]>
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. | |
9 | * | |
10 | * TODO | |
d342894c | 11 | * - IPv6 (not in RFC) |
12 | */ | |
13 | ||
14 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
15 | ||
16 | #include <linux/kernel.h> | |
17 | #include <linux/types.h> | |
18 | #include <linux/module.h> | |
19 | #include <linux/errno.h> | |
20 | #include <linux/slab.h> | |
21 | #include <linux/skbuff.h> | |
22 | #include <linux/rculist.h> | |
23 | #include <linux/netdevice.h> | |
24 | #include <linux/in.h> | |
25 | #include <linux/ip.h> | |
26 | #include <linux/udp.h> | |
27 | #include <linux/igmp.h> | |
28 | #include <linux/etherdevice.h> | |
29 | #include <linux/if_ether.h> | |
d342894c | 30 | #include <linux/hash.h> |
1b13c97f | 31 | #include <linux/ethtool.h> |
e4f67add DS |
32 | #include <net/arp.h> |
33 | #include <net/ndisc.h> | |
d342894c | 34 | #include <net/ip.h> |
c5441932 | 35 | #include <net/ip_tunnels.h> |
d342894c | 36 | #include <net/icmp.h> |
37 | #include <net/udp.h> | |
38 | #include <net/rtnetlink.h> | |
39 | #include <net/route.h> | |
40 | #include <net/dsfield.h> | |
41 | #include <net/inet_ecn.h> | |
42 | #include <net/net_namespace.h> | |
43 | #include <net/netns/generic.h> | |
44 | ||
45 | #define VXLAN_VERSION "0.1" | |
46 | ||
47 | #define VNI_HASH_BITS 10 | |
48 | #define VNI_HASH_SIZE (1<<VNI_HASH_BITS) | |
49 | #define FDB_HASH_BITS 8 | |
50 | #define FDB_HASH_SIZE (1<<FDB_HASH_BITS) | |
51 | #define FDB_AGE_DEFAULT 300 /* 5 min */ | |
52 | #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */ | |
53 | ||
54 | #define VXLAN_N_VID (1u << 24) | |
55 | #define VXLAN_VID_MASK (VXLAN_N_VID - 1) | |
52b702ff AD |
56 | /* IP header + UDP + VXLAN + Ethernet header */ |
57 | #define VXLAN_HEADROOM (20 + 8 + 8 + 14) | |
d342894c | 58 | |
59 | #define VXLAN_FLAGS 0x08000000 /* struct vxlanhdr.vx_flags required value. */ | |
60 | ||
61 | /* VXLAN protocol header */ | |
62 | struct vxlanhdr { | |
63 | __be32 vx_flags; | |
64 | __be32 vx_vni; | |
65 | }; | |
66 | ||
23c578bf | 67 | /* UDP port for VXLAN traffic. |
68 | * The IANA assigned port is 4789, but the Linux default is 8472 | |
69 | * for compatability with early adopters. | |
70 | */ | |
d342894c | 71 | static unsigned int vxlan_port __read_mostly = 8472; |
72 | module_param_named(udp_port, vxlan_port, uint, 0444); | |
73 | MODULE_PARM_DESC(udp_port, "Destination UDP port"); | |
74 | ||
75 | static bool log_ecn_error = true; | |
76 | module_param(log_ecn_error, bool, 0644); | |
77 | MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN"); | |
78 | ||
79 | /* per-net private data for this module */ | |
80 | static unsigned int vxlan_net_id; | |
81 | struct vxlan_net { | |
82 | struct socket *sock; /* UDP encap socket */ | |
83 | struct hlist_head vni_list[VNI_HASH_SIZE]; | |
84 | }; | |
85 | ||
6681712d DS |
86 | struct vxlan_rdst { |
87 | struct rcu_head rcu; | |
88 | __be32 remote_ip; | |
89 | __be16 remote_port; | |
90 | u32 remote_vni; | |
91 | u32 remote_ifindex; | |
92 | struct vxlan_rdst *remote_next; | |
93 | }; | |
94 | ||
d342894c | 95 | /* Forwarding table entry */ |
96 | struct vxlan_fdb { | |
97 | struct hlist_node hlist; /* linked list of entries */ | |
98 | struct rcu_head rcu; | |
99 | unsigned long updated; /* jiffies */ | |
100 | unsigned long used; | |
6681712d | 101 | struct vxlan_rdst remote; |
d342894c | 102 | u16 state; /* see ndm_state */ |
ae884082 | 103 | u8 flags; /* see ndm_flags */ |
d342894c | 104 | u8 eth_addr[ETH_ALEN]; |
105 | }; | |
106 | ||
d342894c | 107 | /* Pseudo network device */ |
108 | struct vxlan_dev { | |
109 | struct hlist_node hlist; | |
110 | struct net_device *dev; | |
c7995c43 | 111 | struct vxlan_rdst default_dst; /* default destination */ |
d342894c | 112 | __be32 saddr; /* source address */ |
823aa873 | 113 | __be16 dst_port; |
05f47d69 | 114 | __u16 port_min; /* source port range */ |
115 | __u16 port_max; | |
d342894c | 116 | __u8 tos; /* TOS override */ |
117 | __u8 ttl; | |
e4f67add | 118 | u32 flags; /* VXLAN_F_* below */ |
d342894c | 119 | |
120 | unsigned long age_interval; | |
121 | struct timer_list age_timer; | |
122 | spinlock_t hash_lock; | |
123 | unsigned int addrcnt; | |
124 | unsigned int addrmax; | |
d342894c | 125 | |
126 | struct hlist_head fdb_head[FDB_HASH_SIZE]; | |
127 | }; | |
128 | ||
e4f67add DS |
129 | #define VXLAN_F_LEARN 0x01 |
130 | #define VXLAN_F_PROXY 0x02 | |
131 | #define VXLAN_F_RSC 0x04 | |
132 | #define VXLAN_F_L2MISS 0x08 | |
133 | #define VXLAN_F_L3MISS 0x10 | |
134 | ||
d342894c | 135 | /* salt for hash table */ |
136 | static u32 vxlan_salt __read_mostly; | |
137 | ||
138 | static inline struct hlist_head *vni_head(struct net *net, u32 id) | |
139 | { | |
140 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
141 | ||
142 | return &vn->vni_list[hash_32(id, VNI_HASH_BITS)]; | |
143 | } | |
144 | ||
145 | /* Look up VNI in a per net namespace table */ | |
146 | static struct vxlan_dev *vxlan_find_vni(struct net *net, u32 id) | |
147 | { | |
148 | struct vxlan_dev *vxlan; | |
d342894c | 149 | |
b67bfe0d | 150 | hlist_for_each_entry_rcu(vxlan, vni_head(net, id), hlist) { |
c7995c43 | 151 | if (vxlan->default_dst.remote_vni == id) |
d342894c | 152 | return vxlan; |
153 | } | |
154 | ||
155 | return NULL; | |
156 | } | |
157 | ||
158 | /* Fill in neighbour message in skbuff. */ | |
159 | static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan, | |
160 | const struct vxlan_fdb *fdb, | |
6681712d DS |
161 | u32 portid, u32 seq, int type, unsigned int flags, |
162 | const struct vxlan_rdst *rdst) | |
d342894c | 163 | { |
164 | unsigned long now = jiffies; | |
165 | struct nda_cacheinfo ci; | |
166 | struct nlmsghdr *nlh; | |
167 | struct ndmsg *ndm; | |
e4f67add | 168 | bool send_ip, send_eth; |
d342894c | 169 | |
170 | nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags); | |
171 | if (nlh == NULL) | |
172 | return -EMSGSIZE; | |
173 | ||
174 | ndm = nlmsg_data(nlh); | |
175 | memset(ndm, 0, sizeof(*ndm)); | |
e4f67add DS |
176 | |
177 | send_eth = send_ip = true; | |
178 | ||
179 | if (type == RTM_GETNEIGH) { | |
180 | ndm->ndm_family = AF_INET; | |
6681712d | 181 | send_ip = rdst->remote_ip != htonl(INADDR_ANY); |
e4f67add DS |
182 | send_eth = !is_zero_ether_addr(fdb->eth_addr); |
183 | } else | |
184 | ndm->ndm_family = AF_BRIDGE; | |
d342894c | 185 | ndm->ndm_state = fdb->state; |
186 | ndm->ndm_ifindex = vxlan->dev->ifindex; | |
ae884082 | 187 | ndm->ndm_flags = fdb->flags; |
d342894c | 188 | ndm->ndm_type = NDA_DST; |
189 | ||
e4f67add | 190 | if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr)) |
d342894c | 191 | goto nla_put_failure; |
192 | ||
6681712d DS |
193 | if (send_ip && nla_put_be32(skb, NDA_DST, rdst->remote_ip)) |
194 | goto nla_put_failure; | |
195 | ||
823aa873 | 196 | if (rdst->remote_port && rdst->remote_port != vxlan->dst_port && |
6681712d DS |
197 | nla_put_be16(skb, NDA_PORT, rdst->remote_port)) |
198 | goto nla_put_failure; | |
c7995c43 | 199 | if (rdst->remote_vni != vxlan->default_dst.remote_vni && |
6681712d DS |
200 | nla_put_be32(skb, NDA_VNI, rdst->remote_vni)) |
201 | goto nla_put_failure; | |
202 | if (rdst->remote_ifindex && | |
203 | nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex)) | |
d342894c | 204 | goto nla_put_failure; |
205 | ||
206 | ci.ndm_used = jiffies_to_clock_t(now - fdb->used); | |
207 | ci.ndm_confirmed = 0; | |
208 | ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated); | |
209 | ci.ndm_refcnt = 0; | |
210 | ||
211 | if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci)) | |
