blob: 65439188c5829e1cce0464e6c34be1c8bb5c18c3 [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
Li RongQing7256eac2016-01-29 09:43:47 +080076static const u8 all_zeros_mac[ETH_ALEN + 2];
Mike Rapoportafbd8ba2013-06-25 16:01:51 +030077
Jiri Benc205f3562015-09-24 13:50:01 +020078static int vxlan_sock_add(struct vxlan_dev *vxlan);
Thomas Graf614732e2015-07-21 10:44:06 +020079
stephen hemminger553675f2013-05-16 11:35:20 +000080/* per-network namespace private data for this module */
81struct vxlan_net {
82 struct list_head vxlan_list;
83 struct hlist_head sock_list[PORT_HASH_SIZE];
Stephen Hemminger1c51a912013-06-17 14:16:11 -070084 spinlock_t sock_lock;
stephen hemminger553675f2013-05-16 11:35:20 +000085};
86
stephen hemmingerd3428942012-10-01 12:32:35 +000087/* Forwarding table entry */
88struct vxlan_fdb {
89 struct hlist_node hlist; /* linked list of entries */
90 struct rcu_head rcu;
91 unsigned long updated; /* jiffies */
92 unsigned long used;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -070093 struct list_head remotes;
Sowmini Varadhan7177a3b2015-07-20 09:54:50 +020094 u8 eth_addr[ETH_ALEN];
stephen hemmingerd3428942012-10-01 12:32:35 +000095 u16 state; /* see ndm_state */
David Stevensae884082013-04-19 00:36:26 +000096 u8 flags; /* see ndm_flags */
stephen hemmingerd3428942012-10-01 12:32:35 +000097};
98
stephen hemmingerd3428942012-10-01 12:32:35 +000099/* salt for hash table */
100static u32 vxlan_salt __read_mostly;
Stephen Hemminger758c57d2013-06-17 14:16:09 -0700101static struct workqueue_struct *vxlan_wq;
stephen hemmingerd3428942012-10-01 12:32:35 +0000102
Thomas Grafee122c72015-07-21 10:43:58 +0200103static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
104{
Thomas Grafe7030872015-07-21 10:44:01 +0200105 return vs->flags & VXLAN_F_COLLECT_METADATA ||
106 ip_tunnel_collect_metadata();
Thomas Grafee122c72015-07-21 10:43:58 +0200107}
108
Cong Wange4c7ed42013-08-31 13:44:33 +0800109#if IS_ENABLED(CONFIG_IPV6)
110static inline
111bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
112{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200113 if (a->sa.sa_family != b->sa.sa_family)
114 return false;
115 if (a->sa.sa_family == AF_INET6)
116 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
117 else
118 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
Cong Wange4c7ed42013-08-31 13:44:33 +0800119}
120
121static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
122{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200123 if (ipa->sa.sa_family == AF_INET6)
124 return ipv6_addr_any(&ipa->sin6.sin6_addr);
125 else
126 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
Cong Wange4c7ed42013-08-31 13:44:33 +0800127}
128
129static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
130{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200131 if (ipa->sa.sa_family == AF_INET6)
132 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
133 else
134 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
Cong Wange4c7ed42013-08-31 13:44:33 +0800135}
136
137static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
138{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200139 if (nla_len(nla) >= sizeof(struct in6_addr)) {
Jiri Benc67b61f62015-03-29 16:59:26 +0200140 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200141 ip->sa.sa_family = AF_INET6;
142 return 0;
143 } else if (nla_len(nla) >= sizeof(__be32)) {
Jiri Benc67b61f62015-03-29 16:59:26 +0200144 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200145 ip->sa.sa_family = AF_INET;
146 return 0;
147 } else {
148 return -EAFNOSUPPORT;
149 }
Cong Wange4c7ed42013-08-31 13:44:33 +0800150}
151
152static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
Jiri Bencf0ef3122015-03-29 16:17:37 +0200153 const union vxlan_addr *ip)
Cong Wange4c7ed42013-08-31 13:44:33 +0800154{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200155 if (ip->sa.sa_family == AF_INET6)
Jiri Benc930345e2015-03-29 16:59:25 +0200156 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200157 else
Jiri Benc930345e2015-03-29 16:59:25 +0200158 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
Cong Wange4c7ed42013-08-31 13:44:33 +0800159}
160
161#else /* !CONFIG_IPV6 */
162
163static inline
164bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
165{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200166 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
Cong Wange4c7ed42013-08-31 13:44:33 +0800167}
168
169static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
170{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200171 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
Cong Wange4c7ed42013-08-31 13:44:33 +0800172}
173
174static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
175{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200176 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
Cong Wange4c7ed42013-08-31 13:44:33 +0800177}
178
179static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
180{
Jiri Bencf0ef3122015-03-29 16:17:37 +0200181 if (nla_len(nla) >= sizeof(struct in6_addr)) {
182 return -EAFNOSUPPORT;
183 } else if (nla_len(nla) >= sizeof(__be32)) {
Jiri Benc67b61f62015-03-29 16:59:26 +0200184 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
Jiri Bencf0ef3122015-03-29 16:17:37 +0200185 ip->sa.sa_family = AF_INET;
186 return 0;
187 } else {
188 return -EAFNOSUPPORT;
189 }
Cong Wange4c7ed42013-08-31 13:44:33 +0800190}
191
192static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
Jiri Bencf0ef3122015-03-29 16:17:37 +0200193 const union vxlan_addr *ip)
Cong Wange4c7ed42013-08-31 13:44:33 +0800194{
Jiri Benc930345e2015-03-29 16:59:25 +0200195 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
Cong Wange4c7ed42013-08-31 13:44:33 +0800196}
197#endif
198
stephen hemminger553675f2013-05-16 11:35:20 +0000199/* Virtual Network hash table head */
200static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id)
201{
202 return &vs->vni_list[hash_32(id, VNI_HASH_BITS)];
203}
204
205/* Socket hash table head */
206static inline struct hlist_head *vs_head(struct net *net, __be16 port)
stephen hemmingerd3428942012-10-01 12:32:35 +0000207{
208 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
209
stephen hemminger553675f2013-05-16 11:35:20 +0000210 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
211}
212
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700213/* First remote destination for a forwarding entry.
214 * Guaranteed to be non-NULL because remotes are never deleted.
215 */
stephen hemminger5ca54612013-08-04 17:17:39 -0700216static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700217{
stephen hemminger5ca54612013-08-04 17:17:39 -0700218 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
219}
220
221static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
222{
223 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700224}
225
Thomas Grafac5132d2015-01-15 03:53:56 +0100226/* Find VXLAN socket based on network namespace, address family and UDP port
227 * and enabled unshareable flags.
228 */
229static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
230 __be16 port, u32 flags)
stephen hemminger553675f2013-05-16 11:35:20 +0000231{
232 struct vxlan_sock *vs;
Tom Herbertaf33c1a2015-01-20 11:23:05 -0800233
234 flags &= VXLAN_F_RCV_FLAGS;
stephen hemminger553675f2013-05-16 11:35:20 +0000235
236 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -0200237 if (inet_sk(vs->sock->sk)->inet_sport == port &&
Jiri Benc705cc622015-08-20 13:56:28 +0200238 vxlan_get_sk_family(vs) == family &&
Tom Herbertaf33c1a2015-01-20 11:23:05 -0800239 vs->flags == flags)
stephen hemminger553675f2013-05-16 11:35:20 +0000240 return vs;
241 }
242 return NULL;
stephen hemmingerd3428942012-10-01 12:32:35 +0000243}
244
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700245static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id)
stephen hemmingerd3428942012-10-01 12:32:35 +0000246{
247 struct vxlan_dev *vxlan;
stephen hemmingerd3428942012-10-01 12:32:35 +0000248
stephen hemminger553675f2013-05-16 11:35:20 +0000249 hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) {
Atzm Watanabec7995c42013-04-16 02:50:52 +0000250 if (vxlan->default_dst.remote_vni == id)
stephen hemmingerd3428942012-10-01 12:32:35 +0000251 return vxlan;
252 }
253
254 return NULL;
255}
256
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700257/* Look up VNI in a per net namespace table */
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -0200258static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id,
Thomas Grafac5132d2015-01-15 03:53:56 +0100259 sa_family_t family, __be16 port,
260 u32 flags)
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700261{
262 struct vxlan_sock *vs;
263
Thomas Grafac5132d2015-01-15 03:53:56 +0100264 vs = vxlan_find_sock(net, family, port, flags);
Pravin B Shelar5cfccc52013-08-19 11:23:02 -0700265 if (!vs)
266 return NULL;
267
268 return vxlan_vs_find_vni(vs, id);
269}
270
stephen hemmingerd3428942012-10-01 12:32:35 +0000271/* Fill in neighbour message in skbuff. */
272static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
Stephen Hemminger234f5b72013-06-17 14:16:41 -0700273 const struct vxlan_fdb *fdb,
274 u32 portid, u32 seq, int type, unsigned int flags,
275 const struct vxlan_rdst *rdst)
stephen hemmingerd3428942012-10-01 12:32:35 +0000276{
277 unsigned long now = jiffies;
278 struct nda_cacheinfo ci;
279 struct nlmsghdr *nlh;
280 struct ndmsg *ndm;
David Stevense4f67ad2012-11-20 02:50:14 +0000281 bool send_ip, send_eth;
stephen hemmingerd3428942012-10-01 12:32:35 +0000282
283 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
284 if (nlh == NULL)
285 return -EMSGSIZE;
286
287 ndm = nlmsg_data(nlh);
288 memset(ndm, 0, sizeof(*ndm));
David Stevense4f67ad2012-11-20 02:50:14 +0000289
290 send_eth = send_ip = true;
291
292 if (type == RTM_GETNEIGH) {
293 ndm->ndm_family = AF_INET;
Cong Wange4c7ed42013-08-31 13:44:33 +0800294 send_ip = !vxlan_addr_any(&rdst->remote_ip);
David Stevense4f67ad2012-11-20 02:50:14 +0000295 send_eth = !is_zero_ether_addr(fdb->eth_addr);
296 } else
297 ndm->ndm_family = AF_BRIDGE;
stephen hemmingerd3428942012-10-01 12:32:35 +0000298 ndm->ndm_state = fdb->state;
299 ndm->ndm_ifindex = vxlan->dev->ifindex;
David Stevensae884082013-04-19 00:36:26 +0000300 ndm->ndm_flags = fdb->flags;
Jun Zhao545469f2014-07-26 00:38:59 +0800301 ndm->ndm_type = RTN_UNICAST;
stephen hemmingerd3428942012-10-01 12:32:35 +0000302
Nicolas Dichtel193523b2015-01-20 15:15:47 +0100303 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
Nicolas Dichtel49670822015-01-26 14:10:53 +0100304 nla_put_s32(skb, NDA_LINK_NETNSID,
Nicolas Dichtel7a0877d2015-05-07 11:02:49 +0200305 peernet2id_alloc(dev_net(vxlan->dev), vxlan->net)))
Nicolas Dichtel193523b2015-01-20 15:15:47 +0100306 goto nla_put_failure;
307
David Stevense4f67ad2012-11-20 02:50:14 +0000308 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
stephen hemmingerd3428942012-10-01 12:32:35 +0000309 goto nla_put_failure;
310
Cong Wange4c7ed42013-08-31 13:44:33 +0800311 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
David Stevens66817122013-03-15 04:35:51 +0000312 goto nla_put_failure;
313
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200314 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
David Stevens66817122013-03-15 04:35:51 +0000315 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
316 goto nla_put_failure;
Atzm Watanabec7995c42013-04-16 02:50:52 +0000317 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
Pravin B Shelar60d9d4c2013-06-20 00:26:31 -0700318 nla_put_u32(skb, NDA_VNI, rdst->remote_vni))
David Stevens66817122013-03-15 04:35:51 +0000319 goto nla_put_failure;
320 if (rdst->remote_ifindex &&
321 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
stephen hemmingerd3428942012-10-01 12:32:35 +0000322 goto nla_put_failure;
323
324 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
325 ci.ndm_confirmed = 0;
326 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
327 ci.ndm_refcnt = 0;
328
329 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
330 goto nla_put_failure;
331
Johannes Berg053c0952015-01-16 22:09:00 +0100332 nlmsg_end(skb, nlh);
333 return 0;
stephen hemmingerd3428942012-10-01 12:32:35 +0000334
335nla_put_failure:
336 nlmsg_cancel(skb, nlh);
337 return -EMSGSIZE;
338}
339
340static inline size_t vxlan_nlmsg_size(void)
341{
342 return NLMSG_ALIGN(sizeof(struct ndmsg))
343 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
Cong Wange4c7ed42013-08-31 13:44:33 +0800344 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
stephen hemminger73cf3312013-04-27 11:31:54 +0000345 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
David Stevens66817122013-03-15 04:35:51 +0000346 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
347 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
Nicolas Dichtel49670822015-01-26 14:10:53 +0100348 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
stephen hemmingerd3428942012-10-01 12:32:35 +0000349 + nla_total_size(sizeof(struct nda_cacheinfo));
350}
351
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200352static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
353 struct vxlan_rdst *rd, int type)
stephen hemmingerd3428942012-10-01 12:32:35 +0000354{
355 struct net *net = dev_net(vxlan->dev);
356 struct sk_buff *skb;
357 int err = -ENOBUFS;
358
359 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
360 if (skb == NULL)
361 goto errout;
362
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200363 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
stephen hemmingerd3428942012-10-01 12:32:35 +0000364 if (err < 0) {
365 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
366 WARN_ON(err == -EMSGSIZE);
367 kfree_skb(skb);
368 goto errout;
369 }
370
371 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
372 return;
373errout:
374 if (err < 0)
375 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
376}
377
Cong Wange4c7ed42013-08-31 13:44:33 +0800378static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
David Stevense4f67ad2012-11-20 02:50:14 +0000379{
380 struct vxlan_dev *vxlan = netdev_priv(dev);
Stephen Hemmingerbb3fd682013-06-17 14:16:40 -0700381 struct vxlan_fdb f = {
382 .state = NUD_STALE,
383 };
384 struct vxlan_rdst remote = {
Cong Wange4c7ed42013-08-31 13:44:33 +0800385 .remote_ip = *ipa, /* goes to NDA_DST */
Stephen Hemmingerbb3fd682013-06-17 14:16:40 -0700386 .remote_vni = VXLAN_N_VID,
387 };
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700388
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200389 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
David Stevense4f67ad2012-11-20 02:50:14 +0000390}
391
392static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
393{
Stephen Hemmingerbb3fd682013-06-17 14:16:40 -0700394 struct vxlan_fdb f = {
395 .state = NUD_STALE,
396 };
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200397 struct vxlan_rdst remote = { };
David Stevense4f67ad2012-11-20 02:50:14 +0000398
David Stevense4f67ad2012-11-20 02:50:14 +0000399 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
400
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200401 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
David Stevense4f67ad2012-11-20 02:50:14 +0000402}
403
stephen hemmingerd3428942012-10-01 12:32:35 +0000404/* Hash Ethernet address */
405static u32 eth_hash(const unsigned char *addr)
406{
407 u64 value = get_unaligned((u64 *)addr);
408
409 /* only want 6 bytes */
410#ifdef __BIG_ENDIAN
stephen hemmingerd3428942012-10-01 12:32:35 +0000411 value >>= 16;
stephen hemminger321fb992012-10-09 20:35:47 +0000412#else
413 value <<= 16;
stephen hemmingerd3428942012-10-01 12:32:35 +0000414#endif
415 return hash_64(value, FDB_HASH_BITS);
416}
417
418/* Hash chain to use given mac address */
419static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
420 const u8 *mac)
421{
422 return &vxlan->fdb_head[eth_hash(mac)];
423}
424
425/* Look up Ethernet address in forwarding table */
Sridhar Samudrala014be2c2013-05-17 06:39:07 +0000426static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
stephen hemmingerd3428942012-10-01 12:32:35 +0000427 const u8 *mac)
stephen hemmingerd3428942012-10-01 12:32:35 +0000428{
429 struct hlist_head *head = vxlan_fdb_head(vxlan, mac);
430 struct vxlan_fdb *f;
stephen hemmingerd3428942012-10-01 12:32:35 +0000431
Sasha Levinb67bfe02013-02-27 17:06:00 -0800432 hlist_for_each_entry_rcu(f, head, hlist) {
Joe Perches7367d0b2013-09-01 11:51:23 -0700433 if (ether_addr_equal(mac, f->eth_addr))
stephen hemmingerd3428942012-10-01 12:32:35 +0000434 return f;
435 }
436
437 return NULL;
438}
439
Sridhar Samudrala014be2c2013-05-17 06:39:07 +0000440static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
441 const u8 *mac)
442{
443 struct vxlan_fdb *f;
444
445 f = __vxlan_find_mac(vxlan, mac);
446 if (f)
447 f->used = jiffies;
448
449 return f;
450}
451
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300452/* caller should hold vxlan->hash_lock */
453static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
Cong Wange4c7ed42013-08-31 13:44:33 +0800454 union vxlan_addr *ip, __be16 port,
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300455 __u32 vni, __u32 ifindex)
456{
457 struct vxlan_rdst *rd;
458
459 list_for_each_entry(rd, &f->remotes, list) {
Cong Wange4c7ed42013-08-31 13:44:33 +0800460 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300461 rd->remote_port == port &&
462 rd->remote_vni == vni &&
463 rd->remote_ifindex == ifindex)
464 return rd;
465 }
466
467 return NULL;
468}
469
Thomas Richter906dc182013-07-19 17:20:07 +0200470/* Replace destination of unicast mac */
471static int vxlan_fdb_replace(struct vxlan_fdb *f,
Cong Wange4c7ed42013-08-31 13:44:33 +0800472 union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex)
Thomas Richter906dc182013-07-19 17:20:07 +0200473{
474 struct vxlan_rdst *rd;
475
476 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
477 if (rd)
478 return 0;
479
480 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
481 if (!rd)
482 return 0;
Cong Wange4c7ed42013-08-31 13:44:33 +0800483 rd->remote_ip = *ip;
Thomas Richter906dc182013-07-19 17:20:07 +0200484 rd->remote_port = port;
485 rd->remote_vni = vni;
486 rd->remote_ifindex = ifindex;
487 return 1;
488}
489
David Stevens66817122013-03-15 04:35:51 +0000490/* Add/update destinations for multicast */
491static int vxlan_fdb_append(struct vxlan_fdb *f,
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200492 union vxlan_addr *ip, __be16 port, __u32 vni,
493 __u32 ifindex, struct vxlan_rdst **rdp)
David Stevens66817122013-03-15 04:35:51 +0000494{
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700495 struct vxlan_rdst *rd;
David Stevens66817122013-03-15 04:35:51 +0000496
Mike Rapoporta5e7c102013-06-25 16:01:52 +0300497 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
498 if (rd)
499 return 0;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700500
David Stevens66817122013-03-15 04:35:51 +0000501 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
502 if (rd == NULL)
503 return -ENOBUFS;
Cong Wange4c7ed42013-08-31 13:44:33 +0800504 rd->remote_ip = *ip;
David Stevens66817122013-03-15 04:35:51 +0000505 rd->remote_port = port;
506 rd->remote_vni = vni;
507 rd->remote_ifindex = ifindex;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700508
509 list_add_tail_rcu(&rd->list, &f->remotes);
510
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200511 *rdp = rd;
David Stevens66817122013-03-15 04:35:51 +0000512 return 1;
513}
514
Tom Herbertdfd86452015-01-12 17:00:38 -0800515static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
516 unsigned int off,
517 struct vxlanhdr *vh, size_t hdrlen,
Tom Herbert0ace2ca2015-02-10 16:30:32 -0800518 u32 data, struct gro_remcsum *grc,
519 bool nopartial)
Tom Herbertdfd86452015-01-12 17:00:38 -0800520{
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700521 size_t start, offset;
Tom Herbertdfd86452015-01-12 17:00:38 -0800522
523 if (skb->remcsum_offload)
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700524 return vh;
Tom Herbertdfd86452015-01-12 17:00:38 -0800525
526 if (!NAPI_GRO_CB(skb)->csum_valid)
527 return NULL;
528
529 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
530 offset = start + ((data & VXLAN_RCO_UDP) ?
