xref: /linux/net/bridge/br_arp_nd_proxy.c (revision 6fa6b5cb60490db2591bb93872b95f72315e5f53)
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 #include <net/ipv6_stubs.h>
21 #if IS_ENABLED(CONFIG_IPV6)
22 #include <net/ip6_checksum.h>
23 #endif
24 
25 #include "br_private.h"
26 
27 void br_recalculate_neigh_suppress_enabled(struct net_bridge *br)
28 {
29 	struct net_bridge_port *p;
30 	bool neigh_suppress = false;
31 
32 	list_for_each_entry(p, &br->port_list, list) {
33 		if (p->flags & (BR_NEIGH_SUPPRESS | BR_NEIGH_VLAN_SUPPRESS)) {
34 			neigh_suppress = true;
35 			break;
36 		}
37 	}
38 
39 	br_opt_toggle(br, BROPT_NEIGH_SUPPRESS_ENABLED, neigh_suppress);
40 }
41 
42 #if IS_ENABLED(CONFIG_INET)
43 static void br_arp_send(struct net_bridge *br, struct net_bridge_port *p,
44 			struct net_device *dev, __be32 dest_ip, __be32 src_ip,
45 			const unsigned char *dest_hw,
46 			const unsigned char *src_hw,
47 			const unsigned char *target_hw,
48 			__be16 vlan_proto, u16 vlan_tci)
49 {
50 	struct net_bridge_vlan_group *vg;
51 	struct sk_buff *skb;
52 	u16 pvid;
53 
54 	netdev_dbg(dev, "arp send dev %s dst %pI4 dst_hw %pM src %pI4 src_hw %pM\n",
55 		   dev->name, &dest_ip, dest_hw, &src_ip, src_hw);
56 
57 	if (!vlan_tci) {
58 		arp_send(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
59 			 dest_hw, src_hw, target_hw);
60 		return;
61 	}
62 
63 	skb = arp_create(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
64 			 dest_hw, src_hw, target_hw);
65 	if (!skb)
66 		return;
67 
68 	if (p)
69 		vg = nbp_vlan_group_rcu(p);
70 	else
71 		vg = br_vlan_group_rcu(br);
72 	pvid = br_get_pvid(vg);
73 	if (pvid == (vlan_tci & VLAN_VID_MASK))
74 		vlan_tci = 0;
75 
76 	if (vlan_tci)
77 		__vlan_hwaccel_put_tag(skb, vlan_proto, vlan_tci);
78 
79 	if (p) {
80 		arp_xmit(skb);
81 	} else {
82 		skb_reset_mac_header(skb);
83 		__skb_pull(skb, skb_network_offset(skb));
84 		skb->ip_summed = CHECKSUM_UNNECESSARY;
85 		skb->pkt_type = PACKET_HOST;
86 
87 		netif_rx(skb);
88 	}
89 }
90 
91 static int br_chk_addr_ip(struct net_device *dev,
92 			  struct netdev_nested_priv *priv)
93 {
94 	__be32 ip = *(__be32 *)priv->data;
95 	struct in_device *in_dev;
96 	__be32 addr = 0;
97 
98 	in_dev = __in_dev_get_rcu(dev);
99 	if (in_dev)
100 		addr = inet_confirm_addr(dev_net(dev), in_dev, 0, ip,
101 					 RT_SCOPE_HOST);
102 
103 	if (addr == ip)
104 		return 1;
105 
106 	return 0;
107 }
108 
109 static bool br_is_local_ip(struct net_device *dev, __be32 ip)
110 {
111 	struct netdev_nested_priv priv = {
112 		.data = (void *)&ip,
113 	};
114 
115 	if (br_chk_addr_ip(dev, &priv))
116 		return true;
117 
118 	/* check if ip is configured on upper dev */
119 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip, &priv))
120 		return true;
121 
122 	return false;
123 }
124 
125 void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br,
126 			      u16 vid, struct net_bridge_port *p)
127 {
128 	struct net_device *dev = br->dev;
129 	struct net_device *vlandev = dev;
130 	struct neighbour *n;
131 	struct arphdr *parp;
132 	u8 *arpptr, *sha;
133 	__be32 sip, tip;
134 
135 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = 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 		    (ipv4_is_zeronet(sip) || sip == tip)) {
169 			/* prevent flooding to neigh suppress ports */
170 			BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
