projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'for-linus-v3.10-rc3' of git://oss.sgi.com/xfs/xfs
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / drivers / net / vxlan.c
CommitLineData
d342894c 1/*
eb5ce439 2 * VXLAN: Virtual eXtensible Local Area Network
d342894c 3 *
3b8df3c6 4 * Copyright (c) 2012-2013 Vyatta Inc.
d342894c 5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 as
8 * published by the Free Software Foundation.
9 *
10 * TODO
d342894c 11 * - IPv6 (not in RFC)
12 */
13
14#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
15
16#include <linux/kernel.h>
17#include <linux/types.h>
18#include <linux/module.h>
19#include <linux/errno.h>
20#include <linux/slab.h>
21#include <linux/skbuff.h>
22#include <linux/rculist.h>
23#include <linux/netdevice.h>
24#include <linux/in.h>
25#include <linux/ip.h>
26#include <linux/udp.h>
27#include <linux/igmp.h>
28#include <linux/etherdevice.h>
29#include <linux/if_ether.h>
d342894c 30#include <linux/hash.h>
1b13c97f 31#include <linux/ethtool.h>
e4f67add
DS		 32#include <net/arp.h>
33#include <net/ndisc.h>
d342894c 34#include <net/ip.h>
c5441932 35#include <net/ip_tunnels.h>
d342894c 36#include <net/icmp.h>
37#include <net/udp.h>
38#include <net/rtnetlink.h>
39#include <net/route.h>
40#include <net/dsfield.h>
41#include <net/inet_ecn.h>
42#include <net/net_namespace.h>
43#include <net/netns/generic.h>
44
45#define VXLAN_VERSION "0.1"
46
47#define VNI_HASH_BITS 10
48#define VNI_HASH_SIZE (1<<VNI_HASH_BITS)
49#define FDB_HASH_BITS 8
50#define FDB_HASH_SIZE (1<<FDB_HASH_BITS)
51#define FDB_AGE_DEFAULT 300 /* 5 min */
52#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
53
54#define VXLAN_N_VID (1u << 24)
55#define VXLAN_VID_MASK (VXLAN_N_VID - 1)
52b702ff
AD		 56/* IP header + UDP + VXLAN + Ethernet header */
57#define VXLAN_HEADROOM (20 + 8 + 8 + 14)
d342894c 58
59#define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */
60
61/* VXLAN protocol header */
62struct vxlanhdr {
63 __be32 vx_flags;
64 __be32 vx_vni;
65};
66
23c578bf 67/* UDP port for VXLAN traffic.
68 * The IANA assigned port is 4789, but the Linux default is 8472
69 * for compatability with early adopters.
70 */
d342894c 71static unsigned int vxlan_port __read_mostly = 8472;
72module_param_named(udp_port, vxlan_port, uint, 0444);
73MODULE_PARM_DESC(udp_port, "Destination UDP port");
74
75static bool log_ecn_error = true;
76module_param(log_ecn_error, bool, 0644);
77MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
78
79/* per-net private data for this module */
80static unsigned int vxlan_net_id;
81struct vxlan_net {
82 struct socket *sock; /* UDP encap socket */
83 struct hlist_head vni_list[VNI_HASH_SIZE];
84};
85
6681712d
DS		 86struct vxlan_rdst {
87 struct rcu_head rcu;
88 __be32 remote_ip;
89 __be16 remote_port;
90 u32 remote_vni;
91 u32 remote_ifindex;
92 struct vxlan_rdst *remote_next;
93};
94
d342894c 95/* Forwarding table entry */
96struct vxlan_fdb {
97 struct hlist_node hlist; /* linked list of entries */
98 struct rcu_head rcu;
99 unsigned long updated; /* jiffies */
100 unsigned long used;
6681712d 101 struct vxlan_rdst remote;
d342894c 102 u16 state; /* see ndm_state */
ae884082 103 u8 flags; /* see ndm_flags */
d342894c 104 u8 eth_addr[ETH_ALEN];
105};
106
d342894c 107/* Pseudo network device */
108struct vxlan_dev {
109 struct hlist_node hlist;
110 struct net_device *dev;
c7995c43 111 struct vxlan_rdst default_dst; /* default destination */
d342894c 112 __be32 saddr; /* source address */
823aa873 113 __be16 dst_port;
05f47d69 114 __u16 port_min; /* source port range */
115 __u16 port_max;
d342894c 116 __u8 tos; /* TOS override */
117 __u8 ttl;
e4f67add 118 u32 flags; /* VXLAN_F_* below */
d342894c 119
120 unsigned long age_interval;
121 struct timer_list age_timer;
122 spinlock_t hash_lock;
123 unsigned int addrcnt;
124 unsigned int addrmax;
d342894c 125
126 struct hlist_head fdb_head[FDB_HASH_SIZE];
127};
128
e4f67add
DS		 129#define VXLAN_F_LEARN 0x01
130#define VXLAN_F_PROXY 0x02
131#define VXLAN_F_RSC 0x04
132#define VXLAN_F_L2MISS 0x08
133#define VXLAN_F_L3MISS 0x10
134
d342894c 135/* salt for hash table */
136static u32 vxlan_salt __read_mostly;
137
138static inline struct hlist_head *vni_head(struct net *net, u32 id)
139{
140 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
141
142 return &vn->vni_list[hash_32(id, VNI_HASH_BITS)];
143}
144
145/* Look up VNI in a per net namespace table */
146static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id)
147{
148 struct vxlan_dev *vxlan;
d342894c 149
b67bfe0d 150 hlist_for_each_entry_rcu(vxlan, vni_head(net, id), hlist) {
c7995c43 151 if (vxlan->default_dst.remote_vni == id)
d342894c 152 return vxlan;
153 }
154
155 return NULL;
156}
157
158/* Fill in neighbour message in skbuff. */
159static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
160 const struct vxlan_fdb *fdb,
6681712d
DS		 161 u32 portid, u32 seq, int type, unsigned int flags,
162 const struct vxlan_rdst *rdst)
d342894c 163{
164 unsigned long now = jiffies;
165 struct nda_cacheinfo ci;
166 struct nlmsghdr *nlh;
167 struct ndmsg *ndm;
e4f67add 168 bool send_ip, send_eth;
d342894c 169
170 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
171 if (nlh == NULL)
172 return -EMSGSIZE;
173
174 ndm = nlmsg_data(nlh);
175 memset(ndm, 0, sizeof(*ndm));
e4f67add
DS		 176
177 send_eth = send_ip = true;
178
179 if (type == RTM_GETNEIGH) {
180 ndm->ndm_family = AF_INET;
6681712d 181 send_ip = rdst->remote_ip != htonl(INADDR_ANY);
e4f67add
DS		 182 send_eth = !is_zero_ether_addr(fdb->eth_addr);
183 } else
184 ndm->ndm_family = AF_BRIDGE;
d342894c 185 ndm->ndm_state = fdb->state;
186 ndm->ndm_ifindex = vxlan->dev->ifindex;
ae884082 187 ndm->ndm_flags = fdb->flags;
d342894c 188 ndm->ndm_type = NDA_DST;
189
e4f67add 190 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
d342894c 191 goto nla_put_failure;
192
6681712d
DS		 193 if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip))
194 goto nla_put_failure;
195
823aa873 196 if (rdst->remote_port && rdst->remote_port != vxlan->dst_port &&
6681712d
DS		 197 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
198 goto nla_put_failure;
c7995c43 199 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
6681712d
DS		 200 nla_put_be32(skb, NDA_VNI, rdst->remote_vni))
201 goto nla_put_failure;
