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> | |
d342894c | 14 | #include <linux/module.h> |
15 | #include <linux/errno.h> | |
16 | #include <linux/slab.h> | |
d342894c | 17 | #include <linux/udp.h> |
18 | #include <linux/igmp.h> | |
d342894c | 19 | #include <linux/if_ether.h> |
1b13c97f | 20 | #include <linux/ethtool.h> |
e4f67add DS |
21 | #include <net/arp.h> |
22 | #include <net/ndisc.h> | |
d342894c | 23 | #include <net/ip.h> |
24 | #include <net/icmp.h> | |
d342894c | 25 | #include <net/rtnetlink.h> |
d342894c | 26 | #include <net/inet_ecn.h> |
27 | #include <net/net_namespace.h> | |
28 | #include <net/netns/generic.h> | |
012a5729 | 29 | #include <net/vxlan.h> |
dc01e7d3 | 30 | #include <net/protocol.h> |
40d29af0 | 31 | |
e4c7ed41 | 32 | #if IS_ENABLED(CONFIG_IPV6) |
e4c7ed41 | 33 | #include <net/ip6_tunnel.h> |
660d98ca | 34 | #include <net/ip6_checksum.h> |
e4c7ed41 | 35 | #endif |
d342894c | 36 | |
37 | #define VXLAN_VERSION "0.1" | |
38 | ||
553675fb | 39 | #define PORT_HASH_BITS 8 |
40 | #define PORT_HASH_SIZE (1<<PORT_HASH_BITS) | |
d342894c | 41 | #define FDB_AGE_DEFAULT 300 /* 5 min */ |
42 | #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ | |
43 | ||
23c578bf | 44 | /* UDP port for VXLAN traffic. |
45 | * The IANA assigned port is 4789, but the Linux default is 8472 | |
234f5b73 | 46 | * for compatibility with early adopters. |
23c578bf | 47 | */ |
9daaa397 SH |
48 | static unsigned short vxlan_port __read_mostly = 8472; |
49 | module_param_named(udp_port, vxlan_port, ushort, 0444); | |
d342894c | 50 | MODULE_PARM_DESC(udp_port, "Destination UDP port"); |
51 | ||
52 | static bool log_ecn_error = true; | |
53 | module_param(log_ecn_error, bool, 0644); | |
54 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); | |
55 | ||
60d9d4c6 | 56 | static int vxlan_net_id; |
0dfbdf41 | 57 | static struct rtnl_link_ops vxlan_link_ops; |
553675fb | 58 | |
7256eac1 | 59 | static const u8 all_zeros_mac[ETH_ALEN + 2]; |
afbd8bae | 60 | |
205f356d | 61 | static int vxlan_sock_add(struct vxlan_dev *vxlan); |
614732ea | 62 | |
553675fb | 63 | /* per-network namespace private data for this module */ |
64 | struct vxlan_net { | |
65 | struct list_head vxlan_list; | |
66 | struct hlist_head sock_list[PORT_HASH_SIZE]; | |
1c51a915 | 67 | spinlock_t sock_lock; |
553675fb | 68 | }; |
69 | ||
d342894c | 70 | /* Forwarding table entry */ |
71 | struct vxlan_fdb { | |
72 | struct hlist_node hlist; /* linked list of entries */ | |
73 | struct rcu_head rcu; | |
74 | unsigned long updated; /* jiffies */ | |
75 | unsigned long used; | |
3e61aa8f | 76 | struct list_head remotes; |
7177a3b0 | 77 | u8 eth_addr[ETH_ALEN]; |
d342894c | 78 | u16 state; /* see ndm_state */ |
ae884082 | 79 | u8 flags; /* see ndm_flags */ |
d342894c | 80 | }; |
81 | ||
d342894c | 82 | /* salt for hash table */ |
83 | static u32 vxlan_salt __read_mostly; | |
84 | ||
ee122c79 TG |
85 | static inline bool vxlan_collect_metadata(struct vxlan_sock *vs) |
86 | { | |
e7030878 TG |
87 | return vs->flags & VXLAN_F_COLLECT_METADATA || |
88 | ip_tunnel_collect_metadata(); | |
ee122c79 TG |
89 | } |
90 | ||
e4c7ed41 CW |
91 | #if IS_ENABLED(CONFIG_IPV6) |
92 | static inline | |
93 | bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b) | |
94 | { | |
f0ef3126 JB |
95 | if (a->sa.sa_family != b->sa.sa_family) |
96 | return false; | |
97 | if (a->sa.sa_family == AF_INET6) | |
98 | return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr); | |
99 | else | |
100 | return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr; | |
e4c7ed41 CW |
101 | } |
102 | ||
103 | static inline bool vxlan_addr_any(const union vxlan_addr *ipa) | |
104 | { | |
f0ef3126 JB |
105 | if (ipa->sa.sa_family == AF_INET6) |
106 | return ipv6_addr_any(&ipa->sin6.sin6_addr); | |
107 | else | |
108 | return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY); | |
e4c7ed41 CW |
109 | } |
110 | ||
111 | static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa) | |
112 | { | |
f0ef3126 JB |
113 | if (ipa->sa.sa_family == AF_INET6) |
114 | return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr); | |
115 | else | |
116 | return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr)); | |
e4c7ed41 CW |
117 | } |
118 | ||
119 | static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla) | |
120 | { | |
f0ef3126 | 121 | if (nla_len(nla) >= sizeof(struct in6_addr)) { |
67b61f6c | 122 | ip->sin6.sin6_addr = nla_get_in6_addr(nla); |
f0ef3126 JB |
123 | ip->sa.sa_family = AF_INET6; |
124 | return 0; | |
125 | } else if (nla_len(nla) >= sizeof(__be32)) { | |
67b61f6c | 126 | ip->sin.sin_addr.s_addr = nla_get_in_addr(nla); |
f0ef3126 JB |
127 | ip->sa.sa_family = AF_INET; |
128 | return 0; | |
129 | } else { | |
130 | return -EAFNOSUPPORT; | |
131 | } | |
e4c7ed41 CW |
132 | } |
133 | ||
134 | static int vxlan_nla_put_addr(struct sk_buff *skb, int attr, | |
f0ef3126 | 135 | const union vxlan_addr *ip) |
e4c7ed41 | 136 | { |
f0ef3126 | 137 | if (ip->sa.sa_family == AF_INET6) |
930345ea | 138 | return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr); |
f0ef3126 | 139 | else |
930345ea | 140 | return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr); |
e4c7ed41 CW |
141 | } |
142 | ||
143 | #else /* !CONFIG_IPV6 */ | |
144 | ||
145 | static inline | |
146 | bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b) | |
147 | { | |
f0ef3126 | 148 | return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr; |
e4c7ed41 CW |
149 | } |
150 | ||
151 | static inline bool vxlan_addr_any(const union vxlan_addr *ipa) | |
152 | { | |
f0ef3126 | 153 | return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY); |
e4c7ed41 CW |
154 | } |
155 | ||
156 | static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa) | |
157 | { | |
f0ef3126 | 158 | return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr)); |
e4c7ed41 CW |
159 | } |
160 | ||
161 | static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla) | |
162 | { | |
f0ef3126 JB |
163 | if (nla_len(nla) >= sizeof(struct in6_addr)) { |
164 | return -EAFNOSUPPORT; | |
165 | } else if (nla_len(nla) >= sizeof(__be32)) { | |
67b61f6c | 166 | ip->sin.sin_addr.s_addr = nla_get_in_addr(nla); |
f0ef3126 JB |
167 | ip->sa.sa_family = AF_INET; |
168 | return 0; | |
169 | } else { | |
170 | return -EAFNOSUPPORT; | |
171 | } | |
e4c7ed41 CW |
172 | } |
173 | ||
174 | static int vxlan_nla_put_addr(struct sk_buff *skb, int attr, | |
f0ef3126 | 175 | const union vxlan_addr *ip) |
e4c7ed41 | 176 | { |
930345ea | 177 | return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr); |
e4c7ed41 CW |
178 | } |
179 | #endif | |
180 | ||
553675fb | 181 | /* Virtual Network hash table head */ |
54bfd872 | 182 | static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni) |
553675fb | 183 | { |
54bfd872 | 184 | return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)]; |
553675fb | 185 | } |
186 | ||
187 | /* Socket hash table head */ | |
188 | static inline struct hlist_head *vs_head(struct net *net, __be16 port) | |
d342894c | 189 | { |
190 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
191 | ||
553675fb | 192 | return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)]; |
193 | } | |
194 | ||
3e61aa8f SH |
195 | /* First remote destination for a forwarding entry. |
196 | * Guaranteed to be non-NULL because remotes are never deleted. | |
197 | */ | |
5ca5461c | 198 | static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb) |
3e61aa8f | 199 | { |
5ca5461c | 200 | return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list); |
201 | } | |
202 | ||
203 | static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb) | |
204 | { | |
205 | return list_first_entry(&fdb->remotes, struct vxlan_rdst, list); | |
3e61aa8f SH |
206 | } |
207 | ||
ac5132d1 TG |
208 | /* Find VXLAN socket based on network namespace, address family and UDP port |
209 | * and enabled unshareable flags. | |
210 | */ | |
211 | static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family, | |
212 | __be16 port, u32 flags) | |
553675fb | 213 | { |
214 | struct vxlan_sock *vs; | |
af33c1ad TH |
215 | |
216 | flags &= VXLAN_F_RCV_FLAGS; | |
553675fb | 217 | |
218 | hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { | |
19ca9fc1 | 219 | if (inet_sk(vs->sock->sk)->inet_sport == port && |
705cc62f | 220 | vxlan_get_sk_family(vs) == family && |
af33c1ad | 221 | vs->flags == flags) |
553675fb | 222 | return vs; |
223 | } | |
224 | return NULL; | |
d342894c | 225 | } |
226 | ||
54bfd872 | 227 | static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni) |
d342894c | 228 | { |
229 | struct vxlan_dev *vxlan; | |
d342894c | 230 | |
b9167b2e JB |
231 | /* For flow based devices, map all packets to VNI 0 */ |
232 | if (vs->flags & VXLAN_F_COLLECT_METADATA) | |
233 | vni = 0; | |
234 | ||
54bfd872 JB |
235 | hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) { |
236 | if (vxlan->default_dst.remote_vni == vni) | |
d342894c | 237 | return vxlan; |
238 | } | |
239 | ||
240 | return NULL; | |
241 | } | |
242 | ||
5cfccc5a | 243 | /* Look up VNI in a per net namespace table */ |
54bfd872 | 244 | static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni, |
ac5132d1 TG |
245 | sa_family_t family, __be16 port, |
246 | u32 flags) | |
5cfccc5a PS |
247 | { |
248 | struct vxlan_sock *vs; | |
249 | ||
ac5132d1 | 250 | vs = vxlan_find_sock(net, family, port, flags); |
5cfccc5a PS |
251 | if (!vs) |
252 | return NULL; | |
253 | ||
54bfd872 | 254 | return vxlan_vs_find_vni(vs, vni); |
5cfccc5a PS |
255 | } |
256 | ||
d342894c | 257 | /* Fill in neighbour message in skbuff. */ |
258 | static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, | |
234f5b73 SH |
259 | const struct vxlan_fdb *fdb, |
260 | u32 portid, u32 seq, int type, unsigned int flags, | |
261 | const struct vxlan_rdst *rdst) | |
d342894c | 262 | { |
263 | unsigned long now = jiffies; | |
264 | struct nda_cacheinfo ci; | |
265 | struct nlmsghdr *nlh; | |
266 | struct ndmsg *ndm; | |
e4f67add | 267 | bool send_ip, send_eth; |
d342894c | 268 | |
269 | nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); | |
270 | if (nlh == NULL) | |
271 | return -EMSGSIZE; | |
272 | ||
273 | ndm = nlmsg_data(nlh); | |
274 | memset(ndm, 0, sizeof(*ndm)); | |
e4f67add DS |
275 | |
276 | send_eth = send_ip = true; | |
277 | ||
278 | if (type == RTM_GETNEIGH) { | |
279 | ndm->ndm_family = AF_INET; | |
e4c7ed41 | 280 | send_ip = !vxlan_addr_any(&rdst->remote_ip); |
e4f67add DS |
281 | send_eth = !is_zero_ether_addr(fdb->eth_addr); |
282 | } else | |
283 | ndm->ndm_family = AF_BRIDGE; | |
d342894c | 284 | ndm->ndm_state = fdb->state; |
285 | ndm->ndm_ifindex = vxlan->dev->ifindex; | |
ae884082 | 286 | ndm->ndm_flags = fdb->flags; |
545469f7 | 287 | ndm->ndm_type = RTN_UNICAST; |
d342894c | 288 | |
193523bf | 289 | if (!net_eq(dev_net(vxlan->dev), vxlan->net) && |
4967082b | 290 | nla_put_s32(skb, NDA_LINK_NETNSID, |
7a0877d4 | 291 | peernet2id_alloc(dev_net(vxlan->dev), vxlan->net))) |
193523bf ND |
292 | goto nla_put_failure; |
293 | ||
e4f67add | 294 | if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr)) |
d342894c | 295 | goto nla_put_failure; |
296 | ||
e4c7ed41 | 297 | if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip)) |
6681712d DS |
298 | goto nla_put_failure; |
299 | ||
0dfbdf41 | 300 | if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port && |
6681712d DS |
301 | nla_put_be16(skb, NDA_PORT, rdst->remote_port)) |
302 | goto nla_put_failure; | |
c7995c43 | 303 | if (rdst->remote_vni != vxlan->default_dst.remote_vni && |
54bfd872 | 304 | nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni))) |
6681712d DS |
305 | goto nla_put_failure; |
306 | if (rdst->remote_ifindex && | |
307 | nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) | |
d342894c | 308 | goto nla_put_failure; |
309 | ||
310 | ci.ndm_used = jiffies_to_clock_t(now - fdb->used); | |
311 | ci.ndm_confirmed = 0; | |
312 | ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated); | |
313 | ci.ndm_refcnt = 0; | |
314 | ||
315 | if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) | |
316 | goto nla_put_failure; | |
317 | ||
053c095a JB |
318 | nlmsg_end(skb, nlh); |
319 | return 0; | |
d342894c | 320 | |
321 | nla_put_failure: | |
322 | nlmsg_cancel(skb, nlh); | |
323 | return -EMSGSIZE; | |
324 | } | |
325 | ||
326 | static inline size_t vxlan_nlmsg_size(void) | |
327 | { | |
328 | return NLMSG_ALIGN(sizeof(struct ndmsg)) | |
329 | + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ | |
e4c7ed41 | 330 | + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ |
73cf3317 | 331 | + nla_total_size(sizeof(__be16)) /* NDA_PORT */ |
6681712d DS |
332 | + nla_total_size(sizeof(__be32)) /* NDA_VNI */ |
333 | + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ | |
4967082b | 334 | + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ |
d342894c | 335 | + nla_total_size(sizeof(struct nda_cacheinfo)); |
336 | } | |
337 | ||
9e4b93f9 ND |
338 | static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, |
339 | struct vxlan_rdst *rd, int type) | |
d342894c | 340 | { |
341 | struct net *net = dev_net(vxlan->dev); | |
342 | struct sk_buff *skb; | |
343 | int err = -ENOBUFS; | |
344 | ||
345 | skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); | |
346 | if (skb == NULL) | |
347 | goto errout; | |
348 | ||
9e4b93f9 | 349 | err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); |
d342894c | 350 | if (err < 0) { |
351 | /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ | |
352 | WARN_ON(err == -EMSGSIZE); | |
353 | kfree_skb(skb); | |
354 | goto errout; | |
355 | } | |
356 | ||
357 | rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); | |
358 | return; | |
359 | errout: | |
360 | if (err < 0) | |
361 | rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); | |
362 | } | |
363 | ||
e4c7ed41 | 364 | static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) |
e4f67add DS |
365 | { |
366 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
bb3fd687 SH |
367 | struct vxlan_fdb f = { |
368 | .state = NUD_STALE, | |
369 | }; | |
370 | struct vxlan_rdst remote = { | |
e4c7ed41 | 371 | .remote_ip = *ipa, /* goes to NDA_DST */ |
54bfd872 | 372 | .remote_vni = cpu_to_be32(VXLAN_N_VID), |
bb3fd687 | 373 | }; |
3e61aa8f | 374 | |
9e4b93f9 | 375 | vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH); |
e4f67add DS |
376 | } |
377 | ||
378 | static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) | |
379 | { | |
bb3fd687 SH |
380 | struct vxlan_fdb f = { |
381 | .state = NUD_STALE, | |
382 | }; | |
9e4b93f9 | 383 | struct vxlan_rdst remote = { }; |
e4f67add | 384 | |
e4f67add DS |
385 | memcpy(f.eth_addr, eth_addr, ETH_ALEN); |
386 | ||
9e4b93f9 | 387 | vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH); |
e4f67add DS |
388 | } |
389 | ||
d342894c | 390 | /* Hash Ethernet address */ |
391 | static u32 eth_hash(const unsigned char *addr) | |
392 | { | |
393 | u64 value = get_unaligned((u64 *)addr); | |
394 | ||
395 | /* only want 6 bytes */ | |
396 | #ifdef __BIG_ENDIAN | |
d342894c | 397 | value >>= 16; |
321fb991 | 398 | #else |
399 | value <<= 16; | |
d342894c | 400 | #endif |
401 | return hash_64(value, FDB_HASH_BITS); | |
402 | } | |
403 | ||
404 | /* Hash chain to use given mac address */ | |
405 | static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, | |
406 | const u8 *mac) | |
407 | { | |
408 | return &vxlan->fdb_head[eth_hash(mac)]; | |
409 | } | |
410 | ||
411 | /* Look up Ethernet address in forwarding table */ | |
014be2c8 | 412 | static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, |
d342894c | 413 | const u8 *mac) |
d342894c | 414 | { |
415 | struct hlist_head *head = vxlan_fdb_head(vxlan, mac); | |
416 | struct vxlan_fdb *f; | |
d342894c | 417 | |
b67bfe0d | 418 | hlist_for_each_entry_rcu(f, head, hlist) { |
7367d0b5 | 419 | if (ether_addr_equal(mac, f->eth_addr)) |
d342894c | 420 | return f; |
421 | } | |
422 | ||
423 | return NULL; | |
424 | } | |
425 | ||
014be2c8 SS |
426 | static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, |
427 | const u8 *mac) | |
428 | { | |
429 | struct vxlan_fdb *f; | |
430 | ||
431 | f = __vxlan_find_mac(vxlan, mac); | |
