]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - drivers/net/vxlan.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
net: hns3: Correct unreasonable code comments
[mirror_ubuntu-bionic-kernel.git] / drivers / net / vxlan.c
1 /*
2 * VXLAN: Virtual eXtensible Local Area Network
3 *
4 * Copyright (c) 2012-2013 Vyatta Inc.
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
11 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
12
13 #include <linux/kernel.h>
14 #include <linux/module.h>
15 #include <linux/errno.h>
16 #include <linux/slab.h>
17 #include <linux/udp.h>
18 #include <linux/igmp.h>
19 #include <linux/inetdevice.h>
20 #include <linux/if_ether.h>
21 #include <linux/ethtool.h>
22 #include <net/arp.h>
23 #include <net/ndisc.h>
24 #include <net/ip.h>
25 #include <net/icmp.h>
26 #include <net/rtnetlink.h>
27 #include <net/inet_ecn.h>
28 #include <net/net_namespace.h>
29 #include <net/netns/generic.h>
30 #include <net/tun_proto.h>
31 #include <net/vxlan.h>
32
33 #if IS_ENABLED(CONFIG_IPV6)
34 #include <net/ip6_tunnel.h>
35 #include <net/ip6_checksum.h>
36 #endif
37
38 #define VXLAN_VERSION "0.1"
39
40 #define PORT_HASH_BITS 8
41 #define PORT_HASH_SIZE (1<<PORT_HASH_BITS)
42 #define FDB_AGE_DEFAULT 300 /* 5 min */
43 #define FDB_AGE_INTERVAL (10 * HZ) /* rescan interval */
44
45 /* UDP port for VXLAN traffic.
46 * The IANA assigned port is 4789, but the Linux default is 8472
47 * for compatibility with early adopters.
48 */
49 static unsigned short vxlan_port __read_mostly = 8472;
50 module_param_named(udp_port, vxlan_port, ushort, 0444);
51 MODULE_PARM_DESC(udp_port, "Destination UDP port");
52
53 static bool log_ecn_error = true;
54 module_param(log_ecn_error, bool, 0644);
55 MODULE_PARM_DESC(log_ecn_error, "Log packets received with corrupted ECN");
56
57 static unsigned int vxlan_net_id;
58 static struct rtnl_link_ops vxlan_link_ops;
59
60 static const u8 all_zeros_mac[ETH_ALEN + 2];
61
62 static int vxlan_sock_add(struct vxlan_dev *vxlan);
63
64 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan);
65
66 /* per-network namespace private data for this module */
67 struct vxlan_net {
68 struct list_head vxlan_list;
69 struct hlist_head sock_list[PORT_HASH_SIZE];
70 spinlock_t sock_lock;
71 };
72
73 /* Forwarding table entry */
74 struct vxlan_fdb {
75 struct hlist_node hlist; /* linked list of entries */
76 struct rcu_head rcu;
77 unsigned long updated; /* jiffies */
78 unsigned long used;
79 struct list_head remotes;
80 u8 eth_addr[ETH_ALEN];
81 u16 state; /* see ndm_state */
82 __be32 vni;
83 u8 flags; /* see ndm_flags */
84 };
85
86 /* salt for hash table */
87 static u32 vxlan_salt __read_mostly;
88
89 static inline bool vxlan_collect_metadata(struct vxlan_sock *vs)
90 {
91 return vs->flags & VXLAN_F_COLLECT_METADATA ||
92 ip_tunnel_collect_metadata();
93 }
94
95 static struct ip_fan_map *vxlan_fan_find_map(struct vxlan_dev *vxlan, __be32 daddr)
96 {
97 struct ip_fan_map *fan_map;
98
99 rcu_read_lock();
100 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
101 if (fan_map->overlay ==
102 (daddr & inet_make_mask(fan_map->overlay_prefix))) {
103 rcu_read_unlock();
104 return fan_map;
105 }
106 }
107 rcu_read_unlock();
108
109 return NULL;
110 }
111
112 static void vxlan_fan_flush_map(struct vxlan_dev *vxlan)
113 {
114 struct ip_fan_map *fan_map;
115
116 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
117 list_del_rcu(&fan_map->list);
118 kfree_rcu(fan_map, rcu);
119 }
120 }
121
122 static int vxlan_fan_del_map(struct vxlan_dev *vxlan, __be32 overlay)
123 {
124 struct ip_fan_map *fan_map;
125
126 fan_map = vxlan_fan_find_map(vxlan, overlay);
127 if (!fan_map)
128 return -ENOENT;
129
130 list_del_rcu(&fan_map->list);
131 kfree_rcu(fan_map, rcu);
132
133 return 0;
134 }
135
136 static int vxlan_fan_add_map(struct vxlan_dev *vxlan, struct ifla_fan_map *map)
137 {
138 __be32 overlay_mask, underlay_mask;
139 struct ip_fan_map *fan_map;
140
141 overlay_mask = inet_make_mask(map->overlay_prefix);
142 underlay_mask = inet_make_mask(map->underlay_prefix);
143
144 netdev_dbg(vxlan->dev, "vfam: map: o %x/%d u %x/%d om %x um %x\n",
145 map->overlay, map->overlay_prefix,
146 map->underlay, map->underlay_prefix,
147 overlay_mask, underlay_mask);
148
149 if ((map->overlay & ~overlay_mask) || (map->underlay & ~underlay_mask))
150 return -EINVAL;
151
152 if (!(map->overlay & overlay_mask) && (map->underlay & underlay_mask))
153 return -EINVAL;
154
155 /* Special case: overlay 0 and underlay 0: flush all mappings */
156 if (!map->overlay && !map->underlay) {
157 vxlan_fan_flush_map(vxlan);
158 return 0;
159 }
160
161 /* Special case: overlay set and underlay 0: clear map for overlay */
162 if (!map->underlay)
163 return vxlan_fan_del_map(vxlan, map->overlay);
164
165 if (vxlan_fan_find_map(vxlan, map->overlay))
166 return -EEXIST;
167
168 fan_map = kmalloc(sizeof(*fan_map), GFP_KERNEL);
169 fan_map->underlay = map->underlay;
170 fan_map->overlay = map->overlay;
171 fan_map->underlay_prefix = map->underlay_prefix;
172 fan_map->overlay_mask = ntohl(overlay_mask);
173 fan_map->overlay_prefix = map->overlay_prefix;
174
175 list_add_tail_rcu(&fan_map->list, &vxlan->fan.fan_maps);
176
177 return 0;
178 }
179
180 static int vxlan_parse_fan_map(struct nlattr *data[], struct vxlan_dev *vxlan)
181 {
182 struct ifla_fan_map *map;
183 struct nlattr *attr;
184 int rem, rv;
185
186 nla_for_each_nested(attr, data[IFLA_IPTUN_FAN_MAP], rem) {
187 map = nla_data(attr);
188 rv = vxlan_fan_add_map(vxlan, map);
189 if (rv)
190 return rv;
191 }
192
193 return 0;
194 }
195
196 static int vxlan_fan_build_rdst(struct vxlan_dev *vxlan, struct sk_buff *skb,
197 struct vxlan_rdst *fan_rdst)
198 {
199 struct ip_fan_map *f_map;
200 union vxlan_addr *va;
201 u32 daddr, underlay;
202 struct arphdr *arp;
203 void *arp_ptr;
204 struct ethhdr *eth;
205 struct iphdr *iph;
206
207 eth = eth_hdr(skb);
208 switch (eth->h_proto) {
209 case htons(ETH_P_IP):
210 iph = ip_hdr(skb);
211 if (!iph)
212 return -EINVAL;
213 daddr = iph->daddr;
214 break;
215 case htons(ETH_P_ARP):
216 arp = arp_hdr(skb);
217 if (!arp)
218 return -EINVAL;
219 arp_ptr = arp + 1;
220 netdev_dbg(vxlan->dev,
221 "vfbr: arp sha %pM sip %pI4 tha %pM tip %pI4\n",
222 arp_ptr, arp_ptr + skb->dev->addr_len,
223 arp_ptr + skb->dev->addr_len + 4,
224 arp_ptr + (skb->dev->addr_len * 2) + 4);
225 arp_ptr += (skb->dev->addr_len * 2) + 4;
226 memcpy(&daddr, arp_ptr, 4);
227 break;
228 default:
229 netdev_dbg(vxlan->dev, "vfbr: unknown eth p %x\n", eth->h_proto);
230 return -EINVAL;
231 }
232
233 f_map = vxlan_fan_find_map(vxlan, daddr);
234 if (!f_map)
235 return -EINVAL;
236
237 daddr = ntohl(daddr);
238 underlay = ntohl(f_map->underlay);
239 if (!underlay)
240 return -EINVAL;
241
242 memset(fan_rdst, 0, sizeof(*fan_rdst));
243 va = &fan_rdst->remote_ip;
244 va->sa.sa_family = AF_INET;
245 fan_rdst->remote_vni = vxlan->default_dst.remote_vni;
246 va->sin.sin_addr.s_addr = htonl(underlay |
247 ((daddr & ~f_map->overlay_mask) >>
248 (32 - f_map->overlay_prefix -
249 (32 - f_map->underlay_prefix))));
250 netdev_dbg(vxlan->dev, "vfbr: daddr %x ul %x dst %x\n",
251 daddr, underlay, va->sin.sin_addr.s_addr);
252
253 return 0;
254 }
255
256 #if IS_ENABLED(CONFIG_IPV6)
257 static inline
258 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
259 {
260 if (a->sa.sa_family != b->sa.sa_family)
261 return false;
262 if (a->sa.sa_family == AF_INET6)
263 return ipv6_addr_equal(&a->sin6.sin6_addr, &b->sin6.sin6_addr);
264 else
265 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
266 }
267
268 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
269 {
270 if (ipa->sa.sa_family == AF_INET6)
271 return ipv6_addr_any(&ipa->sin6.sin6_addr);
272 else
273 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
274 }
275
276 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
277 {
278 if (ipa->sa.sa_family == AF_INET6)
279 return ipv6_addr_is_multicast(&ipa->sin6.sin6_addr);
280 else
281 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
282 }
283
284 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
285 {
286 if (nla_len(nla) >= sizeof(struct in6_addr)) {
287 ip->sin6.sin6_addr = nla_get_in6_addr(nla);
288 ip->sa.sa_family = AF_INET6;
289 return 0;
290 } else if (nla_len(nla) >= sizeof(__be32)) {
291 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
292 ip->sa.sa_family = AF_INET;
293 return 0;
294 } else {
295 return -EAFNOSUPPORT;
296 }
297 }
298
299 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
300 const union vxlan_addr *ip)
301 {
302 if (ip->sa.sa_family == AF_INET6)
303 return nla_put_in6_addr(skb, attr, &ip->sin6.sin6_addr);
304 else
305 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
306 }
307
308 #else /* !CONFIG_IPV6 */
309
310 static inline
311 bool vxlan_addr_equal(const union vxlan_addr *a, const union vxlan_addr *b)
312 {
313 return a->sin.sin_addr.s_addr == b->sin.sin_addr.s_addr;
314 }
315
316 static inline bool vxlan_addr_any(const union vxlan_addr *ipa)
317 {
318 return ipa->sin.sin_addr.s_addr == htonl(INADDR_ANY);
319 }
320
321 static inline bool vxlan_addr_multicast(const union vxlan_addr *ipa)
322 {
323 return IN_MULTICAST(ntohl(ipa->sin.sin_addr.s_addr));
324 }
325
326 static int vxlan_nla_get_addr(union vxlan_addr *ip, struct nlattr *nla)
327 {
328 if (nla_len(nla) >= sizeof(struct in6_addr)) {
329 return -EAFNOSUPPORT;
330 } else if (nla_len(nla) >= sizeof(__be32)) {
331 ip->sin.sin_addr.s_addr = nla_get_in_addr(nla);
332 ip->sa.sa_family = AF_INET;
333 return 0;
334 } else {
335 return -EAFNOSUPPORT;
336 }
337 }
338
339 static int vxlan_nla_put_addr(struct sk_buff *skb, int attr,
340 const union vxlan_addr *ip)
341 {
342 return nla_put_in_addr(skb, attr, ip->sin.sin_addr.s_addr);
343 }
344 #endif
345
346 /* Virtual Network hash table head */
347 static inline struct hlist_head *vni_head(struct vxlan_sock *vs, __be32 vni)
348 {
349 return &vs->vni_list[hash_32((__force u32)vni, VNI_HASH_BITS)];
350 }
351
352 /* Socket hash table head */
353 static inline struct hlist_head *vs_head(struct net *net, __be16 port)
354 {
355 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
356
357 return &vn->sock_list[hash_32(ntohs(port), PORT_HASH_BITS)];
358 }
359
360 /* First remote destination for a forwarding entry.
361 * Guaranteed to be non-NULL because remotes are never deleted.
362 */
363 static inline struct vxlan_rdst *first_remote_rcu(struct vxlan_fdb *fdb)
364 {
365 return list_entry_rcu(fdb->remotes.next, struct vxlan_rdst, list);
366 }
367
368 static inline struct vxlan_rdst *first_remote_rtnl(struct vxlan_fdb *fdb)
369 {
370 return list_first_entry(&fdb->remotes, struct vxlan_rdst, list);
371 }
372
373 /* Find VXLAN socket based on network namespace, address family and UDP port
374 * and enabled unshareable flags.
375 */
376 static struct vxlan_sock *vxlan_find_sock(struct net *net, sa_family_t family,
377 __be16 port, u32 flags)
378 {
379 struct vxlan_sock *vs;
380
381 flags &= VXLAN_F_RCV_FLAGS;
382
383 hlist_for_each_entry_rcu(vs, vs_head(net, port), hlist) {
384 if (inet_sk(vs->sock->sk)->inet_sport == port &&
385 vxlan_get_sk_family(vs) == family &&
386 vs->flags == flags)
387 return vs;
388 }
389 return NULL;
390 }
391
392 static struct vxlan_dev *vxlan_vs_find_vni(struct vxlan_sock *vs, int ifindex,
393 __be32 vni)
394 {
395 struct vxlan_dev_node *node;
396
397 /* For flow based devices, map all packets to VNI 0 */
398 if (vs->flags & VXLAN_F_COLLECT_METADATA)
399 vni = 0;
400
401 hlist_for_each_entry_rcu(node, vni_head(vs, vni), hlist) {
402 if (node->vxlan->default_dst.remote_vni != vni)
403 continue;
404
405 if (IS_ENABLED(CONFIG_IPV6)) {
406 const struct vxlan_config *cfg = &node->vxlan->cfg;
407
408 if ((cfg->flags & VXLAN_F_IPV6_LINKLOCAL) &&
409 cfg->remote_ifindex != ifindex)
410 continue;
411 }
412
413 return node->vxlan;
414 }
415
416 return NULL;
417 }
418
419 /* Look up VNI in a per net namespace table */
420 static struct vxlan_dev *vxlan_find_vni(struct net *net, int ifindex,
421 __be32 vni, sa_family_t family,
422 __be16 port, u32 flags)
423 {
424 struct vxlan_sock *vs;
425
426 vs = vxlan_find_sock(net, family, port, flags);
427 if (!vs)
428 return NULL;
429
430 return vxlan_vs_find_vni(vs, ifindex, vni);
431 }
432
433 /* Fill in neighbour message in skbuff. */
434 static int vxlan_fdb_info(struct sk_buff *skb, struct vxlan_dev *vxlan,
435 const struct vxlan_fdb *fdb,
436 u32 portid, u32 seq, int type, unsigned int flags,
437 const struct vxlan_rdst *rdst)
438 {
439 unsigned long now = jiffies;
440 struct nda_cacheinfo ci;
441 struct nlmsghdr *nlh;
442 struct ndmsg *ndm;
443 bool send_ip, send_eth;
444
445 nlh = nlmsg_put(skb, portid, seq, type, sizeof(*ndm), flags);
446 if (nlh == NULL)
447 return -EMSGSIZE;
448
449 ndm = nlmsg_data(nlh);
450 memset(ndm, 0, sizeof(*ndm));
451
452 send_eth = send_ip = true;
453
454 if (type == RTM_GETNEIGH) {
455 send_ip = !vxlan_addr_any(&rdst->remote_ip);
456 send_eth = !is_zero_ether_addr(fdb->eth_addr);
457 ndm->ndm_family = send_ip ? rdst->remote_ip.sa.sa_family : AF_INET;
458 } else
459 ndm->ndm_family = AF_BRIDGE;
460 ndm->ndm_state = fdb->state;
461 ndm->ndm_ifindex = vxlan->dev->ifindex;
462 ndm->ndm_flags = fdb->flags;
463 ndm->ndm_type = RTN_UNICAST;
464
465 if (!net_eq(dev_net(vxlan->dev), vxlan->net) &&
466 nla_put_s32(skb, NDA_LINK_NETNSID,
467 peernet2id(dev_net(vxlan->dev), vxlan->net)))
468 goto nla_put_failure;
469
470 if (send_eth && nla_put(skb, NDA_LLADDR, ETH_ALEN, &fdb->eth_addr))
471 goto nla_put_failure;
472
473 if (send_ip && vxlan_nla_put_addr(skb, NDA_DST, &rdst->remote_ip))
474 goto nla_put_failure;
475
476 if (rdst->remote_port && rdst->remote_port != vxlan->cfg.dst_port &&
477 nla_put_be16(skb, NDA_PORT, rdst->remote_port))
478 goto nla_put_failure;
479 if (rdst->remote_vni != vxlan->default_dst.remote_vni &&
480 nla_put_u32(skb, NDA_VNI, be32_to_cpu(rdst->remote_vni)))
481 goto nla_put_failure;
482 if ((vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) && fdb->vni &&
483 nla_put_u32(skb, NDA_SRC_VNI,
484 be32_to_cpu(fdb->vni)))
485 goto nla_put_failure;
486 if (rdst->remote_ifindex &&
487 nla_put_u32(skb, NDA_IFINDEX, rdst->remote_ifindex))
488 goto nla_put_failure;
489
490 ci.ndm_used = jiffies_to_clock_t(now - fdb->used);
491 ci.ndm_confirmed = 0;
492 ci.ndm_updated = jiffies_to_clock_t(now - fdb->updated);
493 ci.ndm_refcnt = 0;
494
495 if (nla_put(skb, NDA_CACHEINFO, sizeof(ci), &ci))
496 goto nla_put_failure;
497
498 nlmsg_end(skb, nlh);
499 return 0;
500
501 nla_put_failure:
502 nlmsg_cancel(skb, nlh);
503 return -EMSGSIZE;
504 }
505
506 static inline size_t vxlan_nlmsg_size(void)
507 {
508 return NLMSG_ALIGN(sizeof(struct ndmsg))
509 + nla_total_size(ETH_ALEN) /* NDA_LLADDR */
510 + nla_total_size(sizeof(struct in6_addr)) /* NDA_DST */
511 + nla_total_size(sizeof(__be16)) /* NDA_PORT */
512 + nla_total_size(sizeof(__be32)) /* NDA_VNI */
513 + nla_total_size(sizeof(__u32)) /* NDA_IFINDEX */
514 + nla_total_size(sizeof(__s32)) /* NDA_LINK_NETNSID */
515 + nla_total_size(sizeof(struct nda_cacheinfo));
516 }
517
518 static void vxlan_fdb_notify(struct vxlan_dev *vxlan, struct vxlan_fdb *fdb,
519 struct vxlan_rdst *rd, int type)
520 {
521 struct net *net = dev_net(vxlan->dev);
522 struct sk_buff *skb;
523 int err = -ENOBUFS;
524
525 skb = nlmsg_new(vxlan_nlmsg_size(), GFP_ATOMIC);
526 if (skb == NULL)
