1 /*
2  * Copyright (c) 2014 Pablo Neira Ayuso <pablo@netfilter.org>
3  *
4  * This program is free software; you can redistribute it and/or modify
5  * it under the terms of the GNU General Public License version 2 as
6  * published by the Free Software Foundation.
7  */
8 
9 #include <linux/kernel.h>
10 #include <linux/init.h>
11 #include <linux/module.h>
12 #include <linux/netlink.h>
13 #include <linux/netfilter.h>
14 #include <linux/netfilter/nf_tables.h>
15 #include <net/netfilter/nf_tables.h>
16 #include <net/netfilter/nft_reject.h>
17 #include <net/netfilter/nf_tables_bridge.h>
18 #include <net/netfilter/ipv4/nf_reject.h>
19 #include <net/netfilter/ipv6/nf_reject.h>
20 #include <linux/ip.h>
21 #include <net/ip.h>
22 #include <net/ip6_checksum.h>
23 #include <linux/netfilter_bridge.h>
24 #include <linux/netfilter_ipv6.h>
25 #include "../br_private.h"
26 
27 static void nft_reject_br_push_etherhdr(struct sk_buff *oldskb,
28 					struct sk_buff *nskb)
29 {
30 	struct ethhdr *eth;
31 
32 	eth = (struct ethhdr *)skb_push(nskb, ETH_HLEN);
33 	skb_reset_mac_header(nskb);
34 	ether_addr_copy(eth->h_source, eth_hdr(oldskb)->h_dest);
35 	ether_addr_copy(eth->h_dest, eth_hdr(oldskb)->h_source);
36 	eth->h_proto = eth_hdr(oldskb)->h_proto;
37 	skb_pull(nskb, ETH_HLEN);
38 }
39 
40 /* We cannot use oldskb->dev, it can be either bridge device (NF_BRIDGE INPUT)
41  * or the bridge port (NF_BRIDGE PREROUTING).
42  */
43 static void nft_reject_br_send_v4_tcp_reset(struct net *net,
44 					    struct sk_buff *oldskb,
45 					    const struct net_device *dev,
46 					    int hook)
47 {
48 	struct sk_buff *nskb;
49 	struct iphdr *niph;
50 	const struct tcphdr *oth;
51 	struct tcphdr _oth;
52 
53 	if (!nft_bridge_iphdr_validate(oldskb))
54 		return;
55 
56 	oth = nf_reject_ip_tcphdr_get(oldskb, &_oth, hook);
57 	if (!oth)
58 		return;
59 
60 	nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct tcphdr) +
61 			 LL_MAX_HEADER, GFP_ATOMIC);
62 	if (!nskb)
63 		return;
64 
65 	skb_reserve(nskb, LL_MAX_HEADER);
66 	niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_TCP,
67 				   net->ipv4.sysctl_ip_default_ttl);
68 	nf_reject_ip_tcphdr_put(nskb, oldskb, oth);
69 	niph->ttl	= net->ipv4.sysctl_ip_default_ttl;
70 	niph->tot_len	= htons(nskb->len);
71 	ip_send_check(niph);
72 
73 	nft_reject_br_push_etherhdr(oldskb, nskb);
74 
75 	br_forward(br_port_get_rcu(dev), nskb, false, true);
76 }
77 
78 static void nft_reject_br_send_v4_unreach(struct net *net,
79 					  struct sk_buff *oldskb,
80 					  const struct net_device *dev,
81 					  int hook, u8 code)
82 {
83 	struct sk_buff *nskb;
84 	struct iphdr *niph;
85 	struct icmphdr *icmph;
86 	unsigned int len;
87 	void *payload;
88 	__wsum csum;
89 	u8 proto;
90 
91 	if (oldskb->csum_bad || !nft_bridge_iphdr_validate(oldskb))
92 		return;
93 
94 	/* IP header checks: fragment. */
95 	if (ip_hdr(oldskb)->frag_off & htons(IP_OFFSET))
96 		return;
97 
98 	/* RFC says return as much as we can without exceeding 576 bytes. */
99 	len = min_t(unsigned int, 536, oldskb->len);
100 
101 	if (!pskb_may_pull(oldskb, len))
102 		return;
103 
104 	if (pskb_trim_rcsum(oldskb, ntohs(ip_hdr(oldskb)->tot_len)))
105 		return;
106 
107 	if (ip_hdr(oldskb)->protocol == IPPROTO_TCP ||
108 	    ip_hdr(oldskb)->protocol == IPPROTO_UDP)
