]> Git Repo - linux.git/blame - drivers/net/vxlan.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: Remove return value from list_netdevice()
[linux.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
4 * Copyright (c) 2012 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 * - use IANA UDP port number (when defined)
12 * - IPv6 (not in RFC)
13 */
14
15#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
16
17#include <linux/kernel.h>
18#include <linux/types.h>
19#include <linux/module.h>
20#include <linux/errno.h>
21#include <linux/slab.h>
22#include <linux/skbuff.h>
23#include <linux/rculist.h>
24#include <linux/netdevice.h>
25#include <linux/in.h>
26#include <linux/ip.h>
27#include <linux/udp.h>
28#include <linux/igmp.h>
29#include <linux/etherdevice.h>
30#include <linux/if_ether.h>
d342894c 31#include <linux/hash.h>
1b13c97f 32#include <linux/ethtool.h>
e4f67add
DS		 33#include <net/arp.h>
34#include <net/ndisc.h>
d342894c 35#include <net/ip.h>
c5441932 36#include <net/ip_tunnels.h>
d342894c 37#include <net/icmp.h>
38#include <net/udp.h>
39#include <net/rtnetlink.h>
40#include <net/route.h>
41#include <net/dsfield.h>
42#include <net/inet_ecn.h>
43#include <net/net_namespace.h>
44#include <net/netns/generic.h>
45
46#define VXLAN_VERSION "0.1"
47
48#define VNI_HASH_BITS 10
49#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
50#define FDB_HASH_BITS 8
51#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
52#define FDB_AGE_DEFAULT 300 /* 5 min */
53#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
54
55#define VXLAN_N_VID (1u << 24)
56#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
52b702ff
AD		 57/* IP header + UDP + VXLAN + Ethernet header */
58#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
d342894c 59
60#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
61
62/* VXLAN protocol header */
63struct vxlanhdr {
64 __be32 vx_flags;
65 __be32 vx_vni;
66};
67
68/* UDP port for VXLAN traffic. */
69static unsigned int vxlan_port __read_mostly = 8472;
70module_param_named(udp_port, vxlan_port, uint, 0444);
71MODULE_PARM_DESC(udp_port, "Destination UDP port");
72
73static bool log_ecn_error = true;
74module_param(log_ecn_error, bool, 0644);
75MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
76
77/* per-net private data for this module */
78static unsigned int vxlan_net_id;
79struct vxlan_net {
80 struct socket *sock; /* UDP encap socket */
81 struct hlist_head vni_list[VNI_HASH_SIZE];
82};
83
6681712d
DS		 84struct vxlan_rdst {
85 struct rcu_head rcu;
86 __be32 remote_ip;
87 __be16 remote_port;
88 u32 remote_vni;
89 u32 remote_ifindex;
90 struct vxlan_rdst *remote_next;
91};
92
d342894c 93/* Forwarding table entry */
94struct vxlan_fdb {
95 struct hlist_node hlist; /* linked list of entries */
96 struct rcu_head rcu;
97 unsigned long updated; /* jiffies */
98 unsigned long used;
6681712d 99 struct vxlan_rdst remote;
d342894c 100 u16 state; /* see ndm_state */
101 u8 eth_addr[ETH_ALEN];
102};
103
d342894c 104/* Pseudo network device */
105struct vxlan_dev {
106 struct hlist_node hlist;
107 struct net_device *dev;
c7995c43 108 struct vxlan_rdst default_dst; /* default destination */
d342894c 109 __be32 saddr; /* source address */
05f47d69 110 __u16 port_min; /* source port range */
111 __u16 port_max;
d342894c 112 __u8 tos; /* TOS override */
113 __u8 ttl;
e4f67add 114 u32 flags; /* VXLAN_F_* below */
d342894c 115
116 unsigned long age_interval;
117 struct timer_list age_timer;
118 spinlock_t hash_lock;
119 unsigned int addrcnt;
120 unsigned int addrmax;
d342894c 121
122 struct hlist_head fdb_head[FDB_HASH_SIZE];
123};
124
e4f67add
DS		 125#define VXLAN_F_LEARN 0x01
126#define VXLAN_F_PROXY 0x02
127#define VXLAN_F_RSC 0x04
128#define VXLAN_F_L2MISS 0x08
129#define VXLAN_F_L3MISS 0x10
130
d342894c 131/* salt for hash table */
132static u32 vxlan_salt __read_mostly;
133
134static inline struct hlist_head *vni_head(struct net *net, u32 id)
135{
136 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
137
138 return &vn->vni_list[hash_32(id, VNI_HASH_BITS)];
139}
140
141/* Look up VNI in a per net namespace table */
142static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id)
143{
144 struct vxlan_dev *vxlan;
d342894c 145
b67bfe0d 146 hlist_for_each_entry_rcu(vxlan, vni_head(net, id), hlist) {
c7995c43 147 if (vxlan->default_dst.remote_vni == id)
d342894c 148 return vxlan;
149 }
150
151 return NULL;
152}
153
154/* Fill in neighbour message in skbuff. */
155static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
156 const struct vxlan_fdb *fdb,
6681712d
DS		 157 u32 portid, u32 seq, int type, unsigned int flags,
158 const struct vxlan_rdst *rdst)
d342894c 159{
160 unsigned long now = jiffies;
161 struct nda_cacheinfo ci;
162 struct nlmsghdr *nlh;
163 struct ndmsg *ndm;
e4f67add 164 bool send_ip, send_eth;
d342894c 165
166 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
167 if (nlh == NULL)
168 return -EMSGSIZE;
169
170 ndm = nlmsg_data(nlh);
171 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 172
173 send_eth = send_ip = true;
174
175 if (type == RTM_GETNEIGH) {
176 ndm->ndm_family = AF_INET;
6681712d 177 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
e4f67add
DS		 178 send_eth = !is_zero_ether_addr(fdb->eth_addr);
179 } else
180 ndm->ndm_family = AF_BRIDGE;
d342894c 181 ndm->ndm_state = fdb->state;
182 ndm->ndm_ifindex = vxlan->dev->ifindex;
183 ndm->ndm_flags = NTF_SELF;
184 ndm->ndm_type = NDA_DST;
185
e4f67add 186 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 187 goto nla_put_failure;
188
6681712d
DS		 189 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
190 goto nla_put_failure;
191
192 if (rdst->remote_port && rdst->remote_port != vxlan_port &&
193 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
194 goto nla_put_failure;
c7995c43 195 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
6681712d
DS		 196 nla_put_be32(skb, NDA_VNI, rdst->remote_vni))
197 goto nla_put_failure;
198 if (rdst->remote_ifindex &&
199 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 200 goto nla_put_failure;
201
