1 /*
2 * Handle firewalling
3 * Linux ethernet bridge
4 *
5 * Authors:
6 * Lennert Buytenhek <buytenh@gnu.org>
7 * Bart De Schuymer <bdschuym@pandora.be>
8 *
9 * This program is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU General Public License
11 * as published by the Free Software Foundation; either version
12 * 2 of the License, or (at your option) any later version.
13 *
14 * Lennert dedicates this file to Kerstin Wurdinger.
15 */
16
17 #include <linux/module.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.h>
20 #include <linux/ip.h>
21 #include <linux/netdevice.h>
22 #include <linux/skbuff.h>
23 #include <linux/if_arp.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <linux/if_pppox.h>
27 #include <linux/ppp_defs.h>
28 #include <linux/netfilter_bridge.h>
29 #include <linux/netfilter_ipv4.h>
30 #include <linux/netfilter_ipv6.h>
31 #include <linux/netfilter_arp.h>
32 #include <linux/in_route.h>
33 #include <linux/rculist.h>
34 #include <linux/inetdevice.h>
35
36 #include <net/ip.h>
37 #include <net/ipv6.h>
38 #include <net/addrconf.h>
39 #include <net/route.h>
40 #include <net/netfilter/br_netfilter.h>
41 #include <net/netns/generic.h>
42
43 #include <linux/uaccess.h>
44 #include "br_private.h"
45 #ifdef CONFIG_SYSCTL
46 #include <linux/sysctl.h>
47 #endif
48
49 static unsigned int brnf_net_id __read_mostly;
50
51 struct brnf_net {
52 bool enabled;
53 };
54
55 #ifdef CONFIG_SYSCTL
56 static struct ctl_table_header *brnf_sysctl_header;
57 static int brnf_call_iptables __read_mostly = 1;
58 static int brnf_call_ip6tables __read_mostly = 1;
59 static int brnf_call_arptables __read_mostly = 1;
60 static int brnf_filter_vlan_tagged __read_mostly;
61 static int brnf_filter_pppoe_tagged __read_mostly;
62 static int brnf_pass_vlan_indev __read_mostly;
63 #else
64 #define brnf_call_iptables 1
65 #define brnf_call_ip6tables 1
66 #define brnf_call_arptables 1
67 #define brnf_filter_vlan_tagged 0
68 #define brnf_filter_pppoe_tagged 0
69 #define brnf_pass_vlan_indev 0
70 #endif
71
72 #define IS_IP(skb) \
73 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
74
75 #define IS_IPV6(skb) \
76 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
77
78 #define IS_ARP(skb) \
79 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
80
vlan_proto(const struct sk_buff * skb)81 static inline __be16 vlan_proto(const struct sk_buff *skb)
82 {
83 if (skb_vlan_tag_present(skb))
84 return skb->protocol;
85 else if (skb->protocol == htons(ETH_P_8021Q))
86 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
87 else
88 return 0;
89 }
90
91 #define IS_VLAN_IP(skb) \
92 (vlan_proto(skb) == htons(ETH_P_IP) && \
93 brnf_filter_vlan_tagged)
94
95 #define IS_VLAN_IPV6(skb) \
96 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
97 brnf_filter_vlan_tagged)
98
99 #define IS_VLAN_ARP(skb) \
100 (vlan_proto(skb) == htons(ETH_P_ARP) && \
101 brnf_filter_vlan_tagged)
102
pppoe_proto(const struct sk_buff * skb)103 static inline __be16 pppoe_proto(const struct sk_buff *skb)
104 {
105 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
106 sizeof(struct pppoe_hdr)));
107 }
108
109 #define IS_PPPOE_IP(skb) \
110 (skb->protocol == htons(ETH_P_PPP_SES) && \
111 pppoe_proto(skb) == htons(PPP_IP) && \
112 brnf_filter_pppoe_tagged)
113
114 #define IS_PPPOE_IPV6(skb) \
115 (skb->protocol == htons(ETH_P_PPP_SES) && \
116 pppoe_proto(skb) == htons(PPP_IPV6) && \
117 brnf_filter_pppoe_tagged)
118
119 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
120 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
121
122 struct brnf_frag_data {
123 char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
124 u8 encap_size;
125 u8 size;
126 u16 vlan_tci;
127 __be16 vlan_proto;
128 };
129
130 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
131
nf_bridge_info_free(struct sk_buff * skb)132 static void nf_bridge_info_free(struct sk_buff *skb)
133 {
134 if (skb->nf_bridge) {
135 nf_bridge_put(skb->nf_bridge);
136 skb->nf_bridge = NULL;
137 }
138 }
139
bridge_parent(const struct net_device * dev)140 static inline struct net_device *bridge_parent(const struct net_device *dev)
141 {
142 struct net_bridge_port *port;
143
144 port = br_port_get_rcu(dev);
