?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 3.11-rc3
[cris-mirror.git] / drivers / net / vxlan.c
bloba5ba8dd7e6bea33d060e9d5306b9eab6bb2095da
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 * TODO
11 * - IPv6 (not in RFC)
12 */
13
14 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15
16 #include <linux/kernel.h>
17 #include <linux/types.h>
18 #include <linux/module.h>
19 #include <linux/errno.h>
20 #include <linux/slab.h>
21 #include <linux/skbuff.h>
22 #include <linux/rculist.h>
23 #include <linux/netdevice.h>
24 #include <linux/in.h>
25 #include <linux/ip.h>
26 #include <linux/udp.h>
27 #include <linux/igmp.h>
28 #include <linux/etherdevice.h>
29 #include <linux/if_ether.h>
30 #include <linux/hash.h>
31 #include <linux/ethtool.h>
32 #include <net/arp.h>
33 #include <net/ndisc.h>
34 #include <net/ip.h>
35 #include <net/ip_tunnels.h>
36 #include <net/icmp.h>
37 #include <net/udp.h>
38 #include <net/rtnetlink.h>
39 #include <net/route.h>
40 #include <net/dsfield.h>
41 #include <net/inet_ecn.h>
42 #include <net/net_namespace.h>
43 #include <net/netns/generic.h>
44
45 #define VXLAN_VERSION "0.1"
46
47 #define PORT_HASH_BITS 8
48 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
49 #define VNI_HASH_BITS 10
50 #define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
51 #define FDB_HASH_BITS 8
52 #define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
53 #define FDB_AGE_DEFAULT 300 /* 5 min */
54 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
55
56 #define VXLAN_N_VID (1u << 24)
57 #define VXLAN_VID_MASK (VXLAN_N_VID - 1)
58 /* IP header + UDP + VXLAN + Ethernet header */
59 #define VXLAN_HEADROOM (20 + 8 + 8 + 14)
60
61 #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
62
63 /* VXLAN protocol header */
64 struct vxlanhdr {
65 __be32 vx_flags;
66 __be32 vx_vni;
67 };
68
69 /* UDP port for VXLAN traffic.
70 * The IANA assigned port is 4789, but the Linux default is 8472
71 * for compatibility with early adopters.
72 */
73 static unsigned short vxlan_port __read_mostly = 8472;
74 module_param_named(udp_port, vxlan_port, ushort, 0444);
75 MODULE_PARM_DESC(udp_port, "Destination UDP port");
76
77 static bool log_ecn_error = true;
78 module_param(log_ecn_error, bool, 0644);
79 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
80
81 static int vxlan_net_id;
82
83 static const u8 all_zeros_mac[ETH_ALEN];
84
85 /* per UDP socket information */
86 struct vxlan_sock {
87 struct hlist_node hlist;
88 struct rcu_head rcu;
89 struct work_struct del_work;
90 atomic_t refcnt;
91 struct socket *sock;
92 struct hlist_head vni_list[VNI_HASH_SIZE];
93 };
94
95 /* per-network namespace private data for this module */
96 struct vxlan_net {
97 struct list_head vxlan_list;
98 struct hlist_head sock_list[PORT_HASH_SIZE];
99 spinlock_t sock_lock;
100 };
101
102 struct vxlan_rdst {
103 __be32 remote_ip;
104 __be16 remote_port;
105 u32 remote_vni;
106 u32 remote_ifindex;
107 struct list_head list;
108 struct rcu_head rcu;
109 };
110
111 /* Forwarding table entry */
112 struct vxlan_fdb {
113 struct hlist_node hlist; /* linked list of entries */
114 struct rcu_head rcu;
115 unsigned long updated; /* jiffies */
116 unsigned long used;
117 struct list_head remotes;
118 u16 state; /* see ndm_state */
119 u8 flags; /* see ndm_flags */
120 u8 eth_addr[ETH_ALEN];
121 };
122
123 /* Pseudo network device */
124 struct vxlan_dev {
125 struct hlist_node hlist; /* vni hash table */
126 struct list_head next; /* vxlan's per namespace list */
127 struct vxlan_sock *vn_sock; /* listening socket */
128 struct net_device *dev;
129 struct vxlan_rdst default_dst; /* default destination */
130 __be32 saddr; /* source address */
131 __be16 dst_port;
132 __u16 port_min; /* source port range */
133 __u16 port_max;
134 __u8 tos; /* TOS override */
135 __u8 ttl;
136 u32 flags; /* VXLAN_F_* below */
137
138 struct work_struct sock_work;
139 struct work_struct igmp_work;
140
141 unsigned long age_interval;
142 struct timer_list age_timer;
143 spinlock_t hash_lock;
144 unsigned int addrcnt;
145 unsigned int addrmax;
146
147 struct hlist_head fdb_head[FDB_HASH_SIZE];
148 };
149
150 #define VXLAN_F_LEARN 0x01
151 #define VXLAN_F_PROXY 0x02
152 #define VXLAN_F_RSC 0x04
153 #define VXLAN_F_L2MISS 0x08
154 #define VXLAN_F_L3MISS 0x10
155
156 /* salt for hash table */
157 static u32 vxlan_salt __read_mostly;
158 static struct workqueue_struct *vxlan_wq;
159
160 static void vxlan_sock_work(struct work_struct *work);
161
162 /* Virtual Network hash table head */
163 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
164 {
165 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
166 }
167
168 /* Socket hash table head */
169 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
170 {
171 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
172
173 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
174 }
175
176 /* First remote destination for a forwarding entry.
177 * Guaranteed to be non-NULL because remotes are never deleted.