527 goto errout;
528
529 err = vxlan_fdb_info(skb, vxlan, fdb, 0, 0, type, 0, rd);
530 if (err < 0) {
531 /* -EMSGSIZE implies BUG in vxlan_nlmsg_size() */
532 WARN_ON(err == -EMSGSIZE);
533 kfree_skb(skb);
534 goto errout;
535 }
536
537 rtnl_notify(skb, net, 0, RTNLGRP_NEIGH, NULL, GFP_ATOMIC);
538 return;
539 errout:
540 if (err < 0)
541 rtnl_set_sk_err(net, RTNLGRP_NEIGH, err);
542 }
543
544 static void vxlan_ip_miss(struct net_device *dev, union vxlan_addr *ipa)
545 {
546 struct vxlan_dev *vxlan = netdev_priv(dev);
547 struct vxlan_fdb f = {
548 .state = NUD_STALE,
549 };
550 struct vxlan_rdst remote = {
551 .remote_ip = *ipa, /* goes to NDA_DST */
552 .remote_vni = cpu_to_be32(VXLAN_N_VID),
553 };
554
555 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
556 }
557
558 static void vxlan_fdb_miss(struct vxlan_dev *vxlan, const u8 eth_addr[ETH_ALEN])
559 {
560 struct vxlan_fdb f = {
561 .state = NUD_STALE,
562 };
563 struct vxlan_rdst remote = { };
564
565 memcpy(f.eth_addr, eth_addr, ETH_ALEN);
566
567 vxlan_fdb_notify(vxlan, &f, &remote, RTM_GETNEIGH);
568 }
569
570 /* Hash Ethernet address */
571 static u32 eth_hash(const unsigned char *addr)
572 {
573 u64 value = get_unaligned((u64 *)addr);
574
575 /* only want 6 bytes */
576 #ifdef __BIG_ENDIAN
577 value >>= 16;
578 #else
579 value <<= 16;
580 #endif
581 return hash_64(value, FDB_HASH_BITS);
582 }
583
584 static u32 eth_vni_hash(const unsigned char *addr, __be32 vni)
585 {
586 /* use 1 byte of OUI and 3 bytes of NIC */
587 u32 key = get_unaligned((u32 *)(addr + 2));
588
589 return jhash_2words(key, vni, vxlan_salt) & (FDB_HASH_SIZE - 1);
590 }
591
592 /* Hash chain to use given mac address */
593 static inline struct hlist_head *vxlan_fdb_head(struct vxlan_dev *vxlan,
594 const u8 *mac, __be32 vni)
595 {
596 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)
597 return &vxlan->fdb_head[eth_vni_hash(mac, vni)];
598 else
599 return &vxlan->fdb_head[eth_hash(mac)];
600 }
601
602 /* Look up Ethernet address in forwarding table */
603 static struct vxlan_fdb *__vxlan_find_mac(struct vxlan_dev *vxlan,
604 const u8 *mac, __be32 vni)
605 {
606 struct hlist_head *head = vxlan_fdb_head(vxlan, mac, vni);
607 struct vxlan_fdb *f;
608
609 hlist_for_each_entry_rcu(f, head, hlist) {
610 if (ether_addr_equal(mac, f->eth_addr)) {
611 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
612 if (vni == f->vni)
613 return f;
614 } else {
615 return f;
616 }
617 }
618 }
619
620 return NULL;
621 }
622
623 static struct vxlan_fdb *vxlan_find_mac(struct vxlan_dev *vxlan,
624 const u8 *mac, __be32 vni)
625 {
626 struct vxlan_fdb *f;
627
628 f = __vxlan_find_mac(vxlan, mac, vni);
629 if (f)
630 f->used = jiffies;
631
632 return f;
633 }
634
635 /* caller should hold vxlan->hash_lock */
636 static struct vxlan_rdst *vxlan_fdb_find_rdst(struct vxlan_fdb *f,
637 union vxlan_addr *ip, __be16 port,
638 __be32 vni, __u32 ifindex)
639 {
640 struct vxlan_rdst *rd;
641
642 list_for_each_entry(rd, &f->remotes, list) {
643 if (vxlan_addr_equal(&rd->remote_ip, ip) &&
644 rd->remote_port == port &&
645 rd->remote_vni == vni &&
646 rd->remote_ifindex == ifindex)
647 return rd;
648 }
649
650 return NULL;
651 }
652
653 /* Replace destination of unicast mac */
654 static int vxlan_fdb_replace(struct vxlan_fdb *f,
655 union vxlan_addr *ip, __be16 port, __be32 vni,
656 __u32 ifindex)
657 {
658 struct vxlan_rdst *rd;
659
660 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
661 if (rd)
662 return 0;
663
664 rd = list_first_entry_or_null(&f->remotes, struct vxlan_rdst, list);
665 if (!rd)
666 return 0;
667
668 dst_cache_reset(&rd->dst_cache);
669 rd->remote_ip = *ip;
670 rd->remote_port = port;
671 rd->remote_vni = vni;
672 rd->remote_ifindex = ifindex;
673 return 1;
674 }
675
676 /* Add/update destinations for multicast */
677 static int vxlan_fdb_append(struct vxlan_fdb *f,
678 union vxlan_addr *ip, __be16 port, __be32 vni,
679 __u32 ifindex, struct vxlan_rdst **rdp)
680 {
681 struct vxlan_rdst *rd;
682
683 rd = vxlan_fdb_find_rdst(f, ip, port, vni, ifindex);
684 if (rd)
685 return 0;
686
687 rd = kmalloc(sizeof(*rd), GFP_ATOMIC);
688 if (rd == NULL)
689 return -ENOBUFS;
690
691 if (dst_cache_init(&rd->dst_cache, GFP_ATOMIC)) {
692 kfree(rd);
693 return -ENOBUFS;
694 }
695
696 rd->remote_ip = *ip;
697 rd->remote_port = port;
698 rd->remote_vni = vni;
699 rd->remote_ifindex = ifindex;
700
701 list_add_tail_rcu(&rd->list, &f->remotes);
702
703 *rdp = rd;
704 return 1;
705 }
706
707 static struct vxlanhdr *vxlan_gro_remcsum(struct sk_buff *skb,
708 unsigned int off,
709 struct vxlanhdr *vh, size_t hdrlen,
710 __be32 vni_field,
711 struct gro_remcsum *grc,
712 bool nopartial)
713 {
714 size_t start, offset;
715
716 if (skb->remcsum_offload)
717 return vh;
718
719 if (!NAPI_GRO_CB(skb)->csum_valid)
720 return NULL;
721
722 start = vxlan_rco_start(vni_field);
723 offset = start + vxlan_rco_offset(vni_field);
724
725 vh = skb_gro_remcsum_process(skb, (void *)vh, off, hdrlen,
726 start, offset, grc, nopartial);
727
728 skb->remcsum_offload = 1;
729
730 return vh;
731 }
732
733 static struct sk_buff **vxlan_gro_receive(struct sock *sk,
734 struct sk_buff **head,
735 struct sk_buff *skb)
736 {
737 struct sk_buff *p, **pp = NULL;
738 struct vxlanhdr *vh, *vh2;
739 unsigned int hlen, off_vx;
740 int flush = 1;
741 struct vxlan_sock *vs = rcu_dereference_sk_user_data(sk);
742 __be32 flags;
743 struct gro_remcsum grc;
744
745 skb_gro_remcsum_init(&grc);
746
747 off_vx = skb_gro_offset(skb);
748 hlen = off_vx + sizeof(*vh);
749 vh = skb_gro_header_fast(skb, off_vx);
750 if (skb_gro_header_hard(skb, hlen)) {
751 vh = skb_gro_header_slow(skb, hlen, off_vx);
752 if (unlikely(!vh))
753 goto out;
754 }
755
756 skb_gro_postpull_rcsum(skb, vh, sizeof(struct vxlanhdr));
757
758 flags = vh->vx_flags;
759
760 if ((flags & VXLAN_HF_RCO) && (vs->flags & VXLAN_F_REMCSUM_RX)) {
761 vh = vxlan_gro_remcsum(skb, off_vx, vh, sizeof(struct vxlanhdr),
762 vh->vx_vni, &grc,
763 !!(vs->flags &
764 VXLAN_F_REMCSUM_NOPARTIAL));
765
766 if (!vh)
767 goto out;
768 }
769
770 skb_gro_pull(skb, sizeof(struct vxlanhdr)); /* pull vxlan header */
771
772 for (p = *head; p; p = p->next) {
773 if (!NAPI_GRO_CB(p)->same_flow)
774 continue;
775
776 vh2 = (struct vxlanhdr *)(p->data + off_vx);
777 if (vh->vx_flags != vh2->vx_flags ||
778 vh->vx_vni != vh2->vx_vni) {
779 NAPI_GRO_CB(p)->same_flow = 0;
780 continue;
781 }
782 }
783
784 pp = call_gro_receive(eth_gro_receive, head, skb);
785 flush = 0;
786
787 out:
788 skb_gro_remcsum_cleanup(skb, &grc);
789 skb->remcsum_offload = 0;
790 NAPI_GRO_CB(skb)->flush |= flush;
791
792 return pp;
793 }
794
795 static int vxlan_gro_complete(struct sock *sk, struct sk_buff *skb, int nhoff)
796 {
797 /* Sets 'skb->inner_mac_header' since we are always called with
798 * 'skb->encapsulation' set.
799 */
800 return eth_gro_complete(skb, nhoff + sizeof(struct vxlanhdr));
801 }
802
803 /* Add new entry to forwarding table -- assumes lock held */
804 static int vxlan_fdb_create(struct vxlan_dev *vxlan,
805 const u8 *mac, union vxlan_addr *ip,
806 __u16 state, __u16 flags,
807 __be16 port, __be32 src_vni, __be32 vni,
808 __u32 ifindex, __u8 ndm_flags)
809 {
810 struct vxlan_rdst *rd = NULL;
811 struct vxlan_fdb *f;
812 int notify = 0;
813 int rc;
814
815 f = __vxlan_find_mac(vxlan, mac, src_vni);
816 if (f) {
817 if (flags & NLM_F_EXCL) {
818 netdev_dbg(vxlan->dev,
819 "lost race to create %pM\n", mac);
820 return -EEXIST;
821 }
822 if (f->state != state) {
823 f->state = state;
824 f->updated = jiffies;
825 notify = 1;
826 }
827 if (f->flags != ndm_flags) {
828 f->flags = ndm_flags;
829 f->updated = jiffies;
830 notify = 1;
831 }
832 if ((flags & NLM_F_REPLACE)) {
833 /* Only change unicasts */
834 if (!(is_multicast_ether_addr(f->eth_addr) ||
835 is_zero_ether_addr(f->eth_addr))) {
836 notify |= vxlan_fdb_replace(f, ip, port, vni,
837 ifindex);
838 } else
839 return -EOPNOTSUPP;
840 }
841 if ((flags & NLM_F_APPEND) &&
842 (is_multicast_ether_addr(f->eth_addr) ||
843 is_zero_ether_addr(f->eth_addr))) {
844 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
845
846 if (rc < 0)
847 return rc;
848 notify |= rc;
849 }
850 } else {
851 if (!(flags & NLM_F_CREATE))
852 return -ENOENT;
853
854 if (vxlan->cfg.addrmax &&
855 vxlan->addrcnt >= vxlan->cfg.addrmax)
856 return -ENOSPC;
857
858 /* Disallow replace to add a multicast entry */
859 if ((flags & NLM_F_REPLACE) &&
860 (is_multicast_ether_addr(mac) || is_zero_ether_addr(mac)))
861 return -EOPNOTSUPP;
862
863 netdev_dbg(vxlan->dev, "add %pM -> %pIS\n", mac, ip);
864 f = kmalloc(sizeof(*f), GFP_ATOMIC);
865 if (!f)
866 return -ENOMEM;
867
868 notify = 1;
869 f->state = state;
870 f->flags = ndm_flags;
871 f->updated = f->used = jiffies;
872 f->vni = src_vni;
873 INIT_LIST_HEAD(&f->remotes);
874 memcpy(f->eth_addr, mac, ETH_ALEN);
875
876 rc = vxlan_fdb_append(f, ip, port, vni, ifindex, &rd);
877 if (rc < 0) {
878 kfree(f);
879 return rc;
880 }
881
882 ++vxlan->addrcnt;
883 hlist_add_head_rcu(&f->hlist,
884 vxlan_fdb_head(vxlan, mac, src_vni));
885 }
886
887 if (notify) {
888 if (rd == NULL)
889 rd = first_remote_rtnl(f);
890 vxlan_fdb_notify(vxlan, f, rd, RTM_NEWNEIGH);
891 }
892
893 return 0;
894 }
895
896 static void vxlan_fdb_free(struct rcu_head *head)
897 {
898 struct vxlan_fdb *f = container_of(head, struct vxlan_fdb, rcu);
899 struct vxlan_rdst *rd, *nd;
900
901 list_for_each_entry_safe(rd, nd, &f->remotes, list) {
902 dst_cache_destroy(&rd->dst_cache);
903 kfree(rd);
904 }
905 kfree(f);
906 }
907
908 static void vxlan_fdb_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f)
909 {
910 netdev_dbg(vxlan->dev,
911 "delete %pM\n", f->eth_addr);
912
913 --vxlan->addrcnt;
914 vxlan_fdb_notify(vxlan, f, first_remote_rtnl(f), RTM_DELNEIGH);
915
916 hlist_del_rcu(&f->hlist);
917 call_rcu(&f->rcu, vxlan_fdb_free);
918 }
919
920 static void vxlan_dst_free(struct rcu_head *head)
921 {
922 struct vxlan_rdst *rd = container_of(head, struct vxlan_rdst, rcu);
923
924 dst_cache_destroy(&rd->dst_cache);
925 kfree(rd);
926 }
927
928 static void vxlan_fdb_dst_destroy(struct vxlan_dev *vxlan, struct vxlan_fdb *f,
929 struct vxlan_rdst *rd)
930 {
931 list_del_rcu(&rd->list);
932 vxlan_fdb_notify(vxlan, f, rd, RTM_DELNEIGH);
933 call_rcu(&rd->rcu, vxlan_dst_free);
934 }
935
936 static int vxlan_fdb_parse(struct nlattr *tb[], struct vxlan_dev *vxlan,
937 union vxlan_addr *ip, __be16 *port, __be32 *src_vni,
938 __be32 *vni, u32 *ifindex)
939 {
940 struct net *net = dev_net(vxlan->dev);
941 int err;
942
943 if (tb[NDA_DST]) {
944 err = vxlan_nla_get_addr(ip, tb[NDA_DST]);
945 if (err)
946 return err;
947 } else {
948 union vxlan_addr *remote = &vxlan->default_dst.remote_ip;
949 if (remote->sa.sa_family == AF_INET) {
950 ip->sin.sin_addr.s_addr = htonl(INADDR_ANY);
951 ip->sa.sa_family = AF_INET;
952 #if IS_ENABLED(CONFIG_IPV6)
953 } else {
954 ip->sin6.sin6_addr = in6addr_any;
955 ip->sa.sa_family = AF_INET6;
956 #endif
957 }
958 }
959
960 if (tb[NDA_PORT]) {
961 if (nla_len(tb[NDA_PORT]) != sizeof(__be16))
962 return -EINVAL;
963 *port = nla_get_be16(tb[NDA_PORT]);
964 } else {
965 *port = vxlan->cfg.dst_port;
966 }
967
968 if (tb[NDA_VNI]) {
969 if (nla_len(tb[NDA_VNI]) != sizeof(u32))
970 return -EINVAL;
971 *vni = cpu_to_be32(nla_get_u32(tb[NDA_VNI]));
972 } else {
973 *vni = vxlan->default_dst.remote_vni;
974 }
975
976 if (tb[NDA_SRC_VNI]) {
977 if (nla_len(tb[NDA_SRC_VNI]) != sizeof(u32))
978 return -EINVAL;
979 *src_vni = cpu_to_be32(nla_get_u32(tb[NDA_SRC_VNI]));
980 } else {
981 *src_vni = vxlan->default_dst.remote_vni;
982 }
983
984 if (tb[NDA_IFINDEX]) {
985 struct net_device *tdev;
986
987 if (nla_len(tb[NDA_IFINDEX]) != sizeof(u32))
988 return -EINVAL;
989 *ifindex = nla_get_u32(tb[NDA_IFINDEX]);
990 tdev = __dev_get_by_index(net, *ifindex);
991 if (!tdev)
992 return -EADDRNOTAVAIL;
993 } else {
994 *ifindex = 0;
995 }
996
997 return 0;
998 }
999
1000 /* Add static entry (via netlink) */
1001 static int vxlan_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
1002 struct net_device *dev,
1003 const unsigned char *addr, u16 vid, u16 flags)
1004 {
1005 struct vxlan_dev *vxlan = netdev_priv(dev);
1006 /* struct net *net = dev_net(vxlan->dev); */
1007 union vxlan_addr ip;
1008 __be16 port;
1009 __be32 src_vni, vni;
1010 u32 ifindex;
1011 int err;
1012
1013 if (!(ndm->ndm_state & (NUD_PERMANENT|NUD_REACHABLE))) {
1014 pr_info("RTM_NEWNEIGH with invalid state %#x\n",
1015 ndm->ndm_state);
1016 return -EINVAL;
1017 }
1018
1019 if (tb[NDA_DST] == NULL)
1020 return -EINVAL;
1021
1022 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1023 if (err)
1024 return err;
1025
1026 if (vxlan->default_dst.remote_ip.sa.sa_family != ip.sa.sa_family)
1027 return -EAFNOSUPPORT;
1028
1029 spin_lock_bh(&vxlan->hash_lock);
1030 err = vxlan_fdb_create(vxlan, addr, &ip, ndm->ndm_state, flags,
1031 port, src_vni, vni, ifindex, ndm->ndm_flags);
1032 spin_unlock_bh(&vxlan->hash_lock);
1033
1034 return err;
1035 }
1036
1037 static int __vxlan_fdb_delete(struct vxlan_dev *vxlan,
1038 const unsigned char *addr, union vxlan_addr ip,
1039 __be16 port, __be32 src_vni, __be32 vni,
1040 u32 ifindex, u16 vid)
1041 {
1042 struct vxlan_fdb *f;
1043 struct vxlan_rdst *rd = NULL;
1044 int err = -ENOENT;
1045
1046 f = vxlan_find_mac(vxlan, addr, src_vni);
1047 if (!f)
1048 return err;
1049
1050 if (!vxlan_addr_any(&ip)) {
1051 rd = vxlan_fdb_find_rdst(f, &ip, port, vni, ifindex);
1052 if (!rd)
1053 goto out;
1054 }
1055
1056 /* remove a destination if it's not the only one on the list,
1057 * otherwise destroy the fdb entry
1058 */
1059 if (rd && !list_is_singular(&f->remotes)) {
1060 vxlan_fdb_dst_destroy(vxlan, f, rd);
1061 goto out;
1062 }
1063
1064 vxlan_fdb_destroy(vxlan, f);
1065
1066 out:
1067 return 0;
1068 }
1069
1070 /* Delete entry (via netlink) */
1071 static int vxlan_fdb_delete(struct ndmsg *ndm, struct nlattr *tb[],
1072 struct net_device *dev,
1073 const unsigned char *addr, u16 vid)
1074 {
1075 struct vxlan_dev *vxlan = netdev_priv(dev);
1076 union vxlan_addr ip;
1077 __be32 src_vni, vni;
1078 __be16 port;
1079 u32 ifindex;
1080 int err;
1081
1082 err = vxlan_fdb_parse(tb, vxlan, &ip, &port, &src_vni, &vni, &ifindex);
1083 if (err)
1084 return err;
1085
1086 spin_lock_bh(&vxlan->hash_lock);
1087 err = __vxlan_fdb_delete(vxlan, addr, ip, port, src_vni, vni, ifindex,
1088 vid);
1089 spin_unlock_bh(&vxlan->hash_lock);
1090
1091 return err;
1092 }
1093
1094 /* Dump forwarding table */
1095 static int vxlan_fdb_dump(struct sk_buff *skb, struct netlink_callback *cb,
1096 struct net_device *dev,
1097 struct net_device *filter_dev, int *idx)
1098 {
1099 struct vxlan_dev *vxlan = netdev_priv(dev);
1100 unsigned int h;
1101 int err = 0;
1102
1103 for (h = 0; h < FDB_HASH_SIZE; ++h) {
1104 struct vxlan_fdb *f;
1105
1106 hlist_for_each_entry_rcu(f, &vxlan->fdb_head[h], hlist) {
1107 struct vxlan_rdst *rd;
1108
1109 list_for_each_entry_rcu(rd, &f->remotes, list) {
1110 if (*idx < cb->args[2])
1111 goto skip;
1112
1113 err = vxlan_fdb_info(skb, vxlan, f,
1114 NETLINK_CB(cb->skb).portid,
1115 cb->nlh->nlmsg_seq,
1116 RTM_NEWNEIGH,
1117 NLM_F_MULTI, rd);
1118 if (err < 0)
1119 goto out;
1120 skip:
1121 *idx += 1;
1122 }
1123 }
1124 }
1125 out:
1126 return err;
1127 }
1128
1129 /* Watch incoming packets to learn mapping between Ethernet address
1130 * and Tunnel endpoint.
