projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
vxlan: Extend vxlan handlers for openvswitch.
[linux-2.6-block.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>
012a5729 44#include <net/vxlan.h>
d342894c 45
46#define VXLAN_VERSION "0.1"
47
553675fb 48#define PORT_HASH_BITS 8
49#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
d342894c 50#define VNI_HASH_BITS 10
51#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
52#define FDB_HASH_BITS 8
53#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
54#define FDB_AGE_DEFAULT 300 /* 5 min */
55#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
56
57#define VXLAN_N_VID (1u << 24)
58#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
52b702ff
AD		 59/* IP header + UDP + VXLAN + Ethernet header */
60#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
7ce04758 61#define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr))
d342894c 62
63#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
64
65/* VXLAN protocol header */
66struct vxlanhdr {
67 __be32 vx_flags;
68 __be32 vx_vni;
69};
70
23c578bf 71/* UDP port for VXLAN traffic.
72 * The IANA assigned port is 4789, but the Linux default is 8472
234f5b73 73 * for compatibility with early adopters.
23c578bf 74 */
9daaa397
SH		 75static unsigned short vxlan_port __read_mostly = 8472;
76module_param_named(udp_port, vxlan_port, ushort, 0444);
d342894c 77MODULE_PARM_DESC(udp_port, "Destination UDP port");
78
79static bool log_ecn_error = true;
80module_param(log_ecn_error, bool, 0644);
81MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
82
60d9d4c6 83static int vxlan_net_id;
553675fb 84
afbd8bae
MR		 85static const u8 all_zeros_mac[ETH_ALEN];
86
553675fb 87/* per-network namespace private data for this module */
88struct vxlan_net {
89 struct list_head vxlan_list;
90 struct hlist_head sock_list[PORT_HASH_SIZE];
1c51a915 91 spinlock_t sock_lock;
553675fb 92};
93
6681712d 94struct vxlan_rdst {
6681712d
DS		 95 __be32 remote_ip;
96 __be16 remote_port;
97 u32 remote_vni;
98 u32 remote_ifindex;
3e61aa8f 99 struct list_head list;
bc7892ba 100 struct rcu_head rcu;
6681712d
DS		 101};
102
d342894c 103/* Forwarding table entry */
104struct vxlan_fdb {
105 struct hlist_node hlist; /* linked list of entries */
106 struct rcu_head rcu;
107 unsigned long updated; /* jiffies */
108 unsigned long used;
3e61aa8f 109 struct list_head remotes;
d342894c 110 u16 state; /* see ndm_state */
ae884082 111 u8 flags; /* see ndm_flags */
d342894c 112 u8 eth_addr[ETH_ALEN];
113};
114
d342894c 115/* Pseudo network device */
116struct vxlan_dev {
553675fb 117 struct hlist_node hlist; /* vni hash table */
118 struct list_head next; /* vxlan's per namespace list */
119 struct vxlan_sock *vn_sock; /* listening socket */
d342894c 120 struct net_device *dev;
c7995c43 121 struct vxlan_rdst default_dst; /* default destination */
d342894c 122 __be32 saddr; /* source address */
823aa873 123 __be16 dst_port;
05f47d69 124 __u16 port_min; /* source port range */
125 __u16 port_max;
d342894c 126 __u8 tos; /* TOS override */
127 __u8 ttl;
e4f67add 128 u32 flags; /* VXLAN_F_* below */
d342894c 129
1c51a915 130 struct work_struct sock_work;
3fc2de2f 131 struct work_struct igmp_join;
132 struct work_struct igmp_leave;
1c51a915 133
d342894c 134 unsigned long age_interval;
135 struct timer_list age_timer;
136 spinlock_t hash_lock;
137 unsigned int addrcnt;
138 unsigned int addrmax;
d342894c 139
140 struct hlist_head fdb_head[FDB_HASH_SIZE];
141};
142
e4f67add
DS		 143#define VXLAN_F_LEARN 0x01
144#define VXLAN_F_PROXY 0x02
145#define VXLAN_F_RSC 0x04
146#define VXLAN_F_L2MISS 0x08
147#define VXLAN_F_L3MISS 0x10
148
d342894c 149/* salt for hash table */
150static u32 vxlan_salt __read_mostly;
758c57d1 151static struct workqueue_struct *vxlan_wq;
d342894c 152
1c51a915
SH		 153static void vxlan_sock_work(struct work_struct *work);
154
553675fb 155/* Virtual Network hash table head */
156static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
157{
158 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
159}
160
161/* Socket hash table head */
162static inline struct hlist_head *vs_head(struct net *net, __be16 port)
d342894c 163{
164 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
165
553675fb 166 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
167}
168
3e61aa8f
SH		 169/* First remote destination for a forwarding entry.
170 * Guaranteed to be non-NULL because remotes are never deleted.
171 */
5ca5461c 172static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
3e61aa8f 173{
5ca5461c 174 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
175}
176
177static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
178{
179 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
3e61aa8f
SH		 180}
181
553675fb 182/* Find VXLAN socket based on network namespace and UDP port */
9c2e24e1 183static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port)
553675fb 184{
185 struct vxlan_sock *vs;
186
187 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
188 if (inet_sk(vs->sock->sk)->inet_sport == port)
189 return vs;
190 }
191 return NULL;
d342894c 192}
193
5cfccc5a 194static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
d342894c 195{
196 struct vxlan_dev *vxlan;
d342894c 197
553675fb 198 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
c7995c43 199 if (vxlan->default_dst.remote_vni == id)
d342894c 200 return vxlan;
201 }
202
203 return NULL;
204}
205
5cfccc5a
PS		 206/* Look up VNI in a per net namespace table */
207static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port)
208{
209 struct vxlan_sock *vs;
210
211 vs = vxlan_find_sock(net, port);
212 if (!vs)
213 return NULL;
214
215 return vxlan_vs_find_vni(vs, id);
216}
217
d342894c 218/* Fill in neighbour message in skbuff. */
219static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
234f5b73
SH		 220 const struct vxlan_fdb *fdb,
221 u32 portid, u32 seq, int type, unsigned int flags,
222 const struct vxlan_rdst *rdst)
d342894c 223{
224 unsigned long now = jiffies;
225 struct nda_cacheinfo ci;
226 struct nlmsghdr *nlh;
227 struct ndmsg *ndm;
e4f67add 228 bool send_ip, send_eth;
d342894c 229
230 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
231 if (nlh == NULL)
232 return -EMSGSIZE;
233
234 ndm = nlmsg_data(nlh);
235 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 236
237 send_eth = send_ip = true;
238
239 if (type == RTM_GETNEIGH) {
240 ndm->ndm_family = AF_INET;
6681712d 241 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
e4f67add
DS		 242 send_eth = !is_zero_ether_addr(fdb->eth_addr);
243 } else
244 ndm->ndm_family = AF_BRIDGE;
d342894c 245 ndm->ndm_state = fdb->state;
246 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 247 ndm->ndm_flags = fdb->flags;
d342894c 248 ndm->ndm_type = NDA_DST;
249
e4f67add 250 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 251 goto nla_put_failure;
252
6681712d
DS		 253 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
254 goto nla_put_failure;
255
823aa873 256 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 257 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
258 goto nla_put_failure;
c7995c43 259 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
60d9d4c6 260 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
6681712d
DS		 261 goto nla_put_failure;
262 if (rdst->remote_ifindex &&
263 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 264 goto nla_put_failure;
265
266 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
267 ci.ndm_confirmed = 0;
268 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
269 ci.ndm_refcnt = 0;
270
