1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * VXLAN: Virtual eXtensible Local Area Network
4 *
5 * Copyright (c) 2012-2013 Vyatta Inc.
6 */
7
8 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
9
10 #include <linux/kernel.h>
11 #include <linux/module.h>
12 #include <linux/errno.h>
13 #include <linux/slab.h>
14 #include <linux/udp.h>
15 #include <linux/igmp.h>
16 #include <linux/if_ether.h>
17 #include <linux/ethtool.h>
18 #include <linux/rhashtable.h>
19 #include <net/arp.h>
20 #include <net/ndisc.h>
21 #include <net/gro.h>
22 #include <net/ipv6_stubs.h>
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/rtnetlink.h>
26 #include <net/inet_ecn.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
29 #include <net/netdev_lock.h>
30 #include <net/tun_proto.h>
31 #include <net/vxlan.h>
32 #include <net/nexthop.h>
33
34 #if IS_ENABLED(CONFIG_IPV6)
35 #include <net/ip6_tunnel.h>
36 #include <net/ip6_checksum.h>
37 #endif
38
39 #include "vxlan_private.h"
40
41 #define VXLAN_VERSION "0.1"
42
43 #define FDB_AGE_DEFAULT 300 /* 5 min */
44 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
45
46 /* UDP port for VXLAN traffic.
47 * The IANA assigned port is 4789, but the Linux default is 8472
48 * for compatibility with early adopters.
49 */
50 static unsigned short vxlan_port __read_mostly = 8472;
51 module_param_named(udp_port, vxlan_port, ushort, 0444);
52 MODULE_PARM_DESC(udp_port, "Destination UDP port");
53
54 static bool log_ecn_error = true;
55 module_param(log_ecn_error, bool, 0644);
56 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
57
58 unsigned int vxlan_net_id;
59
60 const u8 all_zeros_mac[ETH_ALEN + 2];
61 static struct rtnl_link_ops vxlan_link_ops;
62
63 static int vxlan_sock_add(struct vxlan_dev *vxlan);
64
65 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
66
67 static const struct rhashtable_params vxlan_fdb_rht_params = {
68 .head_offset = offsetof(struct vxlan_fdb, rhnode),
69 .key_offset = offsetof(struct vxlan_fdb, key),
70 .key_len = sizeof(struct vxlan_fdb_key),
71 .automatic_shrinking = true,
72 };
73
vxlan_collect_metadata(struct vxlan_sock * vs)74 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
75 {
76 return vs->flags & VXLAN_F_COLLECT_METADATA ||
77 ip_tunnel_collect_metadata();
78 }
79
80 /* Find VXLAN socket based on network namespace, address family, UDP port,
81 * enabled unshareable flags and socket device binding (see l3mdev with
82 * non-default VRF).
83 */
vxlan_find_sock(struct net * net,sa_family_t family,__be16 port,u32 flags,int ifindex)84 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
85 __be16 port, u32 flags, int ifindex)
86 {
87 struct vxlan_sock *vs;
88
89 flags &= VXLAN_F_RCV_FLAGS;
90
91 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
92 if (inet_sk(vs->sock->sk)->inet_sport == port &&
93 vxlan_get_sk_family(vs) == family &&
94 vs->flags == flags &&
95 vs->sock->sk->sk_bound_dev_if == ifindex)
96 return vs;
97 }
98 return NULL;
99 }
100
vxlan_vs_find_vni(struct vxlan_sock * vs,int ifindex,__be32 vni,struct vxlan_vni_node ** vninode)101 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs,
102 int ifindex, __be32 vni,
103 struct vxlan_vni_node **vninode)
104 {
105 struct vxlan_vni_node *vnode;
106 struct vxlan_dev_node *node;
107
108 /* For flow based devices, map all packets to VNI 0 */
109 if (vs->flags & VXLAN_F_COLLECT_METADATA &&
110 !(vs->flags & VXLAN_F_VNIFILTER))
111 vni = 0;
112
113 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
114 if (!node->vxlan)
115 continue;
116 vnode = NULL;
117 if (node->vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
118 vnode = vxlan_vnifilter_lookup(node->vxlan, vni);
119 if (!vnode)
120 continue;
121 } else if (node->vxlan->default_dst.remote_vni != vni) {
122 continue;
123 }
124
125 if (IS_ENABLED(CONFIG_IPV6)) {
126 const struct vxlan_config *cfg = &node->vxlan->cfg;
127
128 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
129 cfg->remote_ifindex != ifindex)
130 continue;
131 }
132
133 if (vninode)
134 *vninode = vnode;
135 return node->vxlan;
136 }
137
138 return NULL;
139 }
140
141 /* Look up VNI in a per net namespace table */
vxlan_find_vni(struct net * net,int ifindex,__be32 vni,sa_family_t family,__be16 port,u32 flags)142 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
143 __be32 vni, sa_family_t family,
144 __be16 port, u32 flags)
145 {
146 struct vxlan_sock *vs;
147
148 vs = vxlan_find_sock(net, family, port, flags, ifindex);
149 if (!vs)
150 return NULL;
151
152 return vxlan_vs_find_vni(vs, ifindex, vni, NULL);
153 }
154
155 /* Fill in neighbour message in skbuff. */
vxlan_fdb_info(struct sk_buff * skb,struct vxlan_dev * vxlan,const struct vxlan_fdb * fdb,u32 portid,u32 seq,int type,unsigned int flags,const struct vxlan_rdst * rdst)156 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
157 const struct vxlan_fdb *fdb,
158 u32 portid, u32 seq, int type, unsigned int flags,
159 const struct vxlan_rdst *rdst)
160 {
161 unsigned long now = jiffies;
162 struct nda_cacheinfo ci;
163 bool send_ip, send_eth;
164 struct nlmsghdr *nlh;
165 struct nexthop *nh;
166 struct ndmsg *ndm;
167 int nh_family;
168 u32 nh_id;
169
170 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
171 if (nlh == NULL)
172 return -EMSGSIZE;
173
174 ndm = nlmsg_data(nlh);
175 memset(ndm, 0, sizeof(*ndm));
176
177 send_eth = send_ip = true;
178
179 rcu_read_lock();
180 nh = rcu_dereference(fdb->nh);
181 if (nh) {
182 nh_family = nexthop_get_family(nh);
183 nh_id = nh->id;
184 }
185 rcu_read_unlock();
186
187 if (type == RTM_GETNEIGH) {
188 if (rdst) {
189 send_ip = !vxlan_addr_any(&rdst->remote_ip);
190 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
191 } else if (nh) {
192 ndm->ndm_family = nh_family;
193 }
194 send_eth = !is_zero_ether_addr(fdb->key.eth_addr);
195 } else
196 ndm->ndm_family = AF_BRIDGE;
197 ndm->ndm_state = fdb->state;
198 ndm->ndm_ifindex = vxlan->dev->ifindex;
199 ndm->ndm_flags = fdb->flags;
200 if (rdst && rdst->offloaded)
201 ndm->ndm_flags |= NTF_OFFLOADED;
202 ndm->ndm_type = RTN_UNICAST;
203
204 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
205 nla_put_s32(skb, NDA_LINK_NETNSID,
206 peernet2id(dev_net(vxlan->dev), vxlan->net)))
207 goto nla_put_failure;
208
209 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->key.eth_addr))
210 goto nla_put_failure;
211 if (nh) {
212 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
213 goto nla_put_failure;
214 } else if (rdst) {
215 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
216 &rdst->remote_ip))
217 goto nla_put_failure;
218
219 if (rdst->remote_port &&
220 rdst->remote_port != vxlan->cfg.dst_port &&
221 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
222 goto nla_put_failure;
223 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
224 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
225 goto nla_put_failure;
226 if (rdst->remote_ifindex &&
227 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
228 goto nla_put_failure;
229 }
230
231 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->key.vni &&
232 nla_put_u32(skb, NDA_SRC_VNI,
233 be32_to_cpu(fdb->key.vni)))
234 goto nla_put_failure;
235
236 ci.ndm_used = jiffies_to_clock_t(now - READ_ONCE(fdb->used));
237 ci.ndm_confirmed = 0;
238 ci.ndm_updated = jiffies_to_clock_t(now - READ_ONCE(fdb->updated));
239 ci.ndm_refcnt = 0;
240
241 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
242 goto nla_put_failure;
243
244 nlmsg_end(skb, nlh);
245 return 0;
246
247 nla_put_failure:
248 nlmsg_cancel(skb, nlh);
249 return -EMSGSIZE;
250 }
251
vxlan_nlmsg_size(void)252 static inline size_t vxlan_nlmsg_size(void)
253 {
254 return NLMSG_ALIGN(sizeof(struct ndmsg))
255 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
256 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
257 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
258 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
259 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
260 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
261 + nla_total_size(sizeof(struct nda_cacheinfo));
262 }
263
__vxlan_fdb_notify(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,struct vxlan_rdst * rd,int type)264 static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
265 struct vxlan_rdst *rd, int type)
266 {
267 struct net *net = dev_net(vxlan->dev);
268 struct sk_buff *skb;
269 int err = -ENOBUFS;
270
271 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
272 if (skb == NULL)
273 goto errout;
274
275 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
276 if (err < 0) {
277 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
278 WARN_ON(err == -EMSGSIZE);
279 kfree_skb(skb);
280 goto errout;
281 }
282
283 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
284 return;
285 errout:
286 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
287 }
288
vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev * vxlan,const struct vxlan_fdb * fdb,const struct vxlan_rdst * rd,struct netlink_ext_ack * extack,struct switchdev_notifier_vxlan_fdb_info * fdb_info)289 static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
290 const struct vxlan_fdb *fdb,
291 const struct vxlan_rdst *rd,
292 struct netlink_ext_ack *extack,
293 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
294 {
295 fdb_info->info.dev = vxlan->dev;
296 fdb_info->info.extack = extack;
297 fdb_info->remote_ip = rd->remote_ip;
298 fdb_info->remote_port = rd->remote_port;
299 fdb_info->remote_vni = rd->remote_vni;
300 fdb_info->remote_ifindex = rd->remote_ifindex;
301 memcpy(fdb_info->eth_addr, fdb->key.eth_addr, ETH_ALEN);
302 fdb_info->vni = fdb->key.vni;
303 fdb_info->offloaded = rd->offloaded;
304 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
305 }
306
vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,struct vxlan_rdst * rd,bool adding,struct netlink_ext_ack * extack)307 static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
308 struct vxlan_fdb *fdb,
309 struct vxlan_rdst *rd,
310 bool adding,
311 struct netlink_ext_ack *extack)
312 {
313 struct switchdev_notifier_vxlan_fdb_info info;
314 enum switchdev_notifier_type notifier_type;
315 int ret;
316
317 if (WARN_ON(!rd))
318 return 0;
319
320 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
321 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
322 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
323 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
324 &info.info, extack);
325 return notifier_to_errno(ret);
326 }
327
vxlan_fdb_notify(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,struct vxlan_rdst * rd,int type,bool swdev_notify,struct netlink_ext_ack * extack)328 static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
329 struct vxlan_rdst *rd, int type, bool swdev_notify,
330 struct netlink_ext_ack *extack)
331 {
332 int err;
333
334 if (swdev_notify && rd) {
335 switch (type) {
336 case RTM_NEWNEIGH:
337 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
338 true, extack);
339 if (err)
340 return err;
341 break;
342 case RTM_DELNEIGH:
343 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
344 false, extack);
345 break;
346 }
347 }
348
349 __vxlan_fdb_notify(vxlan, fdb, rd, type);
350 return 0;
351 }
352
vxlan_ip_miss(struct net_device * dev,union vxlan_addr * ipa)353 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
354 {
355 struct vxlan_dev *vxlan = netdev_priv(dev);
356 struct vxlan_fdb f = {
357 .state = NUD_STALE,
358 };
359 struct vxlan_rdst remote = {
360 .remote_ip = *ipa, /* goes to NDA_DST */
361 .remote_vni = cpu_to_be32(VXLAN_N_VID),
362 };
363
364 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
365 }
366
vxlan_fdb_miss(struct vxlan_dev * vxlan,const u8 eth_addr[ETH_ALEN])367 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
368 {
369 struct vxlan_fdb f = {
370 .state = NUD_STALE,
371 };
372 struct vxlan_rdst remote = { };
373
374 memcpy(f.key.eth_addr, eth_addr, ETH_ALEN);
375
376 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
377 }
378
379 /* Look up Ethernet address in forwarding table */
vxlan_find_mac_rcu(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)380 static struct vxlan_fdb *vxlan_find_mac_rcu(struct vxlan_dev *vxlan,
381 const u8 *mac, __be32 vni)
382 {
383 struct vxlan_fdb_key key;
384
385 memset(&key, 0, sizeof(key));
386 memcpy(key.eth_addr, mac, sizeof(key.eth_addr));
387 if (!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA))
388 key.vni = vxlan->default_dst.remote_vni;
389 else
390 key.vni = vni;
391
392 return rhashtable_lookup(&vxlan->fdb_hash_tbl, &key,
393 vxlan_fdb_rht_params);
394 }
395
vxlan_find_mac_tx(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)396 static struct vxlan_fdb *vxlan_find_mac_tx(struct vxlan_dev *vxlan,
397 const u8 *mac, __be32 vni)
398 {
399 struct vxlan_fdb *f;
400
401 f = vxlan_find_mac_rcu(vxlan, mac, vni);
402 if (f) {
403 unsigned long now = jiffies;
404
405 if (READ_ONCE(f->used) != now)
406 WRITE_ONCE(f->used, now);
407 }
408
409 return f;
410 }
411
vxlan_find_mac(struct vxlan_dev * vxlan,const u8 * mac,__be32 vni)412 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
413 const u8 *mac, __be32 vni)
414 {
415 struct vxlan_fdb *f;
416
417 lockdep_assert_held_once(&vxlan->hash_lock);
418
419 rcu_read_lock();
420 f = vxlan_find_mac_rcu(vxlan, mac, vni);
421 rcu_read_unlock();
422
423 return f;
424 }
425
426 /* caller should hold vxlan->hash_lock */
vxlan_fdb_find_rdst(struct vxlan_fdb * f,union vxlan_addr * ip,__be16 port,__be32 vni,__u32 ifindex)427 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
428 union vxlan_addr *ip, __be16 port,
429 __be32 vni, __u32 ifindex)
430 {
431 struct vxlan_rdst *rd;
432
433 list_for_each_entry(rd, &f->remotes, list) {
434 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
435 rd->remote_port == port &&
436 rd->remote_vni == vni &&
437 rd->remote_ifindex == ifindex)
438 return rd;
439 }
440
441 return NULL;
442 }
443
vxlan_fdb_find_uc(struct net_device * dev,const u8 * mac,__be32 vni,struct switchdev_notifier_vxlan_fdb_info * fdb_info)444 int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
445 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
446 {
447 struct vxlan_dev *vxlan = netdev_priv(dev);
448 u8 eth_addr[ETH_ALEN + 2] = { 0 };
449 struct vxlan_rdst *rdst = NULL;
450 struct vxlan_fdb *f;
451 int rc = 0;
452
453 if (is_multicast_ether_addr(mac) ||
454 is_zero_ether_addr(mac))
455 return -EINVAL;
456
457 ether_addr_copy(eth_addr, mac);
458
459 rcu_read_lock();
460
461 f = vxlan_find_mac_rcu(vxlan, eth_addr, vni);
462 if (f)
463 rdst = first_remote_rcu(f);
464 if (!rdst) {
465 rc = -ENOENT;
466 goto out;
467 }
468
469 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
470
471 out:
472 rcu_read_unlock();
473 return rc;
474 }
475 EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
476
vxlan_fdb_notify_one(struct notifier_block * nb,const struct vxlan_dev * vxlan,const struct vxlan_fdb * f,const struct vxlan_rdst * rdst,struct netlink_ext_ack * extack)477 static int vxlan_fdb_notify_one(struct notifier_block *nb,
478 const struct vxlan_dev *vxlan,
479 const struct vxlan_fdb *f,
480 const struct vxlan_rdst *rdst,
481 struct netlink_ext_ack *extack)
482 {
483 struct switchdev_notifier_vxlan_fdb_info fdb_info;
484 int rc;
485
486 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
487 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
488 &fdb_info);
489 return notifier_to_errno(rc);
490 }
491
vxlan_fdb_replay(const struct net_device * dev,__be32 vni,struct notifier_block * nb,struct netlink_ext_ack * extack)492 int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
493 struct notifier_block *nb,
494 struct netlink_ext_ack *extack)
495 {
496 struct vxlan_dev *vxlan;
497 struct vxlan_rdst *rdst;
498 struct vxlan_fdb *f;
499 int rc = 0;
500
501 if (!netif_is_vxlan(dev))
502 return -EINVAL;
503 vxlan = netdev_priv(dev);
504
505 spin_lock_bh(&vxlan->hash_lock);
506 hlist_for_each_entry(f, &vxlan->fdb_list, fdb_node) {
507 if (f->key.vni == vni) {
508 list_for_each_entry(rdst, &f->remotes, list) {
509 rc = vxlan_fdb_notify_one(nb, vxlan, f, rdst,
510 extack);
511 if (rc)
512 goto unlock;
513 }
514 }
515 }
516 spin_unlock_bh(&vxlan->hash_lock);
517 return 0;
518
519 unlock:
520 spin_unlock_bh(&vxlan->hash_lock);
521 return rc;
522 }
523 EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
524
vxlan_fdb_clear_offload(const struct net_device * dev,__be32 vni)525 void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
526 {
527 struct vxlan_dev *vxlan;
528 struct vxlan_rdst *rdst;
529 struct vxlan_fdb *f;
530
531 if (!netif_is_vxlan(dev))
532 return;
533 vxlan = netdev_priv(dev);
534
535 spin_lock_bh(&vxlan->hash_lock);
536 hlist_for_each_entry(f, &vxlan->fdb_list, fdb_node) {
537 if (f->key.vni == vni) {
538 list_for_each_entry(rdst, &f->remotes, list)
539 rdst->offloaded = false;
540 }
541 }
542 spin_unlock_bh(&vxlan->hash_lock);
543
544 }
545 EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
546
547 /* Replace destination of unicast mac */
vxlan_fdb_replace(struct vxlan_fdb * f,union vxlan_addr * ip,__be16 port,__be32 vni,__u32 ifindex,struct vxlan_rdst * oldrd)548 static int vxlan_fdb_replace(struct vxlan_fdb *f,
549 union vxlan_addr *ip, __be16 port, __be32 vni,
550 __u32 ifindex, struct vxlan_rdst *oldrd)
551 {
552 struct vxlan_rdst *rd;
553
554 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
555 if (rd)
556 return 0;
557
558 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
559 if (!rd)
560 return 0;
561
562 *oldrd = *rd;
563 dst_cache_reset(&rd->dst_cache);
564 rd->remote_ip = *ip;
565 rd->remote_port = port;
566 rd->remote_vni = vni;
567 rd->remote_ifindex = ifindex;
568 rd->offloaded = false;
569 return 1;
570 }
571
572 /* Add/update destinations for multicast */
vxlan_fdb_append(struct vxlan_fdb * f,union vxlan_addr * ip,__be16 port,__be32 vni,__u32 ifindex,struct vxlan_rdst ** rdp)573 static int vxlan_fdb_append(struct vxlan_fdb *f,
574 union vxlan_addr *ip, __be16 port, __be32 vni,
575 __u32 ifindex, struct vxlan_rdst **rdp)
576 {
577 struct vxlan_rdst *rd;
578
579 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
580 if (rd)
581 return 0;
582
583 rd = kmalloc_obj(*rd, GFP_ATOMIC);
584 if (rd == NULL)
585 return -ENOMEM;
586
587 /* The driver can work correctly without a dst cache, so do not treat
588 * dst cache initialization errors as fatal.
589 */
590 dst_cache_init(&rd->dst_cache, GFP_ATOMIC | __GFP_NOWARN);
591
592 rd->remote_ip = *ip;
593 rd->remote_port = port;
594 rd->offloaded = false;
595 rd->remote_vni = vni;
596 rd->remote_ifindex = ifindex;
597
598 list_add_tail_rcu(&rd->list, &f->remotes);
599
600 *rdp = rd;
601 return 1;
602 }
603
vxlan_parse_gpe_proto(const struct vxlanhdr * hdr,__be16 * protocol)604 static bool vxlan_parse_gpe_proto(const struct vxlanhdr *hdr, __be16 *protocol)
605 {
606 const struct vxlanhdr_gpe *gpe = (const struct vxlanhdr_gpe *)hdr;
607
608 /* Need to have Next Protocol set for interfaces in GPE mode. */
609 if (!gpe->np_applied)
610 return false;
611 /* "The initial version is 0. If a receiver does not support the
612 * version indicated it MUST drop the packet.
613 */
614 if (gpe->version != 0)
615 return false;
616 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
617 * processing MUST occur." However, we don't implement OAM
618 * processing, thus drop the packet.