145 return port ? port->br->dev : NULL;
146 }
147
nf_bridge_unshare(struct sk_buff * skb)148 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
149 {
150 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
151
152 if (refcount_read(&nf_bridge->use) > 1) {
153 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
154
155 if (tmp) {
156 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
157 refcount_set(&tmp->use, 1);
158 }
159 nf_bridge_put(nf_bridge);
160 nf_bridge = tmp;
161 }
162 return nf_bridge;
163 }
164
nf_bridge_encap_header_len(const struct sk_buff * skb)165 unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
166 {
167 switch (skb->protocol) {
168 case __cpu_to_be16(ETH_P_8021Q):
169 return VLAN_HLEN;
170 case __cpu_to_be16(ETH_P_PPP_SES):
171 return PPPOE_SES_HLEN;
172 default:
173 return 0;
174 }
175 }
176
nf_bridge_pull_encap_header(struct sk_buff * skb)177 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
178 {
179 unsigned int len = nf_bridge_encap_header_len(skb);
180
181 skb_pull(skb, len);
182 skb->network_header += len;
183 }
184
nf_bridge_pull_encap_header_rcsum(struct sk_buff * skb)185 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
186 {
187 unsigned int len = nf_bridge_encap_header_len(skb);
188
189 skb_pull_rcsum(skb, len);
190 skb->network_header += len;
191 }
192
193 /* When handing a packet over to the IP layer
194 * check whether we have a skb that is in the
195 * expected format
196 */
197
br_validate_ipv4(struct net * net,struct sk_buff * skb)198 static int br_validate_ipv4(struct net *net, struct sk_buff *skb)
199 {
200 const struct iphdr *iph;
201 u32 len;
202
203 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
204 goto inhdr_error;
205
206 iph = ip_hdr(skb);
207
208 /* Basic sanity checks */
209 if (iph->ihl < 5 || iph->version != 4)
210 goto inhdr_error;
211
212 if (!pskb_may_pull(skb, iph->ihl*4))
213 goto inhdr_error;
214
215 iph = ip_hdr(skb);
216 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
217 goto inhdr_error;
218
219 len = ntohs(iph->tot_len);
220 if (skb->len < len) {
221 __IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS);
222 goto drop;
223 } else if (len < (iph->ihl*4))
224 goto inhdr_error;
225
226 if (pskb_trim_rcsum(skb, len)) {
227 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
228 goto drop;
229 }
230
231 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
232 /* We should really parse IP options here but until
233 * somebody who actually uses IP options complains to
234 * us we'll just silently ignore the options because
235 * we're lazy!
236 */
237 return 0;
238
239 inhdr_error:
240 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
241 drop:
242 return -1;
243 }
244
nf_bridge_update_protocol(struct sk_buff * skb)245 void nf_bridge_update_protocol(struct sk_buff *skb)
246 {
247 switch (skb->nf_bridge->orig_proto) {
248 case BRNF_PROTO_8021Q:
249 skb->protocol = htons(ETH_P_8021Q);
250 break;
251 case BRNF_PROTO_PPPOE:
252 skb->protocol = htons(ETH_P_PPP_SES);
253 break;
254 case BRNF_PROTO_UNCHANGED:
255 break;
256 }
257 }
258
259 /* Obtain the correct destination MAC address, while preserving the original
260 * source MAC address. If we already know this address, we just copy it. If we
261 * don't, we use the neighbour framework to find out. In both cases, we make
262 * sure that br_handle_frame_finish() is called afterwards.
263 */
br_nf_pre_routing_finish_bridge(struct net * net,struct sock * sk,struct sk_buff * skb)264 int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb)
265 {
266 struct neighbour *neigh;
267 struct dst_entry *dst;
268
269 skb->dev = bridge_parent(skb->dev);
270 if (!skb->dev)
271 goto free_skb;
272 dst = skb_dst(skb);
273 neigh = dst_neigh_lookup_skb(dst, skb);
274 if (neigh) {
275 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
276 int ret;
277
278 if ((neigh->nud_state & NUD_CONNECTED) && neigh->hh.hh_len) {
279 neigh_hh_bridge(&neigh->hh, skb);
280 skb->dev = nf_bridge->physindev;
281 ret = br_handle_frame_finish(net, sk, skb);
282 } else {
283 /* the neighbour function below overwrites the complete
284 * MAC header, so we save the Ethernet source address and
285 * protocol number.
