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/ipv4/nf_reject.h>
18 #include <net/netfilter/ipv6/nf_reject.h>
19 #include <linux/ip.h>
20 #include <net/ip.h>
21 #include <net/ip6_checksum.h>
22 #include <linux/netfilter_bridge.h>
23 #include <linux/netfilter_ipv6.h>
24 #include "../br_private.h"
25
26 static void nft_reject_br_push_etherhdr(struct sk_buff *oldskb,
27 struct sk_buff *nskb)
28 {
29 struct ethhdr *eth;
30
31 eth = skb_push(nskb, ETH_HLEN);
32 skb_reset_mac_header(nskb);
33 ether_addr_copy(eth->h_source, eth_hdr(oldskb)->h_dest);
34 ether_addr_copy(eth->h_dest, eth_hdr(oldskb)->h_source);
35 eth->h_proto = eth_hdr(oldskb)->h_proto;
36 skb_pull(nskb, ETH_HLEN);
37 }
38
39 static int nft_bridge_iphdr_validate(struct sk_buff *skb)
40 {
41 struct iphdr *iph;
42 u32 len;
43
44 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
45 return 0;
46
47 iph = ip_hdr(skb);
48 if (iph->ihl < 5 || iph->version != 4)
49 return 0;
50
51 len = ntohs(iph->tot_len);
52 if (skb->len < len)
53 return 0;
54 else if (len < (iph->ihl*4))
55 return 0;
56
57 if (!pskb_may_pull(skb, iph->ihl*4))
58 return 0;
59
60 return 1;
61 }
62
63 /* We cannot use oldskb->dev, it can be either bridge device (NF_BRIDGE INPUT)
64 * or the bridge port (NF_BRIDGE PREROUTING).
65 */
66 static void nft_reject_br_send_v4_tcp_reset(struct net *net,
67 struct sk_buff *oldskb,
68 const struct net_device *dev,
69 int hook)
70 {
71 struct sk_buff *nskb;
72 struct iphdr *niph;
73 const struct tcphdr *oth;
74 struct tcphdr _oth;
75
76 if (!nft_bridge_iphdr_validate(oldskb))
77 return;
78
79 oth = nf_reject_ip_tcphdr_get(oldskb, &_oth, hook);
80 if (!oth)
81 return;
82
83 nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct tcphdr) +
84 LL_MAX_HEADER, GFP_ATOMIC);
85 if (!nskb)
86 return;
87
88 skb_reserve(nskb, LL_MAX_HEADER);
89 niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_TCP,
90 net->ipv4.sysctl_ip_default_ttl);
91 nf_reject_ip_tcphdr_put(nskb, oldskb, oth);
92 niph->ttl = net->ipv4.sysctl_ip_default_ttl;
93 niph->tot_len = htons(nskb->len);
94 ip_send_check(niph);
95
96 nft_reject_br_push_etherhdr(oldskb, nskb);
97
98 br_forward(br_port_get_rcu(dev), nskb, false, true);
99 }
100
101 static void nft_reject_br_send_v4_unreach(struct net *net,
102 struct sk_buff *oldskb,
103 const struct net_device *dev,
104 int hook, u8 code)
105 {
106 struct sk_buff *nskb;
107 struct iphdr *niph;
108 struct icmphdr *icmph;
109 unsigned int len;
110 __wsum csum;
111 u8 proto;
112
113 if (!nft_bridge_iphdr_validate(oldskb))
114 return;
115
116 /* IP header checks: fragment. */
117 if (ip_hdr(oldskb)->frag_off & htons(IP_OFFSET))
118 return;
119
120 /* RFC says return as much as we can without exceeding 576 bytes. */
121 len = min_t(unsigned int, 536, oldskb->len);
122
123 if (!pskb_may_pull(oldskb, len))
124 return;
125
126 if (pskb_trim_rcsum(oldskb, ntohs(ip_hdr(oldskb)->tot_len)))
127 return;
128
129 if (ip_hdr(oldskb)->protocol == IPPROTO_TCP ||
130 ip_hdr(oldskb)->protocol == IPPROTO_UDP)
131 proto = ip_hdr(oldskb)->protocol;
132 else
133 proto = 0;
134
135 if (!skb_csum_unnecessary(oldskb) &&
136 nf_ip_checksum(oldskb, hook, ip_hdrlen(oldskb), proto))
137 return;
138