271 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
272 goto nla_put_failure;
273
274 return nlmsg_end(skb, nlh);
275
276nla_put_failure:
277 nlmsg_cancel(skb, nlh);
278 return -EMSGSIZE;
279}
280
281static inline size_t vxlan_nlmsg_size(void)
282{
283 return NLMSG_ALIGN(sizeof(struct ndmsg))
284 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
285 + nla_total_size(sizeof(__be32)) /* NDA_DST */
73cf3317 286 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 287 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
288 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 289 + nla_total_size(sizeof(struct nda_cacheinfo));
290}
291
292static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
3e61aa8f 293 struct vxlan_fdb *fdb, int type)
d342894c 294{
295 struct net *net = dev_net(vxlan->dev);
296 struct sk_buff *skb;
297 int err = -ENOBUFS;
298
299 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
300 if (skb == NULL)
301 goto errout;
302
5ca5461c 303 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0,
304 first_remote_rtnl(fdb));
d342894c 305 if (err < 0) {
306 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
307 WARN_ON(err == -EMSGSIZE);
308 kfree_skb(skb);
309 goto errout;
310 }
311
312 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
313 return;
314errout:
315 if (err < 0)
316 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
317}
318
e4f67add
DS		 319static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
320{
321 struct vxlan_dev *vxlan = netdev_priv(dev);
bb3fd687
SH		 322 struct vxlan_fdb f = {
323 .state = NUD_STALE,
324 };
325 struct vxlan_rdst remote = {
326 .remote_ip = ipa, /* goes to NDA_DST */
327 .remote_vni = VXLAN_N_VID,
328 };
3e61aa8f
SH		 329
330 INIT_LIST_HEAD(&f.remotes);
331 list_add_rcu(&remote.list, &f.remotes);
e4f67add
DS		 332
333 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
334}
335
336static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
337{
bb3fd687
SH		 338 struct vxlan_fdb f = {
339 .state = NUD_STALE,
340 };
e4f67add 341
3e61aa8f 342 INIT_LIST_HEAD(&f.remotes);
e4f67add
DS		 343 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
344
345 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
346}
347
d342894c 348/* Hash Ethernet address */
349static u32 eth_hash(const unsigned char *addr)
350{
351 u64 value = get_unaligned((u64 *)addr);
352
353 /* only want 6 bytes */
354#ifdef __BIG_ENDIAN
d342894c 355 value >>= 16;
321fb991 356#else
357 value <<= 16;
d342894c 358#endif
359 return hash_64(value, FDB_HASH_BITS);
360}
361
362/* Hash chain to use given mac address */
363static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
364 const u8 *mac)
365{
366 return &vxlan->fdb_head[eth_hash(mac)];
367}
368
369/* Look up Ethernet address in forwarding table */
014be2c8 370static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 371 const u8 *mac)
372
373{
374 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
375 struct vxlan_fdb *f;
d342894c 376
b67bfe0d 377 hlist_for_each_entry_rcu(f, head, hlist) {
d342894c 378 if (compare_ether_addr(mac, f->eth_addr) == 0)
379 return f;
380 }
381
382 return NULL;
383}
384
014be2c8
SS		 385static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
386 const u8 *mac)
387{
388 struct vxlan_fdb *f;
389
390 f = __vxlan_find_mac(vxlan, mac);
391 if (f)
392 f->used = jiffies;
393
394 return f;
395}
396
a5e7c10a
MR		 397/* caller should hold vxlan->hash_lock */
398static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
399 __be32 ip, __be16 port,
400 __u32 vni, __u32 ifindex)
6681712d 401{
3e61aa8f 402 struct vxlan_rdst *rd;
6681712d 403
3e61aa8f 404 list_for_each_entry(rd, &f->remotes, list) {
6681712d
DS		 405 if (rd->remote_ip == ip &&
406 rd->remote_port == port &&
407 rd->remote_vni == vni &&
408 rd->remote_ifindex == ifindex)
a5e7c10a 409 return rd;
6681712d 410 }
3e61aa8f 411
a5e7c10a
MR		 412 return NULL;
413}
414
906dc186
TR		 415/* Replace destination of unicast mac */
416static int vxlan_fdb_replace(struct vxlan_fdb *f,
417 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
418{
419 struct vxlan_rdst *rd;
420
421 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
422 if (rd)
423 return 0;
424
425 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
426 if (!rd)
427 return 0;
428 rd->remote_ip = ip;
429 rd->remote_port = port;
430 rd->remote_vni = vni;
431 rd->remote_ifindex = ifindex;
432 return 1;
433}
434
a5e7c10a
MR		 435/* Add/update destinations for multicast */
436static int vxlan_fdb_append(struct vxlan_fdb *f,
437 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
438{
439 struct vxlan_rdst *rd;
440
441 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
442 if (rd)
443 return 0;
444
6681712d
DS		 445 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
446 if (rd == NULL)
447 return -ENOBUFS;
448 rd->remote_ip = ip;
449 rd->remote_port = port;
450 rd->remote_vni = vni;
451 rd->remote_ifindex = ifindex;
3e61aa8f
SH		 452
453 list_add_tail_rcu(&rd->list, &f->remotes);
454
6681712d
DS		 455 return 1;
456}
457
d342894c 458/* Add new entry to forwarding table -- assumes lock held */
459static int vxlan_fdb_create(struct vxlan_dev *vxlan,
460 const u8 *mac, __be32 ip,
6681712d 461 __u16 state, __u16 flags,
73cf3317 462 __be16 port, __u32 vni, __u32 ifindex,
ae884082 463 __u8 ndm_flags)
d342894c 464{
465 struct vxlan_fdb *f;
466 int notify = 0;
467
014be2c8 468 f = __vxlan_find_mac(vxlan, mac);
d342894c 469 if (f) {
470 if (flags & NLM_F_EXCL) {
471 netdev_dbg(vxlan->dev,
472 "lost race to create %pM\n", mac);
473 return -EEXIST;
474 }
475 if (f->state != state) {
476 f->state = state;
477 f->updated = jiffies;
478 notify = 1;
479 }
ae884082
DS		 480 if (f->flags != ndm_flags) {
481 f->flags = ndm_flags;
482 f->updated = jiffies;
483 notify = 1;
484 }
906dc186
TR		 485 if ((flags & NLM_F_REPLACE)) {
486 /* Only change unicasts */
487 if (!(is_multicast_ether_addr(f->eth_addr) ||
488 is_zero_ether_addr(f->eth_addr))) {
489 int rc = vxlan_fdb_replace(f, ip, port, vni,
490 ifindex);
491
492 if (rc < 0)
493 return rc;
494 notify |= rc;
495 } else
496 return -EOPNOTSUPP;
497 }
6681712d 498 if ((flags & NLM_F_APPEND) &&
58e4c767
MR		 499 (is_multicast_ether_addr(f->eth_addr) ||
500 is_zero_ether_addr(f->eth_addr))) {
6681712d
DS		 501 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
502
503 if (rc < 0)
504 return rc;
505 notify |= rc;
506 }
d342894c 507 } else {
508 if (!(flags & NLM_F_CREATE))
509 return -ENOENT;
510
511 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
512 return -ENOSPC;
513
906dc186
TR		 514 /* Disallow replace to add a multicast entry */
515 if ((flags & NLM_F_REPLACE) &&
516 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
517 return -EOPNOTSUPP;
518
d342894c 519 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
520 f = kmalloc(sizeof(*f), GFP_ATOMIC);
521 if (!f)
522 return -ENOMEM;
523
524 notify = 1;
d342894c 525 f->state = state;
ae884082 526 f->flags = ndm_flags;
d342894c 527 f->updated = f->used = jiffies;
3e61aa8f 528 INIT_LIST_HEAD(&f->remotes);
d342894c 529 memcpy(f->eth_addr, mac, ETH_ALEN);
530
3e61aa8f
SH		 531 vxlan_fdb_append(f, ip, port, vni, ifindex);
532
d342894c 533 ++vxlan->addrcnt;
534 hlist_add_head_rcu(&f->hlist,
535 vxlan_fdb_head(vxlan, mac));
536 }
537
538 if (notify)
539 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
540
541 return 0;
542}
543
bc7892ba
MR		 544static void vxlan_fdb_free_rdst(struct rcu_head *head)
545{
546 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
547 kfree(rd);
548}
549
6706c82e 550static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 551{
