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 struct nf_bridge_info *nf_bridge_info_get(const struct sk_buff *skb)
129 return skb->nf_bridge;
132 static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
134 struct net_bridge_port *port;
136 port = br_port_get_rcu(dev);
137 return port ? &port->br->fake_rtable : NULL;
140 static inline struct net_device *bridge_parent(const struct net_device *dev)
142 struct net_bridge_port *port;
144 port = br_port_get_rcu(dev);
145 return port ? port->br->dev : NULL;
148 static inline struct nf_bridge_info *nf_bridge_alloc(struct sk_buff *skb)
150 skb->nf_bridge = kzalloc(sizeof(struct nf_bridge_info), GFP_ATOMIC);
151 if (likely(skb->nf_bridge))
152 atomic_set(&(skb->nf_bridge->use), 1);
154 return skb->nf_bridge;
157 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
159 struct nf_bridge_info *nf_bridge = skb->nf_bridge;
161 if (atomic_read(&nf_bridge->use) > 1) {
162 struct nf_bridge_info *tmp = nf_bridge_alloc(skb);
165 memcpy(tmp, nf_bridge, sizeof(struct nf_bridge_info));
166 atomic_set(&tmp->use, 1);
168 nf_bridge_put(nf_bridge);
174 static unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
176 switch (skb->protocol) {
177 case __cpu_to_be16(ETH_P_8021Q):
179 case __cpu_to_be16(ETH_P_PPP_SES):
180 return PPPOE_SES_HLEN;
186 static inline void nf_bridge_push_encap_header(struct sk_buff *skb)
188 unsigned int len = nf_bridge_encap_header_len(skb);
191 skb->network_header -= len;
194 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
196 unsigned int len = nf_bridge_encap_header_len(skb);
199 skb->network_header += len;
202 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
204 unsigned int len = nf_bridge_encap_header_len(skb);
206 skb_pull_rcsum(skb, len);
207 skb->network_header += len;
210 /* When handing a packet over to the IP layer
211 * check whether we have a skb that is in the
215 static int br_parse_ip_options(struct sk_buff *skb)
217 const struct iphdr *iph;
218 struct net_device *dev = skb->dev;
221 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
226 /* Basic sanity checks */
227 if (iph->ihl < 5 || iph->version != 4)
230 if (!pskb_may_pull(skb, iph->ihl*4))
234 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
237 len = ntohs(iph->tot_len);
238 if (skb->len < len) {
239 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INTRUNCATEDPKTS);
241 } else if (len < (iph->ihl*4))
244 if (pskb_trim_rcsum(skb, len)) {
245 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INDISCARDS);
249 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
250 /* We should really parse IP options here but until
251 * somebody who actually uses IP options complains to
252 * us we'll just silently ignore the options because
258 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_INHDRERRORS);
263 static void nf_bridge_update_protocol(struct sk_buff *skb)
265 switch (skb->nf_bridge->orig_proto) {
266 case BRNF_PROTO_8021Q:
267 skb->protocol = htons(ETH_P_8021Q);
269 case BRNF_PROTO_PPPOE:
270 skb->protocol = htons(ETH_P_PPP_SES);
272 case BRNF_PROTO_UNCHANGED:
277 /* PF_BRIDGE/PRE_ROUTING *********************************************/
278 /* Undo the changes made for ip6tables PREROUTING and continue the
279 * bridge PRE_ROUTING hook. */
280 static int br_nf_pre_routing_finish_ipv6(struct sk_buff *skb)
282 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
285 if (nf_bridge->pkt_otherhost) {
286 skb->pkt_type = PACKET_OTHERHOST;
287 nf_bridge->pkt_otherhost = false;
289 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
291 rt = bridge_parent_rtable(nf_bridge->physindev);
296 skb_dst_set_noref(skb, &rt->dst);
298 skb->dev = nf_bridge->physindev;
299 nf_bridge_update_protocol(skb);
300 nf_bridge_push_encap_header(skb);
301 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
302 br_handle_frame_finish, 1);
307 /* Obtain the correct destination MAC address, while preserving the original
308 * source MAC address. If we already know this address, we just copy it. If we
309 * don't, we use the neighbour framework to find out. In both cases, we make
310 * sure that br_handle_frame_finish() is called afterwards.