202 if (rdst->remote_ifindex &&
203 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
d342894c 204 goto nla_put_failure;
205
206 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
207 ci.ndm_confirmed = 0;
208 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
209 ci.ndm_refcnt = 0;
210
211 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
212 goto nla_put_failure;
213
214 return nlmsg_end(skb, nlh);
215
216nla_put_failure:
217 nlmsg_cancel(skb, nlh);
218 return -EMSGSIZE;
219}
220
221static inline size_t vxlan_nlmsg_size(void)
222{
223 return NLMSG_ALIGN(sizeof(struct ndmsg))
224 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
225 + nla_total_size(sizeof(__be32)) /* NDA_DST */
73cf3317 226 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
6681712d
DS		 227 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
228 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
d342894c 229 + nla_total_size(sizeof(struct nda_cacheinfo));
230}
231
232static void vxlan_fdb_notify(struct vxlan_dev *vxlan,
233 const struct vxlan_fdb *fdb, int type)
234{
235 struct net *net = dev_net(vxlan->dev);
236 struct sk_buff *skb;
237 int err = -ENOBUFS;
238
239 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
240 if (skb == NULL)
241 goto errout;
242
6681712d 243 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, &fdb->remote);
d342894c 244 if (err < 0) {
245 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
246 WARN_ON(err == -EMSGSIZE);
247 kfree_skb(skb);
248 goto errout;
249 }
250
251 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
252 return;
253errout:
254 if (err < 0)
255 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
256}
257
e4f67add
DS		 258static void vxlan_ip_miss(struct net_device *dev, __be32 ipa)
259{
260 struct vxlan_dev *vxlan = netdev_priv(dev);
261 struct vxlan_fdb f;
262
263 memset(&f, 0, sizeof f);
264 f.state = NUD_STALE;
6681712d
DS		 265 f.remote.remote_ip = ipa; /* goes to NDA_DST */
266 f.remote.remote_vni = VXLAN_N_VID;
e4f67add
DS		 267
268 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
269}
270
271static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
272{
273 struct vxlan_fdb f;
274
275 memset(&f, 0, sizeof f);
276 f.state = NUD_STALE;
277 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
278
279 vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH);
280}
281
d342894c 282/* Hash Ethernet address */
283static u32 eth_hash(const unsigned char *addr)
284{
285 u64 value = get_unaligned((u64 *)addr);
286
287 /* only want 6 bytes */
288#ifdef __BIG_ENDIAN
d342894c 289 value >>= 16;
321fb991 290#else
291 value <<= 16;
d342894c 292#endif
293 return hash_64(value, FDB_HASH_BITS);
294}
295
296/* Hash chain to use given mac address */
297static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
298 const u8 *mac)
299{
300 return &vxlan->fdb_head[eth_hash(mac)];
301}
302
303/* Look up Ethernet address in forwarding table */
014be2c8 304static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
d342894c 305 const u8 *mac)
306
307{
308 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
309 struct vxlan_fdb *f;
d342894c 310
b67bfe0d 311 hlist_for_each_entry_rcu(f, head, hlist) {
d342894c 312 if (compare_ether_addr(mac, f->eth_addr) == 0)
313 return f;
314 }
315
316 return NULL;
317}
318
014be2c8
SS		 319static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
320 const u8 *mac)
321{
322 struct vxlan_fdb *f;
323
324 f = __vxlan_find_mac(vxlan, mac);
325 if (f)
326 f->used = jiffies;
327
328 return f;
329}
330
6681712d
DS		 331/* Add/update destinations for multicast */
332static int vxlan_fdb_append(struct vxlan_fdb *f,
73cf3317 333 __be32 ip, __be16 port, __u32 vni, __u32 ifindex)
6681712d
DS		 334{
335 struct vxlan_rdst *rd_prev, *rd;
336
337 rd_prev = NULL;
338 for (rd = &f->remote; rd; rd = rd->remote_next) {
339 if (rd->remote_ip == ip &&
340 rd->remote_port == port &&
341 rd->remote_vni == vni &&
342 rd->remote_ifindex == ifindex)
343 return 0;
344 rd_prev = rd;
345 }
346 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
347 if (rd == NULL)
348 return -ENOBUFS;
349 rd->remote_ip = ip;
350 rd->remote_port = port;
351 rd->remote_vni = vni;
352 rd->remote_ifindex = ifindex;
353 rd->remote_next = NULL;
354 rd_prev->remote_next = rd;
355 return 1;
356}
357
d342894c 358/* Add new entry to forwarding table -- assumes lock held */
359static int vxlan_fdb_create(struct vxlan_dev *vxlan,
360 const u8 *mac, __be32 ip,
6681712d 361 __u16 state, __u16 flags,
73cf3317 362 __be16 port, __u32 vni, __u32 ifindex,
ae884082 363 __u8 ndm_flags)
d342894c 364{
365 struct vxlan_fdb *f;
366 int notify = 0;
367
014be2c8 368 f = __vxlan_find_mac(vxlan, mac);
d342894c 369 if (f) {
370 if (flags & NLM_F_EXCL) {
371 netdev_dbg(vxlan->dev,
372 "lost race to create %pM\n", mac);
373 return -EEXIST;
374 }
375 if (f->state != state) {
376 f->state = state;
377 f->updated = jiffies;
378 notify = 1;
379 }
ae884082
DS		 380 if (f->flags != ndm_flags) {
381 f->flags = ndm_flags;
382 f->updated = jiffies;
383 notify = 1;
384 }
6681712d
DS		 385 if ((flags & NLM_F_APPEND) &&
386 is_multicast_ether_addr(f->eth_addr)) {
387 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex);
388
389 if (rc < 0)
390 return rc;
391 notify |= rc;
392 }
d342894c 393 } else {
394 if (!(flags & NLM_F_CREATE))
395 return -ENOENT;
396
397 if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax)
398 return -ENOSPC;
399
400 netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip);
401 f = kmalloc(sizeof(*f), GFP_ATOMIC);
402 if (!f)
403 return -ENOMEM;
404
405 notify = 1;
6681712d
DS		 406 f->remote.remote_ip = ip;
407 f->remote.remote_port = port;
408 f->remote.remote_vni = vni;
409 f->remote.remote_ifindex = ifindex;
410 f->remote.remote_next = NULL;
d342894c 411 f->state = state;
ae884082 412 f->flags = ndm_flags;
d342894c 413 f->updated = f->used = jiffies;
414 memcpy(f->eth_addr, mac, ETH_ALEN);
415
416 ++vxlan->addrcnt;
417 hlist_add_head_rcu(&f->hlist,
418 vxlan_fdb_head(vxlan, mac));
419 }
420
421 if (notify)
422 vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH);
423
424 return 0;
425}
426
6706c82e 427static void vxlan_fdb_free(struct rcu_head *head)
6681712d
DS		 428{
429 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
430
431 while (f->remote.remote_next) {
