1/*
2 * This is a module which is used for rejecting packets.
3 * Added support for customized reject packets (Jozsef Kadlecsik).
4 */
5#include <linux/config.h>
6#include <linux/module.h>
7#include <linux/skbuff.h>
8#include <linux/ip.h>
9#include <linux/udp.h>
10#include <linux/icmp.h>
11#include <net/icmp.h>
12#include <net/ip.h>
13#include <net/tcp.h>
14struct in_device;
15#include <net/route.h>
16#include <linux/netfilter_ipv4/ip_tables.h>
17#include <linux/netfilter_ipv4/ipt_REJECT.h>
18
19#define DEBUGP(format, args...)
20
21/* If the original packet is part of a connection, but the connection
22   is not confirmed, our manufactured reply will not be associated
23   with it, so we need to do this manually. */
24static void connection_attach(struct sk_buff *new_skb, struct nf_ct_info *nfct)
25{
26	void (*attach)(struct sk_buff *, struct nf_ct_info *);
27
28	/* Avoid module unload race with ip_ct_attach being NULLed out */
29	if (nfct && (attach = ip_ct_attach) != NULL)
30		attach(new_skb, nfct);
31}
32
33/* Send RST reply */
34static void send_reset(struct sk_buff *oldskb, int local)
35{
36	struct sk_buff *nskb;
37	struct tcphdr *otcph, *tcph;
38	struct rtable *rt;
39	unsigned int otcplen;
40	u_int16_t tmp_port;
41	u_int32_t tmp_addr;
42	int needs_ack;
43
44	/* IP header checks: fragment, too short. */
45	if (oldskb->nh.iph->frag_off & htons(IP_OFFSET)
46	    || oldskb->len < (oldskb->nh.iph->ihl<<2) + sizeof(struct tcphdr))
47		return;
48
49	otcph = (struct tcphdr *)((u_int32_t*)oldskb->nh.iph + oldskb->nh.iph->ihl);
50	otcplen = oldskb->len - oldskb->nh.iph->ihl*4;
51
52	/* No RST for RST. */
53	if (otcph->rst)
54		return;
55
56	/* Check checksum. */
57	if (tcp_v4_check(otcph, otcplen, oldskb->nh.iph->saddr,
58			 oldskb->nh.iph->daddr,
59			 csum_partial((char *)otcph, otcplen, 0)) != 0)
60		return;
61
62	/* Copy skb (even if skb is about to be dropped, we can't just
63           clone it because there may be other things, such as tcpdump,
64           interested in it) */
65	nskb = skb_copy(oldskb, GFP_ATOMIC);
66	if (!nskb)
67		return;
68
69	/* This packet will not be the same as the other: clear nf fields */
70	nf_conntrack_put(nskb->nfct);
71	nskb->nfct = NULL;
72	nskb->nfcache = 0;
73#ifdef CONFIG_NETFILTER_DEBUG
74	nskb->nf_debug = 0;
75#endif
76	nskb->nfmark = 0;
77
78	tcph = (struct tcphdr *)((u_int32_t*)nskb->nh.iph + nskb->nh.iph->ihl);
79
80	/* Swap source and dest */
81	tmp_addr = nskb->nh.iph->saddr;
82	nskb->nh.iph->saddr = nskb->nh.iph->daddr;
83	nskb->nh.iph->daddr = tmp_addr;
84	tmp_port = tcph->source;
85	tcph->source = tcph->dest;
86	tcph->dest = tmp_port;
87
88	/* Truncate to length (no data) */
89	tcph->doff = sizeof(struct tcphdr)/4;
90	skb_trim(nskb, nskb->nh.iph->ihl*4 + sizeof(struct tcphdr));
91	nskb->nh.iph->tot_len = htons(nskb->len);
92
93	if (tcph->ack) {
94		needs_ack = 0;
95		tcph->seq = otcph->ack_seq;
96		tcph->ack_seq = 0;
97	} else {
98		needs_ack = 1;
99		tcph->ack_seq = htonl(ntohl(otcph->seq) + otcph->syn + otcph->fin
100				      + otcplen - (otcph->doff<<2));
101		tcph->seq = 0;
102	}
103
104	/* Reset flags */
105	((u_int8_t *)tcph)[13] = 0;
106	tcph->rst = 1;
107	tcph->ack = needs_ack;
108
109	tcph->window = 0;
110	tcph->urg_ptr = 0;
111
112	/* Adjust TCP checksum */
113	tcph->check = 0;
114	tcph->check = tcp_v4_check(tcph, sizeof(struct tcphdr),
115				   nskb->nh.iph->saddr,
116				   nskb->nh.iph->daddr,
117				   csum_partial((char *)tcph,
118						sizeof(struct tcphdr), 0));
119
120	/* Adjust IP TTL, DF */
121	nskb->nh.iph->ttl = MAXTTL;
122	/* Set DF, id = 0 */
123	nskb->nh.iph->frag_off = htons(IP_DF);
124	nskb->nh.iph->id = 0;
125
126	/* Adjust IP checksum */
127	nskb->nh.iph->check = 0;
128	nskb->nh.iph->check = ip_fast_csum((unsigned char *)nskb->nh.iph,
