Commit | Line | Data |
---|---|---|
d342894c | 1 | /* |
eb5ce439 | 2 | * VXLAN: Virtual eXtensible Local Area Network |
d342894c | 3 | * |
3b8df3c6 | 4 | * Copyright (c) 2012-2013 Vyatta Inc. |
d342894c | 5 | * |
6 | * This program is free software; you can redistribute it and/or modify | |
7 | * it under the terms of the GNU General Public License version 2 as | |
8 | * published by the Free Software Foundation. | |
d342894c | 9 | */ |
10 | ||
11 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
12 | ||
13 | #include <linux/kernel.h> | |
14 | #include <linux/types.h> | |
15 | #include <linux/module.h> | |
16 | #include <linux/errno.h> | |
17 | #include <linux/slab.h> | |
18 | #include <linux/skbuff.h> | |
19 | #include <linux/rculist.h> | |
20 | #include <linux/netdevice.h> | |
21 | #include <linux/in.h> | |
22 | #include <linux/ip.h> | |
23 | #include <linux/udp.h> | |
24 | #include <linux/igmp.h> | |
25 | #include <linux/etherdevice.h> | |
26 | #include <linux/if_ether.h> | |
1eaa8178 | 27 | #include <linux/if_vlan.h> |
d342894c | 28 | #include <linux/hash.h> |
1b13c97f | 29 | #include <linux/ethtool.h> |
e4f67add DS |
30 | #include <net/arp.h> |
31 | #include <net/ndisc.h> | |
d342894c | 32 | #include <net/ip.h> |
c5441932 | 33 | #include <net/ip_tunnels.h> |
d342894c | 34 | #include <net/icmp.h> |
35 | #include <net/udp.h> | |
3ee64f39 | 36 | #include <net/udp_tunnel.h> |
d342894c | 37 | #include <net/rtnetlink.h> |
38 | #include <net/route.h> | |
39 | #include <net/dsfield.h> | |
40 | #include <net/inet_ecn.h> | |
41 | #include <net/net_namespace.h> | |
42 | #include <net/netns/generic.h> | |
012a5729 | 43 | #include <net/vxlan.h> |
dc01e7d3 | 44 | #include <net/protocol.h> |
40d29af0 | 45 | |
e4c7ed41 CW |
46 | #if IS_ENABLED(CONFIG_IPV6) |
47 | #include <net/ipv6.h> | |
48 | #include <net/addrconf.h> | |
49 | #include <net/ip6_tunnel.h> | |
660d98ca | 50 | #include <net/ip6_checksum.h> |
e4c7ed41 | 51 | #endif |
ee122c79 | 52 | #include <net/dst_metadata.h> |
d342894c | 53 | |
54 | #define VXLAN_VERSION "0.1" | |
55 | ||
553675fb | 56 | #define PORT_HASH_BITS 8 |
57 | #define PORT_HASH_SIZE (1<<PORT_HASH_BITS) | |
d342894c | 58 | #define FDB_AGE_DEFAULT 300 /* 5 min */ |
59 | #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ | |
60 | ||
23c578bf | 61 | /* UDP port for VXLAN traffic. |
62 | * The IANA assigned port is 4789, but the Linux default is 8472 | |
234f5b73 | 63 | * for compatibility with early adopters. |
23c578bf | 64 | */ |
9daaa397 SH |
65 | static unsigned short vxlan_port __read_mostly = 8472; |
66 | module_param_named(udp_port, vxlan_port, ushort, 0444); | |
d342894c | 67 | MODULE_PARM_DESC(udp_port, "Destination UDP port"); |
68 | ||
69 | static bool log_ecn_error = true; | |
70 | module_param(log_ecn_error, bool, 0644); | |
71 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); | |
72 | ||
60d9d4c6 | 73 | static int vxlan_net_id; |
0dfbdf41 | 74 | static struct rtnl_link_ops vxlan_link_ops; |
553675fb | 75 | |
7256eac1 | 76 | static const u8 all_zeros_mac[ETH_ALEN + 2]; |
afbd8bae | 77 | |
205f356d | 78 | static int vxlan_sock_add(struct vxlan_dev *vxlan); |
614732ea | 79 | |
553675fb | 80 | /* per-network namespace private data for this module */ |
81 | struct vxlan_net { | |
82 | struct list_head vxlan_list; | |
83 | struct hlist_head sock_list[PORT_HASH_SIZE]; | |
1c51a915 | 84 | spinlock_t sock_lock; |
553675fb | 85 | }; |
86 | ||
d342894c | 87 | /* Forwarding table entry */ |
88 | struct vxlan_fdb { | |
89 | struct hlist_node hlist; /* linked list of entries */ | |
90 | struct rcu_head rcu; | |
91 | unsigned long updated; /* jiffies */ | |
92 | unsigned long used; | |
3e61aa8f | 93 | struct list_head remotes; |
7177a3b0 | 94 | u8 eth_addr[ETH_ALEN]; |
d342894c | 95 | u16 state; /* see ndm_state */ |
ae884082 | 96 | u8 flags; /* see ndm_flags */ |
d342894c | 97 | }; |
98 | ||
d342894c | 99 | /* salt for hash table */ |
100 | static u32 vxlan_salt __read_mostly; | |
758c57d1 | 101 | static struct workqueue_struct *vxlan_wq; |
d342894c | 102 | |
ee122c79 TG |
103 | static inline bool vxlan_collect_metadata(struct vxlan_sock *vs) |
104 | { | |
e7030878 TG |
105 | return vs->flags & VXLAN_F_COLLECT_METADATA || |
106 | ip_tunnel_collect_metadata(); | |
ee122c79 TG |
107 | } |
108 | ||
e4c7ed41 CW |
109 | #if IS_ENABLED(CONFIG_IPV6) |
110 | static inline | |
111 | bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b) | |
112 | { | |
f0ef3126 JB |
113 | if (a->sa.sa_family != b->sa.sa_family) |
114 | return false; | |
115 | if (a->sa.sa_family == AF_INET6) | |
116 | return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr); | |
117 | else | |
118 | return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr; | |
e4c7ed41 CW |
119 | } |
120 | ||
121 | static inline bool vxlan_addr_any(const union vxlan_addr *ipa) | |
122 | { | |
f0ef3126 JB |
123 | if (ipa->sa.sa_family == AF_INET6) |
124 | return ipv6_addr_any(&ipa->sin6.sin6_addr); | |
125 | else | |
126 | return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY); | |
e4c7ed41 CW |
127 | } |
128 | ||
129 | static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa) | |
130 | { | |
f0ef3126 JB |
131 | if (ipa->sa.sa_family == AF_INET6) |
132 | return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr); | |
133 | else | |
134 | return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr)); | |
e4c7ed41 CW |
135 | } |
136 | ||
137 | static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla) | |
138 | { | |
f0ef3126 | 139 | if (nla_len(nla) >= sizeof(struct in6_addr)) { |
67b61f6c | 140 | ip->sin6.sin6_addr = nla_get_in6_addr(nla); |
f0ef3126 JB |
141 | ip->sa.sa_family = AF_INET6; |
142 | return 0; | |
143 | } else if (nla_len(nla) >= sizeof(__be32)) { | |
67b61f6c | 144 | ip->sin.sin_addr.s_addr = nla_get_in_addr(nla); |
f0ef3126 JB |
145 | ip->sa.sa_family = AF_INET; |
146 | return 0; | |
147 | } else { | |
148 | return -EAFNOSUPPORT; | |
149 | } | |
e4c7ed41 CW |
150 | } |
151 | ||
152 | static int vxlan_nla_put_addr(struct sk_buff *skb, int attr, | |
f0ef3126 | 153 | const union vxlan_addr *ip) |
e4c7ed41 | 154 | { |
f0ef3126 | 155 | if (ip->sa.sa_family == AF_INET6) |
930345ea | 156 | return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr); |
f0ef3126 | 157 | else |
930345ea | 158 | return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr); |
e4c7ed41 CW |
159 | } |
160 | ||
161 | #else /* !CONFIG_IPV6 */ | |
162 | ||
163 | static inline | |
164 | bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b) | |
165 | { | |
f0ef3126 | 166 | return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr; |
e4c7ed41 CW |
167 | } |
168 | ||
169 | static inline bool vxlan_addr_any(const union vxlan_addr *ipa) | |
170 | { | |
f0ef3126 | 171 | return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY); |
e4c7ed41 CW |
172 | } |
173 | ||
174 | static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa) | |
175 | { | |
f0ef3126 | 176 | return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr)); |
e4c7ed41 CW |
177 | } |
178 | ||
179 | static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla) | |
180 | { | |
f0ef3126 JB |
181 | if (nla_len(nla) >= sizeof(struct in6_addr)) { |
182 | return -EAFNOSUPPORT; | |
183 | } else if (nla_len(nla) >= sizeof(__be32)) { | |
67b61f6c | 184 | ip->sin.sin_addr.s_addr = nla_get_in_addr(nla); |
f0ef3126 JB |
185 | ip->sa.sa_family = AF_INET; |
186 | return 0; | |
187 | } else { | |
188 | return -EAFNOSUPPORT; | |
189 | } | |
e4c7ed41 CW |
190 | } |
191 | ||
192 | static int vxlan_nla_put_addr(struct sk_buff *skb, int attr, | |
f0ef3126 | 193 | const union vxlan_addr *ip) |
e4c7ed41 | 194 | { |
930345ea | 195 | return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr); |
e4c7ed41 CW |
196 | } |
197 | #endif | |
198 | ||
553675fb | 199 | /* Virtual Network hash table head */ |
54bfd872 | 200 | static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni) |
553675fb | 201 | { |
54bfd872 | 202 | return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)]; |
553675fb | 203 | } |
204 | ||
205 | /* Socket hash table head */ | |
206 | static inline struct hlist_head *vs_head(struct net *net, __be16 port) | |
d342894c | 207 | { |
208 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
209 | ||
553675fb | 210 | return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)]; |
211 | } | |
212 | ||
3e61aa8f SH |
213 | /* First remote destination for a forwarding entry. |
214 | * Guaranteed to be non-NULL because remotes are never deleted. | |
215 | */ | |
5ca5461c | 216 | static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb) |
3e61aa8f | 217 | { |
5ca5461c | 218 | return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list); |
219 | } | |
220 | ||
221 | static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb) | |
222 | { | |
223 | return list_first_entry(&fdb->remotes, struct vxlan_rdst, list); | |
3e61aa8f SH |
224 | } |
225 | ||
ac5132d1 TG |
226 | /* Find VXLAN socket based on network namespace, address family and UDP port |
227 | * and enabled unshareable flags. | |
228 | */ | |
229 | static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family, | |
230 | __be16 port, u32 flags) | |
553675fb | 231 | { |
232 | struct vxlan_sock *vs; | |
af33c1ad TH |
233 | |
234 | flags &= VXLAN_F_RCV_FLAGS; | |
553675fb | 235 | |
236 | hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) { | |
19ca9fc1 | 237 | if (inet_sk(vs->sock->sk)->inet_sport == port && |
705cc62f | 238 | vxlan_get_sk_family(vs) == family && |
af33c1ad | 239 | vs->flags == flags) |
553675fb | 240 | return vs; |
241 | } | |
242 | return NULL; | |
d342894c | 243 | } |
244 | ||
54bfd872 | 245 | static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, __be32 vni) |
d342894c | 246 | { |
247 | struct vxlan_dev *vxlan; | |
d342894c | 248 | |
54bfd872 JB |
249 | hlist_for_each_entry_rcu(vxlan, vni_head(vs, vni), hlist) { |
250 | if (vxlan->default_dst.remote_vni == vni) | |
d342894c | 251 | return vxlan; |
252 | } | |
253 | ||
254 | return NULL; | |
255 | } | |
256 | ||
5cfccc5a | 257 | /* Look up VNI in a per net namespace table */ |
54bfd872 | 258 | static struct vxlan_dev *vxlan_find_vni(struct net *net, __be32 vni, |
ac5132d1 TG |
259 | sa_family_t family, __be16 port, |
260 | u32 flags) | |
5cfccc5a PS |
261 | { |
262 | struct vxlan_sock *vs; | |
263 | ||
ac5132d1 | 264 | vs = vxlan_find_sock(net, family, port, flags); |
5cfccc5a PS |
265 | if (!vs) |
266 | return NULL; | |
267 | ||
54bfd872 | 268 | return vxlan_vs_find_vni(vs, vni); |
5cfccc5a PS |
269 | } |
270 | ||
d342894c | 271 | /* Fill in neighbour message in skbuff. */ |
272 | static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, | |
234f5b73 SH |
273 | const struct vxlan_fdb *fdb, |
274 | u32 portid, u32 seq, int type, unsigned int flags, | |
275 | const struct vxlan_rdst *rdst) | |
d342894c | 276 | { |
277 | unsigned long now = jiffies; | |
278 | struct nda_cacheinfo ci; | |
279 | struct nlmsghdr *nlh; | |
280 | struct ndmsg *ndm; | |
e4f67add | 281 | bool send_ip, send_eth; |
d342894c | 282 | |
283 | nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); | |
284 | if (nlh == NULL) | |
285 | return -EMSGSIZE; | |
286 | ||
287 | ndm = nlmsg_data(nlh); | |
288 | memset(ndm, 0, sizeof(*ndm)); | |
e4f67add DS |
289 | |
290 | send_eth = send_ip = true; | |
291 | ||
292 | if (type == RTM_GETNEIGH) { | |
293 | ndm->ndm_family = AF_INET; | |
e4c7ed41 | 294 | send_ip = !vxlan_addr_any(&rdst->remote_ip); |
e4f67add DS |
295 | send_eth = !is_zero_ether_addr(fdb->eth_addr); |
296 | } else | |
297 | ndm->ndm_family = AF_BRIDGE; | |
d342894c | 298 | ndm->ndm_state = fdb->state; |
299 | ndm->ndm_ifindex = vxlan->dev->ifindex; | |
ae884082 | 300 | ndm->ndm_flags = fdb->flags; |
545469f7 | 301 | ndm->ndm_type = RTN_UNICAST; |
d342894c | 302 | |
193523bf | 303 | if (!net_eq(dev_net(vxlan->dev), vxlan->net) && |
4967082b | 304 | nla_put_s32(skb, NDA_LINK_NETNSID, |
7a0877d4 | 305 | peernet2id_alloc(dev_net(vxlan->dev), vxlan->net))) |
193523bf ND |
306 | goto nla_put_failure; |
307 | ||
e4f67add | 308 | if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr)) |
d342894c | 309 | goto nla_put_failure; |
310 | ||
e4c7ed41 | 311 | if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip)) |
6681712d DS |
312 | goto nla_put_failure; |
313 | ||
0dfbdf41 | 314 | if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port && |
6681712d DS |
315 | nla_put_be16(skb, NDA_PORT, rdst->remote_port)) |
316 | goto nla_put_failure; | |
c7995c43 | 317 | if (rdst->remote_vni != vxlan->default_dst.remote_vni && |
54bfd872 | 318 | nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni))) |
6681712d DS |
319 | goto nla_put_failure; |
320 | if (rdst->remote_ifindex && | |
321 | nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) | |
d342894c | 322 | goto nla_put_failure; |
323 | ||
324 | ci.ndm_used = jiffies_to_clock_t(now - fdb->used); | |
325 | ci.ndm_confirmed = 0; | |
326 | ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated); | |
327 | ci.ndm_refcnt = 0; | |
328 | ||
329 | if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) | |
330 | goto nla_put_failure; | |
331 | ||
053c095a JB |
332 | nlmsg_end(skb, nlh); |
333 | return 0; | |
d342894c | 334 | |
335 | nla_put_failure: | |
336 | nlmsg_cancel(skb, nlh); | |
337 | return -EMSGSIZE; | |
338 | } | |
339 | ||
340 | static inline size_t vxlan_nlmsg_size(void) | |
341 | { | |
342 | return NLMSG_ALIGN(sizeof(struct ndmsg)) | |
343 | + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ | |
e4c7ed41 | 344 | + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */ |
73cf3317 | 345 | + nla_total_size(sizeof(__be16)) /* NDA_PORT */ |
6681712d DS |
346 | + nla_total_size(sizeof(__be32)) /* NDA_VNI */ |
347 | + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ | |
4967082b | 348 | + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */ |
d342894c | 349 | + nla_total_size(sizeof(struct nda_cacheinfo)); |
350 | } | |
351 | ||
9e4b93f9 ND |
352 | static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb, |
353 | struct vxlan_rdst *rd, int type) | |
d342894c | 354 | { |
355 | struct net *net = dev_net(vxlan->dev); | |
356 | struct sk_buff *skb; | |
357 | int err = -ENOBUFS; | |
358 | ||
359 | skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); | |
360 | if (skb == NULL) | |
361 | goto errout; | |
362 | ||
9e4b93f9 | 363 | err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd); |
d342894c | 364 | if (err < 0) { |
365 | /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ | |
366 | WARN_ON(err == -EMSGSIZE); | |
367 | kfree_skb(skb); | |
368 | goto errout; | |
369 | } | |
370 | ||
371 | rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); | |
372 | return; | |
373 | errout: | |
374 | if (err < 0) | |
375 | rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); | |
376 | } | |
377 | ||
e4c7ed41 | 378 | static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa) |
e4f67add DS |
379 | { |
380 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
bb3fd687 SH |
381 | struct vxlan_fdb f = { |
382 | .state = NUD_STALE, | |
383 | }; | |
384 | struct vxlan_rdst remote = { | |
e4c7ed41 | 385 | .remote_ip = *ipa, /* goes to NDA_DST */ |
54bfd872 | 386 | .remote_vni = cpu_to_be32(VXLAN_N_VID), |
bb3fd687 | 387 | }; |
3e61aa8f | 388 | |
9e4b93f9 | 389 | vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH); |
e4f67add DS |
390 | } |
391 | ||
