xref: /linux/net/bridge/br_netfilter_hooks.c (revision 1d227fcc72223cbdd34d0ce13541cbaab5e0d72f)
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/dst_metadata.h>
37 #include <net/route.h>
38 #include <net/netfilter/br_netfilter.h>
39 #include <net/netns/generic.h>
40 #include <net/inet_dscp.h>
41 
42 #include <linux/uaccess.h>
43 #include "br_private.h"
44 #ifdef CONFIG_SYSCTL
45 #include <linux/sysctl.h>
46 #endif
47 
48 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
49 #include <net/netfilter/nf_conntrack_core.h>
50 #endif
51 
52 static unsigned int brnf_net_id __read_mostly;
53 
54 struct brnf_net {
55 	bool enabled;
56 
57 #ifdef CONFIG_SYSCTL
58 	struct ctl_table_header *ctl_hdr;
59 #endif
60 
61 	/* default value is 1 */
62 	int call_iptables;
63 	int call_ip6tables;
64 	int call_arptables;
65 
66 	/* default value is 0 */
67 	int filter_vlan_tagged;
68 	int filter_pppoe_tagged;
69 	int pass_vlan_indev;
70 };
71 
72 #define IS_IP(skb) \
73 	(!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IP))
74 
75 #define IS_IPV6(skb) \
76 	(!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_IPV6))
77 
78 #define IS_ARP(skb) \
79 	(!skb_vlan_tag_present(skb) && skb->protocol == htons(ETH_P_ARP))
80 
vlan_proto(const struct sk_buff * skb)81 static inline __be16 vlan_proto(const struct sk_buff *skb)
82 {
83 	if (skb_vlan_tag_present(skb))
84 		return skb->protocol;
85 	else if (skb->protocol == htons(ETH_P_8021Q))
86 		return vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
87 	else
88 		return 0;
89 }
90 
is_vlan_ip(const struct sk_buff * skb,const struct net * net)91 static inline bool is_vlan_ip(const struct sk_buff *skb, const struct net *net)
92 {
93 	struct brnf_net *brnet = net_generic(net, brnf_net_id);
94 
95 	return vlan_proto(skb) == htons(ETH_P_IP) && brnet->filter_vlan_tagged;
96 }
97 
is_vlan_ipv6(const struct sk_buff * skb,const struct net * net)98 static inline bool is_vlan_ipv6(const struct sk_buff *skb,
99 				const struct net *net)
100 {
101 	struct brnf_net *brnet = net_generic(net, brnf_net_id);
102 
103 	return vlan_proto(skb) == htons(ETH_P_IPV6) &&
104 	       brnet->filter_vlan_tagged;
105 }
106 
is_vlan_arp(const struct sk_buff * skb,const struct net * net)107 static inline bool is_vlan_arp(const struct sk_buff *skb, const struct net *net)
108 {
109 	struct brnf_net *brnet = net_generic(net, brnf_net_id);
110 
111 	return vlan_proto(skb) == htons(ETH_P_ARP) && brnet->filter_vlan_tagged;
112 }
113 
pppoe_proto(const struct sk_buff * skb)114 static inline __be16 pppoe_proto(const struct sk_buff *skb)
115 {
116 	return *((__be16 *)(skb_mac_header(skb) + ETH_HLEN +
117 			    sizeof(struct pppoe_hdr)));
118 }
119 
is_pppoe_ip(const struct sk_buff * skb,const struct net * net)120 static inline bool is_pppoe_ip(const struct sk_buff *skb, const struct net *net)
121 {
122 	struct brnf_net *brnet = net_generic(net, brnf_net_id);
123 
124 	return skb->protocol == htons(ETH_P_PPP_SES) &&
125 	       pppoe_proto(skb) == htons(PPP_IP) && brnet->filter_pppoe_tagged;
126 }
127 
is_pppoe_ipv6(const struct sk_buff * skb,const struct net * net)128 static inline bool is_pppoe_ipv6(const struct sk_buff *skb,
129 				 const struct net *net)
130 {
131 	struct brnf_net *brnet = net_generic(net, brnf_net_id);
132 
133 	return skb->protocol == htons(ETH_P_PPP_SES) &&
134 	       pppoe_proto(skb) == htons(PPP_IPV6) &&
135 	       brnet->filter_pppoe_tagged;
136 }
137 
138 /* largest possible L2 header, see br_nf_dev_queue_xmit() */
139 #define NF_BRIDGE_MAX_MAC_HEADER_LENGTH (PPPOE_SES_HLEN + ETH_HLEN)
140 
141 struct brnf_frag_data {
142 	local_lock_t bh_lock;
143 	char mac[NF_BRIDGE_MAX_MAC_HEADER_LENGTH];
144 	u8 encap_size;
145 	u8 size;
146 	u16 vlan_tci;
147 	__be16 vlan_proto;
148 };
149 
150 static DEFINE_PER_CPU(struct brnf_frag_data, brnf_frag_data_storage) = {
151 	.bh_lock = INIT_LOCAL_LOCK(bh_lock),
152 };
153 
nf_bridge_info_free(struct sk_buff * skb)154 static void nf_bridge_info_free(struct sk_buff *skb)
155 {
156 	skb_ext_del(skb, SKB_EXT_BRIDGE_NF);
157 }
158 
bridge_parent(const struct net_device * dev)159 static inline struct net_device *bridge_parent(const struct net_device *dev)
160 {
161 	struct net_bridge_port *port;
162 
163 	port = br_port_get_rcu(dev);
164 	return port ? port->br->dev : NULL;
165 }
166 
nf_bridge_unshare(struct sk_buff * skb)167 static inline struct nf_bridge_info *nf_bridge_unshare(struct sk_buff *skb)
168 {
169 	return skb_ext_add(skb, SKB_EXT_BRIDGE_NF);
170 }
171 
nf_bridge_encap_header_len(const struct sk_buff * skb)172 unsigned int nf_bridge_encap_header_len(const struct sk_buff *skb)
173 {
174 	switch (skb->protocol) {
175 	case __cpu_to_be16(ETH_P_8021Q):
176 		return VLAN_HLEN;
177 	case __cpu_to_be16(ETH_P_PPP_SES):
178 		return PPPOE_SES_HLEN;
179 	default:
180 		return 0;
181 	}
182 }
183 
nf_bridge_pull_encap_header(struct sk_buff * skb)184 static inline void nf_bridge_pull_encap_header(struct sk_buff *skb)
185 {
186 	unsigned int len = nf_bridge_encap_header_len(skb);
187 
188 	skb_pull(skb, len);
189 	skb->network_header += len;
190 }
191 
nf_bridge_pull_encap_header_rcsum(struct sk_buff * skb)192 static inline void nf_bridge_pull_encap_header_rcsum(struct sk_buff *skb)
193 {
194 	unsigned int len = nf_bridge_encap_header_len(skb);
195 
196 	skb_pull_rcsum(skb, len);
197 	skb->network_header += len;
198 }
199 
200 /* When handing a packet over to the IP layer
201  * check whether we have a skb that is in the
202  * expected format
203  */
204 
br_validate_ipv4(struct net * net,struct sk_buff * skb)205 static int br_validate_ipv4(struct net *net, struct sk_buff *skb)
206 {
207 	const struct iphdr *iph;
208 	u32 len;
209 
210 	if (!pskb_may_pull(skb, sizeof(struct iphdr)))
211 		goto inhdr_error;
212 
213 	iph = ip_hdr(skb);
214 
215 	/* Basic sanity checks */
216 	if (iph->ihl < 5 || iph->version != 4)
217 		goto inhdr_error;
218 
219 	if (!pskb_may_pull(skb, iph->ihl*4))
220 		goto inhdr_error;
221 
222 	iph = ip_hdr(skb);
223 	if (unlikely(ip_fast_csum((u8 *)iph, iph->ihl)))
224 		goto csum_error;
225 
226 	len = skb_ip_totlen(skb);
227 	if (skb->len < len) {
228 		__IP_INC_STATS(net, IPSTATS_MIB_INTRUNCATEDPKTS);
229 		goto drop;
230 	} else if (len < (iph->ihl*4))
231 		goto inhdr_error;
232 
233 	if (pskb_trim_rcsum(skb, len)) {
234 		__IP_INC_STATS(net, IPSTATS_MIB_INDISCARDS);
235 		goto drop;
236 	}
237 
238 	memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
239 	/* We should really parse IP options here but until
240 	 * somebody who actually uses IP options complains to
241 	 * us we'll just silently ignore the options because
242 	 * we're lazy!
