Loading...
Note: File does not exist in v6.8.
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 <net/arp.h>
19#include <net/ndisc.h>
20#include <net/ipv6_stubs.h>
21#include <net/ip.h>
22#include <net/icmp.h>
23#include <net/rtnetlink.h>
24#include <net/inet_ecn.h>
25#include <net/net_namespace.h>
26#include <net/netns/generic.h>
27#include <net/tun_proto.h>
28#include <net/vxlan.h>
29#include <net/nexthop.h>
30
31#if IS_ENABLED(CONFIG_IPV6)
32#include <net/ip6_tunnel.h>
33#include <net/ip6_checksum.h>
34#endif
35
36#define VXLAN_VERSION "0.1"
37
38#define PORT_HASH_BITS 8
39#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
40#define FDB_AGE_DEFAULT 300 /* 5 min */
41#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
42
43/* UDP port for VXLAN traffic.
44 * The IANA assigned port is 4789, but the Linux default is 8472
45 * for compatibility with early adopters.
46 */
47static unsigned short vxlan_port __read_mostly = 8472;
48module_param_named(udp_port, vxlan_port, ushort, 0444);
49MODULE_PARM_DESC(udp_port, "Destination UDP port");
50
51static bool log_ecn_error = true;
52module_param(log_ecn_error, bool, 0644);
53MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
54
55static unsigned int vxlan_net_id;
56static struct rtnl_link_ops vxlan_link_ops;
57
58static const u8 all_zeros_mac[ETH_ALEN + 2];
59
60static int vxlan_sock_add(struct vxlan_dev *vxlan);
61
62static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
63
64/* per-network namespace private data for this module */
65struct vxlan_net {
66 struct list_head vxlan_list;
67 struct hlist_head sock_list[PORT_HASH_SIZE];
68 spinlock_t sock_lock;
69 struct notifier_block nexthop_notifier_block;
70};
71
72/* Forwarding table entry */
73struct vxlan_fdb {
74 struct hlist_node hlist; /* linked list of entries */
75 struct rcu_head rcu;
76 unsigned long updated; /* jiffies */
77 unsigned long used;
78 struct list_head remotes;
79 u8 eth_addr[ETH_ALEN];
80 u16 state; /* see ndm_state */
81 __be32 vni;
82 u16 flags; /* see ndm_flags and below */
83 struct list_head nh_list;
84 struct nexthop __rcu *nh;
85 struct vxlan_dev __rcu *vdev;
86};
87
88#define NTF_VXLAN_ADDED_BY_USER 0x100
89
90/* salt for hash table */
91static u32 vxlan_salt __read_mostly;
92
93static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
94{
95 return vs->flags & VXLAN_F_COLLECT_METADATA ||
96 ip_tunnel_collect_metadata();
97}
98
99#if IS_ENABLED(CONFIG_IPV6)
100static inline
101bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
102{
103 if (a->sa.sa_family != b->sa.sa_family)
104 return false;
105 if (a->sa.sa_family == AF_INET6)
106 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
107 else
108 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
109}
110
111static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
112{
113 if (nla_len(nla) >= sizeof(struct in6_addr)) {
114 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
115 ip->sa.sa_family = AF_INET6;
116 return 0;
117 } else if (nla_len(nla) >= sizeof(__be32)) {
118 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
119 ip->sa.sa_family = AF_INET;
120 return 0;
121 } else {
122 return -EAFNOSUPPORT;
123 }
124}
125
126static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
127 const union vxlan_addr *ip)
128{
129 if (ip->sa.sa_family == AF_INET6)
130 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
131 else
132 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
133}
134
135#else /* !CONFIG_IPV6 */
136
137static inline
138bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
139{
140 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
141}
142
143static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
144{
145 if (nla_len(nla) >= sizeof(struct in6_addr)) {
146 return -EAFNOSUPPORT;
147 } else if (nla_len(nla) >= sizeof(__be32)) {
148 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
149 ip->sa.sa_family = AF_INET;
150 return 0;
151 } else {
152 return -EAFNOSUPPORT;
153 }
154}
155
156static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
157 const union vxlan_addr *ip)
158{
159 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
160}
161#endif
162
163/* Virtual Network hash table head */
164static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
165{
166 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
167}
168
169/* Socket hash table head */
170static inline struct hlist_head *vs_head(struct net *net, __be16 port)
171{
172 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
173
174 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
175}
176
177/* First remote destination for a forwarding entry.
178 * Guaranteed to be non-NULL because remotes are never deleted.
179 */
180static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
181{
182 if (rcu_access_pointer(fdb->nh))
183 return NULL;
184 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
185}
186
187static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
188{
189 if (rcu_access_pointer(fdb->nh))
190 return NULL;
191 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
192}
193
194/* Find VXLAN socket based on network namespace, address family, UDP port,
195 * enabled unshareable flags and socket device binding (see l3mdev with
196 * non-default VRF).
197 */
198static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
199 __be16 port, u32 flags, int ifindex)
200{
201 struct vxlan_sock *vs;
202
203 flags &= VXLAN_F_RCV_FLAGS;
204
205 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
206 if (inet_sk(vs->sock->sk)->inet_sport == port &&
207 vxlan_get_sk_family(vs) == family &&
208 vs->flags == flags &&
209 vs->sock->sk->sk_bound_dev_if == ifindex)
210 return vs;
211 }
212 return NULL;
213}
214
215static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
216 __be32 vni)
217{
218 struct vxlan_dev_node *node;
219
220 /* For flow based devices, map all packets to VNI 0 */
221 if (vs->flags & VXLAN_F_COLLECT_METADATA)
222 vni = 0;
223
224 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
225 if (node->vxlan->default_dst.remote_vni != vni)
226 continue;
227
228 if (IS_ENABLED(CONFIG_IPV6)) {
229 const struct vxlan_config *cfg = &node->vxlan->cfg;
230
231 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
232 cfg->remote_ifindex != ifindex)
233 continue;
234 }
235
236 return node->vxlan;
237 }
238
239 return NULL;
240}
241
242/* Look up VNI in a per net namespace table */
243static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
244 __be32 vni, sa_family_t family,
245 __be16 port, u32 flags)
246{
247 struct vxlan_sock *vs;
248
249 vs = vxlan_find_sock(net, family, port, flags, ifindex);
250 if (!vs)
251 return NULL;
252
253 return vxlan_vs_find_vni(vs, ifindex, vni);
254}
255
256/* Fill in neighbour message in skbuff. */
257static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
258 const struct vxlan_fdb *fdb,
259 u32 portid, u32 seq, int type, unsigned int flags,
260 const struct vxlan_rdst *rdst)
261{
262 unsigned long now = jiffies;
263 struct nda_cacheinfo ci;
264 bool send_ip, send_eth;
265 struct nlmsghdr *nlh;
266 struct nexthop *nh;
267 struct ndmsg *ndm;
268 int nh_family;
269 u32 nh_id;
270
271 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
272 if (nlh == NULL)
273 return -EMSGSIZE;
274
275 ndm = nlmsg_data(nlh);
276 memset(ndm, 0, sizeof(*ndm));
277
278 send_eth = send_ip = true;
279
280 rcu_read_lock();
281 nh = rcu_dereference(fdb->nh);
282 if (nh) {
283 nh_family = nexthop_get_family(nh);
284 nh_id = nh->id;
285 }
286 rcu_read_unlock();
287
288 if (type == RTM_GETNEIGH) {
289 if (rdst) {
290 send_ip = !vxlan_addr_any(&rdst->remote_ip);
291 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
292 } else if (nh) {
293 ndm->ndm_family = nh_family;
294 }
295 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 } else
297 ndm->ndm_family = AF_BRIDGE;
298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
300 ndm->ndm_flags = fdb->flags;
301 if (rdst && rdst->offloaded)
302 ndm->ndm_flags |= NTF_OFFLOADED;
303 ndm->ndm_type = RTN_UNICAST;
304
305 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
306 nla_put_s32(skb, NDA_LINK_NETNSID,
307 peernet2id(dev_net(vxlan->dev), vxlan->net)))
308 goto nla_put_failure;
309
310 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
311 goto nla_put_failure;
312 if (nh) {
313 if (nla_put_u32(skb, NDA_NH_ID, nh_id))
314 goto nla_put_failure;
315 } else if (rdst) {
316 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST,
317 &rdst->remote_ip))
318 goto nla_put_failure;
319
320 if (rdst->remote_port &&
321 rdst->remote_port != vxlan->cfg.dst_port &&
322 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
323 goto nla_put_failure;
324 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
325 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
326 goto nla_put_failure;
327 if (rdst->remote_ifindex &&
328 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
329 goto nla_put_failure;
330 }
331
332 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
333 nla_put_u32(skb, NDA_SRC_VNI,
334 be32_to_cpu(fdb->vni)))
335 goto nla_put_failure;
336
337 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
338 ci.ndm_confirmed = 0;
339 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
340 ci.ndm_refcnt = 0;
341
342 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
343 goto nla_put_failure;
344
345 nlmsg_end(skb, nlh);
346 return 0;
347
348nla_put_failure:
349 nlmsg_cancel(skb, nlh);
350 return -EMSGSIZE;
351}
352
353static inline size_t vxlan_nlmsg_size(void)
354{
355 return NLMSG_ALIGN(sizeof(struct ndmsg))
356 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
357 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
358 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
359 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
360 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
361 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
362 + nla_total_size(sizeof(struct nda_cacheinfo));
363}
364
365static void __vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
366 struct vxlan_rdst *rd, int type)
367{
368 struct net *net = dev_net(vxlan->dev);
369 struct sk_buff *skb;
370 int err = -ENOBUFS;
371
372 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
373 if (skb == NULL)
374 goto errout;
375
376 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
377 if (err < 0) {
378 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
379 WARN_ON(err == -EMSGSIZE);
380 kfree_skb(skb);
381 goto errout;
382 }
383
384 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
385 return;
386errout:
387 if (err < 0)
388 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
389}
390
391static void vxlan_fdb_switchdev_notifier_info(const struct vxlan_dev *vxlan,
392 const struct vxlan_fdb *fdb,
393 const struct vxlan_rdst *rd,
394 struct netlink_ext_ack *extack,
395 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
396{
397 fdb_info->info.dev = vxlan->dev;
398 fdb_info->info.extack = extack;
399 fdb_info->remote_ip = rd->remote_ip;
400 fdb_info->remote_port = rd->remote_port;
401 fdb_info->remote_vni = rd->remote_vni;
402 fdb_info->remote_ifindex = rd->remote_ifindex;
403 memcpy(fdb_info->eth_addr, fdb->eth_addr, ETH_ALEN);
404 fdb_info->vni = fdb->vni;
405 fdb_info->offloaded = rd->offloaded;
406 fdb_info->added_by_user = fdb->flags & NTF_VXLAN_ADDED_BY_USER;
407}
408
409static int vxlan_fdb_switchdev_call_notifiers(struct vxlan_dev *vxlan,
410 struct vxlan_fdb *fdb,
411 struct vxlan_rdst *rd,
412 bool adding,
413 struct netlink_ext_ack *extack)
414{
415 struct switchdev_notifier_vxlan_fdb_info info;
416 enum switchdev_notifier_type notifier_type;
417 int ret;
418
419 if (WARN_ON(!rd))
420 return 0;
421
422 notifier_type = adding ? SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE
423 : SWITCHDEV_VXLAN_FDB_DEL_TO_DEVICE;
424 vxlan_fdb_switchdev_notifier_info(vxlan, fdb, rd, NULL, &info);
425 ret = call_switchdev_notifiers(notifier_type, vxlan->dev,
426 &info.info, extack);
427 return notifier_to_errno(ret);
428}
429
430static int vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
431 struct vxlan_rdst *rd, int type, bool swdev_notify,
432 struct netlink_ext_ack *extack)
433{
434 int err;
435
436 if (swdev_notify && rd) {
437 switch (type) {
438 case RTM_NEWNEIGH:
439 err = vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
440 true, extack);
441 if (err)
442 return err;
443 break;
444 case RTM_DELNEIGH:
445 vxlan_fdb_switchdev_call_notifiers(vxlan, fdb, rd,
446 false, extack);
447 break;
448 }
449 }
450
451 __vxlan_fdb_notify(vxlan, fdb, rd, type);
452 return 0;
453}
454
455static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
456{
457 struct vxlan_dev *vxlan = netdev_priv(dev);
458 struct vxlan_fdb f = {
459 .state = NUD_STALE,
460 };
461 struct vxlan_rdst remote = {
462 .remote_ip = *ipa, /* goes to NDA_DST */
463 .remote_vni = cpu_to_be32(VXLAN_N_VID),
464 };
465
466 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
467}
468
469static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
470{
471 struct vxlan_fdb f = {
472 .state = NUD_STALE,
473 };
474 struct vxlan_rdst remote = { };
475
476 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
477
478 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH, true, NULL);
479}
480
481/* Hash Ethernet address */
482static u32 eth_hash(const unsigned char *addr)
483{
484 u64 value = get_unaligned((u64 *)addr);
485
486 /* only want 6 bytes */
487#ifdef __BIG_ENDIAN
488 value >>= 16;
489#else
490 value <<= 16;
491#endif
492 return hash_64(value, FDB_HASH_BITS);
493}
494
495static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
496{
497 /* use 1 byte of OUI and 3 bytes of NIC */
498 u32 key = get_unaligned((u32 *)(addr + 2));
499
500 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
501}
502
503static u32 fdb_head_index(struct vxlan_dev *vxlan, const u8 *mac, __be32 vni)
504{
505 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
506 return eth_vni_hash(mac, vni);
507 else
508 return eth_hash(mac);
509}
510
511/* Hash chain to use given mac address */
512static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
513 const u8 *mac, __be32 vni)
514{
515 return &vxlan->fdb_head[fdb_head_index(vxlan, mac, vni)];
516}
517
518/* Look up Ethernet address in forwarding table */
519static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
520 const u8 *mac, __be32 vni)
521{
522 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
523 struct vxlan_fdb *f;
524
525 hlist_for_each_entry_rcu(f, head, hlist) {
526 if (ether_addr_equal(mac, f->eth_addr)) {
527 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
528 if (vni == f->vni)
529 return f;
530 } else {
531 return f;
532 }
533 }
534 }
535
536 return NULL;
537}
538
539static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
540 const u8 *mac, __be32 vni)
541{
542 struct vxlan_fdb *f;
543
544 f = __vxlan_find_mac(vxlan, mac, vni);
545 if (f && f->used != jiffies)
546 f->used = jiffies;
547
548 return f;
549}
550
551/* caller should hold vxlan->hash_lock */
552static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
553 union vxlan_addr *ip, __be16 port,
554 __be32 vni, __u32 ifindex)
555{
556 struct vxlan_rdst *rd;
557
558 list_for_each_entry(rd, &f->remotes, list) {
559 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
560 rd->remote_port == port &&
561 rd->remote_vni == vni &&
562 rd->remote_ifindex == ifindex)
563 return rd;
564 }
565
566 return NULL;
567}
568
569int vxlan_fdb_find_uc(struct net_device *dev, const u8 *mac, __be32 vni,
570 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
571{
572 struct vxlan_dev *vxlan = netdev_priv(dev);
573 u8 eth_addr[ETH_ALEN + 2] = { 0 };
574 struct vxlan_rdst *rdst;
575 struct vxlan_fdb *f;
576 int rc = 0;
577
578 if (is_multicast_ether_addr(mac) ||
579 is_zero_ether_addr(mac))
580 return -EINVAL;
581
582 ether_addr_copy(eth_addr, mac);
583
584 rcu_read_lock();
585
586 f = __vxlan_find_mac(vxlan, eth_addr, vni);
587 if (!f) {
588 rc = -ENOENT;
589 goto out;
590 }
591
592 rdst = first_remote_rcu(f);
593 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, NULL, fdb_info);
594
595out:
596 rcu_read_unlock();
597 return rc;
598}
599EXPORT_SYMBOL_GPL(vxlan_fdb_find_uc);
600
601static int vxlan_fdb_notify_one(struct notifier_block *nb,
602 const struct vxlan_dev *vxlan,
603 const struct vxlan_fdb *f,
604 const struct vxlan_rdst *rdst,
605 struct netlink_ext_ack *extack)
606{
607 struct switchdev_notifier_vxlan_fdb_info fdb_info;
608 int rc;
609
610 vxlan_fdb_switchdev_notifier_info(vxlan, f, rdst, extack, &fdb_info);
611 rc = nb->notifier_call(nb, SWITCHDEV_VXLAN_FDB_ADD_TO_DEVICE,
612 &fdb_info);
613 return notifier_to_errno(rc);
614}
615
616int vxlan_fdb_replay(const struct net_device *dev, __be32 vni,
617 struct notifier_block *nb,
618 struct netlink_ext_ack *extack)
619{
620 struct vxlan_dev *vxlan;
621 struct vxlan_rdst *rdst;
622 struct vxlan_fdb *f;
623 unsigned int h;
624 int rc = 0;
625
626 if (!netif_is_vxlan(dev))
627 return -EINVAL;
628 vxlan = netdev_priv(dev);
629
630 for (h = 0; h < FDB_HASH_SIZE; ++h) {
631 spin_lock_bh(&vxlan->hash_lock[h]);
632 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist) {
633 if (f->vni == vni) {
634 list_for_each_entry(rdst, &f->remotes, list) {
635 rc = vxlan_fdb_notify_one(nb, vxlan,
636 f, rdst,
637 extack);
638 if (rc)
639 goto unlock;
640 }
641 }
642 }
643 spin_unlock_bh(&vxlan->hash_lock[h]);
644 }
645 return 0;
646
647unlock:
648 spin_unlock_bh(&vxlan->hash_lock[h]);
649 return rc;
650}
651EXPORT_SYMBOL_GPL(vxlan_fdb_replay);
652
653void vxlan_fdb_clear_offload(const struct net_device *dev, __be32 vni)
654{
655 struct vxlan_dev *vxlan;
656 struct vxlan_rdst *rdst;
657 struct vxlan_fdb *f;
658 unsigned int h;
659
660 if (!netif_is_vxlan(dev))
661 return;
662 vxlan = netdev_priv(dev);
663
664 for (h = 0; h < FDB_HASH_SIZE; ++h) {
665 spin_lock_bh(&vxlan->hash_lock[h]);
666 hlist_for_each_entry(f, &vxlan->fdb_head[h], hlist)
667 if (f->vni == vni)
668 list_for_each_entry(rdst, &f->remotes, list)
669 rdst->offloaded = false;
670 spin_unlock_bh(&vxlan->hash_lock[h]);
671 }
672
673}
674EXPORT_SYMBOL_GPL(vxlan_fdb_clear_offload);
675
676/* Replace destination of unicast mac */
677static int vxlan_fdb_replace(struct vxlan_fdb *f,
678 union vxlan_addr *ip, __be16 port, __be32 vni,
679 __u32 ifindex, struct vxlan_rdst *oldrd)
680{
681 struct vxlan_rdst *rd;
682
683 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
684 if (rd)
685 return 0;
686
687 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
688 if (!rd)
689 return 0;
690
691 *oldrd = *rd;
692 dst_cache_reset(&rd->dst_cache);
693 rd->remote_ip = *ip;
694 rd->remote_port = port;
695 rd->remote_vni = vni;
696 rd->remote_ifindex = ifindex;
697 rd->offloaded = false;
698 return 1;
699}
700
701/* Add/update destinations for multicast */
702static int vxlan_fdb_append(struct vxlan_fdb *f,
703 union vxlan_addr *ip, __be16 port, __be32 vni,
704 __u32 ifindex, struct vxlan_rdst **rdp)
705{
706 struct vxlan_rdst *rd;
707
708 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
709 if (rd)
710 return 0;
711
712 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
713 if (rd == NULL)
714 return -ENOBUFS;
715
716 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
717 kfree(rd);
718 return -ENOBUFS;
719 }
720
721 rd->remote_ip = *ip;
722 rd->remote_port = port;
723 rd->offloaded = false;
724 rd->remote_vni = vni;
725 rd->remote_ifindex = ifindex;
726
727 list_add_tail_rcu(&rd->list, &f->remotes);
728
729 *rdp = rd;
730 return 1;
731}
732
733static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
734 unsigned int off,
735 struct vxlanhdr *vh, size_t hdrlen,
736 __be32 vni_field,
737 struct gro_remcsum *grc,
738 bool nopartial)
739{
740 size_t start, offset;
741
742 if (skb->remcsum_offload)
743 return vh;
744
745 if (!NAPI_GRO_CB(skb)->csum_valid)
746 return NULL;
747
748 start = vxlan_rco_start(vni_field);
749 offset = start + vxlan_rco_offset(vni_field);
750
751 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
752 start, offset, grc, nopartial);
753
754 skb->remcsum_offload = 1;
755
756 return vh;
757}
758
759static struct sk_buff *vxlan_gro_receive(struct sock *sk,
760 struct list_head *head,
761 struct sk_buff *skb)
762{
763 struct sk_buff *pp = NULL;
764 struct sk_buff *p;
765 struct vxlanhdr *vh, *vh2;
766 unsigned int hlen, off_vx;
767 int flush = 1;
768 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
769 __be32 flags;
770 struct gro_remcsum grc;
771
772 skb_gro_remcsum_init(&grc);
773
774 off_vx = skb_gro_offset(skb);
775 hlen = off_vx + sizeof(*vh);
776 vh = skb_gro_header_fast(skb, off_vx);
777 if (skb_gro_header_hard(skb, hlen)) {
778 vh = skb_gro_header_slow(skb, hlen, off_vx);
779 if (unlikely(!vh))
780 goto out;
781 }
782
783 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
784
785 flags = vh->vx_flags;
786
787 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
788 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
789 vh->vx_vni, &grc,
790 !!(vs->flags &
791 VXLAN_F_REMCSUM_NOPARTIAL));
792
793 if (!vh)
794 goto out;
795 }
796
797 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
798
799 list_for_each_entry(p, head, list) {
800 if (!NAPI_GRO_CB(p)->same_flow)
801 continue;
802
803 vh2 = (struct vxlanhdr *)(p->data + off_vx);
804 if (vh->vx_flags != vh2->vx_flags ||
805 vh->vx_vni != vh2->vx_vni) {
806 NAPI_GRO_CB(p)->same_flow = 0;
807 continue;
808 }
809 }
810
811 pp = call_gro_receive(eth_gro_receive, head, skb);
812 flush = 0;
813
814out:
815 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
816
817 return pp;
818}
819
820static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
821{
822 /* Sets 'skb->inner_mac_header' since we are always called with
823 * 'skb->encapsulation' set.
