Commit | Line | Data |
---|---|---|
d342894c | 1 | /* |
eb5ce439 | 2 | * VXLAN: Virtual eXtensible Local Area Network |
d342894c | 3 | * |
3b8df3c6 | 4 | * Copyright (c) 2012-2013 Vyatta Inc. |
d342894c | 5 | * |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License version 2 as | |
8 | * published by the Free Software Foundation. | |
d342894c | 9 | */ |
10 | ||
11 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
12 | ||
13 | #include <linux/kernel.h> | |
14 | #include <linux/types.h> | |
15 | #include <linux/module.h> | |
16 | #include <linux/errno.h> | |
17 | #include <linux/slab.h> | |
18 | #include <linux/skbuff.h> | |
19 | #include <linux/rculist.h> | |
20 | #include <linux/netdevice.h> | |
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> | |
1eaa8178 | 27 | #include <linux/if_vlan.h> |
d342894c | 28 | #include <linux/hash.h> |
1b13c97f | 29 | #include <linux/ethtool.h> |
e4f67add DS |
30 | #include <net/arp.h> |
31 | #include <net/ndisc.h> | |
d342894c | 32 | #include <net/ip.h> |
c5441932 | 33 | #include <net/ip_tunnels.h> |
d342894c | 34 | #include <net/icmp.h> |
35 | #include <net/udp.h> | |
36 | #include <net/rtnetlink.h> | |
37 | #include <net/route.h> | |
38 | #include <net/dsfield.h> | |
39 | #include <net/inet_ecn.h> | |
40 | #include <net/net_namespace.h> | |
41 | #include <net/netns/generic.h> | |
012a5729 | 42 | #include <net/vxlan.h> |
dc01e7d3 | 43 | #include <net/protocol.h> |
e4c7ed41 CW |
44 | #if IS_ENABLED(CONFIG_IPV6) |
45 | #include <net/ipv6.h> | |
46 | #include <net/addrconf.h> | |
47 | #include <net/ip6_tunnel.h> | |
660d98ca | 48 | #include <net/ip6_checksum.h> |
e4c7ed41 | 49 | #endif |
d342894c | 50 | |
51 | #define VXLAN_VERSION "0.1" | |
52 | ||
553675fb | 53 | #define PORT_HASH_BITS 8 |
54 | #define PORT_HASH_SIZE (1<<PORT_HASH_BITS) | |
d342894c | 55 | #define VNI_HASH_BITS 10 |
56 | #define VNI_HASH_SIZE (1<<VNI_HASH_BITS) | |
57 | #define FDB_HASH_BITS 8 | |
58 | #define FDB_HASH_SIZE (1<<FDB_HASH_BITS) | |
59 | #define FDB_AGE_DEFAULT 300 /* 5 min */ | |
60 | #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ | |
61 | ||
62 | #define VXLAN_N_VID (1u << 24) | |
63 | #define VXLAN_VID_MASK (VXLAN_N_VID - 1) | |
7ce04758 | 64 | #define VXLAN_HLEN (sizeof(struct udphdr) + sizeof(struct vxlanhdr)) |
d342894c | 65 | |
66 | #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */ | |
67 | ||
68 | /* VXLAN protocol header */ | |
69 | struct vxlanhdr { | |
70 | __be32 vx_flags; | |
71 | __be32 vx_vni; | |
72 | }; | |
73 | ||
23c578bf | 74 | /* UDP port for VXLAN traffic. |
75 | * The IANA assigned port is 4789, but the Linux default is 8472 | |
234f5b73 | 76 | * for compatibility with early adopters. |
23c578bf | 77 | */ |
9daaa397 SH |
78 | static unsigned short vxlan_port __read_mostly = 8472; |
79 | module_param_named(udp_port, vxlan_port, ushort, 0444); | |
d342894c | 80 | MODULE_PARM_DESC(udp_port, "Destination UDP port"); |
81 | ||
82 | static bool log_ecn_error = true; | |
83 | module_param(log_ecn_error, bool, 0644); | |
84 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); | |
85 | ||
60d9d4c6 | 86 | static int vxlan_net_id; |
553675fb | 87 | |
afbd8bae MR |
88 | static const u8 all_zeros_mac[ETH_ALEN]; |
89 | ||
553675fb | 90 | /* per-network namespace private data for this module */ |
91 | struct vxlan_net { | |
92 | struct list_head vxlan_list; | |
93 | struct hlist_head sock_list[PORT_HASH_SIZE]; | |
1c51a915 | 94 | spinlock_t sock_lock; |
553675fb | 95 | }; |
96 | ||
e4c7ed41 CW |
97 | union vxlan_addr { |
98 | struct sockaddr_in sin; | |
99 | struct sockaddr_in6 sin6; | |
100 | struct sockaddr sa; | |
101 | }; | |
102 | ||
6681712d | 103 | struct vxlan_rdst { |
e4c7ed41 | 104 | union vxlan_addr remote_ip; |
6681712d DS |
105 | __be16 remote_port; |
106 | u32 remote_vni; | |
107 | u32 remote_ifindex; | |
3e61aa8f | 108 | struct list_head list; |
bc7892ba | 109 | struct rcu_head rcu; |
6681712d DS |
110 | }; |
111 | ||
d342894c | 112 | /* Forwarding table entry */ |
113 | struct vxlan_fdb { | |
114 | struct hlist_node hlist; /* linked list of entries */ | |
115 | struct rcu_head rcu; | |
116 | unsigned long updated; /* jiffies */ | |
117 | unsigned long used; | |
3e61aa8f | 118 | struct list_head remotes; |
d342894c | 119 | u16 state; /* see ndm_state */ |
ae884082 | 120 | u8 flags; /* see ndm_flags */ |
d342894c | 121 | u8 eth_addr[ETH_ALEN]; |
122 | }; | |
123 | ||
d342894c | 124 | /* Pseudo network device */ |
125 | struct vxlan_dev { | |
553675fb | 126 | struct hlist_node hlist; /* vni hash table */ |
127 | struct list_head next; /* vxlan's per namespace list */ | |
128 | struct vxlan_sock *vn_sock; /* listening socket */ | |
d342894c | 129 | struct net_device *dev; |
c7995c43 | 130 | struct vxlan_rdst default_dst; /* default destination */ |
e4c7ed41 | 131 | union vxlan_addr saddr; /* source address */ |
823aa873 | 132 | __be16 dst_port; |
05f47d69 | 133 | __u16 port_min; /* source port range */ |
134 | __u16 port_max; | |
d342894c | 135 | __u8 tos; /* TOS override */ |
136 | __u8 ttl; | |
e4f67add | 137 | u32 flags; /* VXLAN_F_* below */ |
d342894c | 138 | |
1c51a915 | 139 | struct work_struct sock_work; |
3fc2de2f | 140 | struct work_struct igmp_join; |
141 | struct work_struct igmp_leave; | |
1c51a915 | 142 | |
d342894c | 143 | unsigned long age_interval; |
144 | struct timer_list age_timer; | |
145 | spinlock_t hash_lock; | |
146 | unsigned int addrcnt; | |
147 | unsigned int addrmax; | |
d342894c | 148 | |
149 | struct hlist_head fdb_head[FDB_HASH_SIZE]; | |
150 | }; | |
151 | ||
e4f67add DS |
152 | #define VXLAN_F_LEARN 0x01 |
153 | #define VXLAN_F_PROXY 0x02 | |
154 | #define VXLAN_F_RSC 0x04 | |
155 | #define VXLAN_F_L2MISS 0x08 | |
156 | #define VXLAN_F_L3MISS 0x10 | |
e4c7ed41 | 157 | #define VXLAN_F_IPV6 0x20 /* internal flag */ |
e4f67add | 158 | |
d342894c | 159 | /* salt for hash table */ |
160 | static u32 vxlan_salt __read_mostly; | |
758c57d1 | 161 | static struct workqueue_struct *vxlan_wq; |
d342894c | 162 | |
1c51a915 SH |
163 | static void vxlan_sock_work(struct work_struct *work); |
164 | ||
e4c7ed41 CW |
165 | #if IS_ENABLED(CONFIG_IPV6) |
166 | static inline | |
167 | bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b) | |
168 | { | |
169 | if (a->sa.sa_family != b->sa.sa_family) | |
170 | return false; | |
171 | if (a->sa.sa_family == AF_INET6) | |
172 | return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr); | |
173 | else | |
174 | return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr; | |
175 | } | |
176 | ||
177 | static inline bool vxlan_addr_any(const union vxlan_addr *ipa) | |
178 | { | |
179 | if (ipa->sa.sa_family == AF_INET6) | |
180 | return ipv6_addr_any(&ipa->sin6.sin6_addr); | |
181 | else | |
182 | return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY); | |
183 | } | |
184 | ||
185 | static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa) | |
186 | { | |
187 | if (ipa->sa.sa_family == AF_INET6) | |
188 | return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr); | |
189 | else | |
190 | return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr)); | |
191 | } | |
192 | ||
193 | static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla) | |
194 | { | |
195 | if (nla_len(nla) >= sizeof(struct in6_addr)) { | |
196 | nla_memcpy(&ip->sin6.sin6_addr, nla, sizeof(struct in6_addr)); | |
197 | ip->sa.sa_family = AF_INET6; | |
198 | return 0; | |
199 | } else if (nla_len(nla) >= sizeof(__be32)) { | |
200 | ip->sin.sin_addr.s_addr = nla_get_be32(nla); | |
201 | ip->sa.sa_family = AF_INET; | |
202 | return 0; | |
203 | } else { | |
204 | return -EAFNOSUPPORT; | |
205 | } | |
206 | } | |
207 | ||
208 | static int vxlan_nla_put_addr(struct sk_buff *skb, int attr, | |
209 | const union vxlan_addr *ip) | |
210 | { | |
211 | if (ip->sa.sa_family == AF_INET6) | |
212 | return nla_put(skb, attr, sizeof(struct in6_addr), &ip->sin6.sin6_addr); | |
213 | else | |
214 | return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr); | |
215 | } | |
216 | ||
217 | #else /* !CONFIG_IPV6 */ | |
218 | ||
219 | static inline | |
220 | bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b) | |
221 | { | |
222 | return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr; | |
223 | } | |
224 | ||
225 | static inline bool vxlan_addr_any(const union vxlan_addr *ipa) | |
226 | { | |
227 | return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY); | |
228 | } | |
229 | ||
230 | static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa) | |
231 | { | |
232 | return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr)); | |
233 | } | |
234 | ||
235 | static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla) | |
236 | { | |
237 | if (nla_len(nla) >= sizeof(struct in6_addr)) { | |
238 | return -EAFNOSUPPORT; | |
239 | } else if (nla_len(nla) >= sizeof(__be32)) { | |
240 | ip->sin.sin_addr.s_addr = nla_get_be32(nla); | |
241 | ip->sa.sa_family = AF_INET; | |
242 | return 0; | |
243 | } else { | |
244 | return -EAFNOSUPPORT; | |
245 | } | |
246 | } | |
247 | ||
248 | static int vxlan_nla_put_addr(struct sk_buff *skb, int attr, | |
249 | const union vxlan_addr *ip) | |
250 | { | |
251 | return nla_put_be32(skb, attr, ip->sin.sin_addr.s_addr); | |
252 | } | |
253 | #endif | |
254 | ||
553675fb | 255 | /* Virtual Network hash table head */ |
256 | static inline struct hlist_head *vni_head(struct vxlan_sock *vs, u32 id) | |
257 | { | |
258 | return &vs->vni_list[hash_32(id, VNI_HASH_BITS)]; | |
259 | } | |
260 | ||
261 | /* Socket hash table head */ | |
262 | static inline struct hlist_head *vs_head(struct net *net, __be16 port) | |
d342894c | 263 | { |
264 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
265 | ||
553675fb | 266 | return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)]; |
267 | } | |
268 | ||
3e61aa8f SH |
269 | /* First remote destination for a forwarding entry. |
270 | * Guaranteed to be non-NULL because remotes are never deleted. | |
271 | */ | |
5ca5461c | 272 | static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb) |
3e61aa8f | 273 | { |
5ca5461c | 274 | return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list); |
275 | } | |
276 | ||
277 | static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb) | |
278 | { | |
279 | return list_first_entry(&fdb->remotes, struct vxlan_rdst, list); | |
3e61aa8f SH |
280 | } |
281 | ||
553675fb | 282 | /* Find VXLAN socket based on network namespace and UDP port */ |
9c2e24e1 | 283 | static struct vxlan_sock *vxlan_find_sock(struct net *net, __be16 port) |
553675fb | 284 | { |
285 | struct vxlan_sock *vs; | |
286 | ||
287 | hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { | |
288 | if (inet_sk(vs->sock->sk)->inet_sport == port) | |
289 | return vs; | |
290 | } | |
291 | return NULL; | |
d342894c | 292 | } |
293 | ||
5cfccc5a | 294 | static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, u32 id) |
d342894c | 295 | { |
296 | struct vxlan_dev *vxlan; | |
d342894c | 297 | |
553675fb | 298 | hlist_for_each_entry_rcu(vxlan, vni_head(vs, id), hlist) { |
c7995c43 | 299 | if (vxlan->default_dst.remote_vni == id) |
d342894c | 300 | return vxlan; |
301 | } | |
302 | ||
303 | return NULL; | |
304 | } | |
305 | ||
5cfccc5a PS |
306 | /* Look up VNI in a per net namespace table */ |
307 | static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id, __be16 port) | |
308 | { | |
309 | struct vxlan_sock *vs; | |
310 | ||
311 | vs = vxlan_find_sock(net, port); | |
312 | if (!vs) | |
313 | return NULL; | |
314 | ||
315 | return vxlan_vs_find_vni(vs, id); | |
316 | } | |
317 | ||
d342894c | 318 | /* Fill in neighbour message in skbuff. */ |
319 | static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, | |
234f5b73 SH |
320 | const struct vxlan_fdb *fdb, |
321 | u32 portid, u32 seq, int type, unsigned int flags, | |
322 | const struct vxlan_rdst *rdst) | |
d342894c | 323 | { |
324 | unsigned long now = jiffies; | |
325 | struct nda_cacheinfo ci; | |
326 | struct nlmsghdr *nlh; | |
327 | struct ndmsg *ndm; | |
e4f67add | 328 | bool send_ip, send_eth; |
d342894c | 329 | |
330 | nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); | |
331 | if (nlh == NULL) | |
332 | return -EMSGSIZE; | |
333 | ||
334 | ndm = nlmsg_data(nlh); | |
335 | memset(ndm, 0, sizeof(*ndm)); | |
e4f67add DS |
336 | |
337 | send_eth = send_ip = true; | |
338 | ||
339 | if (type == RTM_GETNEIGH) { | |
340 | ndm->ndm_family = AF_INET; | |
e4c7ed41 | 341 | send_ip = !vxlan_addr_any(&rdst->remote_ip); |
e4f67add DS |
342 | send_eth = !is_zero_ether_addr(fdb->eth_addr); |
343 | } else | |
344 | ndm->ndm_family = AF_BRIDGE; | |
d342894c | 345 | ndm->ndm_state = fdb->state; |
346 | ndm->ndm_ifindex = vxlan->dev->ifindex; | |
ae884082 | 347 | ndm->ndm_flags = fdb->flags; |
d342894c | 348 | ndm->ndm_type = NDA_DST; |
349 | ||
e4f67add | 350 | if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr)) |
d342894c | 351 | goto nla_put_failure; |
352 | ||
e4c7ed41 | 353 | if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip)) |
6681712d DS |
354 | goto nla_put_failure; |
355 | ||
823aa873 | 356 | if (rdst->remote_port && rdst->remote_port != vxlan->dst_port && |
6681712d DS |
357 | nla_put_be16(skb, NDA_PORT, rdst->remote_port)) |
