1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * Handle firewalling
4 * Linux ethernet bridge
5 *
6 * Authors:
7 * Lennert Buytenhek <buytenh@gnu.org>
8 * Bart De Schuymer <bdschuym@pandora.be>
9 *
10 * Lennert dedicates this file to Kerstin Wurdinger.
11 */
12
13 #include <linux/module.h>
14 #include <linux/kernel.h>
15 #include <linux/slab.h>
16 #include <linux/ip.h>
17 #include <linux/netdevice.h>
18 #include <linux/skbuff.h>
19 #include <linux/if_arp.h>
20 #include <linux/if_ether.h>
21 #include <linux/if_vlan.h>
22 #include <linux/if_pppox.h>
23 #include <linux/ppp_defs.h>
24 #include <linux/netfilter_bridge.h>
25 #include <uapi/linux/netfilter_bridge.h>
26 #include <linux/netfilter_ipv4.h>
27 #include <linux/netfilter_ipv6.h>
28 #include <linux/netfilter_arp.h>
29 #include <linux/in_route.h>
30 #include <linux/rculist.h>
31 #include <linux/inetdevice.h>
32
33 #include <net/ip.h>
34 #include <net/ipv6.h>
35 #include <net/addrconf.h>
36 #include <net/route.h>
37 #include <net/netfilter/br_netfilter.h>
38 #include <net/netns/generic.h>
39
40 #include <linux/uaccess.h>
41 #include "br_private.h"
42 #ifdef CONFIG_SYSCTL
43 #include <linux/sysctl.h>
44 #endif
45
46 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
47 #include <net/netfilter/nf_conntrack_core.h>
48 #endif
49
50 static unsigned int brnf_net_id __read_mostly;
51
52 struct brnf_net {
53 bool enabled;
54
55 #ifdef CONFIG_SYSCTL
56 struct ctl_table_header *ctl_hdr;
57 #endif
58
59 /* default value is 1 */
60 int call_iptables;
61 int call_ip6tables;
62 int call_arptables;
63
64 /* default value is 0 */
65 int filter_vlan_tagged;
66 int filter_pppoe_tagged;
67 int pass_vlan_indev;
68 };
69
70 #define IS_IP(skb) \
71 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
72
73 #define IS_IPV6(skb) \
74 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
75
76 #define IS_ARP(skb) \
77 (!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
78
79 static inline __be16 vlan_proto(const struct sk_buff *skb)
80 {
81 if (skb_vlan_tag_present(skb))
82 return skb->protocol;
83 else if (skb->protocol == htons(ETH_P_8021Q))
84 return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
85 else
86 return 0;
87 }
88
89 static inline bool is_vlan_ip(const struct sk_buff *skb, const struct net *net)
90 {
91 struct brnf_net *brnet = net_generic(net, brnf_net_id);
92
93 return vlan_proto(skb) == htons(ETH_P_IP) && brnet->filter_vlan_tagged;
94 }
95
96 static inline bool is_vlan_ipv6(const struct sk_buff *skb,
97 const struct net *net)
98 {
99 struct brnf_net *brnet = net_generic(net, brnf_net_id);
100
101 return vlan_proto(skb) == htons(ETH_P_IPV6) &&
102 brnet->filter_vlan_tagged;
103 }
104
105 static inline bool is_vlan_arp(const struct sk_buff *skb, const struct net *net)
106 {
107 struct brnf_net *brnet = net_generic(net, brnf_net_id);
108
109 return vlan_proto(skb) == htons(ETH_P_ARP) && brnet->filter_vlan_tagged;
110 }
111
112 static inline __be16 pppoe_proto(const struct sk_buff *skb)
113 {
114 return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
115 sizeof(struct pppoe_hdr)));
116 }
117
118 static inline bool is_pppoe_ip(const struct sk_buff *skb, const struct net *net)
119 {
120 struct brnf_net *brnet = net_generic(net, brnf_net_id);
121
122 return skb->protocol == htons(ETH_P_PPP_SES) &&
123 pppoe_proto(skb) == htons(PPP_IP) && brnet->filter_pppoe_tagged;
124 }
125
126 static inline bool is_pppoe_ipv6(const struct sk_buff *skb,
127 const struct net *net)
128 {
129 struct brnf_net *brnet = net_generic(net, brnf_net_id);
130
131 return skb->protocol == htons(ETH_P_PPP_SES) &&
132 pppoe_proto(skb) == htons(PPP_IPV6) &&
133 brnet->filter_pppoe_tagged;
134 }
135
136 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
137 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
138
139 struct brnf_frag_data {
140 local_lock_t bh_lock;
141 char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
142 u8 encap_size;
143 u8 size;
144 u16 vlan_tci;
145 __be16 vlan_proto;
146 };
147
148 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage) = {
149 .bh_lock = INIT_LOCAL_LOCK(bh_lock),
150 };
151
152 static void nf_bridge_info_free(struct sk_buff *skb)
153 {
154 skb_ext_del(skb, SKB_EXT_BRIDGE_NF);
155 }
156
157 static inline struct net_device *bridge_parent(const struct net_device *dev)
158 {
159 struct net_bridge_port *port;
160
161 port = br_port_get_rcu(dev);
162 return port ? port->br->dev : NULL;
163 }
164
165 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
166 {
167 return skb_ext_add(skb, SKB_EXT_BRIDGE_NF);
168 }
169
170 unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
171 {
172 switch (skb->protocol) {
173 case __cpu_to_be16(ETH_P_8021Q):
174 return VLAN_HLEN;
175 case __cpu_to_be16(ETH_P_PPP_SES):
176 return PPPOE_SES_HLEN;
177 default:
178 return 0;
179 }
180 }
181
182 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
183 {
184 unsigned int len = nf_bridge_encap_header_len(skb);
185
186 skb_pull(skb, len);
187 skb->network_header += len;
188 }
189
190 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
191 {
192 unsigned int len = nf_bridge_encap_header_len(skb);
193
194 skb_pull_rcsum(skb, len);
195 skb->network_header += len;
196 }
197
198 /* When handing a packet over to the IP layer
199 * check whether we have a skb that is in the
200 * expected format
201 */
202
203 static int br_validate_ipv4(struct net *net, struct sk_buff *skb)
204 {
205 const struct iphdr *iph;
206 u32 len;
207
208 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
209 goto inhdr_error;
210
211 iph = ip_hdr(skb);
212
213 /* Basic sanity checks */
214 if (iph->ihl < 5 || iph->version != 4)
215 goto inhdr_error;
216
217 if (!pskb_may_pull(skb, iph->ihl*4))
218 goto inhdr_error;
219
220 iph = ip_hdr(skb);
221 if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
222 goto csum_error;
223
224 len = skb_ip_totlen(skb);
225 if (skb->len < len) {
226 __IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS);
227 goto drop;
228 } else if (len < (iph->ihl*4))
229 goto inhdr_error;
230
231 if (pskb_trim_rcsum(skb, len)) {
232 __IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
233 goto drop;
234 }
235
236 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
237 /* We should really parse IP options here but until
238 * somebody who actually uses IP options complains to
239 * us we'll just silently ignore the options because
240 * we're lazy!
