]> git.proxmox.com Git - mirror_ovs.git/blame - datapath/linux/compat/vxlan.c
projects / mirror_ovs.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
cirrus: Use FreeBSD 12.2.
[mirror_ovs.git] / datapath / linux / compat / vxlan.c
CommitLineData
1b7ee51f 1/*
e23775f2 2 * VXLAN: Virtual eXtensible Local Area Network
1b7ee51f 3 *
e23775f2 4 * Copyright (c) 2012-2013 Vyatta Inc.
1b7ee51f 5 *
e23775f2
PS		 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.
1b7ee51f
PS		 9 */
10
11#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13#include <linux/kernel.h>
14#include <linux/types.h>
15#include <linux/module.h>
16#include <linux/errno.h>
17#include <linux/slab.h>
18#include <linux/skbuff.h>
19#include <linux/rculist.h>
20#include <linux/netdevice.h>
f2252c61 21#include <linux/netdev_features.h>
1b7ee51f
PS		 22#include <linux/in.h>
23#include <linux/ip.h>
24#include <linux/udp.h>
25#include <linux/igmp.h>
26#include <linux/etherdevice.h>
27#include <linux/if_ether.h>
28#include <linux/if_vlan.h>
29#include <linux/hash.h>
30#include <linux/ethtool.h>
31#include <net/arp.h>
f2252c61 32#include <net/dst_metadata.h>
1b7ee51f
PS		 33#include <net/ndisc.h>
34#include <net/ip.h>
35#include <net/ip_tunnels.h>
36#include <net/icmp.h>
37#include <net/udp.h>
131971ce 38#include <net/udp_tunnel.h>
1b7ee51f
PS		 39#include <net/rtnetlink.h>
40#include <net/route.h>
41#include <net/dsfield.h>
42#include <net/inet_ecn.h>
43#include <net/net_namespace.h>
44#include <net/netns/generic.h>
e23775f2 45#include <net/protocol.h>
f2252c61 46
e23775f2
PS		 47#if IS_ENABLED(CONFIG_IPV6)
48#include <net/ipv6.h>
49#include <net/addrconf.h>
50#include <net/ip6_tunnel.h>
51#include <net/ip6_checksum.h>
f2252c61 52#include <net/ip6_route.h>
e23775f2 53#endif
1b7ee51f 54
b80837ac 55#include <net/tun_proto.h>
0643a78b 56#include <net/vxlan.h>
1b7ee51f 57#include "gso.h"
e23775f2 58#include "vport-netdev.h"
f2252c61 59#include "compat.h"
e23775f2 60
f2252c61 61#ifndef USE_UPSTREAM_TUNNEL
e23775f2
PS		 62#define VXLAN_VERSION "0.1"
63
64#define PORT_HASH_BITS 8
65#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
66#define FDB_AGE_DEFAULT 300 /* 5 min */
67#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
68
e23775f2
PS		 69/* UDP port for VXLAN traffic.
70 * The IANA assigned port is 4789, but the Linux default is 8472
71 * for compatibility with early adopters.
72 */
73static unsigned short vxlan_port __read_mostly = 8472;
74module_param_named(udp_port, vxlan_port, ushort, 0444);
75MODULE_PARM_DESC(udp_port, "Destination UDP port");
76
77static int vxlan_net_id;
78static struct rtnl_link_ops vxlan_link_ops;
1b7ee51f 79
f2252c61 80static const u8 all_zeros_mac[ETH_ALEN + 2];
b22a8d87 81
f2252c61 82static int vxlan_sock_add(struct vxlan_dev *vxlan);
e23775f2
PS		 83
84/* per-network namespace private data for this module */
85struct vxlan_net {
86 struct list_head vxlan_list;
87 struct hlist_head sock_list[PORT_HASH_SIZE];
88 spinlock_t sock_lock;
1b7ee51f
PS		 89};
90
e23775f2
PS		 91/* Forwarding table entry */
92struct vxlan_fdb {
93 struct hlist_node hlist; /* linked list of entries */
94 struct rcu_head rcu;
95 unsigned long updated; /* jiffies */
96 unsigned long used;
97 struct list_head remotes;
98 u8 eth_addr[ETH_ALEN];
99 u16 state; /* see ndm_state */
100 u8 flags; /* see ndm_flags */
101};
102
103/* salt for hash table */
104static u32 vxlan_salt __read_mostly;
e23775f2
PS		 105
106static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
107{
108 return vs->flags & VXLAN_F_COLLECT_METADATA ||
109 ip_tunnel_collect_metadata();
110}
111
112#if IS_ENABLED(CONFIG_IPV6)
113static inline
114bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
115{
116 if (a->sa.sa_family != b->sa.sa_family)
117 return false;
118 if (a->sa.sa_family == AF_INET6)
119 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
120 else
121 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
122}
123
124static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
125{
126 if (ipa->sa.sa_family == AF_INET6)
127 return ipv6_addr_any(&ipa->sin6.sin6_addr);
128 else
129 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
130}
131
132static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
133{
134 if (ipa->sa.sa_family == AF_INET6)
135 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
136 else
137 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
138}
139
140#else /* !CONFIG_IPV6 */
141
142static inline
143bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
144{
145 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
146}
147
148static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
149{
150 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
151}
152
153static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
154{
155 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
156}
e23775f2
PS		 157#endif
158
159/* Virtual Network hash table head */
f2252c61 160static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
e23775f2 161{
f2252c61 162 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
e23775f2
PS		 163}
164
165/* Socket hash table head */
166static inline struct hlist_head *vs_head(struct net *net, __be16 port)
167{
168 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
169
170 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
171}
172
e23775f2
PS		 173/* Find VXLAN socket based on network namespace, address family and UDP port
174 * and enabled unshareable flags.
175 */
176static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
177 __be16 port, u32 flags)
1b7ee51f 178{
1b7ee51f 179 struct vxlan_sock *vs;
1b7ee51f 180
e23775f2 181 flags &= VXLAN_F_RCV_FLAGS;
1b7ee51f 182
e23775f2 183 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
8063e095 184 if (inet_sk(vs->sock->sk)->inet_sport == port &&
e23775f2
PS		 185 vxlan_get_sk_family(vs) == family &&
186 vs->flags == flags)
187 return vs;
188 }
189 return NULL;
190}
ababf424 191
f2252c61 192static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni)
e23775f2
PS		 193{
194 struct vxlan_dev *vxlan;
195
f2252c61
PS		 196 /* For flow based devices, map all packets to VNI 0 */
197 if (vs->flags & VXLAN_F_COLLECT_METADATA)
198 vni = 0;
199
200 hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) {
201 if (vxlan->default_dst.remote_vni == vni)
e23775f2 202 return vxlan;
1b7ee51f
PS		 203 }
204
e23775f2
PS		 205 return NULL;
206}
1b7ee51f 207
e23775f2 208/* Look up VNI in a per net namespace table */
f2252c61 209static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni,
e23775f2
PS		 210 sa_family_t family, __be16 port,
211 u32 flags)
212{
213 struct vxlan_sock *vs;
214
215 vs = vxlan_find_sock(net, family, port, flags);
1b7ee51f 216 if (!vs)
e23775f2 217 return NULL;
1b7ee51f 218
f2252c61 219 return vxlan_vs_find_vni(vs, vni);
e23775f2 220}
3174a818 221
f2252c61
PS		 222static int vxlan_fdb_create(struct vxlan_dev *vxlan,
223 const u8 *mac, union vxlan_addr *ip,
224 __u16 state, __u16 flags,
0643a78b 225 __be16 port, __be32 vni, __u32 ifindex,
f2252c61 226 __u8 ndm_flags)
e23775f2
PS		 227{
228 return -EINVAL;
229}
3174a818 230
f2252c61
PS		 231static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
232{
233
234}
235
e23775f2
PS		 236static inline size_t vxlan_nlmsg_size(void)
237{
238 return NLMSG_ALIGN(sizeof(struct ndmsg))
239 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
240 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
241 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
242 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
243 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
244 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
245 + nla_total_size(sizeof(struct nda_cacheinfo));
246}
3174a818 247
e23775f2
PS		 248#ifdef HAVE_UDP_OFFLOAD
249#ifdef HAVE_NETIF_F_GSO_TUNNEL_REMCSUM
f2252c61 250
e23775f2
PS		 251static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
252 unsigned int off,
253 struct vxlanhdr *vh, size_t hdrlen,
f2252c61
PS		 254 __be32 vni_field,
255 struct gro_remcsum *grc,
e23775f2
PS		 256 bool nopartial)
257{
258 size_t start, offset;
259
260 if (skb->remcsum_offload)
261 return vh;
262
263 if (!NAPI_GRO_CB(skb)->csum_valid)
264 return NULL;
265
f2252c61
PS		 266 start = vxlan_rco_start(vni_field);
267 offset = start + vxlan_rco_offset(vni_field);
e23775f2
PS		 268
269 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
270 start, offset, grc, nopartial);
271
272 skb->remcsum_offload = 1;
273
274 return vh;
275}
276#else
277static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
f2252c61
PS		 278 unsigned int off,
279 struct vxlanhdr *vh, size_t hdrlen,
280 u32 data, struct gro_remcsum *grc,
281 bool nopartial)
e23775f2
PS		 282{
283 return NULL;
3174a818 284}
e23775f2 285#endif
3174a818 286
e23775f2
PS		 287#ifndef HAVE_UDP_OFFLOAD_ARG_UOFF
288static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
f2252c61 289 struct sk_buff *skb)
e23775f2
PS		 290#else
291static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
f2252c61
PS		 292 struct sk_buff *skb,
293 struct udp_offload *uoff)
e23775f2 294#endif
1b7ee51f 295{
e23775f2
PS		 296#ifdef HAVE_UDP_OFFLOAD_ARG_UOFF
297 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
f2252c61 298 udp_offloads);
e23775f2
PS		 299#else
300 struct vxlan_sock *vs = NULL;
301#endif
302 struct sk_buff *p, **pp = NULL;
303 struct vxlanhdr *vh, *vh2;
304 unsigned int hlen, off_vx;
305 int flush = 1;
f2252c61 306 __be32 flags;
e23775f2 307 struct gro_remcsum grc;
1b7ee51f 308
e23775f2 309 skb_gro_remcsum_init(&grc);
1b7ee51f 310
e23775f2
PS		 311 off_vx = skb_gro_offset(skb);
312 hlen = off_vx + sizeof(*vh);
313 vh = skb_gro_header_fast(skb, off_vx);
314 if (skb_gro_header_hard(skb, hlen)) {
315 vh = skb_gro_header_slow(skb, hlen, off_vx);
316 if (unlikely(!vh))
317 goto out;
93258bd7 318 }
1b7ee51f 319
e23775f2 320 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
1b7ee51f 321
f2252c61 322 flags = vh->vx_flags;
3174a818 323
9989631e 324 if ((flags & VXLAN_HF_RCO) && vs && (vs->flags & VXLAN_F_REMCSUM_RX)) {
e23775f2 325 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
f2252c61 326 vh->vx_vni, &grc,
e23775f2
PS		 327 !!(vs->flags &
328 VXLAN_F_REMCSUM_NOPARTIAL));
1b7ee51f 329
e23775f2
PS		 330 if (!vh)
331 goto out;
332 }
1b7ee51f 333
e23775f2 334 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
e4bafe59 335
e23775f2
PS		 336 for (p = *head; p; p = p->next) {
337 if (!NAPI_GRO_CB(p)->same_flow)
338 continue;
a109c9fb 339
e23775f2
PS		 340 vh2 = (struct vxlanhdr *)(p->data + off_vx);
341 if (vh->vx_flags != vh2->vx_flags ||
342 vh->vx_vni != vh2->vx_vni) {
343 NAPI_GRO_CB(p)->same_flow = 0;
344 continue;
345 }
346 }
347
348 pp = eth_gro_receive(head, skb);
f2252c61 349 flush = 0;
e23775f2
PS		 350
351out:
352 skb_gro_remcsum_cleanup(skb, &grc);
353 NAPI_GRO_CB(skb)->flush |= flush;
354
355 return pp;
a109c9fb
PS		 356}
357
e23775f2
PS		 358#ifndef HAVE_UDP_OFFLOAD_ARG_UOFF
359static int vxlan_gro_complete(struct sk_buff *skb, int nhoff)
360#else
361static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
f2252c61 362 struct udp_offload *uoff)
e23775f2 363#endif
a109c9fb 364{
cc3af88e
PS		 365 /* Sets 'skb->inner_mac_header' since we are always called with
366 * 'skb->encapsulation' set.
