3 * Linux ethernet bridge
6 * Lennert Buytenhek <buytenh@gnu.org>
7 * Bart De Schuymer <bdschuym@pandora.be>
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.
14 * Lennert dedicates this file to Kerstin Wurdinger.
17 #include <linux/module.h>
18 #include <linux/kernel.h>
19 #include <linux/slab.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>
37 #include <net/route.h>
38 #include <net/netfilter/br_netfilter.h>
40 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
41 #include <net/netfilter/nf_conntrack.h>
44 #include <asm/uaccess.h>
45 #include "br_private.h"
47 #include <linux/sysctl.h>
51 static struct ctl_table_header *brnf_sysctl_header;
52 static int brnf_call_iptables __read_mostly = 1;
53 static int brnf_call_ip6tables __read_mostly = 1;
54 static int brnf_call_arptables __read_mostly = 1;
55 static int brnf_filter_vlan_tagged __read_mostly = 0;
56 static int brnf_filter_pppoe_tagged __read_mostly = 0;
57 static int brnf_pass_vlan_indev __read_mostly = 0;
59 #define brnf_call_iptables 1
60 #define brnf_call_ip6tables 1
61 #define brnf_call_arptables 1
62 #define brnf_filter_vlan_tagged 0
63 #define brnf_filter_pppoe_tagged 0
64 #define brnf_pass_vlan_indev 0
68 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
70 #define IS_IPV6(skb) \
71 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
74 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
76 static inline __be16 vlan_proto(const struct sk_buff *skb)
78 if (skb_vlan_tag_present(skb))
80 else if (skb->protocol == htons(ETH_P_8021Q))
81 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
86 #define IS_VLAN_IP(skb) \
87 (vlan_proto(skb) == htons(ETH_P_IP) && \
88 brnf_filter_vlan_tagged)
90 #define IS_VLAN_IPV6(skb) \
91 (vlan_proto(skb) == htons(ETH_P_IPV6) && \
92 brnf_filter_vlan_tagged)
94 #define IS_VLAN_ARP(skb) \
95 (vlan_proto(skb) == htons(ETH_P_ARP) && \
96 brnf_filter_vlan_tagged)
98 static inline __be16 pppoe_proto(const struct sk_buff *skb)
100 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
101 sizeof(struct pppoe_hdr)));
104 #define IS_PPPOE_IP(skb) \
105 (skb->protocol == htons(ETH_P_PPP_SES) && \
106 pppoe_proto(skb) == htons(PPP_IP) && \
107 brnf_filter_pppoe_tagged)
109 #define IS_PPPOE_IPV6(skb) \
110 (skb->protocol == htons(ETH_P_PPP_SES) && \
111 pppoe_proto(skb) == htons(PPP_IPV6) && \
112 brnf_filter_pppoe_tagged)
114 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
115 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
117 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
118 struct brnf_frag_data {
119 char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
124 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage);
127 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
129 struct net_bridge_port *port;
131 port = br_port_get_rcu(dev);
132 return port ? &port->br->fake_rtable : NULL;
135 static inline struct net_device *bridge_parent(const struct net_device *dev)
137 struct net_bridge_port *port;
139 port = br_port_get_rcu(dev);
140 return port ? port->br->dev : NULL;
143 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
145 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
146 if (likely(skb->nf_bridge))
147 atomic_set(&(skb->nf_bridge->use), 1);
149 return skb->nf_bridge;
152 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
154 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
156 if (atomic_read(&nf_bridge->use) > 1) {
157 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
160 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
161 atomic_set(&tmp->use, 1);
163 nf_bridge_put(nf_bridge);
169 static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
171 switch (skb->protocol) {
172 case __cpu_to_be16(ETH_P_8021Q):
174 case __cpu_to_be16(ETH_P_PPP_SES):
175 return PPPOE_SES_HLEN;
181 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
183 unsigned int len = nf_bridge_encap_header_len(skb);
186 skb->network_header -= len;
189 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
191 unsigned int len = nf_bridge_encap_header_len(skb);
194 skb->network_header += len;
197 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
199 unsigned int len = nf_bridge_encap_header_len(skb);
201 skb_pull_rcsum(skb, len);
202 skb->network_header += len;
205 /* When handing a packet over to the IP layer
206 * check whether we have a skb that is in the
210 static int br_parse_ip_options(struct sk_buff *skb)
212 const struct iphdr *iph;
213 struct net_device *dev = skb->dev;
216 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