1131 * Return true if packet is bogus and should be dropped.
1132 */
1133 static bool vxlan_snoop(struct net_device *dev,
1134 union vxlan_addr *src_ip, const u8 *src_mac,
1135 u32 src_ifindex, __be32 vni)
1136 {
1137 struct vxlan_dev *vxlan = netdev_priv(dev);
1138 struct vxlan_fdb *f;
1139 u32 ifindex = 0;
1140
1141 #if IS_ENABLED(CONFIG_IPV6)
1142 if (src_ip->sa.sa_family == AF_INET6 &&
1143 (ipv6_addr_type(&src_ip->sin6.sin6_addr) & IPV6_ADDR_LINKLOCAL))
1144 ifindex = src_ifindex;
1145 #endif
1146
1147 f = vxlan_find_mac(vxlan, src_mac, vni);
1148 if (likely(f)) {
1149 struct vxlan_rdst *rdst = first_remote_rcu(f);
1150
1151 if (likely(vxlan_addr_equal(&rdst->remote_ip, src_ip) &&
1152 rdst->remote_ifindex == ifindex))
1153 return false;
1154
1155 /* Don't migrate static entries, drop packets */
1156 if (f->state & (NUD_PERMANENT | NUD_NOARP))
1157 return true;
1158
1159 if (net_ratelimit())
1160 netdev_info(dev,
1161 "%pM migrated from %pIS to %pIS\n",
1162 src_mac, &rdst->remote_ip.sa, &src_ip->sa);
1163
1164 rdst->remote_ip = *src_ip;
1165 f->updated = jiffies;
1166 vxlan_fdb_notify(vxlan, f, rdst, RTM_NEWNEIGH);
1167 } else {
1168 /* learned new entry */
1169 spin_lock(&vxlan->hash_lock);
1170
1171 /* close off race between vxlan_flush and incoming packets */
1172 if (netif_running(dev))
1173 vxlan_fdb_create(vxlan, src_mac, src_ip,
1174 NUD_REACHABLE,
1175 NLM_F_EXCL|NLM_F_CREATE,
1176 vxlan->cfg.dst_port,
1177 vni,
1178 vxlan->default_dst.remote_vni,
1179 ifindex, NTF_SELF);
1180 spin_unlock(&vxlan->hash_lock);
1181 }
1182
1183 return false;
1184 }
1185
1186 /* See if multicast group is already in use by other ID */
1187 static bool vxlan_group_used(struct vxlan_net *vn, struct vxlan_dev *dev)
1188 {
1189 struct vxlan_dev *vxlan;
1190 struct vxlan_sock *sock4;
1191 #if IS_ENABLED(CONFIG_IPV6)
1192 struct vxlan_sock *sock6;
1193 #endif
1194 unsigned short family = dev->default_dst.remote_ip.sa.sa_family;
1195
1196 sock4 = rtnl_dereference(dev->vn4_sock);
1197
1198 /* The vxlan_sock is only used by dev, leaving group has
1199 * no effect on other vxlan devices.
1200 */
1201 if (family == AF_INET && sock4 && refcount_read(&sock4->refcnt) == 1)
1202 return false;
1203 #if IS_ENABLED(CONFIG_IPV6)
1204 sock6 = rtnl_dereference(dev->vn6_sock);
1205 if (family == AF_INET6 && sock6 && refcount_read(&sock6->refcnt) == 1)
1206 return false;
1207 #endif
1208
1209 list_for_each_entry(vxlan, &vn->vxlan_list, next) {
1210 if (!netif_running(vxlan->dev) || vxlan == dev)
1211 continue;
1212
1213 if (family == AF_INET &&
1214 rtnl_dereference(vxlan->vn4_sock) != sock4)
1215 continue;
1216 #if IS_ENABLED(CONFIG_IPV6)
1217 if (family == AF_INET6 &&
1218 rtnl_dereference(vxlan->vn6_sock) != sock6)
1219 continue;
1220 #endif
1221
1222 if (!vxlan_addr_equal(&vxlan->default_dst.remote_ip,
1223 &dev->default_dst.remote_ip))
1224 continue;
1225
1226 if (vxlan->default_dst.remote_ifindex !=
1227 dev->default_dst.remote_ifindex)
1228 continue;
1229
1230 return true;
1231 }
1232
1233 return false;
1234 }
1235
1236 static bool __vxlan_sock_release_prep(struct vxlan_sock *vs)
1237 {
1238 struct vxlan_net *vn;
1239
1240 if (!vs)
1241 return false;
1242 if (!refcount_dec_and_test(&vs->refcnt))
1243 return false;
1244
1245 vn = net_generic(sock_net(vs->sock->sk), vxlan_net_id);
1246 spin_lock(&vn->sock_lock);
1247 hlist_del_rcu(&vs->hlist);
1248 udp_tunnel_notify_del_rx_port(vs->sock,
1249 (vs->flags & VXLAN_F_GPE) ?
1250 UDP_TUNNEL_TYPE_VXLAN_GPE :
1251 UDP_TUNNEL_TYPE_VXLAN);
1252 spin_unlock(&vn->sock_lock);
1253
1254 return true;
1255 }
1256
1257 static void vxlan_sock_release(struct vxlan_dev *vxlan)
1258 {
1259 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1260 #if IS_ENABLED(CONFIG_IPV6)
1261 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1262
1263 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
1264 #endif
1265
1266 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
1267 synchronize_net();
1268
1269 vxlan_vs_del_dev(vxlan);
1270
1271 if (__vxlan_sock_release_prep(sock4)) {
1272 udp_tunnel_sock_release(sock4->sock);
1273 kfree(sock4);
1274 }
1275
1276 #if IS_ENABLED(CONFIG_IPV6)
1277 if (__vxlan_sock_release_prep(sock6)) {
1278 udp_tunnel_sock_release(sock6->sock);
1279 kfree(sock6);
1280 }
1281 #endif
1282 }
1283
1284 /* Update multicast group membership when first VNI on
1285 * multicast address is brought up
1286 */
1287 static int vxlan_igmp_join(struct vxlan_dev *vxlan)
1288 {
1289 struct sock *sk;
1290 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1291 int ifindex = vxlan->default_dst.remote_ifindex;
1292 int ret = -EINVAL;
1293
1294 if (ip->sa.sa_family == AF_INET) {
1295 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1296 struct ip_mreqn mreq = {
1297 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1298 .imr_ifindex = ifindex,
1299 };
1300
1301 sk = sock4->sock->sk;
1302 lock_sock(sk);
1303 ret = ip_mc_join_group(sk, &mreq);
1304 release_sock(sk);
1305 #if IS_ENABLED(CONFIG_IPV6)
1306 } else {
1307 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1308
1309 sk = sock6->sock->sk;
1310 lock_sock(sk);
1311 ret = ipv6_stub->ipv6_sock_mc_join(sk, ifindex,
1312 &ip->sin6.sin6_addr);
1313 release_sock(sk);
1314 #endif
1315 }
1316
1317 return ret;
1318 }
1319
1320 /* Inverse of vxlan_igmp_join when last VNI is brought down */
1321 static int vxlan_igmp_leave(struct vxlan_dev *vxlan)
1322 {
1323 struct sock *sk;
1324 union vxlan_addr *ip = &vxlan->default_dst.remote_ip;
1325 int ifindex = vxlan->default_dst.remote_ifindex;
1326 int ret = -EINVAL;
1327
1328 if (ip->sa.sa_family == AF_INET) {
1329 struct vxlan_sock *sock4 = rtnl_dereference(vxlan->vn4_sock);
1330 struct ip_mreqn mreq = {
1331 .imr_multiaddr.s_addr = ip->sin.sin_addr.s_addr,
1332 .imr_ifindex = ifindex,
1333 };
1334
1335 sk = sock4->sock->sk;
1336 lock_sock(sk);
1337 ret = ip_mc_leave_group(sk, &mreq);
1338 release_sock(sk);
1339 #if IS_ENABLED(CONFIG_IPV6)
1340 } else {
1341 struct vxlan_sock *sock6 = rtnl_dereference(vxlan->vn6_sock);
1342
1343 sk = sock6->sock->sk;
1344 lock_sock(sk);
1345 ret = ipv6_stub->ipv6_sock_mc_drop(sk, ifindex,
1346 &ip->sin6.sin6_addr);
1347 release_sock(sk);
1348 #endif
1349 }
1350
1351 return ret;
1352 }
1353
1354 static bool vxlan_remcsum(struct vxlanhdr *unparsed,
1355 struct sk_buff *skb, u32 vxflags)
1356 {
1357 size_t start, offset;
1358
1359 if (!(unparsed->vx_flags & VXLAN_HF_RCO) || skb->remcsum_offload)
1360 goto out;
1361
1362 start = vxlan_rco_start(unparsed->vx_vni);
1363 offset = start + vxlan_rco_offset(unparsed->vx_vni);
1364
1365 if (!pskb_may_pull(skb, offset + sizeof(u16)))
1366 return false;
1367
1368 skb_remcsum_process(skb, (void *)(vxlan_hdr(skb) + 1), start, offset,
1369 !!(vxflags & VXLAN_F_REMCSUM_NOPARTIAL));
1370 out:
1371 unparsed->vx_flags &= ~VXLAN_HF_RCO;
1372 unparsed->vx_vni &= VXLAN_VNI_MASK;
1373 return true;
1374 }
1375
1376 static void vxlan_parse_gbp_hdr(struct vxlanhdr *unparsed,
1377 struct sk_buff *skb, u32 vxflags,
1378 struct vxlan_metadata *md)
1379 {
1380 struct vxlanhdr_gbp *gbp = (struct vxlanhdr_gbp *)unparsed;
1381 struct metadata_dst *tun_dst;
1382
1383 if (!(unparsed->vx_flags & VXLAN_HF_GBP))
1384 goto out;
1385
1386 md->gbp = ntohs(gbp->policy_id);
1387
1388 tun_dst = (struct metadata_dst *)skb_dst(skb);
1389 if (tun_dst) {
1390 tun_dst->u.tun_info.key.tun_flags |= TUNNEL_VXLAN_OPT;
1391 tun_dst->u.tun_info.options_len = sizeof(*md);
1392 }
1393 if (gbp->dont_learn)
1394 md->gbp |= VXLAN_GBP_DONT_LEARN;
1395
1396 if (gbp->policy_applied)
1397 md->gbp |= VXLAN_GBP_POLICY_APPLIED;
1398
1399 /* In flow-based mode, GBP is carried in dst_metadata */
1400 if (!(vxflags & VXLAN_F_COLLECT_METADATA))
1401 skb->mark = md->gbp;
1402 out:
1403 unparsed->vx_flags &= ~VXLAN_GBP_USED_BITS;
1404 }
1405
1406 static bool vxlan_parse_gpe_hdr(struct vxlanhdr *unparsed,
1407 __be16 *protocol,
1408 struct sk_buff *skb, u32 vxflags)
1409 {
1410 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)unparsed;
1411
1412 /* Need to have Next Protocol set for interfaces in GPE mode. */
1413 if (!gpe->np_applied)
1414 return false;
1415 /* "The initial version is 0. If a receiver does not support the
1416 * version indicated it MUST drop the packet.
1417 */
1418 if (gpe->version != 0)
1419 return false;
1420 /* "When the O bit is set to 1, the packet is an OAM packet and OAM
1421 * processing MUST occur." However, we don't implement OAM
1422 * processing, thus drop the packet.
1423 */
1424 if (gpe->oam_flag)
1425 return false;
1426
1427 *protocol = tun_p_to_eth_p(gpe->next_protocol);
1428 if (!*protocol)
1429 return false;
1430
1431 unparsed->vx_flags &= ~VXLAN_GPE_USED_BITS;
1432 return true;
1433 }
1434
1435 static bool vxlan_set_mac(struct vxlan_dev *vxlan,
1436 struct vxlan_sock *vs,
1437 struct sk_buff *skb, __be32 vni)
1438 {
1439 union vxlan_addr saddr;
1440 u32 ifindex = skb->dev->ifindex;
1441
1442 skb_reset_mac_header(skb);
1443 skb->protocol = eth_type_trans(skb, vxlan->dev);
1444 skb_postpull_rcsum(skb, eth_hdr(skb), ETH_HLEN);
1445
1446 /* Ignore packet loops (and multicast echo) */
1447 if (ether_addr_equal(eth_hdr(skb)->h_source, vxlan->dev->dev_addr))
1448 return false;
1449
1450 /* Get address from the outer IP header */
1451 if (vxlan_get_sk_family(vs) == AF_INET) {
1452 saddr.sin.sin_addr.s_addr = ip_hdr(skb)->saddr;
1453 saddr.sa.sa_family = AF_INET;
1454 #if IS_ENABLED(CONFIG_IPV6)
1455 } else {
1456 saddr.sin6.sin6_addr = ipv6_hdr(skb)->saddr;
1457 saddr.sa.sa_family = AF_INET6;
1458 #endif
1459 }
1460
1461 if ((vxlan->cfg.flags & VXLAN_F_LEARN) &&
1462 vxlan_snoop(skb->dev, &saddr, eth_hdr(skb)->h_source, ifindex, vni))
1463 return false;
1464
1465 return true;
1466 }
1467
1468 static bool vxlan_ecn_decapsulate(struct vxlan_sock *vs, void *oiph,
1469 struct sk_buff *skb)
1470 {
1471 int err = 0;
1472
1473 if (vxlan_get_sk_family(vs) == AF_INET)
1474 err = IP_ECN_decapsulate(oiph, skb);
1475 #if IS_ENABLED(CONFIG_IPV6)
1476 else
1477 err = IP6_ECN_decapsulate(oiph, skb);
1478 #endif
1479
1480 if (unlikely(err) && log_ecn_error) {
1481 if (vxlan_get_sk_family(vs) == AF_INET)
1482 net_info_ratelimited("non-ECT from %pI4 with TOS=%#x\n",
1483 &((struct iphdr *)oiph)->saddr,
1484 ((struct iphdr *)oiph)->tos);
1485 else
1486 net_info_ratelimited("non-ECT from %pI6\n",
1487 &((struct ipv6hdr *)oiph)->saddr);
1488 }
1489 return err <= 1;
1490 }
1491
1492 /* Callback from net/ipv4/udp.c to receive packets */
1493 static int vxlan_rcv(struct sock *sk, struct sk_buff *skb)
1494 {
1495 struct pcpu_sw_netstats *stats;
1496 struct vxlan_dev *vxlan;
1497 struct vxlan_sock *vs;
1498 struct vxlanhdr unparsed;
1499 struct vxlan_metadata _md;
1500 struct vxlan_metadata *md = &_md;
1501 __be16 protocol = htons(ETH_P_TEB);
1502 bool raw_proto = false;
1503 void *oiph;
1504 __be32 vni = 0;
1505
1506 /* Need UDP and VXLAN header to be present */
1507 if (!pskb_may_pull(skb, VXLAN_HLEN))
1508 goto drop;
1509
1510 unparsed = *vxlan_hdr(skb);
1511 /* VNI flag always required to be set */
1512 if (!(unparsed.vx_flags & VXLAN_HF_VNI)) {
1513 netdev_dbg(skb->dev, "invalid vxlan flags=%#x vni=%#x\n",
1514 ntohl(vxlan_hdr(skb)->vx_flags),
1515 ntohl(vxlan_hdr(skb)->vx_vni));
1516 /* Return non vxlan pkt */
1517 goto drop;
1518 }
1519 unparsed.vx_flags &= ~VXLAN_HF_VNI;
1520 unparsed.vx_vni &= ~VXLAN_VNI_MASK;
1521
1522 vs = rcu_dereference_sk_user_data(sk);
1523 if (!vs)
1524 goto drop;
1525
1526 vni = vxlan_vni(vxlan_hdr(skb)->vx_vni);
1527
1528 vxlan = vxlan_vs_find_vni(vs, skb->dev->ifindex, vni);
1529 if (!vxlan)
1530 goto drop;
1531
1532 /* For backwards compatibility, only allow reserved fields to be
1533 * used by VXLAN extensions if explicitly requested.
1534 */
1535 if (vs->flags & VXLAN_F_GPE) {
1536 if (!vxlan_parse_gpe_hdr(&unparsed, &protocol, skb, vs->flags))
1537 goto drop;
1538 raw_proto = true;
1539 }
1540
1541 if (__iptunnel_pull_header(skb, VXLAN_HLEN, protocol, raw_proto,
1542 !net_eq(vxlan->net, dev_net(vxlan->dev))))
1543 goto drop;
1544
1545 if (vxlan_collect_metadata(vs)) {
1546 struct metadata_dst *tun_dst;
1547
1548 tun_dst = udp_tun_rx_dst(skb, vxlan_get_sk_family(vs), TUNNEL_KEY,
1549 key32_to_tunnel_id(vni), sizeof(*md));
1550
1551 if (!tun_dst)
1552 goto drop;
1553
1554 md = ip_tunnel_info_opts(&tun_dst->u.tun_info);
1555
1556 skb_dst_set(skb, (struct dst_entry *)tun_dst);
1557 } else {
1558 memset(md, 0, sizeof(*md));
1559 }
1560
1561 if (vs->flags & VXLAN_F_REMCSUM_RX)
1562 if (!vxlan_remcsum(&unparsed, skb, vs->flags))
1563 goto drop;
1564 if (vs->flags & VXLAN_F_GBP)
1565 vxlan_parse_gbp_hdr(&unparsed, skb, vs->flags, md);
1566 /* Note that GBP and GPE can never be active together. This is
1567 * ensured in vxlan_dev_configure.
1568 */
1569
1570 if (unparsed.vx_flags || unparsed.vx_vni) {
1571 /* If there are any unprocessed flags remaining treat
1572 * this as a malformed packet. This behavior diverges from
1573 * VXLAN RFC (RFC7348) which stipulates that bits in reserved
1574 * in reserved fields are to be ignored. The approach here
1575 * maintains compatibility with previous stack code, and also
1576 * is more robust and provides a little more security in
1577 * adding extensions to VXLAN.
