projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'v4.5-rc5' into devel
[deliverable/linux.git] / drivers / net / vxlan.c
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/types.h>
15 #include <linux/module.h>
16 #include <linux/errno.h>
17 #include <linux/slab.h>
18 #include <linux/skbuff.h>
19 #include <linux/rculist.h>
20 #include <linux/netdevice.h>
21 #include <linux/in.h>
22 #include <linux/ip.h>
23 #include <linux/udp.h>
24 #include <linux/igmp.h>
25 #include <linux/etherdevice.h>
26 #include <linux/if_ether.h>
27 #include <linux/if_vlan.h>
28 #include <linux/hash.h>
29 #include <linux/ethtool.h>
30 #include <net/arp.h>
31 #include <net/ndisc.h>
32 #include <net/ip.h>
33 #include <net/ip_tunnels.h>
34 #include <net/icmp.h>
35 #include <net/udp.h>
36 #include <net/udp_tunnel.h>
37 #include <net/rtnetlink.h>
38 #include <net/route.h>
39 #include <net/dsfield.h>
40 #include <net/inet_ecn.h>
41 #include <net/net_namespace.h>
42 #include <net/netns/generic.h>
43 #include <net/vxlan.h>
44 #include <net/protocol.h>
45 #include <net/udp_tunnel.h>
46 #if IS_ENABLED(CONFIG_IPV6)
47 #include <net/ipv6.h>
48 #include <net/addrconf.h>
49 #include <net/ip6_tunnel.h>
50 #include <net/ip6_checksum.h>
51 #endif
52 #include <net/dst_metadata.h>
53
54 #define VXLAN_VERSION "0.1"
55
56 #define PORT_HASH_BITS 8
57 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
58 #define FDB_AGE_DEFAULT 300 /* 5 min */
59 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
61 /* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
63 * for compatibility with early adopters.
64 */
65 static unsigned short vxlan_port __read_mostly = 8472;
66 module_param_named(udp_port, vxlan_port, ushort, 0444);
67 MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69 static bool log_ecn_error = true;
70 module_param(log_ecn_error, bool, 0644);
71 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
73 static int vxlan_net_id;
74 static struct rtnl_link_ops vxlan_link_ops;
75
76 static const u8 all_zeros_mac[ETH_ALEN + 2];
77
78 static int vxlan_sock_add(struct vxlan_dev *vxlan);
79
80 /* per-network namespace private data for this module */
81 struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
84 spinlock_t sock_lock;
85 };
86
87 /* Forwarding table entry */
88 struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
93 struct list_head remotes;
94 u8 eth_addr[ETH_ALEN];
95 u16 state; /* see ndm_state */
96 u8 flags; /* see ndm_flags */
97 };
98
99 /* salt for hash table */
100 static u32 vxlan_salt __read_mostly;
101 static struct workqueue_struct *vxlan_wq;
102
103 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104 {
105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
107 }
108
109 #if IS_ENABLED(CONFIG_IPV6)
110 static inline
111 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112 {
113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
119 }
120
121 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122 {
123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
127 }
128
129 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130 {
131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
135 }
136
137 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138 {
139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
150 }
151
152 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
153 const union vxlan_addr *ip)
154 {
155 if (ip->sa.sa_family == AF_INET6)
156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
157 else
158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
159 }
160
161 #else /* !CONFIG_IPV6 */
162
163 static inline
164 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165 {
166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
167 }
168
169 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170 {
171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
172 }
173
174 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175 {
176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
177 }
178
179 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180 {
181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
190 }
191
192 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
193 const union vxlan_addr *ip)
194 {
195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
196 }
197 #endif
198
199 /* Virtual Network hash table head */
200 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
201 {
202 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
203 }
204
205 /* Socket hash table head */
206 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
207 {
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211 }
212
213 /* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
216 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
217 {
218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219 }
220
221 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222 {
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
224 }
225
226 /* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
231 {
232 struct vxlan_sock *vs;
233
234 flags &= VXLAN_F_RCV_FLAGS;
235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
238 vxlan_get_sk_family(vs) == family &&
239 vs->flags == flags)
240 return vs;
241 }
242 return NULL;
243 }
244
245 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
246 {
247 struct vxlan_dev *vxlan;
248
249 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
250 if (vxlan->default_dst.remote_vni == id)
251 return 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, u32 id,
259 sa_family_t family, __be16 port,
260 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, id);
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 ndm->ndm_family = AF_INET;
294 send_ip = !vxlan_addr_any(&rdst->remote_ip);
295 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 } else
297 ndm->ndm_family = AF_BRIDGE;
298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
300 ndm->ndm_flags = fdb->flags;
301 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_alloc(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, rdst->remote_vni))
319 goto nla_put_failure;
320 if (rdst->remote_ifindex &&
321 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
322 goto nla_put_failure;
323
324 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
325 ci.ndm_confirmed = 0;
326 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
327 ci.ndm_refcnt = 0;
328
329 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
330 goto nla_put_failure;
331
332 nlmsg_end(skb, nlh);
333 return 0;
334
335 nla_put_failure:
336 nlmsg_cancel(skb, nlh);
337 return -EMSGSIZE;
338 }
339
340 static inline size_t vxlan_nlmsg_size(void)
341 {
342 return NLMSG_ALIGN(sizeof(struct ndmsg))
343 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
344 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
345 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
346 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
347 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
348 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
349 + nla_total_size(sizeof(struct nda_cacheinfo));
350 }
351
352 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
353 struct vxlan_rdst *rd, int type)
354 {
355 struct net *net = dev_net(vxlan->dev);
356 struct sk_buff *skb;
357 int err = -ENOBUFS;
358
359 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
360 if (skb == NULL)
361 goto errout;
362
363 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
364 if (err < 0) {
365 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
366 WARN_ON(err == -EMSGSIZE);
367 kfree_skb(skb);
368 goto errout;
369 }
370
371 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
372 return;
373 errout:
374 if (err < 0)
375 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
376 }
377
378 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
379 {
380 struct vxlan_dev *vxlan = netdev_priv(dev);
381 struct vxlan_fdb f = {
382 .state = NUD_STALE,
383 };
384 struct vxlan_rdst remote = {
385 .remote_ip = *ipa, /* goes to NDA_DST */
386 .remote_vni = VXLAN_N_VID,
387 };
388
389 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
390 }
391
392 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
393 {
394 struct vxlan_fdb f = {
395 .state = NUD_STALE,
396 };
397 struct vxlan_rdst remote = { };
398
399 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
400
401 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
402 }
403
404 /* Hash Ethernet address */
405 static u32 eth_hash(const unsigned char *addr)
406 {
407 u64 value = get_unaligned((u64 *)addr);
408
409 /* only want 6 bytes */
410 #ifdef __BIG_ENDIAN
411 value >>= 16;
412 #else
413 value <<= 16;
414 #endif
415 return hash_64(value, FDB_HASH_BITS);
416 }
417
418 /* Hash chain to use given mac address */
419 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
420 const u8 *mac)
421 {
422 return &vxlan->fdb_head[eth_hash(mac)];
423 }
424
425 /* Look up Ethernet address in forwarding table */
426 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
427 const u8 *mac)
428 {
429 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
430 struct vxlan_fdb *f;
431
432 hlist_for_each_entry_rcu(f, head, hlist) {
433 if (ether_addr_equal(mac, f->eth_addr))
434 return f;
435 }
436
437 return NULL;
438 }
439
440 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
441 const u8 *mac)
442 {
443 struct vxlan_fdb *f;
444
445 f = __vxlan_find_mac(vxlan, mac);
446 if (f)
447 f->used = jiffies;
448
449 return f;
450 }
451
452 /* caller should hold vxlan->hash_lock */
453 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
454 union vxlan_addr *ip, __be16 port,
455 __u32 vni, __u32 ifindex)
456 {
457 struct vxlan_rdst *rd;
458
459 list_for_each_entry(rd, &f->remotes, list) {
460 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
461 rd->remote_port == port &&
462 rd->remote_vni == vni &&
463 rd->remote_ifindex == ifindex)
464 return rd;
465 }
466
467 return NULL;
468 }
469
470 /* Replace destination of unicast mac */
471 static int vxlan_fdb_replace(struct vxlan_fdb *f,
472 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
473 {
474 struct vxlan_rdst *rd;
475
476 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
477 if (rd)
478 return 0;
479
480 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
481 if (!rd)
482 return 0;
483 rd->remote_ip = *ip;
484 rd->remote_port = port;
485 rd->remote_vni = vni;
486 rd->remote_ifindex = ifindex;
487 return 1;
488 }
489
490 /* Add/update destinations for multicast */
491 static int vxlan_fdb_append(struct vxlan_fdb *f,
492 union vxlan_addr *ip, __be16 port, __u32 vni,
493 __u32 ifindex, struct vxlan_rdst **rdp)
494 {
495 struct vxlan_rdst *rd;
496
497 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
498 if (rd)
499 return 0;
500
501 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
502 if (rd == NULL)
503 return -ENOBUFS;
504 rd->remote_ip = *ip;
505 rd->remote_port = port;
506 rd->remote_vni = vni;
507 rd->remote_ifindex = ifindex;
508
509 list_add_tail_rcu(&rd->list, &f->remotes);
510
511 *rdp = rd;
512 return 1;
513 }
514
515 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
516 unsigned int off,
517 struct vxlanhdr *vh, size_t hdrlen,
518 u32 data, struct gro_remcsum *grc,
519 bool nopartial)
520 {
521 size_t start, offset;
522
523 if (skb->remcsum_offload)
524 return vh;
525
526 if (!NAPI_GRO_CB(skb)->csum_valid)
527 return NULL;
528
529 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
530 offset = start + ((data & VXLAN_RCO_UDP) ?