202 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
203 ci.ndm_confirmed = 0;
204 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
205 ci.ndm_refcnt = 0;
206
207 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
208 goto nla_put_failure;
209
210 return nlmsg_end(skb, nlh);
211
212nla_put_failure:
213 nlmsg_cancel(skb, nlh);
214 return -EMSGSIZE;
215}
216
217static inline size_t vxlan_nlmsg_size(void)
218{
219 return NLMSG_ALIGN(sizeof(struct ndmsg))
220 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
221 + nla_total_size(sizeof(__be32)) /* NDA_DST */
6681712d
DS		 222 + nla_total_size(sizeof(__be32)) /* NDA_PORT */
223 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
224 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 225 + nla_total_size(sizeof(struct nda_cacheinfo));
226}
227
228static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
229 const struct vxlan_fdb *fdb, int type)
230{
231 struct net *net = dev_net(vxlan->dev);
232 struct sk_buff *skb;
233 int err = -ENOBUFS;
234
235 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
236 if (skb == NULL)
237 goto errout;
238
6681712d 239 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, &fdb->remote);
d342894c 240 if (err < 0) {
241 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
242 WARN_ON(err == -EMSGSIZE);
243 kfree_skb(skb);
244 goto errout;
245 }
246
247 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
248 return;
249errout:
250 if (err < 0)
251 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
252}
253
e4f67add
DS		 254static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
255{
256 struct vxlan_dev *vxlan = netdev_priv(dev);
257 struct vxlan_fdb f;
258
259 memset(&f, 0, sizeof f);
260 f.state = NUD_STALE;
6681712d
DS		 261 f.remote.remote_ip = ipa; /* goes to NDA_DST */
262 f.remote.remote_vni = VXLAN_N_VID;
e4f67add
DS		 263
264 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
265}
266
267static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
268{
269 struct vxlan_fdb f;
270
271 memset(&f, 0, sizeof f);
272 f.state = NUD_STALE;
273 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
274
275 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
276}
277
d342894c 278/* Hash Ethernet address */
279static u32 eth_hash(const unsigned char *addr)
280{
281 u64 value = get_unaligned((u64 *)addr);
282
283 /* only want 6 bytes */
284#ifdef __BIG_ENDIAN
d342894c 285 value >>= 16;
321fb991 286#else
287 value <<= 16;
d342894c 288#endif
289 return hash_64(value, FDB_HASH_BITS);
290}
291
292/* Hash chain to use given mac address */
293static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
294 const u8 *mac)
295{
296 return &vxlan->fdb_head[eth_hash(mac)];
297}
298
299/* Look up Ethernet address in forwarding table */
300static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
301 const u8 *mac)
302
303{
304 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
305 struct vxlan_fdb *f;
d342894c 306
b67bfe0d 307 hlist_for_each_entry_rcu(f, head, hlist) {
d342894c 308 if (compare_ether_addr(mac, f->eth_addr) == 0)
309 return f;
310 }
311
312 return NULL;
313}
314
6681712d
DS		 315/* Add/update destinations for multicast */
316static int vxlan_fdb_append(struct vxlan_fdb *f,
317 __be32 ip, __u32 port, __u32 vni, __u32 ifindex)
318{
319 struct vxlan_rdst *rd_prev, *rd;
320
321 rd_prev = NULL;
322 for (rd = &f->remote; rd; rd = rd->remote_next) {
323 if (rd->remote_ip == ip &&
324 rd->remote_port == port &&
325 rd->remote_vni == vni &&
326 rd->remote_ifindex == ifindex)
327 return 0;
328 rd_prev = rd;
329 }
330 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
331 if (rd == NULL)
332 return -ENOBUFS;
333 rd->remote_ip = ip;
334 rd->remote_port = port;
335 rd->remote_vni = vni;
336 rd->remote_ifindex = ifindex;
337 rd->remote_next = NULL;
338 rd_prev->remote_next = rd;
339 return 1;
340}
341
d342894c 342/* Add new entry to forwarding table -- assumes lock held */
343static int vxlan_fdb_create(struct vxlan_dev *vxlan,
344 const u8 *mac, __be32 ip,
6681712d
DS		 345 __u16 state, __u16 flags,
346 __u32 port, __u32 vni, __u32 ifindex)
d342894c 347{
348 struct vxlan_fdb *f;
349 int notify = 0;
350
351 f = vxlan_find_mac(vxlan, mac);
352 if (f) {
353 if (flags & NLM_F_EXCL) {
354 netdev_dbg(vxlan->dev,
355 "lost race to create %pM\n", mac);
356 return -EEXIST;
357 }
358 if (f->state != state) {
359 f->state = state;
360 f->updated = jiffies;
361 notify = 1;
362 }
6681712d
DS		 363 if ((flags & NLM_F_APPEND) &&
364 is_multicast_ether_addr(f->eth_addr)) {
365 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
366
367 if (rc < 0)
368 return rc;
369 notify |= rc;
370 }
d342894c 371 } else {
372 if (!(flags & NLM_F_CREATE))
373 return -ENOENT;
374
375 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
376 return -ENOSPC;
377
378 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
379 f = kmalloc(sizeof(*f), GFP_ATOMIC);
380 if (!f)
381 return -ENOMEM;
382
383 notify = 1;
6681712d
DS		 384 f->remote.remote_ip = ip;
385 f->remote.remote_port = port;
386 f->remote.remote_vni = vni;
387 f->remote.remote_ifindex = ifindex;
388 f->remote.remote_next = NULL;
d342894c 389 f->state = state;
390 f->updated = f->used = jiffies;
391 memcpy(f->eth_addr, mac, ETH_ALEN);
392
393 ++vxlan->addrcnt;
394 hlist_add_head_rcu(&f->hlist,
395 vxlan_fdb_head(vxlan, mac));
396 }
397
398 if (notify)
399 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
400
401 return 0;
402}
403
6706c82e 404static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 405{
406 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
407
408 while (f->remote.remote_next) {
409 struct vxlan_rdst *rd = f->remote.remote_next;
410
411 f->remote.remote_next = rd->remote_next;
412 kfree(rd);
413 }
414 kfree(f);
415}
416
d342894c 417static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
418{
419 netdev_dbg(vxlan->dev,
420 "delete %pM\n", f->eth_addr);
421
422 --vxlan->addrcnt;
423 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
424
425 hlist_del_rcu(&f->hlist);
6681712d 426 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 427}
428