531 offsetof(struct udphdr, check) :
532 offsetof(struct tcphdr, check));
533
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700534 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
535 start, offset, grc, nopartial);
Tom Herbertdfd86452015-01-12 17:00:38 -0800536
537 skb->remcsum_offload = 1;
538
539 return vh;
540}
541
Tom Herberta2b12f32015-01-12 17:00:37 -0800542static struct sk_buff **vxlan_gro_receive(struct sk_buff **head,
543 struct sk_buff *skb,
544 struct udp_offload *uoff)
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200545{
546 struct sk_buff *p, **pp = NULL;
547 struct vxlanhdr *vh, *vh2;
Jesse Gross9b174d82014-12-30 19:10:15 -0800548 unsigned int hlen, off_vx;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200549 int flush = 1;
Tom Herbertdfd86452015-01-12 17:00:38 -0800550 struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock,
551 udp_offloads);
552 u32 flags;
Tom Herbert26c4f7d2015-02-10 16:30:27 -0800553 struct gro_remcsum grc;
554
555 skb_gro_remcsum_init(&grc);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200556
557 off_vx = skb_gro_offset(skb);
558 hlen = off_vx + sizeof(*vh);
559 vh = skb_gro_header_fast(skb, off_vx);
560 if (skb_gro_header_hard(skb, hlen)) {
561 vh = skb_gro_header_slow(skb, hlen, off_vx);
562 if (unlikely(!vh))
563 goto out;
564 }
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200565
Tom Herbertdfd86452015-01-12 17:00:38 -0800566 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
567
568 flags = ntohl(vh->vx_flags);
569
570 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
571 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
Tom Herbert0ace2ca2015-02-10 16:30:32 -0800572 ntohl(vh->vx_vni), &grc,
573 !!(vs->flags &
574 VXLAN_F_REMCSUM_NOPARTIAL));
Tom Herbertdfd86452015-01-12 17:00:38 -0800575
576 if (!vh)
577 goto out;
578 }
579
Tom Herbertb7fe10e2015-08-19 17:07:32 -0700580 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
581
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200582 flush = 0;
583
584 for (p = *head; p; p = p->next) {
585 if (!NAPI_GRO_CB(p)->same_flow)
586 continue;
587
588 vh2 = (struct vxlanhdr *)(p->data + off_vx);
Thomas Graf35114942015-01-15 03:53:55 +0100589 if (vh->vx_flags != vh2->vx_flags ||
590 vh->vx_vni != vh2->vx_vni) {
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200591 NAPI_GRO_CB(p)->same_flow = 0;
592 continue;
593 }
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200594 }
595
Jesse Gross9b174d82014-12-30 19:10:15 -0800596 pp = eth_gro_receive(head, skb);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200597
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200598out:
Tom Herbert26c4f7d2015-02-10 16:30:27 -0800599 skb_gro_remcsum_cleanup(skb, &grc);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200600 NAPI_GRO_CB(skb)->flush |= flush;
601
602 return pp;
603}
604
Tom Herberta2b12f32015-01-12 17:00:37 -0800605static int vxlan_gro_complete(struct sk_buff *skb, int nhoff,
606 struct udp_offload *uoff)
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200607{
Jesse Grosscfdf1e12014-11-10 11:45:13 -0800608 udp_tunnel_gro_complete(skb, nhoff);
609
Jesse Gross9b174d82014-12-30 19:10:15 -0800610 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200611}
612
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700613/* Notify netdevs that UDP port started listening */
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200614static void vxlan_notify_add_rx_port(struct vxlan_sock *vs)
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700615{
616 struct net_device *dev;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200617 struct sock *sk = vs->sock->sk;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700618 struct net *net = sock_net(sk);
Jiri Benc705cc622015-08-20 13:56:28 +0200619 sa_family_t sa_family = vxlan_get_sk_family(vs);
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700620 __be16 port = inet_sk(sk)->inet_sport;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200621 int err;
622
623 if (sa_family == AF_INET) {
Hannes Frederic Sowa787d7ac2016-01-07 14:28:39 +0100624 err = udp_add_offload(net, &vs->udp_offloads);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200625 if (err)
626 pr_warn("vxlan: udp_add_offload failed with status %d\n", err);
627 }
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700628
629 rcu_read_lock();
630 for_each_netdev_rcu(net, dev) {
631 if (dev->netdev_ops->ndo_add_vxlan_port)
632 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
633 port);
634 }
635 rcu_read_unlock();
636}
637
638/* Notify netdevs that UDP port is no more listening */
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200639static void vxlan_notify_del_rx_port(struct vxlan_sock *vs)
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700640{
641 struct net_device *dev;
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200642 struct sock *sk = vs->sock->sk;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700643 struct net *net = sock_net(sk);
Jiri Benc705cc622015-08-20 13:56:28 +0200644 sa_family_t sa_family = vxlan_get_sk_family(vs);
Joseph Gasparakis35e42372013-09-13 07:34:13 -0700645 __be16 port = inet_sk(sk)->inet_sport;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700646
647 rcu_read_lock();
648 for_each_netdev_rcu(net, dev) {
649 if (dev->netdev_ops->ndo_del_vxlan_port)
650 dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family,
651 port);
652 }
653 rcu_read_unlock();
Or Gerlitzdc01e7d2014-01-20 13:59:21 +0200654
655 if (sa_family == AF_INET)
656 udp_del_offload(&vs->udp_offloads);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -0700657}
658
stephen hemmingerd3428942012-10-01 12:32:35 +0000659/* Add new entry to forwarding table -- assumes lock held */
660static int vxlan_fdb_create(struct vxlan_dev *vxlan,
Cong Wange4c7ed42013-08-31 13:44:33 +0800661 const u8 *mac, union vxlan_addr *ip,
David Stevens66817122013-03-15 04:35:51 +0000662 __u16 state, __u16 flags,
stephen hemminger73cf3312013-04-27 11:31:54 +0000663 __be16 port, __u32 vni, __u32 ifindex,
David Stevensae884082013-04-19 00:36:26 +0000664 __u8 ndm_flags)
stephen hemmingerd3428942012-10-01 12:32:35 +0000665{
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200666 struct vxlan_rdst *rd = NULL;
stephen hemmingerd3428942012-10-01 12:32:35 +0000667 struct vxlan_fdb *f;
668 int notify = 0;
669
Sridhar Samudrala014be2c2013-05-17 06:39:07 +0000670 f = __vxlan_find_mac(vxlan, mac);
stephen hemmingerd3428942012-10-01 12:32:35 +0000671 if (f) {
672 if (flags & NLM_F_EXCL) {
673 netdev_dbg(vxlan->dev,
674 "lost race to create %pM\n", mac);
675 return -EEXIST;
676 }
677 if (f->state != state) {
678 f->state = state;
679 f->updated = jiffies;
680 notify = 1;
681 }
David Stevensae884082013-04-19 00:36:26 +0000682 if (f->flags != ndm_flags) {
683 f->flags = ndm_flags;
684 f->updated = jiffies;
685 notify = 1;
686 }
Thomas Richter906dc182013-07-19 17:20:07 +0200687 if ((flags & NLM_F_REPLACE)) {
688 /* Only change unicasts */
689 if (!(is_multicast_ether_addr(f->eth_addr) ||
690 is_zero_ether_addr(f->eth_addr))) {
Li RongQing60840422015-04-22 15:49:10 +0800691 notify |= vxlan_fdb_replace(f, ip, port, vni,
Thomas Richter906dc182013-07-19 17:20:07 +0200692 ifindex);
Thomas Richter906dc182013-07-19 17:20:07 +0200693 } else
694 return -EOPNOTSUPP;
695 }
David Stevens66817122013-03-15 04:35:51 +0000696 if ((flags & NLM_F_APPEND) &&
Mike Rapoport58e4c762013-06-25 16:01:56 +0300697 (is_multicast_ether_addr(f->eth_addr) ||
698 is_zero_ether_addr(f->eth_addr))) {
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200699 int rc = vxlan_fdb_append(f, ip, port, vni, ifindex,
700 &rd);
David Stevens66817122013-03-15 04:35:51 +0000701
702 if (rc < 0)
703 return rc;
704 notify |= rc;
705 }
stephen hemmingerd3428942012-10-01 12:32:35 +0000706 } else {
707 if (!(flags & NLM_F_CREATE))
708 return -ENOENT;
709
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200710 if (vxlan->cfg.addrmax &&
711 vxlan->addrcnt >= vxlan->cfg.addrmax)
stephen hemmingerd3428942012-10-01 12:32:35 +0000712 return -ENOSPC;
713
Thomas Richter906dc182013-07-19 17:20:07 +0200714 /* Disallow replace to add a multicast entry */
715 if ((flags & NLM_F_REPLACE) &&
716 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
717 return -EOPNOTSUPP;
718
Cong Wange4c7ed42013-08-31 13:44:33 +0800719 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
stephen hemmingerd3428942012-10-01 12:32:35 +0000720 f = kmalloc(sizeof(*f), GFP_ATOMIC);
721 if (!f)
722 return -ENOMEM;
723
724 notify = 1;
stephen hemmingerd3428942012-10-01 12:32:35 +0000725 f->state = state;
David Stevensae884082013-04-19 00:36:26 +0000726 f->flags = ndm_flags;
stephen hemmingerd3428942012-10-01 12:32:35 +0000727 f->updated = f->used = jiffies;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700728 INIT_LIST_HEAD(&f->remotes);
stephen hemmingerd3428942012-10-01 12:32:35 +0000729 memcpy(f->eth_addr, mac, ETH_ALEN);
730
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200731 vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700732
stephen hemmingerd3428942012-10-01 12:32:35 +0000733 ++vxlan->addrcnt;
734 hlist_add_head_rcu(&f->hlist,
735 vxlan_fdb_head(vxlan, mac));
736 }
737
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200738 if (notify) {
739 if (rd == NULL)
740 rd = first_remote_rtnl(f);
741 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
742 }
stephen hemmingerd3428942012-10-01 12:32:35 +0000743
744 return 0;
745}
746
Wei Yongjun6706c822013-04-11 19:00:35 +0000747static void vxlan_fdb_free(struct rcu_head *head)
David Stevens66817122013-03-15 04:35:51 +0000748{
749 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700750 struct vxlan_rdst *rd, *nd;
David Stevens66817122013-03-15 04:35:51 +0000751
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700752 list_for_each_entry_safe(rd, nd, &f->remotes, list)
David Stevens66817122013-03-15 04:35:51 +0000753 kfree(rd);
David Stevens66817122013-03-15 04:35:51 +0000754 kfree(f);
755}
756
stephen hemmingerd3428942012-10-01 12:32:35 +0000757static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
758{
759 netdev_dbg(vxlan->dev,
760 "delete %pM\n", f->eth_addr);
761
762 --vxlan->addrcnt;
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200763 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
stephen hemmingerd3428942012-10-01 12:32:35 +0000764
765 hlist_del_rcu(&f->hlist);
David Stevens66817122013-03-15 04:35:51 +0000766 call_rcu(&f->rcu, vxlan_fdb_free);
stephen hemmingerd3428942012-10-01 12:32:35 +0000767}
768
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300769static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
Cong Wange4c7ed42013-08-31 13:44:33 +0800770 union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex)
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300771{
772 struct net *net = dev_net(vxlan->dev);
Cong Wange4c7ed42013-08-31 13:44:33 +0800773 int err;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300774
775 if (tb[NDA_DST]) {
Cong Wange4c7ed42013-08-31 13:44:33 +0800776 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
777 if (err)
778 return err;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300779 } else {
Cong Wange4c7ed42013-08-31 13:44:33 +0800780 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
781 if (remote->sa.sa_family == AF_INET) {
782 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
783 ip->sa.sa_family = AF_INET;
784#if IS_ENABLED(CONFIG_IPV6)
785 } else {
786 ip->sin6.sin6_addr = in6addr_any;
787 ip->sa.sa_family = AF_INET6;
788#endif
789 }
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300790 }
791
792 if (tb[NDA_PORT]) {
793 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
794 return -EINVAL;
795 *port = nla_get_be16(tb[NDA_PORT]);
796 } else {
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200797 *port = vxlan->cfg.dst_port;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300798 }
799
800 if (tb[NDA_VNI]) {
801 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
802 return -EINVAL;
803 *vni = nla_get_u32(tb[NDA_VNI]);
804 } else {
805 *vni = vxlan->default_dst.remote_vni;
806 }
807
808 if (tb[NDA_IFINDEX]) {
809 struct net_device *tdev;
810
811 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
812 return -EINVAL;
813 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
Ying Xue73763942014-01-15 10:23:41 +0800814 tdev = __dev_get_by_index(net, *ifindex);
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300815 if (!tdev)
816 return -EADDRNOTAVAIL;
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300817 } else {
818 *ifindex = 0;
819 }
820
821 return 0;
822}
823
stephen hemmingerd3428942012-10-01 12:32:35 +0000824/* Add static entry (via netlink) */
825static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
826 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100827 const unsigned char *addr, u16 vid, u16 flags)
stephen hemmingerd3428942012-10-01 12:32:35 +0000828{
829 struct vxlan_dev *vxlan = netdev_priv(dev);
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300830 /* struct net *net = dev_net(vxlan->dev); */
Cong Wange4c7ed42013-08-31 13:44:33 +0800831 union vxlan_addr ip;
stephen hemminger73cf3312013-04-27 11:31:54 +0000832 __be16 port;
833 u32 vni, ifindex;
stephen hemmingerd3428942012-10-01 12:32:35 +0000834 int err;
835
836 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
837 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
838 ndm->ndm_state);
839 return -EINVAL;
840 }
841
842 if (tb[NDA_DST] == NULL)
843 return -EINVAL;
844
Mike Rapoportf0b074b2013-06-25 16:01:53 +0300845 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
846 if (err)
847 return err;
David Stevens66817122013-03-15 04:35:51 +0000848
Mike Rapoport5933a7b2014-04-01 09:23:01 +0300849 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
850 return -EAFNOSUPPORT;
851
stephen hemmingerd3428942012-10-01 12:32:35 +0000852 spin_lock_bh(&vxlan->hash_lock);
Cong Wange4c7ed42013-08-31 13:44:33 +0800853 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
stephen hemminger73cf3312013-04-27 11:31:54 +0000854 port, vni, ifindex, ndm->ndm_flags);
stephen hemmingerd3428942012-10-01 12:32:35 +0000855 spin_unlock_bh(&vxlan->hash_lock);
856
857 return err;
858}
859
860/* Delete entry (via netlink) */
Vlad Yasevich1690be62013-02-13 12:00:18 +0000861static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
862 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100863 const unsigned char *addr, u16 vid)
stephen hemmingerd3428942012-10-01 12:32:35 +0000864{
865 struct vxlan_dev *vxlan = netdev_priv(dev);
866 struct vxlan_fdb *f;
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300867 struct vxlan_rdst *rd = NULL;
Cong Wange4c7ed42013-08-31 13:44:33 +0800868 union vxlan_addr ip;
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300869 __be16 port;
870 u32 vni, ifindex;
871 int err;
872
873 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex);
874 if (err)
875 return err;
876
877 err = -ENOENT;
stephen hemmingerd3428942012-10-01 12:32:35 +0000878
879 spin_lock_bh(&vxlan->hash_lock);
880 f = vxlan_find_mac(vxlan, addr);
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300881 if (!f)
882 goto out;
883
Cong Wange4c7ed42013-08-31 13:44:33 +0800884 if (!vxlan_addr_any(&ip)) {
885 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300886 if (!rd)
887 goto out;
stephen hemmingerd3428942012-10-01 12:32:35 +0000888 }
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300889
890 err = 0;
891
892 /* remove a destination if it's not the only one on the list,
893 * otherwise destroy the fdb entry
894 */
895 if (rd && !list_is_singular(&f->remotes)) {
896 list_del_rcu(&rd->list);
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200897 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
Wei Yongjundcdc7a52013-08-17 07:32:09 +0800898 kfree_rcu(rd, rcu);
Mike Rapoportbc7892b2013-06-25 16:01:54 +0300899 goto out;
900 }
901
902 vxlan_fdb_destroy(vxlan, f);
903
904out:
stephen hemmingerd3428942012-10-01 12:32:35 +0000905 spin_unlock_bh(&vxlan->hash_lock);
906
907 return err;
908}
909
910/* Dump forwarding table */
911static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -0400912 struct net_device *dev,
913 struct net_device *filter_dev, int idx)
stephen hemmingerd3428942012-10-01 12:32:35 +0000914{
915 struct vxlan_dev *vxlan = netdev_priv(dev);
916 unsigned int h;
917
918 for (h = 0; h < FDB_HASH_SIZE; ++h) {
919 struct vxlan_fdb *f;
stephen hemmingerd3428942012-10-01 12:32:35 +0000920 int err;
921
Sasha Levinb67bfe02013-02-27 17:06:00 -0800922 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
David Stevens66817122013-03-15 04:35:51 +0000923 struct vxlan_rdst *rd;
stephen hemmingerd3428942012-10-01 12:32:35 +0000924
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700925 list_for_each_entry_rcu(rd, &f->remotes, list) {
Atzm Watanabe07a51cd2015-08-10 23:39:09 +0900926 if (idx < cb->args[0])
927 goto skip;
928
David Stevens66817122013-03-15 04:35:51 +0000929 err = vxlan_fdb_info(skb, vxlan, f,
930 NETLINK_CB(cb->skb).portid,
931 cb->nlh->nlmsg_seq,
932 RTM_NEWNEIGH,
933 NLM_F_MULTI, rd);
934 if (err < 0)
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700935 goto out;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700936skip:
Atzm Watanabe07a51cd2015-08-10 23:39:09 +0900937 ++idx;
938 }
stephen hemmingerd3428942012-10-01 12:32:35 +0000939 }
940 }
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700941out:
stephen hemmingerd3428942012-10-01 12:32:35 +0000942 return idx;
943}
944
945/* Watch incoming packets to learn mapping between Ethernet address
946 * and Tunnel endpoint.