392 | static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) | |
393 | { | |
bb3fd687 SH |
394 | struct vxlan_fdb f = { |
395 | .state = NUD_STALE, | |
396 | }; | |
9e4b93f9 | 397 | struct vxlan_rdst remote = { }; |
e4f67add | 398 | |
e4f67add DS |
399 | memcpy(f.eth_addr, eth_addr, ETH_ALEN); |
400 | ||
9e4b93f9 | 401 | vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH); |
e4f67add DS |
402 | } |
403 | ||
d342894c | 404 | /* Hash Ethernet address */ |
405 | static u32 eth_hash(const unsigned char *addr) | |
406 | { | |
407 | u64 value = get_unaligned((u64 *)addr); | |
408 | ||
409 | /* only want 6 bytes */ | |
410 | #ifdef __BIG_ENDIAN | |
d342894c | 411 | value >>= 16; |
321fb991 | 412 | #else |
413 | value <<= 16; | |
d342894c | 414 | #endif |
415 | return hash_64(value, FDB_HASH_BITS); | |
416 | } | |
417 | ||
418 | /* Hash chain to use given mac address */ | |
419 | static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, | |
420 | const u8 *mac) | |
421 | { | |
422 | return &vxlan->fdb_head[eth_hash(mac)]; | |
423 | } | |
424 | ||
425 | /* Look up Ethernet address in forwarding table */ | |
014be2c8 | 426 | static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, |
d342894c | 427 | const u8 *mac) |
d342894c | 428 | { |
429 | struct hlist_head *head = vxlan_fdb_head(vxlan, mac); | |
430 | struct vxlan_fdb *f; | |
d342894c | 431 | |
b67bfe0d | 432 | hlist_for_each_entry_rcu(f, head, hlist) { |
7367d0b5 | 433 | if (ether_addr_equal(mac, f->eth_addr)) |
d342894c | 434 | return f; |
435 | } | |
436 | ||
437 | return NULL; | |
438 | } | |
439 | ||
014be2c8 SS |
440 | static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, |
441 | const u8 *mac) | |
442 | { | |
443 | struct vxlan_fdb *f; | |
444 | ||
445 | f = __vxlan_find_mac(vxlan, mac); | |
446 | if (f) | |
447 | f->used = jiffies; | |
448 | ||
449 | return f; | |
450 | } | |
451 | ||
a5e7c10a MR |
452 | /* caller should hold vxlan->hash_lock */ |
453 | static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f, | |
e4c7ed41 | 454 | union vxlan_addr *ip, __be16 port, |
54bfd872 | 455 | __be32 vni, __u32 ifindex) |
6681712d | 456 | { |
3e61aa8f | 457 | struct vxlan_rdst *rd; |
6681712d | 458 | |
3e61aa8f | 459 | list_for_each_entry(rd, &f->remotes, list) { |
e4c7ed41 | 460 | if (vxlan_addr_equal(&rd->remote_ip, ip) && |
6681712d DS |
461 | rd->remote_port == port && |
462 | rd->remote_vni == vni && | |
463 | rd->remote_ifindex == ifindex) | |
a5e7c10a | 464 | return rd; |
6681712d | 465 | } |
3e61aa8f | 466 | |
a5e7c10a MR |
467 | return NULL; |
468 | } | |
469 | ||
906dc186 TR |
470 | /* Replace destination of unicast mac */ |
471 | static int vxlan_fdb_replace(struct vxlan_fdb *f, | |
54bfd872 JB |
472 | union vxlan_addr *ip, __be16 port, __be32 vni, |
473 | __u32 ifindex) | |
906dc186 TR |
474 | { |
475 | struct vxlan_rdst *rd; | |
476 | ||
477 | rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); | |
478 | if (rd) | |
479 | return 0; | |
480 | ||
481 | rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list); | |
482 | if (!rd) | |
483 | return 0; | |
0c1d70af PA |
484 | |
485 | dst_cache_reset(&rd->dst_cache); | |
e4c7ed41 | 486 | rd->remote_ip = *ip; |
906dc186 TR |
487 | rd->remote_port = port; |
488 | rd->remote_vni = vni; | |
489 | rd->remote_ifindex = ifindex; | |
490 | return 1; | |
491 | } | |
492 | ||
a5e7c10a MR |
493 | /* Add/update destinations for multicast */ |
494 | static int vxlan_fdb_append(struct vxlan_fdb *f, | |
54bfd872 | 495 | union vxlan_addr *ip, __be16 port, __be32 vni, |
9e4b93f9 | 496 | __u32 ifindex, struct vxlan_rdst **rdp) |
a5e7c10a MR |
497 | { |
498 | struct vxlan_rdst *rd; | |
499 | ||
500 | rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex); | |
501 | if (rd) | |
502 | return 0; | |
503 | ||
6681712d DS |
504 | rd = kmalloc(sizeof(*rd), GFP_ATOMIC); |
505 | if (rd == NULL) | |
506 | return -ENOBUFS; | |
0c1d70af PA |
507 | |
508 | if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) { | |
509 | kfree(rd); | |
510 | return -ENOBUFS; | |
511 | } | |
512 | ||
e4c7ed41 | 513 | rd->remote_ip = *ip; |
6681712d DS |
514 | rd->remote_port = port; |
515 | rd->remote_vni = vni; | |
516 | rd->remote_ifindex = ifindex; | |
3e61aa8f SH |
517 | |
518 | list_add_tail_rcu(&rd->list, &f->remotes); | |
519 | ||
9e4b93f9 | 520 | *rdp = rd; |
6681712d DS |
521 | return 1; |
522 | } | |
523 | ||
dfd8645e TH |
524 | static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb, |
525 | unsigned int off, | |
526 | struct vxlanhdr *vh, size_t hdrlen, | |
54bfd872 JB |
527 | __be32 vni_field, |
528 | struct gro_remcsum *grc, | |
0ace2ca8 | 529 | bool nopartial) |
dfd8645e | 530 | { |
b7fe10e5 | 531 | size_t start, offset; |
dfd8645e TH |
532 | |
533 | if (skb->remcsum_offload) | |
b7fe10e5 | 534 | return vh; |
dfd8645e TH |
535 | |
536 | if (!NAPI_GRO_CB(skb)->csum_valid) | |
537 | return NULL; | |
538 | ||
54bfd872 JB |
539 | start = vxlan_rco_start(vni_field); |
540 | offset = start + vxlan_rco_offset(vni_field); | |
dfd8645e | 541 | |
b7fe10e5 TH |
542 | vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen, |
543 | start, offset, grc, nopartial); | |
dfd8645e TH |
544 | |
545 | skb->remcsum_offload = 1; | |
546 | ||
547 | return vh; | |
548 | } | |
549 | ||
a2b12f3c TH |
550 | static struct sk_buff **vxlan_gro_receive(struct sk_buff **head, |
551 | struct sk_buff *skb, | |
552 | struct udp_offload *uoff) | |
dc01e7d3 OG |
553 | { |
554 | struct sk_buff *p, **pp = NULL; | |
555 | struct vxlanhdr *vh, *vh2; | |
9b174d88 | 556 | unsigned int hlen, off_vx; |
dc01e7d3 | 557 | int flush = 1; |
dfd8645e TH |
558 | struct vxlan_sock *vs = container_of(uoff, struct vxlan_sock, |
559 | udp_offloads); | |
54bfd872 | 560 | __be32 flags; |
26c4f7da TH |
561 | struct gro_remcsum grc; |
562 | ||
563 | skb_gro_remcsum_init(&grc); | |
dc01e7d3 OG |
564 | |
565 | off_vx = skb_gro_offset(skb); | |
566 | hlen = off_vx + sizeof(*vh); | |
567 | vh = skb_gro_header_fast(skb, off_vx); | |
568 | if (skb_gro_header_hard(skb, hlen)) { | |
569 | vh = skb_gro_header_slow(skb, hlen, off_vx); | |
570 | if (unlikely(!vh)) | |
571 | goto out; | |
572 | } | |
dc01e7d3 | 573 | |
dfd8645e TH |
574 | skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr)); |
575 | ||
54bfd872 | 576 | flags = vh->vx_flags; |
dfd8645e TH |
577 | |
578 | if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { | |
579 | vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr), | |
54bfd872 | 580 | vh->vx_vni, &grc, |
0ace2ca8 TH |
581 | !!(vs->flags & |
582 | VXLAN_F_REMCSUM_NOPARTIAL)); | |
dfd8645e TH |
583 | |
584 | if (!vh) | |
585 | goto out; | |
586 | } | |
587 | ||
b7fe10e5 TH |
588 | skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */ |
589 | ||
dc01e7d3 OG |
590 | flush = 0; |
591 | ||
592 | for (p = *head; p; p = p->next) { | |
593 | if (!NAPI_GRO_CB(p)->same_flow) | |
594 | continue; | |
595 | ||
596 | vh2 = (struct vxlanhdr *)(p->data + off_vx); | |
3511494c TG |
597 | if (vh->vx_flags != vh2->vx_flags || |
598 | vh->vx_vni != vh2->vx_vni) { | |
dc01e7d3 OG |
599 | NAPI_GRO_CB(p)->same_flow = 0; |
600 | continue; | |
601 | } | |
dc01e7d3 OG |
602 | } |
603 | ||
9b174d88 | 604 | pp = eth_gro_receive(head, skb); |
dc01e7d3 | 605 | |
dc01e7d3 | 606 | out: |
26c4f7da | 607 | skb_gro_remcsum_cleanup(skb, &grc); |
dc01e7d3 OG |
608 | NAPI_GRO_CB(skb)->flush |= flush; |
609 | ||
610 | return pp; | |
611 | } | |
612 | ||
a2b12f3c TH |
613 | static int vxlan_gro_complete(struct sk_buff *skb, int nhoff, |
614 | struct udp_offload *uoff) | |
dc01e7d3 | 615 | { |
cfdf1e1b JG |
616 | udp_tunnel_gro_complete(skb, nhoff); |
617 | ||
9b174d88 | 618 | return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr)); |
dc01e7d3 OG |
619 | } |
620 | ||
53cf5275 | 621 | /* Notify netdevs that UDP port started listening */ |
dc01e7d3 | 622 | static void vxlan_notify_add_rx_port(struct vxlan_sock *vs) |
53cf5275 JG |
623 | { |
624 | struct net_device *dev; | |
dc01e7d3 | 625 | struct sock *sk = vs->sock->sk; |
53cf5275 | 626 | struct net *net = sock_net(sk); |
705cc62f | 627 | sa_family_t sa_family = vxlan_get_sk_family(vs); |
35e42379 | 628 | __be16 port = inet_sk(sk)->inet_sport; |
dc01e7d3 OG |
629 | int err; |
630 | ||
631 | if (sa_family == AF_INET) { | |
787d7ac3 | 632 | err = udp_add_offload(net, &vs->udp_offloads); |
dc01e7d3 OG |
633 | if (err) |
634 | pr_warn("vxlan: udp_add_offload failed with status %d\n", err); | |
635 | } | |
53cf5275 JG |
636 | |
637 | rcu_read_lock(); | |
638 | for_each_netdev_rcu(net, dev) { | |
639 | if (dev->netdev_ops->ndo_add_vxlan_port) | |
640 | dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family, | |
641 | port); | |
642 | } | |
643 | rcu_read_unlock(); | |
644 | } | |
645 | ||
646 | /* Notify netdevs that UDP port is no more listening */ | |
dc01e7d3 | 647 | static void vxlan_notify_del_rx_port(struct vxlan_sock *vs) |
53cf5275 JG |
648 | { |
649 | struct net_device *dev; | |
dc01e7d3 | 650 | struct sock *sk = vs->sock->sk; |
53cf5275 | 651 | struct net *net = sock_net(sk); |
705cc62f | 652 | sa_family_t sa_family = vxlan_get_sk_family(vs); |
35e42379 | 653 | __be16 port = inet_sk(sk)->inet_sport; |
53cf5275 JG |
654 | |
655 | rcu_read_lock(); | |
656 | for_each_netdev_rcu(net, dev) { | |
657 | if (dev->netdev_ops->ndo_del_vxlan_port) | |
658 | dev->netdev_ops->ndo_del_vxlan_port(dev, sa_family, | |
659 | port); | |
660 | } | |
661 | rcu_read_unlock(); | |
dc01e7d3 OG |
662 | |
663 | if (sa_family == AF_INET) | |
664 | udp_del_offload(&vs->udp_offloads); | |
53cf5275 JG |
665 | } |
666 | ||
d342894c | 667 | /* Add new entry to forwarding table -- assumes lock held */ |
668 | static int vxlan_fdb_create(struct vxlan_dev *vxlan, | |
e4c7ed41 | 669 | const u8 *mac, union vxlan_addr *ip, |
6681712d | 670 | __u16 state, __u16 flags, |
54bfd872 | 671 | __be16 port, __be32 vni, __u32 ifindex, |
ae884082 | 672 | __u8 ndm_flags) |
d342894c | 673 | { |
9e4b93f9 | 674 | struct vxlan_rdst *rd = NULL; |
d342894c | 675 | struct vxlan_fdb *f; |
676 | int notify = 0; | |
677 | ||
014be2c8 | 678 | f = __vxlan_find_mac(vxlan, mac); |
d342894c | 679 | if (f) { |
680 | if (flags & NLM_F_EXCL) { | |
681 | netdev_dbg(vxlan->dev, | |
682 | "lost race to create %pM\n", mac); | |
683 | return -EEXIST; | |
684 | } | |
685 | if (f->state != state) { | |
686 | f->state = state; | |
687 | f->updated = jiffies; | |
688 | notify = 1; | |
689 | } | |
ae884082 DS |
690 | if (f->flags != ndm_flags) { |
691 | f->flags = ndm_flags; | |
692 | f->updated = jiffies; | |
693 | notify = 1; | |
694 | } | |
906dc186 TR |
695 | if ((flags & NLM_F_REPLACE)) { |
696 | /* Only change unicasts */ | |
697 | if (!(is_multicast_ether_addr(f->eth_addr) || | |
698 | is_zero_ether_addr(f->eth_addr))) { | |
60840429 | 699 | notify |= vxlan_fdb_replace(f, ip, port, vni, |
906dc186 | 700 | ifindex); |
906dc186 TR |
701 | } else |
702 | return -EOPNOTSUPP; | |
703 | } | |
6681712d | 704 | if ((flags & NLM_F_APPEND) && |
58e4c767 MR |
705 | (is_multicast_ether_addr(f->eth_addr) || |
706 | is_zero_ether_addr(f->eth_addr))) { | |
9e4b93f9 ND |
707 | int rc = vxlan_fdb_append(f, ip, port, vni, ifindex, |
708 | &rd); | |
6681712d DS |
709 | |
710 | if (rc < 0) | |
711 | return rc; | |
712 | notify |= rc; | |
713 | } | |
d342894c | 714 | } else { |
715 | if (!(flags & NLM_F_CREATE)) | |
716 | return -ENOENT; | |
717 | ||
0dfbdf41 TG |
718 | if (vxlan->cfg.addrmax && |
719 | vxlan->addrcnt >= vxlan->cfg.addrmax) | |
d342894c | 720 | return -ENOSPC; |
721 | ||
906dc186 TR |
722 | /* Disallow replace to add a multicast entry */ |
723 | if ((flags & NLM_F_REPLACE) && | |
724 | (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac))) | |
725 | return -EOPNOTSUPP; | |
726 | ||
e4c7ed41 | 727 | netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip); |
d342894c | 728 | f = kmalloc(sizeof(*f), GFP_ATOMIC); |
729 | if (!f) | |
730 | return -ENOMEM; | |
731 | ||
732 | notify = 1; | |
d342894c | 733 | f->state = state; |
ae884082 | 734 | f->flags = ndm_flags; |
d342894c | 735 | f->updated = f->used = jiffies; |
3e61aa8f | 736 | INIT_LIST_HEAD(&f->remotes); |
d342894c | 737 | memcpy(f->eth_addr, mac, ETH_ALEN); |
738 | ||
9e4b93f9 | 739 | vxlan_fdb_append(f, ip, port, vni, ifindex, &rd); |
3e61aa8f | 740 | |
d342894c | 741 | ++vxlan->addrcnt; |
742 | hlist_add_head_rcu(&f->hlist, | |
743 | vxlan_fdb_head(vxlan, mac)); | |
744 | } | |
745 | ||
9e4b93f9 ND |
746 | if (notify) { |
747 | if (rd == NULL) | |
748 | rd = first_remote_rtnl(f); | |
749 | vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH); | |
750 | } | |
d342894c | 751 | |
752 | return 0; | |
753 | } | |
754 | ||
6706c82e | 755 | static void vxlan_fdb_free(struct rcu_head *head) |
6681712d DS |
756 | { |
757 | struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); | |
3e61aa8f | 758 | struct vxlan_rdst *rd, *nd; |
6681712d | 759 | |
0c1d70af PA |
760 | list_for_each_entry_safe(rd, nd, &f->remotes, list) { |
761 | dst_cache_destroy(&rd->dst_cache); | |
6681712d | 762 | kfree(rd); |
0c1d70af | 763 | } |
6681712d DS |
764 | kfree(f); |
765 | } | |
766 | ||
d342894c | 767 | static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f) |
768 | { | |
769 | netdev_dbg(vxlan->dev, | |
770 | "delete %pM\n", f->eth_addr); | |
771 | ||
772 | --vxlan->addrcnt; | |
9e4b93f9 | 773 | vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH); |
d342894c | 774 | |
775 | hlist_del_rcu(&f->hlist); | |
6681712d | 776 | call_rcu(&f->rcu, vxlan_fdb_free); |
d342894c | 777 | } |
778 | ||
f0b074be | 779 | static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan, |
54bfd872 JB |
780 | union vxlan_addr *ip, __be16 *port, __be32 *vni, |
781 | u32 *ifindex) | |
d342894c | 782 | { |
6681712d | 783 | struct net *net = dev_net(vxlan->dev); |
e4c7ed41 | 784 | int err; |
d342894c | 785 | |
f0b074be | 786 | if (tb[NDA_DST]) { |
e4c7ed41 CW |
787 | err = vxlan_nla_get_addr(ip, tb[NDA_DST]); |
788 | if (err) | |
789 | return err; | |
f0b074be | 790 | } else { |
e4c7ed41 CW |
791 | union vxlan_addr *remote = &vxlan->default_dst.remote_ip; |
792 | if (remote->sa.sa_family == AF_INET) { | |
793 | ip->sin.sin_addr.s_addr = htonl(INADDR_ANY); | |
794 | ip->sa.sa_family = AF_INET; | |
795 | #if IS_ENABLED(CONFIG_IPV6) | |
796 | } else { | |
797 | ip->sin6.sin6_addr = in6addr_any; | |
798 | ip->sa.sa_family = AF_INET6; | |
799 | #endif | |
800 | } | |
f0b074be | 801 | } |
d342894c | 802 | |
6681712d | 803 | if (tb[NDA_PORT]) { |
73cf3317 | 804 | if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) |
6681712d | 805 | return -EINVAL; |
f0b074be MR |
806 | *port = nla_get_be16(tb[NDA_PORT]); |
807 | } else { | |
0dfbdf41 | 808 | *port = vxlan->cfg.dst_port; |
f0b074be | 809 | } |
6681712d DS |
810 | |
811 | if (tb[NDA_VNI]) { | |
812 | if (nla_len(tb[NDA_VNI]) != sizeof(u32)) | |
813 | return -EINVAL; | |
54bfd872 | 814 | *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI])); |
f0b074be MR |
815 | } else { |
816 | *vni = vxlan->default_dst.remote_vni; | |
817 | } | |
6681712d DS |
818 | |
819 | if (tb[NDA_IFINDEX]) { | |
5abb0029 | 820 | struct net_device *tdev; |
6681712d DS |
821 | |
822 | if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) | |
823 | return -EINVAL; | |
f0b074be | 824 | *ifindex = nla_get_u32(tb[NDA_IFINDEX]); |
73763949 | 825 | tdev = __dev_get_by_index(net, *ifindex); |
5abb0029 | 826 | if (!tdev) |
6681712d | 827 | return -EADDRNOTAVAIL; |
f0b074be MR |
828 | } else { |
829 | *ifindex = 0; | |
830 | } | |
831 | ||
832 | return 0; | |