429/* Add static entry (via netlink) */
430static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
431 struct net_device *dev,
432 const unsigned char *addr, u16 flags)
433{
434 struct vxlan_dev *vxlan = netdev_priv(dev);
6681712d 435 struct net *net = dev_net(vxlan->dev);
d342894c 436 __be32 ip;
6681712d 437 u32 port, vni, ifindex;
d342894c 438 int err;
439
440 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
441 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
442 ndm->ndm_state);
443 return -EINVAL;
444 }
445
446 if (tb[NDA_DST] == NULL)
447 return -EINVAL;
448
449 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
450 return -EAFNOSUPPORT;
451
452 ip = nla_get_be32(tb[NDA_DST]);
453
6681712d
DS		 454 if (tb[NDA_PORT]) {
455 if (nla_len(tb[NDA_PORT]) != sizeof(u32))
456 return -EINVAL;
457 port = nla_get_u32(tb[NDA_PORT]);
458 } else
459 port = vxlan_port;
460
461 if (tb[NDA_VNI]) {
462 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
463 return -EINVAL;
464 vni = nla_get_u32(tb[NDA_VNI]);
465 } else
c7995c43 466 vni = vxlan->default_dst.remote_vni;
6681712d
DS		 467
468 if (tb[NDA_IFINDEX]) {
5abb0029 469 struct net_device *tdev;
6681712d
DS		 470
471 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
472 return -EINVAL;
473 ifindex = nla_get_u32(tb[NDA_IFINDEX]);
5abb0029
PS		 474 tdev = dev_get_by_index(net, ifindex);
475 if (!tdev)
6681712d 476 return -EADDRNOTAVAIL;
5abb0029 477 dev_put(tdev);
6681712d
DS		 478 } else
479 ifindex = 0;
480
d342894c 481 spin_lock_bh(&vxlan->hash_lock);
6681712d
DS		 482 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags, port,
483 vni, ifindex);
d342894c 484 spin_unlock_bh(&vxlan->hash_lock);
485
486 return err;
487}
488
489/* Delete entry (via netlink) */
1690be63
VY		 490static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
491 struct net_device *dev,
d342894c 492 const unsigned char *addr)
493{
494 struct vxlan_dev *vxlan = netdev_priv(dev);
495 struct vxlan_fdb *f;
496 int err = -ENOENT;
497
498 spin_lock_bh(&vxlan->hash_lock);
499 f = vxlan_find_mac(vxlan, addr);
500 if (f) {
501 vxlan_fdb_destroy(vxlan, f);
502 err = 0;
503 }
504 spin_unlock_bh(&vxlan->hash_lock);
505
506 return err;
507}
508
509/* Dump forwarding table */
510static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
511 struct net_device *dev, int idx)
512{
513 struct vxlan_dev *vxlan = netdev_priv(dev);
514 unsigned int h;
515
516 for (h = 0; h < FDB_HASH_SIZE; ++h) {
517 struct vxlan_fdb *f;
d342894c 518 int err;
519
b67bfe0d 520 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d
DS		 521 struct vxlan_rdst *rd;
522 for (rd = &f->remote; rd; rd = rd->remote_next) {
523 if (idx < cb->args[0])
524 goto skip;
525
526 err = vxlan_fdb_info(skb, vxlan, f,
527 NETLINK_CB(cb->skb).portid,
528 cb->nlh->nlmsg_seq,
529 RTM_NEWNEIGH,
530 NLM_F_MULTI, rd);
531 if (err < 0)
532 break;
d342894c 533skip:
6681712d
DS		 534 ++idx;
535 }
d342894c 536 }
537 }
538
539 return idx;
540}
541
542/* Watch incoming packets to learn mapping between Ethernet address
543 * and Tunnel endpoint.
544 */
545static void vxlan_snoop(struct net_device *dev,
546 __be32 src_ip, const u8 *src_mac)
547{
548 struct vxlan_dev *vxlan = netdev_priv(dev);
549 struct vxlan_fdb *f;
550 int err;
551
552 f = vxlan_find_mac(vxlan, src_mac);
553 if (likely(f)) {
554 f->used = jiffies;
6681712d 555 if (likely(f->remote.remote_ip == src_ip))
d342894c 556 return;
557
558 if (net_ratelimit())
559 netdev_info(dev,
560 "%pM migrated from %pI4 to %pI4\n",
6681712d 561 src_mac, &f->remote.remote_ip, &src_ip);
d342894c 562
6681712d 563 f->remote.remote_ip = src_ip;
d342894c 564 f->updated = jiffies;
565 } else {
566 /* learned new entry */
567 spin_lock(&vxlan->hash_lock);
568 err = vxlan_fdb_create(vxlan, src_mac, src_ip,
569 NUD_REACHABLE,
6681712d 570 NLM_F_EXCL|NLM_F_CREATE,
c7995c43 571 vxlan_port, vxlan->default_dst.remote_vni, 0);
d342894c 572 spin_unlock(&vxlan->hash_lock);
573 }
574}
575
576
577/* See if multicast group is already in use by other ID */
578static bool vxlan_group_used(struct vxlan_net *vn,
579 const struct vxlan_dev *this)
580{
581 const struct vxlan_dev *vxlan;
d342894c 582 unsigned h;
583
584 for (h = 0; h < VNI_HASH_SIZE; ++h)
b67bfe0d 585 hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist) {
d342894c 586 if (vxlan == this)
587 continue;
588
589 if (!netif_running(vxlan->dev))
590 continue;
591
c7995c43 592 if (vxlan->default_dst.remote_ip == this->default_dst.remote_ip)
d342894c 593 return true;
594 }
595
596 return false;
597}
598
599/* kernel equivalent to IP_ADD_MEMBERSHIP */
600static int vxlan_join_group(struct net_device *dev)
601{
602 struct vxlan_dev *vxlan = netdev_priv(dev);
603 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
604 struct sock *sk = vn->sock->sk;
605 struct ip_mreqn mreq = {
c7995c43
AW		 606 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
607 .imr_ifindex = vxlan->default_dst.remote_ifindex,
d342894c 608 };
609 int err;
610
611 /* Already a member of group */
612 if (vxlan_group_used(vn, vxlan))
613 return 0;
614
615 /* Need to drop RTNL to call multicast join */
616 rtnl_unlock();
617 lock_sock(sk);
618 err = ip_mc_join_group(sk, &mreq);
619 release_sock(sk);
620 rtnl_lock();
621
622 return err;
623}
624
625
626/* kernel equivalent to IP_DROP_MEMBERSHIP */
627static int vxlan_leave_group(struct net_device *dev)
628{
629 struct vxlan_dev *vxlan = netdev_priv(dev);
630 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
631 int err = 0;
632 struct sock *sk = vn->sock->sk;
633 struct ip_mreqn mreq = {
c7995c43
AW		 634 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
635 .imr_ifindex = vxlan->default_dst.remote_ifindex,
d342894c 636 };
637
638 /* Only leave group when last vxlan is done. */
639 if (vxlan_group_used(vn, vxlan))
640 return 0;
641
642 /* Need to drop RTNL to call multicast leave */
643 rtnl_unlock();
644 lock_sock(sk);
645 err = ip_mc_leave_group(sk, &mreq);
646 release_sock(sk);
647 rtnl_lock();
648
649 return err;
650}
651
652/* Callback from net/ipv4/udp.c to receive packets */
653static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
