xref: /linux/net/bridge/br_arp_nd_proxy.c (revision 4ce06406958b67fdddcc2e6948237dd6ff6ba112)
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 
136 	if ((dev->flags & IFF_NOARP) ||
137 	    !pskb_may_pull(skb, arp_hdr_len(dev)))
138 		return;
139 
140 	parp = arp_hdr(skb);
141 
142 	if (parp->ar_pro != htons(ETH_P_IP) ||
143 	    parp->ar_hln != dev->addr_len ||
144 	    parp->ar_pln != 4)
145 		return;
146 
147 	arpptr = (u8 *)parp + sizeof(struct arphdr);
148 	sha = arpptr;
149 	arpptr += dev->addr_len;	/* sha */
150 	memcpy(&sip, arpptr, sizeof(sip));
151 	arpptr += sizeof(sip);
152 	arpptr += dev->addr_len;	/* tha */
153 	memcpy(&tip, arpptr, sizeof(tip));
154 
155 	if (ipv4_is_loopback(tip) ||
156 	    ipv4_is_multicast(tip))
157 		return;
158 
159 	if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
160 		if (br_is_neigh_suppress_enabled(p, vid))
161 			return;
162 		if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
163 		    parp->ar_op == htons(ARPOP_REQUEST))
164 			return;
165 		if (parp->ar_op != htons(ARPOP_RREQUEST) &&
166 		    parp->ar_op != htons(ARPOP_RREPLY) &&
167 		    (ipv4_is_zeronet(sip) || sip == tip)) {
168 			/* prevent flooding to neigh suppress ports */
169 			BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
170 			return;
171 		}
172 	}
173 
174 	if (parp->ar_op != htons(ARPOP_REQUEST))
175 		return;
176 
177 	if (vid != 0) {
178 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
179 						   vid);
180 		if (!vlandev)
181 			return;
182 	}
183 
184 	if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
185 	    br_is_local_ip(vlandev, tip)) {
186 		/* its our local ip, so don't proxy reply
187 		 * and don't forward to neigh suppress ports
188 		 */
189 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
190 		return;
191 	}
192 
193 	n = neigh_lookup(&arp_tbl, &tip, vlandev);
194 	if (n) {
195 		struct net_bridge_fdb_entry *f;
196 
197 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
198 			neigh_release(n);
199 			return;
200 		}
201 
202 		f = br_fdb_find_rcu(br, n->ha, vid);
203 		if (f) {
204 			bool replied = false;
205 
206 			if ((p && (p->flags & BR_PROXYARP)) ||
207 			    (f->dst && (f->dst->flags & BR_PROXYARP_WIFI)) ||
208 			    br_is_neigh_suppress_enabled(f->dst, vid)) {
209 				if (!vid)
210 					br_arp_send(br, p, skb->dev, sip, tip,
211 						    sha, n->ha, sha, 0, 0);
212 				else
213 					br_arp_send(br, p, skb->dev, sip, tip,
214 						    sha, n->ha, sha,
215 						    skb->vlan_proto,
216 						    skb_vlan_tag_get(skb));
217 				replied = true;
218 			}
219 
220 			/* If we have replied or as long as we know the
221 			 * mac, indicate to arp replied
222 			 */
223 			if (replied ||
224 			    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
225 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
226 		}
227 
228 		neigh_release(n);
229 	}
230 }
231 #endif
232 
233 #if IS_ENABLED(CONFIG_IPV6)
234 struct nd_msg *br_is_nd_neigh_msg(const struct sk_buff *skb, struct nd_msg *msg)
235 {
236 	struct nd_msg *m;
237 
238 	m = skb_header_pointer(skb, skb_network_offset(skb) +
239 			       sizeof(struct ipv6hdr), sizeof(*msg), msg);
240 	if (!m)
241 		return NULL;
242 
243 	if (m->icmph.icmp6_code != 0 ||
244 	    (m->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION &&
245 	     m->icmph.icmp6_type != NDISC_NEIGHBOUR_ADVERTISEMENT))
246 		return NULL;
247 
248 	return m;
249 }
250 
251 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
252 		       struct sk_buff *request, struct neighbour *n,
253 		       __be16 vlan_proto, u16 vlan_tci, struct nd_msg *ns)
254 {
255 	struct net_device *dev = request->dev;
256 	struct net_bridge_vlan_group *vg;