286 */
287 skb_copy_from_linear_data_offset(skb,
288 -(ETH_HLEN-ETH_ALEN),
289 nf_bridge->neigh_header,
290 ETH_HLEN-ETH_ALEN);
291 /* tell br_dev_xmit to continue with forwarding */
292 nf_bridge->bridged_dnat = 1;
293 /* FIXME Need to refragment */
294 ret = neigh->output(neigh, skb);
295 }
296 neigh_release(neigh);
297 return ret;
298 }
299 free_skb:
300 kfree_skb(skb);
301 return 0;
302 }
303
304 static inline bool
br_nf_ipv4_daddr_was_changed(const struct sk_buff * skb,const struct nf_bridge_info * nf_bridge)305 br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb,
306 const struct nf_bridge_info *nf_bridge)
307 {
308 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
309 }
310
311 /* This requires some explaining. If DNAT has taken place,
312 * we will need to fix up the destination Ethernet address.
313 * This is also true when SNAT takes place (for the reply direction).
314 *
315 * There are two cases to consider:
316 * 1. The packet was DNAT'ed to a device in the same bridge
317 * port group as it was received on. We can still bridge
318 * the packet.
319 * 2. The packet was DNAT'ed to a different device, either
320 * a non-bridged device or another bridge port group.
321 * The packet will need to be routed.
322 *
323 * The correct way of distinguishing between these two cases is to
324 * call ip_route_input() and to look at skb->dst->dev, which is
325 * changed to the destination device if ip_route_input() succeeds.
326 *
327 * Let's first consider the case that ip_route_input() succeeds:
328 *
329 * If the output device equals the logical bridge device the packet
330 * came in on, we can consider this bridging. The corresponding MAC
331 * address will be obtained in br_nf_pre_routing_finish_bridge.
332 * Otherwise, the packet is considered to be routed and we just
333 * change the destination MAC address so that the packet will
334 * later be passed up to the IP stack to be routed. For a redirected
335 * packet, ip_route_input() will give back the localhost as output device,
336 * which differs from the bridge device.
337 *
338 * Let's now consider the case that ip_route_input() fails:
339 *
340 * This can be because the destination address is martian, in which case
341 * the packet will be dropped.
342 * If IP forwarding is disabled, ip_route_input() will fail, while
343 * ip_route_output_key() can return success. The source
344 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
345 * thinks we're handling a locally generated packet and won't care
346 * if IP forwarding is enabled. If the output device equals the logical bridge
347 * device, we proceed as if ip_route_input() succeeded. If it differs from the
348 * logical bridge port or if ip_route_output_key() fails we drop the packet.
349 */
br_nf_pre_routing_finish(struct net * net,struct sock * sk,struct sk_buff * skb)350 static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
351 {
352 struct net_device *dev = skb->dev;
353 struct iphdr *iph = ip_hdr(skb);
354 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
355 struct rtable *rt;
356 int err;
357
358 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
359
360 if (nf_bridge->pkt_otherhost) {
361 skb->pkt_type = PACKET_OTHERHOST;
362 nf_bridge->pkt_otherhost = false;
363 }
364 nf_bridge->in_prerouting = 0;
365 if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) {
366 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
367 struct in_device *in_dev = __in_dev_get_rcu(dev);
368
369 /* If err equals -EHOSTUNREACH the error is due to a
370 * martian destination or due to the fact that
371 * forwarding is disabled. For most martian packets,
372 * ip_route_output_key() will fail. It won't fail for 2 types of
373 * martian destinations: loopback destinations and destination
374 * 0.0.0.0. In both cases the packet will be dropped because the
375 * destination is the loopback device and not the bridge. */
376 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
377 goto free_skb;
378
379 rt = ip_route_output(net, iph->daddr, 0,
380 RT_TOS(iph->tos), 0);
381 if (!IS_ERR(rt)) {
382 /* - Bridged-and-DNAT'ed traffic doesn't
383 * require ip_forwarding. */
384 if (rt->dst.dev == dev) {
385 skb_dst_set(skb, &rt->dst);
386 goto bridged_dnat;
387 }
388 ip_rt_put(rt);
389 }
390 free_skb:
391 kfree_skb(skb);
392 return 0;
393 } else {
394 if (skb_dst(skb)->dev == dev) {
395 bridged_dnat:
396 skb->dev = nf_bridge->physindev;
397 nf_bridge_update_protocol(skb);
398 nf_bridge_push_encap_header(skb);
399 br_nf_hook_thresh(NF_BR_PRE_ROUTING,
400 net, sk, skb, skb->dev,
401 NULL,
402 br_nf_pre_routing_finish_bridge);
403 return 0;
404 }
405 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
406 skb->pkt_type = PACKET_HOST;
407 }
408 } else {
409 rt = bridge_parent_rtable(nf_bridge->physindev);
410 if (!rt) {