129					   nskb->nh.iph->ihl);
130
131	/* Routing: if not headed for us, route won't like source */
132	if (ip_route_output(&rt, nskb->nh.iph->daddr,
133			    local ? nskb->nh.iph->saddr : 0,
134			    RT_TOS(nskb->nh.iph->tos) | RTO_CONN,
135			    0) != 0)
136		goto free_nskb;
137
138	dst_release(nskb->dst);
139	nskb->dst = &rt->u.dst;
140
141	/* "Never happens" */
142	if (nskb->len > nskb->dst->pmtu)
143		goto free_nskb;
144
145	connection_attach(nskb, oldskb->nfct);
146
147	NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
148		ip_finish_output);
149	return;
150
151 free_nskb:
152	kfree_skb(nskb);
153}
154
155static void send_unreach(struct sk_buff *skb_in, int code)
156{
157	struct iphdr *iph;
158	struct udphdr *udph;
159	struct icmphdr *icmph;
160	struct sk_buff *nskb;
161	u32 saddr;
162	u8 tos;
163	int hh_len, length;
164	struct rtable *rt = (struct rtable*)skb_in->dst;
165	unsigned char *data;
166
167	if (!rt)
168		return;
169
170	if (!xrlim_allow(&rt->u.dst, 1*HZ))
171		return;
172
173	iph = skb_in->nh.iph;
174
175	/* No replies to physical multicast/broadcast */
176	if (skb_in->pkt_type!=PACKET_HOST)
177		return;
178
179	/* Now check at the protocol level */
180	if (rt->rt_flags&(RTCF_BROADCAST|RTCF_MULTICAST))
181		return;
182
183	/* Only reply to fragment 0. */
184	if (iph->frag_off&htons(IP_OFFSET))
185		return;
186
187	/* if UDP checksum is set, verify it's correct */
188	if (iph->protocol == IPPROTO_UDP
189	    && skb_in->tail-(u8*)iph >= sizeof(struct udphdr)) {
190		int datalen = skb_in->len - (iph->ihl<<2);
191		udph = (struct udphdr *)((char *)iph + (iph->ihl<<2));
192		if (udph->check
193		    && csum_tcpudp_magic(iph->saddr, iph->daddr,
194		                         datalen, IPPROTO_UDP,
195		                         csum_partial((char *)udph, datalen,
196		                                      0)) != 0)
197			return;
198	}
199
200	/* If we send an ICMP error to an ICMP error a mess would result.. */
201	if (iph->protocol == IPPROTO_ICMP
202	    && skb_in->tail-(u8*)iph >= sizeof(struct icmphdr)) {
203		icmph = (struct icmphdr *)((char *)iph + (iph->ihl<<2));
204		/* Between echo-reply (0) and timestamp (13),
205		   everything except echo-request (8) is an error.
206		   Also, anything greater than NR_ICMP_TYPES is
207		   unknown, and hence should be treated as an error... */
208		if ((icmph->type < ICMP_TIMESTAMP
209		     && icmph->type != ICMP_ECHOREPLY
210		     && icmph->type != ICMP_ECHO)
211		    || icmph->type > NR_ICMP_TYPES)
212			return;
213	}
214
215	saddr = iph->daddr;
216	if (!(rt->rt_flags & RTCF_LOCAL))
217		saddr = 0;
218
219	tos = (iph->tos & IPTOS_TOS_MASK) | IPTOS_PREC_INTERNETCONTROL;
220
221	if (ip_route_output(&rt, iph->saddr, saddr, RT_TOS(tos), 0))
222		return;
223
224	/* RFC says return as much as we can without exceeding 576 bytes. */
225	length = skb_in->len + sizeof(struct iphdr) + sizeof(struct icmphdr);
226
227	if (length > rt->u.dst.pmtu)
228		length = rt->u.dst.pmtu;
229	if (length > 576)
230		length = 576;
231
232	hh_len = (rt->u.dst.dev->hard_header_len + 15)&~15;
233
234	nskb = alloc_skb(hh_len+15+length, GFP_ATOMIC);
235	if (!nskb) {
236		ip_rt_put(rt);
237		return;
238	}
239
240	nskb->priority = 0;
241	nskb->dst = &rt->u.dst;
242	skb_reserve(nskb, hh_len);
243
244	/* Set up IP header */
245	iph = nskb->nh.iph
246		= (struct iphdr *)skb_put(nskb, sizeof(struct iphdr));
247	iph->version=4;
248	iph->ihl=5;
249	iph->tos=tos;
250	iph->tot_len = htons(length);
251
252	/* PMTU discovery never applies to ICMP packets. */
253	iph->frag_off = 0;
254
255	iph->ttl = MAXTTL;
256	ip_select_ident(iph, &rt->u.dst, NULL);
257	iph->protocol=IPPROTO_ICMP;
258	iph->saddr=rt->rt_src;
259	iph->daddr=rt->rt_dst;
260	iph->check=0;
261	iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