824 */
825 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
826}
827
828static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan, const u8 *mac,
829 __u16 state, __be32 src_vni,
830 __u16 ndm_flags)
831{
832 struct vxlan_fdb *f;
833
834 f = kmalloc(sizeof(*f), GFP_ATOMIC);
835 if (!f)
836 return NULL;
837 f->state = state;
838 f->flags = ndm_flags;
839 f->updated = f->used = jiffies;
840 f->vni = src_vni;
841 f->nh = NULL;
842 RCU_INIT_POINTER(f->vdev, vxlan);
843 INIT_LIST_HEAD(&f->nh_list);
844 INIT_LIST_HEAD(&f->remotes);
845 memcpy(f->eth_addr, mac, ETH_ALEN);
846
847 return f;
848}
849
850static void vxlan_fdb_insert(struct vxlan_dev *vxlan, const u8 *mac,
851 __be32 src_vni, struct vxlan_fdb *f)
852{
853 ++vxlan->addrcnt;
854 hlist_add_head_rcu(&f->hlist,
855 vxlan_fdb_head(vxlan, mac, src_vni));
856}
857
858static int vxlan_fdb_nh_update(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
859 u32 nhid, struct netlink_ext_ack *extack)
860{
861 struct nexthop *old_nh = rtnl_dereference(fdb->nh);
862 struct nexthop *nh;
863 int err = -EINVAL;
864
865 if (old_nh && old_nh->id == nhid)
866 return 0;
867
868 nh = nexthop_find_by_id(vxlan->net, nhid);
869 if (!nh) {
870 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
871 goto err_inval;
872 }
873
874 if (nh) {
875 if (!nexthop_get(nh)) {
876 NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
877 nh = NULL;
878 goto err_inval;
879 }
880 if (!nexthop_is_fdb(nh)) {
881 NL_SET_ERR_MSG(extack, "Nexthop is not a fdb nexthop");
882 goto err_inval;
883 }
884
885 if (!nexthop_is_multipath(nh)) {
886 NL_SET_ERR_MSG(extack, "Nexthop is not a multipath group");
887 goto err_inval;
888 }
889
890 /* check nexthop group family */
891 switch (vxlan->default_dst.remote_ip.sa.sa_family) {
892 case AF_INET:
893 if (!nexthop_has_v4(nh)) {
894 err = -EAFNOSUPPORT;
895 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
896 goto err_inval;
897 }
898 break;
899 case AF_INET6:
900 if (nexthop_has_v4(nh)) {
901 err = -EAFNOSUPPORT;
902 NL_SET_ERR_MSG(extack, "Nexthop group family not supported");
903 goto err_inval;
904 }
905 }
906 }
907
908 if (old_nh) {
909 list_del_rcu(&fdb->nh_list);
910 nexthop_put(old_nh);
911 }
912 rcu_assign_pointer(fdb->nh, nh);
913 list_add_tail_rcu(&fdb->nh_list, &nh->fdb_list);
914 return 1;
915
916err_inval:
917 if (nh)
918 nexthop_put(nh);
919 return err;
920}
921
922static int vxlan_fdb_create(struct vxlan_dev *vxlan,
923 const u8 *mac, union vxlan_addr *ip,
924 __u16 state, __be16 port, __be32 src_vni,
925 __be32 vni, __u32 ifindex, __u16 ndm_flags,
926 u32 nhid, struct vxlan_fdb **fdb,
927 struct netlink_ext_ack *extack)
928{
929 struct vxlan_rdst *rd = NULL;
930 struct vxlan_fdb *f;
931 int rc;
932
933 if (vxlan->cfg.addrmax &&
934 vxlan->addrcnt >= vxlan->cfg.addrmax)
935 return -ENOSPC;
936
937 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
938 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
939 if (!f)
940 return -ENOMEM;
941
942 if (nhid)
943 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
944 else
945 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
946 if (rc < 0)
947 goto errout;
948
949 *fdb = f;
950
951 return 0;
952
953errout:
954 kfree(f);
955 return rc;
956}
957
958static void __vxlan_fdb_free(struct vxlan_fdb *f)
959{
960 struct vxlan_rdst *rd, *nd;
961 struct nexthop *nh;
962
963 nh = rcu_dereference_raw(f->nh);
964 if (nh) {
965 rcu_assign_pointer(f->nh, NULL);
966 rcu_assign_pointer(f->vdev, NULL);
967 nexthop_put(nh);
968 }
969
970 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
971 dst_cache_destroy(&rd->dst_cache);
972 kfree(rd);
973 }
974 kfree(f);
975}
976
977static void vxlan_fdb_free(struct rcu_head *head)
978{
979 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
980
981 __vxlan_fdb_free(f);
982}
983
984static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
985 bool do_notify, bool swdev_notify)
986{
987 struct vxlan_rdst *rd;
988
989 netdev_dbg(vxlan->dev, "delete %pM\n", f->eth_addr);
990
991 --vxlan->addrcnt;
992 if (do_notify) {
993 if (rcu_access_pointer(f->nh))
994 vxlan_fdb_notify(vxlan, f, NULL, RTM_DELNEIGH,
995 swdev_notify, NULL);
996 else
997 list_for_each_entry(rd, &f->remotes, list)
998 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH,
999 swdev_notify, NULL);
1000 }
1001
1002 hlist_del_rcu(&f->hlist);
1003 list_del_rcu(&f->nh_list);
1004 call_rcu(&f->rcu, vxlan_fdb_free);
1005}
1006
1007static void vxlan_dst_free(struct rcu_head *head)
1008{
1009 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
1010
1011 dst_cache_destroy(&rd->dst_cache);
1012 kfree(rd);
1013}
1014
1015static int vxlan_fdb_update_existing(struct vxlan_dev *vxlan,
1016 union vxlan_addr *ip,
1017 __u16 state, __u16 flags,
1018 __be16 port, __be32 vni,
1019 __u32 ifindex, __u16 ndm_flags,
1020 struct vxlan_fdb *f, u32 nhid,
1021 bool swdev_notify,
1022 struct netlink_ext_ack *extack)
1023{
1024 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1025 struct vxlan_rdst *rd = NULL;
1026 struct vxlan_rdst oldrd;
1027 int notify = 0;
1028 int rc = 0;
1029 int err;
1030
1031 if (nhid && !rcu_access_pointer(f->nh)) {
1032 NL_SET_ERR_MSG(extack,
1033 "Cannot replace an existing non nexthop fdb with a nexthop");
1034 return -EOPNOTSUPP;
1035 }
1036
1037 if (nhid && (flags & NLM_F_APPEND)) {
1038 NL_SET_ERR_MSG(extack,
1039 "Cannot append to a nexthop fdb");
1040 return -EOPNOTSUPP;
1041 }
1042
1043 /* Do not allow an externally learned entry to take over an entry added
1044 * by the user.
1045 */
1046 if (!(fdb_flags & NTF_EXT_LEARNED) ||
1047 !(f->flags & NTF_VXLAN_ADDED_BY_USER)) {
1048 if (f->state != state) {
1049 f->state = state;
1050 f->updated = jiffies;
1051 notify = 1;
1052 }
1053 if (f->flags != fdb_flags) {
1054 f->flags = fdb_flags;
1055 f->updated = jiffies;
1056 notify = 1;
1057 }
1058 }
1059
1060 if ((flags & NLM_F_REPLACE)) {
1061 /* Only change unicasts */
1062 if (!(is_multicast_ether_addr(f->eth_addr) ||
1063 is_zero_ether_addr(f->eth_addr))) {
1064 if (nhid) {
1065 rc = vxlan_fdb_nh_update(vxlan, f, nhid, extack);
1066 if (rc < 0)
1067 return rc;
1068 } else {
1069 rc = vxlan_fdb_replace(f, ip, port, vni,
1070 ifindex, &oldrd);
1071 }
1072 notify |= rc;
1073 } else {
1074 NL_SET_ERR_MSG(extack, "Cannot replace non-unicast fdb entries");
1075 return -EOPNOTSUPP;
1076 }
1077 }
1078 if ((flags & NLM_F_APPEND) &&
1079 (is_multicast_ether_addr(f->eth_addr) ||
1080 is_zero_ether_addr(f->eth_addr))) {
1081 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
1082
1083 if (rc < 0)
1084 return rc;
1085 notify |= rc;
1086 }
1087
1088 if (ndm_flags & NTF_USE)
1089 f->used = jiffies;
1090
1091 if (notify) {
1092 if (rd == NULL)
1093 rd = first_remote_rtnl(f);
1094
1095 err = vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH,
1096 swdev_notify, extack);
1097 if (err)
1098 goto err_notify;
1099 }
1100
1101 return 0;
1102
1103err_notify:
1104 if (nhid)
1105 return err;
1106 if ((flags & NLM_F_REPLACE) && rc)
1107 *rd = oldrd;
1108 else if ((flags & NLM_F_APPEND) && rc) {
1109 list_del_rcu(&rd->list);
1110 call_rcu(&rd->rcu, vxlan_dst_free);
1111 }
1112 return err;
1113}
1114
1115static int vxlan_fdb_update_create(struct vxlan_dev *vxlan,
1116 const u8 *mac, union vxlan_addr *ip,
1117 __u16 state, __u16 flags,
1118 __be16 port, __be32 src_vni, __be32 vni,
1119 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1120 bool swdev_notify,
1121 struct netlink_ext_ack *extack)
1122{
1123 __u16 fdb_flags = (ndm_flags & ~NTF_USE);
1124 struct vxlan_fdb *f;
1125 int rc;
1126
1127 /* Disallow replace to add a multicast entry */
1128 if ((flags & NLM_F_REPLACE) &&
1129 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
1130 return -EOPNOTSUPP;
1131
1132 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
1133 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
1134 vni, ifindex, fdb_flags, nhid, &f, extack);
1135 if (rc < 0)
1136 return rc;
1137
1138 vxlan_fdb_insert(vxlan, mac, src_vni, f);
1139 rc = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH,
1140 swdev_notify, extack);
1141 if (rc)
1142 goto err_notify;
1143
1144 return 0;
1145
1146err_notify:
1147 vxlan_fdb_destroy(vxlan, f, false, false);
1148 return rc;
1149}
1150
1151/* Add new entry to forwarding table -- assumes lock held */
1152static int vxlan_fdb_update(struct vxlan_dev *vxlan,
1153 const u8 *mac, union vxlan_addr *ip,
1154 __u16 state, __u16 flags,
1155 __be16 port, __be32 src_vni, __be32 vni,
1156 __u32 ifindex, __u16 ndm_flags, u32 nhid,
1157 bool swdev_notify,
1158 struct netlink_ext_ack *extack)
1159{
1160 struct vxlan_fdb *f;
1161
1162 f = __vxlan_find_mac(vxlan, mac, src_vni);
1163 if (f) {
1164 if (flags & NLM_F_EXCL) {
1165 netdev_dbg(vxlan->dev,
1166 "lost race to create %pM\n", mac);
1167 return -EEXIST;
1168 }
1169
1170 return vxlan_fdb_update_existing(vxlan, ip, state, flags, port,
1171 vni, ifindex, ndm_flags, f,
1172 nhid, swdev_notify, extack);
1173 } else {
1174 if (!(flags & NLM_F_CREATE))
1175 return -ENOENT;
1176
1177 return vxlan_fdb_update_create(vxlan, mac, ip, state, flags,
1178 port, src_vni, vni, ifindex,
1179 ndm_flags, nhid, swdev_notify,
1180 extack);
1181 }
1182}
1183
1184static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
1185 struct vxlan_rdst *rd, bool swdev_notify)
1186{
1187 list_del_rcu(&rd->list);
1188 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH, swdev_notify, NULL);
1189 call_rcu(&rd->rcu, vxlan_dst_free);
1190}
1191
1192static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
1193 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
1194 __be32 *vni, u32 *ifindex, u32 *nhid)
1195{
1196 struct net *net = dev_net(vxlan->dev);
1197 int err;
1198
1199 if (tb[NDA_NH_ID] && (tb[NDA_DST] || tb[NDA_VNI] || tb[NDA_IFINDEX] ||
1200 tb[NDA_PORT]))
1201 return -EINVAL;
1202
1203 if (tb[NDA_DST]) {
1204 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
1205 if (err)
1206 return err;
1207 } else {
1208 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
1209
1210 if (remote->sa.sa_family == AF_INET) {
1211 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
1212 ip->sa.sa_family = AF_INET;
1213#if IS_ENABLED(CONFIG_IPV6)
1214 } else {
1215 ip->sin6.sin6_addr = in6addr_any;
1216 ip->sa.sa_family = AF_INET6;
1217#endif
1218 }
1219 }
1220
1221 if (tb[NDA_PORT]) {
1222 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
1223 return -EINVAL;
1224 *port = nla_get_be16(tb[NDA_PORT]);
1225 } else {
1226 *port = vxlan->cfg.dst_port;
1227 }
1228
1229 if (tb[NDA_VNI]) {
1230 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
1231 return -EINVAL;
1232 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1233 } else {
1234 *vni = vxlan->default_dst.remote_vni;
1235 }
1236
1237 if (tb[NDA_SRC_VNI]) {
1238 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
1239 return -EINVAL;
1240 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
1241 } else {
1242 *src_vni = vxlan->default_dst.remote_vni;
1243 }
1244
1245 if (tb[NDA_IFINDEX]) {
1246 struct net_device *tdev;
1247
1248 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
1249 return -EINVAL;
1250 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
1251 tdev = __dev_get_by_index(net, *ifindex);
1252 if (!tdev)
1253 return -EADDRNOTAVAIL;
1254 } else {
1255 *ifindex = 0;
1256 }
1257
1258 if (tb[NDA_NH_ID])
1259 *nhid = nla_get_u32(tb[NDA_NH_ID]);
1260 else
1261 *nhid = 0;
1262
1263 return 0;
1264}
1265
1266/* Add static entry (via netlink) */
1267static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1268 struct net_device *dev,
1269 const unsigned char *addr, u16 vid, u16 flags,
1270 struct netlink_ext_ack *extack)
1271{
1272 struct vxlan_dev *vxlan = netdev_priv(dev);
1273 /* struct net *net = dev_net(vxlan->dev); */
1274 union vxlan_addr ip;
1275 __be16 port;
1276 __be32 src_vni, vni;
1277 u32 ifindex, nhid;
1278 u32 hash_index;
1279 int err;
1280
1281 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1282 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1283 ndm->ndm_state);
1284 return -EINVAL;
1285 }
1286
1287 if (!tb || (!tb[NDA_DST] && !tb[NDA_NH_ID]))
1288 return -EINVAL;
1289
1290 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1291 &nhid);
1292 if (err)
1293 return err;
1294
1295 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1296 return -EAFNOSUPPORT;
1297
1298 hash_index = fdb_head_index(vxlan, addr, src_vni);
1299 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1300 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
1301 port, src_vni, vni, ifindex,
1302 ndm->ndm_flags | NTF_VXLAN_ADDED_BY_USER,
1303 nhid, true, extack);
1304 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1305
1306 return err;
1307}
1308
1309static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1310 const unsigned char *addr, union vxlan_addr ip,
1311 __be16 port, __be32 src_vni, __be32 vni,
1312 u32 ifindex, bool swdev_notify)
1313{
1314 struct vxlan_rdst *rd = NULL;
1315 struct vxlan_fdb *f;
1316 int err = -ENOENT;
1317
1318 f = vxlan_find_mac(vxlan, addr, src_vni);
1319 if (!f)
1320 return err;
1321
1322 if (!vxlan_addr_any(&ip)) {
1323 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1324 if (!rd)
1325 goto out;
1326 }
1327
1328 /* remove a destination if it's not the only one on the list,
1329 * otherwise destroy the fdb entry
1330 */
1331 if (rd && !list_is_singular(&f->remotes)) {
1332 vxlan_fdb_dst_destroy(vxlan, f, rd, swdev_notify);
1333 goto out;
1334 }
1335
1336 vxlan_fdb_destroy(vxlan, f, true, swdev_notify);
1337
1338out:
1339 return 0;
1340}
1341
1342/* Delete entry (via netlink) */
1343static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1344 struct net_device *dev,
1345 const unsigned char *addr, u16 vid)
1346{
1347 struct vxlan_dev *vxlan = netdev_priv(dev);
1348 union vxlan_addr ip;
1349 __be32 src_vni, vni;
1350 u32 ifindex, nhid;
1351 u32 hash_index;
1352 __be16 port;
1353 int err;
1354
1355 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex,
1356 &nhid);
1357 if (err)
1358 return err;
1359
1360 hash_index = fdb_head_index(vxlan, addr, src_vni);
1361 spin_lock_bh(&vxlan->hash_lock[hash_index]);
1362 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1363 true);
1364 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
1365
1366 return err;
1367}
1368
1369/* Dump forwarding table */
1370static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1371 struct net_device *dev,
1372 struct net_device *filter_dev, int *idx)
1373{
1374 struct vxlan_dev *vxlan = netdev_priv(dev);
1375 unsigned int h;
1376 int err = 0;
1377
1378 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1379 struct vxlan_fdb *f;
1380
1381 rcu_read_lock();
