]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - drivers/net/vxlan.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
bridge: fix rcu check warning in multicast port group
[mirror_ubuntu-bionic-kernel.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 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
d342894c 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>
d342894c 30#include <linux/hash.h>
1b13c97f 31#include <linux/ethtool.h>
e4f67add
DS		 32#include <net/arp.h>
33#include <net/ndisc.h>
d342894c 34#include <net/ip.h>
c5441932 35#include <net/ip_tunnels.h>
d342894c 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
553675fb 47#define PORT_HASH_BITS 8
48#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 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)
52b702ff
AD		 58/* IP header + UDP + VXLAN + Ethernet header */
59#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
d342894c 60
61#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
62
63/* VXLAN protocol header */
64struct vxlanhdr {
65 __be32 vx_flags;
66 __be32 vx_vni;
67};
68
23c578bf 69/* UDP port for VXLAN traffic.
70 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 71 * for compatibility with early adopters.
23c578bf 72 */
9daaa397
SH		 73static unsigned short vxlan_port __read_mostly = 8472;
74module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 75MODULE_PARM_DESC(udp_port, "Destination UDP port");
76
77static bool log_ecn_error = true;
78module_param(log_ecn_error, bool, 0644);
79MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
80
60d9d4c6 81static int vxlan_net_id;
553675fb 82
afbd8bae
MR		 83static const u8 all_zeros_mac[ETH_ALEN];
84
553675fb 85/* per UDP socket information */
86struct vxlan_sock {
87 struct hlist_node hlist;
88 struct rcu_head rcu;
89 struct work_struct del_work;
7c47cedf 90 atomic_t refcnt;
553675fb 91 struct socket *sock;
d342894c 92 struct hlist_head vni_list[VNI_HASH_SIZE];
93};
94
553675fb 95/* per-network namespace private data for this module */
96struct vxlan_net {
97 struct list_head vxlan_list;
98 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 99 spinlock_t sock_lock;
553675fb 100};
101
6681712d 102struct vxlan_rdst {
6681712d
DS		 103 __be32 remote_ip;
104 __be16 remote_port;
105 u32 remote_vni;
106 u32 remote_ifindex;
3e61aa8f 107 struct list_head list;
bc7892ba 108 struct rcu_head rcu;
6681712d
DS		 109};
110
d342894c 111/* Forwarding table entry */
112struct 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;
3e61aa8f 117 struct list_head remotes;
d342894c 118 u16 state; /* see ndm_state */
ae884082 119 u8 flags; /* see ndm_flags */
d342894c 120 u8 eth_addr[ETH_ALEN];
121};
122
d342894c 123/* Pseudo network device */
124struct vxlan_dev {
553675fb 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 */
d342894c 128 struct net_device *dev;
c7995c43 129 struct vxlan_rdst default_dst; /* default destination */
d342894c 130 __be32 saddr; /* source address */
823aa873 131 __be16 dst_port;
05f47d69 132 __u16 port_min; /* source port range */
133 __u16 port_max;
d342894c 134 __u8 tos; /* TOS override */
135 __u8 ttl;
e4f67add 136 u32 flags; /* VXLAN_F_* below */
d342894c 137
1c51a915 138 struct work_struct sock_work;
3fc2de2f 139 struct work_struct igmp_join;
140 struct work_struct igmp_leave;
1c51a915 141
d342894c 142 unsigned long age_interval;
143 struct timer_list age_timer;
144 spinlock_t hash_lock;
145 unsigned int addrcnt;
146 unsigned int addrmax;
d342894c 147
148 struct hlist_head fdb_head[FDB_HASH_SIZE];
149};
150
e4f67add
DS		 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
d342894c 157/* salt for hash table */
158static u32 vxlan_salt __read_mostly;
758c57d1 159static struct workqueue_struct *vxlan_wq;
d342894c 160
1c51a915
SH		 161static void vxlan_sock_work(struct work_struct *work);
162
553675fb 163/* Virtual Network hash table head */
164static 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 */
170static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 171{
172 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
173
553675fb 174 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
175}
176
3e61aa8f
SH		 177/* First remote destination for a forwarding entry.
178 * Guaranteed to be non-NULL because remotes are never deleted.
