projects / deliverable / linux.git / blob
? search:


c3dd6e496327d1dca0886b2140d485ac27f194ed
[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 * 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_join;
140 struct work_struct igmp_leave;
141
142 unsigned long age_interval;
143 struct timer_list age_timer;
144 spinlock_t hash_lock;
145 unsigned int addrcnt;
146 unsigned int addrmax;
147
148 struct hlist_head fdb_head[FDB_HASH_SIZE];
149 };
150
151 #define VXLAN_F_LEARN 0x01
152 #define VXLAN_F_PROXY 0x02
153 #define VXLAN_F_RSC 0x04
154 #define VXLAN_F_L2MISS 0x08
155 #define VXLAN_F_L3MISS 0x10
156
157 /* salt for hash table */
158 static u32 vxlan_salt __read_mostly;
159 static struct workqueue_struct *vxlan_wq;
160
161 static void vxlan_sock_work(struct work_struct *work);
162
163 /* Virtual Network hash table head */
164 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
165 {
166 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
167 }
168
169 /* Socket hash table head */
170 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
171 {
172 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
173
174 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
175 }
176
177 /* First remote destination for a forwarding entry.
178 * Guaranteed to be non-NULL because remotes are never deleted.
179 */
180 static inline struct vxlan_rdst *first_remote(struct vxlan_fdb *fdb)
181 {
182 return list_first_or_null_rcu(&fdb->remotes, struct vxlan_rdst, list);
183 }
184
185 /* Find VXLAN socket based on network namespace and UDP port */
186 static struct vxlan_sock *vxlan_find_port(struct net *net, __be16 port)
187 {
188 struct vxlan_sock *vs;
189
190 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
191 if (inet_sk(vs->sock->sk)->inet_sport == port)
192 return vs;
193 }
194 return NULL;
195 }
196
197 /* Look up VNI in a per net namespace table */
198 static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
199 {
200 struct vxlan_sock *vs;
201 struct vxlan_dev *vxlan;
202
203 vs = vxlan_find_port(net, port);
204 if (!vs)
205 return NULL;
206
207 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
208 if (vxlan->default_dst.remote_vni == id)
209 return vxlan;
210 }
211
212 return NULL;
213 }
214
215 /* Fill in neighbour message in skbuff. */
216 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
217 const struct vxlan_fdb *fdb,
218 u32 portid, u32 seq, int type, unsigned int flags,
219 const struct vxlan_rdst *rdst)
220 {
221 unsigned long now = jiffies;
222 struct nda_cacheinfo ci;
223 struct nlmsghdr *nlh;
224 struct ndmsg *ndm;
225 bool send_ip, send_eth;
226
227 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
228 if (nlh == NULL)
229 return -EMSGSIZE;
230
231 ndm = nlmsg_data(nlh);
232 memset(ndm, 0, sizeof(*ndm));
233
234 send_eth = send_ip = true;
235
236 if (type == RTM_GETNEIGH) {
237 ndm->ndm_family = AF_INET;
238 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
239 send_eth = !is_zero_ether_addr(fdb->eth_addr);
240 } else
241 ndm->ndm_family = AF_BRIDGE;
242 ndm->ndm_state = fdb->state;
243 ndm->ndm_ifindex = vxlan->dev->ifindex;
244 ndm->ndm_flags = fdb->flags;
245 ndm->ndm_type = NDA_DST;
246
247 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
248 goto nla_put_failure;
249
250 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
251 goto nla_put_failure;
252
253 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
254 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
255 goto nla_put_failure;
256 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
257 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
258 goto nla_put_failure;
259 if (rdst->remote_ifindex &&
260 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
261 goto nla_put_failure;
262
263 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
264 ci.ndm_confirmed = 0;
265 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
266 ci.ndm_refcnt = 0;
267
268 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
269 goto nla_put_failure;
270
271 return nlmsg_end(skb, nlh);
272
273 nla_put_failure:
274 nlmsg_cancel(skb, nlh);
275 return -EMSGSIZE;
276 }
277
278 static inline size_t vxlan_nlmsg_size(void)
279 {
280 return NLMSG_ALIGN(sizeof(struct ndmsg))
281 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
282 + nla_total_size(sizeof(__be32)) /* NDA_DST */
283 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
284 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
285 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
286 + nla_total_size(sizeof(struct nda_cacheinfo));
287 }
288
289 static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
290 struct vxlan_fdb *fdb, int type)
291 {
292 struct net *net = dev_net(vxlan->dev);
293 struct sk_buff *skb;
294 int err = -ENOBUFS;
295
296 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
297 if (skb == NULL)
298 goto errout;
299
300 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, first_remote(fdb));
301 if (err < 0) {
302 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
303 WARN_ON(err == -EMSGSIZE);
304 kfree_skb(skb);
305 goto errout;
306 }
307
308 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
309 return;
310 errout:
311 if (err < 0)
312 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
313 }
314
315 static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
316 {
317 struct vxlan_dev *vxlan = netdev_priv(dev);
318 struct vxlan_fdb f = {
319 .state = NUD_STALE,
320 };
321 struct vxlan_rdst remote = {
322 .remote_ip = ipa, /* goes to NDA_DST */
323 .remote_vni = VXLAN_N_VID,
324 };
325
326 INIT_LIST_HEAD(&f.remotes);
327 list_add_rcu(&remote.list, &f.remotes);
328
329 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
330 }
331
332 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
333 {
334 struct vxlan_fdb f = {
335 .state = NUD_STALE,
336 };
337
338 INIT_LIST_HEAD(&f.remotes);
339 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
340
341 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
342 }
343
344 /* Hash Ethernet address */
345 static u32 eth_hash(const unsigned char *addr)
346 {
347 u64 value = get_unaligned((u64 *)addr);
348
349 /* only want 6 bytes */
350 #ifdef __BIG_ENDIAN
351 value >>= 16;
352 #else
353 value <<= 16;
354 #endif
355 return hash_64(value, FDB_HASH_BITS);
356 }
357
358 /* Hash chain to use given mac address */
359 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
360 const u8 *mac)
361 {
362 return &vxlan->fdb_head[eth_hash(mac)];
363 }
364
365 /* Look up Ethernet address in forwarding table */
366 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
367 const u8 *mac)
368
369 {
370 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
371 struct vxlan_fdb *f;
372
373 hlist_for_each_entry_rcu(f, head, hlist) {
374 if (compare_ether_addr(mac, f->eth_addr) == 0)
375 return f;
376 }
377
378 return NULL;
379 }
380
381 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
382 const u8 *mac)
383 {
384 struct vxlan_fdb *f;
385
386 f = __vxlan_find_mac(vxlan, mac);
387 if (f)
388 f->used = jiffies;
389
390 return f;
391 }
392
393 /* caller should hold vxlan->hash_lock */
394 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
395 __be32 ip, __be16 port,
396 __u32 vni, __u32 ifindex)
397 {
398 struct vxlan_rdst *rd;
399
400 list_for_each_entry(rd, &f->remotes, list) {
401 if (rd->remote_ip == ip &&
402 rd->remote_port == port &&
403 rd->remote_vni == vni &&
404 rd->remote_ifindex == ifindex)
405 return rd;
406 }
407
408 return NULL;
409 }
410
411 /* Replace destination of unicast mac */
412 static int vxlan_fdb_replace(struct vxlan_fdb *f,