1578 */
1579 goto drop;
1580 }
1581
1582 if (!raw_proto) {
1583 if (!vxlan_set_mac(vxlan, vs, skb, vni))
1584 goto drop;
1585 } else {
1586 skb_reset_mac_header(skb);
1587 skb->dev = vxlan->dev;
1588 skb->pkt_type = PACKET_HOST;
1589 }
1590
1591 oiph = skb_network_header(skb);
1592 skb_reset_network_header(skb);
1593
1594 if (!vxlan_ecn_decapsulate(vs, oiph, skb)) {
1595 ++vxlan->dev->stats.rx_frame_errors;
1596 ++vxlan->dev->stats.rx_errors;
1597 goto drop;
1598 }
1599
1600 stats = this_cpu_ptr(vxlan->dev->tstats);
1601 u64_stats_update_begin(&stats->syncp);
1602 stats->rx_packets++;
1603 stats->rx_bytes += skb->len;
1604 u64_stats_update_end(&stats->syncp);
1605
1606 gro_cells_receive(&vxlan->gro_cells, skb);
1607 return 0;
1608
1609 drop:
1610 /* Consume bad packet */
1611 kfree_skb(skb);
1612 return 0;
1613 }
1614
1615 static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1616 {
1617 struct vxlan_dev *vxlan = netdev_priv(dev);
1618 struct arphdr *parp;
1619 u8 *arpptr, *sha;
1620 __be32 sip, tip;
1621 struct neighbour *n;
1622
1623 if (dev->flags & IFF_NOARP)
1624 goto out;
1625
1626 if (!pskb_may_pull(skb, arp_hdr_len(dev))) {
1627 dev->stats.tx_dropped++;
1628 goto out;
1629 }
1630 parp = arp_hdr(skb);
1631
1632 if ((parp->ar_hrd != htons(ARPHRD_ETHER) &&
1633 parp->ar_hrd != htons(ARPHRD_IEEE802)) ||
1634 parp->ar_pro != htons(ETH_P_IP) ||
1635 parp->ar_op != htons(ARPOP_REQUEST) ||
1636 parp->ar_hln != dev->addr_len ||
1637 parp->ar_pln != 4)
1638 goto out;
1639 arpptr = (u8 *)parp + sizeof(struct arphdr);
1640 sha = arpptr;
1641 arpptr += dev->addr_len; /* sha */
1642 memcpy(&sip, arpptr, sizeof(sip));
1643 arpptr += sizeof(sip);
1644 arpptr += dev->addr_len; /* tha */
1645 memcpy(&tip, arpptr, sizeof(tip));
1646
1647 if (ipv4_is_loopback(tip) ||
1648 ipv4_is_multicast(tip))
1649 goto out;
1650
1651 n = neigh_lookup(&arp_tbl, &tip, dev);
1652
1653 if (n) {
1654 struct vxlan_fdb *f;
1655 struct sk_buff *reply;
1656
1657 if (!(n->nud_state & NUD_CONNECTED)) {
1658 neigh_release(n);
1659 goto out;
1660 }
1661
1662 f = vxlan_find_mac(vxlan, n->ha, vni);
1663 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1664 /* bridge-local neighbor */
1665 neigh_release(n);
1666 goto out;
1667 }
1668
1669 reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
1670 n->ha, sha);
1671
1672 neigh_release(n);
1673
1674 if (reply == NULL)
1675 goto out;
1676
1677 skb_reset_mac_header(reply);
1678 __skb_pull(reply, skb_network_offset(reply));
1679 reply->ip_summed = CHECKSUM_UNNECESSARY;
1680 reply->pkt_type = PACKET_HOST;
1681
1682 if (netif_rx_ni(reply) == NET_RX_DROP)
1683 dev->stats.rx_dropped++;
1684 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1685 union vxlan_addr ipa = {
1686 .sin.sin_addr.s_addr = tip,
1687 .sin.sin_family = AF_INET,
1688 };
1689
1690 vxlan_ip_miss(dev, &ipa);
1691 }
1692 out:
1693 consume_skb(skb);
1694 return NETDEV_TX_OK;
1695 }
1696
1697 #if IS_ENABLED(CONFIG_IPV6)
1698 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
1699 struct neighbour *n, bool isrouter)
1700 {
1701 struct net_device *dev = request->dev;
1702 struct sk_buff *reply;
1703 struct nd_msg *ns, *na;
1704 struct ipv6hdr *pip6;
1705 u8 *daddr;
1706 int na_olen = 8; /* opt hdr + ETH_ALEN for target */
1707 int ns_olen;
1708 int i, len;
1709
1710 if (dev == NULL || !pskb_may_pull(request, request->len))
1711 return NULL;
1712
1713 len = LL_RESERVED_SPACE(dev) + sizeof(struct ipv6hdr) +
1714 sizeof(*na) + na_olen + dev->needed_tailroom;
1715 reply = alloc_skb(len, GFP_ATOMIC);
1716 if (reply == NULL)
1717 return NULL;
1718
1719 reply->protocol = htons(ETH_P_IPV6);
1720 reply->dev = dev;
1721 skb_reserve(reply, LL_RESERVED_SPACE(request->dev));
1722 skb_push(reply, sizeof(struct ethhdr));
1723 skb_reset_mac_header(reply);
1724
1725 ns = (struct nd_msg *)(ipv6_hdr(request) + 1);
1726
1727 daddr = eth_hdr(request)->h_source;
1728 ns_olen = request->len - skb_network_offset(request) -
1729 sizeof(struct ipv6hdr) - sizeof(*ns);
1730 for (i = 0; i < ns_olen-1; i += (ns->opt[i+1]<<3)) {
1731 if (ns->opt[i] == ND_OPT_SOURCE_LL_ADDR) {
1732 daddr = ns->opt + i + sizeof(struct nd_opt_hdr);
1733 break;
1734 }
1735 }
1736
1737 /* Ethernet header */
1738 ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
1739 ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
1740 eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
1741 reply->protocol = htons(ETH_P_IPV6);
1742
1743 skb_pull(reply, sizeof(struct ethhdr));
1744 skb_reset_network_header(reply);
1745 skb_put(reply, sizeof(struct ipv6hdr));
1746
1747 /* IPv6 header */
1748
1749 pip6 = ipv6_hdr(reply);
1750 memset(pip6, 0, sizeof(struct ipv6hdr));
1751 pip6->version = 6;
1752 pip6->priority = ipv6_hdr(request)->priority;
1753 pip6->nexthdr = IPPROTO_ICMPV6;
1754 pip6->hop_limit = 255;
1755 pip6->daddr = ipv6_hdr(request)->saddr;
1756 pip6->saddr = *(struct in6_addr *)n->primary_key;
1757
1758 skb_pull(reply, sizeof(struct ipv6hdr));
1759 skb_reset_transport_header(reply);
1760
1761 /* Neighbor Advertisement */
1762 na = skb_put_zero(reply, sizeof(*na) + na_olen);
1763 na->icmph.icmp6_type = NDISC_NEIGHBOUR_ADVERTISEMENT;
1764 na->icmph.icmp6_router = isrouter;
1765 na->icmph.icmp6_override = 1;
1766 na->icmph.icmp6_solicited = 1;
1767 na->target = ns->target;
1768 ether_addr_copy(&na->opt[2], n->ha);
1769 na->opt[0] = ND_OPT_TARGET_LL_ADDR;
1770 na->opt[1] = na_olen >> 3;
1771
1772 na->icmph.icmp6_cksum = csum_ipv6_magic(&pip6->saddr,
1773 &pip6->daddr, sizeof(*na)+na_olen, IPPROTO_ICMPV6,
1774 csum_partial(na, sizeof(*na)+na_olen, 0));
1775
1776 pip6->payload_len = htons(sizeof(*na)+na_olen);
1777
1778 skb_push(reply, sizeof(struct ipv6hdr));
1779
1780 reply->ip_summed = CHECKSUM_UNNECESSARY;
1781
1782 return reply;
1783 }
1784
1785 static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
1786 {
1787 struct vxlan_dev *vxlan = netdev_priv(dev);
1788 const struct in6_addr *daddr;
1789 const struct ipv6hdr *iphdr;
1790 struct inet6_dev *in6_dev;
1791 struct neighbour *n;
1792 struct nd_msg *msg;
1793
1794 in6_dev = __in6_dev_get(dev);
1795 if (!in6_dev)
1796 goto out;
1797
1798 iphdr = ipv6_hdr(skb);
1799 daddr = &iphdr->daddr;
1800 msg = (struct nd_msg *)(iphdr + 1);
1801
1802 if (ipv6_addr_loopback(daddr) ||
1803 ipv6_addr_is_multicast(&msg->target))
1804 goto out;
1805
1806 n = neigh_lookup(ipv6_stub->nd_tbl, &msg->target, dev);
1807
1808 if (n) {
1809 struct vxlan_fdb *f;
1810 struct sk_buff *reply;
1811
1812 if (!(n->nud_state & NUD_CONNECTED)) {
1813 neigh_release(n);
1814 goto out;
1815 }
1816
1817 f = vxlan_find_mac(vxlan, n->ha, vni);
1818 if (f && vxlan_addr_any(&(first_remote_rcu(f)->remote_ip))) {
1819 /* bridge-local neighbor */
1820 neigh_release(n);
1821 goto out;
1822 }
1823
1824 reply = vxlan_na_create(skb, n,
1825 !!(f ? f->flags & NTF_ROUTER : 0));
1826
1827 neigh_release(n);
1828
1829 if (reply == NULL)
1830 goto out;
1831
1832 if (netif_rx_ni(reply) == NET_RX_DROP)
1833 dev->stats.rx_dropped++;
1834
1835 } else if (vxlan->cfg.flags & VXLAN_F_L3MISS) {
1836 union vxlan_addr ipa = {
1837 .sin6.sin6_addr = msg->target,
1838 .sin6.sin6_family = AF_INET6,
1839 };
1840
1841 vxlan_ip_miss(dev, &ipa);
1842 }
1843
1844 out:
1845 consume_skb(skb);
1846 return NETDEV_TX_OK;
1847 }
1848 #endif
1849
1850 static bool route_shortcircuit(struct net_device *dev, struct sk_buff *skb)
1851 {
1852 struct vxlan_dev *vxlan = netdev_priv(dev);
1853 struct neighbour *n;
1854
1855 if (is_multicast_ether_addr(eth_hdr(skb)->h_dest))
1856 return false;
1857
1858 n = NULL;
1859 switch (ntohs(eth_hdr(skb)->h_proto)) {
1860 case ETH_P_IP:
1861 {
1862 struct iphdr *pip;
1863
1864 if (!pskb_may_pull(skb, sizeof(struct iphdr)))
1865 return false;
1866 pip = ip_hdr(skb);
1867 n = neigh_lookup(&arp_tbl, &pip->daddr, dev);
1868 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1869 union vxlan_addr ipa = {
1870 .sin.sin_addr.s_addr = pip->daddr,
1871 .sin.sin_family = AF_INET,
1872 };
1873
1874 vxlan_ip_miss(dev, &ipa);
1875 return false;
1876 }
1877
1878 break;
1879 }
1880 #if IS_ENABLED(CONFIG_IPV6)
1881 case ETH_P_IPV6:
1882 {
1883 struct ipv6hdr *pip6;
1884
1885 if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
1886 return false;
1887 pip6 = ipv6_hdr(skb);
1888 n = neigh_lookup(ipv6_stub->nd_tbl, &pip6->daddr, dev);
1889 if (!n && (vxlan->cfg.flags & VXLAN_F_L3MISS)) {
1890 union vxlan_addr ipa = {
1891 .sin6.sin6_addr = pip6->daddr,
1892 .sin6.sin6_family = AF_INET6,
1893 };
1894
1895 vxlan_ip_miss(dev, &ipa);
1896 return false;
1897 }
1898
1899 break;
1900 }
1901 #endif
1902 default:
1903 return false;
1904 }
1905
1906 if (n) {
1907 bool diff;
1908
1909 diff = !ether_addr_equal(eth_hdr(skb)->h_dest, n->ha);
1910 if (diff) {
1911 memcpy(eth_hdr(skb)->h_source, eth_hdr(skb)->h_dest,
1912 dev->addr_len);
1913 memcpy(eth_hdr(skb)->h_dest, n->ha, dev->addr_len);
1914 }
1915 neigh_release(n);
1916 return diff;
1917 }
1918
1919 return false;
1920 }
1921
1922 static void vxlan_build_gbp_hdr(struct vxlanhdr *vxh, u32 vxflags,
1923 struct vxlan_metadata *md)
1924 {
1925 struct vxlanhdr_gbp *gbp;
1926
1927 if (!md->gbp)
1928 return;
1929
1930 gbp = (struct vxlanhdr_gbp *)vxh;
1931 vxh->vx_flags |= VXLAN_HF_GBP;
1932
1933 if (md->gbp & VXLAN_GBP_DONT_LEARN)
1934 gbp->dont_learn = 1;
1935
1936 if (md->gbp & VXLAN_GBP_POLICY_APPLIED)
1937 gbp->policy_applied = 1;
1938
1939 gbp->policy_id = htons(md->gbp & VXLAN_GBP_ID_MASK);
1940 }
1941
1942 static int vxlan_build_gpe_hdr(struct vxlanhdr *vxh, u32 vxflags,
1943 __be16 protocol)
1944 {
1945 struct vxlanhdr_gpe *gpe = (struct vxlanhdr_gpe *)vxh;
1946
1947 gpe->np_applied = 1;
1948 gpe->next_protocol = tun_p_from_eth_p(protocol);
1949 if (!gpe->next_protocol)
1950 return -EPFNOSUPPORT;
1951 return 0;
1952 }
1953
1954 static int vxlan_build_skb(struct sk_buff *skb, struct dst_entry *dst,
1955 int iphdr_len, __be32 vni,
1956 struct vxlan_metadata *md, u32 vxflags,
1957 bool udp_sum)
1958 {
1959 struct vxlanhdr *vxh;
1960 int min_headroom;
1961 int err;
1962 int type = udp_sum ? SKB_GSO_UDP_TUNNEL_CSUM : SKB_GSO_UDP_TUNNEL;
1963 __be16 inner_protocol = htons(ETH_P_TEB);
1964
1965 if ((vxflags & VXLAN_F_REMCSUM_TX) &&
1966 skb->ip_summed == CHECKSUM_PARTIAL) {
1967 int csum_start = skb_checksum_start_offset(skb);
1968
1969 if (csum_start <= VXLAN_MAX_REMCSUM_START &&
1970 !(csum_start & VXLAN_RCO_SHIFT_MASK) &&
1971 (skb->csum_offset == offsetof(struct udphdr, check) ||
1972 skb->csum_offset == offsetof(struct tcphdr, check)))
1973 type |= SKB_GSO_TUNNEL_REMCSUM;
1974 }
1975
1976 min_headroom = LL_RESERVED_SPACE(dst->dev) + dst->header_len
1977 + VXLAN_HLEN + iphdr_len;
1978
1979 /* Need space for new headers (invalidates iph ptr) */
1980 err = skb_cow_head(skb, min_headroom);
1981 if (unlikely(err))
1982 return err;
1983
1984 err = iptunnel_handle_offloads(skb, type);
1985 if (err)
1986 return err;
1987
1988 vxh = __skb_push(skb, sizeof(*vxh));
1989 vxh->vx_flags = VXLAN_HF_VNI;
1990 vxh->vx_vni = vxlan_vni_field(vni);
1991
1992 if (type & SKB_GSO_TUNNEL_REMCSUM) {
1993 unsigned int start;
1994
1995 start = skb_checksum_start_offset(skb) - sizeof(struct vxlanhdr);
1996 vxh->vx_vni |= vxlan_compute_rco(start, skb->csum_offset);
1997 vxh->vx_flags |= VXLAN_HF_RCO;
1998
1999 if (!skb_is_gso(skb)) {
2000 skb->ip_summed = CHECKSUM_NONE;
2001 skb->encapsulation = 0;
2002 }
2003 }
2004
2005 if (vxflags & VXLAN_F_GBP)
2006 vxlan_build_gbp_hdr(vxh, vxflags, md);
2007 if (vxflags & VXLAN_F_GPE) {
2008 err = vxlan_build_gpe_hdr(vxh, vxflags, skb->protocol);
2009 if (err < 0)
2010 return err;
2011 inner_protocol = skb->protocol;
2012 }
2013
2014 skb_set_inner_protocol(skb, inner_protocol);
2015 return 0;
2016 }
2017
2018 static struct rtable *vxlan_get_route(struct vxlan_dev *vxlan, struct net_device *dev,
2019 struct vxlan_sock *sock4,
2020 struct sk_buff *skb, int oif, u8 tos,
2021 __be32 daddr, __be32 *saddr, __be16 dport, __be16 sport,
2022 struct dst_cache *dst_cache,
2023 const struct ip_tunnel_info *info)
2024 {
2025 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2026 struct rtable *rt = NULL;
2027 struct flowi4 fl4;
2028
2029 if (!sock4)
2030 return ERR_PTR(-EIO);
2031
2032 if (tos && !info)
2033 use_cache = false;
2034 if (use_cache) {
2035 rt = dst_cache_get_ip4(dst_cache, saddr);
2036 if (rt)
2037 return rt;
2038 }
2039
2040 memset(&fl4, 0, sizeof(fl4));
2041 fl4.flowi4_oif = oif;
2042 fl4.flowi4_tos = RT_TOS(tos);
2043 fl4.flowi4_mark = skb->mark;
2044 fl4.flowi4_proto = IPPROTO_UDP;
2045 fl4.daddr = daddr;
2046 fl4.saddr = *saddr;
2047 fl4.fl4_dport = dport;
2048 fl4.fl4_sport = sport;
2049
2050 rt = ip_route_output_key(vxlan->net, &fl4);
2051 if (likely(!IS_ERR(rt))) {
2052 if (rt->dst.dev == dev) {
2053 netdev_dbg(dev, "circular route to %pI4\n", &daddr);
2054 ip_rt_put(rt);
2055 return ERR_PTR(-ELOOP);
2056 }
2057
2058 *saddr = fl4.saddr;
2059 if (use_cache)
2060 dst_cache_set_ip4(dst_cache, &rt->dst, fl4.saddr);
2061 } else {
2062 netdev_dbg(dev, "no route to %pI4\n", &daddr);
2063 return ERR_PTR(-ENETUNREACH);
2064 }
2065 return rt;
2066 }
2067
2068 #if IS_ENABLED(CONFIG_IPV6)
2069 static struct dst_entry *vxlan6_get_route(struct vxlan_dev *vxlan,
2070 struct net_device *dev,
2071 struct vxlan_sock *sock6,
2072 struct sk_buff *skb, int oif, u8 tos,
2073 __be32 label,
2074 const struct in6_addr *daddr,
2075 struct in6_addr *saddr,
2076 __be16 dport, __be16 sport,
2077 struct dst_cache *dst_cache,
2078 const struct ip_tunnel_info *info)
2079 {
2080 bool use_cache = ip_tunnel_dst_cache_usable(skb, info);
2081 struct dst_entry *ndst;
2082 struct flowi6 fl6;
2083 int err;
2084
2085 if (!sock6)
2086 return ERR_PTR(-EIO);
2087
2088 if (tos && !info)
2089 use_cache = false;
2090 if (use_cache) {
2091 ndst = dst_cache_get_ip6(dst_cache, saddr);
2092 if (ndst)
2093 return ndst;
2094 }
2095
2096 memset(&fl6, 0, sizeof(fl6));
2097 fl6.flowi6_oif = oif;
2098 fl6.daddr = *daddr;
2099 fl6.saddr = *saddr;
2100 fl6.flowlabel = ip6_make_flowinfo(RT_TOS(tos), label);
2101 fl6.flowi6_mark = skb->mark;
2102 fl6.flowi6_proto = IPPROTO_UDP;
2103 fl6.fl6_dport = dport;
2104 fl6.fl6_sport = sport;
2105
2106 err = ipv6_stub->ipv6_dst_lookup(vxlan->net,
2107 sock6->sock->sk,
2108 &ndst, &fl6);
2109 if (unlikely(err < 0)) {
2110 netdev_dbg(dev, "no route to %pI6\n", daddr);
2111 return ERR_PTR(-ENETUNREACH);
2112 }
2113
2114 if (unlikely(ndst->dev == dev)) {
2115 netdev_dbg(dev, "circular route to %pI6\n", daddr);
2116 dst_release(ndst);
2117 return ERR_PTR(-ELOOP);
2118 }
2119
2120 *saddr = fl6.saddr;
2121 if (use_cache)
2122 dst_cache_set_ip6(dst_cache, ndst, saddr);
2123 return ndst;
2124 }
2125 #endif
2126
2127 /* Bypass encapsulation if the destination is local */
2128 static void vxlan_encap_bypass(struct sk_buff *skb, struct vxlan_dev *src_vxlan,
2129 struct vxlan_dev *dst_vxlan, __be32 vni)
2130 {
2131 struct pcpu_sw_netstats *tx_stats, *rx_stats;
2132 union vxlan_addr loopback;
2133 union vxlan_addr *remote_ip = &dst_vxlan->default_dst.remote_ip;
2134 struct net_device *dev = skb->dev;
2135 int len = skb->len;
2136
2137 tx_stats = this_cpu_ptr(src_vxlan->dev->tstats);
2138 rx_stats = this_cpu_ptr(dst_vxlan->dev->tstats);
2139 skb->pkt_type = PACKET_HOST;
2140 skb->encapsulation = 0;
2141 skb->dev = dst_vxlan->dev;
2142 __skb_pull(skb, skb_network_offset(skb));
2143
2144 if (remote_ip->sa.sa_family == AF_INET) {
2145 loopback.sin.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
2146 loopback.sa.sa_family = AF_INET;
2147 #if IS_ENABLED(CONFIG_IPV6)
2148 } else {
2149 loopback.sin6.sin6_addr = in6addr_loopback;
2150 loopback.sa.sa_family = AF_INET6;
2151 #endif
2152 }
2153
2154 if (dst_vxlan->cfg.flags & VXLAN_F_LEARN)
2155 vxlan_snoop(skb->dev, &loopback, eth_hdr(skb)->h_source, 0,
2156 vni);
2157
2158 u64_stats_update_begin(&tx_stats->syncp);
2159 tx_stats->tx_packets++;
2160 tx_stats->tx_bytes += len;
2161 u64_stats_update_end(&tx_stats->syncp);
2162
2163 if (netif_rx(skb) == NET_RX_SUCCESS) {
2164 u64_stats_update_begin(&rx_stats->syncp);
2165 rx_stats->rx_packets++;
2166 rx_stats->rx_bytes += len;
2167 u64_stats_update_end(&rx_stats->syncp);
2168 } else {
2169 dev->stats.rx_dropped++;
2170 }
2171 }
2172
2173 static int encap_bypass_if_local(struct sk_buff *skb, struct net_device *dev,
2174 struct vxlan_dev *vxlan,
2175 union vxlan_addr *daddr,
2176 __be16 dst_port, int dst_ifindex, __be32 vni,
2177 struct dst_entry *dst,
2178 u32 rt_flags)
2179 {
2180 #if IS_ENABLED(CONFIG_IPV6)
2181 /* IPv6 rt-flags are checked against RTF_LOCAL, but the value of
2182 * RTF_LOCAL is equal to RTCF_LOCAL. So to keep code simple
2183 * we can use RTCF_LOCAL which works for ipv4 and ipv6 route entry.