139 nskb = alloc_skb(sizeof(struct iphdr) + sizeof(struct icmphdr) +
140 LL_MAX_HEADER + len, GFP_ATOMIC);
141 if (!nskb)
142 return;
143
144 skb_reserve(nskb, LL_MAX_HEADER);
145 niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_ICMP,
146 net->ipv4.sysctl_ip_default_ttl);
147
148 skb_reset_transport_header(nskb);
149 icmph = skb_put_zero(nskb, sizeof(struct icmphdr));
150 icmph->type = ICMP_DEST_UNREACH;
151 icmph->code = code;
152
153 skb_put_data(nskb, skb_network_header(oldskb), len);
154
155 csum = csum_partial((void *)icmph, len + sizeof(struct icmphdr), 0);
156 icmph->checksum = csum_fold(csum);
157
158 niph->tot_len = htons(nskb->len);
159 ip_send_check(niph);
160
161 nft_reject_br_push_etherhdr(oldskb, nskb);
162
163 br_forward(br_port_get_rcu(dev), nskb, false, true);
164 }
165
166 static int nft_bridge_ip6hdr_validate(struct sk_buff *skb)
167 {
168 struct ipv6hdr *hdr;
169 u32 pkt_len;
170
171 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
172 return 0;
173
174 hdr = ipv6_hdr(skb);
175 if (hdr->version != 6)
176 return 0;
177
178 pkt_len = ntohs(hdr->payload_len);
179 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
180 return 0;
181
182 return 1;
183 }
184
185 static void nft_reject_br_send_v6_tcp_reset(struct net *net,
186 struct sk_buff *oldskb,
187 const struct net_device *dev,
188 int hook)
189 {
190 struct sk_buff *nskb;
191 const struct tcphdr *oth;
192 struct tcphdr _oth;
193 unsigned int otcplen;
194 struct ipv6hdr *nip6h;
195
196 if (!nft_bridge_ip6hdr_validate(oldskb))
197 return;
198
199 oth = nf_reject_ip6_tcphdr_get(oldskb, &_oth, &otcplen, hook);
200 if (!oth)
201 return;
202
203 nskb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(struct tcphdr) +
204 LL_MAX_HEADER, GFP_ATOMIC);
205 if (!nskb)
206 return;
207
208 skb_reserve(nskb, LL_MAX_HEADER);
209 nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_TCP,
210 net->ipv6.devconf_all->hop_limit);
211 nf_reject_ip6_tcphdr_put(nskb, oldskb, oth, otcplen);
212 nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
213
214 nft_reject_br_push_etherhdr(oldskb, nskb);
215
216 br_forward(br_port_get_rcu(dev), nskb, false, true);
217 }
218
219 static bool reject6_br_csum_ok(struct sk_buff *skb, int hook)
220 {
221 const struct ipv6hdr *ip6h = ipv6_hdr(skb);
222 int thoff;
223 __be16 fo;
224 u8 proto = ip6h->nexthdr;
225
226 if (skb_csum_unnecessary(skb))
227 return true;
228
229 if (ip6h->payload_len &&
230 pskb_trim_rcsum(skb, ntohs(ip6h->payload_len) + sizeof(*ip6h)))
231 return false;
232
233 thoff = ipv6_skip_exthdr(skb, ((u8*)(ip6h+1) - skb->data), &proto, &fo);
234 if (thoff < 0 || thoff >= skb->len || (fo & htons(~0x7)) != 0)
235 return false;
236
237 return nf_ip6_checksum(skb, hook, thoff, proto) == 0;
238 }
239
240 static void nft_reject_br_send_v6_unreach(struct net *net,
241 struct sk_buff *oldskb,
242 const struct net_device *dev,
243 int hook, u8 code)
244 {
245 struct sk_buff *nskb;
246 struct ipv6hdr *nip6h;
247 struct icmp6hdr *icmp6h;
248 unsigned int len;
249
250 if (!nft_bridge_ip6hdr_validate(oldskb))
251 return;
252
253 /* Include "As much of invoking packet as possible without the ICMPv6
254 * packet exceeding the minimum IPv6 MTU" in the ICMP payload.