833 | } | |
834 | ||
835 | /* Add static entry (via netlink) */ | |
836 | static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], | |
837 | struct net_device *dev, | |
f6f6424b | 838 | const unsigned char *addr, u16 vid, u16 flags) |
f0b074be MR |
839 | { |
840 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
841 | /* struct net *net = dev_net(vxlan->dev); */ | |
e4c7ed41 | 842 | union vxlan_addr ip; |
f0b074be | 843 | __be16 port; |
54bfd872 JB |
844 | __be32 vni; |
845 | u32 ifindex; | |
f0b074be MR |
846 | int err; |
847 | ||
848 | if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { | |
849 | pr_info("RTM_NEWNEIGH with invalid state %#x\n", | |
850 | ndm->ndm_state); | |
851 | return -EINVAL; | |
852 | } | |
853 | ||
854 | if (tb[NDA_DST] == NULL) | |
855 | return -EINVAL; | |
856 | ||
857 | err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex); | |
858 | if (err) | |
859 | return err; | |
6681712d | 860 | |
5933a7bb MR |
861 | if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family) |
862 | return -EAFNOSUPPORT; | |
863 | ||
d342894c | 864 | spin_lock_bh(&vxlan->hash_lock); |
e4c7ed41 | 865 | err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags, |
73cf3317 | 866 | port, vni, ifindex, ndm->ndm_flags); |
d342894c | 867 | spin_unlock_bh(&vxlan->hash_lock); |
868 | ||
869 | return err; | |
870 | } | |
871 | ||
872 | /* Delete entry (via netlink) */ | |
1690be63 VY |
873 | static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], |
874 | struct net_device *dev, | |
f6f6424b | 875 | const unsigned char *addr, u16 vid) |
d342894c | 876 | { |
877 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
878 | struct vxlan_fdb *f; | |
bc7892ba | 879 | struct vxlan_rdst *rd = NULL; |
e4c7ed41 | 880 | union vxlan_addr ip; |
bc7892ba | 881 | __be16 port; |
54bfd872 JB |
882 | __be32 vni; |
883 | u32 ifindex; | |
bc7892ba MR |
884 | int err; |
885 | ||
886 | err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &vni, &ifindex); | |
887 | if (err) | |
888 | return err; | |
889 | ||
890 | err = -ENOENT; | |
d342894c | 891 | |
892 | spin_lock_bh(&vxlan->hash_lock); | |
893 | f = vxlan_find_mac(vxlan, addr); | |
bc7892ba MR |
894 | if (!f) |
895 | goto out; | |
896 | ||
e4c7ed41 CW |
897 | if (!vxlan_addr_any(&ip)) { |
898 | rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex); | |
bc7892ba MR |
899 | if (!rd) |
900 | goto out; | |
901 | } | |
902 | ||
903 | err = 0; | |
904 | ||
905 | /* remove a destination if it's not the only one on the list, | |
906 | * otherwise destroy the fdb entry | |
907 | */ | |
908 | if (rd && !list_is_singular(&f->remotes)) { | |
909 | list_del_rcu(&rd->list); | |
9e4b93f9 | 910 | vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH); |
dcdc7a59 | 911 | kfree_rcu(rd, rcu); |
bc7892ba | 912 | goto out; |
d342894c | 913 | } |
bc7892ba MR |
914 | |
915 | vxlan_fdb_destroy(vxlan, f); | |
916 | ||
917 | out: | |
d342894c | 918 | spin_unlock_bh(&vxlan->hash_lock); |
919 | ||
920 | return err; | |
921 | } | |
922 | ||
923 | /* Dump forwarding table */ | |
924 | static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, | |
5d5eacb3 JHS |
925 | struct net_device *dev, |
926 | struct net_device *filter_dev, int idx) | |
d342894c | 927 | { |
928 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
929 | unsigned int h; | |
930 | ||
931 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
932 | struct vxlan_fdb *f; | |
d342894c | 933 | int err; |
934 | ||
b67bfe0d | 935 | hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { |
6681712d | 936 | struct vxlan_rdst *rd; |
6681712d | 937 | |
3e61aa8f | 938 | list_for_each_entry_rcu(rd, &f->remotes, list) { |
07a51cd3 AW |
939 | if (idx < cb->args[0]) |
940 | goto skip; | |
941 | ||
6681712d DS |
942 | err = vxlan_fdb_info(skb, vxlan, f, |
943 | NETLINK_CB(cb->skb).portid, | |
944 | cb->nlh->nlmsg_seq, | |
945 | RTM_NEWNEIGH, | |
946 | NLM_F_MULTI, rd); | |
947 | if (err < 0) | |
3e61aa8f | 948 | goto out; |
3e61aa8f | 949 | skip: |
07a51cd3 AW |
950 | ++idx; |
951 | } | |
d342894c | 952 | } |
953 | } | |
3e61aa8f | 954 | out: |
d342894c | 955 | return idx; |
956 | } | |
957 | ||
958 | /* Watch incoming packets to learn mapping between Ethernet address | |
959 | * and Tunnel endpoint. | |
c4b49512 | 960 | * Return true if packet is bogus and should be dropped. |
d342894c | 961 | */ |
26a41ae6 | 962 | static bool vxlan_snoop(struct net_device *dev, |
e4c7ed41 | 963 | union vxlan_addr *src_ip, const u8 *src_mac) |
d342894c | 964 | { |
965 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
966 | struct vxlan_fdb *f; | |
d342894c | 967 | |
968 | f = vxlan_find_mac(vxlan, src_mac); | |
969 | if (likely(f)) { | |
5ca5461c | 970 | struct vxlan_rdst *rdst = first_remote_rcu(f); |
3e61aa8f | 971 | |
e4c7ed41 | 972 | if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip))) |
26a41ae6 | 973 | return false; |
974 | ||
975 | /* Don't migrate static entries, drop packets */ | |
eb064c3b | 976 | if (f->state & NUD_NOARP) |
26a41ae6 | 977 | return true; |
d342894c | 978 | |
979 | if (net_ratelimit()) | |
980 | netdev_info(dev, | |
e4c7ed41 | 981 | "%pM migrated from %pIS to %pIS\n", |
a4870f79 | 982 | src_mac, &rdst->remote_ip.sa, &src_ip->sa); |
d342894c | 983 | |
e4c7ed41 | 984 | rdst->remote_ip = *src_ip; |
d342894c | 985 | f->updated = jiffies; |
9e4b93f9 | 986 | vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH); |
d342894c | 987 | } else { |
988 | /* learned new entry */ | |
989 | spin_lock(&vxlan->hash_lock); | |
3bf74b1a | 990 | |
991 | /* close off race between vxlan_flush and incoming packets */ | |
992 | if (netif_running(dev)) | |
993 | vxlan_fdb_create(vxlan, src_mac, src_ip, | |
994 | NUD_REACHABLE, | |
995 | NLM_F_EXCL|NLM_F_CREATE, | |
0dfbdf41 | 996 | vxlan->cfg.dst_port, |
3bf74b1a | 997 | vxlan->default_dst.remote_vni, |
998 | 0, NTF_SELF); | |
d342894c | 999 | spin_unlock(&vxlan->hash_lock); |
1000 | } | |
26a41ae6 | 1001 | |
1002 | return false; | |
d342894c | 1003 | } |
1004 | ||
d342894c | 1005 | /* See if multicast group is already in use by other ID */ |
95ab0991 | 1006 | static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev) |
d342894c | 1007 | { |
553675fb | 1008 | struct vxlan_dev *vxlan; |
b1be00a6 | 1009 | unsigned short family = dev->default_dst.remote_ip.sa.sa_family; |
d342894c | 1010 | |
95ab0991 G |
1011 | /* The vxlan_sock is only used by dev, leaving group has |
1012 | * no effect on other vxlan devices. | |
1013 | */ | |
b1be00a6 JB |
1014 | if (family == AF_INET && dev->vn4_sock && |
1015 | atomic_read(&dev->vn4_sock->refcnt) == 1) | |
95ab0991 | 1016 | return false; |
b1be00a6 JB |
1017 | #if IS_ENABLED(CONFIG_IPV6) |
1018 | if (family == AF_INET6 && dev->vn6_sock && | |
1019 | atomic_read(&dev->vn6_sock->refcnt) == 1) | |
1020 | return false; | |
1021 | #endif | |
95ab0991 | 1022 | |
553675fb | 1023 | list_for_each_entry(vxlan, &vn->vxlan_list, next) { |
95ab0991 | 1024 | if (!netif_running(vxlan->dev) || vxlan == dev) |
553675fb | 1025 | continue; |
d342894c | 1026 | |
b1be00a6 | 1027 | if (family == AF_INET && vxlan->vn4_sock != dev->vn4_sock) |
95ab0991 | 1028 | continue; |
b1be00a6 JB |
1029 | #if IS_ENABLED(CONFIG_IPV6) |
1030 | if (family == AF_INET6 && vxlan->vn6_sock != dev->vn6_sock) | |
1031 | continue; | |
1032 | #endif | |
95ab0991 G |
1033 | |
1034 | if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip, | |
1035 | &dev->default_dst.remote_ip)) | |
1036 | continue; | |
1037 | ||
1038 | if (vxlan->default_dst.remote_ifindex != | |
1039 | dev->default_dst.remote_ifindex) | |
1040 | continue; | |
1041 | ||
1042 | return true; | |
553675fb | 1043 | } |
d342894c | 1044 | |
1045 | return false; | |
1046 | } | |
1047 | ||
b1be00a6 | 1048 | static void __vxlan_sock_release(struct vxlan_sock *vs) |
7c47cedf | 1049 | { |
b1be00a6 | 1050 | struct vxlan_net *vn; |
012a5729 | 1051 | |
b1be00a6 JB |
1052 | if (!vs) |
1053 | return; | |
7c47cedf SH |
1054 | if (!atomic_dec_and_test(&vs->refcnt)) |
1055 | return; | |
d342894c | 1056 | |
b1be00a6 | 1057 | vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id); |
1c51a915 | 1058 | spin_lock(&vn->sock_lock); |
7c47cedf | 1059 | hlist_del_rcu(&vs->hlist); |
dc01e7d3 | 1060 | vxlan_notify_del_rx_port(vs); |
1c51a915 SH |
1061 | spin_unlock(&vn->sock_lock); |
1062 | ||
7c47cedf | 1063 | queue_work(vxlan_wq, &vs->del_work); |
d342894c | 1064 | } |
1065 | ||
b1be00a6 JB |
1066 | static void vxlan_sock_release(struct vxlan_dev *vxlan) |
1067 | { | |
1068 | __vxlan_sock_release(vxlan->vn4_sock); | |
1069 | #if IS_ENABLED(CONFIG_IPV6) | |
1070 | __vxlan_sock_release(vxlan->vn6_sock); | |
1071 | #endif | |
1072 | } | |
1073 | ||
56ef9c90 | 1074 | /* Update multicast group membership when first VNI on |
c4b49512 | 1075 | * multicast address is brought up |
7c47cedf | 1076 | */ |
56ef9c90 | 1077 | static int vxlan_igmp_join(struct vxlan_dev *vxlan) |
d342894c | 1078 | { |
b1be00a6 | 1079 | struct sock *sk; |
e4c7ed41 CW |
1080 | union vxlan_addr *ip = &vxlan->default_dst.remote_ip; |
1081 | int ifindex = vxlan->default_dst.remote_ifindex; | |
149d7549 | 1082 | int ret = -EINVAL; |
d342894c | 1083 | |
e4c7ed41 CW |
1084 | if (ip->sa.sa_family == AF_INET) { |
1085 | struct ip_mreqn mreq = { | |
1086 | .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr, | |
1087 | .imr_ifindex = ifindex, | |
1088 | }; | |
1089 | ||
b1be00a6 JB |
1090 | sk = vxlan->vn4_sock->sock->sk; |
1091 | lock_sock(sk); | |
56ef9c90 | 1092 | ret = ip_mc_join_group(sk, &mreq); |
b1be00a6 | 1093 | release_sock(sk); |
e4c7ed41 CW |
1094 | #if IS_ENABLED(CONFIG_IPV6) |
1095 | } else { | |
b1be00a6 JB |
1096 | sk = vxlan->vn6_sock->sock->sk; |
1097 | lock_sock(sk); | |
56ef9c90 MRL |
1098 | ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex, |
1099 | &ip->sin6.sin6_addr); | |
b1be00a6 | 1100 | release_sock(sk); |
e4c7ed41 CW |
1101 | #endif |
1102 | } | |
3fc2de2f | 1103 | |
56ef9c90 | 1104 | return ret; |
3fc2de2f | 1105 | } |
1106 | ||
1107 | /* Inverse of vxlan_igmp_join when last VNI is brought down */ | |
56ef9c90 | 1108 | static int vxlan_igmp_leave(struct vxlan_dev *vxlan) |
3fc2de2f | 1109 | { |
b1be00a6 | 1110 | struct sock *sk; |
e4c7ed41 CW |
1111 | union vxlan_addr *ip = &vxlan->default_dst.remote_ip; |
1112 | int ifindex = vxlan->default_dst.remote_ifindex; | |
149d7549 | 1113 | int ret = -EINVAL; |
3fc2de2f | 1114 | |
e4c7ed41 CW |
1115 | if (ip->sa.sa_family == AF_INET) { |
1116 | struct ip_mreqn mreq = { | |
1117 | .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr, | |
1118 | .imr_ifindex = ifindex, | |
1119 | }; | |
1120 | ||
b1be00a6 JB |
1121 | sk = vxlan->vn4_sock->sock->sk; |
1122 | lock_sock(sk); | |
56ef9c90 | 1123 | ret = ip_mc_leave_group(sk, &mreq); |
b1be00a6 | 1124 | release_sock(sk); |
e4c7ed41 CW |
1125 | #if IS_ENABLED(CONFIG_IPV6) |
1126 | } else { | |
b1be00a6 JB |
1127 | sk = vxlan->vn6_sock->sock->sk; |
1128 | lock_sock(sk); | |
56ef9c90 MRL |
1129 | ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex, |
1130 | &ip->sin6.sin6_addr); | |
b1be00a6 | 1131 | release_sock(sk); |
e4c7ed41 CW |
1132 | #endif |
1133 | } | |
d342894c | 1134 | |
56ef9c90 | 1135 | return ret; |
d342894c | 1136 | } |
1137 | ||
dfd8645e | 1138 | static struct vxlanhdr *vxlan_remcsum(struct sk_buff *skb, struct vxlanhdr *vh, |
54bfd872 JB |
1139 | size_t hdrlen, __be32 vni_field, |
1140 | bool nopartial) | |
dfd8645e TH |
1141 | { |
1142 | size_t start, offset, plen; | |
dfd8645e | 1143 | |
b7fe10e5 TH |
1144 | if (skb->remcsum_offload) |
1145 | return vh; | |
1146 | ||
54bfd872 JB |
1147 | start = vxlan_rco_start(vni_field); |
1148 | offset = start + vxlan_rco_offset(vni_field); | |
dfd8645e TH |
1149 | |
1150 | plen = hdrlen + offset + sizeof(u16); | |
1151 | ||
1152 | if (!pskb_may_pull(skb, plen)) | |
1153 | return NULL; | |
1154 | ||
1155 | vh = (struct vxlanhdr *)(udp_hdr(skb) + 1); | |
1156 | ||
0ace2ca8 TH |
1157 | skb_remcsum_process(skb, (void *)vh + hdrlen, start, offset, |
1158 | nopartial); | |
dfd8645e TH |
1159 | |
1160 | return vh; | |
1161 | } | |
1162 | ||
614732ea | 1163 | static void vxlan_rcv(struct vxlan_sock *vs, struct sk_buff *skb, |
54bfd872 | 1164 | struct vxlan_metadata *md, __be32 vni, |
614732ea TG |
1165 | struct metadata_dst *tun_dst) |
1166 | { | |
1167 | struct iphdr *oip = NULL; | |
1168 | struct ipv6hdr *oip6 = NULL; | |
1169 | struct vxlan_dev *vxlan; | |
1170 | struct pcpu_sw_netstats *stats; | |
1171 | union vxlan_addr saddr; | |
1172 | int err = 0; | |
614732ea TG |
1173 | |
1174 | /* For flow based devices, map all packets to VNI 0 */ | |
da8b43c0 | 1175 | if (vs->flags & VXLAN_F_COLLECT_METADATA) |
614732ea TG |
1176 | vni = 0; |
1177 | ||
1178 | /* Is this VNI defined? */ | |
1179 | vxlan = vxlan_vs_find_vni(vs, vni); | |
1180 | if (!vxlan) | |
1181 | goto drop; | |
1182 | ||
614732ea TG |
1183 | skb_reset_mac_header(skb); |
1184 | skb_scrub_packet(skb, !net_eq(vxlan->net, dev_net(vxlan->dev))); | |
1185 | skb->protocol = eth_type_trans(skb, vxlan->dev); | |
1186 | skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN); | |
1187 | ||
1188 | /* Ignore packet loops (and multicast echo) */ | |
1189 | if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr)) | |
1190 | goto drop; | |
1191 | ||
ce212d0f JB |
1192 | /* Get data from the outer IP header */ |
1193 | if (vxlan_get_sk_family(vs) == AF_INET) { | |
614732ea TG |
1194 | oip = ip_hdr(skb); |
1195 | saddr.sin.sin_addr.s_addr = oip->saddr; | |
1196 | saddr.sa.sa_family = AF_INET; | |
1197 | #if IS_ENABLED(CONFIG_IPV6) | |
1198 | } else { | |
1199 | oip6 = ipv6_hdr(skb); | |
1200 | saddr.sin6.sin6_addr = oip6->saddr; | |
1201 | saddr.sa.sa_family = AF_INET6; | |
1202 | #endif | |
1203 | } | |
1204 | ||
1205 | if (tun_dst) { | |
1206 | skb_dst_set(skb, (struct dst_entry *)tun_dst); | |
1207 | tun_dst = NULL; | |
1208 | } | |
1209 | ||
1210 | if ((vxlan->flags & VXLAN_F_LEARN) && | |
1211 | vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source)) | |
1212 | goto drop; | |
1213 | ||
1214 | skb_reset_network_header(skb); | |
1215 | /* In flow-based mode, GBP is carried in dst_metadata */ | |
da8b43c0 | 1216 | if (!(vs->flags & VXLAN_F_COLLECT_METADATA)) |
614732ea TG |
1217 | skb->mark = md->gbp; |
1218 | ||
1219 | if (oip6) | |
1220 | err = IP6_ECN_decapsulate(oip6, skb); | |
1221 | if (oip) | |
1222 | err = IP_ECN_decapsulate(oip, skb); | |
1223 | ||
1224 | if (unlikely(err)) { | |
1225 | if (log_ecn_error) { | |
1226 | if (oip6) | |
1227 | net_info_ratelimited("non-ECT from %pI6\n", | |
1228 | &oip6->saddr); | |
1229 | if (oip) | |
1230 | net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", | |
1231 | &oip->saddr, oip->tos); | |
1232 | } | |
1233 | if (err > 1) { | |
1234 | ++vxlan->dev->stats.rx_frame_errors; | |
1235 | ++vxlan->dev->stats.rx_errors; | |
1236 | goto drop; | |
1237 | } | |
1238 | } | |
1239 | ||
1240 | stats = this_cpu_ptr(vxlan->dev->tstats); | |
1241 | u64_stats_update_begin(&stats->syncp); | |
1242 | stats->rx_packets++; | |
1243 | stats->rx_bytes += skb->len; | |
1244 | u64_stats_update_end(&stats->syncp); | |
1245 | ||
58ce31cc | 1246 | gro_cells_receive(&vxlan->gro_cells, skb); |
614732ea TG |
1247 | |
1248 | return; | |
1249 | drop: | |
1250 | if (tun_dst) | |
1251 | dst_release((struct dst_entry *)tun_dst); | |
1252 | ||
1253 | /* Consume bad packet */ | |
1254 | kfree_skb(skb); | |
1255 | } | |
1256 | ||
d342894c | 1257 | /* Callback from net/ipv4/udp.c to receive packets */ |