367 */
e23775f2 368 udp_tunnel_gro_complete(skb, nhoff);
a109c9fb 369
e23775f2 370 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
a109c9fb 371}
f2252c61 372#endif
a109c9fb 373
e23775f2
PS		 374/* Notify netdevs that UDP port started listening */
375static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
131971ce 376{
e23775f2
PS		 377 struct net_device *dev;
378 struct sock *sk = vs->sock->sk;
379 struct net *net = sock_net(sk);
380 sa_family_t sa_family = vxlan_get_sk_family(vs);
131971ce 381
f2252c61
PS		 382
383 if (sa_family == AF_INET) {
384 int err;
385
edf6cd31 386 err = udp_add_offload(net, &vs->udp_offloads);
f2252c61
PS		 387 if (err)
388 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
389 }
390
edf6cd31
PS		 391 rcu_read_lock();
392 for_each_netdev_rcu(net, dev) {
393#ifdef HAVE_NDO_ADD_VXLAN_PORT
394 __be16 port = inet_sk(sk)->inet_sport;
395
396 if (dev->netdev_ops->ndo_add_vxlan_port)
397 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
398 port);
e2e11c89 399#elif defined(HAVE_NDO_UDP_TUNNEL_ADD)
400 struct udp_tunnel_info ti;
401 if (vs->flags & VXLAN_F_GPE)
402 ti.type = UDP_TUNNEL_TYPE_VXLAN_GPE;
403 else
404 ti.type = UDP_TUNNEL_TYPE_VXLAN;
405 ti.sa_family = sa_family;
406 ti.port = inet_sk(sk)->inet_sport;
407
408 if (dev->netdev_ops->ndo_udp_tunnel_add)
409 dev->netdev_ops->ndo_udp_tunnel_add(dev, &ti);
f2252c61 410#endif
edf6cd31
PS		 411 }
412 rcu_read_unlock();
e23775f2 413}
131971ce 414
e23775f2
PS		 415/* Notify netdevs that UDP port is no more listening */
416static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
417{
418 struct net_device *dev;
419 struct sock *sk = vs->sock->sk;
420 struct net *net = sock_net(sk);
421 sa_family_t sa_family = vxlan_get_sk_family(vs);
131971ce 422
e23775f2
PS		 423 rcu_read_lock();
424 for_each_netdev_rcu(net, dev) {
edf6cd31
PS		 425#ifdef HAVE_NDO_ADD_VXLAN_PORT
426 __be16 port = inet_sk(sk)->inet_sport;
427
e23775f2
PS		 428 if (dev->netdev_ops->ndo_del_vxlan_port)
429 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
edf6cd31 430 port);
e2e11c89 431#elif defined(HAVE_NDO_UDP_TUNNEL_ADD)
432 struct udp_tunnel_info ti;
433 if (vs->flags & VXLAN_F_GPE)
434 ti.type = UDP_TUNNEL_TYPE_VXLAN_GPE;
435 else
436 ti.type = UDP_TUNNEL_TYPE_VXLAN;
437 ti.port = inet_sk(sk)->inet_sport;
438 ti.sa_family = sa_family;
439
440 if (dev->netdev_ops->ndo_udp_tunnel_del)
441 dev->netdev_ops->ndo_udp_tunnel_del(dev, &ti);
edf6cd31 442#endif
e23775f2
PS		 443 }
444 rcu_read_unlock();
131971ce 445
f2252c61 446 if (sa_family == AF_INET) {
e23775f2 447 udp_del_offload(&vs->udp_offloads);
edf6cd31 448 }
1b7ee51f
PS		 449}
450
e23775f2
PS		 451/* See if multicast group is already in use by other ID */
452static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1b7ee51f 453{
e23775f2 454 struct vxlan_dev *vxlan;
76e671ea
PS		 455 struct vxlan_sock *sock4;
456 struct vxlan_sock *sock6 = NULL;
f2252c61 457 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
e23775f2 458
76e671ea
PS		 459 sock4 = rtnl_dereference(dev->vn4_sock);
460
e23775f2
PS		 461 /* The vxlan_sock is only used by dev, leaving group has
462 * no effect on other vxlan devices.
463 */
76e671ea 464 if (family == AF_INET && sock4 && atomic_read(&sock4->refcnt) == 1)
f2252c61
PS		 465 return false;
466#if IS_ENABLED(CONFIG_IPV6)
76e671ea
PS		 467 sock6 = rtnl_dereference(dev->vn6_sock);
468 if (family == AF_INET6 && sock6 && atomic_read(&sock6->refcnt) == 1)
e23775f2 469 return false;
f2252c61 470#endif
1b7ee51f 471
e23775f2
PS		 472 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
473 if (!netif_running(vxlan->dev) || vxlan == dev)
474 continue;
475
76e671ea
PS		 476 if (family == AF_INET &&
477 rtnl_dereference(vxlan->vn4_sock) != sock4)
e23775f2 478 continue;
f2252c61 479#if IS_ENABLED(CONFIG_IPV6)
76e671ea
PS		 480 if (family == AF_INET6 &&
481 rtnl_dereference(vxlan->vn6_sock) != sock6)
f2252c61
PS		 482 continue;
483#endif
e23775f2
PS		 484
485 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
486 &dev->default_dst.remote_ip))
487 continue;
488
489 if (vxlan->default_dst.remote_ifindex !=
490 dev->default_dst.remote_ifindex)
491 continue;
492
493 return true;
494 }
495
496 return false;
1b7ee51f 497}
29c71cfa 498
f2252c61 499static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
e23775f2 500{
f2252c61 501 struct vxlan_net *vn;
e23775f2 502
f2252c61
PS		 503 if (!vs)
504 return false;
e23775f2 505 if (!atomic_dec_and_test(&vs->refcnt))
f2252c61 506 return false;
e23775f2 507
f2252c61 508 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
e23775f2
PS		 509 spin_lock(&vn->sock_lock);
510 hlist_del_rcu(&vs->hlist);
e23775f2 511 vxlan_notify_del_rx_port(vs);
e23775f2
PS		 512 spin_unlock(&vn->sock_lock);
513
f2252c61
PS		 514 return true;
515}
516
517static void vxlan_sock_release(struct vxlan_dev *vxlan)
518{
76e671ea 519 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
f2252c61 520#if IS_ENABLED(CONFIG_IPV6)
76e671ea
PS		 521 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
522
523 rcu_assign_pointer(vxlan->vn6_sock, NULL);
f2252c61
PS		 524#endif
525
76e671ea 526 rcu_assign_pointer(vxlan->vn4_sock, NULL);
f2252c61
PS		 527 synchronize_net();
528
76e671ea
PS		 529 if (__vxlan_sock_release_prep(sock4)) {
530 udp_tunnel_sock_release(sock4->sock);
531 kfree(sock4);
f2252c61
PS		 532 }
533
534#if IS_ENABLED(CONFIG_IPV6)
76e671ea
PS		 535 if (__vxlan_sock_release_prep(sock6)) {
536 udp_tunnel_sock_release(sock6->sock);
537 kfree(sock6);
f2252c61
PS		 538 }
539#endif
e23775f2
PS		 540}
541
542/* Update multicast group membership when first VNI on
543 * multicast address is brought up
544 */
545static int vxlan_igmp_join(struct vxlan_dev *vxlan)
546{
547 return -EINVAL;
548}
549
550/* Inverse of vxlan_igmp_join when last VNI is brought down */
551static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
552{
553 return -EINVAL;
554}
555
f2252c61
PS		 556static bool vxlan_remcsum(struct vxlanhdr *unparsed,
557 struct sk_buff *skb, u32 vxflags)
e23775f2 558{
f2252c61
PS		 559#ifndef USE_UPSTREAM_TUNNEL
560 return false;
561#else
562 size_t start, offset;
e23775f2 563
f2252c61
PS		 564 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
565 goto out;
e23775f2 566
f2252c61
PS		 567 start = vxlan_rco_start(unparsed->vx_vni);
568 offset = start + vxlan_rco_offset(unparsed->vx_vni);
e23775f2 569
f2252c61
PS		 570 if (!pskb_may_pull(skb, offset + sizeof(u16)))
571 return false;
e23775f2 572
f2252c61
PS		 573 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
574 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
575out:
576 unparsed->vx_flags &= ~VXLAN_HF_RCO;
577 unparsed->vx_vni &= VXLAN_VNI_MASK;
578 return true;
e23775f2 579#endif
f2252c61 580}
e23775f2 581
f2252c61
PS		 582static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
583 struct sk_buff *skb, u32 vxflags,
584 struct vxlan_metadata *md)
e23775f2 585{
f2252c61
PS		 586 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
587 struct metadata_dst *tun_dst;
e23775f2 588
f2252c61
PS		 589 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
590 goto out;
e23775f2 591
f2252c61 592 md->gbp = ntohs(gbp->policy_id);
e23775f2 593
f2252c61 594 tun_dst = (struct metadata_dst *)skb_dst(skb);
e23775f2 595 if (tun_dst) {
f2252c61
PS		 596 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
597 tun_dst->u.tun_info.options_len = sizeof(*md);
e23775f2 598 }
f2252c61
PS		 599 if (gbp->dont_learn)
600 md->gbp |= VXLAN_GBP_DONT_LEARN;
e23775f2 601
f2252c61
PS		 602 if (gbp->policy_applied)
603 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
e23775f2 604
e23775f2 605 /* In flow-based mode, GBP is carried in dst_metadata */
f2252c61 606 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
e23775f2 607 skb->mark = md->gbp;
f2252c61
PS		 608out:
609 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
610}
e23775f2 611
f2252c61
PS		 612static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
613 __be16 *protocol,
614 struct sk_buff *skb, u32 vxflags)
615{
616 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
e23775f2 617
f2252c61
PS		 618 /* Need to have Next Protocol set for interfaces in GPE mode. */
619 if (!gpe->np_applied)
620 return false;
621 /* "The initial version is 0. If a receiver does not support the
622 * version indicated it MUST drop the packet.