358 | goto nla_put_failure; | |
c7995c43 | 359 | if (rdst->remote_vni != vxlan->default_dst.remote_vni && |
60d9d4c6 | 360 | nla_put_u32(skb, NDA_VNI, rdst->remote_vni)) |
6681712d DS |
361 | goto nla_put_failure; |
362 | if (rdst->remote_ifindex && | |
363 | nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) | |
d342894c | 364 | goto nla_put_failure; |
365 | ||
366 | ci.ndm_used = jiffies_to_clock_t(now - fdb->used); | |
367 | ci.ndm_confirmed = 0; | |
368 | ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated); | |
369 | ci.ndm_refcnt = 0; | |
370 | ||
371 | if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) | |
372 | goto nla_put_failure; | |
373 | ||
374 | return nlmsg_end(skb, nlh); | |
375 | ||
376 | nla_put_failure: | |
377 | nlmsg_cancel(skb, nlh); | |
378 | return -EMSGSIZE; | |
379 | } | |
380 | ||
381 | static inline size_t vxlan_nlmsg_size(void) | |
382 | { | |
383 | return NLMSG_ALIGN(sizeof(struct ndmsg)) | |
384 | + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ | |
e4c7ed41 | 385 | + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ |
73cf3317 | 386 | + nla_total_size(sizeof(__be16)) /* NDA_PORT */ |
6681712d DS |
387 | + nla_total_size(sizeof(__be32)) /* NDA_VNI */ |
388 | + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ | |
d342894c | 389 | + nla_total_size(sizeof(struct nda_cacheinfo)); |
390 | } | |
391 | ||
392 | static void vxlan_fdb_notify(struct vxlan_dev *vxlan, | |
3e61aa8f | 393 | struct vxlan_fdb *fdb, int type) |
d342894c | 394 | { |
395 | struct net *net = dev_net(vxlan->dev); | |
396 | struct sk_buff *skb; | |
397 | int err = -ENOBUFS; | |
398 | ||
399 | skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); | |
400 | if (skb == NULL) | |
401 | goto errout; | |
402 | ||
5ca5461c | 403 | err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, |
404 | first_remote_rtnl(fdb)); | |
d342894c | 405 | if (err < 0) { |
406 | /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ | |
407 | WARN_ON(err == -EMSGSIZE); | |
408 | kfree_skb(skb); | |
409 | goto errout; | |
410 | } | |
411 | ||
412 | rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); | |
413 | return; | |
414 | errout: | |
415 | if (err < 0) | |
416 | rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); | |
417 | } | |
418 | ||
e4c7ed41 | 419 | static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) |
e4f67add DS |
420 | { |
421 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
bb3fd687 SH |
422 | struct vxlan_fdb f = { |
423 | .state = NUD_STALE, | |
424 | }; | |
425 | struct vxlan_rdst remote = { | |
e4c7ed41 | 426 | .remote_ip = *ipa, /* goes to NDA_DST */ |
bb3fd687 SH |
427 | .remote_vni = VXLAN_N_VID, |
428 | }; | |
3e61aa8f SH |
429 | |
430 | INIT_LIST_HEAD(&f.remotes); | |
431 | list_add_rcu(&remote.list, &f.remotes); | |
e4f67add DS |
432 | |
433 | vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH); | |
434 | } | |
435 | ||
436 | static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) | |
437 | { | |
bb3fd687 SH |
438 | struct vxlan_fdb f = { |
439 | .state = NUD_STALE, | |
440 | }; | |
e4f67add | 441 | |
3e61aa8f | 442 | INIT_LIST_HEAD(&f.remotes); |
e4f67add DS |
443 | memcpy(f.eth_addr, eth_addr, ETH_ALEN); |
444 | ||
445 | vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH); | |
446 | } | |
447 | ||
d342894c | 448 | /* Hash Ethernet address */ |
449 | static u32 eth_hash(const unsigned char *addr) | |
450 | { | |
451 | u64 value = get_unaligned((u64 *)addr); | |
452 | ||
453 | /* only want 6 bytes */ | |
454 | #ifdef __BIG_ENDIAN | |
d342894c | 455 | value >>= 16; |
321fb991 | 456 | #else |
457 | value <<= 16; | |
d342894c | 458 | #endif |
459 | return hash_64(value, FDB_HASH_BITS); | |
460 | } | |
461 | ||
462 | /* Hash chain to use given mac address */ | |
463 | static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, | |
464 | const u8 *mac) | |
465 | { | |
466 | return &vxlan->fdb_head[eth_hash(mac)]; | |
467 | } | |
468 | ||
469 | /* Look up Ethernet address in forwarding table */ | |
014be2c8 | 470 | static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, |
d342894c | 471 | const u8 *mac) |
472 | ||
473 | { | |
474 | struct hlist_head *head = vxlan_fdb_head(vxlan, mac); | |
475 | struct vxlan_fdb *f; | |
d342894c | 476 | |
b67bfe0d | 477 | hlist_for_each_entry_rcu(f, head, hlist) { |
7367d0b5 | 478 | if (ether_addr_equal(mac, f->eth_addr)) |
d342894c | 479 | return f; |
480 | } | |
481 | ||
482 | return NULL; | |
483 | } | |
484 | ||
014be2c8 SS |
485 | static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, |
486 | const u8 *mac) | |
487 | { | |
488 | struct vxlan_fdb *f; | |
489 | ||
490 | f = __vxlan_find_mac(vxlan, mac); | |
491 | if (f) | |
492 | f->used = jiffies; | |
493 | ||
494 | return f; | |
495 | } | |
496 | ||
a5e7c10a MR |
497 | /* caller should hold vxlan->hash_lock */ |
498 | static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, | |
e4c7ed41 | 499 | union vxlan_addr *ip, __be16 port, |
a5e7c10a | 500 | __u32 vni, __u32 ifindex) |
6681712d | 501 | { |
3e61aa8f | 502 | struct vxlan_rdst *rd; |
6681712d | 503 | |
3e61aa8f | 504 | list_for_each_entry(rd, &f->remotes, list) { |
e4c7ed41 | 505 | if (vxlan_addr_equal(&rd->remote_ip, ip) && |
6681712d DS |
506 | rd->remote_port == port && |
507 | rd->remote_vni == vni && | |
508 | rd->remote_ifindex == ifindex) | |
a5e7c10a | 509 | return rd; |
6681712d | 510 | } |
3e61aa8f | 511 | |
a5e7c10a MR |
512 | return NULL; |
513 | } | |
514 | ||
906dc186 TR |
515 | /* Replace destination of unicast mac */ |
516 | static int vxlan_fdb_replace(struct vxlan_fdb *f, | |
e4c7ed41 | 517 | union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex) |
906dc186 TR |
518 | { |
519 | struct vxlan_rdst *rd; | |
520 | ||
521 | rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); | |
522 | if (rd) | |
523 | return 0; | |
524 | ||
525 | rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); | |
526 | if (!rd) | |
527 | return 0; | |
e4c7ed41 | 528 | rd->remote_ip = *ip; |
906dc186 TR |
529 | rd->remote_port = port; |
530 | rd->remote_vni = vni; | |
531 | rd->remote_ifindex = ifindex; | |
532 | return 1; | |
533 | } | |
534 | ||
a5e7c10a MR |
535 | /* Add/update destinations for multicast */ |
536 | static int vxlan_fdb_append(struct vxlan_fdb *f, | |
e4c7ed41 | 537 | union vxlan_addr *ip, __be16 port, __u32 vni, __u32 ifindex) |
a5e7c10a MR |
538 | { |
539 | struct vxlan_rdst *rd; | |
540 | ||
541 | rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); | |
542 | if (rd) | |
543 | return 0; | |
544 | ||
6681712d DS |
545 | rd = kmalloc(sizeof(*rd), GFP_ATOMIC); |
546 | if (rd == NULL) | |
547 | return -ENOBUFS; | |
e4c7ed41 | 548 | rd->remote_ip = *ip; |
6681712d DS |
549 | rd->remote_port = port; |
550 | rd->remote_vni = vni; | |
551 | rd->remote_ifindex = ifindex; | |
3e61aa8f SH |
552 | |
553 | list_add_tail_rcu(&rd->list, &f->remotes); | |
554 | ||
6681712d DS |
555 | return 1; |
556 | } | |
557 | ||
dc01e7d3 OG |
558 | static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, struct sk_buff *skb) |
559 | { | |
560 | struct sk_buff *p, **pp = NULL; | |
561 | struct vxlanhdr *vh, *vh2; | |
562 | struct ethhdr *eh, *eh2; | |
563 | unsigned int hlen, off_vx, off_eth; | |
564 | const struct packet_offload *ptype; | |
565 | __be16 type; | |
566 | int flush = 1; | |
567 | ||
568 | off_vx = skb_gro_offset(skb); | |
569 | hlen = off_vx + sizeof(*vh); | |
570 | vh = skb_gro_header_fast(skb, off_vx); | |
571 | if (skb_gro_header_hard(skb, hlen)) { | |
572 | vh = skb_gro_header_slow(skb, hlen, off_vx); | |
573 | if (unlikely(!vh)) | |
574 | goto out; | |
575 | } | |
576 | skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ | |
577 | ||
578 | off_eth = skb_gro_offset(skb); | |
579 | hlen = off_eth + sizeof(*eh); | |
580 | eh = skb_gro_header_fast(skb, off_eth); | |
581 | if (skb_gro_header_hard(skb, hlen)) { | |
582 | eh = skb_gro_header_slow(skb, hlen, off_eth); | |
583 | if (unlikely(!eh)) | |
584 | goto out; | |
585 | } | |
586 | ||
587 | flush = 0; | |
588 | ||
589 | for (p = *head; p; p = p->next) { | |
590 | if (!NAPI_GRO_CB(p)->same_flow) | |
591 | continue; | |
592 | ||
593 | vh2 = (struct vxlanhdr *)(p->data + off_vx); | |
594 | eh2 = (struct ethhdr *)(p->data + off_eth); | |
595 | if (vh->vx_vni != vh2->vx_vni || compare_ether_header(eh, eh2)) { | |
596 | NAPI_GRO_CB(p)->same_flow = 0; | |
597 | continue; | |
598 | } | |
599 | goto found; | |
600 | } | |
601 | ||
602 | found: | |
603 | type = eh->h_proto; | |
604 | ||
605 | rcu_read_lock(); | |
606 | ptype = gro_find_receive_by_type(type); | |
607 | if (ptype == NULL) { | |
608 | flush = 1; | |
609 | goto out_unlock; | |
610 | } | |
611 | ||
612 | skb_gro_pull(skb, sizeof(*eh)); /* pull inner eth header */ | |
613 | pp = ptype->callbacks.gro_receive(head, skb); | |
614 | ||
615 | out_unlock: | |
616 | rcu_read_unlock(); | |
617 | out: | |
618 | NAPI_GRO_CB(skb)->flush |= flush; | |
619 | ||
620 | return pp; | |
621 | } | |
622 | ||
623 | static int vxlan_gro_complete(struct sk_buff *skb, int nhoff) | |
624 | { | |
625 | struct ethhdr *eh; | |
626 | struct packet_offload *ptype; | |
627 | __be16 type; | |
628 | int vxlan_len = sizeof(struct vxlanhdr) + sizeof(struct ethhdr); | |
629 | int err = -ENOSYS; | |
630 | ||
631 | eh = (struct ethhdr *)(skb->data + nhoff + sizeof(struct vxlanhdr)); | |
632 | type = eh->h_proto; | |
633 | ||
634 | rcu_read_lock(); | |
635 | ptype = gro_find_complete_by_type(type); | |
636 | if (ptype != NULL) | |
637 | err = ptype->callbacks.gro_complete(skb, nhoff + vxlan_len); | |
638 | ||
639 | rcu_read_unlock(); | |
640 | return err; | |
641 | } | |
642 | ||
53cf5275 | 643 | /* Notify netdevs that UDP port started listening */ |
dc01e7d3 | 644 | static void vxlan_notify_add_rx_port(struct vxlan_sock *vs) |
53cf5275 JG |
645 | { |
646 | struct net_device *dev; | |
dc01e7d3 | 647 | struct sock *sk = vs->sock->sk; |
53cf5275 JG |
648 | struct net *net = sock_net(sk); |
649 | sa_family_t sa_family = sk->sk_family; | |
35e42379 | 650 | __be16 port = inet_sk(sk)->inet_sport; |
dc01e7d3 OG |
651 | int err; |
652 | ||
653 | if (sa_family == AF_INET) { | |
654 | err = udp_add_offload(&vs->udp_offloads); | |
655 | if (err) | |
656 | pr_warn("vxlan: udp_add_offload failed with status %d\n", err); | |
657 | } | |
53cf5275 JG |
658 | |
659 | rcu_read_lock(); | |
660 | for_each_netdev_rcu(net, dev) { | |
661 | if (dev->netdev_ops->ndo_add_vxlan_port) | |
662 | dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family, | |
663 | port); | |
664 | } | |
665 | rcu_read_unlock(); | |
666 | } | |
667 | ||
668 | /* Notify netdevs that UDP port is no more listening */ | |
dc01e7d3 | 669 | static void vxlan_notify_del_rx_port(struct vxlan_sock *vs) |
53cf5275 JG |
670 | { |
671 | struct net_device *dev; | |
dc01e7d3 | 672 | struct sock *sk = vs->sock->sk; |
53cf5275 JG |
673 | struct net *net = sock_net(sk); |
674 | sa_family_t sa_family = sk->sk_family; | |
35e42379 | 675 | __be16 port = inet_sk(sk)->inet_sport; |
53cf5275 JG |
676 | |
677 | rcu_read_lock(); | |
678 | for_each_netdev_rcu(net, dev) { | |
679 | if (dev->netdev_ops->ndo_del_vxlan_port) | |
680 | dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family, | |
681 | port); | |
682 | } | |
683 | rcu_read_unlock(); | |
dc01e7d3 OG |
684 | |
685 | if (sa_family == AF_INET) | |
686 | udp_del_offload(&vs->udp_offloads); | |
53cf5275 JG |
687 | } |
688 | ||
d342894c | 689 | /* Add new entry to forwarding table -- assumes lock held */ |
690 | static int vxlan_fdb_create(struct vxlan_dev *vxlan, | |
e4c7ed41 | 691 | const u8 *mac, union vxlan_addr *ip, |
6681712d | 692 | __u16 state, __u16 flags, |
73cf3317 | 693 | __be16 port, __u32 vni, __u32 ifindex, |
ae884082 | 694 | __u8 ndm_flags) |
d342894c | 695 | { |
696 | struct vxlan_fdb *f; | |
697 | int notify = 0; | |
698 | ||
014be2c8 | 699 | f = __vxlan_find_mac(vxlan, mac); |
d342894c | 700 | if (f) { |
701 | if (flags & NLM_F_EXCL) { | |
702 | netdev_dbg(vxlan->dev, | |
703 | "lost race to create %pM\n", mac); | |
704 | return -EEXIST; | |
705 | } | |
706 | if (f->state != state) { | |
707 | f->state = state; | |
708 | f->updated = jiffies; | |
709 | notify = 1; | |
710 | } | |
ae884082 DS |
711 | if (f->flags != ndm_flags) { |
712 | f->flags = ndm_flags; | |
713 | f->updated = jiffies; | |
714 | notify = 1; | |
715 | } | |
906dc186 TR |
716 | if ((flags & NLM_F_REPLACE)) { |
717 | /* Only change unicasts */ | |
718 | if (!(is_multicast_ether_addr(f->eth_addr) || | |
719 | is_zero_ether_addr(f->eth_addr))) { | |
720 | int rc = vxlan_fdb_replace(f, ip, port, vni, | |
721 | ifindex); | |
722 | ||
723 | if (rc < 0) | |
724 | return rc; | |
725 | notify |= rc; | |
726 | } else | |
727 | return -EOPNOTSUPP; | |
728 | } | |
6681712d | 729 | if ((flags & NLM_F_APPEND) && |
58e4c767 MR |
730 | (is_multicast_ether_addr(f->eth_addr) || |
731 | is_zero_ether_addr(f->eth_addr))) { | |
6681712d DS |
732 | int rc = vxlan_fdb_append(f, ip, port, vni, ifindex); |
733 | ||
734 | if (rc < 0) | |
735 | return rc; | |
736 | notify |= rc; | |
737 | } | |
d342894c | 738 | } else { |
739 | if (!(flags & NLM_F_CREATE)) | |
740 | return -ENOENT; | |
741 | ||
742 | if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax) | |