1382 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1383 struct vxlan_rdst *rd;
1384
1385 if (rcu_access_pointer(f->nh)) {
1386 if (*idx < cb->args[2])
1387 goto skip_nh;
1388 err = vxlan_fdb_info(skb, vxlan, f,
1389 NETLINK_CB(cb->skb).portid,
1390 cb->nlh->nlmsg_seq,
1391 RTM_NEWNEIGH,
1392 NLM_F_MULTI, NULL);
1393 if (err < 0) {
1394 rcu_read_unlock();
1395 goto out;
1396 }
1397skip_nh:
1398 *idx += 1;
1399 continue;
1400 }
1401
1402 list_for_each_entry_rcu(rd, &f->remotes, list) {
1403 if (*idx < cb->args[2])
1404 goto skip;
1405
1406 err = vxlan_fdb_info(skb, vxlan, f,
1407 NETLINK_CB(cb->skb).portid,
1408 cb->nlh->nlmsg_seq,
1409 RTM_NEWNEIGH,
1410 NLM_F_MULTI, rd);
1411 if (err < 0) {
1412 rcu_read_unlock();
1413 goto out;
1414 }
1415skip:
1416 *idx += 1;
1417 }
1418 }
1419 rcu_read_unlock();
1420 }
1421out:
1422 return err;
1423}
1424
1425static int vxlan_fdb_get(struct sk_buff *skb,
1426 struct nlattr *tb[],
1427 struct net_device *dev,
1428 const unsigned char *addr,
1429 u16 vid, u32 portid, u32 seq,
1430 struct netlink_ext_ack *extack)
1431{
1432 struct vxlan_dev *vxlan = netdev_priv(dev);
1433 struct vxlan_fdb *f;
1434 __be32 vni;
1435 int err;
1436
1437 if (tb[NDA_VNI])
1438 vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
1439 else
1440 vni = vxlan->default_dst.remote_vni;
1441
1442 rcu_read_lock();
1443
1444 f = __vxlan_find_mac(vxlan, addr, vni);
1445 if (!f) {
1446 NL_SET_ERR_MSG(extack, "Fdb entry not found");
1447 err = -ENOENT;
1448 goto errout;
1449 }
1450
1451 err = vxlan_fdb_info(skb, vxlan, f, portid, seq,
1452 RTM_NEWNEIGH, 0, first_remote_rcu(f));
1453errout:
1454 rcu_read_unlock();
1455 return err;
1456}
1457
1458/* Watch incoming packets to learn mapping between Ethernet address
1459 * and Tunnel endpoint.
1460 * Return true if packet is bogus and should be dropped.
1461 */
1462static bool vxlan_snoop(struct net_device *dev,
1463 union vxlan_addr *src_ip, const u8 *src_mac,
1464 u32 src_ifindex, __be32 vni)
1465{
1466 struct vxlan_dev *vxlan = netdev_priv(dev);
1467 struct vxlan_fdb *f;
1468 u32 ifindex = 0;
1469
1470#if IS_ENABLED(CONFIG_IPV6)
1471 if (src_ip->sa.sa_family == AF_INET6 &&
1472 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1473 ifindex = src_ifindex;
1474#endif
1475
1476 f = vxlan_find_mac(vxlan, src_mac, vni);
1477 if (likely(f)) {
1478 struct vxlan_rdst *rdst = first_remote_rcu(f);
1479
1480 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1481 rdst->remote_ifindex == ifindex))
1482 return false;
1483
1484 /* Don't migrate static entries, drop packets */
1485 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1486 return true;
1487
1488 /* Don't override an fdb with nexthop with a learnt entry */
1489 if (rcu_access_pointer(f->nh))
1490 return true;
1491
1492 if (net_ratelimit())
1493 netdev_info(dev,
1494 "%pM migrated from %pIS to %pIS\n",
1495 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1496
1497 rdst->remote_ip = *src_ip;
1498 f->updated = jiffies;
1499 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH, true, NULL);
1500 } else {
1501 u32 hash_index = fdb_head_index(vxlan, src_mac, vni);
1502
1503 /* learned new entry */
1504 spin_lock(&vxlan->hash_lock[hash_index]);
1505
1506 /* close off race between vxlan_flush and incoming packets */
1507 if (netif_running(dev))
1508 vxlan_fdb_update(vxlan, src_mac, src_ip,
1509 NUD_REACHABLE,
1510 NLM_F_EXCL|NLM_F_CREATE,
1511 vxlan->cfg.dst_port,
1512 vni,
1513 vxlan->default_dst.remote_vni,
1514 ifindex, NTF_SELF, 0, true, NULL);
1515 spin_unlock(&vxlan->hash_lock[hash_index]);
1516 }
1517
1518 return false;
1519}
1520
1521/* See if multicast group is already in use by other ID */
1522static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1523{
1524 struct vxlan_dev *vxlan;
1525 struct vxlan_sock *sock4;
1526#if IS_ENABLED(CONFIG_IPV6)
1527 struct vxlan_sock *sock6;
1528#endif
1529 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1530
1531 sock4 = rtnl_dereference(dev->vn4_sock);
1532
1533 /* The vxlan_sock is only used by dev, leaving group has
1534 * no effect on other vxlan devices.
1535 */
1536 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1537 return false;
1538#if IS_ENABLED(CONFIG_IPV6)
1539 sock6 = rtnl_dereference(dev->vn6_sock);
1540 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1541 return false;
1542#endif
1543
1544 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1545 if (!netif_running(vxlan->dev) || vxlan == dev)
1546 continue;
1547
1548 if (family == AF_INET &&
1549 rtnl_dereference(vxlan->vn4_sock) != sock4)
1550 continue;
1551#if IS_ENABLED(CONFIG_IPV6)
1552 if (family == AF_INET6 &&
1553 rtnl_dereference(vxlan->vn6_sock) != sock6)
1554 continue;
1555#endif
1556
1557 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1558 &dev->default_dst.remote_ip))
1559 continue;
1560
1561 if (vxlan->default_dst.remote_ifindex !=
1562 dev->default_dst.remote_ifindex)
1563 continue;
1564
1565 return true;
1566 }
1567
1568 return false;
1569}
1570
1571static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1572{
1573 struct vxlan_net *vn;
1574
1575 if (!vs)
1576 return false;
1577 if (!refcount_dec_and_test(&vs->refcnt))
1578 return false;
1579
1580 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1581 spin_lock(&vn->sock_lock);
1582 hlist_del_rcu(&vs->hlist);
1583 udp_tunnel_notify_del_rx_port(vs->sock,
1584 (vs->flags & VXLAN_F_GPE) ?
1585 UDP_TUNNEL_TYPE_VXLAN_GPE :
1586 UDP_TUNNEL_TYPE_VXLAN);
1587 spin_unlock(&vn->sock_lock);
1588
1589 return true;
1590}
1591
1592static void vxlan_sock_release(struct vxlan_dev *vxlan)
1593{
1594 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1595#if IS_ENABLED(CONFIG_IPV6)
1596 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1597
1598 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1599#endif
1600
1601 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1602 synchronize_net();
1603
1604 vxlan_vs_del_dev(vxlan);
1605
1606 if (__vxlan_sock_release_prep(sock4)) {
1607 udp_tunnel_sock_release(sock4->sock);
1608 kfree(sock4);
1609 }
1610
1611#if IS_ENABLED(CONFIG_IPV6)
1612 if (__vxlan_sock_release_prep(sock6)) {
1613 udp_tunnel_sock_release(sock6->sock);
1614 kfree(sock6);
1615 }
1616#endif
1617}
1618
1619/* Update multicast group membership when first VNI on
1620 * multicast address is brought up
1621 */
1622static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1623{
1624 struct sock *sk;
1625 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1626 int ifindex = vxlan->default_dst.remote_ifindex;
1627 int ret = -EINVAL;
1628
1629 if (ip->sa.sa_family == AF_INET) {
1630 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1631 struct ip_mreqn mreq = {
1632 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1633 .imr_ifindex = ifindex,
1634 };
1635
1636 sk = sock4->sock->sk;
1637 lock_sock(sk);
1638 ret = ip_mc_join_group(sk, &mreq);
1639 release_sock(sk);
1640#if IS_ENABLED(CONFIG_IPV6)
1641 } else {
1642 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1643
1644 sk = sock6->sock->sk;
1645 lock_sock(sk);
1646 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1647 &ip->sin6.sin6_addr);
1648 release_sock(sk);
1649#endif
1650 }
1651
1652 return ret;
1653}
1654
1655/* Inverse of vxlan_igmp_join when last VNI is brought down */
1656static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1657{
1658 struct sock *sk;
1659 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1660 int ifindex = vxlan->default_dst.remote_ifindex;
1661 int ret = -EINVAL;
1662
1663 if (ip->sa.sa_family == AF_INET) {
1664 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1665 struct ip_mreqn mreq = {
1666 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1667 .imr_ifindex = ifindex,
1668 };
1669
1670 sk = sock4->sock->sk;
1671 lock_sock(sk);
1672 ret = ip_mc_leave_group(sk, &mreq);
1673 release_sock(sk);
1674#if IS_ENABLED(CONFIG_IPV6)
1675 } else {
1676 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1677
1678 sk = sock6->sock->sk;
1679 lock_sock(sk);
1680 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1681 &ip->sin6.sin6_addr);
1682 release_sock(sk);
1683#endif
1684 }
1685
1686 return ret;
1687}
1688
1689static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1690 struct sk_buff *skb, u32 vxflags)
1691{
1692 size_t start, offset;
1693
1694 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1695 goto out;
1696
1697 start = vxlan_rco_start(unparsed->vx_vni);
1698 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1699
1700 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1701 return false;
1702
1703 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1704 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1705out:
1706 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1707 unparsed->vx_vni &= VXLAN_VNI_MASK;
1708 return true;
1709}
1710
1711static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1712 struct sk_buff *skb, u32 vxflags,
1713 struct vxlan_metadata *md)
1714{
1715 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1716 struct metadata_dst *tun_dst;
1717
1718 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1719 goto out;
1720
1721 md->gbp = ntohs(gbp->policy_id);
1722
1723 tun_dst = (struct metadata_dst *)skb_dst(skb);
1724 if (tun_dst) {
1725 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1726 tun_dst->u.tun_info.options_len = sizeof(*md);
1727 }
1728 if (gbp->dont_learn)
1729 md->gbp |= VXLAN_GBP_DONT_LEARN;
1730
1731 if (gbp->policy_applied)
1732 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1733
1734 /* In flow-based mode, GBP is carried in dst_metadata */
1735 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1736 skb->mark = md->gbp;
1737out:
1738 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1739}
1740
1741static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1742 __be16 *protocol,
1743 struct sk_buff *skb, u32 vxflags)
1744{
1745 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1746
1747 /* Need to have Next Protocol set for interfaces in GPE mode. */
1748 if (!gpe->np_applied)
1749 return false;
1750 /* "The initial version is 0. If a receiver does not support the
1751 * version indicated it MUST drop the packet.
1752 */
1753 if (gpe->version != 0)
1754 return false;
1755 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1756 * processing MUST occur." However, we don't implement OAM
1757 * processing, thus drop the packet.
1758 */
1759 if (gpe->oam_flag)
1760 return false;
1761
1762 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1763 if (!*protocol)
1764 return false;
1765
1766 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1767 return true;
1768}
1769
1770static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1771 struct vxlan_sock *vs,
1772 struct sk_buff *skb, __be32 vni)
1773{
1774 union vxlan_addr saddr;
1775 u32 ifindex = skb->dev->ifindex;
1776
1777 skb_reset_mac_header(skb);
1778 skb->protocol = eth_type_trans(skb, vxlan->dev);
1779 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1780
1781 /* Ignore packet loops (and multicast echo) */
1782 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1783 return false;
1784
1785 /* Get address from the outer IP header */
1786 if (vxlan_get_sk_family(vs) == AF_INET) {
1787 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1788 saddr.sa.sa_family = AF_INET;
1789#if IS_ENABLED(CONFIG_IPV6)
1790 } else {
1791 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1792 saddr.sa.sa_family = AF_INET6;
1793#endif
1794 }
1795
1796 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1797 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1798 return false;
1799
1800 return true;
1801}
1802
1803static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1804 struct sk_buff *skb)
1805{
1806 int err = 0;
1807
1808 if (vxlan_get_sk_family(vs) == AF_INET)
1809 err = IP_ECN_decapsulate(oiph, skb);
1810#if IS_ENABLED(CONFIG_IPV6)
1811 else
1812 err = IP6_ECN_decapsulate(oiph, skb);
1813#endif
1814
1815 if (unlikely(err) && log_ecn_error) {
1816 if (vxlan_get_sk_family(vs) == AF_INET)
1817 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1818 &((struct iphdr *)oiph)->saddr,
1819 ((struct iphdr *)oiph)->tos);
1820 else
1821 net_info_ratelimited("non-ECT from %pI6\n",
1822 &((struct ipv6hdr *)oiph)->saddr);
1823 }
1824 return err <= 1;
1825}
1826
1827/* Callback from net/ipv4/udp.c to receive packets */
1828static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1829{
1830 struct vxlan_dev *vxlan;
1831 struct vxlan_sock *vs;
1832 struct vxlanhdr unparsed;
1833 struct vxlan_metadata _md;
1834 struct vxlan_metadata *md = &_md;
1835 __be16 protocol = htons(ETH_P_TEB);
1836 bool raw_proto = false;
1837 void *oiph;
1838 __be32 vni = 0;
1839
1840 /* Need UDP and VXLAN header to be present */
1841 if (!pskb_may_pull(skb, VXLAN_HLEN))
1842 goto drop;
1843
1844 unparsed = *vxlan_hdr(skb);
1845 /* VNI flag always required to be set */
1846 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1847 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1848 ntohl(vxlan_hdr(skb)->vx_flags),
1849 ntohl(vxlan_hdr(skb)->vx_vni));
1850 /* Return non vxlan pkt */
1851 goto drop;
1852 }
1853 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1854 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1855
1856 vs = rcu_dereference_sk_user_data(sk);
1857 if (!vs)
1858 goto drop;
1859
1860 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1861
1862 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1863 if (!vxlan)
1864 goto drop;
1865
1866 /* For backwards compatibility, only allow reserved fields to be
1867 * used by VXLAN extensions if explicitly requested.
1868 */
1869 if (vs->flags & VXLAN_F_GPE) {
1870 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1871 goto drop;
1872 raw_proto = true;
1873 }
1874
1875 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1876 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1877 goto drop;
1878
1879 if (vs->flags & VXLAN_F_REMCSUM_RX)
1880 if (unlikely(!vxlan_remcsum(&unparsed, skb, vs->flags)))
1881 goto drop;
1882
1883 if (vxlan_collect_metadata(vs)) {
1884 struct metadata_dst *tun_dst;
1885
1886 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1887 key32_to_tunnel_id(vni), sizeof(*md));
1888
1889 if (!tun_dst)
1890 goto drop;
1891
1892 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1893
1894 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1895 } else {
1896 memset(md, 0, sizeof(*md));
1897 }
1898
1899 if (vs->flags & VXLAN_F_GBP)
1900 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1901 /* Note that GBP and GPE can never be active together. This is
1902 * ensured in vxlan_dev_configure.