1258 | static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb) | |
1259 | { | |
ee122c79 | 1260 | struct metadata_dst *tun_dst = NULL; |
5cfccc5a | 1261 | struct vxlan_sock *vs; |
54bfd872 | 1262 | __be32 flags, vni_field; |
ee122c79 TG |
1263 | struct vxlan_metadata _md; |
1264 | struct vxlan_metadata *md = &_md; | |
d342894c | 1265 | |
d342894c | 1266 | /* Need Vxlan and inner Ethernet header to be present */ |
7ce04758 | 1267 | if (!pskb_may_pull(skb, VXLAN_HLEN)) |
d342894c | 1268 | goto error; |
1269 | ||
54bfd872 JB |
1270 | flags = vxlan_hdr(skb)->vx_flags; |
1271 | vni_field = vxlan_hdr(skb)->vx_vni; | |
3bf39475 TH |
1272 | |
1273 | if (flags & VXLAN_HF_VNI) { | |
1274 | flags &= ~VXLAN_HF_VNI; | |
1275 | } else { | |
1276 | /* VNI flag always required to be set */ | |
1277 | goto bad_flags; | |
d342894c | 1278 | } |
1279 | ||
5cfccc5a PS |
1280 | if (iptunnel_pull_header(skb, VXLAN_HLEN, htons(ETH_P_TEB))) |
1281 | goto drop; | |
1282 | ||
559835ea | 1283 | vs = rcu_dereference_sk_user_data(sk); |
5cfccc5a | 1284 | if (!vs) |
d342894c | 1285 | goto drop; |
d342894c | 1286 | |
dfd8645e | 1287 | if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) { |
54bfd872 JB |
1288 | if (!vxlan_remcsum(skb, vxlan_hdr(skb), sizeof(struct vxlanhdr), |
1289 | vni_field, | |
d4ac05ff | 1290 | !!(vs->flags & VXLAN_F_REMCSUM_NOPARTIAL))) |
dfd8645e TH |
1291 | goto drop; |
1292 | ||
1293 | flags &= ~VXLAN_HF_RCO; | |
54bfd872 | 1294 | vni_field &= VXLAN_VNI_MASK; |
dfd8645e TH |
1295 | } |
1296 | ||
ee122c79 | 1297 | if (vxlan_collect_metadata(vs)) { |
c29a70d2 | 1298 | tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY, |
54bfd872 | 1299 | vxlan_vni(vni_field), sizeof(*md)); |
c29a70d2 | 1300 | |
ee122c79 TG |
1301 | if (!tun_dst) |
1302 | goto drop; | |
1303 | ||
0f1b7354 | 1304 | md = ip_tunnel_info_opts(&tun_dst->u.tun_info); |
ee122c79 TG |
1305 | } else { |
1306 | memset(md, 0, sizeof(*md)); | |
1307 | } | |
1308 | ||
3511494c TG |
1309 | /* For backwards compatibility, only allow reserved fields to be |
1310 | * used by VXLAN extensions if explicitly requested. | |
1311 | */ | |
1312 | if ((flags & VXLAN_HF_GBP) && (vs->flags & VXLAN_F_GBP)) { | |
1313 | struct vxlanhdr_gbp *gbp; | |
1314 | ||
d4ac05ff | 1315 | gbp = (struct vxlanhdr_gbp *)vxlan_hdr(skb); |
ee122c79 TG |
1316 | md->gbp = ntohs(gbp->policy_id); |
1317 | ||
1318 | if (tun_dst) | |
0f1b7354 | 1319 | tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT; |
3511494c TG |
1320 | |
1321 | if (gbp->dont_learn) | |
ee122c79 | 1322 | md->gbp |= VXLAN_GBP_DONT_LEARN; |
3511494c TG |
1323 | |
1324 | if (gbp->policy_applied) | |
ee122c79 | 1325 | md->gbp |= VXLAN_GBP_POLICY_APPLIED; |
3511494c TG |
1326 | |
1327 | flags &= ~VXLAN_GBP_USED_BITS; | |
1328 | } | |
1329 | ||
54bfd872 | 1330 | if (flags || vni_field & ~VXLAN_VNI_MASK) { |
3bf39475 TH |
1331 | /* If there are any unprocessed flags remaining treat |
1332 | * this as a malformed packet. This behavior diverges from | |
1333 | * VXLAN RFC (RFC7348) which stipulates that bits in reserved | |
1334 | * in reserved fields are to be ignored. The approach here | |
c4b49512 | 1335 | * maintains compatibility with previous stack code, and also |
3bf39475 TH |
1336 | * is more robust and provides a little more security in |
1337 | * adding extensions to VXLAN. | |
1338 | */ | |
1339 | ||
1340 | goto bad_flags; | |
1341 | } | |
1342 | ||
54bfd872 | 1343 | vxlan_rcv(vs, skb, md, vxlan_vni(vni_field), tun_dst); |
5cfccc5a PS |
1344 | return 0; |
1345 | ||
1346 | drop: | |
1347 | /* Consume bad packet */ | |
1348 | kfree_skb(skb); | |
1349 | return 0; | |
1350 | ||
3bf39475 TH |
1351 | bad_flags: |
1352 | netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", | |
d4ac05ff JB |
1353 | ntohl(vxlan_hdr(skb)->vx_flags), |
1354 | ntohl(vxlan_hdr(skb)->vx_vni)); | |
3bf39475 | 1355 | |
5cfccc5a | 1356 | error: |
ee122c79 TG |
1357 | if (tun_dst) |
1358 | dst_release((struct dst_entry *)tun_dst); | |
1359 | ||
5cfccc5a PS |
1360 | /* Return non vxlan pkt */ |
1361 | return 1; | |
1362 | } | |
1363 | ||
e4f67add DS |
1364 | static int arp_reduce(struct net_device *dev, struct sk_buff *skb) |
1365 | { | |
1366 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1367 | struct arphdr *parp; | |
1368 | u8 *arpptr, *sha; | |
1369 | __be32 sip, tip; | |
1370 | struct neighbour *n; | |
1371 | ||
1372 | if (dev->flags & IFF_NOARP) | |
1373 | goto out; | |
1374 | ||
1375 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) { | |
1376 | dev->stats.tx_dropped++; | |
1377 | goto out; | |
1378 | } | |
1379 | parp = arp_hdr(skb); | |
1380 | ||
1381 | if ((parp->ar_hrd != htons(ARPHRD_ETHER) && | |
1382 | parp->ar_hrd != htons(ARPHRD_IEEE802)) || | |
1383 | parp->ar_pro != htons(ETH_P_IP) || | |
1384 | parp->ar_op != htons(ARPOP_REQUEST) || | |
1385 | parp->ar_hln != dev->addr_len || | |
1386 | parp->ar_pln != 4) | |
1387 | goto out; | |
1388 | arpptr = (u8 *)parp + sizeof(struct arphdr); | |
1389 | sha = arpptr; | |
1390 | arpptr += dev->addr_len; /* sha */ | |
1391 | memcpy(&sip, arpptr, sizeof(sip)); | |
1392 | arpptr += sizeof(sip); | |
1393 | arpptr += dev->addr_len; /* tha */ | |
1394 | memcpy(&tip, arpptr, sizeof(tip)); | |
1395 | ||
1396 | if (ipv4_is_loopback(tip) || | |
1397 | ipv4_is_multicast(tip)) | |
1398 | goto out; | |
1399 | ||
1400 | n = neigh_lookup(&arp_tbl, &tip, dev); | |
1401 | ||
1402 | if (n) { | |
e4f67add DS |
1403 | struct vxlan_fdb *f; |
1404 | struct sk_buff *reply; | |
1405 | ||
1406 | if (!(n->nud_state & NUD_CONNECTED)) { | |
1407 | neigh_release(n); | |
1408 | goto out; | |
1409 | } | |
1410 | ||
1411 | f = vxlan_find_mac(vxlan, n->ha); | |
e4c7ed41 | 1412 | if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { |
e4f67add DS |
1413 | /* bridge-local neighbor */ |
1414 | neigh_release(n); | |
1415 | goto out; | |
1416 | } | |
1417 | ||
1418 | reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, | |
1419 | n->ha, sha); | |
1420 | ||
1421 | neigh_release(n); | |
1422 | ||
7346135d DS |
1423 | if (reply == NULL) |
1424 | goto out; | |
1425 | ||
e4f67add DS |
1426 | skb_reset_mac_header(reply); |
1427 | __skb_pull(reply, skb_network_offset(reply)); | |
1428 | reply->ip_summed = CHECKSUM_UNNECESSARY; | |
1429 | reply->pkt_type = PACKET_HOST; | |
1430 | ||
1431 | if (netif_rx_ni(reply) == NET_RX_DROP) | |
1432 | dev->stats.rx_dropped++; | |
e4c7ed41 CW |
1433 | } else if (vxlan->flags & VXLAN_F_L3MISS) { |
1434 | union vxlan_addr ipa = { | |
1435 | .sin.sin_addr.s_addr = tip, | |
a45e92a5 | 1436 | .sin.sin_family = AF_INET, |
e4c7ed41 CW |
1437 | }; |
1438 | ||
1439 | vxlan_ip_miss(dev, &ipa); | |
1440 | } | |
e4f67add DS |
1441 | out: |
1442 | consume_skb(skb); | |
1443 | return NETDEV_TX_OK; | |
1444 | } | |
1445 | ||
f564f45c | 1446 | #if IS_ENABLED(CONFIG_IPV6) |
4b29dba9 DS |
1447 | static struct sk_buff *vxlan_na_create(struct sk_buff *request, |
1448 | struct neighbour *n, bool isrouter) | |
1449 | { | |
1450 | struct net_device *dev = request->dev; | |
1451 | struct sk_buff *reply; | |
1452 | struct nd_msg *ns, *na; | |
1453 | struct ipv6hdr *pip6; | |
1454 | u8 *daddr; | |
1455 | int na_olen = 8; /* opt hdr + ETH_ALEN for target */ | |
1456 | int ns_olen; | |
1457 | int i, len; | |
1458 | ||
1459 | if (dev == NULL) | |
1460 | return NULL; | |
1461 | ||
1462 | len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) + | |
1463 | sizeof(*na) + na_olen + dev->needed_tailroom; | |
1464 | reply = alloc_skb(len, GFP_ATOMIC); | |
1465 | if (reply == NULL) | |
1466 | return NULL; | |
1467 | ||
1468 | reply->protocol = htons(ETH_P_IPV6); | |
1469 | reply->dev = dev; | |
1470 | skb_reserve(reply, LL_RESERVED_SPACE(request->dev)); | |
1471 | skb_push(reply, sizeof(struct ethhdr)); | |
1472 | skb_set_mac_header(reply, 0); | |
1473 | ||
1474 | ns = (struct nd_msg *)skb_transport_header(request); | |
1475 | ||
1476 | daddr = eth_hdr(request)->h_source; | |
1477 | ns_olen = request->len - skb_transport_offset(request) - sizeof(*ns); | |
1478 | for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) { | |
1479 | if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) { | |
1480 | daddr = ns->opt + i + sizeof(struct nd_opt_hdr); | |
1481 | break; | |
1482 | } | |
1483 | } | |
1484 | ||
1485 | /* Ethernet header */ | |
1486 | ether_addr_copy(eth_hdr(reply)->h_dest, daddr); | |
1487 | ether_addr_copy(eth_hdr(reply)->h_source, n->ha); | |
1488 | eth_hdr(reply)->h_proto = htons(ETH_P_IPV6); | |
1489 | reply->protocol = htons(ETH_P_IPV6); | |
1490 | ||
1491 | skb_pull(reply, sizeof(struct ethhdr)); | |
1492 | skb_set_network_header(reply, 0); | |
1493 | skb_put(reply, sizeof(struct ipv6hdr)); | |
1494 | ||
1495 | /* IPv6 header */ | |
1496 | ||
1497 | pip6 = ipv6_hdr(reply); | |
1498 | memset(pip6, 0, sizeof(struct ipv6hdr)); | |
1499 | pip6->version = 6; | |
1500 | pip6->priority = ipv6_hdr(request)->priority; | |
1501 | pip6->nexthdr = IPPROTO_ICMPV6; | |
1502 | pip6->hop_limit = 255; | |
1503 | pip6->daddr = ipv6_hdr(request)->saddr; | |
1504 | pip6->saddr = *(struct in6_addr *)n->primary_key; | |
1505 | ||
1506 | skb_pull(reply, sizeof(struct ipv6hdr)); | |
1507 | skb_set_transport_header(reply, 0); | |
1508 | ||
1509 | na = (struct nd_msg *)skb_put(reply, sizeof(*na) + na_olen); | |
1510 | ||
1511 | /* Neighbor Advertisement */ | |
1512 | memset(na, 0, sizeof(*na)+na_olen); | |
1513 | na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT; | |
1514 | na->icmph.icmp6_router = isrouter; | |
1515 | na->icmph.icmp6_override = 1; | |
1516 | na->icmph.icmp6_solicited = 1; | |
1517 | na->target = ns->target; | |
1518 | ether_addr_copy(&na->opt[2], n->ha); | |
1519 | na->opt[0] = ND_OPT_TARGET_LL_ADDR; | |
1520 | na->opt[1] = na_olen >> 3; | |
1521 | ||
1522 | na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr, | |
1523 | &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6, | |
1524 | csum_partial(na, sizeof(*na)+na_olen, 0)); | |
1525 | ||
1526 | pip6->payload_len = htons(sizeof(*na)+na_olen); | |
1527 | ||
1528 | skb_push(reply, sizeof(struct ipv6hdr)); | |
1529 | ||
1530 | reply->ip_summed = CHECKSUM_UNNECESSARY; | |
1531 | ||
1532 | return reply; | |
1533 | } | |
1534 | ||
f564f45c CW |
1535 | static int neigh_reduce(struct net_device *dev, struct sk_buff *skb) |
1536 | { | |
1537 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
4b29dba9 | 1538 | struct nd_msg *msg; |
f564f45c CW |
1539 | const struct ipv6hdr *iphdr; |
1540 | const struct in6_addr *saddr, *daddr; | |
4b29dba9 DS |
1541 | struct neighbour *n; |
1542 | struct inet6_dev *in6_dev; | |
f564f45c CW |
1543 | |
1544 | in6_dev = __in6_dev_get(dev); | |
1545 | if (!in6_dev) | |
1546 | goto out; | |
1547 | ||
f564f45c CW |
1548 | iphdr = ipv6_hdr(skb); |
1549 | saddr = &iphdr->saddr; | |
1550 | daddr = &iphdr->daddr; | |
1551 | ||
f564f45c CW |
1552 | msg = (struct nd_msg *)skb_transport_header(skb); |
1553 | if (msg->icmph.icmp6_code != 0 || | |
1554 | msg->icmph.icmp6_type != NDISC_NEIGHBOUR_SOLICITATION) | |
1555 | goto out; | |
1556 | ||
4b29dba9 DS |
1557 | if (ipv6_addr_loopback(daddr) || |
1558 | ipv6_addr_is_multicast(&msg->target)) | |
1559 | goto out; | |
1560 | ||
1561 | n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev); | |
f564f45c CW |
1562 | |
1563 | if (n) { | |
1564 | struct vxlan_fdb *f; | |
4b29dba9 | 1565 | struct sk_buff *reply; |
f564f45c CW |
1566 | |
1567 | if (!(n->nud_state & NUD_CONNECTED)) { | |
1568 | neigh_release(n); | |
1569 | goto out; | |
1570 | } | |
1571 | ||
1572 | f = vxlan_find_mac(vxlan, n->ha); | |
1573 | if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) { | |
1574 | /* bridge-local neighbor */ | |
1575 | neigh_release(n); | |
1576 | goto out; | |
1577 | } | |
1578 | ||
4b29dba9 DS |
1579 | reply = vxlan_na_create(skb, n, |
1580 | !!(f ? f->flags & NTF_ROUTER : 0)); | |
1581 | ||
f564f45c | 1582 | neigh_release(n); |
4b29dba9 DS |
1583 | |
1584 | if (reply == NULL) | |
1585 | goto out; | |
1586 | ||
1587 | if (netif_rx_ni(reply) == NET_RX_DROP) | |
1588 | dev->stats.rx_dropped++; | |
1589 | ||
f564f45c | 1590 | } else if (vxlan->flags & VXLAN_F_L3MISS) { |
4b29dba9 DS |
1591 | union vxlan_addr ipa = { |
1592 | .sin6.sin6_addr = msg->target, | |
a45e92a5 | 1593 | .sin6.sin6_family = AF_INET6, |
4b29dba9 DS |
1594 | }; |
1595 | ||
f564f45c CW |
1596 | vxlan_ip_miss(dev, &ipa); |
1597 | } | |
1598 | ||
1599 | out: | |
1600 | consume_skb(skb); | |
1601 | return NETDEV_TX_OK; | |
1602 | } | |
1603 | #endif | |
1604 | ||
e4f67add DS |
1605 | static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) |
1606 | { | |
1607 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1608 | struct neighbour *n; | |
e4f67add DS |
1609 | |
1610 | if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) | |
1611 | return false; | |
1612 | ||
1613 | n = NULL; | |
1614 | switch (ntohs(eth_hdr(skb)->h_proto)) { | |
1615 | case ETH_P_IP: | |
e15a00aa CW |
1616 | { |
1617 | struct iphdr *pip; | |
1618 | ||
e4f67add DS |
1619 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) |
1620 | return false; | |
1621 | pip = ip_hdr(skb); | |
1622 | n = neigh_lookup(&arp_tbl, &pip->daddr, dev); | |
e4c7ed41 CW |
1623 | if (!n && (vxlan->flags & VXLAN_F_L3MISS)) { |
1624 | union vxlan_addr ipa = { | |
1625 | .sin.sin_addr.s_addr = pip->daddr, | |
a45e92a5 | 1626 | .sin.sin_family = AF_INET, |
e4c7ed41 CW |
1627 | }; |
1628 | ||
1629 | vxlan_ip_miss(dev, &ipa); | |
1630 | return false; | |
1631 | } | |
1632 | ||
e4f67add | 1633 | break; |
e15a00aa CW |
1634 | } |
1635 | #if IS_ENABLED(CONFIG_IPV6) | |
1636 | case ETH_P_IPV6: | |
1637 | { | |
1638 | struct ipv6hdr *pip6; | |
1639 | ||
1640 | if (!pskb_may_pull(skb, sizeof(struct ipv6hdr))) | |
1641 | return false; | |
1642 | pip6 = ipv6_hdr(skb); | |
1643 | n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev); | |
1644 | if (!n && (vxlan->flags & VXLAN_F_L3MISS)) { | |
1645 | union vxlan_addr ipa = { | |
1646 | .sin6.sin6_addr = pip6->daddr, | |
a45e92a5 | 1647 | .sin6.sin6_family = AF_INET6, |
e15a00aa CW |
1648 | }; |
1649 | ||
1650 | vxlan_ip_miss(dev, &ipa); | |
1651 | return false; | |
1652 | } | |
1653 | ||
1654 | break; | |
1655 | } | |
1656 | #endif | |
e4f67add DS |
1657 | default: |
1658 | return false; | |
1659 | } | |
1660 | ||
1661 | if (n) { | |
1662 | bool diff; | |
1663 | ||
7367d0b5 | 1664 | diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha); |
e4f67add DS |
1665 | if (diff) { |
1666 | memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, | |
1667 | dev->addr_len); | |
1668 | memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); | |
1669 | } | |
1670 | neigh_release(n); | |
1671 | return diff; | |
e4c7ed41 CW |
1672 | } |
1673 | ||
e4f67add DS |
1674 | return false; |
1675 | } | |
1676 | ||
af33c1ad | 1677 | static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags, |
3511494c TG |
1678 | struct vxlan_metadata *md) |
1679 | { | |
1680 | struct vxlanhdr_gbp *gbp; | |
1681 | ||
db79a621 TG |
1682 | if (!md->gbp) |
1683 | return; | |
1684 | ||
3511494c | 1685 | gbp = (struct vxlanhdr_gbp *)vxh; |
54bfd872 | 1686 | vxh->vx_flags |= VXLAN_HF_GBP; |
3511494c TG |
1687 | |
1688 | if (md->gbp & VXLAN_GBP_DONT_LEARN) | |
1689 | gbp->dont_learn = 1; | |
1690 | ||
1691 | if (md->gbp & VXLAN_GBP_POLICY_APPLIED) | |
1692 | gbp->policy_applied = 1; | |
1693 | ||
1694 | gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK); | |
1695 | } | |
1696 | ||
f491e56d JB |
1697 | static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst, |