552 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
3e61aa8f 553 struct vxlan_rdst *rd, *nd;
6681712d 554
3e61aa8f 555 list_for_each_entry_safe(rd, nd, &f->remotes, list)
6681712d 556 kfree(rd);
6681712d
DS		 557 kfree(f);
558}
559
d342894c 560static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
561{
562 netdev_dbg(vxlan->dev,
563 "delete %pM\n", f->eth_addr);
564
565 --vxlan->addrcnt;
566 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
567
568 hlist_del_rcu(&f->hlist);
6681712d 569 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 570}
571
f0b074be
MR		 572static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
573 __be32 *ip, __be16 *port, u32 *vni, u32 *ifindex)
d342894c 574{
6681712d 575 struct net *net = dev_net(vxlan->dev);
d342894c 576
f0b074be
MR		 577 if (tb[NDA_DST]) {
578 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
579 return -EAFNOSUPPORT;
d342894c 580
f0b074be
MR		 581 *ip = nla_get_be32(tb[NDA_DST]);
582 } else {
583 *ip = htonl(INADDR_ANY);
584 }
d342894c 585
6681712d 586 if (tb[NDA_PORT]) {
73cf3317 587 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 588 return -EINVAL;
f0b074be
MR		 589 *port = nla_get_be16(tb[NDA_PORT]);
590 } else {
591 *port = vxlan->dst_port;
592 }
6681712d
DS		 593
594 if (tb[NDA_VNI]) {
595 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
596 return -EINVAL;
f0b074be
MR		 597 *vni = nla_get_u32(tb[NDA_VNI]);
598 } else {
599 *vni = vxlan->default_dst.remote_vni;
600 }
6681712d
DS		 601
602 if (tb[NDA_IFINDEX]) {
5abb0029 603 struct net_device *tdev;
6681712d
DS		 604
605 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
606 return -EINVAL;
f0b074be
MR		 607 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
608 tdev = dev_get_by_index(net, *ifindex);
5abb0029 609 if (!tdev)
6681712d 610 return -EADDRNOTAVAIL;
5abb0029 611 dev_put(tdev);
f0b074be
MR		 612 } else {
613 *ifindex = 0;
614 }
615
616 return 0;
617}
618
619/* Add static entry (via netlink) */
620static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
621 struct net_device *dev,
622 const unsigned char *addr, u16 flags)
623{
624 struct vxlan_dev *vxlan = netdev_priv(dev);
625 /* struct net *net = dev_net(vxlan->dev); */
626 __be32 ip;
627 __be16 port;
628 u32 vni, ifindex;
629 int err;
630
631 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
632 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
633 ndm->ndm_state);
634 return -EINVAL;
635 }
636
637 if (tb[NDA_DST] == NULL)
638 return -EINVAL;
639
640 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
641 if (err)
642 return err;
6681712d 643
d342894c 644 spin_lock_bh(&vxlan->hash_lock);
73cf3317 645 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags,
646 port, vni, ifindex, ndm->ndm_flags);
d342894c 647 spin_unlock_bh(&vxlan->hash_lock);
648
649 return err;
650}
651
652/* Delete entry (via netlink) */
1690be63
VY		 653static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
654 struct net_device *dev,
d342894c 655 const unsigned char *addr)
656{
657 struct vxlan_dev *vxlan = netdev_priv(dev);
658 struct vxlan_fdb *f;
bc7892ba
MR		 659 struct vxlan_rdst *rd = NULL;
660 __be32 ip;
661 __be16 port;
662 u32 vni, ifindex;
663 int err;
664
665 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
666 if (err)
667 return err;
668
669 err = -ENOENT;
d342894c 670
671 spin_lock_bh(&vxlan->hash_lock);
672 f = vxlan_find_mac(vxlan, addr);
bc7892ba
MR		 673 if (!f)
674 goto out;
675
676 if (ip != htonl(INADDR_ANY)) {
677 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
678 if (!rd)
679 goto out;
680 }
681
682 err = 0;
683
684 /* remove a destination if it's not the only one on the list,
685 * otherwise destroy the fdb entry
686 */
687 if (rd && !list_is_singular(&f->remotes)) {
688 list_del_rcu(&rd->list);
689 call_rcu(&rd->rcu, vxlan_fdb_free_rdst);
690 goto out;
d342894c 691 }
bc7892ba
MR		 692
693 vxlan_fdb_destroy(vxlan, f);
694
695out:
d342894c 696 spin_unlock_bh(&vxlan->hash_lock);
697
698 return err;
699}
700
701/* Dump forwarding table */
702static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
703 struct net_device *dev, int idx)
704{
705 struct vxlan_dev *vxlan = netdev_priv(dev);
706 unsigned int h;
707
708 for (h = 0; h < FDB_HASH_SIZE; ++h) {
709 struct vxlan_fdb *f;
d342894c 710 int err;
711
b67bfe0d 712 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d 713 struct vxlan_rdst *rd;
6681712d 714
3e61aa8f
SH		 715 if (idx < cb->args[0])
716 goto skip;
717
718 list_for_each_entry_rcu(rd, &f->remotes, list) {
6681712d
DS		 719 err = vxlan_fdb_info(skb, vxlan, f,
720 NETLINK_CB(cb->skb).portid,
721 cb->nlh->nlmsg_seq,
722 RTM_NEWNEIGH,
723 NLM_F_MULTI, rd);
724 if (err < 0)
3e61aa8f 725 goto out;
6681712d 726 }
3e61aa8f
SH		 727skip:
728 ++idx;
d342894c 729 }
730 }
3e61aa8f 731out:
d342894c 732 return idx;
733}
734
735/* Watch incoming packets to learn mapping between Ethernet address
736 * and Tunnel endpoint.
26a41ae6 737 * Return true if packet is bogus and should be droppped.
d342894c 738 */
26a41ae6 739static bool vxlan_snoop(struct net_device *dev,
d342894c 740 __be32 src_ip, const u8 *src_mac)
741{
742 struct vxlan_dev *vxlan = netdev_priv(dev);
743 struct vxlan_fdb *f;
d342894c 744
745 f = vxlan_find_mac(vxlan, src_mac);
746 if (likely(f)) {
5ca5461c 747 struct vxlan_rdst *rdst = first_remote_rcu(f);
3e61aa8f
SH		 748
749 if (likely(rdst->remote_ip == src_ip))
26a41ae6 750 return false;
751
752 /* Don't migrate static entries, drop packets */
eb064c3b 753 if (f->state & NUD_NOARP)
26a41ae6 754 return true;
d342894c 755
756 if (net_ratelimit())
757 netdev_info(dev,
758 "%pM migrated from %pI4 to %pI4\n",
3e61aa8f 759 src_mac, &rdst->remote_ip, &src_ip);
d342894c 760
3e61aa8f 761 rdst->remote_ip = src_ip;
d342894c 762 f->updated = jiffies;
8385f50a 763 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
d342894c 764 } else {
765 /* learned new entry */
766 spin_lock(&vxlan->hash_lock);
3bf74b1a 767
768 /* close off race between vxlan_flush and incoming packets */
769 if (netif_running(dev))
770 vxlan_fdb_create(vxlan, src_mac, src_ip,
771 NUD_REACHABLE,
772 NLM_F_EXCL|NLM_F_CREATE,
773 vxlan->dst_port,
774 vxlan->default_dst.remote_vni,
775 0, NTF_SELF);
d342894c 776 spin_unlock(&vxlan->hash_lock);
777 }
26a41ae6 778
779 return false;
d342894c 780}
781
d342894c 782/* See if multicast group is already in use by other ID */
7c47cedf 783static bool vxlan_group_used(struct vxlan_net *vn, __be32 remote_ip)
d342894c 784{
553675fb 785 struct vxlan_dev *vxlan;
d342894c 786
553675fb 787 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
553675fb 788 if (!netif_running(vxlan->dev))
789 continue;
d342894c 790
7c47cedf 791 if (vxlan->default_dst.remote_ip == remote_ip)
553675fb 792 return true;
793 }
d342894c 794
795 return false;
796}
797
7c47cedf 798static void vxlan_sock_hold(struct vxlan_sock *vs)
d342894c 799{
7c47cedf
SH		 800 atomic_inc(&vs->refcnt);
801}
d342894c 802
012a5729 803void vxlan_sock_release(struct vxlan_sock *vs)
7c47cedf 804{
012a5729
PS		 805 struct vxlan_net *vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
806
7c47cedf
SH		 807 if (!atomic_dec_and_test(&vs->refcnt))
808 return;
d342894c 809
1c51a915 810 spin_lock(&vn->sock_lock);
7c47cedf 811 hlist_del_rcu(&vs->hlist);
1c51a915
SH		 812 spin_unlock(&vn->sock_lock);
813
7c47cedf 814 queue_work(vxlan_wq, &vs->del_work);
d342894c 815}
012a5729 816EXPORT_SYMBOL_GPL(vxlan_sock_release);
d342894c 817
3fc2de2f 818/* Callback to update multicast group membership when first VNI on
819 * multicast asddress is brought up
820 * Done as workqueue because ip_mc_join_group acquires RTNL.