413 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
414 {
415 struct vxlan_rdst *rd;
416
417 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
418 if (rd)
419 return 0;
420
421 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
422 if (!rd)
423 return 0;
424 rd->remote_ip = ip;
425 rd->remote_port = port;
426 rd->remote_vni = vni;
427 rd->remote_ifindex = ifindex;
428 return 1;
429 }
430
431 /* Add/update destinations for multicast */
432 static int vxlan_fdb_append(struct vxlan_fdb *f,
433 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
434 {
435 struct vxlan_rdst *rd;
436
437 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
438 if (rd)
439 return 0;
440
441 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
442 if (rd == NULL)
443 return -ENOBUFS;
444 rd->remote_ip = ip;
445 rd->remote_port = port;
446 rd->remote_vni = vni;
447 rd->remote_ifindex = ifindex;
448
449 list_add_tail_rcu(&rd->list, &f->remotes);
450
451 return 1;
452 }
453
454 /* Add new entry to forwarding table -- assumes lock held */
455 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
456 const u8 *mac, __be32 ip,
457 __u16 state, __u16 flags,
458 __be16 port, __u32 vni, __u32 ifindex,
459 __u8 ndm_flags)
460 {
461 struct vxlan_fdb *f;
462 int notify = 0;
463
464 f = __vxlan_find_mac(vxlan, mac);
465 if (f) {
466 if (flags & NLM_F_EXCL) {
467 netdev_dbg(vxlan->dev,
468 "lost race to create %pM\n", mac);
469 return -EEXIST;
470 }
471 if (f->state != state) {
472 f->state = state;
473 f->updated = jiffies;
474 notify = 1;
475 }
476 if (f->flags != ndm_flags) {
477 f->flags = ndm_flags;
478 f->updated = jiffies;
479 notify = 1;
480 }
481 if ((flags & NLM_F_REPLACE)) {
482 /* Only change unicasts */
483 if (!(is_multicast_ether_addr(f->eth_addr) ||
484 is_zero_ether_addr(f->eth_addr))) {
485 int rc = vxlan_fdb_replace(f, ip, port, vni,
486 ifindex);
487
488 if (rc < 0)
489 return rc;
490 notify |= rc;
491 } else
492 return -EOPNOTSUPP;
493 }
494 if ((flags & NLM_F_APPEND) &&
495 (is_multicast_ether_addr(f->eth_addr) ||
496 is_zero_ether_addr(f->eth_addr))) {
497 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
498
499 if (rc < 0)
500 return rc;
501 notify |= rc;
502 }
503 } else {
504 if (!(flags & NLM_F_CREATE))
505 return -ENOENT;
506
507 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
508 return -ENOSPC;
509
510 /* Disallow replace to add a multicast entry */
511 if ((flags & NLM_F_REPLACE) &&
512 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
513 return -EOPNOTSUPP;
514
515 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
516 f = kmalloc(sizeof(*f), GFP_ATOMIC);
517 if (!f)
518 return -ENOMEM;
519
520 notify = 1;
521 f->state = state;
522 f->flags = ndm_flags;
523 f->updated = f->used = jiffies;
524 INIT_LIST_HEAD(&f->remotes);
525 memcpy(f->eth_addr, mac, ETH_ALEN);
526
527 vxlan_fdb_append(f, ip, port, vni, ifindex);
528
529 ++vxlan->addrcnt;
530 hlist_add_head_rcu(&f->hlist,
531 vxlan_fdb_head(vxlan, mac));
532 }
533
534 if (notify)
535 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
536
537 return 0;
538 }
539
540 static void vxlan_fdb_free_rdst(struct rcu_head *head)
541 {
542 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
543 kfree(rd);
544 }
545
546 static void vxlan_fdb_free(struct rcu_head *head)
547 {
548 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
549 struct vxlan_rdst *rd, *nd;
550
551 list_for_each_entry_safe(rd, nd, &f->remotes, list)
552 kfree(rd);
553 kfree(f);
554 }
555
556 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
557 {
558 netdev_dbg(vxlan->dev,
559 "delete %pM\n", f->eth_addr);
560
561 --vxlan->addrcnt;
562 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
563
564 hlist_del_rcu(&f->hlist);
565 call_rcu(&f->rcu, vxlan_fdb_free);
566 }
567
568 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
569 __be32 *ip, __be16 *port, u32 *vni, u32 *ifindex)
570 {
571 struct net *net = dev_net(vxlan->dev);
572
573 if (tb[NDA_DST]) {
574 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
575 return -EAFNOSUPPORT;
576
577 *ip = nla_get_be32(tb[NDA_DST]);
578 } else {
579 *ip = htonl(INADDR_ANY);
580 }
581
582 if (tb[NDA_PORT]) {
583 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
584 return -EINVAL;
585 *port = nla_get_be16(tb[NDA_PORT]);
586 } else {
587 *port = vxlan->dst_port;
588 }
589
590 if (tb[NDA_VNI]) {
591 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
592 return -EINVAL;
593 *vni = nla_get_u32(tb[NDA_VNI]);
594 } else {
595 *vni = vxlan->default_dst.remote_vni;
596 }
597
598 if (tb[NDA_IFINDEX]) {
599 struct net_device *tdev;
600
601 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
602 return -EINVAL;
603 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
604 tdev = dev_get_by_index(net, *ifindex);
605 if (!tdev)
606 return -EADDRNOTAVAIL;
607 dev_put(tdev);
608 } else {
609 *ifindex = 0;
610 }
611
612 return 0;
613 }
614
615 /* Add static entry (via netlink) */
616 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
617 struct net_device *dev,
618 const unsigned char *addr, u16 flags)
619 {
620 struct vxlan_dev *vxlan = netdev_priv(dev);
621 /* struct net *net = dev_net(vxlan->dev); */
622 __be32 ip;
623 __be16 port;
624 u32 vni, ifindex;
625 int err;
626
627 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
628 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
629 ndm->ndm_state);
630 return -EINVAL;
631 }
632
633 if (tb[NDA_DST] == NULL)
634 return -EINVAL;
635
636 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
637 if (err)
638 return err;
639
640 spin_lock_bh(&vxlan->hash_lock);
641 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags,
642 port, vni, ifindex, ndm->ndm_flags);
643 spin_unlock_bh(&vxlan->hash_lock);
644
645 return err;
646 }
647
648 /* Delete entry (via netlink) */
649 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
650 struct net_device *dev,
651 const unsigned char *addr)
652 {
653 struct vxlan_dev *vxlan = netdev_priv(dev);
654 struct vxlan_fdb *f;
655 struct vxlan_rdst *rd = NULL;
656 __be32 ip;
657 __be16 port;
658 u32 vni, ifindex;
659 int err;
660
661 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
662 if (err)
663 return err;
664
665 err = -ENOENT;
666
667 spin_lock_bh(&vxlan->hash_lock);
668 f = vxlan_find_mac(vxlan, addr);
669 if (!f)
670 goto out;
671
672 if (ip != htonl(INADDR_ANY)) {
673 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
674 if (!rd)
675 goto out;
676 }
677
678 err = 0;
679
680 /* remove a destination if it's not the only one on the list,
681 * otherwise destroy the fdb entry
682 */
683 if (rd && !list_is_singular(&f->remotes)) {
684 list_del_rcu(&rd->list);
685 call_rcu(&rd->rcu, vxlan_fdb_free_rdst);
686 goto out;
687 }
688
689 vxlan_fdb_destroy(vxlan, f);
690
691 out:
692 spin_unlock_bh(&vxlan->hash_lock);
693
694 return err;
695 }
696
697 /* Dump forwarding table */
698 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
699 struct net_device *dev, int idx)
700 {
701 struct vxlan_dev *vxlan = netdev_priv(dev);
702 unsigned int h;
703
704 for (h = 0; h < FDB_HASH_SIZE; ++h) {
705 struct vxlan_fdb *f;
706 int err;
707
708 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
709 struct vxlan_rdst *rd;
710
711 if (idx < cb->args[0])
712 goto skip;
713
714 list_for_each_entry_rcu(rd, &f->remotes, list) {
715 err = vxlan_fdb_info(skb, vxlan, f,
716 NETLINK_CB(cb->skb).portid,
717 cb->nlh->nlmsg_seq,
718 RTM_NEWNEIGH,
719 NLM_F_MULTI, rd);
720 if (err < 0)
721 goto out;
722 }
723 skip:
724 ++idx;
725 }
726 }
727 out:
728 return idx;
729 }
730
731 /* Watch incoming packets to learn mapping between Ethernet address
732 * and Tunnel endpoint.