312 static int br_nf_pre_routing_finish_bridge(struct sk_buff *skb)
314 struct neighbour *neigh;
315 struct dst_entry *dst;
317 skb->dev = bridge_parent(skb->dev);
321 neigh = dst_neigh_lookup_skb(dst, skb);
323 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
326 if (neigh->hh.hh_len) {
327 neigh_hh_bridge(&neigh->hh, skb);
328 skb->dev = nf_bridge->physindev;
329 ret = br_handle_frame_finish(skb);
331 /* the neighbour function below overwrites the complete
332 * MAC header, so we save the Ethernet source address and
335 skb_copy_from_linear_data_offset(skb,
336 -(ETH_HLEN-ETH_ALEN),
337 nf_bridge->neigh_header,
339 /* tell br_dev_xmit to continue with forwarding */
340 nf_bridge->mask |= BRNF_BRIDGED_DNAT;
341 /* FIXME Need to refragment */
342 ret = neigh->output(neigh, skb);
344 neigh_release(neigh);
352 static bool dnat_took_place(const struct sk_buff *skb)
354 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
355 enum ip_conntrack_info ctinfo;
358 ct = nf_ct_get(skb, &ctinfo);
359 if (!ct || nf_ct_is_untracked(ct))
362 return test_bit(IPS_DST_NAT_BIT, &ct->status);
368 /* This requires some explaining. If DNAT has taken place,
369 * we will need to fix up the destination Ethernet address.
371 * There are two cases to consider:
372 * 1. The packet was DNAT'ed to a device in the same bridge
373 * port group as it was received on. We can still bridge
375 * 2. The packet was DNAT'ed to a different device, either
376 * a non-bridged device or another bridge port group.
377 * The packet will need to be routed.
379 * The correct way of distinguishing between these two cases is to
380 * call ip_route_input() and to look at skb->dst->dev, which is
381 * changed to the destination device if ip_route_input() succeeds.
383 * Let's first consider the case that ip_route_input() succeeds:
385 * If the output device equals the logical bridge device the packet
386 * came in on, we can consider this bridging. The corresponding MAC
387 * address will be obtained in br_nf_pre_routing_finish_bridge.
388 * Otherwise, the packet is considered to be routed and we just
389 * change the destination MAC address so that the packet will
390 * later be passed up to the IP stack to be routed. For a redirected
391 * packet, ip_route_input() will give back the localhost as output device,
392 * which differs from the bridge device.
394 * Let's now consider the case that ip_route_input() fails:
396 * This can be because the destination address is martian, in which case
397 * the packet will be dropped.
398 * If IP forwarding is disabled, ip_route_input() will fail, while
399 * ip_route_output_key() can return success. The source
400 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
401 * thinks we're handling a locally generated packet and won't care
402 * if IP forwarding is enabled. If the output device equals the logical bridge
403 * device, we proceed as if ip_route_input() succeeded. If it differs from the
404 * logical bridge port or if ip_route_output_key() fails we drop the packet.