179 */
180static inline struct vxlan_rdst *first_remote(struct vxlan_fdb *fdb)
181{
182 return list_first_or_null_rcu(&fdb->remotes, struct vxlan_rdst, list);
183}
184
553675fb 185/* Find VXLAN socket based on network namespace and UDP port */
186static 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;
d342894c 195}
196
197/* Look up VNI in a per net namespace table */
553675fb 198static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
d342894c 199{
553675fb 200 struct vxlan_sock *vs;
d342894c 201 struct vxlan_dev *vxlan;
d342894c 202
553675fb 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) {
c7995c43 208 if (vxlan->default_dst.remote_vni == id)
d342894c 209 return vxlan;
210 }
211
212 return NULL;
213}
214
215/* Fill in neighbour message in skbuff. */
216static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 217 const struct vxlan_fdb *fdb,
218 u32 portid, u32 seq, int type, unsigned int flags,
219 const struct vxlan_rdst *rdst)
d342894c 220{
221 unsigned long now = jiffies;
222 struct nda_cacheinfo ci;
223 struct nlmsghdr *nlh;
224 struct ndmsg *ndm;
e4f67add 225 bool send_ip, send_eth;
d342894c 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));
e4f67add
DS		 233
234 send_eth = send_ip = true;
235
236 if (type == RTM_GETNEIGH) {
237 ndm->ndm_family = AF_INET;
6681712d 238 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
e4f67add
DS		 239 send_eth = !is_zero_ether_addr(fdb->eth_addr);
240 } else
241 ndm->ndm_family = AF_BRIDGE;
d342894c 242 ndm->ndm_state = fdb->state;
243 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 244 ndm->ndm_flags = fdb->flags;
d342894c 245 ndm->ndm_type = NDA_DST;
246
e4f67add 247 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 248 goto nla_put_failure;
249
6681712d
DS		 250 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
251 goto nla_put_failure;
252
823aa873 253 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 254 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
255 goto nla_put_failure;
c7995c43 256 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 257 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 258 goto nla_put_failure;
259 if (rdst->remote_ifindex &&
260 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 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
273nla_put_failure:
274 nlmsg_cancel(skb, nlh);
275 return -EMSGSIZE;
276}
277
278static 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 */
73cf3317 283 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 284 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
285 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 286 + nla_total_size(sizeof(struct nda_cacheinfo));
287}
288
289static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
3e61aa8f 290 struct vxlan_fdb *fdb, int type)
d342894c 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
3e61aa8f 300 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, first_remote(fdb));
d342894c 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;
310errout:
311 if (err < 0)
312 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
313}
314
e4f67add
DS		 315static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
316{
317 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 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 };
3e61aa8f
SH		 325
326 INIT_LIST_HEAD(&f.remotes);
327 list_add_rcu(&remote.list, &f.remotes);
e4f67add
DS		 328
329 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
330}
331
332static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
333{
bb3fd687
SH		 334 struct vxlan_fdb f = {
335 .state = NUD_STALE,
336 };
e4f67add 337
3e61aa8f 338 INIT_LIST_HEAD(&f.remotes);
e4f67add
DS		 339 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
340
341 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
342}
343
d342894c 344/* Hash Ethernet address */
345static 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
d342894c 351 value >>= 16;
321fb991 352#else
353 value <<= 16;
d342894c 354#endif
355 return hash_64(value, FDB_HASH_BITS);
356}
357
358/* Hash chain to use given mac address */
359static 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 */
014be2c8 366static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 367 const u8 *mac)
368
369{
370 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
371 struct vxlan_fdb *f;
d342894c 372
b67bfe0d 373 hlist_for_each_entry_rcu(f, head, hlist) {
d342894c 374 if (compare_ether_addr(mac, f->eth_addr) == 0)
375 return f;
376 }
377
378 return NULL;
379}
380
014be2c8
SS		 381static 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
a5e7c10a
MR		 393/* caller should hold vxlan->hash_lock */
394static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
395 __be32 ip, __be16 port,
396 __u32 vni, __u32 ifindex)
6681712d 397{
3e61aa8f 398 struct vxlan_rdst *rd;
6681712d 399
3e61aa8f 400 list_for_each_entry(rd, &f->remotes, list) {
6681712d
DS		 401 if (rd->remote_ip == ip &&
402 rd->remote_port == port &&
403 rd->remote_vni == vni &&
404 rd->remote_ifindex == ifindex)
a5e7c10a 405 return rd;
6681712d 406 }
3e61aa8f 407
a5e7c10a
MR		 408 return NULL;
409}
410
906dc186
TR		 411/* Replace destination of unicast mac */
412static 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
a5e7c10a
MR		 431/* Add/update destinations for multicast */
