?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.19.133
[linux/fpc-iii.git] / drivers / net / vxlan.c
blob09f0b53b2b773fe1ea9f370a398c07822ff5242a
1 /*
2 * VXLAN: Virtual eXtensible Local Area Network
3 *
4 * Copyright (c) 2012-2013 Vyatta Inc.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 */
10
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/errno.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <linux/igmp.h>
19 #include <linux/if_ether.h>
20 #include <linux/ethtool.h>
21 #include <net/arp.h>
22 #include <net/ndisc.h>
23 #include <net/ip.h>
24 #include <net/icmp.h>
25 #include <net/rtnetlink.h>
26 #include <net/inet_ecn.h>
27 #include <net/net_namespace.h>
28 #include <net/netns/generic.h>
29 #include <net/tun_proto.h>
30 #include <net/vxlan.h>
31
32 #if IS_ENABLED(CONFIG_IPV6)
33 #include <net/ip6_tunnel.h>
34 #include <net/ip6_checksum.h>
35 #endif
36
37 #define VXLAN_VERSION "0.1"
38
39 #define PORT_HASH_BITS 8
40 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
41 #define FDB_AGE_DEFAULT 300 /* 5 min */
42 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
43
44 /* UDP port for VXLAN traffic.
45 * The IANA assigned port is 4789, but the Linux default is 8472
46 * for compatibility with early adopters.
47 */
48 static unsigned short vxlan_port __read_mostly = 8472;
49 module_param_named(udp_port, vxlan_port, ushort, 0444);
50 MODULE_PARM_DESC(udp_port, "Destination UDP port");
51
52 static bool log_ecn_error = true;
53 module_param(log_ecn_error, bool, 0644);
54 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
55
56 static unsigned int vxlan_net_id;
57 static struct rtnl_link_ops vxlan_link_ops;
58
59 static const u8 all_zeros_mac[ETH_ALEN + 2];
60
61 static int vxlan_sock_add(struct vxlan_dev *vxlan);
62
63 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
64
65 /* per-network namespace private data for this module */
66 struct vxlan_net {
67 struct list_head vxlan_list;
68 struct hlist_head sock_list[PORT_HASH_SIZE];
69 spinlock_t sock_lock;
70 };
71
72 /* Forwarding table entry */
73 struct 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 u8 flags; /* see ndm_flags */
83 };
84
85 /* salt for hash table */
86 static u32 vxlan_salt __read_mostly;
87
88 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
89 {
90 return vs->flags & VXLAN_F_COLLECT_METADATA ||
91 ip_tunnel_collect_metadata();
92 }
93
94 #if IS_ENABLED(CONFIG_IPV6)
95 static inline
96 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
97 {
98 if (a->sa.sa_family != b->sa.sa_family)
99 return false;
100 if (a->sa.sa_family == AF_INET6)
101 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
102 else
103 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
104 }
105
106 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
107 {
108 if (ipa->sa.sa_family == AF_INET6)
109 return ipv6_addr_any(&ipa->sin6.sin6_addr);
110 else
111 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
112 }
113
114 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
115 {
116 if (ipa->sa.sa_family == AF_INET6)
117 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
118 else
119 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
120 }
121
122 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
123 {
124 if (nla_len(nla) >= sizeof(struct in6_addr)) {
125 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
126 ip->sa.sa_family = AF_INET6;
127 return 0;
128 } else if (nla_len(nla) >= sizeof(__be32)) {
129 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
130 ip->sa.sa_family = AF_INET;
131 return 0;
132 } else {
133 return -EAFNOSUPPORT;
134 }
135 }
136
137 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
138 const union vxlan_addr *ip)
139 {
140 if (ip->sa.sa_family == AF_INET6)
141 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
142 else
143 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
144 }
145
146 #else /* !CONFIG_IPV6 */
147
148 static inline
149 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
150 {
151 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
152 }
153
154 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
155 {
156 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
157 }
158
159 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
160 {
161 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
162 }
163
164 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
165 {
166 if (nla_len(nla) >= sizeof(struct in6_addr)) {
167 return -EAFNOSUPPORT;
168 } else if (nla_len(nla) >= sizeof(__be32)) {
169 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
170 ip->sa.sa_family = AF_INET;
171 return 0;
172 } else {
173 return -EAFNOSUPPORT;
174 }
175 }
176
177 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
178 const union vxlan_addr *ip)
179 {
180 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
181 }
182 #endif
183
184 /* Virtual Network hash table head */
185 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
186 {
187 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
188 }
189
190 /* Socket hash table head */
191 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
192 {
193 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
194
195 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
196 }
197
198 /* First remote destination for a forwarding entry.
199 * Guaranteed to be non-NULL because remotes are never deleted.
200 */
201 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
202 {
203 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
204 }
205
206 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
207 {
208 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
209 }
210
211 /* Find VXLAN socket based on network namespace, address family and UDP port
212 * and enabled unshareable flags.
213 */
214 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
215 __be16 port, u32 flags)
216 {
217 struct vxlan_sock *vs;
218
219 flags &= VXLAN_F_RCV_FLAGS;
220
221 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
222 if (inet_sk(vs->sock->sk)->inet_sport == port &&
223 vxlan_get_sk_family(vs) == family &&
224 vs->flags == flags)
225 return vs;
226 }
227 return NULL;
228 }
229
230 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
231 __be32 vni)
232 {
233 struct vxlan_dev_node *node;
234
235 /* For flow based devices, map all packets to VNI 0 */
236 if (vs->flags & VXLAN_F_COLLECT_METADATA)
237 vni = 0;
238
239 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
240 if (node->vxlan->default_dst.remote_vni != vni)
241 continue;
242
243 if (IS_ENABLED(CONFIG_IPV6)) {
244 const struct vxlan_config *cfg = &node->vxlan->cfg;
245
246 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
247 cfg->remote_ifindex != ifindex)
248 continue;
249 }
250
251 return node->vxlan;
252 }
253
254 return NULL;
255 }
256
257 /* Look up VNI in a per net namespace table */
258 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
259 __be32 vni, sa_family_t family,
260 __be16 port, u32 flags)
261 {
262 struct vxlan_sock *vs;
263
264 vs = vxlan_find_sock(net, family, port, flags);
265 if (!vs)
266 return NULL;
267
268 return vxlan_vs_find_vni(vs, ifindex, vni);
269 }
270
271 /* Fill in neighbour message in skbuff. */
272 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
273 const struct vxlan_fdb *fdb,
274 u32 portid, u32 seq, int type, unsigned int flags,
275 const struct vxlan_rdst *rdst)
276 {
277 unsigned long now = jiffies;
278 struct nda_cacheinfo ci;
279 struct nlmsghdr *nlh;
280 struct ndmsg *ndm;
281 bool send_ip, send_eth;
282
283 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
284 if (nlh == NULL)
285 return -EMSGSIZE;
286
287 ndm = nlmsg_data(nlh);
288 memset(ndm, 0, sizeof(*ndm));
289
290 send_eth = send_ip = true;
291
292 if (type == RTM_GETNEIGH) {
293 send_ip = !vxlan_addr_any(&rdst->remote_ip);
294 send_eth = !is_zero_ether_addr(fdb->eth_addr);
295 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
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 ndm->ndm_type = RTN_UNICAST;
302
303 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
304 nla_put_s32(skb, NDA_LINK_NETNSID,
305 peernet2id(dev_net(vxlan->dev), vxlan->net)))
306 goto nla_put_failure;
307
308 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
309 goto nla_put_failure;
310
311 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
312 goto nla_put_failure;
313
314 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
315 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
316 goto nla_put_failure;
317 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
318 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
319 goto nla_put_failure;
320 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
321 nla_put_u32(skb, NDA_SRC_VNI,
322 be32_to_cpu(fdb->vni)))
323 goto nla_put_failure;
324 if (rdst->remote_ifindex &&
325 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
326 goto nla_put_failure;
327
328 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
329 ci.ndm_confirmed = 0;
330 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
331 ci.ndm_refcnt = 0;
332
333 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
334 goto nla_put_failure;
335
336 nlmsg_end(skb, nlh);
337 return 0;
338
339 nla_put_failure:
340 nlmsg_cancel(skb, nlh);
341 return -EMSGSIZE;
342 }
343
344 static inline size_t vxlan_nlmsg_size(void)
345 {
346 return NLMSG_ALIGN(sizeof(struct ndmsg))
347 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
348 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
349 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
350 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
351 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
352 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
353 + nla_total_size(sizeof(struct nda_cacheinfo));
354 }
355
356 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
357 struct vxlan_rdst *rd, int type)
358 {
359 struct net *net = dev_net(vxlan->dev);
360 struct sk_buff *skb;
361 int err = -ENOBUFS;
362
363 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
364 if (skb == NULL)
365 goto errout;
366
367 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
368 if (err < 0) {
369 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
370 WARN_ON(err == -EMSGSIZE);
371 kfree_skb(skb);
372 goto errout;
373 }
374
375 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
376 return;
377 errout:
378 if (err < 0)
379 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
380 }
381
382 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
383 {
384 struct vxlan_dev *vxlan = netdev_priv(dev);
385 struct vxlan_fdb f = {
386 .state = NUD_STALE,
387 };
388 struct vxlan_rdst remote = {
389 .remote_ip = *ipa, /* goes to NDA_DST */
390 .remote_vni = cpu_to_be32(VXLAN_N_VID),
391 };
392
393 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
394 }
395
396 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
397 {
398 struct vxlan_fdb f = {
399 .state = NUD_STALE,
400 };
401 struct vxlan_rdst remote = { };
402
403 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
404
405 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
406 }
407
408 /* Hash Ethernet address */
409 static u32 eth_hash(const unsigned char *addr)
410 {
411 u64 value = get_unaligned((u64 *)addr);
412
413 /* only want 6 bytes */
414 #ifdef __BIG_ENDIAN
415 value >>= 16;
416 #else
417 value <<= 16;
418 #endif
419 return hash_64(value, FDB_HASH_BITS);
420 }
421
422 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
423 {
424 /* use 1 byte of OUI and 3 bytes of NIC */
425 u32 key = get_unaligned((u32 *)(addr + 2));
426
427 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
428 }
429
430 /* Hash chain to use given mac address */
431 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
432 const u8 *mac, __be32 vni)
433 {
434 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
435 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
436 else
437 return &vxlan->fdb_head[eth_hash(mac)];
438 }
439
440 /* Look up Ethernet address in forwarding table */
441 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
442 const u8 *mac, __be32 vni)
443 {
444 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
445 struct vxlan_fdb *f;
446
447 hlist_for_each_entry_rcu(f, head, hlist) {
448 if (ether_addr_equal(mac, f->eth_addr)) {
449 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
450 if (vni == f->vni)
451 return f;
452 } else {
453 return f;
454 }
455 }
456 }
457
458 return NULL;
459 }
460
461 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
462 const u8 *mac, __be32 vni)
463 {
464 struct vxlan_fdb *f;
465
466 f = __vxlan_find_mac(vxlan, mac, vni);
467 if (f)
468 f->used = jiffies;
469
470 return f;
471 }
472
473 /* caller should hold vxlan->hash_lock */
474 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
475 union vxlan_addr *ip, __be16 port,
476 __be32 vni, __u32 ifindex)
477 {
478 struct vxlan_rdst *rd;
479
480 list_for_each_entry(rd, &f->remotes, list) {
481 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
482 rd->remote_port == port &&
483 rd->remote_vni == vni &&
484 rd->remote_ifindex == ifindex)
485 return rd;
486 }
487
488 return NULL;
489 }
490
491 /* Replace destination of unicast mac */
492 static int vxlan_fdb_replace(struct vxlan_fdb *f,
493 union vxlan_addr *ip, __be16 port, __be32 vni,
494 __u32 ifindex)
495 {
496 struct vxlan_rdst *rd;
497
498 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
499 if (rd)
500 return 0;
501
502 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
503 if (!rd)
504 return 0;
505
506 dst_cache_reset(&rd->dst_cache);
507 rd->remote_ip = *ip;
508 rd->remote_port = port;
509 rd->remote_vni = vni;
510 rd->remote_ifindex = ifindex;
511 return 1;
512 }
513
514 /* Add/update destinations for multicast */
515 static int vxlan_fdb_append(struct vxlan_fdb *f,
516 union vxlan_addr *ip, __be16 port, __be32 vni,
517 __u32 ifindex, struct vxlan_rdst **rdp)
518 {
519 struct vxlan_rdst *rd;
520
521 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
522 if (rd)
523 return 0;
524
525 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
526 if (rd == NULL)
527 return -ENOBUFS;
528
529 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
530 kfree(rd);
531 return -ENOBUFS;
532 }
533
534 rd->remote_ip = *ip;
535 rd->remote_port = port;
536 rd->remote_vni = vni;
537 rd->remote_ifindex = ifindex;
538
539 list_add_tail_rcu(&rd->list, &f->remotes);
540
541 *rdp = rd;
542 return 1;
543 }
544
545 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
546 unsigned int off,
547 struct vxlanhdr *vh, size_t hdrlen,
548 __be32 vni_field,
549 struct gro_remcsum *grc,
550 bool nopartial)
551 {
552 size_t start, offset;
553
554 if (skb->remcsum_offload)
555 return vh;
556
557 if (!NAPI_GRO_CB(skb)->csum_valid)
558 return NULL;
559
560 start = vxlan_rco_start(vni_field);
561 offset = start + vxlan_rco_offset(vni_field);
562
563 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
564 start, offset, grc, nopartial);
565
566 skb->remcsum_offload = 1;
567
568 return vh;
569 }
570
571 static struct sk_buff *vxlan_gro_receive(struct sock *sk,
572 struct list_head *head,
573 struct sk_buff *skb)
574 {
575 struct sk_buff *pp = NULL;
576 struct sk_buff *p;
577 struct vxlanhdr *vh, *vh2;
578 unsigned int hlen, off_vx;
579 int flush = 1;
580 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
581 __be32 flags;
582 struct gro_remcsum grc;
583
584 skb_gro_remcsum_init(&grc);
585
586 off_vx = skb_gro_offset(skb);
587 hlen = off_vx + sizeof(*vh);
588 vh = skb_gro_header_fast(skb, off_vx);
589 if (skb_gro_header_hard(skb, hlen)) {
590 vh = skb_gro_header_slow(skb, hlen, off_vx);
591 if (unlikely(!vh))
592 goto out;
593 }
594
595 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
596
597 flags = vh->vx_flags;
598
599 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
600 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
601 vh->vx_vni, &grc,
602 !!(vs->flags &
603 VXLAN_F_REMCSUM_NOPARTIAL));
604
605 if (!vh)
606 goto out;
607 }
608
609 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
610
611 list_for_each_entry(p, head, list) {
612 if (!NAPI_GRO_CB(p)->same_flow)
613 continue;
614
615 vh2 = (struct vxlanhdr *)(p->data + off_vx);
616 if (vh->vx_flags != vh2->vx_flags ||
617 vh->vx_vni != vh2->vx_vni) {
618 NAPI_GRO_CB(p)->same_flow = 0;
619 continue;
620 }
621 }
622
623 pp = call_gro_receive(eth_gro_receive, head, skb);
624 flush = 0;
625
626 out:
627 skb_gro_flush_final_remcsum(skb, pp, flush, &grc);
628
629 return pp;
630 }
631
632 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
633 {
634 /* Sets 'skb->inner_mac_header' since we are always called with
635 * 'skb->encapsulation' set.