623 */
624 if (gpe->version != 0)
625 return false;
626 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
627 * processing MUST occur." However, we don't implement OAM
628 * processing, thus drop the packet.
629 */
630 if (gpe->oam_flag)
631 return false;
632
b80837ac
YY		 633 *protocol = tun_p_to_eth_p(gpe->next_protocol);
634 if (!*protocol)
f2252c61 635 return false;
e23775f2 636
f2252c61
PS		 637 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
638 return true;
639}
e23775f2 640
f2252c61
PS		 641static bool vxlan_set_mac(struct vxlan_dev *vxlan,
642 struct vxlan_sock *vs,
643 struct sk_buff *skb)
644{
645 return true;
646}
647
648static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
649 struct sk_buff *skb)
650{
651 int err = 0;
652
653 if (vxlan_get_sk_family(vs) == AF_INET)
654 err = IP_ECN_decapsulate(oiph, skb);
655#if IS_ENABLED(CONFIG_IPV6)
656 else
657 err = IP6_ECN_decapsulate(oiph, skb);
658#endif
659 return err <= 1;
e23775f2
PS		 660}
661
662/* Callback from net/ipv4/udp.c to receive packets */
f2252c61 663static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
e23775f2 664{
e23775f2
PS		 665 union {
666 struct metadata_dst dst;
f2252c61 667 char buf[sizeof(struct metadata_dst) + sizeof(struct vxlan_metadata)];
e23775f2
PS		 668 } buf;
669
f2252c61
PS		 670 struct pcpu_sw_netstats *stats;
671 struct vxlan_dev *vxlan;
672 struct vxlan_sock *vs;
673 struct vxlanhdr unparsed;
674 struct vxlan_metadata _md;
675 struct vxlan_metadata *md = &_md;
676 __be16 protocol = htons(ETH_P_TEB);
677 bool raw_proto = false;
678 void *oiph;
e23775f2 679
f2252c61
PS		 680 /* Need UDP and VXLAN header to be present */
681 if (!pskb_may_pull(skb, VXLAN_HLEN))
2b3397eb 682 goto drop;
f2252c61
PS		 683
684 unparsed = *vxlan_hdr(skb);
685 /* VNI flag always required to be set */
686 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
687 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
688 ntohl(vxlan_hdr(skb)->vx_flags),
689 ntohl(vxlan_hdr(skb)->vx_vni));
690 /* Return non vxlan pkt */
2b3397eb 691 goto drop;
e23775f2
PS		 692 }
693
f2252c61
PS		 694 unparsed.vx_flags &= ~VXLAN_HF_VNI;
695 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
e23775f2
PS		 696
697 vs = rcu_dereference_sk_user_data(sk);
698 if (!vs)
699 goto drop;
700
f2252c61
PS		 701#if IS_ENABLED(CONFIG_IPV6)
702#ifdef OVS_CHECK_UDP_TUNNEL_ZERO_CSUM
703 if (vxlan_get_sk_family(vs) == AF_INET6 &&
704 !udp_hdr(skb)->check &&
705 !(vs->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) {
706 udp6_csum_zero_error(skb);
707 goto drop;
e23775f2
PS		 708 }
709#endif
f2252c61
PS		 710#endif
711 vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni));
712 if (!vxlan)
713 goto drop;
e23775f2
PS		 714
715 /* For backwards compatibility, only allow reserved fields to be
716 * used by VXLAN extensions if explicitly requested.
717 */
f2252c61
PS		 718 if (vs->flags & VXLAN_F_GPE) {
719 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
720 goto drop;
721 raw_proto = true;
722 }
e23775f2 723
f2252c61
PS		 724 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
725 !net_eq(vxlan->net, dev_net(vxlan->dev))))
726 goto drop;
727
728 if (vxlan_collect_metadata(vs)) {
729 __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
730 struct metadata_dst *tun_dst;
e23775f2 731
f2252c61 732 tun_dst = &buf.dst;
a5d59bf3 733 ovs_udp_tun_rx_dst(tun_dst, skb,
f2252c61
PS		 734 vxlan_get_sk_family(vs), TUNNEL_KEY,
735 vxlan_vni_to_tun_id(vni), sizeof(*md));
e23775f2 736
f2252c61
PS		 737 if (!tun_dst)
738 goto drop;
e23775f2 739
f2252c61 740 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
e23775f2 741
f2252c61
PS		 742 ovs_skb_dst_set(skb, (struct dst_entry *)tun_dst);
743 } else {
744 memset(md, 0, sizeof(*md));
e23775f2
PS		 745 }
746
f2252c61
PS		 747 if (vs->flags & VXLAN_F_REMCSUM_RX)
748 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
749 goto drop;
750
751 if (vs->flags & VXLAN_F_GBP)
752 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
753 /* Note that GBP and GPE can never be active together. This is
754 * ensured in vxlan_dev_configure.
755 */
756
757 if (unparsed.vx_flags || unparsed.vx_vni) {
e23775f2
PS		 758 /* If there are any unprocessed flags remaining treat
759 * this as a malformed packet. This behavior diverges from
760 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
761 * in reserved fields are to be ignored. The approach here
762 * maintains compatibility with previous stack code, and also
763 * is more robust and provides a little more security in
764 * adding extensions to VXLAN.
765 */
f2252c61
PS		 766 goto drop;
767 }
768
769 if (!raw_proto) {
770 if (!vxlan_set_mac(vxlan, vs, skb))
771 goto drop;
772 skb_reset_mac_header(skb);
773 skb->protocol = eth_type_trans(skb, vxlan->dev);
774 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
775 } else {
776 skb_reset_mac_header(skb);
777 skb->dev = vxlan->dev;
778 skb->pkt_type = PACKET_HOST;
779 }
780
781 oiph = skb_network_header(skb);
782 skb_reset_network_header(skb);
e23775f2 783
f2252c61
PS		 784 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
785 ++vxlan->dev->stats.rx_frame_errors;
786 ++vxlan->dev->stats.rx_errors;
787 goto drop;
e23775f2
PS		 788 }
789
f2252c61
PS		 790 stats = this_cpu_ptr(vxlan->dev->tstats);
791 u64_stats_update_begin(&stats->syncp);
792 stats->rx_packets++;
793 stats->rx_bytes += skb->len;
794 u64_stats_update_end(&stats->syncp);
795
796 netdev_port_receive(skb, skb_tunnel_info(skb));
e23775f2
PS		 797 return 0;
798
799drop:
800 /* Consume bad packet */
801 kfree_skb(skb);
802 return 0;
e23775f2
PS		 803}
804
805static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
806 struct vxlan_metadata *md)
807{
808 struct vxlanhdr_gbp *gbp;
809
810 if (!md->gbp)
811 return;
812
813 gbp = (struct vxlanhdr_gbp *)vxh;
f2252c61 814 vxh->vx_flags |= VXLAN_HF_GBP;
e23775f2
PS		 815
816 if (md->gbp & VXLAN_GBP_DONT_LEARN)
817 gbp->dont_learn = 1;
818
819 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
820 gbp->policy_applied = 1;
821
822 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
823}
824
f2252c61
PS		 825static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
826 __be16 protocol)
e23775f2 827{
f2252c61 828 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
e23775f2 829
f2252c61 830 gpe->np_applied = 1;
b80837ac
YY		 831 gpe->next_protocol = tun_p_from_eth_p(protocol);
832 if (!gpe->next_protocol)
833 return -EPFNOSUPPORT;
834 return 0;
e23775f2 835}
e23775f2 836
f2252c61
PS		 837static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
838 int iphdr_len, __be32 vni,
839 struct vxlan_metadata *md, u32 vxflags,
840 bool udp_sum)
e23775f2 841{
cf5789aa 842 void (*fix_segment)(struct sk_buff *);
e23775f2
PS		 843 struct vxlanhdr *vxh;
844 int min_headroom;
845 int err;
f2252c61
PS		 846 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
847 __be16 inner_protocol = htons(ETH_P_TEB);
e23775f2
PS		 848
849 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
850 skb->ip_summed == CHECKSUM_PARTIAL) {
851 int csum_start = skb_checksum_start_offset(skb);
852
853 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
854 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
855 (skb->csum_offset == offsetof(struct udphdr, check) ||
f2252c61 856 skb->csum_offset == offsetof(struct tcphdr, check)))
e23775f2 857 type |= SKB_GSO_TUNNEL_REMCSUM;
e23775f2
PS		 858 }
859
f2252c61
PS		 860 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
861 + VXLAN_HLEN + iphdr_len
e23775f2
PS		 862 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
863
864 /* Need space for new headers (invalidates iph ptr) */
865 err = skb_cow_head(skb, min_headroom);
f2252c61
PS		 866 if (unlikely(err))
867 goto out_free;
e23775f2 868
436d36db
GR		 869 if (skb_vlan_tag_present(skb))
870 skb = __vlan_hwaccel_push_inside(skb);
e23775f2
PS		 871 if (WARN_ON(!skb))
872 return -ENOMEM;
873
cf5789aa 874 type |= udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
b416d163 875#ifndef USE_UPSTREAM_TUNNEL_GSO
db4b6975 876 fix_segment = !udp_sum ? ovs_udp_gso : ovs_udp_csum_gso;
b416d163
PS		 877#else
878 fix_segment = NULL;
879#endif
2a94b571 880 err = ovs_iptunnel_handle_offloads(skb, type, fix_segment);
f2252c61
PS		 881 if (err)
882 goto out_free;
883
e23775f2 884 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
f2252c61
PS		 885 vxh->vx_flags = VXLAN_HF_VNI;
886 vxh->vx_vni = vxlan_vni_field(vni);
e23775f2
PS		 887
888 if (type & SKB_GSO_TUNNEL_REMCSUM) {
f2252c61 889 unsigned int start;
e23775f2 890
f2252c61
PS		 891 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
892 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
893 vxh->vx_flags |= VXLAN_HF_RCO;
e23775f2
PS		 894
895 if (!skb_is_gso(skb)) {
896 skb->ip_summed = CHECKSUM_NONE;
e23775f2 897 skb->encapsulation = 0;
e23775f2
PS		 898 }
899 }
f2252c61 900
e23775f2
PS		 901 if (vxflags & VXLAN_F_GBP)
902 vxlan_build_gbp_hdr(vxh, vxflags, md);
f2252c61
PS		 903 if (vxflags & VXLAN_F_GPE) {
904 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
905 if (err < 0)
906 goto out_free;
907 inner_protocol = skb->protocol;
908 }
e23775f2 909
f2252c61 910 ovs_skb_set_inner_protocol(skb, inner_protocol);
43dd2fce 911 return 0;
f2252c61
PS		 912
913out_free:
914 kfree_skb(skb);
915 return err;
916}
917
918static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan,
919 struct sk_buff *skb, int oif, u8 tos,
920 __be32 daddr, __be32 *saddr,
d69a9596 921 __be16 dport, __be16 sport,
e02c0ed7 922 struct dst_cache *dst_cache,
f2252c61
PS		 923 const struct ip_tunnel_info *info)