221 /* Basic sanity checks */
222 if (iph->ihl < 5 || iph->version != 4)
225 if (!pskb_may_pull(skb, iph->ihl*4))
229 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
232 len = ntohs(iph->tot_len);
233 if (skb->len < len) {
234 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
236 } else if (len < (iph->ihl*4))
239 if (pskb_trim_rcsum(skb, len)) {
240 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
244 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
245 /* We should really parse IP options here but until
246 * somebody who actually uses IP options complains to
247 * us we'll just silently ignore the options because
253 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
258 static void nf_bridge_update_protocol(struct sk_buff *skb)
260 if (skb->nf_bridge->mask & BRNF_8021Q)
261 skb->protocol = htons(ETH_P_8021Q);
262 else if (skb->nf_bridge->mask & BRNF_PPPoE)
263 skb->protocol = htons(ETH_P_PPP_SES);
266 /* PF_BRIDGE/PRE_ROUTING *********************************************/
267 /* Undo the changes made for ip6tables PREROUTING and continue the
268 * bridge PRE_ROUTING hook. */
269 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
271 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
274 if (nf_bridge->mask & BRNF_PKT_TYPE) {
275 skb->pkt_type = PACKET_OTHERHOST;
276 nf_bridge->mask ^= BRNF_PKT_TYPE;
278 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
280 rt = bridge_parent_rtable(nf_bridge->physindev);
285 skb_dst_set_noref(skb, &rt->dst);
287 skb->dev = nf_bridge->physindev;
288 nf_bridge_update_protocol(skb);
289 nf_bridge_push_encap_header(skb);
290 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
291 br_handle_frame_finish, 1);
296 /* Obtain the correct destination MAC address, while preserving the original
297 * source MAC address. If we already know this address, we just copy it. If we
298 * don't, we use the neighbour framework to find out. In both cases, we make
299 * sure that br_handle_frame_finish() is called afterwards.
301 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
303 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
304 struct neighbour *neigh;
305 struct dst_entry *dst;
307 skb->dev = bridge_parent(skb->dev);
311 neigh = dst_neigh_lookup_skb(dst, skb);
315 if (neigh->hh.hh_len) {
316 neigh_hh_bridge(&neigh->hh, skb);
317 skb->dev = nf_bridge->physindev;
318 ret = br_handle_frame_finish(skb);
320 /* the neighbour function below overwrites the complete
321 * MAC header, so we save the Ethernet source address and
324 skb_copy_from_linear_data_offset(skb,
325 -(ETH_HLEN-ETH_ALEN),
326 nf_bridge->neigh_header,
328 /* tell br_dev_xmit to continue with forwarding */
329 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
330 /* FIXME Need to refragment */
331 ret = neigh->output(neigh, skb);
333 neigh_release(neigh);
341 static bool dnat_took_place(const struct sk_buff *skb)
343 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
344 enum ip_conntrack_info ctinfo;
347 ct = nf_ct_get(skb, &ctinfo);
348 if (!ct || nf_ct_is_untracked(ct))
351 return test_bit(IPS_DST_NAT_BIT, &ct->status);
357 /* This requires some explaining. If DNAT has taken place,
358 * we will need to fix up the destination Ethernet address.
360 * There are two cases to consider:
361 * 1. The packet was DNAT'ed to a device in the same bridge
362 * port group as it was received on. We can still bridge
364 * 2. The packet was DNAT'ed to a different device, either
365 * a non-bridged device or another bridge port group.
366 * The packet will need to be routed.
368 * The correct way of distinguishing between these two cases is to
369 * call ip_route_input() and to look at skb->dst->dev, which is
370 * changed to the destination device if ip_route_input() succeeds.
372 * Let's first consider the case that ip_route_input() succeeds:
374 * If the output device equals the logical bridge device the packet
375 * came in on, we can consider this bridging. The corresponding MAC
376 * address will be obtained in br_nf_pre_routing_finish_bridge.
377 * Otherwise, the packet is considered to be routed and we just
378 * change the destination MAC address so that the packet will
379 * later be passed up to the IP stack to be routed. For a redirected
380 * packet, ip_route_input() will give back the localhost as output device,
381 * which differs from the bridge device.
383 * Let's now consider the case that ip_route_input() fails:
385 * This can be because the destination address is martian, in which case
386 * the packet will be dropped.
