr, iph->daddr,
73                                    RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
74                         dst_release(&rt->u.dst);
75                         return NULL;
76                 }
77                 dst_release(&rt->u.dst);
78                 rt = (struct rtable *)skb->dst;
79                 skb->dst = odst;
80
81                 fl.nl_u.ip4_u.daddr = iph->saddr;
82                 fl.nl_u.ip4_u.saddr = iph->daddr;
83                 fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
84         }
85
86         if (rt->u.dst.error) {
87                 dst_release(&rt->u.dst);
88                 return NULL;
89         }
90
91         fl.proto = IPPROTO_TCP;
92         fl.fl_ip_sport = tcph->dest;
93         fl.fl_ip_dport = tcph->source;
94
95         if (xfrm_lookup((struct dst_entry **)&rt, &fl, NULL, 0)) {
96                 dst_release(&rt->u.dst);
97                 rt = NULL;
98         }
99
100         return rt;
101 }
102
103 /* Send RST reply */
104 static void send_reset(struct sk_buff *oldskb, int hook)
105 {
106         struct sk_buff *nskb;
107         struct tcphdr _otcph, *oth, *tcph;
108         struct rtable *rt;
109         u_int16_t tmp_port;
110         u_int32_t tmp_addr;
111         int needs_ack;
112         int hh_len;
113
114         /* IP header checks: fragment. */
115         if (oldskb->nh.iph->frag_off & htons(IP_OFFSET))
116                 return;
117
118         oth = skb_header_pointer(oldskb, oldskb->nh.iph->ihl * 4,
119                                  sizeof(_otcph), &_otcph);
120         if (oth == NULL)
121                 return;
122
123         /* No RST for RST. */
124         if (oth->rst)
125                 return;
126
127         /* FIXME: Check checksum --RR */
128         if ((rt = route_reverse(oldskb, oth, hook)) == NULL)
129                 return;
130
131         hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
132
133         /* We need a linear, writeable skb.  We also need to expand
134            headroom in case hh_len of incoming interface < hh_len of
135            outgoing interface */
136         nskb = skb_copy_expand(oldskb, hh_len, skb_tailroom(oldskb),
137                                GFP_ATOMIC);
138         if (!nskb) {
139                 dst_release(&rt->u.dst);
140                 return;
141         }
142
143         dst_release(nskb->dst);
144         nskb->dst = &rt->u.dst;
145
146         /* This packet will not be the same as the other: clear nf fields */
147         nf_reset(nskb);
148         nskb->nfcache = 0;
149         nskb->nfmark = 0;
150 #ifdef CONFIG_BRIDGE_NETFILTER
151         nf_bridge_put(nskb->nf_bridge);
152         nskb->nf_bridge = NULL;
153 #endif
154
155         tcph = (struct tcphdr *)((u_int32_t*)nskb->nh.iph + nskb->nh.iph->ihl);
156
157         /* Swap source and dest */
158         tmp_addr = nskb->nh.iph->saddr;
159         nskb->nh.iph->saddr = nskb->nh.iph->daddr;
160         nskb->nh.iph->daddr = tmp_addr;
161         tmp_port = tcph->source;
162         tcph->source = tcph->dest;
163         tcph->dest = tmp_port;
164
165         /* Truncate to length (no data) */
166         tcph->doff = sizeof(struct tcphdr)/4;
167         skb_trim(nskb, nskb->nh.iph->ihl*4 + sizeof(struct tcphdr));
168         nskb->nh.iph->tot_len = htons(nskb->len);
169
170         if (tcph->ack) {
171                 needs_ack = 0;
172                 tcph->seq = oth->ack_seq;
173                 tcph->ack_seq = 0;
174         } else {
175                 needs_ack = 1;
176                 tcph->ack_seq = htonl(ntohl(oth->seq) + oth->syn + oth->fin
177                                       + oldskb->len - oldskb->nh.iph->ihl*4
178                                       - (oth->doff<<2));
179                 tcph->seq = 0;
180         }
181
182         /* Reset flags */
183         ((u_int8_t *)tcph)[13] = 0;
184         tcph->rst = 1;
185         tcph->ack = needs_ack;
186
187         tcph->window = 0;
188         tcph->urg_ptr = 0;
189
190         /* Adjust TCP checksum */
191         tcph->check = 0;
192         tcph->check = tcp_v4_check(tcph, sizeof(struct tcphdr),
193                                    nskb->nh.iph->saddr,
194                                    nskb->nh.iph->daddr,
195                                    csum_partial((char *)tcph,
196                                                 sizeof(struct tcphdr), 0));
197
198         /* Adjust IP TTL, DF */
199         nskb->nh.iph->ttl = MAXTTL;
200         /* Set DF, id = 0 */
201         nskb->nh.iph->frag_off = htons(IP_DF);
202         nskb->nh.iph->id = 0;
203
204         /* Adjust IP checksum */
205         nskb->nh.iph->check = 0;
206         nskb->nh.iph->check = ip_fast_csum((unsigned char *)nskb->nh.iph, 
207                                            nskb->nh.iph->ihl);
208
209         /* "Never happens" */
210         if (nskb->len > dst_mtu(nskb->dst))
211                 goto free_nskb;
212
213         nf_ct_attach(nskb, oldskb);
214
215         NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
216                 dst_output);
217         return;
218
219  free_nskb:
220         kfree_skb(nskb);
221 }
222
223 static inline void send_unreach(struct sk_buff *skb_in, int code)
224 {
225         icmp_send(skb_in, ICMP_DEST_UNREACH, code, 0);
226 }       
227
228 static unsigned int reject(struct sk_buff **pskb,
229                            const struct net_device *in,
230                            const struct net_device *out,
231                            unsigned int hooknum,
232                            const void *targinfo,
233                            void *userinfo)
234 {
235         const struct ipt_reject_info *reject = targinfo;
236
237         /* Our naive response construction doesn't deal with IP
238            options, and probably shouldn't try. */
239         if ((*pskb)->nh.iph->ihl<<2 != sizeof(struct iphdr))
240                 return NF_DROP;
241
242         /* WARNING: This code causes reentry within iptables.