243 	 */
244 	return 0;
245 
246 csum_error:
247 	__IP_INC_STATS(net, IPSTATS_MIB_CSUMERRORS);
248 inhdr_error:
249 	__IP_INC_STATS(net, IPSTATS_MIB_INHDRERRORS);
250 drop:
251 	return -1;
252 }
253 
nf_bridge_update_protocol(struct sk_buff * skb)254 void nf_bridge_update_protocol(struct sk_buff *skb)
255 {
256 	const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
257 
258 	switch (nf_bridge->orig_proto) {
259 	case BRNF_PROTO_8021Q:
260 		skb->protocol = htons(ETH_P_8021Q);
261 		break;
262 	case BRNF_PROTO_PPPOE:
263 		skb->protocol = htons(ETH_P_PPP_SES);
264 		break;
265 	case BRNF_PROTO_UNCHANGED:
266 		break;
267 	}
268 }
269 
270 /* Obtain the correct destination MAC address, while preserving the original
271  * source MAC address. If we already know this address, we just copy it. If we
272  * don't, we use the neighbour framework to find out. In both cases, we make
273  * sure that br_handle_frame_finish() is called afterwards.
274  */
br_nf_pre_routing_finish_bridge(struct net * net,struct sock * sk,struct sk_buff * skb)275 int br_nf_pre_routing_finish_bridge(struct net *net, struct sock *sk, struct sk_buff *skb)
276 {
277 	struct neighbour *neigh;
278 	struct dst_entry *dst;
279 
280 	skb->dev = bridge_parent(skb->dev);
281 	if (!skb->dev)
282 		goto free_skb;
283 	dst = skb_dst(skb);
284 	neigh = dst_neigh_lookup_skb(dst, skb);
285 	if (neigh) {
286 		struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
287 		int ret;
288 
289 		if ((READ_ONCE(neigh->nud_state) & NUD_CONNECTED) &&
290 		    READ_ONCE(neigh->hh.hh_len)) {
291 			struct net_device *br_indev;
292 
293 			br_indev = nf_bridge_get_physindev(skb, net);
294 			if (!br_indev) {
295 				neigh_release(neigh);
296 				goto free_skb;
297 			}
298 
299 			neigh_hh_bridge(&neigh->hh, skb);
300 			skb->dev = br_indev;
301 
302 			ret = br_handle_frame_finish(net, sk, skb);
303 		} else {
304 			/* the neighbour function below overwrites the complete
305 			 * MAC header, so we save the Ethernet source address and
306 			 * protocol number.
307 			 */
308 			skb_copy_from_linear_data_offset(skb,
309 							 -(ETH_HLEN-ETH_ALEN),
310 							 nf_bridge->neigh_header,
311 							 ETH_HLEN-ETH_ALEN);
312 			/* tell br_dev_xmit to continue with forwarding */
313 			nf_bridge->bridged_dnat = 1;
314 			/* FIXME Need to refragment */
315 			ret = READ_ONCE(neigh->output)(neigh, skb);
316 		}
317 		neigh_release(neigh);
318 		return ret;
319 	}
320 free_skb:
321 	kfree_skb(skb);
322 	return 0;
323 }
324 
325 static inline bool
br_nf_ipv4_daddr_was_changed(const struct sk_buff * skb,const struct nf_bridge_info * nf_bridge)326 br_nf_ipv4_daddr_was_changed(const struct sk_buff *skb,
327 			     const struct nf_bridge_info *nf_bridge)
328 {
329 	return ip_hdr(skb)->daddr != nf_bridge->ipv4_daddr;
330 }
331 
332 /* This requires some explaining. If DNAT has taken place,
333  * we will need to fix up the destination Ethernet address.
334  * This is also true when SNAT takes place (for the reply direction).
335  *
336  * There are two cases to consider:
337  * 1. The packet was DNAT'ed to a device in the same bridge
338  *    port group as it was received on. We can still bridge
339  *    the packet.
340  * 2. The packet was DNAT'ed to a different device, either
341  *    a non-bridged device or another bridge port group.
342  *    The packet will need to be routed.
343  *
344  * The correct way of distinguishing between these two cases is to
345  * call ip_route_input() and to look at skb->dst->dev, which is
346  * changed to the destination device if ip_route_input() succeeds.
347  *
348  * Let's first consider the case that ip_route_input() succeeds:
349  *
350  * If the output device equals the logical bridge device the packet
351  * came in on, we can consider this bridging. The corresponding MAC
352  * address will be obtained in br_nf_pre_routing_finish_bridge.
353  * Otherwise, the packet is considered to be routed and we just
354  * change the destination MAC address so that the packet will
355  * later be passed up to the IP stack to be routed. For a redirected
356  * packet, ip_route_input() will give back the localhost as output device,
357  * which differs from the bridge device.
358  *
359  * Let's now consider the case that ip_route_input() fails:
360  *
361  * This can be because the destination address is martian, in which case
362  * the packet will be dropped.
363  * If IP forwarding is disabled, ip_route_input() will fail, while
364  * ip_route_output_key() can return success. The source
365  * address for ip_route_output_key() is set to zero, so ip_route_output_key()
366  * thinks we're handling a locally generated packet and won't care
367  * if IP forwarding is enabled. If the output device equals the logical bridge
368  * device, we proceed as if ip_route_input() succeeded. If it differs from the
369  * logical bridge port or if ip_route_output_key() fails we drop the packet.