432 struct vxlan_rdst *rd = f->remote.remote_next;
433
434 f->remote.remote_next = rd->remote_next;
435 kfree(rd);
436 }
437 kfree(f);
438}
439
d342894c 440static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
441{
442 netdev_dbg(vxlan->dev,
443 "delete %pM\n", f->eth_addr);
444
445 --vxlan->addrcnt;
446 vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH);
447
448 hlist_del_rcu(&f->hlist);
6681712d 449 call_rcu(&f->rcu, vxlan_fdb_free);
d342894c 450}
451
452/* Add static entry (via netlink) */
453static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
454 struct net_device *dev,
455 const unsigned char *addr, u16 flags)
456{
457 struct vxlan_dev *vxlan = netdev_priv(dev);
6681712d 458 struct net *net = dev_net(vxlan->dev);
d342894c 459 __be32 ip;
73cf3317 460 __be16 port;
461 u32 vni, ifindex;
d342894c 462 int err;
463
464 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
465 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
466 ndm->ndm_state);
467 return -EINVAL;
468 }
469
470 if (tb[NDA_DST] == NULL)
471 return -EINVAL;
472
473 if (nla_len(tb[NDA_DST]) != sizeof(__be32))
474 return -EAFNOSUPPORT;
475
476 ip = nla_get_be32(tb[NDA_DST]);
477
6681712d 478 if (tb[NDA_PORT]) {
73cf3317 479 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
6681712d 480 return -EINVAL;
73cf3317 481 port = nla_get_be16(tb[NDA_PORT]);
6681712d 482 } else
823aa873 483 port = vxlan->dst_port;
6681712d
DS		 484
485 if (tb[NDA_VNI]) {
486 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
487 return -EINVAL;
488 vni = nla_get_u32(tb[NDA_VNI]);
489 } else
c7995c43 490 vni = vxlan->default_dst.remote_vni;
6681712d
DS		 491
492 if (tb[NDA_IFINDEX]) {
5abb0029 493 struct net_device *tdev;
6681712d
DS		 494
495 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
496 return -EINVAL;
497 ifindex = nla_get_u32(tb[NDA_IFINDEX]);
5abb0029
PS		 498 tdev = dev_get_by_index(net, ifindex);
499 if (!tdev)
6681712d 500 return -EADDRNOTAVAIL;
5abb0029 501 dev_put(tdev);
6681712d
DS		 502 } else
503 ifindex = 0;
504
d342894c 505 spin_lock_bh(&vxlan->hash_lock);
73cf3317 506 err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags,
507 port, vni, ifindex, ndm->ndm_flags);
d342894c 508 spin_unlock_bh(&vxlan->hash_lock);
509
510 return err;
511}
512
513/* Delete entry (via netlink) */
1690be63
VY		 514static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
515 struct net_device *dev,
d342894c 516 const unsigned char *addr)
517{
518 struct vxlan_dev *vxlan = netdev_priv(dev);
519 struct vxlan_fdb *f;
520 int err = -ENOENT;
521
522 spin_lock_bh(&vxlan->hash_lock);
523 f = vxlan_find_mac(vxlan, addr);
524 if (f) {
525 vxlan_fdb_destroy(vxlan, f);
526 err = 0;
527 }
528 spin_unlock_bh(&vxlan->hash_lock);
529
530 return err;
531}
532
533/* Dump forwarding table */
534static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
535 struct net_device *dev, int idx)
536{
537 struct vxlan_dev *vxlan = netdev_priv(dev);
538 unsigned int h;
539
540 for (h = 0; h < FDB_HASH_SIZE; ++h) {
541 struct vxlan_fdb *f;
d342894c 542 int err;
543
b67bfe0d 544 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
6681712d
DS		 545 struct vxlan_rdst *rd;
546 for (rd = &f->remote; rd; rd = rd->remote_next) {
547 if (idx < cb->args[0])
548 goto skip;
549
550 err = vxlan_fdb_info(skb, vxlan, f,
551 NETLINK_CB(cb->skb).portid,
552 cb->nlh->nlmsg_seq,
553 RTM_NEWNEIGH,
554 NLM_F_MULTI, rd);
555 if (err < 0)
556 break;
d342894c 557skip:
6681712d
DS		 558 ++idx;
559 }
d342894c 560 }
561 }
562
563 return idx;
564}
565
566/* Watch incoming packets to learn mapping between Ethernet address
567 * and Tunnel endpoint.
568 */
569static void vxlan_snoop(struct net_device *dev,
570 __be32 src_ip, const u8 *src_mac)
571{
572 struct vxlan_dev *vxlan = netdev_priv(dev);
573 struct vxlan_fdb *f;
574 int err;
575
576 f = vxlan_find_mac(vxlan, src_mac);
577 if (likely(f)) {
6681712d 578 if (likely(f->remote.remote_ip == src_ip))
d342894c 579 return;
580
581 if (net_ratelimit())
582 netdev_info(dev,
583 "%pM migrated from %pI4 to %pI4\n",
6681712d 584 src_mac, &f->remote.remote_ip, &src_ip);
d342894c 585
6681712d 586 f->remote.remote_ip = src_ip;
d342894c 587 f->updated = jiffies;
588 } else {
589 /* learned new entry */
590 spin_lock(&vxlan->hash_lock);
591 err = vxlan_fdb_create(vxlan, src_mac, src_ip,
592 NUD_REACHABLE,
6681712d 593 NLM_F_EXCL|NLM_F_CREATE,
823aa873 594 vxlan->dst_port,
ae884082
DS		 595 vxlan->default_dst.remote_vni,
596 0, NTF_SELF);
d342894c 597 spin_unlock(&vxlan->hash_lock);
598 }
599}
600
601
602/* See if multicast group is already in use by other ID */
603static bool vxlan_group_used(struct vxlan_net *vn,
604 const struct vxlan_dev *this)
605{
606 const struct vxlan_dev *vxlan;
d342894c 607 unsigned h;
608
609 for (h = 0; h < VNI_HASH_SIZE; ++h)
b67bfe0d 610 hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist) {
d342894c 611 if (vxlan == this)
612 continue;
613
614 if (!netif_running(vxlan->dev))
615 continue;
616
c7995c43 617 if (vxlan->default_dst.remote_ip == this->default_dst.remote_ip)
d342894c 618 return true;
619 }
620
621 return false;
622}
623
624/* kernel equivalent to IP_ADD_MEMBERSHIP */
625static int vxlan_join_group(struct net_device *dev)
626{
627 struct vxlan_dev *vxlan = netdev_priv(dev);
628 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
629 struct sock *sk = vn->sock->sk;
630 struct ip_mreqn mreq = {
c7995c43
AW		 631 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
632 .imr_ifindex = vxlan->default_dst.remote_ifindex,
d342894c 633 };
634 int err;
635
636 /* Already a member of group */
637 if (vxlan_group_used(vn, vxlan))
638 return 0;
639
640 /* Need to drop RTNL to call multicast join */
641 rtnl_unlock();
642 lock_sock(sk);
643 err = ip_mc_join_group(sk, &mreq);
644 release_sock(sk);
645 rtnl_lock();
646
647 return err;
648}
649
650
651/* kernel equivalent to IP_DROP_MEMBERSHIP */
652static int vxlan_leave_group(struct net_device *dev)
653{
654 struct vxlan_dev *vxlan = netdev_priv(dev);
655 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
656 int err = 0;
657 struct sock *sk = vn->sock->sk;
658 struct ip_mreqn mreq = {
c7995c43
AW		 659 .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip,
660 .imr_ifindex = vxlan->default_dst.remote_ifindex,
d342894c 661 };
662
663 /* Only leave group when last vxlan is done. */