387 * If IP forwarding is disabled, ip_route_input() will fail, while
388 * ip_route_output_key() can return success. The source
389 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
390 * thinks we're handling a locally generated packet and won't care
391 * if IP forwarding is enabled. If the output device equals the logical bridge
392 * device, we proceed as if ip_route_input() succeeded. If it differs from the
393 * logical bridge port or if ip_route_output_key() fails we drop the packet.
395 static int br_nf_pre_routing_finish(struct sk_buff *skb)
397 struct net_device *dev = skb->dev;
398 struct iphdr *iph = ip_hdr(skb);
399 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
404 frag_max_size = IPCB(skb)->frag_max_size;
405 BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size;
407 if (nf_bridge->mask & BRNF_PKT_TYPE) {
408 skb->pkt_type = PACKET_OTHERHOST;
409 nf_bridge->mask ^= BRNF_PKT_TYPE;
411 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
412 if (dnat_took_place(skb)) {
413 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
414 struct in_device *in_dev = __in_dev_get_rcu(dev);
416 /* If err equals -EHOSTUNREACH the error is due to a
417 * martian destination or due to the fact that
418 * forwarding is disabled. For most martian packets,
419 * ip_route_output_key() will fail. It won't fail for 2 types of
420 * martian destinations: loopback destinations and destination
421 * 0.0.0.0. In both cases the packet will be dropped because the
422 * destination is the loopback device and not the bridge. */
423 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
426 rt = ip_route_output(dev_net(dev), iph->daddr, 0,
427 RT_TOS(iph->tos), 0);
429 /* - Bridged-and-DNAT'ed traffic doesn't
430 * require ip_forwarding. */
431 if (rt->dst.dev == dev) {
432 skb_dst_set(skb, &rt->dst);
441 if (skb_dst(skb)->dev == dev) {
443 skb->dev = nf_bridge->physindev;
444 nf_bridge_update_protocol(skb);
445 nf_bridge_push_encap_header(skb);
446 NF_HOOK_THRESH(NFPROTO_BRIDGE,
449 br_nf_pre_routing_finish_bridge,
453 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
454 skb->pkt_type = PACKET_HOST;
457 rt = bridge_parent_rtable(nf_bridge->physindev);
462 skb_dst_set_noref(skb, &rt->dst);
465 skb->dev = nf_bridge->physindev;
466 nf_bridge_update_protocol(skb);
467 nf_bridge_push_encap_header(skb);
468 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
469 br_handle_frame_finish, 1);
474 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
476 struct net_device *vlan, *br;
478 br = bridge_parent(dev);
479 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
482 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
483 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
485 return vlan ? vlan : br;
488 /* Some common code for IPv4/IPv6 */
489 static struct net_device *setup_pre_routing(struct sk_buff *skb)
491 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
493 if (skb->pkt_type == PACKET_OTHERHOST) {
494 skb->pkt_type = PACKET_HOST;
495 nf_bridge->mask |= BRNF_PKT_TYPE;
498 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
499 nf_bridge->physindev = skb->dev;
500 skb->dev = brnf_get_logical_dev(skb, skb->dev);
501 if (skb->protocol == htons(ETH_P_8021Q))
502 nf_bridge->mask |= BRNF_8021Q;
503 else if (skb->protocol == htons(ETH_P_PPP_SES))
504 nf_bridge->mask |= BRNF_PPPoE;
506 /* Must drop socket now because of tproxy. */
511 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
512 static int check_hbh_len(struct sk_buff *skb)
514 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
516 const unsigned char *nh = skb_network_header(skb);
518 int len = (raw[1] + 1) << 3;
520 if ((raw + len) - skb->data > skb_headlen(skb))
527 int optlen = nh[off + 1] + 2;
538 if (nh[off + 1] != 4 || (off & 3) != 2)
540 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
541 if (pkt_len <= IPV6_MAXPLEN ||
542 ipv6_hdr(skb)->payload_len)
544 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
546 if (pskb_trim_rcsum(skb,
547 pkt_len + sizeof(struct ipv6hdr)))
549 nh = skb_network_header(skb);
566 /* Replicate the checks that IPv6 does on packet reception and pass the packet
567 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
568 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
570 const struct net_device *in,
571 const struct net_device *out,
572 int (*okfn)(struct sk_buff *))
574 const struct ipv6hdr *hdr;
577 if (skb->len < sizeof(struct ipv6hdr))
580 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
585 if (hdr->version != 6)
588 pkt_len = ntohs(hdr->payload_len);
590 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
591 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
593 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
596 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
599 nf_bridge_put(skb->nf_bridge);
600 if (!nf_bridge_alloc(skb))
602 if (!setup_pre_routing(skb))
605 skb->protocol = htons(ETH_P_IPV6);
606 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
607 br_nf_pre_routing_finish_ipv6);
612 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
613 * Replicate the checks that IPv4 does on packet reception.