411 kfree_skb(skb);
412 return 0;
413 }
414 skb_dst_set_noref(skb, &rt->dst);
415 }
416
417 skb->dev = nf_bridge->physindev;
418 nf_bridge_update_protocol(skb);
419 nf_bridge_push_encap_header(skb);
420 br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL,
421 br_handle_frame_finish);
422 return 0;
423 }
424
brnf_get_logical_dev(struct sk_buff * skb,const struct net_device * dev)425 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
426 {
427 struct net_device *vlan, *br;
428
429 br = bridge_parent(dev);
430 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
431 return br;
432
433 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
434 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
435
436 return vlan ? vlan : br;
437 }
438
439 /* Some common code for IPv4/IPv6 */
setup_pre_routing(struct sk_buff * skb)440 struct net_device *setup_pre_routing(struct sk_buff *skb)
441 {
442 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
443
444 if (skb->pkt_type == PACKET_OTHERHOST) {
445 skb->pkt_type = PACKET_HOST;
446 nf_bridge->pkt_otherhost = true;
447 }
448
449 nf_bridge->in_prerouting = 1;
450 nf_bridge->physindev = skb->dev;
451 skb->dev = brnf_get_logical_dev(skb, skb->dev);
452
453 if (skb->protocol == htons(ETH_P_8021Q))
454 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
455 else if (skb->protocol == htons(ETH_P_PPP_SES))
456 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
457
458 /* Must drop socket now because of tproxy. */
459 skb_orphan(skb);
460 return skb->dev;
461 }
462
463 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
464 * Replicate the checks that IPv4 does on packet reception.
465 * Set skb->dev to the bridge device (i.e. parent of the
466 * receiving device) to make netfilter happy, the REDIRECT
467 * target in particular. Save the original destination IP
468 * address to be able to detect DNAT afterwards. */
br_nf_pre_routing(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)469 static unsigned int br_nf_pre_routing(void *priv,
470 struct sk_buff *skb,
471 const struct nf_hook_state *state)
472 {
473 struct nf_bridge_info *nf_bridge;
474 struct net_bridge_port *p;
475 struct net_bridge *br;
476 __u32 len = nf_bridge_encap_header_len(skb);
477
478 if (unlikely(!pskb_may_pull(skb, len)))
479 return NF_DROP;
480
481 p = br_port_get_rcu(state->in);
482 if (p == NULL)
483 return NF_DROP;
484 br = p->br;
485
486 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
487 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
488 return NF_ACCEPT;
489
490 nf_bridge_pull_encap_header_rcsum(skb);
491 return br_nf_pre_routing_ipv6(priv, skb, state);
492 }
493
494 if (!brnf_call_iptables && !br->nf_call_iptables)
495 return NF_ACCEPT;
496
497 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
498 return NF_ACCEPT;
499
500 nf_bridge_pull_encap_header_rcsum(skb);
501
502 if (br_validate_ipv4(state->net, skb))
503 return NF_DROP;
504
505 nf_bridge_put(skb->nf_bridge);
506 if (!nf_bridge_alloc(skb))
507 return NF_DROP;
508 if (!setup_pre_routing(skb))
509 return NF_DROP;
510
511 nf_bridge = nf_bridge_info_get(skb);
512 nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
513
514 skb->protocol = htons(ETH_P_IP);
515 skb->transport_header = skb->network_header + ip_hdr(skb)->ihl * 4;
516
517 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb,
518 skb->dev, NULL,
519 br_nf_pre_routing_finish);
520
521 return NF_STOLEN;
522 }
523
524
525 /* PF_BRIDGE/FORWARD *************************************************/
br_nf_forward_finish(struct net * net,struct sock * sk,struct sk_buff * skb)526 static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
527 {
528 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
529 struct net_device *in;
530
531 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
532
533 if (skb->protocol == htons(ETH_P_IP))
534 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
535
536 if (skb->protocol == htons(ETH_P_IPV6))
537 nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
538
539 in = nf_bridge->physindev;
540 if (nf_bridge->pkt_otherhost) {
541 skb->pkt_type = PACKET_OTHERHOST;
542 nf_bridge->pkt_otherhost = false;
543 }
544 nf_bridge_update_protocol(skb);
545 } else {
546 in = *((struct net_device **)(skb->cb));
547 }
548 nf_bridge_push_encap_header(skb);
549
550 br_nf_hook_thresh(NF_BR_FORWARD, net, sk, skb, in, skb->dev,
551 br_forward_finish);
552 return 0;
553 }
554
555
556 /* This is the 'purely bridged' case. For IP, we pass the packet to
557 * netfilter with indev and outdev set to the bridge device,
558 * but we are still able to filter on the 'real' indev/outdev
559 * because of the physdev module. For ARP, indev and outdev are the
560 * bridge ports. */
br_nf_forward_ip(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)561 static unsigned int br_nf_forward_ip(void *priv,