924{
ec7d36ff 925 bool use_cache = (dst_cache && ip_tunnel_dst_cache_usable(skb, info));
f2252c61
PS		 926 struct rtable *rt = NULL;
927 struct flowi4 fl4;
928
e02c0ed7
PS		 929 if (tos && !info)
930 use_cache = false;
931 if (use_cache) {
932 rt = dst_cache_get_ip4(dst_cache, saddr);
933 if (rt)
934 return rt;
935 }
936
f2252c61
PS		 937 memset(&fl4, 0, sizeof(fl4));
938 fl4.flowi4_oif = oif;
939 fl4.flowi4_tos = RT_TOS(tos);
940 fl4.flowi4_mark = skb->mark;
941 fl4.flowi4_proto = IPPROTO_UDP;
942 fl4.daddr = daddr;
885f0013 943 fl4.saddr = *saddr;
d69a9596
QX		 944 fl4.fl4_dport = dport;
945 fl4.fl4_sport = sport;
f2252c61
PS		 946
947 rt = ip_route_output_key(vxlan->net, &fl4);
948 if (!IS_ERR(rt)) {
949 *saddr = fl4.saddr;
e02c0ed7
PS		 950 if (use_cache)
951 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
f2252c61
PS		 952 }
953 return rt;
954}
955
956#if IS_ENABLED(CONFIG_IPV6)
957static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
958 struct sk_buff *skb, int oif, u8 tos,
959 __be32 label,
960 const struct in6_addr *daddr,
961 struct in6_addr *saddr,
d69a9596 962 __be16 dport, __be16 sport,
e02c0ed7 963 struct dst_cache *dst_cache,
f2252c61
PS		 964 const struct ip_tunnel_info *info)
965{
76e671ea 966 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
ec7d36ff 967 bool use_cache = (dst_cache && ip_tunnel_dst_cache_usable(skb, info));
f2252c61
PS		 968 struct dst_entry *ndst;
969 struct flowi6 fl6;
33f9c873
GR		 970#if !defined(HAVE_IPV6_STUB_WITH_DST_ENTRY) || \
971 !defined(HAVE_IPV6_DST_LOOKUP_FLOW)
f2252c61 972 int err;
33f9c873 973#endif
f2252c61 974
76e671ea
PS		 975 if (!sock6)
976 return ERR_PTR(-EIO);
977
e02c0ed7
PS		 978 if (tos && !info)
979 use_cache = false;
980 if (use_cache) {
981 ndst = dst_cache_get_ip6(dst_cache, saddr);
982 if (ndst)
983 return ndst;
984 }
985
f2252c61
PS		 986 memset(&fl6, 0, sizeof(fl6));
987 fl6.flowi6_oif = oif;
988 fl6.daddr = *daddr;
885f0013 989 fl6.saddr = *saddr;
f2252c61
PS		 990 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
991 fl6.flowi6_mark = skb->mark;
992 fl6.flowi6_proto = IPPROTO_UDP;
d69a9596
QX		 993 fl6.fl6_dport = dport;
994 fl6.fl6_sport = sport;
f2252c61 995
33f9c873
GR		 996#if defined(HAVE_IPV6_STUB_WITH_DST_ENTRY) && defined(HAVE_IPV6_DST_LOOKUP_FLOW)
997#ifdef HAVE_IPV6_DST_LOOKUP_FLOW_NET
998 ndst = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
999 &fl6, NULL);
1000#else
1001 ndst = ipv6_stub->ipv6_dst_lookup_flow(sock6->sock->sk, &fl6, NULL);
1002#endif
1003 if (unlikely(IS_ERR(ndst))) {
1004#elif defined(HAVE_IPV6_DST_LOOKUP_FLOW_NET)
5519e384
YHW		 1005 err = ipv6_stub->ipv6_dst_lookup_flow(vxlan->net, sock6->sock->sk,
1006 &ndst, &fl6);
1007#elif defined(HAVE_IPV6_DST_LOOKUP_FLOW)
1008 err = ipv6_stub->ipv6_dst_lookup_flow(sock6->sock->sk, &ndst, &fl6);
1009#elif defined(HAVE_IPV6_DST_LOOKUP_NET)
1010 err = ipv6_stub->ipv6_dst_lookup(vxlan->net, sock6->sock->sk,
f2252c61 1011 &ndst, &fl6);
5519e384 1012#elif defined(HAVE_IPV6_STUB)
f2252c61
PS		 1013 err = ipv6_stub->ipv6_dst_lookup(vxlan->vn6_sock->sock->sk,
1014 &ndst, &fl6);
1015#else
1016 err = ip6_dst_lookup(vxlan->vn6_sock->sock->sk, &ndst, &fl6);
f2252c61 1017#endif
33f9c873
GR		 1018#if defined(HAVE_IPV6_STUB_WITH_DST_ENTRY) && defined(HAVE_IPV6_DST_LOOKUP_FLOW)
1019 return ERR_PTR(-ENETUNREACH);
1020 }
1021#else
f2252c61
PS		 1022 if (err < 0)
1023 return ERR_PTR(err);
33f9c873 1024#endif
f2252c61
PS		 1025
1026 *saddr = fl6.saddr;
e02c0ed7
PS		 1027 if (use_cache)
1028 dst_cache_set_ip6(dst_cache, ndst, saddr);
f2252c61
PS		 1029 return ndst;
1030}
1031#endif
1032
1033/* Bypass encapsulation if the destination is local */
1034static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1035 struct vxlan_dev *dst_vxlan)
1036{
1037 skb->dev->stats.rx_dropped++;
1038 kfree_skb(skb);
e23775f2
PS		 1039}
1040
1041static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1042 struct vxlan_rdst *rdst, bool did_rsc)
1043{
e02c0ed7 1044 struct dst_cache *dst_cache;
e23775f2
PS		 1045 struct ip_tunnel_info *info;
1046 struct vxlan_dev *vxlan = netdev_priv(dev);
f2252c61 1047 struct sock *sk;
e23775f2
PS		 1048 struct rtable *rt = NULL;
1049 const struct iphdr *old_iph;
e23775f2 1050 union vxlan_addr *dst;
885f0013
PS		 1051 union vxlan_addr remote_ip, local_ip;
1052 union vxlan_addr *src;
e23775f2
PS		 1053 struct vxlan_metadata _md;
1054 struct vxlan_metadata *md = &_md;
1055 __be16 src_port = 0, dst_port;
f2252c61 1056 __be32 vni, label;
e23775f2
PS		 1057 __be16 df = 0;
1058 __u8 tos, ttl;
1059 int err;
1060 u32 flags = vxlan->flags;
f2252c61
PS		 1061 bool udp_sum = false;
1062 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
e23775f2
PS		 1063
1064 info = skb_tunnel_info(skb);
1065
1066 if (rdst) {
1067 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
1068 vni = rdst->remote_vni;
1069 dst = &rdst->remote_ip;
885f0013 1070 src = &vxlan->cfg.saddr;
e02c0ed7 1071 dst_cache = &rdst->dst_cache;
e23775f2
PS		 1072 } else {
1073 if (!info) {
1074 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1075 dev->name);
1076 goto drop;
1077 }
e23775f2 1078 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
f2252c61
PS		 1079 vni = vxlan_tun_id_to_vni(info->key.tun_id);
1080 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
885f0013 1081 if (remote_ip.sa.sa_family == AF_INET) {
e23775f2 1082 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
885f0013
PS		 1083 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
1084 } else {
e23775f2 1085 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
885f0013
PS		 1086 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
1087 }
e23775f2 1088 dst = &remote_ip;
885f0013 1089 src = &local_ip;
e02c0ed7 1090 dst_cache = &info->dst_cache;
e23775f2
PS		 1091 }
1092
1093 if (vxlan_addr_any(dst)) {
1094 if (did_rsc) {
1095 /* short-circuited back to local bridge */
f2252c61
PS		 1096 vxlan_encap_bypass(skb, vxlan, vxlan);
1097 return;
e23775f2
PS		 1098 }
1099 goto drop;
1100 }
1101
1102 old_iph = ip_hdr(skb);
1103
1104 ttl = vxlan->cfg.ttl;
1105 if (!ttl && vxlan_addr_multicast(dst))
1106 ttl = 1;
1107
1108 tos = vxlan->cfg.tos;
1109 if (tos == 1)
1110 tos = ip_tunnel_get_dsfield(old_iph, skb);
1111
3259c4ff 1112 label = vxlan->cfg.label;
e23775f2
PS		 1113 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1114 vxlan->cfg.port_max, true);
1115
1116 if (info) {
e23775f2
PS		 1117 ttl = info->key.ttl;
1118 tos = info->key.tos;
3259c4ff 1119 label = info->key.label;
f2252c61 1120 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
e23775f2 1121
3378775c
PJV		 1122 if (info->options_len &&
1123 info->key.tun_flags & TUNNEL_VXLAN_OPT)
e23775f2
PS		 1124 md = ip_tunnel_info_opts(info);
1125 } else {
1126 md->gbp = skb->mark;
1127 }
1128
1129 if (dst->sa.sa_family == AF_INET) {
76e671ea
PS		 1130 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
1131
1132 if (!sock4)
f2252c61 1133 goto drop;
76e671ea 1134 sk = sock4->sock->sk;
e23775f2 1135
f2252c61
PS		 1136 rt = vxlan_get_route(vxlan, skb,
1137 rdst ? rdst->remote_ifindex : 0, tos,
885f0013
PS		 1138 dst->sin.sin_addr.s_addr,
1139 &src->sin.sin_addr.s_addr,
d69a9596 1140 dst_port, src_port,
e02c0ed7 1141 dst_cache, info);
e23775f2
PS		 1142 if (IS_ERR(rt)) {
1143 netdev_dbg(dev, "no route to %pI4\n",
1144 &dst->sin.sin_addr.s_addr);
1145 dev->stats.tx_carrier_errors++;
1146 goto tx_error;
1147 }
1148
8063e095 1149 if (rt->dst.dev == dev) {
e23775f2
PS		 1150 netdev_dbg(dev, "circular route to %pI4\n",
1151 &dst->sin.sin_addr.s_addr);
1152 dev->stats.collisions++;
1153 goto rt_tx_error;
1154 }
1155
1156 /* Bypass encapsulation if the destination is local */
b1c74f35 1157 if (!info && rt->rt_flags & RTCF_LOCAL &&
e23775f2
PS		 1158 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1159 struct vxlan_dev *dst_vxlan;
1160
1161 ip_rt_put(rt);
1162 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1163 dst->sa.sa_family, dst_port,
1164 vxlan->flags);
1165 if (!dst_vxlan)
1166 goto tx_error;
f2252c61
PS		 1167 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1168 return;
e23775f2
PS		 1169 }
1170
f2252c61
PS		 1171 if (!info)
1172 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
1173 else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
1174 df = htons(IP_DF);
1175
e23775f2 1176 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
8063e095 1177 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
f2252c61
PS		 1178 err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr),
1179 vni, md, flags, udp_sum);
1180 if (err < 0)
1181 goto xmit_tx_error;
1182
885f0013 1183 udp_tunnel_xmit_skb(rt, sk, skb, src->sin.sin_addr.s_addr,
f2252c61
PS		 1184 dst->sin.sin_addr.s_addr, tos, ttl, df,
1185 src_port, dst_port, xnet, !udp_sum);
e23775f2
PS		 1186#if IS_ENABLED(CONFIG_IPV6)
1187 } else {
76e671ea 1188 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
e23775f2 1189 struct dst_entry *ndst;
e23775f2
PS		 1190 u32 rt6i_flags;
1191
76e671ea 1192 if (!sock6)
f2252c61 1193 goto drop;
76e671ea 1194 sk = sock6->sock->sk;
e23775f2 1195
f2252c61
PS		 1196 ndst = vxlan6_get_route(vxlan, skb,
1197 rdst ? rdst->remote_ifindex : 0, tos,
885f0013
PS		 1198 label, &dst->sin6.sin6_addr,
1199 &src->sin6.sin6_addr,
d69a9596 1200 dst_port, src_port,
e02c0ed7 1201 dst_cache, info);
f2252c61 1202 if (IS_ERR(ndst)) {
e23775f2
PS		 1203 netdev_dbg(dev, "no route to %pI6\n",
1204 &dst->sin6.sin6_addr);
1205 dev->stats.tx_carrier_errors++;
1206 goto tx_error;
1207 }
1208
1209 if (ndst->dev == dev) {
1210 netdev_dbg(dev, "circular route to %pI6\n",
