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]
7 /* (C) 1999-2001 Paul `Rusty' Russell
8 * (C) 2002-2004 Netfilter Core Team <coreteam@netfilter.org>
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.
15 #include <linux/config.h>
16 #include <linux/module.h>
17 #include <linux/skbuff.h>
19 #include <linux/udp.h>
20 #include <linux/icmp.h>
24 #include <net/route.h>
25 #include <linux/netfilter_ipv4/ip_tables.h>
26 #include <linux/netfilter_ipv4/ipt_REJECT.h>
27 #ifdef CONFIG_BRIDGE_NETFILTER
28 #include <linux/netfilter_bridge.h>
31 MODULE_LICENSE("GPL");
32 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
33 MODULE_DESCRIPTION("iptables REJECT target module");
38 #define DEBUGP(format, args...)
41 /* If the original packet is part of a connection, but the connection
42 is not confirmed, our manufactured reply will not be associated
43 with it, so we need to do this manually. */
44 static void connection_attach(struct sk_buff *new_skb, struct nf_ct_info *nfct)
46 void (*attach)(struct sk_buff *, struct nf_ct_info *);
48 /* Avoid module unload race with ip_ct_attach being NULLed out */
49 if (nfct && (attach = ip_ct_attach) != NULL) {
50 mb(); /* Just to be sure: must be read before executing this */
51 attach(new_skb, nfct);
55 static inline struct rtable *route_reverse(struct sk_buff *skb, int hook)
57 struct iphdr *iph = skb->nh.iph;
58 struct dst_entry *odst;
62 /* We don't require ip forwarding to be enabled to be able to
63 * send a RST reply for bridged traffic. */
64 if (hook != NF_IP_FORWARD
65 #ifdef CONFIG_BRIDGE_NETFILTER
66 || (skb->nf_bridge && skb->nf_bridge->mask & BRNF_BRIDGED)
69 fl.nl_u.ip4_u.daddr = iph->saddr;
70 if (hook == NF_IP_LOCAL_IN)
71 fl.nl_u.ip4_u.saddr = iph->daddr;
72 fl.nl_u.ip4_u.tos = RT_TOS(iph->tos);
74 if (ip_route_output_key(&rt, &fl) != 0)
77 /* non-local src, find valid iif to satisfy
78 * rp-filter when calling ip_route_input. */
79 fl.nl_u.ip4_u.daddr = iph->daddr;
80 if (ip_route_output_key(&rt, &fl) != 0)
84 if (ip_route_input(skb, iph->saddr, iph->daddr,
85 RT_TOS(iph->tos), rt->u.dst.dev) != 0) {
86 dst_release(&rt->u.dst);
89 dst_release(&rt->u.dst);
90 rt = (struct rtable *)skb->dst;
94 if (rt->u.dst.error) {
95 dst_release(&rt->u.dst);
103 static void send_reset(struct sk_buff *oldskb, int hook)
105 struct sk_buff *nskb;
106 struct tcphdr otcph, *tcph;
113 /* IP header checks: fragment. */
114 if (oldskb->nh.iph->frag_off & htons(IP_OFFSET))
117 if (skb_copy_bits(oldskb, oldskb->nh.iph->ihl*4,
118 &otcph, sizeof(otcph)) < 0)
121 /* No RST for RST. */
125 /* FIXME: Check checksum --RR */
126 if ((rt = route_reverse(oldskb, hook)) == NULL)
129 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
131 /* We need a linear, writeable skb. We also need to expand
132 headroom in case hh_len of incoming interface < hh_len of
133 outgoing interface */
134 nskb = skb_copy_expand(oldskb, hh_len, skb_tailroom(oldskb),
137 dst_release(&rt->u.dst);
