blob: bbac1d35ed4e37450a3282b4b7f9dc41dabcd4e8 [file] [log] [blame]
stephen hemmingerd3428942012-10-01 12:32:35 +00001/*
Rami Roseneb5ce432012-11-13 13:29:15 +00002 * VXLAN: Virtual eXtensible Local Area Network
stephen hemmingerd3428942012-10-01 12:32:35 +00003 *
stephen hemminger3b8df3c2013-04-27 11:31:52 +00004 * Copyright (c) 2012-2013 Vyatta Inc.
stephen hemmingerd3428942012-10-01 12:32:35 +00005 *
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.
stephen hemmingerd3428942012-10-01 12:32:35 +00009 */
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>
21#include <linux/in.h>
22#include <linux/ip.h>
23#include <linux/udp.h>
24#include <linux/igmp.h>
25#include <linux/etherdevice.h>
26#include <linux/if_ether.h>
Pravin B Shelar1eaa8172013-08-19 11:23:29 -070027#include <linux/if_vlan.h>
stephen hemmingerd3428942012-10-01 12:32:35 +000028#include <linux/hash.h>
Yan Burman1b13c972013-01-29 23:43:07 +000029#include <linux/ethtool.h>
David Stevense4f67ad2012-11-20 02:50:14 +000030#include <net/arp.h>
31#include <net/ndisc.h>
stephen hemmingerd3428942012-10-01 12:32:35 +000032#include <net/ip.h>
Pravin B Shelarc5441932013-03-25 14:49:35 +000033#include <net/ip_tunnels.h>
stephen hemmingerd3428942012-10-01 12:32:35 +000034#include <net/icmp.h>
35#include <net/udp.h>
Tom Herbert3ee64f32014-07-13 19:49:42 -070036#include <net/udp_tunnel.h>
stephen hemmingerd3428942012-10-01 12:32:35 +000037#include <net/rtnetlink.h>
38#include <net/route.h>
39#include <net/dsfield.h>
40#include <net/inet_ecn.h>
41#include <net/net_namespace.h>
42#include <net/netns/generic.h>
Pravin B Shelar012a5722013-08-19 11:23:07 -070043#include <net/vxlan.h>
Or Gerlitzdc01e7d2014-01-20 13:59:21 +020044#include <net/protocol.h>
Andy Zhouacbf74a2014-09-16 17:31:18 -070045#include <net/udp_tunnel.h>
Cong Wange4c7ed42013-08-31 13:44:33 +080046#if IS_ENABLED(CONFIG_IPV6)
47#include <net/ipv6.h>
48#include <net/addrconf.h>
49#include <net/ip6_tunnel.h>
Cong Wang660d98c2013-09-02 10:06:52 +080050#include <net/ip6_checksum.h>
Cong Wange4c7ed42013-08-31 13:44:33 +080051#endif
Thomas Grafee122c72015-07-21 10:43:58 +020052#include <net/dst_metadata.h>
stephen hemmingerd3428942012-10-01 12:32:35 +000053
54#define VXLAN_VERSION "0.1"
55
stephen hemminger553675f2013-05-16 11:35:20 +000056#define PORT_HASH_BITS 8
57#define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
stephen hemmingerd3428942012-10-01 12:32:35 +000058#define FDB_AGE_DEFAULT 300 /* 5 min */
59#define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
60
stephen hemminger23c578b2013-04-27 11:31:53 +000061/* UDP port for VXLAN traffic.
62 * The IANA assigned port is 4789, but the Linux default is 8472
Stephen Hemminger234f5b72013-06-17 14:16:41 -070063 * for compatibility with early adopters.
stephen hemminger23c578b2013-04-27 11:31:53 +000064 */
Stephen Hemminger9daaa392013-06-17 14:16:12 -070065static unsigned short vxlan_port __read_mostly = 8472;
66module_param_named(udp_port, vxlan_port, ushort, 0444);
stephen hemmingerd3428942012-10-01 12:32:35 +000067MODULE_PARM_DESC(udp_port, "Destination UDP port");
68
69static bool log_ecn_error = true;
70module_param(log_ecn_error, bool, 0644);
71MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
72
Pravin B Shelar60d9d4c2013-06-20 00:26:31 -070073static int vxlan_net_id;
Thomas Graf0dfbdf42015-07-21 10:44:02 +020074static struct rtnl_link_ops vxlan_link_ops;
stephen hemminger553675f2013-05-16 11:35:20 +000075
Mike Rapoportafbd8ba2013-06-25 16:01:51 +030076static const u8 all_zeros_mac[ETH_ALEN];
77
Thomas Graf614732e2015-07-21 10:44:06 +020078static struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
79 bool no_share, u32 flags);
80
stephen hemminger553675f2013-05-16 11:35:20 +000081/* per-network namespace private data for this module */
82struct vxlan_net {
83 struct list_head vxlan_list;
84 struct hlist_head sock_list[PORT_HASH_SIZE];
Stephen Hemminger1c51a912013-06-17 14:16:11 -070085 spinlock_t sock_lock;
stephen hemminger553675f2013-05-16 11:35:20 +000086};
87
stephen hemmingerd3428942012-10-01 12:32:35 +000088/* Forwarding table entry */
89struct vxlan_fdb {
90 struct hlist_node hlist; /* linked list of entries */
91 struct rcu_head rcu;
92 unsigned long updated; /* jiffies */
93 unsigned long used;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -070094 struct list_head remotes;
Sowmini Varadhan7177a3b2015-07-20 09:54:50 +020095 u8 eth_addr[ETH_ALEN];
stephen hemmingerd3428942012-10-01 12:32:35 +000096 u16 state; /* see ndm_state */
David Stevensae884082013-04-19 00:36:26 +000097 u8 flags; /* see ndm_flags */
stephen hemmingerd3428942012-10-01 12:32:35 +000098};
99
stephen hemmingerd3428942012-10-01 12:32:35 +0000100/* salt for hash table */
101static u32 vxlan_salt __read_mostly;
Stephen Hemminger758c57d2013-06-17 14:16:09 -0700102static struct workqueue_struct *vxlan_wq;
stephen hemmingerd3428942012-10-01 12:32:35 +0000103
Thomas Grafee122c72015-07-21 10:43:58 +0200104static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
105{
Thomas Grafe7030872015-07-21 10:44:01 +0200106 return vs->flags & VXLAN_F_COLLECT_METADATA ||
107 ip_tunnel_collect_metadata();
Thomas Grafee122c72015-07-21 10:43:58 +0200108}
109
Cong Wange4c7ed42013-08-31 13:44:33 +0800110#if IS_ENABLED(CONFIG_IPV6)
111static inline
112bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
113{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200114 if (a->sa.sa_family != b->sa.sa_family)
115 return false;
116 if (a->sa.sa_family == AF_INET6)
117 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
118 else
119 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
Cong Wange4c7ed42013-08-31 13:44:33 +0800120}
121
122static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
123{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200124 if (ipa->sa.sa_family == AF_INET6)
125 return ipv6_addr_any(&ipa->sin6.sin6_addr);
126 else
127 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
Cong Wange4c7ed42013-08-31 13:44:33 +0800128}
129
130static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
131{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200132 if (ipa->sa.sa_family == AF_INET6)
133 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
134 else
135 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
Cong Wange4c7ed42013-08-31 13:44:33 +0800136}
137
138static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
139{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200140 if (nla_len(nla) >= sizeof(struct in6_addr)) {
Jiri Benc67b61f62015-03-29 16:59:26 +0200141 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200142 ip->sa.sa_family = AF_INET6;
143 return 0;
144 } else if (nla_len(nla) >= sizeof(__be32)) {
Jiri Benc67b61f62015-03-29 16:59:26 +0200145 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200146 ip->sa.sa_family = AF_INET;
147 return 0;
148 } else {
149 return -EAFNOSUPPORT;
150 }
Cong Wange4c7ed42013-08-31 13:44:33 +0800151}
152
153static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
Jiri Bencf0ef3122015-03-29 16:17:37 +0200154 const union vxlan_addr *ip)
Cong Wange4c7ed42013-08-31 13:44:33 +0800155{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200156 if (ip->sa.sa_family == AF_INET6)
Jiri Benc930345e2015-03-29 16:59:25 +0200157 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200158 else
Jiri Benc930345e2015-03-29 16:59:25 +0200159 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
Cong Wange4c7ed42013-08-31 13:44:33 +0800160}
161
162#else /* !CONFIG_IPV6 */
163
164static inline
165bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
166{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200167 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
Cong Wange4c7ed42013-08-31 13:44:33 +0800168}
169
170static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
171{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200172 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
Cong Wange4c7ed42013-08-31 13:44:33 +0800173}
174
175static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
176{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200177 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
Cong Wange4c7ed42013-08-31 13:44:33 +0800178}
179
180static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
181{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200182 if (nla_len(nla) >= sizeof(struct in6_addr)) {
183 return -EAFNOSUPPORT;
184 } else if (nla_len(nla) >= sizeof(__be32)) {
Jiri Benc67b61f62015-03-29 16:59:26 +0200185 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200186 ip->sa.sa_family = AF_INET;
187 return 0;
188 } else {
189 return -EAFNOSUPPORT;
190 }
Cong Wange4c7ed42013-08-31 13:44:33 +0800191}
192
193static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
Jiri Bencf0ef3122015-03-29 16:17:37 +0200194 const union vxlan_addr *ip)
Cong Wange4c7ed42013-08-31 13:44:33 +0800195{
Jiri Benc930345e2015-03-29 16:59:25 +0200196 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
Cong Wange4c7ed42013-08-31 13:44:33 +0800197}
198#endif
199
stephen hemminger553675f2013-05-16 11:35:20 +0000200/* Virtual Network hash table head */
201static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
202{
203 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
204}
205
206/* Socket hash table head */
207static inline struct hlist_head *vs_head(struct net *net, __be16 port)
stephen hemmingerd3428942012-10-01 12:32:35 +0000208{
209 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
210
stephen hemminger553675f2013-05-16 11:35:20 +0000211 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
212}
213
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700214/* First remote destination for a forwarding entry.
215 * Guaranteed to be non-NULL because remotes are never deleted.
216 */
stephen hemminger5ca54612013-08-04 17:17:39 -0700217static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700218{
stephen hemminger5ca54612013-08-04 17:17:39 -0700219 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
220}
221
222static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
223{
224 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700225}
226
Thomas Grafac5132d2015-01-15 03:53:56 +0100227/* Find VXLAN socket based on network namespace, address family and UDP port
228 * and enabled unshareable flags.
229 */
230static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
231 __be16 port, u32 flags)
stephen hemminger553675f2013-05-16 11:35:20 +0000232{
233 struct vxlan_sock *vs;
Tom Herbertaf33c1a2015-01-20 11:23:05 -0800234
235 flags &= VXLAN_F_RCV_FLAGS;
stephen hemminger553675f2013-05-16 11:35:20 +0000236
237 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -0200238 if (inet_sk(vs->sock->sk)->inet_sport == port &&
Jiri Benc705cc622015-08-20 13:56:28 +0200239 vxlan_get_sk_family(vs) == family &&
Tom Herbertaf33c1a2015-01-20 11:23:05 -0800240 vs->flags == flags)
stephen hemminger553675f2013-05-16 11:35:20 +0000241 return vs;
242 }
243 return NULL;
stephen hemmingerd3428942012-10-01 12:32:35 +0000244}
245
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700246static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
stephen hemmingerd3428942012-10-01 12:32:35 +0000247{
248 struct vxlan_dev *vxlan;
stephen hemmingerd3428942012-10-01 12:32:35 +0000249
stephen hemminger553675f2013-05-16 11:35:20 +0000250 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
Atzm Watanabec7995c42013-04-16 02:50:52 +0000251 if (vxlan->default_dst.remote_vni == id)
stephen hemmingerd3428942012-10-01 12:32:35 +0000252 return vxlan;
253 }
254
255 return NULL;
256}
257
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700258/* Look up VNI in a per net namespace table */
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -0200259static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
Thomas Grafac5132d2015-01-15 03:53:56 +0100260 sa_family_t family, __be16 port,
261 u32 flags)
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700262{
263 struct vxlan_sock *vs;
264
Thomas Grafac5132d2015-01-15 03:53:56 +0100265 vs = vxlan_find_sock(net, family, port, flags);
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700266 if (!vs)
267 return NULL;
268
269 return vxlan_vs_find_vni(vs, id);
270}
271
stephen hemmingerd3428942012-10-01 12:32:35 +0000272/* Fill in neighbour message in skbuff. */
273static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
Stephen Hemminger234f5b72013-06-17 14:16:41 -0700274 const struct vxlan_fdb *fdb,
275 u32 portid, u32 seq, int type, unsigned int flags,
276 const struct vxlan_rdst *rdst)
stephen hemmingerd3428942012-10-01 12:32:35 +0000277{
278 unsigned long now = jiffies;
279 struct nda_cacheinfo ci;
280 struct nlmsghdr *nlh;
281 struct ndmsg *ndm;
David Stevense4f67ad2012-11-20 02:50:14 +0000282 bool send_ip, send_eth;
stephen hemmingerd3428942012-10-01 12:32:35 +0000283
284 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
285 if (nlh == NULL)
286 return -EMSGSIZE;
287
288 ndm = nlmsg_data(nlh);
289 memset(ndm, 0, sizeof(*ndm));
David Stevense4f67ad2012-11-20 02:50:14 +0000290
291 send_eth = send_ip = true;
292
293 if (type == RTM_GETNEIGH) {
294 ndm->ndm_family = AF_INET;
Cong Wange4c7ed42013-08-31 13:44:33 +0800295 send_ip = !vxlan_addr_any(&rdst->remote_ip);
David Stevense4f67ad2012-11-20 02:50:14 +0000296 send_eth = !is_zero_ether_addr(fdb->eth_addr);
297 } else
298 ndm->ndm_family = AF_BRIDGE;
stephen hemmingerd3428942012-10-01 12:32:35 +0000299 ndm->ndm_state = fdb->state;
300 ndm->ndm_ifindex = vxlan->dev->ifindex;
David Stevensae884082013-04-19 00:36:26 +0000301 ndm->ndm_flags = fdb->flags;
Jun Zhao545469f2014-07-26 00:38:59 +0800302 ndm->ndm_type = RTN_UNICAST;
stephen hemmingerd3428942012-10-01 12:32:35 +0000303
Nicolas Dichtel193523b2015-01-20 15:15:47 +0100304 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
Nicolas Dichtel49670822015-01-26 14:10:53 +0100305 nla_put_s32(skb, NDA_LINK_NETNSID,
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200306 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
Nicolas Dichtel193523b2015-01-20 15:15:47 +0100307 goto nla_put_failure;
308
David Stevense4f67ad2012-11-20 02:50:14 +0000309 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
stephen hemmingerd3428942012-10-01 12:32:35 +0000310 goto nla_put_failure;
311
Cong Wange4c7ed42013-08-31 13:44:33 +0800312 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
David Stevens66817122013-03-15 04:35:51 +0000313 goto nla_put_failure;
314
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200315 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
David Stevens66817122013-03-15 04:35:51 +0000316 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
317 goto nla_put_failure;
Atzm Watanabec7995c42013-04-16 02:50:52 +0000318 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
Pravin B Shelar60d9d4c2013-06-20 00:26:31 -0700319 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
David Stevens66817122013-03-15 04:35:51 +0000320 goto nla_put_failure;
321 if (rdst->remote_ifindex &&
322 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
stephen hemmingerd3428942012-10-01 12:32:35 +0000323 goto nla_put_failure;
324
325 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
326 ci.ndm_confirmed = 0;
327 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
328 ci.ndm_refcnt = 0;
329
330 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
331 goto nla_put_failure;
332
Johannes Berg053c0952015-01-16 22:09:00 +0100333 nlmsg_end(skb, nlh);
334 return 0;
stephen hemmingerd3428942012-10-01 12:32:35 +0000335
336nla_put_failure:
337 nlmsg_cancel(skb, nlh);
338 return -EMSGSIZE;
339}
340
341static inline size_t vxlan_nlmsg_size(void)
342{
343 return NLMSG_ALIGN(sizeof(struct ndmsg))
344 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
Cong Wange4c7ed42013-08-31 13:44:33 +0800345 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
stephen hemminger73cf3312013-04-27 11:31:54 +0000346 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
David Stevens66817122013-03-15 04:35:51 +0000347 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
348 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
Nicolas Dichtel49670822015-01-26 14:10:53 +0100349 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
stephen hemmingerd3428942012-10-01 12:32:35 +0000350 + nla_total_size(sizeof(struct nda_cacheinfo));
351}
352
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200353static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
354 struct vxlan_rdst *rd, int type)
stephen hemmingerd3428942012-10-01 12:32:35 +0000355{
356 struct net *net = dev_net(vxlan->dev);
357 struct sk_buff *skb;
358 int err = -ENOBUFS;
359
360 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
361 if (skb == NULL)
362 goto errout;
363
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200364 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
stephen hemmingerd3428942012-10-01 12:32:35 +0000365 if (err < 0) {
366 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
367 WARN_ON(err == -EMSGSIZE);
368 kfree_skb(skb);
369 goto errout;
370 }
371
372 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
373 return;
374errout:
375 if (err < 0)
376 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
377}
378
Cong Wange4c7ed42013-08-31 13:44:33 +0800379static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
David Stevense4f67ad2012-11-20 02:50:14 +0000380{
381 struct vxlan_dev *vxlan = netdev_priv(dev);
Stephen Hemmingerbb3fd682013-06-17 14:16:40 -0700382 struct vxlan_fdb f = {
383 .state = NUD_STALE,
384 };
385 struct vxlan_rdst remote = {
Cong Wange4c7ed42013-08-31 13:44:33 +0800386 .remote_ip = *ipa, /* goes to NDA_DST */
Stephen Hemmingerbb3fd682013-06-17 14:16:40 -0700387 .remote_vni = VXLAN_N_VID,
388 };
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700389
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200390 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
David Stevense4f67ad2012-11-20 02:50:14 +0000391}
392
393static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
394{
Stephen Hemmingerbb3fd682013-06-17 14:16:40 -0700395 struct vxlan_fdb f = {
396 .state = NUD_STALE,
397 };
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200398 struct vxlan_rdst remote = { };
David Stevense4f67ad2012-11-20 02:50:14 +0000399
David Stevense4f67ad2012-11-20 02:50:14 +0000400 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
401
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200402 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
David Stevense4f67ad2012-11-20 02:50:14 +0000403}
404
stephen hemmingerd3428942012-10-01 12:32:35 +0000405/* Hash Ethernet address */
406static u32 eth_hash(const unsigned char *addr)
407{
408 u64 value = get_unaligned((u64 *)addr);
409
410 /* only want 6 bytes */
411#ifdef __BIG_ENDIAN
stephen hemmingerd3428942012-10-01 12:32:35 +0000412 value >>= 16;
stephen hemminger321fb992012-10-09 20:35:47 +0000413#else
414 value <<= 16;
stephen hemmingerd3428942012-10-01 12:32:35 +0000415#endif
416 return hash_64(value, FDB_HASH_BITS);
417}
418
419/* Hash chain to use given mac address */
420static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
421 const u8 *mac)
422{
423 return &vxlan->fdb_head[eth_hash(mac)];
424}
425
426/* Look up Ethernet address in forwarding table */
Sridhar Samudrala014be2c2013-05-17 06:39:07 +0000427static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
stephen hemmingerd3428942012-10-01 12:32:35 +0000428 const u8 *mac)
stephen hemmingerd3428942012-10-01 12:32:35 +0000429{
430 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
431 struct vxlan_fdb *f;
stephen hemmingerd3428942012-10-01 12:32:35 +0000432
Sasha Levinb67bfe02013-02-27 17:06:00 -0800433 hlist_for_each_entry_rcu(f, head, hlist) {
Joe Perches7367d0b2013-09-01 11:51:23 -0700434 if (ether_addr_equal(mac, f->eth_addr))
stephen hemmingerd3428942012-10-01 12:32:35 +0000435 return f;
436 }
437
438 return NULL;
439}
440
Sridhar Samudrala014be2c2013-05-17 06:39:07 +0000441static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
442 const u8 *mac)
443{
444 struct vxlan_fdb *f;
445
446 f = __vxlan_find_mac(vxlan, mac);
447 if (f)
448 f->used = jiffies;
449
450 return f;
451}
452
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300453/* caller should hold vxlan->hash_lock */
454static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
Cong Wange4c7ed42013-08-31 13:44:33 +0800455 union vxlan_addr *ip, __be16 port,
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300456 __u32 vni, __u32 ifindex)