1903 */
1904
1905 if (unparsed.vx_flags || unparsed.vx_vni) {
1906 /* If there are any unprocessed flags remaining treat
1907 * this as a malformed packet. This behavior diverges from
1908 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1909 * in reserved fields are to be ignored. The approach here
1910 * maintains compatibility with previous stack code, and also
1911 * is more robust and provides a little more security in
1912 * adding extensions to VXLAN.
1913 */
1914 goto drop;
1915 }
1916
1917 if (!raw_proto) {
1918 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1919 goto drop;
1920 } else {
1921 skb_reset_mac_header(skb);
1922 skb->dev = vxlan->dev;
1923 skb->pkt_type = PACKET_HOST;
1924 }
1925
1926 oiph = skb_network_header(skb);
1927 skb_reset_network_header(skb);
1928
1929 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1930 ++vxlan->dev->stats.rx_frame_errors;
1931 ++vxlan->dev->stats.rx_errors;
1932 goto drop;
1933 }
1934
1935 rcu_read_lock();
1936
1937 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1938 rcu_read_unlock();
1939 atomic_long_inc(&vxlan->dev->rx_dropped);
1940 goto drop;
1941 }
1942
1943 dev_sw_netstats_rx_add(vxlan->dev, skb->len);
1944 gro_cells_receive(&vxlan->gro_cells, skb);
1945
1946 rcu_read_unlock();
1947
1948 return 0;
1949
1950drop:
1951 /* Consume bad packet */
1952 kfree_skb(skb);
1953 return 0;
1954}
1955
1956/* Callback from net/ipv{4,6}/udp.c to check that we have a VNI for errors */
1957static int vxlan_err_lookup(struct sock *sk, struct sk_buff *skb)
1958{
1959 struct vxlan_dev *vxlan;
1960 struct vxlan_sock *vs;
1961 struct vxlanhdr *hdr;
1962 __be32 vni;
1963
1964 if (!pskb_may_pull(skb, skb_transport_offset(skb) + VXLAN_HLEN))
1965 return -EINVAL;
1966
1967 hdr = vxlan_hdr(skb);
1968
1969 if (!(hdr->vx_flags & VXLAN_HF_VNI))
1970 return -EINVAL;
1971
1972 vs = rcu_dereference_sk_user_data(sk);
1973 if (!vs)
1974 return -ENOENT;
1975
1976 vni = vxlan_vni(hdr->vx_vni);
1977 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1978 if (!vxlan)
1979 return -ENOENT;
1980
1981 return 0;
1982}
1983
1984static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1985{
1986 struct vxlan_dev *vxlan = netdev_priv(dev);
1987 struct arphdr *parp;
1988 u8 *arpptr, *sha;
1989 __be32 sip, tip;
1990 struct neighbour *n;
1991
1992 if (dev->flags & IFF_NOARP)
1993 goto out;
1994
1995 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1996 dev->stats.tx_dropped++;
1997 goto out;
1998 }
1999 parp = arp_hdr(skb);
2000
2001 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
2002 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
2003 parp->ar_pro != htons(ETH_P_IP) ||
2004 parp->ar_op != htons(ARPOP_REQUEST) ||
2005 parp->ar_hln != dev->addr_len ||
2006 parp->ar_pln != 4)
2007 goto out;
2008 arpptr = (u8 *)parp + sizeof(struct arphdr);
2009 sha = arpptr;
2010 arpptr += dev->addr_len; /* sha */
2011 memcpy(&sip, arpptr, sizeof(sip));
2012 arpptr += sizeof(sip);
2013 arpptr += dev->addr_len; /* tha */
2014 memcpy(&tip, arpptr, sizeof(tip));
2015
2016 if (ipv4_is_loopback(tip) ||
2017 ipv4_is_multicast(tip))
2018 goto out;
2019
2020 n = neigh_lookup(&arp_tbl, &tip, dev);
2021
2022 if (n) {
2023 struct vxlan_fdb *f;
2024 struct sk_buff *reply;
2025
2026 if (!(n->nud_state & NUD_CONNECTED)) {
2027 neigh_release(n);
2028 goto out;
2029 }
2030
2031 f = vxlan_find_mac(vxlan, n->ha, vni);
2032 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2033 /* bridge-local neighbor */
2034 neigh_release(n);
2035 goto out;
2036 }
2037
2038 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
2039 n->ha, sha);
2040
2041 neigh_release(n);
2042
2043 if (reply == NULL)
2044 goto out;
2045
2046 skb_reset_mac_header(reply);
2047 __skb_pull(reply, skb_network_offset(reply));
2048 reply->ip_summed = CHECKSUM_UNNECESSARY;
2049 reply->pkt_type = PACKET_HOST;
2050
2051 if (netif_rx_ni(reply) == NET_RX_DROP)
2052 dev->stats.rx_dropped++;
2053 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2054 union vxlan_addr ipa = {
2055 .sin.sin_addr.s_addr = tip,
2056 .sin.sin_family = AF_INET,
2057 };
2058
2059 vxlan_ip_miss(dev, &ipa);
2060 }
2061out:
2062 consume_skb(skb);
2063 return NETDEV_TX_OK;
2064}
2065
2066#if IS_ENABLED(CONFIG_IPV6)
2067static struct sk_buff *vxlan_na_create(struct sk_buff *request,
2068 struct neighbour *n, bool isrouter)
2069{
2070 struct net_device *dev = request->dev;
2071 struct sk_buff *reply;
2072 struct nd_msg *ns, *na;
2073 struct ipv6hdr *pip6;
2074 u8 *daddr;
2075 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
2076 int ns_olen;
2077 int i, len;
2078
2079 if (dev == NULL || !pskb_may_pull(request, request->len))
2080 return NULL;
2081
2082 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
2083 sizeof(*na) + na_olen + dev->needed_tailroom;
2084 reply = alloc_skb(len, GFP_ATOMIC);
2085 if (reply == NULL)
2086 return NULL;
2087
2088 reply->protocol = htons(ETH_P_IPV6);
2089 reply->dev = dev;
2090 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
2091 skb_push(reply, sizeof(struct ethhdr));
2092 skb_reset_mac_header(reply);
2093
2094 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
2095
2096 daddr = eth_hdr(request)->h_source;
2097 ns_olen = request->len - skb_network_offset(request) -
2098 sizeof(struct ipv6hdr) - sizeof(*ns);
2099 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
2100 if (!ns->opt[i + 1]) {
2101 kfree_skb(reply);
2102 return NULL;
2103 }
2104 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
2105 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
2106 break;
2107 }
2108 }
2109
2110 /* Ethernet header */
2111 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
2112 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
2113 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
2114 reply->protocol = htons(ETH_P_IPV6);
2115
2116 skb_pull(reply, sizeof(struct ethhdr));
2117 skb_reset_network_header(reply);
2118 skb_put(reply, sizeof(struct ipv6hdr));
2119
2120 /* IPv6 header */
2121
2122 pip6 = ipv6_hdr(reply);
2123 memset(pip6, 0, sizeof(struct ipv6hdr));
2124 pip6->version = 6;
2125 pip6->priority = ipv6_hdr(request)->priority;
2126 pip6->nexthdr = IPPROTO_ICMPV6;
2127 pip6->hop_limit = 255;
2128 pip6->daddr = ipv6_hdr(request)->saddr;
2129 pip6->saddr = *(struct in6_addr *)n->primary_key;
2130
2131 skb_pull(reply, sizeof(struct ipv6hdr));
2132 skb_reset_transport_header(reply);
2133
2134 /* Neighbor Advertisement */
2135 na = skb_put_zero(reply, sizeof(*na) + na_olen);
2136 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
2137 na->icmph.icmp6_router = isrouter;
2138 na->icmph.icmp6_override = 1;
2139 na->icmph.icmp6_solicited = 1;
2140 na->target = ns->target;
2141 ether_addr_copy(&na->opt[2], n->ha);
2142 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
2143 na->opt[1] = na_olen >> 3;
2144
2145 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
2146 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
2147 csum_partial(na, sizeof(*na)+na_olen, 0));
2148
2149 pip6->payload_len = htons(sizeof(*na)+na_olen);
2150
2151 skb_push(reply, sizeof(struct ipv6hdr));
2152
2153 reply->ip_summed = CHECKSUM_UNNECESSARY;
2154
2155 return reply;
2156}
2157
2158static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
2159{
2160 struct vxlan_dev *vxlan = netdev_priv(dev);
2161 const struct in6_addr *daddr;
2162 const struct ipv6hdr *iphdr;
2163 struct inet6_dev *in6_dev;
2164 struct neighbour *n;
2165 struct nd_msg *msg;
2166
2167 rcu_read_lock();
2168 in6_dev = __in6_dev_get(dev);
2169 if (!in6_dev)
2170 goto out;
2171
2172 iphdr = ipv6_hdr(skb);
2173 daddr = &iphdr->daddr;
2174 msg = (struct nd_msg *)(iphdr + 1);
2175
2176 if (ipv6_addr_loopback(daddr) ||
2177 ipv6_addr_is_multicast(&msg->target))
2178 goto out;
2179
2180 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
2181
2182 if (n) {
2183 struct vxlan_fdb *f;
2184 struct sk_buff *reply;
2185
2186 if (!(n->nud_state & NUD_CONNECTED)) {
2187 neigh_release(n);
2188 goto out;
2189 }
2190
2191 f = vxlan_find_mac(vxlan, n->ha, vni);
2192 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
2193 /* bridge-local neighbor */
2194 neigh_release(n);
2195 goto out;
2196 }
2197
2198 reply = vxlan_na_create(skb, n,
2199 !!(f ? f->flags & NTF_ROUTER : 0));
2200
2201 neigh_release(n);
2202
2203 if (reply == NULL)
2204 goto out;
2205
2206 if (netif_rx_ni(reply) == NET_RX_DROP)
2207 dev->stats.rx_dropped++;
2208
2209 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
2210 union vxlan_addr ipa = {
2211 .sin6.sin6_addr = msg->target,
2212 .sin6.sin6_family = AF_INET6,
2213 };
2214
2215 vxlan_ip_miss(dev, &ipa);
2216 }
2217
2218out:
2219 rcu_read_unlock();
2220 consume_skb(skb);
2221 return NETDEV_TX_OK;
2222}
2223#endif
2224
2225static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
2226{
2227 struct vxlan_dev *vxlan = netdev_priv(dev);
2228 struct neighbour *n;
2229
2230 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
2231 return false;
2232
2233 n = NULL;
2234 switch (ntohs(eth_hdr(skb)->h_proto)) {
2235 case ETH_P_IP:
2236 {
2237 struct iphdr *pip;
2238
2239 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
2240 return false;
2241 pip = ip_hdr(skb);
2242 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
2243 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2244 union vxlan_addr ipa = {
2245 .sin.sin_addr.s_addr = pip->daddr,
2246 .sin.sin_family = AF_INET,
2247 };
2248
2249 vxlan_ip_miss(dev, &ipa);
2250 return false;
2251 }
2252
2253 break;
2254 }
2255#if IS_ENABLED(CONFIG_IPV6)
2256 case ETH_P_IPV6:
2257 {
2258 struct ipv6hdr *pip6;
2259
2260 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
2261 return false;
2262 pip6 = ipv6_hdr(skb);
2263 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
2264 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
2265 union vxlan_addr ipa = {
2266 .sin6.sin6_addr = pip6->daddr,
2267 .sin6.sin6_family = AF_INET6,
2268 };
2269
2270 vxlan_ip_miss(dev, &ipa);
2271 return false;
2272 }
2273
2274 break;
2275 }
2276#endif
2277 default:
2278 return false;
2279 }
2280
2281 if (n) {
2282 bool diff;
2283
2284 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
2285 if (diff) {
2286 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
2287 dev->addr_len);
2288 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
2289 }
2290 neigh_release(n);
2291 return diff;
2292 }
2293
2294 return false;
2295}
2296
2297static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
2298 struct vxlan_metadata *md)
2299{
2300 struct vxlanhdr_gbp *gbp;
2301
2302 if (!md->gbp)
2303 return;
2304
2305 gbp = (struct vxlanhdr_gbp *)vxh;
2306 vxh->vx_flags |= VXLAN_HF_GBP;
2307
2308 if (md->gbp & VXLAN_GBP_DONT_LEARN)
2309 gbp->dont_learn = 1;
2310
2311 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
2312 gbp->policy_applied = 1;
2313
2314 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
2315}
2316
2317static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
2318 __be16 protocol)
2319{
2320 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
2321
2322 gpe->np_applied = 1;
2323 gpe->next_protocol = tun_p_from_eth_p(protocol);
2324 if (!gpe->next_protocol)
2325 return -EPFNOSUPPORT;
2326 return 0;
2327}
2328
2329static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
2330 int iphdr_len, __be32 vni,
2331 struct vxlan_metadata *md, u32 vxflags,
2332 bool udp_sum)
2333{
2334 struct vxlanhdr *vxh;
2335 int min_headroom;
2336 int err;
2337 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
2338 __be16 inner_protocol = htons(ETH_P_TEB);
2339
2340 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
2341 skb->ip_summed == CHECKSUM_PARTIAL) {
2342 int csum_start = skb_checksum_start_offset(skb);
2343
2344 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
2345 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
2346 (skb->csum_offset == offsetof(struct udphdr, check) ||
2347 skb->csum_offset == offsetof(struct tcphdr, check)))
2348 type |= SKB_GSO_TUNNEL_REMCSUM;
2349 }
2350
2351 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
2352 + VXLAN_HLEN + iphdr_len;
2353
2354 /* Need space for new headers (invalidates iph ptr) */
2355 err = skb_cow_head(skb, min_headroom);
2356 if (unlikely(err))
2357 return err;
2358
2359 err = iptunnel_handle_offloads(skb, type);
2360 if (err)
2361 return err;
2362
2363 vxh = __skb_push(skb, sizeof(*vxh));
2364 vxh->vx_flags = VXLAN_HF_VNI;
2365 vxh->vx_vni = vxlan_vni_field(vni);
2366
2367 if (type & SKB_GSO_TUNNEL_REMCSUM) {
2368 unsigned int start;
2369
2370 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
2371 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
2372 vxh->vx_flags |= VXLAN_HF_RCO;
2373
2374 if (!skb_is_gso(skb)) {
2375 skb->ip_summed = CHECKSUM_NONE;
2376 skb->encapsulation = 0;
2377 }
2378 }
2379
2380 if (vxflags & VXLAN_F_GBP)
2381 vxlan_build_gbp_hdr(vxh, vxflags, md);
2382 if (vxflags & VXLAN_F_GPE) {
2383 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2384 if (err < 0)
2385 return err;
2386 inner_protocol = skb->protocol;
2387 }
2388
2389 skb_set_inner_protocol(skb, inner_protocol);
2390 return 0;
2391}
2392
2393static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2394 struct vxlan_sock *sock4,
2395 struct sk_buff *skb, int oif, u8 tos,
2396 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2397 struct dst_cache *dst_cache,
2398 const struct ip_tunnel_info *info)
2399{
2400 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2401 struct rtable *rt = NULL;
2402 struct flowi4 fl4;
2403
2404 if (!sock4)
2405 return ERR_PTR(-EIO);
2406
2407 if (tos && !info)
2408 use_cache = false;
2409 if (use_cache) {
2410 rt = dst_cache_get_ip4(dst_cache, saddr);
2411 if (rt)
2412 return rt;
2413 }
2414
2415 memset(&fl4, 0, sizeof(fl4));
2416 fl4.flowi4_oif = oif;
2417 fl4.flowi4_tos = RT_TOS(tos);
2418 fl4.flowi4_mark = skb->mark;
2419 fl4.flowi4_proto = IPPROTO_UDP;
2420 fl4.daddr = daddr;
2421 fl4.saddr = *saddr;
2422 fl4.fl4_dport = dport;
2423 fl4.fl4_sport = sport;
2424
2425 rt = ip_route_output_key(vxlan->net, &fl4);
2426 if (!IS_ERR(rt)) {
2427 if (rt->dst.dev == dev) {
2428 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2429 ip_rt_put(rt);
2430 return ERR_PTR(-ELOOP);
2431 }
2432
2433 *saddr = fl4.saddr;
2434 if (use_cache)
2435 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2436 } else {
2437 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2438 return ERR_PTR(-ENETUNREACH);
2439 }
2440 return rt;
2441}
2442
2443#if IS_ENABLED(CONFIG_IPV6)
2444static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2445 struct net_device *dev,
2446 struct vxlan_sock *sock6,
2447 struct sk_buff *skb, int oif, u8 tos,
2448 __be32 label,
2449 const struct in6_addr *daddr,
2450 struct in6_addr *saddr,
2451 __be16 dport, __be16 sport,
2452 struct dst_cache *dst_cache,
2453 const struct ip_tunnel_info *info)
2454{
2455 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2456 struct dst_entry *ndst;
2457 struct flowi6 fl6;
2458
2459 if (!sock6)
2460 return ERR_PTR(-EIO);
2461
2462 if (tos && !info)
2463 use_cache = false;
2464 if (use_cache) {
2465 ndst = dst_cache_get_ip6(dst_cache, saddr);
2466 if (ndst)
2467 return ndst;
2468 }
2469
2470 memset(&fl6, 0, sizeof(fl6));
2471 fl6.flowi6_oif = oif;
2472 fl6.daddr = *daddr;
2473 fl6.saddr = *saddr;
2474 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2475 fl6.flowi6_mark = skb->mark;
2476 fl6.flowi6_proto = IPPROTO_UDP;
2477 fl6.fl6_dport = dport;
2478 fl6.fl6_sport = sport;
2479
2480 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
2481 &fl6, NULL);
2482 if (IS_ERR(ndst)) {
2483 netdev_dbg(dev, "no route to %pI6\n", daddr);
2484 return ERR_PTR(-ENETUNREACH);
2485 }
2486
2487 if (unlikely(ndst->dev == dev)) {
2488 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2489 dst_release(ndst);
2490 return ERR_PTR(-ELOOP);
2491 }
2492
2493 *saddr = fl6.saddr;
2494 if (use_cache)
2495 dst_cache_set_ip6(dst_cache, ndst, saddr);
2496 return ndst;
2497}
2498#endif
2499
2500/* Bypass encapsulation if the destination is local */
2501static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2502 struct vxlan_dev *dst_vxlan, __be32 vni,
2503 bool snoop)
2504{
2505 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2506 union vxlan_addr loopback;
2507 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2508 struct net_device *dev;
2509 int len = skb->len;
2510
2511 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2512 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2513 skb->pkt_type = PACKET_HOST;
2514 skb->encapsulation = 0;