7c47cedf 821 */
3fc2de2f 822static void vxlan_igmp_join(struct work_struct *work)
d342894c 823{
3fc2de2f 824 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join);
7c47cedf
SH		 825 struct vxlan_sock *vs = vxlan->vn_sock;
826 struct sock *sk = vs->sock->sk;
d342894c 827 struct ip_mreqn mreq = {
c7995c43
AW		 828 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
829 .imr_ifindex = vxlan->default_dst.remote_ifindex,
d342894c 830 };
831
d342894c 832 lock_sock(sk);
3fc2de2f 833 ip_mc_join_group(sk, &mreq);
834 release_sock(sk);
835
012a5729 836 vxlan_sock_release(vs);
3fc2de2f 837 dev_put(vxlan->dev);
838}
839
840/* Inverse of vxlan_igmp_join when last VNI is brought down */
841static void vxlan_igmp_leave(struct work_struct *work)
842{
843 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave);
3fc2de2f 844 struct vxlan_sock *vs = vxlan->vn_sock;
845 struct sock *sk = vs->sock->sk;
846 struct ip_mreqn mreq = {
847 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
848 .imr_ifindex = vxlan->default_dst.remote_ifindex,
849 };
850
851 lock_sock(sk);
852 ip_mc_leave_group(sk, &mreq);
d342894c 853 release_sock(sk);
d342894c 854
012a5729 855 vxlan_sock_release(vs);
7c47cedf 856 dev_put(vxlan->dev);
d342894c 857}
858
859/* Callback from net/ipv4/udp.c to receive packets */
860static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
861{
5cfccc5a 862 struct vxlan_sock *vs;
d342894c 863 struct vxlanhdr *vxh;
553675fb 864 __be16 port;
d342894c 865
d342894c 866 /* Need Vxlan and inner Ethernet header to be present */
7ce04758 867 if (!pskb_may_pull(skb, VXLAN_HLEN))
d342894c 868 goto error;
869
7ce04758
PS		 870 /* Return packets with reserved bits set */
871 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
d342894c 872 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
873 (vxh->vx_vni & htonl(0xff))) {
874 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
875 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
876 goto error;
877 }
878
5cfccc5a
PS		 879 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
880 goto drop;
881
553675fb 882 port = inet_sk(sk)->inet_sport;
5cfccc5a
PS		 883
884 vs = vxlan_find_sock(sock_net(sk), port);
885 if (!vs)
d342894c 886 goto drop;
d342894c 887
5cfccc5a
PS		 888 vs->rcv(vs, skb, vxh->vx_vni);
889 return 0;
890
891drop:
892 /* Consume bad packet */
893 kfree_skb(skb);
894 return 0;
895
896error:
897 /* Return non vxlan pkt */
898 return 1;
899}
900
901static void vxlan_rcv(struct vxlan_sock *vs,
902 struct sk_buff *skb, __be32 vx_vni)
903{
904 struct iphdr *oip;
905 struct vxlan_dev *vxlan;
906 struct pcpu_tstats *stats;
907 __u32 vni;
908 int err;
909
910 vni = ntohl(vx_vni) >> 8;
911 /* Is this VNI defined? */
912 vxlan = vxlan_vs_find_vni(vs, vni);
913 if (!vxlan)
d342894c 914 goto drop;
d342894c 915
e4f67add 916 skb_reset_mac_header(skb);
d342894c 917 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 918
919 /* Ignore packet loops (and multicast echo) */
920 if (compare_ether_addr(eth_hdr(skb)->h_source,
921 vxlan->dev->dev_addr) == 0)
922 goto drop;
923
7ce04758
PS		 924 /* Re-examine inner Ethernet packet */
925 oip = ip_hdr(skb);
26a41ae6 926 if ((vxlan->flags & VXLAN_F_LEARN) &&
927 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source))
928 goto drop;
d342894c 929
d342894c 930 skb_reset_network_header(skb);
0afb1666
JG		 931
932 /* If the NIC driver gave us an encapsulated packet with
933 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
934 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
935 * for us. Otherwise force the upper layers to verify it.
936 */
937 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
938 !(vxlan->dev->features & NETIF_F_RXCSUM))
939 skb->ip_summed = CHECKSUM_NONE;
940
941 skb->encapsulation = 0;
d342894c 942
943 err = IP_ECN_decapsulate(oip, skb);
944 if (unlikely(err)) {
945 if (log_ecn_error)
946 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
947 &oip->saddr, oip->tos);
948 if (err > 1) {
949 ++vxlan->dev->stats.rx_frame_errors;
950 ++vxlan->dev->stats.rx_errors;
951 goto drop;
952 }
953 }
954
e8171045 955 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 956 u64_stats_update_begin(&stats->syncp);
957 stats->rx_packets++;
958 stats->rx_bytes += skb->len;
959 u64_stats_update_end(&stats->syncp);
960
961 netif_rx(skb);
962
5cfccc5a 963 return;
d342894c 964drop:
965 /* Consume bad packet */
966 kfree_skb(skb);
d342894c 967}
968
e4f67add
DS		 969static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
970{
971 struct vxlan_dev *vxlan = netdev_priv(dev);
972 struct arphdr *parp;
973 u8 *arpptr, *sha;
974 __be32 sip, tip;
975 struct neighbour *n;
976
977 if (dev->flags & IFF_NOARP)
978 goto out;
979
980 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
981 dev->stats.tx_dropped++;
982 goto out;
983 }
984 parp = arp_hdr(skb);
985
986 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
987 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
988 parp->ar_pro != htons(ETH_P_IP) ||
989 parp->ar_op != htons(ARPOP_REQUEST) ||
990 parp->ar_hln != dev->addr_len ||
991 parp->ar_pln != 4)
992 goto out;
993 arpptr = (u8 *)parp + sizeof(struct arphdr);
994 sha = arpptr;
995 arpptr += dev->addr_len; /* sha */
996 memcpy(&sip, arpptr, sizeof(sip));
997 arpptr += sizeof(sip);
998 arpptr += dev->addr_len; /* tha */
999 memcpy(&tip, arpptr, sizeof(tip));
1000
1001 if (ipv4_is_loopback(tip) ||
1002 ipv4_is_multicast(tip))
1003 goto out;
1004
1005 n = neigh_lookup(&arp_tbl, &tip, dev);
1006
1007 if (n) {
e4f67add
DS		 1008 struct vxlan_fdb *f;
1009 struct sk_buff *reply;
1010
1011 if (!(n->nud_state & NUD_CONNECTED)) {
1012 neigh_release(n);
1013 goto out;
1014 }
1015
1016 f = vxlan_find_mac(vxlan, n->ha);
5ca5461c 1017 if (f && first_remote_rcu(f)->remote_ip == htonl(INADDR_ANY)) {
e4f67add
DS		 1018 /* bridge-local neighbor */
1019 neigh_release(n);
1020 goto out;
1021 }
1022
1023 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1024 n->ha, sha);
1025
1026 neigh_release(n);
1027
1028 skb_reset_mac_header(reply);
1029 __skb_pull(reply, skb_network_offset(reply));
1030 reply->ip_summed = CHECKSUM_UNNECESSARY;
1031 reply->pkt_type = PACKET_HOST;
1032
1033 if (netif_rx_ni(reply) == NET_RX_DROP)
1034 dev->stats.rx_dropped++;
1035 } else if (vxlan->flags & VXLAN_F_L3MISS)
1036 vxlan_ip_miss(dev, tip);
1037out:
1038 consume_skb(skb);
1039 return NETDEV_TX_OK;
1040}
1041
1042static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1043{
1044 struct vxlan_dev *vxlan = netdev_priv(dev);
1045 struct neighbour *n;
1046 struct iphdr *pip;
1047
1048 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1049 return false;
1050
1051 n = NULL;
1052 switch (ntohs(eth_hdr(skb)->h_proto)) {
1053 case ETH_P_IP:
1054 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1055 return false;
1056 pip = ip_hdr(skb);