654{
655 struct iphdr *oip;
656 struct vxlanhdr *vxh;
657 struct vxlan_dev *vxlan;
e8171045 658 struct pcpu_tstats *stats;
d342894c 659 __u32 vni;
660 int err;
661
662 /* pop off outer UDP header */
663 __skb_pull(skb, sizeof(struct udphdr));
664
665 /* Need Vxlan and inner Ethernet header to be present */
666 if (!pskb_may_pull(skb, sizeof(struct vxlanhdr)))
667 goto error;
668
669 /* Drop packets with reserved bits set */
670 vxh = (struct vxlanhdr *) skb->data;
671 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
672 (vxh->vx_vni & htonl(0xff))) {
673 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
674 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
675 goto error;
676 }
677
678 __skb_pull(skb, sizeof(struct vxlanhdr));
d342894c 679
680 /* Is this VNI defined? */
681 vni = ntohl(vxh->vx_vni) >> 8;
682 vxlan = vxlan_find_vni(sock_net(sk), vni);
683 if (!vxlan) {
684 netdev_dbg(skb->dev, "unknown vni %d\n", vni);
685 goto drop;
686 }
687
688 if (!pskb_may_pull(skb, ETH_HLEN)) {
689 vxlan->dev->stats.rx_length_errors++;
690 vxlan->dev->stats.rx_errors++;
691 goto drop;
692 }
693
e4f67add
DS		 694 skb_reset_mac_header(skb);
695
d342894c 696 /* Re-examine inner Ethernet packet */
697 oip = ip_hdr(skb);
698 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 699
700 /* Ignore packet loops (and multicast echo) */
701 if (compare_ether_addr(eth_hdr(skb)->h_source,
702 vxlan->dev->dev_addr) == 0)
703 goto drop;
704
e4f67add 705 if (vxlan->flags & VXLAN_F_LEARN)
d342894c 706 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source);
707
708 __skb_tunnel_rx(skb, vxlan->dev);
709 skb_reset_network_header(skb);
0afb1666
JG		 710
711 /* If the NIC driver gave us an encapsulated packet with
712 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
713 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
714 * for us. Otherwise force the upper layers to verify it.
715 */
716 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
717 !(vxlan->dev->features & NETIF_F_RXCSUM))
718 skb->ip_summed = CHECKSUM_NONE;
719
720 skb->encapsulation = 0;
d342894c 721
722 err = IP_ECN_decapsulate(oip, skb);
723 if (unlikely(err)) {
724 if (log_ecn_error)
725 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
726 &oip->saddr, oip->tos);
727 if (err > 1) {
728 ++vxlan->dev->stats.rx_frame_errors;
729 ++vxlan->dev->stats.rx_errors;
730 goto drop;
731 }
732 }
733
e8171045 734 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 735 u64_stats_update_begin(&stats->syncp);
736 stats->rx_packets++;
737 stats->rx_bytes += skb->len;
738 u64_stats_update_end(&stats->syncp);
739
740 netif_rx(skb);
741
742 return 0;
743error:
744 /* Put UDP header back */
745 __skb_push(skb, sizeof(struct udphdr));
746
747 return 1;
748drop:
749 /* Consume bad packet */
750 kfree_skb(skb);
751 return 0;
752}
753
e4f67add
DS		 754static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
755{
756 struct vxlan_dev *vxlan = netdev_priv(dev);
757 struct arphdr *parp;
758 u8 *arpptr, *sha;
759 __be32 sip, tip;
760 struct neighbour *n;
761
762 if (dev->flags & IFF_NOARP)
763 goto out;
764
765 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
766 dev->stats.tx_dropped++;
767 goto out;
768 }
769 parp = arp_hdr(skb);
770
771 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
772 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
773 parp->ar_pro != htons(ETH_P_IP) ||
774 parp->ar_op != htons(ARPOP_REQUEST) ||
775 parp->ar_hln != dev->addr_len ||
776 parp->ar_pln != 4)
777 goto out;
778 arpptr = (u8 *)parp + sizeof(struct arphdr);
779 sha = arpptr;
780 arpptr += dev->addr_len; /* sha */
781 memcpy(&sip, arpptr, sizeof(sip));
782 arpptr += sizeof(sip);
783 arpptr += dev->addr_len; /* tha */
784 memcpy(&tip, arpptr, sizeof(tip));
785
786 if (ipv4_is_loopback(tip) ||
787 ipv4_is_multicast(tip))
788 goto out;
789
790 n = neigh_lookup(&arp_tbl, &tip, dev);
791
792 if (n) {
e4f67add
DS		 793 struct vxlan_fdb *f;
794 struct sk_buff *reply;
795
796 if (!(n->nud_state & NUD_CONNECTED)) {
797 neigh_release(n);
798 goto out;
799 }
800
801 f = vxlan_find_mac(vxlan, n->ha);
6681712d 802 if (f && f->remote.remote_ip == htonl(INADDR_ANY)) {
e4f67add
DS		 803 /* bridge-local neighbor */
804 neigh_release(n);
805 goto out;
806 }
807
808 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
809 n->ha, sha);
810
811 neigh_release(n);
812
813 skb_reset_mac_header(reply);
814 __skb_pull(reply, skb_network_offset(reply));
815 reply->ip_summed = CHECKSUM_UNNECESSARY;
816 reply->pkt_type = PACKET_HOST;
817
818 if (netif_rx_ni(reply) == NET_RX_DROP)
819 dev->stats.rx_dropped++;
820 } else if (vxlan->flags & VXLAN_F_L3MISS)
821 vxlan_ip_miss(dev, tip);
822out:
823 consume_skb(skb);
824 return NETDEV_TX_OK;
825}
826
827static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
828{
829 struct vxlan_dev *vxlan = netdev_priv(dev);
830 struct neighbour *n;
831 struct iphdr *pip;
832
833 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
834 return false;
835
836 n = NULL;
837 switch (ntohs(eth_hdr(skb)->h_proto)) {
838 case ETH_P_IP:
839 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
840 return false;
841 pip = ip_hdr(skb);
842 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
843 break;
844 default:
845 return false;
846 }
847
848 if (n) {
849 bool diff;
850
851 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
852 if (diff) {
853 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
854 dev->addr_len);
855 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
856 }
857 neigh_release(n);
858 return diff;
859 } else if (vxlan->flags & VXLAN_F_L3MISS)
860 vxlan_ip_miss(dev, pip->daddr);
861 return false;
862}
863
1cad8715 864static void vxlan_sock_free(struct sk_buff *skb)
865{
866 sock_put(skb->sk);
867}
868
869/* On transmit, associate with the tunnel socket */
870static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
871{
872 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
873 struct sock *sk = vn->sock->sk;
874
875 skb_orphan(skb);
876 sock_hold(sk);
877 skb->sk = sk;
878 skb->destructor = vxlan_sock_free;