614 * Set skb->dev to the bridge device (i.e. parent of the
615 * receiving device) to make netfilter happy, the REDIRECT
616 * target in particular. Save the original destination IP
617 * address to be able to detect DNAT afterwards. */
618 static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
620 const struct net_device *in,
621 const struct net_device *out,
622 int (*okfn)(struct sk_buff *))
624 struct net_bridge_port *p;
625 struct net_bridge *br;
626 __u32 len = nf_bridge_encap_header_len(skb);
628 if (unlikely(!pskb_may_pull(skb, len)))
631 p = br_port_get_rcu(in);
636 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
637 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
640 nf_bridge_pull_encap_header_rcsum(skb);
641 return br_nf_pre_routing_ipv6(ops, skb, in, out, okfn);
644 if (!brnf_call_iptables && !br->nf_call_iptables)
647 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
650 nf_bridge_pull_encap_header_rcsum(skb);
652 if (br_parse_ip_options(skb))
655 nf_bridge_put(skb->nf_bridge);
656 if (!nf_bridge_alloc(skb))
658 if (!setup_pre_routing(skb))
661 skb->protocol = htons(ETH_P_IP);
663 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
664 br_nf_pre_routing_finish);
670 /* PF_BRIDGE/LOCAL_IN ************************************************/
671 /* The packet is locally destined, which requires a real
672 * dst_entry, so detach the fake one. On the way up, the
673 * packet would pass through PRE_ROUTING again (which already
674 * took place when the packet entered the bridge), but we
675 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
676 * prevent this from happening. */
677 static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
679 const struct net_device *in,
680 const struct net_device *out,
681 int (*okfn)(struct sk_buff *))
683 br_drop_fake_rtable(skb);
687 /* PF_BRIDGE/FORWARD *************************************************/
688 static int br_nf_forward_finish(struct sk_buff *skb)
690 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
691 struct net_device *in;
693 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
696 if (skb->protocol == htons(ETH_P_IP)) {
697 frag_max_size = IPCB(skb)->frag_max_size;
698 BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size;
701 in = nf_bridge->physindev;
702 if (nf_bridge->mask & BRNF_PKT_TYPE) {
703 skb->pkt_type = PACKET_OTHERHOST;
704 nf_bridge->mask ^= BRNF_PKT_TYPE;
706 nf_bridge_update_protocol(skb);
708 in = *((struct net_device **)(skb->cb));
710 nf_bridge_push_encap_header(skb);
712 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
713 skb->dev, br_forward_finish, 1);
718 /* This is the 'purely bridged' case. For IP, we pass the packet to
719 * netfilter with indev and outdev set to the bridge device,
720 * but we are still able to filter on the 'real' indev/outdev
721 * because of the physdev module. For ARP, indev and outdev are the
723 static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
725 const struct net_device *in,
726 const struct net_device *out,
727 int (*okfn)(struct sk_buff *))
729 struct nf_bridge_info *nf_bridge;
730 struct net_device *parent;
736 /* Need exclusive nf_bridge_info since we might have multiple
737 * different physoutdevs. */
738 if (!nf_bridge_unshare(skb))
741 parent = bridge_parent(out);
745 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
747 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
752 nf_bridge_pull_encap_header(skb);
754 nf_bridge = skb->nf_bridge;
755 if (skb->pkt_type == PACKET_OTHERHOST) {
756 skb->pkt_type = PACKET_HOST;
757 nf_bridge->mask |= BRNF_PKT_TYPE;
760 if (pf == NFPROTO_IPV4) {
761 int frag_max = BR_INPUT_SKB_CB(skb)->frag_max_size;
763 if (br_parse_ip_options(skb))
766 IPCB(skb)->frag_max_size = frag_max;
769 nf_bridge->physoutdev = skb->dev;
770 if (pf == NFPROTO_IPV4)
771 skb->protocol = htons(ETH_P_IP);
773 skb->protocol = htons(ETH_P_IPV6);
775 NF_HOOK(pf, NF_INET_FORWARD, skb, brnf_get_logical_dev(skb, in), parent,
776 br_nf_forward_finish);