Simon Hormanc4b49512015-04-02 11:17:58 +0900947 * Return true if packet is bogus and should be dropped.
stephen hemmingerd3428942012-10-01 12:32:35 +0000948 */
stephen hemminger26a41ae62013-06-17 12:09:58 -0700949static bool vxlan_snoop(struct net_device *dev,
Cong Wange4c7ed42013-08-31 13:44:33 +0800950 union vxlan_addr *src_ip, const u8 *src_mac)
stephen hemmingerd3428942012-10-01 12:32:35 +0000951{
952 struct vxlan_dev *vxlan = netdev_priv(dev);
953 struct vxlan_fdb *f;
stephen hemmingerd3428942012-10-01 12:32:35 +0000954
955 f = vxlan_find_mac(vxlan, src_mac);
956 if (likely(f)) {
stephen hemminger5ca54612013-08-04 17:17:39 -0700957 struct vxlan_rdst *rdst = first_remote_rcu(f);
Stephen Hemminger3e61aa82013-06-17 14:16:12 -0700958
Cong Wange4c7ed42013-08-31 13:44:33 +0800959 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip)))
stephen hemminger26a41ae62013-06-17 12:09:58 -0700960 return false;
961
962 /* Don't migrate static entries, drop packets */
stephen hemmingereb064c32013-06-18 14:27:01 -0700963 if (f->state & NUD_NOARP)
stephen hemminger26a41ae62013-06-17 12:09:58 -0700964 return true;
stephen hemmingerd3428942012-10-01 12:32:35 +0000965
966 if (net_ratelimit())
967 netdev_info(dev,
Cong Wange4c7ed42013-08-31 13:44:33 +0800968 "%pM migrated from %pIS to %pIS\n",
Rasmus Villemoesa4870f72015-02-07 03:17:31 +0100969 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
stephen hemmingerd3428942012-10-01 12:32:35 +0000970
Cong Wange4c7ed42013-08-31 13:44:33 +0800971 rdst->remote_ip = *src_ip;
stephen hemmingerd3428942012-10-01 12:32:35 +0000972 f->updated = jiffies;
Nicolas Dichtel9e4b93f2014-04-22 15:01:30 +0200973 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
stephen hemmingerd3428942012-10-01 12:32:35 +0000974 } else {
975 /* learned new entry */
976 spin_lock(&vxlan->hash_lock);
stephen hemminger3bf74b12013-06-17 12:09:57 -0700977
978 /* close off race between vxlan_flush and incoming packets */
979 if (netif_running(dev))
980 vxlan_fdb_create(vxlan, src_mac, src_ip,
981 NUD_REACHABLE,
982 NLM_F_EXCL|NLM_F_CREATE,
Thomas Graf0dfbdf42015-07-21 10:44:02 +0200983 vxlan->cfg.dst_port,
stephen hemminger3bf74b12013-06-17 12:09:57 -0700984 vxlan->default_dst.remote_vni,
985 0, NTF_SELF);
stephen hemmingerd3428942012-10-01 12:32:35 +0000986 spin_unlock(&vxlan->hash_lock);
987 }
stephen hemminger26a41ae62013-06-17 12:09:58 -0700988
989 return false;
stephen hemmingerd3428942012-10-01 12:32:35 +0000990}
991
stephen hemmingerd3428942012-10-01 12:32:35 +0000992/* See if multicast group is already in use by other ID */
Gao feng95ab0992013-12-10 16:37:33 +0800993static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
stephen hemmingerd3428942012-10-01 12:32:35 +0000994{
stephen hemminger553675f2013-05-16 11:35:20 +0000995 struct vxlan_dev *vxlan;
Jiri Bencb1be00a2015-09-24 13:50:02 +0200996 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
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 */
Jiri Bencb1be00a2015-09-24 13:50:02 +02001001 if (family == AF_INET && dev->vn4_sock &&
1002 atomic_read(&dev->vn4_sock->refcnt) == 1)
Gao feng95ab0992013-12-10 16:37:33 +08001003 return false;
Jiri Bencb1be00a2015-09-24 13:50:02 +02001004#if IS_ENABLED(CONFIG_IPV6)
1005 if (family == AF_INET6 && dev->vn6_sock &&
1006 atomic_read(&dev->vn6_sock->refcnt) == 1)
1007 return false;
1008#endif
Gao feng95ab0992013-12-10 16:37:33 +08001009
stephen hemminger553675f2013-05-16 11:35:20 +00001010 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
Gao feng95ab0992013-12-10 16:37:33 +08001011 if (!netif_running(vxlan->dev) || vxlan == dev)
stephen hemminger553675f2013-05-16 11:35:20 +00001012 continue;
stephen hemmingerd3428942012-10-01 12:32:35 +00001013
Jiri Bencb1be00a2015-09-24 13:50:02 +02001014 if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock)
Gao feng95ab0992013-12-10 16:37:33 +08001015 continue;
Jiri Bencb1be00a2015-09-24 13:50:02 +02001016#if IS_ENABLED(CONFIG_IPV6)
1017 if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock)
1018 continue;
1019#endif
Gao feng95ab0992013-12-10 16:37:33 +08001020
1021 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1022 &dev->default_dst.remote_ip))
1023 continue;
1024
1025 if (vxlan->default_dst.remote_ifindex !=
1026 dev->default_dst.remote_ifindex)
1027 continue;
1028
1029 return true;
stephen hemminger553675f2013-05-16 11:35:20 +00001030 }
stephen hemmingerd3428942012-10-01 12:32:35 +00001031
1032 return false;
1033}
1034
Jiri Bencb1be00a2015-09-24 13:50:02 +02001035static void __vxlan_sock_release(struct vxlan_sock *vs)
stephen hemmingerd3428942012-10-01 12:32:35 +00001036{
Jiri Bencb1be00a2015-09-24 13:50:02 +02001037 struct vxlan_net *vn;
Pravin B Shelar012a5722013-08-19 11:23:07 -07001038
Jiri Bencb1be00a2015-09-24 13:50:02 +02001039 if (!vs)
1040 return;
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001041 if (!atomic_dec_and_test(&vs->refcnt))
1042 return;
1043
Jiri Bencb1be00a2015-09-24 13:50:02 +02001044 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
Stephen Hemminger1c51a912013-06-17 14:16:11 -07001045 spin_lock(&vn->sock_lock);
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001046 hlist_del_rcu(&vs->hlist);
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02001047 vxlan_notify_del_rx_port(vs);
Stephen Hemminger1c51a912013-06-17 14:16:11 -07001048 spin_unlock(&vn->sock_lock);
1049
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001050 queue_work(vxlan_wq, &vs->del_work);
1051}
1052
Jiri Bencb1be00a2015-09-24 13:50:02 +02001053static void vxlan_sock_release(struct vxlan_dev *vxlan)
1054{
1055 __vxlan_sock_release(vxlan->vn4_sock);
1056#if IS_ENABLED(CONFIG_IPV6)
1057 __vxlan_sock_release(vxlan->vn6_sock);
1058#endif
1059}
1060
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001061/* Update multicast group membership when first VNI on
Simon Hormanc4b49512015-04-02 11:17:58 +09001062 * multicast address is brought up
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001063 */
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001064static int vxlan_igmp_join(struct vxlan_dev *vxlan)
Stephen Hemminger7c47ced2013-06-17 14:16:10 -07001065{
Jiri Bencb1be00a2015-09-24 13:50:02 +02001066 struct sock *sk;
Cong Wange4c7ed42013-08-31 13:44:33 +08001067 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1068 int ifindex = vxlan->default_dst.remote_ifindex;
Marcelo Ricardo Leitner149d7542015-03-20 10:26:21 -03001069 int ret = -EINVAL;
stephen hemmingerd3428942012-10-01 12:32:35 +00001070
Cong Wange4c7ed42013-08-31 13:44:33 +08001071 if (ip->sa.sa_family == AF_INET) {
1072 struct ip_mreqn mreq = {
1073 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1074 .imr_ifindex = ifindex,
1075 };
1076
Jiri Bencb1be00a2015-09-24 13:50:02 +02001077 sk = vxlan->vn4_sock->sock->sk;
1078 lock_sock(sk);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001079 ret = ip_mc_join_group(sk, &mreq);
Jiri Bencb1be00a2015-09-24 13:50:02 +02001080 release_sock(sk);
Cong Wange4c7ed42013-08-31 13:44:33 +08001081#if IS_ENABLED(CONFIG_IPV6)
1082 } else {
Jiri Bencb1be00a2015-09-24 13:50:02 +02001083 sk = vxlan->vn6_sock->sock->sk;
1084 lock_sock(sk);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001085 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1086 &ip->sin6.sin6_addr);
Jiri Bencb1be00a2015-09-24 13:50:02 +02001087 release_sock(sk);
Cong Wange4c7ed42013-08-31 13:44:33 +08001088#endif
1089 }
stephen hemminger3fc2de22013-07-18 08:40:15 -07001090
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001091 return ret;
stephen hemminger3fc2de22013-07-18 08:40:15 -07001092}
1093
1094/* Inverse of vxlan_igmp_join when last VNI is brought down */
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001095static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
stephen hemminger3fc2de22013-07-18 08:40:15 -07001096{
Jiri Bencb1be00a2015-09-24 13:50:02 +02001097 struct sock *sk;
Cong Wange4c7ed42013-08-31 13:44:33 +08001098 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1099 int ifindex = vxlan->default_dst.remote_ifindex;
Marcelo Ricardo Leitner149d7542015-03-20 10:26:21 -03001100 int ret = -EINVAL;
stephen hemminger3fc2de22013-07-18 08:40:15 -07001101
Cong Wange4c7ed42013-08-31 13:44:33 +08001102 if (ip->sa.sa_family == AF_INET) {
1103 struct ip_mreqn mreq = {
1104 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1105 .imr_ifindex = ifindex,
1106 };
1107
Jiri Bencb1be00a2015-09-24 13:50:02 +02001108 sk = vxlan->vn4_sock->sock->sk;
1109 lock_sock(sk);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001110 ret = ip_mc_leave_group(sk, &mreq);
Jiri Bencb1be00a2015-09-24 13:50:02 +02001111 release_sock(sk);
Cong Wange4c7ed42013-08-31 13:44:33 +08001112#if IS_ENABLED(CONFIG_IPV6)
1113 } else {
Jiri Bencb1be00a2015-09-24 13:50:02 +02001114 sk = vxlan->vn6_sock->sock->sk;
1115 lock_sock(sk);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001116 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1117 &ip->sin6.sin6_addr);
Jiri Bencb1be00a2015-09-24 13:50:02 +02001118 release_sock(sk);
Cong Wange4c7ed42013-08-31 13:44:33 +08001119#endif
1120 }
stephen hemmingerd3428942012-10-01 12:32:35 +00001121
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03001122 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00001123}
1124
Tom Herbertdfd86452015-01-12 17:00:38 -08001125static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh,
Tom Herbert0ace2ca2015-02-10 16:30:32 -08001126 size_t hdrlen, u32 data, bool nopartial)
Tom Herbertdfd86452015-01-12 17:00:38 -08001127{
1128 size_t start, offset, plen;
Tom Herbertdfd86452015-01-12 17:00:38 -08001129
Tom Herbertb7fe10e2015-08-19 17:07:32 -07001130 if (skb->remcsum_offload)
1131 return vh;
1132
Tom Herbertdfd86452015-01-12 17:00:38 -08001133 start = (data & VXLAN_RCO_MASK) << VXLAN_RCO_SHIFT;
1134 offset = start + ((data & VXLAN_RCO_UDP) ?
1135 offsetof(struct udphdr, check) :
1136 offsetof(struct tcphdr, check));
1137
1138 plen = hdrlen + offset + sizeof(u16);
1139
1140 if (!pskb_may_pull(skb, plen))
1141 return NULL;
1142
1143 vh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
1144
Tom Herbert0ace2ca2015-02-10 16:30:32 -08001145 skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset,
1146 nopartial);
Tom Herbertdfd86452015-01-12 17:00:38 -08001147
1148 return vh;
1149}
1150
Thomas Graf614732e2015-07-21 10:44:06 +02001151static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb,
1152 struct vxlan_metadata *md, u32 vni,
1153 struct metadata_dst *tun_dst)
1154{
1155 struct iphdr *oip = NULL;
1156 struct ipv6hdr *oip6 = NULL;
1157 struct vxlan_dev *vxlan;
1158 struct pcpu_sw_netstats *stats;
1159 union vxlan_addr saddr;
1160 int err = 0;
Thomas Graf614732e2015-07-21 10:44:06 +02001161
1162 /* For flow based devices, map all packets to VNI 0 */
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07001163 if (vs->flags & VXLAN_F_COLLECT_METADATA)
Thomas Graf614732e2015-07-21 10:44:06 +02001164 vni = 0;
1165
1166 /* Is this VNI defined? */
1167 vxlan = vxlan_vs_find_vni(vs, vni);
1168 if (!vxlan)
1169 goto drop;
1170
Thomas Graf614732e2015-07-21 10:44:06 +02001171 skb_reset_mac_header(skb);
1172 skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev)));
1173 skb->protocol = eth_type_trans(skb, vxlan->dev);
1174 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1175
1176 /* Ignore packet loops (and multicast echo) */
1177 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1178 goto drop;
1179
Jiri Bencce212d02015-12-07 16:29:08 +01001180 /* Get data from the outer IP header */
1181 if (vxlan_get_sk_family(vs) == AF_INET) {
Thomas Graf614732e2015-07-21 10:44:06 +02001182 oip = ip_hdr(skb);
1183 saddr.sin.sin_addr.s_addr = oip->saddr;
1184 saddr.sa.sa_family = AF_INET;
1185#if IS_ENABLED(CONFIG_IPV6)
1186 } else {
1187 oip6 = ipv6_hdr(skb);
1188 saddr.sin6.sin6_addr = oip6->saddr;
1189 saddr.sa.sa_family = AF_INET6;
1190#endif
1191 }
1192
1193 if (tun_dst) {
1194 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1195 tun_dst = NULL;
1196 }
1197
1198 if ((vxlan->flags & VXLAN_F_LEARN) &&
1199 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source))
1200 goto drop;
1201
1202 skb_reset_network_header(skb);
1203 /* In flow-based mode, GBP is carried in dst_metadata */
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07001204 if (!(vs->flags & VXLAN_F_COLLECT_METADATA))
Thomas Graf614732e2015-07-21 10:44:06 +02001205 skb->mark = md->gbp;
1206
1207 if (oip6)
1208 err = IP6_ECN_decapsulate(oip6, skb);
1209 if (oip)
1210 err = IP_ECN_decapsulate(oip, skb);
1211
1212 if (unlikely(err)) {
1213 if (log_ecn_error) {
1214 if (oip6)
1215 net_info_ratelimited("non-ECT from %pI6\n",
1216 &oip6->saddr);
1217 if (oip)
1218 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1219 &oip->saddr, oip->tos);
1220 }
1221 if (err > 1) {
1222 ++vxlan->dev->stats.rx_frame_errors;
1223 ++vxlan->dev->stats.rx_errors;
1224 goto drop;
1225 }
1226 }
1227
1228 stats = this_cpu_ptr(vxlan->dev->tstats);
1229 u64_stats_update_begin(&stats->syncp);
1230 stats->rx_packets++;
1231 stats->rx_bytes += skb->len;
1232 u64_stats_update_end(&stats->syncp);
1233
Tom Herbert58ce31c2015-08-19 17:07:33 -07001234 gro_cells_receive(&vxlan->gro_cells, skb);
Thomas Graf614732e2015-07-21 10:44:06 +02001235
1236 return;
1237drop:
1238 if (tun_dst)
1239 dst_release((struct dst_entry *)tun_dst);
1240
1241 /* Consume bad packet */
1242 kfree_skb(skb);
1243}
1244
stephen hemmingerd3428942012-10-01 12:32:35 +00001245/* Callback from net/ipv4/udp.c to receive packets */
1246static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb)
1247{
Thomas Grafee122c72015-07-21 10:43:58 +02001248 struct metadata_dst *tun_dst = NULL;
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001249 struct vxlan_sock *vs;
stephen hemmingerd3428942012-10-01 12:32:35 +00001250 struct vxlanhdr *vxh;
Tom Herbert3bf39472015-01-08 12:31:18 -08001251 u32 flags, vni;
Thomas Grafee122c72015-07-21 10:43:58 +02001252 struct vxlan_metadata _md;
1253 struct vxlan_metadata *md = &_md;
stephen hemmingerd3428942012-10-01 12:32:35 +00001254
stephen hemmingerd3428942012-10-01 12:32:35 +00001255 /* Need Vxlan and inner Ethernet header to be present */
Pravin B Shelar7ce04752013-08-19 11:22:54 -07001256 if (!pskb_may_pull(skb, VXLAN_HLEN))
stephen hemmingerd3428942012-10-01 12:32:35 +00001257 goto error;
1258
Pravin B Shelar7ce04752013-08-19 11:22:54 -07001259 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
Tom Herbert3bf39472015-01-08 12:31:18 -08001260 flags = ntohl(vxh->vx_flags);
1261 vni = ntohl(vxh->vx_vni);
1262
1263 if (flags & VXLAN_HF_VNI) {
1264 flags &= ~VXLAN_HF_VNI;
1265 } else {
1266 /* VNI flag always required to be set */
1267 goto bad_flags;
stephen hemmingerd3428942012-10-01 12:32:35 +00001268 }
1269
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001270 if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB)))
stephen hemmingerd3428942012-10-01 12:32:35 +00001271 goto drop;
Tom Herbertdfd86452015-01-12 17:00:38 -08001272 vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1);
stephen hemmingerd3428942012-10-01 12:32:35 +00001273
Pravin B Shelar559835ea2013-09-24 10:25:40 -07001274 vs = rcu_dereference_sk_user_data(sk);
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001275 if (!vs)
stephen hemmingerd3428942012-10-01 12:32:35 +00001276 goto drop;
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001277
Tom Herbertdfd86452015-01-12 17:00:38 -08001278 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
Tom Herbert0ace2ca2015-02-10 16:30:32 -08001279 vxh = vxlan_remcsum(skb, vxh, sizeof(struct vxlanhdr), vni,
1280 !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL));
Tom Herbertdfd86452015-01-12 17:00:38 -08001281 if (!vxh)
1282 goto drop;
1283
1284 flags &= ~VXLAN_HF_RCO;
Alexey Kodanev40fb70f2015-03-13 19:13:53 +03001285 vni &= VXLAN_VNI_MASK;
Tom Herbertdfd86452015-01-12 17:00:38 -08001286 }
1287
Thomas Grafee122c72015-07-21 10:43:58 +02001288 if (vxlan_collect_metadata(vs)) {
Pravin B Shelarc29a70d2015-08-26 23:46:50 -07001289 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1290 cpu_to_be64(vni >> 8), sizeof(*md));
1291
Thomas Grafee122c72015-07-21 10:43:58 +02001292 if (!tun_dst)
1293 goto drop;
1294
Geert Uytterhoeven0f1b7352015-09-04 12:49:32 +02001295 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
Thomas Grafee122c72015-07-21 10:43:58 +02001296 } else {
1297 memset(md, 0, sizeof(*md));
1298 }
1299
Thomas Graf35114942015-01-15 03:53:55 +01001300 /* For backwards compatibility, only allow reserved fields to be
1301 * used by VXLAN extensions if explicitly requested.