733 * Return true if packet is bogus and should be droppped.
734 */
735 static bool vxlan_snoop(struct net_device *dev,
736 __be32 src_ip, const u8 *src_mac)
737 {
738 struct vxlan_dev *vxlan = netdev_priv(dev);
739 struct vxlan_fdb *f;
740
741 f = vxlan_find_mac(vxlan, src_mac);
742 if (likely(f)) {
743 struct vxlan_rdst *rdst = first_remote(f);
744
745 if (likely(rdst->remote_ip == src_ip))
746 return false;
747
748 /* Don't migrate static entries, drop packets */
749 if (f->state & NUD_NOARP)
750 return true;
751
752 if (net_ratelimit())
753 netdev_info(dev,
754 "%pM migrated from %pI4 to %pI4\n",
755 src_mac, &rdst->remote_ip, &src_ip);
756
757 rdst->remote_ip = src_ip;
758 f->updated = jiffies;
759 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
760 } else {
761 /* learned new entry */
762 spin_lock(&vxlan->hash_lock);
763
764 /* close off race between vxlan_flush and incoming packets */
765 if (netif_running(dev))
766 vxlan_fdb_create(vxlan, src_mac, src_ip,
767 NUD_REACHABLE,
768 NLM_F_EXCL|NLM_F_CREATE,
769 vxlan->dst_port,
770 vxlan->default_dst.remote_vni,
771 0, NTF_SELF);
772 spin_unlock(&vxlan->hash_lock);
773 }
774
775 return false;
776 }
777
778 /* See if multicast group is already in use by other ID */
779 static bool vxlan_group_used(struct vxlan_net *vn, __be32 remote_ip)
780 {
781 struct vxlan_dev *vxlan;
782
783 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
784 if (!netif_running(vxlan->dev))
785 continue;
786
787 if (vxlan->default_dst.remote_ip == remote_ip)
788 return true;
789 }
790
791 return false;
792 }
793
794 static void vxlan_sock_hold(struct vxlan_sock *vs)
795 {
796 atomic_inc(&vs->refcnt);
797 }
798
799 static void vxlan_sock_release(struct vxlan_net *vn, struct vxlan_sock *vs)
800 {
801 if (!atomic_dec_and_test(&vs->refcnt))
802 return;
803
804 spin_lock(&vn->sock_lock);
805 hlist_del_rcu(&vs->hlist);
806 spin_unlock(&vn->sock_lock);
807
808 queue_work(vxlan_wq, &vs->del_work);
809 }
810
811 /* Callback to update multicast group membership when first VNI on
812 * multicast asddress is brought up
813 * Done as workqueue because ip_mc_join_group acquires RTNL.
814 */
815 static void vxlan_igmp_join(struct work_struct *work)
816 {
817 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
818 struct vxlan_net *vn = net_generic(dev_net(vxlan->dev), vxlan_net_id);
819 struct vxlan_sock *vs = vxlan->vn_sock;
820 struct sock *sk = vs->sock->sk;
821 struct ip_mreqn mreq = {
822 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
823 .imr_ifindex = vxlan->default_dst.remote_ifindex,
824 };
825
826 lock_sock(sk);
827 ip_mc_join_group(sk, &mreq);
828 release_sock(sk);
829
830 vxlan_sock_release(vn, vs);
831 dev_put(vxlan->dev);
832 }
833
834 /* Inverse of vxlan_igmp_join when last VNI is brought down */
835 static void vxlan_igmp_leave(struct work_struct *work)
836 {
837 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
838 struct vxlan_net *vn = net_generic(dev_net(vxlan->dev), vxlan_net_id);
839 struct vxlan_sock *vs = vxlan->vn_sock;
840 struct sock *sk = vs->sock->sk;
841 struct ip_mreqn mreq = {
842 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
843 .imr_ifindex = vxlan->default_dst.remote_ifindex,
844 };
845
846 lock_sock(sk);
847 ip_mc_leave_group(sk, &mreq);
848 release_sock(sk);
849
850 vxlan_sock_release(vn, vs);
851 dev_put(vxlan->dev);
852 }
853
854 /* Callback from net/ipv4/udp.c to receive packets */
855 static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
856 {
857 struct iphdr *oip;
858 struct vxlanhdr *vxh;
859 struct vxlan_dev *vxlan;
860 struct pcpu_tstats *stats;
861 __be16 port;
862 __u32 vni;
863 int err;
864
865 /* pop off outer UDP header */
866 __skb_pull(skb, sizeof(struct udphdr));
867
868 /* Need Vxlan and inner Ethernet header to be present */
869 if (!pskb_may_pull(skb, sizeof(struct vxlanhdr)))
870 goto error;
871
872 /* Drop packets with reserved bits set */
873 vxh = (struct vxlanhdr *) skb->data;
874 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
875 (vxh->vx_vni & htonl(0xff))) {
876 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
877 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
878 goto error;
879 }
880
881 __skb_pull(skb, sizeof(struct vxlanhdr));
882
883 /* Is this VNI defined? */
884 vni = ntohl(vxh->vx_vni) >> 8;
885 port = inet_sk(sk)->inet_sport;
886 vxlan = vxlan_find_vni(sock_net(sk), vni, port);
887 if (!vxlan) {
888 netdev_dbg(skb->dev, "unknown vni %d port %u\n",
889 vni, ntohs(port));
890 goto drop;
891 }
892
893 if (!pskb_may_pull(skb, ETH_HLEN)) {
894 vxlan->dev->stats.rx_length_errors++;
895 vxlan->dev->stats.rx_errors++;
896 goto drop;
897 }
898
899 skb_reset_mac_header(skb);
900
901 /* Re-examine inner Ethernet packet */
902 oip = ip_hdr(skb);
903 skb->protocol = eth_type_trans(skb, vxlan->dev);
904
905 /* Ignore packet loops (and multicast echo) */
906 if (compare_ether_addr(eth_hdr(skb)->h_source,
907 vxlan->dev->dev_addr) == 0)
908 goto drop;
909
910 if ((vxlan->flags & VXLAN_F_LEARN) &&
911 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source))
912 goto drop;
913
914 __skb_tunnel_rx(skb, vxlan->dev);
915 skb_reset_network_header(skb);
916
917 /* If the NIC driver gave us an encapsulated packet with
918 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
919 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
920 * for us. Otherwise force the upper layers to verify it.