781 static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
783 const struct net_device *in,
784 const struct net_device *out,
785 int (*okfn)(struct sk_buff *))
787 struct net_bridge_port *p;
788 struct net_bridge *br;
789 struct net_device **d = (struct net_device **)(skb->cb);
791 p = br_port_get_rcu(out);
796 if (!brnf_call_arptables && !br->nf_call_arptables)
800 if (!IS_VLAN_ARP(skb))
802 nf_bridge_pull_encap_header(skb);
805 if (arp_hdr(skb)->ar_pln != 4) {
806 if (IS_VLAN_ARP(skb))
807 nf_bridge_push_encap_header(skb);
810 *d = (struct net_device *)in;
811 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
812 (struct net_device *)out, br_nf_forward_finish);
817 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
818 static int br_nf_push_frag_xmit(struct sk_buff *skb)
820 struct brnf_frag_data *data;
823 data = this_cpu_ptr(&brnf_frag_data_storage);
824 err = skb_cow_head(skb, data->size);
831 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
832 __skb_push(skb, data->encap_size);
834 return br_dev_queue_push_xmit(skb);
837 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
841 unsigned int mtu_reserved;
843 if (skb_is_gso(skb) || skb->protocol != htons(ETH_P_IP))
844 return br_dev_queue_push_xmit(skb);
846 mtu_reserved = nf_bridge_mtu_reduction(skb);
847 /* This is wrong! We should preserve the original fragment
848 * boundaries by preserving frag_list rather than refragmenting.
850 if (skb->len + mtu_reserved > skb->dev->mtu) {
851 struct brnf_frag_data *data;
853 frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
854 if (br_parse_ip_options(skb))
855 /* Drop invalid packet */
857 IPCB(skb)->frag_max_size = frag_max_size;
859 nf_bridge_update_protocol(skb);
861 data = this_cpu_ptr(&brnf_frag_data_storage);
862 data->encap_size = nf_bridge_encap_header_len(skb);
863 data->size = ETH_HLEN + data->encap_size;
865 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
868 ret = ip_fragment(skb, br_nf_push_frag_xmit);
870 ret = br_dev_queue_push_xmit(skb);
876 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
878 return br_dev_queue_push_xmit(skb);
882 /* PF_BRIDGE/POST_ROUTING ********************************************/
883 static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
885 const struct net_device *in,
886 const struct net_device *out,
887 int (*okfn)(struct sk_buff *))
889 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
890 struct net_device *realoutdev = bridge_parent(skb->dev);
893 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
894 * on a bridge, but was delivered locally and is now being routed:
896 * POST_ROUTING was already invoked from the ip stack.
898 if (!nf_bridge || !nf_bridge->physoutdev)
904 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
906 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
911 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
912 * about the value of skb->pkt_type. */
913 if (skb->pkt_type == PACKET_OTHERHOST) {
914 skb->pkt_type = PACKET_HOST;
915 nf_bridge->mask |= BRNF_PKT_TYPE;
918 nf_bridge_pull_encap_header(skb);
919 if (pf == NFPROTO_IPV4)
920 skb->protocol = htons(ETH_P_IP);
922 skb->protocol = htons(ETH_P_IPV6);
924 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
925 br_nf_dev_queue_xmit);
930 /* IP/SABOTAGE *****************************************************/
931 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
932 * for the second time. */
933 static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
935 const struct net_device *in,
936 const struct net_device *out,
937 int (*okfn)(struct sk_buff *))
939 if (skb->nf_bridge &&
940 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
947 /* This is called when br_netfilter has called into iptables/netfilter,
948 * and DNAT has taken place on a bridge-forwarded packet.
950 * neigh->output has created a new MAC header, with local br0 MAC
953 * This restores the original MAC saddr of the bridged packet
954 * before invoking bridge forward logic to transmit the packet.