562 struct sk_buff *skb,
563 const struct nf_hook_state *state)
564 {
565 struct nf_bridge_info *nf_bridge;
566 struct net_device *parent;
567 u_int8_t pf;
568
569 if (!skb->nf_bridge)
570 return NF_ACCEPT;
571
572 /* Need exclusive nf_bridge_info since we might have multiple
573 * different physoutdevs. */
574 if (!nf_bridge_unshare(skb))
575 return NF_DROP;
576
577 nf_bridge = nf_bridge_info_get(skb);
578 if (!nf_bridge)
579 return NF_DROP;
580
581 parent = bridge_parent(state->out);
582 if (!parent)
583 return NF_DROP;
584
585 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
586 pf = NFPROTO_IPV4;
587 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
588 pf = NFPROTO_IPV6;
589 else
590 return NF_ACCEPT;
591
592 nf_bridge_pull_encap_header(skb);
593
594 if (skb->pkt_type == PACKET_OTHERHOST) {
595 skb->pkt_type = PACKET_HOST;
596 nf_bridge->pkt_otherhost = true;
597 }
598
599 if (pf == NFPROTO_IPV4) {
600 if (br_validate_ipv4(state->net, skb))
601 return NF_DROP;
602 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
603 }
604
605 if (pf == NFPROTO_IPV6) {
606 if (br_validate_ipv6(state->net, skb))
607 return NF_DROP;
608 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
609 }
610
611 nf_bridge->physoutdev = skb->dev;
612 if (pf == NFPROTO_IPV4)
613 skb->protocol = htons(ETH_P_IP);
614 else
615 skb->protocol = htons(ETH_P_IPV6);
616
617 NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb,
618 brnf_get_logical_dev(skb, state->in),
619 parent, br_nf_forward_finish);
620
621 return NF_STOLEN;
622 }
623
br_nf_forward_arp(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)624 static unsigned int br_nf_forward_arp(void *priv,
625 struct sk_buff *skb,
626 const struct nf_hook_state *state)
627 {
628 struct net_bridge_port *p;
629 struct net_bridge *br;
630 struct net_device **d = (struct net_device **)(skb->cb);
631
632 p = br_port_get_rcu(state->out);
633 if (p == NULL)
634 return NF_ACCEPT;
635 br = p->br;
636
637 if (!brnf_call_arptables && !br->nf_call_arptables)
638 return NF_ACCEPT;
639
640 if (!IS_ARP(skb)) {
641 if (!IS_VLAN_ARP(skb))
642 return NF_ACCEPT;
643 nf_bridge_pull_encap_header(skb);
644 }
645
646 if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr))))
647 return NF_DROP;
648
649 if (arp_hdr(skb)->ar_pln != 4) {
650 if (IS_VLAN_ARP(skb))
651 nf_bridge_push_encap_header(skb);
652 return NF_ACCEPT;
653 }
654 *d = state->in;
655 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb,
656 state->in, state->out, br_nf_forward_finish);
657
658 return NF_STOLEN;
659 }
660
br_nf_push_frag_xmit(struct net * net,struct sock * sk,struct sk_buff * skb)661 static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
662 {
663 struct brnf_frag_data *data;
664 int err;
665
666 data = this_cpu_ptr(&brnf_frag_data_storage);
667 err = skb_cow_head(skb, data->size);
668
669 if (err) {
670 kfree_skb(skb);
671 return 0;
672 }
673
674 if (data->vlan_tci) {
675 skb->vlan_tci = data->vlan_tci;
676 skb->vlan_proto = data->vlan_proto;
677 }
678
679 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
680 __skb_push(skb, data->encap_size);
681
682 nf_bridge_info_free(skb);
683 return br_dev_queue_push_xmit(net, sk, skb);
684 }
685
686 static int
br_nf_ip_fragment(struct net * net,struct sock * sk,struct sk_buff * skb,int (* output)(struct net *,struct sock *,struct sk_buff *))687 br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
688 int (*output)(struct net *, struct sock *, struct sk_buff *))
689 {
690 unsigned int mtu = ip_skb_dst_mtu(sk, skb);
691 struct iphdr *iph = ip_hdr(skb);
692
693 if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
694 (IPCB(skb)->frag_max_size &&
695 IPCB(skb)->frag_max_size > mtu))) {
696 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
697 kfree_skb(skb);
698 return -EMSGSIZE;
699 }
700
701 return ip_do_fragment(net, sk, skb, output);
702 }
703
nf_bridge_mtu_reduction(const struct sk_buff * skb)704 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
705 {
706 if (skb->nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
707 return PPPOE_SES_HLEN;
708 return 0;
709 }
710
br_nf_dev_queue_xmit(struct net * net,struct sock * sk,struct sk_buff * skb)711 static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
712 {
713 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
714 unsigned int mtu, mtu_reserved;
715
716 mtu_reserved = nf_bridge_mtu_reduction(skb);
717 mtu = skb->dev->mtu;
718
719 if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu)
720 mtu = nf_bridge->frag_max_size;
721
722 if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) {
723 nf_bridge_info_free(skb);
724 return br_dev_queue_push_xmit(net, sk, skb);
725 }
726
727 /* This is wrong! We should preserve the original fragment
728 * boundaries by preserving frag_list rather than refragmenting.