241 */
242 return 0;
243
244 csum_error:
245 __IP_INC_STATS(net, IPSTATS_MIB_CSUMERRORS);
246 inhdr_error:
247 __IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
248 drop:
249 return -1;
250 }
251
252 void nf_bridge_update_protocol(struct sk_buff *skb)
253 {
254 const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
255
256 switch (nf_bridge->orig_proto) {
257 case BRNF_PROTO_8021Q:
258 skb->protocol = htons(ETH_P_8021Q);
259 break;
260 case BRNF_PROTO_PPPOE:
261 skb->protocol = htons(ETH_P_PPP_SES);
262 break;
263 case BRNF_PROTO_UNCHANGED:
264 break;
265 }
266 }
267
268 /* Obtain the correct destination MAC address, while preserving the original
269 * source MAC address. If we already know this address, we just copy it. If we
270 * don't, we use the neighbour framework to find out. In both cases, we make
271 * sure that br_handle_frame_finish() is called afterwards.
272 */
273 int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb)
274 {
275 struct neighbour *neigh;
276 struct dst_entry *dst;
277
278 skb->dev = bridge_parent(skb->dev);
279 if (!skb->dev)
280 goto free_skb;
281 dst = skb_dst(skb);
282 neigh = dst_neigh_lookup_skb(dst, skb);
283 if (neigh) {
284 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
285 int ret;
286
287 if ((READ_ONCE(neigh->nud_state) & NUD_CONNECTED) &&
288 READ_ONCE(neigh->hh.hh_len)) {
289 struct net_device *br_indev;
290
291 br_indev = nf_bridge_get_physindev(skb, net);
292 if (!br_indev) {
293 neigh_release(neigh);
294 goto free_skb;
295 }
296
297 neigh_hh_bridge(&neigh->hh, skb);
298 skb->dev = br_indev;
299
300 ret = br_handle_frame_finish(net, sk, skb);
301 } else {
302 /* the neighbour function below overwrites the complete
303 * MAC header, so we save the Ethernet source address and
304 * protocol number.
305 */
306 skb_copy_from_linear_data_offset(skb,
307 -(ETH_HLEN-ETH_ALEN),
308 nf_bridge->neigh_header,
309 ETH_HLEN-ETH_ALEN);
310 /* tell br_dev_xmit to continue with forwarding */
311 nf_bridge->bridged_dnat = 1;
312 /* FIXME Need to refragment */
313 ret = READ_ONCE(neigh->output)(neigh, skb);
314 }
315 neigh_release(neigh);
316 return ret;
317 }
318 free_skb:
319 kfree_skb(skb);
320 return 0;
321 }
322
323 static inline bool
324 br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb,
325 const struct nf_bridge_info *nf_bridge)
326 {
327 return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
328 }
329
330 /* This requires some explaining. If DNAT has taken place,
331 * we will need to fix up the destination Ethernet address.
332 * This is also true when SNAT takes place (for the reply direction).
333 *
334 * There are two cases to consider:
335 * 1. The packet was DNAT'ed to a device in the same bridge
336 * port group as it was received on. We can still bridge
337 * the packet.
338 * 2. The packet was DNAT'ed to a different device, either
339 * a non-bridged device or another bridge port group.
340 * The packet will need to be routed.
341 *
342 * The correct way of distinguishing between these two cases is to
343 * call ip_route_input() and to look at skb->dst->dev, which is
344 * changed to the destination device if ip_route_input() succeeds.
345 *
346 * Let's first consider the case that ip_route_input() succeeds:
347 *
348 * If the output device equals the logical bridge device the packet
349 * came in on, we can consider this bridging. The corresponding MAC
350 * address will be obtained in br_nf_pre_routing_finish_bridge.
351 * Otherwise, the packet is considered to be routed and we just
352 * change the destination MAC address so that the packet will
353 * later be passed up to the IP stack to be routed. For a redirected
354 * packet, ip_route_input() will give back the localhost as output device,
355 * which differs from the bridge device.
356 *
357 * Let's now consider the case that ip_route_input() fails:
358 *
359 * This can be because the destination address is martian, in which case
360 * the packet will be dropped.