619 */
620 if (gpe->oam_flag)
621 return false;
622
623 *protocol = tun_p_to_eth_p(gpe->next_protocol);
624 if (!*protocol)
625 return false;
626
627 return true;
628 }
629
vxlan_gro_remcsum(struct sk_buff * skb,unsigned int off,struct vxlanhdr * vh,size_t hdrlen,__be32 vni_field,struct gro_remcsum * grc,bool nopartial)630 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
631 unsigned int off,
632 struct vxlanhdr *vh, size_t hdrlen,
633 __be32 vni_field,
634 struct gro_remcsum *grc,
635 bool nopartial)
636 {
637 size_t start, offset;
638
639 if (skb->remcsum_offload)
640 return vh;
641
642 if (!NAPI_GRO_CB(skb)->csum_valid)
643 return NULL;
644
645 start = vxlan_rco_start(vni_field);
646 offset = start + vxlan_rco_offset(vni_field);
647
648 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
649 start, offset, grc, nopartial);
650
651 skb->remcsum_offload = 1;
652
653 return vh;
654 }
655
vxlan_gro_prepare_receive(struct sock * sk,struct list_head * head,struct sk_buff * skb,struct gro_remcsum * grc)656 static struct vxlanhdr *vxlan_gro_prepare_receive(struct sock *sk,
657 struct list_head *head,
658 struct sk_buff *skb,
659 struct gro_remcsum *grc)
660 {
661 struct sk_buff *p;
662 struct vxlanhdr *vh, *vh2;
663 unsigned int hlen, off_vx;
664 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
665 __be32 flags;
666
667 skb_gro_remcsum_init(grc);
668
669 off_vx = skb_gro_offset(skb);
670 hlen = off_vx + sizeof(*vh);
671 vh = skb_gro_header(skb, hlen, off_vx);
672 if (unlikely(!vh))
673 return NULL;
674
675 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
676
677 flags = vh->vx_flags;
678
679 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
680 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
681 vh->vx_vni, grc,
682 !!(vs->flags &
683 VXLAN_F_REMCSUM_NOPARTIAL));
684
685 if (!vh)
686 return NULL;
687 }
688
689 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
690
691 list_for_each_entry(p, head, list) {
692 if (!NAPI_GRO_CB(p)->same_flow)
693 continue;
694
695 vh2 = (struct vxlanhdr *)(p->data + off_vx);
696 if (vh->vx_flags != vh2->vx_flags ||
697 vh->vx_vni != vh2->vx_vni) {
698 NAPI_GRO_CB(p)->same_flow = 0;
699 continue;
700 }
701 }
702
703 return vh;
704 }
705
vxlan_gro_receive(struct sock * sk,struct list_head * head,struct sk_buff * skb)706 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
707 struct list_head *head,
708 struct sk_buff *skb)
709 {
710 struct sk_buff *pp = NULL;
711 struct gro_remcsum grc;
712 int flush = 1;
713
714 if (vxlan_gro_prepare_receive(sk, head, skb, &grc)) {
715 pp = call_gro_receive(eth_gro_receive, head, skb);
716 flush = 0;
717 }
718 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
719 return pp;
720 }
721
vxlan_gpe_gro_receive(struct sock * sk,struct list_head * head,struct sk_buff * skb)722 static struct sk_buff *vxlan_gpe_gro_receive(struct sock *sk,
723 struct list_head *head,
724 struct sk_buff *skb)
725 {
726 const struct packet_offload *ptype;
727 struct sk_buff *pp = NULL;
728 struct gro_remcsum grc;
729 struct vxlanhdr *vh;
730 __be16 protocol;
731 int flush = 1;
732
733 vh = vxlan_gro_prepare_receive(sk, head, skb, &grc);
734 if (vh) {
735 if (!vxlan_parse_gpe_proto(vh, &protocol))
736 goto out;
737 ptype = gro_find_receive_by_type(protocol);
738 if (!ptype)
739 goto out;
740 pp = call_gro_receive(ptype->callbacks.gro_receive, head, skb);
741 flush = 0;
742 }
743 out:
744 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
745 return pp;
746 }
747
vxlan_gro_complete(struct sock * sk,struct sk_buff * skb,int nhoff)748 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
749 {
750 /* Sets 'skb->inner_mac_header' since we are always called with
751 * 'skb->encapsulation' set.
752 */
753 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
754 }
755
vxlan_gpe_gro_complete(struct sock * sk,struct sk_buff * skb,int nhoff)756 static int vxlan_gpe_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
757 {
758 struct vxlanhdr *vh = (struct vxlanhdr *)(skb->data + nhoff);
759 const struct packet_offload *ptype;
760 int err = -ENOSYS;
761 __be16 protocol;
762
763 if (!vxlan_parse_gpe_proto(vh, &protocol))
764 return err;
765 ptype = gro_find_complete_by_type(protocol);
766 if (ptype)
767 err = ptype->callbacks.gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
768 return err;
769 }
770
vxlan_fdb_alloc(struct vxlan_dev * vxlan,const u8 * mac,__u16 state,__be32 src_vni,__u16 ndm_flags)771 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
772 __u16 state, __be32 src_vni,
773 __u16 ndm_flags)
774 {
775 struct vxlan_fdb *f;
776
777 f = kmalloc_obj(*f, GFP_ATOMIC);
778 if (!f)
779 return NULL;
780 memset(&f->key, 0, sizeof(f->key));
781 f->state = state;
782 f->flags = ndm_flags;
783 f->updated = f->used = jiffies;
784 f->key.vni = src_vni;
785 f->nh = NULL;
786 RCU_INIT_POINTER(f->vdev, vxlan);
787 INIT_LIST_HEAD(&f->nh_list);
788 INIT_LIST_HEAD(&f->remotes);
789 memcpy(f->key.eth_addr, mac, ETH_ALEN);
790
791 return f;
792 }
793
vxlan_fdb_nh_update(struct vxlan_dev * vxlan,struct vxlan_fdb * fdb,u32 nhid,struct netlink_ext_ack * extack)794 static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
795 u32 nhid, struct netlink_ext_ack *extack)
796 {
797 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
798 struct nexthop *nh;
799 int err = -EINVAL;
800
801 if (old_nh && old_nh->id == nhid)
802 return 0;
803
804 nh = nexthop_find_by_id(vxlan->net, nhid);
805 if (!nh) {
806 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
807 goto err_inval;
808 }
809
810 if (!nexthop_get(nh)) {
811 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
812 nh = NULL;
813 goto err_inval;
814 }
815 if (!nexthop_is_fdb(nh)) {
816 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
817 goto err_inval;
818 }
819
820 if (!nexthop_is_multipath(nh)) {
821 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
822 goto err_inval;
823 }
824
825 /* check nexthop group family */
826 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
827 case AF_INET:
828 if (!nexthop_has_v4(nh)) {
829 err = -EAFNOSUPPORT;
830 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
831 goto err_inval;
832 }
833 break;
834 case AF_INET6:
835 if (nexthop_has_v4(nh)) {
836 err = -EAFNOSUPPORT;
837 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
838 goto err_inval;
839 }
840 }
841
842 if (old_nh) {
843 list_del_rcu(&fdb->nh_list);
844 nexthop_put(old_nh);
845 }
846 rcu_assign_pointer(fdb->nh, nh);
847 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
848 return 1;
849
850 err_inval:
851 if (nh)
852 nexthop_put(nh);
853 return err;
854 }
855
vxlan_fdb_create(struct vxlan_dev * vxlan,const u8 * mac,union vxlan_addr * ip,__u16 state,__be16 port,__be32 src_vni,__be32 vni,__u32 ifindex,__u16 ndm_flags,u32 nhid,struct vxlan_fdb ** fdb,struct netlink_ext_ack * extack)856 int vxlan_fdb_create(struct vxlan_dev *vxlan,
857 const u8 *mac, union vxlan_addr *ip,
858 __u16 state, __be16 port, __be32 src_vni,
859 __be32 vni, __u32 ifindex, __u16 ndm_flags,
860 u32 nhid, struct vxlan_fdb **fdb,
861 struct netlink_ext_ack *extack)
862 {
863 struct vxlan_rdst *rd = NULL;
864 struct vxlan_fdb *f;
865 int rc;
866
867 if (vxlan->cfg.addrmax &&
868 vxlan->addrcnt >= vxlan->cfg.addrmax)
869 return -ENOSPC;
870
871 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
872 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
873 if (!f)
874 return -ENOMEM;
875
876 if (nhid)
877 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
878 else
879 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
880 if (rc < 0)
881 goto errout;
882
883 rc = rhashtable_lookup_insert_fast(&vxlan->fdb_hash_tbl, &f->rhnode,
884 vxlan_fdb_rht_params);
885 if (rc)
886 goto destroy_remote;
887
888 ++vxlan->addrcnt;
889 hlist_add_head_rcu(&f->fdb_node, &vxlan->fdb_list);
890
891 *fdb = f;
892
893 return 0;
894
895 destroy_remote:
896 if (rcu_access_pointer(f->nh)) {
897 list_del_rcu(&f->nh_list);
898 nexthop_put(rtnl_dereference(f->nh));
899 } else {
900 list_del(&rd->list);
901 dst_cache_destroy(&rd->dst_cache);
902 kfree(rd);
903 }
904 errout:
905 kfree(f);
906 return rc;
907 }
908
__vxlan_fdb_free(struct vxlan_fdb * f)909 static void __vxlan_fdb_free(struct vxlan_fdb *f)
910 {
911 struct vxlan_rdst *rd, *nd;
912 struct nexthop *nh;
913
914 nh = rcu_dereference_raw(f->nh);
915 if (nh) {
916 rcu_assign_pointer(f->nh, NULL);
917 rcu_assign_pointer(f->vdev, NULL);
918 nexthop_put(nh);
919 }
920
921 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
922 dst_cache_destroy(&rd->dst_cache);
923 kfree(rd);
924 }
925 kfree(f);
926 }
927
vxlan_fdb_free(struct rcu_head * head)928 static void vxlan_fdb_free(struct rcu_head *head)
929 {
930 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
931
932 __vxlan_fdb_free(f);
933 }
934
vxlan_fdb_destroy(struct vxlan_dev * vxlan,struct vxlan_fdb * f,bool do_notify,bool swdev_notify)935 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
936 bool do_notify, bool swdev_notify)
937 {
938 struct vxlan_rdst *rd;
939
940 netdev_dbg(vxlan->dev, "delete %pM\n", f->key.eth_addr);
941
942 --vxlan->addrcnt;
943 if (do_notify) {
944 if (rcu_access_pointer(f->nh))
945 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
946 swdev_notify, NULL);
947 else
948 list_for_each_entry(rd, &f->remotes, list)
949 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
950 swdev_notify, NULL);
951 }
952
953 hlist_del_init_rcu(&f->fdb_node);
954 rhashtable_remove_fast(&vxlan->fdb_hash_tbl, &f->rhnode,
955 vxlan_fdb_rht_params);
956 list_del_rcu(&f->nh_list);
957 call_rcu(&f->rcu, vxlan_fdb_free);
958 }
959
vxlan_dst_free(struct rcu_head * head)960 static void vxlan_dst_free(struct rcu_head *head)
961 {
962 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
963
964 dst_cache_destroy(&rd->dst_cache);
965 kfree(rd);
966 }
967
vxlan_fdb_update_existing(struct vxlan_dev * vxlan,union vxlan_addr * ip,__u16 state,__u16 flags,__be16 port,__be32 vni,__u32 ifindex,__u16 ndm_flags,struct vxlan_fdb * f,u32 nhid,bool swdev_notify,struct netlink_ext_ack * extack)968 static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
969 union vxlan_addr *ip,
970 __u16 state, __u16 flags,
971 __be16 port, __be32 vni,
972 __u32 ifindex, __u16 ndm_flags,
973 struct vxlan_fdb *f, u32 nhid,
974 bool swdev_notify,
975 struct netlink_ext_ack *extack)
976 {
977 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
978 struct vxlan_rdst *rd = NULL;
979 struct vxlan_rdst oldrd;
980 int notify = 0;
981 int rc = 0;
982 int err;
983
984 if (nhid && !rcu_access_pointer(f->nh)) {
985 NL_SET_ERR_MSG(extack,
986 "Cannot replace an existing non nexthop fdb with a nexthop");
987 return -EOPNOTSUPP;
988 }
989
990 if (nhid && (flags & NLM_F_APPEND)) {
991 NL_SET_ERR_MSG(extack,
992 "Cannot append to a nexthop fdb");
993 return -EOPNOTSUPP;
994 }
995
996 /* Do not allow an externally learned entry to take over an entry added
997 * by the user.
998 */
999 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1000 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1001 if (f->state != state) {
1002 f->state = state;
1003 notify = 1;
1004 }
1005 if (f->flags != fdb_flags) {
1006 f->flags = fdb_flags;
1007 notify = 1;
1008 }
1009 }
1010
1011 if ((flags & NLM_F_REPLACE)) {
1012 /* Only change unicasts */
1013 if (!(is_multicast_ether_addr(f->key.eth_addr) ||
1014 is_zero_ether_addr(f->key.eth_addr))) {
1015 if (nhid) {
1016 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1017 if (rc < 0)
1018 return rc;
1019 } else {
1020 rc = vxlan_fdb_replace(f, ip, port, vni,
1021 ifindex, &oldrd);
1022 }
1023 notify |= rc;
1024 } else {
1025 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1026 return -EOPNOTSUPP;
1027 }
1028 }
1029 if ((flags & NLM_F_APPEND) &&
1030 (is_multicast_ether_addr(f->key.eth_addr) ||
1031 is_zero_ether_addr(f->key.eth_addr))) {
1032 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1033
1034 if (rc < 0)
1035 return rc;
1036 notify |= rc;
1037 }
1038
1039 if (ndm_flags & NTF_USE)
1040 WRITE_ONCE(f->updated, jiffies);
1041
1042 if (notify) {
1043 if (rd == NULL)
1044 rd = first_remote_rtnl(f);
1045
1046 WRITE_ONCE(f->updated, jiffies);
1047 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1048 swdev_notify, extack);
1049 if (err)
1050 goto err_notify;
1051 }
1052
1053 return 0;
1054
1055 err_notify:
1056 if (nhid)
1057 return err;
1058 if ((flags & NLM_F_REPLACE) && rc)
1059 *rd = oldrd;
1060 else if ((flags & NLM_F_APPEND) && rc) {
1061 list_del_rcu(&rd->list);
1062 call_rcu(&rd->rcu, vxlan_dst_free);
1063 }
1064 return err;
1065 }
1066
vxlan_fdb_update_create(struct vxlan_dev * vxlan,const u8 * mac,union vxlan_addr * ip,__u16 state,__u16 flags,__be16 port,__be32 src_vni,__be32 vni,__u32 ifindex,__u16 ndm_flags,u32 nhid,bool swdev_notify,struct netlink_ext_ack * extack)1067 static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1068 const u8 *mac, union vxlan_addr *ip,
1069 __u16 state, __u16 flags,
1070 __be16 port, __be32 src_vni, __be32 vni,
1071 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1072 bool swdev_notify,
1073 struct netlink_ext_ack *extack)
1074 {
1075 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1076 struct vxlan_fdb *f;
1077 int rc;
1078
1079 /* Disallow replace to add a multicast entry */
1080 if ((flags & NLM_F_REPLACE) &&
1081 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1082 return -EOPNOTSUPP;
1083
1084 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1085 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1086 vni, ifindex, fdb_flags, nhid, &f, extack);
1087 if (rc < 0)
1088 return rc;
1089
1090 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1091 swdev_notify, extack);
1092 if (rc)
1093 goto err_notify;
1094
1095 return 0;
1096
1097 err_notify:
1098 vxlan_fdb_destroy(vxlan, f, false, false);
1099 return rc;
1100 }
1101
1102 /* Add new entry to forwarding table -- assumes lock held */
vxlan_fdb_update(struct vxlan_dev * vxlan,const u8 * mac,union vxlan_addr * ip,__u16 state,__u16 flags,__be16 port,__be32 src_vni,__be32 vni,__u32 ifindex,__u16 ndm_flags,u32 nhid,bool swdev_notify,struct netlink_ext_ack * extack)1103 int vxlan_fdb_update(struct vxlan_dev *vxlan,
1104 const u8 *mac, union vxlan_addr *ip,
1105 __u16 state, __u16 flags,
1106 __be16 port, __be32 src_vni, __be32 vni,
1107 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1108 bool swdev_notify,
1109 struct netlink_ext_ack *extack)
1110 {
1111 struct vxlan_fdb *f;
1112
1113 f = vxlan_find_mac(vxlan, mac, src_vni);
1114 if (f) {
1115 if (flags & NLM_F_EXCL) {
1116 netdev_dbg(vxlan->dev,
1117 "lost race to create %pM\n", mac);
1118 return -EEXIST;
1119 }
1120
1121 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1122 vni, ifindex, ndm_flags, f,
1123 nhid, swdev_notify, extack);
1124 } else {
1125 if (!(flags & NLM_F_CREATE))
1126 return -ENOENT;
1127
1128 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1129 port, src_vni, vni, ifindex,
1130 ndm_flags, nhid, swdev_notify,
1131 extack);
1132 }
1133 }
1134
vxlan_fdb_dst_destroy(struct vxlan_dev * vxlan,struct vxlan_fdb * f,struct vxlan_rdst * rd,bool swdev_notify)1135 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1136 struct vxlan_rdst *rd, bool swdev_notify)
1137 {
1138 list_del_rcu(&rd->list);
1139 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1140 call_rcu(&rd->rcu, vxlan_dst_free);
1141 }
1142
vxlan_fdb_parse(struct nlattr * tb[],struct vxlan_dev * vxlan,union vxlan_addr * ip,__be16 * port,__be32 * src_vni,__be32 * vni,u32 * ifindex,u32 * nhid,struct netlink_ext_ack * extack)1143 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1144 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1145 __be32 *vni, u32 *ifindex, u32 *nhid,
1146 struct netlink_ext_ack *extack)
1147 {
1148 struct net *net = dev_net(vxlan->dev);
1149 int err;
1150
1151 if (tb[NDA_NH_ID] &&
1152 (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] || tb[NDA_PORT])) {
1153 NL_SET_ERR_MSG(extack, "DST, VNI, ifindex and port are mutually exclusive with NH_ID");
1154 return -EINVAL;
1155 }
1156
1157 if (tb[NDA_DST]) {
1158 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1159 if (err) {
1160 NL_SET_ERR_MSG(extack, "Unsupported address family");
1161 return err;
1162 }
1163 } else {
1164 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1165
1166 if (remote->sa.sa_family == AF_INET) {
1167 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1168 ip->sa.sa_family = AF_INET;
1169 #if IS_ENABLED(CONFIG_IPV6)
1170 } else {
1171 ip->sin6.sin6_addr = in6addr_any;
1172 ip->sa.sa_family = AF_INET6;
1173 #endif
1174 }
1175 }
1176
1177 if (tb[NDA_PORT]) {
1178 if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) {
1179 NL_SET_ERR_MSG(extack, "Invalid vxlan port");
1180 return -EINVAL;
1181 }
1182 *port = nla_get_be16(tb[NDA_PORT]);
1183 } else {
1184 *port = vxlan->cfg.dst_port;
1185 }
1186
1187 if (tb[NDA_VNI]) {
1188 if (nla_len(tb[NDA_VNI]) != sizeof(u32)) {
1189 NL_SET_ERR_MSG(extack, "Invalid vni");
1190 return -EINVAL;
1191 }
1192 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1193 } else {
1194 *vni = vxlan->default_dst.remote_vni;
1195 }
1196
1197 if (tb[NDA_SRC_VNI]) {
1198 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32)) {
1199 NL_SET_ERR_MSG(extack, "Invalid src vni");
1200 return -EINVAL;
1201 }
1202 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1203 } else {
1204 *src_vni = vxlan->default_dst.remote_vni;
1205 }
1206
1207 if (tb[NDA_IFINDEX]) {
1208 struct net_device *tdev;
1209
1210 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) {
1211 NL_SET_ERR_MSG(extack, "Invalid ifindex");
1212 return -EINVAL;
1213 }
1214 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1215 tdev = __dev_get_by_index(net, *ifindex);
1216 if (!tdev) {
1217 NL_SET_ERR_MSG(extack, "Device not found");
1218 return -EADDRNOTAVAIL;
1219 }
1220 } else {
1221 *ifindex = 0;
1222 }
1223
1224 *nhid = nla_get_u32_default(tb[NDA_NH_ID], 0);
1225
1226 return 0;
1227 }
1228
1229 /* Add static entry (via netlink) */
vxlan_fdb_add(struct ndmsg * ndm,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 vid,u16 flags,bool * notified,struct netlink_ext_ack * extack)1230 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1231 struct net_device *dev,
1232 const unsigned char *addr, u16 vid, u16 flags,
1233 bool *notified, struct netlink_ext_ack *extack)
1234 {
1235 struct vxlan_dev *vxlan = netdev_priv(dev);
1236 /* struct net *net = dev_net(vxlan->dev); */
1237 union vxlan_addr ip;
1238 __be16 port;
1239 __be32 src_vni, vni;
1240 u32 ifindex, nhid;
1241 int err;
1242
1243 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1244 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1245 ndm->ndm_state);
1246 return -EINVAL;
1247 }
1248
1249 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1250 return -EINVAL;
1251
1252 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1253 &nhid, extack);
1254 if (err)
1255 return err;
1256
1257 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1258 return -EAFNOSUPPORT;
1259
1260 spin_lock_bh(&vxlan->hash_lock);
1261 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1262 port, src_vni, vni, ifindex,
1263 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1264 nhid, true, extack);
1265 spin_unlock_bh(&vxlan->hash_lock);
1266
1267 if (!err)
1268 *notified = true;
1269
1270 return err;
1271 }
1272
__vxlan_fdb_delete(struct vxlan_dev * vxlan,const unsigned char * addr,union vxlan_addr ip,__be16 port,__be32 src_vni,__be32 vni,u32 ifindex,bool swdev_notify)1273 int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1274 const unsigned char *addr, union vxlan_addr ip,
1275 __be16 port, __be32 src_vni, __be32 vni,
1276 u32 ifindex, bool swdev_notify)
1277 {
1278 struct vxlan_rdst *rd = NULL;
1279 struct vxlan_fdb *f;
1280 int err = -ENOENT;
1281
1282 f = vxlan_find_mac(vxlan, addr, src_vni);
1283 if (!f)
1284 return err;
1285
1286 if (!vxlan_addr_any(&ip)) {
1287 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1288 if (!rd)
1289 goto out;
1290 }
1291
1292 /* remove a destination if it's not the only one on the list,
1293 * otherwise destroy the fdb entry
1294 */
1295 if (rd && !list_is_singular(&f->remotes)) {
1296 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1297 goto out;
1298 }
1299
1300 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1301
1302 out:
1303 return 0;
1304 }
1305
1306 /* Delete entry (via netlink) */
vxlan_fdb_delete(struct ndmsg * ndm,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 vid,bool * notified,struct netlink_ext_ack * extack)1307 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1308 struct net_device *dev,
1309 const unsigned char *addr, u16 vid, bool *notified,
1310 struct netlink_ext_ack *extack)
1311 {
1312 struct vxlan_dev *vxlan = netdev_priv(dev);
1313 union vxlan_addr ip;
1314 __be32 src_vni, vni;
1315 u32 ifindex, nhid;
1316 __be16 port;
1317 int err;
1318
1319 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1320 &nhid, extack);
1321 if (err)
1322 return err;
1323
1324 spin_lock_bh(&vxlan->hash_lock);
1325 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1326 true);
1327 spin_unlock_bh(&vxlan->hash_lock);
1328
1329 if (!err)
1330 *notified = true;
1331
1332 return err;
1333 }
1334
1335 /* Dump forwarding table */
vxlan_fdb_dump(struct sk_buff * skb,struct netlink_callback * cb,struct net_device * dev,struct net_device * filter_dev,int * idx)1336 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1337 struct net_device *dev,
1338 struct net_device *filter_dev, int *idx)
1339 {
1340 struct ndo_fdb_dump_context *ctx = (void *)cb->ctx;
1341 struct vxlan_dev *vxlan = netdev_priv(dev);
1342 struct vxlan_fdb *f;
1343 int err = 0;
1344
1345 rcu_read_lock();
1346 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) {
1347 struct vxlan_rdst *rd;
1348
1349 if (rcu_access_pointer(f->nh)) {
1350 if (*idx < ctx->fdb_idx)
1351 goto skip_nh;
1352 err = vxlan_fdb_info(skb, vxlan, f,
1353 NETLINK_CB(cb->skb).portid,
1354 cb->nlh->nlmsg_seq,
1355 RTM_NEWNEIGH, NLM_F_MULTI, NULL);
1356 if (err < 0) {
1357 rcu_read_unlock();
1358 goto out;
1359 }
1360 skip_nh:
1361 *idx += 1;
1362 continue;
1363 }
1364
1365 list_for_each_entry_rcu(rd, &f->remotes, list) {
1366 if (*idx < ctx->fdb_idx)
1367 goto skip;
1368
1369 err = vxlan_fdb_info(skb, vxlan, f,
1370 NETLINK_CB(cb->skb).portid,
1371 cb->nlh->nlmsg_seq,
1372 RTM_NEWNEIGH, NLM_F_MULTI, rd);
1373 if (err < 0) {
1374 rcu_read_unlock();
1375 goto out;
1376 }
1377 skip:
1378 *idx += 1;
1379 }
1380 }
1381 rcu_read_unlock();
1382 out:
1383 return err;
1384 }
1385
vxlan_fdb_get(struct sk_buff * skb,struct nlattr * tb[],struct net_device * dev,const unsigned char * addr,u16 vid,u32 portid,u32 seq,struct netlink_ext_ack * extack)1386 static int vxlan_fdb_get(struct sk_buff *skb,
1387 struct nlattr *tb[],
1388 struct net_device *dev,
1389 const unsigned char *addr,
1390 u16 vid, u32 portid, u32 seq,
1391 struct netlink_ext_ack *extack)
1392 {
1393 struct vxlan_dev *vxlan = netdev_priv(dev);
1394 struct vxlan_fdb *f;
1395 __be32 vni;
1396 int err;
1397
1398 if (tb[NDA_VNI])
1399 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1400 else
1401 vni = vxlan->default_dst.remote_vni;
1402
1403 rcu_read_lock();
1404
1405 f = vxlan_find_mac_rcu(vxlan, addr, vni);
1406 if (!f) {
1407 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1408 err = -ENOENT;
1409 goto errout;
1410 }
1411
1412 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1413 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1414 errout:
1415 rcu_read_unlock();
1416 return err;
1417 }
1418
1419 /* Watch incoming packets to learn mapping between Ethernet address
1420 * and Tunnel endpoint.