2515 skb->dev = dst_vxlan->dev;
2516 __skb_pull(skb, skb_network_offset(skb));
2517
2518 if (remote_ip->sa.sa_family == AF_INET) {
2519 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2520 loopback.sa.sa_family = AF_INET;
2521#if IS_ENABLED(CONFIG_IPV6)
2522 } else {
2523 loopback.sin6.sin6_addr = in6addr_loopback;
2524 loopback.sa.sa_family = AF_INET6;
2525#endif
2526 }
2527
2528 rcu_read_lock();
2529 dev = skb->dev;
2530 if (unlikely(!(dev->flags & IFF_UP))) {
2531 kfree_skb(skb);
2532 goto drop;
2533 }
2534
2535 if ((dst_vxlan->cfg.flags & VXLAN_F_LEARN) && snoop)
2536 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2537
2538 u64_stats_update_begin(&tx_stats->syncp);
2539 tx_stats->tx_packets++;
2540 tx_stats->tx_bytes += len;
2541 u64_stats_update_end(&tx_stats->syncp);
2542
2543 if (netif_rx(skb) == NET_RX_SUCCESS) {
2544 u64_stats_update_begin(&rx_stats->syncp);
2545 rx_stats->rx_packets++;
2546 rx_stats->rx_bytes += len;
2547 u64_stats_update_end(&rx_stats->syncp);
2548 } else {
2549drop:
2550 dev->stats.rx_dropped++;
2551 }
2552 rcu_read_unlock();
2553}
2554
2555static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2556 struct vxlan_dev *vxlan,
2557 union vxlan_addr *daddr,
2558 __be16 dst_port, int dst_ifindex, __be32 vni,
2559 struct dst_entry *dst,
2560 u32 rt_flags)
2561{
2562#if IS_ENABLED(CONFIG_IPV6)
2563 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2564 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2565 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2566 */
2567 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2568#endif
2569 /* Bypass encapsulation if the destination is local */
2570 if (rt_flags & RTCF_LOCAL &&
2571 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2572 struct vxlan_dev *dst_vxlan;
2573
2574 dst_release(dst);
2575 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2576 daddr->sa.sa_family, dst_port,
2577 vxlan->cfg.flags);
2578 if (!dst_vxlan) {
2579 dev->stats.tx_errors++;
2580 kfree_skb(skb);
2581
2582 return -ENOENT;
2583 }
2584 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni, true);
2585 return 1;
2586 }
2587
2588 return 0;
2589}
2590
2591static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2592 __be32 default_vni, struct vxlan_rdst *rdst,
2593 bool did_rsc)
2594{
2595 struct dst_cache *dst_cache;
2596 struct ip_tunnel_info *info;
2597 struct vxlan_dev *vxlan = netdev_priv(dev);
2598 const struct iphdr *old_iph = ip_hdr(skb);
2599 union vxlan_addr *dst;
2600 union vxlan_addr remote_ip, local_ip;
2601 struct vxlan_metadata _md;
2602 struct vxlan_metadata *md = &_md;
2603 __be16 src_port = 0, dst_port;
2604 struct dst_entry *ndst = NULL;
2605 __be32 vni, label;
2606 __u8 tos, ttl;
2607 int ifindex;
2608 int err;
2609 u32 flags = vxlan->cfg.flags;
2610 bool udp_sum = false;
2611 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2612
2613 info = skb_tunnel_info(skb);
2614
2615 if (rdst) {
2616 dst = &rdst->remote_ip;
2617 if (vxlan_addr_any(dst)) {
2618 if (did_rsc) {
2619 /* short-circuited back to local bridge */
2620 vxlan_encap_bypass(skb, vxlan, vxlan,
2621 default_vni, true);
2622 return;
2623 }
2624 goto drop;
2625 }
2626
2627 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2628 vni = (rdst->remote_vni) ? : default_vni;
2629 ifindex = rdst->remote_ifindex;
2630 local_ip = vxlan->cfg.saddr;
2631 dst_cache = &rdst->dst_cache;
2632 md->gbp = skb->mark;
2633 if (flags & VXLAN_F_TTL_INHERIT) {
2634 ttl = ip_tunnel_get_ttl(old_iph, skb);
2635 } else {
2636 ttl = vxlan->cfg.ttl;
2637 if (!ttl && vxlan_addr_multicast(dst))
2638 ttl = 1;
2639 }
2640
2641 tos = vxlan->cfg.tos;
2642 if (tos == 1)
2643 tos = ip_tunnel_get_dsfield(old_iph, skb);
2644
2645 if (dst->sa.sa_family == AF_INET)
2646 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2647 else
2648 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2649 label = vxlan->cfg.label;
2650 } else {
2651 if (!info) {
2652 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2653 dev->name);
2654 goto drop;
2655 }
2656 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2657 if (remote_ip.sa.sa_family == AF_INET) {
2658 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2659 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2660 } else {
2661 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2662 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2663 }
2664 dst = &remote_ip;
2665 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2666 vni = tunnel_id_to_key32(info->key.tun_id);
2667 ifindex = 0;
2668 dst_cache = &info->dst_cache;
2669 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2670 if (info->options_len < sizeof(*md))
2671 goto drop;
2672 md = ip_tunnel_info_opts(info);
2673 }
2674 ttl = info->key.ttl;
2675 tos = info->key.tos;
2676 label = info->key.label;
2677 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2678 }
2679 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2680 vxlan->cfg.port_max, true);
2681
2682 rcu_read_lock();
2683 if (dst->sa.sa_family == AF_INET) {
2684 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2685 struct rtable *rt;
2686 __be16 df = 0;
2687
2688 if (!ifindex)
2689 ifindex = sock4->sock->sk->sk_bound_dev_if;
2690
2691 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2692 dst->sin.sin_addr.s_addr,
2693 &local_ip.sin.sin_addr.s_addr,
2694 dst_port, src_port,
2695 dst_cache, info);
2696 if (IS_ERR(rt)) {
2697 err = PTR_ERR(rt);
2698 goto tx_error;
2699 }
2700
2701 if (!info) {
2702 /* Bypass encapsulation if the destination is local */
2703 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2704 dst_port, ifindex, vni,
2705 &rt->dst, rt->rt_flags);
2706 if (err)
2707 goto out_unlock;
2708
2709 if (vxlan->cfg.df == VXLAN_DF_SET) {
2710 df = htons(IP_DF);
2711 } else if (vxlan->cfg.df == VXLAN_DF_INHERIT) {
2712 struct ethhdr *eth = eth_hdr(skb);
2713
2714 if (ntohs(eth->h_proto) == ETH_P_IPV6 ||
2715 (ntohs(eth->h_proto) == ETH_P_IP &&
2716 old_iph->frag_off & htons(IP_DF)))
2717 df = htons(IP_DF);
2718 }
2719 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2720 df = htons(IP_DF);
2721 }
2722
2723 ndst = &rt->dst;
2724 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM,
2725 netif_is_any_bridge_port(dev));
2726 if (err < 0) {
2727 goto tx_error;
2728 } else if (err) {
2729 if (info) {
2730 struct ip_tunnel_info *unclone;
2731 struct in_addr src, dst;
2732
2733 unclone = skb_tunnel_info_unclone(skb);
2734 if (unlikely(!unclone))
2735 goto tx_error;
2736
2737 src = remote_ip.sin.sin_addr;
2738 dst = local_ip.sin.sin_addr;
2739 unclone->key.u.ipv4.src = src.s_addr;
2740 unclone->key.u.ipv4.dst = dst.s_addr;
2741 }
2742 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2743 dst_release(ndst);
2744 goto out_unlock;
2745 }
2746
2747 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2748 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2749 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2750 vni, md, flags, udp_sum);
2751 if (err < 0)
2752 goto tx_error;
2753
2754 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2755 dst->sin.sin_addr.s_addr, tos, ttl, df,
2756 src_port, dst_port, xnet, !udp_sum);
2757#if IS_ENABLED(CONFIG_IPV6)
2758 } else {
2759 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2760
2761 if (!ifindex)
2762 ifindex = sock6->sock->sk->sk_bound_dev_if;
2763
2764 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2765 label, &dst->sin6.sin6_addr,
2766 &local_ip.sin6.sin6_addr,
2767 dst_port, src_port,
2768 dst_cache, info);
2769 if (IS_ERR(ndst)) {
2770 err = PTR_ERR(ndst);
2771 ndst = NULL;
2772 goto tx_error;
2773 }
2774
2775 if (!info) {
2776 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2777
2778 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2779 dst_port, ifindex, vni,
2780 ndst, rt6i_flags);
2781 if (err)
2782 goto out_unlock;
2783 }
2784
2785 err = skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM,
2786 netif_is_any_bridge_port(dev));
2787 if (err < 0) {
2788 goto tx_error;
2789 } else if (err) {
2790 if (info) {
2791 struct ip_tunnel_info *unclone;
2792 struct in6_addr src, dst;
2793
2794 unclone = skb_tunnel_info_unclone(skb);
2795 if (unlikely(!unclone))
2796 goto tx_error;
2797
2798 src = remote_ip.sin6.sin6_addr;
2799 dst = local_ip.sin6.sin6_addr;
2800 unclone->key.u.ipv6.src = src;
2801 unclone->key.u.ipv6.dst = dst;
2802 }
2803
2804 vxlan_encap_bypass(skb, vxlan, vxlan, vni, false);
2805 dst_release(ndst);
2806 goto out_unlock;
2807 }
2808
2809 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2810 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2811 skb_scrub_packet(skb, xnet);
2812 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2813 vni, md, flags, udp_sum);
2814 if (err < 0)
2815 goto tx_error;
2816
2817 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2818 &local_ip.sin6.sin6_addr,
2819 &dst->sin6.sin6_addr, tos, ttl,
2820 label, src_port, dst_port, !udp_sum);
2821#endif
2822 }
2823out_unlock:
2824 rcu_read_unlock();
2825 return;
2826
2827drop:
2828 dev->stats.tx_dropped++;
2829 dev_kfree_skb(skb);
2830 return;
2831
2832tx_error:
2833 rcu_read_unlock();
2834 if (err == -ELOOP)
2835 dev->stats.collisions++;
2836 else if (err == -ENETUNREACH)
2837 dev->stats.tx_carrier_errors++;
2838 dst_release(ndst);
2839 dev->stats.tx_errors++;
2840 kfree_skb(skb);
2841}
2842
2843static void vxlan_xmit_nh(struct sk_buff *skb, struct net_device *dev,
2844 struct vxlan_fdb *f, __be32 vni, bool did_rsc)
2845{
2846 struct vxlan_rdst nh_rdst;
2847 struct nexthop *nh;
2848 bool do_xmit;
2849 u32 hash;
2850
2851 memset(&nh_rdst, 0, sizeof(struct vxlan_rdst));
2852 hash = skb_get_hash(skb);
2853
2854 rcu_read_lock();
2855 nh = rcu_dereference(f->nh);
2856 if (!nh) {
2857 rcu_read_unlock();
2858 goto drop;
2859 }
2860 do_xmit = vxlan_fdb_nh_path_select(nh, hash, &nh_rdst);
2861 rcu_read_unlock();
2862
2863 if (likely(do_xmit))
2864 vxlan_xmit_one(skb, dev, vni, &nh_rdst, did_rsc);
2865 else
2866 goto drop;
2867
2868 return;
2869
2870drop:
2871 dev->stats.tx_dropped++;
2872 dev_kfree_skb(skb);
2873}
2874
2875/* Transmit local packets over Vxlan
2876 *
2877 * Outer IP header inherits ECN and DF from inner header.
2878 * Outer UDP destination is the VXLAN assigned port.
2879 * source port is based on hash of flow
2880 */
2881static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2882{
2883 struct vxlan_dev *vxlan = netdev_priv(dev);
2884 struct vxlan_rdst *rdst, *fdst = NULL;
2885 const struct ip_tunnel_info *info;
2886 bool did_rsc = false;
2887 struct vxlan_fdb *f;
2888 struct ethhdr *eth;
2889 __be32 vni = 0;
2890
2891 info = skb_tunnel_info(skb);
2892
2893 skb_reset_mac_header(skb);
2894
2895 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2896 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2897 info->mode & IP_TUNNEL_INFO_TX) {
2898 vni = tunnel_id_to_key32(info->key.tun_id);
2899 } else {
2900 if (info && info->mode & IP_TUNNEL_INFO_TX)
2901 vxlan_xmit_one(skb, dev, vni, NULL, false);
2902 else
2903 kfree_skb(skb);
2904 return NETDEV_TX_OK;
2905 }
2906 }
2907
2908 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2909 eth = eth_hdr(skb);
2910 if (ntohs(eth->h_proto) == ETH_P_ARP)
2911 return arp_reduce(dev, skb, vni);
2912#if IS_ENABLED(CONFIG_IPV6)
2913 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2914 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2915 sizeof(struct nd_msg)) &&
2916 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2917 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2918
2919 if (m->icmph.icmp6_code == 0 &&
2920 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2921 return neigh_reduce(dev, skb, vni);
2922 }
2923#endif
2924 }
2925
2926 eth = eth_hdr(skb);
2927 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2928 did_rsc = false;
2929
2930 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2931 (ntohs(eth->h_proto) == ETH_P_IP ||
2932 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2933 did_rsc = route_shortcircuit(dev, skb);
2934 if (did_rsc)
2935 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2936 }
2937
2938 if (f == NULL) {
2939 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2940 if (f == NULL) {
2941 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2942 !is_multicast_ether_addr(eth->h_dest))
2943 vxlan_fdb_miss(vxlan, eth->h_dest);
2944
2945 dev->stats.tx_dropped++;
2946 kfree_skb(skb);
2947 return NETDEV_TX_OK;
2948 }
2949 }
2950
2951 if (rcu_access_pointer(f->nh)) {
2952 vxlan_xmit_nh(skb, dev, f,
2953 (vni ? : vxlan->default_dst.remote_vni), did_rsc);
2954 } else {
2955 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2956 struct sk_buff *skb1;
2957
2958 if (!fdst) {
2959 fdst = rdst;
2960 continue;
2961 }
2962 skb1 = skb_clone(skb, GFP_ATOMIC);
2963 if (skb1)
2964 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2965 }
2966 if (fdst)
2967 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2968 else
2969 kfree_skb(skb);
2970 }
2971
2972 return NETDEV_TX_OK;
2973}
2974
2975/* Walk the forwarding table and purge stale entries */
2976static void vxlan_cleanup(struct timer_list *t)
2977{
2978 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2979 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2980 unsigned int h;
2981
2982 if (!netif_running(vxlan->dev))
2983 return;
2984
2985 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2986 struct hlist_node *p, *n;
2987
2988 spin_lock(&vxlan->hash_lock[h]);
2989 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2990 struct vxlan_fdb *f
2991 = container_of(p, struct vxlan_fdb, hlist);
2992 unsigned long timeout;
2993
2994 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2995 continue;
2996
2997 if (f->flags & NTF_EXT_LEARNED)
2998 continue;
2999
3000 timeout = f->used + vxlan->cfg.age_interval * HZ;
3001 if (time_before_eq(timeout, jiffies)) {
3002 netdev_dbg(vxlan->dev,
3003 "garbage collect %pM\n",
3004 f->eth_addr);
3005 f->state = NUD_STALE;
3006 vxlan_fdb_destroy(vxlan, f, true, true);
3007 } else if (time_before(timeout, next_timer))
3008 next_timer = timeout;
3009 }
3010 spin_unlock(&vxlan->hash_lock[h]);
3011 }
3012
3013 mod_timer(&vxlan->age_timer, next_timer);
3014}
3015
3016static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
3017{
3018 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3019
3020 spin_lock(&vn->sock_lock);
3021 hlist_del_init_rcu(&vxlan->hlist4.hlist);
3022#if IS_ENABLED(CONFIG_IPV6)
3023 hlist_del_init_rcu(&vxlan->hlist6.hlist);
3024#endif
3025 spin_unlock(&vn->sock_lock);
3026}
3027
3028static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
3029 struct vxlan_dev_node *node)
3030{
3031 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3032 __be32 vni = vxlan->default_dst.remote_vni;
3033
3034 node->vxlan = vxlan;
3035 spin_lock(&vn->sock_lock);
3036 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
3037 spin_unlock(&vn->sock_lock);
3038}
3039
3040/* Setup stats when device is created */
3041static int vxlan_init(struct net_device *dev)
3042{
3043 struct vxlan_dev *vxlan = netdev_priv(dev);
3044 int err;
3045
3046 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
3047 if (!dev->tstats)
3048 return -ENOMEM;
3049
3050 err = gro_cells_init(&vxlan->gro_cells, dev);
3051 if (err) {
3052 free_percpu(dev->tstats);
3053 return err;
3054 }
3055
3056 return 0;
3057}
3058
3059static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
3060{
3061 struct vxlan_fdb *f;
3062 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, vni);
3063
3064 spin_lock_bh(&vxlan->hash_lock[hash_index]);
3065 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
3066 if (f)
3067 vxlan_fdb_destroy(vxlan, f, true, true);
3068 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
3069}
3070
3071static void vxlan_uninit(struct net_device *dev)
3072{
3073 struct vxlan_dev *vxlan = netdev_priv(dev);
3074
3075 gro_cells_destroy(&vxlan->gro_cells);
3076
3077 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
3078
3079 free_percpu(dev->tstats);
3080}
3081
3082/* Start ageing timer and join group when device is brought up */
3083static int vxlan_open(struct net_device *dev)
3084{
3085 struct vxlan_dev *vxlan = netdev_priv(dev);
3086 int ret;
3087
3088 ret = vxlan_sock_add(vxlan);
3089 if (ret < 0)
3090 return ret;
3091
3092 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
3093 ret = vxlan_igmp_join(vxlan);
3094 if (ret == -EADDRINUSE)
3095 ret = 0;
3096 if (ret) {
3097 vxlan_sock_release(vxlan);
3098 return ret;
3099 }
3100 }
3101
3102 if (vxlan->cfg.age_interval)