212 | goto nla_put_failure; | |
213 | ||
214 | return nlmsg_end(skb, nlh); | |
215 | ||
216 | nla_put_failure: | |
217 | nlmsg_cancel(skb, nlh); | |
218 | return -EMSGSIZE; | |
219 | } | |
220 | ||
221 | static inline size_t vxlan_nlmsg_size(void) | |
222 | { | |
223 | return NLMSG_ALIGN(sizeof(struct ndmsg)) | |
224 | + nla_total_size(ETH_ALEN) /* NDA_LLADDR */ | |
225 | + nla_total_size(sizeof(__be32)) /* NDA_DST */ | |
73cf3317 | 226 | + nla_total_size(sizeof(__be16)) /* NDA_PORT */ |
6681712d DS |
227 | + nla_total_size(sizeof(__be32)) /* NDA_VNI */ |
228 | + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */ | |
d342894c | 229 | + nla_total_size(sizeof(struct nda_cacheinfo)); |
230 | } | |
231 | ||
232 | static void vxlan_fdb_notify(struct vxlan_dev *vxlan, | |
233 | const struct vxlan_fdb *fdb, int type) | |
234 | { | |
235 | struct net *net = dev_net(vxlan->dev); | |
236 | struct sk_buff *skb; | |
237 | int err = -ENOBUFS; | |
238 | ||
239 | skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC); | |
240 | if (skb == NULL) | |
241 | goto errout; | |
242 | ||
6681712d | 243 | err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, &fdb->remote); |
d342894c | 244 | if (err < 0) { |
245 | /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */ | |
246 | WARN_ON(err == -EMSGSIZE); | |
247 | kfree_skb(skb); | |
248 | goto errout; | |
249 | } | |
250 | ||
251 | rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC); | |
252 | return; | |
253 | errout: | |
254 | if (err < 0) | |
255 | rtnl_set_sk_err(net, RTNLGRP_NEIGH, err); | |
256 | } | |
257 | ||
e4f67add DS |
258 | static void vxlan_ip_miss(struct net_device *dev, __be32 ipa) |
259 | { | |
260 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
261 | struct vxlan_fdb f; | |
262 | ||
263 | memset(&f, 0, sizeof f); | |
264 | f.state = NUD_STALE; | |
6681712d DS |
265 | f.remote.remote_ip = ipa; /* goes to NDA_DST */ |
266 | f.remote.remote_vni = VXLAN_N_VID; | |
e4f67add DS |
267 | |
268 | vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH); | |
269 | } | |
270 | ||
271 | static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN]) | |
272 | { | |
273 | struct vxlan_fdb f; | |
274 | ||
275 | memset(&f, 0, sizeof f); | |
276 | f.state = NUD_STALE; | |
277 | memcpy(f.eth_addr, eth_addr, ETH_ALEN); | |
278 | ||
279 | vxlan_fdb_notify(vxlan, &f, RTM_GETNEIGH); | |
280 | } | |
281 | ||
d342894c | 282 | /* Hash Ethernet address */ |
283 | static u32 eth_hash(const unsigned char *addr) | |
284 | { | |
285 | u64 value = get_unaligned((u64 *)addr); | |
286 | ||
287 | /* only want 6 bytes */ | |
288 | #ifdef __BIG_ENDIAN | |
d342894c | 289 | value >>= 16; |
321fb991 | 290 | #else |
291 | value <<= 16; | |
d342894c | 292 | #endif |
293 | return hash_64(value, FDB_HASH_BITS); | |
294 | } | |
295 | ||
296 | /* Hash chain to use given mac address */ | |
297 | static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan, | |
298 | const u8 *mac) | |
299 | { | |
300 | return &vxlan->fdb_head[eth_hash(mac)]; | |
301 | } | |
302 | ||
303 | /* Look up Ethernet address in forwarding table */ | |
014be2c8 | 304 | static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan, |
d342894c | 305 | const u8 *mac) |
306 | ||
307 | { | |
308 | struct hlist_head *head = vxlan_fdb_head(vxlan, mac); | |
309 | struct vxlan_fdb *f; | |
d342894c | 310 | |
b67bfe0d | 311 | hlist_for_each_entry_rcu(f, head, hlist) { |
d342894c | 312 | if (compare_ether_addr(mac, f->eth_addr) == 0) |
313 | return f; | |
314 | } | |
315 | ||
316 | return NULL; | |
317 | } | |
318 | ||
014be2c8 SS |
319 | static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan, |
320 | const u8 *mac) | |
321 | { | |
322 | struct vxlan_fdb *f; | |
323 | ||
324 | f = __vxlan_find_mac(vxlan, mac); | |
325 | if (f) | |
326 | f->used = jiffies; | |
327 | ||
328 | return f; | |
329 | } | |
330 | ||
6681712d DS |
331 | /* Add/update destinations for multicast */ |
332 | static int vxlan_fdb_append(struct vxlan_fdb *f, | |
73cf3317 | 333 | __be32 ip, __be16 port, __u32 vni, __u32 ifindex) |
6681712d DS |
334 | { |
335 | struct vxlan_rdst *rd_prev, *rd; | |
336 | ||
337 | rd_prev = NULL; | |
338 | for (rd = &f->remote; rd; rd = rd->remote_next) { | |
339 | if (rd->remote_ip == ip && | |
340 | rd->remote_port == port && | |
341 | rd->remote_vni == vni && | |
342 | rd->remote_ifindex == ifindex) | |
343 | return 0; | |
344 | rd_prev = rd; | |
345 | } | |
346 | rd = kmalloc(sizeof(*rd), GFP_ATOMIC); | |
347 | if (rd == NULL) | |
348 | return -ENOBUFS; | |
349 | rd->remote_ip = ip; | |
350 | rd->remote_port = port; | |
351 | rd->remote_vni = vni; | |
352 | rd->remote_ifindex = ifindex; | |
353 | rd->remote_next = NULL; | |
354 | rd_prev->remote_next = rd; | |
355 | return 1; | |
356 | } | |
357 | ||
d342894c | 358 | /* Add new entry to forwarding table -- assumes lock held */ |
359 | static int vxlan_fdb_create(struct vxlan_dev *vxlan, | |
360 | const u8 *mac, __be32 ip, | |
6681712d | 361 | __u16 state, __u16 flags, |
73cf3317 | 362 | __be16 port, __u32 vni, __u32 ifindex, |
ae884082 | 363 | __u8 ndm_flags) |
d342894c | 364 | { |
365 | struct vxlan_fdb *f; | |
366 | int notify = 0; | |
367 | ||
014be2c8 | 368 | f = __vxlan_find_mac(vxlan, mac); |
d342894c | 369 | if (f) { |
370 | if (flags & NLM_F_EXCL) { | |
371 | netdev_dbg(vxlan->dev, | |
372 | "lost race to create %pM\n", mac); | |
373 | return -EEXIST; | |
374 | } | |
375 | if (f->state != state) { | |
376 | f->state = state; | |
377 | f->updated = jiffies; | |
378 | notify = 1; | |
379 | } | |
ae884082 DS |
380 | if (f->flags != ndm_flags) { |
381 | f->flags = ndm_flags; | |
382 | f->updated = jiffies; | |
383 | notify = 1; | |
384 | } | |
6681712d DS |
385 | if ((flags & NLM_F_APPEND) && |
386 | is_multicast_ether_addr(f->eth_addr)) { | |
387 | int rc = vxlan_fdb_append(f, ip, port, vni, ifindex); | |
388 | ||
389 | if (rc < 0) | |
390 | return rc; | |
391 | notify |= rc; | |
392 | } | |
d342894c | 393 | } else { |
394 | if (!(flags & NLM_F_CREATE)) | |
395 | return -ENOENT; | |
396 | ||
397 | if (vxlan->addrmax && vxlan->addrcnt >= vxlan->addrmax) | |
398 | return -ENOSPC; | |
399 | ||
400 | netdev_dbg(vxlan->dev, "add %pM -> %pI4\n", mac, &ip); | |
401 | f = kmalloc(sizeof(*f), GFP_ATOMIC); | |
402 | if (!f) | |
403 | return -ENOMEM; | |
404 | ||
405 | notify = 1; | |
6681712d DS |
406 | f->remote.remote_ip = ip; |
407 | f->remote.remote_port = port; | |
408 | f->remote.remote_vni = vni; | |
409 | f->remote.remote_ifindex = ifindex; | |
410 | f->remote.remote_next = NULL; | |
d342894c | 411 | f->state = state; |
ae884082 | 412 | f->flags = ndm_flags; |
d342894c | 413 | f->updated = f->used = jiffies; |
414 | memcpy(f->eth_addr, mac, ETH_ALEN); | |
415 | ||
416 | ++vxlan->addrcnt; | |
417 | hlist_add_head_rcu(&f->hlist, | |
418 | vxlan_fdb_head(vxlan, mac)); | |
419 | } | |
420 | ||
421 | if (notify) | |
422 | vxlan_fdb_notify(vxlan, f, RTM_NEWNEIGH); | |
423 | ||
424 | return 0; | |
425 | } | |
426 | ||
6706c82e | 427 | static void vxlan_fdb_free(struct rcu_head *head) |
6681712d DS |
428 | { |
429 | struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu); | |
430 | ||
431 | while (f->remote.remote_next) { | |
432 | struct vxlan_rdst *rd = f->remote.remote_next; | |
433 | ||
434 | f->remote.remote_next = rd->remote_next; | |
435 | kfree(rd); | |
436 | } | |
437 | kfree(f); | |
438 | } | |
439 | ||
d342894c | 440 | static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f) |
441 | { | |
442 | netdev_dbg(vxlan->dev, | |