636 */
637 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
638 }
639
640 static struct vxlan_fdb *vxlan_fdb_alloc(struct vxlan_dev *vxlan,
641 const u8 *mac, __u16 state,
642 __be32 src_vni, __u8 ndm_flags)
643 {
644 struct vxlan_fdb *f;
645
646 f = kmalloc(sizeof(*f), GFP_ATOMIC);
647 if (!f)
648 return NULL;
649 f->state = state;
650 f->flags = ndm_flags;
651 f->updated = f->used = jiffies;
652 f->vni = src_vni;
653 INIT_LIST_HEAD(&f->remotes);
654 memcpy(f->eth_addr, mac, ETH_ALEN);
655
656 return f;
657 }
658
659 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
660 const u8 *mac, union vxlan_addr *ip,
661 __u16 state, __be16 port, __be32 src_vni,
662 __be32 vni, __u32 ifindex, __u8 ndm_flags,
663 struct vxlan_fdb **fdb)
664 {
665 struct vxlan_rdst *rd = NULL;
666 struct vxlan_fdb *f;
667 int rc;
668
669 if (vxlan->cfg.addrmax &&
670 vxlan->addrcnt >= vxlan->cfg.addrmax)
671 return -ENOSPC;
672
673 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
674 f = vxlan_fdb_alloc(vxlan, mac, state, src_vni, ndm_flags);
675 if (!f)
676 return -ENOMEM;
677
678 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
679 if (rc < 0) {
680 kfree(f);
681 return rc;
682 }
683
684 ++vxlan->addrcnt;
685 hlist_add_head_rcu(&f->hlist,
686 vxlan_fdb_head(vxlan, mac, src_vni));
687
688 *fdb = f;
689
690 return 0;
691 }
692
693 /* Add new entry to forwarding table -- assumes lock held */
694 static int vxlan_fdb_update(struct vxlan_dev *vxlan,
695 const u8 *mac, union vxlan_addr *ip,
696 __u16 state, __u16 flags,
697 __be16 port, __be32 src_vni, __be32 vni,
698 __u32 ifindex, __u8 ndm_flags)
699 {
700 struct vxlan_rdst *rd = NULL;
701 struct vxlan_fdb *f;
702 int notify = 0;
703 int rc;
704
705 f = __vxlan_find_mac(vxlan, mac, src_vni);
706 if (f) {
707 if (flags & NLM_F_EXCL) {
708 netdev_dbg(vxlan->dev,
709 "lost race to create %pM\n", mac);
710 return -EEXIST;
711 }
712 if (f->state != state) {
713 f->state = state;
714 f->updated = jiffies;
715 notify = 1;
716 }
717 if (f->flags != ndm_flags) {
718 f->flags = ndm_flags;
719 f->updated = jiffies;
720 notify = 1;
721 }
722 if ((flags & NLM_F_REPLACE)) {
723 /* Only change unicasts */
724 if (!(is_multicast_ether_addr(f->eth_addr) ||
725 is_zero_ether_addr(f->eth_addr))) {
726 notify |= vxlan_fdb_replace(f, ip, port, vni,
727 ifindex);
728 } else
729 return -EOPNOTSUPP;
730 }
731 if ((flags & NLM_F_APPEND) &&
732 (is_multicast_ether_addr(f->eth_addr) ||
733 is_zero_ether_addr(f->eth_addr))) {
734 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
735
736 if (rc < 0)
737 return rc;
738 notify |= rc;
739 }
740 } else {
741 if (!(flags & NLM_F_CREATE))
742 return -ENOENT;
743
744 /* Disallow replace to add a multicast entry */
745 if ((flags & NLM_F_REPLACE) &&
746 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
747 return -EOPNOTSUPP;
748
749 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
750 rc = vxlan_fdb_create(vxlan, mac, ip, state, port, src_vni,
751 vni, ifindex, ndm_flags, &f);
752 if (rc < 0)
753 return rc;
754 notify = 1;
755 }
756
757 if (notify) {
758 if (rd == NULL)
759 rd = first_remote_rtnl(f);
760 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
761 }
762
763 return 0;
764 }
765
766 static void vxlan_fdb_free(struct rcu_head *head)
767 {
768 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
769 struct vxlan_rdst *rd, *nd;
770
771 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
772 dst_cache_destroy(&rd->dst_cache);
773 kfree(rd);
774 }
775 kfree(f);
776 }
777
778 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
779 bool do_notify)
780 {
781 netdev_dbg(vxlan->dev,
782 "delete %pM\n", f->eth_addr);
783
784 --vxlan->addrcnt;
785 if (do_notify)
786 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
787
788 hlist_del_rcu(&f->hlist);
789 call_rcu(&f->rcu, vxlan_fdb_free);
790 }
791
792 static void vxlan_dst_free(struct rcu_head *head)
793 {
794 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
795
796 dst_cache_destroy(&rd->dst_cache);
797 kfree(rd);
798 }
799
800 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
801 struct vxlan_rdst *rd)
802 {
803 list_del_rcu(&rd->list);
804 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
805 call_rcu(&rd->rcu, vxlan_dst_free);
806 }
807
808 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
809 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
810 __be32 *vni, u32 *ifindex)
811 {
812 struct net *net = dev_net(vxlan->dev);
813 int err;
814
815 if (tb[NDA_DST]) {
816 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
817 if (err)
818 return err;
819 } else {
820 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
821 if (remote->sa.sa_family == AF_INET) {
822 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
823 ip->sa.sa_family = AF_INET;
824 #if IS_ENABLED(CONFIG_IPV6)
825 } else {
826 ip->sin6.sin6_addr = in6addr_any;
827 ip->sa.sa_family = AF_INET6;
828 #endif
829 }
830 }
831
832 if (tb[NDA_PORT]) {
833 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
834 return -EINVAL;
835 *port = nla_get_be16(tb[NDA_PORT]);
836 } else {
837 *port = vxlan->cfg.dst_port;
838 }
839
840 if (tb[NDA_VNI]) {
841 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
842 return -EINVAL;
843 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
844 } else {
845 *vni = vxlan->default_dst.remote_vni;
846 }
847
848 if (tb[NDA_SRC_VNI]) {
849 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
850 return -EINVAL;
851 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
852 } else {
853 *src_vni = vxlan->default_dst.remote_vni;
854 }
855
856 if (tb[NDA_IFINDEX]) {
857 struct net_device *tdev;
858
859 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
860 return -EINVAL;
861 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
862 tdev = __dev_get_by_index(net, *ifindex);
863 if (!tdev)
864 return -EADDRNOTAVAIL;
865 } else {
866 *ifindex = 0;
867 }
868
869 return 0;
870 }
871
872 /* Add static entry (via netlink) */
873 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
874 struct net_device *dev,
875 const unsigned char *addr, u16 vid, u16 flags)
876 {
877 struct vxlan_dev *vxlan = netdev_priv(dev);
878 /* struct net *net = dev_net(vxlan->dev); */
879 union vxlan_addr ip;
880 __be16 port;
881 __be32 src_vni, vni;
882 u32 ifindex;
883 int err;
884
885 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
886 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
887 ndm->ndm_state);
888 return -EINVAL;
889 }
890
891 if (tb[NDA_DST] == NULL)
892 return -EINVAL;
893
894 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
895 if (err)
896 return err;
897
898 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
899 return -EAFNOSUPPORT;
900
901 spin_lock_bh(&vxlan->hash_lock);
902 err = vxlan_fdb_update(vxlan, addr, &ip, ndm->ndm_state, flags,
903 port, src_vni, vni, ifindex, ndm->ndm_flags);
904 spin_unlock_bh(&vxlan->hash_lock);
905
906 return err;
907 }
908
909 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
910 const unsigned char *addr, union vxlan_addr ip,
911 __be16 port, __be32 src_vni, __be32 vni,
912 u32 ifindex, u16 vid)
913 {
914 struct vxlan_fdb *f;
915 struct vxlan_rdst *rd = NULL;
916 int err = -ENOENT;
917
918 f = vxlan_find_mac(vxlan, addr, src_vni);
919 if (!f)
920 return err;
921
922 if (!vxlan_addr_any(&ip)) {
923 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
924 if (!rd)
925 goto out;
926 }
927
928 /* remove a destination if it's not the only one on the list,
929 * otherwise destroy the fdb entry
930 */
931 if (rd && !list_is_singular(&f->remotes)) {
932 vxlan_fdb_dst_destroy(vxlan, f, rd);
933 goto out;
934 }
935
936 vxlan_fdb_destroy(vxlan, f, true);
937
938 out:
939 return 0;
940 }
941
942 /* Delete entry (via netlink) */
943 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
944 struct net_device *dev,
945 const unsigned char *addr, u16 vid)
946 {
947 struct vxlan_dev *vxlan = netdev_priv(dev);
948 union vxlan_addr ip;
949 __be32 src_vni, vni;
950 __be16 port;
951 u32 ifindex;
952 int err;
953
954 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
955 if (err)
956 return err;
957
958 spin_lock_bh(&vxlan->hash_lock);
959 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
960 vid);
961 spin_unlock_bh(&vxlan->hash_lock);
962
963 return err;
964 }
965
966 /* Dump forwarding table */
967 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
968 struct net_device *dev,
969 struct net_device *filter_dev, int *idx)
970 {
971 struct vxlan_dev *vxlan = netdev_priv(dev);
972 unsigned int h;
973 int err = 0;
974
975 for (h = 0; h < FDB_HASH_SIZE; ++h) {
976 struct vxlan_fdb *f;
977
978 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
979 struct vxlan_rdst *rd;
980
981 list_for_each_entry_rcu(rd, &f->remotes, list) {
982 if (*idx < cb->args[2])
983 goto skip;
984
985 err = vxlan_fdb_info(skb, vxlan, f,
986 NETLINK_CB(cb->skb).portid,
987 cb->nlh->nlmsg_seq,
988 RTM_NEWNEIGH,
989 NLM_F_MULTI, rd);
990 if (err < 0)
991 goto out;
992 skip:
993 *idx += 1;
994 }
995 }
996 }
997 out:
998 return err;
999 }
1000
1001 /* Watch incoming packets to learn mapping between Ethernet address
1002 * and Tunnel endpoint.
1003 * Return true if packet is bogus and should be dropped.
1004 */
1005 static bool vxlan_snoop(struct net_device *dev,
1006 union vxlan_addr *src_ip, const u8 *src_mac,
1007 u32 src_ifindex, __be32 vni)
1008 {
1009 struct vxlan_dev *vxlan = netdev_priv(dev);
1010 struct vxlan_fdb *f;
1011 u32 ifindex = 0;
1012
1013 #if IS_ENABLED(CONFIG_IPV6)
1014 if (src_ip->sa.sa_family == AF_INET6 &&
1015 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1016 ifindex = src_ifindex;
1017 #endif
1018
1019 f = vxlan_find_mac(vxlan, src_mac, vni);
1020 if (likely(f)) {
1021 struct vxlan_rdst *rdst = first_remote_rcu(f);
1022
1023 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1024 rdst->remote_ifindex == ifindex))
1025 return false;
1026
1027 /* Don't migrate static entries, drop packets */
1028 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1029 return true;
1030
1031 if (net_ratelimit())
1032 netdev_info(dev,
1033 "%pM migrated from %pIS to %pIS\n",
1034 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1035
1036 rdst->remote_ip = *src_ip;
1037 f->updated = jiffies;
1038 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
1039 } else {
1040 /* learned new entry */
1041 spin_lock(&vxlan->hash_lock);
1042
1043 /* close off race between vxlan_flush and incoming packets */
1044 if (netif_running(dev))
1045 vxlan_fdb_update(vxlan, src_mac, src_ip,
1046 NUD_REACHABLE,
1047 NLM_F_EXCL|NLM_F_CREATE,
1048 vxlan->cfg.dst_port,
1049 vni,
1050 vxlan->default_dst.remote_vni,
1051 ifindex, NTF_SELF);
1052 spin_unlock(&vxlan->hash_lock);
1053 }
1054
1055 return false;
1056 }
1057
1058 /* See if multicast group is already in use by other ID */
1059 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1060 {
1061 struct vxlan_dev *vxlan;
1062 struct vxlan_sock *sock4;
1063 #if IS_ENABLED(CONFIG_IPV6)
1064 struct vxlan_sock *sock6;
1065 #endif
1066 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1067
1068 sock4 = rtnl_dereference(dev->vn4_sock);
1069
1070 /* The vxlan_sock is only used by dev, leaving group has
1071 * no effect on other vxlan devices.