1302 */
1303 if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) {
1304 struct vxlanhdr_gbp *gbp;
1305
1306 gbp = (struct vxlanhdr_gbp *)vxh;
Thomas Grafee122c72015-07-21 10:43:58 +02001307 md->gbp = ntohs(gbp->policy_id);
1308
1309 if (tun_dst)
Geert Uytterhoeven0f1b7352015-09-04 12:49:32 +02001310 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
Thomas Graf35114942015-01-15 03:53:55 +01001311
1312 if (gbp->dont_learn)
Thomas Grafee122c72015-07-21 10:43:58 +02001313 md->gbp |= VXLAN_GBP_DONT_LEARN;
Thomas Graf35114942015-01-15 03:53:55 +01001314
1315 if (gbp->policy_applied)
Thomas Grafee122c72015-07-21 10:43:58 +02001316 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
Thomas Graf35114942015-01-15 03:53:55 +01001317
1318 flags &= ~VXLAN_GBP_USED_BITS;
1319 }
1320
Alexey Kodanev40fb70f2015-03-13 19:13:53 +03001321 if (flags || vni & ~VXLAN_VNI_MASK) {
Tom Herbert3bf39472015-01-08 12:31:18 -08001322 /* If there are any unprocessed flags remaining treat
1323 * this as a malformed packet. This behavior diverges from
1324 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1325 * in reserved fields are to be ignored. The approach here
Simon Hormanc4b49512015-04-02 11:17:58 +09001326 * maintains compatibility with previous stack code, and also
Tom Herbert3bf39472015-01-08 12:31:18 -08001327 * is more robust and provides a little more security in
1328 * adding extensions to VXLAN.
1329 */
1330
1331 goto bad_flags;
1332 }
1333
Thomas Graf614732e2015-07-21 10:44:06 +02001334 vxlan_rcv(vs, skb, md, vni >> 8, tun_dst);
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001335 return 0;
1336
1337drop:
1338 /* Consume bad packet */
1339 kfree_skb(skb);
1340 return 0;
1341
Tom Herbert3bf39472015-01-08 12:31:18 -08001342bad_flags:
1343 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1344 ntohl(vxh->vx_flags), ntohl(vxh->vx_vni));
1345
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001346error:
Thomas Grafee122c72015-07-21 10:43:58 +02001347 if (tun_dst)
1348 dst_release((struct dst_entry *)tun_dst);
1349
Pravin B Shelar5cfccc52013-08-19 11:23:02 -07001350 /* Return non vxlan pkt */
1351 return 1;
1352}
1353
David Stevense4f67ad2012-11-20 02:50:14 +00001354static int arp_reduce(struct net_device *dev, struct sk_buff *skb)
1355{
1356 struct vxlan_dev *vxlan = netdev_priv(dev);
1357 struct arphdr *parp;
1358 u8 *arpptr, *sha;
1359 __be32 sip, tip;
1360 struct neighbour *n;
1361
1362 if (dev->flags & IFF_NOARP)
1363 goto out;
1364
1365 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1366 dev->stats.tx_dropped++;
1367 goto out;
1368 }
1369 parp = arp_hdr(skb);
1370
1371 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1372 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1373 parp->ar_pro != htons(ETH_P_IP) ||
1374 parp->ar_op != htons(ARPOP_REQUEST) ||
1375 parp->ar_hln != dev->addr_len ||
1376 parp->ar_pln != 4)
1377 goto out;
1378 arpptr = (u8 *)parp + sizeof(struct arphdr);
1379 sha = arpptr;
1380 arpptr += dev->addr_len; /* sha */
1381 memcpy(&sip, arpptr, sizeof(sip));
1382 arpptr += sizeof(sip);
1383 arpptr += dev->addr_len; /* tha */
1384 memcpy(&tip, arpptr, sizeof(tip));
1385
1386 if (ipv4_is_loopback(tip) ||
1387 ipv4_is_multicast(tip))
1388 goto out;
1389
1390 n = neigh_lookup(&arp_tbl, &tip, dev);
1391
1392 if (n) {
David Stevense4f67ad2012-11-20 02:50:14 +00001393 struct vxlan_fdb *f;
1394 struct sk_buff *reply;
1395
1396 if (!(n->nud_state & NUD_CONNECTED)) {
1397 neigh_release(n);
1398 goto out;
1399 }
1400
1401 f = vxlan_find_mac(vxlan, n->ha);
Cong Wange4c7ed42013-08-31 13:44:33 +08001402 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
David Stevense4f67ad2012-11-20 02:50:14 +00001403 /* bridge-local neighbor */
1404 neigh_release(n);
1405 goto out;
1406 }
1407
1408 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1409 n->ha, sha);
1410
1411 neigh_release(n);
1412
David Stevens73461352014-03-18 12:32:29 -04001413 if (reply == NULL)
1414 goto out;
1415
David Stevense4f67ad2012-11-20 02:50:14 +00001416 skb_reset_mac_header(reply);
1417 __skb_pull(reply, skb_network_offset(reply));
1418 reply->ip_summed = CHECKSUM_UNNECESSARY;
1419 reply->pkt_type = PACKET_HOST;
1420
1421 if (netif_rx_ni(reply) == NET_RX_DROP)
1422 dev->stats.rx_dropped++;
Cong Wange4c7ed42013-08-31 13:44:33 +08001423 } else if (vxlan->flags & VXLAN_F_L3MISS) {
1424 union vxlan_addr ipa = {
1425 .sin.sin_addr.s_addr = tip,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001426 .sin.sin_family = AF_INET,
Cong Wange4c7ed42013-08-31 13:44:33 +08001427 };
1428
1429 vxlan_ip_miss(dev, &ipa);
1430 }
David Stevense4f67ad2012-11-20 02:50:14 +00001431out:
1432 consume_skb(skb);
1433 return NETDEV_TX_OK;
1434}
1435
Cong Wangf564f452013-08-31 13:44:36 +08001436#if IS_ENABLED(CONFIG_IPV6)
David Stevens4b29dba2014-03-24 10:39:58 -04001437static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1438 struct neighbour *n, bool isrouter)
1439{
1440 struct net_device *dev = request->dev;
1441 struct sk_buff *reply;
1442 struct nd_msg *ns, *na;
1443 struct ipv6hdr *pip6;
1444 u8 *daddr;
1445 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1446 int ns_olen;
1447 int i, len;
1448
1449 if (dev == NULL)
1450 return NULL;
1451
1452 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1453 sizeof(*na) + na_olen + dev->needed_tailroom;
1454 reply = alloc_skb(len, GFP_ATOMIC);
1455 if (reply == NULL)
1456 return NULL;
1457
1458 reply->protocol = htons(ETH_P_IPV6);
1459 reply->dev = dev;
1460 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1461 skb_push(reply, sizeof(struct ethhdr));
1462 skb_set_mac_header(reply, 0);
1463
1464 ns = (struct nd_msg *)skb_transport_header(request);
1465
1466 daddr = eth_hdr(request)->h_source;
1467 ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns);
1468 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1469 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1470 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1471 break;
1472 }
1473 }
1474
1475 /* Ethernet header */
1476 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1477 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1478 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1479 reply->protocol = htons(ETH_P_IPV6);
1480
1481 skb_pull(reply, sizeof(struct ethhdr));
1482 skb_set_network_header(reply, 0);
1483 skb_put(reply, sizeof(struct ipv6hdr));
1484
1485 /* IPv6 header */
1486
1487 pip6 = ipv6_hdr(reply);
1488 memset(pip6, 0, sizeof(struct ipv6hdr));
1489 pip6->version = 6;
1490 pip6->priority = ipv6_hdr(request)->priority;
1491 pip6->nexthdr = IPPROTO_ICMPV6;
1492 pip6->hop_limit = 255;
1493 pip6->daddr = ipv6_hdr(request)->saddr;
1494 pip6->saddr = *(struct in6_addr *)n->primary_key;
1495
1496 skb_pull(reply, sizeof(struct ipv6hdr));
1497 skb_set_transport_header(reply, 0);
1498
1499 na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen);
1500
1501 /* Neighbor Advertisement */
1502 memset(na, 0, sizeof(*na)+na_olen);
1503 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1504 na->icmph.icmp6_router = isrouter;
1505 na->icmph.icmp6_override = 1;
1506 na->icmph.icmp6_solicited = 1;
1507 na->target = ns->target;
1508 ether_addr_copy(&na->opt[2], n->ha);
1509 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1510 na->opt[1] = na_olen >> 3;
1511
1512 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1513 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1514 csum_partial(na, sizeof(*na)+na_olen, 0));
1515
1516 pip6->payload_len = htons(sizeof(*na)+na_olen);
1517
1518 skb_push(reply, sizeof(struct ipv6hdr));
1519
1520 reply->ip_summed = CHECKSUM_UNNECESSARY;
1521
1522 return reply;
1523}
1524
Cong Wangf564f452013-08-31 13:44:36 +08001525static int neigh_reduce(struct net_device *dev, struct sk_buff *skb)
1526{
1527 struct vxlan_dev *vxlan = netdev_priv(dev);
David Stevens4b29dba2014-03-24 10:39:58 -04001528 struct nd_msg *msg;
Cong Wangf564f452013-08-31 13:44:36 +08001529 const struct ipv6hdr *iphdr;
1530 const struct in6_addr *saddr, *daddr;
David Stevens4b29dba2014-03-24 10:39:58 -04001531 struct neighbour *n;
1532 struct inet6_dev *in6_dev;
Cong Wangf564f452013-08-31 13:44:36 +08001533
1534 in6_dev = __in6_dev_get(dev);
1535 if (!in6_dev)
1536 goto out;
1537
Cong Wangf564f452013-08-31 13:44:36 +08001538 iphdr = ipv6_hdr(skb);
1539 saddr = &iphdr->saddr;
1540 daddr = &iphdr->daddr;
1541
Cong Wangf564f452013-08-31 13:44:36 +08001542 msg = (struct nd_msg *)skb_transport_header(skb);
1543 if (msg->icmph.icmp6_code != 0 ||
1544 msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION)
1545 goto out;
1546
David Stevens4b29dba2014-03-24 10:39:58 -04001547 if (ipv6_addr_loopback(daddr) ||
1548 ipv6_addr_is_multicast(&msg->target))
1549 goto out;
1550
1551 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
Cong Wangf564f452013-08-31 13:44:36 +08001552
1553 if (n) {
1554 struct vxlan_fdb *f;
David Stevens4b29dba2014-03-24 10:39:58 -04001555 struct sk_buff *reply;
Cong Wangf564f452013-08-31 13:44:36 +08001556
1557 if (!(n->nud_state & NUD_CONNECTED)) {
1558 neigh_release(n);
1559 goto out;
1560 }
1561
1562 f = vxlan_find_mac(vxlan, n->ha);
1563 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1564 /* bridge-local neighbor */
1565 neigh_release(n);
1566 goto out;
1567 }
1568
David Stevens4b29dba2014-03-24 10:39:58 -04001569 reply = vxlan_na_create(skb, n,
1570 !!(f ? f->flags & NTF_ROUTER : 0));
1571
Cong Wangf564f452013-08-31 13:44:36 +08001572 neigh_release(n);
David Stevens4b29dba2014-03-24 10:39:58 -04001573
1574 if (reply == NULL)
1575 goto out;
1576
1577 if (netif_rx_ni(reply) == NET_RX_DROP)
1578 dev->stats.rx_dropped++;
1579
Cong Wangf564f452013-08-31 13:44:36 +08001580 } else if (vxlan->flags & VXLAN_F_L3MISS) {
David Stevens4b29dba2014-03-24 10:39:58 -04001581 union vxlan_addr ipa = {
1582 .sin6.sin6_addr = msg->target,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001583 .sin6.sin6_family = AF_INET6,
David Stevens4b29dba2014-03-24 10:39:58 -04001584 };
1585
Cong Wangf564f452013-08-31 13:44:36 +08001586 vxlan_ip_miss(dev, &ipa);
1587 }
1588
1589out:
1590 consume_skb(skb);
1591 return NETDEV_TX_OK;
1592}
1593#endif
1594
David Stevense4f67ad2012-11-20 02:50:14 +00001595static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1596{
1597 struct vxlan_dev *vxlan = netdev_priv(dev);
1598 struct neighbour *n;
David Stevense4f67ad2012-11-20 02:50:14 +00001599
1600 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1601 return false;
1602
1603 n = NULL;
1604 switch (ntohs(eth_hdr(skb)->h_proto)) {
1605 case ETH_P_IP:
Cong Wange15a00a2013-08-31 13:44:34 +08001606 {
1607 struct iphdr *pip;
1608
David Stevense4f67ad2012-11-20 02:50:14 +00001609 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1610 return false;
1611 pip = ip_hdr(skb);
1612 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
Cong Wange4c7ed42013-08-31 13:44:33 +08001613 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1614 union vxlan_addr ipa = {
1615 .sin.sin_addr.s_addr = pip->daddr,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001616 .sin.sin_family = AF_INET,
Cong Wange4c7ed42013-08-31 13:44:33 +08001617 };
1618
1619 vxlan_ip_miss(dev, &ipa);
1620 return false;
1621 }
1622
David Stevense4f67ad2012-11-20 02:50:14 +00001623 break;
Cong Wange15a00a2013-08-31 13:44:34 +08001624 }
1625#if IS_ENABLED(CONFIG_IPV6)
1626 case ETH_P_IPV6:
1627 {
1628 struct ipv6hdr *pip6;
1629
1630 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1631 return false;
1632 pip6 = ipv6_hdr(skb);
1633 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1634 if (!n && (vxlan->flags & VXLAN_F_L3MISS)) {
1635 union vxlan_addr ipa = {
1636 .sin6.sin6_addr = pip6->daddr,
Gerhard Stenzela45e92a2014-08-22 21:34:16 +02001637 .sin6.sin6_family = AF_INET6,
Cong Wange15a00a2013-08-31 13:44:34 +08001638 };
1639
1640 vxlan_ip_miss(dev, &ipa);
1641 return false;
1642 }
1643
1644 break;
1645 }
1646#endif
David Stevense4f67ad2012-11-20 02:50:14 +00001647 default:
1648 return false;
1649 }
1650
1651 if (n) {
1652 bool diff;
1653
Joe Perches7367d0b2013-09-01 11:51:23 -07001654 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
David Stevense4f67ad2012-11-20 02:50:14 +00001655 if (diff) {
1656 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1657 dev->addr_len);
1658 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1659 }
1660 neigh_release(n);
1661 return diff;
Cong Wange4c7ed42013-08-31 13:44:33 +08001662 }
1663
David Stevense4f67ad2012-11-20 02:50:14 +00001664 return false;
1665}
1666
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001667static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
Thomas Graf35114942015-01-15 03:53:55 +01001668 struct vxlan_metadata *md)
1669{
1670 struct vxlanhdr_gbp *gbp;
1671
Thomas Grafdb79a622015-02-04 17:00:04 +01001672 if (!md->gbp)
1673 return;
1674
Thomas Graf35114942015-01-15 03:53:55 +01001675 gbp = (struct vxlanhdr_gbp *)vxh;
1676 vxh->vx_flags |= htonl(VXLAN_HF_GBP);
1677
1678 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1679 gbp->dont_learn = 1;
1680
1681 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1682 gbp->policy_applied = 1;
1683
1684 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1685}
1686
Cong Wange4c7ed42013-08-31 13:44:33 +08001687#if IS_ENABLED(CONFIG_IPV6)
David Miller79b16aa2015-04-05 22:19:09 -04001688static int vxlan6_xmit_skb(struct dst_entry *dst, struct sock *sk,
1689 struct sk_buff *skb,
Cong Wange4c7ed42013-08-31 13:44:33 +08001690 struct net_device *dev, struct in6_addr *saddr,
1691 struct in6_addr *daddr, __u8 prio, __u8 ttl,
Thomas Graf6b6948d2015-07-22 17:08:42 +02001692 __be16 src_port, __be16 dst_port, __be32 vni,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001693 struct vxlan_metadata *md, bool xnet, u32 vxflags)
Cong Wange4c7ed42013-08-31 13:44:33 +08001694{
Cong Wange4c7ed42013-08-31 13:44:33 +08001695 struct vxlanhdr *vxh;
Cong Wange4c7ed42013-08-31 13:44:33 +08001696 int min_headroom;
1697 int err;
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001698 bool udp_sum = !(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX);
Tom Herbertdfd86452015-01-12 17:00:38 -08001699 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1700 u16 hdrlen = sizeof(struct vxlanhdr);
Cong Wange4c7ed42013-08-31 13:44:33 +08001701
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001702 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
Tom Herbertdfd86452015-01-12 17:00:38 -08001703 skb->ip_summed == CHECKSUM_PARTIAL) {
1704 int csum_start = skb_checksum_start_offset(skb);
1705
1706 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1707 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1708 (skb->csum_offset == offsetof(struct udphdr, check) ||
1709 skb->csum_offset == offsetof(struct tcphdr, check))) {
1710 udp_sum = false;
1711 type |= SKB_GSO_TUNNEL_REMCSUM;
1712 }
1713 }
1714
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02001715 skb_scrub_packet(skb, xnet);
Nicolas Dichtel963a88b2013-09-02 15:34:57 +02001716
Cong Wange4c7ed42013-08-31 13:44:33 +08001717 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1718 + VXLAN_HLEN + sizeof(struct ipv6hdr)
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001719 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Cong Wange4c7ed42013-08-31 13:44:33 +08001720
1721 /* Need space for new headers (invalidates iph ptr) */
1722 err = skb_cow_head(skb, min_headroom);
Pravin B Shelar74f47272014-12-23 16:20:36 -08001723 if (unlikely(err)) {
1724 kfree_skb(skb);
1725 goto err;
1726 }
Cong Wange4c7ed42013-08-31 13:44:33 +08001727
Jiri Pirko59682502014-11-19 14:04:59 +01001728 skb = vlan_hwaccel_push_inside(skb);
Pravin B Shelar74f47272014-12-23 16:20:36 -08001729 if (WARN_ON(!skb)) {
1730 err = -ENOMEM;
1731 goto err;
1732 }
Cong Wange4c7ed42013-08-31 13:44:33 +08001733
Jesse Grossb736a622015-04-09 11:19:14 -07001734 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1735 if (IS_ERR(skb)) {
1736 err = -EINVAL;
1737 goto err;
1738 }
1739
Cong Wange4c7ed42013-08-31 13:44:33 +08001740 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
Tom Herbert3bf39472015-01-08 12:31:18 -08001741 vxh->vx_flags = htonl(VXLAN_HF_VNI);
Thomas Graf614732e2015-07-21 10:44:06 +02001742 vxh->vx_vni = vni;
Cong Wange4c7ed42013-08-31 13:44:33 +08001743
Tom Herbertdfd86452015-01-12 17:00:38 -08001744 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1745 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1746 VXLAN_RCO_SHIFT;
1747
1748 if (skb->csum_offset == offsetof(struct udphdr, check))
1749 data |= VXLAN_RCO_UDP;
1750
1751 vxh->vx_vni |= htonl(data);
1752 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1753
1754 if (!skb_is_gso(skb)) {
1755 skb->ip_summed = CHECKSUM_NONE;
1756 skb->encapsulation = 0;
1757 }
1758 }
1759
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001760 if (vxflags & VXLAN_F_GBP)
1761 vxlan_build_gbp_hdr(vxh, vxflags, md);
Thomas Graf35114942015-01-15 03:53:55 +01001762
Tom Herbert996c9fd2014-09-29 20:22:33 -07001763 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1764
David Miller79b16aa2015-04-05 22:19:09 -04001765 udp_tunnel6_xmit_skb(dst, sk, skb, dev, saddr, daddr, prio,
Tom Herbertd998f8e2015-01-20 11:23:04 -08001766 ttl, src_port, dst_port,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001767 !!(vxflags & VXLAN_F_UDP_ZERO_CSUM6_TX));
Cong Wange4c7ed42013-08-31 13:44:33 +08001768 return 0;
Pravin B Shelar74f47272014-12-23 16:20:36 -08001769err:
1770 dst_release(dst);
1771 return err;
Cong Wange4c7ed42013-08-31 13:44:33 +08001772}