443 | "delete %pM\n", f->eth_addr); | |
444 | ||
445 | --vxlan->addrcnt; | |
446 | vxlan_fdb_notify(vxlan, f, RTM_DELNEIGH); | |
447 | ||
448 | hlist_del_rcu(&f->hlist); | |
6681712d | 449 | call_rcu(&f->rcu, vxlan_fdb_free); |
d342894c | 450 | } |
451 | ||
452 | /* Add static entry (via netlink) */ | |
453 | static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[], | |
454 | struct net_device *dev, | |
455 | const unsigned char *addr, u16 flags) | |
456 | { | |
457 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
6681712d | 458 | struct net *net = dev_net(vxlan->dev); |
d342894c | 459 | __be32 ip; |
73cf3317 | 460 | __be16 port; |
461 | u32 vni, ifindex; | |
d342894c | 462 | int err; |
463 | ||
464 | if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) { | |
465 | pr_info("RTM_NEWNEIGH with invalid state %#x\n", | |
466 | ndm->ndm_state); | |
467 | return -EINVAL; | |
468 | } | |
469 | ||
470 | if (tb[NDA_DST] == NULL) | |
471 | return -EINVAL; | |
472 | ||
473 | if (nla_len(tb[NDA_DST]) != sizeof(__be32)) | |
474 | return -EAFNOSUPPORT; | |
475 | ||
476 | ip = nla_get_be32(tb[NDA_DST]); | |
477 | ||
6681712d | 478 | if (tb[NDA_PORT]) { |
73cf3317 | 479 | if (nla_len(tb[NDA_PORT]) != sizeof(__be16)) |
6681712d | 480 | return -EINVAL; |
73cf3317 | 481 | port = nla_get_be16(tb[NDA_PORT]); |
6681712d | 482 | } else |
823aa873 | 483 | port = vxlan->dst_port; |
6681712d DS |
484 | |
485 | if (tb[NDA_VNI]) { | |
486 | if (nla_len(tb[NDA_VNI]) != sizeof(u32)) | |
487 | return -EINVAL; | |
488 | vni = nla_get_u32(tb[NDA_VNI]); | |
489 | } else | |
c7995c43 | 490 | vni = vxlan->default_dst.remote_vni; |
6681712d DS |
491 | |
492 | if (tb[NDA_IFINDEX]) { | |
5abb0029 | 493 | struct net_device *tdev; |
6681712d DS |
494 | |
495 | if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32)) | |
496 | return -EINVAL; | |
497 | ifindex = nla_get_u32(tb[NDA_IFINDEX]); | |
5abb0029 PS |
498 | tdev = dev_get_by_index(net, ifindex); |
499 | if (!tdev) | |
6681712d | 500 | return -EADDRNOTAVAIL; |
5abb0029 | 501 | dev_put(tdev); |
6681712d DS |
502 | } else |
503 | ifindex = 0; | |
504 | ||
d342894c | 505 | spin_lock_bh(&vxlan->hash_lock); |
73cf3317 | 506 | err = vxlan_fdb_create(vxlan, addr, ip, ndm->ndm_state, flags, |
507 | port, vni, ifindex, ndm->ndm_flags); | |
d342894c | 508 | spin_unlock_bh(&vxlan->hash_lock); |
509 | ||
510 | return err; | |
511 | } | |
512 | ||
513 | /* Delete entry (via netlink) */ | |
1690be63 VY |
514 | static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[], |
515 | struct net_device *dev, | |
d342894c | 516 | const unsigned char *addr) |
517 | { | |
518 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
519 | struct vxlan_fdb *f; | |
520 | int err = -ENOENT; | |
521 | ||
522 | spin_lock_bh(&vxlan->hash_lock); | |
523 | f = vxlan_find_mac(vxlan, addr); | |
524 | if (f) { | |
525 | vxlan_fdb_destroy(vxlan, f); | |
526 | err = 0; | |
527 | } | |
528 | spin_unlock_bh(&vxlan->hash_lock); | |
529 | ||
530 | return err; | |
531 | } | |
532 | ||
533 | /* Dump forwarding table */ | |
534 | static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb, | |
535 | struct net_device *dev, int idx) | |
536 | { | |
537 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
538 | unsigned int h; | |
539 | ||
540 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
541 | struct vxlan_fdb *f; | |
d342894c | 542 | int err; |
543 | ||
b67bfe0d | 544 | hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) { |
6681712d DS |
545 | struct vxlan_rdst *rd; |
546 | for (rd = &f->remote; rd; rd = rd->remote_next) { | |
547 | if (idx < cb->args[0]) | |
548 | goto skip; | |
549 | ||
550 | err = vxlan_fdb_info(skb, vxlan, f, | |
551 | NETLINK_CB(cb->skb).portid, | |
552 | cb->nlh->nlmsg_seq, | |
553 | RTM_NEWNEIGH, | |
554 | NLM_F_MULTI, rd); | |
555 | if (err < 0) | |
556 | break; | |
d342894c | 557 | skip: |
6681712d DS |
558 | ++idx; |
559 | } | |
d342894c | 560 | } |
561 | } | |
562 | ||
563 | return idx; | |
564 | } | |
565 | ||
566 | /* Watch incoming packets to learn mapping between Ethernet address | |
567 | * and Tunnel endpoint. | |
26a41ae6 | 568 | * Return true if packet is bogus and should be droppped. |
d342894c | 569 | */ |
26a41ae6 | 570 | static bool vxlan_snoop(struct net_device *dev, |
d342894c | 571 | __be32 src_ip, const u8 *src_mac) |
572 | { | |
573 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
574 | struct vxlan_fdb *f; | |
d342894c | 575 | |
576 | f = vxlan_find_mac(vxlan, src_mac); | |
577 | if (likely(f)) { | |
6681712d | 578 | if (likely(f->remote.remote_ip == src_ip)) |
26a41ae6 | 579 | return false; |
580 | ||
581 | /* Don't migrate static entries, drop packets */ | |
582 | if (!(f->flags & NTF_SELF)) | |
583 | return true; | |
d342894c | 584 | |
585 | if (net_ratelimit()) | |
586 | netdev_info(dev, | |
587 | "%pM migrated from %pI4 to %pI4\n", | |
6681712d | 588 | src_mac, &f->remote.remote_ip, &src_ip); |
d342894c | 589 | |
6681712d | 590 | f->remote.remote_ip = src_ip; |
d342894c | 591 | f->updated = jiffies; |
592 | } else { | |
593 | /* learned new entry */ | |
594 | spin_lock(&vxlan->hash_lock); | |
3bf74b1a | 595 | |
596 | /* close off race between vxlan_flush and incoming packets */ | |
597 | if (netif_running(dev)) | |
598 | vxlan_fdb_create(vxlan, src_mac, src_ip, | |
599 | NUD_REACHABLE, | |
600 | NLM_F_EXCL|NLM_F_CREATE, | |
601 | vxlan->dst_port, | |
602 | vxlan->default_dst.remote_vni, | |
603 | 0, NTF_SELF); | |
d342894c | 604 | spin_unlock(&vxlan->hash_lock); |
605 | } | |
26a41ae6 | 606 | |
607 | return false; | |
d342894c | 608 | } |
609 | ||
610 | ||
611 | /* See if multicast group is already in use by other ID */ | |
612 | static bool vxlan_group_used(struct vxlan_net *vn, | |
613 | const struct vxlan_dev *this) | |
614 | { | |
615 | const struct vxlan_dev *vxlan; | |
d342894c | 616 | unsigned h; |
617 | ||
618 | for (h = 0; h < VNI_HASH_SIZE; ++h) | |
b67bfe0d | 619 | hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist) { |
d342894c | 620 | if (vxlan == this) |
621 | continue; | |
622 | ||
623 | if (!netif_running(vxlan->dev)) | |
624 | continue; | |
625 | ||
c7995c43 | 626 | if (vxlan->default_dst.remote_ip == this->default_dst.remote_ip) |
d342894c | 627 | return true; |
628 | } | |
629 | ||
630 | return false; | |
631 | } | |
632 | ||
633 | /* kernel equivalent to IP_ADD_MEMBERSHIP */ | |
634 | static int vxlan_join_group(struct net_device *dev) | |
635 | { | |
636 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
637 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); | |
638 | struct sock *sk = vn->sock->sk; | |
639 | struct ip_mreqn mreq = { | |
c7995c43 AW |
640 | .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip, |
641 | .imr_ifindex = vxlan->default_dst.remote_ifindex, | |
d342894c | 642 | }; |
643 | int err; | |
644 | ||
645 | /* Already a member of group */ | |
646 | if (vxlan_group_used(vn, vxlan)) | |
647 | return 0; | |
648 | ||
649 | /* Need to drop RTNL to call multicast join */ | |
650 | rtnl_unlock(); | |
651 | lock_sock(sk); | |
652 | err = ip_mc_join_group(sk, &mreq); | |
653 | release_sock(sk); | |
654 | rtnl_lock(); | |
655 | ||
656 | return err; | |
657 | } | |
658 | ||
659 | ||
660 | /* kernel equivalent to IP_DROP_MEMBERSHIP */ | |
661 | static int vxlan_leave_group(struct net_device *dev) | |
662 | { | |
663 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
664 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); | |
665 | int err = 0; | |
666 | struct sock *sk = vn->sock->sk; | |
667 | struct ip_mreqn mreq = { | |
