]> bbs.cooldavid.org Git - net-next-2.6.git/blame_incremental - net/bridge/br_netfilter.c
git://bbs.cooldavid.org / net-next-2.6.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
ip: ip_ra_control() rcu fix
[net-next-2.6.git] / net / bridge / br_netfilter.c
... / ...
CommitLineData
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/inetdevice.h>
34
35#include <net/ip.h>
36#include <net/ipv6.h>
37#include <net/route.h>
38
39#include <asm/uaccess.h>
40#include "br_private.h"
41#ifdef CONFIG_SYSCTL
42#include <linux/sysctl.h>
43#endif
44
45#define skb_origaddr(skb) (((struct bridge_skb_cb *) \
46 (skb->nf_bridge->data))->daddr.ipv4)
47#define store_orig_dstaddr(skb) (skb_origaddr(skb) = ip_hdr(skb)->daddr)
48#define dnat_took_place(skb) (skb_origaddr(skb) != ip_hdr(skb)->daddr)
49
50#ifdef CONFIG_SYSCTL
51static struct ctl_table_header *brnf_sysctl_header;
52static int brnf_call_iptables __read_mostly = 1;
53static int brnf_call_ip6tables __read_mostly = 1;
54static int brnf_call_arptables __read_mostly = 1;
55static int brnf_filter_vlan_tagged __read_mostly = 0;
56static int brnf_filter_pppoe_tagged __read_mostly = 0;
57#else
58#define brnf_filter_vlan_tagged 0
59#define brnf_filter_pppoe_tagged 0
60#endif
61
62static inline __be16 vlan_proto(const struct sk_buff *skb)
63{
64 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
65}
66
67#define IS_VLAN_IP(skb) \
68 (skb->protocol == htons(ETH_P_8021Q) && \
69 vlan_proto(skb) == htons(ETH_P_IP) && \
70 brnf_filter_vlan_tagged)
71
72#define IS_VLAN_IPV6(skb) \
73 (skb->protocol == htons(ETH_P_8021Q) && \
74 vlan_proto(skb) == htons(ETH_P_IPV6) &&\
75 brnf_filter_vlan_tagged)
76
77#define IS_VLAN_ARP(skb) \
78 (skb->protocol == htons(ETH_P_8021Q) && \
79 vlan_proto(skb) == htons(ETH_P_ARP) && \
80 brnf_filter_vlan_tagged)
81
82static inline __be16 pppoe_proto(const struct sk_buff *skb)
83{
84 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
85 sizeof(struct pppoe_hdr)));
86}
87
88#define IS_PPPOE_IP(skb) \
89 (skb->protocol == htons(ETH_P_PPP_SES) && \
90 pppoe_proto(skb) == htons(PPP_IP) && \
91 brnf_filter_pppoe_tagged)
92
93#define IS_PPPOE_IPV6(skb) \
94 (skb->protocol == htons(ETH_P_PPP_SES) && \
95 pppoe_proto(skb) == htons(PPP_IPV6) && \
96 brnf_filter_pppoe_tagged)
97
98static void fake_update_pmtu(struct dst_entry *dst, u32 mtu)
99{
100}
101
102static struct dst_ops fake_dst_ops = {
103 .family = AF_INET,
104 .protocol = cpu_to_be16(ETH_P_IP),
105 .update_pmtu = fake_update_pmtu,
106 .entries = ATOMIC_INIT(0),
107};
108
109/*
110 * Initialize bogus route table used to keep netfilter happy.
111 * Currently, we fill in the PMTU entry because netfilter
112 * refragmentation needs it, and the rt_flags entry because
113 * ipt_REJECT needs it. Future netfilter modules might
114 * require us to fill additional fields.