432 | if (f) | |
433 | f->used = jiffies; | |
434 | ||
435 | return f; | |
436 | } | |
437 | ||
a5e7c10a MR |
438 | /* caller should hold vxlan->hash_lock */ |
439 | static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, | |
e4c7ed41 | 440 | union vxlan_addr *ip, __be16 port, |
54bfd872 | 441 | __be32 vni, __u32 ifindex) |
6681712d | 442 | { |
3e61aa8f | 443 | struct vxlan_rdst *rd; |
6681712d | 444 | |
3e61aa8f | 445 | list_for_each_entry(rd, &f->remotes, list) { |
e4c7ed41 | 446 | if (vxlan_addr_equal(&rd->remote_ip, ip) && |
6681712d DS |
447 | rd->remote_port == port && |
448 | rd->remote_vni == vni && | |
449 | rd->remote_ifindex == ifindex) | |
a5e7c10a | 450 | return rd; |
6681712d | 451 | } |
3e61aa8f | 452 | |
a5e7c10a MR |
453 | return NULL; |
454 | } | |
455 | ||
906dc186 TR |
456 | /* Replace destination of unicast mac */ |
457 | static int vxlan_fdb_replace(struct vxlan_fdb *f, | |
54bfd872 JB |
458 | union vxlan_addr *ip, __be16 port, __be32 vni, |
459 | __u32 ifindex) | |
906dc186 TR |
460 | { |
461 | struct vxlan_rdst *rd; | |
462 | ||
463 | rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); | |
464 | if (rd) | |
465 | return 0; | |
466 | ||
467 | rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); | |
468 | if (!rd) | |
469 | return 0; | |
0c1d70af PA |
470 | |
471 | dst_cache_reset(&rd->dst_cache); | |
e4c7ed41 | 472 | rd->remote_ip = *ip; |
906dc186 TR |
473 | rd->remote_port = port; |
474 | rd->remote_vni = vni; | |
475 | rd->remote_ifindex = ifindex; | |
476 | return 1; | |
477 | } | |
478 | ||
a5e7c10a MR |
479 | /* Add/update destinations for multicast */ |
480 | static int vxlan_fdb_append(struct vxlan_fdb *f, | |
54bfd872 | 481 | union vxlan_addr *ip, __be16 port, __be32 vni, |
9e4b93f9 | 482 | __u32 ifindex, struct vxlan_rdst **rdp) |
a5e7c10a MR |
483 | { |
484 | struct vxlan_rdst *rd; | |
485 | ||
486 | rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); | |
487 | if (rd) | |
488 | return 0; | |
489 | ||
6681712d DS |
490 | rd = kmalloc(sizeof(*rd), GFP_ATOMIC); |
491 | if (rd == NULL) | |
492 | return -ENOBUFS; | |
0c1d70af PA |
493 | |
494 | if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) { | |
495 | kfree(rd); | |
496 | return -ENOBUFS; | |
497 | } | |
498 | ||
e4c7ed41 | 499 | rd->remote_ip = *ip; |
6681712d DS |
500 | rd->remote_port = port; |
501 | rd->remote_vni = vni; | |
502 | rd->remote_ifindex = ifindex; | |
3e61aa8f SH |
503 | |
504 | list_add_tail_rcu(&rd->list, &f->remotes); | |
505 | ||
9e4b93f9 | 506 | *rdp = rd; |
6681712d DS |
507 | return 1; |
508 | } | |
509 | ||
dfd8645e TH |
510 | static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, |
511 | unsigned int off, | |
512 | struct vxlanhdr *vh, size_t hdrlen, | |
54bfd872 JB |
513 | __be32 vni_field, |
514 | struct gro_remcsum *grc, | |
0ace2ca8 | 515 | bool nopartial) |
dfd8645e | 516 | { |
b7fe10e5 | 517 | size_t start, offset; |
dfd8645e TH |
518 | |
519 | if (skb->remcsum_offload) | |
b7fe10e5 | 520 | return vh; |
dfd8645e TH |
521 | |
522 | if (!NAPI_GRO_CB(skb)->csum_valid) | |
523 | return NULL; | |
524 | ||
54bfd872 JB |
525 | start = vxlan_rco_start(vni_field); |
526 | offset = start + vxlan_rco_offset(vni_field); | |
dfd8645e | 527 | |
b7fe10e5 TH |
528 | vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, |
529 | start, offset, grc, nopartial); | |
dfd8645e TH |
530 | |
531 | skb->remcsum_offload = 1; | |
532 | ||
533 | return vh; | |
534 | } | |
535 | ||
5602c48c TH |
536 | static struct sk_buff **vxlan_gro_receive(struct sock *sk, |
537 | struct sk_buff **head, | |
538 | struct sk_buff *skb) | |
dc01e7d3 OG |
539 | { |
540 | struct sk_buff *p, **pp = NULL; | |
541 | struct vxlanhdr *vh, *vh2; | |
9b174d88 | 542 | unsigned int hlen, off_vx; |
dc01e7d3 | 543 | int flush = 1; |
5602c48c | 544 | struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk); |
54bfd872 | 545 | __be32 flags; |
26c4f7da TH |
546 | struct gro_remcsum grc; |
547 | ||
548 | skb_gro_remcsum_init(&grc); | |
dc01e7d3 OG |
549 | |
550 | off_vx = skb_gro_offset(skb); | |
551 | hlen = off_vx + sizeof(*vh); | |
552 | vh = skb_gro_header_fast(skb, off_vx); | |
553 | if (skb_gro_header_hard(skb, hlen)) { | |
554 | vh = skb_gro_header_slow(skb, hlen, off_vx); | |
555 | if (unlikely(!vh)) | |
556 | goto out; | |
557 | } | |
dc01e7d3 | 558 | |
dfd8645e TH |
559 | skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); |
560 | ||
54bfd872 | 561 | flags = vh->vx_flags; |
dfd8645e TH |
562 | |
563 | if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { | |
564 | vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), | |
54bfd872 | 565 | vh->vx_vni, &grc, |
0ace2ca8 TH |
566 | !!(vs->flags & |
567 | VXLAN_F_REMCSUM_NOPARTIAL)); | |
dfd8645e TH |
568 | |
569 | if (!vh) | |
570 | goto out; | |
571 | } | |
572 | ||
b7fe10e5 TH |
573 | skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ |
574 | ||
dc01e7d3 OG |
575 | for (p = *head; p; p = p->next) { |
576 | if (!NAPI_GRO_CB(p)->same_flow) | |
577 | continue; | |
578 | ||
579 | vh2 = (struct vxlanhdr *)(p->data + off_vx); | |
3511494c TG |
580 | if (vh->vx_flags != vh2->vx_flags || |
581 | vh->vx_vni != vh2->vx_vni) { | |
dc01e7d3 OG |
582 | NAPI_GRO_CB(p)->same_flow = 0; |
583 | continue; | |
584 | } | |
dc01e7d3 OG |
585 | } |
586 | ||
9b174d88 | 587 | pp = eth_gro_receive(head, skb); |
c194cf93 | 588 | flush = 0; |
dc01e7d3 | 589 | |
dc01e7d3 | 590 | out: |
26c4f7da | 591 | skb_gro_remcsum_cleanup(skb, &grc); |
dc01e7d3 OG |
592 | NAPI_GRO_CB(skb)->flush |= flush; |
593 | ||
594 | return pp; | |
595 | } | |
596 | ||
5602c48c | 597 | static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff) |
dc01e7d3 | 598 | { |
229740c6 JR |
599 | /* Sets 'skb->inner_mac_header' since we are always called with |
600 | * 'skb->encapsulation' set. | |
601 | */ | |
9b174d88 | 602 | return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); |
dc01e7d3 OG |
603 | } |
604 | ||
d342894c | 605 | /* Add new entry to forwarding table -- assumes lock held */ |
606 | static int vxlan_fdb_create(struct vxlan_dev *vxlan, | |
e4c7ed41 | 607 | const u8 *mac, union vxlan_addr *ip, |
6681712d | 608 | __u16 state, __u16 flags, |
54bfd872 | 609 | __be16 port, __be32 vni, __u32 ifindex, |
ae884082 | 610 | __u8 ndm_flags) |
d342894c | 611 | { |
9e4b93f9 | 612 | struct vxlan_rdst *rd = NULL; |
d342894c | 613 | struct vxlan_fdb *f; |
614 | int notify = 0; | |
615 | ||
014be2c8 | 616 | f = __vxlan_find_mac(vxlan, mac); |
d342894c | 617 | if (f) { |
618 | if (flags & NLM_F_EXCL) { | |
619 | netdev_dbg(vxlan->dev, | |
620 | "lost race to create %pM\n", mac); | |
621 | return -EEXIST; | |
622 | } | |
623 | if (f->state != state) { | |
624 | f->state = state; | |
625 | f->updated = jiffies; | |
626 | notify = 1; | |
627 | } | |
ae884082 DS |
628 | if (f->flags != ndm_flags) { |
629 | f->flags = ndm_flags; | |
630 | f->updated = jiffies; | |
631 | notify = 1; | |
632 | } | |
906dc186 TR |
633 | if ((flags & NLM_F_REPLACE)) { |
634 | /* Only change unicasts */ | |
635 | if (!(is_multicast_ether_addr(f->eth_addr) || | |
636 | is_zero_ether_addr(f->eth_addr))) { | |
60840429 | 637 | notify |= vxlan_fdb_replace(f, ip, port, vni, |
906dc186 | 638 | ifindex); |
906dc186 TR |
639 | } else |
640 | return -EOPNOTSUPP; | |
641 | } | |
6681712d | 642 | if ((flags & NLM_F_APPEND) && |
58e4c767 MR |
643 | (is_multicast_ether_addr(f->eth_addr) || |
644 | is_zero_ether_addr(f->eth_addr))) { | |
9e4b93f9 ND |
645 | int rc = vxlan_fdb_append(f, ip, port, vni, ifindex, |
646 | &rd); | |
6681712d DS |
647 | |
648 | if (rc < 0) | |
649 | return rc; | |
650 | notify |= rc; | |
651 | } | |
d342894c | 652 | } else { |
653 | if (!(flags & NLM_F_CREATE)) | |
654 | return -ENOENT; | |
655 | ||
0dfbdf41 TG |
656 | if (vxlan->cfg.addrmax && |
657 | vxlan->addrcnt >= vxlan->cfg.addrmax) | |
d342894c | 658 | return -ENOSPC; |
659 | ||
906dc186 TR |
660 | /* Disallow replace to add a multicast entry */ |
661 | if ((flags & NLM_F_REPLACE) && | |
662 | (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) | |
663 | return -EOPNOTSUPP; | |
664 | ||
e4c7ed41 | 665 | netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); |
d342894c | 666 | f = kmalloc(sizeof(*f), GFP_ATOMIC); |
667 | if (!f) | |
668 | return -ENOMEM; | |
669 | ||
670 | notify = 1; | |
d342894c | 671 | f->state = state; |
ae884082 | 672 | f->flags = ndm_flags; |
d342894c | 673 | f->updated = f->used = jiffies; |
3e61aa8f | 674 | INIT_LIST_HEAD(&f->remotes); |
d342894c | 675 | memcpy(f->eth_addr, mac, ETH_ALEN); |
676 | ||
9e4b93f9 | 677 | vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); |
3e61aa8f | 678 | |
d342894c | 679 | ++vxlan->addrcnt; |
680 | hlist_add_head_rcu(&f->hlist, | |
681 | vxlan_fdb_head(vxlan, mac)); | |
682 | } | |
683 | ||
9e4b93f9 ND |
684 | if (notify) { |
685 | if (rd == NULL) | |
686 | rd = first_remote_rtnl(f); | |
687 | vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH); | |
688 | } | |
d342894c | 689 | |
690 | return 0; | |
691 | } | |
692 | ||
6706c82e | 693 | static void vxlan_fdb_free(struct rcu_head *head) |
6681712d DS |
694 | { |
695 | struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); | |
3e61aa8f | 696 | struct vxlan_rdst *rd, *nd; |
6681712d | 697 | |
0c1d70af PA |
698 | list_for_each_entry_safe(rd, nd, &f->remotes, list) { |
699 | dst_cache_destroy(&rd->dst_cache); | |
6681712d | 700 | kfree(rd); |
0c1d70af | 701 | } |
6681712d DS |
702 | kfree(f); |
703 | } | |
704 | ||
d342894c | 705 | static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f) |
706 | { | |
707 | netdev_dbg(vxlan->dev, | |
708 | "delete %pM\n", f->eth_addr); | |
709 | ||
710 | --vxlan->addrcnt; | |
9e4b93f9 | 711 | vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH); |
d342894c | 712 | |
713 | hlist_del_rcu(&f->hlist); | |
6681712d | 714 | call_rcu(&f->rcu, vxlan_fdb_free); |
d342894c | 715 | } |
716 | ||
f0b074be | 717 | static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, |
54bfd872 JB |
718 | union vxlan_addr *ip, __be16 *port, __be32 *vni, |
719 | u32 *ifindex) | |
d342894c | 720 | { |
6681712d | 721 | struct net *net = dev_net(vxlan->dev); |
e4c7ed41 | 722 | int err; |
d342894c | 723 | |
f0b074be | 724 | if (tb[NDA_DST]) { |
e4c7ed41 CW |
725 | err = vxlan_nla_get_addr(ip, tb[NDA_DST]); |
726 | if (err) | |
727 | return err; | |
f0b074be | 728 | } else { |
e4c7ed41 CW |
729 | union vxlan_addr *remote = &vxlan->default_dst.remote_ip; |
730 | if (remote->sa.sa_family == AF_INET) { | |
731 | ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); | |
732 | ip->sa.sa_family = AF_INET; | |
733 | #if IS_ENABLED(CONFIG_IPV6) | |
734 | } else { | |
735 | ip->sin6.sin6_addr = in6addr_any; | |
736 | ip->sa.sa_family = AF_INET6; | |
737 | #endif | |
738 | } | |
f0b074be | 739 | } |
d342894c | 740 | |
6681712d | 741 | if (tb[NDA_PORT]) { |
73cf3317 | 742 | if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) |
6681712d | 743 | return -EINVAL; |
f0b074be MR |
744 | *port = nla_get_be16(tb[NDA_PORT]); |
745 | } else { | |
0dfbdf41 | 746 | *port = vxlan->cfg.dst_port; |
f0b074be | 747 | } |
6681712d DS |
748 | |
749 | if (tb[NDA_VNI]) { | |
750 | if (nla_len(tb[NDA_VNI]) != sizeof(u32)) | |
751 | return -EINVAL; | |
54bfd872 | 752 | *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); |
f0b074be MR |
753 | } else { |
754 | *vni = vxlan->default_dst.remote_vni; | |
755 | } | |
6681712d DS |
756 | |
757 | if (tb[NDA_IFINDEX]) { | |
5abb0029 | 758 | struct net_device *tdev; |
6681712d DS |
759 | |
760 | if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) | |
761 | return -EINVAL; | |
f0b074be | 762 | *ifindex = nla_get_u32(tb[NDA_IFINDEX]); |
73763949 | 763 | tdev = __dev_get_by_index(net, *ifindex); |
5abb0029 | 764 | if (!tdev) |
6681712d | 765 | return -EADDRNOTAVAIL; |
f0b074be MR |
766 | } else { |
767 | *ifindex = 0; | |
768 | } | |
769 | ||
770 | return 0; | |
771 | } | |
772 | ||
773 | /* Add static entry (via netlink) */ | |
774 | static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], | |
775 | struct net_device *dev, | |
f6f6424b | 776 | const unsigned char *addr, u16 vid, u16 flags) |
f0b074be MR |
777 | { |
778 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
779 | /* struct net *net = dev_net(vxlan->dev); */ | |
e4c7ed41 | 780 | union vxlan_addr ip; |
f0b074be | 781 | __be16 port; |
54bfd872 JB |
782 | __be32 vni; |
783 | u32 ifindex; | |
f0b074be MR |
784 | int err; |
785 | ||
786 | if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { | |
787 | pr_info("RTM_NEWNEIGH with invalid state %#x\n", | |
788 | ndm->ndm_state); | |
789 | return -EINVAL; | |
790 | } | |
791 | ||
792 | if (tb[NDA_DST] == NULL) | |
793 | return -EINVAL; | |
794 | ||
795 | err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex); | |
796 | if (err) | |
797 | return err; | |
6681712d | 798 | |
5933a7bb MR |
799 | if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) |
800 | return -EAFNOSUPPORT; | |
801 | ||
d342894c | 802 | spin_lock_bh(&vxlan->hash_lock); |
e4c7ed41 | 803 | err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags, |
73cf3317 | 804 | port, vni, ifindex, ndm->ndm_flags); |
d342894c | 805 | spin_unlock_bh(&vxlan->hash_lock); |
806 | ||
807 | return err; | |
808 | } | |
809 | ||
810 | /* Delete entry (via netlink) */ | |
1690be63 VY |
811 | static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], |
812 | struct net_device *dev, | |
f6f6424b | 813 | const unsigned char *addr, u16 vid) |
d342894c | 814 | { |
815 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
816 | struct vxlan_fdb *f; | |
bc7892ba | 817 | struct vxlan_rdst *rd = NULL; |
e4c7ed41 | 818 | union vxlan_addr ip; |
bc7892ba | 819 | __be16 port; |
54bfd872 JB |
820 | __be32 vni; |
821 | u32 ifindex; | |
bc7892ba MR |
822 | int err; |
823 | ||
824 | err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex); | |
825 | if (err) | |
826 | return err; | |
827 | ||
828 | err = -ENOENT; | |
d342894c | 829 | |
830 | spin_lock_bh(&vxlan->hash_lock); | |
831 | f = vxlan_find_mac(vxlan, addr); | |
bc7892ba MR |
832 | if (!f) |
833 | goto out; | |
834 | ||
e4c7ed41 CW |
835 | if (!vxlan_addr_any(&ip)) { |
836 | rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); | |
bc7892ba MR |
837 | if (!rd) |
838 | goto out; | |
839 | } | |
840 | ||
841 | err = 0; | |
842 | ||
843 | /* remove a destination if it's not the only one on the list, | |
844 | * otherwise destroy the fdb entry | |
845 | */ | |
846 | if (rd && !list_is_singular(&f->remotes)) { | |
847 | list_del_rcu(&rd->list); | |
9e4b93f9 | 848 | vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH); |
dcdc7a59 | 849 | kfree_rcu(rd, rcu); |
bc7892ba | 850 | goto out; |
d342894c | 851 | } |
bc7892ba MR |
852 | |
853 | vxlan_fdb_destroy(vxlan, f); | |
854 | ||
855 | out: | |
d342894c | 856 | spin_unlock_bh(&vxlan->hash_lock); |
857 | ||
858 | return err; | |
859 | } | |
860 | ||
861 | /* Dump forwarding table */ | |
862 | static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, | |
5d5eacb3 JHS |
863 | struct net_device *dev, |
864 | struct net_device *filter_dev, int idx) | |
d342894c | 865 | { |
866 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
867 | unsigned int h; | |
868 | ||
869 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
870 | struct vxlan_fdb *f; | |
d342894c | 871 | int err; |
872 | ||
b67bfe0d | 873 | hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { |
6681712d | 874 | struct vxlan_rdst *rd; |
6681712d | 875 | |
3e61aa8f | 876 | list_for_each_entry_rcu(rd, &f->remotes, list) { |
07a51cd3 AW |
877 | if (idx < cb->args[0]) |
878 | goto skip; | |
879 | ||
6681712d DS |
880 | err = vxlan_fdb_info(skb, vxlan, f, |
881 | NETLINK_CB(cb->skb).portid, | |
882 | cb->nlh->nlmsg_seq, | |
883 | RTM_NEWNEIGH, | |
884 | NLM_F_MULTI, rd); | |
472681d5 MM |
885 | if (err < 0) { |
886 | cb->args[1] = err; | |
3e61aa8f | 887 | goto out; |
472681d5 | 888 | } |
3e61aa8f | 889 | skip: |
07a51cd3 AW |
890 | ++idx; |
891 | } | |
d342894c | 892 | } |
893 | } | |
3e61aa8f | 894 | out: |
d342894c | 895 | return idx; |
896 | } | |
897 | ||
898 | /* Watch incoming packets to learn mapping between Ethernet address | |
899 | * and Tunnel endpoint. | |