c7995c43 AW |
668 | .imr_multiaddr.s_addr = vxlan->default_dst.remote_ip, |
669 | .imr_ifindex = vxlan->default_dst.remote_ifindex, | |
d342894c | 670 | }; |
671 | ||
672 | /* Only leave group when last vxlan is done. */ | |
673 | if (vxlan_group_used(vn, vxlan)) | |
674 | return 0; | |
675 | ||
676 | /* Need to drop RTNL to call multicast leave */ | |
677 | rtnl_unlock(); | |
678 | lock_sock(sk); | |
679 | err = ip_mc_leave_group(sk, &mreq); | |
680 | release_sock(sk); | |
681 | rtnl_lock(); | |
682 | ||
683 | return err; | |
684 | } | |
685 | ||
686 | /* Callback from net/ipv4/udp.c to receive packets */ | |
687 | static int vxlan_udp_encap_recv(struct sock *sk, struct sk_buff *skb) | |
688 | { | |
689 | struct iphdr *oip; | |
690 | struct vxlanhdr *vxh; | |
691 | struct vxlan_dev *vxlan; | |
e8171045 | 692 | struct pcpu_tstats *stats; |
d342894c | 693 | __u32 vni; |
694 | int err; | |
695 | ||
696 | /* pop off outer UDP header */ | |
697 | __skb_pull(skb, sizeof(struct udphdr)); | |
698 | ||
699 | /* Need Vxlan and inner Ethernet header to be present */ | |
700 | if (!pskb_may_pull(skb, sizeof(struct vxlanhdr))) | |
701 | goto error; | |
702 | ||
703 | /* Drop packets with reserved bits set */ | |
704 | vxh = (struct vxlanhdr *) skb->data; | |
705 | if (vxh->vx_flags != htonl(VXLAN_FLAGS) || | |
706 | (vxh->vx_vni & htonl(0xff))) { | |
707 | netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n", | |
708 | ntohl(vxh->vx_flags), ntohl(vxh->vx_vni)); | |
709 | goto error; | |
710 | } | |
711 | ||
712 | __skb_pull(skb, sizeof(struct vxlanhdr)); | |
d342894c | 713 | |
714 | /* Is this VNI defined? */ | |
715 | vni = ntohl(vxh->vx_vni) >> 8; | |
716 | vxlan = vxlan_find_vni(sock_net(sk), vni); | |
717 | if (!vxlan) { | |
718 | netdev_dbg(skb->dev, "unknown vni %d\n", vni); | |
719 | goto drop; | |
720 | } | |
721 | ||
722 | if (!pskb_may_pull(skb, ETH_HLEN)) { | |
723 | vxlan->dev->stats.rx_length_errors++; | |
724 | vxlan->dev->stats.rx_errors++; | |
725 | goto drop; | |
726 | } | |
727 | ||
e4f67add DS |
728 | skb_reset_mac_header(skb); |
729 | ||
d342894c | 730 | /* Re-examine inner Ethernet packet */ |
731 | oip = ip_hdr(skb); | |
732 | skb->protocol = eth_type_trans(skb, vxlan->dev); | |
d342894c | 733 | |
734 | /* Ignore packet loops (and multicast echo) */ | |
735 | if (compare_ether_addr(eth_hdr(skb)->h_source, | |
736 | vxlan->dev->dev_addr) == 0) | |
737 | goto drop; | |
738 | ||
26a41ae6 | 739 | if ((vxlan->flags & VXLAN_F_LEARN) && |
740 | vxlan_snoop(skb->dev, oip->saddr, eth_hdr(skb)->h_source)) | |
741 | goto drop; | |
d342894c | 742 | |
743 | __skb_tunnel_rx(skb, vxlan->dev); | |
744 | skb_reset_network_header(skb); | |
0afb1666 JG |
745 | |
746 | /* If the NIC driver gave us an encapsulated packet with | |
747 | * CHECKSUM_UNNECESSARY and Rx checksum feature is enabled, | |
748 | * leave the CHECKSUM_UNNECESSARY, the device checksummed it | |
749 | * for us. Otherwise force the upper layers to verify it. | |
750 | */ | |
751 | if (skb->ip_summed != CHECKSUM_UNNECESSARY || !skb->encapsulation || | |
752 | !(vxlan->dev->features & NETIF_F_RXCSUM)) | |
753 | skb->ip_summed = CHECKSUM_NONE; | |
754 | ||
755 | skb->encapsulation = 0; | |
d342894c | 756 | |
757 | err = IP_ECN_decapsulate(oip, skb); | |
758 | if (unlikely(err)) { | |
759 | if (log_ecn_error) | |
760 | net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n", | |
761 | &oip->saddr, oip->tos); | |
762 | if (err > 1) { | |
763 | ++vxlan->dev->stats.rx_frame_errors; | |
764 | ++vxlan->dev->stats.rx_errors; | |
765 | goto drop; | |
766 | } | |
767 | } | |
768 | ||
e8171045 | 769 | stats = this_cpu_ptr(vxlan->dev->tstats); |
d342894c | 770 | u64_stats_update_begin(&stats->syncp); |
771 | stats->rx_packets++; | |
772 | stats->rx_bytes += skb->len; | |
773 | u64_stats_update_end(&stats->syncp); | |
774 | ||
775 | netif_rx(skb); | |
776 | ||
777 | return 0; | |
778 | error: | |
779 | /* Put UDP header back */ | |
780 | __skb_push(skb, sizeof(struct udphdr)); | |
781 | ||
782 | return 1; | |
783 | drop: | |
784 | /* Consume bad packet */ | |
785 | kfree_skb(skb); | |
786 | return 0; | |
787 | } | |
788 | ||
e4f67add DS |
789 | static int arp_reduce(struct net_device *dev, struct sk_buff *skb) |
790 | { | |
791 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
792 | struct arphdr *parp; | |
793 | u8 *arpptr, *sha; | |
794 | __be32 sip, tip; | |
795 | struct neighbour *n; | |
796 | ||
797 | if (dev->flags & IFF_NOARP) | |
798 | goto out; | |
799 | ||
800 | if (!pskb_may_pull(skb, arp_hdr_len(dev))) { | |
801 | dev->stats.tx_dropped++; | |
802 | goto out; | |
803 | } | |
804 | parp = arp_hdr(skb); | |
805 | ||
806 | if ((parp->ar_hrd != htons(ARPHRD_ETHER) && | |
807 | parp->ar_hrd != htons(ARPHRD_IEEE802)) || | |
808 | parp->ar_pro != htons(ETH_P_IP) || | |
809 | parp->ar_op != htons(ARPOP_REQUEST) || | |
810 | parp->ar_hln != dev->addr_len || | |
811 | parp->ar_pln != 4) | |
812 | goto out; | |
813 | arpptr = (u8 *)parp + sizeof(struct arphdr); | |
814 | sha = arpptr; | |
815 | arpptr += dev->addr_len; /* sha */ | |
816 | memcpy(&sip, arpptr, sizeof(sip)); | |
817 | arpptr += sizeof(sip); | |
818 | arpptr += dev->addr_len; /* tha */ | |
819 | memcpy(&tip, arpptr, sizeof(tip)); | |
820 | ||
821 | if (ipv4_is_loopback(tip) || | |
822 | ipv4_is_multicast(tip)) | |
823 | goto out; | |
824 | ||
825 | n = neigh_lookup(&arp_tbl, &tip, dev); | |
826 | ||
827 | if (n) { | |
e4f67add DS |
828 | struct vxlan_fdb *f; |
829 | struct sk_buff *reply; | |
830 | ||
831 | if (!(n->nud_state & NUD_CONNECTED)) { | |
832 | neigh_release(n); | |
833 | goto out; | |
834 | } | |
835 | ||
836 | f = vxlan_find_mac(vxlan, n->ha); | |
6681712d | 837 | if (f && f->remote.remote_ip == htonl(INADDR_ANY)) { |
e4f67add DS |
838 | /* bridge-local neighbor */ |
839 | neigh_release(n); | |
840 | goto out; | |
841 | } | |
842 | ||
843 | reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha, | |
844 | n->ha, sha); | |
845 | ||
846 | neigh_release(n); | |
847 | ||
848 | skb_reset_mac_header(reply); | |
849 | __skb_pull(reply, skb_network_offset(reply)); | |
850 | reply->ip_summed = CHECKSUM_UNNECESSARY; | |
851 | reply->pkt_type = PACKET_HOST; | |
852 | ||
853 | if (netif_rx_ni(reply) == NET_RX_DROP) | |
854 | dev->stats.rx_dropped++; | |
855 | } else if (vxlan->flags & VXLAN_F_L3MISS) | |
856 | vxlan_ip_miss(dev, tip); | |
857 | out: | |
858 | consume_skb(skb); | |
859 | return NETDEV_TX_OK; | |
860 | } | |
861 | ||
862 | static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb) | |
863 | { | |
864 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
865 | struct neighbour *n; | |
866 | struct iphdr *pip; | |
867 | ||
868 | if (is_multicast_ether_addr(eth_hdr(skb)->h_dest)) | |
869 | return false; | |
870 | ||
871 | n = NULL; | |
872 | switch (ntohs(eth_hdr(skb)->h_proto)) { | |
873 | case ETH_P_IP: | |
874 | if (!pskb_may_pull(skb, sizeof(struct iphdr))) | |
875 | return false; | |
876 | pip = ip_hdr(skb); | |
877 | n = neigh_lookup(&arp_tbl, &pip->daddr, dev); | |
878 | break; | |
879 | default: | |
880 | return false; | |
881 | } | |
882 | ||
883 | if (n) { | |
884 | bool diff; | |
885 | ||
886 | diff = compare_ether_addr(eth_hdr(skb)->h_dest, n->ha) != 0; | |
887 | if (diff) { | |
888 | memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest, | |
889 | dev->addr_len); | |
890 | memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len); | |
891 | } | |
892 | neigh_release(n); | |
893 | return diff; | |
894 | } else if (vxlan->flags & VXLAN_F_L3MISS) | |