879}
880
05f47d69 881/* Compute source port for outgoing packet
882 * first choice to use L4 flow hash since it will spread
883 * better and maybe available from hardware
884 * secondary choice is to use jhash on the Ethernet header
885 */
886static u16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
887{
888 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
889 u32 hash;
890
891 hash = skb_get_rxhash(skb);
892 if (!hash)
893 hash = jhash(skb->data, 2 * ETH_ALEN,
894 (__force u32) skb->protocol);
895
896 return (((u64) hash * range) >> 32) + vxlan->port_min;
897}
898
05c0db08
PS		 899static int handle_offloads(struct sk_buff *skb)
900{
901 if (skb_is_gso(skb)) {
902 int err = skb_unclone(skb, GFP_ATOMIC);
903 if (unlikely(err))
904 return err;
905
906 skb_shinfo(skb)->gso_type |= (SKB_GSO_UDP_TUNNEL | SKB_GSO_UDP);
907 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
908 skb->ip_summed = CHECKSUM_NONE;
909
910 return 0;
911}
912
9dcc71e1
SS		 913/* Bypass encapsulation if the destination is local */
914static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
915 struct vxlan_dev *dst_vxlan)
916{
917 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
918 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
919
920 skb->pkt_type = PACKET_HOST;
921 skb->encapsulation = 0;
922 skb->dev = dst_vxlan->dev;
923 __skb_pull(skb, skb_network_offset(skb));
924
925 if (dst_vxlan->flags & VXLAN_F_LEARN)
9d9f163c
MR		 926 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
927 eth_hdr(skb)->h_source);
9dcc71e1
SS		 928
929 u64_stats_update_begin(&tx_stats->syncp);
930 tx_stats->tx_packets++;
931 tx_stats->tx_bytes += skb->len;
932 u64_stats_update_end(&tx_stats->syncp);
933
934 if (netif_rx(skb) == NET_RX_SUCCESS) {
935 u64_stats_update_begin(&rx_stats->syncp);
936 rx_stats->rx_packets++;
937 rx_stats->rx_bytes += skb->len;
938 u64_stats_update_end(&rx_stats->syncp);
939 } else {
940 skb->dev->stats.rx_dropped++;
941 }
942}
943
6681712d
DS		 944static netdev_tx_t vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
945 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 946{
947 struct vxlan_dev *vxlan = netdev_priv(dev);
948 struct rtable *rt;
d342894c 949 const struct iphdr *old_iph;
950 struct iphdr *iph;
951 struct vxlanhdr *vxh;
952 struct udphdr *uh;
953 struct flowi4 fl4;
d342894c 954 __be32 dst;
6681712d
DS		 955 __u16 src_port, dst_port;
956 u32 vni;
d342894c 957 __be16 df = 0;
958 __u8 tos, ttl;
d342894c 959
6681712d
DS		 960 dst_port = rdst->remote_port ? rdst->remote_port : vxlan_port;
961 vni = rdst->remote_vni;
962 dst = rdst->remote_ip;
e4f67add
DS		 963
964 if (!dst) {
965 if (did_rsc) {
e4f67add 966 /* short-circuited back to local bridge */
9dcc71e1 967 vxlan_encap_bypass(skb, vxlan, vxlan);
e4f67add
DS		 968 return NETDEV_TX_OK;
969 }
ef59febe 970 goto drop;
e4f67add 971 }
ef59febe 972
d6727fe3
JG		 973 if (!skb->encapsulation) {
974 skb_reset_inner_headers(skb);
975 skb->encapsulation = 1;
976 }
977
d342894c 978 /* Need space for new headers (invalidates iph ptr) */
979 if (skb_cow_head(skb, VXLAN_HEADROOM))
980 goto drop;
981
d342894c 982 old_iph = ip_hdr(skb);
983
d342894c 984 ttl = vxlan->ttl;
985 if (!ttl && IN_MULTICAST(ntohl(dst)))
986 ttl = 1;
987
988 tos = vxlan->tos;
989 if (tos == 1)
206aaafc 990 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 991
05f47d69 992 src_port = vxlan_src_port(vxlan, skb);
d342894c 993
ca78f181 994 memset(&fl4, 0, sizeof(fl4));
6681712d 995 fl4.flowi4_oif = rdst->remote_ifindex;
ca78f181 996 fl4.flowi4_tos = RT_TOS(tos);
997 fl4.daddr = dst;
998 fl4.saddr = vxlan->saddr;
999
1000 rt = ip_route_output_key(dev_net(dev), &fl4);
d342894c 1001 if (IS_ERR(rt)) {
1002 netdev_dbg(dev, "no route to %pI4\n", &dst);
1003 dev->stats.tx_carrier_errors++;
1004 goto tx_error;
1005 }
1006
1007 if (rt->dst.dev == dev) {
1008 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1009 ip_rt_put(rt);
1010 dev->stats.collisions++;
1011 goto tx_error;
1012 }
1013
9dcc71e1 1014 /* Bypass encapsulation if the destination is local */
ab09a6d0
MR		 1015 if (rt->rt_flags & RTCF_LOCAL &&
1016 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
9dcc71e1
SS		 1017 struct vxlan_dev *dst_vxlan;
1018
1019 ip_rt_put(rt);
1020 dst_vxlan = vxlan_find_vni(dev_net(dev), vni);
1021 if (!dst_vxlan)
1022 goto tx_error;
1023 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1024 return NETDEV_TX_OK;
1025 }
1026
d342894c 1027 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1028 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1029 IPSKB_REROUTED);
1030 skb_dst_drop(skb);
1031 skb_dst_set(skb, &rt->dst);
1032
1033 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1034 vxh->vx_flags = htonl(VXLAN_FLAGS);
6681712d 1035 vxh->vx_vni = htonl(vni << 8);
d342894c 1036
1037 __skb_push(skb, sizeof(*uh));
1038 skb_reset_transport_header(skb);
1039 uh = udp_hdr(skb);
1040
6681712d 1041 uh->dest = htons(dst_port);
05f47d69 1042 uh->source = htons(src_port);
d342894c 1043
1044 uh->len = htons(skb->len);
1045 uh->check = 0;
1046
1047 __skb_push(skb, sizeof(*iph));
1048 skb_reset_network_header(skb);
1049 iph = ip_hdr(skb);
1050 iph->version = 4;
1051 iph->ihl = sizeof(struct iphdr) >> 2;
1052 iph->frag_off = df;
1053 iph->protocol = IPPROTO_UDP;
206aaafc 1054 iph->tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
ca78f181 1055 iph->daddr = dst;
d342894c 1056 iph->saddr = fl4.saddr;
1057 iph->ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
8dc98eb2 1058 tunnel_ip_select_ident(skb, old_iph, &rt->dst);
d342894c 1059
88c4c066
ZM		 1060 nf_reset(skb);
1061
1cad8715 1062 vxlan_set_owner(dev, skb);
1063
05c0db08
PS		 1064 if (handle_offloads(skb))
1065 goto drop;
d342894c 1066
6aed0c8b 1067 iptunnel_xmit(skb, dev);
d342894c 1068 return NETDEV_TX_OK;
1069
1070drop:
1071 dev->stats.tx_dropped++;
1072 goto tx_free;
1073
1074tx_error:
1075 dev->stats.tx_errors++;
1076tx_free:
1077 dev_kfree_skb(skb);
1078 return NETDEV_TX_OK;
1079}
1080
6681712d
DS		 1081/* Transmit local packets over Vxlan
1082 *
1083 * Outer IP header inherits ECN and DF from inner header.