c4b49512 | 900 | * Return true if packet is bogus and should be dropped. |
d342894c | 901 | */ |
26a41ae6 | 902 | static bool vxlan_snoop(struct net_device *dev, |
e4c7ed41 | 903 | union vxlan_addr *src_ip, const u8 *src_mac) |
d342894c | 904 | { |
905 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
906 | struct vxlan_fdb *f; | |
d342894c | 907 | |
908 | f = vxlan_find_mac(vxlan, src_mac); | |
909 | if (likely(f)) { | |
5ca5461c | 910 | struct vxlan_rdst *rdst = first_remote_rcu(f); |
3e61aa8f | 911 | |
e4c7ed41 | 912 | if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip))) |
26a41ae6 | 913 | return false; |
914 | ||
915 | /* Don't migrate static entries, drop packets */ | |
eb064c3b | 916 | if (f->state & NUD_NOARP) |
26a41ae6 | 917 | return true; |
d342894c | 918 | |
919 | if (net_ratelimit()) | |
920 | netdev_info(dev, | |
e4c7ed41 | 921 | "%pM migrated from %pIS to %pIS\n", |
a4870f79 | 922 | src_mac, &rdst->remote_ip.sa, &src_ip->sa); |
d342894c | 923 | |
e4c7ed41 | 924 | rdst->remote_ip = *src_ip; |
d342894c | 925 | f->updated = jiffies; |
9e4b93f9 | 926 | vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH); |
d342894c | 927 | } else { |
928 | /* learned new entry */ | |
929 | spin_lock(&vxlan->hash_lock); | |
3bf74b1a | 930 | |
931 | /* close off race between vxlan_flush and incoming packets */ | |
932 | if (netif_running(dev)) | |
933 | vxlan_fdb_create(vxlan, src_mac, src_ip, | |
934 | NUD_REACHABLE, | |
935 | NLM_F_EXCL|NLM_F_CREATE, | |
0dfbdf41 | 936 | vxlan->cfg.dst_port, |
3bf74b1a | 937 | vxlan->default_dst.remote_vni, |
938 | 0, NTF_SELF); | |
d342894c | 939 | spin_unlock(&vxlan->hash_lock); |
940 | } | |
26a41ae6 | 941 | |
942 | return false; | |
d342894c | 943 | } |
944 | ||
d342894c | 945 | /* See if multicast group is already in use by other ID */ |
95ab0991 | 946 | static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev) |
d342894c | 947 | { |
553675fb | 948 | struct vxlan_dev *vxlan; |
b1be00a6 | 949 | unsigned short family = dev->default_dst.remote_ip.sa.sa_family; |
d342894c | 950 | |
95ab0991 G |
951 | /* The vxlan_sock is only used by dev, leaving group has |
952 | * no effect on other vxlan devices. | |
953 | */ | |
b1be00a6 JB |
954 | if (family == AF_INET && dev->vn4_sock && |
955 | atomic_read(&dev->vn4_sock->refcnt) == 1) | |
95ab0991 | 956 | return false; |
b1be00a6 JB |
957 | #if IS_ENABLED(CONFIG_IPV6) |
958 | if (family == AF_INET6 && dev->vn6_sock && | |
959 | atomic_read(&dev->vn6_sock->refcnt) == 1) | |
960 | return false; | |
961 | #endif | |
95ab0991 | 962 | |
553675fb | 963 | list_for_each_entry(vxlan, &vn->vxlan_list, next) { |
95ab0991 | 964 | if (!netif_running(vxlan->dev) || vxlan == dev) |
553675fb | 965 | continue; |
d342894c | 966 | |
b1be00a6 | 967 | if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock) |
95ab0991 | 968 | continue; |
b1be00a6 JB |
969 | #if IS_ENABLED(CONFIG_IPV6) |
970 | if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock) | |
971 | continue; | |
972 | #endif | |
95ab0991 G |
973 | |
974 | if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip, | |
975 | &dev->default_dst.remote_ip)) | |
976 | continue; | |
977 | ||
978 | if (vxlan->default_dst.remote_ifindex != | |
979 | dev->default_dst.remote_ifindex) | |
980 | continue; | |
981 | ||
982 | return true; | |
553675fb | 983 | } |
d342894c | 984 | |
985 | return false; | |
986 | } | |
987 | ||
544a773a | 988 | static bool __vxlan_sock_release_prep(struct vxlan_sock *vs) |
7c47cedf | 989 | { |
b1be00a6 | 990 | struct vxlan_net *vn; |
012a5729 | 991 | |
b1be00a6 | 992 | if (!vs) |
544a773a | 993 | return false; |
7c47cedf | 994 | if (!atomic_dec_and_test(&vs->refcnt)) |
544a773a | 995 | return false; |
d342894c | 996 | |
b1be00a6 | 997 | vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id); |
1c51a915 | 998 | spin_lock(&vn->sock_lock); |
7c47cedf | 999 | hlist_del_rcu(&vs->hlist); |
e7b3db5e | 1000 | udp_tunnel_notify_del_rx_port(vs->sock, |
b9adcd69 AD |
1001 | (vs->flags & VXLAN_F_GPE) ? |
1002 | UDP_TUNNEL_TYPE_VXLAN_GPE : | |
e7b3db5e | 1003 | UDP_TUNNEL_TYPE_VXLAN); |
1c51a915 SH |
1004 | spin_unlock(&vn->sock_lock); |
1005 | ||
544a773a | 1006 | return true; |
d342894c | 1007 | } |
1008 | ||
b1be00a6 JB |
1009 | static void vxlan_sock_release(struct vxlan_dev *vxlan) |
1010 | { | |
544a773a | 1011 | bool ipv4 = __vxlan_sock_release_prep(vxlan->vn4_sock); |
b1be00a6 | 1012 | #if IS_ENABLED(CONFIG_IPV6) |
544a773a HFS |
1013 | bool ipv6 = __vxlan_sock_release_prep(vxlan->vn6_sock); |
1014 | #endif | |
1015 | ||
1016 | synchronize_net(); | |
1017 | ||
1018 | if (ipv4) { | |
1019 | udp_tunnel_sock_release(vxlan->vn4_sock->sock); | |
1020 | kfree(vxlan->vn4_sock); | |
1021 | } | |
1022 | ||
1023 | #if IS_ENABLED(CONFIG_IPV6) | |
1024 | if (ipv6) { | |
1025 | udp_tunnel_sock_release(vxlan->vn6_sock->sock); | |
1026 | kfree(vxlan->vn6_sock); | |
1027 | } | |
b1be00a6 JB |
1028 | #endif |
1029 | } | |
1030 | ||
56ef9c90 | 1031 | /* Update multicast group membership when first VNI on |
c4b49512 | 1032 | * multicast address is brought up |
7c47cedf | 1033 | */ |
56ef9c90 | 1034 | static int vxlan_igmp_join(struct vxlan_dev *vxlan) |
d342894c | 1035 | { |
b1be00a6 | 1036 | struct sock *sk; |
e4c7ed41 CW |
1037 | union vxlan_addr *ip = &vxlan->default_dst.remote_ip; |
1038 | int ifindex = vxlan->default_dst.remote_ifindex; | |
149d7549 | 1039 | int ret = -EINVAL; |
d342894c | 1040 | |
e4c7ed41 CW |
1041 | if (ip->sa.sa_family == AF_INET) { |
1042 | struct ip_mreqn mreq = { | |
1043 | .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr, | |
1044 | .imr_ifindex = ifindex, | |
1045 | }; | |
1046 | ||
b1be00a6 JB |
1047 | sk = vxlan->vn4_sock->sock->sk; |
1048 | lock_sock(sk); | |
56ef9c90 | 1049 | ret = ip_mc_join_group(sk, &mreq); |
b1be00a6 | 1050 | release_sock(sk); |
e4c7ed41 CW |
1051 | #if IS_ENABLED(CONFIG_IPV6) |
1052 | } else { | |
b1be00a6 JB |
1053 | sk = vxlan->vn6_sock->sock->sk; |
1054 | lock_sock(sk); | |
56ef9c90 MRL |
1055 | ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex, |
1056 | &ip->sin6.sin6_addr); | |
b1be00a6 | 1057 | release_sock(sk); |
e4c7ed41 CW |
1058 | #endif |
1059 | } | |
3fc2de2f | 1060 | |
56ef9c90 | 1061 | return ret; |
3fc2de2f | 1062 | } |
1063 | ||
1064 | /* Inverse of vxlan_igmp_join when last VNI is brought down */ | |
56ef9c90 | 1065 | static int vxlan_igmp_leave(struct vxlan_dev *vxlan) |
3fc2de2f | 1066 | { |
b1be00a6 | 1067 | struct sock *sk; |
e4c7ed41 CW |
1068 | union vxlan_addr *ip = &vxlan->default_dst.remote_ip; |
1069 | int ifindex = vxlan->default_dst.remote_ifindex; | |
149d7549 | 1070 | int ret = -EINVAL; |
3fc2de2f | 1071 | |
e4c7ed41 CW |
1072 | if (ip->sa.sa_family == AF_INET) { |
1073 | struct ip_mreqn mreq = { | |
1074 | .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr, | |
1075 | .imr_ifindex = ifindex, | |
1076 | }; | |
1077 | ||
b1be00a6 JB |
1078 | sk = vxlan->vn4_sock->sock->sk; |
1079 | lock_sock(sk); | |
56ef9c90 | 1080 | ret = ip_mc_leave_group(sk, &mreq); |
b1be00a6 | 1081 | release_sock(sk); |
e4c7ed41 CW |
1082 | #if IS_ENABLED(CONFIG_IPV6) |
1083 | } else { | |
b1be00a6 JB |
1084 | sk = vxlan->vn6_sock->sock->sk; |
1085 | lock_sock(sk); | |
56ef9c90 MRL |
1086 | ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex, |
1087 | &ip->sin6.sin6_addr); | |
b1be00a6 | 1088 | release_sock(sk); |
e4c7ed41 CW |
1089 | #endif |
1090 | } | |
d342894c | 1091 | |
56ef9c90 | 1092 | return ret; |
d342894c | 1093 | } |
1094 | ||
f14ecebb JB |
1095 | static bool vxlan_remcsum(struct vxlanhdr *unparsed, |
1096 | struct sk_buff *skb, u32 vxflags) | |
dfd8645e | 1097 | { |
7d34fa75 | 1098 | size_t start, offset; |
dfd8645e | 1099 | |
f14ecebb JB |
1100 | if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload) |
1101 | goto out; | |
b7fe10e5 | 1102 | |
f14ecebb JB |
1103 | start = vxlan_rco_start(unparsed->vx_vni); |
1104 | offset = start + vxlan_rco_offset(unparsed->vx_vni); | |
dfd8645e | 1105 | |
7d34fa75 | 1106 | if (!pskb_may_pull(skb, offset + sizeof(u16))) |
be5cfeab | 1107 | return false; |
dfd8645e | 1108 | |
be5cfeab JB |
1109 | skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset, |
1110 | !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL)); | |
f14ecebb JB |
1111 | out: |
1112 | unparsed->vx_flags &= ~VXLAN_HF_RCO; | |
1113 | unparsed->vx_vni &= VXLAN_VNI_MASK; | |
be5cfeab | 1114 | return true; |
dfd8645e TH |
1115 | } |
1116 | ||
f14ecebb | 1117 | static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed, |
64f87d36 | 1118 | struct sk_buff *skb, u32 vxflags, |
10a5af23 | 1119 | struct vxlan_metadata *md) |
3288af08 | 1120 | { |
f14ecebb | 1121 | struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed; |
10a5af23 | 1122 | struct metadata_dst *tun_dst; |
f14ecebb JB |
1123 | |
1124 | if (!(unparsed->vx_flags & VXLAN_HF_GBP)) | |
1125 | goto out; | |
3288af08 | 1126 | |
3288af08 JB |
1127 | md->gbp = ntohs(gbp->policy_id); |
1128 | ||
10a5af23 | 1129 | tun_dst = (struct metadata_dst *)skb_dst(skb); |
810813c4 | 1130 | if (tun_dst) { |
3288af08 | 1131 | tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT; |
810813c4 DM |
1132 | tun_dst->u.tun_info.options_len = sizeof(*md); |
1133 | } | |
3288af08 JB |
1134 | if (gbp->dont_learn) |
1135 | md->gbp |= VXLAN_GBP_DONT_LEARN; | |
1136 | ||
1137 | if (gbp->policy_applied) | |
1138 | md->gbp |= VXLAN_GBP_POLICY_APPLIED; | |
f14ecebb | 1139 | |
64f87d36 JB |
1140 | /* In flow-based mode, GBP is carried in dst_metadata */ |
1141 | if (!(vxflags & VXLAN_F_COLLECT_METADATA)) | |
1142 | skb->mark = md->gbp; | |
f14ecebb JB |
1143 | out: |
1144 | unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS; | |
3288af08 JB |
1145 | } |
1146 | ||
e1e5314d | 1147 | static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed, |
61618eea | 1148 | __be16 *protocol, |
e1e5314d JB |
1149 | struct sk_buff *skb, u32 vxflags) |
1150 | { | |
1151 | struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed; | |
1152 | ||
1153 | /* Need to have Next Protocol set for interfaces in GPE mode. */ | |
1154 | if (!gpe->np_applied) | |
1155 | return false; | |
1156 | /* "The initial version is 0. If a receiver does not support the | |
1157 | * version indicated it MUST drop the packet. | |
1158 | */ | |
1159 | if (gpe->version != 0) | |
1160 | return false; | |
1161 | /* "When the O bit is set to 1, the packet is an OAM packet and OAM | |
1162 | * processing MUST occur." However, we don't implement OAM | |
1163 | * processing, thus drop the packet. | |
1164 | */ | |
1165 | if (gpe->oam_flag) | |
1166 | return false; | |
1167 | ||
1168 | switch (gpe->next_protocol) { | |
1169 | case VXLAN_GPE_NP_IPV4: | |
1170 | *protocol = htons(ETH_P_IP); | |
1171 | break; | |
1172 | case VXLAN_GPE_NP_IPV6: | |
1173 | *protocol = htons(ETH_P_IPV6); | |
1174 | break; | |
1175 | case VXLAN_GPE_NP_ETHERNET: | |
1176 | *protocol = htons(ETH_P_TEB); | |
1177 | break; | |
1178 | default: | |
1179 | return false; | |
1180 | } | |
1181 | ||
1182 | unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS; | |
1183 | return true; | |
1184 | } | |
1185 | ||
1ab016e2 JB |
1186 | static bool vxlan_set_mac(struct vxlan_dev *vxlan, |
1187 | struct vxlan_sock *vs, | |
1188 | struct sk_buff *skb) | |
614732ea | 1189 | { |
614732ea | 1190 | union vxlan_addr saddr; |
614732ea | 1191 | |
614732ea | 1192 | skb_reset_mac_header(skb); |
614732ea TG |
1193 | skb->protocol = eth_type_trans(skb, vxlan->dev); |
1194 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); | |
1195 | ||
1196 | /* Ignore packet loops (and multicast echo) */ | |
1197 | if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) | |
1ab016e2 | 1198 | return false; |
614732ea | 1199 | |
760c6805 | 1200 | /* Get address from the outer IP header */ |
ce212d0f | 1201 | if (vxlan_get_sk_family(vs) == AF_INET) { |
1ab016e2 | 1202 | saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr; |
614732ea TG |
1203 | saddr.sa.sa_family = AF_INET; |
1204 | #if IS_ENABLED(CONFIG_IPV6) | |
1205 | } else { | |
1ab016e2 | 1206 | saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr; |
614732ea TG |
1207 | saddr.sa.sa_family = AF_INET6; |
1208 | #endif | |
1209 | } | |
1210 | ||
1ab016e2 JB |
1211 | if ((vxlan->flags & VXLAN_F_LEARN) && |
1212 | vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source)) | |
1213 | return false; | |
1214 | ||
1215 | return true; | |
1216 | } | |
1217 | ||
760c6805 JB |
1218 | static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph, |
1219 | struct sk_buff *skb) | |
1220 | { | |
1221 | int err = 0; | |
1222 | ||
1223 | if (vxlan_get_sk_family(vs) == AF_INET) | |
1224 | err = IP_ECN_decapsulate(oiph, skb); | |
1225 | #if IS_ENABLED(CONFIG_IPV6) | |
1226 | else | |
1227 | err = IP6_ECN_decapsulate(oiph, skb); | |
1228 | #endif | |
1229 | ||
1230 | if (unlikely(err) && log_ecn_error) { | |
1231 | if (vxlan_get_sk_family(vs) == AF_INET) | |
1232 | net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", | |
1233 | &((struct iphdr *)oiph)->saddr, | |
1234 | ((struct iphdr *)oiph)->tos); | |
1235 | else | |
1236 | net_info_ratelimited("non-ECT from %pI6\n", | |
1237 | &((struct ipv6hdr *)oiph)->saddr); | |
1238 | } | |
1239 | return err <= 1; | |
1240 | } | |
1241 | ||
d342894c | 1242 | /* Callback from net/ipv4/udp.c to receive packets */ |
f2d1968e | 1243 | static int vxlan_rcv(struct sock *sk, struct sk_buff *skb) |
d342894c | 1244 | { |
f2d1968e | 1245 | struct pcpu_sw_netstats *stats; |
c9e78efb | 1246 | struct vxlan_dev *vxlan; |
5cfccc5a | 1247 | struct vxlan_sock *vs; |
f14ecebb | 1248 | struct vxlanhdr unparsed; |
ee122c79 TG |
1249 | struct vxlan_metadata _md; |
1250 | struct vxlan_metadata *md = &_md; | |
61618eea | 1251 | __be16 protocol = htons(ETH_P_TEB); |
e1e5314d | 1252 | bool raw_proto = false; |
f2d1968e | 1253 | void *oiph; |
d342894c | 1254 | |
e1e5314d | 1255 | /* Need UDP and VXLAN header to be present */ |
7ce04758 | 1256 | if (!pskb_may_pull(skb, VXLAN_HLEN)) |
e5aed006 | 1257 | goto drop; |
d342894c | 1258 | |
f14ecebb | 1259 | unparsed = *vxlan_hdr(skb); |
288b01c8 JB |
1260 | /* VNI flag always required to be set */ |
1261 | if (!(unparsed.vx_flags & VXLAN_HF_VNI)) { | |
1262 | netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", | |
1263 | ntohl(vxlan_hdr(skb)->vx_flags), | |
1264 | ntohl(vxlan_hdr(skb)->vx_vni)); | |
1265 | /* Return non vxlan pkt */ | |
e5aed006 | 1266 | goto drop; |
d342894c | 1267 | } |
288b01c8 JB |
1268 | unparsed.vx_flags &= ~VXLAN_HF_VNI; |
1269 | unparsed.vx_vni &= ~VXLAN_VNI_MASK; | |
d342894c | 1270 | |
559835ea | 1271 | vs = rcu_dereference_sk_user_data(sk); |
5cfccc5a | 1272 | if (!vs) |
d342894c | 1273 | goto drop; |
d342894c | 1274 | |
c9e78efb JB |
1275 | vxlan = vxlan_vs_find_vni(vs, vxlan_vni(vxlan_hdr(skb)->vx_vni)); |
1276 | if (!vxlan) | |
1277 | goto drop; | |
1278 | ||
e1e5314d JB |
1279 | /* For backwards compatibility, only allow reserved fields to be |
1280 | * used by VXLAN extensions if explicitly requested. | |
1281 | */ | |
1282 | if (vs->flags & VXLAN_F_GPE) { | |
1283 | if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags)) | |
1284 | goto drop; | |
1285 | raw_proto = true; | |
1286 | } | |
1287 | ||
1288 | if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto, | |
1289 | !net_eq(vxlan->net, dev_net(vxlan->dev)))) | |
1290 | goto drop; | |
c9e78efb | 1291 | |
ee122c79 | 1292 | if (vxlan_collect_metadata(vs)) { |
07dabf20 | 1293 | __be32 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni); |
10a5af23 | 1294 | struct metadata_dst *tun_dst; |
07dabf20 | 1295 | |
c29a70d2 | 1296 | tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY, |
07dabf20 | 1297 | vxlan_vni_to_tun_id(vni), sizeof(*md)); |
c29a70d2 | 1298 | |
ee122c79 TG |
1299 | if (!tun_dst) |
1300 | goto drop; | |
1301 | ||
0f1b7354 | 1302 | md = ip_tunnel_info_opts(&tun_dst->u.tun_info); |
10a5af23 JB |
1303 | |
1304 | skb_dst_set(skb, (struct dst_entry *)tun_dst); | |
ee122c79 TG |
1305 | } else { |
1306 | memset(md, 0, sizeof(*md)); | |
1307 | } | |
1308 | ||
f14ecebb JB |
1309 | if (vs->flags & VXLAN_F_REMCSUM_RX) |
1310 | if (!vxlan_remcsum(&unparsed, skb, vs->flags)) | |
1311 | goto drop; | |