171 			return;
172 		}
173 	}
174 
175 	if (parp->ar_op != htons(ARPOP_REQUEST))
176 		return;
177 
178 	if (vid != 0) {
179 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
180 						   vid);
181 		if (!vlandev)
182 			return;
183 	}
184 
185 	if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
186 	    br_is_local_ip(vlandev, tip)) {
187 		/* its our local ip, so don't proxy reply
188 		 * and don't forward to neigh suppress ports
189 		 */
190 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
191 		return;
192 	}
193 
194 	n = neigh_lookup(&arp_tbl, &tip, vlandev);
195 	if (n) {
196 		struct net_bridge_fdb_entry *f;
197 
198 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
199 			neigh_release(n);
200 			return;
201 		}
202 
203 		f = br_fdb_find_rcu(br, n->ha, vid);
204 		if (f) {
205 			bool replied = false;
206 
207 			if ((p && (p->flags & BR_PROXYARP)) ||
208 			    (f->dst && (f->dst->flags & BR_PROXYARP_WIFI)) ||
209 			    br_is_neigh_suppress_enabled(f->dst, vid)) {
210 				if (!vid)
211 					br_arp_send(br, p, skb->dev, sip, tip,
212 						    sha, n->ha, sha, 0, 0);
213 				else
214 					br_arp_send(br, p, skb->dev, sip, tip,
215 						    sha, n->ha, sha,
216 						    skb->vlan_proto,
217 						    skb_vlan_tag_get(skb));
218 				replied = true;
219 			}
220 
221 			/* If we have replied or as long as we know the
222 			 * mac, indicate to arp replied
223 			 */
224 			if (replied ||
225 			    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
226 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
227 		}
228 
229 		neigh_release(n);
230 	}
231 }
232 #endif
233 
234 #if IS_ENABLED(CONFIG_IPV6)
235 struct nd_msg *br_is_nd_neigh_msg(const struct sk_buff *skb, struct nd_msg *msg)
236 {
237 	struct nd_msg *m;
238 
239 	m = skb_header_pointer(skb, skb_network_offset(skb) +
240 			       sizeof(struct ipv6hdr), sizeof(*msg), msg);
241 	if (!m)
242 		return NULL;
243 
244 	if (m->icmph.icmp6_code != 0 ||
245 	    (m->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION &&
246 	     m->icmph.icmp6_type != NDISC_NEIGHBOUR_ADVERTISEMENT))
247 		return NULL;
248 
249 	return m;
250 }
251 
252 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
253 		       struct sk_buff *request, struct neighbour *n,
254 		       __be16 vlan_proto, u16 vlan_tci)
255 {
256 	struct net_device *dev = request->dev;
257 	struct net_bridge_vlan_group *vg;
258 	struct nd_msg *na, *ns;
259 	struct sk_buff *reply;
260 	struct ipv6hdr *pip6;
261 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
262 	int ns_olen;
263 	int i, len;
264 	u8 *daddr;
265 	u16 pvid;
266 
267 	if (!dev || skb_linearize(request))
268 		return;
269 
270 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
271 		sizeof(*na) + na_olen + dev->needed_tailroom;
272 
273 	reply = alloc_skb(len, GFP_ATOMIC);
274 	if (!reply)
275 		return;
276 
277 	reply->protocol = htons(ETH_P_IPV6);
278 	reply->dev = dev;
279 	skb_reserve(reply, LL_RESERVED_SPACE(dev));
280 	skb_push(reply, sizeof(struct ethhdr));
281 	skb_set_mac_header(reply, 0);
282 
283 	daddr = eth_hdr(request)->h_source;
284 	ns = (struct nd_msg *)(skb_network_header(request) +
285 			       sizeof(struct ipv6hdr));
286 
287 	/* Do we need option processing ? */
288 	ns_olen = request->len - (skb_network_offset(request) +