370  */
br_nf_pre_routing_finish(struct net * net,struct sock * sk,struct sk_buff * skb)371 static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
372 {
373 	struct net_device *dev = skb->dev, *br_indev;
374 	struct iphdr *iph = ip_hdr(skb);
375 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
376 	struct rtable *rt;
377 	int err;
378 
379 	br_indev = nf_bridge_get_physindev(skb, net);
380 	if (!br_indev) {
381 		kfree_skb(skb);
382 		return 0;
383 	}
384 
385 	nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
386 
387 	if (nf_bridge->pkt_otherhost) {
388 		skb->pkt_type = PACKET_OTHERHOST;
389 		nf_bridge->pkt_otherhost = false;
390 	}
391 	nf_bridge->in_prerouting = 0;
392 	if (br_nf_ipv4_daddr_was_changed(skb, nf_bridge)) {
393 		if ((err = ip_route_input(skb, iph->daddr, iph->saddr, iph->tos, dev))) {
394 			struct in_device *in_dev = __in_dev_get_rcu(dev);
395 
396 			/* If err equals -EHOSTUNREACH the error is due to a
397 			 * martian destination or due to the fact that
398 			 * forwarding is disabled. For most martian packets,
399 			 * ip_route_output_key() will fail. It won't fail for 2 types of
400 			 * martian destinations: loopback destinations and destination
401 			 * 0.0.0.0. In both cases the packet will be dropped because the
402 			 * destination is the loopback device and not the bridge. */
403 			if (err != -EHOSTUNREACH || !in_dev || IN_DEV_FORWARD(in_dev))
404 				goto free_skb;
405 
406 			rt = ip_route_output(net, iph->daddr, 0,
407 					     iph->tos & INET_DSCP_MASK, 0,
408 					     RT_SCOPE_UNIVERSE);
409 			if (!IS_ERR(rt)) {
410 				/* - Bridged-and-DNAT'ed traffic doesn't
411 				 *   require ip_forwarding. */
412 				if (rt->dst.dev == dev) {
413 					skb_dst_drop(skb);
414 					skb_dst_set(skb, &rt->dst);
415 					goto bridged_dnat;
416 				}
417 				ip_rt_put(rt);
418 			}
419 free_skb:
420 			kfree_skb(skb);
421 			return 0;
422 		} else {
423 			if (skb_dst(skb)->dev == dev) {
424 bridged_dnat:
425 				skb->dev = br_indev;
426 				nf_bridge_update_protocol(skb);
427 				nf_bridge_push_encap_header(skb);
428 				br_nf_hook_thresh(NF_BR_PRE_ROUTING,
429 						  net, sk, skb, skb->dev,
430 						  NULL,
431 						  br_nf_pre_routing_finish_bridge);
432 				return 0;
433 			}
434 			ether_addr_copy(eth_hdr(skb)->h_dest, dev->dev_addr);
435 			skb->pkt_type = PACKET_HOST;
436 		}
437 	} else {
438 		rt = bridge_parent_rtable(br_indev);
439 		if (!rt) {
440 			kfree_skb(skb);
441 			return 0;
442 		}
443 		skb_dst_drop(skb);
444 		skb_dst_set_noref(skb, &rt->dst);
445 	}
446 
447 	skb->dev = br_indev;
448 	nf_bridge_update_protocol(skb);
449 	nf_bridge_push_encap_header(skb);
450 	br_nf_hook_thresh(NF_BR_PRE_ROUTING, net, sk, skb, skb->dev, NULL,
451 			  br_handle_frame_finish);
452 	return 0;
453 }
454 
brnf_get_logical_dev(struct sk_buff * skb,const struct net_device * dev,const struct net * net)455 static struct net_device *brnf_get_logical_dev(struct sk_buff *skb,
456 					       const struct net_device *dev,
457 					       const struct net *net)
458 {
459 	struct net_device *vlan, *br;
460 	struct brnf_net *brnet = net_generic(net, brnf_net_id);
461 
462 	br = bridge_parent(dev);
463 
464 	if (brnet->pass_vlan_indev == 0 || !skb_vlan_tag_present(skb))
465 		return br;
466 
467 	vlan = __vlan_find_dev_deep_rcu(br, skb->vlan_proto,
468 				    skb_vlan_tag_get(skb) & VLAN_VID_MASK);
469 
470 	return vlan ? vlan : br;
471 }
472 
473 /* Some common code for IPv4/IPv6 */
setup_pre_routing(struct sk_buff * skb,const struct net * net)474 struct net_device *setup_pre_routing(struct sk_buff *skb, const struct net *net)
475 {
476 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
477 
478 	if (skb->pkt_type == PACKET_OTHERHOST) {
479 		skb->pkt_type = PACKET_HOST;
480 		nf_bridge->pkt_otherhost = true;
481 	}
482 
483 	nf_bridge->in_prerouting = 1;
484 	nf_bridge->physinif = skb->dev->ifindex;
485 	skb->dev = brnf_get_logical_dev(skb, skb->dev, net);
486 
487 	if (skb->protocol == htons(ETH_P_8021Q))
488 		nf_bridge->orig_proto = BRNF_PROTO_8021Q;
489 	else if (skb->protocol == htons(ETH_P_PPP_SES))
490 		nf_bridge->orig_proto = BRNF_PROTO_PPPOE;
491 
492 	/* Must drop socket now because of tproxy. */
493 	skb_orphan(skb);
494 	return skb->dev;
495 }
496 
497 /* Direct IPv6 traffic to br_nf_pre_routing_ipv6.
498  * Replicate the checks that IPv4 does on packet reception.