664 if (vxlan_group_used(vn, vxlan))
665 return 0;
666
667 /* Need to drop RTNL to call multicast leave */
668 rtnl_unlock();
669 lock_sock(sk);
670 err = ip_mc_leave_group(sk, &mreq);
671 release_sock(sk);
672 rtnl_lock();
673
674 return err;
675}
676
677/* Callback from net/ipv4/udp.c to receive packets */
678static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
679{
680 struct iphdr *oip;
681 struct vxlanhdr *vxh;
682 struct vxlan_dev *vxlan;
e8171045 683 struct pcpu_tstats *stats;
d342894c 684 __u32 vni;
685 int err;
686
687 /* pop off outer UDP header */
688 __skb_pull(skb, sizeof(struct udphdr));
689
690 /* Need Vxlan and inner Ethernet header to be present */
691 if (!pskb_may_pull(skb, sizeof(struct vxlanhdr)))
692 goto error;
693
694 /* Drop packets with reserved bits set */
695 vxh = (struct vxlanhdr *) skb->data;
696 if (vxh->vx_flags != htonl(VXLAN_FLAGS) ||
697 (vxh->vx_vni & htonl(0xff))) {
698 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
699 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
700 goto error;
701 }
702
703 __skb_pull(skb, sizeof(struct vxlanhdr));
d342894c 704
705 /* Is this VNI defined? */
706 vni = ntohl(vxh->vx_vni) >> 8;
707 vxlan = vxlan_find_vni(sock_net(sk), vni);
708 if (!vxlan) {
709 netdev_dbg(skb->dev, "unknown vni %d\n", vni);
710 goto drop;
711 }
712
713 if (!pskb_may_pull(skb, ETH_HLEN)) {
714 vxlan->dev->stats.rx_length_errors++;
715 vxlan->dev->stats.rx_errors++;
716 goto drop;
717 }
718
e4f67add
DS		 719 skb_reset_mac_header(skb);
720
d342894c 721 /* Re-examine inner Ethernet packet */
722 oip = ip_hdr(skb);
723 skb->protocol = eth_type_trans(skb, vxlan->dev);
d342894c 724
725 /* Ignore packet loops (and multicast echo) */
726 if (compare_ether_addr(eth_hdr(skb)->h_source,
727 vxlan->dev->dev_addr) == 0)
728 goto drop;
729
e4f67add 730 if (vxlan->flags & VXLAN_F_LEARN)
d342894c 731 vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source);
732
733 __skb_tunnel_rx(skb, vxlan->dev);
734 skb_reset_network_header(skb);
0afb1666
JG		 735
736 /* If the NIC driver gave us an encapsulated packet with
737 * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled,
738 * leave the CHECKSUM_UNNECESSARY, the device checksummed it
739 * for us. Otherwise force the upper layers to verify it.
740 */
741 if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation ||
742 !(vxlan->dev->features & NETIF_F_RXCSUM))
743 skb->ip_summed = CHECKSUM_NONE;
744
745 skb->encapsulation = 0;
d342894c 746
747 err = IP_ECN_decapsulate(oip, skb);
748 if (unlikely(err)) {
749 if (log_ecn_error)
750 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
751 &oip->saddr, oip->tos);
752 if (err > 1) {
753 ++vxlan->dev->stats.rx_frame_errors;
754 ++vxlan->dev->stats.rx_errors;
755 goto drop;
756 }
757 }
758
e8171045 759 stats = this_cpu_ptr(vxlan->dev->tstats);
d342894c 760 u64_stats_update_begin(&stats->syncp);
761 stats->rx_packets++;
762 stats->rx_bytes += skb->len;
763 u64_stats_update_end(&stats->syncp);
764
765 netif_rx(skb);
766
767 return 0;
768error:
769 /* Put UDP header back */
770 __skb_push(skb, sizeof(struct udphdr));
771
772 return 1;
773drop:
774 /* Consume bad packet */
775 kfree_skb(skb);
776 return 0;
777}
778
e4f67add
DS		 779static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
780{
781 struct vxlan_dev *vxlan = netdev_priv(dev);
782 struct arphdr *parp;
783 u8 *arpptr, *sha;
784 __be32 sip, tip;
785 struct neighbour *n;
786
787 if (dev->flags & IFF_NOARP)
788 goto out;
789
790 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
791 dev->stats.tx_dropped++;
792 goto out;
793 }
794 parp = arp_hdr(skb);
795
796 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
797 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
798 parp->ar_pro != htons(ETH_P_IP) ||
799 parp->ar_op != htons(ARPOP_REQUEST) ||
800 parp->ar_hln != dev->addr_len ||
801 parp->ar_pln != 4)
802 goto out;
803 arpptr = (u8 *)parp + sizeof(struct arphdr);
804 sha = arpptr;
805 arpptr += dev->addr_len; /* sha */
806 memcpy(&sip, arpptr, sizeof(sip));
807 arpptr += sizeof(sip);
808 arpptr += dev->addr_len; /* tha */
809 memcpy(&tip, arpptr, sizeof(tip));
810
811 if (ipv4_is_loopback(tip) ||
812 ipv4_is_multicast(tip))
813 goto out;
814
815 n = neigh_lookup(&arp_tbl, &tip, dev);
816
817 if (n) {
e4f67add
DS		 818 struct vxlan_fdb *f;
819 struct sk_buff *reply;
820
821 if (!(n->nud_state & NUD_CONNECTED)) {
822 neigh_release(n);
823 goto out;
824 }
825
826 f = vxlan_find_mac(vxlan, n->ha);
6681712d 827 if (f && f->remote.remote_ip == htonl(INADDR_ANY)) {
e4f67add
DS		 828 /* bridge-local neighbor */
829 neigh_release(n);
830 goto out;
831 }
832
833 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
834 n->ha, sha);
835
836 neigh_release(n);
837
838 skb_reset_mac_header(reply);
839 __skb_pull(reply, skb_network_offset(reply));
840 reply->ip_summed = CHECKSUM_UNNECESSARY;
841 reply->pkt_type = PACKET_HOST;
842
843 if (netif_rx_ni(reply) == NET_RX_DROP)
844 dev->stats.rx_dropped++;
845 } else if (vxlan->flags & VXLAN_F_L3MISS)
846 vxlan_ip_miss(dev, tip);
847out:
848 consume_skb(skb);
849 return NETDEV_TX_OK;
850}
851
852static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
853{
854 struct vxlan_dev *vxlan = netdev_priv(dev);
855 struct neighbour *n;
856 struct iphdr *pip;
857
858 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
859 return false;
860
861 n = NULL;
862 switch (ntohs(eth_hdr(skb)->h_proto)) {
863 case ETH_P_IP:
864 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
865 return false;
866 pip = ip_hdr(skb);
867 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
868 break;
869 default:
870 return false;
871 }
872
873 if (n) {
874 bool diff;
875
876 diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0;
877 if (diff) {
878 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
879 dev->addr_len);
880 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
881 }
882 neigh_release(n);
883 return diff;
884 } else if (vxlan->flags & VXLAN_F_L3MISS)
885 vxlan_ip_miss(dev, pip->daddr);
886 return false;
887}
888
1cad8715 889static void vxlan_sock_free(struct sk_buff *skb)
890{
891 sock_put(skb->sk);
892}
893
894/* On transmit, associate with the tunnel socket */
895static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb)
896{
897 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