178 */
179 static inline struct vxlan_rdst *first_remote(struct vxlan_fdb *fdb)
180 {
181 return list_first_or_null_rcu(&fdb->remotes, struct vxlan_rdst, list);
182 }
183
184 /* Find VXLAN socket based on network namespace and UDP port */
185 static struct vxlan_sock *vxlan_find_port(struct net *net, __be16 port)
186 {
187 struct vxlan_sock *vs;
188
189 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
190 if (inet_sk(vs->sock->sk)->inet_sport == port)
191 return vs;
192 }
193 return NULL;
194 }
195
196 /* Look up VNI in a per net namespace table */
197 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
198 {
199 struct vxlan_sock *vs;
200 struct vxlan_dev *vxlan;
201
202 vs = vxlan_find_port(net, port);
203 if (!vs)
204 return NULL;
205
206 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
207 if (vxlan->default_dst.remote_vni == id)
208 return vxlan;
209 }
210
211 return NULL;
212 }
213
214 /* Fill in neighbour message in skbuff. */
215 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
216 const struct vxlan_fdb *fdb,
217 u32 portid, u32 seq, int type, unsigned int flags,
218 const struct vxlan_rdst *rdst)
219 {
220 unsigned long now = jiffies;
221 struct nda_cacheinfo ci;
222 struct nlmsghdr *nlh;
223 struct ndmsg *ndm;
224 bool send_ip, send_eth;
225
226 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
227 if (nlh == NULL)
228 return -EMSGSIZE;
229
230 ndm = nlmsg_data(nlh);
231 memset(ndm, 0, sizeof(*ndm));
232
233 send_eth = send_ip = true;
234
235 if (type == RTM_GETNEIGH) {
236 ndm->ndm_family = AF_INET;
237 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
238 send_eth = !is_zero_ether_addr(fdb->eth_addr);
239 } else
240 ndm->ndm_family = AF_BRIDGE;
241 ndm->ndm_state = fdb->state;
242 ndm->ndm_ifindex = vxlan->dev->ifindex;
243 ndm->ndm_flags = fdb->flags;
244 ndm->ndm_type = NDA_DST;
245
246 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
247 goto nla_put_failure;
248
249 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
250 goto nla_put_failure;
251
252 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
253 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
254 goto nla_put_failure;
255 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
256 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
257 goto nla_put_failure;
258 if (rdst->remote_ifindex &&
259 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
260 goto nla_put_failure;
261
262 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
263 ci.ndm_confirmed = 0;
264 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
265 ci.ndm_refcnt = 0;
266
267 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
268 goto nla_put_failure;
269
270 return nlmsg_end(skb, nlh);
271
272 nla_put_failure:
273 nlmsg_cancel(skb, nlh);
274 return -EMSGSIZE;
275 }
276
277 static inline size_t vxlan_nlmsg_size(void)
278 {
279 return NLMSG_ALIGN(sizeof(struct ndmsg))
280 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
281 + nla_total_size(sizeof(__be32)) /* NDA_DST */
282 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
283 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
284 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
285 + nla_total_size(sizeof(struct nda_cacheinfo));
286 }
287
288 static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
289 struct vxlan_fdb *fdb, int type)
290 {
291 struct net *net = dev_net(vxlan->dev);
292 struct sk_buff *skb;
293 int err = -ENOBUFS;
294
295 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
296 if (skb == NULL)
297 goto errout;
298
299 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, first_remote(fdb));
300 if (err < 0) {
301 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
302 WARN_ON(err == -EMSGSIZE);
303 kfree_skb(skb);
304 goto errout;
305 }
306
307 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
308 return;
309 errout:
310 if (err < 0)
311 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
312 }
313
314 static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
315 {
316 struct vxlan_dev *vxlan = netdev_priv(dev);
317 struct vxlan_fdb f = {
318 .state = NUD_STALE,
319 };
320 struct vxlan_rdst remote = {
321 .remote_ip = ipa, /* goes to NDA_DST */
322 .remote_vni = VXLAN_N_VID,
323 };
324
325 INIT_LIST_HEAD(&f.remotes);
326 list_add_rcu(&remote.list, &f.remotes);
327
328 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
329 }
330
331 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
332 {
333 struct vxlan_fdb f = {
334 .state = NUD_STALE,
335 };
336
337 INIT_LIST_HEAD(&f.remotes);
338 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
339
340 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
341 }
342
343 /* Hash Ethernet address */
344 static u32 eth_hash(const unsigned char *addr)
345 {
346 u64 value = get_unaligned((u64 *)addr);
347
348 /* only want 6 bytes */
349 #ifdef __BIG_ENDIAN
350 value >>= 16;
351 #else
352 value <<= 16;
353 #endif
354 return hash_64(value, FDB_HASH_BITS);
355 }
356
357 /* Hash chain to use given mac address */
358 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
359 const u8 *mac)
360 {
361 return &vxlan->fdb_head[eth_hash(mac)];
362 }
363
364 /* Look up Ethernet address in forwarding table */
365 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
366 const u8 *mac)
367
368 {
369 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
370 struct vxlan_fdb *f;
371
372 hlist_for_each_entry_rcu(f, head, hlist) {
373 if (compare_ether_addr(mac, f->eth_addr) == 0)
374 return f;
375 }
376
377 return NULL;
378 }
379
380 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
381 const u8 *mac)
382 {
383 struct vxlan_fdb *f;
384
385 f = __vxlan_find_mac(vxlan, mac);
386 if (f)
387 f->used = jiffies;
388
389 return f;
390 }
391
392 /* caller should hold vxlan->hash_lock */
393 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
394 __be32 ip, __be16 port,
395 __u32 vni, __u32 ifindex)
396 {
397 struct vxlan_rdst *rd;
398
399 list_for_each_entry(rd, &f->remotes, list) {
400 if (rd->remote_ip == ip &&
401 rd->remote_port == port &&
402 rd->remote_vni == vni &&
403 rd->remote_ifindex == ifindex)
404 return rd;
405 }
406
407 return NULL;
408 }
409
410 /* Add/update destinations for multicast */
411 static int vxlan_fdb_append(struct vxlan_fdb *f,
412 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
413 {
414 struct vxlan_rdst *rd;
415
416 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
417 if (rd)
418 return 0;
419
420 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
421 if (rd == NULL)
422 return -ENOBUFS;
423 rd->remote_ip = ip;
424 rd->remote_port = port;
425 rd->remote_vni = vni;
426 rd->remote_ifindex = ifindex;
427
428 list_add_tail_rcu(&rd->list, &f->remotes);
429
430 return 1;
431 }
432
433 /* Add new entry to forwarding table -- assumes lock held */
434 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
435 const u8 *mac, __be32 ip,
436 __u16 state, __u16 flags,
437 __be16 port, __u32 vni, __u32 ifindex,
438 __u8 ndm_flags)
439 {
440 struct vxlan_fdb *f;
441 int notify = 0;
442
443 f = __vxlan_find_mac(vxlan, mac);
444 if (f) {
445 if (flags & NLM_F_EXCL) {
446 netdev_dbg(vxlan->dev,
447 "lost race to create %pM\n", mac);
448 return -EEXIST;
449 }
450 if (f->state != state) {
451 f->state = state;
452 f->updated = jiffies;
453 notify = 1;
454 }
455 if (f->flags != ndm_flags) {
456 f->flags = ndm_flags;
457 f->updated = jiffies;
458 notify = 1;
459 }
460 if ((flags & NLM_F_APPEND) &&
461 (is_multicast_ether_addr(f->eth_addr) ||