289 				  sizeof(struct ipv6hdr)) - sizeof(*ns);
290 	for (i = 0; i < ns_olen - 1; i += (ns->opt[i + 1] << 3)) {
291 		if (!ns->opt[i + 1] || i + (ns->opt[i + 1] << 3) > ns_olen) {
292 			kfree_skb(reply);
293 			return;
294 		}
295 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
296 			if ((ns->opt[i + 1] << 3) >=
297 			    sizeof(struct nd_opt_hdr) + ETH_ALEN)
298 				daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
299 			break;
300 		}
301 	}
302 
303 	/* Ethernet header */
304 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
305 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
306 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
307 	reply->protocol = htons(ETH_P_IPV6);
308 
309 	skb_pull(reply, sizeof(struct ethhdr));
310 	skb_set_network_header(reply, 0);
311 	skb_put(reply, sizeof(struct ipv6hdr));
312 
313 	/* IPv6 header */
314 	pip6 = ipv6_hdr(reply);
315 	memset(pip6, 0, sizeof(struct ipv6hdr));
316 	pip6->version = 6;
317 	pip6->priority = ipv6_hdr(request)->priority;
318 	pip6->nexthdr = IPPROTO_ICMPV6;
319 	pip6->hop_limit = 255;
320 	pip6->daddr = ipv6_hdr(request)->saddr;
321 	pip6->saddr = *(struct in6_addr *)n->primary_key;
322 
323 	skb_pull(reply, sizeof(struct ipv6hdr));
324 	skb_set_transport_header(reply, 0);
325 
326 	na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
327 
328 	/* Neighbor Advertisement */
329 	memset(na, 0, sizeof(*na) + na_olen);
330 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
331 	na->icmph.icmp6_router = (n->flags & NTF_ROUTER) ? 1 : 0;
332 	na->icmph.icmp6_override = 1;
333 	na->icmph.icmp6_solicited = 1;
334 	na->target = ns->target;
335 	ether_addr_copy(&na->opt[2], n->ha);
336 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
337 	na->opt[1] = na_olen >> 3;
338 
339 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
340 						&pip6->daddr,
341 						sizeof(*na) + na_olen,
342 						IPPROTO_ICMPV6,
343 						csum_partial(na, sizeof(*na) + na_olen, 0));
344 
345 	pip6->payload_len = htons(sizeof(*na) + na_olen);
346 
347 	skb_push(reply, sizeof(struct ipv6hdr));
348 	skb_push(reply, sizeof(struct ethhdr));
349 
350 	reply->ip_summed = CHECKSUM_UNNECESSARY;
351 
352 	if (p)
353 		vg = nbp_vlan_group_rcu(p);
354 	else
355 		vg = br_vlan_group_rcu(br);
356 	pvid = br_get_pvid(vg);
357 	if (pvid == (vlan_tci & VLAN_VID_MASK))
358 		vlan_tci = 0;
359 
360 	if (vlan_tci)
361 		__vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
362 
363 	netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
364 		   dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha);
365 
366 	if (p) {
367 		dev_queue_xmit(reply);
368 	} else {
369 		skb_reset_mac_header(reply);
370 		__skb_pull(reply, skb_network_offset(reply));
371 		reply->ip_summed = CHECKSUM_UNNECESSARY;
372 		reply->pkt_type = PACKET_HOST;
373 
374 		netif_rx(reply);
375 	}
376 }
377 
378 static int br_chk_addr_ip6(struct net_device *dev,
379 			   struct netdev_nested_priv *priv)
380 {
381 	struct in6_addr *addr = (struct in6_addr *)priv->data;
382 
383 	if (ipv6_chk_addr(dev_net(dev), addr, dev, 0))
384 		return 1;
385 
386 	return 0;
387 }
388 
389 static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr)
390 
391 {
392 	struct netdev_nested_priv priv = {
393 		.data = (void *)addr,
394 	};
395 
396 	if (br_chk_addr_ip6(dev, &priv))
397 		return true;