109 		proto = ip_hdr(oldskb)->protocol;
110 	else
111 		proto = 0;
112 
113 	if (!skb_csum_unnecessary(oldskb) &&
114 	    nf_ip_checksum(oldskb, hook, ip_hdrlen(oldskb), proto))
115 		return;
116 
117 	nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct icmphdr) +
118 			 LL_MAX_HEADER + len, GFP_ATOMIC);
119 	if (!nskb)
120 		return;
121 
122 	skb_reserve(nskb, LL_MAX_HEADER);
123 	niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_ICMP,
124 				   net->ipv4.sysctl_ip_default_ttl);
125 
126 	skb_reset_transport_header(nskb);
127 	icmph = (struct icmphdr *)skb_put(nskb, sizeof(struct icmphdr));
128 	memset(icmph, 0, sizeof(*icmph));
129 	icmph->type     = ICMP_DEST_UNREACH;
130 	icmph->code	= code;
131 
132 	payload = skb_put(nskb, len);
133 	memcpy(payload, skb_network_header(oldskb), len);
134 
135 	csum = csum_partial((void *)icmph, len + sizeof(struct icmphdr), 0);
136 	icmph->checksum = csum_fold(csum);
137 
138 	niph->tot_len	= htons(nskb->len);
139 	ip_send_check(niph);
140 
141 	nft_reject_br_push_etherhdr(oldskb, nskb);
142 
143 	br_forward(br_port_get_rcu(dev), nskb, false, true);
144 }
145 
146 static void nft_reject_br_send_v6_tcp_reset(struct net *net,
147 					    struct sk_buff *oldskb,
148 					    const struct net_device *dev,
149 					    int hook)
150 {
151 	struct sk_buff *nskb;
152 	const struct tcphdr *oth;
153 	struct tcphdr _oth;
154 	unsigned int otcplen;
155 	struct ipv6hdr *nip6h;
156 
157 	if (!nft_bridge_ip6hdr_validate(oldskb))
158 		return;
159 
160 	oth = nf_reject_ip6_tcphdr_get(oldskb, &_oth, &otcplen, hook);
161 	if (!oth)
162 		return;
163 
164 	nskb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(struct tcphdr) +
165 			 LL_MAX_HEADER, GFP_ATOMIC);
166 	if (!nskb)
167 		return;
168 
169 	skb_reserve(nskb, LL_MAX_HEADER);
170 	nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_TCP,
171 				     net->ipv6.devconf_all->hop_limit);
172 	nf_reject_ip6_tcphdr_put(nskb, oldskb, oth, otcplen);
173 	nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
174 
175 	nft_reject_br_push_etherhdr(oldskb, nskb);
176 
177 	br_forward(br_port_get_rcu(dev), nskb, false, true);
178 }
179 
180 static bool reject6_br_csum_ok(struct sk_buff *skb, int hook)
181 {
182 	const struct ipv6hdr *ip6h = ipv6_hdr(skb);
183 	int thoff;
184 	__be16 fo;
185 	u8 proto = ip6h->nexthdr;
186 
187 	if (skb->csum_bad)
188 		return false;
189 
190 	if (skb_csum_unnecessary(skb))
191 		return true;
192 
193 	if (ip6h->payload_len &&
194 	    pskb_trim_rcsum(skb, ntohs(ip6h->payload_len) + sizeof(*ip6h)))
195 		return false;
196 
197 	thoff = ipv6_skip_exthdr(skb, ((u8*)(ip6h+1) - skb->data), &proto, &fo);
198 	if (thoff < 0 || thoff >= skb->len || (fo & htons(~0x7)) != 0)
199 		return false;
200 
201 	return nf_ip6_checksum(skb, hook, thoff, proto) == 0;
202 }
203 
204 static void nft_reject_br_send_v6_unreach(struct net *net,
205 					  struct sk_buff *oldskb,
206 					  const struct net_device *dev,
207 					  int hook, u8 code)
208 {
209 	struct sk_buff *nskb;
210 	struct ipv6hdr *nip6h;
211 	struct icmp6hdr *icmp6h;
212 	unsigned int len;
213 	void *payload;
214 
215 	if (!nft_bridge_ip6hdr_validate(oldskb))
216 		return;
217 
218 	/* Include "As much of invoking packet as possible without the ICMPv6
219 	 * packet exceeding the minimum IPv6 MTU" in the ICMP payload.