1698 | int iphdr_len, __be32 vni, | |
1699 | struct vxlan_metadata *md, u32 vxflags, | |
b4ed5cad | 1700 | bool udp_sum) |
e4c7ed41 | 1701 | { |
e4c7ed41 | 1702 | struct vxlanhdr *vxh; |
e4c7ed41 CW |
1703 | int min_headroom; |
1704 | int err; | |
dfd8645e | 1705 | int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL; |
dfd8645e | 1706 | |
af33c1ad | 1707 | if ((vxflags & VXLAN_F_REMCSUM_TX) && |
dfd8645e TH |
1708 | skb->ip_summed == CHECKSUM_PARTIAL) { |
1709 | int csum_start = skb_checksum_start_offset(skb); | |
1710 | ||
1711 | if (csum_start <= VXLAN_MAX_REMCSUM_START && | |
1712 | !(csum_start & VXLAN_RCO_SHIFT_MASK) && | |
1713 | (skb->csum_offset == offsetof(struct udphdr, check) || | |
b5708501 | 1714 | skb->csum_offset == offsetof(struct tcphdr, check))) |
dfd8645e | 1715 | type |= SKB_GSO_TUNNEL_REMCSUM; |
dfd8645e | 1716 | } |
e4c7ed41 | 1717 | |
e4c7ed41 | 1718 | min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len |
f491e56d | 1719 | + VXLAN_HLEN + iphdr_len |
df8a39de | 1720 | + (skb_vlan_tag_present(skb) ? VLAN_HLEN : 0); |
649c5b8b PS |
1721 | |
1722 | /* Need space for new headers (invalidates iph ptr) */ | |
1723 | err = skb_cow_head(skb, min_headroom); | |
74f47278 PS |
1724 | if (unlikely(err)) { |
1725 | kfree_skb(skb); | |
649c5b8b | 1726 | return err; |
74f47278 | 1727 | } |
649c5b8b | 1728 | |
5968250c JP |
1729 | skb = vlan_hwaccel_push_inside(skb); |
1730 | if (WARN_ON(!skb)) | |
1731 | return -ENOMEM; | |
1eaa8178 | 1732 | |
6fa79666 | 1733 | skb = iptunnel_handle_offloads(skb, type); |
b736a623 JG |
1734 | if (IS_ERR(skb)) |
1735 | return PTR_ERR(skb); | |
1736 | ||
49560532 | 1737 | vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh)); |
54bfd872 JB |
1738 | vxh->vx_flags = VXLAN_HF_VNI; |
1739 | vxh->vx_vni = vxlan_vni_field(vni); | |
49560532 | 1740 | |
dfd8645e | 1741 | if (type & SKB_GSO_TUNNEL_REMCSUM) { |
54bfd872 | 1742 | unsigned int start; |
dfd8645e | 1743 | |
54bfd872 JB |
1744 | start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr); |
1745 | vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset); | |
1746 | vxh->vx_flags |= VXLAN_HF_RCO; | |
dfd8645e TH |
1747 | |
1748 | if (!skb_is_gso(skb)) { | |
1749 | skb->ip_summed = CHECKSUM_NONE; | |
1750 | skb->encapsulation = 0; | |
1751 | } | |
1752 | } | |
1753 | ||
af33c1ad TH |
1754 | if (vxflags & VXLAN_F_GBP) |
1755 | vxlan_build_gbp_hdr(vxh, vxflags, md); | |
3511494c | 1756 | |
996c9fd1 | 1757 | skb_set_inner_protocol(skb, htons(ETH_P_TEB)); |
039f5062 | 1758 | return 0; |
49560532 | 1759 | } |
49560532 | 1760 | |
1a8496ba JB |
1761 | static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, |
1762 | struct sk_buff *skb, int oif, u8 tos, | |
0c1d70af PA |
1763 | __be32 daddr, __be32 *saddr, |
1764 | struct dst_cache *dst_cache, | |
1765 | struct ip_tunnel_info *info) | |
1a8496ba JB |
1766 | { |
1767 | struct rtable *rt = NULL; | |
0c1d70af | 1768 | bool use_cache = false; |
1a8496ba JB |
1769 | struct flowi4 fl4; |
1770 | ||
0c1d70af PA |
1771 | /* when the ip_tunnel_info is availble, the tos used for lookup is |
1772 | * packet independent, so we can use the cache | |
1773 | */ | |
d71785ff | 1774 | if (!skb->mark && (!tos || info)) { |
0c1d70af PA |
1775 | use_cache = true; |
1776 | rt = dst_cache_get_ip4(dst_cache, saddr); | |
1777 | if (rt) | |
1778 | return rt; | |
1779 | } | |
1780 | ||
1a8496ba JB |
1781 | memset(&fl4, 0, sizeof(fl4)); |
1782 | fl4.flowi4_oif = oif; | |
1783 | fl4.flowi4_tos = RT_TOS(tos); | |
1784 | fl4.flowi4_mark = skb->mark; | |
1785 | fl4.flowi4_proto = IPPROTO_UDP; | |
1786 | fl4.daddr = daddr; | |
1787 | fl4.saddr = vxlan->cfg.saddr.sin.sin_addr.s_addr; | |
1788 | ||
1789 | rt = ip_route_output_key(vxlan->net, &fl4); | |
0c1d70af | 1790 | if (!IS_ERR(rt)) { |
1a8496ba | 1791 | *saddr = fl4.saddr; |
0c1d70af PA |
1792 | if (use_cache) |
1793 | dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr); | |
1794 | } | |
1a8496ba JB |
1795 | return rt; |
1796 | } | |
1797 | ||
e5d4b29f JB |
1798 | #if IS_ENABLED(CONFIG_IPV6) |
1799 | static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan, | |
1800 | struct sk_buff *skb, int oif, | |
1801 | const struct in6_addr *daddr, | |
0c1d70af PA |
1802 | struct in6_addr *saddr, |
1803 | struct dst_cache *dst_cache) | |
e5d4b29f JB |
1804 | { |
1805 | struct dst_entry *ndst; | |
1806 | struct flowi6 fl6; | |
1807 | int err; | |
1808 | ||
d71785ff | 1809 | if (!skb->mark) { |
0c1d70af PA |
1810 | ndst = dst_cache_get_ip6(dst_cache, saddr); |
1811 | if (ndst) | |
1812 | return ndst; | |
1813 | } | |
1814 | ||
e5d4b29f JB |
1815 | memset(&fl6, 0, sizeof(fl6)); |
1816 | fl6.flowi6_oif = oif; | |
1817 | fl6.daddr = *daddr; | |
1818 | fl6.saddr = vxlan->cfg.saddr.sin6.sin6_addr; | |
1819 | fl6.flowi6_mark = skb->mark; | |
1820 | fl6.flowi6_proto = IPPROTO_UDP; | |
1821 | ||
1822 | err = ipv6_stub->ipv6_dst_lookup(vxlan->net, | |
1823 | vxlan->vn6_sock->sock->sk, | |
1824 | &ndst, &fl6); | |
1825 | if (err < 0) | |
1826 | return ERR_PTR(err); | |
1827 | ||
1828 | *saddr = fl6.saddr; | |
d71785ff | 1829 | if (!skb->mark) |
0c1d70af | 1830 | dst_cache_set_ip6(dst_cache, ndst, saddr); |
e5d4b29f JB |
1831 | return ndst; |
1832 | } | |
1833 | #endif | |
1834 | ||
9dcc71e1 SS |
1835 | /* Bypass encapsulation if the destination is local */ |
1836 | static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, | |
1837 | struct vxlan_dev *dst_vxlan) | |
1838 | { | |
8f84985f | 1839 | struct pcpu_sw_netstats *tx_stats, *rx_stats; |
e4c7ed41 CW |
1840 | union vxlan_addr loopback; |
1841 | union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip; | |
ce6502a8 LR |
1842 | struct net_device *dev = skb->dev; |
1843 | int len = skb->len; | |
9dcc71e1 | 1844 | |
8f84985f LR |
1845 | tx_stats = this_cpu_ptr(src_vxlan->dev->tstats); |
1846 | rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats); | |
9dcc71e1 SS |
1847 | skb->pkt_type = PACKET_HOST; |
1848 | skb->encapsulation = 0; | |
1849 | skb->dev = dst_vxlan->dev; | |
1850 | __skb_pull(skb, skb_network_offset(skb)); | |
1851 | ||
e4c7ed41 CW |
1852 | if (remote_ip->sa.sa_family == AF_INET) { |
1853 | loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK); | |
1854 | loopback.sa.sa_family = AF_INET; | |
1855 | #if IS_ENABLED(CONFIG_IPV6) | |
1856 | } else { | |
1857 | loopback.sin6.sin6_addr = in6addr_loopback; | |
1858 | loopback.sa.sa_family = AF_INET6; | |
1859 | #endif | |
1860 | } | |
1861 | ||
9dcc71e1 | 1862 | if (dst_vxlan->flags & VXLAN_F_LEARN) |
e4c7ed41 | 1863 | vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source); |
9dcc71e1 SS |
1864 | |
1865 | u64_stats_update_begin(&tx_stats->syncp); | |
1866 | tx_stats->tx_packets++; | |
ce6502a8 | 1867 | tx_stats->tx_bytes += len; |
9dcc71e1 SS |
1868 | u64_stats_update_end(&tx_stats->syncp); |
1869 | ||
1870 | if (netif_rx(skb) == NET_RX_SUCCESS) { | |
1871 | u64_stats_update_begin(&rx_stats->syncp); | |
1872 | rx_stats->rx_packets++; | |
ce6502a8 | 1873 | rx_stats->rx_bytes += len; |
9dcc71e1 SS |
1874 | u64_stats_update_end(&rx_stats->syncp); |
1875 | } else { | |
ce6502a8 | 1876 | dev->stats.rx_dropped++; |
9dcc71e1 SS |
1877 | } |
1878 | } | |
1879 | ||
4ad16930 SH |
1880 | static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, |
1881 | struct vxlan_rdst *rdst, bool did_rsc) | |
d342894c | 1882 | { |
d71785ff | 1883 | struct dst_cache *dst_cache; |
3093fbe7 | 1884 | struct ip_tunnel_info *info; |
d342894c | 1885 | struct vxlan_dev *vxlan = netdev_priv(dev); |
b1be00a6 | 1886 | struct sock *sk; |
e4c7ed41 | 1887 | struct rtable *rt = NULL; |
d342894c | 1888 | const struct iphdr *old_iph; |
e4c7ed41 | 1889 | union vxlan_addr *dst; |
ee122c79 TG |
1890 | union vxlan_addr remote_ip; |
1891 | struct vxlan_metadata _md; | |
1892 | struct vxlan_metadata *md = &_md; | |
e4c7ed41 | 1893 | __be16 src_port = 0, dst_port; |
54bfd872 | 1894 | __be32 vni; |
d342894c | 1895 | __be16 df = 0; |
1896 | __u8 tos, ttl; | |
0e6fbc5b | 1897 | int err; |
ee122c79 | 1898 | u32 flags = vxlan->flags; |
b4ed5cad | 1899 | bool udp_sum = false; |
f491e56d | 1900 | bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev)); |
d342894c | 1901 | |
61adedf3 | 1902 | info = skb_tunnel_info(skb); |
3093fbe7 | 1903 | |
ee122c79 | 1904 | if (rdst) { |
0dfbdf41 | 1905 | dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port; |
ee122c79 TG |
1906 | vni = rdst->remote_vni; |
1907 | dst = &rdst->remote_ip; | |
d71785ff | 1908 | dst_cache = &rdst->dst_cache; |
ee122c79 TG |
1909 | } else { |
1910 | if (!info) { | |
1911 | WARN_ONCE(1, "%s: Missing encapsulation instructions\n", | |
1912 | dev->name); | |
1913 | goto drop; | |
1914 | } | |
0dfbdf41 | 1915 | dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port; |
54bfd872 | 1916 | vni = vxlan_tun_id_to_vni(info->key.tun_id); |
b1be00a6 JB |
1917 | remote_ip.sa.sa_family = ip_tunnel_info_af(info); |
1918 | if (remote_ip.sa.sa_family == AF_INET) | |
a725e514 JB |
1919 | remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst; |
1920 | else | |
1921 | remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst; | |
ee122c79 | 1922 | dst = &remote_ip; |
d71785ff | 1923 | dst_cache = &info->dst_cache; |
ee122c79 | 1924 | } |
e4f67add | 1925 | |
e4c7ed41 | 1926 | if (vxlan_addr_any(dst)) { |
e4f67add | 1927 | if (did_rsc) { |
e4f67add | 1928 | /* short-circuited back to local bridge */ |
9dcc71e1 | 1929 | vxlan_encap_bypass(skb, vxlan, vxlan); |
4ad16930 | 1930 | return; |
e4f67add | 1931 | } |
ef59febe | 1932 | goto drop; |
e4f67add | 1933 | } |
ef59febe | 1934 | |
d342894c | 1935 | old_iph = ip_hdr(skb); |
1936 | ||
0dfbdf41 | 1937 | ttl = vxlan->cfg.ttl; |
e4c7ed41 | 1938 | if (!ttl && vxlan_addr_multicast(dst)) |
d342894c | 1939 | ttl = 1; |
1940 | ||
0dfbdf41 | 1941 | tos = vxlan->cfg.tos; |
d342894c | 1942 | if (tos == 1) |
206aaafc | 1943 | tos = ip_tunnel_get_dsfield(old_iph, skb); |
d342894c | 1944 | |
0dfbdf41 TG |
1945 | src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, |
1946 | vxlan->cfg.port_max, true); | |
d342894c | 1947 | |
a725e514 | 1948 | if (info) { |
a725e514 JB |
1949 | ttl = info->key.ttl; |
1950 | tos = info->key.tos; | |
b4ed5cad | 1951 | udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM); |
a725e514 JB |
1952 | |
1953 | if (info->options_len) | |
4c222798 | 1954 | md = ip_tunnel_info_opts(info); |
a725e514 JB |
1955 | } else { |
1956 | md->gbp = skb->mark; | |
1957 | } | |
1958 | ||
e4c7ed41 | 1959 | if (dst->sa.sa_family == AF_INET) { |
1a8496ba JB |
1960 | __be32 saddr; |
1961 | ||
b1be00a6 JB |
1962 | if (!vxlan->vn4_sock) |
1963 | goto drop; | |
1964 | sk = vxlan->vn4_sock->sock->sk; | |
1965 | ||
35e2d115 JG |
1966 | if (info) { |
1967 | if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) | |
1968 | df = htons(IP_DF); | |
b4ed5cad JB |
1969 | } else { |
1970 | udp_sum = !!(flags & VXLAN_F_UDP_CSUM); | |
35e2d115 | 1971 | } |
ee122c79 | 1972 | |
1a8496ba JB |
1973 | rt = vxlan_get_route(vxlan, skb, |
1974 | rdst ? rdst->remote_ifindex : 0, tos, | |
0c1d70af | 1975 | dst->sin.sin_addr.s_addr, &saddr, |
d71785ff | 1976 | dst_cache, info); |
e4c7ed41 CW |
1977 | if (IS_ERR(rt)) { |
1978 | netdev_dbg(dev, "no route to %pI4\n", | |
1979 | &dst->sin.sin_addr.s_addr); | |
1980 | dev->stats.tx_carrier_errors++; | |
1981 | goto tx_error; | |
1982 | } | |
d342894c | 1983 | |
e4c7ed41 CW |
1984 | if (rt->dst.dev == dev) { |
1985 | netdev_dbg(dev, "circular route to %pI4\n", | |
1986 | &dst->sin.sin_addr.s_addr); | |
1987 | dev->stats.collisions++; | |
fffc15a5 | 1988 | goto rt_tx_error; |
e4c7ed41 CW |
1989 | } |
1990 | ||
1991 | /* Bypass encapsulation if the destination is local */ | |
1992 | if (rt->rt_flags & RTCF_LOCAL && | |
1993 | !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { | |
1994 | struct vxlan_dev *dst_vxlan; | |
1995 | ||
1996 | ip_rt_put(rt); | |
19ca9fc1 | 1997 | dst_vxlan = vxlan_find_vni(vxlan->net, vni, |
ac5132d1 TG |
1998 | dst->sa.sa_family, dst_port, |
1999 | vxlan->flags); | |
e4c7ed41 CW |
2000 | if (!dst_vxlan) |
2001 | goto tx_error; | |
2002 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
2003 | return; | |
2004 | } | |
2005 | ||
2006 | tos = ip_tunnel_ecn_encap(tos, old_iph, skb); | |
2007 | ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); | |
f491e56d | 2008 | err = vxlan_build_skb(skb, &rt->dst, sizeof(struct iphdr), |
54bfd872 | 2009 | vni, md, flags, udp_sum); |
f491e56d JB |
2010 | if (err < 0) |
2011 | goto xmit_tx_error; | |
2012 | ||
2013 | udp_tunnel_xmit_skb(rt, sk, skb, saddr, | |
2014 | dst->sin.sin_addr.s_addr, tos, ttl, df, | |
2015 | src_port, dst_port, xnet, !udp_sum); | |
e4c7ed41 CW |
2016 | #if IS_ENABLED(CONFIG_IPV6) |
2017 | } else { | |
e4c7ed41 | 2018 | struct dst_entry *ndst; |
e5d4b29f | 2019 | struct in6_addr saddr; |
6f264e47 | 2020 | u32 rt6i_flags; |
e4c7ed41 | 2021 | |
b1be00a6 JB |
2022 | if (!vxlan->vn6_sock) |
2023 | goto drop; | |
2024 | sk = vxlan->vn6_sock->sock->sk; | |
2025 | ||
e5d4b29f JB |
2026 | ndst = vxlan6_get_route(vxlan, skb, |
2027 | rdst ? rdst->remote_ifindex : 0, | |
0c1d70af | 2028 | &dst->sin6.sin6_addr, &saddr, |
d71785ff | 2029 | dst_cache); |
e5d4b29f | 2030 | if (IS_ERR(ndst)) { |
e4c7ed41 CW |
2031 | netdev_dbg(dev, "no route to %pI6\n", |
2032 | &dst->sin6.sin6_addr); | |
2033 | dev->stats.tx_carrier_errors++; | |
9dcc71e1 | 2034 | goto tx_error; |
e4c7ed41 | 2035 | } |
d342894c | 2036 | |
e4c7ed41 CW |
2037 | if (ndst->dev == dev) { |
2038 | netdev_dbg(dev, "circular route to %pI6\n", | |
2039 | &dst->sin6.sin6_addr); | |
2040 | dst_release(ndst); | |
2041 | dev->stats.collisions++; | |
2042 | goto tx_error; | |
2043 | } | |
0e6fbc5b | 2044 | |
e4c7ed41 | 2045 | /* Bypass encapsulation if the destination is local */ |
6f264e47 JB |
2046 | rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags; |
2047 | if (rt6i_flags & RTF_LOCAL && | |
2048 | !(rt6i_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { | |
e4c7ed41 CW |
2049 | struct vxlan_dev *dst_vxlan; |
2050 | ||
2051 | dst_release(ndst); | |
19ca9fc1 | 2052 | dst_vxlan = vxlan_find_vni(vxlan->net, vni, |
ac5132d1 TG |
2053 | dst->sa.sa_family, dst_port, |
2054 | vxlan->flags); | |
e4c7ed41 CW |
2055 | if (!dst_vxlan) |
2056 | goto tx_error; | |
2057 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
2058 | return; | |
2059 | } | |
49560532 | 2060 | |
b4ed5cad JB |
2061 | if (!info) |
2062 | udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX); | |
35e2d115 | 2063 | |
e4c7ed41 | 2064 | ttl = ttl ? : ip6_dst_hoplimit(ndst); |
f491e56d JB |
2065 | skb_scrub_packet(skb, xnet); |
2066 | err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr), | |
54bfd872 | 2067 | vni, md, flags, udp_sum); |
f491e56d JB |
2068 | if (err < 0) { |
2069 | dst_release(ndst); | |
2070 | return; | |
2071 | } | |
2072 | udp_tunnel6_xmit_skb(ndst, sk, skb, dev, | |
2073 | &saddr, &dst->sin6.sin6_addr, | |
2074 | 0, ttl, src_port, dst_port, !udp_sum); | |
e4c7ed41 CW |
2075 | #endif |
2076 | } | |
0e6fbc5b | 2077 | |
4ad16930 | 2078 | return; |
d342894c | 2079 | |
2080 | drop: | |
2081 | dev->stats.tx_dropped++; | |
2082 | goto tx_free; | |
2083 | ||
f491e56d JB |
2084 | xmit_tx_error: |
2085 | /* skb is already freed. */ | |
2086 | skb = NULL; | |
49560532 PS |
2087 | rt_tx_error: |
2088 | ip_rt_put(rt); | |
d342894c | 2089 | tx_error: |
2090 | dev->stats.tx_errors++; | |
2091 | tx_free: | |
2092 | dev_kfree_skb(skb); | |
d342894c | 2093 | } |
2094 | ||
6681712d DS |
2095 | /* Transmit local packets over Vxlan |