1312 | if (vs->flags & VXLAN_F_GBP) | |
10a5af23 | 1313 | vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md); |
e1e5314d JB |
1314 | /* Note that GBP and GPE can never be active together. This is |
1315 | * ensured in vxlan_dev_configure. | |
1316 | */ | |
3511494c | 1317 | |
f14ecebb | 1318 | if (unparsed.vx_flags || unparsed.vx_vni) { |
3bf39475 TH |
1319 | /* If there are any unprocessed flags remaining treat |
1320 | * this as a malformed packet. This behavior diverges from | |
1321 | * VXLAN RFC (RFC7348) which stipulates that bits in reserved | |
1322 | * in reserved fields are to be ignored. The approach here | |
c4b49512 | 1323 | * maintains compatibility with previous stack code, and also |
3bf39475 TH |
1324 | * is more robust and provides a little more security in |
1325 | * adding extensions to VXLAN. | |
1326 | */ | |
288b01c8 | 1327 | goto drop; |
3bf39475 TH |
1328 | } |
1329 | ||
e1e5314d JB |
1330 | if (!raw_proto) { |
1331 | if (!vxlan_set_mac(vxlan, vs, skb)) | |
1332 | goto drop; | |
1333 | } else { | |
8be0cfa4 | 1334 | skb_reset_mac_header(skb); |
e1e5314d JB |
1335 | skb->dev = vxlan->dev; |
1336 | skb->pkt_type = PACKET_HOST; | |
1337 | } | |
f2d1968e | 1338 | |
f2d1968e JB |
1339 | oiph = skb_network_header(skb); |
1340 | skb_reset_network_header(skb); | |
1341 | ||
1342 | if (!vxlan_ecn_decapsulate(vs, oiph, skb)) { | |
1343 | ++vxlan->dev->stats.rx_frame_errors; | |
1344 | ++vxlan->dev->stats.rx_errors; | |
1345 | goto drop; | |
1346 | } | |
1347 | ||
1348 | stats = this_cpu_ptr(vxlan->dev->tstats); | |
1349 | u64_stats_update_begin(&stats->syncp); | |
1350 | stats->rx_packets++; | |
1351 | stats->rx_bytes += skb->len; | |
1352 | u64_stats_update_end(&stats->syncp); | |
1353 | ||
1354 | gro_cells_receive(&vxlan->gro_cells, skb); | |
5cfccc5a PS |
1355 | return 0; |
1356 | ||
1357 | drop: | |
288b01c8 JB |
1358 | /* Consume bad packet */ |
1359 | kfree_skb(skb); | |
1360 | return 0; | |
5cfccc5a PS |
1361 | } |
1362 | ||
e4f67add DS |
1363 | static int arp_reduce(struct net_device *dev, struct sk_buff *skb) |
1364 | { | |
1365 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1366 | struct arphdr *parp; | |
1367 | u8 *arpptr, *sha; | |
1368 | __be32 sip, tip; | |
1369 | struct neighbour *n; | |
1370 | ||
1371 | if (dev->flags & IFF_NOARP) | |
1372 | goto out; | |
1373 | ||
1374 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) { | |
1375 | dev->stats.tx_dropped++; | |
1376 | goto out; | |
1377 | } | |
1378 | parp = arp_hdr(skb); | |
1379 | ||
1380 | if ((parp->ar_hrd != htons(ARPHRD_ETHER) && | |
1381 | parp->ar_hrd != htons(ARPHRD_IEEE802)) || | |
1382 | parp->ar_pro != htons(ETH_P_IP) || | |
1383 | parp->ar_op != htons(ARPOP_REQUEST) || | |
1384 | parp->ar_hln != dev->addr_len || | |
1385 | parp->ar_pln != 4) | |
1386 | goto out; | |
1387 | arpptr = (u8 *)parp + sizeof(struct arphdr); | |
1388 | sha = arpptr; | |
1389 | arpptr += dev->addr_len; /* sha */ | |
1390 | memcpy(&sip, arpptr, sizeof(sip)); | |
1391 | arpptr += sizeof(sip); | |
1392 | arpptr += dev->addr_len; /* tha */ | |
1393 | memcpy(&tip, arpptr, sizeof(tip)); | |
1394 | ||
1395 | if (ipv4_is_loopback(tip) || | |
1396 | ipv4_is_multicast(tip)) | |
1397 | goto out; | |
1398 | ||
1399 | n = neigh_lookup(&arp_tbl, &tip, dev); | |
1400 | ||
1401 | if (n) { | |
e4f67add DS |
1402 | struct vxlan_fdb *f; |
1403 | struct sk_buff *reply; | |
1404 | ||
1405 | if (!(n->nud_state & NUD_CONNECTED)) { | |
1406 | neigh_release(n); | |
1407 | goto out; | |
1408 | } | |
1409 | ||
1410 | f = vxlan_find_mac(vxlan, n->ha); | |
e4c7ed41 | 1411 | if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { |
e4f67add DS |
1412 | /* bridge-local neighbor */ |
1413 | neigh_release(n); | |
1414 | goto out; | |
1415 | } | |
1416 | ||
1417 | reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, | |
1418 | n->ha, sha); | |
1419 | ||
1420 | neigh_release(n); | |
1421 | ||
7346135d DS |
1422 | if (reply == NULL) |
1423 | goto out; | |
1424 | ||
e4f67add DS |
1425 | skb_reset_mac_header(reply); |
1426 | __skb_pull(reply, skb_network_offset(reply)); | |
1427 | reply->ip_summed = CHECKSUM_UNNECESSARY; | |
1428 | reply->pkt_type = PACKET_HOST; | |
1429 | ||
1430 | if (netif_rx_ni(reply) == NET_RX_DROP) | |
1431 | dev->stats.rx_dropped++; | |
e4c7ed41 CW |
1432 | } else if (vxlan->flags & VXLAN_F_L3MISS) { |
1433 | union vxlan_addr ipa = { | |
1434 | .sin.sin_addr.s_addr = tip, | |
a45e92a5 | 1435 | .sin.sin_family = AF_INET, |
e4c7ed41 CW |
1436 | }; |
1437 | ||
1438 | vxlan_ip_miss(dev, &ipa); | |
1439 | } | |
e4f67add DS |
1440 | out: |
1441 | consume_skb(skb); | |
1442 | return NETDEV_TX_OK; | |
1443 | } | |
1444 | ||
f564f45c | 1445 | #if IS_ENABLED(CONFIG_IPV6) |
4b29dba9 DS |
1446 | static struct sk_buff *vxlan_na_create(struct sk_buff *request, |
1447 | struct neighbour *n, bool isrouter) | |
1448 | { | |
1449 | struct net_device *dev = request->dev; | |
1450 | struct sk_buff *reply; | |
1451 | struct nd_msg *ns, *na; | |
1452 | struct ipv6hdr *pip6; | |
1453 | u8 *daddr; | |
1454 | int na_olen = 8; /* opt hdr + ETH_ALEN for target */ | |
1455 | int ns_olen; | |
1456 | int i, len; | |
1457 | ||
1458 | if (dev == NULL) | |
1459 | return NULL; | |
1460 | ||
1461 | len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + | |
1462 | sizeof(*na) + na_olen + dev->needed_tailroom; | |
1463 | reply = alloc_skb(len, GFP_ATOMIC); | |
1464 | if (reply == NULL) | |
1465 | return NULL; | |
1466 | ||
1467 | reply->protocol = htons(ETH_P_IPV6); | |
1468 | reply->dev = dev; | |
1469 | skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); | |
1470 | skb_push(reply, sizeof(struct ethhdr)); | |
6297b91c | 1471 | skb_reset_mac_header(reply); |
4b29dba9 DS |
1472 | |
1473 | ns = (struct nd_msg *)skb_transport_header(request); | |
1474 | ||
1475 | daddr = eth_hdr(request)->h_source; | |
1476 | ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns); | |
1477 | for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { | |
1478 | if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { | |
1479 | daddr = ns->opt + i + sizeof(struct nd_opt_hdr); | |
1480 | break; | |
1481 | } | |
1482 | } | |
1483 | ||
1484 | /* Ethernet header */ | |
1485 | ether_addr_copy(eth_hdr(reply)->h_dest, daddr); | |
1486 | ether_addr_copy(eth_hdr(reply)->h_source, n->ha); | |
1487 | eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); | |
1488 | reply->protocol = htons(ETH_P_IPV6); | |
1489 | ||
1490 | skb_pull(reply, sizeof(struct ethhdr)); | |
6297b91c | 1491 | skb_reset_network_header(reply); |
4b29dba9 DS |
1492 | skb_put(reply, sizeof(struct ipv6hdr)); |
1493 | ||
1494 | /* IPv6 header */ | |
1495 | ||
1496 | pip6 = ipv6_hdr(reply); | |
1497 | memset(pip6, 0, sizeof(struct ipv6hdr)); | |
1498 | pip6->version = 6; | |
1499 | pip6->priority = ipv6_hdr(request)->priority; | |
1500 | pip6->nexthdr = IPPROTO_ICMPV6; | |
1501 | pip6->hop_limit = 255; | |
1502 | pip6->daddr = ipv6_hdr(request)->saddr; | |
1503 | pip6->saddr = *(struct in6_addr *)n->primary_key; | |
1504 | ||
1505 | skb_pull(reply, sizeof(struct ipv6hdr)); | |
6297b91c | 1506 | skb_reset_transport_header(reply); |
4b29dba9 DS |
1507 | |
1508 | na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen); | |
1509 | ||
1510 | /* Neighbor Advertisement */ | |
1511 | memset(na, 0, sizeof(*na)+na_olen); | |
1512 | na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; | |
1513 | na->icmph.icmp6_router = isrouter; | |
1514 | na->icmph.icmp6_override = 1; | |
1515 | na->icmph.icmp6_solicited = 1; | |
1516 | na->target = ns->target; | |
1517 | ether_addr_copy(&na->opt[2], n->ha); | |
1518 | na->opt[0] = ND_OPT_TARGET_LL_ADDR; | |
1519 | na->opt[1] = na_olen >> 3; | |
1520 | ||
1521 | na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, | |
1522 | &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, | |
1523 | csum_partial(na, sizeof(*na)+na_olen, 0)); | |
1524 | ||
1525 | pip6->payload_len = htons(sizeof(*na)+na_olen); | |
1526 | ||
1527 | skb_push(reply, sizeof(struct ipv6hdr)); | |
1528 | ||
1529 | reply->ip_summed = CHECKSUM_UNNECESSARY; | |
1530 | ||
1531 | return reply; | |
1532 | } | |
1533 | ||
f564f45c CW |
1534 | static int neigh_reduce(struct net_device *dev, struct sk_buff *skb) |
1535 | { | |
1536 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
4b29dba9 | 1537 | struct nd_msg *msg; |
f564f45c CW |
1538 | const struct ipv6hdr *iphdr; |
1539 | const struct in6_addr *saddr, *daddr; | |
4b29dba9 DS |
1540 | struct neighbour *n; |
1541 | struct inet6_dev *in6_dev; | |
f564f45c CW |
1542 | |
1543 | in6_dev = __in6_dev_get(dev); | |
1544 | if (!in6_dev) | |
1545 | goto out; | |
1546 | ||
f564f45c CW |
1547 | iphdr = ipv6_hdr(skb); |
1548 | saddr = &iphdr->saddr; | |
1549 | daddr = &iphdr->daddr; | |
1550 | ||
f564f45c CW |
1551 | msg = (struct nd_msg *)skb_transport_header(skb); |
1552 | if (msg->icmph.icmp6_code != 0 || | |
1553 | msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION) | |
1554 | goto out; | |
1555 | ||
4b29dba9 DS |
1556 | if (ipv6_addr_loopback(daddr) || |
1557 | ipv6_addr_is_multicast(&msg->target)) | |
1558 | goto out; | |
1559 | ||
1560 | n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); | |
f564f45c CW |
1561 | |
1562 | if (n) { | |
1563 | struct vxlan_fdb *f; | |
4b29dba9 | 1564 | struct sk_buff *reply; |
f564f45c CW |
1565 | |
1566 | if (!(n->nud_state & NUD_CONNECTED)) { | |
1567 | neigh_release(n); | |
1568 | goto out; | |
1569 | } | |
1570 | ||
1571 | f = vxlan_find_mac(vxlan, n->ha); | |
1572 | if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { | |
1573 | /* bridge-local neighbor */ | |
1574 | neigh_release(n); | |
1575 | goto out; | |
1576 | } | |
1577 | ||
4b29dba9 DS |
1578 | reply = vxlan_na_create(skb, n, |
1579 | !!(f ? f->flags & NTF_ROUTER : 0)); | |
1580 | ||
f564f45c | 1581 | neigh_release(n); |
4b29dba9 DS |
1582 | |
1583 | if (reply == NULL) | |
1584 | goto out; | |
1585 | ||
1586 | if (netif_rx_ni(reply) == NET_RX_DROP) | |
1587 | dev->stats.rx_dropped++; | |
1588 | ||
f564f45c | 1589 | } else if (vxlan->flags & VXLAN_F_L3MISS) { |
4b29dba9 DS |
1590 | union vxlan_addr ipa = { |
1591 | .sin6.sin6_addr = msg->target, | |
a45e92a5 | 1592 | .sin6.sin6_family = AF_INET6, |
4b29dba9 DS |
1593 | }; |
1594 | ||
f564f45c CW |
1595 | vxlan_ip_miss(dev, &ipa); |
1596 | } | |
1597 | ||
1598 | out: | |
1599 | consume_skb(skb); | |
1600 | return NETDEV_TX_OK; | |
1601 | } | |
1602 | #endif | |
1603 | ||
e4f67add DS |
1604 | static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) |
1605 | { | |
1606 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1607 | struct neighbour *n; | |
e4f67add DS |
1608 | |
1609 | if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) | |
1610 | return false; | |
1611 | ||
1612 | n = NULL; | |
1613 | switch (ntohs(eth_hdr(skb)->h_proto)) { | |
1614 | case ETH_P_IP: | |
e15a00aa CW |
1615 | { |
1616 | struct iphdr *pip; | |
1617 | ||
e4f67add DS |
1618 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
1619 | return false; | |
1620 | pip = ip_hdr(skb); | |
1621 | n = neigh_lookup(&arp_tbl, &pip->daddr, dev); | |
e4c7ed41 CW |
1622 | if (!n && (vxlan->flags & VXLAN_F_L3MISS)) { |
1623 | union vxlan_addr ipa = { | |
1624 | .sin.sin_addr.s_addr = pip->daddr, | |
a45e92a5 | 1625 | .sin.sin_family = AF_INET, |
e4c7ed41 CW |
1626 | }; |
1627 | ||
1628 | vxlan_ip_miss(dev, &ipa); | |
1629 | return false; | |
1630 | } | |
1631 | ||
e4f67add | 1632 | break; |
e15a00aa CW |
1633 | } |
1634 | #if IS_ENABLED(CONFIG_IPV6) | |
1635 | case ETH_P_IPV6: | |
1636 | { | |
1637 | struct ipv6hdr *pip6; | |
1638 | ||
1639 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
1640 | return false; | |
1641 | pip6 = ipv6_hdr(skb); | |
1642 | n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); | |
1643 | if (!n && (vxlan->flags & VXLAN_F_L3MISS)) { | |
1644 | union vxlan_addr ipa = { | |
1645 | .sin6.sin6_addr = pip6->daddr, | |
a45e92a5 | 1646 | .sin6.sin6_family = AF_INET6, |
e15a00aa CW |
1647 | }; |
1648 | ||
1649 | vxlan_ip_miss(dev, &ipa); | |
1650 | return false; | |
1651 | } | |
1652 | ||
1653 | break; | |
1654 | } | |
1655 | #endif | |
e4f67add DS |
1656 | default: |
1657 | return false; | |
1658 | } | |
1659 | ||
1660 | if (n) { | |
1661 | bool diff; | |
1662 | ||
7367d0b5 | 1663 | diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); |
e4f67add DS |
1664 | if (diff) { |
1665 | memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, | |
1666 | dev->addr_len); | |
1667 | memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); | |
1668 | } | |
1669 | neigh_release(n); | |
1670 | return diff; | |
e4c7ed41 CW |
1671 | } |
1672 | ||
e4f67add DS |
1673 | return false; |
1674 | } | |
1675 | ||
af33c1ad | 1676 | static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags, |
3511494c TG |
1677 | struct vxlan_metadata *md) |
1678 | { | |
1679 | struct vxlanhdr_gbp *gbp; | |
1680 | ||
db79a621 TG |
1681 | if (!md->gbp) |
1682 | return; | |
1683 | ||
3511494c | 1684 | gbp = (struct vxlanhdr_gbp *)vxh; |
54bfd872 | 1685 | vxh->vx_flags |= VXLAN_HF_GBP; |
3511494c TG |
1686 | |
1687 | if (md->gbp & VXLAN_GBP_DONT_LEARN) | |
1688 | gbp->dont_learn = 1; | |
1689 | ||
1690 | if (md->gbp & VXLAN_GBP_POLICY_APPLIED) | |
1691 | gbp->policy_applied = 1; | |
1692 | ||
1693 | gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK); | |
1694 | } | |
1695 | ||
e1e5314d JB |
1696 | static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags, |
1697 | __be16 protocol) | |
1698 | { | |
1699 | struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh; | |
1700 | ||
1701 | gpe->np_applied = 1; | |
1702 | ||
1703 | switch (protocol) { | |
1704 | case htons(ETH_P_IP): | |
1705 | gpe->next_protocol = VXLAN_GPE_NP_IPV4; | |
1706 | return 0; | |
1707 | case htons(ETH_P_IPV6): | |
1708 | gpe->next_protocol = VXLAN_GPE_NP_IPV6; | |
1709 | return 0; | |
1710 | case htons(ETH_P_TEB): | |
1711 | gpe->next_protocol = VXLAN_GPE_NP_ETHERNET; | |
1712 | return 0; | |
1713 | } | |
1714 | return -EPFNOSUPPORT; | |
1715 | } | |
1716 | ||
f491e56d JB |
1717 | static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, |
1718 | int iphdr_len, __be32 vni, | |
1719 | struct vxlan_metadata *md, u32 vxflags, | |
b4ed5cad | 1720 | bool udp_sum) |
e4c7ed41 | 1721 | { |
e4c7ed41 | 1722 | struct vxlanhdr *vxh; |
e4c7ed41 CW |
1723 | int min_headroom; |
1724 | int err; | |
dfd8645e | 1725 | int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; |
e1e5314d | 1726 | __be16 inner_protocol = htons(ETH_P_TEB); |
dfd8645e | 1727 | |
af33c1ad | 1728 | if ((vxflags & VXLAN_F_REMCSUM_TX) && |
dfd8645e TH |
1729 | skb->ip_summed == CHECKSUM_PARTIAL) { |
1730 | int csum_start = skb_checksum_start_offset(skb); | |
1731 | ||
1732 | if (csum_start <= VXLAN_MAX_REMCSUM_START && | |
1733 | !(csum_start & VXLAN_RCO_SHIFT_MASK) && | |
1734 | (skb->csum_offset == offsetof(struct udphdr, check) || | |
b5708501 | 1735 | skb->csum_offset == offsetof(struct tcphdr, check))) |
dfd8645e | 1736 | type |= SKB_GSO_TUNNEL_REMCSUM; |
dfd8645e | 1737 | } |
e4c7ed41 | 1738 | |
e4c7ed41 | 1739 | min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len |
f491e56d | 1740 | + VXLAN_HLEN + iphdr_len |
df8a39de | 1741 | + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0); |
649c5b8b PS |
1742 | |
1743 | /* Need space for new headers (invalidates iph ptr) */ | |
1744 | err = skb_cow_head(skb, min_headroom); | |
e1e5314d JB |
1745 | if (unlikely(err)) |
1746 | goto out_free; | |
649c5b8b | 1747 | |
5968250c JP |
1748 | skb = vlan_hwaccel_push_inside(skb); |
1749 | if (WARN_ON(!skb)) | |
1750 | return -ENOMEM; | |
1eaa8178 | 1751 | |
aed069df AD |
1752 | err = iptunnel_handle_offloads(skb, type); |
1753 | if (err) | |
1754 | goto out_free; | |
b736a623 | 1755 | |
49560532 | 1756 | vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh)); |
54bfd872 JB |
1757 | vxh->vx_flags = VXLAN_HF_VNI; |
1758 | vxh->vx_vni = vxlan_vni_field(vni); | |
49560532 | 1759 | |
dfd8645e | 1760 | if (type & SKB_GSO_TUNNEL_REMCSUM) { |
54bfd872 | 1761 | unsigned int start; |
dfd8645e | 1762 | |
54bfd872 JB |
1763 | start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); |