361 * If IP forwarding is disabled, ip_route_input() will fail, while
362 * ip_route_output_key() can return success. The source
363 * address for ip_route_output_key() is set to zero, so ip_route_output_key()
364 * thinks we're handling a locally generated packet and won't care
365 * if IP forwarding is enabled. If the output device equals the logical bridge
366 * device, we proceed as if ip_route_input() succeeded. If it differs from the
367 * logical bridge port or if ip_route_output_key() fails we drop the packet.
368 */
369 static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
370 {
371 struct net_device *dev = skb->dev, *br_indev;
372 struct iphdr *iph = ip_hdr(skb);
373 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
374 struct rtable *rt;
375 int err;
376
377 br_indev = nf_bridge_get_physindev(skb, net);
378 if (!br_indev) {
379 kfree_skb(skb);
380 return 0;
381 }
382
383 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
384
385 if (nf_bridge->pkt_otherhost) {
386 skb->pkt_type = PACKET_OTHERHOST;
387 nf_bridge->pkt_otherhost = false;
388 }
389 nf_bridge->in_prerouting = 0;
390 if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) {
391 if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
392 struct in_device *in_dev = __in_dev_get_rcu(dev);
393
394 /* If err equals -EHOSTUNREACH the error is due to a
395 * martian destination or due to the fact that
396 * forwarding is disabled. For most martian packets,
397 * ip_route_output_key() will fail. It won't fail for 2 types of
398 * martian destinations: loopback destinations and destination
399 * 0.0.0.0. In both cases the packet will be dropped because the
400 * destination is the loopback device and not the bridge. */
401 if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
402 goto free_skb;
403
404 rt = ip_route_output(net, iph->daddr, 0,
405 RT_TOS(iph->tos), 0,
406 RT_SCOPE_UNIVERSE);
407 if (!IS_ERR(rt)) {
408 /* - Bridged-and-DNAT'ed traffic doesn't
409 * require ip_forwarding. */
410 if (rt->dst.dev == dev) {
411 skb_dst_drop(skb);
412 skb_dst_set(skb, &rt->dst);
413 goto bridged_dnat;
414 }
415 ip_rt_put(rt);
416 }
417 free_skb:
418 kfree_skb(skb);
419 return 0;
420 } else {
421 if (skb_dst(skb)->dev == dev) {
422 bridged_dnat:
423 skb->dev = br_indev;
424 nf_bridge_update_protocol(skb);
425 nf_bridge_push_encap_header(skb);
426 br_nf_hook_thresh(NF_BR_PRE_ROUTING,
427 net, sk, skb, skb->dev,
428 NULL,
429 br_nf_pre_routing_finish_bridge);
430 return 0;
431 }
432 ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
433 skb->pkt_type = PACKET_HOST;
434 }
435 } else {
436 rt = bridge_parent_rtable(br_indev);
437 if (!rt) {
438 kfree_skb(skb);
439 return 0;
440 }
441 skb_dst_drop(skb);
442 skb_dst_set_noref(skb, &rt->dst);
443 }
444
445 skb->dev = br_indev;
446 nf_bridge_update_protocol(skb);
447 nf_bridge_push_encap_header(skb);
448 br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL,
449 br_handle_frame_finish);
450 return 0;
451 }
452
453 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb,
454 const struct net_device *dev,
455 const struct net *net)
456 {
457 struct net_device *vlan, *br;
458 struct brnf_net *brnet = net_generic(net, brnf_net_id);
459
460 br = bridge_parent(dev);
461
462 if (brnet->pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
463 return br;
464
465 vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
466 skb_vlan_tag_get(skb) & VLAN_VID_MASK);
467
468 return vlan ? vlan : br;
469 }
470
471 /* Some common code for IPv4/IPv6 */
472 struct net_device *setup_pre_routing(struct sk_buff *skb, const struct net *net)
473 {
474 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
475
476 if (skb->pkt_type == PACKET_OTHERHOST) {
477 skb->pkt_type = PACKET_HOST;
478 nf_bridge->pkt_otherhost = true;
479 }
480
481 nf_bridge->in_prerouting = 1;
482 nf_bridge->physinif = skb->dev->ifindex;
483 skb->dev = brnf_get_logical_dev(skb, skb->dev, net);
484
485 if (skb->protocol == htons(ETH_P_8021Q))
486 nf_bridge->orig_proto = BRNF_PROTO_8021Q;
487 else if (skb->protocol == htons(ETH_P_PPP_SES))
488 nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
489
490 /* Must drop socket now because of tproxy. */
491 skb_orphan(skb);
492 return skb->dev;
493 }
494
495 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
496 * Replicate the checks that IPv4 does on packet reception.