1421 */
vxlan_snoop(struct net_device * dev,union vxlan_addr * src_ip,const u8 * src_mac,u32 src_ifindex,__be32 vni)1422 static enum skb_drop_reason vxlan_snoop(struct net_device *dev,
1423 union vxlan_addr *src_ip,
1424 const u8 *src_mac, u32 src_ifindex,
1425 __be32 vni)
1426 {
1427 struct vxlan_dev *vxlan = netdev_priv(dev);
1428 struct vxlan_fdb *f;
1429 u32 ifindex = 0;
1430
1431 /* Ignore packets from invalid src-address */
1432 if (!is_valid_ether_addr(src_mac))
1433 return SKB_DROP_REASON_MAC_INVALID_SOURCE;
1434
1435 #if IS_ENABLED(CONFIG_IPV6)
1436 if (src_ip->sa.sa_family == AF_INET6 &&
1437 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1438 ifindex = src_ifindex;
1439 #endif
1440
1441 f = vxlan_find_mac_rcu(vxlan, src_mac, vni);
1442 if (likely(f)) {
1443 struct vxlan_rdst *rdst = first_remote_rcu(f);
1444 unsigned long now = jiffies;
1445
1446 if (READ_ONCE(f->updated) != now)
1447 WRITE_ONCE(f->updated, now);
1448
1449 /* Don't override an fdb with nexthop with a learnt entry */
1450 if (rcu_access_pointer(f->nh))
1451 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS;
1452
1453 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1454 rdst->remote_ifindex == ifindex))
1455 return SKB_NOT_DROPPED_YET;
1456
1457 /* Don't migrate static entries, drop packets */
1458 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1459 return SKB_DROP_REASON_VXLAN_ENTRY_EXISTS;
1460
1461 if (net_ratelimit())
1462 netdev_info(dev,
1463 "%pM migrated from %pIS to %pIS\n",
1464 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1465
1466 rdst->remote_ip = *src_ip;
1467 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1468 } else {
1469 /* learned new entry */
1470 spin_lock(&vxlan->hash_lock);
1471
1472 /* close off race between vxlan_flush and incoming packets */
1473 if (netif_running(dev))
1474 vxlan_fdb_update(vxlan, src_mac, src_ip,
1475 NUD_REACHABLE,
1476 NLM_F_EXCL|NLM_F_CREATE,
1477 vxlan->cfg.dst_port,
1478 vni,
1479 vxlan->default_dst.remote_vni,
1480 ifindex, NTF_SELF, 0, true, NULL);
1481 spin_unlock(&vxlan->hash_lock);
1482 }
1483
1484 return SKB_NOT_DROPPED_YET;
1485 }
1486
__vxlan_sock_release_prep(struct vxlan_sock * vs)1487 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1488 {
1489 ASSERT_RTNL();
1490
1491 if (!vs)
1492 return false;
1493 if (!refcount_dec_and_test(&vs->refcnt))
1494 return false;
1495
1496 hlist_del_rcu(&vs->hlist);
1497 udp_tunnel_notify_del_rx_port(vs->sock,
1498 (vs->flags & VXLAN_F_GPE) ?
1499 UDP_TUNNEL_TYPE_VXLAN_GPE :
1500 UDP_TUNNEL_TYPE_VXLAN);
1501
1502 return true;
1503 }
1504
vxlan_sock_release(struct vxlan_dev * vxlan)1505 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1506 {
1507 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1508 #if IS_ENABLED(CONFIG_IPV6)
1509 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1510
1511 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1512 #endif
1513
1514 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1515 synchronize_net();
1516
1517 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
1518 vxlan_vs_del_vnigrp(vxlan);
1519 else
1520 vxlan_vs_del_dev(vxlan);
1521
1522 if (__vxlan_sock_release_prep(sock4)) {
1523 udp_tunnel_sock_release(sock4->sock);
1524 kfree(sock4);
1525 }
1526
1527 #if IS_ENABLED(CONFIG_IPV6)
1528 if (__vxlan_sock_release_prep(sock6)) {
1529 udp_tunnel_sock_release(sock6->sock);
1530 kfree(sock6);
1531 }
1532 #endif
1533 }
1534
vxlan_remcsum(struct sk_buff * skb,u32 vxflags)1535 static enum skb_drop_reason vxlan_remcsum(struct sk_buff *skb, u32 vxflags)
1536 {
1537 const struct vxlanhdr *vh = vxlan_hdr(skb);
1538 enum skb_drop_reason reason;
1539 size_t start, offset;
1540
1541 if (!(vh->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1542 return SKB_NOT_DROPPED_YET;
1543
1544 start = vxlan_rco_start(vh->vx_vni);
1545 offset = start + vxlan_rco_offset(vh->vx_vni);
1546
1547 reason = pskb_may_pull_reason(skb, offset + sizeof(u16));
1548 if (reason)
1549 return reason;
1550
1551 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1552 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1553 return SKB_NOT_DROPPED_YET;
1554 }
1555
vxlan_parse_gbp_hdr(struct sk_buff * skb,u32 vxflags,struct vxlan_metadata * md)1556 static void vxlan_parse_gbp_hdr(struct sk_buff *skb, u32 vxflags,
1557 struct vxlan_metadata *md)
1558 {
1559 const struct vxlanhdr *vh = vxlan_hdr(skb);
1560 const struct vxlanhdr_gbp *gbp;
1561 struct metadata_dst *tun_dst;
1562
1563 gbp = (const struct vxlanhdr_gbp *)vh;
1564
1565 if (!(vh->vx_flags & VXLAN_HF_GBP))
1566 return;
1567
1568 md->gbp = ntohs(gbp->policy_id);
1569
1570 tun_dst = (struct metadata_dst *)skb_dst(skb);
1571 if (tun_dst) {
1572 __set_bit(IP_TUNNEL_VXLAN_OPT_BIT,
1573 tun_dst->u.tun_info.key.tun_flags);
1574 tun_dst->u.tun_info.options_len = sizeof(*md);
1575 }
1576 if (gbp->dont_learn)
1577 md->gbp |= VXLAN_GBP_DONT_LEARN;
1578
1579 if (gbp->policy_applied)
1580 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1581
1582 /* In flow-based mode, GBP is carried in dst_metadata */
1583 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1584 skb->mark = md->gbp;
1585 }
1586
vxlan_set_mac(struct vxlan_dev * vxlan,struct vxlan_sock * vs,struct sk_buff * skb,__be32 vni)1587 static enum skb_drop_reason vxlan_set_mac(struct vxlan_dev *vxlan,
1588 struct vxlan_sock *vs,
1589 struct sk_buff *skb, __be32 vni)
1590 {
1591 union vxlan_addr saddr;
1592 u32 ifindex = skb->dev->ifindex;
1593
1594 skb_reset_mac_header(skb);
1595 skb->protocol = eth_type_trans(skb, vxlan->dev);
1596 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1597
1598 /* Ignore packet loops (and multicast echo) */
1599 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1600 return SKB_DROP_REASON_LOCAL_MAC;
1601
1602 /* Get address from the outer IP header */
1603 if (vxlan_get_sk_family(vs) == AF_INET) {
1604 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1605 saddr.sa.sa_family = AF_INET;
1606 #if IS_ENABLED(CONFIG_IPV6)
1607 } else {
1608 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1609 saddr.sa.sa_family = AF_INET6;
1610 #endif
1611 }
1612
1613 if (!(vxlan->cfg.flags & VXLAN_F_LEARN))
1614 return SKB_NOT_DROPPED_YET;
1615
1616 return vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source,
1617 ifindex, vni);
1618 }
1619
vxlan_ecn_decapsulate(struct vxlan_sock * vs,void * oiph,struct sk_buff * skb)1620 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1621 struct sk_buff *skb)
1622 {
1623 int err = 0;
1624
1625 if (vxlan_get_sk_family(vs) == AF_INET)
1626 err = IP_ECN_decapsulate(oiph, skb);
1627 #if IS_ENABLED(CONFIG_IPV6)
1628 else
1629 err = IP6_ECN_decapsulate(oiph, skb);
1630 #endif
1631
1632 if (unlikely(err) && log_ecn_error) {
1633 if (vxlan_get_sk_family(vs) == AF_INET)
1634 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1635 &((struct iphdr *)oiph)->saddr,
1636 ((struct iphdr *)oiph)->tos);
1637 else
1638 net_info_ratelimited("non-ECT from %pI6\n",
1639 &((struct ipv6hdr *)oiph)->saddr);
1640 }
1641 return err <= 1;
1642 }
1643
vxlan_rcv(struct sock * sk,struct sk_buff * skb)1644 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1645 {
1646 struct vxlan_vni_node *vninode = NULL;
1647 const struct vxlanhdr *vh;
1648 struct vxlan_dev *vxlan;
1649 struct vxlan_sock *vs;
1650 struct vxlan_metadata _md;
1651 struct vxlan_metadata *md = &_md;
1652 __be16 protocol = htons(ETH_P_TEB);
1653 enum skb_drop_reason reason;
1654 bool raw_proto = false;
1655 void *oiph;
1656 __be32 vni = 0;
1657 int nh;
1658
1659 /* Need UDP and VXLAN header to be present */
1660 reason = pskb_may_pull_reason(skb, VXLAN_HLEN);
1661 if (reason)
1662 goto drop;
1663
1664 vh = vxlan_hdr(skb);
1665 /* VNI flag always required to be set */
1666 if (!(vh->vx_flags & VXLAN_HF_VNI)) {
1667 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1668 ntohl(vh->vx_flags), ntohl(vh->vx_vni));
1669 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR;
1670 /* Return non vxlan pkt */
1671 goto drop;
1672 }
1673
1674 vs = rcu_dereference_sk_user_data(sk);
1675 if (!vs)
1676 goto drop;
1677
1678 vni = vxlan_vni(vh->vx_vni);
1679
1680 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, &vninode);
1681 if (!vxlan) {
1682 reason = SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND;
1683 goto drop;
1684 }
1685
1686 if (vh->vx_flags & vxlan->cfg.reserved_bits.vx_flags ||
1687 vh->vx_vni & vxlan->cfg.reserved_bits.vx_vni) {
1688 /* If the header uses bits besides those enabled by the
1689 * netdevice configuration, treat this as a malformed packet.
1690 * This behavior diverges from VXLAN RFC (RFC7348) which
1691 * stipulates that bits in reserved in reserved fields are to be
1692 * ignored. The approach here maintains compatibility with
1693 * previous stack code, and also is more robust and provides a
1694 * little more security in adding extensions to VXLAN.
1695 */
1696 reason = SKB_DROP_REASON_VXLAN_INVALID_HDR;
1697 DEV_STATS_INC(vxlan->dev, rx_frame_errors);
1698 DEV_STATS_INC(vxlan->dev, rx_errors);
1699 vxlan_vnifilter_count(vxlan, vni, vninode,
1700 VXLAN_VNI_STATS_RX_ERRORS, 0);
1701 goto drop;
1702 }
1703
1704 if (vxlan->cfg.flags & VXLAN_F_GPE) {
1705 if (!vxlan_parse_gpe_proto(vh, &protocol))
1706 goto drop;
1707 raw_proto = true;
1708 }
1709
1710 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1711 !net_eq(vxlan->net, dev_net(vxlan->dev)))) {
1712 reason = SKB_DROP_REASON_NOMEM;
1713 goto drop;
1714 }
1715
1716 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_RX) {
1717 reason = vxlan_remcsum(skb, vxlan->cfg.flags);
1718 if (unlikely(reason))
1719 goto drop;
1720 }
1721
1722 if (vxlan_collect_metadata(vs)) {
1723 IP_TUNNEL_DECLARE_FLAGS(flags) = { };
1724 struct metadata_dst *tun_dst;
1725
1726 __set_bit(IP_TUNNEL_KEY_BIT, flags);
1727 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), flags,
1728 key32_to_tunnel_id(vni), sizeof(*md));
1729
1730 if (!tun_dst) {
1731 reason = SKB_DROP_REASON_NOMEM;
1732 goto drop;
1733 }
1734
1735 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1736
1737 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1738 } else {
1739 memset(md, 0, sizeof(*md));
1740 }
1741
1742 if (vxlan->cfg.flags & VXLAN_F_GBP)
1743 vxlan_parse_gbp_hdr(skb, vxlan->cfg.flags, md);
1744 /* Note that GBP and GPE can never be active together. This is
1745 * ensured in vxlan_dev_configure.
1746 */
1747
1748 if (!raw_proto) {
1749 reason = vxlan_set_mac(vxlan, vs, skb, vni);
1750 if (reason)
1751 goto drop;
1752 } else {
1753 skb_reset_mac_header(skb);
1754 skb->dev = vxlan->dev;
1755 skb->pkt_type = PACKET_HOST;
1756 }
1757
1758 /* Save offset of outer header relative to skb->head,
1759 * because we are going to reset the network header to the inner header
1760 * and might change skb->head.