432static 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
6681712d
DS		 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;
3e61aa8f
SH		 448
449 list_add_tail_rcu(&rd->list, &f->remotes);
450
6681712d
DS		 451 return 1;
452}
453
d342894c 454/* Add new entry to forwarding table -- assumes lock held */
455static int vxlan_fdb_create(struct vxlan_dev *vxlan,
456 const u8 *mac, __be32 ip,
6681712d 457 __u16 state, __u16 flags,
73cf3317 458 __be16 port, __u32 vni, __u32 ifindex,
ae884082 459 __u8 ndm_flags)
d342894c 460{
461 struct vxlan_fdb *f;
462 int notify = 0;
463
014be2c8 464 f = __vxlan_find_mac(vxlan, mac);
d342894c 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 }
ae884082
DS		 476 if (f->flags != ndm_flags) {
477 f->flags = ndm_flags;
478 f->updated = jiffies;
479 notify = 1;
480 }
906dc186
TR		 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 }
6681712d 494 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 495 (is_multicast_ether_addr(f->eth_addr) ||
496 is_zero_ether_addr(f->eth_addr))) {
6681712d
DS		 497 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
498
499 if (rc < 0)
500 return rc;
501 notify |= rc;
502 }
d342894c 503 } else {
504 if (!(flags & NLM_F_CREATE))
505 return -ENOENT;
506
507 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
508 return -ENOSPC;
509
906dc186
TR		 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
d342894c 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;
d342894c 521 f->state = state;
ae884082 522 f->flags = ndm_flags;
d342894c 523 f->updated = f->used = jiffies;
3e61aa8f 524 INIT_LIST_HEAD(&f->remotes);
d342894c 525 memcpy(f->eth_addr, mac, ETH_ALEN);
526
3e61aa8f
SH		 527 vxlan_fdb_append(f, ip, port, vni, ifindex);
528
d342894c 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
bc7892ba
MR		 540static 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
6706c82e 546static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 547{
548 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 549 struct vxlan_rdst *rd, *nd;
6681712d 550
3e61aa8f 551 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 552 kfree(rd);
6681712d
DS		 553 kfree(f);
554}
555
d342894c 556static 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);
6681712d 565 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 566}
567
f0b074be
MR		 568static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
569 __be32 *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 570{
6681712d 571 struct net *net = dev_net(vxlan->dev);
d342894c 572
f0b074be
MR		 573 if (tb[NDA_DST]) {
574 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
575 return -EAFNOSUPPORT;
d342894c 576
f0b074be
MR		 577 *ip = nla_get_be32(tb[NDA_DST]);
578 } else {
579 *ip = htonl(INADDR_ANY);
580 }
d342894c 581
6681712d 582 if (tb[NDA_PORT]) {
73cf3317 583 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 584 return -EINVAL;
f0b074be
MR		 585 *port = nla_get_be16(tb[NDA_PORT]);
586 } else {
587 *port = vxlan->dst_port;
588 }
6681712d
DS		 589
590 if (tb[NDA_VNI]) {
591 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
592 return -EINVAL;
f0b074be
MR		 593 *vni = nla_get_u32(tb[NDA_VNI]);
594 } else {
595 *vni = vxlan->default_dst.remote_vni;
596 }
6681712d
DS		 597
598 if (tb[NDA_IFINDEX]) {
5abb0029 599 struct net_device *tdev;
6681712d
DS		 600
601 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
602 return -EINVAL;
f0b074be
MR		 603 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
604 tdev = dev_get_by_index(net, *ifindex);
5abb0029 605 if (!tdev)
6681712d 606 return -EADDRNOTAVAIL;
5abb0029 607 dev_put(tdev);
f0b074be
MR		 608 } else {
609 *ifindex = 0;
610 }
611
612 return 0;
613}
614
615/* Add static entry (via netlink) */
616static 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;
6681712d 639
d342894c 640 spin_lock_bh(&vxlan->hash_lock);
73cf3317 641 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags,
642 port, vni, ifindex, ndm->ndm_flags);
d342894c 643 spin_unlock_bh(&vxlan->hash_lock);
644
645 return err;
646}
647
648/* Delete entry (via netlink) */
1690be63
VY		 649static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
650 struct net_device *dev,
d342894c 651 const unsigned char *addr)
652{
653 struct vxlan_dev *vxlan = netdev_priv(dev);
654 struct vxlan_fdb *f;
bc7892ba
MR		 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;
d342894c 666
667 spin_lock_bh(&vxlan->hash_lock);
668 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 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;
d342894c 687 }
bc7892ba
MR		 688
689 vxlan_fdb_destroy(vxlan, f);
690
691out:
d342894c 692 spin_unlock_bh(&vxlan->hash_lock);
693
694 return err;
695}
696
697/* Dump forwarding table */
698static 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;
d342894c 706 int err;
707
b67bfe0d 708 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 709 struct vxlan_rdst *rd;
6681712d 710
3e61aa8f
SH		 711 if (idx < cb->args[0])
712 goto skip;
713
714 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 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)
3e61aa8f 721 goto out;
6681712d 722 }
3e61aa8f
SH		 723skip:
724 ++idx;
d342894c 725 }
726 }
3e61aa8f 727out:
d342894c 728 return idx;
729}
730
731/* Watch incoming packets to learn mapping between Ethernet address
732 * and Tunnel endpoint.