531 offsetof(struct udphdr, check) :
532 offsetof(struct tcphdr, check));
533
534 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
535 start, offset, grc, nopartial);
536
537 skb->remcsum_offload = 1;
538
539 return vh;
540 }
541
542 static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
543 struct sk_buff *skb,
544 struct udp_offload *uoff)
545 {
546 struct sk_buff *p, **pp = NULL;
547 struct vxlanhdr *vh, *vh2;
548 unsigned int hlen, off_vx;
549 int flush = 1;
550 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
551 udp_offloads);
552 u32 flags;
553 struct gro_remcsum grc;
554
555 skb_gro_remcsum_init(&grc);
556
557 off_vx = skb_gro_offset(skb);
558 hlen = off_vx + sizeof(*vh);
559 vh = skb_gro_header_fast(skb, off_vx);
560 if (skb_gro_header_hard(skb, hlen)) {
561 vh = skb_gro_header_slow(skb, hlen, off_vx);
562 if (unlikely(!vh))
563 goto out;
564 }
565
566 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
567
568 flags = ntohl(vh->vx_flags);
569
570 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
571 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
572 ntohl(vh->vx_vni), &grc,
573 !!(vs->flags &
574 VXLAN_F_REMCSUM_NOPARTIAL));
575
576 if (!vh)
577 goto out;
578 }
579
580 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
581
582 flush = 0;
583
584 for (p = *head; p; p = p->next) {
585 if (!NAPI_GRO_CB(p)->same_flow)
586 continue;
587
588 vh2 = (struct vxlanhdr *)(p->data + off_vx);
589 if (vh->vx_flags != vh2->vx_flags ||
590 vh->vx_vni != vh2->vx_vni) {
591 NAPI_GRO_CB(p)->same_flow = 0;
592 continue;
593 }
594 }
595
596 pp = eth_gro_receive(head, skb);
597
598 out:
599 skb_gro_remcsum_cleanup(skb, &grc);
600 NAPI_GRO_CB(skb)->flush |= flush;
601
602 return pp;
603 }
604
605 static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
606 struct udp_offload *uoff)
607 {
608 udp_tunnel_gro_complete(skb, nhoff);
609
610 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
611 }
612
613 /* Notify netdevs that UDP port started listening */
614 static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
615 {
616 struct net_device *dev;
617 struct sock *sk = vs->sock->sk;
618 struct net *net = sock_net(sk);
619 sa_family_t sa_family = vxlan_get_sk_family(vs);
620 __be16 port = inet_sk(sk)->inet_sport;
621 int err;
622
623 if (sa_family == AF_INET) {
624 err = udp_add_offload(net, &vs->udp_offloads);
625 if (err)
626 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
627 }
628
629 rcu_read_lock();
630 for_each_netdev_rcu(net, dev) {
631 if (dev->netdev_ops->ndo_add_vxlan_port)
632 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
633 port);
634 }
635 rcu_read_unlock();
636 }
637
638 /* Notify netdevs that UDP port is no more listening */
639 static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
640 {
641 struct net_device *dev;
642 struct sock *sk = vs->sock->sk;
643 struct net *net = sock_net(sk);
644 sa_family_t sa_family = vxlan_get_sk_family(vs);
645 __be16 port = inet_sk(sk)->inet_sport;
646
647 rcu_read_lock();
648 for_each_netdev_rcu(net, dev) {
649 if (dev->netdev_ops->ndo_del_vxlan_port)
650 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
651 port);
652 }
653 rcu_read_unlock();
654
655 if (sa_family == AF_INET)
656 udp_del_offload(&vs->udp_offloads);
657 }
658
659 /* Add new entry to forwarding table -- assumes lock held */
660 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
661 const u8 *mac, union vxlan_addr *ip,
662 __u16 state, __u16 flags,
663 __be16 port, __u32 vni, __u32 ifindex,
664 __u8 ndm_flags)
665 {
666 struct vxlan_rdst *rd = NULL;
667 struct vxlan_fdb *f;
668 int notify = 0;
669
670 f = __vxlan_find_mac(vxlan, mac);
671 if (f) {
672 if (flags & NLM_F_EXCL) {
673 netdev_dbg(vxlan->dev,
674 "lost race to create %pM\n", mac);
675 return -EEXIST;
676 }
677 if (f->state != state) {
678 f->state = state;
679 f->updated = jiffies;
680 notify = 1;
681 }
682 if (f->flags != ndm_flags) {
683 f->flags = ndm_flags;
684 f->updated = jiffies;
685 notify = 1;
686 }
687 if ((flags & NLM_F_REPLACE)) {
688 /* Only change unicasts */
689 if (!(is_multicast_ether_addr(f->eth_addr) ||
690 is_zero_ether_addr(f->eth_addr))) {
691 notify |= vxlan_fdb_replace(f, ip, port, vni,
692 ifindex);
693 } else
694 return -EOPNOTSUPP;
695 }
696 if ((flags & NLM_F_APPEND) &&
697 (is_multicast_ether_addr(f->eth_addr) ||
698 is_zero_ether_addr(f->eth_addr))) {
699 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
700 &rd);
701
702 if (rc < 0)
703 return rc;
704 notify |= rc;
705 }
706 } else {
707 if (!(flags & NLM_F_CREATE))
708 return -ENOENT;
709
710 if (vxlan->cfg.addrmax &&
711 vxlan->addrcnt >= vxlan->cfg.addrmax)
712 return -ENOSPC;
713
714 /* Disallow replace to add a multicast entry */
715 if ((flags & NLM_F_REPLACE) &&
716 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
717 return -EOPNOTSUPP;
718
719 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
720 f = kmalloc(sizeof(*f), GFP_ATOMIC);
721 if (!f)
722 return -ENOMEM;
723
724 notify = 1;
725 f->state = state;
726 f->flags = ndm_flags;
727 f->updated = f->used = jiffies;
728 INIT_LIST_HEAD(&f->remotes);
729 memcpy(f->eth_addr, mac, ETH_ALEN);
730
731 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
732
733 ++vxlan->addrcnt;
734 hlist_add_head_rcu(&f->hlist,
735 vxlan_fdb_head(vxlan, mac));
736 }
737
738 if (notify) {
739 if (rd == NULL)
740 rd = first_remote_rtnl(f);
741 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
742 }
743
744 return 0;
745 }
746
747 static void vxlan_fdb_free(struct rcu_head *head)
748 {
749 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
750 struct vxlan_rdst *rd, *nd;
751
752 list_for_each_entry_safe(rd, nd, &f->remotes, list)
753 kfree(rd);
754 kfree(f);
755 }
756
757 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
758 {
759 netdev_dbg(vxlan->dev,
760 "delete %pM\n", f->eth_addr);
761
762 --vxlan->addrcnt;
763 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
764
765 hlist_del_rcu(&f->hlist);
766 call_rcu(&f->rcu, vxlan_fdb_free);
767 }
768
769 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
770 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
771 {
772 struct net *net = dev_net(vxlan->dev);
773 int err;
774
775 if (tb[NDA_DST]) {
776 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
777 if (err)
778 return err;
779 } else {
780 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
781 if (remote->sa.sa_family == AF_INET) {
782 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
783 ip->sa.sa_family = AF_INET;
784 #if IS_ENABLED(CONFIG_IPV6)
785 } else {
786 ip->sin6.sin6_addr = in6addr_any;
787 ip->sa.sa_family = AF_INET6;
788 #endif
789 }
790 }
791
792 if (tb[NDA_PORT]) {
793 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
794 return -EINVAL;
795 *port = nla_get_be16(tb[NDA_PORT]);
796 } else {
797 *port = vxlan->cfg.dst_port;
798 }
799
800 if (tb[NDA_VNI]) {
801 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
802 return -EINVAL;
803 *vni = nla_get_u32(tb[NDA_VNI]);
804 } else {
805 *vni = vxlan->default_dst.remote_vni;
806 }
807
808 if (tb[NDA_IFINDEX]) {
809 struct net_device *tdev;
810
811 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
812 return -EINVAL;
813 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
814 tdev = __dev_get_by_index(net, *ifindex);
815 if (!tdev)
816 return -EADDRNOTAVAIL;
817 } else {
818 *ifindex = 0;
819 }
820
821 return 0;
822 }
823
824 /* Add static entry (via netlink) */
825 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
826 struct net_device *dev,
827 const unsigned char *addr, u16 vid, u16 flags)
828 {
829 struct vxlan_dev *vxlan = netdev_priv(dev);
830 /* struct net *net = dev_net(vxlan->dev); */
831 union vxlan_addr ip;
832 __be16 port;
833 u32 vni, ifindex;
834 int err;
835
836 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
837 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
838 ndm->ndm_state);
839 return -EINVAL;
840 }
841
842 if (tb[NDA_DST] == NULL)
843 return -EINVAL;
844
845 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
846 if (err)
847 return err;
848
849 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
850 return -EAFNOSUPPORT;
851
852 spin_lock_bh(&vxlan->hash_lock);
853 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
854 port, vni, ifindex, ndm->ndm_flags);
855 spin_unlock_bh(&vxlan->hash_lock);
856
857 return err;
858 }
859
860 /* Delete entry (via netlink) */
861 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
862 struct net_device *dev,
863 const unsigned char *addr, u16 vid)
864 {
865 struct vxlan_dev *vxlan = netdev_priv(dev);
866 struct vxlan_fdb *f;
867 struct vxlan_rdst *rd = NULL;
868 union vxlan_addr ip;
869 __be16 port;
870 u32 vni, ifindex;
871 int err;
872
873 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
874 if (err)
875 return err;
876
877 err = -ENOENT;
878
879 spin_lock_bh(&vxlan->hash_lock);
880 f = vxlan_find_mac(vxlan, addr);
881 if (!f)
882 goto out;
883
884 if (!vxlan_addr_any(&ip)) {
885 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
886 if (!rd)
887 goto out;
888 }
889
890 err = 0;
891
892 /* remove a destination if it's not the only one on the list,
893 * otherwise destroy the fdb entry
894 */
895 if (rd && !list_is_singular(&f->remotes)) {
896 list_del_rcu(&rd->list);
897 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
898 kfree_rcu(rd, rcu);
899 goto out;
900 }
901
902 vxlan_fdb_destroy(vxlan, f);
903
904 out:
905 spin_unlock_bh(&vxlan->hash_lock);
906
907 return err;
908 }
909
910 /* Dump forwarding table */
911 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
912 struct net_device *dev,
913 struct net_device *filter_dev, int idx)
914 {
915 struct vxlan_dev *vxlan = netdev_priv(dev);
916 unsigned int h;
917
918 for (h = 0; h < FDB_HASH_SIZE; ++h) {
919 struct vxlan_fdb *f;
920 int err;
921
922 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
923 struct vxlan_rdst *rd;
924
925 list_for_each_entry_rcu(rd, &f->remotes, list) {
926 if (idx < cb->args[0])
927 goto skip;
928
929 err = vxlan_fdb_info(skb, vxlan, f,
930 NETLINK_CB(cb->skb).portid,
931 cb->nlh->nlmsg_seq,
932 RTM_NEWNEIGH,
933 NLM_F_MULTI, rd);
934 if (err < 0)
935 goto out;
936 skip:
937 ++idx;
938 }
939 }
940 }
941 out:
942 return idx;
943 }
944
945 /* Watch incoming packets to learn mapping between Ethernet address
946 * and Tunnel endpoint.
947 * Return true if packet is bogus and should be dropped.
948 */
949 static bool vxlan_snoop(struct net_device *dev,
950 union vxlan_addr *src_ip, const u8 *src_mac)
951 {
952 struct vxlan_dev *vxlan = netdev_priv(dev);
953 struct vxlan_fdb *f;
954
955 f = vxlan_find_mac(vxlan, src_mac);
956 if (likely(f)) {
957 struct vxlan_rdst *rdst = first_remote_rcu(f);
958
959 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
960 return false;
961
962 /* Don't migrate static entries, drop packets */
963 if (f->state & NUD_NOARP)
964 return true;
965
966 if (net_ratelimit())
967 netdev_info(dev,
968 "%pM migrated from %pIS to %pIS\n",
969 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
970
971 rdst->remote_ip = *src_ip;
972 f->updated = jiffies;
973 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
974 } else {
975 /* learned new entry */
976 spin_lock(&vxlan->hash_lock);
977
978 /* close off race between vxlan_flush and incoming packets */
979 if (netif_running(dev))
980 vxlan_fdb_create(vxlan, src_mac, src_ip,
981 NUD_REACHABLE,
982 NLM_F_EXCL|NLM_F_CREATE,
983 vxlan->cfg.dst_port,
984 vxlan->default_dst.remote_vni,
985 0, NTF_SELF);
986 spin_unlock(&vxlan->hash_lock);
987 }
988
989 return false;
990 }
991
992 /* See if multicast group is already in use by other ID */
993 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
994 {
995 struct vxlan_dev *vxlan;
996 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
997
998 /* The vxlan_sock is only used by dev, leaving group has
999 * no effect on other vxlan devices.