1761 */
1762 nh = skb_network_header(skb) - skb->head;
1763
1764 skb_reset_network_header(skb);
1765
1766 reason = pskb_inet_may_pull_reason(skb);
1767 if (reason) {
1768 DEV_STATS_INC(vxlan->dev, rx_length_errors);
1769 DEV_STATS_INC(vxlan->dev, rx_errors);
1770 vxlan_vnifilter_count(vxlan, vni, vninode,
1771 VXLAN_VNI_STATS_RX_ERRORS, 0);
1772 goto drop;
1773 }
1774
1775 /* Get the outer header. */
1776 oiph = skb->head + nh;
1777
1778 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1779 reason = SKB_DROP_REASON_IP_TUNNEL_ECN;
1780 DEV_STATS_INC(vxlan->dev, rx_frame_errors);
1781 DEV_STATS_INC(vxlan->dev, rx_errors);
1782 vxlan_vnifilter_count(vxlan, vni, vninode,
1783 VXLAN_VNI_STATS_RX_ERRORS, 0);
1784 goto drop;
1785 }
1786
1787 rcu_read_lock();
1788
1789 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1790 rcu_read_unlock();
1791 dev_dstats_rx_dropped(vxlan->dev);
1792 vxlan_vnifilter_count(vxlan, vni, vninode,
1793 VXLAN_VNI_STATS_RX_DROPS, 0);
1794 reason = SKB_DROP_REASON_DEV_READY;
1795 goto drop;
1796 }
1797
1798 dev_dstats_rx_add(vxlan->dev, skb->len);
1799 vxlan_vnifilter_count(vxlan, vni, vninode, VXLAN_VNI_STATS_RX, skb->len);
1800 gro_cells_receive(&vxlan->gro_cells, skb);
1801
1802 rcu_read_unlock();
1803
1804 return 0;
1805
1806 drop:
1807 reason = reason ?: SKB_DROP_REASON_NOT_SPECIFIED;
1808 /* Consume bad packet */
1809 kfree_skb_reason(skb, reason);
1810 return 0;
1811 }
1812
vxlan_err_lookup(struct sock * sk,struct sk_buff * skb)1813 static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1814 {
1815 struct vxlan_dev *vxlan;
1816 struct vxlan_sock *vs;
1817 struct vxlanhdr *hdr;
1818 __be32 vni;
1819
1820 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1821 return -EINVAL;
1822
1823 hdr = vxlan_hdr(skb);
1824
1825 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1826 return -EINVAL;
1827
1828 vs = rcu_dereference_sk_user_data(sk);
1829 if (!vs)
1830 return -ENOENT;
1831
1832 vni = vxlan_vni(hdr->vx_vni);
1833 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni, NULL);
1834 if (!vxlan)
1835 return -ENOENT;
1836
1837 return 0;
1838 }
1839
arp_reduce(struct net_device * dev,struct sk_buff * skb,__be32 vni)1840 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1841 {
1842 struct vxlan_dev *vxlan = netdev_priv(dev);
1843 struct arphdr *parp;
1844 u8 *arpptr, *sha;
1845 __be32 sip, tip;
1846 struct neighbour *n;
1847
1848 if (dev->flags & IFF_NOARP)
1849 goto out;
1850
1851 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1852 dev_dstats_tx_dropped(dev);
1853 vxlan_vnifilter_count(vxlan, vni, NULL,
1854 VXLAN_VNI_STATS_TX_DROPS, 0);
1855 goto out;
1856 }
1857 parp = arp_hdr(skb);
1858
1859 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1860 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1861 parp->ar_pro != htons(ETH_P_IP) ||
1862 parp->ar_op != htons(ARPOP_REQUEST) ||
1863 parp->ar_hln != dev->addr_len ||
1864 parp->ar_pln != 4)
1865 goto out;
1866 arpptr = (u8 *)parp + sizeof(struct arphdr);
1867 sha = arpptr;
1868 arpptr += dev->addr_len; /* sha */
1869 memcpy(&sip, arpptr, sizeof(sip));
1870 arpptr += sizeof(sip);
1871 arpptr += dev->addr_len; /* tha */
1872 memcpy(&tip, arpptr, sizeof(tip));
1873
1874 if (ipv4_is_loopback(tip) ||
1875 ipv4_is_multicast(tip))
1876 goto out;
1877
1878 n = neigh_lookup(&arp_tbl, &tip, dev);
1879
1880 if (n) {
1881 struct vxlan_rdst *rdst = NULL;
1882 struct vxlan_fdb *f;
1883 struct sk_buff *reply;
1884
1885 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
1886 neigh_release(n);
1887 goto out;
1888 }
1889
1890 rcu_read_lock();
1891 f = vxlan_find_mac_tx(vxlan, n->ha, vni);
1892 if (f)
1893 rdst = first_remote_rcu(f);
1894 if (rdst && vxlan_addr_any(&rdst->remote_ip)) {
1895 /* bridge-local neighbor */
1896 neigh_release(n);
1897 rcu_read_unlock();
1898 goto out;
1899 }
1900 rcu_read_unlock();
1901
1902 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1903 n->ha, sha);
1904
1905 neigh_release(n);
1906
1907 if (reply == NULL)
1908 goto out;
1909
1910 skb_reset_mac_header(reply);
1911 __skb_pull(reply, skb_network_offset(reply));
1912 reply->ip_summed = CHECKSUM_UNNECESSARY;
1913 reply->pkt_type = PACKET_HOST;
1914
1915 if (netif_rx(reply) == NET_RX_DROP) {
1916 dev_dstats_rx_dropped(dev);
1917 vxlan_vnifilter_count(vxlan, vni, NULL,
1918 VXLAN_VNI_STATS_RX_DROPS, 0);
1919 }
1920
1921 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1922 union vxlan_addr ipa = {
1923 .sin.sin_addr.s_addr = tip,
1924 .sin.sin_family = AF_INET,
1925 };
1926
1927 vxlan_ip_miss(dev, &ipa);
1928 }
1929 out:
1930 consume_skb(skb);
1931 return NETDEV_TX_OK;
1932 }
1933
1934 #if IS_ENABLED(CONFIG_IPV6)
vxlan_na_create(struct sk_buff * request,struct neighbour * n,bool isrouter)1935 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1936 struct neighbour *n, bool isrouter)
1937 {
1938 struct net_device *dev = request->dev;
1939 struct sk_buff *reply;
1940 struct nd_msg *ns, *na;
1941 struct ipv6hdr *pip6;
1942 u8 *daddr;
1943 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1944 int ns_olen;
1945 int i, len;
1946
1947 if (dev == NULL || !pskb_may_pull(request, request->len))
1948 return NULL;
1949
1950 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1951 sizeof(*na) + na_olen + dev->needed_tailroom;
1952 reply = alloc_skb(len, GFP_ATOMIC);
1953 if (reply == NULL)
1954 return NULL;
1955
1956 reply->protocol = htons(ETH_P_IPV6);
1957 reply->dev = dev;
1958 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1959 skb_push(reply, sizeof(struct ethhdr));
1960 skb_reset_mac_header(reply);
1961
1962 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1963
1964 daddr = eth_hdr(request)->h_source;
1965 ns_olen = request->len - skb_network_offset(request) -
1966 sizeof(struct ipv6hdr) - sizeof(*ns);
1967 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1968 if (!ns->opt[i + 1]) {
1969 kfree_skb(reply);
1970 return NULL;
1971 }
1972 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1973 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1974 break;
1975 }
1976 }
1977
1978 /* Ethernet header */
1979 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1980 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1981 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1982 reply->protocol = htons(ETH_P_IPV6);
1983
1984 skb_pull(reply, sizeof(struct ethhdr));
1985 skb_reset_network_header(reply);
1986 skb_put(reply, sizeof(struct ipv6hdr));
1987
1988 /* IPv6 header */
1989
1990 pip6 = ipv6_hdr(reply);
1991 memset(pip6, 0, sizeof(struct ipv6hdr));
1992 pip6->version = 6;
1993 pip6->priority = ipv6_hdr(request)->priority;
1994 pip6->nexthdr = IPPROTO_ICMPV6;
1995 pip6->hop_limit = 255;
1996 pip6->daddr = ipv6_hdr(request)->saddr;
1997 pip6->saddr = *(struct in6_addr *)n->primary_key;
1998
1999 skb_pull(reply, sizeof(struct ipv6hdr));
2000 skb_reset_transport_header(reply);
2001
2002 /* Neighbor Advertisement */
2003 na = skb_put_zero(reply, sizeof(*na) + na_olen);
2004 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2005 na->icmph.icmp6_router = isrouter;
2006 na->icmph.icmp6_override = 1;
2007 na->icmph.icmp6_solicited = 1;
2008 na->target = ns->target;
2009 ether_addr_copy(&na->opt[2], n->ha);
2010 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2011 na->opt[1] = na_olen >> 3;
2012
2013 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2014 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2015 csum_partial(na, sizeof(*na)+na_olen, 0));
2016
2017 pip6->payload_len = htons(sizeof(*na)+na_olen);
2018
2019 skb_push(reply, sizeof(struct ipv6hdr));
2020
2021 reply->ip_summed = CHECKSUM_UNNECESSARY;
2022
2023 return reply;
2024 }
2025
neigh_reduce(struct net_device * dev,struct sk_buff * skb,__be32 vni)2026 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2027 {
2028 struct vxlan_dev *vxlan = netdev_priv(dev);
2029 const struct in6_addr *daddr;
2030 const struct ipv6hdr *iphdr;
2031 struct inet6_dev *in6_dev;
2032 struct neighbour *n;
2033 struct nd_msg *msg;
2034
2035 rcu_read_lock();
2036 in6_dev = __in6_dev_get(dev);
2037 if (!in6_dev)
2038 goto out;
2039
2040 iphdr = ipv6_hdr(skb);
2041 daddr = &iphdr->daddr;
2042 msg = (struct nd_msg *)(iphdr + 1);
2043
2044 if (ipv6_addr_loopback(daddr) ||
2045 ipv6_addr_is_multicast(&msg->target))
2046 goto out;
2047
2048 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2049
2050 if (n) {
2051 struct vxlan_rdst *rdst = NULL;
2052 struct vxlan_fdb *f;
2053 struct sk_buff *reply;
2054
2055 if (!(READ_ONCE(n->nud_state) & NUD_CONNECTED)) {
2056 neigh_release(n);
2057 goto out;
2058 }
2059
2060 f = vxlan_find_mac_tx(vxlan, n->ha, vni);
2061 if (f)
2062 rdst = first_remote_rcu(f);
2063 if (rdst && vxlan_addr_any(&rdst->remote_ip)) {
2064 /* bridge-local neighbor */
2065 neigh_release(n);
2066 goto out;
2067 }
2068
2069 reply = vxlan_na_create(skb, n,
2070 !!(f ? f->flags & NTF_ROUTER : 0));
2071
2072 neigh_release(n);
2073
2074 if (reply == NULL)
2075 goto out;
2076
2077 if (netif_rx(reply) == NET_RX_DROP) {
2078 dev_dstats_rx_dropped(dev);
2079 vxlan_vnifilter_count(vxlan, vni, NULL,
2080 VXLAN_VNI_STATS_RX_DROPS, 0);
2081 }
2082 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2083 union vxlan_addr ipa = {
2084 .sin6.sin6_addr = msg->target,
2085 .sin6.sin6_family = AF_INET6,
2086 };
2087
2088 vxlan_ip_miss(dev, &ipa);
2089 }
2090
2091 out:
2092 rcu_read_unlock();
2093 consume_skb(skb);
2094 return NETDEV_TX_OK;
2095 }
2096 #endif
2097
route_shortcircuit(struct net_device * dev,struct sk_buff * skb)2098 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2099 {
2100 struct vxlan_dev *vxlan = netdev_priv(dev);
2101 struct neighbour *n;
2102
2103 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2104 return false;
2105
2106 n = NULL;
2107 switch (ntohs(eth_hdr(skb)->h_proto)) {
2108 case ETH_P_IP:
2109 {
2110 struct iphdr *pip;
2111
2112 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2113 return false;
2114 pip = ip_hdr(skb);
2115 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2116 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2117 union vxlan_addr ipa = {
2118 .sin.sin_addr.s_addr = pip->daddr,
2119 .sin.sin_family = AF_INET,
2120 };
2121
2122 vxlan_ip_miss(dev, &ipa);
2123 return false;
2124 }
2125
2126 break;
2127 }
2128 #if IS_ENABLED(CONFIG_IPV6)
2129 case ETH_P_IPV6:
2130 {
2131 struct ipv6hdr *pip6;
2132
2133 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2134 return false;
2135 pip6 = ipv6_hdr(skb);
2136 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2137 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2138 union vxlan_addr ipa = {
2139 .sin6.sin6_addr = pip6->daddr,
2140 .sin6.sin6_family = AF_INET6,
2141 };
2142
2143 vxlan_ip_miss(dev, &ipa);
2144 return false;
2145 }
2146
2147 break;
2148 }
2149 #endif
2150 default:
2151 return false;
2152 }
2153
2154 if (n) {
2155 bool diff;
2156
2157 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2158 if (diff) {
2159 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2160 dev->addr_len);
2161 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2162 }
2163 neigh_release(n);
2164 return diff;
2165 }
2166
2167 return false;
2168 }
2169
vxlan_build_gpe_hdr(struct vxlanhdr * vxh,__be16 protocol)2170 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, __be16 protocol)
2171 {
2172 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2173
2174 gpe->np_applied = 1;
2175 gpe->next_protocol = tun_p_from_eth_p(protocol);
2176 if (!gpe->next_protocol)
2177 return -EPFNOSUPPORT;
2178 return 0;
2179 }
2180
vxlan_build_skb(struct sk_buff * skb,struct dst_entry * dst,int iphdr_len,__be32 vni,struct vxlan_metadata * md,u32 vxflags,bool udp_sum)2181 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2182 int iphdr_len, __be32 vni,
2183 struct vxlan_metadata *md, u32 vxflags,
2184 bool udp_sum)
2185 {
2186 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2187 __be16 inner_protocol = htons(ETH_P_TEB);
2188 struct vxlanhdr *vxh;
2189 bool double_encap;
2190 int min_headroom;
2191 int err;
2192
2193 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2194 skb->ip_summed == CHECKSUM_PARTIAL) {
2195 int csum_start = skb_checksum_start_offset(skb);
2196
2197 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2198 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2199 (skb->csum_offset == offsetof(struct udphdr, check) ||
2200 skb->csum_offset == offsetof(struct tcphdr, check)))
2201 type |= SKB_GSO_TUNNEL_REMCSUM;
2202 }
2203
2204 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2205 + VXLAN_HLEN + iphdr_len;
2206
2207 /* Need space for new headers (invalidates iph ptr) */
2208 err = skb_cow_head(skb, min_headroom);
2209 if (unlikely(err))
2210 return err;
2211
2212 double_encap = udp_tunnel_handle_partial(skb);
2213 err = iptunnel_handle_offloads(skb, type);
2214 if (err)
2215 return err;
2216
2217 vxh = __skb_push(skb, sizeof(*vxh));
2218 vxh->vx_flags = VXLAN_HF_VNI;
2219 vxh->vx_vni = vxlan_vni_field(vni);
2220
2221 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2222 unsigned int start;
2223
2224 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2225 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2226 vxh->vx_flags |= VXLAN_HF_RCO;
2227
2228 if (!skb_is_gso(skb)) {
2229 skb->ip_summed = CHECKSUM_NONE;
2230 skb->encapsulation = 0;
2231 }
2232 }
2233
2234 if (vxflags & VXLAN_F_GBP)
2235 vxlan_build_gbp_hdr(vxh, md);
2236 if (vxflags & VXLAN_F_GPE) {
2237 err = vxlan_build_gpe_hdr(vxh, skb->protocol);
2238 if (err < 0)
2239 return err;
2240 inner_protocol = skb->protocol;
2241 }
2242
2243 udp_tunnel_set_inner_protocol(skb, double_encap, inner_protocol);
2244 return 0;
2245 }
2246
2247 /* Bypass encapsulation if the destination is local */
vxlan_encap_bypass(struct sk_buff * skb,struct vxlan_dev * src_vxlan,struct vxlan_dev * dst_vxlan,__be32 vni,bool snoop)2248 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2249 struct vxlan_dev *dst_vxlan, __be32 vni,
2250 bool snoop)
2251 {
2252 union vxlan_addr loopback;
2253 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2254 unsigned int len = skb->len;
2255 struct net_device *dev;
2256
2257 skb->pkt_type = PACKET_HOST;
2258 skb->encapsulation = 0;
2259 skb->dev = dst_vxlan->dev;
2260 __skb_pull(skb, skb_network_offset(skb));
2261
2262 if (remote_ip->sa.sa_family == AF_INET) {
2263 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2264 loopback.sa.sa_family = AF_INET;
2265 #if IS_ENABLED(CONFIG_IPV6)
2266 } else {
2267 loopback.sin6.sin6_addr = in6addr_loopback;
2268 loopback.sa.sa_family = AF_INET6;
2269 #endif
2270 }
2271
2272 rcu_read_lock();
2273 dev = skb->dev;
2274 if (unlikely(!(dev->flags & IFF_UP))) {
2275 kfree_skb_reason(skb, SKB_DROP_REASON_DEV_READY);
2276 goto drop;
2277 }
2278
2279 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2280 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2281
2282 dev_dstats_tx_add(src_vxlan->dev, len);
2283 vxlan_vnifilter_count(src_vxlan, vni, NULL, VXLAN_VNI_STATS_TX, len);
2284
2285 if (__netif_rx(skb) == NET_RX_SUCCESS) {
2286 dev_dstats_rx_add(dst_vxlan->dev, len);
2287 vxlan_vnifilter_count(dst_vxlan, vni, NULL, VXLAN_VNI_STATS_RX,
2288 len);
2289 } else {
2290 drop:
2291 dev_dstats_rx_dropped(dev);
2292 vxlan_vnifilter_count(dst_vxlan, vni, NULL,
2293 VXLAN_VNI_STATS_RX_DROPS, 0);
2294 }
2295 rcu_read_unlock();
2296 }
2297
encap_bypass_if_local(struct sk_buff * skb,struct net_device * dev,struct vxlan_dev * vxlan,int addr_family,__be16 dst_port,int dst_ifindex,__be32 vni,struct dst_entry * dst,u32 rt_flags)2298 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2299 struct vxlan_dev *vxlan,
2300 int addr_family,
2301 __be16 dst_port, int dst_ifindex, __be32 vni,
2302 struct dst_entry *dst,
2303 u32 rt_flags)
2304 {
2305 #if IS_ENABLED(CONFIG_IPV6)
2306 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2307 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2308 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2309 */
2310 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2311 #endif
2312 /* Bypass encapsulation if the destination is local */
2313 if (rt_flags & RTCF_LOCAL &&
2314 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST)) &&
2315 vxlan->cfg.flags & VXLAN_F_LOCALBYPASS) {
2316 struct vxlan_dev *dst_vxlan;
2317
2318 dst_release(dst);
2319 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2320 addr_family, dst_port,
2321 vxlan->cfg.flags);
2322 if (!dst_vxlan) {
2323 DEV_STATS_INC(dev, tx_errors);
2324 vxlan_vnifilter_count(vxlan, vni, NULL,
2325 VXLAN_VNI_STATS_TX_ERRORS, 0);
2326 kfree_skb_reason(skb, SKB_DROP_REASON_VXLAN_VNI_NOT_FOUND);
2327
2328 return -ENOENT;
2329 }
2330 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2331 return 1;
2332 }
2333
2334 return 0;
2335 }
2336
vxlan_xmit_one(struct sk_buff * skb,struct net_device * dev,__be32 default_vni,struct vxlan_rdst * rdst,bool did_rsc)2337 void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2338 __be32 default_vni, struct vxlan_rdst *rdst, bool did_rsc)
2339 {
2340 struct dst_cache *dst_cache;
2341 struct ip_tunnel_info *info;
2342 struct ip_tunnel_key *pkey;
2343 struct ip_tunnel_key key;
2344 struct vxlan_dev *vxlan = netdev_priv(dev);
2345 const struct iphdr *old_iph;
2346 struct vxlan_metadata _md;
2347 struct vxlan_metadata *md = &_md;
2348 unsigned int pkt_len = skb->len;
2349 __be16 src_port = 0, dst_port;
2350 struct dst_entry *ndst = NULL;
2351 int addr_family;
2352 __u8 tos, ttl;
2353 int ifindex;
2354 int err = 0;
2355 u32 flags = vxlan->cfg.flags;
2356 bool use_cache;
2357 bool udp_sum = false;
2358 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2359 enum skb_drop_reason reason;
2360 bool no_eth_encap;
2361 __be32 vni = 0;
2362
2363 no_eth_encap = flags & VXLAN_F_GPE && skb->protocol != htons(ETH_P_TEB);
2364 reason = skb_vlan_inet_prepare(skb, no_eth_encap);
2365 if (reason)
2366 goto drop;
2367
2368 reason = SKB_DROP_REASON_NOT_SPECIFIED;
2369 old_iph = ip_hdr(skb);
2370
2371 info = skb_tunnel_info(skb);
2372 use_cache = ip_tunnel_dst_cache_usable(skb, info);
2373
2374 if (rdst) {
2375 memset(&key, 0, sizeof(key));
2376 pkey = &key;
2377
2378 if (vxlan_addr_any(&rdst->remote_ip)) {
2379 if (did_rsc) {
2380 /* short-circuited back to local bridge */
2381 vxlan_encap_bypass(skb, vxlan, vxlan,
2382 default_vni, true);
2383 return;
2384 }
2385 goto drop;
2386 }
2387
2388 addr_family = vxlan->cfg.saddr.sa.sa_family;
2389 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2390 vni = (rdst->remote_vni) ? : default_vni;
2391 ifindex = rdst->remote_ifindex;
2392
2393 if (addr_family == AF_INET) {
2394 key.u.ipv4.src = vxlan->cfg.saddr.sin.sin_addr.s_addr;
2395 key.u.ipv4.dst = rdst->remote_ip.sin.sin_addr.s_addr;
2396 } else {
2397 key.u.ipv6.src = vxlan->cfg.saddr.sin6.sin6_addr;
2398 key.u.ipv6.dst = rdst->remote_ip.sin6.sin6_addr;
2399 }
2400
2401 dst_cache = &rdst->dst_cache;
2402 md->gbp = skb->mark;
2403 if (flags & VXLAN_F_TTL_INHERIT) {
2404 ttl = ip_tunnel_get_ttl(old_iph, skb);
2405 } else {
2406 ttl = vxlan->cfg.ttl;
2407 if (!ttl && vxlan_addr_multicast(&rdst->remote_ip))
2408 ttl = 1;
2409 }
2410 tos = vxlan->cfg.tos;
2411 if (tos == 1)
2412 tos = ip_tunnel_get_dsfield(old_iph, skb);
2413 if (tos && !info)
2414 use_cache = false;
2415
2416 if (addr_family == AF_INET)
2417 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2418 else
2419 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2420 #if IS_ENABLED(CONFIG_IPV6)
2421 switch (vxlan->cfg.label_policy) {
2422 case VXLAN_LABEL_FIXED:
2423 key.label = vxlan->cfg.label;
2424 break;
2425 case VXLAN_LABEL_INHERIT:
2426 key.label = ip_tunnel_get_flowlabel(old_iph, skb);
2427 break;
2428 default:
2429 DEBUG_NET_WARN_ON_ONCE(1);
2430 goto drop;
2431 }
2432 #endif
2433 } else {
2434 if (!info) {
2435 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2436 dev->name);
2437 goto drop;
2438 }
2439 pkey = &info->key;
2440 addr_family = ip_tunnel_info_af(info);
2441 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2442 vni = tunnel_id_to_key32(info->key.tun_id);
2443 ifindex = 0;
2444 dst_cache = &info->dst_cache;
2445 if (test_bit(IP_TUNNEL_VXLAN_OPT_BIT, info->key.tun_flags)) {