895 | vxlan_ip_miss(dev, pip->daddr); | |
896 | return false; | |
897 | } | |
898 | ||
1cad8715 | 899 | static void vxlan_sock_free(struct sk_buff *skb) |
900 | { | |
901 | sock_put(skb->sk); | |
902 | } | |
903 | ||
904 | /* On transmit, associate with the tunnel socket */ | |
905 | static void vxlan_set_owner(struct net_device *dev, struct sk_buff *skb) | |
906 | { | |
907 | struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id); | |
908 | struct sock *sk = vn->sock->sk; | |
909 | ||
910 | skb_orphan(skb); | |
911 | sock_hold(sk); | |
912 | skb->sk = sk; | |
913 | skb->destructor = vxlan_sock_free; | |
914 | } | |
915 | ||
05f47d69 | 916 | /* Compute source port for outgoing packet |
917 | * first choice to use L4 flow hash since it will spread | |
918 | * better and maybe available from hardware | |
919 | * secondary choice is to use jhash on the Ethernet header | |
920 | */ | |
7d836a76 | 921 | static __be16 vxlan_src_port(const struct vxlan_dev *vxlan, struct sk_buff *skb) |
05f47d69 | 922 | { |
923 | unsigned int range = (vxlan->port_max - vxlan->port_min) + 1; | |
924 | u32 hash; | |
925 | ||
926 | hash = skb_get_rxhash(skb); | |
927 | if (!hash) | |
928 | hash = jhash(skb->data, 2 * ETH_ALEN, | |
929 | (__force u32) skb->protocol); | |
930 | ||
7d836a76 | 931 | return htons((((u64) hash * range) >> 32) + vxlan->port_min); |
05f47d69 | 932 | } |
933 | ||
05c0db08 PS |
934 | static int handle_offloads(struct sk_buff *skb) |
935 | { | |
936 | if (skb_is_gso(skb)) { | |
937 | int err = skb_unclone(skb, GFP_ATOMIC); | |
938 | if (unlikely(err)) | |
939 | return err; | |
940 | ||
f6ace502 | 941 | skb_shinfo(skb)->gso_type |= SKB_GSO_UDP_TUNNEL; |
05c0db08 PS |
942 | } else if (skb->ip_summed != CHECKSUM_PARTIAL) |
943 | skb->ip_summed = CHECKSUM_NONE; | |
944 | ||
945 | return 0; | |
946 | } | |
947 | ||
9dcc71e1 SS |
948 | /* Bypass encapsulation if the destination is local */ |
949 | static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan, | |
950 | struct vxlan_dev *dst_vxlan) | |
951 | { | |
952 | struct pcpu_tstats *tx_stats = this_cpu_ptr(src_vxlan->dev->tstats); | |
953 | struct pcpu_tstats *rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats); | |
954 | ||
955 | skb->pkt_type = PACKET_HOST; | |
956 | skb->encapsulation = 0; | |
957 | skb->dev = dst_vxlan->dev; | |
958 | __skb_pull(skb, skb_network_offset(skb)); | |
959 | ||
960 | if (dst_vxlan->flags & VXLAN_F_LEARN) | |
9d9f163c MR |
961 | vxlan_snoop(skb->dev, htonl(INADDR_LOOPBACK), |
962 | eth_hdr(skb)->h_source); | |
9dcc71e1 SS |
963 | |
964 | u64_stats_update_begin(&tx_stats->syncp); | |
965 | tx_stats->tx_packets++; | |
966 | tx_stats->tx_bytes += skb->len; | |
967 | u64_stats_update_end(&tx_stats->syncp); | |
968 | ||
969 | if (netif_rx(skb) == NET_RX_SUCCESS) { | |
970 | u64_stats_update_begin(&rx_stats->syncp); | |
971 | rx_stats->rx_packets++; | |
972 | rx_stats->rx_bytes += skb->len; | |
973 | u64_stats_update_end(&rx_stats->syncp); | |
974 | } else { | |
975 | skb->dev->stats.rx_dropped++; | |
976 | } | |
977 | } | |
978 | ||
6681712d DS |
979 | static netdev_tx_t vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev, |
980 | struct vxlan_rdst *rdst, bool did_rsc) | |
d342894c | 981 | { |
982 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
983 | struct rtable *rt; | |
d342894c | 984 | const struct iphdr *old_iph; |
985 | struct iphdr *iph; | |
986 | struct vxlanhdr *vxh; | |
987 | struct udphdr *uh; | |
988 | struct flowi4 fl4; | |
d342894c | 989 | __be32 dst; |
7d836a76 | 990 | __be16 src_port, dst_port; |
6681712d | 991 | u32 vni; |
d342894c | 992 | __be16 df = 0; |
993 | __u8 tos, ttl; | |
d342894c | 994 | |
823aa873 | 995 | dst_port = rdst->remote_port ? rdst->remote_port : vxlan->dst_port; |
6681712d DS |
996 | vni = rdst->remote_vni; |
997 | dst = rdst->remote_ip; | |
e4f67add DS |
998 | |
999 | if (!dst) { | |
1000 | if (did_rsc) { | |
e4f67add | 1001 | /* short-circuited back to local bridge */ |
9dcc71e1 | 1002 | vxlan_encap_bypass(skb, vxlan, vxlan); |
e4f67add DS |
1003 | return NETDEV_TX_OK; |
1004 | } | |
ef59febe | 1005 | goto drop; |
e4f67add | 1006 | } |
ef59febe | 1007 | |
d6727fe3 JG |
1008 | if (!skb->encapsulation) { |
1009 | skb_reset_inner_headers(skb); | |
1010 | skb->encapsulation = 1; | |
1011 | } | |
1012 | ||
d342894c | 1013 | /* Need space for new headers (invalidates iph ptr) */ |
1014 | if (skb_cow_head(skb, VXLAN_HEADROOM)) | |
1015 | goto drop; | |
1016 | ||
d342894c | 1017 | old_iph = ip_hdr(skb); |
1018 | ||
d342894c | 1019 | ttl = vxlan->ttl; |
1020 | if (!ttl && IN_MULTICAST(ntohl(dst))) | |
1021 | ttl = 1; | |
1022 | ||
1023 | tos = vxlan->tos; | |
1024 | if (tos == 1) | |
206aaafc | 1025 | tos = ip_tunnel_get_dsfield(old_iph, skb); |
d342894c | 1026 | |
05f47d69 | 1027 | src_port = vxlan_src_port(vxlan, skb); |
d342894c | 1028 | |
ca78f181 | 1029 | memset(&fl4, 0, sizeof(fl4)); |
6681712d | 1030 | fl4.flowi4_oif = rdst->remote_ifindex; |
ca78f181 | 1031 | fl4.flowi4_tos = RT_TOS(tos); |
1032 | fl4.daddr = dst; | |
1033 | fl4.saddr = vxlan->saddr; | |
1034 | ||
1035 | rt = ip_route_output_key(dev_net(dev), &fl4); | |
d342894c | 1036 | if (IS_ERR(rt)) { |
1037 | netdev_dbg(dev, "no route to %pI4\n", &dst); | |
1038 | dev->stats.tx_carrier_errors++; | |
1039 | goto tx_error; | |
1040 | } | |
1041 | ||
1042 | if (rt->dst.dev == dev) { | |
1043 | netdev_dbg(dev, "circular route to %pI4\n", &dst); | |
1044 | ip_rt_put(rt); | |
1045 | dev->stats.collisions++; | |
1046 | goto tx_error; | |
1047 | } | |
1048 | ||
9dcc71e1 | 1049 | /* Bypass encapsulation if the destination is local */ |
ab09a6d0 MR |
1050 | if (rt->rt_flags & RTCF_LOCAL && |
1051 | !(rt->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) { | |
9dcc71e1 SS |
1052 | struct vxlan_dev *dst_vxlan; |
1053 | ||
1054 | ip_rt_put(rt); | |
1055 | dst_vxlan = vxlan_find_vni(dev_net(dev), vni); | |
1056 | if (!dst_vxlan) | |
1057 | goto tx_error; | |
1058 | vxlan_encap_bypass(skb, vxlan, dst_vxlan); | |
1059 | return NETDEV_TX_OK; | |
1060 | } | |
1061 | ||
d342894c | 1062 | memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt)); |
1063 | IPCB(skb)->flags &= ~(IPSKB_XFRM_TUNNEL_SIZE | IPSKB_XFRM_TRANSFORMED | | |
1064 | IPSKB_REROUTED); | |
1065 | skb_dst_drop(skb); | |
1066 | skb_dst_set(skb, &rt->dst); | |
1067 | ||
1068 | vxh = (struct vxlanhdr *) __skb_push(skb, sizeof(*vxh)); | |
1069 | vxh->vx_flags = htonl(VXLAN_FLAGS); | |
6681712d | 1070 | vxh->vx_vni = htonl(vni << 8); |
d342894c | 1071 | |
1072 | __skb_push(skb, sizeof(*uh)); | |
1073 | skb_reset_transport_header(skb); | |
1074 | uh = udp_hdr(skb); | |
1075 | ||
73cf3317 | 1076 | uh->dest = dst_port; |
7d836a76 | 1077 | uh->source = src_port; |
d342894c | 1078 | |
1079 | uh->len = htons(skb->len); | |
1080 | uh->check = 0; | |
1081 | ||
1082 | __skb_push(skb, sizeof(*iph)); | |
1083 | skb_reset_network_header(skb); | |
1084 | iph = ip_hdr(skb); | |
1085 | iph->version = 4; | |
1086 | iph->ihl = sizeof(struct iphdr) >> 2; | |
1087 | iph->frag_off = df; | |
1088 | iph->protocol = IPPROTO_UDP; | |
206aaafc | 1089 | iph->tos = ip_tunnel_ecn_encap(tos, old_iph, skb); |
ca78f181 | 1090 | iph->daddr = dst; |
d342894c | 1091 | iph->saddr = fl4.saddr; |
1092 | iph->ttl = ttl ? : ip4_dst_hoplimit(&rt->dst); | |
8dc98eb2 | 1093 | tunnel_ip_select_ident(skb, old_iph, &rt->dst); |
d342894c | 1094 | |
88c4c066 ZM |
1095 | nf_reset(skb); |
1096 | ||
1cad8715 | 1097 | vxlan_set_owner(dev, skb); |
1098 | ||
05c0db08 PS |
1099 | if (handle_offloads(skb)) |
1100 | goto drop; | |
