xref: /linux/net/bridge/br_arp_nd_proxy.c (revision 6a20b34fe3b31b292078bc79ec18a2ab0d9f7719)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  *  Handle bridge arp/nd proxy/suppress
4  *
5  *  Copyright (C) 2017 Cumulus Networks
6  *  Copyright (c) 2017 Roopa Prabhu <roopa@cumulusnetworks.com>
7  *
8  *  Authors:
9  *	Roopa Prabhu <roopa@cumulusnetworks.com>
10  */
11 
12 #include <linux/kernel.h>
13 #include <linux/netdevice.h>
14 #include <linux/etherdevice.h>
15 #include <linux/neighbour.h>
16 #include <net/arp.h>
17 #include <linux/if_vlan.h>
18 #include <linux/inetdevice.h>
19 #include <net/addrconf.h>
20 #if IS_ENABLED(CONFIG_IPV6)
21 #include <net/ip6_checksum.h>
22 #endif
23 
24 #include "br_private.h"
25 
26 void br_recalculate_neigh_suppress_enabled(struct net_bridge *br)
27 {
28 	struct net_bridge_port *p;
29 	bool neigh_suppress = false;
30 
31 	list_for_each_entry(p, &br->port_list, list) {
32 		if (p->flags & (BR_NEIGH_SUPPRESS | BR_NEIGH_VLAN_SUPPRESS)) {
33 			neigh_suppress = true;
34 			break;
35 		}
36 	}
37 
38 	br_opt_toggle(br, BROPT_NEIGH_SUPPRESS_ENABLED, neigh_suppress);
39 }
40 
41 #if IS_ENABLED(CONFIG_INET)
42 static void br_arp_send(struct net_bridge *br, struct net_bridge_port *p,
43 			struct net_device *dev, __be32 dest_ip, __be32 src_ip,
44 			const unsigned char *dest_hw,
45 			const unsigned char *src_hw,
46 			const unsigned char *target_hw,
47 			__be16 vlan_proto, u16 vlan_tci)
48 {
49 	struct net_bridge_vlan_group *vg;
50 	struct sk_buff *skb;
51 	u16 pvid;
52 
53 	netdev_dbg(dev, "arp send dev %s dst %pI4 dst_hw %pM src %pI4 src_hw %pM\n",
54 		   dev->name, &dest_ip, dest_hw, &src_ip, src_hw);
55 
56 	if (!vlan_tci) {
57 		arp_send(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
58 			 dest_hw, src_hw, target_hw);
59 		return;
60 	}
61 
62 	skb = arp_create(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
63 			 dest_hw, src_hw, target_hw);
64 	if (!skb)
65 		return;
66 
67 	if (p)
68 		vg = nbp_vlan_group_rcu(p);
69 	else
70 		vg = br_vlan_group_rcu(br);
71 	pvid = br_get_pvid(vg);
72 	if (pvid == (vlan_tci & VLAN_VID_MASK))
73 		vlan_tci = 0;
74 
75 	if (vlan_tci)
76 		__vlan_hwaccel_put_tag(skb, vlan_proto, vlan_tci);
77 
78 	if (p) {
79 		arp_xmit(skb);
80 	} else {
81 		skb_reset_mac_header(skb);
82 		__skb_pull(skb, skb_network_offset(skb));
83 		skb->ip_summed = CHECKSUM_UNNECESSARY;
84 		skb->pkt_type = PACKET_HOST;
85 
86 		netif_rx(skb);
87 	}
88 }
89 
90 static int br_chk_addr_ip(struct net_device *dev,
91 			  struct netdev_nested_priv *priv)
92 {
93 	__be32 ip = *(__be32 *)priv->data;
94 	struct in_device *in_dev;
95 	__be32 addr = 0;
96 
97 	in_dev = __in_dev_get_rcu(dev);
98 	if (in_dev)
99 		addr = inet_confirm_addr(dev_net(dev), in_dev, 0, ip,
100 					 RT_SCOPE_HOST);
101 
102 	if (addr == ip)
103 		return 1;
104 
105 	return 0;
106 }
107 
108 static bool br_is_local_ip(struct net_device *dev, __be32 ip)
109 {
110 	struct netdev_nested_priv priv = {
111 		.data = (void *)&ip,
112 	};
113 
114 	if (br_chk_addr_ip(dev, &priv))
115 		return true;
116 
117 	/* check if ip is configured on upper dev */
118 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip, &priv))
119 		return true;
120 
121 	return false;
122 }
123 
124 void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br,
125 			      u16 vid, struct net_bridge_port *p)
126 {
127 	struct net_device *dev = br->dev;
128 	struct net_device *vlandev = dev;
129 	struct neighbour *n;
130 	struct arphdr *parp;
131 	u8 *arpptr, *sha;
132 	__be32 sip, tip;
133 
134 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
135 	BR_INPUT_SKB_CB(skb)->grat_arp = 0;
136 
137 	if ((dev->flags & IFF_NOARP) ||
138 	    !pskb_may_pull(skb, arp_hdr_len(dev)))
139 		return;
140 
141 	parp = arp_hdr(skb);
142 
143 	if (parp->ar_pro != htons(ETH_P_IP) ||
144 	    parp->ar_hln != dev->addr_len ||
145 	    parp->ar_pln != 4)
146 		return;
147 