1211 &dst->sin6.sin6_addr);
1212 dst_release(ndst);
1213 dev->stats.collisions++;
1214 goto tx_error;
1215 }
1216
1217 /* Bypass encapsulation if the destination is local */
1218 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
b1c74f35 1219 if (!info && rt6i_flags & RTF_LOCAL &&
e23775f2
PS		 1220 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1221 struct vxlan_dev *dst_vxlan;
1222
1223 dst_release(ndst);
1224 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
1225 dst->sa.sa_family, dst_port,
1226 vxlan->flags);
1227 if (!dst_vxlan)
1228 goto tx_error;
f2252c61
PS		 1229 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1230 return;
e23775f2
PS		 1231 }
1232
f2252c61
PS		 1233 if (!info)
1234 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
1235
1236 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
e23775f2 1237 ttl = ttl ? : ip6_dst_hoplimit(ndst);
f2252c61
PS		 1238 skb_scrub_packet(skb, xnet);
1239 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
1240 vni, md, flags, udp_sum);
1241 if (err < 0) {
1242 dst_release(ndst);
1243 return;
1244 }
1245 udp_tunnel6_xmit_skb(ndst, sk, skb, dev,
885f0013
PS		 1246 &src->sin6.sin6_addr,
1247 &dst->sin6.sin6_addr, tos, ttl,
f2252c61 1248 label, src_port, dst_port, !udp_sum);
e23775f2
PS		 1249#endif
1250 }
1251
1252 return;
1253
1254drop:
1255 dev->stats.tx_dropped++;
1256 goto tx_free;
1257
f2252c61
PS		 1258xmit_tx_error:
1259 /* skb is already freed. */
1260 skb = NULL;
e23775f2
PS		 1261rt_tx_error:
1262 ip_rt_put(rt);
1263tx_error:
1264 dev->stats.tx_errors++;
1265tx_free:
1266 dev_kfree_skb(skb);
1267}
1268
1269/* Transmit local packets over Vxlan
1270 *
1271 * Outer IP header inherits ECN and DF from inner header.
1272 * Outer UDP destination is the VXLAN assigned port.
1273 * source port is based on hash of flow
1274 */
1275netdev_tx_t rpl_vxlan_xmit(struct sk_buff *skb)
1276{
1277 struct net_device *dev = skb->dev;
1278 struct vxlan_dev *vxlan = netdev_priv(dev);
1279 const struct ip_tunnel_info *info;
1280
1281 info = skb_tunnel_info(skb);
e23775f2 1282 skb_reset_mac_header(skb);
f2252c61
PS		 1283 if (vxlan->flags & VXLAN_F_COLLECT_METADATA) {
1284 if (info && info->mode & IP_TUNNEL_INFO_TX) {
1285 vxlan_xmit_one(skb, dev, NULL, false);
1286 return NETDEV_TX_OK;
1287 }
e23775f2 1288 }
f2252c61
PS		 1289
1290 dev->stats.tx_dropped++;
e23775f2
PS		 1291 kfree_skb(skb);
1292 return NETDEV_TX_OK;
1293}
f2252c61 1294EXPORT_SYMBOL_GPL(rpl_vxlan_xmit);
e23775f2
PS		 1295
1296/* Walk the forwarding table and purge stale entries */
7f63d830 1297#ifdef HAVE_INIT_TIMER_DEFERRABLE
e23775f2
PS		 1298static void vxlan_cleanup(unsigned long arg)
1299{
1300 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
7f63d830
YS		 1301#else
1302static void vxlan_cleanup(struct timer_list *t)
1303{
1304 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
1305#endif
e23775f2
PS		 1306 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
1307 unsigned int h;
1308
1309 if (!netif_running(vxlan->dev))
1310 return;
1311
1312 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1313 struct hlist_node *p, *n;
1314
1315 spin_lock_bh(&vxlan->hash_lock);
1316 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1317 struct vxlan_fdb *f
1318 = container_of(p, struct vxlan_fdb, hlist);
1319 unsigned long timeout;
1320
1321 if (f->state & NUD_PERMANENT)
1322 continue;
1323
1324 timeout = f->used + vxlan->cfg.age_interval * HZ;
1325 if (time_before_eq(timeout, jiffies)) {
1326 netdev_dbg(vxlan->dev,
1327 "garbage collect %pM\n",
1328 f->eth_addr);
1329 f->state = NUD_STALE;
1330 vxlan_fdb_destroy(vxlan, f);
1331 } else if (time_before(timeout, next_timer))
1332 next_timer = timeout;
1333 }
1334 spin_unlock_bh(&vxlan->hash_lock);
1335 }
1336
1337 mod_timer(&vxlan->age_timer, next_timer);
1338}
1339
1340static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
1341{
1342 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
f2252c61 1343 __be32 vni = vxlan->default_dst.remote_vni;
e23775f2 1344
e23775f2
PS		 1345 spin_lock(&vn->sock_lock);
1346 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
1347 spin_unlock(&vn->sock_lock);
1348}
1349
1350/* Setup stats when device is created */
e23775f2
PS		 1351static int vxlan_init(struct net_device *dev)
1352{
f2252c61 1353 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
e23775f2
PS		 1354 if (!dev->tstats)
1355 return -ENOMEM;
1356
1357 return 0;
1358}
e23775f2
PS		 1359
1360static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
1361{
e23775f2
PS		 1362}
1363
e23775f2
PS		 1364static void vxlan_uninit(struct net_device *dev)
1365{
1366 struct vxlan_dev *vxlan = netdev_priv(dev);
1367
1368 vxlan_fdb_delete_default(vxlan);
1369
1370 free_percpu(dev->tstats);
1371}
e23775f2
PS		 1372
1373/* Start ageing timer and join group when device is brought up */
1374static int vxlan_open(struct net_device *dev)
1375{
1376 struct vxlan_dev *vxlan = netdev_priv(dev);
f2252c61 1377 int ret;
e23775f2 1378
f2252c61
PS		 1379 ret = vxlan_sock_add(vxlan);
1380 if (ret < 0)
1381 return ret;
e23775f2
PS		 1382
1383 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
1384 ret = vxlan_igmp_join(vxlan);
1385 if (ret == -EADDRINUSE)
1386 ret = 0;
1387 if (ret) {
f2252c61 1388 vxlan_sock_release(vxlan);
e23775f2
PS		 1389 return ret;
1390 }
1391 }
1392
1393 if (vxlan->cfg.age_interval)
1394 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
1395
1396 return ret;
1397}
1398
1399/* Purge the forwarding table */
1400static void vxlan_flush(struct vxlan_dev *vxlan)
1401{
1402 unsigned int h;
1403
1404 spin_lock_bh(&vxlan->hash_lock);
1405 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1406 struct hlist_node *p, *n;
e23775f2
PS		 1407 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
1408 struct vxlan_fdb *f
1409 = container_of(p, struct vxlan_fdb, hlist);
1410 /* the all_zeros_mac entry is deleted at vxlan_uninit */
1411 if (!is_zero_ether_addr(f->eth_addr))
1412 vxlan_fdb_destroy(vxlan, f);
1413 }
1414 }
1415 spin_unlock_bh(&vxlan->hash_lock);
1416}
1417
1418/* Cleanup timer and forwarding table on shutdown */
1419static int vxlan_stop(struct net_device *dev)
1420{
1421 struct vxlan_dev *vxlan = netdev_priv(dev);
1422 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
e23775f2
PS		 1423 int ret = 0;
1424
1425 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
1426 !vxlan_group_used(vn, vxlan))
1427 ret = vxlan_igmp_leave(vxlan);
1428
1429 del_timer_sync(&vxlan->age_timer);
1430
1431 vxlan_flush(vxlan);
f2252c61 1432 vxlan_sock_release(vxlan);
e23775f2
PS		 1433
1434 return ret;
1435}
1436
1437/* Stub, nothing needs to be done. */
1438static void vxlan_set_multicast_list(struct net_device *dev)
1439{
1440}
1441
27e39e3e
DW		 1442static int __vxlan_change_mtu(struct net_device *dev,
1443 struct net_device *lowerdev,
1444 struct vxlan_rdst *dst, int new_mtu, bool strict)
e23775f2 1445{
27e39e3e 1446 int max_mtu = IP_MAX_MTU;
e23775f2 1447
27e39e3e
DW		 1448 if (lowerdev)
1449 max_mtu = lowerdev->mtu;
e23775f2
PS		 1450
1451 if (dst->remote_ip.sa.sa_family == AF_INET6)
27e39e3e 1452 max_mtu -= VXLAN6_HEADROOM;
e23775f2 1453 else
27e39e3e 1454 max_mtu -= VXLAN_HEADROOM;
e23775f2 1455
27e39e3e 1456 if (new_mtu < 68)
e23775f2
PS		 1457 return -EINVAL;
1458
27e39e3e
DW		 1459 if (new_mtu > max_mtu) {
1460 if (strict)
1461 return -EINVAL;
1462
1463 new_mtu = max_mtu;
1464 }
1465
e23775f2
PS		 1466 dev->mtu = new_mtu;
1467 return 0;
1468}
1469
27e39e3e
DW		 1470static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
1471{
1472 struct vxlan_dev *vxlan = netdev_priv(dev);
1473 struct vxlan_rdst *dst = &vxlan->default_dst;
1474 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
1475 dst->remote_ifindex);
1476 return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true);
1477}
1478
f2252c61 1479int ovs_vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
aad7cb91
PS		 1480{
1481 struct vxlan_dev *vxlan = netdev_priv(dev);
f2252c61
PS		 1482 struct ip_tunnel_info *info = skb_tunnel_info(skb);
1483 __be16 sport, dport;
aad7cb91 1484
f2252c61
PS		 1485 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1486 vxlan->cfg.port_max, true);
1487 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
aad7cb91 1488
f2252c61 1489 if (ip_tunnel_info_af(info) == AF_INET) {
76e671ea 1490 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
f2252c61
PS		 1491 struct rtable *rt;
1492
76e671ea 1493 if (!sock4)
f2252c61
PS		 1494 return -EINVAL;
1495 rt = vxlan_get_route(vxlan, skb, 0, info->key.tos,
1496 info->key.u.ipv4.dst,
d69a9596
QX		 1497 &info->key.u.ipv4.src,
1498 dport, sport, NULL, info);
f2252c61
PS		 1499 if (IS_ERR(rt))
1500 return PTR_ERR(rt);
1501 ip_rt_put(rt);
1502 } else {
1503#if IS_ENABLED(CONFIG_IPV6)
1504 struct dst_entry *ndst;
aad7cb91 1505
f2252c61
PS		 1506 ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos,
1507 info->key.label, &info->key.u.ipv6.dst,
d69a9596
QX		 1508 &info->key.u.ipv6.src,
1509 dport, sport, NULL, info);
f2252c61
PS		 1510 if (IS_ERR(ndst))
1511 return PTR_ERR(ndst);
1512 dst_release(ndst);
1513#else /* !CONFIG_IPV6 */
1514 return -EPFNOSUPPORT;
1515#endif
1516 }
aad7cb91
PS		 1517 info->key.tp_src = sport;
1518 info->key.tp_dst = dport;
1519 return 0;
1520}
f2252c61 1521EXPORT_SYMBOL_GPL(ovs_vxlan_fill_metadata_dst);
aad7cb91 1522
f2252c61 1523static netdev_tx_t vxlan_dev_xmit(struct sk_buff *skb, struct net_device *dev)
aad7cb91 1524{
f2252c61
PS		 1525 /* Drop All packets coming from networking stack. OVS-CB is
1526 * not initialized for these packets.