3103 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
3104
3105 return ret;
3106}
3107
3108/* Purge the forwarding table */
3109static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
3110{
3111 unsigned int h;
3112
3113 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3114 struct hlist_node *p, *n;
3115
3116 spin_lock_bh(&vxlan->hash_lock[h]);
3117 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
3118 struct vxlan_fdb *f
3119 = container_of(p, struct vxlan_fdb, hlist);
3120 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
3121 continue;
3122 /* the all_zeros_mac entry is deleted at vxlan_uninit */
3123 if (is_zero_ether_addr(f->eth_addr) &&
3124 f->vni == vxlan->cfg.vni)
3125 continue;
3126 vxlan_fdb_destroy(vxlan, f, true, true);
3127 }
3128 spin_unlock_bh(&vxlan->hash_lock[h]);
3129 }
3130}
3131
3132/* Cleanup timer and forwarding table on shutdown */
3133static int vxlan_stop(struct net_device *dev)
3134{
3135 struct vxlan_dev *vxlan = netdev_priv(dev);
3136 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3137 int ret = 0;
3138
3139 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
3140 !vxlan_group_used(vn, vxlan))
3141 ret = vxlan_igmp_leave(vxlan);
3142
3143 del_timer_sync(&vxlan->age_timer);
3144
3145 vxlan_flush(vxlan, false);
3146 vxlan_sock_release(vxlan);
3147
3148 return ret;
3149}
3150
3151/* Stub, nothing needs to be done. */
3152static void vxlan_set_multicast_list(struct net_device *dev)
3153{
3154}
3155
3156static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
3157{
3158 struct vxlan_dev *vxlan = netdev_priv(dev);
3159 struct vxlan_rdst *dst = &vxlan->default_dst;
3160 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3161 dst->remote_ifindex);
3162 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
3163
3164 /* This check is different than dev->max_mtu, because it looks at
3165 * the lowerdev->mtu, rather than the static dev->max_mtu
3166 */
3167 if (lowerdev) {
3168 int max_mtu = lowerdev->mtu -
3169 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
3170 if (new_mtu > max_mtu)
3171 return -EINVAL;
3172 }
3173
3174 dev->mtu = new_mtu;
3175 return 0;
3176}
3177
3178static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
3179{
3180 struct vxlan_dev *vxlan = netdev_priv(dev);
3181 struct ip_tunnel_info *info = skb_tunnel_info(skb);
3182 __be16 sport, dport;
3183
3184 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
3185 vxlan->cfg.port_max, true);
3186 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
3187
3188 if (ip_tunnel_info_af(info) == AF_INET) {
3189 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
3190 struct rtable *rt;
3191
3192 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
3193 info->key.u.ipv4.dst,
3194 &info->key.u.ipv4.src, dport, sport,
3195 &info->dst_cache, info);
3196 if (IS_ERR(rt))
3197 return PTR_ERR(rt);
3198 ip_rt_put(rt);
3199 } else {
3200#if IS_ENABLED(CONFIG_IPV6)
3201 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
3202 struct dst_entry *ndst;
3203
3204 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
3205 info->key.label, &info->key.u.ipv6.dst,
3206 &info->key.u.ipv6.src, dport, sport,
3207 &info->dst_cache, info);
3208 if (IS_ERR(ndst))
3209 return PTR_ERR(ndst);
3210 dst_release(ndst);
3211#else /* !CONFIG_IPV6 */
3212 return -EPFNOSUPPORT;
3213#endif
3214 }
3215 info->key.tp_src = sport;
3216 info->key.tp_dst = dport;
3217 return 0;
3218}
3219
3220static const struct net_device_ops vxlan_netdev_ether_ops = {
3221 .ndo_init = vxlan_init,
3222 .ndo_uninit = vxlan_uninit,
3223 .ndo_open = vxlan_open,
3224 .ndo_stop = vxlan_stop,
3225 .ndo_start_xmit = vxlan_xmit,
3226 .ndo_get_stats64 = dev_get_tstats64,
3227 .ndo_set_rx_mode = vxlan_set_multicast_list,
3228 .ndo_change_mtu = vxlan_change_mtu,
3229 .ndo_validate_addr = eth_validate_addr,
3230 .ndo_set_mac_address = eth_mac_addr,
3231 .ndo_fdb_add = vxlan_fdb_add,
3232 .ndo_fdb_del = vxlan_fdb_delete,
3233 .ndo_fdb_dump = vxlan_fdb_dump,
3234 .ndo_fdb_get = vxlan_fdb_get,
3235 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3236 .ndo_change_proto_down = dev_change_proto_down_generic,
3237};
3238
3239static const struct net_device_ops vxlan_netdev_raw_ops = {
3240 .ndo_init = vxlan_init,
3241 .ndo_uninit = vxlan_uninit,
3242 .ndo_open = vxlan_open,
3243 .ndo_stop = vxlan_stop,
3244 .ndo_start_xmit = vxlan_xmit,
3245 .ndo_get_stats64 = dev_get_tstats64,
3246 .ndo_change_mtu = vxlan_change_mtu,
3247 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
3248};
3249
3250/* Info for udev, that this is a virtual tunnel endpoint */
3251static struct device_type vxlan_type = {
3252 .name = "vxlan",
3253};
3254
3255/* Calls the ndo_udp_tunnel_add of the caller in order to
3256 * supply the listening VXLAN udp ports. Callers are expected
3257 * to implement the ndo_udp_tunnel_add.
3258 */
3259static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
3260{
3261 struct vxlan_sock *vs;
3262 struct net *net = dev_net(dev);
3263 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3264 unsigned int i;
3265
3266 spin_lock(&vn->sock_lock);
3267 for (i = 0; i < PORT_HASH_SIZE; ++i) {
3268 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
3269 unsigned short type;
3270
3271 if (vs->flags & VXLAN_F_GPE)
3272 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
3273 else
3274 type = UDP_TUNNEL_TYPE_VXLAN;
3275
3276 if (push)
3277 udp_tunnel_push_rx_port(dev, vs->sock, type);
3278 else
3279 udp_tunnel_drop_rx_port(dev, vs->sock, type);
3280 }
3281 }
3282 spin_unlock(&vn->sock_lock);
3283}
3284
3285/* Initialize the device structure. */
3286static void vxlan_setup(struct net_device *dev)
3287{
3288 struct vxlan_dev *vxlan = netdev_priv(dev);
3289 unsigned int h;
3290
3291 eth_hw_addr_random(dev);
3292 ether_setup(dev);
3293
3294 dev->needs_free_netdev = true;
3295 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
3296
3297 dev->features |= NETIF_F_LLTX;
3298 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3299 dev->features |= NETIF_F_RXCSUM;
3300 dev->features |= NETIF_F_GSO_SOFTWARE;
3301
3302 dev->vlan_features = dev->features;
3303 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_FRAGLIST;
3304 dev->hw_features |= NETIF_F_RXCSUM;
3305 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
3306 netif_keep_dst(dev);
3307 dev->priv_flags |= IFF_NO_QUEUE;
3308
3309 /* MTU range: 68 - 65535 */
3310 dev->min_mtu = ETH_MIN_MTU;
3311 dev->max_mtu = ETH_MAX_MTU;
3312
3313 INIT_LIST_HEAD(&vxlan->next);
3314
3315 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
3316
3317 vxlan->dev = dev;
3318
3319 for (h = 0; h < FDB_HASH_SIZE; ++h) {
3320 spin_lock_init(&vxlan->hash_lock[h]);
3321 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
3322 }
3323}
3324
3325static void vxlan_ether_setup(struct net_device *dev)
3326{
3327 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
3328 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
3329 dev->netdev_ops = &vxlan_netdev_ether_ops;
3330}
3331
3332static void vxlan_raw_setup(struct net_device *dev)
3333{
3334 dev->header_ops = NULL;
3335 dev->type = ARPHRD_NONE;
3336 dev->hard_header_len = 0;
3337 dev->addr_len = 0;
3338 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
3339 dev->netdev_ops = &vxlan_netdev_raw_ops;
3340}
3341
3342static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
3343 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
3344 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
3345 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
3346 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
3347 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
3348 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
3349 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
3350 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
3351 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
3352 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
3353 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
3354 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
3355 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
3356 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
3357 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
3358 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
3359 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
3360 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
3361 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
3362 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
3363 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
3364 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
3365 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
3366 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
3367 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
3368 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
3369 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
3370 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
3371 [IFLA_VXLAN_DF] = { .type = NLA_U8 },
3372};
3373
3374static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
3375 struct netlink_ext_ack *extack)
3376{
3377 if (tb[IFLA_ADDRESS]) {
3378 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
3379 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3380 "Provided link layer address is not Ethernet");
3381 return -EINVAL;
3382 }
3383
3384 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
3385 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
3386 "Provided Ethernet address is not unicast");
3387 return -EADDRNOTAVAIL;
3388 }
3389 }
3390
3391 if (tb[IFLA_MTU]) {
3392 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
3393
3394 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
3395 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
3396 "MTU must be between 68 and 65535");
3397 return -EINVAL;
3398 }
3399 }
3400
3401 if (!data) {
3402 NL_SET_ERR_MSG(extack,
3403 "Required attributes not provided to perform the operation");
3404 return -EINVAL;
3405 }
3406
3407 if (data[IFLA_VXLAN_ID]) {
3408 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
3409
3410 if (id >= VXLAN_N_VID) {
3411 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_ID],
3412 "VXLAN ID must be lower than 16777216");
3413 return -ERANGE;
3414 }
3415 }
3416
3417 if (data[IFLA_VXLAN_PORT_RANGE]) {
3418 const struct ifla_vxlan_port_range *p
3419 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3420
3421 if (ntohs(p->high) < ntohs(p->low)) {
3422 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_PORT_RANGE],
3423 "Invalid source port range");
3424 return -EINVAL;
3425 }
3426 }
3427
3428 if (data[IFLA_VXLAN_DF]) {
3429 enum ifla_vxlan_df df = nla_get_u8(data[IFLA_VXLAN_DF]);
3430
3431 if (df < 0 || df > VXLAN_DF_MAX) {
3432 NL_SET_ERR_MSG_ATTR(extack, data[IFLA_VXLAN_DF],
3433 "Invalid DF attribute");
3434 return -EINVAL;
3435 }
3436 }
3437
3438 return 0;
3439}
3440
3441static void vxlan_get_drvinfo(struct net_device *netdev,
3442 struct ethtool_drvinfo *drvinfo)
3443{
3444 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
3445 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
3446}
3447
3448static int vxlan_get_link_ksettings(struct net_device *dev,
3449 struct ethtool_link_ksettings *cmd)
3450{
3451 struct vxlan_dev *vxlan = netdev_priv(dev);
3452 struct vxlan_rdst *dst = &vxlan->default_dst;
3453 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
3454 dst->remote_ifindex);
3455
3456 if (!lowerdev) {
3457 cmd->base.duplex = DUPLEX_UNKNOWN;
3458 cmd->base.port = PORT_OTHER;
3459 cmd->base.speed = SPEED_UNKNOWN;
3460
3461 return 0;
3462 }
3463
3464 return __ethtool_get_link_ksettings(lowerdev, cmd);
3465}
3466
3467static const struct ethtool_ops vxlan_ethtool_ops = {
3468 .get_drvinfo = vxlan_get_drvinfo,
3469 .get_link = ethtool_op_get_link,
3470 .get_link_ksettings = vxlan_get_link_ksettings,
3471};
3472
3473static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
3474 __be16 port, u32 flags, int ifindex)
3475{
3476 struct socket *sock;
3477 struct udp_port_cfg udp_conf;
3478 int err;
3479
3480 memset(&udp_conf, 0, sizeof(udp_conf));
3481
3482 if (ipv6) {
3483 udp_conf.family = AF_INET6;
3484 udp_conf.use_udp6_rx_checksums =
3485 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
3486 udp_conf.ipv6_v6only = 1;
3487 } else {
3488 udp_conf.family = AF_INET;
3489 }
3490
3491 udp_conf.local_udp_port = port;
3492 udp_conf.bind_ifindex = ifindex;
3493
3494 /* Open UDP socket */
3495 err = udp_sock_create(net, &udp_conf, &sock);
3496 if (err < 0)
3497 return ERR_PTR(err);
3498
3499 udp_allow_gso(sock->sk);
3500 return sock;
3501}
3502
3503/* Create new listen socket if needed */
3504static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
3505 __be16 port, u32 flags,
3506 int ifindex)
3507{
3508 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3509 struct vxlan_sock *vs;
3510 struct socket *sock;
3511 unsigned int h;
3512 struct udp_tunnel_sock_cfg tunnel_cfg;
3513
3514 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3515 if (!vs)
3516 return ERR_PTR(-ENOMEM);
3517
3518 for (h = 0; h < VNI_HASH_SIZE; ++h)
3519 INIT_HLIST_HEAD(&vs->vni_list[h]);
3520
3521 sock = vxlan_create_sock(net, ipv6, port, flags, ifindex);
3522 if (IS_ERR(sock)) {
3523 kfree(vs);
3524 return ERR_CAST(sock);
3525 }
3526
3527 vs->sock = sock;
3528 refcount_set(&vs->refcnt, 1);
3529 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3530
3531 spin_lock(&vn->sock_lock);
3532 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3533 udp_tunnel_notify_add_rx_port(sock,
3534 (vs->flags & VXLAN_F_GPE) ?
3535 UDP_TUNNEL_TYPE_VXLAN_GPE :
3536 UDP_TUNNEL_TYPE_VXLAN);
3537 spin_unlock(&vn->sock_lock);
3538
3539 /* Mark socket as an encapsulation socket. */
3540 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3541 tunnel_cfg.sk_user_data = vs;
3542 tunnel_cfg.encap_type = 1;
3543 tunnel_cfg.encap_rcv = vxlan_rcv;
3544 tunnel_cfg.encap_err_lookup = vxlan_err_lookup;
3545 tunnel_cfg.encap_destroy = NULL;
3546 tunnel_cfg.gro_receive = vxlan_gro_receive;
3547 tunnel_cfg.gro_complete = vxlan_gro_complete;
3548
3549 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3550
3551 return vs;
3552}
3553
3554static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3555{
3556 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3557 struct vxlan_sock *vs = NULL;
3558 struct vxlan_dev_node *node;
3559 int l3mdev_index = 0;
3560
3561 if (vxlan->cfg.remote_ifindex)
3562 l3mdev_index = l3mdev_master_upper_ifindex_by_index(
3563 vxlan->net, vxlan->cfg.remote_ifindex);
3564
3565 if (!vxlan->cfg.no_share) {
3566 spin_lock(&vn->sock_lock);
3567 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3568 vxlan->cfg.dst_port, vxlan->cfg.flags,
3569 l3mdev_index);
3570 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3571 spin_unlock(&vn->sock_lock);
3572 return -EBUSY;
3573 }
3574 spin_unlock(&vn->sock_lock);
3575 }
3576 if (!vs)
3577 vs = vxlan_socket_create(vxlan->net, ipv6,
3578 vxlan->cfg.dst_port, vxlan->cfg.flags,
3579 l3mdev_index);
3580 if (IS_ERR(vs))
3581 return PTR_ERR(vs);
3582#if IS_ENABLED(CONFIG_IPV6)
3583 if (ipv6) {
3584 rcu_assign_pointer(vxlan->vn6_sock, vs);
3585 node = &vxlan->hlist6;
3586 } else
3587#endif
3588 {
3589 rcu_assign_pointer(vxlan->vn4_sock, vs);
3590 node = &vxlan->hlist4;
3591 }
3592 vxlan_vs_add_dev(vs, vxlan, node);
3593 return 0;
3594}
3595
3596static int vxlan_sock_add(struct vxlan_dev *vxlan)
3597{
3598 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3599 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3600 bool ipv4 = !ipv6 || metadata;
3601 int ret = 0;
3602
3603 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3604#if IS_ENABLED(CONFIG_IPV6)
3605 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3606 if (ipv6) {
3607 ret = __vxlan_sock_add(vxlan, true);
3608 if (ret < 0 && ret != -EAFNOSUPPORT)
3609 ipv4 = false;
3610 }
3611#endif
3612 if (ipv4)
3613 ret = __vxlan_sock_add(vxlan, false);
3614 if (ret < 0)
3615 vxlan_sock_release(vxlan);
3616 return ret;
3617}
3618
3619static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3620 struct net_device **lower,
3621 struct vxlan_dev *old,
3622 struct netlink_ext_ack *extack)
3623{
3624 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3625 struct vxlan_dev *tmp;
3626 bool use_ipv6 = false;
3627
3628 if (conf->flags & VXLAN_F_GPE) {
3629 /* For now, allow GPE only together with
3630 * COLLECT_METADATA. This can be relaxed later; in such
3631 * case, the other side of the PtP link will have to be
3632 * provided.