898 struct sock *sk = vn->sock->sk;
899
900 skb_orphan(skb);
901 sock_hold(sk);
902 skb->sk = sk;
903 skb->destructor = vxlan_sock_free;
904}
905
05f47d69 906/* Compute source port for outgoing packet
907 * first choice to use L4 flow hash since it will spread
908 * better and maybe available from hardware
909 * secondary choice is to use jhash on the Ethernet header
910 */
7d836a76 911static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb)
05f47d69 912{
913 unsigned int range = (vxlan->port_max - vxlan->port_min) + 1;
914 u32 hash;
915
916 hash = skb_get_rxhash(skb);
917 if (!hash)
918 hash = jhash(skb->data, 2 * ETH_ALEN,
919 (__force u32) skb->protocol);
920
7d836a76 921 return htons((((u64) hash * range) >> 32) + vxlan->port_min);
05f47d69 922}
923
05c0db08
PS		 924static int handle_offloads(struct sk_buff *skb)
925{
926 if (skb_is_gso(skb)) {
927 int err = skb_unclone(skb, GFP_ATOMIC);
928 if (unlikely(err))
929 return err;
930
f6ace502 931 skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL;
05c0db08
PS		 932 } else if (skb->ip_summed != CHECKSUM_PARTIAL)
933 skb->ip_summed = CHECKSUM_NONE;
934
935 return 0;
936}
937
9dcc71e1
SS		 938/* Bypass encapsulation if the destination is local */
939static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
940 struct vxlan_dev *dst_vxlan)
941{
942 struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
943 struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
944
945 skb->pkt_type = PACKET_HOST;
946 skb->encapsulation = 0;
947 skb->dev = dst_vxlan->dev;
948 __skb_pull(skb, skb_network_offset(skb));
949
950 if (dst_vxlan->flags & VXLAN_F_LEARN)
9d9f163c
MR		 951 vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK),
952 eth_hdr(skb)->h_source);
9dcc71e1
SS		 953
954 u64_stats_update_begin(&tx_stats->syncp);
955 tx_stats->tx_packets++;
956 tx_stats->tx_bytes += skb->len;
957 u64_stats_update_end(&tx_stats->syncp);
958
959 if (netif_rx(skb) == NET_RX_SUCCESS) {
960 u64_stats_update_begin(&rx_stats->syncp);
961 rx_stats->rx_packets++;
962 rx_stats->rx_bytes += skb->len;
963 u64_stats_update_end(&rx_stats->syncp);
964 } else {
965 skb->dev->stats.rx_dropped++;
966 }
967}
968
6681712d
DS		 969static netdev_tx_t vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
970 struct vxlan_rdst *rdst, bool did_rsc)
d342894c 971{
972 struct vxlan_dev *vxlan = netdev_priv(dev);
973 struct rtable *rt;
d342894c 974 const struct iphdr *old_iph;
975 struct iphdr *iph;
976 struct vxlanhdr *vxh;
977 struct udphdr *uh;
978 struct flowi4 fl4;
d342894c 979 __be32 dst;
7d836a76 980 __be16 src_port, dst_port;
6681712d 981 u32 vni;
d342894c 982 __be16 df = 0;
983 __u8 tos, ttl;
d342894c 984
823aa873 985 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port;
6681712d
DS		 986 vni = rdst->remote_vni;
987 dst = rdst->remote_ip;
e4f67add
DS		 988
989 if (!dst) {
990 if (did_rsc) {
e4f67add 991 /* short-circuited back to local bridge */
9dcc71e1 992 vxlan_encap_bypass(skb, vxlan, vxlan);
e4f67add
DS		 993 return NETDEV_TX_OK;
994 }
ef59febe 995 goto drop;
e4f67add 996 }
ef59febe 997
d6727fe3
JG		 998 if (!skb->encapsulation) {
999 skb_reset_inner_headers(skb);
1000 skb->encapsulation = 1;
1001 }
1002
d342894c 1003 /* Need space for new headers (invalidates iph ptr) */
1004 if (skb_cow_head(skb, VXLAN_HEADROOM))
1005 goto drop;
1006
d342894c 1007 old_iph = ip_hdr(skb);
1008
d342894c 1009 ttl = vxlan->ttl;
1010 if (!ttl && IN_MULTICAST(ntohl(dst)))
1011 ttl = 1;
1012
1013 tos = vxlan->tos;
1014 if (tos == 1)
206aaafc 1015 tos = ip_tunnel_get_dsfield(old_iph, skb);
d342894c 1016
05f47d69 1017 src_port = vxlan_src_port(vxlan, skb);
d342894c 1018
ca78f181 1019 memset(&fl4, 0, sizeof(fl4));
6681712d 1020 fl4.flowi4_oif = rdst->remote_ifindex;
ca78f181 1021 fl4.flowi4_tos = RT_TOS(tos);
1022 fl4.daddr = dst;
1023 fl4.saddr = vxlan->saddr;
1024
1025 rt = ip_route_output_key(dev_net(dev), &fl4);
d342894c 1026 if (IS_ERR(rt)) {
1027 netdev_dbg(dev, "no route to %pI4\n", &dst);
1028 dev->stats.tx_carrier_errors++;
1029 goto tx_error;
1030 }
1031
1032 if (rt->dst.dev == dev) {
1033 netdev_dbg(dev, "circular route to %pI4\n", &dst);
1034 ip_rt_put(rt);
1035 dev->stats.collisions++;
1036 goto tx_error;
1037 }
1038
9dcc71e1 1039 /* Bypass encapsulation if the destination is local */
ab09a6d0
MR		 1040 if (rt->rt_flags & RTCF_LOCAL &&
1041 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
9dcc71e1
SS		 1042 struct vxlan_dev *dst_vxlan;
1043
1044 ip_rt_put(rt);
1045 dst_vxlan = vxlan_find_vni(dev_net(dev), vni);
1046 if (!dst_vxlan)
1047 goto tx_error;
1048 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1049 return NETDEV_TX_OK;
1050 }
1051
d342894c 1052 memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1053 IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED |
1054 IPSKB_REROUTED);
1055 skb_dst_drop(skb);
1056 skb_dst_set(skb, &rt->dst);
1057
1058 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
1059 vxh->vx_flags = htonl(VXLAN_FLAGS);
6681712d 1060 vxh->vx_vni = htonl(vni << 8);
d342894c 1061
1062 __skb_push(skb, sizeof(*uh));
1063 skb_reset_transport_header(skb);
1064 uh = udp_hdr(skb);
1065
73cf3317 1066 uh->dest = dst_port;
7d836a76 1067 uh->source = src_port;
d342894c 1068
1069 uh->len = htons(skb->len);
1070 uh->check = 0;
1071
1072 __skb_push(skb, sizeof(*iph));
1073 skb_reset_network_header(skb);
1074 iph = ip_hdr(skb);
1075 iph->version = 4;
1076 iph->ihl = sizeof(struct iphdr) >> 2;
1077 iph->frag_off = df;
1078 iph->protocol = IPPROTO_UDP;
206aaafc 1079 iph->tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
ca78f181 1080 iph->daddr = dst;
d342894c 1081 iph->saddr = fl4.saddr;
1082 iph->ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
8dc98eb2 1083 tunnel_ip_select_ident(skb, old_iph, &rt->dst);
d342894c 1084
88c4c066
ZM		 1085 nf_reset(skb);
1086
1cad8715 1087 vxlan_set_owner(dev, skb);
1088
05c0db08
PS		 1089 if (handle_offloads(skb))
1090 goto drop;
d342894c 1091
6aed0c8b 1092 iptunnel_xmit(skb, dev);
d342894c 1093 return NETDEV_TX_OK;
1094
1095drop:
1096 dev->stats.tx_dropped++;
1097 goto tx_free;
1098
1099tx_error:
1100 dev->stats.tx_errors++;
1101tx_free:
1102 dev_kfree_skb(skb);
1103 return NETDEV_TX_OK;
1104}
1105
6681712d
DS		 1106/* Transmit local packets over Vxlan
1107 *
1108 * Outer IP header inherits ECN and DF from inner header.