115 */
116void br_netfilter_rtable_init(struct net_bridge *br)
117{
118 struct rtable *rt = &br->fake_rtable;
119
120 atomic_set(&rt->u.dst.__refcnt, 1);
121 rt->u.dst.dev = br->dev;
122 rt->u.dst.path = &rt->u.dst;
123 rt->u.dst.metrics[RTAX_MTU - 1] = 1500;
124 rt->u.dst.flags = DST_NOXFRM;
125 rt->u.dst.ops = &fake_dst_ops;
126}
127
128static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
129{
130 struct net_bridge_port *port = rcu_dereference(dev->br_port);
131
132 return port ? &port->br->fake_rtable : NULL;
133}
134
135static inline struct net_device *bridge_parent(const struct net_device *dev)
136{
137 struct net_bridge_port *port = rcu_dereference(dev->br_port);
138
139 return port ? port->br->dev : NULL;
140}
141
142static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
143{
144 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
145 if (likely(skb->nf_bridge))
146 atomic_set(&(skb->nf_bridge->use), 1);
147
148 return skb->nf_bridge;
149}
150
151static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
152{
153 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
154
155 if (atomic_read(&nf_bridge->use) > 1) {
156 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
157
158 if (tmp) {
159 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
160 atomic_set(&tmp->use, 1);
161 nf_bridge_put(nf_bridge);
162 }
163 nf_bridge = tmp;
164 }
165 return nf_bridge;
166}
167
168static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
169{
170 unsigned int len = nf_bridge_encap_header_len(skb);
171
172 skb_push(skb, len);
173 skb->network_header -= len;
174}
175
176static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
177{
178 unsigned int len = nf_bridge_encap_header_len(skb);
179
180 skb_pull(skb, len);
181 skb->network_header += len;
182}
183
184static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
185{
186 unsigned int len = nf_bridge_encap_header_len(skb);
187
188 skb_pull_rcsum(skb, len);
189 skb->network_header += len;
190}
191
192static inline void nf_bridge_save_header(struct sk_buff *skb)
193{
194 int header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
195
196 skb_copy_from_linear_data_offset(skb, -header_size,
197 skb->nf_bridge->data, header_size);
198}
199
200static inline void nf_bridge_update_protocol(struct sk_buff *skb)
201{
202 if (skb->nf_bridge->mask & BRNF_8021Q)
203 skb->protocol = htons(ETH_P_8021Q);
204 else if (skb->nf_bridge->mask & BRNF_PPPoE)
205 skb->protocol = htons(ETH_P_PPP_SES);
206}
207
208/* Fill in the header for fragmented IP packets handled by
209 * the IPv4 connection tracking code.
210 */
211int nf_bridge_copy_header(struct sk_buff *skb)
212{
213 int err;
214 unsigned int header_size;
215
216 nf_bridge_update_protocol(skb);
217 header_size = ETH_HLEN + nf_bridge_encap_header_len(skb);
218 err = skb_cow_head(skb, header_size);
219 if (err)
220 return err;
221
222 skb_copy_to_linear_data_offset(skb, -header_size,
223 skb->nf_bridge->data, header_size);
224 __skb_push(skb, nf_bridge_encap_header_len(skb));
225 return 0;
226}
227
228/* PF_BRIDGE/PRE_ROUTING *********************************************/
229/* Undo the changes made for ip6tables PREROUTING and continue the
230 * bridge PRE_ROUTING hook. */
231static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
232{
233 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
234 struct rtable *rt;
235
236 if (nf_bridge->mask & BRNF_PKT_TYPE) {
237 skb->pkt_type = PACKET_OTHERHOST;
238 nf_bridge->mask ^= BRNF_PKT_TYPE;
239 }
240 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
241
242 rt = bridge_parent_rtable(nf_bridge->physindev);
243 if (!rt) {
244 kfree_skb(skb);
245 return 0;
246 }
247 dst_hold(&rt->u.dst);
248 skb_dst_set(skb, &rt->u.dst);
249
250 skb->dev = nf_bridge->physindev;
251 nf_bridge_update_protocol(skb);
252 nf_bridge_push_encap_header(skb);
253 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
254 br_handle_frame_finish, 1);
255
256 return 0;
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 */
264static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
265{
266 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
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 if (dst->hh) {
274 neigh_hh_bridge(dst->hh, skb);
275 skb->dev = nf_bridge->physindev;
276 return br_handle_frame_finish(skb);
277 } else if (dst->neighbour) {
278 /* the neighbour function below overwrites the complete
279 * MAC header, so we save the Ethernet source address and
280 * protocol number. */
281 skb_copy_from_linear_data_offset(skb, -(ETH_HLEN-ETH_ALEN), skb->nf_bridge->data, ETH_HLEN-ETH_ALEN);
282 /* tell br_dev_xmit to continue with forwarding */
283 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
284 return dst->neighbour->output(skb);
285 }
286free_skb:
287 kfree_skb(skb);
288 return 0;
289}
290
291/* This requires some explaining. If DNAT has taken place,
292 * we will need to fix up the destination Ethernet address.