1057 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1058 break;
1059 default:
1060 return false;
1061 }
1062
1063 if (n) {
1064 bool diff;
1065
1066 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
1067 if (diff) {
1068 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1069 dev->addr_len);
1070 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1071 }
1072 neigh_release(n);
1073 return diff;
1074 } else if (vxlan->flags & VXLAN_F_L3MISS)
1075 vxlan_ip_miss(dev, pip->daddr);
1076 return false;
1077}
1078
553675fb 1079static void vxlan_sock_put(struct sk_buff *skb)
1cad8715 1080{
1081 sock_put(skb->sk);
1082}
1083
1084/* On transmit, associate with the tunnel socket */
1085static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
1086{
553675fb 1087 struct vxlan_dev *vxlan = netdev_priv(dev);
1088 struct sock *sk = vxlan->vn_sock->sock->sk;
1cad8715 1089
1090 skb_orphan(skb);
1091 sock_hold(sk);
1092 skb->sk = sk;
553675fb 1093 skb->destructor = vxlan_sock_put;
1cad8715 1094}
1095
05f47d69 1096/* Compute source port for outgoing packet
1097 * first choice to use L4 flow hash since it will spread
1098 * better and maybe available from hardware
1099 * secondary choice is to use jhash on the Ethernet header
1100 */
7d836a76 1101static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
05f47d69 1102{
1103 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
1104 u32 hash;
1105
1106 hash = skb_get_rxhash(skb);
1107 if (!hash)
1108 hash = jhash(skb->data, 2 * ETH_ALEN,
1109 (__force u32) skb->protocol);
1110
7d836a76 1111 return htons((((u64) hash * range) >> 32) + vxlan->port_min);
05f47d69 1112}
1113
05c0db08
PS		 1114static int handle_offloads(struct sk_buff *skb)
1115{
1116 if (skb_is_gso(skb)) {
1117 int err = skb_unclone(skb, GFP_ATOMIC);
1118 if (unlikely(err))
1119 return err;
1120
f6ace502 1121 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 1122 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
1123 skb->ip_summed = CHECKSUM_NONE;
1124
1125 return 0;
1126}
1127
9dcc71e1
SS		 1128/* Bypass encapsulation if the destination is local */
1129static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1130 struct vxlan_dev *dst_vxlan)
1131{
1132 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1133 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
1134
1135 skb->pkt_type = PACKET_HOST;
1136 skb->encapsulation = 0;
1137 skb->dev = dst_vxlan->dev;
1138 __skb_pull(skb, skb_network_offset(skb));
1139
1140 if (dst_vxlan->flags & VXLAN_F_LEARN)
9d9f163c
MR		 1141 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
1142 eth_hdr(skb)->h_source);
9dcc71e1
SS		 1143
1144 u64_stats_update_begin(&tx_stats->syncp);
1145 tx_stats->tx_packets++;
1146 tx_stats->tx_bytes += skb->len;
1147 u64_stats_update_end(&tx_stats->syncp);
1148
1149 if (netif_rx(skb) == NET_RX_SUCCESS) {
1150 u64_stats_update_begin(&rx_stats->syncp);
1151 rx_stats->rx_packets++;
1152 rx_stats->rx_bytes += skb->len;
1153 u64_stats_update_end(&rx_stats->syncp);
1154 } else {
1155 skb->dev->stats.rx_dropped++;
1156 }
1157}
1158
4ad16930
SH		 1159static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1160 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 1161{
1162 struct vxlan_dev *vxlan = netdev_priv(dev);
1163 struct rtable *rt;
d342894c 1164 const struct iphdr *old_iph;
d342894c 1165 struct vxlanhdr *vxh;
1166 struct udphdr *uh;
1167 struct flowi4 fl4;
d342894c 1168 __be32 dst;
7d836a76 1169 __be16 src_port, dst_port;
234f5b73 1170 u32 vni;
d342894c 1171 __be16 df = 0;
1172 __u8 tos, ttl;
0e6fbc5b 1173 int err;
d342894c 1174
823aa873 1175 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d
DS		 1176 vni = rdst->remote_vni;
1177 dst = rdst->remote_ip;
e4f67add
DS		 1178
1179 if (!dst) {
1180 if (did_rsc) {
e4f67add 1181 /* short-circuited back to local bridge */
9dcc71e1 1182 vxlan_encap_bypass(skb, vxlan, vxlan);
4ad16930 1183 return;
e4f67add 1184 }
ef59febe 1185 goto drop;
e4f67add 1186 }
ef59febe 1187
d6727fe3
JG		 1188 if (!skb->encapsulation) {
1189 skb_reset_inner_headers(skb);
1190 skb->encapsulation = 1;
1191 }
1192
d342894c 1193 /* Need space for new headers (invalidates iph ptr) */
1194 if (skb_cow_head(skb, VXLAN_HEADROOM))
1195 goto drop;
1196
d342894c 1197 old_iph = ip_hdr(skb);
1198
d342894c 1199 ttl = vxlan->ttl;
1200 if (!ttl && IN_MULTICAST(ntohl(dst)))
1201 ttl = 1;
1202
1203 tos = vxlan->tos;
1204 if (tos == 1)
206aaafc 1205 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1206
05f47d69 1207 src_port = vxlan_src_port(vxlan, skb);
d342894c 1208
ca78f181 1209 memset(&fl4, 0, sizeof(fl4));
6681712d 1210 fl4.flowi4_oif = rdst->remote_ifindex;
ca78f181 1211 fl4.flowi4_tos = RT_TOS(tos);
1212 fl4.daddr = dst;
1213 fl4.saddr = vxlan->saddr;
1214
1215 rt = ip_route_output_key(dev_net(dev), &fl4);
d342894c 1216 if (IS_ERR(rt)) {
1217 netdev_dbg(dev, "no route to %pI4\n", &dst);
1218 dev->stats.tx_carrier_errors++;
1219 goto tx_error;
1220 }
1221
1222 if (rt->dst.dev == dev) {
1223 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1224 ip_rt_put(rt);
1225 dev->stats.collisions++;
1226 goto tx_error;
1227 }
1228
9dcc71e1 1229 /* Bypass encapsulation if the destination is local */
ab09a6d0
MR		 1230 if (rt->rt_flags & RTCF_LOCAL &&
1231 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
9dcc71e1
SS		 1232 struct vxlan_dev *dst_vxlan;
1233
1234 ip_rt_put(rt);
553675fb 1235 dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port);
9dcc71e1
SS		 1236 if (!dst_vxlan)
1237 goto tx_error;
1238 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
4ad16930 1239 return;
9dcc71e1 1240 }
d342894c 1241 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1242 vxh->vx_flags = htonl(VXLAN_FLAGS);
6681712d 1243 vxh->vx_vni = htonl(vni << 8);
d342894c 1244
1245 __skb_push(skb, sizeof(*uh));
1246 skb_reset_transport_header(skb);
1247 uh = udp_hdr(skb);
1248
73cf3317 1249 uh->dest = dst_port;
7d836a76 1250 uh->source = src_port;
d342894c 1251
1252 uh->len = htons(skb->len);
1253 uh->check = 0;
1254
1cad8715 1255 vxlan_set_owner(dev, skb);
1256
05c0db08
PS		 1257 if (handle_offloads(skb))
1258 goto drop;
d342894c 1259
0e6fbc5b
PS		 1260 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1261 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
1262
1263 err = iptunnel_xmit(dev_net(dev), rt, skb, fl4.saddr, dst,
1264 IPPROTO_UDP, tos, ttl, df);
1265 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
1266
4ad16930 1267 return;
d342894c 1268
1269drop:
1270 dev->stats.tx_dropped++;
1271 goto tx_free;
1272
1273tx_error:
1274 dev->stats.tx_errors++;
1275tx_free:
1276 dev_kfree_skb(skb);
d342894c 1277}
1278
6681712d
DS		 1279/* Transmit local packets over Vxlan
1280 *
1281 * Outer IP header inherits ECN and DF from inner header.
1282 * Outer UDP destination is the VXLAN assigned port.