462 is_zero_ether_addr(f->eth_addr))) {
463 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
464
465 if (rc < 0)
466 return rc;
467 notify |= rc;
468 }
469 } else {
470 if (!(flags & NLM_F_CREATE))
471 return -ENOENT;
472
473 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
474 return -ENOSPC;
475
476 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
477 f = kmalloc(sizeof(*f), GFP_ATOMIC);
478 if (!f)
479 return -ENOMEM;
480
481 notify = 1;
482 f->state = state;
483 f->flags = ndm_flags;
484 f->updated = f->used = jiffies;
485 INIT_LIST_HEAD(&f->remotes);
486 memcpy(f->eth_addr, mac, ETH_ALEN);
487
488 vxlan_fdb_append(f, ip, port, vni, ifindex);
489
490 ++vxlan->addrcnt;
491 hlist_add_head_rcu(&f->hlist,
492 vxlan_fdb_head(vxlan, mac));
493 }
494
495 if (notify)
496 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
497
498 return 0;
499 }
500
501 static void vxlan_fdb_free_rdst(struct rcu_head *head)
502 {
503 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
504 kfree(rd);
505 }
506
507 static void vxlan_fdb_free(struct rcu_head *head)
508 {
509 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
510 struct vxlan_rdst *rd, *nd;
511
512 list_for_each_entry_safe(rd, nd, &f->remotes, list)
513 kfree(rd);
514 kfree(f);
515 }
516
517 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
518 {
519 netdev_dbg(vxlan->dev,
520 "delete %pM\n", f->eth_addr);
521
522 --vxlan->addrcnt;
523 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
524
525 hlist_del_rcu(&f->hlist);
526 call_rcu(&f->rcu, vxlan_fdb_free);
527 }
528
529 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
530 __be32 *ip, __be16 *port, u32 *vni, u32 *ifindex)
531 {
532 struct net *net = dev_net(vxlan->dev);
533
534 if (tb[NDA_DST]) {
535 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
536 return -EAFNOSUPPORT;
537
538 *ip = nla_get_be32(tb[NDA_DST]);
539 } else {
540 *ip = htonl(INADDR_ANY);
541 }
542
543 if (tb[NDA_PORT]) {
544 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
545 return -EINVAL;
546 *port = nla_get_be16(tb[NDA_PORT]);
547 } else {
548 *port = vxlan->dst_port;
549 }
550
551 if (tb[NDA_VNI]) {
552 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
553 return -EINVAL;
554 *vni = nla_get_u32(tb[NDA_VNI]);
555 } else {
556 *vni = vxlan->default_dst.remote_vni;
557 }
558
559 if (tb[NDA_IFINDEX]) {
560 struct net_device *tdev;
561
562 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
563 return -EINVAL;
564 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
565 tdev = dev_get_by_index(net, *ifindex);
566 if (!tdev)
567 return -EADDRNOTAVAIL;
568 dev_put(tdev);
569 } else {
570 *ifindex = 0;
571 }
572
573 return 0;
574 }
575
576 /* Add static entry (via netlink) */
577 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
578 struct net_device *dev,
579 const unsigned char *addr, u16 flags)
580 {
581 struct vxlan_dev *vxlan = netdev_priv(dev);
582 /* struct net *net = dev_net(vxlan->dev); */
583 __be32 ip;
584 __be16 port;
585 u32 vni, ifindex;
586 int err;
587
588 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
589 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
590 ndm->ndm_state);
591 return -EINVAL;
592 }
593
594 if (tb[NDA_DST] == NULL)
595 return -EINVAL;
596
597 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
598 if (err)
599 return err;
600
601 spin_lock_bh(&vxlan->hash_lock);
602 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags,
603 port, vni, ifindex, ndm->ndm_flags);
604 spin_unlock_bh(&vxlan->hash_lock);
605
606 return err;
607 }
608
609 /* Delete entry (via netlink) */
610 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
611 struct net_device *dev,
612 const unsigned char *addr)
613 {
614 struct vxlan_dev *vxlan = netdev_priv(dev);
615 struct vxlan_fdb *f;
616 struct vxlan_rdst *rd = NULL;
617 __be32 ip;
618 __be16 port;
619 u32 vni, ifindex;
620 int err;
621
622 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
623 if (err)
624 return err;
625
626 err = -ENOENT;
627
628 spin_lock_bh(&vxlan->hash_lock);
629 f = vxlan_find_mac(vxlan, addr);
630 if (!f)
631 goto out;
632
633 if (ip != htonl(INADDR_ANY)) {
634 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
635 if (!rd)
636 goto out;
637 }
638
639 err = 0;
640
641 /* remove a destination if it's not the only one on the list,
642 * otherwise destroy the fdb entry
643 */
644 if (rd && !list_is_singular(&f->remotes)) {
645 list_del_rcu(&rd->list);
646 call_rcu(&rd->rcu, vxlan_fdb_free_rdst);
647 goto out;
648 }
649
650 vxlan_fdb_destroy(vxlan, f);
651
652 out:
653 spin_unlock_bh(&vxlan->hash_lock);
654
655 return err;
656 }
657
658 /* Dump forwarding table */
659 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
660 struct net_device *dev, int idx)
661 {
662 struct vxlan_dev *vxlan = netdev_priv(dev);
663 unsigned int h;
664
665 for (h = 0; h < FDB_HASH_SIZE; ++h) {
666 struct vxlan_fdb *f;
667 int err;
668
669 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
670 struct vxlan_rdst *rd;
671
672 if (idx < cb->args[0])
673 goto skip;
674
675 list_for_each_entry_rcu(rd, &f->remotes, list) {
676 err = vxlan_fdb_info(skb, vxlan, f,
677 NETLINK_CB(cb->skb).portid,
678 cb->nlh->nlmsg_seq,
679 RTM_NEWNEIGH,
680 NLM_F_MULTI, rd);
681 if (err < 0)
682 goto out;
683 }
684 skip:
685 ++idx;
686 }
687 }
688 out:
689 return idx;
690 }
691
692 /* Watch incoming packets to learn mapping between Ethernet address
693 * and Tunnel endpoint.
694 * Return true if packet is bogus and should be droppped.
695 */
696 static bool vxlan_snoop(struct net_device *dev,
697 __be32 src_ip, const u8 *src_mac)
698 {
699 struct vxlan_dev *vxlan = netdev_priv(dev);
700 struct vxlan_fdb *f;
701
702 f = vxlan_find_mac(vxlan, src_mac);
703 if (likely(f)) {
704 struct vxlan_rdst *rdst = first_remote(f);
705
706 if (likely(rdst->remote_ip == src_ip))
707 return false;
708
709 /* Don't migrate static entries, drop packets */
710 if (f->state & NUD_NOARP)
711 return true;
712
713 if (net_ratelimit())
714 netdev_info(dev,
715 "%pM migrated from %pI4 to %pI4\n",
716 src_mac, &rdst->remote_ip, &src_ip);
717
718 rdst->remote_ip = src_ip;
719 f->updated = jiffies;
720 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
721 } else {
722 /* learned new entry */
723 spin_lock(&vxlan->hash_lock);
724
725 /* close off race between vxlan_flush and incoming packets */
726 if (netif_running(dev))
727 vxlan_fdb_create(vxlan, src_mac, src_ip,
728 NUD_REACHABLE,
729 NLM_F_EXCL|NLM_F_CREATE,
730 vxlan->dst_port,
731 vxlan->default_dst.remote_vni,
732 0, NTF_SELF);
733 spin_unlock(&vxlan->hash_lock);
734 }
735
736 return false;
737 }
738
739
740 /* See if multicast group is already in use by other ID */
741 static bool vxlan_group_used(struct vxlan_net *vn, __be32 remote_ip)
742 {
743 struct vxlan_dev *vxlan;
744
745 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
746 if (!netif_running(vxlan->dev))
747 continue;
748
749 if (vxlan->default_dst.remote_ip == remote_ip)
750 return true;
751 }
752
753 return false;
754 }
755
756 static void vxlan_sock_hold(struct vxlan_sock *vs)
757 {
758 atomic_inc(&vs->refcnt);
759 }
760
761 static void vxlan_sock_release(struct vxlan_net *vn, struct vxlan_sock *vs)
762 {
763 if (!atomic_dec_and_test(&vs->refcnt))
764 return;
765
766 spin_lock(&vn->sock_lock);
767 hlist_del_rcu(&vs->hlist);
768 spin_unlock(&vn->sock_lock);
769
770 queue_work(vxlan_wq, &vs->del_work);
771 }
772
773 /* Callback to update multicast group membership.
774 * Scheduled when vxlan goes up/down.