457{
458 struct vxlan_rdst *rd;
459
460 list_for_each_entry(rd, &f->remotes, list) {
Cong Wange4c7ed42013-08-31 13:44:33 +0800461 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300462 rd->remote_port == port &&
463 rd->remote_vni == vni &&
464 rd->remote_ifindex == ifindex)
465 return rd;
466 }
467
468 return NULL;
469}
470
Thomas Richter906dc182013-07-19 17:20:07 +0200471/* Replace destination of unicast mac */
472static int vxlan_fdb_replace(struct vxlan_fdb *f,
Cong Wange4c7ed42013-08-31 13:44:33 +0800473 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
Thomas Richter906dc182013-07-19 17:20:07 +0200474{
475 struct vxlan_rdst *rd;
476
477 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
478 if (rd)
479 return 0;
480
481 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
482 if (!rd)
483 return 0;
Cong Wange4c7ed42013-08-31 13:44:33 +0800484 rd->remote_ip = *ip;
Thomas Richter906dc182013-07-19 17:20:07 +0200485 rd->remote_port = port;
486 rd->remote_vni = vni;
487 rd->remote_ifindex = ifindex;
488 return 1;
489}
490
David Stevens66817122013-03-15 04:35:51 +0000491/* Add/update destinations for multicast */
492static int vxlan_fdb_append(struct vxlan_fdb *f,
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200493 union vxlan_addr *ip, __be16 port, __u32 vni,
494 __u32 ifindex, struct vxlan_rdst **rdp)
David Stevens66817122013-03-15 04:35:51 +0000495{
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700496 struct vxlan_rdst *rd;
David Stevens66817122013-03-15 04:35:51 +0000497
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300498 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
499 if (rd)
500 return 0;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700501
David Stevens66817122013-03-15 04:35:51 +0000502 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
503 if (rd == NULL)
504 return -ENOBUFS;
Cong Wange4c7ed42013-08-31 13:44:33 +0800505 rd->remote_ip = *ip;
David Stevens66817122013-03-15 04:35:51 +0000506 rd->remote_port = port;
507 rd->remote_vni = vni;
508 rd->remote_ifindex = ifindex;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700509
510 list_add_tail_rcu(&rd->list, &f->remotes);
511
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200512 *rdp = rd;
David Stevens66817122013-03-15 04:35:51 +0000513 return 1;
514}
515
Tom Herbertdfd86452015-01-12 17:00:38 -0800516static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
517 unsigned int off,
518 struct vxlanhdr *vh, size_t hdrlen,
Tom Herbert0ace2ca2015-02-10 16:30:32 -0800519 u32 data, struct gro_remcsum *grc,
520 bool nopartial)
Tom Herbertdfd86452015-01-12 17:00:38 -0800521{
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700522 size_t start, offset;
Tom Herbertdfd86452015-01-12 17:00:38 -0800523
524 if (skb->remcsum_offload)
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700525 return vh;
Tom Herbertdfd86452015-01-12 17:00:38 -0800526
527 if (!NAPI_GRO_CB(skb)->csum_valid)
528 return NULL;
529
530 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
531 offset = start + ((data & VXLAN_RCO_UDP) ?
532 offsetof(struct udphdr, check) :
533 offsetof(struct tcphdr, check));
534
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700535 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
536 start, offset, grc, nopartial);
Tom Herbertdfd86452015-01-12 17:00:38 -0800537
538 skb->remcsum_offload = 1;
539
540 return vh;
541}
542
Tom Herberta2b12f32015-01-12 17:00:37 -0800543static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
544 struct sk_buff *skb,
545 struct udp_offload *uoff)
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200546{
547 struct sk_buff *p, **pp = NULL;
548 struct vxlanhdr *vh, *vh2;
Jesse Gross9b174d82014-12-30 19:10:15 -0800549 unsigned int hlen, off_vx;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200550 int flush = 1;
Tom Herbertdfd86452015-01-12 17:00:38 -0800551 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
552 udp_offloads);
553 u32 flags;
Tom Herbert26c4f7d2015-02-10 16:30:27 -0800554 struct gro_remcsum grc;
555
556 skb_gro_remcsum_init(&grc);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200557
558 off_vx = skb_gro_offset(skb);
559 hlen = off_vx + sizeof(*vh);
560 vh = skb_gro_header_fast(skb, off_vx);
561 if (skb_gro_header_hard(skb, hlen)) {
562 vh = skb_gro_header_slow(skb, hlen, off_vx);
563 if (unlikely(!vh))
564 goto out;
565 }
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200566
Tom Herbertdfd86452015-01-12 17:00:38 -0800567 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
568
569 flags = ntohl(vh->vx_flags);
570
571 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
572 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
Tom Herbert0ace2ca2015-02-10 16:30:32 -0800573 ntohl(vh->vx_vni), &grc,
574 !!(vs->flags &
575 VXLAN_F_REMCSUM_NOPARTIAL));
Tom Herbertdfd86452015-01-12 17:00:38 -0800576
577 if (!vh)
578 goto out;
579 }
580
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700581 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
582
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200583 flush = 0;
584
585 for (p = *head; p; p = p->next) {
586 if (!NAPI_GRO_CB(p)->same_flow)
587 continue;
588
589 vh2 = (struct vxlanhdr *)(p->data + off_vx);
Thomas Graf35114942015-01-15 03:53:55 +0100590 if (vh->vx_flags != vh2->vx_flags ||
591 vh->vx_vni != vh2->vx_vni) {
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200592 NAPI_GRO_CB(p)->same_flow = 0;
593 continue;
594 }
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200595 }
596
Jesse Gross9b174d82014-12-30 19:10:15 -0800597 pp = eth_gro_receive(head, skb);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200598
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200599out:
Tom Herbert26c4f7d2015-02-10 16:30:27 -0800600 skb_gro_remcsum_cleanup(skb, &grc);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200601 NAPI_GRO_CB(skb)->flush |= flush;
602
603 return pp;
604}
605
Tom Herberta2b12f32015-01-12 17:00:37 -0800606static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
607 struct udp_offload *uoff)
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200608{
Jesse Grosscfdf1e12014-11-10 11:45:13 -0800609 udp_tunnel_gro_complete(skb, nhoff);
610
Jesse Gross9b174d82014-12-30 19:10:15 -0800611 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200612}
613
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700614/* Notify netdevs that UDP port started listening */
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200615static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700616{
617 struct net_device *dev;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200618 struct sock *sk = vs->sock->sk;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700619 struct net *net = sock_net(sk);
Jiri Benc705cc622015-08-20 13:56:28 +0200620 sa_family_t sa_family = vxlan_get_sk_family(vs);
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700621 __be16 port = inet_sk(sk)->inet_sport;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200622 int err;
623
624 if (sa_family == AF_INET) {
625 err = udp_add_offload(&vs->udp_offloads);
626 if (err)
627 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
628 }
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700629
630 rcu_read_lock();
631 for_each_netdev_rcu(net, dev) {
632 if (dev->netdev_ops->ndo_add_vxlan_port)
633 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
634 port);
635 }
636 rcu_read_unlock();
637}
638
639/* Notify netdevs that UDP port is no more listening */
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200640static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700641{
642 struct net_device *dev;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200643 struct sock *sk = vs->sock->sk;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700644 struct net *net = sock_net(sk);
Jiri Benc705cc622015-08-20 13:56:28 +0200645 sa_family_t sa_family = vxlan_get_sk_family(vs);
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700646 __be16 port = inet_sk(sk)->inet_sport;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700647
648 rcu_read_lock();
649 for_each_netdev_rcu(net, dev) {
650 if (dev->netdev_ops->ndo_del_vxlan_port)
651 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
652 port);
653 }
654 rcu_read_unlock();
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200655
656 if (sa_family == AF_INET)
657 udp_del_offload(&vs->udp_offloads);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700658}
659
stephen hemmingerd3428942012-10-01 12:32:35 +0000660/* Add new entry to forwarding table -- assumes lock held */
661static int vxlan_fdb_create(struct vxlan_dev *vxlan,
Cong Wange4c7ed42013-08-31 13:44:33 +0800662 const u8 *mac, union vxlan_addr *ip,
David Stevens66817122013-03-15 04:35:51 +0000663 __u16 state, __u16 flags,
stephen hemminger73cf3312013-04-27 11:31:54 +0000664 __be16 port, __u32 vni, __u32 ifindex,
David Stevensae884082013-04-19 00:36:26 +0000665 __u8 ndm_flags)
stephen hemmingerd3428942012-10-01 12:32:35 +0000666{
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200667 struct vxlan_rdst *rd = NULL;
stephen hemmingerd3428942012-10-01 12:32:35 +0000668 struct vxlan_fdb *f;
669 int notify = 0;
670
Sridhar Samudrala014be2c2013-05-17 06:39:07 +0000671 f = __vxlan_find_mac(vxlan, mac);
stephen hemmingerd3428942012-10-01 12:32:35 +0000672 if (f) {
673 if (flags & NLM_F_EXCL) {
674 netdev_dbg(vxlan->dev,
675 "lost race to create %pM\n", mac);
676 return -EEXIST;
677 }
678 if (f->state != state) {
679 f->state = state;
680 f->updated = jiffies;
681 notify = 1;
682 }
David Stevensae884082013-04-19 00:36:26 +0000683 if (f->flags != ndm_flags) {
684 f->flags = ndm_flags;
685 f->updated = jiffies;
686 notify = 1;
687 }
Thomas Richter906dc182013-07-19 17:20:07 +0200688 if ((flags & NLM_F_REPLACE)) {
689 /* Only change unicasts */
690 if (!(is_multicast_ether_addr(f->eth_addr) ||
691 is_zero_ether_addr(f->eth_addr))) {
Li RongQing60840422015-04-22 15:49:10 +0800692 notify |= vxlan_fdb_replace(f, ip, port, vni,
Thomas Richter906dc182013-07-19 17:20:07 +0200693 ifindex);
Thomas Richter906dc182013-07-19 17:20:07 +0200694 } else
695 return -EOPNOTSUPP;
696 }
David Stevens66817122013-03-15 04:35:51 +0000697 if ((flags & NLM_F_APPEND) &&
Mike Rapoport58e4c762013-06-25 16:01:56 +0300698 (is_multicast_ether_addr(f->eth_addr) ||
699 is_zero_ether_addr(f->eth_addr))) {
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200700 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
701 &rd);
David Stevens66817122013-03-15 04:35:51 +0000702
703 if (rc < 0)
704 return rc;
705 notify |= rc;
706 }
stephen hemmingerd3428942012-10-01 12:32:35 +0000707 } else {
708 if (!(flags & NLM_F_CREATE))
709 return -ENOENT;
710
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200711 if (vxlan->cfg.addrmax &&
712 vxlan->addrcnt >= vxlan->cfg.addrmax)
stephen hemmingerd3428942012-10-01 12:32:35 +0000713 return -ENOSPC;
714
Thomas Richter906dc182013-07-19 17:20:07 +0200715 /* Disallow replace to add a multicast entry */
716 if ((flags & NLM_F_REPLACE) &&
717 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
718 return -EOPNOTSUPP;
719
Cong Wange4c7ed42013-08-31 13:44:33 +0800720 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
stephen hemmingerd3428942012-10-01 12:32:35 +0000721 f = kmalloc(sizeof(*f), GFP_ATOMIC);
722 if (!f)
723 return -ENOMEM;
724
725 notify = 1;
stephen hemmingerd3428942012-10-01 12:32:35 +0000726 f->state = state;
David Stevensae884082013-04-19 00:36:26 +0000727 f->flags = ndm_flags;
stephen hemmingerd3428942012-10-01 12:32:35 +0000728 f->updated = f->used = jiffies;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700729 INIT_LIST_HEAD(&f->remotes);
stephen hemmingerd3428942012-10-01 12:32:35 +0000730 memcpy(f->eth_addr, mac, ETH_ALEN);
731
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200732 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700733
stephen hemmingerd3428942012-10-01 12:32:35 +0000734 ++vxlan->addrcnt;
735 hlist_add_head_rcu(&f->hlist,
736 vxlan_fdb_head(vxlan, mac));
737 }
738
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200739 if (notify) {
740 if (rd == NULL)
741 rd = first_remote_rtnl(f);
742 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
743 }
stephen hemmingerd3428942012-10-01 12:32:35 +0000744
745 return 0;
746}
747
Wei Yongjun6706c822013-04-11 19:00:35 +0000748static void vxlan_fdb_free(struct rcu_head *head)
David Stevens66817122013-03-15 04:35:51 +0000749{
750 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700751 struct vxlan_rdst *rd, *nd;
David Stevens66817122013-03-15 04:35:51 +0000752
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700753 list_for_each_entry_safe(rd, nd, &f->remotes, list)
David Stevens66817122013-03-15 04:35:51 +0000754 kfree(rd);
David Stevens66817122013-03-15 04:35:51 +0000755 kfree(f);
756}
757
stephen hemmingerd3428942012-10-01 12:32:35 +0000758static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
759{
760 netdev_dbg(vxlan->dev,
761 "delete %pM\n", f->eth_addr);
762
763 --vxlan->addrcnt;
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200764 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
stephen hemmingerd3428942012-10-01 12:32:35 +0000765
766 hlist_del_rcu(&f->hlist);
David Stevens66817122013-03-15 04:35:51 +0000767 call_rcu(&f->rcu, vxlan_fdb_free);
stephen hemmingerd3428942012-10-01 12:32:35 +0000768}
769
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300770static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
Cong Wange4c7ed42013-08-31 13:44:33 +0800771 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300772{
773 struct net *net = dev_net(vxlan->dev);
Cong Wange4c7ed42013-08-31 13:44:33 +0800774 int err;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300775
776 if (tb[NDA_DST]) {
Cong Wange4c7ed42013-08-31 13:44:33 +0800777 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
778 if (err)
779 return err;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300780 } else {
Cong Wange4c7ed42013-08-31 13:44:33 +0800781 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
782 if (remote->sa.sa_family == AF_INET) {
783 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
784 ip->sa.sa_family = AF_INET;
785#if IS_ENABLED(CONFIG_IPV6)
786 } else {
787 ip->sin6.sin6_addr = in6addr_any;
788 ip->sa.sa_family = AF_INET6;
789#endif
790 }
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300791 }
792
793 if (tb[NDA_PORT]) {
794 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
795 return -EINVAL;
796 *port = nla_get_be16(tb[NDA_PORT]);
797 } else {
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200798 *port = vxlan->cfg.dst_port;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300799 }
800
801 if (tb[NDA_VNI]) {
802 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
803 return -EINVAL;
804 *vni = nla_get_u32(tb[NDA_VNI]);
805 } else {
806 *vni = vxlan->default_dst.remote_vni;
807 }
808
809 if (tb[NDA_IFINDEX]) {
810 struct net_device *tdev;
811
812 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
813 return -EINVAL;
814 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
Ying Xue73763942014-01-15 10:23:41 +0800815 tdev = __dev_get_by_index(net, *ifindex);
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300816 if (!tdev)
817 return -EADDRNOTAVAIL;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300818 } else {
819 *ifindex = 0;
820 }
821
822 return 0;
823}
824
stephen hemmingerd3428942012-10-01 12:32:35 +0000825/* Add static entry (via netlink) */
826static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
827 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100828 const unsigned char *addr, u16 vid, u16 flags)
stephen hemmingerd3428942012-10-01 12:32:35 +0000829{
830 struct vxlan_dev *vxlan = netdev_priv(dev);
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300831 /* struct net *net = dev_net(vxlan->dev); */
Cong Wange4c7ed42013-08-31 13:44:33 +0800832 union vxlan_addr ip;
stephen hemminger73cf3312013-04-27 11:31:54 +0000833 __be16 port;
834 u32 vni, ifindex;
stephen hemmingerd3428942012-10-01 12:32:35 +0000835 int err;
836
837 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
838 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
839 ndm->ndm_state);
840 return -EINVAL;
841 }
842
843 if (tb[NDA_DST] == NULL)
844 return -EINVAL;
845
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300846 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
847 if (err)
848 return err;
David Stevens66817122013-03-15 04:35:51 +0000849
Mike Rapoport5933a7b2014-04-01 09:23:01 +0300850 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
851 return -EAFNOSUPPORT;
852
stephen hemmingerd3428942012-10-01 12:32:35 +0000853 spin_lock_bh(&vxlan->hash_lock);
Cong Wange4c7ed42013-08-31 13:44:33 +0800854 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
stephen hemminger73cf3312013-04-27 11:31:54 +0000855 port, vni, ifindex, ndm->ndm_flags);
stephen hemmingerd3428942012-10-01 12:32:35 +0000856 spin_unlock_bh(&vxlan->hash_lock);
857
858 return err;
859}
860
861/* Delete entry (via netlink) */
Vlad Yasevich1690be62013-02-13 12:00:18 +0000862static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
863 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100864 const unsigned char *addr, u16 vid)
stephen hemmingerd3428942012-10-01 12:32:35 +0000865{
866 struct vxlan_dev *vxlan = netdev_priv(dev);
867 struct vxlan_fdb *f;
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300868 struct vxlan_rdst *rd = NULL;
Cong Wange4c7ed42013-08-31 13:44:33 +0800869 union vxlan_addr ip;
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300870 __be16 port;
871 u32 vni, ifindex;
872 int err;
873
874 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
875 if (err)
876 return err;
877
878 err = -ENOENT;
stephen hemmingerd3428942012-10-01 12:32:35 +0000879
880 spin_lock_bh(&vxlan->hash_lock);
881 f = vxlan_find_mac(vxlan, addr);
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300882 if (!f)
883 goto out;
884
Cong Wange4c7ed42013-08-31 13:44:33 +0800885 if (!vxlan_addr_any(&ip)) {
886 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300887 if (!rd)
888 goto out;
stephen hemmingerd3428942012-10-01 12:32:35 +0000889 }
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300890
891 err = 0;
892
893 /* remove a destination if it's not the only one on the list,
894 * otherwise destroy the fdb entry
895 */
896 if (rd && !list_is_singular(&f->remotes)) {
897 list_del_rcu(&rd->list);
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200898 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
Wei Yongjundcdc7a52013-08-17 07:32:09 +0800899 kfree_rcu(rd, rcu);
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300900 goto out;
901 }
902
903 vxlan_fdb_destroy(vxlan, f);
904
905out:
stephen hemmingerd3428942012-10-01 12:32:35 +0000906 spin_unlock_bh(&vxlan->hash_lock);
907
908 return err;
909}
910
911/* Dump forwarding table */
912static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -0400913 struct net_device *dev,
914 struct net_device *filter_dev, int idx)
stephen hemmingerd3428942012-10-01 12:32:35 +0000915{
916 struct vxlan_dev *vxlan = netdev_priv(dev);
917 unsigned int h;
918
919 for (h = 0; h < FDB_HASH_SIZE; ++h) {
920 struct vxlan_fdb *f;
stephen hemmingerd3428942012-10-01 12:32:35 +0000921 int err;
922
Sasha Levinb67bfe02013-02-27 17:06:00 -0800923 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
David Stevens66817122013-03-15 04:35:51 +0000924 struct vxlan_rdst *rd;
stephen hemmingerd3428942012-10-01 12:32:35 +0000925
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700926 list_for_each_entry_rcu(rd, &f->remotes, list) {
Atzm Watanabe07a51cd2015-08-10 23:39:09 +0900927 if (idx < cb->args[0])
928 goto skip;
929
David Stevens66817122013-03-15 04:35:51 +0000930 err = vxlan_fdb_info(skb, vxlan, f,
931 NETLINK_CB(cb->skb).portid,
932 cb->nlh->nlmsg_seq,
933 RTM_NEWNEIGH,
934 NLM_F_MULTI, rd);
935 if (err < 0)
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700936 goto out;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700937skip:
Atzm Watanabe07a51cd2015-08-10 23:39:09 +0900938 ++idx;
939 }
stephen hemmingerd3428942012-10-01 12:32:35 +0000940 }
941 }
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700942out:
stephen hemmingerd3428942012-10-01 12:32:35 +0000943 return idx;
944}
945
946/* Watch incoming packets to learn mapping between Ethernet address
947 * and Tunnel endpoint.