1773#endif
1774
Thomas Graf614732e2015-07-21 10:44:06 +02001775static int vxlan_xmit_skb(struct rtable *rt, struct sock *sk, struct sk_buff *skb,
1776 __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df,
Thomas Graf6b6948d2015-07-22 17:08:42 +02001777 __be16 src_port, __be16 dst_port, __be32 vni,
Thomas Graf614732e2015-07-21 10:44:06 +02001778 struct vxlan_metadata *md, bool xnet, u32 vxflags)
Pravin B Shelar49560532013-08-19 11:23:17 -07001779{
1780 struct vxlanhdr *vxh;
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001781 int min_headroom;
Pravin B Shelar49560532013-08-19 11:23:17 -07001782 int err;
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001783 bool udp_sum = !!(vxflags & VXLAN_F_UDP_CSUM);
Tom Herbertdfd86452015-01-12 17:00:38 -08001784 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1785 u16 hdrlen = sizeof(struct vxlanhdr);
Pravin B Shelar49560532013-08-19 11:23:17 -07001786
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001787 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
Tom Herbertdfd86452015-01-12 17:00:38 -08001788 skb->ip_summed == CHECKSUM_PARTIAL) {
1789 int csum_start = skb_checksum_start_offset(skb);
1790
1791 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1792 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1793 (skb->csum_offset == offsetof(struct udphdr, check) ||
1794 skb->csum_offset == offsetof(struct tcphdr, check))) {
1795 udp_sum = false;
1796 type |= SKB_GSO_TUNNEL_REMCSUM;
1797 }
1798 }
1799
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001800 min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07001801 + VXLAN_HLEN + sizeof(struct iphdr)
Jiri Pirkodf8a39d2015-01-13 17:13:44 +01001802 + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0);
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001803
1804 /* Need space for new headers (invalidates iph ptr) */
1805 err = skb_cow_head(skb, min_headroom);
Pravin B Shelar74f47272014-12-23 16:20:36 -08001806 if (unlikely(err)) {
1807 kfree_skb(skb);
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001808 return err;
Pravin B Shelar74f47272014-12-23 16:20:36 -08001809 }
Pravin B Shelar649c5b82013-08-19 11:23:22 -07001810
Jiri Pirko59682502014-11-19 14:04:59 +01001811 skb = vlan_hwaccel_push_inside(skb);
1812 if (WARN_ON(!skb))
1813 return -ENOMEM;
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07001814
Jesse Grossb736a622015-04-09 11:19:14 -07001815 skb = iptunnel_handle_offloads(skb, udp_sum, type);
1816 if (IS_ERR(skb))
1817 return PTR_ERR(skb);
1818
Pravin B Shelar49560532013-08-19 11:23:17 -07001819 vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh));
Tom Herbert3bf39472015-01-08 12:31:18 -08001820 vxh->vx_flags = htonl(VXLAN_HF_VNI);
Thomas Graf614732e2015-07-21 10:44:06 +02001821 vxh->vx_vni = vni;
Pravin B Shelar49560532013-08-19 11:23:17 -07001822
Tom Herbertdfd86452015-01-12 17:00:38 -08001823 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1824 u32 data = (skb_checksum_start_offset(skb) - hdrlen) >>
1825 VXLAN_RCO_SHIFT;
1826
1827 if (skb->csum_offset == offsetof(struct udphdr, check))
1828 data |= VXLAN_RCO_UDP;
1829
1830 vxh->vx_vni |= htonl(data);
1831 vxh->vx_flags |= htonl(VXLAN_HF_RCO);
1832
1833 if (!skb_is_gso(skb)) {
1834 skb->ip_summed = CHECKSUM_NONE;
1835 skb->encapsulation = 0;
1836 }
1837 }
1838
Tom Herbertaf33c1a2015-01-20 11:23:05 -08001839 if (vxflags & VXLAN_F_GBP)
1840 vxlan_build_gbp_hdr(vxh, vxflags, md);
Thomas Graf35114942015-01-15 03:53:55 +01001841
Tom Herbert996c9fd2014-09-29 20:22:33 -07001842 skb_set_inner_protocol(skb, htons(ETH_P_TEB));
1843
Pravin B Shelar039f5062015-12-24 14:34:54 -08001844 udp_tunnel_xmit_skb(rt, sk, skb, src, dst, tos, ttl, df,
1845 src_port, dst_port, xnet,
1846 !(vxflags & VXLAN_F_UDP_CSUM));
1847 return 0;
Pravin B Shelar49560532013-08-19 11:23:17 -07001848}
Pravin B Shelar49560532013-08-19 11:23:17 -07001849
Jiri Bence5d4b292015-12-07 13:04:30 +01001850#if IS_ENABLED(CONFIG_IPV6)
1851static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
1852 struct sk_buff *skb, int oif,
1853 const struct in6_addr *daddr,
1854 struct in6_addr *saddr)
1855{
1856 struct dst_entry *ndst;
1857 struct flowi6 fl6;
1858 int err;
1859
1860 memset(&fl6, 0, sizeof(fl6));
1861 fl6.flowi6_oif = oif;
1862 fl6.daddr = *daddr;
1863 fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr;
1864 fl6.flowi6_mark = skb->mark;
1865 fl6.flowi6_proto = IPPROTO_UDP;
1866
1867 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
1868 vxlan->vn6_sock->sock->sk,
1869 &ndst, &fl6);
1870 if (err < 0)
1871 return ERR_PTR(err);
1872
1873 *saddr = fl6.saddr;
1874 return ndst;
1875}
1876#endif
1877
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001878/* Bypass encapsulation if the destination is local */
1879static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
1880 struct vxlan_dev *dst_vxlan)
1881{
Li RongQing8f849852014-01-04 13:57:59 +08001882 struct pcpu_sw_netstats *tx_stats, *rx_stats;
Cong Wange4c7ed42013-08-31 13:44:33 +08001883 union vxlan_addr loopback;
1884 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
Li RongQingce6502a2014-10-16 08:49:41 +08001885 struct net_device *dev = skb->dev;
1886 int len = skb->len;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001887
Li RongQing8f849852014-01-04 13:57:59 +08001888 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
1889 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001890 skb->pkt_type = PACKET_HOST;
1891 skb->encapsulation = 0;
1892 skb->dev = dst_vxlan->dev;
1893 __skb_pull(skb, skb_network_offset(skb));
1894
Cong Wange4c7ed42013-08-31 13:44:33 +08001895 if (remote_ip->sa.sa_family == AF_INET) {
1896 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
1897 loopback.sa.sa_family = AF_INET;
1898#if IS_ENABLED(CONFIG_IPV6)
1899 } else {
1900 loopback.sin6.sin6_addr = in6addr_loopback;
1901 loopback.sa.sa_family = AF_INET6;
1902#endif
1903 }
1904
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001905 if (dst_vxlan->flags & VXLAN_F_LEARN)
Cong Wange4c7ed42013-08-31 13:44:33 +08001906 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source);
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001907
1908 u64_stats_update_begin(&tx_stats->syncp);
1909 tx_stats->tx_packets++;
Li RongQingce6502a2014-10-16 08:49:41 +08001910 tx_stats->tx_bytes += len;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001911 u64_stats_update_end(&tx_stats->syncp);
1912
1913 if (netif_rx(skb) == NET_RX_SUCCESS) {
1914 u64_stats_update_begin(&rx_stats->syncp);
1915 rx_stats->rx_packets++;
Li RongQingce6502a2014-10-16 08:49:41 +08001916 rx_stats->rx_bytes += len;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001917 u64_stats_update_end(&rx_stats->syncp);
1918 } else {
Li RongQingce6502a2014-10-16 08:49:41 +08001919 dev->stats.rx_dropped++;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001920 }
1921}
1922
Stephen Hemminger4ad16932013-06-17 14:16:11 -07001923static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
1924 struct vxlan_rdst *rdst, bool did_rsc)
stephen hemmingerd3428942012-10-01 12:32:35 +00001925{
Thomas Graf3093fbe2015-07-21 10:44:00 +02001926 struct ip_tunnel_info *info;
stephen hemmingerd3428942012-10-01 12:32:35 +00001927 struct vxlan_dev *vxlan = netdev_priv(dev);
Jiri Bencb1be00a2015-09-24 13:50:02 +02001928 struct sock *sk;
Cong Wange4c7ed42013-08-31 13:44:33 +08001929 struct rtable *rt = NULL;
stephen hemmingerd3428942012-10-01 12:32:35 +00001930 const struct iphdr *old_iph;
stephen hemmingerd3428942012-10-01 12:32:35 +00001931 struct flowi4 fl4;
Cong Wange4c7ed42013-08-31 13:44:33 +08001932 union vxlan_addr *dst;
Thomas Grafee122c72015-07-21 10:43:58 +02001933 union vxlan_addr remote_ip;
1934 struct vxlan_metadata _md;
1935 struct vxlan_metadata *md = &_md;
Cong Wange4c7ed42013-08-31 13:44:33 +08001936 __be16 src_port = 0, dst_port;
Stephen Hemminger234f5b72013-06-17 14:16:41 -07001937 u32 vni;
stephen hemmingerd3428942012-10-01 12:32:35 +00001938 __be16 df = 0;
1939 __u8 tos, ttl;
Pravin B Shelar0e6fbc52013-06-17 17:49:56 -07001940 int err;
Thomas Grafee122c72015-07-21 10:43:58 +02001941 u32 flags = vxlan->flags;
stephen hemmingerd3428942012-10-01 12:32:35 +00001942
Jiri Benc61adedf2015-08-20 13:56:25 +02001943 info = skb_tunnel_info(skb);
Thomas Graf3093fbe2015-07-21 10:44:00 +02001944
Thomas Grafee122c72015-07-21 10:43:58 +02001945 if (rdst) {
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001946 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
Thomas Grafee122c72015-07-21 10:43:58 +02001947 vni = rdst->remote_vni;
1948 dst = &rdst->remote_ip;
1949 } else {
1950 if (!info) {
1951 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
1952 dev->name);
1953 goto drop;
1954 }
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001955 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
Thomas Grafee122c72015-07-21 10:43:58 +02001956 vni = be64_to_cpu(info->key.tun_id);
Jiri Bencb1be00a2015-09-24 13:50:02 +02001957 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
1958 if (remote_ip.sa.sa_family == AF_INET)
Jiri Benca725e512015-08-20 13:56:30 +02001959 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
1960 else
1961 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
Thomas Grafee122c72015-07-21 10:43:58 +02001962 dst = &remote_ip;
1963 }
David Stevense4f67ad2012-11-20 02:50:14 +00001964
Cong Wange4c7ed42013-08-31 13:44:33 +08001965 if (vxlan_addr_any(dst)) {
David Stevense4f67ad2012-11-20 02:50:14 +00001966 if (did_rsc) {
David Stevense4f67ad2012-11-20 02:50:14 +00001967 /* short-circuited back to local bridge */
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00001968 vxlan_encap_bypass(skb, vxlan, vxlan);
Stephen Hemminger4ad16932013-06-17 14:16:11 -07001969 return;
David Stevense4f67ad2012-11-20 02:50:14 +00001970 }
stephen hemmingeref59feb2012-10-09 20:35:46 +00001971 goto drop;
David Stevense4f67ad2012-11-20 02:50:14 +00001972 }
stephen hemmingeref59feb2012-10-09 20:35:46 +00001973
stephen hemmingerd3428942012-10-01 12:32:35 +00001974 old_iph = ip_hdr(skb);
1975
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001976 ttl = vxlan->cfg.ttl;
Cong Wange4c7ed42013-08-31 13:44:33 +08001977 if (!ttl && vxlan_addr_multicast(dst))
stephen hemmingerd3428942012-10-01 12:32:35 +00001978 ttl = 1;
1979
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001980 tos = vxlan->cfg.tos;
stephen hemmingerd3428942012-10-01 12:32:35 +00001981 if (tos == 1)
Pravin B Shelar206aaaf2013-03-25 14:49:53 +00001982 tos = ip_tunnel_get_dsfield(old_iph, skb);
stephen hemmingerd3428942012-10-01 12:32:35 +00001983
Thomas Graf0dfbdf42015-07-21 10:44:02 +02001984 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
1985 vxlan->cfg.port_max, true);
stephen hemmingerd3428942012-10-01 12:32:35 +00001986
Jiri Benca725e512015-08-20 13:56:30 +02001987 if (info) {
Jiri Benca725e512015-08-20 13:56:30 +02001988 ttl = info->key.ttl;
1989 tos = info->key.tos;
1990
1991 if (info->options_len)
Pravin B Shelar4c222792015-08-30 18:09:38 -07001992 md = ip_tunnel_info_opts(info);
Jiri Benca725e512015-08-20 13:56:30 +02001993 } else {
1994 md->gbp = skb->mark;
1995 }
1996
Cong Wange4c7ed42013-08-31 13:44:33 +08001997 if (dst->sa.sa_family == AF_INET) {
Jiri Bencb1be00a2015-09-24 13:50:02 +02001998 if (!vxlan->vn4_sock)
1999 goto drop;
2000 sk = vxlan->vn4_sock->sock->sk;
2001
Jesse Gross35e2d112016-01-20 16:22:47 -08002002 if (info) {
2003 if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT)
2004 df = htons(IP_DF);
2005
2006 if (info->key.tun_flags & TUNNEL_CSUM)
2007 flags |= VXLAN_F_UDP_CSUM;
2008 else
2009 flags &= ~VXLAN_F_UDP_CSUM;
2010 }
Thomas Grafee122c72015-07-21 10:43:58 +02002011
Cong Wange4c7ed42013-08-31 13:44:33 +08002012 memset(&fl4, 0, sizeof(fl4));
Thomas Grafee122c72015-07-21 10:43:58 +02002013 fl4.flowi4_oif = rdst ? rdst->remote_ifindex : 0;
Cong Wange4c7ed42013-08-31 13:44:33 +08002014 fl4.flowi4_tos = RT_TOS(tos);
Thomas Graf239fb792015-05-05 15:09:21 +02002015 fl4.flowi4_mark = skb->mark;
2016 fl4.flowi4_proto = IPPROTO_UDP;
Cong Wange4c7ed42013-08-31 13:44:33 +08002017 fl4.daddr = dst->sin.sin_addr.s_addr;
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002018 fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr;
stephen hemmingerca78f182012-10-09 20:35:48 +00002019
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002020 rt = ip_route_output_key(vxlan->net, &fl4);
Cong Wange4c7ed42013-08-31 13:44:33 +08002021 if (IS_ERR(rt)) {
2022 netdev_dbg(dev, "no route to %pI4\n",
2023 &dst->sin.sin_addr.s_addr);
2024 dev->stats.tx_carrier_errors++;
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00002025 goto tx_error;
Cong Wange4c7ed42013-08-31 13:44:33 +08002026 }
2027
2028 if (rt->dst.dev == dev) {
2029 netdev_dbg(dev, "circular route to %pI4\n",
2030 &dst->sin.sin_addr.s_addr);
2031 dev->stats.collisions++;
Fan Dufffc15a2013-12-09 10:33:53 +08002032 goto rt_tx_error;
Cong Wange4c7ed42013-08-31 13:44:33 +08002033 }
2034
2035 /* Bypass encapsulation if the destination is local */
2036 if (rt->rt_flags & RTCF_LOCAL &&
2037 !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2038 struct vxlan_dev *dst_vxlan;
2039
2040 ip_rt_put(rt);
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -02002041 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
Thomas Grafac5132d2015-01-15 03:53:56 +01002042 dst->sa.sa_family, dst_port,
2043 vxlan->flags);
Cong Wange4c7ed42013-08-31 13:44:33 +08002044 if (!dst_vxlan)
2045 goto tx_error;
2046 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2047 return;
2048 }
2049
2050 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2051 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
David Miller79b16aa2015-04-05 22:19:09 -04002052 err = vxlan_xmit_skb(rt, sk, skb, fl4.saddr,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002053 dst->sin.sin_addr.s_addr, tos, ttl, df,
Thomas Graf614732e2015-07-21 10:44:06 +02002054 src_port, dst_port, htonl(vni << 8), md,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002055 !net_eq(vxlan->net, dev_net(vxlan->dev)),
Thomas Grafee122c72015-07-21 10:43:58 +02002056 flags);
Pravin B Shelar74f47272014-12-23 16:20:36 -08002057 if (err < 0) {
2058 /* skb is already freed. */
2059 skb = NULL;
Cong Wange4c7ed42013-08-31 13:44:33 +08002060 goto rt_tx_error;
Pravin B Shelar74f47272014-12-23 16:20:36 -08002061 }
Cong Wange4c7ed42013-08-31 13:44:33 +08002062#if IS_ENABLED(CONFIG_IPV6)
2063 } else {
Cong Wange4c7ed42013-08-31 13:44:33 +08002064 struct dst_entry *ndst;
Jiri Bence5d4b292015-12-07 13:04:30 +01002065 struct in6_addr saddr;
Jiri Benc6f264e42015-08-20 13:56:29 +02002066 u32 rt6i_flags;
Cong Wange4c7ed42013-08-31 13:44:33 +08002067
Jiri Bencb1be00a2015-09-24 13:50:02 +02002068 if (!vxlan->vn6_sock)
2069 goto drop;
2070 sk = vxlan->vn6_sock->sock->sk;
2071
Jiri Bence5d4b292015-12-07 13:04:30 +01002072 ndst = vxlan6_get_route(vxlan, skb,
2073 rdst ? rdst->remote_ifindex : 0,
2074 &dst->sin6.sin6_addr, &saddr);
2075 if (IS_ERR(ndst)) {
Cong Wange4c7ed42013-08-31 13:44:33 +08002076 netdev_dbg(dev, "no route to %pI6\n",
2077 &dst->sin6.sin6_addr);
2078 dev->stats.tx_carrier_errors++;
2079 goto tx_error;
2080 }
2081
2082 if (ndst->dev == dev) {
2083 netdev_dbg(dev, "circular route to %pI6\n",
2084 &dst->sin6.sin6_addr);
2085 dst_release(ndst);
2086 dev->stats.collisions++;
2087 goto tx_error;
2088 }
2089
2090 /* Bypass encapsulation if the destination is local */
Jiri Benc6f264e42015-08-20 13:56:29 +02002091 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2092 if (rt6i_flags & RTF_LOCAL &&
2093 !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
Cong Wange4c7ed42013-08-31 13:44:33 +08002094 struct vxlan_dev *dst_vxlan;
2095
2096 dst_release(ndst);
Marcelo Leitner19ca9fc2014-11-13 14:43:08 -02002097 dst_vxlan = vxlan_find_vni(vxlan->net, vni,
Thomas Grafac5132d2015-01-15 03:53:56 +01002098 dst->sa.sa_family, dst_port,
2099 vxlan->flags);
Cong Wange4c7ed42013-08-31 13:44:33 +08002100 if (!dst_vxlan)
2101 goto tx_error;
2102 vxlan_encap_bypass(skb, vxlan, dst_vxlan);
2103 return;
2104 }
2105
Jesse Gross35e2d112016-01-20 16:22:47 -08002106 if (info) {
2107 if (info->key.tun_flags & TUNNEL_CSUM)
2108 flags &= ~VXLAN_F_UDP_ZERO_CSUM6_TX;
2109 else
2110 flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2111 }
2112
Cong Wange4c7ed42013-08-31 13:44:33 +08002113 ttl = ttl ? : ip6_dst_hoplimit(ndst);
Jiri Bence5d4b292015-12-07 13:04:30 +01002114 err = vxlan6_xmit_skb(ndst, sk, skb, dev, &saddr, &dst->sin6.sin6_addr,
Thomas Graf614732e2015-07-21 10:44:06 +02002115 0, ttl, src_port, dst_port, htonl(vni << 8), md,
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002116 !net_eq(vxlan->net, dev_net(vxlan->dev)),
Jiri Benca725e512015-08-20 13:56:30 +02002117 flags);
Cong Wange4c7ed42013-08-31 13:44:33 +08002118#endif
Sridhar Samudrala9dcc71e2013-04-02 12:31:52 +00002119 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002120
Stephen Hemminger4ad16932013-06-17 14:16:11 -07002121 return;
stephen hemmingerd3428942012-10-01 12:32:35 +00002122
2123drop:
2124 dev->stats.tx_dropped++;
2125 goto tx_free;
2126
Pravin B Shelar49560532013-08-19 11:23:17 -07002127rt_tx_error:
2128 ip_rt_put(rt);
stephen hemmingerd3428942012-10-01 12:32:35 +00002129tx_error:
2130 dev->stats.tx_errors++;
2131tx_free:
2132 dev_kfree_skb(skb);
stephen hemmingerd3428942012-10-01 12:32:35 +00002133}
2134
David Stevens66817122013-03-15 04:35:51 +00002135/* Transmit local packets over Vxlan
2136 *
2137 * Outer IP header inherits ECN and DF from inner header.