743 | return -ENOSPC; | |
744 | ||
906dc186 TR |
745 | /* Disallow replace to add a multicast entry */ |
746 | if ((flags & NLM_F_REPLACE) && | |
747 | (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) | |
748 | return -EOPNOTSUPP; | |
749 | ||
e4c7ed41 | 750 | netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); |
d342894c | 751 | f = kmalloc(sizeof(*f), GFP_ATOMIC); |
752 | if (!f) | |
753 | return -ENOMEM; | |
754 | ||
755 | notify = 1; | |
d342894c | 756 | f->state = state; |
ae884082 | 757 | f->flags = ndm_flags; |
d342894c | 758 | f->updated = f->used = jiffies; |
3e61aa8f | 759 | INIT_LIST_HEAD(&f->remotes); |
d342894c | 760 | memcpy(f->eth_addr, mac, ETH_ALEN); |
761 | ||
3e61aa8f SH |
762 | vxlan_fdb_append(f, ip, port, vni, ifindex); |
763 | ||
d342894c | 764 | ++vxlan->addrcnt; |
765 | hlist_add_head_rcu(&f->hlist, | |
766 | vxlan_fdb_head(vxlan, mac)); | |
767 | } | |
768 | ||
769 | if (notify) | |
770 | vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH); | |
771 | ||
772 | return 0; | |
773 | } | |
774 | ||
6706c82e | 775 | static void vxlan_fdb_free(struct rcu_head *head) |
6681712d DS |
776 | { |
777 | struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); | |
3e61aa8f | 778 | struct vxlan_rdst *rd, *nd; |
6681712d | 779 | |
3e61aa8f | 780 | list_for_each_entry_safe(rd, nd, &f->remotes, list) |
6681712d | 781 | kfree(rd); |
6681712d DS |
782 | kfree(f); |
783 | } | |
784 | ||
d342894c | 785 | static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f) |
786 | { | |
787 | netdev_dbg(vxlan->dev, | |
788 | "delete %pM\n", f->eth_addr); | |
789 | ||
790 | --vxlan->addrcnt; | |
791 | vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH); | |
792 | ||
793 | hlist_del_rcu(&f->hlist); | |
6681712d | 794 | call_rcu(&f->rcu, vxlan_fdb_free); |
d342894c | 795 | } |
796 | ||
f0b074be | 797 | static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, |
e4c7ed41 | 798 | union vxlan_addr *ip, __be16 *port, u32 *vni, u32 *ifindex) |
d342894c | 799 | { |
6681712d | 800 | struct net *net = dev_net(vxlan->dev); |
e4c7ed41 | 801 | int err; |
d342894c | 802 | |
f0b074be | 803 | if (tb[NDA_DST]) { |
e4c7ed41 CW |
804 | err = vxlan_nla_get_addr(ip, tb[NDA_DST]); |
805 | if (err) | |
806 | return err; | |
f0b074be | 807 | } else { |
e4c7ed41 CW |
808 | union vxlan_addr *remote = &vxlan->default_dst.remote_ip; |
809 | if (remote->sa.sa_family == AF_INET) { | |
810 | ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); | |
811 | ip->sa.sa_family = AF_INET; | |
812 | #if IS_ENABLED(CONFIG_IPV6) | |
813 | } else { | |
814 | ip->sin6.sin6_addr = in6addr_any; | |
815 | ip->sa.sa_family = AF_INET6; | |
816 | #endif | |
817 | } | |
f0b074be | 818 | } |
d342894c | 819 | |
6681712d | 820 | if (tb[NDA_PORT]) { |
73cf3317 | 821 | if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) |
6681712d | 822 | return -EINVAL; |
f0b074be MR |
823 | *port = nla_get_be16(tb[NDA_PORT]); |
824 | } else { | |
825 | *port = vxlan->dst_port; | |
826 | } | |
6681712d DS |
827 | |
828 | if (tb[NDA_VNI]) { | |
829 | if (nla_len(tb[NDA_VNI]) != sizeof(u32)) | |
830 | return -EINVAL; | |
f0b074be MR |
831 | *vni = nla_get_u32(tb[NDA_VNI]); |
832 | } else { | |
833 | *vni = vxlan->default_dst.remote_vni; | |
834 | } | |
6681712d DS |
835 | |
836 | if (tb[NDA_IFINDEX]) { | |
5abb0029 | 837 | struct net_device *tdev; |
6681712d DS |
838 | |
839 | if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) | |
840 | return -EINVAL; | |
f0b074be | 841 | *ifindex = nla_get_u32(tb[NDA_IFINDEX]); |
73763949 | 842 | tdev = __dev_get_by_index(net, *ifindex); |
5abb0029 | 843 | if (!tdev) |
6681712d | 844 | return -EADDRNOTAVAIL; |
f0b074be MR |
845 | } else { |
846 | *ifindex = 0; | |
847 | } | |
848 | ||
849 | return 0; | |
850 | } | |
851 | ||
852 | /* Add static entry (via netlink) */ | |
853 | static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], | |
854 | struct net_device *dev, | |
855 | const unsigned char *addr, u16 flags) | |
856 | { | |
857 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
858 | /* struct net *net = dev_net(vxlan->dev); */ | |
e4c7ed41 | 859 | union vxlan_addr ip; |
f0b074be MR |
860 | __be16 port; |
861 | u32 vni, ifindex; | |
862 | int err; | |
863 | ||
864 | if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { | |
865 | pr_info("RTM_NEWNEIGH with invalid state %#x\n", | |
866 | ndm->ndm_state); | |
867 | return -EINVAL; | |
868 | } | |
869 | ||
870 | if (tb[NDA_DST] == NULL) | |
871 | return -EINVAL; | |
872 | ||
873 | err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex); | |
874 | if (err) | |
875 | return err; | |
6681712d | 876 | |
d342894c | 877 | spin_lock_bh(&vxlan->hash_lock); |
e4c7ed41 | 878 | err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags, |
73cf3317 | 879 | port, vni, ifindex, ndm->ndm_flags); |
d342894c | 880 | spin_unlock_bh(&vxlan->hash_lock); |
881 | ||
882 | return err; | |
883 | } | |
884 | ||
885 | /* Delete entry (via netlink) */ | |
1690be63 VY |
886 | static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], |
887 | struct net_device *dev, | |
d342894c | 888 | const unsigned char *addr) |
889 | { | |
890 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
891 | struct vxlan_fdb *f; | |
bc7892ba | 892 | struct vxlan_rdst *rd = NULL; |
e4c7ed41 | 893 | union vxlan_addr ip; |
bc7892ba MR |
894 | __be16 port; |
895 | u32 vni, ifindex; | |
896 | int err; | |
897 | ||
898 | err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex); | |
899 | if (err) | |
900 | return err; | |
901 | ||
902 | err = -ENOENT; | |
d342894c | 903 | |
904 | spin_lock_bh(&vxlan->hash_lock); | |
905 | f = vxlan_find_mac(vxlan, addr); | |
bc7892ba MR |
906 | if (!f) |
907 | goto out; | |
908 | ||
e4c7ed41 CW |
909 | if (!vxlan_addr_any(&ip)) { |
910 | rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); | |
bc7892ba MR |
911 | if (!rd) |
912 | goto out; | |
913 | } | |
914 | ||
915 | err = 0; | |
916 | ||
917 | /* remove a destination if it's not the only one on the list, | |
918 | * otherwise destroy the fdb entry | |
919 | */ | |
920 | if (rd && !list_is_singular(&f->remotes)) { | |
921 | list_del_rcu(&rd->list); | |
dcdc7a59 | 922 | kfree_rcu(rd, rcu); |
bc7892ba | 923 | goto out; |
d342894c | 924 | } |
bc7892ba MR |
925 | |
926 | vxlan_fdb_destroy(vxlan, f); | |
927 | ||
928 | out: | |
d342894c | 929 | spin_unlock_bh(&vxlan->hash_lock); |
930 | ||
931 | return err; | |
932 | } | |
933 | ||
934 | /* Dump forwarding table */ | |
935 | static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, | |
936 | struct net_device *dev, int idx) | |
937 | { | |
938 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
939 | unsigned int h; | |
940 | ||
941 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
942 | struct vxlan_fdb *f; | |
d342894c | 943 | int err; |
944 | ||
b67bfe0d | 945 | hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { |
6681712d | 946 | struct vxlan_rdst *rd; |
6681712d | 947 | |
3e61aa8f SH |
948 | if (idx < cb->args[0]) |
949 | goto skip; | |
950 | ||
951 | list_for_each_entry_rcu(rd, &f->remotes, list) { | |
6681712d DS |
952 | err = vxlan_fdb_info(skb, vxlan, f, |
953 | NETLINK_CB(cb->skb).portid, | |
954 | cb->nlh->nlmsg_seq, | |
955 | RTM_NEWNEIGH, | |
956 | NLM_F_MULTI, rd); | |
957 | if (err < 0) | |
3e61aa8f | 958 | goto out; |
6681712d | 959 | } |
3e61aa8f SH |
960 | skip: |
961 | ++idx; | |
d342894c | 962 | } |
963 | } | |
3e61aa8f | 964 | out: |
d342894c | 965 | return idx; |
966 | } | |
967 | ||
968 | /* Watch incoming packets to learn mapping between Ethernet address | |
969 | * and Tunnel endpoint. | |
26a41ae6 | 970 | * Return true if packet is bogus and should be droppped. |
d342894c | 971 | */ |
26a41ae6 | 972 | static bool vxlan_snoop(struct net_device *dev, |
e4c7ed41 | 973 | union vxlan_addr *src_ip, const u8 *src_mac) |
d342894c | 974 | { |
975 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
976 | struct vxlan_fdb *f; | |
d342894c | 977 | |
978 | f = vxlan_find_mac(vxlan, src_mac); | |
979 | if (likely(f)) { | |
5ca5461c | 980 | struct vxlan_rdst *rdst = first_remote_rcu(f); |
3e61aa8f | 981 | |
e4c7ed41 | 982 | if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip))) |
26a41ae6 | 983 | return false; |
984 | ||
985 | /* Don't migrate static entries, drop packets */ | |
eb064c3b | 986 | if (f->state & NUD_NOARP) |
26a41ae6 | 987 | return true; |
d342894c | 988 | |
989 | if (net_ratelimit()) | |
990 | netdev_info(dev, | |
e4c7ed41 | 991 | "%pM migrated from %pIS to %pIS\n", |
3e61aa8f | 992 | src_mac, &rdst->remote_ip, &src_ip); |
d342894c | 993 | |
e4c7ed41 | 994 | rdst->remote_ip = *src_ip; |
d342894c | 995 | f->updated = jiffies; |
8385f50a | 996 | vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH); |
d342894c | 997 | } else { |
998 | /* learned new entry */ | |
999 | spin_lock(&vxlan->hash_lock); | |
3bf74b1a | 1000 | |
1001 | /* close off race between vxlan_flush and incoming packets */ | |
1002 | if (netif_running(dev)) | |
1003 | vxlan_fdb_create(vxlan, src_mac, src_ip, | |
1004 | NUD_REACHABLE, | |
1005 | NLM_F_EXCL|NLM_F_CREATE, | |
1006 | vxlan->dst_port, | |
1007 | vxlan->default_dst.remote_vni, | |
1008 | 0, NTF_SELF); | |
d342894c | 1009 | spin_unlock(&vxlan->hash_lock); |
1010 | } | |
26a41ae6 | 1011 | |
1012 | return false; | |
d342894c | 1013 | } |
1014 | ||
d342894c | 1015 | /* See if multicast group is already in use by other ID */ |
95ab0991 | 1016 | static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev) |
d342894c | 1017 | { |
553675fb | 1018 | struct vxlan_dev *vxlan; |
d342894c | 1019 | |
95ab0991 G |
1020 | /* The vxlan_sock is only used by dev, leaving group has |
1021 | * no effect on other vxlan devices. | |
1022 | */ | |
1023 | if (atomic_read(&dev->vn_sock->refcnt) == 1) | |
1024 | return false; | |
1025 | ||
553675fb | 1026 | list_for_each_entry(vxlan, &vn->vxlan_list, next) { |
95ab0991 | 1027 | if (!netif_running(vxlan->dev) || vxlan == dev) |
553675fb | 1028 | continue; |
d342894c | 1029 | |
95ab0991 G |
1030 | if (vxlan->vn_sock != dev->vn_sock) |
1031 | continue; | |
1032 | ||
1033 | if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip, | |
1034 | &dev->default_dst.remote_ip)) | |
1035 | continue; | |
1036 | ||
1037 | if (vxlan->default_dst.remote_ifindex != | |
1038 | dev->default_dst.remote_ifindex) | |
1039 | continue; | |
1040 | ||
1041 | return true; | |
553675fb | 1042 | } |
d342894c | 1043 | |
1044 | return false; | |
1045 | } | |
1046 | ||
7c47cedf | 1047 | static void vxlan_sock_hold(struct vxlan_sock *vs) |
d342894c | 1048 | { |
7c47cedf SH |
1049 | atomic_inc(&vs->refcnt); |
1050 | } | |
d342894c | 1051 | |
012a5729 | 1052 | void vxlan_sock_release(struct vxlan_sock *vs) |
7c47cedf | 1053 | { |
53cf5275 JG |
1054 | struct sock *sk = vs->sock->sk; |
1055 | struct net *net = sock_net(sk); | |
1056 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
012a5729 | 1057 | |
7c47cedf SH |
1058 | if (!atomic_dec_and_test(&vs->refcnt)) |
1059 | return; | |
d342894c | 1060 | |
1c51a915 | 1061 | spin_lock(&vn->sock_lock); |
7c47cedf | 1062 | hlist_del_rcu(&vs->hlist); |
559835ea | 1063 | rcu_assign_sk_user_data(vs->sock->sk, NULL); |
dc01e7d3 | 1064 | vxlan_notify_del_rx_port(vs); |
1c51a915 SH |
1065 | spin_unlock(&vn->sock_lock); |
1066 | ||
7c47cedf | 1067 | queue_work(vxlan_wq, &vs->del_work); |
d342894c | 1068 | } |
012a5729 | 1069 | EXPORT_SYMBOL_GPL(vxlan_sock_release); |
d342894c | 1070 | |
3fc2de2f | 1071 | /* Callback to update multicast group membership when first VNI on |
1072 | * multicast asddress is brought up | |
1073 | * Done as workqueue because ip_mc_join_group acquires RTNL. | |
7c47cedf | 1074 | */ |
3fc2de2f | 1075 | static void vxlan_igmp_join(struct work_struct *work) |
d342894c | 1076 | { |
3fc2de2f | 1077 | struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_join); |
7c47cedf SH |
1078 | struct vxlan_sock *vs = vxlan->vn_sock; |
1079 | struct sock *sk = vs->sock->sk; | |
e4c7ed41 CW |
1080 | union vxlan_addr *ip = &vxlan->default_dst.remote_ip; |
1081 | int ifindex = vxlan->default_dst.remote_ifindex; | |
d342894c | 1082 | |
d342894c | 1083 | lock_sock(sk); |
e4c7ed41 CW |
1084 | if (ip->sa.sa_family == AF_INET) { |
1085 | struct ip_mreqn mreq = { | |
1086 | .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr, | |
1087 | .imr_ifindex = ifindex, | |
1088 | }; | |
1089 | ||
1090 | ip_mc_join_group(sk, &mreq); | |
1091 | #if IS_ENABLED(CONFIG_IPV6) | |
1092 | } else { | |
1093 | ipv6_stub->ipv6_sock_mc_join(sk, ifindex, | |
1094 | &ip->sin6.sin6_addr); | |
1095 | #endif | |
1096 | } | |
3fc2de2f | 1097 | release_sock(sk); |
1098 | ||
012a5729 | 1099 | vxlan_sock_release(vs); |
3fc2de2f | 1100 | dev_put(vxlan->dev); |
1101 | } | |
1102 | ||
1103 | /* Inverse of vxlan_igmp_join when last VNI is brought down */ | |
1104 | static void vxlan_igmp_leave(struct work_struct *work) | |
1105 | { | |
1106 | struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, igmp_leave); | |
3fc2de2f | 1107 | struct vxlan_sock *vs = vxlan->vn_sock; |
1108 | struct sock *sk = vs->sock->sk; | |
e4c7ed41 CW |
1109 | union vxlan_addr *ip = &vxlan->default_dst.remote_ip; |
1110 | int ifindex = vxlan->default_dst.remote_ifindex; | |