775 */
776 static void vxlan_igmp_work(struct work_struct *work)
777 {
778 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_work);
779 struct vxlan_net *vn = net_generic(dev_net(vxlan->dev), vxlan_net_id);
780 struct vxlan_sock *vs = vxlan->vn_sock;
781 struct sock *sk = vs->sock->sk;
782 struct ip_mreqn mreq = {
783 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
784 .imr_ifindex = vxlan->default_dst.remote_ifindex,
785 };
786
787 lock_sock(sk);
788 if (vxlan_group_used(vn, vxlan->default_dst.remote_ip))
789 ip_mc_join_group(sk, &mreq);
790 else
791 ip_mc_leave_group(sk, &mreq);
792 release_sock(sk);
793
794 vxlan_sock_release(vn, vs);
795 dev_put(vxlan->dev);
796 }
797
798 /* Callback from net/ipv4/udp.c to receive packets */
799 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
800 {
801 struct iphdr *oip;
802 struct vxlanhdr *vxh;
803 struct vxlan_dev *vxlan;
804 struct pcpu_tstats *stats;
805 __be16 port;
806 __u32 vni;
807 int err;
808
809 /* pop off outer UDP header */
810 __skb_pull(skb, sizeof(struct udphdr));
811
812 /* Need Vxlan and inner Ethernet header to be present */
813 if (!pskb_may_pull(skb, sizeof(struct vxlanhdr)))
814 goto error;
815
816 /* Drop packets with reserved bits set */
817 vxh = (struct vxlanhdr *) skb->data;
818 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
819 (vxh->vx_vni & htonl(0xff))) {
820 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
821 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
822 goto error;
823 }
824
825 __skb_pull(skb, sizeof(struct vxlanhdr));
826
827 /* Is this VNI defined? */
828 vni = ntohl(vxh->vx_vni) >> 8;
829 port = inet_sk(sk)->inet_sport;
830 vxlan = vxlan_find_vni(sock_net(sk), vni, port);
831 if (!vxlan) {
832 netdev_dbg(skb->dev, "unknown vni %d port %u\n",
833 vni, ntohs(port));
834 goto drop;
835 }
836
837 if (!pskb_may_pull(skb, ETH_HLEN)) {
838 vxlan->dev->stats.rx_length_errors++;
839 vxlan->dev->stats.rx_errors++;
840 goto drop;
841 }
842
843 skb_reset_mac_header(skb);
844
845 /* Re-examine inner Ethernet packet */
846 oip = ip_hdr(skb);
847 skb->protocol = eth_type_trans(skb, vxlan->dev);
848
849 /* Ignore packet loops (and multicast echo) */
850 if (compare_ether_addr(eth_hdr(skb)->h_source,
851 vxlan->dev->dev_addr) == 0)
852 goto drop;
853
854 if ((vxlan->flags & VXLAN_F_LEARN) &&
855 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source))
856 goto drop;
857
858 __skb_tunnel_rx(skb, vxlan->dev);
859 skb_reset_network_header(skb);
860
861 /* If the NIC driver gave us an encapsulated packet with
862 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
863 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
864 * for us. Otherwise force the upper layers to verify it.
865 */
866 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
867 !(vxlan->dev->features & NETIF_F_RXCSUM))
868 skb->ip_summed = CHECKSUM_NONE;
869
870 skb->encapsulation = 0;
871
872 err = IP_ECN_decapsulate(oip, skb);
873 if (unlikely(err)) {
874 if (log_ecn_error)
875 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
876 &oip->saddr, oip->tos);
877 if (err > 1) {
878 ++vxlan->dev->stats.rx_frame_errors;
879 ++vxlan->dev->stats.rx_errors;
880 goto drop;
881 }
882 }
883
884 stats = this_cpu_ptr(vxlan->dev->tstats);
885 u64_stats_update_begin(&stats->syncp);
886 stats->rx_packets++;
887 stats->rx_bytes += skb->len;
888 u64_stats_update_end(&stats->syncp);
889
890 netif_rx(skb);
891
892 return 0;
893 error:
894 /* Put UDP header back */
895 __skb_push(skb, sizeof(struct udphdr));
896
897 return 1;
898 drop:
899 /* Consume bad packet */
900 kfree_skb(skb);
901 return 0;
902 }
903
904 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
905 {
906 struct vxlan_dev *vxlan = netdev_priv(dev);
907 struct arphdr *parp;
908 u8 *arpptr, *sha;
909 __be32 sip, tip;
910 struct neighbour *n;
911
912 if (dev->flags & IFF_NOARP)
913 goto out;
914
915 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
916 dev->stats.tx_dropped++;
917 goto out;
918 }
919 parp = arp_hdr(skb);
920
921 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
922 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
923 parp->ar_pro != htons(ETH_P_IP) ||
924 parp->ar_op != htons(ARPOP_REQUEST) ||
925 parp->ar_hln != dev->addr_len ||
926 parp->ar_pln != 4)
927 goto out;
928 arpptr = (u8 *)parp + sizeof(struct arphdr);
929 sha = arpptr;
930 arpptr += dev->addr_len; /* sha */
931 memcpy(&sip, arpptr, sizeof(sip));
932 arpptr += sizeof(sip);
933 arpptr += dev->addr_len; /* tha */
934 memcpy(&tip, arpptr, sizeof(tip));
935
936 if (ipv4_is_loopback(tip) ||
937 ipv4_is_multicast(tip))
938 goto out;
939
940 n = neigh_lookup(&arp_tbl, &tip, dev);
941
942 if (n) {
943 struct vxlan_fdb *f;
944 struct sk_buff *reply;
945
946 if (!(n->nud_state & NUD_CONNECTED)) {
947 neigh_release(n);
948 goto out;
949 }
950
951 f = vxlan_find_mac(vxlan, n->ha);
952 if (f && first_remote(f)->remote_ip == htonl(INADDR_ANY)) {
953 /* bridge-local neighbor */
954 neigh_release(n);
955 goto out;
956 }
957
958 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
959 n->ha, sha);
960
961 neigh_release(n);
962
963 skb_reset_mac_header(reply);
964 __skb_pull(reply, skb_network_offset(reply));
965 reply->ip_summed = CHECKSUM_UNNECESSARY;
966 reply->pkt_type = PACKET_HOST;
967
968 if (netif_rx_ni(reply) == NET_RX_DROP)
969 dev->stats.rx_dropped++;
970 } else if (vxlan->flags & VXLAN_F_L3MISS)
971 vxlan_ip_miss(dev, tip);
972 out:
973 consume_skb(skb);
974 return NETDEV_TX_OK;
975 }
976
977 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
978 {
979 struct vxlan_dev *vxlan = netdev_priv(dev);
980 struct neighbour *n;
981 struct iphdr *pip;
982
983 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
984 return false;
985
986 n = NULL;
987 switch (ntohs(eth_hdr(skb)->h_proto)) {
988 case ETH_P_IP:
989 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
990 return false;
991 pip = ip_hdr(skb);
992 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
993 break;
994 default:
995 return false;