2138 * Outer UDP destination is the VXLAN assigned port.
2139 * source port is based on hash of flow
2140 */
2141static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2142{
2143 struct vxlan_dev *vxlan = netdev_priv(dev);
Thomas Graf3093fbe2015-07-21 10:44:00 +02002144 const struct ip_tunnel_info *info;
David Stevens66817122013-03-15 04:35:51 +00002145 struct ethhdr *eth;
2146 bool did_rsc = false;
Eric Dumazet8f646c92014-01-06 09:54:31 -08002147 struct vxlan_rdst *rdst, *fdst = NULL;
David Stevens66817122013-03-15 04:35:51 +00002148 struct vxlan_fdb *f;
David Stevens66817122013-03-15 04:35:51 +00002149
Jiri Benc61adedf2015-08-20 13:56:25 +02002150 info = skb_tunnel_info(skb);
Thomas Graf3093fbe2015-07-21 10:44:00 +02002151
David Stevens66817122013-03-15 04:35:51 +00002152 skb_reset_mac_header(skb);
2153 eth = eth_hdr(skb);
2154
Cong Wangf564f452013-08-31 13:44:36 +08002155 if ((vxlan->flags & VXLAN_F_PROXY)) {
2156 if (ntohs(eth->h_proto) == ETH_P_ARP)
2157 return arp_reduce(dev, skb);
2158#if IS_ENABLED(CONFIG_IPV6)
2159 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
Li RongQing91269e32014-10-16 09:17:18 +08002160 pskb_may_pull(skb, sizeof(struct ipv6hdr)
2161 + sizeof(struct nd_msg)) &&
Cong Wangf564f452013-08-31 13:44:36 +08002162 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2163 struct nd_msg *msg;
2164
2165 msg = (struct nd_msg *)skb_transport_header(skb);
2166 if (msg->icmph.icmp6_code == 0 &&
2167 msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2168 return neigh_reduce(dev, skb);
2169 }
Li RongQing7a9f5262014-10-17 14:06:16 +08002170 eth = eth_hdr(skb);
Cong Wangf564f452013-08-31 13:44:36 +08002171#endif
2172 }
David Stevens66817122013-03-15 04:35:51 +00002173
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07002174 if (vxlan->flags & VXLAN_F_COLLECT_METADATA &&
Jiri Benc46fa0622015-08-28 20:48:19 +02002175 info && info->mode & IP_TUNNEL_INFO_TX) {
Thomas Grafee122c72015-07-21 10:43:58 +02002176 vxlan_xmit_one(skb, dev, NULL, false);
2177 return NETDEV_TX_OK;
2178 }
2179
David Stevens66817122013-03-15 04:35:51 +00002180 f = vxlan_find_mac(vxlan, eth->h_dest);
David Stevensae884082013-04-19 00:36:26 +00002181 did_rsc = false;
2182
2183 if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) &&
Cong Wange15a00a2013-08-31 13:44:34 +08002184 (ntohs(eth->h_proto) == ETH_P_IP ||
2185 ntohs(eth->h_proto) == ETH_P_IPV6)) {
David Stevensae884082013-04-19 00:36:26 +00002186 did_rsc = route_shortcircuit(dev, skb);
2187 if (did_rsc)
2188 f = vxlan_find_mac(vxlan, eth->h_dest);
2189 }
2190
David Stevens66817122013-03-15 04:35:51 +00002191 if (f == NULL) {
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002192 f = vxlan_find_mac(vxlan, all_zeros_mac);
2193 if (f == NULL) {
2194 if ((vxlan->flags & VXLAN_F_L2MISS) &&
2195 !is_multicast_ether_addr(eth->h_dest))
2196 vxlan_fdb_miss(vxlan, eth->h_dest);
David Stevens66817122013-03-15 04:35:51 +00002197
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002198 dev->stats.tx_dropped++;
Eric Dumazet8f646c92014-01-06 09:54:31 -08002199 kfree_skb(skb);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002200 return NETDEV_TX_OK;
Stephen Hemminger3e61aa82013-06-17 14:16:12 -07002201 }
David Stevens66817122013-03-15 04:35:51 +00002202 }
2203
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002204 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2205 struct sk_buff *skb1;
2206
Eric Dumazet8f646c92014-01-06 09:54:31 -08002207 if (!fdst) {
2208 fdst = rdst;
2209 continue;
2210 }
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002211 skb1 = skb_clone(skb, GFP_ATOMIC);
2212 if (skb1)
2213 vxlan_xmit_one(skb1, dev, rdst, did_rsc);
2214 }
2215
Eric Dumazet8f646c92014-01-06 09:54:31 -08002216 if (fdst)
2217 vxlan_xmit_one(skb, dev, fdst, did_rsc);
2218 else
2219 kfree_skb(skb);
Stephen Hemminger4ad16932013-06-17 14:16:11 -07002220 return NETDEV_TX_OK;
David Stevens66817122013-03-15 04:35:51 +00002221}
2222
stephen hemmingerd3428942012-10-01 12:32:35 +00002223/* Walk the forwarding table and purge stale entries */
2224static void vxlan_cleanup(unsigned long arg)
2225{
2226 struct vxlan_dev *vxlan = (struct vxlan_dev *) arg;
2227 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2228 unsigned int h;
2229
2230 if (!netif_running(vxlan->dev))
2231 return;
2232
stephen hemmingerd3428942012-10-01 12:32:35 +00002233 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2234 struct hlist_node *p, *n;
Sorin Dumitru14e1d0f2015-05-26 10:42:04 +03002235
2236 spin_lock_bh(&vxlan->hash_lock);
stephen hemmingerd3428942012-10-01 12:32:35 +00002237 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2238 struct vxlan_fdb *f
2239 = container_of(p, struct vxlan_fdb, hlist);
2240 unsigned long timeout;
2241
stephen hemminger3c172862012-10-26 06:24:34 +00002242 if (f->state & NUD_PERMANENT)
stephen hemmingerd3428942012-10-01 12:32:35 +00002243 continue;
2244
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002245 timeout = f->used + vxlan->cfg.age_interval * HZ;
stephen hemmingerd3428942012-10-01 12:32:35 +00002246 if (time_before_eq(timeout, jiffies)) {
2247 netdev_dbg(vxlan->dev,
2248 "garbage collect %pM\n",
2249 f->eth_addr);
2250 f->state = NUD_STALE;
2251 vxlan_fdb_destroy(vxlan, f);
2252 } else if (time_before(timeout, next_timer))
2253 next_timer = timeout;
2254 }
Sorin Dumitru14e1d0f2015-05-26 10:42:04 +03002255 spin_unlock_bh(&vxlan->hash_lock);
stephen hemmingerd3428942012-10-01 12:32:35 +00002256 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002257
2258 mod_timer(&vxlan->age_timer, next_timer);
2259}
2260
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002261static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan)
2262{
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002263 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002264 __u32 vni = vxlan->default_dst.remote_vni;
2265
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002266 spin_lock(&vn->sock_lock);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002267 hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni));
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002268 spin_unlock(&vn->sock_lock);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002269}
2270
stephen hemmingerd3428942012-10-01 12:32:35 +00002271/* Setup stats when device is created */
2272static int vxlan_init(struct net_device *dev)
2273{
WANG Cong1c213bd2014-02-13 11:46:28 -08002274 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
Pravin B Shelare8171042013-03-25 14:49:46 +00002275 if (!dev->tstats)
stephen hemmingerd3428942012-10-01 12:32:35 +00002276 return -ENOMEM;
2277
2278 return 0;
2279}
2280
Stephen Hemmingerba609e92013-06-25 17:06:01 -07002281static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan)
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002282{
2283 struct vxlan_fdb *f;
2284
2285 spin_lock_bh(&vxlan->hash_lock);
2286 f = __vxlan_find_mac(vxlan, all_zeros_mac);
2287 if (f)
2288 vxlan_fdb_destroy(vxlan, f);
2289 spin_unlock_bh(&vxlan->hash_lock);
2290}
2291
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002292static void vxlan_uninit(struct net_device *dev)
2293{
2294 struct vxlan_dev *vxlan = netdev_priv(dev);
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002295
Stephen Hemmingerba609e92013-06-25 17:06:01 -07002296 vxlan_fdb_delete_default(vxlan);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002297
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002298 free_percpu(dev->tstats);
2299}
2300
stephen hemmingerd3428942012-10-01 12:32:35 +00002301/* Start ageing timer and join group when device is brought up */
2302static int vxlan_open(struct net_device *dev)
2303{
2304 struct vxlan_dev *vxlan = netdev_priv(dev);
Jiri Benc205f3562015-09-24 13:50:01 +02002305 int ret;
Stephen Hemminger1c51a912013-06-17 14:16:11 -07002306
Jiri Benc205f3562015-09-24 13:50:01 +02002307 ret = vxlan_sock_add(vxlan);
2308 if (ret < 0)
2309 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00002310
Gao feng79d4a942013-12-10 16:37:32 +08002311 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002312 ret = vxlan_igmp_join(vxlan);
Marcelo Ricardo Leitnerbef00572015-08-25 20:22:35 -03002313 if (ret == -EADDRINUSE)
2314 ret = 0;
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002315 if (ret) {
Jiri Benc205f3562015-09-24 13:50:01 +02002316 vxlan_sock_release(vxlan);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002317 return ret;
2318 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002319 }
2320
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002321 if (vxlan->cfg.age_interval)
stephen hemmingerd3428942012-10-01 12:32:35 +00002322 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2323
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002324 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00002325}
2326
2327/* Purge the forwarding table */
2328static void vxlan_flush(struct vxlan_dev *vxlan)
2329{
Cong Wang31fec5a2013-05-27 22:35:52 +00002330 unsigned int h;
stephen hemmingerd3428942012-10-01 12:32:35 +00002331
2332 spin_lock_bh(&vxlan->hash_lock);
2333 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2334 struct hlist_node *p, *n;
2335 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2336 struct vxlan_fdb *f
2337 = container_of(p, struct vxlan_fdb, hlist);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002338 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2339 if (!is_zero_ether_addr(f->eth_addr))
2340 vxlan_fdb_destroy(vxlan, f);
stephen hemmingerd3428942012-10-01 12:32:35 +00002341 }
2342 }
2343 spin_unlock_bh(&vxlan->hash_lock);
2344}
2345
2346/* Cleanup timer and forwarding table on shutdown */
2347static int vxlan_stop(struct net_device *dev)
2348{
2349 struct vxlan_dev *vxlan = netdev_priv(dev);
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002350 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002351 int ret = 0;
stephen hemmingerd3428942012-10-01 12:32:35 +00002352
Marcelo Ricardo Leitner24c0e682015-03-23 16:23:12 -03002353 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
WANG Congf13b1682015-04-08 14:48:30 -07002354 !vxlan_group_used(vn, vxlan))
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002355 ret = vxlan_igmp_leave(vxlan);
stephen hemmingerd3428942012-10-01 12:32:35 +00002356
2357 del_timer_sync(&vxlan->age_timer);
2358
2359 vxlan_flush(vxlan);
Jiri Benc205f3562015-09-24 13:50:01 +02002360 vxlan_sock_release(vxlan);
stephen hemmingerd3428942012-10-01 12:32:35 +00002361
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002362 return ret;
stephen hemmingerd3428942012-10-01 12:32:35 +00002363}
2364
stephen hemmingerd3428942012-10-01 12:32:35 +00002365/* Stub, nothing needs to be done. */
2366static void vxlan_set_multicast_list(struct net_device *dev)
2367{
2368}
2369
Daniel Borkmann345010b2013-12-18 00:21:08 +01002370static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2371{
2372 struct vxlan_dev *vxlan = netdev_priv(dev);
2373 struct vxlan_rdst *dst = &vxlan->default_dst;
2374 struct net_device *lowerdev;
2375 int max_mtu;
2376
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002377 lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex);
Daniel Borkmann345010b2013-12-18 00:21:08 +01002378 if (lowerdev == NULL)
2379 return eth_change_mtu(dev, new_mtu);
2380
2381 if (dst->remote_ip.sa.sa_family == AF_INET6)
2382 max_mtu = lowerdev->mtu - VXLAN6_HEADROOM;
2383 else
2384 max_mtu = lowerdev->mtu - VXLAN_HEADROOM;
2385
2386 if (new_mtu < 68 || new_mtu > max_mtu)
2387 return -EINVAL;
2388
2389 dev->mtu = new_mtu;
2390 return 0;
2391}
2392
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002393static int egress_ipv4_tun_info(struct net_device *dev, struct sk_buff *skb,
2394 struct ip_tunnel_info *info,
2395 __be16 sport, __be16 dport)
2396{
2397 struct vxlan_dev *vxlan = netdev_priv(dev);
2398 struct rtable *rt;
2399 struct flowi4 fl4;
2400
2401 memset(&fl4, 0, sizeof(fl4));
2402 fl4.flowi4_tos = RT_TOS(info->key.tos);
2403 fl4.flowi4_mark = skb->mark;
2404 fl4.flowi4_proto = IPPROTO_UDP;
2405 fl4.daddr = info->key.u.ipv4.dst;
2406
2407 rt = ip_route_output_key(vxlan->net, &fl4);
2408 if (IS_ERR(rt))
2409 return PTR_ERR(rt);
2410 ip_rt_put(rt);
2411
2412 info->key.u.ipv4.src = fl4.saddr;
2413 info->key.tp_src = sport;
2414 info->key.tp_dst = dport;
2415 return 0;
2416}
2417
2418static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2419{
2420 struct vxlan_dev *vxlan = netdev_priv(dev);
2421 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2422 __be16 sport, dport;
2423
2424 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2425 vxlan->cfg.port_max, true);
2426 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2427
Jiri Benc239e9442015-12-07 13:04:31 +01002428 if (ip_tunnel_info_af(info) == AF_INET) {
2429 if (!vxlan->vn4_sock)
2430 return -EINVAL;
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002431 return egress_ipv4_tun_info(dev, skb, info, sport, dport);
Jiri Benc239e9442015-12-07 13:04:31 +01002432 } else {
2433#if IS_ENABLED(CONFIG_IPV6)
2434 struct dst_entry *ndst;
2435
2436 if (!vxlan->vn6_sock)
2437 return -EINVAL;
2438 ndst = vxlan6_get_route(vxlan, skb, 0,
2439 &info->key.u.ipv6.dst,
2440 &info->key.u.ipv6.src);
2441 if (IS_ERR(ndst))
2442 return PTR_ERR(ndst);
2443 dst_release(ndst);
2444
2445 info->key.tp_src = sport;
2446 info->key.tp_dst = dport;
2447#else /* !CONFIG_IPV6 */
2448 return -EPFNOSUPPORT;
2449#endif
2450 }
2451 return 0;
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002452}
2453
stephen hemmingerd3428942012-10-01 12:32:35 +00002454static const struct net_device_ops vxlan_netdev_ops = {
2455 .ndo_init = vxlan_init,
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002456 .ndo_uninit = vxlan_uninit,
stephen hemmingerd3428942012-10-01 12:32:35 +00002457 .ndo_open = vxlan_open,
2458 .ndo_stop = vxlan_stop,
2459 .ndo_start_xmit = vxlan_xmit,
Pravin B Shelare8171042013-03-25 14:49:46 +00002460 .ndo_get_stats64 = ip_tunnel_get_stats64,
stephen hemmingerd3428942012-10-01 12:32:35 +00002461 .ndo_set_rx_mode = vxlan_set_multicast_list,
Daniel Borkmann345010b2013-12-18 00:21:08 +01002462 .ndo_change_mtu = vxlan_change_mtu,
stephen hemmingerd3428942012-10-01 12:32:35 +00002463 .ndo_validate_addr = eth_validate_addr,
2464 .ndo_set_mac_address = eth_mac_addr,
2465 .ndo_fdb_add = vxlan_fdb_add,
2466 .ndo_fdb_del = vxlan_fdb_delete,
2467 .ndo_fdb_dump = vxlan_fdb_dump,
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002468 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
stephen hemmingerd3428942012-10-01 12:32:35 +00002469};
2470
2471/* Info for udev, that this is a virtual tunnel endpoint */
2472static struct device_type vxlan_type = {
2473 .name = "vxlan",
2474};
2475
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002476/* Calls the ndo_add_vxlan_port of the caller in order to
Joseph Gasparakis35e42372013-09-13 07:34:13 -07002477 * supply the listening VXLAN udp ports. Callers are expected
2478 * to implement the ndo_add_vxlan_port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002479 */
2480void vxlan_get_rx_port(struct net_device *dev)
2481{
2482 struct vxlan_sock *vs;
2483 struct net *net = dev_net(dev);
2484 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2485 sa_family_t sa_family;
Joseph Gasparakis35e42372013-09-13 07:34:13 -07002486 __be16 port;
2487 unsigned int i;
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002488
2489 spin_lock(&vn->sock_lock);
2490 for (i = 0; i < PORT_HASH_SIZE; ++i) {
Joseph Gasparakis35e42372013-09-13 07:34:13 -07002491 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2492 port = inet_sk(vs->sock->sk)->inet_sport;
Jiri Benc705cc622015-08-20 13:56:28 +02002493 sa_family = vxlan_get_sk_family(vs);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07002494 dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family,
2495 port);
2496 }
2497 }
2498 spin_unlock(&vn->sock_lock);
2499}
2500EXPORT_SYMBOL_GPL(vxlan_get_rx_port);
2501
stephen hemmingerd3428942012-10-01 12:32:35 +00002502/* Initialize the device structure. */
2503static void vxlan_setup(struct net_device *dev)
2504{
2505 struct vxlan_dev *vxlan = netdev_priv(dev);
Cong Wang31fec5a2013-05-27 22:35:52 +00002506 unsigned int h;
stephen hemmingerd3428942012-10-01 12:32:35 +00002507
2508 eth_hw_addr_random(dev);
2509 ether_setup(dev);
2510
2511 dev->netdev_ops = &vxlan_netdev_ops;
Stephen Hemmingerebf40632013-06-17 14:16:11 -07002512 dev->destructor = free_netdev;
stephen hemmingerd3428942012-10-01 12:32:35 +00002513 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2514
stephen hemmingerd3428942012-10-01 12:32:35 +00002515 dev->features |= NETIF_F_LLTX;
Joseph Gasparakisd6727fe2012-12-07 14:14:16 +00002516 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
Joseph Gasparakis0afb1662012-12-07 14:14:18 +00002517 dev->features |= NETIF_F_RXCSUM;
Pravin B Shelar05c0db02013-03-07 13:22:36 +00002518 dev->features |= NETIF_F_GSO_SOFTWARE;
Joseph Gasparakis0afb1662012-12-07 14:14:18 +00002519
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07002520 dev->vlan_features = dev->features;
2521 dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
Joseph Gasparakis0afb1662012-12-07 14:14:18 +00002522 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
Pravin B Shelar05c0db02013-03-07 13:22:36 +00002523 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
Pravin B Shelar1eaa8172013-08-19 11:23:29 -07002524 dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX;
Eric Dumazet02875872014-10-05 18:38:35 -07002525 netif_keep_dst(dev);
Phil Sutter22dba392015-08-18 10:30:35 +02002526 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE;
stephen hemmingerd3428942012-10-01 12:32:35 +00002527
stephen hemminger553675f2013-05-16 11:35:20 +00002528 INIT_LIST_HEAD(&vxlan->next);
stephen hemmingerd3428942012-10-01 12:32:35 +00002529 spin_lock_init(&vxlan->hash_lock);
2530
2531 init_timer_deferrable(&vxlan->age_timer);