3fc2de2f | 1111 | |
1112 | lock_sock(sk); | |
e4c7ed41 CW |
1113 | if (ip->sa.sa_family == AF_INET) { |
1114 | struct ip_mreqn mreq = { | |
1115 | .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr, | |
1116 | .imr_ifindex = ifindex, | |
1117 | }; | |
1118 | ||
1119 | ip_mc_leave_group(sk, &mreq); | |
1120 | #if IS_ENABLED(CONFIG_IPV6) | |
1121 | } else { | |
1122 | ipv6_stub->ipv6_sock_mc_drop(sk, ifindex, | |
1123 | &ip->sin6.sin6_addr); | |
1124 | #endif | |
1125 | } | |
1126 | ||
d342894c | 1127 | release_sock(sk); |
d342894c | 1128 | |
012a5729 | 1129 | vxlan_sock_release(vs); |
7c47cedf | 1130 | dev_put(vxlan->dev); |
d342894c | 1131 | } |
1132 | ||
1133 | /* Callback from net/ipv4/udp.c to receive packets */ | |
1134 | static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb) | |
1135 | { | |
5cfccc5a | 1136 | struct vxlan_sock *vs; |
d342894c | 1137 | struct vxlanhdr *vxh; |
553675fb | 1138 | __be16 port; |
d342894c | 1139 | |
d342894c | 1140 | /* Need Vxlan and inner Ethernet header to be present */ |
7ce04758 | 1141 | if (!pskb_may_pull(skb, VXLAN_HLEN)) |
d342894c | 1142 | goto error; |
1143 | ||
7ce04758 PS |
1144 | /* Return packets with reserved bits set */ |
1145 | vxh = (struct vxlanhdr *)(udp_hdr(skb) + 1); | |
d342894c | 1146 | if (vxh->vx_flags != htonl(VXLAN_FLAGS) || |
1147 | (vxh->vx_vni & htonl(0xff))) { | |
1148 | netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", | |
1149 | ntohl(vxh->vx_flags), ntohl(vxh->vx_vni)); | |
1150 | goto error; | |
1151 | } | |
1152 | ||
5cfccc5a PS |
1153 | if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB))) |
1154 | goto drop; | |
1155 | ||
553675fb | 1156 | port = inet_sk(sk)->inet_sport; |
5cfccc5a | 1157 | |
559835ea | 1158 | vs = rcu_dereference_sk_user_data(sk); |
5cfccc5a | 1159 | if (!vs) |
d342894c | 1160 | goto drop; |
d342894c | 1161 | |
d0bc6555 OG |
1162 | /* If the NIC driver gave us an encapsulated packet |
1163 | * with the encapsulation mark, the device checksummed it | |
1164 | * for us. Otherwise force the upper layers to verify it. | |
1165 | */ | |
1166 | if ((skb->ip_summed != CHECKSUM_UNNECESSARY && skb->ip_summed != CHECKSUM_PARTIAL) || | |
1167 | !skb->encapsulation) | |
1168 | skb->ip_summed = CHECKSUM_NONE; | |
1169 | ||
1170 | skb->encapsulation = 0; | |
1171 | ||
5cfccc5a PS |
1172 | vs->rcv(vs, skb, vxh->vx_vni); |
1173 | return 0; | |
1174 | ||
1175 | drop: | |
1176 | /* Consume bad packet */ | |
1177 | kfree_skb(skb); | |
1178 | return 0; | |
1179 | ||
1180 | error: | |
1181 | /* Return non vxlan pkt */ | |
1182 | return 1; | |
1183 | } | |
1184 | ||
1185 | static void vxlan_rcv(struct vxlan_sock *vs, | |
1186 | struct sk_buff *skb, __be32 vx_vni) | |
1187 | { | |
e4c7ed41 CW |
1188 | struct iphdr *oip = NULL; |
1189 | struct ipv6hdr *oip6 = NULL; | |
5cfccc5a | 1190 | struct vxlan_dev *vxlan; |
8f84985f | 1191 | struct pcpu_sw_netstats *stats; |
e4c7ed41 | 1192 | union vxlan_addr saddr; |
5cfccc5a | 1193 | __u32 vni; |
e4c7ed41 CW |
1194 | int err = 0; |
1195 | union vxlan_addr *remote_ip; | |
5cfccc5a PS |
1196 | |
1197 | vni = ntohl(vx_vni) >> 8; | |
1198 | /* Is this VNI defined? */ | |
1199 | vxlan = vxlan_vs_find_vni(vs, vni); | |
1200 | if (!vxlan) | |
d342894c | 1201 | goto drop; |
d342894c | 1202 | |
e4c7ed41 | 1203 | remote_ip = &vxlan->default_dst.remote_ip; |
e4f67add | 1204 | skb_reset_mac_header(skb); |
d342894c | 1205 | skb->protocol = eth_type_trans(skb, vxlan->dev); |
d342894c | 1206 | |
1207 | /* Ignore packet loops (and multicast echo) */ | |
7367d0b5 | 1208 | if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) |
d342894c | 1209 | goto drop; |
1210 | ||
7ce04758 | 1211 | /* Re-examine inner Ethernet packet */ |
e4c7ed41 CW |
1212 | if (remote_ip->sa.sa_family == AF_INET) { |
1213 | oip = ip_hdr(skb); | |
1214 | saddr.sin.sin_addr.s_addr = oip->saddr; | |
1215 | saddr.sa.sa_family = AF_INET; | |
1216 | #if IS_ENABLED(CONFIG_IPV6) | |
1217 | } else { | |
1218 | oip6 = ipv6_hdr(skb); | |
1219 | saddr.sin6.sin6_addr = oip6->saddr; | |
1220 | saddr.sa.sa_family = AF_INET6; | |
1221 | #endif | |
1222 | } | |
1223 | ||
26a41ae6 | 1224 | if ((vxlan->flags & VXLAN_F_LEARN) && |
e4c7ed41 | 1225 | vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source)) |
26a41ae6 | 1226 | goto drop; |
d342894c | 1227 | |
d342894c | 1228 | skb_reset_network_header(skb); |
0afb1666 | 1229 | |
e4c7ed41 CW |
1230 | if (oip6) |
1231 | err = IP6_ECN_decapsulate(oip6, skb); | |
1232 | if (oip) | |
1233 | err = IP_ECN_decapsulate(oip, skb); | |
1234 | ||
d342894c | 1235 | if (unlikely(err)) { |
e4c7ed41 CW |
1236 | if (log_ecn_error) { |
1237 | if (oip6) | |
1238 | net_info_ratelimited("non-ECT from %pI6\n", | |
1239 | &oip6->saddr); | |
1240 | if (oip) | |
1241 | net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", | |
1242 | &oip->saddr, oip->tos); | |
1243 | } | |
d342894c | 1244 | if (err > 1) { |
1245 | ++vxlan->dev->stats.rx_frame_errors; | |
1246 | ++vxlan->dev->stats.rx_errors; | |
1247 | goto drop; | |
1248 | } | |
1249 | } | |
1250 | ||
e8171045 | 1251 | stats = this_cpu_ptr(vxlan->dev->tstats); |
d342894c | 1252 | u64_stats_update_begin(&stats->syncp); |
1253 | stats->rx_packets++; | |
1254 | stats->rx_bytes += skb->len; | |
1255 | u64_stats_update_end(&stats->syncp); | |
1256 | ||
1257 | netif_rx(skb); | |
1258 | ||
5cfccc5a | 1259 | return; |
d342894c | 1260 | drop: |
1261 | /* Consume bad packet */ | |
1262 | kfree_skb(skb); | |
d342894c | 1263 | } |
1264 | ||
e4f67add DS |
1265 | static int arp_reduce(struct net_device *dev, struct sk_buff *skb) |
1266 | { | |
1267 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1268 | struct arphdr *parp; | |
1269 | u8 *arpptr, *sha; | |
1270 | __be32 sip, tip; | |
1271 | struct neighbour *n; | |
1272 | ||
1273 | if (dev->flags & IFF_NOARP) | |
1274 | goto out; | |
1275 | ||
1276 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) { | |
1277 | dev->stats.tx_dropped++; | |
1278 | goto out; | |
1279 | } | |
1280 | parp = arp_hdr(skb); | |
1281 | ||
1282 | if ((parp->ar_hrd != htons(ARPHRD_ETHER) && | |
1283 | parp->ar_hrd != htons(ARPHRD_IEEE802)) || | |
1284 | parp->ar_pro != htons(ETH_P_IP) || | |
1285 | parp->ar_op != htons(ARPOP_REQUEST) || | |
1286 | parp->ar_hln != dev->addr_len || | |
1287 | parp->ar_pln != 4) | |
1288 | goto out; | |
1289 | arpptr = (u8 *)parp + sizeof(struct arphdr); | |
1290 | sha = arpptr; | |
1291 | arpptr += dev->addr_len; /* sha */ | |
1292 | memcpy(&sip, arpptr, sizeof(sip)); | |
1293 | arpptr += sizeof(sip); | |
1294 | arpptr += dev->addr_len; /* tha */ | |
1295 | memcpy(&tip, arpptr, sizeof(tip)); | |
1296 | ||
1297 | if (ipv4_is_loopback(tip) || | |
1298 | ipv4_is_multicast(tip)) | |
1299 | goto out; | |
1300 | ||
1301 | n = neigh_lookup(&arp_tbl, &tip, dev); | |
1302 | ||
1303 | if (n) { | |
e4f67add DS |
1304 | struct vxlan_fdb *f; |
1305 | struct sk_buff *reply; | |
1306 | ||
1307 | if (!(n->nud_state & NUD_CONNECTED)) { | |
1308 | neigh_release(n); | |
1309 | goto out; | |
1310 | } | |
1311 | ||
1312 | f = vxlan_find_mac(vxlan, n->ha); | |
e4c7ed41 | 1313 | if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { |
e4f67add DS |
1314 | /* bridge-local neighbor */ |
1315 | neigh_release(n); | |
1316 | goto out; | |
1317 | } | |
1318 | ||
1319 | reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, | |
1320 | n->ha, sha); | |
1321 | ||
1322 | neigh_release(n); | |
1323 | ||
1324 | skb_reset_mac_header(reply); | |
1325 | __skb_pull(reply, skb_network_offset(reply)); | |
1326 | reply->ip_summed = CHECKSUM_UNNECESSARY; | |
1327 | reply->pkt_type = PACKET_HOST; | |
1328 | ||
1329 | if (netif_rx_ni(reply) == NET_RX_DROP) | |
1330 | dev->stats.rx_dropped++; | |
e4c7ed41 CW |
1331 | } else if (vxlan->flags & VXLAN_F_L3MISS) { |
1332 | union vxlan_addr ipa = { | |
1333 | .sin.sin_addr.s_addr = tip, | |
1334 | .sa.sa_family = AF_INET, | |
1335 | }; | |
1336 | ||
1337 | vxlan_ip_miss(dev, &ipa); | |
1338 | } | |
e4f67add DS |
1339 | out: |
1340 | consume_skb(skb); | |
1341 | return NETDEV_TX_OK; | |
1342 | } | |
1343 | ||
f564f45c CW |
1344 | #if IS_ENABLED(CONFIG_IPV6) |
1345 | static int neigh_reduce(struct net_device *dev, struct sk_buff *skb) | |
1346 | { | |
1347 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1348 | struct neighbour *n; | |
1349 | union vxlan_addr ipa; | |
1350 | const struct ipv6hdr *iphdr; | |
1351 | const struct in6_addr *saddr, *daddr; | |
1352 | struct nd_msg *msg; | |
1353 | struct inet6_dev *in6_dev = NULL; | |
1354 | ||
1355 | in6_dev = __in6_dev_get(dev); | |
1356 | if (!in6_dev) | |
1357 | goto out; | |
1358 | ||
1359 | if (!pskb_may_pull(skb, skb->len)) | |
1360 | goto out; | |
1361 | ||
1362 | iphdr = ipv6_hdr(skb); | |
1363 | saddr = &iphdr->saddr; | |
1364 | daddr = &iphdr->daddr; | |
1365 | ||
1366 | if (ipv6_addr_loopback(daddr) || | |
1367 | ipv6_addr_is_multicast(daddr)) | |
1368 | goto out; | |
1369 | ||
1370 | msg = (struct nd_msg *)skb_transport_header(skb); | |
1371 | if (msg->icmph.icmp6_code != 0 || | |
1372 | msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION) | |
1373 | goto out; | |
1374 | ||
1375 | n = neigh_lookup(ipv6_stub->nd_tbl, daddr, dev); | |
1376 | ||
1377 | if (n) { | |
1378 | struct vxlan_fdb *f; | |
1379 | ||
1380 | if (!(n->nud_state & NUD_CONNECTED)) { | |
1381 | neigh_release(n); | |
1382 | goto out; | |
1383 | } | |
1384 | ||
1385 | f = vxlan_find_mac(vxlan, n->ha); | |
1386 | if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { | |
1387 | /* bridge-local neighbor */ | |
1388 | neigh_release(n); | |
1389 | goto out; | |
1390 | } | |
1391 | ||
1392 | ipv6_stub->ndisc_send_na(dev, n, saddr, &msg->target, | |
1393 | !!in6_dev->cnf.forwarding, | |
1394 | true, false, false); | |
1395 | neigh_release(n); | |
1396 | } else if (vxlan->flags & VXLAN_F_L3MISS) { | |
1397 | ipa.sin6.sin6_addr = *daddr; | |
1398 | ipa.sa.sa_family = AF_INET6; | |
1399 | vxlan_ip_miss(dev, &ipa); | |
1400 | } | |
1401 | ||
1402 | out: | |
1403 | consume_skb(skb); | |
1404 | return NETDEV_TX_OK; | |
1405 | } | |
1406 | #endif | |
1407 | ||
e4f67add DS |
1408 | static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) |
1409 | { | |
1410 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1411 | struct neighbour *n; | |
e4f67add DS |
1412 | |
1413 | if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) | |
1414 | return false; | |
1415 | ||
1416 | n = NULL; | |
1417 | switch (ntohs(eth_hdr(skb)->h_proto)) { | |
1418 | case ETH_P_IP: | |
e15a00aa CW |
1419 | { |
1420 | struct iphdr *pip; | |
1421 | ||
e4f67add DS |
1422 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
1423 | return false; | |
1424 | pip = ip_hdr(skb); | |
1425 | n = neigh_lookup(&arp_tbl, &pip->daddr, dev); | |
e4c7ed41 CW |
1426 | if (!n && (vxlan->flags & VXLAN_F_L3MISS)) { |
1427 | union vxlan_addr ipa = { | |
1428 | .sin.sin_addr.s_addr = pip->daddr, | |
1429 | .sa.sa_family = AF_INET, | |
1430 | }; | |
1431 | ||
1432 | vxlan_ip_miss(dev, &ipa); | |
1433 | return false; | |
1434 | } | |
1435 | ||
e4f67add | 1436 | break; |
e15a00aa CW |
1437 | } |
1438 | #if IS_ENABLED(CONFIG_IPV6) | |
1439 | case ETH_P_IPV6: | |
1440 | { | |
1441 | struct ipv6hdr *pip6; | |
1442 | ||
1443 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
1444 | return false; | |
1445 | pip6 = ipv6_hdr(skb); | |
1446 | n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); | |
1447 | if (!n && (vxlan->flags & VXLAN_F_L3MISS)) { | |
1448 | union vxlan_addr ipa = { | |
1449 | .sin6.sin6_addr = pip6->daddr, | |
1450 | .sa.sa_family = AF_INET6, | |
1451 | }; | |
1452 | ||
1453 | vxlan_ip_miss(dev, &ipa); | |
1454 | return false; | |
1455 | } | |
1456 | ||
1457 | break; | |
1458 | } | |
1459 | #endif | |
e4f67add DS |
1460 | default: |
1461 | return false; | |
1462 | } | |
1463 | ||
1464 | if (n) { | |
1465 | bool diff; | |
1466 | ||
7367d0b5 | 1467 | diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); |
e4f67add DS |
1468 | if (diff) { |
1469 | memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, | |
1470 | dev->addr_len); | |
1471 | memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); | |
1472 | } | |
1473 | neigh_release(n); | |
1474 | return diff; | |
e4c7ed41 CW |
1475 | } |
1476 | ||
e4f67add DS |
1477 | return false; |
1478 | } | |
1479 | ||
05f47d69 | 1480 | /* Compute source port for outgoing packet |
1481 | * first choice to use L4 flow hash since it will spread | |
1482 | * better and maybe available from hardware | |
1483 | * secondary choice is to use jhash on the Ethernet header | |
1484 | */ | |
49560532 | 1485 | __be16 vxlan_src_port(__u16 port_min, __u16 port_max, struct sk_buff *skb) |
05f47d69 | 1486 | { |
49560532 | 1487 | unsigned int range = (port_max - port_min) + 1; |
05f47d69 | 1488 | u32 hash; |
1489 | ||
3958afa1 | 1490 | hash = skb_get_hash(skb); |
05f47d69 | 1491 | if (!hash) |
1492 | hash = jhash(skb->data, 2 * ETH_ALEN, | |
1493 | (__force u32) skb->protocol); | |
1494 | ||
49560532 | 1495 | return htons((((u64) hash * range) >> 32) + port_min); |
05f47d69 | 1496 | } |
49560532 | 1497 | EXPORT_SYMBOL_GPL(vxlan_src_port); |
05f47d69 | 1498 | |
05c0db08 PS |
1499 | static int handle_offloads(struct sk_buff *skb) |
1500 | { | |