2096 | * | |
2097 | * Outer IP header inherits ECN and DF from inner header. | |
2098 | * Outer UDP destination is the VXLAN assigned port. | |
2099 | * source port is based on hash of flow | |
2100 | */ | |
2101 | static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) | |
2102 | { | |
2103 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
3093fbe7 | 2104 | const struct ip_tunnel_info *info; |
6681712d DS |
2105 | struct ethhdr *eth; |
2106 | bool did_rsc = false; | |
8f646c92 | 2107 | struct vxlan_rdst *rdst, *fdst = NULL; |
6681712d | 2108 | struct vxlan_fdb *f; |
6681712d | 2109 | |
61adedf3 | 2110 | info = skb_tunnel_info(skb); |
3093fbe7 | 2111 | |
6681712d DS |
2112 | skb_reset_mac_header(skb); |
2113 | eth = eth_hdr(skb); | |
2114 | ||
f564f45c CW |
2115 | if ((vxlan->flags & VXLAN_F_PROXY)) { |
2116 | if (ntohs(eth->h_proto) == ETH_P_ARP) | |
2117 | return arp_reduce(dev, skb); | |
2118 | #if IS_ENABLED(CONFIG_IPV6) | |
2119 | else if (ntohs(eth->h_proto) == ETH_P_IPV6 && | |
91269e39 LR |
2120 | pskb_may_pull(skb, sizeof(struct ipv6hdr) |
2121 | + sizeof(struct nd_msg)) && | |
f564f45c CW |
2122 | ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) { |
2123 | struct nd_msg *msg; | |
2124 | ||
2125 | msg = (struct nd_msg *)skb_transport_header(skb); | |
2126 | if (msg->icmph.icmp6_code == 0 && | |
2127 | msg->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION) | |
2128 | return neigh_reduce(dev, skb); | |
2129 | } | |
7a9f526f | 2130 | eth = eth_hdr(skb); |
f564f45c CW |
2131 | #endif |
2132 | } | |
6681712d | 2133 | |
da8b43c0 | 2134 | if (vxlan->flags & VXLAN_F_COLLECT_METADATA && |
46fa062a | 2135 | info && info->mode & IP_TUNNEL_INFO_TX) { |
ee122c79 TG |
2136 | vxlan_xmit_one(skb, dev, NULL, false); |
2137 | return NETDEV_TX_OK; | |
2138 | } | |
2139 | ||
6681712d | 2140 | f = vxlan_find_mac(vxlan, eth->h_dest); |
ae884082 DS |
2141 | did_rsc = false; |
2142 | ||
2143 | if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) && | |
e15a00aa CW |
2144 | (ntohs(eth->h_proto) == ETH_P_IP || |
2145 | ntohs(eth->h_proto) == ETH_P_IPV6)) { | |
ae884082 DS |
2146 | did_rsc = route_shortcircuit(dev, skb); |
2147 | if (did_rsc) | |
2148 | f = vxlan_find_mac(vxlan, eth->h_dest); | |
2149 | } | |
2150 | ||
6681712d | 2151 | if (f == NULL) { |
afbd8bae MR |
2152 | f = vxlan_find_mac(vxlan, all_zeros_mac); |
2153 | if (f == NULL) { | |
2154 | if ((vxlan->flags & VXLAN_F_L2MISS) && | |
2155 | !is_multicast_ether_addr(eth->h_dest)) | |
2156 | vxlan_fdb_miss(vxlan, eth->h_dest); | |
2157 | ||
2158 | dev->stats.tx_dropped++; | |
8f646c92 | 2159 | kfree_skb(skb); |
afbd8bae MR |
2160 | return NETDEV_TX_OK; |
2161 | } | |
2162 | } | |
6681712d | 2163 | |
afbd8bae MR |
2164 | list_for_each_entry_rcu(rdst, &f->remotes, list) { |
2165 | struct sk_buff *skb1; | |
6681712d | 2166 | |
8f646c92 ED |
2167 | if (!fdst) { |
2168 | fdst = rdst; | |
2169 | continue; | |
2170 | } | |
afbd8bae MR |
2171 | skb1 = skb_clone(skb, GFP_ATOMIC); |
2172 | if (skb1) | |
2173 | vxlan_xmit_one(skb1, dev, rdst, did_rsc); | |
6681712d DS |
2174 | } |
2175 | ||
8f646c92 ED |
2176 | if (fdst) |
2177 | vxlan_xmit_one(skb, dev, fdst, did_rsc); | |
2178 | else | |
2179 | kfree_skb(skb); | |
4ad16930 | 2180 | return NETDEV_TX_OK; |
6681712d DS |
2181 | } |
2182 | ||
d342894c | 2183 | /* Walk the forwarding table and purge stale entries */ |
2184 | static void vxlan_cleanup(unsigned long arg) | |
2185 | { | |
2186 | struct vxlan_dev *vxlan = (struct vxlan_dev *) arg; | |
2187 | unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; | |
2188 | unsigned int h; | |
2189 | ||
2190 | if (!netif_running(vxlan->dev)) | |
2191 | return; | |
2192 | ||
d342894c | 2193 | for (h = 0; h < FDB_HASH_SIZE; ++h) { |
2194 | struct hlist_node *p, *n; | |
14e1d0fa SD |
2195 | |
2196 | spin_lock_bh(&vxlan->hash_lock); | |
d342894c | 2197 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { |
2198 | struct vxlan_fdb *f | |
2199 | = container_of(p, struct vxlan_fdb, hlist); | |
2200 | unsigned long timeout; | |
2201 | ||
3c172868 | 2202 | if (f->state & NUD_PERMANENT) |
d342894c | 2203 | continue; |
2204 | ||
0dfbdf41 | 2205 | timeout = f->used + vxlan->cfg.age_interval * HZ; |
d342894c | 2206 | if (time_before_eq(timeout, jiffies)) { |
2207 | netdev_dbg(vxlan->dev, | |
2208 | "garbage collect %pM\n", | |
2209 | f->eth_addr); | |
2210 | f->state = NUD_STALE; | |
2211 | vxlan_fdb_destroy(vxlan, f); | |
2212 | } else if (time_before(timeout, next_timer)) | |
2213 | next_timer = timeout; | |
2214 | } | |
14e1d0fa | 2215 | spin_unlock_bh(&vxlan->hash_lock); |
d342894c | 2216 | } |
d342894c | 2217 | |
2218 | mod_timer(&vxlan->age_timer, next_timer); | |
2219 | } | |
2220 | ||
9c2e24e1 PS |
2221 | static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan) |
2222 | { | |
56ef9c90 | 2223 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
54bfd872 | 2224 | __be32 vni = vxlan->default_dst.remote_vni; |
9c2e24e1 | 2225 | |
56ef9c90 | 2226 | spin_lock(&vn->sock_lock); |
9c2e24e1 | 2227 | hlist_add_head_rcu(&vxlan->hlist, vni_head(vs, vni)); |
56ef9c90 | 2228 | spin_unlock(&vn->sock_lock); |
9c2e24e1 PS |
2229 | } |
2230 | ||
d342894c | 2231 | /* Setup stats when device is created */ |
2232 | static int vxlan_init(struct net_device *dev) | |
2233 | { | |
1c213bd2 | 2234 | dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats); |
e8171045 | 2235 | if (!dev->tstats) |
d342894c | 2236 | return -ENOMEM; |
2237 | ||
2238 | return 0; | |
2239 | } | |
2240 | ||
ba609e9b | 2241 | static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan) |
afbd8bae MR |
2242 | { |
2243 | struct vxlan_fdb *f; | |
2244 | ||
2245 | spin_lock_bh(&vxlan->hash_lock); | |
2246 | f = __vxlan_find_mac(vxlan, all_zeros_mac); | |
2247 | if (f) | |
2248 | vxlan_fdb_destroy(vxlan, f); | |
2249 | spin_unlock_bh(&vxlan->hash_lock); | |
2250 | } | |
2251 | ||
ebf4063e SH |
2252 | static void vxlan_uninit(struct net_device *dev) |
2253 | { | |
2254 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
ebf4063e | 2255 | |
ba609e9b | 2256 | vxlan_fdb_delete_default(vxlan); |
afbd8bae | 2257 | |
ebf4063e SH |
2258 | free_percpu(dev->tstats); |
2259 | } | |
2260 | ||
d342894c | 2261 | /* Start ageing timer and join group when device is brought up */ |
2262 | static int vxlan_open(struct net_device *dev) | |
2263 | { | |
2264 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
205f356d | 2265 | int ret; |
56ef9c90 | 2266 | |
205f356d JB |
2267 | ret = vxlan_sock_add(vxlan); |
2268 | if (ret < 0) | |
2269 | return ret; | |
d342894c | 2270 | |
79d4a94f | 2271 | if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) { |
56ef9c90 | 2272 | ret = vxlan_igmp_join(vxlan); |
bef0057b MRL |
2273 | if (ret == -EADDRINUSE) |
2274 | ret = 0; | |
56ef9c90 | 2275 | if (ret) { |
205f356d | 2276 | vxlan_sock_release(vxlan); |
56ef9c90 MRL |
2277 | return ret; |
2278 | } | |
d342894c | 2279 | } |
2280 | ||
0dfbdf41 | 2281 | if (vxlan->cfg.age_interval) |
d342894c | 2282 | mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); |
2283 | ||
56ef9c90 | 2284 | return ret; |
d342894c | 2285 | } |
2286 | ||
2287 | /* Purge the forwarding table */ | |
2288 | static void vxlan_flush(struct vxlan_dev *vxlan) | |
2289 | { | |
31fec5aa | 2290 | unsigned int h; |
d342894c | 2291 | |
2292 | spin_lock_bh(&vxlan->hash_lock); | |
2293 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
2294 | struct hlist_node *p, *n; | |
2295 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { | |
2296 | struct vxlan_fdb *f | |
2297 | = container_of(p, struct vxlan_fdb, hlist); | |
afbd8bae MR |
2298 | /* the all_zeros_mac entry is deleted at vxlan_uninit */ |
2299 | if (!is_zero_ether_addr(f->eth_addr)) | |
2300 | vxlan_fdb_destroy(vxlan, f); | |
d342894c | 2301 | } |
2302 | } | |
2303 | spin_unlock_bh(&vxlan->hash_lock); | |
2304 | } | |
2305 | ||
2306 | /* Cleanup timer and forwarding table on shutdown */ | |
2307 | static int vxlan_stop(struct net_device *dev) | |
2308 | { | |
2309 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
f01ec1c0 | 2310 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
56ef9c90 | 2311 | int ret = 0; |
d342894c | 2312 | |
24c0e683 | 2313 | if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) && |
f13b1689 | 2314 | !vxlan_group_used(vn, vxlan)) |
56ef9c90 | 2315 | ret = vxlan_igmp_leave(vxlan); |
d342894c | 2316 | |
2317 | del_timer_sync(&vxlan->age_timer); | |
2318 | ||
2319 | vxlan_flush(vxlan); | |
205f356d | 2320 | vxlan_sock_release(vxlan); |
d342894c | 2321 | |
56ef9c90 | 2322 | return ret; |
d342894c | 2323 | } |
2324 | ||
d342894c | 2325 | /* Stub, nothing needs to be done. */ |
2326 | static void vxlan_set_multicast_list(struct net_device *dev) | |
2327 | { | |
2328 | } | |
2329 | ||
345010b5 DB |
2330 | static int vxlan_change_mtu(struct net_device *dev, int new_mtu) |
2331 | { | |
2332 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
2333 | struct vxlan_rdst *dst = &vxlan->default_dst; | |
2334 | struct net_device *lowerdev; | |
2335 | int max_mtu; | |
2336 | ||
f01ec1c0 | 2337 | lowerdev = __dev_get_by_index(vxlan->net, dst->remote_ifindex); |
345010b5 DB |
2338 | if (lowerdev == NULL) |
2339 | return eth_change_mtu(dev, new_mtu); | |
2340 | ||
2341 | if (dst->remote_ip.sa.sa_family == AF_INET6) | |
2342 | max_mtu = lowerdev->mtu - VXLAN6_HEADROOM; | |
2343 | else | |
2344 | max_mtu = lowerdev->mtu - VXLAN_HEADROOM; | |
2345 | ||
2346 | if (new_mtu < 68 || new_mtu > max_mtu) | |
2347 | return -EINVAL; | |
2348 | ||
2349 | dev->mtu = new_mtu; | |
2350 | return 0; | |
2351 | } | |
2352 | ||
fc4099f1 PS |
2353 | static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb) |
2354 | { | |
2355 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
2356 | struct ip_tunnel_info *info = skb_tunnel_info(skb); | |
2357 | __be16 sport, dport; | |
2358 | ||
2359 | sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min, | |
2360 | vxlan->cfg.port_max, true); | |
2361 | dport = info->key.tp_dst ? : vxlan->cfg.dst_port; | |
2362 | ||
239e944f | 2363 | if (ip_tunnel_info_af(info) == AF_INET) { |
1a8496ba JB |
2364 | struct rtable *rt; |
2365 | ||
239e944f JB |
2366 | if (!vxlan->vn4_sock) |
2367 | return -EINVAL; | |
1a8496ba JB |
2368 | rt = vxlan_get_route(vxlan, skb, 0, info->key.tos, |
2369 | info->key.u.ipv4.dst, | |
0c1d70af | 2370 | &info->key.u.ipv4.src, NULL, info); |
1a8496ba JB |
2371 | if (IS_ERR(rt)) |
2372 | return PTR_ERR(rt); | |
2373 | ip_rt_put(rt); | |
239e944f JB |
2374 | } else { |
2375 | #if IS_ENABLED(CONFIG_IPV6) | |
2376 | struct dst_entry *ndst; | |
2377 | ||
2378 | if (!vxlan->vn6_sock) | |
2379 | return -EINVAL; | |
2380 | ndst = vxlan6_get_route(vxlan, skb, 0, | |
2381 | &info->key.u.ipv6.dst, | |
0c1d70af | 2382 | &info->key.u.ipv6.src, NULL); |
239e944f JB |
2383 | if (IS_ERR(ndst)) |
2384 | return PTR_ERR(ndst); | |
2385 | dst_release(ndst); | |
239e944f JB |
2386 | #else /* !CONFIG_IPV6 */ |
2387 | return -EPFNOSUPPORT; | |
2388 | #endif | |
2389 | } | |
1a8496ba JB |
2390 | info->key.tp_src = sport; |
2391 | info->key.tp_dst = dport; | |
239e944f | 2392 | return 0; |
fc4099f1 PS |
2393 | } |
2394 | ||
d342894c | 2395 | static const struct net_device_ops vxlan_netdev_ops = { |
2396 | .ndo_init = vxlan_init, | |
ebf4063e | 2397 | .ndo_uninit = vxlan_uninit, |
d342894c | 2398 | .ndo_open = vxlan_open, |
2399 | .ndo_stop = vxlan_stop, | |
2400 | .ndo_start_xmit = vxlan_xmit, | |
e8171045 | 2401 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
d342894c | 2402 | .ndo_set_rx_mode = vxlan_set_multicast_list, |
345010b5 | 2403 | .ndo_change_mtu = vxlan_change_mtu, |
d342894c | 2404 | .ndo_validate_addr = eth_validate_addr, |
2405 | .ndo_set_mac_address = eth_mac_addr, | |
2406 | .ndo_fdb_add = vxlan_fdb_add, | |
2407 | .ndo_fdb_del = vxlan_fdb_delete, | |
2408 | .ndo_fdb_dump = vxlan_fdb_dump, | |
fc4099f1 | 2409 | .ndo_fill_metadata_dst = vxlan_fill_metadata_dst, |
d342894c | 2410 | }; |
2411 | ||
2412 | /* Info for udev, that this is a virtual tunnel endpoint */ | |
2413 | static struct device_type vxlan_type = { | |
2414 | .name = "vxlan", | |
2415 | }; | |
2416 | ||
53cf5275 | 2417 | /* Calls the ndo_add_vxlan_port of the caller in order to |
35e42379 JG |
2418 | * supply the listening VXLAN udp ports. Callers are expected |
2419 | * to implement the ndo_add_vxlan_port. | |
53cf5275 JG |
2420 | */ |
2421 | void vxlan_get_rx_port(struct net_device *dev) | |
2422 | { | |
2423 | struct vxlan_sock *vs; | |
2424 | struct net *net = dev_net(dev); | |
2425 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
2426 | sa_family_t sa_family; | |
35e42379 JG |
2427 | __be16 port; |
2428 | unsigned int i; | |
53cf5275 JG |
2429 | |
2430 | spin_lock(&vn->sock_lock); | |
2431 | for (i = 0; i < PORT_HASH_SIZE; ++i) { | |
35e42379 JG |
2432 | hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) { |
2433 | port = inet_sk(vs->sock->sk)->inet_sport; | |
705cc62f | 2434 | sa_family = vxlan_get_sk_family(vs); |
53cf5275 JG |
2435 | dev->netdev_ops->ndo_add_vxlan_port(dev, sa_family, |
2436 | port); | |
2437 | } | |
2438 | } | |
2439 | spin_unlock(&vn->sock_lock); | |
2440 | } | |
2441 | EXPORT_SYMBOL_GPL(vxlan_get_rx_port); | |
2442 | ||
d342894c | 2443 | /* Initialize the device structure. */ |
2444 | static void vxlan_setup(struct net_device *dev) | |
2445 | { | |
2446 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
31fec5aa | 2447 | unsigned int h; |
d342894c | 2448 | |
2449 | eth_hw_addr_random(dev); | |
2450 | ether_setup(dev); | |
2451 | ||
2452 | dev->netdev_ops = &vxlan_netdev_ops; | |
ebf4063e | 2453 | dev->destructor = free_netdev; |
d342894c | 2454 | SET_NETDEV_DEVTYPE(dev, &vxlan_type); |
2455 | ||
d342894c | 2456 | dev->features |= NETIF_F_LLTX; |
d6727fe3 | 2457 | dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; |
0afb1666 | 2458 | dev->features |= NETIF_F_RXCSUM; |
05c0db08 | 2459 | dev->features |= NETIF_F_GSO_SOFTWARE; |
0afb1666 | 2460 | |
1eaa8178 PS |
2461 | dev->vlan_features = dev->features; |
2462 | dev->features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; | |
0afb1666 | 2463 | dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM; |
05c0db08 | 2464 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; |
1eaa8178 | 2465 | dev->hw_features |= NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_STAG_TX; |
02875878 | 2466 | netif_keep_dst(dev); |
22dba393 | 2467 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE | IFF_NO_QUEUE; |
d342894c | 2468 | |
553675fb | 2469 | INIT_LIST_HEAD(&vxlan->next); |
d342894c | 2470 | spin_lock_init(&vxlan->hash_lock); |
2471 | ||
2472 | init_timer_deferrable(&vxlan->age_timer); | |
2473 | vxlan->age_timer.function = vxlan_cleanup; | |
2474 | vxlan->age_timer.data = (unsigned long) vxlan; | |
2475 | ||
0dfbdf41 | 2476 | vxlan->cfg.dst_port = htons(vxlan_port); |
05f47d69 | 2477 | |
d342894c | 2478 | vxlan->dev = dev; |
2479 | ||
58ce31cc TH |
2480 | gro_cells_init(&vxlan->gro_cells, dev); |
2481 | ||
d342894c | 2482 | for (h = 0; h < FDB_HASH_SIZE; ++h) |
2483 | INIT_HLIST_HEAD(&vxlan->fdb_head[h]); | |
2484 | } | |
2485 | ||
2486 | static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { | |
2487 | [IFLA_VXLAN_ID] = { .type = NLA_U32 }, | |
5d174dd8 | 2488 | [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, |
e4c7ed41 | 2489 | [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) }, |
d342894c | 2490 | [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, |