1109 * Outer UDP destination is the VXLAN assigned port.
1110 * source port is based on hash of flow
1111 */
1112static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
1113{
1114 struct vxlan_dev *vxlan = netdev_priv(dev);
1115 struct ethhdr *eth;
1116 bool did_rsc = false;
c7995c43 1117 struct vxlan_rdst *rdst0, *rdst;
6681712d
DS		 1118 struct vxlan_fdb *f;
1119 int rc1, rc;
1120
1121 skb_reset_mac_header(skb);
1122 eth = eth_hdr(skb);
1123
1124 if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP)
1125 return arp_reduce(dev, skb);
6681712d
DS		 1126
1127 f = vxlan_find_mac(vxlan, eth->h_dest);
ae884082
DS		 1128 did_rsc = false;
1129
1130 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
1131 ntohs(eth->h_proto) == ETH_P_IP) {
1132 did_rsc = route_shortcircuit(dev, skb);
1133 if (did_rsc)
1134 f = vxlan_find_mac(vxlan, eth->h_dest);
1135 }
1136
6681712d 1137 if (f == NULL) {
c7995c43
AW		 1138 rdst0 = &vxlan->default_dst;
1139
1140 if (rdst0->remote_ip == htonl(INADDR_ANY) &&
6681712d
DS		 1141 (vxlan->flags & VXLAN_F_L2MISS) &&
1142 !is_multicast_ether_addr(eth->h_dest))
1143 vxlan_fdb_miss(vxlan, eth->h_dest);
1144 } else
1145 rdst0 = &f->remote;
1146
1147 rc = NETDEV_TX_OK;
1148
1149 /* if there are multiple destinations, send copies */
1150 for (rdst = rdst0->remote_next; rdst; rdst = rdst->remote_next) {
1151 struct sk_buff *skb1;
1152
1153 skb1 = skb_clone(skb, GFP_ATOMIC);
1154 rc1 = vxlan_xmit_one(skb1, dev, rdst, did_rsc);
1155 if (rc == NETDEV_TX_OK)
1156 rc = rc1;
1157 }
1158
1159 rc1 = vxlan_xmit_one(skb, dev, rdst0, did_rsc);
1160 if (rc == NETDEV_TX_OK)
1161 rc = rc1;
1162 return rc;
1163}
1164
d342894c 1165/* Walk the forwarding table and purge stale entries */
1166static void vxlan_cleanup(unsigned long arg)
1167{
1168 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
1169 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1170 unsigned int h;
1171
1172 if (!netif_running(vxlan->dev))
1173 return;
1174
1175 spin_lock_bh(&vxlan->hash_lock);
1176 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1177 struct hlist_node *p, *n;
1178 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1179 struct vxlan_fdb *f
1180 = container_of(p, struct vxlan_fdb, hlist);
1181 unsigned long timeout;
1182
3c172868 1183 if (f->state & NUD_PERMANENT)
d342894c 1184 continue;
1185
1186 timeout = f->used + vxlan->age_interval * HZ;
1187 if (time_before_eq(timeout, jiffies)) {
1188 netdev_dbg(vxlan->dev,
1189 "garbage collect %pM\n",
1190 f->eth_addr);
1191 f->state = NUD_STALE;
1192 vxlan_fdb_destroy(vxlan, f);
1193 } else if (time_before(timeout, next_timer))
1194 next_timer = timeout;
1195 }
1196 }
1197 spin_unlock_bh(&vxlan->hash_lock);
1198
1199 mod_timer(&vxlan->age_timer, next_timer);
1200}
1201
1202/* Setup stats when device is created */
1203static int vxlan_init(struct net_device *dev)
1204{
e8171045
PS		 1205 dev->tstats = alloc_percpu(struct pcpu_tstats);
1206 if (!dev->tstats)
d342894c 1207 return -ENOMEM;
1208
1209 return 0;
1210}
1211
1212/* Start ageing timer and join group when device is brought up */
1213static int vxlan_open(struct net_device *dev)
1214{
1215 struct vxlan_dev *vxlan = netdev_priv(dev);
1216 int err;
1217
c7995c43 1218 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) {
d342894c 1219 err = vxlan_join_group(dev);
1220 if (err)
1221 return err;
1222 }
1223
1224 if (vxlan->age_interval)
1225 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1226
1227 return 0;
1228}
1229
1230/* Purge the forwarding table */
1231static void vxlan_flush(struct vxlan_dev *vxlan)
1232{
1233 unsigned h;
1234
1235 spin_lock_bh(&vxlan->hash_lock);
1236 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1237 struct hlist_node *p, *n;
1238 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1239 struct vxlan_fdb *f
1240 = container_of(p, struct vxlan_fdb, hlist);
1241 vxlan_fdb_destroy(vxlan, f);
1242 }
1243 }
1244 spin_unlock_bh(&vxlan->hash_lock);
1245}
1246
1247/* Cleanup timer and forwarding table on shutdown */
1248static int vxlan_stop(struct net_device *dev)
1249{
1250 struct vxlan_dev *vxlan = netdev_priv(dev);
1251
c7995c43 1252 if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip)))
d342894c 1253 vxlan_leave_group(dev);
1254
1255 del_timer_sync(&vxlan->age_timer);
1256
1257 vxlan_flush(vxlan);
1258
1259 return 0;
1260}
1261
d342894c 1262/* Stub, nothing needs to be done. */
1263static void vxlan_set_multicast_list(struct net_device *dev)
1264{
1265}
1266
1267static const struct net_device_ops vxlan_netdev_ops = {
1268 .ndo_init = vxlan_init,
1269 .ndo_open = vxlan_open,
1270 .ndo_stop = vxlan_stop,
1271 .ndo_start_xmit = vxlan_xmit,
e8171045 1272 .ndo_get_stats64 = ip_tunnel_get_stats64,
d342894c 1273 .ndo_set_rx_mode = vxlan_set_multicast_list,
1274 .ndo_change_mtu = eth_change_mtu,
1275 .ndo_validate_addr = eth_validate_addr,
1276 .ndo_set_mac_address = eth_mac_addr,
1277 .ndo_fdb_add = vxlan_fdb_add,
1278 .ndo_fdb_del = vxlan_fdb_delete,
1279 .ndo_fdb_dump = vxlan_fdb_dump,
1280};
1281
1282/* Info for udev, that this is a virtual tunnel endpoint */
1283static struct device_type vxlan_type = {
1284 .name = "vxlan",
1285};
1286
1287static void vxlan_free(struct net_device *dev)
1288{
e8171045 1289 free_percpu(dev->tstats);
d342894c 1290 free_netdev(dev);
1291}
1292
1293/* Initialize the device structure. */
1294static void vxlan_setup(struct net_device *dev)
1295{
1296 struct vxlan_dev *vxlan = netdev_priv(dev);
1297 unsigned h;
05f47d69 1298 int low, high;
d342894c 1299
1300 eth_hw_addr_random(dev);
1301 ether_setup(dev);
2840bf22 1302 dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM;
d342894c 1303
1304 dev->netdev_ops = &vxlan_netdev_ops;
1305 dev->destructor = vxlan_free;
1306 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1307
1308 dev->tx_queue_len = 0;
1309 dev->features |= NETIF_F_LLTX;
1310 dev->features |= NETIF_F_NETNS_LOCAL;
d6727fe3 1311 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
0afb1666 1312 dev->features |= NETIF_F_RXCSUM;
05c0db08 1313 dev->features |= NETIF_F_GSO_SOFTWARE;
0afb1666
JG		 1314
1315 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
05c0db08 1316 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