729 */
730 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) &&
731 skb->protocol == htons(ETH_P_IP)) {
732 struct brnf_frag_data *data;
733
734 if (br_validate_ipv4(net, skb))
735 goto drop;
736
737 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
738
739 nf_bridge_update_protocol(skb);
740
741 data = this_cpu_ptr(&brnf_frag_data_storage);
742
743 data->vlan_tci = skb->vlan_tci;
744 data->vlan_proto = skb->vlan_proto;
745 data->encap_size = nf_bridge_encap_header_len(skb);
746 data->size = ETH_HLEN + data->encap_size;
747
748 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
749 data->size);
750
751 return br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit);
752 }
753 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) &&
754 skb->protocol == htons(ETH_P_IPV6)) {
755 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
756 struct brnf_frag_data *data;
757
758 if (br_validate_ipv6(net, skb))
759 goto drop;
760
761 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
762
763 nf_bridge_update_protocol(skb);
764
765 data = this_cpu_ptr(&brnf_frag_data_storage);
766 data->encap_size = nf_bridge_encap_header_len(skb);
767 data->size = ETH_HLEN + data->encap_size;
768
769 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
770 data->size);
771
772 if (v6ops)
773 return v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit);
774
775 kfree_skb(skb);
776 return -EMSGSIZE;
777 }
778 nf_bridge_info_free(skb);
779 return br_dev_queue_push_xmit(net, sk, skb);
780 drop:
781 kfree_skb(skb);
782 return 0;
783 }
784
785 /* PF_BRIDGE/POST_ROUTING ********************************************/
br_nf_post_routing(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)786 static unsigned int br_nf_post_routing(void *priv,
787 struct sk_buff *skb,
788 const struct nf_hook_state *state)
789 {
790 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
791 struct net_device *realoutdev = bridge_parent(skb->dev);
792 u_int8_t pf;
793
794 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
795 * on a bridge, but was delivered locally and is now being routed:
796 *
797 * POST_ROUTING was already invoked from the ip stack.
798 */
799 if (!nf_bridge || !nf_bridge->physoutdev)
800 return NF_ACCEPT;
801
802 if (!realoutdev)
803 return NF_DROP;
804
805 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
806 pf = NFPROTO_IPV4;
807 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
808 pf = NFPROTO_IPV6;
809 else
810 return NF_ACCEPT;
811
812 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
813 * about the value of skb->pkt_type. */
814 if (skb->pkt_type == PACKET_OTHERHOST) {
815 skb->pkt_type = PACKET_HOST;
816 nf_bridge->pkt_otherhost = true;
817 }
818
819 nf_bridge_pull_encap_header(skb);
820 if (pf == NFPROTO_IPV4)
821 skb->protocol = htons(ETH_P_IP);
822 else
823 skb->protocol = htons(ETH_P_IPV6);
824
825 NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb,
826 NULL, realoutdev,
827 br_nf_dev_queue_xmit);
828
829 return NF_STOLEN;
830 }
831
832 /* IP/SABOTAGE *****************************************************/
833 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
834 * for the second time. */
ip_sabotage_in(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)835 static unsigned int ip_sabotage_in(void *priv,
836 struct sk_buff *skb,
837 const struct nf_hook_state *state)
838 {
839 if (skb->nf_bridge && !skb->nf_bridge->in_prerouting &&
840 !netif_is_l3_master(skb->dev)) {
841 state->okfn(state->net, state->sk, skb);
842 return NF_STOLEN;
843 }
844
845 return NF_ACCEPT;
846 }
847
848 /* This is called when br_netfilter has called into iptables/netfilter,
849 * and DNAT has taken place on a bridge-forwarded packet.
850 *
851 * neigh->output has created a new MAC header, with local br0 MAC
852 * as saddr.
853 *
854 * This restores the original MAC saddr of the bridged packet
855 * before invoking bridge forward logic to transmit the packet.