1000 */
1001 if (family == AF_INET && dev->vn4_sock &&
1002 atomic_read(&dev->vn4_sock->refcnt) == 1)
1003 return false;
1004 #if IS_ENABLED(CONFIG_IPV6)
1005 if (family == AF_INET6 && dev->vn6_sock &&
1006 atomic_read(&dev->vn6_sock->refcnt) == 1)
1007 return false;
1008 #endif
1009
1010 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1011 if (!netif_running(vxlan->dev) || vxlan == dev)
1012 continue;
1013
1014 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
1015 continue;
1016 #if IS_ENABLED(CONFIG_IPV6)
1017 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1018 continue;
1019 #endif
1020
1021 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1022 &dev->default_dst.remote_ip))
1023 continue;
1024
1025 if (vxlan->default_dst.remote_ifindex !=
1026 dev->default_dst.remote_ifindex)
1027 continue;
1028
1029 return true;
1030 }
1031
1032 return false;
1033 }
1034
1035 static void __vxlan_sock_release(struct vxlan_sock *vs)
1036 {
1037 struct vxlan_net *vn;
1038
1039 if (!vs)
1040 return;
1041 if (!atomic_dec_and_test(&vs->refcnt))
1042 return;
1043
1044 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1045 spin_lock(&vn->sock_lock);
1046 hlist_del_rcu(&vs->hlist);
1047 vxlan_notify_del_rx_port(vs);
1048 spin_unlock(&vn->sock_lock);
1049
1050 queue_work(vxlan_wq, &vs->del_work);
1051 }
1052
1053 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1054 {
1055 __vxlan_sock_release(vxlan->vn4_sock);
1056 #if IS_ENABLED(CONFIG_IPV6)
1057 __vxlan_sock_release(vxlan->vn6_sock);
1058 #endif
1059 }
1060
1061 /* Update multicast group membership when first VNI on
1062 * multicast address is brought up
1063 */
1064 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1065 {
1066 struct sock *sk;
1067 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1068 int ifindex = vxlan->default_dst.remote_ifindex;
1069 int ret = -EINVAL;
1070
1071 if (ip->sa.sa_family == AF_INET) {
1072 struct ip_mreqn mreq = {
1073 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1074 .imr_ifindex = ifindex,
1075 };
1076
1077 sk = vxlan->vn4_sock->sock->sk;
1078 lock_sock(sk);
1079 ret = ip_mc_join_group(sk, &mreq);
1080 release_sock(sk);
1081 #if IS_ENABLED(CONFIG_IPV6)
1082 } else {
1083 sk = vxlan->vn6_sock->sock->sk;
1084 lock_sock(sk);
1085 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1086 &ip->sin6.sin6_addr);
1087 release_sock(sk);
1088 #endif
1089 }
1090
1091 return ret;
1092 }
1093
1094 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1095 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1096 {
1097 struct sock *sk;
1098 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1099 int ifindex = vxlan->default_dst.remote_ifindex;
1100 int ret = -EINVAL;
1101
1102 if (ip->sa.sa_family == AF_INET) {
1103 struct ip_mreqn mreq = {
1104 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1105 .imr_ifindex = ifindex,
1106 };
1107
1108 sk = vxlan->vn4_sock->sock->sk;
1109 lock_sock(sk);
1110 ret = ip_mc_leave_group(sk, &mreq);
1111 release_sock(sk);
1112 #if IS_ENABLED(CONFIG_IPV6)
1113 } else {
1114 sk = vxlan->vn6_sock->sock->sk;
1115 lock_sock(sk);
1116 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1117 &ip->sin6.sin6_addr);
1118 release_sock(sk);
1119 #endif
1120 }
1121
1122 return ret;
1123 }
1124
1125 static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
1126 size_t hdrlen, u32 data, bool nopartial)
1127 {
1128 size_t start, offset, plen;
1129
1130 if (skb->remcsum_offload)
1131 return vh;
1132
1133 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1134 offset = start + ((data & VXLAN_RCO_UDP) ?
1135 offsetof(struct udphdr, check) :
1136 offsetof(struct tcphdr, check));
1137
1138 plen = hdrlen + offset + sizeof(u16);
1139
1140 if (!pskb_may_pull(skb, plen))
1141 return NULL;
1142
1143 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1144
1145 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1146 nopartial);
1147
1148 return vh;
1149 }
1150
1151 static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1152 struct vxlan_metadata *md, u32 vni,
1153 struct metadata_dst *tun_dst)
1154 {
1155 struct iphdr *oip = NULL;
1156 struct ipv6hdr *oip6 = NULL;
1157 struct vxlan_dev *vxlan;
1158 struct pcpu_sw_netstats *stats;
1159 union vxlan_addr saddr;
1160 int err = 0;
1161
1162 /* For flow based devices, map all packets to VNI 0 */
1163 if (vs->flags & VXLAN_F_COLLECT_METADATA)
1164 vni = 0;
1165
1166 /* Is this VNI defined? */
1167 vxlan = vxlan_vs_find_vni(vs, vni);
1168 if (!vxlan)
1169 goto drop;
1170
1171 skb_reset_mac_header(skb);
1172 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1173 skb->protocol = eth_type_trans(skb, vxlan->dev);
1174 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1175
1176 /* Ignore packet loops (and multicast echo) */
1177 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1178 goto drop;
1179
1180 /* Get data from the outer IP header */
1181 if (vxlan_get_sk_family(vs) == AF_INET) {
1182 oip = ip_hdr(skb);
1183 saddr.sin.sin_addr.s_addr = oip->saddr;
1184 saddr.sa.sa_family = AF_INET;
1185 #if IS_ENABLED(CONFIG_IPV6)
1186 } else {
1187 oip6 = ipv6_hdr(skb);
1188 saddr.sin6.sin6_addr = oip6->saddr;
1189 saddr.sa.sa_family = AF_INET6;
1190 #endif
1191 }
1192
1193 if (tun_dst) {
1194 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1195 tun_dst = NULL;
1196 }
1197
1198 if ((vxlan->flags & VXLAN_F_LEARN) &&
1199 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1200 goto drop;
1201
1202 skb_reset_network_header(skb);
1203 /* In flow-based mode, GBP is carried in dst_metadata */
1204 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
1205 skb->mark = md->gbp;
1206
1207 if (oip6)
1208 err = IP6_ECN_decapsulate(oip6, skb);
1209 if (oip)
1210 err = IP_ECN_decapsulate(oip, skb);
1211
1212 if (unlikely(err)) {
1213 if (log_ecn_error) {
1214 if (oip6)
1215 net_info_ratelimited("non-ECT from %pI6\n",
1216 &oip6->saddr);
1217 if (oip)
1218 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1219 &oip->saddr, oip->tos);
1220 }
1221 if (err > 1) {
1222 ++vxlan->dev->stats.rx_frame_errors;
1223 ++vxlan->dev->stats.rx_errors;
1224 goto drop;
1225 }
1226 }
1227
1228 stats = this_cpu_ptr(vxlan->dev->tstats);
1229 u64_stats_update_begin(&stats->syncp);
1230 stats->rx_packets++;
1231 stats->rx_bytes += skb->len;
1232 u64_stats_update_end(&stats->syncp);
1233
1234 gro_cells_receive(&vxlan->gro_cells, skb);
1235
1236 return;
1237 drop:
1238 if (tun_dst)
1239 dst_release((struct dst_entry *)tun_dst);
1240
1241 /* Consume bad packet */
1242 kfree_skb(skb);
1243 }
1244
1245 /* Callback from net/ipv4/udp.c to receive packets */
1246 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1247 {
1248 struct metadata_dst *tun_dst = NULL;
1249 struct vxlan_sock *vs;
1250 struct vxlanhdr *vxh;
1251 u32 flags, vni;
1252 struct vxlan_metadata _md;
1253 struct vxlan_metadata *md = &_md;
1254
1255 /* Need Vxlan and inner Ethernet header to be present */
1256 if (!pskb_may_pull(skb, VXLAN_HLEN))
1257 goto error;
1258
1259 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1260 flags = ntohl(vxh->vx_flags);
1261 vni = ntohl(vxh->vx_vni);
1262
1263 if (flags & VXLAN_HF_VNI) {
1264 flags &= ~VXLAN_HF_VNI;
1265 } else {
1266 /* VNI flag always required to be set */
1267 goto bad_flags;
1268 }
1269
1270 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
1271 goto drop;
1272 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1273
1274 vs = rcu_dereference_sk_user_data(sk);
1275 if (!vs)
1276 goto drop;
1277
1278 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
1279 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1280 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
1281 if (!vxh)
1282 goto drop;
1283
1284 flags &= ~VXLAN_HF_RCO;
1285 vni &= VXLAN_VNI_MASK;
1286 }
1287
1288 if (vxlan_collect_metadata(vs)) {
1289 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1290 cpu_to_be64(vni >> 8), sizeof(*md));
1291
1292 if (!tun_dst)
1293 goto drop;
1294
1295 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1296 } else {
1297 memset(md, 0, sizeof(*md));
1298 }
1299
1300 /* For backwards compatibility, only allow reserved fields to be
1301 * used by VXLAN extensions if explicitly requested.
1302 */
1303 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1304 struct vxlanhdr_gbp *gbp;
1305
1306 gbp = (struct vxlanhdr_gbp *)vxh;
1307 md->gbp = ntohs(gbp->policy_id);
1308
1309 if (tun_dst)
1310 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1311
1312 if (gbp->dont_learn)
1313 md->gbp |= VXLAN_GBP_DONT_LEARN;
1314
1315 if (gbp->policy_applied)
1316 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1317
1318 flags &= ~VXLAN_GBP_USED_BITS;
1319 }
1320
1321 if (flags || vni & ~VXLAN_VNI_MASK) {
1322 /* If there are any unprocessed flags remaining treat
1323 * this as a malformed packet. This behavior diverges from
1324 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1325 * in reserved fields are to be ignored. The approach here
1326 * maintains compatibility with previous stack code, and also
1327 * is more robust and provides a little more security in
1328 * adding extensions to VXLAN.