d342894c 1317 dev->priv_flags &= ~IFF_XMIT_DST_RELEASE;
6602d007 1318 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
d342894c 1319
1320 spin_lock_init(&vxlan->hash_lock);
1321
1322 init_timer_deferrable(&vxlan->age_timer);
1323 vxlan->age_timer.function = vxlan_cleanup;
1324 vxlan->age_timer.data = (unsigned long) vxlan;
1325
05f47d69 1326 inet_get_local_port_range(&low, &high);
1327 vxlan->port_min = low;
1328 vxlan->port_max = high;
823aa873 1329 vxlan->dst_port = htons(vxlan_port);
05f47d69 1330
d342894c 1331 vxlan->dev = dev;
1332
1333 for (h = 0; h < FDB_HASH_SIZE; ++h)
1334 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1335}
1336
1337static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
1338 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
5d174dd8 1339 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
d342894c 1340 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
1341 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
1342 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1343 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1344 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1345 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1346 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
05f47d69 1347 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
e4f67add
DS		 1348 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1349 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1350 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1351 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
823aa873 1352 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
d342894c 1353};
1354
1355static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
1356{
1357 if (tb[IFLA_ADDRESS]) {
1358 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1359 pr_debug("invalid link address (not ethernet)\n");
1360 return -EINVAL;
1361 }
1362
1363 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1364 pr_debug("invalid all zero ethernet address\n");
1365 return -EADDRNOTAVAIL;
1366 }
1367 }
1368
1369 if (!data)
1370 return -EINVAL;
1371
1372 if (data[IFLA_VXLAN_ID]) {
1373 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1374 if (id >= VXLAN_VID_MASK)
1375 return -ERANGE;
1376 }
1377
05f47d69 1378 if (data[IFLA_VXLAN_PORT_RANGE]) {
1379 const struct ifla_vxlan_port_range *p
1380 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1381
1382 if (ntohs(p->high) < ntohs(p->low)) {
1383 pr_debug("port range %u .. %u not valid\n",
1384 ntohs(p->low), ntohs(p->high));
1385 return -EINVAL;
1386 }
1387 }
1388
d342894c 1389 return 0;
1390}
1391
1b13c97f
YB		 1392static void vxlan_get_drvinfo(struct net_device *netdev,
1393 struct ethtool_drvinfo *drvinfo)
1394{
1395 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1396 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1397}
1398
1399static const struct ethtool_ops vxlan_ethtool_ops = {
1400 .get_drvinfo = vxlan_get_drvinfo,
1401 .get_link = ethtool_op_get_link,
1402};
1403
d342894c 1404static int vxlan_newlink(struct net *net, struct net_device *dev,
1405 struct nlattr *tb[], struct nlattr *data[])
1406{
1407 struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1408 struct vxlan_rdst *dst = &vxlan->default_dst;
d342894c 1409 __u32 vni;
1410 int err;
1411
1412 if (!data[IFLA_VXLAN_ID])
1413 return -EINVAL;
1414
1415 vni = nla_get_u32(data[IFLA_VXLAN_ID]);
1416 if (vxlan_find_vni(net, vni)) {
1417 pr_info("duplicate VNI %u\n", vni);
1418 return -EEXIST;
1419 }
c7995c43 1420 dst->remote_vni = vni;
d342894c 1421
5d174dd8 1422 if (data[IFLA_VXLAN_GROUP])
1423 dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]);
d342894c 1424
1425 if (data[IFLA_VXLAN_LOCAL])
1426 vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]);
1427
34e02aa1 1428 if (data[IFLA_VXLAN_LINK] &&
c7995c43 1429 (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) {
34e02aa1 1430 struct net_device *lowerdev
c7995c43 1431 = __dev_get_by_index(net, dst->remote_ifindex);
34e02aa1 1432
1433 if (!lowerdev) {
c7995c43 1434 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
34e02aa1 1435 return -ENODEV;
1436 }
d342894c 1437
34e02aa1 1438 if (!tb[IFLA_MTU])
d342894c 1439 dev->mtu = lowerdev->mtu - VXLAN_HEADROOM;
1ba56fb4
AD		 1440
1441 /* update header length based on lower device */
1442 dev->hard_header_len = lowerdev->hard_header_len +
1443 VXLAN_HEADROOM;
d342894c 1444 }
1445
1446 if (data[IFLA_VXLAN_TOS])
1447 vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
1448
afb97186
VB		 1449 if (data[IFLA_VXLAN_TTL])
1450 vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
1451
d342894c 1452 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
e4f67add 1453 vxlan->flags |= VXLAN_F_LEARN;
d342894c 1454
1455 if (data[IFLA_VXLAN_AGEING])
1456 vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
1457 else
1458 vxlan->age_interval = FDB_AGE_DEFAULT;
1459
e4f67add
DS		 1460 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
1461 vxlan->flags |= VXLAN_F_PROXY;
1462
1463 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
1464 vxlan->flags |= VXLAN_F_RSC;
1465
1466 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
1467 vxlan->flags |= VXLAN_F_L2MISS;
1468
1469 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
1470 vxlan->flags |= VXLAN_F_L3MISS;
1471
d342894c 1472 if (data[IFLA_VXLAN_LIMIT])
1473 vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
1474
05f47d69 1475 if (data[IFLA_VXLAN_PORT_RANGE]) {
1476 const struct ifla_vxlan_port_range *p
1477 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1478 vxlan->port_min = ntohs(p->low);
1479 vxlan->port_max = ntohs(p->high);
1480 }
1481
823aa873 1482 if (data[IFLA_VXLAN_PORT])
1483 vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
1484
1b13c97f
YB		 1485 SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops);
1486
d342894c 1487 err = register_netdevice(dev);
1488 if (!err)
c7995c43 1489 hlist_add_head_rcu(&vxlan->hlist, vni_head(net, dst->remote_vni));
d342894c 1490
1491 return err;
1492}
1493
1494static void vxlan_dellink(struct net_device *dev, struct list_head *head)
1495{
1496 struct vxlan_dev *vxlan = netdev_priv(dev);
1497
1498 hlist_del_rcu(&vxlan->hlist);
1499
1500 unregister_netdevice_queue(dev, head);
1501}