148 	arpptr = (u8 *)parp + sizeof(struct arphdr);
149 	sha = arpptr;
150 	arpptr += dev->addr_len;	/* sha */
151 	memcpy(&sip, arpptr, sizeof(sip));
152 	arpptr += sizeof(sip);
153 	arpptr += dev->addr_len;	/* tha */
154 	memcpy(&tip, arpptr, sizeof(tip));
155 
156 	if (ipv4_is_loopback(tip) ||
157 	    ipv4_is_multicast(tip))
158 		return;
159 
160 	if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
161 		if (br_is_neigh_suppress_enabled(p, vid))
162 			return;
163 		if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
164 		    parp->ar_op == htons(ARPOP_REQUEST))
165 			return;
166 		if (parp->ar_op != htons(ARPOP_RREQUEST) &&
167 		    parp->ar_op != htons(ARPOP_RREPLY) &&
168 		    sip == tip) {
169 			/* prevent flooding to neigh suppress ports */
170 			BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
171 			BR_INPUT_SKB_CB(skb)->grat_arp = 1;
172 			return;
173 		}
174 	}
175 
176 	if (parp->ar_op != htons(ARPOP_REQUEST))
177 		return;
178 
179 	if (vid != 0) {
180 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
181 						   vid);
182 		if (!vlandev)
183 			return;
184 	}
185 
186 	if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
187 	    br_is_local_ip(vlandev, tip)) {
188 		/* its our local ip, so don't proxy reply
189 		 * and don't forward to neigh suppress ports
190 		 */
191 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
192 		return;
193 	}
194 
195 	n = neigh_lookup(&arp_tbl, &tip, vlandev);
196 	if (n) {
197 		struct net_bridge_fdb_entry *f;
198 
199 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
200 			neigh_release(n);
201 			return;
202 		}
203 
204 		f = br_fdb_find_rcu(br, n->ha, vid);
205 		if (f) {
206 			const struct net_bridge_port *dst = READ_ONCE(f->dst);
207 			bool replied = false;
208 
209 			if ((p && (p->flags & BR_PROXYARP)) ||
210 			    (dst && (dst->flags & BR_PROXYARP_WIFI)) ||
211 			    br_is_neigh_suppress_enabled(dst, vid)) {
212 				if (!vid)
213 					br_arp_send(br, p, skb->dev, sip, tip,
214 						    sha, n->ha, sha, 0, 0);
215 				else
216 					br_arp_send(br, p, skb->dev, sip, tip,
217 						    sha, n->ha, sha,
218 						    skb->vlan_proto,
219 						    skb_vlan_tag_get(skb));
220 				replied = true;
221 			}
222 
223 			/* If we have replied or as long as we know the
224 			 * mac, indicate to arp replied
225 			 */
226 			if (replied ||
227 			    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
228 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
229 		}
230 
231 		neigh_release(n);
232 	}
233 }
234 #endif
235 
236 #if IS_ENABLED(CONFIG_IPV6)
237 struct nd_msg *br_is_nd_neigh_msg(const struct sk_buff *skb, struct nd_msg *msg)
238 {
239 	struct nd_msg *m;
240 
241 	m = skb_header_pointer(skb, skb_network_offset(skb) +
242 			       sizeof(struct ipv6hdr), sizeof(*msg), msg);
243 	if (!m)
244 		return NULL;
245 
246 	if (m->icmph.icmp6_code != 0 ||
247 	    (m->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION &&
248 	     m->icmph.icmp6_type != NDISC_NEIGHBOUR_ADVERTISEMENT))
249 		return NULL;
250 
251 	return m;
252 }
253 
254 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
255 		       struct sk_buff *request, struct neighbour *n,
256 		       __be16 vlan_proto, u16 vlan_tci)
257 {
258 	struct net_device *dev = request->dev;
259 	struct net_bridge_vlan_group *vg;
260 	struct nd_msg *na, *ns;
261 	struct sk_buff *reply;
262 	struct ipv6hdr *pip6;
263 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
264 	int ns_olen;
265 	int i, len;
266 	u8 *daddr;
267 	bool dad;
268 	u16 pvid;
269 
270 	if (!dev || skb_linearize(request))
271 		return;
272 
273 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
274 		sizeof(*na) + na_olen + dev->needed_tailroom;
275 
276 	reply = alloc_skb(len, GFP_ATOMIC);
277 	if (!reply)
278 		return;
279 
280 	reply->protocol = htons(ETH_P_IPV6);
281 	reply->dev = dev;
282 	skb_reserve(reply, LL_RESERVED_SPACE(dev));