255 */
256 len = min_t(unsigned int, 1220, oldskb->len);
257
258 if (!pskb_may_pull(oldskb, len))
259 return;
260
261 if (!reject6_br_csum_ok(oldskb, hook))
262 return;
263
264 nskb = alloc_skb(sizeof(struct ipv6hdr) + sizeof(struct icmp6hdr) +
265 LL_MAX_HEADER + len, GFP_ATOMIC);
266 if (!nskb)
267 return;
268
269 skb_reserve(nskb, LL_MAX_HEADER);
270 nip6h = nf_reject_ip6hdr_put(nskb, oldskb, IPPROTO_ICMPV6,
271 net->ipv6.devconf_all->hop_limit);
272
273 skb_reset_transport_header(nskb);
274 icmp6h = skb_put_zero(nskb, sizeof(struct icmp6hdr));
275 icmp6h->icmp6_type = ICMPV6_DEST_UNREACH;
276 icmp6h->icmp6_code = code;
277
278 skb_put_data(nskb, skb_network_header(oldskb), len);
279 nip6h->payload_len = htons(nskb->len - sizeof(struct ipv6hdr));
280
281 icmp6h->icmp6_cksum =
282 csum_ipv6_magic(&nip6h->saddr, &nip6h->daddr,
283 nskb->len - sizeof(struct ipv6hdr),
284 IPPROTO_ICMPV6,
285 csum_partial(icmp6h,
286 nskb->len - sizeof(struct ipv6hdr),
287 0));
288
289 nft_reject_br_push_etherhdr(oldskb, nskb);
290
291 br_forward(br_port_get_rcu(dev), nskb, false, true);
292 }
293
294 static void nft_reject_bridge_eval(const struct nft_expr *expr,
295 struct nft_regs *regs,
296 const struct nft_pktinfo *pkt)
297 {
298 struct nft_reject *priv = nft_expr_priv(expr);
299 const unsigned char *dest = eth_hdr(pkt->skb)->h_dest;
300
301 if (is_broadcast_ether_addr(dest) ||
302 is_multicast_ether_addr(dest))
303 goto out;
304
305 switch (eth_hdr(pkt->skb)->h_proto) {
306 case htons(ETH_P_IP):
307 switch (priv->type) {
308 case NFT_REJECT_ICMP_UNREACH:
309 nft_reject_br_send_v4_unreach(nft_net(pkt), pkt->skb,
310 nft_in(pkt),
311 nft_hook(pkt),
312 priv->icmp_code);
313 break;
314 case NFT_REJECT_TCP_RST:
315 nft_reject_br_send_v4_tcp_reset(nft_net(pkt), pkt->skb,
316 nft_in(pkt),
317 nft_hook(pkt));
318 break;
319 case NFT_REJECT_ICMPX_UNREACH:
320 nft_reject_br_send_v4_unreach(nft_net(pkt), pkt->skb,
321 nft_in(pkt),
322 nft_hook(pkt),
323 nft_reject_icmp_code(priv->icmp_code));
324 break;
325 }
326 break;
327 case htons(ETH_P_IPV6):
328 switch (priv->type) {
329 case NFT_REJECT_ICMP_UNREACH:
330 nft_reject_br_send_v6_unreach(nft_net(pkt), pkt->skb,
331 nft_in(pkt),
332 nft_hook(pkt),
333 priv->icmp_code);
334 break;
335 case NFT_REJECT_TCP_RST:
336 nft_reject_br_send_v6_tcp_reset(nft_net(pkt), pkt->skb,
337 nft_in(pkt),
338 nft_hook(pkt));
339 break;
340 case NFT_REJECT_ICMPX_UNREACH:
341 nft_reject_br_send_v6_unreach(nft_net(pkt), pkt->skb,
342 nft_in(pkt),
343 nft_hook(pkt),
344 nft_reject_icmpv6_code(priv->icmp_code));
345 break;
346 }
347 break;
348 default:
349 /* No explicit way to reject this protocol, drop it. */
350 break;
351 }
352 out:
353 regs->verdict.code = NF_DROP;
354 }
355
356 static int nft_reject_bridge_validate(const struct nft_ctx *ctx,
