xref: /openbmc/linux/net/bridge/br_arp_nd_proxy.c (revision 98ddec80) 
1 /*
2  *  Handle bridge arp/nd proxy/suppress
3  *
4  *  Copyright (C) 2017 Cumulus Networks
5  *  Copyright (c) 2017 Roopa Prabhu <roopa@cumulusnetworks.com>
6  *
7  *  Authors:
8  *	Roopa Prabhu <roopa@cumulusnetworks.com>
9  *
10  *  This program is free software; you can redistribute it and/or
11  *  modify it under the terms of the GNU General Public License
12  *  as published by the Free Software Foundation; either version
13  *  2 of the License, or (at your option) any later version.
14  */
15 
16 #include <linux/kernel.h>
17 #include <linux/netdevice.h>
18 #include <linux/etherdevice.h>
19 #include <linux/neighbour.h>
20 #include <net/arp.h>
21 #include <linux/if_vlan.h>
22 #include <linux/inetdevice.h>
23 #include <net/addrconf.h>
24 #if IS_ENABLED(CONFIG_IPV6)
25 #include <net/ip6_checksum.h>
26 #endif
27 
28 #include "br_private.h"
29 
30 void br_recalculate_neigh_suppress_enabled(struct net_bridge *br)
31 {
32 	struct net_bridge_port *p;
33 	bool neigh_suppress = false;
34 
35 	list_for_each_entry(p, &br->port_list, list) {
36 		if (p->flags & BR_NEIGH_SUPPRESS) {
37 			neigh_suppress = true;
38 			break;
39 		}
40 	}
41 
42 	br->neigh_suppress_enabled = neigh_suppress;
43 }
44 
45 #if IS_ENABLED(CONFIG_INET)
46 static void br_arp_send(struct net_bridge *br, struct net_bridge_port *p,
47 			struct net_device *dev, __be32 dest_ip, __be32 src_ip,
48 			const unsigned char *dest_hw,
49 			const unsigned char *src_hw,
50 			const unsigned char *target_hw,
51 			__be16 vlan_proto, u16 vlan_tci)
52 {
53 	struct net_bridge_vlan_group *vg;
54 	struct sk_buff *skb;
55 	u16 pvid;
56 
57 	netdev_dbg(dev, "arp send dev %s dst %pI4 dst_hw %pM src %pI4 src_hw %pM\n",
58 		   dev->name, &dest_ip, dest_hw, &src_ip, src_hw);
59 
60 	if (!vlan_tci) {
61 		arp_send(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
62 			 dest_hw, src_hw, target_hw);
63 		return;
64 	}
65 
66 	skb = arp_create(ARPOP_REPLY, ETH_P_ARP, dest_ip, dev, src_ip,
67 			 dest_hw, src_hw, target_hw);
68 	if (!skb)
69 		return;
70 
71 	if (p)
72 		vg = nbp_vlan_group_rcu(p);
73 	else
74 		vg = br_vlan_group_rcu(br);
75 	pvid = br_get_pvid(vg);
76 	if (pvid == (vlan_tci & VLAN_VID_MASK))
77 		vlan_tci = 0;
78 
79 	if (vlan_tci)
80 		__vlan_hwaccel_put_tag(skb, vlan_proto, vlan_tci);
81 
82 	if (p) {
83 		arp_xmit(skb);
84 	} else {
85 		skb_reset_mac_header(skb);
86 		__skb_pull(skb, skb_network_offset(skb));
87 		skb->ip_summed = CHECKSUM_UNNECESSARY;
88 		skb->pkt_type = PACKET_HOST;
89 
90 		netif_rx_ni(skb);
91 	}
92 }
93 
94 static int br_chk_addr_ip(struct net_device *dev, void *data)
95 {
96 	__be32 ip = *(__be32 *)data;
97 	struct in_device *in_dev;
98 	__be32 addr = 0;
99 
100 	in_dev = __in_dev_get_rcu(dev);
101 	if (in_dev)
102 		addr = inet_confirm_addr(dev_net(dev), in_dev, 0, ip,
103 					 RT_SCOPE_HOST);
104 
105 	if (addr == ip)
106 		return 1;
107 
108 	return 0;
109 }
110 
111 static bool br_is_local_ip(struct net_device *dev, __be32 ip)
112 {
113 	if (br_chk_addr_ip(dev, &ip))
114 		return true;
115 
116 	/* check if ip is configured on upper dev */
117 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip, &ip))
118 		return true;
119 
120 	return false;
121 }
122 
123 void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br,