497 * Set skb->dev to the bridge device (i.e. parent of the
498 * receiving device) to make netfilter happy, the REDIRECT
499 * target in particular. Save the original destination IP
500 * address to be able to detect DNAT afterwards. */
501 static unsigned int br_nf_pre_routing(void *priv,
502 struct sk_buff *skb,
503 const struct nf_hook_state *state)
504 {
505 struct nf_bridge_info *nf_bridge;
506 struct net_bridge_port *p;
507 struct net_bridge *br;
508 __u32 len = nf_bridge_encap_header_len(skb);
509 struct brnf_net *brnet;
510
511 if (unlikely(!pskb_may_pull(skb, len)))
512 return NF_DROP_REASON(skb, SKB_DROP_REASON_PKT_TOO_SMALL, 0);
513
514 p = br_port_get_rcu(state->in);
515 if (p == NULL)
516 return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
517 br = p->br;
518
519 brnet = net_generic(state->net, brnf_net_id);
520 if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) ||
521 is_pppoe_ipv6(skb, state->net)) {
522 if (!brnet->call_ip6tables &&
523 !br_opt_get(br, BROPT_NF_CALL_IP6TABLES))
524 return NF_ACCEPT;
525 if (!ipv6_mod_enabled()) {
526 pr_warn_once("Module ipv6 is disabled, so call_ip6tables is not supported.");
527 return NF_DROP_REASON(skb, SKB_DROP_REASON_IPV6DISABLED, 0);
528 }
529
530 nf_bridge_pull_encap_header_rcsum(skb);
531 return br_nf_pre_routing_ipv6(priv, skb, state);
532 }
533
534 if (!brnet->call_iptables && !br_opt_get(br, BROPT_NF_CALL_IPTABLES))
535 return NF_ACCEPT;
536
537 if (!IS_IP(skb) && !is_vlan_ip(skb, state->net) &&
538 !is_pppoe_ip(skb, state->net))
539 return NF_ACCEPT;
540
541 nf_bridge_pull_encap_header_rcsum(skb);
542
543 if (br_validate_ipv4(state->net, skb))
544 return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0);
545
546 if (!nf_bridge_alloc(skb))
547 return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0);
548 if (!setup_pre_routing(skb, state->net))
549 return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
550
551 nf_bridge = nf_bridge_info_get(skb);
552 nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
553
554 skb->protocol = htons(ETH_P_IP);
555 skb->transport_header = skb->network_header + ip_hdr(skb)->ihl * 4;
556
557 NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb,
558 skb->dev, NULL,
559 br_nf_pre_routing_finish);
560
561 return NF_STOLEN;
562 }
563
564 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
565 /* conntracks' nf_confirm logic cannot handle cloned skbs referencing
566 * the same nf_conn entry, which will happen for multicast (broadcast)
567 * Frames on bridges.
568 *
569 * Example:
570 * macvlan0
571 * br0
572 * ethX ethY
573 *
574 * ethX (or Y) receives multicast or broadcast packet containing
575 * an IP packet, not yet in conntrack table.
576 *
577 * 1. skb passes through bridge and fake-ip (br_netfilter)Prerouting.
578 * -> skb->_nfct now references a unconfirmed entry
579 * 2. skb is broad/mcast packet. bridge now passes clones out on each bridge
580 * interface.
581 * 3. skb gets passed up the stack.
582 * 4. In macvlan case, macvlan driver retains clone(s) of the mcast skb
583 * and schedules a work queue to send them out on the lower devices.
584 *
585 * The clone skb->_nfct is not a copy, it is the same entry as the
586 * original skb. The macvlan rx handler then returns RX_HANDLER_PASS.
587 * 5. Normal conntrack hooks (in NF_INET_LOCAL_IN) confirm the orig skb.
588 *
589 * The Macvlan broadcast worker and normal confirm path will race.
590 *
591 * This race will not happen if step 2 already confirmed a clone. In that
592 * case later steps perform skb_clone() with skb->_nfct already confirmed (in
593 * hash table). This works fine.
594 *
595 * But such confirmation won't happen when eb/ip/nftables rules dropped the
596 * packets before they reached the nf_confirm step in postrouting.
597 *
598 * Work around this problem by explicit confirmation of the entry at
599 * LOCAL_IN time, before upper layer has a chance to clone the unconfirmed
600 * entry.
601 *
602 */
603 static unsigned int br_nf_local_in(void *priv,
604 struct sk_buff *skb,
605 const struct nf_hook_state *state)
606 {
607 bool promisc = BR_INPUT_SKB_CB(skb)->promisc;
608 struct nf_conntrack *nfct = skb_nfct(skb);
609 const struct nf_ct_hook *ct_hook;
610 struct nf_conn *ct;
611 int ret;
612
613 if (promisc) {
614 nf_reset_ct(skb);
615 return NF_ACCEPT;
616 }
617
618 if (!nfct || skb->pkt_type == PACKET_HOST)
619 return NF_ACCEPT;
620
621 ct = container_of(nfct, struct nf_conn, ct_general);
622 if (likely(nf_ct_is_confirmed(ct)))
623 return NF_ACCEPT;
624
625 WARN_ON_ONCE(skb_shared(skb));
626 WARN_ON_ONCE(refcount_read(&nfct->use) != 1);
627
628 /* We can't call nf_confirm here, it would create a dependency
629 * on nf_conntrack module.