141 dst_release(nskb->dst);
142 nskb->dst = &rt->u.dst;
144 /* This packet will not be the same as the other: clear nf fields */
148 #ifdef CONFIG_BRIDGE_NETFILTER
149 nf_bridge_put(nskb->nf_bridge);
150 nskb->nf_bridge = NULL;
153 tcph = (struct tcphdr *)((u_int32_t*)nskb->nh.iph + nskb->nh.iph->ihl);
155 /* Swap source and dest */
156 tmp_addr = nskb->nh.iph->saddr;
157 nskb->nh.iph->saddr = nskb->nh.iph->daddr;
158 nskb->nh.iph->daddr = tmp_addr;
159 tmp_port = tcph->source;
160 tcph->source = tcph->dest;
161 tcph->dest = tmp_port;
163 /* Truncate to length (no data) */
164 tcph->doff = sizeof(struct tcphdr)/4;
165 skb_trim(nskb, nskb->nh.iph->ihl*4 + sizeof(struct tcphdr));
166 nskb->nh.iph->tot_len = htons(nskb->len);
170 tcph->seq = otcph.ack_seq;
174 tcph->ack_seq = htonl(ntohl(otcph.seq) + otcph.syn + otcph.fin
175 + oldskb->len - oldskb->nh.iph->ihl*4
181 ((u_int8_t *)tcph)[13] = 0;
183 tcph->ack = needs_ack;
188 /* Adjust TCP checksum */
190 tcph->check = tcp_v4_check(tcph, sizeof(struct tcphdr),
193 csum_partial((char *)tcph,
194 sizeof(struct tcphdr), 0));
196 /* Adjust IP TTL, DF */
197 nskb->nh.iph->ttl = MAXTTL;
199 nskb->nh.iph->frag_off = htons(IP_DF);
200 nskb->nh.iph->id = 0;
202 /* Adjust IP checksum */
203 nskb->nh.iph->check = 0;
204 nskb->nh.iph->check = ip_fast_csum((unsigned char *)nskb->nh.iph,
207 /* "Never happens" */
208 if (nskb->len > dst_pmtu(nskb->dst))
211 connection_attach(nskb, oldskb->nfct);
213 NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
221 static void send_unreach(struct sk_buff *skb_in, int code)
225 struct icmphdr *icmph;
226 struct sk_buff *nskb;
230 struct rtable *rt = (struct rtable*)skb_in->dst;
236 /* FIXME: Use sysctl number. --RR */
237 if (!xrlim_allow(&rt->u.dst, 1*HZ))
240 iph = skb_in->nh.iph;
242 /* No replies to physical multicast/broadcast */
243 if (skb_in->pkt_type!=PACKET_HOST)
246 /* Now check at the protocol level */
247 if (rt->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST))
250 /* Only reply to fragment 0. */
251 if (iph->frag_off&htons(IP_OFFSET))
254 /* Ensure we have at least 8 bytes of proto header. */
255 if (skb_in->len < skb_in->nh.iph->ihl*4 + 8)
258 /* if UDP checksum is set, verify it's correct */
259 if (iph->protocol == IPPROTO_UDP
260 && skb_in->tail-(u8*)iph >= sizeof(struct udphdr)) {
261 int datalen = skb_in->len - (iph->ihl<<2);
262 udph = (struct udphdr *)((char *)iph + (iph->ihl<<2));
264 && csum_tcpudp_magic(iph->saddr, iph->daddr,
265 datalen, IPPROTO_UDP,
266 csum_partial((char *)udph, datalen,
271 /* If we send an ICMP error to an ICMP error a mess would result.. */
272 if (iph->protocol == IPPROTO_ICMP
273 && skb_in->tail-(u8*)iph >= sizeof(struct icmphdr)) {
274 icmph = (struct icmphdr *)((char *)iph + (iph->ihl<<2));
276 if (skb_copy_bits(skb_in, skb_in->nh.iph->ihl*4,
277 icmph, sizeof(*icmph)) < 0)
280 /* Between echo-reply (0) and timestamp (13),
281 everything except echo-request (8) is an error.