26a41ae6 733 * Return true if packet is bogus and should be droppped.
d342894c 734 */
26a41ae6 735static bool vxlan_snoop(struct net_device *dev,
d342894c 736 __be32 src_ip, const u8 *src_mac)
737{
738 struct vxlan_dev *vxlan = netdev_priv(dev);
739 struct vxlan_fdb *f;
d342894c 740
741 f = vxlan_find_mac(vxlan, src_mac);
742 if (likely(f)) {
3e61aa8f
SH		 743 struct vxlan_rdst *rdst = first_remote(f);
744
745 if (likely(rdst->remote_ip == src_ip))
26a41ae6 746 return false;
747
748 /* Don't migrate static entries, drop packets */
eb064c3b 749 if (f->state & NUD_NOARP)
26a41ae6 750 return true;
d342894c 751
752 if (net_ratelimit())
753 netdev_info(dev,
754 "%pM migrated from %pI4 to %pI4\n",
3e61aa8f 755 src_mac, &rdst->remote_ip, &src_ip);
d342894c 756
3e61aa8f 757 rdst->remote_ip = src_ip;
d342894c 758 f->updated = jiffies;
8385f50a 759 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
d342894c 760 } else {
761 /* learned new entry */
762 spin_lock(&vxlan->hash_lock);
3bf74b1a 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);
d342894c 772 spin_unlock(&vxlan->hash_lock);
773 }
26a41ae6 774
775 return false;
d342894c 776}
777
d342894c 778/* See if multicast group is already in use by other ID */
7c47cedf 779static bool vxlan_group_used(struct vxlan_net *vn, __be32 remote_ip)
d342894c 780{
553675fb 781 struct vxlan_dev *vxlan;
d342894c 782
553675fb 783 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
553675fb 784 if (!netif_running(vxlan->dev))
785 continue;
d342894c 786
7c47cedf 787 if (vxlan->default_dst.remote_ip == remote_ip)
553675fb 788 return true;
789 }
d342894c 790
791 return false;
792}
793
7c47cedf 794static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 795{
7c47cedf
SH		 796 atomic_inc(&vs->refcnt);
797}
d342894c 798
1c51a915 799static void vxlan_sock_release(struct vxlan_net *vn, struct vxlan_sock *vs)
7c47cedf
SH		 800{
801 if (!atomic_dec_and_test(&vs->refcnt))
802 return;
d342894c 803
1c51a915 804 spin_lock(&vn->sock_lock);
7c47cedf 805 hlist_del_rcu(&vs->hlist);
1c51a915
SH		 806 spin_unlock(&vn->sock_lock);
807
7c47cedf 808 queue_work(vxlan_wq, &vs->del_work);
d342894c 809}
810
3fc2de2f 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.