630 */
631 ct_hook = rcu_dereference(nf_ct_hook);
632 if (!ct_hook) {
633 skb->_nfct = 0ul;
634 nf_conntrack_put(nfct);
635 return NF_ACCEPT;
636 }
637
638 nf_bridge_pull_encap_header(skb);
639 ret = ct_hook->confirm(skb);
640 switch (ret & NF_VERDICT_MASK) {
641 case NF_STOLEN:
642 return NF_STOLEN;
643 default:
644 nf_bridge_push_encap_header(skb);
645 break;
646 }
647
648 ct = container_of(nfct, struct nf_conn, ct_general);
649 WARN_ON_ONCE(!nf_ct_is_confirmed(ct));
650
651 return ret;
652 }
653 #endif
654
655 /* PF_BRIDGE/FORWARD *************************************************/
656 static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
657 {
658 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
659 struct net_device *in;
660
661 if (!IS_ARP(skb) && !is_vlan_arp(skb, net)) {
662
663 if (skb->protocol == htons(ETH_P_IP))
664 nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
665
666 if (skb->protocol == htons(ETH_P_IPV6))
667 nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
668
669 in = nf_bridge_get_physindev(skb, net);
670 if (!in) {
671 kfree_skb(skb);
672 return 0;
673 }
674 if (nf_bridge->pkt_otherhost) {
675 skb->pkt_type = PACKET_OTHERHOST;
676 nf_bridge->pkt_otherhost = false;
677 }
678 nf_bridge_update_protocol(skb);
679 } else {
680 in = *((struct net_device **)(skb->cb));
681 }
682 nf_bridge_push_encap_header(skb);
683
684 br_nf_hook_thresh(NF_BR_FORWARD, net, sk, skb, in, skb->dev,
685 br_forward_finish);
686 return 0;
687 }
688
689
690 static unsigned int br_nf_forward_ip(struct sk_buff *skb,
691 const struct nf_hook_state *state,
692 u8 pf)
693 {
694 struct nf_bridge_info *nf_bridge;
695 struct net_device *parent;
696
697 nf_bridge = nf_bridge_info_get(skb);
698 if (!nf_bridge)
699 return NF_ACCEPT;
700
701 /* Need exclusive nf_bridge_info since we might have multiple
702 * different physoutdevs. */
703 if (!nf_bridge_unshare(skb))
704 return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0);
705
706 nf_bridge = nf_bridge_info_get(skb);
707 if (!nf_bridge)
708 return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0);
709
710 parent = bridge_parent(state->out);
711 if (!parent)
712 return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
713
714 nf_bridge_pull_encap_header(skb);
715
716 if (skb->pkt_type == PACKET_OTHERHOST) {
717 skb->pkt_type = PACKET_HOST;
718 nf_bridge->pkt_otherhost = true;
719 }
720
721 if (pf == NFPROTO_IPV4) {
722 if (br_validate_ipv4(state->net, skb))
723 return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0);
724 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
725 skb->protocol = htons(ETH_P_IP);
726 } else if (pf == NFPROTO_IPV6) {
727 if (br_validate_ipv6(state->net, skb))
728 return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0);
729 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
730 skb->protocol = htons(ETH_P_IPV6);
731 } else {
732 WARN_ON_ONCE(1);
733 return NF_DROP;
734 }
735
736 nf_bridge->physoutdev = skb->dev;
737
738 NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb,
739 brnf_get_logical_dev(skb, state->in, state->net),
740 parent, br_nf_forward_finish);
741
742 return NF_STOLEN;
743 }
744
745 static unsigned int br_nf_forward_arp(struct sk_buff *skb,
746 const struct nf_hook_state *state)
747 {
748 struct net_bridge_port *p;
749 struct net_bridge *br;
750 struct net_device **d = (struct net_device **)(skb->cb);
751 struct brnf_net *brnet;
752
753 p = br_port_get_rcu(state->out);
754 if (p == NULL)
755 return NF_ACCEPT;
756 br = p->br;
757
758 brnet = net_generic(state->net, brnf_net_id);
759 if (!brnet->call_arptables && !br_opt_get(br, BROPT_NF_CALL_ARPTABLES))
760 return NF_ACCEPT;
761
762 if (is_vlan_arp(skb, state->net))
763 nf_bridge_pull_encap_header(skb);
764
765 if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr))))
766 return NF_DROP_REASON(skb, SKB_DROP_REASON_PKT_TOO_SMALL, 0);
767
768 if (arp_hdr(skb)->ar_pln != 4) {
769 if (is_vlan_arp(skb, state->net))
770 nf_bridge_push_encap_header(skb);
771 return NF_ACCEPT;
772 }
773 *d = state->in;
774 NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb,
775 state->in, state->out, br_nf_forward_finish);
776
777 return NF_STOLEN;
778 }
779
780 /* This is the 'purely bridged' case. For IP, we pass the packet to
781 * netfilter with indev and outdev set to the bridge device,
782 * but we are still able to filter on the 'real' indev/outdev
783 * because of the physdev module. For ARP, indev and outdev are the
784 * bridge ports.
785 */
786 static unsigned int br_nf_forward(void *priv,
787 struct sk_buff *skb,
788 const struct nf_hook_state *state)
789 {
790 if (IS_IP(skb) || is_vlan_ip(skb, state->net) ||
791 is_pppoe_ip(skb, state->net))
792 return br_nf_forward_ip(skb, state, NFPROTO_IPV4);
793 if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) ||
794 is_pppoe_ipv6(skb, state->net))
795 return br_nf_forward_ip(skb, state, NFPROTO_IPV6);
796 if (IS_ARP(skb) || is_vlan_arp(skb, state->net))
797 return br_nf_forward_arp(skb, state);
798
799 return NF_ACCEPT;
800 }
801
802 static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
803 {
804 struct brnf_frag_data *data;
805 int err;
806
807 data = this_cpu_ptr(&brnf_frag_data_storage);
808 err = skb_cow_head(skb, data->size);
809
810 if (err) {
811 kfree_skb(skb);
812 return 0;
813 }
814
815 if (data->vlan_proto)
816 __vlan_hwaccel_put_tag(skb, data->vlan_proto, data->vlan_tci);
817
818 skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
819 __skb_push(skb, data->encap_size);
820
821 nf_bridge_info_free(skb);
822 return br_dev_queue_push_xmit(net, sk, skb);
823 }
824
825 static int
826 br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
827 int (*output)(struct net *, struct sock *, struct sk_buff *))
828 {
829 unsigned int mtu = ip_skb_dst_mtu(sk, skb);
830 struct iphdr *iph = ip_hdr(skb);
831
832 if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
833 (IPCB(skb)->frag_max_size &&
834 IPCB(skb)->frag_max_size > mtu))) {
835 IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
836 kfree_skb(skb);
837 return -EMSGSIZE;
838 }
839
840 return ip_do_fragment(net, sk, skb, output);
841 }
842
843 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
844 {
845 const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
846
847 if (nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
848 return PPPOE_SES_HLEN;
849 return 0;
850 }
851
852 static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
853 {
854 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
855 unsigned int mtu, mtu_reserved;
856 int ret;
857
858 mtu_reserved = nf_bridge_mtu_reduction(skb);
859 mtu = skb->dev->mtu;
860
861 if (nf_bridge->pkt_otherhost) {
862 skb->pkt_type = PACKET_OTHERHOST;
863 nf_bridge->pkt_otherhost = false;
864 }
865
866 if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu)
867 mtu = nf_bridge->frag_max_size;
868
869 nf_bridge_update_protocol(skb);
870 nf_bridge_push_encap_header(skb);
871
872 if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) {
873 nf_bridge_info_free(skb);
874 return br_dev_queue_push_xmit(net, sk, skb);
875 }
876
877 /* This is wrong! We should preserve the original fragment
878 * boundaries by preserving frag_list rather than refragmenting.