1329 */
1330
1331 goto bad_flags;
1332 }
1333
1334 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
1335 return 0;
1336
1337 drop:
1338 /* Consume bad packet */
1339 kfree_skb(skb);
1340 return 0;
1341
1342 bad_flags:
1343 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1344 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1345
1346 error:
1347 if (tun_dst)
1348 dst_release((struct dst_entry *)tun_dst);
1349
1350 /* Return non vxlan pkt */
1351 return 1;
1352 }
1353
1354 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1355 {
1356 struct vxlan_dev *vxlan = netdev_priv(dev);
1357 struct arphdr *parp;
1358 u8 *arpptr, *sha;
1359 __be32 sip, tip;
1360 struct neighbour *n;
1361
1362 if (dev->flags & IFF_NOARP)
1363 goto out;
1364
1365 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1366 dev->stats.tx_dropped++;
1367 goto out;
1368 }
1369 parp = arp_hdr(skb);
1370
1371 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1372 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1373 parp->ar_pro != htons(ETH_P_IP) ||
1374 parp->ar_op != htons(ARPOP_REQUEST) ||
1375 parp->ar_hln != dev->addr_len ||
1376 parp->ar_pln != 4)
1377 goto out;
1378 arpptr = (u8 *)parp + sizeof(struct arphdr);
1379 sha = arpptr;
1380 arpptr += dev->addr_len; /* sha */
1381 memcpy(&sip, arpptr, sizeof(sip));
1382 arpptr += sizeof(sip);
1383 arpptr += dev->addr_len; /* tha */
1384 memcpy(&tip, arpptr, sizeof(tip));
1385
1386 if (ipv4_is_loopback(tip) ||
1387 ipv4_is_multicast(tip))
1388 goto out;
1389
1390 n = neigh_lookup(&arp_tbl, &tip, dev);
1391
1392 if (n) {
1393 struct vxlan_fdb *f;
1394 struct sk_buff *reply;
1395
1396 if (!(n->nud_state & NUD_CONNECTED)) {
1397 neigh_release(n);
1398 goto out;
1399 }
1400
1401 f = vxlan_find_mac(vxlan, n->ha);
1402 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1403 /* bridge-local neighbor */
1404 neigh_release(n);
1405 goto out;
1406 }
1407
1408 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1409 n->ha, sha);
1410
1411 neigh_release(n);
1412
1413 if (reply == NULL)
1414 goto out;
1415
1416 skb_reset_mac_header(reply);
1417 __skb_pull(reply, skb_network_offset(reply));
1418 reply->ip_summed = CHECKSUM_UNNECESSARY;
1419 reply->pkt_type = PACKET_HOST;
1420
1421 if (netif_rx_ni(reply) == NET_RX_DROP)
1422 dev->stats.rx_dropped++;
1423 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1424 union vxlan_addr ipa = {
1425 .sin.sin_addr.s_addr = tip,
1426 .sin.sin_family = AF_INET,
1427 };
1428
1429 vxlan_ip_miss(dev, &ipa);
1430 }
1431 out:
1432 consume_skb(skb);
1433 return NETDEV_TX_OK;
1434 }
1435
1436 #if IS_ENABLED(CONFIG_IPV6)
1437 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1438 struct neighbour *n, bool isrouter)
1439 {
1440 struct net_device *dev = request->dev;
1441 struct sk_buff *reply;
1442 struct nd_msg *ns, *na;
1443 struct ipv6hdr *pip6;
1444 u8 *daddr;
1445 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1446 int ns_olen;
1447 int i, len;
1448
1449 if (dev == NULL)
1450 return NULL;
1451
1452 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1453 sizeof(*na) + na_olen + dev->needed_tailroom;
1454 reply = alloc_skb(len, GFP_ATOMIC);
1455 if (reply == NULL)
1456 return NULL;
1457
1458 reply->protocol = htons(ETH_P_IPV6);
1459 reply->dev = dev;
1460 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1461 skb_push(reply, sizeof(struct ethhdr));
1462 skb_set_mac_header(reply, 0);
1463
1464 ns = (struct nd_msg *)skb_transport_header(request);
1465
1466 daddr = eth_hdr(request)->h_source;
1467 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1468 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1469 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1470 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1471 break;
1472 }
1473 }
1474
1475 /* Ethernet header */
1476 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1477 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1478 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1479 reply->protocol = htons(ETH_P_IPV6);
1480
1481 skb_pull(reply, sizeof(struct ethhdr));
1482 skb_set_network_header(reply, 0);
1483 skb_put(reply, sizeof(struct ipv6hdr));
1484
1485 /* IPv6 header */
1486
1487 pip6 = ipv6_hdr(reply);
1488 memset(pip6, 0, sizeof(struct ipv6hdr));
1489 pip6->version = 6;
1490 pip6->priority = ipv6_hdr(request)->priority;
1491 pip6->nexthdr = IPPROTO_ICMPV6;
1492 pip6->hop_limit = 255;
1493 pip6->daddr = ipv6_hdr(request)->saddr;
1494 pip6->saddr = *(struct in6_addr *)n->primary_key;
1495
1496 skb_pull(reply, sizeof(struct ipv6hdr));
1497 skb_set_transport_header(reply, 0);
1498
1499 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1500
1501 /* Neighbor Advertisement */
1502 memset(na, 0, sizeof(*na)+na_olen);
1503 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1504 na->icmph.icmp6_router = isrouter;
1505 na->icmph.icmp6_override = 1;
1506 na->icmph.icmp6_solicited = 1;
1507 na->target = ns->target;
1508 ether_addr_copy(&na->opt[2], n->ha);
1509 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1510 na->opt[1] = na_olen >> 3;
1511
1512 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1513 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1514 csum_partial(na, sizeof(*na)+na_olen, 0));
1515
1516 pip6->payload_len = htons(sizeof(*na)+na_olen);
1517
1518 skb_push(reply, sizeof(struct ipv6hdr));
1519
1520 reply->ip_summed = CHECKSUM_UNNECESSARY;
1521
1522 return reply;
1523 }
1524
1525 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1526 {
1527 struct vxlan_dev *vxlan = netdev_priv(dev);
1528 struct nd_msg *msg;
1529 const struct ipv6hdr *iphdr;
1530 const struct in6_addr *saddr, *daddr;
1531 struct neighbour *n;
1532 struct inet6_dev *in6_dev;
1533
1534 in6_dev = __in6_dev_get(dev);
1535 if (!in6_dev)
1536 goto out;
1537
1538 iphdr = ipv6_hdr(skb);
1539 saddr = &iphdr->saddr;
1540 daddr = &iphdr->daddr;
1541
1542 msg = (struct nd_msg *)skb_transport_header(skb);
1543 if (msg->icmph.icmp6_code != 0 ||
1544 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1545 goto out;
1546
1547 if (ipv6_addr_loopback(daddr) ||
1548 ipv6_addr_is_multicast(&msg->target))
1549 goto out;
1550
1551 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1552
1553 if (n) {
1554 struct vxlan_fdb *f;
1555 struct sk_buff *reply;
1556
1557 if (!(n->nud_state & NUD_CONNECTED)) {
1558 neigh_release(n);
1559 goto out;
1560 }
1561
1562 f = vxlan_find_mac(vxlan, n->ha);
1563 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1564 /* bridge-local neighbor */
1565 neigh_release(n);
1566 goto out;
1567 }
1568
1569 reply = vxlan_na_create(skb, n,
1570 !!(f ? f->flags & NTF_ROUTER : 0));
1571
1572 neigh_release(n);
1573
1574 if (reply == NULL)
1575 goto out;
1576
1577 if (netif_rx_ni(reply) == NET_RX_DROP)
1578 dev->stats.rx_dropped++;
1579
1580 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1581 union vxlan_addr ipa = {
1582 .sin6.sin6_addr = msg->target,
1583 .sin6.sin6_family = AF_INET6,
1584 };
1585
1586 vxlan_ip_miss(dev, &ipa);
1587 }
1588
1589 out:
1590 consume_skb(skb);
1591 return NETDEV_TX_OK;
1592 }
1593 #endif
1594
1595 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1596 {
1597 struct vxlan_dev *vxlan = netdev_priv(dev);
1598 struct neighbour *n;
1599
1600 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1601 return false;
1602
1603 n = NULL;
1604 switch (ntohs(eth_hdr(skb)->h_proto)) {
1605 case ETH_P_IP:
1606 {
1607 struct iphdr *pip;
1608
1609 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1610 return false;
1611 pip = ip_hdr(skb);
1612 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1613 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1614 union vxlan_addr ipa = {
1615 .sin.sin_addr.s_addr = pip->daddr,
1616 .sin.sin_family = AF_INET,
1617 };
1618
1619 vxlan_ip_miss(dev, &ipa);
1620 return false;
1621 }
1622
1623 break;
1624 }
1625 #if IS_ENABLED(CONFIG_IPV6)
1626 case ETH_P_IPV6:
1627 {
1628 struct ipv6hdr *pip6;
1629
1630 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1631 return false;
1632 pip6 = ipv6_hdr(skb);
1633 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1634 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1635 union vxlan_addr ipa = {
1636 .sin6.sin6_addr = pip6->daddr,
1637 .sin6.sin6_family = AF_INET6,
1638 };
1639
1640 vxlan_ip_miss(dev, &ipa);
1641 return false;
1642 }
1643
1644 break;
1645 }
1646 #endif
1647 default:
1648 return false;
1649 }
1650
1651 if (n) {
1652 bool diff;
1653
1654 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1655 if (diff) {
1656 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1657 dev->addr_len);
1658 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1659 }
1660 neigh_release(n);
1661 return diff;
1662 }
1663
1664 return false;
1665 }
1666
1667 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1668 struct vxlan_metadata *md)
1669 {
1670 struct vxlanhdr_gbp *gbp;
1671
1672 if (!md->gbp)
1673 return;
1674
1675 gbp = (struct vxlanhdr_gbp *)vxh;
1676 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1677
1678 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1679 gbp->dont_learn = 1;
1680
1681 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1682 gbp->policy_applied = 1;
1683
1684 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1685 }
1686
1687 #if IS_ENABLED(CONFIG_IPV6)
1688 static int vxlan6_xmit_skb(struct dst_entry *dst, struct sock *sk,
1689 struct sk_buff *skb,
1690 struct net_device *dev, struct in6_addr *saddr,
1691 struct in6_addr *daddr, __u8 prio, __u8 ttl,
1692 __be16 src_port, __be16 dst_port, __be32 vni,
1693 struct vxlan_metadata *md, bool xnet, u32 vxflags)
1694 {
1695 struct vxlanhdr *vxh;
1696 int min_headroom;
1697 int err;
1698 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
1699 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1700 u16 hdrlen = sizeof(struct vxlanhdr);
1701
1702 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1703 skb->ip_summed == CHECKSUM_PARTIAL) {
1704 int csum_start = skb_checksum_start_offset(skb);
1705
1706 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1707 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1708 (skb->csum_offset == offsetof(struct udphdr, check) ||
1709 skb->csum_offset == offsetof(struct tcphdr, check))) {
1710 udp_sum = false;
1711 type |= SKB_GSO_TUNNEL_REMCSUM;
1712 }
1713 }
1714
1715 skb_scrub_packet(skb, xnet);
1716
1717 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1718 + VXLAN_HLEN + sizeof(struct ipv6hdr)
1719 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
1720
1721 /* Need space for new headers (invalidates iph ptr) */
1722 err = skb_cow_head(skb, min_headroom);
1723 if (unlikely(err)) {
1724 kfree_skb(skb);
1725 goto err;
1726 }
1727
1728 skb = vlan_hwaccel_push_inside(skb);
1729 if (WARN_ON(!skb)) {
1730 err = -ENOMEM;
1731 goto err;
1732 }
1733
1734 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1735 if (IS_ERR(skb)) {
1736 err = -EINVAL;
1737 goto err;
1738 }
1739
1740 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1741 vxh->vx_flags = htonl(VXLAN_HF_VNI);
1742 vxh->vx_vni = vni;
1743
1744 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1745 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1746 VXLAN_RCO_SHIFT;