2184 */
2185 BUILD_BUG_ON(RTCF_LOCAL != RTF_LOCAL);
2186 #endif
2187 /* Bypass encapsulation if the destination is local */
2188 if (rt_flags & RTCF_LOCAL &&
2189 !(rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))) {
2190 struct vxlan_dev *dst_vxlan;
2191
2192 dst_release(dst);
2193 dst_vxlan = vxlan_find_vni(vxlan->net, dst_ifindex, vni,
2194 daddr->sa.sa_family, dst_port,
2195 vxlan->cfg.flags);
2196 if (!dst_vxlan) {
2197 dev->stats.tx_errors++;
2198 kfree_skb(skb);
2199
2200 return -ENOENT;
2201 }
2202 vxlan_encap_bypass(skb, vxlan, dst_vxlan, vni);
2203 return 1;
2204 }
2205
2206 return 0;
2207 }
2208
2209 static void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
2210 __be32 default_vni, struct vxlan_rdst *rdst,
2211 bool did_rsc)
2212 {
2213 struct dst_cache *dst_cache;
2214 struct ip_tunnel_info *info;
2215 struct vxlan_dev *vxlan = netdev_priv(dev);
2216 const struct iphdr *old_iph = ip_hdr(skb);
2217 union vxlan_addr *dst;
2218 union vxlan_addr remote_ip, local_ip;
2219 struct vxlan_metadata _md;
2220 struct vxlan_metadata *md = &_md;
2221 __be16 src_port = 0, dst_port;
2222 struct dst_entry *ndst = NULL;
2223 __be32 vni, label;
2224 __u8 tos, ttl;
2225 int ifindex;
2226 int err;
2227 u32 flags = vxlan->cfg.flags;
2228 bool udp_sum = false;
2229 bool xnet = !net_eq(vxlan->net, dev_net(vxlan->dev));
2230
2231 info = skb_tunnel_info(skb);
2232
2233 if (rdst) {
2234 dst = &rdst->remote_ip;
2235 if (vxlan_addr_any(dst)) {
2236 if (did_rsc) {
2237 /* short-circuited back to local bridge */
2238 vxlan_encap_bypass(skb, vxlan, vxlan, default_vni);
2239 return;
2240 }
2241 goto drop;
2242 }
2243
2244 dst_port = rdst->remote_port ? rdst->remote_port : vxlan->cfg.dst_port;
2245 vni = (rdst->remote_vni) ? : default_vni;
2246 ifindex = rdst->remote_ifindex;
2247 local_ip = vxlan->cfg.saddr;
2248 dst_cache = &rdst->dst_cache;
2249 md->gbp = skb->mark;
2250 ttl = vxlan->cfg.ttl;
2251 if (!ttl && vxlan_addr_multicast(dst))
2252 ttl = 1;
2253
2254 tos = vxlan->cfg.tos;
2255 if (tos == 1)
2256 tos = ip_tunnel_get_dsfield(old_iph, skb);
2257
2258 if (dst->sa.sa_family == AF_INET)
2259 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM_TX);
2260 else
2261 udp_sum = !(flags & VXLAN_F_UDP_ZERO_CSUM6_TX);
2262 label = vxlan->cfg.label;
2263 } else {
2264 if (!info) {
2265 WARN_ONCE(1, "%s: Missing encapsulation instructions\n",
2266 dev->name);
2267 goto drop;
2268 }
2269 remote_ip.sa.sa_family = ip_tunnel_info_af(info);
2270 if (remote_ip.sa.sa_family == AF_INET) {
2271 remote_ip.sin.sin_addr.s_addr = info->key.u.ipv4.dst;
2272 local_ip.sin.sin_addr.s_addr = info->key.u.ipv4.src;
2273 } else {
2274 remote_ip.sin6.sin6_addr = info->key.u.ipv6.dst;
2275 local_ip.sin6.sin6_addr = info->key.u.ipv6.src;
2276 }
2277 dst = &remote_ip;
2278 dst_port = info->key.tp_dst ? : vxlan->cfg.dst_port;
2279 vni = tunnel_id_to_key32(info->key.tun_id);
2280 ifindex = 0;
2281 dst_cache = &info->dst_cache;
2282 if (info->options_len)
2283 md = ip_tunnel_info_opts(info);
2284 ttl = info->key.ttl;
2285 tos = info->key.tos;
2286 label = info->key.label;
2287 udp_sum = !!(info->key.tun_flags & TUNNEL_CSUM);
2288 }
2289 src_port = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2290 vxlan->cfg.port_max, true);
2291
2292 rcu_read_lock();
2293 if (dst->sa.sa_family == AF_INET) {
2294 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2295 struct rtable *rt;
2296 __be16 df = 0;
2297
2298 rt = vxlan_get_route(vxlan, dev, sock4, skb, ifindex, tos,
2299 dst->sin.sin_addr.s_addr,
2300 &local_ip.sin.sin_addr.s_addr,
2301 dst_port, src_port,
2302 dst_cache, info);
2303 if (IS_ERR(rt)) {
2304 err = PTR_ERR(rt);
2305 goto tx_error;
2306 }
2307
2308 if (fan_has_map(&vxlan->fan) && rt->rt_flags & RTCF_LOCAL) {
2309 netdev_dbg(dev, "discard fan to localhost %pI4\n",
2310 &dst->sin.sin_addr.s_addr);
2311 ip_rt_put(rt);
2312 goto tx_free;
2313 }
2314
2315 /* Bypass encapsulation if the destination is local */
2316 if (!info) {
2317 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2318 dst_port, ifindex, vni,
2319 &rt->dst, rt->rt_flags);
2320 if (err)
2321 goto out_unlock;
2322 } else if (info->key.tun_flags & TUNNEL_DONT_FRAGMENT) {
2323 df = htons(IP_DF);
2324 }
2325
2326 ndst = &rt->dst;
2327 if (skb_dst(skb)) {
2328 int mtu = dst_mtu(ndst) - VXLAN_HEADROOM;
2329
2330 skb_dst_update_pmtu(skb, mtu);
2331 }
2332
2333 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2334 ttl = ttl ? : ip4_dst_hoplimit(&rt->dst);
2335 err = vxlan_build_skb(skb, ndst, sizeof(struct iphdr),
2336 vni, md, flags, udp_sum);
2337 if (err < 0)
2338 goto tx_error;
2339
2340 udp_tunnel_xmit_skb(rt, sock4->sock->sk, skb, local_ip.sin.sin_addr.s_addr,
2341 dst->sin.sin_addr.s_addr, tos, ttl, df,
2342 src_port, dst_port, xnet, !udp_sum);
2343 #if IS_ENABLED(CONFIG_IPV6)
2344 } else {
2345 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2346
2347 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, ifindex, tos,
2348 label, &dst->sin6.sin6_addr,
2349 &local_ip.sin6.sin6_addr,
2350 dst_port, src_port,
2351 dst_cache, info);
2352 if (IS_ERR(ndst)) {
2353 err = PTR_ERR(ndst);
2354 ndst = NULL;
2355 goto tx_error;
2356 }
2357
2358 if (!info) {
2359 u32 rt6i_flags = ((struct rt6_info *)ndst)->rt6i_flags;
2360
2361 err = encap_bypass_if_local(skb, dev, vxlan, dst,
2362 dst_port, ifindex, vni,
2363 ndst, rt6i_flags);
2364 if (err)
2365 goto out_unlock;
2366 }
2367
2368 if (skb_dst(skb)) {
2369 int mtu = dst_mtu(ndst) - VXLAN6_HEADROOM;
2370
2371 skb_dst_update_pmtu(skb, mtu);
2372 }
2373
2374 tos = ip_tunnel_ecn_encap(tos, old_iph, skb);
2375 ttl = ttl ? : ip6_dst_hoplimit(ndst);
2376 skb_scrub_packet(skb, xnet);
2377 err = vxlan_build_skb(skb, ndst, sizeof(struct ipv6hdr),
2378 vni, md, flags, udp_sum);
2379 if (err < 0)
2380 goto tx_error;
2381
2382 udp_tunnel6_xmit_skb(ndst, sock6->sock->sk, skb, dev,
2383 &local_ip.sin6.sin6_addr,
2384 &dst->sin6.sin6_addr, tos, ttl,
2385 label, src_port, dst_port, !udp_sum);
2386 #endif
2387 }
2388 out_unlock:
2389 rcu_read_unlock();
2390 return;
2391
2392 drop:
2393 dev->stats.tx_dropped++;
2394 dev_kfree_skb(skb);
2395 return;
2396
2397 tx_error:
2398 rcu_read_unlock();
2399 if (err == -ELOOP)
2400 dev->stats.collisions++;
2401 else if (err == -ENETUNREACH)
2402 dev->stats.tx_carrier_errors++;
2403 dst_release(ndst);
2404 dev->stats.tx_errors++;
2405 tx_free:
2406 kfree_skb(skb);
2407 }
2408
2409 /* Transmit local packets over Vxlan
2410 *
2411 * Outer IP header inherits ECN and DF from inner header.
2412 * Outer UDP destination is the VXLAN assigned port.
2413 * source port is based on hash of flow
2414 */
2415 static netdev_tx_t vxlan_xmit(struct sk_buff *skb, struct net_device *dev)
2416 {
2417 struct vxlan_dev *vxlan = netdev_priv(dev);
2418 struct vxlan_rdst *rdst, *fdst = NULL;
2419 const struct ip_tunnel_info *info;
2420 bool did_rsc = false;
2421 struct vxlan_fdb *f;
2422 struct ethhdr *eth;
2423 __be32 vni = 0;
2424
2425 info = skb_tunnel_info(skb);
2426
2427 skb_reset_mac_header(skb);
2428
2429 if (vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA) {
2430 if (info && info->mode & IP_TUNNEL_INFO_BRIDGE &&
2431 info->mode & IP_TUNNEL_INFO_TX) {
2432 vni = tunnel_id_to_key32(info->key.tun_id);
2433 } else {
2434 if (info && info->mode & IP_TUNNEL_INFO_TX)
2435 vxlan_xmit_one(skb, dev, vni, NULL, false);
2436 else
2437 kfree_skb(skb);
2438 return NETDEV_TX_OK;
2439 }
2440 }
2441
2442 if (vxlan->cfg.flags & VXLAN_F_PROXY) {
2443 eth = eth_hdr(skb);
2444 if (ntohs(eth->h_proto) == ETH_P_ARP)
2445 return arp_reduce(dev, skb, vni);
2446 #if IS_ENABLED(CONFIG_IPV6)
2447 else if (ntohs(eth->h_proto) == ETH_P_IPV6 &&
2448 pskb_may_pull(skb, sizeof(struct ipv6hdr) +
2449 sizeof(struct nd_msg)) &&
2450 ipv6_hdr(skb)->nexthdr == IPPROTO_ICMPV6) {
2451 struct nd_msg *m = (struct nd_msg *)(ipv6_hdr(skb) + 1);
2452
2453 if (m->icmph.icmp6_code == 0 &&
2454 m->icmph.icmp6_type == NDISC_NEIGHBOUR_SOLICITATION)
2455 return neigh_reduce(dev, skb, vni);
2456 }
2457 #endif
2458 }
2459
2460 if (fan_has_map(&vxlan->fan)) {
2461 struct vxlan_rdst fan_rdst;
2462
2463 netdev_dbg(vxlan->dev, "vxlan_xmit p %x d %pM\n",
2464 eth->h_proto, eth->h_dest);
2465 if (vxlan_fan_build_rdst(vxlan, skb, &fan_rdst)) {
2466 dev->stats.tx_dropped++;
2467 kfree_skb(skb);
2468 return NETDEV_TX_OK;
2469 }
2470 vxlan_xmit_one(skb, dev, vni, &fan_rdst, 0);
2471 return NETDEV_TX_OK;
2472 }
2473
2474 eth = eth_hdr(skb);
2475 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2476 did_rsc = false;
2477
2478 if (f && (f->flags & NTF_ROUTER) && (vxlan->cfg.flags & VXLAN_F_RSC) &&
2479 (ntohs(eth->h_proto) == ETH_P_IP ||
2480 ntohs(eth->h_proto) == ETH_P_IPV6)) {
2481 did_rsc = route_shortcircuit(dev, skb);
2482 if (did_rsc)
2483 f = vxlan_find_mac(vxlan, eth->h_dest, vni);
2484 }
2485
2486 if (f == NULL) {
2487 f = vxlan_find_mac(vxlan, all_zeros_mac, vni);
2488 if (f == NULL) {
2489 if ((vxlan->cfg.flags & VXLAN_F_L2MISS) &&
2490 !is_multicast_ether_addr(eth->h_dest))
2491 vxlan_fdb_miss(vxlan, eth->h_dest);
2492
2493 dev->stats.tx_dropped++;
2494 kfree_skb(skb);
2495 return NETDEV_TX_OK;
2496 }
2497 }
2498
2499 list_for_each_entry_rcu(rdst, &f->remotes, list) {
2500 struct sk_buff *skb1;
2501
2502 if (!fdst) {
2503 fdst = rdst;
2504 continue;
2505 }
2506 skb1 = skb_clone(skb, GFP_ATOMIC);
2507 if (skb1)
2508 vxlan_xmit_one(skb1, dev, vni, rdst, did_rsc);
2509 }
2510
2511 if (fdst)
2512 vxlan_xmit_one(skb, dev, vni, fdst, did_rsc);
2513 else
2514 kfree_skb(skb);
2515 return NETDEV_TX_OK;
2516 }
2517
2518 /* Walk the forwarding table and purge stale entries */
2519 static void vxlan_cleanup(struct timer_list *t)
2520 {
2521 struct vxlan_dev *vxlan = from_timer(vxlan, t, age_timer);
2522 unsigned long next_timer = jiffies + FDB_AGE_INTERVAL;
2523 unsigned int h;
2524
2525 if (!netif_running(vxlan->dev))
2526 return;
2527
2528 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2529 struct hlist_node *p, *n;
2530
2531 spin_lock_bh(&vxlan->hash_lock);
2532 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2533 struct vxlan_fdb *f
2534 = container_of(p, struct vxlan_fdb, hlist);
2535 unsigned long timeout;
2536
2537 if (f->state & (NUD_PERMANENT | NUD_NOARP))
2538 continue;
2539
2540 if (f->flags & NTF_EXT_LEARNED)
2541 continue;
2542
2543 timeout = f->used + vxlan->cfg.age_interval * HZ;
2544 if (time_before_eq(timeout, jiffies)) {
2545 netdev_dbg(vxlan->dev,
2546 "garbage collect %pM\n",
2547 f->eth_addr);
2548 f->state = NUD_STALE;
2549 vxlan_fdb_destroy(vxlan, f);
2550 } else if (time_before(timeout, next_timer))
2551 next_timer = timeout;
2552 }
2553 spin_unlock_bh(&vxlan->hash_lock);
2554 }
2555
2556 mod_timer(&vxlan->age_timer, next_timer);
2557 }
2558
2559 static void vxlan_vs_del_dev(struct vxlan_dev *vxlan)
2560 {
2561 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2562
2563 spin_lock(&vn->sock_lock);
2564 hlist_del_init_rcu(&vxlan->hlist4.hlist);
2565 #if IS_ENABLED(CONFIG_IPV6)
2566 hlist_del_init_rcu(&vxlan->hlist6.hlist);
2567 #endif
2568 spin_unlock(&vn->sock_lock);
2569 }
2570
2571 static void vxlan_vs_add_dev(struct vxlan_sock *vs, struct vxlan_dev *vxlan,
2572 struct vxlan_dev_node *node)
2573 {
2574 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2575 __be32 vni = vxlan->default_dst.remote_vni;
2576
2577 node->vxlan = vxlan;
2578 spin_lock(&vn->sock_lock);
2579 hlist_add_head_rcu(&node->hlist, vni_head(vs, vni));
2580 spin_unlock(&vn->sock_lock);
2581 }
2582
2583 /* Setup stats when device is created */
2584 static int vxlan_init(struct net_device *dev)
2585 {
2586 dev->tstats = netdev_alloc_pcpu_stats(struct pcpu_sw_netstats);
2587 if (!dev->tstats)
2588 return -ENOMEM;
2589
2590 return 0;
2591 }
2592
2593 static void vxlan_fdb_delete_default(struct vxlan_dev *vxlan, __be32 vni)
2594 {
2595 struct vxlan_fdb *f;
2596
2597 spin_lock_bh(&vxlan->hash_lock);
2598 f = __vxlan_find_mac(vxlan, all_zeros_mac, vni);
2599 if (f)
2600 vxlan_fdb_destroy(vxlan, f);
2601 spin_unlock_bh(&vxlan->hash_lock);
2602 }
2603
2604 static void vxlan_uninit(struct net_device *dev)
2605 {
2606 struct vxlan_dev *vxlan = netdev_priv(dev);
2607
2608 vxlan_fdb_delete_default(vxlan, vxlan->cfg.vni);
2609
2610 free_percpu(dev->tstats);
2611 }
2612
2613 /* Start ageing timer and join group when device is brought up */
2614 static int vxlan_open(struct net_device *dev)
2615 {
2616 struct vxlan_dev *vxlan = netdev_priv(dev);
2617 int ret;
2618
2619 ret = vxlan_sock_add(vxlan);
2620 if (ret < 0)
2621 return ret;
2622
2623 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip)) {
2624 ret = vxlan_igmp_join(vxlan);
2625 if (ret == -EADDRINUSE)
2626 ret = 0;
2627 if (ret) {
2628 vxlan_sock_release(vxlan);
2629 return ret;
2630 }
2631 }
2632
2633 if (vxlan->cfg.age_interval)
2634 mod_timer(&vxlan->age_timer, jiffies + FDB_AGE_INTERVAL);
2635
2636 return ret;
2637 }
2638
2639 /* Purge the forwarding table */
2640 static void vxlan_flush(struct vxlan_dev *vxlan, bool do_all)
2641 {
2642 unsigned int h;
2643
2644 spin_lock_bh(&vxlan->hash_lock);
2645 for (h = 0; h < FDB_HASH_SIZE; ++h) {
2646 struct hlist_node *p, *n;
2647 hlist_for_each_safe(p, n, &vxlan->fdb_head[h]) {
2648 struct vxlan_fdb *f
2649 = container_of(p, struct vxlan_fdb, hlist);
2650 if (!do_all && (f->state & (NUD_PERMANENT | NUD_NOARP)))
2651 continue;
2652 /* the all_zeros_mac entry is deleted at vxlan_uninit */
2653 if (!is_zero_ether_addr(f->eth_addr))
2654 vxlan_fdb_destroy(vxlan, f);
2655 }
2656 }
2657 spin_unlock_bh(&vxlan->hash_lock);
2658 }
2659
2660 /* Cleanup timer and forwarding table on shutdown */
2661 static int vxlan_stop(struct net_device *dev)
2662 {
2663 struct vxlan_dev *vxlan = netdev_priv(dev);
2664 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
2665 int ret = 0;
2666
2667 if (vxlan_addr_multicast(&vxlan->default_dst.remote_ip) &&
2668 !vxlan_group_used(vn, vxlan))
2669 ret = vxlan_igmp_leave(vxlan);
2670