7c47cedf 814 */
3fc2de2f 815static void vxlan_igmp_join(struct work_struct *work)
d342894c 816{
3fc2de2f 817 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 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;
d342894c 821 struct ip_mreqn mreq = {
c7995c43
AW		 822 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
823 .imr_ifindex = vxlan->default_dst.remote_ifindex,
d342894c 824 };
825
d342894c 826 lock_sock(sk);
3fc2de2f 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 */
835static 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);
d342894c 848 release_sock(sk);
d342894c 849
1c51a915 850 vxlan_sock_release(vn, vs);
7c47cedf 851 dev_put(vxlan->dev);
d342894c 852}
853
854/* Callback from net/ipv4/udp.c to receive packets */
855static 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;
e8171045 860 struct pcpu_tstats *stats;
553675fb 861 __be16 port;
d342894c 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));
d342894c 882
883 /* Is this VNI defined? */
884 vni = ntohl(vxh->vx_vni) >> 8;
553675fb 885 port = inet_sk(sk)->inet_sport;
886 vxlan = vxlan_find_vni(sock_net(sk), vni, port);
d342894c 887 if (!vxlan) {
553675fb 888 netdev_dbg(skb->dev, "unknown vni %d port %u\n",
889 vni, ntohs(port));
d342894c 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
e4f67add
DS		 899 skb_reset_mac_header(skb);
900
d342894c 901 /* Re-examine inner Ethernet packet */
902 oip = ip_hdr(skb);
903 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 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
26a41ae6 910 if ((vxlan->flags & VXLAN_F_LEARN) &&
911 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source))
912 goto drop;
d342894c 913
914 __skb_tunnel_rx(skb, vxlan->dev);
915 skb_reset_network_header(skb);
0afb1666
JG		 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;
d342894c 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
e8171045 940 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 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;
949error:
950 /* Put UDP header back */
951 __skb_push(skb, sizeof(struct udphdr));
952
953 return 1;
954drop:
955 /* Consume bad packet */
956 kfree_skb(skb);
957 return 0;
958}
959
e4f67add
DS		 960static 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) {
e4f67add
DS		 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);
3e61aa8f 1008 if (f && first_remote(f)->remote_ip == htonl(INADDR_ANY)) {
e4f67add
DS		 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);
1028out:
1029 consume_skb(skb);
1030 return NETDEV_TX_OK;
1031}
1032
1033static 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
553675fb 1070static void vxlan_sock_put(struct sk_buff *skb)
1cad8715 1071{
1072 sock_put(skb->sk);
1073}
1074
1075/* On transmit, associate with the tunnel socket */
1076static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
1077{
553675fb 1078 struct vxlan_dev *vxlan = netdev_priv(dev);
1079 struct sock *sk = vxlan->vn_sock->sock->sk;
1cad8715 1080
1081 skb_orphan(skb);
1082 sock_hold(sk);
1083 skb->sk = sk;
553675fb 1084 skb->destructor = vxlan_sock_put;
1cad8715 1085}
1086
05f47d69 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 */
7d836a76 1092static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
05f47d69 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
7d836a76 1102 return htons((((u64) hash * range) >> 32) + vxlan->port_min);
05f47d69 1103}
1104
05c0db08
PS		 1105static 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
f6ace502 1112 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 1113 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1114 skb->ip_summed = CHECKSUM_NONE;
1115
1116 return 0;
1117}
1118
9dcc71e1
SS		 1119/* Bypass encapsulation if the destination is local */
1120static 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)
9d9f163c
MR		 1132 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
1133 eth_hdr(skb)->h_source);
9dcc71e1
SS		 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
4ad16930
SH		 1150static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1151 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1152{
1153 struct vxlan_dev *vxlan = netdev_priv(dev);
1154 struct rtable *rt;
d342894c 1155 const struct iphdr *old_iph;
d342894c 1156 struct vxlanhdr *vxh;
1157 struct udphdr *uh;
1158 struct flowi4 fl4;
d342894c 1159 __be32 dst;
7d836a76 1160 __be16 src_port, dst_port;
234f5b73 1161 u32 vni;
d342894c 1162 __be16 df = 0;
1163 __u8 tos, ttl;
0e6fbc5b 1164 int err;
d342894c 1165
823aa873 1166 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d
DS		 1167 vni = rdst->remote_vni;
1168 dst = rdst->remote_ip;
e4f67add
DS		 1169
1170 if (!dst) {
1171 if (did_rsc) {
e4f67add 1172 /* short-circuited back to local bridge */
9dcc71e1 1173 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1174 return;
e4f67add 1175 }
ef59febe 1176 goto drop;
e4f67add 1177 }
ef59febe 1178
d6727fe3
JG		 1179 if (!skb->encapsulation) {
1180 skb_reset_inner_headers(skb);
1181 skb->encapsulation = 1;
1182 }
1183
d342894c 1184 /* Need space for new headers (invalidates iph ptr) */
1185 if (skb_cow_head(skb, VXLAN_HEADROOM))
1186 goto drop;
1187
d342894c 1188 old_iph = ip_hdr(skb);
1189
d342894c 1190 ttl = vxlan->ttl;
1191 if (!ttl && IN_MULTICAST(ntohl(dst)))
1192 ttl = 1;