1502
1503static size_t vxlan_get_size(const struct net_device *dev)
1504{
1505
1506 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
5d174dd8 1507 nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */
d342894c 1508 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
1509 nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */
1510 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
1511 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
1512 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
e4f67add
DS		 1513 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
1514 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
1515 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
1516 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
d342894c 1517 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
1518 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
05f47d69 1519 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
823aa873 1520 nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */
d342894c 1521 0;
1522}
1523
1524static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
1525{
1526 const struct vxlan_dev *vxlan = netdev_priv(dev);
c7995c43 1527 const struct vxlan_rdst *dst = &vxlan->default_dst;
05f47d69 1528 struct ifla_vxlan_port_range ports = {
1529 .low = htons(vxlan->port_min),
1530 .high = htons(vxlan->port_max),
1531 };
d342894c 1532
c7995c43 1533 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
d342894c 1534 goto nla_put_failure;
1535
5d174dd8 1536 if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip))
d342894c 1537 goto nla_put_failure;
1538
c7995c43 1539 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
d342894c 1540 goto nla_put_failure;
1541
7c41c42c 1542 if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr))
d342894c 1543 goto nla_put_failure;
1544
1545 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) ||
1546 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) ||
e4f67add
DS		 1547 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
1548 !!(vxlan->flags & VXLAN_F_LEARN)) ||
1549 nla_put_u8(skb, IFLA_VXLAN_PROXY,
1550 !!(vxlan->flags & VXLAN_F_PROXY)) ||
1551 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
1552 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
1553 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
1554 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
1555 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
d342894c 1556 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) ||
823aa873 1557 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) ||
1558 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port))
d342894c 1559 goto nla_put_failure;
1560
05f47d69 1561 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
1562 goto nla_put_failure;
1563
d342894c 1564 return 0;
1565
1566nla_put_failure:
1567 return -EMSGSIZE;
1568}
1569
1570static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
1571 .kind = "vxlan",
1572 .maxtype = IFLA_VXLAN_MAX,
1573 .policy = vxlan_policy,
1574 .priv_size = sizeof(struct vxlan_dev),
1575 .setup = vxlan_setup,
1576 .validate = vxlan_validate,
1577 .newlink = vxlan_newlink,
1578 .dellink = vxlan_dellink,
1579 .get_size = vxlan_get_size,
1580 .fill_info = vxlan_fill_info,
1581};
1582
1583static __net_init int vxlan_init_net(struct net *net)
1584{
1585 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1586 struct sock *sk;
1587 struct sockaddr_in vxlan_addr = {
1588 .sin_family = AF_INET,
1589 .sin_addr.s_addr = htonl(INADDR_ANY),
1590 };
1591 int rc;
1592 unsigned h;
1593
1594 /* Create UDP socket for encapsulation receive. */
1595 rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vn->sock);
1596 if (rc < 0) {
1597 pr_debug("UDP socket create failed\n");
1598 return rc;
1599 }
bfe1b9b1 1600 /* Put in proper namespace */
1601 sk = vn->sock->sk;
1602 sk_change_net(sk, net);
d342894c 1603
1604 vxlan_addr.sin_port = htons(vxlan_port);
1605
1606 rc = kernel_bind(vn->sock, (struct sockaddr *) &vxlan_addr,
1607 sizeof(vxlan_addr));
1608 if (rc < 0) {
1609 pr_debug("bind for UDP socket %pI4:%u (%d)\n",
1610 &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc);
bfe1b9b1 1611 sk_release_kernel(sk);
d342894c 1612 vn->sock = NULL;
1613 return rc;
1614 }
1615
1616 /* Disable multicast loopback */
d342894c 1617 inet_sk(sk)->mc_loop = 0;
1618
1619 /* Mark socket as an encapsulation socket. */
1620 udp_sk(sk)->encap_type = 1;
1621 udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv;
1622 udp_encap_enable();
1623
1624 for (h = 0; h < VNI_HASH_SIZE; ++h)
1625 INIT_HLIST_HEAD(&vn->vni_list[h]);
1626
1627 return 0;
1628}
1629
1630static __net_exit void vxlan_exit_net(struct net *net)
1631{
1632 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
9cb6cb7e
ZM		 1633 struct vxlan_dev *vxlan;
1634 unsigned h;
1635
1636 rtnl_lock();
1637 for (h = 0; h < VNI_HASH_SIZE; ++h)
1638 hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist)
1639 dev_close(vxlan->dev);
1640 rtnl_unlock();
d342894c 1641
1642 if (vn->sock) {
bfe1b9b1 1643 sk_release_kernel(vn->sock->sk);
d342894c 1644 vn->sock = NULL;
1645 }
1646}
1647
1648static struct pernet_operations vxlan_net_ops = {
1649 .init = vxlan_init_net,
1650 .exit = vxlan_exit_net,
1651 .id = &vxlan_net_id,
1652 .size = sizeof(struct vxlan_net),
1653};
1654
1655static int __init vxlan_init_module(void)
1656{
1657 int rc;
1658
1659 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
1660
1661 rc = register_pernet_device(&vxlan_net_ops);
1662 if (rc)
1663 goto out1;
1664
1665 rc = rtnl_link_register(&vxlan_link_ops);
1666 if (rc)
1667 goto out2;
1668
1669 return 0;
1670
1671out2:
1672 unregister_pernet_device(&vxlan_net_ops);
1673out1:
1674 return rc;
1675}
1676module_init(vxlan_init_module);
1677
1678static void __exit vxlan_cleanup_module(void)
1679{
1680 rtnl_link_unregister(&vxlan_link_ops);
1681 unregister_pernet_device(&vxlan_net_ops);
6681712d 1682 rcu_barrier();
d342894c 1683}
1684module_exit(vxlan_cleanup_module);
1685
1686MODULE_LICENSE("GPL");
1687MODULE_VERSION(VXLAN_VERSION);
3b8df3c6 1688MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
d342894c 1689MODULE_ALIAS_RTNL_LINK("vxlan");
kernel source for telekom puls (alcatel/mediatek)