1072 */
1073 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1074 return false;
1075 #if IS_ENABLED(CONFIG_IPV6)
1076 sock6 = rtnl_dereference(dev->vn6_sock);
1077 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1078 return false;
1079 #endif
1080
1081 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1082 if (!netif_running(vxlan->dev) || vxlan == dev)
1083 continue;
1084
1085 if (family == AF_INET &&
1086 rtnl_dereference(vxlan->vn4_sock) != sock4)
1087 continue;
1088 #if IS_ENABLED(CONFIG_IPV6)
1089 if (family == AF_INET6 &&
1090 rtnl_dereference(vxlan->vn6_sock) != sock6)
1091 continue;
1092 #endif
1093
1094 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1095 &dev->default_dst.remote_ip))
1096 continue;
1097
1098 if (vxlan->default_dst.remote_ifindex !=
1099 dev->default_dst.remote_ifindex)
1100 continue;
1101
1102 return true;
1103 }
1104
1105 return false;
1106 }
1107
1108 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1109 {
1110 struct vxlan_net *vn;
1111
1112 if (!vs)
1113 return false;
1114 if (!refcount_dec_and_test(&vs->refcnt))
1115 return false;
1116
1117 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1118 spin_lock(&vn->sock_lock);
1119 hlist_del_rcu(&vs->hlist);
1120 udp_tunnel_notify_del_rx_port(vs->sock,
1121 (vs->flags & VXLAN_F_GPE) ?
1122 UDP_TUNNEL_TYPE_VXLAN_GPE :
1123 UDP_TUNNEL_TYPE_VXLAN);
1124 spin_unlock(&vn->sock_lock);
1125
1126 return true;
1127 }
1128
1129 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1130 {
1131 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1132 #if IS_ENABLED(CONFIG_IPV6)
1133 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1134
1135 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1136 #endif
1137
1138 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1139 synchronize_net();
1140
1141 vxlan_vs_del_dev(vxlan);
1142
1143 if (__vxlan_sock_release_prep(sock4)) {
1144 udp_tunnel_sock_release(sock4->sock);
1145 kfree(sock4);
1146 }
1147
1148 #if IS_ENABLED(CONFIG_IPV6)
1149 if (__vxlan_sock_release_prep(sock6)) {
1150 udp_tunnel_sock_release(sock6->sock);
1151 kfree(sock6);
1152 }
1153 #endif
1154 }
1155
1156 /* Update multicast group membership when first VNI on
1157 * multicast address is brought up
1158 */
1159 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1160 {
1161 struct sock *sk;
1162 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1163 int ifindex = vxlan->default_dst.remote_ifindex;
1164 int ret = -EINVAL;
1165
1166 if (ip->sa.sa_family == AF_INET) {
1167 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1168 struct ip_mreqn mreq = {
1169 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1170 .imr_ifindex = ifindex,
1171 };
1172
1173 sk = sock4->sock->sk;
1174 lock_sock(sk);
1175 ret = ip_mc_join_group(sk, &mreq);
1176 release_sock(sk);
1177 #if IS_ENABLED(CONFIG_IPV6)
1178 } else {
1179 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1180
1181 sk = sock6->sock->sk;
1182 lock_sock(sk);
1183 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1184 &ip->sin6.sin6_addr);
1185 release_sock(sk);
1186 #endif
1187 }
1188
1189 return ret;
1190 }
1191
1192 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1193 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1194 {
1195 struct sock *sk;
1196 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1197 int ifindex = vxlan->default_dst.remote_ifindex;
1198 int ret = -EINVAL;
1199
1200 if (ip->sa.sa_family == AF_INET) {
1201 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1202 struct ip_mreqn mreq = {
1203 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1204 .imr_ifindex = ifindex,
1205 };
1206
1207 sk = sock4->sock->sk;
1208 lock_sock(sk);
1209 ret = ip_mc_leave_group(sk, &mreq);
1210 release_sock(sk);
1211 #if IS_ENABLED(CONFIG_IPV6)
1212 } else {
1213 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1214
1215 sk = sock6->sock->sk;
1216 lock_sock(sk);
1217 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1218 &ip->sin6.sin6_addr);
1219 release_sock(sk);
1220 #endif
1221 }
1222
1223 return ret;
1224 }
1225
1226 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1227 struct sk_buff *skb, u32 vxflags)
1228 {
1229 size_t start, offset;
1230
1231 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1232 goto out;
1233
1234 start = vxlan_rco_start(unparsed->vx_vni);
1235 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1236
1237 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1238 return false;
1239
1240 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1241 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1242 out:
1243 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1244 unparsed->vx_vni &= VXLAN_VNI_MASK;
1245 return true;
1246 }
1247
1248 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1249 struct sk_buff *skb, u32 vxflags,
1250 struct vxlan_metadata *md)
1251 {
1252 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1253 struct metadata_dst *tun_dst;
1254
1255 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1256 goto out;
1257
1258 md->gbp = ntohs(gbp->policy_id);
1259
1260 tun_dst = (struct metadata_dst *)skb_dst(skb);
1261 if (tun_dst) {
1262 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1263 tun_dst->u.tun_info.options_len = sizeof(*md);
1264 }
1265 if (gbp->dont_learn)
1266 md->gbp |= VXLAN_GBP_DONT_LEARN;
1267
1268 if (gbp->policy_applied)
1269 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1270
1271 /* In flow-based mode, GBP is carried in dst_metadata */
1272 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1273 skb->mark = md->gbp;
1274 out:
1275 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1276 }
1277
1278 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1279 __be16 *protocol,
1280 struct sk_buff *skb, u32 vxflags)
1281 {
1282 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1283
1284 /* Need to have Next Protocol set for interfaces in GPE mode. */
1285 if (!gpe->np_applied)
1286 return false;
1287 /* "The initial version is 0. If a receiver does not support the
1288 * version indicated it MUST drop the packet.
1289 */
1290 if (gpe->version != 0)
1291 return false;
1292 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1293 * processing MUST occur." However, we don't implement OAM
1294 * processing, thus drop the packet.
1295 */
1296 if (gpe->oam_flag)
1297 return false;
1298
1299 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1300 if (!*protocol)
1301 return false;
1302
1303 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1304 return true;
1305 }
1306
1307 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1308 struct vxlan_sock *vs,
1309 struct sk_buff *skb, __be32 vni)
1310 {
1311 union vxlan_addr saddr;
1312 u32 ifindex = skb->dev->ifindex;
1313
1314 skb_reset_mac_header(skb);
1315 skb->protocol = eth_type_trans(skb, vxlan->dev);
1316 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1317
1318 /* Ignore packet loops (and multicast echo) */
1319 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1320 return false;
1321
1322 /* Get address from the outer IP header */
1323 if (vxlan_get_sk_family(vs) == AF_INET) {
1324 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1325 saddr.sa.sa_family = AF_INET;
1326 #if IS_ENABLED(CONFIG_IPV6)
1327 } else {
1328 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1329 saddr.sa.sa_family = AF_INET6;
1330 #endif
1331 }
1332
1333 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1334 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1335 return false;
1336
1337 return true;
1338 }
1339
1340 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1341 struct sk_buff *skb)
1342 {
1343 int err = 0;
1344
1345 if (vxlan_get_sk_family(vs) == AF_INET)
1346 err = IP_ECN_decapsulate(oiph, skb);
1347 #if IS_ENABLED(CONFIG_IPV6)
1348 else
1349 err = IP6_ECN_decapsulate(oiph, skb);
1350 #endif
1351
1352 if (unlikely(err) && log_ecn_error) {
1353 if (vxlan_get_sk_family(vs) == AF_INET)
1354 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1355 &((struct iphdr *)oiph)->saddr,
1356 ((struct iphdr *)oiph)->tos);
1357 else
1358 net_info_ratelimited("non-ECT from %pI6\n",
1359 &((struct ipv6hdr *)oiph)->saddr);
1360 }
1361 return err <= 1;
1362 }
1363
1364 /* Callback from net/ipv4/udp.c to receive packets */
1365 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1366 {
1367 struct pcpu_sw_netstats *stats;
1368 struct vxlan_dev *vxlan;
1369 struct vxlan_sock *vs;
1370 struct vxlanhdr unparsed;
1371 struct vxlan_metadata _md;
1372 struct vxlan_metadata *md = &_md;
1373 __be16 protocol = htons(ETH_P_TEB);
1374 bool raw_proto = false;
1375 void *oiph;
1376 __be32 vni = 0;
1377
1378 /* Need UDP and VXLAN header to be present */
1379 if (!pskb_may_pull(skb, VXLAN_HLEN))
1380 goto drop;
1381
1382 unparsed = *vxlan_hdr(skb);
1383 /* VNI flag always required to be set */
1384 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1385 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1386 ntohl(vxlan_hdr(skb)->vx_flags),
1387 ntohl(vxlan_hdr(skb)->vx_vni));
1388 /* Return non vxlan pkt */
1389 goto drop;
1390 }
1391 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1392 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1393
1394 vs = rcu_dereference_sk_user_data(sk);
1395 if (!vs)
1396 goto drop;
1397
1398 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1399
1400 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1401 if (!vxlan)
1402 goto drop;
1403
1404 /* For backwards compatibility, only allow reserved fields to be
1405 * used by VXLAN extensions if explicitly requested.
1406 */
1407 if (vs->flags & VXLAN_F_GPE) {
1408 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1409 goto drop;
1410 raw_proto = true;
1411 }
1412
1413 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1414 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1415 goto drop;
1416
1417 if (vxlan_collect_metadata(vs)) {
1418 struct metadata_dst *tun_dst;
1419
1420 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1421 key32_to_tunnel_id(vni), sizeof(*md));
1422
1423 if (!tun_dst)
1424 goto drop;
1425
1426 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1427
1428 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1429 } else {
1430 memset(md, 0, sizeof(*md));
1431 }
1432
1433 if (vs->flags & VXLAN_F_REMCSUM_RX)
1434 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1435 goto drop;
1436 if (vs->flags & VXLAN_F_GBP)
1437 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1438 /* Note that GBP and GPE can never be active together. This is
1439 * ensured in vxlan_dev_configure.
1440 */
1441
1442 if (unparsed.vx_flags || unparsed.vx_vni) {
1443 /* If there are any unprocessed flags remaining treat
1444 * this as a malformed packet. This behavior diverges from
1445 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1446 * in reserved fields are to be ignored. The approach here
1447 * maintains compatibility with previous stack code, and also
1448 * is more robust and provides a little more security in
1449 * adding extensions to VXLAN.