243            This means that the iptables jump stack is now crap.  We
244            must return an absolute verdict. --RR */
245         switch (reject->with) {
246         case IPT_ICMP_NET_UNREACHABLE:
247                 send_unreach(*pskb, ICMP_NET_UNREACH);
248                 break;
249         case IPT_ICMP_HOST_UNREACHABLE:
250                 send_unreach(*pskb, ICMP_HOST_UNREACH);
251                 break;
252         case IPT_ICMP_PROT_UNREACHABLE:
253                 send_unreach(*pskb, ICMP_PROT_UNREACH);
254                 break;
255         case IPT_ICMP_PORT_UNREACHABLE:
256                 send_unreach(*pskb, ICMP_PORT_UNREACH);
257                 break;
258         case IPT_ICMP_NET_PROHIBITED:
259                 send_unreach(*pskb, ICMP_NET_ANO);
260                 break;
261         case IPT_ICMP_HOST_PROHIBITED:
262                 send_unreach(*pskb, ICMP_HOST_ANO);
263                 break;
264         case IPT_ICMP_ADMIN_PROHIBITED:
265                 send_unreach(*pskb, ICMP_PKT_FILTERED);
266                 break;
267         case IPT_TCP_RESET:
268                 send_reset(*pskb, hooknum);
269         case IPT_ICMP_ECHOREPLY:
270                 /* Doesn't happen. */
271                 break;
272         }
273
274         return NF_DROP;
275 }
276
277 static int check(const char *tablename,
278                  const struct ipt_entry *e,
279                  void *targinfo,
280                  unsigned int targinfosize,
281                  unsigned int hook_mask)
282 {
283         const struct ipt_reject_info *rejinfo = targinfo;
284
285         if (targinfosize != IPT_ALIGN(sizeof(struct ipt_reject_info))) {
286                 DEBUGP("REJECT: targinfosize %u != 0\n", targinfosize);
287                 return 0;
288         }
289
290         /* Only allow these for packet filtering. */
291         if (strcmp(tablename, "filter") != 0) {
292                 DEBUGP("REJECT: bad table `%s'.\n", tablename);
293                 return 0;
294         }
295         if ((hook_mask & ~((1 << NF_IP_LOCAL_IN)
296                            | (1 << NF_IP_FORWARD)
297                            | (1 << NF_IP_LOCAL_OUT))) != 0) {
298                 DEBUGP("REJECT: bad hook mask %X\n", hook_mask);
299                 return 0;
300         }
301
302         if (rejinfo->with == IPT_ICMP_ECHOREPLY) {
303                 printk("REJECT: ECHOREPLY no longer supported.\n");
304                 return 0;
305         } else if (rejinfo->with == IPT_TCP_RESET) {
306                 /* Must specify that it's a TCP packet */
307                 if (e->ip.proto != IPPROTO_TCP
308                     || (e->ip.invflags & IPT_INV_PROTO)) {
309                         DEBUGP("REJECT: TCP_RESET invalid for non-tcp\n");
310                         return 0;
311                 }
312         }
313
314         return 1;
315 }
316
317 static struct ipt_target ipt_reject_reg = {
318         .name           = "REJECT",
319         .target         = reject,
320         .checkentry     = check,
321         .me             = THIS_MODULE,
322 };
323
324 static int __init init(void)
325 {
326         return ipt_register_target(&ipt_reject_reg);
327 }
328
329 static void __exit fini(void)
330 {
331         ipt_unregister_target(&ipt_reject_reg);
332 }
333
334 module_init(init);
335 module_exit(fini);