283 	skb_push(reply, sizeof(struct ethhdr));
284 	skb_set_mac_header(reply, 0);
285 
286 	daddr = eth_hdr(request)->h_source;
287 	ns = (struct nd_msg *)(skb_network_header(request) +
288 			       sizeof(struct ipv6hdr));
289 
290 	/* Do we need option processing ? */
291 	ns_olen = request->len - (skb_network_offset(request) +
292 				  sizeof(struct ipv6hdr)) - sizeof(*ns);
293 	for (i = 0; i < ns_olen - 1; i += (ns->opt[i + 1] << 3)) {
294 		if (!ns->opt[i + 1] || i + (ns->opt[i + 1] << 3) > ns_olen) {
295 			kfree_skb(reply);
296 			return;
297 		}
298 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
299 			if ((ns->opt[i + 1] << 3) >=
300 			    sizeof(struct nd_opt_hdr) + ETH_ALEN)
301 				daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
302 			break;
303 		}
304 	}
305 
306 	dad = ipv6_addr_any(&ipv6_hdr(request)->saddr);
307 
308 	/* Ethernet header */
309 	if (dad)
310 		ipv6_eth_mc_map(&in6addr_linklocal_allnodes, eth_hdr(reply)->h_dest);
311 	else
312 		ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
313 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
314 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
315 	reply->protocol = htons(ETH_P_IPV6);
316 
317 	skb_pull(reply, sizeof(struct ethhdr));
318 	skb_set_network_header(reply, 0);
319 	skb_put(reply, sizeof(struct ipv6hdr));
320 
321 	/* IPv6 header */
322 	pip6 = ipv6_hdr(reply);
323 	memset(pip6, 0, sizeof(struct ipv6hdr));
324 	pip6->version = 6;
325 	pip6->priority = ipv6_hdr(request)->priority;
326 	pip6->nexthdr = IPPROTO_ICMPV6;
327 	pip6->hop_limit = 255;
328 	pip6->daddr = dad ? in6addr_linklocal_allnodes : ipv6_hdr(request)->saddr;
329 	pip6->saddr = *(struct in6_addr *)n->primary_key;
330 
331 	skb_pull(reply, sizeof(struct ipv6hdr));
332 	skb_set_transport_header(reply, 0);
333 
334 	na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
335 
336 	/* Neighbor Advertisement */
337 	memset(na, 0, sizeof(*na) + na_olen);
338 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
339 	na->icmph.icmp6_router = (n->flags & NTF_ROUTER) ? 1 : 0;
340 	na->icmph.icmp6_override = 1;
341 	na->icmph.icmp6_solicited = dad ? 0 : 1;
342 	na->target = ns->target;
343 	ether_addr_copy(&na->opt[2], n->ha);
344 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
345 	na->opt[1] = na_olen >> 3;
346 
347 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
348 						&pip6->daddr,
349 						sizeof(*na) + na_olen,
350 						IPPROTO_ICMPV6,
351 						csum_partial(na, sizeof(*na) + na_olen, 0));
352 
353 	pip6->payload_len = htons(sizeof(*na) + na_olen);
354 
355 	skb_push(reply, sizeof(struct ipv6hdr));
356 	skb_push(reply, sizeof(struct ethhdr));
357 
358 	reply->ip_summed = CHECKSUM_UNNECESSARY;
359 
360 	if (p)
361 		vg = nbp_vlan_group_rcu(p);
362 	else
363 		vg = br_vlan_group_rcu(br);
364 	pvid = br_get_pvid(vg);
365 	if (pvid == (vlan_tci & VLAN_VID_MASK))
366 		vlan_tci = 0;
367 
368 	if (vlan_tci)
369 		__vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
370 
371 	netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
372 		   dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha);
373 
374 	if (p) {
375 		dev_queue_xmit(reply);
376 	} else {
377 		skb_reset_mac_header(reply);
378 		__skb_pull(reply, skb_network_offset(reply));
379 		reply->ip_summed = CHECKSUM_UNNECESSARY;
380 		reply->pkt_type = PACKET_HOST;
381 
382 		netif_rx(reply);
383 	}
384 }
385 
386 static int br_chk_addr_ip6(struct net_device *dev,
387 			   struct netdev_nested_priv *priv)
388 {
389 	struct in6_addr *addr = (struct in6_addr *)priv->data;
390 
391 	if (ipv6_chk_addr(dev_net(dev), addr, dev, 0))
392 		return 1;
393 
394 	return 0;
395 }
396 
397 static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr)
398 
399 {
400 	struct netdev_nested_priv priv = {
401 		.data = (void *)addr,
402 	};
403 
404 	if (br_chk_addr_ip6(dev, &priv))
405 		return true;
406 
407 	/* check if ip is configured on upper dev */