921 */
922 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
923 !(vxlan->dev->features & NETIF_F_RXCSUM))
924 skb->ip_summed = CHECKSUM_NONE;
925
926 skb->encapsulation = 0;
927
928 err = IP_ECN_decapsulate(oip, skb);
929 if (unlikely(err)) {
930 if (log_ecn_error)
931 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
932 &oip->saddr, oip->tos);
933 if (err > 1) {
934 ++vxlan->dev->stats.rx_frame_errors;
935 ++vxlan->dev->stats.rx_errors;
936 goto drop;
937 }
938 }
939
940 stats = this_cpu_ptr(vxlan->dev->tstats);
941 u64_stats_update_begin(&stats->syncp);
942 stats->rx_packets++;
943 stats->rx_bytes += skb->len;
944 u64_stats_update_end(&stats->syncp);
945
946 netif_rx(skb);
947
948 return 0;
949 error:
950 /* Put UDP header back */
951 __skb_push(skb, sizeof(struct udphdr));
952
953 return 1;
954 drop:
955 /* Consume bad packet */
956 kfree_skb(skb);
957 return 0;
958 }
959
960 static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
961 {
962 struct vxlan_dev *vxlan = netdev_priv(dev);
963 struct arphdr *parp;
964 u8 *arpptr, *sha;
965 __be32 sip, tip;
966 struct neighbour *n;
967
968 if (dev->flags & IFF_NOARP)
969 goto out;
970
971 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
972 dev->stats.tx_dropped++;
973 goto out;
974 }
975 parp = arp_hdr(skb);
976
977 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
978 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
979 parp->ar_pro != htons(ETH_P_IP) ||
980 parp->ar_op != htons(ARPOP_REQUEST) ||
981 parp->ar_hln != dev->addr_len ||
982 parp->ar_pln != 4)
983 goto out;
984 arpptr = (u8 *)parp + sizeof(struct arphdr);
985 sha = arpptr;
986 arpptr += dev->addr_len; /* sha */
987 memcpy(&sip, arpptr, sizeof(sip));
988 arpptr += sizeof(sip);
989 arpptr += dev->addr_len; /* tha */
990 memcpy(&tip, arpptr, sizeof(tip));
991
992 if (ipv4_is_loopback(tip) ||
993 ipv4_is_multicast(tip))
994 goto out;
995
996 n = neigh_lookup(&arp_tbl, &tip, dev);
997
998 if (n) {
999 struct vxlan_fdb *f;
1000 struct sk_buff *reply;
1001
1002 if (!(n->nud_state & NUD_CONNECTED)) {
1003 neigh_release(n);
1004 goto out;
1005 }
1006
1007 f = vxlan_find_mac(vxlan, n->ha);
1008 if (f && first_remote(f)->remote_ip == htonl(INADDR_ANY)) {
1009 /* bridge-local neighbor */
1010 neigh_release(n);
1011 goto out;
1012 }
1013
1014 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1015 n->ha, sha);
1016
1017 neigh_release(n);
1018
1019 skb_reset_mac_header(reply);
1020 __skb_pull(reply, skb_network_offset(reply));
1021 reply->ip_summed = CHECKSUM_UNNECESSARY;
1022 reply->pkt_type = PACKET_HOST;
1023
1024 if (netif_rx_ni(reply) == NET_RX_DROP)
1025 dev->stats.rx_dropped++;
1026 } else if (vxlan->flags & VXLAN_F_L3MISS)
1027 vxlan_ip_miss(dev, tip);
1028 out:
1029 consume_skb(skb);
1030 return NETDEV_TX_OK;
1031 }
1032
1033 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1034 {
1035 struct vxlan_dev *vxlan = netdev_priv(dev);
1036 struct neighbour *n;
1037 struct iphdr *pip;
1038
1039 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1040 return false;
1041
1042 n = NULL;
1043 switch (ntohs(eth_hdr(skb)->h_proto)) {
1044 case ETH_P_IP:
1045 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1046 return false;
1047 pip = ip_hdr(skb);
1048 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1049 break;
1050 default:
1051 return false;
1052 }
1053
1054 if (n) {
1055 bool diff;
1056
1057 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
1058 if (diff) {
1059 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1060 dev->addr_len);
1061 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1062 }
1063 neigh_release(n);
1064 return diff;
1065 } else if (vxlan->flags & VXLAN_F_L3MISS)
1066 vxlan_ip_miss(dev, pip->daddr);
1067 return false;
1068 }
1069
1070 static void vxlan_sock_put(struct sk_buff *skb)
1071 {
1072 sock_put(skb->sk);
1073 }
1074
1075 /* On transmit, associate with the tunnel socket */
1076 static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
1077 {
1078 struct vxlan_dev *vxlan = netdev_priv(dev);
1079 struct sock *sk = vxlan->vn_sock->sock->sk;
1080
1081 skb_orphan(skb);
1082 sock_hold(sk);
1083 skb->sk = sk;
1084 skb->destructor = vxlan_sock_put;
1085 }
1086
1087 /* Compute source port for outgoing packet
1088 * first choice to use L4 flow hash since it will spread
1089 * better and maybe available from hardware
1090 * secondary choice is to use jhash on the Ethernet header
1091 */
1092 static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
1093 {
1094 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
1095 u32 hash;
1096
1097 hash = skb_get_rxhash(skb);
1098 if (!hash)
1099 hash = jhash(skb->data, 2 * ETH_ALEN,