293 *
294 * There are two cases to consider:
295 * 1. The packet was DNAT'ed to a device in the same bridge
296 * port group as it was received on. We can still bridge
297 * the packet.
298 * 2. The packet was DNAT'ed to a different device, either
299 * a non-bridged device or another bridge port group.
300 * The packet will need to be routed.
301 *
302 * The correct way of distinguishing between these two cases is to
303 * call ip_route_input() and to look at skb->dst->dev, which is
304 * changed to the destination device if ip_route_input() succeeds.
305 *
306 * Let's first consider the case that ip_route_input() succeeds:
307 *
308 * If the output device equals the logical bridge device the packet
309 * came in on, we can consider this bridging. The corresponding MAC
310 * address will be obtained in br_nf_pre_routing_finish_bridge.
311 * Otherwise, the packet is considered to be routed and we just
312 * change the destination MAC address so that the packet will
313 * later be passed up to the IP stack to be routed. For a redirected
314 * packet, ip_route_input() will give back the localhost as output device,
315 * which differs from the bridge device.
316 *
317 * Let's now consider the case that ip_route_input() fails:
318 *
319 * This can be because the destination address is martian, in which case
320 * the packet will be dropped.
321 * If IP forwarding is disabled, ip_route_input() will fail, while
322 * ip_route_output_key() can return success. The source
323 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
324 * thinks we're handling a locally generated packet and won't care
325 * if IP forwarding is enabled. If the output device equals the logical bridge
326 * device, we proceed as if ip_route_input() succeeded. If it differs from the
327 * logical bridge port or if ip_route_output_key() fails we drop the packet.
328 */
329static int br_nf_pre_routing_finish(struct sk_buff *skb)
330{
331 struct net_device *dev = skb->dev;
332 struct iphdr *iph = ip_hdr(skb);
333 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
334 struct rtable *rt;
335 int err;
336
337 if (nf_bridge->mask & BRNF_PKT_TYPE) {
338 skb->pkt_type = PACKET_OTHERHOST;
339 nf_bridge->mask ^= BRNF_PKT_TYPE;
340 }
341 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
342 if (dnat_took_place(skb)) {
343 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
344 struct flowi fl = {
345 .nl_u = {
346 .ip4_u = {
347 .daddr = iph->daddr,
348 .saddr = 0,
349 .tos = RT_TOS(iph->tos) },
350 },
351 .proto = 0,
352 };
353 struct in_device *in_dev = __in_dev_get_rcu(dev);
354
355 /* If err equals -EHOSTUNREACH the error is due to a
356 * martian destination or due to the fact that
357 * forwarding is disabled. For most martian packets,
358 * ip_route_output_key() will fail. It won't fail for 2 types of
359 * martian destinations: loopback destinations and destination
360 * 0.0.0.0. In both cases the packet will be dropped because the
361 * destination is the loopback device and not the bridge. */
362 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
363 goto free_skb;
364
365 if (!ip_route_output_key(dev_net(dev), &rt, &fl)) {
366 /* - Bridged-and-DNAT'ed traffic doesn't
367 * require ip_forwarding. */
368 if (((struct dst_entry *)rt)->dev == dev) {
369 skb_dst_set(skb, (struct dst_entry *)rt);
370 goto bridged_dnat;
371 }
372 dst_release((struct dst_entry *)rt);
373 }
374free_skb:
375 kfree_skb(skb);
376 return 0;
377 } else {
378 if (skb_dst(skb)->dev == dev) {
379bridged_dnat:
380 skb->dev = nf_bridge->physindev;
381 nf_bridge_update_protocol(skb);
382 nf_bridge_push_encap_header(skb);
383 NF_HOOK_THRESH(NFPROTO_BRIDGE,
384 NF_BR_PRE_ROUTING,
385 skb, skb->dev, NULL,
386 br_nf_pre_routing_finish_bridge,
387 1);
388 return 0;
389 }