2491 | [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, | |
e4c7ed41 | 2492 | [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) }, |
d342894c | 2493 | [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, |
2494 | [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, | |
2495 | [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, | |
2496 | [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, | |
2497 | [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, | |
05f47d69 | 2498 | [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, |
e4f67add DS |
2499 | [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, |
2500 | [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, | |
2501 | [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, | |
2502 | [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, | |
f8a9b1bc | 2503 | [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 }, |
823aa873 | 2504 | [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, |
5c91ae08 TH |
2505 | [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 }, |
2506 | [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 }, | |
2507 | [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 }, | |
dfd8645e TH |
2508 | [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 }, |
2509 | [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 }, | |
3511494c | 2510 | [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, }, |
0ace2ca8 | 2511 | [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG }, |
d342894c | 2512 | }; |
2513 | ||
2514 | static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[]) | |
2515 | { | |
2516 | if (tb[IFLA_ADDRESS]) { | |
2517 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { | |
2518 | pr_debug("invalid link address (not ethernet)\n"); | |
2519 | return -EINVAL; | |
2520 | } | |
2521 | ||
2522 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { | |
2523 | pr_debug("invalid all zero ethernet address\n"); | |
2524 | return -EADDRNOTAVAIL; | |
2525 | } | |
2526 | } | |
2527 | ||
2528 | if (!data) | |
2529 | return -EINVAL; | |
2530 | ||
2531 | if (data[IFLA_VXLAN_ID]) { | |
2532 | __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); | |
2533 | if (id >= VXLAN_VID_MASK) | |
2534 | return -ERANGE; | |
2535 | } | |
2536 | ||
05f47d69 | 2537 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
2538 | const struct ifla_vxlan_port_range *p | |
2539 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
2540 | ||
2541 | if (ntohs(p->high) < ntohs(p->low)) { | |
2542 | pr_debug("port range %u .. %u not valid\n", | |
2543 | ntohs(p->low), ntohs(p->high)); | |
2544 | return -EINVAL; | |
2545 | } | |
2546 | } | |
2547 | ||
d342894c | 2548 | return 0; |
2549 | } | |
2550 | ||
1b13c97f YB |
2551 | static void vxlan_get_drvinfo(struct net_device *netdev, |
2552 | struct ethtool_drvinfo *drvinfo) | |
2553 | { | |
2554 | strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); | |
2555 | strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); | |
2556 | } | |
2557 | ||
2558 | static const struct ethtool_ops vxlan_ethtool_ops = { | |
2559 | .get_drvinfo = vxlan_get_drvinfo, | |
2560 | .get_link = ethtool_op_get_link, | |
2561 | }; | |
2562 | ||
553675fb | 2563 | static void vxlan_del_work(struct work_struct *work) |
2564 | { | |
2565 | struct vxlan_sock *vs = container_of(work, struct vxlan_sock, del_work); | |
acbf74a7 | 2566 | udp_tunnel_sock_release(vs->sock); |
553675fb | 2567 | kfree_rcu(vs, rcu); |
2568 | } | |
2569 | ||
3ee64f39 TH |
2570 | static struct socket *vxlan_create_sock(struct net *net, bool ipv6, |
2571 | __be16 port, u32 flags) | |
553675fb | 2572 | { |
e4c7ed41 | 2573 | struct socket *sock; |
3ee64f39 TH |
2574 | struct udp_port_cfg udp_conf; |
2575 | int err; | |
e4c7ed41 | 2576 | |
3ee64f39 | 2577 | memset(&udp_conf, 0, sizeof(udp_conf)); |
e4c7ed41 | 2578 | |
3ee64f39 TH |
2579 | if (ipv6) { |
2580 | udp_conf.family = AF_INET6; | |
3ee64f39 | 2581 | udp_conf.use_udp6_rx_checksums = |
3dc2b6a8 | 2582 | !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX); |
a43a9ef6 | 2583 | udp_conf.ipv6_v6only = 1; |
3ee64f39 TH |
2584 | } else { |
2585 | udp_conf.family = AF_INET; | |
e4c7ed41 | 2586 | } |
e4c7ed41 | 2587 | |
3ee64f39 | 2588 | udp_conf.local_udp_port = port; |
553675fb | 2589 | |
3ee64f39 TH |
2590 | /* Open UDP socket */ |
2591 | err = udp_sock_create(net, &udp_conf, &sock); | |
2592 | if (err < 0) | |
2593 | return ERR_PTR(err); | |
e4c7ed41 | 2594 | |
39deb2c7 | 2595 | return sock; |
e4c7ed41 CW |
2596 | } |
2597 | ||
2598 | /* Create new listen socket if needed */ | |
b1be00a6 JB |
2599 | static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6, |
2600 | __be16 port, u32 flags) | |
e4c7ed41 CW |
2601 | { |
2602 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
2603 | struct vxlan_sock *vs; | |
2604 | struct socket *sock; | |
e4c7ed41 | 2605 | unsigned int h; |
acbf74a7 | 2606 | struct udp_tunnel_sock_cfg tunnel_cfg; |
e4c7ed41 | 2607 | |
dc01e7d3 | 2608 | vs = kzalloc(sizeof(*vs), GFP_KERNEL); |
e4c7ed41 CW |
2609 | if (!vs) |
2610 | return ERR_PTR(-ENOMEM); | |
2611 | ||
2612 | for (h = 0; h < VNI_HASH_SIZE; ++h) | |
2613 | INIT_HLIST_HEAD(&vs->vni_list[h]); | |
2614 | ||
2615 | INIT_WORK(&vs->del_work, vxlan_del_work); | |
2616 | ||
3ee64f39 | 2617 | sock = vxlan_create_sock(net, ipv6, port, flags); |
39deb2c7 | 2618 | if (IS_ERR(sock)) { |
56ef9c90 MRL |
2619 | pr_info("Cannot bind port %d, err=%ld\n", ntohs(port), |
2620 | PTR_ERR(sock)); | |
553675fb | 2621 | kfree(vs); |
e50fddc8 | 2622 | return ERR_CAST(sock); |
553675fb | 2623 | } |
e4c7ed41 CW |
2624 | |
2625 | vs->sock = sock; | |
9c2e24e1 | 2626 | atomic_set(&vs->refcnt, 1); |
af33c1ad | 2627 | vs->flags = (flags & VXLAN_F_RCV_FLAGS); |
553675fb | 2628 | |
dc01e7d3 OG |
2629 | /* Initialize the vxlan udp offloads structure */ |
2630 | vs->udp_offloads.port = port; | |
2631 | vs->udp_offloads.callbacks.gro_receive = vxlan_gro_receive; | |
2632 | vs->udp_offloads.callbacks.gro_complete = vxlan_gro_complete; | |
2633 | ||
9c2e24e1 PS |
2634 | spin_lock(&vn->sock_lock); |
2635 | hlist_add_head_rcu(&vs->hlist, vs_head(net, port)); | |
dc01e7d3 | 2636 | vxlan_notify_add_rx_port(vs); |
9c2e24e1 | 2637 | spin_unlock(&vn->sock_lock); |
553675fb | 2638 | |
2639 | /* Mark socket as an encapsulation socket. */ | |
acbf74a7 AZ |
2640 | tunnel_cfg.sk_user_data = vs; |
2641 | tunnel_cfg.encap_type = 1; | |
2642 | tunnel_cfg.encap_rcv = vxlan_udp_encap_recv; | |
2643 | tunnel_cfg.encap_destroy = NULL; | |
2644 | ||
2645 | setup_udp_tunnel_sock(net, sock, &tunnel_cfg); | |
e4c7ed41 | 2646 | |
9c2e24e1 PS |
2647 | return vs; |
2648 | } | |
2649 | ||
b1be00a6 | 2650 | static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6) |
9c2e24e1 | 2651 | { |
205f356d JB |
2652 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
2653 | struct vxlan_sock *vs = NULL; | |
9c2e24e1 | 2654 | |
205f356d | 2655 | if (!vxlan->cfg.no_share) { |
56ef9c90 | 2656 | spin_lock(&vn->sock_lock); |
205f356d JB |
2657 | vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET, |
2658 | vxlan->cfg.dst_port, vxlan->flags); | |
2659 | if (vs && !atomic_add_unless(&vs->refcnt, 1, 0)) { | |
56ef9c90 | 2660 | spin_unlock(&vn->sock_lock); |
205f356d | 2661 | return -EBUSY; |
56ef9c90 MRL |
2662 | } |
2663 | spin_unlock(&vn->sock_lock); | |
2664 | } | |
205f356d | 2665 | if (!vs) |
b1be00a6 JB |
2666 | vs = vxlan_socket_create(vxlan->net, ipv6, |
2667 | vxlan->cfg.dst_port, vxlan->flags); | |
205f356d JB |
2668 | if (IS_ERR(vs)) |
2669 | return PTR_ERR(vs); | |
b1be00a6 JB |
2670 | #if IS_ENABLED(CONFIG_IPV6) |
2671 | if (ipv6) | |
2672 | vxlan->vn6_sock = vs; | |
2673 | else | |
2674 | #endif | |
2675 | vxlan->vn4_sock = vs; | |
205f356d JB |
2676 | vxlan_vs_add_dev(vs, vxlan); |
2677 | return 0; | |
553675fb | 2678 | } |
2679 | ||
b1be00a6 JB |
2680 | static int vxlan_sock_add(struct vxlan_dev *vxlan) |
2681 | { | |
2682 | bool ipv6 = vxlan->flags & VXLAN_F_IPV6; | |
2683 | bool metadata = vxlan->flags & VXLAN_F_COLLECT_METADATA; | |
2684 | int ret = 0; | |
2685 | ||
2686 | vxlan->vn4_sock = NULL; | |
2687 | #if IS_ENABLED(CONFIG_IPV6) | |
2688 | vxlan->vn6_sock = NULL; | |
2689 | if (ipv6 || metadata) | |
2690 | ret = __vxlan_sock_add(vxlan, true); | |
2691 | #endif | |
2692 | if (!ret && (!ipv6 || metadata)) | |
2693 | ret = __vxlan_sock_add(vxlan, false); | |
2694 | if (ret < 0) | |
2695 | vxlan_sock_release(vxlan); | |
2696 | return ret; | |
2697 | } | |
2698 | ||
0dfbdf41 TG |
2699 | static int vxlan_dev_configure(struct net *src_net, struct net_device *dev, |
2700 | struct vxlan_config *conf) | |
d342894c | 2701 | { |
33564bbb | 2702 | struct vxlan_net *vn = net_generic(src_net, vxlan_net_id); |
07b9b37c | 2703 | struct vxlan_dev *vxlan = netdev_priv(dev), *tmp; |
c7995c43 | 2704 | struct vxlan_rdst *dst = &vxlan->default_dst; |
b1be00a6 | 2705 | unsigned short needed_headroom = ETH_HLEN; |
d342894c | 2706 | int err; |
e4c7ed41 | 2707 | bool use_ipv6 = false; |
0dfbdf41 | 2708 | __be16 default_port = vxlan->cfg.dst_port; |
d342894c | 2709 | |
33564bbb | 2710 | vxlan->net = src_net; |
f01ec1c0 | 2711 | |
0dfbdf41 | 2712 | dst->remote_vni = conf->vni; |
d342894c | 2713 | |
0dfbdf41 | 2714 | memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip)); |
d342894c | 2715 | |
0dfbdf41 TG |
2716 | /* Unless IPv6 is explicitly requested, assume IPv4 */ |
2717 | if (!dst->remote_ip.sa.sa_family) | |
2718 | dst->remote_ip.sa.sa_family = AF_INET; | |
e4c7ed41 | 2719 | |
0dfbdf41 | 2720 | if (dst->remote_ip.sa.sa_family == AF_INET6 || |
057ba29b JB |
2721 | vxlan->cfg.saddr.sa.sa_family == AF_INET6) { |
2722 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2723 | return -EPFNOSUPPORT; | |
e4c7ed41 | 2724 | use_ipv6 = true; |
b1be00a6 | 2725 | vxlan->flags |= VXLAN_F_IPV6; |
057ba29b | 2726 | } |
d342894c | 2727 | |
0dfbdf41 | 2728 | if (conf->remote_ifindex) { |
34e02aa1 | 2729 | struct net_device *lowerdev |
0dfbdf41 TG |
2730 | = __dev_get_by_index(src_net, conf->remote_ifindex); |
2731 | ||
2732 | dst->remote_ifindex = conf->remote_ifindex; | |
34e02aa1 | 2733 | |
2734 | if (!lowerdev) { | |
c7995c43 | 2735 | pr_info("ifindex %d does not exist\n", dst->remote_ifindex); |
34e02aa1 | 2736 | return -ENODEV; |
2737 | } | |
d342894c | 2738 | |
e4c7ed41 CW |
2739 | #if IS_ENABLED(CONFIG_IPV6) |
2740 | if (use_ipv6) { | |
2741 | struct inet6_dev *idev = __in6_dev_get(lowerdev); | |
2742 | if (idev && idev->cnf.disable_ipv6) { | |
2743 | pr_info("IPv6 is disabled via sysctl\n"); | |
2744 | return -EPERM; | |
2745 | } | |
e4c7ed41 CW |
2746 | } |
2747 | #endif | |
2748 | ||
0dfbdf41 | 2749 | if (!conf->mtu) |
e4c7ed41 | 2750 | dev->mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM); |
1ba56fb4 | 2751 | |
b1be00a6 | 2752 | needed_headroom = lowerdev->hard_header_len; |
9dc2ad10 | 2753 | } |
d342894c | 2754 | |
b1be00a6 JB |
2755 | if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA) |
2756 | needed_headroom += VXLAN6_HEADROOM; | |
2757 | else | |
2758 | needed_headroom += VXLAN_HEADROOM; | |
2759 | dev->needed_headroom = needed_headroom; | |
2760 | ||
0dfbdf41 TG |
2761 | memcpy(&vxlan->cfg, conf, sizeof(*conf)); |
2762 | if (!vxlan->cfg.dst_port) | |
2763 | vxlan->cfg.dst_port = default_port; | |
2764 | vxlan->flags |= conf->flags; | |
2765 | ||
2766 | if (!vxlan->cfg.age_interval) | |
2767 | vxlan->cfg.age_interval = FDB_AGE_DEFAULT; | |
2768 | ||
07b9b37c ND |
2769 | list_for_each_entry(tmp, &vn->vxlan_list, next) { |
2770 | if (tmp->cfg.vni == conf->vni && | |
2771 | (tmp->default_dst.remote_ip.sa.sa_family == AF_INET6 || | |
2772 | tmp->cfg.saddr.sa.sa_family == AF_INET6) == use_ipv6 && | |
2773 | tmp->cfg.dst_port == vxlan->cfg.dst_port && | |
2774 | (tmp->flags & VXLAN_F_RCV_FLAGS) == | |
2775 | (vxlan->flags & VXLAN_F_RCV_FLAGS)) | |
0dfbdf41 | 2776 | return -EEXIST; |
07b9b37c | 2777 | } |
0dfbdf41 TG |
2778 | |
2779 | dev->ethtool_ops = &vxlan_ethtool_ops; | |
2780 | ||
2781 | /* create an fdb entry for a valid default destination */ | |
2782 | if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) { | |
2783 | err = vxlan_fdb_create(vxlan, all_zeros_mac, | |
2784 | &vxlan->default_dst.remote_ip, | |
2785 | NUD_REACHABLE|NUD_PERMANENT, | |
2786 | NLM_F_EXCL|NLM_F_CREATE, | |
2787 | vxlan->cfg.dst_port, | |
2788 | vxlan->default_dst.remote_vni, | |
2789 | vxlan->default_dst.remote_ifindex, | |
2790 | NTF_SELF); | |
2791 | if (err) | |
2792 | return err; | |
2793 | } | |
2794 | ||
2795 | err = register_netdevice(dev); | |
2796 | if (err) { | |
2797 | vxlan_fdb_delete_default(vxlan); | |
2798 | return err; | |
2799 | } | |
2800 | ||
2801 | list_add(&vxlan->next, &vn->vxlan_list); | |
2802 | ||
2803 | return 0; | |
2804 | } | |
2805 | ||
2806 | struct net_device *vxlan_dev_create(struct net *net, const char *name, | |
2807 | u8 name_assign_type, struct vxlan_config *conf) | |
2808 | { | |
2809 | struct nlattr *tb[IFLA_MAX+1]; | |
2810 | struct net_device *dev; | |
2811 | int err; | |
2812 | ||
2813 | memset(&tb, 0, sizeof(tb)); | |
2814 | ||
2815 | dev = rtnl_create_link(net, name, name_assign_type, | |
2816 | &vxlan_link_ops, tb); | |
2817 | if (IS_ERR(dev)) | |
2818 | return dev; | |
2819 | ||
2820 | err = vxlan_dev_configure(net, dev, conf); | |
2821 | if (err < 0) { | |
2822 | free_netdev(dev); | |
2823 | return ERR_PTR(err); | |
2824 | } | |
2825 | ||
2826 | return dev; | |
2827 | } | |
2828 | EXPORT_SYMBOL_GPL(vxlan_dev_create); | |
2829 | ||
2830 | static int vxlan_newlink(struct net *src_net, struct net_device *dev, | |
2831 | struct nlattr *tb[], struct nlattr *data[]) | |
2832 | { | |
2833 | struct vxlan_config conf; | |
2834 | int err; | |
2835 | ||
0dfbdf41 | 2836 | memset(&conf, 0, sizeof(conf)); |
e277de5f JG |
2837 | |
2838 | if (data[IFLA_VXLAN_ID]) | |
54bfd872 | 2839 | conf.vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID])); |
0dfbdf41 TG |
2840 | |
2841 | if (data[IFLA_VXLAN_GROUP]) { | |
2842 | conf.remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]); | |
2843 | } else if (data[IFLA_VXLAN_GROUP6]) { | |
2844 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2845 | return -EPFNOSUPPORT; | |
2846 | ||
2847 | conf.remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]); | |
2848 | conf.remote_ip.sa.sa_family = AF_INET6; | |
2849 | } | |
2850 | ||
2851 | if (data[IFLA_VXLAN_LOCAL]) { | |
2852 | conf.saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]); | |
2853 | conf.saddr.sa.sa_family = AF_INET; | |
2854 | } else if (data[IFLA_VXLAN_LOCAL6]) { | |
2855 | if (!IS_ENABLED(CONFIG_IPV6)) | |
2856 | return -EPFNOSUPPORT; | |
2857 | ||
2858 | /* TODO: respect scope id */ | |
2859 | conf.saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]); | |
2860 | conf.saddr.sa.sa_family = AF_INET6; | |
2861 | } | |
2862 | ||
2863 | if (data[IFLA_VXLAN_LINK]) | |
2864 | conf.remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]); | |
2865 | ||
d342894c | 2866 | if (data[IFLA_VXLAN_TOS]) |
0dfbdf41 | 2867 | conf.tos = nla_get_u8(data[IFLA_VXLAN_TOS]); |
d342894c | 2868 | |
afb97186 | 2869 | if (data[IFLA_VXLAN_TTL]) |
0dfbdf41 | 2870 | conf.ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); |
afb97186 | 2871 | |
d342894c | 2872 | if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING])) |
0dfbdf41 | 2873 | conf.flags |= VXLAN_F_LEARN; |
d342894c | 2874 | |
2875 | if (data[IFLA_VXLAN_AGEING]) | |
0dfbdf41 | 2876 | conf.age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); |