Simon Hormanc4b49512015-04-02 11:17:58 +0900948 * Return true if packet is bogus and should be dropped.
stephen hemmingerd3428942012-10-01 12:32:35 +0000949 */
stephen hemminger26a41ae62013-06-17 12:09:58 -0700950static bool vxlan_snoop(struct net_device *dev,
Cong Wange4c7ed42013-08-31 13:44:33 +0800951 union vxlan_addr *src_ip, const u8 *src_mac)
stephen hemmingerd3428942012-10-01 12:32:35 +0000952{
953 struct vxlan_dev *vxlan = netdev_priv(dev);
954 struct vxlan_fdb *f;
stephen hemmingerd3428942012-10-01 12:32:35 +0000955
956 f = vxlan_find_mac(vxlan, src_mac);
957 if (likely(f)) {
stephen hemminger5ca54612013-08-04 17:17:39 -0700958 struct vxlan_rdst *rdst = first_remote_rcu(f);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700959
Cong Wange4c7ed42013-08-31 13:44:33 +0800960 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
stephen hemminger26a41ae62013-06-17 12:09:58 -0700961 return false;
962
963 /* Don't migrate static entries, drop packets */
stephen hemmingereb064c32013-06-18 14:27:01 -0700964 if (f->state & NUD_NOARP)
stephen hemminger26a41ae62013-06-17 12:09:58 -0700965 return true;
stephen hemmingerd3428942012-10-01 12:32:35 +0000966
967 if (net_ratelimit())
968 netdev_info(dev,
Cong Wange4c7ed42013-08-31 13:44:33 +0800969 "%pM migrated from %pIS to %pIS\n",
Rasmus Villemoesa4870f72015-02-07 03:17:31 +0100970 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
stephen hemmingerd3428942012-10-01 12:32:35 +0000971
Cong Wange4c7ed42013-08-31 13:44:33 +0800972 rdst->remote_ip = *src_ip;
stephen hemmingerd3428942012-10-01 12:32:35 +0000973 f->updated = jiffies;
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200974 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
stephen hemmingerd3428942012-10-01 12:32:35 +0000975 } else {
976 /* learned new entry */
977 spin_lock(&vxlan->hash_lock);
stephen hemminger3bf74b12013-06-17 12:09:57 -0700978
979 /* close off race between vxlan_flush and incoming packets */
980 if (netif_running(dev))
981 vxlan_fdb_create(vxlan, src_mac, src_ip,
982 NUD_REACHABLE,
983 NLM_F_EXCL|NLM_F_CREATE,
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200984 vxlan->cfg.dst_port,
stephen hemminger3bf74b12013-06-17 12:09:57 -0700985 vxlan->default_dst.remote_vni,
986 0, NTF_SELF);
stephen hemmingerd3428942012-10-01 12:32:35 +0000987 spin_unlock(&vxlan->hash_lock);
988 }
stephen hemminger26a41ae62013-06-17 12:09:58 -0700989
990 return false;
stephen hemmingerd3428942012-10-01 12:32:35 +0000991}
992
stephen hemmingerd3428942012-10-01 12:32:35 +0000993/* See if multicast group is already in use by other ID */
Gao feng95ab0992013-12-10 16:37:33 +0800994static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
stephen hemmingerd3428942012-10-01 12:32:35 +0000995{
stephen hemminger553675f2013-05-16 11:35:20 +0000996 struct vxlan_dev *vxlan;
stephen hemmingerd3428942012-10-01 12:32:35 +0000997
Gao feng95ab0992013-12-10 16:37:33 +0800998 /* The vxlan_sock is only used by dev, leaving group has
999 * no effect on other vxlan devices.
1000 */
1001 if (atomic_read(&dev->vn_sock->refcnt) == 1)
1002 return false;
1003
stephen hemminger553675f2013-05-16 11:35:20 +00001004 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
Gao feng95ab0992013-12-10 16:37:33 +08001005 if (!netif_running(vxlan->dev) || vxlan == dev)
stephen hemminger553675f2013-05-16 11:35:20 +00001006 continue;
stephen hemmingerd3428942012-10-01 12:32:35 +00001007
Gao feng95ab0992013-12-10 16:37:33 +08001008 if (vxlan->vn_sock != dev->vn_sock)
1009 continue;
1010
1011 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1012 &dev->default_dst.remote_ip))
1013 continue;
1014
1015 if (vxlan->default_dst.remote_ifindex !=
1016 dev->default_dst.remote_ifindex)
1017 continue;
1018
1019 return true;
stephen hemminger553675f2013-05-16 11:35:20 +00001020 }
stephen hemmingerd3428942012-10-01 12:32:35 +00001021
1022 return false;
1023}
1024
Thomas Graf614732e2015-07-21 10:44:06 +02001025static void vxlan_sock_release(struct vxlan_sock *vs)
stephen hemmingerd3428942012-10-01 12:32:35 +00001026{
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001027 struct sock *sk = vs->sock->sk;
1028 struct net *net = sock_net(sk);
1029 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
Pravin B Shelar012a5722013-08-19 11:23:07 -07001030
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001031 if (!atomic_dec_and_test(&vs->refcnt))
1032 return;
1033
Stephen Hemminger1c51a912013-06-17 14:16:11 -07001034 spin_lock(&vn->sock_lock);
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001035 hlist_del_rcu(&vs->hlist);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02001036 vxlan_notify_del_rx_port(vs);
Stephen Hemminger1c51a912013-06-17 14:16:11 -07001037 spin_unlock(&vn->sock_lock);
1038
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001039 queue_work(vxlan_wq, &vs->del_work);
1040}
1041
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001042/* Update multicast group membership when first VNI on
Simon Hormanc4b49512015-04-02 11:17:58 +09001043 * multicast address is brought up
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001044 */
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001045static int vxlan_igmp_join(struct vxlan_dev *vxlan)
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001046{
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001047 struct vxlan_sock *vs = vxlan->vn_sock;
1048 struct sock *sk = vs->sock->sk;
Cong Wange4c7ed42013-08-31 13:44:33 +08001049 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1050 int ifindex = vxlan->default_dst.remote_ifindex;
Marcelo Ricardo Leitner149d7542015-03-20 10:26:21 -03001051 int ret = -EINVAL;
stephen hemmingerd3428942012-10-01 12:32:35 +00001052
stephen hemmingerd3428942012-10-01 12:32:35 +00001053 lock_sock(sk);
Cong Wange4c7ed42013-08-31 13:44:33 +08001054 if (ip->sa.sa_family == AF_INET) {
1055 struct ip_mreqn mreq = {
1056 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1057 .imr_ifindex = ifindex,
1058 };
1059
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001060 ret = ip_mc_join_group(sk, &mreq);
Cong Wange4c7ed42013-08-31 13:44:33 +08001061#if IS_ENABLED(CONFIG_IPV6)
1062 } else {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001063 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1064 &ip->sin6.sin6_addr);
Cong Wange4c7ed42013-08-31 13:44:33 +08001065#endif
1066 }
stephen hemminger3fc2de22013-07-18 08:40:15 -07001067 release_sock(sk);
1068
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001069 return ret;
stephen hemminger3fc2de22013-07-18 08:40:15 -07001070}
1071
1072/* Inverse of vxlan_igmp_join when last VNI is brought down */
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001073static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
stephen hemminger3fc2de22013-07-18 08:40:15 -07001074{
stephen hemminger3fc2de22013-07-18 08:40:15 -07001075 struct vxlan_sock *vs = vxlan->vn_sock;
1076 struct sock *sk = vs->sock->sk;
Cong Wange4c7ed42013-08-31 13:44:33 +08001077 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1078 int ifindex = vxlan->default_dst.remote_ifindex;
Marcelo Ricardo Leitner149d7542015-03-20 10:26:21 -03001079 int ret = -EINVAL;
stephen hemminger3fc2de22013-07-18 08:40:15 -07001080
1081 lock_sock(sk);
Cong Wange4c7ed42013-08-31 13:44:33 +08001082 if (ip->sa.sa_family == AF_INET) {
1083 struct ip_mreqn mreq = {
1084 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1085 .imr_ifindex = ifindex,
1086 };
1087
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001088 ret = ip_mc_leave_group(sk, &mreq);
Cong Wange4c7ed42013-08-31 13:44:33 +08001089#if IS_ENABLED(CONFIG_IPV6)
1090 } else {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001091 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1092 &ip->sin6.sin6_addr);
Cong Wange4c7ed42013-08-31 13:44:33 +08001093#endif
1094 }
stephen hemmingerd3428942012-10-01 12:32:35 +00001095 release_sock(sk);
stephen hemmingerd3428942012-10-01 12:32:35 +00001096
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001097 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00001098}
1099
Tom Herbertdfd86452015-01-12 17:00:38 -08001100static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
Tom Herbert0ace2ca2015-02-10 16:30:32 -08001101 size_t hdrlen, u32 data, bool nopartial)
Tom Herbertdfd86452015-01-12 17:00:38 -08001102{
1103 size_t start, offset, plen;
Tom Herbertdfd86452015-01-12 17:00:38 -08001104
Tom Herbertb7fe10e2015-08-19 17:07:32 -07001105 if (skb->remcsum_offload)
1106 return vh;
1107
Tom Herbertdfd86452015-01-12 17:00:38 -08001108 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1109 offset = start + ((data & VXLAN_RCO_UDP) ?
1110 offsetof(struct udphdr, check) :
1111 offsetof(struct tcphdr, check));
1112
1113 plen = hdrlen + offset + sizeof(u16);
1114
1115 if (!pskb_may_pull(skb, plen))
1116 return NULL;
1117
1118 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1119
Tom Herbert0ace2ca2015-02-10 16:30:32 -08001120 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1121 nopartial);
Tom Herbertdfd86452015-01-12 17:00:38 -08001122
1123 return vh;
1124}
1125
Thomas Graf614732e2015-07-21 10:44:06 +02001126static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1127 struct vxlan_metadata *md, u32 vni,
1128 struct metadata_dst *tun_dst)
1129{
1130 struct iphdr *oip = NULL;
1131 struct ipv6hdr *oip6 = NULL;
1132 struct vxlan_dev *vxlan;
1133 struct pcpu_sw_netstats *stats;
1134 union vxlan_addr saddr;
1135 int err = 0;
1136 union vxlan_addr *remote_ip;
1137
1138 /* For flow based devices, map all packets to VNI 0 */
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07001139 if (vs->flags & VXLAN_F_COLLECT_METADATA)
Thomas Graf614732e2015-07-21 10:44:06 +02001140 vni = 0;
1141
1142 /* Is this VNI defined? */
1143 vxlan = vxlan_vs_find_vni(vs, vni);
1144 if (!vxlan)
1145 goto drop;
1146
1147 remote_ip = &vxlan->default_dst.remote_ip;
1148 skb_reset_mac_header(skb);
1149 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1150 skb->protocol = eth_type_trans(skb, vxlan->dev);
1151 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1152
1153 /* Ignore packet loops (and multicast echo) */
1154 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1155 goto drop;
1156
1157 /* Re-examine inner Ethernet packet */
1158 if (remote_ip->sa.sa_family == AF_INET) {
1159 oip = ip_hdr(skb);
1160 saddr.sin.sin_addr.s_addr = oip->saddr;
1161 saddr.sa.sa_family = AF_INET;
1162#if IS_ENABLED(CONFIG_IPV6)
1163 } else {
1164 oip6 = ipv6_hdr(skb);
1165 saddr.sin6.sin6_addr = oip6->saddr;
1166 saddr.sa.sa_family = AF_INET6;
1167#endif
1168 }
1169
1170 if (tun_dst) {
1171 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1172 tun_dst = NULL;
1173 }
1174
1175 if ((vxlan->flags & VXLAN_F_LEARN) &&
1176 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1177 goto drop;
1178
1179 skb_reset_network_header(skb);
1180 /* In flow-based mode, GBP is carried in dst_metadata */
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07001181 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
Thomas Graf614732e2015-07-21 10:44:06 +02001182 skb->mark = md->gbp;
1183
1184 if (oip6)
1185 err = IP6_ECN_decapsulate(oip6, skb);
1186 if (oip)
1187 err = IP_ECN_decapsulate(oip, skb);
1188
1189 if (unlikely(err)) {
1190 if (log_ecn_error) {
1191 if (oip6)
1192 net_info_ratelimited("non-ECT from %pI6\n",
1193 &oip6->saddr);
1194 if (oip)
1195 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1196 &oip->saddr, oip->tos);
1197 }
1198 if (err > 1) {
1199 ++vxlan->dev->stats.rx_frame_errors;
1200 ++vxlan->dev->stats.rx_errors;
1201 goto drop;
1202 }
1203 }
1204
1205 stats = this_cpu_ptr(vxlan->dev->tstats);
1206 u64_stats_update_begin(&stats->syncp);
1207 stats->rx_packets++;
1208 stats->rx_bytes += skb->len;
1209 u64_stats_update_end(&stats->syncp);
1210
Tom Herbert58ce31c2015-08-19 17:07:33 -07001211 gro_cells_receive(&vxlan->gro_cells, skb);
Thomas Graf614732e2015-07-21 10:44:06 +02001212
1213 return;
1214drop:
1215 if (tun_dst)
1216 dst_release((struct dst_entry *)tun_dst);
1217
1218 /* Consume bad packet */
1219 kfree_skb(skb);
1220}
1221
stephen hemmingerd3428942012-10-01 12:32:35 +00001222/* Callback from net/ipv4/udp.c to receive packets */
1223static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1224{
Thomas Grafee122c72015-07-21 10:43:58 +02001225 struct metadata_dst *tun_dst = NULL;
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001226 struct vxlan_sock *vs;
stephen hemmingerd3428942012-10-01 12:32:35 +00001227 struct vxlanhdr *vxh;
Tom Herbert3bf39472015-01-08 12:31:18 -08001228 u32 flags, vni;
Thomas Grafee122c72015-07-21 10:43:58 +02001229 struct vxlan_metadata _md;
1230 struct vxlan_metadata *md = &_md;
stephen hemmingerd3428942012-10-01 12:32:35 +00001231
stephen hemmingerd3428942012-10-01 12:32:35 +00001232 /* Need Vxlan and inner Ethernet header to be present */
Pravin B Shelar7ce04752013-08-19 11:22:54 -07001233 if (!pskb_may_pull(skb, VXLAN_HLEN))
stephen hemmingerd3428942012-10-01 12:32:35 +00001234 goto error;
1235
Pravin B Shelar7ce04752013-08-19 11:22:54 -07001236 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
Tom Herbert3bf39472015-01-08 12:31:18 -08001237 flags = ntohl(vxh->vx_flags);
1238 vni = ntohl(vxh->vx_vni);
1239
1240 if (flags & VXLAN_HF_VNI) {
1241 flags &= ~VXLAN_HF_VNI;
1242 } else {
1243 /* VNI flag always required to be set */
1244 goto bad_flags;
stephen hemmingerd3428942012-10-01 12:32:35 +00001245 }
1246
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001247 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
stephen hemmingerd3428942012-10-01 12:32:35 +00001248 goto drop;
Tom Herbertdfd86452015-01-12 17:00:38 -08001249 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
stephen hemmingerd3428942012-10-01 12:32:35 +00001250
Pravin B Shelar559835ea2013-09-24 10:25:40 -07001251 vs = rcu_dereference_sk_user_data(sk);
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001252 if (!vs)
stephen hemmingerd3428942012-10-01 12:32:35 +00001253 goto drop;
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001254
Tom Herbertdfd86452015-01-12 17:00:38 -08001255 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
Tom Herbert0ace2ca2015-02-10 16:30:32 -08001256 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1257 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
Tom Herbertdfd86452015-01-12 17:00:38 -08001258 if (!vxh)
1259 goto drop;
1260
1261 flags &= ~VXLAN_HF_RCO;
Alexey Kodanev40fb70f2015-03-13 19:13:53 +03001262 vni &= VXLAN_VNI_MASK;
Tom Herbertdfd86452015-01-12 17:00:38 -08001263 }
1264
Thomas Grafee122c72015-07-21 10:43:58 +02001265 if (vxlan_collect_metadata(vs)) {
Pravin B Shelarc29a70d2015-08-26 23:46:50 -07001266 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1267 cpu_to_be64(vni >> 8), sizeof(*md));
1268
Thomas Grafee122c72015-07-21 10:43:58 +02001269 if (!tun_dst)
1270 goto drop;
1271
Geert Uytterhoeven0f1b7352015-09-04 12:49:32 +02001272 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
Thomas Grafee122c72015-07-21 10:43:58 +02001273 } else {
1274 memset(md, 0, sizeof(*md));
1275 }
1276
Thomas Graf35114942015-01-15 03:53:55 +01001277 /* For backwards compatibility, only allow reserved fields to be
1278 * used by VXLAN extensions if explicitly requested.