390 memcpy(eth_hdr(skb)->h_dest, dev->dev_addr, ETH_ALEN);
391 skb->pkt_type = PACKET_HOST;
392 }
393 } else {
394 rt = bridge_parent_rtable(nf_bridge->physindev);
395 if (!rt) {
396 kfree_skb(skb);
397 return 0;
398 }
399 dst_hold(&rt->u.dst);
400 skb_dst_set(skb, &rt->u.dst);
401 }
402
403 skb->dev = nf_bridge->physindev;
404 nf_bridge_update_protocol(skb);
405 nf_bridge_push_encap_header(skb);
406 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
407 br_handle_frame_finish, 1);
408
409 return 0;
410}
411
412/* Some common code for IPv4/IPv6 */
413static struct net_device *setup_pre_routing(struct sk_buff *skb)
414{
415 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
416
417 if (skb->pkt_type == PACKET_OTHERHOST) {
418 skb->pkt_type = PACKET_HOST;
419 nf_bridge->mask |= BRNF_PKT_TYPE;
420 }
421
422 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
423 nf_bridge->physindev = skb->dev;
424 skb->dev = bridge_parent(skb->dev);
425 if (skb->protocol == htons(ETH_P_8021Q))
426 nf_bridge->mask |= BRNF_8021Q;
427 else if (skb->protocol == htons(ETH_P_PPP_SES))
428 nf_bridge->mask |= BRNF_PPPoE;
429
430 return skb->dev;
431}
432
433/* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
434static int check_hbh_len(struct sk_buff *skb)
435{
436 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
437 u32 pkt_len;
438 const unsigned char *nh = skb_network_header(skb);
439 int off = raw - nh;
440 int len = (raw[1] + 1) << 3;
441
442 if ((raw + len) - skb->data > skb_headlen(skb))
443 goto bad;
444
445 off += 2;
446 len -= 2;
447
448 while (len > 0) {
449 int optlen = nh[off + 1] + 2;
450
451 switch (nh[off]) {
452 case IPV6_TLV_PAD0:
453 optlen = 1;
454 break;
455
456 case IPV6_TLV_PADN:
457 break;
458
459 case IPV6_TLV_JUMBO:
460 if (nh[off + 1] != 4 || (off & 3) != 2)
461 goto bad;
462 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
463 if (pkt_len <= IPV6_MAXPLEN ||
464 ipv6_hdr(skb)->payload_len)
465 goto bad;
466 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
467 goto bad;
468 if (pskb_trim_rcsum(skb,
469 pkt_len + sizeof(struct ipv6hdr)))
470 goto bad;
471 nh = skb_network_header(skb);
472 break;
473 default:
474 if (optlen > len)
475 goto bad;
476 break;
477 }
478 off += optlen;
479 len -= optlen;
480 }
481 if (len == 0)
482 return 0;
483bad:
484 return -1;
485
486}
487
488/* Replicate the checks that IPv6 does on packet reception and pass the packet
489 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
490static unsigned int br_nf_pre_routing_ipv6(unsigned int hook,
491 struct sk_buff *skb,
492 const struct net_device *in,
493 const struct net_device *out,
494 int (*okfn)(struct sk_buff *))
495{
496 struct ipv6hdr *hdr;
497 u32 pkt_len;
498
499 if (skb->len < sizeof(struct ipv6hdr))
500 goto inhdr_error;
501
502 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
503 goto inhdr_error;
504
505 hdr = ipv6_hdr(skb);
506
507 if (hdr->version != 6)
508 goto inhdr_error;
509
510 pkt_len = ntohs(hdr->payload_len);
511
512 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
513 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
514 goto inhdr_error;
515 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
516 goto inhdr_error;
517 }
518 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
519 goto inhdr_error;
520
521 nf_bridge_put(skb->nf_bridge);
522 if (!nf_bridge_alloc(skb))
523 return NF_DROP;
524 if (!setup_pre_routing(skb))
525 return NF_DROP;
526
527 skb->protocol = htons(ETH_P_IPV6);
528 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
529 br_nf_pre_routing_finish_ipv6);
530
531 return NF_STOLEN;
532
533inhdr_error:
534 return NF_DROP;
535}
536
537/* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
538 * Replicate the checks that IPv4 does on packet reception.