1747
1748 if (skb->csum_offset == offsetof(struct udphdr, check))
1749 data |= VXLAN_RCO_UDP;
1750
1751 vxh->vx_vni |= htonl(data);
1752 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1753
1754 if (!skb_is_gso(skb)) {
1755 skb->ip_summed = CHECKSUM_NONE;
1756 skb->encapsulation = 0;
1757 }
1758 }
1759
1760 if (vxflags & VXLAN_F_GBP)
1761 vxlan_build_gbp_hdr(vxh, vxflags, md);
1762
1763 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1764
1765 udp_tunnel6_xmit_skb(dst, sk, skb, dev, saddr, daddr, prio,
1766 ttl, src_port, dst_port,
1767 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
1768 return 0;
1769 err:
1770 dst_release(dst);
1771 return err;
1772 }
1773 #endif
1774
1775 static int vxlan_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
1776 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
1777 __be16 src_port, __be16 dst_port, __be32 vni,
1778 struct vxlan_metadata *md, bool xnet, u32 vxflags)
1779 {
1780 struct vxlanhdr *vxh;
1781 int min_headroom;
1782 int err;
1783 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
1784 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1785 u16 hdrlen = sizeof(struct vxlanhdr);
1786
1787 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1788 skb->ip_summed == CHECKSUM_PARTIAL) {
1789 int csum_start = skb_checksum_start_offset(skb);
1790
1791 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1792 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1793 (skb->csum_offset == offsetof(struct udphdr, check) ||
1794 skb->csum_offset == offsetof(struct tcphdr, check))) {
1795 udp_sum = false;
1796 type |= SKB_GSO_TUNNEL_REMCSUM;
1797 }
1798 }
1799
1800 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
1801 + VXLAN_HLEN + sizeof(struct iphdr)
1802 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
1803
1804 /* Need space for new headers (invalidates iph ptr) */
1805 err = skb_cow_head(skb, min_headroom);
1806 if (unlikely(err)) {
1807 kfree_skb(skb);
1808 return err;
1809 }
1810
1811 skb = vlan_hwaccel_push_inside(skb);
1812 if (WARN_ON(!skb))
1813 return -ENOMEM;
1814
1815 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1816 if (IS_ERR(skb))
1817 return PTR_ERR(skb);
1818
1819 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1820 vxh->vx_flags = htonl(VXLAN_HF_VNI);
1821 vxh->vx_vni = vni;
1822
1823 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1824 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1825 VXLAN_RCO_SHIFT;
1826
1827 if (skb->csum_offset == offsetof(struct udphdr, check))
1828 data |= VXLAN_RCO_UDP;
1829
1830 vxh->vx_vni |= htonl(data);
1831 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1832
1833 if (!skb_is_gso(skb)) {
1834 skb->ip_summed = CHECKSUM_NONE;
1835 skb->encapsulation = 0;
1836 }
1837 }
1838
1839 if (vxflags & VXLAN_F_GBP)
1840 vxlan_build_gbp_hdr(vxh, vxflags, md);
1841
1842 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1843
1844 udp_tunnel_xmit_skb(rt, sk, skb, src, dst, tos, ttl, df,
1845 src_port, dst_port, xnet,
1846 !(vxflags & VXLAN_F_UDP_CSUM));
1847 return 0;
1848 }
1849
1850 #if IS_ENABLED(CONFIG_IPV6)
1851 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1852 struct sk_buff *skb, int oif,
1853 const struct in6_addr *daddr,
1854 struct in6_addr *saddr)
1855 {
1856 struct dst_entry *ndst;
1857 struct flowi6 fl6;
1858 int err;
1859
1860 memset(&fl6, 0, sizeof(fl6));
1861 fl6.flowi6_oif = oif;
1862 fl6.daddr = *daddr;
1863 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1864 fl6.flowi6_mark = skb->mark;
1865 fl6.flowi6_proto = IPPROTO_UDP;
1866
1867 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1868 vxlan->vn6_sock->sock->sk,
1869 &ndst, &fl6);
1870 if (err < 0)
1871 return ERR_PTR(err);
1872
1873 *saddr = fl6.saddr;
1874 return ndst;
1875 }
1876 #endif
1877
1878 /* Bypass encapsulation if the destination is local */
1879 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1880 struct vxlan_dev *dst_vxlan)
1881 {
1882 struct pcpu_sw_netstats *tx_stats, *rx_stats;
1883 union vxlan_addr loopback;
1884 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
1885 struct net_device *dev = skb->dev;
1886 int len = skb->len;
1887
1888 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1889 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1890 skb->pkt_type = PACKET_HOST;
1891 skb->encapsulation = 0;
1892 skb->dev = dst_vxlan->dev;
1893 __skb_pull(skb, skb_network_offset(skb));
1894
1895 if (remote_ip->sa.sa_family == AF_INET) {
1896 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1897 loopback.sa.sa_family = AF_INET;
1898 #if IS_ENABLED(CONFIG_IPV6)
1899 } else {
1900 loopback.sin6.sin6_addr = in6addr_loopback;
1901 loopback.sa.sa_family = AF_INET6;
1902 #endif
1903 }
1904
1905 if (dst_vxlan->flags & VXLAN_F_LEARN)
1906 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
1907
1908 u64_stats_update_begin(&tx_stats->syncp);
1909 tx_stats->tx_packets++;
1910 tx_stats->tx_bytes += len;
1911 u64_stats_update_end(&tx_stats->syncp);
1912
1913 if (netif_rx(skb) == NET_RX_SUCCESS) {
1914 u64_stats_update_begin(&rx_stats->syncp);
1915 rx_stats->rx_packets++;
1916 rx_stats->rx_bytes += len;
1917 u64_stats_update_end(&rx_stats->syncp);
1918 } else {
1919 dev->stats.rx_dropped++;
1920 }
1921 }
1922
1923 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1924 struct vxlan_rdst *rdst, bool did_rsc)
1925 {
1926 struct ip_tunnel_info *info;
1927 struct vxlan_dev *vxlan = netdev_priv(dev);
1928 struct sock *sk;
1929 struct rtable *rt = NULL;
1930 const struct iphdr *old_iph;
1931 struct flowi4 fl4;
1932 union vxlan_addr *dst;
1933 union vxlan_addr remote_ip;
1934 struct vxlan_metadata _md;
1935 struct vxlan_metadata *md = &_md;
1936 __be16 src_port = 0, dst_port;
1937 u32 vni;
1938 __be16 df = 0;
1939 __u8 tos, ttl;
1940 int err;
1941 u32 flags = vxlan->flags;
1942
1943 info = skb_tunnel_info(skb);
1944
1945 if (rdst) {
1946 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
1947 vni = rdst->remote_vni;
1948 dst = &rdst->remote_ip;
1949 } else {
1950 if (!info) {
1951 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1952 dev->name);
1953 goto drop;
1954 }
1955 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
1956 vni = be64_to_cpu(info->key.tun_id);
1957 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1958 if (remote_ip.sa.sa_family == AF_INET)
1959 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1960 else
1961 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
1962 dst = &remote_ip;
1963 }
1964
1965 if (vxlan_addr_any(dst)) {
1966 if (did_rsc) {
1967 /* short-circuited back to local bridge */
1968 vxlan_encap_bypass(skb, vxlan, vxlan);
1969 return;
1970 }
1971 goto drop;
1972 }
1973
1974 old_iph = ip_hdr(skb);
1975
1976 ttl = vxlan->cfg.ttl;
1977 if (!ttl && vxlan_addr_multicast(dst))
1978 ttl = 1;
1979
1980 tos = vxlan->cfg.tos;
1981 if (tos == 1)
1982 tos = ip_tunnel_get_dsfield(old_iph, skb);
1983
1984 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1985 vxlan->cfg.port_max, true);
1986
1987 if (info) {
1988 ttl = info->key.ttl;
1989 tos = info->key.tos;
1990
1991 if (info->options_len)
1992 md = ip_tunnel_info_opts(info);
1993 } else {
1994 md->gbp = skb->mark;
1995 }
1996
1997 if (dst->sa.sa_family == AF_INET) {
1998 if (!vxlan->vn4_sock)
1999 goto drop;
2000 sk = vxlan->vn4_sock->sock->sk;
2001
2002 if (info) {
2003 if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2004 df = htons(IP_DF);
2005
2006 if (info->key.tun_flags & TUNNEL_CSUM)
2007 flags |= VXLAN_F_UDP_CSUM;
2008 else
2009 flags &= ~VXLAN_F_UDP_CSUM;
2010 }
2011
2012 memset(&fl4, 0, sizeof(fl4));
2013 fl4.flowi4_oif = rdst ? rdst->remote_ifindex : 0;
2014 fl4.flowi4_tos = RT_TOS(tos);
2015 fl4.flowi4_mark = skb->mark;
2016 fl4.flowi4_proto = IPPROTO_UDP;
2017 fl4.daddr = dst->sin.sin_addr.s_addr;
2018 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
2019
2020 rt = ip_route_output_key(vxlan->net, &fl4);
2021 if (IS_ERR(rt)) {
2022 netdev_dbg(dev, "no route to %pI4\n",
2023 &dst->sin.sin_addr.s_addr);
2024 dev->stats.tx_carrier_errors++;
2025 goto tx_error;
2026 }
2027
2028 if (rt->dst.dev == dev) {
2029 netdev_dbg(dev, "circular route to %pI4\n",
2030 &dst->sin.sin_addr.s_addr);
2031 dev->stats.collisions++;
2032 goto rt_tx_error;
2033 }
2034
2035 /* Bypass encapsulation if the destination is local */
2036 if (rt->rt_flags & RTCF_LOCAL &&
2037 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2038 struct vxlan_dev *dst_vxlan;
2039
2040 ip_rt_put(rt);
2041 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2042 dst->sa.sa_family, dst_port,
2043 vxlan->flags);
2044 if (!dst_vxlan)
2045 goto tx_error;
2046 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2047 return;
2048 }
2049
2050 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2051 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2052 err = vxlan_xmit_skb(rt, sk, skb, fl4.saddr,
2053 dst->sin.sin_addr.s_addr, tos, ttl, df,
2054 src_port, dst_port, htonl(vni << 8), md,
2055 !net_eq(vxlan->net, dev_net(vxlan->dev)),
2056 flags);
2057 if (err < 0) {
2058 /* skb is already freed. */
2059 skb = NULL;
2060 goto rt_tx_error;
2061 }
2062 #if IS_ENABLED(CONFIG_IPV6)
2063 } else {
2064 struct dst_entry *ndst;
2065 struct in6_addr saddr;
2066 u32 rt6i_flags;
2067
2068 if (!vxlan->vn6_sock)
2069 goto drop;
2070 sk = vxlan->vn6_sock->sock->sk;
2071
2072 ndst = vxlan6_get_route(vxlan, skb,
2073 rdst ? rdst->remote_ifindex : 0,
2074 &dst->sin6.sin6_addr, &saddr);
2075 if (IS_ERR(ndst)) {
2076 netdev_dbg(dev, "no route to %pI6\n",
2077 &dst->sin6.sin6_addr);
2078 dev->stats.tx_carrier_errors++;
2079 goto tx_error;
2080 }
2081
2082 if (ndst->dev == dev) {
2083 netdev_dbg(dev, "circular route to %pI6\n",
2084 &dst->sin6.sin6_addr);
2085 dst_release(ndst);
2086 dev->stats.collisions++;
2087 goto tx_error;
2088 }
2089
2090 /* Bypass encapsulation if the destination is local */
2091 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2092 if (rt6i_flags & RTF_LOCAL &&
2093 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2094 struct vxlan_dev *dst_vxlan;
2095
2096 dst_release(ndst);
2097 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
2098 dst->sa.sa_family, dst_port,
2099 vxlan->flags);
2100 if (!dst_vxlan)
2101 goto tx_error;
2102 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2103 return;
2104 }
2105
2106 if (info) {
2107 if (info->key.tun_flags & TUNNEL_CSUM)
2108 flags &= ~VXLAN_F_UDP_ZERO_CSUM6_TX;
2109 else
2110 flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2111 }
2112
2113 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2114 err = vxlan6_xmit_skb(ndst, sk, skb, dev, &saddr, &dst->sin6.sin6_addr,
2115 0, ttl, src_port, dst_port, htonl(vni << 8), md,
2116 !net_eq(vxlan->net, dev_net(vxlan->dev)),
2117 flags);
2118 #endif
2119 }
2120
2121 return;
2122
2123 drop:
2124 dev->stats.tx_dropped++;
2125 goto tx_free;
2126
2127 rt_tx_error:
2128 ip_rt_put(rt);
2129 tx_error:
2130 dev->stats.tx_errors++;
2131 tx_free:
2132 dev_kfree_skb(skb);
2133 }
2134
2135 /* Transmit local packets over Vxlan
2136 *
2137 * Outer IP header inherits ECN and DF from inner header.
2138 * Outer UDP destination is the VXLAN assigned port.