499  * Set skb->dev to the bridge device (i.e. parent of the
500  * receiving device) to make netfilter happy, the REDIRECT
501  * target in particular.  Save the original destination IP
502  * address to be able to detect DNAT afterwards. */
br_nf_pre_routing(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)503 static unsigned int br_nf_pre_routing(void *priv,
504 				      struct sk_buff *skb,
505 				      const struct nf_hook_state *state)
506 {
507 	struct nf_bridge_info *nf_bridge;
508 	struct net_bridge_port *p;
509 	struct net_bridge *br;
510 	__u32 len = nf_bridge_encap_header_len(skb);
511 	struct brnf_net *brnet;
512 
513 	if (unlikely(!pskb_may_pull(skb, len)))
514 		return NF_DROP_REASON(skb, SKB_DROP_REASON_PKT_TOO_SMALL, 0);
515 
516 	p = br_port_get_rcu(state->in);
517 	if (p == NULL)
518 		return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
519 	br = p->br;
520 
521 	brnet = net_generic(state->net, brnf_net_id);
522 	if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) ||
523 	    is_pppoe_ipv6(skb, state->net)) {
524 		if (!brnet->call_ip6tables &&
525 		    !br_opt_get(br, BROPT_NF_CALL_IP6TABLES))
526 			return NF_ACCEPT;
527 		if (!ipv6_mod_enabled()) {
528 			pr_warn_once("Module ipv6 is disabled, so call_ip6tables is not supported.");
529 			return NF_DROP_REASON(skb, SKB_DROP_REASON_IPV6DISABLED, 0);
530 		}
531 
532 		nf_bridge_pull_encap_header_rcsum(skb);
533 		return br_nf_pre_routing_ipv6(priv, skb, state);
534 	}
535 
536 	if (!brnet->call_iptables && !br_opt_get(br, BROPT_NF_CALL_IPTABLES))
537 		return NF_ACCEPT;
538 
539 	if (!IS_IP(skb) && !is_vlan_ip(skb, state->net) &&
540 	    !is_pppoe_ip(skb, state->net))
541 		return NF_ACCEPT;
542 
543 	nf_bridge_pull_encap_header_rcsum(skb);
544 
545 	if (br_validate_ipv4(state->net, skb))
546 		return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0);
547 
548 	if (!nf_bridge_alloc(skb))
549 		return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0);
550 	if (!setup_pre_routing(skb, state->net))
551 		return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
552 
553 	nf_bridge = nf_bridge_info_get(skb);
554 	nf_bridge->ipv4_daddr = ip_hdr(skb)->daddr;
555 
556 	skb->protocol = htons(ETH_P_IP);
557 	skb->transport_header = skb->network_header + ip_hdr(skb)->ihl * 4;
558 
559 	NF_HOOK(NFPROTO_IPV4, NF_INET_PRE_ROUTING, state->net, state->sk, skb,
560 		skb->dev, NULL,
561 		br_nf_pre_routing_finish);
562 
563 	return NF_STOLEN;
564 }
565 
566 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
567 /* conntracks' nf_confirm logic cannot handle cloned skbs referencing
568  * the same nf_conn entry, which will happen for multicast (broadcast)
569  * Frames on bridges.
570  *
571  * Example:
572  *      macvlan0
573  *      br0
574  *  ethX  ethY
575  *
576  * ethX (or Y) receives multicast or broadcast packet containing
577  * an IP packet, not yet in conntrack table.
578  *
579  * 1. skb passes through bridge and fake-ip (br_netfilter)Prerouting.
580  *    -> skb->_nfct now references a unconfirmed entry
581  * 2. skb is broad/mcast packet. bridge now passes clones out on each bridge
582  *    interface.
583  * 3. skb gets passed up the stack.
584  * 4. In macvlan case, macvlan driver retains clone(s) of the mcast skb
585  *    and schedules a work queue to send them out on the lower devices.
586  *
587  *    The clone skb->_nfct is not a copy, it is the same entry as the
588  *    original skb.  The macvlan rx handler then returns RX_HANDLER_PASS.
589  * 5. Normal conntrack hooks (in NF_INET_LOCAL_IN) confirm the orig skb.
590  *
591  * The Macvlan broadcast worker and normal confirm path will race.
592  *
593  * This race will not happen if step 2 already confirmed a clone. In that
594  * case later steps perform skb_clone() with skb->_nfct already confirmed (in
595  * hash table).  This works fine.
596  *
597  * But such confirmation won't happen when eb/ip/nftables rules dropped the
598  * packets before they reached the nf_confirm step in postrouting.
599  *
600  * Work around this problem by explicit confirmation of the entry at
601  * LOCAL_IN time, before upper layer has a chance to clone the unconfirmed
602  * entry.
603  *
604  */
br_nf_local_in(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)605 static unsigned int br_nf_local_in(void *priv,
606 				   struct sk_buff *skb,
607 				   const struct nf_hook_state *state)
608 {
609 	bool promisc = BR_INPUT_SKB_CB(skb)->promisc;
610 	struct nf_conntrack *nfct = skb_nfct(skb);
611 	const struct nf_ct_hook *ct_hook;
612 	struct nf_conn *ct;
613 	int ret;
614 
615 	if (promisc) {
616 		nf_reset_ct(skb);
617 		return NF_ACCEPT;
618 	}
619 
620 	if (!nfct || skb->pkt_type == PACKET_HOST)
621 		return NF_ACCEPT;
622 
623 	ct = container_of(nfct, struct nf_conn, ct_general);
624 	if (likely(nf_ct_is_confirmed(ct)))
625 		return NF_ACCEPT;
626 
627 	if (WARN_ON_ONCE(refcount_read(&nfct->use) != 1)) {
628 		nf_reset_ct(skb);
629 		return NF_ACCEPT;
630 	}
631 
632 	WARN_ON_ONCE(skb_shared(skb));
633 
634 	/* We can't call nf_confirm here, it would create a dependency
635 	 * on nf_conntrack module.
636 	 */
637 	ct_hook = rcu_dereference(nf_ct_hook);
638 	if (!ct_hook) {
639 		skb->_nfct = 0ul;
640 		nf_conntrack_put(nfct);
641 		return NF_ACCEPT;
642 	}
643 
644 	nf_bridge_pull_encap_header(skb);
645 	ret = ct_hook->confirm(skb);
646 	switch (ret & NF_VERDICT_MASK) {
647 	case NF_STOLEN:
648 		return NF_STOLEN;
649 	default:
650 		nf_bridge_push_encap_header(skb);
651 		break;
652 	}
653 
654 	ct = container_of(nfct, struct nf_conn, ct_general);
655 	WARN_ON_ONCE(!nf_ct_is_confirmed(ct));
656 
657 	return ret;
658 }
659 #endif
660 
661 /* PF_BRIDGE/FORWARD *************************************************/
br_nf_forward_finish(struct net * net,struct sock * sk,struct sk_buff * skb)662 static int br_nf_forward_finish(struct net *net, struct sock *sk, struct sk_buff *skb)
663 {
664 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
665 	struct net_device *in;
666 
667 	if (!IS_ARP(skb) && !is_vlan_arp(skb, net)) {
668 
669 		if (skb->protocol == htons(ETH_P_IP))
670 			nf_bridge->frag_max_size = IPCB(skb)->frag_max_size;
671 
672 		if (skb->protocol == htons(ETH_P_IPV6))
673 			nf_bridge->frag_max_size = IP6CB(skb)->frag_max_size;
674 
675 		in = nf_bridge_get_physindev(skb, net);
676 		if (!in) {
677 			kfree_skb(skb);
678 			return 0;
679 		}
680 		if (nf_bridge->pkt_otherhost) {
681 			skb->pkt_type = PACKET_OTHERHOST;
682 			nf_bridge->pkt_otherhost = false;
683 		}
684 		nf_bridge_update_protocol(skb);
685 	} else {
686 		in = *((struct net_device **)(skb->cb));
687 	}
688 	nf_bridge_push_encap_header(skb);
689 
690 	br_nf_hook_thresh(NF_BR_FORWARD, net, sk, skb, in, skb->dev,
691 			  br_forward_finish);