1764 | vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); | |
1765 | vxh->vx_flags |= VXLAN_HF_RCO; | |
dfd8645e TH |
1766 | |
1767 | if (!skb_is_gso(skb)) { | |
1768 | skb->ip_summed = CHECKSUM_NONE; | |
1769 | skb->encapsulation = 0; | |
1770 | } | |
1771 | } | |
1772 | ||
af33c1ad TH |
1773 | if (vxflags & VXLAN_F_GBP) |
1774 | vxlan_build_gbp_hdr(vxh, vxflags, md); | |
e1e5314d JB |
1775 | if (vxflags & VXLAN_F_GPE) { |
1776 | err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol); | |
1777 | if (err < 0) | |
1778 | goto out_free; | |
1779 | inner_protocol = skb->protocol; | |
1780 | } | |
3511494c | 1781 | |
e1e5314d | 1782 | skb_set_inner_protocol(skb, inner_protocol); |
039f5062 | 1783 | return 0; |
e1e5314d JB |
1784 | |
1785 | out_free: | |
1786 | kfree_skb(skb); | |
1787 | return err; | |
49560532 | 1788 | } |
49560532 | 1789 | |
1a8496ba JB |
1790 | static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, |
1791 | struct sk_buff *skb, int oif, u8 tos, | |
0c1d70af PA |
1792 | __be32 daddr, __be32 *saddr, |
1793 | struct dst_cache *dst_cache, | |
db3c6139 | 1794 | const struct ip_tunnel_info *info) |
1a8496ba | 1795 | { |
db3c6139 | 1796 | bool use_cache = ip_tunnel_dst_cache_usable(skb, info); |
1a8496ba JB |
1797 | struct rtable *rt = NULL; |
1798 | struct flowi4 fl4; | |
1799 | ||
db3c6139 DB |
1800 | if (tos && !info) |
1801 | use_cache = false; | |
1802 | if (use_cache) { | |
0c1d70af PA |
1803 | rt = dst_cache_get_ip4(dst_cache, saddr); |
1804 | if (rt) | |
1805 | return rt; | |
1806 | } | |
1807 | ||
1a8496ba JB |
1808 | memset(&fl4, 0, sizeof(fl4)); |
1809 | fl4.flowi4_oif = oif; | |
1810 | fl4.flowi4_tos = RT_TOS(tos); | |
1811 | fl4.flowi4_mark = skb->mark; | |
1812 | fl4.flowi4_proto = IPPROTO_UDP; | |
1813 | fl4.daddr = daddr; | |
272d96a5 | 1814 | fl4.saddr = *saddr; |
1a8496ba JB |
1815 | |
1816 | rt = ip_route_output_key(vxlan->net, &fl4); | |
0c1d70af | 1817 | if (!IS_ERR(rt)) { |
1a8496ba | 1818 | *saddr = fl4.saddr; |
0c1d70af PA |
1819 | if (use_cache) |
1820 | dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr); | |
1821 | } | |
1a8496ba JB |
1822 | return rt; |
1823 | } | |
1824 | ||
e5d4b29f JB |
1825 | #if IS_ENABLED(CONFIG_IPV6) |
1826 | static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan, | |
1400615d | 1827 | struct sk_buff *skb, int oif, u8 tos, |
e7f70af1 | 1828 | __be32 label, |
e5d4b29f | 1829 | const struct in6_addr *daddr, |
0c1d70af | 1830 | struct in6_addr *saddr, |
db3c6139 DB |
1831 | struct dst_cache *dst_cache, |
1832 | const struct ip_tunnel_info *info) | |
e5d4b29f | 1833 | { |
db3c6139 | 1834 | bool use_cache = ip_tunnel_dst_cache_usable(skb, info); |
e5d4b29f JB |
1835 | struct dst_entry *ndst; |
1836 | struct flowi6 fl6; | |
1837 | int err; | |
1838 | ||
1400615d DB |
1839 | if (tos && !info) |
1840 | use_cache = false; | |
db3c6139 | 1841 | if (use_cache) { |
0c1d70af PA |
1842 | ndst = dst_cache_get_ip6(dst_cache, saddr); |
1843 | if (ndst) | |
1844 | return ndst; | |
1845 | } | |
1846 | ||
e5d4b29f JB |
1847 | memset(&fl6, 0, sizeof(fl6)); |
1848 | fl6.flowi6_oif = oif; | |
1849 | fl6.daddr = *daddr; | |
272d96a5 | 1850 | fl6.saddr = *saddr; |
eaa93bf4 | 1851 | fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label); |
e5d4b29f JB |
1852 | fl6.flowi6_mark = skb->mark; |
1853 | fl6.flowi6_proto = IPPROTO_UDP; | |
1854 | ||
1855 | err = ipv6_stub->ipv6_dst_lookup(vxlan->net, | |
1856 | vxlan->vn6_sock->sock->sk, | |
1857 | &ndst, &fl6); | |
1858 | if (err < 0) | |
1859 | return ERR_PTR(err); | |
1860 | ||
1861 | *saddr = fl6.saddr; | |
db3c6139 | 1862 | if (use_cache) |
0c1d70af | 1863 | dst_cache_set_ip6(dst_cache, ndst, saddr); |
e5d4b29f JB |
1864 | return ndst; |
1865 | } | |
1866 | #endif | |
1867 | ||
9dcc71e1 SS |
1868 | /* Bypass encapsulation if the destination is local */ |
1869 | static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, | |
1870 | struct vxlan_dev *dst_vxlan) | |
1871 | { | |
8f84985f | 1872 | struct pcpu_sw_netstats *tx_stats, *rx_stats; |
e4c7ed41 CW |
1873 | union vxlan_addr loopback; |
1874 | union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; | |
ce6502a8 LR |
1875 | struct net_device *dev = skb->dev; |
1876 | int len = skb->len; | |
9dcc71e1 | 1877 | |
8f84985f LR |
1878 | tx_stats = this_cpu_ptr(src_vxlan->dev->tstats); |
1879 | rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats); | |
9dcc71e1 SS |
1880 | skb->pkt_type = PACKET_HOST; |
1881 | skb->encapsulation = 0; | |
1882 | skb->dev = dst_vxlan->dev; | |
1883 | __skb_pull(skb, skb_network_offset(skb)); | |
1884 | ||
e4c7ed41 CW |
1885 | if (remote_ip->sa.sa_family == AF_INET) { |
1886 | loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); | |
1887 | loopback.sa.sa_family = AF_INET; | |
1888 | #if IS_ENABLED(CONFIG_IPV6) | |
1889 | } else { | |
1890 | loopback.sin6.sin6_addr = in6addr_loopback; | |
1891 | loopback.sa.sa_family = AF_INET6; | |
1892 | #endif | |
1893 | } | |
1894 | ||
9dcc71e1 | 1895 | if (dst_vxlan->flags & VXLAN_F_LEARN) |
e4c7ed41 | 1896 | vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source); |
9dcc71e1 SS |
1897 | |
1898 | u64_stats_update_begin(&tx_stats->syncp); | |
1899 | tx_stats->tx_packets++; | |
ce6502a8 | 1900 | tx_stats->tx_bytes += len; |
9dcc71e1 SS |
1901 | u64_stats_update_end(&tx_stats->syncp); |
1902 | ||
1903 | if (netif_rx(skb) == NET_RX_SUCCESS) { | |
1904 | u64_stats_update_begin(&rx_stats->syncp); | |
1905 | rx_stats->rx_packets++; | |
ce6502a8 | 1906 | rx_stats->rx_bytes += len; |
9dcc71e1 SS |
1907 | u64_stats_update_end(&rx_stats->syncp); |
1908 | } else { | |
ce6502a8 | 1909 | dev->stats.rx_dropped++; |
9dcc71e1 SS |
1910 | } |
1911 | } | |
1912 | ||
4ad16930 SH |
1913 | static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, |
1914 | struct vxlan_rdst *rdst, bool did_rsc) | |
d342894c | 1915 | { |
d71785ff | 1916 | struct dst_cache *dst_cache; |
3093fbe7 | 1917 | struct ip_tunnel_info *info; |
d342894c | 1918 | struct vxlan_dev *vxlan = netdev_priv(dev); |
b1be00a6 | 1919 | struct sock *sk; |
e4c7ed41 | 1920 | struct rtable *rt = NULL; |
d342894c | 1921 | const struct iphdr *old_iph; |
e4c7ed41 | 1922 | union vxlan_addr *dst; |
272d96a5 | 1923 | union vxlan_addr remote_ip, local_ip; |
1924 | union vxlan_addr *src; | |
ee122c79 TG |
1925 | struct vxlan_metadata _md; |
1926 | struct vxlan_metadata *md = &_md; | |
e4c7ed41 | 1927 | __be16 src_port = 0, dst_port; |
e7f70af1 | 1928 | __be32 vni, label; |
d342894c | 1929 | __be16 df = 0; |
1930 | __u8 tos, ttl; | |
0e6fbc5b | 1931 | int err; |
ee122c79 | 1932 | u32 flags = vxlan->flags; |
b4ed5cad | 1933 | bool udp_sum = false; |
f491e56d | 1934 | bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); |
d342894c | 1935 | |
61adedf3 | 1936 | info = skb_tunnel_info(skb); |
3093fbe7 | 1937 | |
ee122c79 | 1938 | if (rdst) { |
0dfbdf41 | 1939 | dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; |
ee122c79 TG |
1940 | vni = rdst->remote_vni; |
1941 | dst = &rdst->remote_ip; | |
272d96a5 | 1942 | src = &vxlan->cfg.saddr; |
d71785ff | 1943 | dst_cache = &rdst->dst_cache; |
ee122c79 TG |
1944 | } else { |
1945 | if (!info) { | |
1946 | WARN_ONCE(1, "%s: Missing encapsulation instructions\n", | |
1947 | dev->name); | |
1948 | goto drop; | |
1949 | } | |
0dfbdf41 | 1950 | dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; |
54bfd872 | 1951 | vni = vxlan_tun_id_to_vni(info->key.tun_id); |
b1be00a6 | 1952 | remote_ip.sa.sa_family = ip_tunnel_info_af(info); |
272d96a5 | 1953 | if (remote_ip.sa.sa_family == AF_INET) { |
a725e514 | 1954 | remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst; |
272d96a5 | 1955 | local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src; |
1956 | } else { | |
a725e514 | 1957 | remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst; |
272d96a5 | 1958 | local_ip.sin6.sin6_addr = info->key.u.ipv6.src; |
1959 | } | |
ee122c79 | 1960 | dst = &remote_ip; |
272d96a5 | 1961 | src = &local_ip; |
d71785ff | 1962 | dst_cache = &info->dst_cache; |
ee122c79 | 1963 | } |
e4f67add | 1964 | |
e4c7ed41 | 1965 | if (vxlan_addr_any(dst)) { |
e4f67add | 1966 | if (did_rsc) { |
e4f67add | 1967 | /* short-circuited back to local bridge */ |
9dcc71e1 | 1968 | vxlan_encap_bypass(skb, vxlan, vxlan); |
4ad16930 | 1969 | return; |
e4f67add | 1970 | } |
ef59febe | 1971 | goto drop; |
e4f67add | 1972 | } |
ef59febe | 1973 | |
d342894c | 1974 | old_iph = ip_hdr(skb); |
1975 | ||
0dfbdf41 | 1976 | ttl = vxlan->cfg.ttl; |
e4c7ed41 | 1977 | if (!ttl && vxlan_addr_multicast(dst)) |
d342894c | 1978 | ttl = 1; |
1979 | ||
0dfbdf41 | 1980 | tos = vxlan->cfg.tos; |
d342894c | 1981 | if (tos == 1) |
206aaafc | 1982 | tos = ip_tunnel_get_dsfield(old_iph, skb); |
d342894c | 1983 | |
e7f70af1 | 1984 | label = vxlan->cfg.label; |
0dfbdf41 TG |
1985 | src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, |
1986 | vxlan->cfg.port_max, true); | |
d342894c | 1987 | |
a725e514 | 1988 | if (info) { |
a725e514 JB |
1989 | ttl = info->key.ttl; |
1990 | tos = info->key.tos; | |
e7f70af1 | 1991 | label = info->key.label; |
b4ed5cad | 1992 | udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); |
a725e514 JB |
1993 | |
1994 | if (info->options_len) | |
4c222798 | 1995 | md = ip_tunnel_info_opts(info); |
a725e514 JB |
1996 | } else { |
1997 | md->gbp = skb->mark; | |
1998 | } | |
1999 | ||
e4c7ed41 | 2000 | if (dst->sa.sa_family == AF_INET) { |
b1be00a6 JB |
2001 | if (!vxlan->vn4_sock) |
2002 | goto drop; | |
2003 | sk = vxlan->vn4_sock->sock->sk; | |
2004 | ||
1a8496ba JB |
2005 | rt = vxlan_get_route(vxlan, skb, |
2006 | rdst ? rdst->remote_ifindex : 0, tos, | |
272d96a5 | 2007 | dst->sin.sin_addr.s_addr, |
2008 | &src->sin.sin_addr.s_addr, | |
d71785ff | 2009 | dst_cache, info); |
e4c7ed41 CW |
2010 | if (IS_ERR(rt)) { |
2011 | netdev_dbg(dev, "no route to %pI4\n", | |
2012 | &dst->sin.sin_addr.s_addr); | |
2013 | dev->stats.tx_carrier_errors++; | |
2014 | goto tx_error; | |
2015 | } | |
d342894c | 2016 | |
e4c7ed41 CW |
2017 | if (rt->dst.dev == dev) { |
2018 | netdev_dbg(dev, "circular route to %pI4\n", | |
2019 | &dst->sin.sin_addr.s_addr); | |
2020 | dev->stats.collisions++; | |
fffc15a5 | 2021 | goto rt_tx_error; |
e4c7ed41 CW |
2022 | } |
2023 | ||
2024 | /* Bypass encapsulation if the destination is local */ | |
bbec7802 | 2025 | if (!info && rt->rt_flags & RTCF_LOCAL && |
e4c7ed41 CW |
2026 | !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { |
2027 | struct vxlan_dev *dst_vxlan; | |
2028 | ||
2029 | ip_rt_put(rt); | |
19ca9fc1 | 2030 | dst_vxlan = vxlan_find_vni(vxlan->net, vni, |
ac5132d1 TG |
2031 | dst->sa.sa_family, dst_port, |
2032 | vxlan->flags); | |
e4c7ed41 CW |
2033 | if (!dst_vxlan) |
2034 | goto tx_error; | |
2035 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
2036 | return; | |
2037 | } | |
2038 | ||
6ceb31ca AD |
2039 | if (!info) |
2040 | udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX); | |
2041 | else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) | |
2042 | df = htons(IP_DF); | |
2043 | ||
e4c7ed41 CW |
2044 | tos = ip_tunnel_ecn_encap(tos, old_iph, skb); |
2045 | ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); | |
f491e56d | 2046 | err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr), |
54bfd872 | 2047 | vni, md, flags, udp_sum); |
f491e56d JB |
2048 | if (err < 0) |
2049 | goto xmit_tx_error; | |
2050 | ||
272d96a5 | 2051 | udp_tunnel_xmit_skb(rt, sk, skb, src->sin.sin_addr.s_addr, |
f491e56d JB |
2052 | dst->sin.sin_addr.s_addr, tos, ttl, df, |
2053 | src_port, dst_port, xnet, !udp_sum); | |
e4c7ed41 CW |
2054 | #if IS_ENABLED(CONFIG_IPV6) |
2055 | } else { | |
e4c7ed41 | 2056 | struct dst_entry *ndst; |
6f264e47 | 2057 | u32 rt6i_flags; |
e4c7ed41 | 2058 | |
b1be00a6 JB |
2059 | if (!vxlan->vn6_sock) |
2060 | goto drop; | |
2061 | sk = vxlan->vn6_sock->sock->sk; | |
2062 | ||
e5d4b29f | 2063 | ndst = vxlan6_get_route(vxlan, skb, |
1400615d | 2064 | rdst ? rdst->remote_ifindex : 0, tos, |
272d96a5 | 2065 | label, &dst->sin6.sin6_addr, |
2066 | &src->sin6.sin6_addr, | |
db3c6139 | 2067 | dst_cache, info); |
e5d4b29f | 2068 | if (IS_ERR(ndst)) { |
e4c7ed41 CW |
2069 | netdev_dbg(dev, "no route to %pI6\n", |
2070 | &dst->sin6.sin6_addr); | |
2071 | dev->stats.tx_carrier_errors++; | |
9dcc71e1 | 2072 | goto tx_error; |
e4c7ed41 | 2073 | } |
d342894c | 2074 | |
e4c7ed41 CW |
2075 | if (ndst->dev == dev) { |
2076 | netdev_dbg(dev, "circular route to %pI6\n", | |
2077 | &dst->sin6.sin6_addr); | |
2078 | dst_release(ndst); | |
2079 | dev->stats.collisions++; | |
2080 | goto tx_error; | |
2081 | } | |
0e6fbc5b | 2082 | |
e4c7ed41 | 2083 | /* Bypass encapsulation if the destination is local */ |
6f264e47 | 2084 | rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags; |
bbec7802 | 2085 | if (!info && rt6i_flags & RTF_LOCAL && |
6f264e47 | 2086 | !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { |
e4c7ed41 CW |
2087 | struct vxlan_dev *dst_vxlan; |
2088 | ||
2089 | dst_release(ndst); | |
19ca9fc1 | 2090 | dst_vxlan = vxlan_find_vni(vxlan->net, vni, |
ac5132d1 TG |
2091 | dst->sa.sa_family, dst_port, |
2092 | vxlan->flags); | |
e4c7ed41 CW |
2093 | if (!dst_vxlan) |
2094 | goto tx_error; | |
2095 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
2096 | return; | |
2097 | } | |
49560532 | 2098 | |
b4ed5cad JB |
2099 | if (!info) |
2100 | udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); | |
35e2d115 | 2101 | |
1400615d | 2102 | tos = ip_tunnel_ecn_encap(tos, old_iph, skb); |
e4c7ed41 | 2103 | ttl = ttl ? : ip6_dst_hoplimit(ndst); |
f491e56d JB |
2104 | skb_scrub_packet(skb, xnet); |
2105 | err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), | |
54bfd872 | 2106 | vni, md, flags, udp_sum); |
f491e56d JB |
2107 | if (err < 0) { |
2108 | dst_release(ndst); | |
2109 | return; | |
2110 | } | |
2111 | udp_tunnel6_xmit_skb(ndst, sk, skb, dev, | |
272d96a5 | 2112 | &src->sin6.sin6_addr, |
2113 | &dst->sin6.sin6_addr, tos, ttl, | |
e7f70af1 | 2114 | label, src_port, dst_port, !udp_sum); |
e4c7ed41 CW |
2115 | #endif |
2116 | } | |
0e6fbc5b | 2117 | |
4ad16930 | 2118 | return; |
d342894c | 2119 | |
2120 | drop: | |
2121 | dev->stats.tx_dropped++; | |
2122 | goto tx_free; | |
2123 | ||
f491e56d JB |
2124 | xmit_tx_error: |
2125 | /* skb is already freed. */ | |
2126 | skb = NULL; | |
49560532 PS |
2127 | rt_tx_error: |
2128 | ip_rt_put(rt); | |
d342894c | 2129 | tx_error: |
2130 | dev->stats.tx_errors++; | |
2131 | tx_free: | |
2132 | dev_kfree_skb(skb); | |
d342894c | 2133 | } |
2134 | ||
6681712d DS |
2135 | /* Transmit local packets over Vxlan |
2136 | * | |
2137 | * Outer IP header inherits ECN and DF from inner header. | |
2138 | * Outer UDP destination is the VXLAN assigned port. | |
2139 | * source port is based on hash of flow | |
2140 | */ | |
2141 | static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) | |
2142 | { | |
2143 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
3093fbe7 | 2144 | const struct ip_tunnel_info *info; |
6681712d DS |
2145 | struct ethhdr *eth; |
2146 | bool did_rsc = false; | |
8f646c92 | 2147 | struct vxlan_rdst *rdst, *fdst = NULL; |
6681712d | 2148 | struct vxlan_fdb *f; |
6681712d | 2149 | |
61adedf3 | 2150 | info = skb_tunnel_info(skb); |
3093fbe7 | 2151 | |
6681712d | 2152 | skb_reset_mac_header(skb); |
6681712d | 2153 | |
47e5d1b0 JB |
2154 | if (vxlan->flags & VXLAN_F_COLLECT_METADATA) { |
2155 | if (info && info->mode & IP_TUNNEL_INFO_TX) | |
2156 | vxlan_xmit_one(skb, dev, NULL, false); | |
2157 | else | |
2158 | kfree_skb(skb); | |
2159 | return NETDEV_TX_OK; | |
2160 | } | |
2161 | ||
2162 | if (vxlan->flags & VXLAN_F_PROXY) { | |
2163 | eth = eth_hdr(skb); | |
f564f45c CW |
2164 | if (ntohs(eth->h_proto) == ETH_P_ARP) |
2165 | return arp_reduce(dev, skb); | |
2166 | #if IS_ENABLED(CONFIG_IPV6) | |
2167 | else if (ntohs(eth->h_proto) == ETH_P_IPV6 && | |
91269e39 LR |
2168 | pskb_may_pull(skb, sizeof(struct ipv6hdr) |
2169 | + sizeof(struct nd_msg)) && | |
f564f45c CW |
2170 | ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { |
2171 | struct nd_msg *msg; | |
2172 | ||
2173 | msg = (struct nd_msg *)skb_transport_header(skb); | |
2174 | if (msg->icmph.icmp6_code == 0 && | |