406 static int br_nf_pre_routing_finish(struct sk_buff *skb)
408 struct net_device *dev = skb->dev;
409 struct iphdr *iph = ip_hdr(skb);
410 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
415 frag_max_size = IPCB(skb)->frag_max_size;
416 BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size;
418 if (nf_bridge->pkt_otherhost) {
419 skb->pkt_type = PACKET_OTHERHOST;
420 nf_bridge->pkt_otherhost = false;
422 nf_bridge->mask ^= BRNF_NF_BRIDGE_PREROUTING;
423 if (dnat_took_place(skb)) {
424 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
425 struct in_device *in_dev = __in_dev_get_rcu(dev);
427 /* If err equals -EHOSTUNREACH the error is due to a
428 * martian destination or due to the fact that
429 * forwarding is disabled. For most martian packets,
430 * ip_route_output_key() will fail. It won't fail for 2 types of
431 * martian destinations: loopback destinations and destination
432 * 0.0.0.0. In both cases the packet will be dropped because the
433 * destination is the loopback device and not the bridge. */
434 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
437 rt = ip_route_output(dev_net(dev), iph->daddr, 0,
438 RT_TOS(iph->tos), 0);
440 /* - Bridged-and-DNAT'ed traffic doesn't
441 * require ip_forwarding. */
442 if (rt->dst.dev == dev) {
443 skb_dst_set(skb, &rt->dst);
452 if (skb_dst(skb)->dev == dev) {
454 skb->dev = nf_bridge->physindev;
455 nf_bridge_update_protocol(skb);
456 nf_bridge_push_encap_header(skb);
457 NF_HOOK_THRESH(NFPROTO_BRIDGE,
460 br_nf_pre_routing_finish_bridge,
464 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
465 skb->pkt_type = PACKET_HOST;
468 rt = bridge_parent_rtable(nf_bridge->physindev);
473 skb_dst_set_noref(skb, &rt->dst);
476 skb->dev = nf_bridge->physindev;
477 nf_bridge_update_protocol(skb);
478 nf_bridge_push_encap_header(skb);
479 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_PRE_ROUTING, skb, skb->dev, NULL,
480 br_handle_frame_finish, 1);
485 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb, const struct net_device *dev)
487 struct net_device *vlan, *br;
489 br = bridge_parent(dev);
490 if (brnf_pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
493 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
494 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
496 return vlan ? vlan : br;
499 /* Some common code for IPv4/IPv6 */
500 static struct net_device *setup_pre_routing(struct sk_buff *skb)
502 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
504 if (skb->pkt_type == PACKET_OTHERHOST) {
505 skb->pkt_type = PACKET_HOST;
506 nf_bridge->pkt_otherhost = true;
509 nf_bridge->mask |= BRNF_NF_BRIDGE_PREROUTING;
510 nf_bridge->physindev = skb->dev;
511 skb->dev = brnf_get_logical_dev(skb, skb->dev);
513 if (skb->protocol == htons(ETH_P_8021Q))
514 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
515 else if (skb->protocol == htons(ETH_P_PPP_SES))
516 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
518 /* Must drop socket now because of tproxy. */
523 /* We only check the length. A bridge shouldn't do any hop-by-hop stuff anyway */
524 static int check_hbh_len(struct sk_buff *skb)
526 unsigned char *raw = (u8 *)(ipv6_hdr(skb) + 1);
528 const unsigned char *nh = skb_network_header(skb);
530 int len = (raw[1] + 1) << 3;
532 if ((raw + len) - skb->data > skb_headlen(skb))
539 int optlen = nh[off + 1] + 2;
550 if (nh[off + 1] != 4 || (off & 3) != 2)
552 pkt_len = ntohl(*(__be32 *) (nh + off + 2));
553 if (pkt_len <= IPV6_MAXPLEN ||
554 ipv6_hdr(skb)->payload_len)
556 if (pkt_len > skb->len - sizeof(struct ipv6hdr))
558 if (pskb_trim_rcsum(skb,
559 pkt_len + sizeof(struct ipv6hdr)))
561 nh = skb_network_header(skb);
578 /* Replicate the checks that IPv6 does on packet reception and pass the packet
579 * to ip6tables, which doesn't support NAT, so things are fairly simple. */
580 static unsigned int br_nf_pre_routing_ipv6(const struct nf_hook_ops *ops,
582 const struct net_device *in,
583 const struct net_device *out,
584 int (*okfn)(struct sk_buff *))
586 const struct ipv6hdr *hdr;
589 if (skb->len < sizeof(struct ipv6hdr))
592 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
597 if (hdr->version != 6)
600 pkt_len = ntohs(hdr->payload_len);
602 if (pkt_len || hdr->nexthdr != NEXTHDR_HOP) {
603 if (pkt_len + sizeof(struct ipv6hdr) > skb->len)
605 if (pskb_trim_rcsum(skb, pkt_len + sizeof(struct ipv6hdr)))
608 if (hdr->nexthdr == NEXTHDR_HOP && check_hbh_len(skb))
611 nf_bridge_put(skb->nf_bridge);
612 if (!nf_bridge_alloc(skb))
614 if (!setup_pre_routing(skb))
617 skb->protocol = htons(ETH_P_IPV6);
618 NF_HOOK(NFPROTO_IPV6, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
619 br_nf_pre_routing_finish_ipv6);
624 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
625 * Replicate the checks that IPv4 does on packet reception.