257 	struct sk_buff *reply;
258 	struct nd_msg *na;
259 	struct ipv6hdr *pip6;
260 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
261 	int ns_olen;
262 	int i, len;
263 	u8 *daddr;
264 	u16 pvid;
265 
266 	if (!dev)
267 		return;
268 
269 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
270 		sizeof(*na) + na_olen + dev->needed_tailroom;
271 
272 	reply = alloc_skb(len, GFP_ATOMIC);
273 	if (!reply)
274 		return;
275 
276 	reply->protocol = htons(ETH_P_IPV6);
277 	reply->dev = dev;
278 	skb_reserve(reply, LL_RESERVED_SPACE(dev));
279 	skb_push(reply, sizeof(struct ethhdr));
280 	skb_set_mac_header(reply, 0);
281 
282 	daddr = eth_hdr(request)->h_source;
283 
284 	/* Do we need option processing ? */
285 	ns_olen = request->len - (skb_network_offset(request) +
286 				  sizeof(struct ipv6hdr)) - sizeof(*ns);
287 	for (i = 0; i < ns_olen - 1; i += (ns->opt[i + 1] << 3)) {
288 		if (!ns->opt[i + 1]) {
289 			kfree_skb(reply);
290 			return;
291 		}
292 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
293 			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
294 			break;
295 		}
296 	}
297 
298 	/* Ethernet header */
299 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
300 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
301 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
302 	reply->protocol = htons(ETH_P_IPV6);
303 
304 	skb_pull(reply, sizeof(struct ethhdr));
305 	skb_set_network_header(reply, 0);
306 	skb_put(reply, sizeof(struct ipv6hdr));
307 
308 	/* IPv6 header */
309 	pip6 = ipv6_hdr(reply);
310 	memset(pip6, 0, sizeof(struct ipv6hdr));
311 	pip6->version = 6;
312 	pip6->priority = ipv6_hdr(request)->priority;
313 	pip6->nexthdr = IPPROTO_ICMPV6;
314 	pip6->hop_limit = 255;
315 	pip6->daddr = ipv6_hdr(request)->saddr;
316 	pip6->saddr = *(struct in6_addr *)n->primary_key;
317 
318 	skb_pull(reply, sizeof(struct ipv6hdr));
319 	skb_set_transport_header(reply, 0);
320 
321 	na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
322 
323 	/* Neighbor Advertisement */
324 	memset(na, 0, sizeof(*na) + na_olen);
325 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
326 	na->icmph.icmp6_router = (n->flags & NTF_ROUTER) ? 1 : 0;
327 	na->icmph.icmp6_override = 1;
328 	na->icmph.icmp6_solicited = 1;
329 	na->target = ns->target;
330 	ether_addr_copy(&na->opt[2], n->ha);
331 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
332 	na->opt[1] = na_olen >> 3;
333 
334 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
335 						&pip6->daddr,
336 						sizeof(*na) + na_olen,
337 						IPPROTO_ICMPV6,
338 						csum_partial(na, sizeof(*na) + na_olen, 0));
339 
340 	pip6->payload_len = htons(sizeof(*na) + na_olen);
341 
342 	skb_push(reply, sizeof(struct ipv6hdr));
343 	skb_push(reply, sizeof(struct ethhdr));
344 
345 	reply->ip_summed = CHECKSUM_UNNECESSARY;
346 
347 	if (p)
348 		vg = nbp_vlan_group_rcu(p);
349 	else
350 		vg = br_vlan_group_rcu(br);
351 	pvid = br_get_pvid(vg);
352 	if (pvid == (vlan_tci & VLAN_VID_MASK))
353 		vlan_tci = 0;
354 
355 	if (vlan_tci)
356 		__vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
357 
358 	netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
359 		   dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha);
360 
361 	if (p) {
362 		dev_queue_xmit(reply);
363 	} else {
364 		skb_reset_mac_header(reply);
365 		__skb_pull(reply, skb_network_offset(reply));
366 		reply->ip_summed = CHECKSUM_UNNECESSARY;
367 		reply->pkt_type = PACKET_HOST;
368 
369 		netif_rx(reply);
370 	}
371 }
372 
373 static int br_chk_addr_ip6(struct net_device *dev,
374 			   struct netdev_nested_priv *priv)
375 {
376 	struct in6_addr *addr = (struct in6_addr *)priv->data;