2175 | msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) | |
2176 | return neigh_reduce(dev, skb); | |
2177 | } | |
2178 | #endif | |
2179 | } | |
6681712d | 2180 | |
47e5d1b0 | 2181 | eth = eth_hdr(skb); |
6681712d | 2182 | f = vxlan_find_mac(vxlan, eth->h_dest); |
ae884082 DS |
2183 | did_rsc = false; |
2184 | ||
2185 | if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) && | |
e15a00aa CW |
2186 | (ntohs(eth->h_proto) == ETH_P_IP || |
2187 | ntohs(eth->h_proto) == ETH_P_IPV6)) { | |
ae884082 DS |
2188 | did_rsc = route_shortcircuit(dev, skb); |
2189 | if (did_rsc) | |
2190 | f = vxlan_find_mac(vxlan, eth->h_dest); | |
2191 | } | |
2192 | ||
6681712d | 2193 | if (f == NULL) { |
afbd8bae MR |
2194 | f = vxlan_find_mac(vxlan, all_zeros_mac); |
2195 | if (f == NULL) { | |
2196 | if ((vxlan->flags & VXLAN_F_L2MISS) && | |
2197 | !is_multicast_ether_addr(eth->h_dest)) | |
2198 | vxlan_fdb_miss(vxlan, eth->h_dest); | |
2199 | ||
2200 | dev->stats.tx_dropped++; | |
8f646c92 | 2201 | kfree_skb(skb); |
afbd8bae MR |
2202 | return NETDEV_TX_OK; |
2203 | } | |
2204 | } | |
6681712d | 2205 | |
afbd8bae MR |
2206 | list_for_each_entry_rcu(rdst, &f->remotes, list) { |
2207 | struct sk_buff *skb1; | |
6681712d | 2208 | |
8f646c92 ED |
2209 | if (!fdst) { |
2210 | fdst = rdst; | |
2211 | continue; | |
2212 | } | |
afbd8bae MR |
2213 | skb1 = skb_clone(skb, GFP_ATOMIC); |
2214 | if (skb1) | |
2215 | vxlan_xmit_one(skb1, dev, rdst, did_rsc); | |
6681712d DS |
2216 | } |
2217 | ||
8f646c92 ED |
2218 | if (fdst) |
2219 | vxlan_xmit_one(skb, dev, fdst, did_rsc); | |
2220 | else | |
2221 | kfree_skb(skb); | |
4ad16930 | 2222 | return NETDEV_TX_OK; |
6681712d DS |
2223 | } |
2224 | ||
d342894c | 2225 | /* Walk the forwarding table and purge stale entries */ |
2226 | static void vxlan_cleanup(unsigned long arg) | |
2227 | { | |
2228 | struct vxlan_dev *vxlan = (struct vxlan_dev *) arg; | |
2229 | unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; | |
2230 | unsigned int h; | |
2231 | ||
2232 | if (!netif_running(vxlan->dev)) | |
2233 | return; | |
2234 | ||
d342894c | 2235 | for (h = 0; h < FDB_HASH_SIZE; ++h) { |
2236 | struct hlist_node *p, *n; | |
14e1d0fa SD |
2237 | |
2238 | spin_lock_bh(&vxlan->hash_lock); | |
d342894c | 2239 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { |
2240 | struct vxlan_fdb *f | |
2241 | = container_of(p, struct vxlan_fdb, hlist); | |
2242 | unsigned long timeout; | |
2243 | ||
3c172868 | 2244 | if (f->state & NUD_PERMANENT) |
d342894c | 2245 | continue; |
2246 | ||
0dfbdf41 | 2247 | timeout = f->used + vxlan->cfg.age_interval * HZ; |
d342894c | 2248 | if (time_before_eq(timeout, jiffies)) { |
2249 | netdev_dbg(vxlan->dev, | |
2250 | "garbage collect %pM\n", | |
2251 | f->eth_addr); | |
2252 | f->state = NUD_STALE; | |
2253 | vxlan_fdb_destroy(vxlan, f); | |
2254 | } else if (time_before(timeout, next_timer)) | |
2255 | next_timer = timeout; | |
2256 | } | |
14e1d0fa | 2257 | spin_unlock_bh(&vxlan->hash_lock); |
d342894c | 2258 | } |
d342894c | 2259 | |
2260 | mod_timer(&vxlan->age_timer, next_timer); | |
2261 | } | |
2262 | ||
9c2e24e1 PS |
2263 | static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan) |
2264 | { | |
56ef9c90 | 2265 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
54bfd872 | 2266 | __be32 vni = vxlan->default_dst.remote_vni; |
9c2e24e1 | 2267 | |
56ef9c90 | 2268 | spin_lock(&vn->sock_lock); |
9c2e24e1 | 2269 | hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni)); |
56ef9c90 | 2270 | spin_unlock(&vn->sock_lock); |
9c2e24e1 PS |
2271 | } |
2272 | ||
d342894c | 2273 | /* Setup stats when device is created */ |
2274 | static int vxlan_init(struct net_device *dev) | |
2275 | { | |
1c213bd2 | 2276 | dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); |
e8171045 | 2277 | if (!dev->tstats) |
d342894c | 2278 | return -ENOMEM; |
2279 | ||
2280 | return 0; | |
2281 | } | |
2282 | ||
ba609e9b | 2283 | static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan) |
afbd8bae MR |
2284 | { |
2285 | struct vxlan_fdb *f; | |
2286 | ||
2287 | spin_lock_bh(&vxlan->hash_lock); | |
2288 | f = __vxlan_find_mac(vxlan, all_zeros_mac); | |
2289 | if (f) | |
2290 | vxlan_fdb_destroy(vxlan, f); | |
2291 | spin_unlock_bh(&vxlan->hash_lock); | |
2292 | } | |
2293 | ||
ebf4063e SH |
2294 | static void vxlan_uninit(struct net_device *dev) |
2295 | { | |
2296 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
ebf4063e | 2297 | |
ba609e9b | 2298 | vxlan_fdb_delete_default(vxlan); |
afbd8bae | 2299 | |
ebf4063e SH |
2300 | free_percpu(dev->tstats); |
2301 | } | |
2302 | ||
d342894c | 2303 | /* Start ageing timer and join group when device is brought up */ |
2304 | static int vxlan_open(struct net_device *dev) | |
2305 | { | |
2306 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
205f356d | 2307 | int ret; |
56ef9c90 | 2308 | |
205f356d JB |
2309 | ret = vxlan_sock_add(vxlan); |
2310 | if (ret < 0) | |
2311 | return ret; | |
d342894c | 2312 | |
79d4a94f | 2313 | if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) { |
56ef9c90 | 2314 | ret = vxlan_igmp_join(vxlan); |
bef0057b MRL |
2315 | if (ret == -EADDRINUSE) |
2316 | ret = 0; | |
56ef9c90 | 2317 | if (ret) { |
205f356d | 2318 | vxlan_sock_release(vxlan); |
56ef9c90 MRL |
2319 | return ret; |
2320 | } | |
d342894c | 2321 | } |
2322 | ||
0dfbdf41 | 2323 | if (vxlan->cfg.age_interval) |
d342894c | 2324 | mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); |
2325 | ||
56ef9c90 | 2326 | return ret; |
d342894c | 2327 | } |
2328 | ||
2329 | /* Purge the forwarding table */ | |
2330 | static void vxlan_flush(struct vxlan_dev *vxlan) | |
2331 | { | |
31fec5aa | 2332 | unsigned int h; |
d342894c | 2333 | |
2334 | spin_lock_bh(&vxlan->hash_lock); | |
2335 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
2336 | struct hlist_node *p, *n; | |
2337 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { | |
2338 | struct vxlan_fdb *f | |
2339 | = container_of(p, struct vxlan_fdb, hlist); | |
afbd8bae MR |
2340 | /* the all_zeros_mac entry is deleted at vxlan_uninit */ |
2341 | if (!is_zero_ether_addr(f->eth_addr)) | |
2342 | vxlan_fdb_destroy(vxlan, f); | |
d342894c | 2343 | } |
2344 | } | |
2345 | spin_unlock_bh(&vxlan->hash_lock); | |
2346 | } | |
2347 | ||
2348 | /* Cleanup timer and forwarding table on shutdown */ | |
2349 | static int vxlan_stop(struct net_device *dev) | |
2350 | { | |
2351 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
f01ec1c0 | 2352 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
56ef9c90 | 2353 | int ret = 0; |
d342894c | 2354 | |
24c0e683 | 2355 | if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) && |
f13b1689 | 2356 | !vxlan_group_used(vn, vxlan)) |
56ef9c90 | 2357 | ret = vxlan_igmp_leave(vxlan); |
d342894c | 2358 | |
2359 | del_timer_sync(&vxlan->age_timer); | |
2360 | ||
2361 | vxlan_flush(vxlan); | |
205f356d | 2362 | vxlan_sock_release(vxlan); |
d342894c | 2363 | |
56ef9c90 | 2364 | return ret; |
d342894c | 2365 | } |
2366 | ||
d342894c | 2367 | /* Stub, nothing needs to be done. */ |
2368 | static void vxlan_set_multicast_list(struct net_device *dev) | |
2369 | { | |
2370 | } | |
2371 | ||
72564b59 DW |
2372 | static int __vxlan_change_mtu(struct net_device *dev, |
2373 | struct net_device *lowerdev, | |
2374 | struct vxlan_rdst *dst, int new_mtu, bool strict) | |
345010b5 | 2375 | { |
72564b59 | 2376 | int max_mtu = IP_MAX_MTU; |
345010b5 | 2377 | |
72564b59 DW |
2378 | if (lowerdev) |
2379 | max_mtu = lowerdev->mtu; | |
345010b5 DB |
2380 | |
2381 | if (dst->remote_ip.sa.sa_family == AF_INET6) | |
72564b59 | 2382 | max_mtu -= VXLAN6_HEADROOM; |
345010b5 | 2383 | else |
72564b59 | 2384 | max_mtu -= VXLAN_HEADROOM; |
345010b5 | 2385 | |
72564b59 | 2386 | if (new_mtu < 68) |
345010b5 DB |
2387 | return -EINVAL; |
2388 | ||
72564b59 DW |
2389 | if (new_mtu > max_mtu) { |
2390 | if (strict) | |
2391 | return -EINVAL; | |
2392 | ||
2393 | new_mtu = max_mtu; | |
2394 | } | |
2395 | ||
345010b5 DB |
2396 | dev->mtu = new_mtu; |
2397 | return 0; | |
2398 | } | |
2399 | ||
72564b59 DW |
2400 | static int vxlan_change_mtu(struct net_device *dev, int new_mtu) |
2401 | { | |
2402 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
2403 | struct vxlan_rdst *dst = &vxlan->default_dst; | |
2404 | struct net_device *lowerdev = __dev_get_by_index(vxlan->net, | |
2405 | dst->remote_ifindex); | |
2406 | return __vxlan_change_mtu(dev, lowerdev, dst, new_mtu, true); | |
2407 | } | |
2408 | ||
fc4099f1 PS |
2409 | static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
2410 | { | |
2411 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
2412 | struct ip_tunnel_info *info = skb_tunnel_info(skb); | |
2413 | __be16 sport, dport; | |
2414 | ||
2415 | sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, | |
2416 | vxlan->cfg.port_max, true); | |
2417 | dport = info->key.tp_dst ? : vxlan->cfg.dst_port; | |
2418 | ||
239e944f | 2419 | if (ip_tunnel_info_af(info) == AF_INET) { |
1a8496ba JB |
2420 | struct rtable *rt; |
2421 | ||
239e944f JB |
2422 | if (!vxlan->vn4_sock) |
2423 | return -EINVAL; | |
1a8496ba JB |
2424 | rt = vxlan_get_route(vxlan, skb, 0, info->key.tos, |
2425 | info->key.u.ipv4.dst, | |
0c1d70af | 2426 | &info->key.u.ipv4.src, NULL, info); |
1a8496ba JB |
2427 | if (IS_ERR(rt)) |
2428 | return PTR_ERR(rt); | |
2429 | ip_rt_put(rt); | |
239e944f JB |
2430 | } else { |
2431 | #if IS_ENABLED(CONFIG_IPV6) | |
2432 | struct dst_entry *ndst; | |
2433 | ||
2434 | if (!vxlan->vn6_sock) | |
2435 | return -EINVAL; | |
1400615d | 2436 | ndst = vxlan6_get_route(vxlan, skb, 0, info->key.tos, |
e7f70af1 | 2437 | info->key.label, &info->key.u.ipv6.dst, |
db3c6139 | 2438 | &info->key.u.ipv6.src, NULL, info); |
239e944f JB |
2439 | if (IS_ERR(ndst)) |
2440 | return PTR_ERR(ndst); | |
2441 | dst_release(ndst); | |
239e944f JB |
2442 | #else /* !CONFIG_IPV6 */ |
2443 | return -EPFNOSUPPORT; | |
2444 | #endif | |
2445 | } | |
1a8496ba JB |
2446 | info->key.tp_src = sport; |
2447 | info->key.tp_dst = dport; | |
239e944f | 2448 | return 0; |
fc4099f1 PS |
2449 | } |
2450 | ||
0c867c9b | 2451 | static const struct net_device_ops vxlan_netdev_ether_ops = { |
d342894c | 2452 | .ndo_init = vxlan_init, |
ebf4063e | 2453 | .ndo_uninit = vxlan_uninit, |
d342894c | 2454 | .ndo_open = vxlan_open, |
2455 | .ndo_stop = vxlan_stop, | |
2456 | .ndo_start_xmit = vxlan_xmit, | |
e8171045 | 2457 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
d342894c | 2458 | .ndo_set_rx_mode = vxlan_set_multicast_list, |
345010b5 | 2459 | .ndo_change_mtu = vxlan_change_mtu, |
d342894c | 2460 | .ndo_validate_addr = eth_validate_addr, |
2461 | .ndo_set_mac_address = eth_mac_addr, | |
2462 | .ndo_fdb_add = vxlan_fdb_add, | |
2463 | .ndo_fdb_del = vxlan_fdb_delete, | |
2464 | .ndo_fdb_dump = vxlan_fdb_dump, | |
fc4099f1 | 2465 | .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, |
d342894c | 2466 | }; |
2467 | ||
e1e5314d JB |
2468 | static const struct net_device_ops vxlan_netdev_raw_ops = { |
2469 | .ndo_init = vxlan_init, | |
2470 | .ndo_uninit = vxlan_uninit, | |
2471 | .ndo_open = vxlan_open, | |
2472 | .ndo_stop = vxlan_stop, | |
2473 | .ndo_start_xmit = vxlan_xmit, | |
2474 | .ndo_get_stats64 = ip_tunnel_get_stats64, | |
2475 | .ndo_change_mtu = vxlan_change_mtu, | |
2476 | .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, | |
2477 | }; | |
2478 | ||
d342894c | 2479 | /* Info for udev, that this is a virtual tunnel endpoint */ |
2480 | static struct device_type vxlan_type = { | |
2481 | .name = "vxlan", | |
2482 | }; | |
2483 | ||
e5de25dc | 2484 | /* Calls the ndo_udp_tunnel_add of the caller in order to |
35e42379 | 2485 | * supply the listening VXLAN udp ports. Callers are expected |
e5de25dc | 2486 | * to implement the ndo_udp_tunnel_add. |
53cf5275 | 2487 | */ |
b7aade15 | 2488 | static void vxlan_push_rx_ports(struct net_device *dev) |
53cf5275 JG |
2489 | { |
2490 | struct vxlan_sock *vs; | |
2491 | struct net *net = dev_net(dev); | |
2492 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
35e42379 | 2493 | unsigned int i; |
53cf5275 JG |
2494 | |
2495 | spin_lock(&vn->sock_lock); | |
2496 | for (i = 0; i < PORT_HASH_SIZE; ++i) { | |
e7b3db5e AD |
2497 | hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) |
2498 | udp_tunnel_push_rx_port(dev, vs->sock, | |
b9adcd69 AD |
2499 | (vs->flags & VXLAN_F_GPE) ? |
2500 | UDP_TUNNEL_TYPE_VXLAN_GPE : | |
e7b3db5e | 2501 | UDP_TUNNEL_TYPE_VXLAN); |
53cf5275 JG |
2502 | } |
2503 | spin_unlock(&vn->sock_lock); | |
2504 | } | |
53cf5275 | 2505 | |
d342894c | 2506 | /* Initialize the device structure. */ |
2507 | static void vxlan_setup(struct net_device *dev) | |
2508 | { | |
2509 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
31fec5aa | 2510 | unsigned int h; |
d342894c | 2511 | |
65226ef8 JB |
2512 | eth_hw_addr_random(dev); |
2513 | ether_setup(dev); | |
2514 | ||
ebf4063e | 2515 | dev->destructor = free_netdev; |
d342894c | 2516 | SET_NETDEV_DEVTYPE(dev, &vxlan_type); |
2517 | ||
d342894c | 2518 | dev->features |= NETIF_F_LLTX; |
d6727fe3 | 2519 | dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; |
0afb1666 | 2520 | dev->features |= NETIF_F_RXCSUM; |
05c0db08 | 2521 | dev->features |= NETIF_F_GSO_SOFTWARE; |
0afb1666 | 2522 | |
1eaa8178 PS |
2523 | dev->vlan_features = dev->features; |
2524 | dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; | |
0afb1666 | 2525 | dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM; |
05c0db08 | 2526 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; |
1eaa8178 | 2527 | dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; |
02875878 | 2528 | netif_keep_dst(dev); |
0c867c9b | 2529 | dev->priv_flags |= IFF_NO_QUEUE; |
d342894c | 2530 | |
553675fb | 2531 | INIT_LIST_HEAD(&vxlan->next); |
d342894c | 2532 | spin_lock_init(&vxlan->hash_lock); |
2533 | ||
2534 | init_timer_deferrable(&vxlan->age_timer); | |
2535 | vxlan->age_timer.function = vxlan_cleanup; | |
2536 | vxlan->age_timer.data = (unsigned long) vxlan; | |
2537 | ||
0dfbdf41 | 2538 | vxlan->cfg.dst_port = htons(vxlan_port); |
05f47d69 | 2539 | |
d342894c | 2540 | vxlan->dev = dev; |
2541 | ||
58ce31cc TH |
2542 | gro_cells_init(&vxlan->gro_cells, dev); |
2543 | ||
d342894c | 2544 | for (h = 0; h < FDB_HASH_SIZE; ++h) |
2545 | INIT_HLIST_HEAD(&vxlan->fdb_head[h]); | |
2546 | } | |
2547 | ||
0c867c9b JB |
2548 | static void vxlan_ether_setup(struct net_device *dev) |
2549 | { | |
0c867c9b JB |
2550 | dev->priv_flags &= ~IFF_TX_SKB_SHARING; |
2551 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; | |
2552 | dev->netdev_ops = &vxlan_netdev_ether_ops; | |
2553 | } | |
2554 | ||
e1e5314d JB |
2555 | static void vxlan_raw_setup(struct net_device *dev) |
2556 | { | |
65226ef8 | 2557 | dev->header_ops = NULL; |
e1e5314d JB |
2558 | dev->type = ARPHRD_NONE; |
2559 | dev->hard_header_len = 0; | |
2560 | dev->addr_len = 0; | |
e1e5314d JB |
2561 | dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST; |
2562 | dev->netdev_ops = &vxlan_netdev_raw_ops; | |
2563 | } | |
2564 | ||
d342894c | 2565 | static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { |
2566 | [IFLA_VXLAN_ID] = { .type = NLA_U32 }, | |
5d174dd8 | 2567 | [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, |
e4c7ed41 | 2568 | [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, |
d342894c | 2569 | [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, |
2570 | [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, | |
e4c7ed41 | 2571 | [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, |
d342894c | 2572 | [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, |
2573 | [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, | |
e7f70af1 | 2574 | [IFLA_VXLAN_LABEL] = { .type = NLA_U32 }, |
d342894c | 2575 | [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, |
2576 | [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, | |
2577 | [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, | |
05f47d69 | 2578 | [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, |
e4f67add DS |
2579 | [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, |
2580 | [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, | |
2581 | [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, | |