879 */
880 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) &&
881 skb->protocol == htons(ETH_P_IP)) {
882 struct brnf_frag_data *data;
883
884 if (br_validate_ipv4(net, skb))
885 goto drop;
886
887 IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
888
889 local_lock_nested_bh(&brnf_frag_data_storage.bh_lock);
890 data = this_cpu_ptr(&brnf_frag_data_storage);
891
892 if (skb_vlan_tag_present(skb)) {
893 data->vlan_tci = skb->vlan_tci;
894 data->vlan_proto = skb->vlan_proto;
895 } else {
896 data->vlan_proto = 0;
897 }
898
899 data->encap_size = nf_bridge_encap_header_len(skb);
900 data->size = ETH_HLEN + data->encap_size;
901
902 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
903 data->size);
904
905 ret = br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit);
906 local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock);
907 return ret;
908 }
909 if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) &&
910 skb->protocol == htons(ETH_P_IPV6)) {
911 const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
912 struct brnf_frag_data *data;
913
914 if (br_validate_ipv6(net, skb))
915 goto drop;
916
917 IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
918
919 local_lock_nested_bh(&brnf_frag_data_storage.bh_lock);
920 data = this_cpu_ptr(&brnf_frag_data_storage);
921 data->encap_size = nf_bridge_encap_header_len(skb);
922 data->size = ETH_HLEN + data->encap_size;
923
924 skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
925 data->size);
926
927 if (v6ops) {
928 ret = v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit);
929 local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock);
930 return ret;
931 }
932 local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock);
933
934 kfree_skb(skb);
935 return -EMSGSIZE;
936 }
937 nf_bridge_info_free(skb);
938 return br_dev_queue_push_xmit(net, sk, skb);
939 drop:
940 kfree_skb(skb);
941 return 0;
942 }
943
944 /* PF_BRIDGE/POST_ROUTING ********************************************/
945 static unsigned int br_nf_post_routing(void *priv,
946 struct sk_buff *skb,
947 const struct nf_hook_state *state)
948 {
949 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
950 struct net_device *realoutdev = bridge_parent(skb->dev);
951 u_int8_t pf;
952
953 /* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
954 * on a bridge, but was delivered locally and is now being routed:
955 *
956 * POST_ROUTING was already invoked from the ip stack.
957 */
958 if (!nf_bridge || !nf_bridge->physoutdev)
959 return NF_ACCEPT;
960
961 if (!realoutdev)
962 return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
963
964 if (IS_IP(skb) || is_vlan_ip(skb, state->net) ||
965 is_pppoe_ip(skb, state->net))
966 pf = NFPROTO_IPV4;
967 else if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) ||
968 is_pppoe_ipv6(skb, state->net))
969 pf = NFPROTO_IPV6;
970 else
971 return NF_ACCEPT;
972
973 if (skb->pkt_type == PACKET_OTHERHOST) {
974 skb->pkt_type = PACKET_HOST;
975 nf_bridge->pkt_otherhost = true;
976 }
977
978 nf_bridge_pull_encap_header(skb);
979 if (pf == NFPROTO_IPV4)
980 skb->protocol = htons(ETH_P_IP);
981 else
982 skb->protocol = htons(ETH_P_IPV6);
983
984 NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb,
985 NULL, realoutdev,
986 br_nf_dev_queue_xmit);
987
988 return NF_STOLEN;
989 }
990
991 /* IP/SABOTAGE *****************************************************/
992 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
993 * for the second time. */
994 static unsigned int ip_sabotage_in(void *priv,
995 struct sk_buff *skb,
996 const struct nf_hook_state *state)
997 {
998 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
999
1000 if (nf_bridge) {
1001 if (nf_bridge->sabotage_in_done)
1002 return NF_ACCEPT;
1003
1004 if (!nf_bridge->in_prerouting &&
1005 !netif_is_l3_master(skb->dev) &&
1006 !netif_is_l3_slave(skb->dev)) {
1007 nf_bridge->sabotage_in_done = 1;
1008 state->okfn(state->net, state->sk, skb);
1009 return NF_STOLEN;
1010 }
1011 }
1012
1013 return NF_ACCEPT;
1014 }
1015
1016 /* This is called when br_netfilter has called into iptables/netfilter,
1017 * and DNAT has taken place on a bridge-forwarded packet.
1018 *
1019 * neigh->output has created a new MAC header, with local br0 MAC
1020 * as saddr.
1021 *
1022 * This restores the original MAC saddr of the bridged packet
1023 * before invoking bridge forward logic to transmit the packet.