1100 (__force u32) skb->protocol);
1101
1102 return htons((((u64) hash * range) >> 32) + vxlan->port_min);
1103 }
1104
1105 static int handle_offloads(struct sk_buff *skb)
1106 {
1107 if (skb_is_gso(skb)) {
1108 int err = skb_unclone(skb, GFP_ATOMIC);
1109 if (unlikely(err))
1110 return err;
1111
1112 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
1113 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1114 skb->ip_summed = CHECKSUM_NONE;
1115
1116 return 0;
1117 }
1118
1119 /* Bypass encapsulation if the destination is local */
1120 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1121 struct vxlan_dev *dst_vxlan)
1122 {
1123 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1124 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1125
1126 skb->pkt_type = PACKET_HOST;
1127 skb->encapsulation = 0;
1128 skb->dev = dst_vxlan->dev;
1129 __skb_pull(skb, skb_network_offset(skb));
1130
1131 if (dst_vxlan->flags & VXLAN_F_LEARN)
1132 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
1133 eth_hdr(skb)->h_source);
1134
1135 u64_stats_update_begin(&tx_stats->syncp);
1136 tx_stats->tx_packets++;
1137 tx_stats->tx_bytes += skb->len;
1138 u64_stats_update_end(&tx_stats->syncp);
1139
1140 if (netif_rx(skb) == NET_RX_SUCCESS) {
1141 u64_stats_update_begin(&rx_stats->syncp);
1142 rx_stats->rx_packets++;
1143 rx_stats->rx_bytes += skb->len;
1144 u64_stats_update_end(&rx_stats->syncp);
1145 } else {
1146 skb->dev->stats.rx_dropped++;
1147 }
1148 }
1149
1150 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1151 struct vxlan_rdst *rdst, bool did_rsc)
1152 {
1153 struct vxlan_dev *vxlan = netdev_priv(dev);
1154 struct rtable *rt;
1155 const struct iphdr *old_iph;
1156 struct vxlanhdr *vxh;
1157 struct udphdr *uh;
1158 struct flowi4 fl4;
1159 __be32 dst;
1160 __be16 src_port, dst_port;
1161 u32 vni;
1162 __be16 df = 0;
1163 __u8 tos, ttl;
1164 int err;
1165
1166 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
1167 vni = rdst->remote_vni;
1168 dst = rdst->remote_ip;
1169
1170 if (!dst) {
1171 if (did_rsc) {
1172 /* short-circuited back to local bridge */
1173 vxlan_encap_bypass(skb, vxlan, vxlan);
1174 return;
1175 }
1176 goto drop;
1177 }
1178
1179 if (!skb->encapsulation) {
1180 skb_reset_inner_headers(skb);
1181 skb->encapsulation = 1;
1182 }
1183
1184 /* Need space for new headers (invalidates iph ptr) */
1185 if (skb_cow_head(skb, VXLAN_HEADROOM))
1186 goto drop;
1187
1188 old_iph = ip_hdr(skb);
1189
1190 ttl = vxlan->ttl;
1191 if (!ttl && IN_MULTICAST(ntohl(dst)))
1192 ttl = 1;
1193
1194 tos = vxlan->tos;
1195 if (tos == 1)
1196 tos = ip_tunnel_get_dsfield(old_iph, skb);
1197
1198 src_port = vxlan_src_port(vxlan, skb);
1199
1200 memset(&fl4, 0, sizeof(fl4));
1201 fl4.flowi4_oif = rdst->remote_ifindex;
1202 fl4.flowi4_tos = RT_TOS(tos);
1203 fl4.daddr = dst;
1204 fl4.saddr = vxlan->saddr;
1205
1206 rt = ip_route_output_key(dev_net(dev), &fl4);
1207 if (IS_ERR(rt)) {
1208 netdev_dbg(dev, "no route to %pI4\n", &dst);
1209 dev->stats.tx_carrier_errors++;
1210 goto tx_error;
1211 }
1212
1213 if (rt->dst.dev == dev) {
1214 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1215 ip_rt_put(rt);
1216 dev->stats.collisions++;
1217 goto tx_error;
1218 }
1219
1220 /* Bypass encapsulation if the destination is local */
1221 if (rt->rt_flags & RTCF_LOCAL &&
1222 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1223 struct vxlan_dev *dst_vxlan;
1224
1225 ip_rt_put(rt);
1226 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
1227 if (!dst_vxlan)
1228 goto tx_error;
1229 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1230 return;
1231 }
1232 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1233 vxh->vx_flags = htonl(VXLAN_FLAGS);
1234 vxh->vx_vni = htonl(vni << 8);
1235
1236 __skb_push(skb, sizeof(*uh));
1237 skb_reset_transport_header(skb);
1238 uh = udp_hdr(skb);
1239
1240 uh->dest = dst_port;
1241 uh->source = src_port;
1242
1243 uh->len = htons(skb->len);
1244 uh->check = 0;
1245
1246 vxlan_set_owner(dev, skb);
1247
1248 if (handle_offloads(skb))
1249 goto drop;
1250
1251 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1252 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1253
1254 err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, dst,
1255 IPPROTO_UDP, tos, ttl, df);
1256 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1257
1258 return;
1259
1260 drop:
1261 dev->stats.tx_dropped++;
1262 goto tx_free;
1263
1264 tx_error:
1265 dev->stats.tx_errors++;
1266 tx_free:
1267 dev_kfree_skb(skb);
1268 }
1269
1270 /* Transmit local packets over Vxlan
1271 *
1272 * Outer IP header inherits ECN and DF from inner header.
1273 * Outer UDP destination is the VXLAN assigned port.