2582 | [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, | |
f8a9b1bc | 2583 | [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, |
823aa873 | 2584 | [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, |
5c91ae08 TH |
2585 | [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, |
2586 | [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, | |
2587 | [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, | |
dfd8645e TH |
2588 | [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, |
2589 | [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, | |
3511494c | 2590 | [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, |
e1e5314d | 2591 | [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, }, |
0ace2ca8 | 2592 | [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, |
d342894c | 2593 | }; |
2594 | ||
2595 | static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[]) | |
2596 | { | |
2597 | if (tb[IFLA_ADDRESS]) { | |
2598 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { | |
2599 | pr_debug("invalid link address (not ethernet)\n"); | |
2600 | return -EINVAL; | |
2601 | } | |
2602 | ||
2603 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { | |
2604 | pr_debug("invalid all zero ethernet address\n"); | |
2605 | return -EADDRNOTAVAIL; | |
2606 | } | |
2607 | } | |
2608 | ||
2609 | if (!data) | |
2610 | return -EINVAL; | |
2611 | ||
2612 | if (data[IFLA_VXLAN_ID]) { | |
2613 | __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); | |
2614 | if (id >= VXLAN_VID_MASK) | |
2615 | return -ERANGE; | |
2616 | } | |
2617 | ||
05f47d69 | 2618 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
2619 | const struct ifla_vxlan_port_range *p | |
2620 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
2621 | ||
2622 | if (ntohs(p->high) < ntohs(p->low)) { | |
2623 | pr_debug("port range %u .. %u not valid\n", | |
2624 | ntohs(p->low), ntohs(p->high)); | |
2625 | return -EINVAL; | |
2626 | } | |
2627 | } | |
2628 | ||
d342894c | 2629 | return 0; |
2630 | } | |
2631 | ||
1b13c97f YB |
2632 | static void vxlan_get_drvinfo(struct net_device *netdev, |
2633 | struct ethtool_drvinfo *drvinfo) | |
2634 | { | |
2635 | strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); | |
2636 | strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); | |
2637 | } | |
2638 | ||
2639 | static const struct ethtool_ops vxlan_ethtool_ops = { | |
2640 | .get_drvinfo = vxlan_get_drvinfo, | |
2641 | .get_link = ethtool_op_get_link, | |
2642 | }; | |
2643 | ||
3ee64f39 TH |
2644 | static struct socket *vxlan_create_sock(struct net *net, bool ipv6, |
2645 | __be16 port, u32 flags) | |
553675fb | 2646 | { |
e4c7ed41 | 2647 | struct socket *sock; |
3ee64f39 TH |
2648 | struct udp_port_cfg udp_conf; |
2649 | int err; | |
e4c7ed41 | 2650 | |
3ee64f39 | 2651 | memset(&udp_conf, 0, sizeof(udp_conf)); |
e4c7ed41 | 2652 | |
3ee64f39 TH |
2653 | if (ipv6) { |
2654 | udp_conf.family = AF_INET6; | |
3ee64f39 | 2655 | udp_conf.use_udp6_rx_checksums = |
3dc2b6a8 | 2656 | !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); |
a43a9ef6 | 2657 | udp_conf.ipv6_v6only = 1; |
3ee64f39 TH |
2658 | } else { |
2659 | udp_conf.family = AF_INET; | |
e4c7ed41 | 2660 | } |
e4c7ed41 | 2661 | |
3ee64f39 | 2662 | udp_conf.local_udp_port = port; |
553675fb | 2663 | |
3ee64f39 TH |
2664 | /* Open UDP socket */ |
2665 | err = udp_sock_create(net, &udp_conf, &sock); | |
2666 | if (err < 0) | |
2667 | return ERR_PTR(err); | |
e4c7ed41 | 2668 | |
39deb2c7 | 2669 | return sock; |
e4c7ed41 CW |
2670 | } |
2671 | ||
2672 | /* Create new listen socket if needed */ | |
b1be00a6 JB |
2673 | static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, |
2674 | __be16 port, u32 flags) | |
e4c7ed41 CW |
2675 | { |
2676 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
2677 | struct vxlan_sock *vs; | |
2678 | struct socket *sock; | |
e4c7ed41 | 2679 | unsigned int h; |
acbf74a7 | 2680 | struct udp_tunnel_sock_cfg tunnel_cfg; |
e4c7ed41 | 2681 | |
dc01e7d3 | 2682 | vs = kzalloc(sizeof(*vs), GFP_KERNEL); |
e4c7ed41 CW |
2683 | if (!vs) |
2684 | return ERR_PTR(-ENOMEM); | |
2685 | ||
2686 | for (h = 0; h < VNI_HASH_SIZE; ++h) | |
2687 | INIT_HLIST_HEAD(&vs->vni_list[h]); | |
2688 | ||
3ee64f39 | 2689 | sock = vxlan_create_sock(net, ipv6, port, flags); |
39deb2c7 | 2690 | if (IS_ERR(sock)) { |
56ef9c90 MRL |
2691 | pr_info("Cannot bind port %d, err=%ld\n", ntohs(port), |
2692 | PTR_ERR(sock)); | |
553675fb | 2693 | kfree(vs); |
e50fddc8 | 2694 | return ERR_CAST(sock); |
553675fb | 2695 | } |
e4c7ed41 CW |
2696 | |
2697 | vs->sock = sock; | |
9c2e24e1 | 2698 | atomic_set(&vs->refcnt, 1); |
af33c1ad | 2699 | vs->flags = (flags & VXLAN_F_RCV_FLAGS); |
553675fb | 2700 | |
9c2e24e1 PS |
2701 | spin_lock(&vn->sock_lock); |
2702 | hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); | |
e7b3db5e | 2703 | udp_tunnel_notify_add_rx_port(sock, |
b9adcd69 AD |
2704 | (vs->flags & VXLAN_F_GPE) ? |
2705 | UDP_TUNNEL_TYPE_VXLAN_GPE : | |
e7b3db5e | 2706 | UDP_TUNNEL_TYPE_VXLAN); |
9c2e24e1 | 2707 | spin_unlock(&vn->sock_lock); |
553675fb | 2708 | |
2709 | /* Mark socket as an encapsulation socket. */ | |
5602c48c | 2710 | memset(&tunnel_cfg, 0, sizeof(tunnel_cfg)); |
acbf74a7 AZ |
2711 | tunnel_cfg.sk_user_data = vs; |
2712 | tunnel_cfg.encap_type = 1; | |
f2d1968e | 2713 | tunnel_cfg.encap_rcv = vxlan_rcv; |
acbf74a7 | 2714 | tunnel_cfg.encap_destroy = NULL; |
5602c48c TH |
2715 | tunnel_cfg.gro_receive = vxlan_gro_receive; |
2716 | tunnel_cfg.gro_complete = vxlan_gro_complete; | |
acbf74a7 AZ |
2717 | |
2718 | setup_udp_tunnel_sock(net, sock, &tunnel_cfg); | |
e4c7ed41 | 2719 | |
9c2e24e1 PS |
2720 | return vs; |
2721 | } | |
2722 | ||
b1be00a6 | 2723 | static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) |
9c2e24e1 | 2724 | { |
205f356d JB |
2725 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
2726 | struct vxlan_sock *vs = NULL; | |
9c2e24e1 | 2727 | |
205f356d | 2728 | if (!vxlan->cfg.no_share) { |
56ef9c90 | 2729 | spin_lock(&vn->sock_lock); |
205f356d JB |
2730 | vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, |
2731 | vxlan->cfg.dst_port, vxlan->flags); | |
2732 | if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) { | |
56ef9c90 | 2733 | spin_unlock(&vn->sock_lock); |
205f356d | 2734 | return -EBUSY; |
56ef9c90 MRL |
2735 | } |
2736 | spin_unlock(&vn->sock_lock); | |
2737 | } | |
205f356d | 2738 | if (!vs) |
b1be00a6 JB |
2739 | vs = vxlan_socket_create(vxlan->net, ipv6, |
2740 | vxlan->cfg.dst_port, vxlan->flags); | |
205f356d JB |
2741 | if (IS_ERR(vs)) |
2742 | return PTR_ERR(vs); | |
b1be00a6 JB |
2743 | #if IS_ENABLED(CONFIG_IPV6) |
2744 | if (ipv6) | |
2745 | vxlan->vn6_sock = vs; | |
2746 | else | |
2747 | #endif | |
2748 | vxlan->vn4_sock = vs; | |
205f356d JB |
2749 | vxlan_vs_add_dev(vs, vxlan); |
2750 | return 0; | |
553675fb | 2751 | } |
2752 | ||
b1be00a6 JB |
2753 | static int vxlan_sock_add(struct vxlan_dev *vxlan) |
2754 | { | |
2755 | bool ipv6 = vxlan->flags & VXLAN_F_IPV6; | |
2756 | bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA; | |
2757 | int ret = 0; | |
2758 | ||
2759 | vxlan->vn4_sock = NULL; | |
2760 | #if IS_ENABLED(CONFIG_IPV6) | |
2761 | vxlan->vn6_sock = NULL; | |
2762 | if (ipv6 || metadata) | |
2763 | ret = __vxlan_sock_add(vxlan, true); | |
2764 | #endif | |
2765 | if (!ret && (!ipv6 || metadata)) | |
2766 | ret = __vxlan_sock_add(vxlan, false); | |
2767 | if (ret < 0) | |
2768 | vxlan_sock_release(vxlan); | |
2769 | return ret; | |
2770 | } | |
2771 | ||
0dfbdf41 TG |
2772 | static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, |
2773 | struct vxlan_config *conf) | |
d342894c | 2774 | { |
33564bbb | 2775 | struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); |
07b9b37c | 2776 | struct vxlan_dev *vxlan = netdev_priv(dev), *tmp; |
c7995c43 | 2777 | struct vxlan_rdst *dst = &vxlan->default_dst; |
b1be00a6 | 2778 | unsigned short needed_headroom = ETH_HLEN; |
d342894c | 2779 | int err; |
e4c7ed41 | 2780 | bool use_ipv6 = false; |
0dfbdf41 | 2781 | __be16 default_port = vxlan->cfg.dst_port; |
7e059158 | 2782 | struct net_device *lowerdev = NULL; |
d342894c | 2783 | |
e1e5314d JB |
2784 | if (conf->flags & VXLAN_F_GPE) { |
2785 | if (conf->flags & ~VXLAN_F_ALLOWED_GPE) | |
2786 | return -EINVAL; | |
2787 | /* For now, allow GPE only together with COLLECT_METADATA. | |
2788 | * This can be relaxed later; in such case, the other side | |
2789 | * of the PtP link will have to be provided. | |
2790 | */ | |
2791 | if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) | |
2792 | return -EINVAL; | |
2793 | ||
2794 | vxlan_raw_setup(dev); | |
2795 | } else { | |
2796 | vxlan_ether_setup(dev); | |
2797 | } | |
0c867c9b | 2798 | |
33564bbb | 2799 | vxlan->net = src_net; |
f01ec1c0 | 2800 | |
0dfbdf41 | 2801 | dst->remote_vni = conf->vni; |
d342894c | 2802 | |
0dfbdf41 | 2803 | memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); |
d342894c | 2804 | |
0dfbdf41 TG |
2805 | /* Unless IPv6 is explicitly requested, assume IPv4 */ |
2806 | if (!dst->remote_ip.sa.sa_family) | |
2807 | dst->remote_ip.sa.sa_family = AF_INET; | |
e4c7ed41 | 2808 | |
0dfbdf41 | 2809 | if (dst->remote_ip.sa.sa_family == AF_INET6 || |
057ba29b JB |
2810 | vxlan->cfg.saddr.sa.sa_family == AF_INET6) { |
2811 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2812 | return -EPFNOSUPPORT; | |
e4c7ed41 | 2813 | use_ipv6 = true; |
b1be00a6 | 2814 | vxlan->flags |= VXLAN_F_IPV6; |
057ba29b | 2815 | } |
d342894c | 2816 | |
e7f70af1 DB |
2817 | if (conf->label && !use_ipv6) { |
2818 | pr_info("label only supported in use with IPv6\n"); | |
2819 | return -EINVAL; | |
2820 | } | |
2821 | ||
0dfbdf41 | 2822 | if (conf->remote_ifindex) { |
7e059158 | 2823 | lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex); |
0dfbdf41 | 2824 | dst->remote_ifindex = conf->remote_ifindex; |
34e02aa1 | 2825 | |
2826 | if (!lowerdev) { | |
c7995c43 | 2827 | pr_info("ifindex %d does not exist\n", dst->remote_ifindex); |
34e02aa1 | 2828 | return -ENODEV; |
2829 | } | |
d342894c | 2830 | |
e4c7ed41 CW |
2831 | #if IS_ENABLED(CONFIG_IPV6) |
2832 | if (use_ipv6) { | |
2833 | struct inet6_dev *idev = __in6_dev_get(lowerdev); | |
2834 | if (idev && idev->cnf.disable_ipv6) { | |
2835 | pr_info("IPv6 is disabled via sysctl\n"); | |
2836 | return -EPERM; | |
2837 | } | |
e4c7ed41 CW |
2838 | } |
2839 | #endif | |
2840 | ||
0dfbdf41 | 2841 | if (!conf->mtu) |
e4c7ed41 | 2842 | dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM); |
1ba56fb4 | 2843 | |
b1be00a6 | 2844 | needed_headroom = lowerdev->hard_header_len; |
9b4cdd51 JB |
2845 | } else if (vxlan_addr_multicast(&dst->remote_ip)) { |
2846 | pr_info("multicast destination requires interface to be specified\n"); | |
2847 | return -EINVAL; | |
9dc2ad10 | 2848 | } |
d342894c | 2849 | |
7e059158 DW |
2850 | if (conf->mtu) { |
2851 | err = __vxlan_change_mtu(dev, lowerdev, dst, conf->mtu, false); | |
2852 | if (err) | |
2853 | return err; | |
2854 | } | |
2855 | ||
b1be00a6 JB |
2856 | if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA) |
2857 | needed_headroom += VXLAN6_HEADROOM; | |
2858 | else | |
2859 | needed_headroom += VXLAN_HEADROOM; | |
2860 | dev->needed_headroom = needed_headroom; | |
2861 | ||
0dfbdf41 | 2862 | memcpy(&vxlan->cfg, conf, sizeof(*conf)); |
e1e5314d JB |
2863 | if (!vxlan->cfg.dst_port) { |
2864 | if (conf->flags & VXLAN_F_GPE) | |
2865 | vxlan->cfg.dst_port = 4790; /* IANA assigned VXLAN-GPE port */ | |
2866 | else | |
2867 | vxlan->cfg.dst_port = default_port; | |
2868 | } | |
0dfbdf41 TG |
2869 | vxlan->flags |= conf->flags; |
2870 | ||
2871 | if (!vxlan->cfg.age_interval) | |
2872 | vxlan->cfg.age_interval = FDB_AGE_DEFAULT; | |
2873 | ||
07b9b37c ND |
2874 | list_for_each_entry(tmp, &vn->vxlan_list, next) { |
2875 | if (tmp->cfg.vni == conf->vni && | |
2876 | (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 || | |
2877 | tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 && | |
2878 | tmp->cfg.dst_port == vxlan->cfg.dst_port && | |
2879 | (tmp->flags & VXLAN_F_RCV_FLAGS) == | |
2880 | (vxlan->flags & VXLAN_F_RCV_FLAGS)) | |
0dfbdf41 | 2881 | return -EEXIST; |
07b9b37c | 2882 | } |
0dfbdf41 TG |
2883 | |
2884 | dev->ethtool_ops = &vxlan_ethtool_ops; | |
2885 | ||
2886 | /* create an fdb entry for a valid default destination */ | |
2887 | if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) { | |
2888 | err = vxlan_fdb_create(vxlan, all_zeros_mac, | |
2889 | &vxlan->default_dst.remote_ip, | |
2890 | NUD_REACHABLE|NUD_PERMANENT, | |
2891 | NLM_F_EXCL|NLM_F_CREATE, | |
2892 | vxlan->cfg.dst_port, | |
2893 | vxlan->default_dst.remote_vni, | |
2894 | vxlan->default_dst.remote_ifindex, | |
2895 | NTF_SELF); | |
2896 | if (err) | |
2897 | return err; | |
2898 | } | |
2899 | ||
2900 | err = register_netdevice(dev); | |
2901 | if (err) { | |
2902 | vxlan_fdb_delete_default(vxlan); | |
2903 | return err; | |
2904 | } | |
2905 | ||
2906 | list_add(&vxlan->next, &vn->vxlan_list); | |
2907 | ||
2908 | return 0; | |
2909 | } | |
2910 | ||
0dfbdf41 TG |
2911 | static int vxlan_newlink(struct net *src_net, struct net_device *dev, |
2912 | struct nlattr *tb[], struct nlattr *data[]) | |
2913 | { | |
2914 | struct vxlan_config conf; | |
2915 | int err; | |
2916 | ||
0dfbdf41 | 2917 | memset(&conf, 0, sizeof(conf)); |
e277de5f JG |
2918 | |
2919 | if (data[IFLA_VXLAN_ID]) | |
54bfd872 | 2920 | conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); |
0dfbdf41 TG |
2921 | |
2922 | if (data[IFLA_VXLAN_GROUP]) { | |
2923 | conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); | |
2924 | } else if (data[IFLA_VXLAN_GROUP6]) { | |
2925 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2926 | return -EPFNOSUPPORT; | |
2927 | ||
2928 | conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); | |
2929 | conf.remote_ip.sa.sa_family = AF_INET6; | |
2930 | } | |
2931 | ||
2932 | if (data[IFLA_VXLAN_LOCAL]) { | |
2933 | conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); | |
2934 | conf.saddr.sa.sa_family = AF_INET; | |
2935 | } else if (data[IFLA_VXLAN_LOCAL6]) { | |
2936 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2937 | return -EPFNOSUPPORT; | |
2938 | ||
2939 | /* TODO: respect scope id */ | |
2940 | conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); | |
2941 | conf.saddr.sa.sa_family = AF_INET6; | |
2942 | } | |
2943 | ||
2944 | if (data[IFLA_VXLAN_LINK]) | |
2945 | conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); | |
2946 | ||
d342894c | 2947 | if (data[IFLA_VXLAN_TOS]) |
0dfbdf41 | 2948 | conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]); |
d342894c | 2949 | |
afb97186 | 2950 | if (data[IFLA_VXLAN_TTL]) |
0dfbdf41 | 2951 | conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); |
afb97186 | 2952 | |
e7f70af1 DB |
2953 | if (data[IFLA_VXLAN_LABEL]) |
2954 | conf.label = nla_get_be32(data[IFLA_VXLAN_LABEL]) & | |
2955 | IPV6_FLOWLABEL_MASK; | |
2956 | ||
d342894c | 2957 | if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING])) |
0dfbdf41 | 2958 | conf.flags |= VXLAN_F_LEARN; |
d342894c | 2959 | |
2960 | if (data[IFLA_VXLAN_AGEING]) | |
0dfbdf41 | 2961 | conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); |
d342894c | 2962 | |
e4f67add | 2963 | if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY])) |
0dfbdf41 | 2964 | conf.flags |= VXLAN_F_PROXY; |
e4f67add DS |
2965 | |
2966 | if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC])) | |
0dfbdf41 | 2967 | conf.flags |= VXLAN_F_RSC; |
e4f67add DS |
2968 | |
2969 | if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS])) | |
0dfbdf41 | 2970 | conf.flags |= VXLAN_F_L2MISS; |
e4f67add DS |
2971 | |
2972 | if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS])) | |
0dfbdf41 | 2973 | conf.flags |= VXLAN_F_L3MISS; |
e4f67add | 2974 | |
d342894c | 2975 | if (data[IFLA_VXLAN_LIMIT]) |
0dfbdf41 | 2976 | conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); |
d342894c | 2977 | |
f8a9b1bc AS |
2978 | if (data[IFLA_VXLAN_COLLECT_METADATA] && |
2979 | nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA])) | |
2980 | conf.flags |= VXLAN_F_COLLECT_METADATA; | |
2981 | ||