1024 */
1025 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
1026 {
1027 struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1028 struct net_device *br_indev;
1029
1030 br_indev = nf_bridge_get_physindev(skb, dev_net(skb->dev));
1031 if (!br_indev) {
1032 kfree_skb(skb);
1033 return;
1034 }
1035
1036 skb_pull(skb, ETH_HLEN);
1037 nf_bridge->bridged_dnat = 0;
1038
1039 BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
1040
1041 skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
1042 nf_bridge->neigh_header,
1043 ETH_HLEN - ETH_ALEN);
1044 skb->dev = br_indev;
1045
1046 nf_bridge->physoutdev = NULL;
1047 br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
1048 }
1049
1050 static int br_nf_dev_xmit(struct sk_buff *skb)
1051 {
1052 const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1053
1054 if (nf_bridge && nf_bridge->bridged_dnat) {
1055 br_nf_pre_routing_finish_bridge_slow(skb);
1056 return 1;
1057 }
1058 return 0;
1059 }
1060
1061 static const struct nf_br_ops br_ops = {
1062 .br_dev_xmit_hook = br_nf_dev_xmit,
1063 };
1064
1065 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
1066 * br_dev_queue_push_xmit is called afterwards */
1067 static const struct nf_hook_ops br_nf_ops[] = {
1068 {
1069 .hook = br_nf_pre_routing,
1070 .pf = NFPROTO_BRIDGE,
1071 .hooknum = NF_BR_PRE_ROUTING,
1072 .priority = NF_BR_PRI_BRNF,
1073 },
1074 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1075 {
1076 .hook = br_nf_local_in,
1077 .pf = NFPROTO_BRIDGE,
1078 .hooknum = NF_BR_LOCAL_IN,
1079 .priority = NF_BR_PRI_LAST,
1080 },
1081 #endif
1082 {
1083 .hook = br_nf_forward,
1084 .pf = NFPROTO_BRIDGE,
1085 .hooknum = NF_BR_FORWARD,
1086 .priority = NF_BR_PRI_BRNF,
1087 },
1088 {
1089 .hook = br_nf_post_routing,
1090 .pf = NFPROTO_BRIDGE,
1091 .hooknum = NF_BR_POST_ROUTING,
1092 .priority = NF_BR_PRI_LAST,
1093 },
1094 {
1095 .hook = ip_sabotage_in,
1096 .pf = NFPROTO_IPV4,
1097 .hooknum = NF_INET_PRE_ROUTING,
1098 .priority = NF_IP_PRI_FIRST,
1099 },
1100 {
1101 .hook = ip_sabotage_in,
1102 .pf = NFPROTO_IPV6,
1103 .hooknum = NF_INET_PRE_ROUTING,
1104 .priority = NF_IP6_PRI_FIRST,
1105 },
1106 };
1107
1108 static int brnf_device_event(struct notifier_block *unused, unsigned long event,
1109 void *ptr)
1110 {
1111 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1112 struct brnf_net *brnet;
1113 struct net *net;
1114 int ret;
1115
1116 if (event != NETDEV_REGISTER || !netif_is_bridge_master(dev))
1117 return NOTIFY_DONE;
1118
1119 ASSERT_RTNL();
1120
1121 net = dev_net(dev);
1122 brnet = net_generic(net, brnf_net_id);
1123 if (brnet->enabled)
1124 return NOTIFY_OK;
1125
1126 ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
1127 if (ret)
1128 return NOTIFY_BAD;
1129
1130 brnet->enabled = true;
1131 return NOTIFY_OK;
1132 }
1133
1134 static struct notifier_block brnf_notifier __read_mostly = {
1135 .notifier_call = brnf_device_event,
1136 };
1137
1138 /* recursively invokes nf_hook_slow (again), skipping already-called
1139 * hooks (< NF_BR_PRI_BRNF).
1140 *
1141 * Called with rcu read lock held.
1142 */
1143 int br_nf_hook_thresh(unsigned int hook, struct net *net,
1144 struct sock *sk, struct sk_buff *skb,
1145 struct net_device *indev,
1146 struct net_device *outdev,
1147 int (*okfn)(struct net *, struct sock *,
1148 struct sk_buff *))
1149 {
1150 const struct nf_hook_entries *e;
1151 struct nf_hook_state state;
1152 struct nf_hook_ops **ops;
1153 unsigned int i;
1154 int ret;
1155
1156 e = rcu_dereference(net->nf.hooks_bridge[hook]);
1157 if (!e)
1158 return okfn(net, sk, skb);
1159
1160 ops = nf_hook_entries_get_hook_ops(e);
1161 for (i = 0; i < e->num_hook_entries; i++) {
1162 /* These hooks have already been called */
1163 if (ops[i]->priority < NF_BR_PRI_BRNF)
1164 continue;
1165
1166 /* These hooks have not been called yet, run them. */
1167 if (ops[i]->priority > NF_BR_PRI_BRNF)
1168 break;
1169
1170 /* take a closer look at NF_BR_PRI_BRNF. */
1171 if (ops[i]->hook == br_nf_pre_routing) {
1172 /* This hook diverted the skb to this function,
1173 * hooks after this have not been run yet.