996 }
997
998 if (n) {
999 bool diff;
1000
1001 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
1002 if (diff) {
1003 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1004 dev->addr_len);
1005 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1006 }
1007 neigh_release(n);
1008 return diff;
1009 } else if (vxlan->flags & VXLAN_F_L3MISS)
1010 vxlan_ip_miss(dev, pip->daddr);
1011 return false;
1012 }
1013
1014 static void vxlan_sock_put(struct sk_buff *skb)
1015 {
1016 sock_put(skb->sk);
1017 }
1018
1019 /* On transmit, associate with the tunnel socket */
1020 static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
1021 {
1022 struct vxlan_dev *vxlan = netdev_priv(dev);
1023 struct sock *sk = vxlan->vn_sock->sock->sk;
1024
1025 skb_orphan(skb);
1026 sock_hold(sk);
1027 skb->sk = sk;
1028 skb->destructor = vxlan_sock_put;
1029 }
1030
1031 /* Compute source port for outgoing packet
1032 * first choice to use L4 flow hash since it will spread
1033 * better and maybe available from hardware
1034 * secondary choice is to use jhash on the Ethernet header
1035 */
1036 static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
1037 {
1038 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
1039 u32 hash;
1040
1041 hash = skb_get_rxhash(skb);
1042 if (!hash)
1043 hash = jhash(skb->data, 2 * ETH_ALEN,
1044 (__force u32) skb->protocol);
1045
1046 return htons((((u64) hash * range) >> 32) + vxlan->port_min);
1047 }
1048
1049 static int handle_offloads(struct sk_buff *skb)
1050 {
1051 if (skb_is_gso(skb)) {
1052 int err = skb_unclone(skb, GFP_ATOMIC);
1053 if (unlikely(err))
1054 return err;
1055
1056 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
1057 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1058 skb->ip_summed = CHECKSUM_NONE;
1059
1060 return 0;
1061 }
1062
1063 /* Bypass encapsulation if the destination is local */
1064 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1065 struct vxlan_dev *dst_vxlan)
1066 {
1067 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1068 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1069
1070 skb->pkt_type = PACKET_HOST;
1071 skb->encapsulation = 0;
1072 skb->dev = dst_vxlan->dev;
1073 __skb_pull(skb, skb_network_offset(skb));
1074
1075 if (dst_vxlan->flags & VXLAN_F_LEARN)
1076 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
1077 eth_hdr(skb)->h_source);
1078
1079 u64_stats_update_begin(&tx_stats->syncp);
1080 tx_stats->tx_packets++;
1081 tx_stats->tx_bytes += skb->len;
1082 u64_stats_update_end(&tx_stats->syncp);
1083
1084 if (netif_rx(skb) == NET_RX_SUCCESS) {
1085 u64_stats_update_begin(&rx_stats->syncp);
1086 rx_stats->rx_packets++;
1087 rx_stats->rx_bytes += skb->len;
1088 u64_stats_update_end(&rx_stats->syncp);
1089 } else {
1090 skb->dev->stats.rx_dropped++;
1091 }
1092 }
1093
1094 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1095 struct vxlan_rdst *rdst, bool did_rsc)
1096 {
1097 struct vxlan_dev *vxlan = netdev_priv(dev);
1098 struct rtable *rt;
1099 const struct iphdr *old_iph;
1100 struct vxlanhdr *vxh;
1101 struct udphdr *uh;
1102 struct flowi4 fl4;
1103 __be32 dst;
1104 __be16 src_port, dst_port;
1105 u32 vni;
1106 __be16 df = 0;
1107 __u8 tos, ttl;
1108 int err;
1109
1110 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1111 vni = rdst->remote_vni;
1112 dst = rdst->remote_ip;
1113
1114 if (!dst) {
1115 if (did_rsc) {
1116 /* short-circuited back to local bridge */
1117 vxlan_encap_bypass(skb, vxlan, vxlan);
1118 return;
1119 }
1120 goto drop;
1121 }
1122
1123 if (!skb->encapsulation) {
1124 skb_reset_inner_headers(skb);
1125 skb->encapsulation = 1;
1126 }
1127
1128 /* Need space for new headers (invalidates iph ptr) */
1129 if (skb_cow_head(skb, VXLAN_HEADROOM))
1130 goto drop;
1131
1132 old_iph = ip_hdr(skb);
1133
1134 ttl = vxlan->ttl;
1135 if (!ttl && IN_MULTICAST(ntohl(dst)))
1136 ttl = 1;
1137
1138 tos = vxlan->tos;
1139 if (tos == 1)
1140 tos = ip_tunnel_get_dsfield(old_iph, skb);
1141
1142 src_port = vxlan_src_port(vxlan, skb);
1143
1144 memset(&fl4, 0, sizeof(fl4));
1145 fl4.flowi4_oif = rdst->remote_ifindex;
1146 fl4.flowi4_tos = RT_TOS(tos);
1147 fl4.daddr = dst;
1148 fl4.saddr = vxlan->saddr;
1149
1150 rt = ip_route_output_key(dev_net(dev), &fl4);
1151 if (IS_ERR(rt)) {
1152 netdev_dbg(dev, "no route to %pI4\n", &dst);
1153 dev->stats.tx_carrier_errors++;
1154 goto tx_error;
1155 }
1156
1157 if (rt->dst.dev == dev) {
1158 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1159 ip_rt_put(rt);
1160 dev->stats.collisions++;
1161 goto tx_error;
1162 }
1163
1164 /* Bypass encapsulation if the destination is local */
1165 if (rt->rt_flags & RTCF_LOCAL &&
1166 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1167 struct vxlan_dev *dst_vxlan;
1168
1169 ip_rt_put(rt);
1170 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1171 if (!dst_vxlan)
1172 goto tx_error;
1173 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1174 return;
1175 }
1176 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1177 vxh->vx_flags = htonl(VXLAN_FLAGS);
1178 vxh->vx_vni = htonl(vni << 8);
1179
1180 __skb_push(skb, sizeof(*uh));
1181 skb_reset_transport_header(skb);
1182 uh = udp_hdr(skb);
1183
1184 uh->dest = dst_port;
1185 uh->source = src_port;
1186
1187 uh->len = htons(skb->len);
1188 uh->check = 0;
1189
1190 vxlan_set_owner(dev, skb);
1191
1192 if (handle_offloads(skb))
1193 goto drop;
1194
1195 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1196 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1197
1198 err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, dst,
1199 IPPROTO_UDP, tos, ttl, df);
1200 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1201
1202 return;
1203
1204 drop:
1205 dev->stats.tx_dropped++;
1206 goto tx_free;
1207
1208 tx_error:
1209 dev->stats.tx_errors++;
1210 tx_free:
1211 dev_kfree_skb(skb);
1212 }
1213
1214 /* Transmit local packets over Vxlan
1215 *
1216 * Outer IP header inherits ECN and DF from inner header.
1217 * Outer UDP destination is the VXLAN assigned port.