377 
378 	if (ipv6_chk_addr(dev_net(dev), addr, dev, 0))
379 		return 1;
380 
381 	return 0;
382 }
383 
384 static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr)
385 
386 {
387 	struct netdev_nested_priv priv = {
388 		.data = (void *)addr,
389 	};
390 
391 	if (br_chk_addr_ip6(dev, &priv))
392 		return true;
393 
394 	/* check if ip is configured on upper dev */
395 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, &priv))
396 		return true;
397 
398 	return false;
399 }
400 
401 void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
402 		       u16 vid, struct net_bridge_port *p, struct nd_msg *msg)
403 {
404 	struct net_device *dev = br->dev;
405 	struct net_device *vlandev = NULL;
406 	struct in6_addr *saddr, *daddr;
407 	struct ipv6hdr *iphdr;
408 	struct neighbour *n;
409 
410 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
411 
412 	if (br_is_neigh_suppress_enabled(p, vid))
413 		return;
414 
415 	if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
416 	    msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
417 		return;
418 
419 	if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT &&
420 	    !msg->icmph.icmp6_solicited) {
421 		/* prevent flooding to neigh suppress ports */
422 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
423 		return;
424 	}
425 
426 	if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
427 		return;
428 
429 	iphdr = ipv6_hdr(skb);
430 	saddr = &iphdr->saddr;
431 	daddr = &iphdr->daddr;
432 
433 	if (ipv6_addr_any(saddr) || !ipv6_addr_cmp(saddr, daddr)) {
434 		/* prevent flooding to neigh suppress ports */
435 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
436 		return;
437 	}
438 
439 	if (vid != 0) {
440 		/* build neigh table lookup on the vlan device */
441 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
442 						   vid);
443 		if (!vlandev)
444 			return;
445 	} else {
446 		vlandev = dev;
447 	}
448 
449 	if (br_is_local_ip6(vlandev, &msg->target)) {
450 		/* its our own ip, so don't proxy reply
451 		 * and don't forward to arp suppress ports
452 		 */
453 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
454 		return;
455 	}
456 
457 	n = neigh_lookup(&nd_tbl, &msg->target, vlandev);
458 	if (n) {
459 		struct net_bridge_fdb_entry *f;
460 
461 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
462 			neigh_release(n);
463 			return;
464 		}
465 
466 		f = br_fdb_find_rcu(br, n->ha, vid);
467 		if (f) {
468 			bool replied = false;
469 
470 			if (br_is_neigh_suppress_enabled(f->dst, vid)) {
471 				if (vid != 0)
472 					br_nd_send(br, p, skb, n,
473 						   skb->vlan_proto,
474 						   skb_vlan_tag_get(skb), msg);
475 				else
476 					br_nd_send(br, p, skb, n, 0, 0, msg);
477 				replied = true;
478 			}
479 
480 			/* If we have replied or as long as we know the
481 			 * mac, indicate to NEIGH_SUPPRESS ports that we
482 			 * have replied
483 			 */
484 			if (replied ||
485 			    br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
486 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
487 		}
488 		neigh_release(n);
489 	}
490 }
491 #endif
492 
493 bool br_is_neigh_suppress_enabled(const struct net_bridge_port *p, u16 vid)
494 {
495 	if (!p)
496 		return false;
497 
498 	if (!vid)
499 		return !!(p->flags & BR_NEIGH_SUPPRESS);
500 
501 	if (p->flags & BR_NEIGH_VLAN_SUPPRESS) {
502 		struct net_bridge_vlan_group *vg = nbp_vlan_group_rcu(p);
503 		struct net_bridge_vlan *v;
504 
505 		v = br_vlan_find(vg, vid);
506 		if (!v)
507 			return false;
508 		return !!(v->priv_flags & BR_VLFLAG_NEIGH_SUPPRESS_ENABLED);
509 	} else {
510 		return !!(p->flags & BR_NEIGH_SUPPRESS);
511 	}
512 }
513