2446 if (info->options_len < sizeof(*md))
2447 goto drop;
2448 md = ip_tunnel_info_opts(info);
2449 }
2450 ttl = info->key.ttl;
2451 tos = info->key.tos;
2452 udp_sum = test_bit(IP_TUNNEL_CSUM_BIT, info->key.tun_flags);
2453 }
2454 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2455 vxlan->cfg.port_max, true);
2456
2457 rcu_read_lock();
2458 if (addr_family == AF_INET) {
2459 struct vxlan_sock *sock4;
2460 u16 ipcb_flags = 0;
2461 struct rtable *rt;
2462 __be16 df = 0;
2463 __be32 saddr;
2464
2465 sock4 = rcu_dereference(vxlan->vn4_sock);
2466 if (unlikely(!sock4)) {
2467 reason = SKB_DROP_REASON_DEV_READY;
2468 goto tx_error;
2469 }
2470
2471 if (!ifindex)
2472 ifindex = sock4->sock->sk->sk_bound_dev_if;
2473
2474 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, ifindex,
2475 &saddr, pkey, src_port, dst_port,
2476 tos, use_cache ? dst_cache : NULL);
2477 if (IS_ERR(rt)) {
2478 err = PTR_ERR(rt);
2479 reason = SKB_DROP_REASON_IP_OUTNOROUTES;
2480 goto tx_error;
2481 }
2482
2483 if (flags & VXLAN_F_MC_ROUTE)
2484 ipcb_flags |= IPSKB_MCROUTE;
2485
2486 if (!info) {
2487 /* Bypass encapsulation if the destination is local */
2488 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET,
2489 dst_port, ifindex, vni,
2490 &rt->dst, rt->rt_flags);
2491 if (err)
2492 goto out_unlock;
2493
2494 if (vxlan->cfg.df == VXLAN_DF_SET) {
2495 df = htons(IP_DF);
2496 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2497 struct ethhdr *eth = eth_hdr(skb);
2498
2499 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2500 (ntohs(eth->h_proto) == ETH_P_IP &&
2501 old_iph->frag_off & htons(IP_DF)))
2502 df = htons(IP_DF);
2503 }
2504 } else if (test_bit(IP_TUNNEL_DONT_FRAGMENT_BIT,
2505 info->key.tun_flags)) {
2506 df = htons(IP_DF);
2507 }
2508
2509 ndst = &rt->dst;
2510 err = skb_tunnel_check_pmtu(skb, ndst, vxlan_headroom(flags & VXLAN_F_GPE),
2511 netif_is_any_bridge_port(dev));
2512 if (err < 0) {
2513 goto tx_error;
2514 } else if (err) {
2515 if (info) {
2516 struct ip_tunnel_info *unclone;
2517
2518 unclone = skb_tunnel_info_unclone(skb);
2519 if (unlikely(!unclone))
2520 goto tx_error;
2521
2522 unclone->key.u.ipv4.src = pkey->u.ipv4.dst;
2523 unclone->key.u.ipv4.dst = saddr;
2524 }
2525 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2526 dst_release(ndst);
2527 goto out_unlock;
2528 }
2529
2530 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2531 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2532 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2533 vni, md, flags, udp_sum);
2534 if (err < 0) {
2535 reason = SKB_DROP_REASON_NOMEM;
2536 goto tx_error;
2537 }
2538
2539 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, saddr,
2540 pkey->u.ipv4.dst, tos, ttl, df,
2541 src_port, dst_port, xnet, !udp_sum,
2542 ipcb_flags);
2543 #if IS_ENABLED(CONFIG_IPV6)
2544 } else {
2545 struct vxlan_sock *sock6;
2546 struct in6_addr saddr;
2547 u16 ip6cb_flags = 0;
2548
2549 sock6 = rcu_dereference(vxlan->vn6_sock);
2550 if (unlikely(!sock6)) {
2551 reason = SKB_DROP_REASON_DEV_READY;
2552 goto tx_error;
2553 }
2554
2555 if (!ifindex)
2556 ifindex = sock6->sock->sk->sk_bound_dev_if;
2557
2558 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
2559 ifindex, &saddr, pkey,
2560 src_port, dst_port, tos,
2561 use_cache ? dst_cache : NULL);
2562 if (IS_ERR(ndst)) {
2563 err = PTR_ERR(ndst);
2564 ndst = NULL;
2565 reason = SKB_DROP_REASON_IP_OUTNOROUTES;
2566 goto tx_error;
2567 }
2568
2569 if (flags & VXLAN_F_MC_ROUTE)
2570 ip6cb_flags |= IP6SKB_MCROUTE;
2571
2572 if (!info) {
2573 u32 rt6i_flags = dst_rt6_info(ndst)->rt6i_flags;
2574
2575 err = encap_bypass_if_local(skb, dev, vxlan, AF_INET6,
2576 dst_port, ifindex, vni,
2577 ndst, rt6i_flags);
2578 if (err)
2579 goto out_unlock;
2580 }
2581
2582 err = skb_tunnel_check_pmtu(skb, ndst,
2583 vxlan_headroom((flags & VXLAN_F_GPE) | VXLAN_F_IPV6),
2584 netif_is_any_bridge_port(dev));
2585 if (err < 0) {
2586 goto tx_error;
2587 } else if (err) {
2588 if (info) {
2589 struct ip_tunnel_info *unclone;
2590
2591 unclone = skb_tunnel_info_unclone(skb);
2592 if (unlikely(!unclone))
2593 goto tx_error;
2594
2595 unclone->key.u.ipv6.src = pkey->u.ipv6.dst;
2596 unclone->key.u.ipv6.dst = saddr;
2597 }
2598
2599 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2600 dst_release(ndst);
2601 goto out_unlock;
2602 }
2603
2604 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2605 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2606 skb_scrub_packet(skb, xnet);
2607 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2608 vni, md, flags, udp_sum);
2609 if (err < 0) {
2610 reason = SKB_DROP_REASON_NOMEM;
2611 goto tx_error;
2612 }
2613
2614 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2615 &saddr, &pkey->u.ipv6.dst, tos, ttl,
2616 pkey->label, src_port, dst_port, !udp_sum,
2617 ip6cb_flags);
2618 #endif
2619 }
2620 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX, pkt_len);
2621 out_unlock:
2622 rcu_read_unlock();
2623 return;
2624
2625 drop:
2626 dev_dstats_tx_dropped(dev);
2627 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_DROPS, 0);
2628 kfree_skb_reason(skb, reason);
2629 return;
2630
2631 tx_error:
2632 rcu_read_unlock();
2633 if (err == -ELOOP)
2634 DEV_STATS_INC(dev, collisions);
2635 else if (err == -ENETUNREACH)
2636 DEV_STATS_INC(dev, tx_carrier_errors);
2637 dst_release(ndst);
2638 DEV_STATS_INC(dev, tx_errors);
2639 vxlan_vnifilter_count(vxlan, vni, NULL, VXLAN_VNI_STATS_TX_ERRORS, 0);
2640 kfree_skb_reason(skb, reason);
2641 }
2642
vxlan_xmit_nh(struct sk_buff * skb,struct net_device * dev,struct vxlan_fdb * f,__be32 vni,bool did_rsc)2643 static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2644 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2645 {
2646 struct vxlan_rdst nh_rdst;
2647 struct nexthop *nh;
2648 bool do_xmit;
2649 u32 hash;
2650
2651 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2652 hash = skb_get_hash(skb);
2653
2654 nh = rcu_dereference(f->nh);
2655 if (!nh)
2656 goto drop;
2657 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2658
2659 if (likely(do_xmit))
2660 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2661 else
2662 goto drop;
2663
2664 return;
2665
2666 drop:
2667 dev_dstats_tx_dropped(dev);
2668 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2669 VXLAN_VNI_STATS_TX_DROPS, 0);
2670 dev_kfree_skb(skb);
2671 }
2672
vxlan_xmit_nhid(struct sk_buff * skb,struct net_device * dev,u32 nhid,__be32 vni)2673 static netdev_tx_t vxlan_xmit_nhid(struct sk_buff *skb, struct net_device *dev,
2674 u32 nhid, __be32 vni)
2675 {
2676 struct vxlan_dev *vxlan = netdev_priv(dev);
2677 struct vxlan_rdst nh_rdst;
2678 struct nexthop *nh;
2679 bool do_xmit;
2680 u32 hash;
2681
2682 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2683 hash = skb_get_hash(skb);
2684
2685 rcu_read_lock();
2686 nh = nexthop_find_by_id(dev_net(dev), nhid);
2687 if (unlikely(!nh || !nexthop_is_fdb(nh) || !nexthop_is_multipath(nh))) {
2688 rcu_read_unlock();
2689 goto drop;
2690 }
2691 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2692 rcu_read_unlock();
2693
2694 if (vxlan->cfg.saddr.sa.sa_family != nh_rdst.remote_ip.sa.sa_family)
2695 goto drop;
2696
2697 if (likely(do_xmit))
2698 vxlan_xmit_one(skb, dev, vni, &nh_rdst, false);
2699 else
2700 goto drop;
2701
2702 return NETDEV_TX_OK;
2703
2704 drop:
2705 dev_dstats_tx_dropped(dev);
2706 vxlan_vnifilter_count(netdev_priv(dev), vni, NULL,
2707 VXLAN_VNI_STATS_TX_DROPS, 0);
2708 dev_kfree_skb(skb);
2709 return NETDEV_TX_OK;
2710 }
2711
2712 /* Transmit local packets over Vxlan
2713 *
2714 * Outer IP header inherits ECN and DF from inner header.
2715 * Outer UDP destination is the VXLAN assigned port.
2716 * source port is based on hash of flow
2717 */
vxlan_xmit(struct sk_buff * skb,struct net_device * dev)2718 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2719 {
2720 struct vxlan_dev *vxlan = netdev_priv(dev);
2721 struct vxlan_rdst *rdst, *fdst = NULL;
2722 const struct ip_tunnel_info *info;
2723 struct vxlan_fdb *f;
2724 struct ethhdr *eth;
2725 __be32 vni = 0;
2726 u32 nhid = 0;
2727 bool did_rsc;
2728
2729 info = skb_tunnel_info(skb);
2730
2731 skb_reset_mac_header(skb);
2732
2733 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2734 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2735 info->mode & IP_TUNNEL_INFO_TX) {
2736 vni = tunnel_id_to_key32(info->key.tun_id);
2737 nhid = info->key.nhid;
2738 } else {
2739 if (info && info->mode & IP_TUNNEL_INFO_TX)
2740 vxlan_xmit_one(skb, dev, vni, NULL, false);
2741 else
2742 kfree_skb_reason(skb, SKB_DROP_REASON_TUNNEL_TXINFO);
2743 return NETDEV_TX_OK;
2744 }
2745 }
2746
2747 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2748 eth = eth_hdr(skb);
2749 if (ntohs(eth->h_proto) == ETH_P_ARP)
2750 return arp_reduce(dev, skb, vni);
2751 #if IS_ENABLED(CONFIG_IPV6)
2752 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2753 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2754 sizeof(struct nd_msg)) &&
2755 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2756 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2757
2758 if (m->icmph.icmp6_code == 0 &&
2759 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2760 return neigh_reduce(dev, skb, vni);
2761 }
2762 #endif
2763 }
2764
2765 if (nhid)
2766 return vxlan_xmit_nhid(skb, dev, nhid, vni);
2767
2768 if (vxlan->cfg.flags & VXLAN_F_MDB) {
2769 struct vxlan_mdb_entry *mdb_entry;
2770
2771 rcu_read_lock();
2772 mdb_entry = vxlan_mdb_entry_skb_get(vxlan, skb, vni);
2773 if (mdb_entry) {
2774 netdev_tx_t ret;
2775
2776 ret = vxlan_mdb_xmit(vxlan, mdb_entry, skb);
2777 rcu_read_unlock();
2778 return ret;
2779 }
2780 rcu_read_unlock();
2781 }
2782
2783 eth = eth_hdr(skb);
2784 rcu_read_lock();
2785 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni);
2786 did_rsc = false;
2787
2788 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2789 (ntohs(eth->h_proto) == ETH_P_IP ||
2790 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2791 did_rsc = route_shortcircuit(dev, skb);
2792 if (did_rsc)
2793 f = vxlan_find_mac_tx(vxlan, eth->h_dest, vni);
2794 }
2795
2796 if (f == NULL) {
2797 f = vxlan_find_mac_tx(vxlan, all_zeros_mac, vni);
2798 if (f == NULL) {
2799 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2800 !is_multicast_ether_addr(eth->h_dest))
2801 vxlan_fdb_miss(vxlan, eth->h_dest);
2802
2803 dev_dstats_tx_dropped(dev);
2804 vxlan_vnifilter_count(vxlan, vni, NULL,
2805 VXLAN_VNI_STATS_TX_DROPS, 0);
2806 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET);
2807 goto out;
2808 }
2809 }
2810
2811 if (rcu_access_pointer(f->nh)) {
2812 vxlan_xmit_nh(skb, dev, f,
2813 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2814 } else {
2815 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2816 struct sk_buff *skb1;
2817
2818 if (!fdst) {
2819 fdst = rdst;
2820 continue;
2821 }
2822 skb1 = skb_clone(skb, GFP_ATOMIC);
2823 if (skb1)
2824 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2825 }
2826 if (fdst)
2827 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2828 else
2829 kfree_skb_reason(skb, SKB_DROP_REASON_NO_TX_TARGET);
2830 }
2831
2832 out:
2833 rcu_read_unlock();
2834 return NETDEV_TX_OK;
2835 }
2836
2837 /* Walk the forwarding table and purge stale entries */
vxlan_cleanup(struct timer_list * t)2838 static void vxlan_cleanup(struct timer_list *t)
2839 {
2840 struct vxlan_dev *vxlan = timer_container_of(vxlan, t, age_timer);
2841 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2842 struct vxlan_fdb *f;
2843
2844 if (!netif_running(vxlan->dev))
2845 return;
2846
2847 rcu_read_lock();
2848 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) {
2849 unsigned long timeout;
2850
2851 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2852 continue;
2853
2854 if (f->flags & NTF_EXT_LEARNED)
2855 continue;
2856
2857 timeout = READ_ONCE(f->updated) + vxlan->cfg.age_interval * HZ;
2858 if (time_before_eq(timeout, jiffies)) {
2859 spin_lock(&vxlan->hash_lock);
2860 if (!hlist_unhashed(&f->fdb_node)) {
2861 netdev_dbg(vxlan->dev, "garbage collect %pM\n",
2862 f->key.eth_addr);
2863 f->state = NUD_STALE;
2864 vxlan_fdb_destroy(vxlan, f, true, true);
2865 }
2866 spin_unlock(&vxlan->hash_lock);
2867 } else if (time_before(timeout, next_timer)) {
2868 next_timer = timeout;
2869 }
2870 }
2871 rcu_read_unlock();
2872
2873 mod_timer(&vxlan->age_timer, next_timer);
2874 }
2875
vxlan_vs_del_dev(struct vxlan_dev * vxlan)2876 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2877 {
2878 ASSERT_RTNL();
2879
2880 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2881 #if IS_ENABLED(CONFIG_IPV6)
2882 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2883 #endif
2884 }
2885
vxlan_vs_add_dev(struct vxlan_sock * vs,struct vxlan_dev * vxlan,struct vxlan_dev_node * node)2886 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2887 struct vxlan_dev_node *node)
2888 {
2889 __be32 vni = vxlan->default_dst.remote_vni;
2890
2891 ASSERT_RTNL();
2892
2893 node->vxlan = vxlan;
2894 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2895 }
2896
2897 /* Setup stats when device is created */
vxlan_init(struct net_device * dev)2898 static int vxlan_init(struct net_device *dev)
2899 {
2900 struct vxlan_dev *vxlan = netdev_priv(dev);
2901 int err;
2902
2903 err = rhashtable_init(&vxlan->fdb_hash_tbl, &vxlan_fdb_rht_params);
2904 if (err)
2905 return err;
2906
2907 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
2908 err = vxlan_vnigroup_init(vxlan);
2909 if (err)
2910 goto err_rhashtable_destroy;
2911 }
2912
2913 err = gro_cells_init(&vxlan->gro_cells, dev);
2914 if (err)
2915 goto err_vnigroup_uninit;
2916
2917 err = vxlan_mdb_init(vxlan);
2918 if (err)
2919 goto err_gro_cells_destroy;
2920
2921 netdev_lockdep_set_classes(dev);
2922 return 0;
2923
2924 err_gro_cells_destroy:
2925 gro_cells_destroy(&vxlan->gro_cells);
2926 err_vnigroup_uninit:
2927 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2928 vxlan_vnigroup_uninit(vxlan);
2929 err_rhashtable_destroy:
2930 rhashtable_destroy(&vxlan->fdb_hash_tbl);
2931 return err;
2932 }
2933
vxlan_uninit(struct net_device * dev)2934 static void vxlan_uninit(struct net_device *dev)
2935 {
2936 struct vxlan_dev *vxlan = netdev_priv(dev);
2937
2938 vxlan_mdb_fini(vxlan);
2939
2940 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER)
2941 vxlan_vnigroup_uninit(vxlan);
2942
2943 gro_cells_destroy(&vxlan->gro_cells);
2944
2945 rhashtable_destroy(&vxlan->fdb_hash_tbl);
2946 }
2947
2948 /* Start ageing timer and join group when device is brought up */
vxlan_open(struct net_device * dev)2949 static int vxlan_open(struct net_device *dev)
2950 {
2951 struct vxlan_dev *vxlan = netdev_priv(dev);
2952 int ret;
2953
2954 ret = vxlan_sock_add(vxlan);
2955 if (ret < 0)
2956 return ret;
2957
2958 ret = vxlan_multicast_join(vxlan);
2959 if (ret) {
2960 vxlan_sock_release(vxlan);
2961 return ret;
2962 }
2963
2964 if (vxlan->cfg.age_interval)
2965 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2966
2967 return ret;
2968 }
2969
2970 struct vxlan_fdb_flush_desc {
2971 bool ignore_default_entry;
2972 unsigned long state;
2973 unsigned long state_mask;
2974 unsigned long flags;
2975 unsigned long flags_mask;
2976 __be32 src_vni;
2977 u32 nhid;
2978 __be32 vni;
2979 __be16 port;
2980 union vxlan_addr dst_ip;
2981 };
2982
vxlan_fdb_is_default_entry(const struct vxlan_fdb * f,const struct vxlan_dev * vxlan)2983 static bool vxlan_fdb_is_default_entry(const struct vxlan_fdb *f,
2984 const struct vxlan_dev *vxlan)
2985 {
2986 return is_zero_ether_addr(f->key.eth_addr) &&
2987 f->key.vni == vxlan->cfg.vni;
2988 }
2989
vxlan_fdb_nhid_matches(const struct vxlan_fdb * f,u32 nhid)2990 static bool vxlan_fdb_nhid_matches(const struct vxlan_fdb *f, u32 nhid)
2991 {
2992 struct nexthop *nh = rtnl_dereference(f->nh);
2993
2994 return nh && nh->id == nhid;
2995 }
2996
vxlan_fdb_flush_matches(const struct vxlan_fdb * f,const struct vxlan_dev * vxlan,const struct vxlan_fdb_flush_desc * desc)2997 static bool vxlan_fdb_flush_matches(const struct vxlan_fdb *f,
2998 const struct vxlan_dev *vxlan,
2999 const struct vxlan_fdb_flush_desc *desc)
3000 {
3001 if (desc->state_mask && (f->state & desc->state_mask) != desc->state)
3002 return false;
3003
3004 if (desc->flags_mask && (f->flags & desc->flags_mask) != desc->flags)
3005 return false;
3006
3007 if (desc->ignore_default_entry && vxlan_fdb_is_default_entry(f, vxlan))
3008 return false;
3009
3010 if (desc->src_vni && f->key.vni != desc->src_vni)
3011 return false;
3012
3013 if (desc->nhid && !vxlan_fdb_nhid_matches(f, desc->nhid))
3014 return false;
3015
3016 return true;
3017 }
3018
3019 static bool
vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc * desc)3020 vxlan_fdb_flush_should_match_remotes(const struct vxlan_fdb_flush_desc *desc)
3021 {
3022 return desc->vni || desc->port || desc->dst_ip.sa.sa_family;
3023 }
3024
3025 static bool
vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc * desc,const struct vxlan_rdst * rd)3026 vxlan_fdb_flush_remote_matches(const struct vxlan_fdb_flush_desc *desc,
3027 const struct vxlan_rdst *rd)
3028 {
3029 if (desc->vni && rd->remote_vni != desc->vni)
3030 return false;
3031
3032 if (desc->port && rd->remote_port != desc->port)
3033 return false;
3034
3035 if (desc->dst_ip.sa.sa_family &&
3036 !vxlan_addr_equal(&rd->remote_ip, &desc->dst_ip))
3037 return false;
3038
3039 return true;
3040 }
3041
3042 static void
vxlan_fdb_flush_match_remotes(struct vxlan_fdb * f,struct vxlan_dev * vxlan,const struct vxlan_fdb_flush_desc * desc,bool * p_destroy_fdb)3043 vxlan_fdb_flush_match_remotes(struct vxlan_fdb *f, struct vxlan_dev *vxlan,
3044 const struct vxlan_fdb_flush_desc *desc,
3045 bool *p_destroy_fdb)
3046 {
3047 bool remotes_flushed = false;
3048 struct vxlan_rdst *rd, *tmp;
3049
3050 list_for_each_entry_safe(rd, tmp, &f->remotes, list) {
3051 if (!vxlan_fdb_flush_remote_matches(desc, rd))
3052 continue;
3053
3054 vxlan_fdb_dst_destroy(vxlan, f, rd, true);
3055 remotes_flushed = true;
3056 }
3057
3058 *p_destroy_fdb = remotes_flushed && list_empty(&f->remotes);
3059 }
3060
3061 /* Purge the forwarding table */
vxlan_flush(struct vxlan_dev * vxlan,const struct vxlan_fdb_flush_desc * desc)3062 static void vxlan_flush(struct vxlan_dev *vxlan,
3063 const struct vxlan_fdb_flush_desc *desc)
3064 {
3065 bool match_remotes = vxlan_fdb_flush_should_match_remotes(desc);
3066 struct vxlan_fdb *f;
3067
3068 rcu_read_lock();
3069 hlist_for_each_entry_rcu(f, &vxlan->fdb_list, fdb_node) {
3070 if (!vxlan_fdb_flush_matches(f, vxlan, desc))
3071 continue;
3072
3073 spin_lock_bh(&vxlan->hash_lock);
3074 if (hlist_unhashed(&f->fdb_node))
3075 goto unlock;
3076
3077 if (match_remotes) {
3078 bool destroy_fdb = false;
3079
3080 vxlan_fdb_flush_match_remotes(f, vxlan, desc,
3081 &destroy_fdb);
3082
3083 if (!destroy_fdb)
3084 goto unlock;
3085 }
3086
3087 vxlan_fdb_destroy(vxlan, f, true, true);
3088 unlock:
3089 spin_unlock_bh(&vxlan->hash_lock);
3090 }
3091 rcu_read_unlock();
3092 }
3093
3094 static const struct nla_policy vxlan_del_bulk_policy[NDA_MAX + 1] = {
3095 [NDA_SRC_VNI] = { .type = NLA_U32 },
3096 [NDA_NH_ID] = { .type = NLA_U32 },
3097 [NDA_VNI] = { .type = NLA_U32 },
3098 [NDA_PORT] = { .type = NLA_U16 },
3099 [NDA_DST] = NLA_POLICY_RANGE(NLA_BINARY, sizeof(struct in_addr),
3100 sizeof(struct in6_addr)),
3101 [NDA_NDM_STATE_MASK] = { .type = NLA_U16 },
3102 [NDA_NDM_FLAGS_MASK] = { .type = NLA_U8 },
3103 };
3104
3105 #define VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS (NTF_MASTER | NTF_SELF)
3106 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES (NUD_PERMANENT | NUD_NOARP)
3107 #define VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS (NTF_EXT_LEARNED | NTF_OFFLOADED | \
3108 NTF_ROUTER)
3109
vxlan_fdb_delete_bulk(struct nlmsghdr * nlh,struct net_device * dev,struct netlink_ext_ack * extack)3110 static int vxlan_fdb_delete_bulk(struct nlmsghdr *nlh, struct net_device *dev,
3111 struct netlink_ext_ack *extack)
3112 {
3113 struct vxlan_dev *vxlan = netdev_priv(dev);
3114 struct vxlan_fdb_flush_desc desc = {};
3115 struct ndmsg *ndm = nlmsg_data(nlh);