1279 */
1280 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1281 struct vxlanhdr_gbp *gbp;
1282
1283 gbp = (struct vxlanhdr_gbp *)vxh;
Thomas Grafee122c72015-07-21 10:43:58 +02001284 md->gbp = ntohs(gbp->policy_id);
1285
1286 if (tun_dst)
Geert Uytterhoeven0f1b7352015-09-04 12:49:32 +02001287 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
Thomas Graf35114942015-01-15 03:53:55 +01001288
1289 if (gbp->dont_learn)
Thomas Grafee122c72015-07-21 10:43:58 +02001290 md->gbp |= VXLAN_GBP_DONT_LEARN;
Thomas Graf35114942015-01-15 03:53:55 +01001291
1292 if (gbp->policy_applied)
Thomas Grafee122c72015-07-21 10:43:58 +02001293 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
Thomas Graf35114942015-01-15 03:53:55 +01001294
1295 flags &= ~VXLAN_GBP_USED_BITS;
1296 }
1297
Alexey Kodanev40fb70f2015-03-13 19:13:53 +03001298 if (flags || vni & ~VXLAN_VNI_MASK) {
Tom Herbert3bf39472015-01-08 12:31:18 -08001299 /* If there are any unprocessed flags remaining treat
1300 * this as a malformed packet. This behavior diverges from
1301 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1302 * in reserved fields are to be ignored. The approach here
Simon Hormanc4b49512015-04-02 11:17:58 +09001303 * maintains compatibility with previous stack code, and also
Tom Herbert3bf39472015-01-08 12:31:18 -08001304 * is more robust and provides a little more security in
1305 * adding extensions to VXLAN.
1306 */
1307
1308 goto bad_flags;
1309 }
1310
Thomas Graf614732e2015-07-21 10:44:06 +02001311 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001312 return 0;
1313
1314drop:
1315 /* Consume bad packet */
1316 kfree_skb(skb);
1317 return 0;
1318
Tom Herbert3bf39472015-01-08 12:31:18 -08001319bad_flags:
1320 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1321 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1322
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001323error:
Thomas Grafee122c72015-07-21 10:43:58 +02001324 if (tun_dst)
1325 dst_release((struct dst_entry *)tun_dst);
1326
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001327 /* Return non vxlan pkt */
1328 return 1;
1329}
1330
David Stevense4f67ad2012-11-20 02:50:14 +00001331static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1332{
1333 struct vxlan_dev *vxlan = netdev_priv(dev);
1334 struct arphdr *parp;
1335 u8 *arpptr, *sha;
1336 __be32 sip, tip;
1337 struct neighbour *n;
1338
1339 if (dev->flags & IFF_NOARP)
1340 goto out;
1341
1342 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1343 dev->stats.tx_dropped++;
1344 goto out;
1345 }
1346 parp = arp_hdr(skb);
1347
1348 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1349 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1350 parp->ar_pro != htons(ETH_P_IP) ||
1351 parp->ar_op != htons(ARPOP_REQUEST) ||
1352 parp->ar_hln != dev->addr_len ||
1353 parp->ar_pln != 4)
1354 goto out;
1355 arpptr = (u8 *)parp + sizeof(struct arphdr);
1356 sha = arpptr;
1357 arpptr += dev->addr_len; /* sha */
1358 memcpy(&sip, arpptr, sizeof(sip));
1359 arpptr += sizeof(sip);
1360 arpptr += dev->addr_len; /* tha */
1361 memcpy(&tip, arpptr, sizeof(tip));
1362
1363 if (ipv4_is_loopback(tip) ||
1364 ipv4_is_multicast(tip))
1365 goto out;
1366
1367 n = neigh_lookup(&arp_tbl, &tip, dev);
1368
1369 if (n) {
David Stevense4f67ad2012-11-20 02:50:14 +00001370 struct vxlan_fdb *f;
1371 struct sk_buff *reply;
1372
1373 if (!(n->nud_state & NUD_CONNECTED)) {
1374 neigh_release(n);
1375 goto out;
1376 }
1377
1378 f = vxlan_find_mac(vxlan, n->ha);
Cong Wange4c7ed42013-08-31 13:44:33 +08001379 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
David Stevense4f67ad2012-11-20 02:50:14 +00001380 /* bridge-local neighbor */
1381 neigh_release(n);
1382 goto out;
1383 }
1384
1385 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1386 n->ha, sha);
1387
1388 neigh_release(n);
1389
David Stevens73461352014-03-18 12:32:29 -04001390 if (reply == NULL)
1391 goto out;
1392
David Stevense4f67ad2012-11-20 02:50:14 +00001393 skb_reset_mac_header(reply);
1394 __skb_pull(reply, skb_network_offset(reply));
1395 reply->ip_summed = CHECKSUM_UNNECESSARY;
1396 reply->pkt_type = PACKET_HOST;
1397
1398 if (netif_rx_ni(reply) == NET_RX_DROP)
1399 dev->stats.rx_dropped++;
Cong Wange4c7ed42013-08-31 13:44:33 +08001400 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1401 union vxlan_addr ipa = {
1402 .sin.sin_addr.s_addr = tip,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001403 .sin.sin_family = AF_INET,
Cong Wange4c7ed42013-08-31 13:44:33 +08001404 };
1405
1406 vxlan_ip_miss(dev, &ipa);
1407 }
David Stevense4f67ad2012-11-20 02:50:14 +00001408out:
1409 consume_skb(skb);
1410 return NETDEV_TX_OK;
1411}
1412
Cong Wangf564f452013-08-31 13:44:36 +08001413#if IS_ENABLED(CONFIG_IPV6)
David Stevens4b29dba2014-03-24 10:39:58 -04001414static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1415 struct neighbour *n, bool isrouter)
1416{
1417 struct net_device *dev = request->dev;
1418 struct sk_buff *reply;
1419 struct nd_msg *ns, *na;
1420 struct ipv6hdr *pip6;
1421 u8 *daddr;
1422 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1423 int ns_olen;
1424 int i, len;
1425
1426 if (dev == NULL)
1427 return NULL;
1428
1429 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1430 sizeof(*na) + na_olen + dev->needed_tailroom;
1431 reply = alloc_skb(len, GFP_ATOMIC);
1432 if (reply == NULL)
1433 return NULL;
1434
1435 reply->protocol = htons(ETH_P_IPV6);
1436 reply->dev = dev;
1437 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1438 skb_push(reply, sizeof(struct ethhdr));
1439 skb_set_mac_header(reply, 0);
1440
1441 ns = (struct nd_msg *)skb_transport_header(request);
1442
1443 daddr = eth_hdr(request)->h_source;
1444 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1445 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1446 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1447 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1448 break;
1449 }
1450 }
1451
1452 /* Ethernet header */
1453 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1454 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1455 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1456 reply->protocol = htons(ETH_P_IPV6);
1457
1458 skb_pull(reply, sizeof(struct ethhdr));
1459 skb_set_network_header(reply, 0);
1460 skb_put(reply, sizeof(struct ipv6hdr));
1461
1462 /* IPv6 header */
1463
1464 pip6 = ipv6_hdr(reply);
1465 memset(pip6, 0, sizeof(struct ipv6hdr));
1466 pip6->version = 6;
1467 pip6->priority = ipv6_hdr(request)->priority;
1468 pip6->nexthdr = IPPROTO_ICMPV6;
1469 pip6->hop_limit = 255;
1470 pip6->daddr = ipv6_hdr(request)->saddr;
1471 pip6->saddr = *(struct in6_addr *)n->primary_key;
1472
1473 skb_pull(reply, sizeof(struct ipv6hdr));
1474 skb_set_transport_header(reply, 0);
1475
1476 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1477
1478 /* Neighbor Advertisement */
1479 memset(na, 0, sizeof(*na)+na_olen);
1480 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1481 na->icmph.icmp6_router = isrouter;
1482 na->icmph.icmp6_override = 1;
1483 na->icmph.icmp6_solicited = 1;
1484 na->target = ns->target;
1485 ether_addr_copy(&na->opt[2], n->ha);
1486 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1487 na->opt[1] = na_olen >> 3;
1488
1489 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1490 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1491 csum_partial(na, sizeof(*na)+na_olen, 0));
1492
1493 pip6->payload_len = htons(sizeof(*na)+na_olen);
1494
1495 skb_push(reply, sizeof(struct ipv6hdr));
1496
1497 reply->ip_summed = CHECKSUM_UNNECESSARY;
1498
1499 return reply;
1500}
1501
Cong Wangf564f452013-08-31 13:44:36 +08001502static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1503{
1504 struct vxlan_dev *vxlan = netdev_priv(dev);
David Stevens4b29dba2014-03-24 10:39:58 -04001505 struct nd_msg *msg;
Cong Wangf564f452013-08-31 13:44:36 +08001506 const struct ipv6hdr *iphdr;
1507 const struct in6_addr *saddr, *daddr;
David Stevens4b29dba2014-03-24 10:39:58 -04001508 struct neighbour *n;
1509 struct inet6_dev *in6_dev;
Cong Wangf564f452013-08-31 13:44:36 +08001510
1511 in6_dev = __in6_dev_get(dev);
1512 if (!in6_dev)
1513 goto out;
1514
Cong Wangf564f452013-08-31 13:44:36 +08001515 iphdr = ipv6_hdr(skb);
1516 saddr = &iphdr->saddr;
1517 daddr = &iphdr->daddr;
1518
Cong Wangf564f452013-08-31 13:44:36 +08001519 msg = (struct nd_msg *)skb_transport_header(skb);
1520 if (msg->icmph.icmp6_code != 0 ||
1521 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1522 goto out;
1523
David Stevens4b29dba2014-03-24 10:39:58 -04001524 if (ipv6_addr_loopback(daddr) ||
1525 ipv6_addr_is_multicast(&msg->target))
1526 goto out;
1527
1528 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
Cong Wangf564f452013-08-31 13:44:36 +08001529
1530 if (n) {
1531 struct vxlan_fdb *f;
David Stevens4b29dba2014-03-24 10:39:58 -04001532 struct sk_buff *reply;
Cong Wangf564f452013-08-31 13:44:36 +08001533
1534 if (!(n->nud_state & NUD_CONNECTED)) {
1535 neigh_release(n);
1536 goto out;
1537 }
1538
1539 f = vxlan_find_mac(vxlan, n->ha);
1540 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1541 /* bridge-local neighbor */
1542 neigh_release(n);
1543 goto out;
1544 }
1545
David Stevens4b29dba2014-03-24 10:39:58 -04001546 reply = vxlan_na_create(skb, n,
1547 !!(f ? f->flags & NTF_ROUTER : 0));
1548
Cong Wangf564f452013-08-31 13:44:36 +08001549 neigh_release(n);
David Stevens4b29dba2014-03-24 10:39:58 -04001550
1551 if (reply == NULL)
1552 goto out;
1553
1554 if (netif_rx_ni(reply) == NET_RX_DROP)
1555 dev->stats.rx_dropped++;
1556
Cong Wangf564f452013-08-31 13:44:36 +08001557 } else if (vxlan->flags & VXLAN_F_L3MISS) {
David Stevens4b29dba2014-03-24 10:39:58 -04001558 union vxlan_addr ipa = {
1559 .sin6.sin6_addr = msg->target,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001560 .sin6.sin6_family = AF_INET6,
David Stevens4b29dba2014-03-24 10:39:58 -04001561 };
1562
Cong Wangf564f452013-08-31 13:44:36 +08001563 vxlan_ip_miss(dev, &ipa);
1564 }
1565
1566out:
1567 consume_skb(skb);
1568 return NETDEV_TX_OK;
1569}
1570#endif
1571
David Stevense4f67ad2012-11-20 02:50:14 +00001572static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1573{
1574 struct vxlan_dev *vxlan = netdev_priv(dev);
1575 struct neighbour *n;
David Stevense4f67ad2012-11-20 02:50:14 +00001576
1577 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1578 return false;
1579
1580 n = NULL;
1581 switch (ntohs(eth_hdr(skb)->h_proto)) {
1582 case ETH_P_IP:
Cong Wange15a00a2013-08-31 13:44:34 +08001583 {
1584 struct iphdr *pip;
1585
David Stevense4f67ad2012-11-20 02:50:14 +00001586 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1587 return false;
1588 pip = ip_hdr(skb);
1589 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
Cong Wange4c7ed42013-08-31 13:44:33 +08001590 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1591 union vxlan_addr ipa = {
1592 .sin.sin_addr.s_addr = pip->daddr,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001593 .sin.sin_family = AF_INET,
Cong Wange4c7ed42013-08-31 13:44:33 +08001594 };
1595
1596 vxlan_ip_miss(dev, &ipa);
1597 return false;
1598 }
1599
David Stevense4f67ad2012-11-20 02:50:14 +00001600 break;
Cong Wange15a00a2013-08-31 13:44:34 +08001601 }
1602#if IS_ENABLED(CONFIG_IPV6)
1603 case ETH_P_IPV6:
1604 {
1605 struct ipv6hdr *pip6;
1606
1607 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1608 return false;
1609 pip6 = ipv6_hdr(skb);
1610 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1611 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1612 union vxlan_addr ipa = {
1613 .sin6.sin6_addr = pip6->daddr,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001614 .sin6.sin6_family = AF_INET6,
Cong Wange15a00a2013-08-31 13:44:34 +08001615 };
1616
1617 vxlan_ip_miss(dev, &ipa);
1618 return false;
1619 }
1620
1621 break;
1622 }
1623#endif
David Stevense4f67ad2012-11-20 02:50:14 +00001624 default:
1625 return false;
1626 }
1627
1628 if (n) {
1629 bool diff;
1630
Joe Perches7367d0b2013-09-01 11:51:23 -07001631 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
David Stevense4f67ad2012-11-20 02:50:14 +00001632 if (diff) {
1633 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1634 dev->addr_len);
1635 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1636 }
1637 neigh_release(n);
1638 return diff;
Cong Wange4c7ed42013-08-31 13:44:33 +08001639 }
1640
David Stevense4f67ad2012-11-20 02:50:14 +00001641 return false;
1642}
1643
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001644static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
Thomas Graf35114942015-01-15 03:53:55 +01001645 struct vxlan_metadata *md)
1646{
1647 struct vxlanhdr_gbp *gbp;
1648
Thomas Grafdb79a622015-02-04 17:00:04 +01001649 if (!md->gbp)
1650 return;
1651
Thomas Graf35114942015-01-15 03:53:55 +01001652 gbp = (struct vxlanhdr_gbp *)vxh;
1653 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1654
1655 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1656 gbp->dont_learn = 1;
1657
1658 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1659 gbp->policy_applied = 1;
1660
1661 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1662}
1663
Cong Wange4c7ed42013-08-31 13:44:33 +08001664#if IS_ENABLED(CONFIG_IPV6)
David Miller79b16aa2015-04-05 22:19:09 -04001665static int vxlan6_xmit_skb(struct dst_entry *dst, struct sock *sk,
1666 struct sk_buff *skb,
Cong Wange4c7ed42013-08-31 13:44:33 +08001667 struct net_device *dev, struct in6_addr *saddr,
1668 struct in6_addr *daddr, __u8 prio, __u8 ttl,
Thomas Graf6b6948d2015-07-22 17:08:42 +02001669 __be16 src_port, __be16 dst_port, __be32 vni,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001670 struct vxlan_metadata *md, bool xnet, u32 vxflags)
Cong Wange4c7ed42013-08-31 13:44:33 +08001671{
Cong Wange4c7ed42013-08-31 13:44:33 +08001672 struct vxlanhdr *vxh;
Cong Wange4c7ed42013-08-31 13:44:33 +08001673 int min_headroom;
1674 int err;
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001675 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
Tom Herbertdfd86452015-01-12 17:00:38 -08001676 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1677 u16 hdrlen = sizeof(struct vxlanhdr);
Cong Wange4c7ed42013-08-31 13:44:33 +08001678
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001679 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
Tom Herbertdfd86452015-01-12 17:00:38 -08001680 skb->ip_summed == CHECKSUM_PARTIAL) {
1681 int csum_start = skb_checksum_start_offset(skb);
1682
1683 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1684 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1685 (skb->csum_offset == offsetof(struct udphdr, check) ||
1686 skb->csum_offset == offsetof(struct tcphdr, check))) {
1687 udp_sum = false;
1688 type |= SKB_GSO_TUNNEL_REMCSUM;
1689 }
1690 }
1691
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02001692 skb_scrub_packet(skb, xnet);
Nicolas Dichtel963a88b2013-09-02 15:34:57 +02001693
Cong Wange4c7ed42013-08-31 13:44:33 +08001694 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1695 + VXLAN_HLEN + sizeof(struct ipv6hdr)
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001696 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Cong Wange4c7ed42013-08-31 13:44:33 +08001697
1698 /* Need space for new headers (invalidates iph ptr) */
1699 err = skb_cow_head(skb, min_headroom);
Pravin B Shelar74f47272014-12-23 16:20:36 -08001700 if (unlikely(err)) {
1701 kfree_skb(skb);
1702 goto err;
1703 }
Cong Wange4c7ed42013-08-31 13:44:33 +08001704
Jiri Pirko59682502014-11-19 14:04:59 +01001705 skb = vlan_hwaccel_push_inside(skb);
Pravin B Shelar74f47272014-12-23 16:20:36 -08001706 if (WARN_ON(!skb)) {
1707 err = -ENOMEM;
1708 goto err;
1709 }
Cong Wange4c7ed42013-08-31 13:44:33 +08001710
Jesse Grossb736a622015-04-09 11:19:14 -07001711 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1712 if (IS_ERR(skb)) {
1713 err = -EINVAL;
1714 goto err;
1715 }
1716
Cong Wange4c7ed42013-08-31 13:44:33 +08001717 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
Tom Herbert3bf39472015-01-08 12:31:18 -08001718 vxh->vx_flags = htonl(VXLAN_HF_VNI);
Thomas Graf614732e2015-07-21 10:44:06 +02001719 vxh->vx_vni = vni;
Cong Wange4c7ed42013-08-31 13:44:33 +08001720
Tom Herbertdfd86452015-01-12 17:00:38 -08001721 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1722 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1723 VXLAN_RCO_SHIFT;
1724
1725 if (skb->csum_offset == offsetof(struct udphdr, check))
1726 data |= VXLAN_RCO_UDP;
1727
1728 vxh->vx_vni |= htonl(data);
1729 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1730
1731 if (!skb_is_gso(skb)) {
1732 skb->ip_summed = CHECKSUM_NONE;
1733 skb->encapsulation = 0;
1734 }
1735 }
1736
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001737 if (vxflags & VXLAN_F_GBP)
1738 vxlan_build_gbp_hdr(vxh, vxflags, md);
Thomas Graf35114942015-01-15 03:53:55 +01001739
Tom Herbert996c9fd2014-09-29 20:22:33 -07001740 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1741
David Miller79b16aa2015-04-05 22:19:09 -04001742 udp_tunnel6_xmit_skb(dst, sk, skb, dev, saddr, daddr, prio,