1274 * source port is based on hash of flow
1275 */
1276 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1277 {
1278 struct vxlan_dev *vxlan = netdev_priv(dev);
1279 struct ethhdr *eth;
1280 bool did_rsc = false;
1281 struct vxlan_rdst *rdst;
1282 struct vxlan_fdb *f;
1283
1284 skb_reset_mac_header(skb);
1285 eth = eth_hdr(skb);
1286
1287 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1288 return arp_reduce(dev, skb);
1289
1290 f = vxlan_find_mac(vxlan, eth->h_dest);
1291 did_rsc = false;
1292
1293 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1294 ntohs(eth->h_proto) == ETH_P_IP) {
1295 did_rsc = route_shortcircuit(dev, skb);
1296 if (did_rsc)
1297 f = vxlan_find_mac(vxlan, eth->h_dest);
1298 }
1299
1300 if (f == NULL) {
1301 f = vxlan_find_mac(vxlan, all_zeros_mac);
1302 if (f == NULL) {
1303 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1304 !is_multicast_ether_addr(eth->h_dest))
1305 vxlan_fdb_miss(vxlan, eth->h_dest);
1306
1307 dev->stats.tx_dropped++;
1308 dev_kfree_skb(skb);
1309 return NETDEV_TX_OK;
1310 }
1311 }
1312
1313 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1314 struct sk_buff *skb1;
1315
1316 skb1 = skb_clone(skb, GFP_ATOMIC);
1317 if (skb1)
1318 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1319 }
1320
1321 dev_kfree_skb(skb);
1322 return NETDEV_TX_OK;
1323 }
1324
1325 /* Walk the forwarding table and purge stale entries */
1326 static void vxlan_cleanup(unsigned long arg)
1327 {
1328 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1329 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1330 unsigned int h;
1331
1332 if (!netif_running(vxlan->dev))
1333 return;
1334
1335 spin_lock_bh(&vxlan->hash_lock);
1336 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1337 struct hlist_node *p, *n;
1338 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1339 struct vxlan_fdb *f
1340 = container_of(p, struct vxlan_fdb, hlist);
1341 unsigned long timeout;
1342
1343 if (f->state & NUD_PERMANENT)
1344 continue;
1345
1346 timeout = f->used + vxlan->age_interval * HZ;
1347 if (time_before_eq(timeout, jiffies)) {
1348 netdev_dbg(vxlan->dev,
1349 "garbage collect %pM\n",
1350 f->eth_addr);
1351 f->state = NUD_STALE;
1352 vxlan_fdb_destroy(vxlan, f);
1353 } else if (time_before(timeout, next_timer))
1354 next_timer = timeout;
1355 }
1356 }
1357 spin_unlock_bh(&vxlan->hash_lock);
1358
1359 mod_timer(&vxlan->age_timer, next_timer);
1360 }
1361
1362 /* Setup stats when device is created */
1363 static int vxlan_init(struct net_device *dev)
1364 {
1365 struct vxlan_dev *vxlan = netdev_priv(dev);
1366 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1367 struct vxlan_sock *vs;
1368 __u32 vni = vxlan->default_dst.remote_vni;
1369
1370 dev->tstats = alloc_percpu(struct pcpu_tstats);
1371 if (!dev->tstats)
1372 return -ENOMEM;
1373
1374 spin_lock(&vn->sock_lock);
1375 vs = vxlan_find_port(dev_net(dev), vxlan->dst_port);
1376 if (vs) {
1377 /* If we have a socket with same port already, reuse it */
1378 atomic_inc(&vs->refcnt);
1379 vxlan->vn_sock = vs;
1380 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1381 } else {
1382 /* otherwise make new socket outside of RTNL */
1383 dev_hold(dev);
1384 queue_work(vxlan_wq, &vxlan->sock_work);
1385 }
1386 spin_unlock(&vn->sock_lock);
1387
1388 return 0;
1389 }
1390
1391 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
1392 {
1393 struct vxlan_fdb *f;
1394
1395 spin_lock_bh(&vxlan->hash_lock);
1396 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1397 if (f)
1398 vxlan_fdb_destroy(vxlan, f);
1399 spin_unlock_bh(&vxlan->hash_lock);
1400 }
1401
1402 static void vxlan_uninit(struct net_device *dev)
1403 {
1404 struct vxlan_dev *vxlan = netdev_priv(dev);
1405 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1406 struct vxlan_sock *vs = vxlan->vn_sock;
1407
1408 vxlan_fdb_delete_default(vxlan);
1409
1410 if (vs)
1411 vxlan_sock_release(vn, vs);
1412 free_percpu(dev->tstats);
1413 }
1414
1415 /* Start ageing timer and join group when device is brought up */
1416 static int vxlan_open(struct net_device *dev)
1417 {
1418 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1419 struct vxlan_dev *vxlan = netdev_priv(dev);
1420 struct vxlan_sock *vs = vxlan->vn_sock;
1421
1422 /* socket hasn't been created */
1423 if (!vs)
1424 return -ENOTCONN;
1425
1426 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)) &&
1427 ! vxlan_group_used(vn, vxlan->default_dst.remote_ip)) {
1428 vxlan_sock_hold(vs);
1429 dev_hold(dev);
1430 queue_work(vxlan_wq, &vxlan->igmp_join);
1431 }
1432
1433 if (vxlan->age_interval)
1434 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1435
1436 return 0;
1437 }
1438
1439 /* Purge the forwarding table */
1440 static void vxlan_flush(struct vxlan_dev *vxlan)
1441 {
1442 unsigned int h;
1443
1444 spin_lock_bh(&vxlan->hash_lock);
1445 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1446 struct hlist_node *p, *n;
1447 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1448 struct vxlan_fdb *f
1449 = container_of(p, struct vxlan_fdb, hlist);
1450 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1451 if (!is_zero_ether_addr(f->eth_addr))
1452 vxlan_fdb_destroy(vxlan, f);
1453 }
1454 }
1455 spin_unlock_bh(&vxlan->hash_lock);
1456 }
1457
1458 /* Cleanup timer and forwarding table on shutdown */
1459 static int vxlan_stop(struct net_device *dev)
1460 {
1461 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1462 struct vxlan_dev *vxlan = netdev_priv(dev);
1463 struct vxlan_sock *vs = vxlan->vn_sock;
1464
1465 if (vs && IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)) &&
1466 ! vxlan_group_used(vn, vxlan->default_dst.remote_ip)) {
1467 vxlan_sock_hold(vs);
1468 dev_hold(dev);
1469 queue_work(vxlan_wq, &vxlan->igmp_leave);
1470 }
1471
1472 del_timer_sync(&vxlan->age_timer);
1473
1474 vxlan_flush(vxlan);
1475
1476 return 0;
1477 }
1478
1479 /* Stub, nothing needs to be done. */
1480 static void vxlan_set_multicast_list(struct net_device *dev)
1481 {
1482 }
1483
1484 static const struct net_device_ops vxlan_netdev_ops = {
1485 .ndo_init = vxlan_init,
1486 .ndo_uninit = vxlan_uninit,
1487 .ndo_open = vxlan_open,
1488 .ndo_stop = vxlan_stop,
1489 .ndo_start_xmit = vxlan_xmit,
1490 .ndo_get_stats64 = ip_tunnel_get_stats64,
1491 .ndo_set_rx_mode = vxlan_set_multicast_list,
1492 .ndo_change_mtu = eth_change_mtu,
1493 .ndo_validate_addr = eth_validate_addr,
1494 .ndo_set_mac_address = eth_mac_addr,
1495 .ndo_fdb_add = vxlan_fdb_add,
1496 .ndo_fdb_del = vxlan_fdb_delete,
1497 .ndo_fdb_dump = vxlan_fdb_dump,
1498 };
1499
1500 /* Info for udev, that this is a virtual tunnel endpoint */
1501 static struct device_type vxlan_type = {
1502 .name = "vxlan",
1503 };
1504
1505 /* Initialize the device structure. */
1506 static void vxlan_setup(struct net_device *dev)
1507 {
1508 struct vxlan_dev *vxlan = netdev_priv(dev);
1509 unsigned int h;
1510 int low, high;
1511
1512 eth_hw_addr_random(dev);
1513 ether_setup(dev);
1514 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
1515
1516 dev->netdev_ops = &vxlan_netdev_ops;
1517 dev->destructor = free_netdev;
1518 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1519
1520 dev->tx_queue_len = 0;
1521 dev->features |= NETIF_F_LLTX;
1522 dev->features |= NETIF_F_NETNS_LOCAL;
1523 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
1524 dev->features |= NETIF_F_RXCSUM;
1525 dev->features |= NETIF_F_GSO_SOFTWARE;