2671 del_timer_sync(&vxlan->age_timer);
2672
2673 vxlan_flush(vxlan, false);
2674 vxlan_sock_release(vxlan);
2675
2676 return ret;
2677 }
2678
2679 /* Stub, nothing needs to be done. */
2680 static void vxlan_set_multicast_list(struct net_device *dev)
2681 {
2682 }
2683
2684 static int vxlan_change_mtu(struct net_device *dev, int new_mtu)
2685 {
2686 struct vxlan_dev *vxlan = netdev_priv(dev);
2687 struct vxlan_rdst *dst = &vxlan->default_dst;
2688 struct net_device *lowerdev = __dev_get_by_index(vxlan->net,
2689 dst->remote_ifindex);
2690 bool use_ipv6 = !!(vxlan->cfg.flags & VXLAN_F_IPV6);
2691
2692 /* This check is different than dev->max_mtu, because it looks at
2693 * the lowerdev->mtu, rather than the static dev->max_mtu
2694 */
2695 if (lowerdev) {
2696 int max_mtu = lowerdev->mtu -
2697 (use_ipv6 ? VXLAN6_HEADROOM : VXLAN_HEADROOM);
2698 if (new_mtu > max_mtu)
2699 return -EINVAL;
2700 }
2701
2702 dev->mtu = new_mtu;
2703 return 0;
2704 }
2705
2706 static int vxlan_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb)
2707 {
2708 struct vxlan_dev *vxlan = netdev_priv(dev);
2709 struct ip_tunnel_info *info = skb_tunnel_info(skb);
2710 __be16 sport, dport;
2711
2712 sport = udp_flow_src_port(dev_net(dev), skb, vxlan->cfg.port_min,
2713 vxlan->cfg.port_max, true);
2714 dport = info->key.tp_dst ? : vxlan->cfg.dst_port;
2715
2716 if (ip_tunnel_info_af(info) == AF_INET) {
2717 struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
2718 struct rtable *rt;
2719
2720 rt = vxlan_get_route(vxlan, dev, sock4, skb, 0, info->key.tos,
2721 info->key.u.ipv4.dst,
2722 &info->key.u.ipv4.src, dport, sport,
2723 &info->dst_cache, info);
2724 if (IS_ERR(rt))
2725 return PTR_ERR(rt);
2726 ip_rt_put(rt);
2727 } else {
2728 #if IS_ENABLED(CONFIG_IPV6)
2729 struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
2730 struct dst_entry *ndst;
2731
2732 ndst = vxlan6_get_route(vxlan, dev, sock6, skb, 0, info->key.tos,
2733 info->key.label, &info->key.u.ipv6.dst,
2734 &info->key.u.ipv6.src, dport, sport,
2735 &info->dst_cache, info);
2736 if (IS_ERR(ndst))
2737 return PTR_ERR(ndst);
2738 dst_release(ndst);
2739 #else /* !CONFIG_IPV6 */
2740 return -EPFNOSUPPORT;
2741 #endif
2742 }
2743 info->key.tp_src = sport;
2744 info->key.tp_dst = dport;
2745 return 0;
2746 }
2747
2748 static const struct net_device_ops vxlan_netdev_ether_ops = {
2749 .ndo_init = vxlan_init,
2750 .ndo_uninit = vxlan_uninit,
2751 .ndo_open = vxlan_open,
2752 .ndo_stop = vxlan_stop,
2753 .ndo_start_xmit = vxlan_xmit,
2754 .ndo_get_stats64 = ip_tunnel_get_stats64,
2755 .ndo_set_rx_mode = vxlan_set_multicast_list,
2756 .ndo_change_mtu = vxlan_change_mtu,
2757 .ndo_validate_addr = eth_validate_addr,
2758 .ndo_set_mac_address = eth_mac_addr,
2759 .ndo_fdb_add = vxlan_fdb_add,
2760 .ndo_fdb_del = vxlan_fdb_delete,
2761 .ndo_fdb_dump = vxlan_fdb_dump,
2762 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2763 };
2764
2765 static const struct net_device_ops vxlan_netdev_raw_ops = {
2766 .ndo_init = vxlan_init,
2767 .ndo_uninit = vxlan_uninit,
2768 .ndo_open = vxlan_open,
2769 .ndo_stop = vxlan_stop,
2770 .ndo_start_xmit = vxlan_xmit,
2771 .ndo_get_stats64 = ip_tunnel_get_stats64,
2772 .ndo_change_mtu = vxlan_change_mtu,
2773 .ndo_fill_metadata_dst = vxlan_fill_metadata_dst,
2774 };
2775
2776 /* Info for udev, that this is a virtual tunnel endpoint */
2777 static struct device_type vxlan_type = {
2778 .name = "vxlan",
2779 };
2780
2781 /* Calls the ndo_udp_tunnel_add of the caller in order to
2782 * supply the listening VXLAN udp ports. Callers are expected
2783 * to implement the ndo_udp_tunnel_add.
2784 */
2785 static void vxlan_offload_rx_ports(struct net_device *dev, bool push)
2786 {
2787 struct vxlan_sock *vs;
2788 struct net *net = dev_net(dev);
2789 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
2790 unsigned int i;
2791
2792 spin_lock(&vn->sock_lock);
2793 for (i = 0; i < PORT_HASH_SIZE; ++i) {
2794 hlist_for_each_entry_rcu(vs, &vn->sock_list[i], hlist) {
2795 unsigned short type;
2796
2797 if (vs->flags & VXLAN_F_GPE)
2798 type = UDP_TUNNEL_TYPE_VXLAN_GPE;
2799 else
2800 type = UDP_TUNNEL_TYPE_VXLAN;
2801
2802 if (push)
2803 udp_tunnel_push_rx_port(dev, vs->sock, type);
2804 else
2805 udp_tunnel_drop_rx_port(dev, vs->sock, type);
2806 }
2807 }
2808 spin_unlock(&vn->sock_lock);
2809 }
2810
2811 /* Initialize the device structure. */
2812 static void vxlan_setup(struct net_device *dev)
2813 {
2814 struct vxlan_dev *vxlan = netdev_priv(dev);
2815 unsigned int h;
2816
2817 eth_hw_addr_random(dev);
2818 ether_setup(dev);
2819
2820 dev->needs_free_netdev = true;
2821 SET_NETDEV_DEVTYPE(dev, &vxlan_type);
2822
2823 dev->features |= NETIF_F_LLTX;
2824 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
2825 dev->features |= NETIF_F_RXCSUM;
2826 dev->features |= NETIF_F_GSO_SOFTWARE;
2827
2828 dev->vlan_features = dev->features;
2829 dev->hw_features |= NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
2830 dev->hw_features |= NETIF_F_GSO_SOFTWARE;
2831 netif_keep_dst(dev);
2832 dev->priv_flags |= IFF_NO_QUEUE;
2833
2834 /* MTU range: 68 - 65535 */
2835 dev->min_mtu = ETH_MIN_MTU;
2836 dev->max_mtu = ETH_MAX_MTU;
2837
2838 INIT_LIST_HEAD(&vxlan->next);
2839 spin_lock_init(&vxlan->hash_lock);
2840
2841 timer_setup(&vxlan->age_timer, vxlan_cleanup, TIMER_DEFERRABLE);
2842
2843 vxlan->dev = dev;
2844
2845 gro_cells_init(&vxlan->gro_cells, dev);
2846
2847 for (h = 0; h < FDB_HASH_SIZE; ++h)
2848 INIT_HLIST_HEAD(&vxlan->fdb_head[h]);
2849
2850 INIT_LIST_HEAD(&vxlan->fan.fan_maps);
2851 }
2852
2853 static void vxlan_ether_setup(struct net_device *dev)
2854 {
2855 dev->priv_flags &= ~IFF_TX_SKB_SHARING;
2856 dev->priv_flags |= IFF_LIVE_ADDR_CHANGE;
2857 dev->netdev_ops = &vxlan_netdev_ether_ops;
2858 }
2859
2860 static void vxlan_raw_setup(struct net_device *dev)
2861 {
2862 dev->header_ops = NULL;
2863 dev->type = ARPHRD_NONE;
2864 dev->hard_header_len = 0;
2865 dev->addr_len = 0;
2866 dev->flags = IFF_POINTOPOINT | IFF_NOARP | IFF_MULTICAST;
2867 dev->netdev_ops = &vxlan_netdev_raw_ops;
2868 }
2869
2870 static const struct nla_policy vxlan_policy[IFLA_VXLAN_MAX + 1] = {
2871 [IFLA_VXLAN_ID] = { .type = NLA_U32 },
2872 [IFLA_VXLAN_GROUP] = { .len = FIELD_SIZEOF(struct iphdr, daddr) },
2873 [IFLA_VXLAN_GROUP6] = { .len = sizeof(struct in6_addr) },
2874 [IFLA_VXLAN_LINK] = { .type = NLA_U32 },
2875 [IFLA_VXLAN_LOCAL] = { .len = FIELD_SIZEOF(struct iphdr, saddr) },
2876 [IFLA_VXLAN_LOCAL6] = { .len = sizeof(struct in6_addr) },
2877 [IFLA_VXLAN_TOS] = { .type = NLA_U8 },
2878 [IFLA_VXLAN_TTL] = { .type = NLA_U8 },
2879 [IFLA_VXLAN_LABEL] = { .type = NLA_U32 },
2880 [IFLA_VXLAN_LEARNING] = { .type = NLA_U8 },
2881 [IFLA_VXLAN_AGEING] = { .type = NLA_U32 },
2882 [IFLA_VXLAN_LIMIT] = { .type = NLA_U32 },
2883 [IFLA_VXLAN_PORT_RANGE] = { .len = sizeof(struct ifla_vxlan_port_range) },
2884 [IFLA_VXLAN_PROXY] = { .type = NLA_U8 },
2885 [IFLA_VXLAN_RSC] = { .type = NLA_U8 },
2886 [IFLA_VXLAN_L2MISS] = { .type = NLA_U8 },
2887 [IFLA_VXLAN_L3MISS] = { .type = NLA_U8 },
2888 [IFLA_VXLAN_COLLECT_METADATA] = { .type = NLA_U8 },
2889 [IFLA_VXLAN_PORT] = { .type = NLA_U16 },
2890 [IFLA_VXLAN_UDP_CSUM] = { .type = NLA_U8 },
2891 [IFLA_VXLAN_UDP_ZERO_CSUM6_TX] = { .type = NLA_U8 },
2892 [IFLA_VXLAN_UDP_ZERO_CSUM6_RX] = { .type = NLA_U8 },
2893 [IFLA_VXLAN_REMCSUM_TX] = { .type = NLA_U8 },
2894 [IFLA_VXLAN_REMCSUM_RX] = { .type = NLA_U8 },
2895 [IFLA_VXLAN_GBP] = { .type = NLA_FLAG, },
2896 [IFLA_VXLAN_GPE] = { .type = NLA_FLAG, },
2897 [IFLA_VXLAN_REMCSUM_NOPARTIAL] = { .type = NLA_FLAG },
2898 };
2899
2900 static int vxlan_validate(struct nlattr *tb[], struct nlattr *data[],
2901 struct netlink_ext_ack *extack)
2902 {
2903 if (tb[IFLA_ADDRESS]) {
2904 if (nla_len(tb[IFLA_ADDRESS]) != ETH_ALEN) {
2905 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2906 "Provided link layer address is not Ethernet");
2907 return -EINVAL;
2908 }
2909
2910 if (!is_valid_ether_addr(nla_data(tb[IFLA_ADDRESS]))) {
2911 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_ADDRESS],
2912 "Provided Ethernet address is not unicast");
2913 return -EADDRNOTAVAIL;
2914 }
2915 }
2916
2917 if (tb[IFLA_MTU]) {
2918 u32 mtu = nla_get_u32(tb[IFLA_MTU]);
2919
2920 if (mtu < ETH_MIN_MTU || mtu > ETH_MAX_MTU) {
2921 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_MTU],
2922 "MTU must be between 68 and 65535");
2923 return -EINVAL;
2924 }
2925 }
2926
2927 if (!data) {
2928 NL_SET_ERR_MSG(extack,
2929 "Required attributes not provided to perform the operation");
2930 return -EINVAL;
2931 }
2932
2933 if (data[IFLA_VXLAN_ID]) {
2934 u32 id = nla_get_u32(data[IFLA_VXLAN_ID]);
2935
2936 if (id >= VXLAN_N_VID) {
2937 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_ID],
2938 "VXLAN ID must be lower than 16777216");
2939 return -ERANGE;
2940 }
2941 }
2942
2943 if (data[IFLA_VXLAN_PORT_RANGE]) {
2944 const struct ifla_vxlan_port_range *p
2945 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
2946
2947 if (ntohs(p->high) < ntohs(p->low)) {
2948 NL_SET_ERR_MSG_ATTR(extack, tb[IFLA_VXLAN_PORT_RANGE],
2949 "Invalid source port range");
2950 return -EINVAL;
2951 }
2952 }
2953
2954 return 0;
2955 }
2956
2957 static void vxlan_get_drvinfo(struct net_device *netdev,
2958 struct ethtool_drvinfo *drvinfo)
2959 {
2960 strlcpy(drvinfo->version, VXLAN_VERSION, sizeof(drvinfo->version));
2961 strlcpy(drvinfo->driver, "vxlan", sizeof(drvinfo->driver));
2962 }
2963
2964 static const struct ethtool_ops vxlan_ethtool_ops = {
2965 .get_drvinfo = vxlan_get_drvinfo,
2966 .get_link = ethtool_op_get_link,
2967 };
2968
2969 static struct socket *vxlan_create_sock(struct net *net, bool ipv6,
2970 __be16 port, u32 flags)
2971 {
2972 struct socket *sock;
2973 struct udp_port_cfg udp_conf;
2974 int err;
2975
2976 memset(&udp_conf, 0, sizeof(udp_conf));
2977
2978 if (ipv6) {
2979 udp_conf.family = AF_INET6;
2980 udp_conf.use_udp6_rx_checksums =
2981 !(flags & VXLAN_F_UDP_ZERO_CSUM6_RX);
2982 udp_conf.ipv6_v6only = 1;
2983 } else {
2984 udp_conf.family = AF_INET;
2985 }
2986
2987 udp_conf.local_udp_port = port;
2988
2989 /* Open UDP socket */
2990 err = udp_sock_create(net, &udp_conf, &sock);
2991 if (err < 0)
2992 return ERR_PTR(err);
2993
2994 return sock;
2995 }
2996
2997 /* Create new listen socket if needed */
2998 static struct vxlan_sock *vxlan_socket_create(struct net *net, bool ipv6,
2999 __be16 port, u32 flags)
3000 {
3001 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3002 struct vxlan_sock *vs;
3003 struct socket *sock;
3004 unsigned int h;
3005 struct udp_tunnel_sock_cfg tunnel_cfg;
3006
3007 vs = kzalloc(sizeof(*vs), GFP_KERNEL);
3008 if (!vs)
3009 return ERR_PTR(-ENOMEM);
3010
3011 for (h = 0; h < VNI_HASH_SIZE; ++h)
3012 INIT_HLIST_HEAD(&vs->vni_list[h]);
3013
3014 sock = vxlan_create_sock(net, ipv6, port, flags);
3015 if (IS_ERR(sock)) {
3016 kfree(vs);
3017 return ERR_CAST(sock);
3018 }
3019
3020 vs->sock = sock;
3021 refcount_set(&vs->refcnt, 1);
3022 vs->flags = (flags & VXLAN_F_RCV_FLAGS);
3023
3024 spin_lock(&vn->sock_lock);
3025 hlist_add_head_rcu(&vs->hlist, vs_head(net, port));
3026 udp_tunnel_notify_add_rx_port(sock,
3027 (vs->flags & VXLAN_F_GPE) ?
3028 UDP_TUNNEL_TYPE_VXLAN_GPE :
3029 UDP_TUNNEL_TYPE_VXLAN);
3030 spin_unlock(&vn->sock_lock);
3031
3032 /* Mark socket as an encapsulation socket. */
3033 memset(&tunnel_cfg, 0, sizeof(tunnel_cfg));
3034 tunnel_cfg.sk_user_data = vs;
3035 tunnel_cfg.encap_type = 1;
3036 tunnel_cfg.encap_rcv = vxlan_rcv;
3037 tunnel_cfg.encap_destroy = NULL;
3038 tunnel_cfg.gro_receive = vxlan_gro_receive;
3039 tunnel_cfg.gro_complete = vxlan_gro_complete;
3040
3041 setup_udp_tunnel_sock(net, sock, &tunnel_cfg);
3042
3043 return vs;
3044 }
3045
3046 static int __vxlan_sock_add(struct vxlan_dev *vxlan, bool ipv6)
3047 {
3048 struct vxlan_net *vn = net_generic(vxlan->net, vxlan_net_id);
3049 struct vxlan_sock *vs = NULL;
3050 struct vxlan_dev_node *node;
3051
3052 if (!vxlan->cfg.no_share) {
3053 spin_lock(&vn->sock_lock);
3054 vs = vxlan_find_sock(vxlan->net, ipv6 ? AF_INET6 : AF_INET,
3055 vxlan->cfg.dst_port, vxlan->cfg.flags);
3056 if (vs && !refcount_inc_not_zero(&vs->refcnt)) {
3057 spin_unlock(&vn->sock_lock);
3058 return -EBUSY;
3059 }
3060 spin_unlock(&vn->sock_lock);
3061 }
3062 if (!vs)
3063 vs = vxlan_socket_create(vxlan->net, ipv6,
3064 vxlan->cfg.dst_port, vxlan->cfg.flags);
3065 if (IS_ERR(vs))
3066 return PTR_ERR(vs);
3067 #if IS_ENABLED(CONFIG_IPV6)
3068 if (ipv6) {
3069 rcu_assign_pointer(vxlan->vn6_sock, vs);
3070 node = &vxlan->hlist6;
3071 } else
3072 #endif
3073 {
3074 rcu_assign_pointer(vxlan->vn4_sock, vs);
3075 node = &vxlan->hlist4;
3076 }
3077 vxlan_vs_add_dev(vs, vxlan, node);
3078 return 0;
3079 }
3080
3081 static int vxlan_sock_add(struct vxlan_dev *vxlan)
3082 {
3083 bool metadata = vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA;
3084 bool ipv6 = vxlan->cfg.flags & VXLAN_F_IPV6 || metadata;
3085 bool ipv4 = !ipv6 || metadata;
3086 int ret = 0;
3087
3088 RCU_INIT_POINTER(vxlan->vn4_sock, NULL);
3089 #if IS_ENABLED(CONFIG_IPV6)
3090 RCU_INIT_POINTER(vxlan->vn6_sock, NULL);
3091 if (ipv6) {
3092 ret = __vxlan_sock_add(vxlan, true);
3093 if (ret < 0 && ret != -EAFNOSUPPORT)
3094 ipv4 = false;
3095 }
3096 #endif
3097 if (ipv4)
3098 ret = __vxlan_sock_add(vxlan, false);
3099 if (ret < 0)
3100 vxlan_sock_release(vxlan);
3101 return ret;
3102 }
3103
3104 static int vxlan_config_validate(struct net *src_net, struct vxlan_config *conf,
3105 struct net_device **lower,
3106 struct vxlan_dev *old,
3107 struct netlink_ext_ack *extack)
3108 {
3109 struct vxlan_net *vn = net_generic(src_net, vxlan_net_id);
3110 struct vxlan_dev *tmp;
3111 bool use_ipv6 = false;
3112
3113 if (conf->flags & VXLAN_F_GPE) {
3114 /* For now, allow GPE only together with
3115 * COLLECT_METADATA. This can be relaxed later; in such
3116 * case, the other side of the PtP link will have to be
3117 * provided.