d342894c | 1101 | |
6aed0c8b | 1102 | iptunnel_xmit(skb, dev); |
d342894c | 1103 | return NETDEV_TX_OK; |
1104 | ||
1105 | drop: | |
1106 | dev->stats.tx_dropped++; | |
1107 | goto tx_free; | |
1108 | ||
1109 | tx_error: | |
1110 | dev->stats.tx_errors++; | |
1111 | tx_free: | |
1112 | dev_kfree_skb(skb); | |
1113 | return NETDEV_TX_OK; | |
1114 | } | |
1115 | ||
6681712d DS |
1116 | /* Transmit local packets over Vxlan |
1117 | * | |
1118 | * Outer IP header inherits ECN and DF from inner header. | |
1119 | * Outer UDP destination is the VXLAN assigned port. | |
1120 | * source port is based on hash of flow | |
1121 | */ | |
1122 | static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev) | |
1123 | { | |
1124 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1125 | struct ethhdr *eth; | |
1126 | bool did_rsc = false; | |
c7995c43 | 1127 | struct vxlan_rdst *rdst0, *rdst; |
6681712d DS |
1128 | struct vxlan_fdb *f; |
1129 | int rc1, rc; | |
1130 | ||
1131 | skb_reset_mac_header(skb); | |
1132 | eth = eth_hdr(skb); | |
1133 | ||
1134 | if ((vxlan->flags & VXLAN_F_PROXY) && ntohs(eth->h_proto) == ETH_P_ARP) | |
1135 | return arp_reduce(dev, skb); | |
6681712d DS |
1136 | |
1137 | f = vxlan_find_mac(vxlan, eth->h_dest); | |
ae884082 DS |
1138 | did_rsc = false; |
1139 | ||
1140 | if (f && (f->flags & NTF_ROUTER) && (vxlan->flags & VXLAN_F_RSC) && | |
1141 | ntohs(eth->h_proto) == ETH_P_IP) { | |
1142 | did_rsc = route_shortcircuit(dev, skb); | |
1143 | if (did_rsc) | |
1144 | f = vxlan_find_mac(vxlan, eth->h_dest); | |
1145 | } | |
1146 | ||
6681712d | 1147 | if (f == NULL) { |
c7995c43 AW |
1148 | rdst0 = &vxlan->default_dst; |
1149 | ||
1150 | if (rdst0->remote_ip == htonl(INADDR_ANY) && | |
6681712d DS |
1151 | (vxlan->flags & VXLAN_F_L2MISS) && |
1152 | !is_multicast_ether_addr(eth->h_dest)) | |
1153 | vxlan_fdb_miss(vxlan, eth->h_dest); | |
1154 | } else | |
1155 | rdst0 = &f->remote; | |
1156 | ||
1157 | rc = NETDEV_TX_OK; | |
1158 | ||
1159 | /* if there are multiple destinations, send copies */ | |
1160 | for (rdst = rdst0->remote_next; rdst; rdst = rdst->remote_next) { | |
1161 | struct sk_buff *skb1; | |
1162 | ||
1163 | skb1 = skb_clone(skb, GFP_ATOMIC); | |
1164 | rc1 = vxlan_xmit_one(skb1, dev, rdst, did_rsc); | |
1165 | if (rc == NETDEV_TX_OK) | |
1166 | rc = rc1; | |
1167 | } | |
1168 | ||
1169 | rc1 = vxlan_xmit_one(skb, dev, rdst0, did_rsc); | |
1170 | if (rc == NETDEV_TX_OK) | |
1171 | rc = rc1; | |
1172 | return rc; | |
1173 | } | |
1174 | ||
d342894c | 1175 | /* Walk the forwarding table and purge stale entries */ |
1176 | static void vxlan_cleanup(unsigned long arg) | |
1177 | { | |
1178 | struct vxlan_dev *vxlan = (struct vxlan_dev *) arg; | |
1179 | unsigned long next_timer = jiffies + FDB_AGE_INTERVAL; | |
1180 | unsigned int h; | |
1181 | ||
1182 | if (!netif_running(vxlan->dev)) | |
1183 | return; | |
1184 | ||
1185 | spin_lock_bh(&vxlan->hash_lock); | |
1186 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
1187 | struct hlist_node *p, *n; | |
1188 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { | |
1189 | struct vxlan_fdb *f | |
1190 | = container_of(p, struct vxlan_fdb, hlist); | |
1191 | unsigned long timeout; | |
1192 | ||
3c172868 | 1193 | if (f->state & NUD_PERMANENT) |
d342894c | 1194 | continue; |
1195 | ||
1196 | timeout = f->used + vxlan->age_interval * HZ; | |
1197 | if (time_before_eq(timeout, jiffies)) { | |
1198 | netdev_dbg(vxlan->dev, | |
1199 | "garbage collect %pM\n", | |
1200 | f->eth_addr); | |
1201 | f->state = NUD_STALE; | |
1202 | vxlan_fdb_destroy(vxlan, f); | |
1203 | } else if (time_before(timeout, next_timer)) | |
1204 | next_timer = timeout; | |
1205 | } | |
1206 | } | |
1207 | spin_unlock_bh(&vxlan->hash_lock); | |
1208 | ||
1209 | mod_timer(&vxlan->age_timer, next_timer); | |
1210 | } | |
1211 | ||
1212 | /* Setup stats when device is created */ | |
1213 | static int vxlan_init(struct net_device *dev) | |
1214 | { | |
e8171045 PS |
1215 | dev->tstats = alloc_percpu(struct pcpu_tstats); |
1216 | if (!dev->tstats) | |
d342894c | 1217 | return -ENOMEM; |
1218 | ||
1219 | return 0; | |
1220 | } | |
1221 | ||
1222 | /* Start ageing timer and join group when device is brought up */ | |
1223 | static int vxlan_open(struct net_device *dev) | |
1224 | { | |
1225 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1226 | int err; | |
1227 | ||
c7995c43 | 1228 | if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) { |
d342894c | 1229 | err = vxlan_join_group(dev); |
1230 | if (err) | |
1231 | return err; | |
1232 | } | |
1233 | ||
1234 | if (vxlan->age_interval) | |
1235 | mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL); | |
1236 | ||
1237 | return 0; | |
1238 | } | |
1239 | ||
1240 | /* Purge the forwarding table */ | |
1241 | static void vxlan_flush(struct vxlan_dev *vxlan) | |
1242 | { | |
1243 | unsigned h; | |
1244 | ||
1245 | spin_lock_bh(&vxlan->hash_lock); | |
1246 | for (h = 0; h < FDB_HASH_SIZE; ++h) { | |
1247 | struct hlist_node *p, *n; | |
1248 | hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) { | |
1249 | struct vxlan_fdb *f | |
1250 | = container_of(p, struct vxlan_fdb, hlist); | |
1251 | vxlan_fdb_destroy(vxlan, f); | |
1252 | } | |
1253 | } | |
1254 | spin_unlock_bh(&vxlan->hash_lock); | |
1255 | } | |
1256 | ||
1257 | /* Cleanup timer and forwarding table on shutdown */ | |
1258 | static int vxlan_stop(struct net_device *dev) | |
1259 | { | |
1260 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1261 | ||
c7995c43 | 1262 | if (IN_MULTICAST(ntohl(vxlan->default_dst.remote_ip))) |
d342894c | 1263 | vxlan_leave_group(dev); |
1264 | ||
1265 | del_timer_sync(&vxlan->age_timer); | |
1266 | ||
1267 | vxlan_flush(vxlan); | |
1268 | ||
1269 | return 0; | |
1270 | } | |
1271 | ||
d342894c | 1272 | /* Stub, nothing needs to be done. */ |
1273 | static void vxlan_set_multicast_list(struct net_device *dev) | |
1274 | { | |
1275 | } | |
1276 | ||
1277 | static const struct net_device_ops vxlan_netdev_ops = { | |
1278 | .ndo_init = vxlan_init, | |
1279 | .ndo_open = vxlan_open, | |
1280 | .ndo_stop = vxlan_stop, | |
1281 | .ndo_start_xmit = vxlan_xmit, | |
e8171045 | 1282 | .ndo_get_stats64 = ip_tunnel_get_stats64, |
d342894c | 1283 | .ndo_set_rx_mode = vxlan_set_multicast_list, |
1284 | .ndo_change_mtu = eth_change_mtu, | |
1285 | .ndo_validate_addr = eth_validate_addr, | |
1286 | .ndo_set_mac_address = eth_mac_addr, | |
1287 | .ndo_fdb_add = vxlan_fdb_add, | |
1288 | .ndo_fdb_del = vxlan_fdb_delete, | |
1289 | .ndo_fdb_dump = vxlan_fdb_dump, | |
1290 | }; | |
1291 | ||
1292 | /* Info for udev, that this is a virtual tunnel endpoint */ | |
1293 | static struct device_type vxlan_type = { | |
1294 | .name = "vxlan", | |
1295 | }; | |
1296 | ||
1297 | static void vxlan_free(struct net_device *dev) | |
1298 | { | |
e8171045 | 1299 | free_percpu(dev->tstats); |
d342894c | 1300 | free_netdev(dev); |
1301 | } | |
1302 | ||
1303 | /* Initialize the device structure. */ | |
1304 | static void vxlan_setup(struct net_device *dev) | |
1305 | { | |
1306 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1307 | unsigned h; | |
05f47d69 | 1308 | int low, high; |
d342894c | 1309 | |
1310 | eth_hw_addr_random(dev); | |
1311 | ether_setup(dev); | |
2840bf22 | 1312 | dev->hard_header_len = ETH_HLEN + VXLAN_HEADROOM; |
d342894c | 1313 | |
1314 | dev->netdev_ops = &vxlan_netdev_ops; | |
1315 | dev->destructor = vxlan_free; | |
1316 | SET_NETDEV_DEVTYPE(dev, &vxlan_type); | |
1317 | ||
1318 | dev->tx_queue_len = 0; | |