856 */
br_nf_pre_routing_finish_bridge_slow(struct sk_buff * skb)857 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
858 {
859 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
860
861 skb_pull(skb, ETH_HLEN);
862 nf_bridge->bridged_dnat = 0;
863
864 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
865
866 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
867 nf_bridge->neigh_header,
868 ETH_HLEN - ETH_ALEN);
869 skb->dev = nf_bridge->physindev;
870
871 nf_bridge->physoutdev = NULL;
872 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
873 }
874
br_nf_dev_xmit(struct sk_buff * skb)875 static int br_nf_dev_xmit(struct sk_buff *skb)
876 {
877 if (skb->nf_bridge && skb->nf_bridge->bridged_dnat) {
878 br_nf_pre_routing_finish_bridge_slow(skb);
879 return 1;
880 }
881 return 0;
882 }
883
884 static const struct nf_br_ops br_ops = {
885 .br_dev_xmit_hook = br_nf_dev_xmit,
886 };
887
888 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
889 * br_dev_queue_push_xmit is called afterwards */
890 static const struct nf_hook_ops br_nf_ops[] = {
891 {
892 .hook = br_nf_pre_routing,
893 .pf = NFPROTO_BRIDGE,
894 .hooknum = NF_BR_PRE_ROUTING,
895 .priority = NF_BR_PRI_BRNF,
896 },
897 {
898 .hook = br_nf_forward_ip,
899 .pf = NFPROTO_BRIDGE,
900 .hooknum = NF_BR_FORWARD,
901 .priority = NF_BR_PRI_BRNF - 1,
902 },
903 {
904 .hook = br_nf_forward_arp,
905 .pf = NFPROTO_BRIDGE,
906 .hooknum = NF_BR_FORWARD,
907 .priority = NF_BR_PRI_BRNF,
908 },
909 {
910 .hook = br_nf_post_routing,
911 .pf = NFPROTO_BRIDGE,
912 .hooknum = NF_BR_POST_ROUTING,
913 .priority = NF_BR_PRI_LAST,
914 },
915 {
916 .hook = ip_sabotage_in,
917 .pf = NFPROTO_IPV4,
918 .hooknum = NF_INET_PRE_ROUTING,
919 .priority = NF_IP_PRI_FIRST,
920 },
921 {
922 .hook = ip_sabotage_in,
923 .pf = NFPROTO_IPV6,
924 .hooknum = NF_INET_PRE_ROUTING,
925 .priority = NF_IP6_PRI_FIRST,
926 },
927 };
928
brnf_device_event(struct notifier_block * unused,unsigned long event,void * ptr)929 static int brnf_device_event(struct notifier_block *unused, unsigned long event,
930 void *ptr)
931 {
932 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
933 struct brnf_net *brnet;
934 struct net *net;
935 int ret;
936
937 if (event != NETDEV_REGISTER || !(dev->priv_flags & IFF_EBRIDGE))
938 return NOTIFY_DONE;
939
940 ASSERT_RTNL();
941
942 net = dev_net(dev);
943 brnet = net_generic(net, brnf_net_id);
944 if (brnet->enabled)
945 return NOTIFY_OK;
946
947 ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
948 if (ret)
949 return NOTIFY_BAD;
950
951 brnet->enabled = true;
952 return NOTIFY_OK;
953 }
954
brnf_exit_net(struct net * net)955 static void __net_exit brnf_exit_net(struct net *net)
956 {
957 struct brnf_net *brnet = net_generic(net, brnf_net_id);
958
959 if (!brnet->enabled)
960 return;
961
962 nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
963 brnet->enabled = false;
964 }
965
966 static struct pernet_operations brnf_net_ops __read_mostly = {
967 .exit = brnf_exit_net,
968 .id = &brnf_net_id,
969 .size = sizeof(struct brnf_net),
970 };
971
972 static struct notifier_block brnf_notifier __read_mostly = {
973 .notifier_call = brnf_device_event,
974 };
975
976 /* recursively invokes nf_hook_slow (again), skipping already-called
977 * hooks (< NF_BR_PRI_BRNF).
978 *
979 * Called with rcu read lock held.
