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>
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>
32 MODULE_LICENSE("GPL");
33 MODULE_AUTHOR("Netfilter Core Team <coreteam@netfilter.org>");
34 MODULE_DESCRIPTION("iptables REJECT target module");
39 #define DEBUGP(format, args...)
42 static inline struct rtable *route_reverse(struct sk_buff *skb,
43 struct tcphdr *tcph, int hook)
45 struct iphdr *iph = skb->nh.iph;
46 struct dst_entry *odst;
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)
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);
62 if (ip_route_output_key(&rt, &fl) != 0)
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)
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);
77 dst_release(&rt->u.dst);
78 rt = (struct rtable *)skb->dst;
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);
86 if (rt->u.dst.error) {
87 dst_release(&rt->u.dst);
91 fl.proto = IPPROTO_TCP;
92 fl.fl_ip_sport = tcph->dest;
93 fl.fl_ip_dport = tcph->source;
95 if (xfrm_lookup((struct dst_entry **)&rt, &fl, NULL, 0)) {
96 dst_release(&rt->u.dst);
104 static void send_reset(struct sk_buff *oldskb, int hook)
106 struct sk_buff *nskb;
107 struct tcphdr _otcph, *oth, *tcph;
114 /* IP header checks: fragment. */
115 if (oldskb->nh.iph->frag_off & htons(IP_OFFSET))
118 oth = skb_header_pointer(oldskb, oldskb->nh.iph->ihl * 4,
119 sizeof(_otcph), &_otcph);
123 /* No RST for RST. */
127 /* FIXME: Check checksum --RR */
128 if ((rt = route_reverse(oldskb, oth, hook)) == NULL)
131 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
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),
139 dst_release(&rt->u.dst);
143 dst_release(nskb->dst);
144 nskb->dst = &rt->u.dst;
146 /* This packet will not be the same as the other: clear nf fields */
150 #ifdef CONFIG_BRIDGE_NETFILTER
151 nf_bridge_put(nskb->nf_bridge);
152 nskb->nf_bridge = NULL;
155 tcph = (struct tcphdr *)((u_int32_t*)nskb->nh.iph + nskb->nh.iph->ihl);
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;
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);
172 tcph->seq = oth->ack_seq;
176 tcph->ack_seq = htonl(ntohl(oth->seq) + oth->syn + oth->fin
177 + oldskb->len - oldskb->nh.iph->ihl*4
183 ((u_int8_t *)tcph)[13] = 0;
185 tcph->ack = needs_ack;
190 /* Adjust TCP checksum */
192 tcph->check = tcp_v4_check(tcph, sizeof(struct tcphdr),
195 csum_partial((char *)tcph,
196 sizeof(struct tcphdr), 0));
198 /* Adjust IP TTL, DF */
199 nskb->nh.iph->ttl = MAXTTL;
201 nskb->nh.iph->frag_off = htons(IP_DF);
202 nskb->nh.iph->id = 0;
204 /* Adjust IP checksum */
205 nskb->nh.iph->check = 0;
206 nskb->nh.iph->check = ip_fast_csum((unsigned char *)nskb->nh.iph,
209 /* "Never happens" */
210 if (nskb->len > dst_pmtu(nskb->dst))
213 nf_ct_attach(nskb, oldskb);
215 NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
223 static void send_unreach(struct sk_buff *skb_in, int code)
227 struct icmphdr *icmph;
228 struct sk_buff *nskb;
232 struct rtable *rt = (struct rtable*)skb_in->dst;
238 /* FIXME: Use sysctl number. --RR */
239 if (!xrlim_allow(&rt->u.dst, 1*HZ))
242 iph = skb_in->nh.iph;
244 /* No replies to physical multicast/broadcast */
245 if (skb_in->pkt_type!=PACKET_HOST)
248 /* Now check at the protocol level */
249 if (rt->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST))
252 /* Only reply to fragment 0. */
253 if (iph->frag_off&htons(IP_OFFSET))
256 /* Ensure we have at least 8 bytes of proto header. */
257 if (skb_in->len < skb_in->nh.iph->ihl*4 + 8)
260 /* if UDP checksum is set, verify it's correct */
261 if (iph->protocol == IPPROTO_UDP
262 && skb_in->tail-(u8*)iph >= sizeof(struct udphdr)) {
263 int datalen = skb_in->len - (iph->ihl<<2);
264 udph = (struct udphdr *)((char *)iph + (iph->ihl<<2));
266 && csum_tcpudp_magic(iph->saddr, iph->daddr,
267 datalen, IPPROTO_UDP,
268 csum_partial((char *)udph, datalen,
273 /* If we send an ICMP error to an ICMP error a mess would result.. */
274 if (iph->protocol == IPPROTO_ICMP
275 && skb_in->tail-(u8*)iph >= sizeof(struct icmphdr)) {
276 icmph = (struct icmphdr *)((char *)iph + (iph->ihl<<2));
278 if (skb_copy_bits(skb_in, skb_in->nh.iph->ihl*4,
279 icmph, sizeof(*icmph)) < 0)
282 /* Between echo-reply (0) and timestamp (13),
283 everything except echo-request (8) is an error.