d342894c | 2877 | |
e4f67add | 2878 | if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY])) |
0dfbdf41 | 2879 | conf.flags |= VXLAN_F_PROXY; |
e4f67add DS |
2880 | |
2881 | if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC])) | |
0dfbdf41 | 2882 | conf.flags |= VXLAN_F_RSC; |
e4f67add DS |
2883 | |
2884 | if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS])) | |
0dfbdf41 | 2885 | conf.flags |= VXLAN_F_L2MISS; |
e4f67add DS |
2886 | |
2887 | if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS])) | |
0dfbdf41 | 2888 | conf.flags |= VXLAN_F_L3MISS; |
e4f67add | 2889 | |
d342894c | 2890 | if (data[IFLA_VXLAN_LIMIT]) |
0dfbdf41 | 2891 | conf.addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); |
d342894c | 2892 | |
f8a9b1bc AS |
2893 | if (data[IFLA_VXLAN_COLLECT_METADATA] && |
2894 | nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA])) | |
2895 | conf.flags |= VXLAN_F_COLLECT_METADATA; | |
2896 | ||
05f47d69 | 2897 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
2898 | const struct ifla_vxlan_port_range *p | |
2899 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
0dfbdf41 TG |
2900 | conf.port_min = ntohs(p->low); |
2901 | conf.port_max = ntohs(p->high); | |
05f47d69 | 2902 | } |
2903 | ||
823aa873 | 2904 | if (data[IFLA_VXLAN_PORT]) |
0dfbdf41 | 2905 | conf.dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); |
823aa873 | 2906 | |
359a0ea9 | 2907 | if (data[IFLA_VXLAN_UDP_CSUM] && nla_get_u8(data[IFLA_VXLAN_UDP_CSUM])) |
0dfbdf41 | 2908 | conf.flags |= VXLAN_F_UDP_CSUM; |
359a0ea9 TH |
2909 | |
2910 | if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX] && | |
2911 | nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX])) | |
0dfbdf41 | 2912 | conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_TX; |
359a0ea9 TH |
2913 | |
2914 | if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX] && | |
2915 | nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX])) | |
0dfbdf41 | 2916 | conf.flags |= VXLAN_F_UDP_ZERO_CSUM6_RX; |
359a0ea9 | 2917 | |
dfd8645e TH |
2918 | if (data[IFLA_VXLAN_REMCSUM_TX] && |
2919 | nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX])) | |
0dfbdf41 | 2920 | conf.flags |= VXLAN_F_REMCSUM_TX; |
dfd8645e TH |
2921 | |
2922 | if (data[IFLA_VXLAN_REMCSUM_RX] && | |
2923 | nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX])) | |
0dfbdf41 | 2924 | conf.flags |= VXLAN_F_REMCSUM_RX; |
dfd8645e | 2925 | |
3511494c | 2926 | if (data[IFLA_VXLAN_GBP]) |
0dfbdf41 | 2927 | conf.flags |= VXLAN_F_GBP; |
3511494c | 2928 | |
0ace2ca8 | 2929 | if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) |
0dfbdf41 | 2930 | conf.flags |= VXLAN_F_REMCSUM_NOPARTIAL; |
0ace2ca8 | 2931 | |
0dfbdf41 TG |
2932 | err = vxlan_dev_configure(src_net, dev, &conf); |
2933 | switch (err) { | |
2934 | case -ENODEV: | |
2935 | pr_info("ifindex %d does not exist\n", conf.remote_ifindex); | |
2936 | break; | |
1b13c97f | 2937 | |
0dfbdf41 TG |
2938 | case -EPERM: |
2939 | pr_info("IPv6 is disabled via sysctl\n"); | |
2940 | break; | |
d342894c | 2941 | |
0dfbdf41 | 2942 | case -EEXIST: |
54bfd872 | 2943 | pr_info("duplicate VNI %u\n", be32_to_cpu(conf.vni)); |
0dfbdf41 | 2944 | break; |
afbd8bae MR |
2945 | } |
2946 | ||
0dfbdf41 | 2947 | return err; |
d342894c | 2948 | } |
2949 | ||
2950 | static void vxlan_dellink(struct net_device *dev, struct list_head *head) | |
2951 | { | |
2952 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
f01ec1c0 | 2953 | struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id); |
d342894c | 2954 | |
fe5c3561 | 2955 | spin_lock(&vn->sock_lock); |
9c2e24e1 PS |
2956 | if (!hlist_unhashed(&vxlan->hlist)) |
2957 | hlist_del_rcu(&vxlan->hlist); | |
fe5c3561 | 2958 | spin_unlock(&vn->sock_lock); |
2959 | ||
58ce31cc | 2960 | gro_cells_destroy(&vxlan->gro_cells); |
553675fb | 2961 | list_del(&vxlan->next); |
d342894c | 2962 | unregister_netdevice_queue(dev, head); |
2963 | } | |
2964 | ||
2965 | static size_t vxlan_get_size(const struct net_device *dev) | |
2966 | { | |
2967 | ||
2968 | return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ | |
e4c7ed41 | 2969 | nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */ |
d342894c | 2970 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ |
e4c7ed41 | 2971 | nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */ |
d342894c | 2972 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ |
2973 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ | |
2974 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ | |
e4f67add DS |
2975 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ |
2976 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ | |
2977 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ | |
2978 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ | |
da8b43c0 | 2979 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */ |
d342894c | 2980 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ |
2981 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ | |
05f47d69 | 2982 | nla_total_size(sizeof(struct ifla_vxlan_port_range)) + |
359a0ea9 TH |
2983 | nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */ |
2984 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */ | |
2985 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */ | |
2986 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */ | |
dfd8645e TH |
2987 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */ |
2988 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */ | |
d342894c | 2989 | 0; |
2990 | } | |
2991 | ||
2992 | static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
2993 | { | |
2994 | const struct vxlan_dev *vxlan = netdev_priv(dev); | |
c7995c43 | 2995 | const struct vxlan_rdst *dst = &vxlan->default_dst; |
05f47d69 | 2996 | struct ifla_vxlan_port_range ports = { |
0dfbdf41 TG |
2997 | .low = htons(vxlan->cfg.port_min), |
2998 | .high = htons(vxlan->cfg.port_max), | |
05f47d69 | 2999 | }; |
d342894c | 3000 | |
54bfd872 | 3001 | if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni))) |
d342894c | 3002 | goto nla_put_failure; |
3003 | ||
e4c7ed41 CW |
3004 | if (!vxlan_addr_any(&dst->remote_ip)) { |
3005 | if (dst->remote_ip.sa.sa_family == AF_INET) { | |
930345ea JB |
3006 | if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP, |
3007 | dst->remote_ip.sin.sin_addr.s_addr)) | |
e4c7ed41 CW |
3008 | goto nla_put_failure; |
3009 | #if IS_ENABLED(CONFIG_IPV6) | |
3010 | } else { | |
930345ea JB |
3011 | if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6, |
3012 | &dst->remote_ip.sin6.sin6_addr)) | |
e4c7ed41 CW |
3013 | goto nla_put_failure; |
3014 | #endif | |
3015 | } | |
3016 | } | |
d342894c | 3017 | |
c7995c43 | 3018 | if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) |
d342894c | 3019 | goto nla_put_failure; |
3020 | ||
0dfbdf41 TG |
3021 | if (!vxlan_addr_any(&vxlan->cfg.saddr)) { |
3022 | if (vxlan->cfg.saddr.sa.sa_family == AF_INET) { | |
930345ea | 3023 | if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL, |
0dfbdf41 | 3024 | vxlan->cfg.saddr.sin.sin_addr.s_addr)) |
e4c7ed41 CW |
3025 | goto nla_put_failure; |
3026 | #if IS_ENABLED(CONFIG_IPV6) | |
3027 | } else { | |
930345ea | 3028 | if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6, |
0dfbdf41 | 3029 | &vxlan->cfg.saddr.sin6.sin6_addr)) |
e4c7ed41 CW |
3030 | goto nla_put_failure; |
3031 | #endif | |
3032 | } | |
3033 | } | |
d342894c | 3034 | |
0dfbdf41 TG |
3035 | if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) || |
3036 | nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) || | |
e4f67add DS |
3037 | nla_put_u8(skb, IFLA_VXLAN_LEARNING, |
3038 | !!(vxlan->flags & VXLAN_F_LEARN)) || | |
3039 | nla_put_u8(skb, IFLA_VXLAN_PROXY, | |
3040 | !!(vxlan->flags & VXLAN_F_PROXY)) || | |
3041 | nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) || | |
3042 | nla_put_u8(skb, IFLA_VXLAN_L2MISS, | |
3043 | !!(vxlan->flags & VXLAN_F_L2MISS)) || | |
3044 | nla_put_u8(skb, IFLA_VXLAN_L3MISS, | |
3045 | !!(vxlan->flags & VXLAN_F_L3MISS)) || | |
da8b43c0 AS |
3046 | nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA, |
3047 | !!(vxlan->flags & VXLAN_F_COLLECT_METADATA)) || | |
0dfbdf41 TG |
3048 | nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) || |
3049 | nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) || | |
3050 | nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) || | |
359a0ea9 TH |
3051 | nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM, |
3052 | !!(vxlan->flags & VXLAN_F_UDP_CSUM)) || | |
3053 | nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, | |
3054 | !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) || | |
3055 | nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, | |
dfd8645e TH |
3056 | !!(vxlan->flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) || |
3057 | nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX, | |
3058 | !!(vxlan->flags & VXLAN_F_REMCSUM_TX)) || | |
3059 | nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX, | |
3060 | !!(vxlan->flags & VXLAN_F_REMCSUM_RX))) | |
d342894c | 3061 | goto nla_put_failure; |
3062 | ||
05f47d69 | 3063 | if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) |
3064 | goto nla_put_failure; | |
3065 | ||
3511494c TG |
3066 | if (vxlan->flags & VXLAN_F_GBP && |
3067 | nla_put_flag(skb, IFLA_VXLAN_GBP)) | |
3068 | goto nla_put_failure; | |
3069 | ||
0ace2ca8 TH |
3070 | if (vxlan->flags & VXLAN_F_REMCSUM_NOPARTIAL && |
3071 | nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL)) | |
3072 | goto nla_put_failure; | |
3073 | ||
d342894c | 3074 | return 0; |
3075 | ||
3076 | nla_put_failure: | |
3077 | return -EMSGSIZE; | |
3078 | } | |
3079 | ||
1728d4fa ND |
3080 | static struct net *vxlan_get_link_net(const struct net_device *dev) |
3081 | { | |
3082 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
3083 | ||
3084 | return vxlan->net; | |
3085 | } | |
3086 | ||
d342894c | 3087 | static struct rtnl_link_ops vxlan_link_ops __read_mostly = { |
3088 | .kind = "vxlan", | |
3089 | .maxtype = IFLA_VXLAN_MAX, | |
3090 | .policy = vxlan_policy, | |
3091 | .priv_size = sizeof(struct vxlan_dev), | |
3092 | .setup = vxlan_setup, | |
3093 | .validate = vxlan_validate, | |
3094 | .newlink = vxlan_newlink, | |
3095 | .dellink = vxlan_dellink, | |
3096 | .get_size = vxlan_get_size, | |
3097 | .fill_info = vxlan_fill_info, | |
1728d4fa | 3098 | .get_link_net = vxlan_get_link_net, |
d342894c | 3099 | }; |
3100 | ||
acaf4e70 DB |
3101 | static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn, |
3102 | struct net_device *dev) | |
3103 | { | |
3104 | struct vxlan_dev *vxlan, *next; | |
3105 | LIST_HEAD(list_kill); | |
3106 | ||
3107 | list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { | |
3108 | struct vxlan_rdst *dst = &vxlan->default_dst; | |
3109 | ||
3110 | /* In case we created vxlan device with carrier | |
3111 | * and we loose the carrier due to module unload | |
3112 | * we also need to remove vxlan device. In other | |
3113 | * cases, it's not necessary and remote_ifindex | |
3114 | * is 0 here, so no matches. | |
3115 | */ | |
3116 | if (dst->remote_ifindex == dev->ifindex) | |
3117 | vxlan_dellink(vxlan->dev, &list_kill); | |
3118 | } | |
3119 | ||
3120 | unregister_netdevice_many(&list_kill); | |
3121 | } | |
3122 | ||
3123 | static int vxlan_lowerdev_event(struct notifier_block *unused, | |
3124 | unsigned long event, void *ptr) | |
3125 | { | |
3126 | struct net_device *dev = netdev_notifier_info_to_dev(ptr); | |
783c1463 | 3127 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); |
acaf4e70 | 3128 | |
783c1463 | 3129 | if (event == NETDEV_UNREGISTER) |
acaf4e70 DB |
3130 | vxlan_handle_lowerdev_unregister(vn, dev); |
3131 | ||
3132 | return NOTIFY_DONE; | |
3133 | } | |
3134 | ||
3135 | static struct notifier_block vxlan_notifier_block __read_mostly = { | |
3136 | .notifier_call = vxlan_lowerdev_event, | |
3137 | }; | |
3138 | ||
d342894c | 3139 | static __net_init int vxlan_init_net(struct net *net) |
3140 | { | |
3141 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
31fec5aa | 3142 | unsigned int h; |
d342894c | 3143 | |
553675fb | 3144 | INIT_LIST_HEAD(&vn->vxlan_list); |
1c51a915 | 3145 | spin_lock_init(&vn->sock_lock); |
d342894c | 3146 | |
553675fb | 3147 | for (h = 0; h < PORT_HASH_SIZE; ++h) |
3148 | INIT_HLIST_HEAD(&vn->sock_list[h]); | |
d342894c | 3149 | |
3150 | return 0; | |
3151 | } | |
3152 | ||
f01ec1c0 ND |
3153 | static void __net_exit vxlan_exit_net(struct net *net) |
3154 | { | |
3155 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
3156 | struct vxlan_dev *vxlan, *next; | |
3157 | struct net_device *dev, *aux; | |
3158 | LIST_HEAD(list); | |
3159 | ||
3160 | rtnl_lock(); | |
3161 | for_each_netdev_safe(net, dev, aux) | |
3162 | if (dev->rtnl_link_ops == &vxlan_link_ops) | |
3163 | unregister_netdevice_queue(dev, &list); | |
3164 | ||
3165 | list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) { | |
3166 | /* If vxlan->dev is in the same netns, it has already been added | |
3167 | * to the list by the previous loop. | |
3168 | */ | |
58ce31cc TH |
3169 | if (!net_eq(dev_net(vxlan->dev), net)) { |
3170 | gro_cells_destroy(&vxlan->gro_cells); | |
13c3ed6a | 3171 | unregister_netdevice_queue(vxlan->dev, &list); |
58ce31cc | 3172 | } |
f01ec1c0 ND |
3173 | } |
3174 | ||
3175 | unregister_netdevice_many(&list); | |
3176 | rtnl_unlock(); | |
3177 | } | |
3178 | ||
d342894c | 3179 | static struct pernet_operations vxlan_net_ops = { |
3180 | .init = vxlan_init_net, | |
f01ec1c0 | 3181 | .exit = vxlan_exit_net, |
d342894c | 3182 | .id = &vxlan_net_id, |
3183 | .size = sizeof(struct vxlan_net), | |
3184 | }; | |
3185 | ||
3186 | static int __init vxlan_init_module(void) | |
3187 | { | |
3188 | int rc; | |
3189 | ||
758c57d1 SH |
3190 | vxlan_wq = alloc_workqueue("vxlan", 0, 0); |
3191 | if (!vxlan_wq) | |
3192 | return -ENOMEM; | |
3193 | ||
d342894c | 3194 | get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); |
3195 | ||
783c1463 | 3196 | rc = register_pernet_subsys(&vxlan_net_ops); |
d342894c | 3197 | if (rc) |
3198 | goto out1; | |
3199 | ||
acaf4e70 | 3200 | rc = register_netdevice_notifier(&vxlan_notifier_block); |
d342894c | 3201 | if (rc) |
3202 | goto out2; | |
3203 | ||
acaf4e70 DB |
3204 | rc = rtnl_link_register(&vxlan_link_ops); |
3205 | if (rc) | |
3206 | goto out3; | |
d342894c | 3207 | |
acaf4e70 DB |
3208 | return 0; |
3209 | out3: | |
3210 | unregister_netdevice_notifier(&vxlan_notifier_block); | |
d342894c | 3211 | out2: |
783c1463 | 3212 | unregister_pernet_subsys(&vxlan_net_ops); |
d342894c | 3213 | out1: |
758c57d1 | 3214 | destroy_workqueue(vxlan_wq); |
d342894c | 3215 | return rc; |
3216 | } | |
7332a13b | 3217 | late_initcall(vxlan_init_module); |
d342894c | 3218 | |
3219 | static void __exit vxlan_cleanup_module(void) | |
3220 | { | |
b7153984 | 3221 | rtnl_link_unregister(&vxlan_link_ops); |
acaf4e70 | 3222 | unregister_netdevice_notifier(&vxlan_notifier_block); |
758c57d1 | 3223 | destroy_workqueue(vxlan_wq); |
783c1463 DB |
3224 | unregister_pernet_subsys(&vxlan_net_ops); |
3225 | /* rcu_barrier() is called by netns */ | |
d342894c | 3226 | } |
3227 | module_exit(vxlan_cleanup_module); | |
3228 | ||
3229 | MODULE_LICENSE("GPL"); | |
3230 | MODULE_VERSION(VXLAN_VERSION); | |
3b8df3c6 | 3231 | MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); |
ead5139a | 3232 | MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic"); |
d342894c | 3233 | MODULE_ALIAS_RTNL_LINK("vxlan"); |