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 #include <net/ndisc.h>
23 #endif
24
25 #include "br_private.h"
26
br_recalculate_neigh_suppress_enabled(struct net_bridge * br)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 (READ_ONCE(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)
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)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
br_chk_addr_ip(struct net_device * dev,struct netdev_nested_priv * priv)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
br_is_local_ip(struct net_device * dev,__be32 ip)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
br_do_proxy_suppress_arp(struct sk_buff * skb,struct net_bridge * br,u16 vid,struct net_bridge_port * p)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 BR_INPUT_SKB_CB(skb)->grat_arp = 0;
137
138 if ((dev->flags & IFF_NOARP) ||
139 !pskb_may_pull(skb, arp_hdr_len(dev)))
140 return;
141
142 parp = arp_hdr(skb);
143
144 if (parp->ar_pro != htons(ETH_P_IP) ||
145 parp->ar_hln != dev->addr_len ||
146 parp->ar_pln != 4)
147 return;
148
149 arpptr = (u8 *)parp + sizeof(struct arphdr);
150 sha = arpptr;
151 arpptr += dev->addr_len; /* sha */
152 memcpy(&sip, arpptr, sizeof(sip));
153 arpptr += sizeof(sip);
154 arpptr += dev->addr_len; /* tha */
155 memcpy(&tip, arpptr, sizeof(tip));
156
157 if (ipv4_is_loopback(tip) ||
158 ipv4_is_multicast(tip))
159 return;
160
161 if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED)) {
162 if (br_is_neigh_suppress_enabled(p, vid))
163 return;
164 if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
165 parp->ar_op == htons(ARPOP_REQUEST))
166 return;
167 if (parp->ar_op != htons(ARPOP_RREQUEST) &&
168 parp->ar_op != htons(ARPOP_RREPLY) &&
169 sip == tip) {
170 /* prevent flooding to neigh suppress ports */
171 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
172 BR_INPUT_SKB_CB(skb)->grat_arp = 1;
173 return;
174 }
175 }
176
177 if (parp->ar_op != htons(ARPOP_REQUEST))
178 return;
179
180 if (vid != 0) {
181 vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
182 vid);
183 if (!vlandev)
184 return;
185 }
186
187 if (br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED) &&
188 br_is_local_ip(vlandev, tip)) {
189 /* its our local ip, so don't proxy reply
190 * and don't forward to neigh suppress ports
191 */
192 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
193 return;
194 }
195
196 n = neigh_lookup(&arp_tbl, &tip, vlandev);
197 if (n) {
198 struct net_bridge_fdb_entry *f;
199 u8 ha[ETH_ALEN] __aligned(2);
200
201 if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
202 neigh_release(n);
203 return;
204 }
205
206 neigh_ha_snapshot(ha, n, n->dev);
207 f = br_fdb_find_rcu(br, ha, vid);
208 if (f) {
209 const struct net_bridge_port *dst = READ_ONCE(f->dst);
210 bool replied = false;
211
212 if ((p && test_bit(BR_PROXYARP_BIT, &p->flags)) ||
213 (dst && test_bit(BR_PROXYARP_WIFI_BIT, &dst->flags)) ||
214 br_is_neigh_suppress_enabled(dst, vid)) {
215 if (!vid)
216 br_arp_send(br, p, skb->dev, sip, tip,
217 sha, ha, sha, 0, 0);
218 else
219 br_arp_send(br, p, skb->dev, sip, tip,
220 sha, ha, sha,
221 skb->vlan_proto,
222 skb_vlan_tag_get(skb));
223 replied = true;
224 }
225
226 /* If we have replied or as long as we know the
227 * mac, indicate to arp replied
228 */
229 if (replied ||
230 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
231 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
232 }
233
234 neigh_release(n);
235 }
236 }
237 #endif
238
239 #if IS_ENABLED(CONFIG_IPV6)
240 /* Validate skb as an NS/NA and linearize it for br_nd_send()'s ND
241 * option parsing; returns the nd_msg, or NULL on failure.
242 */
br_is_nd_neigh_msg(struct sk_buff * skb)243 struct nd_msg *br_is_nd_neigh_msg(struct sk_buff *skb)
244 {
245 if (ndisc_check_ns_na(skb))
246 return NULL;
247
248 if (skb_linearize(skb))
249 return NULL;
250
251 return (struct nd_msg *)skb_transport_header(skb);
252 }
253
br_nd_send(struct net_bridge * br,struct net_bridge_port * p,struct sk_buff * request,struct neighbour * n,u8 * ha,__be16 vlan_proto,u16 vlan_tci)254 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
255 struct sk_buff *request, struct neighbour *n, u8 *ha,
256 __be16 vlan_proto, u16 vlan_tci)
257 {
258 struct net_device *dev = request->dev;
259 struct net_bridge_vlan_group *vg;
260 struct ndisc_options ndopts;
261 struct nd_msg *na, *ns;
262 struct sk_buff *reply;
263 struct ipv6hdr *pip6;
264 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
265 int ns_olen;
266 u8 *daddr;
267 bool dad;
268 u16 pvid;
269 int len;
270
271 if (!dev)
272 return;
273
274 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
275 sizeof(*na) + na_olen + dev->needed_tailroom;
276
277 reply = alloc_skb(len, GFP_ATOMIC);
278 if (!reply)
279 return;
280
281 reply->protocol = htons(ETH_P_IPV6);
282 reply->dev = dev;
283 skb_reserve(reply, LL_RESERVED_SPACE(dev));
284 skb_push(reply, sizeof(struct ethhdr));
285 skb_set_mac_header(reply, 0);
286
287 daddr = eth_hdr(request)->h_source;
288 ns = (struct nd_msg *)skb_transport_header(request);
289
290 /* Derive the option length from the IPv6 payload length so that any
291 * trailing L2 padding in the skb is not parsed as ND options.