408 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, &priv))
409 		return true;
410 
411 	return false;
412 }
413 
414 void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
415 		       u16 vid, struct net_bridge_port *p, struct nd_msg *msg)
416 {
417 	struct net_device *dev = br->dev;
418 	struct net_device *vlandev = NULL;
419 	struct in6_addr *saddr, *daddr;
420 	struct ipv6hdr *iphdr;
421 	struct neighbour *n;
422 
423 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
424 	BR_INPUT_SKB_CB(skb)->grat_arp = 0;
425 
426 	if (br_is_neigh_suppress_enabled(p, vid))
427 		return;
428 
429 	if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
430 	    msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
431 		return;
432 
433 	if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT &&
434 	    !msg->icmph.icmp6_solicited) {
435 		/* prevent flooding to neigh suppress ports */
436 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
437 		BR_INPUT_SKB_CB(skb)->grat_arp = 1;
438 		return;
439 	}
440 
441 	if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
442 		return;
443 
444 	iphdr = ipv6_hdr(skb);
445 	saddr = &iphdr->saddr;
446 	daddr = &iphdr->daddr;
447 
448 	if (!ipv6_addr_cmp(saddr, daddr)) {
449 		/* prevent flooding to neigh suppress ports */
450 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
451 		return;
452 	}
453 
454 	if (vid != 0) {
455 		/* build neigh table lookup on the vlan device */
456 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
457 						   vid);
458 		if (!vlandev)
459 			return;
460 	} else {
461 		vlandev = dev;
462 	}
463 
464 	if (br_is_local_ip6(vlandev, &msg->target)) {
465 		/* its our own ip, so don't proxy reply
466 		 * and don't forward to arp suppress ports
467 		 */
468 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
469 		return;
470 	}
471 
472 	n = neigh_lookup(&nd_tbl, &msg->target, vlandev);
473 	if (n) {
474 		struct net_bridge_fdb_entry *f;
475 
476 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
477 			neigh_release(n);
478 			return;
479 		}
480 
481 		f = br_fdb_find_rcu(br, n->ha, vid);
482 		if (f) {
483 			const struct net_bridge_port *dst = READ_ONCE(f->dst);
484 			bool replied = false;
485 
486 			if (br_is_neigh_suppress_enabled(dst, vid)) {
487 				if (vid != 0)
488 					br_nd_send(br, p, skb, n,
489 						   skb->vlan_proto,
490 						   skb_vlan_tag_get(skb));
491 				else
492 					br_nd_send(br, p, skb, n, 0, 0);
493 				replied = true;
494 			}
495 
496 			/* If we have replied or as long as we know the
497 			 * mac, indicate to NEIGH_SUPPRESS ports that we
498 			 * have replied
499 			 */
500 			if (replied ||
501 			    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
502 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
503 		}
504 		neigh_release(n);
505 	}
506 }
507 #endif
508 
509 bool br_is_neigh_suppress_enabled(const struct net_bridge_port *p, u16 vid)
510 {
511 	if (!p)
512 		return false;
513 
514 	if (!vid)
515 		return !!(p->flags & BR_NEIGH_SUPPRESS);
516 
517 	if (p->flags & BR_NEIGH_VLAN_SUPPRESS) {
518 		struct net_bridge_vlan_group *vg = nbp_vlan_group_rcu(p);
519 		struct net_bridge_vlan *v;
520 
521 		v = br_vlan_find(vg, vid);
522 		if (!v)
523 			return false;
524 		return !!(v->priv_flags & BR_VLFLAG_NEIGH_SUPPRESS_ENABLED);
525 	} else {
526 		return !!(p->flags & BR_NEIGH_SUPPRESS);
527 	}
528 }
529 
530 bool br_is_neigh_forward_grat_enabled(const struct net_bridge_port *p, u16 vid)
531 {
532 	if (!vid)
533 		return !!(p->flags & BR_NEIGH_FORWARD_GRAT);
534 
535 	if (p->flags & BR_NEIGH_VLAN_SUPPRESS) {
536 		struct net_bridge_vlan_group *vg = nbp_vlan_group_rcu(p);
537 		struct net_bridge_vlan *v;
538 
539 		v = br_vlan_find(vg, vid);
540 		if (!v)
541 			return false;
542 		return !!(v->priv_flags & BR_VLFLAG_NEIGH_FORWARD_GRAT_ENABLED);
543 	} else {
544 		return !!(p->flags & BR_NEIGH_FORWARD_GRAT);
545 	}
546 }
547