3118 */
3119 if ((conf->flags & ~VXLAN_F_ALLOWED_GPE) ||
3120 !(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3121 NL_SET_ERR_MSG(extack,
3122 "VXLAN GPE does not support this combination of attributes");
3123 return -EINVAL;
3124 }
3125 }
3126
3127 if (!conf->remote_ip.sa.sa_family && !conf->saddr.sa.sa_family) {
3128 /* Unless IPv6 is explicitly requested, assume IPv4 */
3129 conf->remote_ip.sa.sa_family = AF_INET;
3130 conf->saddr.sa.sa_family = AF_INET;
3131 } else if (!conf->remote_ip.sa.sa_family) {
3132 conf->remote_ip.sa.sa_family = conf->saddr.sa.sa_family;
3133 } else if (!conf->saddr.sa.sa_family) {
3134 conf->saddr.sa.sa_family = conf->remote_ip.sa.sa_family;
3135 }
3136
3137 if (conf->saddr.sa.sa_family != conf->remote_ip.sa.sa_family) {
3138 NL_SET_ERR_MSG(extack,
3139 "Local and remote address must be from the same family");
3140 return -EINVAL;
3141 }
3142
3143 if (vxlan_addr_multicast(&conf->saddr)) {
3144 NL_SET_ERR_MSG(extack, "Local address cannot be multicast");
3145 return -EINVAL;
3146 }
3147
3148 if (conf->saddr.sa.sa_family == AF_INET6) {
3149 if (!IS_ENABLED(CONFIG_IPV6)) {
3150 NL_SET_ERR_MSG(extack,
3151 "IPv6 support not enabled in the kernel");
3152 return -EPFNOSUPPORT;
3153 }
3154 use_ipv6 = true;
3155 conf->flags |= VXLAN_F_IPV6;
3156
3157 if (!(conf->flags & VXLAN_F_COLLECT_METADATA)) {
3158 int local_type =
3159 ipv6_addr_type(&conf->saddr.sin6.sin6_addr);
3160 int remote_type =
3161 ipv6_addr_type(&conf->remote_ip.sin6.sin6_addr);
3162
3163 if (local_type & IPV6_ADDR_LINKLOCAL) {
3164 if (!(remote_type & IPV6_ADDR_LINKLOCAL) &&
3165 (remote_type != IPV6_ADDR_ANY)) {
3166 NL_SET_ERR_MSG(extack,
3167 "Invalid combination of local and remote address scopes");
3168 return -EINVAL;
3169 }
3170
3171 conf->flags |= VXLAN_F_IPV6_LINKLOCAL;
3172 } else {
3173 if (remote_type ==
3174 (IPV6_ADDR_UNICAST | IPV6_ADDR_LINKLOCAL)) {
3175 NL_SET_ERR_MSG(extack,
3176 "Invalid combination of local and remote address scopes");
3177 return -EINVAL;
3178 }
3179
3180 conf->flags &= ~VXLAN_F_IPV6_LINKLOCAL;
3181 }
3182 }
3183 }
3184
3185 if (conf->label && !use_ipv6) {
3186 NL_SET_ERR_MSG(extack,
3187 "Label attribute only applies to IPv6 VXLAN devices");
3188 return -EINVAL;
3189 }
3190
3191 if (conf->remote_ifindex) {
3192 struct net_device *lowerdev;
3193
3194 lowerdev = __dev_get_by_index(src_net, conf->remote_ifindex);
3195 if (!lowerdev) {
3196 NL_SET_ERR_MSG(extack,
3197 "Invalid local interface, device not found");
3198 return -ENODEV;
3199 }
3200
3201 #if IS_ENABLED(CONFIG_IPV6)
3202 if (use_ipv6) {
3203 struct inet6_dev *idev = __in6_dev_get(lowerdev);
3204 if (idev && idev->cnf.disable_ipv6) {
3205 NL_SET_ERR_MSG(extack,
3206 "IPv6 support disabled by administrator");
3207 return -EPERM;
3208 }
3209 }
3210 #endif
3211
3212 *lower = lowerdev;
3213 } else {
3214 if (vxlan_addr_multicast(&conf->remote_ip)) {
3215 NL_SET_ERR_MSG(extack,
3216 "Local interface required for multicast remote destination");
3217
3218 return -EINVAL;
3219 }
3220
3221 #if IS_ENABLED(CONFIG_IPV6)
3222 if (conf->flags & VXLAN_F_IPV6_LINKLOCAL) {
3223 NL_SET_ERR_MSG(extack,
3224 "Local interface required for link-local local/remote addresses");
3225 return -EINVAL;
3226 }
3227 #endif
3228
3229 *lower = NULL;
3230 }
3231
3232 if (!conf->dst_port) {
3233 if (conf->flags & VXLAN_F_GPE)
3234 conf->dst_port = htons(4790); /* IANA VXLAN-GPE port */
3235 else
3236 conf->dst_port = htons(vxlan_port);
3237 }
3238
3239 if (!conf->age_interval)
3240 conf->age_interval = FDB_AGE_DEFAULT;
3241
3242 list_for_each_entry(tmp, &vn->vxlan_list, next) {
3243 if (tmp == old)
3244 continue;
3245
3246 if (tmp->cfg.vni != conf->vni)
3247 continue;
3248 if (tmp->cfg.dst_port != conf->dst_port)
3249 continue;
3250 if ((tmp->cfg.flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)) !=
3251 (conf->flags & (VXLAN_F_RCV_FLAGS | VXLAN_F_IPV6)))
3252 continue;
3253
3254 if ((conf->flags & VXLAN_F_IPV6_LINKLOCAL) &&
3255 tmp->cfg.remote_ifindex != conf->remote_ifindex)
3256 continue;
3257
3258 NL_SET_ERR_MSG(extack,
3259 "A VXLAN device with the specified VNI already exists");
3260 return -EEXIST;
3261 }
3262
3263 return 0;
3264 }
3265
3266 static void vxlan_config_apply(struct net_device *dev,
3267 struct vxlan_config *conf,
3268 struct net_device *lowerdev,
3269 struct net *src_net,
3270 bool changelink)
3271 {
3272 struct vxlan_dev *vxlan = netdev_priv(dev);
3273 struct vxlan_rdst *dst = &vxlan->default_dst;
3274 unsigned short needed_headroom = ETH_HLEN;
3275 bool use_ipv6 = !!(conf->flags & VXLAN_F_IPV6);
3276 int max_mtu = ETH_MAX_MTU;
3277
3278 if (!changelink) {
3279 if (conf->flags & VXLAN_F_GPE)
3280 vxlan_raw_setup(dev);
3281 else
3282 vxlan_ether_setup(dev);
3283
3284 if (conf->mtu)
3285 dev->mtu = conf->mtu;
3286
3287 vxlan->net = src_net;
3288 }
3289
3290 dst->remote_vni = conf->vni;
3291
3292 memcpy(&dst->remote_ip, &conf->remote_ip, sizeof(conf->remote_ip));
3293
3294 if (lowerdev) {
3295 dst->remote_ifindex = conf->remote_ifindex;
3296
3297 dev->gso_max_size = lowerdev->gso_max_size;
3298 dev->gso_max_segs = lowerdev->gso_max_segs;
3299
3300 needed_headroom = lowerdev->hard_header_len;
3301
3302 max_mtu = lowerdev->mtu - (use_ipv6 ? VXLAN6_HEADROOM :
3303 VXLAN_HEADROOM);
3304 if (max_mtu < ETH_MIN_MTU)
3305 max_mtu = ETH_MIN_MTU;
3306
3307 if (!changelink && !conf->mtu)
3308 dev->mtu = max_mtu;
3309 }
3310
3311 if (dev->mtu > max_mtu)
3312 dev->mtu = max_mtu;
3313
3314 if (use_ipv6 || conf->flags & VXLAN_F_COLLECT_METADATA)
3315 needed_headroom += VXLAN6_HEADROOM;
3316 else
3317 needed_headroom += VXLAN_HEADROOM;
3318 dev->needed_headroom = needed_headroom;
3319
3320 memcpy(&vxlan->cfg, conf, sizeof(*conf));
3321 }
3322
3323 static int vxlan_dev_configure(struct net *src_net, struct net_device *dev,
3324 struct vxlan_config *conf, bool changelink,
3325 struct netlink_ext_ack *extack)
3326 {
3327 struct vxlan_dev *vxlan = netdev_priv(dev);
3328 struct net_device *lowerdev;
3329 int ret;
3330
3331 ret = vxlan_config_validate(src_net, conf, &lowerdev, vxlan, extack);
3332 if (ret)
3333 return ret;
3334
3335 vxlan_config_apply(dev, conf, lowerdev, src_net, changelink);
3336
3337 return 0;
3338 }
3339
3340 static int __vxlan_dev_create(struct net *net, struct net_device *dev,
3341 struct vxlan_config *conf,
3342 struct netlink_ext_ack *extack)
3343 {
3344 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3345 struct vxlan_dev *vxlan = netdev_priv(dev);
3346 int err;
3347
3348 err = vxlan_dev_configure(net, dev, conf, false, extack);
3349 if (err)
3350 return err;
3351
3352 dev->ethtool_ops = &vxlan_ethtool_ops;
3353
3354 /* create an fdb entry for a valid default destination */
3355 if (!vxlan_addr_any(&vxlan->default_dst.remote_ip)) {
3356 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3357 &vxlan->default_dst.remote_ip,
3358 NUD_REACHABLE | NUD_PERMANENT,
3359 NLM_F_EXCL | NLM_F_CREATE,
3360 vxlan->cfg.dst_port,
3361 vxlan->default_dst.remote_vni,
3362 vxlan->default_dst.remote_vni,
3363 vxlan->default_dst.remote_ifindex,
3364 NTF_SELF);
3365 if (err)
3366 return err;
3367 }
3368
3369 err = register_netdevice(dev);
3370 if (err) {
3371 vxlan_fdb_delete_default(vxlan, vxlan->default_dst.remote_vni);
3372 return err;
3373 }
3374
3375 list_add(&vxlan->next, &vn->vxlan_list);
3376 return 0;
3377 }
3378
3379 static int vxlan_nl2conf(struct nlattr *tb[], struct nlattr *data[],
3380 struct net_device *dev, struct vxlan_config *conf,
3381 bool changelink)
3382 {
3383 struct vxlan_dev *vxlan = netdev_priv(dev);
3384 int err;
3385
3386 memset(conf, 0, sizeof(*conf));
3387
3388 /* if changelink operation, start with old existing cfg */
3389 if (changelink)
3390 memcpy(conf, &vxlan->cfg, sizeof(*conf));
3391
3392 if (data[IFLA_VXLAN_ID]) {
3393 __be32 vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3394
3395 if (changelink && (vni != conf->vni))
3396 return -EOPNOTSUPP;
3397 conf->vni = cpu_to_be32(nla_get_u32(data[IFLA_VXLAN_ID]));
3398 }
3399
3400 if (data[IFLA_VXLAN_GROUP]) {
3401 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET))
3402 return -EOPNOTSUPP;
3403
3404 conf->remote_ip.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_GROUP]);
3405 conf->remote_ip.sa.sa_family = AF_INET;
3406 } else if (data[IFLA_VXLAN_GROUP6]) {
3407 if (!IS_ENABLED(CONFIG_IPV6))
3408 return -EPFNOSUPPORT;
3409
3410 if (changelink && (conf->remote_ip.sa.sa_family != AF_INET6))
3411 return -EOPNOTSUPP;
3412
3413 conf->remote_ip.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_GROUP6]);
3414 conf->remote_ip.sa.sa_family = AF_INET6;
3415 }
3416
3417 if (data[IFLA_VXLAN_FAN_MAP]) {
3418 err = vxlan_parse_fan_map(data, vxlan);
3419 if (err)
3420 return err;
3421 }
3422
3423 if (data[IFLA_VXLAN_LOCAL]) {
3424 if (changelink && (conf->saddr.sa.sa_family != AF_INET))
3425 return -EOPNOTSUPP;
3426
3427 conf->saddr.sin.sin_addr.s_addr = nla_get_in_addr(data[IFLA_VXLAN_LOCAL]);
3428 conf->saddr.sa.sa_family = AF_INET;
3429 } else if (data[IFLA_VXLAN_LOCAL6]) {
3430 if (!IS_ENABLED(CONFIG_IPV6))
3431 return -EPFNOSUPPORT;
3432
3433 if (changelink && (conf->saddr.sa.sa_family != AF_INET6))
3434 return -EOPNOTSUPP;
3435
3436 /* TODO: respect scope id */
3437 conf->saddr.sin6.sin6_addr = nla_get_in6_addr(data[IFLA_VXLAN_LOCAL6]);
3438 conf->saddr.sa.sa_family = AF_INET6;
3439 }
3440
3441 if (data[IFLA_VXLAN_LINK])
3442 conf->remote_ifindex = nla_get_u32(data[IFLA_VXLAN_LINK]);
3443
3444 if (data[IFLA_VXLAN_TOS])
3445 conf->tos = nla_get_u8(data[IFLA_VXLAN_TOS]);
3446
3447 if (data[IFLA_VXLAN_TTL])
3448 conf->ttl = nla_get_u8(data[IFLA_VXLAN_TTL]);
3449
3450 if (data[IFLA_VXLAN_LABEL])
3451 conf->label = nla_get_be32(data[IFLA_VXLAN_LABEL]) &
3452 IPV6_FLOWLABEL_MASK;
3453
3454 if (data[IFLA_VXLAN_LEARNING]) {
3455 if (nla_get_u8(data[IFLA_VXLAN_LEARNING]))
3456 conf->flags |= VXLAN_F_LEARN;
3457 else
3458 conf->flags &= ~VXLAN_F_LEARN;
3459 } else if (!changelink) {
3460 /* default to learn on a new device */
3461 conf->flags |= VXLAN_F_LEARN;
3462 }
3463
3464 if (data[IFLA_VXLAN_AGEING]) {
3465 if (changelink)
3466 return -EOPNOTSUPP;
3467 conf->age_interval = nla_get_u32(data[IFLA_VXLAN_AGEING]);
3468 }
3469
3470 if (data[IFLA_VXLAN_PROXY]) {
3471 if (changelink)
3472 return -EOPNOTSUPP;
3473 if (nla_get_u8(data[IFLA_VXLAN_PROXY]))
3474 conf->flags |= VXLAN_F_PROXY;
3475 }
3476
3477 if (data[IFLA_VXLAN_RSC]) {
3478 if (changelink)
3479 return -EOPNOTSUPP;
3480 if (nla_get_u8(data[IFLA_VXLAN_RSC]))
3481 conf->flags |= VXLAN_F_RSC;
3482 }
3483
3484 if (data[IFLA_VXLAN_L2MISS]) {
3485 if (changelink)
3486 return -EOPNOTSUPP;
3487 if (nla_get_u8(data[IFLA_VXLAN_L2MISS]))
3488 conf->flags |= VXLAN_F_L2MISS;
3489 }
3490
3491 if (data[IFLA_VXLAN_L3MISS]) {
3492 if (changelink)
3493 return -EOPNOTSUPP;
3494 if (nla_get_u8(data[IFLA_VXLAN_L3MISS]))
3495 conf->flags |= VXLAN_F_L3MISS;
3496 }
3497
3498 if (data[IFLA_VXLAN_LIMIT]) {
3499 if (changelink)
3500 return -EOPNOTSUPP;
3501 conf->addrmax = nla_get_u32(data[IFLA_VXLAN_LIMIT]);
3502 }
3503
3504 if (data[IFLA_VXLAN_COLLECT_METADATA]) {
3505 if (changelink)
3506 return -EOPNOTSUPP;
3507 if (nla_get_u8(data[IFLA_VXLAN_COLLECT_METADATA]))
3508 conf->flags |= VXLAN_F_COLLECT_METADATA;
3509 }
3510
3511 if (data[IFLA_VXLAN_PORT_RANGE]) {
3512 if (!changelink) {
3513 const struct ifla_vxlan_port_range *p
3514 = nla_data(data[IFLA_VXLAN_PORT_RANGE]);
3515 conf->port_min = ntohs(p->low);
3516 conf->port_max = ntohs(p->high);
3517 } else {
3518 return -EOPNOTSUPP;
3519 }
3520 }
3521
3522 if (data[IFLA_VXLAN_PORT]) {
3523 if (changelink)
3524 return -EOPNOTSUPP;
3525 conf->dst_port = nla_get_be16(data[IFLA_VXLAN_PORT]);
3526 }
3527
3528 if (data[IFLA_VXLAN_UDP_CSUM]) {
3529 if (changelink)
3530 return -EOPNOTSUPP;
3531 if (!nla_get_u8(data[IFLA_VXLAN_UDP_CSUM]))
3532 conf->flags |= VXLAN_F_UDP_ZERO_CSUM_TX;
3533 }
3534
3535 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]) {
3536 if (changelink)
3537 return -EOPNOTSUPP;
3538 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_TX]))
3539 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_TX;
3540 }
3541
3542 if (data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]) {
3543 if (changelink)
3544 return -EOPNOTSUPP;
3545 if (nla_get_u8(data[IFLA_VXLAN_UDP_ZERO_CSUM6_RX]))
3546 conf->flags |= VXLAN_F_UDP_ZERO_CSUM6_RX;
3547 }
3548
3549 if (data[IFLA_VXLAN_REMCSUM_TX]) {
3550 if (changelink)
3551 return -EOPNOTSUPP;
3552 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_TX]))
3553 conf->flags |= VXLAN_F_REMCSUM_TX;
3554 }
3555
3556 if (data[IFLA_VXLAN_REMCSUM_RX]) {
3557 if (changelink)
3558 return -EOPNOTSUPP;
3559 if (nla_get_u8(data[IFLA_VXLAN_REMCSUM_RX]))
3560 conf->flags |= VXLAN_F_REMCSUM_RX;
3561 }
3562
3563 if (data[IFLA_VXLAN_GBP]) {
3564 if (changelink)
3565 return -EOPNOTSUPP;
3566 conf->flags |= VXLAN_F_GBP;
3567 }
3568
3569 if (data[IFLA_VXLAN_GPE]) {
3570 if (changelink)
3571 return -EOPNOTSUPP;
3572 conf->flags |= VXLAN_F_GPE;
3573 }
3574
3575 if (data[IFLA_VXLAN_REMCSUM_NOPARTIAL]) {
3576 if (changelink)
3577 return -EOPNOTSUPP;
3578 conf->flags |= VXLAN_F_REMCSUM_NOPARTIAL;
3579 }
3580
3581 if (tb[IFLA_MTU]) {
3582 if (changelink)
3583 return -EOPNOTSUPP;
3584 conf->mtu = nla_get_u32(tb[IFLA_MTU]);
3585 }
3586
3587 return 0;
3588 }
3589
3590 static int vxlan_newlink(struct net *src_net, struct net_device *dev,
3591 struct nlattr *tb[], struct nlattr *data[],
3592 struct netlink_ext_ack *extack)
3593 {
3594 struct vxlan_config conf;
3595 int err;
3596
3597 err = vxlan_nl2conf(tb, data, dev, &conf, false);
3598 if (err)
3599 return err;
3600
3601 return __vxlan_dev_create(src_net, dev, &conf, extack);
3602 }
3603
3604 static int vxlan_changelink(struct net_device *dev, struct nlattr *tb[],
3605 struct nlattr *data[],
3606 struct netlink_ext_ack *extack)