124 			      u16 vid, struct net_bridge_port *p)
125 {
126 	struct net_device *dev = br->dev;
127 	struct net_device *vlandev = dev;
128 	struct neighbour *n;
129 	struct arphdr *parp;
130 	u8 *arpptr, *sha;
131 	__be32 sip, tip;
132 
133 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = false;
134 
135 	if ((dev->flags & IFF_NOARP) ||
136 	    !pskb_may_pull(skb, arp_hdr_len(dev)))
137 		return;
138 
139 	parp = arp_hdr(skb);
140 
141 	if (parp->ar_pro != htons(ETH_P_IP) ||
142 	    parp->ar_hln != dev->addr_len ||
143 	    parp->ar_pln != 4)
144 		return;
145 
146 	arpptr = (u8 *)parp + sizeof(struct arphdr);
147 	sha = arpptr;
148 	arpptr += dev->addr_len;	/* sha */
149 	memcpy(&sip, arpptr, sizeof(sip));
150 	arpptr += sizeof(sip);
151 	arpptr += dev->addr_len;	/* tha */
152 	memcpy(&tip, arpptr, sizeof(tip));
153 
154 	if (ipv4_is_loopback(tip) ||
155 	    ipv4_is_multicast(tip))
156 		return;
157 
158 	if (br->neigh_suppress_enabled) {
159 		if (p && (p->flags & BR_NEIGH_SUPPRESS))
160 			return;
161 		if (ipv4_is_zeronet(sip) || sip == tip) {
162 			/* prevent flooding to neigh suppress ports */
163 			BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
164 			return;
165 		}
166 	}
167 
168 	if (parp->ar_op != htons(ARPOP_REQUEST))
169 		return;
170 
171 	if (vid != 0) {
172 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
173 						   vid);
174 		if (!vlandev)
175 			return;
176 	}
177 
178 	if (br->neigh_suppress_enabled && br_is_local_ip(vlandev, tip)) {
179 		/* its our local ip, so don't proxy reply
180 		 * and don't forward to neigh suppress ports
181 		 */
182 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
183 		return;
184 	}
185 
186 	n = neigh_lookup(&arp_tbl, &tip, vlandev);
187 	if (n) {
188 		struct net_bridge_fdb_entry *f;
189 
190 		if (!(n->nud_state & NUD_VALID)) {
191 			neigh_release(n);
192 			return;
193 		}
194 
195 		f = br_fdb_find_rcu(br, n->ha, vid);
196 		if (f) {
197 			bool replied = false;
198 
199 			if ((p && (p->flags & BR_PROXYARP)) ||
200 			    (f->dst && (f->dst->flags & (BR_PROXYARP_WIFI |
201 							 BR_NEIGH_SUPPRESS)))) {
202 				if (!vid)
203 					br_arp_send(br, p, skb->dev, sip, tip,
204 						    sha, n->ha, sha, 0, 0);
205 				else
206 					br_arp_send(br, p, skb->dev, sip, tip,
207 						    sha, n->ha, sha,
208 						    skb->vlan_proto,
209 						    skb_vlan_tag_get(skb));
210 				replied = true;
211 			}
212 
213 			/* If we have replied or as long as we know the
214 			 * mac, indicate to arp replied
215 			 */
216 			if (replied || br->neigh_suppress_enabled)
217 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
218 		}
219 
220 		neigh_release(n);
221 	}
222 }
223 #endif
224 
225 #if IS_ENABLED(CONFIG_IPV6)
226 struct nd_msg *br_is_nd_neigh_msg(struct sk_buff *skb, struct nd_msg *msg)
227 {
228 	struct nd_msg *m;
229 
230 	m = skb_header_pointer(skb, skb_network_offset(skb) +
231 			       sizeof(struct ipv6hdr), sizeof(*msg), msg);
232 	if (!m)
233 		return NULL;
234 
235 	if (m->icmph.icmp6_code != 0 ||
236 	    (m->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION &&
237 	     m->icmph.icmp6_type != NDISC_NEIGHBOUR_ADVERTISEMENT))
238 		return NULL;
239 
240 	return m;
241 }
242 
243 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