626 * Set skb->dev to the bridge device (i.e. parent of the
627 * receiving device) to make netfilter happy, the REDIRECT
628 * target in particular. Save the original destination IP
629 * address to be able to detect DNAT afterwards. */
630 static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
632 const struct net_device *in,
633 const struct net_device *out,
634 int (*okfn)(struct sk_buff *))
636 struct net_bridge_port *p;
637 struct net_bridge *br;
638 __u32 len = nf_bridge_encap_header_len(skb);
640 if (unlikely(!pskb_may_pull(skb, len)))
643 p = br_port_get_rcu(in);
648 if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb)) {
649 if (!brnf_call_ip6tables && !br->nf_call_ip6tables)
652 nf_bridge_pull_encap_header_rcsum(skb);
653 return br_nf_pre_routing_ipv6(ops, skb, in, out, okfn);
656 if (!brnf_call_iptables && !br->nf_call_iptables)
659 if (!IS_IP(skb) && !IS_VLAN_IP(skb) && !IS_PPPOE_IP(skb))
662 nf_bridge_pull_encap_header_rcsum(skb);
664 if (br_parse_ip_options(skb))
667 nf_bridge_put(skb->nf_bridge);
668 if (!nf_bridge_alloc(skb))
670 if (!setup_pre_routing(skb))
673 skb->protocol = htons(ETH_P_IP);
675 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, skb, skb->dev, NULL,
676 br_nf_pre_routing_finish);
682 /* PF_BRIDGE/LOCAL_IN ************************************************/
683 /* The packet is locally destined, which requires a real
684 * dst_entry, so detach the fake one. On the way up, the
685 * packet would pass through PRE_ROUTING again (which already
686 * took place when the packet entered the bridge), but we
687 * register an IPv4 PRE_ROUTING 'sabotage' hook that will
688 * prevent this from happening. */
689 static unsigned int br_nf_local_in(const struct nf_hook_ops *ops,
691 const struct net_device *in,
692 const struct net_device *out,
693 int (*okfn)(struct sk_buff *))
695 br_drop_fake_rtable(skb);
699 /* PF_BRIDGE/FORWARD *************************************************/
700 static int br_nf_forward_finish(struct sk_buff *skb)
702 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
703 struct net_device *in;
705 if (!IS_ARP(skb) && !IS_VLAN_ARP(skb)) {
708 if (skb->protocol == htons(ETH_P_IP)) {
709 frag_max_size = IPCB(skb)->frag_max_size;
710 BR_INPUT_SKB_CB(skb)->frag_max_size = frag_max_size;
713 in = nf_bridge->physindev;
714 if (nf_bridge->pkt_otherhost) {
715 skb->pkt_type = PACKET_OTHERHOST;
716 nf_bridge->pkt_otherhost = false;
718 nf_bridge_update_protocol(skb);
720 in = *((struct net_device **)(skb->cb));
722 nf_bridge_push_encap_header(skb);
724 NF_HOOK_THRESH(NFPROTO_BRIDGE, NF_BR_FORWARD, skb, in,
725 skb->dev, br_forward_finish, 1);
730 /* This is the 'purely bridged' case. For IP, we pass the packet to