3116 struct nlattr *tb[NDA_MAX + 1];
3117 u8 ndm_flags;
3118 int err;
3119
3120 ndm_flags = ndm->ndm_flags & ~VXLAN_FDB_FLUSH_IGNORED_NDM_FLAGS;
3121
3122 err = nlmsg_parse(nlh, sizeof(*ndm), tb, NDA_MAX, vxlan_del_bulk_policy,
3123 extack);
3124 if (err)
3125 return err;
3126
3127 if (ndm_flags & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_FLAGS) {
3128 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm flag bits set");
3129 return -EINVAL;
3130 }
3131 if (ndm->ndm_state & ~VXLAN_FDB_FLUSH_ALLOWED_NDM_STATES) {
3132 NL_SET_ERR_MSG(extack, "Unsupported fdb flush ndm state bits set");
3133 return -EINVAL;
3134 }
3135
3136 desc.state = ndm->ndm_state;
3137 desc.flags = ndm_flags;
3138
3139 if (tb[NDA_NDM_STATE_MASK])
3140 desc.state_mask = nla_get_u16(tb[NDA_NDM_STATE_MASK]);
3141
3142 if (tb[NDA_NDM_FLAGS_MASK])
3143 desc.flags_mask = nla_get_u8(tb[NDA_NDM_FLAGS_MASK]);
3144
3145 if (tb[NDA_SRC_VNI])
3146 desc.src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
3147
3148 if (tb[NDA_NH_ID])
3149 desc.nhid = nla_get_u32(tb[NDA_NH_ID]);
3150
3151 if (tb[NDA_VNI])
3152 desc.vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
3153
3154 if (tb[NDA_PORT])
3155 desc.port = nla_get_be16(tb[NDA_PORT]);
3156
3157 if (tb[NDA_DST]) {
3158 union vxlan_addr ip;
3159
3160 err = vxlan_nla_get_addr(&ip, tb[NDA_DST]);
3161 if (err) {
3162 NL_SET_ERR_MSG_ATTR(extack, tb[NDA_DST],
3163 "Unsupported address family");
3164 return err;
3165 }
3166 desc.dst_ip = ip;
3167 }
3168
3169 vxlan_flush(vxlan, &desc);
3170
3171 return 0;
3172 }
3173
3174 /* Cleanup timer and forwarding table on shutdown */
vxlan_stop(struct net_device * dev)3175 static int vxlan_stop(struct net_device *dev)
3176 {
3177 struct vxlan_dev *vxlan = netdev_priv(dev);
3178 struct vxlan_fdb_flush_desc desc = {
3179 /* Default entry is deleted at vxlan_dellink. */
3180 .ignore_default_entry = true,
3181 .state = 0,
3182 .state_mask = NUD_PERMANENT | NUD_NOARP,
3183 };
3184
3185 vxlan_multicast_leave(vxlan);
3186
3187 timer_delete_sync(&vxlan->age_timer);
3188
3189 vxlan_flush(vxlan, &desc);
3190 vxlan_sock_release(vxlan);
3191
3192 return 0;
3193 }
3194
3195 /* Stub, nothing needs to be done. */
vxlan_set_multicast_list(struct net_device * dev)3196 static void vxlan_set_multicast_list(struct net_device *dev)
3197 {
3198 }
3199
vxlan_change_mtu(struct net_device * dev,int new_mtu)3200 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3201 {
3202 struct vxlan_dev *vxlan = netdev_priv(dev);
3203 struct vxlan_rdst *dst = &vxlan->default_dst;
3204 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3205 dst->remote_ifindex);
3206
3207 /* This check is different than dev->max_mtu, because it looks at
3208 * the lowerdev->mtu, rather than the static dev->max_mtu
3209 */
3210 if (lowerdev) {
3211 int max_mtu = lowerdev->mtu - vxlan_headroom(vxlan->cfg.flags);
3212 if (new_mtu > max_mtu)
3213 return -EINVAL;
3214 }
3215
3216 WRITE_ONCE(dev->mtu, new_mtu);
3217 return 0;
3218 }
3219
vxlan_fill_metadata_dst(struct net_device * dev,struct sk_buff * skb)3220 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3221 {
3222 struct vxlan_dev *vxlan = netdev_priv(dev);
3223 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3224 __be16 sport, dport;
3225
3226 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3227 vxlan->cfg.port_max, true);
3228 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3229
3230 if (ip_tunnel_info_af(info) == AF_INET) {
3231 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3232 struct rtable *rt;
3233
3234 if (!sock4)
3235 return -EIO;
3236
3237 rt = udp_tunnel_dst_lookup(skb, dev, vxlan->net, 0,
3238 &info->key.u.ipv4.src,
3239 &info->key,
3240 sport, dport, info->key.tos,
3241 &info->dst_cache);
3242 if (IS_ERR(rt))
3243 return PTR_ERR(rt);
3244 ip_rt_put(rt);
3245 } else {
3246 #if IS_ENABLED(CONFIG_IPV6)
3247 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3248 struct dst_entry *ndst;
3249
3250 if (!sock6)
3251 return -EIO;
3252
3253 ndst = udp_tunnel6_dst_lookup(skb, dev, vxlan->net, sock6->sock,
3254 0, &info->key.u.ipv6.src,
3255 &info->key,
3256 sport, dport, info->key.tos,
3257 &info->dst_cache);
3258 if (IS_ERR(ndst))
3259 return PTR_ERR(ndst);
3260 dst_release(ndst);
3261 #else /* !CONFIG_IPV6 */
3262 return -EPFNOSUPPORT;
3263 #endif
3264 }
3265 info->key.tp_src = sport;
3266 info->key.tp_dst = dport;
3267 return 0;
3268 }
3269
3270 static const struct net_device_ops vxlan_netdev_ether_ops = {
3271 .ndo_init = vxlan_init,
3272 .ndo_uninit = vxlan_uninit,
3273 .ndo_open = vxlan_open,
3274 .ndo_stop = vxlan_stop,
3275 .ndo_start_xmit = vxlan_xmit,
3276 .ndo_set_rx_mode = vxlan_set_multicast_list,
3277 .ndo_change_mtu = vxlan_change_mtu,
3278 .ndo_validate_addr = eth_validate_addr,
3279 .ndo_set_mac_address = eth_mac_addr,
3280 .ndo_fdb_add = vxlan_fdb_add,
3281 .ndo_fdb_del = vxlan_fdb_delete,
3282 .ndo_fdb_del_bulk = vxlan_fdb_delete_bulk,
3283 .ndo_fdb_dump = vxlan_fdb_dump,
3284 .ndo_fdb_get = vxlan_fdb_get,
3285 .ndo_mdb_add = vxlan_mdb_add,
3286 .ndo_mdb_del = vxlan_mdb_del,
3287 .ndo_mdb_del_bulk = vxlan_mdb_del_bulk,
3288 .ndo_mdb_dump = vxlan_mdb_dump,
3289 .ndo_mdb_get = vxlan_mdb_get,
3290 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3291 };
3292
3293 static const struct net_device_ops vxlan_netdev_raw_ops = {
3294 .ndo_init = vxlan_init,
3295 .ndo_uninit = vxlan_uninit,
3296 .ndo_open = vxlan_open,
3297 .ndo_stop = vxlan_stop,
3298 .ndo_start_xmit = vxlan_xmit,
3299 .ndo_change_mtu = vxlan_change_mtu,
3300 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3301 };
3302
3303 /* Info for udev, that this is a virtual tunnel endpoint */
3304 static const struct device_type vxlan_type = {
3305 .name = "vxlan",
3306 };
3307
3308 /* Calls the ndo_udp_tunnel_add of the caller in order to
3309 * supply the listening VXLAN udp ports. Callers are expected
3310 * to implement the ndo_udp_tunnel_add.
3311 */
vxlan_offload_rx_ports(struct net_device * dev,bool push)3312 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3313 {
3314 struct vxlan_sock *vs;
3315 struct net *net = dev_net(dev);
3316 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3317 unsigned int i;
3318
3319 ASSERT_RTNL();
3320
3321 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3322 hlist_for_each_entry(vs, &vn->sock_list[i], hlist) {
3323 unsigned short type;
3324
3325 if (vs->flags & VXLAN_F_GPE)
3326 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3327 else
3328 type = UDP_TUNNEL_TYPE_VXLAN;
3329
3330 if (push)
3331 udp_tunnel_push_rx_port(dev, vs->sock, type);
3332 else
3333 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3334 }
3335 }
3336 }
3337
3338 /* Initialize the device structure. */
vxlan_setup(struct net_device * dev)3339 static void vxlan_setup(struct net_device *dev)
3340 {
3341 struct vxlan_dev *vxlan = netdev_priv(dev);
3342
3343 eth_hw_addr_random(dev);
3344 ether_setup(dev);
3345
3346 dev->needs_free_netdev = true;
3347 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3348
3349 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3350 dev->features |= NETIF_F_RXCSUM;
3351 dev->features |= NETIF_F_GSO_SOFTWARE;
3352
3353 /* Partial features are disabled by default. */
3354 dev->vlan_features = dev->features;
3355 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3356 dev->hw_features |= NETIF_F_RXCSUM;
3357 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3358 dev->hw_features |= UDP_TUNNEL_PARTIAL_FEATURES;
3359 dev->hw_features |= NETIF_F_GSO_PARTIAL;
3360
3361 dev->hw_enc_features = dev->hw_features;
3362 dev->gso_partial_features = UDP_TUNNEL_PARTIAL_FEATURES;
3363 dev->mangleid_features = NETIF_F_GSO_PARTIAL;
3364
3365 netif_keep_dst(dev);
3366 dev->priv_flags |= IFF_NO_QUEUE;
3367 dev->change_proto_down = true;
3368 dev->lltx = true;
3369
3370 /* MTU range: 68 - 65535 */
3371 dev->min_mtu = ETH_MIN_MTU;
3372 dev->max_mtu = ETH_MAX_MTU;
3373
3374 dev->pcpu_stat_type = NETDEV_PCPU_STAT_DSTATS;
3375 INIT_LIST_HEAD(&vxlan->next);
3376 spin_lock_init(&vxlan->hash_lock);
3377
3378 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3379
3380 vxlan->dev = dev;
3381
3382 INIT_HLIST_HEAD(&vxlan->fdb_list);
3383 }
3384
vxlan_ether_setup(struct net_device * dev)3385 static void vxlan_ether_setup(struct net_device *dev)
3386 {
3387 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3388 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3389 dev->netdev_ops = &vxlan_netdev_ether_ops;
3390 }
3391
vxlan_raw_setup(struct net_device * dev)3392 static void vxlan_raw_setup(struct net_device *dev)
3393 {
3394 dev->header_ops = NULL;
3395 dev->type = ARPHRD_NONE;
3396 dev->hard_header_len = 0;
3397 dev->addr_len = 0;
3398 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3399 dev->netdev_ops = &vxlan_netdev_raw_ops;
3400 }
3401
3402 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3403 [IFLA_VXLAN_UNSPEC] = { .strict_start_type = IFLA_VXLAN_LOCALBYPASS },
3404 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3405 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3406 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3407 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3408 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3409 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3410 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3411 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3412 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3413 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3414 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3415 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3416 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3417 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3418 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3419 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3420 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3421 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3422 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3423 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3424 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3425 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3426 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3427 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3428 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3429 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3430 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3431 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3432 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3433 [IFLA_VXLAN_VNIFILTER] = { .type = NLA_U8 },
3434 [IFLA_VXLAN_LOCALBYPASS] = NLA_POLICY_MAX(NLA_U8, 1),
3435 [IFLA_VXLAN_LABEL_POLICY] = NLA_POLICY_MAX(NLA_U32, VXLAN_LABEL_MAX),
3436 [IFLA_VXLAN_RESERVED_BITS] = NLA_POLICY_EXACT_LEN(sizeof(struct vxlanhdr)),
3437 [IFLA_VXLAN_MC_ROUTE] = NLA_POLICY_MAX(NLA_U8, 1),
3438 };
3439
vxlan_validate(struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)3440 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3441 struct netlink_ext_ack *extack)
3442 {
3443 if (tb[IFLA_ADDRESS]) {
3444 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3445 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3446 "Provided link layer address is not Ethernet");
3447 return -EINVAL;
3448 }
3449
3450 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3451 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3452 "Provided Ethernet address is not unicast");
3453 return -EADDRNOTAVAIL;
3454 }
3455 }
3456
3457 if (tb[IFLA_MTU]) {
3458 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3459
3460 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3461 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3462 "MTU must be between 68 and 65535");
3463 return -EINVAL;
3464 }
3465 }
3466
3467 if (!data) {
3468 NL_SET_ERR_MSG(extack,
3469 "Required attributes not provided to perform the operation");
3470 return -EINVAL;
3471 }
3472
3473 if (data[IFLA_VXLAN_ID]) {
3474 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3475
3476 if (id >= VXLAN_N_VID) {
3477 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3478 "VXLAN ID must be lower than 16777216");
3479 return -ERANGE;
3480 }
3481 }
3482
3483 if (data[IFLA_VXLAN_PORT_RANGE]) {
3484 const struct ifla_vxlan_port_range *p
3485 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3486
3487 if (ntohs(p->high) < ntohs(p->low)) {
3488 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3489 "Invalid source port range");
3490 return -EINVAL;
3491 }
3492 }
3493
3494 if (data[IFLA_VXLAN_DF]) {
3495 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3496
3497 if (df < 0 || df > VXLAN_DF_MAX) {
3498 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3499 "Invalid DF attribute");
3500 return -EINVAL;
3501 }
3502 }
3503
3504 return 0;
3505 }
3506
vxlan_get_drvinfo(struct net_device * netdev,struct ethtool_drvinfo * drvinfo)3507 static void vxlan_get_drvinfo(struct net_device *netdev,
3508 struct ethtool_drvinfo *drvinfo)
3509 {
3510 strscpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3511 strscpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3512 }
3513
vxlan_get_link_ksettings(struct net_device * dev,struct ethtool_link_ksettings * cmd)3514 static int vxlan_get_link_ksettings(struct net_device *dev,
3515 struct ethtool_link_ksettings *cmd)
3516 {
3517 struct vxlan_dev *vxlan = netdev_priv(dev);
3518 struct vxlan_rdst *dst = &vxlan->default_dst;
3519 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3520 dst->remote_ifindex);
3521
3522 if (!lowerdev) {
3523 cmd->base.duplex = DUPLEX_UNKNOWN;
3524 cmd->base.port = PORT_OTHER;
3525 cmd->base.speed = SPEED_UNKNOWN;
3526
3527 return 0;
3528 }
3529
3530 return __ethtool_get_link_ksettings(lowerdev, cmd);
3531 }
3532
3533 static const struct ethtool_ops vxlan_ethtool_ops = {
3534 .get_drvinfo = vxlan_get_drvinfo,
3535 .get_link = ethtool_op_get_link,
3536 .get_link_ksettings = vxlan_get_link_ksettings,
3537 };
3538
vxlan_create_sock(struct net * net,bool ipv6,__be16 port,u32 flags,int ifindex)3539 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3540 __be16 port, u32 flags, int ifindex)
3541 {
3542 struct socket *sock;
3543 struct udp_port_cfg udp_conf;
3544 int err;
3545
3546 memset(&udp_conf, 0, sizeof(udp_conf));
3547
3548 if (ipv6) {
3549 udp_conf.family = AF_INET6;
3550 udp_conf.use_udp6_rx_checksums =
3551 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3552 udp_conf.ipv6_v6only = 1;
3553 } else {
3554 udp_conf.family = AF_INET;
3555 }
3556
3557 udp_conf.local_udp_port = port;
3558 udp_conf.bind_ifindex = ifindex;
3559
3560 /* Open UDP socket */
3561 err = udp_sock_create(net, &udp_conf, &sock);
3562 if (err < 0)
3563 return ERR_PTR(err);
3564
3565 udp_allow_gso(sock->sk);
3566 return sock;
3567 }
3568
3569 /* Create new listen socket if needed */
vxlan_socket_create(struct net * net,bool ipv6,__be16 port,u32 flags,int ifindex)3570 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3571 __be16 port, u32 flags,
3572 int ifindex)
3573 {
3574 struct vxlan_sock *vs;
3575 struct socket *sock;
3576 unsigned int h;
3577 struct udp_tunnel_sock_cfg tunnel_cfg;
3578
3579 ASSERT_RTNL();
3580
3581 vs = kzalloc_obj(*vs);
3582 if (!vs)
3583 return ERR_PTR(-ENOMEM);
3584
3585 for (h = 0; h < VNI_HASH_SIZE; ++h)
3586 INIT_HLIST_HEAD(&vs->vni_list[h]);
3587
3588 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3589 if (IS_ERR(sock)) {
3590 kfree(vs);
3591 return ERR_CAST(sock);
3592 }
3593
3594 vs->sock = sock;
3595 refcount_set(&vs->refcnt, 1);
3596 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3597
3598 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3599 udp_tunnel_notify_add_rx_port(sock,
3600 (vs->flags & VXLAN_F_GPE) ?
3601 UDP_TUNNEL_TYPE_VXLAN_GPE :
3602 UDP_TUNNEL_TYPE_VXLAN);
3603
3604 /* Mark socket as an encapsulation socket. */
3605 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3606 tunnel_cfg.sk_user_data = vs;
3607 tunnel_cfg.encap_type = 1;
3608 tunnel_cfg.encap_rcv = vxlan_rcv;
3609 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3610 tunnel_cfg.encap_destroy = NULL;
3611 if (vs->flags & VXLAN_F_GPE) {
3612 tunnel_cfg.gro_receive = vxlan_gpe_gro_receive;
3613 tunnel_cfg.gro_complete = vxlan_gpe_gro_complete;
3614 } else {
3615 tunnel_cfg.gro_receive = vxlan_gro_receive;
3616 tunnel_cfg.gro_complete = vxlan_gro_complete;
3617 }
3618
3619 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3620
3621 return vs;
3622 }
3623
__vxlan_sock_add(struct vxlan_dev * vxlan,bool ipv6)3624 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3625 {
3626 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3627 struct vxlan_sock *vs = NULL;
3628 struct vxlan_dev_node *node;
3629 int l3mdev_index = 0;
3630
3631 ASSERT_RTNL();
3632
3633 if (vxlan->cfg.remote_ifindex)
3634 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3635 vxlan->net, vxlan->cfg.remote_ifindex);
3636
3637 if (!vxlan->cfg.no_share) {
3638 rcu_read_lock();
3639 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3640 vxlan->cfg.dst_port, vxlan->cfg.flags,
3641 l3mdev_index);
3642 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3643 rcu_read_unlock();
3644 return -EBUSY;
3645 }
3646 rcu_read_unlock();
3647 }
3648 if (!vs)
3649 vs = vxlan_socket_create(vxlan->net, ipv6,
3650 vxlan->cfg.dst_port, vxlan->cfg.flags,
3651 l3mdev_index);
3652 if (IS_ERR(vs))
3653 return PTR_ERR(vs);
3654 #if IS_ENABLED(CONFIG_IPV6)
3655 if (ipv6) {
3656 rcu_assign_pointer(vxlan->vn6_sock, vs);
3657 node = &vxlan->hlist6;
3658 } else
3659 #endif
3660 {
3661 rcu_assign_pointer(vxlan->vn4_sock, vs);
3662 node = &vxlan->hlist4;
3663 }
3664
3665 if (metadata && (vxlan->cfg.flags & VXLAN_F_VNIFILTER))
3666 vxlan_vs_add_vnigrp(vxlan, vs, ipv6);
3667 else
3668 vxlan_vs_add_dev(vs, vxlan, node);
3669
3670 return 0;
3671 }
3672
vxlan_sock_add(struct vxlan_dev * vxlan)3673 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3674 {
3675 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3676 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3677 bool ipv4 = !ipv6 || metadata;
3678 int ret = 0;
3679
3680 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3681 #if IS_ENABLED(CONFIG_IPV6)
3682 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3683 if (ipv6) {
3684 ret = __vxlan_sock_add(vxlan, true);
3685 if (ret < 0 && ret != -EAFNOSUPPORT)
3686 ipv4 = false;
3687 }
3688 #endif
3689 if (ipv4)
3690 ret = __vxlan_sock_add(vxlan, false);
3691 if (ret < 0)
3692 vxlan_sock_release(vxlan);
3693 return ret;
3694 }
3695
vxlan_vni_in_use(struct net * src_net,struct vxlan_dev * vxlan,struct vxlan_config * conf,__be32 vni)3696 int vxlan_vni_in_use(struct net *src_net, struct vxlan_dev *vxlan,
3697 struct vxlan_config *conf, __be32 vni)
3698 {
3699 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3700 struct vxlan_dev *tmp;
3701
3702 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3703 if (tmp == vxlan)
3704 continue;
3705 if (tmp->cfg.flags & VXLAN_F_VNIFILTER) {
3706 if (!vxlan_vnifilter_lookup(tmp, vni))
3707 continue;
3708 } else if (tmp->cfg.vni != vni) {
3709 continue;
3710 }
3711 if (tmp->cfg.dst_port != conf->dst_port)
3712 continue;
3713 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3714 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3715 continue;
3716
3717 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3718 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3719 continue;
3720
3721 return -EEXIST;
3722 }
3723
3724 return 0;
3725 }
3726
vxlan_config_validate(struct net * src_net,struct vxlan_config * conf,struct net_device ** lower,struct vxlan_dev * old,struct netlink_ext_ack * extack)3727 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3728 struct net_device **lower,
3729 struct vxlan_dev *old,
3730 struct netlink_ext_ack *extack)
3731 {
3732 bool use_ipv6 = false;
3733
3734 if (conf->flags & VXLAN_F_GPE) {
3735 /* For now, allow GPE only together with
3736 * COLLECT_METADATA. This can be relaxed later; in such
3737 * case, the other side of the PtP link will have to be
3738 * provided.