1527 */
aad7cb91 1528
f2252c61
PS		 1529 dev_kfree_skb(skb);
1530 dev->stats.tx_dropped++;
1531 return NETDEV_TX_OK;
aad7cb91 1532}
aad7cb91 1533
f2252c61 1534static const struct net_device_ops vxlan_netdev_ether_ops = {
e23775f2
PS		 1535 .ndo_init = vxlan_init,
1536 .ndo_uninit = vxlan_uninit,
e23775f2
PS		 1537 .ndo_open = vxlan_open,
1538 .ndo_stop = vxlan_stop,
1539 .ndo_start_xmit = vxlan_dev_xmit,
f2252c61 1540 .ndo_get_stats64 = ip_tunnel_get_stats64,
e23775f2 1541 .ndo_set_rx_mode = vxlan_set_multicast_list,
39ca3383 1542#ifdef HAVE_RHEL7_MAX_MTU
3ab8a26e 1543 .ndo_size = sizeof(struct net_device_ops),
39ca3383
YHW		 1544 .extended.ndo_change_mtu = vxlan_change_mtu,
1545#else
e23775f2 1546 .ndo_change_mtu = vxlan_change_mtu,
39ca3383 1547#endif
e23775f2
PS		 1548 .ndo_validate_addr = eth_validate_addr,
1549 .ndo_set_mac_address = eth_mac_addr,
aad7cb91 1550#ifdef HAVE_NDO_FILL_METADATA_DST
f2252c61
PS		 1551 .ndo_fill_metadata_dst = ovs_vxlan_fill_metadata_dst,
1552#endif
1553};
1554
1555static const struct net_device_ops vxlan_netdev_raw_ops = {
1556 .ndo_init = vxlan_init,
1557 .ndo_uninit = vxlan_uninit,
1558 .ndo_open = vxlan_open,
1559 .ndo_stop = vxlan_stop,
1560 .ndo_start_xmit = vxlan_dev_xmit,
1561 .ndo_get_stats64 = ip_tunnel_get_stats64,
39ca3383 1562#ifdef HAVE_RHEL7_MAX_MTU
3ab8a26e 1563 .ndo_size = sizeof(struct net_device_ops),
39ca3383
YHW		 1564 .extended.ndo_change_mtu = vxlan_change_mtu,
1565#else
f2252c61 1566 .ndo_change_mtu = vxlan_change_mtu,
39ca3383 1567#endif
f2252c61
PS		 1568#ifdef HAVE_NDO_FILL_METADATA_DST
1569 .ndo_fill_metadata_dst = ovs_vxlan_fill_metadata_dst,
aad7cb91 1570#endif
e23775f2
PS		 1571};
1572
1573/* Info for udev, that this is a virtual tunnel endpoint */
1574static struct device_type vxlan_type = {
1575 .name = "vxlan",
1576};
1577
e2e11c89 1578/* Calls the ndo_add_vxlan_port or ndo_udp_tunnel_add of the caller
1579 * in order to supply the listening VXLAN udp ports. Callers are
1580 * expected to implement the ndo_add_vxlan_port.
f2252c61
PS		 1581 */
1582static void vxlan_push_rx_ports(struct net_device *dev)
1583{
1584#ifdef HAVE_NDO_ADD_VXLAN_PORT
1585 struct vxlan_sock *vs;
1586 struct net *net = dev_net(dev);
1587 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1588 sa_family_t sa_family;
1589 __be16 port;
1590 unsigned int i;
1591
1592 if (!dev->netdev_ops->ndo_add_vxlan_port)
1593 return;
1594
1595 spin_lock(&vn->sock_lock);
1596 for (i = 0; i < PORT_HASH_SIZE; ++i) {
1597 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
1598 port = inet_sk(vs->sock->sk)->inet_sport;
1599 sa_family = vxlan_get_sk_family(vs);
1600 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
1601 port);
1602 }
1603 }
1604 spin_unlock(&vn->sock_lock);
e2e11c89 1605#elif defined(HAVE_NDO_UDP_TUNNEL_ADD)
1606 struct vxlan_sock *vs;
1607 struct net *net = dev_net(dev);
1608 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1609 unsigned int i;
1610
1611 if (!dev->netdev_ops->ndo_udp_tunnel_add)
1612 return;
1613
1614 spin_lock(&vn->sock_lock);
1615 for (i = 0; i < PORT_HASH_SIZE; ++i) {
1616 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
1617 struct udp_tunnel_info ti;
1618 if (vs->flags & VXLAN_F_GPE)
1619 ti.type = UDP_TUNNEL_TYPE_VXLAN_GPE;
1620 else
1621 ti.type = UDP_TUNNEL_TYPE_VXLAN;
1622 ti.port = inet_sk(vs->sock->sk)->inet_sport;
1623 ti.sa_family = vxlan_get_sk_family(vs);
1624
1625 dev->netdev_ops->ndo_udp_tunnel_add(dev, &ti);
1626 }
1627 }
1628 spin_unlock(&vn->sock_lock);
f2252c61
PS		 1629#endif
1630}
1631
e23775f2
PS		 1632/* Initialize the device structure. */
1633static void vxlan_setup(struct net_device *dev)
1634{
1635 struct vxlan_dev *vxlan = netdev_priv(dev);
1636 unsigned int h;
1637
1638 eth_hw_addr_random(dev);
1639 ether_setup(dev);
1640
436d36db 1641#ifndef HAVE_NEEDS_FREE_NETDEV
e23775f2 1642 dev->destructor = free_netdev;
436d36db
GR		 1643#else
1644 dev->needs_free_netdev = true;
1645#endif
e23775f2
PS		 1646 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
1647
1648 dev->features |= NETIF_F_LLTX;
1649 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
1650 dev->features |= NETIF_F_RXCSUM;
1651 dev->features |= NETIF_F_GSO_SOFTWARE;
1652
1653 dev->vlan_features = dev->features;
1654 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
e23775f2
PS		 1655 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
1656 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
1657 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
e23775f2
PS		 1658#if 0
1659 netif_keep_dst(dev);
1660#endif
f2252c61 1661 dev->priv_flags |= IFF_NO_QUEUE;
e23775f2
PS		 1662
1663 INIT_LIST_HEAD(&vxlan->next);
1664 spin_lock_init(&vxlan->hash_lock);
1665
7f63d830 1666#ifdef HAVE_INIT_TIMER_DEFERRABLE
e23775f2
PS		 1667 init_timer_deferrable(&vxlan->age_timer);
1668 vxlan->age_timer.function = vxlan_cleanup;
1669 vxlan->age_timer.data = (unsigned long) vxlan;
7f63d830
YS		 1670#else
1671 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
1672#endif
e23775f2
PS		 1673
1674 vxlan->cfg.dst_port = htons(vxlan_port);
1675
1676 vxlan->dev = dev;
1677
1678 for (h = 0; h < FDB_HASH_SIZE; ++h)
1679 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
1680}
1681
f2252c61
PS		 1682static void vxlan_ether_setup(struct net_device *dev)
1683{
1684 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
1685 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
1686 dev->netdev_ops = &vxlan_netdev_ether_ops;
1687}
1688
1689static void vxlan_raw_setup(struct net_device *dev)
1690{
1691 dev->header_ops = NULL;
1692 dev->type = ARPHRD_NONE;
1693 dev->hard_header_len = 0;
1694 dev->addr_len = 0;
1695 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
1696 dev->netdev_ops = &vxlan_netdev_raw_ops;
1697}
1698
e23775f2 1699static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
f2252c61 1700 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
9cfa4718 1701 [IFLA_VXLAN_GROUP] = { .len = sizeof_field(struct iphdr, daddr) },
f2252c61
PS		 1702 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
1703 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
9cfa4718 1704 [IFLA_VXLAN_LOCAL] = { .len = sizeof_field(struct iphdr, saddr) },
f2252c61
PS		 1705 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
1706 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
1707 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
1708 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
1709 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
1710 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
1711 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
1712 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
1713 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
1714 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
1715 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
1716 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
1717 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
e23775f2 1718 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
f2252c61
PS		 1719 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
1720 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
1721 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
1722 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
1723 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
1724 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
1725 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
1726 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
e23775f2
PS		 1727};
1728
82b7e6d1 1729#ifdef HAVE_RTNLOP_VALIDATE_WITH_EXTACK
436d36db
GR		 1730static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
1731 struct netlink_ext_ack *extack)
1732#else
e23775f2 1733static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
436d36db 1734#endif
e23775f2
PS		 1735{
1736 if (tb[IFLA_ADDRESS]) {
1737 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
1738 pr_debug("invalid link address (not ethernet)\n");
1739 return -EINVAL;
1740 }
1741
1742 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
1743 pr_debug("invalid all zero ethernet address\n");
1744 return -EADDRNOTAVAIL;
1745 }
1746 }
1747
1748 if (!data)
1749 return -EINVAL;
1750
1751 if (data[IFLA_VXLAN_ID]) {
1752 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
1753 if (id >= VXLAN_VID_MASK)
1754 return -ERANGE;
1755 }
1756
1757 if (data[IFLA_VXLAN_PORT_RANGE]) {
1758 const struct ifla_vxlan_port_range *p
1759 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
1760
1761 if (ntohs(p->high) < ntohs(p->low)) {
1762 pr_debug("port range %u .. %u not valid\n",
1763 ntohs(p->low), ntohs(p->high));
1764 return -EINVAL;
1765 }
1766 }
1767
1768 return 0;
1769}
1770
1771static void vxlan_get_drvinfo(struct net_device *netdev,
1772 struct ethtool_drvinfo *drvinfo)
1773{
1774 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
1775 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
1776}
1777
1778static const struct ethtool_ops vxlan_ethtool_ops = {
1779 .get_drvinfo = vxlan_get_drvinfo,
1780 .get_link = ethtool_op_get_link,
1781};
1782
e23775f2
PS		 1783static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
1784 __be16 port, u32 flags)
1785{
1786 struct socket *sock;
1787 struct udp_port_cfg udp_conf;
1788 int err;
1789
1790 memset(&udp_conf, 0, sizeof(udp_conf));
1791
1792 if (ipv6) {
1793 udp_conf.family = AF_INET6;
1794 udp_conf.use_udp6_rx_checksums =
1795 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
1796 udp_conf.ipv6_v6only = 1;
1797 } else {
1798 udp_conf.family = AF_INET;
1799 }
1800
1801 udp_conf.local_udp_port = port;