731 * netfilter with indev and outdev set to the bridge device,
732 * but we are still able to filter on the 'real' indev/outdev
733 * because of the physdev module. For ARP, indev and outdev are the
735 static unsigned int br_nf_forward_ip(const struct nf_hook_ops *ops,
737 const struct net_device *in,
738 const struct net_device *out,
739 int (*okfn)(struct sk_buff *))
741 struct nf_bridge_info *nf_bridge;
742 struct net_device *parent;
748 /* Need exclusive nf_bridge_info since we might have multiple
749 * different physoutdevs. */
750 if (!nf_bridge_unshare(skb))
753 nf_bridge = nf_bridge_info_get(skb);
757 parent = bridge_parent(out);
761 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
763 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
768 nf_bridge_pull_encap_header(skb);
770 if (skb->pkt_type == PACKET_OTHERHOST) {
771 skb->pkt_type = PACKET_HOST;
772 nf_bridge->pkt_otherhost = true;
775 if (pf == NFPROTO_IPV4) {
776 int frag_max = BR_INPUT_SKB_CB(skb)->frag_max_size;
778 if (br_parse_ip_options(skb))
781 IPCB(skb)->frag_max_size = frag_max;
784 nf_bridge->physoutdev = skb->dev;
785 if (pf == NFPROTO_IPV4)
786 skb->protocol = htons(ETH_P_IP);
788 skb->protocol = htons(ETH_P_IPV6);
790 NF_HOOK(pf, NF_INET_FORWARD, skb, brnf_get_logical_dev(skb, in), parent,
791 br_nf_forward_finish);
796 static unsigned int br_nf_forward_arp(const struct nf_hook_ops *ops,
798 const struct net_device *in,
799 const struct net_device *out,
800 int (*okfn)(struct sk_buff *))
802 struct net_bridge_port *p;
803 struct net_bridge *br;
804 struct net_device **d = (struct net_device **)(skb->cb);
806 p = br_port_get_rcu(out);
811 if (!brnf_call_arptables && !br->nf_call_arptables)
815 if (!IS_VLAN_ARP(skb))
817 nf_bridge_pull_encap_header(skb);
820 if (arp_hdr(skb)->ar_pln != 4) {
821 if (IS_VLAN_ARP(skb))
822 nf_bridge_push_encap_header(skb);
825 *d = (struct net_device *)in;
826 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, skb, (struct net_device *)in,
827 (struct net_device *)out, br_nf_forward_finish);
832 #if IS_ENABLED(CONFIG_NF_DEFRAG_IPV4)
833 static int br_nf_push_frag_xmit(struct sk_buff *skb)
835 struct brnf_frag_data *data;
838 data = this_cpu_ptr(&brnf_frag_data_storage);
839 err = skb_cow_head(skb, data->size);
846 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
847 __skb_push(skb, data->encap_size);
849 return br_dev_queue_push_xmit(skb);
852 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
856 unsigned int mtu_reserved;
858 if (skb_is_gso(skb) || skb->protocol != htons(ETH_P_IP))
859 return br_dev_queue_push_xmit(skb);
861 mtu_reserved = nf_bridge_mtu_reduction(skb);
862 /* This is wrong! We should preserve the original fragment
863 * boundaries by preserving frag_list rather than refragmenting.