262
263	/* Set up ICMP header. */
264	icmph = nskb->h.icmph
265		= (struct icmphdr *)skb_put(nskb, sizeof(struct icmphdr));
266	icmph->type = ICMP_DEST_UNREACH;
267	icmph->code = code;
268	icmph->un.gateway = 0;
269	icmph->checksum = 0;
270
271	/* Copy as much of original packet as will fit */
272	data = skb_put(nskb,
273		       length - sizeof(struct iphdr) - sizeof(struct icmphdr));
274	memcpy(data, skb_in->nh.iph,
275	       length - sizeof(struct iphdr) - sizeof(struct icmphdr));
276	icmph->checksum = ip_compute_csum((unsigned char *)icmph,
277					  length - sizeof(struct iphdr));
278
279	connection_attach(nskb, skb_in->nfct);
280
281	NF_HOOK(PF_INET, NF_IP_LOCAL_OUT, nskb, NULL, nskb->dst->dev,
282		ip_finish_output);
283}
284
285static unsigned int reject(struct sk_buff **pskb,
286			   unsigned int hooknum,
287			   const struct net_device *in,
288			   const struct net_device *out,
289			   const void *targinfo,
290			   void *userinfo)
291{
292	const struct ipt_reject_info *reject = targinfo;
293
294	/* Our naive response construction doesn't deal with IP
295           options, and probably shouldn't try. */
296	if ((*pskb)->nh.iph->ihl<<2 != sizeof(struct iphdr))
297		return NF_DROP;
298
299	/* WARNING: This code causes reentry within iptables.
300	   This means that the iptables jump stack is now crap.  We
301	   must return an absolute verdict. --RR */
302    	switch (reject->with) {
303    	case IPT_ICMP_NET_UNREACHABLE:
304    		send_unreach(*pskb, ICMP_NET_UNREACH);
305    		break;
306    	case IPT_ICMP_HOST_UNREACHABLE:
307    		send_unreach(*pskb, ICMP_HOST_UNREACH);
308    		break;
309    	case IPT_ICMP_PROT_UNREACHABLE:
310    		send_unreach(*pskb, ICMP_PROT_UNREACH);
311    		break;
312    	case IPT_ICMP_PORT_UNREACHABLE:
313    		send_unreach(*pskb, ICMP_PORT_UNREACH);
314    		break;
315    	case IPT_ICMP_NET_PROHIBITED:
316    		send_unreach(*pskb, ICMP_NET_ANO);
317    		break;
318	case IPT_ICMP_HOST_PROHIBITED:
319    		send_unreach(*pskb, ICMP_HOST_ANO);
320    		break;
321	case IPT_TCP_RESET:
322		send_reset(*pskb, hooknum == NF_IP_LOCAL_IN);
323	case IPT_ICMP_ECHOREPLY:
324		/* Doesn't happen. */
325		break;
326	}
327
328	return NF_DROP;
329}
330
331static int check(const char *tablename,
332		 const struct ipt_entry *e,
333		 void *targinfo,
334		 unsigned int targinfosize,
335		 unsigned int hook_mask)
336{
337 	const struct ipt_reject_info *rejinfo = targinfo;
338
339 	if (targinfosize != IPT_ALIGN(sizeof(struct ipt_reject_info))) {
340  		DEBUGP("REJECT: targinfosize %u != 0\n", targinfosize);
341  		return 0;
342  	}
343
344	/* Only allow these for packet filtering. */
345	if (strcmp(tablename, "filter") != 0) {
346		DEBUGP("REJECT: bad table `%s'.\n", tablename);
347		return 0;
348	}
349	if ((hook_mask & ~((1 << NF_IP_LOCAL_IN)
350			   | (1 << NF_IP_FORWARD)
351			   | (1 << NF_IP_LOCAL_OUT))) != 0) {
352		DEBUGP("REJECT: bad hook mask %X\n", hook_mask);
353		return 0;
354	}
355
356	if (rejinfo->with == IPT_ICMP_ECHOREPLY) {
357		printk("REJECT: ECHOREPLY no longer supported.\n");
358		return 0;
359	} else if (rejinfo->with == IPT_TCP_RESET) {
360		/* Must specify that it's a TCP packet */
361		if (e->ip.proto != IPPROTO_TCP
362		    || (e->ip.invflags & IPT_INV_PROTO)) {
363			DEBUGP("REJECT: TCP_RESET illegal for non-tcp\n");
364			return 0;
365		}
366	}
367
368	return 1;
369}
370
371static struct ipt_target ipt_reject_reg
372= { { NULL, NULL }, "REJECT", reject, check, NULL, THIS_MODULE };
373
374static int __init init(void)
375{
376	if (ipt_register_target(&ipt_reject_reg))
377		return -EINVAL;
378	return 0;
379}
380
381static void __exit fini(void)
382{
383	ipt_unregister_target(&ipt_reject_reg);
384}
385
386module_init(init);
387module_exit(fini);
388MODULE_LICENSE("GPL");
389