1501 | if (skb_is_gso(skb)) { | |
1502 | int err = skb_unclone(skb, GFP_ATOMIC); | |
1503 | if (unlikely(err)) | |
1504 | return err; | |
1505 | ||
f6ace502 | 1506 | skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL; |
05c0db08 PS |
1507 | } else if (skb->ip_summed != CHECKSUM_PARTIAL) |
1508 | skb->ip_summed = CHECKSUM_NONE; | |
1509 | ||
1510 | return 0; | |
1511 | } | |
1512 | ||
e4c7ed41 | 1513 | #if IS_ENABLED(CONFIG_IPV6) |
11796187 | 1514 | static int vxlan6_xmit_skb(struct vxlan_sock *vs, |
e4c7ed41 CW |
1515 | struct dst_entry *dst, struct sk_buff *skb, |
1516 | struct net_device *dev, struct in6_addr *saddr, | |
1517 | struct in6_addr *daddr, __u8 prio, __u8 ttl, | |
1518 | __be16 src_port, __be16 dst_port, __be32 vni) | |
1519 | { | |
1520 | struct ipv6hdr *ip6h; | |
1521 | struct vxlanhdr *vxh; | |
1522 | struct udphdr *uh; | |
1523 | int min_headroom; | |
1524 | int err; | |
1525 | ||
1526 | if (!skb->encapsulation) { | |
1527 | skb_reset_inner_headers(skb); | |
1528 | skb->encapsulation = 1; | |
1529 | } | |
1530 | ||
963a88b3 ND |
1531 | skb_scrub_packet(skb, false); |
1532 | ||
e4c7ed41 CW |
1533 | min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len |
1534 | + VXLAN_HLEN + sizeof(struct ipv6hdr) | |
1535 | + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0); | |
1536 | ||
1537 | /* Need space for new headers (invalidates iph ptr) */ | |
1538 | err = skb_cow_head(skb, min_headroom); | |
1539 | if (unlikely(err)) | |
1540 | return err; | |
1541 | ||
1542 | if (vlan_tx_tag_present(skb)) { | |
1543 | if (WARN_ON(!__vlan_put_tag(skb, | |
1544 | skb->vlan_proto, | |
1545 | vlan_tx_tag_get(skb)))) | |
1546 | return -ENOMEM; | |
1547 | ||
1548 | skb->vlan_tci = 0; | |
1549 | } | |
1550 | ||
1551 | vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh)); | |
1552 | vxh->vx_flags = htonl(VXLAN_FLAGS); | |
1553 | vxh->vx_vni = vni; | |
1554 | ||
1555 | __skb_push(skb, sizeof(*uh)); | |
1556 | skb_reset_transport_header(skb); | |
1557 | uh = udp_hdr(skb); | |
1558 | ||
1559 | uh->dest = dst_port; | |
1560 | uh->source = src_port; | |
1561 | ||
1562 | uh->len = htons(skb->len); | |
1563 | uh->check = 0; | |
1564 | ||
1565 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); | |
1566 | IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | | |
1567 | IPSKB_REROUTED); | |
e4c7ed41 CW |
1568 | skb_dst_set(skb, dst); |
1569 | ||
1570 | if (!skb_is_gso(skb) && !(dst->dev->features & NETIF_F_IPV6_CSUM)) { | |
1571 | __wsum csum = skb_checksum(skb, 0, skb->len, 0); | |
1572 | skb->ip_summed = CHECKSUM_UNNECESSARY; | |
1573 | uh->check = csum_ipv6_magic(saddr, daddr, skb->len, | |
1574 | IPPROTO_UDP, csum); | |
1575 | if (uh->check == 0) | |
1576 | uh->check = CSUM_MANGLED_0; | |
1577 | } else { | |
1578 | skb->ip_summed = CHECKSUM_PARTIAL; | |
1579 | skb->csum_start = skb_transport_header(skb) - skb->head; | |
1580 | skb->csum_offset = offsetof(struct udphdr, check); | |
1581 | uh->check = ~csum_ipv6_magic(saddr, daddr, | |
1582 | skb->len, IPPROTO_UDP, 0); | |
1583 | } | |
1584 | ||
1585 | __skb_push(skb, sizeof(*ip6h)); | |
1586 | skb_reset_network_header(skb); | |
1587 | ip6h = ipv6_hdr(skb); | |
1588 | ip6h->version = 6; | |
1589 | ip6h->priority = prio; | |
1590 | ip6h->flow_lbl[0] = 0; | |
1591 | ip6h->flow_lbl[1] = 0; | |
1592 | ip6h->flow_lbl[2] = 0; | |
1593 | ip6h->payload_len = htons(skb->len); | |
1594 | ip6h->nexthdr = IPPROTO_UDP; | |
1595 | ip6h->hop_limit = ttl; | |
1596 | ip6h->daddr = *daddr; | |
1597 | ip6h->saddr = *saddr; | |
1598 | ||
e4c7ed41 CW |
1599 | err = handle_offloads(skb); |
1600 | if (err) | |
1601 | return err; | |
1602 | ||
1603 | ip6tunnel_xmit(skb, dev); | |
1604 | return 0; | |
1605 | } | |
1606 | #endif | |
1607 | ||
11796187 | 1608 | int vxlan_xmit_skb(struct vxlan_sock *vs, |
49560532 PS |
1609 | struct rtable *rt, struct sk_buff *skb, |
1610 | __be32 src, __be32 dst, __u8 tos, __u8 ttl, __be16 df, | |
1611 | __be16 src_port, __be16 dst_port, __be32 vni) | |
1612 | { | |
1613 | struct vxlanhdr *vxh; | |
1614 | struct udphdr *uh; | |
649c5b8b | 1615 | int min_headroom; |
49560532 PS |
1616 | int err; |
1617 | ||
1618 | if (!skb->encapsulation) { | |
1619 | skb_reset_inner_headers(skb); | |
1620 | skb->encapsulation = 1; | |
1621 | } | |
1622 | ||
649c5b8b | 1623 | min_headroom = LL_RESERVED_SPACE(rt->dst.dev) + rt->dst.header_len |
1eaa8178 PS |
1624 | + VXLAN_HLEN + sizeof(struct iphdr) |
1625 | + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0); | |
649c5b8b PS |
1626 | |
1627 | /* Need space for new headers (invalidates iph ptr) */ | |
1628 | err = skb_cow_head(skb, min_headroom); | |
1629 | if (unlikely(err)) | |
1630 | return err; | |
1631 | ||
1eaa8178 PS |
1632 | if (vlan_tx_tag_present(skb)) { |
1633 | if (WARN_ON(!__vlan_put_tag(skb, | |
1634 | skb->vlan_proto, | |
1635 | vlan_tx_tag_get(skb)))) | |
1636 | return -ENOMEM; | |
1637 | ||
1638 | skb->vlan_tci = 0; | |
1639 | } | |
1640 | ||
49560532 PS |
1641 | vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh)); |
1642 | vxh->vx_flags = htonl(VXLAN_FLAGS); | |
1643 | vxh->vx_vni = vni; | |
1644 | ||
1645 | __skb_push(skb, sizeof(*uh)); | |
1646 | skb_reset_transport_header(skb); | |
1647 | uh = udp_hdr(skb); | |
1648 | ||
1649 | uh->dest = dst_port; | |
1650 | uh->source = src_port; | |
1651 | ||
1652 | uh->len = htons(skb->len); | |
1653 | uh->check = 0; | |
1654 | ||
49560532 PS |
1655 | err = handle_offloads(skb); |
1656 | if (err) | |
1657 | return err; | |
1658 | ||
963a88b3 ND |
1659 | return iptunnel_xmit(rt, skb, src, dst, IPPROTO_UDP, tos, ttl, df, |
1660 | false); | |
49560532 PS |
1661 | } |
1662 | EXPORT_SYMBOL_GPL(vxlan_xmit_skb); | |
1663 | ||
9dcc71e1 SS |
1664 | /* Bypass encapsulation if the destination is local */ |
1665 | static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, | |
1666 | struct vxlan_dev *dst_vxlan) | |
1667 | { | |
8f84985f | 1668 | struct pcpu_sw_netstats *tx_stats, *rx_stats; |
e4c7ed41 CW |
1669 | union vxlan_addr loopback; |
1670 | union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; | |
9dcc71e1 | 1671 | |
8f84985f LR |
1672 | tx_stats = this_cpu_ptr(src_vxlan->dev->tstats); |
1673 | rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats); | |
9dcc71e1 SS |
1674 | skb->pkt_type = PACKET_HOST; |
1675 | skb->encapsulation = 0; | |
1676 | skb->dev = dst_vxlan->dev; | |
1677 | __skb_pull(skb, skb_network_offset(skb)); | |
1678 | ||
e4c7ed41 CW |
1679 | if (remote_ip->sa.sa_family == AF_INET) { |
1680 | loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); | |
1681 | loopback.sa.sa_family = AF_INET; | |
1682 | #if IS_ENABLED(CONFIG_IPV6) | |
1683 | } else { | |
1684 | loopback.sin6.sin6_addr = in6addr_loopback; | |
1685 | loopback.sa.sa_family = AF_INET6; | |
1686 | #endif | |
1687 | } | |
1688 | ||
9dcc71e1 | 1689 | if (dst_vxlan->flags & VXLAN_F_LEARN) |
e4c7ed41 | 1690 | vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source); |
9dcc71e1 SS |
1691 | |
1692 | u64_stats_update_begin(&tx_stats->syncp); | |
1693 | tx_stats->tx_packets++; | |
1694 | tx_stats->tx_bytes += skb->len; | |
1695 | u64_stats_update_end(&tx_stats->syncp); | |
1696 | ||
1697 | if (netif_rx(skb) == NET_RX_SUCCESS) { | |
1698 | u64_stats_update_begin(&rx_stats->syncp); | |
1699 | rx_stats->rx_packets++; | |
1700 | rx_stats->rx_bytes += skb->len; | |
1701 | u64_stats_update_end(&rx_stats->syncp); | |
1702 | } else { | |
1703 | skb->dev->stats.rx_dropped++; | |
1704 | } | |
1705 | } | |
1706 | ||
4ad16930 SH |
1707 | static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, |
1708 | struct vxlan_rdst *rdst, bool did_rsc) | |
d342894c | 1709 | { |
1710 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
e4c7ed41 | 1711 | struct rtable *rt = NULL; |
d342894c | 1712 | const struct iphdr *old_iph; |
d342894c | 1713 | struct flowi4 fl4; |
e4c7ed41 CW |
1714 | union vxlan_addr *dst; |
1715 | __be16 src_port = 0, dst_port; | |
234f5b73 | 1716 | u32 vni; |
d342894c | 1717 | __be16 df = 0; |
1718 | __u8 tos, ttl; | |
0e6fbc5b | 1719 | int err; |
d342894c | 1720 | |
823aa873 | 1721 | dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port; |
6681712d | 1722 | vni = rdst->remote_vni; |
e4c7ed41 | 1723 | dst = &rdst->remote_ip; |
e4f67add | 1724 | |
e4c7ed41 | 1725 | if (vxlan_addr_any(dst)) { |
e4f67add | 1726 | if (did_rsc) { |
e4f67add | 1727 | /* short-circuited back to local bridge */ |
9dcc71e1 | 1728 | vxlan_encap_bypass(skb, vxlan, vxlan); |
4ad16930 | 1729 | return; |
e4f67add | 1730 | } |
ef59febe | 1731 | goto drop; |
e4f67add | 1732 | } |
ef59febe | 1733 | |
d342894c | 1734 | old_iph = ip_hdr(skb); |
1735 | ||
d342894c | 1736 | ttl = vxlan->ttl; |
e4c7ed41 | 1737 | if (!ttl && vxlan_addr_multicast(dst)) |
d342894c | 1738 | ttl = 1; |
1739 | ||
1740 | tos = vxlan->tos; | |
1741 | if (tos == 1) | |
206aaafc | 1742 | tos = ip_tunnel_get_dsfield(old_iph, skb); |
d342894c | 1743 | |
49560532 | 1744 | src_port = vxlan_src_port(vxlan->port_min, vxlan->port_max, skb); |
d342894c | 1745 | |
e4c7ed41 CW |
1746 | if (dst->sa.sa_family == AF_INET) { |
1747 | memset(&fl4, 0, sizeof(fl4)); | |
1748 | fl4.flowi4_oif = rdst->remote_ifindex; | |
1749 | fl4.flowi4_tos = RT_TOS(tos); | |
1750 | fl4.daddr = dst->sin.sin_addr.s_addr; | |
1751 | fl4.saddr = vxlan->saddr.sin.sin_addr.s_addr; | |
1752 | ||
1753 | rt = ip_route_output_key(dev_net(dev), &fl4); | |
1754 | if (IS_ERR(rt)) { | |
1755 | netdev_dbg(dev, "no route to %pI4\n", | |
1756 | &dst->sin.sin_addr.s_addr); | |
1757 | dev->stats.tx_carrier_errors++; | |
1758 | goto tx_error; | |
1759 | } | |
d342894c | 1760 | |
e4c7ed41 CW |
1761 | if (rt->dst.dev == dev) { |
1762 | netdev_dbg(dev, "circular route to %pI4\n", | |
1763 | &dst->sin.sin_addr.s_addr); | |
1764 | dev->stats.collisions++; | |
fffc15a5 | 1765 | goto rt_tx_error; |
e4c7ed41 CW |
1766 | } |
1767 | ||
1768 | /* Bypass encapsulation if the destination is local */ | |
1769 | if (rt->rt_flags & RTCF_LOCAL && | |
1770 | !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { | |
1771 | struct vxlan_dev *dst_vxlan; | |
1772 | ||
1773 | ip_rt_put(rt); | |
1774 | dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port); | |
1775 | if (!dst_vxlan) | |
1776 | goto tx_error; | |
1777 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
1778 | return; | |
1779 | } | |
1780 | ||
1781 | tos = ip_tunnel_ecn_encap(tos, old_iph, skb); | |
1782 | ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); | |
d342894c | 1783 | |
11796187 | 1784 | err = vxlan_xmit_skb(vxlan->vn_sock, rt, skb, |
e4c7ed41 CW |
1785 | fl4.saddr, dst->sin.sin_addr.s_addr, |
1786 | tos, ttl, df, src_port, dst_port, | |
1787 | htonl(vni << 8)); | |
9dcc71e1 | 1788 | |
e4c7ed41 CW |
1789 | if (err < 0) |
1790 | goto rt_tx_error; | |
1791 | iptunnel_xmit_stats(err, &dev->stats, dev->tstats); | |
1792 | #if IS_ENABLED(CONFIG_IPV6) | |
1793 | } else { | |
1794 | struct sock *sk = vxlan->vn_sock->sock->sk; | |
1795 | struct dst_entry *ndst; | |
1796 | struct flowi6 fl6; | |
1797 | u32 flags; | |
1798 | ||
1799 | memset(&fl6, 0, sizeof(fl6)); | |
1800 | fl6.flowi6_oif = rdst->remote_ifindex; | |
1801 | fl6.daddr = dst->sin6.sin6_addr; | |
1802 | fl6.saddr = vxlan->saddr.sin6.sin6_addr; | |
8c1bb79f | 1803 | fl6.flowi6_proto = IPPROTO_UDP; |
e4c7ed41 CW |
1804 | |
1805 | if (ipv6_stub->ipv6_dst_lookup(sk, &ndst, &fl6)) { | |
1806 | netdev_dbg(dev, "no route to %pI6\n", | |
1807 | &dst->sin6.sin6_addr); | |
1808 | dev->stats.tx_carrier_errors++; | |
9dcc71e1 | 1809 | goto tx_error; |
e4c7ed41 | 1810 | } |
d342894c | 1811 | |
e4c7ed41 CW |
1812 | if (ndst->dev == dev) { |
1813 | netdev_dbg(dev, "circular route to %pI6\n", | |
1814 | &dst->sin6.sin6_addr); | |
1815 | dst_release(ndst); | |
1816 | dev->stats.collisions++; | |
1817 | goto tx_error; | |
1818 | } | |
0e6fbc5b | 1819 | |
e4c7ed41 CW |
1820 | /* Bypass encapsulation if the destination is local */ |
1821 | flags = ((struct rt6_info *)ndst)->rt6i_flags; | |
1822 | if (flags & RTF_LOCAL && | |
1823 | !(flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { | |
1824 | struct vxlan_dev *dst_vxlan; | |
1825 | ||
1826 | dst_release(ndst); | |
1827 | dst_vxlan = vxlan_find_vni(dev_net(dev), vni, dst_port); | |
1828 | if (!dst_vxlan) | |
1829 | goto tx_error; | |
1830 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
1831 | return; | |
1832 | } | |
49560532 | 1833 | |
e4c7ed41 CW |
1834 | ttl = ttl ? : ip6_dst_hoplimit(ndst); |
1835 | ||
11796187 | 1836 | err = vxlan6_xmit_skb(vxlan->vn_sock, ndst, skb, |
e4c7ed41 CW |
1837 | dev, &fl6.saddr, &fl6.daddr, 0, ttl, |
1838 | src_port, dst_port, htonl(vni << 8)); | |
1839 | #endif | |
1840 | } | |
0e6fbc5b | 1841 | |
4ad16930 | 1842 | return; |
d342894c | 1843 | |
1844 | drop: | |
1845 | dev->stats.tx_dropped++; | |
1846 | goto tx_free; | |
1847 | ||
49560532 PS |
1848 | rt_tx_error: |
1849 | ip_rt_put(rt); | |
d342894c | 1850 | tx_error: |
1851 | dev->stats.tx_errors++; | |
1852 | tx_free: | |
1853 | dev_kfree_skb(skb); | |
d342894c | 1854 | } |
1855 | ||
6681712d DS |
1856 | /* Transmit local packets over Vxlan |
1857 | * | |
1858 | * Outer IP header inherits ECN and DF from inner header. | |
1859 | * Outer UDP destination is the VXLAN assigned port. | |
1860 | * source port is based on hash of flow | |