2532 vxlan->age_timer.function = vxlan_cleanup;
2533 vxlan->age_timer.data = (unsigned long) vxlan;
2534
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002535 vxlan->cfg.dst_port = htons(vxlan_port);
stephen hemminger05f47d62012-10-09 20:35:50 +00002536
stephen hemmingerd3428942012-10-01 12:32:35 +00002537 vxlan->dev = dev;
2538
Tom Herbert58ce31c2015-08-19 17:07:33 -07002539 gro_cells_init(&vxlan->gro_cells, dev);
2540
stephen hemmingerd3428942012-10-01 12:32:35 +00002541 for (h = 0; h < FDB_HASH_SIZE; ++h)
2542 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2543}
2544
2545static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2546 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
stephen hemminger5d174dd2013-04-27 11:31:55 +00002547 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
Cong Wange4c7ed42013-08-31 13:44:33 +08002548 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
stephen hemmingerd3428942012-10-01 12:32:35 +00002549 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2550 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
Cong Wange4c7ed42013-08-31 13:44:33 +08002551 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
stephen hemmingerd3428942012-10-01 12:32:35 +00002552 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2553 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2554 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2555 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2556 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
stephen hemminger05f47d62012-10-09 20:35:50 +00002557 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
David Stevense4f67ad2012-11-20 02:50:14 +00002558 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2559 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2560 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2561 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
Alexei Starovoitovf8a9b1b2015-07-30 20:10:22 -07002562 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
stephen hemminger823aa872013-04-27 11:31:57 +00002563 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
Tom Herbert5c91ae02014-11-06 18:06:01 -08002564 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2565 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2566 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
Tom Herbertdfd86452015-01-12 17:00:38 -08002567 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2568 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
Thomas Graf35114942015-01-15 03:53:55 +01002569 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
Tom Herbert0ace2ca2015-02-10 16:30:32 -08002570 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
stephen hemmingerd3428942012-10-01 12:32:35 +00002571};
2572
2573static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[])
2574{
2575 if (tb[IFLA_ADDRESS]) {
2576 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2577 pr_debug("invalid link address (not ethernet)\n");
2578 return -EINVAL;
2579 }
2580
2581 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2582 pr_debug("invalid all zero ethernet address\n");
2583 return -EADDRNOTAVAIL;
2584 }
2585 }
2586
2587 if (!data)
2588 return -EINVAL;
2589
2590 if (data[IFLA_VXLAN_ID]) {
2591 __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2592 if (id >= VXLAN_VID_MASK)
2593 return -ERANGE;
2594 }
2595
stephen hemminger05f47d62012-10-09 20:35:50 +00002596 if (data[IFLA_VXLAN_PORT_RANGE]) {
2597 const struct ifla_vxlan_port_range *p
2598 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2599
2600 if (ntohs(p->high) < ntohs(p->low)) {
2601 pr_debug("port range %u .. %u not valid\n",
2602 ntohs(p->low), ntohs(p->high));
2603 return -EINVAL;
2604 }
2605 }
2606
stephen hemmingerd3428942012-10-01 12:32:35 +00002607 return 0;
2608}
2609
Yan Burman1b13c972013-01-29 23:43:07 +00002610static void vxlan_get_drvinfo(struct net_device *netdev,
2611 struct ethtool_drvinfo *drvinfo)
2612{
2613 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2614 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2615}
2616
2617static const struct ethtool_ops vxlan_ethtool_ops = {
2618 .get_drvinfo = vxlan_get_drvinfo,
2619 .get_link = ethtool_op_get_link,
2620};
2621
stephen hemminger553675f2013-05-16 11:35:20 +00002622static void vxlan_del_work(struct work_struct *work)
2623{
2624 struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work);
Andy Zhouacbf74a2014-09-16 17:31:18 -07002625 udp_tunnel_sock_release(vs->sock);
stephen hemminger553675f2013-05-16 11:35:20 +00002626 kfree_rcu(vs, rcu);
2627}
2628
Tom Herbert3ee64f32014-07-13 19:49:42 -07002629static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2630 __be16 port, u32 flags)
stephen hemminger553675f2013-05-16 11:35:20 +00002631{
Cong Wange4c7ed42013-08-31 13:44:33 +08002632 struct socket *sock;
Tom Herbert3ee64f32014-07-13 19:49:42 -07002633 struct udp_port_cfg udp_conf;
2634 int err;
Cong Wange4c7ed42013-08-31 13:44:33 +08002635
Tom Herbert3ee64f32014-07-13 19:49:42 -07002636 memset(&udp_conf, 0, sizeof(udp_conf));
2637
2638 if (ipv6) {
2639 udp_conf.family = AF_INET6;
Tom Herbert3ee64f32014-07-13 19:49:42 -07002640 udp_conf.use_udp6_rx_checksums =
Alexander Duyck3dc2b6a2014-11-24 20:08:38 -08002641 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
Jiri Benca43a9ef2015-08-28 20:48:22 +02002642 udp_conf.ipv6_v6only = 1;
Tom Herbert3ee64f32014-07-13 19:49:42 -07002643 } else {
2644 udp_conf.family = AF_INET;
Cong Wange4c7ed42013-08-31 13:44:33 +08002645 }
2646
Tom Herbert3ee64f32014-07-13 19:49:42 -07002647 udp_conf.local_udp_port = port;
Cong Wange4c7ed42013-08-31 13:44:33 +08002648
Tom Herbert3ee64f32014-07-13 19:49:42 -07002649 /* Open UDP socket */
2650 err = udp_sock_create(net, &udp_conf, &sock);
2651 if (err < 0)
2652 return ERR_PTR(err);
Cong Wange4c7ed42013-08-31 13:44:33 +08002653
Zhi Yong Wu39deb2c2013-10-28 14:01:48 +08002654 return sock;
Cong Wange4c7ed42013-08-31 13:44:33 +08002655}
2656
2657/* Create new listen socket if needed */
Jiri Bencb1be00a2015-09-24 13:50:02 +02002658static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2659 __be16 port, u32 flags)
Cong Wange4c7ed42013-08-31 13:44:33 +08002660{
2661 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2662 struct vxlan_sock *vs;
2663 struct socket *sock;
Cong Wang31fec5a2013-05-27 22:35:52 +00002664 unsigned int h;
Andy Zhouacbf74a2014-09-16 17:31:18 -07002665 struct udp_tunnel_sock_cfg tunnel_cfg;
stephen hemminger553675f2013-05-16 11:35:20 +00002666
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02002667 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
Cong Wange4c7ed42013-08-31 13:44:33 +08002668 if (!vs)
stephen hemminger553675f2013-05-16 11:35:20 +00002669 return ERR_PTR(-ENOMEM);
2670
2671 for (h = 0; h < VNI_HASH_SIZE; ++h)
2672 INIT_HLIST_HEAD(&vs->vni_list[h]);
2673
2674 INIT_WORK(&vs->del_work, vxlan_del_work);
2675
Tom Herbert3ee64f32014-07-13 19:49:42 -07002676 sock = vxlan_create_sock(net, ipv6, port, flags);
Zhi Yong Wu39deb2c2013-10-28 14:01:48 +08002677 if (IS_ERR(sock)) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002678 pr_info("Cannot bind port %d, err=%ld\n", ntohs(port),
2679 PTR_ERR(sock));
stephen hemminger553675f2013-05-16 11:35:20 +00002680 kfree(vs);
Duan Jionge50fddc2013-11-01 13:09:43 +08002681 return ERR_CAST(sock);
stephen hemminger553675f2013-05-16 11:35:20 +00002682 }
2683
Cong Wange4c7ed42013-08-31 13:44:33 +08002684 vs->sock = sock;
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002685 atomic_set(&vs->refcnt, 1);
Tom Herbertaf33c1a2015-01-20 11:23:05 -08002686 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
stephen hemminger553675f2013-05-16 11:35:20 +00002687
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02002688 /* Initialize the vxlan udp offloads structure */
2689 vs->udp_offloads.port = port;
2690 vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive;
2691 vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete;
2692
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002693 spin_lock(&vn->sock_lock);
2694 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
Or Gerlitzdc01e7d2014-01-20 13:59:21 +02002695 vxlan_notify_add_rx_port(vs);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002696 spin_unlock(&vn->sock_lock);
stephen hemminger553675f2013-05-16 11:35:20 +00002697
2698 /* Mark socket as an encapsulation socket. */
Andy Zhouacbf74a2014-09-16 17:31:18 -07002699 tunnel_cfg.sk_user_data = vs;
2700 tunnel_cfg.encap_type = 1;
2701 tunnel_cfg.encap_rcv = vxlan_udp_encap_recv;
2702 tunnel_cfg.encap_destroy = NULL;
2703
2704 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
Cong Wange4c7ed42013-08-31 13:44:33 +08002705
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002706 return vs;
2707}
stephen hemminger553675f2013-05-16 11:35:20 +00002708
Jiri Bencb1be00a2015-09-24 13:50:02 +02002709static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002710{
Jiri Benc205f3562015-09-24 13:50:01 +02002711 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2712 struct vxlan_sock *vs = NULL;
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07002713
Jiri Benc205f3562015-09-24 13:50:01 +02002714 if (!vxlan->cfg.no_share) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002715 spin_lock(&vn->sock_lock);
Jiri Benc205f3562015-09-24 13:50:01 +02002716 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
2717 vxlan->cfg.dst_port, vxlan->flags);
2718 if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) {
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002719 spin_unlock(&vn->sock_lock);
Jiri Benc205f3562015-09-24 13:50:01 +02002720 return -EBUSY;
Marcelo Ricardo Leitner56ef9c92015-03-18 14:50:44 -03002721 }
2722 spin_unlock(&vn->sock_lock);
2723 }
Jiri Benc205f3562015-09-24 13:50:01 +02002724 if (!vs)
Jiri Bencb1be00a2015-09-24 13:50:02 +02002725 vs = vxlan_socket_create(vxlan->net, ipv6,
2726 vxlan->cfg.dst_port, vxlan->flags);
Jiri Benc205f3562015-09-24 13:50:01 +02002727 if (IS_ERR(vs))
2728 return PTR_ERR(vs);
Jiri Bencb1be00a2015-09-24 13:50:02 +02002729#if IS_ENABLED(CONFIG_IPV6)
2730 if (ipv6)
2731 vxlan->vn6_sock = vs;
2732 else
2733#endif
2734 vxlan->vn4_sock = vs;
Jiri Benc205f3562015-09-24 13:50:01 +02002735 vxlan_vs_add_dev(vs, vxlan);
2736 return 0;
stephen hemminger553675f2013-05-16 11:35:20 +00002737}
2738
Jiri Bencb1be00a2015-09-24 13:50:02 +02002739static int vxlan_sock_add(struct vxlan_dev *vxlan)
2740{
2741 bool ipv6 = vxlan->flags & VXLAN_F_IPV6;
2742 bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA;
2743 int ret = 0;
2744
2745 vxlan->vn4_sock = NULL;
2746#if IS_ENABLED(CONFIG_IPV6)
2747 vxlan->vn6_sock = NULL;
2748 if (ipv6 || metadata)
2749 ret = __vxlan_sock_add(vxlan, true);
2750#endif
2751 if (!ret && (!ipv6 || metadata))
2752 ret = __vxlan_sock_add(vxlan, false);
2753 if (ret < 0)
2754 vxlan_sock_release(vxlan);
2755 return ret;
2756}
2757
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002758static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
2759 struct vxlan_config *conf)
stephen hemmingerd3428942012-10-01 12:32:35 +00002760{
Nicolas Dichtel33564bb2015-01-26 22:28:14 +01002761 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
Nicolas Dichtel07b9b372016-01-07 11:26:53 +01002762 struct vxlan_dev *vxlan = netdev_priv(dev), *tmp;
Atzm Watanabec7995c42013-04-16 02:50:52 +00002763 struct vxlan_rdst *dst = &vxlan->default_dst;
Jiri Bencb1be00a2015-09-24 13:50:02 +02002764 unsigned short needed_headroom = ETH_HLEN;
stephen hemmingerd3428942012-10-01 12:32:35 +00002765 int err;
Cong Wange4c7ed42013-08-31 13:44:33 +08002766 bool use_ipv6 = false;
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002767 __be16 default_port = vxlan->cfg.dst_port;
stephen hemmingerd3428942012-10-01 12:32:35 +00002768
Nicolas Dichtel33564bb2015-01-26 22:28:14 +01002769 vxlan->net = src_net;
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02002770
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002771 dst->remote_vni = conf->vni;
2772
2773 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
stephen hemmingerd3428942012-10-01 12:32:35 +00002774
Mike Rapoport5933a7b2014-04-01 09:23:01 +03002775 /* Unless IPv6 is explicitly requested, assume IPv4 */
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002776 if (!dst->remote_ip.sa.sa_family)
2777 dst->remote_ip.sa.sa_family = AF_INET;
stephen hemmingerd3428942012-10-01 12:32:35 +00002778
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002779 if (dst->remote_ip.sa.sa_family == AF_INET6 ||
Jiri Benc057ba292015-09-17 16:11:11 +02002780 vxlan->cfg.saddr.sa.sa_family == AF_INET6) {
2781 if (!IS_ENABLED(CONFIG_IPV6))
2782 return -EPFNOSUPPORT;
Cong Wange4c7ed42013-08-31 13:44:33 +08002783 use_ipv6 = true;
Jiri Bencb1be00a2015-09-24 13:50:02 +02002784 vxlan->flags |= VXLAN_F_IPV6;
Jiri Benc057ba292015-09-17 16:11:11 +02002785 }
Cong Wange4c7ed42013-08-31 13:44:33 +08002786
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002787 if (conf->remote_ifindex) {
stephen hemminger34e02aa2012-10-09 20:35:53 +00002788 struct net_device *lowerdev
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002789 = __dev_get_by_index(src_net, conf->remote_ifindex);
2790
2791 dst->remote_ifindex = conf->remote_ifindex;
stephen hemmingerd3428942012-10-01 12:32:35 +00002792
stephen hemminger34e02aa2012-10-09 20:35:53 +00002793 if (!lowerdev) {
Atzm Watanabec7995c42013-04-16 02:50:52 +00002794 pr_info("ifindex %d does not exist\n", dst->remote_ifindex);
stephen hemminger34e02aa2012-10-09 20:35:53 +00002795 return -ENODEV;
stephen hemmingerd3428942012-10-01 12:32:35 +00002796 }
stephen hemminger34e02aa2012-10-09 20:35:53 +00002797
Cong Wange4c7ed42013-08-31 13:44:33 +08002798#if IS_ENABLED(CONFIG_IPV6)
2799 if (use_ipv6) {
2800 struct inet6_dev *idev = __in6_dev_get(lowerdev);
2801 if (idev && idev->cnf.disable_ipv6) {
2802 pr_info("IPv6 is disabled via sysctl\n");
2803 return -EPERM;
2804 }
Cong Wange4c7ed42013-08-31 13:44:33 +08002805 }
2806#endif
2807
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002808 if (!conf->mtu)
Cong Wange4c7ed42013-08-31 13:44:33 +08002809 dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
Alexander Duyck1ba56fb2012-11-13 13:10:59 +00002810
Jiri Bencb1be00a2015-09-24 13:50:02 +02002811 needed_headroom = lowerdev->hard_header_len;
Jiri Benc9dc2ad12015-09-17 16:11:10 +02002812 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002813
Jiri Bencb1be00a2015-09-24 13:50:02 +02002814 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
2815 needed_headroom += VXLAN6_HEADROOM;
2816 else
2817 needed_headroom += VXLAN_HEADROOM;
2818 dev->needed_headroom = needed_headroom;
2819
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002820 memcpy(&vxlan->cfg, conf, sizeof(*conf));
2821 if (!vxlan->cfg.dst_port)
2822 vxlan->cfg.dst_port = default_port;
2823 vxlan->flags |= conf->flags;
stephen hemmingerd3428942012-10-01 12:32:35 +00002824
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002825 if (!vxlan->cfg.age_interval)
2826 vxlan->cfg.age_interval = FDB_AGE_DEFAULT;
Vincent Bernatafb97182012-10-30 10:27:16 +00002827
Nicolas Dichtel07b9b372016-01-07 11:26:53 +01002828 list_for_each_entry(tmp, &vn->vxlan_list, next) {
2829 if (tmp->cfg.vni == conf->vni &&
2830 (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 ||
2831 tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 &&
2832 tmp->cfg.dst_port == vxlan->cfg.dst_port &&
2833 (tmp->flags & VXLAN_F_RCV_FLAGS) ==
2834 (vxlan->flags & VXLAN_F_RCV_FLAGS))
stephen hemminger553675f2013-05-16 11:35:20 +00002835 return -EEXIST;
Nicolas Dichtel07b9b372016-01-07 11:26:53 +01002836 }
stephen hemminger553675f2013-05-16 11:35:20 +00002837
Wilfried Klaebe7ad24ea2014-05-11 00:12:32 +00002838 dev->ethtool_ops = &vxlan_ethtool_ops;
Yan Burman1b13c972013-01-29 23:43:07 +00002839
Sridhar Samudrala2936b6a2013-09-17 12:12:40 -07002840 /* create an fdb entry for a valid default destination */
2841 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
2842 err = vxlan_fdb_create(vxlan, all_zeros_mac,
2843 &vxlan->default_dst.remote_ip,
2844 NUD_REACHABLE|NUD_PERMANENT,
2845 NLM_F_EXCL|NLM_F_CREATE,
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002846 vxlan->cfg.dst_port,
Sridhar Samudrala2936b6a2013-09-17 12:12:40 -07002847 vxlan->default_dst.remote_vni,
2848 vxlan->default_dst.remote_ifindex,
2849 NTF_SELF);
2850 if (err)
2851 return err;
2852 }
stephen hemmingerd3428942012-10-01 12:32:35 +00002853
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002854 err = register_netdevice(dev);
2855 if (err) {
Stephen Hemmingerba609e92013-06-25 17:06:01 -07002856 vxlan_fdb_delete_default(vxlan);
Mike Rapoportafbd8ba2013-06-25 16:01:51 +03002857 return err;
2858 }
2859
stephen hemminger553675f2013-05-16 11:35:20 +00002860 list_add(&vxlan->next, &vn->vxlan_list);
stephen hemminger553675f2013-05-16 11:35:20 +00002861
2862 return 0;
stephen hemmingerd3428942012-10-01 12:32:35 +00002863}
2864
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002865struct net_device *vxlan_dev_create(struct net *net, const char *name,
2866 u8 name_assign_type, struct vxlan_config *conf)
2867{
2868 struct nlattr *tb[IFLA_MAX+1];
2869 struct net_device *dev;
2870 int err;
2871
2872 memset(&tb, 0, sizeof(tb));
2873
2874 dev = rtnl_create_link(net, name, name_assign_type,
2875 &vxlan_link_ops, tb);
2876 if (IS_ERR(dev))
2877 return dev;
2878
2879 err = vxlan_dev_configure(net, dev, conf);
2880 if (err < 0) {
2881 free_netdev(dev);
2882 return ERR_PTR(err);
2883 }
2884
2885 return dev;
2886}
2887EXPORT_SYMBOL_GPL(vxlan_dev_create);
2888
2889static int vxlan_newlink(struct net *src_net, struct net_device *dev,
2890 struct nlattr *tb[], struct nlattr *data[])
2891{
2892 struct vxlan_config conf;
2893 int err;
2894
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002895 memset(&conf, 0, sizeof(conf));
Jesse Grosse277de52015-10-16 16:36:00 -07002896
2897 if (data[IFLA_VXLAN_ID])