865 if (skb->len + mtu_reserved > skb->dev->mtu) {
866 struct brnf_frag_data *data;
868 frag_max_size = BR_INPUT_SKB_CB(skb)->frag_max_size;
869 if (br_parse_ip_options(skb))
870 /* Drop invalid packet */
872 IPCB(skb)->frag_max_size = frag_max_size;
874 nf_bridge_update_protocol(skb);
876 data = this_cpu_ptr(&brnf_frag_data_storage);
877 data->encap_size = nf_bridge_encap_header_len(skb);
878 data->size = ETH_HLEN + data->encap_size;
880 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
883 ret = ip_fragment(skb, br_nf_push_frag_xmit);
885 ret = br_dev_queue_push_xmit(skb);
891 static int br_nf_dev_queue_xmit(struct sk_buff *skb)
893 return br_dev_queue_push_xmit(skb);
897 /* PF_BRIDGE/POST_ROUTING ********************************************/
898 static unsigned int br_nf_post_routing(const struct nf_hook_ops *ops,
900 const struct net_device *in,
901 const struct net_device *out,
902 int (*okfn)(struct sk_buff *))
904 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
905 struct net_device *realoutdev = bridge_parent(skb->dev);
908 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
909 * on a bridge, but was delivered locally and is now being routed:
911 * POST_ROUTING was already invoked from the ip stack.
913 if (!nf_bridge || !nf_bridge->physoutdev)
919 if (IS_IP(skb) || IS_VLAN_IP(skb) || IS_PPPOE_IP(skb))
921 else if (IS_IPV6(skb) || IS_VLAN_IPV6(skb) || IS_PPPOE_IPV6(skb))
926 /* We assume any code from br_dev_queue_push_xmit onwards doesn't care
927 * about the value of skb->pkt_type. */
928 if (skb->pkt_type == PACKET_OTHERHOST) {
929 skb->pkt_type = PACKET_HOST;
930 nf_bridge->pkt_otherhost = true;
933 nf_bridge_pull_encap_header(skb);
934 if (pf == NFPROTO_IPV4)
935 skb->protocol = htons(ETH_P_IP);
937 skb->protocol = htons(ETH_P_IPV6);
939 NF_HOOK(pf, NF_INET_POST_ROUTING, skb, NULL, realoutdev,
940 br_nf_dev_queue_xmit);
945 /* IP/SABOTAGE *****************************************************/
946 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
947 * for the second time. */
948 static unsigned int ip_sabotage_in(const struct nf_hook_ops *ops,
950 const struct net_device *in,
951 const struct net_device *out,
952 int (*okfn)(struct sk_buff *))
954 if (skb->nf_bridge &&
955 !(skb->nf_bridge->mask & BRNF_NF_BRIDGE_PREROUTING)) {
962 /* This is called when br_netfilter has called into iptables/netfilter,
963 * and DNAT has taken place on a bridge-forwarded packet.
965 * neigh->output has created a new MAC header, with local br0 MAC
968 * This restores the original MAC saddr of the bridged packet
969 * before invoking bridge forward logic to transmit the packet.
971 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
973 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
975 skb_pull(skb, ETH_HLEN);
976 nf_bridge->mask &= ~BRNF_BRIDGED_DNAT;
978 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
980 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
981 nf_bridge->neigh_header,
982 ETH_HLEN - ETH_ALEN);
983 skb->dev = nf_bridge->physindev;
984 br_handle_frame_finish(skb);
987 static int br_nf_dev_xmit(struct sk_buff *skb)
989 if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_BRIDGED_DNAT)) {
990 br_nf_pre_routing_finish_bridge_slow(skb);
996 static const struct nf_br_ops br_ops = {
997 .br_dev_xmit_hook = br_nf_dev_xmit,
1000 void br_netfilter_enable(void)
1003 EXPORT_SYMBOL_GPL(br_netfilter_enable);