1802
1803 /* Open UDP socket */
1804 err = udp_sock_create(net, &udp_conf, &sock);
1805 if (err < 0)
1806 return ERR_PTR(err);
1807
1808 return sock;
1809}
1810
1811/* Create new listen socket if needed */
f2252c61
PS		 1812static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
1813 __be16 port, u32 flags)
e23775f2
PS		 1814{
1815 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
1816 struct vxlan_sock *vs;
1817 struct socket *sock;
1818 unsigned int h;
e23775f2
PS		 1819 struct udp_tunnel_sock_cfg tunnel_cfg;
1820
1821 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
1822 if (!vs)
1823 return ERR_PTR(-ENOMEM);
1824
1825 for (h = 0; h < VNI_HASH_SIZE; ++h)
1826 INIT_HLIST_HEAD(&vs->vni_list[h]);
1827
e23775f2
PS		 1828 sock = vxlan_create_sock(net, ipv6, port, flags);
1829 if (IS_ERR(sock)) {
e23775f2
PS		 1830 kfree(vs);
1831 return ERR_CAST(sock);
1832 }
1833
1834 vs->sock = sock;
1835 atomic_set(&vs->refcnt, 1);
1836 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
1837
e23775f2
PS		 1838#ifdef HAVE_UDP_OFFLOAD
1839 vs->udp_offloads.port = port;
1840 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
1841 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
e23775f2
PS		 1842#endif
1843
1844 spin_lock(&vn->sock_lock);
1845 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
f2252c61 1846 vxlan_notify_add_rx_port(vs);
e23775f2
PS		 1847 spin_unlock(&vn->sock_lock);
1848
1849 /* Mark socket as an encapsulation socket. */
f2252c61 1850 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
e23775f2
PS		 1851 tunnel_cfg.sk_user_data = vs;
1852 tunnel_cfg.encap_type = 1;
f2252c61 1853 tunnel_cfg.encap_rcv = vxlan_rcv;
e23775f2 1854 tunnel_cfg.encap_destroy = NULL;
f2252c61
PS		 1855#ifdef HAVE_UDP_TUNNEL_SOCK_CFG_GRO_RECEIVE
1856 tunnel_cfg.gro_receive = vxlan_gro_receive;
1857 tunnel_cfg.gro_complete = vxlan_gro_complete;
1858#endif
e23775f2
PS		 1859 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
1860
1861 return vs;
1862}
1863
f2252c61 1864static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
e23775f2 1865{
f2252c61
PS		 1866 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
1867 struct vxlan_sock *vs = NULL;
e23775f2 1868
f2252c61 1869 if (!vxlan->cfg.no_share) {
e23775f2 1870 spin_lock(&vn->sock_lock);
f2252c61
PS		 1871 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
1872 vxlan->cfg.dst_port, vxlan->flags);
1873 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
e23775f2 1874 spin_unlock(&vn->sock_lock);
f2252c61 1875 return -EBUSY;
e23775f2
PS		 1876 }
1877 spin_unlock(&vn->sock_lock);
1878 }
f2252c61
PS		 1879 if (!vs)
1880 vs = vxlan_socket_create(vxlan->net, ipv6,
1881 vxlan->cfg.dst_port, vxlan->flags);
1882 if (IS_ERR(vs))
1883 return PTR_ERR(vs);
1884#if IS_ENABLED(CONFIG_IPV6)
1885 if (ipv6)
76e671ea 1886 rcu_assign_pointer(vxlan->vn6_sock, vs);
f2252c61
PS		 1887 else
1888#endif
76e671ea 1889 rcu_assign_pointer(vxlan->vn4_sock, vs);
f2252c61
PS		 1890 vxlan_vs_add_dev(vs, vxlan);
1891 return 0;
1892}
e23775f2 1893
f2252c61
PS		 1894static int vxlan_sock_add(struct vxlan_dev *vxlan)
1895{
f2252c61 1896 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
86e4311b
JB		 1897 bool ipv6 = vxlan->flags & VXLAN_F_IPV6 || metadata;
1898 bool ipv4 = !ipv6 || metadata;
f2252c61
PS		 1899 int ret = 0;
1900
76e671ea 1901 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
f2252c61 1902#if IS_ENABLED(CONFIG_IPV6)
76e671ea 1903 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
86e4311b 1904 if (ipv6) {
f2252c61 1905 ret = __vxlan_sock_add(vxlan, true);
86e4311b
JB		 1906 if (ret < 0 && ret != -EAFNOSUPPORT)
1907 ipv4 = false;
1908 }
f2252c61 1909#endif
86e4311b 1910 if (ipv4)
f2252c61
PS		 1911 ret = __vxlan_sock_add(vxlan, false);
1912 if (ret < 0)
1913 vxlan_sock_release(vxlan);
1914 return ret;
e23775f2
PS		 1915}
1916
1917static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
1918 struct vxlan_config *conf)
1919{
1920 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
f2252c61 1921 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
e23775f2 1922 struct vxlan_rdst *dst = &vxlan->default_dst;
f2252c61 1923 unsigned short needed_headroom = ETH_HLEN;
e23775f2
PS		 1924 int err;
1925 bool use_ipv6 = false;
1926 __be16 default_port = vxlan->cfg.dst_port;
06f1a61a 1927 struct net_device *lowerdev = NULL;
e23775f2 1928
f2252c61
PS		 1929 if (conf->flags & VXLAN_F_GPE) {
1930 if (conf->flags & ~VXLAN_F_ALLOWED_GPE)
1931 return -EINVAL;
1932 /* For now, allow GPE only together with COLLECT_METADATA.
1933 * This can be relaxed later; in such case, the other side
1934 * of the PtP link will have to be provided.
1935 */
1936 if (!(conf->flags & VXLAN_F_COLLECT_METADATA))
1937 return -EINVAL;
1938
1939 vxlan_raw_setup(dev);
1940 } else {
1941 vxlan_ether_setup(dev);
1942 }
1943
e23775f2
PS		 1944 vxlan->net = src_net;
1945
1946 dst->remote_vni = conf->vni;
1947
1948 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
1949
1950 /* Unless IPv6 is explicitly requested, assume IPv4 */
1951 if (!dst->remote_ip.sa.sa_family)
1952 dst->remote_ip.sa.sa_family = AF_INET;
1953
1954 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
1955 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
1956 if (!IS_ENABLED(CONFIG_IPV6))
1957 return -EPFNOSUPPORT;
1958 use_ipv6 = true;
f2252c61
PS		 1959 vxlan->flags |= VXLAN_F_IPV6;
1960 }
1961
1962 if (conf->label && !use_ipv6) {
1963 pr_info("label only supported in use with IPv6\n");
1964 return -EINVAL;
e23775f2
PS		 1965 }
1966
1967 if (conf->remote_ifindex) {
06f1a61a 1968 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
e23775f2
PS		 1969 dst->remote_ifindex = conf->remote_ifindex;
1970
1971 if (!lowerdev) {
1972 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
1973 return -ENODEV;
1974 }
1975
1976#if IS_ENABLED(CONFIG_IPV6)
1977 if (use_ipv6) {
1978 struct inet6_dev *idev = __in6_dev_get(lowerdev);
1979 if (idev && idev->cnf.disable_ipv6) {
1980 pr_info("IPv6 is disabled via sysctl\n");
1981 return -EPERM;
1982 }
e23775f2
PS		 1983 }
1984#endif
1985
1986 if (!conf->mtu)
1987 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
1988
f2252c61 1989 needed_headroom = lowerdev->hard_header_len;
e23775f2
PS		 1990 }
1991
06f1a61a
DW		 1992 if (conf->mtu) {
1993 err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false);
1994 if (err)
1995 return err;
1996 }
1997
f2252c61
PS		 1998 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
1999 needed_headroom += VXLAN6_HEADROOM;
2000 else
2001 needed_headroom += VXLAN_HEADROOM;
2002 dev->needed_headroom = needed_headroom;
2003
e23775f2 2004 memcpy(&vxlan->cfg, conf, sizeof(*conf));
f2252c61
PS		 2005 if (!vxlan->cfg.dst_port) {
2006 if (conf->flags & VXLAN_F_GPE)
2007 vxlan->cfg.dst_port = 4790; /* IANA assigned VXLAN-GPE port */
2008 else
2009 vxlan->cfg.dst_port = default_port;
2010 }
e23775f2
PS		 2011 vxlan->flags |= conf->flags;
2012
2013 if (!vxlan->cfg.age_interval)
2014 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
2015
f2252c61
PS		 2016 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2017 if (tmp->cfg.vni == conf->vni &&
2018 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2019 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2020 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2021 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2022 (vxlan->flags & VXLAN_F_RCV_FLAGS))
e23775f2 2023 return -EEXIST;
f2252c61 2024 }
e23775f2
PS		 2025
2026 dev->ethtool_ops = &vxlan_ethtool_ops;
2027
2028 /* create an fdb entry for a valid default destination */
2029 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2030 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2031 &vxlan->default_dst.remote_ip,
2032 NUD_REACHABLE|NUD_PERMANENT,
2033 NLM_F_EXCL|NLM_F_CREATE,
2034 vxlan->cfg.dst_port,
2035 vxlan->default_dst.remote_vni,
2036 vxlan->default_dst.remote_ifindex,
2037 NTF_SELF);
2038 if (err)
2039 return err;
2040 }
2041
2042 err = register_netdevice(dev);
2043 if (err) {
2044 vxlan_fdb_delete_default(vxlan);
2045 return err;
2046 }
2047
2048 list_add(&vxlan->next, &vn->vxlan_list);
2049
2050 return 0;
2051}
2052
436d36db
GR		 2053#ifdef HAVE_EXT_ACK_IN_RTNL_LINKOPS
2054static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2055 struct nlattr *tb[], struct nlattr *data[],
2056 struct netlink_ext_ack *extack)
2057#else
e23775f2
PS		 2058static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2059 struct nlattr *tb[], struct nlattr *data[])
436d36db 2060#endif
e23775f2 2061{
f2252c61 2062 pr_info("unsupported operation\n");
e23775f2
PS		 2063 return -EINVAL;
2064}
2065
e23775f2 2066static void vxlan_dellink(struct net_device *dev, struct list_head *head)
e23775f2
PS		 2067{
2068 struct vxlan_dev *vxlan = netdev_priv(dev);
2069 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2070
2071 spin_lock(&vn->sock_lock);
2072 if (!hlist_unhashed(&vxlan->hlist))
2073 hlist_del_rcu(&vxlan->hlist);
2074 spin_unlock(&vn->sock_lock);
2075
2076 list_del(&vxlan->next);
2077 unregister_netdevice_queue(dev, head);
2078}
2079
2080static size_t vxlan_get_size(const struct net_device *dev)
2081{
2082
2083 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
2084 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
2085 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
2086 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
2087 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2088 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
f2252c61 2089 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
e23775f2
PS		 2090 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