539 * Set skb->dev to the bridge device (i.e. parent of the
540 * receiving device) to make netfilter happy, the REDIRECT
541 * target in particular. Save the original destination IP
542 * address to be able to detect DNAT afterwards. */
543static unsigned int br_nf_pre_routing(unsigned int hook, struct sk_buff *skb,
544 const struct net_device *in,
545 const struct net_device *out,
546 int (*okfn)(struct sk_buff *))
547{
548 struct iphdr *iph;
549 __u32 len = nf_bridge_encap_header_len(skb);
550
551 if (unlikely(!pskb_may_pull(skb, len)))
552 goto out;
553
554 if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
555 IS_PPPOE_IPV6(skb)) {
556#ifdef CONFIG_SYSCTL
557 if (!brnf_call_ip6tables)
558 return NF_ACCEPT;
559#endif
560 nf_bridge_pull_encap_header_rcsum(skb);
561 return br_nf_pre_routing_ipv6(hook, skb, in, out, okfn);
562 }
563#ifdef CONFIG_SYSCTL
564 if (!brnf_call_iptables)
565 return NF_ACCEPT;
566#endif
567
568 if (skb->protocol != htons(ETH_P_IP) && !IS_VLAN_IP(skb) &&
569 !IS_PPPOE_IP(skb))
570 return NF_ACCEPT;
571
572 nf_bridge_pull_encap_header_rcsum(skb);
573
574 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
575 goto inhdr_error;
576
577 iph = ip_hdr(skb);
578 if (iph->ihl < 5 || iph->version != 4)
579 goto inhdr_error;
580
581 if (!pskb_may_pull(skb, 4 * iph->ihl))
582 goto inhdr_error;
583
584 iph = ip_hdr(skb);
585 if (ip_fast_csum((__u8 *) iph, iph->ihl) != 0)
586 goto inhdr_error;
587
588 len = ntohs(iph->tot_len);
589 if (skb->len < len || len < 4 * iph->ihl)
590 goto inhdr_error;
591
592 pskb_trim_rcsum(skb, len);
593
594 nf_bridge_put(skb->nf_bridge);
595 if (!nf_bridge_alloc(skb))
596 return NF_DROP;
597 if (!setup_pre_routing(skb))
598 return NF_DROP;
599 store_orig_dstaddr(skb);
600 skb->protocol = htons(ETH_P_IP);
601
602 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
603 br_nf_pre_routing_finish);
604
605 return NF_STOLEN;
606
607inhdr_error:
608// IP_INC_STATS_BH(IpInHdrErrors);
609out:
610 return NF_DROP;
611}
612
613
614/* PF_BRIDGE/LOCAL_IN ************************************************/
615/* The packet is locally destined, which requires a real
616 * dst_entry, so detach the fake one. On the way up, the
617 * packet would pass through PRE_ROUTING again (which already
618 * took place when the packet entered the bridge), but we
619 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
620 * prevent this from happening. */
621static unsigned int br_nf_local_in(unsigned int hook, struct sk_buff *skb,
622 const struct net_device *in,
623 const struct net_device *out,
624 int (*okfn)(struct sk_buff *))
625{
626 struct rtable *rt = skb_rtable(skb);
627
628 if (rt && rt == bridge_parent_rtable(in))
629 skb_dst_drop(skb);
630
631 return NF_ACCEPT;
632}
633
634/* PF_BRIDGE/FORWARD *************************************************/
635static int br_nf_forward_finish(struct sk_buff *skb)
636{
637 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