05f47d69 | 2982 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
2983 | const struct ifla_vxlan_port_range *p | |
2984 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
0dfbdf41 TG |
2985 | conf.port_min = ntohs(p->low); |
2986 | conf.port_max = ntohs(p->high); | |
05f47d69 | 2987 | } |
2988 | ||
823aa873 | 2989 | if (data[IFLA_VXLAN_PORT]) |
0dfbdf41 | 2990 | conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); |
823aa873 | 2991 | |
6ceb31ca AD |
2992 | if (data[IFLA_VXLAN_UDP_CSUM] && |
2993 | !nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) | |
2994 | conf.flags |= VXLAN_F_UDP_ZERO_CSUM_TX; | |
359a0ea9 TH |
2995 | |
2996 | if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] && | |
2997 | nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX])) | |
0dfbdf41 | 2998 | conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX; |
359a0ea9 TH |
2999 | |
3000 | if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] && | |
3001 | nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX])) | |
0dfbdf41 | 3002 | conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX; |
359a0ea9 | 3003 | |
dfd8645e TH |
3004 | if (data[IFLA_VXLAN_REMCSUM_TX] && |
3005 | nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX])) | |
0dfbdf41 | 3006 | conf.flags |= VXLAN_F_REMCSUM_TX; |
dfd8645e TH |
3007 | |
3008 | if (data[IFLA_VXLAN_REMCSUM_RX] && | |
3009 | nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX])) | |
0dfbdf41 | 3010 | conf.flags |= VXLAN_F_REMCSUM_RX; |
dfd8645e | 3011 | |
3511494c | 3012 | if (data[IFLA_VXLAN_GBP]) |
0dfbdf41 | 3013 | conf.flags |= VXLAN_F_GBP; |
3511494c | 3014 | |
e1e5314d JB |
3015 | if (data[IFLA_VXLAN_GPE]) |
3016 | conf.flags |= VXLAN_F_GPE; | |
3017 | ||
0ace2ca8 | 3018 | if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) |
0dfbdf41 | 3019 | conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL; |
0ace2ca8 | 3020 | |
ce577668 CH |
3021 | if (tb[IFLA_MTU]) |
3022 | conf.mtu = nla_get_u32(tb[IFLA_MTU]); | |
3023 | ||
0dfbdf41 TG |
3024 | err = vxlan_dev_configure(src_net, dev, &conf); |
3025 | switch (err) { | |
3026 | case -ENODEV: | |
3027 | pr_info("ifindex %d does not exist\n", conf.remote_ifindex); | |
3028 | break; | |
1b13c97f | 3029 | |
0dfbdf41 TG |
3030 | case -EPERM: |
3031 | pr_info("IPv6 is disabled via sysctl\n"); | |
3032 | break; | |
d342894c | 3033 | |
0dfbdf41 | 3034 | case -EEXIST: |
54bfd872 | 3035 | pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni)); |
0dfbdf41 | 3036 | break; |
e1e5314d JB |
3037 | |
3038 | case -EINVAL: | |
3039 | pr_info("unsupported combination of extensions\n"); | |
3040 | break; | |
afbd8bae MR |
3041 | } |
3042 | ||
0dfbdf41 | 3043 | return err; |
d342894c | 3044 | } |
3045 | ||
3046 | static void vxlan_dellink(struct net_device *dev, struct list_head *head) | |
3047 | { | |
3048 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
f01ec1c0 | 3049 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
d342894c | 3050 | |
fe5c3561 | 3051 | spin_lock(&vn->sock_lock); |
9c2e24e1 PS |
3052 | if (!hlist_unhashed(&vxlan->hlist)) |
3053 | hlist_del_rcu(&vxlan->hlist); | |
fe5c3561 | 3054 | spin_unlock(&vn->sock_lock); |
3055 | ||
58ce31cc | 3056 | gro_cells_destroy(&vxlan->gro_cells); |
553675fb | 3057 | list_del(&vxlan->next); |
d342894c | 3058 | unregister_netdevice_queue(dev, head); |
3059 | } | |
3060 | ||
3061 | static size_t vxlan_get_size(const struct net_device *dev) | |
3062 | { | |
3063 | ||
3064 | return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ | |
e4c7ed41 | 3065 | nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ |
d342894c | 3066 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ |
e4c7ed41 | 3067 | nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ |
d342894c | 3068 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ |
3069 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ | |
e7f70af1 | 3070 | nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */ |
d342894c | 3071 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ |
e4f67add DS |
3072 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ |
3073 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ | |
3074 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ | |
3075 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ | |
da8b43c0 | 3076 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ |
d342894c | 3077 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ |
3078 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ | |
05f47d69 | 3079 | nla_total_size(sizeof(struct ifla_vxlan_port_range)) + |
359a0ea9 TH |
3080 | nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ |
3081 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ | |
3082 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ | |
3083 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ | |
dfd8645e TH |
3084 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ |
3085 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ | |
d342894c | 3086 | 0; |
3087 | } | |
3088 | ||
3089 | static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
3090 | { | |
3091 | const struct vxlan_dev *vxlan = netdev_priv(dev); | |
c7995c43 | 3092 | const struct vxlan_rdst *dst = &vxlan->default_dst; |
05f47d69 | 3093 | struct ifla_vxlan_port_range ports = { |
0dfbdf41 TG |
3094 | .low = htons(vxlan->cfg.port_min), |
3095 | .high = htons(vxlan->cfg.port_max), | |
05f47d69 | 3096 | }; |
d342894c | 3097 | |
54bfd872 | 3098 | if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) |
d342894c | 3099 | goto nla_put_failure; |
3100 | ||
e4c7ed41 CW |
3101 | if (!vxlan_addr_any(&dst->remote_ip)) { |
3102 | if (dst->remote_ip.sa.sa_family == AF_INET) { | |
930345ea JB |
3103 | if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, |
3104 | dst->remote_ip.sin.sin_addr.s_addr)) | |
e4c7ed41 CW |
3105 | goto nla_put_failure; |
3106 | #if IS_ENABLED(CONFIG_IPV6) | |
3107 | } else { | |
930345ea JB |
3108 | if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, |
3109 | &dst->remote_ip.sin6.sin6_addr)) | |
e4c7ed41 CW |
3110 | goto nla_put_failure; |
3111 | #endif | |
3112 | } | |
3113 | } | |
d342894c | 3114 | |
c7995c43 | 3115 | if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) |
d342894c | 3116 | goto nla_put_failure; |
3117 | ||
0dfbdf41 TG |
3118 | if (!vxlan_addr_any(&vxlan->cfg.saddr)) { |
3119 | if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { | |
930345ea | 3120 | if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, |
0dfbdf41 | 3121 | vxlan->cfg.saddr.sin.sin_addr.s_addr)) |
e4c7ed41 CW |
3122 | goto nla_put_failure; |
3123 | #if IS_ENABLED(CONFIG_IPV6) | |
3124 | } else { | |
930345ea | 3125 | if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, |
0dfbdf41 | 3126 | &vxlan->cfg.saddr.sin6.sin6_addr)) |
e4c7ed41 CW |
3127 | goto nla_put_failure; |
3128 | #endif | |
3129 | } | |
3130 | } | |
d342894c | 3131 | |
0dfbdf41 TG |
3132 | if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || |
3133 | nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || | |
e7f70af1 | 3134 | nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) || |
e4f67add DS |
3135 | nla_put_u8(skb, IFLA_VXLAN_LEARNING, |
3136 | !!(vxlan->flags & VXLAN_F_LEARN)) || | |
3137 | nla_put_u8(skb, IFLA_VXLAN_PROXY, | |
3138 | !!(vxlan->flags & VXLAN_F_PROXY)) || | |
3139 | nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) || | |
3140 | nla_put_u8(skb, IFLA_VXLAN_L2MISS, | |
3141 | !!(vxlan->flags & VXLAN_F_L2MISS)) || | |
3142 | nla_put_u8(skb, IFLA_VXLAN_L3MISS, | |
3143 | !!(vxlan->flags & VXLAN_F_L3MISS)) || | |
da8b43c0 AS |
3144 | nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, |
3145 | !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) || | |
0dfbdf41 TG |
3146 | nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || |
3147 | nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || | |
3148 | nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || | |
359a0ea9 | 3149 | nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, |
6ceb31ca | 3150 | !(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM_TX)) || |
359a0ea9 TH |
3151 | nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, |
3152 | !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || | |
3153 | nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, | |
dfd8645e TH |
3154 | !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || |
3155 | nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, | |
3156 | !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) || | |
3157 | nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, | |
3158 | !!(vxlan->flags & VXLAN_F_REMCSUM_RX))) | |
d342894c | 3159 | goto nla_put_failure; |
3160 | ||
05f47d69 | 3161 | if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) |
3162 | goto nla_put_failure; | |
3163 | ||
3511494c TG |
3164 | if (vxlan->flags & VXLAN_F_GBP && |
3165 | nla_put_flag(skb, IFLA_VXLAN_GBP)) | |
3166 | goto nla_put_failure; | |
3167 | ||
e1e5314d JB |
3168 | if (vxlan->flags & VXLAN_F_GPE && |
3169 | nla_put_flag(skb, IFLA_VXLAN_GPE)) | |
3170 | goto nla_put_failure; | |
3171 | ||
0ace2ca8 TH |
3172 | if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL && |
3173 | nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) | |
3174 | goto nla_put_failure; | |
3175 | ||
d342894c | 3176 | return 0; |
3177 | ||
3178 | nla_put_failure: | |
3179 | return -EMSGSIZE; | |
3180 | } | |
3181 | ||
1728d4fa ND |
3182 | static struct net *vxlan_get_link_net(const struct net_device *dev) |
3183 | { | |
3184 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
3185 | ||
3186 | return vxlan->net; | |
3187 | } | |
3188 | ||
d342894c | 3189 | static struct rtnl_link_ops vxlan_link_ops __read_mostly = { |
3190 | .kind = "vxlan", | |
3191 | .maxtype = IFLA_VXLAN_MAX, | |
3192 | .policy = vxlan_policy, | |
3193 | .priv_size = sizeof(struct vxlan_dev), | |
3194 | .setup = vxlan_setup, | |
3195 | .validate = vxlan_validate, | |
3196 | .newlink = vxlan_newlink, | |
3197 | .dellink = vxlan_dellink, | |
3198 | .get_size = vxlan_get_size, | |
3199 | .fill_info = vxlan_fill_info, | |
1728d4fa | 3200 | .get_link_net = vxlan_get_link_net, |
d342894c | 3201 | }; |
3202 | ||
cf5da330 ND |
3203 | struct net_device *vxlan_dev_create(struct net *net, const char *name, |
3204 | u8 name_assign_type, | |
3205 | struct vxlan_config *conf) | |
3206 | { | |
3207 | struct nlattr *tb[IFLA_MAX + 1]; | |
3208 | struct net_device *dev; | |
3209 | int err; | |
3210 | ||
3211 | memset(&tb, 0, sizeof(tb)); | |
3212 | ||
3213 | dev = rtnl_create_link(net, name, name_assign_type, | |
3214 | &vxlan_link_ops, tb); | |
3215 | if (IS_ERR(dev)) | |
3216 | return dev; | |
3217 | ||
3218 | err = vxlan_dev_configure(net, dev, conf); | |
3219 | if (err < 0) { | |
3220 | free_netdev(dev); | |
3221 | return ERR_PTR(err); | |
3222 | } | |
3223 | ||
3224 | err = rtnl_configure_link(dev, NULL); | |
3225 | if (err < 0) { | |
3226 | LIST_HEAD(list_kill); | |
3227 | ||
3228 | vxlan_dellink(dev, &list_kill); | |
3229 | unregister_netdevice_many(&list_kill); | |
3230 | return ERR_PTR(err); | |
3231 | } | |
3232 | ||
3233 | return dev; | |
3234 | } | |
3235 | EXPORT_SYMBOL_GPL(vxlan_dev_create); | |
3236 | ||
acaf4e70 DB |
3237 | static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, |
3238 | struct net_device *dev) | |
3239 | { | |
3240 | struct vxlan_dev *vxlan, *next; | |
3241 | LIST_HEAD(list_kill); | |
3242 | ||
3243 | list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { | |
3244 | struct vxlan_rdst *dst = &vxlan->default_dst; | |
3245 | ||
3246 | /* In case we created vxlan device with carrier | |
3247 | * and we loose the carrier due to module unload | |
3248 | * we also need to remove vxlan device. In other | |
3249 | * cases, it's not necessary and remote_ifindex | |
3250 | * is 0 here, so no matches. | |
3251 | */ | |
3252 | if (dst->remote_ifindex == dev->ifindex) | |
3253 | vxlan_dellink(vxlan->dev, &list_kill); | |
3254 | } | |
3255 | ||
3256 | unregister_netdevice_many(&list_kill); | |
3257 | } | |
3258 | ||
b7aade15 HFS |
3259 | static int vxlan_netdevice_event(struct notifier_block *unused, |
3260 | unsigned long event, void *ptr) | |
acaf4e70 DB |
3261 | { |
3262 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
783c1463 | 3263 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); |
acaf4e70 | 3264 | |
783c1463 | 3265 | if (event == NETDEV_UNREGISTER) |
acaf4e70 | 3266 | vxlan_handle_lowerdev_unregister(vn, dev); |
7c46a640 | 3267 | else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO) |
b7aade15 | 3268 | vxlan_push_rx_ports(dev); |
acaf4e70 DB |
3269 | |
3270 | return NOTIFY_DONE; | |
3271 | } | |
3272 | ||
3273 | static struct notifier_block vxlan_notifier_block __read_mostly = { | |
b7aade15 | 3274 | .notifier_call = vxlan_netdevice_event, |
acaf4e70 DB |
3275 | }; |
3276 | ||
d342894c | 3277 | static __net_init int vxlan_init_net(struct net *net) |
3278 | { | |
3279 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
31fec5aa | 3280 | unsigned int h; |
d342894c | 3281 | |
553675fb | 3282 | INIT_LIST_HEAD(&vn->vxlan_list); |
1c51a915 | 3283 | spin_lock_init(&vn->sock_lock); |
d342894c | 3284 | |
553675fb | 3285 | for (h = 0; h < PORT_HASH_SIZE; ++h) |
3286 | INIT_HLIST_HEAD(&vn->sock_list[h]); | |
d342894c | 3287 | |
3288 | return 0; | |
3289 | } | |
3290 | ||
f01ec1c0 ND |
3291 | static void __net_exit vxlan_exit_net(struct net *net) |
3292 | { | |
3293 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
3294 | struct vxlan_dev *vxlan, *next; | |
3295 | struct net_device *dev, *aux; | |
3296 | LIST_HEAD(list); | |
3297 | ||
3298 | rtnl_lock(); | |
3299 | for_each_netdev_safe(net, dev, aux) | |
3300 | if (dev->rtnl_link_ops == &vxlan_link_ops) | |
3301 | unregister_netdevice_queue(dev, &list); | |
3302 | ||
3303 | list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { | |
3304 | /* If vxlan->dev is in the same netns, it has already been added | |
3305 | * to the list by the previous loop. | |
3306 | */ | |
58ce31cc TH |
3307 | if (!net_eq(dev_net(vxlan->dev), net)) { |
3308 | gro_cells_destroy(&vxlan->gro_cells); | |
13c3ed6a | 3309 | unregister_netdevice_queue(vxlan->dev, &list); |
58ce31cc | 3310 | } |
f01ec1c0 ND |
3311 | } |
3312 | ||
3313 | unregister_netdevice_many(&list); | |
3314 | rtnl_unlock(); | |
3315 | } | |
3316 | ||
d342894c | 3317 | static struct pernet_operations vxlan_net_ops = { |
3318 | .init = vxlan_init_net, | |
f01ec1c0 | 3319 | .exit = vxlan_exit_net, |
d342894c | 3320 | .id = &vxlan_net_id, |
3321 | .size = sizeof(struct vxlan_net), | |
3322 | }; | |
3323 | ||
3324 | static int __init vxlan_init_module(void) | |
3325 | { | |
3326 | int rc; | |
3327 | ||
3328 | get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); | |
3329 | ||
783c1463 | 3330 | rc = register_pernet_subsys(&vxlan_net_ops); |
d342894c | 3331 | if (rc) |
3332 | goto out1; | |
3333 | ||
acaf4e70 | 3334 | rc = register_netdevice_notifier(&vxlan_notifier_block); |
d342894c | 3335 | if (rc) |
3336 | goto out2; | |
3337 | ||
acaf4e70 DB |
3338 | rc = rtnl_link_register(&vxlan_link_ops); |
3339 | if (rc) | |
3340 | goto out3; | |
d342894c | 3341 | |
acaf4e70 DB |
3342 | return 0; |
3343 | out3: | |
3344 | unregister_netdevice_notifier(&vxlan_notifier_block); | |
d342894c | 3345 | out2: |
783c1463 | 3346 | unregister_pernet_subsys(&vxlan_net_ops); |
d342894c | 3347 | out1: |
3348 | return rc; | |
3349 | } | |
7332a13b | 3350 | late_initcall(vxlan_init_module); |
d342894c | 3351 | |
3352 | static void __exit vxlan_cleanup_module(void) | |
3353 | { | |
b7153984 | 3354 | rtnl_link_unregister(&vxlan_link_ops); |
acaf4e70 | 3355 | unregister_netdevice_notifier(&vxlan_notifier_block); |
783c1463 DB |
3356 | unregister_pernet_subsys(&vxlan_net_ops); |
3357 | /* rcu_barrier() is called by netns */ | |
d342894c | 3358 | } |
3359 | module_exit(vxlan_cleanup_module); | |
3360 | ||
3361 | MODULE_LICENSE("GPL"); | |
3362 | MODULE_VERSION(VXLAN_VERSION); | |
3b8df3c6 | 3363 | MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); |
ead5139a | 3364 | MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); |
d342894c | 3365 | MODULE_ALIAS_RTNL_LINK("vxlan"); |