292 */
293 ns_olen = ntohs(ipv6_hdr(request)->payload_len) - sizeof(*ns);
294 if (!ndisc_parse_options(dev, ns->opt, ns_olen, &ndopts)) {
295 kfree_skb(reply);
296 return;
297 }
298
299 if (ndopts.nd_opts_src_lladdr) {
300 u8 *lladdr;
301
302 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_src_lladdr, dev);
303 if (lladdr)
304 daddr = lladdr;
305 }
306
307 dad = ipv6_addr_any(&ipv6_hdr(request)->saddr);
308
309 /* Ethernet header */
310 if (dad)
311 ipv6_eth_mc_map(&in6addr_linklocal_allnodes, eth_hdr(reply)->h_dest);
312 else
313 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
314 ether_addr_copy(eth_hdr(reply)->h_source, ha);
315 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
316 reply->protocol = htons(ETH_P_IPV6);
317
318 skb_pull(reply, sizeof(struct ethhdr));
319 skb_set_network_header(reply, 0);
320 skb_put(reply, sizeof(struct ipv6hdr));
321
322 /* IPv6 header */
323 pip6 = ipv6_hdr(reply);
324 memset(pip6, 0, sizeof(struct ipv6hdr));
325 pip6->version = 6;
326 pip6->priority = ipv6_hdr(request)->priority;
327 pip6->nexthdr = IPPROTO_ICMPV6;
328 pip6->hop_limit = 255;
329 pip6->daddr = dad ? in6addr_linklocal_allnodes : ipv6_hdr(request)->saddr;
330 pip6->saddr = *(struct in6_addr *)n->primary_key;
331
332 skb_pull(reply, sizeof(struct ipv6hdr));
333 skb_set_transport_header(reply, 0);
334
335 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
336
337 /* Neighbor Advertisement */
338 memset(na, 0, sizeof(*na) + na_olen);
339 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
340 na->icmph.icmp6_router = (n->flags & NTF_ROUTER) ? 1 : 0;
341 na->icmph.icmp6_override = 1;
342 na->icmph.icmp6_solicited = dad ? 0 : 1;
343 na->target = ns->target;
344 ether_addr_copy(&na->opt[2], ha);
345 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
346 na->opt[1] = na_olen >> 3;
347
348 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
349 &pip6->daddr,
350 sizeof(*na) + na_olen,
351 IPPROTO_ICMPV6,
352 csum_partial(na, sizeof(*na) + na_olen, 0));
353
354 pip6->payload_len = htons(sizeof(*na) + na_olen);
355
356 skb_push(reply, sizeof(struct ipv6hdr));
357 skb_push(reply, sizeof(struct ethhdr));
358
359 reply->ip_summed = CHECKSUM_UNNECESSARY;
360
361 if (p)
362 vg = nbp_vlan_group_rcu(p);
363 else
364 vg = br_vlan_group_rcu(br);
365 pvid = br_get_pvid(vg);
366 if (pvid == (vlan_tci & VLAN_VID_MASK))
367 vlan_tci = 0;
368
369 if (vlan_tci)
370 __vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
371
372 netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
373 dev->name, &pip6->daddr, daddr, &pip6->saddr, ha);
374
375 if (p) {
376 dev_queue_xmit(reply);
377 } else {
378 skb_reset_mac_header(reply);
379 __skb_pull(reply, skb_network_offset(reply));
380 reply->ip_summed = CHECKSUM_UNNECESSARY;
381 reply->pkt_type = PACKET_HOST;
382
383 netif_rx(reply);
384 }
385 }
386
br_chk_addr_ip6(struct net_device * dev,struct netdev_nested_priv * priv)387 static int br_chk_addr_ip6(struct net_device *dev,
388 struct netdev_nested_priv *priv)
389 {
390 struct in6_addr *addr = (struct in6_addr *)priv->data;
391
392 if (ipv6_chk_addr(dev_net(dev), addr, dev, 0))
393 return 1;
394
395 return 0;
396 }
397
br_is_local_ip6(struct net_device * dev,struct in6_addr * addr)398 static bool br_is_local_ip6(struct net_device *dev, struct in6_addr *addr)
399
400 {
401 struct netdev_nested_priv priv = {
402 .data = (void *)addr,
403 };
404
405 if (br_chk_addr_ip6(dev, &priv))
406 return true;
407
408 /* check if ip is configured on upper dev */
409 if (netdev_walk_all_upper_dev_rcu(dev, br_chk_addr_ip6, &priv))
410 return true;
411
412 return false;
413 }
414