3633 */
3634 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3635 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3636 NL_SET_ERR_MSG(extack,
3637 "VXLAN GPE does not support this combination of attributes");
3638 return -EINVAL;
3639 }
3640 }
3641
3642 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3643 /* Unless IPv6 is explicitly requested, assume IPv4 */
3644 conf->remote_ip.sa.sa_family = AF_INET;
3645 conf->saddr.sa.sa_family = AF_INET;
3646 } else if (!conf->remote_ip.sa.sa_family) {
3647 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3648 } else if (!conf->saddr.sa.sa_family) {
3649 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3650 }
3651
3652 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3653 NL_SET_ERR_MSG(extack,
3654 "Local and remote address must be from the same family");
3655 return -EINVAL;
3656 }
3657
3658 if (vxlan_addr_multicast(&conf->saddr)) {
3659 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3660 return -EINVAL;
3661 }
3662
3663 if (conf->saddr.sa.sa_family == AF_INET6) {
3664 if (!IS_ENABLED(CONFIG_IPV6)) {
3665 NL_SET_ERR_MSG(extack,
3666 "IPv6 support not enabled in the kernel");
3667 return -EPFNOSUPPORT;
3668 }
3669 use_ipv6 = true;
3670 conf->flags |= VXLAN_F_IPV6;
3671
3672 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3673 int local_type =
3674 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3675 int remote_type =
3676 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3677
3678 if (local_type & IPV6_ADDR_LINKLOCAL) {
3679 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3680 (remote_type != IPV6_ADDR_ANY)) {
3681 NL_SET_ERR_MSG(extack,
3682 "Invalid combination of local and remote address scopes");
3683 return -EINVAL;
3684 }
3685
3686 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3687 } else {
3688 if (remote_type ==
3689 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3690 NL_SET_ERR_MSG(extack,
3691 "Invalid combination of local and remote address scopes");
3692 return -EINVAL;
3693 }
3694
3695 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3696 }
3697 }
3698 }
3699
3700 if (conf->label && !use_ipv6) {
3701 NL_SET_ERR_MSG(extack,
3702 "Label attribute only applies to IPv6 VXLAN devices");
3703 return -EINVAL;
3704 }
3705
3706 if (conf->remote_ifindex) {
3707 struct net_device *lowerdev;
3708
3709 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3710 if (!lowerdev) {
3711 NL_SET_ERR_MSG(extack,
3712 "Invalid local interface, device not found");
3713 return -ENODEV;
3714 }
3715
3716#if IS_ENABLED(CONFIG_IPV6)
3717 if (use_ipv6) {
3718 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3719
3720 if (idev && idev->cnf.disable_ipv6) {
3721 NL_SET_ERR_MSG(extack,
3722 "IPv6 support disabled by administrator");
3723 return -EPERM;
3724 }
3725 }
3726#endif
3727
3728 *lower = lowerdev;
3729 } else {
3730 if (vxlan_addr_multicast(&conf->remote_ip)) {
3731 NL_SET_ERR_MSG(extack,
3732 "Local interface required for multicast remote destination");
3733
3734 return -EINVAL;
3735 }
3736
3737#if IS_ENABLED(CONFIG_IPV6)
3738 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3739 NL_SET_ERR_MSG(extack,
3740 "Local interface required for link-local local/remote addresses");
3741 return -EINVAL;
3742 }
3743#endif
3744
3745 *lower = NULL;
3746 }
3747
3748 if (!conf->dst_port) {
3749 if (conf->flags & VXLAN_F_GPE)
3750 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3751 else
3752 conf->dst_port = htons(vxlan_port);
3753 }
3754
3755 if (!conf->age_interval)
3756 conf->age_interval = FDB_AGE_DEFAULT;
3757
3758 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3759 if (tmp == old)
3760 continue;
3761
3762 if (tmp->cfg.vni != conf->vni)
3763 continue;
3764 if (tmp->cfg.dst_port != conf->dst_port)
3765 continue;
3766 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3767 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3768 continue;
3769
3770 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3771 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3772 continue;
3773
3774 NL_SET_ERR_MSG(extack,
3775 "A VXLAN device with the specified VNI already exists");
3776 return -EEXIST;
3777 }
3778
3779 return 0;
3780}
3781
3782static void vxlan_config_apply(struct net_device *dev,
3783 struct vxlan_config *conf,
3784 struct net_device *lowerdev,
3785 struct net *src_net,
3786 bool changelink)
3787{
3788 struct vxlan_dev *vxlan = netdev_priv(dev);
3789 struct vxlan_rdst *dst = &vxlan->default_dst;
3790 unsigned short needed_headroom = ETH_HLEN;
3791 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3792 int max_mtu = ETH_MAX_MTU;
3793
3794 if (!changelink) {
3795 if (conf->flags & VXLAN_F_GPE)
3796 vxlan_raw_setup(dev);
3797 else
3798 vxlan_ether_setup(dev);
3799
3800 if (conf->mtu)
3801 dev->mtu = conf->mtu;
3802
3803 vxlan->net = src_net;
3804 }
3805
3806 dst->remote_vni = conf->vni;
3807
3808 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3809
3810 if (lowerdev) {
3811 dst->remote_ifindex = conf->remote_ifindex;
3812
3813 dev->gso_max_size = lowerdev->gso_max_size;
3814 dev->gso_max_segs = lowerdev->gso_max_segs;
3815
3816 needed_headroom = lowerdev->hard_header_len;
3817 needed_headroom += lowerdev->needed_headroom;
3818
3819 dev->needed_tailroom = lowerdev->needed_tailroom;
3820
3821 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3822 VXLAN_HEADROOM);
3823 if (max_mtu < ETH_MIN_MTU)
3824 max_mtu = ETH_MIN_MTU;
3825
3826 if (!changelink && !conf->mtu)
3827 dev->mtu = max_mtu;
3828 }
3829
3830 if (dev->mtu > max_mtu)
3831 dev->mtu = max_mtu;
3832
3833 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3834 needed_headroom += VXLAN6_HEADROOM;
3835 else
3836 needed_headroom += VXLAN_HEADROOM;
3837 dev->needed_headroom = needed_headroom;
3838
3839 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3840}
3841
3842static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3843 struct vxlan_config *conf, bool changelink,
3844 struct netlink_ext_ack *extack)
3845{
3846 struct vxlan_dev *vxlan = netdev_priv(dev);
3847 struct net_device *lowerdev;
3848 int ret;
3849
3850 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3851 if (ret)
3852 return ret;
3853
3854 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3855
3856 return 0;
3857}
3858
3859static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3860 struct vxlan_config *conf,
3861 struct netlink_ext_ack *extack)
3862{
3863 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3864 struct vxlan_dev *vxlan = netdev_priv(dev);
3865 struct net_device *remote_dev = NULL;
3866 struct vxlan_fdb *f = NULL;
3867 bool unregister = false;
3868 struct vxlan_rdst *dst;
3869 int err;
3870
3871 dst = &vxlan->default_dst;
3872 err = vxlan_dev_configure(net, dev, conf, false, extack);
3873 if (err)
3874 return err;
3875
3876 dev->ethtool_ops = &vxlan_ethtool_ops;
3877
3878 /* create an fdb entry for a valid default destination */
3879 if (!vxlan_addr_any(&dst->remote_ip)) {
3880 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3881 &dst->remote_ip,
3882 NUD_REACHABLE | NUD_PERMANENT,
3883 vxlan->cfg.dst_port,
3884 dst->remote_vni,
3885 dst->remote_vni,
3886 dst->remote_ifindex,
3887 NTF_SELF, 0, &f, extack);
3888 if (err)
3889 return err;
3890 }
3891
3892 err = register_netdevice(dev);
3893 if (err)
3894 goto errout;
3895 unregister = true;
3896
3897 if (dst->remote_ifindex) {
3898 remote_dev = __dev_get_by_index(net, dst->remote_ifindex);
3899 if (!remote_dev) {
3900 err = -ENODEV;
3901 goto errout;
3902 }
3903
3904 err = netdev_upper_dev_link(remote_dev, dev, extack);
3905 if (err)
3906 goto errout;
3907 }
3908
3909 err = rtnl_configure_link(dev, NULL);
3910 if (err < 0)
3911 goto unlink;
3912
3913 if (f) {
3914 vxlan_fdb_insert(vxlan, all_zeros_mac, dst->remote_vni, f);
3915
3916 /* notify default fdb entry */
3917 err = vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f),
3918 RTM_NEWNEIGH, true, extack);
3919 if (err) {
3920 vxlan_fdb_destroy(vxlan, f, false, false);
3921 if (remote_dev)
3922 netdev_upper_dev_unlink(remote_dev, dev);
3923 goto unregister;
3924 }
3925 }
3926
3927 list_add(&vxlan->next, &vn->vxlan_list);
3928 if (remote_dev)
3929 dst->remote_dev = remote_dev;
3930 return 0;
3931unlink:
3932 if (remote_dev)
3933 netdev_upper_dev_unlink(remote_dev, dev);
3934errout:
3935 /* unregister_netdevice() destroys the default FDB entry with deletion
3936 * notification. But the addition notification was not sent yet, so
3937 * destroy the entry by hand here.
3938 */
3939 if (f)
3940 __vxlan_fdb_free(f);
3941unregister:
3942 if (unregister)
3943 unregister_netdevice(dev);
3944 return err;
3945}
3946
3947/* Set/clear flags based on attribute */
3948static int vxlan_nl2flag(struct vxlan_config *conf, struct nlattr *tb[],
3949 int attrtype, unsigned long mask, bool changelink,
3950 bool changelink_supported,
3951 struct netlink_ext_ack *extack)
3952{
3953 unsigned long flags;
3954
3955 if (!tb[attrtype])
3956 return 0;
3957
3958 if (changelink && !changelink_supported) {
3959 vxlan_flag_attr_error(attrtype, extack);
3960 return -EOPNOTSUPP;
3961 }
3962
3963 if (vxlan_policy[attrtype].type == NLA_FLAG)
3964 flags = conf->flags | mask;
3965 else if (nla_get_u8(tb[attrtype]))
3966 flags = conf->flags | mask;
3967 else
3968 flags = conf->flags & ~mask;
3969
3970 conf->flags = flags;
3971
3972 return 0;
3973}
3974
3975static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3976 struct net_device *dev, struct vxlan_config *conf,
3977 bool changelink, struct netlink_ext_ack *extack)
3978{
3979 struct vxlan_dev *vxlan = netdev_priv(dev);
3980 int err = 0;
3981
3982 memset(conf, 0, sizeof(*conf));
3983
3984 /* if changelink operation, start with old existing cfg */
3985 if (changelink)
3986 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3987
3988 if (data[IFLA_VXLAN_ID]) {
3989 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3990
3991 if (changelink && (vni != conf->vni)) {
3992 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID], "Cannot change VNI");
3993 return -EOPNOTSUPP;
3994 }
3995 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3996 }
3997
3998 if (data[IFLA_VXLAN_GROUP]) {
3999 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET)) {
4000 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP], "New group address family does not match old group");
4001 return -EOPNOTSUPP;
4002 }
4003
4004 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
4005 conf->remote_ip.sa.sa_family = AF_INET;
4006 } else if (data[IFLA_VXLAN_GROUP6]) {
4007 if (!IS_ENABLED(CONFIG_IPV6)) {
4008 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "IPv6 support not enabled in the kernel");
4009 return -EPFNOSUPPORT;
4010 }
4011
4012 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6)) {
4013 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_GROUP6], "New group address family does not match old group");
4014 return -EOPNOTSUPP;
4015 }
4016
4017 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
4018 conf->remote_ip.sa.sa_family = AF_INET6;
4019 }
4020
4021 if (data[IFLA_VXLAN_LOCAL]) {
4022 if (changelink && (conf->saddr.sa.sa_family != AF_INET)) {
4023 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL], "New local address family does not match old");
4024 return -EOPNOTSUPP;
4025 }
4026
4027 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
4028 conf->saddr.sa.sa_family = AF_INET;
4029 } else if (data[IFLA_VXLAN_LOCAL6]) {
4030 if (!IS_ENABLED(CONFIG_IPV6)) {
4031 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "IPv6 support not enabled in the kernel");
4032 return -EPFNOSUPPORT;
4033 }
4034
4035 if (changelink && (conf->saddr.sa.sa_family != AF_INET6)) {
4036 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LOCAL6], "New local address family does not match old");
4037 return -EOPNOTSUPP;
4038 }
4039
4040 /* TODO: respect scope id */
4041 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
4042 conf->saddr.sa.sa_family = AF_INET6;
4043 }
4044
4045 if (data[IFLA_VXLAN_LINK])
4046 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
4047
4048 if (data[IFLA_VXLAN_TOS])
4049 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
4050
4051 if (data[IFLA_VXLAN_TTL])
4052 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
4053
4054 if (data[IFLA_VXLAN_TTL_INHERIT]) {
4055 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_TTL_INHERIT,
4056 VXLAN_F_TTL_INHERIT, changelink, false,
4057 extack);
4058 if (err)
4059 return err;
4060
4061 }
4062
4063 if (data[IFLA_VXLAN_LABEL])
4064 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
4065 IPV6_FLOWLABEL_MASK;
4066
4067 if (data[IFLA_VXLAN_LEARNING]) {
4068 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_LEARNING,
4069 VXLAN_F_LEARN, changelink, true,
4070 extack);
4071 if (err)
4072 return err;
4073 } else if (!changelink) {
4074 /* default to learn on a new device */
4075 conf->flags |= VXLAN_F_LEARN;
4076 }
4077
4078 if (data[IFLA_VXLAN_AGEING])
4079 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
4080
4081 if (data[IFLA_VXLAN_PROXY]) {
4082 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_PROXY,
4083 VXLAN_F_PROXY, changelink, false,
4084 extack);
4085 if (err)
4086 return err;
4087 }
4088
4089 if (data[IFLA_VXLAN_RSC]) {
4090 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_RSC,
4091 VXLAN_F_RSC, changelink, false,
4092 extack);
4093 if (err)
4094 return err;
4095 }
4096
4097 if (data[IFLA_VXLAN_L2MISS]) {
4098 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L2MISS,
4099 VXLAN_F_L2MISS, changelink, false,
4100 extack);
4101 if (err)
4102 return err;
4103 }
4104
4105 if (data[IFLA_VXLAN_L3MISS]) {
4106 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_L3MISS,
4107 VXLAN_F_L3MISS, changelink, false,
4108 extack);
4109 if (err)
4110 return err;
4111 }
4112
4113 if (data[IFLA_VXLAN_LIMIT]) {
4114 if (changelink) {
4115 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_LIMIT],
4116 "Cannot change limit");
4117 return -EOPNOTSUPP;
4118 }
4119 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
4120 }
4121
4122 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
4123 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_COLLECT_METADATA,
4124 VXLAN_F_COLLECT_METADATA, changelink, false,
4125 extack);
4126 if (err)
4127 return err;
4128 }
4129
4130 if (data[IFLA_VXLAN_PORT_RANGE]) {
4131 if (!changelink) {
4132 const struct ifla_vxlan_port_range *p
4133 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
4134 conf->port_min = ntohs(p->low);
4135 conf->port_max = ntohs(p->high);
4136 } else {
4137 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
4138 "Cannot change port range");
4139 return -EOPNOTSUPP;
4140 }
4141 }
4142
4143 if (data[IFLA_VXLAN_PORT]) {
4144 if (changelink) {
4145 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT],
4146 "Cannot change port");
4147 return -EOPNOTSUPP;
4148 }
4149 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
4150 }
4151
4152 if (data[IFLA_VXLAN_UDP_CSUM]) {
4153 if (changelink) {
4154 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_UDP_CSUM],
4155 "Cannot change UDP_CSUM flag");
4156 return -EOPNOTSUPP;
4157 }
4158 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
4159 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
4160 }
4161
4162 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
4163 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4164 VXLAN_F_UDP_ZERO_CSUM6_TX, changelink,
4165 false, extack);
4166 if (err)
4167 return err;
4168 }
4169
4170 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
4171 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4172 VXLAN_F_UDP_ZERO_CSUM6_RX, changelink,
4173 false, extack);
4174 if (err)
4175 return err;
4176 }
4177
4178 if (data[IFLA_VXLAN_REMCSUM_TX]) {
4179 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_TX,
4180 VXLAN_F_REMCSUM_TX, changelink, false,
4181 extack);
4182 if (err)
4183 return err;
4184 }
4185
4186 if (data[IFLA_VXLAN_REMCSUM_RX]) {
4187 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_RX,
4188 VXLAN_F_REMCSUM_RX, changelink, false,
4189 extack);
4190 if (err)
4191 return err;
4192 }
4193
4194 if (data[IFLA_VXLAN_GBP]) {
4195 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GBP,
4196 VXLAN_F_GBP, changelink, false, extack);
4197 if (err)
4198 return err;
4199 }
4200
4201 if (data[IFLA_VXLAN_GPE]) {
4202 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_GPE,
4203 VXLAN_F_GPE, changelink, false,
4204 extack);
4205 if (err)
4206 return err;
4207 }
4208
4209 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
4210 err = vxlan_nl2flag(conf, data, IFLA_VXLAN_REMCSUM_NOPARTIAL,
4211 VXLAN_F_REMCSUM_NOPARTIAL, changelink,
4212 false, extack);
4213 if (err)
4214 return err;
4215 }
4216
4217 if (tb[IFLA_MTU]) {
4218 if (changelink) {
4219 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
4220 "Cannot change mtu");
4221 return -EOPNOTSUPP;
4222 }
4223 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
4224 }
4225
4226 if (data[IFLA_VXLAN_DF])
4227 conf->df = nla_get_u8(data[IFLA_VXLAN_DF]);
4228
4229 return 0;
4230}
4231
4232static int vxlan_newlink(struct net *src_net, struct net_device *dev,
4233 struct nlattr *tb[], struct nlattr *data[],
4234 struct netlink_ext_ack *extack)
4235{
4236 struct vxlan_config conf;
4237 int err;
4238
4239 err = vxlan_nl2conf(tb, data, dev, &conf, false, extack);
4240 if (err)
4241 return err;
4242
4243 return __vxlan_dev_create(src_net, dev, &conf, extack);
4244}
4245
4246static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
4247 struct nlattr *data[],
4248 struct netlink_ext_ack *extack)
4249{
4250 struct vxlan_dev *vxlan = netdev_priv(dev);
4251 struct net_device *lowerdev;
4252 struct vxlan_config conf;
4253 struct vxlan_rdst *dst;
4254 int err;
4255
4256 dst = &vxlan->default_dst;
4257 err = vxlan_nl2conf(tb, data, dev, &conf, true, extack);
4258 if (err)
4259 return err;
4260
4261 err = vxlan_config_validate(vxlan->net, &conf, &lowerdev,
4262 vxlan, extack);
4263 if (err)
4264 return err;