220 	 */
221 	len = min_t(unsigned int, 1220, oldskb->len);
222 
223 	if (!pskb_may_pull(oldskb, len))
224 		return;
225 
226 	if (!reject6_br_csum_ok(oldskb, hook))
227 		return;
228 
229 	nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct icmp6hdr) +
230 			 LL_MAX_HEADER + len, GFP_ATOMIC);
231 	if (!nskb)
232 		return;
233 
234 	skb_reserve(nskb, LL_MAX_HEADER);
235 	nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_ICMPV6,
236 				     net->ipv6.devconf_all->hop_limit);
237 
238 	skb_reset_transport_header(nskb);
239 	icmp6h = (struct icmp6hdr *)skb_put(nskb, sizeof(struct icmp6hdr));
240 	memset(icmp6h, 0, sizeof(*icmp6h));
241 	icmp6h->icmp6_type = ICMPV6_DEST_UNREACH;
242 	icmp6h->icmp6_code = code;
243 
244 	payload = skb_put(nskb, len);
245 	memcpy(payload, skb_network_header(oldskb), len);
246 	nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
247 
248 	icmp6h->icmp6_cksum =
249 		csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr,
250 				nskb->len - sizeof(struct ipv6hdr),
251 				IPPROTO_ICMPV6,
252 				csum_partial(icmp6h,
253 					     nskb->len - sizeof(struct ipv6hdr),
254 					     0));
255 
256 	nft_reject_br_push_etherhdr(oldskb, nskb);
257 
258 	br_forward(br_port_get_rcu(dev), nskb, false, true);
259 }
260 
261 static void nft_reject_bridge_eval(const struct nft_expr *expr,
262 				   struct nft_regs *regs,
263 				   const struct nft_pktinfo *pkt)
264 {
265 	struct nft_reject *priv = nft_expr_priv(expr);
266 	const unsigned char *dest = eth_hdr(pkt->skb)->h_dest;
267 
268 	if (is_broadcast_ether_addr(dest) ||
269 	    is_multicast_ether_addr(dest))
270 		goto out;
271 
272 	switch (eth_hdr(pkt->skb)->h_proto) {
273 	case htons(ETH_P_IP):
274 		switch (priv->type) {
275 		case NFT_REJECT_ICMP_UNREACH:
276 			nft_reject_br_send_v4_unreach(pkt->net, pkt->skb,
277 						      pkt->in, pkt->hook,
278 						      priv->icmp_code);
279 			break;
280 		case NFT_REJECT_TCP_RST:
281 			nft_reject_br_send_v4_tcp_reset(pkt->net, pkt->skb,
282 							pkt->in, pkt->hook);
283 			break;
284 		case NFT_REJECT_ICMPX_UNREACH:
285 			nft_reject_br_send_v4_unreach(pkt->net, pkt->skb,
286 						      pkt->in, pkt->hook,
287 						      nft_reject_icmp_code(priv->icmp_code));
288 			break;
289 		}
290 		break;
291 	case htons(ETH_P_IPV6):
292 		switch (priv->type) {
293 		case NFT_REJECT_ICMP_UNREACH:
294 			nft_reject_br_send_v6_unreach(pkt->net, pkt->skb,
295 						      pkt->in, pkt->hook,
296 						      priv->icmp_code);
297 			break;
298 		case NFT_REJECT_TCP_RST:
299 			nft_reject_br_send_v6_tcp_reset(pkt->net, pkt->skb,
300 							pkt->in, pkt->hook);
301 			break;
302 		case NFT_REJECT_ICMPX_UNREACH:
303 			nft_reject_br_send_v6_unreach(pkt->net, pkt->skb,
304 						      pkt->in, pkt->hook,
305 						      nft_reject_icmpv6_code(priv->icmp_code));
306 			break;
307 		}
308 		break;
309 	default:
310 		/* No explicit way to reject this protocol, drop it. */
311 		break;
312 	}
313 out:
314 	regs->verdict.code = NF_DROP;