1174 */
1175 i++;
1176 break;
1177 }
1178 }
1179
1180 nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev,
1181 sk, net, okfn);
1182
1183 ret = nf_hook_slow(skb, &state, e, i);
1184 if (ret == 1)
1185 ret = okfn(net, sk, skb);
1186
1187 return ret;
1188 }
1189
1190 #ifdef CONFIG_SYSCTL
1191 static
1192 int brnf_sysctl_call_tables(const struct ctl_table *ctl, int write,
1193 void *buffer, size_t *lenp, loff_t *ppos)
1194 {
1195 int ret;
1196
1197 ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1198
1199 if (write && *(int *)(ctl->data))
1200 *(int *)(ctl->data) = 1;
1201 return ret;
1202 }
1203
1204 static struct ctl_table brnf_table[] = {
1205 {
1206 .procname = "bridge-nf-call-arptables",
1207 .maxlen = sizeof(int),
1208 .mode = 0644,
1209 .proc_handler = brnf_sysctl_call_tables,
1210 },
1211 {
1212 .procname = "bridge-nf-call-iptables",
1213 .maxlen = sizeof(int),
1214 .mode = 0644,
1215 .proc_handler = brnf_sysctl_call_tables,
1216 },
1217 {
1218 .procname = "bridge-nf-call-ip6tables",
1219 .maxlen = sizeof(int),
1220 .mode = 0644,
1221 .proc_handler = brnf_sysctl_call_tables,
1222 },
1223 {
1224 .procname = "bridge-nf-filter-vlan-tagged",
1225 .maxlen = sizeof(int),
1226 .mode = 0644,
1227 .proc_handler = brnf_sysctl_call_tables,
1228 },
1229 {
1230 .procname = "bridge-nf-filter-pppoe-tagged",
1231 .maxlen = sizeof(int),
1232 .mode = 0644,
1233 .proc_handler = brnf_sysctl_call_tables,
1234 },
1235 {
1236 .procname = "bridge-nf-pass-vlan-input-dev",
1237 .maxlen = sizeof(int),
1238 .mode = 0644,
1239 .proc_handler = brnf_sysctl_call_tables,
1240 },
1241 };
1242
1243 static inline void br_netfilter_sysctl_default(struct brnf_net *brnf)
1244 {
1245 brnf->call_iptables = 1;
1246 brnf->call_ip6tables = 1;
1247 brnf->call_arptables = 1;
1248 brnf->filter_vlan_tagged = 0;
1249 brnf->filter_pppoe_tagged = 0;
1250 brnf->pass_vlan_indev = 0;
1251 }
1252
1253 static int br_netfilter_sysctl_init_net(struct net *net)
1254 {
1255 struct ctl_table *table = brnf_table;
1256 struct brnf_net *brnet;
1257
1258 if (!net_eq(net, &init_net)) {
1259 table = kmemdup(table, sizeof(brnf_table), GFP_KERNEL);
1260 if (!table)
1261 return -ENOMEM;
1262 }
1263
1264 brnet = net_generic(net, brnf_net_id);
1265 table[0].data = &brnet->call_arptables;
1266 table[1].data = &brnet->call_iptables;
1267 table[2].data = &brnet->call_ip6tables;
1268 table[3].data = &brnet->filter_vlan_tagged;
1269 table[4].data = &brnet->filter_pppoe_tagged;
1270 table[5].data = &brnet->pass_vlan_indev;
1271
1272 br_netfilter_sysctl_default(brnet);
1273
1274 brnet->ctl_hdr = register_net_sysctl_sz(net, "net/bridge", table,
1275 ARRAY_SIZE(brnf_table));
1276 if (!brnet->ctl_hdr) {
1277 if (!net_eq(net, &init_net))
1278 kfree(table);
1279
1280 return -ENOMEM;
1281 }
1282
1283 return 0;
1284 }
1285
1286 static void br_netfilter_sysctl_exit_net(struct net *net,
1287 struct brnf_net *brnet)
1288 {
1289 const struct ctl_table *table = brnet->ctl_hdr->ctl_table_arg;
1290
1291 unregister_net_sysctl_table(brnet->ctl_hdr);
1292 if (!net_eq(net, &init_net))
1293 kfree(table);
1294 }
1295
1296 static int __net_init brnf_init_net(struct net *net)
1297 {
1298 return br_netfilter_sysctl_init_net(net);
1299 }
1300 #endif
1301
1302 static void __net_exit brnf_exit_net(struct net *net)
1303 {
1304 struct brnf_net *brnet;
1305
1306 brnet = net_generic(net, brnf_net_id);
1307 if (brnet->enabled) {
1308 nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
1309 brnet->enabled = false;
1310 }
1311
1312 #ifdef CONFIG_SYSCTL
1313 br_netfilter_sysctl_exit_net(net, brnet);
1314 #endif
1315 }
1316
1317 static struct pernet_operations brnf_net_ops __read_mostly = {
1318 #ifdef CONFIG_SYSCTL
1319 .init = brnf_init_net,
1320 #endif
1321 .exit = brnf_exit_net,
1322 .id = &brnf_net_id,
1323 .size = sizeof(struct brnf_net),
1324 };
1325
1326 static int __init br_netfilter_init(void)
1327 {
1328 int ret;
1329
1330 ret = register_pernet_subsys(&brnf_net_ops);
1331 if (ret < 0)
1332 return ret;
1333
1334 ret = register_netdevice_notifier(&brnf_notifier);
1335 if (ret < 0) {
1336 unregister_pernet_subsys(&brnf_net_ops);
1337 return ret;
1338 }
1339
1340 RCU_INIT_POINTER(nf_br_ops, &br_ops);
1341 printk(KERN_NOTICE "Bridge firewalling registered\n");
1342 return 0;
1343 }
1344
1345 static void __exit br_netfilter_fini(void)
1346 {
1347 RCU_INIT_POINTER(nf_br_ops, NULL);
1348 unregister_netdevice_notifier(&brnf_notifier);
1349 unregister_pernet_subsys(&brnf_net_ops);
1350 }
1351
1352 module_init(br_netfilter_init);
1353 module_exit(br_netfilter_fini);
1354
1355 MODULE_LICENSE("GPL");
1356 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1357 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1358 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");
1359
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.