1218 * source port is based on hash of flow
1219 */
1220 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1221 {
1222 struct vxlan_dev *vxlan = netdev_priv(dev);
1223 struct ethhdr *eth;
1224 bool did_rsc = false;
1225 struct vxlan_rdst *rdst;
1226 struct vxlan_fdb *f;
1227
1228 skb_reset_mac_header(skb);
1229 eth = eth_hdr(skb);
1230
1231 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1232 return arp_reduce(dev, skb);
1233
1234 f = vxlan_find_mac(vxlan, eth->h_dest);
1235 did_rsc = false;
1236
1237 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1238 ntohs(eth->h_proto) == ETH_P_IP) {
1239 did_rsc = route_shortcircuit(dev, skb);
1240 if (did_rsc)
1241 f = vxlan_find_mac(vxlan, eth->h_dest);
1242 }
1243
1244 if (f == NULL) {
1245 f = vxlan_find_mac(vxlan, all_zeros_mac);
1246 if (f == NULL) {
1247 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1248 !is_multicast_ether_addr(eth->h_dest))
1249 vxlan_fdb_miss(vxlan, eth->h_dest);
1250
1251 dev->stats.tx_dropped++;
1252 dev_kfree_skb(skb);
1253 return NETDEV_TX_OK;
1254 }
1255 }
1256
1257 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1258 struct sk_buff *skb1;
1259
1260 skb1 = skb_clone(skb, GFP_ATOMIC);
1261 if (skb1)
1262 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1263 }
1264
1265 dev_kfree_skb(skb);
1266 return NETDEV_TX_OK;
1267 }
1268
1269 /* Walk the forwarding table and purge stale entries */
1270 static void vxlan_cleanup(unsigned long arg)
1271 {
1272 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1273 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1274 unsigned int h;
1275
1276 if (!netif_running(vxlan->dev))
1277 return;
1278
1279 spin_lock_bh(&vxlan->hash_lock);
1280 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1281 struct hlist_node *p, *n;
1282 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1283 struct vxlan_fdb *f
1284 = container_of(p, struct vxlan_fdb, hlist);
1285 unsigned long timeout;
1286
1287 if (f->state & NUD_PERMANENT)
1288 continue;
1289
1290 timeout = f->used + vxlan->age_interval * HZ;
1291 if (time_before_eq(timeout, jiffies)) {
1292 netdev_dbg(vxlan->dev,
1293 "garbage collect %pM\n",
1294 f->eth_addr);
1295 f->state = NUD_STALE;
1296 vxlan_fdb_destroy(vxlan, f);
1297 } else if (time_before(timeout, next_timer))
1298 next_timer = timeout;
1299 }
1300 }
1301 spin_unlock_bh(&vxlan->hash_lock);
1302
1303 mod_timer(&vxlan->age_timer, next_timer);
1304 }
1305
1306 /* Setup stats when device is created */
1307 static int vxlan_init(struct net_device *dev)
1308 {
1309 struct vxlan_dev *vxlan = netdev_priv(dev);
1310 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1311 struct vxlan_sock *vs;
1312 __u32 vni = vxlan->default_dst.remote_vni;
1313
1314 dev->tstats = alloc_percpu(struct pcpu_tstats);
1315 if (!dev->tstats)
1316 return -ENOMEM;
1317
1318 spin_lock(&vn->sock_lock);
1319 vs = vxlan_find_port(dev_net(dev), vxlan->dst_port);
1320 if (vs) {
1321 /* If we have a socket with same port already, reuse it */
1322 atomic_inc(&vs->refcnt);
1323 vxlan->vn_sock = vs;
1324 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1325 } else {
1326 /* otherwise make new socket outside of RTNL */
1327 dev_hold(dev);
1328 queue_work(vxlan_wq, &vxlan->sock_work);
1329 }
1330 spin_unlock(&vn->sock_lock);
1331
1332 return 0;
1333 }
1334
1335 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
1336 {
1337 struct vxlan_fdb *f;
1338
1339 spin_lock_bh(&vxlan->hash_lock);
1340 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1341 if (f)
1342 vxlan_fdb_destroy(vxlan, f);
1343 spin_unlock_bh(&vxlan->hash_lock);
1344 }
1345
1346 static void vxlan_uninit(struct net_device *dev)
1347 {
1348 struct vxlan_dev *vxlan = netdev_priv(dev);
1349 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1350 struct vxlan_sock *vs = vxlan->vn_sock;
1351
1352 vxlan_fdb_delete_default(vxlan);
1353
1354 if (vs)
1355 vxlan_sock_release(vn, vs);
1356 free_percpu(dev->tstats);
1357 }
1358
1359 /* Start ageing timer and join group when device is brought up */
1360 static int vxlan_open(struct net_device *dev)
1361 {
1362 struct vxlan_dev *vxlan = netdev_priv(dev);
1363 struct vxlan_sock *vs = vxlan->vn_sock;
1364
1365 /* socket hasn't been created */
1366 if (!vs)
1367 return -ENOTCONN;
1368
1369 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
1370 vxlan_sock_hold(vs);
1371 dev_hold(dev);
1372 queue_work(vxlan_wq, &vxlan->igmp_work);
1373 }
1374
1375 if (vxlan->age_interval)
1376 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1377
1378 return 0;
1379 }
1380
1381 /* Purge the forwarding table */
1382 static void vxlan_flush(struct vxlan_dev *vxlan)
1383 {
1384 unsigned int h;
1385
1386 spin_lock_bh(&vxlan->hash_lock);
1387 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1388 struct hlist_node *p, *n;
1389 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1390 struct vxlan_fdb *f
1391 = container_of(p, struct vxlan_fdb, hlist);
1392 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1393 if (!is_zero_ether_addr(f->eth_addr))
1394 vxlan_fdb_destroy(vxlan, f);
1395 }
1396 }
1397 spin_unlock_bh(&vxlan->hash_lock);
1398 }
1399
1400 /* Cleanup timer and forwarding table on shutdown */
1401 static int vxlan_stop(struct net_device *dev)
1402 {
1403 struct vxlan_dev *vxlan = netdev_priv(dev);
1404 struct vxlan_sock *vs = vxlan->vn_sock;
1405
1406 if (vs && IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
1407 vxlan_sock_hold(vs);
1408 dev_hold(dev);
1409 queue_work(vxlan_wq, &vxlan->igmp_work);
1410 }
1411
1412 del_timer_sync(&vxlan->age_timer);
1413
1414 vxlan_flush(vxlan);
1415
1416 return 0;
1417 }
1418
1419 /* Stub, nothing needs to be done. */
1420 static void vxlan_set_multicast_list(struct net_device *dev)
1421 {
1422 }
1423
1424 static const struct net_device_ops vxlan_netdev_ops = {
1425 .ndo_init = vxlan_init,
1426 .ndo_uninit = vxlan_uninit,
1427 .ndo_open = vxlan_open,
1428 .ndo_stop = vxlan_stop,
1429 .ndo_start_xmit = vxlan_xmit,
1430 .ndo_get_stats64 = ip_tunnel_get_stats64,
1431 .ndo_set_rx_mode = vxlan_set_multicast_list,
1432 .ndo_change_mtu = eth_change_mtu,
1433 .ndo_validate_addr = eth_validate_addr,
1434 .ndo_set_mac_address = eth_mac_addr,
1435 .ndo_fdb_add = vxlan_fdb_add,
1436 .ndo_fdb_del = vxlan_fdb_delete,
1437 .ndo_fdb_dump = vxlan_fdb_dump,
1438 };
1439
1440 /* Info for udev, that this is a virtual tunnel endpoint */
1441 static struct device_type vxlan_type = {
1442 .name = "vxlan",
1443 };
1444
1445 /* Initialize the device structure. */
1446 static void vxlan_setup(struct net_device *dev)
1447 {
1448 struct vxlan_dev *vxlan = netdev_priv(dev);
1449 unsigned int h;
1450 int low, high;
1451
1452 eth_hw_addr_random(dev);
1453 ether_setup(dev);
1454 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
1455
1456 dev->netdev_ops = &vxlan_netdev_ops;
1457 dev->destructor = free_netdev;
1458 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1459
1460 dev->tx_queue_len = 0;
1461 dev->features |= NETIF_F_LLTX;
1462 dev->features |= NETIF_F_NETNS_LOCAL;
1463 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
1464 dev->features |= NETIF_F_RXCSUM;
1465 dev->features |= NETIF_F_GSO_SOFTWARE;
1466
1467 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