956 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
958 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
960 skb_pull(skb, ETH_HLEN);
961 nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
963 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
965 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
966 nf_bridge->neigh_header,
967 ETH_HLEN - ETH_ALEN);
968 skb->dev = nf_bridge->physindev;
969 br_handle_frame_finish(skb);
972 static int br_nf_dev_xmit(struct sk_buff *skb)
974 if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
975 br_nf_pre_routing_finish_bridge_slow(skb);
981 static const struct nf_br_ops br_ops = {
982 .br_dev_xmit_hook = br_nf_dev_xmit,
985 void br_netfilter_enable(void)
988 EXPORT_SYMBOL_GPL(br_netfilter_enable);
990 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
991 * br_dev_queue_push_xmit is called afterwards */
992 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
994 .hook = br_nf_pre_routing,
995 .owner = THIS_MODULE,
996 .pf = NFPROTO_BRIDGE,
997 .hooknum = NF_BR_PRE_ROUTING,
998 .priority = NF_BR_PRI_BRNF,
1001 .hook = br_nf_local_in,
1002 .owner = THIS_MODULE,
1003 .pf = NFPROTO_BRIDGE,
1004 .hooknum = NF_BR_LOCAL_IN,
1005 .priority = NF_BR_PRI_BRNF,
1008 .hook = br_nf_forward_ip,
1009 .owner = THIS_MODULE,
1010 .pf = NFPROTO_BRIDGE,
1011 .hooknum = NF_BR_FORWARD,
1012 .priority = NF_BR_PRI_BRNF - 1,
1015 .hook = br_nf_forward_arp,
1016 .owner = THIS_MODULE,
1017 .pf = NFPROTO_BRIDGE,
1018 .hooknum = NF_BR_FORWARD,
1019 .priority = NF_BR_PRI_BRNF,
1022 .hook = br_nf_post_routing,
1023 .owner = THIS_MODULE,
1024 .pf = NFPROTO_BRIDGE,
1025 .hooknum = NF_BR_POST_ROUTING,
1026 .priority = NF_BR_PRI_LAST,
1029 .hook = ip_sabotage_in,
1030 .owner = THIS_MODULE,
1032 .hooknum = NF_INET_PRE_ROUTING,
1033 .priority = NF_IP_PRI_FIRST,
1036 .hook = ip_sabotage_in,
1037 .owner = THIS_MODULE,
1039 .hooknum = NF_INET_PRE_ROUTING,
1040 .priority = NF_IP6_PRI_FIRST,
1044 #ifdef CONFIG_SYSCTL
1046 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1047 void __user *buffer, size_t *lenp, loff_t *ppos)
1051 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1053 if (write && *(int *)(ctl->data))
1054 *(int *)(ctl->data) = 1;
1058 static struct ctl_table brnf_table[] = {
1060 .procname = "bridge-nf-call-arptables",
1061 .data = &brnf_call_arptables,
1062 .maxlen = sizeof(int),
1064 .proc_handler = brnf_sysctl_call_tables,
1067 .procname = "bridge-nf-call-iptables",
1068 .data = &brnf_call_iptables,
1069 .maxlen = sizeof(int),
1071 .proc_handler = brnf_sysctl_call_tables,
1074 .procname = "bridge-nf-call-ip6tables",
1075 .data = &brnf_call_ip6tables,
1076 .maxlen = sizeof(int),
1078 .proc_handler = brnf_sysctl_call_tables,
1081 .procname = "bridge-nf-filter-vlan-tagged",
1082 .data = &brnf_filter_vlan_tagged,
1083 .maxlen = sizeof(int),
1085 .proc_handler = brnf_sysctl_call_tables,
1088 .procname = "bridge-nf-filter-pppoe-tagged",
1089 .data = &brnf_filter_pppoe_tagged,
1090 .maxlen = sizeof(int),
1092 .proc_handler = brnf_sysctl_call_tables,
1095 .procname = "bridge-nf-pass-vlan-input-dev",
1096 .data = &brnf_pass_vlan_indev,
1097 .maxlen = sizeof(int),
1099 .proc_handler = brnf_sysctl_call_tables,
1105 static int __init br_netfilter_init(void)
1109 ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1113 #ifdef CONFIG_SYSCTL
1114 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1115 if (brnf_sysctl_header == NULL) {
1117 "br_netfilter: can't register to sysctl.\n");
1118 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1122 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1123 printk(KERN_NOTICE "Bridge firewalling registered\n");
1127 static void __exit br_netfilter_fini(void)
1129 RCU_INIT_POINTER(nf_br_ops, NULL);
1130 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1131 #ifdef CONFIG_SYSCTL
1132 unregister_net_sysctl_table(brnf_sysctl_header);
1136 module_init(br_netfilter_init);
1137 module_exit(br_netfilter_fini);
1139 MODULE_LICENSE("GPL");
1140 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1141 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1142 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");