Tom Herbertd998f8e2015-01-20 11:23:04 -08001743 ttl, src_port, dst_port,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001744 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
Cong Wange4c7ed42013-08-31 13:44:33 +08001745 return 0;
Pravin B Shelar74f47272014-12-23 16:20:36 -08001746err:
1747 dst_release(dst);
1748 return err;
Cong Wange4c7ed42013-08-31 13:44:33 +08001749}
1750#endif
1751
Thomas Graf614732e2015-07-21 10:44:06 +02001752static int vxlan_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
1753 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
Thomas Graf6b6948d2015-07-22 17:08:42 +02001754 __be16 src_port, __be16 dst_port, __be32 vni,
Thomas Graf614732e2015-07-21 10:44:06 +02001755 struct vxlan_metadata *md, bool xnet, u32 vxflags)
Pravin B Shelar49560532013-08-19 11:23:17 -07001756{
1757 struct vxlanhdr *vxh;
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001758 int min_headroom;
Pravin B Shelar49560532013-08-19 11:23:17 -07001759 int err;
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001760 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
Tom Herbertdfd86452015-01-12 17:00:38 -08001761 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1762 u16 hdrlen = sizeof(struct vxlanhdr);
Pravin B Shelar49560532013-08-19 11:23:17 -07001763
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001764 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
Tom Herbertdfd86452015-01-12 17:00:38 -08001765 skb->ip_summed == CHECKSUM_PARTIAL) {
1766 int csum_start = skb_checksum_start_offset(skb);
1767
1768 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1769 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1770 (skb->csum_offset == offsetof(struct udphdr, check) ||
1771 skb->csum_offset == offsetof(struct tcphdr, check))) {
1772 udp_sum = false;
1773 type |= SKB_GSO_TUNNEL_REMCSUM;
1774 }
1775 }
1776
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001777 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07001778 + VXLAN_HLEN + sizeof(struct iphdr)
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001779 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001780
1781 /* Need space for new headers (invalidates iph ptr) */
1782 err = skb_cow_head(skb, min_headroom);
Pravin B Shelar74f47272014-12-23 16:20:36 -08001783 if (unlikely(err)) {
1784 kfree_skb(skb);
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001785 return err;
Pravin B Shelar74f47272014-12-23 16:20:36 -08001786 }
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001787
Jiri Pirko59682502014-11-19 14:04:59 +01001788 skb = vlan_hwaccel_push_inside(skb);
1789 if (WARN_ON(!skb))
1790 return -ENOMEM;
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07001791
Jesse Grossb736a622015-04-09 11:19:14 -07001792 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1793 if (IS_ERR(skb))
1794 return PTR_ERR(skb);
1795
Pravin B Shelar49560532013-08-19 11:23:17 -07001796 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
Tom Herbert3bf39472015-01-08 12:31:18 -08001797 vxh->vx_flags = htonl(VXLAN_HF_VNI);
Thomas Graf614732e2015-07-21 10:44:06 +02001798 vxh->vx_vni = vni;
Pravin B Shelar49560532013-08-19 11:23:17 -07001799
Tom Herbertdfd86452015-01-12 17:00:38 -08001800 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1801 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1802 VXLAN_RCO_SHIFT;
1803
1804 if (skb->csum_offset == offsetof(struct udphdr, check))
1805 data |= VXLAN_RCO_UDP;
1806
1807 vxh->vx_vni |= htonl(data);
1808 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1809
1810 if (!skb_is_gso(skb)) {
1811 skb->ip_summed = CHECKSUM_NONE;
1812 skb->encapsulation = 0;
1813 }
1814 }
1815
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001816 if (vxflags & VXLAN_F_GBP)
1817 vxlan_build_gbp_hdr(vxh, vxflags, md);
Thomas Graf35114942015-01-15 03:53:55 +01001818
Tom Herbert996c9fd2014-09-29 20:22:33 -07001819 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1820
David Miller79b16aa2015-04-05 22:19:09 -04001821 return udp_tunnel_xmit_skb(rt, sk, skb, src, dst, tos,
Tom Herbertd998f8e2015-01-20 11:23:04 -08001822 ttl, df, src_port, dst_port, xnet,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001823 !(vxflags & VXLAN_F_UDP_CSUM));
Pravin B Shelar49560532013-08-19 11:23:17 -07001824}
Pravin B Shelar49560532013-08-19 11:23:17 -07001825
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001826/* Bypass encapsulation if the destination is local */
1827static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1828 struct vxlan_dev *dst_vxlan)
1829{
Li RongQing8f849852014-01-04 13:57:59 +08001830 struct pcpu_sw_netstats *tx_stats, *rx_stats;
Cong Wange4c7ed42013-08-31 13:44:33 +08001831 union vxlan_addr loopback;
1832 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
Li RongQingce6502a2014-10-16 08:49:41 +08001833 struct net_device *dev = skb->dev;
1834 int len = skb->len;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001835
Li RongQing8f849852014-01-04 13:57:59 +08001836 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1837 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001838 skb->pkt_type = PACKET_HOST;
1839 skb->encapsulation = 0;
1840 skb->dev = dst_vxlan->dev;
1841 __skb_pull(skb, skb_network_offset(skb));
1842
Cong Wange4c7ed42013-08-31 13:44:33 +08001843 if (remote_ip->sa.sa_family == AF_INET) {
1844 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1845 loopback.sa.sa_family = AF_INET;
1846#if IS_ENABLED(CONFIG_IPV6)
1847 } else {
1848 loopback.sin6.sin6_addr = in6addr_loopback;
1849 loopback.sa.sa_family = AF_INET6;
1850#endif
1851 }
1852
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001853 if (dst_vxlan->flags & VXLAN_F_LEARN)
Cong Wange4c7ed42013-08-31 13:44:33 +08001854 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001855
1856 u64_stats_update_begin(&tx_stats->syncp);
1857 tx_stats->tx_packets++;
Li RongQingce6502a2014-10-16 08:49:41 +08001858 tx_stats->tx_bytes += len;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001859 u64_stats_update_end(&tx_stats->syncp);
1860
1861 if (netif_rx(skb) == NET_RX_SUCCESS) {
1862 u64_stats_update_begin(&rx_stats->syncp);
1863 rx_stats->rx_packets++;
Li RongQingce6502a2014-10-16 08:49:41 +08001864 rx_stats->rx_bytes += len;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001865 u64_stats_update_end(&rx_stats->syncp);
1866 } else {
Li RongQingce6502a2014-10-16 08:49:41 +08001867 dev->stats.rx_dropped++;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001868 }
1869}
1870
Stephen Hemminger4ad16932013-06-17 14:16:11 -07001871static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1872 struct vxlan_rdst *rdst, bool did_rsc)
stephen hemmingerd3428942012-10-01 12:32:35 +00001873{
Thomas Graf3093fbe2015-07-21 10:44:00 +02001874 struct ip_tunnel_info *info;
stephen hemmingerd3428942012-10-01 12:32:35 +00001875 struct vxlan_dev *vxlan = netdev_priv(dev);
David Miller79b16aa2015-04-05 22:19:09 -04001876 struct sock *sk = vxlan->vn_sock->sock->sk;
Jiri Benca725e512015-08-20 13:56:30 +02001877 unsigned short family = vxlan_get_sk_family(vxlan->vn_sock);
Cong Wange4c7ed42013-08-31 13:44:33 +08001878 struct rtable *rt = NULL;
stephen hemmingerd3428942012-10-01 12:32:35 +00001879 const struct iphdr *old_iph;
stephen hemmingerd3428942012-10-01 12:32:35 +00001880 struct flowi4 fl4;
Cong Wange4c7ed42013-08-31 13:44:33 +08001881 union vxlan_addr *dst;
Thomas Grafee122c72015-07-21 10:43:58 +02001882 union vxlan_addr remote_ip;
1883 struct vxlan_metadata _md;
1884 struct vxlan_metadata *md = &_md;
Cong Wange4c7ed42013-08-31 13:44:33 +08001885 __be16 src_port = 0, dst_port;
Stephen Hemminger234f5b72013-06-17 14:16:41 -07001886 u32 vni;
stephen hemmingerd3428942012-10-01 12:32:35 +00001887 __be16 df = 0;
1888 __u8 tos, ttl;
Pravin B Shelar0e6fbc52013-06-17 17:49:56 -07001889 int err;
Thomas Grafee122c72015-07-21 10:43:58 +02001890 u32 flags = vxlan->flags;
stephen hemmingerd3428942012-10-01 12:32:35 +00001891
Jiri Benc61adedf2015-08-20 13:56:25 +02001892 info = skb_tunnel_info(skb);
Thomas Graf3093fbe2015-07-21 10:44:00 +02001893
Thomas Grafee122c72015-07-21 10:43:58 +02001894 if (rdst) {
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001895 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
Thomas Grafee122c72015-07-21 10:43:58 +02001896 vni = rdst->remote_vni;
1897 dst = &rdst->remote_ip;
1898 } else {
1899 if (!info) {
1900 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1901 dev->name);
1902 goto drop;
1903 }
Jiri Benc7f9562a2015-08-28 20:48:20 +02001904 if (family != ip_tunnel_info_af(info))
1905 goto drop;
Thomas Grafee122c72015-07-21 10:43:58 +02001906
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001907 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
Thomas Grafee122c72015-07-21 10:43:58 +02001908 vni = be64_to_cpu(info->key.tun_id);
Jiri Benca725e512015-08-20 13:56:30 +02001909 remote_ip.sa.sa_family = family;
1910 if (family == AF_INET)
1911 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1912 else
1913 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
Thomas Grafee122c72015-07-21 10:43:58 +02001914 dst = &remote_ip;
1915 }
David Stevense4f67ad2012-11-20 02:50:14 +00001916
Cong Wange4c7ed42013-08-31 13:44:33 +08001917 if (vxlan_addr_any(dst)) {
David Stevense4f67ad2012-11-20 02:50:14 +00001918 if (did_rsc) {
David Stevense4f67ad2012-11-20 02:50:14 +00001919 /* short-circuited back to local bridge */
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001920 vxlan_encap_bypass(skb, vxlan, vxlan);
Stephen Hemminger4ad16932013-06-17 14:16:11 -07001921 return;
David Stevense4f67ad2012-11-20 02:50:14 +00001922 }
stephen hemmingeref59feb2012-10-09 20:35:46 +00001923 goto drop;
David Stevense4f67ad2012-11-20 02:50:14 +00001924 }
stephen hemmingeref59feb2012-10-09 20:35:46 +00001925
stephen hemmingerd3428942012-10-01 12:32:35 +00001926 old_iph = ip_hdr(skb);
1927
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001928 ttl = vxlan->cfg.ttl;
Cong Wange4c7ed42013-08-31 13:44:33 +08001929 if (!ttl && vxlan_addr_multicast(dst))
stephen hemmingerd3428942012-10-01 12:32:35 +00001930 ttl = 1;
1931
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001932 tos = vxlan->cfg.tos;
stephen hemmingerd3428942012-10-01 12:32:35 +00001933 if (tos == 1)
Pravin B Shelar206aaaf2013-03-25 14:49:53 +00001934 tos = ip_tunnel_get_dsfield(old_iph, skb);
stephen hemmingerd3428942012-10-01 12:32:35 +00001935
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001936 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1937 vxlan->cfg.port_max, true);
stephen hemmingerd3428942012-10-01 12:32:35 +00001938
Jiri Benca725e512015-08-20 13:56:30 +02001939 if (info) {
1940 if (info->key.tun_flags & TUNNEL_CSUM)
1941 flags |= VXLAN_F_UDP_CSUM;
1942 else
1943 flags &= ~VXLAN_F_UDP_CSUM;
1944
1945 ttl = info->key.ttl;
1946 tos = info->key.tos;
1947
1948 if (info->options_len)
Pravin B Shelar4c222792015-08-30 18:09:38 -07001949 md = ip_tunnel_info_opts(info);
Jiri Benca725e512015-08-20 13:56:30 +02001950 } else {
1951 md->gbp = skb->mark;
1952 }
1953
Cong Wange4c7ed42013-08-31 13:44:33 +08001954 if (dst->sa.sa_family == AF_INET) {
Jiri Benca725e512015-08-20 13:56:30 +02001955 if (info && (info->key.tun_flags & TUNNEL_DONT_FRAGMENT))
1956 df = htons(IP_DF);
Thomas Grafee122c72015-07-21 10:43:58 +02001957
Cong Wange4c7ed42013-08-31 13:44:33 +08001958 memset(&fl4, 0, sizeof(fl4));
Thomas Grafee122c72015-07-21 10:43:58 +02001959 fl4.flowi4_oif = rdst ? rdst->remote_ifindex : 0;
Cong Wange4c7ed42013-08-31 13:44:33 +08001960 fl4.flowi4_tos = RT_TOS(tos);
Thomas Graf239fb792015-05-05 15:09:21 +02001961 fl4.flowi4_mark = skb->mark;
1962 fl4.flowi4_proto = IPPROTO_UDP;
Cong Wange4c7ed42013-08-31 13:44:33 +08001963 fl4.daddr = dst->sin.sin_addr.s_addr;
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001964 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
stephen hemmingerca78f182012-10-09 20:35:48 +00001965
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02001966 rt = ip_route_output_key(vxlan->net, &fl4);
Cong Wange4c7ed42013-08-31 13:44:33 +08001967 if (IS_ERR(rt)) {
1968 netdev_dbg(dev, "no route to %pI4\n",
1969 &dst->sin.sin_addr.s_addr);
1970 dev->stats.tx_carrier_errors++;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001971 goto tx_error;
Cong Wange4c7ed42013-08-31 13:44:33 +08001972 }
1973
1974 if (rt->dst.dev == dev) {
1975 netdev_dbg(dev, "circular route to %pI4\n",
1976 &dst->sin.sin_addr.s_addr);
1977 dev->stats.collisions++;
Fan Dufffc15a2013-12-09 10:33:53 +08001978 goto rt_tx_error;
Cong Wange4c7ed42013-08-31 13:44:33 +08001979 }
1980
1981 /* Bypass encapsulation if the destination is local */
1982 if (rt->rt_flags & RTCF_LOCAL &&
1983 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
1984 struct vxlan_dev *dst_vxlan;
1985
1986 ip_rt_put(rt);
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -02001987 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
Thomas Grafac5132d2015-01-15 03:53:56 +01001988 dst->sa.sa_family, dst_port,
1989 vxlan->flags);
Cong Wange4c7ed42013-08-31 13:44:33 +08001990 if (!dst_vxlan)
1991 goto tx_error;
1992 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
1993 return;
1994 }
1995
1996 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
1997 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
David Miller79b16aa2015-04-05 22:19:09 -04001998 err = vxlan_xmit_skb(rt, sk, skb, fl4.saddr,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001999 dst->sin.sin_addr.s_addr, tos, ttl, df,
Thomas Graf614732e2015-07-21 10:44:06 +02002000 src_port, dst_port, htonl(vni << 8), md,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002001 !net_eq(vxlan->net, dev_net(vxlan->dev)),
Thomas Grafee122c72015-07-21 10:43:58 +02002002 flags);
Pravin B Shelar74f47272014-12-23 16:20:36 -08002003 if (err < 0) {
2004 /* skb is already freed. */
2005 skb = NULL;
Cong Wange4c7ed42013-08-31 13:44:33 +08002006 goto rt_tx_error;
Pravin B Shelar74f47272014-12-23 16:20:36 -08002007 }
2008
Cong Wange4c7ed42013-08-31 13:44:33 +08002009 iptunnel_xmit_stats(err, &dev->stats, dev->tstats);
2010#if IS_ENABLED(CONFIG_IPV6)
2011 } else {
Cong Wange4c7ed42013-08-31 13:44:33 +08002012 struct dst_entry *ndst;
2013 struct flowi6 fl6;
Jiri Benc6f264e42015-08-20 13:56:29 +02002014 u32 rt6i_flags;
Cong Wange4c7ed42013-08-31 13:44:33 +08002015
2016 memset(&fl6, 0, sizeof(fl6));
Thomas Grafee122c72015-07-21 10:43:58 +02002017 fl6.flowi6_oif = rdst ? rdst->remote_ifindex : 0;
Cong Wange4c7ed42013-08-31 13:44:33 +08002018 fl6.daddr = dst->sin6.sin6_addr;
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002019 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
Thomas Graf239fb792015-05-05 15:09:21 +02002020 fl6.flowi6_mark = skb->mark;
Cong Wang8c1bb792013-09-02 10:06:51 +08002021 fl6.flowi6_proto = IPPROTO_UDP;
Cong Wange4c7ed42013-08-31 13:44:33 +08002022
Roopa Prabhu343d60a2015-07-30 13:34:53 -07002023 if (ipv6_stub->ipv6_dst_lookup(vxlan->net, sk, &ndst, &fl6)) {
Cong Wange4c7ed42013-08-31 13:44:33 +08002024 netdev_dbg(dev, "no route to %pI6\n",
2025 &dst->sin6.sin6_addr);
2026 dev->stats.tx_carrier_errors++;
2027 goto tx_error;
2028 }
2029
2030 if (ndst->dev == dev) {
2031 netdev_dbg(dev, "circular route to %pI6\n",
2032 &dst->sin6.sin6_addr);
2033 dst_release(ndst);
2034 dev->stats.collisions++;
2035 goto tx_error;
2036 }
2037
2038 /* Bypass encapsulation if the destination is local */
Jiri Benc6f264e42015-08-20 13:56:29 +02002039 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2040 if (rt6i_flags & RTF_LOCAL &&
2041 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
Cong Wange4c7ed42013-08-31 13:44:33 +08002042 struct vxlan_dev *dst_vxlan;
2043
2044 dst_release(ndst);
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -02002045 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
Thomas Grafac5132d2015-01-15 03:53:56 +01002046 dst->sa.sa_family, dst_port,
2047 vxlan->flags);
Cong Wange4c7ed42013-08-31 13:44:33 +08002048 if (!dst_vxlan)
2049 goto tx_error;
2050 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2051 return;
2052 }
2053
2054 ttl = ttl ? : ip6_dst_hoplimit(ndst);
David Miller79b16aa2015-04-05 22:19:09 -04002055 err = vxlan6_xmit_skb(ndst, sk, skb, dev, &fl6.saddr, &fl6.daddr,
Thomas Graf614732e2015-07-21 10:44:06 +02002056 0, ttl, src_port, dst_port, htonl(vni << 8), md,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002057 !net_eq(vxlan->net, dev_net(vxlan->dev)),
Jiri Benca725e512015-08-20 13:56:30 +02002058 flags);
Cong Wange4c7ed42013-08-31 13:44:33 +08002059#endif
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00002060 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002061
Stephen Hemminger4ad16932013-06-17 14:16:11 -07002062 return;
stephen hemmingerd3428942012-10-01 12:32:35 +00002063
2064drop:
2065 dev->stats.tx_dropped++;
2066 goto tx_free;
2067
Pravin B Shelar49560532013-08-19 11:23:17 -07002068rt_tx_error:
2069 ip_rt_put(rt);
stephen hemmingerd3428942012-10-01 12:32:35 +00002070tx_error:
2071 dev->stats.tx_errors++;
2072tx_free:
2073 dev_kfree_skb(skb);
stephen hemmingerd3428942012-10-01 12:32:35 +00002074}
2075
David Stevens66817122013-03-15 04:35:51 +00002076/* Transmit local packets over Vxlan
2077 *
2078 * Outer IP header inherits ECN and DF from inner header.