1861 | */ | |
1862 | static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) | |
1863 | { | |
1864 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1865 | struct ethhdr *eth; | |
1866 | bool did_rsc = false; | |
8f646c92 | 1867 | struct vxlan_rdst *rdst, *fdst = NULL; |
6681712d | 1868 | struct vxlan_fdb *f; |
6681712d DS |
1869 | |
1870 | skb_reset_mac_header(skb); | |
1871 | eth = eth_hdr(skb); | |
1872 | ||
f564f45c CW |
1873 | if ((vxlan->flags & VXLAN_F_PROXY)) { |
1874 | if (ntohs(eth->h_proto) == ETH_P_ARP) | |
1875 | return arp_reduce(dev, skb); | |
1876 | #if IS_ENABLED(CONFIG_IPV6) | |
1877 | else if (ntohs(eth->h_proto) == ETH_P_IPV6 && | |
1878 | skb->len >= sizeof(struct ipv6hdr) + sizeof(struct nd_msg) && | |
1879 | ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { | |
1880 | struct nd_msg *msg; | |
1881 | ||
1882 | msg = (struct nd_msg *)skb_transport_header(skb); | |
1883 | if (msg->icmph.icmp6_code == 0 && | |
1884 | msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) | |
1885 | return neigh_reduce(dev, skb); | |
1886 | } | |
1887 | #endif | |
1888 | } | |
6681712d DS |
1889 | |
1890 | f = vxlan_find_mac(vxlan, eth->h_dest); | |
ae884082 DS |
1891 | did_rsc = false; |
1892 | ||
1893 | if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) && | |
e15a00aa CW |
1894 | (ntohs(eth->h_proto) == ETH_P_IP || |
1895 | ntohs(eth->h_proto) == ETH_P_IPV6)) { | |
ae884082 DS |
1896 | did_rsc = route_shortcircuit(dev, skb); |
1897 | if (did_rsc) | |
1898 | f = vxlan_find_mac(vxlan, eth->h_dest); | |
1899 | } | |
1900 | ||
6681712d | 1901 | if (f == NULL) { |
afbd8bae MR |
1902 | f = vxlan_find_mac(vxlan, all_zeros_mac); |
1903 | if (f == NULL) { | |
1904 | if ((vxlan->flags & VXLAN_F_L2MISS) && | |
1905 | !is_multicast_ether_addr(eth->h_dest)) | |
1906 | vxlan_fdb_miss(vxlan, eth->h_dest); | |
1907 | ||
1908 | dev->stats.tx_dropped++; | |
8f646c92 | 1909 | kfree_skb(skb); |
afbd8bae MR |
1910 | return NETDEV_TX_OK; |
1911 | } | |
1912 | } | |
6681712d | 1913 | |
afbd8bae MR |
1914 | list_for_each_entry_rcu(rdst, &f->remotes, list) { |
1915 | struct sk_buff *skb1; | |
6681712d | 1916 | |
8f646c92 ED |
1917 | if (!fdst) { |
1918 | fdst = rdst; | |
1919 | continue; | |
1920 | } | |
afbd8bae MR |
1921 | skb1 = skb_clone(skb, GFP_ATOMIC); |
1922 | if (skb1) | |
1923 | vxlan_xmit_one(skb1, dev, rdst, did_rsc); | |
6681712d DS |
1924 | } |
1925 | ||
8f646c92 ED |
1926 | if (fdst) |
1927 | vxlan_xmit_one(skb, dev, fdst, did_rsc); | |
1928 | else | |
1929 | kfree_skb(skb); | |
4ad16930 | 1930 | return NETDEV_TX_OK; |
6681712d DS |
1931 | } |
1932 | ||
d342894c | 1933 | /* Walk the forwarding table and purge stale entries */ |
1934 | static void vxlan_cleanup(unsigned long arg) | |
1935 | { | |
1936 | struct vxlan_dev *vxlan = (struct vxlan_dev *) arg; | |
1937 | unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; | |
1938 | unsigned int h; | |
1939 | ||
1940 | if (!netif_running(vxlan->dev)) | |
1941 | return; | |
1942 | ||
1943 | spin_lock_bh(&vxlan->hash_lock); | |
1944 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
1945 | struct hlist_node *p, *n; | |
1946 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { | |
1947 | struct vxlan_fdb *f | |
1948 | = container_of(p, struct vxlan_fdb, hlist); | |
1949 | unsigned long timeout; | |
1950 | ||
3c172868 | 1951 | if (f->state & NUD_PERMANENT) |
d342894c | 1952 | continue; |
1953 | ||
1954 | timeout = f->used + vxlan->age_interval * HZ; | |
1955 | if (time_before_eq(timeout, jiffies)) { | |
1956 | netdev_dbg(vxlan->dev, | |
1957 | "garbage collect %pM\n", | |
1958 | f->eth_addr); | |
1959 | f->state = NUD_STALE; | |
1960 | vxlan_fdb_destroy(vxlan, f); | |
1961 | } else if (time_before(timeout, next_timer)) | |
1962 | next_timer = timeout; | |
1963 | } | |
1964 | } | |
1965 | spin_unlock_bh(&vxlan->hash_lock); | |
1966 | ||
1967 | mod_timer(&vxlan->age_timer, next_timer); | |
1968 | } | |
1969 | ||
9c2e24e1 PS |
1970 | static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan) |
1971 | { | |
1972 | __u32 vni = vxlan->default_dst.remote_vni; | |
1973 | ||
1974 | vxlan->vn_sock = vs; | |
1975 | hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni)); | |
1976 | } | |
1977 | ||
d342894c | 1978 | /* Setup stats when device is created */ |
1979 | static int vxlan_init(struct net_device *dev) | |
1980 | { | |
1c51a915 SH |
1981 | struct vxlan_dev *vxlan = netdev_priv(dev); |
1982 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); | |
1983 | struct vxlan_sock *vs; | |
827da44c | 1984 | int i; |
1c51a915 | 1985 | |
8f84985f | 1986 | dev->tstats = alloc_percpu(struct pcpu_sw_netstats); |
e8171045 | 1987 | if (!dev->tstats) |
d342894c | 1988 | return -ENOMEM; |
1989 | ||
827da44c | 1990 | for_each_possible_cpu(i) { |
8f84985f | 1991 | struct pcpu_sw_netstats *vxlan_stats; |
827da44c JS |
1992 | vxlan_stats = per_cpu_ptr(dev->tstats, i); |
1993 | u64_stats_init(&vxlan_stats->syncp); | |
1994 | } | |
1995 | ||
1996 | ||
1c51a915 | 1997 | spin_lock(&vn->sock_lock); |
9c2e24e1 | 1998 | vs = vxlan_find_sock(dev_net(dev), vxlan->dst_port); |
1c51a915 SH |
1999 | if (vs) { |
2000 | /* If we have a socket with same port already, reuse it */ | |
2001 | atomic_inc(&vs->refcnt); | |
9c2e24e1 | 2002 | vxlan_vs_add_dev(vs, vxlan); |
1c51a915 SH |
2003 | } else { |
2004 | /* otherwise make new socket outside of RTNL */ | |
2005 | dev_hold(dev); | |
2006 | queue_work(vxlan_wq, &vxlan->sock_work); | |
2007 | } | |
2008 | spin_unlock(&vn->sock_lock); | |
2009 | ||
d342894c | 2010 | return 0; |
2011 | } | |
2012 | ||
ba609e9b | 2013 | static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan) |
afbd8bae MR |
2014 | { |
2015 | struct vxlan_fdb *f; | |
2016 | ||
2017 | spin_lock_bh(&vxlan->hash_lock); | |
2018 | f = __vxlan_find_mac(vxlan, all_zeros_mac); | |
2019 | if (f) | |
2020 | vxlan_fdb_destroy(vxlan, f); | |
2021 | spin_unlock_bh(&vxlan->hash_lock); | |
2022 | } | |
2023 | ||
ebf4063e SH |
2024 | static void vxlan_uninit(struct net_device *dev) |
2025 | { | |
2026 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
ebf4063e SH |
2027 | struct vxlan_sock *vs = vxlan->vn_sock; |
2028 | ||
ba609e9b | 2029 | vxlan_fdb_delete_default(vxlan); |
afbd8bae | 2030 | |
ebf4063e | 2031 | if (vs) |
012a5729 | 2032 | vxlan_sock_release(vs); |
ebf4063e SH |
2033 | free_percpu(dev->tstats); |
2034 | } | |
2035 | ||
d342894c | 2036 | /* Start ageing timer and join group when device is brought up */ |
2037 | static int vxlan_open(struct net_device *dev) | |
2038 | { | |
2039 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1c51a915 SH |
2040 | struct vxlan_sock *vs = vxlan->vn_sock; |
2041 | ||
2042 | /* socket hasn't been created */ | |
2043 | if (!vs) | |
2044 | return -ENOTCONN; | |
d342894c | 2045 | |
79d4a94f | 2046 | if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) { |
1c51a915 | 2047 | vxlan_sock_hold(vs); |
7c47cedf | 2048 | dev_hold(dev); |
3fc2de2f | 2049 | queue_work(vxlan_wq, &vxlan->igmp_join); |
d342894c | 2050 | } |
2051 | ||
2052 | if (vxlan->age_interval) | |
2053 | mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); | |
2054 | ||
2055 | return 0; | |
2056 | } | |
2057 | ||
2058 | /* Purge the forwarding table */ | |
2059 | static void vxlan_flush(struct vxlan_dev *vxlan) | |
2060 | { | |
31fec5aa | 2061 | unsigned int h; |
d342894c | 2062 | |
2063 | spin_lock_bh(&vxlan->hash_lock); | |
2064 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
2065 | struct hlist_node *p, *n; | |
2066 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { | |
2067 | struct vxlan_fdb *f | |
2068 | = container_of(p, struct vxlan_fdb, hlist); | |
afbd8bae MR |
2069 | /* the all_zeros_mac entry is deleted at vxlan_uninit */ |
2070 | if (!is_zero_ether_addr(f->eth_addr)) | |
2071 | vxlan_fdb_destroy(vxlan, f); | |
d342894c | 2072 | } |
2073 | } | |
2074 | spin_unlock_bh(&vxlan->hash_lock); | |
2075 | } | |
2076 | ||
2077 | /* Cleanup timer and forwarding table on shutdown */ | |
2078 | static int vxlan_stop(struct net_device *dev) | |
2079 | { | |
3fc2de2f | 2080 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); |
d342894c | 2081 | struct vxlan_dev *vxlan = netdev_priv(dev); |
1c51a915 | 2082 | struct vxlan_sock *vs = vxlan->vn_sock; |
d342894c | 2083 | |
e4c7ed41 | 2084 | if (vs && vxlan_addr_multicast(&vxlan->default_dst.remote_ip) && |
95ab0991 | 2085 | !vxlan_group_used(vn, vxlan)) { |
1c51a915 | 2086 | vxlan_sock_hold(vs); |
7c47cedf | 2087 | dev_hold(dev); |
3fc2de2f | 2088 | queue_work(vxlan_wq, &vxlan->igmp_leave); |
7c47cedf | 2089 | } |
d342894c | 2090 | |
2091 | del_timer_sync(&vxlan->age_timer); | |
2092 | ||
2093 | vxlan_flush(vxlan); | |
2094 | ||
2095 | return 0; | |
2096 | } | |
2097 | ||
d342894c | 2098 | /* Stub, nothing needs to be done. */ |
2099 | static void vxlan_set_multicast_list(struct net_device *dev) | |
2100 | { | |
2101 | } | |
2102 | ||
345010b5 DB |
2103 | static int vxlan_change_mtu(struct net_device *dev, int new_mtu) |
2104 | { | |
2105 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
2106 | struct vxlan_rdst *dst = &vxlan->default_dst; | |
2107 | struct net_device *lowerdev; | |
2108 | int max_mtu; | |
2109 | ||
2110 | lowerdev = __dev_get_by_index(dev_net(dev), dst->remote_ifindex); | |
2111 | if (lowerdev == NULL) | |
2112 | return eth_change_mtu(dev, new_mtu); | |
2113 | ||
2114 | if (dst->remote_ip.sa.sa_family == AF_INET6) | |
2115 | max_mtu = lowerdev->mtu - VXLAN6_HEADROOM; | |
2116 | else | |
2117 | max_mtu = lowerdev->mtu - VXLAN_HEADROOM; | |
2118 | ||
2119 | if (new_mtu < 68 || new_mtu > max_mtu) | |
2120 | return -EINVAL; | |
2121 | ||
2122 | dev->mtu = new_mtu; | |
2123 | return 0; | |
2124 | } | |
2125 | ||
d342894c | 2126 | static const struct net_device_ops vxlan_netdev_ops = { |
2127 | .ndo_init = vxlan_init, | |
ebf4063e | 2128 | .ndo_uninit = vxlan_uninit, |
d342894c | 2129 | .ndo_open = vxlan_open, |
2130 | .ndo_stop = vxlan_stop, | |
2131 | .ndo_start_xmit = vxlan_xmit, | |
e8171045 | 2132 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
d342894c | 2133 | .ndo_set_rx_mode = vxlan_set_multicast_list, |
345010b5 | 2134 | .ndo_change_mtu = vxlan_change_mtu, |
d342894c | 2135 | .ndo_validate_addr = eth_validate_addr, |
2136 | .ndo_set_mac_address = eth_mac_addr, | |
2137 | .ndo_fdb_add = vxlan_fdb_add, | |
2138 | .ndo_fdb_del = vxlan_fdb_delete, | |
2139 | .ndo_fdb_dump = vxlan_fdb_dump, | |
2140 | }; | |
2141 | ||
2142 | /* Info for udev, that this is a virtual tunnel endpoint */ | |
2143 | static struct device_type vxlan_type = { | |
2144 | .name = "vxlan", | |
2145 | }; | |
2146 | ||
53cf5275 | 2147 | /* Calls the ndo_add_vxlan_port of the caller in order to |
35e42379 JG |
2148 | * supply the listening VXLAN udp ports. Callers are expected |
2149 | * to implement the ndo_add_vxlan_port. | |
53cf5275 JG |
2150 | */ |
2151 | void vxlan_get_rx_port(struct net_device *dev) | |
2152 | { | |
2153 | struct vxlan_sock *vs; | |
2154 | struct net *net = dev_net(dev); | |
2155 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
2156 | sa_family_t sa_family; | |
35e42379 JG |
2157 | __be16 port; |
2158 | unsigned int i; | |
53cf5275 JG |
2159 | |
2160 | spin_lock(&vn->sock_lock); | |
2161 | for (i = 0; i < PORT_HASH_SIZE; ++i) { | |
35e42379 JG |
2162 | hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { |
2163 | port = inet_sk(vs->sock->sk)->inet_sport; | |
53cf5275 JG |
2164 | sa_family = vs->sock->sk->sk_family; |
2165 | dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family, | |
2166 | port); | |
2167 | } | |
2168 | } | |
2169 | spin_unlock(&vn->sock_lock); | |
2170 | } | |
2171 | EXPORT_SYMBOL_GPL(vxlan_get_rx_port); | |
2172 | ||
d342894c | 2173 | /* Initialize the device structure. */ |
2174 | static void vxlan_setup(struct net_device *dev) | |
2175 | { | |
2176 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
31fec5aa | 2177 | unsigned int h; |
05f47d69 | 2178 | int low, high; |
d342894c | 2179 | |
2180 | eth_hw_addr_random(dev); | |
2181 | ether_setup(dev); | |
e4c7ed41 CW |
2182 | if (vxlan->default_dst.remote_ip.sa.sa_family == AF_INET6) |
2183 | dev->hard_header_len = ETH_HLEN + VXLAN6_HEADROOM; | |
2184 | else | |
2185 | dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM; | |
d342894c | 2186 | |
2187 | dev->netdev_ops = &vxlan_netdev_ops; | |
ebf4063e | 2188 | dev->destructor = free_netdev; |
d342894c | 2189 | SET_NETDEV_DEVTYPE(dev, &vxlan_type); |
2190 | ||
2191 | dev->tx_queue_len = 0; | |
2192 | dev->features |= NETIF_F_LLTX; | |
2193 | dev->features |= NETIF_F_NETNS_LOCAL; | |
d6727fe3 | 2194 | dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; |
0afb1666 | 2195 | dev->features |= NETIF_F_RXCSUM; |
05c0db08 | 2196 | dev->features |= NETIF_F_GSO_SOFTWARE; |
0afb1666 | 2197 | |
1eaa8178 PS |
2198 | dev->vlan_features = dev->features; |
2199 | dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; | |
0afb1666 | 2200 | dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM; |
05c0db08 | 2201 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; |
1eaa8178 | 2202 | dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; |
d342894c | 2203 | dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