244 		       struct sk_buff *request, struct neighbour *n,
245 		       __be16 vlan_proto, u16 vlan_tci, struct nd_msg *ns)
246 {
247 	struct net_device *dev = request->dev;
248 	struct net_bridge_vlan_group *vg;
249 	struct sk_buff *reply;
250 	struct nd_msg *na;
251 	struct ipv6hdr *pip6;
252 	int na_olen = 8; /* opt hdr + ETH_ALEN for target */
253 	int ns_olen;
254 	int i, len;
255 	u8 *daddr;
256 	u16 pvid;
257 
258 	if (!dev)
259 		return;
260 
261 	len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
262 		sizeof(*na) + na_olen + dev->needed_tailroom;
263 
264 	reply = alloc_skb(len, GFP_ATOMIC);
265 	if (!reply)
266 		return;
267 
268 	reply->protocol = htons(ETH_P_IPV6);
269 	reply->dev = dev;
270 	skb_reserve(reply, LL_RESERVED_SPACE(dev));
271 	skb_push(reply, sizeof(struct ethhdr));
272 	skb_set_mac_header(reply, 0);
273 
274 	daddr = eth_hdr(request)->h_source;
275 
276 	/* Do we need option processing ? */
277 	ns_olen = request->len - (skb_network_offset(request) +
278 				  sizeof(struct ipv6hdr)) - sizeof(*ns);
279 	for (i = 0; i < ns_olen - 1; i += (ns->opt[i + 1] << 3)) {
280 		if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
281 			daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
282 			break;
283 		}
284 	}
285 
286 	/* Ethernet header */
287 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
288 	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
289 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
290 	reply->protocol = htons(ETH_P_IPV6);
291 
292 	skb_pull(reply, sizeof(struct ethhdr));
293 	skb_set_network_header(reply, 0);
294 	skb_put(reply, sizeof(struct ipv6hdr));
295 
296 	/* IPv6 header */
297 	pip6 = ipv6_hdr(reply);
298 	memset(pip6, 0, sizeof(struct ipv6hdr));
299 	pip6->version = 6;
300 	pip6->priority = ipv6_hdr(request)->priority;
301 	pip6->nexthdr = IPPROTO_ICMPV6;
302 	pip6->hop_limit = 255;
303 	pip6->daddr = ipv6_hdr(request)->saddr;
304 	pip6->saddr = *(struct in6_addr *)n->primary_key;
305 
306 	skb_pull(reply, sizeof(struct ipv6hdr));
307 	skb_set_transport_header(reply, 0);
308 
309 	na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
310 
311 	/* Neighbor Advertisement */
312 	memset(na, 0, sizeof(*na) + na_olen);
313 	na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
314 	na->icmph.icmp6_router = 0; /* XXX: should be 1 ? */
315 	na->icmph.icmp6_override = 1;
316 	na->icmph.icmp6_solicited = 1;
317 	na->target = ns->target;
318 	ether_addr_copy(&na->opt[2], n->ha);
319 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
320 	na->opt[1] = na_olen >> 3;
321 
322 	na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
323 						&pip6->daddr,
324 						sizeof(*na) + na_olen,
325 						IPPROTO_ICMPV6,
326 						csum_partial(na, sizeof(*na) + na_olen, 0));
327 
328 	pip6->payload_len = htons(sizeof(*na) + na_olen);
329 
330 	skb_push(reply, sizeof(struct ipv6hdr));
331 	skb_push(reply, sizeof(struct ethhdr));
332 
333 	reply->ip_summed = CHECKSUM_UNNECESSARY;
334 
335 	if (p)
336 		vg = nbp_vlan_group_rcu(p);
337 	else
338 		vg = br_vlan_group_rcu(br);
339 	pvid = br_get_pvid(vg);
340 	if (pvid == (vlan_tci & VLAN_VID_MASK))
341 		vlan_tci = 0;
342 
343 	if (vlan_tci)
344 		__vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
345 
346 	netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