315 }
316 
317 static int nft_reject_bridge_validate(const struct nft_ctx *ctx,
318 				      const struct nft_expr *expr,
319 				      const struct nft_data **data)
320 {
321 	return nft_chain_validate_hooks(ctx->chain, (1 << NF_BR_PRE_ROUTING) |
322 						    (1 << NF_BR_LOCAL_IN));
323 }
324 
325 static int nft_reject_bridge_init(const struct nft_ctx *ctx,
326 				  const struct nft_expr *expr,
327 				  const struct nlattr * const tb[])
328 {
329 	struct nft_reject *priv = nft_expr_priv(expr);
330 	int icmp_code, err;
331 
332 	err = nft_reject_bridge_validate(ctx, expr, NULL);
333 	if (err < 0)
334 		return err;
335 
336 	if (tb[NFTA_REJECT_TYPE] == NULL)
337 		return -EINVAL;
338 
339 	priv->type = ntohl(nla_get_be32(tb[NFTA_REJECT_TYPE]));
340 	switch (priv->type) {
341 	case NFT_REJECT_ICMP_UNREACH:
342 	case NFT_REJECT_ICMPX_UNREACH:
343 		if (tb[NFTA_REJECT_ICMP_CODE] == NULL)
344 			return -EINVAL;
345 
346 		icmp_code = nla_get_u8(tb[NFTA_REJECT_ICMP_CODE]);
347 		if (priv->type == NFT_REJECT_ICMPX_UNREACH &&
348 		    icmp_code > NFT_REJECT_ICMPX_MAX)
349 			return -EINVAL;
350 
351 		priv->icmp_code = icmp_code;
352 		break;
353 	case NFT_REJECT_TCP_RST:
354 		break;
355 	default:
356 		return -EINVAL;
357 	}
358 	return 0;
359 }
360 
361 static int nft_reject_bridge_dump(struct sk_buff *skb,
362 				  const struct nft_expr *expr)
363 {
364 	const struct nft_reject *priv = nft_expr_priv(expr);
365 
366 	if (nla_put_be32(skb, NFTA_REJECT_TYPE, htonl(priv->type)))
367 		goto nla_put_failure;
368 
369 	switch (priv->type) {
370 	case NFT_REJECT_ICMP_UNREACH:
371 	case NFT_REJECT_ICMPX_UNREACH:
372 		if (nla_put_u8(skb, NFTA_REJECT_ICMP_CODE, priv->icmp_code))
373 			goto nla_put_failure;
374 		break;
375 	default:
376 		break;
377 	}
378 
379 	return 0;
380 
381 nla_put_failure:
382 	return -1;
383 }
384 
385 static struct nft_expr_type nft_reject_bridge_type;
386 static const struct nft_expr_ops nft_reject_bridge_ops = {
387 	.type		= &nft_reject_bridge_type,
388 	.size		= NFT_EXPR_SIZE(sizeof(struct nft_reject)),
389 	.eval		= nft_reject_bridge_eval,
390 	.init		= nft_reject_bridge_init,
391 	.dump		= nft_reject_bridge_dump,
392 	.validate	= nft_reject_bridge_validate,
393 };
394 
395 static struct nft_expr_type nft_reject_bridge_type __read_mostly = {
396 	.family		= NFPROTO_BRIDGE,
397 	.name		= "reject",
398 	.ops		= &nft_reject_bridge_ops,
399 	.policy		= nft_reject_policy,
400 	.maxattr	= NFTA_REJECT_MAX,
401 	.owner		= THIS_MODULE,
402 };
403 
404 static int __init nft_reject_bridge_module_init(void)
405 {
406 	return nft_register_expr(&nft_reject_bridge_type);
407 }
408 
409 static void __exit nft_reject_bridge_module_exit(void)
410 {
411 	nft_unregister_expr(&nft_reject_bridge_type);
412 }
413 
414 module_init(nft_reject_bridge_module_init);
415 module_exit(nft_reject_bridge_module_exit);
416 
417 MODULE_LICENSE("GPL");
418 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
419 MODULE_ALIAS_NFT_AF_EXPR(AF_BRIDGE, "reject");
420