1450 */
1451 goto drop;
1452 }
1453
1454 if (!raw_proto) {
1455 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1456 goto drop;
1457 } else {
1458 skb_reset_mac_header(skb);
1459 skb->dev = vxlan->dev;
1460 skb->pkt_type = PACKET_HOST;
1461 }
1462
1463 oiph = skb_network_header(skb);
1464 skb_reset_network_header(skb);
1465
1466 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1467 ++vxlan->dev->stats.rx_frame_errors;
1468 ++vxlan->dev->stats.rx_errors;
1469 goto drop;
1470 }
1471
1472 rcu_read_lock();
1473
1474 if (unlikely(!(vxlan->dev->flags & IFF_UP))) {
1475 rcu_read_unlock();
1476 atomic_long_inc(&vxlan->dev->rx_dropped);
1477 goto drop;
1478 }
1479
1480 stats = this_cpu_ptr(vxlan->dev->tstats);
1481 u64_stats_update_begin(&stats->syncp);
1482 stats->rx_packets++;
1483 stats->rx_bytes += skb->len;
1484 u64_stats_update_end(&stats->syncp);
1485
1486 gro_cells_receive(&vxlan->gro_cells, skb);
1487
1488 rcu_read_unlock();
1489
1490 return 0;
1491
1492 drop:
1493 /* Consume bad packet */
1494 kfree_skb(skb);
1495 return 0;
1496 }
1497
1498 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1499 {
1500 struct vxlan_dev *vxlan = netdev_priv(dev);
1501 struct arphdr *parp;
1502 u8 *arpptr, *sha;
1503 __be32 sip, tip;
1504 struct neighbour *n;
1505
1506 if (dev->flags & IFF_NOARP)
1507 goto out;
1508
1509 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1510 dev->stats.tx_dropped++;
1511 goto out;
1512 }
1513 parp = arp_hdr(skb);
1514
1515 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1516 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1517 parp->ar_pro != htons(ETH_P_IP) ||
1518 parp->ar_op != htons(ARPOP_REQUEST) ||
1519 parp->ar_hln != dev->addr_len ||
1520 parp->ar_pln != 4)
1521 goto out;
1522 arpptr = (u8 *)parp + sizeof(struct arphdr);
1523 sha = arpptr;
1524 arpptr += dev->addr_len; /* sha */
1525 memcpy(&sip, arpptr, sizeof(sip));
1526 arpptr += sizeof(sip);
1527 arpptr += dev->addr_len; /* tha */
1528 memcpy(&tip, arpptr, sizeof(tip));
1529
1530 if (ipv4_is_loopback(tip) ||
1531 ipv4_is_multicast(tip))
1532 goto out;
1533
1534 n = neigh_lookup(&arp_tbl, &tip, dev);
1535
1536 if (n) {
1537 struct vxlan_fdb *f;
1538 struct sk_buff *reply;
1539
1540 if (!(n->nud_state & NUD_CONNECTED)) {
1541 neigh_release(n);
1542 goto out;
1543 }
1544
1545 f = vxlan_find_mac(vxlan, n->ha, vni);
1546 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1547 /* bridge-local neighbor */
1548 neigh_release(n);
1549 goto out;
1550 }
1551
1552 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1553 n->ha, sha);
1554
1555 neigh_release(n);
1556
1557 if (reply == NULL)
1558 goto out;
1559
1560 skb_reset_mac_header(reply);
1561 __skb_pull(reply, skb_network_offset(reply));
1562 reply->ip_summed = CHECKSUM_UNNECESSARY;
1563 reply->pkt_type = PACKET_HOST;
1564
1565 if (netif_rx_ni(reply) == NET_RX_DROP)
1566 dev->stats.rx_dropped++;
1567 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1568 union vxlan_addr ipa = {
1569 .sin.sin_addr.s_addr = tip,
1570 .sin.sin_family = AF_INET,
1571 };
1572
1573 vxlan_ip_miss(dev, &ipa);
1574 }
1575 out:
1576 consume_skb(skb);
1577 return NETDEV_TX_OK;
1578 }
1579
1580 #if IS_ENABLED(CONFIG_IPV6)
1581 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1582 struct neighbour *n, bool isrouter)
1583 {
1584 struct net_device *dev = request->dev;
1585 struct sk_buff *reply;
1586 struct nd_msg *ns, *na;
1587 struct ipv6hdr *pip6;
1588 u8 *daddr;
1589 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1590 int ns_olen;
1591 int i, len;
1592
1593 if (dev == NULL || !pskb_may_pull(request, request->len))
1594 return NULL;
1595
1596 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1597 sizeof(*na) + na_olen + dev->needed_tailroom;
1598 reply = alloc_skb(len, GFP_ATOMIC);
1599 if (reply == NULL)
1600 return NULL;
1601
1602 reply->protocol = htons(ETH_P_IPV6);
1603 reply->dev = dev;
1604 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1605 skb_push(reply, sizeof(struct ethhdr));
1606 skb_reset_mac_header(reply);
1607
1608 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1609
1610 daddr = eth_hdr(request)->h_source;
1611 ns_olen = request->len - skb_network_offset(request) -
1612 sizeof(struct ipv6hdr) - sizeof(*ns);
1613 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1614 if (!ns->opt[i + 1]) {
1615 kfree_skb(reply);
1616 return NULL;
1617 }
1618 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1619 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1620 break;
1621 }
1622 }
1623
1624 /* Ethernet header */
1625 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1626 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1627 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1628 reply->protocol = htons(ETH_P_IPV6);
1629
1630 skb_pull(reply, sizeof(struct ethhdr));
1631 skb_reset_network_header(reply);
1632 skb_put(reply, sizeof(struct ipv6hdr));
1633
1634 /* IPv6 header */
1635
1636 pip6 = ipv6_hdr(reply);
1637 memset(pip6, 0, sizeof(struct ipv6hdr));
1638 pip6->version = 6;
1639 pip6->priority = ipv6_hdr(request)->priority;
1640 pip6->nexthdr = IPPROTO_ICMPV6;
1641 pip6->hop_limit = 255;
1642 pip6->daddr = ipv6_hdr(request)->saddr;
1643 pip6->saddr = *(struct in6_addr *)n->primary_key;
1644
1645 skb_pull(reply, sizeof(struct ipv6hdr));
1646 skb_reset_transport_header(reply);
1647
1648 /* Neighbor Advertisement */
1649 na = skb_put_zero(reply, sizeof(*na) + na_olen);
1650 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1651 na->icmph.icmp6_router = isrouter;
1652 na->icmph.icmp6_override = 1;
1653 na->icmph.icmp6_solicited = 1;
1654 na->target = ns->target;
1655 ether_addr_copy(&na->opt[2], n->ha);
1656 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1657 na->opt[1] = na_olen >> 3;
1658
1659 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1660 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1661 csum_partial(na, sizeof(*na)+na_olen, 0));
1662
1663 pip6->payload_len = htons(sizeof(*na)+na_olen);
1664
1665 skb_push(reply, sizeof(struct ipv6hdr));
1666
1667 reply->ip_summed = CHECKSUM_UNNECESSARY;
1668
1669 return reply;
1670 }
1671
1672 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1673 {
1674 struct vxlan_dev *vxlan = netdev_priv(dev);
1675 const struct in6_addr *daddr;
1676 const struct ipv6hdr *iphdr;
1677 struct inet6_dev *in6_dev;
1678 struct neighbour *n;
1679 struct nd_msg *msg;
1680
1681 in6_dev = __in6_dev_get(dev);
1682 if (!in6_dev)
1683 goto out;
1684
1685 iphdr = ipv6_hdr(skb);
1686 daddr = &iphdr->daddr;
1687 msg = (struct nd_msg *)(iphdr + 1);
1688
1689 if (ipv6_addr_loopback(daddr) ||
1690 ipv6_addr_is_multicast(&msg->target))
1691 goto out;
1692
1693 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1694
1695 if (n) {
1696 struct vxlan_fdb *f;
1697 struct sk_buff *reply;
1698
1699 if (!(n->nud_state & NUD_CONNECTED)) {
1700 neigh_release(n);
1701 goto out;
1702 }
1703
1704 f = vxlan_find_mac(vxlan, n->ha, vni);
1705 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1706 /* bridge-local neighbor */
1707 neigh_release(n);
1708 goto out;
1709 }
1710
1711 reply = vxlan_na_create(skb, n,
1712 !!(f ? f->flags & NTF_ROUTER : 0));
1713
1714 neigh_release(n);
1715
1716 if (reply == NULL)
1717 goto out;
1718
1719 if (netif_rx_ni(reply) == NET_RX_DROP)
1720 dev->stats.rx_dropped++;
1721
1722 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1723 union vxlan_addr ipa = {
1724 .sin6.sin6_addr = msg->target,
1725 .sin6.sin6_family = AF_INET6,
1726 };
1727
1728 vxlan_ip_miss(dev, &ipa);
1729 }
1730
1731 out:
1732 consume_skb(skb);
1733 return NETDEV_TX_OK;
1734 }
1735 #endif
1736
1737 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1738 {
1739 struct vxlan_dev *vxlan = netdev_priv(dev);
1740 struct neighbour *n;
1741
1742 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1743 return false;
1744
1745 n = NULL;
1746 switch (ntohs(eth_hdr(skb)->h_proto)) {
1747 case ETH_P_IP:
1748 {
1749 struct iphdr *pip;
1750
1751 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1752 return false;
1753 pip = ip_hdr(skb);
1754 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1755 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1756 union vxlan_addr ipa = {
1757 .sin.sin_addr.s_addr = pip->daddr,
1758 .sin.sin_family = AF_INET,
1759 };
1760
1761 vxlan_ip_miss(dev, &ipa);
1762 return false;
1763 }
1764
1765 break;
1766 }
1767 #if IS_ENABLED(CONFIG_IPV6)
1768 case ETH_P_IPV6:
1769 {
1770 struct ipv6hdr *pip6;
1771
1772 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1773 return false;
1774 pip6 = ipv6_hdr(skb);
1775 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1776 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1777 union vxlan_addr ipa = {
1778 .sin6.sin6_addr = pip6->daddr,
1779 .sin6.sin6_family = AF_INET6,
1780 };
1781
1782 vxlan_ip_miss(dev, &ipa);
1783 return false;
1784 }
1785
1786 break;
1787 }
1788 #endif
1789 default:
1790 return false;
1791 }
1792
1793 if (n) {
1794 bool diff;
1795
1796 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1797 if (diff) {
1798 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1799 dev->addr_len);
1800 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1801 }
1802 neigh_release(n);
1803 return diff;
1804 }
1805
1806 return false;
1807 }
1808
1809 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1810 struct vxlan_metadata *md)
1811 {
1812 struct vxlanhdr_gbp *gbp;
1813
1814 if (!md->gbp)
1815 return;
1816
1817 gbp = (struct vxlanhdr_gbp *)vxh;
1818 vxh->vx_flags |= VXLAN_HF_GBP;
1819
1820 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1821 gbp->dont_learn = 1;
1822
1823 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1824 gbp->policy_applied = 1;
1825
1826 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1827 }
1828
1829 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1830 __be16 protocol)
1831 {
1832 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1833
1834 gpe->np_applied = 1;
1835 gpe->next_protocol = tun_p_from_eth_p(protocol);
1836 if (!gpe->next_protocol)
1837 return -EPFNOSUPPORT;
1838 return 0;
1839 }
1840
1841 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1842 int iphdr_len, __be32 vni,
1843 struct vxlan_metadata *md, u32 vxflags,
1844 bool udp_sum)
1845 {
1846 struct vxlanhdr *vxh;
1847 int min_headroom;
1848 int err;
1849 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1850 __be16 inner_protocol = htons(ETH_P_TEB);
1851
1852 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1853 skb->ip_summed == CHECKSUM_PARTIAL) {
1854 int csum_start = skb_checksum_start_offset(skb);
1855
1856 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1857 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1858 (skb->csum_offset == offsetof(struct udphdr, check) ||
1859 skb->csum_offset == offsetof(struct tcphdr, check)))
1860 type |= SKB_GSO_TUNNEL_REMCSUM;
1861 }
1862
1863 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1864 + VXLAN_HLEN + iphdr_len;
1865
1866 /* Need space for new headers (invalidates iph ptr) */
1867 err = skb_cow_head(skb, min_headroom);
1868 if (unlikely(err))
1869 return err;
1870
1871 err = iptunnel_handle_offloads(skb, type);
1872 if (err)
1873 return err;
1874
1875 vxh = __skb_push(skb, sizeof(*vxh));
1876 vxh->vx_flags = VXLAN_HF_VNI;
1877 vxh->vx_vni = vxlan_vni_field(vni);
1878
1879 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1880 unsigned int start;
1881
1882 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1883 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1884 vxh->vx_flags |= VXLAN_HF_RCO;
1885
1886 if (!skb_is_gso(skb)) {
1887 skb->ip_summed = CHECKSUM_NONE;
1888 skb->encapsulation = 0;
1889 }
1890 }
1891
1892 if (vxflags & VXLAN_F_GBP)
1893 vxlan_build_gbp_hdr(vxh, vxflags, md);
1894 if (vxflags & VXLAN_F_GPE) {
1895 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
1896 if (err < 0)
1897 return err;
1898 inner_protocol = skb->protocol;
1899 }
1900
1901 skb_set_inner_protocol(skb, inner_protocol);
1902 return 0;
1903 }
1904
1905 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
1906 struct vxlan_sock *sock4,
1907 struct sk_buff *skb, int oif, u8 tos,
1908 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
1909 struct dst_cache *dst_cache,
1910 const struct ip_tunnel_info *info)
1911 {
1912 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1913 struct rtable *rt = NULL;
1914 struct flowi4 fl4;
1915
1916 if (!sock4)
1917 return ERR_PTR(-EIO);
1918
1919 if (tos && !info)
1920 use_cache = false;
1921 if (use_cache) {
1922 rt = dst_cache_get_ip4(dst_cache, saddr);
1923 if (rt)
1924 return rt;
1925 }
1926
1927 memset(&fl4, 0, sizeof(fl4));
1928 fl4.flowi4_oif = oif;
1929 fl4.flowi4_tos = RT_TOS(tos);
1930 fl4.flowi4_mark = skb->mark;
1931 fl4.flowi4_proto = IPPROTO_UDP;
1932 fl4.daddr = daddr;
1933 fl4.saddr = *saddr;
1934 fl4.fl4_dport = dport;
1935 fl4.fl4_sport = sport;
1936
1937 rt = ip_route_output_key(vxlan->net, &fl4);
1938 if (likely(!IS_ERR(rt))) {
1939 if (rt->dst.dev == dev) {
1940 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
1941 ip_rt_put(rt);
1942 return ERR_PTR(-ELOOP);
1943 }
1944
1945 *saddr = fl4.saddr;
1946 if (use_cache)
1947 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
1948 } else {
1949 netdev_dbg(dev, "no route to %pI4\n", &daddr);
1950 return ERR_PTR(-ENETUNREACH);
1951 }
1952 return rt;
1953 }
1954
1955 #if IS_ENABLED(CONFIG_IPV6)
1956 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1957 struct net_device *dev,
1958 struct vxlan_sock *sock6,
1959 struct sk_buff *skb, int oif, u8 tos,
1960 __be32 label,
1961 const struct in6_addr *daddr,
1962 struct in6_addr *saddr,
1963 __be16 dport, __be16 sport,
1964 struct dst_cache *dst_cache,
1965 const struct ip_tunnel_info *info)
1966 {
1967 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
1968 struct dst_entry *ndst;
1969 struct flowi6 fl6;
1970
1971 if (!sock6)
1972 return ERR_PTR(-EIO);
1973