282 Also, anything greater than NR_ICMP_TYPES is
283 unknown, and hence should be treated as an error... */
284 if ((icmph->type < ICMP_TIMESTAMP
285 && icmph->type != ICMP_ECHOREPLY
286 && icmph->type != ICMP_ECHO)
287 || icmph->type > NR_ICMP_TYPES)
292 if (!(rt->rt_flags & RTCF_LOCAL))
295 tos = (iph->tos & IPTOS_TOS_MASK) | IPTOS_PREC_INTERNETCONTROL;
298 struct flowi fl = { .nl_u = { .ip4_u =
299 { .daddr = skb_in->nh.iph->saddr,
301 .tos = RT_TOS(tos) } } };
302 if (ip_route_output_key(&rt, &fl))
305 /* RFC says return as much as we can without exceeding 576 bytes. */
306 length = skb_in->len + sizeof(struct iphdr) + sizeof(struct icmphdr);
308 if (length > dst_pmtu(&rt->u.dst))
309 length = dst_pmtu(&rt->u.dst);
313 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
315 nskb = alloc_skb(hh_len + length, GFP_ATOMIC);
322 nskb->dst = &rt->u.dst;
323 skb_reserve(nskb, hh_len);
325 /* Set up IP header */
327 = (struct iphdr *)skb_put(nskb, sizeof(struct iphdr));
331 iph->tot_len = htons(length);
333 /* PMTU discovery never applies to ICMP packets. */
337 ip_select_ident(iph, &rt->u.dst, NULL);
338 iph->protocol=IPPROTO_ICMP;
339 iph->saddr=rt->rt_src;
340 iph->daddr=rt->rt_dst;
342 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
344 /* Set up ICMP header. */
345 icmph = nskb->h.icmph
346 = (struct icmphdr *)skb_put(nskb, sizeof(struct icmphdr));
347 icmph->type = ICMP_DEST_UNREACH;
349 icmph->un.gateway = 0;
352 /* Copy as much of original packet as will fit */
354 length - sizeof(struct iphdr) - sizeof(struct icmphdr));
356 skb_copy_bits(skb_in, 0, data,
357 length - sizeof(struct iphdr) - sizeof(struct icmphdr));
359 icmph->checksum = ip_compute_csum((unsigned char *)icmph,
360 length - sizeof(struct iphdr));
362 connection_attach(nskb, skb_in->nfct);
364 NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
368 static unsigned int reject(struct sk_buff **pskb,
369 const struct net_device *in,
370 const struct net_device *out,
371 unsigned int hooknum,
372 const void *targinfo,
375 const struct ipt_reject_info *reject = targinfo;
377 /* Our naive response construction doesn't deal with IP
378 options, and probably shouldn't try. */
379 if ((*pskb)->nh.iph->ihl<<2 != sizeof(struct iphdr))
382 /* WARNING: This code causes reentry within iptables.
383 This means that the iptables jump stack is now crap. We
384 must return an absolute verdict. --RR */
385 switch (reject->with) {
386 case IPT_ICMP_NET_UNREACHABLE:
387 send_unreach(*pskb, ICMP_NET_UNREACH);
389 case IPT_ICMP_HOST_UNREACHABLE:
390 send_unreach(*pskb, ICMP_HOST_UNREACH);
392 case IPT_ICMP_PROT_UNREACHABLE:
393 send_unreach(*pskb, ICMP_PROT_UNREACH);
395 case IPT_ICMP_PORT_UNREACHABLE:
396 send_unreach(*pskb, ICMP_PORT_UNREACH);
398 case IPT_ICMP_NET_PROHIBITED:
399 send_unreach(*pskb, ICMP_NET_ANO);
401 case IPT_ICMP_HOST_PROHIBITED:
402 send_unreach(*pskb, ICMP_HOST_ANO);
404 case IPT_ICMP_ADMIN_PROHIBITED:
405 send_unreach(*pskb, ICMP_PKT_FILTERED);
408 send_reset(*pskb, hooknum);
409 case IPT_ICMP_ECHOREPLY:
410 /* Doesn't happen. */
417 static int check(const char *tablename,
418 const struct ipt_entry *e,
420 unsigned int targinfosize,
421 unsigned int hook_mask)
423 const struct ipt_reject_info *rejinfo = targinfo;
425 if (targinfosize != IPT_ALIGN(sizeof(struct ipt_reject_info))) {
426 DEBUGP("REJECT: targinfosize %u != 0\n", targinfosize);
430 /* Only allow these for packet filtering. */
431 if (strcmp(tablename, "filter") != 0) {
432 DEBUGP("REJECT: bad table `%s'.\n", tablename);
435 if ((hook_mask & ~((1 << NF_IP_LOCAL_IN)
436 | (1 << NF_IP_FORWARD)
437 | (1 << NF_IP_LOCAL_OUT))) != 0) {
438 DEBUGP("REJECT: bad hook mask %X\n", hook_mask);
442 if (rejinfo->with == IPT_ICMP_ECHOREPLY) {
443 printk("REJECT: ECHOREPLY no longer supported.\n");
445 } else if (rejinfo->with == IPT_TCP_RESET) {
446 /* Must specify that it's a TCP packet */
447 if (e->ip.proto != IPPROTO_TCP
448 || (e->ip.invflags & IPT_INV_PROTO)) {
449 DEBUGP("REJECT: TCP_RESET invalid for non-tcp\n");
457 static struct ipt_target ipt_reject_reg = {
464 static int __init init(void)
466 return ipt_register_target(&ipt_reject_reg);
469 static void __exit fini(void)
471 ipt_unregister_target(&ipt_reject_reg);