1468 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1469 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1470 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1471
1472 INIT_LIST_HEAD(&vxlan->next);
1473 spin_lock_init(&vxlan->hash_lock);
1474 INIT_WORK(&vxlan->igmp_work, vxlan_igmp_work);
1475 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
1476
1477 init_timer_deferrable(&vxlan->age_timer);
1478 vxlan->age_timer.function = vxlan_cleanup;
1479 vxlan->age_timer.data = (unsigned long) vxlan;
1480
1481 inet_get_local_port_range(&low, &high);
1482 vxlan->port_min = low;
1483 vxlan->port_max = high;
1484 vxlan->dst_port = htons(vxlan_port);
1485
1486 vxlan->dev = dev;
1487
1488 for (h = 0; h < FDB_HASH_SIZE; ++h)
1489 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1490 }
1491
1492 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1493 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
1494 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
1495 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1496 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1497 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1498 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1499 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1500 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1501 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
1502 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
1503 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1504 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1505 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1506 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
1507 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
1508 };
1509
1510 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1511 {
1512 if (tb[IFLA_ADDRESS]) {
1513 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1514 pr_debug("invalid link address (not ethernet)\n");
1515 return -EINVAL;
1516 }
1517
1518 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1519 pr_debug("invalid all zero ethernet address\n");
1520 return -EADDRNOTAVAIL;
1521 }
1522 }
1523
1524 if (!data)
1525 return -EINVAL;
1526
1527 if (data[IFLA_VXLAN_ID]) {
1528 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1529 if (id >= VXLAN_VID_MASK)
1530 return -ERANGE;
1531 }
1532
1533 if (data[IFLA_VXLAN_PORT_RANGE]) {
1534 const struct ifla_vxlan_port_range *p
1535 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1536
1537 if (ntohs(p->high) < ntohs(p->low)) {
1538 pr_debug("port range %u .. %u not valid\n",
1539 ntohs(p->low), ntohs(p->high));
1540 return -EINVAL;
1541 }
1542 }
1543
1544 return 0;
1545 }
1546
1547 static void vxlan_get_drvinfo(struct net_device *netdev,
1548 struct ethtool_drvinfo *drvinfo)
1549 {
1550 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1551 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1552 }
1553
1554 static const struct ethtool_ops vxlan_ethtool_ops = {
1555 .get_drvinfo = vxlan_get_drvinfo,
1556 .get_link = ethtool_op_get_link,
1557 };
1558
1559 static void vxlan_del_work(struct work_struct *work)
1560 {
1561 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
1562
1563 sk_release_kernel(vs->sock->sk);
1564 kfree_rcu(vs, rcu);
1565 }
1566
1567 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port)
1568 {
1569 struct vxlan_sock *vs;
1570 struct sock *sk;
1571 struct sockaddr_in vxlan_addr = {
1572 .sin_family = AF_INET,
1573 .sin_addr.s_addr = htonl(INADDR_ANY),
1574 .sin_port = port,
1575 };
1576 int rc;
1577 unsigned int h;
1578
1579 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
1580 if (!vs)
1581 return ERR_PTR(-ENOMEM);
1582
1583 for (h = 0; h < VNI_HASH_SIZE; ++h)
1584 INIT_HLIST_HEAD(&vs->vni_list[h]);
1585
1586 INIT_WORK(&vs->del_work, vxlan_del_work);
1587
1588 /* Create UDP socket for encapsulation receive. */
1589 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vs->sock);
1590 if (rc < 0) {
1591 pr_debug("UDP socket create failed\n");
1592 kfree(vs);
1593 return ERR_PTR(rc);
1594 }
1595
1596 /* Put in proper namespace */
1597 sk = vs->sock->sk;
1598 sk_change_net(sk, net);
1599
1600 rc = kernel_bind(vs->sock, (struct sockaddr *) &vxlan_addr,
1601 sizeof(vxlan_addr));
1602 if (rc < 0) {
1603 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1604 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
1605 sk_release_kernel(sk);
1606 kfree(vs);
1607 return ERR_PTR(rc);
1608 }
1609
1610 /* Disable multicast loopback */
1611 inet_sk(sk)->mc_loop = 0;
1612
1613 /* Mark socket as an encapsulation socket. */
1614 udp_sk(sk)->encap_type = 1;
1615 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1616 udp_encap_enable();
1617 atomic_set(&vs->refcnt, 1);
1618
1619 return vs;
1620 }
1621
1622 /* Scheduled at device creation to bind to a socket */
1623 static void vxlan_sock_work(struct work_struct *work)
1624 {
1625 struct vxlan_dev *vxlan
1626 = container_of(work, struct vxlan_dev, sock_work);
1627 struct net_device *dev = vxlan->dev;
1628 struct net *net = dev_net(dev);
1629 __u32 vni = vxlan->default_dst.remote_vni;
1630 __be16 port = vxlan->dst_port;
1631 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1632 struct vxlan_sock *nvs, *ovs;
1633
1634 nvs = vxlan_socket_create(net, port);
1635 if (IS_ERR(nvs)) {
1636 netdev_err(vxlan->dev, "Can not create UDP socket, %ld\n",
1637 PTR_ERR(nvs));
1638 goto out;
1639 }
1640
1641 spin_lock(&vn->sock_lock);
1642 /* Look again to see if can reuse socket */
1643 ovs = vxlan_find_port(net, port);
1644 if (ovs) {
1645 atomic_inc(&ovs->refcnt);
1646 vxlan->vn_sock = ovs;
1647 hlist_add_head_rcu(&vxlan->hlist, vni_head(ovs, vni));
1648 spin_unlock(&vn->sock_lock);
1649
1650 sk_release_kernel(nvs->sock->sk);
1651 kfree(nvs);
1652 } else {
1653 vxlan->vn_sock = nvs;
1654 hlist_add_head_rcu(&nvs->hlist, vs_head(net, port));
1655 hlist_add_head_rcu(&vxlan->hlist, vni_head(nvs, vni));
1656 spin_unlock(&vn->sock_lock);
1657 }
1658 out:
1659 dev_put(dev);
1660 }
1661
1662 static int vxlan_newlink(struct net *net, struct net_device *dev,
1663 struct nlattr *tb[], struct nlattr *data[])
1664 {
1665 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1666 struct vxlan_dev *vxlan = netdev_priv(dev);
1667 struct vxlan_rdst *dst = &vxlan->default_dst;
1668 __u32 vni;
1669 int err;
1670
1671 if (!data[IFLA_VXLAN_ID])
1672 return -EINVAL;
1673
1674 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
1675 dst->remote_vni = vni;
1676
1677 if (data[IFLA_VXLAN_GROUP])
1678 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]);
1679
1680 if (data[IFLA_VXLAN_LOCAL])
1681 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1682
1683 if (data[IFLA_VXLAN_LINK] &&
1684 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
1685 struct net_device *lowerdev
1686 = __dev_get_by_index(net, dst->remote_ifindex);
1687
1688 if (!lowerdev) {
1689 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
1690 return -ENODEV;
1691 }
1692
1693 if (!tb[IFLA_MTU])
1694 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1695
1696 /* update header length based on lower device */
1697 dev->hard_header_len = lowerdev->hard_header_len +
1698 VXLAN_HEADROOM;
1699 }
1700
1701 if (data[IFLA_VXLAN_TOS])