br_do_suppress_nd(struct sk_buff * skb,struct net_bridge * br,u16 vid,struct net_bridge_port * p,struct nd_msg * msg)415 void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
416 u16 vid, struct net_bridge_port *p, struct nd_msg *msg)
417 {
418 struct net_device *dev = br->dev;
419 struct net_device *vlandev = NULL;
420 struct in6_addr *saddr, *daddr;
421 struct ipv6hdr *iphdr;
422 struct neighbour *n;
423
424 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 0;
425 BR_INPUT_SKB_CB(skb)->grat_arp = 0;
426
427 if (br_is_neigh_suppress_enabled(p, vid))
428 return;
429
430 if (is_unicast_ether_addr(eth_hdr(skb)->h_dest) &&
431 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
432 return;
433
434 if (msg->icmph.icmp6_type == NDISC_NEIGHBOUR_ADVERTISEMENT &&
435 !msg->icmph.icmp6_solicited) {
436 /* prevent flooding to neigh suppress ports */
437 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
438 BR_INPUT_SKB_CB(skb)->grat_arp = 1;
439 return;
440 }
441
442 if (msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
443 return;
444
445 iphdr = ipv6_hdr(skb);
446 saddr = &iphdr->saddr;
447 daddr = &iphdr->daddr;
448
449 if (!ipv6_addr_cmp(saddr, daddr)) {
450 /* prevent flooding to neigh suppress ports */
451 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
452 return;
453 }
454
455 if (vid != 0) {
456 /* build neigh table lookup on the vlan device */
457 vlandev = __vlan_find_dev_deep_rcu(br->dev, skb->vlan_proto,
458 vid);
459 if (!vlandev)
460 return;
461 } else {
462 vlandev = dev;
463 }
464
465 if (br_is_local_ip6(vlandev, &msg->target)) {
466 /* its our own ip, so don't proxy reply
467 * and don't forward to arp suppress ports
468 */
469 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
470 return;
471 }
472
473 n = neigh_lookup(&nd_tbl, &msg->target, vlandev);
474 if (n) {
475 struct net_bridge_fdb_entry *f;
476 u8 ha[ETH_ALEN] __aligned(2);
477
478 if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
479 neigh_release(n);
480 return;
481 }
482
483 neigh_ha_snapshot(ha, n, n->dev);
484 f = br_fdb_find_rcu(br, ha, vid);
485 if (f) {
486 const struct net_bridge_port *dst = READ_ONCE(f->dst);
487 bool replied = false;
488
489 if (br_is_neigh_suppress_enabled(dst, vid)) {
490 if (vid != 0)
491 br_nd_send(br, p, skb, n, ha,
492 skb->vlan_proto,
493 skb_vlan_tag_get(skb));
494 else
495 br_nd_send(br, p, skb, n, ha, 0, 0);
496 replied = true;
497 }
498
499 /* If we have replied or as long as we know the
500 * mac, indicate to NEIGH_SUPPRESS ports that we
501 * have replied
502 */
503 if (replied ||
504 br_opt_get(br, BROPT_NEIGH_SUPPRESS_ENABLED))
505 BR_INPUT_SKB_CB(skb)->proxyarp_replied = 1;
506 }
507 neigh_release(n);
508 }
509 }
510 #endif
511
br_is_neigh_suppress_enabled(const struct net_bridge_port * p,u16 vid)512 bool br_is_neigh_suppress_enabled(const struct net_bridge_port *p, u16 vid)
513 {
514 if (!p)
515 return false;
516
517 if (vid && test_bit(BR_NEIGH_VLAN_SUPPRESS_BIT, &p->flags)) {
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 }
526 return test_bit(BR_NEIGH_SUPPRESS_BIT, &p->flags);
527 }
528
br_is_neigh_forward_grat_enabled(const struct net_bridge_port * p,u16 vid)529 bool br_is_neigh_forward_grat_enabled(const struct net_bridge_port *p, u16 vid)
530 {
531 if (vid && test_bit(BR_NEIGH_VLAN_SUPPRESS_BIT, &p->flags)) {
532 struct net_bridge_vlan_group *vg = nbp_vlan_group_rcu(p);
533 struct net_bridge_vlan *v;
534
535 v = br_vlan_find(vg, vid);
536 if (!v)
537 return false;
538 return !!(v->priv_flags & BR_VLFLAG_NEIGH_FORWARD_GRAT_ENABLED);
539 }
540 return test_bit(BR_NEIGH_FORWARD_GRAT_BIT, &p->flags);
541 }
542