6602d007 | 2204 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
d342894c | 2205 | |
553675fb | 2206 | INIT_LIST_HEAD(&vxlan->next); |
d342894c | 2207 | spin_lock_init(&vxlan->hash_lock); |
3fc2de2f | 2208 | INIT_WORK(&vxlan->igmp_join, vxlan_igmp_join); |
2209 | INIT_WORK(&vxlan->igmp_leave, vxlan_igmp_leave); | |
1c51a915 | 2210 | INIT_WORK(&vxlan->sock_work, vxlan_sock_work); |
d342894c | 2211 | |
2212 | init_timer_deferrable(&vxlan->age_timer); | |
2213 | vxlan->age_timer.function = vxlan_cleanup; | |
2214 | vxlan->age_timer.data = (unsigned long) vxlan; | |
2215 | ||
0bbf87d8 | 2216 | inet_get_local_port_range(dev_net(dev), &low, &high); |
05f47d69 | 2217 | vxlan->port_min = low; |
2218 | vxlan->port_max = high; | |
823aa873 | 2219 | vxlan->dst_port = htons(vxlan_port); |
05f47d69 | 2220 | |
d342894c | 2221 | vxlan->dev = dev; |
2222 | ||
2223 | for (h = 0; h < FDB_HASH_SIZE; ++h) | |
2224 | INIT_HLIST_HEAD(&vxlan->fdb_head[h]); | |
2225 | } | |
2226 | ||
2227 | static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { | |
2228 | [IFLA_VXLAN_ID] = { .type = NLA_U32 }, | |
5d174dd8 | 2229 | [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, |
e4c7ed41 | 2230 | [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, |
d342894c | 2231 | [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, |
2232 | [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, | |
e4c7ed41 | 2233 | [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, |
d342894c | 2234 | [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, |
2235 | [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, | |
2236 | [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, | |
2237 | [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, | |
2238 | [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, | |
05f47d69 | 2239 | [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, |
e4f67add DS |
2240 | [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, |
2241 | [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, | |
2242 | [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, | |
2243 | [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, | |
823aa873 | 2244 | [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, |
d342894c | 2245 | }; |
2246 | ||
2247 | static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[]) | |
2248 | { | |
2249 | if (tb[IFLA_ADDRESS]) { | |
2250 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { | |
2251 | pr_debug("invalid link address (not ethernet)\n"); | |
2252 | return -EINVAL; | |
2253 | } | |
2254 | ||
2255 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { | |
2256 | pr_debug("invalid all zero ethernet address\n"); | |
2257 | return -EADDRNOTAVAIL; | |
2258 | } | |
2259 | } | |
2260 | ||
2261 | if (!data) | |
2262 | return -EINVAL; | |
2263 | ||
2264 | if (data[IFLA_VXLAN_ID]) { | |
2265 | __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); | |
2266 | if (id >= VXLAN_VID_MASK) | |
2267 | return -ERANGE; | |
2268 | } | |
2269 | ||
05f47d69 | 2270 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
2271 | const struct ifla_vxlan_port_range *p | |
2272 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
2273 | ||
2274 | if (ntohs(p->high) < ntohs(p->low)) { | |
2275 | pr_debug("port range %u .. %u not valid\n", | |
2276 | ntohs(p->low), ntohs(p->high)); | |
2277 | return -EINVAL; | |
2278 | } | |
2279 | } | |
2280 | ||
d342894c | 2281 | return 0; |
2282 | } | |
2283 | ||
1b13c97f YB |
2284 | static void vxlan_get_drvinfo(struct net_device *netdev, |
2285 | struct ethtool_drvinfo *drvinfo) | |
2286 | { | |
2287 | strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); | |
2288 | strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); | |
2289 | } | |
2290 | ||
2291 | static const struct ethtool_ops vxlan_ethtool_ops = { | |
2292 | .get_drvinfo = vxlan_get_drvinfo, | |
2293 | .get_link = ethtool_op_get_link, | |
2294 | }; | |
2295 | ||
553675fb | 2296 | static void vxlan_del_work(struct work_struct *work) |
2297 | { | |
2298 | struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work); | |
2299 | ||
2300 | sk_release_kernel(vs->sock->sk); | |
2301 | kfree_rcu(vs, rcu); | |
2302 | } | |
2303 | ||
e4c7ed41 CW |
2304 | #if IS_ENABLED(CONFIG_IPV6) |
2305 | /* Create UDP socket for encapsulation receive. AF_INET6 socket | |
2306 | * could be used for both IPv4 and IPv6 communications, but | |
2307 | * users may set bindv6only=1. | |
2308 | */ | |
39deb2c7 | 2309 | static struct socket *create_v6_sock(struct net *net, __be16 port) |
553675fb | 2310 | { |
553675fb | 2311 | struct sock *sk; |
e4c7ed41 CW |
2312 | struct socket *sock; |
2313 | struct sockaddr_in6 vxlan_addr = { | |
2314 | .sin6_family = AF_INET6, | |
2315 | .sin6_port = port, | |
2316 | }; | |
2317 | int rc, val = 1; | |
2318 | ||
2319 | rc = sock_create_kern(AF_INET6, SOCK_DGRAM, IPPROTO_UDP, &sock); | |
2320 | if (rc < 0) { | |
2321 | pr_debug("UDPv6 socket create failed\n"); | |
39deb2c7 | 2322 | return ERR_PTR(rc); |
e4c7ed41 CW |
2323 | } |
2324 | ||
2325 | /* Put in proper namespace */ | |
2326 | sk = sock->sk; | |
2327 | sk_change_net(sk, net); | |
2328 | ||
2329 | kernel_setsockopt(sock, SOL_IPV6, IPV6_V6ONLY, | |
2330 | (char *)&val, sizeof(val)); | |
2331 | rc = kernel_bind(sock, (struct sockaddr *)&vxlan_addr, | |
2332 | sizeof(struct sockaddr_in6)); | |
2333 | if (rc < 0) { | |
2334 | pr_debug("bind for UDPv6 socket %pI6:%u (%d)\n", | |
2335 | &vxlan_addr.sin6_addr, ntohs(vxlan_addr.sin6_port), rc); | |
2336 | sk_release_kernel(sk); | |
39deb2c7 | 2337 | return ERR_PTR(rc); |
e4c7ed41 CW |
2338 | } |
2339 | /* At this point, IPv6 module should have been loaded in | |
2340 | * sock_create_kern(). | |
2341 | */ | |
2342 | BUG_ON(!ipv6_stub); | |
2343 | ||
e4c7ed41 CW |
2344 | /* Disable multicast loopback */ |
2345 | inet_sk(sk)->mc_loop = 0; | |
39deb2c7 | 2346 | return sock; |
e4c7ed41 CW |
2347 | } |
2348 | ||
2349 | #else | |
2350 | ||
39deb2c7 | 2351 | static struct socket *create_v6_sock(struct net *net, __be16 port) |
e4c7ed41 | 2352 | { |
39deb2c7 | 2353 | return ERR_PTR(-EPFNOSUPPORT); |
e4c7ed41 CW |
2354 | } |
2355 | #endif | |
2356 | ||
39deb2c7 | 2357 | static struct socket *create_v4_sock(struct net *net, __be16 port) |
e4c7ed41 CW |
2358 | { |
2359 | struct sock *sk; | |
2360 | struct socket *sock; | |
553675fb | 2361 | struct sockaddr_in vxlan_addr = { |
2362 | .sin_family = AF_INET, | |
2363 | .sin_addr.s_addr = htonl(INADDR_ANY), | |
bb3fd687 | 2364 | .sin_port = port, |
553675fb | 2365 | }; |
2366 | int rc; | |
553675fb | 2367 | |
2368 | /* Create UDP socket for encapsulation receive. */ | |
e4c7ed41 | 2369 | rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &sock); |
553675fb | 2370 | if (rc < 0) { |
2371 | pr_debug("UDP socket create failed\n"); | |
39deb2c7 | 2372 | return ERR_PTR(rc); |
553675fb | 2373 | } |
2374 | ||
2375 | /* Put in proper namespace */ | |
e4c7ed41 | 2376 | sk = sock->sk; |
553675fb | 2377 | sk_change_net(sk, net); |
2378 | ||
e4c7ed41 | 2379 | rc = kernel_bind(sock, (struct sockaddr *) &vxlan_addr, |
553675fb | 2380 | sizeof(vxlan_addr)); |
2381 | if (rc < 0) { | |
2382 | pr_debug("bind for UDP socket %pI4:%u (%d)\n", | |
2383 | &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc); | |
2384 | sk_release_kernel(sk); | |
39deb2c7 | 2385 | return ERR_PTR(rc); |
e4c7ed41 CW |
2386 | } |
2387 | ||
e4c7ed41 CW |
2388 | /* Disable multicast loopback */ |
2389 | inet_sk(sk)->mc_loop = 0; | |
39deb2c7 | 2390 | return sock; |
e4c7ed41 CW |
2391 | } |
2392 | ||
2393 | /* Create new listen socket if needed */ | |
2394 | static struct vxlan_sock *vxlan_socket_create(struct net *net, __be16 port, | |
2395 | vxlan_rcv_t *rcv, void *data, bool ipv6) | |
2396 | { | |
2397 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
2398 | struct vxlan_sock *vs; | |
2399 | struct socket *sock; | |
2400 | struct sock *sk; | |
e4c7ed41 CW |
2401 | unsigned int h; |
2402 | ||
dc01e7d3 | 2403 | vs = kzalloc(sizeof(*vs), GFP_KERNEL); |
e4c7ed41 CW |
2404 | if (!vs) |
2405 | return ERR_PTR(-ENOMEM); | |
2406 | ||
2407 | for (h = 0; h < VNI_HASH_SIZE; ++h) | |
2408 | INIT_HLIST_HEAD(&vs->vni_list[h]); | |
2409 | ||
2410 | INIT_WORK(&vs->del_work, vxlan_del_work); | |
2411 | ||
2412 | if (ipv6) | |
39deb2c7 | 2413 | sock = create_v6_sock(net, port); |
e4c7ed41 | 2414 | else |
39deb2c7 ZYW |
2415 | sock = create_v4_sock(net, port); |
2416 | if (IS_ERR(sock)) { | |
553675fb | 2417 | kfree(vs); |
e50fddc8 | 2418 | return ERR_CAST(sock); |
553675fb | 2419 | } |
e4c7ed41 CW |
2420 | |
2421 | vs->sock = sock; | |
2422 | sk = sock->sk; | |
9c2e24e1 | 2423 | atomic_set(&vs->refcnt, 1); |
5cfccc5a | 2424 | vs->rcv = rcv; |
012a5729 | 2425 | vs->data = data; |
559835ea | 2426 | rcu_assign_sk_user_data(vs->sock->sk, vs); |
553675fb | 2427 | |
dc01e7d3 OG |
2428 | /* Initialize the vxlan udp offloads structure */ |
2429 | vs->udp_offloads.port = port; | |
2430 | vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive; | |
2431 | vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete; | |
2432 | ||
9c2e24e1 PS |
2433 | spin_lock(&vn->sock_lock); |
2434 | hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); | |
dc01e7d3 | 2435 | vxlan_notify_add_rx_port(vs); |
9c2e24e1 | 2436 | spin_unlock(&vn->sock_lock); |
553675fb | 2437 | |
2438 | /* Mark socket as an encapsulation socket. */ | |
2439 | udp_sk(sk)->encap_type = 1; | |
2440 | udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv; | |
e4c7ed41 CW |
2441 | #if IS_ENABLED(CONFIG_IPV6) |
2442 | if (ipv6) | |
2443 | ipv6_stub->udpv6_encap_enable(); | |
2444 | else | |
2445 | #endif | |
2446 | udp_encap_enable(); | |
2447 | ||
9c2e24e1 PS |
2448 | return vs; |
2449 | } | |
2450 | ||
012a5729 PS |
2451 | struct vxlan_sock *vxlan_sock_add(struct net *net, __be16 port, |
2452 | vxlan_rcv_t *rcv, void *data, | |
e4c7ed41 | 2453 | bool no_share, bool ipv6) |
9c2e24e1 PS |
2454 | { |
2455 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
2456 | struct vxlan_sock *vs; | |
2457 | ||
e4c7ed41 | 2458 | vs = vxlan_socket_create(net, port, rcv, data, ipv6); |
9c2e24e1 PS |
2459 | if (!IS_ERR(vs)) |
2460 | return vs; | |
553675fb | 2461 | |
012a5729 PS |
2462 | if (no_share) /* Return error if sharing is not allowed. */ |
2463 | return vs; | |
2464 | ||
9c2e24e1 PS |
2465 | spin_lock(&vn->sock_lock); |
2466 | vs = vxlan_find_sock(net, port); | |
5cfccc5a PS |
2467 | if (vs) { |
2468 | if (vs->rcv == rcv) | |
2469 | atomic_inc(&vs->refcnt); | |
2470 | else | |
2471 | vs = ERR_PTR(-EBUSY); | |
2472 | } | |
2473 | spin_unlock(&vn->sock_lock); | |
2474 | ||
2475 | if (!vs) | |
9c2e24e1 PS |
2476 | vs = ERR_PTR(-EINVAL); |
2477 | ||
553675fb | 2478 | return vs; |
2479 | } | |
012a5729 | 2480 | EXPORT_SYMBOL_GPL(vxlan_sock_add); |
553675fb | 2481 | |
1c51a915 SH |
2482 | /* Scheduled at device creation to bind to a socket */ |
2483 | static void vxlan_sock_work(struct work_struct *work) | |
2484 | { | |
9c2e24e1 PS |
2485 | struct vxlan_dev *vxlan = container_of(work, struct vxlan_dev, sock_work); |
2486 | struct net *net = dev_net(vxlan->dev); | |
1c51a915 | 2487 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); |
9c2e24e1 PS |
2488 | __be16 port = vxlan->dst_port; |
2489 | struct vxlan_sock *nvs; | |
1c51a915 | 2490 | |
e4c7ed41 | 2491 | nvs = vxlan_sock_add(net, port, vxlan_rcv, NULL, false, vxlan->flags & VXLAN_F_IPV6); |
1c51a915 | 2492 | spin_lock(&vn->sock_lock); |
9c2e24e1 PS |
2493 | if (!IS_ERR(nvs)) |
2494 | vxlan_vs_add_dev(nvs, vxlan); | |
2495 | spin_unlock(&vn->sock_lock); | |
2496 | ||
2497 | dev_put(vxlan->dev); | |
1c51a915 SH |
2498 | } |
2499 | ||
d342894c | 2500 | static int vxlan_newlink(struct net *net, struct net_device *dev, |
2501 | struct nlattr *tb[], struct nlattr *data[]) | |
2502 | { | |
553675fb | 2503 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); |
d342894c | 2504 | struct vxlan_dev *vxlan = netdev_priv(dev); |
c7995c43 | 2505 | struct vxlan_rdst *dst = &vxlan->default_dst; |
d342894c | 2506 | __u32 vni; |
2507 | int err; | |
e4c7ed41 | 2508 | bool use_ipv6 = false; |
d342894c | 2509 | |
2510 | if (!data[IFLA_VXLAN_ID]) | |
2511 | return -EINVAL; | |
2512 | ||
2513 | vni = nla_get_u32(data[IFLA_VXLAN_ID]); | |
c7995c43 | 2514 | dst->remote_vni = vni; |
d342894c | 2515 | |
e4c7ed41 CW |
2516 | if (data[IFLA_VXLAN_GROUP]) { |
2517 | dst->remote_ip.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_GROUP]); | |
2518 | dst->remote_ip.sa.sa_family = AF_INET; | |
2519 | } else if (data[IFLA_VXLAN_GROUP6]) { | |
2520 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2521 | return -EPFNOSUPPORT; | |
2522 | ||
2523 | nla_memcpy(&dst->remote_ip.sin6.sin6_addr, data[IFLA_VXLAN_GROUP6], | |
2524 | sizeof(struct in6_addr)); | |
2525 | dst->remote_ip.sa.sa_family = AF_INET6; | |
2526 | use_ipv6 = true; | |
2527 | } | |
d342894c | 2528 | |
e4c7ed41 CW |
2529 | if (data[IFLA_VXLAN_LOCAL]) { |
2530 | vxlan->saddr.sin.sin_addr.s_addr = nla_get_be32(data[IFLA_VXLAN_LOCAL]); | |
2531 | vxlan->saddr.sa.sa_family = AF_INET; | |
2532 | } else if (data[IFLA_VXLAN_LOCAL6]) { | |
2533 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2534 | return -EPFNOSUPPORT; | |
2535 | ||
2536 | /* TODO: respect scope id */ | |