1319 | dev->features |= NETIF_F_LLTX; | |
1320 | dev->features |= NETIF_F_NETNS_LOCAL; | |
d6727fe3 | 1321 | dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; |
0afb1666 | 1322 | dev->features |= NETIF_F_RXCSUM; |
05c0db08 | 1323 | dev->features |= NETIF_F_GSO_SOFTWARE; |
0afb1666 JG |
1324 | |
1325 | dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM; | |
05c0db08 | 1326 | dev->hw_features |= NETIF_F_GSO_SOFTWARE; |
d342894c | 1327 | dev->priv_flags &= ~IFF_XMIT_DST_RELEASE; |
6602d007 | 1328 | dev->priv_flags |= IFF_LIVE_ADDR_CHANGE; |
d342894c | 1329 | |
1330 | spin_lock_init(&vxlan->hash_lock); | |
1331 | ||
1332 | init_timer_deferrable(&vxlan->age_timer); | |
1333 | vxlan->age_timer.function = vxlan_cleanup; | |
1334 | vxlan->age_timer.data = (unsigned long) vxlan; | |
1335 | ||
05f47d69 | 1336 | inet_get_local_port_range(&low, &high); |
1337 | vxlan->port_min = low; | |
1338 | vxlan->port_max = high; | |
823aa873 | 1339 | vxlan->dst_port = htons(vxlan_port); |
05f47d69 | 1340 | |
d342894c | 1341 | vxlan->dev = dev; |
1342 | ||
1343 | for (h = 0; h < FDB_HASH_SIZE; ++h) | |
1344 | INIT_HLIST_HEAD(&vxlan->fdb_head[h]); | |
1345 | } | |
1346 | ||
1347 | static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = { | |
1348 | [IFLA_VXLAN_ID] = { .type = NLA_U32 }, | |
5d174dd8 | 1349 | [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) }, |
d342894c | 1350 | [IFLA_VXLAN_LINK] = { .type = NLA_U32 }, |
1351 | [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) }, | |
1352 | [IFLA_VXLAN_TOS] = { .type = NLA_U8 }, | |
1353 | [IFLA_VXLAN_TTL] = { .type = NLA_U8 }, | |
1354 | [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 }, | |
1355 | [IFLA_VXLAN_AGEING] = { .type = NLA_U32 }, | |
1356 | [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 }, | |
05f47d69 | 1357 | [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) }, |
e4f67add DS |
1358 | [IFLA_VXLAN_PROXY] = { .type = NLA_U8 }, |
1359 | [IFLA_VXLAN_RSC] = { .type = NLA_U8 }, | |
1360 | [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 }, | |
1361 | [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 }, | |
823aa873 | 1362 | [IFLA_VXLAN_PORT] = { .type = NLA_U16 }, |
d342894c | 1363 | }; |
1364 | ||
1365 | static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[]) | |
1366 | { | |
1367 | if (tb[IFLA_ADDRESS]) { | |
1368 | if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) { | |
1369 | pr_debug("invalid link address (not ethernet)\n"); | |
1370 | return -EINVAL; | |
1371 | } | |
1372 | ||
1373 | if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) { | |
1374 | pr_debug("invalid all zero ethernet address\n"); | |
1375 | return -EADDRNOTAVAIL; | |
1376 | } | |
1377 | } | |
1378 | ||
1379 | if (!data) | |
1380 | return -EINVAL; | |
1381 | ||
1382 | if (data[IFLA_VXLAN_ID]) { | |
1383 | __u32 id = nla_get_u32(data[IFLA_VXLAN_ID]); | |
1384 | if (id >= VXLAN_VID_MASK) | |
1385 | return -ERANGE; | |
1386 | } | |
1387 | ||
05f47d69 | 1388 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
1389 | const struct ifla_vxlan_port_range *p | |
1390 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
1391 | ||
1392 | if (ntohs(p->high) < ntohs(p->low)) { | |
1393 | pr_debug("port range %u .. %u not valid\n", | |
1394 | ntohs(p->low), ntohs(p->high)); | |
1395 | return -EINVAL; | |
1396 | } | |
1397 | } | |
1398 | ||
d342894c | 1399 | return 0; |
1400 | } | |
1401 | ||
1b13c97f YB |
1402 | static void vxlan_get_drvinfo(struct net_device *netdev, |
1403 | struct ethtool_drvinfo *drvinfo) | |
1404 | { | |
1405 | strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version)); | |
1406 | strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver)); | |
1407 | } | |
1408 | ||
1409 | static const struct ethtool_ops vxlan_ethtool_ops = { | |
1410 | .get_drvinfo = vxlan_get_drvinfo, | |
1411 | .get_link = ethtool_op_get_link, | |
1412 | }; | |
1413 | ||
d342894c | 1414 | static int vxlan_newlink(struct net *net, struct net_device *dev, |
1415 | struct nlattr *tb[], struct nlattr *data[]) | |
1416 | { | |
1417 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
c7995c43 | 1418 | struct vxlan_rdst *dst = &vxlan->default_dst; |
d342894c | 1419 | __u32 vni; |
1420 | int err; | |
1421 | ||
1422 | if (!data[IFLA_VXLAN_ID]) | |
1423 | return -EINVAL; | |
1424 | ||
1425 | vni = nla_get_u32(data[IFLA_VXLAN_ID]); | |
1426 | if (vxlan_find_vni(net, vni)) { | |
1427 | pr_info("duplicate VNI %u\n", vni); | |
1428 | return -EEXIST; | |
1429 | } | |
c7995c43 | 1430 | dst->remote_vni = vni; |
d342894c | 1431 | |
5d174dd8 | 1432 | if (data[IFLA_VXLAN_GROUP]) |
1433 | dst->remote_ip = nla_get_be32(data[IFLA_VXLAN_GROUP]); | |
d342894c | 1434 | |
1435 | if (data[IFLA_VXLAN_LOCAL]) | |
1436 | vxlan->saddr = nla_get_be32(data[IFLA_VXLAN_LOCAL]); | |
1437 | ||
34e02aa1 | 1438 | if (data[IFLA_VXLAN_LINK] && |
c7995c43 | 1439 | (dst->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]))) { |
34e02aa1 | 1440 | struct net_device *lowerdev |
c7995c43 | 1441 | = __dev_get_by_index(net, dst->remote_ifindex); |
34e02aa1 | 1442 | |
1443 | if (!lowerdev) { | |
c7995c43 | 1444 | pr_info("ifindex %d does not exist\n", dst->remote_ifindex); |
34e02aa1 | 1445 | return -ENODEV; |
1446 | } | |
d342894c | 1447 | |
34e02aa1 | 1448 | if (!tb[IFLA_MTU]) |
d342894c | 1449 | dev->mtu = lowerdev->mtu - VXLAN_HEADROOM; |
1ba56fb4 AD |
1450 | |
1451 | /* update header length based on lower device */ | |
1452 | dev->hard_header_len = lowerdev->hard_header_len + | |
1453 | VXLAN_HEADROOM; | |
d342894c | 1454 | } |
1455 | ||
1456 | if (data[IFLA_VXLAN_TOS]) | |
1457 | vxlan->tos = nla_get_u8(data[IFLA_VXLAN_TOS]); | |
1458 | ||
afb97186 VB |
1459 | if (data[IFLA_VXLAN_TTL]) |
1460 | vxlan->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]); | |
1461 | ||
d342894c | 1462 | if (!data[IFLA_VXLAN_LEARNING] || nla_get_u8(data[IFLA_VXLAN_LEARNING])) |
e4f67add | 1463 | vxlan->flags |= VXLAN_F_LEARN; |
d342894c | 1464 | |
1465 | if (data[IFLA_VXLAN_AGEING]) | |
1466 | vxlan->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]); | |
1467 | else | |
1468 | vxlan->age_interval = FDB_AGE_DEFAULT; | |
1469 | ||
e4f67add DS |
1470 | if (data[IFLA_VXLAN_PROXY] && nla_get_u8(data[IFLA_VXLAN_PROXY])) |
1471 | vxlan->flags |= VXLAN_F_PROXY; | |
1472 | ||
1473 | if (data[IFLA_VXLAN_RSC] && nla_get_u8(data[IFLA_VXLAN_RSC])) | |
1474 | vxlan->flags |= VXLAN_F_RSC; | |
1475 | ||
1476 | if (data[IFLA_VXLAN_L2MISS] && nla_get_u8(data[IFLA_VXLAN_L2MISS])) | |
1477 | vxlan->flags |= VXLAN_F_L2MISS; | |
1478 | ||
1479 | if (data[IFLA_VXLAN_L3MISS] && nla_get_u8(data[IFLA_VXLAN_L3MISS])) | |
1480 | vxlan->flags |= VXLAN_F_L3MISS; | |
1481 | ||
d342894c | 1482 | if (data[IFLA_VXLAN_LIMIT]) |
1483 | vxlan->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]); | |
1484 | ||
05f47d69 | 1485 | if (data[IFLA_VXLAN_PORT_RANGE]) { |
1486 | const struct ifla_vxlan_port_range *p | |
1487 | = nla_data(data[IFLA_VXLAN_PORT_RANGE]); | |
1488 | vxlan->port_min = ntohs(p->low); | |
1489 | vxlan->port_max = ntohs(p->high); | |
1490 | } | |
1491 | ||
823aa873 | 1492 | if (data[IFLA_VXLAN_PORT]) |
1493 | vxlan->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]); | |
1494 | ||
1b13c97f YB |
1495 | SET_ETHTOOL_OPS(dev, &vxlan_ethtool_ops); |
1496 | ||
d342894c | 1497 | err = register_netdevice(dev); |
1498 | if (!err) | |
c7995c43 | 1499 | hlist_add_head_rcu(&vxlan->hlist, vni_head(net, dst->remote_vni)); |
d342894c | 1500 | |
1501 | return err; | |
1502 | } | |
1503 | ||