638 struct net_device *in;
639
640 if (skb->protocol != htons(ETH_P_ARP) && !IS_VLAN_ARP(skb)) {
641 in = nf_bridge->physindev;
642 if (nf_bridge->mask & BRNF_PKT_TYPE) {
643 skb->pkt_type = PACKET_OTHERHOST;
644 nf_bridge->mask ^= BRNF_PKT_TYPE;
645 }
646 nf_bridge_update_protocol(skb);
647 } else {
648 in = *((struct net_device **)(skb->cb));
649 }
650 nf_bridge_push_encap_header(skb);
651
652 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
653 skb->dev, br_forward_finish, 1);
654 return 0;
655}
656
657/* This is the 'purely bridged' case. For IP, we pass the packet to
658 * netfilter with indev and outdev set to the bridge device,
659 * but we are still able to filter on the 'real' indev/outdev
660 * because of the physdev module. For ARP, indev and outdev are the
661 * bridge ports. */
662static unsigned int br_nf_forward_ip(unsigned int hook, struct sk_buff *skb,
663 const struct net_device *in,
664 const struct net_device *out,
665 int (*okfn)(struct sk_buff *))
666{
667 struct nf_bridge_info *nf_bridge;
668 struct net_device *parent;
669 u_int8_t pf;
670
671 if (!skb->nf_bridge)
672 return NF_ACCEPT;
673
674 /* Need exclusive nf_bridge_info since we might have multiple
675 * different physoutdevs. */
676 if (!nf_bridge_unshare(skb))
677 return NF_DROP;
678
679 parent = bridge_parent(out);
680 if (!parent)
681 return NF_DROP;
682
683 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
684 IS_PPPOE_IP(skb))
685 pf = PF_INET;
686 else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
687 IS_PPPOE_IPV6(skb))
688 pf = PF_INET6;
689 else
690 return NF_ACCEPT;
691
692 nf_bridge_pull_encap_header(skb);
693
694 nf_bridge = skb->nf_bridge;
695 if (skb->pkt_type == PACKET_OTHERHOST) {
696 skb->pkt_type = PACKET_HOST;
697 nf_bridge->mask |= BRNF_PKT_TYPE;
698 }
699
700 /* The physdev module checks on this */
701 nf_bridge->mask |= BRNF_BRIDGED;
702 nf_bridge->physoutdev = skb->dev;
703 if (pf == PF_INET)
704 skb->protocol = htons(ETH_P_IP);
705 else
706 skb->protocol = htons(ETH_P_IPV6);
707
708 NF_HOOK(pf, NF_INET_FORWARD, skb, bridge_parent(in), parent,
709 br_nf_forward_finish);
710
711 return NF_STOLEN;
712}
713
714static unsigned int br_nf_forward_arp(unsigned int hook, struct sk_buff *skb,
715 const struct net_device *in,
716 const struct net_device *out,
717 int (*okfn)(struct sk_buff *))
718{
719 struct net_device **d = (struct net_device **)(skb->cb);
720
721#ifdef CONFIG_SYSCTL
722 if (!brnf_call_arptables)
723 return NF_ACCEPT;
724#endif
725
726 if (skb->protocol != htons(ETH_P_ARP)) {
727 if (!IS_VLAN_ARP(skb))
728 return NF_ACCEPT;
729 nf_bridge_pull_encap_header(skb);
730 }
731
732 if (arp_hdr(skb)->ar_pln != 4) {
733 if (IS_VLAN_ARP(skb))
734 nf_bridge_push_encap_header(skb);
735 return NF_ACCEPT;
736 }
737 *d = (struct net_device *)in;
738 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
739 (struct net_device *)out, br_nf_forward_finish);