2537 | nla_memcpy(&vxlan->saddr.sin6.sin6_addr, data[IFLA_VXLAN_LOCAL6], | |
2538 | sizeof(struct in6_addr)); | |
2539 | vxlan->saddr.sa.sa_family = AF_INET6; | |
2540 | use_ipv6 = true; | |
2541 | } | |
d342894c | 2542 | |
34e02aa1 | 2543 | if (data[IFLA_VXLAN_LINK] && |
c7995c43 | 2544 | (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) { |
34e02aa1 | 2545 | struct net_device *lowerdev |
c7995c43 | 2546 | = __dev_get_by_index(net, dst->remote_ifindex); |
34e02aa1 | 2547 | |
2548 | if (!lowerdev) { | |
c7995c43 | 2549 | pr_info("ifindex %d does not exist\n", dst->remote_ifindex); |
34e02aa1 | 2550 | return -ENODEV; |
2551 | } | |
d342894c | 2552 | |
e4c7ed41 CW |
2553 | #if IS_ENABLED(CONFIG_IPV6) |
2554 | if (use_ipv6) { | |
2555 | struct inet6_dev *idev = __in6_dev_get(lowerdev); | |
2556 | if (idev && idev->cnf.disable_ipv6) { | |
2557 | pr_info("IPv6 is disabled via sysctl\n"); | |
2558 | return -EPERM; | |
2559 | } | |
2560 | vxlan->flags |= VXLAN_F_IPV6; | |
2561 | } | |
2562 | #endif | |
2563 | ||
34e02aa1 | 2564 | if (!tb[IFLA_MTU]) |
e4c7ed41 | 2565 | dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM); |
1ba56fb4 AD |
2566 | |
2567 | /* update header length based on lower device */ | |
2568 | dev->hard_header_len = lowerdev->hard_header_len + | |
e4c7ed41 | 2569 | (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM); |
7bda701e | 2570 | } else if (use_ipv6) |
2571 | vxlan->flags |= VXLAN_F_IPV6; | |
d342894c | 2572 | |
2573 | if (data[IFLA_VXLAN_TOS]) | |
2574 | vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); | |
2575 | ||
afb97186 VB |
2576 | if (data[IFLA_VXLAN_TTL]) |
2577 | vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); | |
2578 | ||
d342894c | 2579 | if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING])) |
e4f67add | 2580 | vxlan->flags |= VXLAN_F_LEARN; |
d342894c | 2581 | |
2582 | if (data[IFLA_VXLAN_AGEING]) | |
2583 | vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); | |
2584 | else | |
2585 | vxlan->age_interval = FDB_AGE_DEFAULT; | |
2586 | ||
e4f67add DS |
2587 | if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY])) |
2588 | vxlan->flags |= VXLAN_F_PROXY; | |
2589 | ||
2590 | if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC])) | |
2591 | vxlan->flags |= VXLAN_F_RSC; | |
2592 | ||
2593 | if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS])) | |
2594 | vxlan->flags |= VXLAN_F_L2MISS; | |
2595 | ||
2596 | if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS])) | |
2597 | vxlan->flags |= VXLAN_F_L3MISS; | |
2598 | ||
d342894c | 2599 | if (data[IFLA_VXLAN_LIMIT]) |
2600 | vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); | |
2601 | ||
05f47d69 | 2602 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
2603 | const struct ifla_vxlan_port_range *p | |
2604 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
2605 | vxlan->port_min = ntohs(p->low); | |
2606 | vxlan->port_max = ntohs(p->high); | |
2607 | } | |
2608 | ||
823aa873 | 2609 | if (data[IFLA_VXLAN_PORT]) |
2610 | vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); | |
2611 | ||
553675fb | 2612 | if (vxlan_find_vni(net, vni, vxlan->dst_port)) { |
2613 | pr_info("duplicate VNI %u\n", vni); | |
2614 | return -EEXIST; | |
2615 | } | |
2616 | ||
1b13c97f YB |
2617 | SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops); |
2618 | ||
2936b6ab SS |
2619 | /* create an fdb entry for a valid default destination */ |
2620 | if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) { | |
2621 | err = vxlan_fdb_create(vxlan, all_zeros_mac, | |
2622 | &vxlan->default_dst.remote_ip, | |
2623 | NUD_REACHABLE|NUD_PERMANENT, | |
2624 | NLM_F_EXCL|NLM_F_CREATE, | |
2625 | vxlan->dst_port, | |
2626 | vxlan->default_dst.remote_vni, | |
2627 | vxlan->default_dst.remote_ifindex, | |
2628 | NTF_SELF); | |
2629 | if (err) | |
2630 | return err; | |
2631 | } | |
d342894c | 2632 | |
afbd8bae MR |
2633 | err = register_netdevice(dev); |
2634 | if (err) { | |
ba609e9b | 2635 | vxlan_fdb_delete_default(vxlan); |
afbd8bae MR |
2636 | return err; |
2637 | } | |
2638 | ||
553675fb | 2639 | list_add(&vxlan->next, &vn->vxlan_list); |
553675fb | 2640 | |
2641 | return 0; | |
d342894c | 2642 | } |
2643 | ||
2644 | static void vxlan_dellink(struct net_device *dev, struct list_head *head) | |
2645 | { | |
fe5c3561 | 2646 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); |
d342894c | 2647 | struct vxlan_dev *vxlan = netdev_priv(dev); |
2648 | ||
fe5c3561 | 2649 | spin_lock(&vn->sock_lock); |
9c2e24e1 PS |
2650 | if (!hlist_unhashed(&vxlan->hlist)) |
2651 | hlist_del_rcu(&vxlan->hlist); | |
fe5c3561 | 2652 | spin_unlock(&vn->sock_lock); |
2653 | ||
553675fb | 2654 | list_del(&vxlan->next); |
d342894c | 2655 | unregister_netdevice_queue(dev, head); |
2656 | } | |
2657 | ||
2658 | static size_t vxlan_get_size(const struct net_device *dev) | |
2659 | { | |
2660 | ||
2661 | return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ | |
e4c7ed41 | 2662 | nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ |
d342894c | 2663 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ |
e4c7ed41 | 2664 | nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ |
d342894c | 2665 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ |
2666 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ | |
2667 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ | |
e4f67add DS |
2668 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ |
2669 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ | |
2670 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ | |
2671 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ | |
d342894c | 2672 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ |
2673 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ | |
05f47d69 | 2674 | nla_total_size(sizeof(struct ifla_vxlan_port_range)) + |
823aa873 | 2675 | nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */ |
d342894c | 2676 | 0; |
2677 | } | |
2678 | ||
2679 | static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
2680 | { | |
2681 | const struct vxlan_dev *vxlan = netdev_priv(dev); | |
c7995c43 | 2682 | const struct vxlan_rdst *dst = &vxlan->default_dst; |
05f47d69 | 2683 | struct ifla_vxlan_port_range ports = { |
2684 | .low = htons(vxlan->port_min), | |
2685 | .high = htons(vxlan->port_max), | |
2686 | }; | |
d342894c | 2687 | |
c7995c43 | 2688 | if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni)) |
d342894c | 2689 | goto nla_put_failure; |
2690 | ||
e4c7ed41 CW |
2691 | if (!vxlan_addr_any(&dst->remote_ip)) { |
2692 | if (dst->remote_ip.sa.sa_family == AF_INET) { | |
2693 | if (nla_put_be32(skb, IFLA_VXLAN_GROUP, | |
2694 | dst->remote_ip.sin.sin_addr.s_addr)) | |
2695 | goto nla_put_failure; | |
2696 | #if IS_ENABLED(CONFIG_IPV6) | |
2697 | } else { | |
2698 | if (nla_put(skb, IFLA_VXLAN_GROUP6, sizeof(struct in6_addr), | |
2699 | &dst->remote_ip.sin6.sin6_addr)) | |
2700 | goto nla_put_failure; | |
2701 | #endif | |
2702 | } | |
2703 | } | |
d342894c | 2704 | |
c7995c43 | 2705 | if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) |
d342894c | 2706 | goto nla_put_failure; |
2707 | ||
e4c7ed41 CW |
2708 | if (!vxlan_addr_any(&vxlan->saddr)) { |
2709 | if (vxlan->saddr.sa.sa_family == AF_INET) { | |
2710 | if (nla_put_be32(skb, IFLA_VXLAN_LOCAL, | |
2711 | vxlan->saddr.sin.sin_addr.s_addr)) | |
2712 | goto nla_put_failure; | |
2713 | #if IS_ENABLED(CONFIG_IPV6) | |
2714 | } else { | |
2715 | if (nla_put(skb, IFLA_VXLAN_LOCAL6, sizeof(struct in6_addr), | |
2716 | &vxlan->saddr.sin6.sin6_addr)) | |
2717 | goto nla_put_failure; | |
2718 | #endif | |
2719 | } | |
2720 | } | |
d342894c | 2721 | |
2722 | if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) || | |
2723 | nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) || | |
e4f67add DS |
2724 | nla_put_u8(skb, IFLA_VXLAN_LEARNING, |
2725 | !!(vxlan->flags & VXLAN_F_LEARN)) || | |
2726 | nla_put_u8(skb, IFLA_VXLAN_PROXY, | |
2727 | !!(vxlan->flags & VXLAN_F_PROXY)) || | |
2728 | nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) || | |
2729 | nla_put_u8(skb, IFLA_VXLAN_L2MISS, | |
2730 | !!(vxlan->flags & VXLAN_F_L2MISS)) || | |
2731 | nla_put_u8(skb, IFLA_VXLAN_L3MISS, | |
2732 | !!(vxlan->flags & VXLAN_F_L3MISS)) || | |
d342894c | 2733 | nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) || |
823aa873 | 2734 | nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) || |
2735 | nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port)) | |
d342894c | 2736 | goto nla_put_failure; |
2737 | ||
05f47d69 | 2738 | if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) |
2739 | goto nla_put_failure; | |
2740 | ||
d342894c | 2741 | return 0; |
2742 | ||
2743 | nla_put_failure: | |
2744 | return -EMSGSIZE; | |
2745 | } | |
2746 | ||
2747 | static struct rtnl_link_ops vxlan_link_ops __read_mostly = { | |
2748 | .kind = "vxlan", | |
2749 | .maxtype = IFLA_VXLAN_MAX, | |
2750 | .policy = vxlan_policy, | |
2751 | .priv_size = sizeof(struct vxlan_dev), | |
2752 | .setup = vxlan_setup, | |
2753 | .validate = vxlan_validate, | |
2754 | .newlink = vxlan_newlink, | |
2755 | .dellink = vxlan_dellink, | |
2756 | .get_size = vxlan_get_size, | |
2757 | .fill_info = vxlan_fill_info, | |
2758 | }; | |
2759 | ||
acaf4e70 DB |
2760 | static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, |
2761 | struct net_device *dev) | |
2762 | { | |
2763 | struct vxlan_dev *vxlan, *next; | |
2764 | LIST_HEAD(list_kill); | |
2765 | ||
2766 | list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { | |
2767 | struct vxlan_rdst *dst = &vxlan->default_dst; | |
2768 | ||
2769 | /* In case we created vxlan device with carrier | |
2770 | * and we loose the carrier due to module unload | |
2771 | * we also need to remove vxlan device. In other | |
2772 | * cases, it's not necessary and remote_ifindex | |
2773 | * is 0 here, so no matches. | |
2774 | */ | |
2775 | if (dst->remote_ifindex == dev->ifindex) | |
2776 | vxlan_dellink(vxlan->dev, &list_kill); | |
2777 | } | |
2778 | ||
2779 | unregister_netdevice_many(&list_kill); | |
2780 | } | |
2781 | ||
2782 | static int vxlan_lowerdev_event(struct notifier_block *unused, | |
2783 | unsigned long event, void *ptr) | |
2784 | { | |
2785 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
783c1463 | 2786 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); |
acaf4e70 | 2787 | |
783c1463 | 2788 | if (event == NETDEV_UNREGISTER) |
acaf4e70 DB |
2789 | vxlan_handle_lowerdev_unregister(vn, dev); |
2790 | ||
2791 | return NOTIFY_DONE; | |
2792 | } | |
2793 | ||
2794 | static struct notifier_block vxlan_notifier_block __read_mostly = { | |
2795 | .notifier_call = vxlan_lowerdev_event, | |
2796 | }; | |
2797 | ||
d342894c | 2798 | static __net_init int vxlan_init_net(struct net *net) |
2799 | { | |
2800 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
31fec5aa | 2801 | unsigned int h; |
d342894c | 2802 | |
553675fb | 2803 | INIT_LIST_HEAD(&vn->vxlan_list); |
1c51a915 | 2804 | spin_lock_init(&vn->sock_lock); |
d342894c | 2805 | |
553675fb | 2806 | for (h = 0; h < PORT_HASH_SIZE; ++h) |
2807 | INIT_HLIST_HEAD(&vn->sock_list[h]); | |
d342894c | 2808 | |
2809 | return 0; | |
2810 | } | |
2811 | ||
d342894c | 2812 | static struct pernet_operations vxlan_net_ops = { |
2813 | .init = vxlan_init_net, | |
d342894c | 2814 | .id = &vxlan_net_id, |
2815 | .size = sizeof(struct vxlan_net), | |
2816 | }; | |
2817 | ||
2818 | static int __init vxlan_init_module(void) | |
2819 | { | |
2820 | int rc; | |
2821 | ||
758c57d1 SH |
2822 | vxlan_wq = alloc_workqueue("vxlan", 0, 0); |
2823 | if (!vxlan_wq) | |
2824 | return -ENOMEM; | |
2825 | ||
d342894c | 2826 | get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); |
2827 | ||
783c1463 | 2828 | rc = register_pernet_subsys(&vxlan_net_ops); |
d342894c | 2829 | if (rc) |
2830 | goto out1; | |
2831 | ||
acaf4e70 | 2832 | rc = register_netdevice_notifier(&vxlan_notifier_block); |
d342894c | 2833 | if (rc) |
2834 | goto out2; | |
2835 | ||
acaf4e70 DB |
2836 | rc = rtnl_link_register(&vxlan_link_ops); |
2837 | if (rc) | |
2838 | goto out3; | |
d342894c | 2839 | |
acaf4e70 DB |
2840 | return 0; |
2841 | out3: | |
2842 | unregister_netdevice_notifier(&vxlan_notifier_block); | |
d342894c | 2843 | out2: |
783c1463 | 2844 | unregister_pernet_subsys(&vxlan_net_ops); |
d342894c | 2845 | out1: |
758c57d1 | 2846 | destroy_workqueue(vxlan_wq); |
d342894c | 2847 | return rc; |
2848 | } | |
7332a13b | 2849 | late_initcall(vxlan_init_module); |
d342894c | 2850 | |
2851 | static void __exit vxlan_cleanup_module(void) | |
2852 | { | |
b7153984 | 2853 | rtnl_link_unregister(&vxlan_link_ops); |
acaf4e70 | 2854 | unregister_netdevice_notifier(&vxlan_notifier_block); |
758c57d1 | 2855 | destroy_workqueue(vxlan_wq); |
783c1463 DB |
2856 | unregister_pernet_subsys(&vxlan_net_ops); |
2857 | /* rcu_barrier() is called by netns */ | |
d342894c | 2858 | } |
2859 | module_exit(vxlan_cleanup_module); | |
2860 | ||
2861 | MODULE_LICENSE("GPL"); | |
2862 | MODULE_VERSION(VXLAN_VERSION); | |
3b8df3c6 | 2863 | MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); |
ead5139a | 2864 | MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); |
d342894c | 2865 | MODULE_ALIAS_RTNL_LINK("vxlan"); |