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
br_recalculate_neigh_suppress_enabled(struct net_bridge * br)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)
br_arp_send(struct net_bridge * br,struct net_bridge_port * p,struct net_device * dev,__be32 dest_ip,__be32 src_ip,const unsigned char * dest_hw,const unsigned char * src_hw,const unsigned char * target_hw,__be16 vlan_proto,u16 vlan_tci)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
br_chk_addr_ip(struct net_device * dev,struct netdev_nested_priv * priv)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
br_is_local_ip(struct net_device * dev,__be32 ip)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
br_do_proxy_suppress_arp(struct sk_buff * skb,struct net_bridge * br,u16 vid,struct net_bridge_port * p)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 const struct net_bridge_port *dst = READ_ONCE(f->dst);
205 bool replied = false;
206
207 if ((p && (p->flags & BR_PROXYARP)) ||
208 (dst && (dst->flags & BR_PROXYARP_WIFI)) ||
209 br_is_neigh_suppress_enabled(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)
br_is_nd_neigh_msg(const struct sk_buff * skb,struct nd_msg * msg)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
br_nd_send(struct net_bridge * br,struct net_bridge_port * p,struct sk_buff * request,struct neighbour * n,__be16 vlan_proto,u16 vlan_tci)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
br_chk_addr_ip6(struct net_device * dev,struct netdev_nested_priv * priv)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
br_is_local_ip6(struct net_device * dev,struct in6_addr * addr)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
br_do_suppress_nd(struct sk_buff * skb,struct net_bridge * br,u16 vid,struct net_bridge_port * p,struct nd_msg * msg)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(&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 const struct net_bridge_port *dst = READ_ONCE(f->dst);
474 bool replied = false;
475
476 if (br_is_neigh_suppress_enabled(dst, vid)) {
477 if (vid != 0)
478 br_nd_send(br, p, skb, n,
479 skb->vlan_proto,
480 skb_vlan_tag_get(skb));
481 else
482 br_nd_send(br, p, skb, n, 0, 0);
483 replied = true;
484 }
485
486 /* If we have replied or as long as we know the
487 * mac, indicate to NEIGH_SUPPRESS ports that we
488 * have replied
489 */
490 if (replied ||
491 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
492 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
493 }
494 neigh_release(n);
495 }
496 }
497 #endif
498
br_is_neigh_suppress_enabled(const struct net_bridge_port * p,u16 vid)499 bool br_is_neigh_suppress_enabled(const struct net_bridge_port *p, u16 vid)
500 {
501 if (!p)
502 return false;
503
504 if (!vid)
505 return !!(p->flags & BR_NEIGH_SUPPRESS);
506
507 if (p->flags & BR_NEIGH_VLAN_SUPPRESS) {
508 struct net_bridge_vlan_group *vg = nbp_vlan_group_rcu(p);
509 struct net_bridge_vlan *v;
510
511 v = br_vlan_find(vg, vid);
512 if (!v)
513 return false;
514 return !!(v->priv_flags & BR_VLFLAG_NEIGH_SUPPRESS_ENABLED);
515 } else {
516 return !!(p->flags & BR_NEIGH_SUPPRESS);
517 }
518 }
519