1974 if (tos && !info)
1975 use_cache = false;
1976 if (use_cache) {
1977 ndst = dst_cache_get_ip6(dst_cache, saddr);
1978 if (ndst)
1979 return ndst;
1980 }
1981
1982 memset(&fl6, 0, sizeof(fl6));
1983 fl6.flowi6_oif = oif;
1984 fl6.daddr = *daddr;
1985 fl6.saddr = *saddr;
1986 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
1987 fl6.flowi6_mark = skb->mark;
1988 fl6.flowi6_proto = IPPROTO_UDP;
1989 fl6.fl6_dport = dport;
1990 fl6.fl6_sport = sport;
1991
1992 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
1993 &fl6, NULL);
1994 if (unlikely(IS_ERR(ndst))) {
1995 netdev_dbg(dev, "no route to %pI6\n", daddr);
1996 return ERR_PTR(-ENETUNREACH);
1997 }
1998
1999 if (unlikely(ndst->dev == dev)) {
2000 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2001 dst_release(ndst);
2002 return ERR_PTR(-ELOOP);
2003 }
2004
2005 *saddr = fl6.saddr;
2006 if (use_cache)
2007 dst_cache_set_ip6(dst_cache, ndst, saddr);
2008 return ndst;
2009 }
2010 #endif
2011
2012 /* Bypass encapsulation if the destination is local */
2013 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2014 struct vxlan_dev *dst_vxlan, __be32 vni)
2015 {
2016 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2017 union vxlan_addr loopback;
2018 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2019 struct net_device *dev;
2020 int len = skb->len;
2021
2022 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2023 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2024 skb->pkt_type = PACKET_HOST;
2025 skb->encapsulation = 0;
2026 skb->dev = dst_vxlan->dev;
2027 __skb_pull(skb, skb_network_offset(skb));
2028
2029 if (remote_ip->sa.sa_family == AF_INET) {
2030 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2031 loopback.sa.sa_family = AF_INET;
2032 #if IS_ENABLED(CONFIG_IPV6)
2033 } else {
2034 loopback.sin6.sin6_addr = in6addr_loopback;
2035 loopback.sa.sa_family = AF_INET6;
2036 #endif
2037 }
2038
2039 rcu_read_lock();
2040 dev = skb->dev;
2041 if (unlikely(!(dev->flags & IFF_UP))) {
2042 kfree_skb(skb);
2043 goto drop;
2044 }
2045
2046 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
2047 vxlan_snoop(dev, &loopback, eth_hdr(skb)->h_source, 0, vni);
2048
2049 u64_stats_update_begin(&tx_stats->syncp);
2050 tx_stats->tx_packets++;
2051 tx_stats->tx_bytes += len;
2052 u64_stats_update_end(&tx_stats->syncp);
2053
2054 if (netif_rx(skb) == NET_RX_SUCCESS) {
2055 u64_stats_update_begin(&rx_stats->syncp);
2056 rx_stats->rx_packets++;
2057 rx_stats->rx_bytes += len;
2058 u64_stats_update_end(&rx_stats->syncp);
2059 } else {
2060 drop:
2061 dev->stats.rx_dropped++;
2062 }
2063 rcu_read_unlock();
2064 }
2065
2066 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2067 struct vxlan_dev *vxlan,
2068 union vxlan_addr *daddr,
2069 __be16 dst_port, int dst_ifindex, __be32 vni,
2070 struct dst_entry *dst,
2071 u32 rt_flags)
2072 {
2073 #if IS_ENABLED(CONFIG_IPV6)
2074 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2075 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2076 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2077 */
2078 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2079 #endif
2080 /* Bypass encapsulation if the destination is local */
2081 if (rt_flags & RTCF_LOCAL &&
2082 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2083 struct vxlan_dev *dst_vxlan;
2084
2085 dst_release(dst);
2086 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2087 daddr->sa.sa_family, dst_port,
2088 vxlan->cfg.flags);
2089 if (!dst_vxlan) {
2090 dev->stats.tx_errors++;
2091 kfree_skb(skb);
2092
2093 return -ENOENT;
2094 }
2095 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2096 return 1;
2097 }
2098
2099 return 0;
2100 }
2101
2102 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2103 __be32 default_vni, struct vxlan_rdst *rdst,
2104 bool did_rsc)
2105 {
2106 struct dst_cache *dst_cache;
2107 struct ip_tunnel_info *info;
2108 struct vxlan_dev *vxlan = netdev_priv(dev);
2109 const struct iphdr *old_iph = ip_hdr(skb);
2110 union vxlan_addr *dst;
2111 union vxlan_addr remote_ip, local_ip;
2112 struct vxlan_metadata _md;
2113 struct vxlan_metadata *md = &_md;
2114 __be16 src_port = 0, dst_port;
2115 struct dst_entry *ndst = NULL;
2116 __be32 vni, label;
2117 __u8 tos, ttl;
2118 int ifindex;
2119 int err;
2120 u32 flags = vxlan->cfg.flags;
2121 bool udp_sum = false;
2122 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2123
2124 info = skb_tunnel_info(skb);
2125
2126 if (rdst) {
2127 dst = &rdst->remote_ip;
2128 if (vxlan_addr_any(dst)) {
2129 if (did_rsc) {
2130 /* short-circuited back to local bridge */
2131 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
2132 return;
2133 }
2134 goto drop;
2135 }
2136
2137 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2138 vni = (rdst->remote_vni) ? : default_vni;
2139 ifindex = rdst->remote_ifindex;
2140 local_ip = vxlan->cfg.saddr;
2141 dst_cache = &rdst->dst_cache;
2142 md->gbp = skb->mark;
2143 if (flags & VXLAN_F_TTL_INHERIT) {
2144 ttl = ip_tunnel_get_ttl(old_iph, skb);
2145 } else {
2146 ttl = vxlan->cfg.ttl;
2147 if (!ttl && vxlan_addr_multicast(dst))
2148 ttl = 1;
2149 }
2150
2151 tos = vxlan->cfg.tos;
2152 if (tos == 1)
2153 tos = ip_tunnel_get_dsfield(old_iph, skb);
2154
2155 if (dst->sa.sa_family == AF_INET)
2156 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2157 else
2158 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2159 label = vxlan->cfg.label;
2160 } else {
2161 if (!info) {
2162 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2163 dev->name);
2164 goto drop;
2165 }
2166 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2167 if (remote_ip.sa.sa_family == AF_INET) {
2168 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2169 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2170 } else {
2171 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2172 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2173 }
2174 dst = &remote_ip;
2175 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2176 vni = tunnel_id_to_key32(info->key.tun_id);
2177 ifindex = 0;
2178 dst_cache = &info->dst_cache;
2179 if (info->key.tun_flags & TUNNEL_VXLAN_OPT) {
2180 if (info->options_len < sizeof(*md))
2181 goto drop;
2182 md = ip_tunnel_info_opts(info);
2183 }
2184 ttl = info->key.ttl;
2185 tos = info->key.tos;
2186 label = info->key.label;
2187 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2188 }
2189 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2190 vxlan->cfg.port_max, true);
2191
2192 rcu_read_lock();
2193 if (dst->sa.sa_family == AF_INET) {
2194 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2195 struct rtable *rt;
2196 __be16 df = 0;
2197
2198 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2199 dst->sin.sin_addr.s_addr,
2200 &local_ip.sin.sin_addr.s_addr,
2201 dst_port, src_port,
2202 dst_cache, info);
2203 if (IS_ERR(rt)) {
2204 err = PTR_ERR(rt);
2205 goto tx_error;
2206 }
2207
2208 /* Bypass encapsulation if the destination is local */
2209 if (!info) {
2210 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2211 dst_port, ifindex, vni,
2212 &rt->dst, rt->rt_flags);
2213 if (err)
2214 goto out_unlock;
2215 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2216 df = htons(IP_DF);
2217 }
2218
2219 ndst = &rt->dst;
2220 skb_tunnel_check_pmtu(skb, ndst, VXLAN_HEADROOM);
2221
2222 tos = ip_tunnel_ecn_encap(RT_TOS(tos), old_iph, skb);
2223 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2224 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2225 vni, md, flags, udp_sum);
2226 if (err < 0)
2227 goto tx_error;
2228
2229 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2230 dst->sin.sin_addr.s_addr, tos, ttl, df,
2231 src_port, dst_port, xnet, !udp_sum);
2232 #if IS_ENABLED(CONFIG_IPV6)
2233 } else {
2234 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2235
2236 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2237 label, &dst->sin6.sin6_addr,
2238 &local_ip.sin6.sin6_addr,
2239 dst_port, src_port,
2240 dst_cache, info);
2241 if (IS_ERR(ndst)) {
2242 err = PTR_ERR(ndst);
2243 ndst = NULL;
2244 goto tx_error;
2245 }
2246
2247 if (!info) {
2248 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2249
2250 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2251 dst_port, ifindex, vni,
2252 ndst, rt6i_flags);
2253 if (err)
2254 goto out_unlock;
2255 }
2256
2257 skb_tunnel_check_pmtu(skb, ndst, VXLAN6_HEADROOM);
2258
2259 tos = ip_tunnel_ecn_encap(RT_TOS(tos), old_iph, skb);
2260 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2261 skb_scrub_packet(skb, xnet);
2262 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2263 vni, md, flags, udp_sum);
2264 if (err < 0)
2265 goto tx_error;
2266
2267 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2268 &local_ip.sin6.sin6_addr,
2269 &dst->sin6.sin6_addr, tos, ttl,
2270 label, src_port, dst_port, !udp_sum);
2271 #endif
2272 }
2273 out_unlock:
2274 rcu_read_unlock();
2275 return;
2276
2277 drop:
2278 dev->stats.tx_dropped++;
2279 dev_kfree_skb(skb);
2280 return;
2281
2282 tx_error:
2283 rcu_read_unlock();
2284 if (err == -ELOOP)
2285 dev->stats.collisions++;
2286 else if (err == -ENETUNREACH)
2287 dev->stats.tx_carrier_errors++;
2288 dst_release(ndst);
2289 dev->stats.tx_errors++;
2290 kfree_skb(skb);
2291 }
2292
2293 /* Transmit local packets over Vxlan
2294 *
2295 * Outer IP header inherits ECN and DF from inner header.
2296 * Outer UDP destination is the VXLAN assigned port.
2297 * source port is based on hash of flow
2298 */
2299 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2300 {
2301 struct vxlan_dev *vxlan = netdev_priv(dev);
2302 struct vxlan_rdst *rdst, *fdst = NULL;
2303 const struct ip_tunnel_info *info;
2304 bool did_rsc = false;
2305 struct vxlan_fdb *f;
2306 struct ethhdr *eth;
2307 __be32 vni = 0;
2308
2309 info = skb_tunnel_info(skb);
2310
2311 skb_reset_mac_header(skb);
2312
2313 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2314 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2315 info->mode & IP_TUNNEL_INFO_TX) {
2316 vni = tunnel_id_to_key32(info->key.tun_id);
2317 } else {
2318 if (info && info->mode & IP_TUNNEL_INFO_TX)
2319 vxlan_xmit_one(skb, dev, vni, NULL, false);
2320 else
2321 kfree_skb(skb);
2322 return NETDEV_TX_OK;
2323 }
2324 }
2325
2326 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2327 eth = eth_hdr(skb);
2328 if (ntohs(eth->h_proto) == ETH_P_ARP)
2329 return arp_reduce(dev, skb, vni);
2330 #if IS_ENABLED(CONFIG_IPV6)
2331 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2332 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2333 sizeof(struct nd_msg)) &&
2334 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2335 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2336
2337 if (m->icmph.icmp6_code == 0 &&
2338 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2339 return neigh_reduce(dev, skb, vni);
2340 }
2341 #endif
2342 }
2343
2344 eth = eth_hdr(skb);
2345 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2346 did_rsc = false;
2347
2348 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2349 (ntohs(eth->h_proto) == ETH_P_IP ||
2350 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2351 did_rsc = route_shortcircuit(dev, skb);
2352 if (did_rsc)
2353 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2354 }
2355
2356 if (f == NULL) {
2357 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2358 if (f == NULL) {
2359 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2360 !is_multicast_ether_addr(eth->h_dest))
2361 vxlan_fdb_miss(vxlan, eth->h_dest);
2362
2363 dev->stats.tx_dropped++;
2364 kfree_skb(skb);
2365 return NETDEV_TX_OK;
2366 }
2367 }
2368
2369 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2370 struct sk_buff *skb1;
2371
2372 if (!fdst) {
2373 fdst = rdst;
2374 continue;
2375 }
2376 skb1 = skb_clone(skb, GFP_ATOMIC);
2377 if (skb1)
2378 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2379 }
2380
2381 if (fdst)
2382 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2383 else
2384 kfree_skb(skb);
2385 return NETDEV_TX_OK;
2386 }
2387
2388 /* Walk the forwarding table and purge stale entries */
2389 static void vxlan_cleanup(struct timer_list *t)
2390 {
2391 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2392 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2393 unsigned int h;
2394
2395 if (!netif_running(vxlan->dev))
2396 return;
2397
2398 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2399 struct hlist_node *p, *n;
2400
2401 spin_lock_bh(&vxlan->hash_lock);
2402 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2403 struct vxlan_fdb *f
2404 = container_of(p, struct vxlan_fdb, hlist);
2405 unsigned long timeout;
2406
2407 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2408 continue;
2409
2410 if (f->flags & NTF_EXT_LEARNED)
2411 continue;
2412
2413 timeout = f->used + vxlan->cfg.age_interval * HZ;
2414 if (time_before_eq(timeout, jiffies)) {
2415 netdev_dbg(vxlan->dev,
2416 "garbage collect %pM\n",
2417 f->eth_addr);
2418 f->state = NUD_STALE;
2419 vxlan_fdb_destroy(vxlan, f, true);
2420 } else if (time_before(timeout, next_timer))
2421 next_timer = timeout;
2422 }
2423 spin_unlock_bh(&vxlan->hash_lock);
2424 }
2425
2426 mod_timer(&vxlan->age_timer, next_timer);
2427 }
2428
2429 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2430 {
2431 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2432
2433 spin_lock(&vn->sock_lock);
2434 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2435 #if IS_ENABLED(CONFIG_IPV6)
2436 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2437 #endif
2438 spin_unlock(&vn->sock_lock);
2439 }
2440
2441 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2442 struct vxlan_dev_node *node)
2443 {
2444 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2445 __be32 vni = vxlan->default_dst.remote_vni;
2446
2447 node->vxlan = vxlan;
2448 spin_lock(&vn->sock_lock);
2449 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2450 spin_unlock(&vn->sock_lock);