4265
4266 if (dst->remote_dev == lowerdev)
4267 lowerdev = NULL;
4268
4269 err = netdev_adjacent_change_prepare(dst->remote_dev, lowerdev, dev,
4270 extack);
4271 if (err)
4272 return err;
4273
4274 /* handle default dst entry */
4275 if (!vxlan_addr_equal(&conf.remote_ip, &dst->remote_ip)) {
4276 u32 hash_index = fdb_head_index(vxlan, all_zeros_mac, conf.vni);
4277
4278 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4279 if (!vxlan_addr_any(&conf.remote_ip)) {
4280 err = vxlan_fdb_update(vxlan, all_zeros_mac,
4281 &conf.remote_ip,
4282 NUD_REACHABLE | NUD_PERMANENT,
4283 NLM_F_APPEND | NLM_F_CREATE,
4284 vxlan->cfg.dst_port,
4285 conf.vni, conf.vni,
4286 conf.remote_ifindex,
4287 NTF_SELF, 0, true, extack);
4288 if (err) {
4289 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4290 netdev_adjacent_change_abort(dst->remote_dev,
4291 lowerdev, dev);
4292 return err;
4293 }
4294 }
4295 if (!vxlan_addr_any(&dst->remote_ip))
4296 __vxlan_fdb_delete(vxlan, all_zeros_mac,
4297 dst->remote_ip,
4298 vxlan->cfg.dst_port,
4299 dst->remote_vni,
4300 dst->remote_vni,
4301 dst->remote_ifindex,
4302 true);
4303 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4304 }
4305
4306 if (conf.age_interval != vxlan->cfg.age_interval)
4307 mod_timer(&vxlan->age_timer, jiffies);
4308
4309 netdev_adjacent_change_commit(dst->remote_dev, lowerdev, dev);
4310 if (lowerdev && lowerdev != dst->remote_dev)
4311 dst->remote_dev = lowerdev;
4312 vxlan_config_apply(dev, &conf, lowerdev, vxlan->net, true);
4313 return 0;
4314}
4315
4316static void vxlan_dellink(struct net_device *dev, struct list_head *head)
4317{
4318 struct vxlan_dev *vxlan = netdev_priv(dev);
4319
4320 vxlan_flush(vxlan, true);
4321
4322 list_del(&vxlan->next);
4323 unregister_netdevice_queue(dev, head);
4324 if (vxlan->default_dst.remote_dev)
4325 netdev_upper_dev_unlink(vxlan->default_dst.remote_dev, dev);
4326}
4327
4328static size_t vxlan_get_size(const struct net_device *dev)
4329{
4330
4331 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
4332 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
4333 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
4334 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
4335 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
4336 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
4337 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
4338 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_DF */
4339 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
4340 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
4341 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
4342 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
4343 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
4344 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
4345 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
4346 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
4347 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
4348 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
4349 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
4350 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
4351 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
4352 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
4353 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
4354 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
4355 0;
4356}
4357
4358static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
4359{
4360 const struct vxlan_dev *vxlan = netdev_priv(dev);
4361 const struct vxlan_rdst *dst = &vxlan->default_dst;
4362 struct ifla_vxlan_port_range ports = {
4363 .low = htons(vxlan->cfg.port_min),
4364 .high = htons(vxlan->cfg.port_max),
4365 };
4366
4367 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
4368 goto nla_put_failure;
4369
4370 if (!vxlan_addr_any(&dst->remote_ip)) {
4371 if (dst->remote_ip.sa.sa_family == AF_INET) {
4372 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
4373 dst->remote_ip.sin.sin_addr.s_addr))
4374 goto nla_put_failure;
4375#if IS_ENABLED(CONFIG_IPV6)
4376 } else {
4377 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
4378 &dst->remote_ip.sin6.sin6_addr))
4379 goto nla_put_failure;
4380#endif
4381 }
4382 }
4383
4384 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
4385 goto nla_put_failure;
4386
4387 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
4388 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
4389 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
4390 vxlan->cfg.saddr.sin.sin_addr.s_addr))
4391 goto nla_put_failure;
4392#if IS_ENABLED(CONFIG_IPV6)
4393 } else {
4394 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
4395 &vxlan->cfg.saddr.sin6.sin6_addr))
4396 goto nla_put_failure;
4397#endif
4398 }
4399 }
4400
4401 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
4402 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
4403 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
4404 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
4405 nla_put_u8(skb, IFLA_VXLAN_DF, vxlan->cfg.df) ||
4406 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
4407 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
4408 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
4409 nla_put_u8(skb, IFLA_VXLAN_PROXY,
4410 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
4411 nla_put_u8(skb, IFLA_VXLAN_RSC,
4412 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
4413 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
4414 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
4415 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
4416 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
4417 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
4418 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
4419 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
4420 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
4421 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
4422 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
4423 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
4424 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
4425 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
4426 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
4427 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
4428 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
4429 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
4430 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
4431 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
4432 goto nla_put_failure;
4433
4434 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
4435 goto nla_put_failure;
4436
4437 if (vxlan->cfg.flags & VXLAN_F_GBP &&
4438 nla_put_flag(skb, IFLA_VXLAN_GBP))
4439 goto nla_put_failure;
4440
4441 if (vxlan->cfg.flags & VXLAN_F_GPE &&
4442 nla_put_flag(skb, IFLA_VXLAN_GPE))
4443 goto nla_put_failure;
4444
4445 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
4446 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
4447 goto nla_put_failure;
4448
4449 return 0;
4450
4451nla_put_failure:
4452 return -EMSGSIZE;
4453}
4454
4455static struct net *vxlan_get_link_net(const struct net_device *dev)
4456{
4457 struct vxlan_dev *vxlan = netdev_priv(dev);
4458
4459 return vxlan->net;
4460}
4461
4462static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
4463 .kind = "vxlan",
4464 .maxtype = IFLA_VXLAN_MAX,
4465 .policy = vxlan_policy,
4466 .priv_size = sizeof(struct vxlan_dev),
4467 .setup = vxlan_setup,
4468 .validate = vxlan_validate,
4469 .newlink = vxlan_newlink,
4470 .changelink = vxlan_changelink,
4471 .dellink = vxlan_dellink,
4472 .get_size = vxlan_get_size,
4473 .fill_info = vxlan_fill_info,
4474 .get_link_net = vxlan_get_link_net,
4475};
4476
4477struct net_device *vxlan_dev_create(struct net *net, const char *name,
4478 u8 name_assign_type,
4479 struct vxlan_config *conf)
4480{
4481 struct nlattr *tb[IFLA_MAX + 1];
4482 struct net_device *dev;
4483 int err;
4484
4485 memset(&tb, 0, sizeof(tb));
4486
4487 dev = rtnl_create_link(net, name, name_assign_type,
4488 &vxlan_link_ops, tb, NULL);
4489 if (IS_ERR(dev))
4490 return dev;
4491
4492 err = __vxlan_dev_create(net, dev, conf, NULL);
4493 if (err < 0) {
4494 free_netdev(dev);
4495 return ERR_PTR(err);
4496 }
4497
4498 err = rtnl_configure_link(dev, NULL);
4499 if (err < 0) {
4500 LIST_HEAD(list_kill);
4501
4502 vxlan_dellink(dev, &list_kill);
4503 unregister_netdevice_many(&list_kill);
4504 return ERR_PTR(err);
4505 }
4506
4507 return dev;
4508}
4509EXPORT_SYMBOL_GPL(vxlan_dev_create);
4510
4511static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
4512 struct net_device *dev)
4513{
4514 struct vxlan_dev *vxlan, *next;
4515 LIST_HEAD(list_kill);
4516
4517 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4518 struct vxlan_rdst *dst = &vxlan->default_dst;
4519
4520 /* In case we created vxlan device with carrier
4521 * and we loose the carrier due to module unload
4522 * we also need to remove vxlan device. In other
4523 * cases, it's not necessary and remote_ifindex
4524 * is 0 here, so no matches.
4525 */
4526 if (dst->remote_ifindex == dev->ifindex)
4527 vxlan_dellink(vxlan->dev, &list_kill);
4528 }
4529
4530 unregister_netdevice_many(&list_kill);
4531}
4532
4533static int vxlan_netdevice_event(struct notifier_block *unused,
4534 unsigned long event, void *ptr)
4535{
4536 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
4537 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
4538
4539 if (event == NETDEV_UNREGISTER)
4540 vxlan_handle_lowerdev_unregister(vn, dev);
4541 else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO)
4542 vxlan_offload_rx_ports(dev, true);
4543 else if (event == NETDEV_UDP_TUNNEL_DROP_INFO)
4544 vxlan_offload_rx_ports(dev, false);
4545
4546 return NOTIFY_DONE;
4547}
4548
4549static struct notifier_block vxlan_notifier_block __read_mostly = {
4550 .notifier_call = vxlan_netdevice_event,
4551};
4552
4553static void
4554vxlan_fdb_offloaded_set(struct net_device *dev,
4555 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4556{
4557 struct vxlan_dev *vxlan = netdev_priv(dev);
4558 struct vxlan_rdst *rdst;
4559 struct vxlan_fdb *f;
4560 u32 hash_index;
4561
4562 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4563
4564 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4565
4566 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4567 if (!f)
4568 goto out;
4569
4570 rdst = vxlan_fdb_find_rdst(f, &fdb_info->remote_ip,
4571 fdb_info->remote_port,
4572 fdb_info->remote_vni,
4573 fdb_info->remote_ifindex);
4574 if (!rdst)
4575 goto out;
4576
4577 rdst->offloaded = fdb_info->offloaded;
4578
4579out:
4580 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4581}
4582
4583static int
4584vxlan_fdb_external_learn_add(struct net_device *dev,
4585 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4586{
4587 struct vxlan_dev *vxlan = netdev_priv(dev);
4588 struct netlink_ext_ack *extack;
4589 u32 hash_index;
4590 int err;
4591
4592 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4593 extack = switchdev_notifier_info_to_extack(&fdb_info->info);
4594
4595 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4596 err = vxlan_fdb_update(vxlan, fdb_info->eth_addr, &fdb_info->remote_ip,
4597 NUD_REACHABLE,
4598 NLM_F_CREATE | NLM_F_REPLACE,
4599 fdb_info->remote_port,
4600 fdb_info->vni,
4601 fdb_info->remote_vni,
4602 fdb_info->remote_ifindex,
4603 NTF_USE | NTF_SELF | NTF_EXT_LEARNED,
4604 0, false, extack);
4605 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4606
4607 return err;
4608}
4609
4610static int
4611vxlan_fdb_external_learn_del(struct net_device *dev,
4612 struct switchdev_notifier_vxlan_fdb_info *fdb_info)
4613{
4614 struct vxlan_dev *vxlan = netdev_priv(dev);
4615 struct vxlan_fdb *f;
4616 u32 hash_index;
4617 int err = 0;
4618
4619 hash_index = fdb_head_index(vxlan, fdb_info->eth_addr, fdb_info->vni);
4620 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4621
4622 f = vxlan_find_mac(vxlan, fdb_info->eth_addr, fdb_info->vni);
4623 if (!f)
4624 err = -ENOENT;
4625 else if (f->flags & NTF_EXT_LEARNED)
4626 err = __vxlan_fdb_delete(vxlan, fdb_info->eth_addr,
4627 fdb_info->remote_ip,
4628 fdb_info->remote_port,
4629 fdb_info->vni,
4630 fdb_info->remote_vni,
4631 fdb_info->remote_ifindex,
4632 false);
4633
4634 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4635
4636 return err;
4637}
4638
4639static int vxlan_switchdev_event(struct notifier_block *unused,
4640 unsigned long event, void *ptr)
4641{
4642 struct net_device *dev = switchdev_notifier_info_to_dev(ptr);
4643 struct switchdev_notifier_vxlan_fdb_info *fdb_info;
4644 int err = 0;
4645
4646 switch (event) {
4647 case SWITCHDEV_VXLAN_FDB_OFFLOADED:
4648 vxlan_fdb_offloaded_set(dev, ptr);
4649 break;
4650 case SWITCHDEV_VXLAN_FDB_ADD_TO_BRIDGE:
4651 fdb_info = ptr;
4652 err = vxlan_fdb_external_learn_add(dev, fdb_info);
4653 if (err) {
4654 err = notifier_from_errno(err);
4655 break;
4656 }
4657 fdb_info->offloaded = true;
4658 vxlan_fdb_offloaded_set(dev, fdb_info);
4659 break;
4660 case SWITCHDEV_VXLAN_FDB_DEL_TO_BRIDGE:
4661 fdb_info = ptr;
4662 err = vxlan_fdb_external_learn_del(dev, fdb_info);
4663 if (err) {
4664 err = notifier_from_errno(err);
4665 break;
4666 }
4667 fdb_info->offloaded = false;
4668 vxlan_fdb_offloaded_set(dev, fdb_info);
4669 break;
4670 }
4671
4672 return err;
4673}
4674
4675static struct notifier_block vxlan_switchdev_notifier_block __read_mostly = {
4676 .notifier_call = vxlan_switchdev_event,
4677};
4678
4679static void vxlan_fdb_nh_flush(struct nexthop *nh)
4680{
4681 struct vxlan_fdb *fdb;
4682 struct vxlan_dev *vxlan;
4683 u32 hash_index;
4684
4685 rcu_read_lock();
4686 list_for_each_entry_rcu(fdb, &nh->fdb_list, nh_list) {
4687 vxlan = rcu_dereference(fdb->vdev);
4688 WARN_ON(!vxlan);
4689 hash_index = fdb_head_index(vxlan, fdb->eth_addr,
4690 vxlan->default_dst.remote_vni);
4691 spin_lock_bh(&vxlan->hash_lock[hash_index]);
4692 if (!hlist_unhashed(&fdb->hlist))
4693 vxlan_fdb_destroy(vxlan, fdb, false, false);
4694 spin_unlock_bh(&vxlan->hash_lock[hash_index]);
4695 }
4696 rcu_read_unlock();
4697}
4698
4699static int vxlan_nexthop_event(struct notifier_block *nb,
4700 unsigned long event, void *ptr)
4701{
4702 struct nh_notifier_info *info = ptr;
4703 struct nexthop *nh;
4704
4705 if (event != NEXTHOP_EVENT_DEL)
4706 return NOTIFY_DONE;
4707
4708 nh = nexthop_find_by_id(info->net, info->id);
4709 if (!nh)
4710 return NOTIFY_DONE;
4711
4712 vxlan_fdb_nh_flush(nh);
4713
4714 return NOTIFY_DONE;
4715}
4716
4717static __net_init int vxlan_init_net(struct net *net)
4718{
4719 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4720 unsigned int h;
4721
4722 INIT_LIST_HEAD(&vn->vxlan_list);
4723 spin_lock_init(&vn->sock_lock);
4724 vn->nexthop_notifier_block.notifier_call = vxlan_nexthop_event;
4725
4726 for (h = 0; h < PORT_HASH_SIZE; ++h)
4727 INIT_HLIST_HEAD(&vn->sock_list[h]);
4728
4729 return register_nexthop_notifier(net, &vn->nexthop_notifier_block,
4730 NULL);
4731}
4732
4733static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
4734{
4735 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4736 struct vxlan_dev *vxlan, *next;
4737 struct net_device *dev, *aux;
4738
4739 for_each_netdev_safe(net, dev, aux)
4740 if (dev->rtnl_link_ops == &vxlan_link_ops)
4741 unregister_netdevice_queue(dev, head);
4742
4743 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
4744 /* If vxlan->dev is in the same netns, it has already been added
4745 * to the list by the previous loop.
4746 */
4747 if (!net_eq(dev_net(vxlan->dev), net))
4748 unregister_netdevice_queue(vxlan->dev, head);
4749 }
4750
4751}
4752
4753static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
4754{
4755 struct net *net;
4756 LIST_HEAD(list);
4757 unsigned int h;
4758
4759 list_for_each_entry(net, net_list, exit_list) {
4760 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4761
4762 unregister_nexthop_notifier(net, &vn->nexthop_notifier_block);
4763 }
4764 rtnl_lock();
4765 list_for_each_entry(net, net_list, exit_list)
4766 vxlan_destroy_tunnels(net, &list);
4767
4768 unregister_netdevice_many(&list);
4769 rtnl_unlock();
4770
4771 list_for_each_entry(net, net_list, exit_list) {
4772 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
4773
4774 for (h = 0; h < PORT_HASH_SIZE; ++h)
4775 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
4776 }
4777}
4778
4779static struct pernet_operations vxlan_net_ops = {
4780 .init = vxlan_init_net,
4781 .exit_batch = vxlan_exit_batch_net,
4782 .id = &vxlan_net_id,
4783 .size = sizeof(struct vxlan_net),
4784};
4785
4786static int __init vxlan_init_module(void)
4787{
4788 int rc;
4789
4790 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
4791
4792 rc = register_pernet_subsys(&vxlan_net_ops);
4793 if (rc)
4794 goto out1;
4795
4796 rc = register_netdevice_notifier(&vxlan_notifier_block);
4797 if (rc)
4798 goto out2;
4799
4800 rc = register_switchdev_notifier(&vxlan_switchdev_notifier_block);
4801 if (rc)
4802 goto out3;
4803
4804 rc = rtnl_link_register(&vxlan_link_ops);
4805 if (rc)
4806 goto out4;
4807
4808 return 0;
4809out4:
4810 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4811out3:
4812 unregister_netdevice_notifier(&vxlan_notifier_block);
4813out2:
4814 unregister_pernet_subsys(&vxlan_net_ops);
4815out1:
4816 return rc;
4817}
4818late_initcall(vxlan_init_module);
4819
4820static void __exit vxlan_cleanup_module(void)
4821{
4822 rtnl_link_unregister(&vxlan_link_ops);
4823 unregister_switchdev_notifier(&vxlan_switchdev_notifier_block);
4824 unregister_netdevice_notifier(&vxlan_notifier_block);
4825 unregister_pernet_subsys(&vxlan_net_ops);
4826 /* rcu_barrier() is called by netns */
4827}
4828module_exit(vxlan_cleanup_module);
4829
4830MODULE_LICENSE("GPL");
4831MODULE_VERSION(VXLAN_VERSION);
4832MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4833MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4834MODULE_ALIAS_RTNL_LINK("vxlan");