2139 * source port is based on hash of flow
2140 */
2141 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2142 {
2143 struct vxlan_dev *vxlan = netdev_priv(dev);
2144 const struct ip_tunnel_info *info;
2145 struct ethhdr *eth;
2146 bool did_rsc = false;
2147 struct vxlan_rdst *rdst, *fdst = NULL;
2148 struct vxlan_fdb *f;
2149
2150 info = skb_tunnel_info(skb);
2151
2152 skb_reset_mac_header(skb);
2153 eth = eth_hdr(skb);
2154
2155 if ((vxlan->flags & VXLAN_F_PROXY)) {
2156 if (ntohs(eth->h_proto) == ETH_P_ARP)
2157 return arp_reduce(dev, skb);
2158 #if IS_ENABLED(CONFIG_IPV6)
2159 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2160 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2161 + sizeof(struct nd_msg)) &&
2162 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2163 struct nd_msg *msg;
2164
2165 msg = (struct nd_msg *)skb_transport_header(skb);
2166 if (msg->icmph.icmp6_code == 0 &&
2167 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2168 return neigh_reduce(dev, skb);
2169 }
2170 eth = eth_hdr(skb);
2171 #endif
2172 }
2173
2174 if (vxlan->flags & VXLAN_F_COLLECT_METADATA &&
2175 info && info->mode & IP_TUNNEL_INFO_TX) {
2176 vxlan_xmit_one(skb, dev, NULL, false);
2177 return NETDEV_TX_OK;
2178 }
2179
2180 f = vxlan_find_mac(vxlan, eth->h_dest);
2181 did_rsc = false;
2182
2183 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
2184 (ntohs(eth->h_proto) == ETH_P_IP ||
2185 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2186 did_rsc = route_shortcircuit(dev, skb);
2187 if (did_rsc)
2188 f = vxlan_find_mac(vxlan, eth->h_dest);
2189 }
2190
2191 if (f == NULL) {
2192 f = vxlan_find_mac(vxlan, all_zeros_mac);
2193 if (f == NULL) {
2194 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2195 !is_multicast_ether_addr(eth->h_dest))
2196 vxlan_fdb_miss(vxlan, eth->h_dest);
2197
2198 dev->stats.tx_dropped++;
2199 kfree_skb(skb);
2200 return NETDEV_TX_OK;
2201 }
2202 }
2203
2204 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2205 struct sk_buff *skb1;
2206
2207 if (!fdst) {
2208 fdst = rdst;
2209 continue;
2210 }
2211 skb1 = skb_clone(skb, GFP_ATOMIC);
2212 if (skb1)
2213 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
2214 }
2215
2216 if (fdst)
2217 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2218 else
2219 kfree_skb(skb);
2220 return NETDEV_TX_OK;
2221 }
2222
2223 /* Walk the forwarding table and purge stale entries */
2224 static void vxlan_cleanup(unsigned long arg)
2225 {
2226 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2227 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2228 unsigned int h;
2229
2230 if (!netif_running(vxlan->dev))
2231 return;
2232
2233 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2234 struct hlist_node *p, *n;
2235
2236 spin_lock_bh(&vxlan->hash_lock);
2237 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2238 struct vxlan_fdb *f
2239 = container_of(p, struct vxlan_fdb, hlist);
2240 unsigned long timeout;
2241
2242 if (f->state & NUD_PERMANENT)
2243 continue;
2244
2245 timeout = f->used + vxlan->cfg.age_interval * HZ;
2246 if (time_before_eq(timeout, jiffies)) {
2247 netdev_dbg(vxlan->dev,
2248 "garbage collect %pM\n",
2249 f->eth_addr);
2250 f->state = NUD_STALE;
2251 vxlan_fdb_destroy(vxlan, f);
2252 } else if (time_before(timeout, next_timer))
2253 next_timer = timeout;
2254 }
2255 spin_unlock_bh(&vxlan->hash_lock);
2256 }
2257
2258 mod_timer(&vxlan->age_timer, next_timer);
2259 }
2260
2261 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2262 {
2263 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2264 __u32 vni = vxlan->default_dst.remote_vni;
2265
2266 spin_lock(&vn->sock_lock);
2267 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
2268 spin_unlock(&vn->sock_lock);
2269 }
2270
2271 /* Setup stats when device is created */
2272 static int vxlan_init(struct net_device *dev)
2273 {
2274 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2275 if (!dev->tstats)
2276 return -ENOMEM;
2277
2278 return 0;
2279 }
2280
2281 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
2282 {
2283 struct vxlan_fdb *f;
2284
2285 spin_lock_bh(&vxlan->hash_lock);
2286 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2287 if (f)
2288 vxlan_fdb_destroy(vxlan, f);
2289 spin_unlock_bh(&vxlan->hash_lock);
2290 }
2291
2292 static void vxlan_uninit(struct net_device *dev)
2293 {
2294 struct vxlan_dev *vxlan = netdev_priv(dev);
2295
2296 vxlan_fdb_delete_default(vxlan);
2297
2298 free_percpu(dev->tstats);
2299 }
2300
2301 /* Start ageing timer and join group when device is brought up */
2302 static int vxlan_open(struct net_device *dev)
2303 {
2304 struct vxlan_dev *vxlan = netdev_priv(dev);
2305 int ret;
2306
2307 ret = vxlan_sock_add(vxlan);
2308 if (ret < 0)
2309 return ret;
2310
2311 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2312 ret = vxlan_igmp_join(vxlan);
2313 if (ret == -EADDRINUSE)
2314 ret = 0;
2315 if (ret) {
2316 vxlan_sock_release(vxlan);
2317 return ret;
2318 }
2319 }
2320
2321 if (vxlan->cfg.age_interval)
2322 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2323
2324 return ret;
2325 }
2326
2327 /* Purge the forwarding table */
2328 static void vxlan_flush(struct vxlan_dev *vxlan)
2329 {
2330 unsigned int h;
2331
2332 spin_lock_bh(&vxlan->hash_lock);
2333 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2334 struct hlist_node *p, *n;
2335 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2336 struct vxlan_fdb *f
2337 = container_of(p, struct vxlan_fdb, hlist);
2338 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2339 if (!is_zero_ether_addr(f->eth_addr))
2340 vxlan_fdb_destroy(vxlan, f);
2341 }
2342 }
2343 spin_unlock_bh(&vxlan->hash_lock);
2344 }
2345
2346 /* Cleanup timer and forwarding table on shutdown */
2347 static int vxlan_stop(struct net_device *dev)
2348 {
2349 struct vxlan_dev *vxlan = netdev_priv(dev);
2350 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2351 int ret = 0;
2352
2353 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2354 !vxlan_group_used(vn, vxlan))
2355 ret = vxlan_igmp_leave(vxlan);
2356
2357 del_timer_sync(&vxlan->age_timer);
2358
2359 vxlan_flush(vxlan);
2360 vxlan_sock_release(vxlan);
2361
2362 return ret;
2363 }
2364
2365 /* Stub, nothing needs to be done. */
2366 static void vxlan_set_multicast_list(struct net_device *dev)
2367 {
2368 }
2369
2370 static int __vxlan_change_mtu(struct net_device *dev,
2371 struct net_device *lowerdev,
2372 struct vxlan_rdst *dst, int new_mtu, bool strict)
2373 {
2374 int max_mtu = IP_MAX_MTU;
2375
2376 if (lowerdev)
2377 max_mtu = lowerdev->mtu;
2378
2379 if (dst->remote_ip.sa.sa_family == AF_INET6)
2380 max_mtu -= VXLAN6_HEADROOM;
2381 else
2382 max_mtu -= VXLAN_HEADROOM;
2383
2384 if (new_mtu < 68)
2385 return -EINVAL;
2386
2387 if (new_mtu > max_mtu) {
2388 if (strict)
2389 return -EINVAL;
2390
2391 new_mtu = max_mtu;
2392 }
2393
2394 dev->mtu = new_mtu;
2395 return 0;
2396 }
2397
2398 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2399 {
2400 struct vxlan_dev *vxlan = netdev_priv(dev);
2401 struct vxlan_rdst *dst = &vxlan->default_dst;
2402 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2403 dst->remote_ifindex);
2404 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
2405 }
2406
2407 static int egress_ipv4_tun_info(struct net_device *dev, struct sk_buff *skb,
2408 struct ip_tunnel_info *info,
2409 __be16 sport, __be16 dport)
2410 {
2411 struct vxlan_dev *vxlan = netdev_priv(dev);
2412 struct rtable *rt;
2413 struct flowi4 fl4;
2414
2415 memset(&fl4, 0, sizeof(fl4));
2416 fl4.flowi4_tos = RT_TOS(info->key.tos);
2417 fl4.flowi4_mark = skb->mark;
2418 fl4.flowi4_proto = IPPROTO_UDP;
2419 fl4.daddr = info->key.u.ipv4.dst;
2420
2421 rt = ip_route_output_key(vxlan->net, &fl4);
2422 if (IS_ERR(rt))
2423 return PTR_ERR(rt);
2424 ip_rt_put(rt);
2425
2426 info->key.u.ipv4.src = fl4.saddr;
2427 info->key.tp_src = sport;
2428 info->key.tp_dst = dport;
2429 return 0;
2430 }
2431
2432 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2433 {
2434 struct vxlan_dev *vxlan = netdev_priv(dev);
2435 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2436 __be16 sport, dport;
2437
2438 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2439 vxlan->cfg.port_max, true);
2440 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2441
2442 if (ip_tunnel_info_af(info) == AF_INET) {
2443 if (!vxlan->vn4_sock)
2444 return -EINVAL;
2445 return egress_ipv4_tun_info(dev, skb, info, sport, dport);
2446 } else {
2447 #if IS_ENABLED(CONFIG_IPV6)
2448 struct dst_entry *ndst;
2449
2450 if (!vxlan->vn6_sock)
2451 return -EINVAL;
2452 ndst = vxlan6_get_route(vxlan, skb, 0,
2453 &info->key.u.ipv6.dst,
2454 &info->key.u.ipv6.src);
2455 if (IS_ERR(ndst))
2456 return PTR_ERR(ndst);
2457 dst_release(ndst);
2458
2459 info->key.tp_src = sport;
2460 info->key.tp_dst = dport;
2461 #else /* !CONFIG_IPV6 */
2462 return -EPFNOSUPPORT;
2463 #endif
2464 }
2465 return 0;
2466 }
2467
2468 static const struct net_device_ops vxlan_netdev_ops = {
2469 .ndo_init = vxlan_init,
2470 .ndo_uninit = vxlan_uninit,
2471 .ndo_open = vxlan_open,
2472 .ndo_stop = vxlan_stop,
2473 .ndo_start_xmit = vxlan_xmit,
2474 .ndo_get_stats64 = ip_tunnel_get_stats64,
2475 .ndo_set_rx_mode = vxlan_set_multicast_list,
2476 .ndo_change_mtu = vxlan_change_mtu,
2477 .ndo_validate_addr = eth_validate_addr,
2478 .ndo_set_mac_address = eth_mac_addr,
2479 .ndo_fdb_add = vxlan_fdb_add,
2480 .ndo_fdb_del = vxlan_fdb_delete,
2481 .ndo_fdb_dump = vxlan_fdb_dump,
2482 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2483 };
2484
2485 /* Info for udev, that this is a virtual tunnel endpoint */
2486 static struct device_type vxlan_type = {
2487 .name = "vxlan",
2488 };
2489
2490 /* Calls the ndo_add_vxlan_port of the caller in order to
2491 * supply the listening VXLAN udp ports. Callers are expected
2492 * to implement the ndo_add_vxlan_port.