357 const struct nft_expr *expr,
358 const struct nft_data **data)
359 {
360 return nft_chain_validate_hooks(ctx->chain, (1 << NF_BR_PRE_ROUTING) |
361 (1 << NF_BR_LOCAL_IN));
362 }
363
364 static int nft_reject_bridge_init(const struct nft_ctx *ctx,
365 const struct nft_expr *expr,
366 const struct nlattr * const tb[])
367 {
368 struct nft_reject *priv = nft_expr_priv(expr);
369 int icmp_code;
370
371 if (tb[NFTA_REJECT_TYPE] == NULL)
372 return -EINVAL;
373
374 priv->type = ntohl(nla_get_be32(tb[NFTA_REJECT_TYPE]));
375 switch (priv->type) {
376 case NFT_REJECT_ICMP_UNREACH:
377 case NFT_REJECT_ICMPX_UNREACH:
378 if (tb[NFTA_REJECT_ICMP_CODE] == NULL)
379 return -EINVAL;
380
381 icmp_code = nla_get_u8(tb[NFTA_REJECT_ICMP_CODE]);
382 if (priv->type == NFT_REJECT_ICMPX_UNREACH &&
383 icmp_code > NFT_REJECT_ICMPX_MAX)
384 return -EINVAL;
385
386 priv->icmp_code = icmp_code;
387 break;
388 case NFT_REJECT_TCP_RST:
389 break;
390 default:
391 return -EINVAL;
392 }
393 return 0;
394 }
395
396 static int nft_reject_bridge_dump(struct sk_buff *skb,
397 const struct nft_expr *expr)
398 {
399 const struct nft_reject *priv = nft_expr_priv(expr);
400
401 if (nla_put_be32(skb, NFTA_REJECT_TYPE, htonl(priv->type)))
402 goto nla_put_failure;
403
404 switch (priv->type) {
405 case NFT_REJECT_ICMP_UNREACH:
406 case NFT_REJECT_ICMPX_UNREACH:
407 if (nla_put_u8(skb, NFTA_REJECT_ICMP_CODE, priv->icmp_code))
408 goto nla_put_failure;
409 break;
410 default:
411 break;
412 }
413
414 return 0;
415
416 nla_put_failure:
417 return -1;
418 }
419
420 static struct nft_expr_type nft_reject_bridge_type;
421 static const struct nft_expr_ops nft_reject_bridge_ops = {
422 .type = &nft_reject_bridge_type,
423 .size = NFT_EXPR_SIZE(sizeof(struct nft_reject)),
424 .eval = nft_reject_bridge_eval,
425 .init = nft_reject_bridge_init,
426 .dump = nft_reject_bridge_dump,
427 .validate = nft_reject_bridge_validate,
428 };
429
430 static struct nft_expr_type nft_reject_bridge_type __read_mostly = {
431 .family = NFPROTO_BRIDGE,
432 .name = "reject",
433 .ops = &nft_reject_bridge_ops,
434 .policy = nft_reject_policy,
435 .maxattr = NFTA_REJECT_MAX,
436 .owner = THIS_MODULE,
437 };
438
439 static int __init nft_reject_bridge_module_init(void)
440 {
441 return nft_register_expr(&nft_reject_bridge_type);
442 }
443
444 static void __exit nft_reject_bridge_module_exit(void)
445 {
446 nft_unregister_expr(&nft_reject_bridge_type);
447 }
448
449 module_init(nft_reject_bridge_module_init);
450 module_exit(nft_reject_bridge_module_exit);
451
452 MODULE_LICENSE("GPL");
453 MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
454 MODULE_ALIAS_NFT_AF_EXPR(AF_BRIDGE, "reject");
455
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.