1084 * Outer UDP destination is the VXLAN assigned port.
1085 * source port is based on hash of flow
1086 */
1087static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1088{
1089 struct vxlan_dev *vxlan = netdev_priv(dev);
1090 struct ethhdr *eth;
1091 bool did_rsc = false;
c7995c43 1092 struct vxlan_rdst *rdst0, *rdst;
6681712d
DS		 1093 struct vxlan_fdb *f;
1094 int rc1, rc;
1095
1096 skb_reset_mac_header(skb);
1097 eth = eth_hdr(skb);
1098
1099 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1100 return arp_reduce(dev, skb);
1101 else if ((vxlan->flags&VXLAN_F_RSC) && ntohs(eth->h_proto) == ETH_P_IP)
1102 did_rsc = route_shortcircuit(dev, skb);
1103
1104 f = vxlan_find_mac(vxlan, eth->h_dest);
1105 if (f == NULL) {
1106 did_rsc = false;
c7995c43
AW		 1107 rdst0 = &vxlan->default_dst;
1108
1109 if (rdst0->remote_ip == htonl(INADDR_ANY) &&
6681712d
DS		 1110 (vxlan->flags & VXLAN_F_L2MISS) &&
1111 !is_multicast_ether_addr(eth->h_dest))
1112 vxlan_fdb_miss(vxlan, eth->h_dest);
1113 } else
1114 rdst0 = &f->remote;
1115
1116 rc = NETDEV_TX_OK;
1117
1118 /* if there are multiple destinations, send copies */
1119 for (rdst = rdst0->remote_next; rdst; rdst = rdst->remote_next) {
1120 struct sk_buff *skb1;
1121
1122 skb1 = skb_clone(skb, GFP_ATOMIC);
1123 rc1 = vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1124 if (rc == NETDEV_TX_OK)
1125 rc = rc1;
1126 }
1127
1128 rc1 = vxlan_xmit_one(skb, dev, rdst0, did_rsc);
1129 if (rc == NETDEV_TX_OK)
1130 rc = rc1;
1131 return rc;
1132}
1133
d342894c 1134/* Walk the forwarding table and purge stale entries */
1135static void vxlan_cleanup(unsigned long arg)
1136{
1137 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1138 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1139 unsigned int h;
1140
1141 if (!netif_running(vxlan->dev))
1142 return;
1143
1144 spin_lock_bh(&vxlan->hash_lock);
1145 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1146 struct hlist_node *p, *n;
1147 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1148 struct vxlan_fdb *f
1149 = container_of(p, struct vxlan_fdb, hlist);
1150 unsigned long timeout;
1151
3c172868 1152 if (f->state & NUD_PERMANENT)
d342894c 1153 continue;
1154
1155 timeout = f->used + vxlan->age_interval * HZ;
1156 if (time_before_eq(timeout, jiffies)) {
1157 netdev_dbg(vxlan->dev,
1158 "garbage collect %pM\n",
1159 f->eth_addr);
1160 f->state = NUD_STALE;
1161 vxlan_fdb_destroy(vxlan, f);
1162 } else if (time_before(timeout, next_timer))
1163 next_timer = timeout;
1164 }
1165 }
1166 spin_unlock_bh(&vxlan->hash_lock);
1167
1168 mod_timer(&vxlan->age_timer, next_timer);
1169}
1170
1171/* Setup stats when device is created */
1172static int vxlan_init(struct net_device *dev)
1173{
e8171045
PS		 1174 dev->tstats = alloc_percpu(struct pcpu_tstats);
1175 if (!dev->tstats)
d342894c 1176 return -ENOMEM;
1177
1178 return 0;
1179}
1180
1181/* Start ageing timer and join group when device is brought up */
1182static int vxlan_open(struct net_device *dev)
1183{
1184 struct vxlan_dev *vxlan = netdev_priv(dev);
1185 int err;
1186
c7995c43 1187 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
d342894c 1188 err = vxlan_join_group(dev);
1189 if (err)
1190 return err;
1191 }
1192
1193 if (vxlan->age_interval)
1194 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1195
1196 return 0;
1197}
1198
1199/* Purge the forwarding table */
1200static void vxlan_flush(struct vxlan_dev *vxlan)
1201{
1202 unsigned h;
1203
1204 spin_lock_bh(&vxlan->hash_lock);
1205 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1206 struct hlist_node *p, *n;
1207 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1208 struct vxlan_fdb *f
1209 = container_of(p, struct vxlan_fdb, hlist);
1210 vxlan_fdb_destroy(vxlan, f);
1211 }
1212 }
1213 spin_unlock_bh(&vxlan->hash_lock);
1214}
1215
1216/* Cleanup timer and forwarding table on shutdown */
1217static int vxlan_stop(struct net_device *dev)
1218{
1219 struct vxlan_dev *vxlan = netdev_priv(dev);
1220
c7995c43 1221 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)))
d342894c 1222 vxlan_leave_group(dev);
1223
1224 del_timer_sync(&vxlan->age_timer);
1225
1226 vxlan_flush(vxlan);
1227
1228 return 0;
1229}
1230
d342894c 1231/* Stub, nothing needs to be done. */
1232static void vxlan_set_multicast_list(struct net_device *dev)
1233{
1234}
1235
1236static const struct net_device_ops vxlan_netdev_ops = {
1237 .ndo_init = vxlan_init,
1238 .ndo_open = vxlan_open,
1239 .ndo_stop = vxlan_stop,
1240 .ndo_start_xmit = vxlan_xmit,
e8171045 1241 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 1242 .ndo_set_rx_mode = vxlan_set_multicast_list,
1243 .ndo_change_mtu = eth_change_mtu,
1244 .ndo_validate_addr = eth_validate_addr,
1245 .ndo_set_mac_address = eth_mac_addr,
1246 .ndo_fdb_add = vxlan_fdb_add,
1247 .ndo_fdb_del = vxlan_fdb_delete,
1248 .ndo_fdb_dump = vxlan_fdb_dump,
1249};
1250
1251/* Info for udev, that this is a virtual tunnel endpoint */
1252static struct device_type vxlan_type = {
1253 .name = "vxlan",
1254};
1255
1256static void vxlan_free(struct net_device *dev)
1257{
e8171045 1258 free_percpu(dev->tstats);
d342894c 1259 free_netdev(dev);
1260}
1261
1262/* Initialize the device structure. */
1263static void vxlan_setup(struct net_device *dev)
1264{
1265 struct vxlan_dev *vxlan = netdev_priv(dev);
1266 unsigned h;
05f47d69 1267 int low, high;
d342894c 1268
1269 eth_hw_addr_random(dev);
1270 ether_setup(dev);
2840bf22 1271 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 1272
1273 dev->netdev_ops = &vxlan_netdev_ops;
1274 dev->destructor = vxlan_free;
1275 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1276
1277 dev->tx_queue_len = 0;
1278 dev->features |= NETIF_F_LLTX;
1279 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 1280 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 1281 dev->features |= NETIF_F_RXCSUM;
05c0db08 1282 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666
JG		 1283
1284 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 1285 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