1283 * source port is based on hash of flow
1284 */
1285static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1286{
1287 struct vxlan_dev *vxlan = netdev_priv(dev);
1288 struct ethhdr *eth;
1289 bool did_rsc = false;
afbd8bae 1290 struct vxlan_rdst *rdst;
6681712d 1291 struct vxlan_fdb *f;
6681712d
DS		 1292
1293 skb_reset_mac_header(skb);
1294 eth = eth_hdr(skb);
1295
1296 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1297 return arp_reduce(dev, skb);
6681712d
DS		 1298
1299 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1300 did_rsc = false;
1301
1302 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1303 ntohs(eth->h_proto) == ETH_P_IP) {
1304 did_rsc = route_shortcircuit(dev, skb);
1305 if (did_rsc)
1306 f = vxlan_find_mac(vxlan, eth->h_dest);
1307 }
1308
6681712d 1309 if (f == NULL) {
afbd8bae
MR		 1310 f = vxlan_find_mac(vxlan, all_zeros_mac);
1311 if (f == NULL) {
1312 if ((vxlan->flags & VXLAN_F_L2MISS) &&
1313 !is_multicast_ether_addr(eth->h_dest))
1314 vxlan_fdb_miss(vxlan, eth->h_dest);
1315
1316 dev->stats.tx_dropped++;
1317 dev_kfree_skb(skb);
1318 return NETDEV_TX_OK;
1319 }
1320 }
6681712d 1321
afbd8bae
MR		 1322 list_for_each_entry_rcu(rdst, &f->remotes, list) {
1323 struct sk_buff *skb1;
6681712d 1324
afbd8bae
MR		 1325 skb1 = skb_clone(skb, GFP_ATOMIC);
1326 if (skb1)
1327 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
6681712d
DS		 1328 }
1329
afbd8bae 1330 dev_kfree_skb(skb);
4ad16930 1331 return NETDEV_TX_OK;
6681712d
DS		 1332}
1333
d342894c 1334/* Walk the forwarding table and purge stale entries */
1335static void vxlan_cleanup(unsigned long arg)
1336{
1337 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1338 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1339 unsigned int h;
1340
1341 if (!netif_running(vxlan->dev))
1342 return;
1343
1344 spin_lock_bh(&vxlan->hash_lock);
1345 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1346 struct hlist_node *p, *n;
1347 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1348 struct vxlan_fdb *f
1349 = container_of(p, struct vxlan_fdb, hlist);
1350 unsigned long timeout;
1351
3c172868 1352 if (f->state & NUD_PERMANENT)
d342894c 1353 continue;
1354
1355 timeout = f->used + vxlan->age_interval * HZ;
1356 if (time_before_eq(timeout, jiffies)) {
1357 netdev_dbg(vxlan->dev,
1358 "garbage collect %pM\n",
1359 f->eth_addr);
1360 f->state = NUD_STALE;
1361 vxlan_fdb_destroy(vxlan, f);
1362 } else if (time_before(timeout, next_timer))
1363 next_timer = timeout;
1364 }
1365 }
1366 spin_unlock_bh(&vxlan->hash_lock);
1367
1368 mod_timer(&vxlan->age_timer, next_timer);
1369}
1370
9c2e24e1
PS		 1371static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1372{
1373 __u32 vni = vxlan->default_dst.remote_vni;
1374
1375 vxlan->vn_sock = vs;
1376 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1377}
1378
d342894c 1379/* Setup stats when device is created */
1380static int vxlan_init(struct net_device *dev)
1381{
1c51a915
SH		 1382 struct vxlan_dev *vxlan = netdev_priv(dev);
1383 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
1384 struct vxlan_sock *vs;
1c51a915 1385
e8171045
PS		 1386 dev->tstats = alloc_percpu(struct pcpu_tstats);
1387 if (!dev->tstats)
d342894c 1388 return -ENOMEM;
1389
1c51a915 1390 spin_lock(&vn->sock_lock);
9c2e24e1 1391 vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port);
1c51a915
SH		 1392 if (vs) {
1393 /* If we have a socket with same port already, reuse it */
1394 atomic_inc(&vs->refcnt);
9c2e24e1 1395 vxlan_vs_add_dev(vs, vxlan);
1c51a915
SH		 1396 } else {
1397 /* otherwise make new socket outside of RTNL */
1398 dev_hold(dev);
1399 queue_work(vxlan_wq, &vxlan->sock_work);
1400 }
1401 spin_unlock(&vn->sock_lock);
1402
d342894c 1403 return 0;
1404}
1405
ba609e9b 1406static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
afbd8bae
MR		 1407{
1408 struct vxlan_fdb *f;
1409
1410 spin_lock_bh(&vxlan->hash_lock);
1411 f = __vxlan_find_mac(vxlan, all_zeros_mac);
1412 if (f)
1413 vxlan_fdb_destroy(vxlan, f);
1414 spin_unlock_bh(&vxlan->hash_lock);
1415}
1416
ebf4063e
SH		 1417static void vxlan_uninit(struct net_device *dev)
1418{
1419 struct vxlan_dev *vxlan = netdev_priv(dev);
ebf4063e
SH		 1420 struct vxlan_sock *vs = vxlan->vn_sock;
1421
ba609e9b 1422 vxlan_fdb_delete_default(vxlan);
afbd8bae 1423
ebf4063e 1424 if (vs)
012a5729 1425 vxlan_sock_release(vs);
ebf4063e
SH		 1426 free_percpu(dev->tstats);
1427}
1428
d342894c 1429/* Start ageing timer and join group when device is brought up */
1430static int vxlan_open(struct net_device *dev)
1431{
3fc2de2f 1432 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1433 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915
SH		 1434 struct vxlan_sock *vs = vxlan->vn_sock;
1435
1436 /* socket hasn't been created */
1437 if (!vs)
1438 return -ENOTCONN;
d342894c 1439
3fc2de2f 1440 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)) &&
614334df 1441 vxlan_group_used(vn, vxlan->default_dst.remote_ip)) {
1c51a915 1442 vxlan_sock_hold(vs);
7c47cedf 1443 dev_hold(dev);
3fc2de2f 1444 queue_work(vxlan_wq, &vxlan->igmp_join);
d342894c 1445 }
1446
1447 if (vxlan->age_interval)
1448 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1449
1450 return 0;
1451}
1452
1453/* Purge the forwarding table */
1454static void vxlan_flush(struct vxlan_dev *vxlan)
1455{
31fec5aa 1456 unsigned int h;
d342894c 1457
1458 spin_lock_bh(&vxlan->hash_lock);
1459 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1460 struct hlist_node *p, *n;
1461 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1462 struct vxlan_fdb *f
1463 = container_of(p, struct vxlan_fdb, hlist);
afbd8bae
MR		 1464 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1465 if (!is_zero_ether_addr(f->eth_addr))
1466 vxlan_fdb_destroy(vxlan, f);
d342894c 1467 }
1468 }
1469 spin_unlock_bh(&vxlan->hash_lock);
1470}
1471
1472/* Cleanup timer and forwarding table on shutdown */
1473static int vxlan_stop(struct net_device *dev)
1474{
3fc2de2f 1475 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1476 struct vxlan_dev *vxlan = netdev_priv(dev);
1c51a915 1477 struct vxlan_sock *vs = vxlan->vn_sock;
d342894c 1478
3fc2de2f 1479 if (vs && IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)) &&
1480 ! vxlan_group_used(vn, vxlan->default_dst.remote_ip)) {
1c51a915 1481 vxlan_sock_hold(vs);
7c47cedf 1482 dev_hold(dev);
3fc2de2f 1483 queue_work(vxlan_wq, &vxlan->igmp_leave);
7c47cedf 1484 }
d342894c 1485
1486 del_timer_sync(&vxlan->age_timer);
1487
1488 vxlan_flush(vxlan);
1489
1490 return 0;
1491}
1492
d342894c 1493/* Stub, nothing needs to be done. */
1494static void vxlan_set_multicast_list(struct net_device *dev)
1495{
1496}
1497
1498static const struct net_device_ops vxlan_netdev_ops = {
1499 .ndo_init = vxlan_init,
ebf4063e 1500 .ndo_uninit = vxlan_uninit,
d342894c 1501 .ndo_open = vxlan_open,
1502 .ndo_stop = vxlan_stop,
1503 .ndo_start_xmit = vxlan_xmit,
e8171045 1504 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 1505 .ndo_set_rx_mode = vxlan_set_multicast_list,
1506 .ndo_change_mtu = eth_change_mtu,
1507 .ndo_validate_addr = eth_validate_addr,
1508 .ndo_set_mac_address = eth_mac_addr,
1509 .ndo_fdb_add = vxlan_fdb_add,
1510 .ndo_fdb_del = vxlan_fdb_delete,
1511 .ndo_fdb_dump = vxlan_fdb_dump,
1512};
1513
1514/* Info for udev, that this is a virtual tunnel endpoint */
1515static struct device_type vxlan_type = {
1516 .name = "vxlan",
1517};
1518
d342894c 1519/* Initialize the device structure. */
1520static void vxlan_setup(struct net_device *dev)
1521{
1522 struct vxlan_dev *vxlan = netdev_priv(dev);
31fec5aa 1523 unsigned int h;
05f47d69 1524 int low, high;
d342894c 1525
1526 eth_hw_addr_random(dev);
1527 ether_setup(dev);
2840bf22 1528 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 1529
1530 dev->netdev_ops = &vxlan_netdev_ops;
ebf4063e 1531 dev->destructor = free_netdev;
d342894c 1532 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1533
1534 dev->tx_queue_len = 0;
1535 dev->features |= NETIF_F_LLTX;
1536 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 1537 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 1538 dev->features |= NETIF_F_RXCSUM;
05c0db08 1539 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666
JG		 1540