740
741 return NF_STOLEN;
742}
743
744#if defined(CONFIG_NF_CONNTRACK_IPV4) || defined(CONFIG_NF_CONNTRACK_IPV4_MODULE)
745static int br_nf_dev_queue_xmit(struct sk_buff *skb)
746{
747 if (skb->nfct != NULL && skb->protocol == htons(ETH_P_IP) &&
748 skb->len + nf_bridge_mtu_reduction(skb) > skb->dev->mtu &&
749 !skb_is_gso(skb))
750 return ip_fragment(skb, br_dev_queue_push_xmit);
751 else
752 return br_dev_queue_push_xmit(skb);
753}
754#else
755static int br_nf_dev_queue_xmit(struct sk_buff *skb)
756{
757 return br_dev_queue_push_xmit(skb);
758}
759#endif
760
761/* PF_BRIDGE/POST_ROUTING ********************************************/
762static unsigned int br_nf_post_routing(unsigned int hook, struct sk_buff *skb,
763 const struct net_device *in,
764 const struct net_device *out,
765 int (*okfn)(struct sk_buff *))
766{
767 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
768 struct net_device *realoutdev = bridge_parent(skb->dev);
769 u_int8_t pf;
770
771 if (!nf_bridge || !(nf_bridge->mask & BRNF_BRIDGED))
772 return NF_ACCEPT;
773
774 if (!realoutdev)
775 return NF_DROP;
776
777 if (skb->protocol == htons(ETH_P_IP) || IS_VLAN_IP(skb) ||
778 IS_PPPOE_IP(skb))
779 pf = PF_INET;
780 else if (skb->protocol == htons(ETH_P_IPV6) || IS_VLAN_IPV6(skb) ||
781 IS_PPPOE_IPV6(skb))
782 pf = PF_INET6;
783 else
784 return NF_ACCEPT;
785
786 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
787 * about the value of skb->pkt_type. */
788 if (skb->pkt_type == PACKET_OTHERHOST) {
789 skb->pkt_type = PACKET_HOST;
790 nf_bridge->mask |= BRNF_PKT_TYPE;
791 }
792
793 nf_bridge_pull_encap_header(skb);
794 nf_bridge_save_header(skb);
795 if (pf == PF_INET)
796 skb->protocol = htons(ETH_P_IP);
797 else
798 skb->protocol = htons(ETH_P_IPV6);
799
800 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
801 br_nf_dev_queue_xmit);
802
803 return NF_STOLEN;
804}
805
806/* IP/SABOTAGE *****************************************************/
807/* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
808 * for the second time. */
809static unsigned int ip_sabotage_in(unsigned int hook, struct sk_buff *skb,
810 const struct net_device *in,
811 const struct net_device *out,
812 int (*okfn)(struct sk_buff *))
813{
814 if (skb->nf_bridge &&
815 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
816 return NF_STOP;
817 }
818
819 return NF_ACCEPT;
820}
821
822/* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
823 * br_dev_queue_push_xmit is called afterwards */
824static struct nf_hook_ops br_nf_ops[] __read_mostly = {
825 {
826 .hook = br_nf_pre_routing,
827 .owner = THIS_MODULE,
828 .pf = PF_BRIDGE,
829 .hooknum = NF_BR_PRE_ROUTING,
830 .priority = NF_BR_PRI_BRNF,
831 },
832 {
833 .hook = br_nf_local_in,
834 .owner = THIS_MODULE,
835 .pf = PF_BRIDGE,
836 .hooknum = NF_BR_LOCAL_IN,
837 .priority = NF_BR_PRI_BRNF,
838 },
839 {
840 .hook = br_nf_forward_ip,