1526
1527 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
1528 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1529 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
1530 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1531
1532 INIT_LIST_HEAD(&vxlan->next);
1533 spin_lock_init(&vxlan->hash_lock);
1534 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
1535 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1536 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
1537
1538 init_timer_deferrable(&vxlan->age_timer);
1539 vxlan->age_timer.function = vxlan_cleanup;
1540 vxlan->age_timer.data = (unsigned long) vxlan;
1541
1542 inet_get_local_port_range(&low, &high);
1543 vxlan->port_min = low;
1544 vxlan->port_max = high;
1545 vxlan->dst_port = htons(vxlan_port);
1546
1547 vxlan->dev = dev;
1548
1549 for (h = 0; h < FDB_HASH_SIZE; ++h)
1550 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1551 }
1552
1553 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1554 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
1555 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
1556 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1557 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1558 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1559 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1560 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1561 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1562 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
1563 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
1564 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1565 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1566 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1567 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
1568 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
1569 };
1570
1571 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1572 {
1573 if (tb[IFLA_ADDRESS]) {
1574 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1575 pr_debug("invalid link address (not ethernet)\n");
1576 return -EINVAL;
1577 }
1578
1579 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1580 pr_debug("invalid all zero ethernet address\n");
1581 return -EADDRNOTAVAIL;
1582 }
1583 }
1584
1585 if (!data)
1586 return -EINVAL;
1587
1588 if (data[IFLA_VXLAN_ID]) {
1589 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1590 if (id >= VXLAN_VID_MASK)
1591 return -ERANGE;
1592 }
1593
1594 if (data[IFLA_VXLAN_PORT_RANGE]) {
1595 const struct ifla_vxlan_port_range *p
1596 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1597
1598 if (ntohs(p->high) < ntohs(p->low)) {
1599 pr_debug("port range %u .. %u not valid\n",
1600 ntohs(p->low), ntohs(p->high));
1601 return -EINVAL;
1602 }
1603 }
1604
1605 return 0;
1606 }
1607
1608 static void vxlan_get_drvinfo(struct net_device *netdev,
1609 struct ethtool_drvinfo *drvinfo)
1610 {
1611 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1612 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1613 }
1614
1615 static const struct ethtool_ops vxlan_ethtool_ops = {
1616 .get_drvinfo = vxlan_get_drvinfo,
1617 .get_link = ethtool_op_get_link,
1618 };
1619
1620 static void vxlan_del_work(struct work_struct *work)
1621 {
1622 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
1623
1624 sk_release_kernel(vs->sock->sk);
1625 kfree_rcu(vs, rcu);
1626 }
1627
1628 static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port)
1629 {
1630 struct vxlan_sock *vs;
1631 struct sock *sk;
1632 struct sockaddr_in vxlan_addr = {
1633 .sin_family = AF_INET,
1634 .sin_addr.s_addr = htonl(INADDR_ANY),
1635 .sin_port = port,
1636 };
1637 int rc;
1638 unsigned int h;
1639
1640 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
1641 if (!vs)
1642 return ERR_PTR(-ENOMEM);
1643
1644 for (h = 0; h < VNI_HASH_SIZE; ++h)
1645 INIT_HLIST_HEAD(&vs->vni_list[h]);
1646
1647 INIT_WORK(&vs->del_work, vxlan_del_work);
1648
1649 /* Create UDP socket for encapsulation receive. */
1650 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vs->sock);
1651 if (rc < 0) {
1652 pr_debug("UDP socket create failed\n");
1653 kfree(vs);
1654 return ERR_PTR(rc);
1655 }
1656
1657 /* Put in proper namespace */
1658 sk = vs->sock->sk;
1659 sk_change_net(sk, net);
1660
1661 rc = kernel_bind(vs->sock, (struct sockaddr *) &vxlan_addr,
1662 sizeof(vxlan_addr));
1663 if (rc < 0) {
1664 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1665 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
1666 sk_release_kernel(sk);
1667 kfree(vs);
1668 return ERR_PTR(rc);
1669 }
1670
1671 /* Disable multicast loopback */
1672 inet_sk(sk)->mc_loop = 0;
1673
1674 /* Mark socket as an encapsulation socket. */
1675 udp_sk(sk)->encap_type = 1;
1676 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1677 udp_encap_enable();
1678 atomic_set(&vs->refcnt, 1);
1679
1680 return vs;
1681 }
1682
1683 /* Scheduled at device creation to bind to a socket */
1684 static void vxlan_sock_work(struct work_struct *work)
1685 {
1686 struct vxlan_dev *vxlan
1687 = container_of(work, struct vxlan_dev, sock_work);
1688 struct net_device *dev = vxlan->dev;
1689 struct net *net = dev_net(dev);
1690 __u32 vni = vxlan->default_dst.remote_vni;
1691 __be16 port = vxlan->dst_port;
1692 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1693 struct vxlan_sock *nvs, *ovs;
1694
1695 nvs = vxlan_socket_create(net, port);
1696 if (IS_ERR(nvs)) {
1697 netdev_err(vxlan->dev, "Can not create UDP socket, %ld\n",
1698 PTR_ERR(nvs));
1699 goto out;
1700 }
1701
1702 spin_lock(&vn->sock_lock);
1703 /* Look again to see if can reuse socket */
1704 ovs = vxlan_find_port(net, port);
1705 if (ovs) {
1706 atomic_inc(&ovs->refcnt);
1707 vxlan->vn_sock = ovs;
1708 hlist_add_head_rcu(&vxlan->hlist, vni_head(ovs, vni));
1709 spin_unlock(&vn->sock_lock);
1710
1711 sk_release_kernel(nvs->sock->sk);
1712 kfree(nvs);
1713 } else {
1714 vxlan->vn_sock = nvs;
1715 hlist_add_head_rcu(&nvs->hlist, vs_head(net, port));
1716 hlist_add_head_rcu(&vxlan->hlist, vni_head(nvs, vni));
1717 spin_unlock(&vn->sock_lock);
1718 }
1719 out:
1720 dev_put(dev);
1721 }
1722
1723 static int vxlan_newlink(struct net *net, struct net_device *dev,
1724 struct nlattr *tb[], struct nlattr *data[])
1725 {
1726 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1727 struct vxlan_dev *vxlan = netdev_priv(dev);
1728 struct vxlan_rdst *dst = &vxlan->default_dst;
1729 __u32 vni;
1730 int err;
1731
1732 if (!data[IFLA_VXLAN_ID])
1733 return -EINVAL;
1734
1735 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
1736 dst->remote_vni = vni;
1737
1738 if (data[IFLA_VXLAN_GROUP])
1739 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]);
1740
1741 if (data[IFLA_VXLAN_LOCAL])
1742 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1743
1744 if (data[IFLA_VXLAN_LINK] &&
1745 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
1746 struct net_device *lowerdev
1747 = __dev_get_by_index(net, dst->remote_ifindex);
1748
1749 if (!lowerdev) {
1750 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
1751 return -ENODEV;
1752 }
1753
1754 if (!tb[IFLA_MTU])
1755 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1756
1757 /* update header length based on lower device */
1758 dev->hard_header_len = lowerdev->hard_header_len +
1759 VXLAN_HEADROOM;
1760 }
1761
1762 if (data[IFLA_VXLAN_TOS])