1193
1194 tos = vxlan->tos;
1195 if (tos == 1)
206aaafc 1196 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1197
05f47d69 1198 src_port = vxlan_src_port(vxlan, skb);
d342894c 1199
ca78f181 1200 memset(&fl4, 0, sizeof(fl4));
6681712d 1201 fl4.flowi4_oif = rdst->remote_ifindex;
ca78f181 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);
d342894c 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
9dcc71e1 1220 /* Bypass encapsulation if the destination is local */
ab09a6d0
MR		 1221 if (rt->rt_flags & RTCF_LOCAL &&
1222 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
9dcc71e1
SS		 1223 struct vxlan_dev *dst_vxlan;
1224
1225 ip_rt_put(rt);
553675fb 1226 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
9dcc71e1
SS		 1227 if (!dst_vxlan)
1228 goto tx_error;
1229 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
4ad16930 1230 return;
9dcc71e1 1231 }
d342894c 1232 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1233 vxh->vx_flags = htonl(VXLAN_FLAGS);
6681712d 1234 vxh->vx_vni = htonl(vni << 8);
d342894c 1235
1236 __skb_push(skb, sizeof(*uh));
1237 skb_reset_transport_header(skb);
1238 uh = udp_hdr(skb);
1239
73cf3317 1240 uh->dest = dst_port;
7d836a76 1241 uh->source = src_port;
d342894c 1242
1243 uh->len = htons(skb->len);
1244 uh->check = 0;
1245
1cad8715 1246 vxlan_set_owner(dev, skb);
1247
05c0db08
PS		 1248 if (handle_offloads(skb))
1249 goto drop;
d342894c 1250
0e6fbc5b
PS		 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
4ad16930 1258 return;
d342894c 1259
1260drop:
1261 dev->stats.tx_dropped++;
1262 goto tx_free;
1263
1264tx_error:
1265 dev->stats.tx_errors++;
1266tx_free:
1267 dev_kfree_skb(skb);
d342894c 1268}
1269
6681712d
DS		 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 */
1276static 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;
afbd8bae 1281 struct vxlan_rdst *rdst;
6681712d 1282 struct vxlan_fdb *f;
6681712d
DS		 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);
6681712d
DS		 1289
1290 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 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
6681712d 1300 if (f == NULL) {
afbd8bae
MR		 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 }
6681712d 1312
afbd8bae
MR		 1313 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1314 struct sk_buff *skb1;
6681712d 1315
afbd8bae
MR		 1316 skb1 = skb_clone(skb, GFP_ATOMIC);
1317 if (skb1)
1318 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1319 }
1320
afbd8bae 1321 dev_kfree_skb(skb);
4ad16930 1322 return NETDEV_TX_OK;
6681712d
DS		 1323}
1324
d342894c 1325/* Walk the forwarding table and purge stale entries */
1326static 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
3c172868 1343 if (f->state & NUD_PERMANENT)
d342894c 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 */
1363static int vxlan_init(struct net_device *dev)
1364{
1c51a915
SH		 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
e8171045
PS		 1370 dev->tstats = alloc_percpu(struct pcpu_tstats);
1371 if (!dev->tstats)
d342894c 1372 return -ENOMEM;
1373
1c51a915
SH		 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
d342894c 1388 return 0;
1389}
1390
ba609e9b 1391static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 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
ebf4063e
SH		 1402static 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
ba609e9b 1408 vxlan_fdb_delete_default(vxlan);
afbd8bae 1409
ebf4063e
SH		 1410 if (vs)
1411 vxlan_sock_release(vn, vs);
1412 free_percpu(dev->tstats);
1413}
1414
d342894c 1415/* Start ageing timer and join group when device is brought up */
1416static int vxlan_open(struct net_device *dev)
1417{
3fc2de2f 1418 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1419 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 1420 struct vxlan_sock *vs = vxlan->vn_sock;
1421
1422 /* socket hasn't been created */
1423 if (!vs)
1424 return -ENOTCONN;
d342894c 1425
3fc2de2f 1426 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)) &&
1427 ! vxlan_group_used(vn, vxlan->default_dst.remote_ip)) {
1c51a915 1428 vxlan_sock_hold(vs);
7c47cedf 1429 dev_hold(dev);
3fc2de2f 1430 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 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 */
1440static void vxlan_flush(struct vxlan_dev *vxlan)
1441{
31fec5aa 1442 unsigned int h;
d342894c 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);
afbd8bae
MR		 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);
d342894c 1453 }
1454 }
1455 spin_unlock_bh(&vxlan->hash_lock);
1456}
1457
1458/* Cleanup timer and forwarding table on shutdown */
1459static int vxlan_stop(struct net_device *dev)
1460{
3fc2de2f 1461 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1462 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915 1463 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 1464
3fc2de2f 1465 if (vs && IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)) &&
1466 ! vxlan_group_used(vn, vxlan->default_dst.remote_ip)) {
1c51a915 1467 vxlan_sock_hold(vs);
7c47cedf 1468 dev_hold(dev);
3fc2de2f 1469 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 1470 }
d342894c 1471
1472 del_timer_sync(&vxlan->age_timer);
1473
1474 vxlan_flush(vxlan);
1475
1476 return 0;
1477}
1478
d342894c 1479/* Stub, nothing needs to be done. */