2898 conf.vni = nla_get_u32(data[IFLA_VXLAN_ID]);
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002899
2900 if (data[IFLA_VXLAN_GROUP]) {
2901 conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
2902 } else if (data[IFLA_VXLAN_GROUP6]) {
2903 if (!IS_ENABLED(CONFIG_IPV6))
2904 return -EPFNOSUPPORT;
2905
2906 conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
2907 conf.remote_ip.sa.sa_family = AF_INET6;
2908 }
2909
2910 if (data[IFLA_VXLAN_LOCAL]) {
2911 conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
2912 conf.saddr.sa.sa_family = AF_INET;
2913 } else if (data[IFLA_VXLAN_LOCAL6]) {
2914 if (!IS_ENABLED(CONFIG_IPV6))
2915 return -EPFNOSUPPORT;
2916
2917 /* TODO: respect scope id */
2918 conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
2919 conf.saddr.sa.sa_family = AF_INET6;
2920 }
2921
2922 if (data[IFLA_VXLAN_LINK])
2923 conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
2924
2925 if (data[IFLA_VXLAN_TOS])
2926 conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
2927
2928 if (data[IFLA_VXLAN_TTL])
2929 conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
2930
2931 if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING]))
2932 conf.flags |= VXLAN_F_LEARN;
2933
2934 if (data[IFLA_VXLAN_AGEING])
2935 conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
2936
2937 if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY]))
2938 conf.flags |= VXLAN_F_PROXY;
2939
2940 if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC]))
2941 conf.flags |= VXLAN_F_RSC;
2942
2943 if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS]))
2944 conf.flags |= VXLAN_F_L2MISS;
2945
2946 if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS]))
2947 conf.flags |= VXLAN_F_L3MISS;
2948
2949 if (data[IFLA_VXLAN_LIMIT])
2950 conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
2951
Alexei Starovoitovf8a9b1b2015-07-30 20:10:22 -07002952 if (data[IFLA_VXLAN_COLLECT_METADATA] &&
2953 nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
2954 conf.flags |= VXLAN_F_COLLECT_METADATA;
2955
Thomas Graf0dfbdf42015-07-21 10:44:02 +02002956 if (data[IFLA_VXLAN_PORT_RANGE]) {
2957 const struct ifla_vxlan_port_range *p
2958 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2959 conf.port_min = ntohs(p->low);
2960 conf.port_max = ntohs(p->high);
2961 }
2962
2963 if (data[IFLA_VXLAN_PORT])
2964 conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
2965
2966 if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
2967 conf.flags |= VXLAN_F_UDP_CSUM;
2968
2969 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] &&
2970 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
2971 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
2972
2973 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] &&
2974 nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
2975 conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
2976
2977 if (data[IFLA_VXLAN_REMCSUM_TX] &&
2978 nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
2979 conf.flags |= VXLAN_F_REMCSUM_TX;
2980
2981 if (data[IFLA_VXLAN_REMCSUM_RX] &&
2982 nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
2983 conf.flags |= VXLAN_F_REMCSUM_RX;
2984
2985 if (data[IFLA_VXLAN_GBP])
2986 conf.flags |= VXLAN_F_GBP;
2987
2988 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL])
2989 conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL;
2990
2991 err = vxlan_dev_configure(src_net, dev, &conf);
2992 switch (err) {
2993 case -ENODEV:
2994 pr_info("ifindex %d does not exist\n", conf.remote_ifindex);
2995 break;
2996
2997 case -EPERM:
2998 pr_info("IPv6 is disabled via sysctl\n");
2999 break;
3000
3001 case -EEXIST:
3002 pr_info("duplicate VNI %u\n", conf.vni);
3003 break;
3004 }
3005
3006 return err;
3007}
3008
stephen hemmingerd3428942012-10-01 12:32:35 +00003009static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3010{
3011 struct vxlan_dev *vxlan = netdev_priv(dev);
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003012 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
stephen hemmingerd3428942012-10-01 12:32:35 +00003013
stephen hemmingerfe5c3562013-07-13 10:18:18 -07003014 spin_lock(&vn->sock_lock);
Pravin B Shelar9c2e24e2013-08-19 11:22:48 -07003015 if (!hlist_unhashed(&vxlan->hlist))
3016 hlist_del_rcu(&vxlan->hlist);
stephen hemmingerfe5c3562013-07-13 10:18:18 -07003017 spin_unlock(&vn->sock_lock);
3018
Tom Herbert58ce31c2015-08-19 17:07:33 -07003019 gro_cells_destroy(&vxlan->gro_cells);
stephen hemminger553675f2013-05-16 11:35:20 +00003020 list_del(&vxlan->next);
stephen hemmingerd3428942012-10-01 12:32:35 +00003021 unregister_netdevice_queue(dev, head);
3022}
3023
3024static size_t vxlan_get_size(const struct net_device *dev)
3025{
3026
3027 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
Cong Wange4c7ed42013-08-31 13:44:33 +08003028 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
stephen hemmingerd3428942012-10-01 12:32:35 +00003029 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
Cong Wange4c7ed42013-08-31 13:44:33 +08003030 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
stephen hemmingerd3428942012-10-01 12:32:35 +00003031 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3032 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3033 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
David Stevense4f67ad2012-11-20 02:50:14 +00003034 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3035 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3036 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3037 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07003038 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
stephen hemmingerd3428942012-10-01 12:32:35 +00003039 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3040 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
stephen hemminger05f47d62012-10-09 20:35:50 +00003041 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
Tom Herbert359a0ea2014-06-04 17:20:29 -07003042 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3043 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3044 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3045 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
Tom Herbertdfd86452015-01-12 17:00:38 -08003046 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3047 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
stephen hemmingerd3428942012-10-01 12:32:35 +00003048 0;
3049}
3050
3051static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3052{
3053 const struct vxlan_dev *vxlan = netdev_priv(dev);
Atzm Watanabec7995c42013-04-16 02:50:52 +00003054 const struct vxlan_rdst *dst = &vxlan->default_dst;
stephen hemminger05f47d62012-10-09 20:35:50 +00003055 struct ifla_vxlan_port_range ports = {
Thomas Graf0dfbdf42015-07-21 10:44:02 +02003056 .low = htons(vxlan->cfg.port_min),
3057 .high = htons(vxlan->cfg.port_max),
stephen hemminger05f47d62012-10-09 20:35:50 +00003058 };
stephen hemmingerd3428942012-10-01 12:32:35 +00003059
Atzm Watanabec7995c42013-04-16 02:50:52 +00003060 if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni))
stephen hemmingerd3428942012-10-01 12:32:35 +00003061 goto nla_put_failure;
3062
Cong Wange4c7ed42013-08-31 13:44:33 +08003063 if (!vxlan_addr_any(&dst->remote_ip)) {
3064 if (dst->remote_ip.sa.sa_family == AF_INET) {
Jiri Benc930345e2015-03-29 16:59:25 +02003065 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3066 dst->remote_ip.sin.sin_addr.s_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08003067 goto nla_put_failure;
3068#if IS_ENABLED(CONFIG_IPV6)
3069 } else {
Jiri Benc930345e2015-03-29 16:59:25 +02003070 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3071 &dst->remote_ip.sin6.sin6_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08003072 goto nla_put_failure;
3073#endif
3074 }
3075 }
stephen hemmingerd3428942012-10-01 12:32:35 +00003076
Atzm Watanabec7995c42013-04-16 02:50:52 +00003077 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
stephen hemmingerd3428942012-10-01 12:32:35 +00003078 goto nla_put_failure;
3079
Thomas Graf0dfbdf42015-07-21 10:44:02 +02003080 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3081 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
Jiri Benc930345e2015-03-29 16:59:25 +02003082 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
Thomas Graf0dfbdf42015-07-21 10:44:02 +02003083 vxlan->cfg.saddr.sin.sin_addr.s_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08003084 goto nla_put_failure;
3085#if IS_ENABLED(CONFIG_IPV6)
3086 } else {
Jiri Benc930345e2015-03-29 16:59:25 +02003087 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
Thomas Graf0dfbdf42015-07-21 10:44:02 +02003088 &vxlan->cfg.saddr.sin6.sin6_addr))
Cong Wange4c7ed42013-08-31 13:44:33 +08003089 goto nla_put_failure;
3090#endif
3091 }
3092 }
stephen hemmingerd3428942012-10-01 12:32:35 +00003093
Thomas Graf0dfbdf42015-07-21 10:44:02 +02003094 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3095 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
David Stevense4f67ad2012-11-20 02:50:14 +00003096 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3097 !!(vxlan->flags & VXLAN_F_LEARN)) ||
3098 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3099 !!(vxlan->flags & VXLAN_F_PROXY)) ||
3100 nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) ||
3101 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3102 !!(vxlan->flags & VXLAN_F_L2MISS)) ||
3103 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3104 !!(vxlan->flags & VXLAN_F_L3MISS)) ||
Alexei Starovoitovda8b43c2015-08-04 22:51:07 -07003105 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3106 !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) ||
Thomas Graf0dfbdf42015-07-21 10:44:02 +02003107 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3108 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3109 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
Tom Herbert359a0ea2014-06-04 17:20:29 -07003110 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3111 !!(vxlan->flags & VXLAN_F_UDP_CSUM)) ||
3112 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3113 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3114 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
Tom Herbertdfd86452015-01-12 17:00:38 -08003115 !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3116 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3117 !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) ||
3118 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3119 !!(vxlan->flags & VXLAN_F_REMCSUM_RX)))
stephen hemmingerd3428942012-10-01 12:32:35 +00003120 goto nla_put_failure;
3121
stephen hemminger05f47d62012-10-09 20:35:50 +00003122 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3123 goto nla_put_failure;
3124
Thomas Graf35114942015-01-15 03:53:55 +01003125 if (vxlan->flags & VXLAN_F_GBP &&
3126 nla_put_flag(skb, IFLA_VXLAN_GBP))
3127 goto nla_put_failure;
3128
Tom Herbert0ace2ca2015-02-10 16:30:32 -08003129 if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3130 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3131 goto nla_put_failure;
3132
stephen hemmingerd3428942012-10-01 12:32:35 +00003133 return 0;
3134
3135nla_put_failure:
3136 return -EMSGSIZE;
3137}
3138
Nicolas Dichtel1728d4f2015-01-15 15:11:17 +01003139static struct net *vxlan_get_link_net(const struct net_device *dev)
3140{
3141 struct vxlan_dev *vxlan = netdev_priv(dev);
3142
3143 return vxlan->net;
3144}
3145
stephen hemmingerd3428942012-10-01 12:32:35 +00003146static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3147 .kind = "vxlan",
3148 .maxtype = IFLA_VXLAN_MAX,
3149 .policy = vxlan_policy,
3150 .priv_size = sizeof(struct vxlan_dev),
3151 .setup = vxlan_setup,
3152 .validate = vxlan_validate,
3153 .newlink = vxlan_newlink,
3154 .dellink = vxlan_dellink,
3155 .get_size = vxlan_get_size,
3156 .fill_info = vxlan_fill_info,
Nicolas Dichtel1728d4f2015-01-15 15:11:17 +01003157 .get_link_net = vxlan_get_link_net,
stephen hemmingerd3428942012-10-01 12:32:35 +00003158};
3159
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003160static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3161 struct net_device *dev)
3162{
3163 struct vxlan_dev *vxlan, *next;
3164 LIST_HEAD(list_kill);
3165
3166 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3167 struct vxlan_rdst *dst = &vxlan->default_dst;
3168
3169 /* In case we created vxlan device with carrier
3170 * and we loose the carrier due to module unload
3171 * we also need to remove vxlan device. In other
3172 * cases, it's not necessary and remote_ifindex
3173 * is 0 here, so no matches.
3174 */
3175 if (dst->remote_ifindex == dev->ifindex)
3176 vxlan_dellink(vxlan->dev, &list_kill);
3177 }
3178
3179 unregister_netdevice_many(&list_kill);
3180}
3181
3182static int vxlan_lowerdev_event(struct notifier_block *unused,
3183 unsigned long event, void *ptr)
3184{
3185 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
Daniel Borkmann783c1462014-01-22 21:07:53 +01003186 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003187
Daniel Borkmann783c1462014-01-22 21:07:53 +01003188 if (event == NETDEV_UNREGISTER)
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003189 vxlan_handle_lowerdev_unregister(vn, dev);
3190
3191 return NOTIFY_DONE;
3192}
3193
3194static struct notifier_block vxlan_notifier_block __read_mostly = {
3195 .notifier_call = vxlan_lowerdev_event,
3196};
3197
stephen hemmingerd3428942012-10-01 12:32:35 +00003198static __net_init int vxlan_init_net(struct net *net)
3199{
3200 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
Cong Wang31fec5a2013-05-27 22:35:52 +00003201 unsigned int h;
stephen hemmingerd3428942012-10-01 12:32:35 +00003202
stephen hemminger553675f2013-05-16 11:35:20 +00003203 INIT_LIST_HEAD(&vn->vxlan_list);
Stephen Hemminger1c51a912013-06-17 14:16:11 -07003204 spin_lock_init(&vn->sock_lock);
stephen hemmingerd3428942012-10-01 12:32:35 +00003205
stephen hemminger553675f2013-05-16 11:35:20 +00003206 for (h = 0; h < PORT_HASH_SIZE; ++h)
3207 INIT_HLIST_HEAD(&vn->sock_list[h]);
stephen hemmingerd3428942012-10-01 12:32:35 +00003208
3209 return 0;
3210}
3211
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003212static void __net_exit vxlan_exit_net(struct net *net)
3213{
3214 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3215 struct vxlan_dev *vxlan, *next;
3216 struct net_device *dev, *aux;
3217 LIST_HEAD(list);
3218
3219 rtnl_lock();
3220 for_each_netdev_safe(net, dev, aux)
3221 if (dev->rtnl_link_ops == &vxlan_link_ops)
3222 unregister_netdevice_queue(dev, &list);
3223
3224 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3225 /* If vxlan->dev is in the same netns, it has already been added
3226 * to the list by the previous loop.
3227 */
Tom Herbert58ce31c2015-08-19 17:07:33 -07003228 if (!net_eq(dev_net(vxlan->dev), net)) {
3229 gro_cells_destroy(&vxlan->gro_cells);
John W. Linville13c3ed62015-05-18 13:51:24 -04003230 unregister_netdevice_queue(vxlan->dev, &list);
Tom Herbert58ce31c2015-08-19 17:07:33 -07003231 }
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003232 }
3233
3234 unregister_netdevice_many(&list);
3235 rtnl_unlock();
3236}
3237
stephen hemmingerd3428942012-10-01 12:32:35 +00003238static struct pernet_operations vxlan_net_ops = {
3239 .init = vxlan_init_net,
Nicolas Dichtelf01ec1c2014-04-24 10:02:49 +02003240 .exit = vxlan_exit_net,
stephen hemmingerd3428942012-10-01 12:32:35 +00003241 .id = &vxlan_net_id,
3242 .size = sizeof(struct vxlan_net),
3243};
3244
3245static int __init vxlan_init_module(void)
3246{
3247 int rc;
3248
Stephen Hemminger758c57d2013-06-17 14:16:09 -07003249 vxlan_wq = alloc_workqueue("vxlan", 0, 0);
3250 if (!vxlan_wq)
3251 return -ENOMEM;
3252
stephen hemmingerd3428942012-10-01 12:32:35 +00003253 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3254
Daniel Borkmann783c1462014-01-22 21:07:53 +01003255 rc = register_pernet_subsys(&vxlan_net_ops);
stephen hemmingerd3428942012-10-01 12:32:35 +00003256 if (rc)
3257 goto out1;
3258
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003259 rc = register_netdevice_notifier(&vxlan_notifier_block);
stephen hemmingerd3428942012-10-01 12:32:35 +00003260 if (rc)
3261 goto out2;
3262
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003263 rc = rtnl_link_register(&vxlan_link_ops);
3264 if (rc)
3265 goto out3;
stephen hemmingerd3428942012-10-01 12:32:35 +00003266
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003267 return 0;
3268out3:
3269 unregister_netdevice_notifier(&vxlan_notifier_block);
stephen hemmingerd3428942012-10-01 12:32:35 +00003270out2:
Daniel Borkmann783c1462014-01-22 21:07:53 +01003271 unregister_pernet_subsys(&vxlan_net_ops);
stephen hemmingerd3428942012-10-01 12:32:35 +00003272out1:
Stephen Hemminger758c57d2013-06-17 14:16:09 -07003273 destroy_workqueue(vxlan_wq);
stephen hemmingerd3428942012-10-01 12:32:35 +00003274 return rc;
3275}
Cong Wang7332a132013-05-27 22:35:53 +00003276late_initcall(vxlan_init_module);
stephen hemmingerd3428942012-10-01 12:32:35 +00003277
3278static void __exit vxlan_cleanup_module(void)
3279{
Stephen Hemmingerb7153982013-06-17 14:16:09 -07003280 rtnl_link_unregister(&vxlan_link_ops);
Daniel Borkmannacaf4e7092014-01-13 18:41:19 +01003281 unregister_netdevice_notifier(&vxlan_notifier_block);
Stephen Hemminger758c57d2013-06-17 14:16:09 -07003282 destroy_workqueue(vxlan_wq);
Daniel Borkmann783c1462014-01-22 21:07:53 +01003283 unregister_pernet_subsys(&vxlan_net_ops);
3284 /* rcu_barrier() is called by netns */
stephen hemmingerd3428942012-10-01 12:32:35 +00003285}
3286module_exit(vxlan_cleanup_module);
3287
3288MODULE_LICENSE("GPL");
3289MODULE_VERSION(VXLAN_VERSION);
stephen hemminger3b8df3c2013-04-27 11:31:52 +00003290MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
Jesse Brandeburgead51392014-01-17 11:00:33 -08003291MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
stephen hemmingerd3428942012-10-01 12:32:35 +00003292MODULE_ALIAS_RTNL_LINK("vxlan");