1702 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1703
1704 if (data[IFLA_VXLAN_TTL])
1705 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1706
1707 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
1708 vxlan->flags |= VXLAN_F_LEARN;
1709
1710 if (data[IFLA_VXLAN_AGEING])
1711 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1712 else
1713 vxlan->age_interval = FDB_AGE_DEFAULT;
1714
1715 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1716 vxlan->flags |= VXLAN_F_PROXY;
1717
1718 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1719 vxlan->flags |= VXLAN_F_RSC;
1720
1721 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1722 vxlan->flags |= VXLAN_F_L2MISS;
1723
1724 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1725 vxlan->flags |= VXLAN_F_L3MISS;
1726
1727 if (data[IFLA_VXLAN_LIMIT])
1728 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1729
1730 if (data[IFLA_VXLAN_PORT_RANGE]) {
1731 const struct ifla_vxlan_port_range *p
1732 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1733 vxlan->port_min = ntohs(p->low);
1734 vxlan->port_max = ntohs(p->high);
1735 }
1736
1737 if (data[IFLA_VXLAN_PORT])
1738 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
1739
1740 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
1741 pr_info("duplicate VNI %u\n", vni);
1742 return -EEXIST;
1743 }
1744
1745 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1746
1747 /* create an fdb entry for default destination */
1748 err = vxlan_fdb_create(vxlan, all_zeros_mac,
1749 vxlan->default_dst.remote_ip,
1750 NUD_REACHABLE|NUD_PERMANENT,
1751 NLM_F_EXCL|NLM_F_CREATE,
1752 vxlan->dst_port, vxlan->default_dst.remote_vni,
1753 vxlan->default_dst.remote_ifindex, NTF_SELF);
1754 if (err)
1755 return err;
1756
1757 err = register_netdevice(dev);
1758 if (err) {
1759 vxlan_fdb_delete_default(vxlan);
1760 return err;
1761 }
1762
1763 list_add(&vxlan->next, &vn->vxlan_list);
1764
1765 return 0;
1766 }
1767
1768 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1769 {
1770 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1771 struct vxlan_dev *vxlan = netdev_priv(dev);
1772
1773 flush_workqueue(vxlan_wq);
1774
1775 spin_lock(&vn->sock_lock);
1776 hlist_del_rcu(&vxlan->hlist);
1777 spin_unlock(&vn->sock_lock);
1778
1779 list_del(&vxlan->next);
1780 unregister_netdevice_queue(dev, head);
1781 }
1782
1783 static size_t vxlan_get_size(const struct net_device *dev)
1784 {
1785
1786 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
1787 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */
1788 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1789 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1790 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1791 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1792 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
1793 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1794 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1795 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1796 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
1797 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1798 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
1799 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
1800 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
1801 0;
1802 }
1803
1804 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1805 {
1806 const struct vxlan_dev *vxlan = netdev_priv(dev);
1807 const struct vxlan_rdst *dst = &vxlan->default_dst;
1808 struct ifla_vxlan_port_range ports = {
1809 .low = htons(vxlan->port_min),
1810 .high = htons(vxlan->port_max),
1811 };
1812
1813 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
1814 goto nla_put_failure;
1815
1816 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip))
1817 goto nla_put_failure;
1818
1819 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
1820 goto nla_put_failure;
1821
1822 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
1823 goto nla_put_failure;
1824
1825 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1826 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
1827 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1828 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1829 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1830 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1831 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1832 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1833 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1834 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1835 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
1836 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
1837 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
1838 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
1839 goto nla_put_failure;
1840
1841 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1842 goto nla_put_failure;
1843
1844 return 0;
1845
1846 nla_put_failure:
1847 return -EMSGSIZE;
1848 }
1849
1850 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1851 .kind = "vxlan",
1852 .maxtype = IFLA_VXLAN_MAX,
1853 .policy = vxlan_policy,
1854 .priv_size = sizeof(struct vxlan_dev),
1855 .setup = vxlan_setup,
1856 .validate = vxlan_validate,
1857 .newlink = vxlan_newlink,
1858 .dellink = vxlan_dellink,
1859 .get_size = vxlan_get_size,
1860 .fill_info = vxlan_fill_info,
1861 };
1862
1863 static __net_init int vxlan_init_net(struct net *net)
1864 {
1865 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1866 unsigned int h;
1867
1868 INIT_LIST_HEAD(&vn->vxlan_list);
1869 spin_lock_init(&vn->sock_lock);
1870
1871 for (h = 0; h < PORT_HASH_SIZE; ++h)
1872 INIT_HLIST_HEAD(&vn->sock_list[h]);
1873
1874 return 0;
1875 }
1876
1877 static __net_exit void vxlan_exit_net(struct net *net)
1878 {
1879 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1880 struct vxlan_dev *vxlan;
1881
1882 rtnl_lock();
1883 list_for_each_entry(vxlan, &vn->vxlan_list, next)
1884 dev_close(vxlan->dev);
1885 rtnl_unlock();
1886 }
1887
1888 static struct pernet_operations vxlan_net_ops = {
1889 .init = vxlan_init_net,
1890 .exit = vxlan_exit_net,
1891 .id = &vxlan_net_id,
1892 .size = sizeof(struct vxlan_net),
1893 };
1894
1895 static int __init vxlan_init_module(void)
1896 {
1897 int rc;
1898
1899 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
1900 if (!vxlan_wq)
1901 return -ENOMEM;
1902
1903 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1904
1905 rc = register_pernet_device(&vxlan_net_ops);
1906 if (rc)
1907 goto out1;
1908
1909 rc = rtnl_link_register(&vxlan_link_ops);
1910 if (rc)
1911 goto out2;
1912
1913 return 0;
1914
1915 out2:
1916 unregister_pernet_device(&vxlan_net_ops);
1917 out1:
1918 destroy_workqueue(vxlan_wq);
1919 return rc;
1920 }
1921 late_initcall(vxlan_init_module);
1922
1923 static void __exit vxlan_cleanup_module(void)
1924 {
1925 rtnl_link_unregister(&vxlan_link_ops);
1926 destroy_workqueue(vxlan_wq);
1927 unregister_pernet_device(&vxlan_net_ops);
1928 rcu_barrier();
1929 }
1930 module_exit(vxlan_cleanup_module);
1931
1932 MODULE_LICENSE("GPL");
1933 MODULE_VERSION(VXLAN_VERSION);
1934 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
1935 MODULE_ALIAS_RTNL_LINK("vxlan");
CRIS linux kernel tree