2079 * Outer UDP destination is the VXLAN assigned port.
2080 * source port is based on hash of flow
2081 */
2082static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2083{
2084 struct vxlan_dev *vxlan = netdev_priv(dev);
Thomas Graf3093fbe2015-07-21 10:44:00 +02002085 const struct ip_tunnel_info *info;
David Stevens66817122013-03-15 04:35:51 +00002086 struct ethhdr *eth;
2087 bool did_rsc = false;
Eric Dumazet8f646c92014-01-06 09:54:31 -08002088 struct vxlan_rdst *rdst, *fdst = NULL;
David Stevens66817122013-03-15 04:35:51 +00002089 struct vxlan_fdb *f;
David Stevens66817122013-03-15 04:35:51 +00002090
Jiri Benc61adedf2015-08-20 13:56:25 +02002091 info = skb_tunnel_info(skb);
Thomas Graf3093fbe2015-07-21 10:44:00 +02002092
David Stevens66817122013-03-15 04:35:51 +00002093 skb_reset_mac_header(skb);
2094 eth = eth_hdr(skb);
2095
Cong Wangf564f452013-08-31 13:44:36 +08002096 if ((vxlan->flags & VXLAN_F_PROXY)) {
2097 if (ntohs(eth->h_proto) == ETH_P_ARP)
2098 return arp_reduce(dev, skb);
2099#if IS_ENABLED(CONFIG_IPV6)
2100 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
Li RongQing91269e32014-10-16 09:17:18 +08002101 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2102 + sizeof(struct nd_msg)) &&
Cong Wangf564f452013-08-31 13:44:36 +08002103 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2104 struct nd_msg *msg;
2105
2106 msg = (struct nd_msg *)skb_transport_header(skb);
2107 if (msg->icmph.icmp6_code == 0 &&
2108 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2109 return neigh_reduce(dev, skb);
2110 }
Li RongQing7a9f5262014-10-17 14:06:16 +08002111 eth = eth_hdr(skb);
Cong Wangf564f452013-08-31 13:44:36 +08002112#endif
2113 }
David Stevens66817122013-03-15 04:35:51 +00002114
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07002115 if (vxlan->flags & VXLAN_F_COLLECT_METADATA &&
Jiri Benc46fa0622015-08-28 20:48:19 +02002116 info && info->mode & IP_TUNNEL_INFO_TX) {
Thomas Grafee122c72015-07-21 10:43:58 +02002117 vxlan_xmit_one(skb, dev, NULL, false);
2118 return NETDEV_TX_OK;
2119 }
2120
David Stevens66817122013-03-15 04:35:51 +00002121 f = vxlan_find_mac(vxlan, eth->h_dest);
David Stevensae884082013-04-19 00:36:26 +00002122 did_rsc = false;
2123
2124 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
Cong Wange15a00a2013-08-31 13:44:34 +08002125 (ntohs(eth->h_proto) == ETH_P_IP ||
2126 ntohs(eth->h_proto) == ETH_P_IPV6)) {
David Stevensae884082013-04-19 00:36:26 +00002127 did_rsc = route_shortcircuit(dev, skb);
2128 if (did_rsc)
2129 f = vxlan_find_mac(vxlan, eth->h_dest);
2130 }
2131
David Stevens66817122013-03-15 04:35:51 +00002132 if (f == NULL) {
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002133 f = vxlan_find_mac(vxlan, all_zeros_mac);
2134 if (f == NULL) {
2135 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2136 !is_multicast_ether_addr(eth->h_dest))
2137 vxlan_fdb_miss(vxlan, eth->h_dest);
David Stevens66817122013-03-15 04:35:51 +00002138
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002139 dev->stats.tx_dropped++;
Eric Dumazet8f646c92014-01-06 09:54:31 -08002140 kfree_skb(skb);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002141 return NETDEV_TX_OK;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -07002142 }
David Stevens66817122013-03-15 04:35:51 +00002143 }
2144
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002145 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2146 struct sk_buff *skb1;
2147
Eric Dumazet8f646c92014-01-06 09:54:31 -08002148 if (!fdst) {
2149 fdst = rdst;
2150 continue;
2151 }
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002152 skb1 = skb_clone(skb, GFP_ATOMIC);
2153 if (skb1)
2154 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
2155 }
2156
Eric Dumazet8f646c92014-01-06 09:54:31 -08002157 if (fdst)
2158 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2159 else
2160 kfree_skb(skb);
Stephen Hemminger4ad16932013-06-17 14:16:11 -07002161 return NETDEV_TX_OK;
David Stevens66817122013-03-15 04:35:51 +00002162}
2163
stephen hemmingerd3428942012-10-01 12:32:35 +00002164/* Walk the forwarding table and purge stale entries */
2165static void vxlan_cleanup(unsigned long arg)
2166{
2167 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2168 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2169 unsigned int h;
2170
2171 if (!netif_running(vxlan->dev))
2172 return;
2173
stephen hemmingerd3428942012-10-01 12:32:35 +00002174 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2175 struct hlist_node *p, *n;
Sorin Dumitru14e1d0f2015-05-26 10:42:04 +03002176
2177 spin_lock_bh(&vxlan->hash_lock);
stephen hemmingerd3428942012-10-01 12:32:35 +00002178 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2179 struct vxlan_fdb *f
2180 = container_of(p, struct vxlan_fdb, hlist);
2181 unsigned long timeout;
2182
stephen hemminger3c172862012-10-26 06:24:34 +00002183 if (f->state & NUD_PERMANENT)
stephen hemmingerd3428942012-10-01 12:32:35 +00002184 continue;
2185
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002186 timeout = f->used + vxlan->cfg.age_interval * HZ;
stephen hemmingerd3428942012-10-01 12:32:35 +00002187 if (time_before_eq(timeout, jiffies)) {
2188 netdev_dbg(vxlan->dev,
2189 "garbage collect %pM\n",
2190 f->eth_addr);
2191 f->state = NUD_STALE;
2192 vxlan_fdb_destroy(vxlan, f);
2193 } else if (time_before(timeout, next_timer))
2194 next_timer = timeout;
2195 }
Sorin Dumitru14e1d0f2015-05-26 10:42:04 +03002196 spin_unlock_bh(&vxlan->hash_lock);
stephen hemmingerd3428942012-10-01 12:32:35 +00002197 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002198
2199 mod_timer(&vxlan->age_timer, next_timer);
2200}
2201
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002202static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2203{
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002204 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002205 __u32 vni = vxlan->default_dst.remote_vni;
2206
2207 vxlan->vn_sock = vs;
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002208 spin_lock(&vn->sock_lock);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002209 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002210 spin_unlock(&vn->sock_lock);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002211}
2212
stephen hemmingerd3428942012-10-01 12:32:35 +00002213/* Setup stats when device is created */
2214static int vxlan_init(struct net_device *dev)
2215{
WANG Cong1c213bd2014-02-13 11:46:28 -08002216 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
Pravin B Shelare8171042013-03-25 14:49:46 +00002217 if (!dev->tstats)
stephen hemmingerd3428942012-10-01 12:32:35 +00002218 return -ENOMEM;
2219
2220 return 0;
2221}
2222
Stephen Hemmingerba609e92013-06-25 17:06:01 -07002223static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002224{
2225 struct vxlan_fdb *f;
2226
2227 spin_lock_bh(&vxlan->hash_lock);
2228 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2229 if (f)
2230 vxlan_fdb_destroy(vxlan, f);
2231 spin_unlock_bh(&vxlan->hash_lock);
2232}
2233
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002234static void vxlan_uninit(struct net_device *dev)
2235{
2236 struct vxlan_dev *vxlan = netdev_priv(dev);
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002237
Stephen Hemmingerba609e92013-06-25 17:06:01 -07002238 vxlan_fdb_delete_default(vxlan);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002239
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002240 free_percpu(dev->tstats);
2241}
2242
stephen hemmingerd3428942012-10-01 12:32:35 +00002243/* Start ageing timer and join group when device is brought up */
2244static int vxlan_open(struct net_device *dev)
2245{
2246 struct vxlan_dev *vxlan = netdev_priv(dev);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002247 struct vxlan_sock *vs;
2248 int ret = 0;
Stephen Hemminger1c51a912013-06-17 14:16:11 -07002249
Thomas Graf614732e2015-07-21 10:44:06 +02002250 vs = vxlan_sock_add(vxlan->net, vxlan->cfg.dst_port,
2251 vxlan->cfg.no_share, vxlan->flags);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002252 if (IS_ERR(vs))
2253 return PTR_ERR(vs);
2254
2255 vxlan_vs_add_dev(vs, vxlan);
stephen hemmingerd3428942012-10-01 12:32:35 +00002256
Gao feng79d4a942013-12-10 16:37:32 +08002257 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002258 ret = vxlan_igmp_join(vxlan);
Marcelo Ricardo Leitnerbef00572015-08-25 20:22:35 -03002259 if (ret == -EADDRINUSE)
2260 ret = 0;
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002261 if (ret) {
2262 vxlan_sock_release(vs);
2263 return ret;
2264 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002265 }
2266
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002267 if (vxlan->cfg.age_interval)
stephen hemmingerd3428942012-10-01 12:32:35 +00002268 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2269
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002270 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00002271}
2272
2273/* Purge the forwarding table */
2274static void vxlan_flush(struct vxlan_dev *vxlan)
2275{
Cong Wang31fec5a2013-05-27 22:35:52 +00002276 unsigned int h;
stephen hemmingerd3428942012-10-01 12:32:35 +00002277
2278 spin_lock_bh(&vxlan->hash_lock);
2279 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2280 struct hlist_node *p, *n;
2281 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2282 struct vxlan_fdb *f
2283 = container_of(p, struct vxlan_fdb, hlist);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002284 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2285 if (!is_zero_ether_addr(f->eth_addr))
2286 vxlan_fdb_destroy(vxlan, f);
stephen hemmingerd3428942012-10-01 12:32:35 +00002287 }
2288 }
2289 spin_unlock_bh(&vxlan->hash_lock);
2290}
2291
2292/* Cleanup timer and forwarding table on shutdown */
2293static int vxlan_stop(struct net_device *dev)
2294{
2295 struct vxlan_dev *vxlan = netdev_priv(dev);
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002296 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
Stephen Hemminger1c51a912013-06-17 14:16:11 -07002297 struct vxlan_sock *vs = vxlan->vn_sock;
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002298 int ret = 0;
stephen hemmingerd3428942012-10-01 12:32:35 +00002299
Marcelo Ricardo Leitner24c0e682015-03-23 16:23:12 -03002300 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
WANG Congf13b1682015-04-08 14:48:30 -07002301 !vxlan_group_used(vn, vxlan))
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002302 ret = vxlan_igmp_leave(vxlan);
stephen hemmingerd3428942012-10-01 12:32:35 +00002303
2304 del_timer_sync(&vxlan->age_timer);
2305
2306 vxlan_flush(vxlan);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002307 vxlan_sock_release(vs);
stephen hemmingerd3428942012-10-01 12:32:35 +00002308
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002309 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00002310}
2311
stephen hemmingerd3428942012-10-01 12:32:35 +00002312/* Stub, nothing needs to be done. */
2313static void vxlan_set_multicast_list(struct net_device *dev)
2314{
2315}
2316
Daniel Borkmann345010b2013-12-18 00:21:08 +01002317static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2318{
2319 struct vxlan_dev *vxlan = netdev_priv(dev);
2320 struct vxlan_rdst *dst = &vxlan->default_dst;
2321 struct net_device *lowerdev;
2322 int max_mtu;
2323
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002324 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
Daniel Borkmann345010b2013-12-18 00:21:08 +01002325 if (lowerdev == NULL)
2326 return eth_change_mtu(dev, new_mtu);
2327
2328 if (dst->remote_ip.sa.sa_family == AF_INET6)
2329 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2330 else
2331 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2332
2333 if (new_mtu < 68 || new_mtu > max_mtu)
2334 return -EINVAL;
2335
2336 dev->mtu = new_mtu;
2337 return 0;
2338}
2339
stephen hemmingerd3428942012-10-01 12:32:35 +00002340static const struct net_device_ops vxlan_netdev_ops = {
2341 .ndo_init = vxlan_init,
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002342 .ndo_uninit = vxlan_uninit,
stephen hemmingerd3428942012-10-01 12:32:35 +00002343 .ndo_open = vxlan_open,
2344 .ndo_stop = vxlan_stop,
2345 .ndo_start_xmit = vxlan_xmit,
Pravin B Shelare8171042013-03-25 14:49:46 +00002346 .ndo_get_stats64 = ip_tunnel_get_stats64,
stephen hemmingerd3428942012-10-01 12:32:35 +00002347 .ndo_set_rx_mode = vxlan_set_multicast_list,
Daniel Borkmann345010b2013-12-18 00:21:08 +01002348 .ndo_change_mtu = vxlan_change_mtu,
stephen hemmingerd3428942012-10-01 12:32:35 +00002349 .ndo_validate_addr = eth_validate_addr,
2350 .ndo_set_mac_address = eth_mac_addr,
2351 .ndo_fdb_add = vxlan_fdb_add,
2352 .ndo_fdb_del = vxlan_fdb_delete,
2353 .ndo_fdb_dump = vxlan_fdb_dump,
2354};
2355
2356/* Info for udev, that this is a virtual tunnel endpoint */
2357static struct device_type vxlan_type = {
2358 .name = "vxlan",
2359};
2360
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002361/* Calls the ndo_add_vxlan_port of the caller in order to
Joseph Gasparakis35e42372013-09-13 07:34:13 -07002362 * supply the listening VXLAN udp ports. Callers are expected
2363 * to implement the ndo_add_vxlan_port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002364 */
2365void vxlan_get_rx_port(struct net_device *dev)
2366{
2367 struct vxlan_sock *vs;
2368 struct net *net = dev_net(dev);
2369 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2370 sa_family_t sa_family;
Joseph Gasparakis35e42372013-09-13 07:34:13 -07002371 __be16 port;
2372 unsigned int i;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002373
2374 spin_lock(&vn->sock_lock);
2375 for (i = 0; i < PORT_HASH_SIZE; ++i) {
Joseph Gasparakis35e42372013-09-13 07:34:13 -07002376 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2377 port = inet_sk(vs->sock->sk)->inet_sport;
Jiri Benc705cc622015-08-20 13:56:28 +02002378 sa_family = vxlan_get_sk_family(vs);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002379 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2380 port);
2381 }
2382 }
2383 spin_unlock(&vn->sock_lock);
2384}
2385EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2386
stephen hemmingerd3428942012-10-01 12:32:35 +00002387/* Initialize the device structure. */
2388static void vxlan_setup(struct net_device *dev)
2389{
2390 struct vxlan_dev *vxlan = netdev_priv(dev);
Cong Wang31fec5a2013-05-27 22:35:52 +00002391 unsigned int h;
stephen hemmingerd3428942012-10-01 12:32:35 +00002392
2393 eth_hw_addr_random(dev);
2394 ether_setup(dev);
2395
2396 dev->netdev_ops = &vxlan_netdev_ops;
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002397 dev->destructor = free_netdev;
stephen hemmingerd3428942012-10-01 12:32:35 +00002398 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2399
stephen hemmingerd3428942012-10-01 12:32:35 +00002400 dev->features |= NETIF_F_LLTX;
Joseph Gasparakisd6727fe2012-12-07 14:14:16 +00002401 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
Joseph Gasparakis0afb1662012-12-07 14:14:18 +00002402 dev->features |= NETIF_F_RXCSUM;
Pravin B Shelar05c0db02013-03-07 13:22:36 +00002403 dev->features |= NETIF_F_GSO_SOFTWARE;
Joseph Gasparakis0afb1662012-12-07 14:14:18 +00002404
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07002405 dev->vlan_features = dev->features;
2406 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
Joseph Gasparakis0afb1662012-12-07 14:14:18 +00002407 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
Pravin B Shelar05c0db02013-03-07 13:22:36 +00002408 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07002409 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
Eric Dumazet02875872014-10-05 18:38:35 -07002410 netif_keep_dst(dev);
Phil Sutter22dba392015-08-18 10:30:35 +02002411 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
stephen hemmingerd3428942012-10-01 12:32:35 +00002412
stephen hemminger553675f2013-05-16 11:35:20 +00002413 INIT_LIST_HEAD(&vxlan->next);
stephen hemmingerd3428942012-10-01 12:32:35 +00002414 spin_lock_init(&vxlan->hash_lock);
2415
2416 init_timer_deferrable(&vxlan->age_timer);
2417 vxlan->age_timer.function = vxlan_cleanup;
2418 vxlan->age_timer.data = (unsigned long) vxlan;
2419
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002420 vxlan->cfg.dst_port = htons(vxlan_port);
stephen hemminger05f47d62012-10-09 20:35:50 +00002421
stephen hemmingerd3428942012-10-01 12:32:35 +00002422 vxlan->dev = dev;
2423
Tom Herbert58ce31c2015-08-19 17:07:33 -07002424 gro_cells_init(&vxlan->gro_cells, dev);
2425
stephen hemmingerd3428942012-10-01 12:32:35 +00002426 for (h = 0; h < FDB_HASH_SIZE; ++h)
2427 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2428}
2429
2430static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2431 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
stephen hemminger5d174dd2013-04-27 11:31:55 +00002432 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