3607 {
3608 struct vxlan_dev *vxlan = netdev_priv(dev);
3609 struct vxlan_rdst *dst = &vxlan->default_dst;
3610 struct vxlan_rdst old_dst;
3611 struct vxlan_config conf;
3612 int err;
3613
3614 err = vxlan_nl2conf(tb, data,
3615 dev, &conf, true);
3616 if (err)
3617 return err;
3618
3619 memcpy(&old_dst, dst, sizeof(struct vxlan_rdst));
3620
3621 err = vxlan_dev_configure(vxlan->net, dev, &conf, true, extack);
3622 if (err)
3623 return err;
3624
3625 /* handle default dst entry */
3626 if (!vxlan_addr_equal(&dst->remote_ip, &old_dst.remote_ip)) {
3627 spin_lock_bh(&vxlan->hash_lock);
3628 if (!vxlan_addr_any(&old_dst.remote_ip))
3629 __vxlan_fdb_delete(vxlan, all_zeros_mac,
3630 old_dst.remote_ip,
3631 vxlan->cfg.dst_port,
3632 old_dst.remote_vni,
3633 old_dst.remote_vni,
3634 old_dst.remote_ifindex, 0);
3635
3636 if (!vxlan_addr_any(&dst->remote_ip)) {
3637 err = vxlan_fdb_create(vxlan, all_zeros_mac,
3638 &dst->remote_ip,
3639 NUD_REACHABLE | NUD_PERMANENT,
3640 NLM_F_CREATE | NLM_F_APPEND,
3641 vxlan->cfg.dst_port,
3642 dst->remote_vni,
3643 dst->remote_vni,
3644 dst->remote_ifindex,
3645 NTF_SELF);
3646 if (err) {
3647 spin_unlock_bh(&vxlan->hash_lock);
3648 return err;
3649 }
3650 }
3651 spin_unlock_bh(&vxlan->hash_lock);
3652 }
3653
3654 return 0;
3655 }
3656
3657 static void vxlan_dellink(struct net_device *dev, struct list_head *head)
3658 {
3659 struct vxlan_dev *vxlan = netdev_priv(dev);
3660
3661 vxlan_flush(vxlan, true);
3662
3663 gro_cells_destroy(&vxlan->gro_cells);
3664 list_del(&vxlan->next);
3665 unregister_netdevice_queue(dev, head);
3666 }
3667
3668 static size_t vxlan_get_size(const struct net_device *dev)
3669 {
3670
3671 return nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_ID */
3672 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_GROUP{6} */
3673 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LINK */
3674 nla_total_size(sizeof(struct in6_addr)) + /* IFLA_VXLAN_LOCAL{6} */
3675 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TTL */
3676 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_TOS */
3677 nla_total_size(sizeof(__be32)) + /* IFLA_VXLAN_LABEL */
3678 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_LEARNING */
3679 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_PROXY */
3680 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_RSC */
3681 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L2MISS */
3682 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_L3MISS */
3683 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_COLLECT_METADATA */
3684 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_AGEING */
3685 nla_total_size(sizeof(__u32)) + /* IFLA_VXLAN_LIMIT */
3686 nla_total_size(sizeof(struct ifla_vxlan_port_range)) +
3687 nla_total_size(sizeof(__be16)) + /* IFLA_VXLAN_PORT */
3688 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_CSUM */
3689 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_TX */
3690 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_UDP_ZERO_CSUM6_RX */
3691 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_TX */
3692 nla_total_size(sizeof(__u8)) + /* IFLA_VXLAN_REMCSUM_RX */
3693 nla_total_size(sizeof(struct ip_fan_map) * 256) +
3694 0;
3695 }
3696
3697 static int vxlan_fill_info(struct sk_buff *skb, const struct net_device *dev)
3698 {
3699 const struct vxlan_dev *vxlan = netdev_priv(dev);
3700 const struct vxlan_rdst *dst = &vxlan->default_dst;
3701 struct ifla_vxlan_port_range ports = {
3702 .low = htons(vxlan->cfg.port_min),
3703 .high = htons(vxlan->cfg.port_max),
3704 };
3705
3706 if (nla_put_u32(skb, IFLA_VXLAN_ID, be32_to_cpu(dst->remote_vni)))
3707 goto nla_put_failure;
3708
3709 if (!vxlan_addr_any(&dst->remote_ip)) {
3710 if (dst->remote_ip.sa.sa_family == AF_INET) {
3711 if (nla_put_in_addr(skb, IFLA_VXLAN_GROUP,
3712 dst->remote_ip.sin.sin_addr.s_addr))
3713 goto nla_put_failure;
3714 #if IS_ENABLED(CONFIG_IPV6)
3715 } else {
3716 if (nla_put_in6_addr(skb, IFLA_VXLAN_GROUP6,
3717 &dst->remote_ip.sin6.sin6_addr))
3718 goto nla_put_failure;
3719 #endif
3720 }
3721 }
3722
3723 if (dst->remote_ifindex && nla_put_u32(skb, IFLA_VXLAN_LINK, dst->remote_ifindex))
3724 goto nla_put_failure;
3725
3726 if (!vxlan_addr_any(&vxlan->cfg.saddr)) {
3727 if (vxlan->cfg.saddr.sa.sa_family == AF_INET) {
3728 if (nla_put_in_addr(skb, IFLA_VXLAN_LOCAL,
3729 vxlan->cfg.saddr.sin.sin_addr.s_addr))
3730 goto nla_put_failure;
3731 #if IS_ENABLED(CONFIG_IPV6)
3732 } else {
3733 if (nla_put_in6_addr(skb, IFLA_VXLAN_LOCAL6,
3734 &vxlan->cfg.saddr.sin6.sin6_addr))
3735 goto nla_put_failure;
3736 #endif
3737 }
3738 }
3739
3740 if (fan_has_map(&vxlan->fan)) {
3741 struct nlattr *fan_nest;
3742 struct ip_fan_map *fan_map;
3743
3744 fan_nest = nla_nest_start(skb, IFLA_VXLAN_FAN_MAP);
3745 if (!fan_nest)
3746 goto nla_put_failure;
3747 list_for_each_entry_rcu(fan_map, &vxlan->fan.fan_maps, list) {
3748 struct ifla_fan_map map;
3749
3750 map.underlay = fan_map->underlay;
3751 map.underlay_prefix = fan_map->underlay_prefix;
3752 map.overlay = fan_map->overlay;
3753 map.overlay_prefix = fan_map->overlay_prefix;
3754 if (nla_put(skb, IFLA_FAN_MAPPING, sizeof(map), &map))
3755 goto nla_put_failure;
3756 }
3757 nla_nest_end(skb, fan_nest);
3758 }
3759
3760 if (nla_put_u8(skb, IFLA_VXLAN_TTL, vxlan->cfg.ttl) ||
3761 nla_put_u8(skb, IFLA_VXLAN_TOS, vxlan->cfg.tos) ||
3762 nla_put_be32(skb, IFLA_VXLAN_LABEL, vxlan->cfg.label) ||
3763 nla_put_u8(skb, IFLA_VXLAN_LEARNING,
3764 !!(vxlan->cfg.flags & VXLAN_F_LEARN)) ||
3765 nla_put_u8(skb, IFLA_VXLAN_PROXY,
3766 !!(vxlan->cfg.flags & VXLAN_F_PROXY)) ||
3767 nla_put_u8(skb, IFLA_VXLAN_RSC,
3768 !!(vxlan->cfg.flags & VXLAN_F_RSC)) ||
3769 nla_put_u8(skb, IFLA_VXLAN_L2MISS,
3770 !!(vxlan->cfg.flags & VXLAN_F_L2MISS)) ||
3771 nla_put_u8(skb, IFLA_VXLAN_L3MISS,
3772 !!(vxlan->cfg.flags & VXLAN_F_L3MISS)) ||
3773 nla_put_u8(skb, IFLA_VXLAN_COLLECT_METADATA,
3774 !!(vxlan->cfg.flags & VXLAN_F_COLLECT_METADATA)) ||
3775 nla_put_u32(skb, IFLA_VXLAN_AGEING, vxlan->cfg.age_interval) ||
3776 nla_put_u32(skb, IFLA_VXLAN_LIMIT, vxlan->cfg.addrmax) ||
3777 nla_put_be16(skb, IFLA_VXLAN_PORT, vxlan->cfg.dst_port) ||
3778 nla_put_u8(skb, IFLA_VXLAN_UDP_CSUM,
3779 !(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM_TX)) ||
3780 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_TX,
3781 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_TX)) ||
3782 nla_put_u8(skb, IFLA_VXLAN_UDP_ZERO_CSUM6_RX,
3783 !!(vxlan->cfg.flags & VXLAN_F_UDP_ZERO_CSUM6_RX)) ||
3784 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_TX,
3785 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_TX)) ||
3786 nla_put_u8(skb, IFLA_VXLAN_REMCSUM_RX,
3787 !!(vxlan->cfg.flags & VXLAN_F_REMCSUM_RX)))
3788 goto nla_put_failure;
3789
3790 if (nla_put(skb, IFLA_VXLAN_PORT_RANGE, sizeof(ports), &ports))
3791 goto nla_put_failure;
3792
3793 if (vxlan->cfg.flags & VXLAN_F_GBP &&
3794 nla_put_flag(skb, IFLA_VXLAN_GBP))
3795 goto nla_put_failure;
3796
3797 if (vxlan->cfg.flags & VXLAN_F_GPE &&
3798 nla_put_flag(skb, IFLA_VXLAN_GPE))
3799 goto nla_put_failure;
3800
3801 if (vxlan->cfg.flags & VXLAN_F_REMCSUM_NOPARTIAL &&
3802 nla_put_flag(skb, IFLA_VXLAN_REMCSUM_NOPARTIAL))
3803 goto nla_put_failure;
3804
3805 return 0;
3806
3807 nla_put_failure:
3808 return -EMSGSIZE;
3809 }
3810
3811 static struct net *vxlan_get_link_net(const struct net_device *dev)
3812 {
3813 struct vxlan_dev *vxlan = netdev_priv(dev);
3814
3815 return vxlan->net;
3816 }
3817
3818 static struct rtnl_link_ops vxlan_link_ops __read_mostly = {
3819 .kind = "vxlan",
3820 .maxtype = IFLA_VXLAN_MAX,
3821 .policy = vxlan_policy,
3822 .priv_size = sizeof(struct vxlan_dev),
3823 .setup = vxlan_setup,
3824 .validate = vxlan_validate,
3825 .newlink = vxlan_newlink,
3826 .changelink = vxlan_changelink,
3827 .dellink = vxlan_dellink,
3828 .get_size = vxlan_get_size,
3829 .fill_info = vxlan_fill_info,
3830 .get_link_net = vxlan_get_link_net,
3831 };
3832
3833 struct net_device *vxlan_dev_create(struct net *net, const char *name,
3834 u8 name_assign_type,
3835 struct vxlan_config *conf)
3836 {
3837 struct nlattr *tb[IFLA_MAX + 1];
3838 struct net_device *dev;
3839 int err;
3840
3841 memset(&tb, 0, sizeof(tb));
3842
3843 dev = rtnl_create_link(net, name, name_assign_type,
3844 &vxlan_link_ops, tb);
3845 if (IS_ERR(dev))
3846 return dev;
3847
3848 err = __vxlan_dev_create(net, dev, conf, NULL);
3849 if (err < 0) {
3850 free_netdev(dev);
3851 return ERR_PTR(err);
3852 }
3853
3854 err = rtnl_configure_link(dev, NULL);
3855 if (err < 0) {
3856 LIST_HEAD(list_kill);
3857
3858 vxlan_dellink(dev, &list_kill);
3859 unregister_netdevice_many(&list_kill);
3860 return ERR_PTR(err);
3861 }
3862
3863 return dev;
3864 }
3865 EXPORT_SYMBOL_GPL(vxlan_dev_create);
3866
3867 static void vxlan_handle_lowerdev_unregister(struct vxlan_net *vn,
3868 struct net_device *dev)
3869 {
3870 struct vxlan_dev *vxlan, *next;
3871 LIST_HEAD(list_kill);
3872
3873 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3874 struct vxlan_rdst *dst = &vxlan->default_dst;
3875
3876 /* In case we created vxlan device with carrier
3877 * and we loose the carrier due to module unload
3878 * we also need to remove vxlan device. In other
3879 * cases, it's not necessary and remote_ifindex
3880 * is 0 here, so no matches.
3881 */
3882 if (dst->remote_ifindex == dev->ifindex)
3883 vxlan_dellink(vxlan->dev, &list_kill);
3884 }
3885
3886 unregister_netdevice_many(&list_kill);
3887 }
3888
3889 static int vxlan_netdevice_event(struct notifier_block *unused,
3890 unsigned long event, void *ptr)
3891 {
3892 struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3893 struct vxlan_net *vn = net_generic(dev_net(dev), vxlan_net_id);
3894
3895 if (event == NETDEV_UNREGISTER) {
3896 vxlan_offload_rx_ports(dev, false);
3897 vxlan_handle_lowerdev_unregister(vn, dev);
3898 } else if (event == NETDEV_REGISTER) {
3899 vxlan_offload_rx_ports(dev, true);
3900 } else if (event == NETDEV_UDP_TUNNEL_PUSH_INFO ||
3901 event == NETDEV_UDP_TUNNEL_DROP_INFO) {
3902 vxlan_offload_rx_ports(dev, event == NETDEV_UDP_TUNNEL_PUSH_INFO);
3903 }
3904
3905 return NOTIFY_DONE;
3906 }
3907
3908 static struct notifier_block vxlan_notifier_block __read_mostly = {
3909 .notifier_call = vxlan_netdevice_event,
3910 };
3911
3912 static __net_init int vxlan_init_net(struct net *net)
3913 {
3914 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3915 unsigned int h;
3916
3917 INIT_LIST_HEAD(&vn->vxlan_list);
3918 spin_lock_init(&vn->sock_lock);
3919
3920 for (h = 0; h < PORT_HASH_SIZE; ++h)
3921 INIT_HLIST_HEAD(&vn->sock_list[h]);
3922
3923 return 0;
3924 }
3925
3926 #ifdef CONFIG_SYSCTL
3927 static struct ctl_table_header *vxlan_fan_header;
3928 static unsigned int vxlan_fan_version = 4;
3929
3930 static struct ctl_table vxlan_fan_sysctls[] = {
3931 {
3932 .procname = "vxlan",
3933 .data = &vxlan_fan_version,
3934 .maxlen = sizeof(vxlan_fan_version),
3935 .mode = 0444,
3936 .proc_handler = proc_dointvec,
3937 },
3938 {},
3939 };
3940 #endif /* CONFIG_SYSCTL */
3941
3942 static void __net_exit vxlan_exit_net(struct net *net)
3943 {
3944 struct vxlan_net *vn = net_generic(net, vxlan_net_id);
3945 struct vxlan_dev *vxlan, *next;
3946 struct net_device *dev, *aux;
3947 unsigned int h;
3948 LIST_HEAD(list);
3949
3950 rtnl_lock();
3951 for_each_netdev_safe(net, dev, aux)
3952 if (dev->rtnl_link_ops == &vxlan_link_ops)
3953 unregister_netdevice_queue(dev, &list);
3954
3955 list_for_each_entry_safe(vxlan, next, &vn->vxlan_list, next) {
3956 /* If vxlan->dev is in the same netns, it has already been added
3957 * to the list by the previous loop.
3958 */
3959 if (!net_eq(dev_net(vxlan->dev), net)) {
3960 gro_cells_destroy(&vxlan->gro_cells);
3961 unregister_netdevice_queue(vxlan->dev, &list);
3962 }
3963 }
3964
3965 unregister_netdevice_many(&list);
3966 rtnl_unlock();
3967
3968 for (h = 0; h < PORT_HASH_SIZE; ++h)
3969 WARN_ON_ONCE(!hlist_empty(&vn->sock_list[h]));
3970 }
3971
3972 static struct pernet_operations vxlan_net_ops = {
3973 .init = vxlan_init_net,
3974 .exit = vxlan_exit_net,
3975 .id = &vxlan_net_id,
3976 .size = sizeof(struct vxlan_net),
3977 };
3978
3979 static int __init vxlan_init_module(void)
3980 {
3981 int rc;
3982
3983 get_random_bytes(&vxlan_salt, sizeof(vxlan_salt));
3984
3985 rc = register_pernet_subsys(&vxlan_net_ops);
3986 if (rc)
3987 goto out1;
3988
3989 rc = register_netdevice_notifier(&vxlan_notifier_block);
3990 if (rc)
3991 goto out2;
3992
3993 rc = rtnl_link_register(&vxlan_link_ops);
3994 if (rc)
3995 goto out3;
3996
3997 #ifdef CONFIG_SYSCTL
3998 vxlan_fan_header = register_net_sysctl(&init_net, "net/fan",
3999 vxlan_fan_sysctls);
4000 if (!vxlan_fan_header) {
4001 rc = -ENOMEM;
4002 goto sysctl_failed;
4003 }
4004 #endif /* CONFIG_SYSCTL */
4005
4006 return 0;
4007 #ifdef CONFIG_SYSCTL
4008 sysctl_failed:
4009 rtnl_link_unregister(&vxlan_link_ops);
4010 #endif /* CONFIG_SYSCTL */
4011 out3:
4012 unregister_netdevice_notifier(&vxlan_notifier_block);
4013 out2:
4014 unregister_pernet_subsys(&vxlan_net_ops);
4015 out1:
4016 return rc;
4017 }
4018 late_initcall(vxlan_init_module);
4019
4020 static void __exit vxlan_cleanup_module(void)
4021 {
4022 #ifdef CONFIG_SYSCTL
4023 unregister_net_sysctl_table(vxlan_fan_header);
4024 #endif /* CONFIG_SYSCTL */
4025 rtnl_link_unregister(&vxlan_link_ops);
4026 unregister_netdevice_notifier(&vxlan_notifier_block);
4027 unregister_pernet_subsys(&vxlan_net_ops);
4028 /* rcu_barrier() is called by netns */
4029 }
4030 module_exit(vxlan_cleanup_module);
4031
4032 MODULE_LICENSE("GPL");
4033 MODULE_VERSION(VXLAN_VERSION);
4034 MODULE_AUTHOR("Stephen Hemminger <stephen@networkplumber.org>");
4035 MODULE_DESCRIPTION("Driver for VXLAN encapsulated traffic");
4036 MODULE_ALIAS_RTNL_LINK("vxlan");
ubuntu-bionic-kernel mirror