2451 }
2452
2453 /* Setup stats when device is created */
2454 static int vxlan_init(struct net_device *dev)
2455 {
2456 struct vxlan_dev *vxlan = netdev_priv(dev);
2457 int err;
2458
2459 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2460 if (!dev->tstats)
2461 return -ENOMEM;
2462
2463 err = gro_cells_init(&vxlan->gro_cells, dev);
2464 if (err) {
2465 free_percpu(dev->tstats);
2466 return err;
2467 }
2468
2469 return 0;
2470 }
2471
2472 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2473 {
2474 struct vxlan_fdb *f;
2475
2476 spin_lock_bh(&vxlan->hash_lock);
2477 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2478 if (f)
2479 vxlan_fdb_destroy(vxlan, f, true);
2480 spin_unlock_bh(&vxlan->hash_lock);
2481 }
2482
2483 static void vxlan_uninit(struct net_device *dev)
2484 {
2485 struct vxlan_dev *vxlan = netdev_priv(dev);
2486
2487 gro_cells_destroy(&vxlan->gro_cells);
2488
2489 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2490
2491 free_percpu(dev->tstats);
2492 }
2493
2494 /* Start ageing timer and join group when device is brought up */
2495 static int vxlan_open(struct net_device *dev)
2496 {
2497 struct vxlan_dev *vxlan = netdev_priv(dev);
2498 int ret;
2499
2500 ret = vxlan_sock_add(vxlan);
2501 if (ret < 0)
2502 return ret;
2503
2504 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2505 ret = vxlan_igmp_join(vxlan);
2506 if (ret == -EADDRINUSE)
2507 ret = 0;
2508 if (ret) {
2509 vxlan_sock_release(vxlan);
2510 return ret;
2511 }
2512 }
2513
2514 if (vxlan->cfg.age_interval)
2515 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2516
2517 return ret;
2518 }
2519
2520 /* Purge the forwarding table */
2521 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2522 {
2523 unsigned int h;
2524
2525 spin_lock_bh(&vxlan->hash_lock);
2526 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2527 struct hlist_node *p, *n;
2528 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2529 struct vxlan_fdb *f
2530 = container_of(p, struct vxlan_fdb, hlist);
2531 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2532 continue;
2533 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2534 if (!is_zero_ether_addr(f->eth_addr))
2535 vxlan_fdb_destroy(vxlan, f, true);
2536 }
2537 }
2538 spin_unlock_bh(&vxlan->hash_lock);
2539 }
2540
2541 /* Cleanup timer and forwarding table on shutdown */
2542 static int vxlan_stop(struct net_device *dev)
2543 {
2544 struct vxlan_dev *vxlan = netdev_priv(dev);
2545 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2546 int ret = 0;
2547
2548 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2549 !vxlan_group_used(vn, vxlan))
2550 ret = vxlan_igmp_leave(vxlan);
2551
2552 del_timer_sync(&vxlan->age_timer);
2553
2554 vxlan_flush(vxlan, false);
2555 vxlan_sock_release(vxlan);
2556
2557 return ret;
2558 }
2559
2560 /* Stub, nothing needs to be done. */
2561 static void vxlan_set_multicast_list(struct net_device *dev)
2562 {
2563 }
2564
2565 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2566 {
2567 struct vxlan_dev *vxlan = netdev_priv(dev);
2568 struct vxlan_rdst *dst = &vxlan->default_dst;
2569 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2570 dst->remote_ifindex);
2571 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
2572
2573 /* This check is different than dev->max_mtu, because it looks at
2574 * the lowerdev->mtu, rather than the static dev->max_mtu
2575 */
2576 if (lowerdev) {
2577 int max_mtu = lowerdev->mtu -
2578 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2579 if (new_mtu > max_mtu)
2580 return -EINVAL;
2581 }
2582
2583 dev->mtu = new_mtu;
2584 return 0;
2585 }
2586
2587 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2588 {
2589 struct vxlan_dev *vxlan = netdev_priv(dev);
2590 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2591 __be16 sport, dport;
2592
2593 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2594 vxlan->cfg.port_max, true);
2595 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2596
2597 if (ip_tunnel_info_af(info) == AF_INET) {
2598 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2599 struct rtable *rt;
2600
2601 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
2602 info->key.u.ipv4.dst,
2603 &info->key.u.ipv4.src, dport, sport,
2604 &info->dst_cache, info);
2605 if (IS_ERR(rt))
2606 return PTR_ERR(rt);
2607 ip_rt_put(rt);
2608 } else {
2609 #if IS_ENABLED(CONFIG_IPV6)
2610 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2611 struct dst_entry *ndst;
2612
2613 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
2614 info->key.label, &info->key.u.ipv6.dst,
2615 &info->key.u.ipv6.src, dport, sport,
2616 &info->dst_cache, info);
2617 if (IS_ERR(ndst))
2618 return PTR_ERR(ndst);
2619 dst_release(ndst);
2620 #else /* !CONFIG_IPV6 */
2621 return -EPFNOSUPPORT;
2622 #endif
2623 }
2624 info->key.tp_src = sport;
2625 info->key.tp_dst = dport;
2626 return 0;
2627 }
2628
2629 static const struct net_device_ops vxlan_netdev_ether_ops = {
2630 .ndo_init = vxlan_init,
2631 .ndo_uninit = vxlan_uninit,
2632 .ndo_open = vxlan_open,
2633 .ndo_stop = vxlan_stop,
2634 .ndo_start_xmit = vxlan_xmit,
2635 .ndo_get_stats64 = ip_tunnel_get_stats64,
2636 .ndo_set_rx_mode = vxlan_set_multicast_list,
2637 .ndo_change_mtu = vxlan_change_mtu,
2638 .ndo_validate_addr = eth_validate_addr,
2639 .ndo_set_mac_address = eth_mac_addr,
2640 .ndo_fdb_add = vxlan_fdb_add,
2641 .ndo_fdb_del = vxlan_fdb_delete,
2642 .ndo_fdb_dump = vxlan_fdb_dump,
2643 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2644 };
2645
2646 static const struct net_device_ops vxlan_netdev_raw_ops = {
2647 .ndo_init = vxlan_init,
2648 .ndo_uninit = vxlan_uninit,
2649 .ndo_open = vxlan_open,
2650 .ndo_stop = vxlan_stop,
2651 .ndo_start_xmit = vxlan_xmit,
2652 .ndo_get_stats64 = ip_tunnel_get_stats64,
2653 .ndo_change_mtu = vxlan_change_mtu,
2654 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2655 };
2656
2657 /* Info for udev, that this is a virtual tunnel endpoint */
2658 static struct device_type vxlan_type = {
2659 .name = "vxlan",
2660 };
2661
2662 /* Calls the ndo_udp_tunnel_add of the caller in order to
2663 * supply the listening VXLAN udp ports. Callers are expected
2664 * to implement the ndo_udp_tunnel_add.
2665 */
2666 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
2667 {
2668 struct vxlan_sock *vs;
2669 struct net *net = dev_net(dev);
2670 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2671 unsigned int i;
2672
2673 spin_lock(&vn->sock_lock);
2674 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2675 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2676 unsigned short type;
2677
2678 if (vs->flags & VXLAN_F_GPE)
2679 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2680 else
2681 type = UDP_TUNNEL_TYPE_VXLAN;
2682
2683 if (push)
2684 udp_tunnel_push_rx_port(dev, vs->sock, type);
2685 else
2686 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2687 }
2688 }
2689 spin_unlock(&vn->sock_lock);
2690 }
2691
2692 /* Initialize the device structure. */
2693 static void vxlan_setup(struct net_device *dev)
2694 {
2695 struct vxlan_dev *vxlan = netdev_priv(dev);
2696 unsigned int h;
2697
2698 eth_hw_addr_random(dev);
2699 ether_setup(dev);
2700
2701 dev->needs_free_netdev = true;
2702 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2703
2704 dev->features |= NETIF_F_LLTX;
2705 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2706 dev->features |= NETIF_F_RXCSUM;
2707 dev->features |= NETIF_F_GSO_SOFTWARE;
2708
2709 dev->vlan_features = dev->features;
2710 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2711 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2712 netif_keep_dst(dev);
2713 dev->priv_flags |= IFF_NO_QUEUE;
2714
2715 /* MTU range: 68 - 65535 */
2716 dev->min_mtu = ETH_MIN_MTU;
2717 dev->max_mtu = ETH_MAX_MTU;
2718
2719 INIT_LIST_HEAD(&vxlan->next);
2720 spin_lock_init(&vxlan->hash_lock);
2721
2722 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
2723
2724 vxlan->dev = dev;
2725
2726 for (h = 0; h < FDB_HASH_SIZE; ++h)
2727 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2728 }
2729
2730 static void vxlan_ether_setup(struct net_device *dev)
2731 {
2732 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2733 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2734 dev->netdev_ops = &vxlan_netdev_ether_ops;
2735 }
2736
2737 static void vxlan_raw_setup(struct net_device *dev)
2738 {
2739 dev->header_ops = NULL;
2740 dev->type = ARPHRD_NONE;
2741 dev->hard_header_len = 0;
2742 dev->addr_len = 0;
2743 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2744 dev->netdev_ops = &vxlan_netdev_raw_ops;
2745 }
2746
2747 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2748 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2749 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2750 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2751 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2752 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2753 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2754 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2755 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2756 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
2757 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2758 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2759 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2760 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2761 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2762 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2763 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2764 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2765 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
2766 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2767 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2768 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2769 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
2770 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2771 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2772 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
2773 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
2774 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
2775 [IFLA_VXLAN_TTL_INHERIT] = { .type = NLA_FLAG },
2776 };
2777
2778 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2779 struct netlink_ext_ack *extack)
2780 {
2781 if (tb[IFLA_ADDRESS]) {
2782 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2783 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2784 "Provided link layer address is not Ethernet");
2785 return -EINVAL;
2786 }
2787
2788 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2789 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2790 "Provided Ethernet address is not unicast");
2791 return -EADDRNOTAVAIL;
2792 }
2793 }
2794
2795 if (tb[IFLA_MTU]) {
2796 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2797
2798 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2799 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2800 "MTU must be between 68 and 65535");
2801 return -EINVAL;
2802 }
2803 }
2804
2805 if (!data) {
2806 NL_SET_ERR_MSG(extack,
2807 "Required attributes not provided to perform the operation");
2808 return -EINVAL;
2809 }
2810
2811 if (data[IFLA_VXLAN_ID]) {
2812 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2813
2814 if (id >= VXLAN_N_VID) {
2815 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2816 "VXLAN ID must be lower than 16777216");
2817 return -ERANGE;
2818 }
2819 }
2820
2821 if (data[IFLA_VXLAN_PORT_RANGE]) {
2822 const struct ifla_vxlan_port_range *p
2823 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2824
2825 if (ntohs(p->high) < ntohs(p->low)) {
2826 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2827 "Invalid source port range");
2828 return -EINVAL;
2829 }
2830 }
2831
2832 return 0;
2833 }
2834
2835 static void vxlan_get_drvinfo(struct net_device *netdev,
2836 struct ethtool_drvinfo *drvinfo)
2837 {
2838 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2839 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2840 }
2841
2842 static const struct ethtool_ops vxlan_ethtool_ops = {
2843 .get_drvinfo = vxlan_get_drvinfo,
2844 .get_link = ethtool_op_get_link,
2845 };
2846
2847 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2848 __be16 port, u32 flags)
2849 {
2850 struct socket *sock;
2851 struct udp_port_cfg udp_conf;
2852 int err;
2853
2854 memset(&udp_conf, 0, sizeof(udp_conf));
2855
2856 if (ipv6) {
2857 udp_conf.family = AF_INET6;
2858 udp_conf.use_udp6_rx_checksums =
2859 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2860 udp_conf.ipv6_v6only = 1;
2861 } else {
2862 udp_conf.family = AF_INET;
2863 }
2864
2865 udp_conf.local_udp_port = port;
2866
2867 /* Open UDP socket */
2868 err = udp_sock_create(net, &udp_conf, &sock);
2869 if (err < 0)
2870 return ERR_PTR(err);
2871
2872 return sock;
2873 }
2874
2875 /* Create new listen socket if needed */
2876 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2877 __be16 port, u32 flags)
2878 {
2879 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2880 struct vxlan_sock *vs;
2881 struct socket *sock;
2882 unsigned int h;
2883 struct udp_tunnel_sock_cfg tunnel_cfg;
2884
2885 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2886 if (!vs)
2887 return ERR_PTR(-ENOMEM);
2888
2889 for (h = 0; h < VNI_HASH_SIZE; ++h)
2890 INIT_HLIST_HEAD(&vs->vni_list[h]);
2891
2892 sock = vxlan_create_sock(net, ipv6, port, flags);
2893 if (IS_ERR(sock)) {
2894 kfree(vs);
2895 return ERR_CAST(sock);
2896 }
2897
2898 vs->sock = sock;
2899 refcount_set(&vs->refcnt, 1);
2900 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2901
2902 spin_lock(&vn->sock_lock);
2903 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2904 udp_tunnel_notify_add_rx_port(sock,
2905 (vs->flags & VXLAN_F_GPE) ?