Cong Wange4c7ed42013-08-31 13:44:33 +08002433 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
stephen hemmingerd3428942012-10-01 12:32:35 +00002434 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2435 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
Cong Wange4c7ed42013-08-31 13:44:33 +08002436 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
stephen hemmingerd3428942012-10-01 12:32:35 +00002437 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2438 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2439 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2440 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2441 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
stephen hemminger05f47d62012-10-09 20:35:50 +00002442 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
David Stevense4f67ad2012-11-20 02:50:14 +00002443 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2444 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2445 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2446 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
Alexei Starovoitovf8a9b1b2015-07-30 20:10:22 -07002447 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
stephen hemminger823aa872013-04-27 11:31:57 +00002448 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
Tom Herbert5c91ae02014-11-06 18:06:01 -08002449 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2450 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2451 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
Tom Herbertdfd86452015-01-12 17:00:38 -08002452 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2453 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
Thomas Graf35114942015-01-15 03:53:55 +01002454 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
Tom Herbert0ace2ca2015-02-10 16:30:32 -08002455 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
stephen hemmingerd3428942012-10-01 12:32:35 +00002456};
2457
2458static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2459{
2460 if (tb[IFLA_ADDRESS]) {
2461 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2462 pr_debug("invalid link address (not ethernet)\n");
2463 return -EINVAL;
2464 }
2465
2466 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2467 pr_debug("invalid all zero ethernet address\n");
2468 return -EADDRNOTAVAIL;
2469 }
2470 }
2471
2472 if (!data)
2473 return -EINVAL;
2474
2475 if (data[IFLA_VXLAN_ID]) {
2476 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2477 if (id >= VXLAN_VID_MASK)
2478 return -ERANGE;
2479 }
2480
stephen hemminger05f47d62012-10-09 20:35:50 +00002481 if (data[IFLA_VXLAN_PORT_RANGE]) {
2482 const struct ifla_vxlan_port_range *p
2483 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2484
2485 if (ntohs(p->high) < ntohs(p->low)) {
2486 pr_debug("port range %u .. %u not valid\n",
2487 ntohs(p->low), ntohs(p->high));
2488 return -EINVAL;
2489 }
2490 }
2491
stephen hemmingerd3428942012-10-01 12:32:35 +00002492 return 0;
2493}
2494
Yan Burman1b13c972013-01-29 23:43:07 +00002495static void vxlan_get_drvinfo(struct net_device *netdev,
2496 struct ethtool_drvinfo *drvinfo)
2497{
2498 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2499 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2500}
2501
2502static const struct ethtool_ops vxlan_ethtool_ops = {
2503 .get_drvinfo = vxlan_get_drvinfo,
2504 .get_link = ethtool_op_get_link,
2505};
2506
stephen hemminger553675f2013-05-16 11:35:20 +00002507static void vxlan_del_work(struct work_struct *work)
2508{
2509 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
Andy Zhouacbf74a2014-09-16 17:31:18 -07002510 udp_tunnel_sock_release(vs->sock);
stephen hemminger553675f2013-05-16 11:35:20 +00002511 kfree_rcu(vs, rcu);
2512}
2513
Tom Herbert3ee64f32014-07-13 19:49:42 -07002514static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2515 __be16 port, u32 flags)
stephen hemminger553675f2013-05-16 11:35:20 +00002516{
Cong Wange4c7ed42013-08-31 13:44:33 +08002517 struct socket *sock;
Tom Herbert3ee64f32014-07-13 19:49:42 -07002518 struct udp_port_cfg udp_conf;
2519 int err;
Cong Wange4c7ed42013-08-31 13:44:33 +08002520
Tom Herbert3ee64f32014-07-13 19:49:42 -07002521 memset(&udp_conf, 0, sizeof(udp_conf));
2522
2523 if (ipv6) {
2524 udp_conf.family = AF_INET6;
Tom Herbert3ee64f32014-07-13 19:49:42 -07002525 udp_conf.use_udp6_rx_checksums =
Alexander Duyck3dc2b6a2014-11-24 20:08:38 -08002526 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
Jiri Benca43a9ef2015-08-28 20:48:22 +02002527 udp_conf.ipv6_v6only = 1;
Tom Herbert3ee64f32014-07-13 19:49:42 -07002528 } else {
2529 udp_conf.family = AF_INET;
Cong Wange4c7ed42013-08-31 13:44:33 +08002530 }
2531
Tom Herbert3ee64f32014-07-13 19:49:42 -07002532 udp_conf.local_udp_port = port;
Cong Wange4c7ed42013-08-31 13:44:33 +08002533
Tom Herbert3ee64f32014-07-13 19:49:42 -07002534 /* Open UDP socket */
2535 err = udp_sock_create(net, &udp_conf, &sock);
2536 if (err < 0)
2537 return ERR_PTR(err);
Cong Wange4c7ed42013-08-31 13:44:33 +08002538
Zhi Yong Wu39deb2c2013-10-28 14:01:48 +08002539 return sock;
Cong Wange4c7ed42013-08-31 13:44:33 +08002540}
2541
2542/* Create new listen socket if needed */
2543static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port,
Tom Herbert359a0ea2014-06-04 17:20:29 -07002544 u32 flags)
Cong Wange4c7ed42013-08-31 13:44:33 +08002545{
2546 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2547 struct vxlan_sock *vs;
2548 struct socket *sock;
Cong Wang31fec5a2013-05-27 22:35:52 +00002549 unsigned int h;
Tom Herbert359a0ea2014-06-04 17:20:29 -07002550 bool ipv6 = !!(flags & VXLAN_F_IPV6);
Andy Zhouacbf74a2014-09-16 17:31:18 -07002551 struct udp_tunnel_sock_cfg tunnel_cfg;
stephen hemminger553675f2013-05-16 11:35:20 +00002552
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02002553 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
Cong Wange4c7ed42013-08-31 13:44:33 +08002554 if (!vs)
stephen hemminger553675f2013-05-16 11:35:20 +00002555 return ERR_PTR(-ENOMEM);
2556
2557 for (h = 0; h < VNI_HASH_SIZE; ++h)
2558 INIT_HLIST_HEAD(&vs->vni_list[h]);
2559
2560 INIT_WORK(&vs->del_work, vxlan_del_work);
2561
Tom Herbert3ee64f32014-07-13 19:49:42 -07002562 sock = vxlan_create_sock(net, ipv6, port, flags);
Zhi Yong Wu39deb2c2013-10-28 14:01:48 +08002563 if (IS_ERR(sock)) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002564 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2565 PTR_ERR(sock));
stephen hemminger553675f2013-05-16 11:35:20 +00002566 kfree(vs);
Duan Jionge50fddc2013-11-01 13:09:43 +08002567 return ERR_CAST(sock);
stephen hemminger553675f2013-05-16 11:35:20 +00002568 }
2569
Cong Wange4c7ed42013-08-31 13:44:33 +08002570 vs->sock = sock;
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002571 atomic_set(&vs->refcnt, 1);
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002572 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
stephen hemminger553675f2013-05-16 11:35:20 +00002573
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02002574 /* Initialize the vxlan udp offloads structure */
2575 vs->udp_offloads.port = port;
2576 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2577 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2578
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002579 spin_lock(&vn->sock_lock);
2580 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02002581 vxlan_notify_add_rx_port(vs);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002582 spin_unlock(&vn->sock_lock);
stephen hemminger553675f2013-05-16 11:35:20 +00002583
2584 /* Mark socket as an encapsulation socket. */
Andy Zhouacbf74a2014-09-16 17:31:18 -07002585 tunnel_cfg.sk_user_data = vs;
2586 tunnel_cfg.encap_type = 1;
2587 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2588 tunnel_cfg.encap_destroy = NULL;
2589
2590 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
Cong Wange4c7ed42013-08-31 13:44:33 +08002591
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002592 return vs;
2593}
stephen hemminger553675f2013-05-16 11:35:20 +00002594
Thomas Graf614732e2015-07-21 10:44:06 +02002595static struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port,
2596 bool no_share, u32 flags)
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002597{
2598 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2599 struct vxlan_sock *vs;
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -02002600 bool ipv6 = flags & VXLAN_F_IPV6;
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002601
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002602 if (!no_share) {
2603 spin_lock(&vn->sock_lock);
2604 vs = vxlan_find_sock(net, ipv6 ? AF_INET6 : AF_INET, port,
2605 flags);
Thomas Graf614732e2015-07-21 10:44:06 +02002606 if (vs) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002607 if (!atomic_add_unless(&vs->refcnt, 1, 0))
2608 vs = ERR_PTR(-EBUSY);
2609 spin_unlock(&vn->sock_lock);
2610 return vs;
2611 }
2612 spin_unlock(&vn->sock_lock);
2613 }
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002614
Thomas Graf614732e2015-07-21 10:44:06 +02002615 return vxlan_socket_create(net, port, flags);
stephen hemminger553675f2013-05-16 11:35:20 +00002616}
2617
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002618static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2619 struct vxlan_config *conf)
stephen hemmingerd3428942012-10-01 12:32:35 +00002620{
Nicolas Dichtel33564bb2015-01-26 22:28:14 +01002621 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
stephen hemmingerd3428942012-10-01 12:32:35 +00002622 struct vxlan_dev *vxlan = netdev_priv(dev);
Atzm Watanabec7995c42013-04-16 02:50:52 +00002623 struct vxlan_rdst *dst = &vxlan->default_dst;
stephen hemmingerd3428942012-10-01 12:32:35 +00002624 int err;
Cong Wange4c7ed42013-08-31 13:44:33 +08002625 bool use_ipv6 = false;
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002626 __be16 default_port = vxlan->cfg.dst_port;
stephen hemmingerd3428942012-10-01 12:32:35 +00002627
Nicolas Dichtel33564bb2015-01-26 22:28:14 +01002628 vxlan->net = src_net;
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002629
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002630 dst->remote_vni = conf->vni;
2631
2632 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
stephen hemmingerd3428942012-10-01 12:32:35 +00002633
Mike Rapoport5933a7b2014-04-01 09:23:01 +03002634 /* Unless IPv6 is explicitly requested, assume IPv4 */
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002635 if (!dst->remote_ip.sa.sa_family)
2636 dst->remote_ip.sa.sa_family = AF_INET;
stephen hemmingerd3428942012-10-01 12:32:35 +00002637
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002638 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
Jiri Benc057ba292015-09-17 16:11:11 +02002639 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2640 if (!IS_ENABLED(CONFIG_IPV6))
2641 return -EPFNOSUPPORT;
Cong Wange4c7ed42013-08-31 13:44:33 +08002642 use_ipv6 = true;
Jiri Benc057ba292015-09-17 16:11:11 +02002643 }
Cong Wange4c7ed42013-08-31 13:44:33 +08002644
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002645 if (conf->remote_ifindex) {
stephen hemminger34e02aa2012-10-09 20:35:53 +00002646 struct net_device *lowerdev
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002647 = __dev_get_by_index(src_net, conf->remote_ifindex);
2648
2649 dst->remote_ifindex = conf->remote_ifindex;
stephen hemmingerd3428942012-10-01 12:32:35 +00002650
stephen hemminger34e02aa2012-10-09 20:35:53 +00002651 if (!lowerdev) {
Atzm Watanabec7995c42013-04-16 02:50:52 +00002652 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
stephen hemminger34e02aa2012-10-09 20:35:53 +00002653 return -ENODEV;
stephen hemmingerd3428942012-10-01 12:32:35 +00002654 }
stephen hemminger34e02aa2012-10-09 20:35:53 +00002655
Cong Wange4c7ed42013-08-31 13:44:33 +08002656#if IS_ENABLED(CONFIG_IPV6)
2657 if (use_ipv6) {
2658 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2659 if (idev && idev->cnf.disable_ipv6) {
2660 pr_info("IPv6 is disabled via sysctl\n");
2661 return -EPERM;
2662 }
2663 vxlan->flags |= VXLAN_F_IPV6;
2664 }
2665#endif
2666
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002667 if (!conf->mtu)
Cong Wange4c7ed42013-08-31 13:44:33 +08002668 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
Alexander Duyck1ba56fb2012-11-13 13:10:59 +00002669
Cong Wang2853af62014-06-12 11:53:10 -07002670 dev->needed_headroom = lowerdev->hard_header_len +
Cong Wange4c7ed42013-08-31 13:44:33 +08002671 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
Jiri Benc9dc2ad12015-09-17 16:11:10 +02002672 } else if (use_ipv6) {
fan.du7bda7012014-01-03 10:18:58 +08002673 vxlan->flags |= VXLAN_F_IPV6;
Jiri Benc9dc2ad12015-09-17 16:11:10 +02002674 dev->needed_headroom = ETH_HLEN + VXLAN6_HEADROOM;
2675 } else {
2676 dev->needed_headroom = ETH_HLEN + VXLAN_HEADROOM;
2677 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002678
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002679 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2680 if (!vxlan->cfg.dst_port)
2681 vxlan->cfg.dst_port = default_port;
2682 vxlan->flags |= conf->flags;
stephen hemmingerd3428942012-10-01 12:32:35 +00002683
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002684 if (!vxlan->cfg.age_interval)
2685 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
Vincent Bernatafb97182012-10-30 10:27:16 +00002686
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002687 if (vxlan_find_vni(src_net, conf->vni, use_ipv6 ? AF_INET6 : AF_INET,
2688 vxlan->cfg.dst_port, vxlan->flags))
stephen hemminger553675f2013-05-16 11:35:20 +00002689 return -EEXIST;
stephen hemminger553675f2013-05-16 11:35:20 +00002690
Wilfried Klaebe7ad24ea2014-05-11 00:12:32 +00002691 dev->ethtool_ops = &vxlan_ethtool_ops;
Yan Burman1b13c972013-01-29 23:43:07 +00002692
Sridhar Samudrala2936b6a2013-09-17 12:12:40 -07002693 /* create an fdb entry for a valid default destination */
2694 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2695 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2696 &vxlan->default_dst.remote_ip,
2697 NUD_REACHABLE|NUD_PERMANENT,
2698 NLM_F_EXCL|NLM_F_CREATE,
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002699 vxlan->cfg.dst_port,
Sridhar Samudrala2936b6a2013-09-17 12:12:40 -07002700 vxlan->default_dst.remote_vni,
2701 vxlan->default_dst.remote_ifindex,
2702 NTF_SELF);
2703 if (err)
2704 return err;
2705 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002706
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002707 err = register_netdevice(dev);
2708 if (err) {
Stephen Hemmingerba609e92013-06-25 17:06:01 -07002709 vxlan_fdb_delete_default(vxlan);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002710 return err;
2711 }
2712
stephen hemminger553675f2013-05-16 11:35:20 +00002713 list_add(&vxlan->next, &vn->vxlan_list);
stephen hemminger553675f2013-05-16 11:35:20 +00002714
2715 return 0;
stephen hemmingerd3428942012-10-01 12:32:35 +00002716}
2717
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002718struct net_device *vxlan_dev_create(struct net *net, const char *name,
2719 u8 name_assign_type, struct vxlan_config *conf)
2720{
2721 struct nlattr *tb[IFLA_MAX+1];
2722 struct net_device *dev;
2723 int err;
2724
2725 memset(&tb, 0, sizeof(tb));
2726
2727 dev = rtnl_create_link(net, name, name_assign_type,
2728 &vxlan_link_ops, tb);
2729 if (IS_ERR(dev))
2730 return dev;
2731
2732 err = vxlan_dev_configure(net, dev, conf);
2733 if (err < 0) {
2734 free_netdev(dev);
2735 return ERR_PTR(err);
2736 }
2737
2738 return dev;
2739}
2740EXPORT_SYMBOL_GPL(vxlan_dev_create);
2741
2742static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2743 struct nlattr *tb[], struct nlattr *data[])
2744{
2745 struct vxlan_config conf;
2746 int err;
2747
2748 if (!data[IFLA_VXLAN_ID])
2749 return -EINVAL;
2750
2751 memset(&conf, 0, sizeof(conf));
2752 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
2753
2754 if (data[IFLA_VXLAN_GROUP]) {
2755 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2756 } else if (data[IFLA_VXLAN_GROUP6]) {
2757 if (!IS_ENABLED(CONFIG_IPV6))
2758 return -EPFNOSUPPORT;
2759
2760 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2761 conf.remote_ip.sa.sa_family = AF_INET6;
2762 }
2763
2764 if (data[IFLA_VXLAN_LOCAL]) {
2765 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2766 conf.saddr.sa.sa_family = AF_INET;
2767 } else if (data[IFLA_VXLAN_LOCAL6]) {
2768 if (!IS_ENABLED(CONFIG_IPV6))
2769 return -EPFNOSUPPORT;
2770
2771 /* TODO: respect scope id */
2772 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2773 conf.saddr.sa.sa_family = AF_INET6;
2774 }
2775
2776 if (data[IFLA_VXLAN_LINK])
2777 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2778
2779 if (data[IFLA_VXLAN_TOS])
2780 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2781
2782 if (data[IFLA_VXLAN_TTL])