1763 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1764
1765 if (data[IFLA_VXLAN_TTL])
1766 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1767
1768 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
1769 vxlan->flags |= VXLAN_F_LEARN;
1770
1771 if (data[IFLA_VXLAN_AGEING])
1772 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1773 else
1774 vxlan->age_interval = FDB_AGE_DEFAULT;
1775
1776 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1777 vxlan->flags |= VXLAN_F_PROXY;
1778
1779 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1780 vxlan->flags |= VXLAN_F_RSC;
1781
1782 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1783 vxlan->flags |= VXLAN_F_L2MISS;
1784
1785 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1786 vxlan->flags |= VXLAN_F_L3MISS;
1787
1788 if (data[IFLA_VXLAN_LIMIT])
1789 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1790
1791 if (data[IFLA_VXLAN_PORT_RANGE]) {
1792 const struct ifla_vxlan_port_range *p
1793 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1794 vxlan->port_min = ntohs(p->low);
1795 vxlan->port_max = ntohs(p->high);
1796 }
1797
1798 if (data[IFLA_VXLAN_PORT])
1799 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
1800
1801 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
1802 pr_info("duplicate VNI %u\n", vni);
1803 return -EEXIST;
1804 }
1805
1806 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1807
1808 /* create an fdb entry for default destination */
1809 err = vxlan_fdb_create(vxlan, all_zeros_mac,
1810 vxlan->default_dst.remote_ip,
1811 NUD_REACHABLE|NUD_PERMANENT,
1812 NLM_F_EXCL|NLM_F_CREATE,
1813 vxlan->dst_port, vxlan->default_dst.remote_vni,
1814 vxlan->default_dst.remote_ifindex, NTF_SELF);
1815 if (err)
1816 return err;
1817
1818 err = register_netdevice(dev);
1819 if (err) {
1820 vxlan_fdb_delete_default(vxlan);
1821 return err;
1822 }
1823
1824 list_add(&vxlan->next, &vn->vxlan_list);
1825
1826 return 0;
1827 }
1828
1829 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1830 {
1831 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1832 struct vxlan_dev *vxlan = netdev_priv(dev);
1833
1834 flush_workqueue(vxlan_wq);
1835
1836 spin_lock(&vn->sock_lock);
1837 hlist_del_rcu(&vxlan->hlist);
1838 spin_unlock(&vn->sock_lock);
1839
1840 list_del(&vxlan->next);
1841 unregister_netdevice_queue(dev, head);
1842 }
1843
1844 static size_t vxlan_get_size(const struct net_device *dev)
1845 {
1846
1847 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
1848 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */
1849 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1850 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1851 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1852 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1853 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
1854 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1855 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1856 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1857 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
1858 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1859 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
1860 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
1861 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
1862 0;
1863 }
1864
1865 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1866 {
1867 const struct vxlan_dev *vxlan = netdev_priv(dev);
1868 const struct vxlan_rdst *dst = &vxlan->default_dst;
1869 struct ifla_vxlan_port_range ports = {
1870 .low = htons(vxlan->port_min),
1871 .high = htons(vxlan->port_max),
1872 };
1873
1874 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
1875 goto nla_put_failure;
1876
1877 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip))
1878 goto nla_put_failure;
1879
1880 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
1881 goto nla_put_failure;
1882
1883 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
1884 goto nla_put_failure;
1885
1886 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1887 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
1888 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1889 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1890 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1891 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1892 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1893 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1894 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1895 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1896 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
1897 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
1898 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
1899 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
1900 goto nla_put_failure;
1901
1902 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1903 goto nla_put_failure;
1904
1905 return 0;
1906
1907 nla_put_failure:
1908 return -EMSGSIZE;
1909 }
1910
1911 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1912 .kind = "vxlan",
1913 .maxtype = IFLA_VXLAN_MAX,
1914 .policy = vxlan_policy,
1915 .priv_size = sizeof(struct vxlan_dev),
1916 .setup = vxlan_setup,
1917 .validate = vxlan_validate,
1918 .newlink = vxlan_newlink,
1919 .dellink = vxlan_dellink,
1920 .get_size = vxlan_get_size,
1921 .fill_info = vxlan_fill_info,
1922 };
1923
1924 static __net_init int vxlan_init_net(struct net *net)
1925 {
1926 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1927 unsigned int h;
1928
1929 INIT_LIST_HEAD(&vn->vxlan_list);
1930 spin_lock_init(&vn->sock_lock);
1931
1932 for (h = 0; h < PORT_HASH_SIZE; ++h)
1933 INIT_HLIST_HEAD(&vn->sock_list[h]);
1934
1935 return 0;
1936 }
1937
1938 static __net_exit void vxlan_exit_net(struct net *net)
1939 {
1940 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1941 struct vxlan_dev *vxlan;
1942 LIST_HEAD(list);
1943
1944 rtnl_lock();
1945 list_for_each_entry(vxlan, &vn->vxlan_list, next)
1946 unregister_netdevice_queue(vxlan->dev, &list);
1947 unregister_netdevice_many(&list);
1948 rtnl_unlock();
1949 }
1950
1951 static struct pernet_operations vxlan_net_ops = {
1952 .init = vxlan_init_net,
1953 .exit = vxlan_exit_net,
1954 .id = &vxlan_net_id,
1955 .size = sizeof(struct vxlan_net),
1956 };
1957
1958 static int __init vxlan_init_module(void)
1959 {
1960 int rc;
1961
1962 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
1963 if (!vxlan_wq)
1964 return -ENOMEM;
1965
1966 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1967
1968 rc = register_pernet_device(&vxlan_net_ops);
1969 if (rc)
1970 goto out1;
1971
1972 rc = rtnl_link_register(&vxlan_link_ops);
1973 if (rc)
1974 goto out2;
1975
1976 return 0;
1977
1978 out2:
1979 unregister_pernet_device(&vxlan_net_ops);
1980 out1:
1981 destroy_workqueue(vxlan_wq);
1982 return rc;
1983 }
1984 late_initcall(vxlan_init_module);
1985
1986 static void __exit vxlan_cleanup_module(void)
1987 {
1988 rtnl_link_unregister(&vxlan_link_ops);
1989 destroy_workqueue(vxlan_wq);
1990 unregister_pernet_device(&vxlan_net_ops);
1991 rcu_barrier();
1992 }
1993 module_exit(vxlan_cleanup_module);
1994
1995 MODULE_LICENSE("GPL");
1996 MODULE_VERSION(VXLAN_VERSION);
1997 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
1998 MODULE_ALIAS_RTNL_LINK("vxlan");
Linux kernel - Deliverables
This page took 0.067714 seconds and 4 git commands to generate.