692 	return 0;
693 }
694 
695 
br_nf_forward_ip(struct sk_buff * skb,const struct nf_hook_state * state,u8 pf)696 static unsigned int br_nf_forward_ip(struct sk_buff *skb,
697 				     const struct nf_hook_state *state,
698 				     u8 pf)
699 {
700 	struct nf_bridge_info *nf_bridge;
701 	struct net_device *parent;
702 
703 	nf_bridge = nf_bridge_info_get(skb);
704 	if (!nf_bridge)
705 		return NF_ACCEPT;
706 
707 	/* Need exclusive nf_bridge_info since we might have multiple
708 	 * different physoutdevs. */
709 	if (!nf_bridge_unshare(skb))
710 		return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0);
711 
712 	nf_bridge = nf_bridge_info_get(skb);
713 	if (!nf_bridge)
714 		return NF_DROP_REASON(skb, SKB_DROP_REASON_NOMEM, 0);
715 
716 	parent = bridge_parent(state->out);
717 	if (!parent)
718 		return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
719 
720 	nf_bridge_pull_encap_header(skb);
721 
722 	if (skb->pkt_type == PACKET_OTHERHOST) {
723 		skb->pkt_type = PACKET_HOST;
724 		nf_bridge->pkt_otherhost = true;
725 	}
726 
727 	if (pf == NFPROTO_IPV4) {
728 		if (br_validate_ipv4(state->net, skb))
729 			return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0);
730 		IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
731 		skb->protocol = htons(ETH_P_IP);
732 	} else if (pf == NFPROTO_IPV6) {
733 		if (br_validate_ipv6(state->net, skb))
734 			return NF_DROP_REASON(skb, SKB_DROP_REASON_IP_INHDR, 0);
735 		IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
736 		skb->protocol = htons(ETH_P_IPV6);
737 	} else {
738 		WARN_ON_ONCE(1);
739 		return NF_DROP;
740 	}
741 
742 	nf_bridge->physoutdev = skb->dev;
743 
744 	NF_HOOK(pf, NF_INET_FORWARD, state->net, NULL, skb,
745 		brnf_get_logical_dev(skb, state->in, state->net),
746 		parent,	br_nf_forward_finish);
747 
748 	return NF_STOLEN;
749 }
750 
br_nf_forward_arp(struct sk_buff * skb,const struct nf_hook_state * state)751 static unsigned int br_nf_forward_arp(struct sk_buff *skb,
752 				      const struct nf_hook_state *state)
753 {
754 	struct net_bridge_port *p;
755 	struct net_bridge *br;
756 	struct net_device **d = (struct net_device **)(skb->cb);
757 	struct brnf_net *brnet;
758 
759 	p = br_port_get_rcu(state->out);
760 	if (p == NULL)
761 		return NF_ACCEPT;
762 	br = p->br;
763 
764 	brnet = net_generic(state->net, brnf_net_id);
765 	if (!brnet->call_arptables && !br_opt_get(br, BROPT_NF_CALL_ARPTABLES))
766 		return NF_ACCEPT;
767 
768 	if (is_vlan_arp(skb, state->net))
769 		nf_bridge_pull_encap_header(skb);
770 
771 	if (unlikely(!pskb_may_pull(skb, sizeof(struct arphdr))))
772 		return NF_DROP_REASON(skb, SKB_DROP_REASON_PKT_TOO_SMALL, 0);
773 
774 	if (arp_hdr(skb)->ar_pln != 4) {
775 		if (is_vlan_arp(skb, state->net))
776 			nf_bridge_push_encap_header(skb);
777 		return NF_ACCEPT;
778 	}
779 	*d = state->in;
780 	NF_HOOK(NFPROTO_ARP, NF_ARP_FORWARD, state->net, state->sk, skb,
781 		state->in, state->out, br_nf_forward_finish);
782 
783 	return NF_STOLEN;
784 }
785 
786 /* This is the 'purely bridged' case.  For IP, we pass the packet to
787  * netfilter with indev and outdev set to the bridge device,
788  * but we are still able to filter on the 'real' indev/outdev
789  * because of the physdev module. For ARP, indev and outdev are the
790  * bridge ports.
791  */
br_nf_forward(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)792 static unsigned int br_nf_forward(void *priv,
793 				  struct sk_buff *skb,
794 				  const struct nf_hook_state *state)
795 {
796 	if (IS_IP(skb) || is_vlan_ip(skb, state->net) ||
797 	    is_pppoe_ip(skb, state->net))
798 		return br_nf_forward_ip(skb, state, NFPROTO_IPV4);
799 	if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) ||
800 	    is_pppoe_ipv6(skb, state->net))
801 		return br_nf_forward_ip(skb, state, NFPROTO_IPV6);
802 	if (IS_ARP(skb) || is_vlan_arp(skb, state->net))
803 		return br_nf_forward_arp(skb, state);
804 
805 	return NF_ACCEPT;
806 }
807 
br_nf_push_frag_xmit(struct net * net,struct sock * sk,struct sk_buff * skb)808 static int br_nf_push_frag_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
809 {
810 	struct brnf_frag_data *data;
811 	int err;
812 
813 	data = this_cpu_ptr(&brnf_frag_data_storage);
814 	err = skb_cow_head(skb, data->size);
815 
816 	if (err) {
817 		kfree_skb(skb);
818 		return 0;
819 	}
820 
821 	if (data->vlan_proto)
822 		__vlan_hwaccel_put_tag(skb, data->vlan_proto, data->vlan_tci);
823 
824 	skb_copy_to_linear_data_offset(skb, -data->size, data->mac, data->size);
825 	__skb_push(skb, data->encap_size);
826 
827 	nf_bridge_info_free(skb);
828 	return br_dev_queue_push_xmit(net, sk, skb);
829 }
830 
831 static int
br_nf_ip_fragment(struct net * net,struct sock * sk,struct sk_buff * skb,int (* output)(struct net *,struct sock *,struct sk_buff *))832 br_nf_ip_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
833 		  int (*output)(struct net *, struct sock *, struct sk_buff *))
834 {
835 	unsigned int mtu = ip_skb_dst_mtu(sk, skb);
836 	struct iphdr *iph = ip_hdr(skb);
837 
838 	if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
839 		     (IPCB(skb)->frag_max_size &&
840 		      IPCB(skb)->frag_max_size > mtu))) {
841 		IP_INC_STATS(net, IPSTATS_MIB_FRAGFAILS);
842 		kfree_skb(skb);
843 		return -EMSGSIZE;
844 	}
845 
846 	return ip_do_fragment(net, sk, skb, output);
847 }
848 
nf_bridge_mtu_reduction(const struct sk_buff * skb)849 static unsigned int nf_bridge_mtu_reduction(const struct sk_buff *skb)
850 {
851 	const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
852 
853 	if (nf_bridge->orig_proto == BRNF_PROTO_PPPOE)
854 		return PPPOE_SES_HLEN;
855 	return 0;
856 }
857 
br_nf_dev_queue_xmit(struct net * net,struct sock * sk,struct sk_buff * skb)858 static int br_nf_dev_queue_xmit(struct net *net, struct sock *sk, struct sk_buff *skb)
859 {
860 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
861 	unsigned int mtu, mtu_reserved;
862 	int ret;
863 
864 	mtu_reserved = nf_bridge_mtu_reduction(skb);
865 	mtu = skb->dev->mtu;
866 
867 	if (nf_bridge->pkt_otherhost) {
868 		skb->pkt_type = PACKET_OTHERHOST;
869 		nf_bridge->pkt_otherhost = false;
870 	}
871 
872 	if (nf_bridge->frag_max_size && nf_bridge->frag_max_size < mtu)
873 		mtu = nf_bridge->frag_max_size;
874 
875 	nf_bridge_update_protocol(skb);
876 	nf_bridge_push_encap_header(skb);
877 
878 	if (skb_is_gso(skb) || skb->len + mtu_reserved <= mtu) {
879 		nf_bridge_info_free(skb);
880 		return br_dev_queue_push_xmit(net, sk, skb);
881 	}
882 
883 	/* Fragmentation on metadata/template dst is not supported */
884 	if (unlikely(!skb_valid_dst(skb)))
885 		goto drop;
886 
887 	/* This is wrong! We should preserve the original fragment
888 	 * boundaries by preserving frag_list rather than refragmenting.
