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 sk_buff *skb)
46 void (*attach)(struct sk_buff *, struct sk_buff *);
48 /* Avoid module unload race with ip_ct_attach being NULLed out */
49 if (skb->nfct && (attach = ip_ct_attach) != NULL) {
50 mb(); /* Just to be sure: must be read before executing this */
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, *oth, *tcph;
113 /* IP header checks: fragment. */
114 if (oldskb->nh.iph->frag_off & htons(IP_OFFSET))
117 oth = skb_header_pointer(oldskb, oldskb->nh.iph->ihl * 4,
118 sizeof(_otcph), &_otcph);
122 /* No RST for RST. */
126 /* FIXME: Check checksum --RR */
127 if ((rt = route_reverse(oldskb, hook)) == NULL)
130 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
132 /* We need a linear, writeable skb. We also need to expand
133 headroom in case hh_len of incoming interface < hh_len of
134 outgoing interface */
135 nskb = skb_copy_expand(oldskb, hh_len, skb_tailroom(oldskb),
138 dst_release(&rt->u.dst);
142 dst_release(nskb->dst);
143 nskb->dst = &rt->u.dst;
145 /* This packet will not be the same as the other: clear nf fields */
149 #ifdef CONFIG_BRIDGE_NETFILTER
150 nf_bridge_put(nskb->nf_bridge);
151 nskb->nf_bridge = NULL;
154 tcph = (struct tcphdr *)((u_int32_t*)nskb->nh.iph + nskb->nh.iph->ihl);
156 /* Swap source and dest */
157 tmp_addr = nskb->nh.iph->saddr;
158 nskb->nh.iph->saddr = nskb->nh.iph->daddr;
159 nskb->nh.iph->daddr = tmp_addr;
160 tmp_port = tcph->source;
161 tcph->source = tcph->dest;
162 tcph->dest = tmp_port;
164 /* Truncate to length (no data) */
165 tcph->doff = sizeof(struct tcphdr)/4;
166 skb_trim(nskb, nskb->nh.iph->ihl*4 + sizeof(struct tcphdr));
167 nskb->nh.iph->tot_len = htons(nskb->len);
171 tcph->seq = oth->ack_seq;
175 tcph->ack_seq = htonl(ntohl(oth->seq) + oth->syn + oth->fin
176 + oldskb->len - oldskb->nh.iph->ihl*4
182 ((u_int8_t *)tcph)[13] = 0;
184 tcph->ack = needs_ack;
189 /* Adjust TCP checksum */
191 tcph->check = tcp_v4_check(tcph, sizeof(struct tcphdr),
194 csum_partial((char *)tcph,
195 sizeof(struct tcphdr), 0));
197 /* Adjust IP TTL, DF */
198 nskb->nh.iph->ttl = MAXTTL;
200 nskb->nh.iph->frag_off = htons(IP_DF);
201 nskb->nh.iph->id = 0;
203 /* Adjust IP checksum */
204 nskb->nh.iph->check = 0;
205 nskb->nh.iph->check = ip_fast_csum((unsigned char *)nskb->nh.iph,
208 /* "Never happens" */
209 if (nskb->len > dst_pmtu(nskb->dst))
212 connection_attach(nskb, oldskb);
214 NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
222 static void send_unreach(struct sk_buff *skb_in, int code)
226 struct icmphdr *icmph;
227 struct sk_buff *nskb;
231 struct rtable *rt = (struct rtable*)skb_in->dst;
237 /* FIXME: Use sysctl number. --RR */
238 if (!xrlim_allow(&rt->u.dst, 1*HZ))
241 iph = skb_in->nh.iph;
243 /* No replies to physical multicast/broadcast */
244 if (skb_in->pkt_type!=PACKET_HOST)
247 /* Now check at the protocol level */
248 if (rt->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST))
251 /* Only reply to fragment 0. */
252 if (iph->frag_off&htons(IP_OFFSET))
255 /* Ensure we have at least 8 bytes of proto header. */
256 if (skb_in->len < skb_in->nh.iph->ihl*4 + 8)
259 /* if UDP checksum is set, verify it's correct */
260 if (iph->protocol == IPPROTO_UDP
261 && skb_in->tail-(u8*)iph >= sizeof(struct udphdr)) {
262 int datalen = skb_in->len - (iph->ihl<<2);
263 udph = (struct udphdr *)((char *)iph + (iph->ihl<<2));
265 && csum_tcpudp_magic(iph->saddr, iph->daddr,
266 datalen, IPPROTO_UDP,
267 csum_partial((char *)udph, datalen,
272 /* If we send an ICMP error to an ICMP error a mess would result.. */
273 if (iph->protocol == IPPROTO_ICMP
274 && skb_in->tail-(u8*)iph >= sizeof(struct icmphdr)) {
275 icmph = (struct icmphdr *)((char *)iph + (iph->ihl<<2));
277 if (skb_copy_bits(skb_in, skb_in->nh.iph->ihl*4,
278 icmph, sizeof(*icmph)) < 0)
281 /* Between echo-reply (0) and timestamp (13),
282 everything except echo-request (8) is an error.