2493 */
2494 void vxlan_get_rx_port(struct net_device *dev)
2495 {
2496 struct vxlan_sock *vs;
2497 struct net *net = dev_net(dev);
2498 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2499 sa_family_t sa_family;
2500 __be16 port;
2501 unsigned int i;
2502
2503 spin_lock(&vn->sock_lock);
2504 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2505 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2506 port = inet_sk(vs->sock->sk)->inet_sport;
2507 sa_family = vxlan_get_sk_family(vs);
2508 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2509 port);
2510 }
2511 }
2512 spin_unlock(&vn->sock_lock);
2513 }
2514 EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2515
2516 /* Initialize the device structure. */
2517 static void vxlan_setup(struct net_device *dev)
2518 {
2519 struct vxlan_dev *vxlan = netdev_priv(dev);
2520 unsigned int h;
2521
2522 eth_hw_addr_random(dev);
2523 ether_setup(dev);
2524
2525 dev->netdev_ops = &vxlan_netdev_ops;
2526 dev->destructor = free_netdev;
2527 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2528
2529 dev->features |= NETIF_F_LLTX;
2530 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2531 dev->features |= NETIF_F_RXCSUM;
2532 dev->features |= NETIF_F_GSO_SOFTWARE;
2533
2534 dev->vlan_features = dev->features;
2535 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2536 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2537 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2538 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
2539 netif_keep_dst(dev);
2540 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
2541
2542 INIT_LIST_HEAD(&vxlan->next);
2543 spin_lock_init(&vxlan->hash_lock);
2544
2545 init_timer_deferrable(&vxlan->age_timer);
2546 vxlan->age_timer.function = vxlan_cleanup;
2547 vxlan->age_timer.data = (unsigned long) vxlan;
2548
2549 vxlan->cfg.dst_port = htons(vxlan_port);
2550
2551 vxlan->dev = dev;
2552
2553 gro_cells_init(&vxlan->gro_cells, dev);
2554
2555 for (h = 0; h < FDB_HASH_SIZE; ++h)
2556 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2557 }
2558
2559 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2560 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2561 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2562 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2563 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2564 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2565 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2566 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2567 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2568 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2569 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2570 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2571 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2572 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2573 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2574 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2575 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2576 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
2577 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2578 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2579 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2580 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
2581 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2582 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2583 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
2584 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
2585 };
2586
2587 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2588 {
2589 if (tb[IFLA_ADDRESS]) {
2590 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2591 pr_debug("invalid link address (not ethernet)\n");
2592 return -EINVAL;
2593 }
2594
2595 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2596 pr_debug("invalid all zero ethernet address\n");
2597 return -EADDRNOTAVAIL;
2598 }
2599 }
2600
2601 if (!data)
2602 return -EINVAL;
2603
2604 if (data[IFLA_VXLAN_ID]) {
2605 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2606 if (id >= VXLAN_VID_MASK)
2607 return -ERANGE;
2608 }
2609
2610 if (data[IFLA_VXLAN_PORT_RANGE]) {
2611 const struct ifla_vxlan_port_range *p
2612 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2613
2614 if (ntohs(p->high) < ntohs(p->low)) {
2615 pr_debug("port range %u .. %u not valid\n",
2616 ntohs(p->low), ntohs(p->high));
2617 return -EINVAL;
2618 }
2619 }
2620
2621 return 0;
2622 }
2623
2624 static void vxlan_get_drvinfo(struct net_device *netdev,
2625 struct ethtool_drvinfo *drvinfo)
2626 {
2627 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2628 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2629 }
2630
2631 static const struct ethtool_ops vxlan_ethtool_ops = {
2632 .get_drvinfo = vxlan_get_drvinfo,
2633 .get_link = ethtool_op_get_link,
2634 };
2635
2636 static void vxlan_del_work(struct work_struct *work)
2637 {
2638 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
2639 udp_tunnel_sock_release(vs->sock);
2640 kfree_rcu(vs, rcu);
2641 }
2642
2643 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2644 __be16 port, u32 flags)
2645 {
2646 struct socket *sock;
2647 struct udp_port_cfg udp_conf;
2648 int err;
2649
2650 memset(&udp_conf, 0, sizeof(udp_conf));
2651
2652 if (ipv6) {
2653 udp_conf.family = AF_INET6;
2654 udp_conf.use_udp6_rx_checksums =
2655 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2656 udp_conf.ipv6_v6only = 1;
2657 } else {
2658 udp_conf.family = AF_INET;
2659 }
2660
2661 udp_conf.local_udp_port = port;
2662
2663 /* Open UDP socket */
2664 err = udp_sock_create(net, &udp_conf, &sock);
2665 if (err < 0)
2666 return ERR_PTR(err);
2667
2668 return sock;
2669 }
2670
2671 /* Create new listen socket if needed */
2672 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2673 __be16 port, u32 flags)
2674 {
2675 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2676 struct vxlan_sock *vs;
2677 struct socket *sock;
2678 unsigned int h;
2679 struct udp_tunnel_sock_cfg tunnel_cfg;
2680
2681 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
2682 if (!vs)
2683 return ERR_PTR(-ENOMEM);
2684
2685 for (h = 0; h < VNI_HASH_SIZE; ++h)
2686 INIT_HLIST_HEAD(&vs->vni_list[h]);
2687
2688 INIT_WORK(&vs->del_work, vxlan_del_work);
2689
2690 sock = vxlan_create_sock(net, ipv6, port, flags);
2691 if (IS_ERR(sock)) {
2692 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2693 PTR_ERR(sock));
2694 kfree(vs);
2695 return ERR_CAST(sock);
2696 }
2697
2698 vs->sock = sock;
2699 atomic_set(&vs->refcnt, 1);
2700 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
2701
2702 /* Initialize the vxlan udp offloads structure */
2703 vs->udp_offloads.port = port;
2704 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2705 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2706
2707 spin_lock(&vn->sock_lock);
2708 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
2709 vxlan_notify_add_rx_port(vs);
2710 spin_unlock(&vn->sock_lock);
2711
2712 /* Mark socket as an encapsulation socket. */
2713 tunnel_cfg.sk_user_data = vs;
2714 tunnel_cfg.encap_type = 1;
2715 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2716 tunnel_cfg.encap_destroy = NULL;
2717
2718 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
2719
2720 return vs;
2721 }
2722
2723 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
2724 {
2725 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2726 struct vxlan_sock *vs = NULL;
2727
2728 if (!vxlan->cfg.no_share) {
2729 spin_lock(&vn->sock_lock);
2730 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2731 vxlan->cfg.dst_port, vxlan->flags);
2732 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
2733 spin_unlock(&vn->sock_lock);
2734 return -EBUSY;
2735 }
2736 spin_unlock(&vn->sock_lock);
2737 }
2738 if (!vs)
2739 vs = vxlan_socket_create(vxlan->net, ipv6,
2740 vxlan->cfg.dst_port, vxlan->flags);
2741 if (IS_ERR(vs))
2742 return PTR_ERR(vs);
2743 #if IS_ENABLED(CONFIG_IPV6)
2744 if (ipv6)
2745 vxlan->vn6_sock = vs;
2746 else
2747 #endif
2748 vxlan->vn4_sock = vs;
2749 vxlan_vs_add_dev(vs, vxlan);
2750 return 0;
2751 }
2752
2753 static int vxlan_sock_add(struct vxlan_dev *vxlan)
2754 {
2755 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2756 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2757 int ret = 0;
2758
2759 vxlan->vn4_sock = NULL;
2760 #if IS_ENABLED(CONFIG_IPV6)
2761 vxlan->vn6_sock = NULL;
2762 if (ipv6 || metadata)
2763 ret = __vxlan_sock_add(vxlan, true);
2764 #endif
2765 if (!ret && (!ipv6 || metadata))
2766 ret = __vxlan_sock_add(vxlan, false);
2767 if (ret < 0)
2768 vxlan_sock_release(vxlan);
2769 return ret;
2770 }
2771
2772 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2773 struct vxlan_config *conf)
2774 {
2775 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
2776 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
2777 struct vxlan_rdst *dst = &vxlan->default_dst;
2778 unsigned short needed_headroom = ETH_HLEN;
2779 int err;
2780 bool use_ipv6 = false;
2781 __be16 default_port = vxlan->cfg.dst_port;
2782 struct net_device *lowerdev = NULL;
2783
2784 vxlan->net = src_net;
2785
2786 dst->remote_vni = conf->vni;
2787
2788 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
2789
2790 /* Unless IPv6 is explicitly requested, assume IPv4 */
2791 if (!dst->remote_ip.sa.sa_family)
2792 dst->remote_ip.sa.sa_family = AF_INET;
2793
2794 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
2795 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2796 if (!IS_ENABLED(CONFIG_IPV6))
2797 return -EPFNOSUPPORT;
2798 use_ipv6 = true;
2799 vxlan->flags |= VXLAN_F_IPV6;
2800 }
2801
2802 if (conf->remote_ifindex) {
2803 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
2804 dst->remote_ifindex = conf->remote_ifindex;
2805
2806 if (!lowerdev) {
2807 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
2808 return -ENODEV;
2809 }
2810
2811 #if IS_ENABLED(CONFIG_IPV6)
2812 if (use_ipv6) {
2813 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2814 if (idev && idev->cnf.disable_ipv6) {
2815 pr_info("IPv6 is disabled via sysctl\n");
2816 return -EPERM;
2817 }
2818 }
2819 #endif
2820
2821 if (!conf->mtu)
2822 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2823
2824 needed_headroom = lowerdev->hard_header_len;
2825 }
2826
2827 if (conf->mtu) {
2828 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
2829 if (err)
2830 return err;
2831 }
2832
2833 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2834 needed_headroom += VXLAN6_HEADROOM;
2835 else
2836 needed_headroom += VXLAN_HEADROOM;
2837 dev->needed_headroom = needed_headroom;
2838
2839 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2840 if (!vxlan->cfg.dst_port)
2841 vxlan->cfg.dst_port = default_port;
2842 vxlan->flags |= conf->flags;
2843
2844 if (!vxlan->cfg.age_interval)
2845 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2846
2847 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2848 if (tmp->cfg.vni == conf->vni &&
2849 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2850 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2851 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2852 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2853 (vxlan->flags & VXLAN_F_RCV_FLAGS))
2854 return -EEXIST;
2855 }
2856
2857 dev->ethtool_ops = &vxlan_ethtool_ops;
2858
2859 /* create an fdb entry for a valid default destination */
2860 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2861 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2862 &vxlan->default_dst.remote_ip,
2863 NUD_REACHABLE|NUD_PERMANENT,
2864 NLM_F_EXCL|NLM_F_CREATE,
2865 vxlan->cfg.dst_port,
2866 vxlan->default_dst.remote_vni,
2867 vxlan->default_dst.remote_ifindex,
2868 NTF_SELF);
2869 if (err)
2870 return err;
2871 }
2872
2873 err = register_netdevice(dev);
2874 if (err) {
2875 vxlan_fdb_delete_default(vxlan);
2876 return err;
2877 }
2878
2879 list_add(&vxlan->next, &vn->vxlan_list);
2880
2881 return 0;
2882 }
2883
2884 struct net_device *vxlan_dev_create(struct net *net, const char *name,
2885 u8 name_assign_type, struct vxlan_config *conf)
2886 {
2887 struct nlattr *tb[IFLA_MAX+1];
2888 struct net_device *dev;
2889 int err;
2890
2891 memset(&tb, 0, sizeof(tb));
2892
2893 dev = rtnl_create_link(net, name, name_assign_type,
2894 &vxlan_link_ops, tb);
2895 if (IS_ERR(dev))
2896 return dev;
2897
2898 err = vxlan_dev_configure(net, dev, conf);
2899 if (err < 0) {
2900 free_netdev(dev);
2901 return ERR_PTR(err);
2902 }
2903
2904 return dev;
2905 }
2906 EXPORT_SYMBOL_GPL(vxlan_dev_create);
2907
2908 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2909 struct nlattr *tb[], struct nlattr *data[])
2910 {
2911 struct vxlan_config conf;
2912 int err;
2913
2914 memset(&conf, 0, sizeof(conf));
2915
2916 if (data[IFLA_VXLAN_ID])
2917 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2918
2919 if (data[IFLA_VXLAN_GROUP]) {