3739 */
3740 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3741 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3742 NL_SET_ERR_MSG(extack,
3743 "VXLAN GPE does not support this combination of attributes");
3744 return -EINVAL;
3745 }
3746 }
3747
3748 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3749 /* Unless IPv6 is explicitly requested, assume IPv4 */
3750 conf->remote_ip.sa.sa_family = AF_INET;
3751 conf->saddr.sa.sa_family = AF_INET;
3752 } else if (!conf->remote_ip.sa.sa_family) {
3753 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3754 } else if (!conf->saddr.sa.sa_family) {
3755 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3756 }
3757
3758 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3759 NL_SET_ERR_MSG(extack,
3760 "Local and remote address must be from the same family");
3761 return -EINVAL;
3762 }
3763
3764 if (vxlan_addr_multicast(&conf->saddr)) {
3765 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3766 return -EINVAL;
3767 }
3768
3769 if (conf->saddr.sa.sa_family == AF_INET6) {
3770 if (!IS_ENABLED(CONFIG_IPV6)) {
3771 NL_SET_ERR_MSG(extack,
3772 "IPv6 support not enabled in the kernel");
3773 return -EPFNOSUPPORT;
3774 }
3775 use_ipv6 = true;
3776 conf->flags |= VXLAN_F_IPV6;
3777
3778 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3779 int local_type =
3780 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3781 int remote_type =
3782 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3783
3784 if (local_type & IPV6_ADDR_LINKLOCAL) {
3785 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3786 (remote_type != IPV6_ADDR_ANY)) {
3787 NL_SET_ERR_MSG(extack,
3788 "Invalid combination of local and remote address scopes");
3789 return -EINVAL;
3790 }
3791
3792 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3793 } else {
3794 if (remote_type ==
3795 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3796 NL_SET_ERR_MSG(extack,
3797 "Invalid combination of local and remote address scopes");
3798 return -EINVAL;
3799 }
3800
3801 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3802 }
3803 }
3804 }
3805
3806 if (conf->label && !use_ipv6) {
3807 NL_SET_ERR_MSG(extack,
3808 "Label attribute only applies to IPv6 VXLAN devices");
3809 return -EINVAL;
3810 }
3811
3812 if (conf->label_policy && !use_ipv6) {
3813 NL_SET_ERR_MSG(extack,
3814 "Label policy only applies to IPv6 VXLAN devices");
3815 return -EINVAL;
3816 }
3817
3818 if (conf->remote_ifindex) {
3819 struct net_device *lowerdev;
3820
3821 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3822 if (!lowerdev) {
3823 NL_SET_ERR_MSG(extack,
3824 "Invalid local interface, device not found");
3825 return -ENODEV;
3826 }
3827
3828 #if IS_ENABLED(CONFIG_IPV6)
3829 if (use_ipv6) {
3830 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3831
3832 if (idev && idev->cnf.disable_ipv6) {
3833 NL_SET_ERR_MSG(extack,
3834 "IPv6 support disabled by administrator");
3835 return -EPERM;
3836 }
3837 }
3838 #endif
3839
3840 *lower = lowerdev;
3841 } else {
3842 if (vxlan_addr_multicast(&conf->remote_ip)) {
3843 NL_SET_ERR_MSG(extack,
3844 "Local interface required for multicast remote destination");
3845
3846 return -EINVAL;
3847 }
3848
3849 #if IS_ENABLED(CONFIG_IPV6)
3850 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3851 NL_SET_ERR_MSG(extack,
3852 "Local interface required for link-local local/remote addresses");
3853 return -EINVAL;
3854 }
3855 #endif
3856
3857 *lower = NULL;
3858 }
3859
3860 if (!conf->dst_port) {
3861 if (conf->flags & VXLAN_F_GPE)
3862 conf->dst_port = htons(IANA_VXLAN_GPE_UDP_PORT);
3863 else
3864 conf->dst_port = htons(vxlan_port);
3865 }
3866
3867 if (!conf->age_interval)
3868 conf->age_interval = FDB_AGE_DEFAULT;
3869
3870 if (vxlan_vni_in_use(src_net, old, conf, conf->vni)) {
3871 NL_SET_ERR_MSG(extack,
3872 "A VXLAN device with the specified VNI already exists");
3873 return -EEXIST;
3874 }
3875
3876 return 0;
3877 }
3878
vxlan_config_apply(struct net_device * dev,struct vxlan_config * conf,struct net_device * lowerdev,struct net * src_net,bool changelink)3879 static void vxlan_config_apply(struct net_device *dev,
3880 struct vxlan_config *conf,
3881 struct net_device *lowerdev,
3882 struct net *src_net,
3883 bool changelink)
3884 {
3885 struct vxlan_dev *vxlan = netdev_priv(dev);
3886 struct vxlan_rdst *dst = &vxlan->default_dst;
3887 unsigned short needed_headroom = ETH_HLEN;
3888 int max_mtu = ETH_MAX_MTU;
3889 u32 flags = conf->flags;
3890
3891 if (!changelink) {
3892 if (flags & VXLAN_F_GPE)
3893 vxlan_raw_setup(dev);
3894 else
3895 vxlan_ether_setup(dev);
3896
3897 if (conf->mtu)
3898 dev->mtu = conf->mtu;
3899
3900 vxlan->net = src_net;
3901 }
3902
3903 dst->remote_vni = conf->vni;
3904
3905 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3906
3907 if (lowerdev) {
3908 dst->remote_ifindex = conf->remote_ifindex;
3909
3910 netif_inherit_tso_max(dev, lowerdev);
3911
3912 needed_headroom = lowerdev->hard_header_len;
3913 needed_headroom += lowerdev->needed_headroom;
3914
3915 dev->needed_tailroom = lowerdev->needed_tailroom;
3916
3917 max_mtu = lowerdev->mtu - vxlan_headroom(flags);
3918 if (max_mtu < ETH_MIN_MTU)
3919 max_mtu = ETH_MIN_MTU;
3920
3921 if (!changelink && !conf->mtu)
3922 dev->mtu = max_mtu;
3923 }
3924
3925 if (dev->mtu > max_mtu)
3926 dev->mtu = max_mtu;
3927
3928 if (flags & VXLAN_F_COLLECT_METADATA)
3929 flags |= VXLAN_F_IPV6;
3930 needed_headroom += vxlan_headroom(flags);
3931 dev->needed_headroom = needed_headroom;
3932
3933 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3934 }
3935
vxlan_dev_configure(struct net * src_net,struct net_device * dev,struct vxlan_config * conf,struct netlink_ext_ack * extack)3936 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3937 struct vxlan_config *conf,
3938 struct netlink_ext_ack *extack)
3939 {
3940 struct vxlan_dev *vxlan = netdev_priv(dev);
3941 struct net_device *lowerdev;
3942 int ret;
3943
3944 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3945 if (ret)
3946 return ret;
3947
3948 vxlan_config_apply(dev, conf, lowerdev, src_net, false);
3949
3950 return 0;
3951 }
3952
__vxlan_dev_create(struct net * net,struct net_device * dev,struct vxlan_config * conf,struct netlink_ext_ack * extack)3953 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3954 struct vxlan_config *conf,
3955 struct netlink_ext_ack *extack)
3956 {
3957 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3958 struct vxlan_dev *vxlan = netdev_priv(dev);
3959 struct net_device *remote_dev = NULL;
3960 struct vxlan_rdst *dst;
3961 int err;
3962
3963 dst = &vxlan->default_dst;
3964 err = vxlan_dev_configure(net, dev, conf, extack);
3965 if (err)
3966 return err;
3967
3968 dev->ethtool_ops = &vxlan_ethtool_ops;
3969
3970 err = register_netdevice(dev);
3971 if (err)
3972 return err;
3973
3974 if (dst->remote_ifindex) {
3975 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3976 if (!remote_dev) {
3977 err = -ENODEV;
3978 goto unregister;
3979 }
3980
3981 err = netdev_upper_dev_link(remote_dev, dev, extack);
3982 if (err)
3983 goto unregister;
3984
3985 dst->remote_dev = remote_dev;
3986 }
3987
3988 err = rtnl_configure_link(dev, NULL, 0, NULL);
3989 if (err < 0)
3990 goto unlink;
3991
3992 /* create an fdb entry for a valid default destination */
3993 if (!vxlan_addr_any(&dst->remote_ip)) {
3994 spin_lock_bh(&vxlan->hash_lock);
3995 err = vxlan_fdb_update(vxlan, all_zeros_mac,
3996 &dst->remote_ip,
3997 NUD_REACHABLE | NUD_PERMANENT,
3998 NLM_F_EXCL | NLM_F_CREATE,
3999 vxlan->cfg.dst_port,
4000 dst->remote_vni,
4001 dst->remote_vni,
4002 dst->remote_ifindex,
4003 NTF_SELF, 0, true, extack);
4004 spin_unlock_bh(&vxlan->hash_lock);
4005 if (err)
4006 goto unlink;
4007 }
4008
4009 list_add(&vxlan->next, &vn->vxlan_list);
4010
4011 return 0;
4012
4013 unlink:
4014 if (remote_dev)
4015 netdev_upper_dev_unlink(remote_dev, dev);
4016 unregister:
4017 unregister_netdevice(dev);
4018 return err;
4019 }
4020
4021 /* Set/clear flags based on attribute */
vxlan_nl2flag(struct vxlan_config * conf,struct nlattr * tb[],int attrtype,unsigned long mask,bool changelink,bool changelink_supported,struct netlink_ext_ack * extack)4022 static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
4023 int attrtype, unsigned long mask, bool changelink,
4024 bool changelink_supported,
4025 struct netlink_ext_ack *extack)
4026 {
4027 unsigned long flags;
4028
4029 if (!tb[attrtype])
4030 return 0;
4031
4032 if (changelink && !changelink_supported) {
4033 vxlan_flag_attr_error(attrtype, extack);
4034 return -EOPNOTSUPP;
4035 }
4036
4037 if (vxlan_policy[attrtype].type == NLA_FLAG)
4038 flags = conf->flags | mask;
4039 else if (nla_get_u8(tb[attrtype]))
4040 flags = conf->flags | mask;
4041 else
4042 flags = conf->flags & ~mask;
4043
4044 conf->flags = flags;
4045
4046 return 0;
4047 }
4048
vxlan_nl2conf(struct nlattr * tb[],struct nlattr * data[],struct net_device * dev,struct vxlan_config * conf,bool changelink,struct netlink_ext_ack * extack)4049 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
4050 struct net_device *dev, struct vxlan_config *conf,
4051 bool changelink, struct netlink_ext_ack *extack)
4052 {
4053 struct vxlanhdr used_bits = {
4054 .vx_flags = VXLAN_HF_VNI,
4055 .vx_vni = VXLAN_VNI_MASK,
4056 };
4057 struct vxlan_dev *vxlan = netdev_priv(dev);
4058 int err = 0;
4059
4060 memset(conf, 0, sizeof(*conf));
4061
4062 /* if changelink operation, start with old existing cfg */
4063 if (changelink)
4064 memcpy(conf, &vxlan->cfg, sizeof(*conf));
4065
4066 if (data[IFLA_VXLAN_ID]) {
4067 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
4068
4069 if (changelink && (vni != conf->vni)) {
4070 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
4071 return -EOPNOTSUPP;
4072 }
4073 conf->vni = vni;
4074 }
4075
4076 if (data[IFLA_VXLAN_GROUP]) {
4077 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4078 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
4079 return -EOPNOTSUPP;
4080 }
4081
4082 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4083 conf->remote_ip.sa.sa_family = AF_INET;
4084 } else if (data[IFLA_VXLAN_GROUP6]) {
4085 if (!IS_ENABLED(CONFIG_IPV6)) {
4086 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4087 return -EPFNOSUPPORT;
4088 }
4089
4090 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4091 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4092 return -EOPNOTSUPP;
4093 }
4094
4095 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4096 conf->remote_ip.sa.sa_family = AF_INET6;
4097 }
4098
4099 if (data[IFLA_VXLAN_LOCAL]) {
4100 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4101 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4102 return -EOPNOTSUPP;
4103 }
4104
4105 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4106 conf->saddr.sa.sa_family = AF_INET;
4107 } else if (data[IFLA_VXLAN_LOCAL6]) {
4108 if (!IS_ENABLED(CONFIG_IPV6)) {
4109 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4110 return -EPFNOSUPPORT;
4111 }
4112
4113 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4114 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4115 return -EOPNOTSUPP;
4116 }
4117
4118 /* TODO: respect scope id */
4119 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4120 conf->saddr.sa.sa_family = AF_INET6;
4121 }
4122
4123 if (data[IFLA_VXLAN_LINK])
4124 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4125
4126 if (data[IFLA_VXLAN_TOS])
4127 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
4128
4129 if (data[IFLA_VXLAN_TTL])
4130 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4131
4132 if (data[IFLA_VXLAN_TTL_INHERIT]) {
4133 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4134 VXLAN_F_TTL_INHERIT, changelink, false,
4135 extack);
4136 if (err)
4137 return err;
4138
4139 }
4140
4141 if (data[IFLA_VXLAN_LABEL])
4142 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4143 IPV6_FLOWLABEL_MASK;
4144 if (data[IFLA_VXLAN_LABEL_POLICY])
4145 conf->label_policy = nla_get_u32(data[IFLA_VXLAN_LABEL_POLICY]);
4146
4147 if (data[IFLA_VXLAN_LEARNING]) {
4148 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4149 VXLAN_F_LEARN, changelink, true,
4150 extack);
4151 if (err)
4152 return err;
4153 } else if (!changelink) {
4154 /* default to learn on a new device */
4155 conf->flags |= VXLAN_F_LEARN;
4156 }
4157
4158 if (data[IFLA_VXLAN_AGEING])
4159 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4160
4161 if (data[IFLA_VXLAN_PROXY]) {
4162 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4163 VXLAN_F_PROXY, changelink, false,
4164 extack);
4165 if (err)
4166 return err;
4167 }
4168
4169 if (data[IFLA_VXLAN_RSC]) {
4170 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4171 VXLAN_F_RSC, changelink, false,
4172 extack);
4173 if (err)
4174 return err;
4175 }
4176
4177 if (data[IFLA_VXLAN_L2MISS]) {
4178 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4179 VXLAN_F_L2MISS, changelink, false,
4180 extack);
4181 if (err)
4182 return err;
4183 }
4184
4185 if (data[IFLA_VXLAN_L3MISS]) {
4186 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4187 VXLAN_F_L3MISS, changelink, false,
4188 extack);
4189 if (err)
4190 return err;
4191 }
4192
4193 if (data[IFLA_VXLAN_LIMIT]) {
4194 if (changelink) {
4195 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4196 "Cannot change limit");
4197 return -EOPNOTSUPP;
4198 }
4199 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4200 }
4201
4202 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4203 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4204 VXLAN_F_COLLECT_METADATA, changelink, false,
4205 extack);
4206 if (err)
4207 return err;
4208 }
4209
4210 if (data[IFLA_VXLAN_PORT_RANGE]) {
4211 if (!changelink) {
4212 const struct ifla_vxlan_port_range *p
4213 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4214 conf->port_min = ntohs(p->low);
4215 conf->port_max = ntohs(p->high);
4216 } else {
4217 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4218 "Cannot change port range");
4219 return -EOPNOTSUPP;
4220 }
4221 }
4222
4223 if (data[IFLA_VXLAN_PORT]) {
4224 if (changelink) {
4225 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4226 "Cannot change port");
4227 return -EOPNOTSUPP;
4228 }
4229 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4230 }
4231
4232 if (data[IFLA_VXLAN_UDP_CSUM]) {
4233 if (changelink) {
4234 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4235 "Cannot change UDP_CSUM flag");
4236 return -EOPNOTSUPP;
4237 }
4238 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4239 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4240 }
4241
4242 if (data[IFLA_VXLAN_LOCALBYPASS]) {
4243 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LOCALBYPASS,
4244 VXLAN_F_LOCALBYPASS, changelink,
4245 true, extack);
4246 if (err)
4247 return err;
4248 } else if (!changelink) {
4249 /* default to local bypass on a new device */
4250 conf->flags |= VXLAN_F_LOCALBYPASS;
4251 }
4252
4253 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4254 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4255 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4256 false, extack);
4257 if (err)
4258 return err;
4259 }
4260
4261 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4262 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4263 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4264 false, extack);
4265 if (err)
4266 return err;
4267 }
4268
4269 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4270 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4271 VXLAN_F_REMCSUM_TX, changelink, false,
4272 extack);
4273 if (err)
4274 return err;
4275 }
4276
4277 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4278 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4279 VXLAN_F_REMCSUM_RX, changelink, false,
4280 extack);
4281 if (err)
4282 return err;
4283 used_bits.vx_flags |= VXLAN_HF_RCO;
4284 used_bits.vx_vni |= ~VXLAN_VNI_MASK;
4285 }
4286
4287 if (data[IFLA_VXLAN_GBP]) {
4288 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4289 VXLAN_F_GBP, changelink, false, extack);
4290 if (err)
4291 return err;
4292 used_bits.vx_flags |= VXLAN_GBP_USED_BITS;
4293 }
4294
4295 if (data[IFLA_VXLAN_GPE]) {
4296 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4297 VXLAN_F_GPE, changelink, false,
4298 extack);
4299 if (err)
4300 return err;
4301
4302 used_bits.vx_flags |= VXLAN_GPE_USED_BITS;
4303 }
4304
4305 if (data[IFLA_VXLAN_RESERVED_BITS]) {
4306 struct vxlanhdr reserved_bits;
4307
4308 if (changelink) {
4309 NL_SET_ERR_MSG_ATTR(extack,
4310 data[IFLA_VXLAN_RESERVED_BITS],
4311 "Cannot change reserved_bits");
4312 return -EOPNOTSUPP;
4313 }
4314
4315 nla_memcpy(&reserved_bits, data[IFLA_VXLAN_RESERVED_BITS],
4316 sizeof(reserved_bits));
4317 if (used_bits.vx_flags & reserved_bits.vx_flags ||
4318 used_bits.vx_vni & reserved_bits.vx_vni) {
4319 __be64 ub_be64, rb_be64;
4320
4321 memcpy(&ub_be64, &used_bits, sizeof(ub_be64));
4322 memcpy(&rb_be64, &reserved_bits, sizeof(rb_be64));
4323
4324 NL_SET_ERR_MSG_ATTR_FMT(extack,
4325 data[IFLA_VXLAN_RESERVED_BITS],
4326 "Used bits %#018llx cannot overlap reserved bits %#018llx",
4327 be64_to_cpu(ub_be64),
4328 be64_to_cpu(rb_be64));
4329 return -EINVAL;
4330 }
4331
4332 conf->reserved_bits = reserved_bits;
4333 } else {
4334 /* For backwards compatibility, only allow reserved fields to be
4335 * used by VXLAN extensions if explicitly requested.