1504 | static void vxlan_dellink(struct net_device *dev, struct list_head *head) | |
1505 | { | |
1506 | struct vxlan_dev *vxlan = netdev_priv(dev); | |
1507 | ||
1508 | hlist_del_rcu(&vxlan->hlist); | |
1509 | ||
1510 | unregister_netdevice_queue(dev, head); | |
1511 | } | |
1512 | ||
1513 | static size_t vxlan_get_size(const struct net_device *dev) | |
1514 | { | |
1515 | ||
1516 | return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */ | |
5d174dd8 | 1517 | nla_total_size(sizeof(__be32)) +/* IFLA_VXLAN_GROUP */ |
d342894c | 1518 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */ |
1519 | nla_total_size(sizeof(__be32))+ /* IFLA_VXLAN_LOCAL */ | |
1520 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */ | |
1521 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */ | |
1522 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */ | |
e4f67add DS |
1523 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */ |
1524 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */ | |
1525 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */ | |
1526 | nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */ | |
d342894c | 1527 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */ |
1528 | nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */ | |
05f47d69 | 1529 | nla_total_size(sizeof(struct ifla_vxlan_port_range)) + |
823aa873 | 1530 | nla_total_size(sizeof(__be16))+ /* IFLA_VXLAN_PORT */ |
d342894c | 1531 | 0; |
1532 | } | |
1533 | ||
1534 | static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev) | |
1535 | { | |
1536 | const struct vxlan_dev *vxlan = netdev_priv(dev); | |
c7995c43 | 1537 | const struct vxlan_rdst *dst = &vxlan->default_dst; |
05f47d69 | 1538 | struct ifla_vxlan_port_range ports = { |
1539 | .low = htons(vxlan->port_min), | |
1540 | .high = htons(vxlan->port_max), | |
1541 | }; | |
d342894c | 1542 | |
c7995c43 | 1543 | if (nla_put_u32(skb, IFLA_VXLAN_ID, dst->remote_vni)) |
d342894c | 1544 | goto nla_put_failure; |
1545 | ||
5d174dd8 | 1546 | if (dst->remote_ip && nla_put_be32(skb, IFLA_VXLAN_GROUP, dst->remote_ip)) |
d342894c | 1547 | goto nla_put_failure; |
1548 | ||
c7995c43 | 1549 | if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex)) |
d342894c | 1550 | goto nla_put_failure; |
1551 | ||
7c41c42c | 1552 | if (vxlan->saddr && nla_put_be32(skb, IFLA_VXLAN_LOCAL, vxlan->saddr)) |
d342894c | 1553 | goto nla_put_failure; |
1554 | ||
1555 | if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->ttl) || | |
1556 | nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->tos) || | |
e4f67add DS |
1557 | nla_put_u8(skb, IFLA_VXLAN_LEARNING, |
1558 | !!(vxlan->flags & VXLAN_F_LEARN)) || | |
1559 | nla_put_u8(skb, IFLA_VXLAN_PROXY, | |
1560 | !!(vxlan->flags & VXLAN_F_PROXY)) || | |
1561 | nla_put_u8(skb, IFLA_VXLAN_RSC, !!(vxlan->flags & VXLAN_F_RSC)) || | |
1562 | nla_put_u8(skb, IFLA_VXLAN_L2MISS, | |
1563 | !!(vxlan->flags & VXLAN_F_L2MISS)) || | |
1564 | nla_put_u8(skb, IFLA_VXLAN_L3MISS, | |
1565 | !!(vxlan->flags & VXLAN_F_L3MISS)) || | |
d342894c | 1566 | nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->age_interval) || |
823aa873 | 1567 | nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->addrmax) || |
1568 | nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->dst_port)) | |
d342894c | 1569 | goto nla_put_failure; |
1570 | ||
05f47d69 | 1571 | if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports)) |
1572 | goto nla_put_failure; | |
1573 | ||
d342894c | 1574 | return 0; |
1575 | ||
1576 | nla_put_failure: | |
1577 | return -EMSGSIZE; | |
1578 | } | |
1579 | ||
1580 | static struct rtnl_link_ops vxlan_link_ops __read_mostly = { | |
1581 | .kind = "vxlan", | |
1582 | .maxtype = IFLA_VXLAN_MAX, | |
1583 | .policy = vxlan_policy, | |
1584 | .priv_size = sizeof(struct vxlan_dev), | |
1585 | .setup = vxlan_setup, | |
1586 | .validate = vxlan_validate, | |
1587 | .newlink = vxlan_newlink, | |
1588 | .dellink = vxlan_dellink, | |
1589 | .get_size = vxlan_get_size, | |
1590 | .fill_info = vxlan_fill_info, | |
1591 | }; | |
1592 | ||
1593 | static __net_init int vxlan_init_net(struct net *net) | |
1594 | { | |
1595 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
1596 | struct sock *sk; | |
1597 | struct sockaddr_in vxlan_addr = { | |
1598 | .sin_family = AF_INET, | |
1599 | .sin_addr.s_addr = htonl(INADDR_ANY), | |
1600 | }; | |
1601 | int rc; | |
1602 | unsigned h; | |
1603 | ||
1604 | /* Create UDP socket for encapsulation receive. */ | |
1605 | rc = sock_create_kern(AF_INET, SOCK_DGRAM, IPPROTO_UDP, &vn->sock); | |
1606 | if (rc < 0) { | |
1607 | pr_debug("UDP socket create failed\n"); | |
1608 | return rc; | |
1609 | } | |
bfe1b9b1 | 1610 | /* Put in proper namespace */ |
1611 | sk = vn->sock->sk; | |
1612 | sk_change_net(sk, net); | |
d342894c | 1613 | |
1614 | vxlan_addr.sin_port = htons(vxlan_port); | |
1615 | ||
1616 | rc = kernel_bind(vn->sock, (struct sockaddr *) &vxlan_addr, | |
1617 | sizeof(vxlan_addr)); | |
1618 | if (rc < 0) { | |
1619 | pr_debug("bind for UDP socket %pI4:%u (%d)\n", | |
1620 | &vxlan_addr.sin_addr, ntohs(vxlan_addr.sin_port), rc); | |
bfe1b9b1 | 1621 | sk_release_kernel(sk); |
d342894c | 1622 | vn->sock = NULL; |
1623 | return rc; | |
1624 | } | |
1625 | ||
1626 | /* Disable multicast loopback */ | |
d342894c | 1627 | inet_sk(sk)->mc_loop = 0; |
1628 | ||
1629 | /* Mark socket as an encapsulation socket. */ | |
1630 | udp_sk(sk)->encap_type = 1; | |
1631 | udp_sk(sk)->encap_rcv = vxlan_udp_encap_recv; | |
1632 | udp_encap_enable(); | |
1633 | ||
1634 | for (h = 0; h < VNI_HASH_SIZE; ++h) | |
1635 | INIT_HLIST_HEAD(&vn->vni_list[h]); | |
1636 | ||
1637 | return 0; | |
1638 | } | |
1639 | ||
1640 | static __net_exit void vxlan_exit_net(struct net *net) | |
1641 | { | |
1642 | struct vxlan_net *vn = net_generic(net, vxlan_net_id); | |
9cb6cb7e ZM |
1643 | struct vxlan_dev *vxlan; |
1644 | unsigned h; | |
1645 | ||
1646 | rtnl_lock(); | |
1647 | for (h = 0; h < VNI_HASH_SIZE; ++h) | |
1648 | hlist_for_each_entry(vxlan, &vn->vni_list[h], hlist) | |
1649 | dev_close(vxlan->dev); | |
1650 | rtnl_unlock(); | |
d342894c | 1651 | |
1652 | if (vn->sock) { | |
bfe1b9b1 | 1653 | sk_release_kernel(vn->sock->sk); |
d342894c | 1654 | vn->sock = NULL; |
1655 | } | |
1656 | } | |
1657 | ||
1658 | static struct pernet_operations vxlan_net_ops = { | |
1659 | .init = vxlan_init_net, | |
1660 | .exit = vxlan_exit_net, | |
1661 | .id = &vxlan_net_id, | |
1662 | .size = sizeof(struct vxlan_net), | |
1663 | }; | |
1664 | ||
1665 | static int __init vxlan_init_module(void) | |
1666 | { | |
1667 | int rc; | |
1668 | ||
1669 | get_random_bytes(&vxlan_salt, sizeof(vxlan_salt)); | |
1670 | ||
1671 | rc = register_pernet_device(&vxlan_net_ops); | |
1672 | if (rc) | |
1673 | goto out1; | |
1674 | ||
1675 | rc = rtnl_link_register(&vxlan_link_ops); | |
1676 | if (rc) | |
1677 | goto out2; | |
1678 | ||
1679 | return 0; | |
1680 | ||
1681 | out2: | |
1682 | unregister_pernet_device(&vxlan_net_ops); | |
1683 | out1: | |
1684 | return rc; | |
1685 | } | |
1686 | module_init(vxlan_init_module); | |
1687 | ||
1688 | static void __exit vxlan_cleanup_module(void) | |
1689 | { | |
1690 | rtnl_link_unregister(&vxlan_link_ops); | |
1691 | unregister_pernet_device(&vxlan_net_ops); | |
6681712d | 1692 | rcu_barrier(); |
d342894c | 1693 | } |
1694 | module_exit(vxlan_cleanup_module); | |
1695 | ||
1696 | MODULE_LICENSE("GPL"); | |
1697 | MODULE_VERSION(VXLAN_VERSION); | |
3b8df3c6 | 1698 | MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>"); |
d342894c | 1699 | MODULE_ALIAS_RTNL_LINK("vxlan"); |