284 Also, anything greater than NR_ICMP_TYPES is
285 unknown, and hence should be treated as an error... */
286 if ((icmph->type < ICMP_TIMESTAMP
287 && icmph->type != ICMP_ECHOREPLY
288 && icmph->type != ICMP_ECHO)
289 || icmph->type > NR_ICMP_TYPES)
294 if (!(rt->rt_flags & RTCF_LOCAL))
297 tos = (iph->tos & IPTOS_TOS_MASK) | IPTOS_PREC_INTERNETCONTROL;
303 .daddr = skb_in->nh.iph->saddr,
308 .proto = IPPROTO_ICMP,
311 .type = ICMP_DEST_UNREACH,
317 if (ip_route_output_key(&rt, &fl))
320 /* RFC says return as much as we can without exceeding 576 bytes. */
321 length = skb_in->len + sizeof(struct iphdr) + sizeof(struct icmphdr);
323 if (length > dst_pmtu(&rt->u.dst))
324 length = dst_pmtu(&rt->u.dst);
328 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
330 nskb = alloc_skb(hh_len + length, GFP_ATOMIC);
337 nskb->dst = &rt->u.dst;
338 skb_reserve(nskb, hh_len);
340 /* Set up IP header */
342 = (struct iphdr *)skb_put(nskb, sizeof(struct iphdr));
346 iph->tot_len = htons(length);
348 /* PMTU discovery never applies to ICMP packets. */
352 ip_select_ident(iph, &rt->u.dst, NULL);
353 iph->protocol=IPPROTO_ICMP;
354 iph->saddr=rt->rt_src;
355 iph->daddr=rt->rt_dst;
357 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
359 /* Set up ICMP header. */
360 icmph = nskb->h.icmph
361 = (struct icmphdr *)skb_put(nskb, sizeof(struct icmphdr));
362 icmph->type = ICMP_DEST_UNREACH;
364 icmph->un.gateway = 0;
367 /* Copy as much of original packet as will fit */
369 length - sizeof(struct iphdr) - sizeof(struct icmphdr));
371 skb_copy_bits(skb_in, 0, data,
372 length - sizeof(struct iphdr) - sizeof(struct icmphdr));
374 icmph->checksum = ip_compute_csum((unsigned char *)icmph,
375 length - sizeof(struct iphdr));
377 nf_ct_attach(nskb, skb_in);
379 NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
383 static unsigned int reject(struct sk_buff **pskb,
384 const struct net_device *in,
385 const struct net_device *out,
386 unsigned int hooknum,
387 const void *targinfo,
390 const struct ipt_reject_info *reject = targinfo;
392 /* Our naive response construction doesn't deal with IP
393 options, and probably shouldn't try. */
394 if ((*pskb)->nh.iph->ihl<<2 != sizeof(struct iphdr))
397 /* WARNING: This code causes reentry within iptables.
398 This means that the iptables jump stack is now crap. We
399 must return an absolute verdict. --RR */
400 switch (reject->with) {
401 case IPT_ICMP_NET_UNREACHABLE:
402 send_unreach(*pskb, ICMP_NET_UNREACH);
404 case IPT_ICMP_HOST_UNREACHABLE:
405 send_unreach(*pskb, ICMP_HOST_UNREACH);
407 case IPT_ICMP_PROT_UNREACHABLE:
408 send_unreach(*pskb, ICMP_PROT_UNREACH);
410 case IPT_ICMP_PORT_UNREACHABLE:
411 send_unreach(*pskb, ICMP_PORT_UNREACH);
413 case IPT_ICMP_NET_PROHIBITED:
414 send_unreach(*pskb, ICMP_NET_ANO);
416 case IPT_ICMP_HOST_PROHIBITED:
417 send_unreach(*pskb, ICMP_HOST_ANO);
419 case IPT_ICMP_ADMIN_PROHIBITED:
420 send_unreach(*pskb, ICMP_PKT_FILTERED);
423 send_reset(*pskb, hooknum);
424 case IPT_ICMP_ECHOREPLY:
425 /* Doesn't happen. */
432 static int check(const char *tablename,
433 const struct ipt_entry *e,
435 unsigned int targinfosize,
436 unsigned int hook_mask)
438 const struct ipt_reject_info *rejinfo = targinfo;
440 if (targinfosize != IPT_ALIGN(sizeof(struct ipt_reject_info))) {
441 DEBUGP("REJECT: targinfosize %u != 0\n", targinfosize);
445 /* Only allow these for packet filtering. */
446 if (strcmp(tablename, "filter") != 0) {
447 DEBUGP("REJECT: bad table `%s'.\n", tablename);
450 if ((hook_mask & ~((1 << NF_IP_LOCAL_IN)
451 | (1 << NF_IP_FORWARD)
452 | (1 << NF_IP_LOCAL_OUT))) != 0) {
453 DEBUGP("REJECT: bad hook mask %X\n", hook_mask);
457 if (rejinfo->with == IPT_ICMP_ECHOREPLY) {
458 printk("REJECT: ECHOREPLY no longer supported.\n");
460 } else if (rejinfo->with == IPT_TCP_RESET) {
461 /* Must specify that it's a TCP packet */
462 if (e->ip.proto != IPPROTO_TCP
463 || (e->ip.invflags & IPT_INV_PROTO)) {
464 DEBUGP("REJECT: TCP_RESET invalid for non-tcp\n");
472 static struct ipt_target ipt_reject_reg = {
479 static int __init init(void)
481 return ipt_register_target(&ipt_reject_reg);
484 static void __exit fini(void)
486 ipt_unregister_target(&ipt_reject_reg);