889 	 */
890 	if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV4) &&
891 	    skb->protocol == htons(ETH_P_IP)) {
892 		struct brnf_frag_data *data;
893 
894 		if (br_validate_ipv4(net, skb))
895 			goto drop;
896 
897 		IPCB(skb)->frag_max_size = nf_bridge->frag_max_size;
898 
899 		local_lock_nested_bh(&brnf_frag_data_storage.bh_lock);
900 		data = this_cpu_ptr(&brnf_frag_data_storage);
901 
902 		if (skb_vlan_tag_present(skb)) {
903 			data->vlan_tci = skb->vlan_tci;
904 			data->vlan_proto = skb->vlan_proto;
905 		} else {
906 			data->vlan_proto = 0;
907 		}
908 
909 		data->encap_size = nf_bridge_encap_header_len(skb);
910 		data->size = ETH_HLEN + data->encap_size;
911 
912 		skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
913 						 data->size);
914 
915 		ret = br_nf_ip_fragment(net, sk, skb, br_nf_push_frag_xmit);
916 		local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock);
917 		return ret;
918 	}
919 	if (IS_ENABLED(CONFIG_NF_DEFRAG_IPV6) &&
920 	    skb->protocol == htons(ETH_P_IPV6)) {
921 		const struct nf_ipv6_ops *v6ops = nf_get_ipv6_ops();
922 		struct brnf_frag_data *data;
923 
924 		if (br_validate_ipv6(net, skb))
925 			goto drop;
926 
927 		IP6CB(skb)->frag_max_size = nf_bridge->frag_max_size;
928 
929 		local_lock_nested_bh(&brnf_frag_data_storage.bh_lock);
930 		data = this_cpu_ptr(&brnf_frag_data_storage);
931 		data->encap_size = nf_bridge_encap_header_len(skb);
932 		data->size = ETH_HLEN + data->encap_size;
933 
934 		skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
935 						 data->size);
936 
937 		if (v6ops) {
938 			ret = v6ops->fragment(net, sk, skb, br_nf_push_frag_xmit);
939 			local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock);
940 			return ret;
941 		}
942 		local_unlock_nested_bh(&brnf_frag_data_storage.bh_lock);
943 
944 		kfree_skb(skb);
945 		return -EMSGSIZE;
946 	}
947 	nf_bridge_info_free(skb);
948 	return br_dev_queue_push_xmit(net, sk, skb);
949  drop:
950 	kfree_skb(skb);
951 	return 0;
952 }
953 
954 /* PF_BRIDGE/POST_ROUTING ********************************************/
br_nf_post_routing(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)955 static unsigned int br_nf_post_routing(void *priv,
956 				       struct sk_buff *skb,
957 				       const struct nf_hook_state *state)
958 {
959 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
960 	struct net_device *realoutdev = bridge_parent(skb->dev);
961 	u_int8_t pf;
962 
963 	/* if nf_bridge is set, but ->physoutdev is NULL, this packet came in
964 	 * on a bridge, but was delivered locally and is now being routed:
965 	 *
966 	 * POST_ROUTING was already invoked from the ip stack.
967 	 */
968 	if (!nf_bridge || !nf_bridge->physoutdev)
969 		return NF_ACCEPT;
970 
971 	if (!realoutdev)
972 		return NF_DROP_REASON(skb, SKB_DROP_REASON_DEV_READY, 0);
973 
974 	if (IS_IP(skb) || is_vlan_ip(skb, state->net) ||
975 	    is_pppoe_ip(skb, state->net))
976 		pf = NFPROTO_IPV4;
977 	else if (IS_IPV6(skb) || is_vlan_ipv6(skb, state->net) ||
978 		 is_pppoe_ipv6(skb, state->net))
979 		pf = NFPROTO_IPV6;
980 	else
981 		return NF_ACCEPT;
982 
983 	if (skb->pkt_type == PACKET_OTHERHOST) {
984 		skb->pkt_type = PACKET_HOST;
985 		nf_bridge->pkt_otherhost = true;
986 	}
987 
988 	nf_bridge_pull_encap_header(skb);
989 	if (pf == NFPROTO_IPV4)
990 		skb->protocol = htons(ETH_P_IP);
991 	else
992 		skb->protocol = htons(ETH_P_IPV6);
993 
994 	NF_HOOK(pf, NF_INET_POST_ROUTING, state->net, state->sk, skb,
995 		NULL, realoutdev,
996 		br_nf_dev_queue_xmit);
997 
998 	return NF_STOLEN;
999 }
1000 
1001 /* IP/SABOTAGE *****************************************************/
1002 /* Don't hand locally destined packets to PF_INET(6)/PRE_ROUTING
1003  * for the second time. */
ip_sabotage_in(void * priv,struct sk_buff * skb,const struct nf_hook_state * state)1004 static unsigned int ip_sabotage_in(void *priv,
1005 				   struct sk_buff *skb,
1006 				   const struct nf_hook_state *state)
1007 {
1008 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1009 
1010 	if (nf_bridge) {
1011 		if (nf_bridge->sabotage_in_done)
1012 			return NF_ACCEPT;
1013 
1014 		if (!nf_bridge->in_prerouting &&
1015 		    !netif_is_l3_master(skb->dev) &&
1016 		    !netif_is_l3_slave(skb->dev)) {
1017 			nf_bridge->sabotage_in_done = 1;
1018 			state->okfn(state->net, state->sk, skb);
1019 			return NF_STOLEN;
1020 		}
1021 	}
1022 
1023 	return NF_ACCEPT;
1024 }
1025 
1026 /* This is called when br_netfilter has called into iptables/netfilter,
1027  * and DNAT has taken place on a bridge-forwarded packet.
1028  *
1029  * neigh->output has created a new MAC header, with local br0 MAC
1030  * as saddr.
1031  *
1032  * This restores the original MAC saddr of the bridged packet
1033  * before invoking bridge forward logic to transmit the packet.