1541 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 1542 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
d342894c 1543 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 1544 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 1545
553675fb 1546 INIT_LIST_HEAD(&vxlan->next);
d342894c 1547 spin_lock_init(&vxlan->hash_lock);
3fc2de2f 1548 INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join);
1549 INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave);
1c51a915 1550 INIT_WORK(&vxlan->sock_work, vxlan_sock_work);
d342894c 1551
1552 init_timer_deferrable(&vxlan->age_timer);
1553 vxlan->age_timer.function = vxlan_cleanup;
1554 vxlan->age_timer.data = (unsigned long) vxlan;
1555
05f47d69 1556 inet_get_local_port_range(&low, &high);
1557 vxlan->port_min = low;
1558 vxlan->port_max = high;
823aa873 1559 vxlan->dst_port = htons(vxlan_port);
05f47d69 1560
d342894c 1561 vxlan->dev = dev;
1562
1563 for (h = 0; h < FDB_HASH_SIZE; ++h)
1564 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1565}
1566
1567static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1568 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 1569 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
d342894c 1570 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1571 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1572 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1573 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1574 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1575 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1576 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 1577 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 1578 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1579 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1580 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1581 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 1582 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 1583};
1584
1585static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1586{
1587 if (tb[IFLA_ADDRESS]) {
1588 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1589 pr_debug("invalid link address (not ethernet)\n");
1590 return -EINVAL;
1591 }
1592
1593 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1594 pr_debug("invalid all zero ethernet address\n");
1595 return -EADDRNOTAVAIL;
1596 }
1597 }
1598
1599 if (!data)
1600 return -EINVAL;
1601
1602 if (data[IFLA_VXLAN_ID]) {
1603 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1604 if (id >= VXLAN_VID_MASK)
1605 return -ERANGE;
1606 }
1607
05f47d69 1608 if (data[IFLA_VXLAN_PORT_RANGE]) {
1609 const struct ifla_vxlan_port_range *p
1610 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1611
1612 if (ntohs(p->high) < ntohs(p->low)) {
1613 pr_debug("port range %u .. %u not valid\n",
1614 ntohs(p->low), ntohs(p->high));
1615 return -EINVAL;
1616 }
1617 }
1618
d342894c 1619 return 0;
1620}
1621
1b13c97f
YB		 1622static void vxlan_get_drvinfo(struct net_device *netdev,
1623 struct ethtool_drvinfo *drvinfo)
1624{
1625 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1626 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1627}
1628
1629static const struct ethtool_ops vxlan_ethtool_ops = {
1630 .get_drvinfo = vxlan_get_drvinfo,
1631 .get_link = ethtool_op_get_link,
1632};
1633
553675fb 1634static void vxlan_del_work(struct work_struct *work)
1635{
1636 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
1637
1638 sk_release_kernel(vs->sock->sk);
1639 kfree_rcu(vs, rcu);
1640}
1641
5cfccc5a 1642static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
012a5729 1643 vxlan_rcv_t *rcv, void *data)
553675fb 1644{
9c2e24e1 1645 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
553675fb 1646 struct vxlan_sock *vs;
1647 struct sock *sk;
1648 struct sockaddr_in vxlan_addr = {
1649 .sin_family = AF_INET,
1650 .sin_addr.s_addr = htonl(INADDR_ANY),
bb3fd687 1651 .sin_port = port,
553675fb 1652 };
1653 int rc;
31fec5aa 1654 unsigned int h;
553675fb 1655
1656 vs = kmalloc(sizeof(*vs), GFP_KERNEL);
9c2e24e1
PS		 1657 if (!vs) {
1658 pr_debug("memory alocation failure\n");
553675fb 1659 return ERR_PTR(-ENOMEM);
9c2e24e1 1660 }
553675fb 1661
1662 for (h = 0; h < VNI_HASH_SIZE; ++h)
1663 INIT_HLIST_HEAD(&vs->vni_list[h]);
1664
1665 INIT_WORK(&vs->del_work, vxlan_del_work);
1666
1667 /* Create UDP socket for encapsulation receive. */
1668 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vs->sock);
1669 if (rc < 0) {
1670 pr_debug("UDP socket create failed\n");
1671 kfree(vs);
1672 return ERR_PTR(rc);
1673 }
1674
1675 /* Put in proper namespace */
1676 sk = vs->sock->sk;
1677 sk_change_net(sk, net);
1678
553675fb 1679 rc = kernel_bind(vs->sock, (struct sockaddr *) &vxlan_addr,
1680 sizeof(vxlan_addr));
1681 if (rc < 0) {
1682 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1683 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
1684 sk_release_kernel(sk);
1685 kfree(vs);
1686 return ERR_PTR(rc);
1687 }
9c2e24e1 1688 atomic_set(&vs->refcnt, 1);
5cfccc5a 1689 vs->rcv = rcv;
012a5729 1690 vs->data = data;
553675fb 1691
1692 /* Disable multicast loopback */
1693 inet_sk(sk)->mc_loop = 0;
9c2e24e1
PS		 1694 spin_lock(&vn->sock_lock);
1695 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
1696 spin_unlock(&vn->sock_lock);
553675fb 1697
1698 /* Mark socket as an encapsulation socket. */
1699 udp_sk(sk)->encap_type = 1;
1700 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1701 udp_encap_enable();
9c2e24e1
PS		 1702 return vs;
1703}
1704
012a5729
PS		 1705struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
1706 vxlan_rcv_t *rcv, void *data,
1707 bool no_share)
9c2e24e1
PS		 1708{
1709 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1710 struct vxlan_sock *vs;
1711
012a5729 1712 vs = vxlan_socket_create(net, port, rcv, data);
9c2e24e1
PS		 1713 if (!IS_ERR(vs))
1714 return vs;
553675fb 1715
012a5729
PS		 1716 if (no_share) /* Return error if sharing is not allowed. */
1717 return vs;
1718
9c2e24e1
PS		 1719 spin_lock(&vn->sock_lock);
1720 vs = vxlan_find_sock(net, port);
5cfccc5a
PS		 1721 if (vs) {
1722 if (vs->rcv == rcv)
1723 atomic_inc(&vs->refcnt);
1724 else
1725 vs = ERR_PTR(-EBUSY);
1726 }
1727 spin_unlock(&vn->sock_lock);
1728
1729 if (!vs)
9c2e24e1
PS		 1730 vs = ERR_PTR(-EINVAL);
1731
553675fb 1732 return vs;
1733}
012a5729 1734EXPORT_SYMBOL_GPL(vxlan_sock_add);
553675fb 1735
1c51a915
SH		 1736/* Scheduled at device creation to bind to a socket */
1737static void vxlan_sock_work(struct work_struct *work)
1738{
9c2e24e1
PS		 1739 struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work);
1740 struct net *net = dev_net(vxlan->dev);
1c51a915 1741 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9c2e24e1
PS		 1742 __be16 port = vxlan->dst_port;
1743 struct vxlan_sock *nvs;
1c51a915 1744
012a5729 1745 nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false);
1c51a915 1746 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 1747 if (!IS_ERR(nvs))
1748 vxlan_vs_add_dev(nvs, vxlan);
1749 spin_unlock(&vn->sock_lock);
1750
1751 dev_put(vxlan->dev);
1c51a915
SH		 1752}
1753
d342894c 1754static int vxlan_newlink(struct net *net, struct net_device *dev,
1755 struct nlattr *tb[], struct nlattr *data[])
1756{
553675fb 1757 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
d342894c 1758 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1759 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 1760 __u32 vni;
1761 int err;
1762
1763 if (!data[IFLA_VXLAN_ID])
1764 return -EINVAL;
1765
1766 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
c7995c43 1767 dst->remote_vni = vni;
d342894c 1768
5d174dd8 1769 if (data[IFLA_VXLAN_GROUP])
1770 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]);
d342894c 1771
1772 if (data[IFLA_VXLAN_LOCAL])
1773 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1774
34e02aa1 1775 if (data[IFLA_VXLAN_LINK] &&
c7995c43 1776 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 1777 struct net_device *lowerdev
c7995c43 1778 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 1779
1780 if (!lowerdev) {
c7995c43 1781 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 1782 return -ENODEV;
1783 }
d342894c 1784
34e02aa1 1785 if (!tb[IFLA_MTU])