283 Also, anything greater than NR_ICMP_TYPES is
284 unknown, and hence should be treated as an error... */
285 if ((icmph->type < ICMP_TIMESTAMP
286 && icmph->type != ICMP_ECHOREPLY
287 && icmph->type != ICMP_ECHO)
288 || icmph->type > NR_ICMP_TYPES)
293 if (!(rt->rt_flags & RTCF_LOCAL))
296 tos = (iph->tos & IPTOS_TOS_MASK) | IPTOS_PREC_INTERNETCONTROL;
299 struct flowi fl = { .nl_u = { .ip4_u =
300 { .daddr = skb_in->nh.iph->saddr,
302 .tos = RT_TOS(tos) } } };
303 if (ip_route_output_key(&rt, &fl))
306 /* RFC says return as much as we can without exceeding 576 bytes. */
307 length = skb_in->len + sizeof(struct iphdr) + sizeof(struct icmphdr);
309 if (length > dst_pmtu(&rt->u.dst))
310 length = dst_pmtu(&rt->u.dst);
314 hh_len = LL_RESERVED_SPACE(rt->u.dst.dev);
316 nskb = alloc_skb(hh_len + length, GFP_ATOMIC);
323 nskb->dst = &rt->u.dst;
324 skb_reserve(nskb, hh_len);
326 /* Set up IP header */
328 = (struct iphdr *)skb_put(nskb, sizeof(struct iphdr));
332 iph->tot_len = htons(length);
334 /* PMTU discovery never applies to ICMP packets. */
338 ip_select_ident(iph, &rt->u.dst, NULL);
339 iph->protocol=IPPROTO_ICMP;
340 iph->saddr=rt->rt_src;
341 iph->daddr=rt->rt_dst;
343 iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
345 /* Set up ICMP header. */
346 icmph = nskb->h.icmph
347 = (struct icmphdr *)skb_put(nskb, sizeof(struct icmphdr));
348 icmph->type = ICMP_DEST_UNREACH;
350 icmph->un.gateway = 0;
353 /* Copy as much of original packet as will fit */
355 length - sizeof(struct iphdr) - sizeof(struct icmphdr));
357 skb_copy_bits(skb_in, 0, data,
358 length - sizeof(struct iphdr) - sizeof(struct icmphdr));
360 icmph->checksum = ip_compute_csum((unsigned char *)icmph,
361 length - sizeof(struct iphdr));
363 connection_attach(nskb, skb_in);
365 NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
369 static unsigned int reject(struct sk_buff **pskb,
370 const struct net_device *in,
371 const struct net_device *out,
372 unsigned int hooknum,
373 const void *targinfo,
376 const struct ipt_reject_info *reject = targinfo;
378 /* Our naive response construction doesn't deal with IP
379 options, and probably shouldn't try. */
380 if ((*pskb)->nh.iph->ihl<<2 != sizeof(struct iphdr))
383 /* WARNING: This code causes reentry within iptables.
384 This means that the iptables jump stack is now crap. We
385 must return an absolute verdict. --RR */
386 switch (reject->with) {
387 case IPT_ICMP_NET_UNREACHABLE:
388 send_unreach(*pskb, ICMP_NET_UNREACH);
390 case IPT_ICMP_HOST_UNREACHABLE:
391 send_unreach(*pskb, ICMP_HOST_UNREACH);
393 case IPT_ICMP_PROT_UNREACHABLE:
394 send_unreach(*pskb, ICMP_PROT_UNREACH);
396 case IPT_ICMP_PORT_UNREACHABLE:
397 send_unreach(*pskb, ICMP_PORT_UNREACH);
399 case IPT_ICMP_NET_PROHIBITED:
400 send_unreach(*pskb, ICMP_NET_ANO);
402 case IPT_ICMP_HOST_PROHIBITED:
403 send_unreach(*pskb, ICMP_HOST_ANO);
405 case IPT_ICMP_ADMIN_PROHIBITED:
406 send_unreach(*pskb, ICMP_PKT_FILTERED);
409 send_reset(*pskb, hooknum);
410 case IPT_ICMP_ECHOREPLY:
411 /* Doesn't happen. */
418 static int check(const char *tablename,
419 const struct ipt_entry *e,
421 unsigned int targinfosize,
422 unsigned int hook_mask)
424 const struct ipt_reject_info *rejinfo = targinfo;
426 if (targinfosize != IPT_ALIGN(sizeof(struct ipt_reject_info))) {
427 DEBUGP("REJECT: targinfosize %u != 0\n", targinfosize);
431 /* Only allow these for packet filtering. */
432 if (strcmp(tablename, "filter") != 0) {
433 DEBUGP("REJECT: bad table `%s'.\n", tablename);
436 if ((hook_mask & ~((1 << NF_IP_LOCAL_IN)
437 | (1 << NF_IP_FORWARD)
438 | (1 << NF_IP_LOCAL_OUT))) != 0) {
439 DEBUGP("REJECT: bad hook mask %X\n", hook_mask);
443 if (rejinfo->with == IPT_ICMP_ECHOREPLY) {
444 printk("REJECT: ECHOREPLY no longer supported.\n");
446 } else if (rejinfo->with == IPT_TCP_RESET) {
447 /* Must specify that it's a TCP packet */
448 if (e->ip.proto != IPPROTO_TCP
449 || (e->ip.invflags & IPT_INV_PROTO)) {
450 DEBUGP("REJECT: TCP_RESET invalid for non-tcp\n");
458 static struct ipt_target ipt_reject_reg = {
465 static int __init init(void)
467 return ipt_register_target(&ipt_reject_reg);
470 static void __exit fini(void)
472 ipt_unregister_target(&ipt_reject_reg);