398 
399 	/* check if ip is configured on upper dev */
400 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, &priv))
401 		return true;
402 
403 	return false;
404 }
405 
406 void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
407 		       u16 vid, struct net_bridge_port *p, struct nd_msg *msg)
408 {
409 	struct net_device *dev = br->dev;
410 	struct net_device *vlandev = NULL;
411 	struct in6_addr *saddr, *daddr;
412 	struct ipv6hdr *iphdr;
413 	struct neighbour *n;
414 
415 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
416 
417 	if (br_is_neigh_suppress_enabled(p, vid))
418 		return;
419 
420 	if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
421 	    msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
422 		return;
423 
424 	if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT &&
425 	    !msg->icmph.icmp6_solicited) {
426 		/* prevent flooding to neigh suppress ports */
427 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
428 		return;
429 	}
430 
431 	if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
432 		return;
433 
434 	iphdr = ipv6_hdr(skb);
435 	saddr = &iphdr->saddr;
436 	daddr = &iphdr->daddr;
437 
438 	if (ipv6_addr_any(saddr) || !ipv6_addr_cmp(saddr, daddr)) {
439 		/* prevent flooding to neigh suppress ports */
440 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
441 		return;
442 	}
443 
444 	if (vid != 0) {
445 		/* build neigh table lookup on the vlan device */
446 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
447 						   vid);
448 		if (!vlandev)
449 			return;
450 	} else {
451 		vlandev = dev;
452 	}
453 
454 	if (br_is_local_ip6(vlandev, &msg->target)) {
455 		/* its our own ip, so don't proxy reply
456 		 * and don't forward to arp suppress ports
457 		 */
458 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
459 		return;
460 	}
461 
462 	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, vlandev);
463 	if (n) {
464 		struct net_bridge_fdb_entry *f;
465 
466 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
467 			neigh_release(n);
468 			return;
469 		}
470 
471 		f = br_fdb_find_rcu(br, n->ha, vid);
472 		if (f) {
473 			bool replied = false;
474 
475 			if (br_is_neigh_suppress_enabled(f->dst, vid)) {
476 				if (vid != 0)
477 					br_nd_send(br, p, skb, n,
478 						   skb->vlan_proto,
479 						   skb_vlan_tag_get(skb));
480 				else
481 					br_nd_send(br, p, skb, n, 0, 0);
482 				replied = true;
483 			}
484 
485 			/* If we have replied or as long as we know the
486 			 * mac, indicate to NEIGH_SUPPRESS ports that we
487 			 * have replied
488 			 */
489 			if (replied ||
490 			    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
491 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
492 		}
493 		neigh_release(n);
494 	}
495 }
496 #endif
497 
498 bool br_is_neigh_suppress_enabled(const struct net_bridge_port *p, u16 vid)
499 {
500 	if (!p)
501 		return false;
502 
503 	if (!vid)
504 		return !!(p->flags & BR_NEIGH_SUPPRESS);
505 
506 	if (p->flags & BR_NEIGH_VLAN_SUPPRESS) {
507 		struct net_bridge_vlan_group *vg = nbp_vlan_group_rcu(p);
508 		struct net_bridge_vlan *v;
509 
510 		v = br_vlan_find(vg, vid);
511 		if (!v)
512 			return false;
513 		return !!(v->priv_flags & BR_VLFLAG_NEIGH_SUPPRESS_ENABLED);
514 	} else {
515 		return !!(p->flags & BR_NEIGH_SUPPRESS);
516 	}
517 }
518