d342894c 1286 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 1287 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 1288
1289 spin_lock_init(&vxlan->hash_lock);
1290
1291 init_timer_deferrable(&vxlan->age_timer);
1292 vxlan->age_timer.function = vxlan_cleanup;
1293 vxlan->age_timer.data = (unsigned long) vxlan;
1294
05f47d69 1295 inet_get_local_port_range(&low, &high);
1296 vxlan->port_min = low;
1297 vxlan->port_max = high;
1298
d342894c 1299 vxlan->dev = dev;
1300
1301 for (h = 0; h < FDB_HASH_SIZE; ++h)
1302 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1303}
1304
1305static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1306 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
c7995c43 1307 [IFLA_VXLAN_REMOTE] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
d342894c 1308 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1309 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1310 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1311 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1312 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1313 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1314 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 1315 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 1316 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1317 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1318 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1319 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
d342894c 1320};
1321
1322static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1323{
1324 if (tb[IFLA_ADDRESS]) {
1325 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1326 pr_debug("invalid link address (not ethernet)\n");
1327 return -EINVAL;
1328 }
1329
1330 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1331 pr_debug("invalid all zero ethernet address\n");
1332 return -EADDRNOTAVAIL;
1333 }
1334 }
1335
1336 if (!data)
1337 return -EINVAL;
1338
1339 if (data[IFLA_VXLAN_ID]) {
1340 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1341 if (id >= VXLAN_VID_MASK)
1342 return -ERANGE;
1343 }
1344
05f47d69 1345 if (data[IFLA_VXLAN_PORT_RANGE]) {
1346 const struct ifla_vxlan_port_range *p
1347 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1348
1349 if (ntohs(p->high) < ntohs(p->low)) {
1350 pr_debug("port range %u .. %u not valid\n",
1351 ntohs(p->low), ntohs(p->high));
1352 return -EINVAL;
1353 }
1354 }
1355
d342894c 1356 return 0;
1357}
1358
1b13c97f
YB		 1359static void vxlan_get_drvinfo(struct net_device *netdev,
1360 struct ethtool_drvinfo *drvinfo)
1361{
1362 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1363 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1364}
1365
1366static const struct ethtool_ops vxlan_ethtool_ops = {
1367 .get_drvinfo = vxlan_get_drvinfo,
1368 .get_link = ethtool_op_get_link,
1369};
1370
d342894c 1371static int vxlan_newlink(struct net *net, struct net_device *dev,
1372 struct nlattr *tb[], struct nlattr *data[])
1373{
1374 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1375 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 1376 __u32 vni;
1377 int err;
1378
1379 if (!data[IFLA_VXLAN_ID])
1380 return -EINVAL;
1381
1382 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
1383 if (vxlan_find_vni(net, vni)) {
1384 pr_info("duplicate VNI %u\n", vni);
1385 return -EEXIST;
1386 }
c7995c43 1387 dst->remote_vni = vni;
d342894c 1388
c7995c43
AW		 1389 if (data[IFLA_VXLAN_REMOTE])
1390 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_REMOTE]);
d342894c 1391
1392 if (data[IFLA_VXLAN_LOCAL])
1393 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1394
34e02aa1 1395 if (data[IFLA_VXLAN_LINK] &&
c7995c43 1396 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 1397 struct net_device *lowerdev
c7995c43 1398 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 1399
1400 if (!lowerdev) {
c7995c43 1401 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 1402 return -ENODEV;
1403 }
d342894c 1404
34e02aa1 1405 if (!tb[IFLA_MTU])
d342894c 1406 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1ba56fb4
AD		 1407
1408 /* update header length based on lower device */
1409 dev->hard_header_len = lowerdev->hard_header_len +
1410 VXLAN_HEADROOM;
d342894c 1411 }
1412
1413 if (data[IFLA_VXLAN_TOS])
1414 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1415
afb97186
VB		 1416 if (data[IFLA_VXLAN_TTL])
1417 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1418
d342894c 1419 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 1420 vxlan->flags |= VXLAN_F_LEARN;
d342894c 1421
1422 if (data[IFLA_VXLAN_AGEING])
1423 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1424 else
1425 vxlan->age_interval = FDB_AGE_DEFAULT;
1426
e4f67add
DS		 1427 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1428 vxlan->flags |= VXLAN_F_PROXY;
1429
1430 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1431 vxlan->flags |= VXLAN_F_RSC;
1432
1433 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1434 vxlan->flags |= VXLAN_F_L2MISS;
1435
1436 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1437 vxlan->flags |= VXLAN_F_L3MISS;
1438
d342894c 1439 if (data[IFLA_VXLAN_LIMIT])
1440 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1441
05f47d69 1442 if (data[IFLA_VXLAN_PORT_RANGE]) {
1443 const struct ifla_vxlan_port_range *p
1444 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1445 vxlan->port_min = ntohs(p->low);
1446 vxlan->port_max = ntohs(p->high);
1447 }
1448
1b13c97f
YB		 1449 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1450
d342894c 1451 err = register_netdevice(dev);
1452 if (!err)
c7995c43 1453 hlist_add_head_rcu(&vxlan->hlist, vni_head(net, dst->remote_vni));
d342894c 1454
1455 return err;
1456}
1457
1458static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1459{
1460 struct vxlan_dev *vxlan = netdev_priv(dev);
1461
1462 hlist_del_rcu(&vxlan->hlist);
1463
1464 unregister_netdevice_queue(dev, head);
1465}
1466