2906 UDP_TUNNEL_TYPE_VXLAN_GPE :
2907 UDP_TUNNEL_TYPE_VXLAN);
2908 spin_unlock(&vn->sock_lock);
2909
2910 /* Mark socket as an encapsulation socket. */
2911 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
2912 tunnel_cfg.sk_user_data = vs;
2913 tunnel_cfg.encap_type = 1;
2914 tunnel_cfg.encap_rcv = vxlan_rcv;
2915 tunnel_cfg.encap_destroy = NULL;
2916 tunnel_cfg.gro_receive = vxlan_gro_receive;
2917 tunnel_cfg.gro_complete = vxlan_gro_complete;
2918
2919 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2920
2921 return vs;
2922 }
2923
2924 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2925 {
2926 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2927 struct vxlan_sock *vs = NULL;
2928 struct vxlan_dev_node *node;
2929
2930 if (!vxlan->cfg.no_share) {
2931 spin_lock(&vn->sock_lock);
2932 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2933 vxlan->cfg.dst_port, vxlan->cfg.flags);
2934 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
2935 spin_unlock(&vn->sock_lock);
2936 return -EBUSY;
2937 }
2938 spin_unlock(&vn->sock_lock);
2939 }
2940 if (!vs)
2941 vs = vxlan_socket_create(vxlan->net, ipv6,
2942 vxlan->cfg.dst_port, vxlan->cfg.flags);
2943 if (IS_ERR(vs))
2944 return PTR_ERR(vs);
2945 #if IS_ENABLED(CONFIG_IPV6)
2946 if (ipv6) {
2947 rcu_assign_pointer(vxlan->vn6_sock, vs);
2948 node = &vxlan->hlist6;
2949 } else
2950 #endif
2951 {
2952 rcu_assign_pointer(vxlan->vn4_sock, vs);
2953 node = &vxlan->hlist4;
2954 }
2955 vxlan_vs_add_dev(vs, vxlan, node);
2956 return 0;
2957 }
2958
2959 static int vxlan_sock_add(struct vxlan_dev *vxlan)
2960 {
2961 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
2962 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
2963 bool ipv4 = !ipv6 || metadata;
2964 int ret = 0;
2965
2966 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
2967 #if IS_ENABLED(CONFIG_IPV6)
2968 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
2969 if (ipv6) {
2970 ret = __vxlan_sock_add(vxlan, true);
2971 if (ret < 0 && ret != -EAFNOSUPPORT)
2972 ipv4 = false;
2973 }
2974 #endif
2975 if (ipv4)
2976 ret = __vxlan_sock_add(vxlan, false);
2977 if (ret < 0)
2978 vxlan_sock_release(vxlan);
2979 return ret;
2980 }
2981
2982 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
2983 struct net_device **lower,
2984 struct vxlan_dev *old,
2985 struct netlink_ext_ack *extack)
2986 {
2987 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
2988 struct vxlan_dev *tmp;
2989 bool use_ipv6 = false;
2990
2991 if (conf->flags & VXLAN_F_GPE) {
2992 /* For now, allow GPE only together with
2993 * COLLECT_METADATA. This can be relaxed later; in such
2994 * case, the other side of the PtP link will have to be
2995 * provided.
2996 */
2997 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
2998 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
2999 NL_SET_ERR_MSG(extack,
3000 "VXLAN GPE does not support this combination of attributes");
3001 return -EINVAL;
3002 }
3003 }
3004
3005 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3006 /* Unless IPv6 is explicitly requested, assume IPv4 */
3007 conf->remote_ip.sa.sa_family = AF_INET;
3008 conf->saddr.sa.sa_family = AF_INET;
3009 } else if (!conf->remote_ip.sa.sa_family) {
3010 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3011 } else if (!conf->saddr.sa.sa_family) {
3012 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3013 }
3014
3015 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3016 NL_SET_ERR_MSG(extack,
3017 "Local and remote address must be from the same family");
3018 return -EINVAL;
3019 }
3020
3021 if (vxlan_addr_multicast(&conf->saddr)) {
3022 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3023 return -EINVAL;
3024 }
3025
3026 if (conf->saddr.sa.sa_family == AF_INET6) {
3027 if (!IS_ENABLED(CONFIG_IPV6)) {
3028 NL_SET_ERR_MSG(extack,
3029 "IPv6 support not enabled in the kernel");
3030 return -EPFNOSUPPORT;
3031 }
3032 use_ipv6 = true;
3033 conf->flags |= VXLAN_F_IPV6;
3034
3035 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3036 int local_type =
3037 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3038 int remote_type =
3039 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3040
3041 if (local_type & IPV6_ADDR_LINKLOCAL) {
3042 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3043 (remote_type != IPV6_ADDR_ANY)) {
3044 NL_SET_ERR_MSG(extack,
3045 "Invalid combination of local and remote address scopes");
3046 return -EINVAL;
3047 }
3048
3049 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3050 } else {
3051 if (remote_type ==
3052 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3053 NL_SET_ERR_MSG(extack,
3054 "Invalid combination of local and remote address scopes");
3055 return -EINVAL;
3056 }
3057
3058 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3059 }
3060 }
3061 }
3062
3063 if (conf->label && !use_ipv6) {
3064 NL_SET_ERR_MSG(extack,
3065 "Label attribute only applies to IPv6 VXLAN devices");
3066 return -EINVAL;
3067 }
3068
3069 if (conf->remote_ifindex) {
3070 struct net_device *lowerdev;
3071
3072 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3073 if (!lowerdev) {
3074 NL_SET_ERR_MSG(extack,
3075 "Invalid local interface, device not found");
3076 return -ENODEV;
3077 }
3078
3079 #if IS_ENABLED(CONFIG_IPV6)
3080 if (use_ipv6) {
3081 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3082 if (idev && idev->cnf.disable_ipv6) {
3083 NL_SET_ERR_MSG(extack,
3084 "IPv6 support disabled by administrator");
3085 return -EPERM;
3086 }
3087 }
3088 #endif
3089
3090 *lower = lowerdev;
3091 } else {
3092 if (vxlan_addr_multicast(&conf->remote_ip)) {
3093 NL_SET_ERR_MSG(extack,
3094 "Local interface required for multicast remote destination");
3095
3096 return -EINVAL;
3097 }
3098
3099 #if IS_ENABLED(CONFIG_IPV6)
3100 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3101 NL_SET_ERR_MSG(extack,
3102 "Local interface required for link-local local/remote addresses");
3103 return -EINVAL;
3104 }
3105 #endif
3106
3107 *lower = NULL;
3108 }
3109
3110 if (!conf->dst_port) {
3111 if (conf->flags & VXLAN_F_GPE)
3112 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3113 else
3114 conf->dst_port = htons(vxlan_port);
3115 }
3116
3117 if (!conf->age_interval)
3118 conf->age_interval = FDB_AGE_DEFAULT;
3119
3120 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3121 if (tmp == old)
3122 continue;
3123
3124 if (tmp->cfg.vni != conf->vni)
3125 continue;
3126 if (tmp->cfg.dst_port != conf->dst_port)
3127 continue;
3128 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3129 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3130 continue;
3131
3132 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3133 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3134 continue;
3135
3136 NL_SET_ERR_MSG(extack,
3137 "A VXLAN device with the specified VNI already exists");
3138 return -EEXIST;
3139 }
3140
3141 return 0;
3142 }
3143
3144 static void vxlan_config_apply(struct net_device *dev,
3145 struct vxlan_config *conf,
3146 struct net_device *lowerdev,
3147 struct net *src_net,
3148 bool changelink)
3149 {
3150 struct vxlan_dev *vxlan = netdev_priv(dev);
3151 struct vxlan_rdst *dst = &vxlan->default_dst;
3152 unsigned short needed_headroom = ETH_HLEN;
3153 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3154 int max_mtu = ETH_MAX_MTU;
3155
3156 if (!changelink) {
3157 if (conf->flags & VXLAN_F_GPE)
3158 vxlan_raw_setup(dev);
3159 else
3160 vxlan_ether_setup(dev);
3161
3162 if (conf->mtu)
3163 dev->mtu = conf->mtu;
3164
3165 vxlan->net = src_net;
3166 }
3167
3168 dst->remote_vni = conf->vni;
3169
3170 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3171
3172 if (lowerdev) {
3173 dst->remote_ifindex = conf->remote_ifindex;
3174
3175 dev->gso_max_size = lowerdev->gso_max_size;
3176 dev->gso_max_segs = lowerdev->gso_max_segs;
3177
3178 needed_headroom = lowerdev->hard_header_len;
3179
3180 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3181 VXLAN_HEADROOM);
3182 if (max_mtu < ETH_MIN_MTU)
3183 max_mtu = ETH_MIN_MTU;
3184
3185 if (!changelink && !conf->mtu)
3186 dev->mtu = max_mtu;
3187 }
3188
3189 if (dev->mtu > max_mtu)
3190 dev->mtu = max_mtu;
3191
3192 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3193 needed_headroom += VXLAN6_HEADROOM;
3194 else
3195 needed_headroom += VXLAN_HEADROOM;
3196 dev->needed_headroom = needed_headroom;
3197
3198 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3199 }
3200
3201 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3202 struct vxlan_config *conf, bool changelink,
3203 struct netlink_ext_ack *extack)
3204 {
3205 struct vxlan_dev *vxlan = netdev_priv(dev);
3206 struct net_device *lowerdev;
3207 int ret;
3208
3209 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3210 if (ret)
3211 return ret;
3212
3213 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3214
3215 return 0;
3216 }
3217
3218 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3219 struct vxlan_config *conf,
3220 struct netlink_ext_ack *extack)
3221 {
3222 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3223 struct vxlan_dev *vxlan = netdev_priv(dev);
3224 struct vxlan_fdb *f = NULL;
3225 bool unregister = false;
3226 int err;
3227
3228 err = vxlan_dev_configure(net, dev, conf, false, extack);
3229 if (err)
3230 return err;
3231
3232 dev->ethtool_ops = &vxlan_ethtool_ops;
3233
3234 /* create an fdb entry for a valid default destination */
3235 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3236 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3237 &vxlan->default_dst.remote_ip,
3238 NUD_REACHABLE | NUD_PERMANENT,
3239 vxlan->cfg.dst_port,
3240 vxlan->default_dst.remote_vni,
3241 vxlan->default_dst.remote_vni,
3242 vxlan->default_dst.remote_ifindex,
3243 NTF_SELF, &f);
3244 if (err)
3245 return err;
3246 }
3247
3248 err = register_netdevice(dev);
3249 if (err)
3250 goto errout;
3251 unregister = true;
3252
3253 err = rtnl_configure_link(dev, NULL);
3254 if (err)
3255 goto errout;
3256
3257 /* notify default fdb entry */
3258 if (f)
3259 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_NEWNEIGH);
3260
3261 list_add(&vxlan->next, &vn->vxlan_list);
3262 return 0;
3263
3264 errout:
3265 /* unregister_netdevice() destroys the default FDB entry with deletion
3266 * notification. But the addition notification was not sent yet, so
3267 * destroy the entry by hand here.