841 .owner = THIS_MODULE,
842 .pf = PF_BRIDGE,
843 .hooknum = NF_BR_FORWARD,
844 .priority = NF_BR_PRI_BRNF - 1,
845 },
846 {
847 .hook = br_nf_forward_arp,
848 .owner = THIS_MODULE,
849 .pf = PF_BRIDGE,
850 .hooknum = NF_BR_FORWARD,
851 .priority = NF_BR_PRI_BRNF,
852 },
853 {
854 .hook = br_nf_post_routing,
855 .owner = THIS_MODULE,
856 .pf = PF_BRIDGE,
857 .hooknum = NF_BR_POST_ROUTING,
858 .priority = NF_BR_PRI_LAST,
859 },
860 {
861 .hook = ip_sabotage_in,
862 .owner = THIS_MODULE,
863 .pf = PF_INET,
864 .hooknum = NF_INET_PRE_ROUTING,
865 .priority = NF_IP_PRI_FIRST,
866 },
867 {
868 .hook = ip_sabotage_in,
869 .owner = THIS_MODULE,
870 .pf = PF_INET6,
871 .hooknum = NF_INET_PRE_ROUTING,
872 .priority = NF_IP6_PRI_FIRST,
873 },
874};
875
876#ifdef CONFIG_SYSCTL
877static
878int brnf_sysctl_call_tables(ctl_table * ctl, int write,
879 void __user * buffer, size_t * lenp, loff_t * ppos)
880{
881 int ret;
882
883 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
884
885 if (write && *(int *)(ctl->data))
886 *(int *)(ctl->data) = 1;
887 return ret;
888}
889
890static ctl_table brnf_table[] = {
891 {
892 .procname = "bridge-nf-call-arptables",
893 .data = &brnf_call_arptables,
894 .maxlen = sizeof(int),
895 .mode = 0644,
896 .proc_handler = brnf_sysctl_call_tables,
897 },
898 {
899 .procname = "bridge-nf-call-iptables",
900 .data = &brnf_call_iptables,
901 .maxlen = sizeof(int),
902 .mode = 0644,
903 .proc_handler = brnf_sysctl_call_tables,
904 },
905 {
906 .procname = "bridge-nf-call-ip6tables",
907 .data = &brnf_call_ip6tables,
908 .maxlen = sizeof(int),
909 .mode = 0644,
910 .proc_handler = brnf_sysctl_call_tables,
911 },
912 {
913 .procname = "bridge-nf-filter-vlan-tagged",
914 .data = &brnf_filter_vlan_tagged,
915 .maxlen = sizeof(int),
916 .mode = 0644,
917 .proc_handler = brnf_sysctl_call_tables,
918 },
919 {
920 .procname = "bridge-nf-filter-pppoe-tagged",
921 .data = &brnf_filter_pppoe_tagged,
922 .maxlen = sizeof(int),
923 .mode = 0644,
924 .proc_handler = brnf_sysctl_call_tables,
925 },
926 { }
927};
928
929static struct ctl_path brnf_path[] = {
930 { .procname = "net", },
931 { .procname = "bridge", },
932 { }
933};
934#endif
935
936int __init br_netfilter_init(void)
937{
938 int ret;
939
940 ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
941 if (ret < 0)
942 return ret;
943#ifdef CONFIG_SYSCTL
944 brnf_sysctl_header = register_sysctl_paths(brnf_path, brnf_table);
945 if (brnf_sysctl_header == NULL) {
946 printk(KERN_WARNING
947 "br_netfilter: can't register to sysctl.\n");
948 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
949 return -ENOMEM;
950 }
951#endif
952 printk(KERN_NOTICE "Bridge firewalling registered\n");
953 return 0;
954}
955
956void br_netfilter_fini(void)
957{
958 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
959#ifdef CONFIG_SYSCTL
960 unregister_sysctl_table(brnf_sysctl_header);
961#endif
962}
Clone of davem's net-next-2.6 tree