d342894c 1786 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1ba56fb4
AD		 1787
1788 /* update header length based on lower device */
1789 dev->hard_header_len = lowerdev->hard_header_len +
1790 VXLAN_HEADROOM;
d342894c 1791 }
1792
1793 if (data[IFLA_VXLAN_TOS])
1794 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1795
afb97186
VB		 1796 if (data[IFLA_VXLAN_TTL])
1797 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1798
d342894c 1799 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 1800 vxlan->flags |= VXLAN_F_LEARN;
d342894c 1801
1802 if (data[IFLA_VXLAN_AGEING])
1803 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1804 else
1805 vxlan->age_interval = FDB_AGE_DEFAULT;
1806
e4f67add
DS		 1807 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1808 vxlan->flags |= VXLAN_F_PROXY;
1809
1810 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1811 vxlan->flags |= VXLAN_F_RSC;
1812
1813 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1814 vxlan->flags |= VXLAN_F_L2MISS;
1815
1816 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1817 vxlan->flags |= VXLAN_F_L3MISS;
1818
d342894c 1819 if (data[IFLA_VXLAN_LIMIT])
1820 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1821
05f47d69 1822 if (data[IFLA_VXLAN_PORT_RANGE]) {
1823 const struct ifla_vxlan_port_range *p
1824 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1825 vxlan->port_min = ntohs(p->low);
1826 vxlan->port_max = ntohs(p->high);
1827 }
1828
823aa873 1829 if (data[IFLA_VXLAN_PORT])
1830 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
1831
553675fb 1832 if (vxlan_find_vni(net, vni, vxlan->dst_port)) {
1833 pr_info("duplicate VNI %u\n", vni);
1834 return -EEXIST;
1835 }
1836
1b13c97f
YB		 1837 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1838
afbd8bae
MR		 1839 /* create an fdb entry for default destination */
1840 err = vxlan_fdb_create(vxlan, all_zeros_mac,
1841 vxlan->default_dst.remote_ip,
1842 NUD_REACHABLE|NUD_PERMANENT,
1843 NLM_F_EXCL|NLM_F_CREATE,
1844 vxlan->dst_port, vxlan->default_dst.remote_vni,
1845 vxlan->default_dst.remote_ifindex, NTF_SELF);
1c51a915 1846 if (err)
553675fb 1847 return err;
d342894c 1848
afbd8bae
MR		 1849 err = register_netdevice(dev);
1850 if (err) {
ba609e9b 1851 vxlan_fdb_delete_default(vxlan);
afbd8bae
MR		 1852 return err;
1853 }
1854
553675fb 1855 list_add(&vxlan->next, &vn->vxlan_list);
553675fb 1856
1857 return 0;
d342894c 1858}
1859
1860static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1861{
fe5c3561 1862 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
d342894c 1863 struct vxlan_dev *vxlan = netdev_priv(dev);
1864
fe5c3561 1865 spin_lock(&vn->sock_lock);
9c2e24e1
PS		 1866 if (!hlist_unhashed(&vxlan->hlist))
1867 hlist_del_rcu(&vxlan->hlist);
fe5c3561 1868 spin_unlock(&vn->sock_lock);
1869
553675fb 1870 list_del(&vxlan->next);
d342894c 1871 unregister_netdevice_queue(dev, head);
1872}
1873
1874static size_t vxlan_get_size(const struct net_device *dev)
1875{
1876
1877 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
5d174dd8 1878 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */
d342894c 1879 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1880 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1881 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1882 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1883 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 1884 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1885 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1886 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1887 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 1888 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1889 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 1890 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 1891 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 1892 0;
1893}
1894
1895static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1896{
1897 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1898 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 1899 struct ifla_vxlan_port_range ports = {
1900 .low = htons(vxlan->port_min),
1901 .high = htons(vxlan->port_max),
1902 };
d342894c 1903
c7995c43 1904 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 1905 goto nla_put_failure;
1906
5d174dd8 1907 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip))
d342894c 1908 goto nla_put_failure;
1909
c7995c43 1910 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 1911 goto nla_put_failure;
1912
7c41c42c 1913 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
d342894c 1914 goto nla_put_failure;
1915
1916 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1917 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 1918 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1919 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1920 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1921 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1922 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1923 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1924 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1925 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1926 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 1927 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 1928 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
1929 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 1930 goto nla_put_failure;
1931
05f47d69 1932 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1933 goto nla_put_failure;
1934
d342894c 1935 return 0;
1936
1937nla_put_failure:
1938 return -EMSGSIZE;
1939}
1940
1941static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1942 .kind = "vxlan",
1943 .maxtype = IFLA_VXLAN_MAX,
1944 .policy = vxlan_policy,
1945 .priv_size = sizeof(struct vxlan_dev),
1946 .setup = vxlan_setup,
1947 .validate = vxlan_validate,
1948 .newlink = vxlan_newlink,
1949 .dellink = vxlan_dellink,
1950 .get_size = vxlan_get_size,
1951 .fill_info = vxlan_fill_info,
1952};
1953
1954static __net_init int vxlan_init_net(struct net *net)
1955{
1956 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
31fec5aa 1957 unsigned int h;
d342894c 1958
553675fb 1959 INIT_LIST_HEAD(&vn->vxlan_list);
1c51a915 1960 spin_lock_init(&vn->sock_lock);
d342894c 1961
553675fb 1962 for (h = 0; h < PORT_HASH_SIZE; ++h)
1963 INIT_HLIST_HEAD(&vn->sock_list[h]);
d342894c 1964
1965 return 0;
1966}
1967
1968static __net_exit void vxlan_exit_net(struct net *net)
1969{
1970 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e 1971 struct vxlan_dev *vxlan;
372675a4 1972 LIST_HEAD(list);
9cb6cb7e
ZM		 1973
1974 rtnl_lock();
553675fb 1975 list_for_each_entry(vxlan, &vn->vxlan_list, next)
372675a4 1976 unregister_netdevice_queue(vxlan->dev, &list);
1977 unregister_netdevice_many(&list);
9cb6cb7e 1978 rtnl_unlock();
d342894c 1979}
1980
1981static struct pernet_operations vxlan_net_ops = {
1982 .init = vxlan_init_net,
1983 .exit = vxlan_exit_net,
1984 .id = &vxlan_net_id,
1985 .size = sizeof(struct vxlan_net),
1986};
1987
1988static int __init vxlan_init_module(void)
1989{
1990 int rc;
1991
758c57d1
SH		 1992 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
1993 if (!vxlan_wq)
1994 return -ENOMEM;
1995
d342894c 1996 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1997
1998 rc = register_pernet_device(&vxlan_net_ops);
1999 if (rc)
2000 goto out1;
2001
2002 rc = rtnl_link_register(&vxlan_link_ops);
2003 if (rc)
2004 goto out2;
2005
2006 return 0;
2007
2008out2:
2009 unregister_pernet_device(&vxlan_net_ops);
2010out1:
758c57d1 2011 destroy_workqueue(vxlan_wq);
d342894c 2012 return rc;
2013}
7332a13b 2014late_initcall(vxlan_init_module);
d342894c 2015
2016static void __exit vxlan_cleanup_module(void)
2017{
b7153984 2018 rtnl_link_unregister(&vxlan_link_ops);
758c57d1 2019 destroy_workqueue(vxlan_wq);
f89e57c4 2020 unregister_pernet_device(&vxlan_net_ops);
6681712d 2021 rcu_barrier();
d342894c 2022}
2023module_exit(vxlan_cleanup_module);
2024
2025MODULE_LICENSE("GPL");
2026MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 2027MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 2028MODULE_ALIAS_RTNL_LINK("vxlan");
Linux 6.x block layer and io_uring tree(s)