2091 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2092 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2093 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2094 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
2095 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
2096 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2097 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
2098 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
2099 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2100 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2101 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2102 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
2103 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2104 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
2105 0;
2106}
2107
2108static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2109{
2110 const struct vxlan_dev *vxlan = netdev_priv(dev);
f2252c61
PS		 2111 const struct vxlan_rdst *dst = &vxlan->default_dst;
2112 struct ifla_vxlan_port_range ports = {
2113 .low = htons(vxlan->cfg.port_min),
2114 .high = htons(vxlan->cfg.port_max),
2115 };
2116
2117 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
2118 goto nla_put_failure;
2119
2120 if (!vxlan_addr_any(&dst->remote_ip)) {
2121 if (dst->remote_ip.sa.sa_family == AF_INET) {
2122 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
2123 dst->remote_ip.sin.sin_addr.s_addr))
2124 goto nla_put_failure;
2125#if IS_ENABLED(CONFIG_IPV6)
2126 } else {
2127 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
2128 &dst->remote_ip.sin6.sin6_addr))
2129 goto nla_put_failure;
2130#endif
2131 }
2132 }
2133
2134 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
2135 goto nla_put_failure;
2136
2137 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
2138 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
2139 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
2140 vxlan->cfg.saddr.sin.sin_addr.s_addr))
2141 goto nla_put_failure;
2142#if IS_ENABLED(CONFIG_IPV6)
2143 } else {
2144 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
2145 &vxlan->cfg.saddr.sin6.sin6_addr))
2146 goto nla_put_failure;
2147#endif
2148 }
2149 }
2150
2151 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
2152 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
2153 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
2154 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2155 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2156 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2157 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2158 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2159 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2160 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2161 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2162 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
2163 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
2164 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
2165 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
2166 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
2167 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
2168 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2169 !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
2170 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2171 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2172 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
2173 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2174 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2175 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2176 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2177 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
2178 goto nla_put_failure;
2179
2180 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2181 goto nla_put_failure;
2182
2183 if (vxlan->flags & VXLAN_F_GBP &&
2184 nla_put_flag(skb, IFLA_VXLAN_GBP))
2185 goto nla_put_failure;
2186
2187 if (vxlan->flags & VXLAN_F_GPE &&
2188 nla_put_flag(skb, IFLA_VXLAN_GPE))
2189 goto nla_put_failure;
e23775f2 2190
f2252c61
PS		 2191 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
2192 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
e23775f2
PS		 2193 goto nla_put_failure;
2194
2195 return 0;
2196
2197nla_put_failure:
2198 return -EMSGSIZE;
2199}
2200
2201#ifdef HAVE_GET_LINK_NET
2202static struct net *vxlan_get_link_net(const struct net_device *dev)
2203{
2204 struct vxlan_dev *vxlan = netdev_priv(dev);
2205
2206 return vxlan->net;
2207}
2208#endif
2209
2210static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
2211 .kind = "ovs_vxlan",
2212 .maxtype = IFLA_VXLAN_MAX,
2213 .policy = vxlan_policy,
2214 .priv_size = sizeof(struct vxlan_dev),
2215 .setup = vxlan_setup,
2216 .validate = vxlan_validate,
2217 .newlink = vxlan_newlink,
2218 .dellink = vxlan_dellink,
2219 .get_size = vxlan_get_size,
2220 .fill_info = vxlan_fill_info,
2221#ifdef HAVE_GET_LINK_NET
2222 .get_link_net = vxlan_get_link_net,
2223#endif
2224};
2225
f6d70869
PS		 2226struct net_device *rpl_vxlan_dev_create(struct net *net, const char *name,
2227 u8 name_assign_type,
2228 struct vxlan_config *conf)
2229{
2230 struct nlattr *tb[IFLA_MAX + 1];
2231 struct net_device *dev;
2232 int err;
2233
2234 memset(&tb, 0, sizeof(tb));
2235
2236 dev = rtnl_create_link(net, name, name_assign_type,
2237 &vxlan_link_ops, tb);
2238 if (IS_ERR(dev))
2239 return dev;
2240
2241 err = vxlan_dev_configure(net, dev, conf);
2242 if (err < 0) {
2243 free_netdev(dev);
2244 return ERR_PTR(err);
2245 }
2246
2247 err = rtnl_configure_link(dev, NULL);
2248 if (err < 0) {
2249 LIST_HEAD(list_kill);
2250
2251 vxlan_dellink(dev, &list_kill);
2252 unregister_netdevice_many(&list_kill);
2253 return ERR_PTR(err);
2254 }
2255
2256 return dev;
2257}
2258EXPORT_SYMBOL_GPL(rpl_vxlan_dev_create);
2259
e23775f2
PS		 2260static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
2261 struct net_device *dev)
2262{
2263 struct vxlan_dev *vxlan, *next;
2264 LIST_HEAD(list_kill);
2265
2266 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2267 struct vxlan_rdst *dst = &vxlan->default_dst;
2268
2269 /* In case we created vxlan device with carrier
2270 * and we loose the carrier due to module unload
2271 * we also need to remove vxlan device. In other
2272 * cases, it's not necessary and remote_ifindex
2273 * is 0 here, so no matches.
2274 */
2275 if (dst->remote_ifindex == dev->ifindex)
e23775f2 2276 vxlan_dellink(vxlan->dev, &list_kill);
e23775f2
PS		 2277 }
2278
2279 unregister_netdevice_many(&list_kill);
2280}
2281
f2252c61
PS		 2282static int vxlan_netdevice_event(struct notifier_block *unused,
2283 unsigned long event, void *ptr)
e23775f2
PS		 2284{
2285 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
2286 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
2287
2288 if (event == NETDEV_UNREGISTER)
2289 vxlan_handle_lowerdev_unregister(vn, dev);
f2252c61
PS		 2290 else if (event == NETDEV_OFFLOAD_PUSH_VXLAN)
2291 vxlan_push_rx_ports(dev);
e23775f2
PS		 2292
2293 return NOTIFY_DONE;
2294}
2295
2296static struct notifier_block vxlan_notifier_block __read_mostly = {
f2252c61 2297 .notifier_call = vxlan_netdevice_event,
e23775f2
PS		 2298};
2299
2300static __net_init int vxlan_init_net(struct net *net)
2301{
2302 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2303 unsigned int h;
2304
2305 INIT_LIST_HEAD(&vn->vxlan_list);
2306 spin_lock_init(&vn->sock_lock);
2307
2308 for (h = 0; h < PORT_HASH_SIZE; ++h)
2309 INIT_HLIST_HEAD(&vn->sock_list[h]);
2310
2311 return 0;
2312}
2313
2314static void __net_exit vxlan_exit_net(struct net *net)
2315{
2316 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2317 struct vxlan_dev *vxlan, *next;
2318 struct net_device *dev, *aux;
2319 LIST_HEAD(list);
2320
2321 rtnl_lock();
2322 for_each_netdev_safe(net, dev, aux)
2323 if (dev->rtnl_link_ops == &vxlan_link_ops)
2324 unregister_netdevice_queue(dev, &list);
2325
2326 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
2327 /* If vxlan->dev is in the same netns, it has already been added
2328 * to the list by the previous loop.
2329 */
f2252c61 2330 if (!net_eq(dev_net(vxlan->dev), net)) {
e23775f2 2331 unregister_netdevice_queue(vxlan->dev, &list);
f2252c61 2332 }
e23775f2
PS		 2333 }
2334
2335 unregister_netdevice_many(&list);
2336 rtnl_unlock();
2337}
2338
2339static struct pernet_operations vxlan_net_ops = {
2340 .init = vxlan_init_net,
2341 .exit = vxlan_exit_net,
2342 .id = &vxlan_net_id,
2343 .size = sizeof(struct vxlan_net),
2344};
2345
e23775f2
PS		 2346int rpl_vxlan_init_module(void)
2347{
2348 int rc;
2349
e23775f2
PS		 2350 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
2351
2352 rc = register_pernet_subsys(&vxlan_net_ops);
2353 if (rc)
2354 goto out1;
2355
2356 rc = register_netdevice_notifier(&vxlan_notifier_block);
2357 if (rc)
2358 goto out2;
2359
2360 rc = rtnl_link_register(&vxlan_link_ops);
2361 if (rc)
2362 goto out3;
2363
2364 pr_info("VxLAN tunneling driver\n");
2365 return 0;
2366out3:
2367 unregister_netdevice_notifier(&vxlan_notifier_block);
2368out2:
2369 unregister_pernet_subsys(&vxlan_net_ops);
2370out1:
258b27d3 2371 pr_err("Error while initializing VxLAN %d\n", rc);
e23775f2
PS		 2372 return rc;
2373}
2374
2375void rpl_vxlan_cleanup_module(void)
2376{
2377 rtnl_link_unregister(&vxlan_link_ops);
2378 unregister_netdevice_notifier(&vxlan_notifier_block);
e23775f2
PS		 2379 unregister_pernet_subsys(&vxlan_net_ops);
2380 /* rcu_barrier() is called by netns */
2381}
2382#endif
openvswitch mirror