1034  */
br_nf_pre_routing_finish_bridge_slow(struct sk_buff * skb)1035 static void br_nf_pre_routing_finish_bridge_slow(struct sk_buff *skb)
1036 {
1037 	struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1038 	struct net_device *br_indev;
1039 
1040 	br_indev = nf_bridge_get_physindev(skb, dev_net(skb->dev));
1041 	if (!br_indev) {
1042 		kfree_skb(skb);
1043 		return;
1044 	}
1045 
1046 	skb_pull(skb, ETH_HLEN);
1047 	nf_bridge->bridged_dnat = 0;
1048 
1049 	BUILD_BUG_ON(sizeof(nf_bridge->neigh_header) != (ETH_HLEN - ETH_ALEN));
1050 
1051 	skb_copy_to_linear_data_offset(skb, -(ETH_HLEN - ETH_ALEN),
1052 				       nf_bridge->neigh_header,
1053 				       ETH_HLEN - ETH_ALEN);
1054 	skb->dev = br_indev;
1055 
1056 	nf_bridge->physoutdev = NULL;
1057 	br_handle_frame_finish(dev_net(skb->dev), NULL, skb);
1058 }
1059 
br_nf_dev_xmit(struct sk_buff * skb)1060 static int br_nf_dev_xmit(struct sk_buff *skb)
1061 {
1062 	const struct nf_bridge_info *nf_bridge = nf_bridge_info_get(skb);
1063 
1064 	if (nf_bridge && nf_bridge->bridged_dnat) {
1065 		br_nf_pre_routing_finish_bridge_slow(skb);
1066 		return 1;
1067 	}
1068 	return 0;
1069 }
1070 
1071 static const struct nf_br_ops br_ops = {
1072 	.br_dev_xmit_hook =	br_nf_dev_xmit,
1073 };
1074 
1075 /* For br_nf_post_routing, we need (prio = NF_BR_PRI_LAST), because
1076  * br_dev_queue_push_xmit is called afterwards */
1077 static const struct nf_hook_ops br_nf_ops[] = {
1078 	{
1079 		.hook = br_nf_pre_routing,
1080 		.pf = NFPROTO_BRIDGE,
1081 		.hooknum = NF_BR_PRE_ROUTING,
1082 		.priority = NF_BR_PRI_BRNF,
1083 	},
1084 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
1085 	{
1086 		.hook = br_nf_local_in,
1087 		.pf = NFPROTO_BRIDGE,
1088 		.hooknum = NF_BR_LOCAL_IN,
1089 		.priority = NF_BR_PRI_LAST,
1090 	},
1091 #endif
1092 	{
1093 		.hook = br_nf_forward,
1094 		.pf = NFPROTO_BRIDGE,
1095 		.hooknum = NF_BR_FORWARD,
1096 		.priority = NF_BR_PRI_BRNF,
1097 	},
1098 	{
1099 		.hook = br_nf_post_routing,
1100 		.pf = NFPROTO_BRIDGE,
1101 		.hooknum = NF_BR_POST_ROUTING,
1102 		.priority = NF_BR_PRI_LAST,
1103 	},
1104 	{
1105 		.hook = ip_sabotage_in,
1106 		.pf = NFPROTO_IPV4,
1107 		.hooknum = NF_INET_PRE_ROUTING,
1108 		.priority = NF_IP_PRI_FIRST,
1109 	},
1110 	{
1111 		.hook = ip_sabotage_in,
1112 		.pf = NFPROTO_IPV6,
1113 		.hooknum = NF_INET_PRE_ROUTING,
1114 		.priority = NF_IP6_PRI_FIRST,
1115 	},
1116 };
1117 
brnf_device_event(struct notifier_block * unused,unsigned long event,void * ptr)1118 static int brnf_device_event(struct notifier_block *unused, unsigned long event,
1119 			     void *ptr)
1120 {
1121 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1122 	struct brnf_net *brnet;
1123 	struct net *net;
1124 	int ret;
1125 
1126 	if (event != NETDEV_REGISTER || !netif_is_bridge_master(dev))
1127 		return NOTIFY_DONE;
1128 
1129 	ASSERT_RTNL();
1130 
1131 	net = dev_net(dev);
1132 	brnet = net_generic(net, brnf_net_id);
1133 	if (brnet->enabled)
1134 		return NOTIFY_OK;
1135 
1136 	ret = nf_register_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
1137 	if (ret)
1138 		return NOTIFY_BAD;
1139 
1140 	brnet->enabled = true;
1141 	return NOTIFY_OK;
1142 }
1143 
1144 static struct notifier_block brnf_notifier __read_mostly = {
1145 	.notifier_call = brnf_device_event,
1146 };
1147 
1148 /* recursively invokes nf_hook_slow (again), skipping already-called
1149  * hooks (< NF_BR_PRI_BRNF).
1150  *
1151  * Called with rcu read lock held.
1152  */
br_nf_hook_thresh(unsigned int hook,struct net * net,struct sock * sk,struct sk_buff * skb,struct net_device * indev,struct net_device * outdev,int (* okfn)(struct net *,struct sock *,struct sk_buff *))1153 int br_nf_hook_thresh(unsigned int hook, struct net *net,
1154 		      struct sock *sk, struct sk_buff *skb,
1155 		      struct net_device *indev,
1156 		      struct net_device *outdev,
1157 		      int (*okfn)(struct net *, struct sock *,
1158 				  struct sk_buff *))
1159 {
1160 	const struct nf_hook_entries *e;
1161 	struct nf_hook_state state;
1162 	struct nf_hook_ops **ops;
1163 	unsigned int i;
1164 	int ret;
1165 
1166 	e = rcu_dereference(net->nf.hooks_bridge[hook]);
1167 	if (!e)
1168 		return okfn(net, sk, skb);
1169 
1170 	ops = nf_hook_entries_get_hook_ops(e);
1171 	for (i = 0; i < e->num_hook_entries; i++) {
1172 		/* These hooks have already been called */
1173 		if (ops[i]->priority < NF_BR_PRI_BRNF)
1174 			continue;
1175 
1176 		/* These hooks have not been called yet, run them. */
1177 		if (ops[i]->priority > NF_BR_PRI_BRNF)
1178 			break;
1179 
1180 		/* take a closer look at NF_BR_PRI_BRNF. */
1181 		if (ops[i]->hook == br_nf_pre_routing) {
1182 			/* This hook diverted the skb to this function,
1183 			 * hooks after this have not been run yet.