347 		   dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha);
348 
349 	if (p) {
350 		dev_queue_xmit(reply);
351 	} else {
352 		skb_reset_mac_header(reply);
353 		__skb_pull(reply, skb_network_offset(reply));
354 		reply->ip_summed = CHECKSUM_UNNECESSARY;
355 		reply->pkt_type = PACKET_HOST;
356 
357 		netif_rx_ni(reply);
358 	}
359 }
360 
361 static int br_chk_addr_ip6(struct net_device *dev, void *data)
362 {
363 	struct in6_addr *addr = (struct in6_addr *)data;
364 
365 	if (ipv6_chk_addr(dev_net(dev), addr, dev, 0))
366 		return 1;
367 
368 	return 0;
369 }
370 
371 static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr)
372 
373 {
374 	if (br_chk_addr_ip6(dev, addr))
375 		return true;
376 
377 	/* check if ip is configured on upper dev */
378 	if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, addr))
379 		return true;
380 
381 	return false;
382 }
383 
384 void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
385 		       u16 vid, struct net_bridge_port *p, struct nd_msg *msg)
386 {
387 	struct net_device *dev = br->dev;
388 	struct net_device *vlandev = NULL;
389 	struct in6_addr *saddr, *daddr;
390 	struct ipv6hdr *iphdr;
391 	struct neighbour *n;
392 
393 	BR_INPUT_SKB_CB(skb)->proxyarp_replied = false;
394 
395 	if (p && (p->flags & BR_NEIGH_SUPPRESS))
396 		return;
397 
398 	if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT &&
399 	    !msg->icmph.icmp6_solicited) {
400 		/* prevent flooding to neigh suppress ports */
401 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
402 		return;
403 	}
404 
405 	if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
406 		return;
407 
408 	iphdr = ipv6_hdr(skb);
409 	saddr = &iphdr->saddr;
410 	daddr = &iphdr->daddr;
411 
412 	if (ipv6_addr_any(saddr) || !ipv6_addr_cmp(saddr, daddr)) {
413 		/* prevent flooding to neigh suppress ports */
414 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
415 		return;
416 	}
417 
418 	if (vid != 0) {
419 		/* build neigh table lookup on the vlan device */
420 		vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
421 						   vid);
422 		if (!vlandev)
423 			return;
424 	} else {
425 		vlandev = dev;
426 	}
427 
428 	if (br_is_local_ip6(vlandev, &msg->target)) {
429 		/* its our own ip, so don't proxy reply
430 		 * and don't forward to arp suppress ports
431 		 */
432 		BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
433 		return;
434 	}
435 
436 	n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, vlandev);
437 	if (n) {
438 		struct net_bridge_fdb_entry *f;
439 
440 		if (!(n->nud_state & NUD_VALID)) {
441 			neigh_release(n);
442 			return;
443 		}
444 
445 		f = br_fdb_find_rcu(br, n->ha, vid);
446 		if (f) {
447 			bool replied = false;
448 
449 			if (f->dst && (f->dst->flags & BR_NEIGH_SUPPRESS)) {
450 				if (vid != 0)
451 					br_nd_send(br, p, skb, n,
452 						   skb->vlan_proto,
453 						   skb_vlan_tag_get(skb), msg);
454 				else
455 					br_nd_send(br, p, skb, n, 0, 0, msg);
456 				replied = true;
457 			}
458 
459 			/* If we have replied or as long as we know the
460 			 * mac, indicate to NEIGH_SUPPRESS ports that we
461 			 * have replied
462 			 */
463 			if (replied || br->neigh_suppress_enabled)
464 				BR_INPUT_SKB_CB(skb)->proxyarp_replied = true;
465 		}
466 		neigh_release(n);
467 	}
468 }
469 #endif
470