1467static size_t vxlan_get_size(const struct net_device *dev)
1468{
1469
1470 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
c7995c43 1471 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_REMOTE */
d342894c 1472 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1473 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1474 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1475 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1476 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 1477 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1478 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1479 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1480 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 1481 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1482 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 1483 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
d342894c 1484 0;
1485}
1486
1487static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1488{
1489 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1490 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 1491 struct ifla_vxlan_port_range ports = {
1492 .low = htons(vxlan->port_min),
1493 .high = htons(vxlan->port_max),
1494 };
d342894c 1495
c7995c43 1496 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 1497 goto nla_put_failure;
1498
c7995c43 1499 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_REMOTE, dst->remote_ip))
d342894c 1500 goto nla_put_failure;
1501
c7995c43 1502 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 1503 goto nla_put_failure;
1504
7c41c42c 1505 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
d342894c 1506 goto nla_put_failure;
1507
1508 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1509 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 1510 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1511 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1512 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1513 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1514 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1515 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1516 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1517 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1518 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 1519 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
1520 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax))
1521 goto nla_put_failure;
1522
05f47d69 1523 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1524 goto nla_put_failure;
1525
d342894c 1526 return 0;
1527
1528nla_put_failure:
1529 return -EMSGSIZE;
1530}
1531
1532static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1533 .kind = "vxlan",
1534 .maxtype = IFLA_VXLAN_MAX,
1535 .policy = vxlan_policy,
1536 .priv_size = sizeof(struct vxlan_dev),
1537 .setup = vxlan_setup,
1538 .validate = vxlan_validate,
1539 .newlink = vxlan_newlink,
1540 .dellink = vxlan_dellink,
1541 .get_size = vxlan_get_size,
1542 .fill_info = vxlan_fill_info,
1543};
1544
1545static __net_init int vxlan_init_net(struct net *net)
1546{
1547 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1548 struct sock *sk;
1549 struct sockaddr_in vxlan_addr = {
1550 .sin_family = AF_INET,
1551 .sin_addr.s_addr = htonl(INADDR_ANY),
1552 };
1553 int rc;
1554 unsigned h;
1555
1556 /* Create UDP socket for encapsulation receive. */
1557 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vn->sock);
1558 if (rc < 0) {
1559 pr_debug("UDP socket create failed\n");
1560 return rc;
1561 }
bfe1b9b1 1562 /* Put in proper namespace */
1563 sk = vn->sock->sk;
1564 sk_change_net(sk, net);
d342894c 1565
1566 vxlan_addr.sin_port = htons(vxlan_port);
1567
1568 rc = kernel_bind(vn->sock, (struct sockaddr *) &vxlan_addr,
1569 sizeof(vxlan_addr));
1570 if (rc < 0) {
1571 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1572 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
bfe1b9b1 1573 sk_release_kernel(sk);
d342894c 1574 vn->sock = NULL;
1575 return rc;
1576 }
1577
1578 /* Disable multicast loopback */
d342894c 1579 inet_sk(sk)->mc_loop = 0;
1580
1581 /* Mark socket as an encapsulation socket. */
1582 udp_sk(sk)->encap_type = 1;
1583 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1584 udp_encap_enable();
1585
1586 for (h = 0; h < VNI_HASH_SIZE; ++h)
1587 INIT_HLIST_HEAD(&vn->vni_list[h]);
1588
1589 return 0;
1590}
1591
1592static __net_exit void vxlan_exit_net(struct net *net)
1593{
1594 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e
ZM		 1595 struct vxlan_dev *vxlan;
1596 unsigned h;
1597
1598 rtnl_lock();
1599 for (h = 0; h < VNI_HASH_SIZE; ++h)
1600 hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist)
1601 dev_close(vxlan->dev);
1602 rtnl_unlock();
d342894c 1603
1604 if (vn->sock) {
bfe1b9b1 1605 sk_release_kernel(vn->sock->sk);
d342894c 1606 vn->sock = NULL;
1607 }
1608}
1609
1610static struct pernet_operations vxlan_net_ops = {
1611 .init = vxlan_init_net,
1612 .exit = vxlan_exit_net,
1613 .id = &vxlan_net_id,
1614 .size = sizeof(struct vxlan_net),
1615};
1616
1617static int __init vxlan_init_module(void)
1618{
1619 int rc;
1620
1621 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1622
1623 rc = register_pernet_device(&vxlan_net_ops);
1624 if (rc)
1625 goto out1;
1626
1627 rc = rtnl_link_register(&vxlan_link_ops);
1628 if (rc)
1629 goto out2;
1630
1631 return 0;
1632
1633out2:
1634 unregister_pernet_device(&vxlan_net_ops);
1635out1:
1636 return rc;
1637}
1638module_init(vxlan_init_module);
1639
1640static void __exit vxlan_cleanup_module(void)
1641{
1642 rtnl_link_unregister(&vxlan_link_ops);
1643 unregister_pernet_device(&vxlan_net_ops);
6681712d 1644 rcu_barrier();
d342894c 1645}
1646module_exit(vxlan_cleanup_module);
1647
1648MODULE_LICENSE("GPL");
1649MODULE_VERSION(VXLAN_VERSION);
1650MODULE_AUTHOR("Stephen Hemminger <[email protected]>");
1651MODULE_ALIAS_RTNL_LINK("vxlan");
Empty description
This page took 0.367425 seconds and 4 git commands to generate.