1184 			 */
1185 			i++;
1186 			break;
1187 		}
1188 	}
1189 
1190 	nf_hook_state_init(&state, hook, NFPROTO_BRIDGE, indev, outdev,
1191 			   sk, net, okfn);
1192 
1193 	ret = nf_hook_slow(skb, &state, e, i);
1194 	if (ret == 1)
1195 		ret = okfn(net, sk, skb);
1196 
1197 	return ret;
1198 }
1199 
1200 #ifdef CONFIG_SYSCTL
1201 static
brnf_sysctl_call_tables(const struct ctl_table * ctl,int write,void * buffer,size_t * lenp,loff_t * ppos)1202 int brnf_sysctl_call_tables(const struct ctl_table *ctl, int write,
1203 			    void *buffer, size_t *lenp, loff_t *ppos)
1204 {
1205 	int ret;
1206 
1207 	ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
1208 
1209 	if (write && *(int *)(ctl->data))
1210 		*(int *)(ctl->data) = 1;
1211 	return ret;
1212 }
1213 
1214 static struct ctl_table brnf_table[] = {
1215 	{
1216 		.procname	= "bridge-nf-call-arptables",
1217 		.maxlen		= sizeof(int),
1218 		.mode		= 0644,
1219 		.proc_handler	= brnf_sysctl_call_tables,
1220 	},
1221 	{
1222 		.procname	= "bridge-nf-call-iptables",
1223 		.maxlen		= sizeof(int),
1224 		.mode		= 0644,
1225 		.proc_handler	= brnf_sysctl_call_tables,
1226 	},
1227 	{
1228 		.procname	= "bridge-nf-call-ip6tables",
1229 		.maxlen		= sizeof(int),
1230 		.mode		= 0644,
1231 		.proc_handler	= brnf_sysctl_call_tables,
1232 	},
1233 	{
1234 		.procname	= "bridge-nf-filter-vlan-tagged",
1235 		.maxlen		= sizeof(int),
1236 		.mode		= 0644,
1237 		.proc_handler	= brnf_sysctl_call_tables,
1238 	},
1239 	{
1240 		.procname	= "bridge-nf-filter-pppoe-tagged",
1241 		.maxlen		= sizeof(int),
1242 		.mode		= 0644,
1243 		.proc_handler	= brnf_sysctl_call_tables,
1244 	},
1245 	{
1246 		.procname	= "bridge-nf-pass-vlan-input-dev",
1247 		.maxlen		= sizeof(int),
1248 		.mode		= 0644,
1249 		.proc_handler	= brnf_sysctl_call_tables,
1250 	},
1251 };
1252 
br_netfilter_sysctl_default(struct brnf_net * brnf)1253 static inline void br_netfilter_sysctl_default(struct brnf_net *brnf)
1254 {
1255 	brnf->call_iptables = 1;
1256 	brnf->call_ip6tables = 1;
1257 	brnf->call_arptables = 1;
1258 	brnf->filter_vlan_tagged = 0;
1259 	brnf->filter_pppoe_tagged = 0;
1260 	brnf->pass_vlan_indev = 0;
1261 }
1262 
br_netfilter_sysctl_init_net(struct net * net)1263 static int br_netfilter_sysctl_init_net(struct net *net)
1264 {
1265 	struct ctl_table *table = brnf_table;
1266 	struct brnf_net *brnet;
1267 
1268 	if (!net_eq(net, &init_net)) {
1269 		table = kmemdup(table, sizeof(brnf_table), GFP_KERNEL);
1270 		if (!table)
1271 			return -ENOMEM;
1272 	}
1273 
1274 	brnet = net_generic(net, brnf_net_id);
1275 	table[0].data = &brnet->call_arptables;
1276 	table[1].data = &brnet->call_iptables;
1277 	table[2].data = &brnet->call_ip6tables;
1278 	table[3].data = &brnet->filter_vlan_tagged;
1279 	table[4].data = &brnet->filter_pppoe_tagged;
1280 	table[5].data = &brnet->pass_vlan_indev;
1281 
1282 	br_netfilter_sysctl_default(brnet);
1283 
1284 	brnet->ctl_hdr = register_net_sysctl_sz(net, "net/bridge", table,
1285 						ARRAY_SIZE(brnf_table));
1286 	if (!brnet->ctl_hdr) {
1287 		if (!net_eq(net, &init_net))
1288 			kfree(table);
1289 
1290 		return -ENOMEM;
1291 	}
1292 
1293 	return 0;
1294 }
1295 
br_netfilter_sysctl_exit_net(struct net * net,struct brnf_net * brnet)1296 static void br_netfilter_sysctl_exit_net(struct net *net,
1297 					 struct brnf_net *brnet)
1298 {
1299 	const struct ctl_table *table = brnet->ctl_hdr->ctl_table_arg;
1300 
1301 	unregister_net_sysctl_table(brnet->ctl_hdr);
1302 	if (!net_eq(net, &init_net))
1303 		kfree(table);
1304 }
1305 
brnf_init_net(struct net * net)1306 static int __net_init brnf_init_net(struct net *net)
1307 {
1308 	return br_netfilter_sysctl_init_net(net);
1309 }
1310 #endif
1311 
brnf_exit_net(struct net * net)1312 static void __net_exit brnf_exit_net(struct net *net)
1313 {
1314 	struct brnf_net *brnet;
1315 
1316 	brnet = net_generic(net, brnf_net_id);
1317 	if (brnet->enabled) {
1318 		nf_unregister_net_hooks(net, br_nf_ops, ARRAY_SIZE(br_nf_ops));
1319 		brnet->enabled = false;
1320 	}
1321 
1322 #ifdef CONFIG_SYSCTL
1323 	br_netfilter_sysctl_exit_net(net, brnet);
1324 #endif
1325 }
1326 
1327 static struct pernet_operations brnf_net_ops __read_mostly = {
1328 #ifdef CONFIG_SYSCTL
1329 	.init = brnf_init_net,
1330 #endif
1331 	.exit = brnf_exit_net,
1332 	.id   = &brnf_net_id,
1333 	.size = sizeof(struct brnf_net),
1334 };
1335 
br_netfilter_init(void)1336 static int __init br_netfilter_init(void)
1337 {
1338 	int ret;
1339 
1340 	ret = register_pernet_subsys(&brnf_net_ops);
1341 	if (ret < 0)
1342 		return ret;
1343 
1344 	ret = register_netdevice_notifier(&brnf_notifier);
1345 	if (ret < 0) {
1346 		unregister_pernet_subsys(&brnf_net_ops);
1347 		return ret;
1348 	}
1349 
1350 	RCU_INIT_POINTER(nf_br_ops, &br_ops);
1351 	printk(KERN_NOTICE "Bridge firewalling registered\n");
1352 	return 0;
1353 }
1354 
br_netfilter_fini(void)1355 static void __exit br_netfilter_fini(void)
1356 {
1357 	RCU_INIT_POINTER(nf_br_ops, NULL);
1358 	unregister_netdevice_notifier(&brnf_notifier);
1359 	unregister_pernet_subsys(&brnf_net_ops);
1360 }
1361 
1362 module_init(br_netfilter_init);
1363 module_exit(br_netfilter_fini);
1364 
1365 MODULE_LICENSE("GPL");
1366 MODULE_AUTHOR("Lennert Buytenhek <buytenh@gnu.org>");
1367 MODULE_AUTHOR("Bart De Schuymer <bdschuym@pandora.be>");
1368 MODULE_DESCRIPTION("Linux ethernet netfilter firewall bridge");
1369