2783 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2784
2785 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2786 conf.flags |= VXLAN_F_LEARN;
2787
2788 if (data[IFLA_VXLAN_AGEING])
2789 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2790
2791 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2792 conf.flags |= VXLAN_F_PROXY;
2793
2794 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2795 conf.flags |= VXLAN_F_RSC;
2796
2797 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2798 conf.flags |= VXLAN_F_L2MISS;
2799
2800 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2801 conf.flags |= VXLAN_F_L3MISS;
2802
2803 if (data[IFLA_VXLAN_LIMIT])
2804 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2805
Alexei Starovoitovf8a9b1b2015-07-30 20:10:22 -07002806 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2807 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2808 conf.flags |= VXLAN_F_COLLECT_METADATA;
2809
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002810 if (data[IFLA_VXLAN_PORT_RANGE]) {
2811 const struct ifla_vxlan_port_range *p
2812 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2813 conf.port_min = ntohs(p->low);
2814 conf.port_max = ntohs(p->high);
2815 }
2816
2817 if (data[IFLA_VXLAN_PORT])
2818 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2819
2820 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2821 conf.flags |= VXLAN_F_UDP_CSUM;
2822
2823 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2824 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2825 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2826
2827 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2828 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2829 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2830
2831 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2832 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2833 conf.flags |= VXLAN_F_REMCSUM_TX;
2834
2835 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2836 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2837 conf.flags |= VXLAN_F_REMCSUM_RX;
2838
2839 if (data[IFLA_VXLAN_GBP])
2840 conf.flags |= VXLAN_F_GBP;
2841
2842 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
2843 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
2844
2845 err = vxlan_dev_configure(src_net, dev, &conf);
2846 switch (err) {
2847 case -ENODEV:
2848 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2849 break;
2850
2851 case -EPERM:
2852 pr_info("IPv6 is disabled via sysctl\n");
2853 break;
2854
2855 case -EEXIST:
2856 pr_info("duplicate VNI %u\n", conf.vni);
2857 break;
2858 }
2859
2860 return err;
2861}
2862
stephen hemmingerd3428942012-10-01 12:32:35 +00002863static void vxlan_dellink(struct net_device *dev, struct list_head *head)
2864{
2865 struct vxlan_dev *vxlan = netdev_priv(dev);
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002866 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
stephen hemmingerd3428942012-10-01 12:32:35 +00002867
stephen hemmingerfe5c3562013-07-13 10:18:18 -07002868 spin_lock(&vn->sock_lock);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002869 if (!hlist_unhashed(&vxlan->hlist))
2870 hlist_del_rcu(&vxlan->hlist);
stephen hemmingerfe5c3562013-07-13 10:18:18 -07002871 spin_unlock(&vn->sock_lock);
2872
Tom Herbert58ce31c2015-08-19 17:07:33 -07002873 gro_cells_destroy(&vxlan->gro_cells);
stephen hemminger553675f2013-05-16 11:35:20 +00002874 list_del(&vxlan->next);
stephen hemmingerd3428942012-10-01 12:32:35 +00002875 unregister_netdevice_queue(dev, head);
2876}
2877
2878static size_t vxlan_get_size(const struct net_device *dev)
2879{
2880
2881 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
Cong Wange4c7ed42013-08-31 13:44:33 +08002882 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
stephen hemmingerd3428942012-10-01 12:32:35 +00002883 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
Cong Wange4c7ed42013-08-31 13:44:33 +08002884 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
stephen hemmingerd3428942012-10-01 12:32:35 +00002885 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
2886 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
2887 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
David Stevense4f67ad2012-11-20 02:50:14 +00002888 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
2889 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
2890 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
2891 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07002892 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
stephen hemmingerd3428942012-10-01 12:32:35 +00002893 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
2894 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
stephen hemminger05f47d62012-10-09 20:35:50 +00002895 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
Tom Herbert359a0ea2014-06-04 17:20:29 -07002896 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
2897 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
2898 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
2899 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
Tom Herbertdfd86452015-01-12 17:00:38 -08002900 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
2901 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
stephen hemmingerd3428942012-10-01 12:32:35 +00002902 0;
2903}
2904
2905static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
2906{
2907 const struct vxlan_dev *vxlan = netdev_priv(dev);
Atzm Watanabec7995c42013-04-16 02:50:52 +00002908 const struct vxlan_rdst *dst = &vxlan->default_dst;
stephen hemminger05f47d62012-10-09 20:35:50 +00002909 struct ifla_vxlan_port_range ports = {
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002910 .low = htons(vxlan->cfg.port_min),
2911 .high = htons(vxlan->cfg.port_max),
stephen hemminger05f47d62012-10-09 20:35:50 +00002912 };
stephen hemmingerd3428942012-10-01 12:32:35 +00002913
Atzm Watanabec7995c42013-04-16 02:50:52 +00002914 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
stephen hemmingerd3428942012-10-01 12:32:35 +00002915 goto nla_put_failure;
2916
Cong Wange4c7ed42013-08-31 13:44:33 +08002917 if (!vxlan_addr_any(&dst->remote_ip)) {
2918 if (dst->remote_ip.sa.sa_family == AF_INET) {
Jiri Benc930345e2015-03-29 16:59:25 +02002919 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
2920 dst->remote_ip.sin.sin_addr.s_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08002921 goto nla_put_failure;
2922#if IS_ENABLED(CONFIG_IPV6)
2923 } else {
Jiri Benc930345e2015-03-29 16:59:25 +02002924 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
2925 &dst->remote_ip.sin6.sin6_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08002926 goto nla_put_failure;
2927#endif
2928 }
2929 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002930
Atzm Watanabec7995c42013-04-16 02:50:52 +00002931 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
stephen hemmingerd3428942012-10-01 12:32:35 +00002932 goto nla_put_failure;
2933
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002934 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
2935 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
Jiri Benc930345e2015-03-29 16:59:25 +02002936 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002937 vxlan->cfg.saddr.sin.sin_addr.s_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08002938 goto nla_put_failure;
2939#if IS_ENABLED(CONFIG_IPV6)
2940 } else {
Jiri Benc930345e2015-03-29 16:59:25 +02002941 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002942 &vxlan->cfg.saddr.sin6.sin6_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08002943 goto nla_put_failure;
2944#endif
2945 }
2946 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002947
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002948 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
2949 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
David Stevense4f67ad2012-11-20 02:50:14 +00002950 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
2951 !!(vxlan->flags & VXLAN_F_LEARN)) ||
2952 nla_put_u8(skb, IFLA_VXLAN_PROXY,
2953 !!(vxlan->flags & VXLAN_F_PROXY)) ||
2954 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
2955 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
2956 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
2957 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
2958 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07002959 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
2960 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002961 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
2962 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
2963 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
Tom Herbert359a0ea2014-06-04 17:20:29 -07002964 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
2965 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
2966 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
2967 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
2968 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
Tom Herbertdfd86452015-01-12 17:00:38 -08002969 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
2970 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
2971 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
2972 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
2973 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
stephen hemmingerd3428942012-10-01 12:32:35 +00002974 goto nla_put_failure;
2975
stephen hemminger05f47d62012-10-09 20:35:50 +00002976 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
2977 goto nla_put_failure;
2978
Thomas Graf35114942015-01-15 03:53:55 +01002979 if (vxlan->flags & VXLAN_F_GBP &&
2980 nla_put_flag(skb, IFLA_VXLAN_GBP))
2981 goto nla_put_failure;
2982
Tom Herbert0ace2ca2015-02-10 16:30:32 -08002983 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
2984 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
2985 goto nla_put_failure;
2986
stephen hemmingerd3428942012-10-01 12:32:35 +00002987 return 0;
2988
2989nla_put_failure:
2990 return -EMSGSIZE;
2991}
2992
Nicolas Dichtel1728d4f2015-01-15 15:11:17 +01002993static struct net *vxlan_get_link_net(const struct net_device *dev)
2994{
2995 struct vxlan_dev *vxlan = netdev_priv(dev);
2996
2997 return vxlan->net;
2998}
2999
stephen hemmingerd3428942012-10-01 12:32:35 +00003000static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3001 .kind = "vxlan",
3002 .maxtype = IFLA_VXLAN_MAX,
3003 .policy = vxlan_policy,
3004 .priv_size = sizeof(struct vxlan_dev),
3005 .setup = vxlan_setup,
3006 .validate = vxlan_validate,
3007 .newlink = vxlan_newlink,
3008 .dellink = vxlan_dellink,
3009 .get_size = vxlan_get_size,
3010 .fill_info = vxlan_fill_info,
Nicolas Dichtel1728d4f2015-01-15 15:11:17 +01003011 .get_link_net = vxlan_get_link_net,
stephen hemmingerd3428942012-10-01 12:32:35 +00003012};
3013
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003014static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3015 struct net_device *dev)
3016{
3017 struct vxlan_dev *vxlan, *next;
3018 LIST_HEAD(list_kill);
3019
3020 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3021 struct vxlan_rdst *dst = &vxlan->default_dst;
3022
3023 /* In case we created vxlan device with carrier
3024 * and we loose the carrier due to module unload
3025 * we also need to remove vxlan device. In other
3026 * cases, it's not necessary and remote_ifindex
3027 * is 0 here, so no matches.
3028 */
3029 if (dst->remote_ifindex == dev->ifindex)
3030 vxlan_dellink(vxlan->dev, &list_kill);
3031 }
3032
3033 unregister_netdevice_many(&list_kill);
3034}
3035
3036static int vxlan_lowerdev_event(struct notifier_block *unused,
3037 unsigned long event, void *ptr)
3038{
3039 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
Daniel Borkmann783c1462014-01-22 21:07:53 +01003040 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003041
Daniel Borkmann783c1462014-01-22 21:07:53 +01003042 if (event == NETDEV_UNREGISTER)
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003043 vxlan_handle_lowerdev_unregister(vn, dev);
3044
3045 return NOTIFY_DONE;
3046}
3047
3048static struct notifier_block vxlan_notifier_block __read_mostly = {
3049 .notifier_call = vxlan_lowerdev_event,
3050};
3051
stephen hemmingerd3428942012-10-01 12:32:35 +00003052static __net_init int vxlan_init_net(struct net *net)
3053{
3054 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
Cong Wang31fec5a2013-05-27 22:35:52 +00003055 unsigned int h;
stephen hemmingerd3428942012-10-01 12:32:35 +00003056
stephen hemminger553675f2013-05-16 11:35:20 +00003057 INIT_LIST_HEAD(&vn->vxlan_list);
Stephen Hemminger1c51a912013-06-17 14:16:11 -07003058 spin_lock_init(&vn->sock_lock);
stephen hemmingerd3428942012-10-01 12:32:35 +00003059
stephen hemminger553675f2013-05-16 11:35:20 +00003060 for (h = 0; h < PORT_HASH_SIZE; ++h)
3061 INIT_HLIST_HEAD(&vn->sock_list[h]);
stephen hemmingerd3428942012-10-01 12:32:35 +00003062
3063 return 0;
3064}
3065
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003066static void __net_exit vxlan_exit_net(struct net *net)
3067{
3068 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3069 struct vxlan_dev *vxlan, *next;
3070 struct net_device *dev, *aux;
3071 LIST_HEAD(list);
3072
3073 rtnl_lock();
3074 for_each_netdev_safe(net, dev, aux)
3075 if (dev->rtnl_link_ops == &vxlan_link_ops)
3076 unregister_netdevice_queue(dev, &list);
3077
3078 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3079 /* If vxlan->dev is in the same netns, it has already been added
3080 * to the list by the previous loop.
3081 */
Tom Herbert58ce31c2015-08-19 17:07:33 -07003082 if (!net_eq(dev_net(vxlan->dev), net)) {
3083 gro_cells_destroy(&vxlan->gro_cells);
John W. Linville13c3ed62015-05-18 13:51:24 -04003084 unregister_netdevice_queue(vxlan->dev, &list);
Tom Herbert58ce31c2015-08-19 17:07:33 -07003085 }
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003086 }
3087
3088 unregister_netdevice_many(&list);
3089 rtnl_unlock();
3090}
3091
stephen hemmingerd3428942012-10-01 12:32:35 +00003092static struct pernet_operations vxlan_net_ops = {
3093 .init = vxlan_init_net,
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003094 .exit = vxlan_exit_net,
stephen hemmingerd3428942012-10-01 12:32:35 +00003095 .id = &vxlan_net_id,
3096 .size = sizeof(struct vxlan_net),
3097};
3098
3099static int __init vxlan_init_module(void)
3100{
3101 int rc;
3102
Stephen Hemminger758c57d2013-06-17 14:16:09 -07003103 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3104 if (!vxlan_wq)
3105 return -ENOMEM;
3106
stephen hemmingerd3428942012-10-01 12:32:35 +00003107 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3108
Daniel Borkmann783c1462014-01-22 21:07:53 +01003109 rc = register_pernet_subsys(&vxlan_net_ops);
stephen hemmingerd3428942012-10-01 12:32:35 +00003110 if (rc)
3111 goto out1;
3112
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003113 rc = register_netdevice_notifier(&vxlan_notifier_block);
stephen hemmingerd3428942012-10-01 12:32:35 +00003114 if (rc)
3115 goto out2;
3116
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003117 rc = rtnl_link_register(&vxlan_link_ops);
3118 if (rc)
3119 goto out3;
stephen hemmingerd3428942012-10-01 12:32:35 +00003120
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003121 return 0;
3122out3:
3123 unregister_netdevice_notifier(&vxlan_notifier_block);
stephen hemmingerd3428942012-10-01 12:32:35 +00003124out2:
Daniel Borkmann783c1462014-01-22 21:07:53 +01003125 unregister_pernet_subsys(&vxlan_net_ops);
stephen hemmingerd3428942012-10-01 12:32:35 +00003126out1:
Stephen Hemminger758c57d2013-06-17 14:16:09 -07003127 destroy_workqueue(vxlan_wq);
stephen hemmingerd3428942012-10-01 12:32:35 +00003128 return rc;
3129}
Cong Wang7332a132013-05-27 22:35:53 +00003130late_initcall(vxlan_init_module);
stephen hemmingerd3428942012-10-01 12:32:35 +00003131
3132static void __exit vxlan_cleanup_module(void)
3133{
Stephen Hemmingerb7153982013-06-17 14:16:09 -07003134 rtnl_link_unregister(&vxlan_link_ops);
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003135 unregister_netdevice_notifier(&vxlan_notifier_block);
Stephen Hemminger758c57d2013-06-17 14:16:09 -07003136 destroy_workqueue(vxlan_wq);
Daniel Borkmann783c1462014-01-22 21:07:53 +01003137 unregister_pernet_subsys(&vxlan_net_ops);
3138 /* rcu_barrier() is called by netns */
stephen hemmingerd3428942012-10-01 12:32:35 +00003139}
3140module_exit(vxlan_cleanup_module);
3141
3142MODULE_LICENSE("GPL");
3143MODULE_VERSION(VXLAN_VERSION);
stephen hemminger3b8df3c2013-04-27 11:31:52 +00003144MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
Jesse Brandeburgead51392014-01-17 11:00:33 -08003145MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
stephen hemmingerd3428942012-10-01 12:32:35 +00003146MODULE_ALIAS_RTNL_LINK("vxlan");