2920 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2921 } else if (data[IFLA_VXLAN_GROUP6]) {
2922 if (!IS_ENABLED(CONFIG_IPV6))
2923 return -EPFNOSUPPORT;
2924
2925 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2926 conf.remote_ip.sa.sa_family = AF_INET6;
2927 }
2928
2929 if (data[IFLA_VXLAN_LOCAL]) {
2930 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2931 conf.saddr.sa.sa_family = AF_INET;
2932 } else if (data[IFLA_VXLAN_LOCAL6]) {
2933 if (!IS_ENABLED(CONFIG_IPV6))
2934 return -EPFNOSUPPORT;
2935
2936 /* TODO: respect scope id */
2937 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2938 conf.saddr.sa.sa_family = AF_INET6;
2939 }
2940
2941 if (data[IFLA_VXLAN_LINK])
2942 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2943
2944 if (data[IFLA_VXLAN_TOS])
2945 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2946
2947 if (data[IFLA_VXLAN_TTL])
2948 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2949
2950 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2951 conf.flags |= VXLAN_F_LEARN;
2952
2953 if (data[IFLA_VXLAN_AGEING])
2954 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2955
2956 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2957 conf.flags |= VXLAN_F_PROXY;
2958
2959 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2960 conf.flags |= VXLAN_F_RSC;
2961
2962 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2963 conf.flags |= VXLAN_F_L2MISS;
2964
2965 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2966 conf.flags |= VXLAN_F_L3MISS;
2967
2968 if (data[IFLA_VXLAN_LIMIT])
2969 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2970
2971 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2972 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2973 conf.flags |= VXLAN_F_COLLECT_METADATA;
2974
2975 if (data[IFLA_VXLAN_PORT_RANGE]) {
2976 const struct ifla_vxlan_port_range *p
2977 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2978 conf.port_min = ntohs(p->low);
2979 conf.port_max = ntohs(p->high);
2980 }
2981
2982 if (data[IFLA_VXLAN_PORT])
2983 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2984
2985 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2986 conf.flags |= VXLAN_F_UDP_CSUM;
2987
2988 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2989 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2990 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2991
2992 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2993 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2994 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2995
2996 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2997 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2998 conf.flags |= VXLAN_F_REMCSUM_TX;
2999
3000 if (data[IFLA_VXLAN_REMCSUM_RX] &&
3001 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3002 conf.flags |= VXLAN_F_REMCSUM_RX;
3003
3004 if (data[IFLA_VXLAN_GBP])
3005 conf.flags |= VXLAN_F_GBP;
3006
3007 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
3008 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3009
3010 err = vxlan_dev_configure(src_net, dev, &conf);
3011 switch (err) {
3012 case -ENODEV:
3013 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
3014 break;
3015
3016 case -EPERM:
3017 pr_info("IPv6 is disabled via sysctl\n");
3018 break;
3019
3020 case -EEXIST:
3021 pr_info("duplicate VNI %u\n", conf.vni);
3022 break;
3023 }
3024
3025 return err;
3026 }
3027
3028 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3029 {
3030 struct vxlan_dev *vxlan = netdev_priv(dev);
3031 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3032
3033 spin_lock(&vn->sock_lock);
3034 if (!hlist_unhashed(&vxlan->hlist))
3035 hlist_del_rcu(&vxlan->hlist);
3036 spin_unlock(&vn->sock_lock);
3037
3038 gro_cells_destroy(&vxlan->gro_cells);
3039 list_del(&vxlan->next);
3040 unregister_netdevice_queue(dev, head);
3041 }
3042
3043 static size_t vxlan_get_size(const struct net_device *dev)
3044 {
3045
3046 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
3047 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
3048 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
3049 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3050 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3051 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3052 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
3053 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3054 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3055 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3056 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
3057 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
3058 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3059 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
3060 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3061 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3062 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3063 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3064 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3065 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3066 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3067 0;
3068 }
3069
3070 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3071 {
3072 const struct vxlan_dev *vxlan = netdev_priv(dev);
3073 const struct vxlan_rdst *dst = &vxlan->default_dst;
3074 struct ifla_vxlan_port_range ports = {
3075 .low = htons(vxlan->cfg.port_min),
3076 .high = htons(vxlan->cfg.port_max),
3077 };
3078
3079 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
3080 goto nla_put_failure;
3081
3082 if (!vxlan_addr_any(&dst->remote_ip)) {
3083 if (dst->remote_ip.sa.sa_family == AF_INET) {
3084 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3085 dst->remote_ip.sin.sin_addr.s_addr))
3086 goto nla_put_failure;
3087 #if IS_ENABLED(CONFIG_IPV6)
3088 } else {
3089 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3090 &dst->remote_ip.sin6.sin6_addr))
3091 goto nla_put_failure;
3092 #endif
3093 }
3094 }
3095
3096 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3097 goto nla_put_failure;
3098
3099 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3100 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3101 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3102 vxlan->cfg.saddr.sin.sin_addr.s_addr))
3103 goto nla_put_failure;
3104 #if IS_ENABLED(CONFIG_IPV6)
3105 } else {
3106 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3107 &vxlan->cfg.saddr.sin6.sin6_addr))
3108 goto nla_put_failure;
3109 #endif
3110 }
3111 }
3112
3113 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3114 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3115 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3116 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3117 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3118 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3119 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3120 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3121 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3122 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3123 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
3124 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3125 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
3126 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3127 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3128 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3129 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3130 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
3131 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3132 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3133 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3134 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3135 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3136 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3137 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3138 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
3139 goto nla_put_failure;
3140
3141 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3142 goto nla_put_failure;
3143
3144 if (vxlan->flags & VXLAN_F_GBP &&
3145 nla_put_flag(skb, IFLA_VXLAN_GBP))
3146 goto nla_put_failure;
3147
3148 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3149 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3150 goto nla_put_failure;
3151
3152 return 0;
3153
3154 nla_put_failure:
3155 return -EMSGSIZE;
3156 }
3157
3158 static struct net *vxlan_get_link_net(const struct net_device *dev)
3159 {
3160 struct vxlan_dev *vxlan = netdev_priv(dev);
3161
3162 return vxlan->net;
3163 }
3164
3165 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3166 .kind = "vxlan",
3167 .maxtype = IFLA_VXLAN_MAX,
3168 .policy = vxlan_policy,
3169 .priv_size = sizeof(struct vxlan_dev),
3170 .setup = vxlan_setup,
3171 .validate = vxlan_validate,
3172 .newlink = vxlan_newlink,
3173 .dellink = vxlan_dellink,
3174 .get_size = vxlan_get_size,
3175 .fill_info = vxlan_fill_info,
3176 .get_link_net = vxlan_get_link_net,
3177 };
3178
3179 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3180 struct net_device *dev)
3181 {
3182 struct vxlan_dev *vxlan, *next;
3183 LIST_HEAD(list_kill);
3184
3185 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3186 struct vxlan_rdst *dst = &vxlan->default_dst;
3187
3188 /* In case we created vxlan device with carrier
3189 * and we loose the carrier due to module unload
3190 * we also need to remove vxlan device. In other
3191 * cases, it's not necessary and remote_ifindex
3192 * is 0 here, so no matches.
3193 */
3194 if (dst->remote_ifindex == dev->ifindex)
3195 vxlan_dellink(vxlan->dev, &list_kill);
3196 }
3197
3198 unregister_netdevice_many(&list_kill);
3199 }
3200
3201 static int vxlan_lowerdev_event(struct notifier_block *unused,
3202 unsigned long event, void *ptr)
3203 {
3204 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3205 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3206
3207 if (event == NETDEV_UNREGISTER)
3208 vxlan_handle_lowerdev_unregister(vn, dev);
3209
3210 return NOTIFY_DONE;
3211 }
3212
3213 static struct notifier_block vxlan_notifier_block __read_mostly = {
3214 .notifier_call = vxlan_lowerdev_event,
3215 };
3216
3217 static __net_init int vxlan_init_net(struct net *net)
3218 {
3219 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3220 unsigned int h;
3221
3222 INIT_LIST_HEAD(&vn->vxlan_list);
3223 spin_lock_init(&vn->sock_lock);
3224
3225 for (h = 0; h < PORT_HASH_SIZE; ++h)
3226 INIT_HLIST_HEAD(&vn->sock_list[h]);
3227
3228 return 0;
3229 }
3230
3231 static void __net_exit vxlan_exit_net(struct net *net)
3232 {
3233 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3234 struct vxlan_dev *vxlan, *next;
3235 struct net_device *dev, *aux;
3236 LIST_HEAD(list);
3237
3238 rtnl_lock();
3239 for_each_netdev_safe(net, dev, aux)
3240 if (dev->rtnl_link_ops == &vxlan_link_ops)
3241 unregister_netdevice_queue(dev, &list);
3242
3243 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3244 /* If vxlan->dev is in the same netns, it has already been added
3245 * to the list by the previous loop.
3246 */
3247 if (!net_eq(dev_net(vxlan->dev), net)) {
3248 gro_cells_destroy(&vxlan->gro_cells);
3249 unregister_netdevice_queue(vxlan->dev, &list);
3250 }
3251 }
3252
3253 unregister_netdevice_many(&list);
3254 rtnl_unlock();
3255 }
3256
3257 static struct pernet_operations vxlan_net_ops = {
3258 .init = vxlan_init_net,
3259 .exit = vxlan_exit_net,
3260 .id = &vxlan_net_id,
3261 .size = sizeof(struct vxlan_net),
3262 };
3263
3264 static int __init vxlan_init_module(void)
3265 {
3266 int rc;
3267
3268 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3269 if (!vxlan_wq)
3270 return -ENOMEM;
3271
3272 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3273
3274 rc = register_pernet_subsys(&vxlan_net_ops);
3275 if (rc)
3276 goto out1;
3277
3278 rc = register_netdevice_notifier(&vxlan_notifier_block);
3279 if (rc)
3280 goto out2;
3281
3282 rc = rtnl_link_register(&vxlan_link_ops);
3283 if (rc)
3284 goto out3;
3285
3286 return 0;
3287 out3:
3288 unregister_netdevice_notifier(&vxlan_notifier_block);
3289 out2:
3290 unregister_pernet_subsys(&vxlan_net_ops);
3291 out1:
3292 destroy_workqueue(vxlan_wq);
3293 return rc;
3294 }
3295 late_initcall(vxlan_init_module);
3296
3297 static void __exit vxlan_cleanup_module(void)
3298 {
3299 rtnl_link_unregister(&vxlan_link_ops);
3300 unregister_netdevice_notifier(&vxlan_notifier_block);
3301 destroy_workqueue(vxlan_wq);
3302 unregister_pernet_subsys(&vxlan_net_ops);
3303 /* rcu_barrier() is called by netns */
3304 }
3305 module_exit(vxlan_cleanup_module);
3306
3307 MODULE_LICENSE("GPL");
3308 MODULE_VERSION(VXLAN_VERSION);
3309 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
3310 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
3311 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.099261 seconds and 5 git commands to generate.