1480static void vxlan_set_multicast_list(struct net_device *dev)
1481{
1482}
1483
1484static const struct net_device_ops vxlan_netdev_ops = {
1485 .ndo_init = vxlan_init,
ebf4063e 1486 .ndo_uninit = vxlan_uninit,
d342894c 1487 .ndo_open = vxlan_open,
1488 .ndo_stop = vxlan_stop,
1489 .ndo_start_xmit = vxlan_xmit,
e8171045 1490 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 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 */
1501static struct device_type vxlan_type = {
1502 .name = "vxlan",
1503};
1504
d342894c 1505/* Initialize the device structure. */
1506static void vxlan_setup(struct net_device *dev)
1507{
1508 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 1509 unsigned int h;
05f47d69 1510 int low, high;
d342894c 1511
1512 eth_hw_addr_random(dev);
1513 ether_setup(dev);
2840bf22 1514 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 1515
1516 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 1517 dev->destructor = free_netdev;
d342894c 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;
d6727fe3 1523 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 1524 dev->features |= NETIF_F_RXCSUM;
05c0db08 1525 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666
JG		 1526
1527 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 1528 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
d342894c 1529 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 1530 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 1531
553675fb 1532 INIT_LIST_HEAD(&vxlan->next);
d342894c 1533 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 1534 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
1535 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 1536 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 1537
1538 init_timer_deferrable(&vxlan->age_timer);
1539 vxlan->age_timer.function = vxlan_cleanup;
1540 vxlan->age_timer.data = (unsigned long) vxlan;
1541
05f47d69 1542 inet_get_local_port_range(&low, &high);
1543 vxlan->port_min = low;
1544 vxlan->port_max = high;
823aa873 1545 vxlan->dst_port = htons(vxlan_port);
05f47d69 1546
d342894c 1547 vxlan->dev = dev;
1548
1549 for (h = 0; h < FDB_HASH_SIZE; ++h)
1550 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1551}
1552
1553static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1554 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 1555 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
d342894c 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 },
05f47d69 1563 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 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 },
823aa873 1568 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 1569};
1570
1571static 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
05f47d69 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
d342894c 1605 return 0;
1606}
1607
1b13c97f
YB		 1608static 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
1615static const struct ethtool_ops vxlan_ethtool_ops = {
1616 .get_drvinfo = vxlan_get_drvinfo,
1617 .get_link = ethtool_op_get_link,
1618};
1619
553675fb 1620static 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
553675fb 1628static 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),
bb3fd687 1635 .sin_port = port,
553675fb 1636 };
1637 int rc;
31fec5aa 1638 unsigned int h;
553675fb 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
553675fb 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();
7c47cedf 1678 atomic_set(&vs->refcnt, 1);
553675fb 1679
553675fb 1680 return vs;
1681}
1682
1c51a915
SH		 1683/* Scheduled at device creation to bind to a socket */
1684static 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 }
1719out:
1720 dev_put(dev);
1721}
1722
d342894c 1723static int vxlan_newlink(struct net *net, struct net_device *dev,
1724 struct nlattr *tb[], struct nlattr *data[])
1725{
553675fb 1726 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 1727 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1728 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 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]);
c7995c43 1736 dst->remote_vni = vni;
d342894c 1737
5d174dd8 1738 if (data[IFLA_VXLAN_GROUP])
1739 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]);
d342894c 1740
1741 if (data[IFLA_VXLAN_LOCAL])
1742 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1743
34e02aa1 1744 if (data[IFLA_VXLAN_LINK] &&
c7995c43 1745 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 1746 struct net_device *lowerdev
c7995c43 1747 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 1748
1749 if (!lowerdev) {
c7995c43 1750 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 1751 return -ENODEV;
1752 }
d342894c 1753
34e02aa1 1754 if (!tb[IFLA_MTU])
d342894c 1755 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1ba56fb4
AD		 1756
1757 /* update header length based on lower device */
1758 dev->hard_header_len = lowerdev->hard_header_len +
1759 VXLAN_HEADROOM;
d342894c 1760 }
1761
1762 if (data[IFLA_VXLAN_TOS])
1763 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1764
afb97186
VB		 1765 if (data[IFLA_VXLAN_TTL])
1766 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1767
d342894c 1768 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 1769 vxlan->flags |= VXLAN_F_LEARN;