3268 */
3269 if (f)
3270 vxlan_fdb_destroy(vxlan, f, false);
3271 if (unregister)
3272 unregister_netdevice(dev);
3273 return err;
3274 }
3275
3276 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3277 struct net_device *dev, struct vxlan_config *conf,
3278 bool changelink)
3279 {
3280 struct vxlan_dev *vxlan = netdev_priv(dev);
3281
3282 memset(conf, 0, sizeof(*conf));
3283
3284 /* if changelink operation, start with old existing cfg */
3285 if (changelink)
3286 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3287
3288 if (data[IFLA_VXLAN_ID]) {
3289 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3290
3291 if (changelink && (vni != conf->vni))
3292 return -EOPNOTSUPP;
3293 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3294 }
3295
3296 if (data[IFLA_VXLAN_GROUP]) {
3297 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3298 return -EOPNOTSUPP;
3299
3300 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3301 conf->remote_ip.sa.sa_family = AF_INET;
3302 } else if (data[IFLA_VXLAN_GROUP6]) {
3303 if (!IS_ENABLED(CONFIG_IPV6))
3304 return -EPFNOSUPPORT;
3305
3306 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3307 return -EOPNOTSUPP;
3308
3309 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3310 conf->remote_ip.sa.sa_family = AF_INET6;
3311 }
3312
3313 if (data[IFLA_VXLAN_LOCAL]) {
3314 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3315 return -EOPNOTSUPP;
3316
3317 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3318 conf->saddr.sa.sa_family = AF_INET;
3319 } else if (data[IFLA_VXLAN_LOCAL6]) {
3320 if (!IS_ENABLED(CONFIG_IPV6))
3321 return -EPFNOSUPPORT;
3322
3323 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3324 return -EOPNOTSUPP;
3325
3326 /* TODO: respect scope id */
3327 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3328 conf->saddr.sa.sa_family = AF_INET6;
3329 }
3330
3331 if (data[IFLA_VXLAN_LINK])
3332 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3333
3334 if (data[IFLA_VXLAN_TOS])
3335 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3336
3337 if (data[IFLA_VXLAN_TTL])
3338 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3339
3340 if (data[IFLA_VXLAN_TTL_INHERIT]) {
3341 if (changelink)
3342 return -EOPNOTSUPP;
3343 conf->flags |= VXLAN_F_TTL_INHERIT;
3344 }
3345
3346 if (data[IFLA_VXLAN_LABEL])
3347 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3348 IPV6_FLOWLABEL_MASK;
3349
3350 if (data[IFLA_VXLAN_LEARNING]) {
3351 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
3352 conf->flags |= VXLAN_F_LEARN;
3353 else
3354 conf->flags &= ~VXLAN_F_LEARN;
3355 } else if (!changelink) {
3356 /* default to learn on a new device */
3357 conf->flags |= VXLAN_F_LEARN;
3358 }
3359
3360 if (data[IFLA_VXLAN_AGEING]) {
3361 if (changelink)
3362 return -EOPNOTSUPP;
3363 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3364 }
3365
3366 if (data[IFLA_VXLAN_PROXY]) {
3367 if (changelink)
3368 return -EOPNOTSUPP;
3369 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3370 conf->flags |= VXLAN_F_PROXY;
3371 }
3372
3373 if (data[IFLA_VXLAN_RSC]) {
3374 if (changelink)
3375 return -EOPNOTSUPP;
3376 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3377 conf->flags |= VXLAN_F_RSC;
3378 }
3379
3380 if (data[IFLA_VXLAN_L2MISS]) {
3381 if (changelink)
3382 return -EOPNOTSUPP;
3383 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3384 conf->flags |= VXLAN_F_L2MISS;
3385 }
3386
3387 if (data[IFLA_VXLAN_L3MISS]) {
3388 if (changelink)
3389 return -EOPNOTSUPP;
3390 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3391 conf->flags |= VXLAN_F_L3MISS;
3392 }
3393
3394 if (data[IFLA_VXLAN_LIMIT]) {
3395 if (changelink)
3396 return -EOPNOTSUPP;
3397 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3398 }
3399
3400 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3401 if (changelink)
3402 return -EOPNOTSUPP;
3403 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3404 conf->flags |= VXLAN_F_COLLECT_METADATA;
3405 }
3406
3407 if (data[IFLA_VXLAN_PORT_RANGE]) {
3408 if (!changelink) {
3409 const struct ifla_vxlan_port_range *p
3410 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3411 conf->port_min = ntohs(p->low);
3412 conf->port_max = ntohs(p->high);
3413 } else {
3414 return -EOPNOTSUPP;
3415 }
3416 }
3417
3418 if (data[IFLA_VXLAN_PORT]) {
3419 if (changelink)
3420 return -EOPNOTSUPP;
3421 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3422 }
3423
3424 if (data[IFLA_VXLAN_UDP_CSUM]) {
3425 if (changelink)
3426 return -EOPNOTSUPP;
3427 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3428 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3429 }
3430
3431 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3432 if (changelink)
3433 return -EOPNOTSUPP;
3434 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3435 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3436 }
3437
3438 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3439 if (changelink)
3440 return -EOPNOTSUPP;
3441 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3442 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3443 }
3444
3445 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3446 if (changelink)
3447 return -EOPNOTSUPP;
3448 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3449 conf->flags |= VXLAN_F_REMCSUM_TX;
3450 }
3451
3452 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3453 if (changelink)
3454 return -EOPNOTSUPP;
3455 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3456 conf->flags |= VXLAN_F_REMCSUM_RX;
3457 }
3458
3459 if (data[IFLA_VXLAN_GBP]) {
3460 if (changelink)
3461 return -EOPNOTSUPP;
3462 conf->flags |= VXLAN_F_GBP;
3463 }
3464
3465 if (data[IFLA_VXLAN_GPE]) {
3466 if (changelink)
3467 return -EOPNOTSUPP;
3468 conf->flags |= VXLAN_F_GPE;
3469 }
3470
3471 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3472 if (changelink)
3473 return -EOPNOTSUPP;
3474 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3475 }
3476
3477 if (tb[IFLA_MTU]) {
3478 if (changelink)
3479 return -EOPNOTSUPP;
3480 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3481 }
3482
3483 return 0;
3484 }
3485
3486 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3487 struct nlattr *tb[], struct nlattr *data[],
3488 struct netlink_ext_ack *extack)
3489 {
3490 struct vxlan_config conf;
3491 int err;
3492
3493 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3494 if (err)
3495 return err;
3496
3497 return __vxlan_dev_create(src_net, dev, &conf, extack);
3498 }
3499
3500 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3501 struct nlattr *data[],
3502 struct netlink_ext_ack *extack)
3503 {
3504 struct vxlan_dev *vxlan = netdev_priv(dev);
3505 struct vxlan_rdst *dst = &vxlan->default_dst;
3506 struct vxlan_rdst old_dst;
3507 struct vxlan_config conf;
3508 int err;
3509
3510 err = vxlan_nl2conf(tb, data,
3511 dev, &conf, true);
3512 if (err)
3513 return err;
3514
3515 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
3516
3517 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
3518 if (err)
3519 return err;
3520
3521 /* handle default dst entry */
3522 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3523 spin_lock_bh(&vxlan->hash_lock);
3524 if (!vxlan_addr_any(&old_dst.remote_ip))
3525 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3526 old_dst.remote_ip,
3527 vxlan->cfg.dst_port,
3528 old_dst.remote_vni,
3529 old_dst.remote_vni,
3530 old_dst.remote_ifindex, 0);
3531
3532 if (!vxlan_addr_any(&dst->remote_ip)) {
3533 err = vxlan_fdb_update(vxlan, all_zeros_mac,
3534 &dst->remote_ip,
3535 NUD_REACHABLE | NUD_PERMANENT,
3536 NLM_F_APPEND | NLM_F_CREATE,
3537 vxlan->cfg.dst_port,
3538 dst->remote_vni,
3539 dst->remote_vni,
3540 dst->remote_ifindex,
3541 NTF_SELF);
3542 if (err) {
3543 spin_unlock_bh(&vxlan->hash_lock);
3544 return err;
3545 }
3546 }
3547 spin_unlock_bh(&vxlan->hash_lock);
3548 }
3549
3550 return 0;
3551 }
3552
3553 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3554 {
3555 struct vxlan_dev *vxlan = netdev_priv(dev);
3556
3557 vxlan_flush(vxlan, true);
3558
3559 list_del(&vxlan->next);
3560 unregister_netdevice_queue(dev, head);
3561 }
3562
3563 static size_t vxlan_get_size(const struct net_device *dev)
3564 {
3565
3566 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
3567 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
3568 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
3569 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3570 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3571 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL_INHERIT */
3572 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3573 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
3574 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
3575 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3576 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3577 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3578 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
3579 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
3580 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3581 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
3582 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3583 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3584 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3585 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3586 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3587 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3588 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3589 0;
3590 }
3591
3592 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3593 {
3594 const struct vxlan_dev *vxlan = netdev_priv(dev);
3595 const struct vxlan_rdst *dst = &vxlan->default_dst;
3596 struct ifla_vxlan_port_range ports = {
3597 .low = htons(vxlan->cfg.port_min),
3598 .high = htons(vxlan->cfg.port_max),
3599 };
3600
3601 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
3602 goto nla_put_failure;
3603
3604 if (!vxlan_addr_any(&dst->remote_ip)) {
3605 if (dst->remote_ip.sa.sa_family == AF_INET) {
3606 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3607 dst->remote_ip.sin.sin_addr.s_addr))
3608 goto nla_put_failure;
3609 #if IS_ENABLED(CONFIG_IPV6)
3610 } else {
3611 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3612 &dst->remote_ip.sin6.sin6_addr))
3613 goto nla_put_failure;
3614 #endif
3615 }
3616 }
3617
3618 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3619 goto nla_put_failure;
3620
3621 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3622 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3623 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3624 vxlan->cfg.saddr.sin.sin_addr.s_addr))
3625 goto nla_put_failure;
3626 #if IS_ENABLED(CONFIG_IPV6)
3627 } else {
3628 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3629 &vxlan->cfg.saddr.sin6.sin6_addr))
3630 goto nla_put_failure;
3631 #endif
3632 }
3633 }
3634
3635 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3636 nla_put_u8(skb, IFLA_VXLAN_TTL_INHERIT,
3637 !!(vxlan->cfg.flags & VXLAN_F_TTL_INHERIT)) ||
3638 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3639 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
3640 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3641 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
3642 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3643 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3644 nla_put_u8(skb, IFLA_VXLAN_RSC,
3645 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
3646 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3647 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
3648 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3649 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
3650 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3651 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
3652 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3653 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3654 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3655 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3656 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
3657 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3658 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3659 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3660 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3661 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3662 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
3663 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3664 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
3665 goto nla_put_failure;
3666
3667 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3668 goto nla_put_failure;
3669
3670 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3671 nla_put_flag(skb, IFLA_VXLAN_GBP))
3672 goto nla_put_failure;
3673
3674 if (vxlan->cfg.flags & VXLAN_F_GPE &&
3675 nla_put_flag(skb, IFLA_VXLAN_GPE))
3676 goto nla_put_failure;
3677
3678 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3679 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3680 goto nla_put_failure;
3681
3682 return 0;
3683
3684 nla_put_failure:
3685 return -EMSGSIZE;
3686 }
3687
3688 static struct net *vxlan_get_link_net(const struct net_device *dev)
3689 {
3690 struct vxlan_dev *vxlan = netdev_priv(dev);
3691
3692 return vxlan->net;
3693 }
3694
3695 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3696 .kind = "vxlan",
3697 .maxtype = IFLA_VXLAN_MAX,
3698 .policy = vxlan_policy,
3699 .priv_size = sizeof(struct vxlan_dev),
3700 .setup = vxlan_setup,
3701 .validate = vxlan_validate,
3702 .newlink = vxlan_newlink,
3703 .changelink = vxlan_changelink,
3704 .dellink = vxlan_dellink,
3705 .get_size = vxlan_get_size,
3706 .fill_info = vxlan_fill_info,
3707 .get_link_net = vxlan_get_link_net,
3708 };
3709
3710 struct net_device *vxlan_dev_create(struct net *net, const char *name,
3711 u8 name_assign_type,
3712 struct vxlan_config *conf)
3713 {
3714 struct nlattr *tb[IFLA_MAX + 1];
3715 struct net_device *dev;
3716 int err;
3717
3718 memset(&tb, 0, sizeof(tb));
3719
3720 dev = rtnl_create_link(net, name, name_assign_type,
3721 &vxlan_link_ops, tb);
3722 if (IS_ERR(dev))
3723 return dev;
3724
3725 err = __vxlan_dev_create(net, dev, conf, NULL);
3726 if (err < 0) {
3727 free_netdev(dev);
3728 return ERR_PTR(err);
3729 }
3730
3731 err = rtnl_configure_link(dev, NULL);
3732 if (err < 0) {
3733 LIST_HEAD(list_kill);
3734
3735 vxlan_dellink(dev, &list_kill);
3736 unregister_netdevice_many(&list_kill);
3737 return ERR_PTR(err);
3738 }
3739
3740 return dev;
3741 }
3742 EXPORT_SYMBOL_GPL(vxlan_dev_create);
3743
3744 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3745 struct net_device *dev)
3746 {
3747 struct vxlan_dev *vxlan, *next;
3748 LIST_HEAD(list_kill);
3749
3750 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3751 struct vxlan_rdst *dst = &vxlan->default_dst;
3752
3753 /* In case we created vxlan device with carrier
3754 * and we loose the carrier due to module unload
3755 * we also need to remove vxlan device. In other
3756 * cases, it's not necessary and remote_ifindex
3757 * is 0 here, so no matches.
3758 */
3759 if (dst->remote_ifindex == dev->ifindex)
3760 vxlan_dellink(vxlan->dev, &list_kill);
3761 }
3762
3763 unregister_netdevice_many(&list_kill);
3764 }
3765
3766 static int vxlan_netdevice_event(struct notifier_block *unused,
3767 unsigned long event, void *ptr)
3768 {
3769 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3770 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3771
3772 if (event == NETDEV_UNREGISTER) {
3773 vxlan_offload_rx_ports(dev, false);
3774 vxlan_handle_lowerdev_unregister(vn, dev);
3775 } else if (event == NETDEV_REGISTER) {
3776 vxlan_offload_rx_ports(dev, true);
3777 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3778 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
3779 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
3780 }
3781
3782 return NOTIFY_DONE;
3783 }
3784
3785 static struct notifier_block vxlan_notifier_block __read_mostly = {
3786 .notifier_call = vxlan_netdevice_event,
3787 };
3788
3789 static __net_init int vxlan_init_net(struct net *net)
3790 {
3791 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3792 unsigned int h;
3793
3794 INIT_LIST_HEAD(&vn->vxlan_list);
3795 spin_lock_init(&vn->sock_lock);
3796
3797 for (h = 0; h < PORT_HASH_SIZE; ++h)
3798 INIT_HLIST_HEAD(&vn->sock_list[h]);
3799
3800 return 0;
3801 }
3802
3803 static void vxlan_destroy_tunnels(struct net *net, struct list_head *head)
3804 {
3805 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3806 struct vxlan_dev *vxlan, *next;
3807 struct net_device *dev, *aux;
3808 unsigned int h;
3809
3810 for_each_netdev_safe(net, dev, aux)
3811 if (dev->rtnl_link_ops == &vxlan_link_ops)
3812 unregister_netdevice_queue(dev, head);
3813
3814 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3815 /* If vxlan->dev is in the same netns, it has already been added
3816 * to the list by the previous loop.
3817 */
3818 if (!net_eq(dev_net(vxlan->dev), net))
3819 unregister_netdevice_queue(vxlan->dev, head);
3820 }
3821
3822 for (h = 0; h < PORT_HASH_SIZE; ++h)
3823 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
3824 }
3825
3826 static void __net_exit vxlan_exit_batch_net(struct list_head *net_list)
3827 {
3828 struct net *net;
3829 LIST_HEAD(list);
3830
3831 rtnl_lock();
3832 list_for_each_entry(net, net_list, exit_list)
3833 vxlan_destroy_tunnels(net, &list);
3834
3835 unregister_netdevice_many(&list);
3836 rtnl_unlock();
3837 }
3838
3839 static struct pernet_operations vxlan_net_ops = {
3840 .init = vxlan_init_net,
3841 .exit_batch = vxlan_exit_batch_net,
3842 .id = &vxlan_net_id,
3843 .size = sizeof(struct vxlan_net),
3844 };
3845
3846 static int __init vxlan_init_module(void)
3847 {
3848 int rc;
3849
3850 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3851
3852 rc = register_pernet_subsys(&vxlan_net_ops);
3853 if (rc)
3854 goto out1;
3855
3856 rc = register_netdevice_notifier(&vxlan_notifier_block);
3857 if (rc)
3858 goto out2;
3859
3860 rc = rtnl_link_register(&vxlan_link_ops);
3861 if (rc)
3862 goto out3;
3863
3864 return 0;
3865 out3:
3866 unregister_netdevice_notifier(&vxlan_notifier_block);
3867 out2:
3868 unregister_pernet_subsys(&vxlan_net_ops);
3869 out1:
3870 return rc;
3871 }
3872 late_initcall(vxlan_init_module);
3873
3874 static void __exit vxlan_cleanup_module(void)
3875 {
3876 rtnl_link_unregister(&vxlan_link_ops);
3877 unregister_netdevice_notifier(&vxlan_notifier_block);
3878 unregister_pernet_subsys(&vxlan_net_ops);
3879 /* rcu_barrier() is called by netns */
3880 }
3881 module_exit(vxlan_cleanup_module);
3882
3883 MODULE_LICENSE("GPL");
3884 MODULE_VERSION(VXLAN_VERSION);
3885 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3886 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3887 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux with added FPC-III support