4336 */
4337 conf->reserved_bits = (struct vxlanhdr) {
4338 .vx_flags = ~used_bits.vx_flags,
4339 .vx_vni = ~used_bits.vx_vni,
4340 };
4341 }
4342
4343 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4344 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4345 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4346 false, extack);
4347 if (err)
4348 return err;
4349 }
4350
4351 if (data[IFLA_VXLAN_MC_ROUTE]) {
4352 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_MC_ROUTE,
4353 VXLAN_F_MC_ROUTE, changelink,
4354 true, extack);
4355 if (err)
4356 return err;
4357 }
4358
4359 if (tb[IFLA_MTU]) {
4360 if (changelink) {
4361 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4362 "Cannot change mtu");
4363 return -EOPNOTSUPP;
4364 }
4365 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4366 }
4367
4368 if (data[IFLA_VXLAN_DF])
4369 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4370
4371 if (data[IFLA_VXLAN_VNIFILTER]) {
4372 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_VNIFILTER,
4373 VXLAN_F_VNIFILTER, changelink, false,
4374 extack);
4375 if (err)
4376 return err;
4377
4378 if ((conf->flags & VXLAN_F_VNIFILTER) &&
4379 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
4380 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_VNIFILTER],
4381 "vxlan vnifilter only valid in collect metadata mode");
4382 return -EINVAL;
4383 }
4384 }
4385
4386 return 0;
4387 }
4388
vxlan_newlink(struct net_device * dev,struct rtnl_newlink_params * params,struct netlink_ext_ack * extack)4389 static int vxlan_newlink(struct net_device *dev,
4390 struct rtnl_newlink_params *params,
4391 struct netlink_ext_ack *extack)
4392 {
4393 struct net *link_net = rtnl_newlink_link_net(params);
4394 struct nlattr **data = params->data;
4395 struct nlattr **tb = params->tb;
4396 struct vxlan_config conf;
4397 int err;
4398
4399 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4400 if (err)
4401 return err;
4402
4403 return __vxlan_dev_create(link_net, dev, &conf, extack);
4404 }
4405
vxlan_changelink(struct net_device * dev,struct nlattr * tb[],struct nlattr * data[],struct netlink_ext_ack * extack)4406 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4407 struct nlattr *data[],
4408 struct netlink_ext_ack *extack)
4409 {
4410 struct vxlan_dev *vxlan = netdev_priv(dev);
4411 bool rem_ip_changed, change_igmp;
4412 struct net_device *lowerdev;
4413 struct vxlan_config conf;
4414 struct vxlan_rdst *dst;
4415 int err;
4416
4417 dst = &vxlan->default_dst;
4418 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4419 if (err)
4420 return err;
4421
4422 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4423 vxlan, extack);
4424 if (err)
4425 return err;
4426
4427 if (dst->remote_dev == lowerdev)
4428 lowerdev = NULL;
4429
4430 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4431 extack);
4432 if (err)
4433 return err;
4434
4435 rem_ip_changed = !vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip);
4436 change_igmp = vxlan->dev->flags & IFF_UP &&
4437 (rem_ip_changed ||
4438 dst->remote_ifindex != conf.remote_ifindex);
4439
4440 /* handle default dst entry */
4441 if (rem_ip_changed) {
4442 spin_lock_bh(&vxlan->hash_lock);
4443 if (!vxlan_addr_any(&conf.remote_ip)) {
4444 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4445 &conf.remote_ip,
4446 NUD_REACHABLE | NUD_PERMANENT,
4447 NLM_F_APPEND | NLM_F_CREATE,
4448 vxlan->cfg.dst_port,
4449 conf.vni, conf.vni,
4450 conf.remote_ifindex,
4451 NTF_SELF, 0, true, extack);
4452 if (err) {
4453 spin_unlock_bh(&vxlan->hash_lock);
4454 netdev_adjacent_change_abort(dst->remote_dev,
4455 lowerdev, dev);
4456 return err;
4457 }
4458 }
4459 if (!vxlan_addr_any(&dst->remote_ip))
4460 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4461 dst->remote_ip,
4462 vxlan->cfg.dst_port,
4463 dst->remote_vni,
4464 dst->remote_vni,
4465 dst->remote_ifindex,
4466 true);
4467 spin_unlock_bh(&vxlan->hash_lock);
4468
4469 /* If vni filtering device, also update fdb entries of
4470 * all vnis that were using default remote ip
4471 */
4472 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER) {
4473 err = vxlan_vnilist_update_group(vxlan, &dst->remote_ip,
4474 &conf.remote_ip, extack);
4475 if (err) {
4476 netdev_adjacent_change_abort(dst->remote_dev,
4477 lowerdev, dev);
4478 return err;
4479 }
4480 }
4481 }
4482
4483 if (change_igmp && vxlan_addr_multicast(&dst->remote_ip))
4484 err = vxlan_multicast_leave(vxlan);
4485
4486 if (conf.age_interval != vxlan->cfg.age_interval)
4487 mod_timer(&vxlan->age_timer, jiffies);
4488
4489 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4490 if (lowerdev && lowerdev != dst->remote_dev)
4491 dst->remote_dev = lowerdev;
4492 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4493
4494 if (!err && change_igmp &&
4495 vxlan_addr_multicast(&dst->remote_ip))
4496 err = vxlan_multicast_join(vxlan);
4497
4498 return err;
4499 }
4500
vxlan_dellink(struct net_device * dev,struct list_head * head)4501 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4502 {
4503 struct vxlan_dev *vxlan = netdev_priv(dev);
4504 struct vxlan_fdb_flush_desc desc = {};
4505
4506 vxlan_flush(vxlan, &desc);
4507
4508 list_del(&vxlan->next);
4509 unregister_netdevice_queue(dev, head);
4510 if (vxlan->default_dst.remote_dev)
4511 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4512 }
4513
vxlan_get_size(const struct net_device * dev)4514 static size_t vxlan_get_size(const struct net_device *dev)
4515 {
4516 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4517 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4518 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4519 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4520 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4521 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4522 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4523 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4524 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4525 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LABEL_POLICY */
4526 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4527 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4528 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4529 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4530 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4531 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4532 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4533 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4534 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4535 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4536 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4537 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4538 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4539 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4540 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LOCALBYPASS */
4541 /* IFLA_VXLAN_PORT_RANGE */
4542 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4543 nla_total_size(0) + /* IFLA_VXLAN_GBP */
4544 nla_total_size(0) + /* IFLA_VXLAN_GPE */
4545 nla_total_size(0) + /* IFLA_VXLAN_REMCSUM_NOPARTIAL */
4546 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_VNIFILTER */
4547 /* IFLA_VXLAN_RESERVED_BITS */
4548 nla_total_size(sizeof(struct vxlanhdr)) +
4549 0;
4550 }
4551
vxlan_fill_info(struct sk_buff * skb,const struct net_device * dev)4552 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4553 {
4554 const struct vxlan_dev *vxlan = netdev_priv(dev);
4555 const struct vxlan_rdst *dst = &vxlan->default_dst;
4556 struct ifla_vxlan_port_range ports = {
4557 .low = htons(vxlan->cfg.port_min),
4558 .high = htons(vxlan->cfg.port_max),
4559 };
4560
4561 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4562 goto nla_put_failure;
4563
4564 if (!vxlan_addr_any(&dst->remote_ip)) {
4565 if (dst->remote_ip.sa.sa_family == AF_INET) {
4566 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4567 dst->remote_ip.sin.sin_addr.s_addr))
4568 goto nla_put_failure;
4569 #if IS_ENABLED(CONFIG_IPV6)
4570 } else {
4571 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4572 &dst->remote_ip.sin6.sin6_addr))
4573 goto nla_put_failure;
4574 #endif
4575 }
4576 }
4577
4578 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4579 goto nla_put_failure;
4580
4581 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4582 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4583 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4584 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4585 goto nla_put_failure;
4586 #if IS_ENABLED(CONFIG_IPV6)
4587 } else {
4588 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4589 &vxlan->cfg.saddr.sin6.sin6_addr))
4590 goto nla_put_failure;
4591 #endif
4592 }
4593 }
4594
4595 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4596 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4597 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4598 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4599 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4600 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4601 nla_put_u32(skb, IFLA_VXLAN_LABEL_POLICY, vxlan->cfg.label_policy) ||
4602 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4603 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4604 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4605 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4606 nla_put_u8(skb, IFLA_VXLAN_RSC,
4607 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4608 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4609 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4610 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4611 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4612 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4613 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4614 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4615 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4616 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4617 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4618 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4619 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4620 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4621 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4622 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4623 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4624 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4625 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4626 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)) ||
4627 nla_put_u8(skb, IFLA_VXLAN_LOCALBYPASS,
4628 !!(vxlan->cfg.flags & VXLAN_F_LOCALBYPASS)))
4629 goto nla_put_failure;
4630
4631 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4632 goto nla_put_failure;
4633
4634 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4635 nla_put_flag(skb, IFLA_VXLAN_GBP))
4636 goto nla_put_failure;
4637
4638 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4639 nla_put_flag(skb, IFLA_VXLAN_GPE))
4640 goto nla_put_failure;
4641
4642 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4643 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4644 goto nla_put_failure;
4645
4646 if (vxlan->cfg.flags & VXLAN_F_VNIFILTER &&
4647 nla_put_u8(skb, IFLA_VXLAN_VNIFILTER,
4648 !!(vxlan->cfg.flags & VXLAN_F_VNIFILTER)))
4649 goto nla_put_failure;
4650
4651 if (nla_put(skb, IFLA_VXLAN_RESERVED_BITS,
4652 sizeof(vxlan->cfg.reserved_bits),
4653 &vxlan->cfg.reserved_bits))
4654 goto nla_put_failure;
4655
4656 return 0;
4657
4658 nla_put_failure:
4659 return -EMSGSIZE;
4660 }
4661
vxlan_get_link_net(const struct net_device * dev)4662 static struct net *vxlan_get_link_net(const struct net_device *dev)
4663 {
4664 struct vxlan_dev *vxlan = netdev_priv(dev);
4665
4666 return READ_ONCE(vxlan->net);
4667 }
4668
4669 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4670 .kind = "vxlan",
4671 .maxtype = IFLA_VXLAN_MAX,
4672 .policy = vxlan_policy,
4673 .priv_size = sizeof(struct vxlan_dev),
4674 .setup = vxlan_setup,
4675 .validate = vxlan_validate,
4676 .newlink = vxlan_newlink,
4677 .changelink = vxlan_changelink,
4678 .dellink = vxlan_dellink,
4679 .get_size = vxlan_get_size,
4680 .fill_info = vxlan_fill_info,
4681 .get_link_net = vxlan_get_link_net,
4682 };
4683
vxlan_dev_create(struct net * net,const char * name,u8 name_assign_type,struct vxlan_config * conf)4684 struct net_device *vxlan_dev_create(struct net *net, const char *name,
4685 u8 name_assign_type,
4686 struct vxlan_config *conf)
4687 {
4688 struct nlattr *tb[IFLA_MAX + 1];
4689 struct net_device *dev;
4690 int err;
4691
4692 memset(&tb, 0, sizeof(tb));
4693
4694 dev = rtnl_create_link(net, name, name_assign_type,
4695 &vxlan_link_ops, tb, NULL);
4696 if (IS_ERR(dev))
4697 return dev;
4698
4699 err = __vxlan_dev_create(net, dev, conf, NULL);
4700 if (err < 0) {
4701 free_netdev(dev);
4702 return ERR_PTR(err);
4703 }
4704
4705 err = rtnl_configure_link(dev, NULL, 0, NULL);
4706 if (err < 0) {
4707 LIST_HEAD(list_kill);
4708
4709 vxlan_dellink(dev, &list_kill);
4710 unregister_netdevice_many(&list_kill);
4711 return ERR_PTR(err);
4712 }
4713
4714 return dev;
4715 }
4716 EXPORT_SYMBOL_GPL(vxlan_dev_create);
4717
vxlan_handle_lowerdev_unregister(struct vxlan_net * vn,struct net_device * dev)4718 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4719 struct net_device *dev)
4720 {
4721 struct vxlan_dev *vxlan, *next;
4722 LIST_HEAD(list_kill);
4723
4724 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4725 struct vxlan_rdst *dst = &vxlan->default_dst;
4726
4727 /* In case we created vxlan device with carrier
4728 * and we loose the carrier due to module unload
4729 * we also need to remove vxlan device. In other
4730 * cases, it's not necessary and remote_ifindex
4731 * is 0 here, so no matches.
4732 */
4733 if (dst->remote_ifindex == dev->ifindex)
4734 vxlan_dellink(vxlan->dev, &list_kill);
4735 }
4736
4737 unregister_netdevice_many(&list_kill);
4738 }
4739
vxlan_netdevice_event(struct notifier_block * unused,unsigned long event,void * ptr)4740 static int vxlan_netdevice_event(struct notifier_block *unused,
4741 unsigned long event, void *ptr)
4742 {
4743 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4744 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4745
4746 if (event == NETDEV_UNREGISTER)
4747 vxlan_handle_lowerdev_unregister(vn, dev);
4748 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4749 vxlan_offload_rx_ports(dev, true);
4750 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4751 vxlan_offload_rx_ports(dev, false);
4752
4753 return NOTIFY_DONE;
4754 }
4755
4756 static struct notifier_block vxlan_notifier_block __read_mostly = {
4757 .notifier_call = vxlan_netdevice_event,
4758 };
4759
4760 static void
vxlan_fdb_offloaded_set(struct net_device * dev,struct switchdev_notifier_vxlan_fdb_info * fdb_info)4761 vxlan_fdb_offloaded_set(struct net_device *dev,
4762 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4763 {
4764 struct vxlan_dev *vxlan = netdev_priv(dev);
4765 struct vxlan_rdst *rdst;
4766 struct vxlan_fdb *f;
4767
4768 spin_lock_bh(&vxlan->hash_lock);
4769
4770 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4771 if (!f)
4772 goto out;
4773
4774 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4775 fdb_info->remote_port,
4776 fdb_info->remote_vni,
4777 fdb_info->remote_ifindex);
4778 if (!rdst)
4779 goto out;
4780
4781 rdst->offloaded = fdb_info->offloaded;
4782
4783 out:
4784 spin_unlock_bh(&vxlan->hash_lock);
4785 }
4786
4787 static int
vxlan_fdb_external_learn_add(struct net_device * dev,struct switchdev_notifier_vxlan_fdb_info * fdb_info)4788 vxlan_fdb_external_learn_add(struct net_device *dev,
4789 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4790 {
4791 struct vxlan_dev *vxlan = netdev_priv(dev);
4792 struct netlink_ext_ack *extack;
4793 int err;
4794
4795 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4796
4797 spin_lock_bh(&vxlan->hash_lock);
4798 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4799 NUD_REACHABLE,
4800 NLM_F_CREATE | NLM_F_REPLACE,
4801 fdb_info->remote_port,
4802 fdb_info->vni,
4803 fdb_info->remote_vni,
4804 fdb_info->remote_ifindex,
4805 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4806 0, false, extack);
4807 spin_unlock_bh(&vxlan->hash_lock);
4808
4809 return err;
4810 }
4811
4812 static int
vxlan_fdb_external_learn_del(struct net_device * dev,struct switchdev_notifier_vxlan_fdb_info * fdb_info)4813 vxlan_fdb_external_learn_del(struct net_device *dev,
4814 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4815 {
4816 struct vxlan_dev *vxlan = netdev_priv(dev);
4817 struct vxlan_fdb *f;
4818 int err = 0;
4819
4820 spin_lock_bh(&vxlan->hash_lock);
4821
4822 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4823 if (!f)
4824 err = -ENOENT;
4825 else if (f->flags & NTF_EXT_LEARNED)
4826 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4827 fdb_info->remote_ip,
4828 fdb_info->remote_port,
4829 fdb_info->vni,
4830 fdb_info->remote_vni,
4831 fdb_info->remote_ifindex,
4832 false);
4833
4834 spin_unlock_bh(&vxlan->hash_lock);
4835
4836 return err;
4837 }
4838
vxlan_switchdev_event(struct notifier_block * unused,unsigned long event,void * ptr)4839 static int vxlan_switchdev_event(struct notifier_block *unused,
4840 unsigned long event, void *ptr)
4841 {
4842 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4843 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4844 int err = 0;
4845
4846 switch (event) {
4847 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4848 vxlan_fdb_offloaded_set(dev, ptr);
4849 break;
4850 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4851 fdb_info = ptr;
4852 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4853 if (err) {
4854 err = notifier_from_errno(err);
4855 break;
4856 }
4857 fdb_info->offloaded = true;
4858 vxlan_fdb_offloaded_set(dev, fdb_info);
4859 break;
4860 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4861 fdb_info = ptr;
4862 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4863 if (err) {
4864 err = notifier_from_errno(err);
4865 break;
4866 }
4867 fdb_info->offloaded = false;
4868 vxlan_fdb_offloaded_set(dev, fdb_info);
4869 break;
4870 }
4871
4872 return err;
4873 }
4874
4875 static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4876 .notifier_call = vxlan_switchdev_event,
4877 };
4878
vxlan_fdb_nh_flush(struct nexthop * nh)4879 static void vxlan_fdb_nh_flush(struct nexthop *nh)
4880 {
4881 struct vxlan_fdb *fdb;
4882 struct vxlan_dev *vxlan;
4883
4884 rcu_read_lock();
4885 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4886 vxlan = rcu_dereference(fdb->vdev);
4887 WARN_ON(!vxlan);
4888 spin_lock_bh(&vxlan->hash_lock);
4889 if (!hlist_unhashed(&fdb->fdb_node))
4890 vxlan_fdb_destroy(vxlan, fdb, false, false);
4891 spin_unlock_bh(&vxlan->hash_lock);
4892 }
4893 rcu_read_unlock();
4894 }
4895
vxlan_nexthop_event(struct notifier_block * nb,unsigned long event,void * ptr)4896 static int vxlan_nexthop_event(struct notifier_block *nb,
4897 unsigned long event, void *ptr)
4898 {
4899 struct nh_notifier_info *info = ptr;
4900 struct nexthop *nh;
4901
4902 if (event != NEXTHOP_EVENT_DEL)
4903 return NOTIFY_DONE;
4904
4905 nh = nexthop_find_by_id(info->net, info->id);
4906 if (!nh)
4907 return NOTIFY_DONE;
4908
4909 vxlan_fdb_nh_flush(nh);
4910
4911 return NOTIFY_DONE;
4912 }
4913
vxlan_init_net(struct net * net)4914 static __net_init int vxlan_init_net(struct net *net)
4915 {
4916 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4917 unsigned int h;
4918
4919 INIT_LIST_HEAD(&vn->vxlan_list);
4920 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4921
4922 for (h = 0; h < PORT_HASH_SIZE; ++h)
4923 INIT_HLIST_HEAD(&vn->sock_list[h]);
4924
4925 return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4926 NULL);
4927 }
4928
vxlan_destroy_tunnels(struct vxlan_net * vn,struct list_head * dev_to_kill)4929 static void __net_exit vxlan_destroy_tunnels(struct vxlan_net *vn,
4930 struct list_head *dev_to_kill)
4931 {
4932 struct vxlan_dev *vxlan, *next;
4933
4934 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next)
4935 vxlan_dellink(vxlan->dev, dev_to_kill);
4936 }
4937
vxlan_exit_rtnl(struct net * net,struct list_head * dev_to_kill)4938 static void __net_exit vxlan_exit_rtnl(struct net *net,
4939 struct list_head *dev_to_kill)
4940 {
4941 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4942
4943 ASSERT_RTNL_NET(net);
4944
4945 __unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4946 vxlan_destroy_tunnels(vn, dev_to_kill);
4947 }
4948
vxlan_exit_net(struct net * net)4949 static void __net_exit vxlan_exit_net(struct net *net)
4950 {
4951 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4952 unsigned int h;
4953
4954 for (h = 0; h < PORT_HASH_SIZE; ++h)
4955 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4956 }
4957
4958 static struct pernet_operations vxlan_net_ops = {
4959 .init = vxlan_init_net,
4960 .exit_rtnl = vxlan_exit_rtnl,
4961 .exit = vxlan_exit_net,
4962 .id = &vxlan_net_id,
4963 .size = sizeof(struct vxlan_net),
4964 };
4965
vxlan_init_module(void)4966 static int __init vxlan_init_module(void)
4967 {
4968 int rc;
4969
4970 rc = register_pernet_subsys(&vxlan_net_ops);
4971 if (rc)
4972 goto out1;
4973
4974 rc = register_netdevice_notifier(&vxlan_notifier_block);
4975 if (rc)
4976 goto out2;
4977
4978 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4979 if (rc)
4980 goto out3;
4981
4982 rc = rtnl_link_register(&vxlan_link_ops);
4983 if (rc)
4984 goto out4;
4985
4986 rc = vxlan_vnifilter_init();
4987 if (rc)
4988 goto out5;
4989
4990 return 0;
4991 out5:
4992 rtnl_link_unregister(&vxlan_link_ops);
4993 out4:
4994 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4995 out3:
4996 unregister_netdevice_notifier(&vxlan_notifier_block);
4997 out2:
4998 unregister_pernet_subsys(&vxlan_net_ops);
4999 out1:
5000 return rc;
5001 }
5002 late_initcall(vxlan_init_module);
5003
vxlan_cleanup_module(void)5004 static void __exit vxlan_cleanup_module(void)
5005 {
5006 vxlan_vnifilter_uninit();
5007 rtnl_link_unregister(&vxlan_link_ops);
5008 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
5009 unregister_netdevice_notifier(&vxlan_notifier_block);
5010 unregister_pernet_subsys(&vxlan_net_ops);
5011 /* rcu_barrier() is called by netns */
5012 }
5013 module_exit(vxlan_cleanup_module);
5014
5015 MODULE_LICENSE("GPL");
5016 MODULE_VERSION(VXLAN_VERSION);
5017 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
5018 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
5019 MODULE_ALIAS_RTNL_LINK("vxlan");
5020