d342894c 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
e4f67add
DS		 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
d342894c 1788 if (data[IFLA_VXLAN_LIMIT])
1789 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1790
05f47d69 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
823aa873 1798 if (data[IFLA_VXLAN_PORT])
1799 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
1800
553675fb 1801 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
1802 pr_info("duplicate VNI %u\n", vni);
1803 return -EEXIST;
1804 }
1805
1b13c97f
YB		 1806 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1807
afbd8bae
MR		 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);
1c51a915 1815 if (err)
553675fb 1816 return err;
d342894c 1817
afbd8bae
MR		 1818 err = register_netdevice(dev);
1819 if (err) {
ba609e9b 1820 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 1821 return err;
1822 }
1823
553675fb 1824 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 1825
1826 return 0;
d342894c 1827}
1828
1829static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1830{
fe5c3561 1831 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1832 struct vxlan_dev *vxlan = netdev_priv(dev);
1833
fe5c3561 1834 flush_workqueue(vxlan_wq);
1835
1836 spin_lock(&vn->sock_lock);
d342894c 1837 hlist_del_rcu(&vxlan->hlist);
fe5c3561 1838 spin_unlock(&vn->sock_lock);
1839
553675fb 1840 list_del(&vxlan->next);
d342894c 1841 unregister_netdevice_queue(dev, head);
1842}
1843
1844static size_t vxlan_get_size(const struct net_device *dev)
1845{
1846
1847 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
5d174dd8 1848 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */
d342894c 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 */
e4f67add
DS		 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 */
d342894c 1858 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1859 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 1860 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 1861 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 1862 0;
1863}
1864
1865static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1866{
1867 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1868 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 1869 struct ifla_vxlan_port_range ports = {
1870 .low = htons(vxlan->port_min),
1871 .high = htons(vxlan->port_max),
1872 };
d342894c 1873
c7995c43 1874 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 1875 goto nla_put_failure;
1876
5d174dd8 1877 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip))
d342894c 1878 goto nla_put_failure;
1879
c7995c43 1880 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 1881 goto nla_put_failure;
1882
7c41c42c 1883 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
d342894c 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) ||
e4f67add
DS		 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)) ||
d342894c 1897 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 1898 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
1899 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 1900 goto nla_put_failure;
1901
05f47d69 1902 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1903 goto nla_put_failure;
1904
d342894c 1905 return 0;
1906
1907nla_put_failure:
1908 return -EMSGSIZE;
1909}
1910
1911static 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
1924static __net_init int vxlan_init_net(struct net *net)
1925{
1926 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 1927 unsigned int h;
d342894c 1928
553675fb 1929 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 1930 spin_lock_init(&vn->sock_lock);
d342894c 1931
553675fb 1932 for (h = 0; h < PORT_HASH_SIZE; ++h)
1933 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 1934
1935 return 0;
1936}
1937
1938static __net_exit void vxlan_exit_net(struct net *net)
1939{
1940 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e 1941 struct vxlan_dev *vxlan;
372675a4 1942 LIST_HEAD(list);
9cb6cb7e
ZM		 1943
1944 rtnl_lock();
553675fb 1945 list_for_each_entry(vxlan, &vn->vxlan_list, next)
372675a4 1946 unregister_netdevice_queue(vxlan->dev, &list);
1947 unregister_netdevice_many(&list);
9cb6cb7e 1948 rtnl_unlock();
d342894c 1949}
1950
1951static 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
1958static int __init vxlan_init_module(void)
1959{
1960 int rc;
1961
758c57d1
SH		 1962 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
1963 if (!vxlan_wq)
1964 return -ENOMEM;
1965
d342894c 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
1978out2:
1979 unregister_pernet_device(&vxlan_net_ops);
1980out1:
758c57d1 1981 destroy_workqueue(vxlan_wq);
d342894c 1982 return rc;
1983}
7332a13b 1984late_initcall(vxlan_init_module);
d342894c 1985
1986static void __exit vxlan_cleanup_module(void)
1987{
b7153984 1988 rtnl_link_unregister(&vxlan_link_ops);
758c57d1 1989 destroy_workqueue(vxlan_wq);
f89e57c4 1990 unregister_pernet_device(&vxlan_net_ops);
6681712d 1991 rcu_barrier();
d342894c 1992}
1993module_exit(vxlan_cleanup_module);
1994
1995MODULE_LICENSE("GPL");
1996MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 1997MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 1998MODULE_ALIAS_RTNL_LINK("vxlan");
ubuntu-bionic-kernel mirror