1005 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
1006 * br_dev_queue_push_xmit is called afterwards */
1007 static struct nf_hook_ops br_nf_ops[] __read_mostly = {
1009 .hook = br_nf_pre_routing,
1010 .owner = THIS_MODULE,
1011 .pf = NFPROTO_BRIDGE,
1012 .hooknum = NF_BR_PRE_ROUTING,
1013 .priority = NF_BR_PRI_BRNF,
1016 .hook = br_nf_local_in,
1017 .owner = THIS_MODULE,
1018 .pf = NFPROTO_BRIDGE,
1019 .hooknum = NF_BR_LOCAL_IN,
1020 .priority = NF_BR_PRI_BRNF,
1023 .hook = br_nf_forward_ip,
1024 .owner = THIS_MODULE,
1025 .pf = NFPROTO_BRIDGE,
1026 .hooknum = NF_BR_FORWARD,
1027 .priority = NF_BR_PRI_BRNF - 1,
1030 .hook = br_nf_forward_arp,
1031 .owner = THIS_MODULE,
1032 .pf = NFPROTO_BRIDGE,
1033 .hooknum = NF_BR_FORWARD,
1034 .priority = NF_BR_PRI_BRNF,
1037 .hook = br_nf_post_routing,
1038 .owner = THIS_MODULE,
1039 .pf = NFPROTO_BRIDGE,
1040 .hooknum = NF_BR_POST_ROUTING,
1041 .priority = NF_BR_PRI_LAST,
1044 .hook = ip_sabotage_in,
1045 .owner = THIS_MODULE,
1047 .hooknum = NF_INET_PRE_ROUTING,
1048 .priority = NF_IP_PRI_FIRST,
1051 .hook = ip_sabotage_in,
1052 .owner = THIS_MODULE,
1054 .hooknum = NF_INET_PRE_ROUTING,
1055 .priority = NF_IP6_PRI_FIRST,
1059 #ifdef CONFIG_SYSCTL
1061 int brnf_sysctl_call_tables(struct ctl_table *ctl, int write,
1062 void __user *buffer, size_t *lenp, loff_t *ppos)
1066 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1068 if (write && *(int *)(ctl->data))
1069 *(int *)(ctl->data) = 1;
1073 static struct ctl_table brnf_table[] = {
1075 .procname = "bridge-nf-call-arptables",
1076 .data = &brnf_call_arptables,
1077 .maxlen = sizeof(int),
1079 .proc_handler = brnf_sysctl_call_tables,
1082 .procname = "bridge-nf-call-iptables",
1083 .data = &brnf_call_iptables,
1084 .maxlen = sizeof(int),
1086 .proc_handler = brnf_sysctl_call_tables,
1089 .procname = "bridge-nf-call-ip6tables",
1090 .data = &brnf_call_ip6tables,
1091 .maxlen = sizeof(int),
1093 .proc_handler = brnf_sysctl_call_tables,
1096 .procname = "bridge-nf-filter-vlan-tagged",
1097 .data = &brnf_filter_vlan_tagged,
1098 .maxlen = sizeof(int),
1100 .proc_handler = brnf_sysctl_call_tables,
1103 .procname = "bridge-nf-filter-pppoe-tagged",
1104 .data = &brnf_filter_pppoe_tagged,
1105 .maxlen = sizeof(int),
1107 .proc_handler = brnf_sysctl_call_tables,
1110 .procname = "bridge-nf-pass-vlan-input-dev",
1111 .data = &brnf_pass_vlan_indev,
1112 .maxlen = sizeof(int),
1114 .proc_handler = brnf_sysctl_call_tables,
1120 static int __init br_netfilter_init(void)
1124 ret = nf_register_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1128 #ifdef CONFIG_SYSCTL
1129 brnf_sysctl_header = register_net_sysctl(&init_net, "net/bridge", brnf_table);
1130 if (brnf_sysctl_header == NULL) {
1132 "br_netfilter: can't register to sysctl.\n");
1133 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1137 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1138 printk(KERN_NOTICE "Bridge firewalling registered\n");
1142 static void __exit br_netfilter_fini(void)
1144 RCU_INIT_POINTER(nf_br_ops, NULL);
1145 nf_unregister_hooks(br_nf_ops, ARRAY_SIZE(br_nf_ops));
1146 #ifdef CONFIG_SYSCTL
1147 unregister_net_sysctl_table(brnf_sysctl_header);
1151 module_init(br_netfilter_init);
1152 module_exit(br_netfilter_fini);
1154 MODULE_LICENSE("GPL");
1155 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1156 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1157 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");