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