980 */
br_nf_hook_thresh(unsigned int hook,struct net * net,struct sock * sk,struct sk_buff * skb,struct net_device * indev,struct net_device * outdev,int (* okfn)(struct net *,struct sock *,struct sk_buff *))981 int br_nf_hook_thresh(unsigned int hook, struct net *net,
982 struct sock *sk, struct sk_buff *skb,
983 struct net_device *indev,
984 struct net_device *outdev,
985 int (*okfn)(struct net *, struct sock *,
986 struct sk_buff *))
987 {
988 const struct nf_hook_entries *e;
989 struct nf_hook_state state;
990 struct nf_hook_ops **ops;
991 unsigned int i;
992 int ret;
993
994 e = rcu_dereference(net->nf.hooks[NFPROTO_BRIDGE][hook]);
995 if (!e)
996 return okfn(net, sk, skb);
997
998 ops = nf_hook_entries_get_hook_ops(e);
999 for (i = 0; i < e->num_hook_entries &&
1000 ops[i]->priority <= NF_BR_PRI_BRNF; i++)
1001 ;
1002
1003 nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev,
1004 sk, net, okfn);
1005
1006 ret = nf_hook_slow(skb, &state, e, i);
1007 if (ret == 1)
1008 ret = okfn(net, sk, skb);
1009
1010 return ret;
1011 }
1012
1013 #ifdef CONFIG_SYSCTL
1014 static
brnf_sysctl_call_tables(struct ctl_table * ctl,int write,void __user * buffer,size_t * lenp,loff_t * ppos)1015 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1016 void __user *buffer, size_t *lenp, loff_t *ppos)
1017 {
1018 int ret;
1019
1020 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1021
1022 if (write && *(int *)(ctl->data))
1023 *(int *)(ctl->data) = 1;
1024 return ret;
1025 }
1026
1027 static struct ctl_table brnf_table[] = {
1028 {
1029 .procname = "bridge-nf-call-arptables",
1030 .data = &brnf_call_arptables,
1031 .maxlen = sizeof(int),
1032 .mode = 0644,
1033 .proc_handler = brnf_sysctl_call_tables,
1034 },
1035 {
1036 .procname = "bridge-nf-call-iptables",
1037 .data = &brnf_call_iptables,
1038 .maxlen = sizeof(int),
1039 .mode = 0644,
1040 .proc_handler = brnf_sysctl_call_tables,
1041 },
1042 {
1043 .procname = "bridge-nf-call-ip6tables",
1044 .data = &brnf_call_ip6tables,
1045 .maxlen = sizeof(int),
1046 .mode = 0644,
1047 .proc_handler = brnf_sysctl_call_tables,
1048 },
1049 {
1050 .procname = "bridge-nf-filter-vlan-tagged",
1051 .data = &brnf_filter_vlan_tagged,
1052 .maxlen = sizeof(int),
1053 .mode = 0644,
1054 .proc_handler = brnf_sysctl_call_tables,
1055 },
1056 {
1057 .procname = "bridge-nf-filter-pppoe-tagged",
1058 .data = &brnf_filter_pppoe_tagged,
1059 .maxlen = sizeof(int),
1060 .mode = 0644,
1061 .proc_handler = brnf_sysctl_call_tables,
1062 },
1063 {
1064 .procname = "bridge-nf-pass-vlan-input-dev",
1065 .data = &brnf_pass_vlan_indev,
1066 .maxlen = sizeof(int),
1067 .mode = 0644,
1068 .proc_handler = brnf_sysctl_call_tables,
1069 },
1070 { }
1071 };
1072 #endif
1073
br_netfilter_init(void)1074 static int __init br_netfilter_init(void)
1075 {
1076 int ret;
1077
1078 ret = register_pernet_subsys(&brnf_net_ops);
1079 if (ret < 0)
1080 return ret;
1081
1082 ret = register_netdevice_notifier(&brnf_notifier);
1083 if (ret < 0) {
1084 unregister_pernet_subsys(&brnf_net_ops);
1085 return ret;
1086 }
1087
1088 #ifdef CONFIG_SYSCTL
1089 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1090 if (brnf_sysctl_header == NULL) {
1091 printk(KERN_WARNING
1092 "br_netfilter: can't register to sysctl.\n");
1093 unregister_netdevice_notifier(&brnf_notifier);
1094 unregister_pernet_subsys(&brnf_net_ops);
1095 return -ENOMEM;
1096 }
1097 #endif
1098 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1099 printk(KERN_NOTICE "Bridge firewalling registered\n");
1100 return 0;
1101 }
1102
br_netfilter_fini(void)1103 static void __exit br_netfilter_fini(void)
1104 {
1105 RCU_INIT_POINTER(nf_br_ops, NULL);
1106 unregister_netdevice_notifier(&brnf_notifier);
1107 unregister_pernet_subsys(&brnf_net_ops);
1108 #ifdef CONFIG_SYSCTL
1109 unregister_net_sysctl_table(brnf_sysctl_header);
1110 #endif
1111 }
1112
1113 module_init(br_netfilter_init);
1114 module_exit(br_netfilter_fini);
1115
1116 MODULE_LICENSE("GPL");
1117 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1118 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1119 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");
1120